[llvm] [PHIElimination] Reuse existing COPY in predecessor basic block (PR #131837)
Guy David via llvm-commits
llvm-commits at lists.llvm.org
Thu Jun 26 07:18:23 PDT 2025
https://github.com/guy-david updated https://github.com/llvm/llvm-project/pull/131837
>From 5294476688127bedb6e14654f1c55b130ea32029 Mon Sep 17 00:00:00 2001
From: Guy David <guyda96 at gmail.com>
Date: Tue, 18 Mar 2025 09:55:38 +0200
Subject: [PATCH 1/3] [PHIElimination] Reuse existing COPY in predecessor basic
block
Simplifies the CFG later on, added a regression test.
This change affects many architectures but the amount of total
instructions in the test cases seems too be slightly lower.
---
llvm/lib/CodeGen/PHIElimination.cpp | 9 ++
.../AArch64/PHIElimination-reuse-copy.mir | 68 +++++++++++++++
.../AArch64/block-layout-regression.mir | 85 +++++++++++++++++++
3 files changed, 162 insertions(+)
create mode 100644 llvm/test/CodeGen/AArch64/PHIElimination-reuse-copy.mir
create mode 100644 llvm/test/CodeGen/AArch64/block-layout-regression.mir
diff --git a/llvm/lib/CodeGen/PHIElimination.cpp b/llvm/lib/CodeGen/PHIElimination.cpp
index a93a89ecaa96e..959a1711727a4 100644
--- a/llvm/lib/CodeGen/PHIElimination.cpp
+++ b/llvm/lib/CodeGen/PHIElimination.cpp
@@ -581,6 +581,15 @@ void PHIEliminationImpl::LowerPHINode(MachineBasicBlock &MBB,
continue;
}
+ // Reuse an existing copy in the block if possible.
+ if (MachineInstr *DefMI = MRI->getUniqueVRegDef(SrcReg)) {
+ if (DefMI->isCopy() && DefMI->getParent() == &opBlock &&
+ MRI->use_empty(SrcReg)) {
+ DefMI->getOperand(0).setReg(IncomingReg);
+ continue;
+ }
+ }
+
// Find a safe location to insert the copy, this may be the first terminator
// in the block (or end()).
MachineBasicBlock::iterator InsertPos =
diff --git a/llvm/test/CodeGen/AArch64/PHIElimination-reuse-copy.mir b/llvm/test/CodeGen/AArch64/PHIElimination-reuse-copy.mir
new file mode 100644
index 0000000000000..75283d9a7b14a
--- /dev/null
+++ b/llvm/test/CodeGen/AArch64/PHIElimination-reuse-copy.mir
@@ -0,0 +1,68 @@
+# NOTE: Assertions have been autogenerated by utils/update_mir_test_checks.py UTC_ARGS: --version 5
+# RUN: llc -run-pass=phi-node-elimination -mtriple=aarch64-linux-gnu -o - %s | FileCheck %s
+
+# Verify that the original COPY in bb.1 is reappropriated as the PHI source in bb.2,
+# instead of creating a new COPY with the same source register.
+
+---
+name: copy_virtual_reg
+tracksRegLiveness: true
+body: |
+ ; CHECK-LABEL: name: copy_virtual_reg
+ ; CHECK: bb.0:
+ ; CHECK-NEXT: successors: %bb.2(0x40000000), %bb.1(0x40000000)
+ ; CHECK-NEXT: liveins: $nzcv, $w0
+ ; CHECK-NEXT: {{ $}}
+ ; CHECK-NEXT: %a:gpr32 = COPY $w0
+ ; CHECK-NEXT: [[DEF:%[0-9]+]]:gpr32 = IMPLICIT_DEF
+ ; CHECK-NEXT: Bcc 8, %bb.2, implicit $nzcv
+ ; CHECK-NEXT: {{ $}}
+ ; CHECK-NEXT: bb.1:
+ ; CHECK-NEXT: successors: %bb.2(0x80000000)
+ ; CHECK-NEXT: {{ $}}
+ ; CHECK-NEXT: [[DEF:%[0-9]+]]:gpr32 = COPY %a
+ ; CHECK-NEXT: {{ $}}
+ ; CHECK-NEXT: bb.2:
+ ; CHECK-NEXT: %c:gpr32 = COPY [[DEF]]
+ ; CHECK-NEXT: %d:gpr32 = COPY %c
+ bb.0:
+ liveins: $nzcv, $w0
+ %a:gpr32 = COPY $w0
+ Bcc 8, %bb.2, implicit $nzcv
+ bb.1:
+ %b:gpr32 = COPY %a:gpr32
+ bb.2:
+ %c:gpr32 = PHI %b:gpr32, %bb.1, undef %undef:gpr32, %bb.0
+ %d:gpr32 = COPY %c:gpr32
+...
+
+---
+name: copy_physical_reg
+tracksRegLiveness: true
+body: |
+ ; CHECK-LABEL: name: copy_physical_reg
+ ; CHECK: bb.0:
+ ; CHECK-NEXT: successors: %bb.2(0x40000000), %bb.1(0x40000000)
+ ; CHECK-NEXT: liveins: $nzcv, $w0
+ ; CHECK-NEXT: {{ $}}
+ ; CHECK-NEXT: [[DEF:%[0-9]+]]:gpr32 = IMPLICIT_DEF
+ ; CHECK-NEXT: Bcc 8, %bb.2, implicit $nzcv
+ ; CHECK-NEXT: {{ $}}
+ ; CHECK-NEXT: bb.1:
+ ; CHECK-NEXT: successors: %bb.2(0x80000000)
+ ; CHECK-NEXT: {{ $}}
+ ; CHECK-NEXT: $x0 = IMPLICIT_DEF
+ ; CHECK-NEXT: [[DEF:%[0-9]+]]:gpr32 = COPY $w0
+ ; CHECK-NEXT: {{ $}}
+ ; CHECK-NEXT: bb.2:
+ ; CHECK-NEXT: %b:gpr32 = COPY [[DEF]]
+ bb.0:
+ liveins: $nzcv, $w0
+ Bcc 8, %bb.2, implicit $nzcv
+ bb.1:
+ $x0 = IMPLICIT_DEF
+ %a:gpr32 = COPY $w0
+ bb.2:
+ %b:gpr32 = PHI %a:gpr32, %bb.1, undef %undef:gpr32, %bb.0
+...
+
diff --git a/llvm/test/CodeGen/AArch64/block-layout-regression.mir b/llvm/test/CodeGen/AArch64/block-layout-regression.mir
new file mode 100644
index 0000000000000..a7ab5c029b142
--- /dev/null
+++ b/llvm/test/CodeGen/AArch64/block-layout-regression.mir
@@ -0,0 +1,85 @@
+# RUN: llc -mtriple=aarch64-linux-gnu -verify-machineinstrs -o - %s \
+# RUN: -start-before=phi-node-elimination -stop-after=branch-relaxation \
+# RUN: | FileCheck %s
+
+# Verify an optimal block layout is produced for the following nested loop, when
+# there's a PHI node in bb.5 that relies on an operand in bb.2.
+# The previous layout used a CBNZX for a null check followed by a unconditional
+# branch to bb.6, instead of a fallthrough.
+
+# [ bb.0 ENTRY ]
+# |
+# v
+# [ bb.1 ] <-------+
+# / \ |
+# v v |
+# [bb.2] [bb.3] <-+ |
+# | | | |
+# | v | |
+# | [bb.4] --+ |
+# | | |
+# v v |
+# [ bb.5 ] |
+# / \ |
+# | v |
+# | [bb.6] -----+
+# | |
+# v v
+# [ bb.7 RET ]
+
+# CHECK-LABEL: test
+# CHECK-NOT: CBNZX
+# CHECK-NOT: B %bb.
+# CHECK-COUNT-2: CBZX
+
+---
+name: test
+tracksRegLiveness: true
+body: |
+ bb.0:
+ successors: %bb.1(0x80000000); %bb.1(100.00%)
+ liveins: $x0, $w1, $x2, $x3
+ %0:gpr64all = IMPLICIT_DEF
+ %1:gpr64common = IMPLICIT_DEF
+ %2:gpr32common = IMPLICIT_DEF
+ %3:gpr32 = IMPLICIT_DEF
+ B %bb.1
+
+ bb.1:
+ successors: %bb.2(0x30000000), %bb.3(0x50000000); %bb.2(37.50%), %bb.3(62.50%)
+ %4:gpr64common = PHI undef %0:gpr64all, %bb.0, %5:gpr64common, %bb.6
+ %6:gpr64 = LDRXui undef %1:gpr64common, 0 :: (load (s64))
+ STRXui killed %4:gpr64common, undef %1:gpr64common, 0 :: (store (s64))
+ CBNZX undef %6:gpr64, %bb.3
+
+ bb.2:
+ successors: %bb.5(0x80000000); %bb.5(100.00%)
+ %7:gpr64all = COPY killed %6:gpr64
+ B %bb.5
+
+ bb.3:
+ successors: %bb.5(0x04000000), %bb.4(0x7c000000); %bb.5(3.12%), %bb.4(96.88%)
+ dead $wzr = SUBSWrr killed undef %3:gpr32, killed undef %2:gpr32common, implicit-def $nzcv
+ Bcc 12, %bb.5, implicit killed undef $nzcv
+ B %bb.4
+
+ bb.4:
+ successors: %bb.5(0x04000000), %bb.3(0x7c000000); %bb.5(3.12%), %bb.3(96.88%)
+ dead $xzr = SUBSXrr killed undef %6:gpr64, killed undef %6:gpr64, implicit-def $nzcv
+ Bcc 1, %bb.3, implicit killed undef $nzcv
+
+ bb.5:
+ successors: %bb.7(0x04000000), %bb.6(0x7c000000); %bb.7(3.12%), %bb.6(96.88%)
+ %5:gpr64common = PHI %7:gpr64all, %bb.2, undef %0:gpr64all, %bb.3, undef %0:gpr64all, %bb.4
+ CBZX undef %5:gpr64common, %bb.7
+ B %bb.6
+
+ bb.6:
+ successors: %bb.7(0x04000000), %bb.1(0x7c000000); %bb.7(3.12%), %bb.1(96.88%)
+ dead $wzr = SUBSWrr killed undef %3:gpr32, killed undef %2:gpr32common, implicit-def $nzcv
+ Bcc 12, %bb.7, implicit killed undef $nzcv
+ B %bb.1
+
+ bb.7:
+ RET_ReallyLR
+...
>From ffdb523e655fcf7ff2ee84fb1546d7d1136034ec Mon Sep 17 00:00:00 2001
From: Guy David <guyda96 at gmail.com>
Date: Tue, 18 Mar 2025 11:04:58 +0200
Subject: [PATCH 2/3] Automatically regenerated tests
---
...aarch64_be-atomic-store-outline_atomics.ll | 16 +-
.../Atomics/aarch64_be-atomic-store-rcpc.ll | 48 +-
.../Atomics/aarch64_be-atomic-store-v8a.ll | 48 +-
.../AArch64/PHIElimination-debugloc.mir | 2 +-
.../AArch64/aarch64-matrix-umull-smull.ll | 2 +-
llvm/test/CodeGen/AArch64/atomicrmw-O0.ll | 48 +-
llvm/test/CodeGen/AArch64/bfis-in-loop.ll | 2 +-
.../AArch64/complex-deinterleaving-crash.ll | 30 +-
...rleaving-reductions-predicated-scalable.ll | 28 +-
.../complex-deinterleaving-reductions.ll | 12 +-
llvm/test/CodeGen/AArch64/phi.ll | 40 +-
llvm/test/CodeGen/AArch64/pr48188.ll | 12 +-
llvm/test/CodeGen/AArch64/ragreedy-csr.ll | 22 +-
.../AArch64/ragreedy-local-interval-cost.ll | 113 +-
llvm/test/CodeGen/AArch64/reduce-or-opt.ll | 24 +-
llvm/test/CodeGen/AArch64/sink-and-fold.ll | 6 +-
llvm/test/CodeGen/AArch64/sve-lsrchain.ll | 14 +-
.../CodeGen/AArch64/sve-ptest-removal-sink.ll | 8 +-
llvm/test/CodeGen/AArch64/swifterror.ll | 16 +-
llvm/test/CodeGen/AArch64/tbl-loops.ll | 16 +-
.../AMDGPU/GlobalISel/atomicrmw_fmax.ll | 146 +-
.../AMDGPU/GlobalISel/atomicrmw_fmin.ll | 146 +-
.../divergence-temporal-divergent-i1.ll | 14 +-
.../CodeGen/AMDGPU/amdgcn.bitcast.1024bit.ll | 46921 ++++++++--------
.../CodeGen/AMDGPU/amdgcn.bitcast.128bit.ll | 189 +-
.../CodeGen/AMDGPU/amdgcn.bitcast.256bit.ll | 388 +-
.../CodeGen/AMDGPU/amdgcn.bitcast.320bit.ll | 442 +-
.../CodeGen/AMDGPU/amdgcn.bitcast.32bit.ll | 57 +-
.../CodeGen/AMDGPU/amdgcn.bitcast.512bit.ll | 4295 +-
.../CodeGen/AMDGPU/amdgcn.bitcast.576bit.ll | 96 +-
.../CodeGen/AMDGPU/amdgcn.bitcast.640bit.ll | 524 +-
.../CodeGen/AMDGPU/amdgcn.bitcast.64bit.ll | 246 +-
.../CodeGen/AMDGPU/amdgcn.bitcast.704bit.ll | 872 +-
.../CodeGen/AMDGPU/amdgcn.bitcast.768bit.ll | 514 +-
.../CodeGen/AMDGPU/amdgcn.bitcast.832bit.ll | 1350 +-
.../CodeGen/AMDGPU/amdgcn.bitcast.896bit.ll | 1136 +-
.../CodeGen/AMDGPU/amdgcn.bitcast.960bit.ll | 2843 +-
.../CodeGen/AMDGPU/amdgcn.bitcast.96bit.ll | 78 +-
.../atomic_optimizations_global_pointer.ll | 2232 +-
.../branch-folding-implicit-def-subreg.ll | 210 +-
.../buffer-fat-pointer-atomicrmw-fadd.ll | 3279 +-
.../buffer-fat-pointer-atomicrmw-fmax.ll | 2252 +-
.../buffer-fat-pointer-atomicrmw-fmin.ll | 2252 +-
llvm/test/CodeGen/AMDGPU/div_i128.ll | 1836 +-
llvm/test/CodeGen/AMDGPU/div_v2i128.ll | 228 +-
.../divergent-branch-uniform-condition.ll | 2 +-
.../CodeGen/AMDGPU/flat-atomicrmw-fmax.ll | 62 +-
.../CodeGen/AMDGPU/flat-atomicrmw-fmin.ll | 62 +-
.../CodeGen/AMDGPU/global-atomicrmw-fadd.ll | 1870 +-
.../CodeGen/AMDGPU/global-atomicrmw-fmax.ll | 1314 +-
.../CodeGen/AMDGPU/global-atomicrmw-fmin.ll | 1314 +-
.../CodeGen/AMDGPU/global-atomicrmw-fsub.ll | 1584 +-
.../AMDGPU/global_atomics_i32_system.ll | 646 +-
.../AMDGPU/global_atomics_i64_system.ll | 922 +-
.../AMDGPU/global_atomics_scan_fadd.ll | 510 +-
.../AMDGPU/global_atomics_scan_fmax.ll | 450 +-
.../AMDGPU/global_atomics_scan_fmin.ll | 450 +-
.../AMDGPU/global_atomics_scan_fsub.ll | 454 +-
.../CodeGen/AMDGPU/indirect-addressing-si.ll | 139 +-
.../AMDGPU/move-to-valu-atomicrmw-system.ll | 34 +-
llvm/test/CodeGen/AMDGPU/mul.ll | 24 +-
llvm/test/CodeGen/AMDGPU/rem_i128.ll | 1740 +-
llvm/test/CodeGen/AMDGPU/sdiv64.ll | 236 +-
llvm/test/CodeGen/AMDGPU/srem64.ll | 236 +-
llvm/test/CodeGen/AMDGPU/udiv64.ll | 212 +-
llvm/test/CodeGen/AMDGPU/urem64.ll | 180 +-
.../test/CodeGen/AMDGPU/vni8-across-blocks.ll | 83 +-
llvm/test/CodeGen/AMDGPU/wave32.ll | 8 +-
llvm/test/CodeGen/ARM/and-cmp0-sink.ll | 22 +-
llvm/test/CodeGen/ARM/cttz.ll | 92 +-
llvm/test/CodeGen/ARM/select-imm.ll | 16 +-
llvm/test/CodeGen/ARM/struct-byval-loop.ll | 16 +-
llvm/test/CodeGen/ARM/swifterror.ll | 308 +-
llvm/test/CodeGen/AVR/bug-81911.ll | 34 +-
.../LoongArch/machinelicm-address-pseudos.ll | 56 +-
.../CodeGen/PowerPC/disable-ctr-ppcf128.ll | 6 +-
llvm/test/CodeGen/PowerPC/sms-phi-2.ll | 13 +-
llvm/test/CodeGen/PowerPC/sms-phi-3.ll | 24 +-
.../PowerPC/stack-restore-with-setjmp.ll | 10 +-
llvm/test/CodeGen/PowerPC/subreg-postra-2.ll | 18 +-
llvm/test/CodeGen/PowerPC/vsx.ll | 2 +-
llvm/test/CodeGen/RISCV/abds.ll | 200 +-
llvm/test/CodeGen/RISCV/machine-pipeliner.ll | 24 +-
.../RISCV/rvv/fixed-vectors-masked-gather.ll | 120 +-
.../fixed-vectors-strided-load-store-asm.ll | 61 +-
.../RISCV/rvv/vxrm-insert-out-of-loop.ll | 26 +-
llvm/test/CodeGen/RISCV/xcvbi.ll | 60 +-
.../LowOverheadLoops/mve-tail-data-types.ll | 96 +-
.../tail-pred-disabled-in-loloops.ll | 44 +-
.../varying-outer-2d-reduction.ll | 32 +-
.../Thumb2/LowOverheadLoops/while-loops.ll | 111 +-
.../test/CodeGen/Thumb2/mve-blockplacement.ll | 21 +-
.../CodeGen/Thumb2/mve-float32regloops.ll | 43 +-
.../Thumb2/mve-laneinterleaving-reduct.ll | 8 +-
llvm/test/CodeGen/Thumb2/mve-memtp-loop.ll | 101 +-
llvm/test/CodeGen/Thumb2/mve-phireg.ll | 14 +-
llvm/test/CodeGen/Thumb2/mve-pipelineloops.ll | 85 +-
llvm/test/CodeGen/Thumb2/mve-postinc-dct.ll | 19 +-
.../CodeGen/Thumb2/mve-postinc-distribute.ll | 17 +-
llvm/test/CodeGen/Thumb2/mve-postinc-lsr.ll | 44 +-
llvm/test/CodeGen/Thumb2/mve-satmul-loops.ll | 33 +-
llvm/test/CodeGen/Thumb2/pr52817.ll | 16 +-
llvm/test/CodeGen/VE/Scalar/br_jt.ll | 38 +-
.../X86/2012-01-10-UndefExceptionEdge.ll | 4 +-
.../CodeGen/X86/AMX/amx-ldtilecfg-insert.ll | 18 +-
llvm/test/CodeGen/X86/AMX/amx-spill-merge.ll | 32 +-
llvm/test/CodeGen/X86/atomic32.ll | 126 +-
llvm/test/CodeGen/X86/atomic64.ll | 35 +-
llvm/test/CodeGen/X86/atomic6432.ll | 72 +-
.../CodeGen/X86/callbr-asm-branch-folding.ll | 8 +-
llvm/test/CodeGen/X86/callbr-asm-kill.mir | 9 +-
...r-breaks-subreg-to-reg-liveness-reduced.ll | 2 +-
llvm/test/CodeGen/X86/combine-pmuldq.ll | 8 +-
llvm/test/CodeGen/X86/fp128-select.ll | 21 +-
llvm/test/CodeGen/X86/madd.ll | 116 +-
llvm/test/CodeGen/X86/masked_load.ll | 27 +-
.../CodeGen/X86/min-legal-vector-width.ll | 30 +-
llvm/test/CodeGen/X86/pcsections-atomics.ll | 296 +-
llvm/test/CodeGen/X86/pr15705.ll | 17 +-
llvm/test/CodeGen/X86/pr32256.ll | 12 +-
llvm/test/CodeGen/X86/pr38795.ll | 15 +-
llvm/test/CodeGen/X86/pr49451.ll | 6 +-
llvm/test/CodeGen/X86/pr63108.ll | 2 +-
llvm/test/CodeGen/X86/sad.ll | 26 +-
llvm/test/CodeGen/X86/sse-scalar-fp-arith.ll | 44 +-
llvm/test/CodeGen/X86/swifterror.ll | 17 +-
.../RISCV/lsr-drop-solution.ll | 18 +-
127 files changed, 44285 insertions(+), 44091 deletions(-)
diff --git a/llvm/test/CodeGen/AArch64/Atomics/aarch64_be-atomic-store-outline_atomics.ll b/llvm/test/CodeGen/AArch64/Atomics/aarch64_be-atomic-store-outline_atomics.ll
index c1c5c53aa7df2..6c300b04508b2 100644
--- a/llvm/test/CodeGen/AArch64/Atomics/aarch64_be-atomic-store-outline_atomics.ll
+++ b/llvm/test/CodeGen/AArch64/Atomics/aarch64_be-atomic-store-outline_atomics.ll
@@ -118,8 +118,8 @@ define dso_local void @store_atomic_i64_aligned_seq_cst(i64 %value, ptr %ptr) {
define dso_local void @store_atomic_i128_aligned_unordered(i128 %value, ptr %ptr) {
; -O0-LABEL: store_atomic_i128_aligned_unordered:
; -O0: bl __aarch64_cas16_relax
-; -O0: subs x10, x10, x11
-; -O0: ccmp x8, x9, #0, eq
+; -O0: subs x9, x0, x9
+; -O0: ccmp x1, x8, #0, eq
;
; -O1-LABEL: store_atomic_i128_aligned_unordered:
; -O1: ldxp xzr, x8, [x2]
@@ -131,8 +131,8 @@ define dso_local void @store_atomic_i128_aligned_unordered(i128 %value, ptr %ptr
define dso_local void @store_atomic_i128_aligned_monotonic(i128 %value, ptr %ptr) {
; -O0-LABEL: store_atomic_i128_aligned_monotonic:
; -O0: bl __aarch64_cas16_relax
-; -O0: subs x10, x10, x11
-; -O0: ccmp x8, x9, #0, eq
+; -O0: subs x9, x0, x9
+; -O0: ccmp x1, x8, #0, eq
;
; -O1-LABEL: store_atomic_i128_aligned_monotonic:
; -O1: ldxp xzr, x8, [x2]
@@ -144,8 +144,8 @@ define dso_local void @store_atomic_i128_aligned_monotonic(i128 %value, ptr %ptr
define dso_local void @store_atomic_i128_aligned_release(i128 %value, ptr %ptr) {
; -O0-LABEL: store_atomic_i128_aligned_release:
; -O0: bl __aarch64_cas16_rel
-; -O0: subs x10, x10, x11
-; -O0: ccmp x8, x9, #0, eq
+; -O0: subs x9, x0, x9
+; -O0: ccmp x1, x8, #0, eq
;
; -O1-LABEL: store_atomic_i128_aligned_release:
; -O1: ldxp xzr, x8, [x2]
@@ -157,8 +157,8 @@ define dso_local void @store_atomic_i128_aligned_release(i128 %value, ptr %ptr)
define dso_local void @store_atomic_i128_aligned_seq_cst(i128 %value, ptr %ptr) {
; -O0-LABEL: store_atomic_i128_aligned_seq_cst:
; -O0: bl __aarch64_cas16_acq_rel
-; -O0: subs x10, x10, x11
-; -O0: ccmp x8, x9, #0, eq
+; -O0: subs x9, x0, x9
+; -O0: ccmp x1, x8, #0, eq
;
; -O1-LABEL: store_atomic_i128_aligned_seq_cst:
; -O1: ldaxp xzr, x8, [x2]
diff --git a/llvm/test/CodeGen/AArch64/Atomics/aarch64_be-atomic-store-rcpc.ll b/llvm/test/CodeGen/AArch64/Atomics/aarch64_be-atomic-store-rcpc.ll
index d1047d84e2956..2a7bbad9d6454 100644
--- a/llvm/test/CodeGen/AArch64/Atomics/aarch64_be-atomic-store-rcpc.ll
+++ b/llvm/test/CodeGen/AArch64/Atomics/aarch64_be-atomic-store-rcpc.ll
@@ -117,13 +117,13 @@ define dso_local void @store_atomic_i64_aligned_seq_cst(i64 %value, ptr %ptr) {
define dso_local void @store_atomic_i128_aligned_unordered(i128 %value, ptr %ptr) {
; -O0-LABEL: store_atomic_i128_aligned_unordered:
-; -O0: ldxp x10, x12, [x9]
+; -O0: ldxp x8, x10, [x13]
+; -O0: cmp x8, x9
; -O0: cmp x10, x11
-; -O0: cmp x12, x13
-; -O0: stxp w8, x14, x15, [x9]
-; -O0: stxp w8, x10, x12, [x9]
-; -O0: subs x12, x12, x13
-; -O0: ccmp x10, x11, #0, eq
+; -O0: stxp w12, x14, x15, [x13]
+; -O0: stxp w12, x8, x10, [x13]
+; -O0: subs x10, x10, x11
+; -O0: ccmp x8, x9, #0, eq
;
; -O1-LABEL: store_atomic_i128_aligned_unordered:
; -O1: ldxp xzr, x8, [x2]
@@ -134,13 +134,13 @@ define dso_local void @store_atomic_i128_aligned_unordered(i128 %value, ptr %ptr
define dso_local void @store_atomic_i128_aligned_monotonic(i128 %value, ptr %ptr) {
; -O0-LABEL: store_atomic_i128_aligned_monotonic:
-; -O0: ldxp x10, x12, [x9]
+; -O0: ldxp x8, x10, [x13]
+; -O0: cmp x8, x9
; -O0: cmp x10, x11
-; -O0: cmp x12, x13
-; -O0: stxp w8, x14, x15, [x9]
-; -O0: stxp w8, x10, x12, [x9]
-; -O0: subs x12, x12, x13
-; -O0: ccmp x10, x11, #0, eq
+; -O0: stxp w12, x14, x15, [x13]
+; -O0: stxp w12, x8, x10, [x13]
+; -O0: subs x10, x10, x11
+; -O0: ccmp x8, x9, #0, eq
;
; -O1-LABEL: store_atomic_i128_aligned_monotonic:
; -O1: ldxp xzr, x8, [x2]
@@ -151,13 +151,13 @@ define dso_local void @store_atomic_i128_aligned_monotonic(i128 %value, ptr %ptr
define dso_local void @store_atomic_i128_aligned_release(i128 %value, ptr %ptr) {
; -O0-LABEL: store_atomic_i128_aligned_release:
-; -O0: ldxp x10, x12, [x9]
+; -O0: ldxp x8, x10, [x13]
+; -O0: cmp x8, x9
; -O0: cmp x10, x11
-; -O0: cmp x12, x13
-; -O0: stlxp w8, x14, x15, [x9]
-; -O0: stlxp w8, x10, x12, [x9]
-; -O0: subs x12, x12, x13
-; -O0: ccmp x10, x11, #0, eq
+; -O0: stlxp w12, x14, x15, [x13]
+; -O0: stlxp w12, x8, x10, [x13]
+; -O0: subs x10, x10, x11
+; -O0: ccmp x8, x9, #0, eq
;
; -O1-LABEL: store_atomic_i128_aligned_release:
; -O1: ldxp xzr, x8, [x2]
@@ -168,13 +168,13 @@ define dso_local void @store_atomic_i128_aligned_release(i128 %value, ptr %ptr)
define dso_local void @store_atomic_i128_aligned_seq_cst(i128 %value, ptr %ptr) {
; -O0-LABEL: store_atomic_i128_aligned_seq_cst:
-; -O0: ldaxp x10, x12, [x9]
+; -O0: ldaxp x8, x10, [x13]
+; -O0: cmp x8, x9
; -O0: cmp x10, x11
-; -O0: cmp x12, x13
-; -O0: stlxp w8, x14, x15, [x9]
-; -O0: stlxp w8, x10, x12, [x9]
-; -O0: subs x12, x12, x13
-; -O0: ccmp x10, x11, #0, eq
+; -O0: stlxp w12, x14, x15, [x13]
+; -O0: stlxp w12, x8, x10, [x13]
+; -O0: subs x10, x10, x11
+; -O0: ccmp x8, x9, #0, eq
;
; -O1-LABEL: store_atomic_i128_aligned_seq_cst:
; -O1: ldaxp xzr, x8, [x2]
diff --git a/llvm/test/CodeGen/AArch64/Atomics/aarch64_be-atomic-store-v8a.ll b/llvm/test/CodeGen/AArch64/Atomics/aarch64_be-atomic-store-v8a.ll
index 1a79c73355143..493bc742f7663 100644
--- a/llvm/test/CodeGen/AArch64/Atomics/aarch64_be-atomic-store-v8a.ll
+++ b/llvm/test/CodeGen/AArch64/Atomics/aarch64_be-atomic-store-v8a.ll
@@ -117,13 +117,13 @@ define dso_local void @store_atomic_i64_aligned_seq_cst(i64 %value, ptr %ptr) {
define dso_local void @store_atomic_i128_aligned_unordered(i128 %value, ptr %ptr) {
; -O0-LABEL: store_atomic_i128_aligned_unordered:
-; -O0: ldxp x10, x12, [x9]
+; -O0: ldxp x8, x10, [x13]
+; -O0: cmp x8, x9
; -O0: cmp x10, x11
-; -O0: cmp x12, x13
-; -O0: stxp w8, x14, x15, [x9]
-; -O0: stxp w8, x10, x12, [x9]
-; -O0: subs x12, x12, x13
-; -O0: ccmp x10, x11, #0, eq
+; -O0: stxp w12, x14, x15, [x13]
+; -O0: stxp w12, x8, x10, [x13]
+; -O0: subs x10, x10, x11
+; -O0: ccmp x8, x9, #0, eq
;
; -O1-LABEL: store_atomic_i128_aligned_unordered:
; -O1: ldxp xzr, x8, [x2]
@@ -134,13 +134,13 @@ define dso_local void @store_atomic_i128_aligned_unordered(i128 %value, ptr %ptr
define dso_local void @store_atomic_i128_aligned_monotonic(i128 %value, ptr %ptr) {
; -O0-LABEL: store_atomic_i128_aligned_monotonic:
-; -O0: ldxp x10, x12, [x9]
+; -O0: ldxp x8, x10, [x13]
+; -O0: cmp x8, x9
; -O0: cmp x10, x11
-; -O0: cmp x12, x13
-; -O0: stxp w8, x14, x15, [x9]
-; -O0: stxp w8, x10, x12, [x9]
-; -O0: subs x12, x12, x13
-; -O0: ccmp x10, x11, #0, eq
+; -O0: stxp w12, x14, x15, [x13]
+; -O0: stxp w12, x8, x10, [x13]
+; -O0: subs x10, x10, x11
+; -O0: ccmp x8, x9, #0, eq
;
; -O1-LABEL: store_atomic_i128_aligned_monotonic:
; -O1: ldxp xzr, x8, [x2]
@@ -151,13 +151,13 @@ define dso_local void @store_atomic_i128_aligned_monotonic(i128 %value, ptr %ptr
define dso_local void @store_atomic_i128_aligned_release(i128 %value, ptr %ptr) {
; -O0-LABEL: store_atomic_i128_aligned_release:
-; -O0: ldxp x10, x12, [x9]
+; -O0: ldxp x8, x10, [x13]
+; -O0: cmp x8, x9
; -O0: cmp x10, x11
-; -O0: cmp x12, x13
-; -O0: stlxp w8, x14, x15, [x9]
-; -O0: stlxp w8, x10, x12, [x9]
-; -O0: subs x12, x12, x13
-; -O0: ccmp x10, x11, #0, eq
+; -O0: stlxp w12, x14, x15, [x13]
+; -O0: stlxp w12, x8, x10, [x13]
+; -O0: subs x10, x10, x11
+; -O0: ccmp x8, x9, #0, eq
;
; -O1-LABEL: store_atomic_i128_aligned_release:
; -O1: ldxp xzr, x8, [x2]
@@ -168,13 +168,13 @@ define dso_local void @store_atomic_i128_aligned_release(i128 %value, ptr %ptr)
define dso_local void @store_atomic_i128_aligned_seq_cst(i128 %value, ptr %ptr) {
; -O0-LABEL: store_atomic_i128_aligned_seq_cst:
-; -O0: ldaxp x10, x12, [x9]
+; -O0: ldaxp x8, x10, [x13]
+; -O0: cmp x8, x9
; -O0: cmp x10, x11
-; -O0: cmp x12, x13
-; -O0: stlxp w8, x14, x15, [x9]
-; -O0: stlxp w8, x10, x12, [x9]
-; -O0: subs x12, x12, x13
-; -O0: ccmp x10, x11, #0, eq
+; -O0: stlxp w12, x14, x15, [x13]
+; -O0: stlxp w12, x8, x10, [x13]
+; -O0: subs x10, x10, x11
+; -O0: ccmp x8, x9, #0, eq
;
; -O1-LABEL: store_atomic_i128_aligned_seq_cst:
; -O1: ldaxp xzr, x8, [x2]
diff --git a/llvm/test/CodeGen/AArch64/PHIElimination-debugloc.mir b/llvm/test/CodeGen/AArch64/PHIElimination-debugloc.mir
index 01c44e3f253bb..993d1c1f1b5f0 100644
--- a/llvm/test/CodeGen/AArch64/PHIElimination-debugloc.mir
+++ b/llvm/test/CodeGen/AArch64/PHIElimination-debugloc.mir
@@ -37,7 +37,7 @@ body: |
bb.1:
%x:gpr32 = COPY $wzr
; Test that the debug location is not copied into bb1!
- ; CHECK: %3:gpr32 = COPY killed %x{{$}}
+ ; CHECK: %3:gpr32 = COPY $wzr
; CHECK-LABEL: bb.2:
bb.2:
%y:gpr32 = PHI %x:gpr32, %bb.1, undef %undef:gpr32, %bb.0, debug-location !14
diff --git a/llvm/test/CodeGen/AArch64/aarch64-matrix-umull-smull.ll b/llvm/test/CodeGen/AArch64/aarch64-matrix-umull-smull.ll
index 8655bb1292ef7..ca1052a769408 100644
--- a/llvm/test/CodeGen/AArch64/aarch64-matrix-umull-smull.ll
+++ b/llvm/test/CodeGen/AArch64/aarch64-matrix-umull-smull.ll
@@ -583,8 +583,8 @@ define i16 @red_mla_dup_ext_u8_s8_s16(ptr noalias nocapture noundef readonly %A,
; CHECK-SD-NEXT: mov w10, w2
; CHECK-SD-NEXT: b.hi .LBB5_4
; CHECK-SD-NEXT: // %bb.2:
-; CHECK-SD-NEXT: mov x11, xzr
; CHECK-SD-NEXT: mov w8, wzr
+; CHECK-SD-NEXT: mov x11, xzr
; CHECK-SD-NEXT: b .LBB5_7
; CHECK-SD-NEXT: .LBB5_3:
; CHECK-SD-NEXT: mov w8, wzr
diff --git a/llvm/test/CodeGen/AArch64/atomicrmw-O0.ll b/llvm/test/CodeGen/AArch64/atomicrmw-O0.ll
index 71e0250b36972..9fd27edae3176 100644
--- a/llvm/test/CodeGen/AArch64/atomicrmw-O0.ll
+++ b/llvm/test/CodeGen/AArch64/atomicrmw-O0.ll
@@ -199,16 +199,16 @@ define i128 @test_rmw_add_128(ptr %dst) {
; NOLSE-NEXT: sub sp, sp, #48
; NOLSE-NEXT: .cfi_def_cfa_offset 48
; NOLSE-NEXT: str x0, [sp, #24] // 8-byte Folded Spill
-; NOLSE-NEXT: ldr x8, [x0, #8]
-; NOLSE-NEXT: ldr x9, [x0]
+; NOLSE-NEXT: ldr x9, [x0, #8]
+; NOLSE-NEXT: ldr x8, [x0]
; NOLSE-NEXT: str x9, [sp, #32] // 8-byte Folded Spill
; NOLSE-NEXT: str x8, [sp, #40] // 8-byte Folded Spill
; NOLSE-NEXT: b .LBB4_1
; NOLSE-NEXT: .LBB4_1: // %atomicrmw.start
; NOLSE-NEXT: // =>This Loop Header: Depth=1
; NOLSE-NEXT: // Child Loop BB4_2 Depth 2
-; NOLSE-NEXT: ldr x13, [sp, #40] // 8-byte Folded Reload
-; NOLSE-NEXT: ldr x11, [sp, #32] // 8-byte Folded Reload
+; NOLSE-NEXT: ldr x13, [sp, #32] // 8-byte Folded Reload
+; NOLSE-NEXT: ldr x11, [sp, #40] // 8-byte Folded Reload
; NOLSE-NEXT: ldr x9, [sp, #24] // 8-byte Folded Reload
; NOLSE-NEXT: adds x14, x11, #1
; NOLSE-NEXT: cinc x15, x13, hs
@@ -238,8 +238,8 @@ define i128 @test_rmw_add_128(ptr %dst) {
; NOLSE-NEXT: str x9, [sp, #16] // 8-byte Folded Spill
; NOLSE-NEXT: subs x12, x12, x13
; NOLSE-NEXT: ccmp x10, x11, #0, eq
-; NOLSE-NEXT: str x9, [sp, #32] // 8-byte Folded Spill
-; NOLSE-NEXT: str x8, [sp, #40] // 8-byte Folded Spill
+; NOLSE-NEXT: str x9, [sp, #40] // 8-byte Folded Spill
+; NOLSE-NEXT: str x8, [sp, #32] // 8-byte Folded Spill
; NOLSE-NEXT: b.ne .LBB4_1
; NOLSE-NEXT: b .LBB4_6
; NOLSE-NEXT: .LBB4_6: // %atomicrmw.end
@@ -253,15 +253,15 @@ define i128 @test_rmw_add_128(ptr %dst) {
; LSE-NEXT: sub sp, sp, #48
; LSE-NEXT: .cfi_def_cfa_offset 48
; LSE-NEXT: str x0, [sp, #24] // 8-byte Folded Spill
-; LSE-NEXT: ldr x8, [x0, #8]
-; LSE-NEXT: ldr x9, [x0]
+; LSE-NEXT: ldr x9, [x0, #8]
+; LSE-NEXT: ldr x8, [x0]
; LSE-NEXT: str x9, [sp, #32] // 8-byte Folded Spill
; LSE-NEXT: str x8, [sp, #40] // 8-byte Folded Spill
; LSE-NEXT: b .LBB4_1
; LSE-NEXT: .LBB4_1: // %atomicrmw.start
; LSE-NEXT: // =>This Inner Loop Header: Depth=1
-; LSE-NEXT: ldr x11, [sp, #40] // 8-byte Folded Reload
-; LSE-NEXT: ldr x10, [sp, #32] // 8-byte Folded Reload
+; LSE-NEXT: ldr x11, [sp, #32] // 8-byte Folded Reload
+; LSE-NEXT: ldr x10, [sp, #40] // 8-byte Folded Reload
; LSE-NEXT: ldr x8, [sp, #24] // 8-byte Folded Reload
; LSE-NEXT: mov x0, x10
; LSE-NEXT: mov x1, x11
@@ -276,8 +276,8 @@ define i128 @test_rmw_add_128(ptr %dst) {
; LSE-NEXT: str x8, [sp, #16] // 8-byte Folded Spill
; LSE-NEXT: subs x11, x8, x11
; LSE-NEXT: ccmp x9, x10, #0, eq
-; LSE-NEXT: str x9, [sp, #32] // 8-byte Folded Spill
-; LSE-NEXT: str x8, [sp, #40] // 8-byte Folded Spill
+; LSE-NEXT: str x9, [sp, #40] // 8-byte Folded Spill
+; LSE-NEXT: str x8, [sp, #32] // 8-byte Folded Spill
; LSE-NEXT: b.ne .LBB4_1
; LSE-NEXT: b .LBB4_2
; LSE-NEXT: .LBB4_2: // %atomicrmw.end
@@ -573,16 +573,16 @@ define i128 @test_rmw_nand_128(ptr %dst) {
; NOLSE-NEXT: sub sp, sp, #48
; NOLSE-NEXT: .cfi_def_cfa_offset 48
; NOLSE-NEXT: str x0, [sp, #24] // 8-byte Folded Spill
-; NOLSE-NEXT: ldr x8, [x0, #8]
-; NOLSE-NEXT: ldr x9, [x0]
+; NOLSE-NEXT: ldr x9, [x0, #8]
+; NOLSE-NEXT: ldr x8, [x0]
; NOLSE-NEXT: str x9, [sp, #32] // 8-byte Folded Spill
; NOLSE-NEXT: str x8, [sp, #40] // 8-byte Folded Spill
; NOLSE-NEXT: b .LBB9_1
; NOLSE-NEXT: .LBB9_1: // %atomicrmw.start
; NOLSE-NEXT: // =>This Loop Header: Depth=1
; NOLSE-NEXT: // Child Loop BB9_2 Depth 2
-; NOLSE-NEXT: ldr x13, [sp, #40] // 8-byte Folded Reload
-; NOLSE-NEXT: ldr x11, [sp, #32] // 8-byte Folded Reload
+; NOLSE-NEXT: ldr x13, [sp, #32] // 8-byte Folded Reload
+; NOLSE-NEXT: ldr x11, [sp, #40] // 8-byte Folded Reload
; NOLSE-NEXT: ldr x9, [sp, #24] // 8-byte Folded Reload
; NOLSE-NEXT: mov w8, w11
; NOLSE-NEXT: mvn w10, w8
@@ -616,8 +616,8 @@ define i128 @test_rmw_nand_128(ptr %dst) {
; NOLSE-NEXT: str x9, [sp, #16] // 8-byte Folded Spill
; NOLSE-NEXT: subs x12, x12, x13
; NOLSE-NEXT: ccmp x10, x11, #0, eq
-; NOLSE-NEXT: str x9, [sp, #32] // 8-byte Folded Spill
-; NOLSE-NEXT: str x8, [sp, #40] // 8-byte Folded Spill
+; NOLSE-NEXT: str x9, [sp, #40] // 8-byte Folded Spill
+; NOLSE-NEXT: str x8, [sp, #32] // 8-byte Folded Spill
; NOLSE-NEXT: b.ne .LBB9_1
; NOLSE-NEXT: b .LBB9_6
; NOLSE-NEXT: .LBB9_6: // %atomicrmw.end
@@ -631,15 +631,15 @@ define i128 @test_rmw_nand_128(ptr %dst) {
; LSE-NEXT: sub sp, sp, #48
; LSE-NEXT: .cfi_def_cfa_offset 48
; LSE-NEXT: str x0, [sp, #24] // 8-byte Folded Spill
-; LSE-NEXT: ldr x8, [x0, #8]
-; LSE-NEXT: ldr x9, [x0]
+; LSE-NEXT: ldr x9, [x0, #8]
+; LSE-NEXT: ldr x8, [x0]
; LSE-NEXT: str x9, [sp, #32] // 8-byte Folded Spill
; LSE-NEXT: str x8, [sp, #40] // 8-byte Folded Spill
; LSE-NEXT: b .LBB9_1
; LSE-NEXT: .LBB9_1: // %atomicrmw.start
; LSE-NEXT: // =>This Inner Loop Header: Depth=1
-; LSE-NEXT: ldr x11, [sp, #40] // 8-byte Folded Reload
-; LSE-NEXT: ldr x10, [sp, #32] // 8-byte Folded Reload
+; LSE-NEXT: ldr x11, [sp, #32] // 8-byte Folded Reload
+; LSE-NEXT: ldr x10, [sp, #40] // 8-byte Folded Reload
; LSE-NEXT: ldr x8, [sp, #24] // 8-byte Folded Reload
; LSE-NEXT: mov x0, x10
; LSE-NEXT: mov x1, x11
@@ -658,8 +658,8 @@ define i128 @test_rmw_nand_128(ptr %dst) {
; LSE-NEXT: str x8, [sp, #16] // 8-byte Folded Spill
; LSE-NEXT: subs x11, x8, x11
; LSE-NEXT: ccmp x9, x10, #0, eq
-; LSE-NEXT: str x9, [sp, #32] // 8-byte Folded Spill
-; LSE-NEXT: str x8, [sp, #40] // 8-byte Folded Spill
+; LSE-NEXT: str x9, [sp, #40] // 8-byte Folded Spill
+; LSE-NEXT: str x8, [sp, #32] // 8-byte Folded Spill
; LSE-NEXT: b.ne .LBB9_1
; LSE-NEXT: b .LBB9_2
; LSE-NEXT: .LBB9_2: // %atomicrmw.end
diff --git a/llvm/test/CodeGen/AArch64/bfis-in-loop.ll b/llvm/test/CodeGen/AArch64/bfis-in-loop.ll
index 43d49da1abd21..b0339222bc2df 100644
--- a/llvm/test/CodeGen/AArch64/bfis-in-loop.ll
+++ b/llvm/test/CodeGen/AArch64/bfis-in-loop.ll
@@ -14,8 +14,8 @@ define i64 @bfis_in_loop_zero() {
; CHECK-LABEL: bfis_in_loop_zero:
; CHECK: // %bb.0: // %entry
; CHECK-NEXT: adrp x9, :got:global
-; CHECK-NEXT: mov x0, xzr
; CHECK-NEXT: mov w8, wzr
+; CHECK-NEXT: mov x0, xzr
; CHECK-NEXT: ldr x9, [x9, :got_lo12:global]
; CHECK-NEXT: mov w10, #65536 // =0x10000
; CHECK-NEXT: ldr x9, [x9]
diff --git a/llvm/test/CodeGen/AArch64/complex-deinterleaving-crash.ll b/llvm/test/CodeGen/AArch64/complex-deinterleaving-crash.ll
index 7542e9c4b8f5b..327d0749c7dbf 100644
--- a/llvm/test/CodeGen/AArch64/complex-deinterleaving-crash.ll
+++ b/llvm/test/CodeGen/AArch64/complex-deinterleaving-crash.ll
@@ -35,10 +35,10 @@ define i32 @check_deinterleaving_has_deinterleave(ptr %a) {
; CHECK-LABEL: check_deinterleaving_has_deinterleave:
; CHECK: // %bb.0: // %entry
; CHECK-NEXT: movi v0.2d, #0000000000000000
-; CHECK-NEXT: movi v1.4s, #1
+; CHECK-NEXT: movi v2.4s, #1
; CHECK-NEXT: add x8, x0, #16
; CHECK-NEXT: movi v3.2d, #0000000000000000
-; CHECK-NEXT: movi v2.2d, #0000000000000000
+; CHECK-NEXT: movi v1.2d, #0000000000000000
; CHECK-NEXT: mov w9, #32 // =0x20
; CHECK-NEXT: movi v4.2d, #0000000000000000
; CHECK-NEXT: movi v5.2d, #0000000000000000
@@ -64,16 +64,16 @@ define i32 @check_deinterleaving_has_deinterleave(ptr %a) {
; CHECK-NEXT: ushll v24.4s, v18.4h, #0
; CHECK-NEXT: ushll2 v18.4s, v18.8h, #0
; CHECK-NEXT: ushll v20.4s, v20.4h, #0
-; CHECK-NEXT: and v21.16b, v21.16b, v1.16b
-; CHECK-NEXT: and v19.16b, v19.16b, v1.16b
-; CHECK-NEXT: and v22.16b, v22.16b, v1.16b
-; CHECK-NEXT: and v17.16b, v17.16b, v1.16b
-; CHECK-NEXT: and v23.16b, v23.16b, v1.16b
-; CHECK-NEXT: and v24.16b, v24.16b, v1.16b
-; CHECK-NEXT: and v18.16b, v18.16b, v1.16b
-; CHECK-NEXT: and v20.16b, v20.16b, v1.16b
+; CHECK-NEXT: and v21.16b, v21.16b, v2.16b
+; CHECK-NEXT: and v19.16b, v19.16b, v2.16b
+; CHECK-NEXT: and v22.16b, v22.16b, v2.16b
+; CHECK-NEXT: and v17.16b, v17.16b, v2.16b
+; CHECK-NEXT: and v23.16b, v23.16b, v2.16b
+; CHECK-NEXT: and v24.16b, v24.16b, v2.16b
+; CHECK-NEXT: and v18.16b, v18.16b, v2.16b
+; CHECK-NEXT: and v20.16b, v20.16b, v2.16b
; CHECK-NEXT: add v4.4s, v4.4s, v19.4s
-; CHECK-NEXT: add v2.4s, v2.4s, v21.4s
+; CHECK-NEXT: add v1.4s, v1.4s, v21.4s
; CHECK-NEXT: add v0.4s, v0.4s, v22.4s
; CHECK-NEXT: add v3.4s, v3.4s, v17.4s
; CHECK-NEXT: add v16.4s, v16.4s, v23.4s
@@ -82,12 +82,12 @@ define i32 @check_deinterleaving_has_deinterleave(ptr %a) {
; CHECK-NEXT: add v7.4s, v7.4s, v18.4s
; CHECK-NEXT: b.ne .LBB1_1
; CHECK-NEXT: // %bb.2: // %middle.block
-; CHECK-NEXT: add v1.4s, v7.4s, v3.4s
+; CHECK-NEXT: add v2.4s, v7.4s, v3.4s
; CHECK-NEXT: add v3.4s, v16.4s, v4.4s
; CHECK-NEXT: add v0.4s, v5.4s, v0.4s
-; CHECK-NEXT: add v2.4s, v6.4s, v2.4s
-; CHECK-NEXT: add v1.4s, v1.4s, v3.4s
-; CHECK-NEXT: add v0.4s, v0.4s, v2.4s
+; CHECK-NEXT: add v1.4s, v6.4s, v1.4s
+; CHECK-NEXT: add v0.4s, v0.4s, v1.4s
+; CHECK-NEXT: add v1.4s, v2.4s, v3.4s
; CHECK-NEXT: add v0.4s, v0.4s, v1.4s
; CHECK-NEXT: addv s0, v0.4s
; CHECK-NEXT: fmov w0, s0
diff --git a/llvm/test/CodeGen/AArch64/complex-deinterleaving-reductions-predicated-scalable.ll b/llvm/test/CodeGen/AArch64/complex-deinterleaving-reductions-predicated-scalable.ll
index 880bd2904154c..69df51a27d10c 100644
--- a/llvm/test/CodeGen/AArch64/complex-deinterleaving-reductions-predicated-scalable.ll
+++ b/llvm/test/CodeGen/AArch64/complex-deinterleaving-reductions-predicated-scalable.ll
@@ -18,8 +18,8 @@ define %"class.std::complex" @complex_mul_v2f64(ptr %a, ptr %b) {
; CHECK-NEXT: mov w8, #100 // =0x64
; CHECK-NEXT: cntd x9
; CHECK-NEXT: whilelo p1.d, xzr, x8
-; CHECK-NEXT: rdvl x10, #2
-; CHECK-NEXT: mov x11, x9
+; CHECK-NEXT: mov x10, x9
+; CHECK-NEXT: rdvl x11, #2
; CHECK-NEXT: ptrue p0.d
; CHECK-NEXT: zip2 z0.d, z1.d, z1.d
; CHECK-NEXT: zip1 z1.d, z1.d, z1.d
@@ -33,16 +33,16 @@ define %"class.std::complex" @complex_mul_v2f64(ptr %a, ptr %b) {
; CHECK-NEXT: ld1d { z4.d }, p2/z, [x1, #1, mul vl]
; CHECK-NEXT: ld1d { z3.d }, p1/z, [x0]
; CHECK-NEXT: ld1d { z5.d }, p1/z, [x1]
-; CHECK-NEXT: add x1, x1, x10
-; CHECK-NEXT: add x0, x0, x10
+; CHECK-NEXT: add x1, x1, x11
+; CHECK-NEXT: add x0, x0, x11
; CHECK-NEXT: fcmla z7.d, p0/m, z4.d, z2.d, #0
; CHECK-NEXT: fcmla z6.d, p0/m, z5.d, z3.d, #0
; CHECK-NEXT: fcmla z7.d, p0/m, z4.d, z2.d, #90
; CHECK-NEXT: fcmla z6.d, p0/m, z5.d, z3.d, #90
; CHECK-NEXT: mov z0.d, p2/m, z7.d
; CHECK-NEXT: mov z1.d, p1/m, z6.d
-; CHECK-NEXT: whilelo p1.d, x11, x8
-; CHECK-NEXT: add x11, x11, x9
+; CHECK-NEXT: whilelo p1.d, x10, x8
+; CHECK-NEXT: add x10, x10, x9
; CHECK-NEXT: b.mi .LBB0_1
; CHECK-NEXT: // %bb.2: // %exit.block
; CHECK-NEXT: uzp1 z2.d, z1.d, z0.d
@@ -217,10 +217,10 @@ define %"class.std::complex" @complex_mul_predicated_x2_v2f64(ptr %a, ptr %b, pt
; CHECK-NEXT: mov w8, #100 // =0x64
; CHECK-NEXT: cntd x9
; CHECK-NEXT: whilelo p1.d, xzr, x8
-; CHECK-NEXT: rdvl x10, #2
-; CHECK-NEXT: cnth x11
+; CHECK-NEXT: mov x10, x9
+; CHECK-NEXT: rdvl x11, #2
; CHECK-NEXT: ptrue p0.d
-; CHECK-NEXT: mov x12, x9
+; CHECK-NEXT: cnth x12
; CHECK-NEXT: zip2 z0.d, z1.d, z1.d
; CHECK-NEXT: zip1 z1.d, z1.d, z1.d
; CHECK-NEXT: .LBB2_1: // %vector.body
@@ -228,7 +228,7 @@ define %"class.std::complex" @complex_mul_predicated_x2_v2f64(ptr %a, ptr %b, pt
; CHECK-NEXT: ld1w { z2.d }, p1/z, [x2]
; CHECK-NEXT: mov z6.d, z1.d
; CHECK-NEXT: mov z7.d, z0.d
-; CHECK-NEXT: add x2, x2, x11
+; CHECK-NEXT: add x2, x2, x12
; CHECK-NEXT: and z2.d, z2.d, #0xffffffff
; CHECK-NEXT: cmpne p1.d, p1/z, z2.d, #0
; CHECK-NEXT: zip2 p2.d, p1.d, p1.d
@@ -237,16 +237,16 @@ define %"class.std::complex" @complex_mul_predicated_x2_v2f64(ptr %a, ptr %b, pt
; CHECK-NEXT: ld1d { z4.d }, p2/z, [x1, #1, mul vl]
; CHECK-NEXT: ld1d { z3.d }, p1/z, [x0]
; CHECK-NEXT: ld1d { z5.d }, p1/z, [x1]
-; CHECK-NEXT: add x1, x1, x10
-; CHECK-NEXT: add x0, x0, x10
+; CHECK-NEXT: add x1, x1, x11
+; CHECK-NEXT: add x0, x0, x11
; CHECK-NEXT: fcmla z7.d, p0/m, z4.d, z2.d, #0
; CHECK-NEXT: fcmla z6.d, p0/m, z5.d, z3.d, #0
; CHECK-NEXT: fcmla z7.d, p0/m, z4.d, z2.d, #90
; CHECK-NEXT: fcmla z6.d, p0/m, z5.d, z3.d, #90
; CHECK-NEXT: mov z0.d, p2/m, z7.d
; CHECK-NEXT: mov z1.d, p1/m, z6.d
-; CHECK-NEXT: whilelo p1.d, x12, x8
-; CHECK-NEXT: add x12, x12, x9
+; CHECK-NEXT: whilelo p1.d, x10, x8
+; CHECK-NEXT: add x10, x10, x9
; CHECK-NEXT: b.mi .LBB2_1
; CHECK-NEXT: // %bb.2: // %exit.block
; CHECK-NEXT: uzp1 z2.d, z1.d, z0.d
diff --git a/llvm/test/CodeGen/AArch64/complex-deinterleaving-reductions.ll b/llvm/test/CodeGen/AArch64/complex-deinterleaving-reductions.ll
index aed3072bb4af3..c977869d2ce95 100644
--- a/llvm/test/CodeGen/AArch64/complex-deinterleaving-reductions.ll
+++ b/llvm/test/CodeGen/AArch64/complex-deinterleaving-reductions.ll
@@ -25,14 +25,14 @@ define dso_local %"struct.std::complex" @complex_mul_v2f64(ptr %a, ptr %b) {
; CHECK-NEXT: ldp q3, q2, [x9]
; CHECK-NEXT: cmp x8, #1600
; CHECK-NEXT: ldp q5, q4, [x10]
-; CHECK-NEXT: fcmla v0.2d, v5.2d, v3.2d, #0
-; CHECK-NEXT: fcmla v1.2d, v4.2d, v2.2d, #0
-; CHECK-NEXT: fcmla v0.2d, v5.2d, v3.2d, #90
-; CHECK-NEXT: fcmla v1.2d, v4.2d, v2.2d, #90
+; CHECK-NEXT: fcmla v1.2d, v5.2d, v3.2d, #0
+; CHECK-NEXT: fcmla v0.2d, v4.2d, v2.2d, #0
+; CHECK-NEXT: fcmla v1.2d, v5.2d, v3.2d, #90
+; CHECK-NEXT: fcmla v0.2d, v4.2d, v2.2d, #90
; CHECK-NEXT: b.ne .LBB0_1
; CHECK-NEXT: // %bb.2: // %middle.block
-; CHECK-NEXT: zip2 v2.2d, v0.2d, v1.2d
-; CHECK-NEXT: zip1 v0.2d, v0.2d, v1.2d
+; CHECK-NEXT: zip2 v2.2d, v1.2d, v0.2d
+; CHECK-NEXT: zip1 v0.2d, v1.2d, v0.2d
; CHECK-NEXT: faddp d0, v0.2d
; CHECK-NEXT: faddp d1, v2.2d
; CHECK-NEXT: ret
diff --git a/llvm/test/CodeGen/AArch64/phi.ll b/llvm/test/CodeGen/AArch64/phi.ll
index 02842c04bf7bf..446c3beeff976 100644
--- a/llvm/test/CodeGen/AArch64/phi.ll
+++ b/llvm/test/CodeGen/AArch64/phi.ll
@@ -131,8 +131,8 @@ define i128 @ti128(i1 %c, ptr %p, i128 %a, i128 %b) {
; CHECK-SD: // %bb.0: // %entry
; CHECK-SD-NEXT: tbz w0, #0, .LBB4_2
; CHECK-SD-NEXT: // %bb.1: // %t
-; CHECK-SD-NEXT: mov x4, x2
; CHECK-SD-NEXT: mov x5, x3
+; CHECK-SD-NEXT: mov x4, x2
; CHECK-SD-NEXT: str wzr, [x1]
; CHECK-SD-NEXT: .LBB4_2: // %e
; CHECK-SD-NEXT: mov x0, x4
@@ -336,9 +336,9 @@ define <3 x i8> @tv3i8(i1 %c, ptr %p, <3 x i8> %a, <3 x i8> %b) {
; CHECK-SD: // %bb.0: // %entry
; CHECK-SD-NEXT: tbz w0, #0, .LBB11_2
; CHECK-SD-NEXT: // %bb.1: // %t
-; CHECK-SD-NEXT: mov w5, w2
-; CHECK-SD-NEXT: mov w6, w3
; CHECK-SD-NEXT: mov w7, w4
+; CHECK-SD-NEXT: mov w6, w3
+; CHECK-SD-NEXT: mov w5, w2
; CHECK-SD-NEXT: str wzr, [x1]
; CHECK-SD-NEXT: .LBB11_2: // %e
; CHECK-SD-NEXT: mov w0, w5
@@ -454,8 +454,8 @@ define <32 x i8> @tv32i8(i1 %c, ptr %p, <32 x i8> %a, <32 x i8> %b) {
; CHECK-SD: // %bb.0: // %entry
; CHECK-SD-NEXT: tbz w0, #0, .LBB15_2
; CHECK-SD-NEXT: // %bb.1: // %t
-; CHECK-SD-NEXT: mov v2.16b, v0.16b
; CHECK-SD-NEXT: mov v3.16b, v1.16b
+; CHECK-SD-NEXT: mov v2.16b, v0.16b
; CHECK-SD-NEXT: str wzr, [x1]
; CHECK-SD-NEXT: .LBB15_2: // %e
; CHECK-SD-NEXT: mov v0.16b, v2.16b
@@ -584,8 +584,8 @@ define <16 x i16> @tv16i16(i1 %c, ptr %p, <16 x i16> %a, <16 x i16> %b) {
; CHECK-SD: // %bb.0: // %entry
; CHECK-SD-NEXT: tbz w0, #0, .LBB20_2
; CHECK-SD-NEXT: // %bb.1: // %t
-; CHECK-SD-NEXT: mov v2.16b, v0.16b
; CHECK-SD-NEXT: mov v3.16b, v1.16b
+; CHECK-SD-NEXT: mov v2.16b, v0.16b
; CHECK-SD-NEXT: str wzr, [x1]
; CHECK-SD-NEXT: .LBB20_2: // %e
; CHECK-SD-NEXT: mov v0.16b, v2.16b
@@ -679,8 +679,8 @@ define <8 x i32> @tv8i32(i1 %c, ptr %p, <8 x i32> %a, <8 x i32> %b) {
; CHECK-SD: // %bb.0: // %entry
; CHECK-SD-NEXT: tbz w0, #0, .LBB24_2
; CHECK-SD-NEXT: // %bb.1: // %t
-; CHECK-SD-NEXT: mov v2.16b, v0.16b
; CHECK-SD-NEXT: mov v3.16b, v1.16b
+; CHECK-SD-NEXT: mov v2.16b, v0.16b
; CHECK-SD-NEXT: str wzr, [x1]
; CHECK-SD-NEXT: .LBB24_2: // %e
; CHECK-SD-NEXT: mov v0.16b, v2.16b
@@ -734,10 +734,10 @@ define <3 x i64> @tv3i64(i1 %c, ptr %p, <3 x i64> %a, <3 x i64> %b) {
; CHECK-SD: // %bb.0: // %entry
; CHECK-SD-NEXT: tbz w0, #0, .LBB26_2
; CHECK-SD-NEXT: // %bb.1: // %t
-; CHECK-SD-NEXT: fmov d3, d0
+; CHECK-SD-NEXT: fmov d5, d2
; CHECK-SD-NEXT: fmov d4, d1
; CHECK-SD-NEXT: str wzr, [x1]
-; CHECK-SD-NEXT: fmov d5, d2
+; CHECK-SD-NEXT: fmov d3, d0
; CHECK-SD-NEXT: .LBB26_2: // %e
; CHECK-SD-NEXT: fmov d0, d3
; CHECK-SD-NEXT: fmov d1, d4
@@ -783,8 +783,8 @@ define <4 x i64> @tv4i64(i1 %c, ptr %p, <4 x i64> %a, <4 x i64> %b) {
; CHECK-SD: // %bb.0: // %entry
; CHECK-SD-NEXT: tbz w0, #0, .LBB27_2
; CHECK-SD-NEXT: // %bb.1: // %t
-; CHECK-SD-NEXT: mov v2.16b, v0.16b
; CHECK-SD-NEXT: mov v3.16b, v1.16b
+; CHECK-SD-NEXT: mov v2.16b, v0.16b
; CHECK-SD-NEXT: str wzr, [x1]
; CHECK-SD-NEXT: .LBB27_2: // %e
; CHECK-SD-NEXT: mov v0.16b, v2.16b
@@ -818,8 +818,8 @@ define <2 x i128> @tv2i128(i1 %c, ptr %p, <2 x i128> %a, <2 x i128> %b) {
; CHECK-SD: // %bb.0: // %entry
; CHECK-SD-NEXT: tbz w0, #0, .LBB28_2
; CHECK-SD-NEXT: // %bb.1: // %t
-; CHECK-SD-NEXT: mov x6, x2
; CHECK-SD-NEXT: mov x7, x3
+; CHECK-SD-NEXT: mov x6, x2
; CHECK-SD-NEXT: str wzr, [x1]
; CHECK-SD-NEXT: b .LBB28_3
; CHECK-SD-NEXT: .LBB28_2:
@@ -883,10 +883,10 @@ define <3 x ptr> @tv3p0(i1 %c, ptr %p, <3 x ptr> %a, <3 x ptr> %b) {
; CHECK-SD: // %bb.0: // %entry
; CHECK-SD-NEXT: tbz w0, #0, .LBB30_2
; CHECK-SD-NEXT: // %bb.1: // %t
-; CHECK-SD-NEXT: fmov d3, d0
+; CHECK-SD-NEXT: fmov d5, d2
; CHECK-SD-NEXT: fmov d4, d1
; CHECK-SD-NEXT: str wzr, [x1]
-; CHECK-SD-NEXT: fmov d5, d2
+; CHECK-SD-NEXT: fmov d3, d0
; CHECK-SD-NEXT: .LBB30_2: // %e
; CHECK-SD-NEXT: fmov d0, d3
; CHECK-SD-NEXT: fmov d1, d4
@@ -932,8 +932,8 @@ define <4 x ptr> @tv4p0(i1 %c, ptr %p, <4 x ptr> %a, <4 x ptr> %b) {
; CHECK-SD: // %bb.0: // %entry
; CHECK-SD-NEXT: tbz w0, #0, .LBB31_2
; CHECK-SD-NEXT: // %bb.1: // %t
-; CHECK-SD-NEXT: mov v2.16b, v0.16b
; CHECK-SD-NEXT: mov v3.16b, v1.16b
+; CHECK-SD-NEXT: mov v2.16b, v0.16b
; CHECK-SD-NEXT: str wzr, [x1]
; CHECK-SD-NEXT: .LBB31_2: // %e
; CHECK-SD-NEXT: mov v0.16b, v2.16b
@@ -1047,8 +1047,8 @@ define <16 x half> @tv16f16(i1 %c, ptr %p, <16 x half> %a, <16 x half> %b) {
; CHECK-SD: // %bb.0: // %entry
; CHECK-SD-NEXT: tbz w0, #0, .LBB36_2
; CHECK-SD-NEXT: // %bb.1: // %t
-; CHECK-SD-NEXT: mov v2.16b, v0.16b
; CHECK-SD-NEXT: mov v3.16b, v1.16b
+; CHECK-SD-NEXT: mov v2.16b, v0.16b
; CHECK-SD-NEXT: str wzr, [x1]
; CHECK-SD-NEXT: .LBB36_2: // %e
; CHECK-SD-NEXT: mov v0.16b, v2.16b
@@ -1142,8 +1142,8 @@ define <8 x float> @tv8f32(i1 %c, ptr %p, <8 x float> %a, <8 x float> %b) {
; CHECK-SD: // %bb.0: // %entry
; CHECK-SD-NEXT: tbz w0, #0, .LBB40_2
; CHECK-SD-NEXT: // %bb.1: // %t
-; CHECK-SD-NEXT: mov v2.16b, v0.16b
; CHECK-SD-NEXT: mov v3.16b, v1.16b
+; CHECK-SD-NEXT: mov v2.16b, v0.16b
; CHECK-SD-NEXT: str wzr, [x1]
; CHECK-SD-NEXT: .LBB40_2: // %e
; CHECK-SD-NEXT: mov v0.16b, v2.16b
@@ -1197,10 +1197,10 @@ define <3 x double> @tv3f64(i1 %c, ptr %p, <3 x double> %a, <3 x double> %b) {
; CHECK-SD: // %bb.0: // %entry
; CHECK-SD-NEXT: tbz w0, #0, .LBB42_2
; CHECK-SD-NEXT: // %bb.1: // %t
-; CHECK-SD-NEXT: fmov d3, d0
+; CHECK-SD-NEXT: fmov d5, d2
; CHECK-SD-NEXT: fmov d4, d1
; CHECK-SD-NEXT: str wzr, [x1]
-; CHECK-SD-NEXT: fmov d5, d2
+; CHECK-SD-NEXT: fmov d3, d0
; CHECK-SD-NEXT: .LBB42_2: // %e
; CHECK-SD-NEXT: fmov d0, d3
; CHECK-SD-NEXT: fmov d1, d4
@@ -1246,8 +1246,8 @@ define <4 x double> @tv4f64(i1 %c, ptr %p, <4 x double> %a, <4 x double> %b) {
; CHECK-SD: // %bb.0: // %entry
; CHECK-SD-NEXT: tbz w0, #0, .LBB43_2
; CHECK-SD-NEXT: // %bb.1: // %t
-; CHECK-SD-NEXT: mov v2.16b, v0.16b
; CHECK-SD-NEXT: mov v3.16b, v1.16b
+; CHECK-SD-NEXT: mov v2.16b, v0.16b
; CHECK-SD-NEXT: str wzr, [x1]
; CHECK-SD-NEXT: .LBB43_2: // %e
; CHECK-SD-NEXT: mov v0.16b, v2.16b
@@ -1281,8 +1281,8 @@ define <2 x fp128> @tv2f128(i1 %c, ptr %p, <2 x fp128> %a, <2 x fp128> %b) {
; CHECK-SD: // %bb.0: // %entry
; CHECK-SD-NEXT: tbz w0, #0, .LBB44_2
; CHECK-SD-NEXT: // %bb.1: // %t
-; CHECK-SD-NEXT: mov v2.16b, v0.16b
; CHECK-SD-NEXT: mov v3.16b, v1.16b
+; CHECK-SD-NEXT: mov v2.16b, v0.16b
; CHECK-SD-NEXT: str wzr, [x1]
; CHECK-SD-NEXT: .LBB44_2: // %e
; CHECK-SD-NEXT: mov v0.16b, v2.16b
@@ -1296,8 +1296,8 @@ define <2 x fp128> @tv2f128(i1 %c, ptr %p, <2 x fp128> %a, <2 x fp128> %b) {
; CHECK-GI-NEXT: mov d4, v1.d[1]
; CHECK-GI-NEXT: mov d5, v0.d[1]
; CHECK-GI-NEXT: str wzr, [x1]
-; CHECK-GI-NEXT: fmov d2, d0
; CHECK-GI-NEXT: fmov d3, d1
+; CHECK-GI-NEXT: fmov d2, d0
; CHECK-GI-NEXT: b .LBB44_3
; CHECK-GI-NEXT: .LBB44_2:
; CHECK-GI-NEXT: mov d4, v3.d[1]
diff --git a/llvm/test/CodeGen/AArch64/pr48188.ll b/llvm/test/CodeGen/AArch64/pr48188.ll
index d01069696572e..634517b099f98 100644
--- a/llvm/test/CodeGen/AArch64/pr48188.ll
+++ b/llvm/test/CodeGen/AArch64/pr48188.ll
@@ -14,17 +14,17 @@ define void @test() nounwind {
; GISEL-NEXT: b .LBB0_1
; GISEL-NEXT: .LBB0_1: // %loop
; GISEL-NEXT: // =>This Inner Loop Header: Depth=1
-; GISEL-NEXT: ldr x8, [sp, #8] // 8-byte Folded Reload
-; GISEL-NEXT: ldr x9, [sp] // 8-byte Folded Reload
-; GISEL-NEXT: str x9, [sp] // 8-byte Folded Spill
-; GISEL-NEXT: str x8, [sp, #8] // 8-byte Folded Spill
+; GISEL-NEXT: ldr x8, [sp] // 8-byte Folded Reload
+; GISEL-NEXT: ldr x9, [sp, #8] // 8-byte Folded Reload
+; GISEL-NEXT: str x9, [sp, #8] // 8-byte Folded Spill
+; GISEL-NEXT: str x8, [sp] // 8-byte Folded Spill
; GISEL-NEXT: b .LBB0_1
;
; SDAG-LABEL: test:
; SDAG: // %bb.0: // %entry
; SDAG-NEXT: sub sp, sp, #16
-; SDAG-NEXT: mov x1, xzr
-; SDAG-NEXT: mov x0, x1
+; SDAG-NEXT: mov x0, xzr
+; SDAG-NEXT: mov x1, x0
; SDAG-NEXT: str x1, [sp] // 8-byte Folded Spill
; SDAG-NEXT: str x0, [sp, #8] // 8-byte Folded Spill
; SDAG-NEXT: b .LBB0_1
diff --git a/llvm/test/CodeGen/AArch64/ragreedy-csr.ll b/llvm/test/CodeGen/AArch64/ragreedy-csr.ll
index 5b501762418ef..921cadc7a7511 100644
--- a/llvm/test/CodeGen/AArch64/ragreedy-csr.ll
+++ b/llvm/test/CodeGen/AArch64/ragreedy-csr.ll
@@ -211,27 +211,27 @@ define fastcc i32 @prune_match(ptr nocapture readonly %a, ptr nocapture readonly
; CHECK-NEXT: cmp w12, #2
; CHECK-NEXT: b.ne LBB0_43
; CHECK-NEXT: ; %bb.35: ; %while.cond130.preheader
-; CHECK-NEXT: ldrb w8, [x9, x11]
-; CHECK-NEXT: cbz w8, LBB0_23
+; CHECK-NEXT: ldrb w12, [x9, x11]
+; CHECK-NEXT: cbz w12, LBB0_23
; CHECK-NEXT: ; %bb.36: ; %land.rhs134.preheader
-; CHECK-NEXT: mov x12, xzr
+; CHECK-NEXT: mov x8, xzr
; CHECK-NEXT: mov w0, #1 ; =0x1
; CHECK-NEXT: b LBB0_38
; CHECK-NEXT: LBB0_37: ; %if.then152
; CHECK-NEXT: ; in Loop: Header=BB0_38 Depth=1
-; CHECK-NEXT: add x8, x9, x12
-; CHECK-NEXT: add x12, x12, #1
-; CHECK-NEXT: add x8, x8, x11
-; CHECK-NEXT: ldrb w8, [x8, #1]
-; CHECK-NEXT: cbz w8, LBB0_43
+; CHECK-NEXT: add x12, x9, x8
+; CHECK-NEXT: add x8, x8, #1
+; CHECK-NEXT: add x12, x12, x11
+; CHECK-NEXT: ldrb w12, [x12, #1]
+; CHECK-NEXT: cbz w12, LBB0_43
; CHECK-NEXT: LBB0_38: ; %land.rhs134
; CHECK-NEXT: ; =>This Inner Loop Header: Depth=1
-; CHECK-NEXT: add x13, x10, x12
+; CHECK-NEXT: add x13, x10, x8
; CHECK-NEXT: ldrb w13, [x13, x11]
; CHECK-NEXT: cbz w13, LBB0_23
; CHECK-NEXT: ; %bb.39: ; %while.body139
; CHECK-NEXT: ; in Loop: Header=BB0_38 Depth=1
-; CHECK-NEXT: cmp w8, w13
+; CHECK-NEXT: cmp w12, w13
; CHECK-NEXT: b.eq LBB0_37
; CHECK-NEXT: ; %bb.40: ; %while.body139
; CHECK-NEXT: ; in Loop: Header=BB0_38 Depth=1
@@ -239,7 +239,7 @@ define fastcc i32 @prune_match(ptr nocapture readonly %a, ptr nocapture readonly
; CHECK-NEXT: b.eq LBB0_37
; CHECK-NEXT: ; %bb.41: ; %while.body139
; CHECK-NEXT: ; in Loop: Header=BB0_38 Depth=1
-; CHECK-NEXT: cmp w8, #94
+; CHECK-NEXT: cmp w12, #94
; CHECK-NEXT: b.eq LBB0_37
; CHECK-NEXT: LBB0_42:
; CHECK-NEXT: mov w0, wzr
diff --git a/llvm/test/CodeGen/AArch64/ragreedy-local-interval-cost.ll b/llvm/test/CodeGen/AArch64/ragreedy-local-interval-cost.ll
index c91de8f3a0a47..8f1c504a7f684 100644
--- a/llvm/test/CodeGen/AArch64/ragreedy-local-interval-cost.ll
+++ b/llvm/test/CodeGen/AArch64/ragreedy-local-interval-cost.ll
@@ -31,14 +31,12 @@ define dso_local void @run_test() local_unnamed_addr uwtable {
; CHECK-NEXT: .cfi_offset b14, -104
; CHECK-NEXT: .cfi_offset b15, -112
; CHECK-NEXT: movi v2.2d, #0000000000000000
-; CHECK-NEXT: // implicit-def: $q1
; CHECK-NEXT: mov x8, xzr
; CHECK-NEXT: adrp x9, B+48
; CHECK-NEXT: add x9, x9, :lo12:B+48
; CHECK-NEXT: adrp x10, A
; CHECK-NEXT: add x10, x10, :lo12:A
; CHECK-NEXT: mov x11, xzr
-; CHECK-NEXT: // kill: killed $q1
; CHECK-NEXT: // implicit-def: $q1
; CHECK-NEXT: mov x12, xzr
; CHECK-NEXT: // implicit-def: $q0
@@ -56,7 +54,7 @@ define dso_local void @run_test() local_unnamed_addr uwtable {
; CHECK-NEXT: // implicit-def: $q22
; CHECK-NEXT: // implicit-def: $q23
; CHECK-NEXT: // implicit-def: $q24
-; CHECK-NEXT: // implicit-def: $q9
+; CHECK-NEXT: // implicit-def: $q25
; CHECK-NEXT: // implicit-def: $q27
; CHECK-NEXT: // implicit-def: $q12
; CHECK-NEXT: // implicit-def: $q28
@@ -66,95 +64,97 @@ define dso_local void @run_test() local_unnamed_addr uwtable {
; CHECK-NEXT: // implicit-def: $q30
; CHECK-NEXT: // implicit-def: $q11
; CHECK-NEXT: // implicit-def: $q31
-; CHECK-NEXT: // implicit-def: $q13
; CHECK-NEXT: // kill: killed $q1
+; CHECK-NEXT: // implicit-def: $q9
+; CHECK-NEXT: // implicit-def: $q13
; CHECK-NEXT: // implicit-def: $q1
; CHECK-NEXT: // kill: killed $q1
; CHECK-NEXT: .LBB0_1: // %for.cond1.preheader
; CHECK-NEXT: // =>This Inner Loop Header: Depth=1
-; CHECK-NEXT: stp q29, q15, [sp] // 32-byte Folded Spill
-; CHECK-NEXT: ldr q15, [x8]
; CHECK-NEXT: ldr x15, [x8]
-; CHECK-NEXT: str q14, [sp, #32] // 16-byte Folded Spill
; CHECK-NEXT: add x20, x10, x11
-; CHECK-NEXT: mov v8.16b, v28.16b
-; CHECK-NEXT: fmov x2, d15
-; CHECK-NEXT: mov x17, v15.d[1]
+; CHECK-NEXT: ldr x6, [x8]
+; CHECK-NEXT: ldr x20, [x20, #128]
+; CHECK-NEXT: stp q25, q29, [sp] // 32-byte Folded Spill
+; CHECK-NEXT: mov v29.16b, v21.16b
+; CHECK-NEXT: stp q15, q14, [sp, #32] // 32-byte Folded Spill
; CHECK-NEXT: ldr q14, [x8]
+; CHECK-NEXT: mov v21.16b, v0.16b
+; CHECK-NEXT: ldr q15, [x8]
+; CHECK-NEXT: mov v8.16b, v28.16b
; CHECK-NEXT: mov v28.16b, v24.16b
-; CHECK-NEXT: mov v24.16b, v20.16b
-; CHECK-NEXT: mov v20.16b, v17.16b
; CHECK-NEXT: fmov x13, d14
; CHECK-NEXT: mov x16, v14.d[1]
-; CHECK-NEXT: mov v17.16b, v5.16b
-; CHECK-NEXT: mul x3, x2, x15
+; CHECK-NEXT: mov v24.16b, v20.16b
+; CHECK-NEXT: fmov x2, d15
+; CHECK-NEXT: mov x17, v15.d[1]
+; CHECK-NEXT: mov v20.16b, v17.16b
; CHECK-NEXT: ldr q14, [x9], #64
+; CHECK-NEXT: mov v17.16b, v5.16b
+; CHECK-NEXT: mul x18, x13, x15
; CHECK-NEXT: ldr q5, [sp, #64] // 16-byte Folded Reload
-; CHECK-NEXT: ldr x6, [x8]
-; CHECK-NEXT: ldr x20, [x20, #128]
-; CHECK-NEXT: mul x1, x17, x15
+; CHECK-NEXT: mov v25.16b, v6.16b
; CHECK-NEXT: mov x14, v14.d[1]
; CHECK-NEXT: fmov x5, d14
-; CHECK-NEXT: mov v29.16b, v21.16b
-; CHECK-NEXT: mov v21.16b, v0.16b
-; CHECK-NEXT: mov v25.16b, v6.16b
-; CHECK-NEXT: mul x18, x13, x15
; CHECK-NEXT: mov v6.16b, v2.16b
+; CHECK-NEXT: mul x3, x2, x15
; CHECK-NEXT: mov v26.16b, v22.16b
-; CHECK-NEXT: fmov d15, x3
; CHECK-NEXT: mov v22.16b, v18.16b
; CHECK-NEXT: mov v18.16b, v7.16b
-; CHECK-NEXT: mul x0, x16, x15
; CHECK-NEXT: mov v7.16b, v3.16b
; CHECK-NEXT: mov v16.16b, v4.16b
+; CHECK-NEXT: mul x0, x16, x15
; CHECK-NEXT: add x11, x11, #8
; CHECK-NEXT: add x12, x12, #1
-; CHECK-NEXT: mov v15.d[1], x1
-; CHECK-NEXT: mul x4, x14, x15
-; CHECK-NEXT: cmp x11, #64
; CHECK-NEXT: fmov d14, x18
+; CHECK-NEXT: cmp x11, #64
+; CHECK-NEXT: mul x1, x17, x15
+; CHECK-NEXT: fmov d15, x3
+; CHECK-NEXT: mul x4, x14, x15
+; CHECK-NEXT: mov v14.d[1], x0
; CHECK-NEXT: mul x15, x5, x15
-; CHECK-NEXT: add v5.2d, v5.2d, v15.2d
+; CHECK-NEXT: mov v15.d[1], x1
; CHECK-NEXT: mul x21, x2, x6
-; CHECK-NEXT: mov v14.d[1], x0
+; CHECK-NEXT: add v5.2d, v5.2d, v14.2d
+; CHECK-NEXT: add v9.2d, v9.2d, v14.2d
; CHECK-NEXT: mul x2, x2, x20
; CHECK-NEXT: fmov d0, x15
-; CHECK-NEXT: str q5, [sp, #64] // 16-byte Folded Spill
-; CHECK-NEXT: ldr q5, [sp, #48] // 16-byte Folded Reload
+; CHECK-NEXT: add v13.2d, v13.2d, v15.2d
; CHECK-NEXT: mul x22, x13, x20
-; CHECK-NEXT: add v5.2d, v5.2d, v14.2d
+; CHECK-NEXT: str q5, [sp, #64] // 16-byte Folded Spill
; CHECK-NEXT: fmov d3, x21
+; CHECK-NEXT: ldp q15, q14, [sp, #32] // 32-byte Folded Reload
; CHECK-NEXT: mul x19, x17, x6
; CHECK-NEXT: mov v0.d[1], x4
+; CHECK-NEXT: mov v5.16b, v13.16b
; CHECK-NEXT: fmov d1, x2
+; CHECK-NEXT: mov v13.16b, v9.16b
+; CHECK-NEXT: ldr q9, [sp, #80] // 16-byte Folded Reload
; CHECK-NEXT: mul x17, x17, x20
-; CHECK-NEXT: str q5, [sp, #48] // 16-byte Folded Spill
-; CHECK-NEXT: add v5.2d, v13.2d, v14.2d
; CHECK-NEXT: fmov d2, x22
-; CHECK-NEXT: ldr q13, [sp, #80] // 16-byte Folded Reload
; CHECK-NEXT: mul x7, x16, x6
-; CHECK-NEXT: ldp q15, q14, [sp, #16] // 32-byte Folded Reload
+; CHECK-NEXT: add v9.2d, v9.2d, v0.2d
; CHECK-NEXT: mov v3.d[1], x19
-; CHECK-NEXT: add v13.2d, v13.2d, v0.2d
; CHECK-NEXT: mul x16, x16, x20
; CHECK-NEXT: mov v1.d[1], x17
+; CHECK-NEXT: str q9, [sp, #80] // 16-byte Folded Spill
+; CHECK-NEXT: mov v9.16b, v13.16b
; CHECK-NEXT: mul x23, x5, x20
-; CHECK-NEXT: str q13, [sp, #80] // 16-byte Folded Spill
; CHECK-NEXT: mov v13.16b, v5.16b
; CHECK-NEXT: mov v5.16b, v17.16b
; CHECK-NEXT: mov v17.16b, v20.16b
; CHECK-NEXT: mov v20.16b, v24.16b
-; CHECK-NEXT: mul x13, x13, x6
; CHECK-NEXT: mov v24.16b, v28.16b
+; CHECK-NEXT: mul x13, x13, x6
; CHECK-NEXT: add v11.2d, v11.2d, v3.2d
+; CHECK-NEXT: add v27.2d, v27.2d, v3.2d
; CHECK-NEXT: mov v2.d[1], x16
; CHECK-NEXT: add v15.2d, v15.2d, v1.2d
-; CHECK-NEXT: add v27.2d, v27.2d, v3.2d
-; CHECK-NEXT: mul x18, x14, x20
; CHECK-NEXT: add v23.2d, v23.2d, v3.2d
+; CHECK-NEXT: mul x18, x14, x20
; CHECK-NEXT: add v19.2d, v19.2d, v3.2d
-; CHECK-NEXT: fmov d4, x23
; CHECK-NEXT: add v10.2d, v10.2d, v3.2d
+; CHECK-NEXT: fmov d4, x23
; CHECK-NEXT: mul x15, x5, x6
; CHECK-NEXT: fmov d0, x13
; CHECK-NEXT: add v14.2d, v14.2d, v2.2d
@@ -164,6 +164,7 @@ define dso_local void @run_test() local_unnamed_addr uwtable {
; CHECK-NEXT: mov v7.16b, v18.16b
; CHECK-NEXT: mov v4.d[1], x18
; CHECK-NEXT: mov v18.16b, v22.16b
+; CHECK-NEXT: mov v6.16b, v25.16b
; CHECK-NEXT: mov v0.d[1], x7
; CHECK-NEXT: fmov d1, x15
; CHECK-NEXT: add v28.2d, v8.2d, v4.2d
@@ -181,38 +182,36 @@ define dso_local void @run_test() local_unnamed_addr uwtable {
; CHECK-NEXT: add v3.2d, v3.2d, v0.2d
; CHECK-NEXT: mov v0.16b, v21.16b
; CHECK-NEXT: mov v21.16b, v29.16b
-; CHECK-NEXT: ldr q29, [sp] // 16-byte Folded Reload
-; CHECK-NEXT: add v9.2d, v9.2d, v1.2d
-; CHECK-NEXT: add v6.2d, v25.2d, v1.2d
+; CHECK-NEXT: add v6.2d, v6.2d, v1.2d
+; CHECK-NEXT: ldp q25, q29, [sp] // 32-byte Folded Reload
; CHECK-NEXT: add v5.2d, v5.2d, v1.2d
-; CHECK-NEXT: add v29.2d, v29.2d, v1.2d
; CHECK-NEXT: add v21.2d, v21.2d, v1.2d
; CHECK-NEXT: add v0.2d, v0.2d, v1.2d
+; CHECK-NEXT: add v29.2d, v29.2d, v1.2d
+; CHECK-NEXT: add v25.2d, v25.2d, v1.2d
; CHECK-NEXT: b.ne .LBB0_1
; CHECK-NEXT: // %bb.2: // %for.cond.cleanup
-; CHECK-NEXT: ldr q1, [sp, #48] // 16-byte Folded Reload
+; CHECK-NEXT: ldr q1, [sp, #64] // 16-byte Folded Reload
; CHECK-NEXT: adrp x8, C
; CHECK-NEXT: add x8, x8, :lo12:C
-; CHECK-NEXT: stp q11, q30, [x8, #80]
+; CHECK-NEXT: stp q31, q11, [x8, #64]
; CHECK-NEXT: ldp x20, x19, [sp, #192] // 16-byte Folded Reload
-; CHECK-NEXT: str q1, [x8]
-; CHECK-NEXT: ldr q1, [sp, #64] // 16-byte Folded Reload
+; CHECK-NEXT: stp q1, q13, [x8]
+; CHECK-NEXT: ldr q1, [sp, #80] // 16-byte Folded Reload
; CHECK-NEXT: ldr x23, [sp, #160] // 8-byte Folded Reload
-; CHECK-NEXT: stp q15, q14, [x8, #144]
+; CHECK-NEXT: stp q30, q29, [x8, #96]
; CHECK-NEXT: ldp x22, x21, [sp, #176] // 16-byte Folded Reload
-; CHECK-NEXT: stp q1, q13, [x8, #16]
-; CHECK-NEXT: ldr q1, [sp, #80] // 16-byte Folded Reload
+; CHECK-NEXT: stp q9, q1, [x8, #32]
+; CHECK-NEXT: ldp d9, d8, [sp, #144] // 16-byte Folded Reload
+; CHECK-NEXT: stp q15, q14, [x8, #144]
+; CHECK-NEXT: ldp d15, d14, [sp, #96] // 16-byte Folded Reload
; CHECK-NEXT: stp q28, q12, [x8, #176]
; CHECK-NEXT: ldp d13, d12, [sp, #112] // 16-byte Folded Reload
-; CHECK-NEXT: stp q1, q31, [x8, #48]
-; CHECK-NEXT: ldp d15, d14, [sp, #96] // 16-byte Folded Reload
-; CHECK-NEXT: stp q9, q24, [x8, #240]
-; CHECK-NEXT: ldp d9, d8, [sp, #144] // 16-byte Folded Reload
; CHECK-NEXT: stp q19, q18, [x8, #336]
; CHECK-NEXT: stp q10, q7, [x8, #400]
; CHECK-NEXT: ldp d11, d10, [sp, #128] // 16-byte Folded Reload
-; CHECK-NEXT: str q29, [x8, #112]
; CHECK-NEXT: str q27, [x8, #208]
+; CHECK-NEXT: stp q25, q24, [x8, #240]
; CHECK-NEXT: stp q23, q22, [x8, #272]
; CHECK-NEXT: stp q21, q20, [x8, #304]
; CHECK-NEXT: stp q6, q17, [x8, #368]
diff --git a/llvm/test/CodeGen/AArch64/reduce-or-opt.ll b/llvm/test/CodeGen/AArch64/reduce-or-opt.ll
index f5df5ea53c990..b3dfab8f69b59 100644
--- a/llvm/test/CodeGen/AArch64/reduce-or-opt.ll
+++ b/llvm/test/CodeGen/AArch64/reduce-or-opt.ll
@@ -95,18 +95,18 @@ define i64 @select_or_reduce_nxv2i1(ptr nocapture noundef readonly %src) {
; CHECK: // %bb.0: // %entry
; CHECK-NEXT: cntd x8
; CHECK-NEXT: ptrue p0.d
-; CHECK-NEXT: mov x9, xzr
-; CHECK-NEXT: neg x10, x8
-; CHECK-NEXT: add x10, x10, #4
+; CHECK-NEXT: mov x10, xzr
+; CHECK-NEXT: neg x9, x8
+; CHECK-NEXT: add x9, x9, #4
; CHECK-NEXT: .LBB2_1: // %vector.body
; CHECK-NEXT: // =>This Inner Loop Header: Depth=1
-; CHECK-NEXT: ld1d { z0.d }, p0/z, [x0, x9, lsl #3]
+; CHECK-NEXT: ld1d { z0.d }, p0/z, [x0, x10, lsl #3]
; CHECK-NEXT: cmpeq p1.d, p0/z, z0.d, #0
; CHECK-NEXT: b.ne .LBB2_3
; CHECK-NEXT: // %bb.2: // %vector.body
; CHECK-NEXT: // in Loop: Header=BB2_1 Depth=1
-; CHECK-NEXT: cmp x10, x9
-; CHECK-NEXT: add x9, x9, x8
+; CHECK-NEXT: cmp x9, x10
+; CHECK-NEXT: add x10, x10, x8
; CHECK-NEXT: b.ne .LBB2_1
; CHECK-NEXT: .LBB2_3: // %middle.split
; CHECK-NEXT: ptest p0, p1.b
@@ -138,18 +138,18 @@ define i64 @br_or_reduce_nxv2i1(ptr nocapture noundef readonly %src, ptr noundef
; CHECK: // %bb.0: // %entry
; CHECK-NEXT: cntd x8
; CHECK-NEXT: ptrue p0.d
-; CHECK-NEXT: mov x9, xzr
-; CHECK-NEXT: neg x10, x8
-; CHECK-NEXT: add x10, x10, #4
+; CHECK-NEXT: mov x10, xzr
+; CHECK-NEXT: neg x9, x8
+; CHECK-NEXT: add x9, x9, #4
; CHECK-NEXT: .LBB3_1: // %vector.body
; CHECK-NEXT: // =>This Inner Loop Header: Depth=1
-; CHECK-NEXT: ld1d { z0.d }, p0/z, [x0, x9, lsl #3]
+; CHECK-NEXT: ld1d { z0.d }, p0/z, [x0, x10, lsl #3]
; CHECK-NEXT: cmpeq p1.d, p0/z, z0.d, #0
; CHECK-NEXT: b.ne .LBB3_3
; CHECK-NEXT: // %bb.2: // %vector.body
; CHECK-NEXT: // in Loop: Header=BB3_1 Depth=1
-; CHECK-NEXT: cmp x10, x9
-; CHECK-NEXT: add x9, x9, x8
+; CHECK-NEXT: cmp x9, x10
+; CHECK-NEXT: add x10, x10, x8
; CHECK-NEXT: b.ne .LBB3_1
; CHECK-NEXT: .LBB3_3: // %middle.split
; CHECK-NEXT: ptest p0, p1.b
diff --git a/llvm/test/CodeGen/AArch64/sink-and-fold.ll b/llvm/test/CodeGen/AArch64/sink-and-fold.ll
index 4d383fefc43c7..163124c0d2757 100644
--- a/llvm/test/CodeGen/AArch64/sink-and-fold.ll
+++ b/llvm/test/CodeGen/AArch64/sink-and-fold.ll
@@ -151,7 +151,7 @@ define void @f4(ptr %a, i64 %n) nounwind "target-features"="+alu-lsl-fast" {
; CHECK-NEXT: .LBB4_3: // %LI
; CHECK-NEXT: // =>This Loop Header: Depth=1
; CHECK-NEXT: // Child Loop BB4_6 Depth 2
-; CHECK-NEXT: mov x21, xzr
+; CHECK-NEXT: mov x8, xzr
; CHECK-NEXT: add x23, x22, #1
; CHECK-NEXT: b .LBB4_6
; CHECK-NEXT: .LBB4_4: // %if.else
@@ -162,13 +162,13 @@ define void @f4(ptr %a, i64 %n) nounwind "target-features"="+alu-lsl-fast" {
; CHECK-NEXT: add x8, x21, #1
; CHECK-NEXT: str w0, [x20, x21, lsl #2]
; CHECK-NEXT: sub x9, x8, #1
-; CHECK-NEXT: mov x21, x8
; CHECK-NEXT: cmp x9, x19
; CHECK-NEXT: b.ge .LBB4_2
; CHECK-NEXT: .LBB4_6: // %LJ
; CHECK-NEXT: // Parent Loop BB4_3 Depth=1
; CHECK-NEXT: // => This Inner Loop Header: Depth=2
-; CHECK-NEXT: ldr w8, [x20, x21, lsl #2]
+; CHECK-NEXT: mov x21, x8
+; CHECK-NEXT: ldr w8, [x20, x8, lsl #2]
; CHECK-NEXT: tbz w8, #31, .LBB4_4
; CHECK-NEXT: // %bb.7: // %if.then
; CHECK-NEXT: // in Loop: Header=BB4_6 Depth=2
diff --git a/llvm/test/CodeGen/AArch64/sve-lsrchain.ll b/llvm/test/CodeGen/AArch64/sve-lsrchain.ll
index d94fa6433bb7f..2fe3001ec0f44 100644
--- a/llvm/test/CodeGen/AArch64/sve-lsrchain.ll
+++ b/llvm/test/CodeGen/AArch64/sve-lsrchain.ll
@@ -14,22 +14,22 @@ define void @test(ptr nocapture noundef readonly %kernel, i32 noundef %kw, float
; CHECK-NEXT: // %bb.2: // %for.body.us.preheader
; CHECK-NEXT: ptrue p0.h
; CHECK-NEXT: add x11, x2, x11, lsl #1
-; CHECK-NEXT: mov w8, wzr
+; CHECK-NEXT: mov x8, xzr
; CHECK-NEXT: ptrue p1.b
-; CHECK-NEXT: mov x9, xzr
+; CHECK-NEXT: mov w9, wzr
; CHECK-NEXT: mov w10, wzr
; CHECK-NEXT: mov x12, #4 // =0x4
; CHECK-NEXT: mov x13, #8 // =0x8
; CHECK-NEXT: .LBB0_3: // %for.body.us
; CHECK-NEXT: // =>This Loop Header: Depth=1
; CHECK-NEXT: // Child Loop BB0_4 Depth 2
-; CHECK-NEXT: add x14, x0, x9, lsl #2
-; CHECK-NEXT: sbfiz x15, x8, #1, #32
+; CHECK-NEXT: add x14, x0, x8, lsl #2
+; CHECK-NEXT: sbfiz x15, x9, #1, #32
; CHECK-NEXT: mov x16, x2
; CHECK-NEXT: ldp s0, s1, [x14]
; CHECK-NEXT: add x15, x15, #8
; CHECK-NEXT: ldp s2, s3, [x14, #8]
-; CHECK-NEXT: ubfiz x14, x8, #1, #32
+; CHECK-NEXT: ubfiz x14, x9, #1, #32
; CHECK-NEXT: fcvt h0, s0
; CHECK-NEXT: fcvt h1, s1
; CHECK-NEXT: fcvt h2, s2
@@ -91,8 +91,8 @@ define void @test(ptr nocapture noundef readonly %kernel, i32 noundef %kw, float
; CHECK-NEXT: // %bb.5: // %while.cond.i..exit_crit_edge.us
; CHECK-NEXT: // in Loop: Header=BB0_3 Depth=1
; CHECK-NEXT: add w10, w10, #1
-; CHECK-NEXT: add x9, x9, #4
-; CHECK-NEXT: add w8, w8, #16
+; CHECK-NEXT: add x8, x8, #4
+; CHECK-NEXT: add w9, w9, #16
; CHECK-NEXT: cmp w10, w1
; CHECK-NEXT: b.ne .LBB0_3
; CHECK-NEXT: .LBB0_6: // %exit78
diff --git a/llvm/test/CodeGen/AArch64/sve-ptest-removal-sink.ll b/llvm/test/CodeGen/AArch64/sve-ptest-removal-sink.ll
index 124f81e7864d1..39fe92aae0619 100644
--- a/llvm/test/CodeGen/AArch64/sve-ptest-removal-sink.ll
+++ b/llvm/test/CodeGen/AArch64/sve-ptest-removal-sink.ll
@@ -11,12 +11,12 @@ define void @test_sink_ptrue_into_ptest(i32 %n) {
; CHECK-NEXT: whilelt p0.s, wzr, w0
; CHECK-NEXT: b.pl .LBB0_3
; CHECK-NEXT: // %bb.1: // %for.body.preheader
-; CHECK-NEXT: mov w8, wzr
-; CHECK-NEXT: cntw x9
+; CHECK-NEXT: mov w9, wzr
+; CHECK-NEXT: cntw x8
; CHECK-NEXT: .LBB0_2: // %for.body
; CHECK-NEXT: // =>This Inner Loop Header: Depth=1
-; CHECK-NEXT: whilelt p0.s, w8, w0
-; CHECK-NEXT: add w8, w8, w9
+; CHECK-NEXT: whilelt p0.s, w9, w0
+; CHECK-NEXT: add w9, w9, w8
; CHECK-NEXT: b.mi .LBB0_2
; CHECK-NEXT: .LBB0_3: // %exit
; CHECK-NEXT: ret
diff --git a/llvm/test/CodeGen/AArch64/swifterror.ll b/llvm/test/CodeGen/AArch64/swifterror.ll
index 07ee87e880aff..9949a48034815 100644
--- a/llvm/test/CodeGen/AArch64/swifterror.ll
+++ b/llvm/test/CodeGen/AArch64/swifterror.ll
@@ -564,10 +564,11 @@ define float @foo_loop(ptr swifterror %error_ptr_ref, i32 %cc, float %cc2) {
; CHECK-O0-AARCH64-NEXT: mov w8, #16 ; =0x10
; CHECK-O0-AARCH64-NEXT: mov w0, w8
; CHECK-O0-AARCH64-NEXT: bl _malloc
-; CHECK-O0-AARCH64-NEXT: mov x9, x0
+; CHECK-O0-AARCH64-NEXT: mov x1, x0
+; CHECK-O0-AARCH64-NEXT: mov x0, x1
+; CHECK-O0-AARCH64-NEXT: str x1, [sp, #8] ; 8-byte Folded Spill
; CHECK-O0-AARCH64-NEXT: mov w8, #1 ; =0x1
-; CHECK-O0-AARCH64-NEXT: strb w8, [x9, #8]
-; CHECK-O0-AARCH64-NEXT: str x0, [sp, #8] ; 8-byte Folded Spill
+; CHECK-O0-AARCH64-NEXT: strb w8, [x0, #8]
; CHECK-O0-AARCH64-NEXT: LBB4_3: ; %bb_cont
; CHECK-O0-AARCH64-NEXT: ; in Loop: Header=BB4_1 Depth=1
; CHECK-O0-AARCH64-NEXT: ldr s0, [sp, #16] ; 4-byte Folded Reload
@@ -605,12 +606,11 @@ define float @foo_loop(ptr swifterror %error_ptr_ref, i32 %cc, float %cc2) {
; CHECK-O0-ARM64_32-NEXT: mov w8, #16 ; =0x10
; CHECK-O0-ARM64_32-NEXT: mov w0, w8
; CHECK-O0-ARM64_32-NEXT: bl _malloc
-; CHECK-O0-ARM64_32-NEXT: mov x9, x0
-; CHECK-O0-ARM64_32-NEXT: ; kill: def $x0 killed $x9
-; CHECK-O0-ARM64_32-NEXT: mov x0, x9
+; CHECK-O0-ARM64_32-NEXT: ; kill: def $x1 killed $x0
+; CHECK-O0-ARM64_32-NEXT: mov x1, x0
+; CHECK-O0-ARM64_32-NEXT: str x1, [sp, #8] ; 8-byte Folded Spill
; CHECK-O0-ARM64_32-NEXT: mov w8, #1 ; =0x1
-; CHECK-O0-ARM64_32-NEXT: strb w8, [x9, #8]
-; CHECK-O0-ARM64_32-NEXT: str x0, [sp, #8] ; 8-byte Folded Spill
+; CHECK-O0-ARM64_32-NEXT: strb w8, [x0, #8]
; CHECK-O0-ARM64_32-NEXT: LBB4_3: ; %bb_cont
; CHECK-O0-ARM64_32-NEXT: ; in Loop: Header=BB4_1 Depth=1
; CHECK-O0-ARM64_32-NEXT: ldr s0, [sp, #16] ; 4-byte Folded Reload
diff --git a/llvm/test/CodeGen/AArch64/tbl-loops.ll b/llvm/test/CodeGen/AArch64/tbl-loops.ll
index b5d64112db727..e73cbf1ee5df0 100644
--- a/llvm/test/CodeGen/AArch64/tbl-loops.ll
+++ b/llvm/test/CodeGen/AArch64/tbl-loops.ll
@@ -17,18 +17,18 @@ define void @loop1(ptr noalias nocapture noundef writeonly %dst, ptr nocapture n
; CHECK-NEXT: .LBB0_3: // %vector.ph
; CHECK-NEXT: add x11, x8, #1
; CHECK-NEXT: mov w8, #1132396544 // =0x437f0000
-; CHECK-NEXT: add x12, x0, #4
+; CHECK-NEXT: add x13, x0, #4
; CHECK-NEXT: and x10, x11, #0x1fffffff8
; CHECK-NEXT: dup v0.4s, w8
-; CHECK-NEXT: add x13, x1, #16
+; CHECK-NEXT: add x14, x1, #16
; CHECK-NEXT: add x8, x1, x10, lsl #2
+; CHECK-NEXT: mov x12, x10
; CHECK-NEXT: add x9, x0, x10
-; CHECK-NEXT: mov x14, x10
; CHECK-NEXT: .LBB0_4: // %vector.body
; CHECK-NEXT: // =>This Inner Loop Header: Depth=1
-; CHECK-NEXT: ldp q1, q2, [x13, #-16]
-; CHECK-NEXT: subs x14, x14, #8
-; CHECK-NEXT: add x13, x13, #32
+; CHECK-NEXT: ldp q1, q2, [x14, #-16]
+; CHECK-NEXT: subs x12, x12, #8
+; CHECK-NEXT: add x14, x14, #32
; CHECK-NEXT: fcmgt v3.4s, v1.4s, v0.4s
; CHECK-NEXT: fcmgt v4.4s, v2.4s, v0.4s
; CHECK-NEXT: fcmlt v5.4s, v1.4s, #0.0
@@ -44,8 +44,8 @@ define void @loop1(ptr noalias nocapture noundef writeonly %dst, ptr nocapture n
; CHECK-NEXT: uzp1 v1.8b, v1.8b, v0.8b
; CHECK-NEXT: uzp1 v2.8b, v2.8b, v0.8b
; CHECK-NEXT: mov v1.s[1], v2.s[0]
-; CHECK-NEXT: stur d1, [x12, #-4]
-; CHECK-NEXT: add x12, x12, #8
+; CHECK-NEXT: stur d1, [x13, #-4]
+; CHECK-NEXT: add x13, x13, #8
; CHECK-NEXT: b.ne .LBB0_4
; CHECK-NEXT: // %bb.5: // %middle.block
; CHECK-NEXT: cmp x11, x10
diff --git a/llvm/test/CodeGen/AMDGPU/GlobalISel/atomicrmw_fmax.ll b/llvm/test/CodeGen/AMDGPU/GlobalISel/atomicrmw_fmax.ll
index 666523c88860c..11afcebfada2e 100644
--- a/llvm/test/CodeGen/AMDGPU/GlobalISel/atomicrmw_fmax.ll
+++ b/llvm/test/CodeGen/AMDGPU/GlobalISel/atomicrmw_fmax.ll
@@ -1670,23 +1670,23 @@ define void @buffer_fat_ptr_agent_atomic_fmax_noret_f32__amdgpu_no_fine_grained_
; GFX942-LABEL: buffer_fat_ptr_agent_atomic_fmax_noret_f32__amdgpu_no_fine_grained_memory:
; GFX942: ; %bb.0:
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
-; GFX942-NEXT: v_mov_b32_e32 v2, s16
-; GFX942-NEXT: buffer_load_dword v1, v2, s[0:3], 0 offen
+; GFX942-NEXT: v_mov_b32_e32 v1, s16
+; GFX942-NEXT: buffer_load_dword v2, v1, s[0:3], 0 offen
; GFX942-NEXT: s_mov_b64 s[4:5], 0
-; GFX942-NEXT: v_max_f32_e32 v3, v0, v0
+; GFX942-NEXT: v_max_f32_e32 v0, v0, v0
; GFX942-NEXT: .LBB13_1: ; %atomicrmw.start
; GFX942-NEXT: ; =>This Inner Loop Header: Depth=1
; GFX942-NEXT: s_waitcnt vmcnt(0)
-; GFX942-NEXT: v_max_f32_e32 v0, v1, v1
-; GFX942-NEXT: v_max_f32_e32 v0, v0, v3
-; GFX942-NEXT: v_mov_b64_e32 v[4:5], v[0:1]
+; GFX942-NEXT: v_mov_b32_e32 v5, v2
+; GFX942-NEXT: v_max_f32_e32 v2, v5, v5
+; GFX942-NEXT: v_max_f32_e32 v4, v2, v0
+; GFX942-NEXT: v_mov_b64_e32 v[2:3], v[4:5]
; GFX942-NEXT: buffer_wbl2 sc1
-; GFX942-NEXT: buffer_atomic_cmpswap v[4:5], v2, s[0:3], 0 offen sc0
+; GFX942-NEXT: buffer_atomic_cmpswap v[2:3], v1, s[0:3], 0 offen sc0
; GFX942-NEXT: s_waitcnt vmcnt(0)
; GFX942-NEXT: buffer_inv sc1
-; GFX942-NEXT: v_cmp_eq_u32_e32 vcc, v4, v1
+; GFX942-NEXT: v_cmp_eq_u32_e32 vcc, v2, v5
; GFX942-NEXT: s_or_b64 s[4:5], vcc, s[4:5]
-; GFX942-NEXT: v_mov_b32_e32 v1, v4
; GFX942-NEXT: s_andn2_b64 exec, exec, s[4:5]
; GFX942-NEXT: s_cbranch_execnz .LBB13_1
; GFX942-NEXT: ; %bb.2: ; %atomicrmw.end
@@ -1718,22 +1718,22 @@ define void @buffer_fat_ptr_agent_atomic_fmax_noret_f32__amdgpu_no_fine_grained_
; GFX90A-LABEL: buffer_fat_ptr_agent_atomic_fmax_noret_f32__amdgpu_no_fine_grained_memory:
; GFX90A: ; %bb.0:
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
-; GFX90A-NEXT: v_mov_b32_e32 v2, s20
-; GFX90A-NEXT: buffer_load_dword v1, v2, s[16:19], 0 offen
+; GFX90A-NEXT: v_mov_b32_e32 v1, s20
+; GFX90A-NEXT: buffer_load_dword v2, v1, s[16:19], 0 offen
; GFX90A-NEXT: s_mov_b64 s[4:5], 0
-; GFX90A-NEXT: v_max_f32_e32 v3, v0, v0
+; GFX90A-NEXT: v_max_f32_e32 v0, v0, v0
; GFX90A-NEXT: .LBB13_1: ; %atomicrmw.start
; GFX90A-NEXT: ; =>This Inner Loop Header: Depth=1
; GFX90A-NEXT: s_waitcnt vmcnt(0)
-; GFX90A-NEXT: v_max_f32_e32 v0, v1, v1
-; GFX90A-NEXT: v_max_f32_e32 v0, v0, v3
-; GFX90A-NEXT: v_pk_mov_b32 v[4:5], v[0:1], v[0:1] op_sel:[0,1]
-; GFX90A-NEXT: buffer_atomic_cmpswap v[4:5], v2, s[16:19], 0 offen glc
+; GFX90A-NEXT: v_mov_b32_e32 v5, v2
+; GFX90A-NEXT: v_max_f32_e32 v2, v5, v5
+; GFX90A-NEXT: v_max_f32_e32 v4, v2, v0
+; GFX90A-NEXT: v_pk_mov_b32 v[2:3], v[4:5], v[4:5] op_sel:[0,1]
+; GFX90A-NEXT: buffer_atomic_cmpswap v[2:3], v1, s[16:19], 0 offen glc
; GFX90A-NEXT: s_waitcnt vmcnt(0)
; GFX90A-NEXT: buffer_wbinvl1
-; GFX90A-NEXT: v_cmp_eq_u32_e32 vcc, v4, v1
+; GFX90A-NEXT: v_cmp_eq_u32_e32 vcc, v2, v5
; GFX90A-NEXT: s_or_b64 s[4:5], vcc, s[4:5]
-; GFX90A-NEXT: v_mov_b32_e32 v1, v4
; GFX90A-NEXT: s_andn2_b64 exec, exec, s[4:5]
; GFX90A-NEXT: s_cbranch_execnz .LBB13_1
; GFX90A-NEXT: ; %bb.2: ; %atomicrmw.end
@@ -1743,23 +1743,23 @@ define void @buffer_fat_ptr_agent_atomic_fmax_noret_f32__amdgpu_no_fine_grained_
; GFX908-LABEL: buffer_fat_ptr_agent_atomic_fmax_noret_f32__amdgpu_no_fine_grained_memory:
; GFX908: ; %bb.0:
; GFX908-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
-; GFX908-NEXT: v_mov_b32_e32 v2, s20
-; GFX908-NEXT: buffer_load_dword v1, v2, s[16:19], 0 offen
+; GFX908-NEXT: v_mov_b32_e32 v3, s20
+; GFX908-NEXT: buffer_load_dword v1, v3, s[16:19], 0 offen
; GFX908-NEXT: s_mov_b64 s[4:5], 0
-; GFX908-NEXT: v_max_f32_e32 v3, v0, v0
+; GFX908-NEXT: v_max_f32_e32 v0, v0, v0
; GFX908-NEXT: .LBB13_1: ; %atomicrmw.start
; GFX908-NEXT: ; =>This Inner Loop Header: Depth=1
; GFX908-NEXT: s_waitcnt vmcnt(0)
-; GFX908-NEXT: v_max_f32_e32 v0, v1, v1
-; GFX908-NEXT: v_max_f32_e32 v0, v0, v3
; GFX908-NEXT: v_mov_b32_e32 v5, v1
-; GFX908-NEXT: v_mov_b32_e32 v4, v0
-; GFX908-NEXT: buffer_atomic_cmpswap v[4:5], v2, s[16:19], 0 offen glc
+; GFX908-NEXT: v_max_f32_e32 v1, v5, v5
+; GFX908-NEXT: v_max_f32_e32 v4, v1, v0
+; GFX908-NEXT: v_mov_b32_e32 v1, v4
+; GFX908-NEXT: v_mov_b32_e32 v2, v5
+; GFX908-NEXT: buffer_atomic_cmpswap v[1:2], v3, s[16:19], 0 offen glc
; GFX908-NEXT: s_waitcnt vmcnt(0)
; GFX908-NEXT: buffer_wbinvl1
-; GFX908-NEXT: v_cmp_eq_u32_e32 vcc, v4, v1
+; GFX908-NEXT: v_cmp_eq_u32_e32 vcc, v1, v5
; GFX908-NEXT: s_or_b64 s[4:5], vcc, s[4:5]
-; GFX908-NEXT: v_mov_b32_e32 v1, v4
; GFX908-NEXT: s_andn2_b64 exec, exec, s[4:5]
; GFX908-NEXT: s_cbranch_execnz .LBB13_1
; GFX908-NEXT: ; %bb.2: ; %atomicrmw.end
@@ -1769,23 +1769,23 @@ define void @buffer_fat_ptr_agent_atomic_fmax_noret_f32__amdgpu_no_fine_grained_
; GFX8-LABEL: buffer_fat_ptr_agent_atomic_fmax_noret_f32__amdgpu_no_fine_grained_memory:
; GFX8: ; %bb.0:
; GFX8-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
-; GFX8-NEXT: v_mov_b32_e32 v2, s20
-; GFX8-NEXT: buffer_load_dword v1, v2, s[16:19], 0 offen
+; GFX8-NEXT: v_mov_b32_e32 v3, s20
+; GFX8-NEXT: buffer_load_dword v1, v3, s[16:19], 0 offen
; GFX8-NEXT: s_mov_b64 s[4:5], 0
-; GFX8-NEXT: v_mul_f32_e32 v3, 1.0, v0
+; GFX8-NEXT: v_mul_f32_e32 v0, 1.0, v0
; GFX8-NEXT: .LBB13_1: ; %atomicrmw.start
; GFX8-NEXT: ; =>This Inner Loop Header: Depth=1
; GFX8-NEXT: s_waitcnt vmcnt(0)
-; GFX8-NEXT: v_mul_f32_e32 v0, 1.0, v1
-; GFX8-NEXT: v_max_f32_e32 v0, v0, v3
; GFX8-NEXT: v_mov_b32_e32 v5, v1
-; GFX8-NEXT: v_mov_b32_e32 v4, v0
-; GFX8-NEXT: buffer_atomic_cmpswap v[4:5], v2, s[16:19], 0 offen glc
+; GFX8-NEXT: v_mul_f32_e32 v1, 1.0, v5
+; GFX8-NEXT: v_max_f32_e32 v4, v1, v0
+; GFX8-NEXT: v_mov_b32_e32 v1, v4
+; GFX8-NEXT: v_mov_b32_e32 v2, v5
+; GFX8-NEXT: buffer_atomic_cmpswap v[1:2], v3, s[16:19], 0 offen glc
; GFX8-NEXT: s_waitcnt vmcnt(0)
; GFX8-NEXT: buffer_wbinvl1
-; GFX8-NEXT: v_cmp_eq_u32_e32 vcc, v4, v1
+; GFX8-NEXT: v_cmp_eq_u32_e32 vcc, v1, v5
; GFX8-NEXT: s_or_b64 s[4:5], vcc, s[4:5]
-; GFX8-NEXT: v_mov_b32_e32 v1, v4
; GFX8-NEXT: s_andn2_b64 exec, exec, s[4:5]
; GFX8-NEXT: s_cbranch_execnz .LBB13_1
; GFX8-NEXT: ; %bb.2: ; %atomicrmw.end
@@ -1986,23 +1986,24 @@ define void @buffer_fat_ptr_agent_atomic_fmax_noret_f64__amdgpu_no_fine_grained_
; GFX12-NEXT: s_wait_bvhcnt 0x0
; GFX12-NEXT: s_wait_kmcnt 0x0
; GFX12-NEXT: v_mov_b32_e32 v6, s16
-; GFX12-NEXT: v_max_num_f64_e32 v[4:5], v[0:1], v[0:1]
+; GFX12-NEXT: v_max_num_f64_e32 v[0:1], v[0:1], v[0:1]
; GFX12-NEXT: s_mov_b32 s4, 0
; GFX12-NEXT: buffer_load_b64 v[2:3], v6, s[0:3], null offen
; GFX12-NEXT: .LBB15_1: ; %atomicrmw.start
; GFX12-NEXT: ; =>This Inner Loop Header: Depth=1
; GFX12-NEXT: s_wait_loadcnt 0x0
-; GFX12-NEXT: v_max_num_f64_e32 v[0:1], v[2:3], v[2:3]
-; GFX12-NEXT: s_wait_storecnt 0x0
-; GFX12-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(SKIP_1) | instid1(VALU_DEP_2)
-; GFX12-NEXT: v_max_num_f64_e32 v[0:1], v[0:1], v[4:5]
; GFX12-NEXT: v_dual_mov_b32 v10, v3 :: v_dual_mov_b32 v9, v2
-; GFX12-NEXT: v_dual_mov_b32 v8, v1 :: v_dual_mov_b32 v7, v0
-; GFX12-NEXT: buffer_atomic_cmpswap_b64 v[7:10], v6, s[0:3], null offen th:TH_ATOMIC_RETURN
+; GFX12-NEXT: s_wait_storecnt 0x0
+; GFX12-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
+; GFX12-NEXT: v_max_num_f64_e32 v[2:3], v[9:10], v[9:10]
+; GFX12-NEXT: v_max_num_f64_e32 v[7:8], v[2:3], v[0:1]
+; GFX12-NEXT: s_delay_alu instid0(VALU_DEP_1)
+; GFX12-NEXT: v_dual_mov_b32 v2, v7 :: v_dual_mov_b32 v3, v8
+; GFX12-NEXT: v_dual_mov_b32 v4, v9 :: v_dual_mov_b32 v5, v10
+; GFX12-NEXT: buffer_atomic_cmpswap_b64 v[2:5], v6, s[0:3], null offen th:TH_ATOMIC_RETURN
; GFX12-NEXT: s_wait_loadcnt 0x0
; GFX12-NEXT: global_inv scope:SCOPE_DEV
-; GFX12-NEXT: v_cmp_eq_u64_e32 vcc_lo, v[7:8], v[2:3]
-; GFX12-NEXT: v_dual_mov_b32 v2, v7 :: v_dual_mov_b32 v3, v8
+; GFX12-NEXT: v_cmp_eq_u64_e32 vcc_lo, v[2:3], v[9:10]
; GFX12-NEXT: s_or_b32 s4, vcc_lo, s4
; GFX12-NEXT: s_wait_alu 0xfffe
; GFX12-NEXT: s_and_not1_b32 exec_lo, exec_lo, s4
@@ -2026,24 +2027,25 @@ define void @buffer_fat_ptr_agent_atomic_fmax_noret_f64__amdgpu_no_fine_grained_
; GFX11: ; %bb.0:
; GFX11-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX11-NEXT: v_mov_b32_e32 v6, s16
-; GFX11-NEXT: v_max_f64 v[4:5], v[0:1], v[0:1]
+; GFX11-NEXT: v_max_f64 v[0:1], v[0:1], v[0:1]
; GFX11-NEXT: s_mov_b32 s4, 0
; GFX11-NEXT: buffer_load_b64 v[2:3], v6, s[0:3], 0 offen
; GFX11-NEXT: .LBB15_1: ; %atomicrmw.start
; GFX11-NEXT: ; =>This Inner Loop Header: Depth=1
; GFX11-NEXT: s_waitcnt vmcnt(0)
-; GFX11-NEXT: v_max_f64 v[0:1], v[2:3], v[2:3]
-; GFX11-NEXT: s_waitcnt_vscnt null, 0x0
-; GFX11-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(SKIP_1) | instid1(VALU_DEP_2)
-; GFX11-NEXT: v_max_f64 v[0:1], v[0:1], v[4:5]
; GFX11-NEXT: v_dual_mov_b32 v10, v3 :: v_dual_mov_b32 v9, v2
-; GFX11-NEXT: v_dual_mov_b32 v8, v1 :: v_dual_mov_b32 v7, v0
-; GFX11-NEXT: buffer_atomic_cmpswap_b64 v[7:10], v6, s[0:3], 0 offen glc
+; GFX11-NEXT: s_waitcnt_vscnt null, 0x0
+; GFX11-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
+; GFX11-NEXT: v_max_f64 v[2:3], v[9:10], v[9:10]
+; GFX11-NEXT: v_max_f64 v[7:8], v[2:3], v[0:1]
+; GFX11-NEXT: s_delay_alu instid0(VALU_DEP_1)
+; GFX11-NEXT: v_dual_mov_b32 v2, v7 :: v_dual_mov_b32 v3, v8
+; GFX11-NEXT: v_dual_mov_b32 v4, v9 :: v_dual_mov_b32 v5, v10
+; GFX11-NEXT: buffer_atomic_cmpswap_b64 v[2:5], v6, s[0:3], 0 offen glc
; GFX11-NEXT: s_waitcnt vmcnt(0)
; GFX11-NEXT: buffer_gl1_inv
; GFX11-NEXT: buffer_gl0_inv
-; GFX11-NEXT: v_cmp_eq_u64_e32 vcc_lo, v[7:8], v[2:3]
-; GFX11-NEXT: v_dual_mov_b32 v2, v7 :: v_dual_mov_b32 v3, v8
+; GFX11-NEXT: v_cmp_eq_u64_e32 vcc_lo, v[2:3], v[9:10]
; GFX11-NEXT: s_or_b32 s4, vcc_lo, s4
; GFX11-NEXT: s_delay_alu instid0(SALU_CYCLE_1)
; GFX11-NEXT: s_and_not1_b32 exec_lo, exec_lo, s4
@@ -2077,24 +2079,24 @@ define void @buffer_fat_ptr_agent_atomic_fmax_noret_f64__amdgpu_no_fine_grained_
; GFX908-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX908-NEXT: v_mov_b32_e32 v6, s20
; GFX908-NEXT: buffer_load_dwordx2 v[2:3], v6, s[16:19], 0 offen
-; GFX908-NEXT: v_max_f64 v[4:5], v[0:1], v[0:1]
+; GFX908-NEXT: v_max_f64 v[0:1], v[0:1], v[0:1]
; GFX908-NEXT: s_mov_b64 s[4:5], 0
; GFX908-NEXT: .LBB15_1: ; %atomicrmw.start
; GFX908-NEXT: ; =>This Inner Loop Header: Depth=1
; GFX908-NEXT: s_waitcnt vmcnt(0)
-; GFX908-NEXT: v_max_f64 v[0:1], v[2:3], v[2:3]
-; GFX908-NEXT: v_max_f64 v[0:1], v[0:1], v[4:5]
; GFX908-NEXT: v_mov_b32_e32 v10, v3
; GFX908-NEXT: v_mov_b32_e32 v9, v2
-; GFX908-NEXT: v_mov_b32_e32 v8, v1
-; GFX908-NEXT: v_mov_b32_e32 v7, v0
-; GFX908-NEXT: buffer_atomic_cmpswap_x2 v[7:10], v6, s[16:19], 0 offen glc
+; GFX908-NEXT: v_max_f64 v[2:3], v[9:10], v[9:10]
+; GFX908-NEXT: v_max_f64 v[7:8], v[2:3], v[0:1]
+; GFX908-NEXT: v_mov_b32_e32 v2, v7
+; GFX908-NEXT: v_mov_b32_e32 v3, v8
+; GFX908-NEXT: v_mov_b32_e32 v4, v9
+; GFX908-NEXT: v_mov_b32_e32 v5, v10
+; GFX908-NEXT: buffer_atomic_cmpswap_x2 v[2:5], v6, s[16:19], 0 offen glc
; GFX908-NEXT: s_waitcnt vmcnt(0)
; GFX908-NEXT: buffer_wbinvl1
-; GFX908-NEXT: v_cmp_eq_u64_e32 vcc, v[7:8], v[2:3]
-; GFX908-NEXT: v_mov_b32_e32 v2, v7
+; GFX908-NEXT: v_cmp_eq_u64_e32 vcc, v[2:3], v[9:10]
; GFX908-NEXT: s_or_b64 s[4:5], vcc, s[4:5]
-; GFX908-NEXT: v_mov_b32_e32 v3, v8
; GFX908-NEXT: s_andn2_b64 exec, exec, s[4:5]
; GFX908-NEXT: s_cbranch_execnz .LBB15_1
; GFX908-NEXT: ; %bb.2: ; %atomicrmw.end
@@ -2106,24 +2108,24 @@ define void @buffer_fat_ptr_agent_atomic_fmax_noret_f64__amdgpu_no_fine_grained_
; GFX8-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX8-NEXT: v_mov_b32_e32 v6, s20
; GFX8-NEXT: buffer_load_dwordx2 v[2:3], v6, s[16:19], 0 offen
-; GFX8-NEXT: v_max_f64 v[4:5], v[0:1], v[0:1]
+; GFX8-NEXT: v_max_f64 v[0:1], v[0:1], v[0:1]
; GFX8-NEXT: s_mov_b64 s[4:5], 0
; GFX8-NEXT: .LBB15_1: ; %atomicrmw.start
; GFX8-NEXT: ; =>This Inner Loop Header: Depth=1
; GFX8-NEXT: s_waitcnt vmcnt(0)
-; GFX8-NEXT: v_max_f64 v[0:1], v[2:3], v[2:3]
-; GFX8-NEXT: v_max_f64 v[0:1], v[0:1], v[4:5]
; GFX8-NEXT: v_mov_b32_e32 v10, v3
; GFX8-NEXT: v_mov_b32_e32 v9, v2
-; GFX8-NEXT: v_mov_b32_e32 v8, v1
-; GFX8-NEXT: v_mov_b32_e32 v7, v0
-; GFX8-NEXT: buffer_atomic_cmpswap_x2 v[7:10], v6, s[16:19], 0 offen glc
+; GFX8-NEXT: v_max_f64 v[2:3], v[9:10], v[9:10]
+; GFX8-NEXT: v_max_f64 v[7:8], v[2:3], v[0:1]
+; GFX8-NEXT: v_mov_b32_e32 v2, v7
+; GFX8-NEXT: v_mov_b32_e32 v3, v8
+; GFX8-NEXT: v_mov_b32_e32 v4, v9
+; GFX8-NEXT: v_mov_b32_e32 v5, v10
+; GFX8-NEXT: buffer_atomic_cmpswap_x2 v[2:5], v6, s[16:19], 0 offen glc
; GFX8-NEXT: s_waitcnt vmcnt(0)
; GFX8-NEXT: buffer_wbinvl1
-; GFX8-NEXT: v_cmp_eq_u64_e32 vcc, v[7:8], v[2:3]
-; GFX8-NEXT: v_mov_b32_e32 v2, v7
+; GFX8-NEXT: v_cmp_eq_u64_e32 vcc, v[2:3], v[9:10]
; GFX8-NEXT: s_or_b64 s[4:5], vcc, s[4:5]
-; GFX8-NEXT: v_mov_b32_e32 v3, v8
; GFX8-NEXT: s_andn2_b64 exec, exec, s[4:5]
; GFX8-NEXT: s_cbranch_execnz .LBB15_1
; GFX8-NEXT: ; %bb.2: ; %atomicrmw.end
diff --git a/llvm/test/CodeGen/AMDGPU/GlobalISel/atomicrmw_fmin.ll b/llvm/test/CodeGen/AMDGPU/GlobalISel/atomicrmw_fmin.ll
index 351502816ae6e..404da6a8a1ef7 100644
--- a/llvm/test/CodeGen/AMDGPU/GlobalISel/atomicrmw_fmin.ll
+++ b/llvm/test/CodeGen/AMDGPU/GlobalISel/atomicrmw_fmin.ll
@@ -1670,23 +1670,23 @@ define void @buffer_fat_ptr_agent_atomic_fmin_noret_f32__amdgpu_no_fine_grained_
; GFX942-LABEL: buffer_fat_ptr_agent_atomic_fmin_noret_f32__amdgpu_no_fine_grained_memory:
; GFX942: ; %bb.0:
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
-; GFX942-NEXT: v_mov_b32_e32 v2, s16
-; GFX942-NEXT: buffer_load_dword v1, v2, s[0:3], 0 offen
+; GFX942-NEXT: v_mov_b32_e32 v1, s16
+; GFX942-NEXT: buffer_load_dword v2, v1, s[0:3], 0 offen
; GFX942-NEXT: s_mov_b64 s[4:5], 0
-; GFX942-NEXT: v_max_f32_e32 v3, v0, v0
+; GFX942-NEXT: v_max_f32_e32 v0, v0, v0
; GFX942-NEXT: .LBB13_1: ; %atomicrmw.start
; GFX942-NEXT: ; =>This Inner Loop Header: Depth=1
; GFX942-NEXT: s_waitcnt vmcnt(0)
-; GFX942-NEXT: v_max_f32_e32 v0, v1, v1
-; GFX942-NEXT: v_min_f32_e32 v0, v0, v3
-; GFX942-NEXT: v_mov_b64_e32 v[4:5], v[0:1]
+; GFX942-NEXT: v_mov_b32_e32 v5, v2
+; GFX942-NEXT: v_max_f32_e32 v2, v5, v5
+; GFX942-NEXT: v_min_f32_e32 v4, v2, v0
+; GFX942-NEXT: v_mov_b64_e32 v[2:3], v[4:5]
; GFX942-NEXT: buffer_wbl2 sc1
-; GFX942-NEXT: buffer_atomic_cmpswap v[4:5], v2, s[0:3], 0 offen sc0
+; GFX942-NEXT: buffer_atomic_cmpswap v[2:3], v1, s[0:3], 0 offen sc0
; GFX942-NEXT: s_waitcnt vmcnt(0)
; GFX942-NEXT: buffer_inv sc1
-; GFX942-NEXT: v_cmp_eq_u32_e32 vcc, v4, v1
+; GFX942-NEXT: v_cmp_eq_u32_e32 vcc, v2, v5
; GFX942-NEXT: s_or_b64 s[4:5], vcc, s[4:5]
-; GFX942-NEXT: v_mov_b32_e32 v1, v4
; GFX942-NEXT: s_andn2_b64 exec, exec, s[4:5]
; GFX942-NEXT: s_cbranch_execnz .LBB13_1
; GFX942-NEXT: ; %bb.2: ; %atomicrmw.end
@@ -1718,22 +1718,22 @@ define void @buffer_fat_ptr_agent_atomic_fmin_noret_f32__amdgpu_no_fine_grained_
; GFX90A-LABEL: buffer_fat_ptr_agent_atomic_fmin_noret_f32__amdgpu_no_fine_grained_memory:
; GFX90A: ; %bb.0:
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
-; GFX90A-NEXT: v_mov_b32_e32 v2, s20
-; GFX90A-NEXT: buffer_load_dword v1, v2, s[16:19], 0 offen
+; GFX90A-NEXT: v_mov_b32_e32 v1, s20
+; GFX90A-NEXT: buffer_load_dword v2, v1, s[16:19], 0 offen
; GFX90A-NEXT: s_mov_b64 s[4:5], 0
-; GFX90A-NEXT: v_max_f32_e32 v3, v0, v0
+; GFX90A-NEXT: v_max_f32_e32 v0, v0, v0
; GFX90A-NEXT: .LBB13_1: ; %atomicrmw.start
; GFX90A-NEXT: ; =>This Inner Loop Header: Depth=1
; GFX90A-NEXT: s_waitcnt vmcnt(0)
-; GFX90A-NEXT: v_max_f32_e32 v0, v1, v1
-; GFX90A-NEXT: v_min_f32_e32 v0, v0, v3
-; GFX90A-NEXT: v_pk_mov_b32 v[4:5], v[0:1], v[0:1] op_sel:[0,1]
-; GFX90A-NEXT: buffer_atomic_cmpswap v[4:5], v2, s[16:19], 0 offen glc
+; GFX90A-NEXT: v_mov_b32_e32 v5, v2
+; GFX90A-NEXT: v_max_f32_e32 v2, v5, v5
+; GFX90A-NEXT: v_min_f32_e32 v4, v2, v0
+; GFX90A-NEXT: v_pk_mov_b32 v[2:3], v[4:5], v[4:5] op_sel:[0,1]
+; GFX90A-NEXT: buffer_atomic_cmpswap v[2:3], v1, s[16:19], 0 offen glc
; GFX90A-NEXT: s_waitcnt vmcnt(0)
; GFX90A-NEXT: buffer_wbinvl1
-; GFX90A-NEXT: v_cmp_eq_u32_e32 vcc, v4, v1
+; GFX90A-NEXT: v_cmp_eq_u32_e32 vcc, v2, v5
; GFX90A-NEXT: s_or_b64 s[4:5], vcc, s[4:5]
-; GFX90A-NEXT: v_mov_b32_e32 v1, v4
; GFX90A-NEXT: s_andn2_b64 exec, exec, s[4:5]
; GFX90A-NEXT: s_cbranch_execnz .LBB13_1
; GFX90A-NEXT: ; %bb.2: ; %atomicrmw.end
@@ -1743,23 +1743,23 @@ define void @buffer_fat_ptr_agent_atomic_fmin_noret_f32__amdgpu_no_fine_grained_
; GFX908-LABEL: buffer_fat_ptr_agent_atomic_fmin_noret_f32__amdgpu_no_fine_grained_memory:
; GFX908: ; %bb.0:
; GFX908-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
-; GFX908-NEXT: v_mov_b32_e32 v2, s20
-; GFX908-NEXT: buffer_load_dword v1, v2, s[16:19], 0 offen
+; GFX908-NEXT: v_mov_b32_e32 v3, s20
+; GFX908-NEXT: buffer_load_dword v1, v3, s[16:19], 0 offen
; GFX908-NEXT: s_mov_b64 s[4:5], 0
-; GFX908-NEXT: v_max_f32_e32 v3, v0, v0
+; GFX908-NEXT: v_max_f32_e32 v0, v0, v0
; GFX908-NEXT: .LBB13_1: ; %atomicrmw.start
; GFX908-NEXT: ; =>This Inner Loop Header: Depth=1
; GFX908-NEXT: s_waitcnt vmcnt(0)
-; GFX908-NEXT: v_max_f32_e32 v0, v1, v1
-; GFX908-NEXT: v_min_f32_e32 v0, v0, v3
; GFX908-NEXT: v_mov_b32_e32 v5, v1
-; GFX908-NEXT: v_mov_b32_e32 v4, v0
-; GFX908-NEXT: buffer_atomic_cmpswap v[4:5], v2, s[16:19], 0 offen glc
+; GFX908-NEXT: v_max_f32_e32 v1, v5, v5
+; GFX908-NEXT: v_min_f32_e32 v4, v1, v0
+; GFX908-NEXT: v_mov_b32_e32 v1, v4
+; GFX908-NEXT: v_mov_b32_e32 v2, v5
+; GFX908-NEXT: buffer_atomic_cmpswap v[1:2], v3, s[16:19], 0 offen glc
; GFX908-NEXT: s_waitcnt vmcnt(0)
; GFX908-NEXT: buffer_wbinvl1
-; GFX908-NEXT: v_cmp_eq_u32_e32 vcc, v4, v1
+; GFX908-NEXT: v_cmp_eq_u32_e32 vcc, v1, v5
; GFX908-NEXT: s_or_b64 s[4:5], vcc, s[4:5]
-; GFX908-NEXT: v_mov_b32_e32 v1, v4
; GFX908-NEXT: s_andn2_b64 exec, exec, s[4:5]
; GFX908-NEXT: s_cbranch_execnz .LBB13_1
; GFX908-NEXT: ; %bb.2: ; %atomicrmw.end
@@ -1769,23 +1769,23 @@ define void @buffer_fat_ptr_agent_atomic_fmin_noret_f32__amdgpu_no_fine_grained_
; GFX8-LABEL: buffer_fat_ptr_agent_atomic_fmin_noret_f32__amdgpu_no_fine_grained_memory:
; GFX8: ; %bb.0:
; GFX8-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
-; GFX8-NEXT: v_mov_b32_e32 v2, s20
-; GFX8-NEXT: buffer_load_dword v1, v2, s[16:19], 0 offen
+; GFX8-NEXT: v_mov_b32_e32 v3, s20
+; GFX8-NEXT: buffer_load_dword v1, v3, s[16:19], 0 offen
; GFX8-NEXT: s_mov_b64 s[4:5], 0
-; GFX8-NEXT: v_mul_f32_e32 v3, 1.0, v0
+; GFX8-NEXT: v_mul_f32_e32 v0, 1.0, v0
; GFX8-NEXT: .LBB13_1: ; %atomicrmw.start
; GFX8-NEXT: ; =>This Inner Loop Header: Depth=1
; GFX8-NEXT: s_waitcnt vmcnt(0)
-; GFX8-NEXT: v_mul_f32_e32 v0, 1.0, v1
-; GFX8-NEXT: v_min_f32_e32 v0, v0, v3
; GFX8-NEXT: v_mov_b32_e32 v5, v1
-; GFX8-NEXT: v_mov_b32_e32 v4, v0
-; GFX8-NEXT: buffer_atomic_cmpswap v[4:5], v2, s[16:19], 0 offen glc
+; GFX8-NEXT: v_mul_f32_e32 v1, 1.0, v5
+; GFX8-NEXT: v_min_f32_e32 v4, v1, v0
+; GFX8-NEXT: v_mov_b32_e32 v1, v4
+; GFX8-NEXT: v_mov_b32_e32 v2, v5
+; GFX8-NEXT: buffer_atomic_cmpswap v[1:2], v3, s[16:19], 0 offen glc
; GFX8-NEXT: s_waitcnt vmcnt(0)
; GFX8-NEXT: buffer_wbinvl1
-; GFX8-NEXT: v_cmp_eq_u32_e32 vcc, v4, v1
+; GFX8-NEXT: v_cmp_eq_u32_e32 vcc, v1, v5
; GFX8-NEXT: s_or_b64 s[4:5], vcc, s[4:5]
-; GFX8-NEXT: v_mov_b32_e32 v1, v4
; GFX8-NEXT: s_andn2_b64 exec, exec, s[4:5]
; GFX8-NEXT: s_cbranch_execnz .LBB13_1
; GFX8-NEXT: ; %bb.2: ; %atomicrmw.end
@@ -1986,23 +1986,24 @@ define void @buffer_fat_ptr_agent_atomic_fmin_noret_f64__amdgpu_no_fine_grained_
; GFX12-NEXT: s_wait_bvhcnt 0x0
; GFX12-NEXT: s_wait_kmcnt 0x0
; GFX12-NEXT: v_mov_b32_e32 v6, s16
-; GFX12-NEXT: v_max_num_f64_e32 v[4:5], v[0:1], v[0:1]
+; GFX12-NEXT: v_max_num_f64_e32 v[0:1], v[0:1], v[0:1]
; GFX12-NEXT: s_mov_b32 s4, 0
; GFX12-NEXT: buffer_load_b64 v[2:3], v6, s[0:3], null offen
; GFX12-NEXT: .LBB15_1: ; %atomicrmw.start
; GFX12-NEXT: ; =>This Inner Loop Header: Depth=1
; GFX12-NEXT: s_wait_loadcnt 0x0
-; GFX12-NEXT: v_max_num_f64_e32 v[0:1], v[2:3], v[2:3]
-; GFX12-NEXT: s_wait_storecnt 0x0
-; GFX12-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(SKIP_1) | instid1(VALU_DEP_2)
-; GFX12-NEXT: v_min_num_f64_e32 v[0:1], v[0:1], v[4:5]
; GFX12-NEXT: v_dual_mov_b32 v10, v3 :: v_dual_mov_b32 v9, v2
-; GFX12-NEXT: v_dual_mov_b32 v8, v1 :: v_dual_mov_b32 v7, v0
-; GFX12-NEXT: buffer_atomic_cmpswap_b64 v[7:10], v6, s[0:3], null offen th:TH_ATOMIC_RETURN
+; GFX12-NEXT: s_wait_storecnt 0x0
+; GFX12-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
+; GFX12-NEXT: v_max_num_f64_e32 v[2:3], v[9:10], v[9:10]
+; GFX12-NEXT: v_min_num_f64_e32 v[7:8], v[2:3], v[0:1]
+; GFX12-NEXT: s_delay_alu instid0(VALU_DEP_1)
+; GFX12-NEXT: v_dual_mov_b32 v2, v7 :: v_dual_mov_b32 v3, v8
+; GFX12-NEXT: v_dual_mov_b32 v4, v9 :: v_dual_mov_b32 v5, v10
+; GFX12-NEXT: buffer_atomic_cmpswap_b64 v[2:5], v6, s[0:3], null offen th:TH_ATOMIC_RETURN
; GFX12-NEXT: s_wait_loadcnt 0x0
; GFX12-NEXT: global_inv scope:SCOPE_DEV
-; GFX12-NEXT: v_cmp_eq_u64_e32 vcc_lo, v[7:8], v[2:3]
-; GFX12-NEXT: v_dual_mov_b32 v2, v7 :: v_dual_mov_b32 v3, v8
+; GFX12-NEXT: v_cmp_eq_u64_e32 vcc_lo, v[2:3], v[9:10]
; GFX12-NEXT: s_or_b32 s4, vcc_lo, s4
; GFX12-NEXT: s_wait_alu 0xfffe
; GFX12-NEXT: s_and_not1_b32 exec_lo, exec_lo, s4
@@ -2026,24 +2027,25 @@ define void @buffer_fat_ptr_agent_atomic_fmin_noret_f64__amdgpu_no_fine_grained_
; GFX11: ; %bb.0:
; GFX11-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX11-NEXT: v_mov_b32_e32 v6, s16
-; GFX11-NEXT: v_max_f64 v[4:5], v[0:1], v[0:1]
+; GFX11-NEXT: v_max_f64 v[0:1], v[0:1], v[0:1]
; GFX11-NEXT: s_mov_b32 s4, 0
; GFX11-NEXT: buffer_load_b64 v[2:3], v6, s[0:3], 0 offen
; GFX11-NEXT: .LBB15_1: ; %atomicrmw.start
; GFX11-NEXT: ; =>This Inner Loop Header: Depth=1
; GFX11-NEXT: s_waitcnt vmcnt(0)
-; GFX11-NEXT: v_max_f64 v[0:1], v[2:3], v[2:3]
-; GFX11-NEXT: s_waitcnt_vscnt null, 0x0
-; GFX11-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(SKIP_1) | instid1(VALU_DEP_2)
-; GFX11-NEXT: v_min_f64 v[0:1], v[0:1], v[4:5]
; GFX11-NEXT: v_dual_mov_b32 v10, v3 :: v_dual_mov_b32 v9, v2
-; GFX11-NEXT: v_dual_mov_b32 v8, v1 :: v_dual_mov_b32 v7, v0
-; GFX11-NEXT: buffer_atomic_cmpswap_b64 v[7:10], v6, s[0:3], 0 offen glc
+; GFX11-NEXT: s_waitcnt_vscnt null, 0x0
+; GFX11-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
+; GFX11-NEXT: v_max_f64 v[2:3], v[9:10], v[9:10]
+; GFX11-NEXT: v_min_f64 v[7:8], v[2:3], v[0:1]
+; GFX11-NEXT: s_delay_alu instid0(VALU_DEP_1)
+; GFX11-NEXT: v_dual_mov_b32 v2, v7 :: v_dual_mov_b32 v3, v8
+; GFX11-NEXT: v_dual_mov_b32 v4, v9 :: v_dual_mov_b32 v5, v10
+; GFX11-NEXT: buffer_atomic_cmpswap_b64 v[2:5], v6, s[0:3], 0 offen glc
; GFX11-NEXT: s_waitcnt vmcnt(0)
; GFX11-NEXT: buffer_gl1_inv
; GFX11-NEXT: buffer_gl0_inv
-; GFX11-NEXT: v_cmp_eq_u64_e32 vcc_lo, v[7:8], v[2:3]
-; GFX11-NEXT: v_dual_mov_b32 v2, v7 :: v_dual_mov_b32 v3, v8
+; GFX11-NEXT: v_cmp_eq_u64_e32 vcc_lo, v[2:3], v[9:10]
; GFX11-NEXT: s_or_b32 s4, vcc_lo, s4
; GFX11-NEXT: s_delay_alu instid0(SALU_CYCLE_1)
; GFX11-NEXT: s_and_not1_b32 exec_lo, exec_lo, s4
@@ -2077,24 +2079,24 @@ define void @buffer_fat_ptr_agent_atomic_fmin_noret_f64__amdgpu_no_fine_grained_
; GFX908-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX908-NEXT: v_mov_b32_e32 v6, s20
; GFX908-NEXT: buffer_load_dwordx2 v[2:3], v6, s[16:19], 0 offen
-; GFX908-NEXT: v_max_f64 v[4:5], v[0:1], v[0:1]
+; GFX908-NEXT: v_max_f64 v[0:1], v[0:1], v[0:1]
; GFX908-NEXT: s_mov_b64 s[4:5], 0
; GFX908-NEXT: .LBB15_1: ; %atomicrmw.start
; GFX908-NEXT: ; =>This Inner Loop Header: Depth=1
; GFX908-NEXT: s_waitcnt vmcnt(0)
-; GFX908-NEXT: v_max_f64 v[0:1], v[2:3], v[2:3]
-; GFX908-NEXT: v_min_f64 v[0:1], v[0:1], v[4:5]
; GFX908-NEXT: v_mov_b32_e32 v10, v3
; GFX908-NEXT: v_mov_b32_e32 v9, v2
-; GFX908-NEXT: v_mov_b32_e32 v8, v1
-; GFX908-NEXT: v_mov_b32_e32 v7, v0
-; GFX908-NEXT: buffer_atomic_cmpswap_x2 v[7:10], v6, s[16:19], 0 offen glc
+; GFX908-NEXT: v_max_f64 v[2:3], v[9:10], v[9:10]
+; GFX908-NEXT: v_min_f64 v[7:8], v[2:3], v[0:1]
+; GFX908-NEXT: v_mov_b32_e32 v2, v7
+; GFX908-NEXT: v_mov_b32_e32 v3, v8
+; GFX908-NEXT: v_mov_b32_e32 v4, v9
+; GFX908-NEXT: v_mov_b32_e32 v5, v10
+; GFX908-NEXT: buffer_atomic_cmpswap_x2 v[2:5], v6, s[16:19], 0 offen glc
; GFX908-NEXT: s_waitcnt vmcnt(0)
; GFX908-NEXT: buffer_wbinvl1
-; GFX908-NEXT: v_cmp_eq_u64_e32 vcc, v[7:8], v[2:3]
-; GFX908-NEXT: v_mov_b32_e32 v2, v7
+; GFX908-NEXT: v_cmp_eq_u64_e32 vcc, v[2:3], v[9:10]
; GFX908-NEXT: s_or_b64 s[4:5], vcc, s[4:5]
-; GFX908-NEXT: v_mov_b32_e32 v3, v8
; GFX908-NEXT: s_andn2_b64 exec, exec, s[4:5]
; GFX908-NEXT: s_cbranch_execnz .LBB15_1
; GFX908-NEXT: ; %bb.2: ; %atomicrmw.end
@@ -2106,24 +2108,24 @@ define void @buffer_fat_ptr_agent_atomic_fmin_noret_f64__amdgpu_no_fine_grained_
; GFX8-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX8-NEXT: v_mov_b32_e32 v6, s20
; GFX8-NEXT: buffer_load_dwordx2 v[2:3], v6, s[16:19], 0 offen
-; GFX8-NEXT: v_max_f64 v[4:5], v[0:1], v[0:1]
+; GFX8-NEXT: v_max_f64 v[0:1], v[0:1], v[0:1]
; GFX8-NEXT: s_mov_b64 s[4:5], 0
; GFX8-NEXT: .LBB15_1: ; %atomicrmw.start
; GFX8-NEXT: ; =>This Inner Loop Header: Depth=1
; GFX8-NEXT: s_waitcnt vmcnt(0)
-; GFX8-NEXT: v_max_f64 v[0:1], v[2:3], v[2:3]
-; GFX8-NEXT: v_min_f64 v[0:1], v[0:1], v[4:5]
; GFX8-NEXT: v_mov_b32_e32 v10, v3
; GFX8-NEXT: v_mov_b32_e32 v9, v2
-; GFX8-NEXT: v_mov_b32_e32 v8, v1
-; GFX8-NEXT: v_mov_b32_e32 v7, v0
-; GFX8-NEXT: buffer_atomic_cmpswap_x2 v[7:10], v6, s[16:19], 0 offen glc
+; GFX8-NEXT: v_max_f64 v[2:3], v[9:10], v[9:10]
+; GFX8-NEXT: v_min_f64 v[7:8], v[2:3], v[0:1]
+; GFX8-NEXT: v_mov_b32_e32 v2, v7
+; GFX8-NEXT: v_mov_b32_e32 v3, v8
+; GFX8-NEXT: v_mov_b32_e32 v4, v9
+; GFX8-NEXT: v_mov_b32_e32 v5, v10
+; GFX8-NEXT: buffer_atomic_cmpswap_x2 v[2:5], v6, s[16:19], 0 offen glc
; GFX8-NEXT: s_waitcnt vmcnt(0)
; GFX8-NEXT: buffer_wbinvl1
-; GFX8-NEXT: v_cmp_eq_u64_e32 vcc, v[7:8], v[2:3]
-; GFX8-NEXT: v_mov_b32_e32 v2, v7
+; GFX8-NEXT: v_cmp_eq_u64_e32 vcc, v[2:3], v[9:10]
; GFX8-NEXT: s_or_b64 s[4:5], vcc, s[4:5]
-; GFX8-NEXT: v_mov_b32_e32 v3, v8
; GFX8-NEXT: s_andn2_b64 exec, exec, s[4:5]
; GFX8-NEXT: s_cbranch_execnz .LBB15_1
; GFX8-NEXT: ; %bb.2: ; %atomicrmw.end
diff --git a/llvm/test/CodeGen/AMDGPU/GlobalISel/divergence-temporal-divergent-i1.ll b/llvm/test/CodeGen/AMDGPU/GlobalISel/divergence-temporal-divergent-i1.ll
index 0da25260e2317..350f9ad882ed0 100644
--- a/llvm/test/CodeGen/AMDGPU/GlobalISel/divergence-temporal-divergent-i1.ll
+++ b/llvm/test/CodeGen/AMDGPU/GlobalISel/divergence-temporal-divergent-i1.ll
@@ -387,11 +387,11 @@ define void @nested_loops_temporal_divergence_both(float %pre.cond.val, i32 %n.i
; GFX10-NEXT: ; =>This Loop Header: Depth=1
; GFX10-NEXT: ; Child Loop BB5_2 Depth 2
; GFX10-NEXT: s_ashr_i32 s7, s6, 31
-; GFX10-NEXT: s_mov_b32 s4, s8
+; GFX10-NEXT: s_mov_b32 s4, 0
; GFX10-NEXT: s_lshl_b64 s[10:11], s[6:7], 2
; GFX10-NEXT: v_mov_b32_e32 v8, s10
; GFX10-NEXT: v_mov_b32_e32 v9, s11
-; GFX10-NEXT: s_mov_b32 s10, 0
+; GFX10-NEXT: s_mov_b32 s10, s8
; GFX10-NEXT: s_mov_b32 s11, 0
; GFX10-NEXT: v_add_co_u32 v8, vcc_lo, v2, v8
; GFX10-NEXT: v_add_co_ci_u32_e32 v9, vcc_lo, v3, v9, vcc_lo
@@ -402,18 +402,18 @@ define void @nested_loops_temporal_divergence_both(float %pre.cond.val, i32 %n.i
; GFX10-NEXT: v_cvt_f32_u32_e32 v8, s11
; GFX10-NEXT: s_mov_b32 s12, exec_lo
; GFX10-NEXT: s_add_i32 s11, s11, 1
-; GFX10-NEXT: s_xor_b32 s4, s4, s12
+; GFX10-NEXT: s_xor_b32 s10, s10, s12
; GFX10-NEXT: s_waitcnt vmcnt(0) lgkmcnt(0)
; GFX10-NEXT: v_cmp_gt_f32_e32 vcc_lo, v8, v0
-; GFX10-NEXT: s_or_b32 s10, vcc_lo, s10
+; GFX10-NEXT: s_or_b32 s4, vcc_lo, s4
; GFX10-NEXT: s_andn2_b32 s9, s9, exec_lo
-; GFX10-NEXT: s_and_b32 s12, exec_lo, s4
+; GFX10-NEXT: s_and_b32 s12, exec_lo, s10
; GFX10-NEXT: s_or_b32 s9, s9, s12
-; GFX10-NEXT: s_andn2_b32 exec_lo, exec_lo, s10
+; GFX10-NEXT: s_andn2_b32 exec_lo, exec_lo, s4
; GFX10-NEXT: s_cbranch_execnz .LBB5_2
; GFX10-NEXT: ; %bb.3: ; %UseInst
; GFX10-NEXT: ; in Loop: Header=BB5_1 Depth=1
-; GFX10-NEXT: s_or_b32 exec_lo, exec_lo, s10
+; GFX10-NEXT: s_or_b32 exec_lo, exec_lo, s4
; GFX10-NEXT: v_mov_b32_e32 v9, s7
; GFX10-NEXT: v_mov_b32_e32 v8, s6
; GFX10-NEXT: v_cmp_lt_u32_e32 vcc_lo, s6, v1
diff --git a/llvm/test/CodeGen/AMDGPU/amdgcn.bitcast.1024bit.ll b/llvm/test/CodeGen/AMDGPU/amdgcn.bitcast.1024bit.ll
index f729de82cb042..67c4b349f9502 100644
--- a/llvm/test/CodeGen/AMDGPU/amdgcn.bitcast.1024bit.ll
+++ b/llvm/test/CodeGen/AMDGPU/amdgcn.bitcast.1024bit.ll
@@ -10475,7 +10475,7 @@ define inreg <128 x i8> @bitcast_v32i32_to_v128i8_scalar(<32 x i32> inreg %a, i3
; GFX11-TRUE16-NEXT: v_readfirstlane_b32 s5, v14
; GFX11-TRUE16-NEXT: v_writelane_b32 v16, s37, 5
; GFX11-TRUE16-NEXT: v_writelane_b32 v17, s101, 5
-; GFX11-TRUE16-NEXT: s_mov_b32 s57, 0
+; GFX11-TRUE16-NEXT: s_mov_b32 vcc_hi, 0
; GFX11-TRUE16-NEXT: s_and_b32 s42, vcc_lo, exec_lo
; GFX11-TRUE16-NEXT: ; implicit-def: $vgpr18 : SGPR spill to VGPR lane
; GFX11-TRUE16-NEXT: ; implicit-def: $vgpr19 : SGPR spill to VGPR lane
@@ -10653,7 +10653,7 @@ define inreg <128 x i8> @bitcast_v32i32_to_v128i8_scalar(<32 x i32> inreg %a, i3
; GFX11-TRUE16-NEXT: s_delay_alu instid0(SALU_CYCLE_1)
; GFX11-TRUE16-NEXT: v_writelane_b32 v18, s42, 0
; GFX11-TRUE16-NEXT: s_lshr_b32 s42, s18, 8
-; GFX11-TRUE16-NEXT: s_and_not1_b32 vcc_lo, exec_lo, s57
+; GFX11-TRUE16-NEXT: s_and_not1_b32 vcc_lo, exec_lo, vcc_hi
; GFX11-TRUE16-NEXT: s_cbranch_vccnz .LBB13_3
; GFX11-TRUE16-NEXT: .LBB13_2: ; %cmp.true
; GFX11-TRUE16-NEXT: s_add_i32 s5, s5, 3
@@ -11321,9 +11321,9 @@ define inreg <128 x i8> @bitcast_v32i32_to_v128i8_scalar(<32 x i32> inreg %a, i3
; GFX11-TRUE16-NEXT: s_waitcnt vmcnt(0)
; GFX11-TRUE16-NEXT: s_setpc_b64 s[30:31]
; GFX11-TRUE16-NEXT: .LBB13_4:
-; GFX11-TRUE16-NEXT: ; implicit-def: $sgpr58_lo16
-; GFX11-TRUE16-NEXT: v_writelane_b32 v18, s58, 0
-; GFX11-TRUE16-NEXT: ; implicit-def: $sgpr58_lo16
+; GFX11-TRUE16-NEXT: ; implicit-def: $sgpr57_lo16
+; GFX11-TRUE16-NEXT: v_writelane_b32 v18, s57, 0
+; GFX11-TRUE16-NEXT: ; implicit-def: $sgpr57_lo16
; GFX11-TRUE16-NEXT: ; implicit-def: $sgpr101_lo16
; GFX11-TRUE16-NEXT: ; implicit-def: $sgpr100_lo16
; GFX11-TRUE16-NEXT: ; implicit-def: $sgpr56_lo16
@@ -11381,72 +11381,72 @@ define inreg <128 x i8> @bitcast_v32i32_to_v128i8_scalar(<32 x i32> inreg %a, i3
; GFX11-TRUE16-NEXT: ; implicit-def: $sgpr72_lo16
; GFX11-TRUE16-NEXT: ; implicit-def: $sgpr62_lo16
; GFX11-TRUE16-NEXT: ; implicit-def: $sgpr60_lo16
-; GFX11-TRUE16-NEXT: v_writelane_b32 v18, s58, 1
-; GFX11-TRUE16-NEXT: ; implicit-def: $sgpr58_lo16
-; GFX11-TRUE16-NEXT: v_writelane_b32 v18, s58, 2
-; GFX11-TRUE16-NEXT: ; implicit-def: $sgpr58_lo16
-; GFX11-TRUE16-NEXT: v_writelane_b32 v18, s58, 3
-; GFX11-TRUE16-NEXT: ; implicit-def: $sgpr58_lo16
-; GFX11-TRUE16-NEXT: v_writelane_b32 v18, s58, 4
-; GFX11-TRUE16-NEXT: ; implicit-def: $sgpr58_lo16
-; GFX11-TRUE16-NEXT: v_writelane_b32 v18, s58, 5
-; GFX11-TRUE16-NEXT: ; implicit-def: $sgpr58_lo16
-; GFX11-TRUE16-NEXT: v_writelane_b32 v18, s58, 6
-; GFX11-TRUE16-NEXT: ; implicit-def: $sgpr58_lo16
-; GFX11-TRUE16-NEXT: v_writelane_b32 v18, s58, 7
-; GFX11-TRUE16-NEXT: ; implicit-def: $sgpr58_lo16
-; GFX11-TRUE16-NEXT: v_writelane_b32 v18, s58, 8
-; GFX11-TRUE16-NEXT: ; implicit-def: $sgpr58_lo16
-; GFX11-TRUE16-NEXT: v_writelane_b32 v18, s58, 9
-; GFX11-TRUE16-NEXT: ; implicit-def: $sgpr58_lo16
-; GFX11-TRUE16-NEXT: v_writelane_b32 v18, s58, 10
-; GFX11-TRUE16-NEXT: ; implicit-def: $sgpr58_lo16
-; GFX11-TRUE16-NEXT: v_writelane_b32 v18, s58, 11
-; GFX11-TRUE16-NEXT: ; implicit-def: $sgpr58_lo16
-; GFX11-TRUE16-NEXT: v_writelane_b32 v18, s58, 12
-; GFX11-TRUE16-NEXT: ; implicit-def: $sgpr58_lo16
-; GFX11-TRUE16-NEXT: v_writelane_b32 v18, s58, 13
-; GFX11-TRUE16-NEXT: ; implicit-def: $sgpr58_lo16
-; GFX11-TRUE16-NEXT: v_writelane_b32 v18, s58, 14
-; GFX11-TRUE16-NEXT: ; implicit-def: $sgpr58_lo16
-; GFX11-TRUE16-NEXT: v_writelane_b32 v18, s58, 15
-; GFX11-TRUE16-NEXT: ; implicit-def: $sgpr58_lo16
-; GFX11-TRUE16-NEXT: v_writelane_b32 v18, s58, 16
-; GFX11-TRUE16-NEXT: ; implicit-def: $sgpr58_lo16
-; GFX11-TRUE16-NEXT: v_writelane_b32 v18, s58, 17
-; GFX11-TRUE16-NEXT: ; implicit-def: $sgpr58_lo16
-; GFX11-TRUE16-NEXT: v_writelane_b32 v18, s58, 18
-; GFX11-TRUE16-NEXT: ; implicit-def: $sgpr58_lo16
-; GFX11-TRUE16-NEXT: v_writelane_b32 v18, s58, 19
-; GFX11-TRUE16-NEXT: ; implicit-def: $sgpr58_lo16
-; GFX11-TRUE16-NEXT: v_writelane_b32 v18, s58, 20
-; GFX11-TRUE16-NEXT: ; implicit-def: $sgpr58_lo16
-; GFX11-TRUE16-NEXT: v_writelane_b32 v18, s58, 21
-; GFX11-TRUE16-NEXT: ; implicit-def: $sgpr58_lo16
-; GFX11-TRUE16-NEXT: v_writelane_b32 v18, s58, 22
-; GFX11-TRUE16-NEXT: ; implicit-def: $sgpr58_lo16
-; GFX11-TRUE16-NEXT: v_writelane_b32 v18, s58, 23
-; GFX11-TRUE16-NEXT: ; implicit-def: $sgpr58_lo16
-; GFX11-TRUE16-NEXT: v_writelane_b32 v18, s58, 24
-; GFX11-TRUE16-NEXT: ; implicit-def: $sgpr58_lo16
-; GFX11-TRUE16-NEXT: v_writelane_b32 v18, s58, 25
-; GFX11-TRUE16-NEXT: ; implicit-def: $sgpr58_lo16
-; GFX11-TRUE16-NEXT: v_writelane_b32 v18, s58, 26
-; GFX11-TRUE16-NEXT: ; implicit-def: $sgpr58_lo16
-; GFX11-TRUE16-NEXT: v_writelane_b32 v18, s58, 27
-; GFX11-TRUE16-NEXT: ; implicit-def: $sgpr58_lo16
-; GFX11-TRUE16-NEXT: v_writelane_b32 v18, s58, 28
-; GFX11-TRUE16-NEXT: ; implicit-def: $sgpr58_lo16
-; GFX11-TRUE16-NEXT: v_writelane_b32 v18, s58, 29
-; GFX11-TRUE16-NEXT: ; implicit-def: $sgpr58_lo16
-; GFX11-TRUE16-NEXT: v_writelane_b32 v18, s58, 30
-; GFX11-TRUE16-NEXT: ; implicit-def: $sgpr58_lo16
-; GFX11-TRUE16-NEXT: v_writelane_b32 v18, s58, 31
-; GFX11-TRUE16-NEXT: ; implicit-def: $sgpr58_lo16
-; GFX11-TRUE16-NEXT: v_writelane_b32 v19, s58, 0
-; GFX11-TRUE16-NEXT: ; implicit-def: $sgpr58_lo16
-; GFX11-TRUE16-NEXT: v_writelane_b32 v19, s58, 1
; GFX11-TRUE16-NEXT: ; implicit-def: $sgpr58_lo16
+; GFX11-TRUE16-NEXT: v_writelane_b32 v18, s57, 1
+; GFX11-TRUE16-NEXT: ; implicit-def: $sgpr57_lo16
+; GFX11-TRUE16-NEXT: v_writelane_b32 v18, s57, 2
+; GFX11-TRUE16-NEXT: ; implicit-def: $sgpr57_lo16
+; GFX11-TRUE16-NEXT: v_writelane_b32 v18, s57, 3
+; GFX11-TRUE16-NEXT: ; implicit-def: $sgpr57_lo16
+; GFX11-TRUE16-NEXT: v_writelane_b32 v18, s57, 4
+; GFX11-TRUE16-NEXT: ; implicit-def: $sgpr57_lo16
+; GFX11-TRUE16-NEXT: v_writelane_b32 v18, s57, 5
+; GFX11-TRUE16-NEXT: ; implicit-def: $sgpr57_lo16
+; GFX11-TRUE16-NEXT: v_writelane_b32 v18, s57, 6
+; GFX11-TRUE16-NEXT: ; implicit-def: $sgpr57_lo16
+; GFX11-TRUE16-NEXT: v_writelane_b32 v18, s57, 7
+; GFX11-TRUE16-NEXT: ; implicit-def: $sgpr57_lo16
+; GFX11-TRUE16-NEXT: v_writelane_b32 v18, s57, 8
+; GFX11-TRUE16-NEXT: ; implicit-def: $sgpr57_lo16
+; GFX11-TRUE16-NEXT: v_writelane_b32 v18, s57, 9
+; GFX11-TRUE16-NEXT: ; implicit-def: $sgpr57_lo16
+; GFX11-TRUE16-NEXT: v_writelane_b32 v18, s57, 10
+; GFX11-TRUE16-NEXT: ; implicit-def: $sgpr57_lo16
+; GFX11-TRUE16-NEXT: v_writelane_b32 v18, s57, 11
+; GFX11-TRUE16-NEXT: ; implicit-def: $sgpr57_lo16
+; GFX11-TRUE16-NEXT: v_writelane_b32 v18, s57, 12
+; GFX11-TRUE16-NEXT: ; implicit-def: $sgpr57_lo16
+; GFX11-TRUE16-NEXT: v_writelane_b32 v18, s57, 13
+; GFX11-TRUE16-NEXT: ; implicit-def: $sgpr57_lo16
+; GFX11-TRUE16-NEXT: v_writelane_b32 v18, s57, 14
+; GFX11-TRUE16-NEXT: ; implicit-def: $sgpr57_lo16
+; GFX11-TRUE16-NEXT: v_writelane_b32 v18, s57, 15
+; GFX11-TRUE16-NEXT: ; implicit-def: $sgpr57_lo16
+; GFX11-TRUE16-NEXT: v_writelane_b32 v18, s57, 16
+; GFX11-TRUE16-NEXT: ; implicit-def: $sgpr57_lo16
+; GFX11-TRUE16-NEXT: v_writelane_b32 v18, s57, 17
+; GFX11-TRUE16-NEXT: ; implicit-def: $sgpr57_lo16
+; GFX11-TRUE16-NEXT: v_writelane_b32 v18, s57, 18
+; GFX11-TRUE16-NEXT: ; implicit-def: $sgpr57_lo16
+; GFX11-TRUE16-NEXT: v_writelane_b32 v18, s57, 19
+; GFX11-TRUE16-NEXT: ; implicit-def: $sgpr57_lo16
+; GFX11-TRUE16-NEXT: v_writelane_b32 v18, s57, 20
+; GFX11-TRUE16-NEXT: ; implicit-def: $sgpr57_lo16
+; GFX11-TRUE16-NEXT: v_writelane_b32 v18, s57, 21
+; GFX11-TRUE16-NEXT: ; implicit-def: $sgpr57_lo16
+; GFX11-TRUE16-NEXT: v_writelane_b32 v18, s57, 22
+; GFX11-TRUE16-NEXT: ; implicit-def: $sgpr57_lo16
+; GFX11-TRUE16-NEXT: v_writelane_b32 v18, s57, 23
+; GFX11-TRUE16-NEXT: ; implicit-def: $sgpr57_lo16
+; GFX11-TRUE16-NEXT: v_writelane_b32 v18, s57, 24
+; GFX11-TRUE16-NEXT: ; implicit-def: $sgpr57_lo16
+; GFX11-TRUE16-NEXT: v_writelane_b32 v18, s57, 25
+; GFX11-TRUE16-NEXT: ; implicit-def: $sgpr57_lo16
+; GFX11-TRUE16-NEXT: v_writelane_b32 v18, s57, 26
+; GFX11-TRUE16-NEXT: ; implicit-def: $sgpr57_lo16
+; GFX11-TRUE16-NEXT: v_writelane_b32 v18, s57, 27
+; GFX11-TRUE16-NEXT: ; implicit-def: $sgpr57_lo16
+; GFX11-TRUE16-NEXT: v_writelane_b32 v18, s57, 28
+; GFX11-TRUE16-NEXT: ; implicit-def: $sgpr57_lo16
+; GFX11-TRUE16-NEXT: v_writelane_b32 v18, s57, 29
+; GFX11-TRUE16-NEXT: ; implicit-def: $sgpr57_lo16
+; GFX11-TRUE16-NEXT: v_writelane_b32 v18, s57, 30
+; GFX11-TRUE16-NEXT: ; implicit-def: $sgpr57_lo16
+; GFX11-TRUE16-NEXT: v_writelane_b32 v18, s57, 31
+; GFX11-TRUE16-NEXT: ; implicit-def: $sgpr57_lo16
+; GFX11-TRUE16-NEXT: v_writelane_b32 v19, s57, 0
+; GFX11-TRUE16-NEXT: ; implicit-def: $sgpr57_lo16
+; GFX11-TRUE16-NEXT: v_writelane_b32 v19, s57, 1
; GFX11-TRUE16-NEXT: v_writelane_b32 v19, s100, 2
; GFX11-TRUE16-NEXT: v_writelane_b32 v19, s101, 3
; GFX11-TRUE16-NEXT: ; implicit-def: $sgpr100_lo16
@@ -11498,7 +11498,7 @@ define inreg <128 x i8> @bitcast_v32i32_to_v128i8_scalar(<32 x i32> inreg %a, i3
; GFX11-FAKE16-NEXT: v_readfirstlane_b32 s5, v14
; GFX11-FAKE16-NEXT: v_writelane_b32 v16, s37, 5
; GFX11-FAKE16-NEXT: v_writelane_b32 v17, s101, 5
-; GFX11-FAKE16-NEXT: s_mov_b32 s101, 0
+; GFX11-FAKE16-NEXT: s_mov_b32 s44, 0
; GFX11-FAKE16-NEXT: s_and_b32 s42, vcc_lo, exec_lo
; GFX11-FAKE16-NEXT: ; implicit-def: $vgpr19 : SGPR spill to VGPR lane
; GFX11-FAKE16-NEXT: ; implicit-def: $vgpr18 : SGPR spill to VGPR lane
@@ -11531,676 +11531,497 @@ define inreg <128 x i8> @bitcast_v32i32_to_v128i8_scalar(<32 x i32> inreg %a, i3
; GFX11-FAKE16-NEXT: v_writelane_b32 v16, s85, 29
; GFX11-FAKE16-NEXT: v_writelane_b32 v16, s86, 30
; GFX11-FAKE16-NEXT: v_writelane_b32 v16, s87, 31
-; GFX11-FAKE16-NEXT: s_cbranch_scc0 .LBB13_2
+; GFX11-FAKE16-NEXT: s_cbranch_scc0 .LBB13_4
; GFX11-FAKE16-NEXT: ; %bb.1: ; %cmp.false
-; GFX11-FAKE16-NEXT: s_lshr_b32 s43, s25, 8
-; GFX11-FAKE16-NEXT: s_lshr_b64 s[62:63], s[4:5], 24
-; GFX11-FAKE16-NEXT: v_writelane_b32 v19, s43, 16
-; GFX11-FAKE16-NEXT: s_lshr_b32 s43, s24, 16
-; GFX11-FAKE16-NEXT: s_lshr_b32 s104, s5, 24
-; GFX11-FAKE16-NEXT: s_lshr_b32 s102, s5, 16
-; GFX11-FAKE16-NEXT: s_lshr_b32 s103, s5, 8
-; GFX11-FAKE16-NEXT: v_writelane_b32 v19, s43, 17
-; GFX11-FAKE16-NEXT: s_lshr_b32 s43, s24, 8
-; GFX11-FAKE16-NEXT: s_lshr_b32 s57, s4, 16
-; GFX11-FAKE16-NEXT: s_lshr_b32 s47, s4, 8
-; GFX11-FAKE16-NEXT: s_lshr_b32 s46, s7, 24
-; GFX11-FAKE16-NEXT: v_writelane_b32 v19, s43, 18
-; GFX11-FAKE16-NEXT: s_lshr_b32 s43, s23, 24
-; GFX11-FAKE16-NEXT: s_lshr_b32 vcc_hi, s7, 16
-; GFX11-FAKE16-NEXT: s_lshr_b32 s34, s7, 8
-; GFX11-FAKE16-NEXT: s_lshr_b32 s69, s6, 16
-; GFX11-FAKE16-NEXT: v_writelane_b32 v19, s43, 19
-; GFX11-FAKE16-NEXT: s_lshr_b32 s43, s23, 16
-; GFX11-FAKE16-NEXT: s_lshr_b32 s56, s6, 8
-; GFX11-FAKE16-NEXT: s_lshr_b32 s35, s9, 24
-; GFX11-FAKE16-NEXT: s_lshr_b32 s36, s9, 16
-; GFX11-FAKE16-NEXT: v_writelane_b32 v19, s43, 20
-; GFX11-FAKE16-NEXT: s_lshr_b32 s43, s23, 8
-; GFX11-FAKE16-NEXT: s_lshr_b32 s37, s9, 8
-; GFX11-FAKE16-NEXT: s_lshr_b32 s38, s8, 16
-; GFX11-FAKE16-NEXT: s_lshr_b32 s39, s8, 8
-; GFX11-FAKE16-NEXT: v_writelane_b32 v19, s43, 21
-; GFX11-FAKE16-NEXT: s_lshr_b32 s43, s22, 16
-; GFX11-FAKE16-NEXT: s_lshr_b32 s48, s11, 24
-; GFX11-FAKE16-NEXT: s_lshr_b32 s49, s11, 16
-; GFX11-FAKE16-NEXT: s_lshr_b32 s50, s11, 8
-; GFX11-FAKE16-NEXT: v_writelane_b32 v19, s43, 22
-; GFX11-FAKE16-NEXT: s_lshr_b32 s43, s22, 8
-; GFX11-FAKE16-NEXT: s_lshr_b32 s51, s10, 16
-; GFX11-FAKE16-NEXT: s_lshr_b32 s52, s10, 8
-; GFX11-FAKE16-NEXT: s_lshr_b32 s53, s13, 24
-; GFX11-FAKE16-NEXT: v_writelane_b32 v19, s43, 23
-; GFX11-FAKE16-NEXT: s_lshr_b32 s43, s21, 24
-; GFX11-FAKE16-NEXT: s_lshr_b32 s54, s13, 16
-; GFX11-FAKE16-NEXT: s_lshr_b32 s55, s13, 8
-; GFX11-FAKE16-NEXT: s_lshr_b32 s64, s12, 16
-; GFX11-FAKE16-NEXT: v_writelane_b32 v19, s43, 24
-; GFX11-FAKE16-NEXT: s_lshr_b32 s43, s21, 16
-; GFX11-FAKE16-NEXT: s_lshr_b32 s65, s12, 8
-; GFX11-FAKE16-NEXT: s_lshr_b32 s66, s15, 24
-; GFX11-FAKE16-NEXT: s_lshr_b32 s67, s15, 16
-; GFX11-FAKE16-NEXT: v_writelane_b32 v19, s43, 25
-; GFX11-FAKE16-NEXT: s_lshr_b32 s43, s21, 8
-; GFX11-FAKE16-NEXT: s_lshr_b32 s68, s15, 8
-; GFX11-FAKE16-NEXT: s_lshr_b32 s59, s14, 16
-; GFX11-FAKE16-NEXT: s_lshr_b32 s58, s14, 8
-; GFX11-FAKE16-NEXT: v_writelane_b32 v19, s43, 26
-; GFX11-FAKE16-NEXT: s_lshr_b32 s43, s20, 16
-; GFX11-FAKE16-NEXT: s_lshr_b32 s70, s41, 24
-; GFX11-FAKE16-NEXT: s_lshr_b32 s71, s41, 16
-; GFX11-FAKE16-NEXT: s_lshr_b32 s60, s41, 8
-; GFX11-FAKE16-NEXT: v_writelane_b32 v19, s43, 27
-; GFX11-FAKE16-NEXT: s_lshr_b32 s43, s20, 8
-; GFX11-FAKE16-NEXT: s_lshr_b32 s80, s40, 16
-; GFX11-FAKE16-NEXT: s_lshr_b32 s61, s40, 8
-; GFX11-FAKE16-NEXT: s_lshr_b32 s81, s29, 24
-; GFX11-FAKE16-NEXT: v_writelane_b32 v19, s43, 28
-; GFX11-FAKE16-NEXT: s_lshr_b32 s43, s19, 24
-; GFX11-FAKE16-NEXT: s_lshr_b32 s82, s29, 16
-; GFX11-FAKE16-NEXT: s_lshr_b32 s83, s29, 8
-; GFX11-FAKE16-NEXT: s_lshr_b32 s84, s28, 16
-; GFX11-FAKE16-NEXT: v_writelane_b32 v19, s43, 29
-; GFX11-FAKE16-NEXT: s_lshr_b32 s43, s19, 16
-; GFX11-FAKE16-NEXT: s_lshr_b32 s85, s28, 8
-; GFX11-FAKE16-NEXT: s_lshr_b32 s86, s27, 24
-; GFX11-FAKE16-NEXT: s_lshr_b32 s72, s27, 16
-; GFX11-FAKE16-NEXT: v_writelane_b32 v19, s43, 30
-; GFX11-FAKE16-NEXT: s_lshr_b32 s43, s19, 8
-; GFX11-FAKE16-NEXT: s_lshr_b32 s87, s27, 8
-; GFX11-FAKE16-NEXT: s_lshr_b32 s73, s26, 16
-; GFX11-FAKE16-NEXT: s_lshr_b32 s96, s26, 8
-; GFX11-FAKE16-NEXT: v_writelane_b32 v19, s43, 31
-; GFX11-FAKE16-NEXT: s_lshr_b32 s43, s18, 16
-; GFX11-FAKE16-NEXT: s_lshr_b32 s97, s25, 24
-; GFX11-FAKE16-NEXT: v_writelane_b32 v18, s43, 0
-; GFX11-FAKE16-NEXT: s_lshr_b32 s43, s18, 8
-; GFX11-FAKE16-NEXT: v_writelane_b32 v19, s62, 14
-; GFX11-FAKE16-NEXT: s_lshr_b32 s42, s25, 16
-; GFX11-FAKE16-NEXT: s_lshr_b32 s74, s2, 16
-; GFX11-FAKE16-NEXT: v_writelane_b32 v18, s43, 1
-; GFX11-FAKE16-NEXT: s_lshr_b32 s43, s17, 24
-; GFX11-FAKE16-NEXT: v_writelane_b32 v19, s63, 15
-; GFX11-FAKE16-NEXT: s_lshr_b64 s[62:63], s[6:7], 24
+; GFX11-FAKE16-NEXT: s_lshr_b32 s42, s5, 24
+; GFX11-FAKE16-NEXT: s_lshr_b64 s[46:47], s[4:5], 24
+; GFX11-FAKE16-NEXT: v_writelane_b32 v19, s42, 4
+; GFX11-FAKE16-NEXT: s_lshr_b32 s42, s5, 16
+; GFX11-FAKE16-NEXT: s_lshr_b32 s102, s29, 8
+; GFX11-FAKE16-NEXT: s_lshr_b32 s103, s28, 16
+; GFX11-FAKE16-NEXT: s_lshr_b32 s104, s28, 8
+; GFX11-FAKE16-NEXT: v_writelane_b32 v19, s42, 5
+; GFX11-FAKE16-NEXT: s_lshr_b32 s42, s5, 8
+; GFX11-FAKE16-NEXT: s_lshr_b32 vcc_hi, s27, 24
+; GFX11-FAKE16-NEXT: s_lshr_b32 s34, s27, 16
+; GFX11-FAKE16-NEXT: s_lshr_b32 s35, s27, 8
+; GFX11-FAKE16-NEXT: v_writelane_b32 v19, s42, 6
+; GFX11-FAKE16-NEXT: s_lshr_b32 s42, s4, 16
+; GFX11-FAKE16-NEXT: s_lshr_b32 s36, s26, 16
+; GFX11-FAKE16-NEXT: s_lshr_b32 s37, s26, 8
+; GFX11-FAKE16-NEXT: s_lshr_b32 s38, s25, 24
+; GFX11-FAKE16-NEXT: v_writelane_b32 v19, s42, 7
+; GFX11-FAKE16-NEXT: s_lshr_b32 s42, s4, 8
+; GFX11-FAKE16-NEXT: s_lshr_b32 s39, s25, 16
+; GFX11-FAKE16-NEXT: s_lshr_b32 s48, s25, 8
+; GFX11-FAKE16-NEXT: s_lshr_b32 s49, s24, 16
+; GFX11-FAKE16-NEXT: v_writelane_b32 v19, s42, 8
+; GFX11-FAKE16-NEXT: s_lshr_b32 s42, s7, 24
+; GFX11-FAKE16-NEXT: s_lshr_b32 s50, s24, 8
+; GFX11-FAKE16-NEXT: s_lshr_b32 s51, s23, 24
+; GFX11-FAKE16-NEXT: s_lshr_b32 s52, s23, 16
+; GFX11-FAKE16-NEXT: v_writelane_b32 v19, s42, 9
+; GFX11-FAKE16-NEXT: s_lshr_b32 s42, s7, 16
+; GFX11-FAKE16-NEXT: s_lshr_b32 s53, s23, 8
+; GFX11-FAKE16-NEXT: s_lshr_b32 s54, s22, 16
+; GFX11-FAKE16-NEXT: s_lshr_b32 s55, s22, 8
+; GFX11-FAKE16-NEXT: v_writelane_b32 v19, s42, 10
+; GFX11-FAKE16-NEXT: s_lshr_b32 s42, s7, 8
+; GFX11-FAKE16-NEXT: s_lshr_b32 s64, s21, 24
+; GFX11-FAKE16-NEXT: s_lshr_b32 s65, s21, 16
+; GFX11-FAKE16-NEXT: s_lshr_b32 s66, s21, 8
+; GFX11-FAKE16-NEXT: v_writelane_b32 v19, s42, 11
+; GFX11-FAKE16-NEXT: s_lshr_b32 s42, s6, 16
+; GFX11-FAKE16-NEXT: s_lshr_b32 s67, s20, 16
+; GFX11-FAKE16-NEXT: s_lshr_b32 s68, s20, 8
+; GFX11-FAKE16-NEXT: s_lshr_b32 s69, s19, 24
+; GFX11-FAKE16-NEXT: v_writelane_b32 v19, s42, 12
+; GFX11-FAKE16-NEXT: s_lshr_b32 s42, s6, 8
+; GFX11-FAKE16-NEXT: s_lshr_b32 s70, s19, 16
+; GFX11-FAKE16-NEXT: s_lshr_b32 s71, s19, 8
+; GFX11-FAKE16-NEXT: s_lshr_b32 s80, s18, 16
+; GFX11-FAKE16-NEXT: v_writelane_b32 v19, s42, 13
+; GFX11-FAKE16-NEXT: s_lshr_b32 s42, s9, 24
+; GFX11-FAKE16-NEXT: s_lshr_b32 s81, s18, 8
+; GFX11-FAKE16-NEXT: s_lshr_b32 s82, s17, 24
+; GFX11-FAKE16-NEXT: s_lshr_b32 s83, s17, 16
+; GFX11-FAKE16-NEXT: v_writelane_b32 v19, s42, 14
+; GFX11-FAKE16-NEXT: s_lshr_b32 s42, s9, 16
+; GFX11-FAKE16-NEXT: s_lshr_b32 s84, s17, 8
+; GFX11-FAKE16-NEXT: s_lshr_b32 s85, s16, 16
+; GFX11-FAKE16-NEXT: s_lshr_b32 s86, s16, 8
+; GFX11-FAKE16-NEXT: v_writelane_b32 v19, s42, 15
+; GFX11-FAKE16-NEXT: s_lshr_b32 s42, s9, 8
+; GFX11-FAKE16-NEXT: s_lshr_b32 s87, s3, 24
+; GFX11-FAKE16-NEXT: s_lshr_b32 s96, s3, 16
+; GFX11-FAKE16-NEXT: s_lshr_b32 s97, s3, 8
+; GFX11-FAKE16-NEXT: v_writelane_b32 v19, s42, 16
+; GFX11-FAKE16-NEXT: s_lshr_b32 s42, s8, 16
+; GFX11-FAKE16-NEXT: s_lshr_b32 s43, s2, 8
; GFX11-FAKE16-NEXT: s_lshr_b32 s98, s1, 24
-; GFX11-FAKE16-NEXT: v_writelane_b32 v18, s43, 2
-; GFX11-FAKE16-NEXT: s_lshr_b32 s43, s17, 16
-; GFX11-FAKE16-NEXT: v_writelane_b32 v19, s62, 12
; GFX11-FAKE16-NEXT: s_lshr_b32 s99, s1, 16
+; GFX11-FAKE16-NEXT: v_writelane_b32 v19, s42, 17
+; GFX11-FAKE16-NEXT: s_lshr_b32 s42, s8, 8
; GFX11-FAKE16-NEXT: s_lshr_b32 s100, s1, 8
-; GFX11-FAKE16-NEXT: v_writelane_b32 v18, s43, 3
-; GFX11-FAKE16-NEXT: s_lshr_b32 s43, s17, 8
-; GFX11-FAKE16-NEXT: v_writelane_b32 v19, s63, 13
-; GFX11-FAKE16-NEXT: s_lshr_b64 s[62:63], s[8:9], 24
-; GFX11-FAKE16-NEXT: s_lshr_b32 s44, s0, 16
-; GFX11-FAKE16-NEXT: v_writelane_b32 v18, s43, 4
-; GFX11-FAKE16-NEXT: s_lshr_b32 s43, s16, 16
-; GFX11-FAKE16-NEXT: v_writelane_b32 v19, s62, 10
-; GFX11-FAKE16-NEXT: s_lshr_b32 s45, s0, 8
+; GFX11-FAKE16-NEXT: s_lshr_b32 s45, s0, 16
+; GFX11-FAKE16-NEXT: s_lshr_b32 s101, s0, 8
+; GFX11-FAKE16-NEXT: v_writelane_b32 v19, s42, 18
+; GFX11-FAKE16-NEXT: s_lshr_b32 s42, s11, 24
+; GFX11-FAKE16-NEXT: s_lshr_b64 s[56:57], s[10:11], 24
+; GFX11-FAKE16-NEXT: s_lshr_b64 s[58:59], s[12:13], 24
+; GFX11-FAKE16-NEXT: s_lshr_b64 s[60:61], s[14:15], 24
+; GFX11-FAKE16-NEXT: v_writelane_b32 v19, s42, 19
+; GFX11-FAKE16-NEXT: s_lshr_b32 s42, s11, 16
+; GFX11-FAKE16-NEXT: s_lshr_b64 s[62:63], s[40:41], 24
+; GFX11-FAKE16-NEXT: s_lshr_b64 s[72:73], s[28:29], 24
; GFX11-FAKE16-NEXT: s_lshr_b64 s[76:77], s[26:27], 24
-; GFX11-FAKE16-NEXT: v_writelane_b32 v18, s43, 5
-; GFX11-FAKE16-NEXT: s_lshr_b32 s43, s16, 8
-; GFX11-FAKE16-NEXT: v_writelane_b32 v19, s63, 11
-; GFX11-FAKE16-NEXT: s_lshr_b64 s[62:63], s[10:11], 24
+; GFX11-FAKE16-NEXT: v_writelane_b32 v19, s42, 20
+; GFX11-FAKE16-NEXT: s_lshr_b32 s42, s11, 8
; GFX11-FAKE16-NEXT: s_lshr_b64 s[88:89], s[24:25], 24
-; GFX11-FAKE16-NEXT: v_writelane_b32 v18, s43, 6
-; GFX11-FAKE16-NEXT: s_lshr_b32 s43, s3, 24
-; GFX11-FAKE16-NEXT: v_writelane_b32 v19, s62, 8
+; GFX11-FAKE16-NEXT: s_lshr_b64 s[74:75], s[22:23], 24
; GFX11-FAKE16-NEXT: s_lshr_b64 s[78:79], s[20:21], 24
+; GFX11-FAKE16-NEXT: v_writelane_b32 v19, s42, 21
+; GFX11-FAKE16-NEXT: s_lshr_b32 s42, s10, 16
; GFX11-FAKE16-NEXT: s_lshr_b64 s[90:91], s[18:19], 24
-; GFX11-FAKE16-NEXT: v_writelane_b32 v18, s43, 7
-; GFX11-FAKE16-NEXT: s_lshr_b32 s43, s3, 16
-; GFX11-FAKE16-NEXT: v_writelane_b32 v19, s63, 9
-; GFX11-FAKE16-NEXT: s_lshr_b64 s[62:63], s[12:13], 24
; GFX11-FAKE16-NEXT: s_lshr_b64 s[92:93], s[16:17], 24
-; GFX11-FAKE16-NEXT: v_writelane_b32 v18, s43, 8
-; GFX11-FAKE16-NEXT: s_lshr_b32 s43, s3, 8
-; GFX11-FAKE16-NEXT: v_writelane_b32 v19, s62, 6
; GFX11-FAKE16-NEXT: s_lshr_b64 s[94:95], s[2:3], 24
+; GFX11-FAKE16-NEXT: v_writelane_b32 v19, s42, 22
+; GFX11-FAKE16-NEXT: s_lshr_b32 s42, s10, 8
; GFX11-FAKE16-NEXT: s_lshr_b64 s[30:31], s[0:1], 24
-; GFX11-FAKE16-NEXT: v_writelane_b32 v18, s43, 9
-; GFX11-FAKE16-NEXT: s_lshr_b32 s43, s2, 8
-; GFX11-FAKE16-NEXT: v_writelane_b32 v19, s63, 7
-; GFX11-FAKE16-NEXT: s_lshr_b64 s[62:63], s[14:15], 24
-; GFX11-FAKE16-NEXT: s_delay_alu instid0(SALU_CYCLE_1) | instskip(SKIP_2) | instid1(SALU_CYCLE_1)
-; GFX11-FAKE16-NEXT: v_writelane_b32 v19, s62, 4
-; GFX11-FAKE16-NEXT: v_writelane_b32 v19, s63, 5
-; GFX11-FAKE16-NEXT: s_lshr_b64 s[62:63], s[40:41], 24
-; GFX11-FAKE16-NEXT: v_writelane_b32 v19, s62, 2
-; GFX11-FAKE16-NEXT: v_writelane_b32 v19, s63, 3
-; GFX11-FAKE16-NEXT: s_lshr_b64 s[62:63], s[28:29], 24
+; GFX11-FAKE16-NEXT: v_writelane_b32 v19, s42, 23
+; GFX11-FAKE16-NEXT: s_lshr_b32 s42, s13, 24
+; GFX11-FAKE16-NEXT: s_delay_alu instid0(SALU_CYCLE_1) | instskip(SKIP_1) | instid1(SALU_CYCLE_1)
+; GFX11-FAKE16-NEXT: v_writelane_b32 v19, s42, 24
+; GFX11-FAKE16-NEXT: s_lshr_b32 s42, s13, 16
+; GFX11-FAKE16-NEXT: v_writelane_b32 v19, s42, 25
+; GFX11-FAKE16-NEXT: s_lshr_b32 s42, s13, 8
+; GFX11-FAKE16-NEXT: s_delay_alu instid0(SALU_CYCLE_1) | instskip(SKIP_1) | instid1(SALU_CYCLE_1)
+; GFX11-FAKE16-NEXT: v_writelane_b32 v19, s42, 26
+; GFX11-FAKE16-NEXT: s_lshr_b32 s42, s12, 16
+; GFX11-FAKE16-NEXT: v_writelane_b32 v19, s42, 27
+; GFX11-FAKE16-NEXT: s_lshr_b32 s42, s12, 8
+; GFX11-FAKE16-NEXT: s_delay_alu instid0(SALU_CYCLE_1) | instskip(SKIP_1) | instid1(SALU_CYCLE_1)
+; GFX11-FAKE16-NEXT: v_writelane_b32 v19, s42, 28
+; GFX11-FAKE16-NEXT: s_lshr_b32 s42, s15, 24
+; GFX11-FAKE16-NEXT: v_writelane_b32 v19, s42, 29
+; GFX11-FAKE16-NEXT: s_lshr_b32 s42, s15, 16
+; GFX11-FAKE16-NEXT: s_delay_alu instid0(SALU_CYCLE_1) | instskip(SKIP_1) | instid1(SALU_CYCLE_1)
+; GFX11-FAKE16-NEXT: v_writelane_b32 v19, s42, 30
+; GFX11-FAKE16-NEXT: s_lshr_b32 s42, s15, 8
+; GFX11-FAKE16-NEXT: v_writelane_b32 v19, s42, 31
+; GFX11-FAKE16-NEXT: s_lshr_b32 s42, s14, 16
; GFX11-FAKE16-NEXT: s_delay_alu instid0(SALU_CYCLE_1)
-; GFX11-FAKE16-NEXT: v_writelane_b32 v19, s62, 0
-; GFX11-FAKE16-NEXT: v_writelane_b32 v19, s63, 1
-; GFX11-FAKE16-NEXT: s_lshr_b64 s[62:63], s[22:23], 24
-; GFX11-FAKE16-NEXT: s_branch .LBB13_3
-; GFX11-FAKE16-NEXT: .LBB13_2:
-; GFX11-FAKE16-NEXT: ; implicit-def: $vcc_hi
-; GFX11-FAKE16-NEXT: ; implicit-def: $vcc_lo
-; GFX11-FAKE16-NEXT: ; implicit-def: $sgpr42
-; GFX11-FAKE16-NEXT: ; kill: killed $sgpr42
-; GFX11-FAKE16-NEXT: s_mov_b32 s101, -1
-; GFX11-FAKE16-NEXT: v_writelane_b32 v19, vcc_lo, 0
-; GFX11-FAKE16-NEXT: ; implicit-def: $sgpr42
-; GFX11-FAKE16-NEXT: ; kill: killed $sgpr42
-; GFX11-FAKE16-NEXT: ; implicit-def: $sgpr42
-; GFX11-FAKE16-NEXT: ; kill: killed $sgpr42
-; GFX11-FAKE16-NEXT: ; implicit-def: $sgpr42
-; GFX11-FAKE16-NEXT: ; kill: killed $sgpr42
-; GFX11-FAKE16-NEXT: ; implicit-def: $sgpr42
-; GFX11-FAKE16-NEXT: ; kill: killed $sgpr42
-; GFX11-FAKE16-NEXT: ; implicit-def: $sgpr42
-; GFX11-FAKE16-NEXT: v_writelane_b32 v19, vcc_hi, 1
-; GFX11-FAKE16-NEXT: ; implicit-def: $vcc_lo
-; GFX11-FAKE16-NEXT: ; kill: killed $sgpr42
-; GFX11-FAKE16-NEXT: ; implicit-def: $sgpr42
-; GFX11-FAKE16-NEXT: ; kill: killed $sgpr42
-; GFX11-FAKE16-NEXT: ; implicit-def: $sgpr42
-; GFX11-FAKE16-NEXT: ; kill: killed $sgpr42
-; GFX11-FAKE16-NEXT: ; implicit-def: $sgpr42
-; GFX11-FAKE16-NEXT: ; kill: killed $sgpr42
-; GFX11-FAKE16-NEXT: ; implicit-def: $sgpr42
-; GFX11-FAKE16-NEXT: ; kill: killed $sgpr42
-; GFX11-FAKE16-NEXT: ; implicit-def: $sgpr42
-; GFX11-FAKE16-NEXT: v_writelane_b32 v19, vcc_lo, 2
-; GFX11-FAKE16-NEXT: ; kill: killed $sgpr42
-; GFX11-FAKE16-NEXT: ; implicit-def: $sgpr42
-; GFX11-FAKE16-NEXT: ; kill: killed $sgpr42
-; GFX11-FAKE16-NEXT: ; implicit-def: $sgpr42
-; GFX11-FAKE16-NEXT: ; kill: killed $sgpr42
-; GFX11-FAKE16-NEXT: ; implicit-def: $sgpr42
-; GFX11-FAKE16-NEXT: ; kill: killed $sgpr42
-; GFX11-FAKE16-NEXT: ; implicit-def: $sgpr42
-; GFX11-FAKE16-NEXT: ; kill: killed $sgpr42
-; GFX11-FAKE16-NEXT: ; implicit-def: $sgpr42
-; GFX11-FAKE16-NEXT: v_writelane_b32 v19, vcc_hi, 3
-; GFX11-FAKE16-NEXT: ; implicit-def: $vcc_lo
-; GFX11-FAKE16-NEXT: ; kill: killed $sgpr42
-; GFX11-FAKE16-NEXT: ; implicit-def: $sgpr42
-; GFX11-FAKE16-NEXT: ; kill: killed $sgpr42
-; GFX11-FAKE16-NEXT: ; implicit-def: $sgpr42
-; GFX11-FAKE16-NEXT: ; kill: killed $sgpr42
-; GFX11-FAKE16-NEXT: ; implicit-def: $sgpr42
-; GFX11-FAKE16-NEXT: ; kill: killed $sgpr42
-; GFX11-FAKE16-NEXT: ; implicit-def: $sgpr42
-; GFX11-FAKE16-NEXT: ; kill: killed $sgpr42
-; GFX11-FAKE16-NEXT: ; implicit-def: $sgpr42
-; GFX11-FAKE16-NEXT: v_writelane_b32 v19, vcc_lo, 4
-; GFX11-FAKE16-NEXT: ; kill: killed $sgpr42
-; GFX11-FAKE16-NEXT: ; implicit-def: $sgpr42
-; GFX11-FAKE16-NEXT: ; kill: killed $sgpr42
-; GFX11-FAKE16-NEXT: ; implicit-def: $sgpr42
-; GFX11-FAKE16-NEXT: ; kill: killed $sgpr42
-; GFX11-FAKE16-NEXT: ; implicit-def: $sgpr42
-; GFX11-FAKE16-NEXT: ; kill: killed $sgpr42
-; GFX11-FAKE16-NEXT: ; implicit-def: $sgpr42
-; GFX11-FAKE16-NEXT: ; kill: killed $sgpr42
-; GFX11-FAKE16-NEXT: ; implicit-def: $sgpr42
-; GFX11-FAKE16-NEXT: v_writelane_b32 v19, vcc_hi, 5
-; GFX11-FAKE16-NEXT: ; implicit-def: $vcc_lo
-; GFX11-FAKE16-NEXT: ; implicit-def: $sgpr45
-; GFX11-FAKE16-NEXT: ; implicit-def: $sgpr44
-; GFX11-FAKE16-NEXT: ; implicit-def: $sgpr30
-; GFX11-FAKE16-NEXT: ; implicit-def: $sgpr100
-; GFX11-FAKE16-NEXT: ; implicit-def: $sgpr99
-; GFX11-FAKE16-NEXT: ; implicit-def: $sgpr98
-; GFX11-FAKE16-NEXT: ; implicit-def: $sgpr43
-; GFX11-FAKE16-NEXT: ; implicit-def: $sgpr74
-; GFX11-FAKE16-NEXT: ; implicit-def: $sgpr94
-; GFX11-FAKE16-NEXT: ; implicit-def: $sgpr92
-; GFX11-FAKE16-NEXT: ; implicit-def: $sgpr90
-; GFX11-FAKE16-NEXT: ; implicit-def: $sgpr78
-; GFX11-FAKE16-NEXT: ; implicit-def: $sgpr62
-; GFX11-FAKE16-NEXT: ; kill: killed $sgpr42
-; GFX11-FAKE16-NEXT: ; implicit-def: $sgpr42
-; GFX11-FAKE16-NEXT: ; implicit-def: $sgpr97
-; GFX11-FAKE16-NEXT: ; implicit-def: $sgpr96
-; GFX11-FAKE16-NEXT: ; implicit-def: $sgpr73
-; GFX11-FAKE16-NEXT: ; implicit-def: $sgpr87
-; GFX11-FAKE16-NEXT: ; implicit-def: $sgpr72
-; GFX11-FAKE16-NEXT: ; implicit-def: $sgpr86
-; GFX11-FAKE16-NEXT: ; implicit-def: $sgpr85
-; GFX11-FAKE16-NEXT: ; implicit-def: $sgpr84
-; GFX11-FAKE16-NEXT: ; implicit-def: $sgpr83
-; GFX11-FAKE16-NEXT: ; implicit-def: $sgpr82
-; GFX11-FAKE16-NEXT: ; implicit-def: $sgpr81
-; GFX11-FAKE16-NEXT: ; implicit-def: $sgpr61
-; GFX11-FAKE16-NEXT: ; implicit-def: $sgpr80
-; GFX11-FAKE16-NEXT: ; implicit-def: $sgpr60
-; GFX11-FAKE16-NEXT: ; implicit-def: $sgpr71
-; GFX11-FAKE16-NEXT: ; implicit-def: $sgpr70
-; GFX11-FAKE16-NEXT: ; implicit-def: $sgpr58
-; GFX11-FAKE16-NEXT: ; implicit-def: $sgpr59
-; GFX11-FAKE16-NEXT: ; implicit-def: $sgpr68
-; GFX11-FAKE16-NEXT: ; implicit-def: $sgpr67
-; GFX11-FAKE16-NEXT: ; implicit-def: $sgpr66
-; GFX11-FAKE16-NEXT: ; implicit-def: $sgpr65
-; GFX11-FAKE16-NEXT: ; implicit-def: $sgpr64
-; GFX11-FAKE16-NEXT: ; implicit-def: $sgpr55
-; GFX11-FAKE16-NEXT: ; implicit-def: $sgpr54
-; GFX11-FAKE16-NEXT: ; implicit-def: $sgpr53
-; GFX11-FAKE16-NEXT: ; implicit-def: $sgpr52
-; GFX11-FAKE16-NEXT: ; implicit-def: $sgpr51
-; GFX11-FAKE16-NEXT: ; implicit-def: $sgpr50
-; GFX11-FAKE16-NEXT: ; implicit-def: $sgpr49
-; GFX11-FAKE16-NEXT: ; implicit-def: $sgpr48
-; GFX11-FAKE16-NEXT: ; implicit-def: $sgpr39
-; GFX11-FAKE16-NEXT: ; implicit-def: $sgpr38
-; GFX11-FAKE16-NEXT: ; implicit-def: $sgpr37
-; GFX11-FAKE16-NEXT: ; implicit-def: $sgpr36
-; GFX11-FAKE16-NEXT: ; implicit-def: $sgpr35
-; GFX11-FAKE16-NEXT: ; implicit-def: $sgpr56
-; GFX11-FAKE16-NEXT: ; implicit-def: $sgpr69
-; GFX11-FAKE16-NEXT: ; implicit-def: $sgpr34
-; GFX11-FAKE16-NEXT: ; implicit-def: $sgpr46
-; GFX11-FAKE16-NEXT: ; implicit-def: $sgpr47
-; GFX11-FAKE16-NEXT: ; implicit-def: $sgpr57
-; GFX11-FAKE16-NEXT: ; implicit-def: $sgpr103
-; GFX11-FAKE16-NEXT: ; implicit-def: $sgpr102
-; GFX11-FAKE16-NEXT: ; implicit-def: $sgpr104
-; GFX11-FAKE16-NEXT: ; implicit-def: $sgpr88
-; GFX11-FAKE16-NEXT: ; implicit-def: $sgpr76
-; GFX11-FAKE16-NEXT: v_writelane_b32 v19, vcc_lo, 6
-; GFX11-FAKE16-NEXT: v_writelane_b32 v19, vcc_hi, 7
-; GFX11-FAKE16-NEXT: ; implicit-def: $vcc_lo
-; GFX11-FAKE16-NEXT: v_writelane_b32 v19, vcc_lo, 8
-; GFX11-FAKE16-NEXT: v_writelane_b32 v19, vcc_hi, 9
-; GFX11-FAKE16-NEXT: ; implicit-def: $vcc_lo
-; GFX11-FAKE16-NEXT: v_writelane_b32 v19, vcc_lo, 10
-; GFX11-FAKE16-NEXT: v_writelane_b32 v19, vcc_hi, 11
-; GFX11-FAKE16-NEXT: ; implicit-def: $vcc_lo
-; GFX11-FAKE16-NEXT: v_writelane_b32 v19, vcc_lo, 12
-; GFX11-FAKE16-NEXT: v_writelane_b32 v19, vcc_hi, 13
-; GFX11-FAKE16-NEXT: ; implicit-def: $vcc_lo
-; GFX11-FAKE16-NEXT: v_writelane_b32 v19, vcc_lo, 14
-; GFX11-FAKE16-NEXT: v_writelane_b32 v19, vcc_hi, 15
-; GFX11-FAKE16-NEXT: .LBB13_3: ; %Flow
-; GFX11-FAKE16-NEXT: s_and_not1_b32 vcc_lo, exec_lo, s101
-; GFX11-FAKE16-NEXT: s_mov_b32 s101, s104
-; GFX11-FAKE16-NEXT: s_mov_b32 s104, s57
-; GFX11-FAKE16-NEXT: s_mov_b32 s57, s69
-; GFX11-FAKE16-NEXT: s_mov_b32 s69, s42
-; GFX11-FAKE16-NEXT: s_cbranch_vccnz .LBB13_5
-; GFX11-FAKE16-NEXT: ; %bb.4: ; %cmp.true
-; GFX11-FAKE16-NEXT: s_add_i32 s25, s25, 3
-; GFX11-FAKE16-NEXT: s_add_i32 s24, s24, 3
-; GFX11-FAKE16-NEXT: s_lshr_b32 s42, s25, 8
-; GFX11-FAKE16-NEXT: s_add_i32 s23, s23, 3
-; GFX11-FAKE16-NEXT: v_writelane_b32 v19, s42, 16
-; GFX11-FAKE16-NEXT: s_lshr_b32 s42, s24, 16
-; GFX11-FAKE16-NEXT: s_add_i32 s22, s22, 3
-; GFX11-FAKE16-NEXT: s_add_i32 s21, s21, 3
-; GFX11-FAKE16-NEXT: s_add_i32 s20, s20, 3
-; GFX11-FAKE16-NEXT: v_writelane_b32 v19, s42, 17
-; GFX11-FAKE16-NEXT: s_lshr_b32 s42, s24, 8
-; GFX11-FAKE16-NEXT: s_add_i32 s19, s19, 3
+; GFX11-FAKE16-NEXT: v_writelane_b32 v18, s42, 0
+; GFX11-FAKE16-NEXT: s_lshr_b32 s42, s14, 8
+; GFX11-FAKE16-NEXT: v_writelane_b32 v19, s46, 2
+; GFX11-FAKE16-NEXT: v_writelane_b32 v18, s42, 1
+; GFX11-FAKE16-NEXT: s_lshr_b32 s42, s41, 24
+; GFX11-FAKE16-NEXT: v_writelane_b32 v19, s47, 3
+; GFX11-FAKE16-NEXT: s_lshr_b64 s[46:47], s[6:7], 24
+; GFX11-FAKE16-NEXT: v_writelane_b32 v18, s42, 2
+; GFX11-FAKE16-NEXT: s_lshr_b32 s42, s41, 16
+; GFX11-FAKE16-NEXT: v_writelane_b32 v19, s46, 0
+; GFX11-FAKE16-NEXT: v_writelane_b32 v18, s42, 3
+; GFX11-FAKE16-NEXT: s_lshr_b32 s42, s41, 8
+; GFX11-FAKE16-NEXT: v_writelane_b32 v19, s47, 1
+; GFX11-FAKE16-NEXT: s_lshr_b64 s[46:47], s[8:9], 24
+; GFX11-FAKE16-NEXT: v_writelane_b32 v18, s42, 4
+; GFX11-FAKE16-NEXT: s_lshr_b32 s42, s40, 16
+; GFX11-FAKE16-NEXT: s_delay_alu instid0(SALU_CYCLE_1) | instskip(SKIP_1) | instid1(SALU_CYCLE_1)
+; GFX11-FAKE16-NEXT: v_writelane_b32 v18, s42, 5
+; GFX11-FAKE16-NEXT: s_lshr_b32 s42, s40, 8
+; GFX11-FAKE16-NEXT: v_writelane_b32 v18, s42, 6
+; GFX11-FAKE16-NEXT: s_lshr_b32 s42, s29, 24
+; GFX11-FAKE16-NEXT: s_delay_alu instid0(SALU_CYCLE_1) | instskip(SKIP_1) | instid1(SALU_CYCLE_1)
+; GFX11-FAKE16-NEXT: v_writelane_b32 v18, s42, 7
+; GFX11-FAKE16-NEXT: s_lshr_b32 s42, s29, 16
+; GFX11-FAKE16-NEXT: v_writelane_b32 v18, s42, 8
+; GFX11-FAKE16-NEXT: s_lshr_b32 s42, s2, 16
+; GFX11-FAKE16-NEXT: s_and_not1_b32 vcc_lo, exec_lo, s44
+; GFX11-FAKE16-NEXT: s_cbranch_vccnz .LBB13_3
+; GFX11-FAKE16-NEXT: .LBB13_2: ; %cmp.true
; GFX11-FAKE16-NEXT: s_add_i32 s5, s5, 3
; GFX11-FAKE16-NEXT: s_add_i32 s4, s4, 3
-; GFX11-FAKE16-NEXT: v_writelane_b32 v19, s42, 18
-; GFX11-FAKE16-NEXT: s_lshr_b32 s42, s23, 24
-; GFX11-FAKE16-NEXT: s_lshr_b64 s[62:63], s[4:5], 24
+; GFX11-FAKE16-NEXT: s_lshr_b32 s42, s5, 24
; GFX11-FAKE16-NEXT: s_add_i32 s7, s7, 3
+; GFX11-FAKE16-NEXT: v_writelane_b32 v19, s42, 4
+; GFX11-FAKE16-NEXT: s_lshr_b32 s42, s5, 16
; GFX11-FAKE16-NEXT: s_add_i32 s6, s6, 3
-; GFX11-FAKE16-NEXT: v_writelane_b32 v19, s42, 19
-; GFX11-FAKE16-NEXT: s_lshr_b32 s42, s23, 16
; GFX11-FAKE16-NEXT: s_add_i32 s9, s9, 3
; GFX11-FAKE16-NEXT: s_add_i32 s8, s8, 3
+; GFX11-FAKE16-NEXT: v_writelane_b32 v19, s42, 5
+; GFX11-FAKE16-NEXT: s_lshr_b32 s42, s5, 8
; GFX11-FAKE16-NEXT: s_add_i32 s11, s11, 3
-; GFX11-FAKE16-NEXT: v_writelane_b32 v19, s42, 20
-; GFX11-FAKE16-NEXT: s_lshr_b32 s42, s23, 8
; GFX11-FAKE16-NEXT: s_add_i32 s10, s10, 3
-; GFX11-FAKE16-NEXT: s_add_i32 s18, s18, 3
; GFX11-FAKE16-NEXT: s_add_i32 s13, s13, 3
-; GFX11-FAKE16-NEXT: v_writelane_b32 v19, s42, 21
-; GFX11-FAKE16-NEXT: s_lshr_b32 s42, s22, 16
+; GFX11-FAKE16-NEXT: v_writelane_b32 v19, s42, 6
+; GFX11-FAKE16-NEXT: s_lshr_b32 s42, s4, 16
; GFX11-FAKE16-NEXT: s_add_i32 s12, s12, 3
-; GFX11-FAKE16-NEXT: s_add_i32 s17, s17, 3
; GFX11-FAKE16-NEXT: s_add_i32 s15, s15, 3
-; GFX11-FAKE16-NEXT: v_writelane_b32 v19, s42, 22
-; GFX11-FAKE16-NEXT: s_lshr_b32 s42, s22, 8
; GFX11-FAKE16-NEXT: s_add_i32 s14, s14, 3
-; GFX11-FAKE16-NEXT: s_add_i32 s16, s16, 3
+; GFX11-FAKE16-NEXT: v_writelane_b32 v19, s42, 7
+; GFX11-FAKE16-NEXT: s_lshr_b32 s42, s4, 8
; GFX11-FAKE16-NEXT: s_add_i32 s41, s41, 3
-; GFX11-FAKE16-NEXT: v_writelane_b32 v19, s42, 23
-; GFX11-FAKE16-NEXT: s_lshr_b32 s42, s21, 24
; GFX11-FAKE16-NEXT: s_add_i32 s40, s40, 3
-; GFX11-FAKE16-NEXT: s_add_i32 s3, s3, 3
+; GFX11-FAKE16-NEXT: s_lshr_b64 s[46:47], s[4:5], 24
+; GFX11-FAKE16-NEXT: v_writelane_b32 v19, s42, 8
+; GFX11-FAKE16-NEXT: s_lshr_b32 s42, s7, 24
; GFX11-FAKE16-NEXT: s_add_i32 s29, s29, 3
-; GFX11-FAKE16-NEXT: v_writelane_b32 v19, s42, 24
-; GFX11-FAKE16-NEXT: s_lshr_b32 s42, s21, 16
-; GFX11-FAKE16-NEXT: s_add_i32 s28, s28, 3
; GFX11-FAKE16-NEXT: s_add_i32 s1, s1, 3
; GFX11-FAKE16-NEXT: s_add_i32 s0, s0, 3
-; GFX11-FAKE16-NEXT: v_writelane_b32 v19, s42, 25
-; GFX11-FAKE16-NEXT: s_lshr_b32 s42, s21, 8
+; GFX11-FAKE16-NEXT: v_writelane_b32 v19, s42, 9
+; GFX11-FAKE16-NEXT: s_lshr_b32 s42, s7, 16
+; GFX11-FAKE16-NEXT: s_add_i32 s3, s3, 3
; GFX11-FAKE16-NEXT: s_add_i32 s2, s2, 3
+; GFX11-FAKE16-NEXT: s_add_i32 s17, s17, 3
+; GFX11-FAKE16-NEXT: v_writelane_b32 v19, s42, 10
+; GFX11-FAKE16-NEXT: s_lshr_b32 s42, s7, 8
+; GFX11-FAKE16-NEXT: s_add_i32 s16, s16, 3
+; GFX11-FAKE16-NEXT: s_add_i32 s19, s19, 3
+; GFX11-FAKE16-NEXT: s_add_i32 s18, s18, 3
+; GFX11-FAKE16-NEXT: v_writelane_b32 v19, s42, 11
+; GFX11-FAKE16-NEXT: s_lshr_b32 s42, s6, 16
+; GFX11-FAKE16-NEXT: s_add_i32 s21, s21, 3
+; GFX11-FAKE16-NEXT: s_add_i32 s20, s20, 3
+; GFX11-FAKE16-NEXT: s_add_i32 s23, s23, 3
+; GFX11-FAKE16-NEXT: v_writelane_b32 v19, s42, 12
+; GFX11-FAKE16-NEXT: s_lshr_b32 s42, s6, 8
+; GFX11-FAKE16-NEXT: s_add_i32 s22, s22, 3
+; GFX11-FAKE16-NEXT: s_add_i32 s25, s25, 3
+; GFX11-FAKE16-NEXT: s_add_i32 s24, s24, 3
+; GFX11-FAKE16-NEXT: v_writelane_b32 v19, s42, 13
+; GFX11-FAKE16-NEXT: s_lshr_b32 s42, s9, 24
; GFX11-FAKE16-NEXT: s_add_i32 s27, s27, 3
; GFX11-FAKE16-NEXT: s_add_i32 s26, s26, 3
+; GFX11-FAKE16-NEXT: s_add_i32 s28, s28, 3
+; GFX11-FAKE16-NEXT: v_writelane_b32 v19, s42, 14
+; GFX11-FAKE16-NEXT: s_lshr_b32 s42, s9, 16
+; GFX11-FAKE16-NEXT: s_lshr_b32 s102, s29, 8
+; GFX11-FAKE16-NEXT: s_lshr_b32 s103, s28, 16
+; GFX11-FAKE16-NEXT: s_lshr_b32 s104, s28, 8
+; GFX11-FAKE16-NEXT: v_writelane_b32 v19, s42, 15
+; GFX11-FAKE16-NEXT: s_lshr_b32 s42, s9, 8
+; GFX11-FAKE16-NEXT: s_lshr_b32 vcc_hi, s27, 24
+; GFX11-FAKE16-NEXT: s_lshr_b32 s34, s27, 16
+; GFX11-FAKE16-NEXT: s_lshr_b32 s35, s27, 8
+; GFX11-FAKE16-NEXT: v_writelane_b32 v19, s42, 16
+; GFX11-FAKE16-NEXT: s_lshr_b32 s42, s8, 16
+; GFX11-FAKE16-NEXT: s_lshr_b32 s36, s26, 16
+; GFX11-FAKE16-NEXT: s_lshr_b32 s37, s26, 8
+; GFX11-FAKE16-NEXT: s_lshr_b32 s38, s25, 24
+; GFX11-FAKE16-NEXT: v_writelane_b32 v19, s42, 17
+; GFX11-FAKE16-NEXT: s_lshr_b32 s42, s8, 8
+; GFX11-FAKE16-NEXT: s_lshr_b32 s39, s25, 16
+; GFX11-FAKE16-NEXT: s_lshr_b32 s48, s25, 8
+; GFX11-FAKE16-NEXT: s_lshr_b32 s49, s24, 16
+; GFX11-FAKE16-NEXT: v_writelane_b32 v19, s42, 18
+; GFX11-FAKE16-NEXT: s_lshr_b32 s42, s11, 24
+; GFX11-FAKE16-NEXT: s_lshr_b32 s50, s24, 8
+; GFX11-FAKE16-NEXT: s_lshr_b32 s51, s23, 24
+; GFX11-FAKE16-NEXT: s_lshr_b32 s52, s23, 16
+; GFX11-FAKE16-NEXT: v_writelane_b32 v19, s42, 19
+; GFX11-FAKE16-NEXT: s_lshr_b32 s42, s11, 16
+; GFX11-FAKE16-NEXT: s_lshr_b32 s53, s23, 8
+; GFX11-FAKE16-NEXT: s_lshr_b32 s54, s22, 16
+; GFX11-FAKE16-NEXT: s_lshr_b32 s55, s22, 8
+; GFX11-FAKE16-NEXT: v_writelane_b32 v19, s42, 20
+; GFX11-FAKE16-NEXT: s_lshr_b32 s42, s11, 8
+; GFX11-FAKE16-NEXT: s_lshr_b32 s64, s21, 24
+; GFX11-FAKE16-NEXT: s_lshr_b32 s65, s21, 16
+; GFX11-FAKE16-NEXT: s_lshr_b32 s66, s21, 8
+; GFX11-FAKE16-NEXT: v_writelane_b32 v19, s42, 21
+; GFX11-FAKE16-NEXT: s_lshr_b32 s42, s10, 16
+; GFX11-FAKE16-NEXT: s_lshr_b32 s67, s20, 16
+; GFX11-FAKE16-NEXT: s_lshr_b32 s68, s20, 8
+; GFX11-FAKE16-NEXT: s_lshr_b32 s69, s19, 24
+; GFX11-FAKE16-NEXT: v_writelane_b32 v19, s42, 22
+; GFX11-FAKE16-NEXT: s_lshr_b32 s42, s10, 8
+; GFX11-FAKE16-NEXT: s_lshr_b32 s70, s19, 16
+; GFX11-FAKE16-NEXT: s_lshr_b32 s71, s19, 8
+; GFX11-FAKE16-NEXT: s_lshr_b32 s80, s18, 16
+; GFX11-FAKE16-NEXT: v_writelane_b32 v19, s42, 23
+; GFX11-FAKE16-NEXT: s_lshr_b32 s42, s13, 24
+; GFX11-FAKE16-NEXT: s_lshr_b32 s81, s18, 8
+; GFX11-FAKE16-NEXT: s_lshr_b32 s82, s17, 24
+; GFX11-FAKE16-NEXT: s_lshr_b32 s83, s17, 16
+; GFX11-FAKE16-NEXT: v_writelane_b32 v19, s42, 24
+; GFX11-FAKE16-NEXT: s_lshr_b32 s42, s13, 16
+; GFX11-FAKE16-NEXT: s_lshr_b32 s84, s17, 8
+; GFX11-FAKE16-NEXT: s_lshr_b32 s85, s16, 16
+; GFX11-FAKE16-NEXT: s_lshr_b32 s86, s16, 8
+; GFX11-FAKE16-NEXT: v_writelane_b32 v19, s42, 25
+; GFX11-FAKE16-NEXT: s_lshr_b32 s42, s13, 8
+; GFX11-FAKE16-NEXT: s_lshr_b32 s87, s3, 24
+; GFX11-FAKE16-NEXT: s_lshr_b32 s96, s3, 16
+; GFX11-FAKE16-NEXT: s_lshr_b32 s97, s3, 8
; GFX11-FAKE16-NEXT: v_writelane_b32 v19, s42, 26
-; GFX11-FAKE16-NEXT: s_lshr_b32 s42, s20, 16
-; GFX11-FAKE16-NEXT: s_lshr_b32 s101, s5, 24
-; GFX11-FAKE16-NEXT: s_lshr_b32 s102, s5, 16
-; GFX11-FAKE16-NEXT: s_lshr_b32 s103, s5, 8
+; GFX11-FAKE16-NEXT: s_lshr_b32 s42, s12, 16
+; GFX11-FAKE16-NEXT: s_lshr_b32 s43, s2, 8
+; GFX11-FAKE16-NEXT: s_lshr_b32 s98, s1, 24
+; GFX11-FAKE16-NEXT: s_lshr_b32 s99, s1, 16
; GFX11-FAKE16-NEXT: v_writelane_b32 v19, s42, 27
-; GFX11-FAKE16-NEXT: s_lshr_b32 s42, s20, 8
-; GFX11-FAKE16-NEXT: s_lshr_b32 s104, s4, 16
-; GFX11-FAKE16-NEXT: s_lshr_b32 s47, s4, 8
-; GFX11-FAKE16-NEXT: s_lshr_b32 s46, s7, 24
+; GFX11-FAKE16-NEXT: s_lshr_b32 s42, s12, 8
+; GFX11-FAKE16-NEXT: s_lshr_b32 s100, s1, 8
+; GFX11-FAKE16-NEXT: s_lshr_b32 s45, s0, 16
+; GFX11-FAKE16-NEXT: s_lshr_b32 s101, s0, 8
; GFX11-FAKE16-NEXT: v_writelane_b32 v19, s42, 28
-; GFX11-FAKE16-NEXT: s_lshr_b32 s42, s19, 24
-; GFX11-FAKE16-NEXT: s_lshr_b32 vcc_hi, s7, 16
-; GFX11-FAKE16-NEXT: s_lshr_b32 s34, s7, 8
-; GFX11-FAKE16-NEXT: s_lshr_b32 s57, s6, 16
+; GFX11-FAKE16-NEXT: s_lshr_b32 s42, s15, 24
+; GFX11-FAKE16-NEXT: s_lshr_b64 s[56:57], s[10:11], 24
+; GFX11-FAKE16-NEXT: s_lshr_b64 s[58:59], s[12:13], 24
+; GFX11-FAKE16-NEXT: s_lshr_b64 s[60:61], s[14:15], 24
; GFX11-FAKE16-NEXT: v_writelane_b32 v19, s42, 29
-; GFX11-FAKE16-NEXT: s_lshr_b32 s42, s19, 16
-; GFX11-FAKE16-NEXT: s_lshr_b32 s56, s6, 8
-; GFX11-FAKE16-NEXT: s_lshr_b32 s35, s9, 24
-; GFX11-FAKE16-NEXT: s_lshr_b32 s36, s9, 16
+; GFX11-FAKE16-NEXT: s_lshr_b32 s42, s15, 16
+; GFX11-FAKE16-NEXT: s_lshr_b64 s[62:63], s[40:41], 24
+; GFX11-FAKE16-NEXT: s_lshr_b64 s[72:73], s[28:29], 24
+; GFX11-FAKE16-NEXT: s_lshr_b64 s[76:77], s[26:27], 24
; GFX11-FAKE16-NEXT: v_writelane_b32 v19, s42, 30
-; GFX11-FAKE16-NEXT: s_lshr_b32 s42, s19, 8
-; GFX11-FAKE16-NEXT: s_lshr_b32 s37, s9, 8
-; GFX11-FAKE16-NEXT: s_lshr_b32 s38, s8, 16
-; GFX11-FAKE16-NEXT: s_lshr_b32 s39, s8, 8
+; GFX11-FAKE16-NEXT: s_lshr_b32 s42, s15, 8
+; GFX11-FAKE16-NEXT: s_lshr_b64 s[88:89], s[24:25], 24
+; GFX11-FAKE16-NEXT: s_lshr_b64 s[74:75], s[22:23], 24
+; GFX11-FAKE16-NEXT: s_lshr_b64 s[78:79], s[20:21], 24
; GFX11-FAKE16-NEXT: v_writelane_b32 v19, s42, 31
-; GFX11-FAKE16-NEXT: s_lshr_b32 s42, s18, 16
-; GFX11-FAKE16-NEXT: s_lshr_b32 s48, s11, 24
+; GFX11-FAKE16-NEXT: s_lshr_b32 s42, s14, 16
+; GFX11-FAKE16-NEXT: s_lshr_b64 s[90:91], s[18:19], 24
; GFX11-FAKE16-NEXT: v_writelane_b32 v18, s42, 0
-; GFX11-FAKE16-NEXT: s_lshr_b32 s42, s18, 8
-; GFX11-FAKE16-NEXT: v_writelane_b32 v19, s62, 14
-; GFX11-FAKE16-NEXT: s_lshr_b32 s49, s11, 16
-; GFX11-FAKE16-NEXT: s_lshr_b32 s50, s11, 8
+; GFX11-FAKE16-NEXT: s_lshr_b32 s42, s14, 8
+; GFX11-FAKE16-NEXT: v_writelane_b32 v19, s46, 2
+; GFX11-FAKE16-NEXT: s_lshr_b64 s[92:93], s[16:17], 24
+; GFX11-FAKE16-NEXT: s_lshr_b64 s[94:95], s[2:3], 24
; GFX11-FAKE16-NEXT: v_writelane_b32 v18, s42, 1
-; GFX11-FAKE16-NEXT: s_lshr_b32 s42, s17, 24
-; GFX11-FAKE16-NEXT: v_writelane_b32 v19, s63, 15
-; GFX11-FAKE16-NEXT: s_lshr_b64 s[62:63], s[6:7], 24
-; GFX11-FAKE16-NEXT: s_lshr_b32 s51, s10, 16
+; GFX11-FAKE16-NEXT: s_lshr_b32 s42, s41, 24
+; GFX11-FAKE16-NEXT: v_writelane_b32 v19, s47, 3
+; GFX11-FAKE16-NEXT: s_lshr_b64 s[46:47], s[6:7], 24
+; GFX11-FAKE16-NEXT: s_lshr_b64 s[30:31], s[0:1], 24
; GFX11-FAKE16-NEXT: v_writelane_b32 v18, s42, 2
-; GFX11-FAKE16-NEXT: s_lshr_b32 s42, s17, 16
-; GFX11-FAKE16-NEXT: v_writelane_b32 v19, s62, 12
-; GFX11-FAKE16-NEXT: s_lshr_b32 s52, s10, 8
-; GFX11-FAKE16-NEXT: s_lshr_b32 s53, s13, 24
+; GFX11-FAKE16-NEXT: s_lshr_b32 s42, s41, 16
+; GFX11-FAKE16-NEXT: v_writelane_b32 v19, s46, 0
; GFX11-FAKE16-NEXT: v_writelane_b32 v18, s42, 3
-; GFX11-FAKE16-NEXT: s_lshr_b32 s42, s17, 8
-; GFX11-FAKE16-NEXT: v_writelane_b32 v19, s63, 13
-; GFX11-FAKE16-NEXT: s_lshr_b64 s[62:63], s[8:9], 24
-; GFX11-FAKE16-NEXT: s_lshr_b32 s54, s13, 16
+; GFX11-FAKE16-NEXT: s_lshr_b32 s42, s41, 8
+; GFX11-FAKE16-NEXT: v_writelane_b32 v19, s47, 1
+; GFX11-FAKE16-NEXT: s_lshr_b64 s[46:47], s[8:9], 24
; GFX11-FAKE16-NEXT: v_writelane_b32 v18, s42, 4
-; GFX11-FAKE16-NEXT: s_lshr_b32 s42, s16, 16
-; GFX11-FAKE16-NEXT: v_writelane_b32 v19, s62, 10
-; GFX11-FAKE16-NEXT: s_lshr_b32 s55, s13, 8
-; GFX11-FAKE16-NEXT: s_lshr_b32 s64, s12, 16
+; GFX11-FAKE16-NEXT: s_lshr_b32 s42, s40, 16
+; GFX11-FAKE16-NEXT: s_delay_alu instid0(SALU_CYCLE_1) | instskip(SKIP_1) | instid1(SALU_CYCLE_1)
; GFX11-FAKE16-NEXT: v_writelane_b32 v18, s42, 5
-; GFX11-FAKE16-NEXT: s_lshr_b32 s42, s16, 8
-; GFX11-FAKE16-NEXT: v_writelane_b32 v19, s63, 11
-; GFX11-FAKE16-NEXT: s_lshr_b64 s[62:63], s[10:11], 24
-; GFX11-FAKE16-NEXT: s_lshr_b32 s65, s12, 8
+; GFX11-FAKE16-NEXT: s_lshr_b32 s42, s40, 8
; GFX11-FAKE16-NEXT: v_writelane_b32 v18, s42, 6
-; GFX11-FAKE16-NEXT: s_lshr_b32 s42, s3, 24
-; GFX11-FAKE16-NEXT: v_writelane_b32 v19, s62, 8
-; GFX11-FAKE16-NEXT: s_lshr_b32 s66, s15, 24
-; GFX11-FAKE16-NEXT: s_lshr_b32 s67, s15, 16
+; GFX11-FAKE16-NEXT: s_lshr_b32 s42, s29, 24
+; GFX11-FAKE16-NEXT: s_delay_alu instid0(SALU_CYCLE_1) | instskip(SKIP_1) | instid1(SALU_CYCLE_1)
; GFX11-FAKE16-NEXT: v_writelane_b32 v18, s42, 7
-; GFX11-FAKE16-NEXT: s_lshr_b32 s42, s3, 16
-; GFX11-FAKE16-NEXT: v_writelane_b32 v19, s63, 9
-; GFX11-FAKE16-NEXT: s_lshr_b64 s[62:63], s[12:13], 24
-; GFX11-FAKE16-NEXT: s_lshr_b32 s68, s15, 8
+; GFX11-FAKE16-NEXT: s_lshr_b32 s42, s29, 16
; GFX11-FAKE16-NEXT: v_writelane_b32 v18, s42, 8
-; GFX11-FAKE16-NEXT: s_lshr_b32 s59, s14, 16
-; GFX11-FAKE16-NEXT: v_writelane_b32 v19, s62, 6
-; GFX11-FAKE16-NEXT: s_lshr_b32 s58, s14, 8
-; GFX11-FAKE16-NEXT: s_lshr_b32 s70, s41, 24
-; GFX11-FAKE16-NEXT: s_lshr_b32 s71, s41, 16
-; GFX11-FAKE16-NEXT: s_lshr_b32 s60, s41, 8
-; GFX11-FAKE16-NEXT: v_writelane_b32 v19, s63, 7
-; GFX11-FAKE16-NEXT: s_lshr_b64 s[62:63], s[14:15], 24
-; GFX11-FAKE16-NEXT: s_lshr_b32 s80, s40, 16
-; GFX11-FAKE16-NEXT: s_lshr_b32 s61, s40, 8
-; GFX11-FAKE16-NEXT: s_lshr_b32 s81, s29, 24
-; GFX11-FAKE16-NEXT: v_writelane_b32 v19, s62, 4
-; GFX11-FAKE16-NEXT: s_lshr_b32 s82, s29, 16
-; GFX11-FAKE16-NEXT: s_lshr_b32 s83, s29, 8
-; GFX11-FAKE16-NEXT: s_lshr_b32 s84, s28, 16
-; GFX11-FAKE16-NEXT: s_lshr_b32 s85, s28, 8
-; GFX11-FAKE16-NEXT: v_writelane_b32 v19, s63, 5
-; GFX11-FAKE16-NEXT: s_lshr_b64 s[62:63], s[40:41], 24
-; GFX11-FAKE16-NEXT: s_lshr_b32 s86, s27, 24
-; GFX11-FAKE16-NEXT: s_lshr_b32 s72, s27, 16
-; GFX11-FAKE16-NEXT: s_lshr_b32 s87, s27, 8
-; GFX11-FAKE16-NEXT: v_writelane_b32 v19, s62, 2
-; GFX11-FAKE16-NEXT: s_lshr_b32 s73, s26, 16
-; GFX11-FAKE16-NEXT: s_lshr_b32 s96, s26, 8
-; GFX11-FAKE16-NEXT: s_lshr_b32 s97, s25, 24
-; GFX11-FAKE16-NEXT: s_lshr_b32 s69, s25, 16
-; GFX11-FAKE16-NEXT: v_writelane_b32 v19, s63, 3
-; GFX11-FAKE16-NEXT: s_lshr_b64 s[62:63], s[28:29], 24
-; GFX11-FAKE16-NEXT: s_lshr_b32 s42, s3, 8
-; GFX11-FAKE16-NEXT: s_lshr_b32 s74, s2, 16
-; GFX11-FAKE16-NEXT: s_lshr_b32 s43, s2, 8
-; GFX11-FAKE16-NEXT: v_writelane_b32 v19, s62, 0
-; GFX11-FAKE16-NEXT: s_lshr_b32 s98, s1, 24
-; GFX11-FAKE16-NEXT: s_lshr_b32 s99, s1, 16
-; GFX11-FAKE16-NEXT: s_lshr_b32 s100, s1, 8
-; GFX11-FAKE16-NEXT: s_lshr_b32 s44, s0, 16
-; GFX11-FAKE16-NEXT: s_lshr_b32 s45, s0, 8
-; GFX11-FAKE16-NEXT: v_writelane_b32 v19, s63, 1
-; GFX11-FAKE16-NEXT: s_lshr_b64 s[76:77], s[26:27], 24
-; GFX11-FAKE16-NEXT: s_lshr_b64 s[88:89], s[24:25], 24
-; GFX11-FAKE16-NEXT: s_lshr_b64 s[62:63], s[22:23], 24
-; GFX11-FAKE16-NEXT: s_lshr_b64 s[78:79], s[20:21], 24
-; GFX11-FAKE16-NEXT: s_lshr_b64 s[90:91], s[18:19], 24
-; GFX11-FAKE16-NEXT: s_lshr_b64 s[92:93], s[16:17], 24
-; GFX11-FAKE16-NEXT: s_lshr_b64 s[94:95], s[2:3], 24
-; GFX11-FAKE16-NEXT: s_lshr_b64 s[30:31], s[0:1], 24
-; GFX11-FAKE16-NEXT: v_writelane_b32 v18, s42, 9
-; GFX11-FAKE16-NEXT: .LBB13_5: ; %end
+; GFX11-FAKE16-NEXT: s_lshr_b32 s42, s2, 16
+; GFX11-FAKE16-NEXT: .LBB13_3: ; %end
+; GFX11-FAKE16-NEXT: s_lshl_b32 s44, s101, 8
+; GFX11-FAKE16-NEXT: s_and_b32 s0, s0, 0xff
+; GFX11-FAKE16-NEXT: s_and_b32 s45, s45, 0xff
+; GFX11-FAKE16-NEXT: s_or_b32 s0, s0, s44
+; GFX11-FAKE16-NEXT: s_lshl_b32 s44, s30, 8
; GFX11-FAKE16-NEXT: s_lshl_b32 s43, s43, 8
+; GFX11-FAKE16-NEXT: s_or_b32 s44, s45, s44
; GFX11-FAKE16-NEXT: s_and_b32 s2, s2, 0xff
-; GFX11-FAKE16-NEXT: s_and_b32 s42, s74, 0xff
-; GFX11-FAKE16-NEXT: s_or_b32 s2, s2, s43
-; GFX11-FAKE16-NEXT: s_lshl_b32 s43, s94, 8
-; GFX11-FAKE16-NEXT: s_and_b32 s2, s2, 0xffff
-; GFX11-FAKE16-NEXT: s_or_b32 s42, s42, s43
-; GFX11-FAKE16-NEXT: s_lshl_b32 s45, s45, 8
-; GFX11-FAKE16-NEXT: s_lshl_b32 s42, s42, 16
-; GFX11-FAKE16-NEXT: s_and_b32 s0, s0, 0xff
-; GFX11-FAKE16-NEXT: s_or_b32 s2, s2, s42
-; GFX11-FAKE16-NEXT: v_readlane_b32 s42, v18, 9
-; GFX11-FAKE16-NEXT: s_or_b32 s0, s0, s45
-; GFX11-FAKE16-NEXT: s_lshl_b32 s45, s30, 8
-; GFX11-FAKE16-NEXT: s_and_b32 s44, s44, 0xff
-; GFX11-FAKE16-NEXT: s_and_b32 s3, s3, 0xff
-; GFX11-FAKE16-NEXT: s_or_b32 s44, s44, s45
-; GFX11-FAKE16-NEXT: s_lshl_b32 s42, s42, 8
; GFX11-FAKE16-NEXT: s_and_b32 s0, s0, 0xffff
; GFX11-FAKE16-NEXT: s_lshl_b32 s44, s44, 16
-; GFX11-FAKE16-NEXT: s_or_b32 s3, s3, s42
-; GFX11-FAKE16-NEXT: v_readlane_b32 s42, v18, 8
-; GFX11-FAKE16-NEXT: v_readlane_b32 s43, v18, 7
+; GFX11-FAKE16-NEXT: s_or_b32 s2, s2, s43
+; GFX11-FAKE16-NEXT: s_lshl_b32 s43, s94, 8
+; GFX11-FAKE16-NEXT: s_and_b32 s42, s42, 0xff
; GFX11-FAKE16-NEXT: s_or_b32 s0, s0, s44
; GFX11-FAKE16-NEXT: s_and_b32 s1, s1, 0xff
; GFX11-FAKE16-NEXT: s_lshl_b32 s44, s100, 8
-; GFX11-FAKE16-NEXT: s_lshl_b32 s45, s98, 8
+; GFX11-FAKE16-NEXT: s_or_b32 s42, s42, s43
; GFX11-FAKE16-NEXT: s_or_b32 s1, s1, s44
; GFX11-FAKE16-NEXT: s_and_b32 s44, s99, 0xff
-; GFX11-FAKE16-NEXT: s_and_b32 s42, s42, 0xff
+; GFX11-FAKE16-NEXT: s_lshl_b32 s45, s98, 8
+; GFX11-FAKE16-NEXT: s_and_b32 s2, s2, 0xffff
+; GFX11-FAKE16-NEXT: s_lshl_b32 s42, s42, 16
; GFX11-FAKE16-NEXT: s_or_b32 s44, s44, s45
-; GFX11-FAKE16-NEXT: s_lshl_b32 s43, s43, 8
+; GFX11-FAKE16-NEXT: s_or_b32 s2, s2, s42
+; GFX11-FAKE16-NEXT: s_and_b32 s3, s3, 0xff
+; GFX11-FAKE16-NEXT: s_lshl_b32 s42, s97, 8
; GFX11-FAKE16-NEXT: s_and_b32 s1, s1, 0xffff
; GFX11-FAKE16-NEXT: s_lshl_b32 s44, s44, 16
-; GFX11-FAKE16-NEXT: s_or_b32 s42, s42, s43
+; GFX11-FAKE16-NEXT: s_or_b32 s3, s3, s42
+; GFX11-FAKE16-NEXT: s_and_b32 s42, s96, 0xff
+; GFX11-FAKE16-NEXT: s_lshl_b32 s43, s87, 8
; GFX11-FAKE16-NEXT: s_or_b32 s1, s1, s44
+; GFX11-FAKE16-NEXT: s_or_b32 s42, s42, s43
; GFX11-FAKE16-NEXT: s_and_b32 s3, s3, 0xffff
; GFX11-FAKE16-NEXT: s_lshl_b32 s42, s42, 16
; GFX11-FAKE16-NEXT: v_dual_mov_b32 v1, s0 :: v_dual_mov_b32 v2, s1
-; GFX11-FAKE16-NEXT: v_readlane_b32 s0, v18, 6
+; GFX11-FAKE16-NEXT: s_lshl_b32 s0, s86, 8
+; GFX11-FAKE16-NEXT: s_and_b32 s1, s16, 0xff
; GFX11-FAKE16-NEXT: s_or_b32 s3, s3, s42
; GFX11-FAKE16-NEXT: s_delay_alu instid0(SALU_CYCLE_1)
; GFX11-FAKE16-NEXT: v_dual_mov_b32 v3, s2 :: v_dual_mov_b32 v4, s3
-; GFX11-FAKE16-NEXT: v_readlane_b32 s2, v18, 5
-; GFX11-FAKE16-NEXT: s_lshl_b32 s0, s0, 8
-; GFX11-FAKE16-NEXT: s_and_b32 s1, s16, 0xff
-; GFX11-FAKE16-NEXT: v_readlane_b32 s3, v18, 2
; GFX11-FAKE16-NEXT: s_or_b32 s0, s1, s0
; GFX11-FAKE16-NEXT: s_lshl_b32 s1, s92, 8
-; GFX11-FAKE16-NEXT: s_and_b32 s2, s2, 0xff
+; GFX11-FAKE16-NEXT: s_and_b32 s2, s85, 0xff
; GFX11-FAKE16-NEXT: s_and_b32 s0, s0, 0xffff
; GFX11-FAKE16-NEXT: s_or_b32 s1, s2, s1
-; GFX11-FAKE16-NEXT: v_readlane_b32 s2, v18, 4
+; GFX11-FAKE16-NEXT: s_lshl_b32 s2, s84, 8
; GFX11-FAKE16-NEXT: s_lshl_b32 s1, s1, 16
-; GFX11-FAKE16-NEXT: s_lshl_b32 s3, s3, 8
+; GFX11-FAKE16-NEXT: s_lshl_b32 s3, s82, 8
; GFX11-FAKE16-NEXT: s_or_b32 s0, s0, s1
; GFX11-FAKE16-NEXT: s_and_b32 s1, s17, 0xff
-; GFX11-FAKE16-NEXT: s_lshl_b32 s2, s2, 8
-; GFX11-FAKE16-NEXT: v_readlane_b32 s16, v18, 0
+; GFX11-FAKE16-NEXT: s_and_b32 s16, s80, 0xff
; GFX11-FAKE16-NEXT: s_or_b32 s1, s1, s2
-; GFX11-FAKE16-NEXT: v_readlane_b32 s2, v18, 3
+; GFX11-FAKE16-NEXT: s_and_b32 s2, s83, 0xff
; GFX11-FAKE16-NEXT: s_and_b32 s1, s1, 0xffff
-; GFX11-FAKE16-NEXT: v_readlane_b32 s17, v19, 29
-; GFX11-FAKE16-NEXT: s_and_b32 s16, s16, 0xff
-; GFX11-FAKE16-NEXT: v_readlane_b32 s100, v17, 4
-; GFX11-FAKE16-NEXT: s_and_b32 s2, s2, 0xff
-; GFX11-FAKE16-NEXT: v_readlane_b32 s99, v17, 3
; GFX11-FAKE16-NEXT: s_or_b32 s2, s2, s3
; GFX11-FAKE16-NEXT: s_and_b32 s3, s18, 0xff
; GFX11-FAKE16-NEXT: s_lshl_b32 s2, s2, 16
-; GFX11-FAKE16-NEXT: s_lshl_b32 s17, s17, 8
+; GFX11-FAKE16-NEXT: s_lshl_b32 s17, s69, 8
; GFX11-FAKE16-NEXT: s_or_b32 s1, s1, s2
-; GFX11-FAKE16-NEXT: v_readlane_b32 s2, v18, 1
+; GFX11-FAKE16-NEXT: s_lshl_b32 s2, s81, 8
; GFX11-FAKE16-NEXT: v_dual_mov_b32 v5, s0 :: v_dual_mov_b32 v6, s1
-; GFX11-FAKE16-NEXT: v_readlane_b32 s0, v19, 28
-; GFX11-FAKE16-NEXT: s_and_b32 s1, s20, 0xff
-; GFX11-FAKE16-NEXT: s_lshl_b32 s2, s2, 8
-; GFX11-FAKE16-NEXT: v_readlane_b32 s18, v19, 19
; GFX11-FAKE16-NEXT: s_or_b32 s2, s3, s2
; GFX11-FAKE16-NEXT: s_lshl_b32 s3, s90, 8
; GFX11-FAKE16-NEXT: s_and_b32 s2, s2, 0xffff
; GFX11-FAKE16-NEXT: s_or_b32 s3, s16, s3
-; GFX11-FAKE16-NEXT: v_readlane_b32 s16, v19, 31
+; GFX11-FAKE16-NEXT: s_lshl_b32 s16, s71, 8
; GFX11-FAKE16-NEXT: s_lshl_b32 s3, s3, 16
-; GFX11-FAKE16-NEXT: s_lshl_b32 s0, s0, 8
+; GFX11-FAKE16-NEXT: s_lshl_b32 s0, s68, 8
; GFX11-FAKE16-NEXT: s_or_b32 s2, s2, s3
; GFX11-FAKE16-NEXT: s_and_b32 s3, s19, 0xff
-; GFX11-FAKE16-NEXT: s_lshl_b32 s16, s16, 8
-; GFX11-FAKE16-NEXT: s_or_b32 s0, s1, s0
+; GFX11-FAKE16-NEXT: s_and_b32 s1, s20, 0xff
; GFX11-FAKE16-NEXT: s_or_b32 s3, s3, s16
-; GFX11-FAKE16-NEXT: v_readlane_b32 s16, v19, 30
+; GFX11-FAKE16-NEXT: s_and_b32 s16, s70, 0xff
; GFX11-FAKE16-NEXT: s_and_b32 s3, s3, 0xffff
-; GFX11-FAKE16-NEXT: s_lshl_b32 s1, s78, 8
-; GFX11-FAKE16-NEXT: s_and_b32 s0, s0, 0xffff
-; GFX11-FAKE16-NEXT: s_lshl_b32 s18, s18, 8
-; GFX11-FAKE16-NEXT: s_and_b32 s16, s16, 0xff
-; GFX11-FAKE16-NEXT: s_lshl_b32 s19, s86, 8
; GFX11-FAKE16-NEXT: s_or_b32 s16, s16, s17
-; GFX11-FAKE16-NEXT: v_readlane_b32 s17, v19, 21
+; GFX11-FAKE16-NEXT: s_or_b32 s0, s1, s0
; GFX11-FAKE16-NEXT: s_lshl_b32 s16, s16, 16
-; GFX11-FAKE16-NEXT: v_readlane_b32 s98, v17, 2
+; GFX11-FAKE16-NEXT: s_lshl_b32 s1, s78, 8
; GFX11-FAKE16-NEXT: s_or_b32 s3, s3, s16
; GFX11-FAKE16-NEXT: s_delay_alu instid0(SALU_CYCLE_1)
; GFX11-FAKE16-NEXT: v_dual_mov_b32 v7, s2 :: v_dual_mov_b32 v8, s3
-; GFX11-FAKE16-NEXT: v_readlane_b32 s2, v19, 27
-; GFX11-FAKE16-NEXT: v_readlane_b32 s3, v19, 24
-; GFX11-FAKE16-NEXT: v_readlane_b32 s16, v19, 22
-; GFX11-FAKE16-NEXT: s_lshl_b32 s17, s17, 8
-; GFX11-FAKE16-NEXT: s_clause 0x1
-; GFX11-FAKE16-NEXT: scratch_store_b128 v0, v[1:4], off
-; GFX11-FAKE16-NEXT: scratch_store_b128 v0, v[5:8], off offset:16
-; GFX11-FAKE16-NEXT: s_and_b32 s2, s2, 0xff
-; GFX11-FAKE16-NEXT: s_lshl_b32 s3, s3, 8
+; GFX11-FAKE16-NEXT: s_and_b32 s2, s67, 0xff
+; GFX11-FAKE16-NEXT: s_and_b32 s0, s0, 0xffff
; GFX11-FAKE16-NEXT: s_or_b32 s1, s2, s1
-; GFX11-FAKE16-NEXT: v_readlane_b32 s2, v19, 26
+; GFX11-FAKE16-NEXT: s_lshl_b32 s2, s66, 8
; GFX11-FAKE16-NEXT: s_lshl_b32 s1, s1, 16
-; GFX11-FAKE16-NEXT: s_and_b32 s16, s16, 0xff
+; GFX11-FAKE16-NEXT: s_lshl_b32 s3, s64, 8
; GFX11-FAKE16-NEXT: s_or_b32 s0, s0, s1
; GFX11-FAKE16-NEXT: s_and_b32 s1, s21, 0xff
-; GFX11-FAKE16-NEXT: s_lshl_b32 s2, s2, 8
-; GFX11-FAKE16-NEXT: v_readlane_b32 s86, v16, 30
+; GFX11-FAKE16-NEXT: s_and_b32 s16, s54, 0xff
; GFX11-FAKE16-NEXT: s_or_b32 s1, s1, s2
-; GFX11-FAKE16-NEXT: v_readlane_b32 s2, v19, 25
+; GFX11-FAKE16-NEXT: s_and_b32 s2, s65, 0xff
; GFX11-FAKE16-NEXT: s_and_b32 s1, s1, 0xffff
-; GFX11-FAKE16-NEXT: v_readlane_b32 s31, v16, 1
-; GFX11-FAKE16-NEXT: v_readlane_b32 s30, v16, 0
-; GFX11-FAKE16-NEXT: s_and_b32 s2, s2, 0xff
-; GFX11-FAKE16-NEXT: s_delay_alu instid0(SALU_CYCLE_1) | instskip(SKIP_2) | instid1(SALU_CYCLE_1)
; GFX11-FAKE16-NEXT: s_or_b32 s2, s2, s3
; GFX11-FAKE16-NEXT: s_and_b32 s3, s22, 0xff
; GFX11-FAKE16-NEXT: s_lshl_b32 s2, s2, 16
+; GFX11-FAKE16-NEXT: s_lshl_b32 s17, s53, 8
; GFX11-FAKE16-NEXT: s_or_b32 s1, s1, s2
-; GFX11-FAKE16-NEXT: v_readlane_b32 s2, v19, 23
-; GFX11-FAKE16-NEXT: v_dual_mov_b32 v9, s0 :: v_dual_mov_b32 v10, s1
-; GFX11-FAKE16-NEXT: v_readlane_b32 s1, v19, 18
-; GFX11-FAKE16-NEXT: s_and_b32 s0, s24, 0xff
-; GFX11-FAKE16-NEXT: s_lshl_b32 s2, s2, 8
-; GFX11-FAKE16-NEXT: s_delay_alu instid0(SALU_CYCLE_1)
+; GFX11-FAKE16-NEXT: s_lshl_b32 s2, s55, 8
+; GFX11-FAKE16-NEXT: s_lshl_b32 s18, s51, 8
; GFX11-FAKE16-NEXT: s_or_b32 s2, s3, s2
-; GFX11-FAKE16-NEXT: s_lshl_b32 s3, s62, 8
+; GFX11-FAKE16-NEXT: s_lshl_b32 s3, s74, 8
; GFX11-FAKE16-NEXT: s_and_b32 s2, s2, 0xffff
; GFX11-FAKE16-NEXT: s_or_b32 s3, s16, s3
; GFX11-FAKE16-NEXT: s_and_b32 s16, s23, 0xff
; GFX11-FAKE16-NEXT: s_lshl_b32 s3, s3, 16
; GFX11-FAKE16-NEXT: s_or_b32 s16, s16, s17
-; GFX11-FAKE16-NEXT: v_readlane_b32 s17, v19, 20
+; GFX11-FAKE16-NEXT: s_and_b32 s17, s52, 0xff
; GFX11-FAKE16-NEXT: s_or_b32 s2, s2, s3
-; GFX11-FAKE16-NEXT: s_and_b32 s3, s16, 0xffff
-; GFX11-FAKE16-NEXT: s_lshl_b32 s1, s1, 8
-; GFX11-FAKE16-NEXT: s_and_b32 s17, s17, 0xff
-; GFX11-FAKE16-NEXT: s_or_b32 s0, s0, s1
; GFX11-FAKE16-NEXT: s_or_b32 s17, s17, s18
-; GFX11-FAKE16-NEXT: s_and_b32 s0, s0, 0xffff
+; GFX11-FAKE16-NEXT: s_and_b32 s3, s16, 0xffff
; GFX11-FAKE16-NEXT: s_lshl_b32 s16, s17, 16
-; GFX11-FAKE16-NEXT: s_lshl_b32 s17, s97, 8
+; GFX11-FAKE16-NEXT: v_dual_mov_b32 v9, s0 :: v_dual_mov_b32 v10, s1
; GFX11-FAKE16-NEXT: s_or_b32 s3, s3, s16
; GFX11-FAKE16-NEXT: s_delay_alu instid0(SALU_CYCLE_1)
; GFX11-FAKE16-NEXT: v_dual_mov_b32 v11, s2 :: v_dual_mov_b32 v12, s3
-; GFX11-FAKE16-NEXT: v_readlane_b32 s2, v19, 17
+; GFX11-FAKE16-NEXT: s_and_b32 s0, s24, 0xff
+; GFX11-FAKE16-NEXT: s_lshl_b32 s1, s50, 8
+; GFX11-FAKE16-NEXT: s_and_b32 s2, s49, 0xff
; GFX11-FAKE16-NEXT: s_lshl_b32 s3, s88, 8
-; GFX11-FAKE16-NEXT: s_and_b32 s16, s69, 0xff
-; GFX11-FAKE16-NEXT: s_and_b32 s18, s72, 0xff
-; GFX11-FAKE16-NEXT: v_readlane_b32 s97, v17, 1
-; GFX11-FAKE16-NEXT: s_and_b32 s2, s2, 0xff
-; GFX11-FAKE16-NEXT: v_readlane_b32 s69, v16, 21
+; GFX11-FAKE16-NEXT: s_or_b32 s0, s0, s1
; GFX11-FAKE16-NEXT: s_or_b32 s1, s2, s3
-; GFX11-FAKE16-NEXT: v_readlane_b32 s3, v19, 16
; GFX11-FAKE16-NEXT: s_and_b32 s2, s25, 0xff
-; GFX11-FAKE16-NEXT: s_lshl_b32 s1, s1, 16
-; GFX11-FAKE16-NEXT: s_delay_alu instid0(SALU_CYCLE_1) | instskip(SKIP_1) | instid1(SALU_CYCLE_1)
-; GFX11-FAKE16-NEXT: s_or_b32 s0, s0, s1
-; GFX11-FAKE16-NEXT: s_lshl_b32 s3, s3, 8
+; GFX11-FAKE16-NEXT: s_lshl_b32 s3, s48, 8
+; GFX11-FAKE16-NEXT: s_and_b32 s16, s39, 0xff
+; GFX11-FAKE16-NEXT: s_lshl_b32 s17, s38, 8
; GFX11-FAKE16-NEXT: s_or_b32 s2, s2, s3
; GFX11-FAKE16-NEXT: s_or_b32 s3, s16, s17
+; GFX11-FAKE16-NEXT: s_and_b32 s0, s0, 0xffff
+; GFX11-FAKE16-NEXT: s_lshl_b32 s1, s1, 16
; GFX11-FAKE16-NEXT: s_and_b32 s2, s2, 0xffff
; GFX11-FAKE16-NEXT: s_lshl_b32 s3, s3, 16
-; GFX11-FAKE16-NEXT: s_and_b32 s16, s73, 0xff
+; GFX11-FAKE16-NEXT: s_or_b32 s0, s0, s1
; GFX11-FAKE16-NEXT: s_or_b32 s1, s2, s3
; GFX11-FAKE16-NEXT: s_and_b32 s2, s26, 0xff
-; GFX11-FAKE16-NEXT: s_lshl_b32 s3, s96, 8
+; GFX11-FAKE16-NEXT: s_lshl_b32 s3, s37, 8
+; GFX11-FAKE16-NEXT: s_and_b32 s16, s36, 0xff
; GFX11-FAKE16-NEXT: s_lshl_b32 s17, s76, 8
; GFX11-FAKE16-NEXT: s_or_b32 s2, s2, s3
; GFX11-FAKE16-NEXT: s_or_b32 s3, s16, s17
; GFX11-FAKE16-NEXT: s_and_b32 s16, s27, 0xff
-; GFX11-FAKE16-NEXT: s_lshl_b32 s17, s87, 8
-; GFX11-FAKE16-NEXT: s_and_b32 s2, s2, 0xffff
+; GFX11-FAKE16-NEXT: s_lshl_b32 s17, s35, 8
+; GFX11-FAKE16-NEXT: s_and_b32 s18, s34, 0xff
+; GFX11-FAKE16-NEXT: s_lshl_b32 s19, vcc_hi, 8
; GFX11-FAKE16-NEXT: s_or_b32 s16, s16, s17
; GFX11-FAKE16-NEXT: s_or_b32 s17, s18, s19
+; GFX11-FAKE16-NEXT: s_and_b32 s2, s2, 0xffff
; GFX11-FAKE16-NEXT: s_lshl_b32 s3, s3, 16
; GFX11-FAKE16-NEXT: s_and_b32 s16, s16, 0xffff
; GFX11-FAKE16-NEXT: s_lshl_b32 s17, s17, 16
; GFX11-FAKE16-NEXT: s_or_b32 s2, s2, s3
; GFX11-FAKE16-NEXT: s_or_b32 s3, s16, s17
-; GFX11-FAKE16-NEXT: v_readlane_b32 s16, v19, 0
+; GFX11-FAKE16-NEXT: v_readlane_b32 s16, v18, 8
+; GFX11-FAKE16-NEXT: v_readlane_b32 s17, v18, 7
+; GFX11-FAKE16-NEXT: s_clause 0x1
+; GFX11-FAKE16-NEXT: scratch_store_b128 v0, v[1:4], off
+; GFX11-FAKE16-NEXT: scratch_store_b128 v0, v[5:8], off offset:16
; GFX11-FAKE16-NEXT: v_dual_mov_b32 v1, s0 :: v_dual_mov_b32 v2, s1
; GFX11-FAKE16-NEXT: v_dual_mov_b32 v3, s2 :: v_dual_mov_b32 v4, s3
; GFX11-FAKE16-NEXT: s_and_b32 s0, s28, 0xff
-; GFX11-FAKE16-NEXT: s_lshl_b32 s1, s85, 8
-; GFX11-FAKE16-NEXT: s_and_b32 s2, s84, 0xff
-; GFX11-FAKE16-NEXT: s_lshl_b32 s3, s16, 8
-; GFX11-FAKE16-NEXT: v_readlane_b32 s17, v19, 1
+; GFX11-FAKE16-NEXT: s_lshl_b32 s1, s104, 8
+; GFX11-FAKE16-NEXT: s_and_b32 s2, s103, 0xff
+; GFX11-FAKE16-NEXT: s_lshl_b32 s3, s72, 8
; GFX11-FAKE16-NEXT: s_or_b32 s0, s0, s1
; GFX11-FAKE16-NEXT: s_or_b32 s1, s2, s3
; GFX11-FAKE16-NEXT: s_and_b32 s2, s29, 0xff
-; GFX11-FAKE16-NEXT: s_lshl_b32 s3, s83, 8
-; GFX11-FAKE16-NEXT: s_and_b32 s16, s82, 0xff
-; GFX11-FAKE16-NEXT: s_lshl_b32 s17, s81, 8
-; GFX11-FAKE16-NEXT: v_readlane_b32 s18, v19, 2
+; GFX11-FAKE16-NEXT: s_lshl_b32 s3, s102, 8
+; GFX11-FAKE16-NEXT: s_and_b32 s16, s16, 0xff
+; GFX11-FAKE16-NEXT: s_lshl_b32 s17, s17, 8
; GFX11-FAKE16-NEXT: s_or_b32 s2, s2, s3
; GFX11-FAKE16-NEXT: s_or_b32 s3, s16, s17
; GFX11-FAKE16-NEXT: s_and_b32 s0, s0, 0xffff
@@ -12209,147 +12030,176 @@ define inreg <128 x i8> @bitcast_v32i32_to_v128i8_scalar(<32 x i32> inreg %a, i3
; GFX11-FAKE16-NEXT: s_lshl_b32 s3, s3, 16
; GFX11-FAKE16-NEXT: s_or_b32 s0, s0, s1
; GFX11-FAKE16-NEXT: s_or_b32 s1, s2, s3
+; GFX11-FAKE16-NEXT: v_readlane_b32 s3, v18, 6
+; GFX11-FAKE16-NEXT: v_readlane_b32 s16, v18, 5
; GFX11-FAKE16-NEXT: s_and_b32 s2, s40, 0xff
-; GFX11-FAKE16-NEXT: s_lshl_b32 s3, s61, 8
-; GFX11-FAKE16-NEXT: s_and_b32 s16, s80, 0xff
-; GFX11-FAKE16-NEXT: s_lshl_b32 s17, s18, 8
-; GFX11-FAKE16-NEXT: v_readlane_b32 s19, v19, 3
+; GFX11-FAKE16-NEXT: s_lshl_b32 s17, s62, 8
+; GFX11-FAKE16-NEXT: v_readlane_b32 s18, v18, 3
+; GFX11-FAKE16-NEXT: s_lshl_b32 s3, s3, 8
+; GFX11-FAKE16-NEXT: s_and_b32 s16, s16, 0xff
; GFX11-FAKE16-NEXT: s_or_b32 s2, s2, s3
; GFX11-FAKE16-NEXT: s_or_b32 s3, s16, s17
+; GFX11-FAKE16-NEXT: v_readlane_b32 s17, v18, 4
+; GFX11-FAKE16-NEXT: v_readlane_b32 s19, v18, 2
; GFX11-FAKE16-NEXT: s_and_b32 s16, s41, 0xff
-; GFX11-FAKE16-NEXT: s_lshl_b32 s17, s60, 8
-; GFX11-FAKE16-NEXT: s_and_b32 s18, s71, 0xff
-; GFX11-FAKE16-NEXT: s_lshl_b32 s19, s70, 8
-; GFX11-FAKE16-NEXT: s_or_b32 s16, s16, s17
-; GFX11-FAKE16-NEXT: s_or_b32 s17, s18, s19
+; GFX11-FAKE16-NEXT: s_and_b32 s18, s18, 0xff
; GFX11-FAKE16-NEXT: s_and_b32 s2, s2, 0xffff
+; GFX11-FAKE16-NEXT: s_lshl_b32 s17, s17, 8
+; GFX11-FAKE16-NEXT: s_lshl_b32 s19, s19, 8
; GFX11-FAKE16-NEXT: s_lshl_b32 s3, s3, 16
+; GFX11-FAKE16-NEXT: s_or_b32 s16, s16, s17
+; GFX11-FAKE16-NEXT: s_or_b32 s17, s18, s19
; GFX11-FAKE16-NEXT: s_and_b32 s16, s16, 0xffff
; GFX11-FAKE16-NEXT: s_lshl_b32 s17, s17, 16
; GFX11-FAKE16-NEXT: s_or_b32 s2, s2, s3
; GFX11-FAKE16-NEXT: s_or_b32 s3, s16, s17
-; GFX11-FAKE16-NEXT: v_readlane_b32 s16, v19, 4
; GFX11-FAKE16-NEXT: v_dual_mov_b32 v5, s0 :: v_dual_mov_b32 v6, s1
; GFX11-FAKE16-NEXT: v_dual_mov_b32 v7, s2 :: v_dual_mov_b32 v8, s3
+; GFX11-FAKE16-NEXT: v_readlane_b32 s1, v18, 1
+; GFX11-FAKE16-NEXT: v_readlane_b32 s2, v18, 0
; GFX11-FAKE16-NEXT: s_and_b32 s0, s14, 0xff
-; GFX11-FAKE16-NEXT: s_lshl_b32 s1, s58, 8
-; GFX11-FAKE16-NEXT: s_and_b32 s2, s59, 0xff
-; GFX11-FAKE16-NEXT: s_lshl_b32 s3, s16, 8
+; GFX11-FAKE16-NEXT: s_lshl_b32 s3, s60, 8
+; GFX11-FAKE16-NEXT: v_readlane_b32 s14, v19, 30
+; GFX11-FAKE16-NEXT: s_lshl_b32 s1, s1, 8
+; GFX11-FAKE16-NEXT: s_and_b32 s2, s2, 0xff
; GFX11-FAKE16-NEXT: s_or_b32 s0, s0, s1
; GFX11-FAKE16-NEXT: s_or_b32 s1, s2, s3
; GFX11-FAKE16-NEXT: s_and_b32 s2, s15, 0xff
-; GFX11-FAKE16-NEXT: s_lshl_b32 s3, s68, 8
-; GFX11-FAKE16-NEXT: s_and_b32 s14, s67, 0xff
-; GFX11-FAKE16-NEXT: s_lshl_b32 s15, s66, 8
-; GFX11-FAKE16-NEXT: s_or_b32 s2, s2, s3
-; GFX11-FAKE16-NEXT: s_or_b32 s3, s14, s15
-; GFX11-FAKE16-NEXT: v_readlane_b32 s14, v19, 6
+; GFX11-FAKE16-NEXT: v_readlane_b32 s3, v19, 31
+; GFX11-FAKE16-NEXT: v_readlane_b32 s15, v19, 29
+; GFX11-FAKE16-NEXT: s_and_b32 s14, s14, 0xff
; GFX11-FAKE16-NEXT: s_and_b32 s0, s0, 0xffff
; GFX11-FAKE16-NEXT: s_lshl_b32 s1, s1, 16
+; GFX11-FAKE16-NEXT: s_lshl_b32 s3, s3, 8
+; GFX11-FAKE16-NEXT: s_lshl_b32 s15, s15, 8
+; GFX11-FAKE16-NEXT: s_or_b32 s2, s2, s3
+; GFX11-FAKE16-NEXT: s_or_b32 s3, s14, s15
; GFX11-FAKE16-NEXT: s_and_b32 s2, s2, 0xffff
; GFX11-FAKE16-NEXT: s_lshl_b32 s3, s3, 16
; GFX11-FAKE16-NEXT: s_or_b32 s0, s0, s1
; GFX11-FAKE16-NEXT: s_or_b32 s1, s2, s3
; GFX11-FAKE16-NEXT: s_and_b32 s2, s12, 0xff
-; GFX11-FAKE16-NEXT: s_lshl_b32 s3, s65, 8
-; GFX11-FAKE16-NEXT: s_and_b32 s12, s64, 0xff
-; GFX11-FAKE16-NEXT: s_lshl_b32 s14, s14, 8
-; GFX11-FAKE16-NEXT: v_readlane_b32 s15, v19, 7
+; GFX11-FAKE16-NEXT: v_readlane_b32 s3, v19, 28
+; GFX11-FAKE16-NEXT: v_readlane_b32 s12, v19, 27
+; GFX11-FAKE16-NEXT: s_lshl_b32 s14, s58, 8
+; GFX11-FAKE16-NEXT: v_readlane_b32 s15, v19, 24
+; GFX11-FAKE16-NEXT: s_clause 0x1
+; GFX11-FAKE16-NEXT: scratch_store_b128 v0, v[9:12], off offset:32
+; GFX11-FAKE16-NEXT: scratch_store_b128 v0, v[1:4], off offset:48
+; GFX11-FAKE16-NEXT: s_lshl_b32 s3, s3, 8
+; GFX11-FAKE16-NEXT: s_and_b32 s12, s12, 0xff
; GFX11-FAKE16-NEXT: s_or_b32 s2, s2, s3
; GFX11-FAKE16-NEXT: s_or_b32 s3, s12, s14
; GFX11-FAKE16-NEXT: s_and_b32 s12, s13, 0xff
-; GFX11-FAKE16-NEXT: s_lshl_b32 s13, s55, 8
-; GFX11-FAKE16-NEXT: s_and_b32 s14, s54, 0xff
-; GFX11-FAKE16-NEXT: s_lshl_b32 s15, s53, 8
-; GFX11-FAKE16-NEXT: s_or_b32 s12, s12, s13
-; GFX11-FAKE16-NEXT: s_or_b32 s13, s14, s15
+; GFX11-FAKE16-NEXT: v_readlane_b32 s13, v19, 26
+; GFX11-FAKE16-NEXT: v_readlane_b32 s14, v19, 25
+; GFX11-FAKE16-NEXT: s_lshl_b32 s15, s15, 8
; GFX11-FAKE16-NEXT: s_and_b32 s2, s2, 0xffff
; GFX11-FAKE16-NEXT: s_lshl_b32 s3, s3, 16
+; GFX11-FAKE16-NEXT: s_lshl_b32 s13, s13, 8
+; GFX11-FAKE16-NEXT: s_and_b32 s14, s14, 0xff
+; GFX11-FAKE16-NEXT: s_or_b32 s12, s12, s13
+; GFX11-FAKE16-NEXT: s_or_b32 s13, s14, s15
; GFX11-FAKE16-NEXT: s_and_b32 s12, s12, 0xffff
; GFX11-FAKE16-NEXT: s_lshl_b32 s13, s13, 16
; GFX11-FAKE16-NEXT: s_or_b32 s2, s2, s3
; GFX11-FAKE16-NEXT: s_or_b32 s3, s12, s13
-; GFX11-FAKE16-NEXT: v_readlane_b32 s12, v19, 8
-; GFX11-FAKE16-NEXT: s_clause 0x1
-; GFX11-FAKE16-NEXT: scratch_store_b128 v0, v[9:12], off offset:32
-; GFX11-FAKE16-NEXT: scratch_store_b128 v0, v[1:4], off offset:48
; GFX11-FAKE16-NEXT: v_dual_mov_b32 v9, s0 :: v_dual_mov_b32 v10, s1
; GFX11-FAKE16-NEXT: v_dual_mov_b32 v11, s2 :: v_dual_mov_b32 v12, s3
+; GFX11-FAKE16-NEXT: v_readlane_b32 s1, v19, 23
+; GFX11-FAKE16-NEXT: v_readlane_b32 s2, v19, 22
; GFX11-FAKE16-NEXT: s_and_b32 s0, s10, 0xff
-; GFX11-FAKE16-NEXT: s_lshl_b32 s1, s52, 8
-; GFX11-FAKE16-NEXT: s_and_b32 s2, s51, 0xff
-; GFX11-FAKE16-NEXT: s_lshl_b32 s3, s12, 8
+; GFX11-FAKE16-NEXT: s_lshl_b32 s3, s56, 8
+; GFX11-FAKE16-NEXT: v_readlane_b32 s10, v19, 20
+; GFX11-FAKE16-NEXT: s_lshl_b32 s1, s1, 8
+; GFX11-FAKE16-NEXT: s_and_b32 s2, s2, 0xff
; GFX11-FAKE16-NEXT: s_or_b32 s0, s0, s1
; GFX11-FAKE16-NEXT: s_or_b32 s1, s2, s3
; GFX11-FAKE16-NEXT: s_and_b32 s2, s11, 0xff
-; GFX11-FAKE16-NEXT: s_lshl_b32 s3, s50, 8
-; GFX11-FAKE16-NEXT: s_and_b32 s10, s49, 0xff
-; GFX11-FAKE16-NEXT: s_lshl_b32 s11, s48, 8
-; GFX11-FAKE16-NEXT: s_or_b32 s2, s2, s3
-; GFX11-FAKE16-NEXT: s_or_b32 s3, s10, s11
-; GFX11-FAKE16-NEXT: v_readlane_b32 s10, v19, 10
+; GFX11-FAKE16-NEXT: v_readlane_b32 s3, v19, 21
+; GFX11-FAKE16-NEXT: v_readlane_b32 s11, v19, 19
+; GFX11-FAKE16-NEXT: s_and_b32 s10, s10, 0xff
; GFX11-FAKE16-NEXT: s_and_b32 s0, s0, 0xffff
; GFX11-FAKE16-NEXT: s_lshl_b32 s1, s1, 16
+; GFX11-FAKE16-NEXT: s_lshl_b32 s3, s3, 8
+; GFX11-FAKE16-NEXT: s_lshl_b32 s11, s11, 8
+; GFX11-FAKE16-NEXT: s_or_b32 s2, s2, s3
+; GFX11-FAKE16-NEXT: s_or_b32 s3, s10, s11
; GFX11-FAKE16-NEXT: s_and_b32 s2, s2, 0xffff
; GFX11-FAKE16-NEXT: s_lshl_b32 s3, s3, 16
; GFX11-FAKE16-NEXT: s_or_b32 s0, s0, s1
; GFX11-FAKE16-NEXT: s_or_b32 s1, s2, s3
; GFX11-FAKE16-NEXT: s_and_b32 s2, s8, 0xff
-; GFX11-FAKE16-NEXT: s_lshl_b32 s3, s39, 8
-; GFX11-FAKE16-NEXT: s_and_b32 s8, s38, 0xff
-; GFX11-FAKE16-NEXT: s_lshl_b32 s10, s10, 8
-; GFX11-FAKE16-NEXT: v_readlane_b32 s11, v19, 11
+; GFX11-FAKE16-NEXT: v_readlane_b32 s3, v19, 18
+; GFX11-FAKE16-NEXT: v_readlane_b32 s8, v19, 17
+; GFX11-FAKE16-NEXT: s_lshl_b32 s10, s46, 8
+; GFX11-FAKE16-NEXT: v_readlane_b32 s11, v19, 14
+; GFX11-FAKE16-NEXT: v_dual_mov_b32 v1, s0 :: v_dual_mov_b32 v2, s1
+; GFX11-FAKE16-NEXT: s_lshl_b32 s3, s3, 8
+; GFX11-FAKE16-NEXT: s_and_b32 s8, s8, 0xff
; GFX11-FAKE16-NEXT: s_or_b32 s2, s2, s3
; GFX11-FAKE16-NEXT: s_or_b32 s3, s8, s10
; GFX11-FAKE16-NEXT: s_and_b32 s8, s9, 0xff
-; GFX11-FAKE16-NEXT: s_lshl_b32 s9, s37, 8
-; GFX11-FAKE16-NEXT: s_and_b32 s10, s36, 0xff
-; GFX11-FAKE16-NEXT: s_lshl_b32 s11, s35, 8
-; GFX11-FAKE16-NEXT: s_or_b32 s8, s8, s9
-; GFX11-FAKE16-NEXT: s_or_b32 s9, s10, s11
+; GFX11-FAKE16-NEXT: v_readlane_b32 s9, v19, 16
+; GFX11-FAKE16-NEXT: v_readlane_b32 s10, v19, 15
+; GFX11-FAKE16-NEXT: s_lshl_b32 s11, s11, 8
; GFX11-FAKE16-NEXT: s_and_b32 s2, s2, 0xffff
; GFX11-FAKE16-NEXT: s_lshl_b32 s3, s3, 16
+; GFX11-FAKE16-NEXT: s_lshl_b32 s9, s9, 8
+; GFX11-FAKE16-NEXT: s_and_b32 s10, s10, 0xff
+; GFX11-FAKE16-NEXT: s_or_b32 s8, s8, s9
+; GFX11-FAKE16-NEXT: s_or_b32 s9, s10, s11
; GFX11-FAKE16-NEXT: s_and_b32 s8, s8, 0xffff
; GFX11-FAKE16-NEXT: s_lshl_b32 s9, s9, 16
; GFX11-FAKE16-NEXT: s_or_b32 s2, s2, s3
; GFX11-FAKE16-NEXT: s_or_b32 s3, s8, s9
-; GFX11-FAKE16-NEXT: v_readlane_b32 s8, v19, 12
-; GFX11-FAKE16-NEXT: v_dual_mov_b32 v1, s0 :: v_dual_mov_b32 v2, s1
+; GFX11-FAKE16-NEXT: s_delay_alu instid0(SALU_CYCLE_1)
; GFX11-FAKE16-NEXT: v_dual_mov_b32 v3, s2 :: v_dual_mov_b32 v4, s3
+; GFX11-FAKE16-NEXT: v_readlane_b32 s1, v19, 13
+; GFX11-FAKE16-NEXT: v_readlane_b32 s2, v19, 12
+; GFX11-FAKE16-NEXT: v_readlane_b32 s8, v19, 0
; GFX11-FAKE16-NEXT: s_and_b32 s0, s6, 0xff
-; GFX11-FAKE16-NEXT: s_lshl_b32 s1, s56, 8
-; GFX11-FAKE16-NEXT: s_and_b32 s2, s57, 0xff
+; GFX11-FAKE16-NEXT: v_readlane_b32 s6, v19, 10
+; GFX11-FAKE16-NEXT: s_lshl_b32 s1, s1, 8
+; GFX11-FAKE16-NEXT: s_and_b32 s2, s2, 0xff
; GFX11-FAKE16-NEXT: s_lshl_b32 s3, s8, 8
; GFX11-FAKE16-NEXT: s_or_b32 s0, s0, s1
; GFX11-FAKE16-NEXT: s_or_b32 s1, s2, s3
; GFX11-FAKE16-NEXT: s_and_b32 s2, s7, 0xff
-; GFX11-FAKE16-NEXT: s_lshl_b32 s3, s34, 8
-; GFX11-FAKE16-NEXT: s_and_b32 s6, vcc_hi, 0xff
-; GFX11-FAKE16-NEXT: s_lshl_b32 s7, s46, 8
-; GFX11-FAKE16-NEXT: s_or_b32 s2, s2, s3
-; GFX11-FAKE16-NEXT: s_or_b32 s3, s6, s7
-; GFX11-FAKE16-NEXT: v_readlane_b32 s6, v19, 14
+; GFX11-FAKE16-NEXT: v_readlane_b32 s3, v19, 11
+; GFX11-FAKE16-NEXT: v_readlane_b32 s7, v19, 9
+; GFX11-FAKE16-NEXT: s_and_b32 s6, s6, 0xff
; GFX11-FAKE16-NEXT: s_and_b32 s0, s0, 0xffff
; GFX11-FAKE16-NEXT: s_lshl_b32 s1, s1, 16
+; GFX11-FAKE16-NEXT: s_lshl_b32 s3, s3, 8
+; GFX11-FAKE16-NEXT: s_lshl_b32 s7, s7, 8
+; GFX11-FAKE16-NEXT: s_or_b32 s2, s2, s3
+; GFX11-FAKE16-NEXT: s_or_b32 s3, s6, s7
; GFX11-FAKE16-NEXT: s_and_b32 s2, s2, 0xffff
; GFX11-FAKE16-NEXT: s_lshl_b32 s3, s3, 16
; GFX11-FAKE16-NEXT: s_or_b32 s0, s0, s1
; GFX11-FAKE16-NEXT: s_or_b32 s1, s2, s3
; GFX11-FAKE16-NEXT: s_and_b32 s2, s4, 0xff
-; GFX11-FAKE16-NEXT: s_lshl_b32 s3, s47, 8
-; GFX11-FAKE16-NEXT: s_and_b32 s4, s104, 0xff
+; GFX11-FAKE16-NEXT: v_readlane_b32 s3, v19, 8
+; GFX11-FAKE16-NEXT: v_readlane_b32 s4, v19, 7
+; GFX11-FAKE16-NEXT: v_readlane_b32 s6, v19, 2
+; GFX11-FAKE16-NEXT: v_readlane_b32 s7, v19, 3
+; GFX11-FAKE16-NEXT: v_readlane_b32 s7, v19, 4
+; GFX11-FAKE16-NEXT: s_lshl_b32 s3, s3, 8
+; GFX11-FAKE16-NEXT: s_and_b32 s4, s4, 0xff
; GFX11-FAKE16-NEXT: s_lshl_b32 s6, s6, 8
-; GFX11-FAKE16-NEXT: v_readlane_b32 s7, v19, 15
; GFX11-FAKE16-NEXT: s_or_b32 s2, s2, s3
; GFX11-FAKE16-NEXT: s_or_b32 s3, s4, s6
; GFX11-FAKE16-NEXT: s_and_b32 s4, s5, 0xff
-; GFX11-FAKE16-NEXT: s_lshl_b32 s5, s103, 8
-; GFX11-FAKE16-NEXT: s_and_b32 s6, s102, 0xff
-; GFX11-FAKE16-NEXT: s_lshl_b32 s7, s101, 8
-; GFX11-FAKE16-NEXT: s_or_b32 s4, s4, s5
-; GFX11-FAKE16-NEXT: s_or_b32 s5, s6, s7
+; GFX11-FAKE16-NEXT: v_readlane_b32 s5, v19, 6
+; GFX11-FAKE16-NEXT: v_readlane_b32 s6, v19, 5
+; GFX11-FAKE16-NEXT: s_lshl_b32 s7, s7, 8
; GFX11-FAKE16-NEXT: s_and_b32 s2, s2, 0xffff
; GFX11-FAKE16-NEXT: s_lshl_b32 s3, s3, 16
+; GFX11-FAKE16-NEXT: s_lshl_b32 s5, s5, 8
+; GFX11-FAKE16-NEXT: s_and_b32 s6, s6, 0xff
+; GFX11-FAKE16-NEXT: s_or_b32 s4, s4, s5
+; GFX11-FAKE16-NEXT: s_or_b32 s5, s6, s7
; GFX11-FAKE16-NEXT: s_and_b32 s4, s4, 0xffff
; GFX11-FAKE16-NEXT: s_lshl_b32 s5, s5, 16
; GFX11-FAKE16-NEXT: s_or_b32 s2, s2, s3
@@ -12357,9 +12207,7 @@ define inreg <128 x i8> @bitcast_v32i32_to_v128i8_scalar(<32 x i32> inreg %a, i3
; GFX11-FAKE16-NEXT: scratch_store_b128 v0, v[5:8], off offset:64
; GFX11-FAKE16-NEXT: v_dual_mov_b32 v5, s0 :: v_dual_mov_b32 v6, s1
; GFX11-FAKE16-NEXT: v_dual_mov_b32 v7, s2 :: v_dual_mov_b32 v8, s3
-; GFX11-FAKE16-NEXT: v_readlane_b32 s17, v19, 5
-; GFX11-FAKE16-NEXT: v_readlane_b32 s13, v19, 9
-; GFX11-FAKE16-NEXT: v_readlane_b32 s9, v19, 13
+; GFX11-FAKE16-NEXT: v_readlane_b32 s9, v19, 1
; GFX11-FAKE16-NEXT: s_clause 0x2
; GFX11-FAKE16-NEXT: scratch_store_b128 v0, v[9:12], off offset:80
; GFX11-FAKE16-NEXT: scratch_store_b128 v0, v[1:4], off offset:96
@@ -12368,8 +12216,13 @@ define inreg <128 x i8> @bitcast_v32i32_to_v128i8_scalar(<32 x i32> inreg %a, i3
; GFX11-FAKE16-NEXT: v_readlane_b32 s103, v17, 7
; GFX11-FAKE16-NEXT: v_readlane_b32 s102, v17, 6
; GFX11-FAKE16-NEXT: v_readlane_b32 s101, v17, 5
+; GFX11-FAKE16-NEXT: v_readlane_b32 s100, v17, 4
+; GFX11-FAKE16-NEXT: v_readlane_b32 s99, v17, 3
+; GFX11-FAKE16-NEXT: v_readlane_b32 s98, v17, 2
+; GFX11-FAKE16-NEXT: v_readlane_b32 s97, v17, 1
; GFX11-FAKE16-NEXT: v_readlane_b32 s96, v17, 0
; GFX11-FAKE16-NEXT: v_readlane_b32 s87, v16, 31
+; GFX11-FAKE16-NEXT: v_readlane_b32 s86, v16, 30
; GFX11-FAKE16-NEXT: v_readlane_b32 s85, v16, 29
; GFX11-FAKE16-NEXT: v_readlane_b32 s84, v16, 28
; GFX11-FAKE16-NEXT: v_readlane_b32 s83, v16, 27
@@ -12378,6 +12231,7 @@ define inreg <128 x i8> @bitcast_v32i32_to_v128i8_scalar(<32 x i32> inreg %a, i3
; GFX11-FAKE16-NEXT: v_readlane_b32 s80, v16, 24
; GFX11-FAKE16-NEXT: v_readlane_b32 s71, v16, 23
; GFX11-FAKE16-NEXT: v_readlane_b32 s70, v16, 22
+; GFX11-FAKE16-NEXT: v_readlane_b32 s69, v16, 21
; GFX11-FAKE16-NEXT: v_readlane_b32 s68, v16, 20
; GFX11-FAKE16-NEXT: v_readlane_b32 s67, v16, 19
; GFX11-FAKE16-NEXT: v_readlane_b32 s66, v16, 18
@@ -12397,6 +12251,8 @@ define inreg <128 x i8> @bitcast_v32i32_to_v128i8_scalar(<32 x i32> inreg %a, i3
; GFX11-FAKE16-NEXT: v_readlane_b32 s36, v16, 4
; GFX11-FAKE16-NEXT: v_readlane_b32 s35, v16, 3
; GFX11-FAKE16-NEXT: v_readlane_b32 s34, v16, 2
+; GFX11-FAKE16-NEXT: v_readlane_b32 s31, v16, 1
+; GFX11-FAKE16-NEXT: v_readlane_b32 s30, v16, 0
; GFX11-FAKE16-NEXT: s_xor_saveexec_b32 s0, -1
; GFX11-FAKE16-NEXT: s_clause 0x3
; GFX11-FAKE16-NEXT: scratch_load_b32 v16, off, s32
@@ -12406,6 +12262,145 @@ define inreg <128 x i8> @bitcast_v32i32_to_v128i8_scalar(<32 x i32> inreg %a, i3
; GFX11-FAKE16-NEXT: s_mov_b32 exec_lo, s0
; GFX11-FAKE16-NEXT: s_waitcnt vmcnt(0)
; GFX11-FAKE16-NEXT: s_setpc_b64 s[30:31]
+; GFX11-FAKE16-NEXT: .LBB13_4:
+; GFX11-FAKE16-NEXT: ; implicit-def: $sgpr46
+; GFX11-FAKE16-NEXT: ; kill: killed $sgpr46
+; GFX11-FAKE16-NEXT: ; implicit-def: $vcc_hi
+; GFX11-FAKE16-NEXT: ; implicit-def: $vcc_lo
+; GFX11-FAKE16-NEXT: ; implicit-def: $sgpr46
+; GFX11-FAKE16-NEXT: ; kill: killed $sgpr46
+; GFX11-FAKE16-NEXT: v_writelane_b32 v19, vcc_lo, 0
+; GFX11-FAKE16-NEXT: ; implicit-def: $sgpr46
+; GFX11-FAKE16-NEXT: ; kill: killed $sgpr46
+; GFX11-FAKE16-NEXT: ; implicit-def: $sgpr101
+; GFX11-FAKE16-NEXT: ; implicit-def: $sgpr45
+; GFX11-FAKE16-NEXT: ; implicit-def: $sgpr30
+; GFX11-FAKE16-NEXT: ; implicit-def: $sgpr100
+; GFX11-FAKE16-NEXT: ; implicit-def: $sgpr99
+; GFX11-FAKE16-NEXT: ; implicit-def: $sgpr98
+; GFX11-FAKE16-NEXT: ; implicit-def: $sgpr43
+; GFX11-FAKE16-NEXT: ; implicit-def: $sgpr42
+; GFX11-FAKE16-NEXT: ; implicit-def: $sgpr94
+; GFX11-FAKE16-NEXT: ; implicit-def: $sgpr97
+; GFX11-FAKE16-NEXT: ; implicit-def: $sgpr96
+; GFX11-FAKE16-NEXT: ; implicit-def: $sgpr87
+; GFX11-FAKE16-NEXT: ; implicit-def: $sgpr86
+; GFX11-FAKE16-NEXT: ; implicit-def: $sgpr85
+; GFX11-FAKE16-NEXT: ; implicit-def: $sgpr92
+; GFX11-FAKE16-NEXT: ; implicit-def: $sgpr84
+; GFX11-FAKE16-NEXT: ; implicit-def: $sgpr83
+; GFX11-FAKE16-NEXT: ; implicit-def: $sgpr82
+; GFX11-FAKE16-NEXT: ; implicit-def: $sgpr81
+; GFX11-FAKE16-NEXT: ; implicit-def: $sgpr80
+; GFX11-FAKE16-NEXT: ; implicit-def: $sgpr90
+; GFX11-FAKE16-NEXT: ; implicit-def: $sgpr71
+; GFX11-FAKE16-NEXT: ; implicit-def: $sgpr70
+; GFX11-FAKE16-NEXT: ; implicit-def: $sgpr69
+; GFX11-FAKE16-NEXT: ; implicit-def: $sgpr68
+; GFX11-FAKE16-NEXT: ; implicit-def: $sgpr67
+; GFX11-FAKE16-NEXT: ; implicit-def: $sgpr78
+; GFX11-FAKE16-NEXT: ; implicit-def: $sgpr66
+; GFX11-FAKE16-NEXT: ; implicit-def: $sgpr65
+; GFX11-FAKE16-NEXT: ; implicit-def: $sgpr64
+; GFX11-FAKE16-NEXT: ; implicit-def: $sgpr55
+; GFX11-FAKE16-NEXT: ; implicit-def: $sgpr54
+; GFX11-FAKE16-NEXT: ; implicit-def: $sgpr74
+; GFX11-FAKE16-NEXT: ; implicit-def: $sgpr53
+; GFX11-FAKE16-NEXT: ; implicit-def: $sgpr52
+; GFX11-FAKE16-NEXT: ; implicit-def: $sgpr51
+; GFX11-FAKE16-NEXT: ; implicit-def: $sgpr50
+; GFX11-FAKE16-NEXT: ; implicit-def: $sgpr49
+; GFX11-FAKE16-NEXT: ; implicit-def: $sgpr48
+; GFX11-FAKE16-NEXT: ; implicit-def: $sgpr39
+; GFX11-FAKE16-NEXT: ; implicit-def: $sgpr38
+; GFX11-FAKE16-NEXT: ; implicit-def: $sgpr37
+; GFX11-FAKE16-NEXT: ; implicit-def: $sgpr36
+; GFX11-FAKE16-NEXT: ; implicit-def: $sgpr35
+; GFX11-FAKE16-NEXT: ; implicit-def: $sgpr34
+; GFX11-FAKE16-NEXT: ; implicit-def: $sgpr104
+; GFX11-FAKE16-NEXT: ; implicit-def: $sgpr103
+; GFX11-FAKE16-NEXT: ; implicit-def: $sgpr102
+; GFX11-FAKE16-NEXT: ; implicit-def: $sgpr88
+; GFX11-FAKE16-NEXT: ; implicit-def: $sgpr76
+; GFX11-FAKE16-NEXT: ; implicit-def: $sgpr72
+; GFX11-FAKE16-NEXT: ; implicit-def: $sgpr62
+; GFX11-FAKE16-NEXT: ; implicit-def: $sgpr60
+; GFX11-FAKE16-NEXT: ; implicit-def: $sgpr58
+; GFX11-FAKE16-NEXT: ; implicit-def: $sgpr56
+; GFX11-FAKE16-NEXT: ; implicit-def: $sgpr46
+; GFX11-FAKE16-NEXT: ; kill: killed $sgpr46
+; GFX11-FAKE16-NEXT: ; implicit-def: $sgpr46
+; GFX11-FAKE16-NEXT: ; kill: killed $sgpr46
+; GFX11-FAKE16-NEXT: ; implicit-def: $sgpr46
+; GFX11-FAKE16-NEXT: ; kill: killed $sgpr46
+; GFX11-FAKE16-NEXT: ; implicit-def: $sgpr46
+; GFX11-FAKE16-NEXT: ; kill: killed $sgpr46
+; GFX11-FAKE16-NEXT: v_writelane_b32 v19, vcc_hi, 1
+; GFX11-FAKE16-NEXT: ; implicit-def: $sgpr46
+; GFX11-FAKE16-NEXT: ; kill: killed $sgpr46
+; GFX11-FAKE16-NEXT: ; implicit-def: $vcc_lo
+; GFX11-FAKE16-NEXT: ; implicit-def: $sgpr46
+; GFX11-FAKE16-NEXT: ; kill: killed $sgpr46
+; GFX11-FAKE16-NEXT: ; implicit-def: $sgpr46
+; GFX11-FAKE16-NEXT: ; kill: killed $sgpr46
+; GFX11-FAKE16-NEXT: ; implicit-def: $sgpr46
+; GFX11-FAKE16-NEXT: ; kill: killed $sgpr46
+; GFX11-FAKE16-NEXT: ; implicit-def: $sgpr46
+; GFX11-FAKE16-NEXT: ; kill: killed $sgpr46
+; GFX11-FAKE16-NEXT: v_writelane_b32 v19, vcc_lo, 2
+; GFX11-FAKE16-NEXT: ; implicit-def: $sgpr46
+; GFX11-FAKE16-NEXT: ; kill: killed $sgpr46
+; GFX11-FAKE16-NEXT: ; implicit-def: $sgpr46
+; GFX11-FAKE16-NEXT: ; kill: killed $sgpr46
+; GFX11-FAKE16-NEXT: ; implicit-def: $sgpr46
+; GFX11-FAKE16-NEXT: ; kill: killed $sgpr46
+; GFX11-FAKE16-NEXT: ; implicit-def: $sgpr46
+; GFX11-FAKE16-NEXT: ; kill: killed $sgpr46
+; GFX11-FAKE16-NEXT: ; implicit-def: $sgpr46
+; GFX11-FAKE16-NEXT: ; kill: killed $sgpr46
+; GFX11-FAKE16-NEXT: v_writelane_b32 v19, vcc_hi, 3
+; GFX11-FAKE16-NEXT: ; implicit-def: $sgpr46
+; GFX11-FAKE16-NEXT: ; kill: killed $sgpr46
+; GFX11-FAKE16-NEXT: ; implicit-def: $sgpr46
+; GFX11-FAKE16-NEXT: ; kill: killed $sgpr46
+; GFX11-FAKE16-NEXT: ; implicit-def: $sgpr46
+; GFX11-FAKE16-NEXT: ; kill: killed $sgpr46
+; GFX11-FAKE16-NEXT: ; implicit-def: $sgpr46
+; GFX11-FAKE16-NEXT: ; kill: killed $sgpr46
+; GFX11-FAKE16-NEXT: ; implicit-def: $sgpr46
+; GFX11-FAKE16-NEXT: ; kill: killed $sgpr46
+; GFX11-FAKE16-NEXT: ; implicit-def: $sgpr46
+; GFX11-FAKE16-NEXT: ; kill: killed $sgpr46
+; GFX11-FAKE16-NEXT: ; implicit-def: $sgpr46
+; GFX11-FAKE16-NEXT: ; kill: killed $sgpr46
+; GFX11-FAKE16-NEXT: ; implicit-def: $sgpr46
+; GFX11-FAKE16-NEXT: ; kill: killed $sgpr46
+; GFX11-FAKE16-NEXT: ; implicit-def: $sgpr46
+; GFX11-FAKE16-NEXT: ; kill: killed $sgpr46
+; GFX11-FAKE16-NEXT: ; implicit-def: $sgpr46
+; GFX11-FAKE16-NEXT: ; kill: killed $sgpr46
+; GFX11-FAKE16-NEXT: ; implicit-def: $sgpr46
+; GFX11-FAKE16-NEXT: ; kill: killed $sgpr46
+; GFX11-FAKE16-NEXT: ; implicit-def: $sgpr46
+; GFX11-FAKE16-NEXT: ; kill: killed $sgpr46
+; GFX11-FAKE16-NEXT: ; implicit-def: $sgpr46
+; GFX11-FAKE16-NEXT: ; kill: killed $sgpr46
+; GFX11-FAKE16-NEXT: ; implicit-def: $sgpr46
+; GFX11-FAKE16-NEXT: ; kill: killed $sgpr46
+; GFX11-FAKE16-NEXT: ; implicit-def: $sgpr46
+; GFX11-FAKE16-NEXT: ; kill: killed $sgpr46
+; GFX11-FAKE16-NEXT: ; implicit-def: $sgpr46
+; GFX11-FAKE16-NEXT: ; kill: killed $sgpr46
+; GFX11-FAKE16-NEXT: ; implicit-def: $sgpr46
+; GFX11-FAKE16-NEXT: ; kill: killed $sgpr46
+; GFX11-FAKE16-NEXT: ; implicit-def: $sgpr46
+; GFX11-FAKE16-NEXT: ; kill: killed $sgpr46
+; GFX11-FAKE16-NEXT: ; implicit-def: $sgpr46
+; GFX11-FAKE16-NEXT: ; kill: killed $sgpr46
+; GFX11-FAKE16-NEXT: ; implicit-def: $sgpr46
+; GFX11-FAKE16-NEXT: ; kill: killed $sgpr46
+; GFX11-FAKE16-NEXT: ; implicit-def: $sgpr46
+; GFX11-FAKE16-NEXT: s_branch .LBB13_2
%cmp = icmp eq i32 %b, 0
br i1 %cmp, label %cmp.true, label %cmp.false
@@ -19705,7 +19700,7 @@ define inreg <32 x i32> @bitcast_v128i8_to_v32i32_scalar(<128 x i8> inreg %a, i3
; VI-NEXT: buffer_store_dword v10, off, s[0:3], s32 offset:476 ; 4-byte Folded Spill
; VI-NEXT: buffer_store_dword v8, off, s[0:3], s32 offset:484 ; 4-byte Folded Spill
; VI-NEXT: buffer_store_dword v6, off, s[0:3], s32 offset:488 ; 4-byte Folded Spill
-; VI-NEXT: buffer_store_dword v0, off, s[0:3], s32 offset:556 ; 4-byte Folded Spill
+; VI-NEXT: buffer_store_dword v0, off, s[0:3], s32 offset:560 ; 4-byte Folded Spill
; VI-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:332
; VI-NEXT: buffer_load_ushort v44, off, s[0:3], s32
; VI-NEXT: buffer_load_ushort v43, off, s[0:3], s32 offset:8
@@ -19731,7 +19726,7 @@ define inreg <32 x i32> @bitcast_v128i8_to_v32i32_scalar(<128 x i8> inreg %a, i3
; VI-NEXT: buffer_load_ushort v37, off, s[0:3], s32 offset:168
; VI-NEXT: buffer_load_ushort v38, off, s[0:3], s32 offset:176
; VI-NEXT: v_lshlrev_b32_e32 v1, 8, v1
-; VI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:560 ; 4-byte Folded Spill
+; VI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:564 ; 4-byte Folded Spill
; VI-NEXT: v_lshlrev_b32_e32 v1, 8, v15
; VI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:548 ; 4-byte Folded Spill
; VI-NEXT: v_lshlrev_b32_e32 v1, 8, v17
@@ -19740,11 +19735,11 @@ define inreg <32 x i32> @bitcast_v128i8_to_v32i32_scalar(<128 x i8> inreg %a, i3
; VI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:540 ; 4-byte Folded Spill
; VI-NEXT: v_lshlrev_b32_e32 v1, 8, v21
; VI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:536 ; 4-byte Folded Spill
-; VI-NEXT: v_lshlrev_b32_e32 v8, 8, v3
-; VI-NEXT: v_lshlrev_b32_e32 v59, 8, v5
-; VI-NEXT: v_lshlrev_b32_e32 v14, 8, v7
-; VI-NEXT: v_lshlrev_b32_e32 v10, 8, v9
-; VI-NEXT: v_lshlrev_b32_e32 v16, 8, v11
+; VI-NEXT: v_lshlrev_b32_e32 v56, 8, v3
+; VI-NEXT: v_lshlrev_b32_e32 v8, 8, v5
+; VI-NEXT: v_lshlrev_b32_e32 v10, 8, v7
+; VI-NEXT: v_lshlrev_b32_e32 v14, 8, v9
+; VI-NEXT: v_lshlrev_b32_e32 v12, 8, v11
; VI-NEXT: v_lshlrev_b32_e32 v6, 8, v13
; VI-NEXT: s_waitcnt vmcnt(14)
; VI-NEXT: v_cmp_ne_u32_e32 vcc, 0, v0
@@ -19765,7 +19760,7 @@ define inreg <32 x i32> @bitcast_v128i8_to_v32i32_scalar(<128 x i8> inreg %a, i3
; VI-NEXT: v_lshlrev_b32_e32 v0, 8, v41
; VI-NEXT: buffer_store_dword v0, off, s[0:3], s32 offset:504 ; 4-byte Folded Spill
; VI-NEXT: v_lshlrev_b32_e32 v0, 8, v40
-; VI-NEXT: buffer_store_dword v0, off, s[0:3], s32 offset:580 ; 4-byte Folded Spill
+; VI-NEXT: buffer_store_dword v0, off, s[0:3], s32 offset:588 ; 4-byte Folded Spill
; VI-NEXT: v_lshlrev_b32_e32 v0, 8, v55
; VI-NEXT: buffer_store_dword v0, off, s[0:3], s32 offset:500 ; 4-byte Folded Spill
; VI-NEXT: v_lshlrev_b32_e32 v0, 8, v54
@@ -19773,22 +19768,22 @@ define inreg <32 x i32> @bitcast_v128i8_to_v32i32_scalar(<128 x i8> inreg %a, i3
; VI-NEXT: v_lshlrev_b32_e32 v0, 8, v53
; VI-NEXT: buffer_store_dword v0, off, s[0:3], s32 offset:496 ; 4-byte Folded Spill
; VI-NEXT: v_lshlrev_b32_e32 v0, 8, v52
-; VI-NEXT: buffer_store_dword v0, off, s[0:3], s32 offset:608 ; 4-byte Folded Spill
+; VI-NEXT: buffer_store_dword v0, off, s[0:3], s32 offset:616 ; 4-byte Folded Spill
; VI-NEXT: v_lshlrev_b32_e32 v0, 8, v51
; VI-NEXT: buffer_store_dword v0, off, s[0:3], s32 offset:492 ; 4-byte Folded Spill
; VI-NEXT: v_lshlrev_b32_e32 v0, 8, v50
-; VI-NEXT: buffer_store_dword v0, off, s[0:3], s32 offset:600 ; 4-byte Folded Spill
+; VI-NEXT: buffer_store_dword v0, off, s[0:3], s32 offset:612 ; 4-byte Folded Spill
; VI-NEXT: v_lshlrev_b32_e32 v0, 8, v49
-; VI-NEXT: buffer_store_dword v0, off, s[0:3], s32 offset:604 ; 4-byte Folded Spill
+; VI-NEXT: buffer_store_dword v0, off, s[0:3], s32 offset:608 ; 4-byte Folded Spill
; VI-NEXT: v_lshlrev_b32_e32 v0, 8, v48
-; VI-NEXT: buffer_store_dword v0, off, s[0:3], s32 offset:596 ; 4-byte Folded Spill
+; VI-NEXT: buffer_store_dword v0, off, s[0:3], s32 offset:604 ; 4-byte Folded Spill
; VI-NEXT: v_lshlrev_b32_e32 v0, 8, v39
; VI-NEXT: buffer_store_dword v0, off, s[0:3], s32 offset:480 ; 4-byte Folded Spill
; VI-NEXT: s_waitcnt vmcnt(14)
; VI-NEXT: v_lshlrev_b32_e32 v0, 8, v22
-; VI-NEXT: buffer_store_dword v0, off, s[0:3], s32 offset:572 ; 4-byte Folded Spill
+; VI-NEXT: buffer_store_dword v0, off, s[0:3], s32 offset:596 ; 4-byte Folded Spill
; VI-NEXT: v_lshlrev_b32_e32 v0, 8, v31
-; VI-NEXT: buffer_store_dword v0, off, s[0:3], s32 offset:588 ; 4-byte Folded Spill
+; VI-NEXT: buffer_store_dword v0, off, s[0:3], s32 offset:600 ; 4-byte Folded Spill
; VI-NEXT: v_lshlrev_b32_e32 v0, 8, v32
; VI-NEXT: buffer_store_dword v0, off, s[0:3], s32 offset:592 ; 4-byte Folded Spill
; VI-NEXT: v_lshlrev_b32_e32 v0, 8, v33
@@ -19811,23 +19806,24 @@ define inreg <32 x i32> @bitcast_v128i8_to_v32i32_scalar(<128 x i8> inreg %a, i3
; VI-NEXT: v_lshlrev_b32_e32 v31, 8, v38
; VI-NEXT: s_and_b64 s[4:5], vcc, exec
; VI-NEXT: s_waitcnt vmcnt(7)
-; VI-NEXT: v_lshlrev_b32_e32 v26, 8, v0
+; VI-NEXT: v_lshlrev_b32_e32 v22, 8, v0
; VI-NEXT: s_waitcnt vmcnt(6)
; VI-NEXT: v_lshlrev_b32_e32 v32, 8, v1
; VI-NEXT: s_waitcnt vmcnt(5)
-; VI-NEXT: v_lshlrev_b32_e32 v54, 8, v13
+; VI-NEXT: v_lshlrev_b32_e32 v23, 8, v13
+; VI-NEXT: s_waitcnt vmcnt(4)
+; VI-NEXT: v_lshlrev_b32_e32 v0, 8, v3
+; VI-NEXT: buffer_store_dword v0, off, s[0:3], s32 offset:580 ; 4-byte Folded Spill
; VI-NEXT: s_waitcnt vmcnt(4)
-; VI-NEXT: v_lshlrev_b32_e32 v49, 8, v3
-; VI-NEXT: s_waitcnt vmcnt(3)
; VI-NEXT: v_lshlrev_b32_e32 v0, 8, v11
-; VI-NEXT: buffer_store_dword v0, off, s[0:3], s32 offset:576 ; 4-byte Folded Spill
-; VI-NEXT: s_waitcnt vmcnt(3)
+; VI-NEXT: buffer_store_dword v0, off, s[0:3], s32 offset:572 ; 4-byte Folded Spill
+; VI-NEXT: s_waitcnt vmcnt(4)
; VI-NEXT: v_lshlrev_b32_e32 v0, 8, v5
-; VI-NEXT: buffer_store_dword v0, off, s[0:3], s32 offset:612 ; 4-byte Folded Spill
-; VI-NEXT: s_waitcnt vmcnt(3)
+; VI-NEXT: buffer_store_dword v0, off, s[0:3], s32 offset:576 ; 4-byte Folded Spill
+; VI-NEXT: s_waitcnt vmcnt(4)
; VI-NEXT: v_lshlrev_b32_e32 v0, 8, v9
; VI-NEXT: buffer_store_dword v0, off, s[0:3], s32 offset:416 ; 4-byte Folded Spill
-; VI-NEXT: s_waitcnt vmcnt(3)
+; VI-NEXT: s_waitcnt vmcnt(4)
; VI-NEXT: v_lshlrev_b32_e32 v0, 8, v7
; VI-NEXT: buffer_store_dword v0, off, s[0:3], s32 offset:412 ; 4-byte Folded Spill
; VI-NEXT: buffer_load_ushort v0, off, s[0:3], s32 offset:248
@@ -19840,37 +19836,37 @@ define inreg <32 x i32> @bitcast_v128i8_to_v32i32_scalar(<128 x i8> inreg %a, i3
; VI-NEXT: buffer_load_ushort v7, off, s[0:3], s32 offset:304
; VI-NEXT: s_waitcnt vmcnt(7)
; VI-NEXT: v_lshlrev_b32_e32 v0, 8, v0
-; VI-NEXT: buffer_store_dword v0, off, s[0:3], s32 offset:564 ; 4-byte Folded Spill
+; VI-NEXT: buffer_store_dword v0, off, s[0:3], s32 offset:568 ; 4-byte Folded Spill
; VI-NEXT: s_waitcnt vmcnt(7)
; VI-NEXT: v_lshlrev_b32_e32 v48, 8, v1
; VI-NEXT: s_waitcnt vmcnt(6)
; VI-NEXT: v_lshlrev_b32_e32 v27, 8, v13
; VI-NEXT: s_waitcnt vmcnt(4)
; VI-NEXT: v_lshlrev_b32_e32 v0, 8, v11
-; VI-NEXT: buffer_store_dword v0, off, s[0:3], s32 offset:568 ; 4-byte Folded Spill
+; VI-NEXT: buffer_store_dword v0, off, s[0:3], s32 offset:556 ; 4-byte Folded Spill
; VI-NEXT: s_waitcnt vmcnt(4)
; VI-NEXT: v_lshlrev_b32_e32 v0, 8, v5
; VI-NEXT: buffer_store_dword v0, off, s[0:3], s32 offset:404 ; 4-byte Folded Spill
-; VI-NEXT: v_lshlrev_b32_e32 v29, 8, v3
+; VI-NEXT: v_lshlrev_b32_e32 v28, 8, v3
; VI-NEXT: buffer_load_ushort v0, off, s[0:3], s32 offset:312
; VI-NEXT: buffer_load_ushort v1, off, s[0:3], s32 offset:320
; VI-NEXT: buffer_load_ushort v3, off, s[0:3], s32 offset:328
; VI-NEXT: buffer_load_ushort v11, off, s[0:3], s32 offset:4
-; VI-NEXT: buffer_load_ushort v12, off, s[0:3], s32 offset:12
+; VI-NEXT: buffer_load_ushort v16, off, s[0:3], s32 offset:12
; VI-NEXT: buffer_load_ushort v50, off, s[0:3], s32 offset:20
-; VI-NEXT: buffer_load_ushort v56, off, s[0:3], s32 offset:28
+; VI-NEXT: buffer_load_ushort v59, off, s[0:3], s32 offset:28
; VI-NEXT: buffer_load_ushort v13, off, s[0:3], s32 offset:36
; VI-NEXT: s_waitcnt vmcnt(11)
; VI-NEXT: v_lshlrev_b32_e32 v5, 8, v7
-; VI-NEXT: v_lshlrev_b32_e32 v9, 8, v9
+; VI-NEXT: v_lshlrev_b32_e32 v29, 8, v9
; VI-NEXT: s_waitcnt vmcnt(7)
; VI-NEXT: v_lshlrev_b32_e32 v7, 8, v0
; VI-NEXT: s_waitcnt vmcnt(6)
; VI-NEXT: v_lshlrev_b32_e32 v0, 8, v1
; VI-NEXT: buffer_store_dword v0, off, s[0:3], s32 offset:552 ; 4-byte Folded Spill
; VI-NEXT: buffer_load_ushort v18, off, s[0:3], s32 offset:44
-; VI-NEXT: buffer_load_ushort v15, off, s[0:3], s32 offset:52
-; VI-NEXT: buffer_load_ushort v17, off, s[0:3], s32 offset:60
+; VI-NEXT: buffer_load_ushort v17, off, s[0:3], s32 offset:52
+; VI-NEXT: buffer_load_ushort v15, off, s[0:3], s32 offset:60
; VI-NEXT: buffer_load_ushort v39, off, s[0:3], s32 offset:68
; VI-NEXT: buffer_load_ushort v38, off, s[0:3], s32 offset:76
; VI-NEXT: buffer_load_ushort v37, off, s[0:3], s32 offset:84
@@ -19878,23 +19874,23 @@ define inreg <32 x i32> @bitcast_v128i8_to_v32i32_scalar(<128 x i8> inreg %a, i3
; VI-NEXT: buffer_load_ushort v36, off, s[0:3], s32 offset:100
; VI-NEXT: buffer_load_ushort v35, off, s[0:3], s32 offset:108
; VI-NEXT: buffer_load_ushort v33, off, s[0:3], s32 offset:116
-; VI-NEXT: buffer_load_ushort v25, off, s[0:3], s32 offset:124
+; VI-NEXT: buffer_load_ushort v54, off, s[0:3], s32 offset:124
; VI-NEXT: buffer_load_ushort v19, off, s[0:3], s32 offset:132
-; VI-NEXT: buffer_load_ushort v51, off, s[0:3], s32 offset:140
+; VI-NEXT: buffer_load_ushort v49, off, s[0:3], s32 offset:140
; VI-NEXT: buffer_load_ushort v20, off, s[0:3], s32 offset:148
-; VI-NEXT: buffer_load_ushort v28, off, s[0:3], s32 offset:156
+; VI-NEXT: buffer_load_ushort v51, off, s[0:3], s32 offset:156
; VI-NEXT: buffer_load_ushort v21, off, s[0:3], s32 offset:164
; VI-NEXT: buffer_load_ushort v34, off, s[0:3], s32 offset:172
-; VI-NEXT: buffer_load_ushort v22, off, s[0:3], s32 offset:180
-; VI-NEXT: buffer_load_ushort v23, off, s[0:3], s32 offset:188
+; VI-NEXT: buffer_load_ushort v25, off, s[0:3], s32 offset:180
+; VI-NEXT: buffer_load_ushort v30, off, s[0:3], s32 offset:188
; VI-NEXT: buffer_load_ushort v43, off, s[0:3], s32 offset:196
-; VI-NEXT: buffer_load_ushort v30, off, s[0:3], s32 offset:204
-; VI-NEXT: buffer_load_ushort v24, off, s[0:3], s32 offset:212
+; VI-NEXT: buffer_load_ushort v24, off, s[0:3], s32 offset:204
+; VI-NEXT: buffer_load_ushort v26, off, s[0:3], s32 offset:212
; VI-NEXT: buffer_load_ushort v46, off, s[0:3], s32 offset:220
; VI-NEXT: buffer_load_ushort v61, off, s[0:3], s32 offset:228
; VI-NEXT: buffer_load_ushort v47, off, s[0:3], s32 offset:236
-; VI-NEXT: buffer_load_ushort v45, off, s[0:3], s32 offset:244
-; VI-NEXT: buffer_load_ushort v58, off, s[0:3], s32 offset:252
+; VI-NEXT: buffer_load_ushort v58, off, s[0:3], s32 offset:244
+; VI-NEXT: buffer_load_ushort v45, off, s[0:3], s32 offset:252
; VI-NEXT: buffer_load_ushort v44, off, s[0:3], s32 offset:260
; VI-NEXT: buffer_load_ushort v63, off, s[0:3], s32 offset:268
; VI-NEXT: buffer_load_ushort v42, off, s[0:3], s32 offset:276
@@ -19907,57 +19903,57 @@ define inreg <32 x i32> @bitcast_v128i8_to_v32i32_scalar(<128 x i8> inreg %a, i3
; VI-NEXT: s_waitcnt vmcnt(14)
; VI-NEXT: v_lshlrev_b32_e32 v0, 8, v3
; VI-NEXT: buffer_store_dword v5, off, s[0:3], s32 offset:400 ; 4-byte Folded Spill
-; VI-NEXT: buffer_store_dword v9, off, s[0:3], s32 offset:616 ; 4-byte Folded Spill
+; VI-NEXT: buffer_store_dword v29, off, s[0:3], s32 offset:620 ; 4-byte Folded Spill
; VI-NEXT: buffer_store_dword v0, off, s[0:3], s32 offset:408 ; 4-byte Folded Spill
-; VI-NEXT: buffer_store_dword v27, off, s[0:3], s32 offset:620 ; 4-byte Folded Spill
-; VI-NEXT: buffer_store_dword v48, off, s[0:3], s32 offset:624 ; 4-byte Folded Spill
+; VI-NEXT: buffer_store_dword v27, off, s[0:3], s32 offset:624 ; 4-byte Folded Spill
+; VI-NEXT: buffer_store_dword v48, off, s[0:3], s32 offset:628 ; 4-byte Folded Spill
; VI-NEXT: s_waitcnt vmcnt(12)
-; VI-NEXT: buffer_store_dword v63, off, s[0:3], s32 offset:628 ; 4-byte Folded Spill
-; VI-NEXT: buffer_store_dword v58, off, s[0:3], s32 offset:632 ; 4-byte Folded Spill
-; VI-NEXT: buffer_store_dword v47, off, s[0:3], s32 offset:636 ; 4-byte Folded Spill
-; VI-NEXT: buffer_store_dword v31, off, s[0:3], s32 offset:640 ; 4-byte Folded Spill
-; VI-NEXT: buffer_store_dword v46, off, s[0:3], s32 offset:644 ; 4-byte Folded Spill
-; VI-NEXT: buffer_store_dword v52, off, s[0:3], s32 offset:648 ; 4-byte Folded Spill
-; VI-NEXT: buffer_store_dword v2, off, s[0:3], s32 offset:652 ; 4-byte Folded Spill
-; VI-NEXT: buffer_store_dword v32, off, s[0:3], s32 offset:656 ; 4-byte Folded Spill
-; VI-NEXT: buffer_store_dword v4, off, s[0:3], s32 offset:660 ; 4-byte Folded Spill
-; VI-NEXT: buffer_store_dword v24, off, s[0:3], s32 offset:664 ; 4-byte Folded Spill
-; VI-NEXT: buffer_store_dword v30, off, s[0:3], s32 offset:668 ; 4-byte Folded Spill
-; VI-NEXT: buffer_store_dword v43, off, s[0:3], s32 offset:672 ; 4-byte Folded Spill
-; VI-NEXT: buffer_store_dword v23, off, s[0:3], s32 offset:676 ; 4-byte Folded Spill
-; VI-NEXT: buffer_store_dword v6, off, s[0:3], s32 offset:680 ; 4-byte Folded Spill
-; VI-NEXT: buffer_store_dword v16, off, s[0:3], s32 offset:684 ; 4-byte Folded Spill
-; VI-NEXT: buffer_store_dword v22, off, s[0:3], s32 offset:688 ; 4-byte Folded Spill
-; VI-NEXT: buffer_store_dword v34, off, s[0:3], s32 offset:692 ; 4-byte Folded Spill
-; VI-NEXT: buffer_store_dword v10, off, s[0:3], s32 offset:696 ; 4-byte Folded Spill
+; VI-NEXT: buffer_store_dword v63, off, s[0:3], s32 offset:632 ; 4-byte Folded Spill
+; VI-NEXT: buffer_store_dword v45, off, s[0:3], s32 offset:636 ; 4-byte Folded Spill
+; VI-NEXT: buffer_store_dword v47, off, s[0:3], s32 offset:640 ; 4-byte Folded Spill
+; VI-NEXT: buffer_store_dword v31, off, s[0:3], s32 offset:644 ; 4-byte Folded Spill
+; VI-NEXT: buffer_store_dword v46, off, s[0:3], s32 offset:648 ; 4-byte Folded Spill
+; VI-NEXT: buffer_store_dword v52, off, s[0:3], s32 offset:652 ; 4-byte Folded Spill
+; VI-NEXT: buffer_store_dword v2, off, s[0:3], s32 offset:656 ; 4-byte Folded Spill
+; VI-NEXT: buffer_store_dword v32, off, s[0:3], s32 offset:660 ; 4-byte Folded Spill
+; VI-NEXT: buffer_store_dword v4, off, s[0:3], s32 offset:664 ; 4-byte Folded Spill
+; VI-NEXT: buffer_store_dword v26, off, s[0:3], s32 offset:668 ; 4-byte Folded Spill
+; VI-NEXT: buffer_store_dword v24, off, s[0:3], s32 offset:672 ; 4-byte Folded Spill
+; VI-NEXT: buffer_store_dword v43, off, s[0:3], s32 offset:676 ; 4-byte Folded Spill
+; VI-NEXT: buffer_store_dword v30, off, s[0:3], s32 offset:680 ; 4-byte Folded Spill
+; VI-NEXT: buffer_store_dword v6, off, s[0:3], s32 offset:684 ; 4-byte Folded Spill
+; VI-NEXT: buffer_store_dword v12, off, s[0:3], s32 offset:688 ; 4-byte Folded Spill
+; VI-NEXT: buffer_store_dword v25, off, s[0:3], s32 offset:692 ; 4-byte Folded Spill
+; VI-NEXT: buffer_store_dword v34, off, s[0:3], s32 offset:696 ; 4-byte Folded Spill
; VI-NEXT: buffer_store_dword v14, off, s[0:3], s32 offset:700 ; 4-byte Folded Spill
-; VI-NEXT: buffer_store_dword v21, off, s[0:3], s32 offset:704 ; 4-byte Folded Spill
-; VI-NEXT: buffer_store_dword v28, off, s[0:3], s32 offset:708 ; 4-byte Folded Spill
-; VI-NEXT: buffer_store_dword v20, off, s[0:3], s32 offset:712 ; 4-byte Folded Spill
-; VI-NEXT: buffer_store_dword v51, off, s[0:3], s32 offset:716 ; 4-byte Folded Spill
-; VI-NEXT: buffer_store_dword v59, off, s[0:3], s32 offset:720 ; 4-byte Folded Spill
-; VI-NEXT: buffer_store_dword v19, off, s[0:3], s32 offset:724 ; 4-byte Folded Spill
-; VI-NEXT: buffer_store_dword v8, off, s[0:3], s32 offset:728 ; 4-byte Folded Spill
-; VI-NEXT: buffer_store_dword v25, off, s[0:3], s32 offset:732 ; 4-byte Folded Spill
-; VI-NEXT: buffer_store_dword v33, off, s[0:3], s32 offset:736 ; 4-byte Folded Spill
-; VI-NEXT: buffer_store_dword v35, off, s[0:3], s32 offset:740 ; 4-byte Folded Spill
-; VI-NEXT: buffer_store_dword v36, off, s[0:3], s32 offset:744 ; 4-byte Folded Spill
-; VI-NEXT: buffer_store_dword v57, off, s[0:3], s32 offset:748 ; 4-byte Folded Spill
-; VI-NEXT: buffer_store_dword v37, off, s[0:3], s32 offset:752 ; 4-byte Folded Spill
-; VI-NEXT: buffer_store_dword v38, off, s[0:3], s32 offset:756 ; 4-byte Folded Spill
-; VI-NEXT: buffer_store_dword v39, off, s[0:3], s32 offset:760 ; 4-byte Folded Spill
-; VI-NEXT: buffer_store_dword v17, off, s[0:3], s32 offset:764 ; 4-byte Folded Spill
+; VI-NEXT: buffer_store_dword v10, off, s[0:3], s32 offset:704 ; 4-byte Folded Spill
+; VI-NEXT: buffer_store_dword v21, off, s[0:3], s32 offset:708 ; 4-byte Folded Spill
+; VI-NEXT: buffer_store_dword v51, off, s[0:3], s32 offset:712 ; 4-byte Folded Spill
+; VI-NEXT: buffer_store_dword v20, off, s[0:3], s32 offset:716 ; 4-byte Folded Spill
+; VI-NEXT: buffer_store_dword v49, off, s[0:3], s32 offset:720 ; 4-byte Folded Spill
+; VI-NEXT: buffer_store_dword v8, off, s[0:3], s32 offset:724 ; 4-byte Folded Spill
+; VI-NEXT: buffer_store_dword v19, off, s[0:3], s32 offset:728 ; 4-byte Folded Spill
+; VI-NEXT: buffer_store_dword v56, off, s[0:3], s32 offset:732 ; 4-byte Folded Spill
+; VI-NEXT: buffer_store_dword v54, off, s[0:3], s32 offset:736 ; 4-byte Folded Spill
+; VI-NEXT: buffer_store_dword v33, off, s[0:3], s32 offset:740 ; 4-byte Folded Spill
+; VI-NEXT: buffer_store_dword v35, off, s[0:3], s32 offset:744 ; 4-byte Folded Spill
+; VI-NEXT: buffer_store_dword v36, off, s[0:3], s32 offset:748 ; 4-byte Folded Spill
+; VI-NEXT: buffer_store_dword v57, off, s[0:3], s32 offset:752 ; 4-byte Folded Spill
+; VI-NEXT: buffer_store_dword v37, off, s[0:3], s32 offset:756 ; 4-byte Folded Spill
+; VI-NEXT: buffer_store_dword v38, off, s[0:3], s32 offset:760 ; 4-byte Folded Spill
+; VI-NEXT: buffer_store_dword v39, off, s[0:3], s32 offset:764 ; 4-byte Folded Spill
; VI-NEXT: buffer_store_dword v15, off, s[0:3], s32 offset:768 ; 4-byte Folded Spill
-; VI-NEXT: buffer_store_dword v18, off, s[0:3], s32 offset:772 ; 4-byte Folded Spill
-; VI-NEXT: buffer_store_dword v13, off, s[0:3], s32 offset:776 ; 4-byte Folded Spill
-; VI-NEXT: buffer_store_dword v56, off, s[0:3], s32 offset:780 ; 4-byte Folded Spill
-; VI-NEXT: buffer_store_dword v50, off, s[0:3], s32 offset:784 ; 4-byte Folded Spill
-; VI-NEXT: buffer_store_dword v12, off, s[0:3], s32 offset:788 ; 4-byte Folded Spill
-; VI-NEXT: buffer_store_dword v11, off, s[0:3], s32 offset:792 ; 4-byte Folded Spill
+; VI-NEXT: buffer_store_dword v17, off, s[0:3], s32 offset:772 ; 4-byte Folded Spill
+; VI-NEXT: buffer_store_dword v18, off, s[0:3], s32 offset:776 ; 4-byte Folded Spill
+; VI-NEXT: buffer_store_dword v13, off, s[0:3], s32 offset:780 ; 4-byte Folded Spill
+; VI-NEXT: buffer_store_dword v59, off, s[0:3], s32 offset:784 ; 4-byte Folded Spill
+; VI-NEXT: buffer_store_dword v50, off, s[0:3], s32 offset:788 ; 4-byte Folded Spill
+; VI-NEXT: buffer_store_dword v16, off, s[0:3], s32 offset:792 ; 4-byte Folded Spill
+; VI-NEXT: buffer_store_dword v11, off, s[0:3], s32 offset:796 ; 4-byte Folded Spill
; VI-NEXT: s_cbranch_scc0 .LBB15_2
; VI-NEXT: ; %bb.1: ; %cmp.false
-; VI-NEXT: v_or_b32_sdwa v0, v2, v8 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
-; VI-NEXT: v_or_b32_sdwa v1, v4, v59 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; VI-NEXT: v_or_b32_sdwa v0, v2, v56 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; VI-NEXT: v_or_b32_sdwa v1, v4, v8 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
; VI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:488 ; 4-byte Folded Reload
; VI-NEXT: buffer_load_dword v3, off, s[0:3], s32 offset:484 ; 4-byte Folded Reload
; VI-NEXT: v_or_b32_sdwa v4, v0, v1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_0 src1_sel:DWORD
@@ -19972,12 +19968,12 @@ define inreg <32 x i32> @bitcast_v128i8_to_v32i32_scalar(<128 x i8> inreg %a, i3
; VI-NEXT: s_lshl_b32 s7, s23, 8
; VI-NEXT: s_lshl_b32 s8, s27, 8
; VI-NEXT: s_waitcnt vmcnt(3)
-; VI-NEXT: v_or_b32_sdwa v2, v2, v14 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; VI-NEXT: v_or_b32_sdwa v2, v2, v10 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
; VI-NEXT: s_waitcnt vmcnt(2)
-; VI-NEXT: v_or_b32_sdwa v3, v3, v10 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; VI-NEXT: v_or_b32_sdwa v3, v3, v14 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
; VI-NEXT: v_or_b32_sdwa v5, v2, v3 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_0 src1_sel:DWORD
; VI-NEXT: s_waitcnt vmcnt(1)
-; VI-NEXT: v_or_b32_sdwa v0, v0, v16 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; VI-NEXT: v_or_b32_sdwa v0, v0, v12 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
; VI-NEXT: s_waitcnt vmcnt(0)
; VI-NEXT: v_or_b32_sdwa v1, v1, v6 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
; VI-NEXT: v_or_b32_sdwa v6, v0, v1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_0 src1_sel:DWORD
@@ -20006,9 +20002,7 @@ define inreg <32 x i32> @bitcast_v128i8_to_v32i32_scalar(<128 x i8> inreg %a, i3
; VI-NEXT: v_or_b32_sdwa v0, v0, v1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
; VI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:432 ; 4-byte Folded Reload
; VI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:528 ; 4-byte Folded Reload
-; VI-NEXT: buffer_store_dword v29, off, s[0:3], s32 offset:796 ; 4-byte Folded Spill
-; VI-NEXT: v_mov_b32_e32 v29, v9
-; VI-NEXT: s_waitcnt vmcnt(1)
+; VI-NEXT: s_waitcnt vmcnt(0)
; VI-NEXT: v_or_b32_sdwa v1, v1, v2 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
; VI-NEXT: v_or_b32_sdwa v9, v0, v1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_0 src1_sel:DWORD
; VI-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:428 ; 4-byte Folded Reload
@@ -20031,15 +20025,15 @@ define inreg <32 x i32> @bitcast_v128i8_to_v32i32_scalar(<128 x i8> inreg %a, i3
; VI-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:508 ; 4-byte Folded Reload
; VI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:504 ; 4-byte Folded Reload
; VI-NEXT: s_waitcnt vmcnt(1)
-; VI-NEXT: v_or_b32_sdwa v0, v12, v0 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; VI-NEXT: v_or_b32_sdwa v0, v16, v0 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
; VI-NEXT: s_waitcnt vmcnt(0)
; VI-NEXT: v_or_b32_sdwa v1, v50, v1 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
; VI-NEXT: v_or_b32_sdwa v12, v0, v1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_0 src1_sel:DWORD
-; VI-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:580 ; 4-byte Folded Reload
+; VI-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:588 ; 4-byte Folded Reload
; VI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:500 ; 4-byte Folded Reload
; VI-NEXT: s_waitcnt vmcnt(1)
; VI-NEXT: v_mov_b32_e32 v50, v0
-; VI-NEXT: v_or_b32_sdwa v0, v56, v0 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; VI-NEXT: v_or_b32_sdwa v0, v59, v0 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
; VI-NEXT: s_waitcnt vmcnt(0)
; VI-NEXT: v_or_b32_sdwa v1, v13, v1 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
; VI-NEXT: v_or_b32_sdwa v13, v0, v1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_0 src1_sel:DWORD
@@ -20049,18 +20043,18 @@ define inreg <32 x i32> @bitcast_v128i8_to_v32i32_scalar(<128 x i8> inreg %a, i3
; VI-NEXT: v_mov_b32_e32 v59, v0
; VI-NEXT: v_or_b32_sdwa v0, v18, v0 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
; VI-NEXT: s_waitcnt vmcnt(0)
-; VI-NEXT: v_or_b32_sdwa v1, v15, v1 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; VI-NEXT: v_or_b32_sdwa v1, v17, v1 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
; VI-NEXT: v_or_b32_sdwa v14, v0, v1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_0 src1_sel:DWORD
-; VI-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:608 ; 4-byte Folded Reload
+; VI-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:616 ; 4-byte Folded Reload
; VI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:492 ; 4-byte Folded Reload
; VI-NEXT: s_waitcnt vmcnt(1)
; VI-NEXT: v_mov_b32_e32 v56, v0
-; VI-NEXT: v_or_b32_sdwa v0, v17, v0 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; VI-NEXT: v_or_b32_sdwa v0, v15, v0 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
; VI-NEXT: s_waitcnt vmcnt(0)
; VI-NEXT: v_or_b32_sdwa v1, v39, v1 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
; VI-NEXT: v_or_b32_sdwa v15, v0, v1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_0 src1_sel:DWORD
-; VI-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:600 ; 4-byte Folded Reload
-; VI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:604 ; 4-byte Folded Reload
+; VI-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:612 ; 4-byte Folded Reload
+; VI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:608 ; 4-byte Folded Reload
; VI-NEXT: s_waitcnt vmcnt(1)
; VI-NEXT: v_mov_b32_e32 v39, v0
; VI-NEXT: v_or_b32_sdwa v0, v38, v0 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
@@ -20068,7 +20062,7 @@ define inreg <32 x i32> @bitcast_v128i8_to_v32i32_scalar(<128 x i8> inreg %a, i3
; VI-NEXT: v_mov_b32_e32 v38, v1
; VI-NEXT: v_or_b32_sdwa v1, v37, v1 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
; VI-NEXT: v_or_b32_sdwa v16, v0, v1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_0 src1_sel:DWORD
-; VI-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:596 ; 4-byte Folded Reload
+; VI-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:604 ; 4-byte Folded Reload
; VI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:480 ; 4-byte Folded Reload
; VI-NEXT: s_waitcnt vmcnt(1)
; VI-NEXT: v_mov_b32_e32 v37, v0
@@ -20076,8 +20070,8 @@ define inreg <32 x i32> @bitcast_v128i8_to_v32i32_scalar(<128 x i8> inreg %a, i3
; VI-NEXT: s_waitcnt vmcnt(0)
; VI-NEXT: v_or_b32_sdwa v1, v36, v1 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
; VI-NEXT: v_or_b32_sdwa v17, v0, v1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_0 src1_sel:DWORD
-; VI-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:572 ; 4-byte Folded Reload
-; VI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:588 ; 4-byte Folded Reload
+; VI-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:596 ; 4-byte Folded Reload
+; VI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:600 ; 4-byte Folded Reload
; VI-NEXT: s_waitcnt vmcnt(1)
; VI-NEXT: v_mov_b32_e32 v36, v0
; VI-NEXT: v_or_b32_sdwa v0, v35, v0 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
@@ -20089,39 +20083,41 @@ define inreg <32 x i32> @bitcast_v128i8_to_v32i32_scalar(<128 x i8> inreg %a, i3
; VI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:468 ; 4-byte Folded Reload
; VI-NEXT: s_waitcnt vmcnt(1)
; VI-NEXT: v_mov_b32_e32 v33, v0
-; VI-NEXT: v_or_b32_sdwa v0, v25, v0 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; VI-NEXT: v_or_b32_sdwa v0, v54, v0 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
; VI-NEXT: s_waitcnt vmcnt(0)
; VI-NEXT: v_or_b32_sdwa v1, v19, v1 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
; VI-NEXT: v_or_b32_sdwa v19, v0, v1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_0 src1_sel:DWORD
; VI-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:456 ; 4-byte Folded Reload
; VI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:444 ; 4-byte Folded Reload
; VI-NEXT: s_waitcnt vmcnt(1)
-; VI-NEXT: v_or_b32_sdwa v0, v51, v0 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; VI-NEXT: v_or_b32_sdwa v0, v49, v0 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
; VI-NEXT: s_waitcnt vmcnt(0)
; VI-NEXT: v_or_b32_sdwa v1, v20, v1 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
; VI-NEXT: v_or_b32_sdwa v20, v0, v1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_0 src1_sel:DWORD
; VI-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:436 ; 4-byte Folded Reload
; VI-NEXT: v_or_b32_sdwa v1, v21, v52 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
-; VI-NEXT: v_mov_b32_e32 v51, v3
; VI-NEXT: s_waitcnt vmcnt(0)
-; VI-NEXT: v_or_b32_sdwa v0, v28, v0 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; VI-NEXT: v_or_b32_sdwa v0, v51, v0 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
; VI-NEXT: v_or_b32_sdwa v21, v0, v1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_0 src1_sel:DWORD
; VI-NEXT: v_or_b32_sdwa v0, v34, v31 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
-; VI-NEXT: v_or_b32_sdwa v1, v22, v26 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; VI-NEXT: v_or_b32_sdwa v1, v25, v22 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; VI-NEXT: v_mov_b32_e32 v34, v22
; VI-NEXT: v_or_b32_sdwa v22, v0, v1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_0 src1_sel:DWORD
-; VI-NEXT: v_or_b32_sdwa v0, v23, v32 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
-; VI-NEXT: v_or_b32_sdwa v1, v43, v54 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; VI-NEXT: v_or_b32_sdwa v0, v30, v32 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; VI-NEXT: v_or_b32_sdwa v1, v43, v23 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; VI-NEXT: v_mov_b32_e32 v32, v23
; VI-NEXT: v_or_b32_sdwa v23, v0, v1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_0 src1_sel:DWORD
-; VI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:576 ; 4-byte Folded Reload
-; VI-NEXT: v_mov_b32_e32 v43, v49
-; VI-NEXT: v_or_b32_sdwa v0, v30, v49 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
-; VI-NEXT: v_mov_b32_e32 v32, v54
-; VI-NEXT: v_mov_b32_e32 v34, v26
+; VI-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:580 ; 4-byte Folded Reload
+; VI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:572 ; 4-byte Folded Reload
+; VI-NEXT: v_mov_b32_e32 v51, v3
+; VI-NEXT: s_waitcnt vmcnt(1)
+; VI-NEXT: v_mov_b32_e32 v43, v0
+; VI-NEXT: v_or_b32_sdwa v0, v24, v0 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
; VI-NEXT: s_waitcnt vmcnt(0)
; VI-NEXT: v_mov_b32_e32 v49, v1
-; VI-NEXT: v_or_b32_sdwa v1, v24, v1 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; VI-NEXT: v_or_b32_sdwa v1, v26, v1 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
; VI-NEXT: v_or_b32_sdwa v24, v0, v1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_0 src1_sel:DWORD
-; VI-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:612 ; 4-byte Folded Reload
+; VI-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:576 ; 4-byte Folded Reload
; VI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:416 ; 4-byte Folded Reload
; VI-NEXT: s_waitcnt vmcnt(1)
; VI-NEXT: v_mov_b32_e32 v54, v0
@@ -20131,28 +20127,26 @@ define inreg <32 x i32> @bitcast_v128i8_to_v32i32_scalar(<128 x i8> inreg %a, i3
; VI-NEXT: v_mov_b32_e32 v46, v61
; VI-NEXT: v_or_b32_sdwa v25, v0, v1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_0 src1_sel:DWORD
; VI-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:412 ; 4-byte Folded Reload
-; VI-NEXT: buffer_load_dword v61, off, s[0:3], s32 offset:564 ; 4-byte Folded Reload
+; VI-NEXT: buffer_load_dword v61, off, s[0:3], s32 offset:568 ; 4-byte Folded Reload
; VI-NEXT: s_waitcnt vmcnt(1)
; VI-NEXT: v_or_b32_sdwa v0, v47, v0 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
; VI-NEXT: s_waitcnt vmcnt(0)
-; VI-NEXT: v_or_b32_sdwa v1, v45, v61 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; VI-NEXT: v_or_b32_sdwa v1, v58, v61 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
; VI-NEXT: v_or_b32_sdwa v26, v0, v1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_0 src1_sel:DWORD
-; VI-NEXT: v_or_b32_sdwa v0, v58, v48 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; VI-NEXT: v_or_b32_sdwa v0, v45, v48 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
; VI-NEXT: v_or_b32_sdwa v1, v44, v27 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
-; VI-NEXT: v_mov_b32_e32 v47, v45
; VI-NEXT: v_or_b32_sdwa v27, v0, v1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_0 src1_sel:DWORD
-; VI-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:796 ; 4-byte Folded Reload
-; VI-NEXT: buffer_load_dword v45, off, s[0:3], s32 offset:568 ; 4-byte Folded Reload
-; VI-NEXT: v_mov_b32_e32 v58, v44
-; VI-NEXT: s_waitcnt vmcnt(1)
-; VI-NEXT: v_mov_b32_e32 v48, v0
-; VI-NEXT: v_or_b32_sdwa v0, v63, v0 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; VI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:556 ; 4-byte Folded Reload
+; VI-NEXT: v_or_b32_sdwa v0, v63, v28 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; VI-NEXT: v_mov_b32_e32 v48, v28
+; VI-NEXT: v_mov_b32_e32 v47, v58
+; VI-NEXT: v_mov_b32_e32 v45, v44
+; VI-NEXT: v_mov_b32_e32 v63, v42
; VI-NEXT: s_waitcnt vmcnt(0)
-; VI-NEXT: v_or_b32_sdwa v1, v42, v45 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; VI-NEXT: v_or_b32_sdwa v1, v42, v1 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
; VI-NEXT: v_or_b32_sdwa v28, v0, v1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_0 src1_sel:DWORD
; VI-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:404 ; 4-byte Folded Reload
; VI-NEXT: v_or_b32_sdwa v1, v40, v29 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
-; VI-NEXT: v_mov_b32_e32 v63, v42
; VI-NEXT: s_waitcnt vmcnt(0)
; VI-NEXT: v_or_b32_sdwa v0, v41, v0 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
; VI-NEXT: v_or_b32_sdwa v29, v0, v1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_0 src1_sel:DWORD
@@ -20168,8 +20162,8 @@ define inreg <32 x i32> @bitcast_v128i8_to_v32i32_scalar(<128 x i8> inreg %a, i3
; VI-NEXT: s_waitcnt vmcnt(0)
; VI-NEXT: v_or_b32_sdwa v1, v53, v1 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
; VI-NEXT: v_or_b32_sdwa v31, v0, v1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_0 src1_sel:DWORD
-; VI-NEXT: buffer_load_dword v52, off, s[0:3], s32 offset:556 ; 4-byte Folded Reload
-; VI-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:560 ; 4-byte Folded Reload
+; VI-NEXT: buffer_load_dword v52, off, s[0:3], s32 offset:560 ; 4-byte Folded Reload
+; VI-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:564 ; 4-byte Folded Reload
; VI-NEXT: s_waitcnt vmcnt(0)
; VI-NEXT: v_mov_b32_e32 v57, v0
; VI-NEXT: v_or_b32_sdwa v0, v52, v0 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
@@ -20203,44 +20197,43 @@ define inreg <32 x i32> @bitcast_v128i8_to_v32i32_scalar(<128 x i8> inreg %a, i3
; VI-NEXT: s_mov_b64 s[4:5], 0
; VI-NEXT: s_branch .LBB15_3
; VI-NEXT: .LBB15_2:
-; VI-NEXT: buffer_load_dword v57, off, s[0:3], s32 offset:560 ; 4-byte Folded Reload
-; VI-NEXT: buffer_load_dword v50, off, s[0:3], s32 offset:580 ; 4-byte Folded Reload
+; VI-NEXT: buffer_load_dword v57, off, s[0:3], s32 offset:564 ; 4-byte Folded Reload
+; VI-NEXT: buffer_load_dword v50, off, s[0:3], s32 offset:588 ; 4-byte Folded Reload
; VI-NEXT: buffer_load_dword v59, off, s[0:3], s32 offset:584 ; 4-byte Folded Reload
-; VI-NEXT: buffer_load_dword v56, off, s[0:3], s32 offset:608 ; 4-byte Folded Reload
-; VI-NEXT: buffer_load_dword v39, off, s[0:3], s32 offset:600 ; 4-byte Folded Reload
-; VI-NEXT: buffer_load_dword v38, off, s[0:3], s32 offset:604 ; 4-byte Folded Reload
-; VI-NEXT: buffer_load_dword v37, off, s[0:3], s32 offset:596 ; 4-byte Folded Reload
-; VI-NEXT: buffer_load_dword v36, off, s[0:3], s32 offset:572 ; 4-byte Folded Reload
-; VI-NEXT: buffer_load_dword v35, off, s[0:3], s32 offset:588 ; 4-byte Folded Reload
+; VI-NEXT: buffer_load_dword v56, off, s[0:3], s32 offset:616 ; 4-byte Folded Reload
+; VI-NEXT: buffer_load_dword v39, off, s[0:3], s32 offset:612 ; 4-byte Folded Reload
+; VI-NEXT: buffer_load_dword v38, off, s[0:3], s32 offset:608 ; 4-byte Folded Reload
+; VI-NEXT: buffer_load_dword v37, off, s[0:3], s32 offset:604 ; 4-byte Folded Reload
+; VI-NEXT: buffer_load_dword v36, off, s[0:3], s32 offset:596 ; 4-byte Folded Reload
+; VI-NEXT: buffer_load_dword v35, off, s[0:3], s32 offset:600 ; 4-byte Folded Reload
; VI-NEXT: buffer_load_dword v33, off, s[0:3], s32 offset:592 ; 4-byte Folded Reload
-; VI-NEXT: v_mov_b32_e32 v32, v54
-; VI-NEXT: v_mov_b32_e32 v43, v49
+; VI-NEXT: buffer_load_dword v43, off, s[0:3], s32 offset:580 ; 4-byte Folded Reload
+; VI-NEXT: buffer_load_dword v49, off, s[0:3], s32 offset:572 ; 4-byte Folded Reload
; VI-NEXT: v_mov_b32_e32 v46, v61
-; VI-NEXT: v_mov_b32_e32 v47, v45
-; VI-NEXT: buffer_load_dword v54, off, s[0:3], s32 offset:612 ; 4-byte Folded Reload
-; VI-NEXT: buffer_load_dword v49, off, s[0:3], s32 offset:576 ; 4-byte Folded Reload
-; VI-NEXT: buffer_load_dword v45, off, s[0:3], s32 offset:568 ; 4-byte Folded Reload
-; VI-NEXT: buffer_load_dword v52, off, s[0:3], s32 offset:556 ; 4-byte Folded Reload
-; VI-NEXT: buffer_load_dword v61, off, s[0:3], s32 offset:564 ; 4-byte Folded Reload
-; VI-NEXT: v_mov_b32_e32 v34, v26
-; VI-NEXT: v_mov_b32_e32 v58, v44
+; VI-NEXT: buffer_load_dword v54, off, s[0:3], s32 offset:576 ; 4-byte Folded Reload
+; VI-NEXT: buffer_load_dword v52, off, s[0:3], s32 offset:560 ; 4-byte Folded Reload
+; VI-NEXT: buffer_load_dword v61, off, s[0:3], s32 offset:568 ; 4-byte Folded Reload
+; VI-NEXT: v_mov_b32_e32 v34, v22
+; VI-NEXT: v_mov_b32_e32 v32, v23
+; VI-NEXT: v_mov_b32_e32 v47, v58
+; VI-NEXT: v_mov_b32_e32 v45, v44
; VI-NEXT: s_waitcnt vmcnt(14)
; VI-NEXT: v_mov_b32_e32 v63, v42
; VI-NEXT: v_mov_b32_e32 v51, v7
-; VI-NEXT: v_mov_b32_e32 v48, v29
+; VI-NEXT: v_mov_b32_e32 v48, v28
; VI-NEXT: s_mov_b64 s[4:5], -1
; VI-NEXT: ; implicit-def: $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7_vgpr8_vgpr9_vgpr10_vgpr11_vgpr12_vgpr13_vgpr14_vgpr15_vgpr16_vgpr17_vgpr18_vgpr19_vgpr20_vgpr21_vgpr22_vgpr23_vgpr24_vgpr25_vgpr26_vgpr27_vgpr28_vgpr29_vgpr30_vgpr31
; VI-NEXT: .LBB15_3: ; %Flow
; VI-NEXT: s_andn2_b64 vcc, exec, s[4:5]
-; VI-NEXT: v_mov_b32_e32 v44, v47
-; VI-NEXT: v_mov_b32_e32 v47, v46
+; VI-NEXT: v_mov_b32_e32 v42, v45
+; VI-NEXT: v_mov_b32_e32 v45, v46
; VI-NEXT: s_waitcnt vmcnt(3)
; VI-NEXT: v_mov_b32_e32 v46, v49
; VI-NEXT: s_cbranch_vccnz .LBB15_5
; VI-NEXT: ; %bb.4: ; %cmp.true
-; VI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:652 ; 4-byte Folded Reload
-; VI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:728 ; 4-byte Folded Reload
-; VI-NEXT: buffer_load_dword v3, off, s[0:3], s32 offset:720 ; 4-byte Folded Reload
+; VI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:656 ; 4-byte Folded Reload
+; VI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:732 ; 4-byte Folded Reload
+; VI-NEXT: buffer_load_dword v3, off, s[0:3], s32 offset:724 ; 4-byte Folded Reload
; VI-NEXT: s_add_i32 s28, s28, 3
; VI-NEXT: s_and_b32 s4, s28, 0xff
; VI-NEXT: s_lshl_b32 s5, s29, 8
@@ -20294,7 +20287,7 @@ define inreg <32 x i32> @bitcast_v128i8_to_v32i32_scalar(<128 x i8> inreg %a, i3
; VI-NEXT: v_add_u32_e32 v1, vcc, 3, v1
; VI-NEXT: s_waitcnt vmcnt(1)
; VI-NEXT: v_or_b32_sdwa v1, v2, v1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:BYTE_0
-; VI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:660 ; 4-byte Folded Reload
+; VI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:664 ; 4-byte Folded Reload
; VI-NEXT: v_add_u32_e32 v1, vcc, 0x300, v1
; VI-NEXT: s_waitcnt vmcnt(0)
; VI-NEXT: v_add_u32_e32 v2, vcc, 3, v2
@@ -20303,8 +20296,8 @@ define inreg <32 x i32> @bitcast_v128i8_to_v32i32_scalar(<128 x i8> inreg %a, i3
; VI-NEXT: v_add_u32_e32 v3, vcc, 0x3000000, v0
; VI-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:488 ; 4-byte Folded Reload
; VI-NEXT: v_add_u32_e32 v4, vcc, 0x3000000, v1
-; VI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:700 ; 4-byte Folded Reload
-; VI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:696 ; 4-byte Folded Reload
+; VI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:704 ; 4-byte Folded Reload
+; VI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:700 ; 4-byte Folded Reload
; VI-NEXT: s_waitcnt vmcnt(2)
; VI-NEXT: v_add_u32_e32 v0, vcc, 3, v0
; VI-NEXT: s_waitcnt vmcnt(1)
@@ -20317,8 +20310,8 @@ define inreg <32 x i32> @bitcast_v128i8_to_v32i32_scalar(<128 x i8> inreg %a, i3
; VI-NEXT: v_or_b32_sdwa v0, v1, v0 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:WORD_0
; VI-NEXT: v_add_u32_e32 v5, vcc, 0x3000000, v0
; VI-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:476 ; 4-byte Folded Reload
-; VI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:684 ; 4-byte Folded Reload
-; VI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:680 ; 4-byte Folded Reload
+; VI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:688 ; 4-byte Folded Reload
+; VI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:684 ; 4-byte Folded Reload
; VI-NEXT: s_waitcnt vmcnt(2)
; VI-NEXT: v_add_u32_e32 v0, vcc, 3, v0
; VI-NEXT: s_waitcnt vmcnt(1)
@@ -20393,29 +20386,29 @@ define inreg <32 x i32> @bitcast_v128i8_to_v32i32_scalar(<128 x i8> inreg %a, i3
; VI-NEXT: v_add_u32_e32 v0, vcc, 3, v0
; VI-NEXT: s_waitcnt vmcnt(1)
; VI-NEXT: v_or_b32_sdwa v0, v1, v0 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:BYTE_0
-; VI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:792 ; 4-byte Folded Reload
+; VI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:796 ; 4-byte Folded Reload
; VI-NEXT: v_add_u32_e32 v0, vcc, 0x300, v0
; VI-NEXT: s_waitcnt vmcnt(0)
; VI-NEXT: v_add_u32_e32 v1, vcc, 3, v1
; VI-NEXT: v_or_b32_sdwa v1, v2, v1 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:BYTE_0
; VI-NEXT: v_or_b32_sdwa v0, v1, v0 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:WORD_0
; VI-NEXT: v_add_u32_e32 v11, vcc, 0x3000000, v0
-; VI-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:788 ; 4-byte Folded Reload
+; VI-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:792 ; 4-byte Folded Reload
; VI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:508 ; 4-byte Folded Reload
; VI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:504 ; 4-byte Folded Reload
; VI-NEXT: s_waitcnt vmcnt(2)
; VI-NEXT: v_add_u32_e32 v0, vcc, 3, v0
; VI-NEXT: s_waitcnt vmcnt(1)
; VI-NEXT: v_or_b32_sdwa v0, v1, v0 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:BYTE_0
-; VI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:784 ; 4-byte Folded Reload
+; VI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:788 ; 4-byte Folded Reload
; VI-NEXT: v_add_u32_e32 v0, vcc, 0x300, v0
; VI-NEXT: s_waitcnt vmcnt(0)
; VI-NEXT: v_add_u32_e32 v1, vcc, 3, v1
; VI-NEXT: v_or_b32_sdwa v1, v2, v1 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:BYTE_0
; VI-NEXT: v_or_b32_sdwa v0, v1, v0 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:WORD_0
; VI-NEXT: v_add_u32_e32 v12, vcc, 0x3000000, v0
-; VI-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:780 ; 4-byte Folded Reload
-; VI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:776 ; 4-byte Folded Reload
+; VI-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:784 ; 4-byte Folded Reload
+; VI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:780 ; 4-byte Folded Reload
; VI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:500 ; 4-byte Folded Reload
; VI-NEXT: s_waitcnt vmcnt(2)
; VI-NEXT: v_add_u32_e32 v0, vcc, 3, v0
@@ -20427,8 +20420,8 @@ define inreg <32 x i32> @bitcast_v128i8_to_v32i32_scalar(<128 x i8> inreg %a, i3
; VI-NEXT: v_add_u32_e32 v0, vcc, 0x300, v0
; VI-NEXT: v_or_b32_sdwa v0, v1, v0 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:WORD_0
; VI-NEXT: v_add_u32_e32 v13, vcc, 0x3000000, v0
-; VI-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:772 ; 4-byte Folded Reload
-; VI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:768 ; 4-byte Folded Reload
+; VI-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:776 ; 4-byte Folded Reload
+; VI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:772 ; 4-byte Folded Reload
; VI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:496 ; 4-byte Folded Reload
; VI-NEXT: s_waitcnt vmcnt(2)
; VI-NEXT: v_add_u32_e32 v0, vcc, 3, v0
@@ -20440,8 +20433,8 @@ define inreg <32 x i32> @bitcast_v128i8_to_v32i32_scalar(<128 x i8> inreg %a, i3
; VI-NEXT: v_add_u32_e32 v0, vcc, 0x300, v0
; VI-NEXT: v_or_b32_sdwa v0, v1, v0 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:WORD_0
; VI-NEXT: v_add_u32_e32 v14, vcc, 0x3000000, v0
-; VI-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:764 ; 4-byte Folded Reload
-; VI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:760 ; 4-byte Folded Reload
+; VI-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:768 ; 4-byte Folded Reload
+; VI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:764 ; 4-byte Folded Reload
; VI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:492 ; 4-byte Folded Reload
; VI-NEXT: s_waitcnt vmcnt(2)
; VI-NEXT: v_add_u32_e32 v0, vcc, 3, v0
@@ -20453,8 +20446,8 @@ define inreg <32 x i32> @bitcast_v128i8_to_v32i32_scalar(<128 x i8> inreg %a, i3
; VI-NEXT: v_add_u32_e32 v0, vcc, 0x300, v0
; VI-NEXT: v_or_b32_sdwa v0, v1, v0 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:WORD_0
; VI-NEXT: v_add_u32_e32 v15, vcc, 0x3000000, v0
-; VI-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:756 ; 4-byte Folded Reload
-; VI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:752 ; 4-byte Folded Reload
+; VI-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:760 ; 4-byte Folded Reload
+; VI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:756 ; 4-byte Folded Reload
; VI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:480 ; 4-byte Folded Reload
; VI-NEXT: s_waitcnt vmcnt(2)
; VI-NEXT: v_add_u32_e32 v0, vcc, 3, v0
@@ -20465,8 +20458,8 @@ define inreg <32 x i32> @bitcast_v128i8_to_v32i32_scalar(<128 x i8> inreg %a, i3
; VI-NEXT: v_add_u32_e32 v0, vcc, 0x300, v0
; VI-NEXT: v_or_b32_sdwa v0, v1, v0 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:WORD_0
; VI-NEXT: v_add_u32_e32 v16, vcc, 0x3000000, v0
-; VI-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:748 ; 4-byte Folded Reload
-; VI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:744 ; 4-byte Folded Reload
+; VI-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:752 ; 4-byte Folded Reload
+; VI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:748 ; 4-byte Folded Reload
; VI-NEXT: s_waitcnt vmcnt(1)
; VI-NEXT: v_add_u32_e32 v0, vcc, 3, v0
; VI-NEXT: v_or_b32_sdwa v0, v37, v0 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:BYTE_0
@@ -20476,8 +20469,8 @@ define inreg <32 x i32> @bitcast_v128i8_to_v32i32_scalar(<128 x i8> inreg %a, i3
; VI-NEXT: v_add_u32_e32 v0, vcc, 0x300, v0
; VI-NEXT: v_or_b32_sdwa v0, v1, v0 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:WORD_0
; VI-NEXT: v_add_u32_e32 v17, vcc, 0x3000000, v0
-; VI-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:740 ; 4-byte Folded Reload
-; VI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:736 ; 4-byte Folded Reload
+; VI-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:744 ; 4-byte Folded Reload
+; VI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:740 ; 4-byte Folded Reload
; VI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:468 ; 4-byte Folded Reload
; VI-NEXT: s_waitcnt vmcnt(2)
; VI-NEXT: v_add_u32_e32 v0, vcc, 3, v0
@@ -20488,8 +20481,8 @@ define inreg <32 x i32> @bitcast_v128i8_to_v32i32_scalar(<128 x i8> inreg %a, i3
; VI-NEXT: v_add_u32_e32 v0, vcc, 0x300, v0
; VI-NEXT: v_or_b32_sdwa v0, v1, v0 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:WORD_0
; VI-NEXT: v_add_u32_e32 v18, vcc, 0x3000000, v0
-; VI-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:732 ; 4-byte Folded Reload
-; VI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:724 ; 4-byte Folded Reload
+; VI-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:736 ; 4-byte Folded Reload
+; VI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:728 ; 4-byte Folded Reload
; VI-NEXT: s_waitcnt vmcnt(1)
; VI-NEXT: v_add_u32_e32 v0, vcc, 3, v0
; VI-NEXT: v_or_b32_sdwa v0, v33, v0 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:BYTE_0
@@ -20499,63 +20492,63 @@ define inreg <32 x i32> @bitcast_v128i8_to_v32i32_scalar(<128 x i8> inreg %a, i3
; VI-NEXT: v_add_u32_e32 v0, vcc, 0x300, v0
; VI-NEXT: v_or_b32_sdwa v0, v1, v0 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:WORD_0
; VI-NEXT: v_add_u32_e32 v19, vcc, 0x3000000, v0
-; VI-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:716 ; 4-byte Folded Reload
+; VI-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:720 ; 4-byte Folded Reload
; VI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:456 ; 4-byte Folded Reload
; VI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:444 ; 4-byte Folded Reload
; VI-NEXT: s_waitcnt vmcnt(2)
; VI-NEXT: v_add_u32_e32 v0, vcc, 3, v0
; VI-NEXT: s_waitcnt vmcnt(1)
; VI-NEXT: v_or_b32_sdwa v0, v1, v0 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:BYTE_0
-; VI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:712 ; 4-byte Folded Reload
+; VI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:716 ; 4-byte Folded Reload
; VI-NEXT: v_add_u32_e32 v0, vcc, 0x300, v0
; VI-NEXT: s_waitcnt vmcnt(0)
; VI-NEXT: v_add_u32_e32 v1, vcc, 3, v1
; VI-NEXT: v_or_b32_sdwa v1, v2, v1 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:BYTE_0
; VI-NEXT: v_or_b32_sdwa v0, v1, v0 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:WORD_0
; VI-NEXT: v_add_u32_e32 v20, vcc, 0x3000000, v0
-; VI-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:708 ; 4-byte Folded Reload
+; VI-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:712 ; 4-byte Folded Reload
; VI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:436 ; 4-byte Folded Reload
-; VI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:648 ; 4-byte Folded Reload
+; VI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:652 ; 4-byte Folded Reload
; VI-NEXT: s_waitcnt vmcnt(2)
; VI-NEXT: v_add_u32_e32 v0, vcc, 3, v0
; VI-NEXT: s_waitcnt vmcnt(1)
; VI-NEXT: v_or_b32_sdwa v0, v1, v0 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:BYTE_0
-; VI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:704 ; 4-byte Folded Reload
+; VI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:708 ; 4-byte Folded Reload
; VI-NEXT: v_add_u32_e32 v0, vcc, 0x300, v0
; VI-NEXT: s_waitcnt vmcnt(0)
; VI-NEXT: v_add_u32_e32 v1, vcc, 3, v1
; VI-NEXT: v_or_b32_sdwa v1, v2, v1 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:BYTE_0
; VI-NEXT: v_or_b32_sdwa v0, v1, v0 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:WORD_0
; VI-NEXT: v_add_u32_e32 v21, vcc, 0x3000000, v0
-; VI-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:692 ; 4-byte Folded Reload
-; VI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:640 ; 4-byte Folded Reload
+; VI-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:696 ; 4-byte Folded Reload
+; VI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:644 ; 4-byte Folded Reload
; VI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:416 ; 4-byte Folded Reload
; VI-NEXT: s_waitcnt vmcnt(2)
; VI-NEXT: v_add_u32_e32 v0, vcc, 3, v0
; VI-NEXT: s_waitcnt vmcnt(1)
; VI-NEXT: v_or_b32_sdwa v0, v1, v0 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:BYTE_0
-; VI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:688 ; 4-byte Folded Reload
+; VI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:692 ; 4-byte Folded Reload
; VI-NEXT: v_add_u32_e32 v0, vcc, 0x300, v0
; VI-NEXT: s_waitcnt vmcnt(0)
; VI-NEXT: v_add_u32_e32 v1, vcc, 3, v1
; VI-NEXT: v_or_b32_sdwa v1, v34, v1 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:BYTE_0
; VI-NEXT: v_or_b32_sdwa v0, v1, v0 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:WORD_0
; VI-NEXT: v_add_u32_e32 v22, vcc, 0x3000000, v0
-; VI-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:676 ; 4-byte Folded Reload
-; VI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:656 ; 4-byte Folded Reload
+; VI-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:680 ; 4-byte Folded Reload
+; VI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:660 ; 4-byte Folded Reload
; VI-NEXT: s_waitcnt vmcnt(1)
; VI-NEXT: v_add_u32_e32 v0, vcc, 3, v0
; VI-NEXT: s_waitcnt vmcnt(0)
; VI-NEXT: v_or_b32_sdwa v0, v1, v0 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:BYTE_0
-; VI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:672 ; 4-byte Folded Reload
+; VI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:676 ; 4-byte Folded Reload
; VI-NEXT: v_add_u32_e32 v0, vcc, 0x300, v0
; VI-NEXT: s_waitcnt vmcnt(0)
; VI-NEXT: v_add_u32_e32 v1, vcc, 3, v1
; VI-NEXT: v_or_b32_sdwa v1, v32, v1 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:BYTE_0
; VI-NEXT: v_or_b32_sdwa v0, v1, v0 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:WORD_0
; VI-NEXT: v_add_u32_e32 v23, vcc, 0x3000000, v0
-; VI-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:668 ; 4-byte Folded Reload
-; VI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:664 ; 4-byte Folded Reload
+; VI-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:672 ; 4-byte Folded Reload
+; VI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:668 ; 4-byte Folded Reload
; VI-NEXT: s_waitcnt vmcnt(1)
; VI-NEXT: v_add_u32_e32 v0, vcc, 3, v0
; VI-NEXT: v_or_b32_sdwa v0, v43, v0 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:BYTE_0
@@ -20565,54 +20558,57 @@ define inreg <32 x i32> @bitcast_v128i8_to_v32i32_scalar(<128 x i8> inreg %a, i3
; VI-NEXT: v_add_u32_e32 v0, vcc, 0x300, v0
; VI-NEXT: v_or_b32_sdwa v0, v1, v0 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:WORD_0
; VI-NEXT: v_add_u32_e32 v24, vcc, 0x3000000, v0
-; VI-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:644 ; 4-byte Folded Reload
-; VI-NEXT: v_add_u32_e32 v1, vcc, 3, v47
+; VI-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:648 ; 4-byte Folded Reload
+; VI-NEXT: v_add_u32_e32 v1, vcc, 3, v45
; VI-NEXT: v_or_b32_sdwa v1, v2, v1 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:BYTE_0
-; VI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:620 ; 4-byte Folded Reload
+; VI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:624 ; 4-byte Folded Reload
; VI-NEXT: s_waitcnt vmcnt(1)
; VI-NEXT: v_add_u32_e32 v0, vcc, 3, v0
; VI-NEXT: v_or_b32_sdwa v0, v54, v0 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:BYTE_0
; VI-NEXT: v_add_u32_e32 v0, vcc, 0x300, v0
; VI-NEXT: v_or_b32_sdwa v0, v1, v0 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:WORD_0
; VI-NEXT: v_add_u32_e32 v25, vcc, 0x3000000, v0
-; VI-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:636 ; 4-byte Folded Reload
+; VI-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:640 ; 4-byte Folded Reload
; VI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:412 ; 4-byte Folded Reload
; VI-NEXT: s_waitcnt vmcnt(1)
; VI-NEXT: v_add_u32_e32 v0, vcc, 3, v0
; VI-NEXT: s_waitcnt vmcnt(0)
; VI-NEXT: v_or_b32_sdwa v0, v1, v0 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:BYTE_0
-; VI-NEXT: v_add_u32_e32 v1, vcc, 3, v44
+; VI-NEXT: v_add_u32_e32 v1, vcc, 3, v47
; VI-NEXT: v_add_u32_e32 v0, vcc, 0x300, v0
; VI-NEXT: v_or_b32_sdwa v1, v61, v1 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:BYTE_0
; VI-NEXT: v_or_b32_sdwa v0, v1, v0 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:WORD_0
; VI-NEXT: v_add_u32_e32 v26, vcc, 0x3000000, v0
-; VI-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:632 ; 4-byte Folded Reload
-; VI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:624 ; 4-byte Folded Reload
+; VI-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:636 ; 4-byte Folded Reload
+; VI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:628 ; 4-byte Folded Reload
; VI-NEXT: s_waitcnt vmcnt(1)
; VI-NEXT: v_add_u32_e32 v0, vcc, 3, v0
; VI-NEXT: s_waitcnt vmcnt(0)
; VI-NEXT: v_or_b32_sdwa v0, v1, v0 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:BYTE_0
-; VI-NEXT: v_add_u32_e32 v1, vcc, 3, v58
+; VI-NEXT: v_add_u32_e32 v1, vcc, 3, v42
; VI-NEXT: v_add_u32_e32 v0, vcc, 0x300, v0
; VI-NEXT: v_or_b32_sdwa v1, v2, v1 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:BYTE_0
; VI-NEXT: v_or_b32_sdwa v0, v1, v0 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:WORD_0
; VI-NEXT: v_add_u32_e32 v27, vcc, 0x3000000, v0
-; VI-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:628 ; 4-byte Folded Reload
+; VI-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:632 ; 4-byte Folded Reload
+; VI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:556 ; 4-byte Folded Reload
; VI-NEXT: v_add_u32_e32 v1, vcc, 3, v63
-; VI-NEXT: v_or_b32_sdwa v1, v45, v1 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:BYTE_0
-; VI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:616 ; 4-byte Folded Reload
; VI-NEXT: s_waitcnt vmcnt(1)
; VI-NEXT: v_add_u32_e32 v0, vcc, 3, v0
; VI-NEXT: v_or_b32_sdwa v0, v48, v0 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:BYTE_0
; VI-NEXT: v_add_u32_e32 v0, vcc, 0x300, v0
+; VI-NEXT: s_waitcnt vmcnt(0)
+; VI-NEXT: v_or_b32_sdwa v1, v2, v1 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:BYTE_0
; VI-NEXT: v_or_b32_sdwa v0, v1, v0 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:WORD_0
; VI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:404 ; 4-byte Folded Reload
+; VI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:620 ; 4-byte Folded Reload
; VI-NEXT: v_add_u32_e32 v28, vcc, 0x3000000, v0
; VI-NEXT: v_add_u32_e32 v0, vcc, 3, v41
-; VI-NEXT: s_waitcnt vmcnt(0)
+; VI-NEXT: s_waitcnt vmcnt(1)
; VI-NEXT: v_or_b32_sdwa v0, v1, v0 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:BYTE_0
; VI-NEXT: v_add_u32_e32 v1, vcc, 3, v40
; VI-NEXT: v_add_u32_e32 v0, vcc, 0x300, v0
+; VI-NEXT: s_waitcnt vmcnt(0)
; VI-NEXT: v_or_b32_sdwa v1, v2, v1 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:BYTE_0
; VI-NEXT: v_or_b32_sdwa v0, v1, v0 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:WORD_0
; VI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:400 ; 4-byte Folded Reload
@@ -20897,7 +20893,7 @@ define inreg <32 x i32> @bitcast_v128i8_to_v32i32_scalar(<128 x i8> inreg %a, i3
; GFX9-NEXT: buffer_load_ushort v31, off, s[0:3], s32 offset:164
; GFX9-NEXT: buffer_load_ushort v23, off, s[0:3], s32 offset:172
; GFX9-NEXT: buffer_load_ushort v50, off, s[0:3], s32 offset:180
-; GFX9-NEXT: buffer_load_ushort v44, off, s[0:3], s32 offset:188
+; GFX9-NEXT: buffer_load_ushort v40, off, s[0:3], s32 offset:188
; GFX9-NEXT: buffer_load_ushort v37, off, s[0:3], s32 offset:196
; GFX9-NEXT: buffer_load_ushort v52, off, s[0:3], s32 offset:204
; GFX9-NEXT: buffer_load_ushort v29, off, s[0:3], s32 offset:212
@@ -20905,11 +20901,11 @@ define inreg <32 x i32> @bitcast_v128i8_to_v32i32_scalar(<128 x i8> inreg %a, i3
; GFX9-NEXT: buffer_load_ushort v48, off, s[0:3], s32 offset:228
; GFX9-NEXT: buffer_load_ushort v39, off, s[0:3], s32 offset:236
; GFX9-NEXT: buffer_load_ushort v45, off, s[0:3], s32 offset:244
-; GFX9-NEXT: buffer_load_ushort v40, off, s[0:3], s32 offset:252
+; GFX9-NEXT: buffer_load_ushort v44, off, s[0:3], s32 offset:252
; GFX9-NEXT: buffer_load_ushort v55, off, s[0:3], s32 offset:260
; GFX9-NEXT: buffer_load_ushort v43, off, s[0:3], s32 offset:268
-; GFX9-NEXT: buffer_load_ushort v36, off, s[0:3], s32 offset:276
-; GFX9-NEXT: buffer_load_ushort v42, off, s[0:3], s32 offset:284
+; GFX9-NEXT: buffer_load_ushort v42, off, s[0:3], s32 offset:276
+; GFX9-NEXT: buffer_load_ushort v36, off, s[0:3], s32 offset:284
; GFX9-NEXT: buffer_load_ushort v41, off, s[0:3], s32 offset:292
; GFX9-NEXT: buffer_load_ushort v32, off, s[0:3], s32 offset:300
; GFX9-NEXT: buffer_load_ushort v35, off, s[0:3], s32 offset:308
@@ -20935,7 +20931,7 @@ define inreg <32 x i32> @bitcast_v128i8_to_v32i32_scalar(<128 x i8> inreg %a, i3
; GFX9-NEXT: buffer_store_dword v29, off, s[0:3], s32 offset:668 ; 4-byte Folded Spill
; GFX9-NEXT: buffer_store_dword v52, off, s[0:3], s32 offset:672 ; 4-byte Folded Spill
; GFX9-NEXT: buffer_store_dword v37, off, s[0:3], s32 offset:676 ; 4-byte Folded Spill
-; GFX9-NEXT: buffer_store_dword v44, off, s[0:3], s32 offset:680 ; 4-byte Folded Spill
+; GFX9-NEXT: buffer_store_dword v40, off, s[0:3], s32 offset:680 ; 4-byte Folded Spill
; GFX9-NEXT: buffer_store_dword v50, off, s[0:3], s32 offset:684 ; 4-byte Folded Spill
; GFX9-NEXT: buffer_store_dword v23, off, s[0:3], s32 offset:688 ; 4-byte Folded Spill
; GFX9-NEXT: buffer_store_dword v31, off, s[0:3], s32 offset:692 ; 4-byte Folded Spill
@@ -21110,7 +21106,7 @@ define inreg <32 x i32> @bitcast_v128i8_to_v32i32_scalar(<128 x i8> inreg %a, i3
; GFX9-NEXT: s_waitcnt vmcnt(1)
; GFX9-NEXT: v_or_b32_sdwa v0, v23, v0 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
; GFX9-NEXT: v_or_b32_sdwa v22, v0, v1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_0 src1_sel:DWORD
-; GFX9-NEXT: v_or_b32_sdwa v0, v44, v58 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; GFX9-NEXT: v_or_b32_sdwa v0, v40, v58 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
; GFX9-NEXT: v_or_b32_sdwa v1, v37, v57 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
; GFX9-NEXT: v_or_b32_sdwa v23, v0, v1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_0 src1_sel:DWORD
; GFX9-NEXT: s_waitcnt vmcnt(0)
@@ -21122,7 +21118,7 @@ define inreg <32 x i32> @bitcast_v128i8_to_v32i32_scalar(<128 x i8> inreg %a, i3
; GFX9-NEXT: v_mov_b32_e32 v37, v57
; GFX9-NEXT: v_mov_b32_e32 v57, v60
; GFX9-NEXT: v_mov_b32_e32 v52, v56
-; GFX9-NEXT: buffer_load_dword v44, off, s[0:3], s32 offset:612 ; 4-byte Folded Reload
+; GFX9-NEXT: buffer_load_dword v40, off, s[0:3], s32 offset:612 ; 4-byte Folded Reload
; GFX9-NEXT: s_waitcnt vmcnt(2)
; GFX9-NEXT: v_mov_b32_e32 v34, v0
; GFX9-NEXT: v_or_b32_sdwa v0, v46, v0 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
@@ -21131,14 +21127,14 @@ define inreg <32 x i32> @bitcast_v128i8_to_v32i32_scalar(<128 x i8> inreg %a, i3
; GFX9-NEXT: v_or_b32_sdwa v25, v0, v1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_0 src1_sel:DWORD
; GFX9-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:440 ; 4-byte Folded Reload
; GFX9-NEXT: s_waitcnt vmcnt(1)
-; GFX9-NEXT: v_or_b32_sdwa v1, v45, v44 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; GFX9-NEXT: v_or_b32_sdwa v1, v45, v40 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
; GFX9-NEXT: s_waitcnt vmcnt(0)
; GFX9-NEXT: v_or_b32_sdwa v0, v39, v0 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
; GFX9-NEXT: v_or_b32_sdwa v26, v0, v1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_0 src1_sel:DWORD
; GFX9-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:436 ; 4-byte Folded Reload
; GFX9-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:432 ; 4-byte Folded Reload
; GFX9-NEXT: s_waitcnt vmcnt(1)
-; GFX9-NEXT: v_or_b32_sdwa v0, v40, v0 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; GFX9-NEXT: v_or_b32_sdwa v0, v44, v0 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
; GFX9-NEXT: s_waitcnt vmcnt(0)
; GFX9-NEXT: v_or_b32_sdwa v1, v55, v1 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
; GFX9-NEXT: v_or_b32_sdwa v27, v0, v1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_0 src1_sel:DWORD
@@ -21148,12 +21144,12 @@ define inreg <32 x i32> @bitcast_v128i8_to_v32i32_scalar(<128 x i8> inreg %a, i3
; GFX9-NEXT: v_mov_b32_e32 v51, v0
; GFX9-NEXT: v_or_b32_sdwa v0, v43, v0 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
; GFX9-NEXT: s_waitcnt vmcnt(0)
-; GFX9-NEXT: v_or_b32_sdwa v1, v36, v1 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; GFX9-NEXT: v_or_b32_sdwa v1, v42, v1 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
; GFX9-NEXT: v_or_b32_sdwa v28, v0, v1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_0 src1_sel:DWORD
; GFX9-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:420 ; 4-byte Folded Reload
; GFX9-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:416 ; 4-byte Folded Reload
; GFX9-NEXT: s_waitcnt vmcnt(1)
-; GFX9-NEXT: v_or_b32_sdwa v0, v42, v0 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; GFX9-NEXT: v_or_b32_sdwa v0, v36, v0 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
; GFX9-NEXT: s_waitcnt vmcnt(0)
; GFX9-NEXT: v_or_b32_sdwa v1, v41, v1 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
; GFX9-NEXT: v_or_b32_sdwa v29, v0, v1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_0 src1_sel:DWORD
@@ -21207,7 +21203,7 @@ define inreg <32 x i32> @bitcast_v128i8_to_v32i32_scalar(<128 x i8> inreg %a, i3
; GFX9-NEXT: .LBB15_2:
; GFX9-NEXT: buffer_load_dword v34, off, s[0:3], s32 offset:616 ; 4-byte Folded Reload
; GFX9-NEXT: buffer_load_dword v51, off, s[0:3], s32 offset:620 ; 4-byte Folded Reload
-; GFX9-NEXT: buffer_load_dword v44, off, s[0:3], s32 offset:612 ; 4-byte Folded Reload
+; GFX9-NEXT: buffer_load_dword v40, off, s[0:3], s32 offset:612 ; 4-byte Folded Reload
; GFX9-NEXT: buffer_load_dword v50, off, s[0:3], s32 offset:608 ; 4-byte Folded Reload
; GFX9-NEXT: buffer_load_dword v37, off, s[0:3], s32 offset:604 ; 4-byte Folded Reload
; GFX9-NEXT: buffer_load_dword v47, off, s[0:3], s32 offset:596 ; 4-byte Folded Reload
@@ -21569,12 +21565,12 @@ define inreg <32 x i32> @bitcast_v128i8_to_v32i32_scalar(<128 x i8> inreg %a, i3
; GFX9-NEXT: s_waitcnt vmcnt(1)
; GFX9-NEXT: v_or_b32_sdwa v0, v1, v0 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:BYTE_0
; GFX9-NEXT: v_add_u32_e32 v1, 3, v45
-; GFX9-NEXT: v_or_b32_sdwa v1, v44, v1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:BYTE_0
+; GFX9-NEXT: v_or_b32_sdwa v1, v40, v1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:BYTE_0
; GFX9-NEXT: v_add_u32_e32 v0, 0x300, v0
; GFX9-NEXT: v_add_u32_sdwa v1, v1, s4 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:DWORD
; GFX9-NEXT: v_or_b32_sdwa v26, v0, v1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_0 src1_sel:DWORD
; GFX9-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:436 ; 4-byte Folded Reload
-; GFX9-NEXT: v_add_u32_e32 v0, 3, v40
+; GFX9-NEXT: v_add_u32_e32 v0, 3, v44
; GFX9-NEXT: s_waitcnt vmcnt(0)
; GFX9-NEXT: v_or_b32_sdwa v0, v1, v0 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:BYTE_0
; GFX9-NEXT: v_add_u32_e32 v1, 3, v55
@@ -21584,7 +21580,7 @@ define inreg <32 x i32> @bitcast_v128i8_to_v32i32_scalar(<128 x i8> inreg %a, i3
; GFX9-NEXT: v_add_u32_sdwa v1, v1, s4 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:DWORD
; GFX9-NEXT: v_or_b32_sdwa v27, v0, v1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_0 src1_sel:DWORD
; GFX9-NEXT: v_add_u32_e32 v0, 3, v43
-; GFX9-NEXT: v_add_u32_e32 v1, 3, v36
+; GFX9-NEXT: v_add_u32_e32 v1, 3, v42
; GFX9-NEXT: v_or_b32_sdwa v0, v51, v0 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:BYTE_0
; GFX9-NEXT: v_add_u32_e32 v0, 0x300, v0
; GFX9-NEXT: s_waitcnt vmcnt(0)
@@ -21593,7 +21589,7 @@ define inreg <32 x i32> @bitcast_v128i8_to_v32i32_scalar(<128 x i8> inreg %a, i3
; GFX9-NEXT: v_or_b32_sdwa v28, v0, v1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_0 src1_sel:DWORD
; GFX9-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:420 ; 4-byte Folded Reload
; GFX9-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:416 ; 4-byte Folded Reload
-; GFX9-NEXT: v_add_u32_e32 v0, 3, v42
+; GFX9-NEXT: v_add_u32_e32 v0, 3, v36
; GFX9-NEXT: s_waitcnt vmcnt(1)
; GFX9-NEXT: v_or_b32_sdwa v0, v1, v0 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:BYTE_0
; GFX9-NEXT: v_add_u32_e32 v1, 3, v41
@@ -24695,8 +24691,23 @@ define inreg <64 x bfloat> @bitcast_v32i32_to_v64bf16_scalar(<32 x i32> inreg %a
; SI-NEXT: s_mov_b32 s72, s74
; SI-NEXT: s_mov_b32 s73, s75
; SI-NEXT: s_mov_b32 s74, s76
-; SI-NEXT: v_readlane_b32 s75, v21, 0
-; SI-NEXT: v_readlane_b32 s76, v21, 1
+; SI-NEXT: s_mov_b32 s75, s77
+; SI-NEXT: s_mov_b32 s76, s78
+; SI-NEXT: s_mov_b32 s77, s79
+; SI-NEXT: s_mov_b32 s78, s88
+; SI-NEXT: s_mov_b32 s79, s89
+; SI-NEXT: s_mov_b32 s88, s90
+; SI-NEXT: s_mov_b32 s89, s91
+; SI-NEXT: s_mov_b32 s90, s92
+; SI-NEXT: s_mov_b32 s91, s93
+; SI-NEXT: s_mov_b32 s92, s94
+; SI-NEXT: s_mov_b32 s93, s95
+; SI-NEXT: s_mov_b32 s94, s30
+; SI-NEXT: s_mov_b32 s95, s31
+; SI-NEXT: s_mov_b32 s30, s34
+; SI-NEXT: s_mov_b32 s31, s35
+; SI-NEXT: v_readlane_b32 s34, v21, 0
+; SI-NEXT: v_readlane_b32 s35, v21, 1
; SI-NEXT: s_cbranch_vccnz .LBB17_5
; SI-NEXT: ; %bb.4: ; %cmp.true
; SI-NEXT: s_add_i32 s16, s16, 3
@@ -24758,22 +24769,22 @@ define inreg <64 x bfloat> @bitcast_v32i32_to_v64bf16_scalar(<32 x i32> inreg %a
; SI-NEXT: s_lshl_b32 s62, s84, 16
; SI-NEXT: s_and_b32 s73, s83, 0xffff0000
; SI-NEXT: s_lshl_b32 s72, s83, 16
-; SI-NEXT: s_and_b32 s77, s82, 0xffff0000
+; SI-NEXT: s_and_b32 s75, s82, 0xffff0000
; SI-NEXT: s_lshl_b32 s74, s82, 16
-; SI-NEXT: s_and_b32 s79, s81, 0xffff0000
-; SI-NEXT: s_lshl_b32 s78, s81, 16
-; SI-NEXT: s_and_b32 s89, s80, 0xffff0000
-; SI-NEXT: s_lshl_b32 s88, s80, 16
-; SI-NEXT: s_and_b32 s91, s71, 0xffff0000
-; SI-NEXT: s_lshl_b32 s90, s71, 16
-; SI-NEXT: s_and_b32 s93, s70, 0xffff0000
-; SI-NEXT: s_lshl_b32 s92, s70, 16
-; SI-NEXT: s_and_b32 s95, s29, 0xffff0000
-; SI-NEXT: s_lshl_b32 s94, s29, 16
-; SI-NEXT: s_and_b32 s31, s28, 0xffff0000
-; SI-NEXT: s_lshl_b32 s30, s28, 16
-; SI-NEXT: s_and_b32 s35, s27, 0xffff0000
-; SI-NEXT: s_lshl_b32 s34, s27, 16
+; SI-NEXT: s_and_b32 s77, s81, 0xffff0000
+; SI-NEXT: s_lshl_b32 s76, s81, 16
+; SI-NEXT: s_and_b32 s79, s80, 0xffff0000
+; SI-NEXT: s_lshl_b32 s78, s80, 16
+; SI-NEXT: s_and_b32 s89, s71, 0xffff0000
+; SI-NEXT: s_lshl_b32 s88, s71, 16
+; SI-NEXT: s_and_b32 s91, s70, 0xffff0000
+; SI-NEXT: s_lshl_b32 s90, s70, 16
+; SI-NEXT: s_and_b32 s93, s29, 0xffff0000
+; SI-NEXT: s_lshl_b32 s92, s29, 16
+; SI-NEXT: s_and_b32 s95, s28, 0xffff0000
+; SI-NEXT: s_lshl_b32 s94, s28, 16
+; SI-NEXT: s_and_b32 s31, s27, 0xffff0000
+; SI-NEXT: s_lshl_b32 s30, s27, 16
; SI-NEXT: s_and_b32 s37, s26, 0xffff0000
; SI-NEXT: s_lshl_b32 s36, s26, 16
; SI-NEXT: s_and_b32 s39, s25, 0xffff0000
@@ -24792,8 +24803,8 @@ define inreg <64 x bfloat> @bitcast_v32i32_to_v64bf16_scalar(<32 x i32> inreg %a
; SI-NEXT: s_lshl_b32 s66, s19, 16
; SI-NEXT: s_and_b32 s69, s18, 0xffff0000
; SI-NEXT: s_lshl_b32 s68, s18, 16
-; SI-NEXT: s_and_b32 s76, s17, 0xffff0000
-; SI-NEXT: s_lshl_b32 s75, s17, 16
+; SI-NEXT: s_and_b32 s35, s17, 0xffff0000
+; SI-NEXT: s_lshl_b32 s34, s17, 16
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_writelane_b32 v21, s6, 2
; SI-NEXT: s_lshl_b32 s6, s16, 16
@@ -24807,9 +24818,9 @@ define inreg <64 x bfloat> @bitcast_v32i32_to_v64bf16_scalar(<32 x i32> inreg %a
; SI-NEXT: v_alignbit_b32 v1, v1, v2, 16
; SI-NEXT: buffer_store_dword v1, v0, s[0:3], 0 offen
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_mul_f32_e64 v1, 1.0, s76
+; SI-NEXT: v_mul_f32_e64 v1, 1.0, s35
; SI-NEXT: v_lshrrev_b32_e32 v1, 16, v1
-; SI-NEXT: v_mul_f32_e64 v2, 1.0, s75
+; SI-NEXT: v_mul_f32_e64 v2, 1.0, s34
; SI-NEXT: v_alignbit_b32 v1, v1, v2, 16
; SI-NEXT: v_add_i32_e32 v2, vcc, 4, v0
; SI-NEXT: buffer_store_dword v1, v2, s[0:3], 0 offen
@@ -24877,57 +24888,57 @@ define inreg <64 x bfloat> @bitcast_v32i32_to_v64bf16_scalar(<32 x i32> inreg %a
; SI-NEXT: v_add_i32_e32 v2, vcc, 40, v0
; SI-NEXT: buffer_store_dword v1, v2, s[0:3], 0 offen
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_mul_f32_e64 v1, 1.0, s35
-; SI-NEXT: v_lshrrev_b32_e32 v1, 16, v1
-; SI-NEXT: v_mul_f32_e64 v2, 1.0, s34
-; SI-NEXT: v_alignbit_b32 v1, v1, v2, 16
-; SI-NEXT: v_add_i32_e32 v2, vcc, 44, v0
-; SI-NEXT: buffer_store_dword v1, v2, s[0:3], 0 offen
-; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_mul_f32_e64 v1, 1.0, s31
; SI-NEXT: v_lshrrev_b32_e32 v1, 16, v1
; SI-NEXT: v_mul_f32_e64 v2, 1.0, s30
; SI-NEXT: v_alignbit_b32 v1, v1, v2, 16
-; SI-NEXT: v_add_i32_e32 v2, vcc, 48, v0
+; SI-NEXT: v_add_i32_e32 v2, vcc, 44, v0
; SI-NEXT: buffer_store_dword v1, v2, s[0:3], 0 offen
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_mul_f32_e64 v1, 1.0, s95
; SI-NEXT: v_lshrrev_b32_e32 v1, 16, v1
; SI-NEXT: v_mul_f32_e64 v2, 1.0, s94
; SI-NEXT: v_alignbit_b32 v1, v1, v2, 16
-; SI-NEXT: v_add_i32_e32 v2, vcc, 52, v0
+; SI-NEXT: v_add_i32_e32 v2, vcc, 48, v0
; SI-NEXT: buffer_store_dword v1, v2, s[0:3], 0 offen
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_mul_f32_e64 v1, 1.0, s93
; SI-NEXT: v_lshrrev_b32_e32 v1, 16, v1
; SI-NEXT: v_mul_f32_e64 v2, 1.0, s92
; SI-NEXT: v_alignbit_b32 v1, v1, v2, 16
-; SI-NEXT: v_add_i32_e32 v2, vcc, 56, v0
+; SI-NEXT: v_add_i32_e32 v2, vcc, 52, v0
; SI-NEXT: buffer_store_dword v1, v2, s[0:3], 0 offen
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_mul_f32_e64 v1, 1.0, s91
; SI-NEXT: v_lshrrev_b32_e32 v1, 16, v1
; SI-NEXT: v_mul_f32_e64 v2, 1.0, s90
; SI-NEXT: v_alignbit_b32 v1, v1, v2, 16
-; SI-NEXT: v_add_i32_e32 v2, vcc, 60, v0
+; SI-NEXT: v_add_i32_e32 v2, vcc, 56, v0
; SI-NEXT: buffer_store_dword v1, v2, s[0:3], 0 offen
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_mul_f32_e64 v1, 1.0, s89
; SI-NEXT: v_lshrrev_b32_e32 v1, 16, v1
; SI-NEXT: v_mul_f32_e64 v2, 1.0, s88
; SI-NEXT: v_alignbit_b32 v1, v1, v2, 16
-; SI-NEXT: v_add_i32_e32 v2, vcc, 64, v0
+; SI-NEXT: v_add_i32_e32 v2, vcc, 60, v0
; SI-NEXT: buffer_store_dword v1, v2, s[0:3], 0 offen
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_mul_f32_e64 v1, 1.0, s79
; SI-NEXT: v_lshrrev_b32_e32 v1, 16, v1
; SI-NEXT: v_mul_f32_e64 v2, 1.0, s78
; SI-NEXT: v_alignbit_b32 v1, v1, v2, 16
-; SI-NEXT: v_add_i32_e32 v2, vcc, 0x44, v0
+; SI-NEXT: v_add_i32_e32 v2, vcc, 64, v0
; SI-NEXT: buffer_store_dword v1, v2, s[0:3], 0 offen
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_mul_f32_e64 v1, 1.0, s77
; SI-NEXT: v_lshrrev_b32_e32 v1, 16, v1
+; SI-NEXT: v_mul_f32_e64 v2, 1.0, s76
+; SI-NEXT: v_alignbit_b32 v1, v1, v2, 16
+; SI-NEXT: v_add_i32_e32 v2, vcc, 0x44, v0
+; SI-NEXT: buffer_store_dword v1, v2, s[0:3], 0 offen
+; SI-NEXT: s_waitcnt expcnt(0)
+; SI-NEXT: v_mul_f32_e64 v1, 1.0, s75
+; SI-NEXT: v_lshrrev_b32_e32 v1, 16, v1
; SI-NEXT: v_mul_f32_e64 v2, 1.0, s74
; SI-NEXT: v_alignbit_b32 v1, v1, v2, 16
; SI-NEXT: v_add_i32_e32 v2, vcc, 0x48, v0
@@ -28367,10 +28378,10 @@ define inreg <32 x i32> @bitcast_v64bf16_to_v32i32_scalar(<64 x bfloat> inreg %a
; SI-NEXT: v_mul_f32_e32 v34, 1.0, v36
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_mul_f32_e64 v0, 1.0, s17
-; SI-NEXT: v_mul_f32_e64 v35, 1.0, s18
+; SI-NEXT: v_mul_f32_e64 v33, 1.0, s18
; SI-NEXT: v_mul_f32_e64 v36, 1.0, s21
; SI-NEXT: v_mul_f32_e64 v42, 1.0, s20
-; SI-NEXT: v_mul_f32_e64 v33, 1.0, s22
+; SI-NEXT: v_mul_f32_e64 v35, 1.0, s22
; SI-NEXT: buffer_store_dword v30, off, s[0:3], s32 offset:196 ; 4-byte Folded Spill
; SI-NEXT: buffer_store_dword v31, off, s[0:3], s32 offset:200 ; 4-byte Folded Spill
; SI-NEXT: buffer_store_dword v29, off, s[0:3], s32 offset:204 ; 4-byte Folded Spill
@@ -28399,9 +28410,9 @@ define inreg <32 x i32> @bitcast_v64bf16_to_v32i32_scalar(<64 x bfloat> inreg %a
; SI-NEXT: buffer_store_dword v9, off, s[0:3], s32 offset:296 ; 4-byte Folded Spill
; SI-NEXT: buffer_store_dword v3, off, s[0:3], s32 offset:300 ; 4-byte Folded Spill
; SI-NEXT: buffer_store_dword v0, off, s[0:3], s32 offset:304 ; 4-byte Folded Spill
-; SI-NEXT: buffer_store_dword v33, off, s[0:3], s32 offset:308 ; 4-byte Folded Spill
-; SI-NEXT: buffer_store_dword v35, off, s[0:3], s32 offset:312 ; 4-byte Folded Spill
-; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:316 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:308 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v33, off, s[0:3], s32 offset:312 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v35, off, s[0:3], s32 offset:316 ; 4-byte Folded Spill
; SI-NEXT: buffer_store_dword v36, off, s[0:3], s32 offset:320 ; 4-byte Folded Spill
; SI-NEXT: buffer_store_dword v42, off, s[0:3], s32 offset:324 ; 4-byte Folded Spill
; SI-NEXT: s_cbranch_scc0 .LBB19_4
@@ -28412,12 +28423,13 @@ define inreg <32 x i32> @bitcast_v64bf16_to_v32i32_scalar(<64 x bfloat> inreg %a
; SI-NEXT: buffer_load_dword v7, off, s[0:3], s32 offset:192 ; 4-byte Folded Reload
; SI-NEXT: v_alignbit_b32 v5, v5, v8, 16
; SI-NEXT: buffer_load_dword v8, off, s[0:3], s32 offset:184 ; 4-byte Folded Reload
-; SI-NEXT: s_waitcnt expcnt(2)
-; SI-NEXT: v_lshrrev_b32_e32 v1, 16, v1
-; SI-NEXT: v_alignbit_b32 v1, v1, v35, 16
+; SI-NEXT: s_waitcnt expcnt(6)
+; SI-NEXT: v_lshrrev_b32_e32 v3, 16, v3
+; SI-NEXT: v_alignbit_b32 v3, v3, v35, 16
; SI-NEXT: v_lshrrev_b32_e32 v4, 16, v4
; SI-NEXT: v_alignbit_b32 v4, v4, v9, 16
; SI-NEXT: buffer_load_dword v9, off, s[0:3], s32 offset:180 ; 4-byte Folded Reload
+; SI-NEXT: s_waitcnt expcnt(5)
; SI-NEXT: v_lshrrev_b32_e32 v0, 16, v0
; SI-NEXT: v_mov_b32_e32 v59, v2
; SI-NEXT: v_alignbit_b32 v0, v0, v2, 16
@@ -28427,10 +28439,11 @@ define inreg <32 x i32> @bitcast_v64bf16_to_v32i32_scalar(<64 x bfloat> inreg %a
; SI-NEXT: v_mov_b32_e32 v47, v10
; SI-NEXT: v_lshrrev_b32_e32 v10, 16, v10
; SI-NEXT: v_mov_b32_e32 v45, v12
-; SI-NEXT: v_lshrrev_b32_e32 v3, 16, v3
-; SI-NEXT: v_alignbit_b32 v3, v3, v33, 16
-; SI-NEXT: v_mov_b32_e32 v33, v14
+; SI-NEXT: v_mov_b32_e32 v44, v14
+; SI-NEXT: s_waitcnt expcnt(4)
+; SI-NEXT: v_lshrrev_b32_e32 v1, 16, v1
; SI-NEXT: v_lshrrev_b32_e32 v13, 16, v13
+; SI-NEXT: v_alignbit_b32 v1, v1, v33, 16
; SI-NEXT: v_mov_b32_e32 v62, v38
; SI-NEXT: v_lshrrev_b32_e32 v15, 16, v15
; SI-NEXT: v_lshrrev_b32_e32 v16, 16, v16
@@ -28470,20 +28483,20 @@ define inreg <32 x i32> @bitcast_v64bf16_to_v32i32_scalar(<64 x bfloat> inreg %a
; SI-NEXT: v_alignbit_b32 v28, v28, v37, 16
; SI-NEXT: v_mov_b32_e32 v37, v34
; SI-NEXT: buffer_load_dword v32, off, s[0:3], s32 offset:148 ; 4-byte Folded Reload
-; SI-NEXT: s_waitcnt vmcnt(3)
+; SI-NEXT: s_waitcnt vmcnt(3) expcnt(2)
; SI-NEXT: v_mov_b32_e32 v35, v7
; SI-NEXT: v_lshrrev_b32_e32 v7, 16, v7
; SI-NEXT: s_waitcnt vmcnt(2)
; SI-NEXT: v_mov_b32_e32 v43, v8
; SI-NEXT: v_alignbit_b32 v7, v7, v8, 16
; SI-NEXT: buffer_load_dword v8, off, s[0:3], s32 offset:188 ; 4-byte Folded Reload
-; SI-NEXT: s_waitcnt vmcnt(2) expcnt(0)
-; SI-NEXT: v_mov_b32_e32 v42, v9
+; SI-NEXT: s_waitcnt vmcnt(2)
+; SI-NEXT: v_mov_b32_e32 v60, v9
; SI-NEXT: s_waitcnt vmcnt(1)
; SI-NEXT: v_lshrrev_b32_e32 v31, 16, v32
; SI-NEXT: v_alignbit_b32 v31, v31, v34, 16
-; SI-NEXT: s_waitcnt vmcnt(0)
-; SI-NEXT: v_mov_b32_e32 v60, v8
+; SI-NEXT: s_waitcnt vmcnt(0) expcnt(0)
+; SI-NEXT: v_mov_b32_e32 v42, v8
; SI-NEXT: v_lshrrev_b32_e32 v8, 16, v8
; SI-NEXT: v_alignbit_b32 v8, v8, v9, 16
; SI-NEXT: v_lshrrev_b32_e32 v9, 16, v11
@@ -28507,7 +28520,7 @@ define inreg <32 x i32> @bitcast_v64bf16_to_v32i32_scalar(<64 x bfloat> inreg %a
; SI-NEXT: v_alignbit_b32 v12, v12, v14, 16
; SI-NEXT: buffer_load_dword v14, off, s[0:3], s32 offset:160 ; 4-byte Folded Reload
; SI-NEXT: s_waitcnt vmcnt(0)
-; SI-NEXT: v_mov_b32_e32 v44, v14
+; SI-NEXT: v_mov_b32_e32 v33, v14
; SI-NEXT: v_alignbit_b32 v13, v13, v14, 16
; SI-NEXT: buffer_load_dword v14, off, s[0:3], s32 offset:152 ; 4-byte Folded Reload
; SI-NEXT: s_waitcnt vmcnt(0)
@@ -28530,7 +28543,7 @@ define inreg <32 x i32> @bitcast_v64bf16_to_v32i32_scalar(<64 x bfloat> inreg %a
; SI-NEXT: .LBB19_2: ; %cmp.true
; SI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:304 ; 4-byte Folded Reload
; SI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:312 ; 4-byte Folded Reload
-; SI-NEXT: buffer_load_dword v3, off, s[0:3], s32 offset:316 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v3, off, s[0:3], s32 offset:308 ; 4-byte Folded Reload
; SI-NEXT: buffer_load_dword v4, off, s[0:3], s32 offset:300 ; 4-byte Folded Reload
; SI-NEXT: v_and_b32_e32 v0, 0xffff0000, v59
; SI-NEXT: v_add_f32_e32 v0, 0x40c00000, v0
@@ -28546,7 +28559,7 @@ define inreg <32 x i32> @bitcast_v64bf16_to_v32i32_scalar(<64 x bfloat> inreg %a
; SI-NEXT: v_and_b32_e32 v8, 0xffff0000, v35
; SI-NEXT: v_add_f32_e32 v8, 0x40c00000, v8
; SI-NEXT: v_lshrrev_b32_e32 v8, 16, v8
-; SI-NEXT: v_and_b32_e32 v9, 0xffff0000, v60
+; SI-NEXT: v_and_b32_e32 v9, 0xffff0000, v42
; SI-NEXT: v_add_f32_e32 v9, 0x40c00000, v9
; SI-NEXT: v_lshrrev_b32_e32 v9, 16, v9
; SI-NEXT: v_and_b32_e32 v10, 0xffff0000, v57
@@ -28558,7 +28571,7 @@ define inreg <32 x i32> @bitcast_v64bf16_to_v32i32_scalar(<64 x bfloat> inreg %a
; SI-NEXT: v_and_b32_e32 v12, 0xffff0000, v45
; SI-NEXT: v_add_f32_e32 v12, 0x40c00000, v12
; SI-NEXT: v_lshrrev_b32_e32 v12, 16, v12
-; SI-NEXT: v_and_b32_e32 v13, 0xffff0000, v33
+; SI-NEXT: v_and_b32_e32 v13, 0xffff0000, v44
; SI-NEXT: v_add_f32_e32 v13, 0x40c00000, v13
; SI-NEXT: v_lshrrev_b32_e32 v13, 16, v13
; SI-NEXT: v_and_b32_e32 v15, 0xffff0000, v36
@@ -28670,7 +28683,7 @@ define inreg <32 x i32> @bitcast_v64bf16_to_v32i32_scalar(<64 x bfloat> inreg %a
; SI-NEXT: v_add_f32_e32 v2, 0x40c00000, v2
; SI-NEXT: v_lshrrev_b32_e32 v3, 16, v3
; SI-NEXT: v_alignbit_b32 v2, v3, v2, 16
-; SI-NEXT: buffer_load_dword v3, off, s[0:3], s32 offset:308 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v3, off, s[0:3], s32 offset:316 ; 4-byte Folded Reload
; SI-NEXT: s_waitcnt vmcnt(0)
; SI-NEXT: v_and_b32_e32 v3, 0xffff0000, v3
; SI-NEXT: v_add_f32_e32 v3, 0x40c00000, v3
@@ -28693,7 +28706,7 @@ define inreg <32 x i32> @bitcast_v64bf16_to_v32i32_scalar(<64 x bfloat> inreg %a
; SI-NEXT: v_and_b32_e32 v7, 0xffff0000, v43
; SI-NEXT: v_add_f32_e32 v7, 0x40c00000, v7
; SI-NEXT: v_alignbit_b32 v7, v8, v7, 16
-; SI-NEXT: v_and_b32_e32 v8, 0xffff0000, v42
+; SI-NEXT: v_and_b32_e32 v8, 0xffff0000, v60
; SI-NEXT: v_add_f32_e32 v8, 0x40c00000, v8
; SI-NEXT: v_alignbit_b32 v8, v9, v8, 16
; SI-NEXT: v_and_b32_e32 v9, 0xffff0000, v58
@@ -28708,7 +28721,7 @@ define inreg <32 x i32> @bitcast_v64bf16_to_v32i32_scalar(<64 x bfloat> inreg %a
; SI-NEXT: v_and_b32_e32 v12, 0xffff0000, v63
; SI-NEXT: v_add_f32_e32 v12, 0x40c00000, v12
; SI-NEXT: v_alignbit_b32 v12, v13, v12, 16
-; SI-NEXT: v_and_b32_e32 v13, 0xffff0000, v44
+; SI-NEXT: v_and_b32_e32 v13, 0xffff0000, v33
; SI-NEXT: v_add_f32_e32 v13, 0x40c00000, v13
; SI-NEXT: v_alignbit_b32 v13, v14, v13, 16
; SI-NEXT: v_and_b32_e32 v14, 0xffff0000, v62
@@ -28794,16 +28807,16 @@ define inreg <32 x i32> @bitcast_v64bf16_to_v32i32_scalar(<64 x bfloat> inreg %a
; SI-NEXT: s_setpc_b64 s[30:31]
; SI-NEXT: .LBB19_4:
; SI-NEXT: buffer_load_dword v43, off, s[0:3], s32 offset:184 ; 4-byte Folded Reload
-; SI-NEXT: s_waitcnt expcnt(3)
+; SI-NEXT: s_waitcnt expcnt(2)
; SI-NEXT: buffer_load_dword v35, off, s[0:3], s32 offset:192 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v60, off, s[0:3], s32 offset:180 ; 4-byte Folded Reload
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: buffer_load_dword v42, off, s[0:3], s32 offset:180 ; 4-byte Folded Reload
-; SI-NEXT: buffer_load_dword v60, off, s[0:3], s32 offset:188 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v42, off, s[0:3], s32 offset:188 ; 4-byte Folded Reload
; SI-NEXT: buffer_load_dword v58, off, s[0:3], s32 offset:176 ; 4-byte Folded Reload
; SI-NEXT: buffer_load_dword v56, off, s[0:3], s32 offset:172 ; 4-byte Folded Reload
; SI-NEXT: buffer_load_dword v46, off, s[0:3], s32 offset:168 ; 4-byte Folded Reload
; SI-NEXT: buffer_load_dword v63, off, s[0:3], s32 offset:164 ; 4-byte Folded Reload
-; SI-NEXT: buffer_load_dword v44, off, s[0:3], s32 offset:160 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v33, off, s[0:3], s32 offset:160 ; 4-byte Folded Reload
; SI-NEXT: v_mov_b32_e32 v61, v53
; SI-NEXT: buffer_load_dword v36, off, s[0:3], s32 offset:152 ; 4-byte Folded Reload
; SI-NEXT: buffer_load_dword v53, off, s[0:3], s32 offset:156 ; 4-byte Folded Reload
@@ -28812,7 +28825,7 @@ define inreg <32 x i32> @bitcast_v64bf16_to_v32i32_scalar(<64 x bfloat> inreg %a
; SI-NEXT: v_mov_b32_e32 v57, v11
; SI-NEXT: v_mov_b32_e32 v47, v10
; SI-NEXT: v_mov_b32_e32 v45, v12
-; SI-NEXT: v_mov_b32_e32 v33, v14
+; SI-NEXT: v_mov_b32_e32 v44, v14
; SI-NEXT: v_mov_b32_e32 v62, v38
; SI-NEXT: v_mov_b32_e32 v38, v39
; SI-NEXT: v_mov_b32_e32 v39, v41
@@ -31878,36 +31891,96 @@ define <64 x half> @bitcast_v32i32_to_v64f16(<32 x i32> %a, i32 %b) {
; SI-NEXT: buffer_load_dword v31, off, s[0:3], s32 offset:8
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: buffer_load_dword v63, off, s[0:3], s32
+; SI-NEXT: ; implicit-def: $vgpr32
+; SI-NEXT: ; kill: killed $vgpr32
+; SI-NEXT: ; implicit-def: $vgpr32
+; SI-NEXT: ; kill: killed $vgpr32
+; SI-NEXT: ; implicit-def: $vgpr32
+; SI-NEXT: ; kill: killed $vgpr32
+; SI-NEXT: ; implicit-def: $vgpr32
+; SI-NEXT: ; kill: killed $vgpr32
+; SI-NEXT: ; implicit-def: $vgpr32
+; SI-NEXT: ; kill: killed $vgpr32
+; SI-NEXT: ; implicit-def: $vgpr32
+; SI-NEXT: ; kill: killed $vgpr32
+; SI-NEXT: ; implicit-def: $vgpr32
+; SI-NEXT: ; kill: killed $vgpr32
+; SI-NEXT: ; implicit-def: $vgpr32
+; SI-NEXT: ; kill: killed $vgpr32
+; SI-NEXT: ; implicit-def: $vgpr32
+; SI-NEXT: ; kill: killed $vgpr32
+; SI-NEXT: ; implicit-def: $vgpr32
+; SI-NEXT: ; kill: killed $vgpr32
+; SI-NEXT: ; implicit-def: $vgpr32
+; SI-NEXT: ; kill: killed $vgpr32
+; SI-NEXT: ; implicit-def: $vgpr32
+; SI-NEXT: ; kill: killed $vgpr32
+; SI-NEXT: ; implicit-def: $vgpr32
+; SI-NEXT: ; kill: killed $vgpr32
+; SI-NEXT: ; implicit-def: $vgpr32
+; SI-NEXT: ; kill: killed $vgpr32
+; SI-NEXT: ; implicit-def: $vgpr32
+; SI-NEXT: ; kill: killed $vgpr32
+; SI-NEXT: ; implicit-def: $vgpr32
+; SI-NEXT: ; kill: killed $vgpr32
+; SI-NEXT: ; implicit-def: $vgpr32
+; SI-NEXT: ; kill: killed $vgpr32
+; SI-NEXT: ; implicit-def: $vgpr32
+; SI-NEXT: ; kill: killed $vgpr32
+; SI-NEXT: ; implicit-def: $vgpr32
+; SI-NEXT: ; kill: killed $vgpr32
+; SI-NEXT: ; implicit-def: $vgpr32
+; SI-NEXT: ; kill: killed $vgpr32
+; SI-NEXT: ; implicit-def: $vgpr32
+; SI-NEXT: ; kill: killed $vgpr32
+; SI-NEXT: ; implicit-def: $vgpr32
+; SI-NEXT: ; kill: killed $vgpr32
+; SI-NEXT: ; implicit-def: $vgpr32
+; SI-NEXT: ; kill: killed $vgpr32
+; SI-NEXT: ; implicit-def: $vgpr32
+; SI-NEXT: ; kill: killed $vgpr32
+; SI-NEXT: ; implicit-def: $vgpr32
+; SI-NEXT: ; kill: killed $vgpr32
+; SI-NEXT: ; implicit-def: $vgpr32
+; SI-NEXT: ; kill: killed $vgpr32
+; SI-NEXT: ; implicit-def: $vgpr32
+; SI-NEXT: ; kill: killed $vgpr32
+; SI-NEXT: ; implicit-def: $vgpr32
+; SI-NEXT: ; kill: killed $vgpr32
+; SI-NEXT: ; implicit-def: $vgpr32
+; SI-NEXT: ; kill: killed $vgpr32
+; SI-NEXT: ; implicit-def: $vgpr32
; SI-NEXT: ; implicit-def: $vgpr60
; SI-NEXT: ; implicit-def: $vgpr61
; SI-NEXT: ; implicit-def: $vgpr58
; SI-NEXT: ; implicit-def: $vgpr59
-; SI-NEXT: ; implicit-def: $vgpr43
+; SI-NEXT: ; implicit-def: $vgpr56
; SI-NEXT: ; implicit-def: $vgpr57
-; SI-NEXT: ; implicit-def: $vgpr42
+; SI-NEXT: ; implicit-def: $vgpr46
; SI-NEXT: ; implicit-def: $vgpr47
-; SI-NEXT: ; implicit-def: $vgpr44
+; SI-NEXT: ; implicit-def: $vgpr40
; SI-NEXT: ; implicit-def: $vgpr45
-; SI-NEXT: ; implicit-def: $vgpr38
+; SI-NEXT: ; implicit-def: $vgpr42
+; SI-NEXT: ; implicit-def: $vgpr43
; SI-NEXT: ; implicit-def: $vgpr41
-; SI-NEXT: ; implicit-def: $vgpr55
-; SI-NEXT: ; implicit-def: $vgpr53
+; SI-NEXT: ; implicit-def: $vgpr38
; SI-NEXT: ; implicit-def: $vgpr51
; SI-NEXT: ; implicit-def: $vgpr49
; SI-NEXT: ; implicit-def: $vgpr39
; SI-NEXT: ; implicit-def: $vgpr37
; SI-NEXT: ; implicit-def: $vgpr35
; SI-NEXT: ; implicit-def: $vgpr33
-; SI-NEXT: ; implicit-def: $vgpr46
-; SI-NEXT: ; implicit-def: $vgpr56
-; SI-NEXT: ; implicit-def: $vgpr50
+; SI-NEXT: ; implicit-def: $vgpr44
+; SI-NEXT: ; kill: killed $vgpr32
; SI-NEXT: ; implicit-def: $vgpr52
+; SI-NEXT: ; implicit-def: $vgpr55
; SI-NEXT: ; implicit-def: $vgpr54
+; SI-NEXT: ; implicit-def: $vgpr50
+; SI-NEXT: ; implicit-def: $vgpr53
; SI-NEXT: ; implicit-def: $vgpr48
-; SI-NEXT: ; implicit-def: $vgpr40
; SI-NEXT: ; implicit-def: $vgpr36
-; SI-NEXT: ; implicit-def: $vgpr34
; SI-NEXT: ; implicit-def: $vgpr32
+; SI-NEXT: ; implicit-def: $vgpr34
; SI-NEXT: s_waitcnt vmcnt(1)
; SI-NEXT: v_cmp_ne_u32_e32 vcc, 0, v31
; SI-NEXT: ; implicit-def: $vgpr31
@@ -31917,173 +31990,99 @@ define <64 x half> @bitcast_v32i32_to_v64f16(<32 x i32> %a, i32 %b) {
; SI-NEXT: ; implicit-def: $vgpr31
; SI-NEXT: ; kill: killed $vgpr31
; SI-NEXT: ; implicit-def: $vgpr31
-; SI-NEXT: ; kill: killed $vgpr31
-; SI-NEXT: ; implicit-def: $vgpr31
-; SI-NEXT: ; kill: killed $vgpr31
-; SI-NEXT: ; implicit-def: $vgpr31
-; SI-NEXT: ; kill: killed $vgpr31
-; SI-NEXT: ; implicit-def: $vgpr31
-; SI-NEXT: ; kill: killed $vgpr31
-; SI-NEXT: ; implicit-def: $vgpr31
-; SI-NEXT: ; kill: killed $vgpr31
-; SI-NEXT: ; implicit-def: $vgpr31
-; SI-NEXT: ; kill: killed $vgpr31
-; SI-NEXT: ; implicit-def: $vgpr31
-; SI-NEXT: ; kill: killed $vgpr31
-; SI-NEXT: ; implicit-def: $vgpr31
-; SI-NEXT: ; kill: killed $vgpr31
-; SI-NEXT: ; implicit-def: $vgpr31
-; SI-NEXT: ; kill: killed $vgpr31
-; SI-NEXT: ; implicit-def: $vgpr31
-; SI-NEXT: ; kill: killed $vgpr31
-; SI-NEXT: ; implicit-def: $vgpr31
-; SI-NEXT: ; kill: killed $vgpr31
-; SI-NEXT: ; implicit-def: $vgpr31
-; SI-NEXT: ; kill: killed $vgpr31
-; SI-NEXT: ; implicit-def: $vgpr31
-; SI-NEXT: ; kill: killed $vgpr31
-; SI-NEXT: ; implicit-def: $vgpr31
-; SI-NEXT: ; kill: killed $vgpr31
-; SI-NEXT: ; implicit-def: $vgpr31
-; SI-NEXT: ; kill: killed $vgpr31
-; SI-NEXT: ; implicit-def: $vgpr31
-; SI-NEXT: ; kill: killed $vgpr31
-; SI-NEXT: ; implicit-def: $vgpr31
-; SI-NEXT: ; kill: killed $vgpr31
-; SI-NEXT: ; implicit-def: $vgpr31
-; SI-NEXT: ; kill: killed $vgpr31
-; SI-NEXT: ; implicit-def: $vgpr31
-; SI-NEXT: ; kill: killed $vgpr31
-; SI-NEXT: ; implicit-def: $vgpr31
-; SI-NEXT: ; kill: killed $vgpr31
-; SI-NEXT: ; implicit-def: $vgpr31
-; SI-NEXT: ; kill: killed $vgpr31
-; SI-NEXT: ; implicit-def: $vgpr31
-; SI-NEXT: ; kill: killed $vgpr31
-; SI-NEXT: ; implicit-def: $vgpr31
-; SI-NEXT: ; kill: killed $vgpr31
-; SI-NEXT: ; implicit-def: $vgpr31
-; SI-NEXT: ; kill: killed $vgpr31
-; SI-NEXT: ; implicit-def: $vgpr31
-; SI-NEXT: ; kill: killed $vgpr31
-; SI-NEXT: ; implicit-def: $vgpr31
-; SI-NEXT: ; kill: killed $vgpr31
-; SI-NEXT: ; implicit-def: $vgpr31
-; SI-NEXT: ; kill: killed $vgpr31
-; SI-NEXT: ; implicit-def: $vgpr31
-; SI-NEXT: ; kill: killed $vgpr31
-; SI-NEXT: ; implicit-def: $vgpr31
-; SI-NEXT: ; kill: killed $vgpr31
-; SI-NEXT: ; implicit-def: $vgpr31
-; SI-NEXT: ; kill: killed $vgpr31
-; SI-NEXT: ; implicit-def: $vgpr31
; SI-NEXT: s_and_saveexec_b64 s[4:5], vcc
; SI-NEXT: s_xor_b64 s[4:5], exec, s[4:5]
; SI-NEXT: s_cbranch_execz .LBB20_2
; SI-NEXT: ; %bb.1: ; %cmp.false
; SI-NEXT: v_lshrrev_b32_e32 v31, 16, v62
-; SI-NEXT: v_cvt_f32_f16_e32 v32, v31
+; SI-NEXT: v_cvt_f32_f16_e32 v34, v31
; SI-NEXT: s_waitcnt vmcnt(0)
; SI-NEXT: v_lshrrev_b32_e32 v31, 16, v63
-; SI-NEXT: v_cvt_f32_f16_e32 v34, v31
+; SI-NEXT: v_cvt_f32_f16_e32 v36, v31
; SI-NEXT: v_lshrrev_b32_e32 v31, 16, v30
-; SI-NEXT: v_cvt_f32_f16_e32 v40, v31
+; SI-NEXT: v_cvt_f32_f16_e32 v53, v31
; SI-NEXT: v_lshrrev_b32_e32 v31, 16, v29
; SI-NEXT: v_cvt_f32_f16_e32 v54, v31
; SI-NEXT: v_lshrrev_b32_e32 v31, 16, v28
-; SI-NEXT: v_cvt_f32_f16_e32 v50, v31
+; SI-NEXT: v_cvt_f32_f16_e32 v52, v31
; SI-NEXT: v_lshrrev_b32_e32 v31, 16, v27
-; SI-NEXT: v_cvt_f32_f16_e32 v46, v31
+; SI-NEXT: v_cvt_f32_f16_e32 v44, v31
; SI-NEXT: v_lshrrev_b32_e32 v31, 16, v26
; SI-NEXT: v_cvt_f32_f16_e32 v31, v31
-; SI-NEXT: v_cvt_f32_f16_e32 v27, v27
-; SI-NEXT: v_cvt_f32_f16_e32 v26, v26
-; SI-NEXT: v_cvt_f32_f16_e32 v36, v63
-; SI-NEXT: buffer_store_dword v31, off, s[0:3], s32 offset:76 ; 4-byte Folded Spill
+; SI-NEXT: v_lshrrev_b32_e32 v32, 16, v8
+; SI-NEXT: v_cvt_f32_f16_e32 v38, v32
+; SI-NEXT: v_lshrrev_b32_e32 v32, 16, v7
+; SI-NEXT: buffer_store_dword v31, off, s[0:3], s32 offset:80 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_lshrrev_b32_e32 v31, 16, v25
; SI-NEXT: v_cvt_f32_f16_e32 v31, v31
-; SI-NEXT: v_cvt_f32_f16_e32 v25, v25
-; SI-NEXT: v_cvt_f32_f16_e32 v48, v30
-; SI-NEXT: v_cvt_f32_f16_e32 v52, v29
-; SI-NEXT: buffer_store_dword v31, off, s[0:3], s32 offset:84 ; 4-byte Folded Spill
+; SI-NEXT: v_cvt_f32_f16_e32 v41, v32
+; SI-NEXT: v_lshrrev_b32_e32 v32, 16, v6
+; SI-NEXT: v_cvt_f32_f16_e32 v43, v32
+; SI-NEXT: buffer_store_dword v31, off, s[0:3], s32 offset:88 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_lshrrev_b32_e32 v31, 16, v24
; SI-NEXT: v_cvt_f32_f16_e32 v31, v31
-; SI-NEXT: v_cvt_f32_f16_e32 v24, v24
-; SI-NEXT: v_cvt_f32_f16_e32 v56, v28
-; SI-NEXT: buffer_store_dword v27, off, s[0:3], s32 offset:80 ; 4-byte Folded Spill
-; SI-NEXT: buffer_store_dword v31, off, s[0:3], s32 offset:92 ; 4-byte Folded Spill
+; SI-NEXT: v_lshrrev_b32_e32 v32, 16, v5
+; SI-NEXT: v_cvt_f32_f16_e32 v45, v32
+; SI-NEXT: v_lshrrev_b32_e32 v32, 16, v4
+; SI-NEXT: buffer_store_dword v31, off, s[0:3], s32 offset:96 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_lshrrev_b32_e32 v31, 16, v23
; SI-NEXT: v_cvt_f32_f16_e32 v31, v31
-; SI-NEXT: v_cvt_f32_f16_e32 v23, v23
-; SI-NEXT: buffer_store_dword v26, off, s[0:3], s32 offset:88 ; 4-byte Folded Spill
-; SI-NEXT: buffer_store_dword v25, off, s[0:3], s32 offset:96 ; 4-byte Folded Spill
-; SI-NEXT: buffer_store_dword v31, off, s[0:3], s32 offset:100 ; 4-byte Folded Spill
+; SI-NEXT: v_cvt_f32_f16_e32 v28, v28
+; SI-NEXT: v_cvt_f32_f16_e32 v27, v27
+; SI-NEXT: v_cvt_f32_f16_e32 v26, v26
+; SI-NEXT: buffer_store_dword v31, off, s[0:3], s32 offset:104 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_lshrrev_b32_e32 v31, 16, v22
; SI-NEXT: v_cvt_f32_f16_e32 v31, v31
-; SI-NEXT: v_cvt_f32_f16_e32 v22, v22
-; SI-NEXT: buffer_store_dword v24, off, s[0:3], s32 offset:104 ; 4-byte Folded Spill
-; SI-NEXT: buffer_store_dword v23, off, s[0:3], s32 offset:108 ; 4-byte Folded Spill
-; SI-NEXT: buffer_store_dword v31, off, s[0:3], s32 offset:112 ; 4-byte Folded Spill
+; SI-NEXT: v_cvt_f32_f16_e32 v25, v25
+; SI-NEXT: v_cvt_f32_f16_e32 v24, v24
+; SI-NEXT: v_cvt_f32_f16_e32 v23, v23
+; SI-NEXT: buffer_store_dword v31, off, s[0:3], s32 offset:116 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_lshrrev_b32_e32 v31, 16, v21
; SI-NEXT: v_cvt_f32_f16_e32 v31, v31
+; SI-NEXT: v_cvt_f32_f16_e32 v22, v22
; SI-NEXT: v_cvt_f32_f16_e32 v21, v21
-; SI-NEXT: buffer_store_dword v22, off, s[0:3], s32 offset:116 ; 4-byte Folded Spill
-; SI-NEXT: v_cvt_f32_f16_e32 v44, v5
-; SI-NEXT: buffer_store_dword v31, off, s[0:3], s32 offset:120 ; 4-byte Folded Spill
+; SI-NEXT: v_cvt_f32_f16_e32 v8, v8
+; SI-NEXT: buffer_store_dword v31, off, s[0:3], s32 offset:124 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_lshrrev_b32_e32 v31, 16, v20
; SI-NEXT: v_cvt_f32_f16_e32 v31, v31
; SI-NEXT: v_cvt_f32_f16_e32 v20, v20
-; SI-NEXT: buffer_store_dword v21, off, s[0:3], s32 offset:124 ; 4-byte Folded Spill
-; SI-NEXT: v_cvt_f32_f16_e32 v42, v4
-; SI-NEXT: buffer_store_dword v31, off, s[0:3], s32 offset:128 ; 4-byte Folded Spill
+; SI-NEXT: v_cvt_f32_f16_e32 v7, v7
+; SI-NEXT: v_cvt_f32_f16_e32 v47, v32
+; SI-NEXT: buffer_store_dword v31, off, s[0:3], s32 offset:132 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_lshrrev_b32_e32 v31, 16, v19
; SI-NEXT: v_cvt_f32_f16_e32 v31, v31
; SI-NEXT: v_cvt_f32_f16_e32 v19, v19
-; SI-NEXT: buffer_store_dword v20, off, s[0:3], s32 offset:132 ; 4-byte Folded Spill
-; SI-NEXT: v_cvt_f32_f16_e32 v43, v3
-; SI-NEXT: buffer_store_dword v31, off, s[0:3], s32 offset:136 ; 4-byte Folded Spill
+; SI-NEXT: v_lshrrev_b32_e32 v32, 16, v3
+; SI-NEXT: v_cvt_f32_f16_e32 v57, v32
+; SI-NEXT: buffer_store_dword v31, off, s[0:3], s32 offset:140 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_lshrrev_b32_e32 v31, 16, v18
; SI-NEXT: v_cvt_f32_f16_e32 v31, v31
; SI-NEXT: v_cvt_f32_f16_e32 v18, v18
-; SI-NEXT: buffer_store_dword v19, off, s[0:3], s32 offset:140 ; 4-byte Folded Spill
-; SI-NEXT: v_cvt_f32_f16_e32 v58, v2
-; SI-NEXT: buffer_store_dword v31, off, s[0:3], s32 offset:144 ; 4-byte Folded Spill
+; SI-NEXT: v_lshrrev_b32_e32 v32, 16, v2
+; SI-NEXT: v_cvt_f32_f16_e32 v59, v32
+; SI-NEXT: buffer_store_dword v31, off, s[0:3], s32 offset:148 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_lshrrev_b32_e32 v31, 16, v17
; SI-NEXT: v_cvt_f32_f16_e32 v31, v31
; SI-NEXT: v_cvt_f32_f16_e32 v17, v17
-; SI-NEXT: buffer_store_dword v18, off, s[0:3], s32 offset:148 ; 4-byte Folded Spill
-; SI-NEXT: v_cvt_f32_f16_e32 v60, v1
-; SI-NEXT: buffer_store_dword v31, off, s[0:3], s32 offset:152 ; 4-byte Folded Spill
+; SI-NEXT: v_lshrrev_b32_e32 v32, 16, v1
+; SI-NEXT: v_cvt_f32_f16_e32 v61, v32
+; SI-NEXT: buffer_store_dword v31, off, s[0:3], s32 offset:156 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_lshrrev_b32_e32 v31, 16, v16
; SI-NEXT: v_cvt_f32_f16_e32 v31, v31
; SI-NEXT: v_cvt_f32_f16_e32 v16, v16
-; SI-NEXT: buffer_store_dword v17, off, s[0:3], s32 offset:156 ; 4-byte Folded Spill
-; SI-NEXT: ; implicit-def: $vgpr17
-; SI-NEXT: ; implicit-def: $vgpr18
-; SI-NEXT: ; implicit-def: $vgpr19
-; SI-NEXT: ; implicit-def: $vgpr20
-; SI-NEXT: ; implicit-def: $vgpr21
-; SI-NEXT: ; implicit-def: $vgpr22
-; SI-NEXT: ; implicit-def: $vgpr23
-; SI-NEXT: ; implicit-def: $vgpr24
-; SI-NEXT: ; implicit-def: $vgpr25
-; SI-NEXT: ; implicit-def: $vgpr26
-; SI-NEXT: ; implicit-def: $vgpr27
-; SI-NEXT: ; implicit-def: $vgpr28
-; SI-NEXT: ; implicit-def: $vgpr29
-; SI-NEXT: ; implicit-def: $vgpr30
-; SI-NEXT: ; implicit-def: $vgpr63
-; SI-NEXT: buffer_store_dword v31, off, s[0:3], s32 offset:160 ; 4-byte Folded Spill
+; SI-NEXT: v_cvt_f32_f16_e32 v32, v62
+; SI-NEXT: v_cvt_f32_f16_e32 v48, v63
+; SI-NEXT: buffer_store_dword v31, off, s[0:3], s32 offset:164 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_lshrrev_b32_e32 v31, 16, v15
; SI-NEXT: v_cvt_f32_f16_e32 v33, v31
@@ -32098,16 +32097,6 @@ define <64 x half> @bitcast_v32i32_to_v64f16(<32 x i32> %a, i32 %b) {
; SI-NEXT: v_lshrrev_b32_e32 v31, 16, v10
; SI-NEXT: v_cvt_f32_f16_e32 v51, v31
; SI-NEXT: v_lshrrev_b32_e32 v31, 16, v9
-; SI-NEXT: v_cvt_f32_f16_e32 v53, v31
-; SI-NEXT: v_lshrrev_b32_e32 v31, 16, v8
-; SI-NEXT: v_cvt_f32_f16_e32 v55, v31
-; SI-NEXT: v_lshrrev_b32_e32 v31, 16, v7
-; SI-NEXT: v_cvt_f32_f16_e32 v41, v31
-; SI-NEXT: v_lshrrev_b32_e32 v31, 16, v6
-; SI-NEXT: v_cvt_f32_f16_e32 v38, v31
-; SI-NEXT: v_lshrrev_b32_e32 v31, 16, v5
-; SI-NEXT: v_cvt_f32_f16_e32 v45, v31
-; SI-NEXT: v_lshrrev_b32_e32 v31, 16, v4
; SI-NEXT: v_cvt_f32_f16_e32 v15, v15
; SI-NEXT: v_cvt_f32_f16_e32 v14, v14
; SI-NEXT: v_cvt_f32_f16_e32 v13, v13
@@ -32115,28 +32104,37 @@ define <64 x half> @bitcast_v32i32_to_v64f16(<32 x i32> %a, i32 %b) {
; SI-NEXT: v_cvt_f32_f16_e32 v11, v11
; SI-NEXT: v_cvt_f32_f16_e32 v10, v10
; SI-NEXT: v_cvt_f32_f16_e32 v9, v9
-; SI-NEXT: v_cvt_f32_f16_e32 v8, v8
-; SI-NEXT: v_cvt_f32_f16_e32 v7, v7
-; SI-NEXT: v_cvt_f32_f16_e32 v6, v6
-; SI-NEXT: v_cvt_f32_f16_e32 v47, v31
-; SI-NEXT: v_lshrrev_b32_e32 v31, 16, v3
-; SI-NEXT: v_cvt_f32_f16_e32 v57, v31
-; SI-NEXT: v_lshrrev_b32_e32 v31, 16, v2
-; SI-NEXT: v_cvt_f32_f16_e32 v59, v31
-; SI-NEXT: v_lshrrev_b32_e32 v31, 16, v1
-; SI-NEXT: v_cvt_f32_f16_e32 v61, v31
-; SI-NEXT: v_cvt_f32_f16_e32 v31, v62
-; SI-NEXT: buffer_store_dword v16, off, s[0:3], s32 offset:164 ; 4-byte Folded Spill
-; SI-NEXT: buffer_store_dword v15, off, s[0:3], s32 offset:168 ; 4-byte Folded Spill
-; SI-NEXT: buffer_store_dword v14, off, s[0:3], s32 offset:172 ; 4-byte Folded Spill
-; SI-NEXT: buffer_store_dword v13, off, s[0:3], s32 offset:176 ; 4-byte Folded Spill
-; SI-NEXT: buffer_store_dword v12, off, s[0:3], s32 offset:180 ; 4-byte Folded Spill
-; SI-NEXT: buffer_store_dword v11, off, s[0:3], s32 offset:184 ; 4-byte Folded Spill
-; SI-NEXT: buffer_store_dword v10, off, s[0:3], s32 offset:188 ; 4-byte Folded Spill
-; SI-NEXT: buffer_store_dword v9, off, s[0:3], s32 offset:192 ; 4-byte Folded Spill
-; SI-NEXT: buffer_store_dword v8, off, s[0:3], s32 offset:196 ; 4-byte Folded Spill
-; SI-NEXT: buffer_store_dword v7, off, s[0:3], s32 offset:200 ; 4-byte Folded Spill
-; SI-NEXT: buffer_store_dword v6, off, s[0:3], s32 offset:204 ; 4-byte Folded Spill
+; SI-NEXT: v_cvt_f32_f16_e32 v31, v31
+; SI-NEXT: v_cvt_f32_f16_e32 v50, v30
+; SI-NEXT: v_cvt_f32_f16_e32 v55, v29
+; SI-NEXT: buffer_store_dword v28, off, s[0:3], s32 offset:76 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v27, off, s[0:3], s32 offset:84 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v26, off, s[0:3], s32 offset:92 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v25, off, s[0:3], s32 offset:100 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v24, off, s[0:3], s32 offset:108 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v23, off, s[0:3], s32 offset:112 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v22, off, s[0:3], s32 offset:120 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v21, off, s[0:3], s32 offset:128 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v20, off, s[0:3], s32 offset:136 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v19, off, s[0:3], s32 offset:144 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v18, off, s[0:3], s32 offset:152 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v17, off, s[0:3], s32 offset:160 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v16, off, s[0:3], s32 offset:168 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v15, off, s[0:3], s32 offset:172 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v14, off, s[0:3], s32 offset:176 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v13, off, s[0:3], s32 offset:180 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v12, off, s[0:3], s32 offset:184 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v11, off, s[0:3], s32 offset:188 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v10, off, s[0:3], s32 offset:192 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v9, off, s[0:3], s32 offset:196 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v8, off, s[0:3], s32 offset:200 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v7, off, s[0:3], s32 offset:204 ; 4-byte Folded Spill
+; SI-NEXT: v_cvt_f32_f16_e32 v42, v6
+; SI-NEXT: v_cvt_f32_f16_e32 v40, v5
+; SI-NEXT: v_cvt_f32_f16_e32 v46, v4
+; SI-NEXT: v_cvt_f32_f16_e32 v56, v3
+; SI-NEXT: v_cvt_f32_f16_e32 v58, v2
+; SI-NEXT: v_cvt_f32_f16_e32 v60, v1
; SI-NEXT: ; implicit-def: $vgpr1
; SI-NEXT: ; implicit-def: $vgpr2
; SI-NEXT: ; implicit-def: $vgpr3
@@ -32153,7 +32151,22 @@ define <64 x half> @bitcast_v32i32_to_v64f16(<32 x i32> %a, i32 %b) {
; SI-NEXT: ; implicit-def: $vgpr14
; SI-NEXT: ; implicit-def: $vgpr15
; SI-NEXT: ; implicit-def: $vgpr16
+; SI-NEXT: ; implicit-def: $vgpr17
+; SI-NEXT: ; implicit-def: $vgpr18
+; SI-NEXT: ; implicit-def: $vgpr19
+; SI-NEXT: ; implicit-def: $vgpr20
+; SI-NEXT: ; implicit-def: $vgpr21
+; SI-NEXT: ; implicit-def: $vgpr22
+; SI-NEXT: ; implicit-def: $vgpr23
+; SI-NEXT: ; implicit-def: $vgpr24
+; SI-NEXT: ; implicit-def: $vgpr25
+; SI-NEXT: ; implicit-def: $vgpr26
+; SI-NEXT: ; implicit-def: $vgpr27
+; SI-NEXT: ; implicit-def: $vgpr28
+; SI-NEXT: ; implicit-def: $vgpr29
+; SI-NEXT: ; implicit-def: $vgpr30
; SI-NEXT: ; implicit-def: $vgpr62
+; SI-NEXT: ; implicit-def: $vgpr63
; SI-NEXT: .LBB20_2: ; %Flow
; SI-NEXT: s_andn2_saveexec_b64 s[4:5], s[4:5]
; SI-NEXT: s_cbranch_execz .LBB20_4
@@ -32161,87 +32174,78 @@ define <64 x half> @bitcast_v32i32_to_v64f16(<32 x i32> %a, i32 %b) {
; SI-NEXT: v_add_i32_e32 v5, vcc, 3, v5
; SI-NEXT: v_add_i32_e32 v33, vcc, 3, v1
; SI-NEXT: v_lshrrev_b32_e32 v1, 16, v5
-; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_add_i32_e32 v6, vcc, 3, v6
; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:208 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_lshrrev_b32_e32 v1, 16, v6
-; SI-NEXT: v_add_i32_e32 v28, vcc, 3, v28
+; SI-NEXT: v_add_i32_e32 v7, vcc, 3, v7
; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:212 ; 4-byte Folded Spill
-; SI-NEXT: v_add_i32_e32 v9, vcc, 3, v9
-; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_lshrrev_b32_e32 v1, 16, v28
-; SI-NEXT: v_lshrrev_b32_e32 v53, 16, v9
-; SI-NEXT: v_cvt_f32_f16_e32 v9, v9
-; SI-NEXT: v_cvt_f32_f16_e32 v1, v1
-; SI-NEXT: v_add_i32_e32 v25, vcc, 3, v25
-; SI-NEXT: v_add_i32_e32 v26, vcc, 3, v26
-; SI-NEXT: v_lshrrev_b32_e32 v48, 16, v25
-; SI-NEXT: v_lshrrev_b32_e32 v50, 16, v26
-; SI-NEXT: buffer_store_dword v9, off, s[0:3], s32 offset:192 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_cvt_f32_f16_e32 v9, v50
-; SI-NEXT: v_mov_b32_e32 v50, v1
-; SI-NEXT: v_cvt_f32_f16_e32 v1, v48
-; SI-NEXT: v_add_i32_e32 v24, vcc, 3, v24
-; SI-NEXT: v_lshrrev_b32_e32 v38, 16, v24
-; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:84 ; 4-byte Folded Spill
-; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_cvt_f32_f16_e32 v1, v38
-; SI-NEXT: v_add_i32_e32 v23, vcc, 3, v23
-; SI-NEXT: v_lshrrev_b32_e32 v56, 16, v23
-; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:92 ; 4-byte Folded Spill
-; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_cvt_f32_f16_e32 v1, v56
-; SI-NEXT: v_add_i32_e32 v22, vcc, 3, v22
-; SI-NEXT: v_lshrrev_b32_e32 v34, 16, v22
-; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:100 ; 4-byte Folded Spill
-; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_cvt_f32_f16_e32 v1, v34
+; SI-NEXT: v_lshrrev_b32_e32 v1, 16, v7
; SI-NEXT: v_add_i32_e32 v21, vcc, 3, v21
-; SI-NEXT: v_lshrrev_b32_e32 v32, 16, v21
-; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:112 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:216 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_cvt_f32_f16_e32 v1, v32
+; SI-NEXT: v_lshrrev_b32_e32 v1, 16, v21
+; SI-NEXT: v_cvt_f32_f16_e32 v1, v1
; SI-NEXT: v_add_i32_e32 v20, vcc, 3, v20
-; SI-NEXT: v_lshrrev_b32_e32 v46, 16, v20
-; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:120 ; 4-byte Folded Spill
+; SI-NEXT: v_lshrrev_b32_e32 v31, 16, v20
+; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:124 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_cvt_f32_f16_e32 v1, v46
+; SI-NEXT: v_cvt_f32_f16_e32 v1, v31
; SI-NEXT: v_add_i32_e32 v19, vcc, 3, v19
; SI-NEXT: v_lshrrev_b32_e32 v61, 16, v19
-; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:128 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:132 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_cvt_f32_f16_e32 v1, v61
; SI-NEXT: v_add_i32_e32 v18, vcc, 3, v18
; SI-NEXT: v_lshrrev_b32_e32 v59, 16, v18
-; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:136 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:140 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_cvt_f32_f16_e32 v1, v59
; SI-NEXT: v_add_i32_e32 v17, vcc, 3, v17
; SI-NEXT: v_lshrrev_b32_e32 v57, 16, v17
-; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:144 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:148 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_cvt_f32_f16_e32 v1, v57
; SI-NEXT: v_add_i32_e32 v16, vcc, 3, v16
; SI-NEXT: v_lshrrev_b32_e32 v47, 16, v16
-; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:152 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:156 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_cvt_f32_f16_e32 v1, v47
-; SI-NEXT: v_add_i32_e32 v7, vcc, 3, v7
+; SI-NEXT: v_add_i32_e32 v9, vcc, 3, v9
+; SI-NEXT: v_lshrrev_b32_e32 v41, 16, v9
+; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:164 ; 4-byte Folded Spill
+; SI-NEXT: s_waitcnt expcnt(0)
+; SI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:216 ; 4-byte Folded Reload
+; SI-NEXT: v_cvt_f32_f16_e32 v31, v41
+; SI-NEXT: v_cvt_f32_f16_e32 v9, v9
+; SI-NEXT: v_add_i32_e32 v26, vcc, 3, v26
+; SI-NEXT: v_lshrrev_b32_e32 v55, 16, v26
+; SI-NEXT: buffer_store_dword v9, off, s[0:3], s32 offset:196 ; 4-byte Folded Spill
+; SI-NEXT: s_waitcnt expcnt(0)
+; SI-NEXT: v_cvt_f32_f16_e32 v9, v55
; SI-NEXT: v_add_i32_e32 v8, vcc, 3, v8
-; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:160 ; 4-byte Folded Spill
+; SI-NEXT: v_add_i32_e32 v25, vcc, 3, v25
+; SI-NEXT: v_lshrrev_b32_e32 v43, 16, v8
+; SI-NEXT: v_lshrrev_b32_e32 v38, 16, v25
+; SI-NEXT: buffer_store_dword v9, off, s[0:3], s32 offset:80 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:212 ; 4-byte Folded Reload
-; SI-NEXT: v_lshrrev_b32_e32 v41, 16, v7
-; SI-NEXT: v_lshrrev_b32_e32 v55, 16, v8
-; SI-NEXT: v_cvt_f32_f16_e32 v8, v8
+; SI-NEXT: v_cvt_f32_f16_e32 v9, v38
+; SI-NEXT: v_cvt_f32_f16_e32 v38, v43
; SI-NEXT: v_cvt_f32_f16_e32 v7, v7
-; SI-NEXT: v_cvt_f32_f16_e32 v6, v6
+; SI-NEXT: s_waitcnt vmcnt(12)
+; SI-NEXT: v_add_i32_e32 v40, vcc, 3, v63
+; SI-NEXT: v_add_i32_e32 v42, vcc, 3, v62
+; SI-NEXT: v_lshrrev_b32_e32 v63, 16, v40
+; SI-NEXT: v_lshrrev_b32_e32 v62, 16, v42
+; SI-NEXT: v_cvt_f32_f16_e32 v8, v8
+; SI-NEXT: v_cvt_f32_f16_e32 v32, v42
+; SI-NEXT: buffer_store_dword v7, off, s[0:3], s32 offset:204 ; 4-byte Folded Spill
+; SI-NEXT: v_cvt_f32_f16_e32 v42, v6
+; SI-NEXT: v_cvt_f32_f16_e32 v6, v62
+; SI-NEXT: s_waitcnt expcnt(0)
+; SI-NEXT: v_cvt_f32_f16_e32 v7, v63
; SI-NEXT: v_add_i32_e32 v27, vcc, 3, v27
-; SI-NEXT: s_waitcnt vmcnt(14)
-; SI-NEXT: v_add_i32_e32 v42, vcc, 3, v63
-; SI-NEXT: v_add_i32_e32 v44, vcc, 3, v62
; SI-NEXT: v_add_i32_e32 v2, vcc, 3, v2
; SI-NEXT: v_add_i32_e32 v3, vcc, 3, v3
; SI-NEXT: v_add_i32_e32 v4, vcc, 3, v4
@@ -32251,39 +32255,43 @@ define <64 x half> @bitcast_v32i32_to_v64f16(<32 x i32> %a, i32 %b) {
; SI-NEXT: v_add_i32_e32 v13, vcc, 3, v13
; SI-NEXT: v_add_i32_e32 v14, vcc, 3, v14
; SI-NEXT: v_add_i32_e32 v15, vcc, 3, v15
-; SI-NEXT: v_lshrrev_b32_e32 v52, 16, v27
+; SI-NEXT: v_add_i32_e32 v22, vcc, 3, v22
+; SI-NEXT: v_add_i32_e32 v23, vcc, 3, v23
+; SI-NEXT: v_add_i32_e32 v24, vcc, 3, v24
+; SI-NEXT: v_lshrrev_b32_e32 v50, 16, v27
+; SI-NEXT: v_add_i32_e32 v28, vcc, 3, v28
; SI-NEXT: v_add_i32_e32 v29, vcc, 3, v29
; SI-NEXT: v_add_i32_e32 v30, vcc, 3, v30
-; SI-NEXT: v_lshrrev_b32_e32 v63, 16, v42
-; SI-NEXT: v_lshrrev_b32_e32 v62, 16, v44
; SI-NEXT: v_lshrrev_b32_e32 v51, 16, v10
; SI-NEXT: v_lshrrev_b32_e32 v49, 16, v11
; SI-NEXT: v_lshrrev_b32_e32 v39, 16, v12
; SI-NEXT: v_lshrrev_b32_e32 v37, 16, v13
; SI-NEXT: v_lshrrev_b32_e32 v35, 16, v14
; SI-NEXT: v_lshrrev_b32_e32 v45, 16, v15
+; SI-NEXT: v_lshrrev_b32_e32 v44, 16, v22
+; SI-NEXT: v_lshrrev_b32_e32 v34, 16, v23
+; SI-NEXT: v_lshrrev_b32_e32 v36, 16, v24
+; SI-NEXT: v_lshrrev_b32_e32 v52, 16, v28
; SI-NEXT: v_lshrrev_b32_e32 v54, 16, v29
-; SI-NEXT: v_lshrrev_b32_e32 v40, 16, v30
-; SI-NEXT: v_cvt_f32_f16_e32 v31, v44
-; SI-NEXT: v_cvt_f32_f16_e32 v36, v42
+; SI-NEXT: v_lshrrev_b32_e32 v53, 16, v30
+; SI-NEXT: v_cvt_f32_f16_e32 v48, v40
; SI-NEXT: v_cvt_f32_f16_e32 v30, v30
; SI-NEXT: v_cvt_f32_f16_e32 v29, v29
-; SI-NEXT: v_cvt_f32_f16_e32 v28, v28
-; SI-NEXT: buffer_store_dword v8, off, s[0:3], s32 offset:196 ; 4-byte Folded Spill
-; SI-NEXT: buffer_store_dword v7, off, s[0:3], s32 offset:200 ; 4-byte Folded Spill
-; SI-NEXT: buffer_store_dword v6, off, s[0:3], s32 offset:204 ; 4-byte Folded Spill
-; SI-NEXT: v_cvt_f32_f16_e32 v44, v5
+; SI-NEXT: buffer_store_dword v8, off, s[0:3], s32 offset:200 ; 4-byte Folded Spill
+; SI-NEXT: v_cvt_f32_f16_e32 v40, v5
+; SI-NEXT: s_waitcnt vmcnt(4)
+; SI-NEXT: v_cvt_f32_f16_e32 v41, v1
+; SI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:212 ; 4-byte Folded Reload
; SI-NEXT: v_lshrrev_b32_e32 v5, 16, v4
-; SI-NEXT: v_cvt_f32_f16_e32 v42, v4
+; SI-NEXT: v_cvt_f32_f16_e32 v46, v4
; SI-NEXT: v_lshrrev_b32_e32 v4, 16, v3
-; SI-NEXT: v_cvt_f32_f16_e32 v43, v3
+; SI-NEXT: v_cvt_f32_f16_e32 v56, v3
; SI-NEXT: v_lshrrev_b32_e32 v3, 16, v2
; SI-NEXT: v_cvt_f32_f16_e32 v58, v2
; SI-NEXT: v_lshrrev_b32_e32 v2, 16, v33
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_cvt_f32_f16_e32 v6, v62
-; SI-NEXT: v_cvt_f32_f16_e32 v7, v63
-; SI-NEXT: v_cvt_f32_f16_e32 v8, v52
+; SI-NEXT: v_cvt_f32_f16_e32 v8, v50
+; SI-NEXT: v_cvt_f32_f16_e32 v28, v28
; SI-NEXT: v_cvt_f32_f16_e32 v27, v27
; SI-NEXT: v_cvt_f32_f16_e32 v26, v26
; SI-NEXT: v_cvt_f32_f16_e32 v25, v25
@@ -32300,53 +32308,58 @@ define <64 x half> @bitcast_v32i32_to_v64f16(<32 x i32> %a, i32 %b) {
; SI-NEXT: v_cvt_f32_f16_e32 v14, v14
; SI-NEXT: v_cvt_f32_f16_e32 v13, v13
; SI-NEXT: v_cvt_f32_f16_e32 v12, v12
-; SI-NEXT: s_waitcnt vmcnt(3)
-; SI-NEXT: v_cvt_f32_f16_e32 v38, v1
-; SI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:208 ; 4-byte Folded Reload
; SI-NEXT: v_cvt_f32_f16_e32 v11, v11
; SI-NEXT: v_cvt_f32_f16_e32 v10, v10
; SI-NEXT: v_cvt_f32_f16_e32 v60, v33
-; SI-NEXT: v_cvt_f32_f16_e32 v40, v40
+; SI-NEXT: v_cvt_f32_f16_e32 v53, v53
; SI-NEXT: v_cvt_f32_f16_e32 v54, v54
+; SI-NEXT: v_cvt_f32_f16_e32 v52, v52
+; SI-NEXT: buffer_store_dword v9, off, s[0:3], s32 offset:88 ; 4-byte Folded Spill
+; SI-NEXT: s_waitcnt expcnt(0)
+; SI-NEXT: v_cvt_f32_f16_e32 v9, v36
+; SI-NEXT: v_mov_b32_e32 v36, v7
+; SI-NEXT: v_cvt_f32_f16_e32 v7, v34
+; SI-NEXT: v_mov_b32_e32 v34, v6
+; SI-NEXT: v_cvt_f32_f16_e32 v6, v44
; SI-NEXT: v_cvt_f32_f16_e32 v33, v45
; SI-NEXT: v_cvt_f32_f16_e32 v35, v35
; SI-NEXT: v_cvt_f32_f16_e32 v37, v37
; SI-NEXT: v_cvt_f32_f16_e32 v39, v39
; SI-NEXT: v_cvt_f32_f16_e32 v49, v49
; SI-NEXT: v_cvt_f32_f16_e32 v51, v51
-; SI-NEXT: v_cvt_f32_f16_e32 v53, v53
-; SI-NEXT: v_cvt_f32_f16_e32 v55, v55
-; SI-NEXT: v_cvt_f32_f16_e32 v41, v41
; SI-NEXT: v_cvt_f32_f16_e32 v47, v5
; SI-NEXT: v_cvt_f32_f16_e32 v57, v4
; SI-NEXT: v_cvt_f32_f16_e32 v59, v3
; SI-NEXT: v_cvt_f32_f16_e32 v61, v2
-; SI-NEXT: v_mov_b32_e32 v52, v29
-; SI-NEXT: v_mov_b32_e32 v48, v30
-; SI-NEXT: v_mov_b32_e32 v56, v28
-; SI-NEXT: v_mov_b32_e32 v34, v7
-; SI-NEXT: v_mov_b32_e32 v32, v6
-; SI-NEXT: v_mov_b32_e32 v46, v8
-; SI-NEXT: buffer_store_dword v27, off, s[0:3], s32 offset:80 ; 4-byte Folded Spill
-; SI-NEXT: buffer_store_dword v26, off, s[0:3], s32 offset:88 ; 4-byte Folded Spill
-; SI-NEXT: buffer_store_dword v25, off, s[0:3], s32 offset:96 ; 4-byte Folded Spill
-; SI-NEXT: buffer_store_dword v24, off, s[0:3], s32 offset:104 ; 4-byte Folded Spill
-; SI-NEXT: buffer_store_dword v23, off, s[0:3], s32 offset:108 ; 4-byte Folded Spill
-; SI-NEXT: buffer_store_dword v22, off, s[0:3], s32 offset:116 ; 4-byte Folded Spill
-; SI-NEXT: buffer_store_dword v21, off, s[0:3], s32 offset:124 ; 4-byte Folded Spill
-; SI-NEXT: buffer_store_dword v20, off, s[0:3], s32 offset:132 ; 4-byte Folded Spill
-; SI-NEXT: buffer_store_dword v19, off, s[0:3], s32 offset:140 ; 4-byte Folded Spill
-; SI-NEXT: buffer_store_dword v18, off, s[0:3], s32 offset:148 ; 4-byte Folded Spill
-; SI-NEXT: buffer_store_dword v17, off, s[0:3], s32 offset:156 ; 4-byte Folded Spill
-; SI-NEXT: buffer_store_dword v16, off, s[0:3], s32 offset:164 ; 4-byte Folded Spill
-; SI-NEXT: buffer_store_dword v15, off, s[0:3], s32 offset:168 ; 4-byte Folded Spill
-; SI-NEXT: buffer_store_dword v14, off, s[0:3], s32 offset:172 ; 4-byte Folded Spill
-; SI-NEXT: buffer_store_dword v13, off, s[0:3], s32 offset:176 ; 4-byte Folded Spill
-; SI-NEXT: buffer_store_dword v12, off, s[0:3], s32 offset:180 ; 4-byte Folded Spill
-; SI-NEXT: buffer_store_dword v11, off, s[0:3], s32 offset:184 ; 4-byte Folded Spill
-; SI-NEXT: buffer_store_dword v10, off, s[0:3], s32 offset:188 ; 4-byte Folded Spill
-; SI-NEXT: buffer_store_dword v9, off, s[0:3], s32 offset:76 ; 4-byte Folded Spill
+; SI-NEXT: v_mov_b32_e32 v50, v30
+; SI-NEXT: v_mov_b32_e32 v55, v29
+; SI-NEXT: v_mov_b32_e32 v44, v8
+; SI-NEXT: buffer_store_dword v28, off, s[0:3], s32 offset:76 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v27, off, s[0:3], s32 offset:84 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v26, off, s[0:3], s32 offset:92 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v25, off, s[0:3], s32 offset:100 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v24, off, s[0:3], s32 offset:108 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v23, off, s[0:3], s32 offset:112 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v22, off, s[0:3], s32 offset:120 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v21, off, s[0:3], s32 offset:128 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v20, off, s[0:3], s32 offset:136 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v19, off, s[0:3], s32 offset:144 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v18, off, s[0:3], s32 offset:152 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v17, off, s[0:3], s32 offset:160 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v16, off, s[0:3], s32 offset:168 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v15, off, s[0:3], s32 offset:172 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v14, off, s[0:3], s32 offset:176 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v13, off, s[0:3], s32 offset:180 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v12, off, s[0:3], s32 offset:184 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v11, off, s[0:3], s32 offset:188 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v10, off, s[0:3], s32 offset:192 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v9, off, s[0:3], s32 offset:96 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v7, off, s[0:3], s32 offset:104 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v6, off, s[0:3], s32 offset:116 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt vmcnt(14)
+; SI-NEXT: v_cvt_f32_f16_e32 v43, v1
+; SI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:208 ; 4-byte Folded Reload
+; SI-NEXT: s_waitcnt vmcnt(0)
; SI-NEXT: v_cvt_f32_f16_e32 v45, v1
; SI-NEXT: .LBB20_4: ; %end
; SI-NEXT: s_or_b64 exec, exec, s[4:5]
@@ -32365,34 +32378,32 @@ define <64 x half> @bitcast_v32i32_to_v64f16(<32 x i32> %a, i32 %b) {
; SI-NEXT: buffer_store_dword v1, v2, s[0:3], 0 offen
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_cvt_f16_f32_e32 v1, v57
-; SI-NEXT: v_cvt_f16_f32_e32 v2, v43
+; SI-NEXT: v_cvt_f16_f32_e32 v2, v56
; SI-NEXT: v_lshlrev_b32_e32 v1, 16, v1
; SI-NEXT: v_or_b32_e32 v1, v2, v1
; SI-NEXT: v_add_i32_e32 v2, vcc, 8, v0
; SI-NEXT: buffer_store_dword v1, v2, s[0:3], 0 offen
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_cvt_f16_f32_e32 v1, v47
-; SI-NEXT: v_cvt_f16_f32_e32 v2, v42
+; SI-NEXT: v_cvt_f16_f32_e32 v2, v46
; SI-NEXT: v_lshlrev_b32_e32 v1, 16, v1
; SI-NEXT: v_or_b32_e32 v1, v2, v1
; SI-NEXT: buffer_store_dword v1, v3, s[0:3], 0 offen
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_cvt_f16_f32_e32 v1, v45
-; SI-NEXT: v_cvt_f16_f32_e32 v2, v44
+; SI-NEXT: v_cvt_f16_f32_e32 v2, v40
; SI-NEXT: v_add_i32_e32 v3, vcc, 16, v0
; SI-NEXT: v_lshlrev_b32_e32 v1, 16, v1
; SI-NEXT: v_or_b32_e32 v1, v2, v1
; SI-NEXT: buffer_store_dword v1, v3, s[0:3], 0 offen
-; SI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:204 ; 4-byte Folded Reload
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_cvt_f16_f32_e32 v1, v38
+; SI-NEXT: v_cvt_f16_f32_e32 v1, v43
+; SI-NEXT: v_cvt_f16_f32_e32 v2, v42
; SI-NEXT: v_add_i32_e32 v3, vcc, 20, v0
; SI-NEXT: v_lshlrev_b32_e32 v1, 16, v1
-; SI-NEXT: s_waitcnt vmcnt(0)
-; SI-NEXT: v_cvt_f16_f32_e32 v2, v2
; SI-NEXT: v_or_b32_e32 v1, v2, v1
; SI-NEXT: buffer_store_dword v1, v3, s[0:3], 0 offen
-; SI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:200 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:204 ; 4-byte Folded Reload
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_cvt_f16_f32_e32 v1, v41
; SI-NEXT: v_add_i32_e32 v3, vcc, 24, v0
@@ -32401,25 +32412,25 @@ define <64 x half> @bitcast_v32i32_to_v64f16(<32 x i32> %a, i32 %b) {
; SI-NEXT: v_cvt_f16_f32_e32 v2, v2
; SI-NEXT: v_or_b32_e32 v1, v2, v1
; SI-NEXT: buffer_store_dword v1, v3, s[0:3], 0 offen
-; SI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:196 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:200 ; 4-byte Folded Reload
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_cvt_f16_f32_e32 v1, v55
+; SI-NEXT: v_cvt_f16_f32_e32 v1, v38
; SI-NEXT: v_add_i32_e32 v3, vcc, 28, v0
; SI-NEXT: v_lshlrev_b32_e32 v1, 16, v1
; SI-NEXT: s_waitcnt vmcnt(0)
; SI-NEXT: v_cvt_f16_f32_e32 v2, v2
; SI-NEXT: v_or_b32_e32 v1, v2, v1
; SI-NEXT: buffer_store_dword v1, v3, s[0:3], 0 offen
-; SI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:192 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:196 ; 4-byte Folded Reload
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_cvt_f16_f32_e32 v1, v53
+; SI-NEXT: v_cvt_f16_f32_e32 v1, v31
; SI-NEXT: v_add_i32_e32 v3, vcc, 32, v0
; SI-NEXT: v_lshlrev_b32_e32 v1, 16, v1
; SI-NEXT: s_waitcnt vmcnt(0)
; SI-NEXT: v_cvt_f16_f32_e32 v2, v2
; SI-NEXT: v_or_b32_e32 v1, v2, v1
; SI-NEXT: buffer_store_dword v1, v3, s[0:3], 0 offen
-; SI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:188 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:192 ; 4-byte Folded Reload
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_cvt_f16_f32_e32 v1, v51
; SI-NEXT: v_add_i32_e32 v3, vcc, 36, v0
@@ -32428,7 +32439,7 @@ define <64 x half> @bitcast_v32i32_to_v64f16(<32 x i32> %a, i32 %b) {
; SI-NEXT: v_cvt_f16_f32_e32 v2, v2
; SI-NEXT: v_or_b32_e32 v1, v2, v1
; SI-NEXT: buffer_store_dword v1, v3, s[0:3], 0 offen
-; SI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:184 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:188 ; 4-byte Folded Reload
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_cvt_f16_f32_e32 v1, v49
; SI-NEXT: v_add_i32_e32 v3, vcc, 40, v0
@@ -32437,7 +32448,7 @@ define <64 x half> @bitcast_v32i32_to_v64f16(<32 x i32> %a, i32 %b) {
; SI-NEXT: v_cvt_f16_f32_e32 v2, v2
; SI-NEXT: v_or_b32_e32 v1, v2, v1
; SI-NEXT: buffer_store_dword v1, v3, s[0:3], 0 offen
-; SI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:180 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:184 ; 4-byte Folded Reload
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_cvt_f16_f32_e32 v1, v39
; SI-NEXT: v_add_i32_e32 v3, vcc, 44, v0
@@ -32446,7 +32457,7 @@ define <64 x half> @bitcast_v32i32_to_v64f16(<32 x i32> %a, i32 %b) {
; SI-NEXT: v_cvt_f16_f32_e32 v2, v2
; SI-NEXT: v_or_b32_e32 v1, v2, v1
; SI-NEXT: buffer_store_dword v1, v3, s[0:3], 0 offen
-; SI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:176 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:180 ; 4-byte Folded Reload
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_cvt_f16_f32_e32 v1, v37
; SI-NEXT: v_add_i32_e32 v3, vcc, 48, v0
@@ -32455,7 +32466,7 @@ define <64 x half> @bitcast_v32i32_to_v64f16(<32 x i32> %a, i32 %b) {
; SI-NEXT: v_cvt_f16_f32_e32 v2, v2
; SI-NEXT: v_or_b32_e32 v1, v2, v1
; SI-NEXT: buffer_store_dword v1, v3, s[0:3], 0 offen
-; SI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:172 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:176 ; 4-byte Folded Reload
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_cvt_f16_f32_e32 v1, v35
; SI-NEXT: v_add_i32_e32 v3, vcc, 52, v0
@@ -32464,7 +32475,7 @@ define <64 x half> @bitcast_v32i32_to_v64f16(<32 x i32> %a, i32 %b) {
; SI-NEXT: v_cvt_f16_f32_e32 v2, v2
; SI-NEXT: v_or_b32_e32 v1, v2, v1
; SI-NEXT: buffer_store_dword v1, v3, s[0:3], 0 offen
-; SI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:168 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:172 ; 4-byte Folded Reload
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_cvt_f16_f32_e32 v1, v33
; SI-NEXT: v_add_i32_e32 v3, vcc, 56, v0
@@ -32474,8 +32485,8 @@ define <64 x half> @bitcast_v32i32_to_v64f16(<32 x i32> %a, i32 %b) {
; SI-NEXT: v_or_b32_e32 v1, v2, v1
; SI-NEXT: buffer_store_dword v1, v3, s[0:3], 0 offen
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:160 ; 4-byte Folded Reload
-; SI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:164 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:164 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:168 ; 4-byte Folded Reload
; SI-NEXT: v_add_i32_e32 v3, vcc, 60, v0
; SI-NEXT: s_waitcnt vmcnt(1)
; SI-NEXT: v_cvt_f16_f32_e32 v1, v1
@@ -32485,8 +32496,8 @@ define <64 x half> @bitcast_v32i32_to_v64f16(<32 x i32> %a, i32 %b) {
; SI-NEXT: v_or_b32_e32 v1, v2, v1
; SI-NEXT: buffer_store_dword v1, v3, s[0:3], 0 offen
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:152 ; 4-byte Folded Reload
-; SI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:156 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:156 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:160 ; 4-byte Folded Reload
; SI-NEXT: v_add_i32_e32 v3, vcc, 64, v0
; SI-NEXT: s_waitcnt vmcnt(1)
; SI-NEXT: v_cvt_f16_f32_e32 v1, v1
@@ -32496,8 +32507,8 @@ define <64 x half> @bitcast_v32i32_to_v64f16(<32 x i32> %a, i32 %b) {
; SI-NEXT: v_or_b32_e32 v1, v2, v1
; SI-NEXT: buffer_store_dword v1, v3, s[0:3], 0 offen
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:144 ; 4-byte Folded Reload
-; SI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:148 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:148 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:152 ; 4-byte Folded Reload
; SI-NEXT: v_add_i32_e32 v3, vcc, 0x44, v0
; SI-NEXT: s_waitcnt vmcnt(1)
; SI-NEXT: v_cvt_f16_f32_e32 v1, v1
@@ -32507,8 +32518,8 @@ define <64 x half> @bitcast_v32i32_to_v64f16(<32 x i32> %a, i32 %b) {
; SI-NEXT: v_or_b32_e32 v1, v2, v1
; SI-NEXT: buffer_store_dword v1, v3, s[0:3], 0 offen
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:136 ; 4-byte Folded Reload
-; SI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:140 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:140 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:144 ; 4-byte Folded Reload
; SI-NEXT: v_add_i32_e32 v3, vcc, 0x48, v0
; SI-NEXT: s_waitcnt vmcnt(1)
; SI-NEXT: v_cvt_f16_f32_e32 v1, v1
@@ -32518,8 +32529,8 @@ define <64 x half> @bitcast_v32i32_to_v64f16(<32 x i32> %a, i32 %b) {
; SI-NEXT: v_or_b32_e32 v1, v2, v1
; SI-NEXT: buffer_store_dword v1, v3, s[0:3], 0 offen
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:128 ; 4-byte Folded Reload
-; SI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:132 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:132 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:136 ; 4-byte Folded Reload
; SI-NEXT: v_add_i32_e32 v3, vcc, 0x4c, v0
; SI-NEXT: s_waitcnt vmcnt(1)
; SI-NEXT: v_cvt_f16_f32_e32 v1, v1
@@ -32529,8 +32540,8 @@ define <64 x half> @bitcast_v32i32_to_v64f16(<32 x i32> %a, i32 %b) {
; SI-NEXT: v_or_b32_e32 v1, v2, v1
; SI-NEXT: buffer_store_dword v1, v3, s[0:3], 0 offen
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:120 ; 4-byte Folded Reload
-; SI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:124 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:124 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:128 ; 4-byte Folded Reload
; SI-NEXT: v_add_i32_e32 v3, vcc, 0x50, v0
; SI-NEXT: s_waitcnt vmcnt(1)
; SI-NEXT: v_cvt_f16_f32_e32 v1, v1
@@ -32540,8 +32551,8 @@ define <64 x half> @bitcast_v32i32_to_v64f16(<32 x i32> %a, i32 %b) {
; SI-NEXT: v_or_b32_e32 v1, v2, v1
; SI-NEXT: buffer_store_dword v1, v3, s[0:3], 0 offen
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:112 ; 4-byte Folded Reload
-; SI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:116 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:116 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:120 ; 4-byte Folded Reload
; SI-NEXT: v_add_i32_e32 v3, vcc, 0x54, v0
; SI-NEXT: s_waitcnt vmcnt(1)
; SI-NEXT: v_cvt_f16_f32_e32 v1, v1
@@ -32551,8 +32562,8 @@ define <64 x half> @bitcast_v32i32_to_v64f16(<32 x i32> %a, i32 %b) {
; SI-NEXT: v_or_b32_e32 v1, v2, v1
; SI-NEXT: buffer_store_dword v1, v3, s[0:3], 0 offen
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:100 ; 4-byte Folded Reload
-; SI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:108 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:104 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:112 ; 4-byte Folded Reload
; SI-NEXT: v_add_i32_e32 v3, vcc, 0x58, v0
; SI-NEXT: s_waitcnt vmcnt(1)
; SI-NEXT: v_cvt_f16_f32_e32 v1, v1
@@ -32562,8 +32573,8 @@ define <64 x half> @bitcast_v32i32_to_v64f16(<32 x i32> %a, i32 %b) {
; SI-NEXT: v_or_b32_e32 v1, v2, v1
; SI-NEXT: buffer_store_dword v1, v3, s[0:3], 0 offen
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:92 ; 4-byte Folded Reload
-; SI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:104 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:96 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:108 ; 4-byte Folded Reload
; SI-NEXT: v_add_i32_e32 v3, vcc, 0x5c, v0
; SI-NEXT: s_waitcnt vmcnt(1)
; SI-NEXT: v_cvt_f16_f32_e32 v1, v1
@@ -32573,8 +32584,8 @@ define <64 x half> @bitcast_v32i32_to_v64f16(<32 x i32> %a, i32 %b) {
; SI-NEXT: v_or_b32_e32 v1, v2, v1
; SI-NEXT: buffer_store_dword v1, v3, s[0:3], 0 offen
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:84 ; 4-byte Folded Reload
-; SI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:96 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:88 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:100 ; 4-byte Folded Reload
; SI-NEXT: v_add_i32_e32 v3, vcc, 0x60, v0
; SI-NEXT: s_waitcnt vmcnt(1)
; SI-NEXT: v_cvt_f16_f32_e32 v1, v1
@@ -32584,8 +32595,8 @@ define <64 x half> @bitcast_v32i32_to_v64f16(<32 x i32> %a, i32 %b) {
; SI-NEXT: v_or_b32_e32 v1, v2, v1
; SI-NEXT: buffer_store_dword v1, v3, s[0:3], 0 offen
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:76 ; 4-byte Folded Reload
-; SI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:88 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:80 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:92 ; 4-byte Folded Reload
; SI-NEXT: v_add_i32_e32 v3, vcc, 0x64, v0
; SI-NEXT: s_waitcnt vmcnt(1)
; SI-NEXT: v_cvt_f16_f32_e32 v1, v1
@@ -32594,46 +32605,48 @@ define <64 x half> @bitcast_v32i32_to_v64f16(<32 x i32> %a, i32 %b) {
; SI-NEXT: v_lshlrev_b32_e32 v1, 16, v1
; SI-NEXT: v_or_b32_e32 v1, v2, v1
; SI-NEXT: buffer_store_dword v1, v3, s[0:3], 0 offen
-; SI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:80 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:84 ; 4-byte Folded Reload
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_cvt_f16_f32_e32 v1, v46
+; SI-NEXT: v_cvt_f16_f32_e32 v1, v44
; SI-NEXT: v_add_i32_e32 v3, vcc, 0x68, v0
; SI-NEXT: v_lshlrev_b32_e32 v1, 16, v1
; SI-NEXT: s_waitcnt vmcnt(0)
; SI-NEXT: v_cvt_f16_f32_e32 v2, v2
; SI-NEXT: v_or_b32_e32 v1, v2, v1
; SI-NEXT: buffer_store_dword v1, v3, s[0:3], 0 offen
+; SI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:76 ; 4-byte Folded Reload
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_cvt_f16_f32_e32 v1, v50
-; SI-NEXT: v_cvt_f16_f32_e32 v2, v56
+; SI-NEXT: v_cvt_f16_f32_e32 v1, v52
; SI-NEXT: v_add_i32_e32 v3, vcc, 0x6c, v0
; SI-NEXT: v_lshlrev_b32_e32 v1, 16, v1
+; SI-NEXT: s_waitcnt vmcnt(0)
+; SI-NEXT: v_cvt_f16_f32_e32 v2, v2
; SI-NEXT: v_or_b32_e32 v1, v2, v1
; SI-NEXT: buffer_store_dword v1, v3, s[0:3], 0 offen
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_cvt_f16_f32_e32 v1, v54
-; SI-NEXT: v_cvt_f16_f32_e32 v2, v52
+; SI-NEXT: v_cvt_f16_f32_e32 v2, v55
; SI-NEXT: v_add_i32_e32 v3, vcc, 0x70, v0
; SI-NEXT: v_lshlrev_b32_e32 v1, 16, v1
; SI-NEXT: v_or_b32_e32 v1, v2, v1
; SI-NEXT: buffer_store_dword v1, v3, s[0:3], 0 offen
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_cvt_f16_f32_e32 v1, v40
-; SI-NEXT: v_cvt_f16_f32_e32 v2, v48
+; SI-NEXT: v_cvt_f16_f32_e32 v1, v53
+; SI-NEXT: v_cvt_f16_f32_e32 v2, v50
; SI-NEXT: v_add_i32_e32 v3, vcc, 0x74, v0
; SI-NEXT: v_lshlrev_b32_e32 v1, 16, v1
; SI-NEXT: v_or_b32_e32 v1, v2, v1
; SI-NEXT: buffer_store_dword v1, v3, s[0:3], 0 offen
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_cvt_f16_f32_e32 v1, v34
-; SI-NEXT: v_cvt_f16_f32_e32 v2, v36
+; SI-NEXT: v_cvt_f16_f32_e32 v1, v36
+; SI-NEXT: v_cvt_f16_f32_e32 v2, v48
; SI-NEXT: v_add_i32_e32 v3, vcc, 0x78, v0
; SI-NEXT: v_lshlrev_b32_e32 v1, 16, v1
; SI-NEXT: v_or_b32_e32 v1, v2, v1
; SI-NEXT: buffer_store_dword v1, v3, s[0:3], 0 offen
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_cvt_f16_f32_e32 v1, v32
-; SI-NEXT: v_cvt_f16_f32_e32 v2, v31
+; SI-NEXT: v_cvt_f16_f32_e32 v1, v34
+; SI-NEXT: v_cvt_f16_f32_e32 v2, v32
; SI-NEXT: v_add_i32_e32 v0, vcc, 0x7c, v0
; SI-NEXT: v_lshlrev_b32_e32 v1, 16, v1
; SI-NEXT: v_or_b32_e32 v1, v2, v1
@@ -37855,24 +37868,24 @@ define inreg <32 x i32> @bitcast_v64i16_to_v32i32_scalar(<64 x i16> inreg %a, i3
; SI-NEXT: buffer_load_dword v10, off, s[0:3], s32 offset:32
; SI-NEXT: buffer_load_dword v46, off, s[0:3], s32 offset:28
; SI-NEXT: buffer_load_dword v12, off, s[0:3], s32 offset:40
-; SI-NEXT: buffer_load_dword v45, off, s[0:3], s32 offset:36
-; SI-NEXT: buffer_load_dword v38, off, s[0:3], s32 offset:48
+; SI-NEXT: buffer_load_dword v59, off, s[0:3], s32 offset:36
+; SI-NEXT: buffer_load_dword v48, off, s[0:3], s32 offset:48
; SI-NEXT: buffer_load_dword v44, off, s[0:3], s32 offset:44
; SI-NEXT: buffer_load_dword v36, off, s[0:3], s32 offset:56
-; SI-NEXT: buffer_load_dword v59, off, s[0:3], s32 offset:52
+; SI-NEXT: buffer_load_dword v43, off, s[0:3], s32 offset:52
; SI-NEXT: buffer_load_dword v34, off, s[0:3], s32 offset:64
; SI-NEXT: buffer_load_dword v42, off, s[0:3], s32 offset:60
; SI-NEXT: buffer_load_dword v32, off, s[0:3], s32 offset:72
-; SI-NEXT: buffer_load_dword v58, off, s[0:3], s32 offset:68
-; SI-NEXT: s_waitcnt expcnt(2)
-; SI-NEXT: v_lshlrev_b32_e32 v61, 16, v1
+; SI-NEXT: buffer_load_dword v38, off, s[0:3], s32 offset:68
+; SI-NEXT: v_lshlrev_b32_e32 v58, 16, v1
; SI-NEXT: v_lshlrev_b32_e32 v55, 16, v3
-; SI-NEXT: v_lshlrev_b32_e32 v50, 16, v5
-; SI-NEXT: v_lshlrev_b32_e32 v43, 16, v7
-; SI-NEXT: v_lshlrev_b32_e32 v41, 16, v9
-; SI-NEXT: v_lshlrev_b32_e32 v40, 16, v11
+; SI-NEXT: s_waitcnt expcnt(2)
+; SI-NEXT: v_lshlrev_b32_e32 v61, 16, v5
+; SI-NEXT: v_lshlrev_b32_e32 v50, 16, v7
+; SI-NEXT: v_lshlrev_b32_e32 v45, 16, v9
+; SI-NEXT: v_lshlrev_b32_e32 v41, 16, v11
; SI-NEXT: v_lshlrev_b32_e32 v13, 16, v13
-; SI-NEXT: v_lshlrev_b32_e32 v48, 16, v15
+; SI-NEXT: v_lshlrev_b32_e32 v40, 16, v15
; SI-NEXT: v_lshlrev_b32_e32 v15, 16, v17
; SI-NEXT: v_lshlrev_b32_e32 v37, 16, v19
; SI-NEXT: v_lshlrev_b32_e32 v17, 16, v21
@@ -37893,23 +37906,23 @@ define inreg <32 x i32> @bitcast_v64i16_to_v32i32_scalar(<64 x i16> inreg %a, i3
; SI-NEXT: s_waitcnt vmcnt(9)
; SI-NEXT: v_lshlrev_b32_e32 v54, 16, v12
; SI-NEXT: s_waitcnt vmcnt(7)
-; SI-NEXT: v_lshlrev_b32_e32 v5, 16, v38
+; SI-NEXT: v_lshlrev_b32_e32 v5, 16, v48
; SI-NEXT: s_waitcnt vmcnt(5) expcnt(1)
; SI-NEXT: v_lshlrev_b32_e32 v62, 16, v36
; SI-NEXT: s_waitcnt vmcnt(3)
; SI-NEXT: v_lshlrev_b32_e32 v3, 16, v34
; SI-NEXT: s_waitcnt vmcnt(1)
-; SI-NEXT: v_lshlrev_b32_e32 v34, 16, v32
-; SI-NEXT: buffer_store_dword v34, off, s[0:3], s32 offset:144 ; 4-byte Folded Spill
+; SI-NEXT: v_lshlrev_b32_e32 v48, 16, v32
+; SI-NEXT: buffer_store_dword v48, off, s[0:3], s32 offset:144 ; 4-byte Folded Spill
; SI-NEXT: buffer_store_dword v3, off, s[0:3], s32 offset:148 ; 4-byte Folded Spill
; SI-NEXT: buffer_store_dword v62, off, s[0:3], s32 offset:152 ; 4-byte Folded Spill
; SI-NEXT: buffer_store_dword v5, off, s[0:3], s32 offset:156 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt vmcnt(4)
-; SI-NEXT: buffer_store_dword v58, off, s[0:3], s32 offset:160 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v38, off, s[0:3], s32 offset:160 ; 4-byte Folded Spill
; SI-NEXT: buffer_store_dword v54, off, s[0:3], s32 offset:164 ; 4-byte Folded Spill
-; SI-NEXT: buffer_store_dword v27, off, s[0:3], s32 offset:168 ; 4-byte Folded Spill
-; SI-NEXT: buffer_store_dword v59, off, s[0:3], s32 offset:172 ; 4-byte Folded Spill
-; SI-NEXT: buffer_store_dword v45, off, s[0:3], s32 offset:176 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v43, off, s[0:3], s32 offset:168 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v27, off, s[0:3], s32 offset:172 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v59, off, s[0:3], s32 offset:176 ; 4-byte Folded Spill
; SI-NEXT: buffer_store_dword v25, off, s[0:3], s32 offset:180 ; 4-byte Folded Spill
; SI-NEXT: buffer_store_dword v47, off, s[0:3], s32 offset:184 ; 4-byte Folded Spill
; SI-NEXT: buffer_store_dword v60, off, s[0:3], s32 offset:188 ; 4-byte Folded Spill
@@ -37922,45 +37935,46 @@ define inreg <32 x i32> @bitcast_v64i16_to_v32i32_scalar(<64 x i16> inreg %a, i3
; SI-NEXT: buffer_store_dword v19, off, s[0:3], s32 offset:216 ; 4-byte Folded Spill
; SI-NEXT: buffer_store_dword v35, off, s[0:3], s32 offset:220 ; 4-byte Folded Spill
; SI-NEXT: buffer_store_dword v22, off, s[0:3], s32 offset:224 ; 4-byte Folded Spill
-; SI-NEXT: buffer_store_dword v17, off, s[0:3], s32 offset:228 ; 4-byte Folded Spill
-; SI-NEXT: buffer_store_dword v0, off, s[0:3], s32 offset:232 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v0, off, s[0:3], s32 offset:228 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v17, off, s[0:3], s32 offset:232 ; 4-byte Folded Spill
; SI-NEXT: buffer_store_dword v37, off, s[0:3], s32 offset:236 ; 4-byte Folded Spill
; SI-NEXT: buffer_store_dword v18, off, s[0:3], s32 offset:240 ; 4-byte Folded Spill
; SI-NEXT: s_cbranch_scc0 .LBB27_4
; SI-NEXT: ; %bb.1: ; %cmp.false
-; SI-NEXT: s_waitcnt expcnt(2)
+; SI-NEXT: s_waitcnt expcnt(3)
; SI-NEXT: v_and_b32_e32 v0, 0xffff, v0
-; SI-NEXT: v_or_b32_e32 v7, v0, v61
+; SI-NEXT: v_or_b32_e32 v7, v0, v58
; SI-NEXT: v_and_b32_e32 v0, 0xffff, v4
-; SI-NEXT: v_or_b32_e32 v9, v0, v50
+; SI-NEXT: v_or_b32_e32 v9, v0, v61
; SI-NEXT: v_and_b32_e32 v0, 0xffff, v6
-; SI-NEXT: v_or_b32_e32 v10, v0, v43
+; SI-NEXT: v_or_b32_e32 v10, v0, v50
; SI-NEXT: v_and_b32_e32 v0, 0xffff, v57
-; SI-NEXT: v_or_b32_e32 v11, v0, v41
+; SI-NEXT: v_or_b32_e32 v11, v0, v45
; SI-NEXT: v_and_b32_e32 v0, 0xffff, v56
-; SI-NEXT: v_or_b32_e32 v12, v0, v40
+; SI-NEXT: v_or_b32_e32 v12, v0, v41
; SI-NEXT: v_and_b32_e32 v0, 0xffff, v49
-; SI-NEXT: v_mov_b32_e32 v52, v57
-; SI-NEXT: v_mov_b32_e32 v57, v40
-; SI-NEXT: v_mov_b32_e32 v40, v49
-; SI-NEXT: v_mov_b32_e32 v49, v13
+; SI-NEXT: v_mov_b32_e32 v36, v41
+; SI-NEXT: v_mov_b32_e32 v41, v13
; SI-NEXT: v_or_b32_e32 v13, v0, v13
; SI-NEXT: v_and_b32_e32 v0, 0xffff, v14
-; SI-NEXT: v_mov_b32_e32 v36, v41
-; SI-NEXT: v_mov_b32_e32 v41, v14
-; SI-NEXT: v_or_b32_e32 v14, v0, v48
-; SI-NEXT: v_and_b32_e32 v0, 0xffff, v16
; SI-NEXT: v_mov_b32_e32 v51, v50
-; SI-NEXT: v_mov_b32_e32 v50, v43
-; SI-NEXT: v_mov_b32_e32 v43, v48
-; SI-NEXT: v_mov_b32_e32 v48, v15
+; SI-NEXT: v_mov_b32_e32 v50, v45
+; SI-NEXT: v_mov_b32_e32 v45, v14
+; SI-NEXT: v_or_b32_e32 v14, v0, v40
+; SI-NEXT: v_and_b32_e32 v0, 0xffff, v16
+; SI-NEXT: v_mov_b32_e32 v52, v57
+; SI-NEXT: v_mov_b32_e32 v57, v49
+; SI-NEXT: v_mov_b32_e32 v49, v40
+; SI-NEXT: v_mov_b32_e32 v40, v15
; SI-NEXT: v_or_b32_e32 v15, v0, v15
; SI-NEXT: v_and_b32_e32 v0, 0xffff, v18
-; SI-NEXT: v_mov_b32_e32 v38, v61
+; SI-NEXT: v_mov_b32_e32 v34, v58
+; SI-NEXT: v_mov_b32_e32 v58, v61
; SI-NEXT: v_mov_b32_e32 v61, v56
; SI-NEXT: v_mov_b32_e32 v56, v16
; SI-NEXT: v_or_b32_e32 v16, v0, v37
; SI-NEXT: v_and_b32_e32 v0, 0xffff, v20
+; SI-NEXT: s_waitcnt expcnt(2)
; SI-NEXT: v_or_b32_e32 v17, v0, v17
; SI-NEXT: v_and_b32_e32 v0, 0xffff, v22
; SI-NEXT: s_waitcnt expcnt(0)
@@ -37994,7 +38008,7 @@ define inreg <32 x i32> @bitcast_v64i16_to_v32i32_scalar(<64 x i16> inreg %a, i3
; SI-NEXT: s_and_b32 s6, s20, 0xffff
; SI-NEXT: s_lshl_b32 s7, s21, 16
; SI-NEXT: v_or_b32_e32 v26, v0, v26
-; SI-NEXT: v_and_b32_e32 v0, 0xffff, v45
+; SI-NEXT: v_and_b32_e32 v0, 0xffff, v59
; SI-NEXT: s_or_b32 s6, s6, s7
; SI-NEXT: s_and_b32 s7, s22, 0xffff
; SI-NEXT: s_lshl_b32 s8, s23, 16
@@ -38005,7 +38019,7 @@ define inreg <32 x i32> @bitcast_v64i16_to_v32i32_scalar(<64 x i16> inreg %a, i3
; SI-NEXT: s_lshl_b32 s9, s25, 16
; SI-NEXT: v_mov_b32_e32 v33, v28
; SI-NEXT: v_or_b32_e32 v28, v0, v5
-; SI-NEXT: v_and_b32_e32 v0, 0xffff, v59
+; SI-NEXT: v_and_b32_e32 v0, 0xffff, v43
; SI-NEXT: s_or_b32 s8, s8, s9
; SI-NEXT: s_and_b32 s9, s26, 0xffff
; SI-NEXT: s_lshl_b32 s10, s27, 16
@@ -38017,7 +38031,7 @@ define inreg <32 x i32> @bitcast_v64i16_to_v32i32_scalar(<64 x i16> inreg %a, i3
; SI-NEXT: s_lshl_b32 s11, s29, 16
; SI-NEXT: v_and_b32_e32 v1, 0xffff, v2
; SI-NEXT: v_or_b32_e32 v30, v0, v3
-; SI-NEXT: v_and_b32_e32 v0, 0xffff, v58
+; SI-NEXT: v_and_b32_e32 v0, 0xffff, v38
; SI-NEXT: s_or_b32 s10, s10, s11
; SI-NEXT: v_mov_b32_e32 v63, v2
; SI-NEXT: v_mov_b32_e32 v32, v55
@@ -38025,9 +38039,9 @@ define inreg <32 x i32> @bitcast_v64i16_to_v32i32_scalar(<64 x i16> inreg %a, i3
; SI-NEXT: v_mov_b32_e32 v55, v4
; SI-NEXT: v_mov_b32_e32 v53, v6
; SI-NEXT: v_mov_b32_e32 v47, v46
-; SI-NEXT: v_mov_b32_e32 v45, v44
-; SI-NEXT: v_mov_b32_e32 v59, v42
-; SI-NEXT: v_or_b32_e32 v31, v0, v34
+; SI-NEXT: v_mov_b32_e32 v59, v44
+; SI-NEXT: v_mov_b32_e32 v43, v42
+; SI-NEXT: v_or_b32_e32 v31, v0, v48
; SI-NEXT: v_mov_b32_e32 v0, s4
; SI-NEXT: v_mov_b32_e32 v1, s5
; SI-NEXT: v_mov_b32_e32 v2, s6
@@ -38037,12 +38051,13 @@ define inreg <32 x i32> @bitcast_v64i16_to_v32i32_scalar(<64 x i16> inreg %a, i3
; SI-NEXT: v_mov_b32_e32 v6, s10
; SI-NEXT: s_cbranch_execnz .LBB27_3
; SI-NEXT: .LBB27_2: ; %cmp.true
-; SI-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:232 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:228 ; 4-byte Folded Reload
; SI-NEXT: v_add_i32_e32 v1, vcc, 3, v63
; SI-NEXT: v_and_b32_e32 v1, 0xffff, v1
; SI-NEXT: v_or_b32_e32 v1, v32, v1
; SI-NEXT: v_add_i32_e32 v8, vcc, 0x30000, v1
; SI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:236 ; 4-byte Folded Reload
+; SI-NEXT: v_mov_b32_e32 v38, v43
; SI-NEXT: s_add_i32 s16, s16, 3
; SI-NEXT: s_and_b32 s4, s16, 0xffff
; SI-NEXT: s_lshl_b32 s5, s17, 16
@@ -38086,42 +38101,42 @@ define inreg <32 x i32> @bitcast_v64i16_to_v32i32_scalar(<64 x i16> inreg %a, i3
; SI-NEXT: s_waitcnt vmcnt(1)
; SI-NEXT: v_add_i32_e32 v0, vcc, 3, v0
; SI-NEXT: v_and_b32_e32 v0, 0xffff, v0
-; SI-NEXT: v_or_b32_e32 v0, v38, v0
+; SI-NEXT: v_or_b32_e32 v0, v34, v0
; SI-NEXT: v_add_i32_e32 v7, vcc, 0x30000, v0
; SI-NEXT: v_add_i32_e32 v0, vcc, 3, v55
; SI-NEXT: v_and_b32_e32 v0, 0xffff, v0
-; SI-NEXT: v_or_b32_e32 v0, v51, v0
+; SI-NEXT: v_or_b32_e32 v0, v58, v0
; SI-NEXT: v_add_i32_e32 v9, vcc, 0x30000, v0
; SI-NEXT: v_add_i32_e32 v0, vcc, 3, v53
; SI-NEXT: v_and_b32_e32 v0, 0xffff, v0
-; SI-NEXT: v_or_b32_e32 v0, v50, v0
+; SI-NEXT: v_or_b32_e32 v0, v51, v0
; SI-NEXT: v_add_i32_e32 v10, vcc, 0x30000, v0
; SI-NEXT: v_add_i32_e32 v0, vcc, 3, v52
; SI-NEXT: v_and_b32_e32 v0, 0xffff, v0
-; SI-NEXT: v_or_b32_e32 v0, v36, v0
+; SI-NEXT: v_or_b32_e32 v0, v50, v0
; SI-NEXT: v_add_i32_e32 v11, vcc, 0x30000, v0
; SI-NEXT: v_add_i32_e32 v0, vcc, 3, v61
; SI-NEXT: v_and_b32_e32 v0, 0xffff, v0
-; SI-NEXT: v_or_b32_e32 v0, v57, v0
+; SI-NEXT: v_or_b32_e32 v0, v36, v0
; SI-NEXT: v_add_i32_e32 v12, vcc, 0x30000, v0
-; SI-NEXT: v_add_i32_e32 v0, vcc, 3, v40
+; SI-NEXT: v_add_i32_e32 v0, vcc, 3, v57
; SI-NEXT: v_and_b32_e32 v0, 0xffff, v0
-; SI-NEXT: v_or_b32_e32 v0, v49, v0
+; SI-NEXT: v_or_b32_e32 v0, v41, v0
; SI-NEXT: v_add_i32_e32 v13, vcc, 0x30000, v0
-; SI-NEXT: v_add_i32_e32 v0, vcc, 3, v41
+; SI-NEXT: v_add_i32_e32 v0, vcc, 3, v45
; SI-NEXT: v_and_b32_e32 v0, 0xffff, v0
-; SI-NEXT: v_or_b32_e32 v0, v43, v0
+; SI-NEXT: v_or_b32_e32 v0, v49, v0
; SI-NEXT: v_add_i32_e32 v14, vcc, 0x30000, v0
; SI-NEXT: v_add_i32_e32 v0, vcc, 3, v56
; SI-NEXT: v_and_b32_e32 v0, 0xffff, v0
-; SI-NEXT: v_or_b32_e32 v0, v48, v0
+; SI-NEXT: v_or_b32_e32 v0, v40, v0
; SI-NEXT: v_add_i32_e32 v15, vcc, 0x30000, v0
; SI-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:240 ; 4-byte Folded Reload
; SI-NEXT: s_waitcnt vmcnt(0)
; SI-NEXT: v_add_i32_e32 v0, vcc, 3, v0
; SI-NEXT: v_and_b32_e32 v0, 0xffff, v0
; SI-NEXT: v_or_b32_e32 v0, v1, v0
-; SI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:228 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:232 ; 4-byte Folded Reload
; SI-NEXT: v_add_i32_e32 v16, vcc, 0x30000, v0
; SI-NEXT: v_add_i32_e32 v0, vcc, 3, v37
; SI-NEXT: v_and_b32_e32 v0, 0xffff, v0
@@ -38183,7 +38198,7 @@ define inreg <32 x i32> @bitcast_v64i16_to_v32i32_scalar(<64 x i16> inreg %a, i3
; SI-NEXT: v_add_i32_e32 v0, vcc, 3, v0
; SI-NEXT: v_and_b32_e32 v0, 0xffff, v0
; SI-NEXT: v_or_b32_e32 v0, v1, v0
-; SI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:168 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:172 ; 4-byte Folded Reload
; SI-NEXT: v_add_i32_e32 v25, vcc, 0x30000, v0
; SI-NEXT: v_add_i32_e32 v0, vcc, 3, v47
; SI-NEXT: v_and_b32_e32 v0, 0xffff, v0
@@ -38199,12 +38214,12 @@ define inreg <32 x i32> @bitcast_v64i16_to_v32i32_scalar(<64 x i16> inreg %a, i3
; SI-NEXT: v_or_b32_e32 v0, v1, v0
; SI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:156 ; 4-byte Folded Reload
; SI-NEXT: v_add_i32_e32 v27, vcc, 0x30000, v0
-; SI-NEXT: v_add_i32_e32 v0, vcc, 3, v45
+; SI-NEXT: v_add_i32_e32 v0, vcc, 3, v59
; SI-NEXT: v_and_b32_e32 v0, 0xffff, v0
; SI-NEXT: s_waitcnt vmcnt(0)
; SI-NEXT: v_or_b32_e32 v0, v1, v0
; SI-NEXT: v_add_i32_e32 v28, vcc, 0x30000, v0
-; SI-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:172 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:168 ; 4-byte Folded Reload
; SI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:152 ; 4-byte Folded Reload
; SI-NEXT: s_waitcnt vmcnt(1)
; SI-NEXT: v_add_i32_e32 v0, vcc, 3, v0
@@ -38213,7 +38228,7 @@ define inreg <32 x i32> @bitcast_v64i16_to_v32i32_scalar(<64 x i16> inreg %a, i3
; SI-NEXT: v_or_b32_e32 v0, v1, v0
; SI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:148 ; 4-byte Folded Reload
; SI-NEXT: v_add_i32_e32 v29, vcc, 0x30000, v0
-; SI-NEXT: v_add_i32_e32 v0, vcc, 3, v59
+; SI-NEXT: v_add_i32_e32 v0, vcc, 3, v38
; SI-NEXT: v_and_b32_e32 v0, 0xffff, v0
; SI-NEXT: s_waitcnt vmcnt(0)
; SI-NEXT: v_or_b32_e32 v0, v1, v0
@@ -38248,26 +38263,26 @@ define inreg <32 x i32> @bitcast_v64i16_to_v32i32_scalar(<64 x i16> inreg %a, i3
; SI-NEXT: s_waitcnt vmcnt(0) expcnt(0)
; SI-NEXT: s_setpc_b64 s[30:31]
; SI-NEXT: .LBB27_4:
-; SI-NEXT: v_mov_b32_e32 v38, v61
+; SI-NEXT: v_mov_b32_e32 v34, v58
; SI-NEXT: v_mov_b32_e32 v32, v55
+; SI-NEXT: v_mov_b32_e32 v58, v61
; SI-NEXT: v_mov_b32_e32 v63, v2
; SI-NEXT: v_mov_b32_e32 v55, v4
; SI-NEXT: v_mov_b32_e32 v53, v6
; SI-NEXT: v_mov_b32_e32 v52, v57
; SI-NEXT: v_mov_b32_e32 v51, v50
; SI-NEXT: v_mov_b32_e32 v61, v56
-; SI-NEXT: v_mov_b32_e32 v50, v43
+; SI-NEXT: v_mov_b32_e32 v50, v45
; SI-NEXT: v_mov_b32_e32 v36, v41
-; SI-NEXT: v_mov_b32_e32 v57, v40
-; SI-NEXT: v_mov_b32_e32 v40, v49
-; SI-NEXT: v_mov_b32_e32 v49, v13
-; SI-NEXT: v_mov_b32_e32 v43, v48
-; SI-NEXT: v_mov_b32_e32 v48, v15
-; SI-NEXT: v_mov_b32_e32 v41, v14
+; SI-NEXT: v_mov_b32_e32 v41, v13
+; SI-NEXT: v_mov_b32_e32 v57, v49
+; SI-NEXT: v_mov_b32_e32 v49, v40
+; SI-NEXT: v_mov_b32_e32 v40, v15
+; SI-NEXT: v_mov_b32_e32 v45, v14
; SI-NEXT: v_mov_b32_e32 v56, v16
; SI-NEXT: v_mov_b32_e32 v47, v46
-; SI-NEXT: v_mov_b32_e32 v45, v44
-; SI-NEXT: v_mov_b32_e32 v59, v42
+; SI-NEXT: v_mov_b32_e32 v59, v44
+; SI-NEXT: v_mov_b32_e32 v43, v42
; SI-NEXT: s_waitcnt expcnt(1)
; SI-NEXT: v_mov_b32_e32 v37, v20
; SI-NEXT: v_mov_b32_e32 v39, v23
@@ -58422,7 +58437,7 @@ define inreg <32 x float> @bitcast_v128i8_to_v32f32_scalar(<128 x i8> inreg %a,
; VI-NEXT: buffer_store_dword v10, off, s[0:3], s32 offset:476 ; 4-byte Folded Spill
; VI-NEXT: buffer_store_dword v8, off, s[0:3], s32 offset:484 ; 4-byte Folded Spill
; VI-NEXT: buffer_store_dword v6, off, s[0:3], s32 offset:488 ; 4-byte Folded Spill
-; VI-NEXT: buffer_store_dword v0, off, s[0:3], s32 offset:556 ; 4-byte Folded Spill
+; VI-NEXT: buffer_store_dword v0, off, s[0:3], s32 offset:560 ; 4-byte Folded Spill
; VI-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:332
; VI-NEXT: buffer_load_ushort v44, off, s[0:3], s32
; VI-NEXT: buffer_load_ushort v43, off, s[0:3], s32 offset:8
@@ -58448,7 +58463,7 @@ define inreg <32 x float> @bitcast_v128i8_to_v32f32_scalar(<128 x i8> inreg %a,
; VI-NEXT: buffer_load_ushort v37, off, s[0:3], s32 offset:168
; VI-NEXT: buffer_load_ushort v38, off, s[0:3], s32 offset:176
; VI-NEXT: v_lshlrev_b32_e32 v1, 8, v1
-; VI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:560 ; 4-byte Folded Spill
+; VI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:564 ; 4-byte Folded Spill
; VI-NEXT: v_lshlrev_b32_e32 v1, 8, v15
; VI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:548 ; 4-byte Folded Spill
; VI-NEXT: v_lshlrev_b32_e32 v1, 8, v17
@@ -58457,11 +58472,11 @@ define inreg <32 x float> @bitcast_v128i8_to_v32f32_scalar(<128 x i8> inreg %a,
; VI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:540 ; 4-byte Folded Spill
; VI-NEXT: v_lshlrev_b32_e32 v1, 8, v21
; VI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:536 ; 4-byte Folded Spill
-; VI-NEXT: v_lshlrev_b32_e32 v8, 8, v3
-; VI-NEXT: v_lshlrev_b32_e32 v59, 8, v5
-; VI-NEXT: v_lshlrev_b32_e32 v14, 8, v7
-; VI-NEXT: v_lshlrev_b32_e32 v10, 8, v9
-; VI-NEXT: v_lshlrev_b32_e32 v16, 8, v11
+; VI-NEXT: v_lshlrev_b32_e32 v56, 8, v3
+; VI-NEXT: v_lshlrev_b32_e32 v8, 8, v5
+; VI-NEXT: v_lshlrev_b32_e32 v10, 8, v7
+; VI-NEXT: v_lshlrev_b32_e32 v14, 8, v9
+; VI-NEXT: v_lshlrev_b32_e32 v12, 8, v11
; VI-NEXT: v_lshlrev_b32_e32 v6, 8, v13
; VI-NEXT: s_waitcnt vmcnt(14)
; VI-NEXT: v_cmp_ne_u32_e32 vcc, 0, v0
@@ -58482,7 +58497,7 @@ define inreg <32 x float> @bitcast_v128i8_to_v32f32_scalar(<128 x i8> inreg %a,
; VI-NEXT: v_lshlrev_b32_e32 v0, 8, v41
; VI-NEXT: buffer_store_dword v0, off, s[0:3], s32 offset:504 ; 4-byte Folded Spill
; VI-NEXT: v_lshlrev_b32_e32 v0, 8, v40
-; VI-NEXT: buffer_store_dword v0, off, s[0:3], s32 offset:580 ; 4-byte Folded Spill
+; VI-NEXT: buffer_store_dword v0, off, s[0:3], s32 offset:588 ; 4-byte Folded Spill
; VI-NEXT: v_lshlrev_b32_e32 v0, 8, v55
; VI-NEXT: buffer_store_dword v0, off, s[0:3], s32 offset:500 ; 4-byte Folded Spill
; VI-NEXT: v_lshlrev_b32_e32 v0, 8, v54
@@ -58490,22 +58505,22 @@ define inreg <32 x float> @bitcast_v128i8_to_v32f32_scalar(<128 x i8> inreg %a,
; VI-NEXT: v_lshlrev_b32_e32 v0, 8, v53
; VI-NEXT: buffer_store_dword v0, off, s[0:3], s32 offset:496 ; 4-byte Folded Spill
; VI-NEXT: v_lshlrev_b32_e32 v0, 8, v52
-; VI-NEXT: buffer_store_dword v0, off, s[0:3], s32 offset:608 ; 4-byte Folded Spill
+; VI-NEXT: buffer_store_dword v0, off, s[0:3], s32 offset:616 ; 4-byte Folded Spill
; VI-NEXT: v_lshlrev_b32_e32 v0, 8, v51
; VI-NEXT: buffer_store_dword v0, off, s[0:3], s32 offset:492 ; 4-byte Folded Spill
; VI-NEXT: v_lshlrev_b32_e32 v0, 8, v50
-; VI-NEXT: buffer_store_dword v0, off, s[0:3], s32 offset:600 ; 4-byte Folded Spill
+; VI-NEXT: buffer_store_dword v0, off, s[0:3], s32 offset:612 ; 4-byte Folded Spill
; VI-NEXT: v_lshlrev_b32_e32 v0, 8, v49
-; VI-NEXT: buffer_store_dword v0, off, s[0:3], s32 offset:604 ; 4-byte Folded Spill
+; VI-NEXT: buffer_store_dword v0, off, s[0:3], s32 offset:608 ; 4-byte Folded Spill
; VI-NEXT: v_lshlrev_b32_e32 v0, 8, v48
-; VI-NEXT: buffer_store_dword v0, off, s[0:3], s32 offset:596 ; 4-byte Folded Spill
+; VI-NEXT: buffer_store_dword v0, off, s[0:3], s32 offset:604 ; 4-byte Folded Spill
; VI-NEXT: v_lshlrev_b32_e32 v0, 8, v39
; VI-NEXT: buffer_store_dword v0, off, s[0:3], s32 offset:480 ; 4-byte Folded Spill
; VI-NEXT: s_waitcnt vmcnt(14)
; VI-NEXT: v_lshlrev_b32_e32 v0, 8, v22
-; VI-NEXT: buffer_store_dword v0, off, s[0:3], s32 offset:572 ; 4-byte Folded Spill
+; VI-NEXT: buffer_store_dword v0, off, s[0:3], s32 offset:596 ; 4-byte Folded Spill
; VI-NEXT: v_lshlrev_b32_e32 v0, 8, v31
-; VI-NEXT: buffer_store_dword v0, off, s[0:3], s32 offset:588 ; 4-byte Folded Spill
+; VI-NEXT: buffer_store_dword v0, off, s[0:3], s32 offset:600 ; 4-byte Folded Spill
; VI-NEXT: v_lshlrev_b32_e32 v0, 8, v32
; VI-NEXT: buffer_store_dword v0, off, s[0:3], s32 offset:592 ; 4-byte Folded Spill
; VI-NEXT: v_lshlrev_b32_e32 v0, 8, v33
@@ -58528,23 +58543,24 @@ define inreg <32 x float> @bitcast_v128i8_to_v32f32_scalar(<128 x i8> inreg %a,
; VI-NEXT: v_lshlrev_b32_e32 v31, 8, v38
; VI-NEXT: s_and_b64 s[4:5], vcc, exec
; VI-NEXT: s_waitcnt vmcnt(7)
-; VI-NEXT: v_lshlrev_b32_e32 v26, 8, v0
+; VI-NEXT: v_lshlrev_b32_e32 v22, 8, v0
; VI-NEXT: s_waitcnt vmcnt(6)
; VI-NEXT: v_lshlrev_b32_e32 v32, 8, v1
; VI-NEXT: s_waitcnt vmcnt(5)
-; VI-NEXT: v_lshlrev_b32_e32 v54, 8, v13
+; VI-NEXT: v_lshlrev_b32_e32 v23, 8, v13
+; VI-NEXT: s_waitcnt vmcnt(4)
+; VI-NEXT: v_lshlrev_b32_e32 v0, 8, v3
+; VI-NEXT: buffer_store_dword v0, off, s[0:3], s32 offset:580 ; 4-byte Folded Spill
; VI-NEXT: s_waitcnt vmcnt(4)
-; VI-NEXT: v_lshlrev_b32_e32 v49, 8, v3
-; VI-NEXT: s_waitcnt vmcnt(3)
; VI-NEXT: v_lshlrev_b32_e32 v0, 8, v11
-; VI-NEXT: buffer_store_dword v0, off, s[0:3], s32 offset:576 ; 4-byte Folded Spill
-; VI-NEXT: s_waitcnt vmcnt(3)
+; VI-NEXT: buffer_store_dword v0, off, s[0:3], s32 offset:572 ; 4-byte Folded Spill
+; VI-NEXT: s_waitcnt vmcnt(4)
; VI-NEXT: v_lshlrev_b32_e32 v0, 8, v5
-; VI-NEXT: buffer_store_dword v0, off, s[0:3], s32 offset:612 ; 4-byte Folded Spill
-; VI-NEXT: s_waitcnt vmcnt(3)
+; VI-NEXT: buffer_store_dword v0, off, s[0:3], s32 offset:576 ; 4-byte Folded Spill
+; VI-NEXT: s_waitcnt vmcnt(4)
; VI-NEXT: v_lshlrev_b32_e32 v0, 8, v9
; VI-NEXT: buffer_store_dword v0, off, s[0:3], s32 offset:416 ; 4-byte Folded Spill
-; VI-NEXT: s_waitcnt vmcnt(3)
+; VI-NEXT: s_waitcnt vmcnt(4)
; VI-NEXT: v_lshlrev_b32_e32 v0, 8, v7
; VI-NEXT: buffer_store_dword v0, off, s[0:3], s32 offset:412 ; 4-byte Folded Spill
; VI-NEXT: buffer_load_ushort v0, off, s[0:3], s32 offset:248
@@ -58557,37 +58573,37 @@ define inreg <32 x float> @bitcast_v128i8_to_v32f32_scalar(<128 x i8> inreg %a,
; VI-NEXT: buffer_load_ushort v7, off, s[0:3], s32 offset:304
; VI-NEXT: s_waitcnt vmcnt(7)
; VI-NEXT: v_lshlrev_b32_e32 v0, 8, v0
-; VI-NEXT: buffer_store_dword v0, off, s[0:3], s32 offset:564 ; 4-byte Folded Spill
+; VI-NEXT: buffer_store_dword v0, off, s[0:3], s32 offset:568 ; 4-byte Folded Spill
; VI-NEXT: s_waitcnt vmcnt(7)
; VI-NEXT: v_lshlrev_b32_e32 v48, 8, v1
; VI-NEXT: s_waitcnt vmcnt(6)
; VI-NEXT: v_lshlrev_b32_e32 v27, 8, v13
; VI-NEXT: s_waitcnt vmcnt(4)
; VI-NEXT: v_lshlrev_b32_e32 v0, 8, v11
-; VI-NEXT: buffer_store_dword v0, off, s[0:3], s32 offset:568 ; 4-byte Folded Spill
+; VI-NEXT: buffer_store_dword v0, off, s[0:3], s32 offset:556 ; 4-byte Folded Spill
; VI-NEXT: s_waitcnt vmcnt(4)
; VI-NEXT: v_lshlrev_b32_e32 v0, 8, v5
; VI-NEXT: buffer_store_dword v0, off, s[0:3], s32 offset:404 ; 4-byte Folded Spill
-; VI-NEXT: v_lshlrev_b32_e32 v29, 8, v3
+; VI-NEXT: v_lshlrev_b32_e32 v28, 8, v3
; VI-NEXT: buffer_load_ushort v0, off, s[0:3], s32 offset:312
; VI-NEXT: buffer_load_ushort v1, off, s[0:3], s32 offset:320
; VI-NEXT: buffer_load_ushort v3, off, s[0:3], s32 offset:328
; VI-NEXT: buffer_load_ushort v11, off, s[0:3], s32 offset:4
-; VI-NEXT: buffer_load_ushort v12, off, s[0:3], s32 offset:12
+; VI-NEXT: buffer_load_ushort v16, off, s[0:3], s32 offset:12
; VI-NEXT: buffer_load_ushort v50, off, s[0:3], s32 offset:20
-; VI-NEXT: buffer_load_ushort v56, off, s[0:3], s32 offset:28
+; VI-NEXT: buffer_load_ushort v59, off, s[0:3], s32 offset:28
; VI-NEXT: buffer_load_ushort v13, off, s[0:3], s32 offset:36
; VI-NEXT: s_waitcnt vmcnt(11)
; VI-NEXT: v_lshlrev_b32_e32 v5, 8, v7
-; VI-NEXT: v_lshlrev_b32_e32 v9, 8, v9
+; VI-NEXT: v_lshlrev_b32_e32 v29, 8, v9
; VI-NEXT: s_waitcnt vmcnt(7)
; VI-NEXT: v_lshlrev_b32_e32 v7, 8, v0
; VI-NEXT: s_waitcnt vmcnt(6)
; VI-NEXT: v_lshlrev_b32_e32 v0, 8, v1
; VI-NEXT: buffer_store_dword v0, off, s[0:3], s32 offset:552 ; 4-byte Folded Spill
; VI-NEXT: buffer_load_ushort v18, off, s[0:3], s32 offset:44
-; VI-NEXT: buffer_load_ushort v15, off, s[0:3], s32 offset:52
-; VI-NEXT: buffer_load_ushort v17, off, s[0:3], s32 offset:60
+; VI-NEXT: buffer_load_ushort v17, off, s[0:3], s32 offset:52
+; VI-NEXT: buffer_load_ushort v15, off, s[0:3], s32 offset:60
; VI-NEXT: buffer_load_ushort v39, off, s[0:3], s32 offset:68
; VI-NEXT: buffer_load_ushort v38, off, s[0:3], s32 offset:76
; VI-NEXT: buffer_load_ushort v37, off, s[0:3], s32 offset:84
@@ -58595,23 +58611,23 @@ define inreg <32 x float> @bitcast_v128i8_to_v32f32_scalar(<128 x i8> inreg %a,
; VI-NEXT: buffer_load_ushort v36, off, s[0:3], s32 offset:100
; VI-NEXT: buffer_load_ushort v35, off, s[0:3], s32 offset:108
; VI-NEXT: buffer_load_ushort v33, off, s[0:3], s32 offset:116
-; VI-NEXT: buffer_load_ushort v25, off, s[0:3], s32 offset:124
+; VI-NEXT: buffer_load_ushort v54, off, s[0:3], s32 offset:124
; VI-NEXT: buffer_load_ushort v19, off, s[0:3], s32 offset:132
-; VI-NEXT: buffer_load_ushort v51, off, s[0:3], s32 offset:140
+; VI-NEXT: buffer_load_ushort v49, off, s[0:3], s32 offset:140
; VI-NEXT: buffer_load_ushort v20, off, s[0:3], s32 offset:148
-; VI-NEXT: buffer_load_ushort v28, off, s[0:3], s32 offset:156
+; VI-NEXT: buffer_load_ushort v51, off, s[0:3], s32 offset:156
; VI-NEXT: buffer_load_ushort v21, off, s[0:3], s32 offset:164
; VI-NEXT: buffer_load_ushort v34, off, s[0:3], s32 offset:172
-; VI-NEXT: buffer_load_ushort v22, off, s[0:3], s32 offset:180
-; VI-NEXT: buffer_load_ushort v23, off, s[0:3], s32 offset:188
+; VI-NEXT: buffer_load_ushort v25, off, s[0:3], s32 offset:180
+; VI-NEXT: buffer_load_ushort v30, off, s[0:3], s32 offset:188
; VI-NEXT: buffer_load_ushort v43, off, s[0:3], s32 offset:196
-; VI-NEXT: buffer_load_ushort v30, off, s[0:3], s32 offset:204
-; VI-NEXT: buffer_load_ushort v24, off, s[0:3], s32 offset:212
+; VI-NEXT: buffer_load_ushort v24, off, s[0:3], s32 offset:204
+; VI-NEXT: buffer_load_ushort v26, off, s[0:3], s32 offset:212
; VI-NEXT: buffer_load_ushort v46, off, s[0:3], s32 offset:220
; VI-NEXT: buffer_load_ushort v61, off, s[0:3], s32 offset:228
; VI-NEXT: buffer_load_ushort v47, off, s[0:3], s32 offset:236
-; VI-NEXT: buffer_load_ushort v45, off, s[0:3], s32 offset:244
-; VI-NEXT: buffer_load_ushort v58, off, s[0:3], s32 offset:252
+; VI-NEXT: buffer_load_ushort v58, off, s[0:3], s32 offset:244
+; VI-NEXT: buffer_load_ushort v45, off, s[0:3], s32 offset:252
; VI-NEXT: buffer_load_ushort v44, off, s[0:3], s32 offset:260
; VI-NEXT: buffer_load_ushort v63, off, s[0:3], s32 offset:268
; VI-NEXT: buffer_load_ushort v42, off, s[0:3], s32 offset:276
@@ -58624,57 +58640,57 @@ define inreg <32 x float> @bitcast_v128i8_to_v32f32_scalar(<128 x i8> inreg %a,
; VI-NEXT: s_waitcnt vmcnt(14)
; VI-NEXT: v_lshlrev_b32_e32 v0, 8, v3
; VI-NEXT: buffer_store_dword v5, off, s[0:3], s32 offset:400 ; 4-byte Folded Spill
-; VI-NEXT: buffer_store_dword v9, off, s[0:3], s32 offset:616 ; 4-byte Folded Spill
+; VI-NEXT: buffer_store_dword v29, off, s[0:3], s32 offset:620 ; 4-byte Folded Spill
; VI-NEXT: buffer_store_dword v0, off, s[0:3], s32 offset:408 ; 4-byte Folded Spill
-; VI-NEXT: buffer_store_dword v27, off, s[0:3], s32 offset:620 ; 4-byte Folded Spill
-; VI-NEXT: buffer_store_dword v48, off, s[0:3], s32 offset:624 ; 4-byte Folded Spill
+; VI-NEXT: buffer_store_dword v27, off, s[0:3], s32 offset:624 ; 4-byte Folded Spill
+; VI-NEXT: buffer_store_dword v48, off, s[0:3], s32 offset:628 ; 4-byte Folded Spill
; VI-NEXT: s_waitcnt vmcnt(12)
-; VI-NEXT: buffer_store_dword v63, off, s[0:3], s32 offset:628 ; 4-byte Folded Spill
-; VI-NEXT: buffer_store_dword v58, off, s[0:3], s32 offset:632 ; 4-byte Folded Spill
-; VI-NEXT: buffer_store_dword v47, off, s[0:3], s32 offset:636 ; 4-byte Folded Spill
-; VI-NEXT: buffer_store_dword v31, off, s[0:3], s32 offset:640 ; 4-byte Folded Spill
-; VI-NEXT: buffer_store_dword v46, off, s[0:3], s32 offset:644 ; 4-byte Folded Spill
-; VI-NEXT: buffer_store_dword v52, off, s[0:3], s32 offset:648 ; 4-byte Folded Spill
-; VI-NEXT: buffer_store_dword v2, off, s[0:3], s32 offset:652 ; 4-byte Folded Spill
-; VI-NEXT: buffer_store_dword v32, off, s[0:3], s32 offset:656 ; 4-byte Folded Spill
-; VI-NEXT: buffer_store_dword v4, off, s[0:3], s32 offset:660 ; 4-byte Folded Spill
-; VI-NEXT: buffer_store_dword v24, off, s[0:3], s32 offset:664 ; 4-byte Folded Spill
-; VI-NEXT: buffer_store_dword v30, off, s[0:3], s32 offset:668 ; 4-byte Folded Spill
-; VI-NEXT: buffer_store_dword v43, off, s[0:3], s32 offset:672 ; 4-byte Folded Spill
-; VI-NEXT: buffer_store_dword v23, off, s[0:3], s32 offset:676 ; 4-byte Folded Spill
-; VI-NEXT: buffer_store_dword v6, off, s[0:3], s32 offset:680 ; 4-byte Folded Spill
-; VI-NEXT: buffer_store_dword v16, off, s[0:3], s32 offset:684 ; 4-byte Folded Spill
-; VI-NEXT: buffer_store_dword v22, off, s[0:3], s32 offset:688 ; 4-byte Folded Spill
-; VI-NEXT: buffer_store_dword v34, off, s[0:3], s32 offset:692 ; 4-byte Folded Spill
-; VI-NEXT: buffer_store_dword v10, off, s[0:3], s32 offset:696 ; 4-byte Folded Spill
+; VI-NEXT: buffer_store_dword v63, off, s[0:3], s32 offset:632 ; 4-byte Folded Spill
+; VI-NEXT: buffer_store_dword v45, off, s[0:3], s32 offset:636 ; 4-byte Folded Spill
+; VI-NEXT: buffer_store_dword v47, off, s[0:3], s32 offset:640 ; 4-byte Folded Spill
+; VI-NEXT: buffer_store_dword v31, off, s[0:3], s32 offset:644 ; 4-byte Folded Spill
+; VI-NEXT: buffer_store_dword v46, off, s[0:3], s32 offset:648 ; 4-byte Folded Spill
+; VI-NEXT: buffer_store_dword v52, off, s[0:3], s32 offset:652 ; 4-byte Folded Spill
+; VI-NEXT: buffer_store_dword v2, off, s[0:3], s32 offset:656 ; 4-byte Folded Spill
+; VI-NEXT: buffer_store_dword v32, off, s[0:3], s32 offset:660 ; 4-byte Folded Spill
+; VI-NEXT: buffer_store_dword v4, off, s[0:3], s32 offset:664 ; 4-byte Folded Spill
+; VI-NEXT: buffer_store_dword v26, off, s[0:3], s32 offset:668 ; 4-byte Folded Spill
+; VI-NEXT: buffer_store_dword v24, off, s[0:3], s32 offset:672 ; 4-byte Folded Spill
+; VI-NEXT: buffer_store_dword v43, off, s[0:3], s32 offset:676 ; 4-byte Folded Spill
+; VI-NEXT: buffer_store_dword v30, off, s[0:3], s32 offset:680 ; 4-byte Folded Spill
+; VI-NEXT: buffer_store_dword v6, off, s[0:3], s32 offset:684 ; 4-byte Folded Spill
+; VI-NEXT: buffer_store_dword v12, off, s[0:3], s32 offset:688 ; 4-byte Folded Spill
+; VI-NEXT: buffer_store_dword v25, off, s[0:3], s32 offset:692 ; 4-byte Folded Spill
+; VI-NEXT: buffer_store_dword v34, off, s[0:3], s32 offset:696 ; 4-byte Folded Spill
; VI-NEXT: buffer_store_dword v14, off, s[0:3], s32 offset:700 ; 4-byte Folded Spill
-; VI-NEXT: buffer_store_dword v21, off, s[0:3], s32 offset:704 ; 4-byte Folded Spill
-; VI-NEXT: buffer_store_dword v28, off, s[0:3], s32 offset:708 ; 4-byte Folded Spill
-; VI-NEXT: buffer_store_dword v20, off, s[0:3], s32 offset:712 ; 4-byte Folded Spill
-; VI-NEXT: buffer_store_dword v51, off, s[0:3], s32 offset:716 ; 4-byte Folded Spill
-; VI-NEXT: buffer_store_dword v59, off, s[0:3], s32 offset:720 ; 4-byte Folded Spill
-; VI-NEXT: buffer_store_dword v19, off, s[0:3], s32 offset:724 ; 4-byte Folded Spill
-; VI-NEXT: buffer_store_dword v8, off, s[0:3], s32 offset:728 ; 4-byte Folded Spill
-; VI-NEXT: buffer_store_dword v25, off, s[0:3], s32 offset:732 ; 4-byte Folded Spill
-; VI-NEXT: buffer_store_dword v33, off, s[0:3], s32 offset:736 ; 4-byte Folded Spill
-; VI-NEXT: buffer_store_dword v35, off, s[0:3], s32 offset:740 ; 4-byte Folded Spill
-; VI-NEXT: buffer_store_dword v36, off, s[0:3], s32 offset:744 ; 4-byte Folded Spill
-; VI-NEXT: buffer_store_dword v57, off, s[0:3], s32 offset:748 ; 4-byte Folded Spill
-; VI-NEXT: buffer_store_dword v37, off, s[0:3], s32 offset:752 ; 4-byte Folded Spill
-; VI-NEXT: buffer_store_dword v38, off, s[0:3], s32 offset:756 ; 4-byte Folded Spill
-; VI-NEXT: buffer_store_dword v39, off, s[0:3], s32 offset:760 ; 4-byte Folded Spill
-; VI-NEXT: buffer_store_dword v17, off, s[0:3], s32 offset:764 ; 4-byte Folded Spill
+; VI-NEXT: buffer_store_dword v10, off, s[0:3], s32 offset:704 ; 4-byte Folded Spill
+; VI-NEXT: buffer_store_dword v21, off, s[0:3], s32 offset:708 ; 4-byte Folded Spill
+; VI-NEXT: buffer_store_dword v51, off, s[0:3], s32 offset:712 ; 4-byte Folded Spill
+; VI-NEXT: buffer_store_dword v20, off, s[0:3], s32 offset:716 ; 4-byte Folded Spill
+; VI-NEXT: buffer_store_dword v49, off, s[0:3], s32 offset:720 ; 4-byte Folded Spill
+; VI-NEXT: buffer_store_dword v8, off, s[0:3], s32 offset:724 ; 4-byte Folded Spill
+; VI-NEXT: buffer_store_dword v19, off, s[0:3], s32 offset:728 ; 4-byte Folded Spill
+; VI-NEXT: buffer_store_dword v56, off, s[0:3], s32 offset:732 ; 4-byte Folded Spill
+; VI-NEXT: buffer_store_dword v54, off, s[0:3], s32 offset:736 ; 4-byte Folded Spill
+; VI-NEXT: buffer_store_dword v33, off, s[0:3], s32 offset:740 ; 4-byte Folded Spill
+; VI-NEXT: buffer_store_dword v35, off, s[0:3], s32 offset:744 ; 4-byte Folded Spill
+; VI-NEXT: buffer_store_dword v36, off, s[0:3], s32 offset:748 ; 4-byte Folded Spill
+; VI-NEXT: buffer_store_dword v57, off, s[0:3], s32 offset:752 ; 4-byte Folded Spill
+; VI-NEXT: buffer_store_dword v37, off, s[0:3], s32 offset:756 ; 4-byte Folded Spill
+; VI-NEXT: buffer_store_dword v38, off, s[0:3], s32 offset:760 ; 4-byte Folded Spill
+; VI-NEXT: buffer_store_dword v39, off, s[0:3], s32 offset:764 ; 4-byte Folded Spill
; VI-NEXT: buffer_store_dword v15, off, s[0:3], s32 offset:768 ; 4-byte Folded Spill
-; VI-NEXT: buffer_store_dword v18, off, s[0:3], s32 offset:772 ; 4-byte Folded Spill
-; VI-NEXT: buffer_store_dword v13, off, s[0:3], s32 offset:776 ; 4-byte Folded Spill
-; VI-NEXT: buffer_store_dword v56, off, s[0:3], s32 offset:780 ; 4-byte Folded Spill
-; VI-NEXT: buffer_store_dword v50, off, s[0:3], s32 offset:784 ; 4-byte Folded Spill
-; VI-NEXT: buffer_store_dword v12, off, s[0:3], s32 offset:788 ; 4-byte Folded Spill
-; VI-NEXT: buffer_store_dword v11, off, s[0:3], s32 offset:792 ; 4-byte Folded Spill
+; VI-NEXT: buffer_store_dword v17, off, s[0:3], s32 offset:772 ; 4-byte Folded Spill
+; VI-NEXT: buffer_store_dword v18, off, s[0:3], s32 offset:776 ; 4-byte Folded Spill
+; VI-NEXT: buffer_store_dword v13, off, s[0:3], s32 offset:780 ; 4-byte Folded Spill
+; VI-NEXT: buffer_store_dword v59, off, s[0:3], s32 offset:784 ; 4-byte Folded Spill
+; VI-NEXT: buffer_store_dword v50, off, s[0:3], s32 offset:788 ; 4-byte Folded Spill
+; VI-NEXT: buffer_store_dword v16, off, s[0:3], s32 offset:792 ; 4-byte Folded Spill
+; VI-NEXT: buffer_store_dword v11, off, s[0:3], s32 offset:796 ; 4-byte Folded Spill
; VI-NEXT: s_cbranch_scc0 .LBB39_2
; VI-NEXT: ; %bb.1: ; %cmp.false
-; VI-NEXT: v_or_b32_sdwa v0, v2, v8 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
-; VI-NEXT: v_or_b32_sdwa v1, v4, v59 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; VI-NEXT: v_or_b32_sdwa v0, v2, v56 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; VI-NEXT: v_or_b32_sdwa v1, v4, v8 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
; VI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:488 ; 4-byte Folded Reload
; VI-NEXT: buffer_load_dword v3, off, s[0:3], s32 offset:484 ; 4-byte Folded Reload
; VI-NEXT: v_or_b32_sdwa v4, v0, v1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_0 src1_sel:DWORD
@@ -58689,12 +58705,12 @@ define inreg <32 x float> @bitcast_v128i8_to_v32f32_scalar(<128 x i8> inreg %a,
; VI-NEXT: s_lshl_b32 s7, s23, 8
; VI-NEXT: s_lshl_b32 s8, s27, 8
; VI-NEXT: s_waitcnt vmcnt(3)
-; VI-NEXT: v_or_b32_sdwa v2, v2, v14 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; VI-NEXT: v_or_b32_sdwa v2, v2, v10 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
; VI-NEXT: s_waitcnt vmcnt(2)
-; VI-NEXT: v_or_b32_sdwa v3, v3, v10 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; VI-NEXT: v_or_b32_sdwa v3, v3, v14 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
; VI-NEXT: v_or_b32_sdwa v5, v2, v3 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_0 src1_sel:DWORD
; VI-NEXT: s_waitcnt vmcnt(1)
-; VI-NEXT: v_or_b32_sdwa v0, v0, v16 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; VI-NEXT: v_or_b32_sdwa v0, v0, v12 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
; VI-NEXT: s_waitcnt vmcnt(0)
; VI-NEXT: v_or_b32_sdwa v1, v1, v6 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
; VI-NEXT: v_or_b32_sdwa v6, v0, v1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_0 src1_sel:DWORD
@@ -58723,9 +58739,7 @@ define inreg <32 x float> @bitcast_v128i8_to_v32f32_scalar(<128 x i8> inreg %a,
; VI-NEXT: v_or_b32_sdwa v0, v0, v1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
; VI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:432 ; 4-byte Folded Reload
; VI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:528 ; 4-byte Folded Reload
-; VI-NEXT: buffer_store_dword v29, off, s[0:3], s32 offset:796 ; 4-byte Folded Spill
-; VI-NEXT: v_mov_b32_e32 v29, v9
-; VI-NEXT: s_waitcnt vmcnt(1)
+; VI-NEXT: s_waitcnt vmcnt(0)
; VI-NEXT: v_or_b32_sdwa v1, v1, v2 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
; VI-NEXT: v_or_b32_sdwa v9, v0, v1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_0 src1_sel:DWORD
; VI-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:428 ; 4-byte Folded Reload
@@ -58748,15 +58762,15 @@ define inreg <32 x float> @bitcast_v128i8_to_v32f32_scalar(<128 x i8> inreg %a,
; VI-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:508 ; 4-byte Folded Reload
; VI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:504 ; 4-byte Folded Reload
; VI-NEXT: s_waitcnt vmcnt(1)
-; VI-NEXT: v_or_b32_sdwa v0, v12, v0 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; VI-NEXT: v_or_b32_sdwa v0, v16, v0 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
; VI-NEXT: s_waitcnt vmcnt(0)
; VI-NEXT: v_or_b32_sdwa v1, v50, v1 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
; VI-NEXT: v_or_b32_sdwa v12, v0, v1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_0 src1_sel:DWORD
-; VI-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:580 ; 4-byte Folded Reload
+; VI-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:588 ; 4-byte Folded Reload
; VI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:500 ; 4-byte Folded Reload
; VI-NEXT: s_waitcnt vmcnt(1)
; VI-NEXT: v_mov_b32_e32 v50, v0
-; VI-NEXT: v_or_b32_sdwa v0, v56, v0 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; VI-NEXT: v_or_b32_sdwa v0, v59, v0 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
; VI-NEXT: s_waitcnt vmcnt(0)
; VI-NEXT: v_or_b32_sdwa v1, v13, v1 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
; VI-NEXT: v_or_b32_sdwa v13, v0, v1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_0 src1_sel:DWORD
@@ -58766,18 +58780,18 @@ define inreg <32 x float> @bitcast_v128i8_to_v32f32_scalar(<128 x i8> inreg %a,
; VI-NEXT: v_mov_b32_e32 v59, v0
; VI-NEXT: v_or_b32_sdwa v0, v18, v0 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
; VI-NEXT: s_waitcnt vmcnt(0)
-; VI-NEXT: v_or_b32_sdwa v1, v15, v1 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; VI-NEXT: v_or_b32_sdwa v1, v17, v1 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
; VI-NEXT: v_or_b32_sdwa v14, v0, v1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_0 src1_sel:DWORD
-; VI-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:608 ; 4-byte Folded Reload
+; VI-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:616 ; 4-byte Folded Reload
; VI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:492 ; 4-byte Folded Reload
; VI-NEXT: s_waitcnt vmcnt(1)
; VI-NEXT: v_mov_b32_e32 v56, v0
-; VI-NEXT: v_or_b32_sdwa v0, v17, v0 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; VI-NEXT: v_or_b32_sdwa v0, v15, v0 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
; VI-NEXT: s_waitcnt vmcnt(0)
; VI-NEXT: v_or_b32_sdwa v1, v39, v1 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
; VI-NEXT: v_or_b32_sdwa v15, v0, v1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_0 src1_sel:DWORD
-; VI-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:600 ; 4-byte Folded Reload
-; VI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:604 ; 4-byte Folded Reload
+; VI-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:612 ; 4-byte Folded Reload
+; VI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:608 ; 4-byte Folded Reload
; VI-NEXT: s_waitcnt vmcnt(1)
; VI-NEXT: v_mov_b32_e32 v39, v0
; VI-NEXT: v_or_b32_sdwa v0, v38, v0 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
@@ -58785,7 +58799,7 @@ define inreg <32 x float> @bitcast_v128i8_to_v32f32_scalar(<128 x i8> inreg %a,
; VI-NEXT: v_mov_b32_e32 v38, v1
; VI-NEXT: v_or_b32_sdwa v1, v37, v1 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
; VI-NEXT: v_or_b32_sdwa v16, v0, v1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_0 src1_sel:DWORD
-; VI-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:596 ; 4-byte Folded Reload
+; VI-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:604 ; 4-byte Folded Reload
; VI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:480 ; 4-byte Folded Reload
; VI-NEXT: s_waitcnt vmcnt(1)
; VI-NEXT: v_mov_b32_e32 v37, v0
@@ -58793,8 +58807,8 @@ define inreg <32 x float> @bitcast_v128i8_to_v32f32_scalar(<128 x i8> inreg %a,
; VI-NEXT: s_waitcnt vmcnt(0)
; VI-NEXT: v_or_b32_sdwa v1, v36, v1 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
; VI-NEXT: v_or_b32_sdwa v17, v0, v1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_0 src1_sel:DWORD
-; VI-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:572 ; 4-byte Folded Reload
-; VI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:588 ; 4-byte Folded Reload
+; VI-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:596 ; 4-byte Folded Reload
+; VI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:600 ; 4-byte Folded Reload
; VI-NEXT: s_waitcnt vmcnt(1)
; VI-NEXT: v_mov_b32_e32 v36, v0
; VI-NEXT: v_or_b32_sdwa v0, v35, v0 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
@@ -58806,39 +58820,41 @@ define inreg <32 x float> @bitcast_v128i8_to_v32f32_scalar(<128 x i8> inreg %a,
; VI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:468 ; 4-byte Folded Reload
; VI-NEXT: s_waitcnt vmcnt(1)
; VI-NEXT: v_mov_b32_e32 v33, v0
-; VI-NEXT: v_or_b32_sdwa v0, v25, v0 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; VI-NEXT: v_or_b32_sdwa v0, v54, v0 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
; VI-NEXT: s_waitcnt vmcnt(0)
; VI-NEXT: v_or_b32_sdwa v1, v19, v1 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
; VI-NEXT: v_or_b32_sdwa v19, v0, v1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_0 src1_sel:DWORD
; VI-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:456 ; 4-byte Folded Reload
; VI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:444 ; 4-byte Folded Reload
; VI-NEXT: s_waitcnt vmcnt(1)
-; VI-NEXT: v_or_b32_sdwa v0, v51, v0 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; VI-NEXT: v_or_b32_sdwa v0, v49, v0 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
; VI-NEXT: s_waitcnt vmcnt(0)
; VI-NEXT: v_or_b32_sdwa v1, v20, v1 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
; VI-NEXT: v_or_b32_sdwa v20, v0, v1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_0 src1_sel:DWORD
; VI-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:436 ; 4-byte Folded Reload
; VI-NEXT: v_or_b32_sdwa v1, v21, v52 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
-; VI-NEXT: v_mov_b32_e32 v51, v3
; VI-NEXT: s_waitcnt vmcnt(0)
-; VI-NEXT: v_or_b32_sdwa v0, v28, v0 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; VI-NEXT: v_or_b32_sdwa v0, v51, v0 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
; VI-NEXT: v_or_b32_sdwa v21, v0, v1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_0 src1_sel:DWORD
; VI-NEXT: v_or_b32_sdwa v0, v34, v31 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
-; VI-NEXT: v_or_b32_sdwa v1, v22, v26 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; VI-NEXT: v_or_b32_sdwa v1, v25, v22 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; VI-NEXT: v_mov_b32_e32 v34, v22
; VI-NEXT: v_or_b32_sdwa v22, v0, v1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_0 src1_sel:DWORD
-; VI-NEXT: v_or_b32_sdwa v0, v23, v32 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
-; VI-NEXT: v_or_b32_sdwa v1, v43, v54 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; VI-NEXT: v_or_b32_sdwa v0, v30, v32 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; VI-NEXT: v_or_b32_sdwa v1, v43, v23 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; VI-NEXT: v_mov_b32_e32 v32, v23
; VI-NEXT: v_or_b32_sdwa v23, v0, v1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_0 src1_sel:DWORD
-; VI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:576 ; 4-byte Folded Reload
-; VI-NEXT: v_mov_b32_e32 v43, v49
-; VI-NEXT: v_or_b32_sdwa v0, v30, v49 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
-; VI-NEXT: v_mov_b32_e32 v32, v54
-; VI-NEXT: v_mov_b32_e32 v34, v26
+; VI-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:580 ; 4-byte Folded Reload
+; VI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:572 ; 4-byte Folded Reload
+; VI-NEXT: v_mov_b32_e32 v51, v3
+; VI-NEXT: s_waitcnt vmcnt(1)
+; VI-NEXT: v_mov_b32_e32 v43, v0
+; VI-NEXT: v_or_b32_sdwa v0, v24, v0 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
; VI-NEXT: s_waitcnt vmcnt(0)
; VI-NEXT: v_mov_b32_e32 v49, v1
-; VI-NEXT: v_or_b32_sdwa v1, v24, v1 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; VI-NEXT: v_or_b32_sdwa v1, v26, v1 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
; VI-NEXT: v_or_b32_sdwa v24, v0, v1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_0 src1_sel:DWORD
-; VI-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:612 ; 4-byte Folded Reload
+; VI-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:576 ; 4-byte Folded Reload
; VI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:416 ; 4-byte Folded Reload
; VI-NEXT: s_waitcnt vmcnt(1)
; VI-NEXT: v_mov_b32_e32 v54, v0
@@ -58848,28 +58864,26 @@ define inreg <32 x float> @bitcast_v128i8_to_v32f32_scalar(<128 x i8> inreg %a,
; VI-NEXT: v_mov_b32_e32 v46, v61
; VI-NEXT: v_or_b32_sdwa v25, v0, v1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_0 src1_sel:DWORD
; VI-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:412 ; 4-byte Folded Reload
-; VI-NEXT: buffer_load_dword v61, off, s[0:3], s32 offset:564 ; 4-byte Folded Reload
+; VI-NEXT: buffer_load_dword v61, off, s[0:3], s32 offset:568 ; 4-byte Folded Reload
; VI-NEXT: s_waitcnt vmcnt(1)
; VI-NEXT: v_or_b32_sdwa v0, v47, v0 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
; VI-NEXT: s_waitcnt vmcnt(0)
-; VI-NEXT: v_or_b32_sdwa v1, v45, v61 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; VI-NEXT: v_or_b32_sdwa v1, v58, v61 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
; VI-NEXT: v_or_b32_sdwa v26, v0, v1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_0 src1_sel:DWORD
-; VI-NEXT: v_or_b32_sdwa v0, v58, v48 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; VI-NEXT: v_or_b32_sdwa v0, v45, v48 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
; VI-NEXT: v_or_b32_sdwa v1, v44, v27 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
-; VI-NEXT: v_mov_b32_e32 v47, v45
; VI-NEXT: v_or_b32_sdwa v27, v0, v1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_0 src1_sel:DWORD
-; VI-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:796 ; 4-byte Folded Reload
-; VI-NEXT: buffer_load_dword v45, off, s[0:3], s32 offset:568 ; 4-byte Folded Reload
-; VI-NEXT: v_mov_b32_e32 v58, v44
-; VI-NEXT: s_waitcnt vmcnt(1)
-; VI-NEXT: v_mov_b32_e32 v48, v0
-; VI-NEXT: v_or_b32_sdwa v0, v63, v0 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; VI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:556 ; 4-byte Folded Reload
+; VI-NEXT: v_or_b32_sdwa v0, v63, v28 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; VI-NEXT: v_mov_b32_e32 v48, v28
+; VI-NEXT: v_mov_b32_e32 v47, v58
+; VI-NEXT: v_mov_b32_e32 v45, v44
+; VI-NEXT: v_mov_b32_e32 v63, v42
; VI-NEXT: s_waitcnt vmcnt(0)
-; VI-NEXT: v_or_b32_sdwa v1, v42, v45 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; VI-NEXT: v_or_b32_sdwa v1, v42, v1 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
; VI-NEXT: v_or_b32_sdwa v28, v0, v1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_0 src1_sel:DWORD
; VI-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:404 ; 4-byte Folded Reload
; VI-NEXT: v_or_b32_sdwa v1, v40, v29 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
-; VI-NEXT: v_mov_b32_e32 v63, v42
; VI-NEXT: s_waitcnt vmcnt(0)
; VI-NEXT: v_or_b32_sdwa v0, v41, v0 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
; VI-NEXT: v_or_b32_sdwa v29, v0, v1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_0 src1_sel:DWORD
@@ -58885,8 +58899,8 @@ define inreg <32 x float> @bitcast_v128i8_to_v32f32_scalar(<128 x i8> inreg %a,
; VI-NEXT: s_waitcnt vmcnt(0)
; VI-NEXT: v_or_b32_sdwa v1, v53, v1 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
; VI-NEXT: v_or_b32_sdwa v31, v0, v1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_0 src1_sel:DWORD
-; VI-NEXT: buffer_load_dword v52, off, s[0:3], s32 offset:556 ; 4-byte Folded Reload
-; VI-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:560 ; 4-byte Folded Reload
+; VI-NEXT: buffer_load_dword v52, off, s[0:3], s32 offset:560 ; 4-byte Folded Reload
+; VI-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:564 ; 4-byte Folded Reload
; VI-NEXT: s_waitcnt vmcnt(0)
; VI-NEXT: v_mov_b32_e32 v57, v0
; VI-NEXT: v_or_b32_sdwa v0, v52, v0 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
@@ -58920,44 +58934,43 @@ define inreg <32 x float> @bitcast_v128i8_to_v32f32_scalar(<128 x i8> inreg %a,
; VI-NEXT: s_mov_b64 s[4:5], 0
; VI-NEXT: s_branch .LBB39_3
; VI-NEXT: .LBB39_2:
-; VI-NEXT: buffer_load_dword v57, off, s[0:3], s32 offset:560 ; 4-byte Folded Reload
-; VI-NEXT: buffer_load_dword v50, off, s[0:3], s32 offset:580 ; 4-byte Folded Reload
+; VI-NEXT: buffer_load_dword v57, off, s[0:3], s32 offset:564 ; 4-byte Folded Reload
+; VI-NEXT: buffer_load_dword v50, off, s[0:3], s32 offset:588 ; 4-byte Folded Reload
; VI-NEXT: buffer_load_dword v59, off, s[0:3], s32 offset:584 ; 4-byte Folded Reload
-; VI-NEXT: buffer_load_dword v56, off, s[0:3], s32 offset:608 ; 4-byte Folded Reload
-; VI-NEXT: buffer_load_dword v39, off, s[0:3], s32 offset:600 ; 4-byte Folded Reload
-; VI-NEXT: buffer_load_dword v38, off, s[0:3], s32 offset:604 ; 4-byte Folded Reload
-; VI-NEXT: buffer_load_dword v37, off, s[0:3], s32 offset:596 ; 4-byte Folded Reload
-; VI-NEXT: buffer_load_dword v36, off, s[0:3], s32 offset:572 ; 4-byte Folded Reload
-; VI-NEXT: buffer_load_dword v35, off, s[0:3], s32 offset:588 ; 4-byte Folded Reload
+; VI-NEXT: buffer_load_dword v56, off, s[0:3], s32 offset:616 ; 4-byte Folded Reload
+; VI-NEXT: buffer_load_dword v39, off, s[0:3], s32 offset:612 ; 4-byte Folded Reload
+; VI-NEXT: buffer_load_dword v38, off, s[0:3], s32 offset:608 ; 4-byte Folded Reload
+; VI-NEXT: buffer_load_dword v37, off, s[0:3], s32 offset:604 ; 4-byte Folded Reload
+; VI-NEXT: buffer_load_dword v36, off, s[0:3], s32 offset:596 ; 4-byte Folded Reload
+; VI-NEXT: buffer_load_dword v35, off, s[0:3], s32 offset:600 ; 4-byte Folded Reload
; VI-NEXT: buffer_load_dword v33, off, s[0:3], s32 offset:592 ; 4-byte Folded Reload
-; VI-NEXT: v_mov_b32_e32 v32, v54
-; VI-NEXT: v_mov_b32_e32 v43, v49
+; VI-NEXT: buffer_load_dword v43, off, s[0:3], s32 offset:580 ; 4-byte Folded Reload
+; VI-NEXT: buffer_load_dword v49, off, s[0:3], s32 offset:572 ; 4-byte Folded Reload
; VI-NEXT: v_mov_b32_e32 v46, v61
-; VI-NEXT: v_mov_b32_e32 v47, v45
-; VI-NEXT: buffer_load_dword v54, off, s[0:3], s32 offset:612 ; 4-byte Folded Reload
-; VI-NEXT: buffer_load_dword v49, off, s[0:3], s32 offset:576 ; 4-byte Folded Reload
-; VI-NEXT: buffer_load_dword v45, off, s[0:3], s32 offset:568 ; 4-byte Folded Reload
-; VI-NEXT: buffer_load_dword v52, off, s[0:3], s32 offset:556 ; 4-byte Folded Reload
-; VI-NEXT: buffer_load_dword v61, off, s[0:3], s32 offset:564 ; 4-byte Folded Reload
-; VI-NEXT: v_mov_b32_e32 v34, v26
-; VI-NEXT: v_mov_b32_e32 v58, v44
+; VI-NEXT: buffer_load_dword v54, off, s[0:3], s32 offset:576 ; 4-byte Folded Reload
+; VI-NEXT: buffer_load_dword v52, off, s[0:3], s32 offset:560 ; 4-byte Folded Reload
+; VI-NEXT: buffer_load_dword v61, off, s[0:3], s32 offset:568 ; 4-byte Folded Reload
+; VI-NEXT: v_mov_b32_e32 v34, v22
+; VI-NEXT: v_mov_b32_e32 v32, v23
+; VI-NEXT: v_mov_b32_e32 v47, v58
+; VI-NEXT: v_mov_b32_e32 v45, v44
; VI-NEXT: s_waitcnt vmcnt(14)
; VI-NEXT: v_mov_b32_e32 v63, v42
; VI-NEXT: v_mov_b32_e32 v51, v7
-; VI-NEXT: v_mov_b32_e32 v48, v29
+; VI-NEXT: v_mov_b32_e32 v48, v28
; VI-NEXT: s_mov_b64 s[4:5], -1
; VI-NEXT: ; implicit-def: $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7_vgpr8_vgpr9_vgpr10_vgpr11_vgpr12_vgpr13_vgpr14_vgpr15_vgpr16_vgpr17_vgpr18_vgpr19_vgpr20_vgpr21_vgpr22_vgpr23_vgpr24_vgpr25_vgpr26_vgpr27_vgpr28_vgpr29_vgpr30_vgpr31
; VI-NEXT: .LBB39_3: ; %Flow
; VI-NEXT: s_andn2_b64 vcc, exec, s[4:5]
-; VI-NEXT: v_mov_b32_e32 v44, v47
-; VI-NEXT: v_mov_b32_e32 v47, v46
+; VI-NEXT: v_mov_b32_e32 v42, v45
+; VI-NEXT: v_mov_b32_e32 v45, v46
; VI-NEXT: s_waitcnt vmcnt(3)
; VI-NEXT: v_mov_b32_e32 v46, v49
; VI-NEXT: s_cbranch_vccnz .LBB39_5
; VI-NEXT: ; %bb.4: ; %cmp.true
-; VI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:652 ; 4-byte Folded Reload
-; VI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:728 ; 4-byte Folded Reload
-; VI-NEXT: buffer_load_dword v3, off, s[0:3], s32 offset:720 ; 4-byte Folded Reload
+; VI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:656 ; 4-byte Folded Reload
+; VI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:732 ; 4-byte Folded Reload
+; VI-NEXT: buffer_load_dword v3, off, s[0:3], s32 offset:724 ; 4-byte Folded Reload
; VI-NEXT: s_add_i32 s28, s28, 3
; VI-NEXT: s_and_b32 s4, s28, 0xff
; VI-NEXT: s_lshl_b32 s5, s29, 8
@@ -59011,7 +59024,7 @@ define inreg <32 x float> @bitcast_v128i8_to_v32f32_scalar(<128 x i8> inreg %a,
; VI-NEXT: v_add_u32_e32 v1, vcc, 3, v1
; VI-NEXT: s_waitcnt vmcnt(1)
; VI-NEXT: v_or_b32_sdwa v1, v2, v1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:BYTE_0
-; VI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:660 ; 4-byte Folded Reload
+; VI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:664 ; 4-byte Folded Reload
; VI-NEXT: v_add_u32_e32 v1, vcc, 0x300, v1
; VI-NEXT: s_waitcnt vmcnt(0)
; VI-NEXT: v_add_u32_e32 v2, vcc, 3, v2
@@ -59020,8 +59033,8 @@ define inreg <32 x float> @bitcast_v128i8_to_v32f32_scalar(<128 x i8> inreg %a,
; VI-NEXT: v_add_u32_e32 v3, vcc, 0x3000000, v0
; VI-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:488 ; 4-byte Folded Reload
; VI-NEXT: v_add_u32_e32 v4, vcc, 0x3000000, v1
-; VI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:700 ; 4-byte Folded Reload
-; VI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:696 ; 4-byte Folded Reload
+; VI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:704 ; 4-byte Folded Reload
+; VI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:700 ; 4-byte Folded Reload
; VI-NEXT: s_waitcnt vmcnt(2)
; VI-NEXT: v_add_u32_e32 v0, vcc, 3, v0
; VI-NEXT: s_waitcnt vmcnt(1)
@@ -59034,8 +59047,8 @@ define inreg <32 x float> @bitcast_v128i8_to_v32f32_scalar(<128 x i8> inreg %a,
; VI-NEXT: v_or_b32_sdwa v0, v1, v0 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:WORD_0
; VI-NEXT: v_add_u32_e32 v5, vcc, 0x3000000, v0
; VI-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:476 ; 4-byte Folded Reload
-; VI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:684 ; 4-byte Folded Reload
-; VI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:680 ; 4-byte Folded Reload
+; VI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:688 ; 4-byte Folded Reload
+; VI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:684 ; 4-byte Folded Reload
; VI-NEXT: s_waitcnt vmcnt(2)
; VI-NEXT: v_add_u32_e32 v0, vcc, 3, v0
; VI-NEXT: s_waitcnt vmcnt(1)
@@ -59110,29 +59123,29 @@ define inreg <32 x float> @bitcast_v128i8_to_v32f32_scalar(<128 x i8> inreg %a,
; VI-NEXT: v_add_u32_e32 v0, vcc, 3, v0
; VI-NEXT: s_waitcnt vmcnt(1)
; VI-NEXT: v_or_b32_sdwa v0, v1, v0 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:BYTE_0
-; VI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:792 ; 4-byte Folded Reload
+; VI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:796 ; 4-byte Folded Reload
; VI-NEXT: v_add_u32_e32 v0, vcc, 0x300, v0
; VI-NEXT: s_waitcnt vmcnt(0)
; VI-NEXT: v_add_u32_e32 v1, vcc, 3, v1
; VI-NEXT: v_or_b32_sdwa v1, v2, v1 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:BYTE_0
; VI-NEXT: v_or_b32_sdwa v0, v1, v0 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:WORD_0
; VI-NEXT: v_add_u32_e32 v11, vcc, 0x3000000, v0
-; VI-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:788 ; 4-byte Folded Reload
+; VI-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:792 ; 4-byte Folded Reload
; VI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:508 ; 4-byte Folded Reload
; VI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:504 ; 4-byte Folded Reload
; VI-NEXT: s_waitcnt vmcnt(2)
; VI-NEXT: v_add_u32_e32 v0, vcc, 3, v0
; VI-NEXT: s_waitcnt vmcnt(1)
; VI-NEXT: v_or_b32_sdwa v0, v1, v0 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:BYTE_0
-; VI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:784 ; 4-byte Folded Reload
+; VI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:788 ; 4-byte Folded Reload
; VI-NEXT: v_add_u32_e32 v0, vcc, 0x300, v0
; VI-NEXT: s_waitcnt vmcnt(0)
; VI-NEXT: v_add_u32_e32 v1, vcc, 3, v1
; VI-NEXT: v_or_b32_sdwa v1, v2, v1 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:BYTE_0
; VI-NEXT: v_or_b32_sdwa v0, v1, v0 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:WORD_0
; VI-NEXT: v_add_u32_e32 v12, vcc, 0x3000000, v0
-; VI-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:780 ; 4-byte Folded Reload
-; VI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:776 ; 4-byte Folded Reload
+; VI-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:784 ; 4-byte Folded Reload
+; VI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:780 ; 4-byte Folded Reload
; VI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:500 ; 4-byte Folded Reload
; VI-NEXT: s_waitcnt vmcnt(2)
; VI-NEXT: v_add_u32_e32 v0, vcc, 3, v0
@@ -59144,8 +59157,8 @@ define inreg <32 x float> @bitcast_v128i8_to_v32f32_scalar(<128 x i8> inreg %a,
; VI-NEXT: v_add_u32_e32 v0, vcc, 0x300, v0
; VI-NEXT: v_or_b32_sdwa v0, v1, v0 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:WORD_0
; VI-NEXT: v_add_u32_e32 v13, vcc, 0x3000000, v0
-; VI-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:772 ; 4-byte Folded Reload
-; VI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:768 ; 4-byte Folded Reload
+; VI-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:776 ; 4-byte Folded Reload
+; VI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:772 ; 4-byte Folded Reload
; VI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:496 ; 4-byte Folded Reload
; VI-NEXT: s_waitcnt vmcnt(2)
; VI-NEXT: v_add_u32_e32 v0, vcc, 3, v0
@@ -59157,8 +59170,8 @@ define inreg <32 x float> @bitcast_v128i8_to_v32f32_scalar(<128 x i8> inreg %a,
; VI-NEXT: v_add_u32_e32 v0, vcc, 0x300, v0
; VI-NEXT: v_or_b32_sdwa v0, v1, v0 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:WORD_0
; VI-NEXT: v_add_u32_e32 v14, vcc, 0x3000000, v0
-; VI-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:764 ; 4-byte Folded Reload
-; VI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:760 ; 4-byte Folded Reload
+; VI-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:768 ; 4-byte Folded Reload
+; VI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:764 ; 4-byte Folded Reload
; VI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:492 ; 4-byte Folded Reload
; VI-NEXT: s_waitcnt vmcnt(2)
; VI-NEXT: v_add_u32_e32 v0, vcc, 3, v0
@@ -59170,8 +59183,8 @@ define inreg <32 x float> @bitcast_v128i8_to_v32f32_scalar(<128 x i8> inreg %a,
; VI-NEXT: v_add_u32_e32 v0, vcc, 0x300, v0
; VI-NEXT: v_or_b32_sdwa v0, v1, v0 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:WORD_0
; VI-NEXT: v_add_u32_e32 v15, vcc, 0x3000000, v0
-; VI-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:756 ; 4-byte Folded Reload
-; VI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:752 ; 4-byte Folded Reload
+; VI-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:760 ; 4-byte Folded Reload
+; VI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:756 ; 4-byte Folded Reload
; VI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:480 ; 4-byte Folded Reload
; VI-NEXT: s_waitcnt vmcnt(2)
; VI-NEXT: v_add_u32_e32 v0, vcc, 3, v0
@@ -59182,8 +59195,8 @@ define inreg <32 x float> @bitcast_v128i8_to_v32f32_scalar(<128 x i8> inreg %a,
; VI-NEXT: v_add_u32_e32 v0, vcc, 0x300, v0
; VI-NEXT: v_or_b32_sdwa v0, v1, v0 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:WORD_0
; VI-NEXT: v_add_u32_e32 v16, vcc, 0x3000000, v0
-; VI-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:748 ; 4-byte Folded Reload
-; VI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:744 ; 4-byte Folded Reload
+; VI-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:752 ; 4-byte Folded Reload
+; VI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:748 ; 4-byte Folded Reload
; VI-NEXT: s_waitcnt vmcnt(1)
; VI-NEXT: v_add_u32_e32 v0, vcc, 3, v0
; VI-NEXT: v_or_b32_sdwa v0, v37, v0 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:BYTE_0
@@ -59193,8 +59206,8 @@ define inreg <32 x float> @bitcast_v128i8_to_v32f32_scalar(<128 x i8> inreg %a,
; VI-NEXT: v_add_u32_e32 v0, vcc, 0x300, v0
; VI-NEXT: v_or_b32_sdwa v0, v1, v0 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:WORD_0
; VI-NEXT: v_add_u32_e32 v17, vcc, 0x3000000, v0
-; VI-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:740 ; 4-byte Folded Reload
-; VI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:736 ; 4-byte Folded Reload
+; VI-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:744 ; 4-byte Folded Reload
+; VI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:740 ; 4-byte Folded Reload
; VI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:468 ; 4-byte Folded Reload
; VI-NEXT: s_waitcnt vmcnt(2)
; VI-NEXT: v_add_u32_e32 v0, vcc, 3, v0
@@ -59205,8 +59218,8 @@ define inreg <32 x float> @bitcast_v128i8_to_v32f32_scalar(<128 x i8> inreg %a,
; VI-NEXT: v_add_u32_e32 v0, vcc, 0x300, v0
; VI-NEXT: v_or_b32_sdwa v0, v1, v0 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:WORD_0
; VI-NEXT: v_add_u32_e32 v18, vcc, 0x3000000, v0
-; VI-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:732 ; 4-byte Folded Reload
-; VI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:724 ; 4-byte Folded Reload
+; VI-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:736 ; 4-byte Folded Reload
+; VI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:728 ; 4-byte Folded Reload
; VI-NEXT: s_waitcnt vmcnt(1)
; VI-NEXT: v_add_u32_e32 v0, vcc, 3, v0
; VI-NEXT: v_or_b32_sdwa v0, v33, v0 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:BYTE_0
@@ -59216,63 +59229,63 @@ define inreg <32 x float> @bitcast_v128i8_to_v32f32_scalar(<128 x i8> inreg %a,
; VI-NEXT: v_add_u32_e32 v0, vcc, 0x300, v0
; VI-NEXT: v_or_b32_sdwa v0, v1, v0 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:WORD_0
; VI-NEXT: v_add_u32_e32 v19, vcc, 0x3000000, v0
-; VI-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:716 ; 4-byte Folded Reload
+; VI-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:720 ; 4-byte Folded Reload
; VI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:456 ; 4-byte Folded Reload
; VI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:444 ; 4-byte Folded Reload
; VI-NEXT: s_waitcnt vmcnt(2)
; VI-NEXT: v_add_u32_e32 v0, vcc, 3, v0
; VI-NEXT: s_waitcnt vmcnt(1)
; VI-NEXT: v_or_b32_sdwa v0, v1, v0 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:BYTE_0
-; VI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:712 ; 4-byte Folded Reload
+; VI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:716 ; 4-byte Folded Reload
; VI-NEXT: v_add_u32_e32 v0, vcc, 0x300, v0
; VI-NEXT: s_waitcnt vmcnt(0)
; VI-NEXT: v_add_u32_e32 v1, vcc, 3, v1
; VI-NEXT: v_or_b32_sdwa v1, v2, v1 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:BYTE_0
; VI-NEXT: v_or_b32_sdwa v0, v1, v0 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:WORD_0
; VI-NEXT: v_add_u32_e32 v20, vcc, 0x3000000, v0
-; VI-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:708 ; 4-byte Folded Reload
+; VI-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:712 ; 4-byte Folded Reload
; VI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:436 ; 4-byte Folded Reload
-; VI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:648 ; 4-byte Folded Reload
+; VI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:652 ; 4-byte Folded Reload
; VI-NEXT: s_waitcnt vmcnt(2)
; VI-NEXT: v_add_u32_e32 v0, vcc, 3, v0
; VI-NEXT: s_waitcnt vmcnt(1)
; VI-NEXT: v_or_b32_sdwa v0, v1, v0 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:BYTE_0
-; VI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:704 ; 4-byte Folded Reload
+; VI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:708 ; 4-byte Folded Reload
; VI-NEXT: v_add_u32_e32 v0, vcc, 0x300, v0
; VI-NEXT: s_waitcnt vmcnt(0)
; VI-NEXT: v_add_u32_e32 v1, vcc, 3, v1
; VI-NEXT: v_or_b32_sdwa v1, v2, v1 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:BYTE_0
; VI-NEXT: v_or_b32_sdwa v0, v1, v0 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:WORD_0
; VI-NEXT: v_add_u32_e32 v21, vcc, 0x3000000, v0
-; VI-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:692 ; 4-byte Folded Reload
-; VI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:640 ; 4-byte Folded Reload
+; VI-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:696 ; 4-byte Folded Reload
+; VI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:644 ; 4-byte Folded Reload
; VI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:416 ; 4-byte Folded Reload
; VI-NEXT: s_waitcnt vmcnt(2)
; VI-NEXT: v_add_u32_e32 v0, vcc, 3, v0
; VI-NEXT: s_waitcnt vmcnt(1)
; VI-NEXT: v_or_b32_sdwa v0, v1, v0 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:BYTE_0
-; VI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:688 ; 4-byte Folded Reload
+; VI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:692 ; 4-byte Folded Reload
; VI-NEXT: v_add_u32_e32 v0, vcc, 0x300, v0
; VI-NEXT: s_waitcnt vmcnt(0)
; VI-NEXT: v_add_u32_e32 v1, vcc, 3, v1
; VI-NEXT: v_or_b32_sdwa v1, v34, v1 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:BYTE_0
; VI-NEXT: v_or_b32_sdwa v0, v1, v0 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:WORD_0
; VI-NEXT: v_add_u32_e32 v22, vcc, 0x3000000, v0
-; VI-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:676 ; 4-byte Folded Reload
-; VI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:656 ; 4-byte Folded Reload
+; VI-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:680 ; 4-byte Folded Reload
+; VI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:660 ; 4-byte Folded Reload
; VI-NEXT: s_waitcnt vmcnt(1)
; VI-NEXT: v_add_u32_e32 v0, vcc, 3, v0
; VI-NEXT: s_waitcnt vmcnt(0)
; VI-NEXT: v_or_b32_sdwa v0, v1, v0 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:BYTE_0
-; VI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:672 ; 4-byte Folded Reload
+; VI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:676 ; 4-byte Folded Reload
; VI-NEXT: v_add_u32_e32 v0, vcc, 0x300, v0
; VI-NEXT: s_waitcnt vmcnt(0)
; VI-NEXT: v_add_u32_e32 v1, vcc, 3, v1
; VI-NEXT: v_or_b32_sdwa v1, v32, v1 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:BYTE_0
; VI-NEXT: v_or_b32_sdwa v0, v1, v0 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:WORD_0
; VI-NEXT: v_add_u32_e32 v23, vcc, 0x3000000, v0
-; VI-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:668 ; 4-byte Folded Reload
-; VI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:664 ; 4-byte Folded Reload
+; VI-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:672 ; 4-byte Folded Reload
+; VI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:668 ; 4-byte Folded Reload
; VI-NEXT: s_waitcnt vmcnt(1)
; VI-NEXT: v_add_u32_e32 v0, vcc, 3, v0
; VI-NEXT: v_or_b32_sdwa v0, v43, v0 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:BYTE_0
@@ -59282,54 +59295,57 @@ define inreg <32 x float> @bitcast_v128i8_to_v32f32_scalar(<128 x i8> inreg %a,
; VI-NEXT: v_add_u32_e32 v0, vcc, 0x300, v0
; VI-NEXT: v_or_b32_sdwa v0, v1, v0 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:WORD_0
; VI-NEXT: v_add_u32_e32 v24, vcc, 0x3000000, v0
-; VI-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:644 ; 4-byte Folded Reload
-; VI-NEXT: v_add_u32_e32 v1, vcc, 3, v47
+; VI-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:648 ; 4-byte Folded Reload
+; VI-NEXT: v_add_u32_e32 v1, vcc, 3, v45
; VI-NEXT: v_or_b32_sdwa v1, v2, v1 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:BYTE_0
-; VI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:620 ; 4-byte Folded Reload
+; VI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:624 ; 4-byte Folded Reload
; VI-NEXT: s_waitcnt vmcnt(1)
; VI-NEXT: v_add_u32_e32 v0, vcc, 3, v0
; VI-NEXT: v_or_b32_sdwa v0, v54, v0 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:BYTE_0
; VI-NEXT: v_add_u32_e32 v0, vcc, 0x300, v0
; VI-NEXT: v_or_b32_sdwa v0, v1, v0 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:WORD_0
; VI-NEXT: v_add_u32_e32 v25, vcc, 0x3000000, v0
-; VI-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:636 ; 4-byte Folded Reload
+; VI-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:640 ; 4-byte Folded Reload
; VI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:412 ; 4-byte Folded Reload
; VI-NEXT: s_waitcnt vmcnt(1)
; VI-NEXT: v_add_u32_e32 v0, vcc, 3, v0
; VI-NEXT: s_waitcnt vmcnt(0)
; VI-NEXT: v_or_b32_sdwa v0, v1, v0 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:BYTE_0
-; VI-NEXT: v_add_u32_e32 v1, vcc, 3, v44
+; VI-NEXT: v_add_u32_e32 v1, vcc, 3, v47
; VI-NEXT: v_add_u32_e32 v0, vcc, 0x300, v0
; VI-NEXT: v_or_b32_sdwa v1, v61, v1 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:BYTE_0
; VI-NEXT: v_or_b32_sdwa v0, v1, v0 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:WORD_0
; VI-NEXT: v_add_u32_e32 v26, vcc, 0x3000000, v0
-; VI-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:632 ; 4-byte Folded Reload
-; VI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:624 ; 4-byte Folded Reload
+; VI-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:636 ; 4-byte Folded Reload
+; VI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:628 ; 4-byte Folded Reload
; VI-NEXT: s_waitcnt vmcnt(1)
; VI-NEXT: v_add_u32_e32 v0, vcc, 3, v0
; VI-NEXT: s_waitcnt vmcnt(0)
; VI-NEXT: v_or_b32_sdwa v0, v1, v0 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:BYTE_0
-; VI-NEXT: v_add_u32_e32 v1, vcc, 3, v58
+; VI-NEXT: v_add_u32_e32 v1, vcc, 3, v42
; VI-NEXT: v_add_u32_e32 v0, vcc, 0x300, v0
; VI-NEXT: v_or_b32_sdwa v1, v2, v1 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:BYTE_0
; VI-NEXT: v_or_b32_sdwa v0, v1, v0 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:WORD_0
; VI-NEXT: v_add_u32_e32 v27, vcc, 0x3000000, v0
-; VI-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:628 ; 4-byte Folded Reload
+; VI-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:632 ; 4-byte Folded Reload
+; VI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:556 ; 4-byte Folded Reload
; VI-NEXT: v_add_u32_e32 v1, vcc, 3, v63
-; VI-NEXT: v_or_b32_sdwa v1, v45, v1 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:BYTE_0
-; VI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:616 ; 4-byte Folded Reload
; VI-NEXT: s_waitcnt vmcnt(1)
; VI-NEXT: v_add_u32_e32 v0, vcc, 3, v0
; VI-NEXT: v_or_b32_sdwa v0, v48, v0 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:BYTE_0
; VI-NEXT: v_add_u32_e32 v0, vcc, 0x300, v0
+; VI-NEXT: s_waitcnt vmcnt(0)
+; VI-NEXT: v_or_b32_sdwa v1, v2, v1 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:BYTE_0
; VI-NEXT: v_or_b32_sdwa v0, v1, v0 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:WORD_0
; VI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:404 ; 4-byte Folded Reload
+; VI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:620 ; 4-byte Folded Reload
; VI-NEXT: v_add_u32_e32 v28, vcc, 0x3000000, v0
; VI-NEXT: v_add_u32_e32 v0, vcc, 3, v41
-; VI-NEXT: s_waitcnt vmcnt(0)
+; VI-NEXT: s_waitcnt vmcnt(1)
; VI-NEXT: v_or_b32_sdwa v0, v1, v0 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:BYTE_0
; VI-NEXT: v_add_u32_e32 v1, vcc, 3, v40
; VI-NEXT: v_add_u32_e32 v0, vcc, 0x300, v0
+; VI-NEXT: s_waitcnt vmcnt(0)
; VI-NEXT: v_or_b32_sdwa v1, v2, v1 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:BYTE_0
; VI-NEXT: v_or_b32_sdwa v0, v1, v0 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:WORD_0
; VI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:400 ; 4-byte Folded Reload
@@ -59614,7 +59630,7 @@ define inreg <32 x float> @bitcast_v128i8_to_v32f32_scalar(<128 x i8> inreg %a,
; GFX9-NEXT: buffer_load_ushort v31, off, s[0:3], s32 offset:164
; GFX9-NEXT: buffer_load_ushort v23, off, s[0:3], s32 offset:172
; GFX9-NEXT: buffer_load_ushort v50, off, s[0:3], s32 offset:180
-; GFX9-NEXT: buffer_load_ushort v44, off, s[0:3], s32 offset:188
+; GFX9-NEXT: buffer_load_ushort v40, off, s[0:3], s32 offset:188
; GFX9-NEXT: buffer_load_ushort v37, off, s[0:3], s32 offset:196
; GFX9-NEXT: buffer_load_ushort v52, off, s[0:3], s32 offset:204
; GFX9-NEXT: buffer_load_ushort v29, off, s[0:3], s32 offset:212
@@ -59622,11 +59638,11 @@ define inreg <32 x float> @bitcast_v128i8_to_v32f32_scalar(<128 x i8> inreg %a,
; GFX9-NEXT: buffer_load_ushort v48, off, s[0:3], s32 offset:228
; GFX9-NEXT: buffer_load_ushort v39, off, s[0:3], s32 offset:236
; GFX9-NEXT: buffer_load_ushort v45, off, s[0:3], s32 offset:244
-; GFX9-NEXT: buffer_load_ushort v40, off, s[0:3], s32 offset:252
+; GFX9-NEXT: buffer_load_ushort v44, off, s[0:3], s32 offset:252
; GFX9-NEXT: buffer_load_ushort v55, off, s[0:3], s32 offset:260
; GFX9-NEXT: buffer_load_ushort v43, off, s[0:3], s32 offset:268
-; GFX9-NEXT: buffer_load_ushort v36, off, s[0:3], s32 offset:276
-; GFX9-NEXT: buffer_load_ushort v42, off, s[0:3], s32 offset:284
+; GFX9-NEXT: buffer_load_ushort v42, off, s[0:3], s32 offset:276
+; GFX9-NEXT: buffer_load_ushort v36, off, s[0:3], s32 offset:284
; GFX9-NEXT: buffer_load_ushort v41, off, s[0:3], s32 offset:292
; GFX9-NEXT: buffer_load_ushort v32, off, s[0:3], s32 offset:300
; GFX9-NEXT: buffer_load_ushort v35, off, s[0:3], s32 offset:308
@@ -59652,7 +59668,7 @@ define inreg <32 x float> @bitcast_v128i8_to_v32f32_scalar(<128 x i8> inreg %a,
; GFX9-NEXT: buffer_store_dword v29, off, s[0:3], s32 offset:668 ; 4-byte Folded Spill
; GFX9-NEXT: buffer_store_dword v52, off, s[0:3], s32 offset:672 ; 4-byte Folded Spill
; GFX9-NEXT: buffer_store_dword v37, off, s[0:3], s32 offset:676 ; 4-byte Folded Spill
-; GFX9-NEXT: buffer_store_dword v44, off, s[0:3], s32 offset:680 ; 4-byte Folded Spill
+; GFX9-NEXT: buffer_store_dword v40, off, s[0:3], s32 offset:680 ; 4-byte Folded Spill
; GFX9-NEXT: buffer_store_dword v50, off, s[0:3], s32 offset:684 ; 4-byte Folded Spill
; GFX9-NEXT: buffer_store_dword v23, off, s[0:3], s32 offset:688 ; 4-byte Folded Spill
; GFX9-NEXT: buffer_store_dword v31, off, s[0:3], s32 offset:692 ; 4-byte Folded Spill
@@ -59827,7 +59843,7 @@ define inreg <32 x float> @bitcast_v128i8_to_v32f32_scalar(<128 x i8> inreg %a,
; GFX9-NEXT: s_waitcnt vmcnt(1)
; GFX9-NEXT: v_or_b32_sdwa v0, v23, v0 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
; GFX9-NEXT: v_or_b32_sdwa v22, v0, v1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_0 src1_sel:DWORD
-; GFX9-NEXT: v_or_b32_sdwa v0, v44, v58 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; GFX9-NEXT: v_or_b32_sdwa v0, v40, v58 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
; GFX9-NEXT: v_or_b32_sdwa v1, v37, v57 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
; GFX9-NEXT: v_or_b32_sdwa v23, v0, v1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_0 src1_sel:DWORD
; GFX9-NEXT: s_waitcnt vmcnt(0)
@@ -59839,7 +59855,7 @@ define inreg <32 x float> @bitcast_v128i8_to_v32f32_scalar(<128 x i8> inreg %a,
; GFX9-NEXT: v_mov_b32_e32 v37, v57
; GFX9-NEXT: v_mov_b32_e32 v57, v60
; GFX9-NEXT: v_mov_b32_e32 v52, v56
-; GFX9-NEXT: buffer_load_dword v44, off, s[0:3], s32 offset:612 ; 4-byte Folded Reload
+; GFX9-NEXT: buffer_load_dword v40, off, s[0:3], s32 offset:612 ; 4-byte Folded Reload
; GFX9-NEXT: s_waitcnt vmcnt(2)
; GFX9-NEXT: v_mov_b32_e32 v34, v0
; GFX9-NEXT: v_or_b32_sdwa v0, v46, v0 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
@@ -59848,14 +59864,14 @@ define inreg <32 x float> @bitcast_v128i8_to_v32f32_scalar(<128 x i8> inreg %a,
; GFX9-NEXT: v_or_b32_sdwa v25, v0, v1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_0 src1_sel:DWORD
; GFX9-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:440 ; 4-byte Folded Reload
; GFX9-NEXT: s_waitcnt vmcnt(1)
-; GFX9-NEXT: v_or_b32_sdwa v1, v45, v44 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; GFX9-NEXT: v_or_b32_sdwa v1, v45, v40 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
; GFX9-NEXT: s_waitcnt vmcnt(0)
; GFX9-NEXT: v_or_b32_sdwa v0, v39, v0 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
; GFX9-NEXT: v_or_b32_sdwa v26, v0, v1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_0 src1_sel:DWORD
; GFX9-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:436 ; 4-byte Folded Reload
; GFX9-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:432 ; 4-byte Folded Reload
; GFX9-NEXT: s_waitcnt vmcnt(1)
-; GFX9-NEXT: v_or_b32_sdwa v0, v40, v0 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; GFX9-NEXT: v_or_b32_sdwa v0, v44, v0 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
; GFX9-NEXT: s_waitcnt vmcnt(0)
; GFX9-NEXT: v_or_b32_sdwa v1, v55, v1 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
; GFX9-NEXT: v_or_b32_sdwa v27, v0, v1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_0 src1_sel:DWORD
@@ -59865,12 +59881,12 @@ define inreg <32 x float> @bitcast_v128i8_to_v32f32_scalar(<128 x i8> inreg %a,
; GFX9-NEXT: v_mov_b32_e32 v51, v0
; GFX9-NEXT: v_or_b32_sdwa v0, v43, v0 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
; GFX9-NEXT: s_waitcnt vmcnt(0)
-; GFX9-NEXT: v_or_b32_sdwa v1, v36, v1 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; GFX9-NEXT: v_or_b32_sdwa v1, v42, v1 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
; GFX9-NEXT: v_or_b32_sdwa v28, v0, v1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_0 src1_sel:DWORD
; GFX9-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:420 ; 4-byte Folded Reload
; GFX9-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:416 ; 4-byte Folded Reload
; GFX9-NEXT: s_waitcnt vmcnt(1)
-; GFX9-NEXT: v_or_b32_sdwa v0, v42, v0 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; GFX9-NEXT: v_or_b32_sdwa v0, v36, v0 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
; GFX9-NEXT: s_waitcnt vmcnt(0)
; GFX9-NEXT: v_or_b32_sdwa v1, v41, v1 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
; GFX9-NEXT: v_or_b32_sdwa v29, v0, v1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_0 src1_sel:DWORD
@@ -59924,7 +59940,7 @@ define inreg <32 x float> @bitcast_v128i8_to_v32f32_scalar(<128 x i8> inreg %a,
; GFX9-NEXT: .LBB39_2:
; GFX9-NEXT: buffer_load_dword v34, off, s[0:3], s32 offset:616 ; 4-byte Folded Reload
; GFX9-NEXT: buffer_load_dword v51, off, s[0:3], s32 offset:620 ; 4-byte Folded Reload
-; GFX9-NEXT: buffer_load_dword v44, off, s[0:3], s32 offset:612 ; 4-byte Folded Reload
+; GFX9-NEXT: buffer_load_dword v40, off, s[0:3], s32 offset:612 ; 4-byte Folded Reload
; GFX9-NEXT: buffer_load_dword v50, off, s[0:3], s32 offset:608 ; 4-byte Folded Reload
; GFX9-NEXT: buffer_load_dword v37, off, s[0:3], s32 offset:604 ; 4-byte Folded Reload
; GFX9-NEXT: buffer_load_dword v47, off, s[0:3], s32 offset:596 ; 4-byte Folded Reload
@@ -60286,12 +60302,12 @@ define inreg <32 x float> @bitcast_v128i8_to_v32f32_scalar(<128 x i8> inreg %a,
; GFX9-NEXT: s_waitcnt vmcnt(1)
; GFX9-NEXT: v_or_b32_sdwa v0, v1, v0 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:BYTE_0
; GFX9-NEXT: v_add_u32_e32 v1, 3, v45
-; GFX9-NEXT: v_or_b32_sdwa v1, v44, v1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:BYTE_0
+; GFX9-NEXT: v_or_b32_sdwa v1, v40, v1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:BYTE_0
; GFX9-NEXT: v_add_u32_e32 v0, 0x300, v0
; GFX9-NEXT: v_add_u32_sdwa v1, v1, s4 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:DWORD
; GFX9-NEXT: v_or_b32_sdwa v26, v0, v1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_0 src1_sel:DWORD
; GFX9-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:436 ; 4-byte Folded Reload
-; GFX9-NEXT: v_add_u32_e32 v0, 3, v40
+; GFX9-NEXT: v_add_u32_e32 v0, 3, v44
; GFX9-NEXT: s_waitcnt vmcnt(0)
; GFX9-NEXT: v_or_b32_sdwa v0, v1, v0 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:BYTE_0
; GFX9-NEXT: v_add_u32_e32 v1, 3, v55
@@ -60301,7 +60317,7 @@ define inreg <32 x float> @bitcast_v128i8_to_v32f32_scalar(<128 x i8> inreg %a,
; GFX9-NEXT: v_add_u32_sdwa v1, v1, s4 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:DWORD
; GFX9-NEXT: v_or_b32_sdwa v27, v0, v1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_0 src1_sel:DWORD
; GFX9-NEXT: v_add_u32_e32 v0, 3, v43
-; GFX9-NEXT: v_add_u32_e32 v1, 3, v36
+; GFX9-NEXT: v_add_u32_e32 v1, 3, v42
; GFX9-NEXT: v_or_b32_sdwa v0, v51, v0 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:BYTE_0
; GFX9-NEXT: v_add_u32_e32 v0, 0x300, v0
; GFX9-NEXT: s_waitcnt vmcnt(0)
@@ -60310,7 +60326,7 @@ define inreg <32 x float> @bitcast_v128i8_to_v32f32_scalar(<128 x i8> inreg %a,
; GFX9-NEXT: v_or_b32_sdwa v28, v0, v1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_0 src1_sel:DWORD
; GFX9-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:420 ; 4-byte Folded Reload
; GFX9-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:416 ; 4-byte Folded Reload
-; GFX9-NEXT: v_add_u32_e32 v0, 3, v42
+; GFX9-NEXT: v_add_u32_e32 v0, 3, v36
; GFX9-NEXT: s_waitcnt vmcnt(1)
; GFX9-NEXT: v_or_b32_sdwa v0, v1, v0 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:BYTE_0
; GFX9-NEXT: v_add_u32_e32 v1, 3, v41
@@ -67130,10 +67146,10 @@ define inreg <32 x float> @bitcast_v64bf16_to_v32f32_scalar(<64 x bfloat> inreg
; SI-NEXT: v_mul_f32_e32 v34, 1.0, v36
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_mul_f32_e64 v0, 1.0, s17
-; SI-NEXT: v_mul_f32_e64 v35, 1.0, s18
+; SI-NEXT: v_mul_f32_e64 v33, 1.0, s18
; SI-NEXT: v_mul_f32_e64 v36, 1.0, s21
; SI-NEXT: v_mul_f32_e64 v42, 1.0, s20
-; SI-NEXT: v_mul_f32_e64 v33, 1.0, s22
+; SI-NEXT: v_mul_f32_e64 v35, 1.0, s22
; SI-NEXT: buffer_store_dword v30, off, s[0:3], s32 offset:196 ; 4-byte Folded Spill
; SI-NEXT: buffer_store_dword v31, off, s[0:3], s32 offset:200 ; 4-byte Folded Spill
; SI-NEXT: buffer_store_dword v29, off, s[0:3], s32 offset:204 ; 4-byte Folded Spill
@@ -67162,9 +67178,9 @@ define inreg <32 x float> @bitcast_v64bf16_to_v32f32_scalar(<64 x bfloat> inreg
; SI-NEXT: buffer_store_dword v9, off, s[0:3], s32 offset:296 ; 4-byte Folded Spill
; SI-NEXT: buffer_store_dword v3, off, s[0:3], s32 offset:300 ; 4-byte Folded Spill
; SI-NEXT: buffer_store_dword v0, off, s[0:3], s32 offset:304 ; 4-byte Folded Spill
-; SI-NEXT: buffer_store_dword v33, off, s[0:3], s32 offset:308 ; 4-byte Folded Spill
-; SI-NEXT: buffer_store_dword v35, off, s[0:3], s32 offset:312 ; 4-byte Folded Spill
-; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:316 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:308 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v33, off, s[0:3], s32 offset:312 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v35, off, s[0:3], s32 offset:316 ; 4-byte Folded Spill
; SI-NEXT: buffer_store_dword v36, off, s[0:3], s32 offset:320 ; 4-byte Folded Spill
; SI-NEXT: buffer_store_dword v42, off, s[0:3], s32 offset:324 ; 4-byte Folded Spill
; SI-NEXT: s_cbranch_scc0 .LBB43_4
@@ -67175,12 +67191,13 @@ define inreg <32 x float> @bitcast_v64bf16_to_v32f32_scalar(<64 x bfloat> inreg
; SI-NEXT: buffer_load_dword v7, off, s[0:3], s32 offset:192 ; 4-byte Folded Reload
; SI-NEXT: v_alignbit_b32 v5, v5, v8, 16
; SI-NEXT: buffer_load_dword v8, off, s[0:3], s32 offset:184 ; 4-byte Folded Reload
-; SI-NEXT: s_waitcnt expcnt(2)
-; SI-NEXT: v_lshrrev_b32_e32 v1, 16, v1
-; SI-NEXT: v_alignbit_b32 v1, v1, v35, 16
+; SI-NEXT: s_waitcnt expcnt(6)
+; SI-NEXT: v_lshrrev_b32_e32 v3, 16, v3
+; SI-NEXT: v_alignbit_b32 v3, v3, v35, 16
; SI-NEXT: v_lshrrev_b32_e32 v4, 16, v4
; SI-NEXT: v_alignbit_b32 v4, v4, v9, 16
; SI-NEXT: buffer_load_dword v9, off, s[0:3], s32 offset:180 ; 4-byte Folded Reload
+; SI-NEXT: s_waitcnt expcnt(5)
; SI-NEXT: v_lshrrev_b32_e32 v0, 16, v0
; SI-NEXT: v_mov_b32_e32 v59, v2
; SI-NEXT: v_alignbit_b32 v0, v0, v2, 16
@@ -67190,10 +67207,11 @@ define inreg <32 x float> @bitcast_v64bf16_to_v32f32_scalar(<64 x bfloat> inreg
; SI-NEXT: v_mov_b32_e32 v47, v10
; SI-NEXT: v_lshrrev_b32_e32 v10, 16, v10
; SI-NEXT: v_mov_b32_e32 v45, v12
-; SI-NEXT: v_lshrrev_b32_e32 v3, 16, v3
-; SI-NEXT: v_alignbit_b32 v3, v3, v33, 16
-; SI-NEXT: v_mov_b32_e32 v33, v14
+; SI-NEXT: v_mov_b32_e32 v44, v14
+; SI-NEXT: s_waitcnt expcnt(4)
+; SI-NEXT: v_lshrrev_b32_e32 v1, 16, v1
; SI-NEXT: v_lshrrev_b32_e32 v13, 16, v13
+; SI-NEXT: v_alignbit_b32 v1, v1, v33, 16
; SI-NEXT: v_mov_b32_e32 v62, v38
; SI-NEXT: v_lshrrev_b32_e32 v15, 16, v15
; SI-NEXT: v_lshrrev_b32_e32 v16, 16, v16
@@ -67233,20 +67251,20 @@ define inreg <32 x float> @bitcast_v64bf16_to_v32f32_scalar(<64 x bfloat> inreg
; SI-NEXT: v_alignbit_b32 v28, v28, v37, 16
; SI-NEXT: v_mov_b32_e32 v37, v34
; SI-NEXT: buffer_load_dword v32, off, s[0:3], s32 offset:148 ; 4-byte Folded Reload
-; SI-NEXT: s_waitcnt vmcnt(3)
+; SI-NEXT: s_waitcnt vmcnt(3) expcnt(2)
; SI-NEXT: v_mov_b32_e32 v35, v7
; SI-NEXT: v_lshrrev_b32_e32 v7, 16, v7
; SI-NEXT: s_waitcnt vmcnt(2)
; SI-NEXT: v_mov_b32_e32 v43, v8
; SI-NEXT: v_alignbit_b32 v7, v7, v8, 16
; SI-NEXT: buffer_load_dword v8, off, s[0:3], s32 offset:188 ; 4-byte Folded Reload
-; SI-NEXT: s_waitcnt vmcnt(2) expcnt(0)
-; SI-NEXT: v_mov_b32_e32 v42, v9
+; SI-NEXT: s_waitcnt vmcnt(2)
+; SI-NEXT: v_mov_b32_e32 v60, v9
; SI-NEXT: s_waitcnt vmcnt(1)
; SI-NEXT: v_lshrrev_b32_e32 v31, 16, v32
; SI-NEXT: v_alignbit_b32 v31, v31, v34, 16
-; SI-NEXT: s_waitcnt vmcnt(0)
-; SI-NEXT: v_mov_b32_e32 v60, v8
+; SI-NEXT: s_waitcnt vmcnt(0) expcnt(0)
+; SI-NEXT: v_mov_b32_e32 v42, v8
; SI-NEXT: v_lshrrev_b32_e32 v8, 16, v8
; SI-NEXT: v_alignbit_b32 v8, v8, v9, 16
; SI-NEXT: v_lshrrev_b32_e32 v9, 16, v11
@@ -67270,7 +67288,7 @@ define inreg <32 x float> @bitcast_v64bf16_to_v32f32_scalar(<64 x bfloat> inreg
; SI-NEXT: v_alignbit_b32 v12, v12, v14, 16
; SI-NEXT: buffer_load_dword v14, off, s[0:3], s32 offset:160 ; 4-byte Folded Reload
; SI-NEXT: s_waitcnt vmcnt(0)
-; SI-NEXT: v_mov_b32_e32 v44, v14
+; SI-NEXT: v_mov_b32_e32 v33, v14
; SI-NEXT: v_alignbit_b32 v13, v13, v14, 16
; SI-NEXT: buffer_load_dword v14, off, s[0:3], s32 offset:152 ; 4-byte Folded Reload
; SI-NEXT: s_waitcnt vmcnt(0)
@@ -67293,7 +67311,7 @@ define inreg <32 x float> @bitcast_v64bf16_to_v32f32_scalar(<64 x bfloat> inreg
; SI-NEXT: .LBB43_2: ; %cmp.true
; SI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:304 ; 4-byte Folded Reload
; SI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:312 ; 4-byte Folded Reload
-; SI-NEXT: buffer_load_dword v3, off, s[0:3], s32 offset:316 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v3, off, s[0:3], s32 offset:308 ; 4-byte Folded Reload
; SI-NEXT: buffer_load_dword v4, off, s[0:3], s32 offset:300 ; 4-byte Folded Reload
; SI-NEXT: v_and_b32_e32 v0, 0xffff0000, v59
; SI-NEXT: v_add_f32_e32 v0, 0x40c00000, v0
@@ -67309,7 +67327,7 @@ define inreg <32 x float> @bitcast_v64bf16_to_v32f32_scalar(<64 x bfloat> inreg
; SI-NEXT: v_and_b32_e32 v8, 0xffff0000, v35
; SI-NEXT: v_add_f32_e32 v8, 0x40c00000, v8
; SI-NEXT: v_lshrrev_b32_e32 v8, 16, v8
-; SI-NEXT: v_and_b32_e32 v9, 0xffff0000, v60
+; SI-NEXT: v_and_b32_e32 v9, 0xffff0000, v42
; SI-NEXT: v_add_f32_e32 v9, 0x40c00000, v9
; SI-NEXT: v_lshrrev_b32_e32 v9, 16, v9
; SI-NEXT: v_and_b32_e32 v10, 0xffff0000, v57
@@ -67321,7 +67339,7 @@ define inreg <32 x float> @bitcast_v64bf16_to_v32f32_scalar(<64 x bfloat> inreg
; SI-NEXT: v_and_b32_e32 v12, 0xffff0000, v45
; SI-NEXT: v_add_f32_e32 v12, 0x40c00000, v12
; SI-NEXT: v_lshrrev_b32_e32 v12, 16, v12
-; SI-NEXT: v_and_b32_e32 v13, 0xffff0000, v33
+; SI-NEXT: v_and_b32_e32 v13, 0xffff0000, v44
; SI-NEXT: v_add_f32_e32 v13, 0x40c00000, v13
; SI-NEXT: v_lshrrev_b32_e32 v13, 16, v13
; SI-NEXT: v_and_b32_e32 v15, 0xffff0000, v36
@@ -67433,7 +67451,7 @@ define inreg <32 x float> @bitcast_v64bf16_to_v32f32_scalar(<64 x bfloat> inreg
; SI-NEXT: v_add_f32_e32 v2, 0x40c00000, v2
; SI-NEXT: v_lshrrev_b32_e32 v3, 16, v3
; SI-NEXT: v_alignbit_b32 v2, v3, v2, 16
-; SI-NEXT: buffer_load_dword v3, off, s[0:3], s32 offset:308 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v3, off, s[0:3], s32 offset:316 ; 4-byte Folded Reload
; SI-NEXT: s_waitcnt vmcnt(0)
; SI-NEXT: v_and_b32_e32 v3, 0xffff0000, v3
; SI-NEXT: v_add_f32_e32 v3, 0x40c00000, v3
@@ -67456,7 +67474,7 @@ define inreg <32 x float> @bitcast_v64bf16_to_v32f32_scalar(<64 x bfloat> inreg
; SI-NEXT: v_and_b32_e32 v7, 0xffff0000, v43
; SI-NEXT: v_add_f32_e32 v7, 0x40c00000, v7
; SI-NEXT: v_alignbit_b32 v7, v8, v7, 16
-; SI-NEXT: v_and_b32_e32 v8, 0xffff0000, v42
+; SI-NEXT: v_and_b32_e32 v8, 0xffff0000, v60
; SI-NEXT: v_add_f32_e32 v8, 0x40c00000, v8
; SI-NEXT: v_alignbit_b32 v8, v9, v8, 16
; SI-NEXT: v_and_b32_e32 v9, 0xffff0000, v58
@@ -67471,7 +67489,7 @@ define inreg <32 x float> @bitcast_v64bf16_to_v32f32_scalar(<64 x bfloat> inreg
; SI-NEXT: v_and_b32_e32 v12, 0xffff0000, v63
; SI-NEXT: v_add_f32_e32 v12, 0x40c00000, v12
; SI-NEXT: v_alignbit_b32 v12, v13, v12, 16
-; SI-NEXT: v_and_b32_e32 v13, 0xffff0000, v44
+; SI-NEXT: v_and_b32_e32 v13, 0xffff0000, v33
; SI-NEXT: v_add_f32_e32 v13, 0x40c00000, v13
; SI-NEXT: v_alignbit_b32 v13, v14, v13, 16
; SI-NEXT: v_and_b32_e32 v14, 0xffff0000, v62
@@ -67557,16 +67575,16 @@ define inreg <32 x float> @bitcast_v64bf16_to_v32f32_scalar(<64 x bfloat> inreg
; SI-NEXT: s_setpc_b64 s[30:31]
; SI-NEXT: .LBB43_4:
; SI-NEXT: buffer_load_dword v43, off, s[0:3], s32 offset:184 ; 4-byte Folded Reload
-; SI-NEXT: s_waitcnt expcnt(3)
+; SI-NEXT: s_waitcnt expcnt(2)
; SI-NEXT: buffer_load_dword v35, off, s[0:3], s32 offset:192 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v60, off, s[0:3], s32 offset:180 ; 4-byte Folded Reload
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: buffer_load_dword v42, off, s[0:3], s32 offset:180 ; 4-byte Folded Reload
-; SI-NEXT: buffer_load_dword v60, off, s[0:3], s32 offset:188 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v42, off, s[0:3], s32 offset:188 ; 4-byte Folded Reload
; SI-NEXT: buffer_load_dword v58, off, s[0:3], s32 offset:176 ; 4-byte Folded Reload
; SI-NEXT: buffer_load_dword v56, off, s[0:3], s32 offset:172 ; 4-byte Folded Reload
; SI-NEXT: buffer_load_dword v46, off, s[0:3], s32 offset:168 ; 4-byte Folded Reload
; SI-NEXT: buffer_load_dword v63, off, s[0:3], s32 offset:164 ; 4-byte Folded Reload
-; SI-NEXT: buffer_load_dword v44, off, s[0:3], s32 offset:160 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v33, off, s[0:3], s32 offset:160 ; 4-byte Folded Reload
; SI-NEXT: v_mov_b32_e32 v61, v53
; SI-NEXT: buffer_load_dword v36, off, s[0:3], s32 offset:152 ; 4-byte Folded Reload
; SI-NEXT: buffer_load_dword v53, off, s[0:3], s32 offset:156 ; 4-byte Folded Reload
@@ -67575,7 +67593,7 @@ define inreg <32 x float> @bitcast_v64bf16_to_v32f32_scalar(<64 x bfloat> inreg
; SI-NEXT: v_mov_b32_e32 v57, v11
; SI-NEXT: v_mov_b32_e32 v47, v10
; SI-NEXT: v_mov_b32_e32 v45, v12
-; SI-NEXT: v_mov_b32_e32 v33, v14
+; SI-NEXT: v_mov_b32_e32 v44, v14
; SI-NEXT: v_mov_b32_e32 v62, v38
; SI-NEXT: v_mov_b32_e32 v38, v39
; SI-NEXT: v_mov_b32_e32 v39, v41
@@ -70641,36 +70659,96 @@ define <64 x half> @bitcast_v32f32_to_v64f16(<32 x float> %a, i32 %b) {
; SI-NEXT: buffer_load_dword v31, off, s[0:3], s32 offset:8
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: buffer_load_dword v63, off, s[0:3], s32
+; SI-NEXT: ; implicit-def: $vgpr32
+; SI-NEXT: ; kill: killed $vgpr32
+; SI-NEXT: ; implicit-def: $vgpr32
+; SI-NEXT: ; kill: killed $vgpr32
+; SI-NEXT: ; implicit-def: $vgpr32
+; SI-NEXT: ; kill: killed $vgpr32
+; SI-NEXT: ; implicit-def: $vgpr32
+; SI-NEXT: ; kill: killed $vgpr32
+; SI-NEXT: ; implicit-def: $vgpr32
+; SI-NEXT: ; kill: killed $vgpr32
+; SI-NEXT: ; implicit-def: $vgpr32
+; SI-NEXT: ; kill: killed $vgpr32
+; SI-NEXT: ; implicit-def: $vgpr32
+; SI-NEXT: ; kill: killed $vgpr32
+; SI-NEXT: ; implicit-def: $vgpr32
+; SI-NEXT: ; kill: killed $vgpr32
+; SI-NEXT: ; implicit-def: $vgpr32
+; SI-NEXT: ; kill: killed $vgpr32
+; SI-NEXT: ; implicit-def: $vgpr32
+; SI-NEXT: ; kill: killed $vgpr32
+; SI-NEXT: ; implicit-def: $vgpr32
+; SI-NEXT: ; kill: killed $vgpr32
+; SI-NEXT: ; implicit-def: $vgpr32
+; SI-NEXT: ; kill: killed $vgpr32
+; SI-NEXT: ; implicit-def: $vgpr32
+; SI-NEXT: ; kill: killed $vgpr32
+; SI-NEXT: ; implicit-def: $vgpr32
+; SI-NEXT: ; kill: killed $vgpr32
+; SI-NEXT: ; implicit-def: $vgpr32
+; SI-NEXT: ; kill: killed $vgpr32
+; SI-NEXT: ; implicit-def: $vgpr32
+; SI-NEXT: ; kill: killed $vgpr32
+; SI-NEXT: ; implicit-def: $vgpr32
+; SI-NEXT: ; kill: killed $vgpr32
+; SI-NEXT: ; implicit-def: $vgpr32
+; SI-NEXT: ; kill: killed $vgpr32
+; SI-NEXT: ; implicit-def: $vgpr32
+; SI-NEXT: ; kill: killed $vgpr32
+; SI-NEXT: ; implicit-def: $vgpr32
+; SI-NEXT: ; kill: killed $vgpr32
+; SI-NEXT: ; implicit-def: $vgpr32
+; SI-NEXT: ; kill: killed $vgpr32
+; SI-NEXT: ; implicit-def: $vgpr32
+; SI-NEXT: ; kill: killed $vgpr32
+; SI-NEXT: ; implicit-def: $vgpr32
+; SI-NEXT: ; kill: killed $vgpr32
+; SI-NEXT: ; implicit-def: $vgpr32
+; SI-NEXT: ; kill: killed $vgpr32
+; SI-NEXT: ; implicit-def: $vgpr32
+; SI-NEXT: ; kill: killed $vgpr32
+; SI-NEXT: ; implicit-def: $vgpr32
+; SI-NEXT: ; kill: killed $vgpr32
+; SI-NEXT: ; implicit-def: $vgpr32
+; SI-NEXT: ; kill: killed $vgpr32
+; SI-NEXT: ; implicit-def: $vgpr32
+; SI-NEXT: ; kill: killed $vgpr32
+; SI-NEXT: ; implicit-def: $vgpr32
+; SI-NEXT: ; kill: killed $vgpr32
+; SI-NEXT: ; implicit-def: $vgpr32
; SI-NEXT: ; implicit-def: $vgpr60
; SI-NEXT: ; implicit-def: $vgpr61
; SI-NEXT: ; implicit-def: $vgpr58
; SI-NEXT: ; implicit-def: $vgpr59
-; SI-NEXT: ; implicit-def: $vgpr43
+; SI-NEXT: ; implicit-def: $vgpr56
; SI-NEXT: ; implicit-def: $vgpr57
-; SI-NEXT: ; implicit-def: $vgpr42
+; SI-NEXT: ; implicit-def: $vgpr46
; SI-NEXT: ; implicit-def: $vgpr47
-; SI-NEXT: ; implicit-def: $vgpr44
+; SI-NEXT: ; implicit-def: $vgpr40
; SI-NEXT: ; implicit-def: $vgpr45
-; SI-NEXT: ; implicit-def: $vgpr38
+; SI-NEXT: ; implicit-def: $vgpr42
+; SI-NEXT: ; implicit-def: $vgpr43
; SI-NEXT: ; implicit-def: $vgpr41
-; SI-NEXT: ; implicit-def: $vgpr55
-; SI-NEXT: ; implicit-def: $vgpr53
+; SI-NEXT: ; implicit-def: $vgpr38
; SI-NEXT: ; implicit-def: $vgpr51
; SI-NEXT: ; implicit-def: $vgpr49
; SI-NEXT: ; implicit-def: $vgpr39
; SI-NEXT: ; implicit-def: $vgpr37
; SI-NEXT: ; implicit-def: $vgpr35
; SI-NEXT: ; implicit-def: $vgpr33
-; SI-NEXT: ; implicit-def: $vgpr46
-; SI-NEXT: ; implicit-def: $vgpr56
-; SI-NEXT: ; implicit-def: $vgpr50
+; SI-NEXT: ; implicit-def: $vgpr44
+; SI-NEXT: ; kill: killed $vgpr32
; SI-NEXT: ; implicit-def: $vgpr52
+; SI-NEXT: ; implicit-def: $vgpr55
; SI-NEXT: ; implicit-def: $vgpr54
+; SI-NEXT: ; implicit-def: $vgpr50
+; SI-NEXT: ; implicit-def: $vgpr53
; SI-NEXT: ; implicit-def: $vgpr48
-; SI-NEXT: ; implicit-def: $vgpr40
; SI-NEXT: ; implicit-def: $vgpr36
-; SI-NEXT: ; implicit-def: $vgpr34
; SI-NEXT: ; implicit-def: $vgpr32
+; SI-NEXT: ; implicit-def: $vgpr34
; SI-NEXT: s_waitcnt vmcnt(1)
; SI-NEXT: v_cmp_ne_u32_e32 vcc, 0, v31
; SI-NEXT: ; implicit-def: $vgpr31
@@ -70680,173 +70758,99 @@ define <64 x half> @bitcast_v32f32_to_v64f16(<32 x float> %a, i32 %b) {
; SI-NEXT: ; implicit-def: $vgpr31
; SI-NEXT: ; kill: killed $vgpr31
; SI-NEXT: ; implicit-def: $vgpr31
-; SI-NEXT: ; kill: killed $vgpr31
-; SI-NEXT: ; implicit-def: $vgpr31
-; SI-NEXT: ; kill: killed $vgpr31
-; SI-NEXT: ; implicit-def: $vgpr31
-; SI-NEXT: ; kill: killed $vgpr31
-; SI-NEXT: ; implicit-def: $vgpr31
-; SI-NEXT: ; kill: killed $vgpr31
-; SI-NEXT: ; implicit-def: $vgpr31
-; SI-NEXT: ; kill: killed $vgpr31
-; SI-NEXT: ; implicit-def: $vgpr31
-; SI-NEXT: ; kill: killed $vgpr31
-; SI-NEXT: ; implicit-def: $vgpr31
-; SI-NEXT: ; kill: killed $vgpr31
-; SI-NEXT: ; implicit-def: $vgpr31
-; SI-NEXT: ; kill: killed $vgpr31
-; SI-NEXT: ; implicit-def: $vgpr31
-; SI-NEXT: ; kill: killed $vgpr31
-; SI-NEXT: ; implicit-def: $vgpr31
-; SI-NEXT: ; kill: killed $vgpr31
-; SI-NEXT: ; implicit-def: $vgpr31
-; SI-NEXT: ; kill: killed $vgpr31
-; SI-NEXT: ; implicit-def: $vgpr31
-; SI-NEXT: ; kill: killed $vgpr31
-; SI-NEXT: ; implicit-def: $vgpr31
-; SI-NEXT: ; kill: killed $vgpr31
-; SI-NEXT: ; implicit-def: $vgpr31
-; SI-NEXT: ; kill: killed $vgpr31
-; SI-NEXT: ; implicit-def: $vgpr31
-; SI-NEXT: ; kill: killed $vgpr31
-; SI-NEXT: ; implicit-def: $vgpr31
-; SI-NEXT: ; kill: killed $vgpr31
-; SI-NEXT: ; implicit-def: $vgpr31
-; SI-NEXT: ; kill: killed $vgpr31
-; SI-NEXT: ; implicit-def: $vgpr31
-; SI-NEXT: ; kill: killed $vgpr31
-; SI-NEXT: ; implicit-def: $vgpr31
-; SI-NEXT: ; kill: killed $vgpr31
-; SI-NEXT: ; implicit-def: $vgpr31
-; SI-NEXT: ; kill: killed $vgpr31
-; SI-NEXT: ; implicit-def: $vgpr31
-; SI-NEXT: ; kill: killed $vgpr31
-; SI-NEXT: ; implicit-def: $vgpr31
-; SI-NEXT: ; kill: killed $vgpr31
-; SI-NEXT: ; implicit-def: $vgpr31
-; SI-NEXT: ; kill: killed $vgpr31
-; SI-NEXT: ; implicit-def: $vgpr31
-; SI-NEXT: ; kill: killed $vgpr31
-; SI-NEXT: ; implicit-def: $vgpr31
-; SI-NEXT: ; kill: killed $vgpr31
-; SI-NEXT: ; implicit-def: $vgpr31
-; SI-NEXT: ; kill: killed $vgpr31
-; SI-NEXT: ; implicit-def: $vgpr31
-; SI-NEXT: ; kill: killed $vgpr31
-; SI-NEXT: ; implicit-def: $vgpr31
-; SI-NEXT: ; kill: killed $vgpr31
-; SI-NEXT: ; implicit-def: $vgpr31
-; SI-NEXT: ; kill: killed $vgpr31
-; SI-NEXT: ; implicit-def: $vgpr31
-; SI-NEXT: ; kill: killed $vgpr31
-; SI-NEXT: ; implicit-def: $vgpr31
; SI-NEXT: s_and_saveexec_b64 s[4:5], vcc
; SI-NEXT: s_xor_b64 s[4:5], exec, s[4:5]
; SI-NEXT: s_cbranch_execz .LBB44_2
; SI-NEXT: ; %bb.1: ; %cmp.false
; SI-NEXT: v_lshrrev_b32_e32 v31, 16, v62
-; SI-NEXT: v_cvt_f32_f16_e32 v32, v31
+; SI-NEXT: v_cvt_f32_f16_e32 v34, v31
; SI-NEXT: s_waitcnt vmcnt(0)
; SI-NEXT: v_lshrrev_b32_e32 v31, 16, v63
-; SI-NEXT: v_cvt_f32_f16_e32 v34, v31
+; SI-NEXT: v_cvt_f32_f16_e32 v36, v31
; SI-NEXT: v_lshrrev_b32_e32 v31, 16, v30
-; SI-NEXT: v_cvt_f32_f16_e32 v40, v31
+; SI-NEXT: v_cvt_f32_f16_e32 v53, v31
; SI-NEXT: v_lshrrev_b32_e32 v31, 16, v29
; SI-NEXT: v_cvt_f32_f16_e32 v54, v31
; SI-NEXT: v_lshrrev_b32_e32 v31, 16, v28
-; SI-NEXT: v_cvt_f32_f16_e32 v50, v31
+; SI-NEXT: v_cvt_f32_f16_e32 v52, v31
; SI-NEXT: v_lshrrev_b32_e32 v31, 16, v27
-; SI-NEXT: v_cvt_f32_f16_e32 v46, v31
+; SI-NEXT: v_cvt_f32_f16_e32 v44, v31
; SI-NEXT: v_lshrrev_b32_e32 v31, 16, v26
; SI-NEXT: v_cvt_f32_f16_e32 v31, v31
-; SI-NEXT: v_cvt_f32_f16_e32 v27, v27
-; SI-NEXT: v_cvt_f32_f16_e32 v26, v26
-; SI-NEXT: v_cvt_f32_f16_e32 v36, v63
-; SI-NEXT: buffer_store_dword v31, off, s[0:3], s32 offset:76 ; 4-byte Folded Spill
+; SI-NEXT: v_lshrrev_b32_e32 v32, 16, v8
+; SI-NEXT: v_cvt_f32_f16_e32 v38, v32
+; SI-NEXT: v_lshrrev_b32_e32 v32, 16, v7
+; SI-NEXT: buffer_store_dword v31, off, s[0:3], s32 offset:80 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_lshrrev_b32_e32 v31, 16, v25
; SI-NEXT: v_cvt_f32_f16_e32 v31, v31
-; SI-NEXT: v_cvt_f32_f16_e32 v25, v25
-; SI-NEXT: v_cvt_f32_f16_e32 v48, v30
-; SI-NEXT: v_cvt_f32_f16_e32 v52, v29
-; SI-NEXT: buffer_store_dword v31, off, s[0:3], s32 offset:84 ; 4-byte Folded Spill
+; SI-NEXT: v_cvt_f32_f16_e32 v41, v32
+; SI-NEXT: v_lshrrev_b32_e32 v32, 16, v6
+; SI-NEXT: v_cvt_f32_f16_e32 v43, v32
+; SI-NEXT: buffer_store_dword v31, off, s[0:3], s32 offset:88 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_lshrrev_b32_e32 v31, 16, v24
; SI-NEXT: v_cvt_f32_f16_e32 v31, v31
-; SI-NEXT: v_cvt_f32_f16_e32 v24, v24
-; SI-NEXT: v_cvt_f32_f16_e32 v56, v28
-; SI-NEXT: buffer_store_dword v27, off, s[0:3], s32 offset:80 ; 4-byte Folded Spill
-; SI-NEXT: buffer_store_dword v31, off, s[0:3], s32 offset:92 ; 4-byte Folded Spill
+; SI-NEXT: v_lshrrev_b32_e32 v32, 16, v5
+; SI-NEXT: v_cvt_f32_f16_e32 v45, v32
+; SI-NEXT: v_lshrrev_b32_e32 v32, 16, v4
+; SI-NEXT: buffer_store_dword v31, off, s[0:3], s32 offset:96 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_lshrrev_b32_e32 v31, 16, v23
; SI-NEXT: v_cvt_f32_f16_e32 v31, v31
-; SI-NEXT: v_cvt_f32_f16_e32 v23, v23
-; SI-NEXT: buffer_store_dword v26, off, s[0:3], s32 offset:88 ; 4-byte Folded Spill
-; SI-NEXT: buffer_store_dword v25, off, s[0:3], s32 offset:96 ; 4-byte Folded Spill
-; SI-NEXT: buffer_store_dword v31, off, s[0:3], s32 offset:100 ; 4-byte Folded Spill
+; SI-NEXT: v_cvt_f32_f16_e32 v28, v28
+; SI-NEXT: v_cvt_f32_f16_e32 v27, v27
+; SI-NEXT: v_cvt_f32_f16_e32 v26, v26
+; SI-NEXT: buffer_store_dword v31, off, s[0:3], s32 offset:104 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_lshrrev_b32_e32 v31, 16, v22
; SI-NEXT: v_cvt_f32_f16_e32 v31, v31
-; SI-NEXT: v_cvt_f32_f16_e32 v22, v22
-; SI-NEXT: buffer_store_dword v24, off, s[0:3], s32 offset:104 ; 4-byte Folded Spill
-; SI-NEXT: buffer_store_dword v23, off, s[0:3], s32 offset:108 ; 4-byte Folded Spill
-; SI-NEXT: buffer_store_dword v31, off, s[0:3], s32 offset:112 ; 4-byte Folded Spill
+; SI-NEXT: v_cvt_f32_f16_e32 v25, v25
+; SI-NEXT: v_cvt_f32_f16_e32 v24, v24
+; SI-NEXT: v_cvt_f32_f16_e32 v23, v23
+; SI-NEXT: buffer_store_dword v31, off, s[0:3], s32 offset:116 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_lshrrev_b32_e32 v31, 16, v21
; SI-NEXT: v_cvt_f32_f16_e32 v31, v31
+; SI-NEXT: v_cvt_f32_f16_e32 v22, v22
; SI-NEXT: v_cvt_f32_f16_e32 v21, v21
-; SI-NEXT: buffer_store_dword v22, off, s[0:3], s32 offset:116 ; 4-byte Folded Spill
-; SI-NEXT: v_cvt_f32_f16_e32 v44, v5
-; SI-NEXT: buffer_store_dword v31, off, s[0:3], s32 offset:120 ; 4-byte Folded Spill
+; SI-NEXT: v_cvt_f32_f16_e32 v8, v8
+; SI-NEXT: buffer_store_dword v31, off, s[0:3], s32 offset:124 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_lshrrev_b32_e32 v31, 16, v20
; SI-NEXT: v_cvt_f32_f16_e32 v31, v31
; SI-NEXT: v_cvt_f32_f16_e32 v20, v20
-; SI-NEXT: buffer_store_dword v21, off, s[0:3], s32 offset:124 ; 4-byte Folded Spill
-; SI-NEXT: v_cvt_f32_f16_e32 v42, v4
-; SI-NEXT: buffer_store_dword v31, off, s[0:3], s32 offset:128 ; 4-byte Folded Spill
+; SI-NEXT: v_cvt_f32_f16_e32 v7, v7
+; SI-NEXT: v_cvt_f32_f16_e32 v47, v32
+; SI-NEXT: buffer_store_dword v31, off, s[0:3], s32 offset:132 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_lshrrev_b32_e32 v31, 16, v19
; SI-NEXT: v_cvt_f32_f16_e32 v31, v31
; SI-NEXT: v_cvt_f32_f16_e32 v19, v19
-; SI-NEXT: buffer_store_dword v20, off, s[0:3], s32 offset:132 ; 4-byte Folded Spill
-; SI-NEXT: v_cvt_f32_f16_e32 v43, v3
-; SI-NEXT: buffer_store_dword v31, off, s[0:3], s32 offset:136 ; 4-byte Folded Spill
+; SI-NEXT: v_lshrrev_b32_e32 v32, 16, v3
+; SI-NEXT: v_cvt_f32_f16_e32 v57, v32
+; SI-NEXT: buffer_store_dword v31, off, s[0:3], s32 offset:140 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_lshrrev_b32_e32 v31, 16, v18
; SI-NEXT: v_cvt_f32_f16_e32 v31, v31
; SI-NEXT: v_cvt_f32_f16_e32 v18, v18
-; SI-NEXT: buffer_store_dword v19, off, s[0:3], s32 offset:140 ; 4-byte Folded Spill
-; SI-NEXT: v_cvt_f32_f16_e32 v58, v2
-; SI-NEXT: buffer_store_dword v31, off, s[0:3], s32 offset:144 ; 4-byte Folded Spill
+; SI-NEXT: v_lshrrev_b32_e32 v32, 16, v2
+; SI-NEXT: v_cvt_f32_f16_e32 v59, v32
+; SI-NEXT: buffer_store_dword v31, off, s[0:3], s32 offset:148 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_lshrrev_b32_e32 v31, 16, v17
; SI-NEXT: v_cvt_f32_f16_e32 v31, v31
; SI-NEXT: v_cvt_f32_f16_e32 v17, v17
-; SI-NEXT: buffer_store_dword v18, off, s[0:3], s32 offset:148 ; 4-byte Folded Spill
-; SI-NEXT: v_cvt_f32_f16_e32 v60, v1
-; SI-NEXT: buffer_store_dword v31, off, s[0:3], s32 offset:152 ; 4-byte Folded Spill
+; SI-NEXT: v_lshrrev_b32_e32 v32, 16, v1
+; SI-NEXT: v_cvt_f32_f16_e32 v61, v32
+; SI-NEXT: buffer_store_dword v31, off, s[0:3], s32 offset:156 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_lshrrev_b32_e32 v31, 16, v16
; SI-NEXT: v_cvt_f32_f16_e32 v31, v31
; SI-NEXT: v_cvt_f32_f16_e32 v16, v16
-; SI-NEXT: buffer_store_dword v17, off, s[0:3], s32 offset:156 ; 4-byte Folded Spill
-; SI-NEXT: ; implicit-def: $vgpr17
-; SI-NEXT: ; implicit-def: $vgpr18
-; SI-NEXT: ; implicit-def: $vgpr19
-; SI-NEXT: ; implicit-def: $vgpr20
-; SI-NEXT: ; implicit-def: $vgpr21
-; SI-NEXT: ; implicit-def: $vgpr22
-; SI-NEXT: ; implicit-def: $vgpr23
-; SI-NEXT: ; implicit-def: $vgpr24
-; SI-NEXT: ; implicit-def: $vgpr25
-; SI-NEXT: ; implicit-def: $vgpr26
-; SI-NEXT: ; implicit-def: $vgpr27
-; SI-NEXT: ; implicit-def: $vgpr28
-; SI-NEXT: ; implicit-def: $vgpr29
-; SI-NEXT: ; implicit-def: $vgpr30
-; SI-NEXT: ; implicit-def: $vgpr63
-; SI-NEXT: buffer_store_dword v31, off, s[0:3], s32 offset:160 ; 4-byte Folded Spill
+; SI-NEXT: v_cvt_f32_f16_e32 v32, v62
+; SI-NEXT: v_cvt_f32_f16_e32 v48, v63
+; SI-NEXT: buffer_store_dword v31, off, s[0:3], s32 offset:164 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_lshrrev_b32_e32 v31, 16, v15
; SI-NEXT: v_cvt_f32_f16_e32 v33, v31
@@ -70861,16 +70865,6 @@ define <64 x half> @bitcast_v32f32_to_v64f16(<32 x float> %a, i32 %b) {
; SI-NEXT: v_lshrrev_b32_e32 v31, 16, v10
; SI-NEXT: v_cvt_f32_f16_e32 v51, v31
; SI-NEXT: v_lshrrev_b32_e32 v31, 16, v9
-; SI-NEXT: v_cvt_f32_f16_e32 v53, v31
-; SI-NEXT: v_lshrrev_b32_e32 v31, 16, v8
-; SI-NEXT: v_cvt_f32_f16_e32 v55, v31
-; SI-NEXT: v_lshrrev_b32_e32 v31, 16, v7
-; SI-NEXT: v_cvt_f32_f16_e32 v41, v31
-; SI-NEXT: v_lshrrev_b32_e32 v31, 16, v6
-; SI-NEXT: v_cvt_f32_f16_e32 v38, v31
-; SI-NEXT: v_lshrrev_b32_e32 v31, 16, v5
-; SI-NEXT: v_cvt_f32_f16_e32 v45, v31
-; SI-NEXT: v_lshrrev_b32_e32 v31, 16, v4
; SI-NEXT: v_cvt_f32_f16_e32 v15, v15
; SI-NEXT: v_cvt_f32_f16_e32 v14, v14
; SI-NEXT: v_cvt_f32_f16_e32 v13, v13
@@ -70878,28 +70872,37 @@ define <64 x half> @bitcast_v32f32_to_v64f16(<32 x float> %a, i32 %b) {
; SI-NEXT: v_cvt_f32_f16_e32 v11, v11
; SI-NEXT: v_cvt_f32_f16_e32 v10, v10
; SI-NEXT: v_cvt_f32_f16_e32 v9, v9
-; SI-NEXT: v_cvt_f32_f16_e32 v8, v8
-; SI-NEXT: v_cvt_f32_f16_e32 v7, v7
-; SI-NEXT: v_cvt_f32_f16_e32 v6, v6
-; SI-NEXT: v_cvt_f32_f16_e32 v47, v31
-; SI-NEXT: v_lshrrev_b32_e32 v31, 16, v3
-; SI-NEXT: v_cvt_f32_f16_e32 v57, v31
-; SI-NEXT: v_lshrrev_b32_e32 v31, 16, v2
-; SI-NEXT: v_cvt_f32_f16_e32 v59, v31
-; SI-NEXT: v_lshrrev_b32_e32 v31, 16, v1
-; SI-NEXT: v_cvt_f32_f16_e32 v61, v31
-; SI-NEXT: v_cvt_f32_f16_e32 v31, v62
-; SI-NEXT: buffer_store_dword v16, off, s[0:3], s32 offset:164 ; 4-byte Folded Spill
-; SI-NEXT: buffer_store_dword v15, off, s[0:3], s32 offset:168 ; 4-byte Folded Spill
-; SI-NEXT: buffer_store_dword v14, off, s[0:3], s32 offset:172 ; 4-byte Folded Spill
-; SI-NEXT: buffer_store_dword v13, off, s[0:3], s32 offset:176 ; 4-byte Folded Spill
-; SI-NEXT: buffer_store_dword v12, off, s[0:3], s32 offset:180 ; 4-byte Folded Spill
-; SI-NEXT: buffer_store_dword v11, off, s[0:3], s32 offset:184 ; 4-byte Folded Spill
-; SI-NEXT: buffer_store_dword v10, off, s[0:3], s32 offset:188 ; 4-byte Folded Spill
-; SI-NEXT: buffer_store_dword v9, off, s[0:3], s32 offset:192 ; 4-byte Folded Spill
-; SI-NEXT: buffer_store_dword v8, off, s[0:3], s32 offset:196 ; 4-byte Folded Spill
-; SI-NEXT: buffer_store_dword v7, off, s[0:3], s32 offset:200 ; 4-byte Folded Spill
-; SI-NEXT: buffer_store_dword v6, off, s[0:3], s32 offset:204 ; 4-byte Folded Spill
+; SI-NEXT: v_cvt_f32_f16_e32 v31, v31
+; SI-NEXT: v_cvt_f32_f16_e32 v50, v30
+; SI-NEXT: v_cvt_f32_f16_e32 v55, v29
+; SI-NEXT: buffer_store_dword v28, off, s[0:3], s32 offset:76 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v27, off, s[0:3], s32 offset:84 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v26, off, s[0:3], s32 offset:92 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v25, off, s[0:3], s32 offset:100 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v24, off, s[0:3], s32 offset:108 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v23, off, s[0:3], s32 offset:112 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v22, off, s[0:3], s32 offset:120 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v21, off, s[0:3], s32 offset:128 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v20, off, s[0:3], s32 offset:136 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v19, off, s[0:3], s32 offset:144 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v18, off, s[0:3], s32 offset:152 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v17, off, s[0:3], s32 offset:160 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v16, off, s[0:3], s32 offset:168 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v15, off, s[0:3], s32 offset:172 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v14, off, s[0:3], s32 offset:176 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v13, off, s[0:3], s32 offset:180 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v12, off, s[0:3], s32 offset:184 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v11, off, s[0:3], s32 offset:188 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v10, off, s[0:3], s32 offset:192 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v9, off, s[0:3], s32 offset:196 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v8, off, s[0:3], s32 offset:200 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v7, off, s[0:3], s32 offset:204 ; 4-byte Folded Spill
+; SI-NEXT: v_cvt_f32_f16_e32 v42, v6
+; SI-NEXT: v_cvt_f32_f16_e32 v40, v5
+; SI-NEXT: v_cvt_f32_f16_e32 v46, v4
+; SI-NEXT: v_cvt_f32_f16_e32 v56, v3
+; SI-NEXT: v_cvt_f32_f16_e32 v58, v2
+; SI-NEXT: v_cvt_f32_f16_e32 v60, v1
; SI-NEXT: ; implicit-def: $vgpr1
; SI-NEXT: ; implicit-def: $vgpr2
; SI-NEXT: ; implicit-def: $vgpr3
@@ -70916,7 +70919,22 @@ define <64 x half> @bitcast_v32f32_to_v64f16(<32 x float> %a, i32 %b) {
; SI-NEXT: ; implicit-def: $vgpr14
; SI-NEXT: ; implicit-def: $vgpr15
; SI-NEXT: ; implicit-def: $vgpr16
+; SI-NEXT: ; implicit-def: $vgpr17
+; SI-NEXT: ; implicit-def: $vgpr18
+; SI-NEXT: ; implicit-def: $vgpr19
+; SI-NEXT: ; implicit-def: $vgpr20
+; SI-NEXT: ; implicit-def: $vgpr21
+; SI-NEXT: ; implicit-def: $vgpr22
+; SI-NEXT: ; implicit-def: $vgpr23
+; SI-NEXT: ; implicit-def: $vgpr24
+; SI-NEXT: ; implicit-def: $vgpr25
+; SI-NEXT: ; implicit-def: $vgpr26
+; SI-NEXT: ; implicit-def: $vgpr27
+; SI-NEXT: ; implicit-def: $vgpr28
+; SI-NEXT: ; implicit-def: $vgpr29
+; SI-NEXT: ; implicit-def: $vgpr30
; SI-NEXT: ; implicit-def: $vgpr62
+; SI-NEXT: ; implicit-def: $vgpr63
; SI-NEXT: .LBB44_2: ; %Flow
; SI-NEXT: s_andn2_saveexec_b64 s[4:5], s[4:5]
; SI-NEXT: s_cbranch_execz .LBB44_4
@@ -70924,89 +70942,80 @@ define <64 x half> @bitcast_v32f32_to_v64f16(<32 x float> %a, i32 %b) {
; SI-NEXT: v_add_f32_e32 v5, 1.0, v5
; SI-NEXT: v_add_f32_e32 v33, 1.0, v1
; SI-NEXT: v_lshrrev_b32_e32 v1, 16, v5
-; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_add_f32_e32 v6, 1.0, v6
; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:208 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_lshrrev_b32_e32 v1, 16, v6
-; SI-NEXT: v_add_f32_e32 v28, 1.0, v28
+; SI-NEXT: v_add_f32_e32 v7, 1.0, v7
; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:212 ; 4-byte Folded Spill
-; SI-NEXT: v_add_f32_e32 v9, 1.0, v9
-; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_lshrrev_b32_e32 v1, 16, v28
-; SI-NEXT: v_lshrrev_b32_e32 v53, 16, v9
-; SI-NEXT: v_cvt_f32_f16_e32 v9, v9
-; SI-NEXT: v_cvt_f32_f16_e32 v1, v1
-; SI-NEXT: v_add_f32_e32 v25, 1.0, v25
-; SI-NEXT: v_add_f32_e32 v26, 1.0, v26
-; SI-NEXT: v_lshrrev_b32_e32 v48, 16, v25
-; SI-NEXT: v_lshrrev_b32_e32 v50, 16, v26
-; SI-NEXT: buffer_store_dword v9, off, s[0:3], s32 offset:192 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_cvt_f32_f16_e32 v9, v50
-; SI-NEXT: v_mov_b32_e32 v50, v1
-; SI-NEXT: v_cvt_f32_f16_e32 v1, v48
-; SI-NEXT: v_add_f32_e32 v24, 1.0, v24
-; SI-NEXT: v_lshrrev_b32_e32 v38, 16, v24
-; SI-NEXT: v_add_f32_e32 v23, 1.0, v23
-; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:84 ; 4-byte Folded Spill
-; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_cvt_f32_f16_e32 v1, v38
-; SI-NEXT: v_lshrrev_b32_e32 v56, 16, v23
-; SI-NEXT: v_add_f32_e32 v22, 1.0, v22
-; SI-NEXT: v_lshrrev_b32_e32 v34, 16, v22
-; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:92 ; 4-byte Folded Spill
-; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_cvt_f32_f16_e32 v1, v56
+; SI-NEXT: v_lshrrev_b32_e32 v1, 16, v7
; SI-NEXT: v_add_f32_e32 v21, 1.0, v21
-; SI-NEXT: v_lshrrev_b32_e32 v32, 16, v21
-; SI-NEXT: v_add_f32_e32 v20, 1.0, v20
-; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:100 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:216 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_cvt_f32_f16_e32 v1, v34
-; SI-NEXT: v_lshrrev_b32_e32 v46, 16, v20
+; SI-NEXT: v_lshrrev_b32_e32 v1, 16, v21
+; SI-NEXT: v_cvt_f32_f16_e32 v1, v1
+; SI-NEXT: v_add_f32_e32 v20, 1.0, v20
+; SI-NEXT: v_lshrrev_b32_e32 v31, 16, v20
; SI-NEXT: v_add_f32_e32 v19, 1.0, v19
-; SI-NEXT: v_lshrrev_b32_e32 v61, 16, v19
-; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:112 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:124 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_cvt_f32_f16_e32 v1, v32
+; SI-NEXT: v_cvt_f32_f16_e32 v1, v31
+; SI-NEXT: v_lshrrev_b32_e32 v61, 16, v19
; SI-NEXT: v_add_f32_e32 v18, 1.0, v18
; SI-NEXT: v_lshrrev_b32_e32 v59, 16, v18
-; SI-NEXT: v_add_f32_e32 v17, 1.0, v17
-; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:120 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:132 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_cvt_f32_f16_e32 v1, v46
+; SI-NEXT: v_cvt_f32_f16_e32 v1, v61
+; SI-NEXT: v_add_f32_e32 v17, 1.0, v17
; SI-NEXT: v_lshrrev_b32_e32 v57, 16, v17
; SI-NEXT: v_add_f32_e32 v16, 1.0, v16
+; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:140 ; 4-byte Folded Spill
+; SI-NEXT: s_waitcnt expcnt(0)
+; SI-NEXT: v_cvt_f32_f16_e32 v1, v59
; SI-NEXT: v_lshrrev_b32_e32 v47, 16, v16
-; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:128 ; 4-byte Folded Spill
+; SI-NEXT: v_add_f32_e32 v9, 1.0, v9
+; SI-NEXT: v_lshrrev_b32_e32 v41, 16, v9
+; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:148 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_cvt_f32_f16_e32 v1, v61
-; SI-NEXT: v_add_f32_e32 v7, 1.0, v7
+; SI-NEXT: v_cvt_f32_f16_e32 v1, v57
+; SI-NEXT: v_cvt_f32_f16_e32 v31, v41
+; SI-NEXT: v_cvt_f32_f16_e32 v9, v9
+; SI-NEXT: v_add_f32_e32 v26, 1.0, v26
+; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:156 ; 4-byte Folded Spill
+; SI-NEXT: s_waitcnt expcnt(0)
+; SI-NEXT: v_cvt_f32_f16_e32 v1, v47
+; SI-NEXT: v_lshrrev_b32_e32 v55, 16, v26
+; SI-NEXT: buffer_store_dword v9, off, s[0:3], s32 offset:196 ; 4-byte Folded Spill
+; SI-NEXT: s_waitcnt expcnt(0)
+; SI-NEXT: v_cvt_f32_f16_e32 v9, v55
+; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:164 ; 4-byte Folded Spill
+; SI-NEXT: s_waitcnt expcnt(0)
+; SI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:216 ; 4-byte Folded Reload
; SI-NEXT: v_add_f32_e32 v8, 1.0, v8
-; SI-NEXT: v_lshrrev_b32_e32 v41, 16, v7
-; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:136 ; 4-byte Folded Spill
+; SI-NEXT: v_add_f32_e32 v25, 1.0, v25
+; SI-NEXT: v_lshrrev_b32_e32 v43, 16, v8
+; SI-NEXT: v_lshrrev_b32_e32 v38, 16, v25
+; SI-NEXT: buffer_store_dword v9, off, s[0:3], s32 offset:80 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_cvt_f32_f16_e32 v1, v59
-; SI-NEXT: v_lshrrev_b32_e32 v55, 16, v8
-; SI-NEXT: v_cvt_f32_f16_e32 v8, v8
+; SI-NEXT: v_cvt_f32_f16_e32 v9, v38
+; SI-NEXT: v_cvt_f32_f16_e32 v38, v43
; SI-NEXT: v_cvt_f32_f16_e32 v7, v7
-; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:144 ; 4-byte Folded Spill
+; SI-NEXT: s_waitcnt vmcnt(12)
+; SI-NEXT: v_add_f32_e32 v40, 1.0, v63
+; SI-NEXT: v_add_f32_e32 v42, 1.0, v62
+; SI-NEXT: v_lshrrev_b32_e32 v63, 16, v40
+; SI-NEXT: v_lshrrev_b32_e32 v62, 16, v42
+; SI-NEXT: v_cvt_f32_f16_e32 v8, v8
+; SI-NEXT: v_cvt_f32_f16_e32 v32, v42
+; SI-NEXT: buffer_store_dword v7, off, s[0:3], s32 offset:204 ; 4-byte Folded Spill
+; SI-NEXT: v_cvt_f32_f16_e32 v42, v6
+; SI-NEXT: v_cvt_f32_f16_e32 v6, v62
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_cvt_f32_f16_e32 v1, v57
-; SI-NEXT: v_cvt_f32_f16_e32 v6, v6
+; SI-NEXT: v_cvt_f32_f16_e32 v7, v63
; SI-NEXT: v_add_f32_e32 v27, 1.0, v27
-; SI-NEXT: s_waitcnt vmcnt(11)
-; SI-NEXT: v_add_f32_e32 v42, 1.0, v63
-; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:152 ; 4-byte Folded Spill
-; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_cvt_f32_f16_e32 v1, v47
-; SI-NEXT: v_add_f32_e32 v44, 1.0, v62
; SI-NEXT: v_add_f32_e32 v2, 1.0, v2
; SI-NEXT: v_add_f32_e32 v3, 1.0, v3
-; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:160 ; 4-byte Folded Spill
-; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:212 ; 4-byte Folded Reload
; SI-NEXT: v_add_f32_e32 v4, 1.0, v4
; SI-NEXT: v_add_f32_e32 v10, 1.0, v10
; SI-NEXT: v_add_f32_e32 v11, 1.0, v11
@@ -71014,39 +71023,40 @@ define <64 x half> @bitcast_v32f32_to_v64f16(<32 x float> %a, i32 %b) {
; SI-NEXT: v_add_f32_e32 v13, 1.0, v13
; SI-NEXT: v_add_f32_e32 v14, 1.0, v14
; SI-NEXT: v_add_f32_e32 v15, 1.0, v15
-; SI-NEXT: v_lshrrev_b32_e32 v52, 16, v27
+; SI-NEXT: v_add_f32_e32 v22, 1.0, v22
+; SI-NEXT: v_add_f32_e32 v23, 1.0, v23
+; SI-NEXT: v_add_f32_e32 v24, 1.0, v24
+; SI-NEXT: v_lshrrev_b32_e32 v50, 16, v27
+; SI-NEXT: v_add_f32_e32 v28, 1.0, v28
; SI-NEXT: v_add_f32_e32 v29, 1.0, v29
; SI-NEXT: v_add_f32_e32 v30, 1.0, v30
-; SI-NEXT: v_lshrrev_b32_e32 v63, 16, v42
-; SI-NEXT: v_lshrrev_b32_e32 v62, 16, v44
; SI-NEXT: v_lshrrev_b32_e32 v51, 16, v10
; SI-NEXT: v_lshrrev_b32_e32 v49, 16, v11
; SI-NEXT: v_lshrrev_b32_e32 v39, 16, v12
; SI-NEXT: v_lshrrev_b32_e32 v37, 16, v13
; SI-NEXT: v_lshrrev_b32_e32 v35, 16, v14
; SI-NEXT: v_lshrrev_b32_e32 v45, 16, v15
+; SI-NEXT: v_lshrrev_b32_e32 v44, 16, v22
+; SI-NEXT: v_lshrrev_b32_e32 v34, 16, v23
+; SI-NEXT: v_lshrrev_b32_e32 v36, 16, v24
+; SI-NEXT: v_lshrrev_b32_e32 v52, 16, v28
; SI-NEXT: v_lshrrev_b32_e32 v54, 16, v29
-; SI-NEXT: v_lshrrev_b32_e32 v40, 16, v30
-; SI-NEXT: v_cvt_f32_f16_e32 v31, v44
-; SI-NEXT: v_cvt_f32_f16_e32 v36, v42
+; SI-NEXT: v_lshrrev_b32_e32 v53, 16, v30
+; SI-NEXT: v_cvt_f32_f16_e32 v48, v40
; SI-NEXT: v_cvt_f32_f16_e32 v30, v30
; SI-NEXT: v_cvt_f32_f16_e32 v29, v29
-; SI-NEXT: v_cvt_f32_f16_e32 v28, v28
-; SI-NEXT: buffer_store_dword v8, off, s[0:3], s32 offset:196 ; 4-byte Folded Spill
-; SI-NEXT: buffer_store_dword v7, off, s[0:3], s32 offset:200 ; 4-byte Folded Spill
-; SI-NEXT: buffer_store_dword v6, off, s[0:3], s32 offset:204 ; 4-byte Folded Spill
-; SI-NEXT: v_cvt_f32_f16_e32 v44, v5
+; SI-NEXT: buffer_store_dword v8, off, s[0:3], s32 offset:200 ; 4-byte Folded Spill
+; SI-NEXT: v_cvt_f32_f16_e32 v40, v5
; SI-NEXT: v_lshrrev_b32_e32 v5, 16, v4
-; SI-NEXT: v_cvt_f32_f16_e32 v42, v4
+; SI-NEXT: v_cvt_f32_f16_e32 v46, v4
; SI-NEXT: v_lshrrev_b32_e32 v4, 16, v3
-; SI-NEXT: v_cvt_f32_f16_e32 v43, v3
+; SI-NEXT: v_cvt_f32_f16_e32 v56, v3
; SI-NEXT: v_lshrrev_b32_e32 v3, 16, v2
; SI-NEXT: v_cvt_f32_f16_e32 v58, v2
; SI-NEXT: v_lshrrev_b32_e32 v2, 16, v33
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_cvt_f32_f16_e32 v6, v62
-; SI-NEXT: v_cvt_f32_f16_e32 v7, v63
-; SI-NEXT: v_cvt_f32_f16_e32 v8, v52
+; SI-NEXT: v_cvt_f32_f16_e32 v8, v50
+; SI-NEXT: v_cvt_f32_f16_e32 v28, v28
; SI-NEXT: v_cvt_f32_f16_e32 v27, v27
; SI-NEXT: v_cvt_f32_f16_e32 v26, v26
; SI-NEXT: v_cvt_f32_f16_e32 v25, v25
@@ -71065,51 +71075,59 @@ define <64 x half> @bitcast_v32f32_to_v64f16(<32 x float> %a, i32 %b) {
; SI-NEXT: v_cvt_f32_f16_e32 v12, v12
; SI-NEXT: v_cvt_f32_f16_e32 v11, v11
; SI-NEXT: v_cvt_f32_f16_e32 v10, v10
+; SI-NEXT: s_waitcnt vmcnt(3)
+; SI-NEXT: v_cvt_f32_f16_e32 v41, v1
+; SI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:212 ; 4-byte Folded Reload
; SI-NEXT: v_cvt_f32_f16_e32 v60, v33
-; SI-NEXT: v_cvt_f32_f16_e32 v40, v40
+; SI-NEXT: v_cvt_f32_f16_e32 v53, v53
; SI-NEXT: v_cvt_f32_f16_e32 v54, v54
+; SI-NEXT: v_cvt_f32_f16_e32 v52, v52
+; SI-NEXT: buffer_store_dword v9, off, s[0:3], s32 offset:88 ; 4-byte Folded Spill
+; SI-NEXT: s_waitcnt expcnt(0)
+; SI-NEXT: v_cvt_f32_f16_e32 v9, v36
+; SI-NEXT: v_mov_b32_e32 v36, v7
+; SI-NEXT: v_cvt_f32_f16_e32 v7, v34
+; SI-NEXT: v_mov_b32_e32 v34, v6
+; SI-NEXT: v_cvt_f32_f16_e32 v6, v44
; SI-NEXT: v_cvt_f32_f16_e32 v33, v45
; SI-NEXT: v_cvt_f32_f16_e32 v35, v35
; SI-NEXT: v_cvt_f32_f16_e32 v37, v37
; SI-NEXT: v_cvt_f32_f16_e32 v39, v39
; SI-NEXT: v_cvt_f32_f16_e32 v49, v49
; SI-NEXT: v_cvt_f32_f16_e32 v51, v51
-; SI-NEXT: v_cvt_f32_f16_e32 v53, v53
-; SI-NEXT: v_cvt_f32_f16_e32 v55, v55
-; SI-NEXT: v_cvt_f32_f16_e32 v41, v41
; SI-NEXT: v_cvt_f32_f16_e32 v47, v5
; SI-NEXT: v_cvt_f32_f16_e32 v57, v4
; SI-NEXT: v_cvt_f32_f16_e32 v59, v3
; SI-NEXT: v_cvt_f32_f16_e32 v61, v2
-; SI-NEXT: v_mov_b32_e32 v52, v29
-; SI-NEXT: v_mov_b32_e32 v48, v30
-; SI-NEXT: v_mov_b32_e32 v56, v28
-; SI-NEXT: v_mov_b32_e32 v34, v7
-; SI-NEXT: v_mov_b32_e32 v32, v6
-; SI-NEXT: v_mov_b32_e32 v46, v8
-; SI-NEXT: s_waitcnt vmcnt(3)
-; SI-NEXT: v_cvt_f32_f16_e32 v38, v1
-; SI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:208 ; 4-byte Folded Reload
-; SI-NEXT: buffer_store_dword v27, off, s[0:3], s32 offset:80 ; 4-byte Folded Spill
-; SI-NEXT: buffer_store_dword v26, off, s[0:3], s32 offset:88 ; 4-byte Folded Spill
-; SI-NEXT: buffer_store_dword v25, off, s[0:3], s32 offset:96 ; 4-byte Folded Spill
-; SI-NEXT: buffer_store_dword v24, off, s[0:3], s32 offset:104 ; 4-byte Folded Spill
-; SI-NEXT: buffer_store_dword v23, off, s[0:3], s32 offset:108 ; 4-byte Folded Spill
-; SI-NEXT: buffer_store_dword v22, off, s[0:3], s32 offset:116 ; 4-byte Folded Spill
-; SI-NEXT: buffer_store_dword v21, off, s[0:3], s32 offset:124 ; 4-byte Folded Spill
-; SI-NEXT: buffer_store_dword v20, off, s[0:3], s32 offset:132 ; 4-byte Folded Spill
-; SI-NEXT: buffer_store_dword v19, off, s[0:3], s32 offset:140 ; 4-byte Folded Spill
-; SI-NEXT: buffer_store_dword v18, off, s[0:3], s32 offset:148 ; 4-byte Folded Spill
-; SI-NEXT: buffer_store_dword v17, off, s[0:3], s32 offset:156 ; 4-byte Folded Spill
-; SI-NEXT: buffer_store_dword v16, off, s[0:3], s32 offset:164 ; 4-byte Folded Spill
-; SI-NEXT: buffer_store_dword v15, off, s[0:3], s32 offset:168 ; 4-byte Folded Spill
-; SI-NEXT: buffer_store_dword v14, off, s[0:3], s32 offset:172 ; 4-byte Folded Spill
-; SI-NEXT: buffer_store_dword v13, off, s[0:3], s32 offset:176 ; 4-byte Folded Spill
-; SI-NEXT: buffer_store_dword v12, off, s[0:3], s32 offset:180 ; 4-byte Folded Spill
-; SI-NEXT: buffer_store_dword v11, off, s[0:3], s32 offset:184 ; 4-byte Folded Spill
-; SI-NEXT: buffer_store_dword v10, off, s[0:3], s32 offset:188 ; 4-byte Folded Spill
-; SI-NEXT: buffer_store_dword v9, off, s[0:3], s32 offset:76 ; 4-byte Folded Spill
+; SI-NEXT: v_mov_b32_e32 v50, v30
+; SI-NEXT: v_mov_b32_e32 v55, v29
+; SI-NEXT: v_mov_b32_e32 v44, v8
+; SI-NEXT: buffer_store_dword v28, off, s[0:3], s32 offset:76 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v27, off, s[0:3], s32 offset:84 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v26, off, s[0:3], s32 offset:92 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v25, off, s[0:3], s32 offset:100 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v24, off, s[0:3], s32 offset:108 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v23, off, s[0:3], s32 offset:112 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v22, off, s[0:3], s32 offset:120 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v21, off, s[0:3], s32 offset:128 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v20, off, s[0:3], s32 offset:136 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v19, off, s[0:3], s32 offset:144 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v18, off, s[0:3], s32 offset:152 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v17, off, s[0:3], s32 offset:160 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v16, off, s[0:3], s32 offset:168 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v15, off, s[0:3], s32 offset:172 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v14, off, s[0:3], s32 offset:176 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v13, off, s[0:3], s32 offset:180 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v12, off, s[0:3], s32 offset:184 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v11, off, s[0:3], s32 offset:188 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v10, off, s[0:3], s32 offset:192 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v9, off, s[0:3], s32 offset:96 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v7, off, s[0:3], s32 offset:104 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v6, off, s[0:3], s32 offset:116 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt vmcnt(14)
+; SI-NEXT: v_cvt_f32_f16_e32 v43, v1
+; SI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:208 ; 4-byte Folded Reload
+; SI-NEXT: s_waitcnt vmcnt(0)
; SI-NEXT: v_cvt_f32_f16_e32 v45, v1
; SI-NEXT: .LBB44_4: ; %end
; SI-NEXT: s_or_b64 exec, exec, s[4:5]
@@ -71128,34 +71146,32 @@ define <64 x half> @bitcast_v32f32_to_v64f16(<32 x float> %a, i32 %b) {
; SI-NEXT: buffer_store_dword v1, v2, s[0:3], 0 offen
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_cvt_f16_f32_e32 v1, v57
-; SI-NEXT: v_cvt_f16_f32_e32 v2, v43
+; SI-NEXT: v_cvt_f16_f32_e32 v2, v56
; SI-NEXT: v_lshlrev_b32_e32 v1, 16, v1
; SI-NEXT: v_or_b32_e32 v1, v2, v1
; SI-NEXT: v_add_i32_e32 v2, vcc, 8, v0
; SI-NEXT: buffer_store_dword v1, v2, s[0:3], 0 offen
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_cvt_f16_f32_e32 v1, v47
-; SI-NEXT: v_cvt_f16_f32_e32 v2, v42
+; SI-NEXT: v_cvt_f16_f32_e32 v2, v46
; SI-NEXT: v_lshlrev_b32_e32 v1, 16, v1
; SI-NEXT: v_or_b32_e32 v1, v2, v1
; SI-NEXT: buffer_store_dword v1, v3, s[0:3], 0 offen
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_cvt_f16_f32_e32 v1, v45
-; SI-NEXT: v_cvt_f16_f32_e32 v2, v44
+; SI-NEXT: v_cvt_f16_f32_e32 v2, v40
; SI-NEXT: v_add_i32_e32 v3, vcc, 16, v0
; SI-NEXT: v_lshlrev_b32_e32 v1, 16, v1
; SI-NEXT: v_or_b32_e32 v1, v2, v1
; SI-NEXT: buffer_store_dword v1, v3, s[0:3], 0 offen
-; SI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:204 ; 4-byte Folded Reload
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_cvt_f16_f32_e32 v1, v38
+; SI-NEXT: v_cvt_f16_f32_e32 v1, v43
+; SI-NEXT: v_cvt_f16_f32_e32 v2, v42
; SI-NEXT: v_add_i32_e32 v3, vcc, 20, v0
; SI-NEXT: v_lshlrev_b32_e32 v1, 16, v1
-; SI-NEXT: s_waitcnt vmcnt(0)
-; SI-NEXT: v_cvt_f16_f32_e32 v2, v2
; SI-NEXT: v_or_b32_e32 v1, v2, v1
; SI-NEXT: buffer_store_dword v1, v3, s[0:3], 0 offen
-; SI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:200 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:204 ; 4-byte Folded Reload
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_cvt_f16_f32_e32 v1, v41
; SI-NEXT: v_add_i32_e32 v3, vcc, 24, v0
@@ -71164,25 +71180,25 @@ define <64 x half> @bitcast_v32f32_to_v64f16(<32 x float> %a, i32 %b) {
; SI-NEXT: v_cvt_f16_f32_e32 v2, v2
; SI-NEXT: v_or_b32_e32 v1, v2, v1
; SI-NEXT: buffer_store_dword v1, v3, s[0:3], 0 offen
-; SI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:196 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:200 ; 4-byte Folded Reload
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_cvt_f16_f32_e32 v1, v55
+; SI-NEXT: v_cvt_f16_f32_e32 v1, v38
; SI-NEXT: v_add_i32_e32 v3, vcc, 28, v0
; SI-NEXT: v_lshlrev_b32_e32 v1, 16, v1
; SI-NEXT: s_waitcnt vmcnt(0)
; SI-NEXT: v_cvt_f16_f32_e32 v2, v2
; SI-NEXT: v_or_b32_e32 v1, v2, v1
; SI-NEXT: buffer_store_dword v1, v3, s[0:3], 0 offen
-; SI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:192 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:196 ; 4-byte Folded Reload
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_cvt_f16_f32_e32 v1, v53
+; SI-NEXT: v_cvt_f16_f32_e32 v1, v31
; SI-NEXT: v_add_i32_e32 v3, vcc, 32, v0
; SI-NEXT: v_lshlrev_b32_e32 v1, 16, v1
; SI-NEXT: s_waitcnt vmcnt(0)
; SI-NEXT: v_cvt_f16_f32_e32 v2, v2
; SI-NEXT: v_or_b32_e32 v1, v2, v1
; SI-NEXT: buffer_store_dword v1, v3, s[0:3], 0 offen
-; SI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:188 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:192 ; 4-byte Folded Reload
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_cvt_f16_f32_e32 v1, v51
; SI-NEXT: v_add_i32_e32 v3, vcc, 36, v0
@@ -71191,7 +71207,7 @@ define <64 x half> @bitcast_v32f32_to_v64f16(<32 x float> %a, i32 %b) {
; SI-NEXT: v_cvt_f16_f32_e32 v2, v2
; SI-NEXT: v_or_b32_e32 v1, v2, v1
; SI-NEXT: buffer_store_dword v1, v3, s[0:3], 0 offen
-; SI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:184 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:188 ; 4-byte Folded Reload
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_cvt_f16_f32_e32 v1, v49
; SI-NEXT: v_add_i32_e32 v3, vcc, 40, v0
@@ -71200,7 +71216,7 @@ define <64 x half> @bitcast_v32f32_to_v64f16(<32 x float> %a, i32 %b) {
; SI-NEXT: v_cvt_f16_f32_e32 v2, v2
; SI-NEXT: v_or_b32_e32 v1, v2, v1
; SI-NEXT: buffer_store_dword v1, v3, s[0:3], 0 offen
-; SI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:180 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:184 ; 4-byte Folded Reload
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_cvt_f16_f32_e32 v1, v39
; SI-NEXT: v_add_i32_e32 v3, vcc, 44, v0
@@ -71209,7 +71225,7 @@ define <64 x half> @bitcast_v32f32_to_v64f16(<32 x float> %a, i32 %b) {
; SI-NEXT: v_cvt_f16_f32_e32 v2, v2
; SI-NEXT: v_or_b32_e32 v1, v2, v1
; SI-NEXT: buffer_store_dword v1, v3, s[0:3], 0 offen
-; SI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:176 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:180 ; 4-byte Folded Reload
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_cvt_f16_f32_e32 v1, v37
; SI-NEXT: v_add_i32_e32 v3, vcc, 48, v0
@@ -71218,7 +71234,7 @@ define <64 x half> @bitcast_v32f32_to_v64f16(<32 x float> %a, i32 %b) {
; SI-NEXT: v_cvt_f16_f32_e32 v2, v2
; SI-NEXT: v_or_b32_e32 v1, v2, v1
; SI-NEXT: buffer_store_dword v1, v3, s[0:3], 0 offen
-; SI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:172 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:176 ; 4-byte Folded Reload
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_cvt_f16_f32_e32 v1, v35
; SI-NEXT: v_add_i32_e32 v3, vcc, 52, v0
@@ -71227,7 +71243,7 @@ define <64 x half> @bitcast_v32f32_to_v64f16(<32 x float> %a, i32 %b) {
; SI-NEXT: v_cvt_f16_f32_e32 v2, v2
; SI-NEXT: v_or_b32_e32 v1, v2, v1
; SI-NEXT: buffer_store_dword v1, v3, s[0:3], 0 offen
-; SI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:168 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:172 ; 4-byte Folded Reload
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_cvt_f16_f32_e32 v1, v33
; SI-NEXT: v_add_i32_e32 v3, vcc, 56, v0
@@ -71237,8 +71253,8 @@ define <64 x half> @bitcast_v32f32_to_v64f16(<32 x float> %a, i32 %b) {
; SI-NEXT: v_or_b32_e32 v1, v2, v1
; SI-NEXT: buffer_store_dword v1, v3, s[0:3], 0 offen
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:160 ; 4-byte Folded Reload
-; SI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:164 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:164 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:168 ; 4-byte Folded Reload
; SI-NEXT: v_add_i32_e32 v3, vcc, 60, v0
; SI-NEXT: s_waitcnt vmcnt(1)
; SI-NEXT: v_cvt_f16_f32_e32 v1, v1
@@ -71248,8 +71264,8 @@ define <64 x half> @bitcast_v32f32_to_v64f16(<32 x float> %a, i32 %b) {
; SI-NEXT: v_or_b32_e32 v1, v2, v1
; SI-NEXT: buffer_store_dword v1, v3, s[0:3], 0 offen
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:152 ; 4-byte Folded Reload
-; SI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:156 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:156 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:160 ; 4-byte Folded Reload
; SI-NEXT: v_add_i32_e32 v3, vcc, 64, v0
; SI-NEXT: s_waitcnt vmcnt(1)
; SI-NEXT: v_cvt_f16_f32_e32 v1, v1
@@ -71259,8 +71275,8 @@ define <64 x half> @bitcast_v32f32_to_v64f16(<32 x float> %a, i32 %b) {
; SI-NEXT: v_or_b32_e32 v1, v2, v1
; SI-NEXT: buffer_store_dword v1, v3, s[0:3], 0 offen
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:144 ; 4-byte Folded Reload
-; SI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:148 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:148 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:152 ; 4-byte Folded Reload
; SI-NEXT: v_add_i32_e32 v3, vcc, 0x44, v0
; SI-NEXT: s_waitcnt vmcnt(1)
; SI-NEXT: v_cvt_f16_f32_e32 v1, v1
@@ -71270,8 +71286,8 @@ define <64 x half> @bitcast_v32f32_to_v64f16(<32 x float> %a, i32 %b) {
; SI-NEXT: v_or_b32_e32 v1, v2, v1
; SI-NEXT: buffer_store_dword v1, v3, s[0:3], 0 offen
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:136 ; 4-byte Folded Reload
-; SI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:140 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:140 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:144 ; 4-byte Folded Reload
; SI-NEXT: v_add_i32_e32 v3, vcc, 0x48, v0
; SI-NEXT: s_waitcnt vmcnt(1)
; SI-NEXT: v_cvt_f16_f32_e32 v1, v1
@@ -71281,8 +71297,8 @@ define <64 x half> @bitcast_v32f32_to_v64f16(<32 x float> %a, i32 %b) {
; SI-NEXT: v_or_b32_e32 v1, v2, v1
; SI-NEXT: buffer_store_dword v1, v3, s[0:3], 0 offen
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:128 ; 4-byte Folded Reload
-; SI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:132 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:132 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:136 ; 4-byte Folded Reload
; SI-NEXT: v_add_i32_e32 v3, vcc, 0x4c, v0
; SI-NEXT: s_waitcnt vmcnt(1)
; SI-NEXT: v_cvt_f16_f32_e32 v1, v1
@@ -71292,8 +71308,8 @@ define <64 x half> @bitcast_v32f32_to_v64f16(<32 x float> %a, i32 %b) {
; SI-NEXT: v_or_b32_e32 v1, v2, v1
; SI-NEXT: buffer_store_dword v1, v3, s[0:3], 0 offen
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:120 ; 4-byte Folded Reload
-; SI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:124 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:124 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:128 ; 4-byte Folded Reload
; SI-NEXT: v_add_i32_e32 v3, vcc, 0x50, v0
; SI-NEXT: s_waitcnt vmcnt(1)
; SI-NEXT: v_cvt_f16_f32_e32 v1, v1
@@ -71303,8 +71319,8 @@ define <64 x half> @bitcast_v32f32_to_v64f16(<32 x float> %a, i32 %b) {
; SI-NEXT: v_or_b32_e32 v1, v2, v1
; SI-NEXT: buffer_store_dword v1, v3, s[0:3], 0 offen
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:112 ; 4-byte Folded Reload
-; SI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:116 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:116 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:120 ; 4-byte Folded Reload
; SI-NEXT: v_add_i32_e32 v3, vcc, 0x54, v0
; SI-NEXT: s_waitcnt vmcnt(1)
; SI-NEXT: v_cvt_f16_f32_e32 v1, v1
@@ -71314,8 +71330,8 @@ define <64 x half> @bitcast_v32f32_to_v64f16(<32 x float> %a, i32 %b) {
; SI-NEXT: v_or_b32_e32 v1, v2, v1
; SI-NEXT: buffer_store_dword v1, v3, s[0:3], 0 offen
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:100 ; 4-byte Folded Reload
-; SI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:108 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:104 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:112 ; 4-byte Folded Reload
; SI-NEXT: v_add_i32_e32 v3, vcc, 0x58, v0
; SI-NEXT: s_waitcnt vmcnt(1)
; SI-NEXT: v_cvt_f16_f32_e32 v1, v1
@@ -71325,8 +71341,8 @@ define <64 x half> @bitcast_v32f32_to_v64f16(<32 x float> %a, i32 %b) {
; SI-NEXT: v_or_b32_e32 v1, v2, v1
; SI-NEXT: buffer_store_dword v1, v3, s[0:3], 0 offen
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:92 ; 4-byte Folded Reload
-; SI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:104 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:96 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:108 ; 4-byte Folded Reload
; SI-NEXT: v_add_i32_e32 v3, vcc, 0x5c, v0
; SI-NEXT: s_waitcnt vmcnt(1)
; SI-NEXT: v_cvt_f16_f32_e32 v1, v1
@@ -71336,8 +71352,8 @@ define <64 x half> @bitcast_v32f32_to_v64f16(<32 x float> %a, i32 %b) {
; SI-NEXT: v_or_b32_e32 v1, v2, v1
; SI-NEXT: buffer_store_dword v1, v3, s[0:3], 0 offen
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:84 ; 4-byte Folded Reload
-; SI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:96 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:88 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:100 ; 4-byte Folded Reload
; SI-NEXT: v_add_i32_e32 v3, vcc, 0x60, v0
; SI-NEXT: s_waitcnt vmcnt(1)
; SI-NEXT: v_cvt_f16_f32_e32 v1, v1
@@ -71347,8 +71363,8 @@ define <64 x half> @bitcast_v32f32_to_v64f16(<32 x float> %a, i32 %b) {
; SI-NEXT: v_or_b32_e32 v1, v2, v1
; SI-NEXT: buffer_store_dword v1, v3, s[0:3], 0 offen
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:76 ; 4-byte Folded Reload
-; SI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:88 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:80 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:92 ; 4-byte Folded Reload
; SI-NEXT: v_add_i32_e32 v3, vcc, 0x64, v0
; SI-NEXT: s_waitcnt vmcnt(1)
; SI-NEXT: v_cvt_f16_f32_e32 v1, v1
@@ -71357,46 +71373,48 @@ define <64 x half> @bitcast_v32f32_to_v64f16(<32 x float> %a, i32 %b) {
; SI-NEXT: v_lshlrev_b32_e32 v1, 16, v1
; SI-NEXT: v_or_b32_e32 v1, v2, v1
; SI-NEXT: buffer_store_dword v1, v3, s[0:3], 0 offen
-; SI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:80 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:84 ; 4-byte Folded Reload
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_cvt_f16_f32_e32 v1, v46
+; SI-NEXT: v_cvt_f16_f32_e32 v1, v44
; SI-NEXT: v_add_i32_e32 v3, vcc, 0x68, v0
; SI-NEXT: v_lshlrev_b32_e32 v1, 16, v1
; SI-NEXT: s_waitcnt vmcnt(0)
; SI-NEXT: v_cvt_f16_f32_e32 v2, v2
; SI-NEXT: v_or_b32_e32 v1, v2, v1
; SI-NEXT: buffer_store_dword v1, v3, s[0:3], 0 offen
+; SI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:76 ; 4-byte Folded Reload
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_cvt_f16_f32_e32 v1, v50
-; SI-NEXT: v_cvt_f16_f32_e32 v2, v56
+; SI-NEXT: v_cvt_f16_f32_e32 v1, v52
; SI-NEXT: v_add_i32_e32 v3, vcc, 0x6c, v0
; SI-NEXT: v_lshlrev_b32_e32 v1, 16, v1
+; SI-NEXT: s_waitcnt vmcnt(0)
+; SI-NEXT: v_cvt_f16_f32_e32 v2, v2
; SI-NEXT: v_or_b32_e32 v1, v2, v1
; SI-NEXT: buffer_store_dword v1, v3, s[0:3], 0 offen
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_cvt_f16_f32_e32 v1, v54
-; SI-NEXT: v_cvt_f16_f32_e32 v2, v52
+; SI-NEXT: v_cvt_f16_f32_e32 v2, v55
; SI-NEXT: v_add_i32_e32 v3, vcc, 0x70, v0
; SI-NEXT: v_lshlrev_b32_e32 v1, 16, v1
; SI-NEXT: v_or_b32_e32 v1, v2, v1
; SI-NEXT: buffer_store_dword v1, v3, s[0:3], 0 offen
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_cvt_f16_f32_e32 v1, v40
-; SI-NEXT: v_cvt_f16_f32_e32 v2, v48
+; SI-NEXT: v_cvt_f16_f32_e32 v1, v53
+; SI-NEXT: v_cvt_f16_f32_e32 v2, v50
; SI-NEXT: v_add_i32_e32 v3, vcc, 0x74, v0
; SI-NEXT: v_lshlrev_b32_e32 v1, 16, v1
; SI-NEXT: v_or_b32_e32 v1, v2, v1
; SI-NEXT: buffer_store_dword v1, v3, s[0:3], 0 offen
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_cvt_f16_f32_e32 v1, v34
-; SI-NEXT: v_cvt_f16_f32_e32 v2, v36
+; SI-NEXT: v_cvt_f16_f32_e32 v1, v36
+; SI-NEXT: v_cvt_f16_f32_e32 v2, v48
; SI-NEXT: v_add_i32_e32 v3, vcc, 0x78, v0
; SI-NEXT: v_lshlrev_b32_e32 v1, 16, v1
; SI-NEXT: v_or_b32_e32 v1, v2, v1
; SI-NEXT: buffer_store_dword v1, v3, s[0:3], 0 offen
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_cvt_f16_f32_e32 v1, v32
-; SI-NEXT: v_cvt_f16_f32_e32 v2, v31
+; SI-NEXT: v_cvt_f16_f32_e32 v1, v34
+; SI-NEXT: v_cvt_f16_f32_e32 v2, v32
; SI-NEXT: v_add_i32_e32 v0, vcc, 0x7c, v0
; SI-NEXT: v_lshlrev_b32_e32 v1, 16, v1
; SI-NEXT: v_or_b32_e32 v1, v2, v1
@@ -71629,12 +71647,12 @@ define inreg <64 x half> @bitcast_v32f32_to_v64f16_scalar(<32 x float> inreg %a,
; SI-NEXT: s_lshr_b32 s4, s12, 16
; SI-NEXT: v_cvt_f32_f16_e32 v32, s4
; SI-NEXT: s_lshr_b32 s4, s13, 16
-; SI-NEXT: v_cvt_f32_f16_e32 v13, s4
+; SI-NEXT: s_waitcnt expcnt(2)
+; SI-NEXT: v_cvt_f32_f16_e32 v61, s4
; SI-NEXT: s_lshr_b32 s4, s14, 16
; SI-NEXT: v_cvt_f32_f16_e32 v11, s4
; SI-NEXT: s_lshr_b32 s4, s15, 16
-; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_cvt_f32_f16_e32 v63, s4
+; SI-NEXT: v_cvt_f32_f16_e32 v18, s4
; SI-NEXT: s_lshr_b32 s4, s40, 16
; SI-NEXT: v_cvt_f32_f16_e32 v21, s4
; SI-NEXT: s_lshr_b32 s4, s41, 16
@@ -71690,8 +71708,8 @@ define inreg <64 x half> @bitcast_v32f32_to_v64f16_scalar(<32 x float> inreg %a,
; SI-NEXT: v_cvt_f32_f16_e32 v6, s6
; SI-NEXT: v_cvt_f32_f16_e32 v7, s7
; SI-NEXT: v_cvt_f32_f16_e32 v9, s8
-; SI-NEXT: v_cvt_f32_f16_e32 v59, s10
-; SI-NEXT: v_cvt_f32_f16_e32 v61, s11
+; SI-NEXT: v_cvt_f32_f16_e32 v13, s10
+; SI-NEXT: v_cvt_f32_f16_e32 v59, s11
; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:72 ; 4-byte Folded Spill
; SI-NEXT: v_cvt_f32_f16_e32 v20, s15
; SI-NEXT: v_cvt_f32_f16_e32 v22, s40
@@ -71714,67 +71732,83 @@ define inreg <64 x half> @bitcast_v32f32_to_v64f16_scalar(<32 x float> inreg %a,
; SI-NEXT: v_cvt_f32_f16_e32 v4, s20
; SI-NEXT: v_cvt_f32_f16_e32 v10, s19
; SI-NEXT: v_cvt_f32_f16_e32 v14, s18
-; SI-NEXT: v_cvt_f32_f16_e32 v18, s17
+; SI-NEXT: v_cvt_f32_f16_e32 v63, s17
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_cvt_f32_f16_e32 v1, s16
; SI-NEXT: s_cbranch_execnz .LBB45_3
; SI-NEXT: .LBB45_2: ; %cmp.true
+; SI-NEXT: v_add_f32_e64 v1, s18, 1.0
+; SI-NEXT: v_lshrrev_b32_e32 v2, 16, v1
+; SI-NEXT: buffer_store_dword v2, off, s[0:3], s32 offset:76 ; 4-byte Folded Spill
+; SI-NEXT: s_waitcnt expcnt(0)
+; SI-NEXT: v_add_f32_e64 v2, s19, 1.0
+; SI-NEXT: v_lshrrev_b32_e32 v3, 16, v2
+; SI-NEXT: v_add_f32_e64 v4, s20, 1.0
+; SI-NEXT: buffer_store_dword v3, off, s[0:3], s32 offset:80 ; 4-byte Folded Spill
+; SI-NEXT: s_waitcnt expcnt(0)
+; SI-NEXT: v_lshrrev_b32_e32 v3, 16, v4
+; SI-NEXT: v_add_f32_e64 v20, s15, 1.0
+; SI-NEXT: buffer_store_dword v3, off, s[0:3], s32 offset:88 ; 4-byte Folded Spill
+; SI-NEXT: v_add_f32_e64 v18, s27, 1.0
+; SI-NEXT: s_waitcnt expcnt(0)
+; SI-NEXT: v_lshrrev_b32_e32 v3, 16, v20
+; SI-NEXT: v_add_f32_e64 v49, s8, 1.0
+; SI-NEXT: v_lshrrev_b32_e32 v40, 16, v18
+; SI-NEXT: v_lshrrev_b32_e32 v51, 16, v49
+; SI-NEXT: v_cvt_f32_f16_e32 v9, v49
+; SI-NEXT: v_cvt_f32_f16_e32 v49, v18
+; SI-NEXT: v_cvt_f32_f16_e32 v18, v3
+; SI-NEXT: buffer_load_dword v3, off, s[0:3], s32 offset:88 ; 4-byte Folded Reload
; SI-NEXT: v_add_f32_e64 v15, s12, 1.0
; SI-NEXT: v_lshrrev_b32_e32 v32, 16, v15
; SI-NEXT: v_cvt_f32_f16_e32 v15, v15
-; SI-NEXT: v_add_f32_e64 v1, s18, 1.0
; SI-NEXT: v_add_f32_e64 v17, s13, 1.0
-; SI-NEXT: v_lshrrev_b32_e32 v2, 16, v1
+; SI-NEXT: v_add_f32_e64 v10, s23, 1.0
+; SI-NEXT: v_add_f32_e64 v16, s26, 1.0
; SI-NEXT: buffer_store_dword v15, off, s[0:3], s32 offset:64 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_cvt_f32_f16_e32 v15, v17
-; SI-NEXT: buffer_store_dword v2, off, s[0:3], s32 offset:76 ; 4-byte Folded Spill
-; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_add_f32_e64 v2, s19, 1.0
-; SI-NEXT: v_lshrrev_b32_e32 v3, 16, v2
+; SI-NEXT: v_add_f32_e64 v23, s29, 1.0
+; SI-NEXT: v_add_f32_e64 v29, s45, 1.0
; SI-NEXT: v_add_f32_e64 v22, s40, 1.0
-; SI-NEXT: buffer_store_dword v3, off, s[0:3], s32 offset:80 ; 4-byte Folded Spill
+; SI-NEXT: v_lshrrev_b32_e32 v56, 16, v10
+; SI-NEXT: v_lshrrev_b32_e32 v42, 16, v16
; SI-NEXT: v_add_f32_e64 v21, s28, 1.0
-; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_lshrrev_b32_e32 v3, 16, v22
+; SI-NEXT: v_lshrrev_b32_e32 v52, 16, v23
+; SI-NEXT: v_lshrrev_b32_e32 v38, 16, v29
+; SI-NEXT: v_lshrrev_b32_e32 v31, 16, v22
; SI-NEXT: v_add_f32_e64 v19, s14, 1.0
-; SI-NEXT: v_lshrrev_b32_e32 v52, 16, v21
+; SI-NEXT: v_lshrrev_b32_e32 v54, 16, v21
; SI-NEXT: v_lshrrev_b32_e32 v11, 16, v19
; SI-NEXT: buffer_store_dword v15, off, s[0:3], s32 offset:68 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_cvt_f32_f16_e32 v15, v19
; SI-NEXT: v_cvt_f32_f16_e32 v19, v21
-; SI-NEXT: v_cvt_f32_f16_e32 v21, v3
-; SI-NEXT: buffer_load_dword v3, off, s[0:3], s32 offset:80 ; 4-byte Folded Reload
+; SI-NEXT: v_cvt_f32_f16_e32 v21, v31
+; SI-NEXT: v_cvt_f32_f16_e32 v31, v38
+; SI-NEXT: v_cvt_f32_f16_e32 v38, v52
+; SI-NEXT: v_cvt_f32_f16_e32 v52, v42
+; SI-NEXT: v_cvt_f32_f16_e32 v42, v56
; SI-NEXT: v_add_f32_e64 v41, s6, 1.0
; SI-NEXT: v_cvt_f32_f16_e32 v7, v41
-; SI-NEXT: v_add_f32_e64 v6, s21, 1.0
-; SI-NEXT: v_add_f32_e64 v10, s23, 1.0
+; SI-NEXT: v_add_f32_e64 v8, s22, 1.0
; SI-NEXT: v_add_f32_e64 v14, s25, 1.0
-; SI-NEXT: v_add_f32_e64 v18, s27, 1.0
-; SI-NEXT: v_add_f32_e64 v23, s29, 1.0
; SI-NEXT: v_add_f32_e64 v27, s46, 1.0
-; SI-NEXT: v_add_f32_e64 v26, s42, 1.0
-; SI-NEXT: v_lshrrev_b32_e32 v57, 16, v6
-; SI-NEXT: v_lshrrev_b32_e32 v46, 16, v10
-; SI-NEXT: v_lshrrev_b32_e32 v42, 16, v14
-; SI-NEXT: v_lshrrev_b32_e32 v54, 16, v18
-; SI-NEXT: v_lshrrev_b32_e32 v50, 16, v23
-; SI-NEXT: v_add_f32_e64 v25, s47, 1.0
-; SI-NEXT: v_lshrrev_b32_e32 v38, 16, v27
-; SI-NEXT: v_lshrrev_b32_e32 v33, 16, v26
-; SI-NEXT: v_lshrrev_b32_e32 v48, 16, v25
-; SI-NEXT: v_lshrrev_b32_e32 v13, 16, v17
+; SI-NEXT: v_add_f32_e64 v24, s41, 1.0
+; SI-NEXT: v_lshrrev_b32_e32 v57, 16, v8
+; SI-NEXT: v_lshrrev_b32_e32 v44, 16, v14
+; SI-NEXT: v_lshrrev_b32_e32 v48, 16, v27
+; SI-NEXT: v_lshrrev_b32_e32 v33, 16, v24
+; SI-NEXT: v_add_f32_e64 v37, s10, 1.0
+; SI-NEXT: v_lshrrev_b32_e32 v39, 16, v37
; SI-NEXT: v_add_f32_e64 v53, s7, 1.0
-; SI-NEXT: v_cvt_f32_f16_e32 v17, v25
-; SI-NEXT: v_cvt_f32_f16_e32 v25, v33
-; SI-NEXT: v_cvt_f32_f16_e32 v33, v38
-; SI-NEXT: v_cvt_f32_f16_e32 v38, v50
-; SI-NEXT: v_cvt_f32_f16_e32 v50, v54
-; SI-NEXT: v_cvt_f32_f16_e32 v54, v42
-; SI-NEXT: v_cvt_f32_f16_e32 v42, v46
-; SI-NEXT: v_cvt_f32_f16_e32 v46, v57
-; SI-NEXT: v_add_f32_e64 v49, s8, 1.0
+; SI-NEXT: v_cvt_f32_f16_e32 v13, v37
+; SI-NEXT: v_cvt_f32_f16_e32 v37, v23
+; SI-NEXT: v_cvt_f32_f16_e32 v23, v33
+; SI-NEXT: v_cvt_f32_f16_e32 v33, v48
+; SI-NEXT: v_cvt_f32_f16_e32 v48, v54
+; SI-NEXT: v_cvt_f32_f16_e32 v54, v44
+; SI-NEXT: v_cvt_f32_f16_e32 v44, v57
; SI-NEXT: v_lshrrev_b32_e32 v55, 16, v53
; SI-NEXT: buffer_store_dword v7, off, s[0:3], s32 offset:84 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
@@ -71782,17 +71816,16 @@ define inreg <64 x half> @bitcast_v32f32_to_v64f16_scalar(<32 x float> inreg %a,
; SI-NEXT: v_cvt_f32_f16_e32 v53, v14
; SI-NEXT: v_cvt_f32_f16_e32 v14, v1
; SI-NEXT: v_add_f32_e64 v1, s17, 1.0
+; SI-NEXT: v_add_f32_e64 v6, s21, 1.0
; SI-NEXT: v_add_f32_e64 v28, s43, 1.0
-; SI-NEXT: v_lshrrev_b32_e32 v51, 16, v49
; SI-NEXT: v_lshrrev_b32_e32 v43, 16, v41
; SI-NEXT: v_add_f32_e64 v45, s9, 1.0
-; SI-NEXT: v_cvt_f32_f16_e32 v9, v49
-; SI-NEXT: v_cvt_f32_f16_e32 v49, v18
; SI-NEXT: v_cvt_f32_f16_e32 v41, v10
; SI-NEXT: v_cvt_f32_f16_e32 v10, v2
; SI-NEXT: v_lshrrev_b32_e32 v2, 16, v1
-; SI-NEXT: v_cvt_f32_f16_e32 v18, v1
+; SI-NEXT: v_cvt_f32_f16_e32 v63, v1
; SI-NEXT: v_add_f32_e64 v1, s16, 1.0
+; SI-NEXT: v_lshrrev_b32_e32 v58, 16, v6
; SI-NEXT: v_lshrrev_b32_e32 v62, 16, v28
; SI-NEXT: v_add_f32_e64 v34, s11, 1.0
; SI-NEXT: v_lshrrev_b32_e32 v47, 16, v45
@@ -71800,35 +71833,24 @@ define inreg <64 x half> @bitcast_v32f32_to_v64f16_scalar(<32 x float> inreg %a,
; SI-NEXT: v_cvt_f32_f16_e32 v45, v6
; SI-NEXT: v_lshrrev_b32_e32 v6, 16, v1
; SI-NEXT: v_lshrrev_b32_e32 v35, 16, v34
-; SI-NEXT: v_cvt_f32_f16_e32 v61, v34
+; SI-NEXT: v_cvt_f32_f16_e32 v59, v34
; SI-NEXT: v_cvt_f32_f16_e32 v34, v27
; SI-NEXT: v_cvt_f32_f16_e32 v27, v62
; SI-NEXT: v_cvt_f32_f16_e32 v62, v6
-; SI-NEXT: v_add_f32_e64 v4, s20, 1.0
-; SI-NEXT: v_add_f32_e64 v8, s22, 1.0
; SI-NEXT: v_add_f32_e64 v12, s24, 1.0
-; SI-NEXT: v_add_f32_e64 v16, s26, 1.0
-; SI-NEXT: v_add_f32_e64 v29, s45, 1.0
+; SI-NEXT: v_add_f32_e64 v25, s47, 1.0
; SI-NEXT: v_add_f32_e64 v30, s44, 1.0
-; SI-NEXT: v_add_f32_e64 v24, s41, 1.0
-; SI-NEXT: v_add_f32_e64 v20, s15, 1.0
-; SI-NEXT: v_add_f32_e64 v37, s10, 1.0
-; SI-NEXT: buffer_load_dword v6, off, s[0:3], s32 offset:84 ; 4-byte Folded Reload
-; SI-NEXT: v_lshrrev_b32_e32 v58, 16, v4
-; SI-NEXT: v_lshrrev_b32_e32 v56, 16, v8
-; SI-NEXT: v_lshrrev_b32_e32 v44, 16, v12
-; SI-NEXT: v_lshrrev_b32_e32 v40, 16, v16
-; SI-NEXT: v_lshrrev_b32_e32 v36, 16, v29
+; SI-NEXT: v_add_f32_e64 v26, s42, 1.0
+; SI-NEXT: s_waitcnt vmcnt(3)
+; SI-NEXT: v_cvt_f32_f16_e32 v56, v3
+; SI-NEXT: buffer_load_dword v3, off, s[0:3], s32 offset:80 ; 4-byte Folded Reload
+; SI-NEXT: v_lshrrev_b32_e32 v46, 16, v12
+; SI-NEXT: v_lshrrev_b32_e32 v50, 16, v25
; SI-NEXT: v_lshrrev_b32_e32 v60, 16, v30
-; SI-NEXT: v_lshrrev_b32_e32 v31, 16, v24
-; SI-NEXT: v_lshrrev_b32_e32 v63, 16, v20
-; SI-NEXT: v_lshrrev_b32_e32 v39, 16, v37
-; SI-NEXT: v_cvt_f32_f16_e32 v59, v37
+; SI-NEXT: v_lshrrev_b32_e32 v36, 16, v26
+; SI-NEXT: v_lshrrev_b32_e32 v61, 16, v17
; SI-NEXT: buffer_store_dword v15, off, s[0:3], s32 offset:72 ; 4-byte Folded Spill
; SI-NEXT: v_cvt_f32_f16_e32 v20, v20
-; SI-NEXT: s_waitcnt vmcnt(3)
-; SI-NEXT: v_cvt_f32_f16_e32 v57, v3
-; SI-NEXT: buffer_load_dword v3, off, s[0:3], s32 offset:76 ; 4-byte Folded Reload
; SI-NEXT: v_cvt_f32_f16_e32 v22, v22
; SI-NEXT: v_cvt_f32_f16_e32 v24, v24
; SI-NEXT: v_cvt_f32_f16_e32 v26, v26
@@ -71836,7 +71858,7 @@ define inreg <64 x half> @bitcast_v32f32_to_v64f16_scalar(<32 x float> inreg %a,
; SI-NEXT: v_cvt_f32_f16_e32 v30, v30
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_cvt_f32_f16_e32 v15, v29
-; SI-NEXT: v_cvt_f32_f16_e32 v37, v23
+; SI-NEXT: v_cvt_f32_f16_e32 v17, v25
; SI-NEXT: v_cvt_f32_f16_e32 v16, v16
; SI-NEXT: v_cvt_f32_f16_e32 v12, v12
; SI-NEXT: v_cvt_f32_f16_e32 v8, v8
@@ -71849,19 +71871,19 @@ define inreg <64 x half> @bitcast_v32f32_to_v64f16_scalar(<32 x float> inreg %a,
; SI-NEXT: v_cvt_f32_f16_e32 v39, v39
; SI-NEXT: v_cvt_f32_f16_e32 v35, v35
; SI-NEXT: v_cvt_f32_f16_e32 v32, v32
-; SI-NEXT: v_cvt_f32_f16_e32 v13, v13
+; SI-NEXT: v_cvt_f32_f16_e32 v61, v61
; SI-NEXT: v_cvt_f32_f16_e32 v11, v11
-; SI-NEXT: v_cvt_f32_f16_e32 v63, v63
-; SI-NEXT: v_cvt_f32_f16_e32 v23, v31
+; SI-NEXT: v_cvt_f32_f16_e32 v25, v36
; SI-NEXT: v_cvt_f32_f16_e32 v29, v60
-; SI-NEXT: v_cvt_f32_f16_e32 v31, v36
-; SI-NEXT: v_cvt_f32_f16_e32 v36, v48
-; SI-NEXT: v_cvt_f32_f16_e32 v48, v52
-; SI-NEXT: v_cvt_f32_f16_e32 v52, v40
-; SI-NEXT: v_cvt_f32_f16_e32 v40, v44
-; SI-NEXT: v_cvt_f32_f16_e32 v44, v56
-; SI-NEXT: v_cvt_f32_f16_e32 v56, v58
+; SI-NEXT: v_cvt_f32_f16_e32 v36, v50
+; SI-NEXT: v_cvt_f32_f16_e32 v50, v40
+; SI-NEXT: v_cvt_f32_f16_e32 v40, v46
+; SI-NEXT: v_cvt_f32_f16_e32 v46, v58
; SI-NEXT: v_cvt_f32_f16_e32 v60, v2
+; SI-NEXT: buffer_load_dword v6, off, s[0:3], s32 offset:84 ; 4-byte Folded Reload
+; SI-NEXT: s_waitcnt vmcnt(2)
+; SI-NEXT: v_cvt_f32_f16_e32 v57, v3
+; SI-NEXT: buffer_load_dword v3, off, s[0:3], s32 offset:76 ; 4-byte Folded Reload
; SI-NEXT: s_waitcnt vmcnt(0)
; SI-NEXT: v_cvt_f32_f16_e32 v58, v3
; SI-NEXT: .LBB45_3: ; %end
@@ -71873,7 +71895,7 @@ define inreg <64 x half> @bitcast_v32f32_to_v64f16_scalar(<32 x float> inreg %a,
; SI-NEXT: buffer_store_dword v1, v0, s[0:3], 0 offen
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_cvt_f16_f32_e32 v1, v60
-; SI-NEXT: v_cvt_f16_f32_e32 v2, v18
+; SI-NEXT: v_cvt_f16_f32_e32 v2, v63
; SI-NEXT: v_lshlrev_b32_e32 v1, 16, v1
; SI-NEXT: v_or_b32_e32 v1, v2, v1
; SI-NEXT: v_add_i32_e32 v2, vcc, 4, v0
@@ -72018,7 +72040,7 @@ define inreg <64 x half> @bitcast_v32f32_to_v64f16_scalar(<32 x float> inreg %a,
; SI-NEXT: v_or_b32_e32 v1, v2, v1
; SI-NEXT: buffer_store_dword v1, v3, s[0:3], 0 offen
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_cvt_f16_f32_e32 v1, v63
+; SI-NEXT: v_cvt_f16_f32_e32 v1, v18
; SI-NEXT: v_cvt_f16_f32_e32 v2, v20
; SI-NEXT: v_add_i32_e32 v3, vcc, 0x58, v0
; SI-NEXT: v_lshlrev_b32_e32 v1, 16, v1
@@ -72035,7 +72057,7 @@ define inreg <64 x half> @bitcast_v32f32_to_v64f16_scalar(<32 x float> inreg %a,
; SI-NEXT: buffer_store_dword v1, v3, s[0:3], 0 offen
; SI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:68 ; 4-byte Folded Reload
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_cvt_f16_f32_e32 v1, v13
+; SI-NEXT: v_cvt_f16_f32_e32 v1, v61
; SI-NEXT: v_add_i32_e32 v3, vcc, 0x60, v0
; SI-NEXT: v_lshlrev_b32_e32 v1, 16, v1
; SI-NEXT: s_waitcnt vmcnt(0)
@@ -72053,14 +72075,14 @@ define inreg <64 x half> @bitcast_v32f32_to_v64f16_scalar(<32 x float> inreg %a,
; SI-NEXT: buffer_store_dword v1, v3, s[0:3], 0 offen
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_cvt_f16_f32_e32 v1, v35
-; SI-NEXT: v_cvt_f16_f32_e32 v2, v61
+; SI-NEXT: v_cvt_f16_f32_e32 v2, v59
; SI-NEXT: v_add_i32_e32 v3, vcc, 0x68, v0
; SI-NEXT: v_lshlrev_b32_e32 v1, 16, v1
; SI-NEXT: v_or_b32_e32 v1, v2, v1
; SI-NEXT: buffer_store_dword v1, v3, s[0:3], 0 offen
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_cvt_f16_f32_e32 v1, v39
-; SI-NEXT: v_cvt_f16_f32_e32 v2, v59
+; SI-NEXT: v_cvt_f16_f32_e32 v2, v13
; SI-NEXT: v_add_i32_e32 v3, vcc, 0x6c, v0
; SI-NEXT: v_lshlrev_b32_e32 v1, 16, v1
; SI-NEXT: v_or_b32_e32 v1, v2, v1
@@ -72117,7 +72139,7 @@ define inreg <64 x half> @bitcast_v32f32_to_v64f16_scalar(<32 x float> inreg %a,
; SI-NEXT: ; implicit-def: $vgpr2
; SI-NEXT: ; implicit-def: $vgpr1
; SI-NEXT: ; implicit-def: $vgpr62
-; SI-NEXT: ; implicit-def: $vgpr18
+; SI-NEXT: ; implicit-def: $vgpr63
; SI-NEXT: ; implicit-def: $vgpr60
; SI-NEXT: ; implicit-def: $vgpr14
; SI-NEXT: ; implicit-def: $vgpr58
@@ -72160,14 +72182,14 @@ define inreg <64 x half> @bitcast_v32f32_to_v64f16_scalar(<32 x float> inreg %a,
; SI-NEXT: ; implicit-def: $vgpr22
; SI-NEXT: ; implicit-def: $vgpr21
; SI-NEXT: ; implicit-def: $vgpr20
-; SI-NEXT: ; implicit-def: $vgpr63
+; SI-NEXT: ; implicit-def: $vgpr18
; SI-NEXT: ; implicit-def: $vgpr11
; SI-NEXT: ; kill: killed $vgpr2
-; SI-NEXT: ; implicit-def: $vgpr13
-; SI-NEXT: ; implicit-def: $vgpr32
; SI-NEXT: ; implicit-def: $vgpr61
-; SI-NEXT: ; implicit-def: $vgpr35
+; SI-NEXT: ; implicit-def: $vgpr32
; SI-NEXT: ; implicit-def: $vgpr59
+; SI-NEXT: ; implicit-def: $vgpr35
+; SI-NEXT: ; implicit-def: $vgpr13
; SI-NEXT: ; implicit-def: $vgpr39
; SI-NEXT: ; implicit-def: $vgpr9
; SI-NEXT: ; implicit-def: $vgpr51
@@ -76571,24 +76593,24 @@ define inreg <32 x float> @bitcast_v64i16_to_v32f32_scalar(<64 x i16> inreg %a,
; SI-NEXT: buffer_load_dword v10, off, s[0:3], s32 offset:32
; SI-NEXT: buffer_load_dword v46, off, s[0:3], s32 offset:28
; SI-NEXT: buffer_load_dword v12, off, s[0:3], s32 offset:40
-; SI-NEXT: buffer_load_dword v45, off, s[0:3], s32 offset:36
-; SI-NEXT: buffer_load_dword v38, off, s[0:3], s32 offset:48
+; SI-NEXT: buffer_load_dword v59, off, s[0:3], s32 offset:36
+; SI-NEXT: buffer_load_dword v48, off, s[0:3], s32 offset:48
; SI-NEXT: buffer_load_dword v44, off, s[0:3], s32 offset:44
; SI-NEXT: buffer_load_dword v36, off, s[0:3], s32 offset:56
-; SI-NEXT: buffer_load_dword v59, off, s[0:3], s32 offset:52
+; SI-NEXT: buffer_load_dword v43, off, s[0:3], s32 offset:52
; SI-NEXT: buffer_load_dword v34, off, s[0:3], s32 offset:64
; SI-NEXT: buffer_load_dword v42, off, s[0:3], s32 offset:60
; SI-NEXT: buffer_load_dword v32, off, s[0:3], s32 offset:72
-; SI-NEXT: buffer_load_dword v58, off, s[0:3], s32 offset:68
-; SI-NEXT: s_waitcnt expcnt(2)
-; SI-NEXT: v_lshlrev_b32_e32 v61, 16, v1
+; SI-NEXT: buffer_load_dword v38, off, s[0:3], s32 offset:68
+; SI-NEXT: v_lshlrev_b32_e32 v58, 16, v1
; SI-NEXT: v_lshlrev_b32_e32 v55, 16, v3
-; SI-NEXT: v_lshlrev_b32_e32 v50, 16, v5
-; SI-NEXT: v_lshlrev_b32_e32 v43, 16, v7
-; SI-NEXT: v_lshlrev_b32_e32 v41, 16, v9
-; SI-NEXT: v_lshlrev_b32_e32 v40, 16, v11
+; SI-NEXT: s_waitcnt expcnt(2)
+; SI-NEXT: v_lshlrev_b32_e32 v61, 16, v5
+; SI-NEXT: v_lshlrev_b32_e32 v50, 16, v7
+; SI-NEXT: v_lshlrev_b32_e32 v45, 16, v9
+; SI-NEXT: v_lshlrev_b32_e32 v41, 16, v11
; SI-NEXT: v_lshlrev_b32_e32 v13, 16, v13
-; SI-NEXT: v_lshlrev_b32_e32 v48, 16, v15
+; SI-NEXT: v_lshlrev_b32_e32 v40, 16, v15
; SI-NEXT: v_lshlrev_b32_e32 v15, 16, v17
; SI-NEXT: v_lshlrev_b32_e32 v37, 16, v19
; SI-NEXT: v_lshlrev_b32_e32 v17, 16, v21
@@ -76609,23 +76631,23 @@ define inreg <32 x float> @bitcast_v64i16_to_v32f32_scalar(<64 x i16> inreg %a,
; SI-NEXT: s_waitcnt vmcnt(9)
; SI-NEXT: v_lshlrev_b32_e32 v54, 16, v12
; SI-NEXT: s_waitcnt vmcnt(7)
-; SI-NEXT: v_lshlrev_b32_e32 v5, 16, v38
+; SI-NEXT: v_lshlrev_b32_e32 v5, 16, v48
; SI-NEXT: s_waitcnt vmcnt(5) expcnt(1)
; SI-NEXT: v_lshlrev_b32_e32 v62, 16, v36
; SI-NEXT: s_waitcnt vmcnt(3)
; SI-NEXT: v_lshlrev_b32_e32 v3, 16, v34
; SI-NEXT: s_waitcnt vmcnt(1)
-; SI-NEXT: v_lshlrev_b32_e32 v34, 16, v32
-; SI-NEXT: buffer_store_dword v34, off, s[0:3], s32 offset:144 ; 4-byte Folded Spill
+; SI-NEXT: v_lshlrev_b32_e32 v48, 16, v32
+; SI-NEXT: buffer_store_dword v48, off, s[0:3], s32 offset:144 ; 4-byte Folded Spill
; SI-NEXT: buffer_store_dword v3, off, s[0:3], s32 offset:148 ; 4-byte Folded Spill
; SI-NEXT: buffer_store_dword v62, off, s[0:3], s32 offset:152 ; 4-byte Folded Spill
; SI-NEXT: buffer_store_dword v5, off, s[0:3], s32 offset:156 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt vmcnt(4)
-; SI-NEXT: buffer_store_dword v58, off, s[0:3], s32 offset:160 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v38, off, s[0:3], s32 offset:160 ; 4-byte Folded Spill
; SI-NEXT: buffer_store_dword v54, off, s[0:3], s32 offset:164 ; 4-byte Folded Spill
-; SI-NEXT: buffer_store_dword v27, off, s[0:3], s32 offset:168 ; 4-byte Folded Spill
-; SI-NEXT: buffer_store_dword v59, off, s[0:3], s32 offset:172 ; 4-byte Folded Spill
-; SI-NEXT: buffer_store_dword v45, off, s[0:3], s32 offset:176 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v43, off, s[0:3], s32 offset:168 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v27, off, s[0:3], s32 offset:172 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v59, off, s[0:3], s32 offset:176 ; 4-byte Folded Spill
; SI-NEXT: buffer_store_dword v25, off, s[0:3], s32 offset:180 ; 4-byte Folded Spill
; SI-NEXT: buffer_store_dword v47, off, s[0:3], s32 offset:184 ; 4-byte Folded Spill
; SI-NEXT: buffer_store_dword v60, off, s[0:3], s32 offset:188 ; 4-byte Folded Spill
@@ -76638,45 +76660,46 @@ define inreg <32 x float> @bitcast_v64i16_to_v32f32_scalar(<64 x i16> inreg %a,
; SI-NEXT: buffer_store_dword v19, off, s[0:3], s32 offset:216 ; 4-byte Folded Spill
; SI-NEXT: buffer_store_dword v35, off, s[0:3], s32 offset:220 ; 4-byte Folded Spill
; SI-NEXT: buffer_store_dword v22, off, s[0:3], s32 offset:224 ; 4-byte Folded Spill
-; SI-NEXT: buffer_store_dword v17, off, s[0:3], s32 offset:228 ; 4-byte Folded Spill
-; SI-NEXT: buffer_store_dword v0, off, s[0:3], s32 offset:232 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v0, off, s[0:3], s32 offset:228 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v17, off, s[0:3], s32 offset:232 ; 4-byte Folded Spill
; SI-NEXT: buffer_store_dword v37, off, s[0:3], s32 offset:236 ; 4-byte Folded Spill
; SI-NEXT: buffer_store_dword v18, off, s[0:3], s32 offset:240 ; 4-byte Folded Spill
; SI-NEXT: s_cbranch_scc0 .LBB51_4
; SI-NEXT: ; %bb.1: ; %cmp.false
-; SI-NEXT: s_waitcnt expcnt(2)
+; SI-NEXT: s_waitcnt expcnt(3)
; SI-NEXT: v_and_b32_e32 v0, 0xffff, v0
-; SI-NEXT: v_or_b32_e32 v7, v0, v61
+; SI-NEXT: v_or_b32_e32 v7, v0, v58
; SI-NEXT: v_and_b32_e32 v0, 0xffff, v4
-; SI-NEXT: v_or_b32_e32 v9, v0, v50
+; SI-NEXT: v_or_b32_e32 v9, v0, v61
; SI-NEXT: v_and_b32_e32 v0, 0xffff, v6
-; SI-NEXT: v_or_b32_e32 v10, v0, v43
+; SI-NEXT: v_or_b32_e32 v10, v0, v50
; SI-NEXT: v_and_b32_e32 v0, 0xffff, v57
-; SI-NEXT: v_or_b32_e32 v11, v0, v41
+; SI-NEXT: v_or_b32_e32 v11, v0, v45
; SI-NEXT: v_and_b32_e32 v0, 0xffff, v56
-; SI-NEXT: v_or_b32_e32 v12, v0, v40
+; SI-NEXT: v_or_b32_e32 v12, v0, v41
; SI-NEXT: v_and_b32_e32 v0, 0xffff, v49
-; SI-NEXT: v_mov_b32_e32 v52, v57
-; SI-NEXT: v_mov_b32_e32 v57, v40
-; SI-NEXT: v_mov_b32_e32 v40, v49
-; SI-NEXT: v_mov_b32_e32 v49, v13
+; SI-NEXT: v_mov_b32_e32 v36, v41
+; SI-NEXT: v_mov_b32_e32 v41, v13
; SI-NEXT: v_or_b32_e32 v13, v0, v13
; SI-NEXT: v_and_b32_e32 v0, 0xffff, v14
-; SI-NEXT: v_mov_b32_e32 v36, v41
-; SI-NEXT: v_mov_b32_e32 v41, v14
-; SI-NEXT: v_or_b32_e32 v14, v0, v48
-; SI-NEXT: v_and_b32_e32 v0, 0xffff, v16
; SI-NEXT: v_mov_b32_e32 v51, v50
-; SI-NEXT: v_mov_b32_e32 v50, v43
-; SI-NEXT: v_mov_b32_e32 v43, v48
-; SI-NEXT: v_mov_b32_e32 v48, v15
+; SI-NEXT: v_mov_b32_e32 v50, v45
+; SI-NEXT: v_mov_b32_e32 v45, v14
+; SI-NEXT: v_or_b32_e32 v14, v0, v40
+; SI-NEXT: v_and_b32_e32 v0, 0xffff, v16
+; SI-NEXT: v_mov_b32_e32 v52, v57
+; SI-NEXT: v_mov_b32_e32 v57, v49
+; SI-NEXT: v_mov_b32_e32 v49, v40
+; SI-NEXT: v_mov_b32_e32 v40, v15
; SI-NEXT: v_or_b32_e32 v15, v0, v15
; SI-NEXT: v_and_b32_e32 v0, 0xffff, v18
-; SI-NEXT: v_mov_b32_e32 v38, v61
+; SI-NEXT: v_mov_b32_e32 v34, v58
+; SI-NEXT: v_mov_b32_e32 v58, v61
; SI-NEXT: v_mov_b32_e32 v61, v56
; SI-NEXT: v_mov_b32_e32 v56, v16
; SI-NEXT: v_or_b32_e32 v16, v0, v37
; SI-NEXT: v_and_b32_e32 v0, 0xffff, v20
+; SI-NEXT: s_waitcnt expcnt(2)
; SI-NEXT: v_or_b32_e32 v17, v0, v17
; SI-NEXT: v_and_b32_e32 v0, 0xffff, v22
; SI-NEXT: s_waitcnt expcnt(0)
@@ -76710,7 +76733,7 @@ define inreg <32 x float> @bitcast_v64i16_to_v32f32_scalar(<64 x i16> inreg %a,
; SI-NEXT: s_and_b32 s6, s20, 0xffff
; SI-NEXT: s_lshl_b32 s7, s21, 16
; SI-NEXT: v_or_b32_e32 v26, v0, v26
-; SI-NEXT: v_and_b32_e32 v0, 0xffff, v45
+; SI-NEXT: v_and_b32_e32 v0, 0xffff, v59
; SI-NEXT: s_or_b32 s6, s6, s7
; SI-NEXT: s_and_b32 s7, s22, 0xffff
; SI-NEXT: s_lshl_b32 s8, s23, 16
@@ -76721,7 +76744,7 @@ define inreg <32 x float> @bitcast_v64i16_to_v32f32_scalar(<64 x i16> inreg %a,
; SI-NEXT: s_lshl_b32 s9, s25, 16
; SI-NEXT: v_mov_b32_e32 v33, v28
; SI-NEXT: v_or_b32_e32 v28, v0, v5
-; SI-NEXT: v_and_b32_e32 v0, 0xffff, v59
+; SI-NEXT: v_and_b32_e32 v0, 0xffff, v43
; SI-NEXT: s_or_b32 s8, s8, s9
; SI-NEXT: s_and_b32 s9, s26, 0xffff
; SI-NEXT: s_lshl_b32 s10, s27, 16
@@ -76733,7 +76756,7 @@ define inreg <32 x float> @bitcast_v64i16_to_v32f32_scalar(<64 x i16> inreg %a,
; SI-NEXT: s_lshl_b32 s11, s29, 16
; SI-NEXT: v_and_b32_e32 v1, 0xffff, v2
; SI-NEXT: v_or_b32_e32 v30, v0, v3
-; SI-NEXT: v_and_b32_e32 v0, 0xffff, v58
+; SI-NEXT: v_and_b32_e32 v0, 0xffff, v38
; SI-NEXT: s_or_b32 s10, s10, s11
; SI-NEXT: v_mov_b32_e32 v63, v2
; SI-NEXT: v_mov_b32_e32 v32, v55
@@ -76741,9 +76764,9 @@ define inreg <32 x float> @bitcast_v64i16_to_v32f32_scalar(<64 x i16> inreg %a,
; SI-NEXT: v_mov_b32_e32 v55, v4
; SI-NEXT: v_mov_b32_e32 v53, v6
; SI-NEXT: v_mov_b32_e32 v47, v46
-; SI-NEXT: v_mov_b32_e32 v45, v44
-; SI-NEXT: v_mov_b32_e32 v59, v42
-; SI-NEXT: v_or_b32_e32 v31, v0, v34
+; SI-NEXT: v_mov_b32_e32 v59, v44
+; SI-NEXT: v_mov_b32_e32 v43, v42
+; SI-NEXT: v_or_b32_e32 v31, v0, v48
; SI-NEXT: v_mov_b32_e32 v0, s4
; SI-NEXT: v_mov_b32_e32 v1, s5
; SI-NEXT: v_mov_b32_e32 v2, s6
@@ -76753,12 +76776,13 @@ define inreg <32 x float> @bitcast_v64i16_to_v32f32_scalar(<64 x i16> inreg %a,
; SI-NEXT: v_mov_b32_e32 v6, s10
; SI-NEXT: s_cbranch_execnz .LBB51_3
; SI-NEXT: .LBB51_2: ; %cmp.true
-; SI-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:232 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:228 ; 4-byte Folded Reload
; SI-NEXT: v_add_i32_e32 v1, vcc, 3, v63
; SI-NEXT: v_and_b32_e32 v1, 0xffff, v1
; SI-NEXT: v_or_b32_e32 v1, v32, v1
; SI-NEXT: v_add_i32_e32 v8, vcc, 0x30000, v1
; SI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:236 ; 4-byte Folded Reload
+; SI-NEXT: v_mov_b32_e32 v38, v43
; SI-NEXT: s_add_i32 s16, s16, 3
; SI-NEXT: s_and_b32 s4, s16, 0xffff
; SI-NEXT: s_lshl_b32 s5, s17, 16
@@ -76802,42 +76826,42 @@ define inreg <32 x float> @bitcast_v64i16_to_v32f32_scalar(<64 x i16> inreg %a,
; SI-NEXT: s_waitcnt vmcnt(1)
; SI-NEXT: v_add_i32_e32 v0, vcc, 3, v0
; SI-NEXT: v_and_b32_e32 v0, 0xffff, v0
-; SI-NEXT: v_or_b32_e32 v0, v38, v0
+; SI-NEXT: v_or_b32_e32 v0, v34, v0
; SI-NEXT: v_add_i32_e32 v7, vcc, 0x30000, v0
; SI-NEXT: v_add_i32_e32 v0, vcc, 3, v55
; SI-NEXT: v_and_b32_e32 v0, 0xffff, v0
-; SI-NEXT: v_or_b32_e32 v0, v51, v0
+; SI-NEXT: v_or_b32_e32 v0, v58, v0
; SI-NEXT: v_add_i32_e32 v9, vcc, 0x30000, v0
; SI-NEXT: v_add_i32_e32 v0, vcc, 3, v53
; SI-NEXT: v_and_b32_e32 v0, 0xffff, v0
-; SI-NEXT: v_or_b32_e32 v0, v50, v0
+; SI-NEXT: v_or_b32_e32 v0, v51, v0
; SI-NEXT: v_add_i32_e32 v10, vcc, 0x30000, v0
; SI-NEXT: v_add_i32_e32 v0, vcc, 3, v52
; SI-NEXT: v_and_b32_e32 v0, 0xffff, v0
-; SI-NEXT: v_or_b32_e32 v0, v36, v0
+; SI-NEXT: v_or_b32_e32 v0, v50, v0
; SI-NEXT: v_add_i32_e32 v11, vcc, 0x30000, v0
; SI-NEXT: v_add_i32_e32 v0, vcc, 3, v61
; SI-NEXT: v_and_b32_e32 v0, 0xffff, v0
-; SI-NEXT: v_or_b32_e32 v0, v57, v0
+; SI-NEXT: v_or_b32_e32 v0, v36, v0
; SI-NEXT: v_add_i32_e32 v12, vcc, 0x30000, v0
-; SI-NEXT: v_add_i32_e32 v0, vcc, 3, v40
+; SI-NEXT: v_add_i32_e32 v0, vcc, 3, v57
; SI-NEXT: v_and_b32_e32 v0, 0xffff, v0
-; SI-NEXT: v_or_b32_e32 v0, v49, v0
+; SI-NEXT: v_or_b32_e32 v0, v41, v0
; SI-NEXT: v_add_i32_e32 v13, vcc, 0x30000, v0
-; SI-NEXT: v_add_i32_e32 v0, vcc, 3, v41
+; SI-NEXT: v_add_i32_e32 v0, vcc, 3, v45
; SI-NEXT: v_and_b32_e32 v0, 0xffff, v0
-; SI-NEXT: v_or_b32_e32 v0, v43, v0
+; SI-NEXT: v_or_b32_e32 v0, v49, v0
; SI-NEXT: v_add_i32_e32 v14, vcc, 0x30000, v0
; SI-NEXT: v_add_i32_e32 v0, vcc, 3, v56
; SI-NEXT: v_and_b32_e32 v0, 0xffff, v0
-; SI-NEXT: v_or_b32_e32 v0, v48, v0
+; SI-NEXT: v_or_b32_e32 v0, v40, v0
; SI-NEXT: v_add_i32_e32 v15, vcc, 0x30000, v0
; SI-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:240 ; 4-byte Folded Reload
; SI-NEXT: s_waitcnt vmcnt(0)
; SI-NEXT: v_add_i32_e32 v0, vcc, 3, v0
; SI-NEXT: v_and_b32_e32 v0, 0xffff, v0
; SI-NEXT: v_or_b32_e32 v0, v1, v0
-; SI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:228 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:232 ; 4-byte Folded Reload
; SI-NEXT: v_add_i32_e32 v16, vcc, 0x30000, v0
; SI-NEXT: v_add_i32_e32 v0, vcc, 3, v37
; SI-NEXT: v_and_b32_e32 v0, 0xffff, v0
@@ -76899,7 +76923,7 @@ define inreg <32 x float> @bitcast_v64i16_to_v32f32_scalar(<64 x i16> inreg %a,
; SI-NEXT: v_add_i32_e32 v0, vcc, 3, v0
; SI-NEXT: v_and_b32_e32 v0, 0xffff, v0
; SI-NEXT: v_or_b32_e32 v0, v1, v0
-; SI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:168 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:172 ; 4-byte Folded Reload
; SI-NEXT: v_add_i32_e32 v25, vcc, 0x30000, v0
; SI-NEXT: v_add_i32_e32 v0, vcc, 3, v47
; SI-NEXT: v_and_b32_e32 v0, 0xffff, v0
@@ -76915,12 +76939,12 @@ define inreg <32 x float> @bitcast_v64i16_to_v32f32_scalar(<64 x i16> inreg %a,
; SI-NEXT: v_or_b32_e32 v0, v1, v0
; SI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:156 ; 4-byte Folded Reload
; SI-NEXT: v_add_i32_e32 v27, vcc, 0x30000, v0
-; SI-NEXT: v_add_i32_e32 v0, vcc, 3, v45
+; SI-NEXT: v_add_i32_e32 v0, vcc, 3, v59
; SI-NEXT: v_and_b32_e32 v0, 0xffff, v0
; SI-NEXT: s_waitcnt vmcnt(0)
; SI-NEXT: v_or_b32_e32 v0, v1, v0
; SI-NEXT: v_add_i32_e32 v28, vcc, 0x30000, v0
-; SI-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:172 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:168 ; 4-byte Folded Reload
; SI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:152 ; 4-byte Folded Reload
; SI-NEXT: s_waitcnt vmcnt(1)
; SI-NEXT: v_add_i32_e32 v0, vcc, 3, v0
@@ -76929,7 +76953,7 @@ define inreg <32 x float> @bitcast_v64i16_to_v32f32_scalar(<64 x i16> inreg %a,
; SI-NEXT: v_or_b32_e32 v0, v1, v0
; SI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:148 ; 4-byte Folded Reload
; SI-NEXT: v_add_i32_e32 v29, vcc, 0x30000, v0
-; SI-NEXT: v_add_i32_e32 v0, vcc, 3, v59
+; SI-NEXT: v_add_i32_e32 v0, vcc, 3, v38
; SI-NEXT: v_and_b32_e32 v0, 0xffff, v0
; SI-NEXT: s_waitcnt vmcnt(0)
; SI-NEXT: v_or_b32_e32 v0, v1, v0
@@ -76964,26 +76988,26 @@ define inreg <32 x float> @bitcast_v64i16_to_v32f32_scalar(<64 x i16> inreg %a,
; SI-NEXT: s_waitcnt vmcnt(0) expcnt(0)
; SI-NEXT: s_setpc_b64 s[30:31]
; SI-NEXT: .LBB51_4:
-; SI-NEXT: v_mov_b32_e32 v38, v61
+; SI-NEXT: v_mov_b32_e32 v34, v58
; SI-NEXT: v_mov_b32_e32 v32, v55
+; SI-NEXT: v_mov_b32_e32 v58, v61
; SI-NEXT: v_mov_b32_e32 v63, v2
; SI-NEXT: v_mov_b32_e32 v55, v4
; SI-NEXT: v_mov_b32_e32 v53, v6
; SI-NEXT: v_mov_b32_e32 v52, v57
; SI-NEXT: v_mov_b32_e32 v51, v50
; SI-NEXT: v_mov_b32_e32 v61, v56
-; SI-NEXT: v_mov_b32_e32 v50, v43
+; SI-NEXT: v_mov_b32_e32 v50, v45
; SI-NEXT: v_mov_b32_e32 v36, v41
-; SI-NEXT: v_mov_b32_e32 v57, v40
-; SI-NEXT: v_mov_b32_e32 v40, v49
-; SI-NEXT: v_mov_b32_e32 v49, v13
-; SI-NEXT: v_mov_b32_e32 v43, v48
-; SI-NEXT: v_mov_b32_e32 v48, v15
-; SI-NEXT: v_mov_b32_e32 v41, v14
+; SI-NEXT: v_mov_b32_e32 v41, v13
+; SI-NEXT: v_mov_b32_e32 v57, v49
+; SI-NEXT: v_mov_b32_e32 v49, v40
+; SI-NEXT: v_mov_b32_e32 v40, v15
+; SI-NEXT: v_mov_b32_e32 v45, v14
; SI-NEXT: v_mov_b32_e32 v56, v16
; SI-NEXT: v_mov_b32_e32 v47, v46
-; SI-NEXT: v_mov_b32_e32 v45, v44
-; SI-NEXT: v_mov_b32_e32 v59, v42
+; SI-NEXT: v_mov_b32_e32 v59, v44
+; SI-NEXT: v_mov_b32_e32 v43, v42
; SI-NEXT: s_waitcnt expcnt(1)
; SI-NEXT: v_mov_b32_e32 v37, v20
; SI-NEXT: v_mov_b32_e32 v39, v23
@@ -85923,7 +85947,7 @@ define inreg <128 x i8> @bitcast_v16i64_to_v128i8_scalar(<16 x i64> inreg %a, i3
; GFX11-TRUE16-NEXT: v_readfirstlane_b32 s5, v14
; GFX11-TRUE16-NEXT: v_writelane_b32 v16, s37, 5
; GFX11-TRUE16-NEXT: v_writelane_b32 v17, s101, 5
-; GFX11-TRUE16-NEXT: s_mov_b32 s57, 0
+; GFX11-TRUE16-NEXT: s_mov_b32 vcc_hi, 0
; GFX11-TRUE16-NEXT: s_and_b32 s42, vcc_lo, exec_lo
; GFX11-TRUE16-NEXT: ; implicit-def: $vgpr18 : SGPR spill to VGPR lane
; GFX11-TRUE16-NEXT: ; implicit-def: $vgpr19 : SGPR spill to VGPR lane
@@ -86101,7 +86125,7 @@ define inreg <128 x i8> @bitcast_v16i64_to_v128i8_scalar(<16 x i64> inreg %a, i3
; GFX11-TRUE16-NEXT: s_delay_alu instid0(SALU_CYCLE_1)
; GFX11-TRUE16-NEXT: v_writelane_b32 v18, s42, 0
; GFX11-TRUE16-NEXT: s_lshr_b32 s42, s18, 8
-; GFX11-TRUE16-NEXT: s_and_not1_b32 vcc_lo, exec_lo, s57
+; GFX11-TRUE16-NEXT: s_and_not1_b32 vcc_lo, exec_lo, vcc_hi
; GFX11-TRUE16-NEXT: s_cbranch_vccnz .LBB57_3
; GFX11-TRUE16-NEXT: .LBB57_2: ; %cmp.true
; GFX11-TRUE16-NEXT: s_add_u32 s0, s0, 3
@@ -86774,9 +86798,9 @@ define inreg <128 x i8> @bitcast_v16i64_to_v128i8_scalar(<16 x i64> inreg %a, i3
; GFX11-TRUE16-NEXT: s_waitcnt vmcnt(0)
; GFX11-TRUE16-NEXT: s_setpc_b64 s[30:31]
; GFX11-TRUE16-NEXT: .LBB57_4:
-; GFX11-TRUE16-NEXT: ; implicit-def: $sgpr58_lo16
-; GFX11-TRUE16-NEXT: v_writelane_b32 v18, s58, 0
-; GFX11-TRUE16-NEXT: ; implicit-def: $sgpr58_lo16
+; GFX11-TRUE16-NEXT: ; implicit-def: $sgpr57_lo16
+; GFX11-TRUE16-NEXT: v_writelane_b32 v18, s57, 0
+; GFX11-TRUE16-NEXT: ; implicit-def: $sgpr57_lo16
; GFX11-TRUE16-NEXT: ; implicit-def: $sgpr101_lo16
; GFX11-TRUE16-NEXT: ; implicit-def: $sgpr100_lo16
; GFX11-TRUE16-NEXT: ; implicit-def: $sgpr56_lo16
@@ -86834,72 +86858,72 @@ define inreg <128 x i8> @bitcast_v16i64_to_v128i8_scalar(<16 x i64> inreg %a, i3
; GFX11-TRUE16-NEXT: ; implicit-def: $sgpr72_lo16
; GFX11-TRUE16-NEXT: ; implicit-def: $sgpr62_lo16
; GFX11-TRUE16-NEXT: ; implicit-def: $sgpr60_lo16
-; GFX11-TRUE16-NEXT: v_writelane_b32 v18, s58, 1
-; GFX11-TRUE16-NEXT: ; implicit-def: $sgpr58_lo16
-; GFX11-TRUE16-NEXT: v_writelane_b32 v18, s58, 2
-; GFX11-TRUE16-NEXT: ; implicit-def: $sgpr58_lo16
-; GFX11-TRUE16-NEXT: v_writelane_b32 v18, s58, 3
-; GFX11-TRUE16-NEXT: ; implicit-def: $sgpr58_lo16
-; GFX11-TRUE16-NEXT: v_writelane_b32 v18, s58, 4
-; GFX11-TRUE16-NEXT: ; implicit-def: $sgpr58_lo16
-; GFX11-TRUE16-NEXT: v_writelane_b32 v18, s58, 5
-; GFX11-TRUE16-NEXT: ; implicit-def: $sgpr58_lo16
-; GFX11-TRUE16-NEXT: v_writelane_b32 v18, s58, 6
-; GFX11-TRUE16-NEXT: ; implicit-def: $sgpr58_lo16
-; GFX11-TRUE16-NEXT: v_writelane_b32 v18, s58, 7
-; GFX11-TRUE16-NEXT: ; implicit-def: $sgpr58_lo16
-; GFX11-TRUE16-NEXT: v_writelane_b32 v18, s58, 8
-; GFX11-TRUE16-NEXT: ; implicit-def: $sgpr58_lo16
-; GFX11-TRUE16-NEXT: v_writelane_b32 v18, s58, 9
-; GFX11-TRUE16-NEXT: ; implicit-def: $sgpr58_lo16
-; GFX11-TRUE16-NEXT: v_writelane_b32 v18, s58, 10
-; GFX11-TRUE16-NEXT: ; implicit-def: $sgpr58_lo16
-; GFX11-TRUE16-NEXT: v_writelane_b32 v18, s58, 11
-; GFX11-TRUE16-NEXT: ; implicit-def: $sgpr58_lo16
-; GFX11-TRUE16-NEXT: v_writelane_b32 v18, s58, 12
-; GFX11-TRUE16-NEXT: ; implicit-def: $sgpr58_lo16
-; GFX11-TRUE16-NEXT: v_writelane_b32 v18, s58, 13
-; GFX11-TRUE16-NEXT: ; implicit-def: $sgpr58_lo16
-; GFX11-TRUE16-NEXT: v_writelane_b32 v18, s58, 14
-; GFX11-TRUE16-NEXT: ; implicit-def: $sgpr58_lo16
-; GFX11-TRUE16-NEXT: v_writelane_b32 v18, s58, 15
-; GFX11-TRUE16-NEXT: ; implicit-def: $sgpr58_lo16
-; GFX11-TRUE16-NEXT: v_writelane_b32 v18, s58, 16
-; GFX11-TRUE16-NEXT: ; implicit-def: $sgpr58_lo16
-; GFX11-TRUE16-NEXT: v_writelane_b32 v18, s58, 17
-; GFX11-TRUE16-NEXT: ; implicit-def: $sgpr58_lo16
-; GFX11-TRUE16-NEXT: v_writelane_b32 v18, s58, 18
-; GFX11-TRUE16-NEXT: ; implicit-def: $sgpr58_lo16
-; GFX11-TRUE16-NEXT: v_writelane_b32 v18, s58, 19
-; GFX11-TRUE16-NEXT: ; implicit-def: $sgpr58_lo16
-; GFX11-TRUE16-NEXT: v_writelane_b32 v18, s58, 20
-; GFX11-TRUE16-NEXT: ; implicit-def: $sgpr58_lo16
-; GFX11-TRUE16-NEXT: v_writelane_b32 v18, s58, 21
-; GFX11-TRUE16-NEXT: ; implicit-def: $sgpr58_lo16
-; GFX11-TRUE16-NEXT: v_writelane_b32 v18, s58, 22
-; GFX11-TRUE16-NEXT: ; implicit-def: $sgpr58_lo16
-; GFX11-TRUE16-NEXT: v_writelane_b32 v18, s58, 23
-; GFX11-TRUE16-NEXT: ; implicit-def: $sgpr58_lo16
-; GFX11-TRUE16-NEXT: v_writelane_b32 v18, s58, 24
-; GFX11-TRUE16-NEXT: ; implicit-def: $sgpr58_lo16
-; GFX11-TRUE16-NEXT: v_writelane_b32 v18, s58, 25
-; GFX11-TRUE16-NEXT: ; implicit-def: $sgpr58_lo16
-; GFX11-TRUE16-NEXT: v_writelane_b32 v18, s58, 26
-; GFX11-TRUE16-NEXT: ; implicit-def: $sgpr58_lo16
-; GFX11-TRUE16-NEXT: v_writelane_b32 v18, s58, 27
-; GFX11-TRUE16-NEXT: ; implicit-def: $sgpr58_lo16
-; GFX11-TRUE16-NEXT: v_writelane_b32 v18, s58, 28
-; GFX11-TRUE16-NEXT: ; implicit-def: $sgpr58_lo16
-; GFX11-TRUE16-NEXT: v_writelane_b32 v18, s58, 29
-; GFX11-TRUE16-NEXT: ; implicit-def: $sgpr58_lo16
-; GFX11-TRUE16-NEXT: v_writelane_b32 v18, s58, 30
-; GFX11-TRUE16-NEXT: ; implicit-def: $sgpr58_lo16
-; GFX11-TRUE16-NEXT: v_writelane_b32 v18, s58, 31
-; GFX11-TRUE16-NEXT: ; implicit-def: $sgpr58_lo16
-; GFX11-TRUE16-NEXT: v_writelane_b32 v19, s58, 0
-; GFX11-TRUE16-NEXT: ; implicit-def: $sgpr58_lo16
-; GFX11-TRUE16-NEXT: v_writelane_b32 v19, s58, 1
; GFX11-TRUE16-NEXT: ; implicit-def: $sgpr58_lo16
+; GFX11-TRUE16-NEXT: v_writelane_b32 v18, s57, 1
+; GFX11-TRUE16-NEXT: ; implicit-def: $sgpr57_lo16
+; GFX11-TRUE16-NEXT: v_writelane_b32 v18, s57, 2
+; GFX11-TRUE16-NEXT: ; implicit-def: $sgpr57_lo16
+; GFX11-TRUE16-NEXT: v_writelane_b32 v18, s57, 3
+; GFX11-TRUE16-NEXT: ; implicit-def: $sgpr57_lo16
+; GFX11-TRUE16-NEXT: v_writelane_b32 v18, s57, 4
+; GFX11-TRUE16-NEXT: ; implicit-def: $sgpr57_lo16
+; GFX11-TRUE16-NEXT: v_writelane_b32 v18, s57, 5
+; GFX11-TRUE16-NEXT: ; implicit-def: $sgpr57_lo16
+; GFX11-TRUE16-NEXT: v_writelane_b32 v18, s57, 6
+; GFX11-TRUE16-NEXT: ; implicit-def: $sgpr57_lo16
+; GFX11-TRUE16-NEXT: v_writelane_b32 v18, s57, 7
+; GFX11-TRUE16-NEXT: ; implicit-def: $sgpr57_lo16
+; GFX11-TRUE16-NEXT: v_writelane_b32 v18, s57, 8
+; GFX11-TRUE16-NEXT: ; implicit-def: $sgpr57_lo16
+; GFX11-TRUE16-NEXT: v_writelane_b32 v18, s57, 9
+; GFX11-TRUE16-NEXT: ; implicit-def: $sgpr57_lo16
+; GFX11-TRUE16-NEXT: v_writelane_b32 v18, s57, 10
+; GFX11-TRUE16-NEXT: ; implicit-def: $sgpr57_lo16
+; GFX11-TRUE16-NEXT: v_writelane_b32 v18, s57, 11
+; GFX11-TRUE16-NEXT: ; implicit-def: $sgpr57_lo16
+; GFX11-TRUE16-NEXT: v_writelane_b32 v18, s57, 12
+; GFX11-TRUE16-NEXT: ; implicit-def: $sgpr57_lo16
+; GFX11-TRUE16-NEXT: v_writelane_b32 v18, s57, 13
+; GFX11-TRUE16-NEXT: ; implicit-def: $sgpr57_lo16
+; GFX11-TRUE16-NEXT: v_writelane_b32 v18, s57, 14
+; GFX11-TRUE16-NEXT: ; implicit-def: $sgpr57_lo16
+; GFX11-TRUE16-NEXT: v_writelane_b32 v18, s57, 15
+; GFX11-TRUE16-NEXT: ; implicit-def: $sgpr57_lo16
+; GFX11-TRUE16-NEXT: v_writelane_b32 v18, s57, 16
+; GFX11-TRUE16-NEXT: ; implicit-def: $sgpr57_lo16
+; GFX11-TRUE16-NEXT: v_writelane_b32 v18, s57, 17
+; GFX11-TRUE16-NEXT: ; implicit-def: $sgpr57_lo16
+; GFX11-TRUE16-NEXT: v_writelane_b32 v18, s57, 18
+; GFX11-TRUE16-NEXT: ; implicit-def: $sgpr57_lo16
+; GFX11-TRUE16-NEXT: v_writelane_b32 v18, s57, 19
+; GFX11-TRUE16-NEXT: ; implicit-def: $sgpr57_lo16
+; GFX11-TRUE16-NEXT: v_writelane_b32 v18, s57, 20
+; GFX11-TRUE16-NEXT: ; implicit-def: $sgpr57_lo16
+; GFX11-TRUE16-NEXT: v_writelane_b32 v18, s57, 21
+; GFX11-TRUE16-NEXT: ; implicit-def: $sgpr57_lo16
+; GFX11-TRUE16-NEXT: v_writelane_b32 v18, s57, 22
+; GFX11-TRUE16-NEXT: ; implicit-def: $sgpr57_lo16
+; GFX11-TRUE16-NEXT: v_writelane_b32 v18, s57, 23
+; GFX11-TRUE16-NEXT: ; implicit-def: $sgpr57_lo16
+; GFX11-TRUE16-NEXT: v_writelane_b32 v18, s57, 24
+; GFX11-TRUE16-NEXT: ; implicit-def: $sgpr57_lo16
+; GFX11-TRUE16-NEXT: v_writelane_b32 v18, s57, 25
+; GFX11-TRUE16-NEXT: ; implicit-def: $sgpr57_lo16
+; GFX11-TRUE16-NEXT: v_writelane_b32 v18, s57, 26
+; GFX11-TRUE16-NEXT: ; implicit-def: $sgpr57_lo16
+; GFX11-TRUE16-NEXT: v_writelane_b32 v18, s57, 27
+; GFX11-TRUE16-NEXT: ; implicit-def: $sgpr57_lo16
+; GFX11-TRUE16-NEXT: v_writelane_b32 v18, s57, 28
+; GFX11-TRUE16-NEXT: ; implicit-def: $sgpr57_lo16
+; GFX11-TRUE16-NEXT: v_writelane_b32 v18, s57, 29
+; GFX11-TRUE16-NEXT: ; implicit-def: $sgpr57_lo16
+; GFX11-TRUE16-NEXT: v_writelane_b32 v18, s57, 30
+; GFX11-TRUE16-NEXT: ; implicit-def: $sgpr57_lo16
+; GFX11-TRUE16-NEXT: v_writelane_b32 v18, s57, 31
+; GFX11-TRUE16-NEXT: ; implicit-def: $sgpr57_lo16
+; GFX11-TRUE16-NEXT: v_writelane_b32 v19, s57, 0
+; GFX11-TRUE16-NEXT: ; implicit-def: $sgpr57_lo16
+; GFX11-TRUE16-NEXT: v_writelane_b32 v19, s57, 1
; GFX11-TRUE16-NEXT: v_writelane_b32 v19, s100, 2
; GFX11-TRUE16-NEXT: v_writelane_b32 v19, s101, 3
; GFX11-TRUE16-NEXT: ; implicit-def: $sgpr100_lo16
@@ -86951,7 +86975,7 @@ define inreg <128 x i8> @bitcast_v16i64_to_v128i8_scalar(<16 x i64> inreg %a, i3
; GFX11-FAKE16-NEXT: v_readfirstlane_b32 s5, v14
; GFX11-FAKE16-NEXT: v_writelane_b32 v16, s37, 5
; GFX11-FAKE16-NEXT: v_writelane_b32 v17, s101, 5
-; GFX11-FAKE16-NEXT: s_mov_b32 s101, 0
+; GFX11-FAKE16-NEXT: s_mov_b32 s44, 0
; GFX11-FAKE16-NEXT: s_and_b32 s42, vcc_lo, exec_lo
; GFX11-FAKE16-NEXT: ; implicit-def: $vgpr19 : SGPR spill to VGPR lane
; GFX11-FAKE16-NEXT: ; implicit-def: $vgpr18 : SGPR spill to VGPR lane
@@ -86984,297 +87008,155 @@ define inreg <128 x i8> @bitcast_v16i64_to_v128i8_scalar(<16 x i64> inreg %a, i3
; GFX11-FAKE16-NEXT: v_writelane_b32 v16, s85, 29
; GFX11-FAKE16-NEXT: v_writelane_b32 v16, s86, 30
; GFX11-FAKE16-NEXT: v_writelane_b32 v16, s87, 31
-; GFX11-FAKE16-NEXT: s_cbranch_scc0 .LBB57_2
+; GFX11-FAKE16-NEXT: s_cbranch_scc0 .LBB57_4
; GFX11-FAKE16-NEXT: ; %bb.1: ; %cmp.false
-; GFX11-FAKE16-NEXT: s_lshr_b32 s43, s25, 8
-; GFX11-FAKE16-NEXT: s_lshr_b64 s[62:63], s[4:5], 24
-; GFX11-FAKE16-NEXT: v_writelane_b32 v19, s43, 16
-; GFX11-FAKE16-NEXT: s_lshr_b32 s43, s24, 16
-; GFX11-FAKE16-NEXT: s_lshr_b32 s104, s5, 24
-; GFX11-FAKE16-NEXT: s_lshr_b32 s102, s5, 16
-; GFX11-FAKE16-NEXT: s_lshr_b32 s103, s5, 8
-; GFX11-FAKE16-NEXT: v_writelane_b32 v19, s43, 17
-; GFX11-FAKE16-NEXT: s_lshr_b32 s43, s24, 8
-; GFX11-FAKE16-NEXT: s_lshr_b32 s57, s4, 16
-; GFX11-FAKE16-NEXT: s_lshr_b32 s47, s4, 8
-; GFX11-FAKE16-NEXT: s_lshr_b32 s46, s7, 24
-; GFX11-FAKE16-NEXT: v_writelane_b32 v19, s43, 18
-; GFX11-FAKE16-NEXT: s_lshr_b32 s43, s23, 24
-; GFX11-FAKE16-NEXT: s_lshr_b32 vcc_hi, s7, 16
-; GFX11-FAKE16-NEXT: s_lshr_b32 s34, s7, 8
-; GFX11-FAKE16-NEXT: s_lshr_b32 s69, s6, 16
-; GFX11-FAKE16-NEXT: v_writelane_b32 v19, s43, 19
-; GFX11-FAKE16-NEXT: s_lshr_b32 s43, s23, 16
-; GFX11-FAKE16-NEXT: s_lshr_b32 s56, s6, 8
-; GFX11-FAKE16-NEXT: s_lshr_b32 s35, s9, 24
-; GFX11-FAKE16-NEXT: s_lshr_b32 s36, s9, 16
-; GFX11-FAKE16-NEXT: v_writelane_b32 v19, s43, 20
-; GFX11-FAKE16-NEXT: s_lshr_b32 s43, s23, 8
-; GFX11-FAKE16-NEXT: s_lshr_b32 s37, s9, 8
-; GFX11-FAKE16-NEXT: s_lshr_b32 s38, s8, 16
-; GFX11-FAKE16-NEXT: s_lshr_b32 s39, s8, 8
-; GFX11-FAKE16-NEXT: v_writelane_b32 v19, s43, 21
-; GFX11-FAKE16-NEXT: s_lshr_b32 s43, s22, 16
-; GFX11-FAKE16-NEXT: s_lshr_b32 s48, s11, 24
-; GFX11-FAKE16-NEXT: s_lshr_b32 s49, s11, 16
-; GFX11-FAKE16-NEXT: s_lshr_b32 s50, s11, 8
-; GFX11-FAKE16-NEXT: v_writelane_b32 v19, s43, 22
-; GFX11-FAKE16-NEXT: s_lshr_b32 s43, s22, 8
-; GFX11-FAKE16-NEXT: s_lshr_b32 s51, s10, 16
-; GFX11-FAKE16-NEXT: s_lshr_b32 s52, s10, 8
-; GFX11-FAKE16-NEXT: s_lshr_b32 s53, s13, 24
-; GFX11-FAKE16-NEXT: v_writelane_b32 v19, s43, 23
-; GFX11-FAKE16-NEXT: s_lshr_b32 s43, s21, 24
-; GFX11-FAKE16-NEXT: s_lshr_b32 s54, s13, 16
-; GFX11-FAKE16-NEXT: s_lshr_b32 s55, s13, 8
-; GFX11-FAKE16-NEXT: s_lshr_b32 s64, s12, 16
-; GFX11-FAKE16-NEXT: v_writelane_b32 v19, s43, 24
-; GFX11-FAKE16-NEXT: s_lshr_b32 s43, s21, 16
-; GFX11-FAKE16-NEXT: s_lshr_b32 s65, s12, 8
-; GFX11-FAKE16-NEXT: s_lshr_b32 s66, s15, 24
-; GFX11-FAKE16-NEXT: s_lshr_b32 s67, s15, 16
-; GFX11-FAKE16-NEXT: v_writelane_b32 v19, s43, 25
-; GFX11-FAKE16-NEXT: s_lshr_b32 s43, s21, 8
-; GFX11-FAKE16-NEXT: s_lshr_b32 s68, s15, 8
-; GFX11-FAKE16-NEXT: s_lshr_b32 s59, s14, 16
-; GFX11-FAKE16-NEXT: s_lshr_b32 s58, s14, 8
-; GFX11-FAKE16-NEXT: v_writelane_b32 v19, s43, 26
-; GFX11-FAKE16-NEXT: s_lshr_b32 s43, s20, 16
-; GFX11-FAKE16-NEXT: s_lshr_b32 s70, s41, 24
-; GFX11-FAKE16-NEXT: s_lshr_b32 s71, s41, 16
-; GFX11-FAKE16-NEXT: s_lshr_b32 s60, s41, 8
-; GFX11-FAKE16-NEXT: v_writelane_b32 v19, s43, 27
-; GFX11-FAKE16-NEXT: s_lshr_b32 s43, s20, 8
-; GFX11-FAKE16-NEXT: s_lshr_b32 s80, s40, 16
-; GFX11-FAKE16-NEXT: s_lshr_b32 s61, s40, 8
-; GFX11-FAKE16-NEXT: s_lshr_b32 s81, s29, 24
-; GFX11-FAKE16-NEXT: v_writelane_b32 v19, s43, 28
-; GFX11-FAKE16-NEXT: s_lshr_b32 s43, s19, 24
-; GFX11-FAKE16-NEXT: s_lshr_b32 s82, s29, 16
-; GFX11-FAKE16-NEXT: s_lshr_b32 s83, s29, 8
-; GFX11-FAKE16-NEXT: s_lshr_b32 s84, s28, 16
-; GFX11-FAKE16-NEXT: v_writelane_b32 v19, s43, 29
-; GFX11-FAKE16-NEXT: s_lshr_b32 s43, s19, 16
-; GFX11-FAKE16-NEXT: s_lshr_b32 s85, s28, 8
-; GFX11-FAKE16-NEXT: s_lshr_b32 s86, s27, 24
-; GFX11-FAKE16-NEXT: s_lshr_b32 s72, s27, 16
-; GFX11-FAKE16-NEXT: v_writelane_b32 v19, s43, 30
-; GFX11-FAKE16-NEXT: s_lshr_b32 s43, s19, 8
-; GFX11-FAKE16-NEXT: s_lshr_b32 s87, s27, 8
-; GFX11-FAKE16-NEXT: s_lshr_b32 s73, s26, 16
-; GFX11-FAKE16-NEXT: s_lshr_b32 s96, s26, 8
-; GFX11-FAKE16-NEXT: v_writelane_b32 v19, s43, 31
-; GFX11-FAKE16-NEXT: s_lshr_b32 s43, s18, 16
-; GFX11-FAKE16-NEXT: s_lshr_b32 s97, s25, 24
-; GFX11-FAKE16-NEXT: v_writelane_b32 v18, s43, 0
-; GFX11-FAKE16-NEXT: s_lshr_b32 s43, s18, 8
-; GFX11-FAKE16-NEXT: v_writelane_b32 v19, s62, 14
-; GFX11-FAKE16-NEXT: s_lshr_b32 s42, s25, 16
-; GFX11-FAKE16-NEXT: s_lshr_b32 s74, s2, 16
-; GFX11-FAKE16-NEXT: v_writelane_b32 v18, s43, 1
-; GFX11-FAKE16-NEXT: s_lshr_b32 s43, s17, 24
-; GFX11-FAKE16-NEXT: v_writelane_b32 v19, s63, 15
-; GFX11-FAKE16-NEXT: s_lshr_b64 s[62:63], s[6:7], 24
+; GFX11-FAKE16-NEXT: s_lshr_b32 s42, s5, 24
+; GFX11-FAKE16-NEXT: s_lshr_b64 s[46:47], s[4:5], 24
+; GFX11-FAKE16-NEXT: v_writelane_b32 v19, s42, 4
+; GFX11-FAKE16-NEXT: s_lshr_b32 s42, s5, 16
+; GFX11-FAKE16-NEXT: s_lshr_b32 s102, s29, 8
+; GFX11-FAKE16-NEXT: s_lshr_b32 s103, s28, 16
+; GFX11-FAKE16-NEXT: s_lshr_b32 s104, s28, 8
+; GFX11-FAKE16-NEXT: v_writelane_b32 v19, s42, 5
+; GFX11-FAKE16-NEXT: s_lshr_b32 s42, s5, 8
+; GFX11-FAKE16-NEXT: s_lshr_b32 vcc_hi, s27, 24
+; GFX11-FAKE16-NEXT: s_lshr_b32 s34, s27, 16
+; GFX11-FAKE16-NEXT: s_lshr_b32 s35, s27, 8
+; GFX11-FAKE16-NEXT: v_writelane_b32 v19, s42, 6
+; GFX11-FAKE16-NEXT: s_lshr_b32 s42, s4, 16
+; GFX11-FAKE16-NEXT: s_lshr_b32 s36, s26, 16
+; GFX11-FAKE16-NEXT: s_lshr_b32 s37, s26, 8
+; GFX11-FAKE16-NEXT: s_lshr_b32 s38, s25, 24
+; GFX11-FAKE16-NEXT: v_writelane_b32 v19, s42, 7
+; GFX11-FAKE16-NEXT: s_lshr_b32 s42, s4, 8
+; GFX11-FAKE16-NEXT: s_lshr_b32 s39, s25, 16
+; GFX11-FAKE16-NEXT: s_lshr_b32 s48, s25, 8
+; GFX11-FAKE16-NEXT: s_lshr_b32 s49, s24, 16
+; GFX11-FAKE16-NEXT: v_writelane_b32 v19, s42, 8
+; GFX11-FAKE16-NEXT: s_lshr_b32 s42, s7, 24
+; GFX11-FAKE16-NEXT: s_lshr_b32 s50, s24, 8
+; GFX11-FAKE16-NEXT: s_lshr_b32 s51, s23, 24
+; GFX11-FAKE16-NEXT: s_lshr_b32 s52, s23, 16
+; GFX11-FAKE16-NEXT: v_writelane_b32 v19, s42, 9
+; GFX11-FAKE16-NEXT: s_lshr_b32 s42, s7, 16
+; GFX11-FAKE16-NEXT: s_lshr_b32 s53, s23, 8
+; GFX11-FAKE16-NEXT: s_lshr_b32 s54, s22, 16
+; GFX11-FAKE16-NEXT: s_lshr_b32 s55, s22, 8
+; GFX11-FAKE16-NEXT: v_writelane_b32 v19, s42, 10
+; GFX11-FAKE16-NEXT: s_lshr_b32 s42, s7, 8
+; GFX11-FAKE16-NEXT: s_lshr_b32 s64, s21, 24
+; GFX11-FAKE16-NEXT: s_lshr_b32 s65, s21, 16
+; GFX11-FAKE16-NEXT: s_lshr_b32 s66, s21, 8
+; GFX11-FAKE16-NEXT: v_writelane_b32 v19, s42, 11
+; GFX11-FAKE16-NEXT: s_lshr_b32 s42, s6, 16
+; GFX11-FAKE16-NEXT: s_lshr_b32 s67, s20, 16
+; GFX11-FAKE16-NEXT: s_lshr_b32 s68, s20, 8
+; GFX11-FAKE16-NEXT: s_lshr_b32 s69, s19, 24
+; GFX11-FAKE16-NEXT: v_writelane_b32 v19, s42, 12
+; GFX11-FAKE16-NEXT: s_lshr_b32 s42, s6, 8
+; GFX11-FAKE16-NEXT: s_lshr_b32 s70, s19, 16
+; GFX11-FAKE16-NEXT: s_lshr_b32 s71, s19, 8
+; GFX11-FAKE16-NEXT: s_lshr_b32 s80, s18, 16
+; GFX11-FAKE16-NEXT: v_writelane_b32 v19, s42, 13
+; GFX11-FAKE16-NEXT: s_lshr_b32 s42, s9, 24
+; GFX11-FAKE16-NEXT: s_lshr_b32 s81, s18, 8
+; GFX11-FAKE16-NEXT: s_lshr_b32 s82, s17, 24
+; GFX11-FAKE16-NEXT: s_lshr_b32 s83, s17, 16
+; GFX11-FAKE16-NEXT: v_writelane_b32 v19, s42, 14
+; GFX11-FAKE16-NEXT: s_lshr_b32 s42, s9, 16
+; GFX11-FAKE16-NEXT: s_lshr_b32 s84, s17, 8
+; GFX11-FAKE16-NEXT: s_lshr_b32 s85, s16, 16
+; GFX11-FAKE16-NEXT: s_lshr_b32 s86, s16, 8
+; GFX11-FAKE16-NEXT: v_writelane_b32 v19, s42, 15
+; GFX11-FAKE16-NEXT: s_lshr_b32 s42, s9, 8
+; GFX11-FAKE16-NEXT: s_lshr_b32 s87, s3, 24
+; GFX11-FAKE16-NEXT: s_lshr_b32 s96, s3, 16
+; GFX11-FAKE16-NEXT: s_lshr_b32 s97, s3, 8
+; GFX11-FAKE16-NEXT: v_writelane_b32 v19, s42, 16
+; GFX11-FAKE16-NEXT: s_lshr_b32 s42, s8, 16
+; GFX11-FAKE16-NEXT: s_lshr_b32 s43, s2, 8
; GFX11-FAKE16-NEXT: s_lshr_b32 s98, s1, 24
-; GFX11-FAKE16-NEXT: v_writelane_b32 v18, s43, 2
-; GFX11-FAKE16-NEXT: s_lshr_b32 s43, s17, 16
-; GFX11-FAKE16-NEXT: v_writelane_b32 v19, s62, 12
; GFX11-FAKE16-NEXT: s_lshr_b32 s99, s1, 16
+; GFX11-FAKE16-NEXT: v_writelane_b32 v19, s42, 17
+; GFX11-FAKE16-NEXT: s_lshr_b32 s42, s8, 8
; GFX11-FAKE16-NEXT: s_lshr_b32 s100, s1, 8
-; GFX11-FAKE16-NEXT: v_writelane_b32 v18, s43, 3
-; GFX11-FAKE16-NEXT: s_lshr_b32 s43, s17, 8
-; GFX11-FAKE16-NEXT: v_writelane_b32 v19, s63, 13
-; GFX11-FAKE16-NEXT: s_lshr_b64 s[62:63], s[8:9], 24
-; GFX11-FAKE16-NEXT: s_lshr_b32 s44, s0, 16
-; GFX11-FAKE16-NEXT: v_writelane_b32 v18, s43, 4
-; GFX11-FAKE16-NEXT: s_lshr_b32 s43, s16, 16
-; GFX11-FAKE16-NEXT: v_writelane_b32 v19, s62, 10
-; GFX11-FAKE16-NEXT: s_lshr_b32 s45, s0, 8
+; GFX11-FAKE16-NEXT: s_lshr_b32 s45, s0, 16
+; GFX11-FAKE16-NEXT: s_lshr_b32 s101, s0, 8
+; GFX11-FAKE16-NEXT: v_writelane_b32 v19, s42, 18
+; GFX11-FAKE16-NEXT: s_lshr_b32 s42, s11, 24
+; GFX11-FAKE16-NEXT: s_lshr_b64 s[56:57], s[10:11], 24
+; GFX11-FAKE16-NEXT: s_lshr_b64 s[58:59], s[12:13], 24
+; GFX11-FAKE16-NEXT: s_lshr_b64 s[60:61], s[14:15], 24
+; GFX11-FAKE16-NEXT: v_writelane_b32 v19, s42, 19
+; GFX11-FAKE16-NEXT: s_lshr_b32 s42, s11, 16
+; GFX11-FAKE16-NEXT: s_lshr_b64 s[62:63], s[40:41], 24
+; GFX11-FAKE16-NEXT: s_lshr_b64 s[72:73], s[28:29], 24
; GFX11-FAKE16-NEXT: s_lshr_b64 s[76:77], s[26:27], 24
-; GFX11-FAKE16-NEXT: v_writelane_b32 v18, s43, 5
-; GFX11-FAKE16-NEXT: s_lshr_b32 s43, s16, 8
-; GFX11-FAKE16-NEXT: v_writelane_b32 v19, s63, 11
-; GFX11-FAKE16-NEXT: s_lshr_b64 s[62:63], s[10:11], 24
+; GFX11-FAKE16-NEXT: v_writelane_b32 v19, s42, 20
+; GFX11-FAKE16-NEXT: s_lshr_b32 s42, s11, 8
; GFX11-FAKE16-NEXT: s_lshr_b64 s[88:89], s[24:25], 24
-; GFX11-FAKE16-NEXT: v_writelane_b32 v18, s43, 6
-; GFX11-FAKE16-NEXT: s_lshr_b32 s43, s3, 24
-; GFX11-FAKE16-NEXT: v_writelane_b32 v19, s62, 8
+; GFX11-FAKE16-NEXT: s_lshr_b64 s[74:75], s[22:23], 24
; GFX11-FAKE16-NEXT: s_lshr_b64 s[78:79], s[20:21], 24
+; GFX11-FAKE16-NEXT: v_writelane_b32 v19, s42, 21
+; GFX11-FAKE16-NEXT: s_lshr_b32 s42, s10, 16
; GFX11-FAKE16-NEXT: s_lshr_b64 s[90:91], s[18:19], 24
-; GFX11-FAKE16-NEXT: v_writelane_b32 v18, s43, 7
-; GFX11-FAKE16-NEXT: s_lshr_b32 s43, s3, 16
-; GFX11-FAKE16-NEXT: v_writelane_b32 v19, s63, 9
-; GFX11-FAKE16-NEXT: s_lshr_b64 s[62:63], s[12:13], 24
; GFX11-FAKE16-NEXT: s_lshr_b64 s[92:93], s[16:17], 24
-; GFX11-FAKE16-NEXT: v_writelane_b32 v18, s43, 8
-; GFX11-FAKE16-NEXT: s_lshr_b32 s43, s3, 8
-; GFX11-FAKE16-NEXT: v_writelane_b32 v19, s62, 6
; GFX11-FAKE16-NEXT: s_lshr_b64 s[94:95], s[2:3], 24
+; GFX11-FAKE16-NEXT: v_writelane_b32 v19, s42, 22
+; GFX11-FAKE16-NEXT: s_lshr_b32 s42, s10, 8
; GFX11-FAKE16-NEXT: s_lshr_b64 s[30:31], s[0:1], 24
-; GFX11-FAKE16-NEXT: v_writelane_b32 v18, s43, 9
-; GFX11-FAKE16-NEXT: s_lshr_b32 s43, s2, 8
-; GFX11-FAKE16-NEXT: v_writelane_b32 v19, s63, 7
-; GFX11-FAKE16-NEXT: s_lshr_b64 s[62:63], s[14:15], 24
-; GFX11-FAKE16-NEXT: s_delay_alu instid0(SALU_CYCLE_1) | instskip(SKIP_2) | instid1(SALU_CYCLE_1)
-; GFX11-FAKE16-NEXT: v_writelane_b32 v19, s62, 4
-; GFX11-FAKE16-NEXT: v_writelane_b32 v19, s63, 5
-; GFX11-FAKE16-NEXT: s_lshr_b64 s[62:63], s[40:41], 24
-; GFX11-FAKE16-NEXT: v_writelane_b32 v19, s62, 2
-; GFX11-FAKE16-NEXT: v_writelane_b32 v19, s63, 3
-; GFX11-FAKE16-NEXT: s_lshr_b64 s[62:63], s[28:29], 24
+; GFX11-FAKE16-NEXT: v_writelane_b32 v19, s42, 23
+; GFX11-FAKE16-NEXT: s_lshr_b32 s42, s13, 24
+; GFX11-FAKE16-NEXT: s_delay_alu instid0(SALU_CYCLE_1) | instskip(SKIP_1) | instid1(SALU_CYCLE_1)
+; GFX11-FAKE16-NEXT: v_writelane_b32 v19, s42, 24
+; GFX11-FAKE16-NEXT: s_lshr_b32 s42, s13, 16
+; GFX11-FAKE16-NEXT: v_writelane_b32 v19, s42, 25
+; GFX11-FAKE16-NEXT: s_lshr_b32 s42, s13, 8
+; GFX11-FAKE16-NEXT: s_delay_alu instid0(SALU_CYCLE_1) | instskip(SKIP_1) | instid1(SALU_CYCLE_1)
+; GFX11-FAKE16-NEXT: v_writelane_b32 v19, s42, 26
+; GFX11-FAKE16-NEXT: s_lshr_b32 s42, s12, 16
+; GFX11-FAKE16-NEXT: v_writelane_b32 v19, s42, 27
+; GFX11-FAKE16-NEXT: s_lshr_b32 s42, s12, 8
+; GFX11-FAKE16-NEXT: s_delay_alu instid0(SALU_CYCLE_1) | instskip(SKIP_1) | instid1(SALU_CYCLE_1)
+; GFX11-FAKE16-NEXT: v_writelane_b32 v19, s42, 28
+; GFX11-FAKE16-NEXT: s_lshr_b32 s42, s15, 24
+; GFX11-FAKE16-NEXT: v_writelane_b32 v19, s42, 29
+; GFX11-FAKE16-NEXT: s_lshr_b32 s42, s15, 16
+; GFX11-FAKE16-NEXT: s_delay_alu instid0(SALU_CYCLE_1) | instskip(SKIP_1) | instid1(SALU_CYCLE_1)
+; GFX11-FAKE16-NEXT: v_writelane_b32 v19, s42, 30
+; GFX11-FAKE16-NEXT: s_lshr_b32 s42, s15, 8
+; GFX11-FAKE16-NEXT: v_writelane_b32 v19, s42, 31
+; GFX11-FAKE16-NEXT: s_lshr_b32 s42, s14, 16
; GFX11-FAKE16-NEXT: s_delay_alu instid0(SALU_CYCLE_1)
-; GFX11-FAKE16-NEXT: v_writelane_b32 v19, s62, 0
-; GFX11-FAKE16-NEXT: v_writelane_b32 v19, s63, 1
-; GFX11-FAKE16-NEXT: s_lshr_b64 s[62:63], s[22:23], 24
-; GFX11-FAKE16-NEXT: s_branch .LBB57_3
-; GFX11-FAKE16-NEXT: .LBB57_2:
-; GFX11-FAKE16-NEXT: ; implicit-def: $vcc_hi
-; GFX11-FAKE16-NEXT: ; implicit-def: $vcc_lo
-; GFX11-FAKE16-NEXT: ; implicit-def: $sgpr42
-; GFX11-FAKE16-NEXT: ; kill: killed $sgpr42
-; GFX11-FAKE16-NEXT: s_mov_b32 s101, -1
-; GFX11-FAKE16-NEXT: v_writelane_b32 v19, vcc_lo, 0
-; GFX11-FAKE16-NEXT: ; implicit-def: $sgpr42
-; GFX11-FAKE16-NEXT: ; kill: killed $sgpr42
-; GFX11-FAKE16-NEXT: ; implicit-def: $sgpr42
-; GFX11-FAKE16-NEXT: ; kill: killed $sgpr42
-; GFX11-FAKE16-NEXT: ; implicit-def: $sgpr42
-; GFX11-FAKE16-NEXT: ; kill: killed $sgpr42
-; GFX11-FAKE16-NEXT: ; implicit-def: $sgpr42
-; GFX11-FAKE16-NEXT: ; kill: killed $sgpr42
-; GFX11-FAKE16-NEXT: ; implicit-def: $sgpr42
-; GFX11-FAKE16-NEXT: v_writelane_b32 v19, vcc_hi, 1
-; GFX11-FAKE16-NEXT: ; implicit-def: $vcc_lo
-; GFX11-FAKE16-NEXT: ; kill: killed $sgpr42
-; GFX11-FAKE16-NEXT: ; implicit-def: $sgpr42
-; GFX11-FAKE16-NEXT: ; kill: killed $sgpr42
-; GFX11-FAKE16-NEXT: ; implicit-def: $sgpr42
-; GFX11-FAKE16-NEXT: ; kill: killed $sgpr42
-; GFX11-FAKE16-NEXT: ; implicit-def: $sgpr42
-; GFX11-FAKE16-NEXT: ; kill: killed $sgpr42
-; GFX11-FAKE16-NEXT: ; implicit-def: $sgpr42
-; GFX11-FAKE16-NEXT: ; kill: killed $sgpr42
-; GFX11-FAKE16-NEXT: ; implicit-def: $sgpr42
-; GFX11-FAKE16-NEXT: v_writelane_b32 v19, vcc_lo, 2
-; GFX11-FAKE16-NEXT: ; kill: killed $sgpr42
-; GFX11-FAKE16-NEXT: ; implicit-def: $sgpr42
-; GFX11-FAKE16-NEXT: ; kill: killed $sgpr42
-; GFX11-FAKE16-NEXT: ; implicit-def: $sgpr42
-; GFX11-FAKE16-NEXT: ; kill: killed $sgpr42
-; GFX11-FAKE16-NEXT: ; implicit-def: $sgpr42
-; GFX11-FAKE16-NEXT: ; kill: killed $sgpr42
-; GFX11-FAKE16-NEXT: ; implicit-def: $sgpr42
-; GFX11-FAKE16-NEXT: ; kill: killed $sgpr42
-; GFX11-FAKE16-NEXT: ; implicit-def: $sgpr42
-; GFX11-FAKE16-NEXT: v_writelane_b32 v19, vcc_hi, 3
-; GFX11-FAKE16-NEXT: ; implicit-def: $vcc_lo
-; GFX11-FAKE16-NEXT: ; kill: killed $sgpr42
-; GFX11-FAKE16-NEXT: ; implicit-def: $sgpr42
-; GFX11-FAKE16-NEXT: ; kill: killed $sgpr42
-; GFX11-FAKE16-NEXT: ; implicit-def: $sgpr42
-; GFX11-FAKE16-NEXT: ; kill: killed $sgpr42
-; GFX11-FAKE16-NEXT: ; implicit-def: $sgpr42
-; GFX11-FAKE16-NEXT: ; kill: killed $sgpr42
-; GFX11-FAKE16-NEXT: ; implicit-def: $sgpr42
-; GFX11-FAKE16-NEXT: ; kill: killed $sgpr42
-; GFX11-FAKE16-NEXT: ; implicit-def: $sgpr42
-; GFX11-FAKE16-NEXT: v_writelane_b32 v19, vcc_lo, 4
-; GFX11-FAKE16-NEXT: ; kill: killed $sgpr42
-; GFX11-FAKE16-NEXT: ; implicit-def: $sgpr42
-; GFX11-FAKE16-NEXT: ; kill: killed $sgpr42
-; GFX11-FAKE16-NEXT: ; implicit-def: $sgpr42
-; GFX11-FAKE16-NEXT: ; kill: killed $sgpr42
-; GFX11-FAKE16-NEXT: ; implicit-def: $sgpr42
-; GFX11-FAKE16-NEXT: ; kill: killed $sgpr42
-; GFX11-FAKE16-NEXT: ; implicit-def: $sgpr42
-; GFX11-FAKE16-NEXT: ; kill: killed $sgpr42
-; GFX11-FAKE16-NEXT: ; implicit-def: $sgpr42
-; GFX11-FAKE16-NEXT: v_writelane_b32 v19, vcc_hi, 5
-; GFX11-FAKE16-NEXT: ; implicit-def: $vcc_lo
-; GFX11-FAKE16-NEXT: ; implicit-def: $sgpr45
-; GFX11-FAKE16-NEXT: ; implicit-def: $sgpr44
-; GFX11-FAKE16-NEXT: ; implicit-def: $sgpr30
-; GFX11-FAKE16-NEXT: ; implicit-def: $sgpr100
-; GFX11-FAKE16-NEXT: ; implicit-def: $sgpr99
-; GFX11-FAKE16-NEXT: ; implicit-def: $sgpr98
-; GFX11-FAKE16-NEXT: ; implicit-def: $sgpr43
-; GFX11-FAKE16-NEXT: ; implicit-def: $sgpr74
-; GFX11-FAKE16-NEXT: ; implicit-def: $sgpr94
-; GFX11-FAKE16-NEXT: ; implicit-def: $sgpr92
-; GFX11-FAKE16-NEXT: ; implicit-def: $sgpr90
-; GFX11-FAKE16-NEXT: ; implicit-def: $sgpr78
-; GFX11-FAKE16-NEXT: ; implicit-def: $sgpr62
-; GFX11-FAKE16-NEXT: ; kill: killed $sgpr42
-; GFX11-FAKE16-NEXT: ; implicit-def: $sgpr42
-; GFX11-FAKE16-NEXT: ; implicit-def: $sgpr97
-; GFX11-FAKE16-NEXT: ; implicit-def: $sgpr96
-; GFX11-FAKE16-NEXT: ; implicit-def: $sgpr73
-; GFX11-FAKE16-NEXT: ; implicit-def: $sgpr87
-; GFX11-FAKE16-NEXT: ; implicit-def: $sgpr72
-; GFX11-FAKE16-NEXT: ; implicit-def: $sgpr86
-; GFX11-FAKE16-NEXT: ; implicit-def: $sgpr85
-; GFX11-FAKE16-NEXT: ; implicit-def: $sgpr84
-; GFX11-FAKE16-NEXT: ; implicit-def: $sgpr83
-; GFX11-FAKE16-NEXT: ; implicit-def: $sgpr82
-; GFX11-FAKE16-NEXT: ; implicit-def: $sgpr81
-; GFX11-FAKE16-NEXT: ; implicit-def: $sgpr61
-; GFX11-FAKE16-NEXT: ; implicit-def: $sgpr80
-; GFX11-FAKE16-NEXT: ; implicit-def: $sgpr60
-; GFX11-FAKE16-NEXT: ; implicit-def: $sgpr71
-; GFX11-FAKE16-NEXT: ; implicit-def: $sgpr70
-; GFX11-FAKE16-NEXT: ; implicit-def: $sgpr58
-; GFX11-FAKE16-NEXT: ; implicit-def: $sgpr59
-; GFX11-FAKE16-NEXT: ; implicit-def: $sgpr68
-; GFX11-FAKE16-NEXT: ; implicit-def: $sgpr67
-; GFX11-FAKE16-NEXT: ; implicit-def: $sgpr66
-; GFX11-FAKE16-NEXT: ; implicit-def: $sgpr65
-; GFX11-FAKE16-NEXT: ; implicit-def: $sgpr64
-; GFX11-FAKE16-NEXT: ; implicit-def: $sgpr55
-; GFX11-FAKE16-NEXT: ; implicit-def: $sgpr54
-; GFX11-FAKE16-NEXT: ; implicit-def: $sgpr53
-; GFX11-FAKE16-NEXT: ; implicit-def: $sgpr52
-; GFX11-FAKE16-NEXT: ; implicit-def: $sgpr51
-; GFX11-FAKE16-NEXT: ; implicit-def: $sgpr50
-; GFX11-FAKE16-NEXT: ; implicit-def: $sgpr49
-; GFX11-FAKE16-NEXT: ; implicit-def: $sgpr48
-; GFX11-FAKE16-NEXT: ; implicit-def: $sgpr39
-; GFX11-FAKE16-NEXT: ; implicit-def: $sgpr38
-; GFX11-FAKE16-NEXT: ; implicit-def: $sgpr37
-; GFX11-FAKE16-NEXT: ; implicit-def: $sgpr36
-; GFX11-FAKE16-NEXT: ; implicit-def: $sgpr35
-; GFX11-FAKE16-NEXT: ; implicit-def: $sgpr56
-; GFX11-FAKE16-NEXT: ; implicit-def: $sgpr69
-; GFX11-FAKE16-NEXT: ; implicit-def: $sgpr34
-; GFX11-FAKE16-NEXT: ; implicit-def: $sgpr46
-; GFX11-FAKE16-NEXT: ; implicit-def: $sgpr47
-; GFX11-FAKE16-NEXT: ; implicit-def: $sgpr57
-; GFX11-FAKE16-NEXT: ; implicit-def: $sgpr103
-; GFX11-FAKE16-NEXT: ; implicit-def: $sgpr102
-; GFX11-FAKE16-NEXT: ; implicit-def: $sgpr104
-; GFX11-FAKE16-NEXT: ; implicit-def: $sgpr88
-; GFX11-FAKE16-NEXT: ; implicit-def: $sgpr76
-; GFX11-FAKE16-NEXT: v_writelane_b32 v19, vcc_lo, 6
-; GFX11-FAKE16-NEXT: v_writelane_b32 v19, vcc_hi, 7
-; GFX11-FAKE16-NEXT: ; implicit-def: $vcc_lo
-; GFX11-FAKE16-NEXT: v_writelane_b32 v19, vcc_lo, 8
-; GFX11-FAKE16-NEXT: v_writelane_b32 v19, vcc_hi, 9
-; GFX11-FAKE16-NEXT: ; implicit-def: $vcc_lo
-; GFX11-FAKE16-NEXT: v_writelane_b32 v19, vcc_lo, 10
-; GFX11-FAKE16-NEXT: v_writelane_b32 v19, vcc_hi, 11
-; GFX11-FAKE16-NEXT: ; implicit-def: $vcc_lo
-; GFX11-FAKE16-NEXT: v_writelane_b32 v19, vcc_lo, 12
-; GFX11-FAKE16-NEXT: v_writelane_b32 v19, vcc_hi, 13
-; GFX11-FAKE16-NEXT: ; implicit-def: $vcc_lo
-; GFX11-FAKE16-NEXT: v_writelane_b32 v19, vcc_lo, 14
-; GFX11-FAKE16-NEXT: v_writelane_b32 v19, vcc_hi, 15
-; GFX11-FAKE16-NEXT: .LBB57_3: ; %Flow
-; GFX11-FAKE16-NEXT: s_and_not1_b32 vcc_lo, exec_lo, s101
-; GFX11-FAKE16-NEXT: s_mov_b32 s101, s104
-; GFX11-FAKE16-NEXT: s_mov_b32 s104, s57
-; GFX11-FAKE16-NEXT: s_mov_b32 s57, s69
-; GFX11-FAKE16-NEXT: s_mov_b32 s69, s42
-; GFX11-FAKE16-NEXT: s_cbranch_vccnz .LBB57_5
-; GFX11-FAKE16-NEXT: ; %bb.4: ; %cmp.true
+; GFX11-FAKE16-NEXT: v_writelane_b32 v18, s42, 0
+; GFX11-FAKE16-NEXT: s_lshr_b32 s42, s14, 8
+; GFX11-FAKE16-NEXT: v_writelane_b32 v19, s46, 2
+; GFX11-FAKE16-NEXT: v_writelane_b32 v18, s42, 1
+; GFX11-FAKE16-NEXT: s_lshr_b32 s42, s41, 24
+; GFX11-FAKE16-NEXT: v_writelane_b32 v19, s47, 3
+; GFX11-FAKE16-NEXT: s_lshr_b64 s[46:47], s[6:7], 24
+; GFX11-FAKE16-NEXT: v_writelane_b32 v18, s42, 2
+; GFX11-FAKE16-NEXT: s_lshr_b32 s42, s41, 16
+; GFX11-FAKE16-NEXT: v_writelane_b32 v19, s46, 0
+; GFX11-FAKE16-NEXT: v_writelane_b32 v18, s42, 3
+; GFX11-FAKE16-NEXT: s_lshr_b32 s42, s41, 8
+; GFX11-FAKE16-NEXT: v_writelane_b32 v19, s47, 1
+; GFX11-FAKE16-NEXT: s_lshr_b64 s[46:47], s[8:9], 24
+; GFX11-FAKE16-NEXT: v_writelane_b32 v18, s42, 4
+; GFX11-FAKE16-NEXT: s_lshr_b32 s42, s40, 16
+; GFX11-FAKE16-NEXT: s_delay_alu instid0(SALU_CYCLE_1) | instskip(SKIP_1) | instid1(SALU_CYCLE_1)
+; GFX11-FAKE16-NEXT: v_writelane_b32 v18, s42, 5
+; GFX11-FAKE16-NEXT: s_lshr_b32 s42, s40, 8
+; GFX11-FAKE16-NEXT: v_writelane_b32 v18, s42, 6
+; GFX11-FAKE16-NEXT: s_lshr_b32 s42, s29, 24
+; GFX11-FAKE16-NEXT: s_delay_alu instid0(SALU_CYCLE_1) | instskip(SKIP_1) | instid1(SALU_CYCLE_1)
+; GFX11-FAKE16-NEXT: v_writelane_b32 v18, s42, 7
+; GFX11-FAKE16-NEXT: s_lshr_b32 s42, s29, 16
+; GFX11-FAKE16-NEXT: v_writelane_b32 v18, s42, 8
+; GFX11-FAKE16-NEXT: s_lshr_b32 s42, s2, 16
+; GFX11-FAKE16-NEXT: s_and_not1_b32 vcc_lo, exec_lo, s44
+; GFX11-FAKE16-NEXT: s_cbranch_vccnz .LBB57_3
+; GFX11-FAKE16-NEXT: .LBB57_2: ; %cmp.true
; GFX11-FAKE16-NEXT: s_add_u32 s0, s0, 3
; GFX11-FAKE16-NEXT: s_addc_u32 s1, s1, 0
; GFX11-FAKE16-NEXT: s_add_u32 s2, s2, 3
@@ -87307,355 +87189,321 @@ define inreg <128 x i8> @bitcast_v16i64_to_v128i8_scalar(<16 x i64> inreg %a, i3
; GFX11-FAKE16-NEXT: s_addc_u32 s7, s7, 0
; GFX11-FAKE16-NEXT: s_add_u32 s4, s4, 3
; GFX11-FAKE16-NEXT: s_addc_u32 s5, s5, 0
-; GFX11-FAKE16-NEXT: s_lshr_b32 s42, s25, 8
-; GFX11-FAKE16-NEXT: s_lshr_b64 s[62:63], s[4:5], 24
+; GFX11-FAKE16-NEXT: s_lshr_b32 s102, s29, 8
+; GFX11-FAKE16-NEXT: s_lshr_b32 s42, s5, 24
+; GFX11-FAKE16-NEXT: s_lshr_b64 s[46:47], s[4:5], 24
+; GFX11-FAKE16-NEXT: v_writelane_b32 v19, s42, 4
+; GFX11-FAKE16-NEXT: s_lshr_b32 s42, s5, 16
+; GFX11-FAKE16-NEXT: s_lshr_b32 s103, s28, 16
+; GFX11-FAKE16-NEXT: s_lshr_b32 s104, s28, 8
+; GFX11-FAKE16-NEXT: s_lshr_b32 vcc_hi, s27, 24
+; GFX11-FAKE16-NEXT: v_writelane_b32 v19, s42, 5
+; GFX11-FAKE16-NEXT: s_lshr_b32 s42, s5, 8
+; GFX11-FAKE16-NEXT: s_lshr_b32 s34, s27, 16
+; GFX11-FAKE16-NEXT: s_lshr_b32 s35, s27, 8
+; GFX11-FAKE16-NEXT: s_lshr_b32 s36, s26, 16
+; GFX11-FAKE16-NEXT: v_writelane_b32 v19, s42, 6
+; GFX11-FAKE16-NEXT: s_lshr_b32 s42, s4, 16
+; GFX11-FAKE16-NEXT: s_lshr_b32 s37, s26, 8
+; GFX11-FAKE16-NEXT: s_lshr_b32 s38, s25, 24
+; GFX11-FAKE16-NEXT: s_lshr_b32 s39, s25, 16
+; GFX11-FAKE16-NEXT: v_writelane_b32 v19, s42, 7
+; GFX11-FAKE16-NEXT: s_lshr_b32 s42, s4, 8
+; GFX11-FAKE16-NEXT: s_lshr_b32 s48, s25, 8
+; GFX11-FAKE16-NEXT: s_lshr_b32 s49, s24, 16
+; GFX11-FAKE16-NEXT: s_lshr_b32 s50, s24, 8
+; GFX11-FAKE16-NEXT: v_writelane_b32 v19, s42, 8
+; GFX11-FAKE16-NEXT: s_lshr_b32 s42, s7, 24
+; GFX11-FAKE16-NEXT: s_lshr_b32 s51, s23, 24
+; GFX11-FAKE16-NEXT: s_lshr_b32 s52, s23, 16
+; GFX11-FAKE16-NEXT: s_lshr_b32 s53, s23, 8
+; GFX11-FAKE16-NEXT: v_writelane_b32 v19, s42, 9
+; GFX11-FAKE16-NEXT: s_lshr_b32 s42, s7, 16
+; GFX11-FAKE16-NEXT: s_lshr_b32 s54, s22, 16
+; GFX11-FAKE16-NEXT: s_lshr_b32 s55, s22, 8
+; GFX11-FAKE16-NEXT: s_lshr_b32 s64, s21, 24
+; GFX11-FAKE16-NEXT: v_writelane_b32 v19, s42, 10
+; GFX11-FAKE16-NEXT: s_lshr_b32 s42, s7, 8
+; GFX11-FAKE16-NEXT: s_lshr_b32 s65, s21, 16
+; GFX11-FAKE16-NEXT: s_lshr_b32 s66, s21, 8
+; GFX11-FAKE16-NEXT: s_lshr_b32 s67, s20, 16
+; GFX11-FAKE16-NEXT: v_writelane_b32 v19, s42, 11
+; GFX11-FAKE16-NEXT: s_lshr_b32 s42, s6, 16
+; GFX11-FAKE16-NEXT: s_lshr_b32 s68, s20, 8
+; GFX11-FAKE16-NEXT: s_lshr_b32 s69, s19, 24
+; GFX11-FAKE16-NEXT: s_lshr_b32 s70, s19, 16
+; GFX11-FAKE16-NEXT: v_writelane_b32 v19, s42, 12
+; GFX11-FAKE16-NEXT: s_lshr_b32 s42, s6, 8
+; GFX11-FAKE16-NEXT: s_lshr_b32 s71, s19, 8
+; GFX11-FAKE16-NEXT: s_lshr_b32 s80, s18, 16
+; GFX11-FAKE16-NEXT: s_lshr_b32 s81, s18, 8
+; GFX11-FAKE16-NEXT: v_writelane_b32 v19, s42, 13
+; GFX11-FAKE16-NEXT: s_lshr_b32 s42, s9, 24
+; GFX11-FAKE16-NEXT: s_lshr_b32 s82, s17, 24
+; GFX11-FAKE16-NEXT: s_lshr_b32 s83, s17, 16
+; GFX11-FAKE16-NEXT: s_lshr_b32 s84, s17, 8
+; GFX11-FAKE16-NEXT: v_writelane_b32 v19, s42, 14
+; GFX11-FAKE16-NEXT: s_lshr_b32 s42, s9, 16
+; GFX11-FAKE16-NEXT: s_lshr_b32 s85, s16, 16
+; GFX11-FAKE16-NEXT: s_lshr_b32 s86, s16, 8
+; GFX11-FAKE16-NEXT: s_lshr_b32 s87, s3, 24
+; GFX11-FAKE16-NEXT: v_writelane_b32 v19, s42, 15
+; GFX11-FAKE16-NEXT: s_lshr_b32 s42, s9, 8
+; GFX11-FAKE16-NEXT: s_lshr_b32 s96, s3, 16
+; GFX11-FAKE16-NEXT: s_lshr_b32 s97, s3, 8
+; GFX11-FAKE16-NEXT: s_lshr_b32 s43, s2, 8
; GFX11-FAKE16-NEXT: v_writelane_b32 v19, s42, 16
-; GFX11-FAKE16-NEXT: s_lshr_b32 s42, s24, 16
-; GFX11-FAKE16-NEXT: s_lshr_b32 s101, s5, 24
-; GFX11-FAKE16-NEXT: s_lshr_b32 s102, s5, 16
-; GFX11-FAKE16-NEXT: s_lshr_b32 s103, s5, 8
+; GFX11-FAKE16-NEXT: s_lshr_b32 s42, s8, 16
+; GFX11-FAKE16-NEXT: s_lshr_b32 s98, s1, 24
+; GFX11-FAKE16-NEXT: s_lshr_b32 s99, s1, 16
+; GFX11-FAKE16-NEXT: s_lshr_b32 s100, s1, 8
; GFX11-FAKE16-NEXT: v_writelane_b32 v19, s42, 17
-; GFX11-FAKE16-NEXT: s_lshr_b32 s42, s24, 8
-; GFX11-FAKE16-NEXT: s_lshr_b32 s104, s4, 16
-; GFX11-FAKE16-NEXT: s_lshr_b32 s47, s4, 8
-; GFX11-FAKE16-NEXT: s_lshr_b32 s46, s7, 24
+; GFX11-FAKE16-NEXT: s_lshr_b32 s42, s8, 8
+; GFX11-FAKE16-NEXT: s_lshr_b32 s45, s0, 16
+; GFX11-FAKE16-NEXT: s_lshr_b32 s101, s0, 8
+; GFX11-FAKE16-NEXT: s_lshr_b64 s[56:57], s[10:11], 24
; GFX11-FAKE16-NEXT: v_writelane_b32 v19, s42, 18
-; GFX11-FAKE16-NEXT: s_lshr_b32 s42, s23, 24
-; GFX11-FAKE16-NEXT: s_lshr_b32 vcc_hi, s7, 16
-; GFX11-FAKE16-NEXT: s_lshr_b32 s34, s7, 8
-; GFX11-FAKE16-NEXT: s_lshr_b32 s57, s6, 16
+; GFX11-FAKE16-NEXT: s_lshr_b32 s42, s11, 24
+; GFX11-FAKE16-NEXT: s_lshr_b64 s[58:59], s[12:13], 24
+; GFX11-FAKE16-NEXT: s_lshr_b64 s[60:61], s[14:15], 24
+; GFX11-FAKE16-NEXT: s_lshr_b64 s[62:63], s[40:41], 24
; GFX11-FAKE16-NEXT: v_writelane_b32 v19, s42, 19
-; GFX11-FAKE16-NEXT: s_lshr_b32 s42, s23, 16
-; GFX11-FAKE16-NEXT: s_lshr_b32 s56, s6, 8
-; GFX11-FAKE16-NEXT: s_lshr_b32 s35, s9, 24
-; GFX11-FAKE16-NEXT: s_lshr_b32 s36, s9, 16
+; GFX11-FAKE16-NEXT: s_lshr_b32 s42, s11, 16
+; GFX11-FAKE16-NEXT: s_lshr_b64 s[72:73], s[28:29], 24
+; GFX11-FAKE16-NEXT: s_lshr_b64 s[76:77], s[26:27], 24
+; GFX11-FAKE16-NEXT: s_lshr_b64 s[88:89], s[24:25], 24
; GFX11-FAKE16-NEXT: v_writelane_b32 v19, s42, 20
-; GFX11-FAKE16-NEXT: s_lshr_b32 s42, s23, 8
-; GFX11-FAKE16-NEXT: s_lshr_b32 s37, s9, 8
-; GFX11-FAKE16-NEXT: s_lshr_b32 s38, s8, 16
-; GFX11-FAKE16-NEXT: s_lshr_b32 s39, s8, 8
+; GFX11-FAKE16-NEXT: s_lshr_b32 s42, s11, 8
+; GFX11-FAKE16-NEXT: s_lshr_b64 s[74:75], s[22:23], 24
+; GFX11-FAKE16-NEXT: s_lshr_b64 s[78:79], s[20:21], 24
+; GFX11-FAKE16-NEXT: s_lshr_b64 s[90:91], s[18:19], 24
; GFX11-FAKE16-NEXT: v_writelane_b32 v19, s42, 21
-; GFX11-FAKE16-NEXT: s_lshr_b32 s42, s22, 16
-; GFX11-FAKE16-NEXT: s_lshr_b32 s48, s11, 24
-; GFX11-FAKE16-NEXT: s_lshr_b32 s49, s11, 16
-; GFX11-FAKE16-NEXT: s_lshr_b32 s50, s11, 8
+; GFX11-FAKE16-NEXT: s_lshr_b32 s42, s10, 16
+; GFX11-FAKE16-NEXT: s_lshr_b64 s[92:93], s[16:17], 24
+; GFX11-FAKE16-NEXT: s_lshr_b64 s[94:95], s[2:3], 24
+; GFX11-FAKE16-NEXT: s_lshr_b64 s[30:31], s[0:1], 24
; GFX11-FAKE16-NEXT: v_writelane_b32 v19, s42, 22
-; GFX11-FAKE16-NEXT: s_lshr_b32 s42, s22, 8
-; GFX11-FAKE16-NEXT: s_lshr_b32 s51, s10, 16
-; GFX11-FAKE16-NEXT: s_lshr_b32 s52, s10, 8
-; GFX11-FAKE16-NEXT: s_lshr_b32 s53, s13, 24
+; GFX11-FAKE16-NEXT: s_lshr_b32 s42, s10, 8
+; GFX11-FAKE16-NEXT: s_delay_alu instid0(SALU_CYCLE_1) | instskip(SKIP_1) | instid1(SALU_CYCLE_1)
; GFX11-FAKE16-NEXT: v_writelane_b32 v19, s42, 23
-; GFX11-FAKE16-NEXT: s_lshr_b32 s42, s21, 24
-; GFX11-FAKE16-NEXT: s_lshr_b32 s54, s13, 16
-; GFX11-FAKE16-NEXT: s_lshr_b32 s55, s13, 8
-; GFX11-FAKE16-NEXT: s_lshr_b32 s64, s12, 16
+; GFX11-FAKE16-NEXT: s_lshr_b32 s42, s13, 24
; GFX11-FAKE16-NEXT: v_writelane_b32 v19, s42, 24
-; GFX11-FAKE16-NEXT: s_lshr_b32 s42, s21, 16
-; GFX11-FAKE16-NEXT: s_lshr_b32 s65, s12, 8
-; GFX11-FAKE16-NEXT: s_lshr_b32 s66, s15, 24
-; GFX11-FAKE16-NEXT: s_lshr_b32 s67, s15, 16
+; GFX11-FAKE16-NEXT: s_lshr_b32 s42, s13, 16
+; GFX11-FAKE16-NEXT: s_delay_alu instid0(SALU_CYCLE_1) | instskip(SKIP_1) | instid1(SALU_CYCLE_1)
; GFX11-FAKE16-NEXT: v_writelane_b32 v19, s42, 25
-; GFX11-FAKE16-NEXT: s_lshr_b32 s42, s21, 8
-; GFX11-FAKE16-NEXT: s_lshr_b32 s68, s15, 8
-; GFX11-FAKE16-NEXT: s_lshr_b32 s59, s14, 16
-; GFX11-FAKE16-NEXT: s_lshr_b32 s58, s14, 8
+; GFX11-FAKE16-NEXT: s_lshr_b32 s42, s13, 8
; GFX11-FAKE16-NEXT: v_writelane_b32 v19, s42, 26
-; GFX11-FAKE16-NEXT: s_lshr_b32 s42, s20, 16
-; GFX11-FAKE16-NEXT: s_lshr_b32 s70, s41, 24
-; GFX11-FAKE16-NEXT: s_lshr_b32 s71, s41, 16
-; GFX11-FAKE16-NEXT: s_lshr_b32 s60, s41, 8
+; GFX11-FAKE16-NEXT: s_lshr_b32 s42, s12, 16
+; GFX11-FAKE16-NEXT: s_delay_alu instid0(SALU_CYCLE_1) | instskip(SKIP_1) | instid1(SALU_CYCLE_1)
; GFX11-FAKE16-NEXT: v_writelane_b32 v19, s42, 27
-; GFX11-FAKE16-NEXT: s_lshr_b32 s42, s20, 8
-; GFX11-FAKE16-NEXT: s_lshr_b32 s80, s40, 16
-; GFX11-FAKE16-NEXT: s_lshr_b32 s61, s40, 8
-; GFX11-FAKE16-NEXT: s_lshr_b32 s81, s29, 24
+; GFX11-FAKE16-NEXT: s_lshr_b32 s42, s12, 8
; GFX11-FAKE16-NEXT: v_writelane_b32 v19, s42, 28
-; GFX11-FAKE16-NEXT: s_lshr_b32 s42, s19, 24
-; GFX11-FAKE16-NEXT: s_lshr_b32 s82, s29, 16
-; GFX11-FAKE16-NEXT: s_lshr_b32 s83, s29, 8
-; GFX11-FAKE16-NEXT: s_lshr_b32 s84, s28, 16
+; GFX11-FAKE16-NEXT: s_lshr_b32 s42, s15, 24
+; GFX11-FAKE16-NEXT: s_delay_alu instid0(SALU_CYCLE_1) | instskip(SKIP_1) | instid1(SALU_CYCLE_1)
; GFX11-FAKE16-NEXT: v_writelane_b32 v19, s42, 29
-; GFX11-FAKE16-NEXT: s_lshr_b32 s42, s19, 16
-; GFX11-FAKE16-NEXT: s_lshr_b32 s85, s28, 8
-; GFX11-FAKE16-NEXT: s_lshr_b32 s86, s27, 24
-; GFX11-FAKE16-NEXT: s_lshr_b32 s72, s27, 16
+; GFX11-FAKE16-NEXT: s_lshr_b32 s42, s15, 16
; GFX11-FAKE16-NEXT: v_writelane_b32 v19, s42, 30
-; GFX11-FAKE16-NEXT: s_lshr_b32 s42, s19, 8
-; GFX11-FAKE16-NEXT: s_lshr_b32 s87, s27, 8
-; GFX11-FAKE16-NEXT: s_lshr_b32 s73, s26, 16
-; GFX11-FAKE16-NEXT: s_lshr_b32 s96, s26, 8
+; GFX11-FAKE16-NEXT: s_lshr_b32 s42, s15, 8
+; GFX11-FAKE16-NEXT: s_delay_alu instid0(SALU_CYCLE_1) | instskip(SKIP_1) | instid1(SALU_CYCLE_1)
; GFX11-FAKE16-NEXT: v_writelane_b32 v19, s42, 31
-; GFX11-FAKE16-NEXT: s_lshr_b32 s42, s18, 16
-; GFX11-FAKE16-NEXT: s_lshr_b32 s97, s25, 24
+; GFX11-FAKE16-NEXT: s_lshr_b32 s42, s14, 16
; GFX11-FAKE16-NEXT: v_writelane_b32 v18, s42, 0
-; GFX11-FAKE16-NEXT: s_lshr_b32 s42, s18, 8
-; GFX11-FAKE16-NEXT: v_writelane_b32 v19, s62, 14
-; GFX11-FAKE16-NEXT: s_lshr_b32 s69, s25, 16
-; GFX11-FAKE16-NEXT: s_lshr_b32 s74, s2, 16
+; GFX11-FAKE16-NEXT: s_lshr_b32 s42, s14, 8
+; GFX11-FAKE16-NEXT: v_writelane_b32 v19, s46, 2
; GFX11-FAKE16-NEXT: v_writelane_b32 v18, s42, 1
-; GFX11-FAKE16-NEXT: s_lshr_b32 s42, s17, 24
-; GFX11-FAKE16-NEXT: v_writelane_b32 v19, s63, 15
-; GFX11-FAKE16-NEXT: s_lshr_b64 s[62:63], s[6:7], 24
-; GFX11-FAKE16-NEXT: s_lshr_b32 s43, s2, 8
+; GFX11-FAKE16-NEXT: s_lshr_b32 s42, s41, 24
+; GFX11-FAKE16-NEXT: v_writelane_b32 v19, s47, 3
+; GFX11-FAKE16-NEXT: s_lshr_b64 s[46:47], s[6:7], 24
; GFX11-FAKE16-NEXT: v_writelane_b32 v18, s42, 2
-; GFX11-FAKE16-NEXT: s_lshr_b32 s42, s17, 16
-; GFX11-FAKE16-NEXT: v_writelane_b32 v19, s62, 12
-; GFX11-FAKE16-NEXT: s_lshr_b32 s98, s1, 24
-; GFX11-FAKE16-NEXT: s_lshr_b32 s99, s1, 16
+; GFX11-FAKE16-NEXT: s_lshr_b32 s42, s41, 16
+; GFX11-FAKE16-NEXT: v_writelane_b32 v19, s46, 0
; GFX11-FAKE16-NEXT: v_writelane_b32 v18, s42, 3
-; GFX11-FAKE16-NEXT: s_lshr_b32 s42, s17, 8
-; GFX11-FAKE16-NEXT: v_writelane_b32 v19, s63, 13
-; GFX11-FAKE16-NEXT: s_lshr_b64 s[62:63], s[8:9], 24
-; GFX11-FAKE16-NEXT: s_lshr_b32 s100, s1, 8
+; GFX11-FAKE16-NEXT: s_lshr_b32 s42, s41, 8
+; GFX11-FAKE16-NEXT: v_writelane_b32 v19, s47, 1
+; GFX11-FAKE16-NEXT: s_lshr_b64 s[46:47], s[8:9], 24
; GFX11-FAKE16-NEXT: v_writelane_b32 v18, s42, 4
-; GFX11-FAKE16-NEXT: s_lshr_b32 s42, s16, 16
-; GFX11-FAKE16-NEXT: v_writelane_b32 v19, s62, 10
-; GFX11-FAKE16-NEXT: s_lshr_b32 s44, s0, 16
-; GFX11-FAKE16-NEXT: s_lshr_b32 s45, s0, 8
+; GFX11-FAKE16-NEXT: s_lshr_b32 s42, s40, 16
+; GFX11-FAKE16-NEXT: s_delay_alu instid0(SALU_CYCLE_1) | instskip(SKIP_1) | instid1(SALU_CYCLE_1)
; GFX11-FAKE16-NEXT: v_writelane_b32 v18, s42, 5
-; GFX11-FAKE16-NEXT: s_lshr_b32 s42, s16, 8
-; GFX11-FAKE16-NEXT: v_writelane_b32 v19, s63, 11
-; GFX11-FAKE16-NEXT: s_lshr_b64 s[62:63], s[10:11], 24
-; GFX11-FAKE16-NEXT: s_lshr_b64 s[76:77], s[26:27], 24
+; GFX11-FAKE16-NEXT: s_lshr_b32 s42, s40, 8
; GFX11-FAKE16-NEXT: v_writelane_b32 v18, s42, 6
-; GFX11-FAKE16-NEXT: s_lshr_b32 s42, s3, 24
-; GFX11-FAKE16-NEXT: v_writelane_b32 v19, s62, 8
-; GFX11-FAKE16-NEXT: s_lshr_b64 s[88:89], s[24:25], 24
-; GFX11-FAKE16-NEXT: s_lshr_b64 s[78:79], s[20:21], 24
+; GFX11-FAKE16-NEXT: s_lshr_b32 s42, s29, 24
+; GFX11-FAKE16-NEXT: s_delay_alu instid0(SALU_CYCLE_1) | instskip(SKIP_1) | instid1(SALU_CYCLE_1)
; GFX11-FAKE16-NEXT: v_writelane_b32 v18, s42, 7
-; GFX11-FAKE16-NEXT: s_lshr_b32 s42, s3, 16
-; GFX11-FAKE16-NEXT: v_writelane_b32 v19, s63, 9
-; GFX11-FAKE16-NEXT: s_lshr_b64 s[62:63], s[12:13], 24
-; GFX11-FAKE16-NEXT: s_lshr_b64 s[90:91], s[18:19], 24
+; GFX11-FAKE16-NEXT: s_lshr_b32 s42, s29, 16
; GFX11-FAKE16-NEXT: v_writelane_b32 v18, s42, 8
-; GFX11-FAKE16-NEXT: s_lshr_b32 s42, s3, 8
-; GFX11-FAKE16-NEXT: v_writelane_b32 v19, s62, 6
-; GFX11-FAKE16-NEXT: s_lshr_b64 s[92:93], s[16:17], 24
-; GFX11-FAKE16-NEXT: s_lshr_b64 s[94:95], s[2:3], 24
-; GFX11-FAKE16-NEXT: s_lshr_b64 s[30:31], s[0:1], 24
-; GFX11-FAKE16-NEXT: v_writelane_b32 v18, s42, 9
-; GFX11-FAKE16-NEXT: v_writelane_b32 v19, s63, 7
-; GFX11-FAKE16-NEXT: s_lshr_b64 s[62:63], s[14:15], 24
-; GFX11-FAKE16-NEXT: s_delay_alu instid0(SALU_CYCLE_1) | instskip(SKIP_2) | instid1(SALU_CYCLE_1)
-; GFX11-FAKE16-NEXT: v_writelane_b32 v19, s62, 4
-; GFX11-FAKE16-NEXT: v_writelane_b32 v19, s63, 5
-; GFX11-FAKE16-NEXT: s_lshr_b64 s[62:63], s[40:41], 24
-; GFX11-FAKE16-NEXT: v_writelane_b32 v19, s62, 2
-; GFX11-FAKE16-NEXT: v_writelane_b32 v19, s63, 3
-; GFX11-FAKE16-NEXT: s_lshr_b64 s[62:63], s[28:29], 24
-; GFX11-FAKE16-NEXT: s_delay_alu instid0(SALU_CYCLE_1)
-; GFX11-FAKE16-NEXT: v_writelane_b32 v19, s62, 0
-; GFX11-FAKE16-NEXT: v_writelane_b32 v19, s63, 1
-; GFX11-FAKE16-NEXT: s_lshr_b64 s[62:63], s[22:23], 24
-; GFX11-FAKE16-NEXT: .LBB57_5: ; %end
+; GFX11-FAKE16-NEXT: s_lshr_b32 s42, s2, 16
+; GFX11-FAKE16-NEXT: .LBB57_3: ; %end
+; GFX11-FAKE16-NEXT: s_lshl_b32 s44, s101, 8
+; GFX11-FAKE16-NEXT: s_and_b32 s0, s0, 0xff
+; GFX11-FAKE16-NEXT: s_and_b32 s45, s45, 0xff
+; GFX11-FAKE16-NEXT: s_or_b32 s0, s0, s44
+; GFX11-FAKE16-NEXT: s_lshl_b32 s44, s30, 8
; GFX11-FAKE16-NEXT: s_lshl_b32 s43, s43, 8
+; GFX11-FAKE16-NEXT: s_or_b32 s44, s45, s44
; GFX11-FAKE16-NEXT: s_and_b32 s2, s2, 0xff
-; GFX11-FAKE16-NEXT: s_and_b32 s42, s74, 0xff
-; GFX11-FAKE16-NEXT: s_or_b32 s2, s2, s43
-; GFX11-FAKE16-NEXT: s_lshl_b32 s43, s94, 8
-; GFX11-FAKE16-NEXT: s_and_b32 s2, s2, 0xffff
-; GFX11-FAKE16-NEXT: s_or_b32 s42, s42, s43
-; GFX11-FAKE16-NEXT: s_lshl_b32 s45, s45, 8
-; GFX11-FAKE16-NEXT: s_lshl_b32 s42, s42, 16
-; GFX11-FAKE16-NEXT: s_and_b32 s0, s0, 0xff
-; GFX11-FAKE16-NEXT: s_or_b32 s2, s2, s42
-; GFX11-FAKE16-NEXT: v_readlane_b32 s42, v18, 9
-; GFX11-FAKE16-NEXT: s_or_b32 s0, s0, s45
-; GFX11-FAKE16-NEXT: s_lshl_b32 s45, s30, 8
-; GFX11-FAKE16-NEXT: s_and_b32 s44, s44, 0xff
-; GFX11-FAKE16-NEXT: s_and_b32 s3, s3, 0xff
-; GFX11-FAKE16-NEXT: s_or_b32 s44, s44, s45
-; GFX11-FAKE16-NEXT: s_lshl_b32 s42, s42, 8
; GFX11-FAKE16-NEXT: s_and_b32 s0, s0, 0xffff
; GFX11-FAKE16-NEXT: s_lshl_b32 s44, s44, 16
-; GFX11-FAKE16-NEXT: s_or_b32 s3, s3, s42
-; GFX11-FAKE16-NEXT: v_readlane_b32 s42, v18, 8
-; GFX11-FAKE16-NEXT: v_readlane_b32 s43, v18, 7
+; GFX11-FAKE16-NEXT: s_or_b32 s2, s2, s43
+; GFX11-FAKE16-NEXT: s_lshl_b32 s43, s94, 8
+; GFX11-FAKE16-NEXT: s_and_b32 s42, s42, 0xff
; GFX11-FAKE16-NEXT: s_or_b32 s0, s0, s44
; GFX11-FAKE16-NEXT: s_and_b32 s1, s1, 0xff
; GFX11-FAKE16-NEXT: s_lshl_b32 s44, s100, 8
-; GFX11-FAKE16-NEXT: s_lshl_b32 s45, s98, 8
+; GFX11-FAKE16-NEXT: s_or_b32 s42, s42, s43
; GFX11-FAKE16-NEXT: s_or_b32 s1, s1, s44
; GFX11-FAKE16-NEXT: s_and_b32 s44, s99, 0xff
-; GFX11-FAKE16-NEXT: s_and_b32 s42, s42, 0xff
+; GFX11-FAKE16-NEXT: s_lshl_b32 s45, s98, 8
+; GFX11-FAKE16-NEXT: s_and_b32 s2, s2, 0xffff
+; GFX11-FAKE16-NEXT: s_lshl_b32 s42, s42, 16
; GFX11-FAKE16-NEXT: s_or_b32 s44, s44, s45
-; GFX11-FAKE16-NEXT: s_lshl_b32 s43, s43, 8
+; GFX11-FAKE16-NEXT: s_or_b32 s2, s2, s42
+; GFX11-FAKE16-NEXT: s_and_b32 s3, s3, 0xff
+; GFX11-FAKE16-NEXT: s_lshl_b32 s42, s97, 8
; GFX11-FAKE16-NEXT: s_and_b32 s1, s1, 0xffff
; GFX11-FAKE16-NEXT: s_lshl_b32 s44, s44, 16
-; GFX11-FAKE16-NEXT: s_or_b32 s42, s42, s43
+; GFX11-FAKE16-NEXT: s_or_b32 s3, s3, s42
+; GFX11-FAKE16-NEXT: s_and_b32 s42, s96, 0xff
+; GFX11-FAKE16-NEXT: s_lshl_b32 s43, s87, 8
; GFX11-FAKE16-NEXT: s_or_b32 s1, s1, s44
+; GFX11-FAKE16-NEXT: s_or_b32 s42, s42, s43
; GFX11-FAKE16-NEXT: s_and_b32 s3, s3, 0xffff
; GFX11-FAKE16-NEXT: s_lshl_b32 s42, s42, 16
; GFX11-FAKE16-NEXT: v_dual_mov_b32 v1, s0 :: v_dual_mov_b32 v2, s1
-; GFX11-FAKE16-NEXT: v_readlane_b32 s0, v18, 6
+; GFX11-FAKE16-NEXT: s_lshl_b32 s0, s86, 8
+; GFX11-FAKE16-NEXT: s_and_b32 s1, s16, 0xff
; GFX11-FAKE16-NEXT: s_or_b32 s3, s3, s42
; GFX11-FAKE16-NEXT: s_delay_alu instid0(SALU_CYCLE_1)
; GFX11-FAKE16-NEXT: v_dual_mov_b32 v3, s2 :: v_dual_mov_b32 v4, s3
-; GFX11-FAKE16-NEXT: v_readlane_b32 s2, v18, 5
-; GFX11-FAKE16-NEXT: s_lshl_b32 s0, s0, 8
-; GFX11-FAKE16-NEXT: s_and_b32 s1, s16, 0xff
-; GFX11-FAKE16-NEXT: v_readlane_b32 s3, v18, 2
; GFX11-FAKE16-NEXT: s_or_b32 s0, s1, s0
; GFX11-FAKE16-NEXT: s_lshl_b32 s1, s92, 8
-; GFX11-FAKE16-NEXT: s_and_b32 s2, s2, 0xff
+; GFX11-FAKE16-NEXT: s_and_b32 s2, s85, 0xff
; GFX11-FAKE16-NEXT: s_and_b32 s0, s0, 0xffff
; GFX11-FAKE16-NEXT: s_or_b32 s1, s2, s1
-; GFX11-FAKE16-NEXT: v_readlane_b32 s2, v18, 4
+; GFX11-FAKE16-NEXT: s_lshl_b32 s2, s84, 8
; GFX11-FAKE16-NEXT: s_lshl_b32 s1, s1, 16
-; GFX11-FAKE16-NEXT: s_lshl_b32 s3, s3, 8
+; GFX11-FAKE16-NEXT: s_lshl_b32 s3, s82, 8
; GFX11-FAKE16-NEXT: s_or_b32 s0, s0, s1
; GFX11-FAKE16-NEXT: s_and_b32 s1, s17, 0xff
-; GFX11-FAKE16-NEXT: s_lshl_b32 s2, s2, 8
-; GFX11-FAKE16-NEXT: v_readlane_b32 s16, v18, 0
+; GFX11-FAKE16-NEXT: s_and_b32 s16, s80, 0xff
; GFX11-FAKE16-NEXT: s_or_b32 s1, s1, s2
-; GFX11-FAKE16-NEXT: v_readlane_b32 s2, v18, 3
+; GFX11-FAKE16-NEXT: s_and_b32 s2, s83, 0xff
; GFX11-FAKE16-NEXT: s_and_b32 s1, s1, 0xffff
-; GFX11-FAKE16-NEXT: v_readlane_b32 s17, v19, 29
-; GFX11-FAKE16-NEXT: s_and_b32 s16, s16, 0xff
-; GFX11-FAKE16-NEXT: v_readlane_b32 s100, v17, 4
-; GFX11-FAKE16-NEXT: s_and_b32 s2, s2, 0xff
-; GFX11-FAKE16-NEXT: v_readlane_b32 s99, v17, 3
; GFX11-FAKE16-NEXT: s_or_b32 s2, s2, s3
; GFX11-FAKE16-NEXT: s_and_b32 s3, s18, 0xff
; GFX11-FAKE16-NEXT: s_lshl_b32 s2, s2, 16
-; GFX11-FAKE16-NEXT: s_lshl_b32 s17, s17, 8
+; GFX11-FAKE16-NEXT: s_lshl_b32 s17, s69, 8
; GFX11-FAKE16-NEXT: s_or_b32 s1, s1, s2
-; GFX11-FAKE16-NEXT: v_readlane_b32 s2, v18, 1
+; GFX11-FAKE16-NEXT: s_lshl_b32 s2, s81, 8
; GFX11-FAKE16-NEXT: v_dual_mov_b32 v5, s0 :: v_dual_mov_b32 v6, s1
-; GFX11-FAKE16-NEXT: v_readlane_b32 s0, v19, 28
-; GFX11-FAKE16-NEXT: s_and_b32 s1, s20, 0xff
-; GFX11-FAKE16-NEXT: s_lshl_b32 s2, s2, 8
-; GFX11-FAKE16-NEXT: v_readlane_b32 s18, v19, 19
; GFX11-FAKE16-NEXT: s_or_b32 s2, s3, s2
; GFX11-FAKE16-NEXT: s_lshl_b32 s3, s90, 8
; GFX11-FAKE16-NEXT: s_and_b32 s2, s2, 0xffff
; GFX11-FAKE16-NEXT: s_or_b32 s3, s16, s3
-; GFX11-FAKE16-NEXT: v_readlane_b32 s16, v19, 31
+; GFX11-FAKE16-NEXT: s_lshl_b32 s16, s71, 8
; GFX11-FAKE16-NEXT: s_lshl_b32 s3, s3, 16
-; GFX11-FAKE16-NEXT: s_lshl_b32 s0, s0, 8
+; GFX11-FAKE16-NEXT: s_lshl_b32 s0, s68, 8
; GFX11-FAKE16-NEXT: s_or_b32 s2, s2, s3
; GFX11-FAKE16-NEXT: s_and_b32 s3, s19, 0xff
-; GFX11-FAKE16-NEXT: s_lshl_b32 s16, s16, 8
-; GFX11-FAKE16-NEXT: s_or_b32 s0, s1, s0
+; GFX11-FAKE16-NEXT: s_and_b32 s1, s20, 0xff
; GFX11-FAKE16-NEXT: s_or_b32 s3, s3, s16
-; GFX11-FAKE16-NEXT: v_readlane_b32 s16, v19, 30
+; GFX11-FAKE16-NEXT: s_and_b32 s16, s70, 0xff
; GFX11-FAKE16-NEXT: s_and_b32 s3, s3, 0xffff
-; GFX11-FAKE16-NEXT: s_lshl_b32 s1, s78, 8
-; GFX11-FAKE16-NEXT: s_and_b32 s0, s0, 0xffff
-; GFX11-FAKE16-NEXT: s_lshl_b32 s18, s18, 8
-; GFX11-FAKE16-NEXT: s_and_b32 s16, s16, 0xff
-; GFX11-FAKE16-NEXT: s_lshl_b32 s19, s86, 8
; GFX11-FAKE16-NEXT: s_or_b32 s16, s16, s17
-; GFX11-FAKE16-NEXT: v_readlane_b32 s17, v19, 21
+; GFX11-FAKE16-NEXT: s_or_b32 s0, s1, s0
; GFX11-FAKE16-NEXT: s_lshl_b32 s16, s16, 16
-; GFX11-FAKE16-NEXT: v_readlane_b32 s98, v17, 2
+; GFX11-FAKE16-NEXT: s_lshl_b32 s1, s78, 8
; GFX11-FAKE16-NEXT: s_or_b32 s3, s3, s16
; GFX11-FAKE16-NEXT: s_delay_alu instid0(SALU_CYCLE_1)
; GFX11-FAKE16-NEXT: v_dual_mov_b32 v7, s2 :: v_dual_mov_b32 v8, s3
-; GFX11-FAKE16-NEXT: v_readlane_b32 s2, v19, 27
-; GFX11-FAKE16-NEXT: v_readlane_b32 s3, v19, 24
-; GFX11-FAKE16-NEXT: v_readlane_b32 s16, v19, 22
-; GFX11-FAKE16-NEXT: s_lshl_b32 s17, s17, 8
-; GFX11-FAKE16-NEXT: s_clause 0x1
-; GFX11-FAKE16-NEXT: scratch_store_b128 v0, v[1:4], off
-; GFX11-FAKE16-NEXT: scratch_store_b128 v0, v[5:8], off offset:16
-; GFX11-FAKE16-NEXT: s_and_b32 s2, s2, 0xff
-; GFX11-FAKE16-NEXT: s_lshl_b32 s3, s3, 8
+; GFX11-FAKE16-NEXT: s_and_b32 s2, s67, 0xff
+; GFX11-FAKE16-NEXT: s_and_b32 s0, s0, 0xffff
; GFX11-FAKE16-NEXT: s_or_b32 s1, s2, s1
-; GFX11-FAKE16-NEXT: v_readlane_b32 s2, v19, 26
+; GFX11-FAKE16-NEXT: s_lshl_b32 s2, s66, 8
; GFX11-FAKE16-NEXT: s_lshl_b32 s1, s1, 16
-; GFX11-FAKE16-NEXT: s_and_b32 s16, s16, 0xff
+; GFX11-FAKE16-NEXT: s_lshl_b32 s3, s64, 8
; GFX11-FAKE16-NEXT: s_or_b32 s0, s0, s1
; GFX11-FAKE16-NEXT: s_and_b32 s1, s21, 0xff
-; GFX11-FAKE16-NEXT: s_lshl_b32 s2, s2, 8
-; GFX11-FAKE16-NEXT: v_readlane_b32 s86, v16, 30
+; GFX11-FAKE16-NEXT: s_and_b32 s16, s54, 0xff
; GFX11-FAKE16-NEXT: s_or_b32 s1, s1, s2
-; GFX11-FAKE16-NEXT: v_readlane_b32 s2, v19, 25
+; GFX11-FAKE16-NEXT: s_and_b32 s2, s65, 0xff
; GFX11-FAKE16-NEXT: s_and_b32 s1, s1, 0xffff
-; GFX11-FAKE16-NEXT: v_readlane_b32 s31, v16, 1
-; GFX11-FAKE16-NEXT: v_readlane_b32 s30, v16, 0
-; GFX11-FAKE16-NEXT: s_and_b32 s2, s2, 0xff
-; GFX11-FAKE16-NEXT: s_delay_alu instid0(SALU_CYCLE_1) | instskip(SKIP_2) | instid1(SALU_CYCLE_1)
; GFX11-FAKE16-NEXT: s_or_b32 s2, s2, s3
; GFX11-FAKE16-NEXT: s_and_b32 s3, s22, 0xff
; GFX11-FAKE16-NEXT: s_lshl_b32 s2, s2, 16
+; GFX11-FAKE16-NEXT: s_lshl_b32 s17, s53, 8
; GFX11-FAKE16-NEXT: s_or_b32 s1, s1, s2
-; GFX11-FAKE16-NEXT: v_readlane_b32 s2, v19, 23
-; GFX11-FAKE16-NEXT: v_dual_mov_b32 v9, s0 :: v_dual_mov_b32 v10, s1
-; GFX11-FAKE16-NEXT: v_readlane_b32 s1, v19, 18
-; GFX11-FAKE16-NEXT: s_and_b32 s0, s24, 0xff
-; GFX11-FAKE16-NEXT: s_lshl_b32 s2, s2, 8
-; GFX11-FAKE16-NEXT: s_delay_alu instid0(SALU_CYCLE_1)
+; GFX11-FAKE16-NEXT: s_lshl_b32 s2, s55, 8
+; GFX11-FAKE16-NEXT: s_lshl_b32 s18, s51, 8
; GFX11-FAKE16-NEXT: s_or_b32 s2, s3, s2
-; GFX11-FAKE16-NEXT: s_lshl_b32 s3, s62, 8
+; GFX11-FAKE16-NEXT: s_lshl_b32 s3, s74, 8
; GFX11-FAKE16-NEXT: s_and_b32 s2, s2, 0xffff
; GFX11-FAKE16-NEXT: s_or_b32 s3, s16, s3
; GFX11-FAKE16-NEXT: s_and_b32 s16, s23, 0xff
; GFX11-FAKE16-NEXT: s_lshl_b32 s3, s3, 16
; GFX11-FAKE16-NEXT: s_or_b32 s16, s16, s17
-; GFX11-FAKE16-NEXT: v_readlane_b32 s17, v19, 20
+; GFX11-FAKE16-NEXT: s_and_b32 s17, s52, 0xff
; GFX11-FAKE16-NEXT: s_or_b32 s2, s2, s3
-; GFX11-FAKE16-NEXT: s_and_b32 s3, s16, 0xffff
-; GFX11-FAKE16-NEXT: s_lshl_b32 s1, s1, 8
-; GFX11-FAKE16-NEXT: s_and_b32 s17, s17, 0xff
-; GFX11-FAKE16-NEXT: s_or_b32 s0, s0, s1
; GFX11-FAKE16-NEXT: s_or_b32 s17, s17, s18
-; GFX11-FAKE16-NEXT: s_and_b32 s0, s0, 0xffff
+; GFX11-FAKE16-NEXT: s_and_b32 s3, s16, 0xffff
; GFX11-FAKE16-NEXT: s_lshl_b32 s16, s17, 16
-; GFX11-FAKE16-NEXT: s_lshl_b32 s17, s97, 8
+; GFX11-FAKE16-NEXT: v_dual_mov_b32 v9, s0 :: v_dual_mov_b32 v10, s1
; GFX11-FAKE16-NEXT: s_or_b32 s3, s3, s16
; GFX11-FAKE16-NEXT: s_delay_alu instid0(SALU_CYCLE_1)
; GFX11-FAKE16-NEXT: v_dual_mov_b32 v11, s2 :: v_dual_mov_b32 v12, s3
-; GFX11-FAKE16-NEXT: v_readlane_b32 s2, v19, 17
+; GFX11-FAKE16-NEXT: s_and_b32 s0, s24, 0xff
+; GFX11-FAKE16-NEXT: s_lshl_b32 s1, s50, 8
+; GFX11-FAKE16-NEXT: s_and_b32 s2, s49, 0xff
; GFX11-FAKE16-NEXT: s_lshl_b32 s3, s88, 8
-; GFX11-FAKE16-NEXT: s_and_b32 s16, s69, 0xff
-; GFX11-FAKE16-NEXT: s_and_b32 s18, s72, 0xff
-; GFX11-FAKE16-NEXT: v_readlane_b32 s97, v17, 1
-; GFX11-FAKE16-NEXT: s_and_b32 s2, s2, 0xff
-; GFX11-FAKE16-NEXT: v_readlane_b32 s69, v16, 21
+; GFX11-FAKE16-NEXT: s_or_b32 s0, s0, s1
; GFX11-FAKE16-NEXT: s_or_b32 s1, s2, s3
-; GFX11-FAKE16-NEXT: v_readlane_b32 s3, v19, 16
; GFX11-FAKE16-NEXT: s_and_b32 s2, s25, 0xff
-; GFX11-FAKE16-NEXT: s_lshl_b32 s1, s1, 16
-; GFX11-FAKE16-NEXT: s_delay_alu instid0(SALU_CYCLE_1) | instskip(SKIP_1) | instid1(SALU_CYCLE_1)
-; GFX11-FAKE16-NEXT: s_or_b32 s0, s0, s1
-; GFX11-FAKE16-NEXT: s_lshl_b32 s3, s3, 8
+; GFX11-FAKE16-NEXT: s_lshl_b32 s3, s48, 8
+; GFX11-FAKE16-NEXT: s_and_b32 s16, s39, 0xff
+; GFX11-FAKE16-NEXT: s_lshl_b32 s17, s38, 8
; GFX11-FAKE16-NEXT: s_or_b32 s2, s2, s3
; GFX11-FAKE16-NEXT: s_or_b32 s3, s16, s17
+; GFX11-FAKE16-NEXT: s_and_b32 s0, s0, 0xffff
+; GFX11-FAKE16-NEXT: s_lshl_b32 s1, s1, 16
; GFX11-FAKE16-NEXT: s_and_b32 s2, s2, 0xffff
; GFX11-FAKE16-NEXT: s_lshl_b32 s3, s3, 16
-; GFX11-FAKE16-NEXT: s_and_b32 s16, s73, 0xff
+; GFX11-FAKE16-NEXT: s_or_b32 s0, s0, s1
; GFX11-FAKE16-NEXT: s_or_b32 s1, s2, s3
; GFX11-FAKE16-NEXT: s_and_b32 s2, s26, 0xff
-; GFX11-FAKE16-NEXT: s_lshl_b32 s3, s96, 8
+; GFX11-FAKE16-NEXT: s_lshl_b32 s3, s37, 8
+; GFX11-FAKE16-NEXT: s_and_b32 s16, s36, 0xff
; GFX11-FAKE16-NEXT: s_lshl_b32 s17, s76, 8
; GFX11-FAKE16-NEXT: s_or_b32 s2, s2, s3
; GFX11-FAKE16-NEXT: s_or_b32 s3, s16, s17
; GFX11-FAKE16-NEXT: s_and_b32 s16, s27, 0xff
-; GFX11-FAKE16-NEXT: s_lshl_b32 s17, s87, 8
-; GFX11-FAKE16-NEXT: s_and_b32 s2, s2, 0xffff
+; GFX11-FAKE16-NEXT: s_lshl_b32 s17, s35, 8
+; GFX11-FAKE16-NEXT: s_and_b32 s18, s34, 0xff
+; GFX11-FAKE16-NEXT: s_lshl_b32 s19, vcc_hi, 8
; GFX11-FAKE16-NEXT: s_or_b32 s16, s16, s17
; GFX11-FAKE16-NEXT: s_or_b32 s17, s18, s19
+; GFX11-FAKE16-NEXT: s_and_b32 s2, s2, 0xffff
; GFX11-FAKE16-NEXT: s_lshl_b32 s3, s3, 16
; GFX11-FAKE16-NEXT: s_and_b32 s16, s16, 0xffff
; GFX11-FAKE16-NEXT: s_lshl_b32 s17, s17, 16
; GFX11-FAKE16-NEXT: s_or_b32 s2, s2, s3
; GFX11-FAKE16-NEXT: s_or_b32 s3, s16, s17
-; GFX11-FAKE16-NEXT: v_readlane_b32 s16, v19, 0
+; GFX11-FAKE16-NEXT: v_readlane_b32 s16, v18, 8
+; GFX11-FAKE16-NEXT: v_readlane_b32 s17, v18, 7
+; GFX11-FAKE16-NEXT: s_clause 0x1
+; GFX11-FAKE16-NEXT: scratch_store_b128 v0, v[1:4], off
+; GFX11-FAKE16-NEXT: scratch_store_b128 v0, v[5:8], off offset:16
; GFX11-FAKE16-NEXT: v_dual_mov_b32 v1, s0 :: v_dual_mov_b32 v2, s1
; GFX11-FAKE16-NEXT: v_dual_mov_b32 v3, s2 :: v_dual_mov_b32 v4, s3
; GFX11-FAKE16-NEXT: s_and_b32 s0, s28, 0xff
-; GFX11-FAKE16-NEXT: s_lshl_b32 s1, s85, 8
-; GFX11-FAKE16-NEXT: s_and_b32 s2, s84, 0xff
-; GFX11-FAKE16-NEXT: s_lshl_b32 s3, s16, 8
-; GFX11-FAKE16-NEXT: v_readlane_b32 s17, v19, 1
+; GFX11-FAKE16-NEXT: s_lshl_b32 s1, s104, 8
+; GFX11-FAKE16-NEXT: s_and_b32 s2, s103, 0xff
+; GFX11-FAKE16-NEXT: s_lshl_b32 s3, s72, 8
; GFX11-FAKE16-NEXT: s_or_b32 s0, s0, s1
; GFX11-FAKE16-NEXT: s_or_b32 s1, s2, s3
; GFX11-FAKE16-NEXT: s_and_b32 s2, s29, 0xff
-; GFX11-FAKE16-NEXT: s_lshl_b32 s3, s83, 8
-; GFX11-FAKE16-NEXT: s_and_b32 s16, s82, 0xff
-; GFX11-FAKE16-NEXT: s_lshl_b32 s17, s81, 8
-; GFX11-FAKE16-NEXT: v_readlane_b32 s18, v19, 2
+; GFX11-FAKE16-NEXT: s_lshl_b32 s3, s102, 8
+; GFX11-FAKE16-NEXT: s_and_b32 s16, s16, 0xff
+; GFX11-FAKE16-NEXT: s_lshl_b32 s17, s17, 8
; GFX11-FAKE16-NEXT: s_or_b32 s2, s2, s3
; GFX11-FAKE16-NEXT: s_or_b32 s3, s16, s17
; GFX11-FAKE16-NEXT: s_and_b32 s0, s0, 0xffff
@@ -87664,147 +87512,176 @@ define inreg <128 x i8> @bitcast_v16i64_to_v128i8_scalar(<16 x i64> inreg %a, i3
; GFX11-FAKE16-NEXT: s_lshl_b32 s3, s3, 16
; GFX11-FAKE16-NEXT: s_or_b32 s0, s0, s1
; GFX11-FAKE16-NEXT: s_or_b32 s1, s2, s3
+; GFX11-FAKE16-NEXT: v_readlane_b32 s3, v18, 6
+; GFX11-FAKE16-NEXT: v_readlane_b32 s16, v18, 5
; GFX11-FAKE16-NEXT: s_and_b32 s2, s40, 0xff
-; GFX11-FAKE16-NEXT: s_lshl_b32 s3, s61, 8
-; GFX11-FAKE16-NEXT: s_and_b32 s16, s80, 0xff
-; GFX11-FAKE16-NEXT: s_lshl_b32 s17, s18, 8
-; GFX11-FAKE16-NEXT: v_readlane_b32 s19, v19, 3
+; GFX11-FAKE16-NEXT: s_lshl_b32 s17, s62, 8
+; GFX11-FAKE16-NEXT: v_readlane_b32 s18, v18, 3
+; GFX11-FAKE16-NEXT: s_lshl_b32 s3, s3, 8
+; GFX11-FAKE16-NEXT: s_and_b32 s16, s16, 0xff
; GFX11-FAKE16-NEXT: s_or_b32 s2, s2, s3
; GFX11-FAKE16-NEXT: s_or_b32 s3, s16, s17
+; GFX11-FAKE16-NEXT: v_readlane_b32 s17, v18, 4
+; GFX11-FAKE16-NEXT: v_readlane_b32 s19, v18, 2
; GFX11-FAKE16-NEXT: s_and_b32 s16, s41, 0xff
-; GFX11-FAKE16-NEXT: s_lshl_b32 s17, s60, 8
-; GFX11-FAKE16-NEXT: s_and_b32 s18, s71, 0xff
-; GFX11-FAKE16-NEXT: s_lshl_b32 s19, s70, 8
-; GFX11-FAKE16-NEXT: s_or_b32 s16, s16, s17
-; GFX11-FAKE16-NEXT: s_or_b32 s17, s18, s19
+; GFX11-FAKE16-NEXT: s_and_b32 s18, s18, 0xff
; GFX11-FAKE16-NEXT: s_and_b32 s2, s2, 0xffff
+; GFX11-FAKE16-NEXT: s_lshl_b32 s17, s17, 8
+; GFX11-FAKE16-NEXT: s_lshl_b32 s19, s19, 8
; GFX11-FAKE16-NEXT: s_lshl_b32 s3, s3, 16
+; GFX11-FAKE16-NEXT: s_or_b32 s16, s16, s17
+; GFX11-FAKE16-NEXT: s_or_b32 s17, s18, s19
; GFX11-FAKE16-NEXT: s_and_b32 s16, s16, 0xffff
; GFX11-FAKE16-NEXT: s_lshl_b32 s17, s17, 16
; GFX11-FAKE16-NEXT: s_or_b32 s2, s2, s3
; GFX11-FAKE16-NEXT: s_or_b32 s3, s16, s17
-; GFX11-FAKE16-NEXT: v_readlane_b32 s16, v19, 4
; GFX11-FAKE16-NEXT: v_dual_mov_b32 v5, s0 :: v_dual_mov_b32 v6, s1
; GFX11-FAKE16-NEXT: v_dual_mov_b32 v7, s2 :: v_dual_mov_b32 v8, s3
+; GFX11-FAKE16-NEXT: v_readlane_b32 s1, v18, 1
+; GFX11-FAKE16-NEXT: v_readlane_b32 s2, v18, 0
; GFX11-FAKE16-NEXT: s_and_b32 s0, s14, 0xff
-; GFX11-FAKE16-NEXT: s_lshl_b32 s1, s58, 8
-; GFX11-FAKE16-NEXT: s_and_b32 s2, s59, 0xff
-; GFX11-FAKE16-NEXT: s_lshl_b32 s3, s16, 8
+; GFX11-FAKE16-NEXT: s_lshl_b32 s3, s60, 8
+; GFX11-FAKE16-NEXT: v_readlane_b32 s14, v19, 30
+; GFX11-FAKE16-NEXT: s_lshl_b32 s1, s1, 8
+; GFX11-FAKE16-NEXT: s_and_b32 s2, s2, 0xff
; GFX11-FAKE16-NEXT: s_or_b32 s0, s0, s1
; GFX11-FAKE16-NEXT: s_or_b32 s1, s2, s3
; GFX11-FAKE16-NEXT: s_and_b32 s2, s15, 0xff
-; GFX11-FAKE16-NEXT: s_lshl_b32 s3, s68, 8
-; GFX11-FAKE16-NEXT: s_and_b32 s14, s67, 0xff
-; GFX11-FAKE16-NEXT: s_lshl_b32 s15, s66, 8
-; GFX11-FAKE16-NEXT: s_or_b32 s2, s2, s3
-; GFX11-FAKE16-NEXT: s_or_b32 s3, s14, s15
-; GFX11-FAKE16-NEXT: v_readlane_b32 s14, v19, 6
+; GFX11-FAKE16-NEXT: v_readlane_b32 s3, v19, 31
+; GFX11-FAKE16-NEXT: v_readlane_b32 s15, v19, 29
+; GFX11-FAKE16-NEXT: s_and_b32 s14, s14, 0xff
; GFX11-FAKE16-NEXT: s_and_b32 s0, s0, 0xffff
; GFX11-FAKE16-NEXT: s_lshl_b32 s1, s1, 16
+; GFX11-FAKE16-NEXT: s_lshl_b32 s3, s3, 8
+; GFX11-FAKE16-NEXT: s_lshl_b32 s15, s15, 8
+; GFX11-FAKE16-NEXT: s_or_b32 s2, s2, s3
+; GFX11-FAKE16-NEXT: s_or_b32 s3, s14, s15
; GFX11-FAKE16-NEXT: s_and_b32 s2, s2, 0xffff
; GFX11-FAKE16-NEXT: s_lshl_b32 s3, s3, 16
; GFX11-FAKE16-NEXT: s_or_b32 s0, s0, s1
; GFX11-FAKE16-NEXT: s_or_b32 s1, s2, s3
; GFX11-FAKE16-NEXT: s_and_b32 s2, s12, 0xff
-; GFX11-FAKE16-NEXT: s_lshl_b32 s3, s65, 8
-; GFX11-FAKE16-NEXT: s_and_b32 s12, s64, 0xff
-; GFX11-FAKE16-NEXT: s_lshl_b32 s14, s14, 8
-; GFX11-FAKE16-NEXT: v_readlane_b32 s15, v19, 7
+; GFX11-FAKE16-NEXT: v_readlane_b32 s3, v19, 28
+; GFX11-FAKE16-NEXT: v_readlane_b32 s12, v19, 27
+; GFX11-FAKE16-NEXT: s_lshl_b32 s14, s58, 8
+; GFX11-FAKE16-NEXT: v_readlane_b32 s15, v19, 24
+; GFX11-FAKE16-NEXT: s_clause 0x1
+; GFX11-FAKE16-NEXT: scratch_store_b128 v0, v[9:12], off offset:32
+; GFX11-FAKE16-NEXT: scratch_store_b128 v0, v[1:4], off offset:48
+; GFX11-FAKE16-NEXT: s_lshl_b32 s3, s3, 8
+; GFX11-FAKE16-NEXT: s_and_b32 s12, s12, 0xff
; GFX11-FAKE16-NEXT: s_or_b32 s2, s2, s3
; GFX11-FAKE16-NEXT: s_or_b32 s3, s12, s14
; GFX11-FAKE16-NEXT: s_and_b32 s12, s13, 0xff
-; GFX11-FAKE16-NEXT: s_lshl_b32 s13, s55, 8
-; GFX11-FAKE16-NEXT: s_and_b32 s14, s54, 0xff
-; GFX11-FAKE16-NEXT: s_lshl_b32 s15, s53, 8
-; GFX11-FAKE16-NEXT: s_or_b32 s12, s12, s13
-; GFX11-FAKE16-NEXT: s_or_b32 s13, s14, s15
+; GFX11-FAKE16-NEXT: v_readlane_b32 s13, v19, 26
+; GFX11-FAKE16-NEXT: v_readlane_b32 s14, v19, 25
+; GFX11-FAKE16-NEXT: s_lshl_b32 s15, s15, 8
; GFX11-FAKE16-NEXT: s_and_b32 s2, s2, 0xffff
; GFX11-FAKE16-NEXT: s_lshl_b32 s3, s3, 16
+; GFX11-FAKE16-NEXT: s_lshl_b32 s13, s13, 8
+; GFX11-FAKE16-NEXT: s_and_b32 s14, s14, 0xff
+; GFX11-FAKE16-NEXT: s_or_b32 s12, s12, s13
+; GFX11-FAKE16-NEXT: s_or_b32 s13, s14, s15
; GFX11-FAKE16-NEXT: s_and_b32 s12, s12, 0xffff
; GFX11-FAKE16-NEXT: s_lshl_b32 s13, s13, 16
; GFX11-FAKE16-NEXT: s_or_b32 s2, s2, s3
; GFX11-FAKE16-NEXT: s_or_b32 s3, s12, s13
-; GFX11-FAKE16-NEXT: v_readlane_b32 s12, v19, 8
-; GFX11-FAKE16-NEXT: s_clause 0x1
-; GFX11-FAKE16-NEXT: scratch_store_b128 v0, v[9:12], off offset:32
-; GFX11-FAKE16-NEXT: scratch_store_b128 v0, v[1:4], off offset:48
; GFX11-FAKE16-NEXT: v_dual_mov_b32 v9, s0 :: v_dual_mov_b32 v10, s1
; GFX11-FAKE16-NEXT: v_dual_mov_b32 v11, s2 :: v_dual_mov_b32 v12, s3
+; GFX11-FAKE16-NEXT: v_readlane_b32 s1, v19, 23
+; GFX11-FAKE16-NEXT: v_readlane_b32 s2, v19, 22
; GFX11-FAKE16-NEXT: s_and_b32 s0, s10, 0xff
-; GFX11-FAKE16-NEXT: s_lshl_b32 s1, s52, 8
-; GFX11-FAKE16-NEXT: s_and_b32 s2, s51, 0xff
-; GFX11-FAKE16-NEXT: s_lshl_b32 s3, s12, 8
+; GFX11-FAKE16-NEXT: s_lshl_b32 s3, s56, 8
+; GFX11-FAKE16-NEXT: v_readlane_b32 s10, v19, 20
+; GFX11-FAKE16-NEXT: s_lshl_b32 s1, s1, 8
+; GFX11-FAKE16-NEXT: s_and_b32 s2, s2, 0xff
; GFX11-FAKE16-NEXT: s_or_b32 s0, s0, s1
; GFX11-FAKE16-NEXT: s_or_b32 s1, s2, s3
; GFX11-FAKE16-NEXT: s_and_b32 s2, s11, 0xff
-; GFX11-FAKE16-NEXT: s_lshl_b32 s3, s50, 8
-; GFX11-FAKE16-NEXT: s_and_b32 s10, s49, 0xff
-; GFX11-FAKE16-NEXT: s_lshl_b32 s11, s48, 8
-; GFX11-FAKE16-NEXT: s_or_b32 s2, s2, s3
-; GFX11-FAKE16-NEXT: s_or_b32 s3, s10, s11
-; GFX11-FAKE16-NEXT: v_readlane_b32 s10, v19, 10
+; GFX11-FAKE16-NEXT: v_readlane_b32 s3, v19, 21
+; GFX11-FAKE16-NEXT: v_readlane_b32 s11, v19, 19
+; GFX11-FAKE16-NEXT: s_and_b32 s10, s10, 0xff
; GFX11-FAKE16-NEXT: s_and_b32 s0, s0, 0xffff
; GFX11-FAKE16-NEXT: s_lshl_b32 s1, s1, 16
+; GFX11-FAKE16-NEXT: s_lshl_b32 s3, s3, 8
+; GFX11-FAKE16-NEXT: s_lshl_b32 s11, s11, 8
+; GFX11-FAKE16-NEXT: s_or_b32 s2, s2, s3
+; GFX11-FAKE16-NEXT: s_or_b32 s3, s10, s11
; GFX11-FAKE16-NEXT: s_and_b32 s2, s2, 0xffff
; GFX11-FAKE16-NEXT: s_lshl_b32 s3, s3, 16
; GFX11-FAKE16-NEXT: s_or_b32 s0, s0, s1
; GFX11-FAKE16-NEXT: s_or_b32 s1, s2, s3
; GFX11-FAKE16-NEXT: s_and_b32 s2, s8, 0xff
-; GFX11-FAKE16-NEXT: s_lshl_b32 s3, s39, 8
-; GFX11-FAKE16-NEXT: s_and_b32 s8, s38, 0xff
-; GFX11-FAKE16-NEXT: s_lshl_b32 s10, s10, 8
-; GFX11-FAKE16-NEXT: v_readlane_b32 s11, v19, 11
+; GFX11-FAKE16-NEXT: v_readlane_b32 s3, v19, 18
+; GFX11-FAKE16-NEXT: v_readlane_b32 s8, v19, 17
+; GFX11-FAKE16-NEXT: s_lshl_b32 s10, s46, 8
+; GFX11-FAKE16-NEXT: v_readlane_b32 s11, v19, 14
+; GFX11-FAKE16-NEXT: v_dual_mov_b32 v1, s0 :: v_dual_mov_b32 v2, s1
+; GFX11-FAKE16-NEXT: s_lshl_b32 s3, s3, 8
+; GFX11-FAKE16-NEXT: s_and_b32 s8, s8, 0xff
; GFX11-FAKE16-NEXT: s_or_b32 s2, s2, s3
; GFX11-FAKE16-NEXT: s_or_b32 s3, s8, s10
; GFX11-FAKE16-NEXT: s_and_b32 s8, s9, 0xff
-; GFX11-FAKE16-NEXT: s_lshl_b32 s9, s37, 8
-; GFX11-FAKE16-NEXT: s_and_b32 s10, s36, 0xff
-; GFX11-FAKE16-NEXT: s_lshl_b32 s11, s35, 8
-; GFX11-FAKE16-NEXT: s_or_b32 s8, s8, s9
-; GFX11-FAKE16-NEXT: s_or_b32 s9, s10, s11
+; GFX11-FAKE16-NEXT: v_readlane_b32 s9, v19, 16
+; GFX11-FAKE16-NEXT: v_readlane_b32 s10, v19, 15
+; GFX11-FAKE16-NEXT: s_lshl_b32 s11, s11, 8
; GFX11-FAKE16-NEXT: s_and_b32 s2, s2, 0xffff
; GFX11-FAKE16-NEXT: s_lshl_b32 s3, s3, 16
+; GFX11-FAKE16-NEXT: s_lshl_b32 s9, s9, 8
+; GFX11-FAKE16-NEXT: s_and_b32 s10, s10, 0xff
+; GFX11-FAKE16-NEXT: s_or_b32 s8, s8, s9
+; GFX11-FAKE16-NEXT: s_or_b32 s9, s10, s11
; GFX11-FAKE16-NEXT: s_and_b32 s8, s8, 0xffff
; GFX11-FAKE16-NEXT: s_lshl_b32 s9, s9, 16
; GFX11-FAKE16-NEXT: s_or_b32 s2, s2, s3
; GFX11-FAKE16-NEXT: s_or_b32 s3, s8, s9
-; GFX11-FAKE16-NEXT: v_readlane_b32 s8, v19, 12
-; GFX11-FAKE16-NEXT: v_dual_mov_b32 v1, s0 :: v_dual_mov_b32 v2, s1
+; GFX11-FAKE16-NEXT: s_delay_alu instid0(SALU_CYCLE_1)
; GFX11-FAKE16-NEXT: v_dual_mov_b32 v3, s2 :: v_dual_mov_b32 v4, s3
+; GFX11-FAKE16-NEXT: v_readlane_b32 s1, v19, 13
+; GFX11-FAKE16-NEXT: v_readlane_b32 s2, v19, 12
+; GFX11-FAKE16-NEXT: v_readlane_b32 s8, v19, 0
; GFX11-FAKE16-NEXT: s_and_b32 s0, s6, 0xff
-; GFX11-FAKE16-NEXT: s_lshl_b32 s1, s56, 8
-; GFX11-FAKE16-NEXT: s_and_b32 s2, s57, 0xff
+; GFX11-FAKE16-NEXT: v_readlane_b32 s6, v19, 10
+; GFX11-FAKE16-NEXT: s_lshl_b32 s1, s1, 8
+; GFX11-FAKE16-NEXT: s_and_b32 s2, s2, 0xff
; GFX11-FAKE16-NEXT: s_lshl_b32 s3, s8, 8
; GFX11-FAKE16-NEXT: s_or_b32 s0, s0, s1
; GFX11-FAKE16-NEXT: s_or_b32 s1, s2, s3
; GFX11-FAKE16-NEXT: s_and_b32 s2, s7, 0xff
-; GFX11-FAKE16-NEXT: s_lshl_b32 s3, s34, 8
-; GFX11-FAKE16-NEXT: s_and_b32 s6, vcc_hi, 0xff
-; GFX11-FAKE16-NEXT: s_lshl_b32 s7, s46, 8
-; GFX11-FAKE16-NEXT: s_or_b32 s2, s2, s3
-; GFX11-FAKE16-NEXT: s_or_b32 s3, s6, s7
-; GFX11-FAKE16-NEXT: v_readlane_b32 s6, v19, 14
+; GFX11-FAKE16-NEXT: v_readlane_b32 s3, v19, 11
+; GFX11-FAKE16-NEXT: v_readlane_b32 s7, v19, 9
+; GFX11-FAKE16-NEXT: s_and_b32 s6, s6, 0xff
; GFX11-FAKE16-NEXT: s_and_b32 s0, s0, 0xffff
; GFX11-FAKE16-NEXT: s_lshl_b32 s1, s1, 16
+; GFX11-FAKE16-NEXT: s_lshl_b32 s3, s3, 8
+; GFX11-FAKE16-NEXT: s_lshl_b32 s7, s7, 8
+; GFX11-FAKE16-NEXT: s_or_b32 s2, s2, s3
+; GFX11-FAKE16-NEXT: s_or_b32 s3, s6, s7
; GFX11-FAKE16-NEXT: s_and_b32 s2, s2, 0xffff
; GFX11-FAKE16-NEXT: s_lshl_b32 s3, s3, 16
; GFX11-FAKE16-NEXT: s_or_b32 s0, s0, s1
; GFX11-FAKE16-NEXT: s_or_b32 s1, s2, s3
; GFX11-FAKE16-NEXT: s_and_b32 s2, s4, 0xff
-; GFX11-FAKE16-NEXT: s_lshl_b32 s3, s47, 8
-; GFX11-FAKE16-NEXT: s_and_b32 s4, s104, 0xff
+; GFX11-FAKE16-NEXT: v_readlane_b32 s3, v19, 8
+; GFX11-FAKE16-NEXT: v_readlane_b32 s4, v19, 7
+; GFX11-FAKE16-NEXT: v_readlane_b32 s6, v19, 2
+; GFX11-FAKE16-NEXT: v_readlane_b32 s7, v19, 3
+; GFX11-FAKE16-NEXT: v_readlane_b32 s7, v19, 4
+; GFX11-FAKE16-NEXT: s_lshl_b32 s3, s3, 8
+; GFX11-FAKE16-NEXT: s_and_b32 s4, s4, 0xff
; GFX11-FAKE16-NEXT: s_lshl_b32 s6, s6, 8
-; GFX11-FAKE16-NEXT: v_readlane_b32 s7, v19, 15
; GFX11-FAKE16-NEXT: s_or_b32 s2, s2, s3
; GFX11-FAKE16-NEXT: s_or_b32 s3, s4, s6
; GFX11-FAKE16-NEXT: s_and_b32 s4, s5, 0xff
-; GFX11-FAKE16-NEXT: s_lshl_b32 s5, s103, 8
-; GFX11-FAKE16-NEXT: s_and_b32 s6, s102, 0xff
-; GFX11-FAKE16-NEXT: s_lshl_b32 s7, s101, 8
-; GFX11-FAKE16-NEXT: s_or_b32 s4, s4, s5
-; GFX11-FAKE16-NEXT: s_or_b32 s5, s6, s7
+; GFX11-FAKE16-NEXT: v_readlane_b32 s5, v19, 6
+; GFX11-FAKE16-NEXT: v_readlane_b32 s6, v19, 5
+; GFX11-FAKE16-NEXT: s_lshl_b32 s7, s7, 8
; GFX11-FAKE16-NEXT: s_and_b32 s2, s2, 0xffff
; GFX11-FAKE16-NEXT: s_lshl_b32 s3, s3, 16
+; GFX11-FAKE16-NEXT: s_lshl_b32 s5, s5, 8
+; GFX11-FAKE16-NEXT: s_and_b32 s6, s6, 0xff
+; GFX11-FAKE16-NEXT: s_or_b32 s4, s4, s5
+; GFX11-FAKE16-NEXT: s_or_b32 s5, s6, s7
; GFX11-FAKE16-NEXT: s_and_b32 s4, s4, 0xffff
; GFX11-FAKE16-NEXT: s_lshl_b32 s5, s5, 16
; GFX11-FAKE16-NEXT: s_or_b32 s2, s2, s3
@@ -87812,9 +87689,7 @@ define inreg <128 x i8> @bitcast_v16i64_to_v128i8_scalar(<16 x i64> inreg %a, i3
; GFX11-FAKE16-NEXT: scratch_store_b128 v0, v[5:8], off offset:64
; GFX11-FAKE16-NEXT: v_dual_mov_b32 v5, s0 :: v_dual_mov_b32 v6, s1
; GFX11-FAKE16-NEXT: v_dual_mov_b32 v7, s2 :: v_dual_mov_b32 v8, s3
-; GFX11-FAKE16-NEXT: v_readlane_b32 s17, v19, 5
-; GFX11-FAKE16-NEXT: v_readlane_b32 s13, v19, 9
-; GFX11-FAKE16-NEXT: v_readlane_b32 s9, v19, 13
+; GFX11-FAKE16-NEXT: v_readlane_b32 s9, v19, 1
; GFX11-FAKE16-NEXT: s_clause 0x2
; GFX11-FAKE16-NEXT: scratch_store_b128 v0, v[9:12], off offset:80
; GFX11-FAKE16-NEXT: scratch_store_b128 v0, v[1:4], off offset:96
@@ -87823,8 +87698,13 @@ define inreg <128 x i8> @bitcast_v16i64_to_v128i8_scalar(<16 x i64> inreg %a, i3
; GFX11-FAKE16-NEXT: v_readlane_b32 s103, v17, 7
; GFX11-FAKE16-NEXT: v_readlane_b32 s102, v17, 6
; GFX11-FAKE16-NEXT: v_readlane_b32 s101, v17, 5
+; GFX11-FAKE16-NEXT: v_readlane_b32 s100, v17, 4
+; GFX11-FAKE16-NEXT: v_readlane_b32 s99, v17, 3
+; GFX11-FAKE16-NEXT: v_readlane_b32 s98, v17, 2
+; GFX11-FAKE16-NEXT: v_readlane_b32 s97, v17, 1
; GFX11-FAKE16-NEXT: v_readlane_b32 s96, v17, 0
; GFX11-FAKE16-NEXT: v_readlane_b32 s87, v16, 31
+; GFX11-FAKE16-NEXT: v_readlane_b32 s86, v16, 30
; GFX11-FAKE16-NEXT: v_readlane_b32 s85, v16, 29
; GFX11-FAKE16-NEXT: v_readlane_b32 s84, v16, 28
; GFX11-FAKE16-NEXT: v_readlane_b32 s83, v16, 27
@@ -87833,6 +87713,7 @@ define inreg <128 x i8> @bitcast_v16i64_to_v128i8_scalar(<16 x i64> inreg %a, i3
; GFX11-FAKE16-NEXT: v_readlane_b32 s80, v16, 24
; GFX11-FAKE16-NEXT: v_readlane_b32 s71, v16, 23
; GFX11-FAKE16-NEXT: v_readlane_b32 s70, v16, 22
+; GFX11-FAKE16-NEXT: v_readlane_b32 s69, v16, 21
; GFX11-FAKE16-NEXT: v_readlane_b32 s68, v16, 20
; GFX11-FAKE16-NEXT: v_readlane_b32 s67, v16, 19
; GFX11-FAKE16-NEXT: v_readlane_b32 s66, v16, 18
@@ -87852,6 +87733,8 @@ define inreg <128 x i8> @bitcast_v16i64_to_v128i8_scalar(<16 x i64> inreg %a, i3
; GFX11-FAKE16-NEXT: v_readlane_b32 s36, v16, 4
; GFX11-FAKE16-NEXT: v_readlane_b32 s35, v16, 3
; GFX11-FAKE16-NEXT: v_readlane_b32 s34, v16, 2
+; GFX11-FAKE16-NEXT: v_readlane_b32 s31, v16, 1
+; GFX11-FAKE16-NEXT: v_readlane_b32 s30, v16, 0
; GFX11-FAKE16-NEXT: s_xor_saveexec_b32 s0, -1
; GFX11-FAKE16-NEXT: s_clause 0x3
; GFX11-FAKE16-NEXT: scratch_load_b32 v16, off, s32
@@ -87861,6 +87744,145 @@ define inreg <128 x i8> @bitcast_v16i64_to_v128i8_scalar(<16 x i64> inreg %a, i3
; GFX11-FAKE16-NEXT: s_mov_b32 exec_lo, s0
; GFX11-FAKE16-NEXT: s_waitcnt vmcnt(0)
; GFX11-FAKE16-NEXT: s_setpc_b64 s[30:31]
+; GFX11-FAKE16-NEXT: .LBB57_4:
+; GFX11-FAKE16-NEXT: ; implicit-def: $sgpr46
+; GFX11-FAKE16-NEXT: ; kill: killed $sgpr46
+; GFX11-FAKE16-NEXT: ; implicit-def: $vcc_hi
+; GFX11-FAKE16-NEXT: ; implicit-def: $vcc_lo
+; GFX11-FAKE16-NEXT: ; implicit-def: $sgpr46
+; GFX11-FAKE16-NEXT: ; kill: killed $sgpr46
+; GFX11-FAKE16-NEXT: v_writelane_b32 v19, vcc_lo, 0
+; GFX11-FAKE16-NEXT: ; implicit-def: $sgpr46
+; GFX11-FAKE16-NEXT: ; kill: killed $sgpr46
+; GFX11-FAKE16-NEXT: ; implicit-def: $sgpr101
+; GFX11-FAKE16-NEXT: ; implicit-def: $sgpr45
+; GFX11-FAKE16-NEXT: ; implicit-def: $sgpr30
+; GFX11-FAKE16-NEXT: ; implicit-def: $sgpr100
+; GFX11-FAKE16-NEXT: ; implicit-def: $sgpr99
+; GFX11-FAKE16-NEXT: ; implicit-def: $sgpr98
+; GFX11-FAKE16-NEXT: ; implicit-def: $sgpr43
+; GFX11-FAKE16-NEXT: ; implicit-def: $sgpr42
+; GFX11-FAKE16-NEXT: ; implicit-def: $sgpr94
+; GFX11-FAKE16-NEXT: ; implicit-def: $sgpr97
+; GFX11-FAKE16-NEXT: ; implicit-def: $sgpr96
+; GFX11-FAKE16-NEXT: ; implicit-def: $sgpr87
+; GFX11-FAKE16-NEXT: ; implicit-def: $sgpr86
+; GFX11-FAKE16-NEXT: ; implicit-def: $sgpr85
+; GFX11-FAKE16-NEXT: ; implicit-def: $sgpr92
+; GFX11-FAKE16-NEXT: ; implicit-def: $sgpr84
+; GFX11-FAKE16-NEXT: ; implicit-def: $sgpr83
+; GFX11-FAKE16-NEXT: ; implicit-def: $sgpr82
+; GFX11-FAKE16-NEXT: ; implicit-def: $sgpr81
+; GFX11-FAKE16-NEXT: ; implicit-def: $sgpr80
+; GFX11-FAKE16-NEXT: ; implicit-def: $sgpr90
+; GFX11-FAKE16-NEXT: ; implicit-def: $sgpr71
+; GFX11-FAKE16-NEXT: ; implicit-def: $sgpr70
+; GFX11-FAKE16-NEXT: ; implicit-def: $sgpr69
+; GFX11-FAKE16-NEXT: ; implicit-def: $sgpr68
+; GFX11-FAKE16-NEXT: ; implicit-def: $sgpr67
+; GFX11-FAKE16-NEXT: ; implicit-def: $sgpr78
+; GFX11-FAKE16-NEXT: ; implicit-def: $sgpr66
+; GFX11-FAKE16-NEXT: ; implicit-def: $sgpr65
+; GFX11-FAKE16-NEXT: ; implicit-def: $sgpr64
+; GFX11-FAKE16-NEXT: ; implicit-def: $sgpr55
+; GFX11-FAKE16-NEXT: ; implicit-def: $sgpr54
+; GFX11-FAKE16-NEXT: ; implicit-def: $sgpr74
+; GFX11-FAKE16-NEXT: ; implicit-def: $sgpr53
+; GFX11-FAKE16-NEXT: ; implicit-def: $sgpr52
+; GFX11-FAKE16-NEXT: ; implicit-def: $sgpr51
+; GFX11-FAKE16-NEXT: ; implicit-def: $sgpr50
+; GFX11-FAKE16-NEXT: ; implicit-def: $sgpr49
+; GFX11-FAKE16-NEXT: ; implicit-def: $sgpr48
+; GFX11-FAKE16-NEXT: ; implicit-def: $sgpr39
+; GFX11-FAKE16-NEXT: ; implicit-def: $sgpr38
+; GFX11-FAKE16-NEXT: ; implicit-def: $sgpr37
+; GFX11-FAKE16-NEXT: ; implicit-def: $sgpr36
+; GFX11-FAKE16-NEXT: ; implicit-def: $sgpr35
+; GFX11-FAKE16-NEXT: ; implicit-def: $sgpr34
+; GFX11-FAKE16-NEXT: ; implicit-def: $sgpr104
+; GFX11-FAKE16-NEXT: ; implicit-def: $sgpr103
+; GFX11-FAKE16-NEXT: ; implicit-def: $sgpr102
+; GFX11-FAKE16-NEXT: ; implicit-def: $sgpr88
+; GFX11-FAKE16-NEXT: ; implicit-def: $sgpr76
+; GFX11-FAKE16-NEXT: ; implicit-def: $sgpr72
+; GFX11-FAKE16-NEXT: ; implicit-def: $sgpr62
+; GFX11-FAKE16-NEXT: ; implicit-def: $sgpr60
+; GFX11-FAKE16-NEXT: ; implicit-def: $sgpr58
+; GFX11-FAKE16-NEXT: ; implicit-def: $sgpr56
+; GFX11-FAKE16-NEXT: ; implicit-def: $sgpr46
+; GFX11-FAKE16-NEXT: ; kill: killed $sgpr46
+; GFX11-FAKE16-NEXT: ; implicit-def: $sgpr46
+; GFX11-FAKE16-NEXT: ; kill: killed $sgpr46
+; GFX11-FAKE16-NEXT: ; implicit-def: $sgpr46
+; GFX11-FAKE16-NEXT: ; kill: killed $sgpr46
+; GFX11-FAKE16-NEXT: ; implicit-def: $sgpr46
+; GFX11-FAKE16-NEXT: ; kill: killed $sgpr46
+; GFX11-FAKE16-NEXT: v_writelane_b32 v19, vcc_hi, 1
+; GFX11-FAKE16-NEXT: ; implicit-def: $sgpr46
+; GFX11-FAKE16-NEXT: ; kill: killed $sgpr46
+; GFX11-FAKE16-NEXT: ; implicit-def: $vcc_lo
+; GFX11-FAKE16-NEXT: ; implicit-def: $sgpr46
+; GFX11-FAKE16-NEXT: ; kill: killed $sgpr46
+; GFX11-FAKE16-NEXT: ; implicit-def: $sgpr46
+; GFX11-FAKE16-NEXT: ; kill: killed $sgpr46
+; GFX11-FAKE16-NEXT: ; implicit-def: $sgpr46
+; GFX11-FAKE16-NEXT: ; kill: killed $sgpr46
+; GFX11-FAKE16-NEXT: ; implicit-def: $sgpr46
+; GFX11-FAKE16-NEXT: ; kill: killed $sgpr46
+; GFX11-FAKE16-NEXT: v_writelane_b32 v19, vcc_lo, 2
+; GFX11-FAKE16-NEXT: ; implicit-def: $sgpr46
+; GFX11-FAKE16-NEXT: ; kill: killed $sgpr46
+; GFX11-FAKE16-NEXT: ; implicit-def: $sgpr46
+; GFX11-FAKE16-NEXT: ; kill: killed $sgpr46
+; GFX11-FAKE16-NEXT: ; implicit-def: $sgpr46
+; GFX11-FAKE16-NEXT: ; kill: killed $sgpr46
+; GFX11-FAKE16-NEXT: ; implicit-def: $sgpr46
+; GFX11-FAKE16-NEXT: ; kill: killed $sgpr46
+; GFX11-FAKE16-NEXT: ; implicit-def: $sgpr46
+; GFX11-FAKE16-NEXT: ; kill: killed $sgpr46
+; GFX11-FAKE16-NEXT: v_writelane_b32 v19, vcc_hi, 3
+; GFX11-FAKE16-NEXT: ; implicit-def: $sgpr46
+; GFX11-FAKE16-NEXT: ; kill: killed $sgpr46
+; GFX11-FAKE16-NEXT: ; implicit-def: $sgpr46
+; GFX11-FAKE16-NEXT: ; kill: killed $sgpr46
+; GFX11-FAKE16-NEXT: ; implicit-def: $sgpr46
+; GFX11-FAKE16-NEXT: ; kill: killed $sgpr46
+; GFX11-FAKE16-NEXT: ; implicit-def: $sgpr46
+; GFX11-FAKE16-NEXT: ; kill: killed $sgpr46
+; GFX11-FAKE16-NEXT: ; implicit-def: $sgpr46
+; GFX11-FAKE16-NEXT: ; kill: killed $sgpr46
+; GFX11-FAKE16-NEXT: ; implicit-def: $sgpr46
+; GFX11-FAKE16-NEXT: ; kill: killed $sgpr46
+; GFX11-FAKE16-NEXT: ; implicit-def: $sgpr46
+; GFX11-FAKE16-NEXT: ; kill: killed $sgpr46
+; GFX11-FAKE16-NEXT: ; implicit-def: $sgpr46
+; GFX11-FAKE16-NEXT: ; kill: killed $sgpr46
+; GFX11-FAKE16-NEXT: ; implicit-def: $sgpr46
+; GFX11-FAKE16-NEXT: ; kill: killed $sgpr46
+; GFX11-FAKE16-NEXT: ; implicit-def: $sgpr46
+; GFX11-FAKE16-NEXT: ; kill: killed $sgpr46
+; GFX11-FAKE16-NEXT: ; implicit-def: $sgpr46
+; GFX11-FAKE16-NEXT: ; kill: killed $sgpr46
+; GFX11-FAKE16-NEXT: ; implicit-def: $sgpr46
+; GFX11-FAKE16-NEXT: ; kill: killed $sgpr46
+; GFX11-FAKE16-NEXT: ; implicit-def: $sgpr46
+; GFX11-FAKE16-NEXT: ; kill: killed $sgpr46
+; GFX11-FAKE16-NEXT: ; implicit-def: $sgpr46
+; GFX11-FAKE16-NEXT: ; kill: killed $sgpr46
+; GFX11-FAKE16-NEXT: ; implicit-def: $sgpr46
+; GFX11-FAKE16-NEXT: ; kill: killed $sgpr46
+; GFX11-FAKE16-NEXT: ; implicit-def: $sgpr46
+; GFX11-FAKE16-NEXT: ; kill: killed $sgpr46
+; GFX11-FAKE16-NEXT: ; implicit-def: $sgpr46
+; GFX11-FAKE16-NEXT: ; kill: killed $sgpr46
+; GFX11-FAKE16-NEXT: ; implicit-def: $sgpr46
+; GFX11-FAKE16-NEXT: ; kill: killed $sgpr46
+; GFX11-FAKE16-NEXT: ; implicit-def: $sgpr46
+; GFX11-FAKE16-NEXT: ; kill: killed $sgpr46
+; GFX11-FAKE16-NEXT: ; implicit-def: $sgpr46
+; GFX11-FAKE16-NEXT: ; kill: killed $sgpr46
+; GFX11-FAKE16-NEXT: ; implicit-def: $sgpr46
+; GFX11-FAKE16-NEXT: s_branch .LBB57_2
%cmp = icmp eq i32 %b, 0
br i1 %cmp, label %cmp.true, label %cmp.false
@@ -95160,7 +95182,7 @@ define inreg <16 x i64> @bitcast_v128i8_to_v16i64_scalar(<128 x i8> inreg %a, i3
; VI-NEXT: buffer_store_dword v10, off, s[0:3], s32 offset:476 ; 4-byte Folded Spill
; VI-NEXT: buffer_store_dword v8, off, s[0:3], s32 offset:484 ; 4-byte Folded Spill
; VI-NEXT: buffer_store_dword v6, off, s[0:3], s32 offset:488 ; 4-byte Folded Spill
-; VI-NEXT: buffer_store_dword v0, off, s[0:3], s32 offset:556 ; 4-byte Folded Spill
+; VI-NEXT: buffer_store_dword v0, off, s[0:3], s32 offset:560 ; 4-byte Folded Spill
; VI-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:332
; VI-NEXT: buffer_load_ushort v44, off, s[0:3], s32
; VI-NEXT: buffer_load_ushort v43, off, s[0:3], s32 offset:8
@@ -95186,7 +95208,7 @@ define inreg <16 x i64> @bitcast_v128i8_to_v16i64_scalar(<128 x i8> inreg %a, i3
; VI-NEXT: buffer_load_ushort v37, off, s[0:3], s32 offset:168
; VI-NEXT: buffer_load_ushort v38, off, s[0:3], s32 offset:176
; VI-NEXT: v_lshlrev_b32_e32 v1, 8, v1
-; VI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:560 ; 4-byte Folded Spill
+; VI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:564 ; 4-byte Folded Spill
; VI-NEXT: v_lshlrev_b32_e32 v1, 8, v15
; VI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:548 ; 4-byte Folded Spill
; VI-NEXT: v_lshlrev_b32_e32 v1, 8, v17
@@ -95195,11 +95217,11 @@ define inreg <16 x i64> @bitcast_v128i8_to_v16i64_scalar(<128 x i8> inreg %a, i3
; VI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:540 ; 4-byte Folded Spill
; VI-NEXT: v_lshlrev_b32_e32 v1, 8, v21
; VI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:536 ; 4-byte Folded Spill
-; VI-NEXT: v_lshlrev_b32_e32 v8, 8, v3
-; VI-NEXT: v_lshlrev_b32_e32 v59, 8, v5
-; VI-NEXT: v_lshlrev_b32_e32 v14, 8, v7
-; VI-NEXT: v_lshlrev_b32_e32 v10, 8, v9
-; VI-NEXT: v_lshlrev_b32_e32 v16, 8, v11
+; VI-NEXT: v_lshlrev_b32_e32 v56, 8, v3
+; VI-NEXT: v_lshlrev_b32_e32 v8, 8, v5
+; VI-NEXT: v_lshlrev_b32_e32 v10, 8, v7
+; VI-NEXT: v_lshlrev_b32_e32 v14, 8, v9
+; VI-NEXT: v_lshlrev_b32_e32 v12, 8, v11
; VI-NEXT: v_lshlrev_b32_e32 v6, 8, v13
; VI-NEXT: s_waitcnt vmcnt(14)
; VI-NEXT: v_cmp_ne_u32_e32 vcc, 0, v0
@@ -95220,7 +95242,7 @@ define inreg <16 x i64> @bitcast_v128i8_to_v16i64_scalar(<128 x i8> inreg %a, i3
; VI-NEXT: v_lshlrev_b32_e32 v0, 8, v41
; VI-NEXT: buffer_store_dword v0, off, s[0:3], s32 offset:504 ; 4-byte Folded Spill
; VI-NEXT: v_lshlrev_b32_e32 v0, 8, v40
-; VI-NEXT: buffer_store_dword v0, off, s[0:3], s32 offset:580 ; 4-byte Folded Spill
+; VI-NEXT: buffer_store_dword v0, off, s[0:3], s32 offset:588 ; 4-byte Folded Spill
; VI-NEXT: v_lshlrev_b32_e32 v0, 8, v55
; VI-NEXT: buffer_store_dword v0, off, s[0:3], s32 offset:500 ; 4-byte Folded Spill
; VI-NEXT: v_lshlrev_b32_e32 v0, 8, v54
@@ -95228,22 +95250,22 @@ define inreg <16 x i64> @bitcast_v128i8_to_v16i64_scalar(<128 x i8> inreg %a, i3
; VI-NEXT: v_lshlrev_b32_e32 v0, 8, v53
; VI-NEXT: buffer_store_dword v0, off, s[0:3], s32 offset:496 ; 4-byte Folded Spill
; VI-NEXT: v_lshlrev_b32_e32 v0, 8, v52
-; VI-NEXT: buffer_store_dword v0, off, s[0:3], s32 offset:608 ; 4-byte Folded Spill
+; VI-NEXT: buffer_store_dword v0, off, s[0:3], s32 offset:616 ; 4-byte Folded Spill
; VI-NEXT: v_lshlrev_b32_e32 v0, 8, v51
; VI-NEXT: buffer_store_dword v0, off, s[0:3], s32 offset:492 ; 4-byte Folded Spill
; VI-NEXT: v_lshlrev_b32_e32 v0, 8, v50
-; VI-NEXT: buffer_store_dword v0, off, s[0:3], s32 offset:600 ; 4-byte Folded Spill
+; VI-NEXT: buffer_store_dword v0, off, s[0:3], s32 offset:612 ; 4-byte Folded Spill
; VI-NEXT: v_lshlrev_b32_e32 v0, 8, v49
-; VI-NEXT: buffer_store_dword v0, off, s[0:3], s32 offset:604 ; 4-byte Folded Spill
+; VI-NEXT: buffer_store_dword v0, off, s[0:3], s32 offset:608 ; 4-byte Folded Spill
; VI-NEXT: v_lshlrev_b32_e32 v0, 8, v48
-; VI-NEXT: buffer_store_dword v0, off, s[0:3], s32 offset:596 ; 4-byte Folded Spill
+; VI-NEXT: buffer_store_dword v0, off, s[0:3], s32 offset:604 ; 4-byte Folded Spill
; VI-NEXT: v_lshlrev_b32_e32 v0, 8, v39
; VI-NEXT: buffer_store_dword v0, off, s[0:3], s32 offset:480 ; 4-byte Folded Spill
; VI-NEXT: s_waitcnt vmcnt(14)
; VI-NEXT: v_lshlrev_b32_e32 v0, 8, v22
-; VI-NEXT: buffer_store_dword v0, off, s[0:3], s32 offset:572 ; 4-byte Folded Spill
+; VI-NEXT: buffer_store_dword v0, off, s[0:3], s32 offset:596 ; 4-byte Folded Spill
; VI-NEXT: v_lshlrev_b32_e32 v0, 8, v31
-; VI-NEXT: buffer_store_dword v0, off, s[0:3], s32 offset:588 ; 4-byte Folded Spill
+; VI-NEXT: buffer_store_dword v0, off, s[0:3], s32 offset:600 ; 4-byte Folded Spill
; VI-NEXT: v_lshlrev_b32_e32 v0, 8, v32
; VI-NEXT: buffer_store_dword v0, off, s[0:3], s32 offset:592 ; 4-byte Folded Spill
; VI-NEXT: v_lshlrev_b32_e32 v0, 8, v33
@@ -95266,23 +95288,24 @@ define inreg <16 x i64> @bitcast_v128i8_to_v16i64_scalar(<128 x i8> inreg %a, i3
; VI-NEXT: v_lshlrev_b32_e32 v31, 8, v38
; VI-NEXT: s_and_b64 s[4:5], vcc, exec
; VI-NEXT: s_waitcnt vmcnt(7)
-; VI-NEXT: v_lshlrev_b32_e32 v26, 8, v0
+; VI-NEXT: v_lshlrev_b32_e32 v22, 8, v0
; VI-NEXT: s_waitcnt vmcnt(6)
; VI-NEXT: v_lshlrev_b32_e32 v32, 8, v1
; VI-NEXT: s_waitcnt vmcnt(5)
-; VI-NEXT: v_lshlrev_b32_e32 v54, 8, v13
+; VI-NEXT: v_lshlrev_b32_e32 v23, 8, v13
+; VI-NEXT: s_waitcnt vmcnt(4)
+; VI-NEXT: v_lshlrev_b32_e32 v0, 8, v3
+; VI-NEXT: buffer_store_dword v0, off, s[0:3], s32 offset:580 ; 4-byte Folded Spill
; VI-NEXT: s_waitcnt vmcnt(4)
-; VI-NEXT: v_lshlrev_b32_e32 v49, 8, v3
-; VI-NEXT: s_waitcnt vmcnt(3)
; VI-NEXT: v_lshlrev_b32_e32 v0, 8, v11
-; VI-NEXT: buffer_store_dword v0, off, s[0:3], s32 offset:576 ; 4-byte Folded Spill
-; VI-NEXT: s_waitcnt vmcnt(3)
+; VI-NEXT: buffer_store_dword v0, off, s[0:3], s32 offset:572 ; 4-byte Folded Spill
+; VI-NEXT: s_waitcnt vmcnt(4)
; VI-NEXT: v_lshlrev_b32_e32 v0, 8, v5
-; VI-NEXT: buffer_store_dword v0, off, s[0:3], s32 offset:612 ; 4-byte Folded Spill
-; VI-NEXT: s_waitcnt vmcnt(3)
+; VI-NEXT: buffer_store_dword v0, off, s[0:3], s32 offset:576 ; 4-byte Folded Spill
+; VI-NEXT: s_waitcnt vmcnt(4)
; VI-NEXT: v_lshlrev_b32_e32 v0, 8, v9
; VI-NEXT: buffer_store_dword v0, off, s[0:3], s32 offset:416 ; 4-byte Folded Spill
-; VI-NEXT: s_waitcnt vmcnt(3)
+; VI-NEXT: s_waitcnt vmcnt(4)
; VI-NEXT: v_lshlrev_b32_e32 v0, 8, v7
; VI-NEXT: buffer_store_dword v0, off, s[0:3], s32 offset:412 ; 4-byte Folded Spill
; VI-NEXT: buffer_load_ushort v0, off, s[0:3], s32 offset:248
@@ -95295,37 +95318,37 @@ define inreg <16 x i64> @bitcast_v128i8_to_v16i64_scalar(<128 x i8> inreg %a, i3
; VI-NEXT: buffer_load_ushort v7, off, s[0:3], s32 offset:304
; VI-NEXT: s_waitcnt vmcnt(7)
; VI-NEXT: v_lshlrev_b32_e32 v0, 8, v0
-; VI-NEXT: buffer_store_dword v0, off, s[0:3], s32 offset:564 ; 4-byte Folded Spill
+; VI-NEXT: buffer_store_dword v0, off, s[0:3], s32 offset:568 ; 4-byte Folded Spill
; VI-NEXT: s_waitcnt vmcnt(7)
; VI-NEXT: v_lshlrev_b32_e32 v48, 8, v1
; VI-NEXT: s_waitcnt vmcnt(6)
; VI-NEXT: v_lshlrev_b32_e32 v27, 8, v13
; VI-NEXT: s_waitcnt vmcnt(4)
; VI-NEXT: v_lshlrev_b32_e32 v0, 8, v11
-; VI-NEXT: buffer_store_dword v0, off, s[0:3], s32 offset:568 ; 4-byte Folded Spill
+; VI-NEXT: buffer_store_dword v0, off, s[0:3], s32 offset:556 ; 4-byte Folded Spill
; VI-NEXT: s_waitcnt vmcnt(4)
; VI-NEXT: v_lshlrev_b32_e32 v0, 8, v5
; VI-NEXT: buffer_store_dword v0, off, s[0:3], s32 offset:404 ; 4-byte Folded Spill
-; VI-NEXT: v_lshlrev_b32_e32 v29, 8, v3
+; VI-NEXT: v_lshlrev_b32_e32 v28, 8, v3
; VI-NEXT: buffer_load_ushort v0, off, s[0:3], s32 offset:312
; VI-NEXT: buffer_load_ushort v1, off, s[0:3], s32 offset:320
; VI-NEXT: buffer_load_ushort v3, off, s[0:3], s32 offset:328
; VI-NEXT: buffer_load_ushort v11, off, s[0:3], s32 offset:4
-; VI-NEXT: buffer_load_ushort v12, off, s[0:3], s32 offset:12
+; VI-NEXT: buffer_load_ushort v16, off, s[0:3], s32 offset:12
; VI-NEXT: buffer_load_ushort v50, off, s[0:3], s32 offset:20
-; VI-NEXT: buffer_load_ushort v56, off, s[0:3], s32 offset:28
+; VI-NEXT: buffer_load_ushort v59, off, s[0:3], s32 offset:28
; VI-NEXT: buffer_load_ushort v13, off, s[0:3], s32 offset:36
; VI-NEXT: s_waitcnt vmcnt(11)
; VI-NEXT: v_lshlrev_b32_e32 v5, 8, v7
-; VI-NEXT: v_lshlrev_b32_e32 v9, 8, v9
+; VI-NEXT: v_lshlrev_b32_e32 v29, 8, v9
; VI-NEXT: s_waitcnt vmcnt(7)
; VI-NEXT: v_lshlrev_b32_e32 v7, 8, v0
; VI-NEXT: s_waitcnt vmcnt(6)
; VI-NEXT: v_lshlrev_b32_e32 v0, 8, v1
; VI-NEXT: buffer_store_dword v0, off, s[0:3], s32 offset:552 ; 4-byte Folded Spill
; VI-NEXT: buffer_load_ushort v18, off, s[0:3], s32 offset:44
-; VI-NEXT: buffer_load_ushort v15, off, s[0:3], s32 offset:52
-; VI-NEXT: buffer_load_ushort v17, off, s[0:3], s32 offset:60
+; VI-NEXT: buffer_load_ushort v17, off, s[0:3], s32 offset:52
+; VI-NEXT: buffer_load_ushort v15, off, s[0:3], s32 offset:60
; VI-NEXT: buffer_load_ushort v39, off, s[0:3], s32 offset:68
; VI-NEXT: buffer_load_ushort v38, off, s[0:3], s32 offset:76
; VI-NEXT: buffer_load_ushort v37, off, s[0:3], s32 offset:84
@@ -95333,23 +95356,23 @@ define inreg <16 x i64> @bitcast_v128i8_to_v16i64_scalar(<128 x i8> inreg %a, i3
; VI-NEXT: buffer_load_ushort v36, off, s[0:3], s32 offset:100
; VI-NEXT: buffer_load_ushort v35, off, s[0:3], s32 offset:108
; VI-NEXT: buffer_load_ushort v33, off, s[0:3], s32 offset:116
-; VI-NEXT: buffer_load_ushort v25, off, s[0:3], s32 offset:124
+; VI-NEXT: buffer_load_ushort v54, off, s[0:3], s32 offset:124
; VI-NEXT: buffer_load_ushort v19, off, s[0:3], s32 offset:132
-; VI-NEXT: buffer_load_ushort v51, off, s[0:3], s32 offset:140
+; VI-NEXT: buffer_load_ushort v49, off, s[0:3], s32 offset:140
; VI-NEXT: buffer_load_ushort v20, off, s[0:3], s32 offset:148
-; VI-NEXT: buffer_load_ushort v28, off, s[0:3], s32 offset:156
+; VI-NEXT: buffer_load_ushort v51, off, s[0:3], s32 offset:156
; VI-NEXT: buffer_load_ushort v21, off, s[0:3], s32 offset:164
; VI-NEXT: buffer_load_ushort v34, off, s[0:3], s32 offset:172
-; VI-NEXT: buffer_load_ushort v22, off, s[0:3], s32 offset:180
-; VI-NEXT: buffer_load_ushort v23, off, s[0:3], s32 offset:188
+; VI-NEXT: buffer_load_ushort v25, off, s[0:3], s32 offset:180
+; VI-NEXT: buffer_load_ushort v30, off, s[0:3], s32 offset:188
; VI-NEXT: buffer_load_ushort v43, off, s[0:3], s32 offset:196
-; VI-NEXT: buffer_load_ushort v30, off, s[0:3], s32 offset:204
-; VI-NEXT: buffer_load_ushort v24, off, s[0:3], s32 offset:212
+; VI-NEXT: buffer_load_ushort v24, off, s[0:3], s32 offset:204
+; VI-NEXT: buffer_load_ushort v26, off, s[0:3], s32 offset:212
; VI-NEXT: buffer_load_ushort v46, off, s[0:3], s32 offset:220
; VI-NEXT: buffer_load_ushort v61, off, s[0:3], s32 offset:228
; VI-NEXT: buffer_load_ushort v47, off, s[0:3], s32 offset:236
-; VI-NEXT: buffer_load_ushort v45, off, s[0:3], s32 offset:244
-; VI-NEXT: buffer_load_ushort v58, off, s[0:3], s32 offset:252
+; VI-NEXT: buffer_load_ushort v58, off, s[0:3], s32 offset:244
+; VI-NEXT: buffer_load_ushort v45, off, s[0:3], s32 offset:252
; VI-NEXT: buffer_load_ushort v44, off, s[0:3], s32 offset:260
; VI-NEXT: buffer_load_ushort v63, off, s[0:3], s32 offset:268
; VI-NEXT: buffer_load_ushort v42, off, s[0:3], s32 offset:276
@@ -95362,57 +95385,57 @@ define inreg <16 x i64> @bitcast_v128i8_to_v16i64_scalar(<128 x i8> inreg %a, i3
; VI-NEXT: s_waitcnt vmcnt(14)
; VI-NEXT: v_lshlrev_b32_e32 v0, 8, v3
; VI-NEXT: buffer_store_dword v5, off, s[0:3], s32 offset:400 ; 4-byte Folded Spill
-; VI-NEXT: buffer_store_dword v9, off, s[0:3], s32 offset:616 ; 4-byte Folded Spill
+; VI-NEXT: buffer_store_dword v29, off, s[0:3], s32 offset:620 ; 4-byte Folded Spill
; VI-NEXT: buffer_store_dword v0, off, s[0:3], s32 offset:408 ; 4-byte Folded Spill
-; VI-NEXT: buffer_store_dword v27, off, s[0:3], s32 offset:620 ; 4-byte Folded Spill
-; VI-NEXT: buffer_store_dword v48, off, s[0:3], s32 offset:624 ; 4-byte Folded Spill
+; VI-NEXT: buffer_store_dword v27, off, s[0:3], s32 offset:624 ; 4-byte Folded Spill
+; VI-NEXT: buffer_store_dword v48, off, s[0:3], s32 offset:628 ; 4-byte Folded Spill
; VI-NEXT: s_waitcnt vmcnt(12)
-; VI-NEXT: buffer_store_dword v63, off, s[0:3], s32 offset:628 ; 4-byte Folded Spill
-; VI-NEXT: buffer_store_dword v58, off, s[0:3], s32 offset:632 ; 4-byte Folded Spill
-; VI-NEXT: buffer_store_dword v47, off, s[0:3], s32 offset:636 ; 4-byte Folded Spill
-; VI-NEXT: buffer_store_dword v31, off, s[0:3], s32 offset:640 ; 4-byte Folded Spill
-; VI-NEXT: buffer_store_dword v46, off, s[0:3], s32 offset:644 ; 4-byte Folded Spill
-; VI-NEXT: buffer_store_dword v52, off, s[0:3], s32 offset:648 ; 4-byte Folded Spill
-; VI-NEXT: buffer_store_dword v2, off, s[0:3], s32 offset:652 ; 4-byte Folded Spill
-; VI-NEXT: buffer_store_dword v32, off, s[0:3], s32 offset:656 ; 4-byte Folded Spill
-; VI-NEXT: buffer_store_dword v4, off, s[0:3], s32 offset:660 ; 4-byte Folded Spill
-; VI-NEXT: buffer_store_dword v24, off, s[0:3], s32 offset:664 ; 4-byte Folded Spill
-; VI-NEXT: buffer_store_dword v30, off, s[0:3], s32 offset:668 ; 4-byte Folded Spill
-; VI-NEXT: buffer_store_dword v43, off, s[0:3], s32 offset:672 ; 4-byte Folded Spill
-; VI-NEXT: buffer_store_dword v23, off, s[0:3], s32 offset:676 ; 4-byte Folded Spill
-; VI-NEXT: buffer_store_dword v6, off, s[0:3], s32 offset:680 ; 4-byte Folded Spill
-; VI-NEXT: buffer_store_dword v16, off, s[0:3], s32 offset:684 ; 4-byte Folded Spill
-; VI-NEXT: buffer_store_dword v22, off, s[0:3], s32 offset:688 ; 4-byte Folded Spill
-; VI-NEXT: buffer_store_dword v34, off, s[0:3], s32 offset:692 ; 4-byte Folded Spill
-; VI-NEXT: buffer_store_dword v10, off, s[0:3], s32 offset:696 ; 4-byte Folded Spill
+; VI-NEXT: buffer_store_dword v63, off, s[0:3], s32 offset:632 ; 4-byte Folded Spill
+; VI-NEXT: buffer_store_dword v45, off, s[0:3], s32 offset:636 ; 4-byte Folded Spill
+; VI-NEXT: buffer_store_dword v47, off, s[0:3], s32 offset:640 ; 4-byte Folded Spill
+; VI-NEXT: buffer_store_dword v31, off, s[0:3], s32 offset:644 ; 4-byte Folded Spill
+; VI-NEXT: buffer_store_dword v46, off, s[0:3], s32 offset:648 ; 4-byte Folded Spill
+; VI-NEXT: buffer_store_dword v52, off, s[0:3], s32 offset:652 ; 4-byte Folded Spill
+; VI-NEXT: buffer_store_dword v2, off, s[0:3], s32 offset:656 ; 4-byte Folded Spill
+; VI-NEXT: buffer_store_dword v32, off, s[0:3], s32 offset:660 ; 4-byte Folded Spill
+; VI-NEXT: buffer_store_dword v4, off, s[0:3], s32 offset:664 ; 4-byte Folded Spill
+; VI-NEXT: buffer_store_dword v26, off, s[0:3], s32 offset:668 ; 4-byte Folded Spill
+; VI-NEXT: buffer_store_dword v24, off, s[0:3], s32 offset:672 ; 4-byte Folded Spill
+; VI-NEXT: buffer_store_dword v43, off, s[0:3], s32 offset:676 ; 4-byte Folded Spill
+; VI-NEXT: buffer_store_dword v30, off, s[0:3], s32 offset:680 ; 4-byte Folded Spill
+; VI-NEXT: buffer_store_dword v6, off, s[0:3], s32 offset:684 ; 4-byte Folded Spill
+; VI-NEXT: buffer_store_dword v12, off, s[0:3], s32 offset:688 ; 4-byte Folded Spill
+; VI-NEXT: buffer_store_dword v25, off, s[0:3], s32 offset:692 ; 4-byte Folded Spill
+; VI-NEXT: buffer_store_dword v34, off, s[0:3], s32 offset:696 ; 4-byte Folded Spill
; VI-NEXT: buffer_store_dword v14, off, s[0:3], s32 offset:700 ; 4-byte Folded Spill
-; VI-NEXT: buffer_store_dword v21, off, s[0:3], s32 offset:704 ; 4-byte Folded Spill
-; VI-NEXT: buffer_store_dword v28, off, s[0:3], s32 offset:708 ; 4-byte Folded Spill
-; VI-NEXT: buffer_store_dword v20, off, s[0:3], s32 offset:712 ; 4-byte Folded Spill
-; VI-NEXT: buffer_store_dword v51, off, s[0:3], s32 offset:716 ; 4-byte Folded Spill
-; VI-NEXT: buffer_store_dword v59, off, s[0:3], s32 offset:720 ; 4-byte Folded Spill
-; VI-NEXT: buffer_store_dword v19, off, s[0:3], s32 offset:724 ; 4-byte Folded Spill
-; VI-NEXT: buffer_store_dword v8, off, s[0:3], s32 offset:728 ; 4-byte Folded Spill
-; VI-NEXT: buffer_store_dword v25, off, s[0:3], s32 offset:732 ; 4-byte Folded Spill
-; VI-NEXT: buffer_store_dword v33, off, s[0:3], s32 offset:736 ; 4-byte Folded Spill
-; VI-NEXT: buffer_store_dword v35, off, s[0:3], s32 offset:740 ; 4-byte Folded Spill
-; VI-NEXT: buffer_store_dword v36, off, s[0:3], s32 offset:744 ; 4-byte Folded Spill
-; VI-NEXT: buffer_store_dword v57, off, s[0:3], s32 offset:748 ; 4-byte Folded Spill
-; VI-NEXT: buffer_store_dword v37, off, s[0:3], s32 offset:752 ; 4-byte Folded Spill
-; VI-NEXT: buffer_store_dword v38, off, s[0:3], s32 offset:756 ; 4-byte Folded Spill
-; VI-NEXT: buffer_store_dword v39, off, s[0:3], s32 offset:760 ; 4-byte Folded Spill
-; VI-NEXT: buffer_store_dword v17, off, s[0:3], s32 offset:764 ; 4-byte Folded Spill
+; VI-NEXT: buffer_store_dword v10, off, s[0:3], s32 offset:704 ; 4-byte Folded Spill
+; VI-NEXT: buffer_store_dword v21, off, s[0:3], s32 offset:708 ; 4-byte Folded Spill
+; VI-NEXT: buffer_store_dword v51, off, s[0:3], s32 offset:712 ; 4-byte Folded Spill
+; VI-NEXT: buffer_store_dword v20, off, s[0:3], s32 offset:716 ; 4-byte Folded Spill
+; VI-NEXT: buffer_store_dword v49, off, s[0:3], s32 offset:720 ; 4-byte Folded Spill
+; VI-NEXT: buffer_store_dword v8, off, s[0:3], s32 offset:724 ; 4-byte Folded Spill
+; VI-NEXT: buffer_store_dword v19, off, s[0:3], s32 offset:728 ; 4-byte Folded Spill
+; VI-NEXT: buffer_store_dword v56, off, s[0:3], s32 offset:732 ; 4-byte Folded Spill
+; VI-NEXT: buffer_store_dword v54, off, s[0:3], s32 offset:736 ; 4-byte Folded Spill
+; VI-NEXT: buffer_store_dword v33, off, s[0:3], s32 offset:740 ; 4-byte Folded Spill
+; VI-NEXT: buffer_store_dword v35, off, s[0:3], s32 offset:744 ; 4-byte Folded Spill
+; VI-NEXT: buffer_store_dword v36, off, s[0:3], s32 offset:748 ; 4-byte Folded Spill
+; VI-NEXT: buffer_store_dword v57, off, s[0:3], s32 offset:752 ; 4-byte Folded Spill
+; VI-NEXT: buffer_store_dword v37, off, s[0:3], s32 offset:756 ; 4-byte Folded Spill
+; VI-NEXT: buffer_store_dword v38, off, s[0:3], s32 offset:760 ; 4-byte Folded Spill
+; VI-NEXT: buffer_store_dword v39, off, s[0:3], s32 offset:764 ; 4-byte Folded Spill
; VI-NEXT: buffer_store_dword v15, off, s[0:3], s32 offset:768 ; 4-byte Folded Spill
-; VI-NEXT: buffer_store_dword v18, off, s[0:3], s32 offset:772 ; 4-byte Folded Spill
-; VI-NEXT: buffer_store_dword v13, off, s[0:3], s32 offset:776 ; 4-byte Folded Spill
-; VI-NEXT: buffer_store_dword v56, off, s[0:3], s32 offset:780 ; 4-byte Folded Spill
-; VI-NEXT: buffer_store_dword v50, off, s[0:3], s32 offset:784 ; 4-byte Folded Spill
-; VI-NEXT: buffer_store_dword v12, off, s[0:3], s32 offset:788 ; 4-byte Folded Spill
-; VI-NEXT: buffer_store_dword v11, off, s[0:3], s32 offset:792 ; 4-byte Folded Spill
+; VI-NEXT: buffer_store_dword v17, off, s[0:3], s32 offset:772 ; 4-byte Folded Spill
+; VI-NEXT: buffer_store_dword v18, off, s[0:3], s32 offset:776 ; 4-byte Folded Spill
+; VI-NEXT: buffer_store_dword v13, off, s[0:3], s32 offset:780 ; 4-byte Folded Spill
+; VI-NEXT: buffer_store_dword v59, off, s[0:3], s32 offset:784 ; 4-byte Folded Spill
+; VI-NEXT: buffer_store_dword v50, off, s[0:3], s32 offset:788 ; 4-byte Folded Spill
+; VI-NEXT: buffer_store_dword v16, off, s[0:3], s32 offset:792 ; 4-byte Folded Spill
+; VI-NEXT: buffer_store_dword v11, off, s[0:3], s32 offset:796 ; 4-byte Folded Spill
; VI-NEXT: s_cbranch_scc0 .LBB59_2
; VI-NEXT: ; %bb.1: ; %cmp.false
-; VI-NEXT: v_or_b32_sdwa v0, v2, v8 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
-; VI-NEXT: v_or_b32_sdwa v1, v4, v59 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; VI-NEXT: v_or_b32_sdwa v0, v2, v56 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; VI-NEXT: v_or_b32_sdwa v1, v4, v8 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
; VI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:488 ; 4-byte Folded Reload
; VI-NEXT: buffer_load_dword v3, off, s[0:3], s32 offset:484 ; 4-byte Folded Reload
; VI-NEXT: v_or_b32_sdwa v4, v0, v1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_0 src1_sel:DWORD
@@ -95427,12 +95450,12 @@ define inreg <16 x i64> @bitcast_v128i8_to_v16i64_scalar(<128 x i8> inreg %a, i3
; VI-NEXT: s_lshl_b32 s7, s23, 8
; VI-NEXT: s_lshl_b32 s8, s27, 8
; VI-NEXT: s_waitcnt vmcnt(3)
-; VI-NEXT: v_or_b32_sdwa v2, v2, v14 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; VI-NEXT: v_or_b32_sdwa v2, v2, v10 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
; VI-NEXT: s_waitcnt vmcnt(2)
-; VI-NEXT: v_or_b32_sdwa v3, v3, v10 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; VI-NEXT: v_or_b32_sdwa v3, v3, v14 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
; VI-NEXT: v_or_b32_sdwa v5, v2, v3 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_0 src1_sel:DWORD
; VI-NEXT: s_waitcnt vmcnt(1)
-; VI-NEXT: v_or_b32_sdwa v0, v0, v16 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; VI-NEXT: v_or_b32_sdwa v0, v0, v12 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
; VI-NEXT: s_waitcnt vmcnt(0)
; VI-NEXT: v_or_b32_sdwa v1, v1, v6 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
; VI-NEXT: v_or_b32_sdwa v6, v0, v1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_0 src1_sel:DWORD
@@ -95461,9 +95484,7 @@ define inreg <16 x i64> @bitcast_v128i8_to_v16i64_scalar(<128 x i8> inreg %a, i3
; VI-NEXT: v_or_b32_sdwa v0, v0, v1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
; VI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:432 ; 4-byte Folded Reload
; VI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:528 ; 4-byte Folded Reload
-; VI-NEXT: buffer_store_dword v29, off, s[0:3], s32 offset:796 ; 4-byte Folded Spill
-; VI-NEXT: v_mov_b32_e32 v29, v9
-; VI-NEXT: s_waitcnt vmcnt(1)
+; VI-NEXT: s_waitcnt vmcnt(0)
; VI-NEXT: v_or_b32_sdwa v1, v1, v2 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
; VI-NEXT: v_or_b32_sdwa v9, v0, v1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_0 src1_sel:DWORD
; VI-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:428 ; 4-byte Folded Reload
@@ -95486,15 +95507,15 @@ define inreg <16 x i64> @bitcast_v128i8_to_v16i64_scalar(<128 x i8> inreg %a, i3
; VI-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:508 ; 4-byte Folded Reload
; VI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:504 ; 4-byte Folded Reload
; VI-NEXT: s_waitcnt vmcnt(1)
-; VI-NEXT: v_or_b32_sdwa v0, v12, v0 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; VI-NEXT: v_or_b32_sdwa v0, v16, v0 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
; VI-NEXT: s_waitcnt vmcnt(0)
; VI-NEXT: v_or_b32_sdwa v1, v50, v1 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
; VI-NEXT: v_or_b32_sdwa v12, v0, v1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_0 src1_sel:DWORD
-; VI-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:580 ; 4-byte Folded Reload
+; VI-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:588 ; 4-byte Folded Reload
; VI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:500 ; 4-byte Folded Reload
; VI-NEXT: s_waitcnt vmcnt(1)
; VI-NEXT: v_mov_b32_e32 v50, v0
-; VI-NEXT: v_or_b32_sdwa v0, v56, v0 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; VI-NEXT: v_or_b32_sdwa v0, v59, v0 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
; VI-NEXT: s_waitcnt vmcnt(0)
; VI-NEXT: v_or_b32_sdwa v1, v13, v1 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
; VI-NEXT: v_or_b32_sdwa v13, v0, v1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_0 src1_sel:DWORD
@@ -95504,18 +95525,18 @@ define inreg <16 x i64> @bitcast_v128i8_to_v16i64_scalar(<128 x i8> inreg %a, i3
; VI-NEXT: v_mov_b32_e32 v59, v0
; VI-NEXT: v_or_b32_sdwa v0, v18, v0 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
; VI-NEXT: s_waitcnt vmcnt(0)
-; VI-NEXT: v_or_b32_sdwa v1, v15, v1 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; VI-NEXT: v_or_b32_sdwa v1, v17, v1 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
; VI-NEXT: v_or_b32_sdwa v14, v0, v1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_0 src1_sel:DWORD
-; VI-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:608 ; 4-byte Folded Reload
+; VI-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:616 ; 4-byte Folded Reload
; VI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:492 ; 4-byte Folded Reload
; VI-NEXT: s_waitcnt vmcnt(1)
; VI-NEXT: v_mov_b32_e32 v56, v0
-; VI-NEXT: v_or_b32_sdwa v0, v17, v0 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; VI-NEXT: v_or_b32_sdwa v0, v15, v0 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
; VI-NEXT: s_waitcnt vmcnt(0)
; VI-NEXT: v_or_b32_sdwa v1, v39, v1 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
; VI-NEXT: v_or_b32_sdwa v15, v0, v1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_0 src1_sel:DWORD
-; VI-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:600 ; 4-byte Folded Reload
-; VI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:604 ; 4-byte Folded Reload
+; VI-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:612 ; 4-byte Folded Reload
+; VI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:608 ; 4-byte Folded Reload
; VI-NEXT: s_waitcnt vmcnt(1)
; VI-NEXT: v_mov_b32_e32 v39, v0
; VI-NEXT: v_or_b32_sdwa v0, v38, v0 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
@@ -95523,7 +95544,7 @@ define inreg <16 x i64> @bitcast_v128i8_to_v16i64_scalar(<128 x i8> inreg %a, i3
; VI-NEXT: v_mov_b32_e32 v38, v1
; VI-NEXT: v_or_b32_sdwa v1, v37, v1 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
; VI-NEXT: v_or_b32_sdwa v16, v0, v1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_0 src1_sel:DWORD
-; VI-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:596 ; 4-byte Folded Reload
+; VI-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:604 ; 4-byte Folded Reload
; VI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:480 ; 4-byte Folded Reload
; VI-NEXT: s_waitcnt vmcnt(1)
; VI-NEXT: v_mov_b32_e32 v37, v0
@@ -95531,8 +95552,8 @@ define inreg <16 x i64> @bitcast_v128i8_to_v16i64_scalar(<128 x i8> inreg %a, i3
; VI-NEXT: s_waitcnt vmcnt(0)
; VI-NEXT: v_or_b32_sdwa v1, v36, v1 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
; VI-NEXT: v_or_b32_sdwa v17, v0, v1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_0 src1_sel:DWORD
-; VI-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:572 ; 4-byte Folded Reload
-; VI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:588 ; 4-byte Folded Reload
+; VI-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:596 ; 4-byte Folded Reload
+; VI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:600 ; 4-byte Folded Reload
; VI-NEXT: s_waitcnt vmcnt(1)
; VI-NEXT: v_mov_b32_e32 v36, v0
; VI-NEXT: v_or_b32_sdwa v0, v35, v0 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
@@ -95544,39 +95565,41 @@ define inreg <16 x i64> @bitcast_v128i8_to_v16i64_scalar(<128 x i8> inreg %a, i3
; VI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:468 ; 4-byte Folded Reload
; VI-NEXT: s_waitcnt vmcnt(1)
; VI-NEXT: v_mov_b32_e32 v33, v0
-; VI-NEXT: v_or_b32_sdwa v0, v25, v0 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; VI-NEXT: v_or_b32_sdwa v0, v54, v0 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
; VI-NEXT: s_waitcnt vmcnt(0)
; VI-NEXT: v_or_b32_sdwa v1, v19, v1 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
; VI-NEXT: v_or_b32_sdwa v19, v0, v1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_0 src1_sel:DWORD
; VI-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:456 ; 4-byte Folded Reload
; VI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:444 ; 4-byte Folded Reload
; VI-NEXT: s_waitcnt vmcnt(1)
-; VI-NEXT: v_or_b32_sdwa v0, v51, v0 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; VI-NEXT: v_or_b32_sdwa v0, v49, v0 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
; VI-NEXT: s_waitcnt vmcnt(0)
; VI-NEXT: v_or_b32_sdwa v1, v20, v1 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
; VI-NEXT: v_or_b32_sdwa v20, v0, v1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_0 src1_sel:DWORD
; VI-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:436 ; 4-byte Folded Reload
; VI-NEXT: v_or_b32_sdwa v1, v21, v52 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
-; VI-NEXT: v_mov_b32_e32 v51, v3
; VI-NEXT: s_waitcnt vmcnt(0)
-; VI-NEXT: v_or_b32_sdwa v0, v28, v0 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; VI-NEXT: v_or_b32_sdwa v0, v51, v0 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
; VI-NEXT: v_or_b32_sdwa v21, v0, v1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_0 src1_sel:DWORD
; VI-NEXT: v_or_b32_sdwa v0, v34, v31 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
-; VI-NEXT: v_or_b32_sdwa v1, v22, v26 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; VI-NEXT: v_or_b32_sdwa v1, v25, v22 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; VI-NEXT: v_mov_b32_e32 v34, v22
; VI-NEXT: v_or_b32_sdwa v22, v0, v1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_0 src1_sel:DWORD
-; VI-NEXT: v_or_b32_sdwa v0, v23, v32 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
-; VI-NEXT: v_or_b32_sdwa v1, v43, v54 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; VI-NEXT: v_or_b32_sdwa v0, v30, v32 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; VI-NEXT: v_or_b32_sdwa v1, v43, v23 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; VI-NEXT: v_mov_b32_e32 v32, v23
; VI-NEXT: v_or_b32_sdwa v23, v0, v1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_0 src1_sel:DWORD
-; VI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:576 ; 4-byte Folded Reload
-; VI-NEXT: v_mov_b32_e32 v43, v49
-; VI-NEXT: v_or_b32_sdwa v0, v30, v49 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
-; VI-NEXT: v_mov_b32_e32 v32, v54
-; VI-NEXT: v_mov_b32_e32 v34, v26
+; VI-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:580 ; 4-byte Folded Reload
+; VI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:572 ; 4-byte Folded Reload
+; VI-NEXT: v_mov_b32_e32 v51, v3
+; VI-NEXT: s_waitcnt vmcnt(1)
+; VI-NEXT: v_mov_b32_e32 v43, v0
+; VI-NEXT: v_or_b32_sdwa v0, v24, v0 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
; VI-NEXT: s_waitcnt vmcnt(0)
; VI-NEXT: v_mov_b32_e32 v49, v1
-; VI-NEXT: v_or_b32_sdwa v1, v24, v1 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; VI-NEXT: v_or_b32_sdwa v1, v26, v1 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
; VI-NEXT: v_or_b32_sdwa v24, v0, v1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_0 src1_sel:DWORD
-; VI-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:612 ; 4-byte Folded Reload
+; VI-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:576 ; 4-byte Folded Reload
; VI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:416 ; 4-byte Folded Reload
; VI-NEXT: s_waitcnt vmcnt(1)
; VI-NEXT: v_mov_b32_e32 v54, v0
@@ -95586,28 +95609,26 @@ define inreg <16 x i64> @bitcast_v128i8_to_v16i64_scalar(<128 x i8> inreg %a, i3
; VI-NEXT: v_mov_b32_e32 v46, v61
; VI-NEXT: v_or_b32_sdwa v25, v0, v1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_0 src1_sel:DWORD
; VI-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:412 ; 4-byte Folded Reload
-; VI-NEXT: buffer_load_dword v61, off, s[0:3], s32 offset:564 ; 4-byte Folded Reload
+; VI-NEXT: buffer_load_dword v61, off, s[0:3], s32 offset:568 ; 4-byte Folded Reload
; VI-NEXT: s_waitcnt vmcnt(1)
; VI-NEXT: v_or_b32_sdwa v0, v47, v0 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
; VI-NEXT: s_waitcnt vmcnt(0)
-; VI-NEXT: v_or_b32_sdwa v1, v45, v61 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; VI-NEXT: v_or_b32_sdwa v1, v58, v61 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
; VI-NEXT: v_or_b32_sdwa v26, v0, v1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_0 src1_sel:DWORD
-; VI-NEXT: v_or_b32_sdwa v0, v58, v48 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; VI-NEXT: v_or_b32_sdwa v0, v45, v48 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
; VI-NEXT: v_or_b32_sdwa v1, v44, v27 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
-; VI-NEXT: v_mov_b32_e32 v47, v45
; VI-NEXT: v_or_b32_sdwa v27, v0, v1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_0 src1_sel:DWORD
-; VI-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:796 ; 4-byte Folded Reload
-; VI-NEXT: buffer_load_dword v45, off, s[0:3], s32 offset:568 ; 4-byte Folded Reload
-; VI-NEXT: v_mov_b32_e32 v58, v44
-; VI-NEXT: s_waitcnt vmcnt(1)
-; VI-NEXT: v_mov_b32_e32 v48, v0
-; VI-NEXT: v_or_b32_sdwa v0, v63, v0 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; VI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:556 ; 4-byte Folded Reload
+; VI-NEXT: v_or_b32_sdwa v0, v63, v28 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; VI-NEXT: v_mov_b32_e32 v48, v28
+; VI-NEXT: v_mov_b32_e32 v47, v58
+; VI-NEXT: v_mov_b32_e32 v45, v44
+; VI-NEXT: v_mov_b32_e32 v63, v42
; VI-NEXT: s_waitcnt vmcnt(0)
-; VI-NEXT: v_or_b32_sdwa v1, v42, v45 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; VI-NEXT: v_or_b32_sdwa v1, v42, v1 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
; VI-NEXT: v_or_b32_sdwa v28, v0, v1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_0 src1_sel:DWORD
; VI-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:404 ; 4-byte Folded Reload
; VI-NEXT: v_or_b32_sdwa v1, v40, v29 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
-; VI-NEXT: v_mov_b32_e32 v63, v42
; VI-NEXT: s_waitcnt vmcnt(0)
; VI-NEXT: v_or_b32_sdwa v0, v41, v0 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
; VI-NEXT: v_or_b32_sdwa v29, v0, v1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_0 src1_sel:DWORD
@@ -95623,8 +95644,8 @@ define inreg <16 x i64> @bitcast_v128i8_to_v16i64_scalar(<128 x i8> inreg %a, i3
; VI-NEXT: s_waitcnt vmcnt(0)
; VI-NEXT: v_or_b32_sdwa v1, v53, v1 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
; VI-NEXT: v_or_b32_sdwa v31, v0, v1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_0 src1_sel:DWORD
-; VI-NEXT: buffer_load_dword v52, off, s[0:3], s32 offset:556 ; 4-byte Folded Reload
-; VI-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:560 ; 4-byte Folded Reload
+; VI-NEXT: buffer_load_dword v52, off, s[0:3], s32 offset:560 ; 4-byte Folded Reload
+; VI-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:564 ; 4-byte Folded Reload
; VI-NEXT: s_waitcnt vmcnt(0)
; VI-NEXT: v_mov_b32_e32 v57, v0
; VI-NEXT: v_or_b32_sdwa v0, v52, v0 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
@@ -95658,44 +95679,43 @@ define inreg <16 x i64> @bitcast_v128i8_to_v16i64_scalar(<128 x i8> inreg %a, i3
; VI-NEXT: s_mov_b64 s[4:5], 0
; VI-NEXT: s_branch .LBB59_3
; VI-NEXT: .LBB59_2:
-; VI-NEXT: buffer_load_dword v57, off, s[0:3], s32 offset:560 ; 4-byte Folded Reload
-; VI-NEXT: buffer_load_dword v50, off, s[0:3], s32 offset:580 ; 4-byte Folded Reload
+; VI-NEXT: buffer_load_dword v57, off, s[0:3], s32 offset:564 ; 4-byte Folded Reload
+; VI-NEXT: buffer_load_dword v50, off, s[0:3], s32 offset:588 ; 4-byte Folded Reload
; VI-NEXT: buffer_load_dword v59, off, s[0:3], s32 offset:584 ; 4-byte Folded Reload
-; VI-NEXT: buffer_load_dword v56, off, s[0:3], s32 offset:608 ; 4-byte Folded Reload
-; VI-NEXT: buffer_load_dword v39, off, s[0:3], s32 offset:600 ; 4-byte Folded Reload
-; VI-NEXT: buffer_load_dword v38, off, s[0:3], s32 offset:604 ; 4-byte Folded Reload
-; VI-NEXT: buffer_load_dword v37, off, s[0:3], s32 offset:596 ; 4-byte Folded Reload
-; VI-NEXT: buffer_load_dword v36, off, s[0:3], s32 offset:572 ; 4-byte Folded Reload
-; VI-NEXT: buffer_load_dword v35, off, s[0:3], s32 offset:588 ; 4-byte Folded Reload
+; VI-NEXT: buffer_load_dword v56, off, s[0:3], s32 offset:616 ; 4-byte Folded Reload
+; VI-NEXT: buffer_load_dword v39, off, s[0:3], s32 offset:612 ; 4-byte Folded Reload
+; VI-NEXT: buffer_load_dword v38, off, s[0:3], s32 offset:608 ; 4-byte Folded Reload
+; VI-NEXT: buffer_load_dword v37, off, s[0:3], s32 offset:604 ; 4-byte Folded Reload
+; VI-NEXT: buffer_load_dword v36, off, s[0:3], s32 offset:596 ; 4-byte Folded Reload
+; VI-NEXT: buffer_load_dword v35, off, s[0:3], s32 offset:600 ; 4-byte Folded Reload
; VI-NEXT: buffer_load_dword v33, off, s[0:3], s32 offset:592 ; 4-byte Folded Reload
-; VI-NEXT: v_mov_b32_e32 v32, v54
-; VI-NEXT: v_mov_b32_e32 v43, v49
+; VI-NEXT: buffer_load_dword v43, off, s[0:3], s32 offset:580 ; 4-byte Folded Reload
+; VI-NEXT: buffer_load_dword v49, off, s[0:3], s32 offset:572 ; 4-byte Folded Reload
; VI-NEXT: v_mov_b32_e32 v46, v61
-; VI-NEXT: v_mov_b32_e32 v47, v45
-; VI-NEXT: buffer_load_dword v54, off, s[0:3], s32 offset:612 ; 4-byte Folded Reload
-; VI-NEXT: buffer_load_dword v49, off, s[0:3], s32 offset:576 ; 4-byte Folded Reload
-; VI-NEXT: buffer_load_dword v45, off, s[0:3], s32 offset:568 ; 4-byte Folded Reload
-; VI-NEXT: buffer_load_dword v52, off, s[0:3], s32 offset:556 ; 4-byte Folded Reload
-; VI-NEXT: buffer_load_dword v61, off, s[0:3], s32 offset:564 ; 4-byte Folded Reload
-; VI-NEXT: v_mov_b32_e32 v34, v26
-; VI-NEXT: v_mov_b32_e32 v58, v44
+; VI-NEXT: buffer_load_dword v54, off, s[0:3], s32 offset:576 ; 4-byte Folded Reload
+; VI-NEXT: buffer_load_dword v52, off, s[0:3], s32 offset:560 ; 4-byte Folded Reload
+; VI-NEXT: buffer_load_dword v61, off, s[0:3], s32 offset:568 ; 4-byte Folded Reload
+; VI-NEXT: v_mov_b32_e32 v34, v22
+; VI-NEXT: v_mov_b32_e32 v32, v23
+; VI-NEXT: v_mov_b32_e32 v47, v58
+; VI-NEXT: v_mov_b32_e32 v45, v44
; VI-NEXT: s_waitcnt vmcnt(14)
; VI-NEXT: v_mov_b32_e32 v63, v42
; VI-NEXT: v_mov_b32_e32 v51, v7
-; VI-NEXT: v_mov_b32_e32 v48, v29
+; VI-NEXT: v_mov_b32_e32 v48, v28
; VI-NEXT: s_mov_b64 s[4:5], -1
; VI-NEXT: ; implicit-def: $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7_vgpr8_vgpr9_vgpr10_vgpr11_vgpr12_vgpr13_vgpr14_vgpr15_vgpr16_vgpr17_vgpr18_vgpr19_vgpr20_vgpr21_vgpr22_vgpr23_vgpr24_vgpr25_vgpr26_vgpr27_vgpr28_vgpr29_vgpr30_vgpr31
; VI-NEXT: .LBB59_3: ; %Flow
; VI-NEXT: s_andn2_b64 vcc, exec, s[4:5]
-; VI-NEXT: v_mov_b32_e32 v44, v47
-; VI-NEXT: v_mov_b32_e32 v47, v46
+; VI-NEXT: v_mov_b32_e32 v42, v45
+; VI-NEXT: v_mov_b32_e32 v45, v46
; VI-NEXT: s_waitcnt vmcnt(3)
; VI-NEXT: v_mov_b32_e32 v46, v49
; VI-NEXT: s_cbranch_vccnz .LBB59_5
; VI-NEXT: ; %bb.4: ; %cmp.true
-; VI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:652 ; 4-byte Folded Reload
-; VI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:728 ; 4-byte Folded Reload
-; VI-NEXT: buffer_load_dword v3, off, s[0:3], s32 offset:720 ; 4-byte Folded Reload
+; VI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:656 ; 4-byte Folded Reload
+; VI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:732 ; 4-byte Folded Reload
+; VI-NEXT: buffer_load_dword v3, off, s[0:3], s32 offset:724 ; 4-byte Folded Reload
; VI-NEXT: s_add_i32 s28, s28, 3
; VI-NEXT: s_and_b32 s4, s28, 0xff
; VI-NEXT: s_lshl_b32 s5, s29, 8
@@ -95749,7 +95769,7 @@ define inreg <16 x i64> @bitcast_v128i8_to_v16i64_scalar(<128 x i8> inreg %a, i3
; VI-NEXT: v_add_u32_e32 v1, vcc, 3, v1
; VI-NEXT: s_waitcnt vmcnt(1)
; VI-NEXT: v_or_b32_sdwa v1, v2, v1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:BYTE_0
-; VI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:660 ; 4-byte Folded Reload
+; VI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:664 ; 4-byte Folded Reload
; VI-NEXT: v_add_u32_e32 v1, vcc, 0x300, v1
; VI-NEXT: s_waitcnt vmcnt(0)
; VI-NEXT: v_add_u32_e32 v2, vcc, 3, v2
@@ -95758,8 +95778,8 @@ define inreg <16 x i64> @bitcast_v128i8_to_v16i64_scalar(<128 x i8> inreg %a, i3
; VI-NEXT: v_add_u32_e32 v3, vcc, 0x3000000, v0
; VI-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:488 ; 4-byte Folded Reload
; VI-NEXT: v_add_u32_e32 v4, vcc, 0x3000000, v1
-; VI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:700 ; 4-byte Folded Reload
-; VI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:696 ; 4-byte Folded Reload
+; VI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:704 ; 4-byte Folded Reload
+; VI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:700 ; 4-byte Folded Reload
; VI-NEXT: s_waitcnt vmcnt(2)
; VI-NEXT: v_add_u32_e32 v0, vcc, 3, v0
; VI-NEXT: s_waitcnt vmcnt(1)
@@ -95772,8 +95792,8 @@ define inreg <16 x i64> @bitcast_v128i8_to_v16i64_scalar(<128 x i8> inreg %a, i3
; VI-NEXT: v_or_b32_sdwa v0, v1, v0 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:WORD_0
; VI-NEXT: v_add_u32_e32 v5, vcc, 0x3000000, v0
; VI-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:476 ; 4-byte Folded Reload
-; VI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:684 ; 4-byte Folded Reload
-; VI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:680 ; 4-byte Folded Reload
+; VI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:688 ; 4-byte Folded Reload
+; VI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:684 ; 4-byte Folded Reload
; VI-NEXT: s_waitcnt vmcnt(2)
; VI-NEXT: v_add_u32_e32 v0, vcc, 3, v0
; VI-NEXT: s_waitcnt vmcnt(1)
@@ -95848,29 +95868,29 @@ define inreg <16 x i64> @bitcast_v128i8_to_v16i64_scalar(<128 x i8> inreg %a, i3
; VI-NEXT: v_add_u32_e32 v0, vcc, 3, v0
; VI-NEXT: s_waitcnt vmcnt(1)
; VI-NEXT: v_or_b32_sdwa v0, v1, v0 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:BYTE_0
-; VI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:792 ; 4-byte Folded Reload
+; VI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:796 ; 4-byte Folded Reload
; VI-NEXT: v_add_u32_e32 v0, vcc, 0x300, v0
; VI-NEXT: s_waitcnt vmcnt(0)
; VI-NEXT: v_add_u32_e32 v1, vcc, 3, v1
; VI-NEXT: v_or_b32_sdwa v1, v2, v1 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:BYTE_0
; VI-NEXT: v_or_b32_sdwa v0, v1, v0 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:WORD_0
; VI-NEXT: v_add_u32_e32 v11, vcc, 0x3000000, v0
-; VI-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:788 ; 4-byte Folded Reload
+; VI-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:792 ; 4-byte Folded Reload
; VI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:508 ; 4-byte Folded Reload
; VI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:504 ; 4-byte Folded Reload
; VI-NEXT: s_waitcnt vmcnt(2)
; VI-NEXT: v_add_u32_e32 v0, vcc, 3, v0
; VI-NEXT: s_waitcnt vmcnt(1)
; VI-NEXT: v_or_b32_sdwa v0, v1, v0 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:BYTE_0
-; VI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:784 ; 4-byte Folded Reload
+; VI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:788 ; 4-byte Folded Reload
; VI-NEXT: v_add_u32_e32 v0, vcc, 0x300, v0
; VI-NEXT: s_waitcnt vmcnt(0)
; VI-NEXT: v_add_u32_e32 v1, vcc, 3, v1
; VI-NEXT: v_or_b32_sdwa v1, v2, v1 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:BYTE_0
; VI-NEXT: v_or_b32_sdwa v0, v1, v0 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:WORD_0
; VI-NEXT: v_add_u32_e32 v12, vcc, 0x3000000, v0
-; VI-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:780 ; 4-byte Folded Reload
-; VI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:776 ; 4-byte Folded Reload
+; VI-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:784 ; 4-byte Folded Reload
+; VI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:780 ; 4-byte Folded Reload
; VI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:500 ; 4-byte Folded Reload
; VI-NEXT: s_waitcnt vmcnt(2)
; VI-NEXT: v_add_u32_e32 v0, vcc, 3, v0
@@ -95882,8 +95902,8 @@ define inreg <16 x i64> @bitcast_v128i8_to_v16i64_scalar(<128 x i8> inreg %a, i3
; VI-NEXT: v_add_u32_e32 v0, vcc, 0x300, v0
; VI-NEXT: v_or_b32_sdwa v0, v1, v0 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:WORD_0
; VI-NEXT: v_add_u32_e32 v13, vcc, 0x3000000, v0
-; VI-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:772 ; 4-byte Folded Reload
-; VI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:768 ; 4-byte Folded Reload
+; VI-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:776 ; 4-byte Folded Reload
+; VI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:772 ; 4-byte Folded Reload
; VI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:496 ; 4-byte Folded Reload
; VI-NEXT: s_waitcnt vmcnt(2)
; VI-NEXT: v_add_u32_e32 v0, vcc, 3, v0
@@ -95895,8 +95915,8 @@ define inreg <16 x i64> @bitcast_v128i8_to_v16i64_scalar(<128 x i8> inreg %a, i3
; VI-NEXT: v_add_u32_e32 v0, vcc, 0x300, v0
; VI-NEXT: v_or_b32_sdwa v0, v1, v0 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:WORD_0
; VI-NEXT: v_add_u32_e32 v14, vcc, 0x3000000, v0
-; VI-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:764 ; 4-byte Folded Reload
-; VI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:760 ; 4-byte Folded Reload
+; VI-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:768 ; 4-byte Folded Reload
+; VI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:764 ; 4-byte Folded Reload
; VI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:492 ; 4-byte Folded Reload
; VI-NEXT: s_waitcnt vmcnt(2)
; VI-NEXT: v_add_u32_e32 v0, vcc, 3, v0
@@ -95908,8 +95928,8 @@ define inreg <16 x i64> @bitcast_v128i8_to_v16i64_scalar(<128 x i8> inreg %a, i3
; VI-NEXT: v_add_u32_e32 v0, vcc, 0x300, v0
; VI-NEXT: v_or_b32_sdwa v0, v1, v0 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:WORD_0
; VI-NEXT: v_add_u32_e32 v15, vcc, 0x3000000, v0
-; VI-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:756 ; 4-byte Folded Reload
-; VI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:752 ; 4-byte Folded Reload
+; VI-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:760 ; 4-byte Folded Reload
+; VI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:756 ; 4-byte Folded Reload
; VI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:480 ; 4-byte Folded Reload
; VI-NEXT: s_waitcnt vmcnt(2)
; VI-NEXT: v_add_u32_e32 v0, vcc, 3, v0
@@ -95920,8 +95940,8 @@ define inreg <16 x i64> @bitcast_v128i8_to_v16i64_scalar(<128 x i8> inreg %a, i3
; VI-NEXT: v_add_u32_e32 v0, vcc, 0x300, v0
; VI-NEXT: v_or_b32_sdwa v0, v1, v0 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:WORD_0
; VI-NEXT: v_add_u32_e32 v16, vcc, 0x3000000, v0
-; VI-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:748 ; 4-byte Folded Reload
-; VI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:744 ; 4-byte Folded Reload
+; VI-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:752 ; 4-byte Folded Reload
+; VI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:748 ; 4-byte Folded Reload
; VI-NEXT: s_waitcnt vmcnt(1)
; VI-NEXT: v_add_u32_e32 v0, vcc, 3, v0
; VI-NEXT: v_or_b32_sdwa v0, v37, v0 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:BYTE_0
@@ -95931,8 +95951,8 @@ define inreg <16 x i64> @bitcast_v128i8_to_v16i64_scalar(<128 x i8> inreg %a, i3
; VI-NEXT: v_add_u32_e32 v0, vcc, 0x300, v0
; VI-NEXT: v_or_b32_sdwa v0, v1, v0 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:WORD_0
; VI-NEXT: v_add_u32_e32 v17, vcc, 0x3000000, v0
-; VI-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:740 ; 4-byte Folded Reload
-; VI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:736 ; 4-byte Folded Reload
+; VI-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:744 ; 4-byte Folded Reload
+; VI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:740 ; 4-byte Folded Reload
; VI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:468 ; 4-byte Folded Reload
; VI-NEXT: s_waitcnt vmcnt(2)
; VI-NEXT: v_add_u32_e32 v0, vcc, 3, v0
@@ -95943,8 +95963,8 @@ define inreg <16 x i64> @bitcast_v128i8_to_v16i64_scalar(<128 x i8> inreg %a, i3
; VI-NEXT: v_add_u32_e32 v0, vcc, 0x300, v0
; VI-NEXT: v_or_b32_sdwa v0, v1, v0 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:WORD_0
; VI-NEXT: v_add_u32_e32 v18, vcc, 0x3000000, v0
-; VI-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:732 ; 4-byte Folded Reload
-; VI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:724 ; 4-byte Folded Reload
+; VI-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:736 ; 4-byte Folded Reload
+; VI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:728 ; 4-byte Folded Reload
; VI-NEXT: s_waitcnt vmcnt(1)
; VI-NEXT: v_add_u32_e32 v0, vcc, 3, v0
; VI-NEXT: v_or_b32_sdwa v0, v33, v0 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:BYTE_0
@@ -95954,63 +95974,63 @@ define inreg <16 x i64> @bitcast_v128i8_to_v16i64_scalar(<128 x i8> inreg %a, i3
; VI-NEXT: v_add_u32_e32 v0, vcc, 0x300, v0
; VI-NEXT: v_or_b32_sdwa v0, v1, v0 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:WORD_0
; VI-NEXT: v_add_u32_e32 v19, vcc, 0x3000000, v0
-; VI-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:716 ; 4-byte Folded Reload
+; VI-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:720 ; 4-byte Folded Reload
; VI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:456 ; 4-byte Folded Reload
; VI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:444 ; 4-byte Folded Reload
; VI-NEXT: s_waitcnt vmcnt(2)
; VI-NEXT: v_add_u32_e32 v0, vcc, 3, v0
; VI-NEXT: s_waitcnt vmcnt(1)
; VI-NEXT: v_or_b32_sdwa v0, v1, v0 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:BYTE_0
-; VI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:712 ; 4-byte Folded Reload
+; VI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:716 ; 4-byte Folded Reload
; VI-NEXT: v_add_u32_e32 v0, vcc, 0x300, v0
; VI-NEXT: s_waitcnt vmcnt(0)
; VI-NEXT: v_add_u32_e32 v1, vcc, 3, v1
; VI-NEXT: v_or_b32_sdwa v1, v2, v1 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:BYTE_0
; VI-NEXT: v_or_b32_sdwa v0, v1, v0 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:WORD_0
; VI-NEXT: v_add_u32_e32 v20, vcc, 0x3000000, v0
-; VI-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:708 ; 4-byte Folded Reload
+; VI-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:712 ; 4-byte Folded Reload
; VI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:436 ; 4-byte Folded Reload
-; VI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:648 ; 4-byte Folded Reload
+; VI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:652 ; 4-byte Folded Reload
; VI-NEXT: s_waitcnt vmcnt(2)
; VI-NEXT: v_add_u32_e32 v0, vcc, 3, v0
; VI-NEXT: s_waitcnt vmcnt(1)
; VI-NEXT: v_or_b32_sdwa v0, v1, v0 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:BYTE_0
-; VI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:704 ; 4-byte Folded Reload
+; VI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:708 ; 4-byte Folded Reload
; VI-NEXT: v_add_u32_e32 v0, vcc, 0x300, v0
; VI-NEXT: s_waitcnt vmcnt(0)
; VI-NEXT: v_add_u32_e32 v1, vcc, 3, v1
; VI-NEXT: v_or_b32_sdwa v1, v2, v1 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:BYTE_0
; VI-NEXT: v_or_b32_sdwa v0, v1, v0 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:WORD_0
; VI-NEXT: v_add_u32_e32 v21, vcc, 0x3000000, v0
-; VI-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:692 ; 4-byte Folded Reload
-; VI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:640 ; 4-byte Folded Reload
+; VI-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:696 ; 4-byte Folded Reload
+; VI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:644 ; 4-byte Folded Reload
; VI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:416 ; 4-byte Folded Reload
; VI-NEXT: s_waitcnt vmcnt(2)
; VI-NEXT: v_add_u32_e32 v0, vcc, 3, v0
; VI-NEXT: s_waitcnt vmcnt(1)
; VI-NEXT: v_or_b32_sdwa v0, v1, v0 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:BYTE_0
-; VI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:688 ; 4-byte Folded Reload
+; VI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:692 ; 4-byte Folded Reload
; VI-NEXT: v_add_u32_e32 v0, vcc, 0x300, v0
; VI-NEXT: s_waitcnt vmcnt(0)
; VI-NEXT: v_add_u32_e32 v1, vcc, 3, v1
; VI-NEXT: v_or_b32_sdwa v1, v34, v1 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:BYTE_0
; VI-NEXT: v_or_b32_sdwa v0, v1, v0 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:WORD_0
; VI-NEXT: v_add_u32_e32 v22, vcc, 0x3000000, v0
-; VI-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:676 ; 4-byte Folded Reload
-; VI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:656 ; 4-byte Folded Reload
+; VI-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:680 ; 4-byte Folded Reload
+; VI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:660 ; 4-byte Folded Reload
; VI-NEXT: s_waitcnt vmcnt(1)
; VI-NEXT: v_add_u32_e32 v0, vcc, 3, v0
; VI-NEXT: s_waitcnt vmcnt(0)
; VI-NEXT: v_or_b32_sdwa v0, v1, v0 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:BYTE_0
-; VI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:672 ; 4-byte Folded Reload
+; VI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:676 ; 4-byte Folded Reload
; VI-NEXT: v_add_u32_e32 v0, vcc, 0x300, v0
; VI-NEXT: s_waitcnt vmcnt(0)
; VI-NEXT: v_add_u32_e32 v1, vcc, 3, v1
; VI-NEXT: v_or_b32_sdwa v1, v32, v1 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:BYTE_0
; VI-NEXT: v_or_b32_sdwa v0, v1, v0 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:WORD_0
; VI-NEXT: v_add_u32_e32 v23, vcc, 0x3000000, v0
-; VI-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:668 ; 4-byte Folded Reload
-; VI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:664 ; 4-byte Folded Reload
+; VI-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:672 ; 4-byte Folded Reload
+; VI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:668 ; 4-byte Folded Reload
; VI-NEXT: s_waitcnt vmcnt(1)
; VI-NEXT: v_add_u32_e32 v0, vcc, 3, v0
; VI-NEXT: v_or_b32_sdwa v0, v43, v0 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:BYTE_0
@@ -96020,54 +96040,57 @@ define inreg <16 x i64> @bitcast_v128i8_to_v16i64_scalar(<128 x i8> inreg %a, i3
; VI-NEXT: v_add_u32_e32 v0, vcc, 0x300, v0
; VI-NEXT: v_or_b32_sdwa v0, v1, v0 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:WORD_0
; VI-NEXT: v_add_u32_e32 v24, vcc, 0x3000000, v0
-; VI-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:644 ; 4-byte Folded Reload
-; VI-NEXT: v_add_u32_e32 v1, vcc, 3, v47
+; VI-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:648 ; 4-byte Folded Reload
+; VI-NEXT: v_add_u32_e32 v1, vcc, 3, v45
; VI-NEXT: v_or_b32_sdwa v1, v2, v1 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:BYTE_0
-; VI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:620 ; 4-byte Folded Reload
+; VI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:624 ; 4-byte Folded Reload
; VI-NEXT: s_waitcnt vmcnt(1)
; VI-NEXT: v_add_u32_e32 v0, vcc, 3, v0
; VI-NEXT: v_or_b32_sdwa v0, v54, v0 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:BYTE_0
; VI-NEXT: v_add_u32_e32 v0, vcc, 0x300, v0
; VI-NEXT: v_or_b32_sdwa v0, v1, v0 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:WORD_0
; VI-NEXT: v_add_u32_e32 v25, vcc, 0x3000000, v0
-; VI-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:636 ; 4-byte Folded Reload
+; VI-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:640 ; 4-byte Folded Reload
; VI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:412 ; 4-byte Folded Reload
; VI-NEXT: s_waitcnt vmcnt(1)
; VI-NEXT: v_add_u32_e32 v0, vcc, 3, v0
; VI-NEXT: s_waitcnt vmcnt(0)
; VI-NEXT: v_or_b32_sdwa v0, v1, v0 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:BYTE_0
-; VI-NEXT: v_add_u32_e32 v1, vcc, 3, v44
+; VI-NEXT: v_add_u32_e32 v1, vcc, 3, v47
; VI-NEXT: v_add_u32_e32 v0, vcc, 0x300, v0
; VI-NEXT: v_or_b32_sdwa v1, v61, v1 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:BYTE_0
; VI-NEXT: v_or_b32_sdwa v0, v1, v0 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:WORD_0
; VI-NEXT: v_add_u32_e32 v26, vcc, 0x3000000, v0
-; VI-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:632 ; 4-byte Folded Reload
-; VI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:624 ; 4-byte Folded Reload
+; VI-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:636 ; 4-byte Folded Reload
+; VI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:628 ; 4-byte Folded Reload
; VI-NEXT: s_waitcnt vmcnt(1)
; VI-NEXT: v_add_u32_e32 v0, vcc, 3, v0
; VI-NEXT: s_waitcnt vmcnt(0)
; VI-NEXT: v_or_b32_sdwa v0, v1, v0 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:BYTE_0
-; VI-NEXT: v_add_u32_e32 v1, vcc, 3, v58
+; VI-NEXT: v_add_u32_e32 v1, vcc, 3, v42
; VI-NEXT: v_add_u32_e32 v0, vcc, 0x300, v0
; VI-NEXT: v_or_b32_sdwa v1, v2, v1 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:BYTE_0
; VI-NEXT: v_or_b32_sdwa v0, v1, v0 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:WORD_0
; VI-NEXT: v_add_u32_e32 v27, vcc, 0x3000000, v0
-; VI-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:628 ; 4-byte Folded Reload
+; VI-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:632 ; 4-byte Folded Reload
+; VI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:556 ; 4-byte Folded Reload
; VI-NEXT: v_add_u32_e32 v1, vcc, 3, v63
-; VI-NEXT: v_or_b32_sdwa v1, v45, v1 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:BYTE_0
-; VI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:616 ; 4-byte Folded Reload
; VI-NEXT: s_waitcnt vmcnt(1)
; VI-NEXT: v_add_u32_e32 v0, vcc, 3, v0
; VI-NEXT: v_or_b32_sdwa v0, v48, v0 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:BYTE_0
; VI-NEXT: v_add_u32_e32 v0, vcc, 0x300, v0
+; VI-NEXT: s_waitcnt vmcnt(0)
+; VI-NEXT: v_or_b32_sdwa v1, v2, v1 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:BYTE_0
; VI-NEXT: v_or_b32_sdwa v0, v1, v0 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:WORD_0
; VI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:404 ; 4-byte Folded Reload
+; VI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:620 ; 4-byte Folded Reload
; VI-NEXT: v_add_u32_e32 v28, vcc, 0x3000000, v0
; VI-NEXT: v_add_u32_e32 v0, vcc, 3, v41
-; VI-NEXT: s_waitcnt vmcnt(0)
+; VI-NEXT: s_waitcnt vmcnt(1)
; VI-NEXT: v_or_b32_sdwa v0, v1, v0 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:BYTE_0
; VI-NEXT: v_add_u32_e32 v1, vcc, 3, v40
; VI-NEXT: v_add_u32_e32 v0, vcc, 0x300, v0
+; VI-NEXT: s_waitcnt vmcnt(0)
; VI-NEXT: v_or_b32_sdwa v1, v2, v1 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:BYTE_0
; VI-NEXT: v_or_b32_sdwa v0, v1, v0 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:WORD_0
; VI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:400 ; 4-byte Folded Reload
@@ -96352,7 +96375,7 @@ define inreg <16 x i64> @bitcast_v128i8_to_v16i64_scalar(<128 x i8> inreg %a, i3
; GFX9-NEXT: buffer_load_ushort v31, off, s[0:3], s32 offset:164
; GFX9-NEXT: buffer_load_ushort v23, off, s[0:3], s32 offset:172
; GFX9-NEXT: buffer_load_ushort v50, off, s[0:3], s32 offset:180
-; GFX9-NEXT: buffer_load_ushort v44, off, s[0:3], s32 offset:188
+; GFX9-NEXT: buffer_load_ushort v40, off, s[0:3], s32 offset:188
; GFX9-NEXT: buffer_load_ushort v37, off, s[0:3], s32 offset:196
; GFX9-NEXT: buffer_load_ushort v52, off, s[0:3], s32 offset:204
; GFX9-NEXT: buffer_load_ushort v29, off, s[0:3], s32 offset:212
@@ -96360,11 +96383,11 @@ define inreg <16 x i64> @bitcast_v128i8_to_v16i64_scalar(<128 x i8> inreg %a, i3
; GFX9-NEXT: buffer_load_ushort v48, off, s[0:3], s32 offset:228
; GFX9-NEXT: buffer_load_ushort v39, off, s[0:3], s32 offset:236
; GFX9-NEXT: buffer_load_ushort v45, off, s[0:3], s32 offset:244
-; GFX9-NEXT: buffer_load_ushort v40, off, s[0:3], s32 offset:252
+; GFX9-NEXT: buffer_load_ushort v44, off, s[0:3], s32 offset:252
; GFX9-NEXT: buffer_load_ushort v55, off, s[0:3], s32 offset:260
; GFX9-NEXT: buffer_load_ushort v43, off, s[0:3], s32 offset:268
-; GFX9-NEXT: buffer_load_ushort v36, off, s[0:3], s32 offset:276
-; GFX9-NEXT: buffer_load_ushort v42, off, s[0:3], s32 offset:284
+; GFX9-NEXT: buffer_load_ushort v42, off, s[0:3], s32 offset:276
+; GFX9-NEXT: buffer_load_ushort v36, off, s[0:3], s32 offset:284
; GFX9-NEXT: buffer_load_ushort v41, off, s[0:3], s32 offset:292
; GFX9-NEXT: buffer_load_ushort v32, off, s[0:3], s32 offset:300
; GFX9-NEXT: buffer_load_ushort v35, off, s[0:3], s32 offset:308
@@ -96390,7 +96413,7 @@ define inreg <16 x i64> @bitcast_v128i8_to_v16i64_scalar(<128 x i8> inreg %a, i3
; GFX9-NEXT: buffer_store_dword v29, off, s[0:3], s32 offset:668 ; 4-byte Folded Spill
; GFX9-NEXT: buffer_store_dword v52, off, s[0:3], s32 offset:672 ; 4-byte Folded Spill
; GFX9-NEXT: buffer_store_dword v37, off, s[0:3], s32 offset:676 ; 4-byte Folded Spill
-; GFX9-NEXT: buffer_store_dword v44, off, s[0:3], s32 offset:680 ; 4-byte Folded Spill
+; GFX9-NEXT: buffer_store_dword v40, off, s[0:3], s32 offset:680 ; 4-byte Folded Spill
; GFX9-NEXT: buffer_store_dword v50, off, s[0:3], s32 offset:684 ; 4-byte Folded Spill
; GFX9-NEXT: buffer_store_dword v23, off, s[0:3], s32 offset:688 ; 4-byte Folded Spill
; GFX9-NEXT: buffer_store_dword v31, off, s[0:3], s32 offset:692 ; 4-byte Folded Spill
@@ -96565,7 +96588,7 @@ define inreg <16 x i64> @bitcast_v128i8_to_v16i64_scalar(<128 x i8> inreg %a, i3
; GFX9-NEXT: s_waitcnt vmcnt(1)
; GFX9-NEXT: v_or_b32_sdwa v0, v23, v0 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
; GFX9-NEXT: v_or_b32_sdwa v22, v0, v1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_0 src1_sel:DWORD
-; GFX9-NEXT: v_or_b32_sdwa v0, v44, v58 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; GFX9-NEXT: v_or_b32_sdwa v0, v40, v58 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
; GFX9-NEXT: v_or_b32_sdwa v1, v37, v57 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
; GFX9-NEXT: v_or_b32_sdwa v23, v0, v1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_0 src1_sel:DWORD
; GFX9-NEXT: s_waitcnt vmcnt(0)
@@ -96577,7 +96600,7 @@ define inreg <16 x i64> @bitcast_v128i8_to_v16i64_scalar(<128 x i8> inreg %a, i3
; GFX9-NEXT: v_mov_b32_e32 v37, v57
; GFX9-NEXT: v_mov_b32_e32 v57, v60
; GFX9-NEXT: v_mov_b32_e32 v52, v56
-; GFX9-NEXT: buffer_load_dword v44, off, s[0:3], s32 offset:612 ; 4-byte Folded Reload
+; GFX9-NEXT: buffer_load_dword v40, off, s[0:3], s32 offset:612 ; 4-byte Folded Reload
; GFX9-NEXT: s_waitcnt vmcnt(2)
; GFX9-NEXT: v_mov_b32_e32 v34, v0
; GFX9-NEXT: v_or_b32_sdwa v0, v46, v0 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
@@ -96586,14 +96609,14 @@ define inreg <16 x i64> @bitcast_v128i8_to_v16i64_scalar(<128 x i8> inreg %a, i3
; GFX9-NEXT: v_or_b32_sdwa v25, v0, v1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_0 src1_sel:DWORD
; GFX9-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:440 ; 4-byte Folded Reload
; GFX9-NEXT: s_waitcnt vmcnt(1)
-; GFX9-NEXT: v_or_b32_sdwa v1, v45, v44 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; GFX9-NEXT: v_or_b32_sdwa v1, v45, v40 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
; GFX9-NEXT: s_waitcnt vmcnt(0)
; GFX9-NEXT: v_or_b32_sdwa v0, v39, v0 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
; GFX9-NEXT: v_or_b32_sdwa v26, v0, v1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_0 src1_sel:DWORD
; GFX9-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:436 ; 4-byte Folded Reload
; GFX9-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:432 ; 4-byte Folded Reload
; GFX9-NEXT: s_waitcnt vmcnt(1)
-; GFX9-NEXT: v_or_b32_sdwa v0, v40, v0 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; GFX9-NEXT: v_or_b32_sdwa v0, v44, v0 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
; GFX9-NEXT: s_waitcnt vmcnt(0)
; GFX9-NEXT: v_or_b32_sdwa v1, v55, v1 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
; GFX9-NEXT: v_or_b32_sdwa v27, v0, v1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_0 src1_sel:DWORD
@@ -96603,12 +96626,12 @@ define inreg <16 x i64> @bitcast_v128i8_to_v16i64_scalar(<128 x i8> inreg %a, i3
; GFX9-NEXT: v_mov_b32_e32 v51, v0
; GFX9-NEXT: v_or_b32_sdwa v0, v43, v0 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
; GFX9-NEXT: s_waitcnt vmcnt(0)
-; GFX9-NEXT: v_or_b32_sdwa v1, v36, v1 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; GFX9-NEXT: v_or_b32_sdwa v1, v42, v1 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
; GFX9-NEXT: v_or_b32_sdwa v28, v0, v1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_0 src1_sel:DWORD
; GFX9-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:420 ; 4-byte Folded Reload
; GFX9-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:416 ; 4-byte Folded Reload
; GFX9-NEXT: s_waitcnt vmcnt(1)
-; GFX9-NEXT: v_or_b32_sdwa v0, v42, v0 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; GFX9-NEXT: v_or_b32_sdwa v0, v36, v0 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
; GFX9-NEXT: s_waitcnt vmcnt(0)
; GFX9-NEXT: v_or_b32_sdwa v1, v41, v1 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
; GFX9-NEXT: v_or_b32_sdwa v29, v0, v1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_0 src1_sel:DWORD
@@ -96662,7 +96685,7 @@ define inreg <16 x i64> @bitcast_v128i8_to_v16i64_scalar(<128 x i8> inreg %a, i3
; GFX9-NEXT: .LBB59_2:
; GFX9-NEXT: buffer_load_dword v34, off, s[0:3], s32 offset:616 ; 4-byte Folded Reload
; GFX9-NEXT: buffer_load_dword v51, off, s[0:3], s32 offset:620 ; 4-byte Folded Reload
-; GFX9-NEXT: buffer_load_dword v44, off, s[0:3], s32 offset:612 ; 4-byte Folded Reload
+; GFX9-NEXT: buffer_load_dword v40, off, s[0:3], s32 offset:612 ; 4-byte Folded Reload
; GFX9-NEXT: buffer_load_dword v50, off, s[0:3], s32 offset:608 ; 4-byte Folded Reload
; GFX9-NEXT: buffer_load_dword v37, off, s[0:3], s32 offset:604 ; 4-byte Folded Reload
; GFX9-NEXT: buffer_load_dword v47, off, s[0:3], s32 offset:596 ; 4-byte Folded Reload
@@ -97024,12 +97047,12 @@ define inreg <16 x i64> @bitcast_v128i8_to_v16i64_scalar(<128 x i8> inreg %a, i3
; GFX9-NEXT: s_waitcnt vmcnt(1)
; GFX9-NEXT: v_or_b32_sdwa v0, v1, v0 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:BYTE_0
; GFX9-NEXT: v_add_u32_e32 v1, 3, v45
-; GFX9-NEXT: v_or_b32_sdwa v1, v44, v1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:BYTE_0
+; GFX9-NEXT: v_or_b32_sdwa v1, v40, v1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:BYTE_0
; GFX9-NEXT: v_add_u32_e32 v0, 0x300, v0
; GFX9-NEXT: v_add_u32_sdwa v1, v1, s4 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:DWORD
; GFX9-NEXT: v_or_b32_sdwa v26, v0, v1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_0 src1_sel:DWORD
; GFX9-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:436 ; 4-byte Folded Reload
-; GFX9-NEXT: v_add_u32_e32 v0, 3, v40
+; GFX9-NEXT: v_add_u32_e32 v0, 3, v44
; GFX9-NEXT: s_waitcnt vmcnt(0)
; GFX9-NEXT: v_or_b32_sdwa v0, v1, v0 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:BYTE_0
; GFX9-NEXT: v_add_u32_e32 v1, 3, v55
@@ -97039,7 +97062,7 @@ define inreg <16 x i64> @bitcast_v128i8_to_v16i64_scalar(<128 x i8> inreg %a, i3
; GFX9-NEXT: v_add_u32_sdwa v1, v1, s4 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:DWORD
; GFX9-NEXT: v_or_b32_sdwa v27, v0, v1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_0 src1_sel:DWORD
; GFX9-NEXT: v_add_u32_e32 v0, 3, v43
-; GFX9-NEXT: v_add_u32_e32 v1, 3, v36
+; GFX9-NEXT: v_add_u32_e32 v1, 3, v42
; GFX9-NEXT: v_or_b32_sdwa v0, v51, v0 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:BYTE_0
; GFX9-NEXT: v_add_u32_e32 v0, 0x300, v0
; GFX9-NEXT: s_waitcnt vmcnt(0)
@@ -97048,7 +97071,7 @@ define inreg <16 x i64> @bitcast_v128i8_to_v16i64_scalar(<128 x i8> inreg %a, i3
; GFX9-NEXT: v_or_b32_sdwa v28, v0, v1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_0 src1_sel:DWORD
; GFX9-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:420 ; 4-byte Folded Reload
; GFX9-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:416 ; 4-byte Folded Reload
-; GFX9-NEXT: v_add_u32_e32 v0, 3, v42
+; GFX9-NEXT: v_add_u32_e32 v0, 3, v36
; GFX9-NEXT: s_waitcnt vmcnt(1)
; GFX9-NEXT: v_or_b32_sdwa v0, v1, v0 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:BYTE_0
; GFX9-NEXT: v_add_u32_e32 v1, 3, v41
@@ -103810,10 +103833,10 @@ define inreg <16 x i64> @bitcast_v64bf16_to_v16i64_scalar(<64 x bfloat> inreg %a
; SI-NEXT: v_mul_f32_e32 v34, 1.0, v36
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_mul_f32_e64 v0, 1.0, s17
-; SI-NEXT: v_mul_f32_e64 v35, 1.0, s18
+; SI-NEXT: v_mul_f32_e64 v33, 1.0, s18
; SI-NEXT: v_mul_f32_e64 v36, 1.0, s21
; SI-NEXT: v_mul_f32_e64 v42, 1.0, s20
-; SI-NEXT: v_mul_f32_e64 v33, 1.0, s22
+; SI-NEXT: v_mul_f32_e64 v35, 1.0, s22
; SI-NEXT: buffer_store_dword v30, off, s[0:3], s32 offset:196 ; 4-byte Folded Spill
; SI-NEXT: buffer_store_dword v31, off, s[0:3], s32 offset:200 ; 4-byte Folded Spill
; SI-NEXT: buffer_store_dword v29, off, s[0:3], s32 offset:204 ; 4-byte Folded Spill
@@ -103842,9 +103865,9 @@ define inreg <16 x i64> @bitcast_v64bf16_to_v16i64_scalar(<64 x bfloat> inreg %a
; SI-NEXT: buffer_store_dword v9, off, s[0:3], s32 offset:296 ; 4-byte Folded Spill
; SI-NEXT: buffer_store_dword v3, off, s[0:3], s32 offset:300 ; 4-byte Folded Spill
; SI-NEXT: buffer_store_dword v0, off, s[0:3], s32 offset:304 ; 4-byte Folded Spill
-; SI-NEXT: buffer_store_dword v33, off, s[0:3], s32 offset:308 ; 4-byte Folded Spill
-; SI-NEXT: buffer_store_dword v35, off, s[0:3], s32 offset:312 ; 4-byte Folded Spill
-; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:316 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:308 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v33, off, s[0:3], s32 offset:312 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v35, off, s[0:3], s32 offset:316 ; 4-byte Folded Spill
; SI-NEXT: buffer_store_dword v36, off, s[0:3], s32 offset:320 ; 4-byte Folded Spill
; SI-NEXT: buffer_store_dword v42, off, s[0:3], s32 offset:324 ; 4-byte Folded Spill
; SI-NEXT: s_cbranch_scc0 .LBB63_4
@@ -103855,12 +103878,13 @@ define inreg <16 x i64> @bitcast_v64bf16_to_v16i64_scalar(<64 x bfloat> inreg %a
; SI-NEXT: buffer_load_dword v7, off, s[0:3], s32 offset:192 ; 4-byte Folded Reload
; SI-NEXT: v_alignbit_b32 v5, v5, v8, 16
; SI-NEXT: buffer_load_dword v8, off, s[0:3], s32 offset:184 ; 4-byte Folded Reload
-; SI-NEXT: s_waitcnt expcnt(2)
-; SI-NEXT: v_lshrrev_b32_e32 v1, 16, v1
-; SI-NEXT: v_alignbit_b32 v1, v1, v35, 16
+; SI-NEXT: s_waitcnt expcnt(6)
+; SI-NEXT: v_lshrrev_b32_e32 v3, 16, v3
+; SI-NEXT: v_alignbit_b32 v3, v3, v35, 16
; SI-NEXT: v_lshrrev_b32_e32 v4, 16, v4
; SI-NEXT: v_alignbit_b32 v4, v4, v9, 16
; SI-NEXT: buffer_load_dword v9, off, s[0:3], s32 offset:180 ; 4-byte Folded Reload
+; SI-NEXT: s_waitcnt expcnt(5)
; SI-NEXT: v_lshrrev_b32_e32 v0, 16, v0
; SI-NEXT: v_mov_b32_e32 v59, v2
; SI-NEXT: v_alignbit_b32 v0, v0, v2, 16
@@ -103870,10 +103894,11 @@ define inreg <16 x i64> @bitcast_v64bf16_to_v16i64_scalar(<64 x bfloat> inreg %a
; SI-NEXT: v_mov_b32_e32 v47, v10
; SI-NEXT: v_lshrrev_b32_e32 v10, 16, v10
; SI-NEXT: v_mov_b32_e32 v45, v12
-; SI-NEXT: v_lshrrev_b32_e32 v3, 16, v3
-; SI-NEXT: v_alignbit_b32 v3, v3, v33, 16
-; SI-NEXT: v_mov_b32_e32 v33, v14
+; SI-NEXT: v_mov_b32_e32 v44, v14
+; SI-NEXT: s_waitcnt expcnt(4)
+; SI-NEXT: v_lshrrev_b32_e32 v1, 16, v1
; SI-NEXT: v_lshrrev_b32_e32 v13, 16, v13
+; SI-NEXT: v_alignbit_b32 v1, v1, v33, 16
; SI-NEXT: v_mov_b32_e32 v62, v38
; SI-NEXT: v_lshrrev_b32_e32 v15, 16, v15
; SI-NEXT: v_lshrrev_b32_e32 v16, 16, v16
@@ -103913,20 +103938,20 @@ define inreg <16 x i64> @bitcast_v64bf16_to_v16i64_scalar(<64 x bfloat> inreg %a
; SI-NEXT: v_alignbit_b32 v28, v28, v37, 16
; SI-NEXT: v_mov_b32_e32 v37, v34
; SI-NEXT: buffer_load_dword v32, off, s[0:3], s32 offset:148 ; 4-byte Folded Reload
-; SI-NEXT: s_waitcnt vmcnt(3)
+; SI-NEXT: s_waitcnt vmcnt(3) expcnt(2)
; SI-NEXT: v_mov_b32_e32 v35, v7
; SI-NEXT: v_lshrrev_b32_e32 v7, 16, v7
; SI-NEXT: s_waitcnt vmcnt(2)
; SI-NEXT: v_mov_b32_e32 v43, v8
; SI-NEXT: v_alignbit_b32 v7, v7, v8, 16
; SI-NEXT: buffer_load_dword v8, off, s[0:3], s32 offset:188 ; 4-byte Folded Reload
-; SI-NEXT: s_waitcnt vmcnt(2) expcnt(0)
-; SI-NEXT: v_mov_b32_e32 v42, v9
+; SI-NEXT: s_waitcnt vmcnt(2)
+; SI-NEXT: v_mov_b32_e32 v60, v9
; SI-NEXT: s_waitcnt vmcnt(1)
; SI-NEXT: v_lshrrev_b32_e32 v31, 16, v32
; SI-NEXT: v_alignbit_b32 v31, v31, v34, 16
-; SI-NEXT: s_waitcnt vmcnt(0)
-; SI-NEXT: v_mov_b32_e32 v60, v8
+; SI-NEXT: s_waitcnt vmcnt(0) expcnt(0)
+; SI-NEXT: v_mov_b32_e32 v42, v8
; SI-NEXT: v_lshrrev_b32_e32 v8, 16, v8
; SI-NEXT: v_alignbit_b32 v8, v8, v9, 16
; SI-NEXT: v_lshrrev_b32_e32 v9, 16, v11
@@ -103950,7 +103975,7 @@ define inreg <16 x i64> @bitcast_v64bf16_to_v16i64_scalar(<64 x bfloat> inreg %a
; SI-NEXT: v_alignbit_b32 v12, v12, v14, 16
; SI-NEXT: buffer_load_dword v14, off, s[0:3], s32 offset:160 ; 4-byte Folded Reload
; SI-NEXT: s_waitcnt vmcnt(0)
-; SI-NEXT: v_mov_b32_e32 v44, v14
+; SI-NEXT: v_mov_b32_e32 v33, v14
; SI-NEXT: v_alignbit_b32 v13, v13, v14, 16
; SI-NEXT: buffer_load_dword v14, off, s[0:3], s32 offset:152 ; 4-byte Folded Reload
; SI-NEXT: s_waitcnt vmcnt(0)
@@ -103973,7 +103998,7 @@ define inreg <16 x i64> @bitcast_v64bf16_to_v16i64_scalar(<64 x bfloat> inreg %a
; SI-NEXT: .LBB63_2: ; %cmp.true
; SI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:304 ; 4-byte Folded Reload
; SI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:312 ; 4-byte Folded Reload
-; SI-NEXT: buffer_load_dword v3, off, s[0:3], s32 offset:316 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v3, off, s[0:3], s32 offset:308 ; 4-byte Folded Reload
; SI-NEXT: buffer_load_dword v4, off, s[0:3], s32 offset:300 ; 4-byte Folded Reload
; SI-NEXT: v_and_b32_e32 v0, 0xffff0000, v59
; SI-NEXT: v_add_f32_e32 v0, 0x40c00000, v0
@@ -103989,7 +104014,7 @@ define inreg <16 x i64> @bitcast_v64bf16_to_v16i64_scalar(<64 x bfloat> inreg %a
; SI-NEXT: v_and_b32_e32 v8, 0xffff0000, v35
; SI-NEXT: v_add_f32_e32 v8, 0x40c00000, v8
; SI-NEXT: v_lshrrev_b32_e32 v8, 16, v8
-; SI-NEXT: v_and_b32_e32 v9, 0xffff0000, v60
+; SI-NEXT: v_and_b32_e32 v9, 0xffff0000, v42
; SI-NEXT: v_add_f32_e32 v9, 0x40c00000, v9
; SI-NEXT: v_lshrrev_b32_e32 v9, 16, v9
; SI-NEXT: v_and_b32_e32 v10, 0xffff0000, v57
@@ -104001,7 +104026,7 @@ define inreg <16 x i64> @bitcast_v64bf16_to_v16i64_scalar(<64 x bfloat> inreg %a
; SI-NEXT: v_and_b32_e32 v12, 0xffff0000, v45
; SI-NEXT: v_add_f32_e32 v12, 0x40c00000, v12
; SI-NEXT: v_lshrrev_b32_e32 v12, 16, v12
-; SI-NEXT: v_and_b32_e32 v13, 0xffff0000, v33
+; SI-NEXT: v_and_b32_e32 v13, 0xffff0000, v44
; SI-NEXT: v_add_f32_e32 v13, 0x40c00000, v13
; SI-NEXT: v_lshrrev_b32_e32 v13, 16, v13
; SI-NEXT: v_and_b32_e32 v15, 0xffff0000, v36
@@ -104113,7 +104138,7 @@ define inreg <16 x i64> @bitcast_v64bf16_to_v16i64_scalar(<64 x bfloat> inreg %a
; SI-NEXT: v_add_f32_e32 v2, 0x40c00000, v2
; SI-NEXT: v_lshrrev_b32_e32 v3, 16, v3
; SI-NEXT: v_alignbit_b32 v2, v3, v2, 16
-; SI-NEXT: buffer_load_dword v3, off, s[0:3], s32 offset:308 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v3, off, s[0:3], s32 offset:316 ; 4-byte Folded Reload
; SI-NEXT: s_waitcnt vmcnt(0)
; SI-NEXT: v_and_b32_e32 v3, 0xffff0000, v3
; SI-NEXT: v_add_f32_e32 v3, 0x40c00000, v3
@@ -104136,7 +104161,7 @@ define inreg <16 x i64> @bitcast_v64bf16_to_v16i64_scalar(<64 x bfloat> inreg %a
; SI-NEXT: v_and_b32_e32 v7, 0xffff0000, v43
; SI-NEXT: v_add_f32_e32 v7, 0x40c00000, v7
; SI-NEXT: v_alignbit_b32 v7, v8, v7, 16
-; SI-NEXT: v_and_b32_e32 v8, 0xffff0000, v42
+; SI-NEXT: v_and_b32_e32 v8, 0xffff0000, v60
; SI-NEXT: v_add_f32_e32 v8, 0x40c00000, v8
; SI-NEXT: v_alignbit_b32 v8, v9, v8, 16
; SI-NEXT: v_and_b32_e32 v9, 0xffff0000, v58
@@ -104151,7 +104176,7 @@ define inreg <16 x i64> @bitcast_v64bf16_to_v16i64_scalar(<64 x bfloat> inreg %a
; SI-NEXT: v_and_b32_e32 v12, 0xffff0000, v63
; SI-NEXT: v_add_f32_e32 v12, 0x40c00000, v12
; SI-NEXT: v_alignbit_b32 v12, v13, v12, 16
-; SI-NEXT: v_and_b32_e32 v13, 0xffff0000, v44
+; SI-NEXT: v_and_b32_e32 v13, 0xffff0000, v33
; SI-NEXT: v_add_f32_e32 v13, 0x40c00000, v13
; SI-NEXT: v_alignbit_b32 v13, v14, v13, 16
; SI-NEXT: v_and_b32_e32 v14, 0xffff0000, v62
@@ -104237,16 +104262,16 @@ define inreg <16 x i64> @bitcast_v64bf16_to_v16i64_scalar(<64 x bfloat> inreg %a
; SI-NEXT: s_setpc_b64 s[30:31]
; SI-NEXT: .LBB63_4:
; SI-NEXT: buffer_load_dword v43, off, s[0:3], s32 offset:184 ; 4-byte Folded Reload
-; SI-NEXT: s_waitcnt expcnt(3)
+; SI-NEXT: s_waitcnt expcnt(2)
; SI-NEXT: buffer_load_dword v35, off, s[0:3], s32 offset:192 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v60, off, s[0:3], s32 offset:180 ; 4-byte Folded Reload
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: buffer_load_dword v42, off, s[0:3], s32 offset:180 ; 4-byte Folded Reload
-; SI-NEXT: buffer_load_dword v60, off, s[0:3], s32 offset:188 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v42, off, s[0:3], s32 offset:188 ; 4-byte Folded Reload
; SI-NEXT: buffer_load_dword v58, off, s[0:3], s32 offset:176 ; 4-byte Folded Reload
; SI-NEXT: buffer_load_dword v56, off, s[0:3], s32 offset:172 ; 4-byte Folded Reload
; SI-NEXT: buffer_load_dword v46, off, s[0:3], s32 offset:168 ; 4-byte Folded Reload
; SI-NEXT: buffer_load_dword v63, off, s[0:3], s32 offset:164 ; 4-byte Folded Reload
-; SI-NEXT: buffer_load_dword v44, off, s[0:3], s32 offset:160 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v33, off, s[0:3], s32 offset:160 ; 4-byte Folded Reload
; SI-NEXT: v_mov_b32_e32 v61, v53
; SI-NEXT: buffer_load_dword v36, off, s[0:3], s32 offset:152 ; 4-byte Folded Reload
; SI-NEXT: buffer_load_dword v53, off, s[0:3], s32 offset:156 ; 4-byte Folded Reload
@@ -104255,7 +104280,7 @@ define inreg <16 x i64> @bitcast_v64bf16_to_v16i64_scalar(<64 x bfloat> inreg %a
; SI-NEXT: v_mov_b32_e32 v57, v11
; SI-NEXT: v_mov_b32_e32 v47, v10
; SI-NEXT: v_mov_b32_e32 v45, v12
-; SI-NEXT: v_mov_b32_e32 v33, v14
+; SI-NEXT: v_mov_b32_e32 v44, v14
; SI-NEXT: v_mov_b32_e32 v62, v38
; SI-NEXT: v_mov_b32_e32 v38, v39
; SI-NEXT: v_mov_b32_e32 v39, v41
@@ -107321,36 +107346,98 @@ define <64 x half> @bitcast_v16i64_to_v64f16(<16 x i64> %a, i32 %b) {
; SI-NEXT: buffer_load_dword v31, off, s[0:3], s32 offset:8
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: buffer_load_dword v63, off, s[0:3], s32
+; SI-NEXT: ; implicit-def: $vgpr32
+; SI-NEXT: ; kill: killed $vgpr32
+; SI-NEXT: ; implicit-def: $vgpr32
+; SI-NEXT: ; kill: killed $vgpr32
+; SI-NEXT: ; implicit-def: $vgpr32
+; SI-NEXT: ; kill: killed $vgpr32
+; SI-NEXT: ; implicit-def: $vgpr32
+; SI-NEXT: ; kill: killed $vgpr32
+; SI-NEXT: ; implicit-def: $vgpr32
+; SI-NEXT: ; kill: killed $vgpr32
+; SI-NEXT: ; implicit-def: $vgpr32
+; SI-NEXT: ; kill: killed $vgpr32
+; SI-NEXT: ; implicit-def: $vgpr32
+; SI-NEXT: ; kill: killed $vgpr32
+; SI-NEXT: ; implicit-def: $vgpr32
+; SI-NEXT: ; kill: killed $vgpr32
+; SI-NEXT: ; implicit-def: $vgpr32
+; SI-NEXT: ; kill: killed $vgpr32
+; SI-NEXT: ; implicit-def: $vgpr32
+; SI-NEXT: ; kill: killed $vgpr32
+; SI-NEXT: ; implicit-def: $vgpr32
+; SI-NEXT: ; kill: killed $vgpr32
+; SI-NEXT: ; implicit-def: $vgpr32
+; SI-NEXT: ; kill: killed $vgpr32
+; SI-NEXT: ; implicit-def: $vgpr32
+; SI-NEXT: ; kill: killed $vgpr32
+; SI-NEXT: ; implicit-def: $vgpr32
+; SI-NEXT: ; kill: killed $vgpr32
+; SI-NEXT: ; implicit-def: $vgpr32
+; SI-NEXT: ; kill: killed $vgpr32
+; SI-NEXT: ; implicit-def: $vgpr32
+; SI-NEXT: ; kill: killed $vgpr32
+; SI-NEXT: ; implicit-def: $vgpr32
+; SI-NEXT: ; kill: killed $vgpr32
+; SI-NEXT: ; implicit-def: $vgpr32
+; SI-NEXT: ; kill: killed $vgpr32
+; SI-NEXT: ; implicit-def: $vgpr32
+; SI-NEXT: ; kill: killed $vgpr32
+; SI-NEXT: ; implicit-def: $vgpr32
+; SI-NEXT: ; kill: killed $vgpr32
+; SI-NEXT: ; implicit-def: $vgpr32
+; SI-NEXT: ; kill: killed $vgpr32
+; SI-NEXT: ; implicit-def: $vgpr32
+; SI-NEXT: ; kill: killed $vgpr32
+; SI-NEXT: ; implicit-def: $vgpr32
+; SI-NEXT: ; kill: killed $vgpr32
+; SI-NEXT: ; implicit-def: $vgpr32
+; SI-NEXT: ; kill: killed $vgpr32
+; SI-NEXT: ; implicit-def: $vgpr32
+; SI-NEXT: ; kill: killed $vgpr32
+; SI-NEXT: ; implicit-def: $vgpr32
+; SI-NEXT: ; kill: killed $vgpr32
+; SI-NEXT: ; implicit-def: $vgpr32
+; SI-NEXT: ; kill: killed $vgpr32
+; SI-NEXT: ; implicit-def: $vgpr32
+; SI-NEXT: ; kill: killed $vgpr32
+; SI-NEXT: ; implicit-def: $vgpr32
+; SI-NEXT: ; kill: killed $vgpr32
+; SI-NEXT: ; implicit-def: $vgpr32
+; SI-NEXT: ; kill: killed $vgpr32
+; SI-NEXT: ; implicit-def: $vgpr32
; SI-NEXT: ; implicit-def: $vgpr60
; SI-NEXT: ; implicit-def: $vgpr61
; SI-NEXT: ; implicit-def: $vgpr58
; SI-NEXT: ; implicit-def: $vgpr59
; SI-NEXT: ; implicit-def: $vgpr56
; SI-NEXT: ; implicit-def: $vgpr57
-; SI-NEXT: ; implicit-def: $vgpr42
+; SI-NEXT: ; implicit-def: $vgpr46
; SI-NEXT: ; implicit-def: $vgpr47
-; SI-NEXT: ; implicit-def: $vgpr44
+; SI-NEXT: ; implicit-def: $vgpr51
; SI-NEXT: ; implicit-def: $vgpr45
+; SI-NEXT: ; implicit-def: $vgpr42
; SI-NEXT: ; implicit-def: $vgpr43
-; SI-NEXT: ; implicit-def: $vgpr41
-; SI-NEXT: ; implicit-def: $vgpr55
; SI-NEXT: ; implicit-def: $vgpr53
-; SI-NEXT: ; implicit-def: $vgpr51
+; SI-NEXT: ; implicit-def: $vgpr41
+; SI-NEXT: ; implicit-def: $vgpr38
+; SI-NEXT: ; implicit-def: $vgpr50
; SI-NEXT: ; implicit-def: $vgpr49
; SI-NEXT: ; implicit-def: $vgpr39
; SI-NEXT: ; implicit-def: $vgpr37
; SI-NEXT: ; implicit-def: $vgpr35
; SI-NEXT: ; implicit-def: $vgpr33
-; SI-NEXT: ; implicit-def: $vgpr52
-; SI-NEXT: ; implicit-def: $vgpr46
+; SI-NEXT: ; kill: killed $vgpr32
+; SI-NEXT: ; implicit-def: $vgpr55
+; SI-NEXT: ; implicit-def: $vgpr44
; SI-NEXT: ; implicit-def: $vgpr54
-; SI-NEXT: ; implicit-def: $vgpr50
-; SI-NEXT: ; implicit-def: $vgpr38
-; SI-NEXT: ; implicit-def: $vgpr48
+; SI-NEXT: ; implicit-def: $vgpr52
; SI-NEXT: ; implicit-def: $vgpr40
+; SI-NEXT: ; implicit-def: $vgpr48
; SI-NEXT: ; implicit-def: $vgpr36
-; SI-NEXT: ; implicit-def: $vgpr34
; SI-NEXT: ; implicit-def: $vgpr32
+; SI-NEXT: ; implicit-def: $vgpr34
; SI-NEXT: s_waitcnt vmcnt(1)
; SI-NEXT: v_cmp_ne_u32_e32 vcc, 0, v31
; SI-NEXT: ; implicit-def: $vgpr31
@@ -107358,175 +107445,104 @@ define <64 x half> @bitcast_v16i64_to_v64f16(<16 x i64> %a, i32 %b) {
; SI-NEXT: ; implicit-def: $vgpr31
; SI-NEXT: ; kill: killed $vgpr31
; SI-NEXT: ; implicit-def: $vgpr31
-; SI-NEXT: ; kill: killed $vgpr31
-; SI-NEXT: ; implicit-def: $vgpr31
-; SI-NEXT: ; kill: killed $vgpr31
-; SI-NEXT: ; implicit-def: $vgpr31
-; SI-NEXT: ; kill: killed $vgpr31
-; SI-NEXT: ; implicit-def: $vgpr31
-; SI-NEXT: ; kill: killed $vgpr31
-; SI-NEXT: ; implicit-def: $vgpr31
-; SI-NEXT: ; kill: killed $vgpr31
-; SI-NEXT: ; implicit-def: $vgpr31
-; SI-NEXT: ; kill: killed $vgpr31
-; SI-NEXT: ; implicit-def: $vgpr31
-; SI-NEXT: ; kill: killed $vgpr31
-; SI-NEXT: ; implicit-def: $vgpr31
-; SI-NEXT: ; kill: killed $vgpr31
-; SI-NEXT: ; implicit-def: $vgpr31
-; SI-NEXT: ; kill: killed $vgpr31
-; SI-NEXT: ; implicit-def: $vgpr31
-; SI-NEXT: ; kill: killed $vgpr31
-; SI-NEXT: ; implicit-def: $vgpr31
-; SI-NEXT: ; kill: killed $vgpr31
-; SI-NEXT: ; implicit-def: $vgpr31
-; SI-NEXT: ; kill: killed $vgpr31
-; SI-NEXT: ; implicit-def: $vgpr31
-; SI-NEXT: ; kill: killed $vgpr31
-; SI-NEXT: ; implicit-def: $vgpr31
-; SI-NEXT: ; kill: killed $vgpr31
-; SI-NEXT: ; implicit-def: $vgpr31
-; SI-NEXT: ; kill: killed $vgpr31
-; SI-NEXT: ; implicit-def: $vgpr31
-; SI-NEXT: ; kill: killed $vgpr31
-; SI-NEXT: ; implicit-def: $vgpr31
-; SI-NEXT: ; kill: killed $vgpr31
-; SI-NEXT: ; implicit-def: $vgpr31
-; SI-NEXT: ; kill: killed $vgpr31
-; SI-NEXT: ; implicit-def: $vgpr31
-; SI-NEXT: ; kill: killed $vgpr31
-; SI-NEXT: ; implicit-def: $vgpr31
-; SI-NEXT: ; kill: killed $vgpr31
-; SI-NEXT: ; implicit-def: $vgpr31
-; SI-NEXT: ; kill: killed $vgpr31
-; SI-NEXT: ; implicit-def: $vgpr31
-; SI-NEXT: ; kill: killed $vgpr31
-; SI-NEXT: ; implicit-def: $vgpr31
-; SI-NEXT: ; kill: killed $vgpr31
-; SI-NEXT: ; implicit-def: $vgpr31
-; SI-NEXT: ; kill: killed $vgpr31
-; SI-NEXT: ; implicit-def: $vgpr31
-; SI-NEXT: ; kill: killed $vgpr31
-; SI-NEXT: ; implicit-def: $vgpr31
-; SI-NEXT: ; kill: killed $vgpr31
-; SI-NEXT: ; implicit-def: $vgpr31
-; SI-NEXT: ; kill: killed $vgpr31
-; SI-NEXT: ; implicit-def: $vgpr31
-; SI-NEXT: ; kill: killed $vgpr31
-; SI-NEXT: ; implicit-def: $vgpr31
-; SI-NEXT: ; kill: killed $vgpr31
-; SI-NEXT: ; implicit-def: $vgpr31
-; SI-NEXT: ; kill: killed $vgpr31
-; SI-NEXT: ; implicit-def: $vgpr31
-; SI-NEXT: ; kill: killed $vgpr31
-; SI-NEXT: ; implicit-def: $vgpr31
; SI-NEXT: s_and_saveexec_b64 s[4:5], vcc
; SI-NEXT: s_xor_b64 s[4:5], exec, s[4:5]
; SI-NEXT: s_cbranch_execz .LBB64_2
; SI-NEXT: ; %bb.1: ; %cmp.false
; SI-NEXT: v_lshrrev_b32_e32 v31, 16, v62
-; SI-NEXT: v_cvt_f32_f16_e32 v32, v31
+; SI-NEXT: v_cvt_f32_f16_e32 v34, v31
; SI-NEXT: s_waitcnt vmcnt(0)
; SI-NEXT: v_lshrrev_b32_e32 v31, 16, v63
-; SI-NEXT: v_cvt_f32_f16_e32 v34, v31
+; SI-NEXT: v_cvt_f32_f16_e32 v36, v31
; SI-NEXT: v_lshrrev_b32_e32 v31, 16, v30
; SI-NEXT: v_cvt_f32_f16_e32 v40, v31
; SI-NEXT: v_lshrrev_b32_e32 v31, 16, v29
-; SI-NEXT: v_cvt_f32_f16_e32 v38, v31
-; SI-NEXT: v_lshrrev_b32_e32 v31, 16, v28
; SI-NEXT: v_cvt_f32_f16_e32 v54, v31
+; SI-NEXT: v_lshrrev_b32_e32 v31, 16, v28
+; SI-NEXT: v_cvt_f32_f16_e32 v55, v31
; SI-NEXT: v_lshrrev_b32_e32 v31, 16, v27
-; SI-NEXT: v_cvt_f32_f16_e32 v52, v31
-; SI-NEXT: v_lshrrev_b32_e32 v31, 16, v26
; SI-NEXT: v_cvt_f32_f16_e32 v31, v31
-; SI-NEXT: v_cvt_f32_f16_e32 v27, v27
-; SI-NEXT: v_cvt_f32_f16_e32 v26, v26
-; SI-NEXT: v_cvt_f32_f16_e32 v36, v63
+; SI-NEXT: v_lshrrev_b32_e32 v32, 16, v8
+; SI-NEXT: v_cvt_f32_f16_e32 v38, v32
+; SI-NEXT: v_lshrrev_b32_e32 v32, 16, v7
; SI-NEXT: buffer_store_dword v31, off, s[0:3], s32 offset:76 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_lshrrev_b32_e32 v31, 16, v25
+; SI-NEXT: v_lshrrev_b32_e32 v31, 16, v26
; SI-NEXT: v_cvt_f32_f16_e32 v31, v31
-; SI-NEXT: v_cvt_f32_f16_e32 v25, v25
-; SI-NEXT: v_cvt_f32_f16_e32 v48, v30
-; SI-NEXT: v_cvt_f32_f16_e32 v50, v29
+; SI-NEXT: v_cvt_f32_f16_e32 v41, v32
+; SI-NEXT: v_lshrrev_b32_e32 v32, 16, v6
+; SI-NEXT: v_cvt_f32_f16_e32 v43, v32
; SI-NEXT: buffer_store_dword v31, off, s[0:3], s32 offset:84 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_lshrrev_b32_e32 v31, 16, v24
+; SI-NEXT: v_lshrrev_b32_e32 v31, 16, v25
; SI-NEXT: v_cvt_f32_f16_e32 v31, v31
-; SI-NEXT: v_cvt_f32_f16_e32 v24, v24
-; SI-NEXT: v_cvt_f32_f16_e32 v46, v28
-; SI-NEXT: buffer_store_dword v27, off, s[0:3], s32 offset:80 ; 4-byte Folded Spill
+; SI-NEXT: v_lshrrev_b32_e32 v32, 16, v5
+; SI-NEXT: v_cvt_f32_f16_e32 v45, v32
+; SI-NEXT: v_lshrrev_b32_e32 v32, 16, v4
; SI-NEXT: buffer_store_dword v31, off, s[0:3], s32 offset:92 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
+; SI-NEXT: v_lshrrev_b32_e32 v31, 16, v24
+; SI-NEXT: v_cvt_f32_f16_e32 v31, v31
+; SI-NEXT: v_cvt_f32_f16_e32 v28, v28
+; SI-NEXT: v_cvt_f32_f16_e32 v27, v27
+; SI-NEXT: v_cvt_f32_f16_e32 v26, v26
+; SI-NEXT: buffer_store_dword v31, off, s[0:3], s32 offset:100 ; 4-byte Folded Spill
+; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_lshrrev_b32_e32 v31, 16, v23
; SI-NEXT: v_cvt_f32_f16_e32 v31, v31
+; SI-NEXT: v_cvt_f32_f16_e32 v25, v25
+; SI-NEXT: v_cvt_f32_f16_e32 v24, v24
; SI-NEXT: v_cvt_f32_f16_e32 v23, v23
-; SI-NEXT: buffer_store_dword v26, off, s[0:3], s32 offset:88 ; 4-byte Folded Spill
-; SI-NEXT: buffer_store_dword v25, off, s[0:3], s32 offset:96 ; 4-byte Folded Spill
-; SI-NEXT: buffer_store_dword v31, off, s[0:3], s32 offset:100 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v31, off, s[0:3], s32 offset:108 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_lshrrev_b32_e32 v31, 16, v22
; SI-NEXT: v_cvt_f32_f16_e32 v31, v31
; SI-NEXT: v_cvt_f32_f16_e32 v22, v22
-; SI-NEXT: buffer_store_dword v24, off, s[0:3], s32 offset:104 ; 4-byte Folded Spill
-; SI-NEXT: buffer_store_dword v23, off, s[0:3], s32 offset:108 ; 4-byte Folded Spill
-; SI-NEXT: buffer_store_dword v31, off, s[0:3], s32 offset:112 ; 4-byte Folded Spill
+; SI-NEXT: v_cvt_f32_f16_e32 v8, v8
+; SI-NEXT: v_cvt_f32_f16_e32 v47, v32
+; SI-NEXT: buffer_store_dword v31, off, s[0:3], s32 offset:120 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_lshrrev_b32_e32 v31, 16, v21
; SI-NEXT: v_cvt_f32_f16_e32 v31, v31
; SI-NEXT: v_cvt_f32_f16_e32 v21, v21
-; SI-NEXT: buffer_store_dword v22, off, s[0:3], s32 offset:116 ; 4-byte Folded Spill
-; SI-NEXT: v_cvt_f32_f16_e32 v44, v5
-; SI-NEXT: buffer_store_dword v31, off, s[0:3], s32 offset:120 ; 4-byte Folded Spill
+; SI-NEXT: v_lshrrev_b32_e32 v32, 16, v3
+; SI-NEXT: v_cvt_f32_f16_e32 v57, v32
+; SI-NEXT: buffer_store_dword v31, off, s[0:3], s32 offset:128 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_lshrrev_b32_e32 v31, 16, v20
; SI-NEXT: v_cvt_f32_f16_e32 v31, v31
; SI-NEXT: v_cvt_f32_f16_e32 v20, v20
-; SI-NEXT: buffer_store_dword v21, off, s[0:3], s32 offset:124 ; 4-byte Folded Spill
-; SI-NEXT: v_cvt_f32_f16_e32 v42, v4
-; SI-NEXT: buffer_store_dword v31, off, s[0:3], s32 offset:128 ; 4-byte Folded Spill
+; SI-NEXT: v_lshrrev_b32_e32 v32, 16, v2
+; SI-NEXT: v_cvt_f32_f16_e32 v59, v32
+; SI-NEXT: buffer_store_dword v31, off, s[0:3], s32 offset:136 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_lshrrev_b32_e32 v31, 16, v19
; SI-NEXT: v_cvt_f32_f16_e32 v31, v31
; SI-NEXT: v_cvt_f32_f16_e32 v19, v19
-; SI-NEXT: buffer_store_dword v20, off, s[0:3], s32 offset:132 ; 4-byte Folded Spill
-; SI-NEXT: v_cvt_f32_f16_e32 v56, v3
-; SI-NEXT: buffer_store_dword v31, off, s[0:3], s32 offset:136 ; 4-byte Folded Spill
+; SI-NEXT: v_lshrrev_b32_e32 v32, 16, v1
+; SI-NEXT: v_cvt_f32_f16_e32 v61, v32
+; SI-NEXT: buffer_store_dword v31, off, s[0:3], s32 offset:144 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_lshrrev_b32_e32 v31, 16, v18
; SI-NEXT: v_cvt_f32_f16_e32 v31, v31
; SI-NEXT: v_cvt_f32_f16_e32 v18, v18
-; SI-NEXT: buffer_store_dword v19, off, s[0:3], s32 offset:140 ; 4-byte Folded Spill
-; SI-NEXT: v_cvt_f32_f16_e32 v58, v2
-; SI-NEXT: buffer_store_dword v31, off, s[0:3], s32 offset:144 ; 4-byte Folded Spill
+; SI-NEXT: v_cvt_f32_f16_e32 v32, v62
+; SI-NEXT: v_cvt_f32_f16_e32 v48, v63
+; SI-NEXT: buffer_store_dword v31, off, s[0:3], s32 offset:152 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_lshrrev_b32_e32 v31, 16, v17
; SI-NEXT: v_cvt_f32_f16_e32 v31, v31
; SI-NEXT: v_cvt_f32_f16_e32 v17, v17
-; SI-NEXT: buffer_store_dword v18, off, s[0:3], s32 offset:148 ; 4-byte Folded Spill
-; SI-NEXT: v_cvt_f32_f16_e32 v60, v1
-; SI-NEXT: buffer_store_dword v31, off, s[0:3], s32 offset:152 ; 4-byte Folded Spill
+; SI-NEXT: v_cvt_f32_f16_e32 v52, v30
+; SI-NEXT: v_cvt_f32_f16_e32 v44, v29
+; SI-NEXT: buffer_store_dword v31, off, s[0:3], s32 offset:160 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_lshrrev_b32_e32 v31, 16, v16
; SI-NEXT: v_cvt_f32_f16_e32 v31, v31
; SI-NEXT: v_cvt_f32_f16_e32 v16, v16
-; SI-NEXT: buffer_store_dword v17, off, s[0:3], s32 offset:156 ; 4-byte Folded Spill
-; SI-NEXT: ; implicit-def: $vgpr17
-; SI-NEXT: ; implicit-def: $vgpr18
-; SI-NEXT: ; implicit-def: $vgpr19
-; SI-NEXT: ; implicit-def: $vgpr20
-; SI-NEXT: ; implicit-def: $vgpr21
-; SI-NEXT: ; implicit-def: $vgpr22
-; SI-NEXT: ; implicit-def: $vgpr23
-; SI-NEXT: ; implicit-def: $vgpr24
-; SI-NEXT: ; implicit-def: $vgpr25
-; SI-NEXT: ; implicit-def: $vgpr26
-; SI-NEXT: ; implicit-def: $vgpr27
-; SI-NEXT: ; implicit-def: $vgpr28
-; SI-NEXT: ; implicit-def: $vgpr29
-; SI-NEXT: ; implicit-def: $vgpr30
-; SI-NEXT: ; implicit-def: $vgpr63
-; SI-NEXT: buffer_store_dword v31, off, s[0:3], s32 offset:160 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v28, off, s[0:3], s32 offset:80 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v27, off, s[0:3], s32 offset:88 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v31, off, s[0:3], s32 offset:168 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_lshrrev_b32_e32 v31, 16, v15
; SI-NEXT: v_cvt_f32_f16_e32 v33, v31
@@ -107539,18 +107555,8 @@ define <64 x half> @bitcast_v16i64_to_v64f16(<16 x i64> %a, i32 %b) {
; SI-NEXT: v_lshrrev_b32_e32 v31, 16, v11
; SI-NEXT: v_cvt_f32_f16_e32 v49, v31
; SI-NEXT: v_lshrrev_b32_e32 v31, 16, v10
-; SI-NEXT: v_cvt_f32_f16_e32 v51, v31
+; SI-NEXT: v_cvt_f32_f16_e32 v50, v31
; SI-NEXT: v_lshrrev_b32_e32 v31, 16, v9
-; SI-NEXT: v_cvt_f32_f16_e32 v53, v31
-; SI-NEXT: v_lshrrev_b32_e32 v31, 16, v8
-; SI-NEXT: v_cvt_f32_f16_e32 v55, v31
-; SI-NEXT: v_lshrrev_b32_e32 v31, 16, v7
-; SI-NEXT: v_cvt_f32_f16_e32 v41, v31
-; SI-NEXT: v_lshrrev_b32_e32 v31, 16, v6
-; SI-NEXT: v_cvt_f32_f16_e32 v43, v31
-; SI-NEXT: v_lshrrev_b32_e32 v31, 16, v5
-; SI-NEXT: v_cvt_f32_f16_e32 v45, v31
-; SI-NEXT: v_lshrrev_b32_e32 v31, 16, v4
; SI-NEXT: v_cvt_f32_f16_e32 v15, v15
; SI-NEXT: v_cvt_f32_f16_e32 v14, v14
; SI-NEXT: v_cvt_f32_f16_e32 v13, v13
@@ -107558,28 +107564,33 @@ define <64 x half> @bitcast_v16i64_to_v64f16(<16 x i64> %a, i32 %b) {
; SI-NEXT: v_cvt_f32_f16_e32 v11, v11
; SI-NEXT: v_cvt_f32_f16_e32 v10, v10
; SI-NEXT: v_cvt_f32_f16_e32 v9, v9
-; SI-NEXT: v_cvt_f32_f16_e32 v8, v8
-; SI-NEXT: v_cvt_f32_f16_e32 v7, v7
-; SI-NEXT: v_cvt_f32_f16_e32 v6, v6
-; SI-NEXT: v_cvt_f32_f16_e32 v47, v31
-; SI-NEXT: v_lshrrev_b32_e32 v31, 16, v3
-; SI-NEXT: v_cvt_f32_f16_e32 v57, v31
-; SI-NEXT: v_lshrrev_b32_e32 v31, 16, v2
-; SI-NEXT: v_cvt_f32_f16_e32 v59, v31
-; SI-NEXT: v_lshrrev_b32_e32 v31, 16, v1
-; SI-NEXT: v_cvt_f32_f16_e32 v61, v31
-; SI-NEXT: v_cvt_f32_f16_e32 v31, v62
-; SI-NEXT: buffer_store_dword v16, off, s[0:3], s32 offset:164 ; 4-byte Folded Spill
-; SI-NEXT: buffer_store_dword v15, off, s[0:3], s32 offset:168 ; 4-byte Folded Spill
-; SI-NEXT: buffer_store_dword v14, off, s[0:3], s32 offset:172 ; 4-byte Folded Spill
-; SI-NEXT: buffer_store_dword v13, off, s[0:3], s32 offset:176 ; 4-byte Folded Spill
-; SI-NEXT: buffer_store_dword v12, off, s[0:3], s32 offset:180 ; 4-byte Folded Spill
-; SI-NEXT: buffer_store_dword v11, off, s[0:3], s32 offset:184 ; 4-byte Folded Spill
-; SI-NEXT: buffer_store_dword v10, off, s[0:3], s32 offset:188 ; 4-byte Folded Spill
-; SI-NEXT: buffer_store_dword v9, off, s[0:3], s32 offset:192 ; 4-byte Folded Spill
-; SI-NEXT: buffer_store_dword v8, off, s[0:3], s32 offset:196 ; 4-byte Folded Spill
-; SI-NEXT: buffer_store_dword v7, off, s[0:3], s32 offset:200 ; 4-byte Folded Spill
-; SI-NEXT: buffer_store_dword v6, off, s[0:3], s32 offset:204 ; 4-byte Folded Spill
+; SI-NEXT: v_cvt_f32_f16_e32 v31, v31
+; SI-NEXT: buffer_store_dword v26, off, s[0:3], s32 offset:96 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v25, off, s[0:3], s32 offset:104 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v24, off, s[0:3], s32 offset:112 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v23, off, s[0:3], s32 offset:116 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v22, off, s[0:3], s32 offset:124 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v21, off, s[0:3], s32 offset:132 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v20, off, s[0:3], s32 offset:140 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v19, off, s[0:3], s32 offset:148 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v18, off, s[0:3], s32 offset:156 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v17, off, s[0:3], s32 offset:164 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v16, off, s[0:3], s32 offset:172 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v15, off, s[0:3], s32 offset:176 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v14, off, s[0:3], s32 offset:180 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v13, off, s[0:3], s32 offset:184 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v12, off, s[0:3], s32 offset:188 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v11, off, s[0:3], s32 offset:192 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v10, off, s[0:3], s32 offset:196 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v9, off, s[0:3], s32 offset:200 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v8, off, s[0:3], s32 offset:204 ; 4-byte Folded Spill
+; SI-NEXT: v_cvt_f32_f16_e32 v53, v7
+; SI-NEXT: v_cvt_f32_f16_e32 v42, v6
+; SI-NEXT: v_cvt_f32_f16_e32 v51, v5
+; SI-NEXT: v_cvt_f32_f16_e32 v46, v4
+; SI-NEXT: v_cvt_f32_f16_e32 v56, v3
+; SI-NEXT: v_cvt_f32_f16_e32 v58, v2
+; SI-NEXT: v_cvt_f32_f16_e32 v60, v1
; SI-NEXT: ; implicit-def: $vgpr1
; SI-NEXT: ; implicit-def: $vgpr2
; SI-NEXT: ; implicit-def: $vgpr3
@@ -107596,7 +107607,22 @@ define <64 x half> @bitcast_v16i64_to_v64f16(<16 x i64> %a, i32 %b) {
; SI-NEXT: ; implicit-def: $vgpr14
; SI-NEXT: ; implicit-def: $vgpr15
; SI-NEXT: ; implicit-def: $vgpr16
+; SI-NEXT: ; implicit-def: $vgpr17
+; SI-NEXT: ; implicit-def: $vgpr18
+; SI-NEXT: ; implicit-def: $vgpr19
+; SI-NEXT: ; implicit-def: $vgpr20
+; SI-NEXT: ; implicit-def: $vgpr21
+; SI-NEXT: ; implicit-def: $vgpr22
+; SI-NEXT: ; implicit-def: $vgpr23
+; SI-NEXT: ; implicit-def: $vgpr24
+; SI-NEXT: ; implicit-def: $vgpr25
+; SI-NEXT: ; implicit-def: $vgpr26
+; SI-NEXT: ; implicit-def: $vgpr27
+; SI-NEXT: ; implicit-def: $vgpr28
+; SI-NEXT: ; implicit-def: $vgpr29
+; SI-NEXT: ; implicit-def: $vgpr30
; SI-NEXT: ; implicit-def: $vgpr62
+; SI-NEXT: ; implicit-def: $vgpr63
; SI-NEXT: .LBB64_2: ; %Flow
; SI-NEXT: s_andn2_saveexec_b64 s[4:5], s[4:5]
; SI-NEXT: s_cbranch_execz .LBB64_4
@@ -107606,9 +107632,9 @@ define <64 x half> @bitcast_v16i64_to_v64f16(<16 x i64> %a, i32 %b) {
; SI-NEXT: v_add_i32_e32 v3, vcc, 3, v3
; SI-NEXT: v_addc_u32_e32 v4, vcc, 0, v4, vcc
; SI-NEXT: v_add_i32_e32 v5, vcc, 3, v5
-; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_addc_u32_e32 v6, vcc, 0, v6, vcc
; SI-NEXT: v_add_i32_e32 v7, vcc, 3, v7
+; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_addc_u32_e32 v8, vcc, 0, v8, vcc
; SI-NEXT: v_add_i32_e32 v9, vcc, 3, v9
; SI-NEXT: v_addc_u32_e32 v10, vcc, 0, v10, vcc
@@ -107619,114 +107645,111 @@ define <64 x half> @bitcast_v16i64_to_v64f16(<16 x i64> %a, i32 %b) {
; SI-NEXT: v_add_i32_e32 v15, vcc, 3, v15
; SI-NEXT: v_addc_u32_e32 v16, vcc, 0, v16, vcc
; SI-NEXT: v_add_i32_e32 v17, vcc, 3, v17
-; SI-NEXT: v_addc_u32_e32 v18, vcc, 0, v18, vcc
-; SI-NEXT: v_add_i32_e32 v19, vcc, 3, v19
-; SI-NEXT: v_addc_u32_e32 v20, vcc, 0, v20, vcc
-; SI-NEXT: v_add_i32_e32 v21, vcc, 3, v21
-; SI-NEXT: v_addc_u32_e32 v22, vcc, 0, v22, vcc
-; SI-NEXT: v_add_i32_e32 v23, vcc, 3, v23
-; SI-NEXT: v_lshrrev_b32_e32 v53, 16, v9
-; SI-NEXT: v_addc_u32_e32 v24, vcc, 0, v24, vcc
-; SI-NEXT: v_cvt_f32_f16_e32 v9, v9
-; SI-NEXT: v_add_i32_e32 v25, vcc, 3, v25
-; SI-NEXT: v_addc_u32_e32 v26, vcc, 0, v26, vcc
-; SI-NEXT: v_lshrrev_b32_e32 v50, 16, v26
-; SI-NEXT: v_add_i32_e32 v27, vcc, 3, v27
-; SI-NEXT: buffer_store_dword v9, off, s[0:3], s32 offset:192 ; 4-byte Folded Spill
-; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_cvt_f32_f16_e32 v9, v50
; SI-NEXT: v_lshrrev_b32_e32 v1, 16, v5
-; SI-NEXT: v_addc_u32_e32 v28, vcc, 0, v28, vcc
+; SI-NEXT: v_addc_u32_e32 v18, vcc, 0, v18, vcc
; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:208 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_lshrrev_b32_e32 v1, 16, v6
-; SI-NEXT: v_add_i32_e32 v29, vcc, 3, v29
+; SI-NEXT: v_add_i32_e32 v19, vcc, 3, v19
; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:212 ; 4-byte Folded Spill
-; SI-NEXT: v_lshrrev_b32_e32 v48, 16, v25
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_lshrrev_b32_e32 v1, 16, v29
-; SI-NEXT: v_cvt_f32_f16_e32 v1, v1
-; SI-NEXT: buffer_store_dword v9, off, s[0:3], s32 offset:76 ; 4-byte Folded Spill
+; SI-NEXT: v_lshrrev_b32_e32 v1, 16, v7
+; SI-NEXT: v_addc_u32_e32 v20, vcc, 0, v20, vcc
+; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:216 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_cvt_f32_f16_e32 v9, v48
-; SI-NEXT: v_lshrrev_b32_e32 v46, 16, v23
-; SI-NEXT: v_lshrrev_b32_e32 v38, 16, v24
-; SI-NEXT: v_lshrrev_b32_e32 v34, 16, v22
-; SI-NEXT: buffer_store_dword v9, off, s[0:3], s32 offset:84 ; 4-byte Folded Spill
+; SI-NEXT: v_lshrrev_b32_e32 v1, 16, v8
+; SI-NEXT: v_add_i32_e32 v21, vcc, 3, v21
+; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:220 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_cvt_f32_f16_e32 v9, v38
-; SI-NEXT: v_mov_b32_e32 v38, v1
-; SI-NEXT: v_cvt_f32_f16_e32 v1, v46
-; SI-NEXT: v_lshrrev_b32_e32 v32, 16, v21
-; SI-NEXT: v_lshrrev_b32_e32 v2, 16, v20
+; SI-NEXT: v_lshrrev_b32_e32 v1, 16, v21
+; SI-NEXT: v_cvt_f32_f16_e32 v1, v1
+; SI-NEXT: v_lshrrev_b32_e32 v31, 16, v20
; SI-NEXT: v_lshrrev_b32_e32 v61, 16, v19
-; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:100 ; 4-byte Folded Spill
-; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_cvt_f32_f16_e32 v1, v34
; SI-NEXT: v_lshrrev_b32_e32 v59, 16, v18
+; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:128 ; 4-byte Folded Spill
+; SI-NEXT: s_waitcnt expcnt(0)
+; SI-NEXT: v_cvt_f32_f16_e32 v1, v31
; SI-NEXT: v_lshrrev_b32_e32 v57, 16, v17
; SI-NEXT: v_lshrrev_b32_e32 v47, 16, v16
-; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:112 ; 4-byte Folded Spill
-; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_cvt_f32_f16_e32 v1, v32
-; SI-NEXT: v_addc_u32_e32 v30, vcc, 0, v30, vcc
-; SI-NEXT: s_waitcnt vmcnt(7)
-; SI-NEXT: v_add_i32_e32 v42, vcc, 3, v63
-; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:120 ; 4-byte Folded Spill
-; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_cvt_f32_f16_e32 v1, v2
-; SI-NEXT: v_lshrrev_b32_e32 v43, 16, v14
-; SI-NEXT: v_addc_u32_e32 v44, vcc, 0, v62, vcc
-; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:128 ; 4-byte Folded Spill
+; SI-NEXT: v_addc_u32_e32 v22, vcc, 0, v22, vcc
+; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:136 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_cvt_f32_f16_e32 v1, v61
-; SI-NEXT: v_lshrrev_b32_e32 v62, 16, v42
-; SI-NEXT: v_lshrrev_b32_e32 v63, 16, v44
-; SI-NEXT: v_cvt_f32_f16_e32 v31, v44
-; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:136 ; 4-byte Folded Spill
+; SI-NEXT: v_add_i32_e32 v23, vcc, 3, v23
+; SI-NEXT: v_addc_u32_e32 v24, vcc, 0, v24, vcc
+; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:144 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_cvt_f32_f16_e32 v1, v59
-; SI-NEXT: v_cvt_f32_f16_e32 v36, v42
-; SI-NEXT: v_cvt_f32_f16_e32 v44, v5
-; SI-NEXT: v_lshrrev_b32_e32 v5, 16, v4
-; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:144 ; 4-byte Folded Spill
+; SI-NEXT: v_add_i32_e32 v25, vcc, 3, v25
+; SI-NEXT: v_lshrrev_b32_e32 v41, 16, v9
+; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:152 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_cvt_f32_f16_e32 v1, v57
-; SI-NEXT: v_cvt_f32_f16_e32 v42, v4
-; SI-NEXT: v_lshrrev_b32_e32 v4, 16, v3
-; SI-NEXT: v_cvt_f32_f16_e32 v56, v3
-; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:152 ; 4-byte Folded Spill
+; SI-NEXT: v_lshrrev_b32_e32 v2, 16, v10
+; SI-NEXT: v_addc_u32_e32 v26, vcc, 0, v26, vcc
+; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:160 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_cvt_f32_f16_e32 v1, v47
+; SI-NEXT: v_cvt_f32_f16_e32 v10, v10
+; SI-NEXT: v_cvt_f32_f16_e32 v9, v9
+; SI-NEXT: v_add_i32_e32 v27, vcc, 3, v27
+; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:168 ; 4-byte Folded Spill
+; SI-NEXT: s_waitcnt expcnt(0)
+; SI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:220 ; 4-byte Folded Reload
+; SI-NEXT: v_addc_u32_e32 v28, vcc, 0, v28, vcc
+; SI-NEXT: v_lshrrev_b32_e32 v50, 16, v27
+; SI-NEXT: v_lshrrev_b32_e32 v52, 16, v28
+; SI-NEXT: buffer_store_dword v10, off, s[0:3], s32 offset:196 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v9, off, s[0:3], s32 offset:200 ; 4-byte Folded Spill
+; SI-NEXT: s_waitcnt expcnt(0)
+; SI-NEXT: v_cvt_f32_f16_e32 v9, v52
+; SI-NEXT: v_cvt_f32_f16_e32 v10, v50
+; SI-NEXT: v_lshrrev_b32_e32 v38, 16, v25
+; SI-NEXT: v_lshrrev_b32_e32 v55, 16, v26
+; SI-NEXT: v_cvt_f32_f16_e32 v31, v41
+; SI-NEXT: buffer_store_dword v10, off, s[0:3], s32 offset:76 ; 4-byte Folded Spill
+; SI-NEXT: s_waitcnt expcnt(0)
+; SI-NEXT: v_cvt_f32_f16_e32 v10, v55
+; SI-NEXT: v_mov_b32_e32 v55, v9
+; SI-NEXT: v_cvt_f32_f16_e32 v9, v38
+; SI-NEXT: v_lshrrev_b32_e32 v43, 16, v14
+; SI-NEXT: v_cvt_f32_f16_e32 v51, v5
+; SI-NEXT: v_lshrrev_b32_e32 v5, 16, v4
+; SI-NEXT: v_cvt_f32_f16_e32 v46, v4
+; SI-NEXT: v_lshrrev_b32_e32 v4, 16, v3
+; SI-NEXT: v_cvt_f32_f16_e32 v56, v3
; SI-NEXT: v_lshrrev_b32_e32 v3, 16, v35
; SI-NEXT: v_cvt_f32_f16_e32 v58, v35
; SI-NEXT: v_cvt_f32_f16_e32 v35, v43
-; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:160 ; 4-byte Folded Spill
-; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:212 ; 4-byte Folded Reload
-; SI-NEXT: v_lshrrev_b32_e32 v41, 16, v7
-; SI-NEXT: v_lshrrev_b32_e32 v55, 16, v8
+; SI-NEXT: v_add_i32_e32 v29, vcc, 3, v29
+; SI-NEXT: v_addc_u32_e32 v30, vcc, 0, v30, vcc
; SI-NEXT: v_cvt_f32_f16_e32 v8, v8
-; SI-NEXT: v_cvt_f32_f16_e32 v7, v7
-; SI-NEXT: v_cvt_f32_f16_e32 v6, v6
-; SI-NEXT: v_lshrrev_b32_e32 v51, 16, v10
+; SI-NEXT: s_waitcnt vmcnt(14)
+; SI-NEXT: v_add_i32_e32 v53, vcc, 3, v63
+; SI-NEXT: v_addc_u32_e32 v42, vcc, 0, v62, vcc
+; SI-NEXT: v_lshrrev_b32_e32 v62, 16, v53
+; SI-NEXT: v_lshrrev_b32_e32 v63, 16, v42
+; SI-NEXT: v_cvt_f32_f16_e32 v48, v53
+; SI-NEXT: buffer_store_dword v8, off, s[0:3], s32 offset:204 ; 4-byte Folded Spill
+; SI-NEXT: v_cvt_f32_f16_e32 v53, v7
+; SI-NEXT: v_cvt_f32_f16_e32 v7, v63
+; SI-NEXT: s_waitcnt expcnt(0)
+; SI-NEXT: v_cvt_f32_f16_e32 v8, v62
; SI-NEXT: v_lshrrev_b32_e32 v49, 16, v11
; SI-NEXT: v_lshrrev_b32_e32 v39, 16, v12
; SI-NEXT: v_lshrrev_b32_e32 v37, 16, v13
; SI-NEXT: v_lshrrev_b32_e32 v45, 16, v15
-; SI-NEXT: v_lshrrev_b32_e32 v52, 16, v27
-; SI-NEXT: v_lshrrev_b32_e32 v54, 16, v28
+; SI-NEXT: v_lshrrev_b32_e32 v44, 16, v22
+; SI-NEXT: v_lshrrev_b32_e32 v34, 16, v23
+; SI-NEXT: v_lshrrev_b32_e32 v36, 16, v24
+; SI-NEXT: v_lshrrev_b32_e32 v54, 16, v29
; SI-NEXT: v_lshrrev_b32_e32 v40, 16, v30
+; SI-NEXT: v_cvt_f32_f16_e32 v32, v42
; SI-NEXT: v_cvt_f32_f16_e32 v30, v30
; SI-NEXT: v_cvt_f32_f16_e32 v29, v29
-; SI-NEXT: v_cvt_f32_f16_e32 v28, v28
-; SI-NEXT: buffer_store_dword v8, off, s[0:3], s32 offset:196 ; 4-byte Folded Spill
-; SI-NEXT: buffer_store_dword v7, off, s[0:3], s32 offset:200 ; 4-byte Folded Spill
-; SI-NEXT: buffer_store_dword v6, off, s[0:3], s32 offset:204 ; 4-byte Folded Spill
-; SI-NEXT: s_waitcnt expcnt(0)
+; SI-NEXT: v_cvt_f32_f16_e32 v42, v6
; SI-NEXT: v_lshrrev_b32_e32 v6, 16, v33
-; SI-NEXT: v_cvt_f32_f16_e32 v7, v63
-; SI-NEXT: v_cvt_f32_f16_e32 v8, v62
+; SI-NEXT: v_cvt_f32_f16_e32 v28, v28
; SI-NEXT: v_cvt_f32_f16_e32 v27, v27
; SI-NEXT: v_cvt_f32_f16_e32 v26, v26
; SI-NEXT: v_cvt_f32_f16_e32 v25, v25
@@ -107744,51 +107767,59 @@ define <64 x half> @bitcast_v16i64_to_v64f16(<16 x i64> %a, i32 %b) {
; SI-NEXT: v_cvt_f32_f16_e32 v13, v13
; SI-NEXT: v_cvt_f32_f16_e32 v12, v12
; SI-NEXT: v_cvt_f32_f16_e32 v11, v11
-; SI-NEXT: v_cvt_f32_f16_e32 v10, v10
; SI-NEXT: v_cvt_f32_f16_e32 v60, v33
; SI-NEXT: v_cvt_f32_f16_e32 v40, v40
; SI-NEXT: v_cvt_f32_f16_e32 v54, v54
-; SI-NEXT: v_cvt_f32_f16_e32 v52, v52
+; SI-NEXT: buffer_store_dword v9, off, s[0:3], s32 offset:92 ; 4-byte Folded Spill
+; SI-NEXT: s_waitcnt expcnt(0)
+; SI-NEXT: v_cvt_f32_f16_e32 v9, v36
+; SI-NEXT: v_mov_b32_e32 v36, v8
+; SI-NEXT: v_cvt_f32_f16_e32 v8, v34
+; SI-NEXT: s_waitcnt vmcnt(5)
+; SI-NEXT: v_cvt_f32_f16_e32 v38, v1
+; SI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:216 ; 4-byte Folded Reload
+; SI-NEXT: v_mov_b32_e32 v34, v7
+; SI-NEXT: v_cvt_f32_f16_e32 v7, v44
; SI-NEXT: v_cvt_f32_f16_e32 v33, v45
; SI-NEXT: v_cvt_f32_f16_e32 v37, v37
; SI-NEXT: v_cvt_f32_f16_e32 v39, v39
; SI-NEXT: v_cvt_f32_f16_e32 v49, v49
-; SI-NEXT: v_cvt_f32_f16_e32 v51, v51
-; SI-NEXT: v_cvt_f32_f16_e32 v53, v53
-; SI-NEXT: v_cvt_f32_f16_e32 v55, v55
-; SI-NEXT: v_cvt_f32_f16_e32 v41, v41
+; SI-NEXT: v_cvt_f32_f16_e32 v50, v2
; SI-NEXT: v_cvt_f32_f16_e32 v47, v5
; SI-NEXT: v_cvt_f32_f16_e32 v57, v4
; SI-NEXT: v_cvt_f32_f16_e32 v59, v3
; SI-NEXT: v_cvt_f32_f16_e32 v61, v6
-; SI-NEXT: v_mov_b32_e32 v50, v29
-; SI-NEXT: v_mov_b32_e32 v48, v30
-; SI-NEXT: v_mov_b32_e32 v46, v28
-; SI-NEXT: v_mov_b32_e32 v34, v8
-; SI-NEXT: v_mov_b32_e32 v32, v7
-; SI-NEXT: buffer_store_dword v27, off, s[0:3], s32 offset:80 ; 4-byte Folded Spill
-; SI-NEXT: buffer_store_dword v26, off, s[0:3], s32 offset:88 ; 4-byte Folded Spill
-; SI-NEXT: buffer_store_dword v25, off, s[0:3], s32 offset:96 ; 4-byte Folded Spill
-; SI-NEXT: buffer_store_dword v24, off, s[0:3], s32 offset:104 ; 4-byte Folded Spill
-; SI-NEXT: buffer_store_dword v23, off, s[0:3], s32 offset:108 ; 4-byte Folded Spill
-; SI-NEXT: buffer_store_dword v22, off, s[0:3], s32 offset:116 ; 4-byte Folded Spill
-; SI-NEXT: buffer_store_dword v21, off, s[0:3], s32 offset:124 ; 4-byte Folded Spill
-; SI-NEXT: buffer_store_dword v20, off, s[0:3], s32 offset:132 ; 4-byte Folded Spill
-; SI-NEXT: buffer_store_dword v19, off, s[0:3], s32 offset:140 ; 4-byte Folded Spill
-; SI-NEXT: buffer_store_dword v18, off, s[0:3], s32 offset:148 ; 4-byte Folded Spill
-; SI-NEXT: buffer_store_dword v17, off, s[0:3], s32 offset:156 ; 4-byte Folded Spill
-; SI-NEXT: buffer_store_dword v16, off, s[0:3], s32 offset:164 ; 4-byte Folded Spill
-; SI-NEXT: buffer_store_dword v15, off, s[0:3], s32 offset:168 ; 4-byte Folded Spill
-; SI-NEXT: buffer_store_dword v14, off, s[0:3], s32 offset:172 ; 4-byte Folded Spill
-; SI-NEXT: buffer_store_dword v13, off, s[0:3], s32 offset:176 ; 4-byte Folded Spill
-; SI-NEXT: buffer_store_dword v12, off, s[0:3], s32 offset:180 ; 4-byte Folded Spill
-; SI-NEXT: buffer_store_dword v11, off, s[0:3], s32 offset:184 ; 4-byte Folded Spill
-; SI-NEXT: buffer_store_dword v10, off, s[0:3], s32 offset:188 ; 4-byte Folded Spill
+; SI-NEXT: v_mov_b32_e32 v52, v30
+; SI-NEXT: v_mov_b32_e32 v44, v29
+; SI-NEXT: buffer_store_dword v28, off, s[0:3], s32 offset:80 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v27, off, s[0:3], s32 offset:88 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v26, off, s[0:3], s32 offset:96 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v25, off, s[0:3], s32 offset:104 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v24, off, s[0:3], s32 offset:112 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v23, off, s[0:3], s32 offset:116 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v22, off, s[0:3], s32 offset:124 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v21, off, s[0:3], s32 offset:132 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v20, off, s[0:3], s32 offset:140 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v19, off, s[0:3], s32 offset:148 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v18, off, s[0:3], s32 offset:156 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v17, off, s[0:3], s32 offset:164 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v16, off, s[0:3], s32 offset:172 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v15, off, s[0:3], s32 offset:176 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v14, off, s[0:3], s32 offset:180 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v13, off, s[0:3], s32 offset:184 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v12, off, s[0:3], s32 offset:188 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v11, off, s[0:3], s32 offset:192 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v10, off, s[0:3], s32 offset:84 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v9, off, s[0:3], s32 offset:100 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v8, off, s[0:3], s32 offset:108 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v7, off, s[0:3], s32 offset:120 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt vmcnt(14)
+; SI-NEXT: v_cvt_f32_f16_e32 v41, v1
+; SI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:212 ; 4-byte Folded Reload
+; SI-NEXT: s_waitcnt vmcnt(0)
; SI-NEXT: v_cvt_f32_f16_e32 v43, v1
; SI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:208 ; 4-byte Folded Reload
-; SI-NEXT: buffer_store_dword v9, off, s[0:3], s32 offset:92 ; 4-byte Folded Spill
-; SI-NEXT: s_waitcnt vmcnt(1)
+; SI-NEXT: s_waitcnt vmcnt(0)
; SI-NEXT: v_cvt_f32_f16_e32 v45, v1
; SI-NEXT: .LBB64_4: ; %end
; SI-NEXT: s_or_b64 exec, exec, s[4:5]
@@ -107814,63 +107845,59 @@ define <64 x half> @bitcast_v16i64_to_v64f16(<16 x i64> %a, i32 %b) {
; SI-NEXT: buffer_store_dword v1, v2, s[0:3], 0 offen
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_cvt_f16_f32_e32 v1, v47
-; SI-NEXT: v_cvt_f16_f32_e32 v2, v42
+; SI-NEXT: v_cvt_f16_f32_e32 v2, v46
; SI-NEXT: v_lshlrev_b32_e32 v1, 16, v1
; SI-NEXT: v_or_b32_e32 v1, v2, v1
; SI-NEXT: buffer_store_dword v1, v3, s[0:3], 0 offen
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_cvt_f16_f32_e32 v1, v45
-; SI-NEXT: v_cvt_f16_f32_e32 v2, v44
+; SI-NEXT: v_cvt_f16_f32_e32 v2, v51
; SI-NEXT: v_add_i32_e32 v3, vcc, 16, v0
; SI-NEXT: v_lshlrev_b32_e32 v1, 16, v1
; SI-NEXT: v_or_b32_e32 v1, v2, v1
; SI-NEXT: buffer_store_dword v1, v3, s[0:3], 0 offen
-; SI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:204 ; 4-byte Folded Reload
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_cvt_f16_f32_e32 v1, v43
+; SI-NEXT: v_cvt_f16_f32_e32 v2, v42
; SI-NEXT: v_add_i32_e32 v3, vcc, 20, v0
; SI-NEXT: v_lshlrev_b32_e32 v1, 16, v1
-; SI-NEXT: s_waitcnt vmcnt(0)
-; SI-NEXT: v_cvt_f16_f32_e32 v2, v2
; SI-NEXT: v_or_b32_e32 v1, v2, v1
; SI-NEXT: buffer_store_dword v1, v3, s[0:3], 0 offen
-; SI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:200 ; 4-byte Folded Reload
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_cvt_f16_f32_e32 v1, v41
+; SI-NEXT: v_cvt_f16_f32_e32 v2, v53
; SI-NEXT: v_add_i32_e32 v3, vcc, 24, v0
; SI-NEXT: v_lshlrev_b32_e32 v1, 16, v1
-; SI-NEXT: s_waitcnt vmcnt(0)
-; SI-NEXT: v_cvt_f16_f32_e32 v2, v2
; SI-NEXT: v_or_b32_e32 v1, v2, v1
; SI-NEXT: buffer_store_dword v1, v3, s[0:3], 0 offen
-; SI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:196 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:204 ; 4-byte Folded Reload
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_cvt_f16_f32_e32 v1, v55
+; SI-NEXT: v_cvt_f16_f32_e32 v1, v38
; SI-NEXT: v_add_i32_e32 v3, vcc, 28, v0
; SI-NEXT: v_lshlrev_b32_e32 v1, 16, v1
; SI-NEXT: s_waitcnt vmcnt(0)
; SI-NEXT: v_cvt_f16_f32_e32 v2, v2
; SI-NEXT: v_or_b32_e32 v1, v2, v1
; SI-NEXT: buffer_store_dword v1, v3, s[0:3], 0 offen
-; SI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:192 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:200 ; 4-byte Folded Reload
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_cvt_f16_f32_e32 v1, v53
+; SI-NEXT: v_cvt_f16_f32_e32 v1, v31
; SI-NEXT: v_add_i32_e32 v3, vcc, 32, v0
; SI-NEXT: v_lshlrev_b32_e32 v1, 16, v1
; SI-NEXT: s_waitcnt vmcnt(0)
; SI-NEXT: v_cvt_f16_f32_e32 v2, v2
; SI-NEXT: v_or_b32_e32 v1, v2, v1
; SI-NEXT: buffer_store_dword v1, v3, s[0:3], 0 offen
-; SI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:188 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:196 ; 4-byte Folded Reload
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_cvt_f16_f32_e32 v1, v51
+; SI-NEXT: v_cvt_f16_f32_e32 v1, v50
; SI-NEXT: v_add_i32_e32 v3, vcc, 36, v0
; SI-NEXT: v_lshlrev_b32_e32 v1, 16, v1
; SI-NEXT: s_waitcnt vmcnt(0)
; SI-NEXT: v_cvt_f16_f32_e32 v2, v2
; SI-NEXT: v_or_b32_e32 v1, v2, v1
; SI-NEXT: buffer_store_dword v1, v3, s[0:3], 0 offen
-; SI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:184 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:192 ; 4-byte Folded Reload
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_cvt_f16_f32_e32 v1, v49
; SI-NEXT: v_add_i32_e32 v3, vcc, 40, v0
@@ -107879,7 +107906,7 @@ define <64 x half> @bitcast_v16i64_to_v64f16(<16 x i64> %a, i32 %b) {
; SI-NEXT: v_cvt_f16_f32_e32 v2, v2
; SI-NEXT: v_or_b32_e32 v1, v2, v1
; SI-NEXT: buffer_store_dword v1, v3, s[0:3], 0 offen
-; SI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:180 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:188 ; 4-byte Folded Reload
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_cvt_f16_f32_e32 v1, v39
; SI-NEXT: v_add_i32_e32 v3, vcc, 44, v0
@@ -107888,7 +107915,7 @@ define <64 x half> @bitcast_v16i64_to_v64f16(<16 x i64> %a, i32 %b) {
; SI-NEXT: v_cvt_f16_f32_e32 v2, v2
; SI-NEXT: v_or_b32_e32 v1, v2, v1
; SI-NEXT: buffer_store_dword v1, v3, s[0:3], 0 offen
-; SI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:176 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:184 ; 4-byte Folded Reload
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_cvt_f16_f32_e32 v1, v37
; SI-NEXT: v_add_i32_e32 v3, vcc, 48, v0
@@ -107897,7 +107924,7 @@ define <64 x half> @bitcast_v16i64_to_v64f16(<16 x i64> %a, i32 %b) {
; SI-NEXT: v_cvt_f16_f32_e32 v2, v2
; SI-NEXT: v_or_b32_e32 v1, v2, v1
; SI-NEXT: buffer_store_dword v1, v3, s[0:3], 0 offen
-; SI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:172 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:180 ; 4-byte Folded Reload
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_cvt_f16_f32_e32 v1, v35
; SI-NEXT: v_add_i32_e32 v3, vcc, 52, v0
@@ -107906,7 +107933,7 @@ define <64 x half> @bitcast_v16i64_to_v64f16(<16 x i64> %a, i32 %b) {
; SI-NEXT: v_cvt_f16_f32_e32 v2, v2
; SI-NEXT: v_or_b32_e32 v1, v2, v1
; SI-NEXT: buffer_store_dword v1, v3, s[0:3], 0 offen
-; SI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:168 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:176 ; 4-byte Folded Reload
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_cvt_f16_f32_e32 v1, v33
; SI-NEXT: v_add_i32_e32 v3, vcc, 56, v0
@@ -107916,9 +107943,20 @@ define <64 x half> @bitcast_v16i64_to_v64f16(<16 x i64> %a, i32 %b) {
; SI-NEXT: v_or_b32_e32 v1, v2, v1
; SI-NEXT: buffer_store_dword v1, v3, s[0:3], 0 offen
; SI-NEXT: s_waitcnt expcnt(0)
+; SI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:168 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:172 ; 4-byte Folded Reload
+; SI-NEXT: v_add_i32_e32 v3, vcc, 60, v0
+; SI-NEXT: s_waitcnt vmcnt(1)
+; SI-NEXT: v_cvt_f16_f32_e32 v1, v1
+; SI-NEXT: s_waitcnt vmcnt(0)
+; SI-NEXT: v_cvt_f16_f32_e32 v2, v2
+; SI-NEXT: v_lshlrev_b32_e32 v1, 16, v1
+; SI-NEXT: v_or_b32_e32 v1, v2, v1
+; SI-NEXT: buffer_store_dword v1, v3, s[0:3], 0 offen
+; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:160 ; 4-byte Folded Reload
; SI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:164 ; 4-byte Folded Reload
-; SI-NEXT: v_add_i32_e32 v3, vcc, 60, v0
+; SI-NEXT: v_add_i32_e32 v3, vcc, 64, v0
; SI-NEXT: s_waitcnt vmcnt(1)
; SI-NEXT: v_cvt_f16_f32_e32 v1, v1
; SI-NEXT: s_waitcnt vmcnt(0)
@@ -107929,7 +107967,7 @@ define <64 x half> @bitcast_v16i64_to_v64f16(<16 x i64> %a, i32 %b) {
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:152 ; 4-byte Folded Reload
; SI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:156 ; 4-byte Folded Reload
-; SI-NEXT: v_add_i32_e32 v3, vcc, 64, v0
+; SI-NEXT: v_add_i32_e32 v3, vcc, 0x44, v0
; SI-NEXT: s_waitcnt vmcnt(1)
; SI-NEXT: v_cvt_f16_f32_e32 v1, v1
; SI-NEXT: s_waitcnt vmcnt(0)
@@ -107940,7 +107978,7 @@ define <64 x half> @bitcast_v16i64_to_v64f16(<16 x i64> %a, i32 %b) {
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:144 ; 4-byte Folded Reload
; SI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:148 ; 4-byte Folded Reload
-; SI-NEXT: v_add_i32_e32 v3, vcc, 0x44, v0
+; SI-NEXT: v_add_i32_e32 v3, vcc, 0x48, v0
; SI-NEXT: s_waitcnt vmcnt(1)
; SI-NEXT: v_cvt_f16_f32_e32 v1, v1
; SI-NEXT: s_waitcnt vmcnt(0)
@@ -107951,7 +107989,7 @@ define <64 x half> @bitcast_v16i64_to_v64f16(<16 x i64> %a, i32 %b) {
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:136 ; 4-byte Folded Reload
; SI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:140 ; 4-byte Folded Reload
-; SI-NEXT: v_add_i32_e32 v3, vcc, 0x48, v0
+; SI-NEXT: v_add_i32_e32 v3, vcc, 0x4c, v0
; SI-NEXT: s_waitcnt vmcnt(1)
; SI-NEXT: v_cvt_f16_f32_e32 v1, v1
; SI-NEXT: s_waitcnt vmcnt(0)
@@ -107962,7 +108000,7 @@ define <64 x half> @bitcast_v16i64_to_v64f16(<16 x i64> %a, i32 %b) {
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:128 ; 4-byte Folded Reload
; SI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:132 ; 4-byte Folded Reload
-; SI-NEXT: v_add_i32_e32 v3, vcc, 0x4c, v0
+; SI-NEXT: v_add_i32_e32 v3, vcc, 0x50, v0
; SI-NEXT: s_waitcnt vmcnt(1)
; SI-NEXT: v_cvt_f16_f32_e32 v1, v1
; SI-NEXT: s_waitcnt vmcnt(0)
@@ -107973,7 +108011,7 @@ define <64 x half> @bitcast_v16i64_to_v64f16(<16 x i64> %a, i32 %b) {
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:120 ; 4-byte Folded Reload
; SI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:124 ; 4-byte Folded Reload
-; SI-NEXT: v_add_i32_e32 v3, vcc, 0x50, v0
+; SI-NEXT: v_add_i32_e32 v3, vcc, 0x54, v0
; SI-NEXT: s_waitcnt vmcnt(1)
; SI-NEXT: v_cvt_f16_f32_e32 v1, v1
; SI-NEXT: s_waitcnt vmcnt(0)
@@ -107982,9 +108020,9 @@ define <64 x half> @bitcast_v16i64_to_v64f16(<16 x i64> %a, i32 %b) {
; SI-NEXT: v_or_b32_e32 v1, v2, v1
; SI-NEXT: buffer_store_dword v1, v3, s[0:3], 0 offen
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:112 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:108 ; 4-byte Folded Reload
; SI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:116 ; 4-byte Folded Reload
-; SI-NEXT: v_add_i32_e32 v3, vcc, 0x54, v0
+; SI-NEXT: v_add_i32_e32 v3, vcc, 0x58, v0
; SI-NEXT: s_waitcnt vmcnt(1)
; SI-NEXT: v_cvt_f16_f32_e32 v1, v1
; SI-NEXT: s_waitcnt vmcnt(0)
@@ -107994,8 +108032,8 @@ define <64 x half> @bitcast_v16i64_to_v64f16(<16 x i64> %a, i32 %b) {
; SI-NEXT: buffer_store_dword v1, v3, s[0:3], 0 offen
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:100 ; 4-byte Folded Reload
-; SI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:108 ; 4-byte Folded Reload
-; SI-NEXT: v_add_i32_e32 v3, vcc, 0x58, v0
+; SI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:112 ; 4-byte Folded Reload
+; SI-NEXT: v_add_i32_e32 v3, vcc, 0x5c, v0
; SI-NEXT: s_waitcnt vmcnt(1)
; SI-NEXT: v_cvt_f16_f32_e32 v1, v1
; SI-NEXT: s_waitcnt vmcnt(0)
@@ -108006,7 +108044,7 @@ define <64 x half> @bitcast_v16i64_to_v64f16(<16 x i64> %a, i32 %b) {
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:92 ; 4-byte Folded Reload
; SI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:104 ; 4-byte Folded Reload
-; SI-NEXT: v_add_i32_e32 v3, vcc, 0x5c, v0
+; SI-NEXT: v_add_i32_e32 v3, vcc, 0x60, v0
; SI-NEXT: s_waitcnt vmcnt(1)
; SI-NEXT: v_cvt_f16_f32_e32 v1, v1
; SI-NEXT: s_waitcnt vmcnt(0)
@@ -108017,7 +108055,7 @@ define <64 x half> @bitcast_v16i64_to_v64f16(<16 x i64> %a, i32 %b) {
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:84 ; 4-byte Folded Reload
; SI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:96 ; 4-byte Folded Reload
-; SI-NEXT: v_add_i32_e32 v3, vcc, 0x60, v0
+; SI-NEXT: v_add_i32_e32 v3, vcc, 0x64, v0
; SI-NEXT: s_waitcnt vmcnt(1)
; SI-NEXT: v_cvt_f16_f32_e32 v1, v1
; SI-NEXT: s_waitcnt vmcnt(0)
@@ -108028,7 +108066,7 @@ define <64 x half> @bitcast_v16i64_to_v64f16(<16 x i64> %a, i32 %b) {
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:76 ; 4-byte Folded Reload
; SI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:88 ; 4-byte Folded Reload
-; SI-NEXT: v_add_i32_e32 v3, vcc, 0x64, v0
+; SI-NEXT: v_add_i32_e32 v3, vcc, 0x68, v0
; SI-NEXT: s_waitcnt vmcnt(1)
; SI-NEXT: v_cvt_f16_f32_e32 v1, v1
; SI-NEXT: s_waitcnt vmcnt(0)
@@ -108038,8 +108076,8 @@ define <64 x half> @bitcast_v16i64_to_v64f16(<16 x i64> %a, i32 %b) {
; SI-NEXT: buffer_store_dword v1, v3, s[0:3], 0 offen
; SI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:80 ; 4-byte Folded Reload
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_cvt_f16_f32_e32 v1, v52
-; SI-NEXT: v_add_i32_e32 v3, vcc, 0x68, v0
+; SI-NEXT: v_cvt_f16_f32_e32 v1, v55
+; SI-NEXT: v_add_i32_e32 v3, vcc, 0x6c, v0
; SI-NEXT: v_lshlrev_b32_e32 v1, 16, v1
; SI-NEXT: s_waitcnt vmcnt(0)
; SI-NEXT: v_cvt_f16_f32_e32 v2, v2
@@ -108047,35 +108085,28 @@ define <64 x half> @bitcast_v16i64_to_v64f16(<16 x i64> %a, i32 %b) {
; SI-NEXT: buffer_store_dword v1, v3, s[0:3], 0 offen
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_cvt_f16_f32_e32 v1, v54
-; SI-NEXT: v_cvt_f16_f32_e32 v2, v46
-; SI-NEXT: v_add_i32_e32 v3, vcc, 0x6c, v0
-; SI-NEXT: v_lshlrev_b32_e32 v1, 16, v1
-; SI-NEXT: v_or_b32_e32 v1, v2, v1
-; SI-NEXT: buffer_store_dword v1, v3, s[0:3], 0 offen
-; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_cvt_f16_f32_e32 v1, v38
-; SI-NEXT: v_cvt_f16_f32_e32 v2, v50
+; SI-NEXT: v_cvt_f16_f32_e32 v2, v44
; SI-NEXT: v_add_i32_e32 v3, vcc, 0x70, v0
; SI-NEXT: v_lshlrev_b32_e32 v1, 16, v1
; SI-NEXT: v_or_b32_e32 v1, v2, v1
; SI-NEXT: buffer_store_dword v1, v3, s[0:3], 0 offen
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_cvt_f16_f32_e32 v1, v40
-; SI-NEXT: v_cvt_f16_f32_e32 v2, v48
+; SI-NEXT: v_cvt_f16_f32_e32 v2, v52
; SI-NEXT: v_add_i32_e32 v3, vcc, 0x74, v0
; SI-NEXT: v_lshlrev_b32_e32 v1, 16, v1
; SI-NEXT: v_or_b32_e32 v1, v2, v1
; SI-NEXT: buffer_store_dword v1, v3, s[0:3], 0 offen
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_cvt_f16_f32_e32 v1, v34
-; SI-NEXT: v_cvt_f16_f32_e32 v2, v36
+; SI-NEXT: v_cvt_f16_f32_e32 v1, v36
+; SI-NEXT: v_cvt_f16_f32_e32 v2, v48
; SI-NEXT: v_add_i32_e32 v3, vcc, 0x78, v0
; SI-NEXT: v_lshlrev_b32_e32 v1, 16, v1
; SI-NEXT: v_or_b32_e32 v1, v2, v1
; SI-NEXT: buffer_store_dword v1, v3, s[0:3], 0 offen
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_cvt_f16_f32_e32 v1, v32
-; SI-NEXT: v_cvt_f16_f32_e32 v2, v31
+; SI-NEXT: v_cvt_f16_f32_e32 v1, v34
+; SI-NEXT: v_cvt_f16_f32_e32 v2, v32
; SI-NEXT: v_add_i32_e32 v0, vcc, 0x7c, v0
; SI-NEXT: v_lshlrev_b32_e32 v1, 16, v1
; SI-NEXT: v_or_b32_e32 v1, v2, v1
@@ -113327,24 +113358,24 @@ define inreg <16 x i64> @bitcast_v64i16_to_v16i64_scalar(<64 x i16> inreg %a, i3
; SI-NEXT: buffer_load_dword v10, off, s[0:3], s32 offset:32
; SI-NEXT: buffer_load_dword v46, off, s[0:3], s32 offset:28
; SI-NEXT: buffer_load_dword v12, off, s[0:3], s32 offset:40
-; SI-NEXT: buffer_load_dword v45, off, s[0:3], s32 offset:36
-; SI-NEXT: buffer_load_dword v38, off, s[0:3], s32 offset:48
+; SI-NEXT: buffer_load_dword v59, off, s[0:3], s32 offset:36
+; SI-NEXT: buffer_load_dword v48, off, s[0:3], s32 offset:48
; SI-NEXT: buffer_load_dword v44, off, s[0:3], s32 offset:44
; SI-NEXT: buffer_load_dword v36, off, s[0:3], s32 offset:56
-; SI-NEXT: buffer_load_dword v59, off, s[0:3], s32 offset:52
+; SI-NEXT: buffer_load_dword v43, off, s[0:3], s32 offset:52
; SI-NEXT: buffer_load_dword v34, off, s[0:3], s32 offset:64
; SI-NEXT: buffer_load_dword v42, off, s[0:3], s32 offset:60
; SI-NEXT: buffer_load_dword v32, off, s[0:3], s32 offset:72
-; SI-NEXT: buffer_load_dword v58, off, s[0:3], s32 offset:68
-; SI-NEXT: s_waitcnt expcnt(2)
-; SI-NEXT: v_lshlrev_b32_e32 v61, 16, v1
+; SI-NEXT: buffer_load_dword v38, off, s[0:3], s32 offset:68
+; SI-NEXT: v_lshlrev_b32_e32 v58, 16, v1
; SI-NEXT: v_lshlrev_b32_e32 v55, 16, v3
-; SI-NEXT: v_lshlrev_b32_e32 v50, 16, v5
-; SI-NEXT: v_lshlrev_b32_e32 v43, 16, v7
-; SI-NEXT: v_lshlrev_b32_e32 v41, 16, v9
-; SI-NEXT: v_lshlrev_b32_e32 v40, 16, v11
+; SI-NEXT: s_waitcnt expcnt(2)
+; SI-NEXT: v_lshlrev_b32_e32 v61, 16, v5
+; SI-NEXT: v_lshlrev_b32_e32 v50, 16, v7
+; SI-NEXT: v_lshlrev_b32_e32 v45, 16, v9
+; SI-NEXT: v_lshlrev_b32_e32 v41, 16, v11
; SI-NEXT: v_lshlrev_b32_e32 v13, 16, v13
-; SI-NEXT: v_lshlrev_b32_e32 v48, 16, v15
+; SI-NEXT: v_lshlrev_b32_e32 v40, 16, v15
; SI-NEXT: v_lshlrev_b32_e32 v15, 16, v17
; SI-NEXT: v_lshlrev_b32_e32 v37, 16, v19
; SI-NEXT: v_lshlrev_b32_e32 v17, 16, v21
@@ -113365,23 +113396,23 @@ define inreg <16 x i64> @bitcast_v64i16_to_v16i64_scalar(<64 x i16> inreg %a, i3
; SI-NEXT: s_waitcnt vmcnt(9)
; SI-NEXT: v_lshlrev_b32_e32 v54, 16, v12
; SI-NEXT: s_waitcnt vmcnt(7)
-; SI-NEXT: v_lshlrev_b32_e32 v5, 16, v38
+; SI-NEXT: v_lshlrev_b32_e32 v5, 16, v48
; SI-NEXT: s_waitcnt vmcnt(5) expcnt(1)
; SI-NEXT: v_lshlrev_b32_e32 v62, 16, v36
; SI-NEXT: s_waitcnt vmcnt(3)
; SI-NEXT: v_lshlrev_b32_e32 v3, 16, v34
; SI-NEXT: s_waitcnt vmcnt(1)
-; SI-NEXT: v_lshlrev_b32_e32 v34, 16, v32
-; SI-NEXT: buffer_store_dword v34, off, s[0:3], s32 offset:144 ; 4-byte Folded Spill
+; SI-NEXT: v_lshlrev_b32_e32 v48, 16, v32
+; SI-NEXT: buffer_store_dword v48, off, s[0:3], s32 offset:144 ; 4-byte Folded Spill
; SI-NEXT: buffer_store_dword v3, off, s[0:3], s32 offset:148 ; 4-byte Folded Spill
; SI-NEXT: buffer_store_dword v62, off, s[0:3], s32 offset:152 ; 4-byte Folded Spill
; SI-NEXT: buffer_store_dword v5, off, s[0:3], s32 offset:156 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt vmcnt(4)
-; SI-NEXT: buffer_store_dword v58, off, s[0:3], s32 offset:160 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v38, off, s[0:3], s32 offset:160 ; 4-byte Folded Spill
; SI-NEXT: buffer_store_dword v54, off, s[0:3], s32 offset:164 ; 4-byte Folded Spill
-; SI-NEXT: buffer_store_dword v27, off, s[0:3], s32 offset:168 ; 4-byte Folded Spill
-; SI-NEXT: buffer_store_dword v59, off, s[0:3], s32 offset:172 ; 4-byte Folded Spill
-; SI-NEXT: buffer_store_dword v45, off, s[0:3], s32 offset:176 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v43, off, s[0:3], s32 offset:168 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v27, off, s[0:3], s32 offset:172 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v59, off, s[0:3], s32 offset:176 ; 4-byte Folded Spill
; SI-NEXT: buffer_store_dword v25, off, s[0:3], s32 offset:180 ; 4-byte Folded Spill
; SI-NEXT: buffer_store_dword v47, off, s[0:3], s32 offset:184 ; 4-byte Folded Spill
; SI-NEXT: buffer_store_dword v60, off, s[0:3], s32 offset:188 ; 4-byte Folded Spill
@@ -113394,45 +113425,46 @@ define inreg <16 x i64> @bitcast_v64i16_to_v16i64_scalar(<64 x i16> inreg %a, i3
; SI-NEXT: buffer_store_dword v19, off, s[0:3], s32 offset:216 ; 4-byte Folded Spill
; SI-NEXT: buffer_store_dword v35, off, s[0:3], s32 offset:220 ; 4-byte Folded Spill
; SI-NEXT: buffer_store_dword v22, off, s[0:3], s32 offset:224 ; 4-byte Folded Spill
-; SI-NEXT: buffer_store_dword v17, off, s[0:3], s32 offset:228 ; 4-byte Folded Spill
-; SI-NEXT: buffer_store_dword v0, off, s[0:3], s32 offset:232 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v0, off, s[0:3], s32 offset:228 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v17, off, s[0:3], s32 offset:232 ; 4-byte Folded Spill
; SI-NEXT: buffer_store_dword v37, off, s[0:3], s32 offset:236 ; 4-byte Folded Spill
; SI-NEXT: buffer_store_dword v18, off, s[0:3], s32 offset:240 ; 4-byte Folded Spill
; SI-NEXT: s_cbranch_scc0 .LBB71_4
; SI-NEXT: ; %bb.1: ; %cmp.false
-; SI-NEXT: s_waitcnt expcnt(2)
+; SI-NEXT: s_waitcnt expcnt(3)
; SI-NEXT: v_and_b32_e32 v0, 0xffff, v0
-; SI-NEXT: v_or_b32_e32 v7, v0, v61
+; SI-NEXT: v_or_b32_e32 v7, v0, v58
; SI-NEXT: v_and_b32_e32 v0, 0xffff, v4
-; SI-NEXT: v_or_b32_e32 v9, v0, v50
+; SI-NEXT: v_or_b32_e32 v9, v0, v61
; SI-NEXT: v_and_b32_e32 v0, 0xffff, v6
-; SI-NEXT: v_or_b32_e32 v10, v0, v43
+; SI-NEXT: v_or_b32_e32 v10, v0, v50
; SI-NEXT: v_and_b32_e32 v0, 0xffff, v57
-; SI-NEXT: v_or_b32_e32 v11, v0, v41
+; SI-NEXT: v_or_b32_e32 v11, v0, v45
; SI-NEXT: v_and_b32_e32 v0, 0xffff, v56
-; SI-NEXT: v_or_b32_e32 v12, v0, v40
+; SI-NEXT: v_or_b32_e32 v12, v0, v41
; SI-NEXT: v_and_b32_e32 v0, 0xffff, v49
-; SI-NEXT: v_mov_b32_e32 v52, v57
-; SI-NEXT: v_mov_b32_e32 v57, v40
-; SI-NEXT: v_mov_b32_e32 v40, v49
-; SI-NEXT: v_mov_b32_e32 v49, v13
+; SI-NEXT: v_mov_b32_e32 v36, v41
+; SI-NEXT: v_mov_b32_e32 v41, v13
; SI-NEXT: v_or_b32_e32 v13, v0, v13
; SI-NEXT: v_and_b32_e32 v0, 0xffff, v14
-; SI-NEXT: v_mov_b32_e32 v36, v41
-; SI-NEXT: v_mov_b32_e32 v41, v14
-; SI-NEXT: v_or_b32_e32 v14, v0, v48
-; SI-NEXT: v_and_b32_e32 v0, 0xffff, v16
; SI-NEXT: v_mov_b32_e32 v51, v50
-; SI-NEXT: v_mov_b32_e32 v50, v43
-; SI-NEXT: v_mov_b32_e32 v43, v48
-; SI-NEXT: v_mov_b32_e32 v48, v15
+; SI-NEXT: v_mov_b32_e32 v50, v45
+; SI-NEXT: v_mov_b32_e32 v45, v14
+; SI-NEXT: v_or_b32_e32 v14, v0, v40
+; SI-NEXT: v_and_b32_e32 v0, 0xffff, v16
+; SI-NEXT: v_mov_b32_e32 v52, v57
+; SI-NEXT: v_mov_b32_e32 v57, v49
+; SI-NEXT: v_mov_b32_e32 v49, v40
+; SI-NEXT: v_mov_b32_e32 v40, v15
; SI-NEXT: v_or_b32_e32 v15, v0, v15
; SI-NEXT: v_and_b32_e32 v0, 0xffff, v18
-; SI-NEXT: v_mov_b32_e32 v38, v61
+; SI-NEXT: v_mov_b32_e32 v34, v58
+; SI-NEXT: v_mov_b32_e32 v58, v61
; SI-NEXT: v_mov_b32_e32 v61, v56
; SI-NEXT: v_mov_b32_e32 v56, v16
; SI-NEXT: v_or_b32_e32 v16, v0, v37
; SI-NEXT: v_and_b32_e32 v0, 0xffff, v20
+; SI-NEXT: s_waitcnt expcnt(2)
; SI-NEXT: v_or_b32_e32 v17, v0, v17
; SI-NEXT: v_and_b32_e32 v0, 0xffff, v22
; SI-NEXT: s_waitcnt expcnt(0)
@@ -113466,7 +113498,7 @@ define inreg <16 x i64> @bitcast_v64i16_to_v16i64_scalar(<64 x i16> inreg %a, i3
; SI-NEXT: s_and_b32 s6, s20, 0xffff
; SI-NEXT: s_lshl_b32 s7, s21, 16
; SI-NEXT: v_or_b32_e32 v26, v0, v26
-; SI-NEXT: v_and_b32_e32 v0, 0xffff, v45
+; SI-NEXT: v_and_b32_e32 v0, 0xffff, v59
; SI-NEXT: s_or_b32 s6, s6, s7
; SI-NEXT: s_and_b32 s7, s22, 0xffff
; SI-NEXT: s_lshl_b32 s8, s23, 16
@@ -113477,7 +113509,7 @@ define inreg <16 x i64> @bitcast_v64i16_to_v16i64_scalar(<64 x i16> inreg %a, i3
; SI-NEXT: s_lshl_b32 s9, s25, 16
; SI-NEXT: v_mov_b32_e32 v33, v28
; SI-NEXT: v_or_b32_e32 v28, v0, v5
-; SI-NEXT: v_and_b32_e32 v0, 0xffff, v59
+; SI-NEXT: v_and_b32_e32 v0, 0xffff, v43
; SI-NEXT: s_or_b32 s8, s8, s9
; SI-NEXT: s_and_b32 s9, s26, 0xffff
; SI-NEXT: s_lshl_b32 s10, s27, 16
@@ -113489,7 +113521,7 @@ define inreg <16 x i64> @bitcast_v64i16_to_v16i64_scalar(<64 x i16> inreg %a, i3
; SI-NEXT: s_lshl_b32 s11, s29, 16
; SI-NEXT: v_and_b32_e32 v1, 0xffff, v2
; SI-NEXT: v_or_b32_e32 v30, v0, v3
-; SI-NEXT: v_and_b32_e32 v0, 0xffff, v58
+; SI-NEXT: v_and_b32_e32 v0, 0xffff, v38
; SI-NEXT: s_or_b32 s10, s10, s11
; SI-NEXT: v_mov_b32_e32 v63, v2
; SI-NEXT: v_mov_b32_e32 v32, v55
@@ -113497,9 +113529,9 @@ define inreg <16 x i64> @bitcast_v64i16_to_v16i64_scalar(<64 x i16> inreg %a, i3
; SI-NEXT: v_mov_b32_e32 v55, v4
; SI-NEXT: v_mov_b32_e32 v53, v6
; SI-NEXT: v_mov_b32_e32 v47, v46
-; SI-NEXT: v_mov_b32_e32 v45, v44
-; SI-NEXT: v_mov_b32_e32 v59, v42
-; SI-NEXT: v_or_b32_e32 v31, v0, v34
+; SI-NEXT: v_mov_b32_e32 v59, v44
+; SI-NEXT: v_mov_b32_e32 v43, v42
+; SI-NEXT: v_or_b32_e32 v31, v0, v48
; SI-NEXT: v_mov_b32_e32 v0, s4
; SI-NEXT: v_mov_b32_e32 v1, s5
; SI-NEXT: v_mov_b32_e32 v2, s6
@@ -113509,12 +113541,13 @@ define inreg <16 x i64> @bitcast_v64i16_to_v16i64_scalar(<64 x i16> inreg %a, i3
; SI-NEXT: v_mov_b32_e32 v6, s10
; SI-NEXT: s_cbranch_execnz .LBB71_3
; SI-NEXT: .LBB71_2: ; %cmp.true
-; SI-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:232 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:228 ; 4-byte Folded Reload
; SI-NEXT: v_add_i32_e32 v1, vcc, 3, v63
; SI-NEXT: v_and_b32_e32 v1, 0xffff, v1
; SI-NEXT: v_or_b32_e32 v1, v32, v1
; SI-NEXT: v_add_i32_e32 v8, vcc, 0x30000, v1
; SI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:236 ; 4-byte Folded Reload
+; SI-NEXT: v_mov_b32_e32 v38, v43
; SI-NEXT: s_add_i32 s16, s16, 3
; SI-NEXT: s_and_b32 s4, s16, 0xffff
; SI-NEXT: s_lshl_b32 s5, s17, 16
@@ -113558,42 +113591,42 @@ define inreg <16 x i64> @bitcast_v64i16_to_v16i64_scalar(<64 x i16> inreg %a, i3
; SI-NEXT: s_waitcnt vmcnt(1)
; SI-NEXT: v_add_i32_e32 v0, vcc, 3, v0
; SI-NEXT: v_and_b32_e32 v0, 0xffff, v0
-; SI-NEXT: v_or_b32_e32 v0, v38, v0
+; SI-NEXT: v_or_b32_e32 v0, v34, v0
; SI-NEXT: v_add_i32_e32 v7, vcc, 0x30000, v0
; SI-NEXT: v_add_i32_e32 v0, vcc, 3, v55
; SI-NEXT: v_and_b32_e32 v0, 0xffff, v0
-; SI-NEXT: v_or_b32_e32 v0, v51, v0
+; SI-NEXT: v_or_b32_e32 v0, v58, v0
; SI-NEXT: v_add_i32_e32 v9, vcc, 0x30000, v0
; SI-NEXT: v_add_i32_e32 v0, vcc, 3, v53
; SI-NEXT: v_and_b32_e32 v0, 0xffff, v0
-; SI-NEXT: v_or_b32_e32 v0, v50, v0
+; SI-NEXT: v_or_b32_e32 v0, v51, v0
; SI-NEXT: v_add_i32_e32 v10, vcc, 0x30000, v0
; SI-NEXT: v_add_i32_e32 v0, vcc, 3, v52
; SI-NEXT: v_and_b32_e32 v0, 0xffff, v0
-; SI-NEXT: v_or_b32_e32 v0, v36, v0
+; SI-NEXT: v_or_b32_e32 v0, v50, v0
; SI-NEXT: v_add_i32_e32 v11, vcc, 0x30000, v0
; SI-NEXT: v_add_i32_e32 v0, vcc, 3, v61
; SI-NEXT: v_and_b32_e32 v0, 0xffff, v0
-; SI-NEXT: v_or_b32_e32 v0, v57, v0
+; SI-NEXT: v_or_b32_e32 v0, v36, v0
; SI-NEXT: v_add_i32_e32 v12, vcc, 0x30000, v0
-; SI-NEXT: v_add_i32_e32 v0, vcc, 3, v40
+; SI-NEXT: v_add_i32_e32 v0, vcc, 3, v57
; SI-NEXT: v_and_b32_e32 v0, 0xffff, v0
-; SI-NEXT: v_or_b32_e32 v0, v49, v0
+; SI-NEXT: v_or_b32_e32 v0, v41, v0
; SI-NEXT: v_add_i32_e32 v13, vcc, 0x30000, v0
-; SI-NEXT: v_add_i32_e32 v0, vcc, 3, v41
+; SI-NEXT: v_add_i32_e32 v0, vcc, 3, v45
; SI-NEXT: v_and_b32_e32 v0, 0xffff, v0
-; SI-NEXT: v_or_b32_e32 v0, v43, v0
+; SI-NEXT: v_or_b32_e32 v0, v49, v0
; SI-NEXT: v_add_i32_e32 v14, vcc, 0x30000, v0
; SI-NEXT: v_add_i32_e32 v0, vcc, 3, v56
; SI-NEXT: v_and_b32_e32 v0, 0xffff, v0
-; SI-NEXT: v_or_b32_e32 v0, v48, v0
+; SI-NEXT: v_or_b32_e32 v0, v40, v0
; SI-NEXT: v_add_i32_e32 v15, vcc, 0x30000, v0
; SI-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:240 ; 4-byte Folded Reload
; SI-NEXT: s_waitcnt vmcnt(0)
; SI-NEXT: v_add_i32_e32 v0, vcc, 3, v0
; SI-NEXT: v_and_b32_e32 v0, 0xffff, v0
; SI-NEXT: v_or_b32_e32 v0, v1, v0
-; SI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:228 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:232 ; 4-byte Folded Reload
; SI-NEXT: v_add_i32_e32 v16, vcc, 0x30000, v0
; SI-NEXT: v_add_i32_e32 v0, vcc, 3, v37
; SI-NEXT: v_and_b32_e32 v0, 0xffff, v0
@@ -113655,7 +113688,7 @@ define inreg <16 x i64> @bitcast_v64i16_to_v16i64_scalar(<64 x i16> inreg %a, i3
; SI-NEXT: v_add_i32_e32 v0, vcc, 3, v0
; SI-NEXT: v_and_b32_e32 v0, 0xffff, v0
; SI-NEXT: v_or_b32_e32 v0, v1, v0
-; SI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:168 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:172 ; 4-byte Folded Reload
; SI-NEXT: v_add_i32_e32 v25, vcc, 0x30000, v0
; SI-NEXT: v_add_i32_e32 v0, vcc, 3, v47
; SI-NEXT: v_and_b32_e32 v0, 0xffff, v0
@@ -113671,12 +113704,12 @@ define inreg <16 x i64> @bitcast_v64i16_to_v16i64_scalar(<64 x i16> inreg %a, i3
; SI-NEXT: v_or_b32_e32 v0, v1, v0
; SI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:156 ; 4-byte Folded Reload
; SI-NEXT: v_add_i32_e32 v27, vcc, 0x30000, v0
-; SI-NEXT: v_add_i32_e32 v0, vcc, 3, v45
+; SI-NEXT: v_add_i32_e32 v0, vcc, 3, v59
; SI-NEXT: v_and_b32_e32 v0, 0xffff, v0
; SI-NEXT: s_waitcnt vmcnt(0)
; SI-NEXT: v_or_b32_e32 v0, v1, v0
; SI-NEXT: v_add_i32_e32 v28, vcc, 0x30000, v0
-; SI-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:172 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:168 ; 4-byte Folded Reload
; SI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:152 ; 4-byte Folded Reload
; SI-NEXT: s_waitcnt vmcnt(1)
; SI-NEXT: v_add_i32_e32 v0, vcc, 3, v0
@@ -113685,7 +113718,7 @@ define inreg <16 x i64> @bitcast_v64i16_to_v16i64_scalar(<64 x i16> inreg %a, i3
; SI-NEXT: v_or_b32_e32 v0, v1, v0
; SI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:148 ; 4-byte Folded Reload
; SI-NEXT: v_add_i32_e32 v29, vcc, 0x30000, v0
-; SI-NEXT: v_add_i32_e32 v0, vcc, 3, v59
+; SI-NEXT: v_add_i32_e32 v0, vcc, 3, v38
; SI-NEXT: v_and_b32_e32 v0, 0xffff, v0
; SI-NEXT: s_waitcnt vmcnt(0)
; SI-NEXT: v_or_b32_e32 v0, v1, v0
@@ -113720,26 +113753,26 @@ define inreg <16 x i64> @bitcast_v64i16_to_v16i64_scalar(<64 x i16> inreg %a, i3
; SI-NEXT: s_waitcnt vmcnt(0) expcnt(0)
; SI-NEXT: s_setpc_b64 s[30:31]
; SI-NEXT: .LBB71_4:
-; SI-NEXT: v_mov_b32_e32 v38, v61
+; SI-NEXT: v_mov_b32_e32 v34, v58
; SI-NEXT: v_mov_b32_e32 v32, v55
+; SI-NEXT: v_mov_b32_e32 v58, v61
; SI-NEXT: v_mov_b32_e32 v63, v2
; SI-NEXT: v_mov_b32_e32 v55, v4
; SI-NEXT: v_mov_b32_e32 v53, v6
; SI-NEXT: v_mov_b32_e32 v52, v57
; SI-NEXT: v_mov_b32_e32 v51, v50
; SI-NEXT: v_mov_b32_e32 v61, v56
-; SI-NEXT: v_mov_b32_e32 v50, v43
+; SI-NEXT: v_mov_b32_e32 v50, v45
; SI-NEXT: v_mov_b32_e32 v36, v41
-; SI-NEXT: v_mov_b32_e32 v57, v40
-; SI-NEXT: v_mov_b32_e32 v40, v49
-; SI-NEXT: v_mov_b32_e32 v49, v13
-; SI-NEXT: v_mov_b32_e32 v43, v48
-; SI-NEXT: v_mov_b32_e32 v48, v15
-; SI-NEXT: v_mov_b32_e32 v41, v14
+; SI-NEXT: v_mov_b32_e32 v41, v13
+; SI-NEXT: v_mov_b32_e32 v57, v49
+; SI-NEXT: v_mov_b32_e32 v49, v40
+; SI-NEXT: v_mov_b32_e32 v40, v15
+; SI-NEXT: v_mov_b32_e32 v45, v14
; SI-NEXT: v_mov_b32_e32 v56, v16
; SI-NEXT: v_mov_b32_e32 v47, v46
-; SI-NEXT: v_mov_b32_e32 v45, v44
-; SI-NEXT: v_mov_b32_e32 v59, v42
+; SI-NEXT: v_mov_b32_e32 v59, v44
+; SI-NEXT: v_mov_b32_e32 v43, v42
; SI-NEXT: s_waitcnt expcnt(1)
; SI-NEXT: v_mov_b32_e32 v37, v20
; SI-NEXT: v_mov_b32_e32 v39, v23
@@ -120323,7 +120356,7 @@ define inreg <128 x i8> @bitcast_v16f64_to_v128i8_scalar(<16 x double> inreg %a,
; VI-NEXT: v_lshrrev_b32_e32 v34, 16, v29
; VI-NEXT: buffer_store_dword v33, off, s[0:3], s32 offset:56 ; 4-byte Folded Spill
; VI-NEXT: v_lshrrev_b32_e32 v33, 8, v11
-; VI-NEXT: buffer_store_dword v34, off, s[0:3], s32 offset:372 ; 4-byte Folded Spill
+; VI-NEXT: buffer_store_dword v34, off, s[0:3], s32 offset:376 ; 4-byte Folded Spill
; VI-NEXT: v_lshrrev_b32_e32 v34, 8, v29
; VI-NEXT: buffer_store_dword v33, off, s[0:3], s32 offset:60 ; 4-byte Folded Spill
; VI-NEXT: v_lshrrev_b32_e32 v33, 8, v1
@@ -120341,7 +120374,7 @@ define inreg <128 x i8> @bitcast_v16f64_to_v128i8_scalar(<16 x double> inreg %a,
; VI-NEXT: v_lshrrev_b32_e32 v39, 24, v30
; VI-NEXT: v_lshrrev_b32_e32 v49, 16, v30
; VI-NEXT: v_lshrrev_b32_e32 v61, 8, v30
-; VI-NEXT: buffer_store_dword v34, off, s[0:3], s32 offset:376 ; 4-byte Folded Spill
+; VI-NEXT: buffer_store_dword v34, off, s[0:3], s32 offset:372 ; 4-byte Folded Spill
; VI-NEXT: v_lshrrev_b32_e32 v36, 24, v32
; VI-NEXT: v_lshrrev_b32_e32 v55, 16, v32
; VI-NEXT: v_lshrrev_b32_e32 v50, 8, v32
@@ -120509,9 +120542,9 @@ define inreg <128 x i8> @bitcast_v16f64_to_v128i8_scalar(<16 x double> inreg %a,
; VI-NEXT: v_mov_b32_e32 v33, s71
; VI-NEXT: buffer_store_dword v33, off, s[0:3], s32 offset:388 ; 4-byte Folded Spill
; VI-NEXT: v_mov_b32_e32 v33, s69
-; VI-NEXT: buffer_store_dword v33, off, s[0:3], s32 offset:372 ; 4-byte Folded Spill
-; VI-NEXT: v_mov_b32_e32 v33, s68
; VI-NEXT: buffer_store_dword v33, off, s[0:3], s32 offset:376 ; 4-byte Folded Spill
+; VI-NEXT: v_mov_b32_e32 v33, s68
+; VI-NEXT: buffer_store_dword v33, off, s[0:3], s32 offset:372 ; 4-byte Folded Spill
; VI-NEXT: v_mov_b32_e32 v33, s67
; VI-NEXT: buffer_store_dword v33, off, s[0:3], s32 offset:352 ; 4-byte Folded Spill
; VI-NEXT: v_mov_b32_e32 v33, s66
@@ -120833,12 +120866,12 @@ define inreg <128 x i8> @bitcast_v16f64_to_v128i8_scalar(<16 x double> inreg %a,
; VI-NEXT: buffer_store_dword v31, v32, s[0:3], 0 offen
; VI-NEXT: buffer_load_dword v31, off, s[0:3], s32 offset:324 ; 4-byte Folded Reload
; VI-NEXT: buffer_load_dword v32, off, s[0:3], s32 offset:328 ; 4-byte Folded Reload
-; VI-NEXT: buffer_load_dword v32, off, s[0:3], s32 offset:372 ; 4-byte Folded Reload
+; VI-NEXT: buffer_load_dword v32, off, s[0:3], s32 offset:376 ; 4-byte Folded Reload
; VI-NEXT: s_waitcnt vmcnt(2)
; VI-NEXT: v_lshlrev_b32_e32 v31, 8, v31
; VI-NEXT: s_waitcnt vmcnt(0)
; VI-NEXT: v_or_b32_sdwa v31, v32, v31 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
-; VI-NEXT: buffer_load_dword v32, off, s[0:3], s32 offset:376 ; 4-byte Folded Reload
+; VI-NEXT: buffer_load_dword v32, off, s[0:3], s32 offset:372 ; 4-byte Folded Reload
; VI-NEXT: s_waitcnt vmcnt(0)
; VI-NEXT: v_lshlrev_b32_e32 v32, 8, v32
; VI-NEXT: v_or_b32_sdwa v29, v29, v32 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
@@ -131843,7 +131876,7 @@ define inreg <16 x double> @bitcast_v128i8_to_v16f64_scalar(<128 x i8> inreg %a,
; VI-NEXT: buffer_store_dword v10, off, s[0:3], s32 offset:476 ; 4-byte Folded Spill
; VI-NEXT: buffer_store_dword v8, off, s[0:3], s32 offset:484 ; 4-byte Folded Spill
; VI-NEXT: buffer_store_dword v6, off, s[0:3], s32 offset:488 ; 4-byte Folded Spill
-; VI-NEXT: buffer_store_dword v0, off, s[0:3], s32 offset:556 ; 4-byte Folded Spill
+; VI-NEXT: buffer_store_dword v0, off, s[0:3], s32 offset:560 ; 4-byte Folded Spill
; VI-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:332
; VI-NEXT: buffer_load_ushort v44, off, s[0:3], s32
; VI-NEXT: buffer_load_ushort v43, off, s[0:3], s32 offset:8
@@ -131869,7 +131902,7 @@ define inreg <16 x double> @bitcast_v128i8_to_v16f64_scalar(<128 x i8> inreg %a,
; VI-NEXT: buffer_load_ushort v37, off, s[0:3], s32 offset:168
; VI-NEXT: buffer_load_ushort v38, off, s[0:3], s32 offset:176
; VI-NEXT: v_lshlrev_b32_e32 v1, 8, v1
-; VI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:560 ; 4-byte Folded Spill
+; VI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:564 ; 4-byte Folded Spill
; VI-NEXT: v_lshlrev_b32_e32 v1, 8, v15
; VI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:548 ; 4-byte Folded Spill
; VI-NEXT: v_lshlrev_b32_e32 v1, 8, v17
@@ -131878,11 +131911,11 @@ define inreg <16 x double> @bitcast_v128i8_to_v16f64_scalar(<128 x i8> inreg %a,
; VI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:540 ; 4-byte Folded Spill
; VI-NEXT: v_lshlrev_b32_e32 v1, 8, v21
; VI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:536 ; 4-byte Folded Spill
-; VI-NEXT: v_lshlrev_b32_e32 v8, 8, v3
-; VI-NEXT: v_lshlrev_b32_e32 v59, 8, v5
-; VI-NEXT: v_lshlrev_b32_e32 v14, 8, v7
-; VI-NEXT: v_lshlrev_b32_e32 v10, 8, v9
-; VI-NEXT: v_lshlrev_b32_e32 v16, 8, v11
+; VI-NEXT: v_lshlrev_b32_e32 v56, 8, v3
+; VI-NEXT: v_lshlrev_b32_e32 v8, 8, v5
+; VI-NEXT: v_lshlrev_b32_e32 v10, 8, v7
+; VI-NEXT: v_lshlrev_b32_e32 v14, 8, v9
+; VI-NEXT: v_lshlrev_b32_e32 v12, 8, v11
; VI-NEXT: v_lshlrev_b32_e32 v6, 8, v13
; VI-NEXT: s_waitcnt vmcnt(14)
; VI-NEXT: v_cmp_ne_u32_e32 vcc, 0, v0
@@ -131903,7 +131936,7 @@ define inreg <16 x double> @bitcast_v128i8_to_v16f64_scalar(<128 x i8> inreg %a,
; VI-NEXT: v_lshlrev_b32_e32 v0, 8, v41
; VI-NEXT: buffer_store_dword v0, off, s[0:3], s32 offset:504 ; 4-byte Folded Spill
; VI-NEXT: v_lshlrev_b32_e32 v0, 8, v40
-; VI-NEXT: buffer_store_dword v0, off, s[0:3], s32 offset:580 ; 4-byte Folded Spill
+; VI-NEXT: buffer_store_dword v0, off, s[0:3], s32 offset:588 ; 4-byte Folded Spill
; VI-NEXT: v_lshlrev_b32_e32 v0, 8, v55
; VI-NEXT: buffer_store_dword v0, off, s[0:3], s32 offset:500 ; 4-byte Folded Spill
; VI-NEXT: v_lshlrev_b32_e32 v0, 8, v54
@@ -131911,22 +131944,22 @@ define inreg <16 x double> @bitcast_v128i8_to_v16f64_scalar(<128 x i8> inreg %a,
; VI-NEXT: v_lshlrev_b32_e32 v0, 8, v53
; VI-NEXT: buffer_store_dword v0, off, s[0:3], s32 offset:496 ; 4-byte Folded Spill
; VI-NEXT: v_lshlrev_b32_e32 v0, 8, v52
-; VI-NEXT: buffer_store_dword v0, off, s[0:3], s32 offset:608 ; 4-byte Folded Spill
+; VI-NEXT: buffer_store_dword v0, off, s[0:3], s32 offset:616 ; 4-byte Folded Spill
; VI-NEXT: v_lshlrev_b32_e32 v0, 8, v51
; VI-NEXT: buffer_store_dword v0, off, s[0:3], s32 offset:492 ; 4-byte Folded Spill
; VI-NEXT: v_lshlrev_b32_e32 v0, 8, v50
-; VI-NEXT: buffer_store_dword v0, off, s[0:3], s32 offset:600 ; 4-byte Folded Spill
+; VI-NEXT: buffer_store_dword v0, off, s[0:3], s32 offset:612 ; 4-byte Folded Spill
; VI-NEXT: v_lshlrev_b32_e32 v0, 8, v49
-; VI-NEXT: buffer_store_dword v0, off, s[0:3], s32 offset:604 ; 4-byte Folded Spill
+; VI-NEXT: buffer_store_dword v0, off, s[0:3], s32 offset:608 ; 4-byte Folded Spill
; VI-NEXT: v_lshlrev_b32_e32 v0, 8, v48
-; VI-NEXT: buffer_store_dword v0, off, s[0:3], s32 offset:596 ; 4-byte Folded Spill
+; VI-NEXT: buffer_store_dword v0, off, s[0:3], s32 offset:604 ; 4-byte Folded Spill
; VI-NEXT: v_lshlrev_b32_e32 v0, 8, v39
; VI-NEXT: buffer_store_dword v0, off, s[0:3], s32 offset:480 ; 4-byte Folded Spill
; VI-NEXT: s_waitcnt vmcnt(14)
; VI-NEXT: v_lshlrev_b32_e32 v0, 8, v22
-; VI-NEXT: buffer_store_dword v0, off, s[0:3], s32 offset:572 ; 4-byte Folded Spill
+; VI-NEXT: buffer_store_dword v0, off, s[0:3], s32 offset:596 ; 4-byte Folded Spill
; VI-NEXT: v_lshlrev_b32_e32 v0, 8, v31
-; VI-NEXT: buffer_store_dword v0, off, s[0:3], s32 offset:588 ; 4-byte Folded Spill
+; VI-NEXT: buffer_store_dword v0, off, s[0:3], s32 offset:600 ; 4-byte Folded Spill
; VI-NEXT: v_lshlrev_b32_e32 v0, 8, v32
; VI-NEXT: buffer_store_dword v0, off, s[0:3], s32 offset:592 ; 4-byte Folded Spill
; VI-NEXT: v_lshlrev_b32_e32 v0, 8, v33
@@ -131949,23 +131982,24 @@ define inreg <16 x double> @bitcast_v128i8_to_v16f64_scalar(<128 x i8> inreg %a,
; VI-NEXT: v_lshlrev_b32_e32 v31, 8, v38
; VI-NEXT: s_and_b64 s[4:5], vcc, exec
; VI-NEXT: s_waitcnt vmcnt(7)
-; VI-NEXT: v_lshlrev_b32_e32 v26, 8, v0
+; VI-NEXT: v_lshlrev_b32_e32 v22, 8, v0
; VI-NEXT: s_waitcnt vmcnt(6)
; VI-NEXT: v_lshlrev_b32_e32 v32, 8, v1
; VI-NEXT: s_waitcnt vmcnt(5)
-; VI-NEXT: v_lshlrev_b32_e32 v54, 8, v13
+; VI-NEXT: v_lshlrev_b32_e32 v23, 8, v13
+; VI-NEXT: s_waitcnt vmcnt(4)
+; VI-NEXT: v_lshlrev_b32_e32 v0, 8, v3
+; VI-NEXT: buffer_store_dword v0, off, s[0:3], s32 offset:580 ; 4-byte Folded Spill
; VI-NEXT: s_waitcnt vmcnt(4)
-; VI-NEXT: v_lshlrev_b32_e32 v49, 8, v3
-; VI-NEXT: s_waitcnt vmcnt(3)
; VI-NEXT: v_lshlrev_b32_e32 v0, 8, v11
-; VI-NEXT: buffer_store_dword v0, off, s[0:3], s32 offset:576 ; 4-byte Folded Spill
-; VI-NEXT: s_waitcnt vmcnt(3)
+; VI-NEXT: buffer_store_dword v0, off, s[0:3], s32 offset:572 ; 4-byte Folded Spill
+; VI-NEXT: s_waitcnt vmcnt(4)
; VI-NEXT: v_lshlrev_b32_e32 v0, 8, v5
-; VI-NEXT: buffer_store_dword v0, off, s[0:3], s32 offset:612 ; 4-byte Folded Spill
-; VI-NEXT: s_waitcnt vmcnt(3)
+; VI-NEXT: buffer_store_dword v0, off, s[0:3], s32 offset:576 ; 4-byte Folded Spill
+; VI-NEXT: s_waitcnt vmcnt(4)
; VI-NEXT: v_lshlrev_b32_e32 v0, 8, v9
; VI-NEXT: buffer_store_dword v0, off, s[0:3], s32 offset:416 ; 4-byte Folded Spill
-; VI-NEXT: s_waitcnt vmcnt(3)
+; VI-NEXT: s_waitcnt vmcnt(4)
; VI-NEXT: v_lshlrev_b32_e32 v0, 8, v7
; VI-NEXT: buffer_store_dword v0, off, s[0:3], s32 offset:412 ; 4-byte Folded Spill
; VI-NEXT: buffer_load_ushort v0, off, s[0:3], s32 offset:248
@@ -131978,37 +132012,37 @@ define inreg <16 x double> @bitcast_v128i8_to_v16f64_scalar(<128 x i8> inreg %a,
; VI-NEXT: buffer_load_ushort v7, off, s[0:3], s32 offset:304
; VI-NEXT: s_waitcnt vmcnt(7)
; VI-NEXT: v_lshlrev_b32_e32 v0, 8, v0
-; VI-NEXT: buffer_store_dword v0, off, s[0:3], s32 offset:564 ; 4-byte Folded Spill
+; VI-NEXT: buffer_store_dword v0, off, s[0:3], s32 offset:568 ; 4-byte Folded Spill
; VI-NEXT: s_waitcnt vmcnt(7)
; VI-NEXT: v_lshlrev_b32_e32 v48, 8, v1
; VI-NEXT: s_waitcnt vmcnt(6)
; VI-NEXT: v_lshlrev_b32_e32 v27, 8, v13
; VI-NEXT: s_waitcnt vmcnt(4)
; VI-NEXT: v_lshlrev_b32_e32 v0, 8, v11
-; VI-NEXT: buffer_store_dword v0, off, s[0:3], s32 offset:568 ; 4-byte Folded Spill
+; VI-NEXT: buffer_store_dword v0, off, s[0:3], s32 offset:556 ; 4-byte Folded Spill
; VI-NEXT: s_waitcnt vmcnt(4)
; VI-NEXT: v_lshlrev_b32_e32 v0, 8, v5
; VI-NEXT: buffer_store_dword v0, off, s[0:3], s32 offset:404 ; 4-byte Folded Spill
-; VI-NEXT: v_lshlrev_b32_e32 v29, 8, v3
+; VI-NEXT: v_lshlrev_b32_e32 v28, 8, v3
; VI-NEXT: buffer_load_ushort v0, off, s[0:3], s32 offset:312
; VI-NEXT: buffer_load_ushort v1, off, s[0:3], s32 offset:320
; VI-NEXT: buffer_load_ushort v3, off, s[0:3], s32 offset:328
; VI-NEXT: buffer_load_ushort v11, off, s[0:3], s32 offset:4
-; VI-NEXT: buffer_load_ushort v12, off, s[0:3], s32 offset:12
+; VI-NEXT: buffer_load_ushort v16, off, s[0:3], s32 offset:12
; VI-NEXT: buffer_load_ushort v50, off, s[0:3], s32 offset:20
-; VI-NEXT: buffer_load_ushort v56, off, s[0:3], s32 offset:28
+; VI-NEXT: buffer_load_ushort v59, off, s[0:3], s32 offset:28
; VI-NEXT: buffer_load_ushort v13, off, s[0:3], s32 offset:36
; VI-NEXT: s_waitcnt vmcnt(11)
; VI-NEXT: v_lshlrev_b32_e32 v5, 8, v7
-; VI-NEXT: v_lshlrev_b32_e32 v9, 8, v9
+; VI-NEXT: v_lshlrev_b32_e32 v29, 8, v9
; VI-NEXT: s_waitcnt vmcnt(7)
; VI-NEXT: v_lshlrev_b32_e32 v7, 8, v0
; VI-NEXT: s_waitcnt vmcnt(6)
; VI-NEXT: v_lshlrev_b32_e32 v0, 8, v1
; VI-NEXT: buffer_store_dword v0, off, s[0:3], s32 offset:552 ; 4-byte Folded Spill
; VI-NEXT: buffer_load_ushort v18, off, s[0:3], s32 offset:44
-; VI-NEXT: buffer_load_ushort v15, off, s[0:3], s32 offset:52
-; VI-NEXT: buffer_load_ushort v17, off, s[0:3], s32 offset:60
+; VI-NEXT: buffer_load_ushort v17, off, s[0:3], s32 offset:52
+; VI-NEXT: buffer_load_ushort v15, off, s[0:3], s32 offset:60
; VI-NEXT: buffer_load_ushort v39, off, s[0:3], s32 offset:68
; VI-NEXT: buffer_load_ushort v38, off, s[0:3], s32 offset:76
; VI-NEXT: buffer_load_ushort v37, off, s[0:3], s32 offset:84
@@ -132016,23 +132050,23 @@ define inreg <16 x double> @bitcast_v128i8_to_v16f64_scalar(<128 x i8> inreg %a,
; VI-NEXT: buffer_load_ushort v36, off, s[0:3], s32 offset:100
; VI-NEXT: buffer_load_ushort v35, off, s[0:3], s32 offset:108
; VI-NEXT: buffer_load_ushort v33, off, s[0:3], s32 offset:116
-; VI-NEXT: buffer_load_ushort v25, off, s[0:3], s32 offset:124
+; VI-NEXT: buffer_load_ushort v54, off, s[0:3], s32 offset:124
; VI-NEXT: buffer_load_ushort v19, off, s[0:3], s32 offset:132
-; VI-NEXT: buffer_load_ushort v51, off, s[0:3], s32 offset:140
+; VI-NEXT: buffer_load_ushort v49, off, s[0:3], s32 offset:140
; VI-NEXT: buffer_load_ushort v20, off, s[0:3], s32 offset:148
-; VI-NEXT: buffer_load_ushort v28, off, s[0:3], s32 offset:156
+; VI-NEXT: buffer_load_ushort v51, off, s[0:3], s32 offset:156
; VI-NEXT: buffer_load_ushort v21, off, s[0:3], s32 offset:164
; VI-NEXT: buffer_load_ushort v34, off, s[0:3], s32 offset:172
-; VI-NEXT: buffer_load_ushort v22, off, s[0:3], s32 offset:180
-; VI-NEXT: buffer_load_ushort v23, off, s[0:3], s32 offset:188
+; VI-NEXT: buffer_load_ushort v25, off, s[0:3], s32 offset:180
+; VI-NEXT: buffer_load_ushort v30, off, s[0:3], s32 offset:188
; VI-NEXT: buffer_load_ushort v43, off, s[0:3], s32 offset:196
-; VI-NEXT: buffer_load_ushort v30, off, s[0:3], s32 offset:204
-; VI-NEXT: buffer_load_ushort v24, off, s[0:3], s32 offset:212
+; VI-NEXT: buffer_load_ushort v24, off, s[0:3], s32 offset:204
+; VI-NEXT: buffer_load_ushort v26, off, s[0:3], s32 offset:212
; VI-NEXT: buffer_load_ushort v46, off, s[0:3], s32 offset:220
; VI-NEXT: buffer_load_ushort v61, off, s[0:3], s32 offset:228
; VI-NEXT: buffer_load_ushort v47, off, s[0:3], s32 offset:236
-; VI-NEXT: buffer_load_ushort v45, off, s[0:3], s32 offset:244
-; VI-NEXT: buffer_load_ushort v58, off, s[0:3], s32 offset:252
+; VI-NEXT: buffer_load_ushort v58, off, s[0:3], s32 offset:244
+; VI-NEXT: buffer_load_ushort v45, off, s[0:3], s32 offset:252
; VI-NEXT: buffer_load_ushort v44, off, s[0:3], s32 offset:260
; VI-NEXT: buffer_load_ushort v63, off, s[0:3], s32 offset:268
; VI-NEXT: buffer_load_ushort v42, off, s[0:3], s32 offset:276
@@ -132045,57 +132079,57 @@ define inreg <16 x double> @bitcast_v128i8_to_v16f64_scalar(<128 x i8> inreg %a,
; VI-NEXT: s_waitcnt vmcnt(14)
; VI-NEXT: v_lshlrev_b32_e32 v0, 8, v3
; VI-NEXT: buffer_store_dword v5, off, s[0:3], s32 offset:400 ; 4-byte Folded Spill
-; VI-NEXT: buffer_store_dword v9, off, s[0:3], s32 offset:616 ; 4-byte Folded Spill
+; VI-NEXT: buffer_store_dword v29, off, s[0:3], s32 offset:620 ; 4-byte Folded Spill
; VI-NEXT: buffer_store_dword v0, off, s[0:3], s32 offset:408 ; 4-byte Folded Spill
-; VI-NEXT: buffer_store_dword v27, off, s[0:3], s32 offset:620 ; 4-byte Folded Spill
-; VI-NEXT: buffer_store_dword v48, off, s[0:3], s32 offset:624 ; 4-byte Folded Spill
+; VI-NEXT: buffer_store_dword v27, off, s[0:3], s32 offset:624 ; 4-byte Folded Spill
+; VI-NEXT: buffer_store_dword v48, off, s[0:3], s32 offset:628 ; 4-byte Folded Spill
; VI-NEXT: s_waitcnt vmcnt(12)
-; VI-NEXT: buffer_store_dword v63, off, s[0:3], s32 offset:628 ; 4-byte Folded Spill
-; VI-NEXT: buffer_store_dword v58, off, s[0:3], s32 offset:632 ; 4-byte Folded Spill
-; VI-NEXT: buffer_store_dword v47, off, s[0:3], s32 offset:636 ; 4-byte Folded Spill
-; VI-NEXT: buffer_store_dword v31, off, s[0:3], s32 offset:640 ; 4-byte Folded Spill
-; VI-NEXT: buffer_store_dword v46, off, s[0:3], s32 offset:644 ; 4-byte Folded Spill
-; VI-NEXT: buffer_store_dword v52, off, s[0:3], s32 offset:648 ; 4-byte Folded Spill
-; VI-NEXT: buffer_store_dword v2, off, s[0:3], s32 offset:652 ; 4-byte Folded Spill
-; VI-NEXT: buffer_store_dword v32, off, s[0:3], s32 offset:656 ; 4-byte Folded Spill
-; VI-NEXT: buffer_store_dword v4, off, s[0:3], s32 offset:660 ; 4-byte Folded Spill
-; VI-NEXT: buffer_store_dword v24, off, s[0:3], s32 offset:664 ; 4-byte Folded Spill
-; VI-NEXT: buffer_store_dword v30, off, s[0:3], s32 offset:668 ; 4-byte Folded Spill
-; VI-NEXT: buffer_store_dword v43, off, s[0:3], s32 offset:672 ; 4-byte Folded Spill
-; VI-NEXT: buffer_store_dword v23, off, s[0:3], s32 offset:676 ; 4-byte Folded Spill
-; VI-NEXT: buffer_store_dword v6, off, s[0:3], s32 offset:680 ; 4-byte Folded Spill
-; VI-NEXT: buffer_store_dword v16, off, s[0:3], s32 offset:684 ; 4-byte Folded Spill
-; VI-NEXT: buffer_store_dword v22, off, s[0:3], s32 offset:688 ; 4-byte Folded Spill
-; VI-NEXT: buffer_store_dword v34, off, s[0:3], s32 offset:692 ; 4-byte Folded Spill
-; VI-NEXT: buffer_store_dword v10, off, s[0:3], s32 offset:696 ; 4-byte Folded Spill
+; VI-NEXT: buffer_store_dword v63, off, s[0:3], s32 offset:632 ; 4-byte Folded Spill
+; VI-NEXT: buffer_store_dword v45, off, s[0:3], s32 offset:636 ; 4-byte Folded Spill
+; VI-NEXT: buffer_store_dword v47, off, s[0:3], s32 offset:640 ; 4-byte Folded Spill
+; VI-NEXT: buffer_store_dword v31, off, s[0:3], s32 offset:644 ; 4-byte Folded Spill
+; VI-NEXT: buffer_store_dword v46, off, s[0:3], s32 offset:648 ; 4-byte Folded Spill
+; VI-NEXT: buffer_store_dword v52, off, s[0:3], s32 offset:652 ; 4-byte Folded Spill
+; VI-NEXT: buffer_store_dword v2, off, s[0:3], s32 offset:656 ; 4-byte Folded Spill
+; VI-NEXT: buffer_store_dword v32, off, s[0:3], s32 offset:660 ; 4-byte Folded Spill
+; VI-NEXT: buffer_store_dword v4, off, s[0:3], s32 offset:664 ; 4-byte Folded Spill
+; VI-NEXT: buffer_store_dword v26, off, s[0:3], s32 offset:668 ; 4-byte Folded Spill
+; VI-NEXT: buffer_store_dword v24, off, s[0:3], s32 offset:672 ; 4-byte Folded Spill
+; VI-NEXT: buffer_store_dword v43, off, s[0:3], s32 offset:676 ; 4-byte Folded Spill
+; VI-NEXT: buffer_store_dword v30, off, s[0:3], s32 offset:680 ; 4-byte Folded Spill
+; VI-NEXT: buffer_store_dword v6, off, s[0:3], s32 offset:684 ; 4-byte Folded Spill
+; VI-NEXT: buffer_store_dword v12, off, s[0:3], s32 offset:688 ; 4-byte Folded Spill
+; VI-NEXT: buffer_store_dword v25, off, s[0:3], s32 offset:692 ; 4-byte Folded Spill
+; VI-NEXT: buffer_store_dword v34, off, s[0:3], s32 offset:696 ; 4-byte Folded Spill
; VI-NEXT: buffer_store_dword v14, off, s[0:3], s32 offset:700 ; 4-byte Folded Spill
-; VI-NEXT: buffer_store_dword v21, off, s[0:3], s32 offset:704 ; 4-byte Folded Spill
-; VI-NEXT: buffer_store_dword v28, off, s[0:3], s32 offset:708 ; 4-byte Folded Spill
-; VI-NEXT: buffer_store_dword v20, off, s[0:3], s32 offset:712 ; 4-byte Folded Spill
-; VI-NEXT: buffer_store_dword v51, off, s[0:3], s32 offset:716 ; 4-byte Folded Spill
-; VI-NEXT: buffer_store_dword v59, off, s[0:3], s32 offset:720 ; 4-byte Folded Spill
-; VI-NEXT: buffer_store_dword v19, off, s[0:3], s32 offset:724 ; 4-byte Folded Spill
-; VI-NEXT: buffer_store_dword v8, off, s[0:3], s32 offset:728 ; 4-byte Folded Spill
-; VI-NEXT: buffer_store_dword v25, off, s[0:3], s32 offset:732 ; 4-byte Folded Spill
-; VI-NEXT: buffer_store_dword v33, off, s[0:3], s32 offset:736 ; 4-byte Folded Spill
-; VI-NEXT: buffer_store_dword v35, off, s[0:3], s32 offset:740 ; 4-byte Folded Spill
-; VI-NEXT: buffer_store_dword v36, off, s[0:3], s32 offset:744 ; 4-byte Folded Spill
-; VI-NEXT: buffer_store_dword v57, off, s[0:3], s32 offset:748 ; 4-byte Folded Spill
-; VI-NEXT: buffer_store_dword v37, off, s[0:3], s32 offset:752 ; 4-byte Folded Spill
-; VI-NEXT: buffer_store_dword v38, off, s[0:3], s32 offset:756 ; 4-byte Folded Spill
-; VI-NEXT: buffer_store_dword v39, off, s[0:3], s32 offset:760 ; 4-byte Folded Spill
-; VI-NEXT: buffer_store_dword v17, off, s[0:3], s32 offset:764 ; 4-byte Folded Spill
+; VI-NEXT: buffer_store_dword v10, off, s[0:3], s32 offset:704 ; 4-byte Folded Spill
+; VI-NEXT: buffer_store_dword v21, off, s[0:3], s32 offset:708 ; 4-byte Folded Spill
+; VI-NEXT: buffer_store_dword v51, off, s[0:3], s32 offset:712 ; 4-byte Folded Spill
+; VI-NEXT: buffer_store_dword v20, off, s[0:3], s32 offset:716 ; 4-byte Folded Spill
+; VI-NEXT: buffer_store_dword v49, off, s[0:3], s32 offset:720 ; 4-byte Folded Spill
+; VI-NEXT: buffer_store_dword v8, off, s[0:3], s32 offset:724 ; 4-byte Folded Spill
+; VI-NEXT: buffer_store_dword v19, off, s[0:3], s32 offset:728 ; 4-byte Folded Spill
+; VI-NEXT: buffer_store_dword v56, off, s[0:3], s32 offset:732 ; 4-byte Folded Spill
+; VI-NEXT: buffer_store_dword v54, off, s[0:3], s32 offset:736 ; 4-byte Folded Spill
+; VI-NEXT: buffer_store_dword v33, off, s[0:3], s32 offset:740 ; 4-byte Folded Spill
+; VI-NEXT: buffer_store_dword v35, off, s[0:3], s32 offset:744 ; 4-byte Folded Spill
+; VI-NEXT: buffer_store_dword v36, off, s[0:3], s32 offset:748 ; 4-byte Folded Spill
+; VI-NEXT: buffer_store_dword v57, off, s[0:3], s32 offset:752 ; 4-byte Folded Spill
+; VI-NEXT: buffer_store_dword v37, off, s[0:3], s32 offset:756 ; 4-byte Folded Spill
+; VI-NEXT: buffer_store_dword v38, off, s[0:3], s32 offset:760 ; 4-byte Folded Spill
+; VI-NEXT: buffer_store_dword v39, off, s[0:3], s32 offset:764 ; 4-byte Folded Spill
; VI-NEXT: buffer_store_dword v15, off, s[0:3], s32 offset:768 ; 4-byte Folded Spill
-; VI-NEXT: buffer_store_dword v18, off, s[0:3], s32 offset:772 ; 4-byte Folded Spill
-; VI-NEXT: buffer_store_dword v13, off, s[0:3], s32 offset:776 ; 4-byte Folded Spill
-; VI-NEXT: buffer_store_dword v56, off, s[0:3], s32 offset:780 ; 4-byte Folded Spill
-; VI-NEXT: buffer_store_dword v50, off, s[0:3], s32 offset:784 ; 4-byte Folded Spill
-; VI-NEXT: buffer_store_dword v12, off, s[0:3], s32 offset:788 ; 4-byte Folded Spill
-; VI-NEXT: buffer_store_dword v11, off, s[0:3], s32 offset:792 ; 4-byte Folded Spill
+; VI-NEXT: buffer_store_dword v17, off, s[0:3], s32 offset:772 ; 4-byte Folded Spill
+; VI-NEXT: buffer_store_dword v18, off, s[0:3], s32 offset:776 ; 4-byte Folded Spill
+; VI-NEXT: buffer_store_dword v13, off, s[0:3], s32 offset:780 ; 4-byte Folded Spill
+; VI-NEXT: buffer_store_dword v59, off, s[0:3], s32 offset:784 ; 4-byte Folded Spill
+; VI-NEXT: buffer_store_dword v50, off, s[0:3], s32 offset:788 ; 4-byte Folded Spill
+; VI-NEXT: buffer_store_dword v16, off, s[0:3], s32 offset:792 ; 4-byte Folded Spill
+; VI-NEXT: buffer_store_dword v11, off, s[0:3], s32 offset:796 ; 4-byte Folded Spill
; VI-NEXT: s_cbranch_scc0 .LBB75_2
; VI-NEXT: ; %bb.1: ; %cmp.false
-; VI-NEXT: v_or_b32_sdwa v0, v2, v8 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
-; VI-NEXT: v_or_b32_sdwa v1, v4, v59 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; VI-NEXT: v_or_b32_sdwa v0, v2, v56 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; VI-NEXT: v_or_b32_sdwa v1, v4, v8 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
; VI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:488 ; 4-byte Folded Reload
; VI-NEXT: buffer_load_dword v3, off, s[0:3], s32 offset:484 ; 4-byte Folded Reload
; VI-NEXT: v_or_b32_sdwa v4, v0, v1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_0 src1_sel:DWORD
@@ -132110,12 +132144,12 @@ define inreg <16 x double> @bitcast_v128i8_to_v16f64_scalar(<128 x i8> inreg %a,
; VI-NEXT: s_lshl_b32 s7, s23, 8
; VI-NEXT: s_lshl_b32 s8, s27, 8
; VI-NEXT: s_waitcnt vmcnt(3)
-; VI-NEXT: v_or_b32_sdwa v2, v2, v14 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; VI-NEXT: v_or_b32_sdwa v2, v2, v10 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
; VI-NEXT: s_waitcnt vmcnt(2)
-; VI-NEXT: v_or_b32_sdwa v3, v3, v10 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; VI-NEXT: v_or_b32_sdwa v3, v3, v14 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
; VI-NEXT: v_or_b32_sdwa v5, v2, v3 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_0 src1_sel:DWORD
; VI-NEXT: s_waitcnt vmcnt(1)
-; VI-NEXT: v_or_b32_sdwa v0, v0, v16 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; VI-NEXT: v_or_b32_sdwa v0, v0, v12 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
; VI-NEXT: s_waitcnt vmcnt(0)
; VI-NEXT: v_or_b32_sdwa v1, v1, v6 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
; VI-NEXT: v_or_b32_sdwa v6, v0, v1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_0 src1_sel:DWORD
@@ -132144,9 +132178,7 @@ define inreg <16 x double> @bitcast_v128i8_to_v16f64_scalar(<128 x i8> inreg %a,
; VI-NEXT: v_or_b32_sdwa v0, v0, v1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
; VI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:432 ; 4-byte Folded Reload
; VI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:528 ; 4-byte Folded Reload
-; VI-NEXT: buffer_store_dword v29, off, s[0:3], s32 offset:796 ; 4-byte Folded Spill
-; VI-NEXT: v_mov_b32_e32 v29, v9
-; VI-NEXT: s_waitcnt vmcnt(1)
+; VI-NEXT: s_waitcnt vmcnt(0)
; VI-NEXT: v_or_b32_sdwa v1, v1, v2 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
; VI-NEXT: v_or_b32_sdwa v9, v0, v1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_0 src1_sel:DWORD
; VI-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:428 ; 4-byte Folded Reload
@@ -132169,15 +132201,15 @@ define inreg <16 x double> @bitcast_v128i8_to_v16f64_scalar(<128 x i8> inreg %a,
; VI-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:508 ; 4-byte Folded Reload
; VI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:504 ; 4-byte Folded Reload
; VI-NEXT: s_waitcnt vmcnt(1)
-; VI-NEXT: v_or_b32_sdwa v0, v12, v0 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; VI-NEXT: v_or_b32_sdwa v0, v16, v0 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
; VI-NEXT: s_waitcnt vmcnt(0)
; VI-NEXT: v_or_b32_sdwa v1, v50, v1 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
; VI-NEXT: v_or_b32_sdwa v12, v0, v1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_0 src1_sel:DWORD
-; VI-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:580 ; 4-byte Folded Reload
+; VI-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:588 ; 4-byte Folded Reload
; VI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:500 ; 4-byte Folded Reload
; VI-NEXT: s_waitcnt vmcnt(1)
; VI-NEXT: v_mov_b32_e32 v50, v0
-; VI-NEXT: v_or_b32_sdwa v0, v56, v0 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; VI-NEXT: v_or_b32_sdwa v0, v59, v0 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
; VI-NEXT: s_waitcnt vmcnt(0)
; VI-NEXT: v_or_b32_sdwa v1, v13, v1 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
; VI-NEXT: v_or_b32_sdwa v13, v0, v1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_0 src1_sel:DWORD
@@ -132187,18 +132219,18 @@ define inreg <16 x double> @bitcast_v128i8_to_v16f64_scalar(<128 x i8> inreg %a,
; VI-NEXT: v_mov_b32_e32 v59, v0
; VI-NEXT: v_or_b32_sdwa v0, v18, v0 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
; VI-NEXT: s_waitcnt vmcnt(0)
-; VI-NEXT: v_or_b32_sdwa v1, v15, v1 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; VI-NEXT: v_or_b32_sdwa v1, v17, v1 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
; VI-NEXT: v_or_b32_sdwa v14, v0, v1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_0 src1_sel:DWORD
-; VI-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:608 ; 4-byte Folded Reload
+; VI-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:616 ; 4-byte Folded Reload
; VI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:492 ; 4-byte Folded Reload
; VI-NEXT: s_waitcnt vmcnt(1)
; VI-NEXT: v_mov_b32_e32 v56, v0
-; VI-NEXT: v_or_b32_sdwa v0, v17, v0 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; VI-NEXT: v_or_b32_sdwa v0, v15, v0 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
; VI-NEXT: s_waitcnt vmcnt(0)
; VI-NEXT: v_or_b32_sdwa v1, v39, v1 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
; VI-NEXT: v_or_b32_sdwa v15, v0, v1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_0 src1_sel:DWORD
-; VI-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:600 ; 4-byte Folded Reload
-; VI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:604 ; 4-byte Folded Reload
+; VI-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:612 ; 4-byte Folded Reload
+; VI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:608 ; 4-byte Folded Reload
; VI-NEXT: s_waitcnt vmcnt(1)
; VI-NEXT: v_mov_b32_e32 v39, v0
; VI-NEXT: v_or_b32_sdwa v0, v38, v0 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
@@ -132206,7 +132238,7 @@ define inreg <16 x double> @bitcast_v128i8_to_v16f64_scalar(<128 x i8> inreg %a,
; VI-NEXT: v_mov_b32_e32 v38, v1
; VI-NEXT: v_or_b32_sdwa v1, v37, v1 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
; VI-NEXT: v_or_b32_sdwa v16, v0, v1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_0 src1_sel:DWORD
-; VI-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:596 ; 4-byte Folded Reload
+; VI-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:604 ; 4-byte Folded Reload
; VI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:480 ; 4-byte Folded Reload
; VI-NEXT: s_waitcnt vmcnt(1)
; VI-NEXT: v_mov_b32_e32 v37, v0
@@ -132214,8 +132246,8 @@ define inreg <16 x double> @bitcast_v128i8_to_v16f64_scalar(<128 x i8> inreg %a,
; VI-NEXT: s_waitcnt vmcnt(0)
; VI-NEXT: v_or_b32_sdwa v1, v36, v1 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
; VI-NEXT: v_or_b32_sdwa v17, v0, v1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_0 src1_sel:DWORD
-; VI-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:572 ; 4-byte Folded Reload
-; VI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:588 ; 4-byte Folded Reload
+; VI-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:596 ; 4-byte Folded Reload
+; VI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:600 ; 4-byte Folded Reload
; VI-NEXT: s_waitcnt vmcnt(1)
; VI-NEXT: v_mov_b32_e32 v36, v0
; VI-NEXT: v_or_b32_sdwa v0, v35, v0 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
@@ -132227,39 +132259,41 @@ define inreg <16 x double> @bitcast_v128i8_to_v16f64_scalar(<128 x i8> inreg %a,
; VI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:468 ; 4-byte Folded Reload
; VI-NEXT: s_waitcnt vmcnt(1)
; VI-NEXT: v_mov_b32_e32 v33, v0
-; VI-NEXT: v_or_b32_sdwa v0, v25, v0 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; VI-NEXT: v_or_b32_sdwa v0, v54, v0 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
; VI-NEXT: s_waitcnt vmcnt(0)
; VI-NEXT: v_or_b32_sdwa v1, v19, v1 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
; VI-NEXT: v_or_b32_sdwa v19, v0, v1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_0 src1_sel:DWORD
; VI-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:456 ; 4-byte Folded Reload
; VI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:444 ; 4-byte Folded Reload
; VI-NEXT: s_waitcnt vmcnt(1)
-; VI-NEXT: v_or_b32_sdwa v0, v51, v0 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; VI-NEXT: v_or_b32_sdwa v0, v49, v0 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
; VI-NEXT: s_waitcnt vmcnt(0)
; VI-NEXT: v_or_b32_sdwa v1, v20, v1 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
; VI-NEXT: v_or_b32_sdwa v20, v0, v1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_0 src1_sel:DWORD
; VI-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:436 ; 4-byte Folded Reload
; VI-NEXT: v_or_b32_sdwa v1, v21, v52 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
-; VI-NEXT: v_mov_b32_e32 v51, v3
; VI-NEXT: s_waitcnt vmcnt(0)
-; VI-NEXT: v_or_b32_sdwa v0, v28, v0 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; VI-NEXT: v_or_b32_sdwa v0, v51, v0 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
; VI-NEXT: v_or_b32_sdwa v21, v0, v1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_0 src1_sel:DWORD
; VI-NEXT: v_or_b32_sdwa v0, v34, v31 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
-; VI-NEXT: v_or_b32_sdwa v1, v22, v26 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; VI-NEXT: v_or_b32_sdwa v1, v25, v22 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; VI-NEXT: v_mov_b32_e32 v34, v22
; VI-NEXT: v_or_b32_sdwa v22, v0, v1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_0 src1_sel:DWORD
-; VI-NEXT: v_or_b32_sdwa v0, v23, v32 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
-; VI-NEXT: v_or_b32_sdwa v1, v43, v54 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; VI-NEXT: v_or_b32_sdwa v0, v30, v32 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; VI-NEXT: v_or_b32_sdwa v1, v43, v23 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; VI-NEXT: v_mov_b32_e32 v32, v23
; VI-NEXT: v_or_b32_sdwa v23, v0, v1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_0 src1_sel:DWORD
-; VI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:576 ; 4-byte Folded Reload
-; VI-NEXT: v_mov_b32_e32 v43, v49
-; VI-NEXT: v_or_b32_sdwa v0, v30, v49 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
-; VI-NEXT: v_mov_b32_e32 v32, v54
-; VI-NEXT: v_mov_b32_e32 v34, v26
+; VI-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:580 ; 4-byte Folded Reload
+; VI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:572 ; 4-byte Folded Reload
+; VI-NEXT: v_mov_b32_e32 v51, v3
+; VI-NEXT: s_waitcnt vmcnt(1)
+; VI-NEXT: v_mov_b32_e32 v43, v0
+; VI-NEXT: v_or_b32_sdwa v0, v24, v0 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
; VI-NEXT: s_waitcnt vmcnt(0)
; VI-NEXT: v_mov_b32_e32 v49, v1
-; VI-NEXT: v_or_b32_sdwa v1, v24, v1 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; VI-NEXT: v_or_b32_sdwa v1, v26, v1 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
; VI-NEXT: v_or_b32_sdwa v24, v0, v1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_0 src1_sel:DWORD
-; VI-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:612 ; 4-byte Folded Reload
+; VI-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:576 ; 4-byte Folded Reload
; VI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:416 ; 4-byte Folded Reload
; VI-NEXT: s_waitcnt vmcnt(1)
; VI-NEXT: v_mov_b32_e32 v54, v0
@@ -132269,28 +132303,26 @@ define inreg <16 x double> @bitcast_v128i8_to_v16f64_scalar(<128 x i8> inreg %a,
; VI-NEXT: v_mov_b32_e32 v46, v61
; VI-NEXT: v_or_b32_sdwa v25, v0, v1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_0 src1_sel:DWORD
; VI-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:412 ; 4-byte Folded Reload
-; VI-NEXT: buffer_load_dword v61, off, s[0:3], s32 offset:564 ; 4-byte Folded Reload
+; VI-NEXT: buffer_load_dword v61, off, s[0:3], s32 offset:568 ; 4-byte Folded Reload
; VI-NEXT: s_waitcnt vmcnt(1)
; VI-NEXT: v_or_b32_sdwa v0, v47, v0 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
; VI-NEXT: s_waitcnt vmcnt(0)
-; VI-NEXT: v_or_b32_sdwa v1, v45, v61 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; VI-NEXT: v_or_b32_sdwa v1, v58, v61 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
; VI-NEXT: v_or_b32_sdwa v26, v0, v1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_0 src1_sel:DWORD
-; VI-NEXT: v_or_b32_sdwa v0, v58, v48 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; VI-NEXT: v_or_b32_sdwa v0, v45, v48 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
; VI-NEXT: v_or_b32_sdwa v1, v44, v27 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
-; VI-NEXT: v_mov_b32_e32 v47, v45
; VI-NEXT: v_or_b32_sdwa v27, v0, v1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_0 src1_sel:DWORD
-; VI-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:796 ; 4-byte Folded Reload
-; VI-NEXT: buffer_load_dword v45, off, s[0:3], s32 offset:568 ; 4-byte Folded Reload
-; VI-NEXT: v_mov_b32_e32 v58, v44
-; VI-NEXT: s_waitcnt vmcnt(1)
-; VI-NEXT: v_mov_b32_e32 v48, v0
-; VI-NEXT: v_or_b32_sdwa v0, v63, v0 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; VI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:556 ; 4-byte Folded Reload
+; VI-NEXT: v_or_b32_sdwa v0, v63, v28 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; VI-NEXT: v_mov_b32_e32 v48, v28
+; VI-NEXT: v_mov_b32_e32 v47, v58
+; VI-NEXT: v_mov_b32_e32 v45, v44
+; VI-NEXT: v_mov_b32_e32 v63, v42
; VI-NEXT: s_waitcnt vmcnt(0)
-; VI-NEXT: v_or_b32_sdwa v1, v42, v45 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; VI-NEXT: v_or_b32_sdwa v1, v42, v1 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
; VI-NEXT: v_or_b32_sdwa v28, v0, v1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_0 src1_sel:DWORD
; VI-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:404 ; 4-byte Folded Reload
; VI-NEXT: v_or_b32_sdwa v1, v40, v29 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
-; VI-NEXT: v_mov_b32_e32 v63, v42
; VI-NEXT: s_waitcnt vmcnt(0)
; VI-NEXT: v_or_b32_sdwa v0, v41, v0 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
; VI-NEXT: v_or_b32_sdwa v29, v0, v1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_0 src1_sel:DWORD
@@ -132306,8 +132338,8 @@ define inreg <16 x double> @bitcast_v128i8_to_v16f64_scalar(<128 x i8> inreg %a,
; VI-NEXT: s_waitcnt vmcnt(0)
; VI-NEXT: v_or_b32_sdwa v1, v53, v1 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
; VI-NEXT: v_or_b32_sdwa v31, v0, v1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_0 src1_sel:DWORD
-; VI-NEXT: buffer_load_dword v52, off, s[0:3], s32 offset:556 ; 4-byte Folded Reload
-; VI-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:560 ; 4-byte Folded Reload
+; VI-NEXT: buffer_load_dword v52, off, s[0:3], s32 offset:560 ; 4-byte Folded Reload
+; VI-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:564 ; 4-byte Folded Reload
; VI-NEXT: s_waitcnt vmcnt(0)
; VI-NEXT: v_mov_b32_e32 v57, v0
; VI-NEXT: v_or_b32_sdwa v0, v52, v0 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
@@ -132341,44 +132373,43 @@ define inreg <16 x double> @bitcast_v128i8_to_v16f64_scalar(<128 x i8> inreg %a,
; VI-NEXT: s_mov_b64 s[4:5], 0
; VI-NEXT: s_branch .LBB75_3
; VI-NEXT: .LBB75_2:
-; VI-NEXT: buffer_load_dword v57, off, s[0:3], s32 offset:560 ; 4-byte Folded Reload
-; VI-NEXT: buffer_load_dword v50, off, s[0:3], s32 offset:580 ; 4-byte Folded Reload
+; VI-NEXT: buffer_load_dword v57, off, s[0:3], s32 offset:564 ; 4-byte Folded Reload
+; VI-NEXT: buffer_load_dword v50, off, s[0:3], s32 offset:588 ; 4-byte Folded Reload
; VI-NEXT: buffer_load_dword v59, off, s[0:3], s32 offset:584 ; 4-byte Folded Reload
-; VI-NEXT: buffer_load_dword v56, off, s[0:3], s32 offset:608 ; 4-byte Folded Reload
-; VI-NEXT: buffer_load_dword v39, off, s[0:3], s32 offset:600 ; 4-byte Folded Reload
-; VI-NEXT: buffer_load_dword v38, off, s[0:3], s32 offset:604 ; 4-byte Folded Reload
-; VI-NEXT: buffer_load_dword v37, off, s[0:3], s32 offset:596 ; 4-byte Folded Reload
-; VI-NEXT: buffer_load_dword v36, off, s[0:3], s32 offset:572 ; 4-byte Folded Reload
-; VI-NEXT: buffer_load_dword v35, off, s[0:3], s32 offset:588 ; 4-byte Folded Reload
+; VI-NEXT: buffer_load_dword v56, off, s[0:3], s32 offset:616 ; 4-byte Folded Reload
+; VI-NEXT: buffer_load_dword v39, off, s[0:3], s32 offset:612 ; 4-byte Folded Reload
+; VI-NEXT: buffer_load_dword v38, off, s[0:3], s32 offset:608 ; 4-byte Folded Reload
+; VI-NEXT: buffer_load_dword v37, off, s[0:3], s32 offset:604 ; 4-byte Folded Reload
+; VI-NEXT: buffer_load_dword v36, off, s[0:3], s32 offset:596 ; 4-byte Folded Reload
+; VI-NEXT: buffer_load_dword v35, off, s[0:3], s32 offset:600 ; 4-byte Folded Reload
; VI-NEXT: buffer_load_dword v33, off, s[0:3], s32 offset:592 ; 4-byte Folded Reload
-; VI-NEXT: v_mov_b32_e32 v32, v54
-; VI-NEXT: v_mov_b32_e32 v43, v49
+; VI-NEXT: buffer_load_dword v43, off, s[0:3], s32 offset:580 ; 4-byte Folded Reload
+; VI-NEXT: buffer_load_dword v49, off, s[0:3], s32 offset:572 ; 4-byte Folded Reload
; VI-NEXT: v_mov_b32_e32 v46, v61
-; VI-NEXT: v_mov_b32_e32 v47, v45
-; VI-NEXT: buffer_load_dword v54, off, s[0:3], s32 offset:612 ; 4-byte Folded Reload
-; VI-NEXT: buffer_load_dword v49, off, s[0:3], s32 offset:576 ; 4-byte Folded Reload
-; VI-NEXT: buffer_load_dword v45, off, s[0:3], s32 offset:568 ; 4-byte Folded Reload
-; VI-NEXT: buffer_load_dword v52, off, s[0:3], s32 offset:556 ; 4-byte Folded Reload
-; VI-NEXT: buffer_load_dword v61, off, s[0:3], s32 offset:564 ; 4-byte Folded Reload
-; VI-NEXT: v_mov_b32_e32 v34, v26
-; VI-NEXT: v_mov_b32_e32 v58, v44
+; VI-NEXT: buffer_load_dword v54, off, s[0:3], s32 offset:576 ; 4-byte Folded Reload
+; VI-NEXT: buffer_load_dword v52, off, s[0:3], s32 offset:560 ; 4-byte Folded Reload
+; VI-NEXT: buffer_load_dword v61, off, s[0:3], s32 offset:568 ; 4-byte Folded Reload
+; VI-NEXT: v_mov_b32_e32 v34, v22
+; VI-NEXT: v_mov_b32_e32 v32, v23
+; VI-NEXT: v_mov_b32_e32 v47, v58
+; VI-NEXT: v_mov_b32_e32 v45, v44
; VI-NEXT: s_waitcnt vmcnt(14)
; VI-NEXT: v_mov_b32_e32 v63, v42
; VI-NEXT: v_mov_b32_e32 v51, v7
-; VI-NEXT: v_mov_b32_e32 v48, v29
+; VI-NEXT: v_mov_b32_e32 v48, v28
; VI-NEXT: s_mov_b64 s[4:5], -1
; VI-NEXT: ; implicit-def: $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7_vgpr8_vgpr9_vgpr10_vgpr11_vgpr12_vgpr13_vgpr14_vgpr15_vgpr16_vgpr17_vgpr18_vgpr19_vgpr20_vgpr21_vgpr22_vgpr23_vgpr24_vgpr25_vgpr26_vgpr27_vgpr28_vgpr29_vgpr30_vgpr31
; VI-NEXT: .LBB75_3: ; %Flow
; VI-NEXT: s_andn2_b64 vcc, exec, s[4:5]
-; VI-NEXT: v_mov_b32_e32 v44, v47
-; VI-NEXT: v_mov_b32_e32 v47, v46
+; VI-NEXT: v_mov_b32_e32 v42, v45
+; VI-NEXT: v_mov_b32_e32 v45, v46
; VI-NEXT: s_waitcnt vmcnt(3)
; VI-NEXT: v_mov_b32_e32 v46, v49
; VI-NEXT: s_cbranch_vccnz .LBB75_5
; VI-NEXT: ; %bb.4: ; %cmp.true
-; VI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:652 ; 4-byte Folded Reload
-; VI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:728 ; 4-byte Folded Reload
-; VI-NEXT: buffer_load_dword v3, off, s[0:3], s32 offset:720 ; 4-byte Folded Reload
+; VI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:656 ; 4-byte Folded Reload
+; VI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:732 ; 4-byte Folded Reload
+; VI-NEXT: buffer_load_dword v3, off, s[0:3], s32 offset:724 ; 4-byte Folded Reload
; VI-NEXT: s_add_i32 s28, s28, 3
; VI-NEXT: s_and_b32 s4, s28, 0xff
; VI-NEXT: s_lshl_b32 s5, s29, 8
@@ -132432,7 +132463,7 @@ define inreg <16 x double> @bitcast_v128i8_to_v16f64_scalar(<128 x i8> inreg %a,
; VI-NEXT: v_add_u32_e32 v1, vcc, 3, v1
; VI-NEXT: s_waitcnt vmcnt(1)
; VI-NEXT: v_or_b32_sdwa v1, v2, v1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:BYTE_0
-; VI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:660 ; 4-byte Folded Reload
+; VI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:664 ; 4-byte Folded Reload
; VI-NEXT: v_add_u32_e32 v1, vcc, 0x300, v1
; VI-NEXT: s_waitcnt vmcnt(0)
; VI-NEXT: v_add_u32_e32 v2, vcc, 3, v2
@@ -132441,8 +132472,8 @@ define inreg <16 x double> @bitcast_v128i8_to_v16f64_scalar(<128 x i8> inreg %a,
; VI-NEXT: v_add_u32_e32 v3, vcc, 0x3000000, v0
; VI-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:488 ; 4-byte Folded Reload
; VI-NEXT: v_add_u32_e32 v4, vcc, 0x3000000, v1
-; VI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:700 ; 4-byte Folded Reload
-; VI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:696 ; 4-byte Folded Reload
+; VI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:704 ; 4-byte Folded Reload
+; VI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:700 ; 4-byte Folded Reload
; VI-NEXT: s_waitcnt vmcnt(2)
; VI-NEXT: v_add_u32_e32 v0, vcc, 3, v0
; VI-NEXT: s_waitcnt vmcnt(1)
@@ -132455,8 +132486,8 @@ define inreg <16 x double> @bitcast_v128i8_to_v16f64_scalar(<128 x i8> inreg %a,
; VI-NEXT: v_or_b32_sdwa v0, v1, v0 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:WORD_0
; VI-NEXT: v_add_u32_e32 v5, vcc, 0x3000000, v0
; VI-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:476 ; 4-byte Folded Reload
-; VI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:684 ; 4-byte Folded Reload
-; VI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:680 ; 4-byte Folded Reload
+; VI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:688 ; 4-byte Folded Reload
+; VI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:684 ; 4-byte Folded Reload
; VI-NEXT: s_waitcnt vmcnt(2)
; VI-NEXT: v_add_u32_e32 v0, vcc, 3, v0
; VI-NEXT: s_waitcnt vmcnt(1)
@@ -132531,29 +132562,29 @@ define inreg <16 x double> @bitcast_v128i8_to_v16f64_scalar(<128 x i8> inreg %a,
; VI-NEXT: v_add_u32_e32 v0, vcc, 3, v0
; VI-NEXT: s_waitcnt vmcnt(1)
; VI-NEXT: v_or_b32_sdwa v0, v1, v0 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:BYTE_0
-; VI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:792 ; 4-byte Folded Reload
+; VI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:796 ; 4-byte Folded Reload
; VI-NEXT: v_add_u32_e32 v0, vcc, 0x300, v0
; VI-NEXT: s_waitcnt vmcnt(0)
; VI-NEXT: v_add_u32_e32 v1, vcc, 3, v1
; VI-NEXT: v_or_b32_sdwa v1, v2, v1 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:BYTE_0
; VI-NEXT: v_or_b32_sdwa v0, v1, v0 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:WORD_0
; VI-NEXT: v_add_u32_e32 v11, vcc, 0x3000000, v0
-; VI-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:788 ; 4-byte Folded Reload
+; VI-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:792 ; 4-byte Folded Reload
; VI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:508 ; 4-byte Folded Reload
; VI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:504 ; 4-byte Folded Reload
; VI-NEXT: s_waitcnt vmcnt(2)
; VI-NEXT: v_add_u32_e32 v0, vcc, 3, v0
; VI-NEXT: s_waitcnt vmcnt(1)
; VI-NEXT: v_or_b32_sdwa v0, v1, v0 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:BYTE_0
-; VI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:784 ; 4-byte Folded Reload
+; VI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:788 ; 4-byte Folded Reload
; VI-NEXT: v_add_u32_e32 v0, vcc, 0x300, v0
; VI-NEXT: s_waitcnt vmcnt(0)
; VI-NEXT: v_add_u32_e32 v1, vcc, 3, v1
; VI-NEXT: v_or_b32_sdwa v1, v2, v1 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:BYTE_0
; VI-NEXT: v_or_b32_sdwa v0, v1, v0 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:WORD_0
; VI-NEXT: v_add_u32_e32 v12, vcc, 0x3000000, v0
-; VI-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:780 ; 4-byte Folded Reload
-; VI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:776 ; 4-byte Folded Reload
+; VI-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:784 ; 4-byte Folded Reload
+; VI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:780 ; 4-byte Folded Reload
; VI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:500 ; 4-byte Folded Reload
; VI-NEXT: s_waitcnt vmcnt(2)
; VI-NEXT: v_add_u32_e32 v0, vcc, 3, v0
@@ -132565,8 +132596,8 @@ define inreg <16 x double> @bitcast_v128i8_to_v16f64_scalar(<128 x i8> inreg %a,
; VI-NEXT: v_add_u32_e32 v0, vcc, 0x300, v0
; VI-NEXT: v_or_b32_sdwa v0, v1, v0 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:WORD_0
; VI-NEXT: v_add_u32_e32 v13, vcc, 0x3000000, v0
-; VI-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:772 ; 4-byte Folded Reload
-; VI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:768 ; 4-byte Folded Reload
+; VI-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:776 ; 4-byte Folded Reload
+; VI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:772 ; 4-byte Folded Reload
; VI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:496 ; 4-byte Folded Reload
; VI-NEXT: s_waitcnt vmcnt(2)
; VI-NEXT: v_add_u32_e32 v0, vcc, 3, v0
@@ -132578,8 +132609,8 @@ define inreg <16 x double> @bitcast_v128i8_to_v16f64_scalar(<128 x i8> inreg %a,
; VI-NEXT: v_add_u32_e32 v0, vcc, 0x300, v0
; VI-NEXT: v_or_b32_sdwa v0, v1, v0 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:WORD_0
; VI-NEXT: v_add_u32_e32 v14, vcc, 0x3000000, v0
-; VI-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:764 ; 4-byte Folded Reload
-; VI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:760 ; 4-byte Folded Reload
+; VI-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:768 ; 4-byte Folded Reload
+; VI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:764 ; 4-byte Folded Reload
; VI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:492 ; 4-byte Folded Reload
; VI-NEXT: s_waitcnt vmcnt(2)
; VI-NEXT: v_add_u32_e32 v0, vcc, 3, v0
@@ -132591,8 +132622,8 @@ define inreg <16 x double> @bitcast_v128i8_to_v16f64_scalar(<128 x i8> inreg %a,
; VI-NEXT: v_add_u32_e32 v0, vcc, 0x300, v0
; VI-NEXT: v_or_b32_sdwa v0, v1, v0 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:WORD_0
; VI-NEXT: v_add_u32_e32 v15, vcc, 0x3000000, v0
-; VI-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:756 ; 4-byte Folded Reload
-; VI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:752 ; 4-byte Folded Reload
+; VI-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:760 ; 4-byte Folded Reload
+; VI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:756 ; 4-byte Folded Reload
; VI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:480 ; 4-byte Folded Reload
; VI-NEXT: s_waitcnt vmcnt(2)
; VI-NEXT: v_add_u32_e32 v0, vcc, 3, v0
@@ -132603,8 +132634,8 @@ define inreg <16 x double> @bitcast_v128i8_to_v16f64_scalar(<128 x i8> inreg %a,
; VI-NEXT: v_add_u32_e32 v0, vcc, 0x300, v0
; VI-NEXT: v_or_b32_sdwa v0, v1, v0 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:WORD_0
; VI-NEXT: v_add_u32_e32 v16, vcc, 0x3000000, v0
-; VI-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:748 ; 4-byte Folded Reload
-; VI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:744 ; 4-byte Folded Reload
+; VI-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:752 ; 4-byte Folded Reload
+; VI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:748 ; 4-byte Folded Reload
; VI-NEXT: s_waitcnt vmcnt(1)
; VI-NEXT: v_add_u32_e32 v0, vcc, 3, v0
; VI-NEXT: v_or_b32_sdwa v0, v37, v0 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:BYTE_0
@@ -132614,8 +132645,8 @@ define inreg <16 x double> @bitcast_v128i8_to_v16f64_scalar(<128 x i8> inreg %a,
; VI-NEXT: v_add_u32_e32 v0, vcc, 0x300, v0
; VI-NEXT: v_or_b32_sdwa v0, v1, v0 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:WORD_0
; VI-NEXT: v_add_u32_e32 v17, vcc, 0x3000000, v0
-; VI-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:740 ; 4-byte Folded Reload
-; VI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:736 ; 4-byte Folded Reload
+; VI-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:744 ; 4-byte Folded Reload
+; VI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:740 ; 4-byte Folded Reload
; VI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:468 ; 4-byte Folded Reload
; VI-NEXT: s_waitcnt vmcnt(2)
; VI-NEXT: v_add_u32_e32 v0, vcc, 3, v0
@@ -132626,8 +132657,8 @@ define inreg <16 x double> @bitcast_v128i8_to_v16f64_scalar(<128 x i8> inreg %a,
; VI-NEXT: v_add_u32_e32 v0, vcc, 0x300, v0
; VI-NEXT: v_or_b32_sdwa v0, v1, v0 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:WORD_0
; VI-NEXT: v_add_u32_e32 v18, vcc, 0x3000000, v0
-; VI-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:732 ; 4-byte Folded Reload
-; VI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:724 ; 4-byte Folded Reload
+; VI-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:736 ; 4-byte Folded Reload
+; VI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:728 ; 4-byte Folded Reload
; VI-NEXT: s_waitcnt vmcnt(1)
; VI-NEXT: v_add_u32_e32 v0, vcc, 3, v0
; VI-NEXT: v_or_b32_sdwa v0, v33, v0 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:BYTE_0
@@ -132637,63 +132668,63 @@ define inreg <16 x double> @bitcast_v128i8_to_v16f64_scalar(<128 x i8> inreg %a,
; VI-NEXT: v_add_u32_e32 v0, vcc, 0x300, v0
; VI-NEXT: v_or_b32_sdwa v0, v1, v0 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:WORD_0
; VI-NEXT: v_add_u32_e32 v19, vcc, 0x3000000, v0
-; VI-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:716 ; 4-byte Folded Reload
+; VI-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:720 ; 4-byte Folded Reload
; VI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:456 ; 4-byte Folded Reload
; VI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:444 ; 4-byte Folded Reload
; VI-NEXT: s_waitcnt vmcnt(2)
; VI-NEXT: v_add_u32_e32 v0, vcc, 3, v0
; VI-NEXT: s_waitcnt vmcnt(1)
; VI-NEXT: v_or_b32_sdwa v0, v1, v0 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:BYTE_0
-; VI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:712 ; 4-byte Folded Reload
+; VI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:716 ; 4-byte Folded Reload
; VI-NEXT: v_add_u32_e32 v0, vcc, 0x300, v0
; VI-NEXT: s_waitcnt vmcnt(0)
; VI-NEXT: v_add_u32_e32 v1, vcc, 3, v1
; VI-NEXT: v_or_b32_sdwa v1, v2, v1 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:BYTE_0
; VI-NEXT: v_or_b32_sdwa v0, v1, v0 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:WORD_0
; VI-NEXT: v_add_u32_e32 v20, vcc, 0x3000000, v0
-; VI-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:708 ; 4-byte Folded Reload
+; VI-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:712 ; 4-byte Folded Reload
; VI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:436 ; 4-byte Folded Reload
-; VI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:648 ; 4-byte Folded Reload
+; VI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:652 ; 4-byte Folded Reload
; VI-NEXT: s_waitcnt vmcnt(2)
; VI-NEXT: v_add_u32_e32 v0, vcc, 3, v0
; VI-NEXT: s_waitcnt vmcnt(1)
; VI-NEXT: v_or_b32_sdwa v0, v1, v0 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:BYTE_0
-; VI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:704 ; 4-byte Folded Reload
+; VI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:708 ; 4-byte Folded Reload
; VI-NEXT: v_add_u32_e32 v0, vcc, 0x300, v0
; VI-NEXT: s_waitcnt vmcnt(0)
; VI-NEXT: v_add_u32_e32 v1, vcc, 3, v1
; VI-NEXT: v_or_b32_sdwa v1, v2, v1 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:BYTE_0
; VI-NEXT: v_or_b32_sdwa v0, v1, v0 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:WORD_0
; VI-NEXT: v_add_u32_e32 v21, vcc, 0x3000000, v0
-; VI-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:692 ; 4-byte Folded Reload
-; VI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:640 ; 4-byte Folded Reload
+; VI-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:696 ; 4-byte Folded Reload
+; VI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:644 ; 4-byte Folded Reload
; VI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:416 ; 4-byte Folded Reload
; VI-NEXT: s_waitcnt vmcnt(2)
; VI-NEXT: v_add_u32_e32 v0, vcc, 3, v0
; VI-NEXT: s_waitcnt vmcnt(1)
; VI-NEXT: v_or_b32_sdwa v0, v1, v0 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:BYTE_0
-; VI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:688 ; 4-byte Folded Reload
+; VI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:692 ; 4-byte Folded Reload
; VI-NEXT: v_add_u32_e32 v0, vcc, 0x300, v0
; VI-NEXT: s_waitcnt vmcnt(0)
; VI-NEXT: v_add_u32_e32 v1, vcc, 3, v1
; VI-NEXT: v_or_b32_sdwa v1, v34, v1 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:BYTE_0
; VI-NEXT: v_or_b32_sdwa v0, v1, v0 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:WORD_0
; VI-NEXT: v_add_u32_e32 v22, vcc, 0x3000000, v0
-; VI-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:676 ; 4-byte Folded Reload
-; VI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:656 ; 4-byte Folded Reload
+; VI-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:680 ; 4-byte Folded Reload
+; VI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:660 ; 4-byte Folded Reload
; VI-NEXT: s_waitcnt vmcnt(1)
; VI-NEXT: v_add_u32_e32 v0, vcc, 3, v0
; VI-NEXT: s_waitcnt vmcnt(0)
; VI-NEXT: v_or_b32_sdwa v0, v1, v0 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:BYTE_0
-; VI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:672 ; 4-byte Folded Reload
+; VI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:676 ; 4-byte Folded Reload
; VI-NEXT: v_add_u32_e32 v0, vcc, 0x300, v0
; VI-NEXT: s_waitcnt vmcnt(0)
; VI-NEXT: v_add_u32_e32 v1, vcc, 3, v1
; VI-NEXT: v_or_b32_sdwa v1, v32, v1 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:BYTE_0
; VI-NEXT: v_or_b32_sdwa v0, v1, v0 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:WORD_0
; VI-NEXT: v_add_u32_e32 v23, vcc, 0x3000000, v0
-; VI-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:668 ; 4-byte Folded Reload
-; VI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:664 ; 4-byte Folded Reload
+; VI-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:672 ; 4-byte Folded Reload
+; VI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:668 ; 4-byte Folded Reload
; VI-NEXT: s_waitcnt vmcnt(1)
; VI-NEXT: v_add_u32_e32 v0, vcc, 3, v0
; VI-NEXT: v_or_b32_sdwa v0, v43, v0 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:BYTE_0
@@ -132703,54 +132734,57 @@ define inreg <16 x double> @bitcast_v128i8_to_v16f64_scalar(<128 x i8> inreg %a,
; VI-NEXT: v_add_u32_e32 v0, vcc, 0x300, v0
; VI-NEXT: v_or_b32_sdwa v0, v1, v0 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:WORD_0
; VI-NEXT: v_add_u32_e32 v24, vcc, 0x3000000, v0
-; VI-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:644 ; 4-byte Folded Reload
-; VI-NEXT: v_add_u32_e32 v1, vcc, 3, v47
+; VI-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:648 ; 4-byte Folded Reload
+; VI-NEXT: v_add_u32_e32 v1, vcc, 3, v45
; VI-NEXT: v_or_b32_sdwa v1, v2, v1 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:BYTE_0
-; VI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:620 ; 4-byte Folded Reload
+; VI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:624 ; 4-byte Folded Reload
; VI-NEXT: s_waitcnt vmcnt(1)
; VI-NEXT: v_add_u32_e32 v0, vcc, 3, v0
; VI-NEXT: v_or_b32_sdwa v0, v54, v0 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:BYTE_0
; VI-NEXT: v_add_u32_e32 v0, vcc, 0x300, v0
; VI-NEXT: v_or_b32_sdwa v0, v1, v0 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:WORD_0
; VI-NEXT: v_add_u32_e32 v25, vcc, 0x3000000, v0
-; VI-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:636 ; 4-byte Folded Reload
+; VI-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:640 ; 4-byte Folded Reload
; VI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:412 ; 4-byte Folded Reload
; VI-NEXT: s_waitcnt vmcnt(1)
; VI-NEXT: v_add_u32_e32 v0, vcc, 3, v0
; VI-NEXT: s_waitcnt vmcnt(0)
; VI-NEXT: v_or_b32_sdwa v0, v1, v0 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:BYTE_0
-; VI-NEXT: v_add_u32_e32 v1, vcc, 3, v44
+; VI-NEXT: v_add_u32_e32 v1, vcc, 3, v47
; VI-NEXT: v_add_u32_e32 v0, vcc, 0x300, v0
; VI-NEXT: v_or_b32_sdwa v1, v61, v1 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:BYTE_0
; VI-NEXT: v_or_b32_sdwa v0, v1, v0 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:WORD_0
; VI-NEXT: v_add_u32_e32 v26, vcc, 0x3000000, v0
-; VI-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:632 ; 4-byte Folded Reload
-; VI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:624 ; 4-byte Folded Reload
+; VI-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:636 ; 4-byte Folded Reload
+; VI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:628 ; 4-byte Folded Reload
; VI-NEXT: s_waitcnt vmcnt(1)
; VI-NEXT: v_add_u32_e32 v0, vcc, 3, v0
; VI-NEXT: s_waitcnt vmcnt(0)
; VI-NEXT: v_or_b32_sdwa v0, v1, v0 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:BYTE_0
-; VI-NEXT: v_add_u32_e32 v1, vcc, 3, v58
+; VI-NEXT: v_add_u32_e32 v1, vcc, 3, v42
; VI-NEXT: v_add_u32_e32 v0, vcc, 0x300, v0
; VI-NEXT: v_or_b32_sdwa v1, v2, v1 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:BYTE_0
; VI-NEXT: v_or_b32_sdwa v0, v1, v0 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:WORD_0
; VI-NEXT: v_add_u32_e32 v27, vcc, 0x3000000, v0
-; VI-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:628 ; 4-byte Folded Reload
+; VI-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:632 ; 4-byte Folded Reload
+; VI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:556 ; 4-byte Folded Reload
; VI-NEXT: v_add_u32_e32 v1, vcc, 3, v63
-; VI-NEXT: v_or_b32_sdwa v1, v45, v1 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:BYTE_0
-; VI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:616 ; 4-byte Folded Reload
; VI-NEXT: s_waitcnt vmcnt(1)
; VI-NEXT: v_add_u32_e32 v0, vcc, 3, v0
; VI-NEXT: v_or_b32_sdwa v0, v48, v0 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:BYTE_0
; VI-NEXT: v_add_u32_e32 v0, vcc, 0x300, v0
+; VI-NEXT: s_waitcnt vmcnt(0)
+; VI-NEXT: v_or_b32_sdwa v1, v2, v1 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:BYTE_0
; VI-NEXT: v_or_b32_sdwa v0, v1, v0 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:WORD_0
; VI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:404 ; 4-byte Folded Reload
+; VI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:620 ; 4-byte Folded Reload
; VI-NEXT: v_add_u32_e32 v28, vcc, 0x3000000, v0
; VI-NEXT: v_add_u32_e32 v0, vcc, 3, v41
-; VI-NEXT: s_waitcnt vmcnt(0)
+; VI-NEXT: s_waitcnt vmcnt(1)
; VI-NEXT: v_or_b32_sdwa v0, v1, v0 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:BYTE_0
; VI-NEXT: v_add_u32_e32 v1, vcc, 3, v40
; VI-NEXT: v_add_u32_e32 v0, vcc, 0x300, v0
+; VI-NEXT: s_waitcnt vmcnt(0)
; VI-NEXT: v_or_b32_sdwa v1, v2, v1 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:BYTE_0
; VI-NEXT: v_or_b32_sdwa v0, v1, v0 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:WORD_0
; VI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:400 ; 4-byte Folded Reload
@@ -133035,7 +133069,7 @@ define inreg <16 x double> @bitcast_v128i8_to_v16f64_scalar(<128 x i8> inreg %a,
; GFX9-NEXT: buffer_load_ushort v31, off, s[0:3], s32 offset:164
; GFX9-NEXT: buffer_load_ushort v23, off, s[0:3], s32 offset:172
; GFX9-NEXT: buffer_load_ushort v50, off, s[0:3], s32 offset:180
-; GFX9-NEXT: buffer_load_ushort v44, off, s[0:3], s32 offset:188
+; GFX9-NEXT: buffer_load_ushort v40, off, s[0:3], s32 offset:188
; GFX9-NEXT: buffer_load_ushort v37, off, s[0:3], s32 offset:196
; GFX9-NEXT: buffer_load_ushort v52, off, s[0:3], s32 offset:204
; GFX9-NEXT: buffer_load_ushort v29, off, s[0:3], s32 offset:212
@@ -133043,11 +133077,11 @@ define inreg <16 x double> @bitcast_v128i8_to_v16f64_scalar(<128 x i8> inreg %a,
; GFX9-NEXT: buffer_load_ushort v48, off, s[0:3], s32 offset:228
; GFX9-NEXT: buffer_load_ushort v39, off, s[0:3], s32 offset:236
; GFX9-NEXT: buffer_load_ushort v45, off, s[0:3], s32 offset:244
-; GFX9-NEXT: buffer_load_ushort v40, off, s[0:3], s32 offset:252
+; GFX9-NEXT: buffer_load_ushort v44, off, s[0:3], s32 offset:252
; GFX9-NEXT: buffer_load_ushort v55, off, s[0:3], s32 offset:260
; GFX9-NEXT: buffer_load_ushort v43, off, s[0:3], s32 offset:268
-; GFX9-NEXT: buffer_load_ushort v36, off, s[0:3], s32 offset:276
-; GFX9-NEXT: buffer_load_ushort v42, off, s[0:3], s32 offset:284
+; GFX9-NEXT: buffer_load_ushort v42, off, s[0:3], s32 offset:276
+; GFX9-NEXT: buffer_load_ushort v36, off, s[0:3], s32 offset:284
; GFX9-NEXT: buffer_load_ushort v41, off, s[0:3], s32 offset:292
; GFX9-NEXT: buffer_load_ushort v32, off, s[0:3], s32 offset:300
; GFX9-NEXT: buffer_load_ushort v35, off, s[0:3], s32 offset:308
@@ -133073,7 +133107,7 @@ define inreg <16 x double> @bitcast_v128i8_to_v16f64_scalar(<128 x i8> inreg %a,
; GFX9-NEXT: buffer_store_dword v29, off, s[0:3], s32 offset:668 ; 4-byte Folded Spill
; GFX9-NEXT: buffer_store_dword v52, off, s[0:3], s32 offset:672 ; 4-byte Folded Spill
; GFX9-NEXT: buffer_store_dword v37, off, s[0:3], s32 offset:676 ; 4-byte Folded Spill
-; GFX9-NEXT: buffer_store_dword v44, off, s[0:3], s32 offset:680 ; 4-byte Folded Spill
+; GFX9-NEXT: buffer_store_dword v40, off, s[0:3], s32 offset:680 ; 4-byte Folded Spill
; GFX9-NEXT: buffer_store_dword v50, off, s[0:3], s32 offset:684 ; 4-byte Folded Spill
; GFX9-NEXT: buffer_store_dword v23, off, s[0:3], s32 offset:688 ; 4-byte Folded Spill
; GFX9-NEXT: buffer_store_dword v31, off, s[0:3], s32 offset:692 ; 4-byte Folded Spill
@@ -133248,7 +133282,7 @@ define inreg <16 x double> @bitcast_v128i8_to_v16f64_scalar(<128 x i8> inreg %a,
; GFX9-NEXT: s_waitcnt vmcnt(1)
; GFX9-NEXT: v_or_b32_sdwa v0, v23, v0 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
; GFX9-NEXT: v_or_b32_sdwa v22, v0, v1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_0 src1_sel:DWORD
-; GFX9-NEXT: v_or_b32_sdwa v0, v44, v58 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; GFX9-NEXT: v_or_b32_sdwa v0, v40, v58 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
; GFX9-NEXT: v_or_b32_sdwa v1, v37, v57 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
; GFX9-NEXT: v_or_b32_sdwa v23, v0, v1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_0 src1_sel:DWORD
; GFX9-NEXT: s_waitcnt vmcnt(0)
@@ -133260,7 +133294,7 @@ define inreg <16 x double> @bitcast_v128i8_to_v16f64_scalar(<128 x i8> inreg %a,
; GFX9-NEXT: v_mov_b32_e32 v37, v57
; GFX9-NEXT: v_mov_b32_e32 v57, v60
; GFX9-NEXT: v_mov_b32_e32 v52, v56
-; GFX9-NEXT: buffer_load_dword v44, off, s[0:3], s32 offset:612 ; 4-byte Folded Reload
+; GFX9-NEXT: buffer_load_dword v40, off, s[0:3], s32 offset:612 ; 4-byte Folded Reload
; GFX9-NEXT: s_waitcnt vmcnt(2)
; GFX9-NEXT: v_mov_b32_e32 v34, v0
; GFX9-NEXT: v_or_b32_sdwa v0, v46, v0 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
@@ -133269,14 +133303,14 @@ define inreg <16 x double> @bitcast_v128i8_to_v16f64_scalar(<128 x i8> inreg %a,
; GFX9-NEXT: v_or_b32_sdwa v25, v0, v1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_0 src1_sel:DWORD
; GFX9-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:440 ; 4-byte Folded Reload
; GFX9-NEXT: s_waitcnt vmcnt(1)
-; GFX9-NEXT: v_or_b32_sdwa v1, v45, v44 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; GFX9-NEXT: v_or_b32_sdwa v1, v45, v40 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
; GFX9-NEXT: s_waitcnt vmcnt(0)
; GFX9-NEXT: v_or_b32_sdwa v0, v39, v0 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
; GFX9-NEXT: v_or_b32_sdwa v26, v0, v1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_0 src1_sel:DWORD
; GFX9-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:436 ; 4-byte Folded Reload
; GFX9-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:432 ; 4-byte Folded Reload
; GFX9-NEXT: s_waitcnt vmcnt(1)
-; GFX9-NEXT: v_or_b32_sdwa v0, v40, v0 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; GFX9-NEXT: v_or_b32_sdwa v0, v44, v0 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
; GFX9-NEXT: s_waitcnt vmcnt(0)
; GFX9-NEXT: v_or_b32_sdwa v1, v55, v1 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
; GFX9-NEXT: v_or_b32_sdwa v27, v0, v1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_0 src1_sel:DWORD
@@ -133286,12 +133320,12 @@ define inreg <16 x double> @bitcast_v128i8_to_v16f64_scalar(<128 x i8> inreg %a,
; GFX9-NEXT: v_mov_b32_e32 v51, v0
; GFX9-NEXT: v_or_b32_sdwa v0, v43, v0 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
; GFX9-NEXT: s_waitcnt vmcnt(0)
-; GFX9-NEXT: v_or_b32_sdwa v1, v36, v1 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; GFX9-NEXT: v_or_b32_sdwa v1, v42, v1 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
; GFX9-NEXT: v_or_b32_sdwa v28, v0, v1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_0 src1_sel:DWORD
; GFX9-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:420 ; 4-byte Folded Reload
; GFX9-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:416 ; 4-byte Folded Reload
; GFX9-NEXT: s_waitcnt vmcnt(1)
-; GFX9-NEXT: v_or_b32_sdwa v0, v42, v0 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; GFX9-NEXT: v_or_b32_sdwa v0, v36, v0 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
; GFX9-NEXT: s_waitcnt vmcnt(0)
; GFX9-NEXT: v_or_b32_sdwa v1, v41, v1 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
; GFX9-NEXT: v_or_b32_sdwa v29, v0, v1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_0 src1_sel:DWORD
@@ -133345,7 +133379,7 @@ define inreg <16 x double> @bitcast_v128i8_to_v16f64_scalar(<128 x i8> inreg %a,
; GFX9-NEXT: .LBB75_2:
; GFX9-NEXT: buffer_load_dword v34, off, s[0:3], s32 offset:616 ; 4-byte Folded Reload
; GFX9-NEXT: buffer_load_dword v51, off, s[0:3], s32 offset:620 ; 4-byte Folded Reload
-; GFX9-NEXT: buffer_load_dword v44, off, s[0:3], s32 offset:612 ; 4-byte Folded Reload
+; GFX9-NEXT: buffer_load_dword v40, off, s[0:3], s32 offset:612 ; 4-byte Folded Reload
; GFX9-NEXT: buffer_load_dword v50, off, s[0:3], s32 offset:608 ; 4-byte Folded Reload
; GFX9-NEXT: buffer_load_dword v37, off, s[0:3], s32 offset:604 ; 4-byte Folded Reload
; GFX9-NEXT: buffer_load_dword v47, off, s[0:3], s32 offset:596 ; 4-byte Folded Reload
@@ -133707,12 +133741,12 @@ define inreg <16 x double> @bitcast_v128i8_to_v16f64_scalar(<128 x i8> inreg %a,
; GFX9-NEXT: s_waitcnt vmcnt(1)
; GFX9-NEXT: v_or_b32_sdwa v0, v1, v0 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:BYTE_0
; GFX9-NEXT: v_add_u32_e32 v1, 3, v45
-; GFX9-NEXT: v_or_b32_sdwa v1, v44, v1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:BYTE_0
+; GFX9-NEXT: v_or_b32_sdwa v1, v40, v1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:BYTE_0
; GFX9-NEXT: v_add_u32_e32 v0, 0x300, v0
; GFX9-NEXT: v_add_u32_sdwa v1, v1, s4 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:DWORD
; GFX9-NEXT: v_or_b32_sdwa v26, v0, v1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_0 src1_sel:DWORD
; GFX9-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:436 ; 4-byte Folded Reload
-; GFX9-NEXT: v_add_u32_e32 v0, 3, v40
+; GFX9-NEXT: v_add_u32_e32 v0, 3, v44
; GFX9-NEXT: s_waitcnt vmcnt(0)
; GFX9-NEXT: v_or_b32_sdwa v0, v1, v0 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:BYTE_0
; GFX9-NEXT: v_add_u32_e32 v1, 3, v55
@@ -133722,7 +133756,7 @@ define inreg <16 x double> @bitcast_v128i8_to_v16f64_scalar(<128 x i8> inreg %a,
; GFX9-NEXT: v_add_u32_sdwa v1, v1, s4 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:DWORD
; GFX9-NEXT: v_or_b32_sdwa v27, v0, v1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_0 src1_sel:DWORD
; GFX9-NEXT: v_add_u32_e32 v0, 3, v43
-; GFX9-NEXT: v_add_u32_e32 v1, 3, v36
+; GFX9-NEXT: v_add_u32_e32 v1, 3, v42
; GFX9-NEXT: v_or_b32_sdwa v0, v51, v0 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:BYTE_0
; GFX9-NEXT: v_add_u32_e32 v0, 0x300, v0
; GFX9-NEXT: s_waitcnt vmcnt(0)
@@ -133731,7 +133765,7 @@ define inreg <16 x double> @bitcast_v128i8_to_v16f64_scalar(<128 x i8> inreg %a,
; GFX9-NEXT: v_or_b32_sdwa v28, v0, v1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_0 src1_sel:DWORD
; GFX9-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:420 ; 4-byte Folded Reload
; GFX9-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:416 ; 4-byte Folded Reload
-; GFX9-NEXT: v_add_u32_e32 v0, 3, v42
+; GFX9-NEXT: v_add_u32_e32 v0, 3, v36
; GFX9-NEXT: s_waitcnt vmcnt(1)
; GFX9-NEXT: v_or_b32_sdwa v0, v1, v0 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:BYTE_0
; GFX9-NEXT: v_add_u32_e32 v1, 3, v41
@@ -136664,19 +136698,20 @@ define inreg <64 x bfloat> @bitcast_v16f64_to_v64bf16_scalar(<16 x double> inreg
; SI-NEXT: v_writelane_b32 v62, s46, 3
; SI-NEXT: s_cbranch_execnz .LBB77_4
; SI-NEXT: .LBB77_2: ; %cmp.true
-; SI-NEXT: v_add_f64 v[19:20], s[12:13], 1.0
+; SI-NEXT: v_add_f64 v[11:12], s[8:9], 1.0
; SI-NEXT: v_add_f64 v[3:4], s[6:7], 1.0
-; SI-NEXT: v_add_f64 v[1:2], s[22:23], 1.0
+; SI-NEXT: v_add_f64 v[15:16], s[10:11], 1.0
; SI-NEXT: v_and_b32_e32 v9, 0xffff0000, v3
; SI-NEXT: v_lshlrev_b32_e32 v6, 16, v3
-; SI-NEXT: v_and_b32_e32 v3, 0xffff0000, v20
+; SI-NEXT: v_and_b32_e32 v3, 0xffff0000, v11
+; SI-NEXT: v_add_f64 v[1:2], s[22:23], 1.0
; SI-NEXT: v_add_f64 v[41:42], s[24:25], 1.0
; SI-NEXT: buffer_store_dword v3, off, s[0:3], s32 offset:56 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_and_b32_e32 v3, 0xffff0000, v19
+; SI-NEXT: v_and_b32_e32 v3, 0xffff0000, v16
; SI-NEXT: buffer_store_dword v3, off, s[0:3], s32 offset:64 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_lshlrev_b32_e32 v3, 16, v19
+; SI-NEXT: v_lshlrev_b32_e32 v3, 16, v16
; SI-NEXT: buffer_store_dword v3, off, s[0:3], s32 offset:60 ; 4-byte Folded Spill
; SI-NEXT: v_and_b32_e32 v54, 0xffff0000, v42
; SI-NEXT: v_lshlrev_b32_e32 v53, 16, v42
@@ -136705,8 +136740,7 @@ define inreg <64 x bfloat> @bitcast_v16f64_to_v64bf16_scalar(<16 x double> inreg
; SI-NEXT: v_add_f64 v[31:32], s[42:43], 1.0
; SI-NEXT: v_add_f64 v[27:28], s[40:41], 1.0
; SI-NEXT: v_add_f64 v[23:24], s[14:15], 1.0
-; SI-NEXT: v_add_f64 v[15:16], s[10:11], 1.0
-; SI-NEXT: v_add_f64 v[11:12], s[8:9], 1.0
+; SI-NEXT: v_add_f64 v[19:20], s[12:13], 1.0
; SI-NEXT: v_add_f64 v[7:8], s[4:5], 1.0
; SI-NEXT: v_add_f64 v[59:60], s[18:19], 1.0
; SI-NEXT: v_and_b32_e32 v13, 0xffff0000, v8
@@ -136715,27 +136749,27 @@ define inreg <64 x bfloat> @bitcast_v16f64_to_v64bf16_scalar(<16 x double> inreg
; SI-NEXT: v_lshlrev_b32_e32 v14, 16, v7
; SI-NEXT: v_and_b32_e32 v21, 0xffff0000, v12
; SI-NEXT: v_lshlrev_b32_e32 v18, 16, v12
-; SI-NEXT: v_and_b32_e32 v25, 0xffff0000, v11
; SI-NEXT: v_lshlrev_b32_e32 v22, 16, v11
-; SI-NEXT: v_and_b32_e32 v29, 0xffff0000, v16
-; SI-NEXT: v_lshlrev_b32_e32 v26, 16, v16
-; SI-NEXT: v_and_b32_e32 v33, 0xffff0000, v15
-; SI-NEXT: v_lshlrev_b32_e32 v30, 16, v15
-; SI-NEXT: v_lshlrev_b32_e32 v34, 16, v20
+; SI-NEXT: v_and_b32_e32 v16, 0xffff0000, v15
+; SI-NEXT: v_lshlrev_b32_e32 v15, 16, v15
+; SI-NEXT: v_and_b32_e32 v12, 0xffff0000, v20
+; SI-NEXT: v_lshlrev_b32_e32 v11, 16, v20
+; SI-NEXT: v_and_b32_e32 v20, 0xffff0000, v19
+; SI-NEXT: v_lshlrev_b32_e32 v19, 16, v19
; SI-NEXT: v_and_b32_e32 v8, 0xffff0000, v24
; SI-NEXT: v_lshlrev_b32_e32 v7, 16, v24
; SI-NEXT: v_and_b32_e32 v24, 0xffff0000, v23
; SI-NEXT: v_lshlrev_b32_e32 v23, 16, v23
-; SI-NEXT: v_and_b32_e32 v16, 0xffff0000, v28
-; SI-NEXT: v_lshlrev_b32_e32 v15, 16, v28
+; SI-NEXT: v_and_b32_e32 v26, 0xffff0000, v28
+; SI-NEXT: v_lshlrev_b32_e32 v25, 16, v28
; SI-NEXT: v_and_b32_e32 v28, 0xffff0000, v27
; SI-NEXT: v_lshlrev_b32_e32 v27, 16, v27
-; SI-NEXT: v_and_b32_e32 v12, 0xffff0000, v32
-; SI-NEXT: v_lshlrev_b32_e32 v11, 16, v32
+; SI-NEXT: v_and_b32_e32 v30, 0xffff0000, v32
+; SI-NEXT: v_lshlrev_b32_e32 v29, 16, v32
; SI-NEXT: v_and_b32_e32 v32, 0xffff0000, v31
; SI-NEXT: v_lshlrev_b32_e32 v31, 16, v31
-; SI-NEXT: v_and_b32_e32 v20, 0xffff0000, v36
-; SI-NEXT: v_lshlrev_b32_e32 v19, 16, v36
+; SI-NEXT: v_and_b32_e32 v34, 0xffff0000, v36
+; SI-NEXT: v_lshlrev_b32_e32 v33, 16, v36
; SI-NEXT: v_and_b32_e32 v36, 0xffff0000, v35
; SI-NEXT: v_lshlrev_b32_e32 v35, 16, v35
; SI-NEXT: v_and_b32_e32 v38, 0xffff0000, v50
@@ -136826,14 +136860,14 @@ define inreg <64 x bfloat> @bitcast_v16f64_to_v64bf16_scalar(<16 x double> inreg
; SI-NEXT: ; kill: killed $sgpr46
; SI-NEXT: s_branch .LBB77_2
; SI-NEXT: .LBB77_4:
-; SI-NEXT: v_mov_b32_e32 v1, s71
+; SI-NEXT: v_mov_b32_e32 v1, s85
; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:56 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_mov_b32_e32 v1, s69
+; SI-NEXT: v_mov_b32_e32 v1, s83
; SI-NEXT: v_readlane_b32 s4, v62, 0
; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:64 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_mov_b32_e32 v1, s68
+; SI-NEXT: v_mov_b32_e32 v1, s82
; SI-NEXT: v_mov_b32_e32 v61, s4
; SI-NEXT: v_readlane_b32 s4, v62, 1
; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:60 ; 4-byte Folded Spill
@@ -136852,27 +136886,27 @@ define inreg <64 x bfloat> @bitcast_v16f64_to_v64bf16_scalar(<16 x double> inreg
; SI-NEXT: v_mov_b32_e32 v14, s96
; SI-NEXT: v_mov_b32_e32 v21, s87
; SI-NEXT: v_mov_b32_e32 v18, s86
-; SI-NEXT: v_mov_b32_e32 v25, s85
; SI-NEXT: v_mov_b32_e32 v22, s84
-; SI-NEXT: v_mov_b32_e32 v29, s83
-; SI-NEXT: v_mov_b32_e32 v26, s82
-; SI-NEXT: v_mov_b32_e32 v33, s81
-; SI-NEXT: v_mov_b32_e32 v30, s80
-; SI-NEXT: v_mov_b32_e32 v34, s70
+; SI-NEXT: v_mov_b32_e32 v16, s81
+; SI-NEXT: v_mov_b32_e32 v15, s80
+; SI-NEXT: v_mov_b32_e32 v12, s71
+; SI-NEXT: v_mov_b32_e32 v11, s70
+; SI-NEXT: v_mov_b32_e32 v20, s69
+; SI-NEXT: v_mov_b32_e32 v19, s68
; SI-NEXT: v_mov_b32_e32 v8, s67
; SI-NEXT: v_mov_b32_e32 v7, s66
; SI-NEXT: v_mov_b32_e32 v24, s65
; SI-NEXT: v_mov_b32_e32 v23, s64
-; SI-NEXT: v_mov_b32_e32 v16, s55
-; SI-NEXT: v_mov_b32_e32 v15, s54
+; SI-NEXT: v_mov_b32_e32 v26, s55
+; SI-NEXT: v_mov_b32_e32 v25, s54
; SI-NEXT: v_mov_b32_e32 v28, s53
; SI-NEXT: v_mov_b32_e32 v27, s52
-; SI-NEXT: v_mov_b32_e32 v12, s51
-; SI-NEXT: v_mov_b32_e32 v11, s50
+; SI-NEXT: v_mov_b32_e32 v30, s51
+; SI-NEXT: v_mov_b32_e32 v29, s50
; SI-NEXT: v_mov_b32_e32 v32, s49
; SI-NEXT: v_mov_b32_e32 v31, s48
-; SI-NEXT: v_mov_b32_e32 v20, s39
-; SI-NEXT: v_mov_b32_e32 v19, s38
+; SI-NEXT: v_mov_b32_e32 v34, s39
+; SI-NEXT: v_mov_b32_e32 v33, s38
; SI-NEXT: v_mov_b32_e32 v36, s37
; SI-NEXT: v_mov_b32_e32 v35, s36
; SI-NEXT: v_mov_b32_e32 v38, s35
@@ -137005,9 +137039,9 @@ define inreg <64 x bfloat> @bitcast_v16f64_to_v64bf16_scalar(<16 x double> inreg
; SI-NEXT: v_add_i32_e32 v2, vcc, 56, v0
; SI-NEXT: buffer_store_dword v1, v2, s[0:3], 0 offen
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_mul_f32_e32 v1, 1.0, v20
+; SI-NEXT: v_mul_f32_e32 v1, 1.0, v34
; SI-NEXT: v_lshrrev_b32_e32 v1, 16, v1
-; SI-NEXT: v_mul_f32_e32 v2, 1.0, v19
+; SI-NEXT: v_mul_f32_e32 v2, 1.0, v33
; SI-NEXT: v_alignbit_b32 v1, v1, v2, 16
; SI-NEXT: v_add_i32_e32 v2, vcc, 60, v0
; SI-NEXT: buffer_store_dword v1, v2, s[0:3], 0 offen
@@ -137019,9 +137053,9 @@ define inreg <64 x bfloat> @bitcast_v16f64_to_v64bf16_scalar(<16 x double> inreg
; SI-NEXT: v_add_i32_e32 v2, vcc, 64, v0
; SI-NEXT: buffer_store_dword v1, v2, s[0:3], 0 offen
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_mul_f32_e32 v1, 1.0, v12
+; SI-NEXT: v_mul_f32_e32 v1, 1.0, v30
; SI-NEXT: v_lshrrev_b32_e32 v1, 16, v1
-; SI-NEXT: v_mul_f32_e32 v2, 1.0, v11
+; SI-NEXT: v_mul_f32_e32 v2, 1.0, v29
; SI-NEXT: v_alignbit_b32 v1, v1, v2, 16
; SI-NEXT: v_add_i32_e32 v2, vcc, 0x44, v0
; SI-NEXT: buffer_store_dword v1, v2, s[0:3], 0 offen
@@ -137033,9 +137067,9 @@ define inreg <64 x bfloat> @bitcast_v16f64_to_v64bf16_scalar(<16 x double> inreg
; SI-NEXT: v_add_i32_e32 v2, vcc, 0x48, v0
; SI-NEXT: buffer_store_dword v1, v2, s[0:3], 0 offen
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_mul_f32_e32 v1, 1.0, v16
+; SI-NEXT: v_mul_f32_e32 v1, 1.0, v26
; SI-NEXT: v_lshrrev_b32_e32 v1, 16, v1
-; SI-NEXT: v_mul_f32_e32 v2, 1.0, v15
+; SI-NEXT: v_mul_f32_e32 v2, 1.0, v25
; SI-NEXT: v_alignbit_b32 v1, v1, v2, 16
; SI-NEXT: v_add_i32_e32 v2, vcc, 0x4c, v0
; SI-NEXT: buffer_store_dword v1, v2, s[0:3], 0 offen
@@ -137054,6 +137088,27 @@ define inreg <64 x bfloat> @bitcast_v16f64_to_v64bf16_scalar(<16 x double> inreg
; SI-NEXT: v_add_i32_e32 v2, vcc, 0x54, v0
; SI-NEXT: buffer_store_dword v1, v2, s[0:3], 0 offen
; SI-NEXT: s_waitcnt expcnt(0)
+; SI-NEXT: v_mul_f32_e32 v1, 1.0, v20
+; SI-NEXT: v_lshrrev_b32_e32 v1, 16, v1
+; SI-NEXT: v_mul_f32_e32 v2, 1.0, v19
+; SI-NEXT: v_alignbit_b32 v1, v1, v2, 16
+; SI-NEXT: v_add_i32_e32 v2, vcc, 0x58, v0
+; SI-NEXT: buffer_store_dword v1, v2, s[0:3], 0 offen
+; SI-NEXT: s_waitcnt expcnt(0)
+; SI-NEXT: v_mul_f32_e32 v1, 1.0, v12
+; SI-NEXT: v_lshrrev_b32_e32 v1, 16, v1
+; SI-NEXT: v_mul_f32_e32 v2, 1.0, v11
+; SI-NEXT: v_alignbit_b32 v1, v1, v2, 16
+; SI-NEXT: v_add_i32_e32 v2, vcc, 0x5c, v0
+; SI-NEXT: buffer_store_dword v1, v2, s[0:3], 0 offen
+; SI-NEXT: s_waitcnt expcnt(0)
+; SI-NEXT: v_mul_f32_e32 v1, 1.0, v16
+; SI-NEXT: v_lshrrev_b32_e32 v1, 16, v1
+; SI-NEXT: v_mul_f32_e32 v2, 1.0, v15
+; SI-NEXT: v_alignbit_b32 v1, v1, v2, 16
+; SI-NEXT: v_add_i32_e32 v2, vcc, 0x60, v0
+; SI-NEXT: buffer_store_dword v1, v2, s[0:3], 0 offen
+; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:64 ; 4-byte Folded Reload
; SI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:60 ; 4-byte Folded Reload
; SI-NEXT: v_readlane_b32 s99, v63, 35
@@ -137098,36 +137153,15 @@ define inreg <64 x bfloat> @bitcast_v16f64_to_v64bf16_scalar(<16 x double> inreg
; SI-NEXT: s_waitcnt vmcnt(0)
; SI-NEXT: v_mul_f32_e32 v2, 1.0, v2
; SI-NEXT: v_alignbit_b32 v1, v1, v2, 16
-; SI-NEXT: v_add_i32_e32 v2, vcc, 0x58, v0
+; SI-NEXT: v_add_i32_e32 v2, vcc, 0x64, v0
; SI-NEXT: buffer_store_dword v1, v2, s[0:3], 0 offen
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:56 ; 4-byte Folded Reload
-; SI-NEXT: v_mul_f32_e32 v2, 1.0, v34
+; SI-NEXT: v_mul_f32_e32 v2, 1.0, v22
; SI-NEXT: s_waitcnt vmcnt(0)
; SI-NEXT: v_mul_f32_e32 v1, 1.0, v1
; SI-NEXT: v_lshrrev_b32_e32 v1, 16, v1
; SI-NEXT: v_alignbit_b32 v1, v1, v2, 16
-; SI-NEXT: v_add_i32_e32 v2, vcc, 0x5c, v0
-; SI-NEXT: buffer_store_dword v1, v2, s[0:3], 0 offen
-; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_mul_f32_e32 v1, 1.0, v33
-; SI-NEXT: v_lshrrev_b32_e32 v1, 16, v1
-; SI-NEXT: v_mul_f32_e32 v2, 1.0, v30
-; SI-NEXT: v_alignbit_b32 v1, v1, v2, 16
-; SI-NEXT: v_add_i32_e32 v2, vcc, 0x60, v0
-; SI-NEXT: buffer_store_dword v1, v2, s[0:3], 0 offen
-; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_mul_f32_e32 v1, 1.0, v29
-; SI-NEXT: v_lshrrev_b32_e32 v1, 16, v1
-; SI-NEXT: v_mul_f32_e32 v2, 1.0, v26
-; SI-NEXT: v_alignbit_b32 v1, v1, v2, 16
-; SI-NEXT: v_add_i32_e32 v2, vcc, 0x64, v0
-; SI-NEXT: buffer_store_dword v1, v2, s[0:3], 0 offen
-; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_mul_f32_e32 v1, 1.0, v25
-; SI-NEXT: v_lshrrev_b32_e32 v1, 16, v1
-; SI-NEXT: v_mul_f32_e32 v2, 1.0, v22
-; SI-NEXT: v_alignbit_b32 v1, v1, v2, 16
; SI-NEXT: v_add_i32_e32 v2, vcc, 0x68, v0
; SI-NEXT: buffer_store_dword v1, v2, s[0:3], 0 offen
; SI-NEXT: s_waitcnt expcnt(0)
@@ -140441,10 +140475,10 @@ define inreg <16 x double> @bitcast_v64bf16_to_v16f64_scalar(<64 x bfloat> inreg
; SI-NEXT: v_mul_f32_e32 v34, 1.0, v36
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_mul_f32_e64 v0, 1.0, s17
-; SI-NEXT: v_mul_f32_e64 v35, 1.0, s18
+; SI-NEXT: v_mul_f32_e64 v33, 1.0, s18
; SI-NEXT: v_mul_f32_e64 v36, 1.0, s21
; SI-NEXT: v_mul_f32_e64 v42, 1.0, s20
-; SI-NEXT: v_mul_f32_e64 v33, 1.0, s22
+; SI-NEXT: v_mul_f32_e64 v35, 1.0, s22
; SI-NEXT: buffer_store_dword v30, off, s[0:3], s32 offset:196 ; 4-byte Folded Spill
; SI-NEXT: buffer_store_dword v31, off, s[0:3], s32 offset:200 ; 4-byte Folded Spill
; SI-NEXT: buffer_store_dword v29, off, s[0:3], s32 offset:204 ; 4-byte Folded Spill
@@ -140473,9 +140507,9 @@ define inreg <16 x double> @bitcast_v64bf16_to_v16f64_scalar(<64 x bfloat> inreg
; SI-NEXT: buffer_store_dword v9, off, s[0:3], s32 offset:296 ; 4-byte Folded Spill
; SI-NEXT: buffer_store_dword v3, off, s[0:3], s32 offset:300 ; 4-byte Folded Spill
; SI-NEXT: buffer_store_dword v0, off, s[0:3], s32 offset:304 ; 4-byte Folded Spill
-; SI-NEXT: buffer_store_dword v33, off, s[0:3], s32 offset:308 ; 4-byte Folded Spill
-; SI-NEXT: buffer_store_dword v35, off, s[0:3], s32 offset:312 ; 4-byte Folded Spill
-; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:316 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:308 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v33, off, s[0:3], s32 offset:312 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v35, off, s[0:3], s32 offset:316 ; 4-byte Folded Spill
; SI-NEXT: buffer_store_dword v36, off, s[0:3], s32 offset:320 ; 4-byte Folded Spill
; SI-NEXT: buffer_store_dword v42, off, s[0:3], s32 offset:324 ; 4-byte Folded Spill
; SI-NEXT: s_cbranch_scc0 .LBB79_4
@@ -140486,12 +140520,13 @@ define inreg <16 x double> @bitcast_v64bf16_to_v16f64_scalar(<64 x bfloat> inreg
; SI-NEXT: buffer_load_dword v7, off, s[0:3], s32 offset:192 ; 4-byte Folded Reload
; SI-NEXT: v_alignbit_b32 v5, v5, v8, 16
; SI-NEXT: buffer_load_dword v8, off, s[0:3], s32 offset:184 ; 4-byte Folded Reload
-; SI-NEXT: s_waitcnt expcnt(2)
-; SI-NEXT: v_lshrrev_b32_e32 v1, 16, v1
-; SI-NEXT: v_alignbit_b32 v1, v1, v35, 16
+; SI-NEXT: s_waitcnt expcnt(6)
+; SI-NEXT: v_lshrrev_b32_e32 v3, 16, v3
+; SI-NEXT: v_alignbit_b32 v3, v3, v35, 16
; SI-NEXT: v_lshrrev_b32_e32 v4, 16, v4
; SI-NEXT: v_alignbit_b32 v4, v4, v9, 16
; SI-NEXT: buffer_load_dword v9, off, s[0:3], s32 offset:180 ; 4-byte Folded Reload
+; SI-NEXT: s_waitcnt expcnt(5)
; SI-NEXT: v_lshrrev_b32_e32 v0, 16, v0
; SI-NEXT: v_mov_b32_e32 v59, v2
; SI-NEXT: v_alignbit_b32 v0, v0, v2, 16
@@ -140501,10 +140536,11 @@ define inreg <16 x double> @bitcast_v64bf16_to_v16f64_scalar(<64 x bfloat> inreg
; SI-NEXT: v_mov_b32_e32 v47, v10
; SI-NEXT: v_lshrrev_b32_e32 v10, 16, v10
; SI-NEXT: v_mov_b32_e32 v45, v12
-; SI-NEXT: v_lshrrev_b32_e32 v3, 16, v3
-; SI-NEXT: v_alignbit_b32 v3, v3, v33, 16
-; SI-NEXT: v_mov_b32_e32 v33, v14
+; SI-NEXT: v_mov_b32_e32 v44, v14
+; SI-NEXT: s_waitcnt expcnt(4)
+; SI-NEXT: v_lshrrev_b32_e32 v1, 16, v1
; SI-NEXT: v_lshrrev_b32_e32 v13, 16, v13
+; SI-NEXT: v_alignbit_b32 v1, v1, v33, 16
; SI-NEXT: v_mov_b32_e32 v62, v38
; SI-NEXT: v_lshrrev_b32_e32 v15, 16, v15
; SI-NEXT: v_lshrrev_b32_e32 v16, 16, v16
@@ -140544,20 +140580,20 @@ define inreg <16 x double> @bitcast_v64bf16_to_v16f64_scalar(<64 x bfloat> inreg
; SI-NEXT: v_alignbit_b32 v28, v28, v37, 16
; SI-NEXT: v_mov_b32_e32 v37, v34
; SI-NEXT: buffer_load_dword v32, off, s[0:3], s32 offset:148 ; 4-byte Folded Reload
-; SI-NEXT: s_waitcnt vmcnt(3)
+; SI-NEXT: s_waitcnt vmcnt(3) expcnt(2)
; SI-NEXT: v_mov_b32_e32 v35, v7
; SI-NEXT: v_lshrrev_b32_e32 v7, 16, v7
; SI-NEXT: s_waitcnt vmcnt(2)
; SI-NEXT: v_mov_b32_e32 v43, v8
; SI-NEXT: v_alignbit_b32 v7, v7, v8, 16
; SI-NEXT: buffer_load_dword v8, off, s[0:3], s32 offset:188 ; 4-byte Folded Reload
-; SI-NEXT: s_waitcnt vmcnt(2) expcnt(0)
-; SI-NEXT: v_mov_b32_e32 v42, v9
+; SI-NEXT: s_waitcnt vmcnt(2)
+; SI-NEXT: v_mov_b32_e32 v60, v9
; SI-NEXT: s_waitcnt vmcnt(1)
; SI-NEXT: v_lshrrev_b32_e32 v31, 16, v32
; SI-NEXT: v_alignbit_b32 v31, v31, v34, 16
-; SI-NEXT: s_waitcnt vmcnt(0)
-; SI-NEXT: v_mov_b32_e32 v60, v8
+; SI-NEXT: s_waitcnt vmcnt(0) expcnt(0)
+; SI-NEXT: v_mov_b32_e32 v42, v8
; SI-NEXT: v_lshrrev_b32_e32 v8, 16, v8
; SI-NEXT: v_alignbit_b32 v8, v8, v9, 16
; SI-NEXT: v_lshrrev_b32_e32 v9, 16, v11
@@ -140581,7 +140617,7 @@ define inreg <16 x double> @bitcast_v64bf16_to_v16f64_scalar(<64 x bfloat> inreg
; SI-NEXT: v_alignbit_b32 v12, v12, v14, 16
; SI-NEXT: buffer_load_dword v14, off, s[0:3], s32 offset:160 ; 4-byte Folded Reload
; SI-NEXT: s_waitcnt vmcnt(0)
-; SI-NEXT: v_mov_b32_e32 v44, v14
+; SI-NEXT: v_mov_b32_e32 v33, v14
; SI-NEXT: v_alignbit_b32 v13, v13, v14, 16
; SI-NEXT: buffer_load_dword v14, off, s[0:3], s32 offset:152 ; 4-byte Folded Reload
; SI-NEXT: s_waitcnt vmcnt(0)
@@ -140604,7 +140640,7 @@ define inreg <16 x double> @bitcast_v64bf16_to_v16f64_scalar(<64 x bfloat> inreg
; SI-NEXT: .LBB79_2: ; %cmp.true
; SI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:304 ; 4-byte Folded Reload
; SI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:312 ; 4-byte Folded Reload
-; SI-NEXT: buffer_load_dword v3, off, s[0:3], s32 offset:316 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v3, off, s[0:3], s32 offset:308 ; 4-byte Folded Reload
; SI-NEXT: buffer_load_dword v4, off, s[0:3], s32 offset:300 ; 4-byte Folded Reload
; SI-NEXT: v_and_b32_e32 v0, 0xffff0000, v59
; SI-NEXT: v_add_f32_e32 v0, 0x40c00000, v0
@@ -140620,7 +140656,7 @@ define inreg <16 x double> @bitcast_v64bf16_to_v16f64_scalar(<64 x bfloat> inreg
; SI-NEXT: v_and_b32_e32 v8, 0xffff0000, v35
; SI-NEXT: v_add_f32_e32 v8, 0x40c00000, v8
; SI-NEXT: v_lshrrev_b32_e32 v8, 16, v8
-; SI-NEXT: v_and_b32_e32 v9, 0xffff0000, v60
+; SI-NEXT: v_and_b32_e32 v9, 0xffff0000, v42
; SI-NEXT: v_add_f32_e32 v9, 0x40c00000, v9
; SI-NEXT: v_lshrrev_b32_e32 v9, 16, v9
; SI-NEXT: v_and_b32_e32 v10, 0xffff0000, v57
@@ -140632,7 +140668,7 @@ define inreg <16 x double> @bitcast_v64bf16_to_v16f64_scalar(<64 x bfloat> inreg
; SI-NEXT: v_and_b32_e32 v12, 0xffff0000, v45
; SI-NEXT: v_add_f32_e32 v12, 0x40c00000, v12
; SI-NEXT: v_lshrrev_b32_e32 v12, 16, v12
-; SI-NEXT: v_and_b32_e32 v13, 0xffff0000, v33
+; SI-NEXT: v_and_b32_e32 v13, 0xffff0000, v44
; SI-NEXT: v_add_f32_e32 v13, 0x40c00000, v13
; SI-NEXT: v_lshrrev_b32_e32 v13, 16, v13
; SI-NEXT: v_and_b32_e32 v15, 0xffff0000, v36
@@ -140744,7 +140780,7 @@ define inreg <16 x double> @bitcast_v64bf16_to_v16f64_scalar(<64 x bfloat> inreg
; SI-NEXT: v_add_f32_e32 v2, 0x40c00000, v2
; SI-NEXT: v_lshrrev_b32_e32 v3, 16, v3
; SI-NEXT: v_alignbit_b32 v2, v3, v2, 16
-; SI-NEXT: buffer_load_dword v3, off, s[0:3], s32 offset:308 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v3, off, s[0:3], s32 offset:316 ; 4-byte Folded Reload
; SI-NEXT: s_waitcnt vmcnt(0)
; SI-NEXT: v_and_b32_e32 v3, 0xffff0000, v3
; SI-NEXT: v_add_f32_e32 v3, 0x40c00000, v3
@@ -140767,7 +140803,7 @@ define inreg <16 x double> @bitcast_v64bf16_to_v16f64_scalar(<64 x bfloat> inreg
; SI-NEXT: v_and_b32_e32 v7, 0xffff0000, v43
; SI-NEXT: v_add_f32_e32 v7, 0x40c00000, v7
; SI-NEXT: v_alignbit_b32 v7, v8, v7, 16
-; SI-NEXT: v_and_b32_e32 v8, 0xffff0000, v42
+; SI-NEXT: v_and_b32_e32 v8, 0xffff0000, v60
; SI-NEXT: v_add_f32_e32 v8, 0x40c00000, v8
; SI-NEXT: v_alignbit_b32 v8, v9, v8, 16
; SI-NEXT: v_and_b32_e32 v9, 0xffff0000, v58
@@ -140782,7 +140818,7 @@ define inreg <16 x double> @bitcast_v64bf16_to_v16f64_scalar(<64 x bfloat> inreg
; SI-NEXT: v_and_b32_e32 v12, 0xffff0000, v63
; SI-NEXT: v_add_f32_e32 v12, 0x40c00000, v12
; SI-NEXT: v_alignbit_b32 v12, v13, v12, 16
-; SI-NEXT: v_and_b32_e32 v13, 0xffff0000, v44
+; SI-NEXT: v_and_b32_e32 v13, 0xffff0000, v33
; SI-NEXT: v_add_f32_e32 v13, 0x40c00000, v13
; SI-NEXT: v_alignbit_b32 v13, v14, v13, 16
; SI-NEXT: v_and_b32_e32 v14, 0xffff0000, v62
@@ -140868,16 +140904,16 @@ define inreg <16 x double> @bitcast_v64bf16_to_v16f64_scalar(<64 x bfloat> inreg
; SI-NEXT: s_setpc_b64 s[30:31]
; SI-NEXT: .LBB79_4:
; SI-NEXT: buffer_load_dword v43, off, s[0:3], s32 offset:184 ; 4-byte Folded Reload
-; SI-NEXT: s_waitcnt expcnt(3)
+; SI-NEXT: s_waitcnt expcnt(2)
; SI-NEXT: buffer_load_dword v35, off, s[0:3], s32 offset:192 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v60, off, s[0:3], s32 offset:180 ; 4-byte Folded Reload
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: buffer_load_dword v42, off, s[0:3], s32 offset:180 ; 4-byte Folded Reload
-; SI-NEXT: buffer_load_dword v60, off, s[0:3], s32 offset:188 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v42, off, s[0:3], s32 offset:188 ; 4-byte Folded Reload
; SI-NEXT: buffer_load_dword v58, off, s[0:3], s32 offset:176 ; 4-byte Folded Reload
; SI-NEXT: buffer_load_dword v56, off, s[0:3], s32 offset:172 ; 4-byte Folded Reload
; SI-NEXT: buffer_load_dword v46, off, s[0:3], s32 offset:168 ; 4-byte Folded Reload
; SI-NEXT: buffer_load_dword v63, off, s[0:3], s32 offset:164 ; 4-byte Folded Reload
-; SI-NEXT: buffer_load_dword v44, off, s[0:3], s32 offset:160 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v33, off, s[0:3], s32 offset:160 ; 4-byte Folded Reload
; SI-NEXT: v_mov_b32_e32 v61, v53
; SI-NEXT: buffer_load_dword v36, off, s[0:3], s32 offset:152 ; 4-byte Folded Reload
; SI-NEXT: buffer_load_dword v53, off, s[0:3], s32 offset:156 ; 4-byte Folded Reload
@@ -140886,7 +140922,7 @@ define inreg <16 x double> @bitcast_v64bf16_to_v16f64_scalar(<64 x bfloat> inreg
; SI-NEXT: v_mov_b32_e32 v57, v11
; SI-NEXT: v_mov_b32_e32 v47, v10
; SI-NEXT: v_mov_b32_e32 v45, v12
-; SI-NEXT: v_mov_b32_e32 v33, v14
+; SI-NEXT: v_mov_b32_e32 v44, v14
; SI-NEXT: v_mov_b32_e32 v62, v38
; SI-NEXT: v_mov_b32_e32 v38, v39
; SI-NEXT: v_mov_b32_e32 v39, v41
@@ -144853,70 +144889,68 @@ define inreg <64 x half> @bitcast_v16f64_to_v64f16_scalar(<16 x double> inreg %a
; SI-NEXT: s_lshr_b32 s46, s5, 16
; SI-NEXT: v_cvt_f32_f16_e32 v10, s46
; SI-NEXT: s_lshr_b32 s46, s4, 16
-; SI-NEXT: v_cvt_f32_f16_e32 v55, s46
+; SI-NEXT: v_cvt_f32_f16_e32 v41, s46
; SI-NEXT: s_lshr_b32 s46, s7, 16
-; SI-NEXT: s_waitcnt expcnt(4)
-; SI-NEXT: v_cvt_f32_f16_e32 v59, s46
-; SI-NEXT: s_lshr_b32 s46, s6, 16
; SI-NEXT: v_cvt_f32_f16_e32 v8, s46
-; SI-NEXT: s_lshr_b32 s46, s9, 16
+; SI-NEXT: s_lshr_b32 s46, s6, 16
; SI-NEXT: v_cvt_f32_f16_e32 v9, s46
-; SI-NEXT: s_lshr_b32 s46, s8, 16
+; SI-NEXT: s_lshr_b32 s46, s9, 16
; SI-NEXT: v_cvt_f32_f16_e32 v16, s46
-; SI-NEXT: s_lshr_b32 s46, s11, 16
+; SI-NEXT: s_lshr_b32 s46, s8, 16
; SI-NEXT: v_cvt_f32_f16_e32 v23, s46
-; SI-NEXT: s_lshr_b32 s46, s10, 16
+; SI-NEXT: s_lshr_b32 s46, s11, 16
; SI-NEXT: v_cvt_f32_f16_e32 v27, s46
-; SI-NEXT: s_lshr_b32 s46, s13, 16
+; SI-NEXT: s_lshr_b32 s46, s10, 16
; SI-NEXT: v_cvt_f32_f16_e32 v31, s46
-; SI-NEXT: s_lshr_b32 s46, s12, 16
+; SI-NEXT: s_lshr_b32 s46, s13, 16
; SI-NEXT: v_cvt_f32_f16_e32 v38, s46
-; SI-NEXT: s_lshr_b32 s46, s15, 16
+; SI-NEXT: s_lshr_b32 s46, s12, 16
; SI-NEXT: v_cvt_f32_f16_e32 v50, s46
-; SI-NEXT: s_lshr_b32 s46, s14, 16
+; SI-NEXT: s_lshr_b32 s46, s15, 16
; SI-NEXT: v_cvt_f32_f16_e32 v54, s46
-; SI-NEXT: s_lshr_b32 s46, s41, 16
+; SI-NEXT: s_lshr_b32 s46, s14, 16
; SI-NEXT: v_cvt_f32_f16_e32 v40, s46
-; SI-NEXT: s_lshr_b32 s46, s40, 16
+; SI-NEXT: s_lshr_b32 s46, s41, 16
; SI-NEXT: v_cvt_f32_f16_e32 v42, s46
+; SI-NEXT: s_lshr_b32 s46, s40, 16
+; SI-NEXT: v_cvt_f32_f16_e32 v56, s46
; SI-NEXT: s_lshr_b32 s46, s43, 16
-; SI-NEXT: s_waitcnt expcnt(2)
-; SI-NEXT: v_cvt_f32_f16_e32 v61, s46
-; SI-NEXT: s_lshr_b32 s46, s42, 16
; SI-NEXT: v_cvt_f32_f16_e32 v12, s46
+; SI-NEXT: s_lshr_b32 s46, s42, 16
+; SI-NEXT: v_cvt_f32_f16_e32 v35, s46
; SI-NEXT: s_lshr_b32 s46, s45, 16
; SI-NEXT: v_cvt_f32_f16_e32 v32, s46
; SI-NEXT: s_lshr_b32 s46, s44, 16
; SI-NEXT: v_cvt_f32_f16_e32 v36, s46
; SI-NEXT: s_lshr_b32 s46, s29, 16
-; SI-NEXT: v_cvt_f32_f16_e32 v28, s46
+; SI-NEXT: v_cvt_f32_f16_e32 v2, s46
; SI-NEXT: s_lshr_b32 s46, s28, 16
; SI-NEXT: v_cvt_f32_f16_e32 v48, s46
; SI-NEXT: s_lshr_b32 s46, s27, 16
-; SI-NEXT: v_cvt_f32_f16_e32 v1, s42
+; SI-NEXT: v_cvt_f32_f16_e32 v1, s7
; SI-NEXT: v_cvt_f32_f16_e32 v24, s46
; SI-NEXT: s_lshr_b32 s46, s26, 16
; SI-NEXT: v_cvt_f32_f16_e32 v52, s46
; SI-NEXT: s_lshr_b32 s46, s25, 16
-; SI-NEXT: v_cvt_f32_f16_e32 v13, s46
+; SI-NEXT: v_cvt_f32_f16_e32 v28, s46
; SI-NEXT: s_lshr_b32 s46, s24, 16
-; SI-NEXT: v_cvt_f32_f16_e32 v20, s46
+; SI-NEXT: v_cvt_f32_f16_e32 v17, s46
; SI-NEXT: s_lshr_b32 s46, s23, 16
; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:64 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_cvt_f32_f16_e32 v1, s45
-; SI-NEXT: v_cvt_f32_f16_e32 v17, s46
+; SI-NEXT: v_cvt_f32_f16_e32 v20, s46
; SI-NEXT: s_lshr_b32 s46, s22, 16
-; SI-NEXT: v_cvt_f32_f16_e32 v35, s46
+; SI-NEXT: v_cvt_f32_f16_e32 v13, s46
; SI-NEXT: s_lshr_b32 s46, s21, 16
; SI-NEXT: v_cvt_f32_f16_e32 v46, s46
; SI-NEXT: s_lshr_b32 s46, s20, 16
-; SI-NEXT: v_cvt_f32_f16_e32 v5, s46
+; SI-NEXT: v_cvt_f32_f16_e32 v45, s46
; SI-NEXT: s_lshr_b32 s46, s19, 16
; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:68 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_cvt_f32_f16_e32 v1, s44
-; SI-NEXT: v_cvt_f32_f16_e32 v57, s46
+; SI-NEXT: v_cvt_f32_f16_e32 v14, s46
; SI-NEXT: s_lshr_b32 s46, s18, 16
; SI-NEXT: v_cvt_f32_f16_e32 v58, s46
; SI-NEXT: s_lshr_b32 s46, s17, 16
@@ -144924,8 +144958,7 @@ define inreg <64 x half> @bitcast_v16f64_to_v64f16_scalar(<16 x double> inreg %a
; SI-NEXT: s_lshr_b32 s46, s16, 16
; SI-NEXT: v_cvt_f32_f16_e32 v62, s46
; SI-NEXT: v_cvt_f32_f16_e32 v7, s5
-; SI-NEXT: v_cvt_f32_f16_e32 v56, s4
-; SI-NEXT: v_cvt_f32_f16_e32 v14, s7
+; SI-NEXT: v_cvt_f32_f16_e32 v5, s4
; SI-NEXT: v_cvt_f32_f16_e32 v15, s6
; SI-NEXT: v_cvt_f32_f16_e32 v6, s9
; SI-NEXT: v_cvt_f32_f16_e32 v18, s8
@@ -144933,11 +144966,12 @@ define inreg <64 x half> @bitcast_v16f64_to_v64f16_scalar(<16 x double> inreg %a
; SI-NEXT: v_cvt_f32_f16_e32 v19, s10
; SI-NEXT: v_cvt_f32_f16_e32 v39, s13
; SI-NEXT: v_cvt_f32_f16_e32 v51, s12
-; SI-NEXT: v_cvt_f32_f16_e32 v41, s15
+; SI-NEXT: v_cvt_f32_f16_e32 v55, s15
; SI-NEXT: v_cvt_f32_f16_e32 v43, s14
-; SI-NEXT: v_cvt_f32_f16_e32 v45, s41
-; SI-NEXT: v_cvt_f32_f16_e32 v44, s40
-; SI-NEXT: v_cvt_f32_f16_e32 v63, s43
+; SI-NEXT: v_cvt_f32_f16_e32 v44, s41
+; SI-NEXT: v_cvt_f32_f16_e32 v57, s40
+; SI-NEXT: v_cvt_f32_f16_e32 v59, s43
+; SI-NEXT: v_cvt_f32_f16_e32 v61, s42
; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:72 ; 4-byte Folded Spill
; SI-NEXT: v_cvt_f32_f16_e32 v37, s29
; SI-NEXT: v_cvt_f32_f16_e32 v22, s28
@@ -144951,7 +144985,7 @@ define inreg <64 x half> @bitcast_v16f64_to_v64f16_scalar(<16 x double> inreg %a
; SI-NEXT: v_cvt_f32_f16_e32 v47, s20
; SI-NEXT: v_cvt_f32_f16_e32 v34, s19
; SI-NEXT: v_cvt_f32_f16_e32 v25, s18
-; SI-NEXT: v_cvt_f32_f16_e32 v2, s17
+; SI-NEXT: v_cvt_f32_f16_e32 v63, s17
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_cvt_f32_f16_e32 v1, s16
; SI-NEXT: s_cbranch_execnz .LBB81_3
@@ -144976,143 +145010,146 @@ define inreg <64 x half> @bitcast_v16f64_to_v64f16_scalar(<16 x double> inreg %a
; SI-NEXT: buffer_store_dword v3, off, s[0:3], s32 offset:96 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_lshrrev_b32_e32 v3, 16, v49
+; SI-NEXT: v_add_f64 v[14:15], s[10:11], 1.0
; SI-NEXT: buffer_store_dword v3, off, s[0:3], s32 offset:100 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_lshrrev_b32_e32 v3, 16, v36
-; SI-NEXT: v_add_f64 v[14:15], s[10:11], 1.0
; SI-NEXT: buffer_store_dword v3, off, s[0:3], s32 offset:104 ; 4-byte Folded Spill
-; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_lshrrev_b32_e32 v3, 16, v37
-; SI-NEXT: buffer_store_dword v3, off, s[0:3], s32 offset:108 ; 4-byte Folded Spill
; SI-NEXT: v_lshrrev_b32_e32 v13, 16, v15
-; SI-NEXT: v_cvt_f32_f16_e32 v23, v13
-; SI-NEXT: buffer_load_dword v13, off, s[0:3], s32 offset:108 ; 4-byte Folded Reload
-; SI-NEXT: v_add_f64 v[6:7], s[6:7], 1.0
-; SI-NEXT: v_add_f64 v[4:5], s[4:5], 1.0
-; SI-NEXT: v_lshrrev_b32_e32 v28, 16, v7
-; SI-NEXT: v_cvt_f32_f16_e32 v59, v28
-; SI-NEXT: v_lshrrev_b32_e32 v32, 16, v4
-; SI-NEXT: v_cvt_f32_f16_e32 v56, v4
-; SI-NEXT: v_cvt_f32_f16_e32 v4, v14
+; SI-NEXT: v_cvt_f32_f16_e32 v27, v13
+; SI-NEXT: buffer_load_dword v13, off, s[0:3], s32 offset:104 ; 4-byte Folded Reload
; SI-NEXT: v_add_f64 v[29:30], s[42:43], 1.0
-; SI-NEXT: v_lshrrev_b32_e32 v24, 16, v6
-; SI-NEXT: v_lshrrev_b32_e32 v57, 16, v29
-; SI-NEXT: buffer_store_dword v4, off, s[0:3], s32 offset:88 ; 4-byte Folded Spill
-; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_cvt_f32_f16_e32 v4, v29
+; SI-NEXT: v_add_f64 v[6:7], s[6:7], 1.0
+; SI-NEXT: v_lshrrev_b32_e32 v58, 16, v29
+; SI-NEXT: v_cvt_f32_f16_e32 v61, v29
; SI-NEXT: v_cvt_f32_f16_e32 v29, v48
-; SI-NEXT: v_cvt_f32_f16_e32 v8, v24
-; SI-NEXT: v_add_f64 v[10:11], s[8:9], 1.0
+; SI-NEXT: v_lshrrev_b32_e32 v24, 16, v6
+; SI-NEXT: v_cvt_f32_f16_e32 v9, v24
; SI-NEXT: v_add_f64 v[25:26], s[40:41], 1.0
-; SI-NEXT: v_lshrrev_b32_e32 v17, 16, v10
-; SI-NEXT: v_lshrrev_b32_e32 v40, 16, v26
-; SI-NEXT: v_cvt_f32_f16_e32 v45, v26
+; SI-NEXT: v_add_f64 v[33:34], s[44:45], 1.0
+; SI-NEXT: v_lshrrev_b32_e32 v42, 16, v26
+; SI-NEXT: v_cvt_f32_f16_e32 v44, v26
; SI-NEXT: v_cvt_f32_f16_e32 v26, v52
-; SI-NEXT: v_cvt_f32_f16_e32 v16, v17
+; SI-NEXT: s_waitcnt expcnt(0)
+; SI-NEXT: v_lshrrev_b32_e32 v3, 16, v37
+; SI-NEXT: buffer_store_dword v3, off, s[0:3], s32 offset:108 ; 4-byte Folded Spill
+; SI-NEXT: s_waitcnt expcnt(0)
+; SI-NEXT: v_lshrrev_b32_e32 v3, 16, v33
+; SI-NEXT: buffer_store_dword v3, off, s[0:3], s32 offset:112 ; 4-byte Folded Spill
+; SI-NEXT: v_add_f64 v[4:5], s[4:5], 1.0
+; SI-NEXT: s_waitcnt expcnt(0)
+; SI-NEXT: v_cvt_f32_f16_e32 v3, v7
+; SI-NEXT: v_lshrrev_b32_e32 v32, 16, v4
+; SI-NEXT: v_lshrrev_b32_e32 v35, 16, v5
+; SI-NEXT: v_cvt_f32_f16_e32 v57, v5
+; SI-NEXT: v_cvt_f32_f16_e32 v5, v4
+; SI-NEXT: v_cvt_f32_f16_e32 v4, v34
; SI-NEXT: v_add_f64 v[21:22], s[14:15], 1.0
-; SI-NEXT: v_add_f64 v[33:34], s[44:45], 1.0
+; SI-NEXT: buffer_store_dword v3, off, s[0:3], s32 offset:64 ; 4-byte Folded Spill
+; SI-NEXT: s_waitcnt expcnt(0)
+; SI-NEXT: v_cvt_f32_f16_e32 v3, v14
; SI-NEXT: v_lshrrev_b32_e32 v62, 16, v30
-; SI-NEXT: v_lshrrev_b32_e32 v54, 16, v21
+; SI-NEXT: v_lshrrev_b32_e32 v40, 16, v21
; SI-NEXT: v_add_f64 v[18:19], s[12:13], 1.0
; SI-NEXT: v_cvt_f32_f16_e32 v43, v21
-; SI-NEXT: v_cvt_f32_f16_e32 v63, v30
+; SI-NEXT: v_cvt_f32_f16_e32 v59, v30
+; SI-NEXT: buffer_store_dword v4, off, s[0:3], s32 offset:68 ; 4-byte Folded Spill
+; SI-NEXT: s_waitcnt expcnt(0)
+; SI-NEXT: v_cvt_f32_f16_e32 v4, v33
; SI-NEXT: v_cvt_f32_f16_e32 v30, v2
; SI-NEXT: v_cvt_f32_f16_e32 v21, v1
; SI-NEXT: v_add_f64 v[1:2], s[18:19], 1.0
-; SI-NEXT: v_cvt_f32_f16_e32 v61, v5
-; SI-NEXT: v_cvt_f32_f16_e32 v62, v62
; SI-NEXT: v_lshrrev_b32_e32 v60, 16, v34
-; SI-NEXT: v_lshrrev_b32_e32 v42, 16, v25
-; SI-NEXT: v_lshrrev_b32_e32 v38, 16, v18
-; SI-NEXT: v_cvt_f32_f16_e32 v12, v7
+; SI-NEXT: v_lshrrev_b32_e32 v56, 16, v25
+; SI-NEXT: v_lshrrev_b32_e32 v50, 16, v18
+; SI-NEXT: v_lshrrev_b32_e32 v28, 16, v7
+; SI-NEXT: buffer_store_dword v3, off, s[0:3], s32 offset:88 ; 4-byte Folded Spill
; SI-NEXT: v_cvt_f32_f16_e32 v51, v18
-; SI-NEXT: v_cvt_f32_f16_e32 v44, v25
-; SI-NEXT: buffer_store_dword v4, off, s[0:3], s32 offset:64 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_cvt_f32_f16_e32 v4, v34
+; SI-NEXT: v_cvt_f32_f16_e32 v3, v25
; SI-NEXT: v_lshrrev_b32_e32 v7, 16, v1
; SI-NEXT: v_lshrrev_b32_e32 v18, 16, v2
; SI-NEXT: v_cvt_f32_f16_e32 v34, v2
; SI-NEXT: v_cvt_f32_f16_e32 v25, v1
; SI-NEXT: v_add_f64 v[1:2], s[16:17], 1.0
-; SI-NEXT: v_lshrrev_b32_e32 v58, 16, v33
-; SI-NEXT: v_lshrrev_b32_e32 v31, 16, v19
-; SI-NEXT: v_lshrrev_b32_e32 v20, 16, v11
-; SI-NEXT: v_lshrrev_b32_e32 v35, 16, v5
-; SI-NEXT: v_cvt_f32_f16_e32 v39, v19
-; SI-NEXT: v_lshrrev_b32_e32 v19, 16, v1
-; SI-NEXT: v_lshrrev_b32_e32 v50, 16, v22
-; SI-NEXT: v_cvt_f32_f16_e32 v3, v10
-; SI-NEXT: v_cvt_f32_f16_e32 v41, v22
-; SI-NEXT: v_cvt_f32_f16_e32 v22, v36
-; SI-NEXT: v_cvt_f32_f16_e32 v10, v35
-; SI-NEXT: v_cvt_f32_f16_e32 v9, v20
-; SI-NEXT: v_cvt_f32_f16_e32 v36, v58
-; SI-NEXT: v_cvt_f32_f16_e32 v58, v7
-; SI-NEXT: v_mov_b32_e32 v7, v61
-; SI-NEXT: v_mov_b32_e32 v61, v62
-; SI-NEXT: v_cvt_f32_f16_e32 v62, v19
-; SI-NEXT: buffer_store_dword v4, off, s[0:3], s32 offset:68 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v4, off, s[0:3], s32 offset:72 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_cvt_f32_f16_e32 v4, v33
-; SI-NEXT: v_cvt_f32_f16_e32 v5, v6
-; SI-NEXT: v_cvt_f32_f16_e32 v57, v57
+; SI-NEXT: v_lshrrev_b32_e32 v4, 16, v2
+; SI-NEXT: v_cvt_f32_f16_e32 v63, v2
+; SI-NEXT: v_cvt_f32_f16_e32 v8, v28
+; SI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:112 ; 4-byte Folded Reload
+; SI-NEXT: v_add_f64 v[10:11], s[8:9], 1.0
+; SI-NEXT: v_lshrrev_b32_e32 v54, 16, v22
+; SI-NEXT: v_lshrrev_b32_e32 v17, 16, v10
+; SI-NEXT: v_cvt_f32_f16_e32 v55, v22
+; SI-NEXT: v_cvt_f32_f16_e32 v22, v36
+; SI-NEXT: v_cvt_f32_f16_e32 v23, v17
+; SI-NEXT: v_lshrrev_b32_e32 v20, 16, v11
+; SI-NEXT: v_cvt_f32_f16_e32 v45, v6
+; SI-NEXT: v_cvt_f32_f16_e32 v12, v10
+; SI-NEXT: v_cvt_f32_f16_e32 v16, v20
+; SI-NEXT: v_cvt_f32_f16_e32 v62, v62
; SI-NEXT: v_add_f64 v[46:47], s[20:21], 1.0
-; SI-NEXT: v_lshrrev_b32_e32 v27, 16, v14
+; SI-NEXT: v_lshrrev_b32_e32 v38, 16, v19
+; SI-NEXT: v_lshrrev_b32_e32 v31, 16, v14
; SI-NEXT: v_cvt_f32_f16_e32 v6, v11
; SI-NEXT: v_cvt_f32_f16_e32 v11, v15
-; SI-NEXT: buffer_store_dword v4, off, s[0:3], s32 offset:72 ; 4-byte Folded Spill
+; SI-NEXT: v_cvt_f32_f16_e32 v39, v19
; SI-NEXT: v_lshrrev_b32_e32 v14, 16, v46
; SI-NEXT: v_lshrrev_b32_e32 v15, 16, v47
-; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_lshrrev_b32_e32 v4, 16, v2
+; SI-NEXT: v_lshrrev_b32_e32 v19, 16, v1
+; SI-NEXT: v_cvt_f32_f16_e32 v33, v47
+; SI-NEXT: v_cvt_f32_f16_e32 v47, v46
+; SI-NEXT: v_cvt_f32_f16_e32 v46, v15
+; SI-NEXT: v_mov_b32_e32 v15, v45
+; SI-NEXT: v_cvt_f32_f16_e32 v45, v14
+; SI-NEXT: v_cvt_f32_f16_e32 v14, v18
+; SI-NEXT: s_waitcnt vmcnt(7)
+; SI-NEXT: v_cvt_f32_f16_e32 v48, v13
+; SI-NEXT: buffer_load_dword v13, off, s[0:3], s32 offset:100 ; 4-byte Folded Reload
+; SI-NEXT: v_mov_b32_e32 v18, v12
+; SI-NEXT: v_mov_b32_e32 v12, v62
+; SI-NEXT: v_cvt_f32_f16_e32 v62, v19
; SI-NEXT: v_cvt_f32_f16_e32 v37, v37
; SI-NEXT: v_cvt_f32_f16_e32 v49, v49
; SI-NEXT: v_cvt_f32_f16_e32 v53, v53
-; SI-NEXT: v_cvt_f32_f16_e32 v33, v47
-; SI-NEXT: s_waitcnt vmcnt(4)
-; SI-NEXT: v_cvt_f32_f16_e32 v28, v13
-; SI-NEXT: buffer_load_dword v13, off, s[0:3], s32 offset:104 ; 4-byte Folded Reload
-; SI-NEXT: v_cvt_f32_f16_e32 v47, v46
-; SI-NEXT: v_cvt_f32_f16_e32 v2, v2
; SI-NEXT: v_cvt_f32_f16_e32 v1, v1
-; SI-NEXT: v_cvt_f32_f16_e32 v55, v32
-; SI-NEXT: v_cvt_f32_f16_e32 v27, v27
+; SI-NEXT: v_cvt_f32_f16_e32 v10, v35
+; SI-NEXT: v_cvt_f32_f16_e32 v41, v32
; SI-NEXT: v_cvt_f32_f16_e32 v31, v31
; SI-NEXT: v_cvt_f32_f16_e32 v38, v38
; SI-NEXT: v_cvt_f32_f16_e32 v50, v50
; SI-NEXT: v_cvt_f32_f16_e32 v54, v54
; SI-NEXT: v_cvt_f32_f16_e32 v40, v40
; SI-NEXT: v_cvt_f32_f16_e32 v42, v42
+; SI-NEXT: v_cvt_f32_f16_e32 v56, v56
+; SI-NEXT: v_cvt_f32_f16_e32 v35, v58
; SI-NEXT: v_cvt_f32_f16_e32 v32, v60
-; SI-NEXT: v_cvt_f32_f16_e32 v46, v15
-; SI-NEXT: v_mov_b32_e32 v15, v5
-; SI-NEXT: v_cvt_f32_f16_e32 v5, v14
-; SI-NEXT: v_mov_b32_e32 v14, v12
-; SI-NEXT: v_mov_b32_e32 v12, v57
-; SI-NEXT: v_cvt_f32_f16_e32 v57, v18
+; SI-NEXT: v_cvt_f32_f16_e32 v58, v7
; SI-NEXT: v_cvt_f32_f16_e32 v60, v4
-; SI-NEXT: v_mov_b32_e32 v18, v3
-; SI-NEXT: buffer_load_dword v17, off, s[0:3], s32 offset:84 ; 4-byte Folded Reload
+; SI-NEXT: v_mov_b32_e32 v7, v57
+; SI-NEXT: v_mov_b32_e32 v57, v3
; SI-NEXT: buffer_load_dword v19, off, s[0:3], s32 offset:88 ; 4-byte Folded Reload
-; SI-NEXT: buffer_load_dword v35, off, s[0:3], s32 offset:76 ; 4-byte Folded Reload
-; SI-NEXT: s_waitcnt vmcnt(3)
-; SI-NEXT: v_cvt_f32_f16_e32 v48, v13
-; SI-NEXT: buffer_load_dword v13, off, s[0:3], s32 offset:100 ; 4-byte Folded Reload
-; SI-NEXT: s_waitcnt vmcnt(3)
-; SI-NEXT: v_cvt_f32_f16_e32 v20, v17
-; SI-NEXT: buffer_load_dword v17, off, s[0:3], s32 offset:80 ; 4-byte Folded Reload
-; SI-NEXT: s_waitcnt vmcnt(2)
-; SI-NEXT: v_cvt_f32_f16_e32 v35, v35
; SI-NEXT: s_waitcnt vmcnt(1)
; SI-NEXT: v_cvt_f32_f16_e32 v24, v13
; SI-NEXT: buffer_load_dword v13, off, s[0:3], s32 offset:96 ; 4-byte Folded Reload
-; SI-NEXT: s_waitcnt vmcnt(1)
-; SI-NEXT: v_cvt_f32_f16_e32 v17, v17
; SI-NEXT: s_waitcnt vmcnt(0)
; SI-NEXT: v_cvt_f32_f16_e32 v52, v13
; SI-NEXT: buffer_load_dword v13, off, s[0:3], s32 offset:92 ; 4-byte Folded Reload
; SI-NEXT: s_waitcnt vmcnt(0)
+; SI-NEXT: v_cvt_f32_f16_e32 v28, v13
+; SI-NEXT: buffer_load_dword v13, off, s[0:3], s32 offset:84 ; 4-byte Folded Reload
+; SI-NEXT: v_cvt_f32_f16_e32 v36, v2
+; SI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:108 ; 4-byte Folded Reload
+; SI-NEXT: s_waitcnt vmcnt(1)
+; SI-NEXT: v_cvt_f32_f16_e32 v17, v13
+; SI-NEXT: buffer_load_dword v13, off, s[0:3], s32 offset:80 ; 4-byte Folded Reload
+; SI-NEXT: s_waitcnt vmcnt(1)
+; SI-NEXT: v_cvt_f32_f16_e32 v2, v2
+; SI-NEXT: s_waitcnt vmcnt(0)
+; SI-NEXT: v_cvt_f32_f16_e32 v20, v13
+; SI-NEXT: buffer_load_dword v13, off, s[0:3], s32 offset:76 ; 4-byte Folded Reload
+; SI-NEXT: s_waitcnt vmcnt(0)
; SI-NEXT: v_cvt_f32_f16_e32 v13, v13
; SI-NEXT: .LBB81_3: ; %end
; SI-NEXT: v_cvt_f16_f32_e32 v3, v62
@@ -145123,7 +145160,7 @@ define inreg <64 x half> @bitcast_v16f64_to_v64f16_scalar(<16 x double> inreg %a
; SI-NEXT: buffer_store_dword v1, v0, s[0:3], 0 offen
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_cvt_f16_f32_e32 v1, v60
-; SI-NEXT: v_cvt_f16_f32_e32 v3, v2
+; SI-NEXT: v_cvt_f16_f32_e32 v3, v63
; SI-NEXT: v_lshlrev_b32_e32 v1, 16, v1
; SI-NEXT: v_or_b32_e32 v1, v3, v1
; SI-NEXT: v_add_i32_e32 v3, vcc, 4, v0
@@ -145136,13 +145173,13 @@ define inreg <64 x half> @bitcast_v16f64_to_v64f16_scalar(<16 x double> inreg %a
; SI-NEXT: v_add_i32_e32 v3, vcc, 8, v0
; SI-NEXT: buffer_store_dword v1, v3, s[0:3], 0 offen
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_cvt_f16_f32_e32 v1, v57
+; SI-NEXT: v_cvt_f16_f32_e32 v1, v14
; SI-NEXT: v_cvt_f16_f32_e32 v3, v34
; SI-NEXT: v_lshlrev_b32_e32 v1, 16, v1
; SI-NEXT: v_or_b32_e32 v1, v3, v1
; SI-NEXT: buffer_store_dword v1, v4, s[0:3], 0 offen
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_cvt_f16_f32_e32 v1, v5
+; SI-NEXT: v_cvt_f16_f32_e32 v1, v45
; SI-NEXT: v_cvt_f16_f32_e32 v3, v47
; SI-NEXT: v_add_i32_e32 v4, vcc, 16, v0
; SI-NEXT: v_lshlrev_b32_e32 v1, 16, v1
@@ -145156,28 +145193,28 @@ define inreg <64 x half> @bitcast_v16f64_to_v64f16_scalar(<16 x double> inreg %a
; SI-NEXT: v_or_b32_e32 v1, v3, v1
; SI-NEXT: buffer_store_dword v1, v4, s[0:3], 0 offen
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_cvt_f16_f32_e32 v1, v35
+; SI-NEXT: v_cvt_f16_f32_e32 v1, v13
; SI-NEXT: v_cvt_f16_f32_e32 v3, v21
; SI-NEXT: v_add_i32_e32 v4, vcc, 24, v0
; SI-NEXT: v_lshlrev_b32_e32 v1, 16, v1
; SI-NEXT: v_or_b32_e32 v1, v3, v1
; SI-NEXT: buffer_store_dword v1, v4, s[0:3], 0 offen
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_cvt_f16_f32_e32 v1, v17
+; SI-NEXT: v_cvt_f16_f32_e32 v1, v20
; SI-NEXT: v_cvt_f16_f32_e32 v3, v30
; SI-NEXT: v_add_i32_e32 v4, vcc, 28, v0
; SI-NEXT: v_lshlrev_b32_e32 v1, 16, v1
; SI-NEXT: v_or_b32_e32 v1, v3, v1
; SI-NEXT: buffer_store_dword v1, v4, s[0:3], 0 offen
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_cvt_f16_f32_e32 v1, v20
+; SI-NEXT: v_cvt_f16_f32_e32 v1, v17
; SI-NEXT: v_cvt_f16_f32_e32 v3, v26
; SI-NEXT: v_add_i32_e32 v4, vcc, 32, v0
; SI-NEXT: v_lshlrev_b32_e32 v1, 16, v1
; SI-NEXT: v_or_b32_e32 v1, v3, v1
; SI-NEXT: buffer_store_dword v1, v4, s[0:3], 0 offen
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_cvt_f16_f32_e32 v1, v13
+; SI-NEXT: v_cvt_f16_f32_e32 v1, v28
; SI-NEXT: v_cvt_f16_f32_e32 v3, v53
; SI-NEXT: v_add_i32_e32 v4, vcc, 36, v0
; SI-NEXT: v_lshlrev_b32_e32 v1, 16, v1
@@ -145205,7 +145242,7 @@ define inreg <64 x half> @bitcast_v16f64_to_v64f16_scalar(<16 x double> inreg %a
; SI-NEXT: v_or_b32_e32 v1, v3, v1
; SI-NEXT: buffer_store_dword v1, v4, s[0:3], 0 offen
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_cvt_f16_f32_e32 v1, v28
+; SI-NEXT: v_cvt_f16_f32_e32 v1, v2
; SI-NEXT: v_cvt_f16_f32_e32 v3, v37
; SI-NEXT: v_add_i32_e32 v4, vcc, 52, v0
; SI-NEXT: v_lshlrev_b32_e32 v1, 16, v1
@@ -145227,111 +145264,111 @@ define inreg <64 x half> @bitcast_v16f64_to_v64f16_scalar(<16 x double> inreg %a
; SI-NEXT: v_lshlrev_b32_e32 v1, 16, v1
; SI-NEXT: s_waitcnt vmcnt(0)
; SI-NEXT: v_cvt_f16_f32_e32 v3, v2
+; SI-NEXT: v_cvt_f16_f32_e32 v2, v6
; SI-NEXT: v_or_b32_e32 v1, v3, v1
; SI-NEXT: buffer_store_dword v1, v4, s[0:3], 0 offen
-; SI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:64 ; 4-byte Folded Reload
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_cvt_f16_f32_e32 v1, v12
+; SI-NEXT: v_cvt_f16_f32_e32 v1, v35
+; SI-NEXT: v_cvt_f16_f32_e32 v3, v61
; SI-NEXT: v_add_i32_e32 v4, vcc, 64, v0
; SI-NEXT: v_lshlrev_b32_e32 v1, 16, v1
-; SI-NEXT: s_waitcnt vmcnt(0)
-; SI-NEXT: v_cvt_f16_f32_e32 v3, v2
-; SI-NEXT: v_cvt_f16_f32_e32 v2, v6
; SI-NEXT: v_or_b32_e32 v1, v3, v1
; SI-NEXT: buffer_store_dword v1, v4, s[0:3], 0 offen
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_cvt_f16_f32_e32 v1, v61
-; SI-NEXT: v_cvt_f16_f32_e32 v3, v63
+; SI-NEXT: v_cvt_f16_f32_e32 v1, v12
+; SI-NEXT: v_cvt_f16_f32_e32 v3, v59
; SI-NEXT: v_add_i32_e32 v4, vcc, 0x44, v0
; SI-NEXT: v_lshlrev_b32_e32 v1, 16, v1
; SI-NEXT: v_or_b32_e32 v1, v3, v1
; SI-NEXT: buffer_store_dword v1, v4, s[0:3], 0 offen
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_cvt_f16_f32_e32 v1, v42
-; SI-NEXT: v_cvt_f16_f32_e32 v3, v44
+; SI-NEXT: v_cvt_f16_f32_e32 v1, v56
+; SI-NEXT: v_cvt_f16_f32_e32 v3, v57
; SI-NEXT: v_add_i32_e32 v4, vcc, 0x48, v0
; SI-NEXT: v_lshlrev_b32_e32 v1, 16, v1
; SI-NEXT: v_or_b32_e32 v1, v3, v1
; SI-NEXT: buffer_store_dword v1, v4, s[0:3], 0 offen
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_cvt_f16_f32_e32 v1, v40
-; SI-NEXT: v_cvt_f16_f32_e32 v3, v45
+; SI-NEXT: v_cvt_f16_f32_e32 v1, v42
+; SI-NEXT: v_cvt_f16_f32_e32 v3, v44
; SI-NEXT: v_add_i32_e32 v4, vcc, 0x4c, v0
; SI-NEXT: v_lshlrev_b32_e32 v1, 16, v1
; SI-NEXT: v_or_b32_e32 v1, v3, v1
; SI-NEXT: buffer_store_dword v1, v4, s[0:3], 0 offen
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_cvt_f16_f32_e32 v1, v54
+; SI-NEXT: v_cvt_f16_f32_e32 v1, v40
; SI-NEXT: v_cvt_f16_f32_e32 v3, v43
; SI-NEXT: v_add_i32_e32 v4, vcc, 0x50, v0
; SI-NEXT: v_lshlrev_b32_e32 v1, 16, v1
; SI-NEXT: v_or_b32_e32 v1, v3, v1
; SI-NEXT: buffer_store_dword v1, v4, s[0:3], 0 offen
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_cvt_f16_f32_e32 v1, v50
-; SI-NEXT: v_cvt_f16_f32_e32 v3, v41
+; SI-NEXT: v_cvt_f16_f32_e32 v1, v54
+; SI-NEXT: v_cvt_f16_f32_e32 v3, v55
; SI-NEXT: v_add_i32_e32 v4, vcc, 0x54, v0
; SI-NEXT: v_lshlrev_b32_e32 v1, 16, v1
; SI-NEXT: v_or_b32_e32 v1, v3, v1
; SI-NEXT: buffer_store_dword v1, v4, s[0:3], 0 offen
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_cvt_f16_f32_e32 v1, v38
+; SI-NEXT: v_cvt_f16_f32_e32 v1, v50
; SI-NEXT: v_cvt_f16_f32_e32 v3, v51
; SI-NEXT: v_add_i32_e32 v4, vcc, 0x58, v0
; SI-NEXT: v_lshlrev_b32_e32 v1, 16, v1
; SI-NEXT: v_or_b32_e32 v1, v3, v1
; SI-NEXT: buffer_store_dword v1, v4, s[0:3], 0 offen
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_cvt_f16_f32_e32 v1, v31
+; SI-NEXT: v_cvt_f16_f32_e32 v1, v38
; SI-NEXT: v_cvt_f16_f32_e32 v3, v39
; SI-NEXT: v_add_i32_e32 v4, vcc, 0x5c, v0
; SI-NEXT: v_lshlrev_b32_e32 v1, 16, v1
; SI-NEXT: v_or_b32_e32 v1, v3, v1
; SI-NEXT: buffer_store_dword v1, v4, s[0:3], 0 offen
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_cvt_f16_f32_e32 v1, v27
+; SI-NEXT: v_cvt_f16_f32_e32 v1, v31
; SI-NEXT: v_cvt_f16_f32_e32 v3, v19
; SI-NEXT: v_add_i32_e32 v4, vcc, 0x60, v0
; SI-NEXT: v_lshlrev_b32_e32 v1, 16, v1
; SI-NEXT: v_or_b32_e32 v1, v3, v1
; SI-NEXT: buffer_store_dword v1, v4, s[0:3], 0 offen
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_cvt_f16_f32_e32 v1, v23
+; SI-NEXT: v_cvt_f16_f32_e32 v1, v27
; SI-NEXT: v_cvt_f16_f32_e32 v3, v11
; SI-NEXT: v_add_i32_e32 v4, vcc, 0x64, v0
; SI-NEXT: v_lshlrev_b32_e32 v1, 16, v1
; SI-NEXT: v_or_b32_e32 v1, v3, v1
; SI-NEXT: buffer_store_dword v1, v4, s[0:3], 0 offen
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_cvt_f16_f32_e32 v1, v16
+; SI-NEXT: v_cvt_f16_f32_e32 v1, v23
; SI-NEXT: v_cvt_f16_f32_e32 v3, v18
; SI-NEXT: v_add_i32_e32 v4, vcc, 0x68, v0
; SI-NEXT: v_lshlrev_b32_e32 v1, 16, v1
; SI-NEXT: v_or_b32_e32 v1, v3, v1
; SI-NEXT: buffer_store_dword v1, v4, s[0:3], 0 offen
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_cvt_f16_f32_e32 v1, v9
+; SI-NEXT: v_cvt_f16_f32_e32 v1, v16
; SI-NEXT: v_add_i32_e32 v3, vcc, 0x6c, v0
; SI-NEXT: v_lshlrev_b32_e32 v1, 16, v1
; SI-NEXT: v_or_b32_e32 v1, v2, v1
; SI-NEXT: buffer_store_dword v1, v3, s[0:3], 0 offen
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_cvt_f16_f32_e32 v1, v8
+; SI-NEXT: v_cvt_f16_f32_e32 v1, v9
; SI-NEXT: v_cvt_f16_f32_e32 v2, v15
; SI-NEXT: v_add_i32_e32 v3, vcc, 0x70, v0
; SI-NEXT: v_lshlrev_b32_e32 v1, 16, v1
; SI-NEXT: v_or_b32_e32 v1, v2, v1
; SI-NEXT: buffer_store_dword v1, v3, s[0:3], 0 offen
+; SI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:64 ; 4-byte Folded Reload
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_cvt_f16_f32_e32 v1, v59
-; SI-NEXT: v_cvt_f16_f32_e32 v2, v14
+; SI-NEXT: v_cvt_f16_f32_e32 v1, v8
; SI-NEXT: v_add_i32_e32 v3, vcc, 0x74, v0
; SI-NEXT: v_lshlrev_b32_e32 v1, 16, v1
+; SI-NEXT: s_waitcnt vmcnt(0)
+; SI-NEXT: v_cvt_f16_f32_e32 v2, v2
; SI-NEXT: v_or_b32_e32 v1, v2, v1
; SI-NEXT: buffer_store_dword v1, v3, s[0:3], 0 offen
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_cvt_f16_f32_e32 v1, v55
-; SI-NEXT: v_cvt_f16_f32_e32 v2, v56
+; SI-NEXT: v_cvt_f16_f32_e32 v1, v41
+; SI-NEXT: v_cvt_f16_f32_e32 v2, v5
; SI-NEXT: v_add_i32_e32 v3, vcc, 0x78, v0
; SI-NEXT: v_lshlrev_b32_e32 v1, 16, v1
; SI-NEXT: v_or_b32_e32 v1, v2, v1
@@ -145367,24 +145404,24 @@ define inreg <64 x half> @bitcast_v16f64_to_v64f16_scalar(<16 x double> inreg %a
; SI-NEXT: ; implicit-def: $vgpr3
; SI-NEXT: ; implicit-def: $vgpr1
; SI-NEXT: ; implicit-def: $vgpr62
-; SI-NEXT: ; implicit-def: $vgpr2
+; SI-NEXT: ; implicit-def: $vgpr63
; SI-NEXT: ; implicit-def: $vgpr60
; SI-NEXT: ; implicit-def: $vgpr25
; SI-NEXT: ; implicit-def: $vgpr58
; SI-NEXT: ; implicit-def: $vgpr34
-; SI-NEXT: ; implicit-def: $vgpr57
+; SI-NEXT: ; implicit-def: $vgpr14
; SI-NEXT: ; implicit-def: $vgpr47
-; SI-NEXT: ; implicit-def: $vgpr5
+; SI-NEXT: ; implicit-def: $vgpr45
; SI-NEXT: ; implicit-def: $vgpr33
; SI-NEXT: ; implicit-def: $vgpr46
; SI-NEXT: ; implicit-def: $vgpr21
-; SI-NEXT: ; implicit-def: $vgpr35
+; SI-NEXT: ; implicit-def: $vgpr13
; SI-NEXT: ; implicit-def: $vgpr30
-; SI-NEXT: ; implicit-def: $vgpr17
-; SI-NEXT: ; implicit-def: $vgpr26
; SI-NEXT: ; implicit-def: $vgpr20
+; SI-NEXT: ; implicit-def: $vgpr26
+; SI-NEXT: ; implicit-def: $vgpr17
; SI-NEXT: ; implicit-def: $vgpr53
-; SI-NEXT: ; implicit-def: $vgpr13
+; SI-NEXT: ; implicit-def: $vgpr28
; SI-NEXT: ; implicit-def: $vgpr29
; SI-NEXT: ; implicit-def: $vgpr52
; SI-NEXT: ; implicit-def: $vgpr49
@@ -145392,39 +145429,39 @@ define inreg <64 x half> @bitcast_v16f64_to_v64f16_scalar(<16 x double> inreg %a
; SI-NEXT: ; implicit-def: $vgpr22
; SI-NEXT: ; implicit-def: $vgpr48
; SI-NEXT: ; implicit-def: $vgpr37
-; SI-NEXT: ; implicit-def: $vgpr28
+; SI-NEXT: ; implicit-def: $vgpr2
; SI-NEXT: ; implicit-def: $vgpr36
; SI-NEXT: ; kill: killed $vgpr3
; SI-NEXT: ; implicit-def: $vgpr32
-; SI-NEXT: ; implicit-def: $vgpr12
-; SI-NEXT: ; implicit-def: $vgpr63
; SI-NEXT: ; implicit-def: $vgpr61
+; SI-NEXT: ; implicit-def: $vgpr35
+; SI-NEXT: ; implicit-def: $vgpr59
+; SI-NEXT: ; implicit-def: $vgpr12
+; SI-NEXT: ; implicit-def: $vgpr57
+; SI-NEXT: ; implicit-def: $vgpr56
; SI-NEXT: ; implicit-def: $vgpr44
; SI-NEXT: ; implicit-def: $vgpr42
-; SI-NEXT: ; implicit-def: $vgpr45
-; SI-NEXT: ; implicit-def: $vgpr40
; SI-NEXT: ; implicit-def: $vgpr43
+; SI-NEXT: ; implicit-def: $vgpr40
+; SI-NEXT: ; implicit-def: $vgpr55
; SI-NEXT: ; implicit-def: $vgpr54
-; SI-NEXT: ; implicit-def: $vgpr41
-; SI-NEXT: ; implicit-def: $vgpr50
; SI-NEXT: ; implicit-def: $vgpr51
-; SI-NEXT: ; implicit-def: $vgpr38
+; SI-NEXT: ; implicit-def: $vgpr50
; SI-NEXT: ; implicit-def: $vgpr39
-; SI-NEXT: ; implicit-def: $vgpr31
+; SI-NEXT: ; implicit-def: $vgpr38
; SI-NEXT: ; implicit-def: $vgpr19
-; SI-NEXT: ; implicit-def: $vgpr27
+; SI-NEXT: ; implicit-def: $vgpr31
; SI-NEXT: ; implicit-def: $vgpr11
-; SI-NEXT: ; implicit-def: $vgpr23
+; SI-NEXT: ; implicit-def: $vgpr27
; SI-NEXT: ; implicit-def: $vgpr18
-; SI-NEXT: ; implicit-def: $vgpr16
+; SI-NEXT: ; implicit-def: $vgpr23
; SI-NEXT: ; implicit-def: $vgpr6
-; SI-NEXT: ; implicit-def: $vgpr9
+; SI-NEXT: ; implicit-def: $vgpr16
; SI-NEXT: ; implicit-def: $vgpr15
+; SI-NEXT: ; implicit-def: $vgpr9
; SI-NEXT: ; implicit-def: $vgpr8
-; SI-NEXT: ; implicit-def: $vgpr14
-; SI-NEXT: ; implicit-def: $vgpr59
-; SI-NEXT: ; implicit-def: $vgpr56
-; SI-NEXT: ; implicit-def: $vgpr55
+; SI-NEXT: ; implicit-def: $vgpr5
+; SI-NEXT: ; implicit-def: $vgpr41
; SI-NEXT: ; implicit-def: $vgpr7
; SI-NEXT: ; implicit-def: $vgpr10
; SI-NEXT: ; implicit-def: $vgpr3
@@ -149698,24 +149735,24 @@ define inreg <16 x double> @bitcast_v64i16_to_v16f64_scalar(<64 x i16> inreg %a,
; SI-NEXT: buffer_load_dword v10, off, s[0:3], s32 offset:32
; SI-NEXT: buffer_load_dword v46, off, s[0:3], s32 offset:28
; SI-NEXT: buffer_load_dword v12, off, s[0:3], s32 offset:40
-; SI-NEXT: buffer_load_dword v45, off, s[0:3], s32 offset:36
-; SI-NEXT: buffer_load_dword v38, off, s[0:3], s32 offset:48
+; SI-NEXT: buffer_load_dword v59, off, s[0:3], s32 offset:36
+; SI-NEXT: buffer_load_dword v48, off, s[0:3], s32 offset:48
; SI-NEXT: buffer_load_dword v44, off, s[0:3], s32 offset:44
; SI-NEXT: buffer_load_dword v36, off, s[0:3], s32 offset:56
-; SI-NEXT: buffer_load_dword v59, off, s[0:3], s32 offset:52
+; SI-NEXT: buffer_load_dword v43, off, s[0:3], s32 offset:52
; SI-NEXT: buffer_load_dword v34, off, s[0:3], s32 offset:64
; SI-NEXT: buffer_load_dword v42, off, s[0:3], s32 offset:60
; SI-NEXT: buffer_load_dword v32, off, s[0:3], s32 offset:72
-; SI-NEXT: buffer_load_dword v58, off, s[0:3], s32 offset:68
-; SI-NEXT: s_waitcnt expcnt(2)
-; SI-NEXT: v_lshlrev_b32_e32 v61, 16, v1
+; SI-NEXT: buffer_load_dword v38, off, s[0:3], s32 offset:68
+; SI-NEXT: v_lshlrev_b32_e32 v58, 16, v1
; SI-NEXT: v_lshlrev_b32_e32 v55, 16, v3
-; SI-NEXT: v_lshlrev_b32_e32 v50, 16, v5
-; SI-NEXT: v_lshlrev_b32_e32 v43, 16, v7
-; SI-NEXT: v_lshlrev_b32_e32 v41, 16, v9
-; SI-NEXT: v_lshlrev_b32_e32 v40, 16, v11
+; SI-NEXT: s_waitcnt expcnt(2)
+; SI-NEXT: v_lshlrev_b32_e32 v61, 16, v5
+; SI-NEXT: v_lshlrev_b32_e32 v50, 16, v7
+; SI-NEXT: v_lshlrev_b32_e32 v45, 16, v9
+; SI-NEXT: v_lshlrev_b32_e32 v41, 16, v11
; SI-NEXT: v_lshlrev_b32_e32 v13, 16, v13
-; SI-NEXT: v_lshlrev_b32_e32 v48, 16, v15
+; SI-NEXT: v_lshlrev_b32_e32 v40, 16, v15
; SI-NEXT: v_lshlrev_b32_e32 v15, 16, v17
; SI-NEXT: v_lshlrev_b32_e32 v37, 16, v19
; SI-NEXT: v_lshlrev_b32_e32 v17, 16, v21
@@ -149736,23 +149773,23 @@ define inreg <16 x double> @bitcast_v64i16_to_v16f64_scalar(<64 x i16> inreg %a,
; SI-NEXT: s_waitcnt vmcnt(9)
; SI-NEXT: v_lshlrev_b32_e32 v54, 16, v12
; SI-NEXT: s_waitcnt vmcnt(7)
-; SI-NEXT: v_lshlrev_b32_e32 v5, 16, v38
+; SI-NEXT: v_lshlrev_b32_e32 v5, 16, v48
; SI-NEXT: s_waitcnt vmcnt(5) expcnt(1)
; SI-NEXT: v_lshlrev_b32_e32 v62, 16, v36
; SI-NEXT: s_waitcnt vmcnt(3)
; SI-NEXT: v_lshlrev_b32_e32 v3, 16, v34
; SI-NEXT: s_waitcnt vmcnt(1)
-; SI-NEXT: v_lshlrev_b32_e32 v34, 16, v32
-; SI-NEXT: buffer_store_dword v34, off, s[0:3], s32 offset:144 ; 4-byte Folded Spill
+; SI-NEXT: v_lshlrev_b32_e32 v48, 16, v32
+; SI-NEXT: buffer_store_dword v48, off, s[0:3], s32 offset:144 ; 4-byte Folded Spill
; SI-NEXT: buffer_store_dword v3, off, s[0:3], s32 offset:148 ; 4-byte Folded Spill
; SI-NEXT: buffer_store_dword v62, off, s[0:3], s32 offset:152 ; 4-byte Folded Spill
; SI-NEXT: buffer_store_dword v5, off, s[0:3], s32 offset:156 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt vmcnt(4)
-; SI-NEXT: buffer_store_dword v58, off, s[0:3], s32 offset:160 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v38, off, s[0:3], s32 offset:160 ; 4-byte Folded Spill
; SI-NEXT: buffer_store_dword v54, off, s[0:3], s32 offset:164 ; 4-byte Folded Spill
-; SI-NEXT: buffer_store_dword v27, off, s[0:3], s32 offset:168 ; 4-byte Folded Spill
-; SI-NEXT: buffer_store_dword v59, off, s[0:3], s32 offset:172 ; 4-byte Folded Spill
-; SI-NEXT: buffer_store_dword v45, off, s[0:3], s32 offset:176 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v43, off, s[0:3], s32 offset:168 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v27, off, s[0:3], s32 offset:172 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v59, off, s[0:3], s32 offset:176 ; 4-byte Folded Spill
; SI-NEXT: buffer_store_dword v25, off, s[0:3], s32 offset:180 ; 4-byte Folded Spill
; SI-NEXT: buffer_store_dword v47, off, s[0:3], s32 offset:184 ; 4-byte Folded Spill
; SI-NEXT: buffer_store_dword v60, off, s[0:3], s32 offset:188 ; 4-byte Folded Spill
@@ -149765,45 +149802,46 @@ define inreg <16 x double> @bitcast_v64i16_to_v16f64_scalar(<64 x i16> inreg %a,
; SI-NEXT: buffer_store_dword v19, off, s[0:3], s32 offset:216 ; 4-byte Folded Spill
; SI-NEXT: buffer_store_dword v35, off, s[0:3], s32 offset:220 ; 4-byte Folded Spill
; SI-NEXT: buffer_store_dword v22, off, s[0:3], s32 offset:224 ; 4-byte Folded Spill
-; SI-NEXT: buffer_store_dword v17, off, s[0:3], s32 offset:228 ; 4-byte Folded Spill
-; SI-NEXT: buffer_store_dword v0, off, s[0:3], s32 offset:232 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v0, off, s[0:3], s32 offset:228 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v17, off, s[0:3], s32 offset:232 ; 4-byte Folded Spill
; SI-NEXT: buffer_store_dword v37, off, s[0:3], s32 offset:236 ; 4-byte Folded Spill
; SI-NEXT: buffer_store_dword v18, off, s[0:3], s32 offset:240 ; 4-byte Folded Spill
; SI-NEXT: s_cbranch_scc0 .LBB87_4
; SI-NEXT: ; %bb.1: ; %cmp.false
-; SI-NEXT: s_waitcnt expcnt(2)
+; SI-NEXT: s_waitcnt expcnt(3)
; SI-NEXT: v_and_b32_e32 v0, 0xffff, v0
-; SI-NEXT: v_or_b32_e32 v7, v0, v61
+; SI-NEXT: v_or_b32_e32 v7, v0, v58
; SI-NEXT: v_and_b32_e32 v0, 0xffff, v4
-; SI-NEXT: v_or_b32_e32 v9, v0, v50
+; SI-NEXT: v_or_b32_e32 v9, v0, v61
; SI-NEXT: v_and_b32_e32 v0, 0xffff, v6
-; SI-NEXT: v_or_b32_e32 v10, v0, v43
+; SI-NEXT: v_or_b32_e32 v10, v0, v50
; SI-NEXT: v_and_b32_e32 v0, 0xffff, v57
-; SI-NEXT: v_or_b32_e32 v11, v0, v41
+; SI-NEXT: v_or_b32_e32 v11, v0, v45
; SI-NEXT: v_and_b32_e32 v0, 0xffff, v56
-; SI-NEXT: v_or_b32_e32 v12, v0, v40
+; SI-NEXT: v_or_b32_e32 v12, v0, v41
; SI-NEXT: v_and_b32_e32 v0, 0xffff, v49
-; SI-NEXT: v_mov_b32_e32 v52, v57
-; SI-NEXT: v_mov_b32_e32 v57, v40
-; SI-NEXT: v_mov_b32_e32 v40, v49
-; SI-NEXT: v_mov_b32_e32 v49, v13
+; SI-NEXT: v_mov_b32_e32 v36, v41
+; SI-NEXT: v_mov_b32_e32 v41, v13
; SI-NEXT: v_or_b32_e32 v13, v0, v13
; SI-NEXT: v_and_b32_e32 v0, 0xffff, v14
-; SI-NEXT: v_mov_b32_e32 v36, v41
-; SI-NEXT: v_mov_b32_e32 v41, v14
-; SI-NEXT: v_or_b32_e32 v14, v0, v48
-; SI-NEXT: v_and_b32_e32 v0, 0xffff, v16
; SI-NEXT: v_mov_b32_e32 v51, v50
-; SI-NEXT: v_mov_b32_e32 v50, v43
-; SI-NEXT: v_mov_b32_e32 v43, v48
-; SI-NEXT: v_mov_b32_e32 v48, v15
+; SI-NEXT: v_mov_b32_e32 v50, v45
+; SI-NEXT: v_mov_b32_e32 v45, v14
+; SI-NEXT: v_or_b32_e32 v14, v0, v40
+; SI-NEXT: v_and_b32_e32 v0, 0xffff, v16
+; SI-NEXT: v_mov_b32_e32 v52, v57
+; SI-NEXT: v_mov_b32_e32 v57, v49
+; SI-NEXT: v_mov_b32_e32 v49, v40
+; SI-NEXT: v_mov_b32_e32 v40, v15
; SI-NEXT: v_or_b32_e32 v15, v0, v15
; SI-NEXT: v_and_b32_e32 v0, 0xffff, v18
-; SI-NEXT: v_mov_b32_e32 v38, v61
+; SI-NEXT: v_mov_b32_e32 v34, v58
+; SI-NEXT: v_mov_b32_e32 v58, v61
; SI-NEXT: v_mov_b32_e32 v61, v56
; SI-NEXT: v_mov_b32_e32 v56, v16
; SI-NEXT: v_or_b32_e32 v16, v0, v37
; SI-NEXT: v_and_b32_e32 v0, 0xffff, v20
+; SI-NEXT: s_waitcnt expcnt(2)
; SI-NEXT: v_or_b32_e32 v17, v0, v17
; SI-NEXT: v_and_b32_e32 v0, 0xffff, v22
; SI-NEXT: s_waitcnt expcnt(0)
@@ -149837,7 +149875,7 @@ define inreg <16 x double> @bitcast_v64i16_to_v16f64_scalar(<64 x i16> inreg %a,
; SI-NEXT: s_and_b32 s6, s20, 0xffff
; SI-NEXT: s_lshl_b32 s7, s21, 16
; SI-NEXT: v_or_b32_e32 v26, v0, v26
-; SI-NEXT: v_and_b32_e32 v0, 0xffff, v45
+; SI-NEXT: v_and_b32_e32 v0, 0xffff, v59
; SI-NEXT: s_or_b32 s6, s6, s7
; SI-NEXT: s_and_b32 s7, s22, 0xffff
; SI-NEXT: s_lshl_b32 s8, s23, 16
@@ -149848,7 +149886,7 @@ define inreg <16 x double> @bitcast_v64i16_to_v16f64_scalar(<64 x i16> inreg %a,
; SI-NEXT: s_lshl_b32 s9, s25, 16
; SI-NEXT: v_mov_b32_e32 v33, v28
; SI-NEXT: v_or_b32_e32 v28, v0, v5
-; SI-NEXT: v_and_b32_e32 v0, 0xffff, v59
+; SI-NEXT: v_and_b32_e32 v0, 0xffff, v43
; SI-NEXT: s_or_b32 s8, s8, s9
; SI-NEXT: s_and_b32 s9, s26, 0xffff
; SI-NEXT: s_lshl_b32 s10, s27, 16
@@ -149860,7 +149898,7 @@ define inreg <16 x double> @bitcast_v64i16_to_v16f64_scalar(<64 x i16> inreg %a,
; SI-NEXT: s_lshl_b32 s11, s29, 16
; SI-NEXT: v_and_b32_e32 v1, 0xffff, v2
; SI-NEXT: v_or_b32_e32 v30, v0, v3
-; SI-NEXT: v_and_b32_e32 v0, 0xffff, v58
+; SI-NEXT: v_and_b32_e32 v0, 0xffff, v38
; SI-NEXT: s_or_b32 s10, s10, s11
; SI-NEXT: v_mov_b32_e32 v63, v2
; SI-NEXT: v_mov_b32_e32 v32, v55
@@ -149868,9 +149906,9 @@ define inreg <16 x double> @bitcast_v64i16_to_v16f64_scalar(<64 x i16> inreg %a,
; SI-NEXT: v_mov_b32_e32 v55, v4
; SI-NEXT: v_mov_b32_e32 v53, v6
; SI-NEXT: v_mov_b32_e32 v47, v46
-; SI-NEXT: v_mov_b32_e32 v45, v44
-; SI-NEXT: v_mov_b32_e32 v59, v42
-; SI-NEXT: v_or_b32_e32 v31, v0, v34
+; SI-NEXT: v_mov_b32_e32 v59, v44
+; SI-NEXT: v_mov_b32_e32 v43, v42
+; SI-NEXT: v_or_b32_e32 v31, v0, v48
; SI-NEXT: v_mov_b32_e32 v0, s4
; SI-NEXT: v_mov_b32_e32 v1, s5
; SI-NEXT: v_mov_b32_e32 v2, s6
@@ -149880,12 +149918,13 @@ define inreg <16 x double> @bitcast_v64i16_to_v16f64_scalar(<64 x i16> inreg %a,
; SI-NEXT: v_mov_b32_e32 v6, s10
; SI-NEXT: s_cbranch_execnz .LBB87_3
; SI-NEXT: .LBB87_2: ; %cmp.true
-; SI-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:232 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:228 ; 4-byte Folded Reload
; SI-NEXT: v_add_i32_e32 v1, vcc, 3, v63
; SI-NEXT: v_and_b32_e32 v1, 0xffff, v1
; SI-NEXT: v_or_b32_e32 v1, v32, v1
; SI-NEXT: v_add_i32_e32 v8, vcc, 0x30000, v1
; SI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:236 ; 4-byte Folded Reload
+; SI-NEXT: v_mov_b32_e32 v38, v43
; SI-NEXT: s_add_i32 s16, s16, 3
; SI-NEXT: s_and_b32 s4, s16, 0xffff
; SI-NEXT: s_lshl_b32 s5, s17, 16
@@ -149929,42 +149968,42 @@ define inreg <16 x double> @bitcast_v64i16_to_v16f64_scalar(<64 x i16> inreg %a,
; SI-NEXT: s_waitcnt vmcnt(1)
; SI-NEXT: v_add_i32_e32 v0, vcc, 3, v0
; SI-NEXT: v_and_b32_e32 v0, 0xffff, v0
-; SI-NEXT: v_or_b32_e32 v0, v38, v0
+; SI-NEXT: v_or_b32_e32 v0, v34, v0
; SI-NEXT: v_add_i32_e32 v7, vcc, 0x30000, v0
; SI-NEXT: v_add_i32_e32 v0, vcc, 3, v55
; SI-NEXT: v_and_b32_e32 v0, 0xffff, v0
-; SI-NEXT: v_or_b32_e32 v0, v51, v0
+; SI-NEXT: v_or_b32_e32 v0, v58, v0
; SI-NEXT: v_add_i32_e32 v9, vcc, 0x30000, v0
; SI-NEXT: v_add_i32_e32 v0, vcc, 3, v53
; SI-NEXT: v_and_b32_e32 v0, 0xffff, v0
-; SI-NEXT: v_or_b32_e32 v0, v50, v0
+; SI-NEXT: v_or_b32_e32 v0, v51, v0
; SI-NEXT: v_add_i32_e32 v10, vcc, 0x30000, v0
; SI-NEXT: v_add_i32_e32 v0, vcc, 3, v52
; SI-NEXT: v_and_b32_e32 v0, 0xffff, v0
-; SI-NEXT: v_or_b32_e32 v0, v36, v0
+; SI-NEXT: v_or_b32_e32 v0, v50, v0
; SI-NEXT: v_add_i32_e32 v11, vcc, 0x30000, v0
; SI-NEXT: v_add_i32_e32 v0, vcc, 3, v61
; SI-NEXT: v_and_b32_e32 v0, 0xffff, v0
-; SI-NEXT: v_or_b32_e32 v0, v57, v0
+; SI-NEXT: v_or_b32_e32 v0, v36, v0
; SI-NEXT: v_add_i32_e32 v12, vcc, 0x30000, v0
-; SI-NEXT: v_add_i32_e32 v0, vcc, 3, v40
+; SI-NEXT: v_add_i32_e32 v0, vcc, 3, v57
; SI-NEXT: v_and_b32_e32 v0, 0xffff, v0
-; SI-NEXT: v_or_b32_e32 v0, v49, v0
+; SI-NEXT: v_or_b32_e32 v0, v41, v0
; SI-NEXT: v_add_i32_e32 v13, vcc, 0x30000, v0
-; SI-NEXT: v_add_i32_e32 v0, vcc, 3, v41
+; SI-NEXT: v_add_i32_e32 v0, vcc, 3, v45
; SI-NEXT: v_and_b32_e32 v0, 0xffff, v0
-; SI-NEXT: v_or_b32_e32 v0, v43, v0
+; SI-NEXT: v_or_b32_e32 v0, v49, v0
; SI-NEXT: v_add_i32_e32 v14, vcc, 0x30000, v0
; SI-NEXT: v_add_i32_e32 v0, vcc, 3, v56
; SI-NEXT: v_and_b32_e32 v0, 0xffff, v0
-; SI-NEXT: v_or_b32_e32 v0, v48, v0
+; SI-NEXT: v_or_b32_e32 v0, v40, v0
; SI-NEXT: v_add_i32_e32 v15, vcc, 0x30000, v0
; SI-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:240 ; 4-byte Folded Reload
; SI-NEXT: s_waitcnt vmcnt(0)
; SI-NEXT: v_add_i32_e32 v0, vcc, 3, v0
; SI-NEXT: v_and_b32_e32 v0, 0xffff, v0
; SI-NEXT: v_or_b32_e32 v0, v1, v0
-; SI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:228 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:232 ; 4-byte Folded Reload
; SI-NEXT: v_add_i32_e32 v16, vcc, 0x30000, v0
; SI-NEXT: v_add_i32_e32 v0, vcc, 3, v37
; SI-NEXT: v_and_b32_e32 v0, 0xffff, v0
@@ -150026,7 +150065,7 @@ define inreg <16 x double> @bitcast_v64i16_to_v16f64_scalar(<64 x i16> inreg %a,
; SI-NEXT: v_add_i32_e32 v0, vcc, 3, v0
; SI-NEXT: v_and_b32_e32 v0, 0xffff, v0
; SI-NEXT: v_or_b32_e32 v0, v1, v0
-; SI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:168 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:172 ; 4-byte Folded Reload
; SI-NEXT: v_add_i32_e32 v25, vcc, 0x30000, v0
; SI-NEXT: v_add_i32_e32 v0, vcc, 3, v47
; SI-NEXT: v_and_b32_e32 v0, 0xffff, v0
@@ -150042,12 +150081,12 @@ define inreg <16 x double> @bitcast_v64i16_to_v16f64_scalar(<64 x i16> inreg %a,
; SI-NEXT: v_or_b32_e32 v0, v1, v0
; SI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:156 ; 4-byte Folded Reload
; SI-NEXT: v_add_i32_e32 v27, vcc, 0x30000, v0
-; SI-NEXT: v_add_i32_e32 v0, vcc, 3, v45
+; SI-NEXT: v_add_i32_e32 v0, vcc, 3, v59
; SI-NEXT: v_and_b32_e32 v0, 0xffff, v0
; SI-NEXT: s_waitcnt vmcnt(0)
; SI-NEXT: v_or_b32_e32 v0, v1, v0
; SI-NEXT: v_add_i32_e32 v28, vcc, 0x30000, v0
-; SI-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:172 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:168 ; 4-byte Folded Reload
; SI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:152 ; 4-byte Folded Reload
; SI-NEXT: s_waitcnt vmcnt(1)
; SI-NEXT: v_add_i32_e32 v0, vcc, 3, v0
@@ -150056,7 +150095,7 @@ define inreg <16 x double> @bitcast_v64i16_to_v16f64_scalar(<64 x i16> inreg %a,
; SI-NEXT: v_or_b32_e32 v0, v1, v0
; SI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:148 ; 4-byte Folded Reload
; SI-NEXT: v_add_i32_e32 v29, vcc, 0x30000, v0
-; SI-NEXT: v_add_i32_e32 v0, vcc, 3, v59
+; SI-NEXT: v_add_i32_e32 v0, vcc, 3, v38
; SI-NEXT: v_and_b32_e32 v0, 0xffff, v0
; SI-NEXT: s_waitcnt vmcnt(0)
; SI-NEXT: v_or_b32_e32 v0, v1, v0
@@ -150091,26 +150130,26 @@ define inreg <16 x double> @bitcast_v64i16_to_v16f64_scalar(<64 x i16> inreg %a,
; SI-NEXT: s_waitcnt vmcnt(0) expcnt(0)
; SI-NEXT: s_setpc_b64 s[30:31]
; SI-NEXT: .LBB87_4:
-; SI-NEXT: v_mov_b32_e32 v38, v61
+; SI-NEXT: v_mov_b32_e32 v34, v58
; SI-NEXT: v_mov_b32_e32 v32, v55
+; SI-NEXT: v_mov_b32_e32 v58, v61
; SI-NEXT: v_mov_b32_e32 v63, v2
; SI-NEXT: v_mov_b32_e32 v55, v4
; SI-NEXT: v_mov_b32_e32 v53, v6
; SI-NEXT: v_mov_b32_e32 v52, v57
; SI-NEXT: v_mov_b32_e32 v51, v50
; SI-NEXT: v_mov_b32_e32 v61, v56
-; SI-NEXT: v_mov_b32_e32 v50, v43
+; SI-NEXT: v_mov_b32_e32 v50, v45
; SI-NEXT: v_mov_b32_e32 v36, v41
-; SI-NEXT: v_mov_b32_e32 v57, v40
-; SI-NEXT: v_mov_b32_e32 v40, v49
-; SI-NEXT: v_mov_b32_e32 v49, v13
-; SI-NEXT: v_mov_b32_e32 v43, v48
-; SI-NEXT: v_mov_b32_e32 v48, v15
-; SI-NEXT: v_mov_b32_e32 v41, v14
+; SI-NEXT: v_mov_b32_e32 v41, v13
+; SI-NEXT: v_mov_b32_e32 v57, v49
+; SI-NEXT: v_mov_b32_e32 v49, v40
+; SI-NEXT: v_mov_b32_e32 v40, v15
+; SI-NEXT: v_mov_b32_e32 v45, v14
; SI-NEXT: v_mov_b32_e32 v56, v16
; SI-NEXT: v_mov_b32_e32 v47, v46
-; SI-NEXT: v_mov_b32_e32 v45, v44
-; SI-NEXT: v_mov_b32_e32 v59, v42
+; SI-NEXT: v_mov_b32_e32 v59, v44
+; SI-NEXT: v_mov_b32_e32 v43, v42
; SI-NEXT: s_waitcnt expcnt(1)
; SI-NEXT: v_mov_b32_e32 v37, v20
; SI-NEXT: v_mov_b32_e32 v39, v23
@@ -157017,6 +157056,7 @@ define inreg <64 x bfloat> @bitcast_v128i8_to_v64bf16_scalar(<128 x i8> inreg %a
; SI-NEXT: buffer_store_dword v41, off, s[0:3], s32 offset:344 ; 4-byte Folded Spill
; SI-NEXT: buffer_store_dword v42, off, s[0:3], s32 offset:348 ; 4-byte Folded Spill
; SI-NEXT: buffer_store_dword v43, off, s[0:3], s32 offset:352 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v44, off, s[0:3], s32 offset:356 ; 4-byte Folded Spill
; SI-NEXT: s_mov_b64 exec, s[4:5]
; SI-NEXT: buffer_store_dword v40, off, s[0:3], s32 offset:340 ; 4-byte Folded Spill
; SI-NEXT: buffer_load_dword v31, off, s[0:3], s32 offset:332
@@ -157027,15 +157067,18 @@ define inreg <64 x bfloat> @bitcast_v128i8_to_v64bf16_scalar(<128 x i8> inreg %a
; SI-NEXT: buffer_load_dword v36, off, s[0:3], s32 offset:312
; SI-NEXT: buffer_load_dword v37, off, s[0:3], s32 offset:308
; SI-NEXT: buffer_load_dword v38, off, s[0:3], s32 offset:304
-; SI-NEXT: ; implicit-def: $vgpr43 : SGPR spill to VGPR lane
-; SI-NEXT: s_mov_b32 s72, s21
-; SI-NEXT: s_waitcnt expcnt(1)
-; SI-NEXT: v_writelane_b32 v43, s19, 0
-; SI-NEXT: v_writelane_b32 v43, s18, 1
-; SI-NEXT: v_writelane_b32 v43, s17, 2
-; SI-NEXT: v_writelane_b32 v43, s16, 3
-; SI-NEXT: s_mov_b32 s60, s24
+; SI-NEXT: ; implicit-def: $vgpr44 : SGPR spill to VGPR lane
+; SI-NEXT: s_waitcnt expcnt(4)
; SI-NEXT: v_writelane_b32 v41, s30, 0
+; SI-NEXT: s_waitcnt expcnt(1)
+; SI-NEXT: v_writelane_b32 v44, s29, 0
+; SI-NEXT: v_writelane_b32 v44, s28, 1
+; SI-NEXT: v_writelane_b32 v44, s27, 2
+; SI-NEXT: v_writelane_b32 v44, s26, 3
+; SI-NEXT: v_writelane_b32 v44, s19, 4
+; SI-NEXT: v_writelane_b32 v44, s18, 5
+; SI-NEXT: v_writelane_b32 v44, s17, 6
+; SI-NEXT: v_writelane_b32 v44, s16, 7
; SI-NEXT: v_writelane_b32 v41, s31, 1
; SI-NEXT: v_writelane_b32 v41, s34, 2
; SI-NEXT: v_writelane_b32 v41, s35, 3
@@ -157059,8 +157102,6 @@ define inreg <64 x bfloat> @bitcast_v128i8_to_v64bf16_scalar(<128 x i8> inreg %a
; SI-NEXT: v_writelane_b32 v41, s69, 21
; SI-NEXT: v_writelane_b32 v41, s70, 22
; SI-NEXT: v_writelane_b32 v41, s71, 23
-; SI-NEXT: s_mov_b32 s77, s28
-; SI-NEXT: s_mov_b32 s76, s27
; SI-NEXT: v_writelane_b32 v41, s80, 24
; SI-NEXT: v_writelane_b32 v41, s81, 25
; SI-NEXT: v_writelane_b32 v41, s82, 26
@@ -157071,100 +157112,92 @@ define inreg <64 x bfloat> @bitcast_v128i8_to_v64bf16_scalar(<128 x i8> inreg %a
; SI-NEXT: v_writelane_b32 v41, s87, 31
; SI-NEXT: v_writelane_b32 v41, s96, 32
; SI-NEXT: v_writelane_b32 v41, s97, 33
+; SI-NEXT: ; implicit-def: $vgpr43 : SGPR spill to VGPR lane
+; SI-NEXT: v_readfirstlane_b32 s30, v16
+; SI-NEXT: v_readfirstlane_b32 s31, v15
+; SI-NEXT: v_readfirstlane_b32 s34, v21
+; SI-NEXT: v_readfirstlane_b32 s35, v22
+; SI-NEXT: v_readfirstlane_b32 s36, v20
+; SI-NEXT: v_readfirstlane_b32 s37, v19
+; SI-NEXT: v_readfirstlane_b32 s38, v25
+; SI-NEXT: v_readfirstlane_b32 s39, v26
+; SI-NEXT: v_readfirstlane_b32 s48, v24
+; SI-NEXT: v_readfirstlane_b32 s49, v23
+; SI-NEXT: v_readfirstlane_b32 s50, v29
+; SI-NEXT: v_readfirstlane_b32 s51, v30
+; SI-NEXT: v_readfirstlane_b32 s52, v28
+; SI-NEXT: v_readfirstlane_b32 s53, v27
; SI-NEXT: v_writelane_b32 v41, s98, 34
; SI-NEXT: v_writelane_b32 v41, s99, 35
-; SI-NEXT: s_mov_b32 s79, s26
-; SI-NEXT: v_readfirstlane_b32 s38, v20
-; SI-NEXT: ; implicit-def: $vgpr42 : SGPR spill to VGPR lane
-; SI-NEXT: v_readfirstlane_b32 s39, v19
-; SI-NEXT: v_writelane_b32 v42, s38, 0
-; SI-NEXT: v_readfirstlane_b32 s48, v25
-; SI-NEXT: v_writelane_b32 v42, s39, 1
-; SI-NEXT: v_readfirstlane_b32 s49, v26
-; SI-NEXT: v_writelane_b32 v42, s48, 2
-; SI-NEXT: v_readfirstlane_b32 s50, v24
-; SI-NEXT: v_writelane_b32 v42, s49, 3
-; SI-NEXT: v_readfirstlane_b32 s51, v23
-; SI-NEXT: v_writelane_b32 v42, s50, 4
-; SI-NEXT: v_readfirstlane_b32 s52, v29
-; SI-NEXT: v_writelane_b32 v42, s51, 5
-; SI-NEXT: v_readfirstlane_b32 s53, v30
-; SI-NEXT: v_writelane_b32 v42, s52, 6
-; SI-NEXT: v_readfirstlane_b32 s54, v28
-; SI-NEXT: v_writelane_b32 v42, s53, 7
-; SI-NEXT: v_readfirstlane_b32 s55, v27
-; SI-NEXT: v_writelane_b32 v42, s54, 8
-; SI-NEXT: v_writelane_b32 v42, s55, 9
+; SI-NEXT: s_mov_b32 s6, s21
; SI-NEXT: v_readfirstlane_b32 s16, v1
; SI-NEXT: v_readfirstlane_b32 s17, v2
; SI-NEXT: v_readfirstlane_b32 s18, v5
; SI-NEXT: v_readfirstlane_b32 s19, v6
-; SI-NEXT: v_readfirstlane_b32 s88, v4
-; SI-NEXT: v_readfirstlane_b32 s89, v3
-; SI-NEXT: v_readfirstlane_b32 s90, v9
+; SI-NEXT: v_readfirstlane_b32 s78, v4
+; SI-NEXT: v_readfirstlane_b32 s79, v3
+; SI-NEXT: v_readfirstlane_b32 s88, v9
+; SI-NEXT: v_readfirstlane_b32 s89, v10
+; SI-NEXT: v_readfirstlane_b32 s90, v8
+; SI-NEXT: v_readfirstlane_b32 s91, v7
+; SI-NEXT: v_readfirstlane_b32 s92, v13
+; SI-NEXT: v_readfirstlane_b32 s93, v14
+; SI-NEXT: v_readfirstlane_b32 s94, v12
; SI-NEXT: s_waitcnt vmcnt(7)
-; SI-NEXT: v_readfirstlane_b32 s6, v31
+; SI-NEXT: v_readfirstlane_b32 s4, v31
; SI-NEXT: buffer_load_dword v31, off, s[0:3], s32 offset:300
; SI-NEXT: buffer_load_dword v39, off, s[0:3], s32 offset:296
; SI-NEXT: buffer_load_dword v48, off, s[0:3], s32 offset:292
; SI-NEXT: buffer_load_dword v49, off, s[0:3], s32 offset:288
; SI-NEXT: buffer_load_dword v50, off, s[0:3], s32 offset:284
; SI-NEXT: buffer_load_dword v51, off, s[0:3], s32 offset:280
+; SI-NEXT: v_writelane_b32 v44, s4, 8
; SI-NEXT: s_waitcnt vmcnt(12)
; SI-NEXT: v_readfirstlane_b32 s4, v32
-; SI-NEXT: v_writelane_b32 v43, s4, 4
+; SI-NEXT: v_writelane_b32 v44, s4, 9
; SI-NEXT: s_waitcnt vmcnt(11)
; SI-NEXT: v_readfirstlane_b32 s4, v33
; SI-NEXT: buffer_load_dword v32, off, s[0:3], s32 offset:276
; SI-NEXT: buffer_load_dword v33, off, s[0:3], s32 offset:272
-; SI-NEXT: v_writelane_b32 v43, s4, 5
+; SI-NEXT: v_writelane_b32 v44, s4, 10
; SI-NEXT: s_waitcnt vmcnt(12)
; SI-NEXT: v_readfirstlane_b32 s4, v34
-; SI-NEXT: v_writelane_b32 v43, s4, 6
+; SI-NEXT: v_writelane_b32 v44, s4, 11
; SI-NEXT: s_waitcnt vmcnt(11)
; SI-NEXT: v_readfirstlane_b32 s4, v35
-; SI-NEXT: v_writelane_b32 v43, s4, 7
+; SI-NEXT: v_writelane_b32 v44, s4, 12
; SI-NEXT: s_waitcnt vmcnt(10)
; SI-NEXT: v_readfirstlane_b32 s4, v36
-; SI-NEXT: v_writelane_b32 v43, s4, 8
+; SI-NEXT: v_writelane_b32 v44, s4, 13
; SI-NEXT: s_waitcnt vmcnt(9)
; SI-NEXT: v_readfirstlane_b32 s4, v37
+; SI-NEXT: v_writelane_b32 v44, s4, 14
; SI-NEXT: buffer_load_dword v34, off, s[0:3], s32 offset:268
; SI-NEXT: buffer_load_dword v35, off, s[0:3], s32 offset:264
; SI-NEXT: buffer_load_dword v36, off, s[0:3], s32 offset:260
; SI-NEXT: buffer_load_dword v37, off, s[0:3], s32 offset:256
-; SI-NEXT: v_writelane_b32 v43, s4, 9
; SI-NEXT: s_waitcnt vmcnt(12)
; SI-NEXT: v_readfirstlane_b32 s4, v38
-; SI-NEXT: v_writelane_b32 v43, s4, 10
-; SI-NEXT: v_readfirstlane_b32 s91, v10
-; SI-NEXT: v_readfirstlane_b32 s92, v8
-; SI-NEXT: v_readfirstlane_b32 s93, v7
-; SI-NEXT: v_readfirstlane_b32 s94, v13
-; SI-NEXT: v_readfirstlane_b32 s95, v14
-; SI-NEXT: v_readfirstlane_b32 s30, v17
-; SI-NEXT: v_readfirstlane_b32 s31, v18
-; SI-NEXT: v_readfirstlane_b32 s34, v16
-; SI-NEXT: v_readfirstlane_b32 s35, v15
-; SI-NEXT: v_readfirstlane_b32 s36, v21
-; SI-NEXT: v_readfirstlane_b32 s37, v22
+; SI-NEXT: v_writelane_b32 v44, s4, 15
+; SI-NEXT: v_readfirstlane_b32 s95, v11
; SI-NEXT: s_waitcnt vmcnt(11)
; SI-NEXT: v_readfirstlane_b32 s4, v31
-; SI-NEXT: v_writelane_b32 v43, s4, 11
+; SI-NEXT: v_writelane_b32 v44, s4, 16
; SI-NEXT: s_waitcnt vmcnt(10)
; SI-NEXT: v_readfirstlane_b32 s4, v39
-; SI-NEXT: v_writelane_b32 v43, s4, 12
+; SI-NEXT: v_writelane_b32 v44, s4, 17
; SI-NEXT: s_waitcnt vmcnt(9)
; SI-NEXT: v_readfirstlane_b32 s4, v48
-; SI-NEXT: v_writelane_b32 v43, s4, 13
+; SI-NEXT: v_writelane_b32 v44, s4, 18
; SI-NEXT: s_waitcnt vmcnt(8)
; SI-NEXT: v_readfirstlane_b32 s4, v49
-; SI-NEXT: v_writelane_b32 v43, s4, 14
+; SI-NEXT: v_writelane_b32 v44, s4, 19
; SI-NEXT: s_waitcnt vmcnt(7)
; SI-NEXT: v_readfirstlane_b32 s4, v50
-; SI-NEXT: v_writelane_b32 v43, s4, 15
+; SI-NEXT: v_writelane_b32 v44, s4, 20
; SI-NEXT: s_waitcnt vmcnt(6)
; SI-NEXT: v_readfirstlane_b32 s4, v51
+; SI-NEXT: v_writelane_b32 v44, s4, 21
; SI-NEXT: buffer_load_dword v31, off, s[0:3], s32 offset:252
; SI-NEXT: buffer_load_dword v38, off, s[0:3], s32 offset:248
; SI-NEXT: buffer_load_dword v39, off, s[0:3], s32 offset:244
@@ -157173,39 +157206,49 @@ define inreg <64 x bfloat> @bitcast_v128i8_to_v64bf16_scalar(<128 x i8> inreg %a
; SI-NEXT: buffer_load_dword v50, off, s[0:3], s32 offset:232
; SI-NEXT: buffer_load_dword v51, off, s[0:3], s32 offset:228
; SI-NEXT: s_waitcnt vmcnt(12)
-; SI-NEXT: v_readfirstlane_b32 s75, v32
+; SI-NEXT: v_readfirstlane_b32 s4, v32
+; SI-NEXT: v_writelane_b32 v44, s4, 22
; SI-NEXT: s_waitcnt vmcnt(11)
-; SI-NEXT: v_readfirstlane_b32 s61, v33
+; SI-NEXT: v_readfirstlane_b32 s4, v33
; SI-NEXT: buffer_load_dword v32, off, s[0:3], s32 offset:224
; SI-NEXT: buffer_load_dword v33, off, s[0:3], s32 offset:220
-; SI-NEXT: v_writelane_b32 v43, s4, 16
+; SI-NEXT: v_writelane_b32 v44, s4, 23
; SI-NEXT: s_waitcnt vmcnt(12)
-; SI-NEXT: v_readfirstlane_b32 s43, v34
+; SI-NEXT: v_readfirstlane_b32 s4, v34
+; SI-NEXT: v_writelane_b32 v44, s4, 24
; SI-NEXT: s_waitcnt vmcnt(11)
-; SI-NEXT: v_readfirstlane_b32 s40, v35
+; SI-NEXT: v_readfirstlane_b32 s4, v35
+; SI-NEXT: v_writelane_b32 v44, s4, 25
; SI-NEXT: s_waitcnt vmcnt(10)
; SI-NEXT: v_readfirstlane_b32 s4, v36
+; SI-NEXT: v_writelane_b32 v44, s4, 26
; SI-NEXT: s_waitcnt vmcnt(9)
-; SI-NEXT: v_readfirstlane_b32 s63, v37
+; SI-NEXT: v_readfirstlane_b32 s4, v37
+; SI-NEXT: v_writelane_b32 v44, s4, 27
; SI-NEXT: buffer_load_dword v34, off, s[0:3], s32 offset:216
; SI-NEXT: buffer_load_dword v35, off, s[0:3], s32 offset:212
; SI-NEXT: buffer_load_dword v36, off, s[0:3], s32 offset:208
; SI-NEXT: buffer_load_dword v37, off, s[0:3], s32 offset:204
-; SI-NEXT: v_writelane_b32 v43, s4, 17
; SI-NEXT: s_waitcnt vmcnt(12)
-; SI-NEXT: v_readfirstlane_b32 s59, v31
+; SI-NEXT: v_readfirstlane_b32 s4, v31
+; SI-NEXT: v_writelane_b32 v44, s4, 28
; SI-NEXT: s_waitcnt vmcnt(11)
-; SI-NEXT: v_readfirstlane_b32 s42, v38
+; SI-NEXT: v_readfirstlane_b32 s4, v38
+; SI-NEXT: v_writelane_b32 v44, s4, 29
; SI-NEXT: s_waitcnt vmcnt(10)
-; SI-NEXT: v_readfirstlane_b32 s73, v39
+; SI-NEXT: v_readfirstlane_b32 s4, v39
+; SI-NEXT: v_writelane_b32 v44, s4, 30
; SI-NEXT: s_waitcnt vmcnt(9)
-; SI-NEXT: v_readfirstlane_b32 s21, v48
+; SI-NEXT: v_readfirstlane_b32 s4, v48
+; SI-NEXT: v_writelane_b32 v44, s4, 31
; SI-NEXT: s_waitcnt vmcnt(8)
-; SI-NEXT: v_readfirstlane_b32 s57, v49
+; SI-NEXT: v_readfirstlane_b32 s4, v49
+; SI-NEXT: v_writelane_b32 v44, s4, 32
; SI-NEXT: s_waitcnt vmcnt(7)
-; SI-NEXT: v_readfirstlane_b32 s13, v50
+; SI-NEXT: v_readfirstlane_b32 s4, v50
+; SI-NEXT: v_writelane_b32 v44, s4, 33
; SI-NEXT: s_waitcnt vmcnt(6)
-; SI-NEXT: v_readfirstlane_b32 s45, v51
+; SI-NEXT: v_readfirstlane_b32 s4, v51
; SI-NEXT: buffer_load_dword v31, off, s[0:3], s32 offset:200
; SI-NEXT: buffer_load_dword v38, off, s[0:3], s32 offset:196
; SI-NEXT: buffer_load_dword v39, off, s[0:3], s32 offset:192
@@ -157213,51 +157256,43 @@ define inreg <64 x bfloat> @bitcast_v128i8_to_v64bf16_scalar(<128 x i8> inreg %a
; SI-NEXT: buffer_load_dword v49, off, s[0:3], s32 offset:184
; SI-NEXT: buffer_load_dword v50, off, s[0:3], s32 offset:180
; SI-NEXT: buffer_load_dword v51, off, s[0:3], s32 offset:176
-; SI-NEXT: s_waitcnt vmcnt(12)
-; SI-NEXT: v_readfirstlane_b32 s47, v32
; SI-NEXT: s_waitcnt vmcnt(11)
-; SI-NEXT: v_readfirstlane_b32 s24, v33
+; SI-NEXT: v_readfirstlane_b32 s58, v33
; SI-NEXT: buffer_load_dword v33, off, s[0:3], s32 offset:172
; SI-NEXT: buffer_load_dword v52, off, s[0:3], s32 offset:168
+; SI-NEXT: v_readfirstlane_b32 s26, v32
+; SI-NEXT: s_waitcnt vmcnt(12)
+; SI-NEXT: v_readfirstlane_b32 s77, v34
+; SI-NEXT: s_waitcnt vmcnt(11)
+; SI-NEXT: v_readfirstlane_b32 s63, v35
+; SI-NEXT: s_waitcnt vmcnt(10)
+; SI-NEXT: v_readfirstlane_b32 s57, v36
+; SI-NEXT: s_waitcnt vmcnt(9)
+; SI-NEXT: v_readfirstlane_b32 s56, v37
; SI-NEXT: buffer_load_dword v53, off, s[0:3], s32 offset:164
; SI-NEXT: buffer_load_dword v54, off, s[0:3], s32 offset:160
; SI-NEXT: buffer_load_dword v55, off, s[0:3], s32 offset:156
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: buffer_load_dword v40, off, s[0:3], s32 offset:152
-; SI-NEXT: s_waitcnt vmcnt(14)
-; SI-NEXT: v_readfirstlane_b32 s78, v34
-; SI-NEXT: v_readfirstlane_b32 s4, v35
-; SI-NEXT: v_writelane_b32 v43, s4, 18
-; SI-NEXT: v_readfirstlane_b32 s4, v36
-; SI-NEXT: v_writelane_b32 v43, s4, 19
-; SI-NEXT: s_waitcnt vmcnt(13)
-; SI-NEXT: v_readfirstlane_b32 s4, v37
-; SI-NEXT: v_writelane_b32 v43, s4, 20
+; SI-NEXT: v_writelane_b32 v44, s4, 34
; SI-NEXT: s_waitcnt vmcnt(12)
-; SI-NEXT: v_readfirstlane_b32 s4, v31
-; SI-NEXT: v_writelane_b32 v43, s4, 21
+; SI-NEXT: v_readfirstlane_b32 s61, v31
; SI-NEXT: s_waitcnt vmcnt(11)
-; SI-NEXT: v_readfirstlane_b32 s4, v38
-; SI-NEXT: v_writelane_b32 v43, s4, 22
+; SI-NEXT: v_readfirstlane_b32 s74, v38
; SI-NEXT: s_waitcnt vmcnt(10)
-; SI-NEXT: v_readfirstlane_b32 s4, v39
-; SI-NEXT: v_writelane_b32 v43, s4, 23
+; SI-NEXT: v_readfirstlane_b32 s76, v39
; SI-NEXT: s_waitcnt vmcnt(9)
-; SI-NEXT: v_readfirstlane_b32 s4, v48
-; SI-NEXT: v_writelane_b32 v43, s4, 24
+; SI-NEXT: v_readfirstlane_b32 s47, v48
; SI-NEXT: s_waitcnt vmcnt(8)
-; SI-NEXT: v_readfirstlane_b32 s4, v49
-; SI-NEXT: v_writelane_b32 v43, s4, 25
+; SI-NEXT: v_readfirstlane_b32 s45, v49
; SI-NEXT: s_waitcnt vmcnt(7)
-; SI-NEXT: v_readfirstlane_b32 s4, v50
-; SI-NEXT: v_writelane_b32 v43, s4, 26
+; SI-NEXT: v_readfirstlane_b32 s60, v50
; SI-NEXT: s_waitcnt vmcnt(6)
-; SI-NEXT: v_readfirstlane_b32 s4, v51
-; SI-NEXT: v_writelane_b32 v43, s4, 27
+; SI-NEXT: v_readfirstlane_b32 s42, v51
; SI-NEXT: buffer_load_dword v31, off, s[0:3], s32 offset:148
; SI-NEXT: buffer_load_dword v32, off, s[0:3], s32 offset:144
; SI-NEXT: s_waitcnt vmcnt(7)
-; SI-NEXT: v_readfirstlane_b32 s4, v33
+; SI-NEXT: v_readfirstlane_b32 s13, v33
; SI-NEXT: buffer_load_dword v33, off, s[0:3], s32 offset:140
; SI-NEXT: buffer_load_dword v34, off, s[0:3], s32 offset:136
; SI-NEXT: buffer_load_dword v35, off, s[0:3], s32 offset:132
@@ -157269,43 +157304,43 @@ define inreg <64 x bfloat> @bitcast_v128i8_to_v64bf16_scalar(<128 x i8> inreg %a
; SI-NEXT: buffer_load_dword v49, off, s[0:3], s32 offset:112
; SI-NEXT: buffer_load_dword v50, off, s[0:3], s32 offset:108
; SI-NEXT: buffer_load_dword v51, off, s[0:3], s32 offset:104
-; SI-NEXT: v_writelane_b32 v43, s4, 28
; SI-NEXT: s_waitcnt vmcnt(14)
-; SI-NEXT: v_readfirstlane_b32 s4, v52
-; SI-NEXT: v_writelane_b32 v43, s4, 29
-; SI-NEXT: v_readfirstlane_b32 s4, v53
-; SI-NEXT: v_writelane_b32 v43, s4, 30
-; SI-NEXT: v_readfirstlane_b32 s4, v54
-; SI-NEXT: v_writelane_b32 v43, s4, 31
-; SI-NEXT: v_readfirstlane_b32 s4, v55
-; SI-NEXT: v_writelane_b32 v43, s4, 32
+; SI-NEXT: v_readfirstlane_b32 s72, v52
+; SI-NEXT: v_readfirstlane_b32 s73, v53
+; SI-NEXT: v_readfirstlane_b32 s44, v55
; SI-NEXT: s_waitcnt vmcnt(13)
; SI-NEXT: v_readfirstlane_b32 s4, v40
-; SI-NEXT: v_writelane_b32 v43, s4, 33
-; SI-NEXT: v_writelane_b32 v43, s22, 34
-; SI-NEXT: v_writelane_b32 v43, s23, 35
-; SI-NEXT: v_writelane_b32 v43, s72, 36
-; SI-NEXT: v_writelane_b32 v43, s20, 37
-; SI-NEXT: v_writelane_b32 v43, s79, 38
-; SI-NEXT: v_writelane_b32 v43, s76, 39
-; SI-NEXT: v_writelane_b32 v43, s25, 40
-; SI-NEXT: v_writelane_b32 v43, s60, 41
-; SI-NEXT: v_writelane_b32 v43, s29, 42
-; SI-NEXT: v_writelane_b32 v43, s77, 43
-; SI-NEXT: v_writelane_b32 v43, s16, 44
-; SI-NEXT: v_writelane_b32 v43, s17, 45
-; SI-NEXT: v_writelane_b32 v43, s18, 46
-; SI-NEXT: v_writelane_b32 v43, s19, 47
-; SI-NEXT: v_writelane_b32 v43, s88, 48
-; SI-NEXT: v_writelane_b32 v43, s89, 49
-; SI-NEXT: v_writelane_b32 v43, s90, 50
-; SI-NEXT: v_writelane_b32 v43, s91, 51
-; SI-NEXT: v_writelane_b32 v43, s92, 52
-; SI-NEXT: v_writelane_b32 v43, s93, 53
-; SI-NEXT: v_writelane_b32 v43, s94, 54
-; SI-NEXT: v_writelane_b32 v43, s95, 55
+; SI-NEXT: v_writelane_b32 v44, s4, 35
+; SI-NEXT: v_writelane_b32 v44, s22, 36
+; SI-NEXT: v_writelane_b32 v44, s23, 37
+; SI-NEXT: v_writelane_b32 v44, s6, 38
+; SI-NEXT: v_writelane_b32 v44, s20, 39
+; SI-NEXT: v_writelane_b32 v44, s25, 40
+; SI-NEXT: v_writelane_b32 v44, s24, 41
+; SI-NEXT: v_writelane_b32 v44, s44, 42
+; SI-NEXT: v_writelane_b32 v44, s72, 43
+; SI-NEXT: v_writelane_b32 v44, s13, 44
+; SI-NEXT: v_writelane_b32 v44, s60, 45
+; SI-NEXT: v_writelane_b32 v44, s73, 46
+; SI-NEXT: v_readfirstlane_b32 s21, v54
+; SI-NEXT: v_writelane_b32 v44, s42, 47
+; SI-NEXT: v_writelane_b32 v44, s21, 48
+; SI-NEXT: v_writelane_b32 v44, s16, 49
+; SI-NEXT: v_writelane_b32 v44, s17, 50
+; SI-NEXT: v_writelane_b32 v44, s18, 51
+; SI-NEXT: v_writelane_b32 v44, s19, 52
+; SI-NEXT: v_writelane_b32 v44, s78, 53
+; SI-NEXT: v_writelane_b32 v44, s79, 54
+; SI-NEXT: v_writelane_b32 v44, s88, 55
+; SI-NEXT: v_writelane_b32 v44, s89, 56
+; SI-NEXT: v_writelane_b32 v44, s90, 57
+; SI-NEXT: v_writelane_b32 v44, s91, 58
+; SI-NEXT: v_writelane_b32 v44, s92, 59
+; SI-NEXT: v_writelane_b32 v44, s93, 60
+; SI-NEXT: v_writelane_b32 v44, s94, 61
+; SI-NEXT: v_writelane_b32 v44, s95, 62
; SI-NEXT: s_waitcnt vmcnt(10)
-; SI-NEXT: v_readfirstlane_b32 s62, v33
+; SI-NEXT: v_readfirstlane_b32 s40, v33
; SI-NEXT: s_waitcnt vmcnt(9)
; SI-NEXT: v_readfirstlane_b32 s10, v34
; SI-NEXT: s_waitcnt vmcnt(8)
@@ -157313,13 +157348,13 @@ define inreg <64 x bfloat> @bitcast_v128i8_to_v64bf16_scalar(<128 x i8> inreg %a
; SI-NEXT: v_readfirstlane_b32 s28, v31
; SI-NEXT: v_readfirstlane_b32 s27, v32
; SI-NEXT: s_waitcnt vmcnt(7)
-; SI-NEXT: v_readfirstlane_b32 s58, v36
+; SI-NEXT: v_readfirstlane_b32 s29, v36
; SI-NEXT: s_waitcnt vmcnt(6)
-; SI-NEXT: v_readfirstlane_b32 s69, v37
+; SI-NEXT: v_readfirstlane_b32 s70, v37
; SI-NEXT: s_waitcnt vmcnt(5)
; SI-NEXT: v_readfirstlane_b32 s14, v38
; SI-NEXT: s_waitcnt vmcnt(4)
-; SI-NEXT: v_readfirstlane_b32 s68, v39
+; SI-NEXT: v_readfirstlane_b32 s69, v39
; SI-NEXT: buffer_load_dword v31, off, s[0:3], s32 offset:100
; SI-NEXT: buffer_load_dword v32, off, s[0:3], s32 offset:96
; SI-NEXT: buffer_load_dword v33, off, s[0:3], s32 offset:92
@@ -157334,42 +157369,50 @@ define inreg <64 x bfloat> @bitcast_v128i8_to_v64bf16_scalar(<128 x i8> inreg %a
; SI-NEXT: s_waitcnt vmcnt(11)
; SI-NEXT: v_readfirstlane_b32 s11, v49
; SI-NEXT: s_waitcnt vmcnt(10)
-; SI-NEXT: v_readfirstlane_b32 s70, v50
+; SI-NEXT: v_readfirstlane_b32 s71, v50
; SI-NEXT: s_waitcnt vmcnt(9)
-; SI-NEXT: v_readfirstlane_b32 s71, v51
+; SI-NEXT: v_readfirstlane_b32 s81, v51
; SI-NEXT: buffer_load_dword v48, off, s[0:3], s32 offset:64
; SI-NEXT: buffer_load_dword v49, off, s[0:3], s32 offset:60
; SI-NEXT: buffer_load_dword v50, off, s[0:3], s32 offset:56
; SI-NEXT: buffer_load_dword v51, off, s[0:3], s32 offset:52
; SI-NEXT: s_and_b64 s[4:5], vcc, exec
-; SI-NEXT: v_readfirstlane_b32 vcc_lo, v12
-; SI-NEXT: v_readfirstlane_b32 vcc_hi, v11
-; SI-NEXT: v_writelane_b32 v43, vcc_lo, 56
-; SI-NEXT: v_writelane_b32 v43, vcc_hi, 57
-; SI-NEXT: v_writelane_b32 v43, s30, 58
-; SI-NEXT: v_writelane_b32 v43, s31, 59
-; SI-NEXT: v_writelane_b32 v43, s34, 60
-; SI-NEXT: v_writelane_b32 v43, s35, 61
-; SI-NEXT: v_writelane_b32 v43, s36, 62
-; SI-NEXT: v_writelane_b32 v43, s37, 63
+; SI-NEXT: v_readfirstlane_b32 vcc_hi, v18
+; SI-NEXT: v_writelane_b32 v43, vcc_hi, 0
+; SI-NEXT: v_writelane_b32 v43, s30, 1
+; SI-NEXT: v_writelane_b32 v43, s31, 2
+; SI-NEXT: v_writelane_b32 v43, s34, 3
+; SI-NEXT: v_writelane_b32 v43, s35, 4
+; SI-NEXT: v_writelane_b32 v43, s36, 5
+; SI-NEXT: v_writelane_b32 v43, s37, 6
+; SI-NEXT: v_writelane_b32 v43, s38, 7
+; SI-NEXT: v_writelane_b32 v43, s39, 8
+; SI-NEXT: v_writelane_b32 v43, s48, 9
+; SI-NEXT: v_writelane_b32 v43, s49, 10
+; SI-NEXT: v_writelane_b32 v43, s50, 11
+; SI-NEXT: v_writelane_b32 v43, s51, 12
+; SI-NEXT: v_writelane_b32 v43, s52, 13
+; SI-NEXT: v_writelane_b32 v43, s53, 14
+; SI-NEXT: v_readfirstlane_b32 vcc_lo, v17
+; SI-NEXT: v_writelane_b32 v44, vcc_lo, 63
; SI-NEXT: s_waitcnt vmcnt(12)
-; SI-NEXT: v_readfirstlane_b32 s74, v31
+; SI-NEXT: v_readfirstlane_b32 s46, v31
; SI-NEXT: s_waitcnt vmcnt(11)
-; SI-NEXT: v_readfirstlane_b32 s46, v32
+; SI-NEXT: v_readfirstlane_b32 s59, v32
; SI-NEXT: s_waitcnt vmcnt(10)
-; SI-NEXT: v_readfirstlane_b32 s96, v33
+; SI-NEXT: v_readfirstlane_b32 s83, v33
; SI-NEXT: s_waitcnt vmcnt(9)
-; SI-NEXT: v_readfirstlane_b32 s98, v34
+; SI-NEXT: v_readfirstlane_b32 s12, v34
; SI-NEXT: s_waitcnt vmcnt(8)
-; SI-NEXT: v_readfirstlane_b32 s41, v35
+; SI-NEXT: v_readfirstlane_b32 s97, v35
; SI-NEXT: s_waitcnt vmcnt(7)
-; SI-NEXT: v_readfirstlane_b32 s56, v36
+; SI-NEXT: v_readfirstlane_b32 s8, v36
; SI-NEXT: s_waitcnt vmcnt(6)
-; SI-NEXT: v_readfirstlane_b32 s87, v37
+; SI-NEXT: v_readfirstlane_b32 s84, v37
; SI-NEXT: s_waitcnt vmcnt(5)
-; SI-NEXT: v_readfirstlane_b32 s99, v38
+; SI-NEXT: v_readfirstlane_b32 s86, v38
; SI-NEXT: s_waitcnt vmcnt(4)
-; SI-NEXT: v_readfirstlane_b32 s81, v39
+; SI-NEXT: v_readfirstlane_b32 s15, v39
; SI-NEXT: buffer_load_dword v31, off, s[0:3], s32 offset:48
; SI-NEXT: buffer_load_dword v32, off, s[0:3], s32 offset:44
; SI-NEXT: buffer_load_dword v33, off, s[0:3], s32 offset:40
@@ -157379,415 +157422,417 @@ define inreg <64 x bfloat> @bitcast_v128i8_to_v64bf16_scalar(<128 x i8> inreg %a
; SI-NEXT: buffer_load_dword v37, off, s[0:3], s32 offset:24
; SI-NEXT: buffer_load_dword v38, off, s[0:3], s32 offset:20
; SI-NEXT: s_waitcnt vmcnt(11)
-; SI-NEXT: v_readfirstlane_b32 s26, v48
+; SI-NEXT: v_readfirstlane_b32 s62, v48
; SI-NEXT: s_waitcnt vmcnt(10)
-; SI-NEXT: v_readfirstlane_b32 s83, v49
+; SI-NEXT: v_readfirstlane_b32 s96, v49
; SI-NEXT: s_waitcnt vmcnt(9)
-; SI-NEXT: v_readfirstlane_b32 s82, v50
+; SI-NEXT: v_readfirstlane_b32 s7, v50
; SI-NEXT: s_waitcnt vmcnt(8)
-; SI-NEXT: v_readfirstlane_b32 s7, v51
+; SI-NEXT: v_readfirstlane_b32 s80, v51
; SI-NEXT: buffer_load_dword v39, off, s[0:3], s32 offset:16
; SI-NEXT: buffer_load_dword v48, off, s[0:3], s32 offset:12
; SI-NEXT: buffer_load_dword v49, off, s[0:3], s32 offset:8
; SI-NEXT: buffer_load_dword v50, off, s[0:3], s32 offset:4
; SI-NEXT: buffer_load_dword v51, off, s[0:3], s32
; SI-NEXT: s_waitcnt vmcnt(12)
-; SI-NEXT: v_readfirstlane_b32 s15, v31
+; SI-NEXT: v_readfirstlane_b32 s41, v31
; SI-NEXT: s_waitcnt vmcnt(11)
-; SI-NEXT: v_readfirstlane_b32 s97, v32
+; SI-NEXT: v_readfirstlane_b32 s98, v32
; SI-NEXT: s_waitcnt vmcnt(10)
-; SI-NEXT: v_readfirstlane_b32 s44, v33
+; SI-NEXT: v_readfirstlane_b32 s99, v33
; SI-NEXT: s_waitcnt vmcnt(9)
; SI-NEXT: v_readfirstlane_b32 s9, v34
; SI-NEXT: s_waitcnt vmcnt(8)
-; SI-NEXT: v_readfirstlane_b32 s80, v35
+; SI-NEXT: v_readfirstlane_b32 s82, v35
; SI-NEXT: s_waitcnt vmcnt(7)
-; SI-NEXT: v_readfirstlane_b32 s86, v36
+; SI-NEXT: v_readfirstlane_b32 s68, v36
; SI-NEXT: s_waitcnt vmcnt(6)
-; SI-NEXT: v_readfirstlane_b32 s85, v37
+; SI-NEXT: v_readfirstlane_b32 s67, v37
; SI-NEXT: s_waitcnt vmcnt(5)
-; SI-NEXT: v_readfirstlane_b32 s8, v38
+; SI-NEXT: v_readfirstlane_b32 s85, v38
; SI-NEXT: s_waitcnt vmcnt(4)
-; SI-NEXT: v_readfirstlane_b32 s12, v39
+; SI-NEXT: v_readfirstlane_b32 s87, v39
; SI-NEXT: s_waitcnt vmcnt(3)
-; SI-NEXT: v_readfirstlane_b32 s65, v48
+; SI-NEXT: v_readfirstlane_b32 s55, v48
; SI-NEXT: s_waitcnt vmcnt(2)
-; SI-NEXT: v_readfirstlane_b32 s64, v49
-; SI-NEXT: v_writelane_b32 v42, s64, 10
-; SI-NEXT: s_waitcnt vmcnt(1)
-; SI-NEXT: v_readfirstlane_b32 s67, v50
-; SI-NEXT: v_writelane_b32 v42, s65, 11
-; SI-NEXT: s_waitcnt vmcnt(0)
-; SI-NEXT: v_readfirstlane_b32 s84, v51
-; SI-NEXT: v_writelane_b32 v42, s67, 12
-; SI-NEXT: v_writelane_b32 v42, s84, 13
-; SI-NEXT: v_writelane_b32 v42, s85, 14
-; SI-NEXT: v_writelane_b32 v42, s86, 15
-; SI-NEXT: v_writelane_b32 v42, s87, 16
-; SI-NEXT: v_writelane_b32 v42, s8, 17
-; SI-NEXT: v_writelane_b32 v42, s99, 18
-; SI-NEXT: v_writelane_b32 v42, s12, 19
-; SI-NEXT: v_writelane_b32 v42, s44, 20
-; SI-NEXT: v_writelane_b32 v42, s97, 21
-; SI-NEXT: v_writelane_b32 v42, s83, 22
-; SI-NEXT: v_writelane_b32 v42, s82, 23
-; SI-NEXT: v_writelane_b32 v42, s98, 24
-; SI-NEXT: v_writelane_b32 v42, s96, 25
-; SI-NEXT: v_writelane_b32 v42, s81, 26
-; SI-NEXT: v_writelane_b32 v42, s9, 27
-; SI-NEXT: v_writelane_b32 v42, s41, 28
-; SI-NEXT: v_writelane_b32 v42, s80, 29
-; SI-NEXT: v_writelane_b32 v42, s7, 30
-; SI-NEXT: v_writelane_b32 v42, s56, 31
-; SI-NEXT: v_writelane_b32 v42, s26, 32
-; SI-NEXT: v_writelane_b32 v42, s15, 33
-; SI-NEXT: v_writelane_b32 v42, s14, 34
-; SI-NEXT: v_writelane_b32 v42, s69, 35
-; SI-NEXT: v_writelane_b32 v42, s71, 36
-; SI-NEXT: v_writelane_b32 v42, s70, 37
-; SI-NEXT: v_writelane_b32 v42, s68, 38
-; SI-NEXT: v_writelane_b32 v42, s74, 39
-; SI-NEXT: v_writelane_b32 v42, s46, 40
-; SI-NEXT: v_writelane_b32 v42, s11, 41
-; SI-NEXT: v_writelane_b32 v42, s10, 42
-; SI-NEXT: v_writelane_b32 v42, s62, 43
-; SI-NEXT: v_writelane_b32 v42, s66, 44
-; SI-NEXT: v_writelane_b32 v42, s58, 45
-; SI-NEXT: v_writelane_b32 v42, s28, 46
-; SI-NEXT: v_writelane_b32 v42, s27, 47
-; SI-NEXT: v_writelane_b32 v42, s78, 48
-; SI-NEXT: v_writelane_b32 v42, s24, 49
+; SI-NEXT: v_readfirstlane_b32 s54, v49
+; SI-NEXT: v_writelane_b32 v43, s54, 15
+; SI-NEXT: s_waitcnt vmcnt(1)
+; SI-NEXT: v_readfirstlane_b32 s64, v50
+; SI-NEXT: v_writelane_b32 v43, s55, 16
+; SI-NEXT: s_waitcnt vmcnt(0)
+; SI-NEXT: v_readfirstlane_b32 s65, v51
+; SI-NEXT: v_writelane_b32 v43, s64, 17
+; SI-NEXT: v_writelane_b32 v43, s65, 18
+; SI-NEXT: v_writelane_b32 v43, s67, 19
+; SI-NEXT: v_writelane_b32 v43, s68, 20
+; SI-NEXT: v_writelane_b32 v43, s84, 21
+; SI-NEXT: v_writelane_b32 v43, s85, 22
+; SI-NEXT: v_writelane_b32 v43, s86, 23
+; SI-NEXT: v_writelane_b32 v43, s87, 24
+; SI-NEXT: v_writelane_b32 v43, s99, 25
+; SI-NEXT: v_writelane_b32 v43, s98, 26
+; SI-NEXT: v_writelane_b32 v43, s96, 27
+; SI-NEXT: v_writelane_b32 v43, s7, 28
+; SI-NEXT: v_writelane_b32 v43, s12, 29
+; SI-NEXT: v_writelane_b32 v43, s83, 30
+; SI-NEXT: v_writelane_b32 v43, s15, 31
+; SI-NEXT: v_writelane_b32 v43, s9, 32
+; SI-NEXT: v_writelane_b32 v43, s97, 33
+; SI-NEXT: v_writelane_b32 v43, s82, 34
+; SI-NEXT: v_writelane_b32 v43, s80, 35
+; SI-NEXT: v_writelane_b32 v43, s8, 36
+; SI-NEXT: v_writelane_b32 v43, s62, 37
+; SI-NEXT: v_writelane_b32 v43, s41, 38
+; SI-NEXT: v_writelane_b32 v43, s14, 39
+; SI-NEXT: v_writelane_b32 v43, s70, 40
+; SI-NEXT: v_writelane_b32 v43, s81, 41
+; SI-NEXT: v_writelane_b32 v43, s71, 42
+; SI-NEXT: v_writelane_b32 v43, s69, 43
+; SI-NEXT: v_writelane_b32 v43, s46, 44
+; SI-NEXT: v_writelane_b32 v43, s59, 45
+; SI-NEXT: v_writelane_b32 v43, s11, 46
+; SI-NEXT: v_writelane_b32 v43, s10, 47
+; SI-NEXT: v_writelane_b32 v43, s40, 48
+; SI-NEXT: v_writelane_b32 v43, s66, 49
+; SI-NEXT: v_writelane_b32 v43, s29, 50
+; SI-NEXT: v_writelane_b32 v43, s28, 51
+; SI-NEXT: v_writelane_b32 v43, s27, 52
+; SI-NEXT: v_writelane_b32 v43, s45, 53
+; SI-NEXT: v_writelane_b32 v43, s47, 54
+; SI-NEXT: v_writelane_b32 v43, s61, 55
; SI-NEXT: s_cbranch_scc0 .LBB89_4
; SI-NEXT: ; %bb.1: ; %cmp.false
-; SI-NEXT: v_readlane_b32 s4, v43, 3
+; SI-NEXT: v_readlane_b32 s4, v44, 7
; SI-NEXT: s_and_b32 s4, s4, 0xff
-; SI-NEXT: v_readlane_b32 s5, v43, 2
+; SI-NEXT: v_readlane_b32 s5, v44, 6
; SI-NEXT: s_lshl_b32 s4, s4, 16
; SI-NEXT: s_lshl_b32 s5, s5, 24
; SI-NEXT: s_or_b32 s4, s5, s4
-; SI-NEXT: v_writelane_b32 v42, s4, 56
-; SI-NEXT: v_readlane_b32 s4, v43, 1
+; SI-NEXT: ; implicit-def: $vgpr42 : SGPR spill to VGPR lane
+; SI-NEXT: v_readlane_b32 s5, v44, 4
+; SI-NEXT: v_writelane_b32 v42, s4, 0
+; SI-NEXT: v_readlane_b32 s4, v44, 5
; SI-NEXT: s_and_b32 s4, s4, 0xff
-; SI-NEXT: v_readlane_b32 s5, v43, 0
; SI-NEXT: s_lshl_b32 s4, s4, 16
; SI-NEXT: s_lshl_b32 s5, s5, 24
; SI-NEXT: s_or_b32 s4, s5, s4
-; SI-NEXT: v_writelane_b32 v42, s4, 57
+; SI-NEXT: v_writelane_b32 v42, s4, 1
; SI-NEXT: s_and_b32 s4, s20, 0xff
-; SI-NEXT: s_lshl_b32 s5, s72, 8
-; SI-NEXT: s_or_b32 s4, s4, s5
+; SI-NEXT: s_lshl_b32 s5, s6, 8
+; SI-NEXT: s_or_b32 s43, s4, s5
; SI-NEXT: s_and_b32 s5, s22, 0xff
; SI-NEXT: s_lshl_b32 s5, s5, 16
-; SI-NEXT: s_mov_b32 s22, s6
; SI-NEXT: s_lshl_b32 s6, s23, 24
-; SI-NEXT: v_writelane_b32 v42, s4, 58
; SI-NEXT: s_or_b32 s4, s6, s5
-; SI-NEXT: s_and_b32 s5, s60, 0xff
+; SI-NEXT: s_and_b32 s5, s24, 0xff
; SI-NEXT: s_lshl_b32 s5, s5, 16
; SI-NEXT: s_lshl_b32 s6, s25, 24
-; SI-NEXT: v_writelane_b32 v42, s4, 59
-; SI-NEXT: s_or_b32 s5, s6, s5
-; SI-NEXT: v_writelane_b32 v42, s5, 60
-; SI-NEXT: s_and_b32 s5, s79, 0xff
+; SI-NEXT: v_writelane_b32 v42, s4, 2
+; SI-NEXT: s_or_b32 s4, s6, s5
+; SI-NEXT: v_readlane_b32 s5, v44, 3
+; SI-NEXT: s_and_b32 s5, s5, 0xff
+; SI-NEXT: v_readlane_b32 s6, v44, 2
; SI-NEXT: s_lshl_b32 s5, s5, 16
-; SI-NEXT: s_lshl_b32 s6, s76, 24
-; SI-NEXT: s_or_b32 s5, s6, s5
-; SI-NEXT: v_writelane_b32 v42, s5, 61
-; SI-NEXT: s_and_b32 s5, s77, 0xff
-; SI-NEXT: s_lshl_b32 s6, s29, 8
-; SI-NEXT: s_or_b32 s5, s5, s6
+; SI-NEXT: s_lshl_b32 s6, s6, 24
+; SI-NEXT: s_or_b32 s25, s6, s5
+; SI-NEXT: v_readlane_b32 s5, v44, 1
+; SI-NEXT: v_readlane_b32 s6, v44, 0
+; SI-NEXT: s_and_b32 s5, s5, 0xff
+; SI-NEXT: s_lshl_b32 s6, s6, 8
+; SI-NEXT: s_or_b32 s24, s5, s6
; SI-NEXT: s_and_b32 s6, s16, 0xff
; SI-NEXT: s_lshl_b32 s6, s6, 16
; SI-NEXT: s_lshl_b32 s16, s17, 24
-; SI-NEXT: s_or_b32 s6, s16, s6
-; SI-NEXT: v_writelane_b32 v42, s6, 62
-; SI-NEXT: s_and_b32 s6, s89, 0xff
+; SI-NEXT: v_writelane_b32 v42, s4, 3
+; SI-NEXT: s_or_b32 s4, s16, s6
+; SI-NEXT: s_and_b32 s6, s79, 0xff
; SI-NEXT: s_lshl_b32 s6, s6, 16
-; SI-NEXT: s_lshl_b32 s16, s88, 24
-; SI-NEXT: s_mov_b32 s4, s47
-; SI-NEXT: s_or_b32 s47, s16, s6
+; SI-NEXT: s_lshl_b32 s16, s78, 24
+; SI-NEXT: s_or_b32 s5, s16, s6
; SI-NEXT: s_and_b32 s6, s18, 0xff
; SI-NEXT: s_lshl_b32 s6, s6, 16
; SI-NEXT: s_lshl_b32 s16, s19, 24
-; SI-NEXT: s_or_b32 s25, s16, s6
-; SI-NEXT: s_and_b32 s6, s93, 0xff
-; SI-NEXT: s_lshl_b32 s16, s92, 8
+; SI-NEXT: s_or_b32 s75, s16, s6
+; SI-NEXT: s_and_b32 s6, s91, 0xff
+; SI-NEXT: s_lshl_b32 s16, s90, 8
; SI-NEXT: s_or_b32 s6, s6, s16
-; SI-NEXT: s_and_b32 s16, s90, 0xff
+; SI-NEXT: s_and_b32 s16, s88, 0xff
; SI-NEXT: s_lshl_b32 s16, s16, 16
-; SI-NEXT: s_lshl_b32 s17, s91, 24
-; SI-NEXT: s_or_b32 s92, s17, s16
-; SI-NEXT: s_and_b32 s16, vcc_hi, 0xff
+; SI-NEXT: s_lshl_b32 s17, s89, 24
+; SI-NEXT: s_or_b32 s78, s17, s16
+; SI-NEXT: s_and_b32 s16, s95, 0xff
; SI-NEXT: s_lshl_b32 s16, s16, 16
-; SI-NEXT: s_lshl_b32 s17, vcc_lo, 24
-; SI-NEXT: s_or_b32 s76, s17, s16
-; SI-NEXT: s_and_b32 s16, s94, 0xff
+; SI-NEXT: s_lshl_b32 s17, s94, 24
+; SI-NEXT: s_mov_b32 s23, s21
+; SI-NEXT: s_or_b32 s21, s17, s16
+; SI-NEXT: s_and_b32 s16, s92, 0xff
; SI-NEXT: s_lshl_b32 s16, s16, 16
-; SI-NEXT: s_lshl_b32 s17, s95, 24
-; SI-NEXT: s_or_b32 s91, s17, s16
-; SI-NEXT: s_and_b32 s16, s35, 0xff
-; SI-NEXT: s_lshl_b32 s17, s34, 8
+; SI-NEXT: s_lshl_b32 s17, s93, 24
+; SI-NEXT: s_or_b32 s79, s17, s16
+; SI-NEXT: s_and_b32 s16, s31, 0xff
+; SI-NEXT: s_lshl_b32 s17, s30, 8
; SI-NEXT: s_or_b32 s16, s16, s17
-; SI-NEXT: s_and_b32 s17, s30, 0xff
+; SI-NEXT: s_and_b32 s17, vcc_lo, 0xff
; SI-NEXT: s_lshl_b32 s17, s17, 16
-; SI-NEXT: s_lshl_b32 s18, s31, 24
-; SI-NEXT: s_or_b32 s77, s18, s17
-; SI-NEXT: s_and_b32 s17, s39, 0xff
+; SI-NEXT: s_lshl_b32 s18, vcc_hi, 24
+; SI-NEXT: s_or_b32 s17, s18, s17
+; SI-NEXT: v_writelane_b32 v43, s17, 56
+; SI-NEXT: s_and_b32 s17, s37, 0xff
; SI-NEXT: s_lshl_b32 s17, s17, 16
-; SI-NEXT: s_lshl_b32 s18, s38, 24
-; SI-NEXT: s_or_b32 s79, s18, s17
-; SI-NEXT: s_and_b32 s17, s36, 0xff
+; SI-NEXT: s_lshl_b32 s18, s36, 24
+; SI-NEXT: s_or_b32 s17, s18, s17
+; SI-NEXT: v_writelane_b32 v43, s17, 58
+; SI-NEXT: s_and_b32 s17, s34, 0xff
; SI-NEXT: s_lshl_b32 s17, s17, 16
-; SI-NEXT: s_lshl_b32 s18, s37, 24
-; SI-NEXT: s_or_b32 s93, s18, s17
-; SI-NEXT: s_and_b32 s17, s51, 0xff
-; SI-NEXT: s_lshl_b32 s18, s50, 8
+; SI-NEXT: s_lshl_b32 s18, s35, 24
+; SI-NEXT: s_or_b32 s17, s18, s17
+; SI-NEXT: v_writelane_b32 v43, s17, 57
+; SI-NEXT: s_and_b32 s17, s49, 0xff
+; SI-NEXT: s_lshl_b32 s18, s48, 8
; SI-NEXT: s_or_b32 s17, s17, s18
-; SI-NEXT: s_and_b32 s18, s48, 0xff
+; SI-NEXT: s_and_b32 s18, s38, 0xff
; SI-NEXT: s_lshl_b32 s18, s18, 16
-; SI-NEXT: s_lshl_b32 s19, s49, 24
-; SI-NEXT: s_or_b32 s89, s19, s18
-; SI-NEXT: s_and_b32 s18, s55, 0xff
+; SI-NEXT: s_lshl_b32 s19, s39, 24
+; SI-NEXT: s_or_b32 s18, s19, s18
+; SI-NEXT: v_writelane_b32 v43, s18, 59
+; SI-NEXT: s_and_b32 s18, s53, 0xff
; SI-NEXT: s_lshl_b32 s18, s18, 16
-; SI-NEXT: s_lshl_b32 s19, s54, 24
-; SI-NEXT: s_or_b32 s31, s19, s18
-; SI-NEXT: s_and_b32 s18, s52, 0xff
+; SI-NEXT: s_lshl_b32 s19, s52, 24
+; SI-NEXT: s_or_b32 s18, s19, s18
+; SI-NEXT: v_writelane_b32 v43, s18, 61
+; SI-NEXT: s_and_b32 s18, s50, 0xff
; SI-NEXT: s_lshl_b32 s18, s18, 16
-; SI-NEXT: s_lshl_b32 s19, s53, 24
-; SI-NEXT: s_or_b32 s94, s19, s18
-; SI-NEXT: s_and_b32 s18, s84, 0xff
-; SI-NEXT: s_lshl_b32 s19, s67, 8
+; SI-NEXT: s_lshl_b32 s19, s51, 24
+; SI-NEXT: s_or_b32 s18, s19, s18
+; SI-NEXT: v_writelane_b32 v43, s18, 60
+; SI-NEXT: s_and_b32 s18, s65, 0xff
+; SI-NEXT: s_lshl_b32 s19, s64, 8
; SI-NEXT: s_or_b32 s18, s18, s19
-; SI-NEXT: s_and_b32 s19, s64, 0xff
+; SI-NEXT: s_and_b32 s19, s54, 0xff
; SI-NEXT: s_lshl_b32 s19, s19, 16
-; SI-NEXT: s_lshl_b32 s20, s65, 24
-; SI-NEXT: s_or_b32 s60, s20, s19
-; SI-NEXT: s_and_b32 s19, s12, 0xff
+; SI-NEXT: s_lshl_b32 s20, s55, 24
+; SI-NEXT: s_or_b32 s19, s20, s19
+; SI-NEXT: v_writelane_b32 v43, s19, 62
+; SI-NEXT: s_and_b32 s19, s87, 0xff
; SI-NEXT: s_lshl_b32 s19, s19, 16
-; SI-NEXT: s_lshl_b32 s20, s8, 24
-; SI-NEXT: s_or_b32 s8, s20, s19
-; SI-NEXT: s_and_b32 s19, s85, 0xff
+; SI-NEXT: s_lshl_b32 s20, s85, 24
+; SI-NEXT: s_or_b32 s19, s20, s19
+; SI-NEXT: v_writelane_b32 v43, s19, 63
+; SI-NEXT: s_and_b32 s19, s67, 0xff
; SI-NEXT: s_lshl_b32 s19, s19, 16
-; SI-NEXT: s_lshl_b32 s20, s86, 24
-; SI-NEXT: s_or_b32 s12, s20, s19
-; SI-NEXT: s_and_b32 s19, s80, 0xff
+; SI-NEXT: s_lshl_b32 s20, s68, 24
+; SI-NEXT: s_or_b32 s95, s20, s19
+; SI-NEXT: s_and_b32 s19, s82, 0xff
; SI-NEXT: s_lshl_b32 s20, s9, 8
; SI-NEXT: s_or_b32 vcc_lo, s19, s20
-; SI-NEXT: s_and_b32 s19, s44, 0xff
+; SI-NEXT: s_and_b32 s19, s99, 0xff
; SI-NEXT: s_lshl_b32 s19, s19, 16
-; SI-NEXT: s_lshl_b32 s20, s97, 24
-; SI-NEXT: s_or_b32 s9, s20, s19
-; SI-NEXT: s_and_b32 s19, s15, 0xff
+; SI-NEXT: s_lshl_b32 s20, s98, 24
+; SI-NEXT: s_or_b32 s30, s20, s19
+; SI-NEXT: s_and_b32 s19, s41, 0xff
; SI-NEXT: s_lshl_b32 s19, s19, 16
-; SI-NEXT: s_lshl_b32 s20, s7, 24
-; SI-NEXT: s_or_b32 s7, s20, s19
-; SI-NEXT: s_and_b32 s19, s82, 0xff
+; SI-NEXT: s_lshl_b32 s20, s80, 24
+; SI-NEXT: s_or_b32 s31, s20, s19
+; SI-NEXT: s_and_b32 s19, s7, 0xff
; SI-NEXT: s_lshl_b32 s19, s19, 16
-; SI-NEXT: s_lshl_b32 s20, s83, 24
-; SI-NEXT: s_or_b32 s23, s20, s19
-; SI-NEXT: s_and_b32 s19, s26, 0xff
-; SI-NEXT: s_lshl_b32 s20, s81, 8
+; SI-NEXT: s_lshl_b32 s20, s96, 24
+; SI-NEXT: s_or_b32 s34, s20, s19
+; SI-NEXT: s_and_b32 s19, s62, 0xff
+; SI-NEXT: s_lshl_b32 s20, s15, 8
; SI-NEXT: s_or_b32 vcc_hi, s19, s20
-; SI-NEXT: s_and_b32 s19, s99, 0xff
-; SI-NEXT: v_writelane_b32 v42, s9, 50
+; SI-NEXT: s_and_b32 s19, s86, 0xff
; SI-NEXT: s_lshl_b32 s19, s19, 16
-; SI-NEXT: s_lshl_b32 s20, s87, 24
-; SI-NEXT: v_writelane_b32 v42, s7, 51
-; SI-NEXT: s_or_b32 s7, s20, s19
-; SI-NEXT: s_and_b32 s19, s56, 0xff
+; SI-NEXT: s_lshl_b32 s20, s84, 24
+; SI-NEXT: s_or_b32 s35, s20, s19
+; SI-NEXT: s_and_b32 s19, s8, 0xff
; SI-NEXT: s_lshl_b32 s19, s19, 16
-; SI-NEXT: s_lshl_b32 s20, s41, 24
-; SI-NEXT: v_writelane_b32 v42, s7, 52
-; SI-NEXT: s_or_b32 s7, s20, s19
-; SI-NEXT: s_and_b32 s19, s98, 0xff
+; SI-NEXT: s_lshl_b32 s20, s97, 24
+; SI-NEXT: s_or_b32 s36, s20, s19
+; SI-NEXT: s_and_b32 s19, s12, 0xff
; SI-NEXT: s_lshl_b32 s19, s19, 16
-; SI-NEXT: s_lshl_b32 s20, s96, 24
-; SI-NEXT: v_writelane_b32 v42, s7, 54
-; SI-NEXT: s_or_b32 s7, s20, s19
-; SI-NEXT: s_and_b32 s19, s46, 0xff
-; SI-NEXT: s_lshl_b32 s20, s74, 8
+; SI-NEXT: s_lshl_b32 s20, s83, 24
+; SI-NEXT: s_or_b32 s37, s20, s19
+; SI-NEXT: s_and_b32 s19, s59, 0xff
+; SI-NEXT: s_lshl_b32 s20, s46, 8
; SI-NEXT: s_or_b32 s84, s19, s20
-; SI-NEXT: s_and_b32 s19, s71, 0xff
+; SI-NEXT: s_and_b32 s19, s81, 0xff
; SI-NEXT: s_lshl_b32 s19, s19, 16
-; SI-NEXT: s_lshl_b32 s20, s70, 24
-; SI-NEXT: s_or_b32 s72, s20, s19
+; SI-NEXT: s_lshl_b32 s20, s71, 24
+; SI-NEXT: s_or_b32 s38, s20, s19
; SI-NEXT: s_and_b32 s19, s11, 0xff
; SI-NEXT: s_lshl_b32 s19, s19, 16
-; SI-NEXT: s_lshl_b32 s20, s68, 24
-; SI-NEXT: v_writelane_b32 v42, s7, 53
-; SI-NEXT: s_or_b32 s7, s20, s19
+; SI-NEXT: s_lshl_b32 s20, s69, 24
+; SI-NEXT: s_or_b32 s39, s20, s19
; SI-NEXT: s_and_b32 s19, s14, 0xff
; SI-NEXT: s_lshl_b32 s19, s19, 16
-; SI-NEXT: s_lshl_b32 s20, s69, 24
-; SI-NEXT: s_or_b32 s9, s20, s19
-; SI-NEXT: s_and_b32 s19, s58, 0xff
+; SI-NEXT: s_lshl_b32 s20, s70, 24
+; SI-NEXT: s_or_b32 s48, s20, s19
+; SI-NEXT: s_and_b32 s19, s29, 0xff
; SI-NEXT: s_lshl_b32 s20, s66, 8
; SI-NEXT: s_or_b32 s85, s19, s20
; SI-NEXT: s_and_b32 s19, s10, 0xff
; SI-NEXT: s_lshl_b32 s19, s19, 16
-; SI-NEXT: s_lshl_b32 s20, s62, 24
+; SI-NEXT: s_lshl_b32 s20, s40, 24
; SI-NEXT: s_or_b32 s49, s20, s19
; SI-NEXT: s_and_b32 s19, s27, 0xff
-; SI-NEXT: v_writelane_b32 v42, s9, 55
; SI-NEXT: s_lshl_b32 s19, s19, 16
; SI-NEXT: s_lshl_b32 s20, s28, 24
-; SI-NEXT: v_readlane_b32 s9, v43, 33
+; SI-NEXT: v_readlane_b32 s7, v44, 35
; SI-NEXT: s_or_b32 s50, s20, s19
-; SI-NEXT: s_and_b32 s19, s9, 0xff
-; SI-NEXT: v_readlane_b32 s9, v43, 32
+; SI-NEXT: s_and_b32 s19, s7, 0xff
; SI-NEXT: s_lshl_b32 s19, s19, 16
-; SI-NEXT: s_lshl_b32 s20, s9, 24
-; SI-NEXT: v_readlane_b32 s9, v43, 31
+; SI-NEXT: s_lshl_b32 s20, s44, 24
; SI-NEXT: s_or_b32 s51, s20, s19
-; SI-NEXT: s_and_b32 s19, s9, 0xff
-; SI-NEXT: v_readlane_b32 s9, v43, 30
-; SI-NEXT: s_lshl_b32 s20, s9, 8
-; SI-NEXT: v_readlane_b32 s9, v43, 29
+; SI-NEXT: s_and_b32 s19, s23, 0xff
+; SI-NEXT: s_lshl_b32 s20, s73, 8
; SI-NEXT: s_or_b32 s86, s19, s20
-; SI-NEXT: s_and_b32 s19, s9, 0xff
-; SI-NEXT: v_readlane_b32 s9, v43, 28
+; SI-NEXT: s_and_b32 s19, s72, 0xff
; SI-NEXT: s_lshl_b32 s19, s19, 16
-; SI-NEXT: s_lshl_b32 s20, s9, 24
-; SI-NEXT: v_readlane_b32 s9, v43, 27
+; SI-NEXT: s_lshl_b32 s20, s13, 24
; SI-NEXT: s_or_b32 s52, s20, s19
-; SI-NEXT: s_and_b32 s19, s9, 0xff
-; SI-NEXT: v_readlane_b32 s9, v43, 26
+; SI-NEXT: s_and_b32 s19, s42, 0xff
; SI-NEXT: s_lshl_b32 s19, s19, 16
-; SI-NEXT: s_lshl_b32 s20, s9, 24
-; SI-NEXT: v_readlane_b32 s9, v43, 25
+; SI-NEXT: s_lshl_b32 s20, s60, 24
; SI-NEXT: s_or_b32 s53, s20, s19
-; SI-NEXT: s_and_b32 s19, s9, 0xff
-; SI-NEXT: v_readlane_b32 s9, v43, 24
+; SI-NEXT: s_and_b32 s19, s45, 0xff
; SI-NEXT: s_lshl_b32 s19, s19, 16
-; SI-NEXT: s_lshl_b32 s20, s9, 24
-; SI-NEXT: v_readlane_b32 s9, v43, 23
+; SI-NEXT: s_lshl_b32 s20, s47, 24
; SI-NEXT: s_or_b32 s54, s20, s19
-; SI-NEXT: s_and_b32 s19, s9, 0xff
-; SI-NEXT: v_readlane_b32 s9, v43, 22
-; SI-NEXT: s_lshl_b32 s20, s9, 8
-; SI-NEXT: v_readlane_b32 s9, v43, 21
+; SI-NEXT: s_and_b32 s19, s76, 0xff
+; SI-NEXT: s_lshl_b32 s20, s74, 8
; SI-NEXT: s_or_b32 s87, s19, s20
-; SI-NEXT: s_and_b32 s19, s9, 0xff
-; SI-NEXT: v_readlane_b32 s9, v43, 20
+; SI-NEXT: s_and_b32 s19, s61, 0xff
; SI-NEXT: s_lshl_b32 s19, s19, 16
-; SI-NEXT: s_lshl_b32 s20, s9, 24
-; SI-NEXT: v_readlane_b32 s9, v43, 19
+; SI-NEXT: s_lshl_b32 s20, s56, 24
; SI-NEXT: s_or_b32 s55, s20, s19
-; SI-NEXT: s_mov_b32 s58, s9
-; SI-NEXT: s_and_b32 s19, s9, 0xff
-; SI-NEXT: v_readlane_b32 s9, v43, 18
+; SI-NEXT: s_and_b32 s19, s57, 0xff
; SI-NEXT: s_lshl_b32 s19, s19, 16
-; SI-NEXT: s_lshl_b32 s20, s9, 24
+; SI-NEXT: s_lshl_b32 s20, s63, 24
; SI-NEXT: s_or_b32 s64, s20, s19
-; SI-NEXT: s_and_b32 s19, s78, 0xff
+; SI-NEXT: s_and_b32 s19, s77, 0xff
; SI-NEXT: s_lshl_b32 s19, s19, 16
-; SI-NEXT: s_lshl_b32 s20, s24, 24
+; SI-NEXT: s_lshl_b32 s20, s58, 24
+; SI-NEXT: v_readlane_b32 s7, v44, 34
; SI-NEXT: s_or_b32 s65, s20, s19
-; SI-NEXT: s_and_b32 s19, s4, 0xff
-; SI-NEXT: s_lshl_b32 s20, s45, 8
+; SI-NEXT: s_and_b32 s19, s26, 0xff
+; SI-NEXT: s_mov_b32 s42, s7
+; SI-NEXT: s_lshl_b32 s20, s7, 8
+; SI-NEXT: v_readlane_b32 s7, v44, 33
+; SI-NEXT: s_mov_b32 s94, s26
; SI-NEXT: s_or_b32 s26, s19, s20
-; SI-NEXT: s_and_b32 s19, s13, 0xff
+; SI-NEXT: s_mov_b32 s47, s7
+; SI-NEXT: s_and_b32 s19, s7, 0xff
+; SI-NEXT: v_readlane_b32 s7, v44, 32
+; SI-NEXT: s_mov_b32 s92, s56
; SI-NEXT: s_lshl_b32 s19, s19, 16
-; SI-NEXT: s_lshl_b32 s20, s57, 24
+; SI-NEXT: s_mov_b32 s56, s7
+; SI-NEXT: s_lshl_b32 s20, s7, 24
+; SI-NEXT: v_readlane_b32 s7, v44, 31
; SI-NEXT: s_or_b32 s66, s20, s19
-; SI-NEXT: s_and_b32 s19, s21, 0xff
+; SI-NEXT: s_mov_b32 s61, s7
+; SI-NEXT: s_and_b32 s19, s7, 0xff
+; SI-NEXT: v_readlane_b32 s7, v44, 30
; SI-NEXT: s_lshl_b32 s19, s19, 16
-; SI-NEXT: s_lshl_b32 s20, s73, 24
+; SI-NEXT: s_mov_b32 s60, s7
+; SI-NEXT: s_lshl_b32 s20, s7, 24
+; SI-NEXT: v_readlane_b32 s7, v44, 29
; SI-NEXT: s_or_b32 s67, s20, s19
-; SI-NEXT: s_and_b32 s19, s42, 0xff
-; SI-NEXT: v_readlane_b32 s88, v43, 17
+; SI-NEXT: s_mov_b32 s68, s7
+; SI-NEXT: s_and_b32 s19, s7, 0xff
+; SI-NEXT: v_readlane_b32 s7, v44, 28
; SI-NEXT: s_lshl_b32 s19, s19, 16
-; SI-NEXT: s_lshl_b32 s20, s59, 24
-; SI-NEXT: s_or_b32 s68, s20, s19
-; SI-NEXT: s_and_b32 s19, s63, 0xff
-; SI-NEXT: s_lshl_b32 s20, s88, 8
+; SI-NEXT: s_mov_b32 s59, s7
+; SI-NEXT: s_lshl_b32 s20, s7, 24
+; SI-NEXT: v_readlane_b32 s7, v44, 27
+; SI-NEXT: s_or_b32 s45, s20, s19
+; SI-NEXT: s_mov_b32 s46, s7
+; SI-NEXT: s_and_b32 s19, s7, 0xff
+; SI-NEXT: v_readlane_b32 s7, v44, 26
+; SI-NEXT: s_mov_b32 s69, s7
+; SI-NEXT: s_lshl_b32 s20, s7, 8
+; SI-NEXT: v_readlane_b32 s7, v44, 25
; SI-NEXT: s_or_b32 s27, s19, s20
-; SI-NEXT: s_and_b32 s19, s40, 0xff
+; SI-NEXT: s_mov_b32 s40, s7
+; SI-NEXT: s_and_b32 s19, s7, 0xff
+; SI-NEXT: v_readlane_b32 s7, v44, 24
+; SI-NEXT: s_mov_b32 s89, s76
+; SI-NEXT: s_mov_b32 s76, s58
; SI-NEXT: s_lshl_b32 s19, s19, 16
-; SI-NEXT: s_lshl_b32 s20, s43, 24
-; SI-NEXT: s_or_b32 s69, s20, s19
-; SI-NEXT: s_and_b32 s19, s61, 0xff
-; SI-NEXT: s_mov_b32 s39, s57
-; SI-NEXT: s_mov_b32 s57, s7
+; SI-NEXT: s_mov_b32 s58, s7
+; SI-NEXT: s_lshl_b32 s20, s7, 24
+; SI-NEXT: v_readlane_b32 s7, v44, 23
+; SI-NEXT: s_mov_b32 s93, s74
+; SI-NEXT: s_mov_b32 s88, s57
+; SI-NEXT: s_or_b32 s57, s20, s19
+; SI-NEXT: s_mov_b32 s74, s7
+; SI-NEXT: s_and_b32 s19, s7, 0xff
+; SI-NEXT: v_readlane_b32 s7, v44, 22
+; SI-NEXT: s_mov_b32 s90, s63
; SI-NEXT: s_lshl_b32 s19, s19, 16
-; SI-NEXT: s_lshl_b32 s20, s75, 24
-; SI-NEXT: v_readlane_b32 s7, v43, 16
+; SI-NEXT: s_mov_b32 s63, s7
+; SI-NEXT: s_lshl_b32 s20, s7, 24
+; SI-NEXT: v_readlane_b32 s7, v44, 21
; SI-NEXT: s_or_b32 s70, s20, s19
-; SI-NEXT: s_mov_b32 s10, s7
+; SI-NEXT: s_mov_b32 s71, s7
; SI-NEXT: s_and_b32 s19, s7, 0xff
-; SI-NEXT: v_readlane_b32 s7, v43, 15
+; SI-NEXT: v_readlane_b32 s7, v44, 20
; SI-NEXT: s_lshl_b32 s19, s19, 16
-; SI-NEXT: s_mov_b32 s71, s7
+; SI-NEXT: s_mov_b32 s81, s7
; SI-NEXT: s_lshl_b32 s20, s7, 24
-; SI-NEXT: v_readlane_b32 s7, v43, 14
-; SI-NEXT: s_or_b32 s62, s20, s19
-; SI-NEXT: s_mov_b32 s15, s7
-; SI-NEXT: s_and_b32 s19, s7, 0xff
-; SI-NEXT: v_readlane_b32 s7, v43, 13
+; SI-NEXT: v_readlane_b32 s7, v44, 19
+; SI-NEXT: s_or_b32 s13, s20, s19
; SI-NEXT: s_mov_b32 s41, s7
+; SI-NEXT: s_and_b32 s19, s7, 0xff
+; SI-NEXT: v_readlane_b32 s7, v44, 18
+; SI-NEXT: s_mov_b32 s14, s7
; SI-NEXT: s_lshl_b32 s20, s7, 8
-; SI-NEXT: v_readlane_b32 s7, v43, 12
+; SI-NEXT: v_readlane_b32 s7, v44, 17
; SI-NEXT: s_or_b32 s29, s19, s20
-; SI-NEXT: s_mov_b32 s14, s7
+; SI-NEXT: s_mov_b32 s10, s7
; SI-NEXT: s_and_b32 s19, s7, 0xff
-; SI-NEXT: v_readlane_b32 s7, v43, 11
+; SI-NEXT: v_readlane_b32 s7, v44, 16
; SI-NEXT: s_lshl_b32 s19, s19, 16
; SI-NEXT: s_mov_b32 s9, s7
; SI-NEXT: s_lshl_b32 s20, s7, 24
-; SI-NEXT: v_readlane_b32 s7, v43, 10
+; SI-NEXT: v_readlane_b32 s7, v44, 15
; SI-NEXT: s_or_b32 s80, s20, s19
-; SI-NEXT: s_mov_b32 s56, s7
+; SI-NEXT: s_mov_b32 s8, s7
; SI-NEXT: s_and_b32 s19, s7, 0xff
-; SI-NEXT: v_readlane_b32 s7, v43, 9
+; SI-NEXT: v_readlane_b32 s7, v44, 14
; SI-NEXT: s_lshl_b32 s19, s19, 16
-; SI-NEXT: s_mov_b32 s81, s7
+; SI-NEXT: s_mov_b32 s15, s7
; SI-NEXT: s_lshl_b32 s20, s7, 24
-; SI-NEXT: v_readlane_b32 s7, v43, 8
-; SI-NEXT: s_or_b32 s11, s20, s19
-; SI-NEXT: s_mov_b32 s82, s7
+; SI-NEXT: v_readlane_b32 s7, v44, 13
+; SI-NEXT: s_mov_b32 s72, s25
+; SI-NEXT: s_or_b32 s25, s20, s19
+; SI-NEXT: s_mov_b32 s83, s7
; SI-NEXT: s_and_b32 s19, s7, 0xff
-; SI-NEXT: v_readlane_b32 s7, v43, 7
+; SI-NEXT: v_readlane_b32 s7, v44, 12
; SI-NEXT: s_lshl_b32 s19, s19, 16
-; SI-NEXT: s_mov_b32 s96, s7
+; SI-NEXT: s_mov_b32 s97, s7
; SI-NEXT: s_lshl_b32 s20, s7, 24
-; SI-NEXT: v_readlane_b32 s7, v43, 6
-; SI-NEXT: s_mov_b32 s36, s63
-; SI-NEXT: s_mov_b32 s63, s93
-; SI-NEXT: s_mov_b32 s93, s61
-; SI-NEXT: s_mov_b32 s61, s91
-; SI-NEXT: s_mov_b32 s91, s75
-; SI-NEXT: s_mov_b32 s75, s92
-; SI-NEXT: s_or_b32 s92, s20, s19
-; SI-NEXT: s_mov_b32 s98, s7
-; SI-NEXT: s_and_b32 s19, s7, 0xff
-; SI-NEXT: v_readlane_b32 s7, v43, 5
+; SI-NEXT: v_readlane_b32 s7, v44, 11
+; SI-NEXT: s_or_b32 s82, s20, s19
; SI-NEXT: s_mov_b32 s44, s7
+; SI-NEXT: s_and_b32 s19, s7, 0xff
+; SI-NEXT: v_readlane_b32 s7, v44, 10
; SI-NEXT: s_lshl_b32 s20, s7, 8
-; SI-NEXT: v_readlane_b32 s7, v43, 4
-; SI-NEXT: s_mov_b32 s48, s13
-; SI-NEXT: s_mov_b32 s13, s94
-; SI-NEXT: s_mov_b32 s94, s21
+; SI-NEXT: v_readlane_b32 s12, v44, 9
+; SI-NEXT: v_readlane_b32 s22, v44, 8
+; SI-NEXT: s_mov_b32 s91, s77
+; SI-NEXT: s_mov_b32 s77, s21
; SI-NEXT: s_or_b32 s21, s19, s20
-; SI-NEXT: s_and_b32 s19, s7, 0xff
-; SI-NEXT: s_mov_b32 s95, s4
+; SI-NEXT: s_and_b32 s19, s12, 0xff
; SI-NEXT: s_lshl_b32 s19, s19, 16
; SI-NEXT: s_lshl_b32 s20, s22, 24
-; SI-NEXT: v_readlane_b32 s4, v42, 58
-; SI-NEXT: s_mov_b32 s46, s45
-; SI-NEXT: s_mov_b32 s34, s73
-; SI-NEXT: s_mov_b32 s73, s12
-; SI-NEXT: s_mov_b32 s37, s42
-; SI-NEXT: s_mov_b32 s38, s59
-; SI-NEXT: s_mov_b32 s59, s8
-; SI-NEXT: s_mov_b32 s30, s88
-; SI-NEXT: s_mov_b32 s88, s31
-; SI-NEXT: s_mov_b32 s78, s40
-; SI-NEXT: s_mov_b32 s31, s43
-; SI-NEXT: s_mov_b32 s12, s7
-; SI-NEXT: s_mov_b32 s7, s22
-; SI-NEXT: s_or_b32 s83, s20, s19
-; SI-NEXT: s_lshl_b32 s20, s4, 16
-; SI-NEXT: s_lshl_b32 s74, s5, 16
+; SI-NEXT: s_mov_b32 s73, s4
+; SI-NEXT: s_mov_b32 s11, s7
+; SI-NEXT: s_mov_b32 s7, s12
+; SI-NEXT: s_mov_b32 s62, s22
+; SI-NEXT: s_or_b32 s28, s20, s19
+; SI-NEXT: s_lshl_b32 s20, s43, 16
+; SI-NEXT: s_lshl_b32 s23, s24, 16
; SI-NEXT: s_lshl_b32 s22, s6, 16
; SI-NEXT: s_lshl_b32 s16, s16, 16
; SI-NEXT: s_lshl_b32 s19, s17, 16
@@ -157795,50 +157840,48 @@ define inreg <64 x bfloat> @bitcast_v128i8_to_v64bf16_scalar(<128 x i8> inreg %a
; SI-NEXT: s_lshl_b32 s17, vcc_lo, 16
; SI-NEXT: s_lshl_b32 s6, vcc_hi, 16
; SI-NEXT: s_lshl_b32 s99, s84, 16
-; SI-NEXT: s_lshl_b32 s8, s85, 16
-; SI-NEXT: s_lshl_b32 s97, s86, 16
-; SI-NEXT: s_lshl_b32 s28, s87, 16
+; SI-NEXT: s_lshl_b32 s98, s85, 16
+; SI-NEXT: s_lshl_b32 s12, s86, 16
+; SI-NEXT: s_lshl_b32 s96, s87, 16
; SI-NEXT: s_lshl_b32 s87, s26, 16
-; SI-NEXT: v_readlane_b32 s26, v42, 56
+; SI-NEXT: v_readlane_b32 s26, v42, 0
; SI-NEXT: s_lshl_b32 s86, s27, 16
-; SI-NEXT: v_readlane_b32 s27, v42, 57
-; SI-NEXT: v_readlane_b32 s35, v42, 61
+; SI-NEXT: v_readlane_b32 s27, v42, 1
; SI-NEXT: s_lshl_b32 s85, s29, 16
-; SI-NEXT: v_readlane_b32 s29, v42, 60
-; SI-NEXT: v_readlane_b32 s24, v42, 59
-; SI-NEXT: v_readlane_b32 s90, v42, 62
+; SI-NEXT: v_readlane_b32 s29, v42, 3
+; SI-NEXT: v_readlane_b32 s24, v42, 2
; SI-NEXT: s_lshl_b32 s84, s21, 16
-; SI-NEXT: s_mov_b32 s21, s47
+; SI-NEXT: s_mov_b32 s21, s5
; SI-NEXT: s_cbranch_execnz .LBB89_3
; SI-NEXT: .LBB89_2: ; %cmp.true
-; SI-NEXT: s_add_i32 s4, s98, 3
+; SI-NEXT: s_add_i32 s4, s44, 3
; SI-NEXT: s_and_b32 s4, s4, 0xff
-; SI-NEXT: s_lshl_b32 s5, s44, 8
-; SI-NEXT: s_add_i32 s6, s12, 3
+; SI-NEXT: s_lshl_b32 s5, s11, 8
+; SI-NEXT: s_add_i32 s6, s7, 3
; SI-NEXT: s_or_b32 s4, s5, s4
; SI-NEXT: s_and_b32 s6, s6, 0xff
-; SI-NEXT: s_lshl_b32 s5, s7, 24
+; SI-NEXT: s_lshl_b32 s5, s62, 24
; SI-NEXT: s_lshl_b32 s6, s6, 16
; SI-NEXT: s_addk_i32 s4, 0x300
; SI-NEXT: s_or_b32 s5, s5, s6
; SI-NEXT: s_and_b32 s4, s4, 0xffff
; SI-NEXT: s_or_b32 s4, s5, s4
-; SI-NEXT: s_add_i32 s5, s56, 3
+; SI-NEXT: s_add_i32 s5, s8, 3
; SI-NEXT: s_and_b32 s5, s5, 0xff
-; SI-NEXT: s_lshl_b32 s6, s81, 8
-; SI-NEXT: s_add_i32 s16, s82, 3
+; SI-NEXT: s_lshl_b32 s6, s15, 8
+; SI-NEXT: s_add_i32 s16, s83, 3
; SI-NEXT: s_or_b32 s5, s6, s5
; SI-NEXT: s_and_b32 s16, s16, 0xff
-; SI-NEXT: s_lshl_b32 s6, s96, 24
+; SI-NEXT: s_lshl_b32 s6, s97, 24
; SI-NEXT: s_lshl_b32 s16, s16, 16
; SI-NEXT: s_addk_i32 s5, 0x300
; SI-NEXT: s_or_b32 s6, s6, s16
; SI-NEXT: s_and_b32 s5, s5, 0xffff
; SI-NEXT: s_or_b32 s5, s6, s5
-; SI-NEXT: s_add_i32 s6, s15, 3
+; SI-NEXT: s_add_i32 s6, s41, 3
; SI-NEXT: s_and_b32 s6, s6, 0xff
-; SI-NEXT: s_lshl_b32 s16, s41, 8
-; SI-NEXT: s_add_i32 s17, s14, 3
+; SI-NEXT: s_lshl_b32 s16, s14, 8
+; SI-NEXT: s_add_i32 s17, s10, 3
; SI-NEXT: s_or_b32 s6, s16, s6
; SI-NEXT: s_and_b32 s17, s17, 0xff
; SI-NEXT: s_lshl_b32 s16, s9, 24
@@ -157847,162 +157890,156 @@ define inreg <64 x bfloat> @bitcast_v128i8_to_v64bf16_scalar(<128 x i8> inreg %a
; SI-NEXT: s_or_b32 s16, s16, s17
; SI-NEXT: s_and_b32 s6, s6, 0xffff
; SI-NEXT: s_or_b32 s6, s16, s6
-; SI-NEXT: s_add_i32 s16, s93, 3
+; SI-NEXT: s_add_i32 s16, s74, 3
; SI-NEXT: s_and_b32 s16, s16, 0xff
-; SI-NEXT: s_lshl_b32 s17, s91, 8
-; SI-NEXT: s_add_i32 s18, s10, 3
+; SI-NEXT: s_lshl_b32 s17, s63, 8
+; SI-NEXT: s_add_i32 s18, s71, 3
; SI-NEXT: s_or_b32 s16, s17, s16
; SI-NEXT: s_and_b32 s18, s18, 0xff
-; SI-NEXT: s_lshl_b32 s17, s71, 24
+; SI-NEXT: s_lshl_b32 s17, s81, 24
; SI-NEXT: s_lshl_b32 s18, s18, 16
; SI-NEXT: s_addk_i32 s16, 0x300
; SI-NEXT: s_or_b32 s17, s17, s18
; SI-NEXT: s_and_b32 s16, s16, 0xffff
; SI-NEXT: s_or_b32 s16, s17, s16
-; SI-NEXT: s_add_i32 s17, s36, 3
+; SI-NEXT: s_add_i32 s17, s46, 3
; SI-NEXT: s_and_b32 s17, s17, 0xff
-; SI-NEXT: s_lshl_b32 s18, s30, 8
-; SI-NEXT: s_add_i32 s19, s78, 3
+; SI-NEXT: s_lshl_b32 s18, s69, 8
+; SI-NEXT: s_add_i32 s19, s40, 3
; SI-NEXT: s_or_b32 s17, s18, s17
; SI-NEXT: s_and_b32 s19, s19, 0xff
-; SI-NEXT: s_lshl_b32 s18, s31, 24
+; SI-NEXT: s_lshl_b32 s18, s58, 24
; SI-NEXT: s_lshl_b32 s19, s19, 16
; SI-NEXT: s_addk_i32 s17, 0x300
; SI-NEXT: s_or_b32 s18, s18, s19
; SI-NEXT: s_and_b32 s17, s17, 0xffff
; SI-NEXT: s_or_b32 s17, s18, s17
-; SI-NEXT: s_add_i32 s18, s94, 3
+; SI-NEXT: s_add_i32 s18, s61, 3
; SI-NEXT: s_and_b32 s18, s18, 0xff
-; SI-NEXT: s_lshl_b32 s19, s34, 8
-; SI-NEXT: s_add_i32 s20, s37, 3
+; SI-NEXT: s_lshl_b32 s19, s60, 8
+; SI-NEXT: s_add_i32 s20, s68, 3
; SI-NEXT: s_or_b32 s18, s19, s18
; SI-NEXT: s_and_b32 s20, s20, 0xff
-; SI-NEXT: s_lshl_b32 s19, s38, 24
+; SI-NEXT: s_lshl_b32 s19, s59, 24
; SI-NEXT: s_lshl_b32 s20, s20, 16
; SI-NEXT: s_addk_i32 s18, 0x300
; SI-NEXT: s_or_b32 s19, s19, s20
; SI-NEXT: s_and_b32 s18, s18, 0xffff
; SI-NEXT: s_or_b32 s18, s19, s18
-; SI-NEXT: s_add_i32 s19, s95, 3
+; SI-NEXT: s_add_i32 s19, s94, 3
; SI-NEXT: s_and_b32 s19, s19, 0xff
-; SI-NEXT: s_lshl_b32 s20, s46, 8
-; SI-NEXT: s_add_i32 s22, s48, 3
+; SI-NEXT: s_lshl_b32 s20, s42, 8
+; SI-NEXT: s_add_i32 s22, s47, 3
; SI-NEXT: s_or_b32 s19, s20, s19
; SI-NEXT: s_and_b32 s22, s22, 0xff
-; SI-NEXT: s_lshl_b32 s20, s39, 24
+; SI-NEXT: s_lshl_b32 s20, s56, 24
; SI-NEXT: s_lshl_b32 s22, s22, 16
; SI-NEXT: s_addk_i32 s19, 0x300
; SI-NEXT: s_or_b32 s20, s20, s22
; SI-NEXT: s_and_b32 s19, s19, 0xffff
; SI-NEXT: s_or_b32 s19, s20, s19
-; SI-NEXT: s_add_i32 s20, s58, 3
-; SI-NEXT: v_readlane_b32 s7, v43, 18
+; SI-NEXT: s_add_i32 s20, s88, 3
; SI-NEXT: s_and_b32 s20, s20, 0xff
-; SI-NEXT: s_lshl_b32 s22, s7, 8
-; SI-NEXT: v_readlane_b32 s7, v42, 49
+; SI-NEXT: s_lshl_b32 s22, s90, 8
+; SI-NEXT: s_add_i32 s23, s91, 3
; SI-NEXT: s_or_b32 s20, s22, s20
-; SI-NEXT: s_lshl_b32 s22, s7, 24
-; SI-NEXT: v_readlane_b32 s7, v42, 48
-; SI-NEXT: s_add_i32 s23, s7, 3
; SI-NEXT: s_and_b32 s23, s23, 0xff
+; SI-NEXT: s_lshl_b32 s22, s76, 24
; SI-NEXT: s_lshl_b32 s23, s23, 16
; SI-NEXT: s_addk_i32 s20, 0x300
; SI-NEXT: s_or_b32 s22, s22, s23
; SI-NEXT: s_and_b32 s20, s20, 0xffff
-; SI-NEXT: v_readlane_b32 s7, v43, 23
; SI-NEXT: s_or_b32 s20, s22, s20
-; SI-NEXT: s_add_i32 s22, s7, 3
-; SI-NEXT: v_readlane_b32 s7, v43, 22
+; SI-NEXT: s_add_i32 s22, s89, 3
+; SI-NEXT: v_readlane_b32 s7, v43, 55
; SI-NEXT: s_and_b32 s22, s22, 0xff
-; SI-NEXT: s_lshl_b32 s23, s7, 8
-; SI-NEXT: v_readlane_b32 s7, v43, 20
-; SI-NEXT: s_or_b32 s22, s23, s22
-; SI-NEXT: s_lshl_b32 s23, s7, 24
-; SI-NEXT: v_readlane_b32 s7, v43, 21
+; SI-NEXT: s_lshl_b32 s23, s93, 8
; SI-NEXT: s_add_i32 s60, s7, 3
+; SI-NEXT: s_or_b32 s22, s23, s22
; SI-NEXT: s_and_b32 s60, s60, 0xff
+; SI-NEXT: s_lshl_b32 s23, s92, 24
; SI-NEXT: s_lshl_b32 s60, s60, 16
; SI-NEXT: s_addk_i32 s22, 0x300
; SI-NEXT: s_or_b32 s23, s23, s60
; SI-NEXT: s_and_b32 s22, s22, 0xffff
-; SI-NEXT: v_readlane_b32 s7, v43, 27
+; SI-NEXT: v_readlane_b32 s7, v44, 47
; SI-NEXT: s_or_b32 s22, s23, s22
; SI-NEXT: s_add_i32 s23, s7, 3
-; SI-NEXT: v_readlane_b32 s7, v43, 26
+; SI-NEXT: v_readlane_b32 s7, v44, 45
; SI-NEXT: s_and_b32 s23, s23, 0xff
; SI-NEXT: s_lshl_b32 s60, s7, 8
-; SI-NEXT: v_readlane_b32 s7, v43, 24
+; SI-NEXT: v_readlane_b32 s7, v43, 54
; SI-NEXT: s_or_b32 s23, s60, s23
; SI-NEXT: s_lshl_b32 s60, s7, 24
-; SI-NEXT: v_readlane_b32 s7, v43, 25
+; SI-NEXT: v_readlane_b32 s7, v43, 53
; SI-NEXT: s_add_i32 s61, s7, 3
; SI-NEXT: s_and_b32 s61, s61, 0xff
; SI-NEXT: s_lshl_b32 s61, s61, 16
; SI-NEXT: s_addk_i32 s23, 0x300
; SI-NEXT: s_or_b32 s60, s60, s61
; SI-NEXT: s_and_b32 s23, s23, 0xffff
-; SI-NEXT: v_readlane_b32 s7, v43, 31
+; SI-NEXT: v_readlane_b32 s7, v44, 48
; SI-NEXT: s_or_b32 s23, s60, s23
; SI-NEXT: s_add_i32 s60, s7, 3
-; SI-NEXT: v_readlane_b32 s7, v43, 30
+; SI-NEXT: v_readlane_b32 s7, v44, 46
; SI-NEXT: s_and_b32 s60, s60, 0xff
; SI-NEXT: s_lshl_b32 s61, s7, 8
-; SI-NEXT: v_readlane_b32 s7, v43, 28
+; SI-NEXT: v_readlane_b32 s7, v44, 44
; SI-NEXT: s_or_b32 s60, s61, s60
; SI-NEXT: s_lshl_b32 s61, s7, 24
-; SI-NEXT: v_readlane_b32 s7, v43, 29
+; SI-NEXT: v_readlane_b32 s7, v44, 43
; SI-NEXT: s_add_i32 s62, s7, 3
-; SI-NEXT: v_readlane_b32 s7, v42, 47
+; SI-NEXT: v_readlane_b32 s7, v43, 52
; SI-NEXT: s_and_b32 s62, s62, 0xff
; SI-NEXT: s_add_i32 s59, s7, 3
-; SI-NEXT: v_readlane_b32 s7, v42, 46
+; SI-NEXT: v_readlane_b32 s7, v43, 51
; SI-NEXT: s_lshl_b32 s62, s62, 16
; SI-NEXT: s_addk_i32 s60, 0x300
; SI-NEXT: s_and_b32 s59, s59, 0xff
; SI-NEXT: s_lshl_b32 s58, s7, 8
-; SI-NEXT: v_readlane_b32 s7, v43, 32
+; SI-NEXT: v_readlane_b32 s7, v44, 42
; SI-NEXT: s_or_b32 s61, s61, s62
; SI-NEXT: s_and_b32 s60, s60, 0xffff
; SI-NEXT: s_or_b32 s58, s58, s59
; SI-NEXT: s_lshl_b32 s59, s7, 24
-; SI-NEXT: v_readlane_b32 s7, v43, 33
+; SI-NEXT: v_readlane_b32 s7, v44, 35
; SI-NEXT: s_or_b32 s60, s61, s60
; SI-NEXT: s_add_i32 s61, s7, 3
-; SI-NEXT: v_readlane_b32 s7, v42, 45
+; SI-NEXT: v_readlane_b32 s7, v43, 50
; SI-NEXT: s_add_i32 s57, s7, 3
-; SI-NEXT: v_readlane_b32 s7, v42, 44
+; SI-NEXT: v_readlane_b32 s7, v43, 49
; SI-NEXT: s_lshl_b32 s56, s7, 8
-; SI-NEXT: v_readlane_b32 s7, v42, 43
+; SI-NEXT: v_readlane_b32 s7, v43, 48
; SI-NEXT: s_lshl_b32 s47, s7, 24
-; SI-NEXT: v_readlane_b32 s7, v42, 42
+; SI-NEXT: v_readlane_b32 s7, v43, 47
; SI-NEXT: s_add_i32 s46, s7, 3
-; SI-NEXT: v_readlane_b32 s7, v42, 41
+; SI-NEXT: v_readlane_b32 s7, v43, 46
; SI-NEXT: s_add_i32 s45, s7, 3
-; SI-NEXT: v_readlane_b32 s7, v42, 38
+; SI-NEXT: v_readlane_b32 s7, v43, 43
; SI-NEXT: s_lshl_b32 s42, s7, 8
-; SI-NEXT: v_readlane_b32 s7, v42, 35
+; SI-NEXT: v_readlane_b32 s7, v43, 40
; SI-NEXT: s_lshl_b32 s15, s7, 24
-; SI-NEXT: v_readlane_b32 s7, v42, 34
+; SI-NEXT: v_readlane_b32 s7, v43, 39
; SI-NEXT: s_and_b32 s45, s45, 0xff
; SI-NEXT: s_add_i32 s14, s7, 3
; SI-NEXT: s_or_b32 s42, s42, s45
; SI-NEXT: s_and_b32 s14, s14, 0xff
; SI-NEXT: s_lshl_b32 s14, s14, 16
; SI-NEXT: s_addk_i32 s42, 0x300
-; SI-NEXT: v_readlane_b32 s7, v42, 40
+; SI-NEXT: v_readlane_b32 s7, v43, 45
; SI-NEXT: s_and_b32 s57, s57, 0xff
; SI-NEXT: s_or_b32 s14, s15, s14
; SI-NEXT: s_and_b32 s15, s42, 0xffff
; SI-NEXT: s_add_i32 s44, s7, 3
-; SI-NEXT: v_readlane_b32 s7, v42, 39
+; SI-NEXT: v_readlane_b32 s7, v43, 44
; SI-NEXT: s_or_b32 s56, s56, s57
; SI-NEXT: s_or_b32 s57, s14, s15
; SI-NEXT: s_and_b32 s14, s44, 0xff
; SI-NEXT: s_lshl_b32 s15, s7, 8
-; SI-NEXT: v_readlane_b32 s7, v42, 37
+; SI-NEXT: v_readlane_b32 s7, v43, 42
; SI-NEXT: s_or_b32 s14, s15, s14
; SI-NEXT: s_lshl_b32 s15, s7, 24
-; SI-NEXT: v_readlane_b32 s7, v42, 36
+; SI-NEXT: v_readlane_b32 s7, v43, 41
; SI-NEXT: s_add_i32 s40, s7, 3
; SI-NEXT: s_and_b32 s61, s61, 0xff
; SI-NEXT: s_and_b32 s40, s40, 0xff
@@ -158017,15 +158054,15 @@ define inreg <64 x bfloat> @bitcast_v128i8_to_v64bf16_scalar(<128 x i8> inreg %a
; SI-NEXT: s_or_b32 s58, s59, s58
; SI-NEXT: s_or_b32 s59, s15, s14
; SI-NEXT: s_add_i32 s14, s6, 0x3000000
-; SI-NEXT: v_readlane_b32 s6, v42, 31
+; SI-NEXT: v_readlane_b32 s6, v43, 36
; SI-NEXT: s_add_i32 s11, s6, 3
-; SI-NEXT: v_readlane_b32 s7, v42, 28
+; SI-NEXT: v_readlane_b32 s7, v43, 33
; SI-NEXT: s_and_b32 s6, s11, 0xff
; SI-NEXT: s_lshl_b32 s8, s7, 8
-; SI-NEXT: v_readlane_b32 s7, v42, 25
+; SI-NEXT: v_readlane_b32 s7, v43, 30
; SI-NEXT: s_or_b32 s6, s8, s6
; SI-NEXT: s_lshl_b32 s8, s7, 24
-; SI-NEXT: v_readlane_b32 s7, v42, 24
+; SI-NEXT: v_readlane_b32 s7, v43, 29
; SI-NEXT: s_add_i32 s24, s7, 3
; SI-NEXT: s_and_b32 s11, s24, 0xff
; SI-NEXT: s_addk_i32 s6, 0x300
@@ -158033,47 +158070,47 @@ define inreg <64 x bfloat> @bitcast_v128i8_to_v64bf16_scalar(<128 x i8> inreg %a
; SI-NEXT: s_and_b32 s6, s6, 0xffff
; SI-NEXT: s_or_b32 s8, s8, s11
; SI-NEXT: s_or_b32 s8, s8, s6
-; SI-NEXT: v_readlane_b32 s6, v42, 32
+; SI-NEXT: v_readlane_b32 s6, v43, 37
; SI-NEXT: s_add_i32 s12, s6, 3
-; SI-NEXT: v_readlane_b32 s7, v42, 26
+; SI-NEXT: v_readlane_b32 s7, v43, 31
; SI-NEXT: s_and_b32 s6, s12, 0xff
; SI-NEXT: s_lshl_b32 s11, s7, 8
-; SI-NEXT: v_readlane_b32 s7, v42, 16
+; SI-NEXT: v_readlane_b32 s7, v43, 21
; SI-NEXT: s_or_b32 s6, s11, s6
; SI-NEXT: s_lshl_b32 s11, s7, 24
-; SI-NEXT: v_readlane_b32 s7, v42, 18
+; SI-NEXT: v_readlane_b32 s7, v43, 23
; SI-NEXT: s_add_i32 s12, s7, 3
; SI-NEXT: s_and_b32 s12, s12, 0xff
; SI-NEXT: s_addk_i32 s6, 0x300
; SI-NEXT: s_lshl_b32 s12, s12, 16
-; SI-NEXT: v_readlane_b32 s7, v42, 33
+; SI-NEXT: v_readlane_b32 s7, v43, 38
; SI-NEXT: s_and_b32 s6, s6, 0xffff
; SI-NEXT: s_or_b32 s11, s11, s12
; SI-NEXT: s_add_i32 s13, s7, 3
-; SI-NEXT: v_readlane_b32 s7, v42, 30
+; SI-NEXT: v_readlane_b32 s7, v43, 35
; SI-NEXT: s_or_b32 s6, s11, s6
; SI-NEXT: s_and_b32 s11, s13, 0xff
; SI-NEXT: s_lshl_b32 s10, s7, 8
-; SI-NEXT: v_readlane_b32 s7, v42, 22
+; SI-NEXT: v_readlane_b32 s7, v43, 27
; SI-NEXT: s_or_b32 s10, s10, s11
; SI-NEXT: s_lshl_b32 s11, s7, 24
-; SI-NEXT: v_readlane_b32 s7, v42, 23
+; SI-NEXT: v_readlane_b32 s7, v43, 28
; SI-NEXT: s_add_i32 s25, s7, 3
; SI-NEXT: s_and_b32 s12, s25, 0xff
; SI-NEXT: s_addk_i32 s10, 0x300
; SI-NEXT: s_lshl_b32 s12, s12, 16
; SI-NEXT: s_and_b32 s10, s10, 0xffff
; SI-NEXT: s_or_b32 s11, s11, s12
-; SI-NEXT: v_readlane_b32 s7, v42, 29
+; SI-NEXT: v_readlane_b32 s7, v43, 34
; SI-NEXT: s_or_b32 s10, s11, s10
; SI-NEXT: s_add_i32 s9, s7, 3
-; SI-NEXT: v_readlane_b32 s7, v42, 27
-; SI-NEXT: v_readlane_b32 s11, v42, 20
+; SI-NEXT: v_readlane_b32 s7, v43, 32
+; SI-NEXT: v_readlane_b32 s11, v43, 25
; SI-NEXT: s_and_b32 s9, s9, 0xff
; SI-NEXT: s_lshl_b32 s7, s7, 8
; SI-NEXT: s_add_i32 s11, s11, 3
; SI-NEXT: s_or_b32 s7, s7, s9
-; SI-NEXT: v_readlane_b32 s9, v42, 21
+; SI-NEXT: v_readlane_b32 s9, v43, 26
; SI-NEXT: s_and_b32 s11, s11, 0xff
; SI-NEXT: s_addk_i32 s7, 0x300
; SI-NEXT: s_lshl_b32 s9, s9, 24
@@ -158081,15 +158118,15 @@ define inreg <64 x bfloat> @bitcast_v128i8_to_v64bf16_scalar(<128 x i8> inreg %a
; SI-NEXT: s_and_b32 s7, s7, 0xffff
; SI-NEXT: s_or_b32 s9, s9, s11
; SI-NEXT: s_or_b32 s7, s9, s7
-; SI-NEXT: v_readlane_b32 s9, v42, 19
+; SI-NEXT: v_readlane_b32 s9, v43, 24
; SI-NEXT: s_add_i32 s21, s9, 3
-; SI-NEXT: v_readlane_b32 s11, v42, 17
-; SI-NEXT: v_readlane_b32 s12, v42, 14
+; SI-NEXT: v_readlane_b32 s11, v43, 22
+; SI-NEXT: v_readlane_b32 s12, v43, 19
; SI-NEXT: s_and_b32 s9, s21, 0xff
; SI-NEXT: s_lshl_b32 s11, s11, 8
; SI-NEXT: s_add_i32 s12, s12, 3
; SI-NEXT: s_or_b32 s9, s11, s9
-; SI-NEXT: v_readlane_b32 s11, v42, 15
+; SI-NEXT: v_readlane_b32 s11, v43, 20
; SI-NEXT: s_and_b32 s12, s12, 0xff
; SI-NEXT: s_addk_i32 s9, 0x300
; SI-NEXT: s_lshl_b32 s11, s11, 24
@@ -158097,15 +158134,15 @@ define inreg <64 x bfloat> @bitcast_v128i8_to_v64bf16_scalar(<128 x i8> inreg %a
; SI-NEXT: s_and_b32 s9, s9, 0xffff
; SI-NEXT: s_or_b32 s11, s11, s12
; SI-NEXT: s_or_b32 s9, s11, s9
-; SI-NEXT: v_readlane_b32 s11, v42, 13
+; SI-NEXT: v_readlane_b32 s11, v43, 18
; SI-NEXT: s_add_i32 s11, s11, 3
-; SI-NEXT: v_readlane_b32 s12, v42, 12
-; SI-NEXT: v_readlane_b32 s13, v42, 10
+; SI-NEXT: v_readlane_b32 s12, v43, 17
+; SI-NEXT: v_readlane_b32 s13, v43, 15
; SI-NEXT: s_and_b32 s11, s11, 0xff
; SI-NEXT: s_lshl_b32 s12, s12, 8
; SI-NEXT: s_add_i32 s13, s13, 3
; SI-NEXT: s_or_b32 s11, s12, s11
-; SI-NEXT: v_readlane_b32 s12, v42, 11
+; SI-NEXT: v_readlane_b32 s12, v43, 16
; SI-NEXT: s_and_b32 s13, s13, 0xff
; SI-NEXT: s_addk_i32 s11, 0x300
; SI-NEXT: s_lshl_b32 s12, s12, 24
@@ -158113,16 +158150,16 @@ define inreg <64 x bfloat> @bitcast_v128i8_to_v64bf16_scalar(<128 x i8> inreg %a
; SI-NEXT: s_and_b32 s11, s11, 0xffff
; SI-NEXT: s_or_b32 s12, s12, s13
; SI-NEXT: s_or_b32 s11, s12, s11
-; SI-NEXT: v_readlane_b32 s12, v42, 9
+; SI-NEXT: v_readlane_b32 s12, v43, 14
; SI-NEXT: s_add_i32 s15, s16, 0x3000000
; SI-NEXT: s_add_i32 s12, s12, 3
-; SI-NEXT: v_readlane_b32 s13, v42, 8
-; SI-NEXT: v_readlane_b32 s16, v42, 6
+; SI-NEXT: v_readlane_b32 s13, v43, 13
+; SI-NEXT: v_readlane_b32 s16, v43, 11
; SI-NEXT: s_and_b32 s12, s12, 0xff
; SI-NEXT: s_lshl_b32 s13, s13, 8
; SI-NEXT: s_add_i32 s16, s16, 3
; SI-NEXT: s_or_b32 s12, s13, s12
-; SI-NEXT: v_readlane_b32 s13, v42, 7
+; SI-NEXT: v_readlane_b32 s13, v43, 12
; SI-NEXT: s_and_b32 s16, s16, 0xff
; SI-NEXT: s_addk_i32 s12, 0x300
; SI-NEXT: s_lshl_b32 s13, s13, 24
@@ -158130,16 +158167,16 @@ define inreg <64 x bfloat> @bitcast_v128i8_to_v64bf16_scalar(<128 x i8> inreg %a
; SI-NEXT: s_and_b32 s12, s12, 0xffff
; SI-NEXT: s_or_b32 s13, s13, s16
; SI-NEXT: s_or_b32 s12, s13, s12
-; SI-NEXT: v_readlane_b32 s13, v42, 5
+; SI-NEXT: v_readlane_b32 s13, v43, 10
; SI-NEXT: s_add_i32 s40, s17, 0x3000000
; SI-NEXT: s_add_i32 s13, s13, 3
-; SI-NEXT: v_readlane_b32 s16, v42, 4
-; SI-NEXT: v_readlane_b32 s17, v42, 2
+; SI-NEXT: v_readlane_b32 s16, v43, 9
+; SI-NEXT: v_readlane_b32 s17, v43, 7
; SI-NEXT: s_and_b32 s13, s13, 0xff
; SI-NEXT: s_lshl_b32 s16, s16, 8
; SI-NEXT: s_add_i32 s17, s17, 3
; SI-NEXT: s_or_b32 s13, s16, s13
-; SI-NEXT: v_readlane_b32 s16, v42, 3
+; SI-NEXT: v_readlane_b32 s16, v43, 8
; SI-NEXT: s_and_b32 s17, s17, 0xff
; SI-NEXT: s_addk_i32 s13, 0x300
; SI-NEXT: s_lshl_b32 s16, s16, 24
@@ -158147,16 +158184,16 @@ define inreg <64 x bfloat> @bitcast_v128i8_to_v64bf16_scalar(<128 x i8> inreg %a
; SI-NEXT: s_and_b32 s13, s13, 0xffff
; SI-NEXT: s_or_b32 s16, s16, s17
; SI-NEXT: s_or_b32 s13, s16, s13
-; SI-NEXT: v_readlane_b32 s16, v42, 1
+; SI-NEXT: v_readlane_b32 s16, v43, 6
; SI-NEXT: s_add_i32 s41, s18, 0x3000000
; SI-NEXT: s_add_i32 s16, s16, 3
-; SI-NEXT: v_readlane_b32 s17, v42, 0
-; SI-NEXT: v_readlane_b32 s18, v43, 62
+; SI-NEXT: v_readlane_b32 s17, v43, 5
+; SI-NEXT: v_readlane_b32 s18, v43, 3
; SI-NEXT: s_and_b32 s16, s16, 0xff
; SI-NEXT: s_lshl_b32 s17, s17, 8
; SI-NEXT: s_add_i32 s18, s18, 3
; SI-NEXT: s_or_b32 s16, s17, s16
-; SI-NEXT: v_readlane_b32 s17, v43, 63
+; SI-NEXT: v_readlane_b32 s17, v43, 4
; SI-NEXT: s_and_b32 s18, s18, 0xff
; SI-NEXT: s_addk_i32 s16, 0x300
; SI-NEXT: s_lshl_b32 s17, s17, 24
@@ -158165,16 +158202,16 @@ define inreg <64 x bfloat> @bitcast_v128i8_to_v64bf16_scalar(<128 x i8> inreg %a
; SI-NEXT: s_or_b32 s17, s17, s18
; SI-NEXT: s_or_b32 s16, s17, s16
; SI-NEXT: s_add_i32 s17, s16, 0x3000000
-; SI-NEXT: v_readlane_b32 s16, v43, 61
+; SI-NEXT: v_readlane_b32 s16, v43, 2
; SI-NEXT: s_add_i32 s42, s19, 0x3000000
; SI-NEXT: s_add_i32 s16, s16, 3
-; SI-NEXT: v_readlane_b32 s18, v43, 60
-; SI-NEXT: v_readlane_b32 s19, v43, 58
+; SI-NEXT: v_readlane_b32 s18, v43, 1
+; SI-NEXT: v_readlane_b32 s19, v44, 63
; SI-NEXT: s_and_b32 s16, s16, 0xff
; SI-NEXT: s_lshl_b32 s18, s18, 8
; SI-NEXT: s_add_i32 s19, s19, 3
; SI-NEXT: s_or_b32 s16, s18, s16
-; SI-NEXT: v_readlane_b32 s18, v43, 59
+; SI-NEXT: v_readlane_b32 s18, v43, 0
; SI-NEXT: s_and_b32 s19, s19, 0xff
; SI-NEXT: s_addk_i32 s16, 0x300
; SI-NEXT: s_lshl_b32 s18, s18, 24
@@ -158182,16 +158219,16 @@ define inreg <64 x bfloat> @bitcast_v128i8_to_v64bf16_scalar(<128 x i8> inreg %a
; SI-NEXT: s_and_b32 s16, s16, 0xffff
; SI-NEXT: s_or_b32 s18, s18, s19
; SI-NEXT: s_or_b32 s16, s18, s16
-; SI-NEXT: v_readlane_b32 s18, v43, 57
+; SI-NEXT: v_readlane_b32 s18, v44, 62
; SI-NEXT: s_add_i32 s43, s20, 0x3000000
; SI-NEXT: s_add_i32 s18, s18, 3
-; SI-NEXT: v_readlane_b32 s19, v43, 56
-; SI-NEXT: v_readlane_b32 s20, v43, 54
+; SI-NEXT: v_readlane_b32 s19, v44, 61
+; SI-NEXT: v_readlane_b32 s20, v44, 59
; SI-NEXT: s_and_b32 s18, s18, 0xff
; SI-NEXT: s_lshl_b32 s19, s19, 8
; SI-NEXT: s_add_i32 s20, s20, 3
; SI-NEXT: s_or_b32 s18, s19, s18
-; SI-NEXT: v_readlane_b32 s19, v43, 55
+; SI-NEXT: v_readlane_b32 s19, v44, 60
; SI-NEXT: s_and_b32 s20, s20, 0xff
; SI-NEXT: s_addk_i32 s18, 0x300
; SI-NEXT: s_lshl_b32 s19, s19, 24
@@ -158199,15 +158236,15 @@ define inreg <64 x bfloat> @bitcast_v128i8_to_v64bf16_scalar(<128 x i8> inreg %a
; SI-NEXT: s_and_b32 s18, s18, 0xffff
; SI-NEXT: s_or_b32 s19, s19, s20
; SI-NEXT: s_or_b32 s18, s19, s18
-; SI-NEXT: v_readlane_b32 s19, v43, 53
+; SI-NEXT: v_readlane_b32 s19, v44, 58
; SI-NEXT: s_add_i32 s19, s19, 3
-; SI-NEXT: v_readlane_b32 s20, v43, 52
-; SI-NEXT: v_readlane_b32 s21, v43, 50
+; SI-NEXT: v_readlane_b32 s20, v44, 57
+; SI-NEXT: v_readlane_b32 s21, v44, 55
; SI-NEXT: s_and_b32 s19, s19, 0xff
; SI-NEXT: s_lshl_b32 s20, s20, 8
; SI-NEXT: s_add_i32 s21, s21, 3
; SI-NEXT: s_or_b32 s19, s20, s19
-; SI-NEXT: v_readlane_b32 s20, v43, 51
+; SI-NEXT: v_readlane_b32 s20, v44, 56
; SI-NEXT: s_and_b32 s21, s21, 0xff
; SI-NEXT: s_addk_i32 s19, 0x300
; SI-NEXT: s_lshl_b32 s20, s20, 24
@@ -158215,16 +158252,16 @@ define inreg <64 x bfloat> @bitcast_v128i8_to_v64bf16_scalar(<128 x i8> inreg %a
; SI-NEXT: s_and_b32 s19, s19, 0xffff
; SI-NEXT: s_or_b32 s20, s20, s21
; SI-NEXT: s_or_b32 s19, s20, s19
-; SI-NEXT: v_readlane_b32 s20, v43, 49
+; SI-NEXT: v_readlane_b32 s20, v44, 54
; SI-NEXT: s_add_i32 s44, s22, 0x3000000
; SI-NEXT: s_add_i32 s20, s20, 3
-; SI-NEXT: v_readlane_b32 s21, v43, 48
-; SI-NEXT: v_readlane_b32 s22, v43, 46
+; SI-NEXT: v_readlane_b32 s21, v44, 53
+; SI-NEXT: v_readlane_b32 s22, v44, 51
; SI-NEXT: s_and_b32 s20, s20, 0xff
; SI-NEXT: s_lshl_b32 s21, s21, 8
; SI-NEXT: s_add_i32 s22, s22, 3
; SI-NEXT: s_or_b32 s20, s21, s20
-; SI-NEXT: v_readlane_b32 s21, v43, 47
+; SI-NEXT: v_readlane_b32 s21, v44, 52
; SI-NEXT: s_and_b32 s22, s22, 0xff
; SI-NEXT: s_addk_i32 s20, 0x300
; SI-NEXT: s_lshl_b32 s21, s21, 24
@@ -158233,16 +158270,16 @@ define inreg <64 x bfloat> @bitcast_v128i8_to_v64bf16_scalar(<128 x i8> inreg %a
; SI-NEXT: s_or_b32 s21, s21, s22
; SI-NEXT: s_or_b32 s20, s21, s20
; SI-NEXT: s_add_i32 s21, s20, 0x3000000
-; SI-NEXT: v_readlane_b32 s20, v43, 43
+; SI-NEXT: v_readlane_b32 s20, v44, 1
; SI-NEXT: s_add_i32 s45, s23, 0x3000000
; SI-NEXT: s_add_i32 s20, s20, 3
-; SI-NEXT: v_readlane_b32 s22, v43, 42
-; SI-NEXT: v_readlane_b32 s23, v43, 44
+; SI-NEXT: v_readlane_b32 s22, v44, 0
+; SI-NEXT: v_readlane_b32 s23, v44, 49
; SI-NEXT: s_and_b32 s20, s20, 0xff
; SI-NEXT: s_lshl_b32 s22, s22, 8
; SI-NEXT: s_add_i32 s23, s23, 3
; SI-NEXT: s_or_b32 s20, s22, s20
-; SI-NEXT: v_readlane_b32 s22, v43, 45
+; SI-NEXT: v_readlane_b32 s22, v44, 50
; SI-NEXT: s_and_b32 s23, s23, 0xff
; SI-NEXT: s_addk_i32 s20, 0x300
; SI-NEXT: s_lshl_b32 s22, s22, 24
@@ -158251,15 +158288,15 @@ define inreg <64 x bfloat> @bitcast_v128i8_to_v64bf16_scalar(<128 x i8> inreg %a
; SI-NEXT: s_or_b32 s22, s22, s23
; SI-NEXT: s_or_b32 s20, s22, s20
; SI-NEXT: s_add_i32 s22, s20, 0x3000000
-; SI-NEXT: v_readlane_b32 s20, v43, 41
+; SI-NEXT: v_readlane_b32 s20, v44, 41
; SI-NEXT: s_add_i32 s20, s20, 3
-; SI-NEXT: v_readlane_b32 s23, v43, 40
-; SI-NEXT: v_readlane_b32 s24, v43, 38
+; SI-NEXT: v_readlane_b32 s23, v44, 40
+; SI-NEXT: v_readlane_b32 s24, v44, 3
; SI-NEXT: s_and_b32 s20, s20, 0xff
; SI-NEXT: s_lshl_b32 s23, s23, 8
; SI-NEXT: s_add_i32 s24, s24, 3
; SI-NEXT: s_or_b32 s20, s23, s20
-; SI-NEXT: v_readlane_b32 s23, v43, 39
+; SI-NEXT: v_readlane_b32 s23, v44, 2
; SI-NEXT: s_and_b32 s24, s24, 0xff
; SI-NEXT: s_addk_i32 s20, 0x300
; SI-NEXT: s_lshl_b32 s23, s23, 24
@@ -158268,134 +158305,136 @@ define inreg <64 x bfloat> @bitcast_v128i8_to_v64bf16_scalar(<128 x i8> inreg %a
; SI-NEXT: s_or_b32 s23, s23, s24
; SI-NEXT: s_or_b32 s20, s23, s20
; SI-NEXT: s_add_i32 s23, s20, 0x3000000
-; SI-NEXT: v_readlane_b32 s20, v43, 37
+; SI-NEXT: v_readlane_b32 s20, v44, 39
; SI-NEXT: s_add_i32 s20, s20, 3
-; SI-NEXT: v_readlane_b32 s24, v43, 36
-; SI-NEXT: v_readlane_b32 s25, v43, 34
+; SI-NEXT: v_readlane_b32 s24, v44, 38
+; SI-NEXT: v_readlane_b32 s25, v44, 36
; SI-NEXT: s_and_b32 s20, s20, 0xff
; SI-NEXT: s_lshl_b32 s24, s24, 8
; SI-NEXT: s_add_i32 s25, s25, 3
; SI-NEXT: s_or_b32 s20, s24, s20
-; SI-NEXT: v_readlane_b32 s24, v43, 35
+; SI-NEXT: v_readlane_b32 s24, v44, 37
; SI-NEXT: s_and_b32 s25, s25, 0xff
; SI-NEXT: s_addk_i32 s20, 0x300
; SI-NEXT: s_lshl_b32 s24, s24, 24
; SI-NEXT: s_lshl_b32 s25, s25, 16
; SI-NEXT: s_and_b32 s20, s20, 0xffff
; SI-NEXT: s_or_b32 s24, s24, s25
-; SI-NEXT: s_and_b32 s46, s46, 0xff
+; SI-NEXT: s_add_i32 s16, s16, 0x3000000
+; SI-NEXT: s_add_i32 s18, s18, 0x3000000
; SI-NEXT: s_or_b32 s20, s24, s20
-; SI-NEXT: v_readlane_b32 s24, v43, 3
-; SI-NEXT: s_lshl_b32 s46, s46, 16
-; SI-NEXT: s_addk_i32 s56, 0x300
+; SI-NEXT: v_readlane_b32 s24, v44, 7
; SI-NEXT: s_add_i32 s24, s24, 3
-; SI-NEXT: v_readlane_b32 s25, v43, 2
-; SI-NEXT: v_readlane_b32 s26, v43, 1
-; SI-NEXT: s_or_b32 s46, s47, s46
-; SI-NEXT: s_and_b32 s47, s56, 0xffff
-; SI-NEXT: s_add_i32 s7, s7, 0x3000000
-; SI-NEXT: s_add_i32 s9, s9, 0x3000000
+; SI-NEXT: v_readlane_b32 s25, v44, 6
+; SI-NEXT: v_readlane_b32 s26, v44, 5
+; SI-NEXT: s_and_b32 s79, s18, 0xffff0000
+; SI-NEXT: s_lshl_b32 s77, s18, 16
+; SI-NEXT: s_and_b32 s18, s16, 0xffff0000
; SI-NEXT: s_and_b32 s24, s24, 0xff
; SI-NEXT: s_lshl_b32 s25, s25, 8
; SI-NEXT: s_add_i32 s26, s26, 3
-; SI-NEXT: s_or_b32 s56, s46, s47
-; SI-NEXT: s_add_i32 s47, s58, 0x3000000
-; SI-NEXT: s_add_i32 s58, s59, 0x3000000
-; SI-NEXT: s_add_i32 s10, s10, 0x3000000
+; SI-NEXT: v_writelane_b32 v43, s18, 56
+; SI-NEXT: s_and_b32 s18, s17, 0xffff0000
+; SI-NEXT: s_and_b32 s46, s46, 0xff
+; SI-NEXT: s_add_i32 s13, s13, 0x3000000
; SI-NEXT: s_or_b32 s24, s25, s24
-; SI-NEXT: v_readlane_b32 s25, v43, 0
+; SI-NEXT: v_readlane_b32 s25, v44, 4
; SI-NEXT: s_and_b32 s26, s26, 0xff
-; SI-NEXT: s_and_b32 s73, s9, 0xffff0000
-; SI-NEXT: s_lshl_b32 s59, s9, 16
-; SI-NEXT: s_and_b32 s9, s7, 0xffff0000
-; SI-NEXT: s_add_i32 s6, s6, 0x3000000
+; SI-NEXT: v_writelane_b32 v43, s18, 57
+; SI-NEXT: s_lshl_b32 s17, s17, 16
+; SI-NEXT: s_lshl_b32 s46, s46, 16
+; SI-NEXT: s_addk_i32 s56, 0x300
+; SI-NEXT: s_add_i32 s12, s12, 0x3000000
+; SI-NEXT: s_add_i32 s19, s19, 0x3000000
; SI-NEXT: s_addk_i32 s24, 0x300
; SI-NEXT: s_lshl_b32 s25, s25, 24
; SI-NEXT: s_lshl_b32 s26, s26, 16
-; SI-NEXT: s_and_b32 s63, s17, 0xffff0000
-; SI-NEXT: s_lshl_b32 s79, s17, 16
-; SI-NEXT: v_writelane_b32 v42, s9, 50
-; SI-NEXT: s_lshl_b32 s17, s7, 16
-; SI-NEXT: s_lshl_b32 s7, s10, 16
-; SI-NEXT: s_add_i32 s8, s8, 0x3000000
+; SI-NEXT: v_writelane_b32 v43, s17, 58
+; SI-NEXT: s_and_b32 s17, s13, 0xffff0000
+; SI-NEXT: s_or_b32 s46, s47, s46
+; SI-NEXT: s_and_b32 s47, s56, 0xffff
; SI-NEXT: s_and_b32 s24, s24, 0xffff
; SI-NEXT: s_or_b32 s25, s25, s26
-; SI-NEXT: v_writelane_b32 v42, s7, 51
-; SI-NEXT: s_and_b32 s7, s6, 0xffff0000
+; SI-NEXT: s_and_b32 s72, s23, 0xffff0000
+; SI-NEXT: s_lshl_b32 s29, s23, 16
+; SI-NEXT: s_and_b32 s73, s22, 0xffff0000
+; SI-NEXT: s_lshl_b32 s23, s22, 16
+; SI-NEXT: s_and_b32 s78, s19, 0xffff0000
+; SI-NEXT: s_lshl_b32 s22, s19, 16
+; SI-NEXT: v_writelane_b32 v43, s17, 59
+; SI-NEXT: s_lshl_b32 s19, s13, 16
+; SI-NEXT: s_and_b32 s13, s12, 0xffff0000
+; SI-NEXT: s_or_b32 s56, s46, s47
+; SI-NEXT: s_add_i32 s11, s11, 0x3000000
; SI-NEXT: s_or_b32 s24, s25, s24
-; SI-NEXT: v_writelane_b32 v42, s7, 52
-; SI-NEXT: s_and_b32 s7, s8, 0xffff0000
+; SI-NEXT: v_writelane_b32 v43, s13, 60
+; SI-NEXT: s_lshl_b32 s12, s12, 16
; SI-NEXT: s_add_i32 s4, s4, 0x3000000
; SI-NEXT: s_add_i32 s5, s5, 0x3000000
; SI-NEXT: s_add_i32 s46, s60, 0x3000000
+; SI-NEXT: s_add_i32 s47, s58, 0x3000000
; SI-NEXT: s_add_i32 s56, s56, 0x3000000
; SI-NEXT: s_add_i32 s57, s57, 0x3000000
-; SI-NEXT: s_add_i32 s11, s11, 0x3000000
-; SI-NEXT: s_add_i32 s12, s12, 0x3000000
-; SI-NEXT: s_add_i32 s13, s13, 0x3000000
-; SI-NEXT: s_add_i32 s16, s16, 0x3000000
-; SI-NEXT: s_add_i32 s18, s18, 0x3000000
-; SI-NEXT: s_add_i32 s19, s19, 0x3000000
+; SI-NEXT: s_add_i32 s58, s59, 0x3000000
+; SI-NEXT: s_add_i32 s8, s8, 0x3000000
+; SI-NEXT: s_add_i32 s6, s6, 0x3000000
+; SI-NEXT: s_add_i32 s10, s10, 0x3000000
+; SI-NEXT: s_add_i32 s7, s7, 0x3000000
+; SI-NEXT: s_add_i32 s9, s9, 0x3000000
; SI-NEXT: s_add_i32 s20, s20, 0x3000000
; SI-NEXT: s_add_i32 s24, s24, 0x3000000
-; SI-NEXT: v_writelane_b32 v42, s7, 53
-; SI-NEXT: s_lshl_b32 s7, s8, 16
+; SI-NEXT: v_writelane_b32 v43, s12, 61
+; SI-NEXT: s_and_b32 s12, s11, 0xffff0000
; SI-NEXT: s_and_b32 s27, s24, 0xffff0000
; SI-NEXT: s_lshl_b32 s26, s24, 16
; SI-NEXT: s_and_b32 s24, s20, 0xffff0000
; SI-NEXT: s_lshl_b32 s20, s20, 16
-; SI-NEXT: s_and_b32 s35, s23, 0xffff0000
-; SI-NEXT: s_lshl_b32 s29, s23, 16
-; SI-NEXT: s_and_b32 s90, s22, 0xffff0000
-; SI-NEXT: s_lshl_b32 s74, s22, 16
-; SI-NEXT: s_and_b32 s25, s21, 0xffff0000
+; SI-NEXT: s_and_b32 s75, s21, 0xffff0000
; SI-NEXT: s_lshl_b32 s21, s21, 16
-; SI-NEXT: s_and_b32 s75, s19, 0xffff0000
-; SI-NEXT: s_lshl_b32 s22, s19, 16
-; SI-NEXT: s_and_b32 s61, s18, 0xffff0000
-; SI-NEXT: s_lshl_b32 s76, s18, 16
-; SI-NEXT: s_and_b32 s77, s16, 0xffff0000
; SI-NEXT: s_lshl_b32 s16, s16, 16
-; SI-NEXT: s_and_b32 s89, s13, 0xffff0000
-; SI-NEXT: s_lshl_b32 s19, s13, 16
-; SI-NEXT: s_and_b32 s13, s12, 0xffff0000
-; SI-NEXT: s_lshl_b32 s88, s12, 16
-; SI-NEXT: s_and_b32 s60, s11, 0xffff0000
+; SI-NEXT: v_writelane_b32 v43, s12, 62
; SI-NEXT: s_lshl_b32 s18, s11, 16
-; SI-NEXT: s_and_b32 s23, s10, 0xffff0000
+; SI-NEXT: s_and_b32 s95, s9, 0xffff0000
+; SI-NEXT: s_lshl_b32 s9, s9, 16
+; SI-NEXT: s_and_b32 s30, s7, 0xffff0000
+; SI-NEXT: s_lshl_b32 s17, s7, 16
+; SI-NEXT: s_and_b32 s34, s10, 0xffff0000
+; SI-NEXT: s_lshl_b32 s31, s10, 16
+; SI-NEXT: s_and_b32 s35, s6, 0xffff0000
; SI-NEXT: s_lshl_b32 s6, s6, 16
-; SI-NEXT: v_writelane_b32 v42, s7, 54
-; SI-NEXT: s_and_b32 s72, s58, 0xffff0000
+; SI-NEXT: s_and_b32 s37, s8, 0xffff0000
+; SI-NEXT: s_lshl_b32 s36, s8, 16
+; SI-NEXT: s_and_b32 s38, s58, 0xffff0000
; SI-NEXT: s_lshl_b32 s99, s58, 16
-; SI-NEXT: s_and_b32 s7, s57, 0xffff0000
-; SI-NEXT: s_lshl_b32 s57, s57, 16
+; SI-NEXT: s_and_b32 s48, s57, 0xffff0000
+; SI-NEXT: s_lshl_b32 s39, s57, 16
; SI-NEXT: s_and_b32 s49, s56, 0xffff0000
-; SI-NEXT: s_lshl_b32 s8, s56, 16
+; SI-NEXT: s_lshl_b32 s98, s56, 16
; SI-NEXT: s_and_b32 s51, s47, 0xffff0000
; SI-NEXT: s_lshl_b32 s50, s47, 16
; SI-NEXT: s_and_b32 s52, s46, 0xffff0000
-; SI-NEXT: s_lshl_b32 s97, s46, 16
+; SI-NEXT: s_lshl_b32 s12, s46, 16
; SI-NEXT: s_and_b32 s54, s45, 0xffff0000
; SI-NEXT: s_lshl_b32 s53, s45, 16
; SI-NEXT: s_and_b32 s55, s44, 0xffff0000
-; SI-NEXT: s_lshl_b32 s28, s44, 16
+; SI-NEXT: s_lshl_b32 s96, s44, 16
; SI-NEXT: s_and_b32 s65, s43, 0xffff0000
; SI-NEXT: s_lshl_b32 s64, s43, 16
; SI-NEXT: s_and_b32 s66, s42, 0xffff0000
; SI-NEXT: s_lshl_b32 s87, s42, 16
-; SI-NEXT: s_and_b32 s68, s41, 0xffff0000
+; SI-NEXT: s_and_b32 s45, s41, 0xffff0000
; SI-NEXT: s_lshl_b32 s67, s41, 16
-; SI-NEXT: s_and_b32 s69, s40, 0xffff0000
+; SI-NEXT: s_and_b32 s57, s40, 0xffff0000
; SI-NEXT: s_lshl_b32 s86, s40, 16
-; SI-NEXT: s_and_b32 s62, s15, 0xffff0000
+; SI-NEXT: s_and_b32 s13, s15, 0xffff0000
; SI-NEXT: s_lshl_b32 s70, s15, 16
; SI-NEXT: s_and_b32 s80, s14, 0xffff0000
; SI-NEXT: s_lshl_b32 s85, s14, 16
-; SI-NEXT: s_and_b32 s92, s5, 0xffff0000
-; SI-NEXT: s_lshl_b32 s11, s5, 16
-; SI-NEXT: s_and_b32 s83, s4, 0xffff0000
+; SI-NEXT: s_and_b32 s82, s5, 0xffff0000
+; SI-NEXT: s_lshl_b32 s25, s5, 16
+; SI-NEXT: s_and_b32 s28, s4, 0xffff0000
; SI-NEXT: s_lshl_b32 s84, s4, 16
-; SI-NEXT: v_writelane_b32 v42, s7, 55
+; SI-NEXT: v_writelane_b32 v43, s9, 63
; SI-NEXT: .LBB89_3: ; %end
; SI-NEXT: v_mul_f32_e64 v1, 1.0, s27
; SI-NEXT: v_lshrrev_b32_e32 v1, 16, v1
@@ -158410,134 +158449,136 @@ define inreg <64 x bfloat> @bitcast_v128i8_to_v64bf16_scalar(<128 x i8> inreg %a
; SI-NEXT: v_add_i32_e32 v2, vcc, 4, v0
; SI-NEXT: buffer_store_dword v1, v2, s[0:3], 0 offen
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_mul_f32_e64 v1, 1.0, s35
+; SI-NEXT: v_mul_f32_e64 v1, 1.0, s72
; SI-NEXT: v_lshrrev_b32_e32 v1, 16, v1
; SI-NEXT: v_mul_f32_e64 v2, 1.0, s29
; SI-NEXT: v_alignbit_b32 v1, v1, v2, 16
; SI-NEXT: v_add_i32_e32 v2, vcc, 8, v0
; SI-NEXT: buffer_store_dword v1, v2, s[0:3], 0 offen
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_mul_f32_e64 v1, 1.0, s90
+; SI-NEXT: v_mul_f32_e64 v1, 1.0, s73
; SI-NEXT: v_lshrrev_b32_e32 v1, 16, v1
-; SI-NEXT: v_mul_f32_e64 v2, 1.0, s74
+; SI-NEXT: v_mul_f32_e64 v2, 1.0, s23
; SI-NEXT: v_alignbit_b32 v1, v1, v2, 16
; SI-NEXT: v_add_i32_e32 v2, vcc, 12, v0
; SI-NEXT: buffer_store_dword v1, v2, s[0:3], 0 offen
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_mul_f32_e64 v1, 1.0, s25
+; SI-NEXT: v_mul_f32_e64 v1, 1.0, s75
; SI-NEXT: v_lshrrev_b32_e32 v1, 16, v1
; SI-NEXT: v_mul_f32_e64 v2, 1.0, s21
; SI-NEXT: v_alignbit_b32 v1, v1, v2, 16
; SI-NEXT: v_add_i32_e32 v2, vcc, 16, v0
; SI-NEXT: buffer_store_dword v1, v2, s[0:3], 0 offen
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_mul_f32_e64 v1, 1.0, s75
+; SI-NEXT: v_mul_f32_e64 v1, 1.0, s78
; SI-NEXT: v_lshrrev_b32_e32 v1, 16, v1
; SI-NEXT: v_mul_f32_e64 v2, 1.0, s22
; SI-NEXT: v_alignbit_b32 v1, v1, v2, 16
; SI-NEXT: v_add_i32_e32 v2, vcc, 20, v0
; SI-NEXT: buffer_store_dword v1, v2, s[0:3], 0 offen
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_mul_f32_e64 v1, 1.0, s61
+; SI-NEXT: v_mul_f32_e64 v1, 1.0, s79
; SI-NEXT: v_lshrrev_b32_e32 v1, 16, v1
-; SI-NEXT: v_mul_f32_e64 v2, 1.0, s76
+; SI-NEXT: v_mul_f32_e64 v2, 1.0, s77
; SI-NEXT: v_alignbit_b32 v1, v1, v2, 16
; SI-NEXT: v_add_i32_e32 v2, vcc, 24, v0
+; SI-NEXT: v_readlane_b32 s4, v43, 56
; SI-NEXT: buffer_store_dword v1, v2, s[0:3], 0 offen
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_mul_f32_e64 v1, 1.0, s77
+; SI-NEXT: v_mul_f32_e64 v1, 1.0, s4
; SI-NEXT: v_lshrrev_b32_e32 v1, 16, v1
; SI-NEXT: v_mul_f32_e64 v2, 1.0, s16
; SI-NEXT: v_alignbit_b32 v1, v1, v2, 16
; SI-NEXT: v_add_i32_e32 v2, vcc, 28, v0
+; SI-NEXT: v_readlane_b32 s4, v43, 57
; SI-NEXT: buffer_store_dword v1, v2, s[0:3], 0 offen
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_mul_f32_e64 v1, 1.0, s63
+; SI-NEXT: v_mul_f32_e64 v1, 1.0, s4
+; SI-NEXT: v_readlane_b32 s4, v43, 58
; SI-NEXT: v_lshrrev_b32_e32 v1, 16, v1
-; SI-NEXT: v_mul_f32_e64 v2, 1.0, s79
+; SI-NEXT: v_mul_f32_e64 v2, 1.0, s4
; SI-NEXT: v_alignbit_b32 v1, v1, v2, 16
; SI-NEXT: v_add_i32_e32 v2, vcc, 32, v0
+; SI-NEXT: v_readlane_b32 s4, v43, 59
; SI-NEXT: buffer_store_dword v1, v2, s[0:3], 0 offen
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_mul_f32_e64 v1, 1.0, s89
+; SI-NEXT: v_mul_f32_e64 v1, 1.0, s4
; SI-NEXT: v_lshrrev_b32_e32 v1, 16, v1
; SI-NEXT: v_mul_f32_e64 v2, 1.0, s19
; SI-NEXT: v_alignbit_b32 v1, v1, v2, 16
; SI-NEXT: v_add_i32_e32 v2, vcc, 36, v0
+; SI-NEXT: v_readlane_b32 s4, v43, 60
; SI-NEXT: buffer_store_dword v1, v2, s[0:3], 0 offen
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_mul_f32_e64 v1, 1.0, s13
+; SI-NEXT: v_mul_f32_e64 v1, 1.0, s4
+; SI-NEXT: v_readlane_b32 s4, v43, 61
; SI-NEXT: v_lshrrev_b32_e32 v1, 16, v1
-; SI-NEXT: v_mul_f32_e64 v2, 1.0, s88
+; SI-NEXT: v_mul_f32_e64 v2, 1.0, s4
; SI-NEXT: v_alignbit_b32 v1, v1, v2, 16
; SI-NEXT: v_add_i32_e32 v2, vcc, 40, v0
+; SI-NEXT: v_readlane_b32 s4, v43, 62
; SI-NEXT: buffer_store_dword v1, v2, s[0:3], 0 offen
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_mul_f32_e64 v1, 1.0, s60
+; SI-NEXT: v_mul_f32_e64 v1, 1.0, s4
; SI-NEXT: v_lshrrev_b32_e32 v1, 16, v1
; SI-NEXT: v_mul_f32_e64 v2, 1.0, s18
; SI-NEXT: v_alignbit_b32 v1, v1, v2, 16
; SI-NEXT: v_add_i32_e32 v2, vcc, 44, v0
; SI-NEXT: buffer_store_dword v1, v2, s[0:3], 0 offen
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_mul_f32_e64 v1, 1.0, s73
+; SI-NEXT: v_mul_f32_e64 v1, 1.0, s95
+; SI-NEXT: v_readlane_b32 s4, v43, 63
; SI-NEXT: v_lshrrev_b32_e32 v1, 16, v1
-; SI-NEXT: v_mul_f32_e64 v2, 1.0, s59
+; SI-NEXT: v_mul_f32_e64 v2, 1.0, s4
; SI-NEXT: v_alignbit_b32 v1, v1, v2, 16
; SI-NEXT: v_add_i32_e32 v2, vcc, 48, v0
-; SI-NEXT: v_readlane_b32 s4, v42, 50
; SI-NEXT: buffer_store_dword v1, v2, s[0:3], 0 offen
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_mul_f32_e64 v1, 1.0, s4
+; SI-NEXT: v_mul_f32_e64 v1, 1.0, s30
; SI-NEXT: v_lshrrev_b32_e32 v1, 16, v1
; SI-NEXT: v_mul_f32_e64 v2, 1.0, s17
; SI-NEXT: v_alignbit_b32 v1, v1, v2, 16
; SI-NEXT: v_add_i32_e32 v2, vcc, 52, v0
; SI-NEXT: buffer_store_dword v1, v2, s[0:3], 0 offen
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_mul_f32_e64 v1, 1.0, s23
-; SI-NEXT: v_readlane_b32 s4, v42, 51
+; SI-NEXT: v_mul_f32_e64 v1, 1.0, s34
; SI-NEXT: v_lshrrev_b32_e32 v1, 16, v1
-; SI-NEXT: v_mul_f32_e64 v2, 1.0, s4
+; SI-NEXT: v_mul_f32_e64 v2, 1.0, s31
; SI-NEXT: v_alignbit_b32 v1, v1, v2, 16
; SI-NEXT: v_add_i32_e32 v2, vcc, 56, v0
-; SI-NEXT: v_readlane_b32 s4, v42, 52
; SI-NEXT: buffer_store_dword v1, v2, s[0:3], 0 offen
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_mul_f32_e64 v1, 1.0, s4
+; SI-NEXT: v_mul_f32_e64 v1, 1.0, s35
; SI-NEXT: v_lshrrev_b32_e32 v1, 16, v1
; SI-NEXT: v_mul_f32_e64 v2, 1.0, s6
; SI-NEXT: v_alignbit_b32 v1, v1, v2, 16
; SI-NEXT: v_add_i32_e32 v2, vcc, 60, v0
-; SI-NEXT: v_readlane_b32 s4, v42, 53
; SI-NEXT: buffer_store_dword v1, v2, s[0:3], 0 offen
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_mul_f32_e64 v1, 1.0, s4
-; SI-NEXT: v_readlane_b32 s4, v42, 54
+; SI-NEXT: v_mul_f32_e64 v1, 1.0, s37
; SI-NEXT: v_lshrrev_b32_e32 v1, 16, v1
-; SI-NEXT: v_mul_f32_e64 v2, 1.0, s4
+; SI-NEXT: v_mul_f32_e64 v2, 1.0, s36
; SI-NEXT: v_alignbit_b32 v1, v1, v2, 16
; SI-NEXT: v_add_i32_e32 v2, vcc, 64, v0
; SI-NEXT: buffer_store_dword v1, v2, s[0:3], 0 offen
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_mul_f32_e64 v1, 1.0, s72
+; SI-NEXT: v_mul_f32_e64 v1, 1.0, s38
; SI-NEXT: v_lshrrev_b32_e32 v1, 16, v1
; SI-NEXT: v_mul_f32_e64 v2, 1.0, s99
; SI-NEXT: v_alignbit_b32 v1, v1, v2, 16
; SI-NEXT: v_add_i32_e32 v2, vcc, 0x44, v0
-; SI-NEXT: v_readlane_b32 s4, v42, 55
; SI-NEXT: buffer_store_dword v1, v2, s[0:3], 0 offen
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_mul_f32_e64 v1, 1.0, s4
+; SI-NEXT: v_mul_f32_e64 v1, 1.0, s48
; SI-NEXT: v_lshrrev_b32_e32 v1, 16, v1
-; SI-NEXT: v_mul_f32_e64 v2, 1.0, s57
+; SI-NEXT: v_mul_f32_e64 v2, 1.0, s39
; SI-NEXT: v_alignbit_b32 v1, v1, v2, 16
; SI-NEXT: v_add_i32_e32 v2, vcc, 0x48, v0
; SI-NEXT: buffer_store_dword v1, v2, s[0:3], 0 offen
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_mul_f32_e64 v1, 1.0, s49
; SI-NEXT: v_lshrrev_b32_e32 v1, 16, v1
-; SI-NEXT: v_mul_f32_e64 v2, 1.0, s8
+; SI-NEXT: v_mul_f32_e64 v2, 1.0, s98
; SI-NEXT: v_alignbit_b32 v1, v1, v2, 16
; SI-NEXT: v_add_i32_e32 v2, vcc, 0x4c, v0
; SI-NEXT: buffer_store_dword v1, v2, s[0:3], 0 offen
@@ -158551,7 +158592,7 @@ define inreg <64 x bfloat> @bitcast_v128i8_to_v64bf16_scalar(<128 x i8> inreg %a
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_mul_f32_e64 v1, 1.0, s52
; SI-NEXT: v_lshrrev_b32_e32 v1, 16, v1
-; SI-NEXT: v_mul_f32_e64 v2, 1.0, s97
+; SI-NEXT: v_mul_f32_e64 v2, 1.0, s12
; SI-NEXT: v_alignbit_b32 v1, v1, v2, 16
; SI-NEXT: v_add_i32_e32 v2, vcc, 0x54, v0
; SI-NEXT: buffer_store_dword v1, v2, s[0:3], 0 offen
@@ -158565,7 +158606,7 @@ define inreg <64 x bfloat> @bitcast_v128i8_to_v64bf16_scalar(<128 x i8> inreg %a
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_mul_f32_e64 v1, 1.0, s55
; SI-NEXT: v_lshrrev_b32_e32 v1, 16, v1
-; SI-NEXT: v_mul_f32_e64 v2, 1.0, s28
+; SI-NEXT: v_mul_f32_e64 v2, 1.0, s96
; SI-NEXT: v_alignbit_b32 v1, v1, v2, 16
; SI-NEXT: v_add_i32_e32 v2, vcc, 0x5c, v0
; SI-NEXT: buffer_store_dword v1, v2, s[0:3], 0 offen
@@ -158584,21 +158625,21 @@ define inreg <64 x bfloat> @bitcast_v128i8_to_v64bf16_scalar(<128 x i8> inreg %a
; SI-NEXT: v_add_i32_e32 v2, vcc, 0x64, v0
; SI-NEXT: buffer_store_dword v1, v2, s[0:3], 0 offen
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_mul_f32_e64 v1, 1.0, s68
+; SI-NEXT: v_mul_f32_e64 v1, 1.0, s45
; SI-NEXT: v_lshrrev_b32_e32 v1, 16, v1
; SI-NEXT: v_mul_f32_e64 v2, 1.0, s67
; SI-NEXT: v_alignbit_b32 v1, v1, v2, 16
; SI-NEXT: v_add_i32_e32 v2, vcc, 0x68, v0
; SI-NEXT: buffer_store_dword v1, v2, s[0:3], 0 offen
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_mul_f32_e64 v1, 1.0, s69
+; SI-NEXT: v_mul_f32_e64 v1, 1.0, s57
; SI-NEXT: v_lshrrev_b32_e32 v1, 16, v1
; SI-NEXT: v_mul_f32_e64 v2, 1.0, s86
; SI-NEXT: v_alignbit_b32 v1, v1, v2, 16
; SI-NEXT: v_add_i32_e32 v2, vcc, 0x6c, v0
; SI-NEXT: buffer_store_dword v1, v2, s[0:3], 0 offen
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_mul_f32_e64 v1, 1.0, s62
+; SI-NEXT: v_mul_f32_e64 v1, 1.0, s13
; SI-NEXT: v_lshrrev_b32_e32 v1, 16, v1
; SI-NEXT: v_mul_f32_e64 v2, 1.0, s70
; SI-NEXT: v_alignbit_b32 v1, v1, v2, 16
@@ -158612,14 +158653,14 @@ define inreg <64 x bfloat> @bitcast_v128i8_to_v64bf16_scalar(<128 x i8> inreg %a
; SI-NEXT: v_add_i32_e32 v2, vcc, 0x74, v0
; SI-NEXT: buffer_store_dword v1, v2, s[0:3], 0 offen
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_mul_f32_e64 v1, 1.0, s92
+; SI-NEXT: v_mul_f32_e64 v1, 1.0, s82
; SI-NEXT: v_lshrrev_b32_e32 v1, 16, v1
-; SI-NEXT: v_mul_f32_e64 v2, 1.0, s11
+; SI-NEXT: v_mul_f32_e64 v2, 1.0, s25
; SI-NEXT: v_alignbit_b32 v1, v1, v2, 16
; SI-NEXT: v_add_i32_e32 v2, vcc, 0x78, v0
; SI-NEXT: buffer_store_dword v1, v2, s[0:3], 0 offen
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_mul_f32_e64 v1, 1.0, s83
+; SI-NEXT: v_mul_f32_e64 v1, 1.0, s28
; SI-NEXT: v_lshrrev_b32_e32 v1, 16, v1
; SI-NEXT: v_mul_f32_e64 v2, 1.0, s84
; SI-NEXT: v_alignbit_b32 v1, v1, v2, 16
@@ -158666,109 +158707,118 @@ define inreg <64 x bfloat> @bitcast_v128i8_to_v64bf16_scalar(<128 x i8> inreg %a
; SI-NEXT: buffer_load_dword v41, off, s[0:3], s32 offset:344 ; 4-byte Folded Reload
; SI-NEXT: buffer_load_dword v42, off, s[0:3], s32 offset:348 ; 4-byte Folded Reload
; SI-NEXT: buffer_load_dword v43, off, s[0:3], s32 offset:352 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v44, off, s[0:3], s32 offset:356 ; 4-byte Folded Reload
; SI-NEXT: s_mov_b64 exec, s[4:5]
; SI-NEXT: s_waitcnt vmcnt(0) expcnt(0)
; SI-NEXT: s_setpc_b64 s[30:31]
; SI-NEXT: .LBB89_4:
-; SI-NEXT: ; implicit-def: $sgpr8
-; SI-NEXT: ; kill: killed $sgpr8
-; SI-NEXT: s_mov_b32 s7, s6
-; SI-NEXT: ; implicit-def: $sgpr8
; SI-NEXT: ; implicit-def: $sgpr6
-; SI-NEXT: ; kill: killed $sgpr8
-; SI-NEXT: v_readlane_b32 s58, v43, 19
-; SI-NEXT: ; implicit-def: $sgpr8
-; SI-NEXT: s_mov_b32 s95, s47
-; SI-NEXT: s_mov_b32 s94, s21
-; SI-NEXT: s_mov_b32 s93, s61
-; SI-NEXT: s_mov_b32 s34, s73
-; SI-NEXT: s_mov_b32 s91, s75
-; SI-NEXT: v_readlane_b32 s56, v43, 10
-; SI-NEXT: s_mov_b32 s36, s63
-; SI-NEXT: s_mov_b32 s38, s59
-; SI-NEXT: s_mov_b32 s37, s42
-; SI-NEXT: v_readlane_b32 s30, v43, 17
-; SI-NEXT: v_readlane_b32 s98, v43, 6
-; SI-NEXT: s_mov_b32 s46, s45
-; SI-NEXT: s_mov_b32 s31, s43
-; SI-NEXT: s_mov_b32 s78, s40
-; SI-NEXT: v_readlane_b32 s15, v43, 14
-; SI-NEXT: s_mov_b32 s39, s57
-; SI-NEXT: s_mov_b32 s48, s13
-; SI-NEXT: v_readlane_b32 s41, v43, 13
-; SI-NEXT: v_readlane_b32 s44, v43, 5
-; SI-NEXT: v_readlane_b32 s9, v43, 11
-; SI-NEXT: v_readlane_b32 s14, v43, 12
-; SI-NEXT: v_readlane_b32 s81, v43, 9
-; SI-NEXT: v_readlane_b32 s10, v43, 16
-; SI-NEXT: v_readlane_b32 s12, v43, 4
-; SI-NEXT: v_readlane_b32 s96, v43, 7
-; SI-NEXT: v_readlane_b32 s82, v43, 8
-; SI-NEXT: v_readlane_b32 s71, v43, 15
+; SI-NEXT: ; kill: killed $sgpr6
+; SI-NEXT: s_mov_b32 s89, s76
+; SI-NEXT: ; implicit-def: $sgpr6
+; SI-NEXT: ; kill: killed $sgpr6
+; SI-NEXT: s_mov_b32 s88, s57
+; SI-NEXT: ; implicit-def: $sgpr6
+; SI-NEXT: ; kill: killed $sgpr6
+; SI-NEXT: s_mov_b32 s94, s26
+; SI-NEXT: ; implicit-def: $sgpr6
+; SI-NEXT: ; kill: killed $sgpr6
+; SI-NEXT: s_mov_b32 s93, s74
+; SI-NEXT: ; implicit-def: $sgpr6
+; SI-NEXT: ; kill: killed $sgpr6
+; SI-NEXT: s_mov_b32 s90, s63
+; SI-NEXT: ; implicit-def: $sgpr6
+; SI-NEXT: ; kill: killed $sgpr6
+; SI-NEXT: s_mov_b32 s91, s77
+; SI-NEXT: ; implicit-def: $sgpr6
+; SI-NEXT: v_readlane_b32 s61, v44, 31
+; SI-NEXT: v_readlane_b32 s74, v44, 23
+; SI-NEXT: v_readlane_b32 s60, v44, 30
+; SI-NEXT: v_readlane_b32 s63, v44, 22
+; SI-NEXT: v_readlane_b32 s8, v44, 15
+; SI-NEXT: v_readlane_b32 s46, v44, 27
+; SI-NEXT: v_readlane_b32 s59, v44, 28
+; SI-NEXT: v_readlane_b32 s68, v44, 29
+; SI-NEXT: s_mov_b32 s92, s56
+; SI-NEXT: v_readlane_b32 s69, v44, 26
+; SI-NEXT: s_mov_b32 s76, s58
+; SI-NEXT: v_readlane_b32 s44, v44, 11
+; SI-NEXT: v_readlane_b32 s42, v44, 34
+; SI-NEXT: v_readlane_b32 s58, v44, 24
+; SI-NEXT: v_readlane_b32 s40, v44, 25
+; SI-NEXT: v_readlane_b32 s41, v44, 19
+; SI-NEXT: v_readlane_b32 s56, v44, 32
+; SI-NEXT: v_readlane_b32 s47, v44, 33
+; SI-NEXT: v_readlane_b32 s14, v44, 18
+; SI-NEXT: v_readlane_b32 s11, v44, 10
+; SI-NEXT: v_readlane_b32 s9, v44, 16
+; SI-NEXT: v_readlane_b32 s10, v44, 17
+; SI-NEXT: v_readlane_b32 s15, v44, 14
+; SI-NEXT: v_readlane_b32 s7, v44, 9
+; SI-NEXT: v_readlane_b32 s62, v44, 8
+; SI-NEXT: v_readlane_b32 s97, v44, 12
+; SI-NEXT: v_readlane_b32 s83, v44, 13
+; SI-NEXT: v_readlane_b32 s81, v44, 20
+; SI-NEXT: v_readlane_b32 s71, v44, 21
; SI-NEXT: ; kill: killed $sgpr6
; SI-NEXT: ; implicit-def: $sgpr6
-; SI-NEXT: ; kill: killed $sgpr8
-; SI-NEXT: ; implicit-def: $sgpr8
; SI-NEXT: ; implicit-def: $sgpr26
; SI-NEXT: ; implicit-def: $sgpr27
; SI-NEXT: ; implicit-def: $sgpr20
; SI-NEXT: ; implicit-def: $sgpr24
; SI-NEXT: ; implicit-def: $sgpr29
-; SI-NEXT: ; implicit-def: $sgpr35
-; SI-NEXT: ; implicit-def: $sgpr74
-; SI-NEXT: ; implicit-def: $sgpr90
+; SI-NEXT: ; implicit-def: $sgpr72
+; SI-NEXT: ; implicit-def: $sgpr23
+; SI-NEXT: ; implicit-def: $sgpr73
; SI-NEXT: ; implicit-def: $sgpr21
-; SI-NEXT: ; implicit-def: $sgpr25
-; SI-NEXT: ; implicit-def: $sgpr22
; SI-NEXT: ; implicit-def: $sgpr75
-; SI-NEXT: ; implicit-def: $sgpr76
-; SI-NEXT: ; implicit-def: $sgpr61
-; SI-NEXT: ; implicit-def: $sgpr16
+; SI-NEXT: ; implicit-def: $sgpr22
+; SI-NEXT: ; implicit-def: $sgpr78
; SI-NEXT: ; implicit-def: $sgpr77
; SI-NEXT: ; implicit-def: $sgpr79
-; SI-NEXT: ; implicit-def: $sgpr63
+; SI-NEXT: ; implicit-def: $sgpr16
; SI-NEXT: ; implicit-def: $sgpr19
-; SI-NEXT: ; implicit-def: $sgpr89
-; SI-NEXT: ; implicit-def: $sgpr88
-; SI-NEXT: ; implicit-def: $sgpr13
; SI-NEXT: ; implicit-def: $sgpr18
-; SI-NEXT: ; implicit-def: $sgpr60
-; SI-NEXT: ; implicit-def: $sgpr59
-; SI-NEXT: ; implicit-def: $sgpr73
-; SI-NEXT: ; implicit-def: $sgpr17
; SI-NEXT: ; kill: killed $sgpr6
-; SI-NEXT: ; implicit-def: $sgpr23
+; SI-NEXT: ; implicit-def: $sgpr95
+; SI-NEXT: ; implicit-def: $sgpr17
+; SI-NEXT: ; implicit-def: $sgpr30
+; SI-NEXT: ; implicit-def: $sgpr31
+; SI-NEXT: ; implicit-def: $sgpr34
; SI-NEXT: ; implicit-def: $sgpr6
+; SI-NEXT: ; implicit-def: $sgpr35
+; SI-NEXT: ; implicit-def: $sgpr36
+; SI-NEXT: ; implicit-def: $sgpr37
; SI-NEXT: ; implicit-def: $sgpr99
-; SI-NEXT: ; implicit-def: $sgpr72
-; SI-NEXT: ; implicit-def: $sgpr57
-; SI-NEXT: ; kill: killed $sgpr8
-; SI-NEXT: ; implicit-def: $sgpr8
+; SI-NEXT: ; implicit-def: $sgpr38
+; SI-NEXT: ; implicit-def: $sgpr39
+; SI-NEXT: ; implicit-def: $sgpr48
+; SI-NEXT: ; implicit-def: $sgpr98
; SI-NEXT: ; implicit-def: $sgpr49
; SI-NEXT: ; implicit-def: $sgpr50
; SI-NEXT: ; implicit-def: $sgpr51
-; SI-NEXT: ; implicit-def: $sgpr97
+; SI-NEXT: ; implicit-def: $sgpr12
; SI-NEXT: ; implicit-def: $sgpr52
; SI-NEXT: ; implicit-def: $sgpr53
; SI-NEXT: ; implicit-def: $sgpr54
-; SI-NEXT: ; implicit-def: $sgpr28
+; SI-NEXT: ; implicit-def: $sgpr96
; SI-NEXT: ; implicit-def: $sgpr55
; SI-NEXT: ; implicit-def: $sgpr64
; SI-NEXT: ; implicit-def: $sgpr65
; SI-NEXT: ; implicit-def: $sgpr87
; SI-NEXT: ; implicit-def: $sgpr66
; SI-NEXT: ; implicit-def: $sgpr67
-; SI-NEXT: ; implicit-def: $sgpr68
+; SI-NEXT: ; implicit-def: $sgpr45
; SI-NEXT: ; implicit-def: $sgpr86
-; SI-NEXT: ; implicit-def: $sgpr69
+; SI-NEXT: ; implicit-def: $sgpr57
; SI-NEXT: ; implicit-def: $sgpr70
-; SI-NEXT: ; implicit-def: $sgpr62
+; SI-NEXT: ; implicit-def: $sgpr13
; SI-NEXT: ; implicit-def: $sgpr85
; SI-NEXT: ; implicit-def: $sgpr80
-; SI-NEXT: ; implicit-def: $sgpr11
-; SI-NEXT: ; implicit-def: $sgpr92
+; SI-NEXT: ; implicit-def: $sgpr25
+; SI-NEXT: ; implicit-def: $sgpr82
; SI-NEXT: ; implicit-def: $sgpr84
-; SI-NEXT: ; implicit-def: $sgpr83
+; SI-NEXT: ; implicit-def: $sgpr28
; SI-NEXT: s_branch .LBB89_2
;
; VI-LABEL: bitcast_v128i8_to_v64bf16_scalar:
@@ -158831,13 +158881,14 @@ define inreg <64 x bfloat> @bitcast_v128i8_to_v64bf16_scalar(<128 x i8> inreg %a
; VI-NEXT: buffer_load_ushort v37, off, s[0:3], s32 offset:168
; VI-NEXT: buffer_load_ushort v38, off, s[0:3], s32 offset:176
; VI-NEXT: v_lshlrev_b32_e32 v1, 8, v1
-; VI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:508 ; 4-byte Folded Spill
+; VI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:504 ; 4-byte Folded Spill
; VI-NEXT: v_lshlrev_b32_e32 v1, 8, v3
; VI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:464 ; 4-byte Folded Spill
; VI-NEXT: v_lshlrev_b32_e32 v1, 8, v5
; VI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:468 ; 4-byte Folded Spill
; VI-NEXT: v_lshlrev_b32_e32 v1, 8, v7
; VI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:472 ; 4-byte Folded Spill
+; VI-NEXT: v_lshlrev_b32_e32 v40, 8, v27
; VI-NEXT: v_lshlrev_b32_e32 v9, 8, v9
; VI-NEXT: v_lshlrev_b32_e32 v11, 8, v11
; VI-NEXT: v_lshlrev_b32_e32 v13, 8, v13
@@ -158849,46 +158900,42 @@ define inreg <64 x bfloat> @bitcast_v128i8_to_v64bf16_scalar(<128 x i8> inreg %a
; VI-NEXT: v_lshlrev_b32_e32 v25, 8, v25
; VI-NEXT: s_waitcnt vmcnt(14)
; VI-NEXT: v_cmp_ne_u32_e32 vcc, 0, v0
-; VI-NEXT: v_lshlrev_b32_e32 v0, 8, v27
-; VI-NEXT: buffer_store_dword v0, off, s[0:3], s32 offset:476 ; 4-byte Folded Spill
; VI-NEXT: v_lshlrev_b32_e32 v0, 8, v29
-; VI-NEXT: buffer_store_dword v0, off, s[0:3], s32 offset:480 ; 4-byte Folded Spill
+; VI-NEXT: buffer_store_dword v0, off, s[0:3], s32 offset:476 ; 4-byte Folded Spill
; VI-NEXT: v_lshlrev_b32_e32 v0, 8, v2
-; VI-NEXT: buffer_store_dword v0, off, s[0:3], s32 offset:484 ; 4-byte Folded Spill
+; VI-NEXT: buffer_store_dword v0, off, s[0:3], s32 offset:480 ; 4-byte Folded Spill
; VI-NEXT: v_lshlrev_b32_e32 v0, 8, v4
-; VI-NEXT: buffer_store_dword v0, off, s[0:3], s32 offset:488 ; 4-byte Folded Spill
+; VI-NEXT: buffer_store_dword v0, off, s[0:3], s32 offset:484 ; 4-byte Folded Spill
; VI-NEXT: v_lshlrev_b32_e32 v0, 8, v6
-; VI-NEXT: buffer_store_dword v0, off, s[0:3], s32 offset:492 ; 4-byte Folded Spill
+; VI-NEXT: buffer_store_dword v0, off, s[0:3], s32 offset:488 ; 4-byte Folded Spill
; VI-NEXT: v_lshlrev_b32_e32 v0, 8, v8
-; VI-NEXT: buffer_store_dword v0, off, s[0:3], s32 offset:496 ; 4-byte Folded Spill
+; VI-NEXT: buffer_store_dword v0, off, s[0:3], s32 offset:492 ; 4-byte Folded Spill
; VI-NEXT: v_lshlrev_b32_e32 v0, 8, v10
-; VI-NEXT: buffer_store_dword v0, off, s[0:3], s32 offset:500 ; 4-byte Folded Spill
+; VI-NEXT: buffer_store_dword v0, off, s[0:3], s32 offset:496 ; 4-byte Folded Spill
; VI-NEXT: v_lshlrev_b32_e32 v0, 8, v12
-; VI-NEXT: buffer_store_dword v0, off, s[0:3], s32 offset:504 ; 4-byte Folded Spill
-; VI-NEXT: v_lshlrev_b32_e32 v0, 8, v26
-; VI-NEXT: buffer_store_dword v0, off, s[0:3], s32 offset:512 ; 4-byte Folded Spill
+; VI-NEXT: buffer_store_dword v0, off, s[0:3], s32 offset:500 ; 4-byte Folded Spill
+; VI-NEXT: v_lshlrev_b32_e32 v0, 8, v24
+; VI-NEXT: buffer_store_dword v0, off, s[0:3], s32 offset:508 ; 4-byte Folded Spill
; VI-NEXT: s_waitcnt vmcnt(14)
-; VI-NEXT: v_lshlrev_b32_e32 v0, 8, v28
-; VI-NEXT: buffer_store_dword v0, off, s[0:3], s32 offset:516 ; 4-byte Folded Spill
; VI-NEXT: v_lshlrev_b32_e32 v0, 8, v30
-; VI-NEXT: buffer_store_dword v0, off, s[0:3], s32 offset:520 ; 4-byte Folded Spill
+; VI-NEXT: buffer_store_dword v0, off, s[0:3], s32 offset:512 ; 4-byte Folded Spill
; VI-NEXT: v_lshlrev_b32_e32 v0, 8, v31
-; VI-NEXT: buffer_store_dword v0, off, s[0:3], s32 offset:560 ; 4-byte Folded Spill
+; VI-NEXT: buffer_store_dword v0, off, s[0:3], s32 offset:556 ; 4-byte Folded Spill
; VI-NEXT: v_lshlrev_b32_e32 v0, 8, v32
-; VI-NEXT: buffer_store_dword v0, off, s[0:3], s32 offset:564 ; 4-byte Folded Spill
-; VI-NEXT: v_lshlrev_b32_e32 v0, 8, v33
-; VI-NEXT: buffer_store_dword v0, off, s[0:3], s32 offset:592 ; 4-byte Folded Spill
-; VI-NEXT: v_lshlrev_b32_e32 v0, 8, v34
; VI-NEXT: buffer_store_dword v0, off, s[0:3], s32 offset:588 ; 4-byte Folded Spill
-; VI-NEXT: v_lshlrev_b32_e32 v0, 8, v35
+; VI-NEXT: v_lshlrev_b32_e32 v0, 8, v33
; VI-NEXT: buffer_store_dword v0, off, s[0:3], s32 offset:584 ; 4-byte Folded Spill
-; VI-NEXT: v_lshlrev_b32_e32 v0, 8, v36
+; VI-NEXT: v_lshlrev_b32_e32 v0, 8, v34
; VI-NEXT: buffer_store_dword v0, off, s[0:3], s32 offset:576 ; 4-byte Folded Spill
-; VI-NEXT: v_lshlrev_b32_e32 v0, 8, v37
+; VI-NEXT: v_lshlrev_b32_e32 v0, 8, v35
; VI-NEXT: buffer_store_dword v0, off, s[0:3], s32 offset:572 ; 4-byte Folded Spill
+; VI-NEXT: v_lshlrev_b32_e32 v0, 8, v36
+; VI-NEXT: buffer_store_dword v0, off, s[0:3], s32 offset:564 ; 4-byte Folded Spill
; VI-NEXT: s_waitcnt vmcnt(14)
+; VI-NEXT: v_lshlrev_b32_e32 v0, 8, v37
+; VI-NEXT: buffer_store_dword v0, off, s[0:3], s32 offset:560 ; 4-byte Folded Spill
; VI-NEXT: v_lshlrev_b32_e32 v0, 8, v38
-; VI-NEXT: buffer_store_dword v0, off, s[0:3], s32 offset:580 ; 4-byte Folded Spill
+; VI-NEXT: buffer_store_dword v0, off, s[0:3], s32 offset:568 ; 4-byte Folded Spill
; VI-NEXT: buffer_load_ushort v0, off, s[0:3], s32 offset:184
; VI-NEXT: buffer_load_ushort v1, off, s[0:3], s32 offset:192
; VI-NEXT: buffer_load_ushort v2, off, s[0:3], s32 offset:200
@@ -158897,34 +158944,37 @@ define inreg <64 x bfloat> @bitcast_v128i8_to_v64bf16_scalar(<128 x i8> inreg %a
; VI-NEXT: buffer_load_ushort v5, off, s[0:3], s32 offset:224
; VI-NEXT: buffer_load_ushort v6, off, s[0:3], s32 offset:232
; VI-NEXT: buffer_load_ushort v7, off, s[0:3], s32 offset:240
-; VI-NEXT: v_lshlrev_b32_e32 v45, 8, v22
-; VI-NEXT: v_lshlrev_b32_e32 v8, 8, v24
+; VI-NEXT: v_lshlrev_b32_e32 v8, 8, v26
+; VI-NEXT: v_lshlrev_b32_e32 v10, 8, v28
; VI-NEXT: s_and_b64 s[4:5], vcc, exec
; VI-NEXT: v_lshlrev_b32_e32 v14, 8, v14
; VI-NEXT: v_lshlrev_b32_e32 v16, 8, v16
; VI-NEXT: v_lshlrev_b32_e32 v18, 8, v18
; VI-NEXT: v_lshlrev_b32_e32 v20, 8, v20
+; VI-NEXT: v_lshlrev_b32_e32 v22, 8, v22
; VI-NEXT: s_waitcnt vmcnt(7)
-; VI-NEXT: v_lshlrev_b32_e32 v22, 8, v0
+; VI-NEXT: v_lshlrev_b32_e32 v24, 8, v0
; VI-NEXT: s_waitcnt vmcnt(6)
; VI-NEXT: v_lshlrev_b32_e32 v0, 8, v1
-; VI-NEXT: buffer_store_dword v0, off, s[0:3], s32 offset:568 ; 4-byte Folded Spill
-; VI-NEXT: s_waitcnt vmcnt(5)
+; VI-NEXT: buffer_store_dword v0, off, s[0:3], s32 offset:592 ; 4-byte Folded Spill
+; VI-NEXT: s_waitcnt vmcnt(6)
+; VI-NEXT: v_lshlrev_b32_e32 v0, 8, v2
+; VI-NEXT: buffer_store_dword v0, off, s[0:3], s32 offset:580 ; 4-byte Folded Spill
+; VI-NEXT: s_waitcnt vmcnt(6)
; VI-NEXT: v_lshlrev_b32_e32 v0, 8, v3
; VI-NEXT: buffer_store_dword v0, off, s[0:3], s32 offset:600 ; 4-byte Folded Spill
-; VI-NEXT: s_waitcnt vmcnt(5)
+; VI-NEXT: s_waitcnt vmcnt(6)
; VI-NEXT: v_lshlrev_b32_e32 v0, 8, v4
; VI-NEXT: buffer_store_dword v0, off, s[0:3], s32 offset:596 ; 4-byte Folded Spill
-; VI-NEXT: s_waitcnt vmcnt(5)
+; VI-NEXT: s_waitcnt vmcnt(6)
; VI-NEXT: v_lshlrev_b32_e32 v0, 8, v5
; VI-NEXT: buffer_store_dword v0, off, s[0:3], s32 offset:608 ; 4-byte Folded Spill
-; VI-NEXT: s_waitcnt vmcnt(5)
+; VI-NEXT: s_waitcnt vmcnt(6)
; VI-NEXT: v_lshlrev_b32_e32 v0, 8, v6
; VI-NEXT: buffer_store_dword v0, off, s[0:3], s32 offset:612 ; 4-byte Folded Spill
-; VI-NEXT: s_waitcnt vmcnt(5)
+; VI-NEXT: s_waitcnt vmcnt(6)
; VI-NEXT: v_lshlrev_b32_e32 v0, 8, v7
; VI-NEXT: buffer_store_dword v0, off, s[0:3], s32 offset:604 ; 4-byte Folded Spill
-; VI-NEXT: v_lshlrev_b32_e32 v24, 8, v2
; VI-NEXT: buffer_load_ushort v0, off, s[0:3], s32 offset:248
; VI-NEXT: buffer_load_ushort v1, off, s[0:3], s32 offset:256
; VI-NEXT: buffer_load_ushort v2, off, s[0:3], s32 offset:264
@@ -158943,6 +158993,11 @@ define inreg <64 x bfloat> @bitcast_v128i8_to_v64bf16_scalar(<128 x i8> inreg %a
; VI-NEXT: s_waitcnt vmcnt(4)
; VI-NEXT: v_lshlrev_b32_e32 v0, 8, v4
; VI-NEXT: buffer_store_dword v0, off, s[0:3], s32 offset:620 ; 4-byte Folded Spill
+; VI-NEXT: s_waitcnt vmcnt(3)
+; VI-NEXT: v_lshlrev_b32_e32 v4, 8, v6
+; VI-NEXT: s_waitcnt vmcnt(2)
+; VI-NEXT: v_lshlrev_b32_e32 v0, 8, v7
+; VI-NEXT: buffer_store_dword v0, off, s[0:3], s32 offset:544 ; 4-byte Folded Spill
; VI-NEXT: buffer_load_ushort v0, off, s[0:3], s32 offset:312
; VI-NEXT: buffer_load_ushort v1, off, s[0:3], s32 offset:320
; VI-NEXT: buffer_load_ushort v2, off, s[0:3], s32 offset:328
@@ -158951,12 +159006,8 @@ define inreg <64 x bfloat> @bitcast_v128i8_to_v64bf16_scalar(<128 x i8> inreg %a
; VI-NEXT: buffer_load_ushort v33, off, s[0:3], s32 offset:20
; VI-NEXT: buffer_load_ushort v34, off, s[0:3], s32 offset:28
; VI-NEXT: buffer_load_ushort v35, off, s[0:3], s32 offset:36
-; VI-NEXT: s_waitcnt vmcnt(11)
-; VI-NEXT: v_lshlrev_b32_e32 v4, 8, v6
; VI-NEXT: v_lshlrev_b32_e32 v28, 8, v3
; VI-NEXT: v_lshlrev_b32_e32 v3, 8, v5
-; VI-NEXT: s_waitcnt vmcnt(10)
-; VI-NEXT: v_lshlrev_b32_e32 v5, 8, v7
; VI-NEXT: s_waitcnt vmcnt(7)
; VI-NEXT: v_lshlrev_b32_e32 v6, 8, v0
; VI-NEXT: buffer_load_ushort v36, off, s[0:3], s32 offset:44
@@ -158965,47 +159016,45 @@ define inreg <64 x bfloat> @bitcast_v128i8_to_v64bf16_scalar(<128 x i8> inreg %a
; VI-NEXT: buffer_load_ushort v39, off, s[0:3], s32 offset:68
; VI-NEXT: buffer_load_ushort v48, off, s[0:3], s32 offset:76
; VI-NEXT: buffer_load_ushort v49, off, s[0:3], s32 offset:84
-; VI-NEXT: buffer_load_ushort v50, off, s[0:3], s32 offset:92
-; VI-NEXT: buffer_load_ushort v51, off, s[0:3], s32 offset:100
-; VI-NEXT: buffer_load_ushort v54, off, s[0:3], s32 offset:108
+; VI-NEXT: buffer_load_ushort v51, off, s[0:3], s32 offset:92
+; VI-NEXT: buffer_load_ushort v53, off, s[0:3], s32 offset:100
+; VI-NEXT: buffer_load_ushort v55, off, s[0:3], s32 offset:108
; VI-NEXT: buffer_load_ushort v41, off, s[0:3], s32 offset:116
; VI-NEXT: buffer_load_ushort v56, off, s[0:3], s32 offset:124
-; VI-NEXT: buffer_load_ushort v55, off, s[0:3], s32 offset:132
-; VI-NEXT: buffer_load_ushort v52, off, s[0:3], s32 offset:140
-; VI-NEXT: buffer_load_ushort v43, off, s[0:3], s32 offset:148
-; VI-NEXT: buffer_load_ushort v53, off, s[0:3], s32 offset:156
-; VI-NEXT: buffer_load_ushort v46, off, s[0:3], s32 offset:164
-; VI-NEXT: buffer_load_ushort v58, off, s[0:3], s32 offset:172
+; VI-NEXT: buffer_load_ushort v50, off, s[0:3], s32 offset:132
+; VI-NEXT: buffer_load_ushort v43, off, s[0:3], s32 offset:140
+; VI-NEXT: buffer_load_ushort v58, off, s[0:3], s32 offset:148
+; VI-NEXT: buffer_load_ushort v52, off, s[0:3], s32 offset:156
+; VI-NEXT: buffer_load_ushort v60, off, s[0:3], s32 offset:164
+; VI-NEXT: buffer_load_ushort v54, off, s[0:3], s32 offset:172
; VI-NEXT: buffer_load_ushort v47, off, s[0:3], s32 offset:180
; VI-NEXT: buffer_load_ushort v63, off, s[0:3], s32 offset:188
-; VI-NEXT: buffer_load_ushort v57, off, s[0:3], s32 offset:196
+; VI-NEXT: buffer_load_ushort v46, off, s[0:3], s32 offset:196
; VI-NEXT: buffer_load_ushort v61, off, s[0:3], s32 offset:204
-; VI-NEXT: buffer_load_ushort v60, off, s[0:3], s32 offset:212
+; VI-NEXT: buffer_load_ushort v57, off, s[0:3], s32 offset:212
; VI-NEXT: buffer_load_ushort v59, off, s[0:3], s32 offset:220
-; VI-NEXT: buffer_load_ushort v40, off, s[0:3], s32 offset:228
-; VI-NEXT: buffer_load_ushort v42, off, s[0:3], s32 offset:236
-; VI-NEXT: buffer_load_ushort v62, off, s[0:3], s32 offset:244
-; VI-NEXT: buffer_load_ushort v44, off, s[0:3], s32 offset:252
+; VI-NEXT: buffer_load_ushort v42, off, s[0:3], s32 offset:228
+; VI-NEXT: buffer_load_ushort v44, off, s[0:3], s32 offset:236
+; VI-NEXT: buffer_load_ushort v45, off, s[0:3], s32 offset:244
+; VI-NEXT: buffer_load_ushort v62, off, s[0:3], s32 offset:252
; VI-NEXT: buffer_load_ushort v0, off, s[0:3], s32 offset:260
; VI-NEXT: s_waitcnt vmcnt(14)
-; VI-NEXT: v_lshlrev_b32_e32 v1, 8, v1
-; VI-NEXT: v_lshlrev_b32_e32 v2, 8, v2
-; VI-NEXT: s_waitcnt vmcnt(0)
-; VI-NEXT: buffer_store_dword v0, off, s[0:3], s32 offset:552 ; 4-byte Folded Spill
-; VI-NEXT: buffer_load_ushort v29, off, s[0:3], s32 offset:268
-; VI-NEXT: buffer_load_ushort v0, off, s[0:3], s32 offset:276
+; VI-NEXT: v_lshlrev_b32_e32 v5, 8, v1
+; VI-NEXT: v_lshlrev_b32_e32 v1, 8, v2
; VI-NEXT: s_waitcnt vmcnt(0)
-; VI-NEXT: buffer_store_dword v0, off, s[0:3], s32 offset:556 ; 4-byte Folded Spill
-; VI-NEXT: buffer_load_ushort v30, off, s[0:3], s32 offset:284
-; VI-NEXT: buffer_load_ushort v31, off, s[0:3], s32 offset:292
-; VI-NEXT: buffer_load_ushort v0, off, s[0:3], s32 offset:300
-; VI-NEXT: buffer_load_ushort v7, off, s[0:3], s32 offset:308
+; VI-NEXT: buffer_store_dword v0, off, s[0:3], s32 offset:548 ; 4-byte Folded Spill
+; VI-NEXT: buffer_load_ushort v0, off, s[0:3], s32 offset:268
; VI-NEXT: s_waitcnt vmcnt(0)
-; VI-NEXT: buffer_store_dword v7, off, s[0:3], s32 offset:528 ; 4-byte Folded Spill
-; VI-NEXT: buffer_load_ushort v7, off, s[0:3], s32 offset:316
-; VI-NEXT: buffer_load_ushort v10, off, s[0:3], s32 offset:324
+; VI-NEXT: buffer_store_dword v0, off, s[0:3], s32 offset:552 ; 4-byte Folded Spill
+; VI-NEXT: buffer_load_ushort v29, off, s[0:3], s32 offset:276
+; VI-NEXT: buffer_load_ushort v31, off, s[0:3], s32 offset:284
+; VI-NEXT: buffer_load_ushort v30, off, s[0:3], s32 offset:292
+; VI-NEXT: buffer_load_ushort v7, off, s[0:3], s32 offset:300
+; VI-NEXT: buffer_load_ushort v0, off, s[0:3], s32 offset:308
; VI-NEXT: s_waitcnt vmcnt(0)
-; VI-NEXT: buffer_store_dword v10, off, s[0:3], s32 offset:540 ; 4-byte Folded Spill
+; VI-NEXT: buffer_store_dword v0, off, s[0:3], s32 offset:520 ; 4-byte Folded Spill
+; VI-NEXT: buffer_load_ushort v0, off, s[0:3], s32 offset:316
+; VI-NEXT: buffer_load_ushort v2, off, s[0:3], s32 offset:324
; VI-NEXT: buffer_store_dword v9, off, s[0:3], s32 offset:624 ; 4-byte Folded Spill
; VI-NEXT: buffer_store_dword v11, off, s[0:3], s32 offset:628 ; 4-byte Folded Spill
; VI-NEXT: buffer_store_dword v13, off, s[0:3], s32 offset:632 ; 4-byte Folded Spill
@@ -159015,46 +159064,50 @@ define inreg <64 x bfloat> @bitcast_v128i8_to_v64bf16_scalar(<128 x i8> inreg %a
; VI-NEXT: buffer_store_dword v21, off, s[0:3], s32 offset:648 ; 4-byte Folded Spill
; VI-NEXT: buffer_store_dword v23, off, s[0:3], s32 offset:652 ; 4-byte Folded Spill
; VI-NEXT: buffer_store_dword v25, off, s[0:3], s32 offset:656 ; 4-byte Folded Spill
-; VI-NEXT: buffer_store_dword v14, off, s[0:3], s32 offset:660 ; 4-byte Folded Spill
-; VI-NEXT: buffer_store_dword v16, off, s[0:3], s32 offset:664 ; 4-byte Folded Spill
-; VI-NEXT: buffer_store_dword v18, off, s[0:3], s32 offset:668 ; 4-byte Folded Spill
-; VI-NEXT: buffer_store_dword v20, off, s[0:3], s32 offset:672 ; 4-byte Folded Spill
-; VI-NEXT: buffer_store_dword v45, off, s[0:3], s32 offset:676 ; 4-byte Folded Spill
-; VI-NEXT: buffer_store_dword v8, off, s[0:3], s32 offset:680 ; 4-byte Folded Spill
+; VI-NEXT: buffer_store_dword v40, off, s[0:3], s32 offset:660 ; 4-byte Folded Spill
+; VI-NEXT: buffer_store_dword v14, off, s[0:3], s32 offset:664 ; 4-byte Folded Spill
+; VI-NEXT: buffer_store_dword v16, off, s[0:3], s32 offset:668 ; 4-byte Folded Spill
+; VI-NEXT: buffer_store_dword v18, off, s[0:3], s32 offset:672 ; 4-byte Folded Spill
+; VI-NEXT: buffer_store_dword v20, off, s[0:3], s32 offset:676 ; 4-byte Folded Spill
+; VI-NEXT: buffer_store_dword v22, off, s[0:3], s32 offset:680 ; 4-byte Folded Spill
; VI-NEXT: buffer_store_dword v12, off, s[0:3], s32 offset:684 ; 4-byte Folded Spill
; VI-NEXT: buffer_store_dword v32, off, s[0:3], s32 offset:688 ; 4-byte Folded Spill
; VI-NEXT: buffer_store_dword v33, off, s[0:3], s32 offset:692 ; 4-byte Folded Spill
; VI-NEXT: buffer_store_dword v34, off, s[0:3], s32 offset:696 ; 4-byte Folded Spill
; VI-NEXT: buffer_store_dword v35, off, s[0:3], s32 offset:700 ; 4-byte Folded Spill
-; VI-NEXT: buffer_store_dword v36, off, s[0:3], s32 offset:704 ; 4-byte Folded Spill
-; VI-NEXT: buffer_store_dword v37, off, s[0:3], s32 offset:708 ; 4-byte Folded Spill
-; VI-NEXT: buffer_store_dword v38, off, s[0:3], s32 offset:712 ; 4-byte Folded Spill
-; VI-NEXT: buffer_store_dword v39, off, s[0:3], s32 offset:716 ; 4-byte Folded Spill
-; VI-NEXT: buffer_store_dword v48, off, s[0:3], s32 offset:720 ; 4-byte Folded Spill
-; VI-NEXT: buffer_store_dword v49, off, s[0:3], s32 offset:724 ; 4-byte Folded Spill
-; VI-NEXT: buffer_store_dword v50, off, s[0:3], s32 offset:728 ; 4-byte Folded Spill
-; VI-NEXT: buffer_store_dword v51, off, s[0:3], s32 offset:732 ; 4-byte Folded Spill
-; VI-NEXT: buffer_store_dword v54, off, s[0:3], s32 offset:736 ; 4-byte Folded Spill
-; VI-NEXT: buffer_store_dword v41, off, s[0:3], s32 offset:740 ; 4-byte Folded Spill
-; VI-NEXT: buffer_store_dword v56, off, s[0:3], s32 offset:744 ; 4-byte Folded Spill
-; VI-NEXT: buffer_store_dword v55, off, s[0:3], s32 offset:748 ; 4-byte Folded Spill
-; VI-NEXT: buffer_store_dword v52, off, s[0:3], s32 offset:752 ; 4-byte Folded Spill
-; VI-NEXT: buffer_store_dword v43, off, s[0:3], s32 offset:756 ; 4-byte Folded Spill
-; VI-NEXT: buffer_store_dword v53, off, s[0:3], s32 offset:760 ; 4-byte Folded Spill
-; VI-NEXT: buffer_store_dword v46, off, s[0:3], s32 offset:764 ; 4-byte Folded Spill
-; VI-NEXT: buffer_store_dword v58, off, s[0:3], s32 offset:768 ; 4-byte Folded Spill
-; VI-NEXT: buffer_store_dword v47, off, s[0:3], s32 offset:772 ; 4-byte Folded Spill
-; VI-NEXT: buffer_store_dword v63, off, s[0:3], s32 offset:776 ; 4-byte Folded Spill
-; VI-NEXT: buffer_store_dword v57, off, s[0:3], s32 offset:780 ; 4-byte Folded Spill
-; VI-NEXT: buffer_store_dword v61, off, s[0:3], s32 offset:784 ; 4-byte Folded Spill
-; VI-NEXT: buffer_store_dword v29, off, s[0:3], s32 offset:788 ; 4-byte Folded Spill
-; VI-NEXT: buffer_store_dword v30, off, s[0:3], s32 offset:792 ; 4-byte Folded Spill
-; VI-NEXT: buffer_store_dword v31, off, s[0:3], s32 offset:796 ; 4-byte Folded Spill
+; VI-NEXT: buffer_store_dword v8, off, s[0:3], s32 offset:704 ; 4-byte Folded Spill
+; VI-NEXT: buffer_store_dword v36, off, s[0:3], s32 offset:708 ; 4-byte Folded Spill
+; VI-NEXT: buffer_store_dword v37, off, s[0:3], s32 offset:712 ; 4-byte Folded Spill
+; VI-NEXT: buffer_store_dword v38, off, s[0:3], s32 offset:716 ; 4-byte Folded Spill
+; VI-NEXT: buffer_store_dword v39, off, s[0:3], s32 offset:720 ; 4-byte Folded Spill
+; VI-NEXT: buffer_store_dword v10, off, s[0:3], s32 offset:724 ; 4-byte Folded Spill
+; VI-NEXT: buffer_store_dword v48, off, s[0:3], s32 offset:728 ; 4-byte Folded Spill
+; VI-NEXT: buffer_store_dword v49, off, s[0:3], s32 offset:732 ; 4-byte Folded Spill
+; VI-NEXT: buffer_store_dword v51, off, s[0:3], s32 offset:736 ; 4-byte Folded Spill
+; VI-NEXT: buffer_store_dword v53, off, s[0:3], s32 offset:740 ; 4-byte Folded Spill
+; VI-NEXT: buffer_store_dword v55, off, s[0:3], s32 offset:744 ; 4-byte Folded Spill
+; VI-NEXT: buffer_store_dword v41, off, s[0:3], s32 offset:748 ; 4-byte Folded Spill
+; VI-NEXT: buffer_store_dword v56, off, s[0:3], s32 offset:752 ; 4-byte Folded Spill
+; VI-NEXT: buffer_store_dword v50, off, s[0:3], s32 offset:756 ; 4-byte Folded Spill
+; VI-NEXT: buffer_store_dword v43, off, s[0:3], s32 offset:760 ; 4-byte Folded Spill
+; VI-NEXT: buffer_store_dword v58, off, s[0:3], s32 offset:764 ; 4-byte Folded Spill
+; VI-NEXT: buffer_store_dword v52, off, s[0:3], s32 offset:768 ; 4-byte Folded Spill
+; VI-NEXT: buffer_store_dword v60, off, s[0:3], s32 offset:772 ; 4-byte Folded Spill
+; VI-NEXT: buffer_store_dword v54, off, s[0:3], s32 offset:776 ; 4-byte Folded Spill
+; VI-NEXT: buffer_store_dword v47, off, s[0:3], s32 offset:780 ; 4-byte Folded Spill
+; VI-NEXT: buffer_store_dword v63, off, s[0:3], s32 offset:784 ; 4-byte Folded Spill
+; VI-NEXT: buffer_store_dword v46, off, s[0:3], s32 offset:788 ; 4-byte Folded Spill
+; VI-NEXT: buffer_store_dword v61, off, s[0:3], s32 offset:792 ; 4-byte Folded Spill
+; VI-NEXT: buffer_store_dword v29, off, s[0:3], s32 offset:796 ; 4-byte Folded Spill
+; VI-NEXT: buffer_store_dword v31, off, s[0:3], s32 offset:800 ; 4-byte Folded Spill
+; VI-NEXT: buffer_store_dword v30, off, s[0:3], s32 offset:804 ; 4-byte Folded Spill
+; VI-NEXT: buffer_store_dword v7, off, s[0:3], s32 offset:516 ; 4-byte Folded Spill
+; VI-NEXT: s_waitcnt vmcnt(14)
; VI-NEXT: buffer_store_dword v0, off, s[0:3], s32 offset:524 ; 4-byte Folded Spill
+; VI-NEXT: buffer_store_dword v2, off, s[0:3], s32 offset:528 ; 4-byte Folded Spill
; VI-NEXT: buffer_store_dword v6, off, s[0:3], s32 offset:532 ; 4-byte Folded Spill
-; VI-NEXT: buffer_store_dword v7, off, s[0:3], s32 offset:536 ; 4-byte Folded Spill
-; VI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:544 ; 4-byte Folded Spill
-; VI-NEXT: buffer_store_dword v2, off, s[0:3], s32 offset:548 ; 4-byte Folded Spill
+; VI-NEXT: buffer_store_dword v5, off, s[0:3], s32 offset:536 ; 4-byte Folded Spill
+; VI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:540 ; 4-byte Folded Spill
; VI-NEXT: s_cbranch_scc0 .LBB89_4
; VI-NEXT: ; %bb.1: ; %cmp.false
; VI-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:400 ; 4-byte Folded Reload
@@ -159071,11 +159124,10 @@ define inreg <64 x bfloat> @bitcast_v128i8_to_v64bf16_scalar(<128 x i8> inreg %a
; VI-NEXT: v_or_b32_sdwa v0, v0, v1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
; VI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:404 ; 4-byte Folded Reload
; VI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:468 ; 4-byte Folded Reload
-; VI-NEXT: buffer_store_dword v3, off, s[0:3], s32 offset:800 ; 4-byte Folded Spill
+; VI-NEXT: buffer_store_dword v3, off, s[0:3], s32 offset:808 ; 4-byte Folded Spill
; VI-NEXT: buffer_load_dword v3, off, s[0:3], s32 offset:472 ; 4-byte Folded Reload
-; VI-NEXT: buffer_store_dword v4, off, s[0:3], s32 offset:808 ; 4-byte Folded Spill
-; VI-NEXT: buffer_store_dword v5, off, s[0:3], s32 offset:804 ; 4-byte Folded Spill
-; VI-NEXT: s_waitcnt vmcnt(4)
+; VI-NEXT: buffer_store_dword v4, off, s[0:3], s32 offset:812 ; 4-byte Folded Spill
+; VI-NEXT: s_waitcnt vmcnt(3)
; VI-NEXT: v_or_b32_sdwa v1, v1, v2 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
; VI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:408 ; 4-byte Folded Reload
; VI-NEXT: v_or_b32_sdwa v4, v0, v1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_0 src1_sel:DWORD
@@ -159099,6 +159151,7 @@ define inreg <64 x bfloat> @bitcast_v128i8_to_v64bf16_scalar(<128 x i8> inreg %a
; VI-NEXT: v_or_b32_sdwa v0, v0, v15 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
; VI-NEXT: s_waitcnt vmcnt(0)
; VI-NEXT: v_or_b32_sdwa v1, v1, v17 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; VI-NEXT: v_mov_b32_e32 v17, v10
; VI-NEXT: v_or_b32_sdwa v7, v0, v1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_0 src1_sel:DWORD
; VI-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:432 ; 4-byte Folded Reload
; VI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:436 ; 4-byte Folded Reload
@@ -159115,38 +159168,43 @@ define inreg <64 x bfloat> @bitcast_v128i8_to_v64bf16_scalar(<128 x i8> inreg %a
; VI-NEXT: v_or_b32_sdwa v1, v1, v25 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
; VI-NEXT: v_or_b32_sdwa v9, v0, v1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_0 src1_sel:DWORD
; VI-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:448 ; 4-byte Folded Reload
-; VI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:476 ; 4-byte Folded Reload
-; VI-NEXT: s_waitcnt vmcnt(0)
-; VI-NEXT: v_or_b32_sdwa v0, v0, v1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
; VI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:452 ; 4-byte Folded Reload
-; VI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:480 ; 4-byte Folded Reload
+; VI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:476 ; 4-byte Folded Reload
+; VI-NEXT: s_waitcnt vmcnt(2)
+; VI-NEXT: v_or_b32_sdwa v0, v0, v40 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
; VI-NEXT: s_waitcnt vmcnt(0)
; VI-NEXT: v_or_b32_sdwa v1, v1, v2 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; VI-NEXT: v_mov_b32_e32 v40, v42
+; VI-NEXT: v_mov_b32_e32 v42, v44
+; VI-NEXT: v_mov_b32_e32 v44, v45
+; VI-NEXT: v_mov_b32_e32 v45, v62
+; VI-NEXT: v_or_b32_sdwa v2, v51, v3 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; VI-NEXT: v_or_b32_sdwa v3, v53, v17 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
; VI-NEXT: v_or_b32_sdwa v10, v0, v1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_0 src1_sel:DWORD
; VI-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:456 ; 4-byte Folded Reload
-; VI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:484 ; 4-byte Folded Reload
-; VI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:512 ; 4-byte Folded Reload
+; VI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:480 ; 4-byte Folded Reload
+; VI-NEXT: v_or_b32_sdwa v17, v2, v3 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_0 src1_sel:DWORD
+; VI-NEXT: buffer_load_dword v62, off, s[0:3], s32 offset:548 ; 4-byte Folded Reload
; VI-NEXT: s_waitcnt vmcnt(1)
; VI-NEXT: v_or_b32_sdwa v0, v0, v1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
-; VI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:488 ; 4-byte Folded Reload
-; VI-NEXT: s_waitcnt vmcnt(1)
-; VI-NEXT: v_or_b32_sdwa v2, v50, v2 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; VI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:484 ; 4-byte Folded Reload
; VI-NEXT: s_waitcnt vmcnt(0)
; VI-NEXT: v_or_b32_sdwa v1, v12, v1 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
; VI-NEXT: v_or_b32_sdwa v11, v0, v1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_0 src1_sel:DWORD
-; VI-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:492 ; 4-byte Folded Reload
-; VI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:496 ; 4-byte Folded Reload
+; VI-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:488 ; 4-byte Folded Reload
+; VI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:492 ; 4-byte Folded Reload
; VI-NEXT: s_waitcnt vmcnt(1)
; VI-NEXT: v_or_b32_sdwa v0, v32, v0 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
; VI-NEXT: s_waitcnt vmcnt(0)
; VI-NEXT: v_or_b32_sdwa v1, v33, v1 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
; VI-NEXT: v_or_b32_sdwa v12, v0, v1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_0 src1_sel:DWORD
-; VI-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:500 ; 4-byte Folded Reload
-; VI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:504 ; 4-byte Folded Reload
+; VI-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:496 ; 4-byte Folded Reload
+; VI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:500 ; 4-byte Folded Reload
; VI-NEXT: s_waitcnt vmcnt(1)
; VI-NEXT: v_or_b32_sdwa v0, v34, v0 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
; VI-NEXT: s_waitcnt vmcnt(0)
; VI-NEXT: v_or_b32_sdwa v1, v35, v1 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; VI-NEXT: v_mov_b32_e32 v34, v24
; VI-NEXT: v_or_b32_sdwa v13, v0, v1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_0 src1_sel:DWORD
; VI-NEXT: v_or_b32_sdwa v0, v36, v14 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
; VI-NEXT: v_or_b32_sdwa v1, v37, v16 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
@@ -159154,77 +159212,74 @@ define inreg <64 x bfloat> @bitcast_v128i8_to_v64bf16_scalar(<128 x i8> inreg %a
; VI-NEXT: v_or_b32_sdwa v0, v38, v18 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
; VI-NEXT: v_or_b32_sdwa v1, v39, v20 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
; VI-NEXT: v_or_b32_sdwa v15, v0, v1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_0 src1_sel:DWORD
-; VI-NEXT: v_or_b32_sdwa v0, v48, v45 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
-; VI-NEXT: v_or_b32_sdwa v1, v49, v3 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
-; VI-NEXT: buffer_load_dword v3, off, s[0:3], s32 offset:516 ; 4-byte Folded Reload
-; VI-NEXT: v_mov_b32_e32 v45, v62
+; VI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:508 ; 4-byte Folded Reload
+; VI-NEXT: v_or_b32_sdwa v0, v48, v22 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; VI-NEXT: s_waitcnt vmcnt(0)
+; VI-NEXT: v_or_b32_sdwa v1, v49, v1 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
; VI-NEXT: v_or_b32_sdwa v16, v0, v1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_0 src1_sel:DWORD
-; VI-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:520 ; 4-byte Folded Reload
-; VI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:560 ; 4-byte Folded Reload
-; VI-NEXT: buffer_load_dword v62, off, s[0:3], s32 offset:552 ; 4-byte Folded Reload
-; VI-NEXT: s_waitcnt vmcnt(3)
-; VI-NEXT: v_or_b32_sdwa v3, v51, v3 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
-; VI-NEXT: v_or_b32_sdwa v17, v2, v3 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_0 src1_sel:DWORD
-; VI-NEXT: s_waitcnt vmcnt(2)
-; VI-NEXT: v_or_b32_sdwa v0, v54, v0 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; VI-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:512 ; 4-byte Folded Reload
+; VI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:556 ; 4-byte Folded Reload
; VI-NEXT: s_waitcnt vmcnt(1)
+; VI-NEXT: v_or_b32_sdwa v0, v55, v0 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; VI-NEXT: s_waitcnt vmcnt(0)
; VI-NEXT: v_mov_b32_e32 v32, v1
; VI-NEXT: v_or_b32_sdwa v1, v41, v1 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
-; VI-NEXT: v_mov_b32_e32 v54, v22
-; VI-NEXT: v_mov_b32_e32 v41, v24
; VI-NEXT: v_or_b32_sdwa v18, v0, v1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_0 src1_sel:DWORD
-; VI-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:564 ; 4-byte Folded Reload
-; VI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:592 ; 4-byte Folded Reload
+; VI-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:588 ; 4-byte Folded Reload
+; VI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:584 ; 4-byte Folded Reload
; VI-NEXT: s_waitcnt vmcnt(1)
-; VI-NEXT: v_mov_b32_e32 v34, v0
+; VI-NEXT: v_mov_b32_e32 v33, v0
; VI-NEXT: v_or_b32_sdwa v0, v56, v0 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
; VI-NEXT: s_waitcnt vmcnt(0)
; VI-NEXT: v_mov_b32_e32 v37, v1
-; VI-NEXT: v_or_b32_sdwa v1, v55, v1 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
-; VI-NEXT: v_mov_b32_e32 v55, v26
+; VI-NEXT: v_or_b32_sdwa v1, v50, v1 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; VI-NEXT: v_mov_b32_e32 v50, v26
; VI-NEXT: v_or_b32_sdwa v19, v0, v1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_0 src1_sel:DWORD
-; VI-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:588 ; 4-byte Folded Reload
-; VI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:584 ; 4-byte Folded Reload
+; VI-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:576 ; 4-byte Folded Reload
+; VI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:572 ; 4-byte Folded Reload
; VI-NEXT: s_waitcnt vmcnt(1)
; VI-NEXT: v_mov_b32_e32 v39, v0
-; VI-NEXT: v_or_b32_sdwa v0, v52, v0 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; VI-NEXT: v_or_b32_sdwa v0, v43, v0 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
; VI-NEXT: s_waitcnt vmcnt(0)
; VI-NEXT: v_mov_b32_e32 v49, v1
-; VI-NEXT: v_or_b32_sdwa v1, v43, v1 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
-; VI-NEXT: v_mov_b32_e32 v43, v27
+; VI-NEXT: v_or_b32_sdwa v1, v58, v1 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
; VI-NEXT: v_or_b32_sdwa v20, v0, v1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_0 src1_sel:DWORD
-; VI-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:576 ; 4-byte Folded Reload
-; VI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:572 ; 4-byte Folded Reload
+; VI-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:564 ; 4-byte Folded Reload
+; VI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:560 ; 4-byte Folded Reload
; VI-NEXT: s_waitcnt vmcnt(1)
; VI-NEXT: v_mov_b32_e32 v51, v0
-; VI-NEXT: v_or_b32_sdwa v0, v53, v0 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; VI-NEXT: v_or_b32_sdwa v0, v52, v0 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
; VI-NEXT: s_waitcnt vmcnt(0)
-; VI-NEXT: v_mov_b32_e32 v35, v1
-; VI-NEXT: v_or_b32_sdwa v1, v46, v1 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
-; VI-NEXT: v_mov_b32_e32 v53, v28
+; VI-NEXT: v_mov_b32_e32 v53, v1
+; VI-NEXT: v_or_b32_sdwa v1, v60, v1 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; VI-NEXT: v_mov_b32_e32 v52, v28
; VI-NEXT: v_or_b32_sdwa v21, v0, v1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_0 src1_sel:DWORD
-; VI-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:580 ; 4-byte Folded Reload
-; VI-NEXT: v_or_b32_sdwa v1, v47, v22 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
-; VI-NEXT: s_waitcnt vmcnt(0)
-; VI-NEXT: v_mov_b32_e32 v33, v0
-; VI-NEXT: v_or_b32_sdwa v0, v58, v0 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
-; VI-NEXT: v_or_b32_sdwa v22, v0, v1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_0 src1_sel:DWORD
; VI-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:568 ; 4-byte Folded Reload
-; VI-NEXT: v_or_b32_sdwa v1, v57, v24 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; VI-NEXT: v_or_b32_sdwa v1, v47, v24 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
; VI-NEXT: s_waitcnt vmcnt(0)
-; VI-NEXT: v_mov_b32_e32 v36, v0
+; VI-NEXT: v_mov_b32_e32 v55, v0
+; VI-NEXT: v_or_b32_sdwa v0, v54, v0 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; VI-NEXT: v_or_b32_sdwa v22, v0, v1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_0 src1_sel:DWORD
+; VI-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:592 ; 4-byte Folded Reload
+; VI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:580 ; 4-byte Folded Reload
+; VI-NEXT: s_waitcnt vmcnt(1)
+; VI-NEXT: v_mov_b32_e32 v35, v0
; VI-NEXT: v_or_b32_sdwa v0, v63, v0 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; VI-NEXT: s_waitcnt vmcnt(0)
+; VI-NEXT: v_mov_b32_e32 v41, v1
+; VI-NEXT: v_or_b32_sdwa v1, v46, v1 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; VI-NEXT: v_mov_b32_e32 v63, v27
+; VI-NEXT: v_mov_b32_e32 v46, v57
; VI-NEXT: v_or_b32_sdwa v23, v0, v1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_0 src1_sel:DWORD
; VI-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:600 ; 4-byte Folded Reload
; VI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:596 ; 4-byte Folded Reload
; VI-NEXT: s_waitcnt vmcnt(1)
-; VI-NEXT: v_mov_b32_e32 v56, v0
+; VI-NEXT: v_mov_b32_e32 v36, v0
; VI-NEXT: v_or_b32_sdwa v0, v61, v0 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
; VI-NEXT: s_waitcnt vmcnt(0)
-; VI-NEXT: v_mov_b32_e32 v58, v1
-; VI-NEXT: v_or_b32_sdwa v1, v60, v1 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
-; VI-NEXT: v_mov_b32_e32 v61, v60
-; VI-NEXT: v_mov_b32_e32 v60, v59
+; VI-NEXT: v_mov_b32_e32 v56, v1
+; VI-NEXT: v_or_b32_sdwa v1, v57, v1 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; VI-NEXT: v_mov_b32_e32 v61, v59
; VI-NEXT: v_or_b32_sdwa v24, v0, v1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_0 src1_sel:DWORD
; VI-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:608 ; 4-byte Folded Reload
; VI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:612 ; 4-byte Folded Reload
@@ -159236,55 +159291,53 @@ define inreg <64 x bfloat> @bitcast_v128i8_to_v64bf16_scalar(<128 x i8> inreg %a
; VI-NEXT: v_or_b32_sdwa v1, v40, v1 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
; VI-NEXT: v_or_b32_sdwa v25, v0, v1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_0 src1_sel:DWORD
; VI-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:604 ; 4-byte Folded Reload
-; VI-NEXT: buffer_load_dword v59, off, s[0:3], s32 offset:556 ; 4-byte Folded Reload
-; VI-NEXT: v_or_b32_sdwa v1, v45, v26 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; VI-NEXT: buffer_load_dword v59, off, s[0:3], s32 offset:552 ; 4-byte Folded Reload
+; VI-NEXT: v_or_b32_sdwa v1, v44, v26 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
; VI-NEXT: s_waitcnt vmcnt(1)
-; VI-NEXT: v_mov_b32_e32 v50, v0
+; VI-NEXT: v_mov_b32_e32 v58, v0
; VI-NEXT: v_or_b32_sdwa v0, v42, v0 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
; VI-NEXT: v_or_b32_sdwa v26, v0, v1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_0 src1_sel:DWORD
; VI-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:616 ; 4-byte Folded Reload
; VI-NEXT: v_or_b32_sdwa v1, v62, v27 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
; VI-NEXT: s_waitcnt vmcnt(0)
-; VI-NEXT: v_mov_b32_e32 v52, v0
-; VI-NEXT: v_or_b32_sdwa v0, v44, v0 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; VI-NEXT: v_mov_b32_e32 v43, v0
+; VI-NEXT: v_or_b32_sdwa v0, v45, v0 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
; VI-NEXT: v_or_b32_sdwa v27, v0, v1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_0 src1_sel:DWORD
; VI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:620 ; 4-byte Folded Reload
-; VI-NEXT: v_or_b32_sdwa v0, v29, v28 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; VI-NEXT: v_or_b32_sdwa v0, v59, v28 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
; VI-NEXT: s_waitcnt vmcnt(0)
-; VI-NEXT: v_mov_b32_e32 v46, v1
-; VI-NEXT: v_or_b32_sdwa v1, v59, v1 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; VI-NEXT: v_mov_b32_e32 v60, v1
+; VI-NEXT: v_or_b32_sdwa v1, v29, v1 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
; VI-NEXT: v_or_b32_sdwa v28, v0, v1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_0 src1_sel:DWORD
-; VI-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:800 ; 4-byte Folded Reload
-; VI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:808 ; 4-byte Folded Reload
+; VI-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:808 ; 4-byte Folded Reload
+; VI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:812 ; 4-byte Folded Reload
; VI-NEXT: s_waitcnt vmcnt(1)
-; VI-NEXT: v_mov_b32_e32 v63, v0
-; VI-NEXT: v_or_b32_sdwa v0, v30, v0 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; VI-NEXT: v_mov_b32_e32 v54, v0
+; VI-NEXT: v_or_b32_sdwa v0, v31, v0 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
; VI-NEXT: s_waitcnt vmcnt(0)
; VI-NEXT: v_mov_b32_e32 v47, v1
-; VI-NEXT: v_or_b32_sdwa v1, v31, v1 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; VI-NEXT: v_or_b32_sdwa v1, v30, v1 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
; VI-NEXT: v_or_b32_sdwa v29, v0, v1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_0 src1_sel:DWORD
-; VI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:804 ; 4-byte Folded Reload
-; VI-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:524 ; 4-byte Folded Reload
-; VI-NEXT: s_waitcnt vmcnt(1)
-; VI-NEXT: v_mov_b32_e32 v57, v1
-; VI-NEXT: s_waitcnt vmcnt(0)
-; VI-NEXT: v_or_b32_sdwa v0, v0, v1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
-; VI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:528 ; 4-byte Folded Reload
+; VI-NEXT: buffer_load_dword v57, off, s[0:3], s32 offset:544 ; 4-byte Folded Reload
+; VI-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:516 ; 4-byte Folded Reload
+; VI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:520 ; 4-byte Folded Reload
; VI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:532 ; 4-byte Folded Reload
+; VI-NEXT: s_waitcnt vmcnt(2)
+; VI-NEXT: v_or_b32_sdwa v0, v0, v57 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
; VI-NEXT: s_waitcnt vmcnt(0)
; VI-NEXT: v_or_b32_sdwa v1, v1, v2 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
; VI-NEXT: v_or_b32_sdwa v30, v0, v1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_0 src1_sel:DWORD
-; VI-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:544 ; 4-byte Folded Reload
+; VI-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:524 ; 4-byte Folded Reload
; VI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:536 ; 4-byte Folded Reload
; VI-NEXT: s_waitcnt vmcnt(0)
-; VI-NEXT: v_or_b32_sdwa v0, v1, v0 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
-; VI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:548 ; 4-byte Folded Reload
-; VI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:540 ; 4-byte Folded Reload
+; VI-NEXT: v_or_b32_sdwa v0, v0, v1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; VI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:540 ; 4-byte Folded Reload
+; VI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:528 ; 4-byte Folded Reload
; VI-NEXT: s_waitcnt vmcnt(0)
; VI-NEXT: v_or_b32_sdwa v1, v2, v1 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
; VI-NEXT: v_or_b32_sdwa v31, v0, v1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_0 src1_sel:DWORD
; VI-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:460 ; 4-byte Folded Reload
-; VI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:508 ; 4-byte Folded Reload
+; VI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:504 ; 4-byte Folded Reload
; VI-NEXT: s_waitcnt vmcnt(0)
; VI-NEXT: v_or_b32_sdwa v0, v0, v1 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
; VI-NEXT: v_or_b32_e32 v3, s4, v0
@@ -159316,12 +159369,10 @@ define inreg <64 x bfloat> @bitcast_v128i8_to_v64bf16_scalar(<128 x i8> inreg %a
; VI-NEXT: v_mov_b32_e32 v2, s6
; VI-NEXT: s_cbranch_execnz .LBB89_3
; VI-NEXT: .LBB89_2: ; %cmp.true
-; VI-NEXT: s_waitcnt vmcnt(0)
-; VI-NEXT: v_add_u32_e32 v3, vcc, 3, v59
-; VI-NEXT: v_or_b32_sdwa v29, v46, v3 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:BYTE_0
-; VI-NEXT: buffer_load_dword v3, off, s[0:3], s32 offset:788 ; 4-byte Folded Reload
-; VI-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:540 ; 4-byte Folded Reload
-; VI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:548 ; 4-byte Folded Reload
+; VI-NEXT: buffer_load_dword v3, off, s[0:3], s32 offset:796 ; 4-byte Folded Reload
+; VI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:540 ; 4-byte Folded Reload
+; VI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:532 ; 4-byte Folded Reload
+; VI-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:528 ; 4-byte Folded Reload
; VI-NEXT: s_add_i32 s28, s28, 3
; VI-NEXT: s_and_b32 s4, s28, 0xff
; VI-NEXT: s_lshl_b32 s5, s29, 8
@@ -159340,165 +159391,147 @@ define inreg <64 x bfloat> @bitcast_v128i8_to_v64bf16_scalar(<128 x i8> inreg %a
; VI-NEXT: s_lshl_b32 s9, s19, 8
; VI-NEXT: s_add_i32 s16, s16, 3
; VI-NEXT: s_lshl_b32 s10, s17, 8
-; VI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:532 ; 4-byte Folded Reload
-; VI-NEXT: s_waitcnt vmcnt(3)
+; VI-NEXT: buffer_load_dword v4, off, s[0:3], s32 offset:512 ; 4-byte Folded Reload
+; VI-NEXT: s_waitcnt vmcnt(4)
; VI-NEXT: v_add_u32_e32 v3, vcc, 3, v3
-; VI-NEXT: v_or_b32_sdwa v26, v53, v3 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:BYTE_0
+; VI-NEXT: v_or_b32_sdwa v28, v60, v3 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:BYTE_0
+; VI-NEXT: v_add_u32_e32 v3, vcc, 3, v59
+; VI-NEXT: v_or_b32_sdwa v25, v52, v3 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:BYTE_0
; VI-NEXT: v_add_u32_e32 v3, vcc, 3, v62
-; VI-NEXT: v_or_b32_sdwa v28, v43, v3 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:BYTE_0
-; VI-NEXT: v_add_u32_e32 v3, vcc, 3, v44
-; VI-NEXT: v_or_b32_sdwa v53, v52, v3 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:BYTE_0
+; VI-NEXT: v_or_b32_sdwa v27, v63, v3 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:BYTE_0
; VI-NEXT: v_add_u32_e32 v3, vcc, 3, v45
-; VI-NEXT: v_or_b32_sdwa v27, v55, v3 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:BYTE_0
+; VI-NEXT: v_or_b32_sdwa v52, v43, v3 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:BYTE_0
+; VI-NEXT: v_add_u32_e32 v3, vcc, 3, v44
+; VI-NEXT: v_or_b32_sdwa v26, v50, v3 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:BYTE_0
; VI-NEXT: v_add_u32_e32 v3, vcc, 3, v42
-; VI-NEXT: v_or_b32_sdwa v52, v50, v3 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:BYTE_0
+; VI-NEXT: v_or_b32_sdwa v63, v58, v3 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:BYTE_0
; VI-NEXT: v_add_u32_e32 v3, vcc, 3, v40
-; VI-NEXT: v_or_b32_sdwa v25, v48, v3 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:BYTE_0
-; VI-NEXT: v_add_u32_e32 v3, vcc, 3, v60
-; VI-NEXT: v_or_b32_sdwa v59, v38, v3 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:BYTE_0
+; VI-NEXT: v_or_b32_sdwa v43, v48, v3 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:BYTE_0
; VI-NEXT: v_add_u32_e32 v3, vcc, 3, v61
-; VI-NEXT: v_or_b32_sdwa v24, v58, v3 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:BYTE_0
-; VI-NEXT: buffer_load_dword v3, off, s[0:3], s32 offset:784 ; 4-byte Folded Reload
+; VI-NEXT: v_or_b32_sdwa v59, v38, v3 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:BYTE_0
+; VI-NEXT: v_add_u32_e32 v3, vcc, 3, v46
+; VI-NEXT: v_or_b32_sdwa v24, v56, v3 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:BYTE_0
+; VI-NEXT: buffer_load_dword v3, off, s[0:3], s32 offset:792 ; 4-byte Folded Reload
; VI-NEXT: s_waitcnt vmcnt(0)
; VI-NEXT: v_add_u32_e32 v3, vcc, 3, v3
-; VI-NEXT: v_or_b32_sdwa v48, v56, v3 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:BYTE_0
-; VI-NEXT: buffer_load_dword v3, off, s[0:3], s32 offset:780 ; 4-byte Folded Reload
+; VI-NEXT: v_or_b32_sdwa v48, v36, v3 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:BYTE_0
+; VI-NEXT: buffer_load_dword v3, off, s[0:3], s32 offset:788 ; 4-byte Folded Reload
; VI-NEXT: v_add_u32_e32 v48, vcc, 0x300, v48
; VI-NEXT: v_or_b32_sdwa v24, v24, v48 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:WORD_0
; VI-NEXT: v_add_u32_e32 v24, vcc, 0x3000000, v24
; VI-NEXT: s_waitcnt vmcnt(0)
; VI-NEXT: v_add_u32_e32 v3, vcc, 3, v3
; VI-NEXT: v_or_b32_sdwa v23, v41, v3 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:BYTE_0
-; VI-NEXT: buffer_load_dword v3, off, s[0:3], s32 offset:776 ; 4-byte Folded Reload
+; VI-NEXT: buffer_load_dword v3, off, s[0:3], s32 offset:784 ; 4-byte Folded Reload
; VI-NEXT: s_waitcnt vmcnt(0)
; VI-NEXT: v_add_u32_e32 v3, vcc, 3, v3
-; VI-NEXT: v_or_b32_sdwa v38, v36, v3 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:BYTE_0
-; VI-NEXT: buffer_load_dword v3, off, s[0:3], s32 offset:772 ; 4-byte Folded Reload
+; VI-NEXT: v_or_b32_sdwa v38, v35, v3 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:BYTE_0
+; VI-NEXT: buffer_load_dword v3, off, s[0:3], s32 offset:780 ; 4-byte Folded Reload
; VI-NEXT: v_add_u32_e32 v38, vcc, 0x300, v38
; VI-NEXT: v_or_b32_sdwa v23, v23, v38 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:WORD_0
; VI-NEXT: v_add_u32_e32 v23, vcc, 0x3000000, v23
; VI-NEXT: s_waitcnt vmcnt(0)
; VI-NEXT: v_add_u32_e32 v3, vcc, 3, v3
-; VI-NEXT: v_or_b32_sdwa v22, v54, v3 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:BYTE_0
-; VI-NEXT: buffer_load_dword v3, off, s[0:3], s32 offset:768 ; 4-byte Folded Reload
+; VI-NEXT: v_or_b32_sdwa v22, v34, v3 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:BYTE_0
+; VI-NEXT: buffer_load_dword v3, off, s[0:3], s32 offset:776 ; 4-byte Folded Reload
; VI-NEXT: s_waitcnt vmcnt(0)
; VI-NEXT: v_add_u32_e32 v3, vcc, 3, v3
-; VI-NEXT: v_or_b32_sdwa v50, v33, v3 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:BYTE_0
-; VI-NEXT: buffer_load_dword v3, off, s[0:3], s32 offset:764 ; 4-byte Folded Reload
-; VI-NEXT: v_add_u32_e32 v50, vcc, 0x300, v50
+; VI-NEXT: v_or_b32_sdwa v36, v55, v3 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:BYTE_0
+; VI-NEXT: buffer_load_dword v3, off, s[0:3], s32 offset:772 ; 4-byte Folded Reload
+; VI-NEXT: v_add_u32_e32 v50, vcc, 0x300, v36
; VI-NEXT: v_or_b32_sdwa v22, v22, v50 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:WORD_0
; VI-NEXT: v_add_u32_e32 v22, vcc, 0x3000000, v22
; VI-NEXT: s_waitcnt vmcnt(0)
; VI-NEXT: v_add_u32_e32 v3, vcc, 3, v3
-; VI-NEXT: v_or_b32_sdwa v21, v35, v3 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:BYTE_0
-; VI-NEXT: buffer_load_dword v3, off, s[0:3], s32 offset:760 ; 4-byte Folded Reload
+; VI-NEXT: v_or_b32_sdwa v21, v53, v3 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:BYTE_0
+; VI-NEXT: buffer_load_dword v3, off, s[0:3], s32 offset:768 ; 4-byte Folded Reload
; VI-NEXT: s_waitcnt vmcnt(0)
; VI-NEXT: v_add_u32_e32 v3, vcc, 3, v3
-; VI-NEXT: v_or_b32_sdwa v54, v51, v3 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:BYTE_0
-; VI-NEXT: buffer_load_dword v3, off, s[0:3], s32 offset:756 ; 4-byte Folded Reload
+; VI-NEXT: v_or_b32_sdwa v53, v51, v3 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:BYTE_0
+; VI-NEXT: buffer_load_dword v3, off, s[0:3], s32 offset:764 ; 4-byte Folded Reload
; VI-NEXT: s_waitcnt vmcnt(0)
; VI-NEXT: v_add_u32_e32 v3, vcc, 3, v3
; VI-NEXT: v_or_b32_sdwa v20, v49, v3 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:BYTE_0
-; VI-NEXT: buffer_load_dword v3, off, s[0:3], s32 offset:752 ; 4-byte Folded Reload
+; VI-NEXT: buffer_load_dword v3, off, s[0:3], s32 offset:760 ; 4-byte Folded Reload
; VI-NEXT: s_waitcnt vmcnt(0)
; VI-NEXT: v_add_u32_e32 v3, vcc, 3, v3
; VI-NEXT: v_or_b32_sdwa v49, v39, v3 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:BYTE_0
-; VI-NEXT: buffer_load_dword v3, off, s[0:3], s32 offset:748 ; 4-byte Folded Reload
-; VI-NEXT: v_add_u32_e32 v0, vcc, 3, v0
-; VI-NEXT: v_or_b32_sdwa v0, v1, v0 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:BYTE_0
-; VI-NEXT: buffer_store_dword v0, off, s[0:3], s32 offset:540 ; 4-byte Folded Spill
-; VI-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:536 ; 4-byte Folded Reload
-; VI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:544 ; 4-byte Folded Reload
+; VI-NEXT: buffer_load_dword v3, off, s[0:3], s32 offset:756 ; 4-byte Folded Reload
; VI-NEXT: v_add_u32_e32 v49, vcc, 0x300, v49
; VI-NEXT: v_or_b32_sdwa v20, v20, v49 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:WORD_0
; VI-NEXT: v_add_u32_e32 v20, vcc, 0x3000000, v20
-; VI-NEXT: s_waitcnt vmcnt(3)
+; VI-NEXT: s_waitcnt vmcnt(0)
; VI-NEXT: v_add_u32_e32 v3, vcc, 3, v3
; VI-NEXT: v_or_b32_sdwa v19, v37, v3 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:BYTE_0
-; VI-NEXT: buffer_load_dword v3, off, s[0:3], s32 offset:744 ; 4-byte Folded Reload
-; VI-NEXT: s_waitcnt vmcnt(2)
-; VI-NEXT: v_add_u32_e32 v0, vcc, 3, v0
-; VI-NEXT: s_waitcnt vmcnt(1)
-; VI-NEXT: v_or_b32_sdwa v0, v1, v0 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:BYTE_0
-; VI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:528 ; 4-byte Folded Reload
-; VI-NEXT: s_waitcnt vmcnt(1)
+; VI-NEXT: buffer_load_dword v3, off, s[0:3], s32 offset:752 ; 4-byte Folded Reload
+; VI-NEXT: s_waitcnt vmcnt(0)
; VI-NEXT: v_add_u32_e32 v3, vcc, 3, v3
-; VI-NEXT: v_or_b32_sdwa v37, v34, v3 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:BYTE_0
-; VI-NEXT: buffer_load_dword v3, off, s[0:3], s32 offset:740 ; 4-byte Folded Reload
+; VI-NEXT: v_or_b32_sdwa v37, v33, v3 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:BYTE_0
+; VI-NEXT: buffer_load_dword v3, off, s[0:3], s32 offset:748 ; 4-byte Folded Reload
; VI-NEXT: v_add_u32_e32 v37, vcc, 0x300, v37
-; VI-NEXT: s_waitcnt vmcnt(1)
-; VI-NEXT: v_add_u32_e32 v1, vcc, 3, v1
-; VI-NEXT: v_or_b32_sdwa v31, v2, v1 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:BYTE_0
-; VI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:524 ; 4-byte Folded Reload
-; VI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:796 ; 4-byte Folded Reload
; VI-NEXT: v_or_b32_sdwa v19, v19, v37 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:WORD_0
; VI-NEXT: v_add_u32_e32 v19, vcc, 0x3000000, v19
-; VI-NEXT: s_waitcnt vmcnt(2)
+; VI-NEXT: s_waitcnt vmcnt(0)
; VI-NEXT: v_add_u32_e32 v3, vcc, 3, v3
; VI-NEXT: v_or_b32_sdwa v18, v32, v3 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:BYTE_0
-; VI-NEXT: buffer_load_dword v3, off, s[0:3], s32 offset:736 ; 4-byte Folded Reload
-; VI-NEXT: buffer_load_dword v4, off, s[0:3], s32 offset:520 ; 4-byte Folded Reload
+; VI-NEXT: buffer_load_dword v3, off, s[0:3], s32 offset:744 ; 4-byte Folded Reload
; VI-NEXT: buffer_load_dword v32, off, s[0:3], s32 offset:460 ; 4-byte Folded Reload
-; VI-NEXT: s_waitcnt vmcnt(4)
-; VI-NEXT: v_add_u32_e32 v1, vcc, 3, v1
-; VI-NEXT: v_or_b32_sdwa v1, v57, v1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:BYTE_0
-; VI-NEXT: s_waitcnt vmcnt(2)
-; VI-NEXT: v_add_u32_e32 v3, vcc, 3, v3
; VI-NEXT: s_waitcnt vmcnt(1)
-; VI-NEXT: v_or_b32_sdwa v57, v4, v3 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:BYTE_0
-; VI-NEXT: buffer_load_dword v3, off, s[0:3], s32 offset:732 ; 4-byte Folded Reload
-; VI-NEXT: buffer_load_dword v4, off, s[0:3], s32 offset:516 ; 4-byte Folded Reload
-; VI-NEXT: v_add_u32_e32 v35, vcc, 0x300, v57
-; VI-NEXT: v_or_b32_sdwa v18, v18, v35 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:WORD_0
+; VI-NEXT: v_add_u32_e32 v3, vcc, 3, v3
+; VI-NEXT: v_or_b32_sdwa v58, v4, v3 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:BYTE_0
+; VI-NEXT: buffer_load_dword v3, off, s[0:3], s32 offset:740 ; 4-byte Folded Reload
+; VI-NEXT: buffer_load_dword v4, off, s[0:3], s32 offset:724 ; 4-byte Folded Reload
+; VI-NEXT: v_add_u32_e32 v36, vcc, 0x300, v58
+; VI-NEXT: v_or_b32_sdwa v18, v18, v36 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:WORD_0
; VI-NEXT: v_add_u32_e32 v18, vcc, 0x3000000, v18
; VI-NEXT: s_waitcnt vmcnt(1)
; VI-NEXT: v_add_u32_e32 v3, vcc, 3, v3
; VI-NEXT: s_waitcnt vmcnt(0)
; VI-NEXT: v_or_b32_sdwa v16, v4, v3 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:BYTE_0
-; VI-NEXT: buffer_load_dword v3, off, s[0:3], s32 offset:728 ; 4-byte Folded Reload
-; VI-NEXT: buffer_load_dword v4, off, s[0:3], s32 offset:512 ; 4-byte Folded Reload
+; VI-NEXT: buffer_load_dword v3, off, s[0:3], s32 offset:736 ; 4-byte Folded Reload
+; VI-NEXT: buffer_load_dword v4, off, s[0:3], s32 offset:704 ; 4-byte Folded Reload
; VI-NEXT: s_waitcnt vmcnt(1)
; VI-NEXT: v_add_u32_e32 v3, vcc, 3, v3
; VI-NEXT: s_waitcnt vmcnt(0)
; VI-NEXT: v_or_b32_sdwa v10, v4, v3 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:BYTE_0
-; VI-NEXT: buffer_load_dword v3, off, s[0:3], s32 offset:724 ; 4-byte Folded Reload
-; VI-NEXT: buffer_load_dword v4, off, s[0:3], s32 offset:680 ; 4-byte Folded Reload
+; VI-NEXT: buffer_load_dword v3, off, s[0:3], s32 offset:732 ; 4-byte Folded Reload
+; VI-NEXT: buffer_load_dword v4, off, s[0:3], s32 offset:508 ; 4-byte Folded Reload
; VI-NEXT: s_waitcnt vmcnt(1)
; VI-NEXT: v_add_u32_e32 v3, vcc, 3, v3
; VI-NEXT: s_waitcnt vmcnt(0)
; VI-NEXT: v_or_b32_sdwa v17, v4, v3 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:BYTE_0
+; VI-NEXT: buffer_load_dword v3, off, s[0:3], s32 offset:728 ; 4-byte Folded Reload
+; VI-NEXT: buffer_load_dword v4, off, s[0:3], s32 offset:680 ; 4-byte Folded Reload
+; VI-NEXT: s_waitcnt vmcnt(1)
+; VI-NEXT: v_add_u32_e32 v3, vcc, 3, v3
+; VI-NEXT: s_waitcnt vmcnt(0)
+; VI-NEXT: v_or_b32_sdwa v11, v4, v3 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:BYTE_0
; VI-NEXT: buffer_load_dword v3, off, s[0:3], s32 offset:720 ; 4-byte Folded Reload
; VI-NEXT: buffer_load_dword v4, off, s[0:3], s32 offset:676 ; 4-byte Folded Reload
; VI-NEXT: s_waitcnt vmcnt(1)
; VI-NEXT: v_add_u32_e32 v3, vcc, 3, v3
; VI-NEXT: s_waitcnt vmcnt(0)
-; VI-NEXT: v_or_b32_sdwa v11, v4, v3 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:BYTE_0
+; VI-NEXT: v_or_b32_sdwa v15, v4, v3 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:BYTE_0
; VI-NEXT: buffer_load_dword v3, off, s[0:3], s32 offset:716 ; 4-byte Folded Reload
; VI-NEXT: buffer_load_dword v4, off, s[0:3], s32 offset:672 ; 4-byte Folded Reload
; VI-NEXT: s_waitcnt vmcnt(1)
; VI-NEXT: v_add_u32_e32 v3, vcc, 3, v3
; VI-NEXT: s_waitcnt vmcnt(0)
-; VI-NEXT: v_or_b32_sdwa v15, v4, v3 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:BYTE_0
+; VI-NEXT: v_or_b32_sdwa v56, v4, v3 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:BYTE_0
; VI-NEXT: buffer_load_dword v3, off, s[0:3], s32 offset:712 ; 4-byte Folded Reload
; VI-NEXT: buffer_load_dword v4, off, s[0:3], s32 offset:668 ; 4-byte Folded Reload
; VI-NEXT: s_waitcnt vmcnt(1)
; VI-NEXT: v_add_u32_e32 v3, vcc, 3, v3
; VI-NEXT: s_waitcnt vmcnt(0)
-; VI-NEXT: v_or_b32_sdwa v56, v4, v3 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:BYTE_0
+; VI-NEXT: v_or_b32_sdwa v14, v4, v3 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:BYTE_0
; VI-NEXT: buffer_load_dword v3, off, s[0:3], s32 offset:708 ; 4-byte Folded Reload
; VI-NEXT: buffer_load_dword v4, off, s[0:3], s32 offset:664 ; 4-byte Folded Reload
; VI-NEXT: s_waitcnt vmcnt(1)
; VI-NEXT: v_add_u32_e32 v3, vcc, 3, v3
; VI-NEXT: s_waitcnt vmcnt(0)
-; VI-NEXT: v_or_b32_sdwa v14, v4, v3 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:BYTE_0
-; VI-NEXT: buffer_load_dword v3, off, s[0:3], s32 offset:704 ; 4-byte Folded Reload
-; VI-NEXT: buffer_load_dword v4, off, s[0:3], s32 offset:660 ; 4-byte Folded Reload
-; VI-NEXT: s_waitcnt vmcnt(1)
-; VI-NEXT: v_add_u32_e32 v3, vcc, 3, v3
-; VI-NEXT: s_waitcnt vmcnt(0)
; VI-NEXT: v_or_b32_sdwa v34, v4, v3 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:BYTE_0
; VI-NEXT: buffer_load_dword v3, off, s[0:3], s32 offset:700 ; 4-byte Folded Reload
-; VI-NEXT: buffer_load_dword v4, off, s[0:3], s32 offset:504 ; 4-byte Folded Reload
+; VI-NEXT: buffer_load_dword v4, off, s[0:3], s32 offset:500 ; 4-byte Folded Reload
; VI-NEXT: v_add_u32_e32 v34, vcc, 0x300, v34
; VI-NEXT: v_or_b32_sdwa v14, v14, v34 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:WORD_0
; VI-NEXT: v_add_u32_e32 v14, vcc, 0x3000000, v14
@@ -159507,67 +159540,78 @@ define inreg <64 x bfloat> @bitcast_v128i8_to_v64bf16_scalar(<128 x i8> inreg %a
; VI-NEXT: s_waitcnt vmcnt(0)
; VI-NEXT: v_or_b32_sdwa v13, v4, v3 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:BYTE_0
; VI-NEXT: buffer_load_dword v3, off, s[0:3], s32 offset:696 ; 4-byte Folded Reload
-; VI-NEXT: buffer_load_dword v4, off, s[0:3], s32 offset:500 ; 4-byte Folded Reload
-; VI-NEXT: s_waitcnt vmcnt(1)
+; VI-NEXT: buffer_load_dword v4, off, s[0:3], s32 offset:496 ; 4-byte Folded Reload
+; VI-NEXT: v_add_u32_e32 v0, vcc, 3, v0
+; VI-NEXT: v_or_b32_sdwa v31, v1, v0 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:BYTE_0
+; VI-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:524 ; 4-byte Folded Reload
+; VI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:536 ; 4-byte Folded Reload
+; VI-NEXT: s_waitcnt vmcnt(3)
; VI-NEXT: v_add_u32_e32 v3, vcc, 3, v3
-; VI-NEXT: s_waitcnt vmcnt(0)
-; VI-NEXT: v_or_b32_sdwa v36, v4, v3 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:BYTE_0
+; VI-NEXT: s_waitcnt vmcnt(2)
+; VI-NEXT: v_or_b32_sdwa v35, v4, v3 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:BYTE_0
; VI-NEXT: buffer_load_dword v3, off, s[0:3], s32 offset:692 ; 4-byte Folded Reload
-; VI-NEXT: buffer_load_dword v4, off, s[0:3], s32 offset:496 ; 4-byte Folded Reload
-; VI-NEXT: v_add_u32_e32 v36, vcc, 0x300, v36
-; VI-NEXT: v_or_b32_sdwa v13, v13, v36 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:WORD_0
-; VI-NEXT: v_add_u32_e32 v36, vcc, 0x300, v26
-; VI-NEXT: v_add_u32_e32 v26, vcc, 0x300, v52
-; VI-NEXT: v_or_b32_sdwa v26, v27, v26 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:WORD_0
-; VI-NEXT: v_add_u32_e32 v52, vcc, 0x300, v54
-; VI-NEXT: v_or_b32_sdwa v21, v21, v52 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:WORD_0
+; VI-NEXT: buffer_load_dword v4, off, s[0:3], s32 offset:492 ; 4-byte Folded Reload
+; VI-NEXT: s_waitcnt vmcnt(3)
+; VI-NEXT: v_add_u32_e32 v0, vcc, 3, v0
+; VI-NEXT: s_waitcnt vmcnt(2)
+; VI-NEXT: v_or_b32_sdwa v0, v1, v0 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:BYTE_0
+; VI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:520 ; 4-byte Folded Reload
+; VI-NEXT: v_add_u32_e32 v35, vcc, 0x300, v35
+; VI-NEXT: v_or_b32_sdwa v13, v13, v35 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:WORD_0
+; VI-NEXT: v_add_u32_e32 v35, vcc, 0x300, v25
+; VI-NEXT: v_add_u32_e32 v25, vcc, 0x300, v59
+; VI-NEXT: v_or_b32_sdwa v25, v43, v25 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:WORD_0
+; VI-NEXT: v_or_b32_sdwa v28, v28, v35 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:WORD_0
; VI-NEXT: v_add_u32_e32 v13, vcc, 0x3000000, v13
-; VI-NEXT: v_add_u32_e32 v21, vcc, 0x3000000, v21
-; VI-NEXT: v_add_u32_e32 v26, vcc, 0x3000000, v26
-; VI-NEXT: s_waitcnt vmcnt(1)
+; VI-NEXT: v_add_u32_e32 v25, vcc, 0x3000000, v25
+; VI-NEXT: v_add_u32_e32 v28, vcc, 0x3000000, v28
+; VI-NEXT: s_waitcnt vmcnt(2)
; VI-NEXT: v_add_u32_e32 v3, vcc, 3, v3
-; VI-NEXT: s_waitcnt vmcnt(0)
+; VI-NEXT: s_waitcnt vmcnt(1)
; VI-NEXT: v_or_b32_sdwa v12, v4, v3 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:BYTE_0
; VI-NEXT: buffer_load_dword v3, off, s[0:3], s32 offset:688 ; 4-byte Folded Reload
-; VI-NEXT: buffer_load_dword v4, off, s[0:3], s32 offset:492 ; 4-byte Folded Reload
-; VI-NEXT: s_waitcnt vmcnt(1)
+; VI-NEXT: buffer_load_dword v4, off, s[0:3], s32 offset:488 ; 4-byte Folded Reload
+; VI-NEXT: s_waitcnt vmcnt(2)
+; VI-NEXT: v_add_u32_e32 v1, vcc, 3, v1
+; VI-NEXT: v_or_b32_sdwa v30, v2, v1 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:BYTE_0
+; VI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:516 ; 4-byte Folded Reload
+; VI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:804 ; 4-byte Folded Reload
+; VI-NEXT: s_waitcnt vmcnt(3)
; VI-NEXT: v_add_u32_e32 v3, vcc, 3, v3
-; VI-NEXT: s_waitcnt vmcnt(0)
+; VI-NEXT: s_waitcnt vmcnt(2)
; VI-NEXT: v_or_b32_sdwa v51, v4, v3 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:BYTE_0
; VI-NEXT: buffer_load_dword v3, off, s[0:3], s32 offset:684 ; 4-byte Folded Reload
-; VI-NEXT: buffer_load_dword v4, off, s[0:3], s32 offset:488 ; 4-byte Folded Reload
+; VI-NEXT: buffer_load_dword v4, off, s[0:3], s32 offset:484 ; 4-byte Folded Reload
+; VI-NEXT: s_waitcnt vmcnt(3)
+; VI-NEXT: v_add_u32_e32 v1, vcc, 3, v1
+; VI-NEXT: v_or_b32_sdwa v1, v57, v1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:BYTE_0
; VI-NEXT: v_add_u32_e32 v51, vcc, 0x300, v51
; VI-NEXT: v_or_b32_sdwa v12, v12, v51 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:WORD_0
-; VI-NEXT: v_add_u32_e32 v51, vcc, 0x300, v59
-; VI-NEXT: v_or_b32_sdwa v25, v25, v51 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:WORD_0
+; VI-NEXT: v_add_u32_e32 v33, vcc, 0x300, v1
+; VI-NEXT: v_add_u32_e32 v51, vcc, 0x300, v63
+; VI-NEXT: v_or_b32_sdwa v26, v26, v51 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:WORD_0
+; VI-NEXT: v_or_b32_sdwa v30, v30, v33 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:WORD_0
; VI-NEXT: v_add_u32_e32 v12, vcc, 0x3000000, v12
-; VI-NEXT: v_add_u32_e32 v25, vcc, 0x3000000, v25
+; VI-NEXT: v_add_u32_e32 v26, vcc, 0x3000000, v26
+; VI-NEXT: v_add_u32_e32 v30, vcc, 0x3000000, v30
; VI-NEXT: s_waitcnt vmcnt(1)
; VI-NEXT: v_add_u32_e32 v3, vcc, 3, v3
; VI-NEXT: s_waitcnt vmcnt(0)
-; VI-NEXT: v_or_b32_sdwa v33, v4, v3 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:BYTE_0
+; VI-NEXT: v_or_b32_sdwa v57, v4, v3 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:BYTE_0
; VI-NEXT: buffer_load_dword v3, off, s[0:3], s32 offset:456 ; 4-byte Folded Reload
-; VI-NEXT: buffer_load_dword v4, off, s[0:3], s32 offset:484 ; 4-byte Folded Reload
+; VI-NEXT: buffer_load_dword v4, off, s[0:3], s32 offset:480 ; 4-byte Folded Reload
; VI-NEXT: s_waitcnt vmcnt(1)
; VI-NEXT: v_add_u32_e32 v3, vcc, 3, v3
; VI-NEXT: s_waitcnt vmcnt(0)
; VI-NEXT: v_or_b32_sdwa v40, v4, v3 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:BYTE_0
; VI-NEXT: buffer_load_dword v3, off, s[0:3], s32 offset:452 ; 4-byte Folded Reload
-; VI-NEXT: buffer_load_dword v4, off, s[0:3], s32 offset:480 ; 4-byte Folded Reload
-; VI-NEXT: v_add_u32_e32 v2, vcc, 3, v2
-; VI-NEXT: v_or_b32_sdwa v30, v47, v2 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:BYTE_0
-; VI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:792 ; 4-byte Folded Reload
-; VI-NEXT: buffer_load_dword v47, off, s[0:3], s32 offset:464 ; 4-byte Folded Reload
-; VI-NEXT: s_waitcnt vmcnt(3)
+; VI-NEXT: buffer_load_dword v4, off, s[0:3], s32 offset:476 ; 4-byte Folded Reload
+; VI-NEXT: s_waitcnt vmcnt(1)
; VI-NEXT: v_add_u32_e32 v3, vcc, 3, v3
-; VI-NEXT: s_waitcnt vmcnt(2)
+; VI-NEXT: s_waitcnt vmcnt(0)
; VI-NEXT: v_or_b32_sdwa v39, v4, v3 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:BYTE_0
; VI-NEXT: buffer_load_dword v3, off, s[0:3], s32 offset:448 ; 4-byte Folded Reload
-; VI-NEXT: buffer_load_dword v4, off, s[0:3], s32 offset:476 ; 4-byte Folded Reload
-; VI-NEXT: s_waitcnt vmcnt(3)
-; VI-NEXT: v_add_u32_e32 v2, vcc, 3, v2
-; VI-NEXT: v_or_b32_sdwa v2, v63, v2 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:BYTE_0
-; VI-NEXT: v_add_u32_e32 v34, vcc, 0x300, v2
+; VI-NEXT: buffer_load_dword v4, off, s[0:3], s32 offset:660 ; 4-byte Folded Reload
; VI-NEXT: s_waitcnt vmcnt(1)
; VI-NEXT: v_add_u32_e32 v3, vcc, 3, v3
; VI-NEXT: s_waitcnt vmcnt(0)
@@ -159591,15 +159635,14 @@ define inreg <64 x bfloat> @bitcast_v128i8_to_v64bf16_scalar(<128 x i8> inreg %a
; VI-NEXT: v_add_u32_e32 v41, vcc, 0x300, v10
; VI-NEXT: v_add_u32_e32 v10, vcc, 0x300, v55
; VI-NEXT: v_or_b32_sdwa v10, v39, v10 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:WORD_0
-; VI-NEXT: v_add_u32_e32 v39, vcc, 0x300, v53
-; VI-NEXT: v_or_b32_sdwa v27, v28, v39 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:WORD_0
-; VI-NEXT: v_or_b32_sdwa v28, v29, v36 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:WORD_0
-; VI-NEXT: v_or_b32_sdwa v29, v30, v34 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:WORD_0
+; VI-NEXT: v_add_u32_e32 v39, vcc, 0x300, v52
+; VI-NEXT: v_add_u32_e32 v52, vcc, 0x300, v53
+; VI-NEXT: v_or_b32_sdwa v21, v21, v52 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:WORD_0
+; VI-NEXT: v_or_b32_sdwa v27, v27, v39 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:WORD_0
; VI-NEXT: v_add_u32_e32 v9, vcc, 0x3000000, v9
; VI-NEXT: v_add_u32_e32 v10, vcc, 0x3000000, v10
+; VI-NEXT: v_add_u32_e32 v21, vcc, 0x3000000, v21
; VI-NEXT: v_add_u32_e32 v27, vcc, 0x3000000, v27
-; VI-NEXT: v_add_u32_e32 v28, vcc, 0x3000000, v28
-; VI-NEXT: v_add_u32_e32 v29, vcc, 0x3000000, v29
; VI-NEXT: s_waitcnt vmcnt(1)
; VI-NEXT: v_add_u32_e32 v3, vcc, 3, v3
; VI-NEXT: s_waitcnt vmcnt(0)
@@ -159615,18 +159658,14 @@ define inreg <64 x bfloat> @bitcast_v128i8_to_v64bf16_scalar(<128 x i8> inreg %a
; VI-NEXT: v_add_u32_e32 v42, vcc, 0x300, v42
; VI-NEXT: v_or_b32_sdwa v8, v8, v42 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:WORD_0
; VI-NEXT: v_add_u32_e32 v42, vcc, 0x300, v11
-; VI-NEXT: v_add_u32_e32 v11, vcc, 0x300, v40
-; VI-NEXT: v_or_b32_sdwa v11, v33, v11 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:WORD_0
-; VI-NEXT: v_add_u32_e32 v33, vcc, 0x300, v1
-; VI-NEXT: v_or_b32_sdwa v30, v31, v33 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:WORD_0
-; VI-NEXT: buffer_load_dword v31, off, s[0:3], s32 offset:540 ; 4-byte Folded Reload
; VI-NEXT: v_or_b32_sdwa v17, v17, v42 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:WORD_0
+; VI-NEXT: v_add_u32_e32 v11, vcc, 0x300, v40
+; VI-NEXT: v_or_b32_sdwa v11, v57, v11 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:WORD_0
; VI-NEXT: v_add_u32_e32 v8, vcc, 0x3000000, v8
; VI-NEXT: v_add_u32_e32 v11, vcc, 0x3000000, v11
-; VI-NEXT: v_add_u32_e32 v30, vcc, 0x3000000, v30
-; VI-NEXT: s_waitcnt vmcnt(2)
-; VI-NEXT: v_add_u32_e32 v3, vcc, 3, v3
; VI-NEXT: s_waitcnt vmcnt(1)
+; VI-NEXT: v_add_u32_e32 v3, vcc, 3, v3
+; VI-NEXT: s_waitcnt vmcnt(0)
; VI-NEXT: v_or_b32_sdwa v7, v4, v3 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:BYTE_0
; VI-NEXT: buffer_load_dword v3, off, s[0:3], s32 offset:424 ; 4-byte Folded Reload
; VI-NEXT: buffer_load_dword v4, off, s[0:3], s32 offset:636 ; 4-byte Folded Reload
@@ -159666,19 +159705,29 @@ define inreg <64 x bfloat> @bitcast_v128i8_to_v64bf16_scalar(<128 x i8> inreg %a
; VI-NEXT: v_or_b32_sdwa v46, v4, v3 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:BYTE_0
; VI-NEXT: buffer_load_dword v3, off, s[0:3], s32 offset:404 ; 4-byte Folded Reload
; VI-NEXT: buffer_load_dword v4, off, s[0:3], s32 offset:468 ; 4-byte Folded Reload
+; VI-NEXT: v_add_u32_e32 v2, vcc, 3, v2
+; VI-NEXT: v_or_b32_sdwa v29, v47, v2 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:BYTE_0
+; VI-NEXT: buffer_load_dword v47, off, s[0:3], s32 offset:464 ; 4-byte Folded Reload
+; VI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:800 ; 4-byte Folded Reload
; VI-NEXT: v_add_u32_e32 v46, vcc, 0x300, v46
; VI-NEXT: v_or_b32_sdwa v5, v5, v46 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:WORD_0
; VI-NEXT: v_add_u32_e32 v5, vcc, 0x3000000, v5
-; VI-NEXT: s_waitcnt vmcnt(1)
+; VI-NEXT: s_waitcnt vmcnt(3)
; VI-NEXT: v_add_u32_e32 v3, vcc, 3, v3
-; VI-NEXT: s_waitcnt vmcnt(0)
+; VI-NEXT: s_waitcnt vmcnt(2)
; VI-NEXT: v_or_b32_sdwa v3, v4, v3 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:BYTE_0
; VI-NEXT: buffer_load_dword v4, off, s[0:3], s32 offset:400 ; 4-byte Folded Reload
+; VI-NEXT: s_waitcnt vmcnt(1)
+; VI-NEXT: v_add_u32_e32 v2, vcc, 3, v2
+; VI-NEXT: v_or_b32_sdwa v2, v54, v2 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:BYTE_0
+; VI-NEXT: v_add_u32_e32 v34, vcc, 0x300, v2
+; VI-NEXT: v_or_b32_sdwa v29, v29, v34 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:WORD_0
+; VI-NEXT: v_add_u32_e32 v29, vcc, 0x3000000, v29
; VI-NEXT: s_waitcnt vmcnt(0)
; VI-NEXT: v_add_u32_e32 v4, vcc, 3, v4
; VI-NEXT: v_or_b32_sdwa v4, v47, v4 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:BYTE_0
; VI-NEXT: v_add_u32_e32 v47, vcc, 3, v32
-; VI-NEXT: buffer_load_dword v32, off, s[0:3], s32 offset:508 ; 4-byte Folded Reload
+; VI-NEXT: buffer_load_dword v32, off, s[0:3], s32 offset:504 ; 4-byte Folded Reload
; VI-NEXT: v_add_u32_e32 v4, vcc, 0x300, v4
; VI-NEXT: v_or_b32_sdwa v4, v3, v4 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:WORD_0
; VI-NEXT: v_add_u32_e32 v4, vcc, 0x3000000, v4
@@ -159745,35 +159794,38 @@ define inreg <64 x bfloat> @bitcast_v128i8_to_v64bf16_scalar(<128 x i8> inreg %a
; VI-NEXT: s_waitcnt vmcnt(0)
; VI-NEXT: s_setpc_b64 s[30:31]
; VI-NEXT: .LBB89_4:
-; VI-NEXT: buffer_load_dword v46, off, s[0:3], s32 offset:620 ; 4-byte Folded Reload
-; VI-NEXT: buffer_load_dword v52, off, s[0:3], s32 offset:616 ; 4-byte Folded Reload
-; VI-NEXT: buffer_load_dword v50, off, s[0:3], s32 offset:604 ; 4-byte Folded Reload
+; VI-NEXT: buffer_load_dword v60, off, s[0:3], s32 offset:620 ; 4-byte Folded Reload
+; VI-NEXT: buffer_load_dword v43, off, s[0:3], s32 offset:616 ; 4-byte Folded Reload
+; VI-NEXT: buffer_load_dword v58, off, s[0:3], s32 offset:604 ; 4-byte Folded Reload
; VI-NEXT: buffer_load_dword v48, off, s[0:3], s32 offset:612 ; 4-byte Folded Reload
; VI-NEXT: buffer_load_dword v38, off, s[0:3], s32 offset:608 ; 4-byte Folded Reload
-; VI-NEXT: buffer_load_dword v58, off, s[0:3], s32 offset:596 ; 4-byte Folded Reload
-; VI-NEXT: buffer_load_dword v56, off, s[0:3], s32 offset:600 ; 4-byte Folded Reload
-; VI-NEXT: buffer_load_dword v36, off, s[0:3], s32 offset:568 ; 4-byte Folded Reload
-; VI-NEXT: buffer_load_dword v33, off, s[0:3], s32 offset:580 ; 4-byte Folded Reload
-; VI-NEXT: buffer_load_dword v35, off, s[0:3], s32 offset:572 ; 4-byte Folded Reload
-; VI-NEXT: buffer_load_dword v51, off, s[0:3], s32 offset:576 ; 4-byte Folded Reload
-; VI-NEXT: buffer_load_dword v49, off, s[0:3], s32 offset:584 ; 4-byte Folded Reload
-; VI-NEXT: buffer_load_dword v39, off, s[0:3], s32 offset:588 ; 4-byte Folded Reload
-; VI-NEXT: buffer_load_dword v37, off, s[0:3], s32 offset:592 ; 4-byte Folded Reload
-; VI-NEXT: buffer_load_dword v34, off, s[0:3], s32 offset:564 ; 4-byte Folded Reload
-; VI-NEXT: buffer_load_dword v32, off, s[0:3], s32 offset:560 ; 4-byte Folded Reload
-; VI-NEXT: v_mov_b32_e32 v61, v60
-; VI-NEXT: v_mov_b32_e32 v60, v59
+; VI-NEXT: buffer_load_dword v56, off, s[0:3], s32 offset:596 ; 4-byte Folded Reload
+; VI-NEXT: buffer_load_dword v36, off, s[0:3], s32 offset:600 ; 4-byte Folded Reload
+; VI-NEXT: buffer_load_dword v41, off, s[0:3], s32 offset:580 ; 4-byte Folded Reload
+; VI-NEXT: buffer_load_dword v35, off, s[0:3], s32 offset:592 ; 4-byte Folded Reload
+; VI-NEXT: buffer_load_dword v55, off, s[0:3], s32 offset:568 ; 4-byte Folded Reload
+; VI-NEXT: buffer_load_dword v53, off, s[0:3], s32 offset:560 ; 4-byte Folded Reload
+; VI-NEXT: buffer_load_dword v51, off, s[0:3], s32 offset:564 ; 4-byte Folded Reload
+; VI-NEXT: buffer_load_dword v49, off, s[0:3], s32 offset:572 ; 4-byte Folded Reload
+; VI-NEXT: buffer_load_dword v39, off, s[0:3], s32 offset:576 ; 4-byte Folded Reload
+; VI-NEXT: buffer_load_dword v37, off, s[0:3], s32 offset:584 ; 4-byte Folded Reload
+; VI-NEXT: buffer_load_dword v33, off, s[0:3], s32 offset:588 ; 4-byte Folded Reload
+; VI-NEXT: buffer_load_dword v32, off, s[0:3], s32 offset:556 ; 4-byte Folded Reload
+; VI-NEXT: v_mov_b32_e32 v61, v59
+; VI-NEXT: v_mov_b32_e32 v46, v57
+; VI-NEXT: buffer_load_dword v57, off, s[0:3], s32 offset:544 ; 4-byte Folded Reload
+; VI-NEXT: v_mov_b32_e32 v40, v42
+; VI-NEXT: v_mov_b32_e32 v42, v44
+; VI-NEXT: v_mov_b32_e32 v44, v45
; VI-NEXT: v_mov_b32_e32 v45, v62
-; VI-NEXT: buffer_load_dword v62, off, s[0:3], s32 offset:552 ; 4-byte Folded Reload
-; VI-NEXT: buffer_load_dword v59, off, s[0:3], s32 offset:556 ; 4-byte Folded Reload
-; VI-NEXT: v_mov_b32_e32 v57, v5
+; VI-NEXT: buffer_load_dword v62, off, s[0:3], s32 offset:548 ; 4-byte Folded Reload
+; VI-NEXT: buffer_load_dword v59, off, s[0:3], s32 offset:552 ; 4-byte Folded Reload
; VI-NEXT: v_mov_b32_e32 v47, v4
-; VI-NEXT: v_mov_b32_e32 v63, v3
-; VI-NEXT: v_mov_b32_e32 v53, v28
-; VI-NEXT: v_mov_b32_e32 v43, v27
-; VI-NEXT: v_mov_b32_e32 v55, v26
-; VI-NEXT: v_mov_b32_e32 v41, v24
-; VI-NEXT: v_mov_b32_e32 v54, v22
+; VI-NEXT: v_mov_b32_e32 v54, v3
+; VI-NEXT: v_mov_b32_e32 v52, v28
+; VI-NEXT: v_mov_b32_e32 v63, v27
+; VI-NEXT: v_mov_b32_e32 v50, v26
+; VI-NEXT: v_mov_b32_e32 v34, v24
; VI-NEXT: ; implicit-def: $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7_vgpr8_vgpr9_vgpr10_vgpr11_vgpr12_vgpr13_vgpr14_vgpr15
; VI-NEXT: ; implicit-def: $vgpr16_vgpr17_vgpr18_vgpr19_vgpr20_vgpr21_vgpr22_vgpr23_vgpr24_vgpr25_vgpr26_vgpr27_vgpr28_vgpr29_vgpr30_vgpr31
; VI-NEXT: s_branch .LBB89_2
@@ -159835,18 +159887,18 @@ define inreg <64 x bfloat> @bitcast_v128i8_to_v64bf16_scalar(<128 x i8> inreg %a
; GFX9-NEXT: v_lshlrev_b32_e32 v1, 8, v29
; GFX9-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:400 ; 4-byte Folded Spill
; GFX9-NEXT: v_lshlrev_b32_e32 v14, 8, v3
-; GFX9-NEXT: v_lshlrev_b32_e32 v46, 8, v5
+; GFX9-NEXT: v_lshlrev_b32_e32 v18, 8, v5
; GFX9-NEXT: v_lshlrev_b32_e32 v22, 8, v7
; GFX9-NEXT: v_lshlrev_b32_e32 v24, 8, v9
; GFX9-NEXT: v_lshlrev_b32_e32 v26, 8, v11
; GFX9-NEXT: v_lshlrev_b32_e32 v20, 8, v13
-; GFX9-NEXT: v_lshlrev_b32_e32 v28, 8, v15
-; GFX9-NEXT: v_lshlrev_b32_e32 v18, 8, v17
-; GFX9-NEXT: v_lshlrev_b32_e32 v17, 8, v25
-; GFX9-NEXT: v_lshlrev_b32_e32 v12, 8, v27
+; GFX9-NEXT: v_lshlrev_b32_e32 v12, 8, v15
+; GFX9-NEXT: v_lshlrev_b32_e32 v17, 8, v17
; GFX9-NEXT: v_lshlrev_b32_e32 v19, 8, v19
; GFX9-NEXT: v_lshlrev_b32_e32 v21, 8, v21
; GFX9-NEXT: v_lshlrev_b32_e32 v23, 8, v23
+; GFX9-NEXT: v_lshlrev_b32_e32 v25, 8, v25
+; GFX9-NEXT: v_lshlrev_b32_e32 v27, 8, v27
; GFX9-NEXT: s_waitcnt vmcnt(24)
; GFX9-NEXT: v_cmp_ne_u32_e32 vcc, 0, v43
; GFX9-NEXT: s_waitcnt vmcnt(23)
@@ -159875,10 +159927,10 @@ define inreg <64 x bfloat> @bitcast_v128i8_to_v64bf16_scalar(<128 x i8> inreg %a
; GFX9-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:608 ; 4-byte Folded Spill
; GFX9-NEXT: s_waitcnt vmcnt(23)
; GFX9-NEXT: v_lshlrev_b32_e32 v1, 8, v52
-; GFX9-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:600 ; 4-byte Folded Spill
+; GFX9-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:604 ; 4-byte Folded Spill
; GFX9-NEXT: s_waitcnt vmcnt(23)
; GFX9-NEXT: v_lshlrev_b32_e32 v1, 8, v51
-; GFX9-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:604 ; 4-byte Folded Spill
+; GFX9-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:600 ; 4-byte Folded Spill
; GFX9-NEXT: s_waitcnt vmcnt(23)
; GFX9-NEXT: v_lshlrev_b32_e32 v1, 8, v50
; GFX9-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:496 ; 4-byte Folded Spill
@@ -159890,7 +159942,7 @@ define inreg <64 x bfloat> @bitcast_v128i8_to_v64bf16_scalar(<128 x i8> inreg %a
; GFX9-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:596 ; 4-byte Folded Spill
; GFX9-NEXT: s_waitcnt vmcnt(23)
; GFX9-NEXT: v_lshlrev_b32_e32 v1, 8, v39
-; GFX9-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:592 ; 4-byte Folded Spill
+; GFX9-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:588 ; 4-byte Folded Spill
; GFX9-NEXT: s_waitcnt vmcnt(23)
; GFX9-NEXT: v_lshlrev_b32_e32 v1, 8, v30
; GFX9-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:448 ; 4-byte Folded Spill
@@ -159938,7 +159990,7 @@ define inreg <64 x bfloat> @bitcast_v128i8_to_v64bf16_scalar(<128 x i8> inreg %a
; GFX9-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:584 ; 4-byte Folded Spill
; GFX9-NEXT: s_waitcnt vmcnt(6)
; GFX9-NEXT: v_lshlrev_b32_e32 v1, 8, v3
-; GFX9-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:588 ; 4-byte Folded Spill
+; GFX9-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:592 ; 4-byte Folded Spill
; GFX9-NEXT: s_waitcnt vmcnt(6)
; GFX9-NEXT: v_lshlrev_b32_e32 v1, 8, v13
; GFX9-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:536 ; 4-byte Folded Spill
@@ -159965,23 +160017,23 @@ define inreg <64 x bfloat> @bitcast_v128i8_to_v64bf16_scalar(<128 x i8> inreg %a
; GFX9-NEXT: v_lshlrev_b32_e32 v11, 8, v11
; GFX9-NEXT: s_waitcnt vmcnt(6)
; GFX9-NEXT: v_lshlrev_b32_e32 v1, 8, v1
-; GFX9-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:556 ; 4-byte Folded Spill
+; GFX9-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:572 ; 4-byte Folded Spill
; GFX9-NEXT: s_waitcnt vmcnt(6)
; GFX9-NEXT: v_lshlrev_b32_e32 v1, 8, v15
-; GFX9-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:572 ; 4-byte Folded Spill
+; GFX9-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:568 ; 4-byte Folded Spill
; GFX9-NEXT: s_waitcnt vmcnt(6)
; GFX9-NEXT: v_lshlrev_b32_e32 v1, 8, v3
; GFX9-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:552 ; 4-byte Folded Spill
; GFX9-NEXT: s_waitcnt vmcnt(6)
; GFX9-NEXT: v_lshlrev_b32_e32 v1, 8, v13
-; GFX9-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:568 ; 4-byte Folded Spill
+; GFX9-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:564 ; 4-byte Folded Spill
; GFX9-NEXT: s_waitcnt vmcnt(6)
; GFX9-NEXT: v_lshlrev_b32_e32 v1, 8, v5
; GFX9-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:548 ; 4-byte Folded Spill
; GFX9-NEXT: s_waitcnt vmcnt(6)
; GFX9-NEXT: v_lshlrev_b32_e32 v1, 8, v9
-; GFX9-NEXT: buffer_store_dword v11, off, s[0:3], s32 offset:564 ; 4-byte Folded Spill
-; GFX9-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:560 ; 4-byte Folded Spill
+; GFX9-NEXT: buffer_store_dword v11, off, s[0:3], s32 offset:560 ; 4-byte Folded Spill
+; GFX9-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:556 ; 4-byte Folded Spill
; GFX9-NEXT: buffer_load_ushort v3, off, s[0:3], s32 offset:312
; GFX9-NEXT: s_nop 0
; GFX9-NEXT: buffer_load_ushort v1, off, s[0:3], s32 offset:320
@@ -159994,48 +160046,49 @@ define inreg <64 x bfloat> @bitcast_v128i8_to_v64bf16_scalar(<128 x i8> inreg %a
; GFX9-NEXT: s_waitcnt vmcnt(15)
; GFX9-NEXT: v_lshlrev_b32_e32 v7, 8, v7
; GFX9-NEXT: s_waitcnt vmcnt(7)
-; GFX9-NEXT: v_lshlrev_b32_e32 v3, 8, v3
+; GFX9-NEXT: v_lshlrev_b32_e32 v9, 8, v3
; GFX9-NEXT: s_waitcnt vmcnt(6)
-; GFX9-NEXT: v_lshlrev_b32_e32 v9, 8, v1
+; GFX9-NEXT: v_lshlrev_b32_e32 v1, 8, v1
+; GFX9-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:520 ; 4-byte Folded Spill
; GFX9-NEXT: buffer_load_ushort v52, off, s[0:3], s32 offset:44
; GFX9-NEXT: buffer_load_ushort v50, off, s[0:3], s32 offset:52
; GFX9-NEXT: buffer_load_ushort v49, off, s[0:3], s32 offset:60
; GFX9-NEXT: buffer_load_ushort v15, off, s[0:3], s32 offset:68
; GFX9-NEXT: buffer_load_ushort v48, off, s[0:3], s32 offset:76
; GFX9-NEXT: buffer_load_ushort v39, off, s[0:3], s32 offset:84
-; GFX9-NEXT: buffer_load_ushort v25, off, s[0:3], s32 offset:92
+; GFX9-NEXT: buffer_load_ushort v40, off, s[0:3], s32 offset:92
; GFX9-NEXT: buffer_load_ushort v34, off, s[0:3], s32 offset:100
; GFX9-NEXT: buffer_load_ushort v33, off, s[0:3], s32 offset:108
-; GFX9-NEXT: buffer_load_ushort v45, off, s[0:3], s32 offset:116
+; GFX9-NEXT: buffer_load_ushort v46, off, s[0:3], s32 offset:116
; GFX9-NEXT: buffer_load_ushort v43, off, s[0:3], s32 offset:124
; GFX9-NEXT: buffer_load_ushort v38, off, s[0:3], s32 offset:132
-; GFX9-NEXT: buffer_load_ushort v56, off, s[0:3], s32 offset:140
+; GFX9-NEXT: buffer_load_ushort v61, off, s[0:3], s32 offset:140
; GFX9-NEXT: buffer_load_ushort v47, off, s[0:3], s32 offset:148
-; GFX9-NEXT: buffer_load_ushort v61, off, s[0:3], s32 offset:156
+; GFX9-NEXT: buffer_load_ushort v62, off, s[0:3], s32 offset:156
; GFX9-NEXT: buffer_load_ushort v1, off, s[0:3], s32 offset:164
-; GFX9-NEXT: s_waitcnt vmcnt(21)
-; GFX9-NEXT: v_lshlrev_b32_e32 v5, 8, v5
+; GFX9-NEXT: s_waitcnt vmcnt(22)
+; GFX9-NEXT: v_lshlrev_b32_e32 v3, 8, v5
; GFX9-NEXT: s_waitcnt vmcnt(0)
; GFX9-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:576 ; 4-byte Folded Spill
-; GFX9-NEXT: buffer_load_ushort v31, off, s[0:3], s32 offset:172
-; GFX9-NEXT: buffer_load_ushort v32, off, s[0:3], s32 offset:180
-; GFX9-NEXT: buffer_load_ushort v51, off, s[0:3], s32 offset:188
-; GFX9-NEXT: buffer_load_ushort v30, off, s[0:3], s32 offset:196
-; GFX9-NEXT: buffer_load_ushort v44, off, s[0:3], s32 offset:204
+; GFX9-NEXT: buffer_load_ushort v57, off, s[0:3], s32 offset:172
+; GFX9-NEXT: buffer_load_ushort v31, off, s[0:3], s32 offset:180
+; GFX9-NEXT: buffer_load_ushort v36, off, s[0:3], s32 offset:188
+; GFX9-NEXT: buffer_load_ushort v37, off, s[0:3], s32 offset:196
+; GFX9-NEXT: buffer_load_ushort v30, off, s[0:3], s32 offset:204
; GFX9-NEXT: buffer_load_ushort v35, off, s[0:3], s32 offset:212
-; GFX9-NEXT: buffer_load_ushort v36, off, s[0:3], s32 offset:220
-; GFX9-NEXT: buffer_load_ushort v37, off, s[0:3], s32 offset:228
+; GFX9-NEXT: buffer_load_ushort v28, off, s[0:3], s32 offset:220
+; GFX9-NEXT: buffer_load_ushort v51, off, s[0:3], s32 offset:228
; GFX9-NEXT: buffer_load_ushort v41, off, s[0:3], s32 offset:236
; GFX9-NEXT: buffer_load_ushort v54, off, s[0:3], s32 offset:244
-; GFX9-NEXT: buffer_load_ushort v29, off, s[0:3], s32 offset:252
-; GFX9-NEXT: buffer_load_ushort v27, off, s[0:3], s32 offset:260
+; GFX9-NEXT: buffer_load_ushort v44, off, s[0:3], s32 offset:252
+; GFX9-NEXT: buffer_load_ushort v45, off, s[0:3], s32 offset:260
; GFX9-NEXT: buffer_load_ushort v60, off, s[0:3], s32 offset:268
-; GFX9-NEXT: buffer_load_ushort v57, off, s[0:3], s32 offset:276
+; GFX9-NEXT: buffer_load_ushort v29, off, s[0:3], s32 offset:276
; GFX9-NEXT: buffer_load_ushort v59, off, s[0:3], s32 offset:284
; GFX9-NEXT: buffer_load_ushort v63, off, s[0:3], s32 offset:292
-; GFX9-NEXT: buffer_load_ushort v62, off, s[0:3], s32 offset:300
-; GFX9-NEXT: buffer_load_ushort v40, off, s[0:3], s32 offset:308
-; GFX9-NEXT: buffer_load_ushort v58, off, s[0:3], s32 offset:316
+; GFX9-NEXT: buffer_load_ushort v58, off, s[0:3], s32 offset:300
+; GFX9-NEXT: buffer_load_ushort v32, off, s[0:3], s32 offset:308
+; GFX9-NEXT: buffer_load_ushort v56, off, s[0:3], s32 offset:316
; GFX9-NEXT: buffer_load_ushort v1, off, s[0:3], s32 offset:324
; GFX9-NEXT: s_nop 0
; GFX9-NEXT: buffer_store_dword v10, off, s[0:3], s32 offset:620 ; 4-byte Folded Spill
@@ -160046,55 +160099,54 @@ define inreg <64 x bfloat> @bitcast_v128i8_to_v64bf16_scalar(<128 x i8> inreg %a
; GFX9-NEXT: buffer_store_dword v0, off, s[0:3], s32 offset:640 ; 4-byte Folded Spill
; GFX9-NEXT: buffer_store_dword v2, off, s[0:3], s32 offset:644 ; 4-byte Folded Spill
; GFX9-NEXT: buffer_store_dword v4, off, s[0:3], s32 offset:648 ; 4-byte Folded Spill
-; GFX9-NEXT: buffer_store_dword v12, off, s[0:3], s32 offset:652 ; 4-byte Folded Spill
+; GFX9-NEXT: buffer_store_dword v27, off, s[0:3], s32 offset:652 ; 4-byte Folded Spill
; GFX9-NEXT: buffer_store_dword v7, off, s[0:3], s32 offset:508 ; 4-byte Folded Spill
; GFX9-NEXT: s_waitcnt vmcnt(28)
-; GFX9-NEXT: buffer_store_dword v32, off, s[0:3], s32 offset:656 ; 4-byte Folded Spill
-; GFX9-NEXT: buffer_store_dword v31, off, s[0:3], s32 offset:660 ; 4-byte Folded Spill
-; GFX9-NEXT: buffer_store_dword v17, off, s[0:3], s32 offset:664 ; 4-byte Folded Spill
+; GFX9-NEXT: buffer_store_dword v31, off, s[0:3], s32 offset:656 ; 4-byte Folded Spill
+; GFX9-NEXT: buffer_store_dword v57, off, s[0:3], s32 offset:660 ; 4-byte Folded Spill
+; GFX9-NEXT: buffer_store_dword v25, off, s[0:3], s32 offset:664 ; 4-byte Folded Spill
; GFX9-NEXT: buffer_store_dword v23, off, s[0:3], s32 offset:668 ; 4-byte Folded Spill
; GFX9-NEXT: s_waitcnt vmcnt(30)
-; GFX9-NEXT: buffer_store_dword v30, off, s[0:3], s32 offset:672 ; 4-byte Folded Spill
-; GFX9-NEXT: buffer_store_dword v51, off, s[0:3], s32 offset:676 ; 4-byte Folded Spill
+; GFX9-NEXT: buffer_store_dword v37, off, s[0:3], s32 offset:672 ; 4-byte Folded Spill
+; GFX9-NEXT: buffer_store_dword v36, off, s[0:3], s32 offset:676 ; 4-byte Folded Spill
; GFX9-NEXT: buffer_store_dword v21, off, s[0:3], s32 offset:680 ; 4-byte Folded Spill
; GFX9-NEXT: buffer_store_dword v19, off, s[0:3], s32 offset:684 ; 4-byte Folded Spill
-; GFX9-NEXT: buffer_store_dword v3, off, s[0:3], s32 offset:512 ; 4-byte Folded Spill
+; GFX9-NEXT: buffer_store_dword v9, off, s[0:3], s32 offset:512 ; 4-byte Folded Spill
; GFX9-NEXT: s_waitcnt vmcnt(33)
; GFX9-NEXT: buffer_store_dword v35, off, s[0:3], s32 offset:688 ; 4-byte Folded Spill
-; GFX9-NEXT: buffer_store_dword v44, off, s[0:3], s32 offset:692 ; 4-byte Folded Spill
-; GFX9-NEXT: buffer_store_dword v18, off, s[0:3], s32 offset:696 ; 4-byte Folded Spill
-; GFX9-NEXT: buffer_store_dword v28, off, s[0:3], s32 offset:700 ; 4-byte Folded Spill
-; GFX9-NEXT: buffer_store_dword v5, off, s[0:3], s32 offset:516 ; 4-byte Folded Spill
+; GFX9-NEXT: buffer_store_dword v30, off, s[0:3], s32 offset:692 ; 4-byte Folded Spill
+; GFX9-NEXT: buffer_store_dword v17, off, s[0:3], s32 offset:696 ; 4-byte Folded Spill
+; GFX9-NEXT: buffer_store_dword v12, off, s[0:3], s32 offset:700 ; 4-byte Folded Spill
+; GFX9-NEXT: buffer_store_dword v3, off, s[0:3], s32 offset:516 ; 4-byte Folded Spill
; GFX9-NEXT: s_waitcnt vmcnt(36)
-; GFX9-NEXT: buffer_store_dword v37, off, s[0:3], s32 offset:704 ; 4-byte Folded Spill
-; GFX9-NEXT: buffer_store_dword v9, off, s[0:3], s32 offset:520 ; 4-byte Folded Spill
-; GFX9-NEXT: buffer_store_dword v36, off, s[0:3], s32 offset:708 ; 4-byte Folded Spill
+; GFX9-NEXT: buffer_store_dword v51, off, s[0:3], s32 offset:704 ; 4-byte Folded Spill
+; GFX9-NEXT: buffer_store_dword v28, off, s[0:3], s32 offset:708 ; 4-byte Folded Spill
; GFX9-NEXT: buffer_store_dword v20, off, s[0:3], s32 offset:712 ; 4-byte Folded Spill
; GFX9-NEXT: buffer_store_dword v26, off, s[0:3], s32 offset:716 ; 4-byte Folded Spill
-; GFX9-NEXT: s_waitcnt vmcnt(39)
+; GFX9-NEXT: s_waitcnt vmcnt(38)
; GFX9-NEXT: buffer_store_dword v54, off, s[0:3], s32 offset:720 ; 4-byte Folded Spill
; GFX9-NEXT: buffer_store_dword v41, off, s[0:3], s32 offset:724 ; 4-byte Folded Spill
; GFX9-NEXT: buffer_store_dword v24, off, s[0:3], s32 offset:728 ; 4-byte Folded Spill
; GFX9-NEXT: buffer_store_dword v22, off, s[0:3], s32 offset:732 ; 4-byte Folded Spill
-; GFX9-NEXT: s_waitcnt vmcnt(41)
-; GFX9-NEXT: buffer_store_dword v27, off, s[0:3], s32 offset:736 ; 4-byte Folded Spill
-; GFX9-NEXT: buffer_store_dword v29, off, s[0:3], s32 offset:740 ; 4-byte Folded Spill
-; GFX9-NEXT: s_waitcnt vmcnt(41)
-; GFX9-NEXT: buffer_store_dword v57, off, s[0:3], s32 offset:744 ; 4-byte Folded Spill
+; GFX9-NEXT: s_waitcnt vmcnt(40)
+; GFX9-NEXT: buffer_store_dword v45, off, s[0:3], s32 offset:736 ; 4-byte Folded Spill
+; GFX9-NEXT: buffer_store_dword v44, off, s[0:3], s32 offset:740 ; 4-byte Folded Spill
+; GFX9-NEXT: s_waitcnt vmcnt(40)
+; GFX9-NEXT: buffer_store_dword v29, off, s[0:3], s32 offset:744 ; 4-byte Folded Spill
; GFX9-NEXT: buffer_store_dword v60, off, s[0:3], s32 offset:748 ; 4-byte Folded Spill
-; GFX9-NEXT: s_waitcnt vmcnt(41)
+; GFX9-NEXT: s_waitcnt vmcnt(40)
; GFX9-NEXT: buffer_store_dword v63, off, s[0:3], s32 offset:752 ; 4-byte Folded Spill
; GFX9-NEXT: buffer_store_dword v59, off, s[0:3], s32 offset:756 ; 4-byte Folded Spill
-; GFX9-NEXT: s_waitcnt vmcnt(41)
-; GFX9-NEXT: buffer_store_dword v40, off, s[0:3], s32 offset:760 ; 4-byte Folded Spill
-; GFX9-NEXT: buffer_store_dword v62, off, s[0:3], s32 offset:764 ; 4-byte Folded Spill
-; GFX9-NEXT: s_waitcnt vmcnt(41)
+; GFX9-NEXT: s_waitcnt vmcnt(40)
+; GFX9-NEXT: buffer_store_dword v32, off, s[0:3], s32 offset:760 ; 4-byte Folded Spill
+; GFX9-NEXT: buffer_store_dword v58, off, s[0:3], s32 offset:764 ; 4-byte Folded Spill
+; GFX9-NEXT: s_waitcnt vmcnt(40)
; GFX9-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:524 ; 4-byte Folded Spill
-; GFX9-NEXT: buffer_store_dword v58, off, s[0:3], s32 offset:768 ; 4-byte Folded Spill
+; GFX9-NEXT: buffer_store_dword v56, off, s[0:3], s32 offset:768 ; 4-byte Folded Spill
; GFX9-NEXT: buffer_store_dword v39, off, s[0:3], s32 offset:772 ; 4-byte Folded Spill
; GFX9-NEXT: buffer_store_dword v48, off, s[0:3], s32 offset:776 ; 4-byte Folded Spill
; GFX9-NEXT: buffer_store_dword v34, off, s[0:3], s32 offset:780 ; 4-byte Folded Spill
-; GFX9-NEXT: buffer_store_dword v25, off, s[0:3], s32 offset:784 ; 4-byte Folded Spill
+; GFX9-NEXT: buffer_store_dword v40, off, s[0:3], s32 offset:784 ; 4-byte Folded Spill
; GFX9-NEXT: buffer_store_dword v15, off, s[0:3], s32 offset:788 ; 4-byte Folded Spill
; GFX9-NEXT: buffer_store_dword v49, off, s[0:3], s32 offset:792 ; 4-byte Folded Spill
; GFX9-NEXT: buffer_store_dword v50, off, s[0:3], s32 offset:796 ; 4-byte Folded Spill
@@ -160104,7 +160156,7 @@ define inreg <64 x bfloat> @bitcast_v128i8_to_v64bf16_scalar(<128 x i8> inreg %a
; GFX9-NEXT: buffer_store_dword v53, off, s[0:3], s32 offset:812 ; 4-byte Folded Spill
; GFX9-NEXT: buffer_store_dword v55, off, s[0:3], s32 offset:816 ; 4-byte Folded Spill
; GFX9-NEXT: buffer_store_dword v42, off, s[0:3], s32 offset:820 ; 4-byte Folded Spill
-; GFX9-NEXT: buffer_store_dword v46, off, s[0:3], s32 offset:824 ; 4-byte Folded Spill
+; GFX9-NEXT: buffer_store_dword v18, off, s[0:3], s32 offset:824 ; 4-byte Folded Spill
; GFX9-NEXT: buffer_store_dword v14, off, s[0:3], s32 offset:828 ; 4-byte Folded Spill
; GFX9-NEXT: buffer_store_dword v11, off, s[0:3], s32 offset:832 ; 4-byte Folded Spill
; GFX9-NEXT: s_cbranch_scc0 .LBB89_2
@@ -160117,7 +160169,7 @@ define inreg <64 x bfloat> @bitcast_v128i8_to_v64bf16_scalar(<128 x i8> inreg %a
; GFX9-NEXT: v_or_b32_sdwa v1, v2, v14 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
; GFX9-NEXT: v_and_b32_e32 v1, 0xffff, v1
; GFX9-NEXT: v_or_b32_sdwa v2, v0, v16 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
-; GFX9-NEXT: v_or_b32_sdwa v4, v4, v46 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; GFX9-NEXT: v_or_b32_sdwa v4, v4, v18 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
; GFX9-NEXT: v_or_b32_sdwa v0, v6, v22 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
; GFX9-NEXT: v_lshl_or_b32 v4, v4, 16, v1
; GFX9-NEXT: v_and_b32_e32 v0, 0xffff, v0
@@ -160154,10 +160206,10 @@ define inreg <64 x bfloat> @bitcast_v128i8_to_v64bf16_scalar(<128 x i8> inreg %a
; GFX9-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:452 ; 4-byte Folded Reload
; GFX9-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:456 ; 4-byte Folded Reload
; GFX9-NEXT: s_waitcnt vmcnt(1)
-; GFX9-NEXT: v_or_b32_sdwa v0, v0, v28 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; GFX9-NEXT: v_or_b32_sdwa v0, v0, v12 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
; GFX9-NEXT: v_and_b32_e32 v0, 0xffff, v0
; GFX9-NEXT: s_waitcnt vmcnt(0)
-; GFX9-NEXT: v_or_b32_sdwa v1, v1, v18 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; GFX9-NEXT: v_or_b32_sdwa v1, v1, v17 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
; GFX9-NEXT: v_lshl_or_b32 v7, v1, 16, v0
; GFX9-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:436 ; 4-byte Folded Reload
; GFX9-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:440 ; 4-byte Folded Reload
@@ -160173,13 +160225,13 @@ define inreg <64 x bfloat> @bitcast_v128i8_to_v64bf16_scalar(<128 x i8> inreg %a
; GFX9-NEXT: v_or_b32_sdwa v0, v0, v23 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
; GFX9-NEXT: v_and_b32_e32 v0, 0xffff, v0
; GFX9-NEXT: s_waitcnt vmcnt(0)
-; GFX9-NEXT: v_or_b32_sdwa v1, v1, v17 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; GFX9-NEXT: v_or_b32_sdwa v1, v1, v25 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
; GFX9-NEXT: v_lshl_or_b32 v9, v1, 16, v0
; GFX9-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:420 ; 4-byte Folded Reload
; GFX9-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:424 ; 4-byte Folded Reload
; GFX9-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:400 ; 4-byte Folded Reload
; GFX9-NEXT: s_waitcnt vmcnt(2)
-; GFX9-NEXT: v_or_b32_sdwa v0, v0, v12 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; GFX9-NEXT: v_or_b32_sdwa v0, v0, v27 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
; GFX9-NEXT: v_and_b32_e32 v0, 0xffff, v0
; GFX9-NEXT: s_waitcnt vmcnt(0)
; GFX9-NEXT: v_or_b32_sdwa v1, v1, v2 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
@@ -160187,7 +160239,7 @@ define inreg <64 x bfloat> @bitcast_v128i8_to_v64bf16_scalar(<128 x i8> inreg %a
; GFX9-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:416 ; 4-byte Folded Reload
; GFX9-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:504 ; 4-byte Folded Reload
; GFX9-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:492 ; 4-byte Folded Reload
-; GFX9-NEXT: buffer_load_dword v16, off, s[0:3], s32 offset:592 ; 4-byte Folded Reload
+; GFX9-NEXT: buffer_load_dword v16, off, s[0:3], s32 offset:588 ; 4-byte Folded Reload
; GFX9-NEXT: s_waitcnt vmcnt(2)
; GFX9-NEXT: v_or_b32_sdwa v0, v0, v1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
; GFX9-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:404 ; 4-byte Folded Reload
@@ -160228,8 +160280,8 @@ define inreg <64 x bfloat> @bitcast_v128i8_to_v64bf16_scalar(<128 x i8> inreg %a
; GFX9-NEXT: v_mov_b32_e32 v52, v1
; GFX9-NEXT: v_or_b32_sdwa v1, v50, v1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
; GFX9-NEXT: v_lshl_or_b32 v14, v1, 16, v0
-; GFX9-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:600 ; 4-byte Folded Reload
-; GFX9-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:604 ; 4-byte Folded Reload
+; GFX9-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:604 ; 4-byte Folded Reload
+; GFX9-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:600 ; 4-byte Folded Reload
; GFX9-NEXT: s_waitcnt vmcnt(1)
; GFX9-NEXT: v_mov_b32_e32 v50, v0
; GFX9-NEXT: v_or_b32_sdwa v0, v49, v0 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
@@ -160247,16 +160299,16 @@ define inreg <64 x bfloat> @bitcast_v128i8_to_v64bf16_scalar(<128 x i8> inreg %a
; GFX9-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:448 ; 4-byte Folded Reload
; GFX9-NEXT: s_waitcnt vmcnt(1)
; GFX9-NEXT: v_mov_b32_e32 v48, v1
-; GFX9-NEXT: v_or_b32_sdwa v1, v25, v1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; GFX9-NEXT: v_or_b32_sdwa v1, v40, v1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
; GFX9-NEXT: v_and_b32_e32 v1, 0xffff, v1
; GFX9-NEXT: v_lshl_or_b32 v17, v17, 16, v1
; GFX9-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:444 ; 4-byte Folded Reload
; GFX9-NEXT: s_waitcnt vmcnt(1)
; GFX9-NEXT: v_or_b32_sdwa v0, v33, v0 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
; GFX9-NEXT: v_and_b32_e32 v0, 0xffff, v0
-; GFX9-NEXT: v_mov_b32_e32 v33, v45
+; GFX9-NEXT: v_mov_b32_e32 v33, v46
; GFX9-NEXT: s_waitcnt vmcnt(0)
-; GFX9-NEXT: v_or_b32_sdwa v1, v45, v1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; GFX9-NEXT: v_or_b32_sdwa v1, v46, v1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
; GFX9-NEXT: v_lshl_or_b32 v18, v1, 16, v0
; GFX9-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:464 ; 4-byte Folded Reload
; GFX9-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:460 ; 4-byte Folded Reload
@@ -160269,7 +160321,7 @@ define inreg <64 x bfloat> @bitcast_v128i8_to_v64bf16_scalar(<128 x i8> inreg %a
; GFX9-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:472 ; 4-byte Folded Reload
; GFX9-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:468 ; 4-byte Folded Reload
; GFX9-NEXT: s_waitcnt vmcnt(1)
-; GFX9-NEXT: v_or_b32_sdwa v0, v56, v0 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; GFX9-NEXT: v_or_b32_sdwa v0, v42, v0 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
; GFX9-NEXT: v_and_b32_e32 v0, 0xffff, v0
; GFX9-NEXT: s_waitcnt vmcnt(0)
; GFX9-NEXT: v_or_b32_sdwa v1, v47, v1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
@@ -160278,7 +160330,7 @@ define inreg <64 x bfloat> @bitcast_v128i8_to_v64bf16_scalar(<128 x i8> inreg %a
; GFX9-NEXT: buffer_load_dword v61, off, s[0:3], s32 offset:576 ; 4-byte Folded Reload
; GFX9-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:476 ; 4-byte Folded Reload
; GFX9-NEXT: s_waitcnt vmcnt(2)
-; GFX9-NEXT: v_or_b32_sdwa v0, v42, v0 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; GFX9-NEXT: v_or_b32_sdwa v0, v62, v0 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
; GFX9-NEXT: v_and_b32_e32 v0, 0xffff, v0
; GFX9-NEXT: s_waitcnt vmcnt(0)
; GFX9-NEXT: v_or_b32_sdwa v1, v61, v1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
@@ -160286,121 +160338,122 @@ define inreg <64 x bfloat> @bitcast_v128i8_to_v64bf16_scalar(<128 x i8> inreg %a
; GFX9-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:528 ; 4-byte Folded Reload
; GFX9-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:532 ; 4-byte Folded Reload
; GFX9-NEXT: s_waitcnt vmcnt(1)
-; GFX9-NEXT: v_or_b32_sdwa v0, v31, v0 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; GFX9-NEXT: v_or_b32_sdwa v0, v57, v0 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
; GFX9-NEXT: v_and_b32_e32 v0, 0xffff, v0
; GFX9-NEXT: s_waitcnt vmcnt(0)
-; GFX9-NEXT: v_or_b32_sdwa v1, v32, v1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; GFX9-NEXT: v_or_b32_sdwa v1, v31, v1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
; GFX9-NEXT: v_lshl_or_b32 v22, v1, 16, v0
; GFX9-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:488 ; 4-byte Folded Reload
; GFX9-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:584 ; 4-byte Folded Reload
; GFX9-NEXT: s_waitcnt vmcnt(1)
-; GFX9-NEXT: v_or_b32_sdwa v0, v51, v0 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; GFX9-NEXT: v_or_b32_sdwa v0, v36, v0 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
; GFX9-NEXT: v_and_b32_e32 v0, 0xffff, v0
; GFX9-NEXT: s_waitcnt vmcnt(0)
; GFX9-NEXT: v_mov_b32_e32 v34, v1
-; GFX9-NEXT: v_or_b32_sdwa v1, v30, v1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; GFX9-NEXT: v_or_b32_sdwa v1, v37, v1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
; GFX9-NEXT: v_lshl_or_b32 v23, v1, 16, v0
-; GFX9-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:588 ; 4-byte Folded Reload
-; GFX9-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:572 ; 4-byte Folded Reload
-; GFX9-NEXT: buffer_load_dword v45, off, s[0:3], s32 offset:536 ; 4-byte Folded Reload
-; GFX9-NEXT: buffer_load_dword v51, off, s[0:3], s32 offset:564 ; 4-byte Folded Reload
-; GFX9-NEXT: buffer_load_dword v32, off, s[0:3], s32 offset:552 ; 4-byte Folded Reload
+; GFX9-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:592 ; 4-byte Folded Reload
+; GFX9-NEXT: buffer_load_dword v57, off, s[0:3], s32 offset:572 ; 4-byte Folded Reload
+; GFX9-NEXT: buffer_load_dword v46, off, s[0:3], s32 offset:536 ; 4-byte Folded Reload
+; GFX9-NEXT: buffer_load_dword v36, off, s[0:3], s32 offset:544 ; 4-byte Folded Reload
+; GFX9-NEXT: buffer_load_dword v37, off, s[0:3], s32 offset:540 ; 4-byte Folded Reload
; GFX9-NEXT: s_waitcnt vmcnt(4)
-; GFX9-NEXT: v_mov_b32_e32 v46, v0
-; GFX9-NEXT: v_or_b32_sdwa v0, v44, v0 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; GFX9-NEXT: v_mov_b32_e32 v40, v0
+; GFX9-NEXT: v_or_b32_sdwa v0, v30, v0 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
; GFX9-NEXT: v_and_b32_e32 v0, 0xffff, v0
; GFX9-NEXT: s_waitcnt vmcnt(2)
-; GFX9-NEXT: v_or_b32_sdwa v1, v35, v45 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; GFX9-NEXT: v_or_b32_sdwa v1, v35, v46 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
; GFX9-NEXT: v_lshl_or_b32 v24, v1, 16, v0
; GFX9-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:580 ; 4-byte Folded Reload
-; GFX9-NEXT: buffer_load_dword v44, off, s[0:3], s32 offset:568 ; 4-byte Folded Reload
-; GFX9-NEXT: v_mov_b32_e32 v35, v45
-; GFX9-NEXT: v_mov_b32_e32 v45, v61
-; GFX9-NEXT: v_mov_b32_e32 v61, v42
+; GFX9-NEXT: s_waitcnt vmcnt(2)
+; GFX9-NEXT: v_or_b32_sdwa v1, v51, v36 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; GFX9-NEXT: buffer_load_dword v51, off, s[0:3], s32 offset:560 ; 4-byte Folded Reload
; GFX9-NEXT: s_waitcnt vmcnt(1)
; GFX9-NEXT: v_mov_b32_e32 v38, v0
-; GFX9-NEXT: v_or_b32_sdwa v0, v36, v0 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
-; GFX9-NEXT: buffer_load_dword v36, off, s[0:3], s32 offset:544 ; 4-byte Folded Reload
+; GFX9-NEXT: v_or_b32_sdwa v0, v28, v0 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
; GFX9-NEXT: v_and_b32_e32 v0, 0xffff, v0
-; GFX9-NEXT: s_waitcnt vmcnt(0)
-; GFX9-NEXT: v_or_b32_sdwa v1, v37, v36 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
-; GFX9-NEXT: buffer_load_dword v37, off, s[0:3], s32 offset:540 ; 4-byte Folded Reload
; GFX9-NEXT: v_lshl_or_b32 v25, v1, 16, v0
-; GFX9-NEXT: v_or_b32_sdwa v1, v54, v51 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
-; GFX9-NEXT: v_mov_b32_e32 v54, v2
-; GFX9-NEXT: s_waitcnt vmcnt(0)
; GFX9-NEXT: v_or_b32_sdwa v0, v41, v37 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
-; GFX9-NEXT: buffer_load_dword v41, off, s[0:3], s32 offset:556 ; 4-byte Folded Reload
; GFX9-NEXT: v_and_b32_e32 v0, 0xffff, v0
-; GFX9-NEXT: v_lshl_or_b32 v26, v1, 16, v0
-; GFX9-NEXT: v_or_b32_sdwa v1, v27, v2 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; GFX9-NEXT: v_mov_b32_e32 v41, v57
; GFX9-NEXT: s_waitcnt vmcnt(0)
-; GFX9-NEXT: v_or_b32_sdwa v0, v29, v41 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; GFX9-NEXT: v_or_b32_sdwa v1, v54, v51 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; GFX9-NEXT: buffer_load_dword v54, off, s[0:3], s32 offset:568 ; 4-byte Folded Reload
+; GFX9-NEXT: v_lshl_or_b32 v26, v1, 16, v0
+; GFX9-NEXT: v_or_b32_sdwa v0, v44, v57 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; GFX9-NEXT: buffer_load_dword v44, off, s[0:3], s32 offset:564 ; 4-byte Folded Reload
; GFX9-NEXT: v_and_b32_e32 v0, 0xffff, v0
+; GFX9-NEXT: s_waitcnt vmcnt(1)
+; GFX9-NEXT: v_or_b32_sdwa v1, v45, v54 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; GFX9-NEXT: buffer_load_dword v45, off, s[0:3], s32 offset:552 ; 4-byte Folded Reload
; GFX9-NEXT: v_lshl_or_b32 v27, v1, 16, v0
-; GFX9-NEXT: v_or_b32_sdwa v0, v60, v32 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; GFX9-NEXT: s_waitcnt vmcnt(0)
+; GFX9-NEXT: v_or_b32_sdwa v0, v60, v45 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
; GFX9-NEXT: buffer_load_dword v60, off, s[0:3], s32 offset:548 ; 4-byte Folded Reload
-; GFX9-NEXT: v_or_b32_sdwa v1, v57, v44 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
-; GFX9-NEXT: buffer_load_dword v57, off, s[0:3], s32 offset:560 ; 4-byte Folded Reload
+; GFX9-NEXT: v_or_b32_sdwa v1, v29, v44 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
; GFX9-NEXT: v_and_b32_e32 v0, 0xffff, v0
; GFX9-NEXT: v_lshl_or_b32 v28, v1, 16, v0
-; GFX9-NEXT: s_waitcnt vmcnt(1)
+; GFX9-NEXT: s_waitcnt vmcnt(0)
; GFX9-NEXT: v_or_b32_sdwa v0, v59, v60 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; GFX9-NEXT: buffer_load_dword v59, off, s[0:3], s32 offset:556 ; 4-byte Folded Reload
; GFX9-NEXT: v_and_b32_e32 v0, 0xffff, v0
; GFX9-NEXT: s_waitcnt vmcnt(0)
-; GFX9-NEXT: v_or_b32_sdwa v1, v63, v57 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; GFX9-NEXT: v_or_b32_sdwa v1, v63, v59 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; GFX9-NEXT: v_mov_b32_e32 v57, v59
; GFX9-NEXT: v_lshl_or_b32 v29, v1, 16, v0
; GFX9-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:508 ; 4-byte Folded Reload
; GFX9-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:512 ; 4-byte Folded Reload
; GFX9-NEXT: s_waitcnt vmcnt(1)
-; GFX9-NEXT: v_or_b32_sdwa v0, v62, v0 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
-; GFX9-NEXT: v_and_b32_e32 v0, 0xffff, v0
+; GFX9-NEXT: v_or_b32_sdwa v0, v58, v0 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
; GFX9-NEXT: s_waitcnt vmcnt(0)
-; GFX9-NEXT: v_or_b32_sdwa v1, v40, v1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; GFX9-NEXT: v_or_b32_sdwa v1, v32, v1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; GFX9-NEXT: v_and_b32_e32 v0, 0xffff, v0
; GFX9-NEXT: v_lshl_or_b32 v30, v1, 16, v0
; GFX9-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:520 ; 4-byte Folded Reload
; GFX9-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:524 ; 4-byte Folded Reload
; GFX9-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:516 ; 4-byte Folded Reload
; GFX9-NEXT: s_waitcnt vmcnt(2)
-; GFX9-NEXT: v_or_b32_sdwa v0, v58, v0 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
-; GFX9-NEXT: v_and_b32_e32 v0, 0xffff, v0
+; GFX9-NEXT: v_or_b32_sdwa v0, v56, v0 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
; GFX9-NEXT: s_waitcnt vmcnt(0)
; GFX9-NEXT: v_or_b32_sdwa v1, v1, v2 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; GFX9-NEXT: v_mov_b32_e32 v2, s6
+; GFX9-NEXT: v_mov_b32_e32 v56, v42
+; GFX9-NEXT: v_and_b32_e32 v0, 0xffff, v0
; GFX9-NEXT: v_lshl_or_b32 v31, v1, 16, v0
; GFX9-NEXT: v_mov_b32_e32 v0, s4
; GFX9-NEXT: v_mov_b32_e32 v1, s5
-; GFX9-NEXT: v_mov_b32_e32 v2, s6
; GFX9-NEXT: s_mov_b64 s[4:5], 0
; GFX9-NEXT: s_branch .LBB89_3
; GFX9-NEXT: .LBB89_2:
; GFX9-NEXT: buffer_load_dword v55, off, s[0:3], s32 offset:616 ; 4-byte Folded Reload
; GFX9-NEXT: buffer_load_dword v53, off, s[0:3], s32 offset:612 ; 4-byte Folded Reload
; GFX9-NEXT: buffer_load_dword v52, off, s[0:3], s32 offset:608 ; 4-byte Folded Reload
-; GFX9-NEXT: buffer_load_dword v50, off, s[0:3], s32 offset:600 ; 4-byte Folded Reload
-; GFX9-NEXT: buffer_load_dword v49, off, s[0:3], s32 offset:604 ; 4-byte Folded Reload
+; GFX9-NEXT: buffer_load_dword v50, off, s[0:3], s32 offset:604 ; 4-byte Folded Reload
+; GFX9-NEXT: buffer_load_dword v49, off, s[0:3], s32 offset:600 ; 4-byte Folded Reload
; GFX9-NEXT: buffer_load_dword v48, off, s[0:3], s32 offset:596 ; 4-byte Folded Reload
-; GFX9-NEXT: buffer_load_dword v39, off, s[0:3], s32 offset:592 ; 4-byte Folded Reload
+; GFX9-NEXT: buffer_load_dword v39, off, s[0:3], s32 offset:588 ; 4-byte Folded Reload
; GFX9-NEXT: buffer_load_dword v38, off, s[0:3], s32 offset:580 ; 4-byte Folded Reload
-; GFX9-NEXT: buffer_load_dword v46, off, s[0:3], s32 offset:588 ; 4-byte Folded Reload
+; GFX9-NEXT: buffer_load_dword v40, off, s[0:3], s32 offset:592 ; 4-byte Folded Reload
; GFX9-NEXT: buffer_load_dword v34, off, s[0:3], s32 offset:584 ; 4-byte Folded Reload
-; GFX9-NEXT: v_mov_b32_e32 v33, v45
-; GFX9-NEXT: buffer_load_dword v35, off, s[0:3], s32 offset:536 ; 4-byte Folded Reload
+; GFX9-NEXT: v_mov_b32_e32 v33, v46
+; GFX9-NEXT: buffer_load_dword v46, off, s[0:3], s32 offset:536 ; 4-byte Folded Reload
; GFX9-NEXT: buffer_load_dword v36, off, s[0:3], s32 offset:544 ; 4-byte Folded Reload
; GFX9-NEXT: buffer_load_dword v37, off, s[0:3], s32 offset:540 ; 4-byte Folded Reload
-; GFX9-NEXT: buffer_load_dword v51, off, s[0:3], s32 offset:564 ; 4-byte Folded Reload
-; GFX9-NEXT: buffer_load_dword v54, off, s[0:3], s32 offset:572 ; 4-byte Folded Reload
-; GFX9-NEXT: buffer_load_dword v41, off, s[0:3], s32 offset:556 ; 4-byte Folded Reload
-; GFX9-NEXT: buffer_load_dword v44, off, s[0:3], s32 offset:568 ; 4-byte Folded Reload
-; GFX9-NEXT: buffer_load_dword v32, off, s[0:3], s32 offset:552 ; 4-byte Folded Reload
-; GFX9-NEXT: buffer_load_dword v57, off, s[0:3], s32 offset:560 ; 4-byte Folded Reload
+; GFX9-NEXT: buffer_load_dword v51, off, s[0:3], s32 offset:560 ; 4-byte Folded Reload
+; GFX9-NEXT: buffer_load_dword v54, off, s[0:3], s32 offset:568 ; 4-byte Folded Reload
+; GFX9-NEXT: buffer_load_dword v41, off, s[0:3], s32 offset:572 ; 4-byte Folded Reload
+; GFX9-NEXT: buffer_load_dword v44, off, s[0:3], s32 offset:564 ; 4-byte Folded Reload
+; GFX9-NEXT: buffer_load_dword v45, off, s[0:3], s32 offset:552 ; 4-byte Folded Reload
+; GFX9-NEXT: v_mov_b32_e32 v56, v61
+; GFX9-NEXT: buffer_load_dword v57, off, s[0:3], s32 offset:556 ; 4-byte Folded Reload
; GFX9-NEXT: buffer_load_dword v60, off, s[0:3], s32 offset:548 ; 4-byte Folded Reload
-; GFX9-NEXT: buffer_load_dword v45, off, s[0:3], s32 offset:576 ; 4-byte Folded Reload
+; GFX9-NEXT: buffer_load_dword v61, off, s[0:3], s32 offset:576 ; 4-byte Folded Reload
; GFX9-NEXT: s_mov_b64 s[4:5], -1
; GFX9-NEXT: ; implicit-def: $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7_vgpr8_vgpr9_vgpr10_vgpr11_vgpr12_vgpr13_vgpr14_vgpr15
; GFX9-NEXT: ; implicit-def: $vgpr16_vgpr17_vgpr18_vgpr19_vgpr20_vgpr21_vgpr22_vgpr23_vgpr24_vgpr25_vgpr26_vgpr27_vgpr28_vgpr29_vgpr30_vgpr31
; GFX9-NEXT: .LBB89_3: ; %Flow
; GFX9-NEXT: buffer_load_dword v42, off, s[0:3], s32 offset:412 ; 4-byte Folded Reload
-; GFX9-NEXT: buffer_load_dword v40, off, s[0:3], s32 offset:408 ; 4-byte Folded Reload
+; GFX9-NEXT: buffer_load_dword v32, off, s[0:3], s32 offset:408 ; 4-byte Folded Reload
; GFX9-NEXT: buffer_load_dword v58, off, s[0:3], s32 offset:404 ; 4-byte Folded Reload
; GFX9-NEXT: buffer_load_dword v63, off, s[0:3], s32 offset:400 ; 4-byte Folded Reload
; GFX9-NEXT: s_andn2_b64 vcc, exec, s[4:5]
@@ -160603,7 +160656,7 @@ define inreg <64 x bfloat> @bitcast_v128i8_to_v64bf16_scalar(<128 x i8> inreg %a
; GFX9-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:816 ; 4-byte Folded Reload
; GFX9-NEXT: s_waitcnt vmcnt(1)
; GFX9-NEXT: v_add_u32_e32 v0, 3, v0
-; GFX9-NEXT: v_or_b32_sdwa v0, v40, v0 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:BYTE_0
+; GFX9-NEXT: v_or_b32_sdwa v0, v32, v0 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:BYTE_0
; GFX9-NEXT: s_waitcnt vmcnt(0)
; GFX9-NEXT: v_add_u32_e32 v1, 3, v1
; GFX9-NEXT: v_or_b32_sdwa v1, v2, v1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:BYTE_0
@@ -160663,11 +160716,11 @@ define inreg <64 x bfloat> @bitcast_v128i8_to_v64bf16_scalar(<128 x i8> inreg %a
; GFX9-NEXT: v_and_b32_e32 v0, 0xffff, v0
; GFX9-NEXT: s_waitcnt vmcnt(0)
; GFX9-NEXT: v_add_u32_e32 v26, 3, v26
-; GFX9-NEXT: v_or_b32_sdwa v48, v46, v26 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:BYTE_0
+; GFX9-NEXT: v_or_b32_sdwa v48, v40, v26 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:BYTE_0
; GFX9-NEXT: buffer_load_dword v26, off, s[0:3], s32 offset:688 ; 4-byte Folded Reload
; GFX9-NEXT: s_waitcnt vmcnt(0)
; GFX9-NEXT: v_add_u32_e32 v26, 3, v26
-; GFX9-NEXT: v_or_b32_sdwa v49, v35, v26 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:BYTE_0
+; GFX9-NEXT: v_or_b32_sdwa v49, v46, v26 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:BYTE_0
; GFX9-NEXT: buffer_load_dword v26, off, s[0:3], s32 offset:676 ; 4-byte Folded Reload
; GFX9-NEXT: s_waitcnt vmcnt(0)
; GFX9-NEXT: v_add_u32_e32 v26, 3, v26
@@ -160702,7 +160755,7 @@ define inreg <64 x bfloat> @bitcast_v128i8_to_v64bf16_scalar(<128 x i8> inreg %a
; GFX9-NEXT: v_or_b32_sdwa v53, v27, v26 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:BYTE_0
; GFX9-NEXT: buffer_load_dword v27, off, s[0:3], s32 offset:480 ; 4-byte Folded Reload
; GFX9-NEXT: v_add_u32_e32 v24, 3, v24
-; GFX9-NEXT: v_add_u32_e32 v26, 3, v61
+; GFX9-NEXT: v_add_u32_e32 v26, 3, v62
; GFX9-NEXT: v_or_b32_sdwa v24, v54, v24 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:BYTE_0
; GFX9-NEXT: v_add_u32_e32 v36, 0x300, v24
; GFX9-NEXT: v_add_u32_e32 v24, 0x300, v48
@@ -160711,7 +160764,7 @@ define inreg <64 x bfloat> @bitcast_v128i8_to_v64bf16_scalar(<128 x i8> inreg %a
; GFX9-NEXT: s_waitcnt vmcnt(0)
; GFX9-NEXT: v_or_b32_sdwa v54, v27, v26 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:BYTE_0
; GFX9-NEXT: buffer_load_dword v27, off, s[0:3], s32 offset:476 ; 4-byte Folded Reload
-; GFX9-NEXT: v_add_u32_e32 v26, 3, v45
+; GFX9-NEXT: v_add_u32_e32 v26, 3, v61
; GFX9-NEXT: v_add_u32_e32 v20, 3, v20
; GFX9-NEXT: v_or_b32_sdwa v20, v57, v20 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:BYTE_0
; GFX9-NEXT: v_add_u32_e32 v34, 0x300, v20
@@ -160720,7 +160773,7 @@ define inreg <64 x bfloat> @bitcast_v128i8_to_v64bf16_scalar(<128 x i8> inreg %a
; GFX9-NEXT: buffer_load_dword v27, off, s[0:3], s32 offset:472 ; 4-byte Folded Reload
; GFX9-NEXT: v_add_u32_e32 v26, 3, v56
; GFX9-NEXT: v_add_u32_e32 v21, 3, v21
-; GFX9-NEXT: v_or_b32_sdwa v21, v32, v21 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:BYTE_0
+; GFX9-NEXT: v_or_b32_sdwa v21, v45, v21 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:BYTE_0
; GFX9-NEXT: v_add_u32_e32 v28, 0x300, v21
; GFX9-NEXT: v_add_u32_e32 v21, 0x300, v54
; GFX9-NEXT: v_and_b32_e32 v21, 0xffff, v21
@@ -162564,29 +162617,30 @@ define <128 x i8> @bitcast_v64bf16_to_v128i8(<64 x bfloat> %a, i32 %b) {
; SI-NEXT: buffer_store_dword v63, off, s[0:3], s32 offset:140 ; 4-byte Folded Spill
; SI-NEXT: buffer_load_dword v31, off, s[0:3], s32 offset:136
; SI-NEXT: buffer_load_dword v32, off, s[0:3], s32 offset:4
-; SI-NEXT: buffer_load_dword v34, off, s[0:3], s32
-; SI-NEXT: buffer_load_dword v35, off, s[0:3], s32 offset:12
-; SI-NEXT: buffer_load_dword v36, off, s[0:3], s32 offset:8
-; SI-NEXT: buffer_load_dword v37, off, s[0:3], s32 offset:20
-; SI-NEXT: buffer_load_dword v38, off, s[0:3], s32 offset:16
-; SI-NEXT: buffer_load_dword v39, off, s[0:3], s32 offset:28
-; SI-NEXT: buffer_load_dword v48, off, s[0:3], s32 offset:24
-; SI-NEXT: buffer_load_dword v49, off, s[0:3], s32 offset:36
+; SI-NEXT: buffer_load_dword v33, off, s[0:3], s32
+; SI-NEXT: buffer_load_dword v34, off, s[0:3], s32 offset:12
+; SI-NEXT: buffer_load_dword v35, off, s[0:3], s32 offset:8
+; SI-NEXT: buffer_load_dword v36, off, s[0:3], s32 offset:20
+; SI-NEXT: buffer_load_dword v37, off, s[0:3], s32 offset:16
+; SI-NEXT: buffer_load_dword v38, off, s[0:3], s32 offset:28
+; SI-NEXT: buffer_load_dword v39, off, s[0:3], s32 offset:24
+; SI-NEXT: buffer_load_dword v48, off, s[0:3], s32 offset:36
; SI-NEXT: buffer_load_dword v50, off, s[0:3], s32 offset:32
-; SI-NEXT: buffer_load_dword v51, off, s[0:3], s32 offset:44
-; SI-NEXT: buffer_load_dword v52, off, s[0:3], s32 offset:40
-; SI-NEXT: buffer_load_dword v53, off, s[0:3], s32 offset:52
-; SI-NEXT: buffer_load_dword v54, off, s[0:3], s32 offset:48
-; SI-NEXT: buffer_load_dword v55, off, s[0:3], s32 offset:60
-; SI-NEXT: buffer_load_dword v40, off, s[0:3], s32 offset:56
-; SI-NEXT: buffer_load_dword v41, off, s[0:3], s32 offset:68
-; SI-NEXT: buffer_load_dword v43, off, s[0:3], s32 offset:64
-; SI-NEXT: buffer_load_dword v44, off, s[0:3], s32 offset:76
-; SI-NEXT: buffer_load_dword v56, off, s[0:3], s32 offset:72
+; SI-NEXT: buffer_load_dword v52, off, s[0:3], s32 offset:44
+; SI-NEXT: buffer_load_dword v54, off, s[0:3], s32 offset:40
+; SI-NEXT: buffer_load_dword v55, off, s[0:3], s32 offset:52
+; SI-NEXT: buffer_load_dword v40, off, s[0:3], s32 offset:48
+; SI-NEXT: buffer_load_dword v41, off, s[0:3], s32 offset:60
+; SI-NEXT: buffer_load_dword v42, off, s[0:3], s32 offset:56
+; SI-NEXT: buffer_load_dword v43, off, s[0:3], s32 offset:68
+; SI-NEXT: buffer_load_dword v44, off, s[0:3], s32 offset:64
+; SI-NEXT: buffer_load_dword v45, off, s[0:3], s32 offset:76
+; SI-NEXT: buffer_load_dword v47, off, s[0:3], s32 offset:72
+; SI-NEXT: s_waitcnt expcnt(3)
+; SI-NEXT: buffer_load_dword v60, off, s[0:3], s32 offset:84
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: buffer_load_dword v63, off, s[0:3], s32 offset:84
-; SI-NEXT: buffer_load_dword v33, off, s[0:3], s32 offset:80
-; SI-NEXT: buffer_load_dword v58, off, s[0:3], s32 offset:92
+; SI-NEXT: buffer_load_dword v63, off, s[0:3], s32 offset:80
+; SI-NEXT: buffer_load_dword v61, off, s[0:3], s32 offset:92
; SI-NEXT: v_mul_f32_e32 v1, 1.0, v1
; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:604 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
@@ -162675,31 +162729,31 @@ define <128 x i8> @bitcast_v64bf16_to_v128i8(<64 x bfloat> %a, i32 %b) {
; SI-NEXT: v_mul_f32_e32 v1, 1.0, v32
; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:616 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_mul_f32_e32 v1, 1.0, v34
+; SI-NEXT: v_mul_f32_e32 v1, 1.0, v33
; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:668 ; 4-byte Folded Spill
; SI-NEXT: buffer_store_dword v2, off, s[0:3], s32 offset:600 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(1)
-; SI-NEXT: v_mul_f32_e32 v1, 1.0, v38
+; SI-NEXT: v_mul_f32_e32 v1, 1.0, v37
; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:680 ; 4-byte Folded Spill
-; SI-NEXT: v_mul_f32_e32 v61, 1.0, v36
+; SI-NEXT: v_mul_f32_e32 v51, 1.0, v39
; SI-NEXT: v_cmp_ne_u32_e32 vcc, 0, v31
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_mul_f32_e32 v1, 1.0, v50
; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:684 ; 4-byte Folded Spill
-; SI-NEXT: v_mul_f32_e32 v57, 1.0, v29
-; SI-NEXT: v_mul_f32_e32 v42, 1.0, v35
+; SI-NEXT: v_mul_f32_e32 v16, 1.0, v29
+; SI-NEXT: v_mul_f32_e32 v24, 1.0, v55
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_mul_f32_e32 v1, 1.0, v54
+; SI-NEXT: v_mul_f32_e32 v1, 1.0, v40
; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:688 ; 4-byte Folded Spill
-; SI-NEXT: v_mul_f32_e32 v62, 1.0, v37
-; SI-NEXT: v_mul_f32_e32 v6, 1.0, v41
+; SI-NEXT: v_mul_f32_e32 v55, 1.0, v41
+; SI-NEXT: v_mul_f32_e32 v56, 1.0, v43
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_mul_f32_e32 v1, 1.0, v43
+; SI-NEXT: v_mul_f32_e32 v1, 1.0, v44
; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:692 ; 4-byte Folded Spill
-; SI-NEXT: v_mul_f32_e32 v9, 1.0, v56
-; SI-NEXT: v_mul_f32_e32 v59, 1.0, v39
+; SI-NEXT: v_mul_f32_e32 v9, 1.0, v47
+; SI-NEXT: v_mul_f32_e32 v62, 1.0, v60
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_mul_f32_e32 v1, 1.0, v33
+; SI-NEXT: v_mul_f32_e32 v1, 1.0, v63
; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:704 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:88
@@ -162710,21 +162764,21 @@ define <128 x i8> @bitcast_v64bf16_to_v128i8(<64 x bfloat> %a, i32 %b) {
; SI-NEXT: buffer_load_dword v7, off, s[0:3], s32 offset:116
; SI-NEXT: buffer_load_dword v8, off, s[0:3], s32 offset:112
; SI-NEXT: buffer_load_dword v10, off, s[0:3], s32 offset:124
-; SI-NEXT: v_mul_f32_e32 v56, 1.0, v58
-; SI-NEXT: ; implicit-def: $vgpr58
-; SI-NEXT: ; kill: killed $vgpr58
-; SI-NEXT: ; implicit-def: $vgpr58
-; SI-NEXT: v_mul_f32_e32 v60, 1.0, v48
-; SI-NEXT: v_mul_f32_e32 v45, 1.0, v49
-; SI-NEXT: v_mul_f32_e32 v49, 1.0, v51
-; SI-NEXT: v_mul_f32_e32 v51, 1.0, v52
-; SI-NEXT: v_mul_f32_e32 v24, 1.0, v53
-; SI-NEXT: v_mul_f32_e32 v46, 1.0, v55
-; SI-NEXT: v_mul_f32_e32 v53, 1.0, v40
-; SI-NEXT: v_mul_f32_e32 v30, 1.0, v44
-; SI-NEXT: v_mul_f32_e32 v19, 1.0, v63
-; SI-NEXT: ; kill: killed $vgpr58
-; SI-NEXT: ; implicit-def: $vgpr58
+; SI-NEXT: v_mul_f32_e32 v6, 1.0, v42
+; SI-NEXT: v_mul_f32_e32 v41, 1.0, v61
+; SI-NEXT: ; implicit-def: $vgpr61
+; SI-NEXT: ; kill: killed $vgpr61
+; SI-NEXT: ; implicit-def: $vgpr61
+; SI-NEXT: v_mul_f32_e32 v58, 1.0, v34
+; SI-NEXT: v_mul_f32_e32 v59, 1.0, v35
+; SI-NEXT: v_mul_f32_e32 v12, 1.0, v36
+; SI-NEXT: v_mul_f32_e32 v49, 1.0, v38
+; SI-NEXT: v_mul_f32_e32 v57, 1.0, v48
+; SI-NEXT: v_mul_f32_e32 v53, 1.0, v52
+; SI-NEXT: v_mul_f32_e32 v46, 1.0, v54
+; SI-NEXT: v_mul_f32_e32 v33, 1.0, v45
+; SI-NEXT: ; kill: killed $vgpr61
+; SI-NEXT: ; implicit-def: $vgpr61
; SI-NEXT: ; implicit-def: $vgpr38
; SI-NEXT: ; implicit-def: $vgpr40
; SI-NEXT: ; implicit-def: $vgpr35
@@ -162748,30 +162802,30 @@ define <128 x i8> @bitcast_v64bf16_to_v128i8(<64 x bfloat> %a, i32 %b) {
; SI-NEXT: ; implicit-def: $vgpr22
; SI-NEXT: ; implicit-def: $vgpr20
; SI-NEXT: ; implicit-def: $vgpr17
-; SI-NEXT: ; implicit-def: $vgpr47
+; SI-NEXT: ; implicit-def: $vgpr27
; SI-NEXT: ; implicit-def: $vgpr14
-; SI-NEXT: ; implicit-def: $vgpr33
-; SI-NEXT: ; kill: killed $vgpr58
-; SI-NEXT: ; implicit-def: $vgpr58
+; SI-NEXT: ; implicit-def: $vgpr63
+; SI-NEXT: ; kill: killed $vgpr61
+; SI-NEXT: ; implicit-def: $vgpr61
; SI-NEXT: s_waitcnt vmcnt(7)
-; SI-NEXT: v_mul_f32_e32 v41, 1.0, v1
+; SI-NEXT: v_mul_f32_e32 v43, 1.0, v1
; SI-NEXT: s_waitcnt vmcnt(6)
-; SI-NEXT: v_mul_f32_e32 v12, 1.0, v2
+; SI-NEXT: v_mul_f32_e32 v47, 1.0, v2
; SI-NEXT: s_waitcnt vmcnt(5)
-; SI-NEXT: v_mul_f32_e32 v27, 1.0, v3
+; SI-NEXT: v_mul_f32_e32 v39, 1.0, v3
; SI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:120
; SI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:132
; SI-NEXT: buffer_load_dword v3, off, s[0:3], s32 offset:128
; SI-NEXT: s_waitcnt vmcnt(7)
-; SI-NEXT: v_mul_f32_e32 v63, 1.0, v4
+; SI-NEXT: v_mul_f32_e32 v19, 1.0, v4
; SI-NEXT: s_waitcnt vmcnt(5)
-; SI-NEXT: v_mul_f32_e32 v36, 1.0, v7
+; SI-NEXT: v_mul_f32_e32 v42, 1.0, v7
; SI-NEXT: ; implicit-def: $vgpr7
; SI-NEXT: ; kill: killed $vgpr7
; SI-NEXT: ; implicit-def: $vgpr7
-; SI-NEXT: v_mul_f32_e32 v16, 1.0, v5
+; SI-NEXT: v_mul_f32_e32 v36, 1.0, v5
; SI-NEXT: s_waitcnt vmcnt(4)
-; SI-NEXT: v_mul_f32_e32 v39, 1.0, v8
+; SI-NEXT: v_mul_f32_e32 v44, 1.0, v8
; SI-NEXT: s_waitcnt vmcnt(3)
; SI-NEXT: v_mul_f32_e32 v21, 1.0, v10
; SI-NEXT: ; kill: killed $vgpr7
@@ -162783,7 +162837,7 @@ define <128 x i8> @bitcast_v64bf16_to_v128i8(<64 x bfloat> %a, i32 %b) {
; SI-NEXT: ; kill: killed $vgpr7
; SI-NEXT: ; implicit-def: $vgpr7
; SI-NEXT: s_waitcnt vmcnt(2)
-; SI-NEXT: v_mul_f32_e32 v43, 1.0, v1
+; SI-NEXT: v_mul_f32_e32 v60, 1.0, v1
; SI-NEXT: ; implicit-def: $vgpr1
; SI-NEXT: ; kill: killed $vgpr1
; SI-NEXT: ; implicit-def: $vgpr1
@@ -162956,9 +163010,9 @@ define <128 x i8> @bitcast_v64bf16_to_v128i8(<64 x bfloat> %a, i32 %b) {
; SI-NEXT: ; kill: killed $vgpr1
; SI-NEXT: ; implicit-def: $vgpr1
; SI-NEXT: s_waitcnt vmcnt(1)
-; SI-NEXT: v_mul_f32_e32 v44, 1.0, v2
+; SI-NEXT: v_mul_f32_e32 v30, 1.0, v2
; SI-NEXT: s_waitcnt vmcnt(0)
-; SI-NEXT: v_mul_f32_e32 v55, 1.0, v3
+; SI-NEXT: v_mul_f32_e32 v45, 1.0, v3
; SI-NEXT: ; kill: killed $vgpr1
; SI-NEXT: ; implicit-def: $vgpr1
; SI-NEXT: ; implicit-def: $vgpr3
@@ -162971,10 +163025,10 @@ define <128 x i8> @bitcast_v64bf16_to_v128i8(<64 x bfloat> %a, i32 %b) {
; SI-NEXT: ; %bb.1: ; %cmp.false
; SI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:600 ; 4-byte Folded Reload
; SI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:604 ; 4-byte Folded Reload
-; SI-NEXT: buffer_load_dword v58, off, s[0:3], s32 offset:264 ; 4-byte Folded Reload
-; SI-NEXT: buffer_load_dword v33, off, s[0:3], s32 offset:272 ; 4-byte Folded Reload
-; SI-NEXT: buffer_load_dword v7, off, s[0:3], s32 offset:688 ; 4-byte Folded Reload
-; SI-NEXT: buffer_load_dword v47, off, s[0:3], s32 offset:280 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v63, off, s[0:3], s32 offset:264 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v61, off, s[0:3], s32 offset:272 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v7, off, s[0:3], s32 offset:692 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v27, off, s[0:3], s32 offset:280 ; 4-byte Folded Reload
; SI-NEXT: s_waitcnt vmcnt(5)
; SI-NEXT: v_lshrrev_b32_e32 v1, 16, v1
; SI-NEXT: s_waitcnt vmcnt(4)
@@ -163004,7 +163058,7 @@ define <128 x i8> @bitcast_v64bf16_to_v128i8(<64 x bfloat> %a, i32 %b) {
; SI-NEXT: s_waitcnt vmcnt(0)
; SI-NEXT: v_alignbit_b32 v26, v1, v2, 16
; SI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:436 ; 4-byte Folded Reload
-; SI-NEXT: v_lshrrev_b32_e32 v1, 16, v58
+; SI-NEXT: v_lshrrev_b32_e32 v1, 16, v63
; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:256 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt vmcnt(1)
; SI-NEXT: v_alignbit_b32 v40, v1, v2, 16
@@ -163016,7 +163070,7 @@ define <128 x i8> @bitcast_v64bf16_to_v128i8(<64 x bfloat> %a, i32 %b) {
; SI-NEXT: s_waitcnt vmcnt(0)
; SI-NEXT: v_alignbit_b32 v23, v1, v2, 16
; SI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:448 ; 4-byte Folded Reload
-; SI-NEXT: v_lshrrev_b32_e32 v1, 16, v33
+; SI-NEXT: v_lshrrev_b32_e32 v1, 16, v61
; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:252 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt vmcnt(1)
; SI-NEXT: v_alignbit_b32 v54, v1, v2, 16
@@ -163028,7 +163082,7 @@ define <128 x i8> @bitcast_v64bf16_to_v128i8(<64 x bfloat> %a, i32 %b) {
; SI-NEXT: s_waitcnt vmcnt(0)
; SI-NEXT: v_alignbit_b32 v18, v1, v2, 16
; SI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:460 ; 4-byte Folded Reload
-; SI-NEXT: v_lshrrev_b32_e32 v1, 16, v47
+; SI-NEXT: v_lshrrev_b32_e32 v1, 16, v27
; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:248 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt vmcnt(1)
; SI-NEXT: v_alignbit_b32 v52, v1, v2, 16
@@ -163037,88 +163091,148 @@ define <128 x i8> @bitcast_v64bf16_to_v128i8(<64 x bfloat> %a, i32 %b) {
; SI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:468 ; 4-byte Folded Reload
; SI-NEXT: s_waitcnt vmcnt(1)
; SI-NEXT: v_lshrrev_b32_e32 v1, 16, v1
-; SI-NEXT: v_alignbit_b32 v15, v1, v57, 16
-; SI-NEXT: buffer_load_dword v57, off, s[0:3], s32 offset:284 ; 4-byte Folded Reload
+; SI-NEXT: v_alignbit_b32 v15, v1, v16, 16
+; SI-NEXT: buffer_load_dword v16, off, s[0:3], s32 offset:284 ; 4-byte Folded Reload
; SI-NEXT: s_waitcnt vmcnt(0)
-; SI-NEXT: v_lshrrev_b32_e32 v1, 16, v57
+; SI-NEXT: v_lshrrev_b32_e32 v1, 16, v16
; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:244 ; 4-byte Folded Spill
; SI-NEXT: v_alignbit_b32 v50, v1, v2, 16
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_lshrrev_b32_e32 v1, 16, v42
-; SI-NEXT: buffer_load_dword v42, off, s[0:3], s32 offset:348 ; 4-byte Folded Reload
+; SI-NEXT: v_lshrrev_b32_e32 v1, 16, v58
+; SI-NEXT: buffer_load_dword v58, off, s[0:3], s32 offset:348 ; 4-byte Folded Reload
; SI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:584 ; 4-byte Folded Reload
-; SI-NEXT: v_alignbit_b32 v13, v1, v61, 16
-; SI-NEXT: ; implicit-def: $vgpr61
+; SI-NEXT: v_alignbit_b32 v13, v1, v59, 16
+; SI-NEXT: ; implicit-def: $vgpr59
; SI-NEXT: s_waitcnt vmcnt(1)
-; SI-NEXT: v_lshrrev_b32_e32 v1, 16, v42
+; SI-NEXT: v_lshrrev_b32_e32 v1, 16, v58
; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:240 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt vmcnt(1)
; SI-NEXT: v_alignbit_b32 v48, v1, v2, 16
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_lshrrev_b32_e32 v1, 16, v59
-; SI-NEXT: buffer_load_dword v59, off, s[0:3], s32 offset:528 ; 4-byte Folded Reload
+; SI-NEXT: v_lshrrev_b32_e32 v1, 16, v49
+; SI-NEXT: buffer_load_dword v49, off, s[0:3], s32 offset:528 ; 4-byte Folded Reload
; SI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:644 ; 4-byte Folded Reload
-; SI-NEXT: v_alignbit_b32 v11, v1, v60, 16
-; SI-NEXT: ; implicit-def: $vgpr60
-; SI-NEXT: s_waitcnt vmcnt(1)
-; SI-NEXT: v_lshrrev_b32_e32 v1, 16, v59
-; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:236 ; 4-byte Folded Spill
+; SI-NEXT: v_alignbit_b32 v11, v1, v51, 16
+; SI-NEXT: buffer_load_dword v51, off, s[0:3], s32 offset:588 ; 4-byte Folded Reload
+; SI-NEXT: s_waitcnt vmcnt(2)
+; SI-NEXT: v_lshrrev_b32_e32 v1, 16, v49
; SI-NEXT: s_waitcnt vmcnt(1)
; SI-NEXT: v_alignbit_b32 v37, v1, v2, 16
-; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_lshrrev_b32_e32 v1, 16, v49
-; SI-NEXT: buffer_load_dword v49, off, s[0:3], s32 offset:588 ; 4-byte Folded Reload
; SI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:656 ; 4-byte Folded Reload
-; SI-NEXT: v_alignbit_b32 v8, v1, v51, 16
-; SI-NEXT: buffer_load_dword v51, off, s[0:3], s32 offset:616 ; 4-byte Folded Reload
+; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:236 ; 4-byte Folded Spill
+; SI-NEXT: s_waitcnt expcnt(0)
+; SI-NEXT: v_lshrrev_b32_e32 v1, 16, v53
+; SI-NEXT: v_alignbit_b32 v8, v1, v46, 16
; SI-NEXT: s_waitcnt vmcnt(2)
-; SI-NEXT: v_lshrrev_b32_e32 v1, 16, v49
-; SI-NEXT: s_waitcnt vmcnt(1)
+; SI-NEXT: v_lshrrev_b32_e32 v1, 16, v51
+; SI-NEXT: buffer_load_dword v53, off, s[0:3], s32 offset:616 ; 4-byte Folded Reload
+; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:232 ; 4-byte Folded Spill
+; SI-NEXT: ; implicit-def: $vgpr46
+; SI-NEXT: s_waitcnt vmcnt(3)
; SI-NEXT: v_alignbit_b32 v34, v1, v2, 16
; SI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:668 ; 4-byte Folded Reload
-; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:232 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_lshrrev_b32_e32 v1, 16, v46
-; SI-NEXT: v_alignbit_b32 v5, v1, v53, 16
-; SI-NEXT: s_waitcnt vmcnt(2)
-; SI-NEXT: v_lshrrev_b32_e32 v1, 16, v51
+; SI-NEXT: v_lshrrev_b32_e32 v1, 16, v55
+; SI-NEXT: v_alignbit_b32 v5, v1, v6, 16
+; SI-NEXT: buffer_load_dword v6, off, s[0:3], s32 offset:688 ; 4-byte Folded Reload
+; SI-NEXT: s_waitcnt vmcnt(3)
+; SI-NEXT: v_lshrrev_b32_e32 v1, 16, v53
; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:228 ; 4-byte Folded Spill
-; SI-NEXT: ; implicit-def: $vgpr46
-; SI-NEXT: ; implicit-def: $vgpr53
+; SI-NEXT: ; implicit-def: $vgpr55
; SI-NEXT: s_waitcnt vmcnt(2)
; SI-NEXT: v_alignbit_b32 v31, v1, v2, 16
; SI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:680 ; 4-byte Folded Reload
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_lshrrev_b32_e32 v1, 16, v30
+; SI-NEXT: v_lshrrev_b32_e32 v1, 16, v33
; SI-NEXT: v_alignbit_b32 v4, v1, v9, 16
-; SI-NEXT: v_lshrrev_b32_e32 v1, 16, v62
-; SI-NEXT: buffer_load_dword v9, off, s[0:3], s32 offset:692 ; 4-byte Folded Reload
+; SI-NEXT: v_lshrrev_b32_e32 v1, 16, v12
; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:224 ; 4-byte Folded Spill
-; SI-NEXT: ; implicit-def: $vgpr30
+; SI-NEXT: v_lshrrev_b32_e32 v9, 24, v63
+; SI-NEXT: buffer_store_dword v9, off, s[0:3], s32 offset:488 ; 4-byte Folded Spill
+; SI-NEXT: ; implicit-def: $vgpr9
+; SI-NEXT: ; kill: killed $vgpr9
+; SI-NEXT: ; implicit-def: $vgpr33
+; SI-NEXT: ; implicit-def: $vgpr9
; SI-NEXT: s_waitcnt vmcnt(2)
; SI-NEXT: v_alignbit_b32 v28, v1, v2, 16
; SI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:684 ; 4-byte Folded Reload
-; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_lshrrev_b32_e32 v1, 16, v56
-; SI-NEXT: v_alignbit_b32 v3, v1, v41, 16
-; SI-NEXT: v_lshrrev_b32_e32 v1, 16, v45
+; SI-NEXT: s_waitcnt expcnt(1)
+; SI-NEXT: v_lshrrev_b32_e32 v1, 16, v41
+; SI-NEXT: v_alignbit_b32 v3, v1, v43, 16
+; SI-NEXT: v_lshrrev_b32_e32 v1, 16, v57
; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:220 ; 4-byte Folded Spill
-; SI-NEXT: ; implicit-def: $vgpr56
; SI-NEXT: ; implicit-def: $vgpr41
+; SI-NEXT: ; implicit-def: $vgpr43
; SI-NEXT: s_waitcnt vmcnt(1)
; SI-NEXT: v_alignbit_b32 v25, v1, v2, 16
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_lshrrev_b32_e32 v1, 16, v63
-; SI-NEXT: v_alignbit_b32 v2, v1, v16, 16
+; SI-NEXT: v_lshrrev_b32_e32 v1, 16, v19
+; SI-NEXT: v_alignbit_b32 v2, v1, v36, 16
; SI-NEXT: v_lshrrev_b32_e32 v1, 16, v24
-; SI-NEXT: v_alignbit_b32 v22, v1, v7, 16
-; SI-NEXT: v_lshrrev_b32_e32 v7, 16, v6
-; SI-NEXT: v_alignbit_b32 v20, v7, v9, 16
-; SI-NEXT: buffer_load_dword v9, off, s[0:3], s32 offset:704 ; 4-byte Folded Reload
-; SI-NEXT: v_lshrrev_b32_e32 v6, 24, v6
+; SI-NEXT: v_alignbit_b32 v22, v1, v6, 16
+; SI-NEXT: v_lshrrev_b32_e32 v6, 16, v56
+; SI-NEXT: v_alignbit_b32 v20, v6, v7, 16
+; SI-NEXT: buffer_load_dword v7, off, s[0:3], s32 offset:704 ; 4-byte Folded Reload
+; SI-NEXT: buffer_store_dword v6, off, s[0:3], s32 offset:212 ; 4-byte Folded Spill
+; SI-NEXT: s_waitcnt expcnt(0)
+; SI-NEXT: v_lshrrev_b32_e32 v6, 16, v62
+; SI-NEXT: buffer_store_dword v6, off, s[0:3], s32 offset:208 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:216 ; 4-byte Folded Spill
+; SI-NEXT: s_waitcnt expcnt(0)
+; SI-NEXT: v_lshrrev_b32_e32 v1, 16, v21
+; SI-NEXT: v_alignbit_b32 v1, v1, v60, 16
+; SI-NEXT: ; implicit-def: $vgpr19
+; SI-NEXT: ; kill: killed $vgpr19
+; SI-NEXT: ; implicit-def: $vgpr19
+; SI-NEXT: ; implicit-def: $vgpr36
+; SI-NEXT: ; implicit-def: $vgpr21
+; SI-NEXT: ; implicit-def: $vgpr60
+; SI-NEXT: s_waitcnt vmcnt(3)
+; SI-NEXT: v_alignbit_b32 v17, v6, v7, 16
+; SI-NEXT: v_lshrrev_b32_e32 v6, 16, v47
+; SI-NEXT: buffer_store_dword v6, off, s[0:3], s32 offset:204 ; 4-byte Folded Spill
+; SI-NEXT: v_alignbit_b32 v14, v6, v39, 16
+; SI-NEXT: s_waitcnt expcnt(0)
+; SI-NEXT: v_lshrrev_b32_e32 v6, 16, v42
+; SI-NEXT: v_alignbit_b32 v10, v6, v44, 16
+; SI-NEXT: v_mov_b32_e32 v63, v6
+; SI-NEXT: v_lshrrev_b32_e32 v6, 24, v61
+; SI-NEXT: buffer_store_dword v6, off, s[0:3], s32 offset:456 ; 4-byte Folded Spill
+; SI-NEXT: s_waitcnt expcnt(0)
+; SI-NEXT: v_lshrrev_b32_e32 v6, 24, v27
+; SI-NEXT: buffer_store_dword v6, off, s[0:3], s32 offset:424 ; 4-byte Folded Spill
+; SI-NEXT: s_waitcnt expcnt(0)
+; SI-NEXT: v_lshrrev_b32_e32 v6, 24, v16
+; SI-NEXT: buffer_store_dword v6, off, s[0:3], s32 offset:400 ; 4-byte Folded Spill
+; SI-NEXT: s_waitcnt expcnt(0)
+; SI-NEXT: v_lshrrev_b32_e32 v6, 24, v58
+; SI-NEXT: buffer_store_dword v6, off, s[0:3], s32 offset:376 ; 4-byte Folded Spill
+; SI-NEXT: s_waitcnt expcnt(0)
+; SI-NEXT: v_lshrrev_b32_e32 v6, 24, v49
+; SI-NEXT: buffer_store_dword v6, off, s[0:3], s32 offset:344 ; 4-byte Folded Spill
+; SI-NEXT: s_waitcnt expcnt(0)
+; SI-NEXT: v_lshrrev_b32_e32 v6, 24, v51
+; SI-NEXT: buffer_store_dword v6, off, s[0:3], s32 offset:320 ; 4-byte Folded Spill
+; SI-NEXT: s_waitcnt expcnt(0)
+; SI-NEXT: v_lshrrev_b32_e32 v6, 24, v53
+; SI-NEXT: buffer_store_dword v6, off, s[0:3], s32 offset:304 ; 4-byte Folded Spill
+; SI-NEXT: s_waitcnt expcnt(0)
+; SI-NEXT: v_lshrrev_b32_e32 v6, 24, v12
+; SI-NEXT: buffer_store_dword v6, off, s[0:3], s32 offset:300 ; 4-byte Folded Spill
+; SI-NEXT: s_waitcnt expcnt(0)
+; SI-NEXT: v_lshrrev_b32_e32 v6, 24, v57
+; SI-NEXT: buffer_store_dword v6, off, s[0:3], s32 offset:296 ; 4-byte Folded Spill
+; SI-NEXT: s_waitcnt expcnt(0)
+; SI-NEXT: v_lshrrev_b32_e32 v6, 24, v24
+; SI-NEXT: buffer_store_dword v6, off, s[0:3], s32 offset:292 ; 4-byte Folded Spill
+; SI-NEXT: s_waitcnt expcnt(0)
+; SI-NEXT: v_lshrrev_b32_e32 v6, 24, v56
; SI-NEXT: buffer_store_dword v6, off, s[0:3], s32 offset:288 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_lshrrev_b32_e32 v6, 24, v44
+; SI-NEXT: v_lshrrev_b32_e32 v6, 24, v42
+; SI-NEXT: buffer_store_dword v6, off, s[0:3], s32 offset:268 ; 4-byte Folded Spill
+; SI-NEXT: s_waitcnt expcnt(0)
+; SI-NEXT: v_lshrrev_b32_e32 v6, 24, v30
; SI-NEXT: buffer_store_dword v6, off, s[0:3], s32 offset:260 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_alignbit_b32 v6, v40, v38, 24
@@ -163221,9 +163335,6 @@ define <128 x i8> @bitcast_v64bf16_to_v128i8(<64 x bfloat> %a, i32 %b) {
; SI-NEXT: buffer_store_dword v6, off, s[0:3], s32 offset:404 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_alignbit_b32 v6, v20, v5, 24
-; SI-NEXT: buffer_store_dword v7, off, s[0:3], s32 offset:212 ; 4-byte Folded Spill
-; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_lshrrev_b32_e32 v7, 16, v19
; SI-NEXT: buffer_store_dword v6, off, s[0:3], s32 offset:368 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_alignbit_b32 v6, v20, v5, 16
@@ -163231,49 +163342,33 @@ define <128 x i8> @bitcast_v64bf16_to_v128i8(<64 x bfloat> %a, i32 %b) {
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_alignbit_b32 v6, v20, v5, 8
; SI-NEXT: buffer_store_dword v6, off, s[0:3], s32 offset:384 ; 4-byte Folded Spill
-; SI-NEXT: buffer_store_dword v7, off, s[0:3], s32 offset:208 ; 4-byte Folded Spill
-; SI-NEXT: s_waitcnt vmcnt(14)
-; SI-NEXT: v_alignbit_b32 v17, v7, v9, 16
-; SI-NEXT: s_waitcnt expcnt(1)
-; SI-NEXT: v_alignbit_b32 v6, v17, v4, 24
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_lshrrev_b32_e32 v7, 16, v12
+; SI-NEXT: v_alignbit_b32 v6, v17, v4, 24
; SI-NEXT: buffer_store_dword v6, off, s[0:3], s32 offset:356 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_alignbit_b32 v6, v17, v4, 16
-; SI-NEXT: v_alignbit_b32 v14, v7, v27, 16
; SI-NEXT: buffer_store_dword v6, off, s[0:3], s32 offset:360 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_alignbit_b32 v6, v17, v4, 8
; SI-NEXT: buffer_store_dword v6, off, s[0:3], s32 offset:364 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_alignbit_b32 v6, v14, v3, 24
-; SI-NEXT: buffer_store_dword v7, off, s[0:3], s32 offset:204 ; 4-byte Folded Spill
-; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_lshrrev_b32_e32 v7, 16, v36
; SI-NEXT: buffer_store_dword v6, off, s[0:3], s32 offset:336 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_alignbit_b32 v6, v14, v3, 16
-; SI-NEXT: v_alignbit_b32 v10, v7, v39, 16
; SI-NEXT: buffer_store_dword v6, off, s[0:3], s32 offset:340 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_alignbit_b32 v6, v14, v3, 8
; SI-NEXT: buffer_store_dword v6, off, s[0:3], s32 offset:352 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_alignbit_b32 v6, v10, v2, 24
-; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:216 ; 4-byte Folded Spill
-; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_lshrrev_b32_e32 v1, 16, v21
-; SI-NEXT: buffer_store_dword v7, off, s[0:3], s32 offset:276 ; 4-byte Folded Spill
-; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_lshrrev_b32_e32 v7, 16, v44
+; SI-NEXT: v_lshrrev_b32_e32 v7, 16, v30
; SI-NEXT: buffer_store_dword v6, off, s[0:3], s32 offset:324 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_alignbit_b32 v6, v10, v2, 16
-; SI-NEXT: v_alignbit_b32 v1, v1, v43, 16
-; SI-NEXT: buffer_store_dword v7, off, s[0:3], s32 offset:268 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v7, off, s[0:3], s32 offset:276 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_alignbit_b32 v7, v7, v55, 16
+; SI-NEXT: v_alignbit_b32 v7, v7, v45, 16
; SI-NEXT: buffer_store_dword v6, off, s[0:3], s32 offset:328 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_alignbit_b32 v6, v10, v2, 8
@@ -163378,173 +163473,135 @@ define <128 x i8> @bitcast_v64bf16_to_v128i8(<64 x bfloat> %a, i32 %b) {
; SI-NEXT: ; implicit-def: $vgpr6
; SI-NEXT: ; kill: killed $vgpr6
; SI-NEXT: ; implicit-def: $vgpr6
-; SI-NEXT: v_lshrrev_b32_e32 v9, 24, v58
; SI-NEXT: ; kill: killed $vgpr6
; SI-NEXT: ; implicit-def: $vgpr6
-; SI-NEXT: buffer_store_dword v9, off, s[0:3], s32 offset:488 ; 4-byte Folded Spill
-; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_lshrrev_b32_e32 v9, 24, v33
; SI-NEXT: ; kill: killed $vgpr6
; SI-NEXT: ; implicit-def: $vgpr6
-; SI-NEXT: buffer_store_dword v9, off, s[0:3], s32 offset:456 ; 4-byte Folded Spill
-; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_lshrrev_b32_e32 v9, 24, v47
; SI-NEXT: ; kill: killed $vgpr6
; SI-NEXT: ; implicit-def: $vgpr6
-; SI-NEXT: buffer_store_dword v9, off, s[0:3], s32 offset:424 ; 4-byte Folded Spill
-; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_lshrrev_b32_e32 v9, 24, v57
; SI-NEXT: ; kill: killed $vgpr6
; SI-NEXT: ; implicit-def: $vgpr6
-; SI-NEXT: buffer_store_dword v9, off, s[0:3], s32 offset:400 ; 4-byte Folded Spill
-; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_lshrrev_b32_e32 v9, 24, v42
; SI-NEXT: ; kill: killed $vgpr6
; SI-NEXT: ; implicit-def: $vgpr6
-; SI-NEXT: buffer_store_dword v9, off, s[0:3], s32 offset:376 ; 4-byte Folded Spill
-; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_lshrrev_b32_e32 v9, 24, v59
; SI-NEXT: ; kill: killed $vgpr6
; SI-NEXT: ; implicit-def: $vgpr6
-; SI-NEXT: buffer_store_dword v9, off, s[0:3], s32 offset:344 ; 4-byte Folded Spill
-; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_lshrrev_b32_e32 v9, 24, v49
; SI-NEXT: ; kill: killed $vgpr6
; SI-NEXT: ; implicit-def: $vgpr6
-; SI-NEXT: buffer_store_dword v9, off, s[0:3], s32 offset:320 ; 4-byte Folded Spill
-; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_lshrrev_b32_e32 v9, 24, v51
; SI-NEXT: ; kill: killed $vgpr6
; SI-NEXT: ; implicit-def: $vgpr6
-; SI-NEXT: buffer_store_dword v9, off, s[0:3], s32 offset:304 ; 4-byte Folded Spill
-; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_lshrrev_b32_e32 v9, 24, v62
; SI-NEXT: ; kill: killed $vgpr6
; SI-NEXT: ; implicit-def: $vgpr6
-; SI-NEXT: buffer_store_dword v9, off, s[0:3], s32 offset:300 ; 4-byte Folded Spill
-; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_lshrrev_b32_e32 v9, 24, v45
; SI-NEXT: ; kill: killed $vgpr6
; SI-NEXT: ; implicit-def: $vgpr6
-; SI-NEXT: buffer_store_dword v9, off, s[0:3], s32 offset:296 ; 4-byte Folded Spill
-; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_lshrrev_b32_e32 v9, 24, v24
; SI-NEXT: ; kill: killed $vgpr6
; SI-NEXT: ; implicit-def: $vgpr6
-; SI-NEXT: buffer_store_dword v9, off, s[0:3], s32 offset:292 ; 4-byte Folded Spill
-; SI-NEXT: v_lshrrev_b32_e32 v33, 24, v12
; SI-NEXT: ; kill: killed $vgpr6
; SI-NEXT: ; implicit-def: $vgpr6
-; SI-NEXT: ; implicit-def: $vgpr9
+; SI-NEXT: v_lshrrev_b32_e32 v27, 24, v62
+; SI-NEXT: v_lshrrev_b32_e32 v61, 24, v47
+; SI-NEXT: ; implicit-def: $vgpr16
+; SI-NEXT: ; implicit-def: $vgpr58
; SI-NEXT: ; implicit-def: $vgpr12
-; SI-NEXT: v_lshrrev_b32_e32 v47, 24, v19
-; SI-NEXT: v_lshrrev_b32_e32 v58, 24, v36
-; SI-NEXT: ; implicit-def: $vgpr57
-; SI-NEXT: ; implicit-def: $vgpr42
-; SI-NEXT: ; implicit-def: $vgpr62
-; SI-NEXT: ; implicit-def: $vgpr59
-; SI-NEXT: ; implicit-def: $vgpr45
; SI-NEXT: ; implicit-def: $vgpr49
; SI-NEXT: ; implicit-def: $vgpr51
+; SI-NEXT: ; implicit-def: $vgpr57
+; SI-NEXT: ; implicit-def: $vgpr53
; SI-NEXT: ; implicit-def: $vgpr24
; SI-NEXT: ; kill: killed $vgpr6
; SI-NEXT: ; implicit-def: $vgpr6
-; SI-NEXT: ; kill: killed $vgpr9
-; SI-NEXT: ; implicit-def: $vgpr9
-; SI-NEXT: ; implicit-def: $vgpr19
-; SI-NEXT: ; kill: killed $vgpr12
-; SI-NEXT: ; implicit-def: $vgpr12
-; SI-NEXT: ; implicit-def: $vgpr27
-; SI-NEXT: ; implicit-def: $vgpr63
-; SI-NEXT: ; implicit-def: $vgpr16
-; SI-NEXT: ; implicit-def: $vgpr36
+; SI-NEXT: ; implicit-def: $vgpr56
+; SI-NEXT: ; implicit-def: $vgpr62
+; SI-NEXT: ; implicit-def: $vgpr47
; SI-NEXT: ; implicit-def: $vgpr39
-; SI-NEXT: ; implicit-def: $vgpr21
-; SI-NEXT: ; implicit-def: $vgpr43
+; SI-NEXT: ; implicit-def: $vgpr42
; SI-NEXT: ; implicit-def: $vgpr44
-; SI-NEXT: ; implicit-def: $vgpr55
+; SI-NEXT: ; implicit-def: $vgpr30
+; SI-NEXT: ; implicit-def: $vgpr45
; SI-NEXT: .LBB90_2: ; %Flow
; SI-NEXT: s_andn2_saveexec_b64 s[4:5], s[4:5]
; SI-NEXT: s_cbranch_execz .LBB90_4
; SI-NEXT: ; %bb.3: ; %cmp.true
+; SI-NEXT: v_and_b32_e32 v11, 0xffff0000, v53
+; SI-NEXT: v_and_b32_e32 v8, 0xffff0000, v46
+; SI-NEXT: v_add_f32_e32 v11, 0x40c00000, v11
+; SI-NEXT: v_add_f32_e32 v8, 0x40c00000, v8
+; SI-NEXT: v_lshrrev_b32_e32 v11, 16, v11
+; SI-NEXT: v_alignbit_b32 v8, v11, v8, 16
+; SI-NEXT: v_and_b32_e32 v11, 0xffff0000, v39
+; SI-NEXT: v_add_f32_e32 v13, 0x40c00000, v11
+; SI-NEXT: v_and_b32_e32 v11, 0xffff0000, v47
+; SI-NEXT: v_and_b32_e32 v14, 0xffff0000, v49
+; SI-NEXT: v_and_b32_e32 v3, 0xffff0000, v19
+; SI-NEXT: v_add_f32_e32 v19, 0x40c00000, v11
+; SI-NEXT: v_and_b32_e32 v11, 0xffff0000, v51
+; SI-NEXT: v_add_f32_e32 v14, 0x40c00000, v14
+; SI-NEXT: v_add_f32_e32 v11, 0x40c00000, v11
+; SI-NEXT: v_lshrrev_b32_e32 v14, 16, v14
+; SI-NEXT: v_alignbit_b32 v11, v14, v11, 16
+; SI-NEXT: v_lshrrev_b32_e32 v14, 16, v19
+; SI-NEXT: buffer_store_dword v14, off, s[0:3], s32 offset:204 ; 4-byte Folded Spill
+; SI-NEXT: s_waitcnt expcnt(0)
+; SI-NEXT: v_alignbit_b32 v14, v14, v13, 16
+; SI-NEXT: buffer_load_dword v13, off, s[0:3], s32 offset:704 ; 4-byte Folded Reload
+; SI-NEXT: v_and_b32_e32 v17, 0xffff0000, v58
; SI-NEXT: v_and_b32_e32 v2, 0xffff0000, v21
-; SI-NEXT: v_and_b32_e32 v1, 0xffff0000, v43
+; SI-NEXT: v_add_f32_e32 v17, 0x40c00000, v17
+; SI-NEXT: v_lshrrev_b32_e32 v17, 16, v17
+; SI-NEXT: v_and_b32_e32 v1, 0xffff0000, v60
; SI-NEXT: v_add_f32_e32 v2, 0x40c00000, v2
; SI-NEXT: v_add_f32_e32 v1, 0x40c00000, v1
; SI-NEXT: v_lshrrev_b32_e32 v2, 16, v2
-; SI-NEXT: v_and_b32_e32 v3, 0xffff0000, v63
; SI-NEXT: v_alignbit_b32 v1, v2, v1, 16
-; SI-NEXT: v_and_b32_e32 v2, 0xffff0000, v16
+; SI-NEXT: v_and_b32_e32 v2, 0xffff0000, v36
; SI-NEXT: v_add_f32_e32 v3, 0x40c00000, v3
; SI-NEXT: v_add_f32_e32 v2, 0x40c00000, v2
; SI-NEXT: v_lshrrev_b32_e32 v3, 16, v3
-; SI-NEXT: v_and_b32_e32 v4, 0xffff0000, v56
+; SI-NEXT: v_and_b32_e32 v4, 0xffff0000, v41
; SI-NEXT: v_alignbit_b32 v2, v3, v2, 16
-; SI-NEXT: v_and_b32_e32 v3, 0xffff0000, v41
+; SI-NEXT: v_and_b32_e32 v3, 0xffff0000, v43
; SI-NEXT: v_add_f32_e32 v4, 0x40c00000, v4
; SI-NEXT: v_add_f32_e32 v3, 0x40c00000, v3
; SI-NEXT: v_lshrrev_b32_e32 v4, 16, v4
-; SI-NEXT: v_and_b32_e32 v5, 0xffff0000, v30
+; SI-NEXT: v_and_b32_e32 v5, 0xffff0000, v33
; SI-NEXT: v_alignbit_b32 v3, v4, v3, 16
; SI-NEXT: v_and_b32_e32 v4, 0xffff0000, v9
; SI-NEXT: v_add_f32_e32 v5, 0x40c00000, v5
; SI-NEXT: v_add_f32_e32 v4, 0x40c00000, v4
; SI-NEXT: v_lshrrev_b32_e32 v5, 16, v5
; SI-NEXT: v_alignbit_b32 v4, v5, v4, 16
-; SI-NEXT: v_and_b32_e32 v5, 0xffff0000, v55
+; SI-NEXT: v_and_b32_e32 v5, 0xffff0000, v45
; SI-NEXT: v_add_f32_e32 v7, 0x40c00000, v5
-; SI-NEXT: v_and_b32_e32 v5, 0xffff0000, v44
-; SI-NEXT: v_and_b32_e32 v8, 0xffff0000, v46
-; SI-NEXT: s_waitcnt expcnt(0)
+; SI-NEXT: v_and_b32_e32 v5, 0xffff0000, v30
+; SI-NEXT: v_and_b32_e32 v12, 0xffff0000, v12
+; SI-NEXT: v_add_f32_e32 v12, 0x40c00000, v12
; SI-NEXT: v_add_f32_e32 v9, 0x40c00000, v5
-; SI-NEXT: v_and_b32_e32 v5, 0xffff0000, v53
-; SI-NEXT: v_add_f32_e32 v8, 0x40c00000, v8
+; SI-NEXT: v_and_b32_e32 v5, 0xffff0000, v6
+; SI-NEXT: v_and_b32_e32 v6, 0xffff0000, v55
+; SI-NEXT: v_add_f32_e32 v6, 0x40c00000, v6
; SI-NEXT: v_add_f32_e32 v5, 0x40c00000, v5
-; SI-NEXT: v_lshrrev_b32_e32 v8, 16, v8
-; SI-NEXT: v_alignbit_b32 v5, v8, v5, 16
-; SI-NEXT: v_lshrrev_b32_e32 v8, 16, v9
-; SI-NEXT: buffer_store_dword v8, off, s[0:3], s32 offset:268 ; 4-byte Folded Spill
-; SI-NEXT: v_alignbit_b32 v7, v8, v7, 16
-; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_and_b32_e32 v8, 0xffff0000, v39
-; SI-NEXT: v_add_f32_e32 v10, 0x40c00000, v8
-; SI-NEXT: v_and_b32_e32 v8, 0xffff0000, v36
-; SI-NEXT: v_and_b32_e32 v11, 0xffff0000, v49
-; SI-NEXT: v_add_f32_e32 v16, 0x40c00000, v8
-; SI-NEXT: v_and_b32_e32 v8, 0xffff0000, v51
-; SI-NEXT: v_add_f32_e32 v11, 0x40c00000, v11
-; SI-NEXT: v_add_f32_e32 v8, 0x40c00000, v8
-; SI-NEXT: v_lshrrev_b32_e32 v11, 16, v11
-; SI-NEXT: v_alignbit_b32 v8, v11, v8, 16
-; SI-NEXT: v_lshrrev_b32_e32 v11, 16, v16
-; SI-NEXT: buffer_store_dword v11, off, s[0:3], s32 offset:276 ; 4-byte Folded Spill
-; SI-NEXT: v_alignbit_b32 v10, v11, v10, 16
-; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_and_b32_e32 v11, 0xffff0000, v27
-; SI-NEXT: v_add_f32_e32 v13, 0x40c00000, v11
-; SI-NEXT: v_and_b32_e32 v11, 0xffff0000, v12
-; SI-NEXT: v_and_b32_e32 v14, 0xffff0000, v59
-; SI-NEXT: v_add_f32_e32 v12, 0x40c00000, v11
-; SI-NEXT: v_and_b32_e32 v11, 0xffff0000, v60
-; SI-NEXT: v_add_f32_e32 v14, 0x40c00000, v14
-; SI-NEXT: v_add_f32_e32 v11, 0x40c00000, v11
-; SI-NEXT: v_lshrrev_b32_e32 v14, 16, v14
-; SI-NEXT: v_alignbit_b32 v11, v14, v11, 16
-; SI-NEXT: v_lshrrev_b32_e32 v14, 16, v12
-; SI-NEXT: buffer_store_dword v14, off, s[0:3], s32 offset:204 ; 4-byte Folded Spill
+; SI-NEXT: v_lshrrev_b32_e32 v6, 16, v6
+; SI-NEXT: v_alignbit_b32 v5, v6, v5, 16
+; SI-NEXT: v_lshrrev_b32_e32 v6, 16, v9
+; SI-NEXT: buffer_store_dword v6, off, s[0:3], s32 offset:276 ; 4-byte Folded Spill
+; SI-NEXT: v_alignbit_b32 v7, v6, v7, 16
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_alignbit_b32 v14, v14, v13, 16
-; SI-NEXT: buffer_load_dword v13, off, s[0:3], s32 offset:704 ; 4-byte Folded Reload
-; SI-NEXT: v_and_b32_e32 v17, 0xffff0000, v42
-; SI-NEXT: v_add_f32_e32 v17, 0x40c00000, v17
-; SI-NEXT: v_lshrrev_b32_e32 v17, 16, v17
-; SI-NEXT: v_and_b32_e32 v6, 0xffff0000, v6
+; SI-NEXT: v_and_b32_e32 v6, 0xffff0000, v44
+; SI-NEXT: v_add_f32_e32 v10, 0x40c00000, v6
+; SI-NEXT: v_and_b32_e32 v6, 0xffff0000, v42
; SI-NEXT: v_add_f32_e32 v6, 0x40c00000, v6
-; SI-NEXT: v_lshrrev_b32_e32 v58, 24, v16
-; SI-NEXT: buffer_load_dword v20, off, s[0:3], s32 offset:716 ; 4-byte Folded Reload
-; SI-NEXT: buffer_load_dword v23, off, s[0:3], s32 offset:708 ; 4-byte Folded Reload
+; SI-NEXT: v_lshrrev_b32_e32 v63, 16, v6
+; SI-NEXT: v_lshrrev_b32_e32 v6, 24, v6
+; SI-NEXT: buffer_store_dword v6, off, s[0:3], s32 offset:268 ; 4-byte Folded Spill
+; SI-NEXT: s_waitcnt expcnt(0)
+; SI-NEXT: v_lshrrev_b32_e32 v6, 24, v9
+; SI-NEXT: buffer_store_dword v6, off, s[0:3], s32 offset:260 ; 4-byte Folded Spill
+; SI-NEXT: v_alignbit_b32 v10, v63, v10, 16
+; SI-NEXT: v_lshrrev_b32_e32 v61, 24, v19
+; SI-NEXT: buffer_load_dword v22, off, s[0:3], s32 offset:708 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v23, off, s[0:3], s32 offset:700 ; 4-byte Folded Reload
; SI-NEXT: buffer_load_dword v26, off, s[0:3], s32 offset:696 ; 4-byte Folded Reload
-; SI-NEXT: buffer_load_dword v29, off, s[0:3], s32 offset:672 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v28, off, s[0:3], s32 offset:672 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v29, off, s[0:3], s32 offset:664 ; 4-byte Folded Reload
; SI-NEXT: buffer_load_dword v32, off, s[0:3], s32 offset:660 ; 4-byte Folded Reload
; SI-NEXT: buffer_load_dword v35, off, s[0:3], s32 offset:640 ; 4-byte Folded Reload
; SI-NEXT: buffer_load_dword v38, off, s[0:3], s32 offset:620 ; 4-byte Folded Reload
@@ -163553,39 +163610,41 @@ define <128 x i8> @bitcast_v64bf16_to_v128i8(<64 x bfloat> %a, i32 %b) {
; SI-NEXT: buffer_load_dword v52, off, s[0:3], s32 offset:280 ; 4-byte Folded Reload
; SI-NEXT: buffer_load_dword v54, off, s[0:3], s32 offset:272 ; 4-byte Folded Reload
; SI-NEXT: buffer_load_dword v40, off, s[0:3], s32 offset:264 ; 4-byte Folded Reload
-; SI-NEXT: s_waitcnt vmcnt(12)
+; SI-NEXT: s_waitcnt vmcnt(14)
; SI-NEXT: v_and_b32_e32 v13, 0xffff0000, v13
; SI-NEXT: v_add_f32_e32 v15, 0x40c00000, v13
-; SI-NEXT: v_and_b32_e32 v13, 0xffff0000, v19
-; SI-NEXT: v_add_f32_e32 v19, 0x40c00000, v13
-; SI-NEXT: v_and_b32_e32 v13, 0xffff0000, v61
+; SI-NEXT: v_and_b32_e32 v13, 0xffff0000, v62
+; SI-NEXT: v_add_f32_e32 v21, 0x40c00000, v13
+; SI-NEXT: v_and_b32_e32 v13, 0xffff0000, v59
; SI-NEXT: v_add_f32_e32 v13, 0x40c00000, v13
; SI-NEXT: v_alignbit_b32 v13, v17, v13, 16
-; SI-NEXT: v_lshrrev_b32_e32 v17, 16, v19
+; SI-NEXT: v_lshrrev_b32_e32 v17, 16, v21
; SI-NEXT: buffer_store_dword v17, off, s[0:3], s32 offset:208 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_alignbit_b32 v17, v17, v15, 16
; SI-NEXT: buffer_load_dword v15, off, s[0:3], s32 offset:692 ; 4-byte Folded Reload
-; SI-NEXT: s_waitcnt vmcnt(13)
-; SI-NEXT: v_and_b32_e32 v20, 0xffff0000, v20
-; SI-NEXT: v_add_f32_e32 v20, 0x40c00000, v20
-; SI-NEXT: v_lshrrev_b32_e32 v20, 16, v20
+; SI-NEXT: s_waitcnt vmcnt(14)
+; SI-NEXT: v_and_b32_e32 v22, 0xffff0000, v22
+; SI-NEXT: v_add_f32_e32 v22, 0x40c00000, v22
+; SI-NEXT: v_lshrrev_b32_e32 v22, 16, v22
; SI-NEXT: s_waitcnt vmcnt(12)
-; SI-NEXT: v_and_b32_e32 v23, 0xffff0000, v23
-; SI-NEXT: v_add_f32_e32 v23, 0x40c00000, v23
-; SI-NEXT: v_lshrrev_b32_e32 v23, 16, v23
-; SI-NEXT: s_waitcnt vmcnt(11)
; SI-NEXT: v_and_b32_e32 v26, 0xffff0000, v26
+; SI-NEXT: v_and_b32_e32 v23, 0xffff0000, v23
; SI-NEXT: v_add_f32_e32 v26, 0x40c00000, v26
+; SI-NEXT: v_add_f32_e32 v23, 0x40c00000, v23
; SI-NEXT: v_lshrrev_b32_e32 v26, 16, v26
-; SI-NEXT: s_waitcnt vmcnt(10)
-; SI-NEXT: v_and_b32_e32 v29, 0xffff0000, v29
-; SI-NEXT: v_add_f32_e32 v29, 0x40c00000, v29
-; SI-NEXT: v_lshrrev_b32_e32 v29, 16, v29
+; SI-NEXT: v_alignbit_b32 v23, v26, v23, 16
+; SI-NEXT: s_waitcnt vmcnt(11)
+; SI-NEXT: v_and_b32_e32 v28, 0xffff0000, v28
+; SI-NEXT: v_add_f32_e32 v28, 0x40c00000, v28
+; SI-NEXT: v_lshrrev_b32_e32 v28, 16, v28
; SI-NEXT: s_waitcnt vmcnt(9)
; SI-NEXT: v_and_b32_e32 v32, 0xffff0000, v32
+; SI-NEXT: v_and_b32_e32 v29, 0xffff0000, v29
; SI-NEXT: v_add_f32_e32 v32, 0x40c00000, v32
+; SI-NEXT: v_add_f32_e32 v29, 0x40c00000, v29
; SI-NEXT: v_lshrrev_b32_e32 v32, 16, v32
+; SI-NEXT: v_alignbit_b32 v29, v32, v29, 16
; SI-NEXT: s_waitcnt vmcnt(8)
; SI-NEXT: v_and_b32_e32 v35, 0xffff0000, v35
; SI-NEXT: v_add_f32_e32 v35, 0x40c00000, v35
@@ -163594,7 +163653,6 @@ define <128 x i8> @bitcast_v64bf16_to_v128i8(<64 x bfloat> %a, i32 %b) {
; SI-NEXT: v_and_b32_e32 v38, 0xffff0000, v38
; SI-NEXT: v_add_f32_e32 v38, 0x40c00000, v38
; SI-NEXT: v_lshrrev_b32_e32 v38, 16, v38
-; SI-NEXT: v_lshrrev_b32_e32 v47, 24, v19
; SI-NEXT: s_waitcnt vmcnt(6)
; SI-NEXT: v_and_b32_e32 v49, 0xffff0000, v49
; SI-NEXT: s_waitcnt vmcnt(5)
@@ -163613,6 +163671,14 @@ define <128 x i8> @bitcast_v64bf16_to_v128i8(<64 x bfloat> %a, i32 %b) {
; SI-NEXT: v_lshrrev_b32_e32 v49, 16, v49
; SI-NEXT: v_lshrrev_b32_e32 v50, 16, v51
; SI-NEXT: v_lshrrev_b32_e32 v52, 16, v53
+; SI-NEXT: s_waitcnt vmcnt(0)
+; SI-NEXT: v_and_b32_e32 v15, 0xffff0000, v15
+; SI-NEXT: v_add_f32_e32 v18, 0x40c00000, v15
+; SI-NEXT: v_and_b32_e32 v15, 0xffff0000, v56
+; SI-NEXT: v_add_f32_e32 v30, 0x40c00000, v15
+; SI-NEXT: v_and_b32_e32 v15, 0xffff0000, v16
+; SI-NEXT: buffer_load_dword v16, off, s[0:3], s32 offset:716 ; 4-byte Folded Reload
+; SI-NEXT: v_add_f32_e32 v15, 0x40c00000, v15
; SI-NEXT: v_lshrrev_b32_e32 v54, 16, v55
; SI-NEXT: v_lshrrev_b32_e32 v40, 16, v41
; SI-NEXT: buffer_store_dword v50, off, s[0:3], s32 offset:244 ; 4-byte Folded Spill
@@ -163620,91 +163686,80 @@ define <128 x i8> @bitcast_v64bf16_to_v128i8(<64 x bfloat> %a, i32 %b) {
; SI-NEXT: buffer_store_dword v54, off, s[0:3], s32 offset:252 ; 4-byte Folded Spill
; SI-NEXT: buffer_store_dword v40, off, s[0:3], s32 offset:256 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt vmcnt(4)
-; SI-NEXT: v_and_b32_e32 v15, 0xffff0000, v15
-; SI-NEXT: v_add_f32_e32 v18, 0x40c00000, v15
-; SI-NEXT: v_and_b32_e32 v15, 0xffff0000, v57
-; SI-NEXT: v_add_f32_e32 v15, 0x40c00000, v15
-; SI-NEXT: v_alignbit_b32 v15, v20, v15, 16
-; SI-NEXT: v_lshrrev_b32_e32 v20, 16, v6
-; SI-NEXT: buffer_store_dword v20, off, s[0:3], s32 offset:212 ; 4-byte Folded Spill
-; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_alignbit_b32 v20, v20, v18, 16
-; SI-NEXT: buffer_load_dword v18, off, s[0:3], s32 offset:688 ; 4-byte Folded Reload
-; SI-NEXT: v_lshrrev_b32_e32 v6, 24, v6
-; SI-NEXT: buffer_store_dword v6, off, s[0:3], s32 offset:288 ; 4-byte Folded Spill
-; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_lshrrev_b32_e32 v6, 24, v9
-; SI-NEXT: buffer_store_dword v6, off, s[0:3], s32 offset:260 ; 4-byte Folded Spill
-; SI-NEXT: s_waitcnt vmcnt(2)
-; SI-NEXT: v_and_b32_e32 v18, 0xffff0000, v18
-; SI-NEXT: v_add_f32_e32 v22, 0x40c00000, v18
+; SI-NEXT: v_and_b32_e32 v16, 0xffff0000, v16
+; SI-NEXT: v_add_f32_e32 v16, 0x40c00000, v16
+; SI-NEXT: v_lshrrev_b32_e32 v16, 16, v16
+; SI-NEXT: v_alignbit_b32 v15, v16, v15, 16
+; SI-NEXT: v_lshrrev_b32_e32 v16, 16, v30
+; SI-NEXT: v_alignbit_b32 v20, v16, v18, 16
; SI-NEXT: v_and_b32_e32 v18, 0xffff0000, v24
-; SI-NEXT: v_add_f32_e32 v21, 0x40c00000, v18
+; SI-NEXT: v_add_f32_e32 v24, 0x40c00000, v18
; SI-NEXT: buffer_load_dword v18, off, s[0:3], s32 offset:712 ; 4-byte Folded Reload
-; SI-NEXT: s_waitcnt vmcnt(0)
+; SI-NEXT: buffer_store_dword v16, off, s[0:3], s32 offset:212 ; 4-byte Folded Spill
+; SI-NEXT: s_waitcnt expcnt(0)
+; SI-NEXT: buffer_load_dword v16, off, s[0:3], s32 offset:688 ; 4-byte Folded Reload
+; SI-NEXT: s_waitcnt vmcnt(2)
; SI-NEXT: v_and_b32_e32 v18, 0xffff0000, v18
; SI-NEXT: v_add_f32_e32 v18, 0x40c00000, v18
-; SI-NEXT: v_alignbit_b32 v18, v23, v18, 16
-; SI-NEXT: v_lshrrev_b32_e32 v23, 16, v21
-; SI-NEXT: buffer_store_dword v23, off, s[0:3], s32 offset:216 ; 4-byte Folded Spill
-; SI-NEXT: v_alignbit_b32 v22, v23, v22, 16
+; SI-NEXT: s_waitcnt vmcnt(0)
+; SI-NEXT: v_and_b32_e32 v16, 0xffff0000, v16
+; SI-NEXT: v_add_f32_e32 v16, 0x40c00000, v16
+; SI-NEXT: v_alignbit_b32 v18, v22, v18, 16
+; SI-NEXT: v_lshrrev_b32_e32 v22, 16, v24
+; SI-NEXT: buffer_store_dword v22, off, s[0:3], s32 offset:216 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: buffer_load_dword v23, off, s[0:3], s32 offset:684 ; 4-byte Folded Reload
-; SI-NEXT: v_lshrrev_b32_e32 v21, 24, v21
-; SI-NEXT: buffer_store_dword v21, off, s[0:3], s32 offset:292 ; 4-byte Folded Spill
-; SI-NEXT: s_waitcnt vmcnt(1)
-; SI-NEXT: v_and_b32_e32 v23, 0xffff0000, v23
-; SI-NEXT: v_add_f32_e32 v25, 0x40c00000, v23
-; SI-NEXT: v_and_b32_e32 v23, 0xffff0000, v45
-; SI-NEXT: v_add_f32_e32 v24, 0x40c00000, v23
-; SI-NEXT: buffer_load_dword v23, off, s[0:3], s32 offset:700 ; 4-byte Folded Reload
+; SI-NEXT: v_alignbit_b32 v22, v22, v16, 16
+; SI-NEXT: buffer_load_dword v16, off, s[0:3], s32 offset:684 ; 4-byte Folded Reload
; SI-NEXT: s_waitcnt vmcnt(0)
-; SI-NEXT: v_and_b32_e32 v23, 0xffff0000, v23
-; SI-NEXT: v_add_f32_e32 v23, 0x40c00000, v23
-; SI-NEXT: v_alignbit_b32 v23, v26, v23, 16
-; SI-NEXT: v_lshrrev_b32_e32 v26, 16, v24
+; SI-NEXT: v_and_b32_e32 v16, 0xffff0000, v16
+; SI-NEXT: v_add_f32_e32 v25, 0x40c00000, v16
+; SI-NEXT: v_and_b32_e32 v16, 0xffff0000, v57
+; SI-NEXT: v_add_f32_e32 v16, 0x40c00000, v16
+; SI-NEXT: v_lshrrev_b32_e32 v26, 16, v16
; SI-NEXT: buffer_store_dword v26, off, s[0:3], s32 offset:220 ; 4-byte Folded Spill
; SI-NEXT: v_alignbit_b32 v25, v26, v25, 16
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: buffer_load_dword v26, off, s[0:3], s32 offset:680 ; 4-byte Folded Reload
-; SI-NEXT: v_lshrrev_b32_e32 v24, 24, v24
-; SI-NEXT: buffer_store_dword v24, off, s[0:3], s32 offset:296 ; 4-byte Folded Spill
-; SI-NEXT: s_waitcnt vmcnt(1)
+; SI-NEXT: s_waitcnt vmcnt(0)
; SI-NEXT: v_and_b32_e32 v26, 0xffff0000, v26
-; SI-NEXT: v_add_f32_e32 v28, 0x40c00000, v26
-; SI-NEXT: v_and_b32_e32 v26, 0xffff0000, v62
; SI-NEXT: v_add_f32_e32 v27, 0x40c00000, v26
; SI-NEXT: buffer_load_dword v26, off, s[0:3], s32 offset:676 ; 4-byte Folded Reload
; SI-NEXT: s_waitcnt vmcnt(0)
; SI-NEXT: v_and_b32_e32 v26, 0xffff0000, v26
; SI-NEXT: v_add_f32_e32 v26, 0x40c00000, v26
-; SI-NEXT: v_alignbit_b32 v26, v29, v26, 16
-; SI-NEXT: v_lshrrev_b32_e32 v29, 16, v27
-; SI-NEXT: buffer_store_dword v29, off, s[0:3], s32 offset:224 ; 4-byte Folded Spill
-; SI-NEXT: v_alignbit_b32 v28, v29, v28, 16
+; SI-NEXT: v_alignbit_b32 v26, v28, v26, 16
+; SI-NEXT: v_lshrrev_b32_e32 v28, 16, v12
+; SI-NEXT: buffer_store_dword v28, off, s[0:3], s32 offset:224 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: buffer_load_dword v29, off, s[0:3], s32 offset:668 ; 4-byte Folded Reload
-; SI-NEXT: v_lshrrev_b32_e32 v27, 24, v27
-; SI-NEXT: buffer_store_dword v27, off, s[0:3], s32 offset:300 ; 4-byte Folded Spill
-; SI-NEXT: s_waitcnt vmcnt(1)
-; SI-NEXT: v_and_b32_e32 v29, 0xffff0000, v29
-; SI-NEXT: v_add_f32_e32 v31, 0x40c00000, v29
-; SI-NEXT: buffer_load_dword v29, off, s[0:3], s32 offset:616 ; 4-byte Folded Reload
-; SI-NEXT: s_waitcnt vmcnt(0)
-; SI-NEXT: v_and_b32_e32 v29, 0xffff0000, v29
-; SI-NEXT: v_add_f32_e32 v30, 0x40c00000, v29
-; SI-NEXT: buffer_load_dword v29, off, s[0:3], s32 offset:664 ; 4-byte Folded Reload
+; SI-NEXT: v_alignbit_b32 v28, v28, v27, 16
+; SI-NEXT: buffer_load_dword v27, off, s[0:3], s32 offset:668 ; 4-byte Folded Reload
+; SI-NEXT: v_lshrrev_b32_e32 v12, 24, v12
+; SI-NEXT: buffer_store_dword v12, off, s[0:3], s32 offset:300 ; 4-byte Folded Spill
+; SI-NEXT: s_waitcnt expcnt(0)
+; SI-NEXT: v_lshrrev_b32_e32 v12, 24, v16
+; SI-NEXT: buffer_store_dword v12, off, s[0:3], s32 offset:296 ; 4-byte Folded Spill
+; SI-NEXT: s_waitcnt expcnt(0)
+; SI-NEXT: v_lshrrev_b32_e32 v12, 24, v24
+; SI-NEXT: buffer_store_dword v12, off, s[0:3], s32 offset:292 ; 4-byte Folded Spill
+; SI-NEXT: s_waitcnt expcnt(0)
+; SI-NEXT: v_lshrrev_b32_e32 v12, 24, v30
+; SI-NEXT: buffer_store_dword v12, off, s[0:3], s32 offset:288 ; 4-byte Folded Spill
+; SI-NEXT: s_waitcnt vmcnt(4)
+; SI-NEXT: v_and_b32_e32 v27, 0xffff0000, v27
+; SI-NEXT: v_add_f32_e32 v31, 0x40c00000, v27
+; SI-NEXT: buffer_load_dword v27, off, s[0:3], s32 offset:616 ; 4-byte Folded Reload
; SI-NEXT: s_waitcnt vmcnt(0)
-; SI-NEXT: v_and_b32_e32 v29, 0xffff0000, v29
-; SI-NEXT: v_add_f32_e32 v29, 0x40c00000, v29
-; SI-NEXT: v_alignbit_b32 v29, v32, v29, 16
-; SI-NEXT: v_lshrrev_b32_e32 v32, 16, v30
+; SI-NEXT: v_and_b32_e32 v27, 0xffff0000, v27
+; SI-NEXT: v_add_f32_e32 v27, 0x40c00000, v27
+; SI-NEXT: v_lshrrev_b32_e32 v32, 16, v27
; SI-NEXT: buffer_store_dword v32, off, s[0:3], s32 offset:228 ; 4-byte Folded Spill
; SI-NEXT: v_alignbit_b32 v31, v32, v31, 16
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: buffer_load_dword v32, off, s[0:3], s32 offset:656 ; 4-byte Folded Reload
-; SI-NEXT: v_lshrrev_b32_e32 v30, 24, v30
-; SI-NEXT: buffer_store_dword v30, off, s[0:3], s32 offset:304 ; 4-byte Folded Spill
+; SI-NEXT: v_lshrrev_b32_e32 v27, 24, v27
+; SI-NEXT: buffer_store_dword v27, off, s[0:3], s32 offset:304 ; 4-byte Folded Spill
+; SI-NEXT: s_waitcnt expcnt(0)
+; SI-NEXT: v_lshrrev_b32_e32 v27, 24, v21
; SI-NEXT: s_waitcnt vmcnt(1)
; SI-NEXT: v_and_b32_e32 v32, 0xffff0000, v32
; SI-NEXT: v_add_f32_e32 v34, 0x40c00000, v32
@@ -163724,8 +163779,6 @@ define <128 x i8> @bitcast_v64bf16_to_v128i8(<64 x bfloat> %a, i32 %b) {
; SI-NEXT: buffer_load_dword v35, off, s[0:3], s32 offset:644 ; 4-byte Folded Reload
; SI-NEXT: v_lshrrev_b32_e32 v33, 24, v33
; SI-NEXT: buffer_store_dword v33, off, s[0:3], s32 offset:320 ; 4-byte Folded Spill
-; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_lshrrev_b32_e32 v33, 24, v12
; SI-NEXT: s_waitcnt vmcnt(1)
; SI-NEXT: v_and_b32_e32 v35, 0xffff0000, v35
; SI-NEXT: v_add_f32_e32 v37, 0x40c00000, v35
@@ -163987,8 +164040,8 @@ define <128 x i8> @bitcast_v64bf16_to_v128i8(<64 x bfloat> %a, i32 %b) {
; SI-NEXT: buffer_store_dword v6, off, s[0:3], s32 offset:372 ; 4-byte Folded Spill
; SI-NEXT: .LBB90_4: ; %end
; SI-NEXT: s_or_b64 exec, exec, s[4:5]
-; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: buffer_load_dword v9, off, s[0:3], s32 offset:592 ; 4-byte Folded Reload
+; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_and_b32_e32 v6, 0xff, v38
; SI-NEXT: v_and_b32_e32 v5, 0xff, v5
; SI-NEXT: v_and_b32_e32 v4, 0xff, v4
@@ -164442,7 +164495,7 @@ define <128 x i8> @bitcast_v64bf16_to_v128i8(<64 x bfloat> %a, i32 %b) {
; SI-NEXT: buffer_load_dword v5, off, s[0:3], s32 offset:444 ; 4-byte Folded Reload
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_and_b32_e32 v4, 0xff, v17
-; SI-NEXT: v_lshlrev_b32_e32 v6, 24, v47
+; SI-NEXT: v_lshlrev_b32_e32 v6, 24, v27
; SI-NEXT: s_waitcnt vmcnt(0)
; SI-NEXT: v_lshlrev_b32_e32 v5, 8, v5
; SI-NEXT: v_or_b32_e32 v4, v4, v5
@@ -164475,7 +164528,7 @@ define <128 x i8> @bitcast_v64bf16_to_v128i8(<64 x bfloat> %a, i32 %b) {
; SI-NEXT: buffer_load_dword v4, off, s[0:3], s32 offset:416 ; 4-byte Folded Reload
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_and_b32_e32 v3, 0xff, v14
-; SI-NEXT: v_lshlrev_b32_e32 v5, 24, v33
+; SI-NEXT: v_lshlrev_b32_e32 v5, 24, v61
; SI-NEXT: s_waitcnt vmcnt(0)
; SI-NEXT: v_lshlrev_b32_e32 v4, 8, v4
; SI-NEXT: v_or_b32_e32 v3, v3, v4
@@ -164506,17 +164559,17 @@ define <128 x i8> @bitcast_v64bf16_to_v128i8(<64 x bfloat> %a, i32 %b) {
; SI-NEXT: v_add_i32_e32 v3, vcc, 0x70, v0
; SI-NEXT: buffer_store_dword v2, v3, s[0:3], 0 offen
; SI-NEXT: buffer_load_dword v3, off, s[0:3], s32 offset:396 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v4, off, s[0:3], s32 offset:268 ; 4-byte Folded Reload
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_and_b32_e32 v2, 0xff, v10
-; SI-NEXT: v_lshlrev_b32_e32 v4, 24, v58
-; SI-NEXT: s_waitcnt vmcnt(0)
+; SI-NEXT: s_waitcnt vmcnt(1)
; SI-NEXT: v_lshlrev_b32_e32 v3, 8, v3
; SI-NEXT: v_or_b32_e32 v2, v2, v3
-; SI-NEXT: buffer_load_dword v3, off, s[0:3], s32 offset:276 ; 4-byte Folded Reload
-; SI-NEXT: v_and_b32_e32 v2, 0xffff, v2
-; SI-NEXT: s_waitcnt vmcnt(0)
-; SI-NEXT: v_and_b32_e32 v3, 0xff, v3
+; SI-NEXT: v_and_b32_e32 v3, 0xff, v63
; SI-NEXT: v_lshlrev_b32_e32 v3, 16, v3
+; SI-NEXT: s_waitcnt vmcnt(0)
+; SI-NEXT: v_lshlrev_b32_e32 v4, 24, v4
+; SI-NEXT: v_and_b32_e32 v2, 0xffff, v2
; SI-NEXT: v_or_b32_e32 v3, v4, v3
; SI-NEXT: v_or_b32_e32 v2, v2, v3
; SI-NEXT: v_add_i32_e32 v3, vcc, 0x74, v0
@@ -164546,7 +164599,7 @@ define <128 x i8> @bitcast_v64bf16_to_v128i8(<64 x bfloat> %a, i32 %b) {
; SI-NEXT: s_waitcnt vmcnt(1)
; SI-NEXT: v_lshlrev_b32_e32 v2, 8, v2
; SI-NEXT: v_or_b32_e32 v1, v1, v2
-; SI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:268 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:276 ; 4-byte Folded Reload
; SI-NEXT: s_waitcnt vmcnt(1)
; SI-NEXT: v_lshlrev_b32_e32 v3, 24, v3
; SI-NEXT: v_and_b32_e32 v1, 0xffff, v1
@@ -164597,12 +164650,65 @@ define <128 x i8> @bitcast_v64bf16_to_v128i8(<64 x bfloat> %a, i32 %b) {
; VI-NEXT: buffer_load_dword v32, off, s[0:3], s32 offset:4
; VI-NEXT: buffer_load_dword v33, off, s[0:3], s32 offset:8
; VI-NEXT: buffer_load_dword v31, off, s[0:3], s32
+; VI-NEXT: ; implicit-def: $vgpr34
+; VI-NEXT: ; kill: killed $vgpr34
+; VI-NEXT: ; implicit-def: $vgpr34
+; VI-NEXT: ; kill: killed $vgpr34
+; VI-NEXT: ; implicit-def: $vgpr34
+; VI-NEXT: ; kill: killed $vgpr34
+; VI-NEXT: ; implicit-def: $vgpr34
+; VI-NEXT: ; kill: killed $vgpr34
+; VI-NEXT: ; implicit-def: $vgpr34
+; VI-NEXT: ; kill: killed $vgpr34
+; VI-NEXT: ; implicit-def: $vgpr34
+; VI-NEXT: ; kill: killed $vgpr34
+; VI-NEXT: ; implicit-def: $vgpr34
+; VI-NEXT: ; kill: killed $vgpr34
+; VI-NEXT: ; implicit-def: $vgpr34
+; VI-NEXT: ; kill: killed $vgpr34
+; VI-NEXT: ; implicit-def: $vgpr34
+; VI-NEXT: ; kill: killed $vgpr34
+; VI-NEXT: ; implicit-def: $vgpr34
+; VI-NEXT: ; kill: killed $vgpr34
+; VI-NEXT: ; implicit-def: $vgpr34
+; VI-NEXT: ; kill: killed $vgpr34
+; VI-NEXT: ; implicit-def: $vgpr34
+; VI-NEXT: ; kill: killed $vgpr34
+; VI-NEXT: ; implicit-def: $vgpr34
+; VI-NEXT: ; kill: killed $vgpr34
+; VI-NEXT: ; implicit-def: $vgpr34
+; VI-NEXT: ; kill: killed $vgpr34
+; VI-NEXT: ; implicit-def: $vgpr34
+; VI-NEXT: ; kill: killed $vgpr34
+; VI-NEXT: ; implicit-def: $vgpr34
+; VI-NEXT: ; kill: killed $vgpr34
+; VI-NEXT: ; implicit-def: $vgpr34
+; VI-NEXT: ; kill: killed $vgpr34
+; VI-NEXT: ; implicit-def: $vgpr34
+; VI-NEXT: ; kill: killed $vgpr34
+; VI-NEXT: ; implicit-def: $vgpr34
+; VI-NEXT: ; kill: killed $vgpr34
+; VI-NEXT: ; implicit-def: $vgpr34
+; VI-NEXT: ; kill: killed $vgpr34
+; VI-NEXT: ; implicit-def: $vgpr34
+; VI-NEXT: ; kill: killed $vgpr34
+; VI-NEXT: ; implicit-def: $vgpr34
+; VI-NEXT: ; kill: killed $vgpr34
+; VI-NEXT: ; implicit-def: $vgpr34
+; VI-NEXT: ; kill: killed $vgpr34
+; VI-NEXT: ; implicit-def: $vgpr34
+; VI-NEXT: ; kill: killed $vgpr34
+; VI-NEXT: ; implicit-def: $vgpr34
+; VI-NEXT: ; kill: killed $vgpr34
+; VI-NEXT: ; implicit-def: $vgpr34
+; VI-NEXT: ; kill: killed $vgpr34
+; VI-NEXT: ; implicit-def: $vgpr34
; VI-NEXT: ; implicit-def: $vgpr35
+; VI-NEXT: ; kill: killed $vgpr34
; VI-NEXT: ; implicit-def: $vgpr45
; VI-NEXT: ; implicit-def: $vgpr34
-; VI-NEXT: ; implicit-def: $vgpr38
-; VI-NEXT: ; implicit-def: $vgpr55
-; VI-NEXT: ; implicit-def: $vgpr54
+; VI-NEXT: ; implicit-def: $vgpr63
+; VI-NEXT: ; implicit-def: $vgpr43
; VI-NEXT: ; implicit-def: $vgpr60
; VI-NEXT: ; implicit-def: $vgpr59
; VI-NEXT: ; implicit-def: $vgpr42
@@ -164615,11 +164721,13 @@ define <128 x i8> @bitcast_v64bf16_to_v128i8(<64 x bfloat> %a, i32 %b) {
; VI-NEXT: ; implicit-def: $vgpr46
; VI-NEXT: ; implicit-def: $vgpr39
; VI-NEXT: ; implicit-def: $vgpr49
-; VI-NEXT: ; implicit-def: $vgpr36
+; VI-NEXT: ; implicit-def: $vgpr38
+; VI-NEXT: ; implicit-def: $vgpr55
+; VI-NEXT: ; implicit-def: $vgpr54
; VI-NEXT: ; implicit-def: $vgpr53
; VI-NEXT: ; implicit-def: $vgpr62
; VI-NEXT: ; implicit-def: $vgpr61
-; VI-NEXT: ; implicit-def: $vgpr63
+; VI-NEXT: ; implicit-def: $vgpr36
; VI-NEXT: ; implicit-def: $vgpr58
; VI-NEXT: ; implicit-def: $vgpr48
; VI-NEXT: ; implicit-def: $vgpr52
@@ -164722,173 +164830,115 @@ define <128 x i8> @bitcast_v64bf16_to_v128i8(<64 x bfloat> %a, i32 %b) {
; VI-NEXT: ; implicit-def: $vgpr33
; VI-NEXT: ; kill: killed $vgpr33
; VI-NEXT: ; implicit-def: $vgpr33
-; VI-NEXT: ; kill: killed $vgpr33
-; VI-NEXT: ; implicit-def: $vgpr33
-; VI-NEXT: ; kill: killed $vgpr33
-; VI-NEXT: ; implicit-def: $vgpr33
-; VI-NEXT: ; kill: killed $vgpr33
-; VI-NEXT: ; implicit-def: $vgpr33
-; VI-NEXT: ; kill: killed $vgpr33
-; VI-NEXT: ; implicit-def: $vgpr33
-; VI-NEXT: ; kill: killed $vgpr33
-; VI-NEXT: ; implicit-def: $vgpr33
-; VI-NEXT: ; kill: killed $vgpr33
-; VI-NEXT: ; implicit-def: $vgpr33
-; VI-NEXT: ; kill: killed $vgpr33
-; VI-NEXT: ; implicit-def: $vgpr33
-; VI-NEXT: ; kill: killed $vgpr33
-; VI-NEXT: ; implicit-def: $vgpr33
-; VI-NEXT: ; kill: killed $vgpr33
-; VI-NEXT: ; implicit-def: $vgpr33
-; VI-NEXT: ; kill: killed $vgpr33
-; VI-NEXT: ; implicit-def: $vgpr33
-; VI-NEXT: ; kill: killed $vgpr33
-; VI-NEXT: ; implicit-def: $vgpr33
-; VI-NEXT: ; kill: killed $vgpr33
-; VI-NEXT: ; implicit-def: $vgpr33
-; VI-NEXT: ; kill: killed $vgpr33
-; VI-NEXT: ; implicit-def: $vgpr33
-; VI-NEXT: ; kill: killed $vgpr33
-; VI-NEXT: ; implicit-def: $vgpr33
-; VI-NEXT: ; kill: killed $vgpr33
-; VI-NEXT: ; implicit-def: $vgpr33
-; VI-NEXT: ; kill: killed $vgpr33
-; VI-NEXT: ; implicit-def: $vgpr33
-; VI-NEXT: ; kill: killed $vgpr33
-; VI-NEXT: ; implicit-def: $vgpr33
-; VI-NEXT: ; kill: killed $vgpr33
-; VI-NEXT: ; implicit-def: $vgpr33
-; VI-NEXT: ; kill: killed $vgpr33
-; VI-NEXT: ; implicit-def: $vgpr33
-; VI-NEXT: ; kill: killed $vgpr33
-; VI-NEXT: ; implicit-def: $vgpr33
-; VI-NEXT: ; kill: killed $vgpr33
-; VI-NEXT: ; implicit-def: $vgpr33
-; VI-NEXT: ; kill: killed $vgpr33
-; VI-NEXT: ; implicit-def: $vgpr33
-; VI-NEXT: ; kill: killed $vgpr33
-; VI-NEXT: ; implicit-def: $vgpr33
-; VI-NEXT: ; kill: killed $vgpr33
-; VI-NEXT: ; implicit-def: $vgpr33
-; VI-NEXT: ; kill: killed $vgpr33
-; VI-NEXT: ; implicit-def: $vgpr33
-; VI-NEXT: ; kill: killed $vgpr33
-; VI-NEXT: ; implicit-def: $vgpr33
-; VI-NEXT: ; kill: killed $vgpr33
-; VI-NEXT: ; implicit-def: $vgpr33
-; VI-NEXT: ; kill: killed $vgpr33
-; VI-NEXT: ; implicit-def: $vgpr33
; VI-NEXT: s_and_saveexec_b64 s[4:5], vcc
; VI-NEXT: s_xor_b64 s[4:5], exec, s[4:5]
; VI-NEXT: s_cbranch_execz .LBB90_2
; VI-NEXT: ; %bb.1: ; %cmp.false
-; VI-NEXT: v_lshrrev_b32_e32 v51, 8, v29
-; VI-NEXT: buffer_store_dword v51, off, s[0:3], s32 offset:400 ; 4-byte Folded Spill
-; VI-NEXT: v_lshrrev_b32_e32 v51, 24, v28
-; VI-NEXT: buffer_store_dword v51, off, s[0:3], s32 offset:240 ; 4-byte Folded Spill
-; VI-NEXT: v_lshrrev_b32_e32 v51, 16, v28
-; VI-NEXT: buffer_store_dword v51, off, s[0:3], s32 offset:248 ; 4-byte Folded Spill
-; VI-NEXT: v_lshrrev_b32_e32 v51, 8, v28
-; VI-NEXT: buffer_store_dword v51, off, s[0:3], s32 offset:252 ; 4-byte Folded Spill
-; VI-NEXT: v_lshrrev_b32_e32 v51, 16, v27
-; VI-NEXT: buffer_store_dword v51, off, s[0:3], s32 offset:260 ; 4-byte Folded Spill
-; VI-NEXT: v_lshrrev_b32_e32 v51, 8, v27
-; VI-NEXT: buffer_store_dword v51, off, s[0:3], s32 offset:268 ; 4-byte Folded Spill
-; VI-NEXT: v_lshrrev_b32_e32 v51, 24, v26
-; VI-NEXT: buffer_store_dword v51, off, s[0:3], s32 offset:276 ; 4-byte Folded Spill
-; VI-NEXT: v_lshrrev_b32_e32 v51, 16, v26
-; VI-NEXT: buffer_store_dword v51, off, s[0:3], s32 offset:284 ; 4-byte Folded Spill
-; VI-NEXT: v_lshrrev_b32_e32 v51, 8, v26
; VI-NEXT: v_lshrrev_b32_e32 v33, 24, v16
-; VI-NEXT: buffer_store_dword v51, off, s[0:3], s32 offset:292 ; 4-byte Folded Spill
-; VI-NEXT: v_lshrrev_b32_e32 v51, 16, v25
; VI-NEXT: buffer_store_dword v33, off, s[0:3], s32 offset:196 ; 4-byte Folded Spill
; VI-NEXT: v_lshrrev_b32_e32 v33, 16, v16
-; VI-NEXT: buffer_store_dword v51, off, s[0:3], s32 offset:296 ; 4-byte Folded Spill
-; VI-NEXT: v_lshrrev_b32_e32 v51, 8, v25
; VI-NEXT: buffer_store_dword v33, off, s[0:3], s32 offset:200 ; 4-byte Folded Spill
; VI-NEXT: v_lshrrev_b32_e32 v33, 8, v16
-; VI-NEXT: buffer_store_dword v51, off, s[0:3], s32 offset:308 ; 4-byte Folded Spill
-; VI-NEXT: v_lshrrev_b32_e32 v51, 24, v24
; VI-NEXT: buffer_store_dword v33, off, s[0:3], s32 offset:204 ; 4-byte Folded Spill
; VI-NEXT: v_lshrrev_b32_e32 v33, 16, v15
-; VI-NEXT: buffer_store_dword v51, off, s[0:3], s32 offset:320 ; 4-byte Folded Spill
-; VI-NEXT: v_lshrrev_b32_e32 v51, 16, v24
; VI-NEXT: buffer_store_dword v33, off, s[0:3], s32 offset:208 ; 4-byte Folded Spill
; VI-NEXT: v_lshrrev_b32_e32 v33, 8, v15
-; VI-NEXT: buffer_store_dword v51, off, s[0:3], s32 offset:324 ; 4-byte Folded Spill
-; VI-NEXT: v_lshrrev_b32_e32 v51, 8, v24
; VI-NEXT: buffer_store_dword v33, off, s[0:3], s32 offset:212 ; 4-byte Folded Spill
; VI-NEXT: v_lshrrev_b32_e32 v33, 24, v14
-; VI-NEXT: buffer_store_dword v51, off, s[0:3], s32 offset:328 ; 4-byte Folded Spill
-; VI-NEXT: v_lshrrev_b32_e32 v51, 16, v23
; VI-NEXT: buffer_store_dword v33, off, s[0:3], s32 offset:216 ; 4-byte Folded Spill
; VI-NEXT: v_lshrrev_b32_e32 v33, 16, v14
-; VI-NEXT: buffer_store_dword v51, off, s[0:3], s32 offset:332 ; 4-byte Folded Spill
-; VI-NEXT: v_lshrrev_b32_e32 v51, 8, v23
; VI-NEXT: buffer_store_dword v33, off, s[0:3], s32 offset:220 ; 4-byte Folded Spill
; VI-NEXT: v_lshrrev_b32_e32 v33, 8, v14
-; VI-NEXT: buffer_store_dword v51, off, s[0:3], s32 offset:336 ; 4-byte Folded Spill
-; VI-NEXT: v_lshrrev_b32_e32 v51, 24, v22
; VI-NEXT: buffer_store_dword v33, off, s[0:3], s32 offset:224 ; 4-byte Folded Spill
; VI-NEXT: v_lshrrev_b32_e32 v33, 16, v13
-; VI-NEXT: buffer_store_dword v51, off, s[0:3], s32 offset:340 ; 4-byte Folded Spill
-; VI-NEXT: v_lshrrev_b32_e32 v51, 16, v22
; VI-NEXT: buffer_store_dword v33, off, s[0:3], s32 offset:228 ; 4-byte Folded Spill
; VI-NEXT: v_lshrrev_b32_e32 v33, 8, v13
-; VI-NEXT: buffer_store_dword v51, off, s[0:3], s32 offset:344 ; 4-byte Folded Spill
-; VI-NEXT: v_lshrrev_b32_e32 v51, 8, v22
; VI-NEXT: buffer_store_dword v33, off, s[0:3], s32 offset:232 ; 4-byte Folded Spill
-; VI-NEXT: v_lshrrev_b32_e32 v33, 16, v5
-; VI-NEXT: buffer_store_dword v51, off, s[0:3], s32 offset:348 ; 4-byte Folded Spill
-; VI-NEXT: v_lshrrev_b32_e32 v51, 16, v21
+; VI-NEXT: v_lshrrev_b32_e32 v33, 24, v12
; VI-NEXT: buffer_store_dword v33, off, s[0:3], s32 offset:236 ; 4-byte Folded Spill
+; VI-NEXT: v_lshrrev_b32_e32 v33, 16, v12
+; VI-NEXT: buffer_store_dword v33, off, s[0:3], s32 offset:360 ; 4-byte Folded Spill
+; VI-NEXT: v_lshrrev_b32_e32 v33, 24, v6
+; VI-NEXT: buffer_store_dword v33, off, s[0:3], s32 offset:248 ; 4-byte Folded Spill
+; VI-NEXT: v_lshrrev_b32_e32 v33, 16, v6
+; VI-NEXT: buffer_store_dword v33, off, s[0:3], s32 offset:252 ; 4-byte Folded Spill
+; VI-NEXT: v_lshrrev_b32_e32 v33, 8, v6
+; VI-NEXT: buffer_store_dword v33, off, s[0:3], s32 offset:260 ; 4-byte Folded Spill
+; VI-NEXT: v_lshrrev_b32_e32 v33, 16, v5
+; VI-NEXT: buffer_store_dword v33, off, s[0:3], s32 offset:268 ; 4-byte Folded Spill
; VI-NEXT: v_lshrrev_b32_e32 v33, 8, v5
-; VI-NEXT: buffer_store_dword v51, off, s[0:3], s32 offset:352 ; 4-byte Folded Spill
-; VI-NEXT: v_lshrrev_b32_e32 v51, 8, v21
-; VI-NEXT: buffer_store_dword v33, off, s[0:3], s32 offset:244 ; 4-byte Folded Spill
+; VI-NEXT: buffer_store_dword v33, off, s[0:3], s32 offset:276 ; 4-byte Folded Spill
; VI-NEXT: v_lshrrev_b32_e32 v33, 24, v4
-; VI-NEXT: buffer_store_dword v51, off, s[0:3], s32 offset:356 ; 4-byte Folded Spill
-; VI-NEXT: v_lshrrev_b32_e32 v51, 24, v20
-; VI-NEXT: buffer_store_dword v33, off, s[0:3], s32 offset:256 ; 4-byte Folded Spill
+; VI-NEXT: buffer_store_dword v33, off, s[0:3], s32 offset:288 ; 4-byte Folded Spill
; VI-NEXT: v_lshrrev_b32_e32 v33, 16, v4
-; VI-NEXT: buffer_store_dword v51, off, s[0:3], s32 offset:360 ; 4-byte Folded Spill
-; VI-NEXT: v_lshrrev_b32_e32 v51, 16, v20
-; VI-NEXT: buffer_store_dword v33, off, s[0:3], s32 offset:264 ; 4-byte Folded Spill
+; VI-NEXT: buffer_store_dword v33, off, s[0:3], s32 offset:296 ; 4-byte Folded Spill
; VI-NEXT: v_lshrrev_b32_e32 v33, 8, v4
-; VI-NEXT: buffer_store_dword v51, off, s[0:3], s32 offset:364 ; 4-byte Folded Spill
-; VI-NEXT: v_lshrrev_b32_e32 v51, 8, v20
-; VI-NEXT: buffer_store_dword v33, off, s[0:3], s32 offset:272 ; 4-byte Folded Spill
+; VI-NEXT: buffer_store_dword v33, off, s[0:3], s32 offset:304 ; 4-byte Folded Spill
; VI-NEXT: v_lshrrev_b32_e32 v33, 16, v3
-; VI-NEXT: buffer_store_dword v51, off, s[0:3], s32 offset:368 ; 4-byte Folded Spill
-; VI-NEXT: v_lshrrev_b32_e32 v51, 16, v19
-; VI-NEXT: buffer_store_dword v33, off, s[0:3], s32 offset:280 ; 4-byte Folded Spill
+; VI-NEXT: buffer_store_dword v33, off, s[0:3], s32 offset:312 ; 4-byte Folded Spill
; VI-NEXT: v_lshrrev_b32_e32 v33, 8, v3
-; VI-NEXT: buffer_store_dword v51, off, s[0:3], s32 offset:372 ; 4-byte Folded Spill
-; VI-NEXT: v_lshrrev_b32_e32 v51, 8, v19
-; VI-NEXT: buffer_store_dword v33, off, s[0:3], s32 offset:288 ; 4-byte Folded Spill
+; VI-NEXT: buffer_store_dword v33, off, s[0:3], s32 offset:320 ; 4-byte Folded Spill
; VI-NEXT: v_lshrrev_b32_e32 v33, 24, v2
-; VI-NEXT: buffer_store_dword v51, off, s[0:3], s32 offset:376 ; 4-byte Folded Spill
-; VI-NEXT: v_lshrrev_b32_e32 v51, 24, v18
-; VI-NEXT: buffer_store_dword v33, off, s[0:3], s32 offset:300 ; 4-byte Folded Spill
+; VI-NEXT: buffer_store_dword v33, off, s[0:3], s32 offset:332 ; 4-byte Folded Spill
; VI-NEXT: v_lshrrev_b32_e32 v33, 16, v2
-; VI-NEXT: buffer_store_dword v51, off, s[0:3], s32 offset:380 ; 4-byte Folded Spill
-; VI-NEXT: v_lshrrev_b32_e32 v51, 16, v18
-; VI-NEXT: buffer_store_dword v33, off, s[0:3], s32 offset:304 ; 4-byte Folded Spill
-; VI-NEXT: v_lshrrev_b32_e32 v33, 8, v2
-; VI-NEXT: buffer_store_dword v51, off, s[0:3], s32 offset:384 ; 4-byte Folded Spill
-; VI-NEXT: v_lshrrev_b32_e32 v51, 8, v18
-; VI-NEXT: buffer_store_dword v33, off, s[0:3], s32 offset:316 ; 4-byte Folded Spill
+; VI-NEXT: buffer_store_dword v33, off, s[0:3], s32 offset:336 ; 4-byte Folded Spill
; VI-NEXT: v_lshrrev_b32_e32 v33, 16, v1
-; VI-NEXT: buffer_store_dword v51, off, s[0:3], s32 offset:388 ; 4-byte Folded Spill
-; VI-NEXT: v_lshrrev_b32_e32 v51, 16, v17
+; VI-NEXT: buffer_store_dword v33, off, s[0:3], s32 offset:344 ; 4-byte Folded Spill
+; VI-NEXT: v_lshrrev_b32_e32 v33, 16, v30
+; VI-NEXT: buffer_store_dword v33, off, s[0:3], s32 offset:240 ; 4-byte Folded Spill
+; VI-NEXT: v_lshrrev_b32_e32 v33, 8, v30
+; VI-NEXT: buffer_store_dword v33, off, s[0:3], s32 offset:244 ; 4-byte Folded Spill
+; VI-NEXT: v_lshrrev_b32_e32 v33, 16, v29
+; VI-NEXT: buffer_store_dword v33, off, s[0:3], s32 offset:256 ; 4-byte Folded Spill
+; VI-NEXT: v_lshrrev_b32_e32 v33, 8, v29
+; VI-NEXT: buffer_store_dword v33, off, s[0:3], s32 offset:264 ; 4-byte Folded Spill
+; VI-NEXT: v_lshrrev_b32_e32 v33, 24, v28
+; VI-NEXT: buffer_store_dword v33, off, s[0:3], s32 offset:272 ; 4-byte Folded Spill
+; VI-NEXT: v_lshrrev_b32_e32 v33, 16, v28
+; VI-NEXT: buffer_store_dword v33, off, s[0:3], s32 offset:280 ; 4-byte Folded Spill
+; VI-NEXT: v_lshrrev_b32_e32 v33, 8, v28
+; VI-NEXT: buffer_store_dword v33, off, s[0:3], s32 offset:284 ; 4-byte Folded Spill
+; VI-NEXT: v_lshrrev_b32_e32 v33, 16, v27
+; VI-NEXT: buffer_store_dword v33, off, s[0:3], s32 offset:292 ; 4-byte Folded Spill
+; VI-NEXT: v_lshrrev_b32_e32 v33, 8, v27
+; VI-NEXT: buffer_store_dword v33, off, s[0:3], s32 offset:300 ; 4-byte Folded Spill
+; VI-NEXT: v_lshrrev_b32_e32 v33, 24, v26
+; VI-NEXT: buffer_store_dword v33, off, s[0:3], s32 offset:308 ; 4-byte Folded Spill
+; VI-NEXT: v_lshrrev_b32_e32 v33, 16, v26
+; VI-NEXT: buffer_store_dword v33, off, s[0:3], s32 offset:316 ; 4-byte Folded Spill
+; VI-NEXT: v_lshrrev_b32_e32 v33, 8, v26
+; VI-NEXT: buffer_store_dword v33, off, s[0:3], s32 offset:324 ; 4-byte Folded Spill
+; VI-NEXT: v_lshrrev_b32_e32 v33, 16, v25
+; VI-NEXT: buffer_store_dword v33, off, s[0:3], s32 offset:328 ; 4-byte Folded Spill
+; VI-NEXT: v_lshrrev_b32_e32 v33, 8, v25
+; VI-NEXT: buffer_store_dword v33, off, s[0:3], s32 offset:340 ; 4-byte Folded Spill
+; VI-NEXT: v_lshrrev_b32_e32 v33, 24, v24
+; VI-NEXT: buffer_store_dword v33, off, s[0:3], s32 offset:348 ; 4-byte Folded Spill
+; VI-NEXT: v_lshrrev_b32_e32 v33, 16, v24
+; VI-NEXT: buffer_store_dword v33, off, s[0:3], s32 offset:352 ; 4-byte Folded Spill
+; VI-NEXT: v_lshrrev_b32_e32 v33, 8, v24
+; VI-NEXT: buffer_store_dword v33, off, s[0:3], s32 offset:356 ; 4-byte Folded Spill
+; VI-NEXT: v_lshrrev_b32_e32 v33, 16, v23
+; VI-NEXT: buffer_store_dword v33, off, s[0:3], s32 offset:364 ; 4-byte Folded Spill
+; VI-NEXT: v_lshrrev_b32_e32 v33, 8, v23
+; VI-NEXT: buffer_store_dword v33, off, s[0:3], s32 offset:368 ; 4-byte Folded Spill
+; VI-NEXT: v_lshrrev_b32_e32 v33, 24, v22
+; VI-NEXT: buffer_store_dword v33, off, s[0:3], s32 offset:372 ; 4-byte Folded Spill
+; VI-NEXT: v_lshrrev_b32_e32 v33, 16, v22
+; VI-NEXT: buffer_store_dword v33, off, s[0:3], s32 offset:376 ; 4-byte Folded Spill
+; VI-NEXT: v_lshrrev_b32_e32 v33, 8, v22
+; VI-NEXT: buffer_store_dword v33, off, s[0:3], s32 offset:380 ; 4-byte Folded Spill
+; VI-NEXT: v_lshrrev_b32_e32 v33, 16, v21
+; VI-NEXT: buffer_store_dword v33, off, s[0:3], s32 offset:384 ; 4-byte Folded Spill
+; VI-NEXT: v_lshrrev_b32_e32 v33, 8, v21
+; VI-NEXT: buffer_store_dword v33, off, s[0:3], s32 offset:388 ; 4-byte Folded Spill
+; VI-NEXT: v_lshrrev_b32_e32 v33, 24, v20
+; VI-NEXT: buffer_store_dword v33, off, s[0:3], s32 offset:392 ; 4-byte Folded Spill
+; VI-NEXT: v_lshrrev_b32_e32 v33, 16, v20
; VI-NEXT: v_lshrrev_b64 v[44:45], 24, v[15:16]
-; VI-NEXT: buffer_store_dword v33, off, s[0:3], s32 offset:312 ; 4-byte Folded Spill
-; VI-NEXT: v_lshrrev_b32_e32 v33, 8, v1
-; VI-NEXT: buffer_store_dword v51, off, s[0:3], s32 offset:392 ; 4-byte Folded Spill
-; VI-NEXT: v_lshrrev_b32_e32 v51, 8, v17
-; VI-NEXT: buffer_store_dword v33, off, s[0:3], s32 offset:404 ; 4-byte Folded Spill
-; VI-NEXT: buffer_store_dword v51, off, s[0:3], s32 offset:396 ; 4-byte Folded Spill
+; VI-NEXT: buffer_store_dword v33, off, s[0:3], s32 offset:396 ; 4-byte Folded Spill
+; VI-NEXT: v_lshrrev_b32_e32 v33, 8, v20
+; VI-NEXT: buffer_store_dword v33, off, s[0:3], s32 offset:400 ; 4-byte Folded Spill
; VI-NEXT: buffer_store_dword v44, off, s[0:3], s32 offset:108 ; 4-byte Folded Spill
; VI-NEXT: buffer_store_dword v45, off, s[0:3], s32 offset:112 ; 4-byte Folded Spill
; VI-NEXT: v_lshrrev_b64 v[44:45], 24, v[13:14]
@@ -164913,17 +164963,15 @@ define <128 x i8> @bitcast_v64bf16_to_v128i8(<64 x bfloat> %a, i32 %b) {
; VI-NEXT: buffer_store_dword v44, off, s[0:3], s32 offset:188 ; 4-byte Folded Spill
; VI-NEXT: buffer_store_dword v45, off, s[0:3], s32 offset:192 ; 4-byte Folded Spill
; VI-NEXT: v_lshrrev_b64 v[44:45], 24, v[31:32]
-; VI-NEXT: v_lshrrev_b32_e32 v46, 24, v12
-; VI-NEXT: buffer_store_dword v44, off, s[0:3], s32 offset:76 ; 4-byte Folded Spill
-; VI-NEXT: buffer_store_dword v45, off, s[0:3], s32 offset:80 ; 4-byte Folded Spill
-; VI-NEXT: v_mov_b32_e32 v45, v46
; VI-NEXT: v_lshrrev_b64 v[46:47], 24, v[29:30]
-; VI-NEXT: v_lshrrev_b32_e32 v63, 8, v11
+; VI-NEXT: v_lshrrev_b32_e32 v51, 8, v11
; VI-NEXT: v_lshrrev_b32_e32 v50, 8, v31
+; VI-NEXT: buffer_store_dword v44, off, s[0:3], s32 offset:76 ; 4-byte Folded Spill
+; VI-NEXT: buffer_store_dword v45, off, s[0:3], s32 offset:80 ; 4-byte Folded Spill
; VI-NEXT: buffer_store_dword v46, off, s[0:3], s32 offset:84 ; 4-byte Folded Spill
; VI-NEXT: buffer_store_dword v47, off, s[0:3], s32 offset:88 ; 4-byte Folded Spill
-; VI-NEXT: v_mov_b32_e32 v46, v63
-; VI-NEXT: v_mov_b32_e32 v63, v50
+; VI-NEXT: v_mov_b32_e32 v46, v51
+; VI-NEXT: v_mov_b32_e32 v45, v50
; VI-NEXT: v_lshrrev_b64 v[50:51], 24, v[27:28]
; VI-NEXT: v_lshrrev_b32_e32 v56, 8, v10
; VI-NEXT: v_lshrrev_b32_e32 v57, 16, v9
@@ -164945,7 +164993,8 @@ define <128 x i8> @bitcast_v64bf16_to_v128i8(<64 x bfloat> %a, i32 %b) {
; VI-NEXT: buffer_store_dword v44, off, s[0:3], s32 offset:144 ; 4-byte Folded Spill
; VI-NEXT: v_lshrrev_b64 v[43:44], 24, v[17:18]
; VI-NEXT: v_lshrrev_b32_e32 v34, 16, v10
-; VI-NEXT: v_lshrrev_b32_e32 v36, 16, v12
+; VI-NEXT: buffer_store_dword v43, off, s[0:3], s32 offset:156 ; 4-byte Folded Spill
+; VI-NEXT: buffer_store_dword v44, off, s[0:3], s32 offset:160 ; 4-byte Folded Spill
; VI-NEXT: v_lshrrev_b32_e32 v49, 8, v12
; VI-NEXT: v_lshrrev_b32_e32 v39, 16, v11
; VI-NEXT: v_lshrrev_b32_e32 v40, 24, v10
@@ -164954,30 +165003,29 @@ define <128 x i8> @bitcast_v64bf16_to_v128i8(<64 x bfloat> %a, i32 %b) {
; VI-NEXT: v_lshrrev_b32_e32 v35, 8, v8
; VI-NEXT: v_lshrrev_b32_e32 v59, 16, v7
; VI-NEXT: v_lshrrev_b32_e32 v60, 8, v7
-; VI-NEXT: v_lshrrev_b32_e32 v54, 24, v6
-; VI-NEXT: v_lshrrev_b32_e32 v55, 16, v6
-; VI-NEXT: v_lshrrev_b32_e32 v38, 8, v6
+; VI-NEXT: v_lshrrev_b32_e32 v63, 8, v1
; VI-NEXT: v_lshrrev_b32_e32 v37, 24, v32
; VI-NEXT: v_lshrrev_b32_e32 v52, 16, v32
; VI-NEXT: v_lshrrev_b32_e32 v48, 8, v32
; VI-NEXT: v_lshrrev_b32_e32 v58, 16, v31
-; VI-NEXT: v_lshrrev_b32_e32 v61, 24, v30
-; VI-NEXT: v_lshrrev_b32_e32 v62, 16, v30
-; VI-NEXT: v_lshrrev_b32_e32 v53, 8, v30
-; VI-NEXT: v_lshrrev_b32_e32 v33, 16, v29
+; VI-NEXT: v_lshrrev_b32_e32 v36, 24, v30
+; VI-NEXT: v_lshrrev_b32_e32 v61, 16, v19
+; VI-NEXT: v_lshrrev_b32_e32 v62, 8, v19
+; VI-NEXT: v_lshrrev_b32_e32 v53, 24, v18
+; VI-NEXT: v_lshrrev_b32_e32 v54, 16, v18
+; VI-NEXT: v_lshrrev_b32_e32 v55, 8, v18
+; VI-NEXT: v_lshrrev_b32_e32 v33, 16, v17
+; VI-NEXT: v_lshrrev_b32_e32 v38, 8, v17
; VI-NEXT: v_mov_b32_e32 v47, v34
-; VI-NEXT: buffer_store_dword v43, off, s[0:3], s32 offset:156 ; 4-byte Folded Spill
-; VI-NEXT: buffer_store_dword v44, off, s[0:3], s32 offset:160 ; 4-byte Folded Spill
+; VI-NEXT: v_lshrrev_b32_e32 v43, 8, v2
; VI-NEXT: .LBB90_2: ; %Flow
; VI-NEXT: s_or_saveexec_b64 s[4:5], s[4:5]
-; VI-NEXT: buffer_load_dword v34, off, s[0:3], s32 offset:400 ; 4-byte Folded Reload
-; VI-NEXT: buffer_load_dword v44, off, s[0:3], s32 offset:404 ; 4-byte Folded Reload
+; VI-NEXT: v_mov_b32_e32 v34, v36
; VI-NEXT: s_xor_b64 exec, exec, s[4:5]
; VI-NEXT: s_cbranch_execz .LBB90_4
; VI-NEXT: ; %bb.3: ; %cmp.true
; VI-NEXT: v_lshlrev_b32_e32 v33, 16, v18
; VI-NEXT: v_add_f32_e32 v33, 0x40c00000, v33
-; VI-NEXT: s_waitcnt vmcnt(1)
; VI-NEXT: v_bfe_u32 v34, v33, 16, 1
; VI-NEXT: v_add_u32_e32 v34, vcc, v34, v33
; VI-NEXT: v_add_u32_e32 v34, vcc, 0x7fff, v34
@@ -165596,109 +165644,112 @@ define <128 x i8> @bitcast_v64bf16_to_v128i8(<64 x bfloat> %a, i32 %b) {
; VI-NEXT: buffer_store_dword v33, off, s[0:3], s32 offset:140 ; 4-byte Folded Spill
; VI-NEXT: buffer_store_dword v34, off, s[0:3], s32 offset:144 ; 4-byte Folded Spill
; VI-NEXT: v_lshrrev_b64 v[33:34], 24, v[17:18]
-; VI-NEXT: v_lshrrev_b32_e32 v43, 24, v28
; VI-NEXT: buffer_store_dword v33, off, s[0:3], s32 offset:156 ; 4-byte Folded Spill
; VI-NEXT: buffer_store_dword v34, off, s[0:3], s32 offset:160 ; 4-byte Folded Spill
-; VI-NEXT: buffer_store_dword v43, off, s[0:3], s32 offset:240 ; 4-byte Folded Spill
-; VI-NEXT: v_lshrrev_b32_e32 v43, 16, v28
-; VI-NEXT: buffer_store_dword v43, off, s[0:3], s32 offset:248 ; 4-byte Folded Spill
-; VI-NEXT: v_lshrrev_b32_e32 v43, 8, v28
-; VI-NEXT: buffer_store_dword v43, off, s[0:3], s32 offset:252 ; 4-byte Folded Spill
-; VI-NEXT: v_lshrrev_b32_e32 v43, 16, v27
-; VI-NEXT: buffer_store_dword v43, off, s[0:3], s32 offset:260 ; 4-byte Folded Spill
-; VI-NEXT: v_lshrrev_b32_e32 v43, 8, v27
-; VI-NEXT: buffer_store_dword v43, off, s[0:3], s32 offset:268 ; 4-byte Folded Spill
-; VI-NEXT: v_lshrrev_b32_e32 v43, 24, v26
-; VI-NEXT: buffer_store_dword v43, off, s[0:3], s32 offset:276 ; 4-byte Folded Spill
-; VI-NEXT: v_lshrrev_b32_e32 v43, 16, v26
-; VI-NEXT: buffer_store_dword v43, off, s[0:3], s32 offset:284 ; 4-byte Folded Spill
-; VI-NEXT: v_lshrrev_b32_e32 v43, 8, v26
; VI-NEXT: v_lshrrev_b32_e32 v33, 24, v16
-; VI-NEXT: buffer_store_dword v43, off, s[0:3], s32 offset:292 ; 4-byte Folded Spill
-; VI-NEXT: v_lshrrev_b32_e32 v43, 16, v25
; VI-NEXT: buffer_store_dword v33, off, s[0:3], s32 offset:196 ; 4-byte Folded Spill
; VI-NEXT: v_lshrrev_b32_e32 v33, 16, v16
-; VI-NEXT: buffer_store_dword v43, off, s[0:3], s32 offset:296 ; 4-byte Folded Spill
-; VI-NEXT: v_lshrrev_b32_e32 v43, 8, v25
; VI-NEXT: buffer_store_dword v33, off, s[0:3], s32 offset:200 ; 4-byte Folded Spill
; VI-NEXT: v_lshrrev_b32_e32 v33, 8, v16
-; VI-NEXT: buffer_store_dword v43, off, s[0:3], s32 offset:308 ; 4-byte Folded Spill
-; VI-NEXT: v_lshrrev_b32_e32 v43, 24, v24
; VI-NEXT: buffer_store_dword v33, off, s[0:3], s32 offset:204 ; 4-byte Folded Spill
; VI-NEXT: v_lshrrev_b32_e32 v33, 16, v15
-; VI-NEXT: buffer_store_dword v43, off, s[0:3], s32 offset:320 ; 4-byte Folded Spill
-; VI-NEXT: v_lshrrev_b32_e32 v43, 16, v24
; VI-NEXT: buffer_store_dword v33, off, s[0:3], s32 offset:208 ; 4-byte Folded Spill
; VI-NEXT: v_lshrrev_b32_e32 v33, 8, v15
-; VI-NEXT: buffer_store_dword v43, off, s[0:3], s32 offset:324 ; 4-byte Folded Spill
-; VI-NEXT: v_lshrrev_b32_e32 v43, 8, v24
; VI-NEXT: buffer_store_dword v33, off, s[0:3], s32 offset:212 ; 4-byte Folded Spill
; VI-NEXT: v_lshrrev_b32_e32 v33, 24, v14
-; VI-NEXT: buffer_store_dword v43, off, s[0:3], s32 offset:328 ; 4-byte Folded Spill
-; VI-NEXT: v_lshrrev_b32_e32 v43, 16, v23
; VI-NEXT: buffer_store_dword v33, off, s[0:3], s32 offset:216 ; 4-byte Folded Spill
; VI-NEXT: v_lshrrev_b32_e32 v33, 16, v14
-; VI-NEXT: buffer_store_dword v43, off, s[0:3], s32 offset:332 ; 4-byte Folded Spill
-; VI-NEXT: v_lshrrev_b32_e32 v43, 8, v23
; VI-NEXT: buffer_store_dword v33, off, s[0:3], s32 offset:220 ; 4-byte Folded Spill
; VI-NEXT: v_lshrrev_b32_e32 v33, 8, v14
-; VI-NEXT: buffer_store_dword v43, off, s[0:3], s32 offset:336 ; 4-byte Folded Spill
-; VI-NEXT: v_lshrrev_b32_e32 v43, 24, v22
; VI-NEXT: buffer_store_dword v33, off, s[0:3], s32 offset:224 ; 4-byte Folded Spill
; VI-NEXT: v_lshrrev_b32_e32 v33, 16, v13
-; VI-NEXT: buffer_store_dword v43, off, s[0:3], s32 offset:340 ; 4-byte Folded Spill
-; VI-NEXT: v_lshrrev_b32_e32 v43, 16, v22
; VI-NEXT: buffer_store_dword v33, off, s[0:3], s32 offset:228 ; 4-byte Folded Spill
; VI-NEXT: v_lshrrev_b32_e32 v33, 8, v13
-; VI-NEXT: buffer_store_dword v43, off, s[0:3], s32 offset:344 ; 4-byte Folded Spill
-; VI-NEXT: v_lshrrev_b32_e32 v43, 8, v22
; VI-NEXT: buffer_store_dword v33, off, s[0:3], s32 offset:232 ; 4-byte Folded Spill
-; VI-NEXT: v_lshrrev_b32_e32 v33, 16, v5
-; VI-NEXT: buffer_store_dword v43, off, s[0:3], s32 offset:348 ; 4-byte Folded Spill
-; VI-NEXT: v_lshrrev_b32_e32 v43, 16, v21
+; VI-NEXT: v_lshrrev_b32_e32 v33, 24, v12
; VI-NEXT: buffer_store_dword v33, off, s[0:3], s32 offset:236 ; 4-byte Folded Spill
+; VI-NEXT: v_lshrrev_b32_e32 v33, 16, v12
+; VI-NEXT: buffer_store_dword v33, off, s[0:3], s32 offset:360 ; 4-byte Folded Spill
+; VI-NEXT: v_lshrrev_b32_e32 v33, 24, v6
+; VI-NEXT: buffer_store_dword v33, off, s[0:3], s32 offset:248 ; 4-byte Folded Spill
+; VI-NEXT: v_lshrrev_b32_e32 v33, 16, v6
+; VI-NEXT: buffer_store_dword v33, off, s[0:3], s32 offset:252 ; 4-byte Folded Spill
+; VI-NEXT: v_lshrrev_b32_e32 v33, 8, v6
+; VI-NEXT: buffer_store_dword v33, off, s[0:3], s32 offset:260 ; 4-byte Folded Spill
+; VI-NEXT: v_lshrrev_b32_e32 v33, 16, v5
+; VI-NEXT: buffer_store_dword v33, off, s[0:3], s32 offset:268 ; 4-byte Folded Spill
; VI-NEXT: v_lshrrev_b32_e32 v33, 8, v5
-; VI-NEXT: buffer_store_dword v43, off, s[0:3], s32 offset:352 ; 4-byte Folded Spill
-; VI-NEXT: v_lshrrev_b32_e32 v43, 8, v21
-; VI-NEXT: buffer_store_dword v33, off, s[0:3], s32 offset:244 ; 4-byte Folded Spill
+; VI-NEXT: buffer_store_dword v33, off, s[0:3], s32 offset:276 ; 4-byte Folded Spill
; VI-NEXT: v_lshrrev_b32_e32 v33, 24, v4
-; VI-NEXT: buffer_store_dword v43, off, s[0:3], s32 offset:356 ; 4-byte Folded Spill
-; VI-NEXT: v_lshrrev_b32_e32 v43, 24, v20
-; VI-NEXT: buffer_store_dword v33, off, s[0:3], s32 offset:256 ; 4-byte Folded Spill
+; VI-NEXT: buffer_store_dword v33, off, s[0:3], s32 offset:288 ; 4-byte Folded Spill
; VI-NEXT: v_lshrrev_b32_e32 v33, 16, v4
-; VI-NEXT: buffer_store_dword v43, off, s[0:3], s32 offset:360 ; 4-byte Folded Spill
-; VI-NEXT: v_lshrrev_b32_e32 v43, 16, v20
-; VI-NEXT: buffer_store_dword v33, off, s[0:3], s32 offset:264 ; 4-byte Folded Spill
+; VI-NEXT: buffer_store_dword v33, off, s[0:3], s32 offset:296 ; 4-byte Folded Spill
; VI-NEXT: v_lshrrev_b32_e32 v33, 8, v4
-; VI-NEXT: buffer_store_dword v43, off, s[0:3], s32 offset:364 ; 4-byte Folded Spill
-; VI-NEXT: v_lshrrev_b32_e32 v43, 8, v20
-; VI-NEXT: buffer_store_dword v33, off, s[0:3], s32 offset:272 ; 4-byte Folded Spill
+; VI-NEXT: buffer_store_dword v33, off, s[0:3], s32 offset:304 ; 4-byte Folded Spill
; VI-NEXT: v_lshrrev_b32_e32 v33, 16, v3
-; VI-NEXT: buffer_store_dword v43, off, s[0:3], s32 offset:368 ; 4-byte Folded Spill
-; VI-NEXT: v_lshrrev_b32_e32 v43, 16, v19
-; VI-NEXT: buffer_store_dword v33, off, s[0:3], s32 offset:280 ; 4-byte Folded Spill
+; VI-NEXT: buffer_store_dword v33, off, s[0:3], s32 offset:312 ; 4-byte Folded Spill
; VI-NEXT: v_lshrrev_b32_e32 v33, 8, v3
-; VI-NEXT: buffer_store_dword v43, off, s[0:3], s32 offset:372 ; 4-byte Folded Spill
-; VI-NEXT: v_lshrrev_b32_e32 v43, 8, v19
-; VI-NEXT: buffer_store_dword v33, off, s[0:3], s32 offset:288 ; 4-byte Folded Spill
+; VI-NEXT: buffer_store_dword v33, off, s[0:3], s32 offset:320 ; 4-byte Folded Spill
; VI-NEXT: v_lshrrev_b32_e32 v33, 24, v2
-; VI-NEXT: buffer_store_dword v43, off, s[0:3], s32 offset:376 ; 4-byte Folded Spill
-; VI-NEXT: v_lshrrev_b32_e32 v43, 24, v18
-; VI-NEXT: buffer_store_dword v33, off, s[0:3], s32 offset:300 ; 4-byte Folded Spill
+; VI-NEXT: buffer_store_dword v33, off, s[0:3], s32 offset:332 ; 4-byte Folded Spill
; VI-NEXT: v_lshrrev_b32_e32 v33, 16, v2
-; VI-NEXT: buffer_store_dword v43, off, s[0:3], s32 offset:380 ; 4-byte Folded Spill
-; VI-NEXT: v_lshrrev_b32_e32 v43, 16, v18
-; VI-NEXT: buffer_store_dword v33, off, s[0:3], s32 offset:304 ; 4-byte Folded Spill
-; VI-NEXT: v_lshrrev_b32_e32 v33, 8, v2
-; VI-NEXT: buffer_store_dword v43, off, s[0:3], s32 offset:384 ; 4-byte Folded Spill
-; VI-NEXT: v_lshrrev_b32_e32 v43, 8, v18
-; VI-NEXT: v_lshrrev_b64 v[56:57], 24, v[25:26]
-; VI-NEXT: buffer_store_dword v33, off, s[0:3], s32 offset:316 ; 4-byte Folded Spill
+; VI-NEXT: buffer_store_dword v33, off, s[0:3], s32 offset:336 ; 4-byte Folded Spill
; VI-NEXT: v_lshrrev_b32_e32 v33, 16, v1
-; VI-NEXT: buffer_store_dword v43, off, s[0:3], s32 offset:388 ; 4-byte Folded Spill
-; VI-NEXT: v_lshrrev_b32_e32 v43, 16, v17
-; VI-NEXT: v_lshrrev_b32_e32 v45, 24, v12
-; VI-NEXT: v_lshrrev_b32_e32 v36, 16, v12
+; VI-NEXT: buffer_store_dword v33, off, s[0:3], s32 offset:344 ; 4-byte Folded Spill
+; VI-NEXT: v_lshrrev_b32_e32 v33, 16, v30
+; VI-NEXT: buffer_store_dword v33, off, s[0:3], s32 offset:240 ; 4-byte Folded Spill
+; VI-NEXT: v_lshrrev_b32_e32 v33, 8, v30
+; VI-NEXT: buffer_store_dword v33, off, s[0:3], s32 offset:244 ; 4-byte Folded Spill
+; VI-NEXT: v_lshrrev_b32_e32 v33, 16, v29
+; VI-NEXT: buffer_store_dword v33, off, s[0:3], s32 offset:256 ; 4-byte Folded Spill
+; VI-NEXT: v_lshrrev_b32_e32 v33, 8, v29
+; VI-NEXT: buffer_store_dword v33, off, s[0:3], s32 offset:264 ; 4-byte Folded Spill
+; VI-NEXT: v_lshrrev_b32_e32 v33, 24, v28
+; VI-NEXT: buffer_store_dword v33, off, s[0:3], s32 offset:272 ; 4-byte Folded Spill
+; VI-NEXT: v_lshrrev_b32_e32 v33, 16, v28
+; VI-NEXT: buffer_store_dword v33, off, s[0:3], s32 offset:280 ; 4-byte Folded Spill
+; VI-NEXT: v_lshrrev_b32_e32 v33, 8, v28
+; VI-NEXT: buffer_store_dword v33, off, s[0:3], s32 offset:284 ; 4-byte Folded Spill
+; VI-NEXT: v_lshrrev_b32_e32 v33, 16, v27
+; VI-NEXT: buffer_store_dword v33, off, s[0:3], s32 offset:292 ; 4-byte Folded Spill
+; VI-NEXT: v_lshrrev_b32_e32 v33, 8, v27
+; VI-NEXT: buffer_store_dword v33, off, s[0:3], s32 offset:300 ; 4-byte Folded Spill
+; VI-NEXT: v_lshrrev_b32_e32 v33, 24, v26
+; VI-NEXT: buffer_store_dword v33, off, s[0:3], s32 offset:308 ; 4-byte Folded Spill
+; VI-NEXT: v_lshrrev_b32_e32 v33, 16, v26
+; VI-NEXT: buffer_store_dword v33, off, s[0:3], s32 offset:316 ; 4-byte Folded Spill
+; VI-NEXT: v_lshrrev_b32_e32 v33, 8, v26
+; VI-NEXT: buffer_store_dword v33, off, s[0:3], s32 offset:324 ; 4-byte Folded Spill
+; VI-NEXT: v_lshrrev_b32_e32 v33, 16, v25
+; VI-NEXT: buffer_store_dword v33, off, s[0:3], s32 offset:328 ; 4-byte Folded Spill
+; VI-NEXT: v_lshrrev_b32_e32 v33, 8, v25
+; VI-NEXT: buffer_store_dword v33, off, s[0:3], s32 offset:340 ; 4-byte Folded Spill
+; VI-NEXT: v_lshrrev_b32_e32 v33, 24, v24
+; VI-NEXT: buffer_store_dword v33, off, s[0:3], s32 offset:348 ; 4-byte Folded Spill
+; VI-NEXT: v_lshrrev_b32_e32 v33, 16, v24
+; VI-NEXT: buffer_store_dword v33, off, s[0:3], s32 offset:352 ; 4-byte Folded Spill
+; VI-NEXT: v_lshrrev_b32_e32 v33, 8, v24
+; VI-NEXT: buffer_store_dword v33, off, s[0:3], s32 offset:356 ; 4-byte Folded Spill
+; VI-NEXT: v_lshrrev_b32_e32 v33, 16, v23
+; VI-NEXT: buffer_store_dword v33, off, s[0:3], s32 offset:364 ; 4-byte Folded Spill
+; VI-NEXT: v_lshrrev_b32_e32 v33, 8, v23
+; VI-NEXT: buffer_store_dword v33, off, s[0:3], s32 offset:368 ; 4-byte Folded Spill
+; VI-NEXT: v_lshrrev_b32_e32 v33, 24, v22
+; VI-NEXT: buffer_store_dword v33, off, s[0:3], s32 offset:372 ; 4-byte Folded Spill
+; VI-NEXT: v_lshrrev_b32_e32 v33, 16, v22
+; VI-NEXT: buffer_store_dword v33, off, s[0:3], s32 offset:376 ; 4-byte Folded Spill
+; VI-NEXT: v_lshrrev_b32_e32 v33, 8, v22
+; VI-NEXT: buffer_store_dword v33, off, s[0:3], s32 offset:380 ; 4-byte Folded Spill
+; VI-NEXT: v_lshrrev_b32_e32 v33, 16, v21
+; VI-NEXT: buffer_store_dword v33, off, s[0:3], s32 offset:384 ; 4-byte Folded Spill
+; VI-NEXT: v_lshrrev_b32_e32 v33, 8, v21
+; VI-NEXT: buffer_store_dword v33, off, s[0:3], s32 offset:388 ; 4-byte Folded Spill
+; VI-NEXT: v_lshrrev_b32_e32 v33, 24, v20
+; VI-NEXT: buffer_store_dword v33, off, s[0:3], s32 offset:392 ; 4-byte Folded Spill
+; VI-NEXT: v_lshrrev_b32_e32 v33, 16, v20
+; VI-NEXT: v_lshrrev_b64 v[56:57], 24, v[25:26]
+; VI-NEXT: buffer_store_dword v33, off, s[0:3], s32 offset:396 ; 4-byte Folded Spill
+; VI-NEXT: v_lshrrev_b32_e32 v33, 8, v20
; VI-NEXT: v_lshrrev_b32_e32 v49, 8, v12
; VI-NEXT: v_lshrrev_b32_e32 v39, 16, v11
; VI-NEXT: v_lshrrev_b32_e32 v46, 8, v11
@@ -165712,69 +165763,63 @@ define <128 x i8> @bitcast_v64bf16_to_v128i8(<64 x bfloat> %a, i32 %b) {
; VI-NEXT: v_lshrrev_b32_e32 v35, 8, v8
; VI-NEXT: v_lshrrev_b32_e32 v59, 16, v7
; VI-NEXT: v_lshrrev_b32_e32 v60, 8, v7
-; VI-NEXT: v_lshrrev_b32_e32 v54, 24, v6
-; VI-NEXT: v_lshrrev_b32_e32 v55, 16, v6
-; VI-NEXT: v_lshrrev_b32_e32 v38, 8, v6
-; VI-NEXT: buffer_store_dword v33, off, s[0:3], s32 offset:312 ; 4-byte Folded Spill
-; VI-NEXT: s_waitcnt vmcnt(14)
-; VI-NEXT: v_lshrrev_b32_e32 v44, 8, v1
+; VI-NEXT: v_lshrrev_b32_e32 v43, 8, v2
+; VI-NEXT: v_lshrrev_b32_e32 v63, 8, v1
; VI-NEXT: v_lshrrev_b32_e32 v37, 24, v32
; VI-NEXT: v_lshrrev_b32_e32 v52, 16, v32
; VI-NEXT: v_lshrrev_b32_e32 v48, 8, v32
; VI-NEXT: v_lshrrev_b32_e32 v58, 16, v31
-; VI-NEXT: v_lshrrev_b32_e32 v63, 8, v31
-; VI-NEXT: v_lshrrev_b32_e32 v61, 24, v30
-; VI-NEXT: v_lshrrev_b32_e32 v62, 16, v30
-; VI-NEXT: v_lshrrev_b32_e32 v53, 8, v30
-; VI-NEXT: v_lshrrev_b32_e32 v33, 16, v29
-; VI-NEXT: v_lshrrev_b32_e32 v34, 8, v29
-; VI-NEXT: buffer_store_dword v43, off, s[0:3], s32 offset:392 ; 4-byte Folded Spill
-; VI-NEXT: v_lshrrev_b32_e32 v43, 8, v17
-; VI-NEXT: buffer_store_dword v43, off, s[0:3], s32 offset:396 ; 4-byte Folded Spill
+; VI-NEXT: v_lshrrev_b32_e32 v45, 8, v31
+; VI-NEXT: v_lshrrev_b32_e32 v34, 24, v30
+; VI-NEXT: buffer_store_dword v33, off, s[0:3], s32 offset:400 ; 4-byte Folded Spill
+; VI-NEXT: v_lshrrev_b32_e32 v61, 16, v19
+; VI-NEXT: v_lshrrev_b32_e32 v62, 8, v19
+; VI-NEXT: v_lshrrev_b32_e32 v53, 24, v18
+; VI-NEXT: v_lshrrev_b32_e32 v54, 16, v18
+; VI-NEXT: v_lshrrev_b32_e32 v55, 8, v18
+; VI-NEXT: v_lshrrev_b32_e32 v33, 16, v17
+; VI-NEXT: v_lshrrev_b32_e32 v38, 8, v17
; VI-NEXT: .LBB90_4: ; %end
; VI-NEXT: s_or_b64 exec, exec, s[4:5]
-; VI-NEXT: buffer_load_dword v43, off, s[0:3], s32 offset:316 ; 4-byte Folded Reload
-; VI-NEXT: s_waitcnt vmcnt(1)
-; VI-NEXT: v_lshlrev_b16_e32 v44, 8, v44
+; VI-NEXT: v_lshlrev_b16_e32 v44, 8, v63
; VI-NEXT: v_or_b32_sdwa v1, v1, v44 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
-; VI-NEXT: s_waitcnt vmcnt(0)
; VI-NEXT: v_lshlrev_b16_e32 v44, 8, v43
; VI-NEXT: v_or_b32_sdwa v2, v2, v44 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
; VI-NEXT: buffer_load_dword v43, off, s[0:3], s32 offset:188 ; 4-byte Folded Reload
; VI-NEXT: buffer_load_dword v44, off, s[0:3], s32 offset:192 ; 4-byte Folded Reload
-; VI-NEXT: buffer_load_dword v44, off, s[0:3], s32 offset:312 ; 4-byte Folded Reload
+; VI-NEXT: buffer_load_dword v36, off, s[0:3], s32 offset:344 ; 4-byte Folded Reload
; VI-NEXT: s_waitcnt vmcnt(2)
; VI-NEXT: v_lshlrev_b16_e32 v43, 8, v43
; VI-NEXT: s_waitcnt vmcnt(0)
-; VI-NEXT: v_or_b32_sdwa v43, v44, v43 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; VI-NEXT: v_or_b32_sdwa v43, v36, v43 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
; VI-NEXT: v_or_b32_sdwa v1, v1, v43 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_0 src1_sel:DWORD
; VI-NEXT: buffer_store_dword v1, v0, s[0:3], 0 offen
-; VI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:300 ; 4-byte Folded Reload
-; VI-NEXT: buffer_load_dword v43, off, s[0:3], s32 offset:304 ; 4-byte Folded Reload
+; VI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:332 ; 4-byte Folded Reload
+; VI-NEXT: buffer_load_dword v36, off, s[0:3], s32 offset:336 ; 4-byte Folded Reload
; VI-NEXT: s_waitcnt vmcnt(1)
; VI-NEXT: v_lshlrev_b16_e32 v1, 8, v1
; VI-NEXT: s_waitcnt vmcnt(0)
-; VI-NEXT: v_or_b32_sdwa v1, v43, v1 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; VI-NEXT: v_or_b32_sdwa v1, v36, v1 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
; VI-NEXT: v_or_b32_sdwa v1, v2, v1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_0 src1_sel:DWORD
; VI-NEXT: v_add_u32_e32 v2, vcc, 4, v0
; VI-NEXT: buffer_store_dword v1, v2, s[0:3], 0 offen
; VI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:180 ; 4-byte Folded Reload
; VI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:184 ; 4-byte Folded Reload
-; VI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:280 ; 4-byte Folded Reload
+; VI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:312 ; 4-byte Folded Reload
; VI-NEXT: s_waitcnt vmcnt(2)
; VI-NEXT: v_lshlrev_b16_e32 v1, 8, v1
; VI-NEXT: s_waitcnt vmcnt(0)
; VI-NEXT: v_or_b32_sdwa v1, v2, v1 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
-; VI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:288 ; 4-byte Folded Reload
+; VI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:320 ; 4-byte Folded Reload
; VI-NEXT: s_waitcnt vmcnt(0)
; VI-NEXT: v_lshlrev_b16_e32 v2, 8, v2
; VI-NEXT: v_or_b32_sdwa v2, v3, v2 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
; VI-NEXT: v_or_b32_sdwa v1, v2, v1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_0 src1_sel:DWORD
; VI-NEXT: v_add_u32_e32 v2, vcc, 8, v0
; VI-NEXT: buffer_store_dword v1, v2, s[0:3], 0 offen
-; VI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:272 ; 4-byte Folded Reload
-; VI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:256 ; 4-byte Folded Reload
-; VI-NEXT: buffer_load_dword v3, off, s[0:3], s32 offset:264 ; 4-byte Folded Reload
+; VI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:304 ; 4-byte Folded Reload
+; VI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:288 ; 4-byte Folded Reload
+; VI-NEXT: buffer_load_dword v3, off, s[0:3], s32 offset:296 ; 4-byte Folded Reload
; VI-NEXT: s_waitcnt vmcnt(2)
; VI-NEXT: v_lshlrev_b16_e32 v1, 8, v1
; VI-NEXT: s_waitcnt vmcnt(1)
@@ -165787,22 +165832,28 @@ define <128 x i8> @bitcast_v64bf16_to_v128i8(<64 x bfloat> %a, i32 %b) {
; VI-NEXT: buffer_store_dword v1, v2, s[0:3], 0 offen
; VI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:172 ; 4-byte Folded Reload
; VI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:176 ; 4-byte Folded Reload
-; VI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:236 ; 4-byte Folded Reload
+; VI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:268 ; 4-byte Folded Reload
; VI-NEXT: s_waitcnt vmcnt(2)
; VI-NEXT: v_lshlrev_b16_e32 v1, 8, v1
; VI-NEXT: s_waitcnt vmcnt(0)
; VI-NEXT: v_or_b32_sdwa v1, v2, v1 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
-; VI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:244 ; 4-byte Folded Reload
+; VI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:276 ; 4-byte Folded Reload
; VI-NEXT: s_waitcnt vmcnt(0)
; VI-NEXT: v_lshlrev_b16_e32 v2, 8, v2
; VI-NEXT: v_or_b32_sdwa v2, v5, v2 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
; VI-NEXT: v_or_b32_sdwa v1, v2, v1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_0 src1_sel:DWORD
; VI-NEXT: v_add_u32_e32 v2, vcc, 16, v0
; VI-NEXT: buffer_store_dword v1, v2, s[0:3], 0 offen
-; VI-NEXT: v_lshlrev_b16_e32 v1, 8, v38
-; VI-NEXT: v_lshlrev_b16_e32 v2, 8, v54
+; VI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:260 ; 4-byte Folded Reload
+; VI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:248 ; 4-byte Folded Reload
+; VI-NEXT: buffer_load_dword v3, off, s[0:3], s32 offset:252 ; 4-byte Folded Reload
+; VI-NEXT: s_waitcnt vmcnt(2)
+; VI-NEXT: v_lshlrev_b16_e32 v1, 8, v1
+; VI-NEXT: s_waitcnt vmcnt(1)
+; VI-NEXT: v_lshlrev_b16_e32 v2, 8, v2
; VI-NEXT: v_or_b32_sdwa v1, v6, v1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
-; VI-NEXT: v_or_b32_sdwa v2, v55, v2 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; VI-NEXT: s_waitcnt vmcnt(0)
+; VI-NEXT: v_or_b32_sdwa v2, v3, v2 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
; VI-NEXT: v_or_b32_sdwa v1, v1, v2 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_0 src1_sel:DWORD
; VI-NEXT: v_add_u32_e32 v2, vcc, 20, v0
; VI-NEXT: buffer_store_dword v1, v2, s[0:3], 0 offen
@@ -165850,10 +165901,14 @@ define <128 x i8> @bitcast_v64bf16_to_v128i8(<64 x bfloat> %a, i32 %b) {
; VI-NEXT: v_or_b32_sdwa v1, v2, v1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_0 src1_sel:DWORD
; VI-NEXT: v_add_u32_e32 v2, vcc, 40, v0
; VI-NEXT: buffer_store_dword v1, v2, s[0:3], 0 offen
+; VI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:236 ; 4-byte Folded Reload
+; VI-NEXT: buffer_load_dword v3, off, s[0:3], s32 offset:360 ; 4-byte Folded Reload
; VI-NEXT: v_lshlrev_b16_e32 v1, 8, v49
-; VI-NEXT: v_lshlrev_b16_e32 v2, 8, v45
; VI-NEXT: v_or_b32_sdwa v1, v12, v1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
-; VI-NEXT: v_or_b32_sdwa v2, v36, v2 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; VI-NEXT: s_waitcnt vmcnt(1)
+; VI-NEXT: v_lshlrev_b16_e32 v2, 8, v2
+; VI-NEXT: s_waitcnt vmcnt(0)
+; VI-NEXT: v_or_b32_sdwa v2, v3, v2 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
; VI-NEXT: v_or_b32_sdwa v1, v1, v2 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_0 src1_sel:DWORD
; VI-NEXT: v_add_u32_e32 v2, vcc, 44, v0
; VI-NEXT: buffer_store_dword v1, v2, s[0:3], 0 offen
@@ -165913,48 +165968,34 @@ define <128 x i8> @bitcast_v64bf16_to_v128i8(<64 x bfloat> %a, i32 %b) {
; VI-NEXT: buffer_store_dword v1, v2, s[0:3], 0 offen
; VI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:156 ; 4-byte Folded Reload
; VI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:160 ; 4-byte Folded Reload
-; VI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:392 ; 4-byte Folded Reload
-; VI-NEXT: s_waitcnt vmcnt(2)
-; VI-NEXT: v_lshlrev_b16_e32 v1, 8, v1
; VI-NEXT: s_waitcnt vmcnt(0)
-; VI-NEXT: v_or_b32_sdwa v1, v2, v1 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
-; VI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:396 ; 4-byte Folded Reload
-; VI-NEXT: s_waitcnt vmcnt(0)
-; VI-NEXT: v_lshlrev_b16_e32 v2, 8, v2
+; VI-NEXT: v_lshlrev_b16_e32 v2, 8, v38
; VI-NEXT: v_or_b32_sdwa v2, v17, v2 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; VI-NEXT: v_lshlrev_b16_e32 v1, 8, v1
+; VI-NEXT: v_or_b32_sdwa v1, v33, v1 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
; VI-NEXT: v_or_b32_sdwa v1, v2, v1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_0 src1_sel:DWORD
; VI-NEXT: v_add_u32_e32 v2, vcc, 64, v0
; VI-NEXT: buffer_store_dword v1, v2, s[0:3], 0 offen
-; VI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:388 ; 4-byte Folded Reload
-; VI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:380 ; 4-byte Folded Reload
-; VI-NEXT: buffer_load_dword v3, off, s[0:3], s32 offset:384 ; 4-byte Folded Reload
-; VI-NEXT: s_waitcnt vmcnt(2)
-; VI-NEXT: v_lshlrev_b16_e32 v1, 8, v1
-; VI-NEXT: s_waitcnt vmcnt(1)
-; VI-NEXT: v_lshlrev_b16_e32 v2, 8, v2
+; VI-NEXT: v_lshlrev_b16_e32 v1, 8, v55
+; VI-NEXT: v_lshlrev_b16_e32 v2, 8, v53
; VI-NEXT: v_or_b32_sdwa v1, v18, v1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
-; VI-NEXT: s_waitcnt vmcnt(0)
-; VI-NEXT: v_or_b32_sdwa v2, v3, v2 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; VI-NEXT: v_or_b32_sdwa v2, v54, v2 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
; VI-NEXT: v_or_b32_sdwa v1, v1, v2 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_0 src1_sel:DWORD
; VI-NEXT: v_add_u32_e32 v2, vcc, 0x44, v0
; VI-NEXT: buffer_store_dword v1, v2, s[0:3], 0 offen
; VI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:140 ; 4-byte Folded Reload
; VI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:144 ; 4-byte Folded Reload
-; VI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:372 ; 4-byte Folded Reload
-; VI-NEXT: s_waitcnt vmcnt(2)
-; VI-NEXT: v_lshlrev_b16_e32 v1, 8, v1
-; VI-NEXT: s_waitcnt vmcnt(0)
-; VI-NEXT: v_or_b32_sdwa v1, v2, v1 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
-; VI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:376 ; 4-byte Folded Reload
; VI-NEXT: s_waitcnt vmcnt(0)
-; VI-NEXT: v_lshlrev_b16_e32 v2, 8, v2
+; VI-NEXT: v_lshlrev_b16_e32 v2, 8, v62
; VI-NEXT: v_or_b32_sdwa v2, v19, v2 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; VI-NEXT: v_lshlrev_b16_e32 v1, 8, v1
+; VI-NEXT: v_or_b32_sdwa v1, v61, v1 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
; VI-NEXT: v_or_b32_sdwa v1, v2, v1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_0 src1_sel:DWORD
; VI-NEXT: v_add_u32_e32 v2, vcc, 0x48, v0
; VI-NEXT: buffer_store_dword v1, v2, s[0:3], 0 offen
-; VI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:368 ; 4-byte Folded Reload
-; VI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:360 ; 4-byte Folded Reload
-; VI-NEXT: buffer_load_dword v3, off, s[0:3], s32 offset:364 ; 4-byte Folded Reload
+; VI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:400 ; 4-byte Folded Reload
+; VI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:392 ; 4-byte Folded Reload
+; VI-NEXT: buffer_load_dword v3, off, s[0:3], s32 offset:396 ; 4-byte Folded Reload
; VI-NEXT: s_waitcnt vmcnt(2)
; VI-NEXT: v_lshlrev_b16_e32 v1, 8, v1
; VI-NEXT: s_waitcnt vmcnt(1)
@@ -165965,10 +166006,10 @@ define <128 x i8> @bitcast_v64bf16_to_v128i8(<64 x bfloat> %a, i32 %b) {
; VI-NEXT: v_or_b32_sdwa v1, v1, v2 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_0 src1_sel:DWORD
; VI-NEXT: v_add_u32_e32 v2, vcc, 0x4c, v0
; VI-NEXT: buffer_store_dword v1, v2, s[0:3], 0 offen
-; VI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:356 ; 4-byte Folded Reload
+; VI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:388 ; 4-byte Folded Reload
; VI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:116 ; 4-byte Folded Reload
; VI-NEXT: buffer_load_dword v3, off, s[0:3], s32 offset:120 ; 4-byte Folded Reload
-; VI-NEXT: buffer_load_dword v3, off, s[0:3], s32 offset:352 ; 4-byte Folded Reload
+; VI-NEXT: buffer_load_dword v3, off, s[0:3], s32 offset:384 ; 4-byte Folded Reload
; VI-NEXT: s_waitcnt vmcnt(2)
; VI-NEXT: v_lshlrev_b16_e32 v2, 8, v2
; VI-NEXT: v_lshlrev_b16_e32 v1, 8, v1
@@ -165978,9 +166019,9 @@ define <128 x i8> @bitcast_v64bf16_to_v128i8(<64 x bfloat> %a, i32 %b) {
; VI-NEXT: v_or_b32_sdwa v1, v1, v2 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_0 src1_sel:DWORD
; VI-NEXT: v_add_u32_e32 v2, vcc, 0x50, v0
; VI-NEXT: buffer_store_dword v1, v2, s[0:3], 0 offen
-; VI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:348 ; 4-byte Folded Reload
-; VI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:340 ; 4-byte Folded Reload
-; VI-NEXT: buffer_load_dword v3, off, s[0:3], s32 offset:344 ; 4-byte Folded Reload
+; VI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:380 ; 4-byte Folded Reload
+; VI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:372 ; 4-byte Folded Reload
+; VI-NEXT: buffer_load_dword v3, off, s[0:3], s32 offset:376 ; 4-byte Folded Reload
; VI-NEXT: s_waitcnt vmcnt(2)
; VI-NEXT: v_lshlrev_b16_e32 v1, 8, v1
; VI-NEXT: s_waitcnt vmcnt(1)
@@ -165991,10 +166032,10 @@ define <128 x i8> @bitcast_v64bf16_to_v128i8(<64 x bfloat> %a, i32 %b) {
; VI-NEXT: v_or_b32_sdwa v1, v1, v2 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_0 src1_sel:DWORD
; VI-NEXT: v_add_u32_e32 v2, vcc, 0x54, v0
; VI-NEXT: buffer_store_dword v1, v2, s[0:3], 0 offen
-; VI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:336 ; 4-byte Folded Reload
+; VI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:368 ; 4-byte Folded Reload
; VI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:100 ; 4-byte Folded Reload
; VI-NEXT: buffer_load_dword v3, off, s[0:3], s32 offset:104 ; 4-byte Folded Reload
-; VI-NEXT: buffer_load_dword v3, off, s[0:3], s32 offset:332 ; 4-byte Folded Reload
+; VI-NEXT: buffer_load_dword v3, off, s[0:3], s32 offset:364 ; 4-byte Folded Reload
; VI-NEXT: s_waitcnt vmcnt(2)
; VI-NEXT: v_lshlrev_b16_e32 v2, 8, v2
; VI-NEXT: v_lshlrev_b16_e32 v1, 8, v1
@@ -166004,9 +166045,9 @@ define <128 x i8> @bitcast_v64bf16_to_v128i8(<64 x bfloat> %a, i32 %b) {
; VI-NEXT: v_or_b32_sdwa v1, v1, v2 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_0 src1_sel:DWORD
; VI-NEXT: v_add_u32_e32 v2, vcc, 0x58, v0
; VI-NEXT: buffer_store_dword v1, v2, s[0:3], 0 offen
-; VI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:328 ; 4-byte Folded Reload
-; VI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:320 ; 4-byte Folded Reload
-; VI-NEXT: buffer_load_dword v3, off, s[0:3], s32 offset:324 ; 4-byte Folded Reload
+; VI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:356 ; 4-byte Folded Reload
+; VI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:348 ; 4-byte Folded Reload
+; VI-NEXT: buffer_load_dword v3, off, s[0:3], s32 offset:352 ; 4-byte Folded Reload
; VI-NEXT: s_waitcnt vmcnt(2)
; VI-NEXT: v_lshlrev_b16_e32 v1, 8, v1
; VI-NEXT: s_waitcnt vmcnt(1)
@@ -166017,8 +166058,8 @@ define <128 x i8> @bitcast_v64bf16_to_v128i8(<64 x bfloat> %a, i32 %b) {
; VI-NEXT: v_or_b32_sdwa v1, v1, v2 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_0 src1_sel:DWORD
; VI-NEXT: v_add_u32_e32 v2, vcc, 0x5c, v0
; VI-NEXT: buffer_store_dword v1, v2, s[0:3], 0 offen
-; VI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:308 ; 4-byte Folded Reload
-; VI-NEXT: buffer_load_dword v3, off, s[0:3], s32 offset:296 ; 4-byte Folded Reload
+; VI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:340 ; 4-byte Folded Reload
+; VI-NEXT: buffer_load_dword v3, off, s[0:3], s32 offset:328 ; 4-byte Folded Reload
; VI-NEXT: v_lshlrev_b16_e32 v2, 8, v56
; VI-NEXT: s_waitcnt vmcnt(1)
; VI-NEXT: v_lshlrev_b16_e32 v1, 8, v1
@@ -166028,9 +166069,9 @@ define <128 x i8> @bitcast_v64bf16_to_v128i8(<64 x bfloat> %a, i32 %b) {
; VI-NEXT: v_or_b32_sdwa v1, v1, v2 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_0 src1_sel:DWORD
; VI-NEXT: v_add_u32_e32 v2, vcc, 0x60, v0
; VI-NEXT: buffer_store_dword v1, v2, s[0:3], 0 offen
-; VI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:292 ; 4-byte Folded Reload
-; VI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:276 ; 4-byte Folded Reload
-; VI-NEXT: buffer_load_dword v3, off, s[0:3], s32 offset:284 ; 4-byte Folded Reload
+; VI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:324 ; 4-byte Folded Reload
+; VI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:308 ; 4-byte Folded Reload
+; VI-NEXT: buffer_load_dword v3, off, s[0:3], s32 offset:316 ; 4-byte Folded Reload
; VI-NEXT: s_waitcnt vmcnt(2)
; VI-NEXT: v_lshlrev_b16_e32 v1, 8, v1
; VI-NEXT: s_waitcnt vmcnt(1)
@@ -166041,10 +166082,10 @@ define <128 x i8> @bitcast_v64bf16_to_v128i8(<64 x bfloat> %a, i32 %b) {
; VI-NEXT: v_or_b32_sdwa v1, v1, v2 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_0 src1_sel:DWORD
; VI-NEXT: v_add_u32_e32 v2, vcc, 0x64, v0
; VI-NEXT: buffer_store_dword v1, v2, s[0:3], 0 offen
-; VI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:268 ; 4-byte Folded Reload
+; VI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:300 ; 4-byte Folded Reload
; VI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:92 ; 4-byte Folded Reload
; VI-NEXT: buffer_load_dword v3, off, s[0:3], s32 offset:96 ; 4-byte Folded Reload
-; VI-NEXT: buffer_load_dword v3, off, s[0:3], s32 offset:260 ; 4-byte Folded Reload
+; VI-NEXT: buffer_load_dword v3, off, s[0:3], s32 offset:292 ; 4-byte Folded Reload
; VI-NEXT: s_waitcnt vmcnt(3)
; VI-NEXT: v_lshlrev_b16_e32 v1, 8, v1
; VI-NEXT: s_waitcnt vmcnt(2)
@@ -166055,9 +166096,9 @@ define <128 x i8> @bitcast_v64bf16_to_v128i8(<64 x bfloat> %a, i32 %b) {
; VI-NEXT: v_or_b32_sdwa v1, v1, v2 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_0 src1_sel:DWORD
; VI-NEXT: v_add_u32_e32 v2, vcc, 0x68, v0
; VI-NEXT: buffer_store_dword v1, v2, s[0:3], 0 offen
-; VI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:252 ; 4-byte Folded Reload
-; VI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:240 ; 4-byte Folded Reload
-; VI-NEXT: buffer_load_dword v3, off, s[0:3], s32 offset:248 ; 4-byte Folded Reload
+; VI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:284 ; 4-byte Folded Reload
+; VI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:272 ; 4-byte Folded Reload
+; VI-NEXT: buffer_load_dword v3, off, s[0:3], s32 offset:280 ; 4-byte Folded Reload
; VI-NEXT: s_waitcnt vmcnt(2)
; VI-NEXT: v_lshlrev_b16_e32 v1, 8, v1
; VI-NEXT: s_waitcnt vmcnt(1)
@@ -166068,26 +166109,34 @@ define <128 x i8> @bitcast_v64bf16_to_v128i8(<64 x bfloat> %a, i32 %b) {
; VI-NEXT: v_or_b32_sdwa v1, v1, v2 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_0 src1_sel:DWORD
; VI-NEXT: v_add_u32_e32 v2, vcc, 0x6c, v0
; VI-NEXT: buffer_store_dword v1, v2, s[0:3], 0 offen
+; VI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:264 ; 4-byte Folded Reload
; VI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:84 ; 4-byte Folded Reload
; VI-NEXT: buffer_load_dword v3, off, s[0:3], s32 offset:88 ; 4-byte Folded Reload
-; VI-NEXT: v_lshlrev_b16_e32 v1, 8, v34
-; VI-NEXT: v_or_b32_sdwa v1, v29, v1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
-; VI-NEXT: s_waitcnt vmcnt(1)
+; VI-NEXT: buffer_load_dword v3, off, s[0:3], s32 offset:256 ; 4-byte Folded Reload
+; VI-NEXT: s_waitcnt vmcnt(3)
+; VI-NEXT: v_lshlrev_b16_e32 v1, 8, v1
+; VI-NEXT: s_waitcnt vmcnt(2)
; VI-NEXT: v_lshlrev_b16_e32 v2, 8, v2
-; VI-NEXT: v_or_b32_sdwa v2, v33, v2 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; VI-NEXT: v_or_b32_sdwa v1, v29, v1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; VI-NEXT: s_waitcnt vmcnt(0)
+; VI-NEXT: v_or_b32_sdwa v2, v3, v2 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
; VI-NEXT: v_or_b32_sdwa v1, v1, v2 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_0 src1_sel:DWORD
; VI-NEXT: v_add_u32_e32 v2, vcc, 0x70, v0
; VI-NEXT: buffer_store_dword v1, v2, s[0:3], 0 offen
-; VI-NEXT: v_lshlrev_b16_e32 v1, 8, v53
-; VI-NEXT: v_lshlrev_b16_e32 v2, 8, v61
+; VI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:244 ; 4-byte Folded Reload
+; VI-NEXT: buffer_load_dword v3, off, s[0:3], s32 offset:240 ; 4-byte Folded Reload
+; VI-NEXT: v_lshlrev_b16_e32 v2, 8, v34
+; VI-NEXT: s_waitcnt vmcnt(1)
+; VI-NEXT: v_lshlrev_b16_e32 v1, 8, v1
; VI-NEXT: v_or_b32_sdwa v1, v30, v1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
-; VI-NEXT: v_or_b32_sdwa v2, v62, v2 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; VI-NEXT: s_waitcnt vmcnt(0)
+; VI-NEXT: v_or_b32_sdwa v2, v3, v2 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
; VI-NEXT: v_or_b32_sdwa v1, v1, v2 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_0 src1_sel:DWORD
; VI-NEXT: v_add_u32_e32 v2, vcc, 0x74, v0
; VI-NEXT: buffer_store_dword v1, v2, s[0:3], 0 offen
; VI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:76 ; 4-byte Folded Reload
; VI-NEXT: buffer_load_dword v3, off, s[0:3], s32 offset:80 ; 4-byte Folded Reload
-; VI-NEXT: v_lshlrev_b16_e32 v1, 8, v63
+; VI-NEXT: v_lshlrev_b16_e32 v1, 8, v45
; VI-NEXT: v_or_b32_sdwa v1, v31, v1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
; VI-NEXT: s_waitcnt vmcnt(1)
; VI-NEXT: v_lshlrev_b16_e32 v2, 8, v2
@@ -166140,714 +166189,716 @@ define <128 x i8> @bitcast_v64bf16_to_v128i8(<64 x bfloat> %a, i32 %b) {
; GFX9-NEXT: buffer_store_dword v61, off, s[0:3], s32 offset:20 ; 4-byte Folded Spill
; GFX9-NEXT: buffer_store_dword v62, off, s[0:3], s32 offset:16 ; 4-byte Folded Spill
; GFX9-NEXT: buffer_store_dword v63, off, s[0:3], s32 offset:12 ; 4-byte Folded Spill
-; GFX9-NEXT: v_mov_b32_e32 v46, v15
; GFX9-NEXT: buffer_load_dword v32, off, s[0:3], s32 offset:4
-; GFX9-NEXT: buffer_load_dword v15, off, s[0:3], s32 offset:8
+; GFX9-NEXT: buffer_load_dword v33, off, s[0:3], s32 offset:8
; GFX9-NEXT: buffer_load_dword v31, off, s[0:3], s32
-; GFX9-NEXT: ; implicit-def: $vgpr33
-; GFX9-NEXT: ; kill: killed $vgpr33
-; GFX9-NEXT: ; implicit-def: $vgpr33
-; GFX9-NEXT: ; kill: killed $vgpr33
-; GFX9-NEXT: ; implicit-def: $vgpr33
-; GFX9-NEXT: ; kill: killed $vgpr33
-; GFX9-NEXT: ; implicit-def: $vgpr33
-; GFX9-NEXT: ; kill: killed $vgpr33
-; GFX9-NEXT: ; implicit-def: $vgpr33
-; GFX9-NEXT: ; implicit-def: $vgpr58
-; GFX9-NEXT: ; kill: killed $vgpr33
-; GFX9-NEXT: ; implicit-def: $vgpr33
-; GFX9-NEXT: ; kill: killed $vgpr58
-; GFX9-NEXT: ; implicit-def: $vgpr58
-; GFX9-NEXT: ; kill: killed $vgpr33
-; GFX9-NEXT: ; implicit-def: $vgpr33
-; GFX9-NEXT: ; kill: killed $vgpr58
-; GFX9-NEXT: ; implicit-def: $vgpr58
-; GFX9-NEXT: ; kill: killed $vgpr33
-; GFX9-NEXT: ; implicit-def: $vgpr33
-; GFX9-NEXT: ; kill: killed $vgpr58
-; GFX9-NEXT: ; implicit-def: $vgpr58
-; GFX9-NEXT: ; kill: killed $vgpr33
-; GFX9-NEXT: ; implicit-def: $vgpr33
-; GFX9-NEXT: ; kill: killed $vgpr58
-; GFX9-NEXT: ; implicit-def: $vgpr58
-; GFX9-NEXT: ; kill: killed $vgpr33
-; GFX9-NEXT: ; implicit-def: $vgpr33
-; GFX9-NEXT: ; kill: killed $vgpr58
-; GFX9-NEXT: ; implicit-def: $vgpr58
-; GFX9-NEXT: ; kill: killed $vgpr33
-; GFX9-NEXT: ; implicit-def: $vgpr33
-; GFX9-NEXT: ; kill: killed $vgpr58
-; GFX9-NEXT: ; implicit-def: $vgpr58
-; GFX9-NEXT: ; kill: killed $vgpr33
-; GFX9-NEXT: ; implicit-def: $vgpr33
-; GFX9-NEXT: ; kill: killed $vgpr58
-; GFX9-NEXT: ; implicit-def: $vgpr58
-; GFX9-NEXT: ; kill: killed $vgpr33
-; GFX9-NEXT: ; implicit-def: $vgpr33
-; GFX9-NEXT: ; kill: killed $vgpr58
-; GFX9-NEXT: ; implicit-def: $vgpr58
-; GFX9-NEXT: ; kill: killed $vgpr33
-; GFX9-NEXT: ; implicit-def: $vgpr33
-; GFX9-NEXT: ; kill: killed $vgpr58
-; GFX9-NEXT: ; implicit-def: $vgpr58
-; GFX9-NEXT: ; kill: killed $vgpr33
-; GFX9-NEXT: ; implicit-def: $vgpr33
-; GFX9-NEXT: ; kill: killed $vgpr58
-; GFX9-NEXT: ; implicit-def: $vgpr58
-; GFX9-NEXT: ; kill: killed $vgpr33
-; GFX9-NEXT: ; implicit-def: $vgpr33
-; GFX9-NEXT: ; kill: killed $vgpr58
-; GFX9-NEXT: ; implicit-def: $vgpr58
-; GFX9-NEXT: ; kill: killed $vgpr33
-; GFX9-NEXT: ; implicit-def: $vgpr33
-; GFX9-NEXT: ; kill: killed $vgpr58
-; GFX9-NEXT: ; implicit-def: $vgpr58
-; GFX9-NEXT: ; kill: killed $vgpr33
-; GFX9-NEXT: ; implicit-def: $vgpr33
-; GFX9-NEXT: ; implicit-def: $vgpr59
-; GFX9-NEXT: ; kill: killed $vgpr58
-; GFX9-NEXT: ; implicit-def: $vgpr58
-; GFX9-NEXT: ; kill: killed $vgpr33
-; GFX9-NEXT: ; implicit-def: $vgpr33
-; GFX9-NEXT: ; kill: killed $vgpr59
-; GFX9-NEXT: ; implicit-def: $vgpr59
-; GFX9-NEXT: ; kill: killed $vgpr58
-; GFX9-NEXT: ; implicit-def: $vgpr58
-; GFX9-NEXT: ; kill: killed $vgpr33
-; GFX9-NEXT: ; implicit-def: $vgpr33
-; GFX9-NEXT: ; kill: killed $vgpr59
-; GFX9-NEXT: ; implicit-def: $vgpr59
-; GFX9-NEXT: ; kill: killed $vgpr58
-; GFX9-NEXT: ; implicit-def: $vgpr58
-; GFX9-NEXT: ; kill: killed $vgpr33
-; GFX9-NEXT: ; implicit-def: $vgpr33
-; GFX9-NEXT: ; kill: killed $vgpr59
-; GFX9-NEXT: ; implicit-def: $vgpr59
-; GFX9-NEXT: ; kill: killed $vgpr58
-; GFX9-NEXT: ; implicit-def: $vgpr58
-; GFX9-NEXT: v_mov_b32_e32 v47, v16
-; GFX9-NEXT: ; kill: killed $vgpr33
-; GFX9-NEXT: ; implicit-def: $vgpr33
-; GFX9-NEXT: ; implicit-def: $vgpr36
-; GFX9-NEXT: ; kill: killed $vgpr59
+; GFX9-NEXT: v_mov_b32_e32 v36, v15
+; GFX9-NEXT: ; implicit-def: $vgpr15
+; GFX9-NEXT: ; kill: killed $vgpr15
+; GFX9-NEXT: ; implicit-def: $vgpr15
+; GFX9-NEXT: ; kill: killed $vgpr15
+; GFX9-NEXT: ; implicit-def: $vgpr15
+; GFX9-NEXT: ; kill: killed $vgpr15
+; GFX9-NEXT: ; implicit-def: $vgpr15
+; GFX9-NEXT: ; kill: killed $vgpr15
+; GFX9-NEXT: ; implicit-def: $vgpr15
+; GFX9-NEXT: ; kill: killed $vgpr15
+; GFX9-NEXT: ; implicit-def: $vgpr15
+; GFX9-NEXT: ; kill: killed $vgpr15
+; GFX9-NEXT: ; implicit-def: $vgpr15
+; GFX9-NEXT: ; kill: killed $vgpr15
+; GFX9-NEXT: ; implicit-def: $vgpr15
+; GFX9-NEXT: ; kill: killed $vgpr15
+; GFX9-NEXT: ; implicit-def: $vgpr15
+; GFX9-NEXT: ; kill: killed $vgpr15
+; GFX9-NEXT: ; implicit-def: $vgpr15
+; GFX9-NEXT: ; kill: killed $vgpr15
+; GFX9-NEXT: ; implicit-def: $vgpr15
+; GFX9-NEXT: ; kill: killed $vgpr15
+; GFX9-NEXT: ; implicit-def: $vgpr15
+; GFX9-NEXT: ; kill: killed $vgpr15
+; GFX9-NEXT: ; implicit-def: $vgpr15
+; GFX9-NEXT: ; kill: killed $vgpr15
+; GFX9-NEXT: ; implicit-def: $vgpr15
+; GFX9-NEXT: ; kill: killed $vgpr15
+; GFX9-NEXT: ; implicit-def: $vgpr15
+; GFX9-NEXT: ; kill: killed $vgpr15
+; GFX9-NEXT: ; implicit-def: $vgpr15
+; GFX9-NEXT: ; kill: killed $vgpr15
+; GFX9-NEXT: ; implicit-def: $vgpr15
+; GFX9-NEXT: ; kill: killed $vgpr15
+; GFX9-NEXT: ; implicit-def: $vgpr15
+; GFX9-NEXT: ; kill: killed $vgpr15
+; GFX9-NEXT: ; implicit-def: $vgpr15
+; GFX9-NEXT: ; kill: killed $vgpr15
+; GFX9-NEXT: ; implicit-def: $vgpr15
+; GFX9-NEXT: ; kill: killed $vgpr15
+; GFX9-NEXT: ; implicit-def: $vgpr15
+; GFX9-NEXT: ; kill: killed $vgpr15
+; GFX9-NEXT: ; implicit-def: $vgpr15
+; GFX9-NEXT: ; kill: killed $vgpr15
+; GFX9-NEXT: ; implicit-def: $vgpr15
+; GFX9-NEXT: ; kill: killed $vgpr15
+; GFX9-NEXT: ; implicit-def: $vgpr15
+; GFX9-NEXT: ; kill: killed $vgpr15
+; GFX9-NEXT: ; implicit-def: $vgpr15
+; GFX9-NEXT: ; kill: killed $vgpr15
+; GFX9-NEXT: ; implicit-def: $vgpr15
+; GFX9-NEXT: ; kill: killed $vgpr15
+; GFX9-NEXT: ; implicit-def: $vgpr15
+; GFX9-NEXT: ; kill: killed $vgpr15
+; GFX9-NEXT: ; implicit-def: $vgpr15
+; GFX9-NEXT: ; kill: killed $vgpr15
+; GFX9-NEXT: ; implicit-def: $vgpr15
+; GFX9-NEXT: ; kill: killed $vgpr15
+; GFX9-NEXT: ; implicit-def: $vgpr15
+; GFX9-NEXT: ; kill: killed $vgpr15
+; GFX9-NEXT: ; implicit-def: $vgpr15
+; GFX9-NEXT: ; kill: killed $vgpr15
+; GFX9-NEXT: ; implicit-def: $vgpr15
+; GFX9-NEXT: ; kill: killed $vgpr15
+; GFX9-NEXT: ; implicit-def: $vgpr15
+; GFX9-NEXT: ; kill: killed $vgpr15
+; GFX9-NEXT: ; implicit-def: $vgpr15
+; GFX9-NEXT: ; kill: killed $vgpr15
+; GFX9-NEXT: ; implicit-def: $vgpr15
+; GFX9-NEXT: ; kill: killed $vgpr15
+; GFX9-NEXT: ; implicit-def: $vgpr15
+; GFX9-NEXT: ; kill: killed $vgpr15
+; GFX9-NEXT: ; implicit-def: $vgpr15
+; GFX9-NEXT: ; kill: killed $vgpr15
+; GFX9-NEXT: ; implicit-def: $vgpr15
+; GFX9-NEXT: ; kill: killed $vgpr15
+; GFX9-NEXT: ; implicit-def: $vgpr15
+; GFX9-NEXT: ; kill: killed $vgpr15
+; GFX9-NEXT: ; implicit-def: $vgpr15
+; GFX9-NEXT: ; kill: killed $vgpr15
+; GFX9-NEXT: ; implicit-def: $vgpr15
+; GFX9-NEXT: ; kill: killed $vgpr15
+; GFX9-NEXT: ; implicit-def: $vgpr15
+; GFX9-NEXT: ; kill: killed $vgpr15
+; GFX9-NEXT: ; implicit-def: $vgpr15
+; GFX9-NEXT: ; kill: killed $vgpr15
+; GFX9-NEXT: ; implicit-def: $vgpr15
+; GFX9-NEXT: ; kill: killed $vgpr15
+; GFX9-NEXT: ; implicit-def: $vgpr15
+; GFX9-NEXT: ; kill: killed $vgpr15
+; GFX9-NEXT: ; implicit-def: $vgpr15
+; GFX9-NEXT: ; kill: killed $vgpr15
+; GFX9-NEXT: ; implicit-def: $vgpr15
+; GFX9-NEXT: ; kill: killed $vgpr15
+; GFX9-NEXT: ; implicit-def: $vgpr15
+; GFX9-NEXT: ; kill: killed $vgpr15
+; GFX9-NEXT: ; implicit-def: $vgpr15
+; GFX9-NEXT: ; kill: killed $vgpr15
+; GFX9-NEXT: ; implicit-def: $vgpr15
+; GFX9-NEXT: ; kill: killed $vgpr15
+; GFX9-NEXT: ; implicit-def: $vgpr15
+; GFX9-NEXT: ; kill: killed $vgpr15
+; GFX9-NEXT: ; implicit-def: $vgpr15
+; GFX9-NEXT: ; kill: killed $vgpr15
+; GFX9-NEXT: ; implicit-def: $vgpr15
+; GFX9-NEXT: v_mov_b32_e32 v37, v16
+; GFX9-NEXT: ; kill: killed $vgpr15
+; GFX9-NEXT: ; implicit-def: $vgpr15
+; GFX9-NEXT: ; implicit-def: $vgpr57
; GFX9-NEXT: ; implicit-def: $vgpr59
-; GFX9-NEXT: ; kill: killed $vgpr58
-; GFX9-NEXT: ; implicit-def: $vgpr58
+; GFX9-NEXT: ; implicit-def: $vgpr54
+; GFX9-NEXT: ; implicit-def: $vgpr46
+; GFX9-NEXT: ; implicit-def: $vgpr42
; GFX9-NEXT: ; implicit-def: $vgpr41
-; GFX9-NEXT: ; implicit-def: $vgpr38
-; GFX9-NEXT: ; implicit-def: $vgpr51
-; GFX9-NEXT: ; implicit-def: $vgpr53
-; GFX9-NEXT: ; implicit-def: $vgpr48
-; GFX9-NEXT: ; implicit-def: $vgpr49
; GFX9-NEXT: ; implicit-def: $vgpr52
-; GFX9-NEXT: ; implicit-def: $vgpr16
-; GFX9-NEXT: ; implicit-def: $vgpr42
-; GFX9-NEXT: ; implicit-def: $vgpr39
-; GFX9-NEXT: ; implicit-def: $vgpr45
; GFX9-NEXT: ; implicit-def: $vgpr40
-; GFX9-NEXT: ; implicit-def: $vgpr54
-; GFX9-NEXT: ; implicit-def: $vgpr62
-; GFX9-NEXT: ; implicit-def: $vgpr61
-; GFX9-NEXT: ; implicit-def: $vgpr60
-; GFX9-NEXT: ; implicit-def: $vgpr35
-; GFX9-NEXT: ; implicit-def: $vgpr44
-; GFX9-NEXT: ; implicit-def: $vgpr34
-; GFX9-NEXT: ; implicit-def: $vgpr43
-; GFX9-NEXT: ; kill: killed $vgpr33
-; GFX9-NEXT: ; implicit-def: $vgpr33
+; GFX9-NEXT: ; implicit-def: $vgpr49
; GFX9-NEXT: ; implicit-def: $vgpr55
+; GFX9-NEXT: ; implicit-def: $vgpr53
+; GFX9-NEXT: ; implicit-def: $vgpr38
+; GFX9-NEXT: ; implicit-def: $vgpr35
+; GFX9-NEXT: ; implicit-def: $vgpr60
+; GFX9-NEXT: ; implicit-def: $vgpr45
; GFX9-NEXT: ; implicit-def: $vgpr56
+; GFX9-NEXT: ; implicit-def: $vgpr43
; GFX9-NEXT: ; implicit-def: $vgpr50
-; GFX9-NEXT: ; kill: killed $vgpr36
-; GFX9-NEXT: ; implicit-def: $vgpr37
-; GFX9-NEXT: ; implicit-def: $vgpr57
-; GFX9-NEXT: ; implicit-def: $vgpr36
+; GFX9-NEXT: ; implicit-def: $vgpr51
+; GFX9-NEXT: ; implicit-def: $vgpr62
; GFX9-NEXT: ; implicit-def: $vgpr63
-; GFX9-NEXT: ; kill: killed $vgpr59
-; GFX9-NEXT: buffer_store_dword v58, off, s[0:3], s32 offset:196 ; 4-byte Folded Spill
-; GFX9-NEXT: s_nop 0
-; GFX9-NEXT: buffer_store_dword v59, off, s[0:3], s32 offset:200 ; 4-byte Folded Spill
; GFX9-NEXT: ; implicit-def: $vgpr58
-; GFX9-NEXT: buffer_store_dword v58, off, s[0:3], s32 offset:188 ; 4-byte Folded Spill
+; GFX9-NEXT: ; implicit-def: $vgpr47
+; GFX9-NEXT: ; implicit-def: $vgpr48
+; GFX9-NEXT: ; implicit-def: $vgpr39
+; GFX9-NEXT: ; implicit-def: $vgpr44
+; GFX9-NEXT: ; implicit-def: $vgpr61
+; GFX9-NEXT: buffer_store_dword v15, off, s[0:3], s32 offset:188 ; 4-byte Folded Spill
; GFX9-NEXT: s_nop 0
-; GFX9-NEXT: buffer_store_dword v59, off, s[0:3], s32 offset:192 ; 4-byte Folded Spill
-; GFX9-NEXT: ; implicit-def: $vgpr58
-; GFX9-NEXT: buffer_store_dword v58, off, s[0:3], s32 offset:172 ; 4-byte Folded Spill
+; GFX9-NEXT: buffer_store_dword v16, off, s[0:3], s32 offset:192 ; 4-byte Folded Spill
+; GFX9-NEXT: ; implicit-def: $vgpr15
+; GFX9-NEXT: buffer_store_dword v15, off, s[0:3], s32 offset:180 ; 4-byte Folded Spill
; GFX9-NEXT: s_nop 0
-; GFX9-NEXT: buffer_store_dword v59, off, s[0:3], s32 offset:176 ; 4-byte Folded Spill
-; GFX9-NEXT: ; implicit-def: $vgpr58
-; GFX9-NEXT: buffer_store_dword v58, off, s[0:3], s32 offset:156 ; 4-byte Folded Spill
+; GFX9-NEXT: buffer_store_dword v16, off, s[0:3], s32 offset:184 ; 4-byte Folded Spill
+; GFX9-NEXT: ; implicit-def: $vgpr15
+; GFX9-NEXT: ; implicit-def: $vgpr16
+; GFX9-NEXT: buffer_store_dword v16, off, s[0:3], s32 offset:156 ; 4-byte Folded Spill
; GFX9-NEXT: s_nop 0
-; GFX9-NEXT: buffer_store_dword v59, off, s[0:3], s32 offset:160 ; 4-byte Folded Spill
-; GFX9-NEXT: ; implicit-def: $vgpr58
-; GFX9-NEXT: buffer_store_dword v58, off, s[0:3], s32 offset:148 ; 4-byte Folded Spill
+; GFX9-NEXT: buffer_store_dword v17, off, s[0:3], s32 offset:160 ; 4-byte Folded Spill
+; GFX9-NEXT: ; implicit-def: $vgpr16
+; GFX9-NEXT: buffer_store_dword v16, off, s[0:3], s32 offset:148 ; 4-byte Folded Spill
; GFX9-NEXT: s_nop 0
-; GFX9-NEXT: buffer_store_dword v59, off, s[0:3], s32 offset:152 ; 4-byte Folded Spill
-; GFX9-NEXT: ; implicit-def: $vgpr58
-; GFX9-NEXT: buffer_store_dword v58, off, s[0:3], s32 offset:140 ; 4-byte Folded Spill
+; GFX9-NEXT: buffer_store_dword v17, off, s[0:3], s32 offset:152 ; 4-byte Folded Spill
+; GFX9-NEXT: ; implicit-def: $vgpr16
+; GFX9-NEXT: buffer_store_dword v16, off, s[0:3], s32 offset:140 ; 4-byte Folded Spill
; GFX9-NEXT: s_nop 0
-; GFX9-NEXT: buffer_store_dword v59, off, s[0:3], s32 offset:144 ; 4-byte Folded Spill
-; GFX9-NEXT: ; implicit-def: $vgpr58
-; GFX9-NEXT: buffer_store_dword v58, off, s[0:3], s32 offset:132 ; 4-byte Folded Spill
+; GFX9-NEXT: buffer_store_dword v17, off, s[0:3], s32 offset:144 ; 4-byte Folded Spill
+; GFX9-NEXT: ; implicit-def: $vgpr16
+; GFX9-NEXT: buffer_store_dword v16, off, s[0:3], s32 offset:132 ; 4-byte Folded Spill
; GFX9-NEXT: s_nop 0
-; GFX9-NEXT: buffer_store_dword v59, off, s[0:3], s32 offset:136 ; 4-byte Folded Spill
-; GFX9-NEXT: ; implicit-def: $vgpr58
-; GFX9-NEXT: buffer_store_dword v58, off, s[0:3], s32 offset:124 ; 4-byte Folded Spill
+; GFX9-NEXT: buffer_store_dword v17, off, s[0:3], s32 offset:136 ; 4-byte Folded Spill
+; GFX9-NEXT: ; implicit-def: $vgpr16
+; GFX9-NEXT: buffer_store_dword v16, off, s[0:3], s32 offset:124 ; 4-byte Folded Spill
; GFX9-NEXT: s_nop 0
-; GFX9-NEXT: buffer_store_dword v59, off, s[0:3], s32 offset:128 ; 4-byte Folded Spill
-; GFX9-NEXT: ; implicit-def: $vgpr58
-; GFX9-NEXT: buffer_store_dword v58, off, s[0:3], s32 offset:180 ; 4-byte Folded Spill
+; GFX9-NEXT: buffer_store_dword v17, off, s[0:3], s32 offset:128 ; 4-byte Folded Spill
+; GFX9-NEXT: ; implicit-def: $vgpr16
+; GFX9-NEXT: buffer_store_dword v16, off, s[0:3], s32 offset:172 ; 4-byte Folded Spill
; GFX9-NEXT: s_nop 0
-; GFX9-NEXT: buffer_store_dword v59, off, s[0:3], s32 offset:184 ; 4-byte Folded Spill
-; GFX9-NEXT: ; implicit-def: $vgpr58
-; GFX9-NEXT: buffer_store_dword v58, off, s[0:3], s32 offset:164 ; 4-byte Folded Spill
+; GFX9-NEXT: buffer_store_dword v17, off, s[0:3], s32 offset:176 ; 4-byte Folded Spill
+; GFX9-NEXT: ; implicit-def: $vgpr16
+; GFX9-NEXT: buffer_store_dword v16, off, s[0:3], s32 offset:164 ; 4-byte Folded Spill
; GFX9-NEXT: s_nop 0
-; GFX9-NEXT: buffer_store_dword v59, off, s[0:3], s32 offset:168 ; 4-byte Folded Spill
-; GFX9-NEXT: ; implicit-def: $vgpr58
-; GFX9-NEXT: buffer_store_dword v58, off, s[0:3], s32 offset:116 ; 4-byte Folded Spill
+; GFX9-NEXT: buffer_store_dword v17, off, s[0:3], s32 offset:168 ; 4-byte Folded Spill
+; GFX9-NEXT: ; implicit-def: $vgpr16
+; GFX9-NEXT: buffer_store_dword v16, off, s[0:3], s32 offset:108 ; 4-byte Folded Spill
; GFX9-NEXT: s_nop 0
-; GFX9-NEXT: buffer_store_dword v59, off, s[0:3], s32 offset:120 ; 4-byte Folded Spill
-; GFX9-NEXT: ; implicit-def: $vgpr58
-; GFX9-NEXT: buffer_store_dword v58, off, s[0:3], s32 offset:108 ; 4-byte Folded Spill
+; GFX9-NEXT: buffer_store_dword v17, off, s[0:3], s32 offset:112 ; 4-byte Folded Spill
+; GFX9-NEXT: ; implicit-def: $vgpr16
+; GFX9-NEXT: buffer_store_dword v16, off, s[0:3], s32 offset:100 ; 4-byte Folded Spill
; GFX9-NEXT: s_nop 0
-; GFX9-NEXT: buffer_store_dword v59, off, s[0:3], s32 offset:112 ; 4-byte Folded Spill
-; GFX9-NEXT: ; implicit-def: $vgpr58
-; GFX9-NEXT: buffer_store_dword v58, off, s[0:3], s32 offset:100 ; 4-byte Folded Spill
+; GFX9-NEXT: buffer_store_dword v17, off, s[0:3], s32 offset:104 ; 4-byte Folded Spill
+; GFX9-NEXT: ; implicit-def: $vgpr16
+; GFX9-NEXT: buffer_store_dword v16, off, s[0:3], s32 offset:92 ; 4-byte Folded Spill
; GFX9-NEXT: s_nop 0
-; GFX9-NEXT: buffer_store_dword v59, off, s[0:3], s32 offset:104 ; 4-byte Folded Spill
-; GFX9-NEXT: ; implicit-def: $vgpr58
-; GFX9-NEXT: buffer_store_dword v58, off, s[0:3], s32 offset:92 ; 4-byte Folded Spill
+; GFX9-NEXT: buffer_store_dword v17, off, s[0:3], s32 offset:96 ; 4-byte Folded Spill
+; GFX9-NEXT: ; implicit-def: $vgpr16
+; GFX9-NEXT: buffer_store_dword v16, off, s[0:3], s32 offset:116 ; 4-byte Folded Spill
; GFX9-NEXT: s_nop 0
-; GFX9-NEXT: buffer_store_dword v59, off, s[0:3], s32 offset:96 ; 4-byte Folded Spill
-; GFX9-NEXT: ; implicit-def: $vgpr58
-; GFX9-NEXT: buffer_store_dword v58, off, s[0:3], s32 offset:84 ; 4-byte Folded Spill
+; GFX9-NEXT: buffer_store_dword v17, off, s[0:3], s32 offset:120 ; 4-byte Folded Spill
+; GFX9-NEXT: ; implicit-def: $vgpr16
+; GFX9-NEXT: buffer_store_dword v16, off, s[0:3], s32 offset:84 ; 4-byte Folded Spill
; GFX9-NEXT: s_nop 0
-; GFX9-NEXT: buffer_store_dword v59, off, s[0:3], s32 offset:88 ; 4-byte Folded Spill
-; GFX9-NEXT: ; implicit-def: $vgpr58
-; GFX9-NEXT: buffer_store_dword v58, off, s[0:3], s32 offset:76 ; 4-byte Folded Spill
+; GFX9-NEXT: buffer_store_dword v17, off, s[0:3], s32 offset:88 ; 4-byte Folded Spill
+; GFX9-NEXT: ; implicit-def: $vgpr16
+; GFX9-NEXT: buffer_store_dword v16, off, s[0:3], s32 offset:76 ; 4-byte Folded Spill
; GFX9-NEXT: s_nop 0
-; GFX9-NEXT: buffer_store_dword v59, off, s[0:3], s32 offset:80 ; 4-byte Folded Spill
-; GFX9-NEXT: s_waitcnt vmcnt(33)
-; GFX9-NEXT: v_cmp_ne_u32_e32 vcc, 0, v15
-; GFX9-NEXT: ; implicit-def: $vgpr15
-; GFX9-NEXT: ; kill: killed $vgpr15
-; GFX9-NEXT: ; implicit-def: $vgpr15
-; GFX9-NEXT: ; kill: killed $vgpr15
-; GFX9-NEXT: ; implicit-def: $vgpr15
-; GFX9-NEXT: ; kill: killed $vgpr15
-; GFX9-NEXT: ; implicit-def: $vgpr15
-; GFX9-NEXT: ; kill: killed $vgpr15
-; GFX9-NEXT: ; implicit-def: $vgpr15
-; GFX9-NEXT: ; kill: killed $vgpr15
-; GFX9-NEXT: ; implicit-def: $vgpr15
-; GFX9-NEXT: ; kill: killed $vgpr15
-; GFX9-NEXT: ; implicit-def: $vgpr15
+; GFX9-NEXT: buffer_store_dword v17, off, s[0:3], s32 offset:80 ; 4-byte Folded Spill
+; GFX9-NEXT: s_waitcnt vmcnt(31)
+; GFX9-NEXT: v_cmp_ne_u32_e32 vcc, 0, v33
; GFX9-NEXT: s_and_saveexec_b64 s[4:5], vcc
; GFX9-NEXT: s_xor_b64 s[4:5], exec, s[4:5]
; GFX9-NEXT: s_cbranch_execz .LBB90_2
; GFX9-NEXT: ; %bb.1: ; %cmp.false
-; GFX9-NEXT: v_lshrrev_b32_e32 v33, 24, v4
-; GFX9-NEXT: buffer_store_dword v33, off, s[0:3], s32 offset:400 ; 4-byte Folded Spill
-; GFX9-NEXT: v_lshrrev_b32_e32 v33, 16, v4
-; GFX9-NEXT: buffer_store_dword v33, off, s[0:3], s32 offset:248 ; 4-byte Folded Spill
-; GFX9-NEXT: v_lshrrev_b32_e32 v33, 16, v2
-; GFX9-NEXT: buffer_store_dword v33, off, s[0:3], s32 offset:256 ; 4-byte Folded Spill
-; GFX9-NEXT: v_lshrrev_b32_e32 v33, 24, v32
-; GFX9-NEXT: buffer_store_dword v33, off, s[0:3], s32 offset:264 ; 4-byte Folded Spill
-; GFX9-NEXT: v_lshrrev_b32_e32 v33, 16, v32
-; GFX9-NEXT: buffer_store_dword v33, off, s[0:3], s32 offset:208 ; 4-byte Folded Spill
-; GFX9-NEXT: v_lshrrev_b32_e32 v33, 8, v32
-; GFX9-NEXT: v_lshrrev_b32_e32 v15, 24, v47
-; GFX9-NEXT: buffer_store_dword v33, off, s[0:3], s32 offset:268 ; 4-byte Folded Spill
-; GFX9-NEXT: s_waitcnt vmcnt(38)
-; GFX9-NEXT: v_lshrrev_b32_e32 v33, 16, v31
-; GFX9-NEXT: buffer_store_dword v15, off, s[0:3], s32 offset:280 ; 4-byte Folded Spill
-; GFX9-NEXT: v_lshrrev_b32_e32 v15, 16, v47
-; GFX9-NEXT: buffer_store_dword v33, off, s[0:3], s32 offset:272 ; 4-byte Folded Spill
-; GFX9-NEXT: v_lshrrev_b32_e32 v33, 8, v31
-; GFX9-NEXT: buffer_store_dword v15, off, s[0:3], s32 offset:204 ; 4-byte Folded Spill
-; GFX9-NEXT: v_lshrrev_b32_e32 v15, 8, v47
-; GFX9-NEXT: buffer_store_dword v33, off, s[0:3], s32 offset:276 ; 4-byte Folded Spill
-; GFX9-NEXT: v_lshrrev_b32_e32 v33, 24, v30
-; GFX9-NEXT: buffer_store_dword v15, off, s[0:3], s32 offset:364 ; 4-byte Folded Spill
-; GFX9-NEXT: v_lshrrev_b32_e32 v15, 16, v46
-; GFX9-NEXT: buffer_store_dword v33, off, s[0:3], s32 offset:284 ; 4-byte Folded Spill
-; GFX9-NEXT: v_lshrrev_b32_e32 v33, 16, v30
-; GFX9-NEXT: buffer_store_dword v15, off, s[0:3], s32 offset:288 ; 4-byte Folded Spill
-; GFX9-NEXT: v_lshrrev_b32_e32 v15, 8, v46
-; GFX9-NEXT: buffer_store_dword v33, off, s[0:3], s32 offset:216 ; 4-byte Folded Spill
-; GFX9-NEXT: v_lshrrev_b32_e32 v33, 8, v30
-; GFX9-NEXT: buffer_store_dword v15, off, s[0:3], s32 offset:376 ; 4-byte Folded Spill
+; GFX9-NEXT: v_lshrrev_b32_e32 v15, 24, v37
+; GFX9-NEXT: buffer_store_dword v15, off, s[0:3], s32 offset:268 ; 4-byte Folded Spill
+; GFX9-NEXT: v_lshrrev_b32_e32 v15, 16, v37
+; GFX9-NEXT: buffer_store_dword v15, off, s[0:3], s32 offset:196 ; 4-byte Folded Spill
+; GFX9-NEXT: v_lshrrev_b32_e32 v15, 8, v37
+; GFX9-NEXT: buffer_store_dword v15, off, s[0:3], s32 offset:356 ; 4-byte Folded Spill
+; GFX9-NEXT: v_lshrrev_b32_e32 v15, 16, v36
+; GFX9-NEXT: buffer_store_dword v15, off, s[0:3], s32 offset:276 ; 4-byte Folded Spill
; GFX9-NEXT: v_lshrrev_b32_e32 v15, 24, v14
-; GFX9-NEXT: buffer_store_dword v33, off, s[0:3], s32 offset:296 ; 4-byte Folded Spill
-; GFX9-NEXT: v_lshrrev_b32_e32 v33, 16, v29
-; GFX9-NEXT: buffer_store_dword v15, off, s[0:3], s32 offset:304 ; 4-byte Folded Spill
+; GFX9-NEXT: buffer_store_dword v15, off, s[0:3], s32 offset:288 ; 4-byte Folded Spill
; GFX9-NEXT: v_lshrrev_b32_e32 v15, 16, v14
-; GFX9-NEXT: buffer_store_dword v33, off, s[0:3], s32 offset:300 ; 4-byte Folded Spill
-; GFX9-NEXT: v_lshrrev_b32_e32 v33, 8, v29
-; GFX9-NEXT: buffer_store_dword v15, off, s[0:3], s32 offset:212 ; 4-byte Folded Spill
+; GFX9-NEXT: buffer_store_dword v15, off, s[0:3], s32 offset:204 ; 4-byte Folded Spill
; GFX9-NEXT: v_lshrrev_b32_e32 v15, 8, v14
-; GFX9-NEXT: buffer_store_dword v33, off, s[0:3], s32 offset:308 ; 4-byte Folded Spill
-; GFX9-NEXT: v_lshrrev_b32_e32 v33, 24, v28
-; GFX9-NEXT: buffer_store_dword v15, off, s[0:3], s32 offset:384 ; 4-byte Folded Spill
+; GFX9-NEXT: buffer_store_dword v15, off, s[0:3], s32 offset:376 ; 4-byte Folded Spill
; GFX9-NEXT: v_lshrrev_b32_e32 v15, 16, v13
-; GFX9-NEXT: buffer_store_dword v33, off, s[0:3], s32 offset:316 ; 4-byte Folded Spill
-; GFX9-NEXT: v_lshrrev_b32_e32 v33, 16, v28
-; GFX9-NEXT: buffer_store_dword v15, off, s[0:3], s32 offset:312 ; 4-byte Folded Spill
+; GFX9-NEXT: buffer_store_dword v15, off, s[0:3], s32 offset:300 ; 4-byte Folded Spill
; GFX9-NEXT: v_lshrrev_b32_e32 v15, 8, v13
-; GFX9-NEXT: buffer_store_dword v33, off, s[0:3], s32 offset:224 ; 4-byte Folded Spill
-; GFX9-NEXT: v_lshrrev_b32_e32 v33, 8, v28
-; GFX9-NEXT: buffer_store_dword v15, off, s[0:3], s32 offset:396 ; 4-byte Folded Spill
-; GFX9-NEXT: v_lshrrev_b32_e32 v15, 24, v12
-; GFX9-NEXT: buffer_store_dword v33, off, s[0:3], s32 offset:320 ; 4-byte Folded Spill
-; GFX9-NEXT: v_lshrrev_b32_e32 v33, 16, v27
-; GFX9-NEXT: buffer_store_dword v15, off, s[0:3], s32 offset:324 ; 4-byte Folded Spill
+; GFX9-NEXT: buffer_store_dword v15, off, s[0:3], s32 offset:384 ; 4-byte Folded Spill
; GFX9-NEXT: v_lshrrev_b32_e32 v15, 16, v12
-; GFX9-NEXT: buffer_store_dword v33, off, s[0:3], s32 offset:328 ; 4-byte Folded Spill
-; GFX9-NEXT: v_lshrrev_b32_e32 v33, 8, v27
-; GFX9-NEXT: buffer_store_dword v15, off, s[0:3], s32 offset:220 ; 4-byte Folded Spill
+; GFX9-NEXT: buffer_store_dword v15, off, s[0:3], s32 offset:212 ; 4-byte Folded Spill
; GFX9-NEXT: v_lshrrev_b32_e32 v15, 8, v12
-; GFX9-NEXT: buffer_store_dword v33, off, s[0:3], s32 offset:332 ; 4-byte Folded Spill
-; GFX9-NEXT: v_lshrrev_b32_e32 v33, 24, v26
-; GFX9-NEXT: buffer_store_dword v15, off, s[0:3], s32 offset:404 ; 4-byte Folded Spill
+; GFX9-NEXT: buffer_store_dword v15, off, s[0:3], s32 offset:408 ; 4-byte Folded Spill
; GFX9-NEXT: v_lshrrev_b32_e32 v15, 16, v11
-; GFX9-NEXT: buffer_store_dword v33, off, s[0:3], s32 offset:340 ; 4-byte Folded Spill
-; GFX9-NEXT: v_lshrrev_b32_e32 v33, 16, v26
-; GFX9-NEXT: buffer_store_dword v15, off, s[0:3], s32 offset:336 ; 4-byte Folded Spill
+; GFX9-NEXT: buffer_store_dword v15, off, s[0:3], s32 offset:324 ; 4-byte Folded Spill
; GFX9-NEXT: v_lshrrev_b32_e32 v15, 24, v10
-; GFX9-NEXT: buffer_store_dword v33, off, s[0:3], s32 offset:236 ; 4-byte Folded Spill
-; GFX9-NEXT: v_lshrrev_b32_e32 v33, 8, v26
-; GFX9-NEXT: buffer_store_dword v15, off, s[0:3], s32 offset:344 ; 4-byte Folded Spill
+; GFX9-NEXT: buffer_store_dword v15, off, s[0:3], s32 offset:332 ; 4-byte Folded Spill
; GFX9-NEXT: v_lshrrev_b32_e32 v15, 16, v10
-; GFX9-NEXT: buffer_store_dword v33, off, s[0:3], s32 offset:348 ; 4-byte Folded Spill
-; GFX9-NEXT: v_lshrrev_b32_e32 v33, 16, v25
-; GFX9-NEXT: buffer_store_dword v15, off, s[0:3], s32 offset:228 ; 4-byte Folded Spill
+; GFX9-NEXT: buffer_store_dword v15, off, s[0:3], s32 offset:220 ; 4-byte Folded Spill
; GFX9-NEXT: v_lshrrev_b32_e32 v15, 16, v9
-; GFX9-NEXT: buffer_store_dword v33, off, s[0:3], s32 offset:352 ; 4-byte Folded Spill
-; GFX9-NEXT: v_lshrrev_b32_e32 v33, 8, v25
-; GFX9-NEXT: buffer_store_dword v15, off, s[0:3], s32 offset:356 ; 4-byte Folded Spill
+; GFX9-NEXT: buffer_store_dword v15, off, s[0:3], s32 offset:344 ; 4-byte Folded Spill
+; GFX9-NEXT: v_lshrrev_b32_e32 v15, 8, v9
+; GFX9-NEXT: buffer_store_dword v15, off, s[0:3], s32 offset:420 ; 4-byte Folded Spill
; GFX9-NEXT: v_lshrrev_b32_e32 v15, 24, v8
-; GFX9-NEXT: v_lshrrev_b32_e32 v16, 24, v6
-; GFX9-NEXT: buffer_store_dword v33, off, s[0:3], s32 offset:360 ; 4-byte Folded Spill
-; GFX9-NEXT: v_lshrrev_b32_e32 v33, 24, v24
-; GFX9-NEXT: v_lshrrev_b32_e32 v58, 16, v18
-; GFX9-NEXT: buffer_store_dword v15, off, s[0:3], s32 offset:372 ; 4-byte Folded Spill
+; GFX9-NEXT: buffer_store_dword v15, off, s[0:3], s32 offset:364 ; 4-byte Folded Spill
; GFX9-NEXT: v_lshrrev_b32_e32 v15, 16, v8
-; GFX9-NEXT: buffer_store_dword v16, off, s[0:3], s32 offset:388 ; 4-byte Folded Spill
-; GFX9-NEXT: v_lshrrev_b32_e32 v16, 16, v6
-; GFX9-NEXT: buffer_store_dword v33, off, s[0:3], s32 offset:368 ; 4-byte Folded Spill
-; GFX9-NEXT: v_lshrrev_b32_e32 v33, 16, v24
-; GFX9-NEXT: buffer_store_dword v58, off, s[0:3], s32 offset:292 ; 4-byte Folded Spill
-; GFX9-NEXT: v_lshrrev_b64 v[58:59], 24, v[46:47]
-; GFX9-NEXT: buffer_store_dword v15, off, s[0:3], s32 offset:232 ; 4-byte Folded Spill
-; GFX9-NEXT: v_lshrrev_b32_e32 v15, 16, v7
-; GFX9-NEXT: buffer_store_dword v16, off, s[0:3], s32 offset:240 ; 4-byte Folded Spill
-; GFX9-NEXT: v_lshrrev_b32_e32 v16, 16, v5
-; GFX9-NEXT: buffer_store_dword v33, off, s[0:3], s32 offset:244 ; 4-byte Folded Spill
-; GFX9-NEXT: v_lshrrev_b32_e32 v33, 16, v22
-; GFX9-NEXT: v_lshrrev_b32_e32 v34, 16, v20
-; GFX9-NEXT: buffer_store_dword v15, off, s[0:3], s32 offset:380 ; 4-byte Folded Spill
-; GFX9-NEXT: buffer_store_dword v16, off, s[0:3], s32 offset:392 ; 4-byte Folded Spill
-; GFX9-NEXT: buffer_store_dword v33, off, s[0:3], s32 offset:252 ; 4-byte Folded Spill
-; GFX9-NEXT: buffer_store_dword v34, off, s[0:3], s32 offset:260 ; 4-byte Folded Spill
-; GFX9-NEXT: buffer_store_dword v58, off, s[0:3], s32 offset:124 ; 4-byte Folded Spill
-; GFX9-NEXT: s_nop 0
-; GFX9-NEXT: buffer_store_dword v59, off, s[0:3], s32 offset:128 ; 4-byte Folded Spill
-; GFX9-NEXT: v_lshrrev_b64 v[58:59], 24, v[13:14]
-; GFX9-NEXT: buffer_store_dword v58, off, s[0:3], s32 offset:132 ; 4-byte Folded Spill
-; GFX9-NEXT: s_nop 0
-; GFX9-NEXT: buffer_store_dword v59, off, s[0:3], s32 offset:136 ; 4-byte Folded Spill
-; GFX9-NEXT: v_lshrrev_b64 v[58:59], 24, v[11:12]
-; GFX9-NEXT: buffer_store_dword v58, off, s[0:3], s32 offset:140 ; 4-byte Folded Spill
-; GFX9-NEXT: s_nop 0
-; GFX9-NEXT: buffer_store_dword v59, off, s[0:3], s32 offset:144 ; 4-byte Folded Spill
-; GFX9-NEXT: v_lshrrev_b64 v[58:59], 24, v[9:10]
-; GFX9-NEXT: buffer_store_dword v58, off, s[0:3], s32 offset:148 ; 4-byte Folded Spill
-; GFX9-NEXT: s_nop 0
-; GFX9-NEXT: buffer_store_dword v59, off, s[0:3], s32 offset:152 ; 4-byte Folded Spill
-; GFX9-NEXT: v_lshrrev_b64 v[58:59], 24, v[7:8]
-; GFX9-NEXT: buffer_store_dword v58, off, s[0:3], s32 offset:156 ; 4-byte Folded Spill
-; GFX9-NEXT: s_nop 0
-; GFX9-NEXT: buffer_store_dword v59, off, s[0:3], s32 offset:160 ; 4-byte Folded Spill
-; GFX9-NEXT: v_lshrrev_b64 v[58:59], 24, v[5:6]
-; GFX9-NEXT: buffer_store_dword v58, off, s[0:3], s32 offset:172 ; 4-byte Folded Spill
-; GFX9-NEXT: s_nop 0
-; GFX9-NEXT: buffer_store_dword v59, off, s[0:3], s32 offset:176 ; 4-byte Folded Spill
-; GFX9-NEXT: v_lshrrev_b64 v[58:59], 24, v[3:4]
-; GFX9-NEXT: buffer_store_dword v58, off, s[0:3], s32 offset:188 ; 4-byte Folded Spill
-; GFX9-NEXT: s_nop 0
-; GFX9-NEXT: buffer_store_dword v59, off, s[0:3], s32 offset:192 ; 4-byte Folded Spill
-; GFX9-NEXT: v_lshrrev_b64 v[58:59], 24, v[1:2]
-; GFX9-NEXT: buffer_store_dword v58, off, s[0:3], s32 offset:196 ; 4-byte Folded Spill
-; GFX9-NEXT: s_nop 0
-; GFX9-NEXT: buffer_store_dword v59, off, s[0:3], s32 offset:200 ; 4-byte Folded Spill
-; GFX9-NEXT: v_lshrrev_b64 v[58:59], 24, v[31:32]
-; GFX9-NEXT: buffer_store_dword v58, off, s[0:3], s32 offset:76 ; 4-byte Folded Spill
-; GFX9-NEXT: s_nop 0
-; GFX9-NEXT: buffer_store_dword v59, off, s[0:3], s32 offset:80 ; 4-byte Folded Spill
-; GFX9-NEXT: v_lshrrev_b64 v[58:59], 24, v[29:30]
-; GFX9-NEXT: buffer_store_dword v58, off, s[0:3], s32 offset:84 ; 4-byte Folded Spill
-; GFX9-NEXT: s_nop 0
-; GFX9-NEXT: buffer_store_dword v59, off, s[0:3], s32 offset:88 ; 4-byte Folded Spill
-; GFX9-NEXT: v_lshrrev_b64 v[58:59], 24, v[27:28]
-; GFX9-NEXT: buffer_store_dword v58, off, s[0:3], s32 offset:92 ; 4-byte Folded Spill
-; GFX9-NEXT: s_nop 0
-; GFX9-NEXT: buffer_store_dword v59, off, s[0:3], s32 offset:96 ; 4-byte Folded Spill
-; GFX9-NEXT: v_lshrrev_b64 v[58:59], 24, v[25:26]
-; GFX9-NEXT: buffer_store_dword v58, off, s[0:3], s32 offset:100 ; 4-byte Folded Spill
-; GFX9-NEXT: s_nop 0
-; GFX9-NEXT: buffer_store_dword v59, off, s[0:3], s32 offset:104 ; 4-byte Folded Spill
-; GFX9-NEXT: v_lshrrev_b64 v[58:59], 24, v[23:24]
-; GFX9-NEXT: buffer_store_dword v58, off, s[0:3], s32 offset:108 ; 4-byte Folded Spill
-; GFX9-NEXT: s_nop 0
-; GFX9-NEXT: buffer_store_dword v59, off, s[0:3], s32 offset:112 ; 4-byte Folded Spill
-; GFX9-NEXT: v_lshrrev_b64 v[58:59], 24, v[21:22]
-; GFX9-NEXT: buffer_store_dword v58, off, s[0:3], s32 offset:116 ; 4-byte Folded Spill
-; GFX9-NEXT: s_nop 0
-; GFX9-NEXT: buffer_store_dword v59, off, s[0:3], s32 offset:120 ; 4-byte Folded Spill
-; GFX9-NEXT: v_lshrrev_b64 v[58:59], 24, v[19:20]
-; GFX9-NEXT: buffer_store_dword v58, off, s[0:3], s32 offset:164 ; 4-byte Folded Spill
-; GFX9-NEXT: s_nop 0
-; GFX9-NEXT: buffer_store_dword v59, off, s[0:3], s32 offset:168 ; 4-byte Folded Spill
-; GFX9-NEXT: v_lshrrev_b64 v[58:59], 24, v[17:18]
-; GFX9-NEXT: v_lshrrev_b32_e32 v54, 8, v11
-; GFX9-NEXT: v_lshrrev_b32_e32 v40, 8, v10
-; GFX9-NEXT: v_lshrrev_b32_e32 v45, 8, v9
-; GFX9-NEXT: v_lshrrev_b32_e32 v39, 8, v8
+; GFX9-NEXT: buffer_store_dword v15, off, s[0:3], s32 offset:224 ; 4-byte Folded Spill
+; GFX9-NEXT: v_lshrrev_b32_e32 v15, 8, v8
+; GFX9-NEXT: buffer_store_dword v15, off, s[0:3], s32 offset:424 ; 4-byte Folded Spill
; GFX9-NEXT: v_lshrrev_b32_e32 v15, 8, v7
-; GFX9-NEXT: v_lshrrev_b32_e32 v42, 8, v6
-; GFX9-NEXT: v_lshrrev_b32_e32 v16, 8, v5
-; GFX9-NEXT: v_lshrrev_b32_e32 v52, 8, v4
-; GFX9-NEXT: v_lshrrev_b32_e32 v49, 16, v3
-; GFX9-NEXT: v_lshrrev_b32_e32 v48, 8, v3
-; GFX9-NEXT: v_lshrrev_b32_e32 v53, 24, v2
-; GFX9-NEXT: v_lshrrev_b32_e32 v51, 8, v2
-; GFX9-NEXT: v_lshrrev_b32_e32 v38, 16, v1
-; GFX9-NEXT: v_lshrrev_b32_e32 v41, 8, v1
-; GFX9-NEXT: v_lshrrev_b32_e32 v63, 8, v24
-; GFX9-NEXT: v_lshrrev_b32_e32 v36, 16, v23
-; GFX9-NEXT: v_lshrrev_b32_e32 v57, 8, v23
-; GFX9-NEXT: v_lshrrev_b32_e32 v37, 24, v22
-; GFX9-NEXT: v_lshrrev_b32_e32 v50, 8, v22
-; GFX9-NEXT: v_lshrrev_b32_e32 v56, 16, v21
-; GFX9-NEXT: v_lshrrev_b32_e32 v55, 8, v21
-; GFX9-NEXT: v_lshrrev_b32_e32 v33, 24, v20
-; GFX9-NEXT: v_lshrrev_b32_e32 v43, 8, v20
-; GFX9-NEXT: v_lshrrev_b32_e32 v34, 16, v19
-; GFX9-NEXT: v_lshrrev_b32_e32 v44, 8, v19
-; GFX9-NEXT: v_lshrrev_b32_e32 v35, 24, v18
-; GFX9-NEXT: v_lshrrev_b32_e32 v60, 8, v18
-; GFX9-NEXT: v_lshrrev_b32_e32 v61, 16, v17
-; GFX9-NEXT: v_lshrrev_b32_e32 v62, 8, v17
-; GFX9-NEXT: buffer_store_dword v58, off, s[0:3], s32 offset:180 ; 4-byte Folded Spill
-; GFX9-NEXT: s_nop 0
-; GFX9-NEXT: buffer_store_dword v59, off, s[0:3], s32 offset:184 ; 4-byte Folded Spill
+; GFX9-NEXT: buffer_store_dword v15, off, s[0:3], s32 offset:428 ; 4-byte Folded Spill
+; GFX9-NEXT: v_lshrrev_b32_e32 v15, 24, v6
+; GFX9-NEXT: buffer_store_dword v15, off, s[0:3], s32 offset:380 ; 4-byte Folded Spill
+; GFX9-NEXT: v_lshrrev_b32_e32 v15, 16, v6
+; GFX9-NEXT: buffer_store_dword v15, off, s[0:3], s32 offset:232 ; 4-byte Folded Spill
+; GFX9-NEXT: v_lshrrev_b32_e32 v15, 24, v4
+; GFX9-NEXT: buffer_store_dword v15, off, s[0:3], s32 offset:396 ; 4-byte Folded Spill
+; GFX9-NEXT: v_lshrrev_b32_e32 v15, 16, v4
+; GFX9-NEXT: buffer_store_dword v15, off, s[0:3], s32 offset:240 ; 4-byte Folded Spill
+; GFX9-NEXT: v_lshrrev_b32_e32 v15, 16, v3
+; GFX9-NEXT: buffer_store_dword v15, off, s[0:3], s32 offset:412 ; 4-byte Folded Spill
+; GFX9-NEXT: v_lshrrev_b32_e32 v15, 16, v2
+; GFX9-NEXT: buffer_store_dword v15, off, s[0:3], s32 offset:248 ; 4-byte Folded Spill
+; GFX9-NEXT: v_lshrrev_b32_e32 v15, 16, v1
+; GFX9-NEXT: buffer_store_dword v15, off, s[0:3], s32 offset:416 ; 4-byte Folded Spill
+; GFX9-NEXT: v_lshrrev_b32_e32 v15, 16, v32
+; GFX9-NEXT: buffer_store_dword v15, off, s[0:3], s32 offset:200 ; 4-byte Folded Spill
+; GFX9-NEXT: v_lshrrev_b32_e32 v15, 8, v32
+; GFX9-NEXT: buffer_store_dword v15, off, s[0:3], s32 offset:312 ; 4-byte Folded Spill
+; GFX9-NEXT: s_waitcnt vmcnt(59)
+; GFX9-NEXT: v_lshrrev_b32_e32 v15, 16, v31
+; GFX9-NEXT: buffer_store_dword v15, off, s[0:3], s32 offset:256 ; 4-byte Folded Spill
+; GFX9-NEXT: v_lshrrev_b32_e32 v15, 8, v31
+; GFX9-NEXT: buffer_store_dword v15, off, s[0:3], s32 offset:260 ; 4-byte Folded Spill
+; GFX9-NEXT: v_lshrrev_b32_e32 v15, 24, v30
+; GFX9-NEXT: buffer_store_dword v15, off, s[0:3], s32 offset:272 ; 4-byte Folded Spill
+; GFX9-NEXT: v_lshrrev_b32_e32 v15, 16, v30
+; GFX9-NEXT: buffer_store_dword v15, off, s[0:3], s32 offset:208 ; 4-byte Folded Spill
+; GFX9-NEXT: v_lshrrev_b32_e32 v15, 8, v30
+; GFX9-NEXT: buffer_store_dword v15, off, s[0:3], s32 offset:284 ; 4-byte Folded Spill
+; GFX9-NEXT: v_lshrrev_b32_e32 v15, 8, v29
+; GFX9-NEXT: buffer_store_dword v15, off, s[0:3], s32 offset:292 ; 4-byte Folded Spill
+; GFX9-NEXT: v_lshrrev_b32_e32 v15, 24, v28
+; GFX9-NEXT: buffer_store_dword v15, off, s[0:3], s32 offset:304 ; 4-byte Folded Spill
+; GFX9-NEXT: v_lshrrev_b32_e32 v15, 16, v28
+; GFX9-NEXT: buffer_store_dword v15, off, s[0:3], s32 offset:216 ; 4-byte Folded Spill
+; GFX9-NEXT: v_lshrrev_b32_e32 v15, 8, v28
+; GFX9-NEXT: buffer_store_dword v15, off, s[0:3], s32 offset:308 ; 4-byte Folded Spill
+; GFX9-NEXT: v_lshrrev_b32_e32 v15, 8, v27
+; GFX9-NEXT: buffer_store_dword v15, off, s[0:3], s32 offset:316 ; 4-byte Folded Spill
+; GFX9-NEXT: v_lshrrev_b32_e32 v15, 24, v26
+; GFX9-NEXT: buffer_store_dword v15, off, s[0:3], s32 offset:328 ; 4-byte Folded Spill
+; GFX9-NEXT: v_lshrrev_b32_e32 v15, 16, v26
+; GFX9-NEXT: buffer_store_dword v15, off, s[0:3], s32 offset:228 ; 4-byte Folded Spill
+; GFX9-NEXT: v_lshrrev_b32_e32 v15, 8, v26
+; GFX9-NEXT: buffer_store_dword v15, off, s[0:3], s32 offset:336 ; 4-byte Folded Spill
+; GFX9-NEXT: v_lshrrev_b32_e32 v15, 16, v25
+; GFX9-NEXT: buffer_store_dword v15, off, s[0:3], s32 offset:340 ; 4-byte Folded Spill
+; GFX9-NEXT: v_lshrrev_b32_e32 v15, 8, v25
+; GFX9-NEXT: buffer_store_dword v15, off, s[0:3], s32 offset:348 ; 4-byte Folded Spill
+; GFX9-NEXT: v_lshrrev_b32_e32 v15, 24, v24
+; GFX9-NEXT: buffer_store_dword v15, off, s[0:3], s32 offset:360 ; 4-byte Folded Spill
+; GFX9-NEXT: v_lshrrev_b32_e32 v15, 16, v24
+; GFX9-NEXT: buffer_store_dword v15, off, s[0:3], s32 offset:236 ; 4-byte Folded Spill
+; GFX9-NEXT: v_lshrrev_b32_e32 v15, 8, v23
+; GFX9-NEXT: buffer_store_dword v15, off, s[0:3], s32 offset:368 ; 4-byte Folded Spill
+; GFX9-NEXT: v_lshrrev_b32_e32 v15, 16, v22
+; GFX9-NEXT: buffer_store_dword v15, off, s[0:3], s32 offset:244 ; 4-byte Folded Spill
+; GFX9-NEXT: v_lshrrev_b32_e32 v15, 8, v22
+; GFX9-NEXT: buffer_store_dword v15, off, s[0:3], s32 offset:388 ; 4-byte Folded Spill
+; GFX9-NEXT: v_lshrrev_b32_e32 v15, 16, v21
+; GFX9-NEXT: buffer_store_dword v15, off, s[0:3], s32 offset:392 ; 4-byte Folded Spill
+; GFX9-NEXT: v_lshrrev_b32_e32 v15, 8, v21
+; GFX9-NEXT: buffer_store_dword v15, off, s[0:3], s32 offset:400 ; 4-byte Folded Spill
+; GFX9-NEXT: v_lshrrev_b32_e32 v15, 16, v20
+; GFX9-NEXT: buffer_store_dword v15, off, s[0:3], s32 offset:252 ; 4-byte Folded Spill
+; GFX9-NEXT: v_lshrrev_b32_e32 v15, 16, v18
+; GFX9-NEXT: buffer_store_dword v15, off, s[0:3], s32 offset:280 ; 4-byte Folded Spill
+; GFX9-NEXT: v_lshrrev_b64 v[15:16], 24, v[36:37]
+; GFX9-NEXT: buffer_store_dword v15, off, s[0:3], s32 offset:124 ; 4-byte Folded Spill
+; GFX9-NEXT: s_nop 0
+; GFX9-NEXT: buffer_store_dword v16, off, s[0:3], s32 offset:128 ; 4-byte Folded Spill
+; GFX9-NEXT: v_lshrrev_b64 v[15:16], 24, v[13:14]
+; GFX9-NEXT: buffer_store_dword v15, off, s[0:3], s32 offset:132 ; 4-byte Folded Spill
+; GFX9-NEXT: s_nop 0
+; GFX9-NEXT: buffer_store_dword v16, off, s[0:3], s32 offset:136 ; 4-byte Folded Spill
+; GFX9-NEXT: v_lshrrev_b64 v[15:16], 24, v[11:12]
+; GFX9-NEXT: buffer_store_dword v15, off, s[0:3], s32 offset:140 ; 4-byte Folded Spill
+; GFX9-NEXT: s_nop 0
+; GFX9-NEXT: buffer_store_dword v16, off, s[0:3], s32 offset:144 ; 4-byte Folded Spill
+; GFX9-NEXT: v_lshrrev_b64 v[15:16], 24, v[9:10]
+; GFX9-NEXT: buffer_store_dword v15, off, s[0:3], s32 offset:148 ; 4-byte Folded Spill
+; GFX9-NEXT: s_nop 0
+; GFX9-NEXT: buffer_store_dword v16, off, s[0:3], s32 offset:152 ; 4-byte Folded Spill
+; GFX9-NEXT: v_lshrrev_b64 v[15:16], 24, v[7:8]
+; GFX9-NEXT: v_lshrrev_b64 v[33:34], 24, v[3:4]
+; GFX9-NEXT: buffer_store_dword v15, off, s[0:3], s32 offset:156 ; 4-byte Folded Spill
+; GFX9-NEXT: s_nop 0
+; GFX9-NEXT: buffer_store_dword v16, off, s[0:3], s32 offset:160 ; 4-byte Folded Spill
+; GFX9-NEXT: buffer_store_dword v33, off, s[0:3], s32 offset:180 ; 4-byte Folded Spill
+; GFX9-NEXT: s_nop 0
+; GFX9-NEXT: buffer_store_dword v34, off, s[0:3], s32 offset:184 ; 4-byte Folded Spill
+; GFX9-NEXT: v_lshrrev_b64 v[33:34], 24, v[1:2]
+; GFX9-NEXT: buffer_store_dword v33, off, s[0:3], s32 offset:188 ; 4-byte Folded Spill
+; GFX9-NEXT: s_nop 0
+; GFX9-NEXT: buffer_store_dword v34, off, s[0:3], s32 offset:192 ; 4-byte Folded Spill
+; GFX9-NEXT: v_lshrrev_b64 v[33:34], 24, v[31:32]
+; GFX9-NEXT: buffer_store_dword v33, off, s[0:3], s32 offset:76 ; 4-byte Folded Spill
+; GFX9-NEXT: s_nop 0
+; GFX9-NEXT: buffer_store_dword v34, off, s[0:3], s32 offset:80 ; 4-byte Folded Spill
+; GFX9-NEXT: v_lshrrev_b64 v[33:34], 24, v[29:30]
+; GFX9-NEXT: buffer_store_dword v33, off, s[0:3], s32 offset:84 ; 4-byte Folded Spill
+; GFX9-NEXT: s_nop 0
+; GFX9-NEXT: buffer_store_dword v34, off, s[0:3], s32 offset:88 ; 4-byte Folded Spill
+; GFX9-NEXT: v_lshrrev_b64 v[33:34], 24, v[27:28]
+; GFX9-NEXT: buffer_store_dword v33, off, s[0:3], s32 offset:116 ; 4-byte Folded Spill
+; GFX9-NEXT: s_nop 0
+; GFX9-NEXT: buffer_store_dword v34, off, s[0:3], s32 offset:120 ; 4-byte Folded Spill
+; GFX9-NEXT: v_lshrrev_b64 v[33:34], 24, v[25:26]
+; GFX9-NEXT: buffer_store_dword v33, off, s[0:3], s32 offset:92 ; 4-byte Folded Spill
+; GFX9-NEXT: s_nop 0
+; GFX9-NEXT: buffer_store_dword v34, off, s[0:3], s32 offset:96 ; 4-byte Folded Spill
+; GFX9-NEXT: v_lshrrev_b64 v[33:34], 24, v[23:24]
+; GFX9-NEXT: buffer_store_dword v33, off, s[0:3], s32 offset:100 ; 4-byte Folded Spill
+; GFX9-NEXT: s_nop 0
+; GFX9-NEXT: buffer_store_dword v34, off, s[0:3], s32 offset:104 ; 4-byte Folded Spill
+; GFX9-NEXT: v_lshrrev_b64 v[33:34], 24, v[21:22]
+; GFX9-NEXT: buffer_store_dword v33, off, s[0:3], s32 offset:108 ; 4-byte Folded Spill
+; GFX9-NEXT: s_nop 0
+; GFX9-NEXT: buffer_store_dword v34, off, s[0:3], s32 offset:112 ; 4-byte Folded Spill
+; GFX9-NEXT: v_lshrrev_b64 v[33:34], 24, v[19:20]
+; GFX9-NEXT: buffer_store_dword v33, off, s[0:3], s32 offset:164 ; 4-byte Folded Spill
+; GFX9-NEXT: s_nop 0
+; GFX9-NEXT: buffer_store_dword v34, off, s[0:3], s32 offset:168 ; 4-byte Folded Spill
+; GFX9-NEXT: v_lshrrev_b64 v[33:34], 24, v[17:18]
+; GFX9-NEXT: v_lshrrev_b32_e32 v35, 8, v36
+; GFX9-NEXT: v_lshrrev_b32_e32 v38, 24, v12
+; GFX9-NEXT: v_lshrrev_b32_e32 v53, 8, v11
+; GFX9-NEXT: v_lshrrev_b32_e32 v55, 8, v10
+; GFX9-NEXT: v_lshrrev_b32_e32 v49, 16, v7
+; GFX9-NEXT: v_lshrrev_b32_e32 v40, 8, v6
+; GFX9-NEXT: v_lshrrev_b32_e32 v52, 16, v5
+; GFX9-NEXT: v_lshrrev_b32_e32 v41, 8, v5
+; GFX9-NEXT: v_lshrrev_b32_e32 v42, 8, v4
+; GFX9-NEXT: v_lshrrev_b32_e32 v46, 8, v3
+; GFX9-NEXT: v_lshrrev_b32_e32 v54, 24, v2
+; GFX9-NEXT: v_lshrrev_b32_e32 v59, 8, v2
+; GFX9-NEXT: v_lshrrev_b32_e32 v57, 8, v1
+; GFX9-NEXT: v_lshrrev_b32_e32 v61, 24, v32
+; GFX9-NEXT: v_lshrrev_b32_e32 v44, 16, v29
+; GFX9-NEXT: v_lshrrev_b32_e32 v39, 16, v27
+; GFX9-NEXT: v_lshrrev_b32_e32 v48, 8, v24
+; GFX9-NEXT: v_lshrrev_b32_e32 v47, 16, v23
+; GFX9-NEXT: v_lshrrev_b32_e32 v58, 24, v22
+; GFX9-NEXT: v_lshrrev_b32_e32 v63, 24, v20
+; GFX9-NEXT: v_lshrrev_b32_e32 v62, 8, v20
+; GFX9-NEXT: v_lshrrev_b32_e32 v51, 16, v19
+; GFX9-NEXT: v_lshrrev_b32_e32 v50, 8, v19
+; GFX9-NEXT: v_lshrrev_b32_e32 v43, 24, v18
+; GFX9-NEXT: v_lshrrev_b32_e32 v56, 8, v18
+; GFX9-NEXT: v_lshrrev_b32_e32 v45, 16, v17
+; GFX9-NEXT: v_lshrrev_b32_e32 v60, 8, v17
+; GFX9-NEXT: v_lshrrev_b64 v[15:16], 24, v[5:6]
+; GFX9-NEXT: buffer_store_dword v33, off, s[0:3], s32 offset:172 ; 4-byte Folded Spill
+; GFX9-NEXT: s_nop 0
+; GFX9-NEXT: buffer_store_dword v34, off, s[0:3], s32 offset:176 ; 4-byte Folded Spill
; GFX9-NEXT: .LBB90_2: ; %Flow
-; GFX9-NEXT: s_or_saveexec_b64 s[4:5], s[4:5]
-; GFX9-NEXT: v_mov_b32_e32 v58, v57
-; GFX9-NEXT: s_xor_b64 exec, exec, s[4:5]
+; GFX9-NEXT: s_andn2_saveexec_b64 s[4:5], s[4:5]
; GFX9-NEXT: s_cbranch_execz .LBB90_4
; GFX9-NEXT: ; %bb.3: ; %cmp.true
-; GFX9-NEXT: v_lshlrev_b32_e32 v15, 16, v18
-; GFX9-NEXT: v_add_f32_e32 v15, 0x40c00000, v15
-; GFX9-NEXT: v_bfe_u32 v16, v15, 16, 1
+; GFX9-NEXT: v_lshlrev_b32_e32 v33, 16, v18
+; GFX9-NEXT: v_add_f32_e32 v33, 0x40c00000, v33
+; GFX9-NEXT: v_and_b32_e32 v18, 0xffff0000, v18
+; GFX9-NEXT: v_bfe_u32 v34, v33, 16, 1
; GFX9-NEXT: s_movk_i32 s6, 0x7fff
-; GFX9-NEXT: v_add3_u32 v16, v16, v15, s6
-; GFX9-NEXT: v_or_b32_e32 v33, 0x400000, v15
-; GFX9-NEXT: v_cmp_u_f32_e32 vcc, v15, v15
-; GFX9-NEXT: v_and_b32_e32 v15, 0xffff0000, v18
-; GFX9-NEXT: v_add_f32_e32 v15, 0x40c00000, v15
-; GFX9-NEXT: v_cndmask_b32_e32 v33, v16, v33, vcc
-; GFX9-NEXT: v_bfe_u32 v16, v15, 16, 1
-; GFX9-NEXT: v_add3_u32 v16, v16, v15, s6
-; GFX9-NEXT: v_or_b32_e32 v18, 0x400000, v15
-; GFX9-NEXT: v_cmp_u_f32_e32 vcc, v15, v15
-; GFX9-NEXT: v_cndmask_b32_e32 v15, v16, v18, vcc
+; GFX9-NEXT: v_add_f32_e32 v18, 0x40c00000, v18
+; GFX9-NEXT: v_add3_u32 v34, v34, v33, s6
+; GFX9-NEXT: v_or_b32_e32 v35, 0x400000, v33
+; GFX9-NEXT: v_cmp_u_f32_e32 vcc, v33, v33
+; GFX9-NEXT: v_bfe_u32 v33, v18, 16, 1
+; GFX9-NEXT: v_cndmask_b32_e32 v39, v34, v35, vcc
+; GFX9-NEXT: v_add3_u32 v33, v33, v18, s6
+; GFX9-NEXT: v_or_b32_e32 v34, 0x400000, v18
+; GFX9-NEXT: v_cmp_u_f32_e32 vcc, v18, v18
+; GFX9-NEXT: v_lshlrev_b32_e32 v18, 16, v17
+; GFX9-NEXT: v_add_f32_e32 v18, 0x40c00000, v18
+; GFX9-NEXT: v_and_b32_e32 v17, 0xffff0000, v17
+; GFX9-NEXT: v_cndmask_b32_e32 v62, v33, v34, vcc
+; GFX9-NEXT: v_bfe_u32 v33, v18, 16, 1
+; GFX9-NEXT: v_add_f32_e32 v17, 0x40c00000, v17
+; GFX9-NEXT: v_add3_u32 v33, v33, v18, s6
+; GFX9-NEXT: v_or_b32_e32 v34, 0x400000, v18
+; GFX9-NEXT: v_cmp_u_f32_e32 vcc, v18, v18
+; GFX9-NEXT: v_bfe_u32 v18, v17, 16, 1
+; GFX9-NEXT: v_cndmask_b32_e32 v15, v33, v34, vcc
+; GFX9-NEXT: v_add3_u32 v18, v18, v17, s6
+; GFX9-NEXT: v_or_b32_e32 v33, 0x400000, v17
+; GFX9-NEXT: v_cmp_u_f32_e32 vcc, v17, v17
; GFX9-NEXT: s_mov_b32 s7, 0x7060302
+; GFX9-NEXT: v_cndmask_b32_e32 v17, v18, v33, vcc
+; GFX9-NEXT: buffer_store_dword v15, off, s[0:3], s32 offset:108 ; 4-byte Folded Spill
+; GFX9-NEXT: v_perm_b32 v15, v17, v15, s7
+; GFX9-NEXT: v_lshlrev_b32_e32 v17, 16, v20
+; GFX9-NEXT: v_add_f32_e32 v17, 0x40c00000, v17
+; GFX9-NEXT: v_bfe_u32 v18, v17, 16, 1
+; GFX9-NEXT: v_add3_u32 v18, v18, v17, s6
+; GFX9-NEXT: v_or_b32_e32 v33, 0x400000, v17
+; GFX9-NEXT: v_cmp_u_f32_e32 vcc, v17, v17
+; GFX9-NEXT: v_and_b32_e32 v17, 0xffff0000, v20
+; GFX9-NEXT: v_add_f32_e32 v17, 0x40c00000, v17
+; GFX9-NEXT: v_cndmask_b32_e32 v63, v18, v33, vcc
+; GFX9-NEXT: v_bfe_u32 v18, v17, 16, 1
+; GFX9-NEXT: v_perm_b32 v16, v62, v39, s7
+; GFX9-NEXT: v_add3_u32 v18, v18, v17, s6
+; GFX9-NEXT: v_or_b32_e32 v20, 0x400000, v17
+; GFX9-NEXT: v_cmp_u_f32_e32 vcc, v17, v17
+; GFX9-NEXT: v_lshlrev_b32_e32 v17, 16, v19
+; GFX9-NEXT: buffer_store_dword v15, off, s[0:3], s32 offset:432 ; 4-byte Folded Spill
+; GFX9-NEXT: s_nop 0
+; GFX9-NEXT: buffer_store_dword v16, off, s[0:3], s32 offset:436 ; 4-byte Folded Spill
+; GFX9-NEXT: v_add_f32_e32 v17, 0x40c00000, v17
+; GFX9-NEXT: v_cndmask_b32_e32 v15, v18, v20, vcc
+; GFX9-NEXT: v_bfe_u32 v18, v17, 16, 1
+; GFX9-NEXT: v_add3_u32 v18, v18, v17, s6
+; GFX9-NEXT: v_or_b32_e32 v20, 0x400000, v17
+; GFX9-NEXT: v_cmp_u_f32_e32 vcc, v17, v17
+; GFX9-NEXT: v_and_b32_e32 v17, 0xffff0000, v19
+; GFX9-NEXT: v_add_f32_e32 v17, 0x40c00000, v17
+; GFX9-NEXT: buffer_store_dword v15, off, s[0:3], s32 offset:76 ; 4-byte Folded Spill
+; GFX9-NEXT: v_perm_b32 v16, v15, v63, s7
+; GFX9-NEXT: v_cndmask_b32_e32 v15, v18, v20, vcc
+; GFX9-NEXT: v_bfe_u32 v18, v17, 16, 1
+; GFX9-NEXT: v_add3_u32 v18, v18, v17, s6
+; GFX9-NEXT: v_or_b32_e32 v19, 0x400000, v17
+; GFX9-NEXT: v_cmp_u_f32_e32 vcc, v17, v17
+; GFX9-NEXT: v_cndmask_b32_e32 v17, v18, v19, vcc
; GFX9-NEXT: buffer_store_dword v15, off, s[0:3], s32 offset:84 ; 4-byte Folded Spill
-; GFX9-NEXT: v_perm_b32 v34, v15, v33, s7
-; GFX9-NEXT: v_lshlrev_b32_e32 v15, 16, v17
-; GFX9-NEXT: v_add_f32_e32 v15, 0x40c00000, v15
-; GFX9-NEXT: v_bfe_u32 v16, v15, 16, 1
-; GFX9-NEXT: v_add3_u32 v16, v16, v15, s6
-; GFX9-NEXT: v_or_b32_e32 v18, 0x400000, v15
-; GFX9-NEXT: v_cmp_u_f32_e32 vcc, v15, v15
-; GFX9-NEXT: v_and_b32_e32 v15, 0xffff0000, v17
-; GFX9-NEXT: v_add_f32_e32 v15, 0x40c00000, v15
-; GFX9-NEXT: v_cndmask_b32_e32 v18, v16, v18, vcc
-; GFX9-NEXT: v_bfe_u32 v16, v15, 16, 1
-; GFX9-NEXT: v_add3_u32 v16, v16, v15, s6
-; GFX9-NEXT: v_or_b32_e32 v17, 0x400000, v15
-; GFX9-NEXT: v_cmp_u_f32_e32 vcc, v15, v15
-; GFX9-NEXT: v_cndmask_b32_e32 v15, v16, v17, vcc
-; GFX9-NEXT: buffer_store_dword v33, off, s[0:3], s32 offset:76 ; 4-byte Folded Spill
-; GFX9-NEXT: v_perm_b32 v33, v15, v18, s7
-; GFX9-NEXT: v_lshlrev_b32_e32 v15, 16, v20
-; GFX9-NEXT: v_add_f32_e32 v15, 0x40c00000, v15
-; GFX9-NEXT: v_bfe_u32 v16, v15, 16, 1
-; GFX9-NEXT: v_add3_u32 v16, v16, v15, s6
-; GFX9-NEXT: v_or_b32_e32 v17, 0x400000, v15
-; GFX9-NEXT: v_cmp_u_f32_e32 vcc, v15, v15
-; GFX9-NEXT: v_and_b32_e32 v15, 0xffff0000, v20
-; GFX9-NEXT: v_add_f32_e32 v15, 0x40c00000, v15
-; GFX9-NEXT: buffer_store_dword v18, off, s[0:3], s32 offset:92 ; 4-byte Folded Spill
-; GFX9-NEXT: v_cndmask_b32_e32 v18, v16, v17, vcc
-; GFX9-NEXT: v_bfe_u32 v16, v15, 16, 1
-; GFX9-NEXT: v_add3_u32 v16, v16, v15, s6
-; GFX9-NEXT: v_or_b32_e32 v17, 0x400000, v15
-; GFX9-NEXT: v_cmp_u_f32_e32 vcc, v15, v15
-; GFX9-NEXT: v_cndmask_b32_e32 v15, v16, v17, vcc
-; GFX9-NEXT: buffer_store_dword v33, off, s[0:3], s32 offset:408 ; 4-byte Folded Spill
+; GFX9-NEXT: v_perm_b32 v15, v17, v15, s7
+; GFX9-NEXT: v_lshlrev_b32_e32 v17, 16, v22
+; GFX9-NEXT: v_add_f32_e32 v17, 0x40c00000, v17
+; GFX9-NEXT: v_bfe_u32 v18, v17, 16, 1
+; GFX9-NEXT: v_add3_u32 v18, v18, v17, s6
+; GFX9-NEXT: v_or_b32_e32 v19, 0x400000, v17
+; GFX9-NEXT: v_cmp_u_f32_e32 vcc, v17, v17
+; GFX9-NEXT: v_and_b32_e32 v17, 0xffff0000, v22
+; GFX9-NEXT: buffer_store_dword v15, off, s[0:3], s32 offset:440 ; 4-byte Folded Spill
; GFX9-NEXT: s_nop 0
-; GFX9-NEXT: buffer_store_dword v34, off, s[0:3], s32 offset:412 ; 4-byte Folded Spill
-; GFX9-NEXT: buffer_store_dword v18, off, s[0:3], s32 offset:100 ; 4-byte Folded Spill
-; GFX9-NEXT: buffer_store_dword v15, off, s[0:3], s32 offset:108 ; 4-byte Folded Spill
-; GFX9-NEXT: v_perm_b32 v18, v15, v18, s7
-; GFX9-NEXT: v_lshlrev_b32_e32 v15, 16, v19
-; GFX9-NEXT: v_add_f32_e32 v15, 0x40c00000, v15
-; GFX9-NEXT: v_bfe_u32 v16, v15, 16, 1
-; GFX9-NEXT: v_add3_u32 v16, v16, v15, s6
-; GFX9-NEXT: v_or_b32_e32 v17, 0x400000, v15
-; GFX9-NEXT: v_cmp_u_f32_e32 vcc, v15, v15
-; GFX9-NEXT: v_and_b32_e32 v15, 0xffff0000, v19
-; GFX9-NEXT: v_add_f32_e32 v15, 0x40c00000, v15
-; GFX9-NEXT: v_cndmask_b32_e32 v20, v16, v17, vcc
-; GFX9-NEXT: v_bfe_u32 v16, v15, 16, 1
-; GFX9-NEXT: v_add3_u32 v16, v16, v15, s6
-; GFX9-NEXT: v_or_b32_e32 v17, 0x400000, v15
-; GFX9-NEXT: v_cmp_u_f32_e32 vcc, v15, v15
-; GFX9-NEXT: v_cndmask_b32_e32 v15, v16, v17, vcc
-; GFX9-NEXT: v_perm_b32 v17, v15, v20, s7
-; GFX9-NEXT: v_lshlrev_b32_e32 v15, 16, v22
-; GFX9-NEXT: v_add_f32_e32 v15, 0x40c00000, v15
-; GFX9-NEXT: buffer_store_dword v20, off, s[0:3], s32 offset:116 ; 4-byte Folded Spill
-; GFX9-NEXT: buffer_store_dword v17, off, s[0:3], s32 offset:416 ; 4-byte Folded Spill
+; GFX9-NEXT: buffer_store_dword v16, off, s[0:3], s32 offset:444 ; 4-byte Folded Spill
+; GFX9-NEXT: v_add_f32_e32 v17, 0x40c00000, v17
+; GFX9-NEXT: v_cndmask_b32_e32 v15, v18, v19, vcc
+; GFX9-NEXT: v_bfe_u32 v18, v17, 16, 1
+; GFX9-NEXT: v_add3_u32 v18, v18, v17, s6
+; GFX9-NEXT: v_or_b32_e32 v19, 0x400000, v17
+; GFX9-NEXT: v_cmp_u_f32_e32 vcc, v17, v17
+; GFX9-NEXT: v_lshlrev_b32_e32 v17, 16, v21
+; GFX9-NEXT: v_add_f32_e32 v17, 0x40c00000, v17
+; GFX9-NEXT: v_cndmask_b32_e32 v16, v18, v19, vcc
+; GFX9-NEXT: v_bfe_u32 v18, v17, 16, 1
+; GFX9-NEXT: v_add3_u32 v18, v18, v17, s6
+; GFX9-NEXT: v_or_b32_e32 v19, 0x400000, v17
+; GFX9-NEXT: v_cmp_u_f32_e32 vcc, v17, v17
+; GFX9-NEXT: v_and_b32_e32 v17, 0xffff0000, v21
+; GFX9-NEXT: v_add_f32_e32 v17, 0x40c00000, v17
+; GFX9-NEXT: v_cndmask_b32_e32 v43, v18, v19, vcc
+; GFX9-NEXT: v_bfe_u32 v18, v17, 16, 1
+; GFX9-NEXT: v_add3_u32 v18, v18, v17, s6
+; GFX9-NEXT: v_or_b32_e32 v19, 0x400000, v17
+; GFX9-NEXT: v_cmp_u_f32_e32 vcc, v17, v17
+; GFX9-NEXT: v_cndmask_b32_e32 v17, v18, v19, vcc
+; GFX9-NEXT: buffer_store_dword v15, off, s[0:3], s32 offset:92 ; 4-byte Folded Spill
+; GFX9-NEXT: buffer_store_dword v16, off, s[0:3], s32 offset:100 ; 4-byte Folded Spill
+; GFX9-NEXT: v_perm_b32 v16, v16, v15, s7
+; GFX9-NEXT: v_perm_b32 v15, v17, v43, s7
+; GFX9-NEXT: v_lshlrev_b32_e32 v17, 16, v24
+; GFX9-NEXT: v_add_f32_e32 v17, 0x40c00000, v17
+; GFX9-NEXT: v_bfe_u32 v18, v17, 16, 1
+; GFX9-NEXT: v_add3_u32 v18, v18, v17, s6
+; GFX9-NEXT: v_or_b32_e32 v19, 0x400000, v17
+; GFX9-NEXT: v_cmp_u_f32_e32 vcc, v17, v17
+; GFX9-NEXT: v_and_b32_e32 v17, 0xffff0000, v24
+; GFX9-NEXT: v_add_f32_e32 v17, 0x40c00000, v17
+; GFX9-NEXT: v_cndmask_b32_e32 v38, v18, v19, vcc
+; GFX9-NEXT: v_bfe_u32 v18, v17, 16, 1
+; GFX9-NEXT: v_add3_u32 v18, v18, v17, s6
+; GFX9-NEXT: v_or_b32_e32 v19, 0x400000, v17
+; GFX9-NEXT: v_cmp_u_f32_e32 vcc, v17, v17
+; GFX9-NEXT: v_lshlrev_b32_e32 v17, 16, v23
+; GFX9-NEXT: v_add_f32_e32 v17, 0x40c00000, v17
+; GFX9-NEXT: v_cndmask_b32_e32 v24, v18, v19, vcc
+; GFX9-NEXT: v_bfe_u32 v18, v17, 16, 1
+; GFX9-NEXT: v_add3_u32 v18, v18, v17, s6
+; GFX9-NEXT: v_or_b32_e32 v19, 0x400000, v17
+; GFX9-NEXT: v_cmp_u_f32_e32 vcc, v17, v17
+; GFX9-NEXT: v_and_b32_e32 v17, 0xffff0000, v23
+; GFX9-NEXT: v_add_f32_e32 v17, 0x40c00000, v17
+; GFX9-NEXT: v_cndmask_b32_e32 v22, v18, v19, vcc
+; GFX9-NEXT: v_bfe_u32 v18, v17, 16, 1
+; GFX9-NEXT: v_add3_u32 v18, v18, v17, s6
+; GFX9-NEXT: v_or_b32_e32 v19, 0x400000, v17
+; GFX9-NEXT: v_cmp_u_f32_e32 vcc, v17, v17
+; GFX9-NEXT: buffer_store_dword v15, off, s[0:3], s32 offset:400 ; 4-byte Folded Spill
; GFX9-NEXT: s_nop 0
-; GFX9-NEXT: buffer_store_dword v18, off, s[0:3], s32 offset:420 ; 4-byte Folded Spill
-; GFX9-NEXT: v_bfe_u32 v16, v15, 16, 1
-; GFX9-NEXT: v_add3_u32 v16, v16, v15, s6
-; GFX9-NEXT: v_or_b32_e32 v17, 0x400000, v15
-; GFX9-NEXT: v_cmp_u_f32_e32 vcc, v15, v15
-; GFX9-NEXT: v_and_b32_e32 v15, 0xffff0000, v22
-; GFX9-NEXT: v_add_f32_e32 v15, 0x40c00000, v15
-; GFX9-NEXT: v_cndmask_b32_e32 v18, v16, v17, vcc
-; GFX9-NEXT: v_bfe_u32 v16, v15, 16, 1
-; GFX9-NEXT: v_add3_u32 v16, v16, v15, s6
-; GFX9-NEXT: v_or_b32_e32 v17, 0x400000, v15
-; GFX9-NEXT: v_cmp_u_f32_e32 vcc, v15, v15
-; GFX9-NEXT: v_cndmask_b32_e32 v15, v16, v17, vcc
-; GFX9-NEXT: buffer_store_dword v18, off, s[0:3], s32 offset:132 ; 4-byte Folded Spill
-; GFX9-NEXT: buffer_store_dword v15, off, s[0:3], s32 offset:124 ; 4-byte Folded Spill
-; GFX9-NEXT: v_perm_b32 v18, v15, v18, s7
-; GFX9-NEXT: v_lshlrev_b32_e32 v15, 16, v21
-; GFX9-NEXT: v_add_f32_e32 v15, 0x40c00000, v15
-; GFX9-NEXT: v_bfe_u32 v16, v15, 16, 1
-; GFX9-NEXT: v_add3_u32 v16, v16, v15, s6
-; GFX9-NEXT: v_or_b32_e32 v17, 0x400000, v15
-; GFX9-NEXT: v_cmp_u_f32_e32 vcc, v15, v15
-; GFX9-NEXT: v_and_b32_e32 v15, 0xffff0000, v21
-; GFX9-NEXT: v_add_f32_e32 v15, 0x40c00000, v15
-; GFX9-NEXT: v_cndmask_b32_e32 v61, v16, v17, vcc
-; GFX9-NEXT: v_bfe_u32 v16, v15, 16, 1
-; GFX9-NEXT: v_add3_u32 v16, v16, v15, s6
-; GFX9-NEXT: v_or_b32_e32 v17, 0x400000, v15
-; GFX9-NEXT: v_cmp_u_f32_e32 vcc, v15, v15
-; GFX9-NEXT: v_cndmask_b32_e32 v15, v16, v17, vcc
-; GFX9-NEXT: v_perm_b32 v17, v15, v61, s7
-; GFX9-NEXT: v_lshlrev_b32_e32 v15, 16, v24
-; GFX9-NEXT: v_add_f32_e32 v15, 0x40c00000, v15
-; GFX9-NEXT: buffer_store_dword v17, off, s[0:3], s32 offset:424 ; 4-byte Folded Spill
+; GFX9-NEXT: buffer_store_dword v16, off, s[0:3], s32 offset:404 ; 4-byte Folded Spill
+; GFX9-NEXT: v_cndmask_b32_e32 v17, v18, v19, vcc
+; GFX9-NEXT: v_perm_b32 v15, v17, v22, s7
+; GFX9-NEXT: v_lshlrev_b32_e32 v17, 16, v26
+; GFX9-NEXT: v_add_f32_e32 v17, 0x40c00000, v17
+; GFX9-NEXT: v_bfe_u32 v18, v17, 16, 1
+; GFX9-NEXT: v_add3_u32 v18, v18, v17, s6
+; GFX9-NEXT: v_or_b32_e32 v19, 0x400000, v17
+; GFX9-NEXT: v_cmp_u_f32_e32 vcc, v17, v17
+; GFX9-NEXT: v_and_b32_e32 v17, 0xffff0000, v26
+; GFX9-NEXT: v_add_f32_e32 v17, 0x40c00000, v17
+; GFX9-NEXT: v_cndmask_b32_e32 v23, v18, v19, vcc
+; GFX9-NEXT: v_bfe_u32 v18, v17, 16, 1
+; GFX9-NEXT: v_add3_u32 v18, v18, v17, s6
+; GFX9-NEXT: v_or_b32_e32 v19, 0x400000, v17
+; GFX9-NEXT: v_cmp_u_f32_e32 vcc, v17, v17
+; GFX9-NEXT: v_lshlrev_b32_e32 v17, 16, v25
+; GFX9-NEXT: v_add_f32_e32 v17, 0x40c00000, v17
+; GFX9-NEXT: v_cndmask_b32_e32 v26, v18, v19, vcc
+; GFX9-NEXT: v_bfe_u32 v18, v17, 16, 1
+; GFX9-NEXT: v_add3_u32 v18, v18, v17, s6
+; GFX9-NEXT: v_or_b32_e32 v19, 0x400000, v17
+; GFX9-NEXT: v_cmp_u_f32_e32 vcc, v17, v17
+; GFX9-NEXT: v_and_b32_e32 v17, 0xffff0000, v25
+; GFX9-NEXT: v_add_f32_e32 v17, 0x40c00000, v17
+; GFX9-NEXT: v_cndmask_b32_e32 v50, v18, v19, vcc
+; GFX9-NEXT: v_bfe_u32 v18, v17, 16, 1
+; GFX9-NEXT: v_perm_b32 v16, v24, v38, s7
+; GFX9-NEXT: v_add3_u32 v18, v18, v17, s6
+; GFX9-NEXT: v_or_b32_e32 v19, 0x400000, v17
+; GFX9-NEXT: v_cmp_u_f32_e32 vcc, v17, v17
+; GFX9-NEXT: buffer_store_dword v15, off, s[0:3], s32 offset:368 ; 4-byte Folded Spill
; GFX9-NEXT: s_nop 0
-; GFX9-NEXT: buffer_store_dword v18, off, s[0:3], s32 offset:428 ; 4-byte Folded Spill
-; GFX9-NEXT: v_bfe_u32 v16, v15, 16, 1
-; GFX9-NEXT: v_add3_u32 v16, v16, v15, s6
-; GFX9-NEXT: v_or_b32_e32 v17, 0x400000, v15
-; GFX9-NEXT: v_cmp_u_f32_e32 vcc, v15, v15
-; GFX9-NEXT: v_and_b32_e32 v15, 0xffff0000, v24
-; GFX9-NEXT: v_add_f32_e32 v15, 0x40c00000, v15
-; GFX9-NEXT: v_cndmask_b32_e32 v18, v16, v17, vcc
-; GFX9-NEXT: v_bfe_u32 v16, v15, 16, 1
-; GFX9-NEXT: v_add3_u32 v16, v16, v15, s6
-; GFX9-NEXT: v_or_b32_e32 v17, 0x400000, v15
-; GFX9-NEXT: v_cmp_u_f32_e32 vcc, v15, v15
-; GFX9-NEXT: v_cndmask_b32_e32 v15, v16, v17, vcc
-; GFX9-NEXT: buffer_store_dword v15, off, s[0:3], s32 offset:148 ; 4-byte Folded Spill
-; GFX9-NEXT: v_perm_b32 v59, v15, v18, s7
-; GFX9-NEXT: v_lshlrev_b32_e32 v15, 16, v23
-; GFX9-NEXT: v_add_f32_e32 v15, 0x40c00000, v15
-; GFX9-NEXT: v_bfe_u32 v16, v15, 16, 1
-; GFX9-NEXT: v_add3_u32 v16, v16, v15, s6
-; GFX9-NEXT: v_or_b32_e32 v17, 0x400000, v15
-; GFX9-NEXT: v_cmp_u_f32_e32 vcc, v15, v15
-; GFX9-NEXT: v_and_b32_e32 v15, 0xffff0000, v23
-; GFX9-NEXT: v_add_f32_e32 v15, 0x40c00000, v15
-; GFX9-NEXT: buffer_store_dword v18, off, s[0:3], s32 offset:140 ; 4-byte Folded Spill
-; GFX9-NEXT: v_cndmask_b32_e32 v18, v16, v17, vcc
-; GFX9-NEXT: v_bfe_u32 v16, v15, 16, 1
-; GFX9-NEXT: v_add3_u32 v16, v16, v15, s6
-; GFX9-NEXT: v_or_b32_e32 v17, 0x400000, v15
-; GFX9-NEXT: v_cmp_u_f32_e32 vcc, v15, v15
-; GFX9-NEXT: v_cndmask_b32_e32 v15, v16, v17, vcc
-; GFX9-NEXT: v_perm_b32 v58, v15, v18, s7
-; GFX9-NEXT: v_lshlrev_b32_e32 v15, 16, v26
-; GFX9-NEXT: v_add_f32_e32 v15, 0x40c00000, v15
-; GFX9-NEXT: v_bfe_u32 v16, v15, 16, 1
-; GFX9-NEXT: v_add3_u32 v16, v16, v15, s6
-; GFX9-NEXT: v_or_b32_e32 v17, 0x400000, v15
-; GFX9-NEXT: v_cmp_u_f32_e32 vcc, v15, v15
-; GFX9-NEXT: v_and_b32_e32 v15, 0xffff0000, v26
-; GFX9-NEXT: v_add_f32_e32 v15, 0x40c00000, v15
-; GFX9-NEXT: buffer_store_dword v18, off, s[0:3], s32 offset:156 ; 4-byte Folded Spill
-; GFX9-NEXT: v_cndmask_b32_e32 v18, v16, v17, vcc
-; GFX9-NEXT: v_bfe_u32 v16, v15, 16, 1
-; GFX9-NEXT: v_add3_u32 v16, v16, v15, s6
-; GFX9-NEXT: v_or_b32_e32 v17, 0x400000, v15
-; GFX9-NEXT: v_cmp_u_f32_e32 vcc, v15, v15
-; GFX9-NEXT: v_cndmask_b32_e32 v15, v16, v17, vcc
-; GFX9-NEXT: buffer_store_dword v15, off, s[0:3], s32 offset:172 ; 4-byte Folded Spill
-; GFX9-NEXT: v_perm_b32 v63, v15, v18, s7
-; GFX9-NEXT: v_lshlrev_b32_e32 v15, 16, v25
-; GFX9-NEXT: v_add_f32_e32 v15, 0x40c00000, v15
-; GFX9-NEXT: v_bfe_u32 v16, v15, 16, 1
-; GFX9-NEXT: v_add3_u32 v16, v16, v15, s6
-; GFX9-NEXT: v_or_b32_e32 v17, 0x400000, v15
-; GFX9-NEXT: v_cmp_u_f32_e32 vcc, v15, v15
-; GFX9-NEXT: v_and_b32_e32 v15, 0xffff0000, v25
-; GFX9-NEXT: v_add_f32_e32 v15, 0x40c00000, v15
-; GFX9-NEXT: buffer_store_dword v18, off, s[0:3], s32 offset:164 ; 4-byte Folded Spill
-; GFX9-NEXT: v_cndmask_b32_e32 v18, v16, v17, vcc
-; GFX9-NEXT: v_bfe_u32 v16, v15, 16, 1
-; GFX9-NEXT: v_add3_u32 v16, v16, v15, s6
-; GFX9-NEXT: v_or_b32_e32 v17, 0x400000, v15
-; GFX9-NEXT: v_cmp_u_f32_e32 vcc, v15, v15
-; GFX9-NEXT: v_cndmask_b32_e32 v15, v16, v17, vcc
-; GFX9-NEXT: v_perm_b32 v62, v15, v18, s7
-; GFX9-NEXT: v_lshlrev_b32_e32 v15, 16, v28
-; GFX9-NEXT: v_add_f32_e32 v15, 0x40c00000, v15
-; GFX9-NEXT: v_bfe_u32 v16, v15, 16, 1
-; GFX9-NEXT: v_add3_u32 v16, v16, v15, s6
-; GFX9-NEXT: v_or_b32_e32 v17, 0x400000, v15
-; GFX9-NEXT: v_cmp_u_f32_e32 vcc, v15, v15
-; GFX9-NEXT: v_and_b32_e32 v15, 0xffff0000, v28
-; GFX9-NEXT: v_add_f32_e32 v15, 0x40c00000, v15
-; GFX9-NEXT: v_cndmask_b32_e32 v60, v16, v17, vcc
-; GFX9-NEXT: v_bfe_u32 v16, v15, 16, 1
-; GFX9-NEXT: v_add3_u32 v16, v16, v15, s6
-; GFX9-NEXT: v_or_b32_e32 v17, 0x400000, v15
-; GFX9-NEXT: v_cmp_u_f32_e32 vcc, v15, v15
-; GFX9-NEXT: v_lshlrev_b32_e32 v15, 16, v27
-; GFX9-NEXT: v_add_f32_e32 v15, 0x40c00000, v15
-; GFX9-NEXT: v_cndmask_b32_e32 v28, v16, v17, vcc
-; GFX9-NEXT: v_bfe_u32 v16, v15, 16, 1
-; GFX9-NEXT: v_add3_u32 v16, v16, v15, s6
-; GFX9-NEXT: v_or_b32_e32 v17, 0x400000, v15
-; GFX9-NEXT: v_cmp_u_f32_e32 vcc, v15, v15
-; GFX9-NEXT: v_and_b32_e32 v15, 0xffff0000, v27
-; GFX9-NEXT: v_add_f32_e32 v15, 0x40c00000, v15
-; GFX9-NEXT: v_cndmask_b32_e32 v26, v16, v17, vcc
-; GFX9-NEXT: v_bfe_u32 v16, v15, 16, 1
-; GFX9-NEXT: v_add3_u32 v16, v16, v15, s6
-; GFX9-NEXT: v_or_b32_e32 v17, 0x400000, v15
-; GFX9-NEXT: v_cmp_u_f32_e32 vcc, v15, v15
-; GFX9-NEXT: v_cndmask_b32_e32 v15, v16, v17, vcc
-; GFX9-NEXT: v_perm_b32 v56, v15, v26, s7
-; GFX9-NEXT: v_lshlrev_b32_e32 v15, 16, v30
-; GFX9-NEXT: v_add_f32_e32 v15, 0x40c00000, v15
-; GFX9-NEXT: v_bfe_u32 v16, v15, 16, 1
-; GFX9-NEXT: v_add3_u32 v16, v16, v15, s6
-; GFX9-NEXT: v_or_b32_e32 v17, 0x400000, v15
-; GFX9-NEXT: v_cmp_u_f32_e32 vcc, v15, v15
-; GFX9-NEXT: v_and_b32_e32 v15, 0xffff0000, v30
-; GFX9-NEXT: v_add_f32_e32 v15, 0x40c00000, v15
-; GFX9-NEXT: v_cndmask_b32_e32 v27, v16, v17, vcc
-; GFX9-NEXT: v_bfe_u32 v16, v15, 16, 1
-; GFX9-NEXT: v_add3_u32 v16, v16, v15, s6
-; GFX9-NEXT: v_or_b32_e32 v17, 0x400000, v15
-; GFX9-NEXT: v_cmp_u_f32_e32 vcc, v15, v15
-; GFX9-NEXT: v_lshlrev_b32_e32 v15, 16, v29
-; GFX9-NEXT: v_add_f32_e32 v15, 0x40c00000, v15
-; GFX9-NEXT: v_cndmask_b32_e32 v30, v16, v17, vcc
-; GFX9-NEXT: v_bfe_u32 v16, v15, 16, 1
-; GFX9-NEXT: v_add3_u32 v16, v16, v15, s6
-; GFX9-NEXT: v_or_b32_e32 v17, 0x400000, v15
-; GFX9-NEXT: v_cmp_u_f32_e32 vcc, v15, v15
-; GFX9-NEXT: v_and_b32_e32 v15, 0xffff0000, v29
-; GFX9-NEXT: v_add_f32_e32 v15, 0x40c00000, v15
-; GFX9-NEXT: v_cndmask_b32_e32 v25, v16, v17, vcc
-; GFX9-NEXT: v_bfe_u32 v16, v15, 16, 1
-; GFX9-NEXT: v_add3_u32 v16, v16, v15, s6
-; GFX9-NEXT: v_or_b32_e32 v17, 0x400000, v15
-; GFX9-NEXT: v_cmp_u_f32_e32 vcc, v15, v15
-; GFX9-NEXT: v_cndmask_b32_e32 v15, v16, v17, vcc
-; GFX9-NEXT: v_perm_b32 v33, v15, v25, s7
-; GFX9-NEXT: v_lshlrev_b32_e32 v15, 16, v32
-; GFX9-NEXT: v_add_f32_e32 v15, 0x40c00000, v15
-; GFX9-NEXT: v_bfe_u32 v16, v15, 16, 1
-; GFX9-NEXT: v_add3_u32 v16, v16, v15, s6
-; GFX9-NEXT: v_or_b32_e32 v17, 0x400000, v15
-; GFX9-NEXT: v_cmp_u_f32_e32 vcc, v15, v15
-; GFX9-NEXT: v_and_b32_e32 v15, 0xffff0000, v32
-; GFX9-NEXT: v_add_f32_e32 v15, 0x40c00000, v15
-; GFX9-NEXT: v_cndmask_b32_e32 v29, v16, v17, vcc
-; GFX9-NEXT: v_bfe_u32 v16, v15, 16, 1
-; GFX9-NEXT: v_add3_u32 v16, v16, v15, s6
-; GFX9-NEXT: v_or_b32_e32 v17, 0x400000, v15
-; GFX9-NEXT: v_cmp_u_f32_e32 vcc, v15, v15
-; GFX9-NEXT: s_waitcnt vmcnt(51)
-; GFX9-NEXT: v_lshlrev_b32_e32 v15, 16, v31
-; GFX9-NEXT: v_add_f32_e32 v15, 0x40c00000, v15
-; GFX9-NEXT: v_cndmask_b32_e32 v32, v16, v17, vcc
-; GFX9-NEXT: v_bfe_u32 v16, v15, 16, 1
-; GFX9-NEXT: v_add3_u32 v16, v16, v15, s6
-; GFX9-NEXT: v_or_b32_e32 v17, 0x400000, v15
-; GFX9-NEXT: v_cmp_u_f32_e32 vcc, v15, v15
-; GFX9-NEXT: v_and_b32_e32 v15, 0xffff0000, v31
-; GFX9-NEXT: v_add_f32_e32 v15, 0x40c00000, v15
-; GFX9-NEXT: v_cndmask_b32_e32 v24, v16, v17, vcc
-; GFX9-NEXT: v_bfe_u32 v16, v15, 16, 1
-; GFX9-NEXT: v_add3_u32 v16, v16, v15, s6
-; GFX9-NEXT: v_or_b32_e32 v17, 0x400000, v15
-; GFX9-NEXT: v_cmp_u_f32_e32 vcc, v15, v15
-; GFX9-NEXT: v_cndmask_b32_e32 v15, v16, v17, vcc
-; GFX9-NEXT: v_perm_b32 v35, v15, v24, s7
-; GFX9-NEXT: v_lshlrev_b32_e32 v15, 16, v2
-; GFX9-NEXT: v_add_f32_e32 v15, 0x40c00000, v15
+; GFX9-NEXT: buffer_store_dword v16, off, s[0:3], s32 offset:372 ; 4-byte Folded Spill
+; GFX9-NEXT: v_cndmask_b32_e32 v17, v18, v19, vcc
+; GFX9-NEXT: v_perm_b32 v15, v17, v50, s7
+; GFX9-NEXT: v_lshlrev_b32_e32 v17, 16, v28
+; GFX9-NEXT: v_add_f32_e32 v17, 0x40c00000, v17
+; GFX9-NEXT: v_bfe_u32 v18, v17, 16, 1
+; GFX9-NEXT: v_add3_u32 v18, v18, v17, s6
+; GFX9-NEXT: v_or_b32_e32 v19, 0x400000, v17
+; GFX9-NEXT: v_cmp_u_f32_e32 vcc, v17, v17
+; GFX9-NEXT: v_and_b32_e32 v17, 0xffff0000, v28
+; GFX9-NEXT: v_add_f32_e32 v17, 0x40c00000, v17
+; GFX9-NEXT: v_cndmask_b32_e32 v25, v18, v19, vcc
+; GFX9-NEXT: v_bfe_u32 v18, v17, 16, 1
+; GFX9-NEXT: v_add3_u32 v18, v18, v17, s6
+; GFX9-NEXT: v_or_b32_e32 v19, 0x400000, v17
+; GFX9-NEXT: v_cmp_u_f32_e32 vcc, v17, v17
+; GFX9-NEXT: v_lshlrev_b32_e32 v17, 16, v27
+; GFX9-NEXT: v_add_f32_e32 v17, 0x40c00000, v17
+; GFX9-NEXT: v_cndmask_b32_e32 v28, v18, v19, vcc
+; GFX9-NEXT: v_bfe_u32 v18, v17, 16, 1
+; GFX9-NEXT: v_add3_u32 v18, v18, v17, s6
+; GFX9-NEXT: v_or_b32_e32 v19, 0x400000, v17
+; GFX9-NEXT: v_cmp_u_f32_e32 vcc, v17, v17
+; GFX9-NEXT: v_and_b32_e32 v17, 0xffff0000, v27
+; GFX9-NEXT: v_add_f32_e32 v17, 0x40c00000, v17
+; GFX9-NEXT: v_cndmask_b32_e32 v51, v18, v19, vcc
+; GFX9-NEXT: v_bfe_u32 v18, v17, 16, 1
+; GFX9-NEXT: v_perm_b32 v16, v26, v23, s7
+; GFX9-NEXT: v_add3_u32 v18, v18, v17, s6
+; GFX9-NEXT: v_or_b32_e32 v19, 0x400000, v17
+; GFX9-NEXT: v_cmp_u_f32_e32 vcc, v17, v17
+; GFX9-NEXT: buffer_store_dword v15, off, s[0:3], s32 offset:348 ; 4-byte Folded Spill
+; GFX9-NEXT: s_nop 0
+; GFX9-NEXT: buffer_store_dword v16, off, s[0:3], s32 offset:352 ; 4-byte Folded Spill
+; GFX9-NEXT: v_cndmask_b32_e32 v17, v18, v19, vcc
+; GFX9-NEXT: v_perm_b32 v15, v17, v51, s7
+; GFX9-NEXT: v_lshlrev_b32_e32 v17, 16, v30
+; GFX9-NEXT: v_add_f32_e32 v17, 0x40c00000, v17
+; GFX9-NEXT: v_bfe_u32 v18, v17, 16, 1
+; GFX9-NEXT: v_add3_u32 v18, v18, v17, s6
+; GFX9-NEXT: v_or_b32_e32 v19, 0x400000, v17
+; GFX9-NEXT: v_cmp_u_f32_e32 vcc, v17, v17
+; GFX9-NEXT: v_and_b32_e32 v17, 0xffff0000, v30
+; GFX9-NEXT: v_add_f32_e32 v17, 0x40c00000, v17
+; GFX9-NEXT: v_cndmask_b32_e32 v27, v18, v19, vcc
+; GFX9-NEXT: v_bfe_u32 v18, v17, 16, 1
+; GFX9-NEXT: v_add3_u32 v18, v18, v17, s6
+; GFX9-NEXT: v_or_b32_e32 v19, 0x400000, v17
+; GFX9-NEXT: v_cmp_u_f32_e32 vcc, v17, v17
+; GFX9-NEXT: v_lshlrev_b32_e32 v17, 16, v29
+; GFX9-NEXT: v_add_f32_e32 v17, 0x40c00000, v17
+; GFX9-NEXT: v_cndmask_b32_e32 v30, v18, v19, vcc
+; GFX9-NEXT: v_bfe_u32 v18, v17, 16, 1
+; GFX9-NEXT: v_add3_u32 v18, v18, v17, s6
+; GFX9-NEXT: v_or_b32_e32 v19, 0x400000, v17
+; GFX9-NEXT: v_cmp_u_f32_e32 vcc, v17, v17
+; GFX9-NEXT: v_and_b32_e32 v17, 0xffff0000, v29
+; GFX9-NEXT: v_add_f32_e32 v17, 0x40c00000, v17
+; GFX9-NEXT: v_cndmask_b32_e32 v48, v18, v19, vcc
+; GFX9-NEXT: v_bfe_u32 v18, v17, 16, 1
+; GFX9-NEXT: v_perm_b32 v16, v28, v25, s7
+; GFX9-NEXT: v_add3_u32 v18, v18, v17, s6
+; GFX9-NEXT: v_or_b32_e32 v19, 0x400000, v17
+; GFX9-NEXT: v_cmp_u_f32_e32 vcc, v17, v17
+; GFX9-NEXT: buffer_store_dword v15, off, s[0:3], s32 offset:316 ; 4-byte Folded Spill
+; GFX9-NEXT: s_nop 0
+; GFX9-NEXT: buffer_store_dword v16, off, s[0:3], s32 offset:320 ; 4-byte Folded Spill
+; GFX9-NEXT: v_cndmask_b32_e32 v17, v18, v19, vcc
+; GFX9-NEXT: v_perm_b32 v15, v17, v48, s7
+; GFX9-NEXT: v_lshlrev_b32_e32 v17, 16, v32
+; GFX9-NEXT: v_add_f32_e32 v17, 0x40c00000, v17
+; GFX9-NEXT: v_bfe_u32 v18, v17, 16, 1
+; GFX9-NEXT: v_add3_u32 v18, v18, v17, s6
+; GFX9-NEXT: v_or_b32_e32 v19, 0x400000, v17
+; GFX9-NEXT: v_cmp_u_f32_e32 vcc, v17, v17
+; GFX9-NEXT: v_and_b32_e32 v17, 0xffff0000, v32
+; GFX9-NEXT: v_add_f32_e32 v17, 0x40c00000, v17
+; GFX9-NEXT: v_cndmask_b32_e32 v29, v18, v19, vcc
+; GFX9-NEXT: v_bfe_u32 v18, v17, 16, 1
+; GFX9-NEXT: v_add3_u32 v18, v18, v17, s6
+; GFX9-NEXT: v_or_b32_e32 v19, 0x400000, v17
+; GFX9-NEXT: v_cmp_u_f32_e32 vcc, v17, v17
+; GFX9-NEXT: s_waitcnt vmcnt(47)
+; GFX9-NEXT: v_lshlrev_b32_e32 v17, 16, v31
+; GFX9-NEXT: v_add_f32_e32 v17, 0x40c00000, v17
+; GFX9-NEXT: v_cndmask_b32_e32 v32, v18, v19, vcc
+; GFX9-NEXT: v_bfe_u32 v18, v17, 16, 1
+; GFX9-NEXT: v_add3_u32 v18, v18, v17, s6
+; GFX9-NEXT: v_or_b32_e32 v19, 0x400000, v17
+; GFX9-NEXT: v_cmp_u_f32_e32 vcc, v17, v17
+; GFX9-NEXT: v_and_b32_e32 v17, 0xffff0000, v31
+; GFX9-NEXT: v_add_f32_e32 v17, 0x40c00000, v17
+; GFX9-NEXT: v_cndmask_b32_e32 v49, v18, v19, vcc
+; GFX9-NEXT: v_bfe_u32 v18, v17, 16, 1
+; GFX9-NEXT: v_perm_b32 v16, v30, v27, s7
+; GFX9-NEXT: v_add3_u32 v18, v18, v17, s6
+; GFX9-NEXT: v_or_b32_e32 v19, 0x400000, v17
+; GFX9-NEXT: v_cmp_u_f32_e32 vcc, v17, v17
+; GFX9-NEXT: buffer_store_dword v15, off, s[0:3], s32 offset:292 ; 4-byte Folded Spill
+; GFX9-NEXT: s_nop 0
+; GFX9-NEXT: buffer_store_dword v16, off, s[0:3], s32 offset:296 ; 4-byte Folded Spill
+; GFX9-NEXT: v_cndmask_b32_e32 v17, v18, v19, vcc
+; GFX9-NEXT: v_perm_b32 v15, v17, v49, s7
+; GFX9-NEXT: v_lshlrev_b32_e32 v17, 16, v2
+; GFX9-NEXT: v_add_f32_e32 v17, 0x40c00000, v17
; GFX9-NEXT: v_and_b32_e32 v2, 0xffff0000, v2
-; GFX9-NEXT: v_bfe_u32 v16, v15, 16, 1
+; GFX9-NEXT: v_bfe_u32 v18, v17, 16, 1
; GFX9-NEXT: v_add_f32_e32 v2, 0x40c00000, v2
-; GFX9-NEXT: v_add3_u32 v16, v16, v15, s6
-; GFX9-NEXT: v_or_b32_e32 v17, 0x400000, v15
-; GFX9-NEXT: v_cmp_u_f32_e32 vcc, v15, v15
-; GFX9-NEXT: v_bfe_u32 v15, v2, 16, 1
-; GFX9-NEXT: v_cndmask_b32_e32 v31, v16, v17, vcc
-; GFX9-NEXT: v_add3_u32 v15, v15, v2, s6
-; GFX9-NEXT: v_or_b32_e32 v16, 0x400000, v2
+; GFX9-NEXT: v_add3_u32 v18, v18, v17, s6
+; GFX9-NEXT: v_or_b32_e32 v19, 0x400000, v17
+; GFX9-NEXT: v_cmp_u_f32_e32 vcc, v17, v17
+; GFX9-NEXT: v_bfe_u32 v17, v2, 16, 1
+; GFX9-NEXT: v_cndmask_b32_e32 v31, v18, v19, vcc
+; GFX9-NEXT: v_add3_u32 v17, v17, v2, s6
+; GFX9-NEXT: v_or_b32_e32 v18, 0x400000, v2
; GFX9-NEXT: v_cmp_u_f32_e32 vcc, v2, v2
; GFX9-NEXT: v_lshlrev_b32_e32 v2, 16, v1
; GFX9-NEXT: v_add_f32_e32 v2, 0x40c00000, v2
; GFX9-NEXT: v_and_b32_e32 v1, 0xffff0000, v1
-; GFX9-NEXT: v_cndmask_b32_e32 v22, v15, v16, vcc
-; GFX9-NEXT: v_bfe_u32 v15, v2, 16, 1
+; GFX9-NEXT: v_cndmask_b32_e32 v21, v17, v18, vcc
+; GFX9-NEXT: v_bfe_u32 v17, v2, 16, 1
; GFX9-NEXT: v_add_f32_e32 v1, 0x40c00000, v1
-; GFX9-NEXT: v_add3_u32 v15, v15, v2, s6
-; GFX9-NEXT: v_or_b32_e32 v16, 0x400000, v2
+; GFX9-NEXT: v_add3_u32 v17, v17, v2, s6
+; GFX9-NEXT: v_or_b32_e32 v18, 0x400000, v2
; GFX9-NEXT: v_cmp_u_f32_e32 vcc, v2, v2
; GFX9-NEXT: v_bfe_u32 v2, v1, 16, 1
-; GFX9-NEXT: v_cndmask_b32_e32 v23, v15, v16, vcc
+; GFX9-NEXT: v_cndmask_b32_e32 v52, v17, v18, vcc
; GFX9-NEXT: v_add3_u32 v2, v2, v1, s6
-; GFX9-NEXT: v_or_b32_e32 v15, 0x400000, v1
+; GFX9-NEXT: v_or_b32_e32 v17, 0x400000, v1
; GFX9-NEXT: v_cmp_u_f32_e32 vcc, v1, v1
-; GFX9-NEXT: v_cndmask_b32_e32 v1, v2, v15, vcc
-; GFX9-NEXT: v_perm_b32 v37, v1, v23, s7
+; GFX9-NEXT: v_cndmask_b32_e32 v1, v2, v17, vcc
+; GFX9-NEXT: v_perm_b32 v60, v1, v52, s7
; GFX9-NEXT: v_lshlrev_b32_e32 v1, 16, v4
; GFX9-NEXT: v_add_f32_e32 v1, 0x40c00000, v1
; GFX9-NEXT: v_bfe_u32 v2, v1, 16, 1
; GFX9-NEXT: v_add3_u32 v2, v2, v1, s6
-; GFX9-NEXT: v_or_b32_e32 v15, 0x400000, v1
+; GFX9-NEXT: v_or_b32_e32 v17, 0x400000, v1
; GFX9-NEXT: v_cmp_u_f32_e32 vcc, v1, v1
; GFX9-NEXT: v_and_b32_e32 v1, 0xffff0000, v4
; GFX9-NEXT: v_add_f32_e32 v1, 0x40c00000, v1
-; GFX9-NEXT: v_cndmask_b32_e32 v21, v2, v15, vcc
+; GFX9-NEXT: v_cndmask_b32_e32 v20, v2, v17, vcc
; GFX9-NEXT: v_bfe_u32 v2, v1, 16, 1
; GFX9-NEXT: v_add3_u32 v2, v2, v1, s6
; GFX9-NEXT: v_or_b32_e32 v4, 0x400000, v1
@@ -166861,13 +166912,13 @@ define <128 x i8> @bitcast_v64bf16_to_v128i8(<64 x bfloat> %a, i32 %b) {
; GFX9-NEXT: v_cmp_u_f32_e32 vcc, v1, v1
; GFX9-NEXT: v_and_b32_e32 v1, 0xffff0000, v3
; GFX9-NEXT: v_add_f32_e32 v1, 0x40c00000, v1
-; GFX9-NEXT: v_cndmask_b32_e32 v20, v2, v4, vcc
+; GFX9-NEXT: v_cndmask_b32_e32 v53, v2, v4, vcc
; GFX9-NEXT: v_bfe_u32 v2, v1, 16, 1
; GFX9-NEXT: v_add3_u32 v2, v2, v1, s6
; GFX9-NEXT: v_or_b32_e32 v3, 0x400000, v1
; GFX9-NEXT: v_cmp_u_f32_e32 vcc, v1, v1
; GFX9-NEXT: v_cndmask_b32_e32 v1, v2, v3, vcc
-; GFX9-NEXT: v_perm_b32 v48, v1, v20, s7
+; GFX9-NEXT: v_perm_b32 v54, v1, v53, s7
; GFX9-NEXT: v_lshlrev_b32_e32 v1, 16, v6
; GFX9-NEXT: v_add_f32_e32 v1, 0x40c00000, v1
; GFX9-NEXT: v_bfe_u32 v2, v1, 16, 1
@@ -166876,7 +166927,6 @@ define <128 x i8> @bitcast_v64bf16_to_v128i8(<64 x bfloat> %a, i32 %b) {
; GFX9-NEXT: v_cmp_u_f32_e32 vcc, v1, v1
; GFX9-NEXT: v_and_b32_e32 v1, 0xffff0000, v6
; GFX9-NEXT: v_add_f32_e32 v1, 0x40c00000, v1
-; GFX9-NEXT: buffer_store_dword v18, off, s[0:3], s32 offset:180 ; 4-byte Folded Spill
; GFX9-NEXT: v_cndmask_b32_e32 v18, v2, v3, vcc
; GFX9-NEXT: v_bfe_u32 v2, v1, 16, 1
; GFX9-NEXT: v_add3_u32 v2, v2, v1, s6
@@ -166897,7 +166947,7 @@ define <128 x i8> @bitcast_v64bf16_to_v128i8(<64 x bfloat> %a, i32 %b) {
; GFX9-NEXT: v_or_b32_e32 v3, 0x400000, v1
; GFX9-NEXT: v_cmp_u_f32_e32 vcc, v1, v1
; GFX9-NEXT: v_cndmask_b32_e32 v1, v2, v3, vcc
-; GFX9-NEXT: v_perm_b32 v50, v1, v17, s7
+; GFX9-NEXT: v_perm_b32 v33, v1, v17, s7
; GFX9-NEXT: v_lshlrev_b32_e32 v1, 16, v8
; GFX9-NEXT: v_add_f32_e32 v1, 0x40c00000, v1
; GFX9-NEXT: v_bfe_u32 v2, v1, 16, 1
@@ -166926,7 +166976,7 @@ define <128 x i8> @bitcast_v64bf16_to_v128i8(<64 x bfloat> %a, i32 %b) {
; GFX9-NEXT: v_or_b32_e32 v7, 0x400000, v1
; GFX9-NEXT: v_cmp_u_f32_e32 vcc, v1, v1
; GFX9-NEXT: v_cndmask_b32_e32 v1, v2, v7, vcc
-; GFX9-NEXT: v_perm_b32 v52, v1, v4, s7
+; GFX9-NEXT: v_perm_b32 v56, v1, v4, s7
; GFX9-NEXT: v_lshlrev_b32_e32 v1, 16, v10
; GFX9-NEXT: v_add_f32_e32 v1, 0x40c00000, v1
; GFX9-NEXT: v_bfe_u32 v2, v1, 16, 1
@@ -166945,480 +166995,515 @@ define <128 x i8> @bitcast_v64bf16_to_v128i8(<64 x bfloat> %a, i32 %b) {
; GFX9-NEXT: v_cndmask_b32_e32 v10, v2, v10, vcc
; GFX9-NEXT: v_bfe_u32 v2, v1, 16, 1
; GFX9-NEXT: v_add3_u32 v2, v2, v1, s6
-; GFX9-NEXT: v_or_b32_e32 v15, 0x400000, v1
+; GFX9-NEXT: v_or_b32_e32 v42, 0x400000, v1
; GFX9-NEXT: v_cmp_u_f32_e32 vcc, v1, v1
; GFX9-NEXT: v_and_b32_e32 v1, 0xffff0000, v9
; GFX9-NEXT: v_add_f32_e32 v1, 0x40c00000, v1
; GFX9-NEXT: v_bfe_u32 v9, v1, 16, 1
-; GFX9-NEXT: v_cndmask_b32_e32 v3, v2, v15, vcc
+; GFX9-NEXT: v_cndmask_b32_e32 v3, v2, v42, vcc
; GFX9-NEXT: v_add3_u32 v9, v9, v1, s6
-; GFX9-NEXT: v_or_b32_e32 v15, 0x400000, v1
+; GFX9-NEXT: v_or_b32_e32 v42, 0x400000, v1
; GFX9-NEXT: v_cmp_u_f32_e32 vcc, v1, v1
-; GFX9-NEXT: v_cndmask_b32_e32 v1, v9, v15, vcc
-; GFX9-NEXT: v_perm_b32 v39, v1, v3, s7
+; GFX9-NEXT: v_cndmask_b32_e32 v1, v9, v42, vcc
+; GFX9-NEXT: v_perm_b32 v40, v1, v3, s7
; GFX9-NEXT: v_lshlrev_b32_e32 v1, 16, v12
; GFX9-NEXT: v_add_f32_e32 v1, 0x40c00000, v1
; GFX9-NEXT: v_bfe_u32 v9, v1, 16, 1
; GFX9-NEXT: v_add3_u32 v9, v9, v1, s6
-; GFX9-NEXT: v_or_b32_e32 v15, 0x400000, v1
+; GFX9-NEXT: v_or_b32_e32 v44, 0x400000, v1
; GFX9-NEXT: v_cmp_u_f32_e32 vcc, v1, v1
; GFX9-NEXT: v_and_b32_e32 v1, 0xffff0000, v12
; GFX9-NEXT: v_add_f32_e32 v1, 0x40c00000, v1
; GFX9-NEXT: v_bfe_u32 v12, v1, 16, 1
-; GFX9-NEXT: v_cndmask_b32_e32 v9, v9, v15, vcc
+; GFX9-NEXT: v_cndmask_b32_e32 v9, v9, v44, vcc
; GFX9-NEXT: v_add3_u32 v12, v12, v1, s6
-; GFX9-NEXT: v_or_b32_e32 v15, 0x400000, v1
+; GFX9-NEXT: v_or_b32_e32 v44, 0x400000, v1
; GFX9-NEXT: v_cmp_u_f32_e32 vcc, v1, v1
; GFX9-NEXT: v_lshlrev_b32_e32 v1, 16, v11
; GFX9-NEXT: v_add_f32_e32 v1, 0x40c00000, v1
-; GFX9-NEXT: v_cndmask_b32_e32 v12, v12, v15, vcc
-; GFX9-NEXT: v_bfe_u32 v15, v1, 16, 1
+; GFX9-NEXT: v_cndmask_b32_e32 v12, v12, v44, vcc
+; GFX9-NEXT: v_bfe_u32 v44, v1, 16, 1
; GFX9-NEXT: v_and_b32_e32 v11, 0xffff0000, v11
-; GFX9-NEXT: v_add3_u32 v15, v15, v1, s6
-; GFX9-NEXT: v_or_b32_e32 v16, 0x400000, v1
+; GFX9-NEXT: v_add3_u32 v44, v44, v1, s6
+; GFX9-NEXT: v_or_b32_e32 v46, 0x400000, v1
; GFX9-NEXT: v_cmp_u_f32_e32 vcc, v1, v1
; GFX9-NEXT: v_add_f32_e32 v11, 0x40c00000, v11
-; GFX9-NEXT: v_cndmask_b32_e32 v2, v15, v16, vcc
-; GFX9-NEXT: v_bfe_u32 v15, v11, 16, 1
-; GFX9-NEXT: v_add3_u32 v15, v15, v11, s6
-; GFX9-NEXT: v_or_b32_e32 v16, 0x400000, v11
+; GFX9-NEXT: v_cndmask_b32_e32 v2, v44, v46, vcc
+; GFX9-NEXT: v_bfe_u32 v44, v11, 16, 1
+; GFX9-NEXT: v_add3_u32 v44, v44, v11, s6
+; GFX9-NEXT: v_or_b32_e32 v46, 0x400000, v11
; GFX9-NEXT: v_cmp_u_f32_e32 vcc, v11, v11
-; GFX9-NEXT: v_cndmask_b32_e32 v11, v15, v16, vcc
-; GFX9-NEXT: v_perm_b32 v54, v11, v2, s7
+; GFX9-NEXT: v_cndmask_b32_e32 v11, v44, v46, vcc
+; GFX9-NEXT: v_perm_b32 v44, v11, v2, s7
; GFX9-NEXT: v_lshlrev_b32_e32 v11, 16, v14
; GFX9-NEXT: v_add_f32_e32 v11, 0x40c00000, v11
-; GFX9-NEXT: v_bfe_u32 v15, v11, 16, 1
+; GFX9-NEXT: v_bfe_u32 v46, v11, 16, 1
; GFX9-NEXT: v_and_b32_e32 v14, 0xffff0000, v14
-; GFX9-NEXT: v_add3_u32 v15, v15, v11, s6
-; GFX9-NEXT: v_or_b32_e32 v16, 0x400000, v11
+; GFX9-NEXT: v_add3_u32 v46, v46, v11, s6
+; GFX9-NEXT: v_or_b32_e32 v47, 0x400000, v11
; GFX9-NEXT: v_cmp_u_f32_e32 vcc, v11, v11
; GFX9-NEXT: v_add_f32_e32 v14, 0x40c00000, v14
-; GFX9-NEXT: v_cndmask_b32_e32 v11, v15, v16, vcc
-; GFX9-NEXT: v_bfe_u32 v15, v14, 16, 1
-; GFX9-NEXT: v_add3_u32 v15, v15, v14, s6
-; GFX9-NEXT: v_or_b32_e32 v16, 0x400000, v14
+; GFX9-NEXT: v_cndmask_b32_e32 v11, v46, v47, vcc
+; GFX9-NEXT: v_bfe_u32 v46, v14, 16, 1
+; GFX9-NEXT: v_add3_u32 v46, v46, v14, s6
+; GFX9-NEXT: v_or_b32_e32 v47, 0x400000, v14
; GFX9-NEXT: v_cmp_u_f32_e32 vcc, v14, v14
-; GFX9-NEXT: v_cndmask_b32_e32 v14, v15, v16, vcc
-; GFX9-NEXT: v_lshlrev_b32_e32 v15, 16, v13
-; GFX9-NEXT: v_add_f32_e32 v15, 0x40c00000, v15
+; GFX9-NEXT: v_cndmask_b32_e32 v14, v46, v47, vcc
+; GFX9-NEXT: v_lshlrev_b32_e32 v46, 16, v13
+; GFX9-NEXT: v_add_f32_e32 v46, 0x40c00000, v46
; GFX9-NEXT: v_and_b32_e32 v13, 0xffff0000, v13
-; GFX9-NEXT: v_bfe_u32 v16, v15, 16, 1
+; GFX9-NEXT: v_bfe_u32 v58, v46, 16, 1
; GFX9-NEXT: v_add_f32_e32 v13, 0x40c00000, v13
-; GFX9-NEXT: v_add3_u32 v16, v16, v15, s6
-; GFX9-NEXT: v_or_b32_e32 v41, 0x400000, v15
-; GFX9-NEXT: v_cmp_u_f32_e32 vcc, v15, v15
-; GFX9-NEXT: v_bfe_u32 v15, v13, 16, 1
-; GFX9-NEXT: v_cndmask_b32_e32 v1, v16, v41, vcc
-; GFX9-NEXT: v_add3_u32 v15, v15, v13, s6
-; GFX9-NEXT: v_or_b32_e32 v16, 0x400000, v13
+; GFX9-NEXT: v_add3_u32 v58, v58, v46, s6
+; GFX9-NEXT: v_or_b32_e32 v59, 0x400000, v46
+; GFX9-NEXT: v_cmp_u_f32_e32 vcc, v46, v46
+; GFX9-NEXT: v_bfe_u32 v46, v13, 16, 1
+; GFX9-NEXT: v_cndmask_b32_e32 v1, v58, v59, vcc
+; GFX9-NEXT: v_add3_u32 v46, v46, v13, s6
+; GFX9-NEXT: v_or_b32_e32 v58, 0x400000, v13
; GFX9-NEXT: v_cmp_u_f32_e32 vcc, v13, v13
-; GFX9-NEXT: v_cndmask_b32_e32 v13, v15, v16, vcc
-; GFX9-NEXT: v_perm_b32 v41, v13, v1, s7
-; GFX9-NEXT: v_lshlrev_b32_e32 v13, 16, v47
+; GFX9-NEXT: v_cndmask_b32_e32 v13, v46, v58, vcc
+; GFX9-NEXT: v_perm_b32 v16, v32, v29, s7
+; GFX9-NEXT: v_perm_b32 v46, v13, v1, s7
+; GFX9-NEXT: v_lshlrev_b32_e32 v13, 16, v37
+; GFX9-NEXT: buffer_store_dword v15, off, s[0:3], s32 offset:260 ; 4-byte Folded Spill
+; GFX9-NEXT: s_nop 0
+; GFX9-NEXT: buffer_store_dword v16, off, s[0:3], s32 offset:264 ; 4-byte Folded Spill
; GFX9-NEXT: v_add_f32_e32 v13, 0x40c00000, v13
-; GFX9-NEXT: v_bfe_u32 v15, v13, 16, 1
-; GFX9-NEXT: v_add3_u32 v15, v15, v13, s6
-; GFX9-NEXT: v_or_b32_e32 v16, 0x400000, v13
+; GFX9-NEXT: v_bfe_u32 v58, v13, 16, 1
+; GFX9-NEXT: v_and_b32_e32 v16, 0xffff0000, v37
+; GFX9-NEXT: v_add3_u32 v58, v58, v13, s6
+; GFX9-NEXT: v_or_b32_e32 v59, 0x400000, v13
; GFX9-NEXT: v_cmp_u_f32_e32 vcc, v13, v13
-; GFX9-NEXT: v_cndmask_b32_e32 v13, v15, v16, vcc
-; GFX9-NEXT: v_and_b32_e32 v15, 0xffff0000, v47
-; GFX9-NEXT: v_add_f32_e32 v15, 0x40c00000, v15
-; GFX9-NEXT: v_bfe_u32 v16, v15, 16, 1
-; GFX9-NEXT: v_add3_u32 v16, v16, v15, s6
-; GFX9-NEXT: v_or_b32_e32 v44, 0x400000, v15
-; GFX9-NEXT: v_cmp_u_f32_e32 vcc, v15, v15
-; GFX9-NEXT: v_cndmask_b32_e32 v44, v16, v44, vcc
-; GFX9-NEXT: v_perm_b32 v16, v44, v13, s7
-; GFX9-NEXT: v_lshrrev_b32_e32 v47, 16, v13
+; GFX9-NEXT: v_add_f32_e32 v16, 0x40c00000, v16
+; GFX9-NEXT: v_cndmask_b32_e32 v13, v58, v59, vcc
+; GFX9-NEXT: v_bfe_u32 v58, v16, 16, 1
+; GFX9-NEXT: v_add3_u32 v58, v58, v16, s6
+; GFX9-NEXT: v_or_b32_e32 v59, 0x400000, v16
+; GFX9-NEXT: v_cmp_u_f32_e32 vcc, v16, v16
+; GFX9-NEXT: v_cndmask_b32_e32 v16, v58, v59, vcc
+; GFX9-NEXT: v_lshlrev_b32_e32 v58, 16, v36
+; GFX9-NEXT: v_add_f32_e32 v58, 0x40c00000, v58
+; GFX9-NEXT: v_bfe_u32 v35, v58, 16, 1
+; GFX9-NEXT: v_add3_u32 v35, v35, v58, s6
+; GFX9-NEXT: v_or_b32_e32 v15, 0x400000, v58
+; GFX9-NEXT: v_cmp_u_f32_e32 vcc, v58, v58
+; GFX9-NEXT: v_cndmask_b32_e32 v15, v35, v15, vcc
+; GFX9-NEXT: v_and_b32_e32 v35, 0xffff0000, v36
+; GFX9-NEXT: v_add_f32_e32 v35, 0x40c00000, v35
+; GFX9-NEXT: v_bfe_u32 v36, v35, 16, 1
+; GFX9-NEXT: v_add3_u32 v36, v36, v35, s6
+; GFX9-NEXT: v_or_b32_e32 v37, 0x400000, v35
+; GFX9-NEXT: v_cmp_u_f32_e32 vcc, v35, v35
+; GFX9-NEXT: v_cndmask_b32_e32 v35, v36, v37, vcc
+; GFX9-NEXT: v_perm_b32 v58, v35, v15, s7
+; GFX9-NEXT: v_lshrrev_b32_e32 v36, 16, v15
+; GFX9-NEXT: v_lshrrev_b32_e32 v15, 16, v32
+; GFX9-NEXT: buffer_store_dword v15, off, s[0:3], s32 offset:200 ; 4-byte Folded Spill
+; GFX9-NEXT: v_lshrrev_b32_e32 v15, 16, v30
+; GFX9-NEXT: buffer_store_dword v15, off, s[0:3], s32 offset:208 ; 4-byte Folded Spill
+; GFX9-NEXT: v_lshrrev_b32_e32 v15, 16, v28
+; GFX9-NEXT: buffer_store_dword v15, off, s[0:3], s32 offset:216 ; 4-byte Folded Spill
+; GFX9-NEXT: v_lshrrev_b32_e32 v15, 16, v26
+; GFX9-NEXT: buffer_store_dword v15, off, s[0:3], s32 offset:228 ; 4-byte Folded Spill
+; GFX9-NEXT: v_lshrrev_b32_e32 v15, 16, v24
+; GFX9-NEXT: buffer_store_dword v15, off, s[0:3], s32 offset:236 ; 4-byte Folded Spill
+; GFX9-NEXT: buffer_load_dword v15, off, s[0:3], s32 offset:100 ; 4-byte Folded Reload
+; GFX9-NEXT: v_lshrrev_b32_e32 v26, 16, v23
+; GFX9-NEXT: v_lshrrev_b32_e32 v23, 16, v22
+; GFX9-NEXT: v_perm_b32 v59, v16, v13, s7
+; GFX9-NEXT: v_lshrrev_b32_e32 v37, 16, v13
; GFX9-NEXT: v_lshrrev_b32_e32 v13, 16, v14
-; GFX9-NEXT: buffer_store_dword v13, off, s[0:3], s32 offset:212 ; 4-byte Folded Spill
+; GFX9-NEXT: buffer_store_dword v13, off, s[0:3], s32 offset:204 ; 4-byte Folded Spill
; GFX9-NEXT: v_lshrrev_b32_e32 v13, 16, v1
; GFX9-NEXT: v_lshrrev_b32_e32 v1, 16, v12
-; GFX9-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:220 ; 4-byte Folded Spill
+; GFX9-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:212 ; 4-byte Folded Spill
; GFX9-NEXT: v_lshrrev_b32_e32 v1, 16, v10
-; GFX9-NEXT: v_perm_b32 v53, v8, v5, s7
-; GFX9-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:228 ; 4-byte Folded Spill
+; GFX9-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:220 ; 4-byte Folded Spill
; GFX9-NEXT: v_lshrrev_b32_e32 v1, 16, v8
+; GFX9-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:224 ; 4-byte Folded Spill
+; GFX9-NEXT: v_lshrrev_b32_e32 v1, 16, v6
+; GFX9-NEXT: v_perm_b32 v55, v19, v20, s7
+; GFX9-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:232 ; 4-byte Folded Spill
+; GFX9-NEXT: v_lshrrev_b32_e32 v1, 16, v19
+; GFX9-NEXT: v_lshrrev_b32_e32 v16, 16, v16
+; GFX9-NEXT: v_perm_b32 v57, v8, v5, s7
+; GFX9-NEXT: buffer_store_dword v16, off, s[0:3], s32 offset:196 ; 4-byte Folded Spill
; GFX9-NEXT: v_lshrrev_b32_e32 v8, 16, v5
; GFX9-NEXT: v_lshrrev_b32_e32 v5, 16, v17
-; GFX9-NEXT: v_lshrrev_b32_e32 v17, 16, v32
-; GFX9-NEXT: buffer_store_dword v17, off, s[0:3], s32 offset:208 ; 4-byte Folded Spill
-; GFX9-NEXT: v_lshrrev_b32_e32 v17, 16, v30
-; GFX9-NEXT: buffer_store_dword v17, off, s[0:3], s32 offset:216 ; 4-byte Folded Spill
-; GFX9-NEXT: v_lshrrev_b32_e32 v17, 16, v28
-; GFX9-NEXT: buffer_store_dword v17, off, s[0:3], s32 offset:224 ; 4-byte Folded Spill
-; GFX9-NEXT: buffer_load_dword v17, off, s[0:3], s32 offset:172 ; 4-byte Folded Reload
-; GFX9-NEXT: v_perm_b32 v34, v30, v27, s7
-; GFX9-NEXT: v_lshrrev_b32_e32 v30, 16, v27
-; GFX9-NEXT: v_lshrrev_b32_e32 v27, 16, v26
-; GFX9-NEXT: v_perm_b32 v36, v32, v29, s7
-; GFX9-NEXT: v_lshrrev_b32_e32 v32, 16, v29
-; GFX9-NEXT: v_lshrrev_b32_e32 v29, 16, v25
-; GFX9-NEXT: v_perm_b32 v38, v22, v31, s7
-; GFX9-NEXT: v_perm_b32 v42, v14, v11, s7
+; GFX9-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:240 ; 4-byte Folded Spill
+; GFX9-NEXT: v_lshrrev_b32_e32 v1, 16, v21
+; GFX9-NEXT: v_perm_b32 v47, v14, v11, s7
+; GFX9-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:248 ; 4-byte Folded Spill
+; GFX9-NEXT: v_perm_b32 v45, v12, v9, s7
+; GFX9-NEXT: v_perm_b32 v41, v10, v7, s7
+; GFX9-NEXT: v_perm_b32 v34, v6, v18, s7
+; GFX9-NEXT: v_lshrrev_b32_e32 v6, 16, v18
+; GFX9-NEXT: v_lshrrev_b32_e32 v24, 16, v38
+; GFX9-NEXT: v_lshrrev_b32_e32 v18, 16, v39
+; GFX9-NEXT: v_mov_b32_e32 v38, v54
+; GFX9-NEXT: v_perm_b32 v61, v21, v31, s7
+; GFX9-NEXT: v_lshrrev_b32_e32 v21, 16, v43
+; GFX9-NEXT: v_mov_b32_e32 v43, v34
+; GFX9-NEXT: v_mov_b32_e32 v39, v55
+; GFX9-NEXT: v_mov_b32_e32 v42, v33
+; GFX9-NEXT: v_lshrrev_b64 v[33:34], 24, v[38:39]
; GFX9-NEXT: v_lshrrev_b32_e32 v14, 16, v11
; GFX9-NEXT: v_lshrrev_b32_e32 v11, 16, v2
; GFX9-NEXT: v_lshrrev_b32_e32 v2, 16, v31
-; GFX9-NEXT: v_lshrrev_b32_e32 v31, 16, v24
-; GFX9-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:232 ; 4-byte Folded Spill
-; GFX9-NEXT: v_lshrrev_b32_e32 v1, 16, v6
-; GFX9-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:240 ; 4-byte Folded Spill
-; GFX9-NEXT: v_lshrrev_b32_e32 v1, 16, v19
-; GFX9-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:248 ; 4-byte Folded Spill
-; GFX9-NEXT: v_lshrrev_b32_e32 v1, 16, v22
-; GFX9-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:256 ; 4-byte Folded Spill
-; GFX9-NEXT: v_lshrrev_b32_e32 v1, 16, v23
-; GFX9-NEXT: v_perm_b32 v55, v12, v9, s7
+; GFX9-NEXT: v_lshrrev_b32_e32 v32, 16, v29
+; GFX9-NEXT: v_lshrrev_b32_e32 v31, 16, v49
+; GFX9-NEXT: v_lshrrev_b32_e32 v29, 16, v48
+; GFX9-NEXT: v_mov_b32_e32 v48, v60
+; GFX9-NEXT: v_mov_b32_e32 v49, v61
+; GFX9-NEXT: v_lshrrev_b32_e32 v30, 16, v27
+; GFX9-NEXT: v_lshrrev_b32_e32 v28, 16, v25
+; GFX9-NEXT: v_lshrrev_b32_e32 v27, 16, v51
+; GFX9-NEXT: v_lshrrev_b32_e32 v25, 16, v50
; GFX9-NEXT: v_lshrrev_b32_e32 v12, 16, v9
; GFX9-NEXT: v_lshrrev_b32_e32 v9, 16, v3
-; GFX9-NEXT: v_lshrrev_b32_e32 v3, 16, v20
-; GFX9-NEXT: v_perm_b32 v49, v19, v21, s7
-; GFX9-NEXT: v_lshlrev_b32_e32 v15, 16, v46
-; GFX9-NEXT: v_add_f32_e32 v15, 0x40c00000, v15
-; GFX9-NEXT: v_bfe_u32 v45, v15, 16, 1
-; GFX9-NEXT: v_add3_u32 v45, v45, v15, s6
-; GFX9-NEXT: v_or_b32_e32 v43, 0x400000, v15
-; GFX9-NEXT: v_cmp_u_f32_e32 vcc, v15, v15
-; GFX9-NEXT: v_and_b32_e32 v15, 0xffff0000, v46
-; GFX9-NEXT: v_add_f32_e32 v15, 0x40c00000, v15
-; GFX9-NEXT: v_cndmask_b32_e32 v43, v45, v43, vcc
-; GFX9-NEXT: v_bfe_u32 v45, v15, 16, 1
-; GFX9-NEXT: v_add3_u32 v45, v45, v15, s6
-; GFX9-NEXT: v_or_b32_e32 v46, 0x400000, v15
-; GFX9-NEXT: v_cmp_u_f32_e32 vcc, v15, v15
-; GFX9-NEXT: v_cndmask_b32_e32 v15, v45, v46, vcc
-; GFX9-NEXT: v_perm_b32 v15, v15, v43, s7
-; GFX9-NEXT: v_lshrrev_b32_e32 v44, 16, v44
-; GFX9-NEXT: buffer_store_dword v44, off, s[0:3], s32 offset:204 ; 4-byte Folded Spill
-; GFX9-NEXT: v_lshrrev_b32_e32 v46, 16, v43
-; GFX9-NEXT: v_lshrrev_b64 v[43:44], 24, v[15:16]
-; GFX9-NEXT: v_perm_b32 v51, v6, v18, s7
-; GFX9-NEXT: v_lshrrev_b32_e32 v6, 16, v18
-; GFX9-NEXT: v_perm_b32 v40, v10, v7, s7
-; GFX9-NEXT: v_perm_b32 v57, v28, v60, s7
+; GFX9-NEXT: v_lshrrev_b32_e32 v3, 16, v53
+; GFX9-NEXT: v_lshrrev_b32_e32 v53, 8, v44
+; GFX9-NEXT: v_lshrrev_b32_e32 v1, 16, v52
+; GFX9-NEXT: v_lshrrev_b32_e32 v55, 8, v41
+; GFX9-NEXT: v_lshrrev_b32_e32 v52, 16, v42
; GFX9-NEXT: v_lshrrev_b32_e32 v10, 16, v7
; GFX9-NEXT: v_lshrrev_b32_e32 v7, 16, v4
-; GFX9-NEXT: v_lshrrev_b32_e32 v4, 16, v21
-; GFX9-NEXT: v_lshrrev_b32_e32 v28, 16, v60
-; GFX9-NEXT: v_lshrrev_b32_e32 v21, 16, v61
-; GFX9-NEXT: v_lshrrev_b32_e32 v45, 8, v39
-; GFX9-NEXT: s_waitcnt vmcnt(5)
-; GFX9-NEXT: v_lshrrev_b32_e32 v17, 16, v17
-; GFX9-NEXT: buffer_store_dword v17, off, s[0:3], s32 offset:236 ; 4-byte Folded Spill
-; GFX9-NEXT: buffer_load_dword v17, off, s[0:3], s32 offset:164 ; 4-byte Folded Reload
-; GFX9-NEXT: s_waitcnt vmcnt(0)
-; GFX9-NEXT: v_lshrrev_b32_e32 v26, 16, v17
-; GFX9-NEXT: buffer_load_dword v17, off, s[0:3], s32 offset:180 ; 4-byte Folded Reload
-; GFX9-NEXT: s_waitcnt vmcnt(0)
-; GFX9-NEXT: v_lshrrev_b32_e32 v25, 16, v17
-; GFX9-NEXT: buffer_load_dword v17, off, s[0:3], s32 offset:148 ; 4-byte Folded Reload
-; GFX9-NEXT: s_waitcnt vmcnt(0)
-; GFX9-NEXT: v_lshrrev_b32_e32 v17, 16, v17
-; GFX9-NEXT: buffer_store_dword v17, off, s[0:3], s32 offset:244 ; 4-byte Folded Spill
-; GFX9-NEXT: buffer_load_dword v17, off, s[0:3], s32 offset:140 ; 4-byte Folded Reload
-; GFX9-NEXT: s_waitcnt vmcnt(0)
-; GFX9-NEXT: v_lshrrev_b32_e32 v24, 16, v17
-; GFX9-NEXT: buffer_load_dword v17, off, s[0:3], s32 offset:156 ; 4-byte Folded Reload
-; GFX9-NEXT: s_waitcnt vmcnt(0)
-; GFX9-NEXT: v_lshrrev_b32_e32 v23, 16, v17
-; GFX9-NEXT: buffer_load_dword v17, off, s[0:3], s32 offset:124 ; 4-byte Folded Reload
-; GFX9-NEXT: s_waitcnt vmcnt(0)
-; GFX9-NEXT: v_lshrrev_b32_e32 v17, 16, v17
-; GFX9-NEXT: buffer_store_dword v17, off, s[0:3], s32 offset:252 ; 4-byte Folded Spill
-; GFX9-NEXT: buffer_load_dword v17, off, s[0:3], s32 offset:132 ; 4-byte Folded Reload
-; GFX9-NEXT: s_waitcnt vmcnt(0)
-; GFX9-NEXT: v_lshrrev_b32_e32 v22, 16, v17
-; GFX9-NEXT: buffer_load_dword v17, off, s[0:3], s32 offset:108 ; 4-byte Folded Reload
-; GFX9-NEXT: s_waitcnt vmcnt(0)
-; GFX9-NEXT: v_lshrrev_b32_e32 v17, 16, v17
-; GFX9-NEXT: buffer_store_dword v17, off, s[0:3], s32 offset:260 ; 4-byte Folded Spill
-; GFX9-NEXT: buffer_load_dword v17, off, s[0:3], s32 offset:100 ; 4-byte Folded Reload
+; GFX9-NEXT: v_lshrrev_b32_e32 v4, 16, v20
+; GFX9-NEXT: v_lshrrev_b32_e32 v20, 16, v63
+; GFX9-NEXT: s_waitcnt vmcnt(8)
+; GFX9-NEXT: v_lshrrev_b32_e32 v15, 16, v15
+; GFX9-NEXT: buffer_store_dword v15, off, s[0:3], s32 offset:244 ; 4-byte Folded Spill
+; GFX9-NEXT: buffer_load_dword v15, off, s[0:3], s32 offset:92 ; 4-byte Folded Reload
; GFX9-NEXT: s_waitcnt vmcnt(0)
-; GFX9-NEXT: v_lshrrev_b32_e32 v20, 16, v17
-; GFX9-NEXT: buffer_load_dword v17, off, s[0:3], s32 offset:116 ; 4-byte Folded Reload
+; GFX9-NEXT: v_lshrrev_b32_e32 v22, 16, v15
+; GFX9-NEXT: buffer_load_dword v15, off, s[0:3], s32 offset:76 ; 4-byte Folded Reload
; GFX9-NEXT: s_waitcnt vmcnt(0)
-; GFX9-NEXT: v_lshrrev_b32_e32 v19, 16, v17
-; GFX9-NEXT: buffer_load_dword v17, off, s[0:3], s32 offset:84 ; 4-byte Folded Reload
+; GFX9-NEXT: v_lshrrev_b32_e32 v15, 16, v15
+; GFX9-NEXT: buffer_store_dword v15, off, s[0:3], s32 offset:252 ; 4-byte Folded Spill
+; GFX9-NEXT: buffer_load_dword v15, off, s[0:3], s32 offset:84 ; 4-byte Folded Reload
; GFX9-NEXT: s_waitcnt vmcnt(0)
-; GFX9-NEXT: v_lshrrev_b32_e32 v17, 16, v17
-; GFX9-NEXT: buffer_store_dword v17, off, s[0:3], s32 offset:292 ; 4-byte Folded Spill
-; GFX9-NEXT: buffer_load_dword v17, off, s[0:3], s32 offset:76 ; 4-byte Folded Reload
+; GFX9-NEXT: v_lshrrev_b32_e32 v19, 16, v15
+; GFX9-NEXT: v_lshrrev_b32_e32 v15, 16, v62
+; GFX9-NEXT: buffer_store_dword v15, off, s[0:3], s32 offset:280 ; 4-byte Folded Spill
+; GFX9-NEXT: buffer_load_dword v15, off, s[0:3], s32 offset:108 ; 4-byte Folded Reload
; GFX9-NEXT: s_waitcnt vmcnt(0)
-; GFX9-NEXT: v_lshrrev_b32_e32 v18, 16, v17
-; GFX9-NEXT: buffer_load_dword v17, off, s[0:3], s32 offset:92 ; 4-byte Folded Reload
+; GFX9-NEXT: v_lshrrev_b32_e32 v17, 16, v15
+; GFX9-NEXT: v_lshrrev_b64 v[15:16], 24, v[58:59]
+; GFX9-NEXT: buffer_store_dword v15, off, s[0:3], s32 offset:124 ; 4-byte Folded Spill
; GFX9-NEXT: s_nop 0
-; GFX9-NEXT: buffer_store_dword v43, off, s[0:3], s32 offset:124 ; 4-byte Folded Spill
+; GFX9-NEXT: buffer_store_dword v16, off, s[0:3], s32 offset:128 ; 4-byte Folded Spill
+; GFX9-NEXT: v_lshrrev_b64 v[15:16], 24, v[46:47]
+; GFX9-NEXT: buffer_store_dword v15, off, s[0:3], s32 offset:132 ; 4-byte Folded Spill
; GFX9-NEXT: s_nop 0
-; GFX9-NEXT: buffer_store_dword v44, off, s[0:3], s32 offset:128 ; 4-byte Folded Spill
-; GFX9-NEXT: v_lshrrev_b64 v[43:44], 24, v[41:42]
-; GFX9-NEXT: buffer_store_dword v43, off, s[0:3], s32 offset:132 ; 4-byte Folded Spill
+; GFX9-NEXT: buffer_store_dword v16, off, s[0:3], s32 offset:136 ; 4-byte Folded Spill
+; GFX9-NEXT: v_lshrrev_b64 v[15:16], 24, v[44:45]
+; GFX9-NEXT: buffer_store_dword v15, off, s[0:3], s32 offset:140 ; 4-byte Folded Spill
; GFX9-NEXT: s_nop 0
-; GFX9-NEXT: buffer_store_dword v44, off, s[0:3], s32 offset:136 ; 4-byte Folded Spill
-; GFX9-NEXT: v_lshrrev_b64 v[43:44], 24, v[54:55]
-; GFX9-NEXT: buffer_store_dword v43, off, s[0:3], s32 offset:140 ; 4-byte Folded Spill
+; GFX9-NEXT: buffer_store_dword v16, off, s[0:3], s32 offset:144 ; 4-byte Folded Spill
+; GFX9-NEXT: v_lshrrev_b64 v[15:16], 24, v[40:41]
+; GFX9-NEXT: buffer_store_dword v15, off, s[0:3], s32 offset:148 ; 4-byte Folded Spill
; GFX9-NEXT: s_nop 0
-; GFX9-NEXT: buffer_store_dword v44, off, s[0:3], s32 offset:144 ; 4-byte Folded Spill
-; GFX9-NEXT: v_lshrrev_b64 v[43:44], 24, v[39:40]
-; GFX9-NEXT: buffer_store_dword v43, off, s[0:3], s32 offset:148 ; 4-byte Folded Spill
+; GFX9-NEXT: buffer_store_dword v16, off, s[0:3], s32 offset:152 ; 4-byte Folded Spill
+; GFX9-NEXT: v_lshrrev_b64 v[15:16], 24, v[56:57]
+; GFX9-NEXT: buffer_store_dword v15, off, s[0:3], s32 offset:156 ; 4-byte Folded Spill
; GFX9-NEXT: s_nop 0
-; GFX9-NEXT: buffer_store_dword v44, off, s[0:3], s32 offset:152 ; 4-byte Folded Spill
-; GFX9-NEXT: v_lshrrev_b64 v[43:44], 24, v[52:53]
-; GFX9-NEXT: buffer_store_dword v43, off, s[0:3], s32 offset:156 ; 4-byte Folded Spill
+; GFX9-NEXT: buffer_store_dword v16, off, s[0:3], s32 offset:160 ; 4-byte Folded Spill
+; GFX9-NEXT: buffer_store_dword v33, off, s[0:3], s32 offset:180 ; 4-byte Folded Spill
; GFX9-NEXT: s_nop 0
-; GFX9-NEXT: buffer_store_dword v44, off, s[0:3], s32 offset:160 ; 4-byte Folded Spill
-; GFX9-NEXT: v_lshrrev_b64 v[43:44], 24, v[50:51]
-; GFX9-NEXT: buffer_store_dword v43, off, s[0:3], s32 offset:172 ; 4-byte Folded Spill
+; GFX9-NEXT: buffer_store_dword v34, off, s[0:3], s32 offset:184 ; 4-byte Folded Spill
+; GFX9-NEXT: v_lshrrev_b64 v[33:34], 24, v[48:49]
+; GFX9-NEXT: buffer_store_dword v33, off, s[0:3], s32 offset:188 ; 4-byte Folded Spill
; GFX9-NEXT: s_nop 0
-; GFX9-NEXT: buffer_store_dword v44, off, s[0:3], s32 offset:176 ; 4-byte Folded Spill
-; GFX9-NEXT: v_lshrrev_b64 v[43:44], 24, v[48:49]
-; GFX9-NEXT: buffer_store_dword v43, off, s[0:3], s32 offset:188 ; 4-byte Folded Spill
+; GFX9-NEXT: buffer_store_dword v34, off, s[0:3], s32 offset:192 ; 4-byte Folded Spill
+; GFX9-NEXT: buffer_load_dword v61, off, s[0:3], s32 offset:260 ; 4-byte Folded Reload
+; GFX9-NEXT: buffer_load_dword v62, off, s[0:3], s32 offset:264 ; 4-byte Folded Reload
+; GFX9-NEXT: v_lshrrev_b64 v[15:16], 24, v[42:43]
+; GFX9-NEXT: v_lshrrev_b32_e32 v16, 24, v59
+; GFX9-NEXT: s_waitcnt vmcnt(0)
+; GFX9-NEXT: v_lshrrev_b64 v[33:34], 24, v[61:62]
+; GFX9-NEXT: buffer_store_dword v33, off, s[0:3], s32 offset:76 ; 4-byte Folded Spill
; GFX9-NEXT: s_nop 0
-; GFX9-NEXT: buffer_store_dword v44, off, s[0:3], s32 offset:192 ; 4-byte Folded Spill
-; GFX9-NEXT: v_lshrrev_b64 v[43:44], 24, v[37:38]
-; GFX9-NEXT: buffer_store_dword v43, off, s[0:3], s32 offset:196 ; 4-byte Folded Spill
+; GFX9-NEXT: buffer_store_dword v34, off, s[0:3], s32 offset:80 ; 4-byte Folded Spill
+; GFX9-NEXT: buffer_load_dword v33, off, s[0:3], s32 offset:292 ; 4-byte Folded Reload
; GFX9-NEXT: s_nop 0
-; GFX9-NEXT: buffer_store_dword v44, off, s[0:3], s32 offset:200 ; 4-byte Folded Spill
-; GFX9-NEXT: v_lshrrev_b64 v[43:44], 24, v[35:36]
-; GFX9-NEXT: buffer_store_dword v43, off, s[0:3], s32 offset:76 ; 4-byte Folded Spill
+; GFX9-NEXT: buffer_load_dword v34, off, s[0:3], s32 offset:296 ; 4-byte Folded Reload
+; GFX9-NEXT: v_lshrrev_b32_e32 v35, 8, v62
+; GFX9-NEXT: s_waitcnt vmcnt(0)
+; GFX9-NEXT: v_lshrrev_b64 v[33:34], 24, v[33:34]
+; GFX9-NEXT: buffer_store_dword v33, off, s[0:3], s32 offset:84 ; 4-byte Folded Spill
; GFX9-NEXT: s_nop 0
-; GFX9-NEXT: buffer_store_dword v44, off, s[0:3], s32 offset:80 ; 4-byte Folded Spill
-; GFX9-NEXT: v_lshrrev_b64 v[43:44], 24, v[33:34]
-; GFX9-NEXT: buffer_store_dword v43, off, s[0:3], s32 offset:84 ; 4-byte Folded Spill
+; GFX9-NEXT: buffer_store_dword v34, off, s[0:3], s32 offset:88 ; 4-byte Folded Spill
+; GFX9-NEXT: buffer_load_dword v33, off, s[0:3], s32 offset:316 ; 4-byte Folded Reload
; GFX9-NEXT: s_nop 0
-; GFX9-NEXT: buffer_store_dword v44, off, s[0:3], s32 offset:88 ; 4-byte Folded Spill
-; GFX9-NEXT: v_lshrrev_b64 v[43:44], 24, v[56:57]
-; GFX9-NEXT: buffer_store_dword v43, off, s[0:3], s32 offset:92 ; 4-byte Folded Spill
+; GFX9-NEXT: buffer_load_dword v34, off, s[0:3], s32 offset:320 ; 4-byte Folded Reload
+; GFX9-NEXT: s_waitcnt vmcnt(0)
+; GFX9-NEXT: v_lshrrev_b64 v[33:34], 24, v[33:34]
+; GFX9-NEXT: buffer_store_dword v33, off, s[0:3], s32 offset:116 ; 4-byte Folded Spill
; GFX9-NEXT: s_nop 0
-; GFX9-NEXT: buffer_store_dword v44, off, s[0:3], s32 offset:96 ; 4-byte Folded Spill
-; GFX9-NEXT: v_lshrrev_b64 v[43:44], 24, v[62:63]
-; GFX9-NEXT: buffer_store_dword v43, off, s[0:3], s32 offset:100 ; 4-byte Folded Spill
+; GFX9-NEXT: buffer_store_dword v34, off, s[0:3], s32 offset:120 ; 4-byte Folded Spill
+; GFX9-NEXT: buffer_load_dword v33, off, s[0:3], s32 offset:348 ; 4-byte Folded Reload
; GFX9-NEXT: s_nop 0
-; GFX9-NEXT: buffer_store_dword v44, off, s[0:3], s32 offset:104 ; 4-byte Folded Spill
-; GFX9-NEXT: v_lshrrev_b64 v[43:44], 24, v[58:59]
-; GFX9-NEXT: buffer_store_dword v43, off, s[0:3], s32 offset:108 ; 4-byte Folded Spill
+; GFX9-NEXT: buffer_load_dword v34, off, s[0:3], s32 offset:352 ; 4-byte Folded Reload
+; GFX9-NEXT: s_waitcnt vmcnt(0)
+; GFX9-NEXT: v_lshrrev_b64 v[33:34], 24, v[33:34]
+; GFX9-NEXT: buffer_store_dword v33, off, s[0:3], s32 offset:92 ; 4-byte Folded Spill
; GFX9-NEXT: s_nop 0
-; GFX9-NEXT: buffer_store_dword v44, off, s[0:3], s32 offset:112 ; 4-byte Folded Spill
-; GFX9-NEXT: buffer_load_dword v60, off, s[0:3], s32 offset:424 ; 4-byte Folded Reload
-; GFX9-NEXT: buffer_load_dword v61, off, s[0:3], s32 offset:428 ; 4-byte Folded Reload
-; GFX9-NEXT: s_waitcnt vmcnt(28)
-; GFX9-NEXT: v_lshrrev_b32_e32 v17, 16, v17
+; GFX9-NEXT: buffer_store_dword v34, off, s[0:3], s32 offset:96 ; 4-byte Folded Spill
+; GFX9-NEXT: buffer_load_dword v33, off, s[0:3], s32 offset:368 ; 4-byte Folded Reload
+; GFX9-NEXT: s_nop 0
+; GFX9-NEXT: buffer_load_dword v34, off, s[0:3], s32 offset:372 ; 4-byte Folded Reload
; GFX9-NEXT: s_waitcnt vmcnt(0)
-; GFX9-NEXT: v_lshrrev_b64 v[43:44], 24, v[60:61]
-; GFX9-NEXT: buffer_store_dword v43, off, s[0:3], s32 offset:116 ; 4-byte Folded Spill
+; GFX9-NEXT: v_lshrrev_b64 v[33:34], 24, v[33:34]
+; GFX9-NEXT: buffer_store_dword v33, off, s[0:3], s32 offset:100 ; 4-byte Folded Spill
; GFX9-NEXT: s_nop 0
-; GFX9-NEXT: buffer_store_dword v44, off, s[0:3], s32 offset:120 ; 4-byte Folded Spill
-; GFX9-NEXT: buffer_load_dword v43, off, s[0:3], s32 offset:416 ; 4-byte Folded Reload
+; GFX9-NEXT: buffer_store_dword v34, off, s[0:3], s32 offset:104 ; 4-byte Folded Spill
+; GFX9-NEXT: buffer_load_dword v33, off, s[0:3], s32 offset:400 ; 4-byte Folded Reload
; GFX9-NEXT: s_nop 0
-; GFX9-NEXT: buffer_load_dword v44, off, s[0:3], s32 offset:420 ; 4-byte Folded Reload
+; GFX9-NEXT: buffer_load_dword v34, off, s[0:3], s32 offset:404 ; 4-byte Folded Reload
; GFX9-NEXT: s_waitcnt vmcnt(0)
-; GFX9-NEXT: v_lshrrev_b64 v[43:44], 24, v[43:44]
-; GFX9-NEXT: buffer_store_dword v43, off, s[0:3], s32 offset:164 ; 4-byte Folded Spill
+; GFX9-NEXT: v_lshrrev_b64 v[33:34], 24, v[33:34]
+; GFX9-NEXT: buffer_store_dword v33, off, s[0:3], s32 offset:108 ; 4-byte Folded Spill
; GFX9-NEXT: s_nop 0
-; GFX9-NEXT: buffer_store_dword v44, off, s[0:3], s32 offset:168 ; 4-byte Folded Spill
-; GFX9-NEXT: buffer_load_dword v43, off, s[0:3], s32 offset:408 ; 4-byte Folded Reload
+; GFX9-NEXT: buffer_store_dword v34, off, s[0:3], s32 offset:112 ; 4-byte Folded Spill
+; GFX9-NEXT: buffer_load_dword v33, off, s[0:3], s32 offset:440 ; 4-byte Folded Reload
; GFX9-NEXT: s_nop 0
-; GFX9-NEXT: buffer_load_dword v44, off, s[0:3], s32 offset:412 ; 4-byte Folded Reload
+; GFX9-NEXT: buffer_load_dword v34, off, s[0:3], s32 offset:444 ; 4-byte Folded Reload
; GFX9-NEXT: s_waitcnt vmcnt(0)
-; GFX9-NEXT: v_lshrrev_b64 v[43:44], 24, v[43:44]
-; GFX9-NEXT: buffer_store_dword v43, off, s[0:3], s32 offset:180 ; 4-byte Folded Spill
+; GFX9-NEXT: v_lshrrev_b64 v[50:51], 24, v[33:34]
+; GFX9-NEXT: buffer_store_dword v50, off, s[0:3], s32 offset:164 ; 4-byte Folded Spill
; GFX9-NEXT: s_nop 0
-; GFX9-NEXT: buffer_store_dword v44, off, s[0:3], s32 offset:184 ; 4-byte Folded Spill
-; GFX9-NEXT: v_lshrrev_b32_e32 v43, 24, v16
-; GFX9-NEXT: v_lshrrev_b32_e32 v16, 8, v16
-; GFX9-NEXT: buffer_store_dword v16, off, s[0:3], s32 offset:364 ; 4-byte Folded Spill
-; GFX9-NEXT: v_lshrrev_b32_e32 v16, 16, v15
-; GFX9-NEXT: v_lshrrev_b32_e32 v15, 8, v15
-; GFX9-NEXT: buffer_store_dword v15, off, s[0:3], s32 offset:376 ; 4-byte Folded Spill
-; GFX9-NEXT: v_lshrrev_b32_e32 v15, 24, v42
-; GFX9-NEXT: buffer_store_dword v15, off, s[0:3], s32 offset:304 ; 4-byte Folded Spill
-; GFX9-NEXT: v_lshrrev_b32_e32 v15, 8, v42
-; GFX9-NEXT: buffer_store_dword v15, off, s[0:3], s32 offset:384 ; 4-byte Folded Spill
-; GFX9-NEXT: v_lshrrev_b32_e32 v15, 16, v41
-; GFX9-NEXT: buffer_store_dword v15, off, s[0:3], s32 offset:312 ; 4-byte Folded Spill
-; GFX9-NEXT: v_lshrrev_b32_e32 v15, 8, v41
-; GFX9-NEXT: buffer_store_dword v15, off, s[0:3], s32 offset:396 ; 4-byte Folded Spill
-; GFX9-NEXT: v_lshrrev_b32_e32 v15, 24, v55
-; GFX9-NEXT: buffer_store_dword v15, off, s[0:3], s32 offset:324 ; 4-byte Folded Spill
-; GFX9-NEXT: v_lshrrev_b32_e32 v15, 8, v55
-; GFX9-NEXT: buffer_store_dword v15, off, s[0:3], s32 offset:404 ; 4-byte Folded Spill
-; GFX9-NEXT: v_lshrrev_b32_e32 v15, 16, v54
-; GFX9-NEXT: buffer_store_dword v15, off, s[0:3], s32 offset:336 ; 4-byte Folded Spill
-; GFX9-NEXT: v_lshrrev_b32_e32 v15, 24, v40
-; GFX9-NEXT: buffer_store_dword v15, off, s[0:3], s32 offset:344 ; 4-byte Folded Spill
-; GFX9-NEXT: v_lshrrev_b32_e32 v15, 16, v39
+; GFX9-NEXT: buffer_store_dword v51, off, s[0:3], s32 offset:168 ; 4-byte Folded Spill
+; GFX9-NEXT: buffer_load_dword v50, off, s[0:3], s32 offset:432 ; 4-byte Folded Reload
+; GFX9-NEXT: s_nop 0
+; GFX9-NEXT: buffer_load_dword v51, off, s[0:3], s32 offset:436 ; 4-byte Folded Reload
+; GFX9-NEXT: s_waitcnt vmcnt(0)
+; GFX9-NEXT: v_lshrrev_b64 v[50:51], 24, v[50:51]
+; GFX9-NEXT: buffer_store_dword v50, off, s[0:3], s32 offset:172 ; 4-byte Folded Spill
+; GFX9-NEXT: s_nop 0
+; GFX9-NEXT: buffer_store_dword v51, off, s[0:3], s32 offset:176 ; 4-byte Folded Spill
+; GFX9-NEXT: buffer_store_dword v16, off, s[0:3], s32 offset:268 ; 4-byte Folded Spill
+; GFX9-NEXT: v_lshrrev_b32_e32 v16, 8, v59
+; GFX9-NEXT: buffer_store_dword v16, off, s[0:3], s32 offset:356 ; 4-byte Folded Spill
+; GFX9-NEXT: v_lshrrev_b32_e32 v16, 16, v58
+; GFX9-NEXT: buffer_store_dword v16, off, s[0:3], s32 offset:276 ; 4-byte Folded Spill
+; GFX9-NEXT: v_lshrrev_b32_e32 v16, 24, v47
; GFX9-NEXT: buffer_store_dword v16, off, s[0:3], s32 offset:288 ; 4-byte Folded Spill
-; GFX9-NEXT: buffer_store_dword v15, off, s[0:3], s32 offset:356 ; 4-byte Folded Spill
-; GFX9-NEXT: v_lshrrev_b32_e32 v15, 24, v53
-; GFX9-NEXT: v_lshrrev_b32_e32 v39, 8, v53
-; GFX9-NEXT: v_lshrrev_b32_e32 v16, 24, v51
-; GFX9-NEXT: v_lshrrev_b32_e32 v42, 8, v51
-; GFX9-NEXT: v_lshrrev_b32_e32 v53, 24, v38
-; GFX9-NEXT: v_lshrrev_b32_e32 v51, 8, v38
-; GFX9-NEXT: v_lshrrev_b32_e32 v38, 16, v37
-; GFX9-NEXT: v_lshrrev_b32_e32 v41, 8, v37
-; GFX9-NEXT: v_lshrrev_b32_e32 v37, 24, v36
-; GFX9-NEXT: v_lshrrev_b32_e32 v36, 8, v36
-; GFX9-NEXT: buffer_store_dword v36, off, s[0:3], s32 offset:268 ; 4-byte Folded Spill
-; GFX9-NEXT: v_lshrrev_b32_e32 v36, 16, v35
-; GFX9-NEXT: v_lshrrev_b32_e32 v35, 8, v35
-; GFX9-NEXT: buffer_store_dword v35, off, s[0:3], s32 offset:276 ; 4-byte Folded Spill
-; GFX9-NEXT: v_lshrrev_b32_e32 v35, 24, v34
-; GFX9-NEXT: v_lshrrev_b32_e32 v34, 8, v34
-; GFX9-NEXT: buffer_store_dword v34, off, s[0:3], s32 offset:296 ; 4-byte Folded Spill
-; GFX9-NEXT: v_lshrrev_b32_e32 v34, 16, v33
-; GFX9-NEXT: v_lshrrev_b32_e32 v33, 8, v33
-; GFX9-NEXT: buffer_store_dword v33, off, s[0:3], s32 offset:308 ; 4-byte Folded Spill
-; GFX9-NEXT: v_lshrrev_b32_e32 v33, 24, v57
-; GFX9-NEXT: buffer_store_dword v33, off, s[0:3], s32 offset:316 ; 4-byte Folded Spill
-; GFX9-NEXT: v_lshrrev_b32_e32 v33, 8, v57
-; GFX9-NEXT: buffer_store_dword v33, off, s[0:3], s32 offset:320 ; 4-byte Folded Spill
-; GFX9-NEXT: v_lshrrev_b32_e32 v33, 16, v56
-; GFX9-NEXT: buffer_store_dword v33, off, s[0:3], s32 offset:328 ; 4-byte Folded Spill
-; GFX9-NEXT: v_lshrrev_b32_e32 v33, 8, v56
-; GFX9-NEXT: buffer_store_dword v33, off, s[0:3], s32 offset:332 ; 4-byte Folded Spill
-; GFX9-NEXT: v_lshrrev_b32_e32 v33, 24, v63
-; GFX9-NEXT: buffer_store_dword v33, off, s[0:3], s32 offset:340 ; 4-byte Folded Spill
-; GFX9-NEXT: v_lshrrev_b32_e32 v33, 8, v63
-; GFX9-NEXT: buffer_store_dword v33, off, s[0:3], s32 offset:348 ; 4-byte Folded Spill
-; GFX9-NEXT: v_lshrrev_b32_e32 v33, 16, v62
-; GFX9-NEXT: buffer_store_dword v16, off, s[0:3], s32 offset:388 ; 4-byte Folded Spill
-; GFX9-NEXT: v_lshrrev_b32_e32 v16, 16, v50
-; GFX9-NEXT: buffer_store_dword v33, off, s[0:3], s32 offset:352 ; 4-byte Folded Spill
-; GFX9-NEXT: v_lshrrev_b32_e32 v33, 8, v62
-; GFX9-NEXT: buffer_store_dword v15, off, s[0:3], s32 offset:372 ; 4-byte Folded Spill
-; GFX9-NEXT: v_lshrrev_b32_e32 v15, 16, v52
-; GFX9-NEXT: buffer_store_dword v16, off, s[0:3], s32 offset:392 ; 4-byte Folded Spill
-; GFX9-NEXT: v_lshrrev_b32_e32 v16, 8, v50
-; GFX9-NEXT: v_lshrrev_b32_e32 v50, 24, v49
-; GFX9-NEXT: buffer_store_dword v33, off, s[0:3], s32 offset:360 ; 4-byte Folded Spill
-; GFX9-NEXT: v_lshrrev_b32_e32 v33, 24, v59
-; GFX9-NEXT: buffer_store_dword v43, off, s[0:3], s32 offset:280 ; 4-byte Folded Spill
-; GFX9-NEXT: buffer_store_dword v15, off, s[0:3], s32 offset:380 ; 4-byte Folded Spill
-; GFX9-NEXT: buffer_store_dword v50, off, s[0:3], s32 offset:400 ; 4-byte Folded Spill
-; GFX9-NEXT: buffer_store_dword v37, off, s[0:3], s32 offset:264 ; 4-byte Folded Spill
-; GFX9-NEXT: buffer_store_dword v36, off, s[0:3], s32 offset:272 ; 4-byte Folded Spill
+; GFX9-NEXT: v_lshrrev_b32_e32 v16, 8, v47
+; GFX9-NEXT: buffer_store_dword v16, off, s[0:3], s32 offset:376 ; 4-byte Folded Spill
+; GFX9-NEXT: v_lshrrev_b32_e32 v16, 16, v46
+; GFX9-NEXT: buffer_store_dword v16, off, s[0:3], s32 offset:300 ; 4-byte Folded Spill
+; GFX9-NEXT: v_lshrrev_b32_e32 v16, 8, v46
+; GFX9-NEXT: buffer_store_dword v16, off, s[0:3], s32 offset:384 ; 4-byte Folded Spill
+; GFX9-NEXT: v_lshrrev_b32_e32 v16, 8, v45
+; GFX9-NEXT: buffer_store_dword v16, off, s[0:3], s32 offset:408 ; 4-byte Folded Spill
+; GFX9-NEXT: v_lshrrev_b32_e32 v16, 16, v44
+; GFX9-NEXT: buffer_store_dword v16, off, s[0:3], s32 offset:324 ; 4-byte Folded Spill
+; GFX9-NEXT: v_lshrrev_b32_e32 v16, 24, v41
+; GFX9-NEXT: buffer_store_dword v16, off, s[0:3], s32 offset:332 ; 4-byte Folded Spill
+; GFX9-NEXT: v_lshrrev_b32_e32 v16, 16, v40
+; GFX9-NEXT: buffer_store_dword v16, off, s[0:3], s32 offset:344 ; 4-byte Folded Spill
+; GFX9-NEXT: v_lshrrev_b32_e32 v16, 8, v40
+; GFX9-NEXT: buffer_store_dword v16, off, s[0:3], s32 offset:420 ; 4-byte Folded Spill
+; GFX9-NEXT: v_lshrrev_b32_e32 v16, 24, v57
+; GFX9-NEXT: buffer_store_dword v16, off, s[0:3], s32 offset:364 ; 4-byte Folded Spill
+; GFX9-NEXT: v_lshrrev_b32_e32 v16, 8, v57
+; GFX9-NEXT: buffer_store_dword v16, off, s[0:3], s32 offset:424 ; 4-byte Folded Spill
+; GFX9-NEXT: v_lshrrev_b32_e32 v16, 8, v56
+; GFX9-NEXT: buffer_store_dword v16, off, s[0:3], s32 offset:428 ; 4-byte Folded Spill
+; GFX9-NEXT: v_lshrrev_b32_e32 v16, 24, v43
+; GFX9-NEXT: v_lshrrev_b32_e32 v40, 8, v43
+; GFX9-NEXT: v_mov_b32_e32 v43, v38
+; GFX9-NEXT: buffer_store_dword v16, off, s[0:3], s32 offset:380 ; 4-byte Folded Spill
+; GFX9-NEXT: v_lshrrev_b32_e32 v16, 24, v39
+; GFX9-NEXT: v_lshrrev_b32_e32 v46, 8, v43
+; GFX9-NEXT: v_mov_b32_e32 v43, v48
+; GFX9-NEXT: buffer_store_dword v16, off, s[0:3], s32 offset:396 ; 4-byte Folded Spill
+; GFX9-NEXT: v_lshrrev_b32_e32 v16, 16, v38
+; GFX9-NEXT: v_lshrrev_b32_e32 v57, 8, v43
+; GFX9-NEXT: buffer_store_dword v35, off, s[0:3], s32 offset:312 ; 4-byte Folded Spill
+; GFX9-NEXT: v_lshrrev_b32_e32 v35, 16, v61
+; GFX9-NEXT: v_mov_b32_e32 v43, v61
+; GFX9-NEXT: buffer_store_dword v16, off, s[0:3], s32 offset:412 ; 4-byte Folded Spill
+; GFX9-NEXT: v_lshrrev_b32_e32 v16, 16, v48
+; GFX9-NEXT: buffer_store_dword v35, off, s[0:3], s32 offset:256 ; 4-byte Folded Spill
+; GFX9-NEXT: v_lshrrev_b32_e32 v35, 8, v43
+; GFX9-NEXT: buffer_store_dword v16, off, s[0:3], s32 offset:416 ; 4-byte Folded Spill
+; GFX9-NEXT: buffer_store_dword v35, off, s[0:3], s32 offset:260 ; 4-byte Folded Spill
+; GFX9-NEXT: buffer_load_dword v43, off, s[0:3], s32 offset:292 ; 4-byte Folded Reload
+; GFX9-NEXT: buffer_load_dword v44, off, s[0:3], s32 offset:296 ; 4-byte Folded Reload
+; GFX9-NEXT: v_lshrrev_b32_e32 v50, 8, v58
+; GFX9-NEXT: v_lshrrev_b32_e32 v58, 24, v45
+; GFX9-NEXT: v_lshrrev_b32_e32 v41, 8, v42
+; GFX9-NEXT: v_lshrrev_b32_e32 v42, 8, v39
+; GFX9-NEXT: v_lshrrev_b32_e32 v16, 24, v62
+; GFX9-NEXT: v_lshrrev_b32_e32 v47, 16, v56
+; GFX9-NEXT: v_lshrrev_b32_e32 v56, 24, v49
+; GFX9-NEXT: v_lshrrev_b32_e32 v59, 8, v49
+; GFX9-NEXT: v_mov_b32_e32 v54, v56
+; GFX9-NEXT: v_mov_b32_e32 v38, v58
+; GFX9-NEXT: s_waitcnt vmcnt(1)
+; GFX9-NEXT: v_lshrrev_b32_e32 v51, 8, v43
+; GFX9-NEXT: s_waitcnt vmcnt(0)
+; GFX9-NEXT: v_lshrrev_b32_e32 v35, 24, v44
+; GFX9-NEXT: buffer_store_dword v35, off, s[0:3], s32 offset:272 ; 4-byte Folded Spill
+; GFX9-NEXT: v_lshrrev_b32_e32 v35, 8, v44
; GFX9-NEXT: buffer_store_dword v35, off, s[0:3], s32 offset:284 ; 4-byte Folded Spill
-; GFX9-NEXT: buffer_store_dword v34, off, s[0:3], s32 offset:300 ; 4-byte Folded Spill
-; GFX9-NEXT: buffer_store_dword v33, off, s[0:3], s32 offset:368 ; 4-byte Folded Spill
-; GFX9-NEXT: v_lshrrev_b32_e32 v63, 8, v59
-; GFX9-NEXT: v_lshrrev_b32_e32 v37, 24, v61
-; GFX9-NEXT: v_lshrrev_b32_e32 v50, 8, v61
-; GFX9-NEXT: v_lshrrev_b32_e32 v56, 16, v60
-; GFX9-NEXT: v_mov_b32_e32 v33, v60
-; GFX9-NEXT: buffer_load_dword v61, off, s[0:3], s32 offset:416 ; 4-byte Folded Reload
-; GFX9-NEXT: buffer_load_dword v62, off, s[0:3], s32 offset:420 ; 4-byte Folded Reload
-; GFX9-NEXT: buffer_load_dword v59, off, s[0:3], s32 offset:408 ; 4-byte Folded Reload
-; GFX9-NEXT: buffer_load_dword v60, off, s[0:3], s32 offset:412 ; 4-byte Folded Reload
-; GFX9-NEXT: v_lshrrev_b32_e32 v54, 8, v54
-; GFX9-NEXT: v_lshrrev_b32_e32 v40, 8, v40
-; GFX9-NEXT: v_lshrrev_b32_e32 v15, 8, v52
-; GFX9-NEXT: v_lshrrev_b32_e32 v52, 8, v49
-; GFX9-NEXT: v_lshrrev_b32_e32 v49, 16, v48
-; GFX9-NEXT: v_lshrrev_b32_e32 v48, 8, v48
-; GFX9-NEXT: v_lshrrev_b32_e32 v36, 16, v58
-; GFX9-NEXT: v_lshrrev_b32_e32 v58, 8, v58
-; GFX9-NEXT: v_lshrrev_b32_e32 v55, 8, v33
-; GFX9-NEXT: s_waitcnt vmcnt(3)
-; GFX9-NEXT: v_lshrrev_b32_e32 v34, 16, v61
-; GFX9-NEXT: s_waitcnt vmcnt(2)
-; GFX9-NEXT: v_lshrrev_b32_e32 v33, 24, v62
-; GFX9-NEXT: v_lshrrev_b32_e32 v43, 8, v62
-; GFX9-NEXT: v_lshrrev_b32_e32 v44, 8, v61
-; GFX9-NEXT: s_waitcnt vmcnt(0)
-; GFX9-NEXT: v_lshrrev_b32_e32 v35, 24, v60
+; GFX9-NEXT: buffer_store_dword v51, off, s[0:3], s32 offset:292 ; 4-byte Folded Spill
+; GFX9-NEXT: v_lshrrev_b32_e32 v35, 16, v43
+; GFX9-NEXT: buffer_load_dword v43, off, s[0:3], s32 offset:316 ; 4-byte Folded Reload
+; GFX9-NEXT: buffer_load_dword v44, off, s[0:3], s32 offset:320 ; 4-byte Folded Reload
+; GFX9-NEXT: s_waitcnt vmcnt(0)
+; GFX9-NEXT: v_lshrrev_b32_e32 v51, 24, v44
+; GFX9-NEXT: buffer_store_dword v51, off, s[0:3], s32 offset:304 ; 4-byte Folded Spill
+; GFX9-NEXT: v_lshrrev_b32_e32 v51, 8, v44
+; GFX9-NEXT: buffer_store_dword v51, off, s[0:3], s32 offset:308 ; 4-byte Folded Spill
+; GFX9-NEXT: v_lshrrev_b32_e32 v51, 16, v43
+; GFX9-NEXT: v_lshrrev_b32_e32 v43, 8, v43
+; GFX9-NEXT: buffer_store_dword v43, off, s[0:3], s32 offset:316 ; 4-byte Folded Spill
+; GFX9-NEXT: buffer_load_dword v44, off, s[0:3], s32 offset:348 ; 4-byte Folded Reload
+; GFX9-NEXT: buffer_load_dword v45, off, s[0:3], s32 offset:352 ; 4-byte Folded Reload
+; GFX9-NEXT: s_waitcnt vmcnt(0)
+; GFX9-NEXT: v_lshrrev_b32_e32 v43, 24, v45
+; GFX9-NEXT: buffer_store_dword v43, off, s[0:3], s32 offset:328 ; 4-byte Folded Spill
+; GFX9-NEXT: v_lshrrev_b32_e32 v43, 8, v45
+; GFX9-NEXT: buffer_store_dword v43, off, s[0:3], s32 offset:336 ; 4-byte Folded Spill
+; GFX9-NEXT: v_lshrrev_b32_e32 v43, 16, v44
+; GFX9-NEXT: buffer_store_dword v43, off, s[0:3], s32 offset:340 ; 4-byte Folded Spill
+; GFX9-NEXT: v_lshrrev_b32_e32 v43, 8, v44
+; GFX9-NEXT: buffer_store_dword v43, off, s[0:3], s32 offset:348 ; 4-byte Folded Spill
+; GFX9-NEXT: buffer_load_dword v43, off, s[0:3], s32 offset:368 ; 4-byte Folded Reload
+; GFX9-NEXT: s_nop 0
+; GFX9-NEXT: buffer_load_dword v44, off, s[0:3], s32 offset:372 ; 4-byte Folded Reload
+; GFX9-NEXT: s_waitcnt vmcnt(1)
+; GFX9-NEXT: v_lshrrev_b32_e32 v39, 8, v43
+; GFX9-NEXT: s_waitcnt vmcnt(0)
+; GFX9-NEXT: v_lshrrev_b32_e32 v45, 24, v44
+; GFX9-NEXT: buffer_store_dword v45, off, s[0:3], s32 offset:360 ; 4-byte Folded Spill
+; GFX9-NEXT: buffer_store_dword v39, off, s[0:3], s32 offset:368 ; 4-byte Folded Spill
+; GFX9-NEXT: buffer_load_dword v62, off, s[0:3], s32 offset:400 ; 4-byte Folded Reload
+; GFX9-NEXT: buffer_load_dword v63, off, s[0:3], s32 offset:404 ; 4-byte Folded Reload
+; GFX9-NEXT: v_lshrrev_b32_e32 v45, 16, v43
+; GFX9-NEXT: v_lshrrev_b32_e32 v48, 8, v44
+; GFX9-NEXT: v_mov_b32_e32 v44, v35
+; GFX9-NEXT: v_mov_b32_e32 v35, v50
+; GFX9-NEXT: v_mov_b32_e32 v39, v51
+; GFX9-NEXT: v_lshrrev_b32_e32 v51, 16, v33
+; GFX9-NEXT: s_waitcnt vmcnt(0)
+; GFX9-NEXT: v_lshrrev_b32_e32 v49, 8, v63
+; GFX9-NEXT: buffer_store_dword v49, off, s[0:3], s32 offset:388 ; 4-byte Folded Spill
+; GFX9-NEXT: v_lshrrev_b32_e32 v49, 16, v62
+; GFX9-NEXT: buffer_store_dword v49, off, s[0:3], s32 offset:392 ; 4-byte Folded Spill
+; GFX9-NEXT: v_lshrrev_b32_e32 v49, 8, v62
+; GFX9-NEXT: buffer_store_dword v49, off, s[0:3], s32 offset:400 ; 4-byte Folded Spill
+; GFX9-NEXT: buffer_load_dword v60, off, s[0:3], s32 offset:432 ; 4-byte Folded Reload
+; GFX9-NEXT: buffer_load_dword v61, off, s[0:3], s32 offset:436 ; 4-byte Folded Reload
+; GFX9-NEXT: v_mov_b32_e32 v49, v33
+; GFX9-NEXT: v_lshrrev_b32_e32 v58, 24, v63
+; GFX9-NEXT: v_lshrrev_b32_e32 v63, 24, v34
+; GFX9-NEXT: v_lshrrev_b32_e32 v62, 8, v34
+; GFX9-NEXT: v_lshrrev_b32_e32 v50, 8, v49
+; GFX9-NEXT: v_mov_b32_e32 v49, v47
+; GFX9-NEXT: v_mov_b32_e32 v47, v45
+; GFX9-NEXT: s_waitcnt vmcnt(1)
+; GFX9-NEXT: v_lshrrev_b32_e32 v45, 16, v60
+; GFX9-NEXT: s_waitcnt vmcnt(0)
+; GFX9-NEXT: v_lshrrev_b32_e32 v43, 24, v61
+; GFX9-NEXT: v_lshrrev_b32_e32 v56, 8, v61
+; GFX9-NEXT: v_mov_b32_e32 v61, v16
; GFX9-NEXT: v_lshrrev_b32_e32 v60, 8, v60
-; GFX9-NEXT: v_lshrrev_b32_e32 v61, 16, v59
-; GFX9-NEXT: v_lshrrev_b32_e32 v62, 8, v59
; GFX9-NEXT: .LBB90_4: ; %end
; GFX9-NEXT: s_or_b64 exec, exec, s[4:5]
-; GFX9-NEXT: v_lshlrev_b16_e32 v15, 8, v15
-; GFX9-NEXT: v_or_b32_sdwa v7, v7, v15 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
-; GFX9-NEXT: v_lshlrev_b16_e32 v15, 8, v39
-; GFX9-NEXT: v_or_b32_sdwa v8, v8, v15 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
-; GFX9-NEXT: v_lshlrev_b16_e32 v15, 8, v45
-; GFX9-NEXT: v_or_b32_sdwa v9, v9, v15 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
-; GFX9-NEXT: v_lshlrev_b16_e32 v15, 8, v40
-; GFX9-NEXT: v_or_b32_sdwa v10, v10, v15 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
-; GFX9-NEXT: v_lshlrev_b16_e32 v15, 8, v54
-; GFX9-NEXT: v_or_b32_sdwa v11, v11, v15 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
-; GFX9-NEXT: buffer_load_dword v15, off, s[0:3], s32 offset:404 ; 4-byte Folded Reload
-; GFX9-NEXT: v_lshlrev_b16_e32 v16, 8, v16
-; GFX9-NEXT: v_lshlrev_b16_e32 v51, 8, v51
-; GFX9-NEXT: v_lshlrev_b16_e32 v48, 8, v48
-; GFX9-NEXT: v_or_b32_sdwa v5, v5, v16 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; GFX9-NEXT: v_lshlrev_b16_e32 v16, 8, v57
+; GFX9-NEXT: v_or_b32_sdwa v1, v1, v16 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; GFX9-NEXT: v_lshlrev_b16_e32 v16, 8, v59
+; GFX9-NEXT: v_or_b32_sdwa v2, v2, v16 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; GFX9-NEXT: v_lshlrev_b16_e32 v16, 8, v46
+; GFX9-NEXT: v_or_b32_sdwa v3, v3, v16 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
; GFX9-NEXT: v_lshlrev_b16_e32 v16, 8, v42
-; GFX9-NEXT: v_or_b32_sdwa v2, v2, v51 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
-; GFX9-NEXT: v_or_b32_sdwa v3, v3, v48 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
-; GFX9-NEXT: v_lshlrev_b16_e32 v48, 8, v52
+; GFX9-NEXT: v_or_b32_sdwa v4, v4, v16 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; GFX9-NEXT: v_lshlrev_b16_e32 v16, 8, v41
+; GFX9-NEXT: v_or_b32_sdwa v5, v5, v16 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; GFX9-NEXT: v_lshlrev_b16_e32 v16, 8, v40
; GFX9-NEXT: v_or_b32_sdwa v6, v6, v16 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
-; GFX9-NEXT: v_lshlrev_b16_e32 v39, 8, v62
-; GFX9-NEXT: v_or_b32_sdwa v17, v17, v39 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
-; GFX9-NEXT: v_lshlrev_b16_e32 v39, 8, v60
-; GFX9-NEXT: v_lshlrev_b16_e32 v41, 8, v41
-; GFX9-NEXT: v_or_b32_sdwa v18, v18, v39 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
-; GFX9-NEXT: v_or_b32_sdwa v1, v1, v41 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
-; GFX9-NEXT: v_or_b32_sdwa v4, v4, v48 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
-; GFX9-NEXT: buffer_load_dword v16, off, s[0:3], s32 offset:364 ; 4-byte Folded Reload
-; GFX9-NEXT: s_waitcnt vmcnt(1)
-; GFX9-NEXT: v_lshlrev_b16_e32 v15, 8, v15
-; GFX9-NEXT: v_or_b32_sdwa v12, v12, v15 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
-; GFX9-NEXT: buffer_load_dword v15, off, s[0:3], s32 offset:396 ; 4-byte Folded Reload
-; GFX9-NEXT: s_waitcnt vmcnt(1)
+; GFX9-NEXT: buffer_load_dword v16, off, s[0:3], s32 offset:428 ; 4-byte Folded Reload
+; GFX9-NEXT: s_waitcnt vmcnt(0)
; GFX9-NEXT: v_lshlrev_b16_e32 v16, 8, v16
-; GFX9-NEXT: v_or_b32_sdwa v16, v47, v16 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; GFX9-NEXT: v_or_b32_sdwa v7, v7, v16 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; GFX9-NEXT: buffer_load_dword v16, off, s[0:3], s32 offset:424 ; 4-byte Folded Reload
; GFX9-NEXT: s_waitcnt vmcnt(0)
-; GFX9-NEXT: v_lshlrev_b16_e32 v15, 8, v15
-; GFX9-NEXT: v_or_b32_sdwa v13, v13, v15 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
-; GFX9-NEXT: buffer_load_dword v15, off, s[0:3], s32 offset:384 ; 4-byte Folded Reload
+; GFX9-NEXT: v_lshlrev_b16_e32 v16, 8, v16
+; GFX9-NEXT: v_or_b32_sdwa v8, v8, v16 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; GFX9-NEXT: buffer_load_dword v16, off, s[0:3], s32 offset:420 ; 4-byte Folded Reload
; GFX9-NEXT: s_waitcnt vmcnt(0)
-; GFX9-NEXT: v_lshlrev_b16_e32 v15, 8, v15
-; GFX9-NEXT: v_or_b32_sdwa v14, v14, v15 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
-; GFX9-NEXT: buffer_load_dword v15, off, s[0:3], s32 offset:376 ; 4-byte Folded Reload
-; GFX9-NEXT: buffer_load_dword v51, off, s[0:3], s32 offset:196 ; 4-byte Folded Reload
-; GFX9-NEXT: buffer_load_dword v52, off, s[0:3], s32 offset:200 ; 4-byte Folded Reload
-; GFX9-NEXT: s_waitcnt vmcnt(1)
-; GFX9-NEXT: v_lshlrev_b16_e32 v39, 8, v51
-; GFX9-NEXT: v_or_b32_sdwa v38, v38, v39 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
-; GFX9-NEXT: v_or_b32_sdwa v1, v1, v38 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_0 src1_sel:DWORD
+; GFX9-NEXT: v_lshlrev_b16_e32 v16, 8, v16
+; GFX9-NEXT: v_or_b32_sdwa v9, v9, v16 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; GFX9-NEXT: v_lshlrev_b16_e32 v16, 8, v55
+; GFX9-NEXT: v_or_b32_sdwa v10, v10, v16 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; GFX9-NEXT: v_lshlrev_b16_e32 v16, 8, v53
+; GFX9-NEXT: v_or_b32_sdwa v11, v11, v16 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; GFX9-NEXT: buffer_load_dword v16, off, s[0:3], s32 offset:408 ; 4-byte Folded Reload
+; GFX9-NEXT: s_waitcnt vmcnt(0)
+; GFX9-NEXT: v_lshlrev_b16_e32 v16, 8, v16
+; GFX9-NEXT: v_or_b32_sdwa v12, v12, v16 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; GFX9-NEXT: buffer_load_dword v16, off, s[0:3], s32 offset:384 ; 4-byte Folded Reload
+; GFX9-NEXT: s_waitcnt vmcnt(0)
+; GFX9-NEXT: v_lshlrev_b16_e32 v16, 8, v16
+; GFX9-NEXT: v_or_b32_sdwa v13, v13, v16 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; GFX9-NEXT: buffer_load_dword v16, off, s[0:3], s32 offset:376 ; 4-byte Folded Reload
+; GFX9-NEXT: s_waitcnt vmcnt(0)
+; GFX9-NEXT: v_lshlrev_b16_e32 v16, 8, v16
+; GFX9-NEXT: v_or_b32_sdwa v14, v14, v16 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; GFX9-NEXT: v_lshlrev_b16_e32 v16, 8, v35
+; GFX9-NEXT: buffer_load_dword v35, off, s[0:3], s32 offset:356 ; 4-byte Folded Reload
+; GFX9-NEXT: buffer_load_dword v33, off, s[0:3], s32 offset:188 ; 4-byte Folded Reload
+; GFX9-NEXT: buffer_load_dword v34, off, s[0:3], s32 offset:192 ; 4-byte Folded Reload
+; GFX9-NEXT: v_or_b32_sdwa v16, v36, v16 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; GFX9-NEXT: v_lshlrev_b16_e32 v36, 8, v60
+; GFX9-NEXT: v_or_b32_sdwa v17, v17, v36 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; GFX9-NEXT: v_lshlrev_b16_e32 v36, 8, v56
+; GFX9-NEXT: v_or_b32_sdwa v18, v18, v36 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; GFX9-NEXT: s_waitcnt vmcnt(1)
+; GFX9-NEXT: v_lshlrev_b16_e32 v36, 8, v33
+; GFX9-NEXT: buffer_load_dword v33, off, s[0:3], s32 offset:416 ; 4-byte Folded Reload
+; GFX9-NEXT: v_lshlrev_b16_e32 v35, 8, v35
+; GFX9-NEXT: v_or_b32_sdwa v35, v37, v35 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; GFX9-NEXT: s_waitcnt vmcnt(0)
+; GFX9-NEXT: v_or_b32_sdwa v36, v33, v36 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; GFX9-NEXT: v_or_b32_sdwa v1, v1, v36 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_0 src1_sel:DWORD
; GFX9-NEXT: buffer_store_dword v1, v0, s[0:3], 0 offen
-; GFX9-NEXT: buffer_load_dword v38, off, s[0:3], s32 offset:256 ; 4-byte Folded Reload
-; GFX9-NEXT: v_lshlrev_b16_e32 v1, 8, v53
-; GFX9-NEXT: v_lshlrev_b16_e32 v15, 8, v15
-; GFX9-NEXT: v_or_b32_sdwa v15, v46, v15 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; GFX9-NEXT: buffer_load_dword v36, off, s[0:3], s32 offset:248 ; 4-byte Folded Reload
+; GFX9-NEXT: v_lshlrev_b16_e32 v1, 8, v54
; GFX9-NEXT: s_waitcnt vmcnt(0)
-; GFX9-NEXT: v_or_b32_sdwa v1, v38, v1 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; GFX9-NEXT: v_or_b32_sdwa v1, v36, v1 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
; GFX9-NEXT: v_or_b32_sdwa v1, v2, v1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_0 src1_sel:DWORD
; GFX9-NEXT: buffer_store_dword v1, v0, s[0:3], 0 offen offset:4
-; GFX9-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:188 ; 4-byte Folded Reload
+; GFX9-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:180 ; 4-byte Folded Reload
; GFX9-NEXT: s_nop 0
-; GFX9-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:192 ; 4-byte Folded Reload
-; GFX9-NEXT: s_waitcnt vmcnt(1)
+; GFX9-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:184 ; 4-byte Folded Reload
+; GFX9-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:412 ; 4-byte Folded Reload
+; GFX9-NEXT: s_waitcnt vmcnt(2)
; GFX9-NEXT: v_lshlrev_b16_e32 v1, 8, v1
-; GFX9-NEXT: v_or_b32_sdwa v1, v49, v1 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; GFX9-NEXT: s_waitcnt vmcnt(0)
+; GFX9-NEXT: v_or_b32_sdwa v1, v2, v1 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
; GFX9-NEXT: v_or_b32_sdwa v1, v3, v1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_0 src1_sel:DWORD
; GFX9-NEXT: buffer_store_dword v1, v0, s[0:3], 0 offen offset:8
-; GFX9-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:400 ; 4-byte Folded Reload
+; GFX9-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:396 ; 4-byte Folded Reload
; GFX9-NEXT: s_nop 0
-; GFX9-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:248 ; 4-byte Folded Reload
+; GFX9-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:240 ; 4-byte Folded Reload
; GFX9-NEXT: s_waitcnt vmcnt(1)
; GFX9-NEXT: v_lshlrev_b16_e32 v1, 8, v1
; GFX9-NEXT: s_waitcnt vmcnt(0)
; GFX9-NEXT: v_or_b32_sdwa v1, v2, v1 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
; GFX9-NEXT: v_or_b32_sdwa v1, v4, v1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_0 src1_sel:DWORD
; GFX9-NEXT: buffer_store_dword v1, v0, s[0:3], 0 offen offset:12
-; GFX9-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:172 ; 4-byte Folded Reload
-; GFX9-NEXT: s_nop 0
-; GFX9-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:176 ; 4-byte Folded Reload
-; GFX9-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:392 ; 4-byte Folded Reload
-; GFX9-NEXT: s_waitcnt vmcnt(2)
-; GFX9-NEXT: v_lshlrev_b16_e32 v1, 8, v1
-; GFX9-NEXT: s_waitcnt vmcnt(0)
-; GFX9-NEXT: v_or_b32_sdwa v1, v2, v1 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; GFX9-NEXT: v_lshlrev_b16_e32 v1, 8, v15
+; GFX9-NEXT: v_or_b32_sdwa v1, v52, v1 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
; GFX9-NEXT: v_or_b32_sdwa v1, v5, v1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_0 src1_sel:DWORD
; GFX9-NEXT: buffer_store_dword v1, v0, s[0:3], 0 offen offset:16
-; GFX9-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:388 ; 4-byte Folded Reload
+; GFX9-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:380 ; 4-byte Folded Reload
; GFX9-NEXT: s_nop 0
-; GFX9-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:240 ; 4-byte Folded Reload
+; GFX9-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:232 ; 4-byte Folded Reload
; GFX9-NEXT: s_waitcnt vmcnt(1)
; GFX9-NEXT: v_lshlrev_b16_e32 v1, 8, v1
; GFX9-NEXT: s_waitcnt vmcnt(0)
@@ -167428,16 +167513,14 @@ define <128 x i8> @bitcast_v64bf16_to_v128i8(<64 x bfloat> %a, i32 %b) {
; GFX9-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:156 ; 4-byte Folded Reload
; GFX9-NEXT: s_nop 0
; GFX9-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:160 ; 4-byte Folded Reload
-; GFX9-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:380 ; 4-byte Folded Reload
-; GFX9-NEXT: s_waitcnt vmcnt(2)
+; GFX9-NEXT: s_waitcnt vmcnt(1)
; GFX9-NEXT: v_lshlrev_b16_e32 v1, 8, v1
-; GFX9-NEXT: s_waitcnt vmcnt(0)
-; GFX9-NEXT: v_or_b32_sdwa v1, v2, v1 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; GFX9-NEXT: v_or_b32_sdwa v1, v49, v1 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
; GFX9-NEXT: v_or_b32_sdwa v1, v7, v1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_0 src1_sel:DWORD
; GFX9-NEXT: buffer_store_dword v1, v0, s[0:3], 0 offen offset:24
-; GFX9-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:372 ; 4-byte Folded Reload
+; GFX9-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:364 ; 4-byte Folded Reload
; GFX9-NEXT: s_nop 0
-; GFX9-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:232 ; 4-byte Folded Reload
+; GFX9-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:224 ; 4-byte Folded Reload
; GFX9-NEXT: s_waitcnt vmcnt(1)
; GFX9-NEXT: v_lshlrev_b16_e32 v1, 8, v1
; GFX9-NEXT: s_waitcnt vmcnt(0)
@@ -167447,16 +167530,16 @@ define <128 x i8> @bitcast_v64bf16_to_v128i8(<64 x bfloat> %a, i32 %b) {
; GFX9-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:148 ; 4-byte Folded Reload
; GFX9-NEXT: s_nop 0
; GFX9-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:152 ; 4-byte Folded Reload
-; GFX9-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:356 ; 4-byte Folded Reload
+; GFX9-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:344 ; 4-byte Folded Reload
; GFX9-NEXT: s_waitcnt vmcnt(2)
; GFX9-NEXT: v_lshlrev_b16_e32 v1, 8, v1
; GFX9-NEXT: s_waitcnt vmcnt(0)
; GFX9-NEXT: v_or_b32_sdwa v1, v2, v1 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
; GFX9-NEXT: v_or_b32_sdwa v1, v9, v1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_0 src1_sel:DWORD
; GFX9-NEXT: buffer_store_dword v1, v0, s[0:3], 0 offen offset:32
-; GFX9-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:344 ; 4-byte Folded Reload
+; GFX9-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:332 ; 4-byte Folded Reload
; GFX9-NEXT: s_nop 0
-; GFX9-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:228 ; 4-byte Folded Reload
+; GFX9-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:220 ; 4-byte Folded Reload
; GFX9-NEXT: s_waitcnt vmcnt(1)
; GFX9-NEXT: v_lshlrev_b16_e32 v1, 8, v1
; GFX9-NEXT: s_waitcnt vmcnt(0)
@@ -167466,18 +167549,15 @@ define <128 x i8> @bitcast_v64bf16_to_v128i8(<64 x bfloat> %a, i32 %b) {
; GFX9-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:140 ; 4-byte Folded Reload
; GFX9-NEXT: s_nop 0
; GFX9-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:144 ; 4-byte Folded Reload
-; GFX9-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:336 ; 4-byte Folded Reload
+; GFX9-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:324 ; 4-byte Folded Reload
; GFX9-NEXT: s_waitcnt vmcnt(2)
; GFX9-NEXT: v_lshlrev_b16_e32 v1, 8, v1
; GFX9-NEXT: s_waitcnt vmcnt(0)
; GFX9-NEXT: v_or_b32_sdwa v1, v2, v1 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
; GFX9-NEXT: v_or_b32_sdwa v1, v11, v1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_0 src1_sel:DWORD
; GFX9-NEXT: buffer_store_dword v1, v0, s[0:3], 0 offen offset:40
-; GFX9-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:324 ; 4-byte Folded Reload
-; GFX9-NEXT: s_nop 0
-; GFX9-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:220 ; 4-byte Folded Reload
-; GFX9-NEXT: s_waitcnt vmcnt(1)
-; GFX9-NEXT: v_lshlrev_b16_e32 v1, 8, v1
+; GFX9-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:212 ; 4-byte Folded Reload
+; GFX9-NEXT: v_lshlrev_b16_e32 v1, 8, v38
; GFX9-NEXT: s_waitcnt vmcnt(0)
; GFX9-NEXT: v_or_b32_sdwa v1, v2, v1 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
; GFX9-NEXT: v_or_b32_sdwa v1, v12, v1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_0 src1_sel:DWORD
@@ -167485,16 +167565,16 @@ define <128 x i8> @bitcast_v64bf16_to_v128i8(<64 x bfloat> %a, i32 %b) {
; GFX9-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:132 ; 4-byte Folded Reload
; GFX9-NEXT: s_nop 0
; GFX9-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:136 ; 4-byte Folded Reload
-; GFX9-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:312 ; 4-byte Folded Reload
+; GFX9-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:300 ; 4-byte Folded Reload
; GFX9-NEXT: s_waitcnt vmcnt(2)
; GFX9-NEXT: v_lshlrev_b16_e32 v1, 8, v1
; GFX9-NEXT: s_waitcnt vmcnt(0)
; GFX9-NEXT: v_or_b32_sdwa v1, v2, v1 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
; GFX9-NEXT: v_or_b32_sdwa v1, v13, v1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_0 src1_sel:DWORD
; GFX9-NEXT: buffer_store_dword v1, v0, s[0:3], 0 offen offset:48
-; GFX9-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:304 ; 4-byte Folded Reload
+; GFX9-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:288 ; 4-byte Folded Reload
; GFX9-NEXT: s_nop 0
-; GFX9-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:212 ; 4-byte Folded Reload
+; GFX9-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:204 ; 4-byte Folded Reload
; GFX9-NEXT: s_waitcnt vmcnt(1)
; GFX9-NEXT: v_lshlrev_b16_e32 v1, 8, v1
; GFX9-NEXT: s_waitcnt vmcnt(0)
@@ -167504,32 +167584,32 @@ define <128 x i8> @bitcast_v64bf16_to_v128i8(<64 x bfloat> %a, i32 %b) {
; GFX9-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:124 ; 4-byte Folded Reload
; GFX9-NEXT: s_nop 0
; GFX9-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:128 ; 4-byte Folded Reload
-; GFX9-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:288 ; 4-byte Folded Reload
+; GFX9-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:276 ; 4-byte Folded Reload
; GFX9-NEXT: s_waitcnt vmcnt(2)
; GFX9-NEXT: v_lshlrev_b16_e32 v1, 8, v1
; GFX9-NEXT: s_waitcnt vmcnt(0)
; GFX9-NEXT: v_or_b32_sdwa v1, v2, v1 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
-; GFX9-NEXT: v_or_b32_sdwa v1, v15, v1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_0 src1_sel:DWORD
+; GFX9-NEXT: v_or_b32_sdwa v1, v16, v1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_0 src1_sel:DWORD
; GFX9-NEXT: buffer_store_dword v1, v0, s[0:3], 0 offen offset:56
-; GFX9-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:280 ; 4-byte Folded Reload
+; GFX9-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:268 ; 4-byte Folded Reload
; GFX9-NEXT: s_nop 0
-; GFX9-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:204 ; 4-byte Folded Reload
+; GFX9-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:196 ; 4-byte Folded Reload
; GFX9-NEXT: s_waitcnt vmcnt(1)
; GFX9-NEXT: v_lshlrev_b16_e32 v1, 8, v1
; GFX9-NEXT: s_waitcnt vmcnt(0)
; GFX9-NEXT: v_or_b32_sdwa v1, v2, v1 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
-; GFX9-NEXT: v_or_b32_sdwa v1, v16, v1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_0 src1_sel:DWORD
+; GFX9-NEXT: v_or_b32_sdwa v1, v35, v1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_0 src1_sel:DWORD
; GFX9-NEXT: buffer_store_dword v1, v0, s[0:3], 0 offen offset:60
-; GFX9-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:180 ; 4-byte Folded Reload
+; GFX9-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:172 ; 4-byte Folded Reload
; GFX9-NEXT: s_nop 0
-; GFX9-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:184 ; 4-byte Folded Reload
+; GFX9-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:176 ; 4-byte Folded Reload
; GFX9-NEXT: s_waitcnt vmcnt(1)
; GFX9-NEXT: v_lshlrev_b16_e32 v1, 8, v1
-; GFX9-NEXT: v_or_b32_sdwa v1, v61, v1 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; GFX9-NEXT: v_or_b32_sdwa v1, v45, v1 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
; GFX9-NEXT: v_or_b32_sdwa v1, v17, v1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_0 src1_sel:DWORD
; GFX9-NEXT: buffer_store_dword v1, v0, s[0:3], 0 offen offset:64
-; GFX9-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:292 ; 4-byte Folded Reload
-; GFX9-NEXT: v_lshlrev_b16_e32 v1, 8, v35
+; GFX9-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:280 ; 4-byte Folded Reload
+; GFX9-NEXT: v_lshlrev_b16_e32 v1, 8, v43
; GFX9-NEXT: s_waitcnt vmcnt(0)
; GFX9-NEXT: v_or_b32_sdwa v1, v2, v1 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
; GFX9-NEXT: v_or_b32_sdwa v1, v18, v1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_0 src1_sel:DWORD
@@ -167538,49 +167618,57 @@ define <128 x i8> @bitcast_v64bf16_to_v128i8(<64 x bfloat> %a, i32 %b) {
; GFX9-NEXT: s_nop 0
; GFX9-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:168 ; 4-byte Folded Reload
; GFX9-NEXT: s_waitcnt vmcnt(0)
-; GFX9-NEXT: v_lshlrev_b16_e32 v2, 8, v44
+; GFX9-NEXT: v_lshlrev_b16_e32 v2, 8, v50
; GFX9-NEXT: v_or_b32_sdwa v2, v19, v2 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
; GFX9-NEXT: v_lshlrev_b16_e32 v1, 8, v1
-; GFX9-NEXT: v_or_b32_sdwa v1, v34, v1 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; GFX9-NEXT: v_or_b32_sdwa v1, v51, v1 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
; GFX9-NEXT: v_or_b32_sdwa v1, v2, v1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_0 src1_sel:DWORD
; GFX9-NEXT: buffer_store_dword v1, v0, s[0:3], 0 offen offset:72
-; GFX9-NEXT: buffer_load_dword v3, off, s[0:3], s32 offset:260 ; 4-byte Folded Reload
-; GFX9-NEXT: v_lshlrev_b16_e32 v1, 8, v43
-; GFX9-NEXT: v_lshlrev_b16_e32 v2, 8, v33
+; GFX9-NEXT: buffer_load_dword v3, off, s[0:3], s32 offset:252 ; 4-byte Folded Reload
+; GFX9-NEXT: v_lshlrev_b16_e32 v1, 8, v62
+; GFX9-NEXT: v_lshlrev_b16_e32 v2, 8, v63
; GFX9-NEXT: v_or_b32_sdwa v1, v20, v1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
; GFX9-NEXT: s_waitcnt vmcnt(0)
; GFX9-NEXT: v_or_b32_sdwa v2, v3, v2 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
; GFX9-NEXT: v_or_b32_sdwa v1, v1, v2 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_0 src1_sel:DWORD
; GFX9-NEXT: buffer_store_dword v1, v0, s[0:3], 0 offen offset:76
-; GFX9-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:116 ; 4-byte Folded Reload
-; GFX9-NEXT: buffer_load_dword v3, off, s[0:3], s32 offset:120 ; 4-byte Folded Reload
-; GFX9-NEXT: v_lshlrev_b16_e32 v1, 8, v55
-; GFX9-NEXT: v_or_b32_sdwa v1, v21, v1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
-; GFX9-NEXT: s_waitcnt vmcnt(1)
+; GFX9-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:400 ; 4-byte Folded Reload
+; GFX9-NEXT: s_nop 0
+; GFX9-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:108 ; 4-byte Folded Reload
+; GFX9-NEXT: buffer_load_dword v3, off, s[0:3], s32 offset:112 ; 4-byte Folded Reload
+; GFX9-NEXT: buffer_load_dword v3, off, s[0:3], s32 offset:392 ; 4-byte Folded Reload
+; GFX9-NEXT: s_waitcnt vmcnt(2)
; GFX9-NEXT: v_lshlrev_b16_e32 v2, 8, v2
-; GFX9-NEXT: v_or_b32_sdwa v2, v56, v2 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; GFX9-NEXT: v_lshlrev_b16_e32 v1, 8, v1
+; GFX9-NEXT: v_or_b32_sdwa v1, v21, v1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; GFX9-NEXT: s_waitcnt vmcnt(0)
+; GFX9-NEXT: v_or_b32_sdwa v2, v3, v2 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
; GFX9-NEXT: v_or_b32_sdwa v1, v1, v2 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_0 src1_sel:DWORD
; GFX9-NEXT: buffer_store_dword v1, v0, s[0:3], 0 offen offset:80
-; GFX9-NEXT: buffer_load_dword v3, off, s[0:3], s32 offset:252 ; 4-byte Folded Reload
-; GFX9-NEXT: v_lshlrev_b16_e32 v1, 8, v50
-; GFX9-NEXT: v_lshlrev_b16_e32 v2, 8, v37
+; GFX9-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:388 ; 4-byte Folded Reload
+; GFX9-NEXT: v_lshlrev_b16_e32 v2, 8, v58
+; GFX9-NEXT: buffer_load_dword v3, off, s[0:3], s32 offset:244 ; 4-byte Folded Reload
+; GFX9-NEXT: s_waitcnt vmcnt(1)
+; GFX9-NEXT: v_lshlrev_b16_e32 v1, 8, v1
; GFX9-NEXT: v_or_b32_sdwa v1, v22, v1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
; GFX9-NEXT: s_waitcnt vmcnt(0)
; GFX9-NEXT: v_or_b32_sdwa v2, v3, v2 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
; GFX9-NEXT: v_or_b32_sdwa v1, v1, v2 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_0 src1_sel:DWORD
; GFX9-NEXT: buffer_store_dword v1, v0, s[0:3], 0 offen offset:84
-; GFX9-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:108 ; 4-byte Folded Reload
-; GFX9-NEXT: buffer_load_dword v3, off, s[0:3], s32 offset:112 ; 4-byte Folded Reload
-; GFX9-NEXT: v_lshlrev_b16_e32 v1, 8, v58
-; GFX9-NEXT: v_or_b32_sdwa v1, v23, v1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; GFX9-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:368 ; 4-byte Folded Reload
+; GFX9-NEXT: s_nop 0
+; GFX9-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:100 ; 4-byte Folded Reload
+; GFX9-NEXT: buffer_load_dword v3, off, s[0:3], s32 offset:104 ; 4-byte Folded Reload
; GFX9-NEXT: s_waitcnt vmcnt(1)
; GFX9-NEXT: v_lshlrev_b16_e32 v2, 8, v2
-; GFX9-NEXT: v_or_b32_sdwa v2, v36, v2 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; GFX9-NEXT: v_lshlrev_b16_e32 v1, 8, v1
+; GFX9-NEXT: v_or_b32_sdwa v1, v23, v1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; GFX9-NEXT: v_or_b32_sdwa v2, v47, v2 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
; GFX9-NEXT: v_or_b32_sdwa v1, v1, v2 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_0 src1_sel:DWORD
; GFX9-NEXT: buffer_store_dword v1, v0, s[0:3], 0 offen offset:88
-; GFX9-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:368 ; 4-byte Folded Reload
-; GFX9-NEXT: buffer_load_dword v3, off, s[0:3], s32 offset:244 ; 4-byte Folded Reload
-; GFX9-NEXT: v_lshlrev_b16_e32 v1, 8, v63
+; GFX9-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:360 ; 4-byte Folded Reload
+; GFX9-NEXT: buffer_load_dword v3, off, s[0:3], s32 offset:236 ; 4-byte Folded Reload
+; GFX9-NEXT: v_lshlrev_b16_e32 v1, 8, v48
; GFX9-NEXT: v_or_b32_sdwa v1, v24, v1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
; GFX9-NEXT: s_waitcnt vmcnt(1)
; GFX9-NEXT: v_lshlrev_b16_e32 v2, 8, v2
@@ -167588,11 +167676,11 @@ define <128 x i8> @bitcast_v64bf16_to_v128i8(<64 x bfloat> %a, i32 %b) {
; GFX9-NEXT: v_or_b32_sdwa v2, v3, v2 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
; GFX9-NEXT: v_or_b32_sdwa v1, v1, v2 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_0 src1_sel:DWORD
; GFX9-NEXT: buffer_store_dword v1, v0, s[0:3], 0 offen offset:92
-; GFX9-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:360 ; 4-byte Folded Reload
+; GFX9-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:348 ; 4-byte Folded Reload
; GFX9-NEXT: s_nop 0
-; GFX9-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:100 ; 4-byte Folded Reload
-; GFX9-NEXT: buffer_load_dword v3, off, s[0:3], s32 offset:104 ; 4-byte Folded Reload
-; GFX9-NEXT: buffer_load_dword v3, off, s[0:3], s32 offset:352 ; 4-byte Folded Reload
+; GFX9-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:92 ; 4-byte Folded Reload
+; GFX9-NEXT: buffer_load_dword v3, off, s[0:3], s32 offset:96 ; 4-byte Folded Reload
+; GFX9-NEXT: buffer_load_dword v3, off, s[0:3], s32 offset:340 ; 4-byte Folded Reload
; GFX9-NEXT: s_waitcnt vmcnt(2)
; GFX9-NEXT: v_lshlrev_b16_e32 v2, 8, v2
; GFX9-NEXT: v_lshlrev_b16_e32 v1, 8, v1
@@ -167601,10 +167689,10 @@ define <128 x i8> @bitcast_v64bf16_to_v128i8(<64 x bfloat> %a, i32 %b) {
; GFX9-NEXT: v_or_b32_sdwa v2, v3, v2 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
; GFX9-NEXT: v_or_b32_sdwa v1, v1, v2 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_0 src1_sel:DWORD
; GFX9-NEXT: buffer_store_dword v1, v0, s[0:3], 0 offen offset:96
-; GFX9-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:348 ; 4-byte Folded Reload
+; GFX9-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:336 ; 4-byte Folded Reload
; GFX9-NEXT: s_nop 0
-; GFX9-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:340 ; 4-byte Folded Reload
-; GFX9-NEXT: buffer_load_dword v3, off, s[0:3], s32 offset:236 ; 4-byte Folded Reload
+; GFX9-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:328 ; 4-byte Folded Reload
+; GFX9-NEXT: buffer_load_dword v3, off, s[0:3], s32 offset:228 ; 4-byte Folded Reload
; GFX9-NEXT: s_waitcnt vmcnt(2)
; GFX9-NEXT: v_lshlrev_b16_e32 v1, 8, v1
; GFX9-NEXT: v_or_b32_sdwa v1, v26, v1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
@@ -167614,23 +167702,21 @@ define <128 x i8> @bitcast_v64bf16_to_v128i8(<64 x bfloat> %a, i32 %b) {
; GFX9-NEXT: v_or_b32_sdwa v2, v3, v2 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
; GFX9-NEXT: v_or_b32_sdwa v1, v1, v2 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_0 src1_sel:DWORD
; GFX9-NEXT: buffer_store_dword v1, v0, s[0:3], 0 offen offset:100
-; GFX9-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:332 ; 4-byte Folded Reload
+; GFX9-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:316 ; 4-byte Folded Reload
; GFX9-NEXT: s_nop 0
-; GFX9-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:92 ; 4-byte Folded Reload
-; GFX9-NEXT: buffer_load_dword v3, off, s[0:3], s32 offset:96 ; 4-byte Folded Reload
-; GFX9-NEXT: buffer_load_dword v3, off, s[0:3], s32 offset:328 ; 4-byte Folded Reload
-; GFX9-NEXT: s_waitcnt vmcnt(2)
+; GFX9-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:116 ; 4-byte Folded Reload
+; GFX9-NEXT: buffer_load_dword v3, off, s[0:3], s32 offset:120 ; 4-byte Folded Reload
+; GFX9-NEXT: s_waitcnt vmcnt(1)
; GFX9-NEXT: v_lshlrev_b16_e32 v2, 8, v2
; GFX9-NEXT: v_lshlrev_b16_e32 v1, 8, v1
; GFX9-NEXT: v_or_b32_sdwa v1, v27, v1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
-; GFX9-NEXT: s_waitcnt vmcnt(0)
-; GFX9-NEXT: v_or_b32_sdwa v2, v3, v2 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; GFX9-NEXT: v_or_b32_sdwa v2, v39, v2 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
; GFX9-NEXT: v_or_b32_sdwa v1, v1, v2 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_0 src1_sel:DWORD
; GFX9-NEXT: buffer_store_dword v1, v0, s[0:3], 0 offen offset:104
-; GFX9-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:320 ; 4-byte Folded Reload
+; GFX9-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:308 ; 4-byte Folded Reload
; GFX9-NEXT: s_nop 0
-; GFX9-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:316 ; 4-byte Folded Reload
-; GFX9-NEXT: buffer_load_dword v3, off, s[0:3], s32 offset:224 ; 4-byte Folded Reload
+; GFX9-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:304 ; 4-byte Folded Reload
+; GFX9-NEXT: buffer_load_dword v3, off, s[0:3], s32 offset:216 ; 4-byte Folded Reload
; GFX9-NEXT: s_waitcnt vmcnt(2)
; GFX9-NEXT: v_lshlrev_b16_e32 v1, 8, v1
; GFX9-NEXT: v_or_b32_sdwa v1, v28, v1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
@@ -167640,23 +167726,21 @@ define <128 x i8> @bitcast_v64bf16_to_v128i8(<64 x bfloat> %a, i32 %b) {
; GFX9-NEXT: v_or_b32_sdwa v2, v3, v2 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
; GFX9-NEXT: v_or_b32_sdwa v1, v1, v2 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_0 src1_sel:DWORD
; GFX9-NEXT: buffer_store_dword v1, v0, s[0:3], 0 offen offset:108
-; GFX9-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:308 ; 4-byte Folded Reload
+; GFX9-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:292 ; 4-byte Folded Reload
; GFX9-NEXT: s_nop 0
; GFX9-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:84 ; 4-byte Folded Reload
; GFX9-NEXT: buffer_load_dword v3, off, s[0:3], s32 offset:88 ; 4-byte Folded Reload
-; GFX9-NEXT: buffer_load_dword v3, off, s[0:3], s32 offset:300 ; 4-byte Folded Reload
-; GFX9-NEXT: s_waitcnt vmcnt(2)
+; GFX9-NEXT: s_waitcnt vmcnt(1)
; GFX9-NEXT: v_lshlrev_b16_e32 v2, 8, v2
; GFX9-NEXT: v_lshlrev_b16_e32 v1, 8, v1
; GFX9-NEXT: v_or_b32_sdwa v1, v29, v1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
-; GFX9-NEXT: s_waitcnt vmcnt(0)
-; GFX9-NEXT: v_or_b32_sdwa v2, v3, v2 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; GFX9-NEXT: v_or_b32_sdwa v2, v44, v2 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
; GFX9-NEXT: v_or_b32_sdwa v1, v1, v2 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_0 src1_sel:DWORD
; GFX9-NEXT: buffer_store_dword v1, v0, s[0:3], 0 offen offset:112
-; GFX9-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:296 ; 4-byte Folded Reload
+; GFX9-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:284 ; 4-byte Folded Reload
; GFX9-NEXT: s_nop 0
-; GFX9-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:284 ; 4-byte Folded Reload
-; GFX9-NEXT: buffer_load_dword v3, off, s[0:3], s32 offset:216 ; 4-byte Folded Reload
+; GFX9-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:272 ; 4-byte Folded Reload
+; GFX9-NEXT: buffer_load_dword v3, off, s[0:3], s32 offset:208 ; 4-byte Folded Reload
; GFX9-NEXT: s_waitcnt vmcnt(2)
; GFX9-NEXT: v_lshlrev_b16_e32 v1, 8, v1
; GFX9-NEXT: v_or_b32_sdwa v1, v30, v1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
@@ -167666,11 +167750,11 @@ define <128 x i8> @bitcast_v64bf16_to_v128i8(<64 x bfloat> %a, i32 %b) {
; GFX9-NEXT: v_or_b32_sdwa v2, v3, v2 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
; GFX9-NEXT: v_or_b32_sdwa v1, v1, v2 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_0 src1_sel:DWORD
; GFX9-NEXT: buffer_store_dword v1, v0, s[0:3], 0 offen offset:116
-; GFX9-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:276 ; 4-byte Folded Reload
+; GFX9-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:260 ; 4-byte Folded Reload
; GFX9-NEXT: s_nop 0
; GFX9-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:76 ; 4-byte Folded Reload
; GFX9-NEXT: buffer_load_dword v3, off, s[0:3], s32 offset:80 ; 4-byte Folded Reload
-; GFX9-NEXT: buffer_load_dword v3, off, s[0:3], s32 offset:272 ; 4-byte Folded Reload
+; GFX9-NEXT: buffer_load_dword v3, off, s[0:3], s32 offset:256 ; 4-byte Folded Reload
; GFX9-NEXT: s_waitcnt vmcnt(2)
; GFX9-NEXT: v_lshlrev_b16_e32 v2, 8, v2
; GFX9-NEXT: v_lshlrev_b16_e32 v1, 8, v1
@@ -167679,15 +167763,12 @@ define <128 x i8> @bitcast_v64bf16_to_v128i8(<64 x bfloat> %a, i32 %b) {
; GFX9-NEXT: v_or_b32_sdwa v2, v3, v2 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
; GFX9-NEXT: v_or_b32_sdwa v1, v1, v2 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_0 src1_sel:DWORD
; GFX9-NEXT: buffer_store_dword v1, v0, s[0:3], 0 offen offset:120
-; GFX9-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:268 ; 4-byte Folded Reload
-; GFX9-NEXT: s_nop 0
-; GFX9-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:264 ; 4-byte Folded Reload
-; GFX9-NEXT: buffer_load_dword v3, off, s[0:3], s32 offset:208 ; 4-byte Folded Reload
-; GFX9-NEXT: s_waitcnt vmcnt(2)
+; GFX9-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:312 ; 4-byte Folded Reload
+; GFX9-NEXT: v_lshlrev_b16_e32 v2, 8, v61
+; GFX9-NEXT: buffer_load_dword v3, off, s[0:3], s32 offset:200 ; 4-byte Folded Reload
+; GFX9-NEXT: s_waitcnt vmcnt(1)
; GFX9-NEXT: v_lshlrev_b16_e32 v1, 8, v1
; GFX9-NEXT: v_or_b32_sdwa v1, v32, v1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
-; GFX9-NEXT: s_waitcnt vmcnt(1)
-; GFX9-NEXT: v_lshlrev_b16_e32 v2, 8, v2
; GFX9-NEXT: s_waitcnt vmcnt(0)
; GFX9-NEXT: v_or_b32_sdwa v2, v3, v2 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
; GFX9-NEXT: v_or_b32_sdwa v1, v1, v2 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_0 src1_sel:DWORD
@@ -167757,117 +167838,117 @@ define <128 x i8> @bitcast_v64bf16_to_v128i8(<64 x bfloat> %a, i32 %b) {
; GFX11-TRUE16-NEXT: scratch_load_b32 v33, off, s32 offset:8
; GFX11-TRUE16-NEXT: scratch_load_b32 v32, off, s32 offset:4
; GFX11-TRUE16-NEXT: scratch_load_b32 v31, off, s32
-; GFX11-TRUE16-NEXT: ; implicit-def: $vgpr144_hi16
+; GFX11-TRUE16-NEXT: ; implicit-def: $vgpr148_hi16
; GFX11-TRUE16-NEXT: ; implicit-def: $vgpr108_lo16
; GFX11-TRUE16-NEXT: ; implicit-def: $vgpr68_lo16
-; GFX11-TRUE16-NEXT: ; implicit-def: $vgpr131_hi16
-; GFX11-TRUE16-NEXT: ; implicit-def: $vgpr107_lo16
; GFX11-TRUE16-NEXT: ; implicit-def: $vgpr132_hi16
+; GFX11-TRUE16-NEXT: ; implicit-def: $vgpr107_lo16
+; GFX11-TRUE16-NEXT: ; implicit-def: $vgpr117_hi16
; GFX11-TRUE16-NEXT: ; implicit-def: $vgpr106_lo16
-; GFX11-TRUE16-NEXT: ; implicit-def: $vgpr162_hi16
+; GFX11-TRUE16-NEXT: ; implicit-def: $vgpr163_hi16
; GFX11-TRUE16-NEXT: ; implicit-def: $vgpr105_lo16
; GFX11-TRUE16-NEXT: ; implicit-def: $vgpr67_lo16
; GFX11-TRUE16-NEXT: ; implicit-def: $vgpr147_hi16
+; GFX11-TRUE16-NEXT: ; implicit-def: $vgpr104_lo16
+; GFX11-TRUE16-NEXT: ; implicit-def: $vgpr128_hi16
; GFX11-TRUE16-NEXT: ; implicit-def: $vgpr95_lo16
-; GFX11-TRUE16-NEXT: ; implicit-def: $vgpr149_hi16
+; GFX11-TRUE16-NEXT: ; implicit-def: $vgpr179_hi16
+; GFX11-TRUE16-NEXT: ; implicit-def: $vgpr94_lo16
+; GFX11-TRUE16-NEXT: ; implicit-def: $vgpr66_lo16
+; GFX11-TRUE16-NEXT: ; implicit-def: $vgpr160_hi16
; GFX11-TRUE16-NEXT: ; implicit-def: $vgpr93_lo16
-; GFX11-TRUE16-NEXT: ; implicit-def: $vgpr180_hi16
+; GFX11-TRUE16-NEXT: ; implicit-def: $vgpr131_hi16
+; GFX11-TRUE16-NEXT: ; implicit-def: $vgpr91_lo16
+; GFX11-TRUE16-NEXT: ; implicit-def: $vgpr42_hi16
; GFX11-TRUE16-NEXT: ; implicit-def: $vgpr90_lo16
-; GFX11-TRUE16-NEXT: ; implicit-def: $vgpr66_lo16
-; GFX11-TRUE16-NEXT: ; implicit-def: $vgpr164_hi16
-; GFX11-TRUE16-NEXT: ; implicit-def: $vgpr88_lo16
-; GFX11-TRUE16-NEXT: ; implicit-def: $vgpr165_hi16
-; GFX11-TRUE16-NEXT: ; implicit-def: $vgpr78_lo16
-; GFX11-TRUE16-NEXT: ; implicit-def: $vgpr46_hi16
-; GFX11-TRUE16-NEXT: ; implicit-def: $vgpr75_lo16
; GFX11-TRUE16-NEXT: ; implicit-def: $vgpr65_lo16
-; GFX11-TRUE16-NEXT: ; implicit-def: $vgpr178_hi16
+; GFX11-TRUE16-NEXT: ; implicit-def: $vgpr167_hi16
+; GFX11-TRUE16-NEXT: ; implicit-def: $vgpr77_lo16
+; GFX11-TRUE16-NEXT: ; implicit-def: $vgpr134_hi16
+; GFX11-TRUE16-NEXT: ; implicit-def: $vgpr76_lo16
+; GFX11-TRUE16-NEXT: ; implicit-def: $vgpr60_hi16
; GFX11-TRUE16-NEXT: ; implicit-def: $vgpr63_lo16
-; GFX11-TRUE16-NEXT: ; implicit-def: $vgpr179_hi16
+; GFX11-TRUE16-NEXT: ; implicit-def: $vgpr64_lo16
+; GFX11-TRUE16-NEXT: ; implicit-def: $vgpr182_hi16
; GFX11-TRUE16-NEXT: ; implicit-def: $vgpr62_lo16
-; GFX11-TRUE16-NEXT: ; implicit-def: $vgpr74_hi16
+; GFX11-TRUE16-NEXT: ; implicit-def: $vgpr145_hi16
; GFX11-TRUE16-NEXT: ; implicit-def: $vgpr57_lo16
-; GFX11-TRUE16-NEXT: ; implicit-def: $vgpr64_lo16
-; GFX11-TRUE16-NEXT: ; implicit-def: $vgpr45_hi16
-; GFX11-TRUE16-NEXT: ; implicit-def: $vgpr56_lo16
-; GFX11-TRUE16-NEXT: ; implicit-def: $vgpr42_hi16
-; GFX11-TRUE16-NEXT: ; implicit-def: $vgpr43_lo16
-; GFX11-TRUE16-NEXT: ; implicit-def: $vgpr89_hi16
-; GFX11-TRUE16-NEXT: ; implicit-def: $vgpr40_lo16
+; GFX11-TRUE16-NEXT: ; implicit-def: $vgpr72_hi16
+; GFX11-TRUE16-NEXT: ; implicit-def: $vgpr47_lo16
; GFX11-TRUE16-NEXT: ; implicit-def: $vgpr52_lo16
-; GFX11-TRUE16-NEXT: ; implicit-def: $vgpr59_hi16
-; GFX11-TRUE16-NEXT: ; implicit-def: $vgpr182_lo16
-; GFX11-TRUE16-NEXT: ; implicit-def: $vgpr60_hi16
+; GFX11-TRUE16-NEXT: ; implicit-def: $vgpr43_hi16
+; GFX11-TRUE16-NEXT: ; implicit-def: $vgpr45_lo16
+; GFX11-TRUE16-NEXT: ; implicit-def: $vgpr150_hi16
+; GFX11-TRUE16-NEXT: ; implicit-def: $vgpr44_lo16
+; GFX11-TRUE16-NEXT: ; implicit-def: $vgpr79_hi16
; GFX11-TRUE16-NEXT: ; implicit-def: $vgpr181_lo16
-; GFX11-TRUE16-NEXT: ; implicit-def: $vgpr94_hi16
-; GFX11-TRUE16-NEXT: ; implicit-def: $vgpr167_lo16
; GFX11-TRUE16-NEXT: ; implicit-def: $vgpr49_lo16
-; GFX11-TRUE16-NEXT: ; implicit-def: $vgpr77_hi16
-; GFX11-TRUE16-NEXT: ; implicit-def: $vgpr163_lo16
-; GFX11-TRUE16-NEXT: ; implicit-def: $vgpr76_hi16
-; GFX11-TRUE16-NEXT: ; implicit-def: $vgpr160_lo16
-; GFX11-TRUE16-NEXT: ; implicit-def: $vgpr104_hi16
-; GFX11-TRUE16-NEXT: ; implicit-def: $vgpr151_lo16
+; GFX11-TRUE16-NEXT: ; implicit-def: $vgpr58_hi16
+; GFX11-TRUE16-NEXT: ; implicit-def: $vgpr178_lo16
+; GFX11-TRUE16-NEXT: ; implicit-def: $vgpr165_hi16
+; GFX11-TRUE16-NEXT: ; implicit-def: $vgpr176_lo16
+; GFX11-TRUE16-NEXT: ; implicit-def: $vgpr88_hi16
+; GFX11-TRUE16-NEXT: ; implicit-def: $vgpr166_lo16
; GFX11-TRUE16-NEXT: ; implicit-def: $vgpr48_lo16
-; GFX11-TRUE16-NEXT: ; implicit-def: $vgpr91_hi16
-; GFX11-TRUE16-NEXT: ; implicit-def: $vgpr146_lo16
-; GFX11-TRUE16-NEXT: ; implicit-def: $vgpr92_hi16
-; GFX11-TRUE16-NEXT: ; implicit-def: $vgpr135_lo16
-; GFX11-TRUE16-NEXT: ; implicit-def: $vgpr80_hi16
-; GFX11-TRUE16-NEXT: ; implicit-def: $vgpr79_lo16
+; GFX11-TRUE16-NEXT: ; implicit-def: $vgpr74_hi16
+; GFX11-TRUE16-NEXT: ; implicit-def: $vgpr161_lo16
+; GFX11-TRUE16-NEXT: ; implicit-def: $vgpr183_hi16
+; GFX11-TRUE16-NEXT: ; implicit-def: $vgpr149_lo16
+; GFX11-TRUE16-NEXT: ; implicit-def: $vgpr83_hi16
+; GFX11-TRUE16-NEXT: ; implicit-def: $vgpr92_lo16
; GFX11-TRUE16-NEXT: ; implicit-def: $vgpr53_lo16
+; GFX11-TRUE16-NEXT: ; implicit-def: $vgpr71_hi16
+; GFX11-TRUE16-NEXT: ; implicit-def: $vgpr89_lo16
; GFX11-TRUE16-NEXT: ; implicit-def: $vgpr55_hi16
+; GFX11-TRUE16-NEXT: ; implicit-def: $vgpr78_lo16
+; GFX11-TRUE16-NEXT: ; implicit-def: $vgpr86_hi16
+; GFX11-TRUE16-NEXT: ; implicit-def: $vgpr75_lo16
+; GFX11-TRUE16-NEXT: ; implicit-def: $vgpr50_lo16
+; GFX11-TRUE16-NEXT: ; implicit-def: $vgpr81_hi16
; GFX11-TRUE16-NEXT: ; implicit-def: $vgpr73_lo16
; GFX11-TRUE16-NEXT: ; implicit-def: $vgpr70_hi16
-; GFX11-TRUE16-NEXT: ; implicit-def: $vgpr72_lo16
-; GFX11-TRUE16-NEXT: ; implicit-def: $vgpr83_hi16
; GFX11-TRUE16-NEXT: ; implicit-def: $vgpr61_lo16
-; GFX11-TRUE16-NEXT: ; implicit-def: $vgpr50_lo16
-; GFX11-TRUE16-NEXT: ; implicit-def: $vgpr81_hi16
-; GFX11-TRUE16-NEXT: ; implicit-def: $vgpr58_lo16
-; GFX11-TRUE16-NEXT: ; implicit-def: $vgpr71_hi16
-; GFX11-TRUE16-NEXT: ; implicit-def: $vgpr47_lo16
-; GFX11-TRUE16-NEXT: ; implicit-def: $vgpr85_hi16
-; GFX11-TRUE16-NEXT: ; implicit-def: $vgpr44_lo16
+; GFX11-TRUE16-NEXT: ; implicit-def: $vgpr97_hi16
+; GFX11-TRUE16-NEXT: ; implicit-def: $vgpr59_lo16
; GFX11-TRUE16-NEXT: ; implicit-def: $vgpr38_lo16
-; GFX11-TRUE16-NEXT: ; implicit-def: $vgpr82_hi16
-; GFX11-TRUE16-NEXT: ; implicit-def: $vgpr41_lo16
; GFX11-TRUE16-NEXT: ; implicit-def: $vgpr84_hi16
-; GFX11-TRUE16-NEXT: ; implicit-def: $vgpr183_lo16
+; GFX11-TRUE16-NEXT: ; implicit-def: $vgpr56_lo16
+; GFX11-TRUE16-NEXT: ; implicit-def: $vgpr80_hi16
+; GFX11-TRUE16-NEXT: ; implicit-def: $vgpr46_lo16
+; GFX11-TRUE16-NEXT: ; implicit-def: $vgpr100_hi16
+; GFX11-TRUE16-NEXT: ; implicit-def: $vgpr41_lo16
+; GFX11-TRUE16-NEXT: ; implicit-def: $vgpr37_lo16
; GFX11-TRUE16-NEXT: ; implicit-def: $vgpr96_hi16
+; GFX11-TRUE16-NEXT: ; implicit-def: $vgpr40_lo16
+; GFX11-TRUE16-NEXT: ; implicit-def: $vgpr82_hi16
+; GFX11-TRUE16-NEXT: ; implicit-def: $vgpr180_lo16
+; GFX11-TRUE16-NEXT: ; implicit-def: $vgpr103_hi16
; GFX11-TRUE16-NEXT: ; implicit-def: $vgpr177_lo16
-; GFX11-TRUE16-NEXT: ; implicit-def: $vgpr37_lo16
-; GFX11-TRUE16-NEXT: ; implicit-def: $vgpr86_hi16
-; GFX11-TRUE16-NEXT: ; implicit-def: $vgpr176_lo16
-; GFX11-TRUE16-NEXT: ; implicit-def: $vgpr87_hi16
-; GFX11-TRUE16-NEXT: ; implicit-def: $vgpr166_lo16
-; GFX11-TRUE16-NEXT: ; implicit-def: $vgpr101_hi16
-; GFX11-TRUE16-NEXT: ; implicit-def: $vgpr161_lo16
; GFX11-TRUE16-NEXT: ; implicit-def: $vgpr36_lo16
-; GFX11-TRUE16-NEXT: ; implicit-def: $vgpr97_hi16
-; GFX11-TRUE16-NEXT: ; implicit-def: $vgpr150_lo16
-; GFX11-TRUE16-NEXT: ; implicit-def: $vgpr98_hi16
-; GFX11-TRUE16-NEXT: ; implicit-def: $vgpr148_lo16
-; GFX11-TRUE16-NEXT: ; implicit-def: $vgpr112_hi16
-; GFX11-TRUE16-NEXT: ; implicit-def: $vgpr145_lo16
-; GFX11-TRUE16-NEXT: ; implicit-def: $vgpr35_lo16
; GFX11-TRUE16-NEXT: ; implicit-def: $vgpr99_hi16
-; GFX11-TRUE16-NEXT: ; implicit-def: $vgpr134_lo16
-; GFX11-TRUE16-NEXT: ; implicit-def: $vgpr100_hi16
-; GFX11-TRUE16-NEXT: ; implicit-def: $vgpr133_lo16
+; GFX11-TRUE16-NEXT: ; implicit-def: $vgpr164_lo16
+; GFX11-TRUE16-NEXT: ; implicit-def: $vgpr85_hi16
+; GFX11-TRUE16-NEXT: ; implicit-def: $vgpr162_lo16
; GFX11-TRUE16-NEXT: ; implicit-def: $vgpr113_hi16
-; GFX11-TRUE16-NEXT: ; implicit-def: $vgpr130_lo16
+; GFX11-TRUE16-NEXT: ; implicit-def: $vgpr151_lo16
+; GFX11-TRUE16-NEXT: ; implicit-def: $vgpr35_lo16
+; GFX11-TRUE16-NEXT: ; implicit-def: $vgpr101_hi16
+; GFX11-TRUE16-NEXT: ; implicit-def: $vgpr146_lo16
+; GFX11-TRUE16-NEXT: ; implicit-def: $vgpr87_hi16
+; GFX11-TRUE16-NEXT: ; implicit-def: $vgpr144_lo16
+; GFX11-TRUE16-NEXT: ; implicit-def: $vgpr115_hi16
+; GFX11-TRUE16-NEXT: ; implicit-def: $vgpr135_lo16
; GFX11-TRUE16-NEXT: ; implicit-def: $vgpr34_lo16
-; GFX11-TRUE16-NEXT: ; implicit-def: $vgpr103_hi16
-; GFX11-TRUE16-NEXT: ; implicit-def: $vgpr129_lo16
-; GFX11-TRUE16-NEXT: ; implicit-def: $vgpr102_hi16
-; GFX11-TRUE16-NEXT: ; implicit-def: $vgpr128_lo16
+; GFX11-TRUE16-NEXT: ; implicit-def: $vgpr112_hi16
+; GFX11-TRUE16-NEXT: ; implicit-def: $vgpr133_lo16
+; GFX11-TRUE16-NEXT: ; implicit-def: $vgpr98_hi16
+; GFX11-TRUE16-NEXT: ; implicit-def: $vgpr130_lo16
; GFX11-TRUE16-NEXT: ; implicit-def: $vgpr116_hi16
-; GFX11-TRUE16-NEXT: ; implicit-def: $vgpr119_lo16
+; GFX11-TRUE16-NEXT: ; implicit-def: $vgpr129_lo16
; GFX11-TRUE16-NEXT: ; implicit-def: $vgpr114_hi16
+; GFX11-TRUE16-NEXT: ; implicit-def: $vgpr119_lo16
+; GFX11-TRUE16-NEXT: ; implicit-def: $vgpr102_hi16
; GFX11-TRUE16-NEXT: ; implicit-def: $vgpr118_lo16
-; GFX11-TRUE16-NEXT: ; implicit-def: $vgpr115_hi16
-; GFX11-TRUE16-NEXT: ; implicit-def: $vgpr117_lo16
; GFX11-TRUE16-NEXT: s_waitcnt vmcnt(2)
; GFX11-TRUE16-NEXT: v_cmp_ne_u32_e32 vcc_lo, 0, v33
; GFX11-TRUE16-NEXT: ; implicit-def: $vgpr33_lo16
@@ -167889,183 +167970,185 @@ define <128 x i8> @bitcast_v64bf16_to_v128i8(<64 x bfloat> %a, i32 %b) {
; GFX11-TRUE16-NEXT: v_lshrrev_b64 v[52:53], 24, v[11:12]
; GFX11-TRUE16-NEXT: v_lshrrev_b64 v[67:68], 24, v[3:4]
; GFX11-TRUE16-NEXT: v_lshrrev_b64 v[37:38], 24, v[23:24]
-; GFX11-TRUE16-NEXT: v_lshrrev_b32_e32 v135, 24, v16
-; GFX11-TRUE16-NEXT: v_lshrrev_b32_e32 v146, 8, v16
-; GFX11-TRUE16-NEXT: v_lshrrev_b32_e32 v151, 8, v15
-; GFX11-TRUE16-NEXT: v_lshrrev_b32_e32 v160, 24, v14
-; GFX11-TRUE16-NEXT: v_lshrrev_b32_e32 v163, 8, v14
-; GFX11-TRUE16-NEXT: v_lshrrev_b32_e32 v167, 8, v13
-; GFX11-TRUE16-NEXT: v_lshrrev_b32_e32 v181, 24, v12
-; GFX11-TRUE16-NEXT: v_lshrrev_b32_e32 v182, 8, v12
-; GFX11-TRUE16-NEXT: v_lshrrev_b32_e32 v40, 8, v11
-; GFX11-TRUE16-NEXT: v_lshrrev_b32_e32 v43, 24, v10
-; GFX11-TRUE16-NEXT: v_lshrrev_b32_e32 v56, 8, v10
-; GFX11-TRUE16-NEXT: v_lshrrev_b32_e32 v57, 8, v9
-; GFX11-TRUE16-NEXT: v_lshrrev_b32_e32 v62, 24, v8
-; GFX11-TRUE16-NEXT: v_lshrrev_b32_e32 v63, 8, v8
-; GFX11-TRUE16-NEXT: v_lshrrev_b32_e32 v75, 8, v7
-; GFX11-TRUE16-NEXT: v_lshrrev_b32_e32 v78, 24, v6
-; GFX11-TRUE16-NEXT: v_lshrrev_b32_e32 v88, 8, v6
-; GFX11-TRUE16-NEXT: v_lshrrev_b32_e32 v90, 8, v5
-; GFX11-TRUE16-NEXT: v_lshrrev_b32_e32 v93, 24, v4
-; GFX11-TRUE16-NEXT: v_lshrrev_b32_e32 v95, 8, v4
+; GFX11-TRUE16-NEXT: v_lshrrev_b32_e32 v149, 24, v16
+; GFX11-TRUE16-NEXT: v_lshrrev_b32_e32 v161, 8, v16
+; GFX11-TRUE16-NEXT: v_lshrrev_b32_e32 v166, 8, v15
+; GFX11-TRUE16-NEXT: v_lshrrev_b32_e32 v176, 24, v14
+; GFX11-TRUE16-NEXT: v_lshrrev_b32_e32 v178, 8, v14
+; GFX11-TRUE16-NEXT: v_lshrrev_b32_e32 v181, 8, v13
+; GFX11-TRUE16-NEXT: v_lshrrev_b32_e32 v44, 24, v12
+; GFX11-TRUE16-NEXT: v_lshrrev_b32_e32 v45, 8, v12
+; GFX11-TRUE16-NEXT: v_lshrrev_b32_e32 v47, 8, v11
+; GFX11-TRUE16-NEXT: v_lshrrev_b32_e32 v57, 24, v10
+; GFX11-TRUE16-NEXT: v_lshrrev_b32_e32 v62, 8, v10
+; GFX11-TRUE16-NEXT: v_lshrrev_b32_e32 v63, 8, v9
+; GFX11-TRUE16-NEXT: v_lshrrev_b32_e32 v76, 24, v8
+; GFX11-TRUE16-NEXT: v_lshrrev_b32_e32 v77, 8, v8
+; GFX11-TRUE16-NEXT: v_lshrrev_b32_e32 v90, 8, v7
+; GFX11-TRUE16-NEXT: v_lshrrev_b32_e32 v91, 24, v6
+; GFX11-TRUE16-NEXT: v_lshrrev_b32_e32 v93, 8, v6
+; GFX11-TRUE16-NEXT: v_lshrrev_b32_e32 v94, 8, v5
+; GFX11-TRUE16-NEXT: v_lshrrev_b32_e32 v95, 24, v4
+; GFX11-TRUE16-NEXT: v_lshrrev_b32_e32 v104, 8, v4
; GFX11-TRUE16-NEXT: v_lshrrev_b32_e32 v105, 8, v3
; GFX11-TRUE16-NEXT: v_lshrrev_b32_e32 v106, 24, v2
; GFX11-TRUE16-NEXT: v_lshrrev_b32_e32 v107, 8, v2
; GFX11-TRUE16-NEXT: v_lshrrev_b32_e32 v108, 8, v1
-; GFX11-TRUE16-NEXT: v_lshrrev_b32_e32 v117, 24, v32
-; GFX11-TRUE16-NEXT: v_lshrrev_b32_e32 v118, 8, v32
-; GFX11-TRUE16-NEXT: v_lshrrev_b32_e32 v119, 8, v31
-; GFX11-TRUE16-NEXT: v_lshrrev_b32_e32 v128, 24, v30
-; GFX11-TRUE16-NEXT: v_lshrrev_b32_e32 v129, 8, v30
-; GFX11-TRUE16-NEXT: v_lshrrev_b32_e32 v130, 8, v29
-; GFX11-TRUE16-NEXT: v_lshrrev_b32_e32 v133, 24, v28
-; GFX11-TRUE16-NEXT: v_lshrrev_b32_e32 v134, 8, v28
-; GFX11-TRUE16-NEXT: v_lshrrev_b32_e32 v145, 8, v27
-; GFX11-TRUE16-NEXT: v_lshrrev_b32_e32 v148, 24, v26
-; GFX11-TRUE16-NEXT: v_lshrrev_b32_e32 v150, 8, v26
-; GFX11-TRUE16-NEXT: v_lshrrev_b32_e32 v161, 8, v25
-; GFX11-TRUE16-NEXT: v_lshrrev_b32_e32 v166, 24, v24
-; GFX11-TRUE16-NEXT: v_lshrrev_b32_e32 v176, 8, v24
-; GFX11-TRUE16-NEXT: v_lshrrev_b32_e32 v177, 8, v23
-; GFX11-TRUE16-NEXT: v_lshrrev_b32_e32 v183, 24, v22
-; GFX11-TRUE16-NEXT: v_lshrrev_b32_e32 v41, 8, v22
-; GFX11-TRUE16-NEXT: v_lshrrev_b32_e32 v44, 8, v21
-; GFX11-TRUE16-NEXT: v_lshrrev_b32_e32 v47, 24, v20
-; GFX11-TRUE16-NEXT: v_lshrrev_b32_e32 v58, 8, v20
-; GFX11-TRUE16-NEXT: v_lshrrev_b32_e32 v61, 8, v19
-; GFX11-TRUE16-NEXT: v_lshrrev_b32_e32 v72, 24, v18
-; GFX11-TRUE16-NEXT: v_lshrrev_b32_e32 v73, 8, v18
-; GFX11-TRUE16-NEXT: v_lshrrev_b32_e32 v79, 8, v17
+; GFX11-TRUE16-NEXT: v_lshrrev_b32_e32 v118, 24, v32
+; GFX11-TRUE16-NEXT: v_lshrrev_b32_e32 v119, 8, v32
+; GFX11-TRUE16-NEXT: v_lshrrev_b32_e32 v129, 8, v31
+; GFX11-TRUE16-NEXT: v_lshrrev_b32_e32 v130, 24, v30
+; GFX11-TRUE16-NEXT: v_lshrrev_b32_e32 v133, 8, v30
+; GFX11-TRUE16-NEXT: v_lshrrev_b32_e32 v135, 8, v29
+; GFX11-TRUE16-NEXT: v_lshrrev_b32_e32 v144, 24, v28
+; GFX11-TRUE16-NEXT: v_lshrrev_b32_e32 v146, 8, v28
+; GFX11-TRUE16-NEXT: v_lshrrev_b32_e32 v151, 8, v27
+; GFX11-TRUE16-NEXT: v_lshrrev_b32_e32 v162, 24, v26
+; GFX11-TRUE16-NEXT: v_lshrrev_b32_e32 v164, 8, v26
+; GFX11-TRUE16-NEXT: v_lshrrev_b32_e32 v177, 8, v25
+; GFX11-TRUE16-NEXT: v_lshrrev_b32_e32 v180, 24, v24
+; GFX11-TRUE16-NEXT: v_lshrrev_b32_e32 v40, 8, v24
+; GFX11-TRUE16-NEXT: v_lshrrev_b32_e32 v41, 8, v23
+; GFX11-TRUE16-NEXT: v_lshrrev_b32_e32 v46, 24, v22
+; GFX11-TRUE16-NEXT: v_lshrrev_b32_e32 v56, 8, v22
+; GFX11-TRUE16-NEXT: v_lshrrev_b32_e32 v59, 8, v21
+; GFX11-TRUE16-NEXT: v_lshrrev_b32_e32 v61, 24, v20
+; GFX11-TRUE16-NEXT: v_lshrrev_b32_e32 v73, 8, v20
+; GFX11-TRUE16-NEXT: v_lshrrev_b32_e32 v75, 8, v19
+; GFX11-TRUE16-NEXT: v_lshrrev_b32_e32 v78, 24, v18
+; GFX11-TRUE16-NEXT: v_lshrrev_b32_e32 v89, 8, v18
+; GFX11-TRUE16-NEXT: v_lshrrev_b32_e32 v92, 8, v17
+; GFX11-TRUE16-NEXT: v_mov_b16_e64 v183.h, v16.h
+; GFX11-TRUE16-NEXT: v_mov_b16_e64 v165.h, v14.h
+; GFX11-TRUE16-NEXT: v_mov_b16_e64 v150.h, v12.h
+; GFX11-TRUE16-NEXT: v_mov_b16_e64 v145.h, v10.h
+; GFX11-TRUE16-NEXT: v_mov_b16_e64 v134.h, v8.h
+; GFX11-TRUE16-NEXT: v_mov_b16_e64 v131.h, v6.h
+; GFX11-TRUE16-NEXT: v_mov_b16_e64 v128.h, v4.h
+; GFX11-TRUE16-NEXT: v_mov_b16_e32 v117.h, v2.h
; GFX11-TRUE16-NEXT: v_lshrrev_b64 v[68:69], 24, v[1:2]
+; GFX11-TRUE16-NEXT: v_mov_b16_e32 v102.h, v32.h
+; GFX11-TRUE16-NEXT: v_mov_b16_e32 v98.h, v30.h
+; GFX11-TRUE16-NEXT: v_mov_b16_e32 v87.h, v28.h
+; GFX11-TRUE16-NEXT: v_mov_b16_e32 v85.h, v26.h
+; GFX11-TRUE16-NEXT: v_mov_b16_e32 v82.h, v24.h
+; GFX11-TRUE16-NEXT: v_mov_b16_e32 v80.h, v22.h
; GFX11-TRUE16-NEXT: v_lshrrev_b64 v[38:39], 24, v[21:22]
+; GFX11-TRUE16-NEXT: v_mov_b16_e32 v70.h, v20.h
; GFX11-TRUE16-NEXT: v_lshrrev_b64 v[50:51], 24, v[19:20]
+; GFX11-TRUE16-NEXT: v_mov_b16_e32 v55.h, v18.h
; GFX11-TRUE16-NEXT: v_lshrrev_b64 v[53:54], 24, v[17:18]
-; GFX11-TRUE16-NEXT: v_mov_b16_e64 v144.h, v1.l
-; GFX11-TRUE16-NEXT: v_mov_b16_e64 v131.h, v2.l
-; GFX11-TRUE16-NEXT: v_mov_b16_e64 v132.h, v2.h
-; GFX11-TRUE16-NEXT: v_mov_b16_e64 v162.h, v3.l
+; GFX11-TRUE16-NEXT: v_mov_b16_e32 v74.h, v16.l
+; GFX11-TRUE16-NEXT: v_mov_b16_e32 v88.h, v15.l
+; GFX11-TRUE16-NEXT: v_mov_b16_e32 v58.h, v14.l
+; GFX11-TRUE16-NEXT: v_mov_b16_e32 v79.h, v13.l
+; GFX11-TRUE16-NEXT: v_mov_b16_e32 v43.h, v12.l
+; GFX11-TRUE16-NEXT: v_mov_b16_e32 v72.h, v11.l
+; GFX11-TRUE16-NEXT: v_mov_b16_e64 v182.h, v10.l
+; GFX11-TRUE16-NEXT: v_mov_b16_e32 v60.h, v9.l
+; GFX11-TRUE16-NEXT: v_mov_b16_e64 v167.h, v8.l
+; GFX11-TRUE16-NEXT: v_mov_b16_e32 v42.h, v7.l
+; GFX11-TRUE16-NEXT: v_mov_b16_e64 v160.h, v6.l
+; GFX11-TRUE16-NEXT: v_mov_b16_e64 v179.h, v5.l
; GFX11-TRUE16-NEXT: v_mov_b16_e64 v147.h, v4.l
-; GFX11-TRUE16-NEXT: v_mov_b16_e64 v149.h, v4.h
-; GFX11-TRUE16-NEXT: v_mov_b16_e64 v180.h, v5.l
-; GFX11-TRUE16-NEXT: v_mov_b16_e64 v164.h, v6.l
-; GFX11-TRUE16-NEXT: v_mov_b16_e64 v165.h, v6.h
-; GFX11-TRUE16-NEXT: v_mov_b16_e32 v46.h, v7.l
-; GFX11-TRUE16-NEXT: v_mov_b16_e64 v178.h, v8.l
-; GFX11-TRUE16-NEXT: v_mov_b16_e64 v179.h, v8.h
-; GFX11-TRUE16-NEXT: v_mov_b16_e32 v74.h, v9.l
-; GFX11-TRUE16-NEXT: v_mov_b16_e32 v45.h, v10.l
-; GFX11-TRUE16-NEXT: v_mov_b16_e32 v42.h, v10.h
-; GFX11-TRUE16-NEXT: v_mov_b16_e32 v89.h, v11.l
-; GFX11-TRUE16-NEXT: v_mov_b16_e32 v59.h, v12.l
-; GFX11-TRUE16-NEXT: v_mov_b16_e32 v60.h, v12.h
-; GFX11-TRUE16-NEXT: v_mov_b16_e32 v94.h, v13.l
-; GFX11-TRUE16-NEXT: v_mov_b16_e32 v77.h, v14.l
-; GFX11-TRUE16-NEXT: v_mov_b16_e32 v76.h, v14.h
-; GFX11-TRUE16-NEXT: v_mov_b16_e32 v104.h, v15.l
-; GFX11-TRUE16-NEXT: v_mov_b16_e32 v91.h, v16.l
-; GFX11-TRUE16-NEXT: v_mov_b16_e32 v92.h, v16.h
-; GFX11-TRUE16-NEXT: v_mov_b16_e32 v80.h, v17.l
-; GFX11-TRUE16-NEXT: v_mov_b16_e32 v55.h, v18.l
-; GFX11-TRUE16-NEXT: v_mov_b16_e32 v70.h, v18.h
-; GFX11-TRUE16-NEXT: v_mov_b16_e32 v83.h, v19.l
-; GFX11-TRUE16-NEXT: v_mov_b16_e32 v81.h, v20.l
-; GFX11-TRUE16-NEXT: v_mov_b16_e32 v71.h, v20.h
-; GFX11-TRUE16-NEXT: v_mov_b16_e32 v85.h, v21.l
-; GFX11-TRUE16-NEXT: v_mov_b16_e32 v82.h, v22.l
-; GFX11-TRUE16-NEXT: v_mov_b16_e32 v84.h, v22.h
-; GFX11-TRUE16-NEXT: v_mov_b16_e32 v96.h, v23.l
-; GFX11-TRUE16-NEXT: v_mov_b16_e32 v86.h, v24.l
-; GFX11-TRUE16-NEXT: v_mov_b16_e32 v87.h, v24.h
-; GFX11-TRUE16-NEXT: v_mov_b16_e32 v101.h, v25.l
-; GFX11-TRUE16-NEXT: v_mov_b16_e32 v97.h, v26.l
-; GFX11-TRUE16-NEXT: v_mov_b16_e32 v98.h, v26.h
-; GFX11-TRUE16-NEXT: v_mov_b16_e32 v112.h, v27.l
-; GFX11-TRUE16-NEXT: v_mov_b16_e32 v99.h, v28.l
-; GFX11-TRUE16-NEXT: v_mov_b16_e32 v100.h, v28.h
-; GFX11-TRUE16-NEXT: v_mov_b16_e32 v113.h, v29.l
-; GFX11-TRUE16-NEXT: v_mov_b16_e32 v103.h, v30.l
-; GFX11-TRUE16-NEXT: v_mov_b16_e32 v102.h, v30.h
-; GFX11-TRUE16-NEXT: v_mov_b16_e32 v116.h, v31.l
+; GFX11-TRUE16-NEXT: v_mov_b16_e64 v163.h, v3.l
+; GFX11-TRUE16-NEXT: v_mov_b16_e64 v132.h, v2.l
+; GFX11-TRUE16-NEXT: v_mov_b16_e64 v148.h, v1.l
; GFX11-TRUE16-NEXT: v_mov_b16_e32 v114.h, v32.l
-; GFX11-TRUE16-NEXT: v_mov_b16_e32 v115.h, v32.h
+; GFX11-TRUE16-NEXT: v_mov_b16_e32 v116.h, v31.l
+; GFX11-TRUE16-NEXT: v_mov_b16_e32 v112.h, v30.l
+; GFX11-TRUE16-NEXT: v_mov_b16_e32 v115.h, v29.l
+; GFX11-TRUE16-NEXT: v_mov_b16_e32 v101.h, v28.l
+; GFX11-TRUE16-NEXT: v_mov_b16_e32 v113.h, v27.l
+; GFX11-TRUE16-NEXT: v_mov_b16_e32 v99.h, v26.l
+; GFX11-TRUE16-NEXT: v_mov_b16_e32 v103.h, v25.l
+; GFX11-TRUE16-NEXT: v_mov_b16_e32 v96.h, v24.l
+; GFX11-TRUE16-NEXT: v_mov_b16_e32 v100.h, v23.l
+; GFX11-TRUE16-NEXT: v_mov_b16_e32 v84.h, v22.l
+; GFX11-TRUE16-NEXT: v_mov_b16_e32 v97.h, v21.l
+; GFX11-TRUE16-NEXT: v_mov_b16_e32 v81.h, v20.l
+; GFX11-TRUE16-NEXT: v_mov_b16_e32 v86.h, v19.l
+; GFX11-TRUE16-NEXT: v_mov_b16_e32 v71.h, v18.l
+; GFX11-TRUE16-NEXT: v_mov_b16_e32 v83.h, v17.l
; GFX11-TRUE16-NEXT: .LBB90_2: ; %Flow
; GFX11-TRUE16-NEXT: s_and_not1_saveexec_b32 s0, s0
; GFX11-TRUE16-NEXT: s_cbranch_execz .LBB90_4
; GFX11-TRUE16-NEXT: ; %bb.3: ; %cmp.true
-; GFX11-TRUE16-NEXT: v_and_b32_e32 v34, 0xffff0000, v17
-; GFX11-TRUE16-NEXT: v_lshlrev_b32_e32 v17, 16, v17
-; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
-; GFX11-TRUE16-NEXT: v_add_f32_e32 v17, 0x40c00000, v17
-; GFX11-TRUE16-NEXT: v_or_b32_e32 v51, 0x400000, v17
; GFX11-TRUE16-NEXT: v_and_b32_e32 v33, 0xffff0000, v18
-; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
-; GFX11-TRUE16-NEXT: v_dual_add_f32 v33, 0x40c00000, v33 :: v_dual_lshlrev_b32 v18, 16, v18
-; GFX11-TRUE16-NEXT: v_add_f32_e32 v18, 0x40c00000, v18
-; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_2) | instskip(NEXT) | instid1(VALU_DEP_2)
+; GFX11-TRUE16-NEXT: v_lshlrev_b32_e32 v18, 16, v18
+; GFX11-TRUE16-NEXT: v_and_b32_e32 v35, 0xffff0000, v20
+; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_3) | instskip(NEXT) | instid1(VALU_DEP_2)
+; GFX11-TRUE16-NEXT: v_dual_add_f32 v33, 0x40c00000, v33 :: v_dual_lshlrev_b32 v20, 16, v20
+; GFX11-TRUE16-NEXT: v_dual_add_f32 v18, 0x40c00000, v18 :: v_dual_add_f32 v35, 0x40c00000, v35
+; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_2) | instskip(NEXT) | instid1(VALU_DEP_3)
+; GFX11-TRUE16-NEXT: v_add_f32_e32 v20, 0x40c00000, v20
; GFX11-TRUE16-NEXT: v_bfe_u32 v36, v33, 16, 1
+; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_3)
; GFX11-TRUE16-NEXT: v_bfe_u32 v37, v18, 16, 1
; GFX11-TRUE16-NEXT: v_or_b32_e32 v39, 0x400000, v18
; GFX11-TRUE16-NEXT: v_cmp_u_f32_e32 vcc_lo, v18, v18
; GFX11-TRUE16-NEXT: v_or_b32_e32 v38, 0x400000, v33
; GFX11-TRUE16-NEXT: v_add3_u32 v36, v36, v33, 0x7fff
; GFX11-TRUE16-NEXT: v_add3_u32 v37, v37, v18, 0x7fff
-; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1)
-; GFX11-TRUE16-NEXT: v_cndmask_b32_e32 v55, v37, v39, vcc_lo
+; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
+; GFX11-TRUE16-NEXT: v_dual_cndmask_b32 v71, v37, v39 :: v_dual_and_b32 v34, 0xffff0000, v17
+; GFX11-TRUE16-NEXT: v_dual_add_f32 v34, 0x40c00000, v34 :: v_dual_lshlrev_b32 v17, 16, v17
; GFX11-TRUE16-NEXT: v_cmp_u_f32_e32 vcc_lo, v33, v33
-; GFX11-TRUE16-NEXT: v_and_b32_e32 v35, 0xffff0000, v20
-; GFX11-TRUE16-NEXT: v_lshlrev_b32_e32 v20, 16, v20
-; GFX11-TRUE16-NEXT: v_bfe_u32 v50, v17, 16, 1
-; GFX11-TRUE16-NEXT: v_mov_b16_e32 v33.l, v55.h
-; GFX11-TRUE16-NEXT: v_cndmask_b32_e32 v70, v36, v38, vcc_lo
-; GFX11-TRUE16-NEXT: v_dual_add_f32 v34, 0x40c00000, v34 :: v_dual_add_f32 v35, 0x40c00000, v35
-; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_4) | instskip(SKIP_2) | instid1(VALU_DEP_4)
-; GFX11-TRUE16-NEXT: v_add3_u32 v37, v50, v17, 0x7fff
-; GFX11-TRUE16-NEXT: v_cmp_u_f32_e32 vcc_lo, v17, v17
-; GFX11-TRUE16-NEXT: v_add_f32_e32 v20, 0x40c00000, v20
+; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_3) | instskip(NEXT) | instid1(VALU_DEP_3)
+; GFX11-TRUE16-NEXT: v_mov_b16_e32 v33.l, v71.h
+; GFX11-TRUE16-NEXT: v_add_f32_e32 v17, 0x40c00000, v17
+; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_4)
; GFX11-TRUE16-NEXT: v_bfe_u32 v48, v34, 16, 1
+; GFX11-TRUE16-NEXT: v_cndmask_b32_e32 v55, v36, v38, vcc_lo
; GFX11-TRUE16-NEXT: v_or_b32_e32 v49, 0x400000, v34
; GFX11-TRUE16-NEXT: v_bfe_u32 v36, v35, 16, 1
-; GFX11-TRUE16-NEXT: v_cndmask_b32_e32 v80, v37, v51, vcc_lo
-; GFX11-TRUE16-NEXT: v_cmp_u_f32_e32 vcc_lo, v34, v34
+; GFX11-TRUE16-NEXT: v_bfe_u32 v50, v17, 16, 1
+; GFX11-TRUE16-NEXT: v_or_b32_e32 v51, 0x400000, v17
+; GFX11-TRUE16-NEXT: v_cmp_u_f32_e32 vcc_lo, v17, v17
; GFX11-TRUE16-NEXT: v_add3_u32 v18, v48, v34, 0x7fff
-; GFX11-TRUE16-NEXT: v_or_b32_e32 v38, 0x400000, v35
+; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_4) | instskip(NEXT) | instid1(VALU_DEP_1)
+; GFX11-TRUE16-NEXT: v_add3_u32 v37, v50, v17, 0x7fff
+; GFX11-TRUE16-NEXT: v_cndmask_b32_e32 v83, v37, v51, vcc_lo
+; GFX11-TRUE16-NEXT: v_cmp_u_f32_e32 vcc_lo, v34, v34
+; GFX11-TRUE16-NEXT: v_and_b32_e32 v51, 0xffff0000, v11
; GFX11-TRUE16-NEXT: v_or_b32_e32 v37, 0x400000, v20
-; GFX11-TRUE16-NEXT: v_mov_b16_e32 v34.l, v80.h
-; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_4) | instskip(SKIP_3) | instid1(VALU_DEP_4)
+; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_4) | instskip(SKIP_4) | instid1(VALU_DEP_4)
+; GFX11-TRUE16-NEXT: v_mov_b16_e32 v34.l, v83.h
; GFX11-TRUE16-NEXT: v_cndmask_b32_e32 v17, v18, v49, vcc_lo
-; GFX11-TRUE16-NEXT: v_bfi_b32 v18, 0xffff, v33, v70
+; GFX11-TRUE16-NEXT: v_bfi_b32 v18, 0xffff, v33, v55
; GFX11-TRUE16-NEXT: v_bfe_u32 v33, v20, 16, 1
; GFX11-TRUE16-NEXT: v_cmp_u_f32_e32 vcc_lo, v20, v20
; GFX11-TRUE16-NEXT: v_bfi_b32 v17, 0xffff, v34, v17
; GFX11-TRUE16-NEXT: v_add3_u32 v34, v36, v35, 0x7fff
-; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_4)
-; GFX11-TRUE16-NEXT: v_add3_u32 v33, v33, v20, 0x7fff
; GFX11-TRUE16-NEXT: v_and_b32_e32 v36, 0xffff0000, v19
-; GFX11-TRUE16-NEXT: v_lshrrev_b32_e32 v72, 24, v18
-; GFX11-TRUE16-NEXT: v_lshrrev_b32_e32 v73, 8, v18
-; GFX11-TRUE16-NEXT: v_lshrrev_b32_e32 v79, 8, v17
+; GFX11-TRUE16-NEXT: v_add3_u32 v33, v33, v20, 0x7fff
+; GFX11-TRUE16-NEXT: v_lshrrev_b32_e32 v78, 24, v18
+; GFX11-TRUE16-NEXT: v_lshrrev_b32_e32 v89, 8, v18
+; GFX11-TRUE16-NEXT: v_lshrrev_b32_e32 v92, 8, v17
+; GFX11-TRUE16-NEXT: v_add_f32_e32 v36, 0x40c00000, v36
+; GFX11-TRUE16-NEXT: v_or_b32_e32 v38, 0x400000, v35
; GFX11-TRUE16-NEXT: v_cndmask_b32_e32 v81, v33, v37, vcc_lo
-; GFX11-TRUE16-NEXT: v_dual_add_f32 v36, 0x40c00000, v36 :: v_dual_lshlrev_b32 v19, 16, v19
; GFX11-TRUE16-NEXT: v_cmp_u_f32_e32 vcc_lo, v35, v35
; GFX11-TRUE16-NEXT: v_and_b32_e32 v35, 0xffff0000, v22
-; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_3) | instskip(NEXT) | instid1(VALU_DEP_4)
-; GFX11-TRUE16-NEXT: v_dual_add_f32 v19, 0x40c00000, v19 :: v_dual_lshlrev_b32 v22, 16, v22
+; GFX11-TRUE16-NEXT: v_lshlrev_b32_e32 v22, 16, v22
+; GFX11-TRUE16-NEXT: v_lshlrev_b32_e32 v19, 16, v19
; GFX11-TRUE16-NEXT: v_bfe_u32 v20, v36, 16, 1
-; GFX11-TRUE16-NEXT: v_cndmask_b32_e32 v71, v34, v38, vcc_lo
+; GFX11-TRUE16-NEXT: v_cndmask_b32_e32 v70, v34, v38, vcc_lo
; GFX11-TRUE16-NEXT: v_or_b32_e32 v38, 0x400000, v36
+; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_4) | instskip(NEXT) | instid1(VALU_DEP_4)
+; GFX11-TRUE16-NEXT: v_dual_add_f32 v22, 0x40c00000, v22 :: v_dual_add_f32 v19, 0x40c00000, v19
+; GFX11-TRUE16-NEXT: v_add3_u32 v20, v20, v36, 0x7fff
; GFX11-TRUE16-NEXT: v_mov_b16_e32 v34.l, v81.h
+; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_3) | instskip(SKIP_2) | instid1(VALU_DEP_3)
; GFX11-TRUE16-NEXT: v_bfe_u32 v33, v19, 16, 1
; GFX11-TRUE16-NEXT: v_or_b32_e32 v37, 0x400000, v19
; GFX11-TRUE16-NEXT: v_cmp_u_f32_e32 vcc_lo, v19, v19
-; GFX11-TRUE16-NEXT: v_add3_u32 v20, v20, v36, 0x7fff
-; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_4) | instskip(NEXT) | instid1(VALU_DEP_1)
; GFX11-TRUE16-NEXT: v_add3_u32 v33, v33, v19, 0x7fff
-; GFX11-TRUE16-NEXT: v_dual_add_f32 v22, 0x40c00000, v22 :: v_dual_cndmask_b32 v83, v33, v37
-; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(SKIP_1) | instid1(VALU_DEP_3)
+; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(SKIP_2) | instid1(VALU_DEP_3)
+; GFX11-TRUE16-NEXT: v_cndmask_b32_e32 v86, v33, v37, vcc_lo
; GFX11-TRUE16-NEXT: v_bfe_u32 v33, v22, 16, 1
; GFX11-TRUE16-NEXT: v_cmp_u_f32_e32 vcc_lo, v36, v36
-; GFX11-TRUE16-NEXT: v_mov_b16_e32 v37.l, v83.h
+; GFX11-TRUE16-NEXT: v_mov_b16_e32 v37.l, v86.h
; GFX11-TRUE16-NEXT: v_cndmask_b32_e32 v36, v20, v38, vcc_lo
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_4) | instskip(SKIP_4) | instid1(VALU_DEP_4)
; GFX11-TRUE16-NEXT: v_add3_u32 v20, v33, v22, 0x7fff
@@ -168073,264 +168156,267 @@ define <128 x i8> @bitcast_v64bf16_to_v128i8(<64 x bfloat> %a, i32 %b) {
; GFX11-TRUE16-NEXT: v_cmp_u_f32_e32 vcc_lo, v22, v22
; GFX11-TRUE16-NEXT: v_and_b32_e32 v38, 0xffff0000, v21
; GFX11-TRUE16-NEXT: v_lshlrev_b32_e32 v21, 16, v21
-; GFX11-TRUE16-NEXT: v_dual_cndmask_b32 v82, v20, v33 :: v_dual_add_f32 v35, 0x40c00000, v35
-; GFX11-TRUE16-NEXT: v_bfi_b32 v20, 0xffff, v34, v71
+; GFX11-TRUE16-NEXT: v_dual_cndmask_b32 v84, v20, v33 :: v_dual_add_f32 v35, 0x40c00000, v35
+; GFX11-TRUE16-NEXT: v_bfi_b32 v20, 0xffff, v34, v70
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_2) | instskip(NEXT) | instid1(VALU_DEP_3)
-; GFX11-TRUE16-NEXT: v_mov_b16_e32 v22.l, v82.h
+; GFX11-TRUE16-NEXT: v_mov_b16_e32 v22.l, v84.h
; GFX11-TRUE16-NEXT: v_bfe_u32 v19, v35, 16, 1
; GFX11-TRUE16-NEXT: v_or_b32_e32 v39, 0x400000, v35
; GFX11-TRUE16-NEXT: v_cmp_u_f32_e32 vcc_lo, v35, v35
-; GFX11-TRUE16-NEXT: v_lshrrev_b32_e32 v47, 24, v20
-; GFX11-TRUE16-NEXT: v_lshrrev_b32_e32 v58, 8, v20
+; GFX11-TRUE16-NEXT: v_lshrrev_b32_e32 v61, 24, v20
+; GFX11-TRUE16-NEXT: v_lshrrev_b32_e32 v73, 8, v20
; GFX11-TRUE16-NEXT: v_add3_u32 v19, v19, v35, 0x7fff
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(SKIP_3) | instid1(VALU_DEP_4)
-; GFX11-TRUE16-NEXT: v_dual_add_f32 v33, 0x40c00000, v38 :: v_dual_cndmask_b32 v84, v19, v39
+; GFX11-TRUE16-NEXT: v_cndmask_b32_e32 v80, v19, v39, vcc_lo
; GFX11-TRUE16-NEXT: v_bfi_b32 v19, 0xffff, v37, v36
; GFX11-TRUE16-NEXT: v_and_b32_e32 v36, 0xffff0000, v24
-; GFX11-TRUE16-NEXT: v_add_f32_e32 v21, 0x40c00000, v21
-; GFX11-TRUE16-NEXT: v_bfe_u32 v35, v33, 16, 1
-; GFX11-TRUE16-NEXT: v_lshlrev_b32_e32 v24, 16, v24
-; GFX11-TRUE16-NEXT: v_or_b32_e32 v38, 0x400000, v33
-; GFX11-TRUE16-NEXT: v_add_f32_e32 v36, 0x40c00000, v36
+; GFX11-TRUE16-NEXT: v_dual_add_f32 v21, 0x40c00000, v21 :: v_dual_lshlrev_b32 v24, 16, v24
+; GFX11-TRUE16-NEXT: v_bfi_b32 v22, 0xffff, v22, v80
+; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_4) | instskip(NEXT) | instid1(VALU_DEP_3)
+; GFX11-TRUE16-NEXT: v_lshrrev_b32_e32 v75, 8, v19
+; GFX11-TRUE16-NEXT: v_add_f32_e32 v24, 0x40c00000, v24
+; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_4)
; GFX11-TRUE16-NEXT: v_bfe_u32 v34, v21, 16, 1
; GFX11-TRUE16-NEXT: v_or_b32_e32 v37, 0x400000, v21
; GFX11-TRUE16-NEXT: v_cmp_u_f32_e32 vcc_lo, v21, v21
-; GFX11-TRUE16-NEXT: v_add3_u32 v35, v35, v33, 0x7fff
-; GFX11-TRUE16-NEXT: v_or_b32_e32 v39, 0x400000, v36
+; GFX11-TRUE16-NEXT: v_lshrrev_b32_e32 v46, 24, v22
+; GFX11-TRUE16-NEXT: v_lshrrev_b32_e32 v56, 8, v22
; GFX11-TRUE16-NEXT: v_add3_u32 v34, v34, v21, 0x7fff
-; GFX11-TRUE16-NEXT: v_bfi_b32 v22, 0xffff, v22, v84
-; GFX11-TRUE16-NEXT: v_lshrrev_b32_e32 v61, 8, v19
-; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_3)
-; GFX11-TRUE16-NEXT: v_cndmask_b32_e32 v85, v34, v37, vcc_lo
+; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(SKIP_2) | instid1(VALU_DEP_2)
+; GFX11-TRUE16-NEXT: v_cndmask_b32_e32 v97, v34, v37, vcc_lo
+; GFX11-TRUE16-NEXT: v_dual_add_f32 v33, 0x40c00000, v38 :: v_dual_add_f32 v36, 0x40c00000, v36
+; GFX11-TRUE16-NEXT: v_bfe_u32 v34, v24, 16, 1
+; GFX11-TRUE16-NEXT: v_bfe_u32 v35, v33, 16, 1
+; GFX11-TRUE16-NEXT: v_or_b32_e32 v38, 0x400000, v33
; GFX11-TRUE16-NEXT: v_cmp_u_f32_e32 vcc_lo, v33, v33
-; GFX11-TRUE16-NEXT: v_add_f32_e32 v24, 0x40c00000, v24
+; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_4) | instskip(SKIP_3) | instid1(VALU_DEP_1)
+; GFX11-TRUE16-NEXT: v_add3_u32 v34, v34, v24, 0x7fff
+; GFX11-TRUE16-NEXT: v_or_b32_e32 v39, 0x400000, v36
+; GFX11-TRUE16-NEXT: v_add3_u32 v35, v35, v33, 0x7fff
; GFX11-TRUE16-NEXT: v_bfe_u32 v33, v36, 16, 1
-; GFX11-TRUE16-NEXT: v_lshrrev_b32_e32 v183, 24, v22
-; GFX11-TRUE16-NEXT: v_lshrrev_b32_e32 v41, 8, v22
+; GFX11-TRUE16-NEXT: v_add3_u32 v33, v33, v36, 0x7fff
+; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_3)
; GFX11-TRUE16-NEXT: v_cndmask_b32_e32 v21, v35, v38, vcc_lo
-; GFX11-TRUE16-NEXT: v_bfe_u32 v34, v24, 16, 1
; GFX11-TRUE16-NEXT: v_or_b32_e32 v38, 0x400000, v24
; GFX11-TRUE16-NEXT: v_cmp_u_f32_e32 vcc_lo, v24, v24
-; GFX11-TRUE16-NEXT: v_add3_u32 v33, v33, v36, 0x7fff
-; GFX11-TRUE16-NEXT: v_mov_b16_e32 v35.l, v85.h
-; GFX11-TRUE16-NEXT: v_add3_u32 v34, v34, v24, 0x7fff
; GFX11-TRUE16-NEXT: v_and_b32_e32 v37, 0xffff0000, v23
; GFX11-TRUE16-NEXT: v_lshlrev_b32_e32 v23, 16, v23
-; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_4) | instskip(NEXT) | instid1(VALU_DEP_4)
-; GFX11-TRUE16-NEXT: v_bfi_b32 v21, 0xffff, v35, v21
-; GFX11-TRUE16-NEXT: v_cndmask_b32_e32 v86, v34, v38, vcc_lo
-; GFX11-TRUE16-NEXT: v_cmp_u_f32_e32 vcc_lo, v36, v36
+; GFX11-TRUE16-NEXT: v_mov_b16_e32 v35.l, v97.h
+; GFX11-TRUE16-NEXT: v_cndmask_b32_e32 v96, v34, v38, vcc_lo
; GFX11-TRUE16-NEXT: v_and_b32_e32 v38, 0xffff0000, v26
+; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_4) | instskip(SKIP_2) | instid1(VALU_DEP_3)
; GFX11-TRUE16-NEXT: v_add_f32_e32 v23, 0x40c00000, v23
; GFX11-TRUE16-NEXT: v_dual_add_f32 v37, 0x40c00000, v37 :: v_dual_lshlrev_b32 v26, 16, v26
-; GFX11-TRUE16-NEXT: v_cndmask_b32_e32 v87, v33, v39, vcc_lo
-; GFX11-TRUE16-NEXT: v_mov_b16_e32 v33.l, v86.h
-; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_4) | instskip(NEXT) | instid1(VALU_DEP_4)
+; GFX11-TRUE16-NEXT: v_cmp_u_f32_e32 vcc_lo, v36, v36
; GFX11-TRUE16-NEXT: v_bfe_u32 v34, v23, 16, 1
+; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_3)
; GFX11-TRUE16-NEXT: v_bfe_u32 v24, v37, 16, 1
; GFX11-TRUE16-NEXT: v_or_b32_e32 v36, 0x400000, v23
+; GFX11-TRUE16-NEXT: v_cndmask_b32_e32 v82, v33, v39, vcc_lo
; GFX11-TRUE16-NEXT: v_cmp_u_f32_e32 vcc_lo, v23, v23
-; GFX11-TRUE16-NEXT: v_or_b32_e32 v39, 0x400000, v37
; GFX11-TRUE16-NEXT: v_add3_u32 v34, v34, v23, 0x7fff
; GFX11-TRUE16-NEXT: v_add3_u32 v24, v24, v37, 0x7fff
+; GFX11-TRUE16-NEXT: v_or_b32_e32 v39, 0x400000, v37
+; GFX11-TRUE16-NEXT: v_mov_b16_e32 v33.l, v96.h
; GFX11-TRUE16-NEXT: v_add_f32_e32 v26, 0x40c00000, v26
-; GFX11-TRUE16-NEXT: v_lshrrev_b32_e32 v44, 8, v21
-; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_4) | instskip(SKIP_2) | instid1(VALU_DEP_3)
-; GFX11-TRUE16-NEXT: v_cndmask_b32_e32 v96, v34, v36, vcc_lo
-; GFX11-TRUE16-NEXT: v_cmp_u_f32_e32 vcc_lo, v37, v37
+; GFX11-TRUE16-NEXT: v_cndmask_b32_e32 v100, v34, v36, vcc_lo
; GFX11-TRUE16-NEXT: v_add_f32_e32 v34, 0x40c00000, v38
-; GFX11-TRUE16-NEXT: v_mov_b16_e32 v36.l, v96.h
-; GFX11-TRUE16-NEXT: v_cndmask_b32_e32 v23, v24, v39, vcc_lo
-; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_3)
+; GFX11-TRUE16-NEXT: v_cmp_u_f32_e32 vcc_lo, v37, v37
+; GFX11-TRUE16-NEXT: v_bfi_b32 v21, 0xffff, v35, v21
+; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_4) | instskip(NEXT) | instid1(VALU_DEP_4)
+; GFX11-TRUE16-NEXT: v_mov_b16_e32 v36.l, v100.h
; GFX11-TRUE16-NEXT: v_bfe_u32 v37, v34, 16, 1
-; GFX11-TRUE16-NEXT: v_bfi_b32 v24, 0xffff, v33, v87
-; GFX11-TRUE16-NEXT: v_bfe_u32 v33, v26, 16, 1
-; GFX11-TRUE16-NEXT: v_cmp_u_f32_e32 vcc_lo, v26, v26
-; GFX11-TRUE16-NEXT: v_bfi_b32 v23, 0xffff, v36, v23
+; GFX11-TRUE16-NEXT: v_cndmask_b32_e32 v23, v24, v39, vcc_lo
+; GFX11-TRUE16-NEXT: v_bfi_b32 v24, 0xffff, v33, v82
+; GFX11-TRUE16-NEXT: v_or_b32_e32 v38, 0x400000, v34
+; GFX11-TRUE16-NEXT: v_lshrrev_b32_e32 v59, 8, v21
; GFX11-TRUE16-NEXT: v_add3_u32 v35, v37, v34, 0x7fff
+; GFX11-TRUE16-NEXT: v_bfi_b32 v23, 0xffff, v36, v23
; GFX11-TRUE16-NEXT: v_and_b32_e32 v36, 0xffff0000, v25
; GFX11-TRUE16-NEXT: v_lshlrev_b32_e32 v25, 16, v25
-; GFX11-TRUE16-NEXT: v_add3_u32 v33, v33, v26, 0x7fff
+; GFX11-TRUE16-NEXT: v_bfe_u32 v33, v26, 16, 1
; GFX11-TRUE16-NEXT: v_or_b32_e32 v37, 0x400000, v26
-; GFX11-TRUE16-NEXT: v_or_b32_e32 v38, 0x400000, v34
-; GFX11-TRUE16-NEXT: v_add_f32_e32 v36, 0x40c00000, v36
-; GFX11-TRUE16-NEXT: v_lshrrev_b32_e32 v166, 24, v24
-; GFX11-TRUE16-NEXT: v_lshrrev_b32_e32 v176, 8, v24
-; GFX11-TRUE16-NEXT: v_cndmask_b32_e32 v97, v33, v37, vcc_lo
-; GFX11-TRUE16-NEXT: v_cmp_u_f32_e32 vcc_lo, v34, v34
+; GFX11-TRUE16-NEXT: v_cmp_u_f32_e32 vcc_lo, v26, v26
+; GFX11-TRUE16-NEXT: v_lshrrev_b32_e32 v180, 24, v24
; GFX11-TRUE16-NEXT: v_add_f32_e32 v25, 0x40c00000, v25
-; GFX11-TRUE16-NEXT: v_bfe_u32 v26, v36, 16, 1
-; GFX11-TRUE16-NEXT: v_lshrrev_b32_e32 v177, 8, v23
-; GFX11-TRUE16-NEXT: v_cndmask_b32_e32 v98, v35, v38, vcc_lo
-; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_4)
+; GFX11-TRUE16-NEXT: v_add3_u32 v33, v33, v26, 0x7fff
+; GFX11-TRUE16-NEXT: v_lshrrev_b32_e32 v40, 8, v24
+; GFX11-TRUE16-NEXT: v_lshrrev_b32_e32 v41, 8, v23
+; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_3) | instskip(SKIP_3) | instid1(VALU_DEP_4)
+; GFX11-TRUE16-NEXT: v_cndmask_b32_e32 v99, v33, v37, vcc_lo
+; GFX11-TRUE16-NEXT: v_cmp_u_f32_e32 vcc_lo, v34, v34
; GFX11-TRUE16-NEXT: v_bfe_u32 v33, v25, 16, 1
; GFX11-TRUE16-NEXT: v_or_b32_e32 v37, 0x400000, v25
+; GFX11-TRUE16-NEXT: v_mov_b16_e32 v34.l, v99.h
+; GFX11-TRUE16-NEXT: v_dual_cndmask_b32 v85, v35, v38 :: v_dual_add_f32 v36, 0x40c00000, v36
+; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_4) | instskip(SKIP_1) | instid1(VALU_DEP_3)
+; GFX11-TRUE16-NEXT: v_add3_u32 v33, v33, v25, 0x7fff
; GFX11-TRUE16-NEXT: v_cmp_u_f32_e32 vcc_lo, v25, v25
-; GFX11-TRUE16-NEXT: v_add3_u32 v26, v26, v36, 0x7fff
+; GFX11-TRUE16-NEXT: v_bfe_u32 v26, v36, 16, 1
; GFX11-TRUE16-NEXT: v_or_b32_e32 v38, 0x400000, v36
-; GFX11-TRUE16-NEXT: v_add3_u32 v33, v33, v25, 0x7fff
-; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(SKIP_4) | instid1(VALU_DEP_3)
-; GFX11-TRUE16-NEXT: v_cndmask_b32_e32 v101, v33, v37, vcc_lo
+; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_4) | instskip(SKIP_1) | instid1(VALU_DEP_4)
+; GFX11-TRUE16-NEXT: v_cndmask_b32_e32 v103, v33, v37, vcc_lo
; GFX11-TRUE16-NEXT: v_cmp_u_f32_e32 vcc_lo, v36, v36
-; GFX11-TRUE16-NEXT: v_and_b32_e32 v35, 0xffff0000, v28
-; GFX11-TRUE16-NEXT: v_lshlrev_b32_e32 v28, 16, v28
-; GFX11-TRUE16-NEXT: v_mov_b16_e32 v34.l, v97.h
-; GFX11-TRUE16-NEXT: v_dual_cndmask_b32 v36, v26, v38 :: v_dual_add_f32 v35, 0x40c00000, v35
-; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_3) | instskip(SKIP_2) | instid1(VALU_DEP_4)
-; GFX11-TRUE16-NEXT: v_add_f32_e32 v28, 0x40c00000, v28
+; GFX11-TRUE16-NEXT: v_add3_u32 v26, v26, v36, 0x7fff
+; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
+; GFX11-TRUE16-NEXT: v_dual_cndmask_b32 v36, v26, v38 :: v_dual_and_b32 v35, 0xffff0000, v28
+; GFX11-TRUE16-NEXT: v_dual_add_f32 v35, 0x40c00000, v35 :: v_dual_lshlrev_b32 v28, 16, v28
; GFX11-TRUE16-NEXT: v_and_b32_e32 v38, 0xffff0000, v27
-; GFX11-TRUE16-NEXT: v_mov_b16_e32 v37.l, v101.h
+; GFX11-TRUE16-NEXT: v_lshlrev_b32_e32 v27, 16, v27
+; GFX11-TRUE16-NEXT: v_mov_b16_e32 v37.l, v103.h
+; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_4) | instskip(SKIP_2) | instid1(VALU_DEP_3)
+; GFX11-TRUE16-NEXT: v_add_f32_e32 v28, 0x40c00000, v28
; GFX11-TRUE16-NEXT: v_bfe_u32 v25, v35, 16, 1
-; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_4) | instskip(SKIP_2) | instid1(VALU_DEP_4)
+; GFX11-TRUE16-NEXT: v_or_b32_e32 v39, 0x400000, v35
; GFX11-TRUE16-NEXT: v_bfe_u32 v33, v28, 16, 1
; GFX11-TRUE16-NEXT: v_cmp_u_f32_e32 vcc_lo, v28, v28
-; GFX11-TRUE16-NEXT: v_or_b32_e32 v39, 0x400000, v35
+; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_4) | instskip(NEXT) | instid1(VALU_DEP_3)
; GFX11-TRUE16-NEXT: v_add3_u32 v25, v25, v35, 0x7fff
-; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_4) | instskip(SKIP_1) | instid1(VALU_DEP_1)
; GFX11-TRUE16-NEXT: v_add3_u32 v26, v33, v28, 0x7fff
; GFX11-TRUE16-NEXT: v_or_b32_e32 v33, 0x400000, v28
-; GFX11-TRUE16-NEXT: v_cndmask_b32_e32 v99, v26, v33, vcc_lo
+; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1)
+; GFX11-TRUE16-NEXT: v_cndmask_b32_e32 v101, v26, v33, vcc_lo
; GFX11-TRUE16-NEXT: v_cmp_u_f32_e32 vcc_lo, v35, v35
-; GFX11-TRUE16-NEXT: v_bfi_b32 v26, 0xffff, v34, v98
-; GFX11-TRUE16-NEXT: v_cndmask_b32_e32 v100, v25, v39, vcc_lo
+; GFX11-TRUE16-NEXT: v_bfi_b32 v26, 0xffff, v34, v85
+; GFX11-TRUE16-NEXT: v_cndmask_b32_e32 v87, v25, v39, vcc_lo
; GFX11-TRUE16-NEXT: v_bfi_b32 v25, 0xffff, v37, v36
; GFX11-TRUE16-NEXT: v_and_b32_e32 v36, 0xffff0000, v30
-; GFX11-TRUE16-NEXT: v_lshlrev_b32_e32 v27, 16, v27
-; GFX11-TRUE16-NEXT: v_lshlrev_b32_e32 v30, 16, v30
-; GFX11-TRUE16-NEXT: v_mov_b16_e32 v28.l, v99.h
-; GFX11-TRUE16-NEXT: v_lshrrev_b32_e32 v148, 24, v26
+; GFX11-TRUE16-NEXT: v_dual_add_f32 v27, 0x40c00000, v27 :: v_dual_lshlrev_b32 v30, 16, v30
+; GFX11-TRUE16-NEXT: v_add_f32_e32 v33, 0x40c00000, v38
+; GFX11-TRUE16-NEXT: v_mov_b16_e32 v28.l, v101.h
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_4) | instskip(NEXT) | instid1(VALU_DEP_4)
-; GFX11-TRUE16-NEXT: v_dual_add_f32 v36, 0x40c00000, v36 :: v_dual_add_f32 v27, 0x40c00000, v27
+; GFX11-TRUE16-NEXT: v_add_f32_e32 v36, 0x40c00000, v36
; GFX11-TRUE16-NEXT: v_add_f32_e32 v30, 0x40c00000, v30
-; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_4) | instskip(SKIP_1) | instid1(VALU_DEP_4)
-; GFX11-TRUE16-NEXT: v_bfi_b32 v28, 0xffff, v28, v100
-; GFX11-TRUE16-NEXT: v_lshrrev_b32_e32 v150, 8, v26
-; GFX11-TRUE16-NEXT: v_or_b32_e32 v39, 0x400000, v36
; GFX11-TRUE16-NEXT: v_bfe_u32 v34, v27, 16, 1
+; GFX11-TRUE16-NEXT: v_bfe_u32 v35, v33, 16, 1
; GFX11-TRUE16-NEXT: v_or_b32_e32 v37, 0x400000, v27
; GFX11-TRUE16-NEXT: v_cmp_u_f32_e32 vcc_lo, v27, v27
-; GFX11-TRUE16-NEXT: v_lshrrev_b32_e32 v133, 24, v28
-; GFX11-TRUE16-NEXT: v_lshrrev_b32_e32 v134, 8, v28
-; GFX11-TRUE16-NEXT: v_add3_u32 v34, v34, v27, 0x7fff
-; GFX11-TRUE16-NEXT: v_add_f32_e32 v33, 0x40c00000, v38
-; GFX11-TRUE16-NEXT: v_lshrrev_b32_e32 v161, 8, v25
-; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_3) | instskip(NEXT) | instid1(VALU_DEP_3)
-; GFX11-TRUE16-NEXT: v_cndmask_b32_e32 v112, v34, v37, vcc_lo
-; GFX11-TRUE16-NEXT: v_bfe_u32 v35, v33, 16, 1
; GFX11-TRUE16-NEXT: v_or_b32_e32 v38, 0x400000, v33
-; GFX11-TRUE16-NEXT: v_cmp_u_f32_e32 vcc_lo, v33, v33
-; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_3) | instskip(SKIP_1) | instid1(VALU_DEP_2)
+; GFX11-TRUE16-NEXT: v_add3_u32 v34, v34, v27, 0x7fff
; GFX11-TRUE16-NEXT: v_add3_u32 v35, v35, v33, 0x7fff
+; GFX11-TRUE16-NEXT: v_or_b32_e32 v39, 0x400000, v36
+; GFX11-TRUE16-NEXT: v_bfi_b32 v28, 0xffff, v28, v87
+; GFX11-TRUE16-NEXT: v_lshrrev_b32_e32 v162, 24, v26
+; GFX11-TRUE16-NEXT: v_cndmask_b32_e32 v113, v34, v37, vcc_lo
+; GFX11-TRUE16-NEXT: v_cmp_u_f32_e32 vcc_lo, v33, v33
+; GFX11-TRUE16-NEXT: v_bfe_u32 v34, v30, 16, 1
; GFX11-TRUE16-NEXT: v_bfe_u32 v33, v36, 16, 1
+; GFX11-TRUE16-NEXT: v_lshrrev_b32_e32 v144, 24, v28
+; GFX11-TRUE16-NEXT: v_lshrrev_b32_e32 v146, 8, v28
; GFX11-TRUE16-NEXT: v_cndmask_b32_e32 v27, v35, v38, vcc_lo
-; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_2)
-; GFX11-TRUE16-NEXT: v_add3_u32 v33, v33, v36, 0x7fff
-; GFX11-TRUE16-NEXT: v_and_b32_e32 v37, 0xffff0000, v29
-; GFX11-TRUE16-NEXT: v_lshlrev_b32_e32 v29, 16, v29
-; GFX11-TRUE16-NEXT: v_bfe_u32 v34, v30, 16, 1
+; GFX11-TRUE16-NEXT: v_add3_u32 v34, v34, v30, 0x7fff
; GFX11-TRUE16-NEXT: v_or_b32_e32 v38, 0x400000, v30
; GFX11-TRUE16-NEXT: v_cmp_u_f32_e32 vcc_lo, v30, v30
-; GFX11-TRUE16-NEXT: v_mov_b16_e32 v35.l, v112.h
-; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_4) | instskip(NEXT) | instid1(VALU_DEP_2)
-; GFX11-TRUE16-NEXT: v_add3_u32 v34, v34, v30, 0x7fff
-; GFX11-TRUE16-NEXT: v_bfi_b32 v27, 0xffff, v35, v27
+; GFX11-TRUE16-NEXT: v_and_b32_e32 v37, 0xffff0000, v29
+; GFX11-TRUE16-NEXT: v_add3_u32 v33, v33, v36, 0x7fff
+; GFX11-TRUE16-NEXT: v_lshrrev_b32_e32 v164, 8, v26
+; GFX11-TRUE16-NEXT: v_lshrrev_b32_e32 v177, 8, v25
+; GFX11-TRUE16-NEXT: v_cndmask_b32_e32 v112, v34, v38, vcc_lo
; GFX11-TRUE16-NEXT: s_waitcnt vmcnt(1)
-; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_2) | instskip(SKIP_4) | instid1(VALU_DEP_4)
-; GFX11-TRUE16-NEXT: v_dual_cndmask_b32 v103, v34, v38 :: v_dual_and_b32 v38, 0xffff0000, v32
-; GFX11-TRUE16-NEXT: v_add_f32_e32 v29, 0x40c00000, v29
+; GFX11-TRUE16-NEXT: v_and_b32_e32 v38, 0xffff0000, v32
+; GFX11-TRUE16-NEXT: v_lshlrev_b32_e32 v29, 16, v29
+; GFX11-TRUE16-NEXT: v_dual_add_f32 v37, 0x40c00000, v37 :: v_dual_lshlrev_b32 v32, 16, v32
; GFX11-TRUE16-NEXT: v_cmp_u_f32_e32 vcc_lo, v36, v36
-; GFX11-TRUE16-NEXT: v_lshlrev_b32_e32 v32, 16, v32
-; GFX11-TRUE16-NEXT: v_lshrrev_b32_e32 v145, 8, v27
+; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_3) | instskip(NEXT) | instid1(VALU_DEP_3)
+; GFX11-TRUE16-NEXT: v_add_f32_e32 v29, 0x40c00000, v29
+; GFX11-TRUE16-NEXT: v_bfe_u32 v30, v37, 16, 1
+; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_4)
+; GFX11-TRUE16-NEXT: v_add_f32_e32 v32, 0x40c00000, v32
+; GFX11-TRUE16-NEXT: v_cndmask_b32_e32 v98, v33, v39, vcc_lo
+; GFX11-TRUE16-NEXT: v_mov_b16_e32 v33.l, v112.h
; GFX11-TRUE16-NEXT: v_bfe_u32 v34, v29, 16, 1
-; GFX11-TRUE16-NEXT: v_dual_cndmask_b32 v102, v33, v39 :: v_dual_add_f32 v37, 0x40c00000, v37
; GFX11-TRUE16-NEXT: v_or_b32_e32 v36, 0x400000, v29
; GFX11-TRUE16-NEXT: v_cmp_u_f32_e32 vcc_lo, v29, v29
-; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_4)
-; GFX11-TRUE16-NEXT: v_add3_u32 v34, v34, v29, 0x7fff
-; GFX11-TRUE16-NEXT: v_mov_b16_e32 v33.l, v103.h
-; GFX11-TRUE16-NEXT: v_bfe_u32 v30, v37, 16, 1
+; GFX11-TRUE16-NEXT: v_add3_u32 v30, v30, v37, 0x7fff
; GFX11-TRUE16-NEXT: v_or_b32_e32 v39, 0x400000, v37
-; GFX11-TRUE16-NEXT: v_add_f32_e32 v32, 0x40c00000, v32
-; GFX11-TRUE16-NEXT: v_cndmask_b32_e32 v113, v34, v36, vcc_lo
+; GFX11-TRUE16-NEXT: v_add3_u32 v34, v34, v29, 0x7fff
+; GFX11-TRUE16-NEXT: v_mov_b16_e32 v35.l, v113.h
+; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_2) | instskip(SKIP_1) | instid1(VALU_DEP_3)
+; GFX11-TRUE16-NEXT: v_dual_cndmask_b32 v115, v34, v36 :: v_dual_add_f32 v34, 0x40c00000, v38
; GFX11-TRUE16-NEXT: v_cmp_u_f32_e32 vcc_lo, v37, v37
-; GFX11-TRUE16-NEXT: v_add3_u32 v30, v30, v37, 0x7fff
-; GFX11-TRUE16-NEXT: v_add_f32_e32 v34, 0x40c00000, v38
-; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_4) | instskip(NEXT) | instid1(VALU_DEP_3)
-; GFX11-TRUE16-NEXT: v_mov_b16_e32 v36.l, v113.h
-; GFX11-TRUE16-NEXT: v_cndmask_b32_e32 v29, v30, v39, vcc_lo
-; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_3) | instskip(SKIP_2) | instid1(VALU_DEP_4)
+; GFX11-TRUE16-NEXT: v_bfi_b32 v27, 0xffff, v35, v27
+; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_3) | instskip(NEXT) | instid1(VALU_DEP_4)
+; GFX11-TRUE16-NEXT: v_mov_b16_e32 v36.l, v115.h
; GFX11-TRUE16-NEXT: v_bfe_u32 v37, v34, 16, 1
-; GFX11-TRUE16-NEXT: v_bfi_b32 v30, 0xffff, v33, v102
-; GFX11-TRUE16-NEXT: v_or_b32_e32 v38, 0x400000, v34
+; GFX11-TRUE16-NEXT: v_cndmask_b32_e32 v29, v30, v39, vcc_lo
+; GFX11-TRUE16-NEXT: v_bfi_b32 v30, 0xffff, v33, v98
+; GFX11-TRUE16-NEXT: v_bfe_u32 v33, v32, 16, 1
+; GFX11-TRUE16-NEXT: v_cmp_u_f32_e32 vcc_lo, v32, v32
+; GFX11-TRUE16-NEXT: v_add3_u32 v35, v37, v34, 0x7fff
; GFX11-TRUE16-NEXT: v_bfi_b32 v29, 0xffff, v36, v29
; GFX11-TRUE16-NEXT: s_waitcnt vmcnt(0)
; GFX11-TRUE16-NEXT: v_and_b32_e32 v36, 0xffff0000, v31
-; GFX11-TRUE16-NEXT: v_lshlrev_b32_e32 v31, 16, v31
-; GFX11-TRUE16-NEXT: v_bfe_u32 v33, v32, 16, 1
-; GFX11-TRUE16-NEXT: v_add3_u32 v35, v37, v34, 0x7fff
-; GFX11-TRUE16-NEXT: v_or_b32_e32 v37, 0x400000, v32
-; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_4) | instskip(NEXT) | instid1(VALU_DEP_4)
-; GFX11-TRUE16-NEXT: v_dual_add_f32 v36, 0x40c00000, v36 :: v_dual_add_f32 v31, 0x40c00000, v31
; GFX11-TRUE16-NEXT: v_add3_u32 v33, v33, v32, 0x7fff
-; GFX11-TRUE16-NEXT: v_cmp_u_f32_e32 vcc_lo, v32, v32
-; GFX11-TRUE16-NEXT: v_lshrrev_b32_e32 v128, 24, v30
-; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_4)
-; GFX11-TRUE16-NEXT: v_bfe_u32 v32, v36, 16, 1
-; GFX11-TRUE16-NEXT: v_lshrrev_b32_e32 v129, 8, v30
-; GFX11-TRUE16-NEXT: v_lshrrev_b32_e32 v130, 8, v29
+; GFX11-TRUE16-NEXT: v_or_b32_e32 v37, 0x400000, v32
+; GFX11-TRUE16-NEXT: v_lshlrev_b32_e32 v31, 16, v31
+; GFX11-TRUE16-NEXT: v_or_b32_e32 v38, 0x400000, v34
+; GFX11-TRUE16-NEXT: v_add_f32_e32 v36, 0x40c00000, v36
+; GFX11-TRUE16-NEXT: v_lshrrev_b32_e32 v130, 24, v30
; GFX11-TRUE16-NEXT: v_cndmask_b32_e32 v114, v33, v37, vcc_lo
; GFX11-TRUE16-NEXT: v_cmp_u_f32_e32 vcc_lo, v34, v34
+; GFX11-TRUE16-NEXT: v_add_f32_e32 v31, 0x40c00000, v31
+; GFX11-TRUE16-NEXT: v_bfe_u32 v32, v36, 16, 1
+; GFX11-TRUE16-NEXT: v_lshrrev_b32_e32 v133, 8, v30
+; GFX11-TRUE16-NEXT: v_mov_b16_e32 v34.l, v114.h
+; GFX11-TRUE16-NEXT: v_cndmask_b32_e32 v102, v35, v38, vcc_lo
; GFX11-TRUE16-NEXT: v_bfe_u32 v33, v31, 16, 1
; GFX11-TRUE16-NEXT: v_or_b32_e32 v37, 0x400000, v31
+; GFX11-TRUE16-NEXT: v_cmp_u_f32_e32 vcc_lo, v31, v31
; GFX11-TRUE16-NEXT: v_add3_u32 v32, v32, v36, 0x7fff
-; GFX11-TRUE16-NEXT: v_cndmask_b32_e32 v115, v35, v38, vcc_lo
-; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_4)
+; GFX11-TRUE16-NEXT: v_or_b32_e32 v38, 0x400000, v36
; GFX11-TRUE16-NEXT: v_add3_u32 v33, v33, v31, 0x7fff
-; GFX11-TRUE16-NEXT: v_cmp_u_f32_e32 vcc_lo, v31, v31
; GFX11-TRUE16-NEXT: v_and_b32_e32 v35, 0xffff0000, v2
-; GFX11-TRUE16-NEXT: v_or_b32_e32 v38, 0x400000, v36
-; GFX11-TRUE16-NEXT: v_mov_b16_e32 v34.l, v114.h
+; GFX11-TRUE16-NEXT: v_lshrrev_b32_e32 v135, 8, v29
+; GFX11-TRUE16-NEXT: v_lshrrev_b32_e32 v151, 8, v27
+; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_4) | instskip(NEXT) | instid1(VALU_DEP_4)
; GFX11-TRUE16-NEXT: v_cndmask_b32_e32 v116, v33, v37, vcc_lo
-; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_4) | instskip(SKIP_1) | instid1(VALU_DEP_3)
; GFX11-TRUE16-NEXT: v_dual_add_f32 v35, 0x40c00000, v35 :: v_dual_lshlrev_b32 v2, 16, v2
; GFX11-TRUE16-NEXT: v_cmp_u_f32_e32 vcc_lo, v36, v36
+; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_3) | instskip(NEXT) | instid1(VALU_DEP_3)
; GFX11-TRUE16-NEXT: v_mov_b16_e32 v37.l, v116.h
-; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_3) | instskip(NEXT) | instid1(VALU_DEP_4)
; GFX11-TRUE16-NEXT: v_add_f32_e32 v2, 0x40c00000, v2
+; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_4) | instskip(SKIP_1) | instid1(VALU_DEP_2)
; GFX11-TRUE16-NEXT: v_bfe_u32 v31, v35, 16, 1
+; GFX11-TRUE16-NEXT: v_or_b32_e32 v39, 0x400000, v35
+; GFX11-TRUE16-NEXT: v_add3_u32 v31, v31, v35, 0x7fff
; GFX11-TRUE16-NEXT: v_cndmask_b32_e32 v36, v32, v38, vcc_lo
; GFX11-TRUE16-NEXT: v_and_b32_e32 v38, 0xffff0000, v1
; GFX11-TRUE16-NEXT: v_lshlrev_b32_e32 v1, 16, v1
; GFX11-TRUE16-NEXT: v_bfe_u32 v33, v2, 16, 1
; GFX11-TRUE16-NEXT: v_cmp_u_f32_e32 vcc_lo, v2, v2
-; GFX11-TRUE16-NEXT: v_add3_u32 v31, v31, v35, 0x7fff
-; GFX11-TRUE16-NEXT: v_or_b32_e32 v39, 0x400000, v35
+; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_3) | instskip(NEXT) | instid1(VALU_DEP_3)
; GFX11-TRUE16-NEXT: v_add_f32_e32 v1, 0x40c00000, v1
; GFX11-TRUE16-NEXT: v_add3_u32 v32, v33, v2, 0x7fff
; GFX11-TRUE16-NEXT: v_or_b32_e32 v33, 0x400000, v2
-; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1)
-; GFX11-TRUE16-NEXT: v_cndmask_b32_e32 v131, v32, v33, vcc_lo
-; GFX11-TRUE16-NEXT: v_add_f32_e32 v33, 0x40c00000, v38
+; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(SKIP_2) | instid1(VALU_DEP_3)
+; GFX11-TRUE16-NEXT: v_dual_cndmask_b32 v132, v32, v33 :: v_dual_add_f32 v33, 0x40c00000, v38
; GFX11-TRUE16-NEXT: v_cmp_u_f32_e32 vcc_lo, v35, v35
-; GFX11-TRUE16-NEXT: v_bfi_b32 v32, 0xffff, v34, v115
-; GFX11-TRUE16-NEXT: v_bfe_u32 v34, v1, 16, 1
-; GFX11-TRUE16-NEXT: v_cndmask_b32_e32 v132, v31, v39, vcc_lo
+; GFX11-TRUE16-NEXT: v_bfi_b32 v32, 0xffff, v34, v102
+; GFX11-TRUE16-NEXT: v_mov_b16_e64 v2.l, v132.h
+; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_4)
+; GFX11-TRUE16-NEXT: v_bfe_u32 v35, v33, 16, 1
+; GFX11-TRUE16-NEXT: v_cndmask_b32_e32 v117, v31, v39, vcc_lo
; GFX11-TRUE16-NEXT: v_bfi_b32 v31, 0xffff, v37, v36
+; GFX11-TRUE16-NEXT: v_or_b32_e32 v38, 0x400000, v33
+; GFX11-TRUE16-NEXT: v_lshrrev_b32_e32 v118, 24, v32
+; GFX11-TRUE16-NEXT: v_add3_u32 v35, v35, v33, 0x7fff
; GFX11-TRUE16-NEXT: v_and_b32_e32 v36, 0xffff0000, v4
; GFX11-TRUE16-NEXT: v_lshlrev_b32_e32 v4, 16, v4
-; GFX11-TRUE16-NEXT: v_bfe_u32 v35, v33, 16, 1
-; GFX11-TRUE16-NEXT: v_add3_u32 v34, v34, v1, 0x7fff
+; GFX11-TRUE16-NEXT: v_bfe_u32 v34, v1, 16, 1
; GFX11-TRUE16-NEXT: v_or_b32_e32 v37, 0x400000, v1
; GFX11-TRUE16-NEXT: v_cmp_u_f32_e32 vcc_lo, v1, v1
-; GFX11-TRUE16-NEXT: v_or_b32_e32 v38, 0x400000, v33
-; GFX11-TRUE16-NEXT: v_add3_u32 v35, v35, v33, 0x7fff
-; GFX11-TRUE16-NEXT: v_mov_b16_e64 v2.l, v131.h
-; GFX11-TRUE16-NEXT: v_lshrrev_b32_e32 v117, 24, v32
-; GFX11-TRUE16-NEXT: v_dual_cndmask_b32 v144, v34, v37 :: v_dual_and_b32 v37, 0xffff0000, v3
+; GFX11-TRUE16-NEXT: v_bfi_b32 v2, 0xffff, v2, v117
+; GFX11-TRUE16-NEXT: v_lshrrev_b32_e32 v119, 8, v32
+; GFX11-TRUE16-NEXT: v_add3_u32 v34, v34, v1, 0x7fff
+; GFX11-TRUE16-NEXT: v_lshrrev_b32_e32 v129, 8, v31
+; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_4) | instskip(SKIP_1) | instid1(VALU_DEP_4)
+; GFX11-TRUE16-NEXT: v_lshrrev_b32_e32 v106, 24, v2
+; GFX11-TRUE16-NEXT: v_lshrrev_b32_e32 v107, 8, v2
+; GFX11-TRUE16-NEXT: v_dual_cndmask_b32 v148, v34, v37 :: v_dual_and_b32 v37, 0xffff0000, v3
; GFX11-TRUE16-NEXT: v_add_f32_e32 v4, 0x40c00000, v4
; GFX11-TRUE16-NEXT: v_cmp_u_f32_e32 vcc_lo, v33, v33
; GFX11-TRUE16-NEXT: v_lshlrev_b32_e32 v3, 16, v3
-; GFX11-TRUE16-NEXT: v_bfi_b32 v2, 0xffff, v2, v132
+; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_4) | instskip(NEXT) | instid1(VALU_DEP_4)
; GFX11-TRUE16-NEXT: v_add_f32_e32 v37, 0x40c00000, v37
; GFX11-TRUE16-NEXT: v_bfe_u32 v34, v4, 16, 1
; GFX11-TRUE16-NEXT: v_dual_cndmask_b32 v1, v35, v38 :: v_dual_add_f32 v36, 0x40c00000, v36
@@ -168342,179 +168428,170 @@ define <128 x i8> @bitcast_v64bf16_to_v128i8(<64 x bfloat> %a, i32 %b) {
; GFX11-TRUE16-NEXT: v_bfe_u32 v33, v36, 16, 1
; GFX11-TRUE16-NEXT: v_or_b32_e32 v39, 0x400000, v36
; GFX11-TRUE16-NEXT: v_bfe_u32 v4, v37, 16, 1
+; GFX11-TRUE16-NEXT: v_mov_b16_e64 v35.l, v148.h
+; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_4)
+; GFX11-TRUE16-NEXT: v_add3_u32 v33, v33, v36, 0x7fff
; GFX11-TRUE16-NEXT: v_cndmask_b32_e32 v147, v34, v38, vcc_lo
; GFX11-TRUE16-NEXT: v_cmp_u_f32_e32 vcc_lo, v36, v36
-; GFX11-TRUE16-NEXT: v_add3_u32 v33, v33, v36, 0x7fff
-; GFX11-TRUE16-NEXT: v_mov_b16_e64 v35.l, v144.h
; GFX11-TRUE16-NEXT: v_add3_u32 v4, v4, v37, 0x7fff
-; GFX11-TRUE16-NEXT: v_lshrrev_b32_e32 v106, 24, v2
-; GFX11-TRUE16-NEXT: v_lshrrev_b32_e32 v107, 8, v2
-; GFX11-TRUE16-NEXT: v_cndmask_b32_e32 v149, v33, v39, vcc_lo
+; GFX11-TRUE16-NEXT: v_bfi_b32 v1, 0xffff, v35, v1
+; GFX11-TRUE16-NEXT: v_cndmask_b32_e32 v128, v33, v39, vcc_lo
; GFX11-TRUE16-NEXT: v_or_b32_e32 v39, 0x400000, v37
; GFX11-TRUE16-NEXT: v_and_b32_e32 v38, 0xffff0000, v6
; GFX11-TRUE16-NEXT: v_lshlrev_b32_e32 v6, 16, v6
; GFX11-TRUE16-NEXT: v_bfe_u32 v34, v3, 16, 1
; GFX11-TRUE16-NEXT: v_or_b32_e32 v36, 0x400000, v3
; GFX11-TRUE16-NEXT: v_cmp_u_f32_e32 vcc_lo, v3, v3
-; GFX11-TRUE16-NEXT: v_bfi_b32 v1, 0xffff, v35, v1
-; GFX11-TRUE16-NEXT: v_add_f32_e32 v6, 0x40c00000, v6
-; GFX11-TRUE16-NEXT: v_add3_u32 v34, v34, v3, 0x7fff
-; GFX11-TRUE16-NEXT: v_lshrrev_b32_e32 v118, 8, v32
-; GFX11-TRUE16-NEXT: v_lshrrev_b32_e32 v119, 8, v31
+; GFX11-TRUE16-NEXT: v_mov_b16_e64 v33.l, v147.h
; GFX11-TRUE16-NEXT: v_lshrrev_b32_e32 v108, 8, v1
-; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_4) | instskip(SKIP_1) | instid1(VALU_DEP_2)
-; GFX11-TRUE16-NEXT: v_cndmask_b32_e32 v162, v34, v36, vcc_lo
+; GFX11-TRUE16-NEXT: v_add3_u32 v34, v34, v3, 0x7fff
+; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(SKIP_1) | instid1(VALU_DEP_2)
+; GFX11-TRUE16-NEXT: v_cndmask_b32_e32 v163, v34, v36, vcc_lo
; GFX11-TRUE16-NEXT: v_cmp_u_f32_e32 vcc_lo, v37, v37
-; GFX11-TRUE16-NEXT: v_mov_b16_e64 v36.l, v162.h
-; GFX11-TRUE16-NEXT: v_cndmask_b32_e32 v3, v4, v39, vcc_lo
-; GFX11-TRUE16-NEXT: v_cmp_u_f32_e32 vcc_lo, v6, v6
+; GFX11-TRUE16-NEXT: v_mov_b16_e64 v36.l, v163.h
+; GFX11-TRUE16-NEXT: v_dual_cndmask_b32 v3, v4, v39 :: v_dual_add_f32 v34, 0x40c00000, v38
+; GFX11-TRUE16-NEXT: v_bfi_b32 v4, 0xffff, v33, v128
; GFX11-TRUE16-NEXT: v_and_b32_e32 v39, 0xffff0000, v7
; GFX11-TRUE16-NEXT: v_lshlrev_b32_e32 v7, 16, v7
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_4)
; GFX11-TRUE16-NEXT: v_bfi_b32 v3, 0xffff, v36, v3
; GFX11-TRUE16-NEXT: v_and_b32_e32 v36, 0xffff0000, v5
-; GFX11-TRUE16-NEXT: v_mov_b16_e64 v33.l, v147.h
-; GFX11-TRUE16-NEXT: v_dual_add_f32 v34, 0x40c00000, v38 :: v_dual_lshlrev_b32 v5, 16, v5
-; GFX11-TRUE16-NEXT: v_add_f32_e32 v7, 0x40c00000, v7
-; GFX11-TRUE16-NEXT: v_lshrrev_b32_e32 v105, 8, v3
-; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_4)
-; GFX11-TRUE16-NEXT: v_bfi_b32 v4, 0xffff, v33, v149
-; GFX11-TRUE16-NEXT: v_bfe_u32 v33, v6, 16, 1
-; GFX11-TRUE16-NEXT: v_add_f32_e32 v5, 0x40c00000, v5
+; GFX11-TRUE16-NEXT: v_dual_add_f32 v6, 0x40c00000, v6 :: v_dual_lshlrev_b32 v5, 16, v5
; GFX11-TRUE16-NEXT: v_bfe_u32 v37, v34, 16, 1
; GFX11-TRUE16-NEXT: v_or_b32_e32 v38, 0x400000, v34
-; GFX11-TRUE16-NEXT: v_lshrrev_b32_e32 v93, 24, v4
-; GFX11-TRUE16-NEXT: v_add3_u32 v33, v33, v6, 0x7fff
-; GFX11-TRUE16-NEXT: v_lshrrev_b32_e32 v95, 8, v4
+; GFX11-TRUE16-NEXT: v_lshrrev_b32_e32 v95, 24, v4
+; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_4)
+; GFX11-TRUE16-NEXT: v_add_f32_e32 v5, 0x40c00000, v5
+; GFX11-TRUE16-NEXT: v_bfe_u32 v33, v6, 16, 1
; GFX11-TRUE16-NEXT: v_add3_u32 v35, v37, v34, 0x7fff
; GFX11-TRUE16-NEXT: v_or_b32_e32 v37, 0x400000, v6
-; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1)
-; GFX11-TRUE16-NEXT: v_cndmask_b32_e32 v164, v33, v37, vcc_lo
+; GFX11-TRUE16-NEXT: v_cmp_u_f32_e32 vcc_lo, v6, v6
+; GFX11-TRUE16-NEXT: v_lshrrev_b32_e32 v104, 8, v4
+; GFX11-TRUE16-NEXT: v_add3_u32 v33, v33, v6, 0x7fff
+; GFX11-TRUE16-NEXT: v_lshrrev_b32_e32 v105, 8, v3
+; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_2)
+; GFX11-TRUE16-NEXT: v_cndmask_b32_e32 v160, v33, v37, vcc_lo
; GFX11-TRUE16-NEXT: v_cmp_u_f32_e32 vcc_lo, v34, v34
-; GFX11-TRUE16-NEXT: v_add_f32_e32 v36, 0x40c00000, v36
; GFX11-TRUE16-NEXT: v_bfe_u32 v33, v5, 16, 1
; GFX11-TRUE16-NEXT: v_or_b32_e32 v37, 0x400000, v5
-; GFX11-TRUE16-NEXT: v_cndmask_b32_e32 v165, v35, v38, vcc_lo
+; GFX11-TRUE16-NEXT: v_cndmask_b32_e32 v131, v35, v38, vcc_lo
; GFX11-TRUE16-NEXT: v_and_b32_e32 v35, 0xffff0000, v8
-; GFX11-TRUE16-NEXT: v_bfe_u32 v6, v36, 16, 1
+; GFX11-TRUE16-NEXT: v_add_f32_e32 v36, 0x40c00000, v36
; GFX11-TRUE16-NEXT: v_add3_u32 v33, v33, v5, 0x7fff
; GFX11-TRUE16-NEXT: v_cmp_u_f32_e32 vcc_lo, v5, v5
+; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_4) | instskip(NEXT) | instid1(VALU_DEP_4)
+; GFX11-TRUE16-NEXT: v_dual_add_f32 v35, 0x40c00000, v35 :: v_dual_lshlrev_b32 v8, 16, v8
+; GFX11-TRUE16-NEXT: v_bfe_u32 v6, v36, 16, 1
; GFX11-TRUE16-NEXT: v_or_b32_e32 v38, 0x400000, v36
-; GFX11-TRUE16-NEXT: v_add_f32_e32 v35, 0x40c00000, v35
-; GFX11-TRUE16-NEXT: v_add3_u32 v6, v6, v36, 0x7fff
-; GFX11-TRUE16-NEXT: v_mov_b16_e64 v34.l, v164.h
-; GFX11-TRUE16-NEXT: v_cndmask_b32_e32 v180, v33, v37, vcc_lo
+; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_3) | instskip(SKIP_1) | instid1(VALU_DEP_4)
+; GFX11-TRUE16-NEXT: v_dual_cndmask_b32 v179, v33, v37 :: v_dual_add_f32 v8, 0x40c00000, v8
; GFX11-TRUE16-NEXT: v_cmp_u_f32_e32 vcc_lo, v36, v36
+; GFX11-TRUE16-NEXT: v_add3_u32 v6, v6, v36, 0x7fff
; GFX11-TRUE16-NEXT: v_bfe_u32 v5, v35, 16, 1
-; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_3) | instskip(SKIP_1) | instid1(VALU_DEP_3)
-; GFX11-TRUE16-NEXT: v_mov_b16_e64 v37.l, v180.h
-; GFX11-TRUE16-NEXT: v_cndmask_b32_e32 v36, v6, v38, vcc_lo
-; GFX11-TRUE16-NEXT: v_add3_u32 v5, v5, v35, 0x7fff
-; GFX11-TRUE16-NEXT: v_or_b32_e32 v38, 0x400000, v35
-; GFX11-TRUE16-NEXT: v_lshlrev_b32_e32 v8, 16, v8
-; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
-; GFX11-TRUE16-NEXT: v_add_f32_e32 v8, 0x40c00000, v8
+; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_4)
+; GFX11-TRUE16-NEXT: v_mov_b16_e64 v37.l, v179.h
; GFX11-TRUE16-NEXT: v_bfe_u32 v33, v8, 16, 1
+; GFX11-TRUE16-NEXT: v_mov_b16_e64 v34.l, v160.h
+; GFX11-TRUE16-NEXT: v_cndmask_b32_e32 v36, v6, v38, vcc_lo
; GFX11-TRUE16-NEXT: v_cmp_u_f32_e32 vcc_lo, v8, v8
-; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_2) | instskip(SKIP_1) | instid1(VALU_DEP_1)
+; GFX11-TRUE16-NEXT: v_add3_u32 v5, v5, v35, 0x7fff
; GFX11-TRUE16-NEXT: v_add3_u32 v6, v33, v8, 0x7fff
; GFX11-TRUE16-NEXT: v_or_b32_e32 v33, 0x400000, v8
-; GFX11-TRUE16-NEXT: v_cndmask_b32_e32 v178, v6, v33, vcc_lo
+; GFX11-TRUE16-NEXT: v_or_b32_e32 v38, 0x400000, v35
+; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_2)
+; GFX11-TRUE16-NEXT: v_cndmask_b32_e32 v167, v6, v33, vcc_lo
; GFX11-TRUE16-NEXT: v_cmp_u_f32_e32 vcc_lo, v35, v35
; GFX11-TRUE16-NEXT: v_add_f32_e32 v33, 0x40c00000, v39
-; GFX11-TRUE16-NEXT: v_bfe_u32 v35, v7, 16, 1
-; GFX11-TRUE16-NEXT: v_bfi_b32 v6, 0xffff, v34, v165
-; GFX11-TRUE16-NEXT: v_and_b32_e32 v39, 0xffff0000, v9
-; GFX11-TRUE16-NEXT: v_cndmask_b32_e32 v179, v5, v38, vcc_lo
+; GFX11-TRUE16-NEXT: v_bfi_b32 v6, 0xffff, v34, v131
+; GFX11-TRUE16-NEXT: v_dual_cndmask_b32 v134, v5, v38 :: v_dual_and_b32 v39, 0xffff0000, v9
; GFX11-TRUE16-NEXT: v_bfi_b32 v5, 0xffff, v37, v36
+; GFX11-TRUE16-NEXT: v_dual_add_f32 v7, 0x40c00000, v7 :: v_dual_lshlrev_b32 v36, 16, v10
; GFX11-TRUE16-NEXT: v_bfe_u32 v34, v33, 16, 1
-; GFX11-TRUE16-NEXT: v_lshlrev_b32_e32 v36, 16, v10
-; GFX11-TRUE16-NEXT: v_add3_u32 v35, v35, v7, 0x7fff
+; GFX11-TRUE16-NEXT: v_or_b32_e32 v38, 0x400000, v33
+; GFX11-TRUE16-NEXT: v_mov_b16_e64 v8.l, v167.h
+; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_4)
+; GFX11-TRUE16-NEXT: v_add_f32_e32 v36, 0x40c00000, v36
+; GFX11-TRUE16-NEXT: v_bfe_u32 v35, v7, 16, 1
; GFX11-TRUE16-NEXT: v_or_b32_e32 v37, 0x400000, v7
; GFX11-TRUE16-NEXT: v_cmp_u_f32_e32 vcc_lo, v7, v7
; GFX11-TRUE16-NEXT: v_add3_u32 v34, v34, v33, 0x7fff
-; GFX11-TRUE16-NEXT: v_or_b32_e32 v38, 0x400000, v33
-; GFX11-TRUE16-NEXT: v_mov_b16_e64 v8.l, v178.h
-; GFX11-TRUE16-NEXT: v_lshrrev_b32_e32 v78, 24, v6
-; GFX11-TRUE16-NEXT: v_cndmask_b32_e32 v46, v35, v37, vcc_lo
-; GFX11-TRUE16-NEXT: v_cmp_u_f32_e32 vcc_lo, v33, v33
-; GFX11-TRUE16-NEXT: v_add_f32_e32 v36, 0x40c00000, v36
-; GFX11-TRUE16-NEXT: v_bfi_b32 v8, 0xffff, v8, v179
-; GFX11-TRUE16-NEXT: v_lshrrev_b32_e32 v88, 8, v6
-; GFX11-TRUE16-NEXT: v_lshrrev_b32_e32 v90, 8, v5
-; GFX11-TRUE16-NEXT: v_cndmask_b32_e32 v33, v34, v38, vcc_lo
+; GFX11-TRUE16-NEXT: v_bfi_b32 v8, 0xffff, v8, v134
+; GFX11-TRUE16-NEXT: v_add3_u32 v35, v35, v7, 0x7fff
; GFX11-TRUE16-NEXT: v_bfe_u32 v7, v36, 16, 1
+; GFX11-TRUE16-NEXT: v_lshrrev_b32_e32 v91, 24, v6
+; GFX11-TRUE16-NEXT: v_lshrrev_b32_e32 v93, 8, v6
+; GFX11-TRUE16-NEXT: v_lshrrev_b32_e32 v76, 24, v8
+; GFX11-TRUE16-NEXT: v_cndmask_b32_e32 v42, v35, v37, vcc_lo
+; GFX11-TRUE16-NEXT: v_cmp_u_f32_e32 vcc_lo, v33, v33
+; GFX11-TRUE16-NEXT: v_add3_u32 v7, v7, v36, 0x7fff
; GFX11-TRUE16-NEXT: v_or_b32_e32 v37, 0x400000, v36
+; GFX11-TRUE16-NEXT: v_lshrrev_b32_e32 v77, 8, v8
+; GFX11-TRUE16-NEXT: v_lshrrev_b32_e32 v94, 8, v5
+; GFX11-TRUE16-NEXT: v_cndmask_b32_e32 v33, v34, v38, vcc_lo
; GFX11-TRUE16-NEXT: v_cmp_u_f32_e32 vcc_lo, v36, v36
-; GFX11-TRUE16-NEXT: v_lshrrev_b32_e32 v62, 24, v8
-; GFX11-TRUE16-NEXT: v_lshrrev_b32_e32 v63, 8, v8
-; GFX11-TRUE16-NEXT: v_add3_u32 v7, v7, v36, 0x7fff
-; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(SKIP_3) | instid1(VALU_DEP_3)
-; GFX11-TRUE16-NEXT: v_cndmask_b32_e32 v45, v7, v37, vcc_lo
+; GFX11-TRUE16-NEXT: v_mov_b16_e32 v34.l, v42.h
+; GFX11-TRUE16-NEXT: v_cndmask_b32_e32 v182, v7, v37, vcc_lo
; GFX11-TRUE16-NEXT: v_lshlrev_b32_e32 v7, 16, v9
; GFX11-TRUE16-NEXT: v_add_f32_e32 v9, 0x40c00000, v39
-; GFX11-TRUE16-NEXT: v_mov_b16_e32 v34.l, v46.h
+; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_2) | instskip(NEXT) | instid1(VALU_DEP_2)
; GFX11-TRUE16-NEXT: v_add_f32_e32 v7, 0x40c00000, v7
-; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_3) | instskip(SKIP_1) | instid1(VALU_DEP_3)
; GFX11-TRUE16-NEXT: v_bfe_u32 v36, v9, 16, 1
; GFX11-TRUE16-NEXT: v_or_b32_e32 v39, 0x400000, v9
+; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_3) | instskip(NEXT) | instid1(VALU_DEP_3)
; GFX11-TRUE16-NEXT: v_or_b32_e32 v49, 0x400000, v7
-; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_3) | instskip(SKIP_1) | instid1(VALU_DEP_1)
; GFX11-TRUE16-NEXT: v_add3_u32 v36, v36, v9, 0x7fff
; GFX11-TRUE16-NEXT: v_and_b32_e32 v10, 0xffff0000, v10
+; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
; GFX11-TRUE16-NEXT: v_add_f32_e32 v10, 0x40c00000, v10
-; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(SKIP_2) | instid1(VALU_DEP_3)
; GFX11-TRUE16-NEXT: v_bfe_u32 v35, v10, 16, 1
; GFX11-TRUE16-NEXT: v_or_b32_e32 v38, 0x400000, v10
; GFX11-TRUE16-NEXT: v_cmp_u_f32_e32 vcc_lo, v10, v10
+; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_3) | instskip(NEXT) | instid1(VALU_DEP_1)
; GFX11-TRUE16-NEXT: v_add3_u32 v35, v35, v10, 0x7fff
-; GFX11-TRUE16-NEXT: v_lshlrev_b32_e32 v10, 16, v12
-; GFX11-TRUE16-NEXT: v_and_b32_e32 v12, 0xffff0000, v12
-; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_3) | instskip(NEXT) | instid1(VALU_DEP_3)
-; GFX11-TRUE16-NEXT: v_cndmask_b32_e32 v42, v35, v38, vcc_lo
+; GFX11-TRUE16-NEXT: v_dual_cndmask_b32 v145, v35, v38 :: v_dual_lshlrev_b32 v10, 16, v12
+; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(SKIP_1) | instid1(VALU_DEP_2)
; GFX11-TRUE16-NEXT: v_add_f32_e32 v37, 0x40c00000, v10
-; GFX11-TRUE16-NEXT: v_mov_b16_e32 v35.l, v45.h
-; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_4) | instskip(NEXT) | instid1(VALU_DEP_3)
-; GFX11-TRUE16-NEXT: v_add_f32_e32 v12, 0x40c00000, v12
+; GFX11-TRUE16-NEXT: v_mov_b16_e64 v35.l, v182.h
; GFX11-TRUE16-NEXT: v_bfe_u32 v38, v37, 16, 1
-; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_3)
-; GFX11-TRUE16-NEXT: v_bfi_b32 v10, 0xffff, v35, v42
-; GFX11-TRUE16-NEXT: v_bfe_u32 v35, v7, 16, 1
; GFX11-TRUE16-NEXT: v_or_b32_e32 v50, 0x400000, v37
; GFX11-TRUE16-NEXT: v_cmp_u_f32_e32 vcc_lo, v37, v37
+; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_4)
+; GFX11-TRUE16-NEXT: v_bfi_b32 v10, 0xffff, v35, v145
+; GFX11-TRUE16-NEXT: v_bfe_u32 v35, v7, 16, 1
; GFX11-TRUE16-NEXT: v_add3_u32 v38, v38, v37, 0x7fff
-; GFX11-TRUE16-NEXT: v_bfe_u32 v48, v12, 16, 1
+; GFX11-TRUE16-NEXT: v_and_b32_e32 v12, 0xffff0000, v12
+; GFX11-TRUE16-NEXT: v_add_f32_e32 v37, 0x40c00000, v51
+; GFX11-TRUE16-NEXT: v_lshrrev_b32_e32 v57, 24, v10
; GFX11-TRUE16-NEXT: v_add3_u32 v35, v35, v7, 0x7fff
-; GFX11-TRUE16-NEXT: v_or_b32_e32 v52, 0x400000, v12
-; GFX11-TRUE16-NEXT: v_lshrrev_b32_e32 v43, 24, v10
-; GFX11-TRUE16-NEXT: v_cndmask_b32_e32 v59, v38, v50, vcc_lo
+; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_4) | instskip(SKIP_3) | instid1(VALU_DEP_4)
+; GFX11-TRUE16-NEXT: v_dual_cndmask_b32 v43, v38, v50 :: v_dual_add_f32 v12, 0x40c00000, v12
; GFX11-TRUE16-NEXT: v_cmp_u_f32_e32 vcc_lo, v7, v7
-; GFX11-TRUE16-NEXT: v_add3_u32 v48, v48, v12, 0x7fff
; GFX11-TRUE16-NEXT: v_and_b32_e32 v38, 0xffff0000, v14
; GFX11-TRUE16-NEXT: v_lshlrev_b32_e32 v14, 16, v14
-; GFX11-TRUE16-NEXT: v_mov_b16_e32 v7.l, v59.h
-; GFX11-TRUE16-NEXT: v_cndmask_b32_e32 v74, v35, v49, vcc_lo
+; GFX11-TRUE16-NEXT: v_mov_b16_e32 v7.l, v43.h
+; GFX11-TRUE16-NEXT: v_bfe_u32 v48, v12, 16, 1
+; GFX11-TRUE16-NEXT: v_or_b32_e32 v52, 0x400000, v12
+; GFX11-TRUE16-NEXT: v_cndmask_b32_e32 v60, v35, v49, vcc_lo
; GFX11-TRUE16-NEXT: v_cmp_u_f32_e32 vcc_lo, v12, v12
-; GFX11-TRUE16-NEXT: v_and_b32_e32 v51, 0xffff0000, v11
+; GFX11-TRUE16-NEXT: v_bfe_u32 v35, v37, 16, 1
+; GFX11-TRUE16-NEXT: v_add3_u32 v48, v48, v12, 0x7fff
; GFX11-TRUE16-NEXT: v_dual_add_f32 v14, 0x40c00000, v14 :: v_dual_lshlrev_b32 v11, 16, v11
-; GFX11-TRUE16-NEXT: v_lshrrev_b32_e32 v56, 8, v10
+; GFX11-TRUE16-NEXT: v_lshrrev_b32_e32 v62, 8, v10
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_3) | instskip(SKIP_1) | instid1(VALU_DEP_4)
-; GFX11-TRUE16-NEXT: v_dual_cndmask_b32 v60, v48, v52 :: v_dual_add_f32 v37, 0x40c00000, v51
+; GFX11-TRUE16-NEXT: v_cndmask_b32_e32 v150, v48, v52, vcc_lo
; GFX11-TRUE16-NEXT: v_cmp_u_f32_e32 vcc_lo, v9, v9
; GFX11-TRUE16-NEXT: v_bfe_u32 v49, v14, 16, 1
-; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_3) | instskip(NEXT) | instid1(VALU_DEP_4)
-; GFX11-TRUE16-NEXT: v_bfi_b32 v12, 0xffff, v7, v60
-; GFX11-TRUE16-NEXT: v_bfe_u32 v35, v37, 16, 1
+; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_3)
+; GFX11-TRUE16-NEXT: v_bfi_b32 v12, 0xffff, v7, v150
; GFX11-TRUE16-NEXT: v_cndmask_b32_e32 v9, v36, v39, vcc_lo
; GFX11-TRUE16-NEXT: v_add_f32_e32 v7, 0x40c00000, v11
-; GFX11-TRUE16-NEXT: v_cmp_u_f32_e32 vcc_lo, v37, v37
-; GFX11-TRUE16-NEXT: v_mov_b16_e32 v36.l, v74.h
; GFX11-TRUE16-NEXT: v_add3_u32 v11, v35, v37, 0x7fff
; GFX11-TRUE16-NEXT: v_or_b32_e32 v35, 0x400000, v37
+; GFX11-TRUE16-NEXT: v_cmp_u_f32_e32 vcc_lo, v37, v37
+; GFX11-TRUE16-NEXT: v_mov_b16_e32 v36.l, v60.h
; GFX11-TRUE16-NEXT: v_bfe_u32 v39, v7, 16, 1
-; GFX11-TRUE16-NEXT: v_lshrrev_b32_e32 v181, 24, v12
-; GFX11-TRUE16-NEXT: v_bfi_b32 v9, 0xffff, v36, v9
-; GFX11-TRUE16-NEXT: v_lshrrev_b32_e32 v182, 8, v12
+; GFX11-TRUE16-NEXT: v_lshrrev_b32_e32 v44, 24, v12
+; GFX11-TRUE16-NEXT: v_lshrrev_b32_e32 v45, 8, v12
; GFX11-TRUE16-NEXT: v_cndmask_b32_e32 v11, v11, v35, vcc_lo
; GFX11-TRUE16-NEXT: v_add_f32_e32 v35, 0x40c00000, v38
; GFX11-TRUE16-NEXT: v_add3_u32 v37, v39, v7, 0x7fff
@@ -168522,73 +168599,75 @@ define <128 x i8> @bitcast_v64bf16_to_v128i8(<64 x bfloat> %a, i32 %b) {
; GFX11-TRUE16-NEXT: v_cmp_u_f32_e32 vcc_lo, v7, v7
; GFX11-TRUE16-NEXT: v_and_b32_e32 v39, 0xffff0000, v13
; GFX11-TRUE16-NEXT: v_bfe_u32 v48, v35, 16, 1
-; GFX11-TRUE16-NEXT: v_lshlrev_b32_e32 v13, 16, v13
-; GFX11-TRUE16-NEXT: v_lshrrev_b64 v[64:65], 24, v[9:10]
-; GFX11-TRUE16-NEXT: v_cndmask_b32_e32 v89, v37, v38, vcc_lo
+; GFX11-TRUE16-NEXT: v_bfi_b32 v9, 0xffff, v36, v9
+; GFX11-TRUE16-NEXT: v_cndmask_b32_e32 v72, v37, v38, vcc_lo
; GFX11-TRUE16-NEXT: v_or_b32_e32 v38, 0x400000, v35
+; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_4)
; GFX11-TRUE16-NEXT: v_add3_u32 v37, v48, v35, 0x7fff
; GFX11-TRUE16-NEXT: v_cmp_u_f32_e32 vcc_lo, v35, v35
-; GFX11-TRUE16-NEXT: v_or_b32_e32 v48, 0x400000, v14
-; GFX11-TRUE16-NEXT: v_lshrrev_b32_e32 v57, 8, v9
-; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_4) | instskip(SKIP_3) | instid1(VALU_DEP_3)
-; GFX11-TRUE16-NEXT: v_dual_cndmask_b32 v76, v37, v38 :: v_dual_and_b32 v37, 0xffff0000, v16
-; GFX11-TRUE16-NEXT: v_dual_add_f32 v7, 0x40c00000, v39 :: v_dual_lshlrev_b32 v16, 16, v16
+; GFX11-TRUE16-NEXT: v_add_f32_e32 v7, 0x40c00000, v39
; GFX11-TRUE16-NEXT: v_add3_u32 v39, v49, v14, 0x7fff
-; GFX11-TRUE16-NEXT: v_cmp_u_f32_e32 vcc_lo, v14, v14
-; GFX11-TRUE16-NEXT: v_add_f32_e32 v16, 0x40c00000, v16
-; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_4) | instskip(SKIP_3) | instid1(VALU_DEP_4)
+; GFX11-TRUE16-NEXT: v_or_b32_e32 v48, 0x400000, v14
+; GFX11-TRUE16-NEXT: v_lshrrev_b64 v[64:65], 24, v[9:10]
+; GFX11-TRUE16-NEXT: v_cndmask_b32_e32 v165, v37, v38, vcc_lo
+; GFX11-TRUE16-NEXT: v_and_b32_e32 v37, 0xffff0000, v16
+; GFX11-TRUE16-NEXT: v_lshlrev_b32_e32 v16, 16, v16
+; GFX11-TRUE16-NEXT: v_lshlrev_b32_e32 v13, 16, v13
; GFX11-TRUE16-NEXT: v_bfe_u32 v49, v7, 16, 1
+; GFX11-TRUE16-NEXT: v_cmp_u_f32_e32 vcc_lo, v14, v14
; GFX11-TRUE16-NEXT: v_or_b32_e32 v35, 0x400000, v7
-; GFX11-TRUE16-NEXT: v_cndmask_b32_e32 v77, v39, v48, vcc_lo
-; GFX11-TRUE16-NEXT: v_cmp_u_f32_e32 vcc_lo, v7, v7
+; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_4) | instskip(NEXT) | instid1(VALU_DEP_4)
+; GFX11-TRUE16-NEXT: v_dual_add_f32 v16, 0x40c00000, v16 :: v_dual_add_f32 v13, 0x40c00000, v13
; GFX11-TRUE16-NEXT: v_add3_u32 v14, v49, v7, 0x7fff
+; GFX11-TRUE16-NEXT: v_cndmask_b32_e32 v58, v39, v48, vcc_lo
+; GFX11-TRUE16-NEXT: v_cmp_u_f32_e32 vcc_lo, v7, v7
+; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_4)
; GFX11-TRUE16-NEXT: v_or_b32_e32 v49, 0x400000, v16
-; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_4) | instskip(NEXT) | instid1(VALU_DEP_3)
-; GFX11-TRUE16-NEXT: v_mov_b16_e32 v38.l, v77.h
-; GFX11-TRUE16-NEXT: v_dual_cndmask_b32 v7, v14, v35 :: v_dual_add_f32 v14, 0x40c00000, v37
-; GFX11-TRUE16-NEXT: v_lshlrev_b32_e32 v37, 16, v15
-; GFX11-TRUE16-NEXT: v_and_b32_e32 v15, 0xffff0000, v15
-; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_3) | instskip(NEXT) | instid1(VALU_DEP_3)
-; GFX11-TRUE16-NEXT: v_bfe_u32 v48, v14, 16, 1
-; GFX11-TRUE16-NEXT: v_add_f32_e32 v37, 0x40c00000, v37
-; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_3) | instskip(NEXT) | instid1(VALU_DEP_2)
-; GFX11-TRUE16-NEXT: v_add_f32_e32 v15, 0x40c00000, v15
-; GFX11-TRUE16-NEXT: v_or_b32_e32 v51, 0x400000, v37
-; GFX11-TRUE16-NEXT: v_add_f32_e32 v13, 0x40c00000, v13
-; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_3) | instskip(NEXT) | instid1(VALU_DEP_2)
-; GFX11-TRUE16-NEXT: v_bfe_u32 v50, v15, 16, 1
; GFX11-TRUE16-NEXT: v_bfe_u32 v39, v13, 16, 1
+; GFX11-TRUE16-NEXT: v_lshrrev_b32_e32 v63, 8, v9
+; GFX11-TRUE16-NEXT: v_mov_b16_e32 v38.l, v58.h
+; GFX11-TRUE16-NEXT: v_cndmask_b32_e32 v7, v14, v35, vcc_lo
; GFX11-TRUE16-NEXT: v_cmp_u_f32_e32 vcc_lo, v13, v13
-; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_2) | instskip(SKIP_2) | instid1(VALU_DEP_2)
; GFX11-TRUE16-NEXT: v_add3_u32 v35, v39, v13, 0x7fff
; GFX11-TRUE16-NEXT: v_or_b32_e32 v39, 0x400000, v13
; GFX11-TRUE16-NEXT: v_bfe_u32 v13, v16, 16, 1
-; GFX11-TRUE16-NEXT: v_cndmask_b32_e32 v94, v35, v39, vcc_lo
-; GFX11-TRUE16-NEXT: v_bfe_u32 v39, v37, 16, 1
-; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_3)
+; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_2) | instskip(NEXT) | instid1(VALU_DEP_2)
+; GFX11-TRUE16-NEXT: v_cndmask_b32_e32 v79, v35, v39, vcc_lo
; GFX11-TRUE16-NEXT: v_add3_u32 v13, v13, v16, 0x7fff
; GFX11-TRUE16-NEXT: v_cmp_u_f32_e32 vcc_lo, v16, v16
+; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_2) | instskip(SKIP_2) | instid1(VALU_DEP_2)
+; GFX11-TRUE16-NEXT: v_cndmask_b32_e32 v74, v13, v49, vcc_lo
+; GFX11-TRUE16-NEXT: v_dual_add_f32 v14, 0x40c00000, v37 :: v_dual_lshlrev_b32 v37, 16, v15
+; GFX11-TRUE16-NEXT: v_and_b32_e32 v15, 0xffff0000, v15
+; GFX11-TRUE16-NEXT: v_bfe_u32 v48, v14, 16, 1
+; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_3) | instskip(NEXT) | instid1(VALU_DEP_3)
+; GFX11-TRUE16-NEXT: v_add_f32_e32 v37, 0x40c00000, v37
+; GFX11-TRUE16-NEXT: v_add_f32_e32 v15, 0x40c00000, v15
+; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_3) | instskip(NEXT) | instid1(VALU_DEP_3)
; GFX11-TRUE16-NEXT: v_add3_u32 v35, v48, v14, 0x7fff
+; GFX11-TRUE16-NEXT: v_bfe_u32 v39, v37, 16, 1
+; GFX11-TRUE16-NEXT: v_or_b32_e32 v51, 0x400000, v37
+; GFX11-TRUE16-NEXT: v_cmp_u_f32_e32 vcc_lo, v37, v37
; GFX11-TRUE16-NEXT: v_or_b32_e32 v48, 0x400000, v14
+; GFX11-TRUE16-NEXT: v_bfe_u32 v50, v15, 16, 1
; GFX11-TRUE16-NEXT: v_add3_u32 v39, v39, v37, 0x7fff
; GFX11-TRUE16-NEXT: v_or_b32_e32 v16, 0x400000, v15
-; GFX11-TRUE16-NEXT: v_cndmask_b32_e32 v91, v13, v49, vcc_lo
-; GFX11-TRUE16-NEXT: v_cmp_u_f32_e32 vcc_lo, v37, v37
+; GFX11-TRUE16-NEXT: v_mov_b16_e32 v37.l, v79.h
+; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_4) | instskip(NEXT) | instid1(VALU_DEP_4)
; GFX11-TRUE16-NEXT: v_add3_u32 v13, v50, v15, 0x7fff
-; GFX11-TRUE16-NEXT: v_mov_b16_e32 v37.l, v94.h
-; GFX11-TRUE16-NEXT: v_cndmask_b32_e32 v104, v39, v51, vcc_lo
+; GFX11-TRUE16-NEXT: v_cndmask_b32_e32 v88, v39, v51, vcc_lo
; GFX11-TRUE16-NEXT: v_cmp_u_f32_e32 vcc_lo, v14, v14
-; GFX11-TRUE16-NEXT: v_mov_b16_e32 v39.l, v89.h
-; GFX11-TRUE16-NEXT: v_bfi_b32 v14, 0xffff, v38, v76
-; GFX11-TRUE16-NEXT: v_cndmask_b32_e32 v92, v35, v48, vcc_lo
+; GFX11-TRUE16-NEXT: v_mov_b16_e32 v39.l, v72.h
+; GFX11-TRUE16-NEXT: v_bfi_b32 v14, 0xffff, v38, v165
+; GFX11-TRUE16-NEXT: v_cndmask_b32_e32 v183, v35, v48, vcc_lo
; GFX11-TRUE16-NEXT: v_cmp_u_f32_e32 vcc_lo, v15, v15
-; GFX11-TRUE16-NEXT: v_mov_b16_e32 v15.l, v104.h
-; GFX11-TRUE16-NEXT: v_mov_b16_e32 v35.l, v91.h
+; GFX11-TRUE16-NEXT: v_mov_b16_e32 v15.l, v88.h
+; GFX11-TRUE16-NEXT: v_mov_b16_e32 v35.l, v74.h
; GFX11-TRUE16-NEXT: v_bfi_b32 v11, 0xffff, v39, v11
-; GFX11-TRUE16-NEXT: v_lshrrev_b32_e32 v160, 24, v14
+; GFX11-TRUE16-NEXT: v_lshrrev_b32_e32 v176, 24, v14
; GFX11-TRUE16-NEXT: v_cndmask_b32_e32 v13, v13, v16, vcc_lo
-; GFX11-TRUE16-NEXT: v_lshrrev_b32_e32 v163, 8, v14
-; GFX11-TRUE16-NEXT: v_bfi_b32 v16, 0xffff, v35, v92
+; GFX11-TRUE16-NEXT: v_lshrrev_b32_e32 v178, 8, v14
+; GFX11-TRUE16-NEXT: v_bfi_b32 v16, 0xffff, v35, v183
; GFX11-TRUE16-NEXT: v_lshrrev_b64 v[52:53], 24, v[11:12]
; GFX11-TRUE16-NEXT: v_lshrrev_b64 v[53:54], 24, v[17:18]
; GFX11-TRUE16-NEXT: v_bfi_b32 v15, 0xffff, v15, v13
@@ -168607,27 +168686,27 @@ define <128 x i8> @bitcast_v64bf16_to_v128i8(<64 x bfloat> %a, i32 %b) {
; GFX11-TRUE16-NEXT: v_lshrrev_b64 v[68:69], 24, v[1:2]
; GFX11-TRUE16-NEXT: v_lshrrev_b64 v[38:39], 24, v[21:22]
; GFX11-TRUE16-NEXT: v_lshrrev_b64 v[50:51], 24, v[19:20]
-; GFX11-TRUE16-NEXT: v_lshrrev_b32_e32 v135, 24, v16
-; GFX11-TRUE16-NEXT: v_lshrrev_b32_e32 v146, 8, v16
-; GFX11-TRUE16-NEXT: v_lshrrev_b32_e32 v151, 8, v15
-; GFX11-TRUE16-NEXT: v_lshrrev_b32_e32 v167, 8, v13
-; GFX11-TRUE16-NEXT: v_lshrrev_b32_e32 v40, 8, v11
-; GFX11-TRUE16-NEXT: v_lshrrev_b32_e32 v75, 8, v7
+; GFX11-TRUE16-NEXT: v_lshrrev_b32_e32 v149, 24, v16
+; GFX11-TRUE16-NEXT: v_lshrrev_b32_e32 v161, 8, v16
+; GFX11-TRUE16-NEXT: v_lshrrev_b32_e32 v166, 8, v15
+; GFX11-TRUE16-NEXT: v_lshrrev_b32_e32 v181, 8, v13
+; GFX11-TRUE16-NEXT: v_lshrrev_b32_e32 v47, 8, v11
+; GFX11-TRUE16-NEXT: v_lshrrev_b32_e32 v90, 8, v7
; GFX11-TRUE16-NEXT: .LBB90_4: ; %end
; GFX11-TRUE16-NEXT: s_or_b32 exec_lo, exec_lo, s0
-; GFX11-TRUE16-NEXT: v_and_b16 v1.l, 0xff, v144.h
+; GFX11-TRUE16-NEXT: v_and_b16 v1.l, 0xff, v148.h
; GFX11-TRUE16-NEXT: v_lshlrev_b16 v2.l, 8, v108.l
; GFX11-TRUE16-NEXT: v_and_b16 v1.h, 0xff, v1.h
; GFX11-TRUE16-NEXT: v_lshlrev_b16 v2.h, 8, v68.l
-; GFX11-TRUE16-NEXT: v_and_b16 v3.l, 0xff, v132.h
+; GFX11-TRUE16-NEXT: v_and_b16 v3.l, 0xff, v117.h
; GFX11-TRUE16-NEXT: v_lshlrev_b16 v4.l, 8, v106.l
; GFX11-TRUE16-NEXT: v_or_b16 v1.l, v1.l, v2.l
-; GFX11-TRUE16-NEXT: v_and_b16 v5.l, 0xff, v149.h
+; GFX11-TRUE16-NEXT: v_and_b16 v5.l, 0xff, v128.h
; GFX11-TRUE16-NEXT: v_or_b16 v2.l, v1.h, v2.h
-; GFX11-TRUE16-NEXT: v_and_b16 v1.h, 0xff, v131.h
+; GFX11-TRUE16-NEXT: v_and_b16 v1.h, 0xff, v132.h
; GFX11-TRUE16-NEXT: v_lshlrev_b16 v2.h, 8, v107.l
; GFX11-TRUE16-NEXT: v_mov_b16_e32 v6.l, v1.l
-; GFX11-TRUE16-NEXT: v_lshlrev_b16 v7.l, 8, v78.l
+; GFX11-TRUE16-NEXT: v_lshlrev_b16 v7.l, 8, v91.l
; GFX11-TRUE16-NEXT: v_mov_b16_e32 v8.l, v2.l
; GFX11-TRUE16-NEXT: v_or_b16 v2.l, v3.l, v4.l
; GFX11-TRUE16-NEXT: v_or_b16 v1.l, v1.h, v2.h
@@ -168636,7 +168715,7 @@ define <128 x i8> @bitcast_v64bf16_to_v128i8(<64 x bfloat> %a, i32 %b) {
; GFX11-TRUE16-NEXT: v_lshlrev_b32_e32 v6, 16, v8
; GFX11-TRUE16-NEXT: v_mov_b16_e32 v10.l, v2.l
; GFX11-TRUE16-NEXT: v_mov_b16_e32 v8.l, v1.l
-; GFX11-TRUE16-NEXT: v_and_b16 v2.l, 0xff, v162.h
+; GFX11-TRUE16-NEXT: v_and_b16 v2.l, 0xff, v163.h
; GFX11-TRUE16-NEXT: v_and_b16 v3.l, 0xff, v3.h
; GFX11-TRUE16-NEXT: v_lshlrev_b16 v3.h, 8, v67.l
; GFX11-TRUE16-NEXT: v_or_b32_e32 v1, v4, v6
@@ -168645,15 +168724,15 @@ define <128 x i8> @bitcast_v64bf16_to_v128i8(<64 x bfloat> %a, i32 %b) {
; GFX11-TRUE16-NEXT: v_or_b16 v4.l, v2.l, v2.h
; GFX11-TRUE16-NEXT: v_or_b16 v3.l, v3.l, v3.h
; GFX11-TRUE16-NEXT: v_and_b16 v3.h, 0xff, v147.h
-; GFX11-TRUE16-NEXT: v_lshlrev_b16 v4.h, 8, v95.l
-; GFX11-TRUE16-NEXT: v_lshlrev_b16 v6.l, 8, v93.l
+; GFX11-TRUE16-NEXT: v_lshlrev_b16 v4.h, 8, v104.l
+; GFX11-TRUE16-NEXT: v_lshlrev_b16 v6.l, 8, v95.l
; GFX11-TRUE16-NEXT: v_or_b32_e32 v2, v8, v10
; GFX11-TRUE16-NEXT: v_mov_b16_e32 v8.l, v4.l
; GFX11-TRUE16-NEXT: v_mov_b16_e32 v10.l, v3.l
; GFX11-TRUE16-NEXT: v_or_b16 v3.l, v3.h, v4.h
; GFX11-TRUE16-NEXT: v_or_b16 v4.l, v5.l, v6.l
-; GFX11-TRUE16-NEXT: v_and_b16 v3.h, 0xff, v180.h
-; GFX11-TRUE16-NEXT: v_lshlrev_b16 v4.h, 8, v90.l
+; GFX11-TRUE16-NEXT: v_and_b16 v3.h, 0xff, v179.h
+; GFX11-TRUE16-NEXT: v_lshlrev_b16 v4.h, 8, v94.l
; GFX11-TRUE16-NEXT: v_and_b32_e32 v6, 0xffff, v8
; GFX11-TRUE16-NEXT: v_lshlrev_b32_e32 v8, 16, v10
; GFX11-TRUE16-NEXT: v_mov_b16_e32 v10.l, v3.l
@@ -168664,10 +168743,10 @@ define <128 x i8> @bitcast_v64bf16_to_v128i8(<64 x bfloat> %a, i32 %b) {
; GFX11-TRUE16-NEXT: v_or_b32_e32 v3, v6, v8
; GFX11-TRUE16-NEXT: v_and_b32_e32 v8, 0xffff, v10
; GFX11-TRUE16-NEXT: v_lshlrev_b32_e32 v10, 16, v12
-; GFX11-TRUE16-NEXT: v_and_b16 v5.h, 0xff, v164.h
+; GFX11-TRUE16-NEXT: v_and_b16 v5.h, 0xff, v160.h
; GFX11-TRUE16-NEXT: v_or_b16 v5.l, v4.h, v5.l
-; GFX11-TRUE16-NEXT: v_lshlrev_b16 v6.l, 8, v88.l
-; GFX11-TRUE16-NEXT: v_and_b16 v6.h, 0xff, v165.h
+; GFX11-TRUE16-NEXT: v_lshlrev_b16 v6.l, 8, v93.l
+; GFX11-TRUE16-NEXT: v_and_b16 v6.h, 0xff, v131.h
; GFX11-TRUE16-NEXT: v_mov_b16_e32 v12.l, v4.l
; GFX11-TRUE16-NEXT: v_or_b32_e32 v4, v8, v10
; GFX11-TRUE16-NEXT: v_mov_b16_e32 v10.l, v5.l
@@ -168676,8 +168755,8 @@ define <128 x i8> @bitcast_v64bf16_to_v128i8(<64 x bfloat> %a, i32 %b) {
; GFX11-TRUE16-NEXT: v_and_b16 v7.l, 0xff, v7.h
; GFX11-TRUE16-NEXT: v_lshlrev_b16 v7.h, 8, v65.l
; GFX11-TRUE16-NEXT: v_and_b32_e32 v8, 0xffff, v12
-; GFX11-TRUE16-NEXT: v_and_b16 v5.h, 0xff, v46.h
-; GFX11-TRUE16-NEXT: v_lshlrev_b16 v6.h, 8, v75.l
+; GFX11-TRUE16-NEXT: v_and_b16 v5.h, 0xff, v42.h
+; GFX11-TRUE16-NEXT: v_lshlrev_b16 v6.h, 8, v90.l
; GFX11-TRUE16-NEXT: v_lshlrev_b32_e32 v10, 16, v10
; GFX11-TRUE16-NEXT: v_or_b16 v7.l, v7.l, v7.h
; GFX11-TRUE16-NEXT: v_mov_b16_e32 v12.l, v5.l
@@ -168685,24 +168764,24 @@ define <128 x i8> @bitcast_v64bf16_to_v128i8(<64 x bfloat> %a, i32 %b) {
; GFX11-TRUE16-NEXT: v_or_b16 v6.l, v5.h, v6.h
; GFX11-TRUE16-NEXT: v_or_b32_e32 v5, v8, v10
; GFX11-TRUE16-NEXT: v_mov_b16_e32 v16.l, v7.l
-; GFX11-TRUE16-NEXT: v_and_b16 v7.l, 0xff, v178.h
-; GFX11-TRUE16-NEXT: v_lshlrev_b16 v7.h, 8, v63.l
-; GFX11-TRUE16-NEXT: v_and_b16 v8.l, 0xff, v179.h
-; GFX11-TRUE16-NEXT: v_lshlrev_b16 v8.h, 8, v62.l
+; GFX11-TRUE16-NEXT: v_and_b16 v7.l, 0xff, v167.h
+; GFX11-TRUE16-NEXT: v_lshlrev_b16 v7.h, 8, v77.l
+; GFX11-TRUE16-NEXT: v_and_b16 v8.l, 0xff, v134.h
+; GFX11-TRUE16-NEXT: v_lshlrev_b16 v8.h, 8, v76.l
; GFX11-TRUE16-NEXT: v_and_b32_e32 v10, 0xffff, v12
; GFX11-TRUE16-NEXT: v_lshlrev_b32_e32 v12, 16, v14
; GFX11-TRUE16-NEXT: v_mov_b16_e32 v14.l, v6.l
; GFX11-TRUE16-NEXT: v_or_b16 v7.l, v7.l, v7.h
; GFX11-TRUE16-NEXT: v_or_b16 v8.l, v8.l, v8.h
-; GFX11-TRUE16-NEXT: v_and_b16 v7.h, 0xff, v74.h
+; GFX11-TRUE16-NEXT: v_and_b16 v7.h, 0xff, v60.h
; GFX11-TRUE16-NEXT: v_or_b32_e32 v6, v10, v12
; GFX11-TRUE16-NEXT: v_and_b32_e32 v12, 0xffff, v14
; GFX11-TRUE16-NEXT: v_lshlrev_b32_e32 v14, 16, v16
-; GFX11-TRUE16-NEXT: v_lshlrev_b16 v8.h, 8, v57.l
+; GFX11-TRUE16-NEXT: v_lshlrev_b16 v8.h, 8, v63.l
; GFX11-TRUE16-NEXT: v_and_b16 v9.l, 0xff, v9.h
; GFX11-TRUE16-NEXT: v_lshlrev_b16 v9.h, 8, v64.l
-; GFX11-TRUE16-NEXT: v_and_b16 v10.l, 0xff, v45.h
-; GFX11-TRUE16-NEXT: v_lshlrev_b16 v10.h, 8, v56.l
+; GFX11-TRUE16-NEXT: v_and_b16 v10.l, 0xff, v182.h
+; GFX11-TRUE16-NEXT: v_lshlrev_b16 v10.h, 8, v62.l
; GFX11-TRUE16-NEXT: v_mov_b16_e32 v16.l, v7.l
; GFX11-TRUE16-NEXT: v_mov_b16_e32 v18.l, v8.l
; GFX11-TRUE16-NEXT: v_or_b16 v7.l, v7.h, v8.h
@@ -168715,25 +168794,25 @@ define <128 x i8> @bitcast_v64bf16_to_v128i8(<64 x bfloat> %a, i32 %b) {
; GFX11-TRUE16-NEXT: v_and_b16 v11.l, 0xff, v11.h
; GFX11-TRUE16-NEXT: v_lshlrev_b16 v11.h, 8, v52.l
; GFX11-TRUE16-NEXT: v_or_b32_e32 v8, v10, v16
-; GFX11-TRUE16-NEXT: v_and_b16 v10.l, 0xff, v89.h
-; GFX11-TRUE16-NEXT: v_lshlrev_b16 v10.h, 8, v40.l
+; GFX11-TRUE16-NEXT: v_and_b16 v10.l, 0xff, v72.h
+; GFX11-TRUE16-NEXT: v_lshlrev_b16 v10.h, 8, v47.l
; GFX11-TRUE16-NEXT: v_or_b32_e32 v7, v12, v14
; GFX11-TRUE16-NEXT: v_and_b32_e32 v16, 0xffff, v18
; GFX11-TRUE16-NEXT: v_lshlrev_b32_e32 v18, 16, v20
; GFX11-TRUE16-NEXT: v_and_b32_e32 v20, 0xffff, v9
-; GFX11-TRUE16-NEXT: v_and_b16 v9.l, 0xff, v42.h
-; GFX11-TRUE16-NEXT: v_lshlrev_b16 v9.h, 8, v43.l
-; GFX11-TRUE16-NEXT: v_and_b16 v13.l, 0xff, v60.h
-; GFX11-TRUE16-NEXT: v_lshlrev_b16 v14.l, 8, v181.l
+; GFX11-TRUE16-NEXT: v_and_b16 v9.l, 0xff, v145.h
+; GFX11-TRUE16-NEXT: v_lshlrev_b16 v9.h, 8, v57.l
+; GFX11-TRUE16-NEXT: v_and_b16 v13.l, 0xff, v150.h
+; GFX11-TRUE16-NEXT: v_lshlrev_b16 v14.l, 8, v44.l
; GFX11-TRUE16-NEXT: v_or_b16 v10.l, v10.l, v10.h
; GFX11-TRUE16-NEXT: v_or_b16 v11.l, v11.l, v11.h
; GFX11-TRUE16-NEXT: s_clause 0x1
; GFX11-TRUE16-NEXT: scratch_store_b128 v0, v[1:4], off
; GFX11-TRUE16-NEXT: scratch_store_b128 v0, v[5:8], off offset:16
-; GFX11-TRUE16-NEXT: v_and_b16 v5.l, 0xff, v94.h
-; GFX11-TRUE16-NEXT: v_lshlrev_b16 v5.h, 8, v167.l
-; GFX11-TRUE16-NEXT: v_and_b16 v12.l, 0xff, v59.h
-; GFX11-TRUE16-NEXT: v_lshlrev_b16 v12.h, 8, v182.l
+; GFX11-TRUE16-NEXT: v_and_b16 v5.l, 0xff, v79.h
+; GFX11-TRUE16-NEXT: v_lshlrev_b16 v5.h, 8, v181.l
+; GFX11-TRUE16-NEXT: v_and_b16 v12.l, 0xff, v43.h
+; GFX11-TRUE16-NEXT: v_lshlrev_b16 v12.h, 8, v45.l
; GFX11-TRUE16-NEXT: v_or_b16 v9.l, v9.l, v9.h
; GFX11-TRUE16-NEXT: v_or_b16 v13.l, v13.l, v14.l
; GFX11-TRUE16-NEXT: v_and_b32_e32 v10, 0xffff, v10
@@ -168745,8 +168824,8 @@ define <128 x i8> @bitcast_v64bf16_to_v128i8(<64 x bfloat> %a, i32 %b) {
; GFX11-TRUE16-NEXT: v_mov_b16_e32 v14.l, v13.l
; GFX11-TRUE16-NEXT: v_lshlrev_b32_e32 v9, 16, v9
; GFX11-TRUE16-NEXT: v_or_b32_e32 v3, v10, v11
-; GFX11-TRUE16-NEXT: v_and_b16 v7.l, 0xff, v77.h
-; GFX11-TRUE16-NEXT: v_lshlrev_b16 v7.h, 8, v163.l
+; GFX11-TRUE16-NEXT: v_and_b16 v7.l, 0xff, v58.h
+; GFX11-TRUE16-NEXT: v_lshlrev_b16 v7.h, 8, v178.l
; GFX11-TRUE16-NEXT: v_or_b16 v6.l, v6.l, v6.h
; GFX11-TRUE16-NEXT: v_and_b32_e32 v10, 0xffff, v5
; GFX11-TRUE16-NEXT: v_and_b16 v5.l, 0xff, v15.h
@@ -168754,10 +168833,10 @@ define <128 x i8> @bitcast_v64bf16_to_v128i8(<64 x bfloat> %a, i32 %b) {
; GFX11-TRUE16-NEXT: v_and_b32_e32 v12, 0xffff, v12
; GFX11-TRUE16-NEXT: v_lshlrev_b32_e32 v14, 16, v14
; GFX11-TRUE16-NEXT: v_or_b32_e32 v2, v20, v9
-; GFX11-TRUE16-NEXT: v_and_b16 v8.l, 0xff, v76.h
-; GFX11-TRUE16-NEXT: v_lshlrev_b16 v8.h, 8, v160.l
-; GFX11-TRUE16-NEXT: v_and_b16 v9.l, 0xff, v104.h
-; GFX11-TRUE16-NEXT: v_lshlrev_b16 v9.h, 8, v151.l
+; GFX11-TRUE16-NEXT: v_and_b16 v8.l, 0xff, v165.h
+; GFX11-TRUE16-NEXT: v_lshlrev_b16 v8.h, 8, v176.l
+; GFX11-TRUE16-NEXT: v_and_b16 v9.l, 0xff, v88.h
+; GFX11-TRUE16-NEXT: v_lshlrev_b16 v9.h, 8, v166.l
; GFX11-TRUE16-NEXT: v_or_b16 v7.l, v7.l, v7.h
; GFX11-TRUE16-NEXT: v_lshlrev_b32_e32 v11, 16, v6
; GFX11-TRUE16-NEXT: v_or_b16 v5.l, v5.l, v5.h
@@ -168765,24 +168844,24 @@ define <128 x i8> @bitcast_v64bf16_to_v128i8(<64 x bfloat> %a, i32 %b) {
; GFX11-TRUE16-NEXT: v_or_b16 v8.l, v8.l, v8.h
; GFX11-TRUE16-NEXT: v_or_b16 v9.l, v9.l, v9.h
; GFX11-TRUE16-NEXT: v_and_b32_e32 v12, 0xffff, v7
-; GFX11-TRUE16-NEXT: v_and_b16 v6.l, 0xff, v91.h
-; GFX11-TRUE16-NEXT: v_lshlrev_b16 v6.h, 8, v146.l
-; GFX11-TRUE16-NEXT: v_and_b16 v7.l, 0xff, v92.h
-; GFX11-TRUE16-NEXT: v_lshlrev_b16 v7.h, 8, v135.l
+; GFX11-TRUE16-NEXT: v_and_b16 v6.l, 0xff, v74.h
+; GFX11-TRUE16-NEXT: v_lshlrev_b16 v6.h, 8, v161.l
+; GFX11-TRUE16-NEXT: v_and_b16 v7.l, 0xff, v183.h
+; GFX11-TRUE16-NEXT: v_lshlrev_b16 v7.h, 8, v149.l
; GFX11-TRUE16-NEXT: v_lshlrev_b32_e32 v15, 16, v5
; GFX11-TRUE16-NEXT: v_or_b32_e32 v5, v10, v11
-; GFX11-TRUE16-NEXT: v_and_b16 v10.l, 0xff, v55.h
-; GFX11-TRUE16-NEXT: v_lshlrev_b16 v10.h, 8, v73.l
+; GFX11-TRUE16-NEXT: v_and_b16 v10.l, 0xff, v71.h
+; GFX11-TRUE16-NEXT: v_lshlrev_b16 v10.h, 8, v89.l
; GFX11-TRUE16-NEXT: v_lshlrev_b32_e32 v13, 16, v8
; GFX11-TRUE16-NEXT: v_and_b32_e32 v14, 0xffff, v9
-; GFX11-TRUE16-NEXT: v_and_b16 v8.l, 0xff, v80.h
-; GFX11-TRUE16-NEXT: v_lshlrev_b16 v8.h, 8, v79.l
+; GFX11-TRUE16-NEXT: v_and_b16 v8.l, 0xff, v83.h
+; GFX11-TRUE16-NEXT: v_lshlrev_b16 v8.h, 8, v92.l
; GFX11-TRUE16-NEXT: v_and_b16 v9.l, 0xff, v17.h
; GFX11-TRUE16-NEXT: v_lshlrev_b16 v9.h, 8, v53.l
; GFX11-TRUE16-NEXT: v_or_b16 v6.l, v6.l, v6.h
; GFX11-TRUE16-NEXT: v_or_b16 v7.l, v7.l, v7.h
-; GFX11-TRUE16-NEXT: v_and_b16 v11.l, 0xff, v70.h
-; GFX11-TRUE16-NEXT: v_lshlrev_b16 v11.h, 8, v72.l
+; GFX11-TRUE16-NEXT: v_and_b16 v11.l, 0xff, v55.h
+; GFX11-TRUE16-NEXT: v_lshlrev_b16 v11.h, 8, v78.l
; GFX11-TRUE16-NEXT: v_or_b16 v10.l, v10.l, v10.h
; GFX11-TRUE16-NEXT: v_or_b32_e32 v1, v16, v18
; GFX11-TRUE16-NEXT: v_or_b16 v8.l, v8.l, v8.h
@@ -168791,24 +168870,24 @@ define <128 x i8> @bitcast_v64bf16_to_v128i8(<64 x bfloat> %a, i32 %b) {
; GFX11-TRUE16-NEXT: v_lshlrev_b32_e32 v17, 16, v7
; GFX11-TRUE16-NEXT: v_or_b32_e32 v6, v12, v13
; GFX11-TRUE16-NEXT: v_or_b32_e32 v7, v14, v15
-; GFX11-TRUE16-NEXT: v_and_b16 v12.l, 0xff, v83.h
-; GFX11-TRUE16-NEXT: v_lshlrev_b16 v12.h, 8, v61.l
+; GFX11-TRUE16-NEXT: v_and_b16 v12.l, 0xff, v86.h
+; GFX11-TRUE16-NEXT: v_lshlrev_b16 v12.h, 8, v75.l
; GFX11-TRUE16-NEXT: v_and_b16 v13.l, 0xff, v19.h
; GFX11-TRUE16-NEXT: v_lshlrev_b16 v13.h, 8, v50.l
; GFX11-TRUE16-NEXT: v_or_b16 v11.l, v11.l, v11.h
; GFX11-TRUE16-NEXT: v_and_b32_e32 v15, 0xffff, v10
-; GFX11-TRUE16-NEXT: v_and_b16 v10.l, 0xff, v71.h
-; GFX11-TRUE16-NEXT: v_lshlrev_b16 v10.h, 8, v47.l
+; GFX11-TRUE16-NEXT: v_and_b16 v10.l, 0xff, v70.h
+; GFX11-TRUE16-NEXT: v_lshlrev_b16 v10.h, 8, v61.l
; GFX11-TRUE16-NEXT: v_and_b32_e32 v18, 0xffff, v8
; GFX11-TRUE16-NEXT: v_lshlrev_b32_e32 v9, 16, v9
; GFX11-TRUE16-NEXT: v_or_b32_e32 v8, v16, v17
; GFX11-TRUE16-NEXT: v_and_b16 v14.l, 0xff, v81.h
-; GFX11-TRUE16-NEXT: v_lshlrev_b16 v14.h, 8, v58.l
+; GFX11-TRUE16-NEXT: v_lshlrev_b16 v14.h, 8, v73.l
; GFX11-TRUE16-NEXT: v_or_b16 v12.l, v12.l, v12.h
; GFX11-TRUE16-NEXT: v_or_b16 v13.l, v13.l, v13.h
; GFX11-TRUE16-NEXT: v_lshlrev_b32_e32 v16, 16, v11
-; GFX11-TRUE16-NEXT: v_and_b16 v11.l, 0xff, v85.h
-; GFX11-TRUE16-NEXT: v_lshlrev_b16 v11.h, 8, v44.l
+; GFX11-TRUE16-NEXT: v_and_b16 v11.l, 0xff, v97.h
+; GFX11-TRUE16-NEXT: v_lshlrev_b16 v11.h, 8, v59.l
; GFX11-TRUE16-NEXT: v_or_b16 v10.l, v10.l, v10.h
; GFX11-TRUE16-NEXT: v_or_b32_e32 v9, v18, v9
; GFX11-TRUE16-NEXT: v_or_b16 v14.l, v14.l, v14.h
@@ -168819,29 +168898,29 @@ define <128 x i8> @bitcast_v64bf16_to_v128i8(<64 x bfloat> %a, i32 %b) {
; GFX11-TRUE16-NEXT: v_or_b16 v11.l, v11.l, v11.h
; GFX11-TRUE16-NEXT: v_lshlrev_b32_e32 v20, 16, v10
; GFX11-TRUE16-NEXT: v_or_b32_e32 v10, v15, v16
-; GFX11-TRUE16-NEXT: v_and_b16 v15.l, 0xff, v96.h
-; GFX11-TRUE16-NEXT: v_lshlrev_b16 v15.h, 8, v177.l
+; GFX11-TRUE16-NEXT: v_and_b16 v15.l, 0xff, v100.h
+; GFX11-TRUE16-NEXT: v_lshlrev_b16 v15.h, 8, v41.l
; GFX11-TRUE16-NEXT: v_and_b32_e32 v19, 0xffff, v14
-; GFX11-TRUE16-NEXT: v_and_b16 v13.l, 0xff, v82.h
-; GFX11-TRUE16-NEXT: v_lshlrev_b16 v13.h, 8, v41.l
-; GFX11-TRUE16-NEXT: v_and_b16 v14.l, 0xff, v84.h
-; GFX11-TRUE16-NEXT: v_lshlrev_b16 v14.h, 8, v183.l
+; GFX11-TRUE16-NEXT: v_and_b16 v13.l, 0xff, v84.h
+; GFX11-TRUE16-NEXT: v_lshlrev_b16 v13.h, 8, v56.l
+; GFX11-TRUE16-NEXT: v_and_b16 v14.l, 0xff, v80.h
+; GFX11-TRUE16-NEXT: v_lshlrev_b16 v14.h, 8, v46.l
; GFX11-TRUE16-NEXT: v_or_b16 v12.l, v12.l, v12.h
; GFX11-TRUE16-NEXT: v_and_b32_e32 v21, 0xffff, v11
; GFX11-TRUE16-NEXT: v_or_b32_e32 v11, v17, v18
; GFX11-TRUE16-NEXT: v_and_b16 v16.l, 0xff, v23.h
; GFX11-TRUE16-NEXT: v_lshlrev_b16 v16.h, 8, v37.l
-; GFX11-TRUE16-NEXT: v_and_b16 v17.l, 0xff, v86.h
-; GFX11-TRUE16-NEXT: v_lshlrev_b16 v17.h, 8, v176.l
+; GFX11-TRUE16-NEXT: v_and_b16 v17.l, 0xff, v96.h
+; GFX11-TRUE16-NEXT: v_lshlrev_b16 v17.h, 8, v40.l
; GFX11-TRUE16-NEXT: v_or_b16 v15.l, v15.l, v15.h
; GFX11-TRUE16-NEXT: v_or_b16 v13.l, v13.l, v13.h
; GFX11-TRUE16-NEXT: v_or_b16 v14.l, v14.l, v14.h
; GFX11-TRUE16-NEXT: v_lshlrev_b32_e32 v22, 16, v12
; GFX11-TRUE16-NEXT: v_or_b32_e32 v12, v19, v20
-; GFX11-TRUE16-NEXT: v_and_b16 v18.l, 0xff, v87.h
-; GFX11-TRUE16-NEXT: v_lshlrev_b16 v18.h, 8, v166.l
-; GFX11-TRUE16-NEXT: v_and_b16 v19.l, 0xff, v101.h
-; GFX11-TRUE16-NEXT: v_lshlrev_b16 v19.h, 8, v161.l
+; GFX11-TRUE16-NEXT: v_and_b16 v18.l, 0xff, v82.h
+; GFX11-TRUE16-NEXT: v_lshlrev_b16 v18.h, 8, v180.l
+; GFX11-TRUE16-NEXT: v_and_b16 v19.l, 0xff, v103.h
+; GFX11-TRUE16-NEXT: v_lshlrev_b16 v19.h, 8, v177.l
; GFX11-TRUE16-NEXT: v_or_b16 v16.l, v16.l, v16.h
; GFX11-TRUE16-NEXT: v_or_b16 v17.l, v17.l, v17.h
; GFX11-TRUE16-NEXT: v_and_b32_e32 v20, 0xffff, v15
@@ -168854,16 +168933,16 @@ define <128 x i8> @bitcast_v64bf16_to_v128i8(<64 x bfloat> %a, i32 %b) {
; GFX11-TRUE16-NEXT: v_or_b16 v19.l, v19.l, v19.h
; GFX11-TRUE16-NEXT: v_lshlrev_b32_e32 v21, 16, v16
; GFX11-TRUE16-NEXT: v_and_b32_e32 v22, 0xffff, v17
-; GFX11-TRUE16-NEXT: v_and_b16 v16.l, 0xff, v97.h
-; GFX11-TRUE16-NEXT: v_lshlrev_b16 v16.h, 8, v150.l
-; GFX11-TRUE16-NEXT: v_and_b16 v17.l, 0xff, v98.h
-; GFX11-TRUE16-NEXT: v_lshlrev_b16 v17.h, 8, v148.l
+; GFX11-TRUE16-NEXT: v_and_b16 v16.l, 0xff, v99.h
+; GFX11-TRUE16-NEXT: v_lshlrev_b16 v16.h, 8, v164.l
+; GFX11-TRUE16-NEXT: v_and_b16 v17.l, 0xff, v85.h
+; GFX11-TRUE16-NEXT: v_lshlrev_b16 v17.h, 8, v162.l
; GFX11-TRUE16-NEXT: v_or_b16 v15.l, v15.l, v15.h
; GFX11-TRUE16-NEXT: v_or_b32_e32 v14, v24, v14
; GFX11-TRUE16-NEXT: v_lshlrev_b32_e32 v23, 16, v18
; GFX11-TRUE16-NEXT: v_and_b32_e32 v24, 0xffff, v19
-; GFX11-TRUE16-NEXT: v_and_b16 v18.l, 0xff, v112.h
-; GFX11-TRUE16-NEXT: v_lshlrev_b16 v18.h, 8, v145.l
+; GFX11-TRUE16-NEXT: v_and_b16 v18.l, 0xff, v113.h
+; GFX11-TRUE16-NEXT: v_lshlrev_b16 v18.h, 8, v151.l
; GFX11-TRUE16-NEXT: v_and_b16 v19.l, 0xff, v27.h
; GFX11-TRUE16-NEXT: v_lshlrev_b16 v19.h, 8, v35.l
; GFX11-TRUE16-NEXT: v_or_b16 v16.l, v16.l, v16.h
@@ -168876,16 +168955,16 @@ define <128 x i8> @bitcast_v64bf16_to_v128i8(<64 x bfloat> %a, i32 %b) {
; GFX11-TRUE16-NEXT: v_or_b32_e32 v15, v20, v21
; GFX11-TRUE16-NEXT: v_or_b32_e32 v16, v22, v23
; GFX11-TRUE16-NEXT: v_or_b32_e32 v17, v24, v25
-; GFX11-TRUE16-NEXT: v_and_b16 v20.l, 0xff, v99.h
-; GFX11-TRUE16-NEXT: v_lshlrev_b16 v20.h, 8, v134.l
-; GFX11-TRUE16-NEXT: v_and_b16 v21.l, 0xff, v100.h
-; GFX11-TRUE16-NEXT: v_lshlrev_b16 v21.h, 8, v133.l
-; GFX11-TRUE16-NEXT: v_and_b16 v22.l, 0xff, v113.h
-; GFX11-TRUE16-NEXT: v_lshlrev_b16 v22.h, 8, v130.l
+; GFX11-TRUE16-NEXT: v_and_b16 v20.l, 0xff, v101.h
+; GFX11-TRUE16-NEXT: v_lshlrev_b16 v20.h, 8, v146.l
+; GFX11-TRUE16-NEXT: v_and_b16 v21.l, 0xff, v87.h
+; GFX11-TRUE16-NEXT: v_lshlrev_b16 v21.h, 8, v144.l
+; GFX11-TRUE16-NEXT: v_and_b16 v22.l, 0xff, v115.h
+; GFX11-TRUE16-NEXT: v_lshlrev_b16 v22.h, 8, v135.l
; GFX11-TRUE16-NEXT: v_and_b16 v23.l, 0xff, v29.h
; GFX11-TRUE16-NEXT: v_lshlrev_b16 v23.h, 8, v34.l
-; GFX11-TRUE16-NEXT: v_and_b16 v24.l, 0xff, v103.h
-; GFX11-TRUE16-NEXT: v_lshlrev_b16 v24.h, 8, v129.l
+; GFX11-TRUE16-NEXT: v_and_b16 v24.l, 0xff, v112.h
+; GFX11-TRUE16-NEXT: v_lshlrev_b16 v24.h, 8, v133.l
; GFX11-TRUE16-NEXT: v_and_b32_e32 v28, 0xffff, v18
; GFX11-TRUE16-NEXT: v_lshlrev_b32_e32 v19, 16, v19
; GFX11-TRUE16-NEXT: v_or_b16 v20.l, v20.l, v20.h
@@ -168900,17 +168979,17 @@ define <128 x i8> @bitcast_v64bf16_to_v128i8(<64 x bfloat> %a, i32 %b) {
; GFX11-TRUE16-NEXT: v_and_b32_e32 v27, 0xffff, v22
; GFX11-TRUE16-NEXT: v_lshlrev_b32_e32 v28, 16, v23
; GFX11-TRUE16-NEXT: v_and_b32_e32 v29, 0xffff, v24
-; GFX11-TRUE16-NEXT: v_and_b16 v20.l, 0xff, v102.h
-; GFX11-TRUE16-NEXT: v_lshlrev_b16 v20.h, 8, v128.l
+; GFX11-TRUE16-NEXT: v_and_b16 v20.l, 0xff, v98.h
+; GFX11-TRUE16-NEXT: v_lshlrev_b16 v20.h, 8, v130.l
; GFX11-TRUE16-NEXT: v_and_b16 v21.l, 0xff, v116.h
-; GFX11-TRUE16-NEXT: v_lshlrev_b16 v21.h, 8, v119.l
+; GFX11-TRUE16-NEXT: v_lshlrev_b16 v21.h, 8, v129.l
; GFX11-TRUE16-NEXT: s_waitcnt vmcnt(0)
; GFX11-TRUE16-NEXT: v_and_b16 v22.l, 0xff, v31.h
; GFX11-TRUE16-NEXT: v_lshlrev_b16 v22.h, 8, v33.l
; GFX11-TRUE16-NEXT: v_and_b16 v23.l, 0xff, v114.h
-; GFX11-TRUE16-NEXT: v_lshlrev_b16 v23.h, 8, v118.l
-; GFX11-TRUE16-NEXT: v_and_b16 v24.l, 0xff, v115.h
-; GFX11-TRUE16-NEXT: v_lshlrev_b16 v24.h, 8, v117.l
+; GFX11-TRUE16-NEXT: v_lshlrev_b16 v23.h, 8, v119.l
+; GFX11-TRUE16-NEXT: v_and_b16 v24.l, 0xff, v102.h
+; GFX11-TRUE16-NEXT: v_lshlrev_b16 v24.h, 8, v118.l
; GFX11-TRUE16-NEXT: v_or_b16 v20.l, v20.l, v20.h
; GFX11-TRUE16-NEXT: v_or_b16 v21.l, v21.l, v21.h
; GFX11-TRUE16-NEXT: v_or_b16 v22.l, v22.l, v22.h
@@ -168978,128 +169057,132 @@ define <128 x i8> @bitcast_v64bf16_to_v128i8(<64 x bfloat> %a, i32 %b) {
; GFX11-FAKE16-LABEL: bitcast_v64bf16_to_v128i8:
; GFX11-FAKE16: ; %bb.0:
; GFX11-FAKE16-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
-; GFX11-FAKE16-NEXT: s_clause 0x15
-; GFX11-FAKE16-NEXT: scratch_store_b32 off, v40, s32 offset:96
-; GFX11-FAKE16-NEXT: scratch_store_b32 off, v41, s32 offset:92
-; GFX11-FAKE16-NEXT: scratch_store_b32 off, v42, s32 offset:88
-; GFX11-FAKE16-NEXT: scratch_store_b32 off, v43, s32 offset:84
-; GFX11-FAKE16-NEXT: scratch_store_b32 off, v44, s32 offset:80
-; GFX11-FAKE16-NEXT: scratch_store_b32 off, v45, s32 offset:76
-; GFX11-FAKE16-NEXT: scratch_store_b32 off, v46, s32 offset:72
-; GFX11-FAKE16-NEXT: scratch_store_b32 off, v47, s32 offset:68
-; GFX11-FAKE16-NEXT: scratch_store_b32 off, v56, s32 offset:64
-; GFX11-FAKE16-NEXT: scratch_store_b32 off, v57, s32 offset:60
-; GFX11-FAKE16-NEXT: scratch_store_b32 off, v58, s32 offset:56
-; GFX11-FAKE16-NEXT: scratch_store_b32 off, v59, s32 offset:52
-; GFX11-FAKE16-NEXT: scratch_store_b32 off, v60, s32 offset:48
-; GFX11-FAKE16-NEXT: scratch_store_b32 off, v61, s32 offset:44
-; GFX11-FAKE16-NEXT: scratch_store_b32 off, v62, s32 offset:40
-; GFX11-FAKE16-NEXT: scratch_store_b32 off, v63, s32 offset:36
-; GFX11-FAKE16-NEXT: scratch_store_b32 off, v72, s32 offset:32
-; GFX11-FAKE16-NEXT: scratch_store_b32 off, v73, s32 offset:28
-; GFX11-FAKE16-NEXT: scratch_store_b32 off, v74, s32 offset:24
-; GFX11-FAKE16-NEXT: scratch_store_b32 off, v75, s32 offset:20
-; GFX11-FAKE16-NEXT: scratch_store_b32 off, v76, s32 offset:16
-; GFX11-FAKE16-NEXT: scratch_store_b32 off, v77, s32 offset:12
+; GFX11-FAKE16-NEXT: s_clause 0x19
+; GFX11-FAKE16-NEXT: scratch_store_b32 off, v40, s32 offset:112
+; GFX11-FAKE16-NEXT: scratch_store_b32 off, v41, s32 offset:108
+; GFX11-FAKE16-NEXT: scratch_store_b32 off, v42, s32 offset:104
+; GFX11-FAKE16-NEXT: scratch_store_b32 off, v43, s32 offset:100
+; GFX11-FAKE16-NEXT: scratch_store_b32 off, v44, s32 offset:96
+; GFX11-FAKE16-NEXT: scratch_store_b32 off, v45, s32 offset:92
+; GFX11-FAKE16-NEXT: scratch_store_b32 off, v46, s32 offset:88
+; GFX11-FAKE16-NEXT: scratch_store_b32 off, v47, s32 offset:84
+; GFX11-FAKE16-NEXT: scratch_store_b32 off, v56, s32 offset:80
+; GFX11-FAKE16-NEXT: scratch_store_b32 off, v57, s32 offset:76
+; GFX11-FAKE16-NEXT: scratch_store_b32 off, v58, s32 offset:72
+; GFX11-FAKE16-NEXT: scratch_store_b32 off, v59, s32 offset:68
+; GFX11-FAKE16-NEXT: scratch_store_b32 off, v60, s32 offset:64
+; GFX11-FAKE16-NEXT: scratch_store_b32 off, v61, s32 offset:60
+; GFX11-FAKE16-NEXT: scratch_store_b32 off, v62, s32 offset:56
+; GFX11-FAKE16-NEXT: scratch_store_b32 off, v63, s32 offset:52
+; GFX11-FAKE16-NEXT: scratch_store_b32 off, v72, s32 offset:48
+; GFX11-FAKE16-NEXT: scratch_store_b32 off, v73, s32 offset:44
+; GFX11-FAKE16-NEXT: scratch_store_b32 off, v74, s32 offset:40
+; GFX11-FAKE16-NEXT: scratch_store_b32 off, v75, s32 offset:36
+; GFX11-FAKE16-NEXT: scratch_store_b32 off, v76, s32 offset:32
+; GFX11-FAKE16-NEXT: scratch_store_b32 off, v77, s32 offset:28
+; GFX11-FAKE16-NEXT: scratch_store_b32 off, v78, s32 offset:24
+; GFX11-FAKE16-NEXT: scratch_store_b32 off, v79, s32 offset:20
+; GFX11-FAKE16-NEXT: scratch_store_b32 off, v88, s32 offset:16
+; GFX11-FAKE16-NEXT: scratch_store_b32 off, v89, s32 offset:12
; GFX11-FAKE16-NEXT: s_clause 0x2
; GFX11-FAKE16-NEXT: scratch_load_b32 v33, off, s32 offset:8
; GFX11-FAKE16-NEXT: scratch_load_b32 v32, off, s32 offset:4
; GFX11-FAKE16-NEXT: scratch_load_b32 v31, off, s32
+; GFX11-FAKE16-NEXT: ; implicit-def: $vgpr89
+; GFX11-FAKE16-NEXT: ; implicit-def: $vgpr88
+; GFX11-FAKE16-NEXT: ; implicit-def: $vgpr99
+; GFX11-FAKE16-NEXT: ; implicit-def: $vgpr78
+; GFX11-FAKE16-NEXT: ; implicit-def: $vgpr180
; GFX11-FAKE16-NEXT: ; implicit-def: $vgpr76
; GFX11-FAKE16-NEXT: ; implicit-def: $vgpr75
-; GFX11-FAKE16-NEXT: ; implicit-def: $vgpr66
; GFX11-FAKE16-NEXT: ; implicit-def: $vgpr73
+; GFX11-FAKE16-NEXT: ; implicit-def: $vgpr98
+; GFX11-FAKE16-NEXT: ; implicit-def: $vgpr72
; GFX11-FAKE16-NEXT: ; implicit-def: $vgpr166
-; GFX11-FAKE16-NEXT: ; implicit-def: $vgpr63
; GFX11-FAKE16-NEXT: ; implicit-def: $vgpr62
-; GFX11-FAKE16-NEXT: ; implicit-def: $vgpr60
-; GFX11-FAKE16-NEXT: ; implicit-def: $vgpr54
+; GFX11-FAKE16-NEXT: ; implicit-def: $vgpr59
; GFX11-FAKE16-NEXT: ; implicit-def: $vgpr58
-; GFX11-FAKE16-NEXT: ; implicit-def: $vgpr160
+; GFX11-FAKE16-NEXT: ; implicit-def: $vgpr86
; GFX11-FAKE16-NEXT: ; implicit-def: $vgpr56
+; GFX11-FAKE16-NEXT: ; implicit-def: $vgpr161
+; GFX11-FAKE16-NEXT: ; implicit-def: $vgpr47
; GFX11-FAKE16-NEXT: ; implicit-def: $vgpr45
-; GFX11-FAKE16-NEXT: ; implicit-def: $vgpr44
-; GFX11-FAKE16-NEXT: ; implicit-def: $vgpr53
+; GFX11-FAKE16-NEXT: ; implicit-def: $vgpr43
+; GFX11-FAKE16-NEXT: ; implicit-def: $vgpr85
; GFX11-FAKE16-NEXT: ; implicit-def: $vgpr42
-; GFX11-FAKE16-NEXT: ; implicit-def: $vgpr147
+; GFX11-FAKE16-NEXT: ; implicit-def: $vgpr146
; GFX11-FAKE16-NEXT: ; implicit-def: $vgpr41
; GFX11-FAKE16-NEXT: ; implicit-def: $vgpr183
; GFX11-FAKE16-NEXT: ; implicit-def: $vgpr181
-; GFX11-FAKE16-NEXT: ; implicit-def: $vgpr52
-; GFX11-FAKE16-NEXT: ; implicit-def: $vgpr180
-; GFX11-FAKE16-NEXT: ; implicit-def: $vgpr132
+; GFX11-FAKE16-NEXT: ; implicit-def: $vgpr81
; GFX11-FAKE16-NEXT: ; implicit-def: $vgpr179
-; GFX11-FAKE16-NEXT: ; implicit-def: $vgpr177
+; GFX11-FAKE16-NEXT: ; implicit-def: $vgpr133
+; GFX11-FAKE16-NEXT: ; implicit-def: $vgpr178
+; GFX11-FAKE16-NEXT: ; implicit-def: $vgpr176
; GFX11-FAKE16-NEXT: ; implicit-def: $vgpr167
-; GFX11-FAKE16-NEXT: ; implicit-def: $vgpr51
-; GFX11-FAKE16-NEXT: ; implicit-def: $vgpr165
-; GFX11-FAKE16-NEXT: ; implicit-def: $vgpr119
+; GFX11-FAKE16-NEXT: ; implicit-def: $vgpr80
; GFX11-FAKE16-NEXT: ; implicit-def: $vgpr164
+; GFX11-FAKE16-NEXT: ; implicit-def: $vgpr131
+; GFX11-FAKE16-NEXT: ; implicit-def: $vgpr163
; GFX11-FAKE16-NEXT: ; implicit-def: $vgpr162
-; GFX11-FAKE16-NEXT: ; implicit-def: $vgpr161
-; GFX11-FAKE16-NEXT: ; implicit-def: $vgpr50
+; GFX11-FAKE16-NEXT: ; implicit-def: $vgpr151
+; GFX11-FAKE16-NEXT: ; implicit-def: $vgpr67
; GFX11-FAKE16-NEXT: ; implicit-def: $vgpr150
-; GFX11-FAKE16-NEXT: ; implicit-def: $vgpr118
+; GFX11-FAKE16-NEXT: ; implicit-def: $vgpr129
; GFX11-FAKE16-NEXT: ; implicit-def: $vgpr149
; GFX11-FAKE16-NEXT: ; implicit-def: $vgpr148
; GFX11-FAKE16-NEXT: ; implicit-def: $vgpr145
-; GFX11-FAKE16-NEXT: ; implicit-def: $vgpr49
+; GFX11-FAKE16-NEXT: ; implicit-def: $vgpr64
; GFX11-FAKE16-NEXT: ; implicit-def: $vgpr144
-; GFX11-FAKE16-NEXT: ; implicit-def: $vgpr116
+; GFX11-FAKE16-NEXT: ; implicit-def: $vgpr103
; GFX11-FAKE16-NEXT: ; implicit-def: $vgpr135
-; GFX11-FAKE16-NEXT: ; implicit-def: $vgpr134
-; GFX11-FAKE16-NEXT: ; implicit-def: $vgpr131
-; GFX11-FAKE16-NEXT: ; implicit-def: $vgpr48
-; GFX11-FAKE16-NEXT: ; implicit-def: $vgpr130
-; GFX11-FAKE16-NEXT: ; implicit-def: $vgpr114
-; GFX11-FAKE16-NEXT: ; implicit-def: $vgpr129
+; GFX11-FAKE16-NEXT: ; implicit-def: $vgpr79
+; GFX11-FAKE16-NEXT: ; implicit-def: $vgpr77
; GFX11-FAKE16-NEXT: ; implicit-def: $vgpr74
-; GFX11-FAKE16-NEXT: ; implicit-def: $vgpr72
-; GFX11-FAKE16-NEXT: ; implicit-def: $vgpr67
+; GFX11-FAKE16-NEXT: ; implicit-def: $vgpr132
+; GFX11-FAKE16-NEXT: ; implicit-def: $vgpr38
+; GFX11-FAKE16-NEXT: ; implicit-def: $vgpr63
; GFX11-FAKE16-NEXT: ; implicit-def: $vgpr61
-; GFX11-FAKE16-NEXT: ; implicit-def: $vgpr69
-; GFX11-FAKE16-NEXT: ; implicit-def: $vgpr59
+; GFX11-FAKE16-NEXT: ; implicit-def: $vgpr36
+; GFX11-FAKE16-NEXT: ; implicit-def: $vgpr60
+; GFX11-FAKE16-NEXT: ; implicit-def: $vgpr160
; GFX11-FAKE16-NEXT: ; implicit-def: $vgpr57
-; GFX11-FAKE16-NEXT: ; implicit-def: $vgpr47
-; GFX11-FAKE16-NEXT: ; implicit-def: $vgpr64
+; GFX11-FAKE16-NEXT: ; implicit-def: $vgpr52
; GFX11-FAKE16-NEXT: ; implicit-def: $vgpr46
-; GFX11-FAKE16-NEXT: ; implicit-def: $vgpr146
-; GFX11-FAKE16-NEXT: ; implicit-def: $vgpr43
-; GFX11-FAKE16-NEXT: ; implicit-def: $vgpr70
+; GFX11-FAKE16-NEXT: ; implicit-def: $vgpr65
+; GFX11-FAKE16-NEXT: ; implicit-def: $vgpr53
+; GFX11-FAKE16-NEXT: ; implicit-def: $vgpr147
+; GFX11-FAKE16-NEXT: ; implicit-def: $vgpr44
+; GFX11-FAKE16-NEXT: ; implicit-def: $vgpr69
; GFX11-FAKE16-NEXT: ; implicit-def: $vgpr40
-; GFX11-FAKE16-NEXT: ; implicit-def: $vgpr38
-; GFX11-FAKE16-NEXT: ; implicit-def: $vgpr71
-; GFX11-FAKE16-NEXT: ; implicit-def: $vgpr133
+; GFX11-FAKE16-NEXT: ; implicit-def: $vgpr54
+; GFX11-FAKE16-NEXT: ; implicit-def: $vgpr70
+; GFX11-FAKE16-NEXT: ; implicit-def: $vgpr134
; GFX11-FAKE16-NEXT: ; implicit-def: $vgpr182
-; GFX11-FAKE16-NEXT: ; implicit-def: $vgpr80
-; GFX11-FAKE16-NEXT: ; implicit-def: $vgpr178
-; GFX11-FAKE16-NEXT: ; implicit-def: $vgpr37
-; GFX11-FAKE16-NEXT: ; implicit-def: $vgpr81
-; GFX11-FAKE16-NEXT: ; implicit-def: $vgpr128
-; GFX11-FAKE16-NEXT: ; implicit-def: $vgpr176
-; GFX11-FAKE16-NEXT: ; implicit-def: $vgpr82
-; GFX11-FAKE16-NEXT: ; implicit-def: $vgpr163
-; GFX11-FAKE16-NEXT: ; implicit-def: $vgpr36
; GFX11-FAKE16-NEXT: ; implicit-def: $vgpr83
-; GFX11-FAKE16-NEXT: ; implicit-def: $vgpr117
-; GFX11-FAKE16-NEXT: ; implicit-def: $vgpr151
+; GFX11-FAKE16-NEXT: ; implicit-def: $vgpr177
+; GFX11-FAKE16-NEXT: ; implicit-def: $vgpr50
; GFX11-FAKE16-NEXT: ; implicit-def: $vgpr84
-; GFX11-FAKE16-NEXT: ; implicit-def: $vgpr103
-; GFX11-FAKE16-NEXT: ; implicit-def: $vgpr35
-; GFX11-FAKE16-NEXT: ; implicit-def: $vgpr85
-; GFX11-FAKE16-NEXT: ; implicit-def: $vgpr115
-; GFX11-FAKE16-NEXT: ; implicit-def: $vgpr102
-; GFX11-FAKE16-NEXT: ; implicit-def: $vgpr86
-; GFX11-FAKE16-NEXT: ; implicit-def: $vgpr101
-; GFX11-FAKE16-NEXT: ; implicit-def: $vgpr87
-; GFX11-FAKE16-NEXT: ; implicit-def: $vgpr113
-; GFX11-FAKE16-NEXT: ; implicit-def: $vgpr100
+; GFX11-FAKE16-NEXT: ; implicit-def: $vgpr130
+; GFX11-FAKE16-NEXT: ; implicit-def: $vgpr165
; GFX11-FAKE16-NEXT: ; implicit-def: $vgpr96
-; GFX11-FAKE16-NEXT: ; implicit-def: $vgpr99
+; GFX11-FAKE16-NEXT: ; implicit-def: $vgpr119
+; GFX11-FAKE16-NEXT: ; implicit-def: $vgpr49
; GFX11-FAKE16-NEXT: ; implicit-def: $vgpr97
+; GFX11-FAKE16-NEXT: ; implicit-def: $vgpr128
+; GFX11-FAKE16-NEXT: ; implicit-def: $vgpr118
+; GFX11-FAKE16-NEXT: ; implicit-def: $vgpr101
+; GFX11-FAKE16-NEXT: ; implicit-def: $vgpr117
+; GFX11-FAKE16-NEXT: ; implicit-def: $vgpr102
+; GFX11-FAKE16-NEXT: ; implicit-def: $vgpr71
+; GFX11-FAKE16-NEXT: ; implicit-def: $vgpr116
; GFX11-FAKE16-NEXT: ; implicit-def: $vgpr112
-; GFX11-FAKE16-NEXT: ; implicit-def: $vgpr98
-; GFX11-FAKE16-NEXT: ; implicit-def: $vgpr34
+; GFX11-FAKE16-NEXT: ; implicit-def: $vgpr115
+; GFX11-FAKE16-NEXT: ; implicit-def: $vgpr113
+; GFX11-FAKE16-NEXT: ; implicit-def: $vgpr39
+; GFX11-FAKE16-NEXT: ; implicit-def: $vgpr114
+; GFX11-FAKE16-NEXT: ; implicit-def: $vgpr48
+; GFX11-FAKE16-NEXT: ; implicit-def: $vgpr35
; GFX11-FAKE16-NEXT: s_waitcnt vmcnt(2)
; GFX11-FAKE16-NEXT: v_cmp_ne_u32_e32 vcc_lo, 0, v33
; GFX11-FAKE16-NEXT: ; implicit-def: $vgpr33
@@ -169108,863 +169191,867 @@ define <128 x i8> @bitcast_v64bf16_to_v128i8(<64 x bfloat> %a, i32 %b) {
; GFX11-FAKE16-NEXT: s_xor_b32 s0, exec_lo, s0
; GFX11-FAKE16-NEXT: s_cbranch_execz .LBB90_2
; GFX11-FAKE16-NEXT: ; %bb.1: ; %cmp.false
-; GFX11-FAKE16-NEXT: v_lshrrev_b64 v[48:49], 24, v[15:16]
-; GFX11-FAKE16-NEXT: v_lshrrev_b64 v[49:50], 24, v[13:14]
+; GFX11-FAKE16-NEXT: v_lshrrev_b64 v[48:49], 24, v[29:30]
+; GFX11-FAKE16-NEXT: v_lshrrev_b64 v[64:65], 24, v[15:16]
+; GFX11-FAKE16-NEXT: v_lshrrev_b64 v[80:81], 24, v[11:12]
+; GFX11-FAKE16-NEXT: v_lshrrev_b64 v[85:86], 24, v[7:8]
+; GFX11-FAKE16-NEXT: v_lshrrev_b64 v[98:99], 24, v[3:4]
; GFX11-FAKE16-NEXT: s_waitcnt vmcnt(0)
-; GFX11-FAKE16-NEXT: v_lshrrev_b64 v[33:34], 24, v[31:32]
-; GFX11-FAKE16-NEXT: v_lshrrev_b64 v[50:51], 24, v[11:12]
-; GFX11-FAKE16-NEXT: v_lshrrev_b64 v[34:35], 24, v[29:30]
-; GFX11-FAKE16-NEXT: v_lshrrev_b64 v[51:52], 24, v[9:10]
-; GFX11-FAKE16-NEXT: v_lshrrev_b64 v[35:36], 24, v[27:28]
-; GFX11-FAKE16-NEXT: v_lshrrev_b64 v[52:53], 24, v[7:8]
-; GFX11-FAKE16-NEXT: v_lshrrev_b64 v[36:37], 24, v[25:26]
-; GFX11-FAKE16-NEXT: v_lshrrev_b64 v[53:54], 24, v[5:6]
-; GFX11-FAKE16-NEXT: v_lshrrev_b64 v[66:67], 24, v[1:2]
-; GFX11-FAKE16-NEXT: v_lshrrev_b64 v[37:38], 24, v[23:24]
-; GFX11-FAKE16-NEXT: v_lshrrev_b32_e32 v129, 24, v16
-; GFX11-FAKE16-NEXT: v_lshrrev_b32_e32 v114, 16, v16
-; GFX11-FAKE16-NEXT: v_lshrrev_b32_e32 v130, 8, v16
-; GFX11-FAKE16-NEXT: v_lshrrev_b32_e32 v131, 16, v15
-; GFX11-FAKE16-NEXT: v_lshrrev_b32_e32 v134, 8, v15
-; GFX11-FAKE16-NEXT: v_lshrrev_b32_e32 v135, 24, v14
-; GFX11-FAKE16-NEXT: v_lshrrev_b32_e32 v116, 16, v14
-; GFX11-FAKE16-NEXT: v_lshrrev_b32_e32 v144, 8, v14
-; GFX11-FAKE16-NEXT: v_lshrrev_b32_e32 v145, 16, v13
-; GFX11-FAKE16-NEXT: v_lshrrev_b32_e32 v148, 8, v13
-; GFX11-FAKE16-NEXT: v_lshrrev_b32_e32 v149, 24, v12
-; GFX11-FAKE16-NEXT: v_lshrrev_b32_e32 v118, 16, v12
-; GFX11-FAKE16-NEXT: v_lshrrev_b32_e32 v150, 8, v12
-; GFX11-FAKE16-NEXT: v_lshrrev_b32_e32 v161, 16, v11
-; GFX11-FAKE16-NEXT: v_lshrrev_b32_e32 v162, 8, v11
-; GFX11-FAKE16-NEXT: v_lshrrev_b32_e32 v164, 24, v10
-; GFX11-FAKE16-NEXT: v_lshrrev_b32_e32 v119, 16, v10
-; GFX11-FAKE16-NEXT: v_lshrrev_b32_e32 v165, 8, v10
-; GFX11-FAKE16-NEXT: v_lshrrev_b32_e32 v167, 16, v9
-; GFX11-FAKE16-NEXT: v_lshrrev_b32_e32 v177, 8, v9
-; GFX11-FAKE16-NEXT: v_lshrrev_b32_e32 v179, 24, v8
-; GFX11-FAKE16-NEXT: v_lshrrev_b32_e32 v132, 16, v8
-; GFX11-FAKE16-NEXT: v_lshrrev_b32_e32 v180, 8, v8
-; GFX11-FAKE16-NEXT: v_lshrrev_b32_e32 v181, 16, v7
-; GFX11-FAKE16-NEXT: v_lshrrev_b32_e32 v183, 8, v7
-; GFX11-FAKE16-NEXT: v_lshrrev_b32_e32 v41, 24, v6
-; GFX11-FAKE16-NEXT: v_lshrrev_b32_e32 v147, 16, v6
-; GFX11-FAKE16-NEXT: v_lshrrev_b32_e32 v42, 8, v6
-; GFX11-FAKE16-NEXT: v_lshrrev_b32_e32 v44, 16, v5
-; GFX11-FAKE16-NEXT: v_lshrrev_b32_e32 v45, 8, v5
-; GFX11-FAKE16-NEXT: v_lshrrev_b32_e32 v56, 24, v4
-; GFX11-FAKE16-NEXT: v_lshrrev_b32_e32 v160, 16, v4
-; GFX11-FAKE16-NEXT: v_lshrrev_b32_e32 v58, 8, v4
-; GFX11-FAKE16-NEXT: v_lshrrev_b32_e32 v60, 16, v3
-; GFX11-FAKE16-NEXT: v_lshrrev_b32_e32 v62, 8, v3
-; GFX11-FAKE16-NEXT: v_lshrrev_b32_e32 v63, 24, v2
-; GFX11-FAKE16-NEXT: v_lshrrev_b32_e32 v166, 16, v2
-; GFX11-FAKE16-NEXT: v_lshrrev_b32_e32 v73, 8, v2
-; GFX11-FAKE16-NEXT: v_lshrrev_b32_e32 v75, 16, v1
-; GFX11-FAKE16-NEXT: v_lshrrev_b32_e32 v76, 8, v1
-; GFX11-FAKE16-NEXT: v_lshrrev_b32_e32 v98, 24, v32
-; GFX11-FAKE16-NEXT: v_lshrrev_b32_e32 v112, 16, v32
-; GFX11-FAKE16-NEXT: v_lshrrev_b32_e32 v97, 8, v32
-; GFX11-FAKE16-NEXT: v_lshrrev_b32_e32 v99, 16, v31
-; GFX11-FAKE16-NEXT: v_lshrrev_b32_e32 v96, 8, v31
-; GFX11-FAKE16-NEXT: v_lshrrev_b32_e32 v100, 24, v30
-; GFX11-FAKE16-NEXT: v_lshrrev_b32_e32 v113, 16, v30
-; GFX11-FAKE16-NEXT: v_lshrrev_b32_e32 v87, 8, v30
-; GFX11-FAKE16-NEXT: v_lshrrev_b32_e32 v101, 16, v29
-; GFX11-FAKE16-NEXT: v_lshrrev_b32_e32 v86, 8, v29
-; GFX11-FAKE16-NEXT: v_lshrrev_b32_e32 v102, 24, v28
-; GFX11-FAKE16-NEXT: v_lshrrev_b32_e32 v115, 16, v28
-; GFX11-FAKE16-NEXT: v_lshrrev_b32_e32 v85, 8, v28
-; GFX11-FAKE16-NEXT: v_lshrrev_b32_e32 v103, 16, v27
-; GFX11-FAKE16-NEXT: v_lshrrev_b32_e32 v84, 8, v27
-; GFX11-FAKE16-NEXT: v_lshrrev_b32_e32 v151, 24, v26
-; GFX11-FAKE16-NEXT: v_lshrrev_b32_e32 v117, 16, v26
-; GFX11-FAKE16-NEXT: v_lshrrev_b32_e32 v83, 8, v26
-; GFX11-FAKE16-NEXT: v_lshrrev_b32_e32 v163, 16, v25
-; GFX11-FAKE16-NEXT: v_lshrrev_b32_e32 v82, 8, v25
-; GFX11-FAKE16-NEXT: v_lshrrev_b32_e32 v176, 24, v24
-; GFX11-FAKE16-NEXT: v_lshrrev_b32_e32 v128, 16, v24
-; GFX11-FAKE16-NEXT: v_lshrrev_b32_e32 v81, 8, v24
-; GFX11-FAKE16-NEXT: v_lshrrev_b32_e32 v178, 16, v23
-; GFX11-FAKE16-NEXT: v_lshrrev_b32_e32 v80, 8, v23
-; GFX11-FAKE16-NEXT: v_lshrrev_b32_e32 v182, 24, v22
-; GFX11-FAKE16-NEXT: v_lshrrev_b32_e32 v133, 16, v22
-; GFX11-FAKE16-NEXT: v_lshrrev_b32_e32 v71, 8, v22
-; GFX11-FAKE16-NEXT: v_lshrrev_b32_e32 v40, 16, v21
-; GFX11-FAKE16-NEXT: v_lshrrev_b32_e32 v70, 8, v21
-; GFX11-FAKE16-NEXT: v_lshrrev_b32_e32 v43, 24, v20
-; GFX11-FAKE16-NEXT: v_lshrrev_b32_e32 v146, 16, v20
-; GFX11-FAKE16-NEXT: v_lshrrev_b32_e32 v46, 8, v20
-; GFX11-FAKE16-NEXT: v_lshrrev_b32_e32 v47, 16, v19
-; GFX11-FAKE16-NEXT: v_lshrrev_b32_e32 v57, 8, v19
-; GFX11-FAKE16-NEXT: v_lshrrev_b32_e32 v59, 24, v18
-; GFX11-FAKE16-NEXT: v_lshrrev_b32_e32 v69, 16, v18
-; GFX11-FAKE16-NEXT: v_lshrrev_b32_e32 v61, 8, v18
-; GFX11-FAKE16-NEXT: v_lshrrev_b32_e32 v72, 16, v17
-; GFX11-FAKE16-NEXT: v_lshrrev_b32_e32 v74, 8, v17
-; GFX11-FAKE16-NEXT: v_lshrrev_b64 v[54:55], 24, v[3:4]
-; GFX11-FAKE16-NEXT: v_lshrrev_b64 v[38:39], 24, v[21:22]
-; GFX11-FAKE16-NEXT: v_lshrrev_b64 v[64:65], 24, v[19:20]
-; GFX11-FAKE16-NEXT: v_lshrrev_b64 v[67:68], 24, v[17:18]
+; GFX11-FAKE16-NEXT: v_lshrrev_b64 v[35:36], 24, v[31:32]
+; GFX11-FAKE16-NEXT: v_lshrrev_b64 v[49:50], 24, v[27:28]
+; GFX11-FAKE16-NEXT: v_lshrrev_b32_e32 v135, 24, v16
+; GFX11-FAKE16-NEXT: v_lshrrev_b32_e32 v103, 16, v16
+; GFX11-FAKE16-NEXT: v_lshrrev_b32_e32 v144, 8, v16
+; GFX11-FAKE16-NEXT: v_lshrrev_b32_e32 v145, 16, v15
+; GFX11-FAKE16-NEXT: v_lshrrev_b32_e32 v148, 8, v15
+; GFX11-FAKE16-NEXT: v_lshrrev_b32_e32 v149, 24, v14
+; GFX11-FAKE16-NEXT: v_lshrrev_b32_e32 v129, 16, v14
+; GFX11-FAKE16-NEXT: v_lshrrev_b32_e32 v150, 8, v14
+; GFX11-FAKE16-NEXT: v_lshrrev_b32_e32 v151, 16, v13
+; GFX11-FAKE16-NEXT: v_lshrrev_b32_e32 v162, 8, v13
+; GFX11-FAKE16-NEXT: v_lshrrev_b32_e32 v163, 24, v12
+; GFX11-FAKE16-NEXT: v_lshrrev_b32_e32 v131, 16, v12
+; GFX11-FAKE16-NEXT: v_lshrrev_b32_e32 v164, 8, v12
+; GFX11-FAKE16-NEXT: v_lshrrev_b32_e32 v167, 16, v11
+; GFX11-FAKE16-NEXT: v_lshrrev_b32_e32 v176, 8, v11
+; GFX11-FAKE16-NEXT: v_lshrrev_b32_e32 v178, 24, v10
+; GFX11-FAKE16-NEXT: v_lshrrev_b32_e32 v133, 16, v10
+; GFX11-FAKE16-NEXT: v_lshrrev_b32_e32 v179, 8, v10
+; GFX11-FAKE16-NEXT: v_lshrrev_b32_e32 v181, 16, v9
+; GFX11-FAKE16-NEXT: v_lshrrev_b32_e32 v183, 8, v9
+; GFX11-FAKE16-NEXT: v_lshrrev_b32_e32 v41, 24, v8
+; GFX11-FAKE16-NEXT: v_lshrrev_b32_e32 v146, 16, v8
+; GFX11-FAKE16-NEXT: v_lshrrev_b32_e32 v42, 8, v8
+; GFX11-FAKE16-NEXT: v_lshrrev_b32_e32 v43, 16, v7
+; GFX11-FAKE16-NEXT: v_lshrrev_b32_e32 v45, 8, v7
+; GFX11-FAKE16-NEXT: v_lshrrev_b32_e32 v47, 24, v6
+; GFX11-FAKE16-NEXT: v_lshrrev_b32_e32 v161, 16, v6
+; GFX11-FAKE16-NEXT: v_lshrrev_b32_e32 v56, 8, v6
+; GFX11-FAKE16-NEXT: v_lshrrev_b32_e32 v58, 16, v5
+; GFX11-FAKE16-NEXT: v_lshrrev_b32_e32 v59, 8, v5
+; GFX11-FAKE16-NEXT: v_lshrrev_b32_e32 v62, 24, v4
+; GFX11-FAKE16-NEXT: v_lshrrev_b32_e32 v166, 16, v4
+; GFX11-FAKE16-NEXT: v_lshrrev_b32_e32 v72, 8, v4
+; GFX11-FAKE16-NEXT: v_lshrrev_b32_e32 v73, 16, v3
+; GFX11-FAKE16-NEXT: v_lshrrev_b32_e32 v75, 8, v3
+; GFX11-FAKE16-NEXT: v_lshrrev_b32_e32 v76, 24, v2
+; GFX11-FAKE16-NEXT: v_lshrrev_b32_e32 v180, 16, v2
+; GFX11-FAKE16-NEXT: v_lshrrev_b32_e32 v78, 8, v2
+; GFX11-FAKE16-NEXT: v_lshrrev_b32_e32 v88, 16, v1
+; GFX11-FAKE16-NEXT: v_lshrrev_b32_e32 v89, 8, v1
+; GFX11-FAKE16-NEXT: v_lshrrev_b32_e32 v114, 24, v32
+; GFX11-FAKE16-NEXT: v_lshrrev_b32_e32 v39, 16, v32
+; GFX11-FAKE16-NEXT: v_lshrrev_b32_e32 v113, 8, v32
+; GFX11-FAKE16-NEXT: v_lshrrev_b32_e32 v115, 16, v31
+; GFX11-FAKE16-NEXT: v_lshrrev_b32_e32 v112, 8, v31
+; GFX11-FAKE16-NEXT: v_lshrrev_b32_e32 v116, 24, v30
+; GFX11-FAKE16-NEXT: v_lshrrev_b32_e32 v71, 16, v30
+; GFX11-FAKE16-NEXT: v_lshrrev_b32_e32 v102, 8, v30
+; GFX11-FAKE16-NEXT: v_lshrrev_b32_e32 v117, 16, v29
+; GFX11-FAKE16-NEXT: v_lshrrev_b32_e32 v101, 8, v29
+; GFX11-FAKE16-NEXT: v_lshrrev_b32_e32 v118, 24, v28
+; GFX11-FAKE16-NEXT: v_lshrrev_b32_e32 v128, 16, v28
+; GFX11-FAKE16-NEXT: v_lshrrev_b32_e32 v97, 8, v28
+; GFX11-FAKE16-NEXT: v_lshrrev_b32_e32 v119, 16, v27
+; GFX11-FAKE16-NEXT: v_lshrrev_b32_e32 v96, 8, v27
+; GFX11-FAKE16-NEXT: v_lshrrev_b32_e32 v165, 24, v26
+; GFX11-FAKE16-NEXT: v_lshrrev_b32_e32 v130, 16, v26
+; GFX11-FAKE16-NEXT: v_lshrrev_b32_e32 v84, 8, v26
+; GFX11-FAKE16-NEXT: v_lshrrev_b32_e32 v177, 16, v25
+; GFX11-FAKE16-NEXT: v_lshrrev_b32_e32 v83, 8, v25
+; GFX11-FAKE16-NEXT: v_lshrrev_b32_e32 v182, 24, v24
+; GFX11-FAKE16-NEXT: v_lshrrev_b32_e32 v134, 16, v24
+; GFX11-FAKE16-NEXT: v_lshrrev_b32_e32 v70, 8, v24
+; GFX11-FAKE16-NEXT: v_lshrrev_b32_e32 v40, 16, v23
+; GFX11-FAKE16-NEXT: v_lshrrev_b32_e32 v69, 8, v23
+; GFX11-FAKE16-NEXT: v_lshrrev_b32_e32 v44, 24, v22
+; GFX11-FAKE16-NEXT: v_lshrrev_b32_e32 v147, 16, v22
+; GFX11-FAKE16-NEXT: v_lshrrev_b32_e32 v53, 8, v22
+; GFX11-FAKE16-NEXT: v_lshrrev_b32_e32 v46, 16, v21
+; GFX11-FAKE16-NEXT: v_lshrrev_b32_e32 v52, 8, v21
+; GFX11-FAKE16-NEXT: v_lshrrev_b32_e32 v57, 24, v20
+; GFX11-FAKE16-NEXT: v_lshrrev_b32_e32 v160, 16, v20
+; GFX11-FAKE16-NEXT: v_lshrrev_b32_e32 v60, 8, v20
+; GFX11-FAKE16-NEXT: v_lshrrev_b32_e32 v61, 16, v19
+; GFX11-FAKE16-NEXT: v_lshrrev_b32_e32 v63, 8, v19
+; GFX11-FAKE16-NEXT: v_lshrrev_b32_e32 v38, 24, v18
+; GFX11-FAKE16-NEXT: v_lshrrev_b32_e32 v132, 16, v18
+; GFX11-FAKE16-NEXT: v_lshrrev_b32_e32 v74, 8, v18
+; GFX11-FAKE16-NEXT: v_lshrrev_b32_e32 v77, 16, v17
+; GFX11-FAKE16-NEXT: v_lshrrev_b32_e32 v79, 8, v17
+; GFX11-FAKE16-NEXT: v_lshrrev_b64 v[67:68], 24, v[13:14]
+; GFX11-FAKE16-NEXT: v_lshrrev_b64 v[81:82], 24, v[9:10]
+; GFX11-FAKE16-NEXT: v_lshrrev_b64 v[86:87], 24, v[5:6]
+; GFX11-FAKE16-NEXT: v_lshrrev_b64 v[99:100], 24, v[1:2]
+; GFX11-FAKE16-NEXT: v_lshrrev_b64 v[50:51], 24, v[25:26]
+; GFX11-FAKE16-NEXT: v_lshrrev_b64 v[54:55], 24, v[23:24]
+; GFX11-FAKE16-NEXT: v_lshrrev_b64 v[65:66], 24, v[21:22]
+; GFX11-FAKE16-NEXT: v_lshrrev_b64 v[36:37], 24, v[19:20]
+; GFX11-FAKE16-NEXT: v_lshrrev_b64 v[33:34], 24, v[17:18]
; GFX11-FAKE16-NEXT: .LBB90_2: ; %Flow
; GFX11-FAKE16-NEXT: s_and_not1_saveexec_b32 s0, s0
; GFX11-FAKE16-NEXT: s_cbranch_execz .LBB90_4
; GFX11-FAKE16-NEXT: ; %bb.3: ; %cmp.true
-; GFX11-FAKE16-NEXT: v_lshlrev_b32_e32 v33, 16, v18
; GFX11-FAKE16-NEXT: v_lshlrev_b32_e32 v34, 16, v17
; GFX11-FAKE16-NEXT: v_and_b32_e32 v17, 0xffff0000, v17
+; GFX11-FAKE16-NEXT: v_lshlrev_b32_e32 v33, 16, v18
+; GFX11-FAKE16-NEXT: s_delay_alu instid0(VALU_DEP_3) | instskip(NEXT) | instid1(VALU_DEP_3)
+; GFX11-FAKE16-NEXT: v_add_f32_e32 v34, 0x40c00000, v34
+; GFX11-FAKE16-NEXT: v_add_f32_e32 v36, 0x40c00000, v17
; GFX11-FAKE16-NEXT: s_delay_alu instid0(VALU_DEP_3) | instskip(NEXT) | instid1(VALU_DEP_2)
; GFX11-FAKE16-NEXT: v_add_f32_e32 v33, 0x40c00000, v33
-; GFX11-FAKE16-NEXT: v_add_f32_e32 v35, 0x40c00000, v17
+; GFX11-FAKE16-NEXT: v_bfe_u32 v48, v36, 16, 1
; GFX11-FAKE16-NEXT: s_delay_alu instid0(VALU_DEP_2) | instskip(SKIP_2) | instid1(VALU_DEP_3)
; GFX11-FAKE16-NEXT: v_bfe_u32 v17, v33, 16, 1
-; GFX11-FAKE16-NEXT: v_or_b32_e32 v36, 0x400000, v33
+; GFX11-FAKE16-NEXT: v_or_b32_e32 v35, 0x400000, v33
; GFX11-FAKE16-NEXT: v_cmp_u_f32_e32 vcc_lo, v33, v33
; GFX11-FAKE16-NEXT: v_add3_u32 v17, v17, v33, 0x7fff
-; GFX11-FAKE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
-; GFX11-FAKE16-NEXT: v_dual_cndmask_b32 v17, v17, v36 :: v_dual_and_b32 v18, 0xffff0000, v18
-; GFX11-FAKE16-NEXT: v_add_f32_e32 v18, 0x40c00000, v18
-; GFX11-FAKE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(SKIP_2) | instid1(VALU_DEP_3)
+; GFX11-FAKE16-NEXT: v_and_b32_e32 v18, 0xffff0000, v18
+; GFX11-FAKE16-NEXT: v_add3_u32 v33, v48, v36, 0x7fff
+; GFX11-FAKE16-NEXT: s_delay_alu instid0(VALU_DEP_2) | instskip(NEXT) | instid1(VALU_DEP_1)
+; GFX11-FAKE16-NEXT: v_dual_cndmask_b32 v17, v17, v35 :: v_dual_add_f32 v18, 0x40c00000, v18
; GFX11-FAKE16-NEXT: v_bfe_u32 v37, v18, 16, 1
; GFX11-FAKE16-NEXT: v_or_b32_e32 v39, 0x400000, v18
; GFX11-FAKE16-NEXT: v_cmp_u_f32_e32 vcc_lo, v18, v18
+; GFX11-FAKE16-NEXT: s_delay_alu instid0(VALU_DEP_3) | instskip(NEXT) | instid1(VALU_DEP_1)
; GFX11-FAKE16-NEXT: v_add3_u32 v37, v37, v18, 0x7fff
-; GFX11-FAKE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(SKIP_1) | instid1(VALU_DEP_2)
-; GFX11-FAKE16-NEXT: v_dual_cndmask_b32 v77, v37, v39 :: v_dual_add_f32 v34, 0x40c00000, v34
-; GFX11-FAKE16-NEXT: v_lshlrev_b32_e32 v37, 16, v20
-; GFX11-FAKE16-NEXT: v_perm_b32 v69, v77, v17, 0x7060302
-; GFX11-FAKE16-NEXT: s_delay_alu instid0(VALU_DEP_3) | instskip(SKIP_2) | instid1(VALU_DEP_4)
+; GFX11-FAKE16-NEXT: v_dual_cndmask_b32 v132, v37, v39 :: v_dual_lshlrev_b32 v37, 16, v20
; GFX11-FAKE16-NEXT: v_bfe_u32 v38, v34, 16, 1
; GFX11-FAKE16-NEXT: v_or_b32_e32 v18, 0x400000, v34
; GFX11-FAKE16-NEXT: v_cmp_u_f32_e32 vcc_lo, v34, v34
-; GFX11-FAKE16-NEXT: v_lshrrev_b32_e32 v59, 24, v69
-; GFX11-FAKE16-NEXT: s_delay_alu instid0(VALU_DEP_4) | instskip(SKIP_2) | instid1(VALU_DEP_2)
+; GFX11-FAKE16-NEXT: v_or_b32_e32 v39, 0x400000, v36
+; GFX11-FAKE16-NEXT: s_delay_alu instid0(VALU_DEP_4) | instskip(SKIP_2) | instid1(VALU_DEP_3)
; GFX11-FAKE16-NEXT: v_add3_u32 v38, v38, v34, 0x7fff
+; GFX11-FAKE16-NEXT: v_perm_b32 v34, v132, v17, 0x7060302
+; GFX11-FAKE16-NEXT: v_lshrrev_b32_e32 v132, 16, v132
+; GFX11-FAKE16-NEXT: v_dual_cndmask_b32 v35, v38, v18 :: v_dual_add_f32 v18, 0x40c00000, v37
+; GFX11-FAKE16-NEXT: v_lshlrev_b32_e32 v37, 16, v19
+; GFX11-FAKE16-NEXT: v_cmp_u_f32_e32 vcc_lo, v36, v36
; GFX11-FAKE16-NEXT: v_and_b32_e32 v20, 0xffff0000, v20
-; GFX11-FAKE16-NEXT: v_lshrrev_b32_e32 v61, 8, v69
-; GFX11-FAKE16-NEXT: v_dual_cndmask_b32 v33, v38, v18 :: v_dual_add_f32 v20, 0x40c00000, v20
-; GFX11-FAKE16-NEXT: v_bfe_u32 v48, v35, 16, 1
-; GFX11-FAKE16-NEXT: v_or_b32_e32 v39, 0x400000, v35
-; GFX11-FAKE16-NEXT: v_add_f32_e32 v18, 0x40c00000, v37
-; GFX11-FAKE16-NEXT: v_cmp_u_f32_e32 vcc_lo, v35, v35
-; GFX11-FAKE16-NEXT: s_delay_alu instid0(VALU_DEP_4) | instskip(NEXT) | instid1(VALU_DEP_3)
-; GFX11-FAKE16-NEXT: v_add3_u32 v36, v48, v35, 0x7fff
-; GFX11-FAKE16-NEXT: v_bfe_u32 v35, v18, 16, 1
-; GFX11-FAKE16-NEXT: v_or_b32_e32 v37, 0x400000, v18
-; GFX11-FAKE16-NEXT: s_delay_alu instid0(VALU_DEP_3) | instskip(SKIP_1) | instid1(VALU_DEP_4)
-; GFX11-FAKE16-NEXT: v_cndmask_b32_e32 v34, v36, v39, vcc_lo
-; GFX11-FAKE16-NEXT: v_lshlrev_b32_e32 v36, 16, v19
-; GFX11-FAKE16-NEXT: v_add3_u32 v35, v35, v18, 0x7fff
-; GFX11-FAKE16-NEXT: v_cmp_u_f32_e32 vcc_lo, v18, v18
; GFX11-FAKE16-NEXT: v_and_b32_e32 v19, 0xffff0000, v19
-; GFX11-FAKE16-NEXT: v_perm_b32 v68, v34, v33, 0x7060302
-; GFX11-FAKE16-NEXT: v_bfe_u32 v34, v20, 16, 1
-; GFX11-FAKE16-NEXT: v_cndmask_b32_e32 v18, v35, v37, vcc_lo
-; GFX11-FAKE16-NEXT: s_delay_alu instid0(VALU_DEP_4) | instskip(NEXT) | instid1(VALU_DEP_3)
-; GFX11-FAKE16-NEXT: v_dual_add_f32 v38, 0x40c00000, v19 :: v_dual_lshlrev_b32 v37, 16, v22
-; GFX11-FAKE16-NEXT: v_add3_u32 v34, v34, v20, 0x7fff
+; GFX11-FAKE16-NEXT: v_bfe_u32 v36, v18, 16, 1
+; GFX11-FAKE16-NEXT: v_add_f32_e32 v37, 0x40c00000, v37
+; GFX11-FAKE16-NEXT: s_delay_alu instid0(VALU_DEP_4) | instskip(SKIP_1) | instid1(VALU_DEP_4)
+; GFX11-FAKE16-NEXT: v_dual_cndmask_b32 v33, v33, v39 :: v_dual_add_f32 v20, 0x40c00000, v20
+; GFX11-FAKE16-NEXT: v_or_b32_e32 v39, 0x400000, v18
+; GFX11-FAKE16-NEXT: v_add3_u32 v36, v36, v18, 0x7fff
+; GFX11-FAKE16-NEXT: v_cmp_u_f32_e32 vcc_lo, v18, v18
+; GFX11-FAKE16-NEXT: v_add_f32_e32 v48, 0x40c00000, v19
+; GFX11-FAKE16-NEXT: v_bfe_u32 v38, v20, 16, 1
; GFX11-FAKE16-NEXT: v_or_b32_e32 v19, 0x400000, v20
+; GFX11-FAKE16-NEXT: v_bfe_u32 v49, v37, 16, 1
+; GFX11-FAKE16-NEXT: v_cndmask_b32_e32 v18, v36, v39, vcc_lo
; GFX11-FAKE16-NEXT: v_cmp_u_f32_e32 vcc_lo, v20, v20
-; GFX11-FAKE16-NEXT: v_add_f32_e32 v36, 0x40c00000, v36
-; GFX11-FAKE16-NEXT: v_bfe_u32 v35, v38, 16, 1
-; GFX11-FAKE16-NEXT: v_lshrrev_b32_e32 v72, 16, v68
-; GFX11-FAKE16-NEXT: v_lshrrev_b32_e32 v74, 8, v68
-; GFX11-FAKE16-NEXT: v_cndmask_b32_e32 v19, v34, v19, vcc_lo
-; GFX11-FAKE16-NEXT: v_bfe_u32 v39, v36, 16, 1
-; GFX11-FAKE16-NEXT: v_or_b32_e32 v34, 0x400000, v36
-; GFX11-FAKE16-NEXT: v_cmp_u_f32_e32 vcc_lo, v36, v36
-; GFX11-FAKE16-NEXT: v_add3_u32 v35, v35, v38, 0x7fff
-; GFX11-FAKE16-NEXT: v_perm_b32 v65, v19, v18, 0x7060302
-; GFX11-FAKE16-NEXT: v_add3_u32 v20, v39, v36, 0x7fff
-; GFX11-FAKE16-NEXT: v_or_b32_e32 v39, 0x400000, v38
-; GFX11-FAKE16-NEXT: v_lshrrev_b32_e32 v146, 16, v19
-; GFX11-FAKE16-NEXT: s_delay_alu instid0(VALU_DEP_4) | instskip(NEXT) | instid1(VALU_DEP_4)
-; GFX11-FAKE16-NEXT: v_lshrrev_b32_e32 v43, 24, v65
-; GFX11-FAKE16-NEXT: v_cndmask_b32_e32 v34, v20, v34, vcc_lo
-; GFX11-FAKE16-NEXT: v_add_f32_e32 v20, 0x40c00000, v37
-; GFX11-FAKE16-NEXT: v_cmp_u_f32_e32 vcc_lo, v38, v38
-; GFX11-FAKE16-NEXT: v_and_b32_e32 v22, 0xffff0000, v22
-; GFX11-FAKE16-NEXT: v_lshlrev_b32_e32 v37, 16, v21
-; GFX11-FAKE16-NEXT: v_lshrrev_b32_e32 v19, 16, v34
-; GFX11-FAKE16-NEXT: v_bfe_u32 v36, v20, 16, 1
-; GFX11-FAKE16-NEXT: v_cndmask_b32_e32 v35, v35, v39, vcc_lo
-; GFX11-FAKE16-NEXT: v_or_b32_e32 v38, 0x400000, v20
+; GFX11-FAKE16-NEXT: v_add3_u32 v38, v38, v20, 0x7fff
+; GFX11-FAKE16-NEXT: v_perm_b32 v33, v33, v35, 0x7060302
+; GFX11-FAKE16-NEXT: v_add3_u32 v20, v49, v37, 0x7fff
+; GFX11-FAKE16-NEXT: v_lshlrev_b32_e32 v49, 16, v22
+; GFX11-FAKE16-NEXT: v_bfe_u32 v36, v48, 16, 1
+; GFX11-FAKE16-NEXT: v_dual_cndmask_b32 v19, v38, v19 :: v_dual_and_b32 v22, 0xffff0000, v22
+; GFX11-FAKE16-NEXT: v_or_b32_e32 v38, 0x400000, v37
+; GFX11-FAKE16-NEXT: v_cmp_u_f32_e32 vcc_lo, v37, v37
+; GFX11-FAKE16-NEXT: s_delay_alu instid0(VALU_DEP_4)
+; GFX11-FAKE16-NEXT: v_add3_u32 v39, v36, v48, 0x7fff
+; GFX11-FAKE16-NEXT: v_or_b32_e32 v50, 0x400000, v48
+; GFX11-FAKE16-NEXT: v_lshrrev_b32_e32 v160, 16, v19
+; GFX11-FAKE16-NEXT: v_lshrrev_b32_e32 v74, 8, v34
+; GFX11-FAKE16-NEXT: v_cndmask_b32_e32 v36, v20, v38, vcc_lo
+; GFX11-FAKE16-NEXT: v_add_f32_e32 v20, 0x40c00000, v49
+; GFX11-FAKE16-NEXT: v_cmp_u_f32_e32 vcc_lo, v48, v48
+; GFX11-FAKE16-NEXT: v_lshlrev_b32_e32 v48, 16, v21
+; GFX11-FAKE16-NEXT: v_lshrrev_b32_e32 v77, 16, v33
+; GFX11-FAKE16-NEXT: v_lshrrev_b32_e32 v79, 8, v33
+; GFX11-FAKE16-NEXT: v_cndmask_b32_e32 v37, v39, v50, vcc_lo
+; GFX11-FAKE16-NEXT: v_bfe_u32 v39, v20, 16, 1
+; GFX11-FAKE16-NEXT: v_or_b32_e32 v50, 0x400000, v20
; GFX11-FAKE16-NEXT: v_cmp_u_f32_e32 vcc_lo, v20, v20
-; GFX11-FAKE16-NEXT: v_add_f32_e32 v37, 0x40c00000, v37
-; GFX11-FAKE16-NEXT: v_add3_u32 v36, v36, v20, 0x7fff
+; GFX11-FAKE16-NEXT: v_perm_b32 v38, v19, v18, 0x7060302
+; GFX11-FAKE16-NEXT: v_perm_b32 v37, v37, v36, 0x7060302
+; GFX11-FAKE16-NEXT: v_add3_u32 v39, v39, v20, 0x7fff
; GFX11-FAKE16-NEXT: v_and_b32_e32 v21, 0xffff0000, v21
-; GFX11-FAKE16-NEXT: v_perm_b32 v64, v35, v34, 0x7060302
-; GFX11-FAKE16-NEXT: v_lshrrev_b32_e32 v46, 8, v65
-; GFX11-FAKE16-NEXT: v_bfe_u32 v48, v37, 16, 1
-; GFX11-FAKE16-NEXT: v_cndmask_b32_e32 v20, v36, v38, vcc_lo
-; GFX11-FAKE16-NEXT: v_dual_add_f32 v22, 0x40c00000, v22 :: v_dual_add_f32 v39, 0x40c00000, v21
-; GFX11-FAKE16-NEXT: v_lshrrev_b32_e32 v47, 16, v64
-; GFX11-FAKE16-NEXT: v_lshrrev_b32_e32 v57, 8, v64
-; GFX11-FAKE16-NEXT: v_lshrrev_b64 v[64:65], 24, v[64:65]
+; GFX11-FAKE16-NEXT: v_lshrrev_b32_e32 v19, 16, v36
+; GFX11-FAKE16-NEXT: v_lshrrev_b32_e32 v57, 24, v38
+; GFX11-FAKE16-NEXT: v_lshrrev_b32_e32 v60, 8, v38
+; GFX11-FAKE16-NEXT: v_cndmask_b32_e32 v20, v39, v50, vcc_lo
+; GFX11-FAKE16-NEXT: v_dual_add_f32 v22, 0x40c00000, v22 :: v_dual_add_f32 v51, 0x40c00000, v21
+; GFX11-FAKE16-NEXT: v_lshlrev_b32_e32 v50, 16, v24
+; GFX11-FAKE16-NEXT: v_and_b32_e32 v24, 0xffff0000, v24
+; GFX11-FAKE16-NEXT: v_lshrrev_b32_e32 v61, 16, v37
; GFX11-FAKE16-NEXT: s_delay_alu instid0(VALU_DEP_4)
-; GFX11-FAKE16-NEXT: v_bfe_u32 v35, v22, 16, 1
+; GFX11-FAKE16-NEXT: v_bfe_u32 v49, v22, 16, 1
; GFX11-FAKE16-NEXT: v_or_b32_e32 v21, 0x400000, v22
; GFX11-FAKE16-NEXT: v_cmp_u_f32_e32 vcc_lo, v22, v22
-; GFX11-FAKE16-NEXT: v_lshlrev_b32_e32 v38, 16, v24
-; GFX11-FAKE16-NEXT: v_bfe_u32 v36, v39, 16, 1
-; GFX11-FAKE16-NEXT: v_add3_u32 v35, v35, v22, 0x7fff
-; GFX11-FAKE16-NEXT: v_add3_u32 v22, v48, v37, 0x7fff
-; GFX11-FAKE16-NEXT: v_or_b32_e32 v48, 0x400000, v39
-; GFX11-FAKE16-NEXT: v_and_b32_e32 v24, 0xffff0000, v24
-; GFX11-FAKE16-NEXT: v_add3_u32 v36, v36, v39, 0x7fff
-; GFX11-FAKE16-NEXT: v_cndmask_b32_e32 v21, v35, v21, vcc_lo
-; GFX11-FAKE16-NEXT: v_or_b32_e32 v35, 0x400000, v37
-; GFX11-FAKE16-NEXT: v_cmp_u_f32_e32 vcc_lo, v37, v37
+; GFX11-FAKE16-NEXT: v_bfe_u32 v39, v51, 16, 1
; GFX11-FAKE16-NEXT: v_add_f32_e32 v24, 0x40c00000, v24
-; GFX11-FAKE16-NEXT: s_delay_alu instid0(VALU_DEP_4)
-; GFX11-FAKE16-NEXT: v_perm_b32 v71, v21, v20, 0x7060302
-; GFX11-FAKE16-NEXT: v_lshrrev_b32_e32 v133, 16, v21
-; GFX11-FAKE16-NEXT: v_dual_cndmask_b32 v35, v22, v35 :: v_dual_add_f32 v22, 0x40c00000, v38
-; GFX11-FAKE16-NEXT: v_cmp_u_f32_e32 vcc_lo, v39, v39
-; GFX11-FAKE16-NEXT: v_lshlrev_b32_e32 v38, 16, v23
+; GFX11-FAKE16-NEXT: v_add3_u32 v49, v49, v22, 0x7fff
+; GFX11-FAKE16-NEXT: v_add_f32_e32 v48, 0x40c00000, v48
+; GFX11-FAKE16-NEXT: v_lshrrev_b32_e32 v63, 8, v37
+; GFX11-FAKE16-NEXT: v_add3_u32 v39, v39, v51, 0x7fff
+; GFX11-FAKE16-NEXT: s_delay_alu instid0(VALU_DEP_4) | instskip(NEXT) | instid1(VALU_DEP_4)
+; GFX11-FAKE16-NEXT: v_cndmask_b32_e32 v21, v49, v21, vcc_lo
+; GFX11-FAKE16-NEXT: v_bfe_u32 v52, v48, 16, 1
+; GFX11-FAKE16-NEXT: v_or_b32_e32 v49, 0x400000, v48
+; GFX11-FAKE16-NEXT: v_cmp_u_f32_e32 vcc_lo, v48, v48
+; GFX11-FAKE16-NEXT: s_delay_alu instid0(VALU_DEP_4) | instskip(NEXT) | instid1(VALU_DEP_4)
+; GFX11-FAKE16-NEXT: v_perm_b32 v53, v21, v20, 0x7060302
+; GFX11-FAKE16-NEXT: v_add3_u32 v22, v52, v48, 0x7fff
+; GFX11-FAKE16-NEXT: v_or_b32_e32 v52, 0x400000, v51
+; GFX11-FAKE16-NEXT: v_lshrrev_b32_e32 v147, 16, v21
+; GFX11-FAKE16-NEXT: s_delay_alu instid0(VALU_DEP_3)
+; GFX11-FAKE16-NEXT: v_cndmask_b32_e32 v48, v22, v49, vcc_lo
+; GFX11-FAKE16-NEXT: v_add_f32_e32 v22, 0x40c00000, v50
+; GFX11-FAKE16-NEXT: v_cmp_u_f32_e32 vcc_lo, v51, v51
+; GFX11-FAKE16-NEXT: v_lshlrev_b32_e32 v50, 16, v23
; GFX11-FAKE16-NEXT: v_and_b32_e32 v23, 0xffff0000, v23
-; GFX11-FAKE16-NEXT: s_delay_alu instid0(VALU_DEP_4)
-; GFX11-FAKE16-NEXT: v_lshrrev_b32_e32 v21, 16, v35
-; GFX11-FAKE16-NEXT: v_bfe_u32 v37, v22, 16, 1
-; GFX11-FAKE16-NEXT: v_cndmask_b32_e32 v36, v36, v48, vcc_lo
-; GFX11-FAKE16-NEXT: v_or_b32_e32 v39, 0x400000, v22
+; GFX11-FAKE16-NEXT: v_lshrrev_b32_e32 v21, 16, v48
+; GFX11-FAKE16-NEXT: v_bfe_u32 v49, v22, 16, 1
+; GFX11-FAKE16-NEXT: s_delay_alu instid0(VALU_DEP_4) | instskip(SKIP_2) | instid1(VALU_DEP_4)
+; GFX11-FAKE16-NEXT: v_dual_cndmask_b32 v39, v39, v52 :: v_dual_add_f32 v50, 0x40c00000, v50
+; GFX11-FAKE16-NEXT: v_or_b32_e32 v51, 0x400000, v22
; GFX11-FAKE16-NEXT: v_cmp_u_f32_e32 vcc_lo, v22, v22
-; GFX11-FAKE16-NEXT: v_add_f32_e32 v38, 0x40c00000, v38
-; GFX11-FAKE16-NEXT: v_add3_u32 v37, v37, v22, 0x7fff
-; GFX11-FAKE16-NEXT: v_perm_b32 v70, v36, v35, 0x7060302
-; GFX11-FAKE16-NEXT: v_add_f32_e32 v48, 0x40c00000, v23
-; GFX11-FAKE16-NEXT: s_delay_alu instid0(VALU_DEP_4) | instskip(NEXT) | instid1(VALU_DEP_4)
-; GFX11-FAKE16-NEXT: v_bfe_u32 v49, v38, 16, 1
-; GFX11-FAKE16-NEXT: v_dual_cndmask_b32 v22, v37, v39 :: v_dual_lshlrev_b32 v39, 16, v26
-; GFX11-FAKE16-NEXT: v_bfe_u32 v36, v24, 16, 1
+; GFX11-FAKE16-NEXT: v_add3_u32 v49, v49, v22, 0x7fff
+; GFX11-FAKE16-NEXT: s_delay_alu instid0(VALU_DEP_4)
+; GFX11-FAKE16-NEXT: v_perm_b32 v52, v39, v48, 0x7060302
+; GFX11-FAKE16-NEXT: v_bfe_u32 v39, v24, 16, 1
+; GFX11-FAKE16-NEXT: v_add_f32_e32 v54, 0x40c00000, v23
; GFX11-FAKE16-NEXT: v_or_b32_e32 v23, 0x400000, v24
+; GFX11-FAKE16-NEXT: v_bfe_u32 v55, v50, 16, 1
+; GFX11-FAKE16-NEXT: v_cndmask_b32_e32 v22, v49, v51, vcc_lo
+; GFX11-FAKE16-NEXT: v_add3_u32 v39, v39, v24, 0x7fff
; GFX11-FAKE16-NEXT: v_cmp_u_f32_e32 vcc_lo, v24, v24
-; GFX11-FAKE16-NEXT: v_bfe_u32 v37, v48, 16, 1
-; GFX11-FAKE16-NEXT: v_and_b32_e32 v26, 0xffff0000, v26
-; GFX11-FAKE16-NEXT: v_add3_u32 v36, v36, v24, 0x7fff
-; GFX11-FAKE16-NEXT: v_add3_u32 v24, v49, v38, 0x7fff
-; GFX11-FAKE16-NEXT: v_or_b32_e32 v49, 0x400000, v48
-; GFX11-FAKE16-NEXT: v_add3_u32 v37, v37, v48, 0x7fff
-; GFX11-FAKE16-NEXT: s_delay_alu instid0(VALU_DEP_4) | instskip(SKIP_3) | instid1(VALU_DEP_4)
-; GFX11-FAKE16-NEXT: v_dual_add_f32 v26, 0x40c00000, v26 :: v_dual_cndmask_b32 v23, v36, v23
-; GFX11-FAKE16-NEXT: v_or_b32_e32 v36, 0x400000, v38
-; GFX11-FAKE16-NEXT: v_cmp_u_f32_e32 vcc_lo, v38, v38
-; GFX11-FAKE16-NEXT: v_lshrrev_b32_e32 v40, 16, v70
-; GFX11-FAKE16-NEXT: v_lshrrev_b32_e32 v128, 16, v23
-; GFX11-FAKE16-NEXT: s_delay_alu instid0(VALU_DEP_4)
-; GFX11-FAKE16-NEXT: v_cndmask_b32_e32 v36, v24, v36, vcc_lo
-; GFX11-FAKE16-NEXT: v_add_f32_e32 v24, 0x40c00000, v39
-; GFX11-FAKE16-NEXT: v_cmp_u_f32_e32 vcc_lo, v48, v48
-; GFX11-FAKE16-NEXT: v_lshlrev_b32_e32 v39, 16, v25
+; GFX11-FAKE16-NEXT: v_bfe_u32 v49, v54, 16, 1
+; GFX11-FAKE16-NEXT: v_add3_u32 v24, v55, v50, 0x7fff
+; GFX11-FAKE16-NEXT: v_lshlrev_b32_e32 v55, 16, v26
+; GFX11-FAKE16-NEXT: v_or_b32_e32 v64, 0x400000, v54
+; GFX11-FAKE16-NEXT: v_cndmask_b32_e32 v23, v39, v23, vcc_lo
+; GFX11-FAKE16-NEXT: v_or_b32_e32 v39, 0x400000, v50
+; GFX11-FAKE16-NEXT: v_cmp_u_f32_e32 vcc_lo, v50, v50
+; GFX11-FAKE16-NEXT: v_add3_u32 v51, v49, v54, 0x7fff
+; GFX11-FAKE16-NEXT: v_lshrrev_b32_e32 v46, 16, v52
+; GFX11-FAKE16-NEXT: v_lshrrev_b32_e32 v134, 16, v23
+; GFX11-FAKE16-NEXT: v_cndmask_b32_e32 v49, v24, v39, vcc_lo
+; GFX11-FAKE16-NEXT: v_add_f32_e32 v24, 0x40c00000, v55
+; GFX11-FAKE16-NEXT: v_cmp_u_f32_e32 vcc_lo, v54, v54
+; GFX11-FAKE16-NEXT: s_delay_alu instid0(VALU_DEP_2)
+; GFX11-FAKE16-NEXT: v_bfe_u32 v50, v24, 16, 1
+; GFX11-FAKE16-NEXT: v_cndmask_b32_e32 v39, v51, v64, vcc_lo
+; GFX11-FAKE16-NEXT: v_lshlrev_b32_e32 v51, 16, v25
; GFX11-FAKE16-NEXT: v_and_b32_e32 v25, 0xffff0000, v25
-; GFX11-FAKE16-NEXT: v_perm_b32 v81, v23, v22, 0x7060302
-; GFX11-FAKE16-NEXT: v_bfe_u32 v38, v24, 16, 1
-; GFX11-FAKE16-NEXT: v_cndmask_b32_e32 v37, v37, v49, vcc_lo
-; GFX11-FAKE16-NEXT: v_add_f32_e32 v39, 0x40c00000, v39
-; GFX11-FAKE16-NEXT: v_or_b32_e32 v48, 0x400000, v24
+; GFX11-FAKE16-NEXT: v_or_b32_e32 v54, 0x400000, v24
+; GFX11-FAKE16-NEXT: v_add3_u32 v50, v50, v24, 0x7fff
; GFX11-FAKE16-NEXT: v_cmp_u_f32_e32 vcc_lo, v24, v24
-; GFX11-FAKE16-NEXT: v_add3_u32 v38, v38, v24, 0x7fff
-; GFX11-FAKE16-NEXT: v_perm_b32 v80, v37, v36, 0x7060302
-; GFX11-FAKE16-NEXT: v_bfe_u32 v37, v26, 16, 1
-; GFX11-FAKE16-NEXT: v_add_f32_e32 v49, 0x40c00000, v25
+; GFX11-FAKE16-NEXT: v_perm_b32 v69, v39, v49, 0x7060302
+; GFX11-FAKE16-NEXT: v_add_f32_e32 v55, 0x40c00000, v25
+; GFX11-FAKE16-NEXT: v_add_f32_e32 v51, 0x40c00000, v51
+; GFX11-FAKE16-NEXT: v_perm_b32 v70, v23, v22, 0x7060302
+; GFX11-FAKE16-NEXT: v_cndmask_b32_e32 v24, v50, v54, vcc_lo
+; GFX11-FAKE16-NEXT: v_lshrrev_b32_e32 v23, 16, v49
+; GFX11-FAKE16-NEXT: v_bfe_u32 v50, v55, 16, 1
+; GFX11-FAKE16-NEXT: v_or_b32_e32 v65, 0x400000, v55
+; GFX11-FAKE16-NEXT: v_bfe_u32 v64, v51, 16, 1
+; GFX11-FAKE16-NEXT: v_lshrrev_b32_e32 v40, 16, v69
+; GFX11-FAKE16-NEXT: s_delay_alu instid0(VALU_DEP_4) | instskip(SKIP_1) | instid1(VALU_DEP_1)
+; GFX11-FAKE16-NEXT: v_add3_u32 v54, v50, v55, 0x7fff
+; GFX11-FAKE16-NEXT: v_and_b32_e32 v26, 0xffff0000, v26
+; GFX11-FAKE16-NEXT: v_add_f32_e32 v26, 0x40c00000, v26
+; GFX11-FAKE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(SKIP_2) | instid1(VALU_DEP_3)
+; GFX11-FAKE16-NEXT: v_bfe_u32 v39, v26, 16, 1
; GFX11-FAKE16-NEXT: v_or_b32_e32 v25, 0x400000, v26
-; GFX11-FAKE16-NEXT: v_bfe_u32 v50, v39, 16, 1
-; GFX11-FAKE16-NEXT: v_cndmask_b32_e32 v24, v38, v48, vcc_lo
-; GFX11-FAKE16-NEXT: v_add3_u32 v37, v37, v26, 0x7fff
; GFX11-FAKE16-NEXT: v_cmp_u_f32_e32 vcc_lo, v26, v26
-; GFX11-FAKE16-NEXT: v_bfe_u32 v38, v49, 16, 1
-; GFX11-FAKE16-NEXT: v_add3_u32 v26, v50, v39, 0x7fff
-; GFX11-FAKE16-NEXT: v_lshlrev_b32_e32 v48, 16, v28
-; GFX11-FAKE16-NEXT: v_or_b32_e32 v50, 0x400000, v49
-; GFX11-FAKE16-NEXT: v_cndmask_b32_e32 v25, v37, v25, vcc_lo
-; GFX11-FAKE16-NEXT: v_or_b32_e32 v37, 0x400000, v39
-; GFX11-FAKE16-NEXT: v_cmp_u_f32_e32 vcc_lo, v39, v39
-; GFX11-FAKE16-NEXT: v_add3_u32 v38, v38, v49, 0x7fff
-; GFX11-FAKE16-NEXT: v_and_b32_e32 v28, 0xffff0000, v28
-; GFX11-FAKE16-NEXT: v_lshrrev_b32_e32 v23, 16, v36
-; GFX11-FAKE16-NEXT: v_lshrrev_b32_e32 v178, 16, v80
-; GFX11-FAKE16-NEXT: v_dual_cndmask_b32 v37, v26, v37 :: v_dual_add_f32 v26, 0x40c00000, v48
-; GFX11-FAKE16-NEXT: v_cmp_u_f32_e32 vcc_lo, v49, v49
+; GFX11-FAKE16-NEXT: v_add3_u32 v39, v39, v26, 0x7fff
+; GFX11-FAKE16-NEXT: v_add3_u32 v26, v64, v51, 0x7fff
+; GFX11-FAKE16-NEXT: v_lshlrev_b32_e32 v64, 16, v28
+; GFX11-FAKE16-NEXT: s_delay_alu instid0(VALU_DEP_3) | instskip(SKIP_2) | instid1(VALU_DEP_3)
+; GFX11-FAKE16-NEXT: v_dual_cndmask_b32 v25, v39, v25 :: v_dual_and_b32 v28, 0xffff0000, v28
+; GFX11-FAKE16-NEXT: v_or_b32_e32 v39, 0x400000, v51
+; GFX11-FAKE16-NEXT: v_cmp_u_f32_e32 vcc_lo, v51, v51
; GFX11-FAKE16-NEXT: v_add_f32_e32 v28, 0x40c00000, v28
-; GFX11-FAKE16-NEXT: v_lshlrev_b32_e32 v48, 16, v27
+; GFX11-FAKE16-NEXT: s_delay_alu instid0(VALU_DEP_4) | instskip(NEXT) | instid1(VALU_DEP_4)
+; GFX11-FAKE16-NEXT: v_lshrrev_b32_e32 v130, 16, v25
+; GFX11-FAKE16-NEXT: v_cndmask_b32_e32 v50, v26, v39, vcc_lo
+; GFX11-FAKE16-NEXT: v_add_f32_e32 v26, 0x40c00000, v64
+; GFX11-FAKE16-NEXT: v_cmp_u_f32_e32 vcc_lo, v55, v55
+; GFX11-FAKE16-NEXT: s_delay_alu instid0(VALU_DEP_2) | instskip(SKIP_3) | instid1(VALU_DEP_4)
+; GFX11-FAKE16-NEXT: v_bfe_u32 v51, v26, 16, 1
+; GFX11-FAKE16-NEXT: v_dual_cndmask_b32 v39, v54, v65 :: v_dual_lshlrev_b32 v54, 16, v27
; GFX11-FAKE16-NEXT: v_and_b32_e32 v27, 0xffff0000, v27
-; GFX11-FAKE16-NEXT: v_bfe_u32 v39, v26, 16, 1
-; GFX11-FAKE16-NEXT: v_cndmask_b32_e32 v38, v38, v50, vcc_lo
-; GFX11-FAKE16-NEXT: v_or_b32_e32 v49, 0x400000, v26
-; GFX11-FAKE16-NEXT: v_add_f32_e32 v48, 0x40c00000, v48
+; GFX11-FAKE16-NEXT: v_or_b32_e32 v55, 0x400000, v26
+; GFX11-FAKE16-NEXT: v_add3_u32 v51, v51, v26, 0x7fff
+; GFX11-FAKE16-NEXT: s_delay_alu instid0(VALU_DEP_4)
+; GFX11-FAKE16-NEXT: v_perm_b32 v83, v39, v50, 0x7060302
+; GFX11-FAKE16-NEXT: v_bfe_u32 v39, v28, 16, 1
+; GFX11-FAKE16-NEXT: v_add_f32_e32 v54, 0x40c00000, v54
; GFX11-FAKE16-NEXT: v_cmp_u_f32_e32 vcc_lo, v26, v26
-; GFX11-FAKE16-NEXT: v_add3_u32 v39, v39, v26, 0x7fff
-; GFX11-FAKE16-NEXT: v_perm_b32 v82, v38, v37, 0x7060302
-; GFX11-FAKE16-NEXT: v_bfe_u32 v38, v28, 16, 1
-; GFX11-FAKE16-NEXT: v_add_f32_e32 v50, 0x40c00000, v27
+; GFX11-FAKE16-NEXT: v_add_f32_e32 v64, 0x40c00000, v27
; GFX11-FAKE16-NEXT: v_or_b32_e32 v27, 0x400000, v28
-; GFX11-FAKE16-NEXT: v_bfe_u32 v51, v48, 16, 1
-; GFX11-FAKE16-NEXT: v_cndmask_b32_e32 v26, v39, v49, vcc_lo
-; GFX11-FAKE16-NEXT: v_add3_u32 v38, v38, v28, 0x7fff
+; GFX11-FAKE16-NEXT: v_add3_u32 v39, v39, v28, 0x7fff
+; GFX11-FAKE16-NEXT: v_bfe_u32 v65, v54, 16, 1
+; GFX11-FAKE16-NEXT: v_cndmask_b32_e32 v26, v51, v55, vcc_lo
; GFX11-FAKE16-NEXT: v_cmp_u_f32_e32 vcc_lo, v28, v28
-; GFX11-FAKE16-NEXT: v_bfe_u32 v39, v50, 16, 1
-; GFX11-FAKE16-NEXT: v_add3_u32 v28, v51, v48, 0x7fff
-; GFX11-FAKE16-NEXT: v_lshlrev_b32_e32 v49, 16, v30
-; GFX11-FAKE16-NEXT: v_or_b32_e32 v51, 0x400000, v50
-; GFX11-FAKE16-NEXT: v_cndmask_b32_e32 v27, v38, v27, vcc_lo
-; GFX11-FAKE16-NEXT: v_or_b32_e32 v38, 0x400000, v48
-; GFX11-FAKE16-NEXT: v_cmp_u_f32_e32 vcc_lo, v48, v48
-; GFX11-FAKE16-NEXT: v_add3_u32 v39, v39, v50, 0x7fff
+; GFX11-FAKE16-NEXT: v_bfe_u32 v51, v64, 16, 1
+; GFX11-FAKE16-NEXT: v_or_b32_e32 v66, 0x400000, v64
+; GFX11-FAKE16-NEXT: v_add3_u32 v28, v65, v54, 0x7fff
+; GFX11-FAKE16-NEXT: v_lshlrev_b32_e32 v65, 16, v30
+; GFX11-FAKE16-NEXT: v_cndmask_b32_e32 v27, v39, v27, vcc_lo
+; GFX11-FAKE16-NEXT: v_or_b32_e32 v39, 0x400000, v54
+; GFX11-FAKE16-NEXT: v_cmp_u_f32_e32 vcc_lo, v54, v54
+; GFX11-FAKE16-NEXT: v_add3_u32 v55, v51, v64, 0x7fff
; GFX11-FAKE16-NEXT: v_and_b32_e32 v30, 0xffff0000, v30
-; GFX11-FAKE16-NEXT: v_perm_b32 v83, v25, v24, 0x7060302
-; GFX11-FAKE16-NEXT: v_cndmask_b32_e32 v38, v28, v38, vcc_lo
-; GFX11-FAKE16-NEXT: v_add_f32_e32 v28, 0x40c00000, v49
-; GFX11-FAKE16-NEXT: v_cmp_u_f32_e32 vcc_lo, v50, v50
-; GFX11-FAKE16-NEXT: v_dual_add_f32 v30, 0x40c00000, v30 :: v_dual_lshlrev_b32 v49, 16, v29
-; GFX11-FAKE16-NEXT: v_and_b32_e32 v29, 0xffff0000, v29
+; GFX11-FAKE16-NEXT: v_perm_b32 v84, v25, v24, 0x7060302
+; GFX11-FAKE16-NEXT: v_lshrrev_b32_e32 v128, 16, v27
+; GFX11-FAKE16-NEXT: v_dual_cndmask_b32 v51, v28, v39 :: v_dual_add_f32 v28, 0x40c00000, v65
+; GFX11-FAKE16-NEXT: v_cmp_u_f32_e32 vcc_lo, v64, v64
+; GFX11-FAKE16-NEXT: v_add_f32_e32 v30, 0x40c00000, v30
+; GFX11-FAKE16-NEXT: v_lshrrev_b32_e32 v165, 24, v84
; GFX11-FAKE16-NEXT: s_delay_alu instid0(VALU_DEP_4)
-; GFX11-FAKE16-NEXT: v_bfe_u32 v48, v28, 16, 1
-; GFX11-FAKE16-NEXT: v_cndmask_b32_e32 v39, v39, v51, vcc_lo
-; GFX11-FAKE16-NEXT: v_or_b32_e32 v50, 0x400000, v28
-; GFX11-FAKE16-NEXT: v_add_f32_e32 v49, 0x40c00000, v49
-; GFX11-FAKE16-NEXT: v_cmp_u_f32_e32 vcc_lo, v28, v28
-; GFX11-FAKE16-NEXT: v_add3_u32 v48, v48, v28, 0x7fff
-; GFX11-FAKE16-NEXT: v_perm_b32 v84, v39, v38, 0x7060302
+; GFX11-FAKE16-NEXT: v_bfe_u32 v54, v28, 16, 1
+; GFX11-FAKE16-NEXT: v_cndmask_b32_e32 v39, v55, v66, vcc_lo
+; GFX11-FAKE16-NEXT: v_lshlrev_b32_e32 v55, 16, v29
+; GFX11-FAKE16-NEXT: v_and_b32_e32 v29, 0xffff0000, v29
+; GFX11-FAKE16-NEXT: v_or_b32_e32 v64, 0x400000, v28
+; GFX11-FAKE16-NEXT: v_add3_u32 v54, v54, v28, 0x7fff
+; GFX11-FAKE16-NEXT: v_perm_b32 v96, v39, v51, 0x7060302
; GFX11-FAKE16-NEXT: v_bfe_u32 v39, v30, 16, 1
-; GFX11-FAKE16-NEXT: v_add_f32_e32 v51, 0x40c00000, v29
+; GFX11-FAKE16-NEXT: v_add_f32_e32 v55, 0x40c00000, v55
+; GFX11-FAKE16-NEXT: v_cmp_u_f32_e32 vcc_lo, v28, v28
+; GFX11-FAKE16-NEXT: v_add_f32_e32 v65, 0x40c00000, v29
; GFX11-FAKE16-NEXT: v_or_b32_e32 v29, 0x400000, v30
-; GFX11-FAKE16-NEXT: v_bfe_u32 v52, v49, 16, 1
-; GFX11-FAKE16-NEXT: v_cndmask_b32_e32 v28, v48, v50, vcc_lo
; GFX11-FAKE16-NEXT: v_add3_u32 v39, v39, v30, 0x7fff
+; GFX11-FAKE16-NEXT: v_bfe_u32 v66, v55, 16, 1
+; GFX11-FAKE16-NEXT: v_cndmask_b32_e32 v28, v54, v64, vcc_lo
; GFX11-FAKE16-NEXT: v_cmp_u_f32_e32 vcc_lo, v30, v30
-; GFX11-FAKE16-NEXT: v_bfe_u32 v48, v51, 16, 1
-; GFX11-FAKE16-NEXT: v_add3_u32 v30, v52, v49, 0x7fff
+; GFX11-FAKE16-NEXT: v_bfe_u32 v54, v65, 16, 1
+; GFX11-FAKE16-NEXT: v_or_b32_e32 v67, 0x400000, v65
+; GFX11-FAKE16-NEXT: v_add3_u32 v30, v66, v55, 0x7fff
; GFX11-FAKE16-NEXT: s_waitcnt vmcnt(1)
-; GFX11-FAKE16-NEXT: v_lshlrev_b32_e32 v50, 16, v32
-; GFX11-FAKE16-NEXT: v_or_b32_e32 v52, 0x400000, v51
-; GFX11-FAKE16-NEXT: v_cndmask_b32_e32 v29, v39, v29, vcc_lo
-; GFX11-FAKE16-NEXT: v_or_b32_e32 v39, 0x400000, v49
-; GFX11-FAKE16-NEXT: v_cmp_u_f32_e32 vcc_lo, v49, v49
-; GFX11-FAKE16-NEXT: v_add3_u32 v48, v48, v51, 0x7fff
+; GFX11-FAKE16-NEXT: v_dual_cndmask_b32 v29, v39, v29 :: v_dual_lshlrev_b32 v66, 16, v32
+; GFX11-FAKE16-NEXT: v_or_b32_e32 v39, 0x400000, v55
+; GFX11-FAKE16-NEXT: v_cmp_u_f32_e32 vcc_lo, v55, v55
+; GFX11-FAKE16-NEXT: v_add3_u32 v64, v54, v65, 0x7fff
; GFX11-FAKE16-NEXT: v_and_b32_e32 v32, 0xffff0000, v32
-; GFX11-FAKE16-NEXT: v_lshrrev_b32_e32 v151, 24, v83
-; GFX11-FAKE16-NEXT: v_dual_cndmask_b32 v39, v30, v39 :: v_dual_add_f32 v30, 0x40c00000, v50
-; GFX11-FAKE16-NEXT: v_cmp_u_f32_e32 vcc_lo, v51, v51
-; GFX11-FAKE16-NEXT: s_delay_alu instid0(VALU_DEP_4)
+; GFX11-FAKE16-NEXT: v_perm_b32 v97, v27, v26, 0x7060302
+; GFX11-FAKE16-NEXT: v_lshrrev_b32_e32 v25, 16, v50
+; GFX11-FAKE16-NEXT: v_cndmask_b32_e32 v54, v30, v39, vcc_lo
+; GFX11-FAKE16-NEXT: v_add_f32_e32 v30, 0x40c00000, v66
+; GFX11-FAKE16-NEXT: v_cmp_u_f32_e32 vcc_lo, v65, v65
; GFX11-FAKE16-NEXT: v_add_f32_e32 v32, 0x40c00000, v32
+; GFX11-FAKE16-NEXT: v_lshrrev_b32_e32 v27, 16, v51
+; GFX11-FAKE16-NEXT: s_delay_alu instid0(VALU_DEP_4)
+; GFX11-FAKE16-NEXT: v_bfe_u32 v55, v30, 16, 1
+; GFX11-FAKE16-NEXT: v_cndmask_b32_e32 v39, v64, v67, vcc_lo
; GFX11-FAKE16-NEXT: s_waitcnt vmcnt(0)
-; GFX11-FAKE16-NEXT: v_lshlrev_b32_e32 v50, 16, v31
+; GFX11-FAKE16-NEXT: v_lshlrev_b32_e32 v64, 16, v31
; GFX11-FAKE16-NEXT: v_and_b32_e32 v31, 0xffff0000, v31
-; GFX11-FAKE16-NEXT: v_bfe_u32 v49, v30, 16, 1
-; GFX11-FAKE16-NEXT: v_cndmask_b32_e32 v48, v48, v52, vcc_lo
-; GFX11-FAKE16-NEXT: v_or_b32_e32 v51, 0x400000, v30
-; GFX11-FAKE16-NEXT: v_add_f32_e32 v50, 0x40c00000, v50
+; GFX11-FAKE16-NEXT: v_or_b32_e32 v65, 0x400000, v30
+; GFX11-FAKE16-NEXT: v_add3_u32 v55, v55, v30, 0x7fff
+; GFX11-FAKE16-NEXT: v_perm_b32 v101, v39, v54, 0x7060302
+; GFX11-FAKE16-NEXT: v_bfe_u32 v39, v32, 16, 1
+; GFX11-FAKE16-NEXT: v_add_f32_e32 v64, 0x40c00000, v64
; GFX11-FAKE16-NEXT: v_cmp_u_f32_e32 vcc_lo, v30, v30
-; GFX11-FAKE16-NEXT: v_add3_u32 v49, v49, v30, 0x7fff
-; GFX11-FAKE16-NEXT: v_perm_b32 v86, v48, v39, 0x7060302
-; GFX11-FAKE16-NEXT: v_bfe_u32 v48, v32, 16, 1
-; GFX11-FAKE16-NEXT: v_add_f32_e32 v52, 0x40c00000, v31
+; GFX11-FAKE16-NEXT: v_add_f32_e32 v66, 0x40c00000, v31
; GFX11-FAKE16-NEXT: v_or_b32_e32 v31, 0x400000, v32
-; GFX11-FAKE16-NEXT: v_bfe_u32 v53, v50, 16, 1
-; GFX11-FAKE16-NEXT: v_cndmask_b32_e32 v30, v49, v51, vcc_lo
-; GFX11-FAKE16-NEXT: v_add3_u32 v48, v48, v32, 0x7fff
+; GFX11-FAKE16-NEXT: v_add3_u32 v39, v39, v32, 0x7fff
+; GFX11-FAKE16-NEXT: v_bfe_u32 v67, v64, 16, 1
+; GFX11-FAKE16-NEXT: v_cndmask_b32_e32 v30, v55, v65, vcc_lo
; GFX11-FAKE16-NEXT: v_cmp_u_f32_e32 vcc_lo, v32, v32
-; GFX11-FAKE16-NEXT: v_bfe_u32 v49, v52, 16, 1
-; GFX11-FAKE16-NEXT: v_add3_u32 v32, v53, v50, 0x7fff
-; GFX11-FAKE16-NEXT: v_lshlrev_b32_e32 v51, 16, v2
-; GFX11-FAKE16-NEXT: v_or_b32_e32 v53, 0x400000, v52
-; GFX11-FAKE16-NEXT: v_cndmask_b32_e32 v31, v48, v31, vcc_lo
-; GFX11-FAKE16-NEXT: v_or_b32_e32 v48, 0x400000, v50
-; GFX11-FAKE16-NEXT: v_cmp_u_f32_e32 vcc_lo, v50, v50
-; GFX11-FAKE16-NEXT: v_add3_u32 v49, v49, v52, 0x7fff
+; GFX11-FAKE16-NEXT: v_bfe_u32 v55, v66, 16, 1
+; GFX11-FAKE16-NEXT: v_or_b32_e32 v68, 0x400000, v66
+; GFX11-FAKE16-NEXT: v_add3_u32 v32, v67, v64, 0x7fff
+; GFX11-FAKE16-NEXT: v_lshlrev_b32_e32 v67, 16, v2
+; GFX11-FAKE16-NEXT: v_cndmask_b32_e32 v31, v39, v31, vcc_lo
+; GFX11-FAKE16-NEXT: v_or_b32_e32 v39, 0x400000, v64
+; GFX11-FAKE16-NEXT: v_cmp_u_f32_e32 vcc_lo, v64, v64
+; GFX11-FAKE16-NEXT: v_add3_u32 v65, v55, v66, 0x7fff
; GFX11-FAKE16-NEXT: v_and_b32_e32 v2, 0xffff0000, v2
-; GFX11-FAKE16-NEXT: v_perm_b32 v85, v27, v26, 0x7060302
-; GFX11-FAKE16-NEXT: v_cndmask_b32_e32 v48, v32, v48, vcc_lo
-; GFX11-FAKE16-NEXT: v_add_f32_e32 v32, 0x40c00000, v51
-; GFX11-FAKE16-NEXT: v_cmp_u_f32_e32 vcc_lo, v52, v52
-; GFX11-FAKE16-NEXT: v_dual_add_f32 v2, 0x40c00000, v2 :: v_dual_lshlrev_b32 v51, 16, v1
-; GFX11-FAKE16-NEXT: v_and_b32_e32 v1, 0xffff0000, v1
-; GFX11-FAKE16-NEXT: s_delay_alu instid0(VALU_DEP_4)
-; GFX11-FAKE16-NEXT: v_bfe_u32 v50, v32, 16, 1
-; GFX11-FAKE16-NEXT: v_cndmask_b32_e32 v49, v49, v53, vcc_lo
-; GFX11-FAKE16-NEXT: v_or_b32_e32 v52, 0x400000, v32
-; GFX11-FAKE16-NEXT: v_add_f32_e32 v51, 0x40c00000, v51
+; GFX11-FAKE16-NEXT: v_perm_b32 v102, v29, v28, 0x7060302
+; GFX11-FAKE16-NEXT: v_cndmask_b32_e32 v55, v32, v39, vcc_lo
+; GFX11-FAKE16-NEXT: v_cmp_u_f32_e32 vcc_lo, v66, v66
+; GFX11-FAKE16-NEXT: v_add_f32_e32 v32, 0x40c00000, v67
+; GFX11-FAKE16-NEXT: v_add_f32_e32 v2, 0x40c00000, v2
+; GFX11-FAKE16-NEXT: v_lshrrev_b64 v[48:49], 24, v[101:102]
+; GFX11-FAKE16-NEXT: v_lshrrev_b64 v[49:50], 24, v[96:97]
+; GFX11-FAKE16-NEXT: v_cndmask_b32_e32 v39, v65, v68, vcc_lo
+; GFX11-FAKE16-NEXT: v_bfe_u32 v64, v32, 16, 1
+; GFX11-FAKE16-NEXT: v_lshlrev_b32_e32 v65, 16, v1
+; GFX11-FAKE16-NEXT: v_or_b32_e32 v66, 0x400000, v32
+; GFX11-FAKE16-NEXT: v_or_b32_e32 v68, 0x400000, v2
+; GFX11-FAKE16-NEXT: v_perm_b32 v112, v39, v55, 0x7060302
+; GFX11-FAKE16-NEXT: v_bfe_u32 v39, v2, 16, 1
+; GFX11-FAKE16-NEXT: v_add3_u32 v64, v64, v32, 0x7fff
+; GFX11-FAKE16-NEXT: v_add_f32_e32 v65, 0x40c00000, v65
; GFX11-FAKE16-NEXT: v_cmp_u_f32_e32 vcc_lo, v32, v32
-; GFX11-FAKE16-NEXT: v_add3_u32 v50, v50, v32, 0x7fff
-; GFX11-FAKE16-NEXT: v_perm_b32 v96, v49, v48, 0x7060302
-; GFX11-FAKE16-NEXT: v_bfe_u32 v49, v2, 16, 1
-; GFX11-FAKE16-NEXT: v_add_f32_e32 v53, 0x40c00000, v1
-; GFX11-FAKE16-NEXT: v_or_b32_e32 v54, 0x400000, v2
-; GFX11-FAKE16-NEXT: v_bfe_u32 v55, v51, 16, 1
-; GFX11-FAKE16-NEXT: v_cndmask_b32_e32 v1, v50, v52, vcc_lo
-; GFX11-FAKE16-NEXT: v_add3_u32 v49, v49, v2, 0x7fff
+; GFX11-FAKE16-NEXT: v_lshrrev_b64 v[50:51], 24, v[83:84]
+; GFX11-FAKE16-NEXT: v_add3_u32 v39, v39, v2, 0x7fff
+; GFX11-FAKE16-NEXT: v_and_b32_e32 v1, 0xffff0000, v1
+; GFX11-FAKE16-NEXT: v_bfe_u32 v71, v65, 16, 1
+; GFX11-FAKE16-NEXT: v_lshrrev_b32_e32 v84, 8, v84
+; GFX11-FAKE16-NEXT: s_delay_alu instid0(VALU_DEP_3) | instskip(SKIP_3) | instid1(VALU_DEP_4)
+; GFX11-FAKE16-NEXT: v_add_f32_e32 v67, 0x40c00000, v1
+; GFX11-FAKE16-NEXT: v_cndmask_b32_e32 v1, v64, v66, vcc_lo
; GFX11-FAKE16-NEXT: v_cmp_u_f32_e32 vcc_lo, v2, v2
-; GFX11-FAKE16-NEXT: v_bfe_u32 v32, v53, 16, 1
-; GFX11-FAKE16-NEXT: v_or_b32_e32 v50, 0x400000, v51
-; GFX11-FAKE16-NEXT: v_perm_b32 v87, v29, v28, 0x7060302
-; GFX11-FAKE16-NEXT: v_cndmask_b32_e32 v2, v49, v54, vcc_lo
-; GFX11-FAKE16-NEXT: v_add3_u32 v49, v55, v51, 0x7fff
-; GFX11-FAKE16-NEXT: v_lshlrev_b32_e32 v54, 16, v4
-; GFX11-FAKE16-NEXT: v_cmp_u_f32_e32 vcc_lo, v51, v51
-; GFX11-FAKE16-NEXT: v_add3_u32 v52, v32, v53, 0x7fff
-; GFX11-FAKE16-NEXT: v_or_b32_e32 v55, 0x400000, v53
+; GFX11-FAKE16-NEXT: v_or_b32_e32 v64, 0x400000, v65
+; GFX11-FAKE16-NEXT: v_bfe_u32 v32, v67, 16, 1
+; GFX11-FAKE16-NEXT: v_cndmask_b32_e32 v2, v39, v68, vcc_lo
+; GFX11-FAKE16-NEXT: v_add3_u32 v39, v71, v65, 0x7fff
+; GFX11-FAKE16-NEXT: v_lshlrev_b32_e32 v68, 16, v4
+; GFX11-FAKE16-NEXT: v_cmp_u_f32_e32 vcc_lo, v65, v65
+; GFX11-FAKE16-NEXT: v_add3_u32 v66, v32, v67, 0x7fff
+; GFX11-FAKE16-NEXT: v_or_b32_e32 v71, 0x400000, v67
; GFX11-FAKE16-NEXT: v_and_b32_e32 v4, 0xffff0000, v4
-; GFX11-FAKE16-NEXT: v_cndmask_b32_e32 v32, v49, v50, vcc_lo
-; GFX11-FAKE16-NEXT: v_add_f32_e32 v49, 0x40c00000, v54
-; GFX11-FAKE16-NEXT: v_cmp_u_f32_e32 vcc_lo, v53, v53
-; GFX11-FAKE16-NEXT: s_delay_alu instid0(VALU_DEP_4) | instskip(NEXT) | instid1(VALU_DEP_3)
+; GFX11-FAKE16-NEXT: v_perm_b32 v115, v2, v1, 0x7060302
+; GFX11-FAKE16-NEXT: v_cndmask_b32_e32 v32, v39, v64, vcc_lo
+; GFX11-FAKE16-NEXT: v_add_f32_e32 v39, 0x40c00000, v68
+; GFX11-FAKE16-NEXT: v_cmp_u_f32_e32 vcc_lo, v67, v67
; GFX11-FAKE16-NEXT: v_add_f32_e32 v4, 0x40c00000, v4
-; GFX11-FAKE16-NEXT: v_bfe_u32 v51, v49, 16, 1
-; GFX11-FAKE16-NEXT: v_cndmask_b32_e32 v50, v52, v55, vcc_lo
-; GFX11-FAKE16-NEXT: v_lshlrev_b32_e32 v52, 16, v3
+; GFX11-FAKE16-NEXT: v_lshrrev_b32_e32 v180, 16, v2
+; GFX11-FAKE16-NEXT: v_lshrrev_b32_e32 v2, 16, v1
+; GFX11-FAKE16-NEXT: v_bfe_u32 v65, v39, 16, 1
+; GFX11-FAKE16-NEXT: v_cndmask_b32_e32 v64, v66, v71, vcc_lo
+; GFX11-FAKE16-NEXT: v_lshlrev_b32_e32 v66, 16, v3
; GFX11-FAKE16-NEXT: v_and_b32_e32 v3, 0xffff0000, v3
-; GFX11-FAKE16-NEXT: v_or_b32_e32 v53, 0x400000, v49
-; GFX11-FAKE16-NEXT: v_add3_u32 v51, v51, v49, 0x7fff
-; GFX11-FAKE16-NEXT: v_perm_b32 v98, v50, v32, 0x7060302
-; GFX11-FAKE16-NEXT: v_bfe_u32 v50, v4, 16, 1
-; GFX11-FAKE16-NEXT: v_add_f32_e32 v52, 0x40c00000, v52
-; GFX11-FAKE16-NEXT: v_cmp_u_f32_e32 vcc_lo, v49, v49
+; GFX11-FAKE16-NEXT: v_or_b32_e32 v67, 0x400000, v39
+; GFX11-FAKE16-NEXT: v_add3_u32 v65, v65, v39, 0x7fff
+; GFX11-FAKE16-NEXT: v_perm_b32 v114, v64, v32, 0x7060302
+; GFX11-FAKE16-NEXT: v_bfe_u32 v64, v4, 16, 1
+; GFX11-FAKE16-NEXT: v_add_f32_e32 v66, 0x40c00000, v66
+; GFX11-FAKE16-NEXT: v_cmp_u_f32_e32 vcc_lo, v39, v39
; GFX11-FAKE16-NEXT: v_add_f32_e32 v3, 0x40c00000, v3
-; GFX11-FAKE16-NEXT: v_or_b32_e32 v54, 0x400000, v4
-; GFX11-FAKE16-NEXT: v_add3_u32 v50, v50, v4, 0x7fff
-; GFX11-FAKE16-NEXT: v_bfe_u32 v55, v52, 16, 1
-; GFX11-FAKE16-NEXT: v_cndmask_b32_e32 v49, v51, v53, vcc_lo
+; GFX11-FAKE16-NEXT: v_or_b32_e32 v68, 0x400000, v4
+; GFX11-FAKE16-NEXT: v_add3_u32 v64, v64, v4, 0x7fff
+; GFX11-FAKE16-NEXT: v_bfe_u32 v71, v66, 16, 1
+; GFX11-FAKE16-NEXT: v_cndmask_b32_e32 v39, v65, v67, vcc_lo
; GFX11-FAKE16-NEXT: v_cmp_u_f32_e32 vcc_lo, v4, v4
-; GFX11-FAKE16-NEXT: v_bfe_u32 v51, v3, 16, 1
-; GFX11-FAKE16-NEXT: v_or_b32_e32 v53, 0x400000, v52
-; GFX11-FAKE16-NEXT: v_lshrrev_b32_e32 v75, 16, v98
-; GFX11-FAKE16-NEXT: v_lshrrev_b32_e32 v76, 8, v98
-; GFX11-FAKE16-NEXT: v_cndmask_b32_e32 v4, v50, v54, vcc_lo
-; GFX11-FAKE16-NEXT: v_add3_u32 v50, v55, v52, 0x7fff
-; GFX11-FAKE16-NEXT: v_lshlrev_b32_e32 v54, 16, v6
-; GFX11-FAKE16-NEXT: v_cmp_u_f32_e32 vcc_lo, v52, v52
-; GFX11-FAKE16-NEXT: v_add3_u32 v51, v51, v3, 0x7fff
-; GFX11-FAKE16-NEXT: v_or_b32_e32 v55, 0x400000, v3
+; GFX11-FAKE16-NEXT: v_bfe_u32 v65, v3, 16, 1
+; GFX11-FAKE16-NEXT: v_or_b32_e32 v67, 0x400000, v66
+; GFX11-FAKE16-NEXT: v_perm_b32 v113, v31, v30, 0x7060302
+; GFX11-FAKE16-NEXT: v_lshrrev_b32_e32 v76, 24, v115
+; GFX11-FAKE16-NEXT: v_cndmask_b32_e32 v4, v64, v68, vcc_lo
+; GFX11-FAKE16-NEXT: v_add3_u32 v64, v71, v66, 0x7fff
+; GFX11-FAKE16-NEXT: v_lshlrev_b32_e32 v68, 16, v6
+; GFX11-FAKE16-NEXT: v_cmp_u_f32_e32 vcc_lo, v66, v66
+; GFX11-FAKE16-NEXT: v_add3_u32 v65, v65, v3, 0x7fff
+; GFX11-FAKE16-NEXT: v_or_b32_e32 v71, 0x400000, v3
; GFX11-FAKE16-NEXT: v_and_b32_e32 v6, 0xffff0000, v6
-; GFX11-FAKE16-NEXT: v_add_f32_e32 v52, 0x40c00000, v54
-; GFX11-FAKE16-NEXT: v_cndmask_b32_e32 v50, v50, v53, vcc_lo
+; GFX11-FAKE16-NEXT: v_add_f32_e32 v66, 0x40c00000, v68
+; GFX11-FAKE16-NEXT: v_cndmask_b32_e32 v64, v64, v67, vcc_lo
; GFX11-FAKE16-NEXT: v_cmp_u_f32_e32 vcc_lo, v3, v3
; GFX11-FAKE16-NEXT: s_delay_alu instid0(VALU_DEP_4)
-; GFX11-FAKE16-NEXT: v_dual_add_f32 v6, 0x40c00000, v6 :: v_dual_lshlrev_b32 v53, 16, v5
+; GFX11-FAKE16-NEXT: v_dual_add_f32 v6, 0x40c00000, v6 :: v_dual_lshlrev_b32 v67, 16, v5
; GFX11-FAKE16-NEXT: v_and_b32_e32 v5, 0xffff0000, v5
-; GFX11-FAKE16-NEXT: v_or_b32_e32 v54, 0x400000, v52
-; GFX11-FAKE16-NEXT: v_cndmask_b32_e32 v3, v51, v55, vcc_lo
-; GFX11-FAKE16-NEXT: v_bfe_u32 v51, v52, 16, 1
-; GFX11-FAKE16-NEXT: v_add_f32_e32 v53, 0x40c00000, v53
-; GFX11-FAKE16-NEXT: v_cmp_u_f32_e32 vcc_lo, v52, v52
+; GFX11-FAKE16-NEXT: v_or_b32_e32 v68, 0x400000, v66
+; GFX11-FAKE16-NEXT: v_cndmask_b32_e32 v3, v65, v71, vcc_lo
+; GFX11-FAKE16-NEXT: v_bfe_u32 v65, v66, 16, 1
+; GFX11-FAKE16-NEXT: v_add_f32_e32 v67, 0x40c00000, v67
+; GFX11-FAKE16-NEXT: v_cmp_u_f32_e32 vcc_lo, v66, v66
; GFX11-FAKE16-NEXT: v_add_f32_e32 v5, 0x40c00000, v5
-; GFX11-FAKE16-NEXT: v_perm_b32 v100, v3, v50, 0x7060302
+; GFX11-FAKE16-NEXT: v_perm_b32 v116, v3, v64, 0x7060302
; GFX11-FAKE16-NEXT: v_bfe_u32 v3, v6, 16, 1
-; GFX11-FAKE16-NEXT: v_add3_u32 v51, v51, v52, 0x7fff
-; GFX11-FAKE16-NEXT: v_or_b32_e32 v55, 0x400000, v6
-; GFX11-FAKE16-NEXT: v_bfe_u32 v66, v53, 16, 1
-; GFX11-FAKE16-NEXT: v_bfe_u32 v52, v5, 16, 1
+; GFX11-FAKE16-NEXT: v_add3_u32 v65, v65, v66, 0x7fff
+; GFX11-FAKE16-NEXT: v_or_b32_e32 v71, 0x400000, v6
+; GFX11-FAKE16-NEXT: v_bfe_u32 v80, v67, 16, 1
+; GFX11-FAKE16-NEXT: v_bfe_u32 v66, v5, 16, 1
; GFX11-FAKE16-NEXT: v_add3_u32 v3, v3, v6, 0x7fff
-; GFX11-FAKE16-NEXT: v_cndmask_b32_e32 v51, v51, v54, vcc_lo
+; GFX11-FAKE16-NEXT: v_cndmask_b32_e32 v65, v65, v68, vcc_lo
; GFX11-FAKE16-NEXT: v_cmp_u_f32_e32 vcc_lo, v6, v6
-; GFX11-FAKE16-NEXT: v_add3_u32 v6, v66, v53, 0x7fff
-; GFX11-FAKE16-NEXT: v_or_b32_e32 v54, 0x400000, v53
-; GFX11-FAKE16-NEXT: v_add3_u32 v52, v52, v5, 0x7fff
-; GFX11-FAKE16-NEXT: v_or_b32_e32 v66, 0x400000, v5
-; GFX11-FAKE16-NEXT: v_cndmask_b32_e32 v3, v3, v55, vcc_lo
-; GFX11-FAKE16-NEXT: v_lshlrev_b32_e32 v55, 16, v8
-; GFX11-FAKE16-NEXT: v_cmp_u_f32_e32 vcc_lo, v53, v53
+; GFX11-FAKE16-NEXT: v_add3_u32 v6, v80, v67, 0x7fff
+; GFX11-FAKE16-NEXT: v_or_b32_e32 v68, 0x400000, v67
+; GFX11-FAKE16-NEXT: v_add3_u32 v66, v66, v5, 0x7fff
+; GFX11-FAKE16-NEXT: v_or_b32_e32 v80, 0x400000, v5
+; GFX11-FAKE16-NEXT: v_cndmask_b32_e32 v3, v3, v71, vcc_lo
+; GFX11-FAKE16-NEXT: v_lshlrev_b32_e32 v71, 16, v8
+; GFX11-FAKE16-NEXT: v_cmp_u_f32_e32 vcc_lo, v67, v67
; GFX11-FAKE16-NEXT: v_and_b32_e32 v8, 0xffff0000, v8
-; GFX11-FAKE16-NEXT: v_perm_b32 v97, v31, v30, 0x7060302
-; GFX11-FAKE16-NEXT: v_perm_b32 v103, v3, v51, 0x7060302
-; GFX11-FAKE16-NEXT: v_lshrrev_b32_e32 v147, 16, v3
-; GFX11-FAKE16-NEXT: v_dual_cndmask_b32 v53, v6, v54 :: v_dual_add_f32 v6, 0x40c00000, v55
+; GFX11-FAKE16-NEXT: v_lshrrev_b32_e32 v1, 16, v32
+; GFX11-FAKE16-NEXT: v_perm_b32 v119, v3, v65, 0x7060302
+; GFX11-FAKE16-NEXT: v_lshrrev_b32_e32 v161, 16, v3
+; GFX11-FAKE16-NEXT: v_dual_cndmask_b32 v67, v6, v68 :: v_dual_add_f32 v6, 0x40c00000, v71
; GFX11-FAKE16-NEXT: v_cmp_u_f32_e32 vcc_lo, v5, v5
; GFX11-FAKE16-NEXT: v_add_f32_e32 v8, 0x40c00000, v8
-; GFX11-FAKE16-NEXT: v_lshlrev_b32_e32 v54, 16, v7
+; GFX11-FAKE16-NEXT: v_lshlrev_b32_e32 v68, 16, v7
; GFX11-FAKE16-NEXT: v_and_b32_e32 v7, 0xffff0000, v7
-; GFX11-FAKE16-NEXT: v_or_b32_e32 v55, 0x400000, v6
-; GFX11-FAKE16-NEXT: v_cndmask_b32_e32 v5, v52, v66, vcc_lo
-; GFX11-FAKE16-NEXT: v_bfe_u32 v52, v6, 16, 1
-; GFX11-FAKE16-NEXT: v_add_f32_e32 v54, 0x40c00000, v54
+; GFX11-FAKE16-NEXT: v_or_b32_e32 v71, 0x400000, v6
+; GFX11-FAKE16-NEXT: v_cndmask_b32_e32 v5, v66, v80, vcc_lo
+; GFX11-FAKE16-NEXT: v_bfe_u32 v66, v6, 16, 1
+; GFX11-FAKE16-NEXT: v_add_f32_e32 v68, 0x40c00000, v68
; GFX11-FAKE16-NEXT: v_cmp_u_f32_e32 vcc_lo, v6, v6
; GFX11-FAKE16-NEXT: v_add_f32_e32 v7, 0x40c00000, v7
-; GFX11-FAKE16-NEXT: v_perm_b32 v102, v5, v53, 0x7060302
+; GFX11-FAKE16-NEXT: v_perm_b32 v118, v5, v67, 0x7060302
; GFX11-FAKE16-NEXT: v_bfe_u32 v5, v8, 16, 1
-; GFX11-FAKE16-NEXT: v_add3_u32 v52, v52, v6, 0x7fff
-; GFX11-FAKE16-NEXT: v_or_b32_e32 v66, 0x400000, v8
-; GFX11-FAKE16-NEXT: v_bfe_u32 v67, v54, 16, 1
-; GFX11-FAKE16-NEXT: v_perm_b32 v99, v2, v1, 0x7060302
+; GFX11-FAKE16-NEXT: v_add3_u32 v66, v66, v6, 0x7fff
+; GFX11-FAKE16-NEXT: v_or_b32_e32 v80, 0x400000, v8
+; GFX11-FAKE16-NEXT: v_bfe_u32 v81, v68, 16, 1
+; GFX11-FAKE16-NEXT: v_perm_b32 v117, v4, v39, 0x7060302
; GFX11-FAKE16-NEXT: v_add3_u32 v5, v5, v8, 0x7fff
-; GFX11-FAKE16-NEXT: v_cndmask_b32_e32 v6, v52, v55, vcc_lo
+; GFX11-FAKE16-NEXT: v_cndmask_b32_e32 v6, v66, v71, vcc_lo
; GFX11-FAKE16-NEXT: v_cmp_u_f32_e32 vcc_lo, v8, v8
-; GFX11-FAKE16-NEXT: v_bfe_u32 v52, v7, 16, 1
-; GFX11-FAKE16-NEXT: v_add3_u32 v8, v67, v54, 0x7fff
-; GFX11-FAKE16-NEXT: v_or_b32_e32 v55, 0x400000, v54
-; GFX11-FAKE16-NEXT: v_or_b32_e32 v67, 0x400000, v7
-; GFX11-FAKE16-NEXT: v_cndmask_b32_e32 v5, v5, v66, vcc_lo
-; GFX11-FAKE16-NEXT: v_lshlrev_b32_e32 v66, 16, v10
-; GFX11-FAKE16-NEXT: v_cmp_u_f32_e32 vcc_lo, v54, v54
-; GFX11-FAKE16-NEXT: v_add3_u32 v52, v52, v7, 0x7fff
+; GFX11-FAKE16-NEXT: v_bfe_u32 v66, v7, 16, 1
+; GFX11-FAKE16-NEXT: v_add3_u32 v8, v81, v68, 0x7fff
+; GFX11-FAKE16-NEXT: v_or_b32_e32 v71, 0x400000, v68
+; GFX11-FAKE16-NEXT: v_or_b32_e32 v81, 0x400000, v7
+; GFX11-FAKE16-NEXT: v_dual_cndmask_b32 v5, v5, v80 :: v_dual_lshlrev_b32 v80, 16, v10
+; GFX11-FAKE16-NEXT: v_cmp_u_f32_e32 vcc_lo, v68, v68
+; GFX11-FAKE16-NEXT: v_add3_u32 v66, v66, v7, 0x7fff
; GFX11-FAKE16-NEXT: v_and_b32_e32 v10, 0xffff0000, v10
-; GFX11-FAKE16-NEXT: v_lshrrev_b32_e32 v166, 16, v2
-; GFX11-FAKE16-NEXT: v_lshrrev_b32_e32 v2, 16, v1
-; GFX11-FAKE16-NEXT: v_cndmask_b32_e32 v54, v8, v55, vcc_lo
-; GFX11-FAKE16-NEXT: v_add_f32_e32 v8, 0x40c00000, v66
+; GFX11-FAKE16-NEXT: v_lshrrev_b32_e32 v32, 16, v30
+; GFX11-FAKE16-NEXT: v_lshrrev_b32_e32 v30, 16, v28
+; GFX11-FAKE16-NEXT: v_cndmask_b32_e32 v68, v8, v71, vcc_lo
+; GFX11-FAKE16-NEXT: v_add_f32_e32 v8, 0x40c00000, v80
; GFX11-FAKE16-NEXT: v_cmp_u_f32_e32 vcc_lo, v7, v7
-; GFX11-FAKE16-NEXT: v_dual_add_f32 v10, 0x40c00000, v10 :: v_dual_lshlrev_b32 v55, 16, v9
+; GFX11-FAKE16-NEXT: v_dual_add_f32 v10, 0x40c00000, v10 :: v_dual_lshlrev_b32 v71, 16, v9
; GFX11-FAKE16-NEXT: v_and_b32_e32 v9, 0xffff0000, v9
; GFX11-FAKE16-NEXT: s_delay_alu instid0(VALU_DEP_4)
-; GFX11-FAKE16-NEXT: v_or_b32_e32 v66, 0x400000, v8
-; GFX11-FAKE16-NEXT: v_cndmask_b32_e32 v7, v52, v67, vcc_lo
-; GFX11-FAKE16-NEXT: v_bfe_u32 v52, v8, 16, 1
-; GFX11-FAKE16-NEXT: v_add_f32_e32 v55, 0x40c00000, v55
+; GFX11-FAKE16-NEXT: v_or_b32_e32 v80, 0x400000, v8
+; GFX11-FAKE16-NEXT: v_cndmask_b32_e32 v7, v66, v81, vcc_lo
+; GFX11-FAKE16-NEXT: v_bfe_u32 v66, v8, 16, 1
+; GFX11-FAKE16-NEXT: v_add_f32_e32 v71, 0x40c00000, v71
; GFX11-FAKE16-NEXT: v_cmp_u_f32_e32 vcc_lo, v8, v8
; GFX11-FAKE16-NEXT: v_add_f32_e32 v9, 0x40c00000, v9
-; GFX11-FAKE16-NEXT: v_perm_b32 v182, v7, v54, 0x7060302
+; GFX11-FAKE16-NEXT: v_perm_b32 v44, v7, v68, 0x7060302
; GFX11-FAKE16-NEXT: v_bfe_u32 v7, v10, 16, 1
-; GFX11-FAKE16-NEXT: v_add3_u32 v52, v52, v8, 0x7fff
-; GFX11-FAKE16-NEXT: v_or_b32_e32 v67, 0x400000, v10
-; GFX11-FAKE16-NEXT: v_bfe_u32 v112, v55, 16, 1
-; GFX11-FAKE16-NEXT: v_perm_b32 v183, v5, v6, 0x7060302
+; GFX11-FAKE16-NEXT: v_add3_u32 v66, v66, v8, 0x7fff
+; GFX11-FAKE16-NEXT: v_or_b32_e32 v81, 0x400000, v10
+; GFX11-FAKE16-NEXT: v_bfe_u32 v82, v71, 16, 1
+; GFX11-FAKE16-NEXT: v_perm_b32 v45, v5, v6, 0x7060302
; GFX11-FAKE16-NEXT: v_add3_u32 v7, v7, v10, 0x7fff
-; GFX11-FAKE16-NEXT: v_cndmask_b32_e32 v8, v52, v66, vcc_lo
+; GFX11-FAKE16-NEXT: v_cndmask_b32_e32 v8, v66, v80, vcc_lo
; GFX11-FAKE16-NEXT: v_cmp_u_f32_e32 vcc_lo, v10, v10
-; GFX11-FAKE16-NEXT: v_bfe_u32 v52, v9, 16, 1
-; GFX11-FAKE16-NEXT: v_add3_u32 v10, v112, v55, 0x7fff
-; GFX11-FAKE16-NEXT: v_or_b32_e32 v66, 0x400000, v55
-; GFX11-FAKE16-NEXT: v_or_b32_e32 v112, 0x400000, v9
-; GFX11-FAKE16-NEXT: v_cndmask_b32_e32 v7, v7, v67, vcc_lo
-; GFX11-FAKE16-NEXT: v_lshlrev_b32_e32 v67, 16, v12
-; GFX11-FAKE16-NEXT: v_cmp_u_f32_e32 vcc_lo, v55, v55
-; GFX11-FAKE16-NEXT: v_add3_u32 v52, v52, v9, 0x7fff
+; GFX11-FAKE16-NEXT: v_bfe_u32 v66, v9, 16, 1
+; GFX11-FAKE16-NEXT: v_add3_u32 v10, v82, v71, 0x7fff
+; GFX11-FAKE16-NEXT: v_or_b32_e32 v80, 0x400000, v71
+; GFX11-FAKE16-NEXT: v_or_b32_e32 v82, 0x400000, v9
+; GFX11-FAKE16-NEXT: v_cndmask_b32_e32 v7, v7, v81, vcc_lo
+; GFX11-FAKE16-NEXT: v_lshlrev_b32_e32 v81, 16, v12
+; GFX11-FAKE16-NEXT: v_cmp_u_f32_e32 vcc_lo, v71, v71
+; GFX11-FAKE16-NEXT: v_add3_u32 v66, v66, v9, 0x7fff
; GFX11-FAKE16-NEXT: v_and_b32_e32 v12, 0xffff0000, v12
-; GFX11-FAKE16-NEXT: v_perm_b32 v101, v4, v49, 0x7060302
-; GFX11-FAKE16-NEXT: v_lshrrev_b32_e32 v160, 16, v4
-; GFX11-FAKE16-NEXT: v_dual_cndmask_b32 v55, v10, v66 :: v_dual_add_f32 v10, 0x40c00000, v67
+; GFX11-FAKE16-NEXT: v_lshrrev_b32_e32 v3, 16, v64
+; GFX11-FAKE16-NEXT: v_lshrrev_b32_e32 v28, 16, v26
+; GFX11-FAKE16-NEXT: v_dual_cndmask_b32 v71, v10, v80 :: v_dual_add_f32 v10, 0x40c00000, v81
; GFX11-FAKE16-NEXT: v_cmp_u_f32_e32 vcc_lo, v9, v9
; GFX11-FAKE16-NEXT: v_add_f32_e32 v12, 0x40c00000, v12
-; GFX11-FAKE16-NEXT: v_lshrrev_b32_e32 v4, 16, v49
-; GFX11-FAKE16-NEXT: v_lshrrev_b32_e32 v3, 16, v50
-; GFX11-FAKE16-NEXT: v_bfe_u32 v66, v10, 16, 1
-; GFX11-FAKE16-NEXT: v_dual_cndmask_b32 v9, v52, v112 :: v_dual_lshlrev_b32 v52, 16, v11
-; GFX11-FAKE16-NEXT: v_bfe_u32 v67, v12, 16, 1
+; GFX11-FAKE16-NEXT: v_lshrrev_b32_e32 v26, 16, v24
+; GFX11-FAKE16-NEXT: v_lshrrev_b32_e32 v24, 16, v22
+; GFX11-FAKE16-NEXT: v_bfe_u32 v80, v10, 16, 1
+; GFX11-FAKE16-NEXT: v_dual_cndmask_b32 v9, v66, v82 :: v_dual_lshlrev_b32 v66, 16, v11
+; GFX11-FAKE16-NEXT: v_bfe_u32 v81, v12, 16, 1
; GFX11-FAKE16-NEXT: v_and_b32_e32 v11, 0xffff0000, v11
; GFX11-FAKE16-NEXT: v_cmp_u_f32_e32 vcc_lo, v10, v10
; GFX11-FAKE16-NEXT: s_delay_alu instid0(VALU_DEP_4)
-; GFX11-FAKE16-NEXT: v_perm_b32 v176, v9, v55, 0x7060302
-; GFX11-FAKE16-NEXT: v_add_f32_e32 v9, 0x40c00000, v52
-; GFX11-FAKE16-NEXT: v_add3_u32 v52, v66, v10, 0x7fff
-; GFX11-FAKE16-NEXT: v_or_b32_e32 v66, 0x400000, v10
-; GFX11-FAKE16-NEXT: v_add3_u32 v67, v67, v12, 0x7fff
-; GFX11-FAKE16-NEXT: v_or_b32_e32 v112, 0x400000, v12
-; GFX11-FAKE16-NEXT: v_bfe_u32 v113, v9, 16, 1
+; GFX11-FAKE16-NEXT: v_perm_b32 v182, v9, v71, 0x7060302
+; GFX11-FAKE16-NEXT: v_add_f32_e32 v9, 0x40c00000, v66
+; GFX11-FAKE16-NEXT: v_add3_u32 v66, v80, v10, 0x7fff
+; GFX11-FAKE16-NEXT: v_or_b32_e32 v80, 0x400000, v10
+; GFX11-FAKE16-NEXT: v_add3_u32 v81, v81, v12, 0x7fff
+; GFX11-FAKE16-NEXT: v_or_b32_e32 v82, 0x400000, v12
+; GFX11-FAKE16-NEXT: v_bfe_u32 v85, v9, 16, 1
; GFX11-FAKE16-NEXT: s_delay_alu instid0(VALU_DEP_4) | instskip(SKIP_2) | instid1(VALU_DEP_4)
-; GFX11-FAKE16-NEXT: v_dual_add_f32 v11, 0x40c00000, v11 :: v_dual_cndmask_b32 v10, v52, v66
+; GFX11-FAKE16-NEXT: v_dual_add_f32 v11, 0x40c00000, v11 :: v_dual_cndmask_b32 v10, v66, v80
; GFX11-FAKE16-NEXT: v_cmp_u_f32_e32 vcc_lo, v12, v12
-; GFX11-FAKE16-NEXT: v_or_b32_e32 v66, 0x400000, v9
-; GFX11-FAKE16-NEXT: v_add3_u32 v52, v113, v9, 0x7fff
-; GFX11-FAKE16-NEXT: v_perm_b32 v177, v7, v8, 0x7060302
-; GFX11-FAKE16-NEXT: v_lshrrev_b32_e32 v1, 16, v32
-; GFX11-FAKE16-NEXT: v_dual_cndmask_b32 v12, v67, v112 :: v_dual_lshlrev_b32 v67, 16, v14
-; GFX11-FAKE16-NEXT: v_bfe_u32 v112, v11, 16, 1
+; GFX11-FAKE16-NEXT: v_or_b32_e32 v80, 0x400000, v9
+; GFX11-FAKE16-NEXT: v_add3_u32 v66, v85, v9, 0x7fff
+; GFX11-FAKE16-NEXT: v_perm_b32 v183, v7, v8, 0x7060302
+; GFX11-FAKE16-NEXT: v_lshrrev_b32_e32 v22, 16, v20
+; GFX11-FAKE16-NEXT: v_cndmask_b32_e32 v12, v81, v82, vcc_lo
+; GFX11-FAKE16-NEXT: v_lshlrev_b32_e32 v81, 16, v14
+; GFX11-FAKE16-NEXT: v_bfe_u32 v82, v11, 16, 1
; GFX11-FAKE16-NEXT: v_cmp_u_f32_e32 vcc_lo, v9, v9
; GFX11-FAKE16-NEXT: v_and_b32_e32 v14, 0xffff0000, v14
-; GFX11-FAKE16-NEXT: v_lshrrev_b32_e32 v32, 16, v30
-; GFX11-FAKE16-NEXT: v_lshrrev_b32_e32 v30, 16, v28
-; GFX11-FAKE16-NEXT: v_lshrrev_b32_e32 v28, 16, v26
-; GFX11-FAKE16-NEXT: v_dual_cndmask_b32 v9, v52, v66 :: v_dual_add_f32 v52, 0x40c00000, v67
-; GFX11-FAKE16-NEXT: v_add3_u32 v66, v112, v11, 0x7fff
-; GFX11-FAKE16-NEXT: v_or_b32_e32 v67, 0x400000, v11
-; GFX11-FAKE16-NEXT: v_lshlrev_b32_e32 v112, 16, v13
+; GFX11-FAKE16-NEXT: v_lshrrev_b32_e32 v20, 16, v18
+; GFX11-FAKE16-NEXT: v_lshrrev_b32_e32 v18, 16, v17
+; GFX11-FAKE16-NEXT: v_lshrrev_b32_e32 v17, 16, v35
+; GFX11-FAKE16-NEXT: v_dual_cndmask_b32 v9, v66, v80 :: v_dual_add_f32 v66, 0x40c00000, v81
+; GFX11-FAKE16-NEXT: v_add3_u32 v80, v82, v11, 0x7fff
+; GFX11-FAKE16-NEXT: v_or_b32_e32 v81, 0x400000, v11
+; GFX11-FAKE16-NEXT: v_lshlrev_b32_e32 v82, 16, v13
; GFX11-FAKE16-NEXT: v_cmp_u_f32_e32 vcc_lo, v11, v11
-; GFX11-FAKE16-NEXT: v_bfe_u32 v113, v52, 16, 1
+; GFX11-FAKE16-NEXT: v_bfe_u32 v85, v66, 16, 1
; GFX11-FAKE16-NEXT: v_add_f32_e32 v14, 0x40c00000, v14
; GFX11-FAKE16-NEXT: v_and_b32_e32 v13, 0xffff0000, v13
-; GFX11-FAKE16-NEXT: v_lshrrev_b32_e32 v26, 16, v24
-; GFX11-FAKE16-NEXT: v_dual_cndmask_b32 v11, v66, v67 :: v_dual_add_f32 v66, 0x40c00000, v112
-; GFX11-FAKE16-NEXT: v_add3_u32 v67, v113, v52, 0x7fff
-; GFX11-FAKE16-NEXT: v_or_b32_e32 v112, 0x400000, v52
-; GFX11-FAKE16-NEXT: v_bfe_u32 v113, v14, 16, 1
-; GFX11-FAKE16-NEXT: v_cmp_u_f32_e32 vcc_lo, v52, v52
-; GFX11-FAKE16-NEXT: v_bfe_u32 v114, v66, 16, 1
+; GFX11-FAKE16-NEXT: v_lshrrev_b64 v[35:36], 24, v[112:113]
+; GFX11-FAKE16-NEXT: v_dual_cndmask_b32 v11, v80, v81 :: v_dual_add_f32 v80, 0x40c00000, v82
+; GFX11-FAKE16-NEXT: v_add3_u32 v81, v85, v66, 0x7fff
+; GFX11-FAKE16-NEXT: v_or_b32_e32 v82, 0x400000, v66
+; GFX11-FAKE16-NEXT: v_bfe_u32 v85, v14, 16, 1
+; GFX11-FAKE16-NEXT: v_cmp_u_f32_e32 vcc_lo, v66, v66
+; GFX11-FAKE16-NEXT: v_bfe_u32 v86, v80, 16, 1
; GFX11-FAKE16-NEXT: v_add_f32_e32 v13, 0x40c00000, v13
-; GFX11-FAKE16-NEXT: v_or_b32_e32 v115, 0x400000, v66
-; GFX11-FAKE16-NEXT: v_perm_b32 v162, v11, v9, 0x7060302
-; GFX11-FAKE16-NEXT: v_cndmask_b32_e32 v52, v67, v112, vcc_lo
-; GFX11-FAKE16-NEXT: v_add3_u32 v67, v113, v14, 0x7fff
-; GFX11-FAKE16-NEXT: v_or_b32_e32 v112, 0x400000, v14
-; GFX11-FAKE16-NEXT: v_add3_u32 v113, v114, v66, 0x7fff
-; GFX11-FAKE16-NEXT: v_lshlrev_b32_e32 v114, 16, v16
+; GFX11-FAKE16-NEXT: v_or_b32_e32 v87, 0x400000, v80
+; GFX11-FAKE16-NEXT: v_perm_b32 v176, v11, v9, 0x7060302
+; GFX11-FAKE16-NEXT: v_cndmask_b32_e32 v66, v81, v82, vcc_lo
+; GFX11-FAKE16-NEXT: v_add3_u32 v81, v85, v14, 0x7fff
+; GFX11-FAKE16-NEXT: v_or_b32_e32 v82, 0x400000, v14
+; GFX11-FAKE16-NEXT: v_add3_u32 v85, v86, v80, 0x7fff
+; GFX11-FAKE16-NEXT: v_lshlrev_b32_e32 v86, 16, v16
; GFX11-FAKE16-NEXT: v_cmp_u_f32_e32 vcc_lo, v14, v14
-; GFX11-FAKE16-NEXT: v_bfe_u32 v116, v13, 16, 1
+; GFX11-FAKE16-NEXT: v_bfe_u32 v98, v13, 16, 1
; GFX11-FAKE16-NEXT: v_and_b32_e32 v16, 0xffff0000, v16
-; GFX11-FAKE16-NEXT: v_lshrrev_b32_e32 v24, 16, v22
-; GFX11-FAKE16-NEXT: v_lshrrev_b32_e32 v22, 16, v20
-; GFX11-FAKE16-NEXT: v_dual_cndmask_b32 v14, v67, v112 :: v_dual_add_f32 v67, 0x40c00000, v114
-; GFX11-FAKE16-NEXT: v_cmp_u_f32_e32 vcc_lo, v66, v66
-; GFX11-FAKE16-NEXT: v_add3_u32 v112, v116, v13, 0x7fff
+; GFX11-FAKE16-NEXT: v_lshrrev_b32_e32 v11, 16, v9
+; GFX11-FAKE16-NEXT: v_lshrrev_b32_e32 v133, 16, v7
+; GFX11-FAKE16-NEXT: v_cndmask_b32_e32 v14, v81, v82, vcc_lo
+; GFX11-FAKE16-NEXT: v_add_f32_e32 v81, 0x40c00000, v86
+; GFX11-FAKE16-NEXT: v_cmp_u_f32_e32 vcc_lo, v80, v80
+; GFX11-FAKE16-NEXT: v_add3_u32 v82, v98, v13, 0x7fff
; GFX11-FAKE16-NEXT: v_add_f32_e32 v16, 0x40c00000, v16
-; GFX11-FAKE16-NEXT: s_delay_alu instid0(VALU_DEP_4)
-; GFX11-FAKE16-NEXT: v_perm_b32 v149, v14, v52, 0x7060302
-; GFX11-FAKE16-NEXT: v_bfe_u32 v114, v67, 16, 1
-; GFX11-FAKE16-NEXT: v_cndmask_b32_e32 v66, v113, v115, vcc_lo
-; GFX11-FAKE16-NEXT: v_or_b32_e32 v113, 0x400000, v13
-; GFX11-FAKE16-NEXT: v_lshlrev_b32_e32 v115, 16, v15
+; GFX11-FAKE16-NEXT: v_perm_b32 v163, v14, v66, 0x7060302
+; GFX11-FAKE16-NEXT: v_bfe_u32 v86, v81, 16, 1
+; GFX11-FAKE16-NEXT: v_cndmask_b32_e32 v80, v85, v87, vcc_lo
+; GFX11-FAKE16-NEXT: v_or_b32_e32 v85, 0x400000, v13
+; GFX11-FAKE16-NEXT: v_lshlrev_b32_e32 v87, 16, v15
; GFX11-FAKE16-NEXT: v_cmp_u_f32_e32 vcc_lo, v13, v13
; GFX11-FAKE16-NEXT: v_and_b32_e32 v15, 0xffff0000, v15
-; GFX11-FAKE16-NEXT: v_perm_b32 v163, v12, v10, 0x7060302
-; GFX11-FAKE16-NEXT: v_lshrrev_b32_e32 v20, 16, v18
-; GFX11-FAKE16-NEXT: v_lshrrev_b32_e32 v18, 16, v17
-; GFX11-FAKE16-NEXT: v_dual_cndmask_b32 v13, v112, v113 :: v_dual_add_f32 v112, 0x40c00000, v115
-; GFX11-FAKE16-NEXT: v_add3_u32 v113, v114, v67, 0x7fff
-; GFX11-FAKE16-NEXT: v_or_b32_e32 v114, 0x400000, v67
-; GFX11-FAKE16-NEXT: v_bfe_u32 v115, v16, 16, 1
-; GFX11-FAKE16-NEXT: v_cmp_u_f32_e32 vcc_lo, v67, v67
+; GFX11-FAKE16-NEXT: v_perm_b32 v177, v12, v10, 0x7060302
+; GFX11-FAKE16-NEXT: v_lshrrev_b32_e32 v131, 16, v12
+; GFX11-FAKE16-NEXT: v_lshrrev_b32_e32 v12, 16, v10
+; GFX11-FAKE16-NEXT: v_dual_cndmask_b32 v13, v82, v85 :: v_dual_add_f32 v82, 0x40c00000, v87
+; GFX11-FAKE16-NEXT: v_add3_u32 v85, v86, v81, 0x7fff
+; GFX11-FAKE16-NEXT: v_or_b32_e32 v86, 0x400000, v81
+; GFX11-FAKE16-NEXT: v_bfe_u32 v87, v16, 16, 1
+; GFX11-FAKE16-NEXT: v_cmp_u_f32_e32 vcc_lo, v81, v81
; GFX11-FAKE16-NEXT: v_add_f32_e32 v15, 0x40c00000, v15
-; GFX11-FAKE16-NEXT: v_bfe_u32 v116, v112, 16, 1
-; GFX11-FAKE16-NEXT: v_or_b32_e32 v117, 0x400000, v112
-; GFX11-FAKE16-NEXT: v_perm_b32 v148, v13, v66, 0x7060302
-; GFX11-FAKE16-NEXT: v_cndmask_b32_e32 v67, v113, v114, vcc_lo
-; GFX11-FAKE16-NEXT: v_add3_u32 v114, v115, v16, 0x7fff
-; GFX11-FAKE16-NEXT: v_or_b32_e32 v115, 0x400000, v16
+; GFX11-FAKE16-NEXT: v_bfe_u32 v98, v82, 16, 1
+; GFX11-FAKE16-NEXT: v_or_b32_e32 v99, 0x400000, v82
+; GFX11-FAKE16-NEXT: v_perm_b32 v162, v13, v80, 0x7060302
+; GFX11-FAKE16-NEXT: v_cndmask_b32_e32 v81, v85, v86, vcc_lo
+; GFX11-FAKE16-NEXT: v_add3_u32 v86, v87, v16, 0x7fff
+; GFX11-FAKE16-NEXT: v_or_b32_e32 v87, 0x400000, v16
; GFX11-FAKE16-NEXT: v_cmp_u_f32_e32 vcc_lo, v16, v16
-; GFX11-FAKE16-NEXT: v_bfe_u32 v113, v15, 16, 1
-; GFX11-FAKE16-NEXT: v_add3_u32 v116, v116, v112, 0x7fff
-; GFX11-FAKE16-NEXT: v_or_b32_e32 v118, 0x400000, v15
-; GFX11-FAKE16-NEXT: v_lshrrev_b32_e32 v17, 16, v33
-; GFX11-FAKE16-NEXT: v_cndmask_b32_e32 v16, v114, v115, vcc_lo
-; GFX11-FAKE16-NEXT: v_cmp_u_f32_e32 vcc_lo, v112, v112
-; GFX11-FAKE16-NEXT: v_add3_u32 v113, v113, v15, 0x7fff
-; GFX11-FAKE16-NEXT: v_lshrrev_b64 v[33:34], 24, v[96:97]
-; GFX11-FAKE16-NEXT: v_lshrrev_b64 v[34:35], 24, v[86:87]
-; GFX11-FAKE16-NEXT: v_perm_b32 v135, v16, v67, 0x7060302
-; GFX11-FAKE16-NEXT: v_cndmask_b32_e32 v112, v116, v117, vcc_lo
-; GFX11-FAKE16-NEXT: v_cmp_u_f32_e32 vcc_lo, v15, v15
-; GFX11-FAKE16-NEXT: v_lshrrev_b32_e32 v116, 16, v14
-; GFX11-FAKE16-NEXT: v_lshrrev_b32_e32 v14, 16, v52
-; GFX11-FAKE16-NEXT: v_lshrrev_b64 v[35:36], 24, v[84:85]
-; GFX11-FAKE16-NEXT: v_lshrrev_b32_e32 v132, 16, v5
-; GFX11-FAKE16-NEXT: v_cndmask_b32_e32 v15, v113, v118, vcc_lo
-; GFX11-FAKE16-NEXT: v_lshrrev_b32_e32 v118, 16, v12
-; GFX11-FAKE16-NEXT: v_lshrrev_b32_e32 v12, 16, v10
+; GFX11-FAKE16-NEXT: v_bfe_u32 v85, v15, 16, 1
+; GFX11-FAKE16-NEXT: v_add3_u32 v98, v98, v82, 0x7fff
+; GFX11-FAKE16-NEXT: v_or_b32_e32 v100, 0x400000, v15
+; GFX11-FAKE16-NEXT: v_lshrrev_b32_e32 v13, 16, v80
+; GFX11-FAKE16-NEXT: v_cndmask_b32_e32 v16, v86, v87, vcc_lo
+; GFX11-FAKE16-NEXT: v_cmp_u_f32_e32 vcc_lo, v82, v82
+; GFX11-FAKE16-NEXT: v_add3_u32 v85, v85, v15, 0x7fff
; GFX11-FAKE16-NEXT: v_lshrrev_b32_e32 v10, 16, v8
; GFX11-FAKE16-NEXT: v_lshrrev_b32_e32 v8, 16, v6
-; GFX11-FAKE16-NEXT: v_perm_b32 v134, v15, v112, 0x7060302
-; GFX11-FAKE16-NEXT: v_lshrrev_b32_e32 v15, 16, v112
-; GFX11-FAKE16-NEXT: v_lshrrev_b32_e32 v112, 16, v31
-; GFX11-FAKE16-NEXT: v_lshrrev_b32_e32 v31, 16, v48
-; GFX11-FAKE16-NEXT: v_lshrrev_b32_e32 v6, 16, v51
-; GFX11-FAKE16-NEXT: v_lshrrev_b64 v[48:49], 24, v[134:135]
-; GFX11-FAKE16-NEXT: v_lshrrev_b64 v[49:50], 24, v[148:149]
-; GFX11-FAKE16-NEXT: v_lshrrev_b64 v[50:51], 24, v[162:163]
-; GFX11-FAKE16-NEXT: v_lshrrev_b64 v[51:52], 24, v[176:177]
-; GFX11-FAKE16-NEXT: v_lshrrev_b32_e32 v5, 16, v53
-; GFX11-FAKE16-NEXT: v_lshrrev_b32_e32 v117, 16, v25
-; GFX11-FAKE16-NEXT: v_lshrrev_b32_e32 v25, 16, v37
-; GFX11-FAKE16-NEXT: v_lshrrev_b64 v[52:53], 24, v[182:183]
-; GFX11-FAKE16-NEXT: v_lshrrev_b64 v[36:37], 24, v[82:83]
-; GFX11-FAKE16-NEXT: v_lshrrev_b32_e32 v114, 16, v16
-; GFX11-FAKE16-NEXT: v_lshrrev_b32_e32 v16, 16, v67
-; GFX11-FAKE16-NEXT: v_lshrrev_b32_e32 v13, 16, v66
-; GFX11-FAKE16-NEXT: v_lshrrev_b32_e32 v119, 16, v7
-; GFX11-FAKE16-NEXT: v_lshrrev_b32_e32 v7, 16, v54
-; GFX11-FAKE16-NEXT: v_lshrrev_b32_e32 v115, 16, v27
-; GFX11-FAKE16-NEXT: v_lshrrev_b32_e32 v27, 16, v38
-; GFX11-FAKE16-NEXT: v_lshrrev_b64 v[53:54], 24, v[102:103]
-; GFX11-FAKE16-NEXT: v_lshrrev_b64 v[66:67], 24, v[98:99]
-; GFX11-FAKE16-NEXT: v_lshrrev_b64 v[37:38], 24, v[80:81]
-; GFX11-FAKE16-NEXT: v_lshrrev_b32_e32 v11, 16, v9
-; GFX11-FAKE16-NEXT: v_lshrrev_b32_e32 v9, 16, v55
-; GFX11-FAKE16-NEXT: v_lshrrev_b32_e32 v113, 16, v29
-; GFX11-FAKE16-NEXT: v_lshrrev_b32_e32 v29, 16, v39
-; GFX11-FAKE16-NEXT: v_lshrrev_b64 v[54:55], 24, v[100:101]
-; GFX11-FAKE16-NEXT: v_lshrrev_b64 v[38:39], 24, v[70:71]
-; GFX11-FAKE16-NEXT: v_lshrrev_b64 v[67:68], 24, v[68:69]
-; GFX11-FAKE16-NEXT: v_lshrrev_b32_e32 v129, 24, v135
-; GFX11-FAKE16-NEXT: v_lshrrev_b32_e32 v130, 8, v135
-; GFX11-FAKE16-NEXT: v_lshrrev_b32_e32 v131, 16, v134
-; GFX11-FAKE16-NEXT: v_lshrrev_b32_e32 v134, 8, v134
+; GFX11-FAKE16-NEXT: v_perm_b32 v149, v16, v81, 0x7060302
+; GFX11-FAKE16-NEXT: v_cndmask_b32_e32 v82, v98, v99, vcc_lo
+; GFX11-FAKE16-NEXT: v_cmp_u_f32_e32 vcc_lo, v15, v15
+; GFX11-FAKE16-NEXT: v_lshrrev_b32_e32 v103, 16, v16
+; GFX11-FAKE16-NEXT: v_lshrrev_b32_e32 v16, 16, v81
+; GFX11-FAKE16-NEXT: v_lshrrev_b32_e32 v6, 16, v65
+; GFX11-FAKE16-NEXT: v_lshrrev_b64 v[80:81], 24, v[176:177]
+; GFX11-FAKE16-NEXT: v_cndmask_b32_e32 v15, v85, v100, vcc_lo
+; GFX11-FAKE16-NEXT: v_lshrrev_b64 v[85:86], 24, v[44:45]
+; GFX11-FAKE16-NEXT: v_lshrrev_b64 v[98:99], 24, v[116:117]
+; GFX11-FAKE16-NEXT: v_lshrrev_b32_e32 v129, 16, v14
+; GFX11-FAKE16-NEXT: v_lshrrev_b32_e32 v14, 16, v66
+; GFX11-FAKE16-NEXT: v_perm_b32 v148, v15, v82, 0x7060302
+; GFX11-FAKE16-NEXT: v_lshrrev_b32_e32 v15, 16, v82
+; GFX11-FAKE16-NEXT: v_lshrrev_b32_e32 v9, 16, v71
+; GFX11-FAKE16-NEXT: v_lshrrev_b32_e32 v146, 16, v5
+; GFX11-FAKE16-NEXT: v_lshrrev_b32_e32 v7, 16, v68
+; GFX11-FAKE16-NEXT: v_lshrrev_b64 v[64:65], 24, v[148:149]
+; GFX11-FAKE16-NEXT: v_lshrrev_b32_e32 v5, 16, v67
+; GFX11-FAKE16-NEXT: v_lshrrev_b32_e32 v166, 16, v4
+; GFX11-FAKE16-NEXT: v_lshrrev_b32_e32 v4, 16, v39
+; GFX11-FAKE16-NEXT: v_lshrrev_b32_e32 v39, 16, v31
+; GFX11-FAKE16-NEXT: v_lshrrev_b32_e32 v31, 16, v55
+; GFX11-FAKE16-NEXT: v_lshrrev_b32_e32 v71, 16, v29
+; GFX11-FAKE16-NEXT: v_lshrrev_b32_e32 v29, 16, v54
+; GFX11-FAKE16-NEXT: v_lshrrev_b64 v[67:68], 24, v[162:163]
+; GFX11-FAKE16-NEXT: v_lshrrev_b64 v[81:82], 24, v[182:183]
+; GFX11-FAKE16-NEXT: v_lshrrev_b64 v[86:87], 24, v[118:119]
+; GFX11-FAKE16-NEXT: v_lshrrev_b64 v[99:100], 24, v[114:115]
+; GFX11-FAKE16-NEXT: v_lshrrev_b64 v[54:55], 24, v[69:70]
+; GFX11-FAKE16-NEXT: v_lshrrev_b64 v[65:66], 24, v[52:53]
+; GFX11-FAKE16-NEXT: v_lshrrev_b64 v[36:37], 24, v[37:38]
+; GFX11-FAKE16-NEXT: v_lshrrev_b32_e32 v38, 24, v34
+; GFX11-FAKE16-NEXT: v_lshrrev_b64 v[33:34], 24, v[33:34]
; GFX11-FAKE16-NEXT: v_lshrrev_b32_e32 v135, 24, v149
; GFX11-FAKE16-NEXT: v_lshrrev_b32_e32 v144, 8, v149
; GFX11-FAKE16-NEXT: v_lshrrev_b32_e32 v145, 16, v148
; GFX11-FAKE16-NEXT: v_lshrrev_b32_e32 v148, 8, v148
; GFX11-FAKE16-NEXT: v_lshrrev_b32_e32 v149, 24, v163
; GFX11-FAKE16-NEXT: v_lshrrev_b32_e32 v150, 8, v163
-; GFX11-FAKE16-NEXT: v_lshrrev_b32_e32 v161, 16, v162
+; GFX11-FAKE16-NEXT: v_lshrrev_b32_e32 v151, 16, v162
; GFX11-FAKE16-NEXT: v_lshrrev_b32_e32 v162, 8, v162
-; GFX11-FAKE16-NEXT: v_lshrrev_b32_e32 v164, 24, v177
-; GFX11-FAKE16-NEXT: v_lshrrev_b32_e32 v165, 8, v177
+; GFX11-FAKE16-NEXT: v_lshrrev_b32_e32 v163, 24, v177
+; GFX11-FAKE16-NEXT: v_lshrrev_b32_e32 v164, 8, v177
; GFX11-FAKE16-NEXT: v_lshrrev_b32_e32 v167, 16, v176
-; GFX11-FAKE16-NEXT: v_lshrrev_b32_e32 v177, 8, v176
-; GFX11-FAKE16-NEXT: v_lshrrev_b32_e32 v179, 24, v183
-; GFX11-FAKE16-NEXT: v_lshrrev_b32_e32 v180, 8, v183
+; GFX11-FAKE16-NEXT: v_lshrrev_b32_e32 v176, 8, v176
+; GFX11-FAKE16-NEXT: v_lshrrev_b32_e32 v178, 24, v183
+; GFX11-FAKE16-NEXT: v_lshrrev_b32_e32 v179, 8, v183
; GFX11-FAKE16-NEXT: v_lshrrev_b32_e32 v181, 16, v182
; GFX11-FAKE16-NEXT: v_lshrrev_b32_e32 v183, 8, v182
-; GFX11-FAKE16-NEXT: v_lshrrev_b32_e32 v41, 24, v103
-; GFX11-FAKE16-NEXT: v_lshrrev_b32_e32 v42, 8, v103
-; GFX11-FAKE16-NEXT: v_lshrrev_b32_e32 v44, 16, v102
-; GFX11-FAKE16-NEXT: v_lshrrev_b32_e32 v45, 8, v102
-; GFX11-FAKE16-NEXT: v_lshrrev_b32_e32 v56, 24, v101
-; GFX11-FAKE16-NEXT: v_lshrrev_b32_e32 v58, 8, v101
-; GFX11-FAKE16-NEXT: v_lshrrev_b32_e32 v60, 16, v100
-; GFX11-FAKE16-NEXT: v_lshrrev_b32_e32 v62, 8, v100
-; GFX11-FAKE16-NEXT: v_lshrrev_b32_e32 v63, 24, v99
-; GFX11-FAKE16-NEXT: v_lshrrev_b32_e32 v73, 8, v99
-; GFX11-FAKE16-NEXT: v_lshrrev_b32_e32 v98, 24, v97
+; GFX11-FAKE16-NEXT: v_lshrrev_b32_e32 v41, 24, v45
+; GFX11-FAKE16-NEXT: v_lshrrev_b32_e32 v42, 8, v45
+; GFX11-FAKE16-NEXT: v_lshrrev_b32_e32 v43, 16, v44
+; GFX11-FAKE16-NEXT: v_lshrrev_b32_e32 v45, 8, v44
+; GFX11-FAKE16-NEXT: v_lshrrev_b32_e32 v47, 24, v119
+; GFX11-FAKE16-NEXT: v_lshrrev_b32_e32 v56, 8, v119
+; GFX11-FAKE16-NEXT: v_lshrrev_b32_e32 v58, 16, v118
+; GFX11-FAKE16-NEXT: v_lshrrev_b32_e32 v59, 8, v118
+; GFX11-FAKE16-NEXT: v_lshrrev_b32_e32 v62, 24, v117
+; GFX11-FAKE16-NEXT: v_lshrrev_b32_e32 v72, 8, v117
+; GFX11-FAKE16-NEXT: v_lshrrev_b32_e32 v73, 16, v116
+; GFX11-FAKE16-NEXT: v_lshrrev_b32_e32 v75, 8, v116
+; GFX11-FAKE16-NEXT: v_lshrrev_b32_e32 v78, 8, v115
+; GFX11-FAKE16-NEXT: v_lshrrev_b32_e32 v88, 16, v114
+; GFX11-FAKE16-NEXT: v_lshrrev_b32_e32 v89, 8, v114
+; GFX11-FAKE16-NEXT: v_lshrrev_b32_e32 v114, 24, v113
+; GFX11-FAKE16-NEXT: v_lshrrev_b32_e32 v113, 8, v113
+; GFX11-FAKE16-NEXT: v_lshrrev_b32_e32 v115, 16, v112
+; GFX11-FAKE16-NEXT: v_lshrrev_b32_e32 v112, 8, v112
+; GFX11-FAKE16-NEXT: v_lshrrev_b32_e32 v116, 24, v102
+; GFX11-FAKE16-NEXT: v_lshrrev_b32_e32 v102, 8, v102
+; GFX11-FAKE16-NEXT: v_lshrrev_b32_e32 v117, 16, v101
+; GFX11-FAKE16-NEXT: v_lshrrev_b32_e32 v101, 8, v101
+; GFX11-FAKE16-NEXT: v_lshrrev_b32_e32 v118, 24, v97
; GFX11-FAKE16-NEXT: v_lshrrev_b32_e32 v97, 8, v97
-; GFX11-FAKE16-NEXT: v_lshrrev_b32_e32 v99, 16, v96
+; GFX11-FAKE16-NEXT: v_lshrrev_b32_e32 v119, 16, v96
; GFX11-FAKE16-NEXT: v_lshrrev_b32_e32 v96, 8, v96
-; GFX11-FAKE16-NEXT: v_lshrrev_b32_e32 v100, 24, v87
-; GFX11-FAKE16-NEXT: v_lshrrev_b32_e32 v87, 8, v87
-; GFX11-FAKE16-NEXT: v_lshrrev_b32_e32 v101, 16, v86
-; GFX11-FAKE16-NEXT: v_lshrrev_b32_e32 v86, 8, v86
-; GFX11-FAKE16-NEXT: v_lshrrev_b32_e32 v102, 24, v85
-; GFX11-FAKE16-NEXT: v_lshrrev_b32_e32 v85, 8, v85
-; GFX11-FAKE16-NEXT: v_lshrrev_b32_e32 v103, 16, v84
-; GFX11-FAKE16-NEXT: v_lshrrev_b32_e32 v84, 8, v84
+; GFX11-FAKE16-NEXT: v_lshrrev_b32_e32 v177, 16, v83
; GFX11-FAKE16-NEXT: v_lshrrev_b32_e32 v83, 8, v83
-; GFX11-FAKE16-NEXT: v_lshrrev_b32_e32 v163, 16, v82
-; GFX11-FAKE16-NEXT: v_lshrrev_b32_e32 v82, 8, v82
-; GFX11-FAKE16-NEXT: v_lshrrev_b32_e32 v176, 24, v81
-; GFX11-FAKE16-NEXT: v_lshrrev_b32_e32 v81, 8, v81
-; GFX11-FAKE16-NEXT: v_lshrrev_b32_e32 v80, 8, v80
-; GFX11-FAKE16-NEXT: v_lshrrev_b32_e32 v182, 24, v71
-; GFX11-FAKE16-NEXT: v_lshrrev_b32_e32 v71, 8, v71
+; GFX11-FAKE16-NEXT: v_lshrrev_b32_e32 v182, 24, v70
; GFX11-FAKE16-NEXT: v_lshrrev_b32_e32 v70, 8, v70
-; GFX11-FAKE16-NEXT: v_lshrrev_b32_e32 v69, 16, v77
+; GFX11-FAKE16-NEXT: v_lshrrev_b32_e32 v69, 8, v69
+; GFX11-FAKE16-NEXT: v_lshrrev_b32_e32 v44, 24, v53
+; GFX11-FAKE16-NEXT: v_lshrrev_b32_e32 v53, 8, v53
+; GFX11-FAKE16-NEXT: v_lshrrev_b32_e32 v52, 8, v52
; GFX11-FAKE16-NEXT: .LBB90_4: ; %end
; GFX11-FAKE16-NEXT: s_or_b32 exec_lo, exec_lo, s0
-; GFX11-FAKE16-NEXT: v_lshlrev_b16 v55, 8, v76
+; GFX11-FAKE16-NEXT: v_lshlrev_b16 v37, 8, v89
; GFX11-FAKE16-NEXT: v_and_b32_e32 v1, 0xff, v1
-; GFX11-FAKE16-NEXT: v_lshlrev_b16 v39, 8, v66
+; GFX11-FAKE16-NEXT: v_lshlrev_b16 v34, 8, v99
; GFX11-FAKE16-NEXT: v_and_b32_e32 v2, 0xff, v2
-; GFX11-FAKE16-NEXT: v_lshlrev_b16 v65, 8, v63
-; GFX11-FAKE16-NEXT: v_lshlrev_b16 v54, 8, v54
-; GFX11-FAKE16-NEXT: v_or_b32_e32 v1, v1, v55
-; GFX11-FAKE16-NEXT: v_and_b32_e32 v55, 0xff, v75
+; GFX11-FAKE16-NEXT: v_lshlrev_b16 v51, 8, v76
; GFX11-FAKE16-NEXT: v_and_b32_e32 v3, 0xff, v3
-; GFX11-FAKE16-NEXT: v_and_b32_e32 v66, 0xff, v60
+; GFX11-FAKE16-NEXT: v_or_b32_e32 v1, v1, v37
+; GFX11-FAKE16-NEXT: v_and_b32_e32 v37, 0xff, v88
+; GFX11-FAKE16-NEXT: v_lshlrev_b16 v55, 8, v75
+; GFX11-FAKE16-NEXT: v_and_b32_e32 v66, 0xff, v73
; GFX11-FAKE16-NEXT: v_and_b32_e32 v4, 0xff, v4
; GFX11-FAKE16-NEXT: v_and_b32_e32 v1, 0xffff, v1
-; GFX11-FAKE16-NEXT: v_or_b32_e32 v39, v55, v39
-; GFX11-FAKE16-NEXT: v_lshlrev_b16 v55, 8, v73
-; GFX11-FAKE16-NEXT: v_or_b32_e32 v54, v66, v54
-; GFX11-FAKE16-NEXT: v_lshlrev_b16 v66, 8, v56
-; GFX11-FAKE16-NEXT: v_and_b32_e32 v5, 0xff, v5
-; GFX11-FAKE16-NEXT: v_lshlrev_b32_e32 v39, 16, v39
-; GFX11-FAKE16-NEXT: v_or_b32_e32 v2, v2, v55
+; GFX11-FAKE16-NEXT: v_or_b32_e32 v34, v37, v34
+; GFX11-FAKE16-NEXT: v_lshlrev_b16 v37, 8, v78
+; GFX11-FAKE16-NEXT: v_or_b32_e32 v3, v3, v55
; GFX11-FAKE16-NEXT: v_and_b32_e32 v55, 0xff, v166
-; GFX11-FAKE16-NEXT: v_lshlrev_b16 v53, 8, v53
+; GFX11-FAKE16-NEXT: v_and_b32_e32 v5, 0xff, v5
+; GFX11-FAKE16-NEXT: v_lshlrev_b32_e32 v34, 16, v34
+; GFX11-FAKE16-NEXT: v_or_b32_e32 v2, v2, v37
+; GFX11-FAKE16-NEXT: v_and_b32_e32 v37, 0xff, v180
+; GFX11-FAKE16-NEXT: v_and_b32_e32 v3, 0xffff, v3
; GFX11-FAKE16-NEXT: v_and_b32_e32 v6, 0xff, v6
-; GFX11-FAKE16-NEXT: v_or_b32_e32 v1, v1, v39
+; GFX11-FAKE16-NEXT: v_or_b32_e32 v1, v1, v34
; GFX11-FAKE16-NEXT: v_and_b32_e32 v2, 0xffff, v2
-; GFX11-FAKE16-NEXT: v_or_b32_e32 v55, v55, v65
-; GFX11-FAKE16-NEXT: v_lshlrev_b16 v65, 8, v62
+; GFX11-FAKE16-NEXT: v_or_b32_e32 v37, v37, v51
+; GFX11-FAKE16-NEXT: v_lshlrev_b16 v51, 8, v98
; GFX11-FAKE16-NEXT: v_and_b32_e32 v7, 0xff, v7
-; GFX11-FAKE16-NEXT: v_lshlrev_b16 v52, 8, v52
; GFX11-FAKE16-NEXT: v_and_b32_e32 v8, 0xff, v8
-; GFX11-FAKE16-NEXT: v_lshlrev_b32_e32 v39, 16, v55
-; GFX11-FAKE16-NEXT: v_or_b32_e32 v3, v3, v65
-; GFX11-FAKE16-NEXT: v_lshlrev_b16 v55, 8, v58
-; GFX11-FAKE16-NEXT: v_and_b32_e32 v65, 0xff, v160
-; GFX11-FAKE16-NEXT: v_lshlrev_b16 v51, 8, v51
-; GFX11-FAKE16-NEXT: v_or_b32_e32 v2, v2, v39
-; GFX11-FAKE16-NEXT: v_and_b32_e32 v3, 0xffff, v3
-; GFX11-FAKE16-NEXT: v_lshlrev_b32_e32 v39, 16, v54
-; GFX11-FAKE16-NEXT: v_or_b32_e32 v4, v4, v55
-; GFX11-FAKE16-NEXT: v_or_b32_e32 v54, v65, v66
-; GFX11-FAKE16-NEXT: v_lshlrev_b16 v55, 8, v45
-; GFX11-FAKE16-NEXT: v_and_b32_e32 v65, 0xff, v44
-; GFX11-FAKE16-NEXT: v_or_b32_e32 v3, v3, v39
+; GFX11-FAKE16-NEXT: v_and_b32_e32 v9, 0xff, v9
+; GFX11-FAKE16-NEXT: v_lshlrev_b32_e32 v34, 16, v37
+; GFX11-FAKE16-NEXT: v_or_b32_e32 v51, v66, v51
+; GFX11-FAKE16-NEXT: v_lshlrev_b16 v37, 8, v72
+; GFX11-FAKE16-NEXT: v_lshlrev_b16 v66, 8, v62
+; GFX11-FAKE16-NEXT: v_and_b32_e32 v10, 0xff, v10
+; GFX11-FAKE16-NEXT: v_or_b32_e32 v2, v2, v34
+; GFX11-FAKE16-NEXT: v_lshlrev_b32_e32 v34, 16, v51
+; GFX11-FAKE16-NEXT: v_or_b32_e32 v4, v4, v37
+; GFX11-FAKE16-NEXT: v_or_b32_e32 v37, v55, v66
+; GFX11-FAKE16-NEXT: v_lshlrev_b16 v51, 8, v59
+; GFX11-FAKE16-NEXT: v_and_b32_e32 v55, 0xff, v58
+; GFX11-FAKE16-NEXT: v_lshlrev_b16 v66, 8, v86
+; GFX11-FAKE16-NEXT: v_or_b32_e32 v3, v3, v34
; GFX11-FAKE16-NEXT: v_and_b32_e32 v4, 0xffff, v4
-; GFX11-FAKE16-NEXT: v_lshlrev_b32_e32 v39, 16, v54
-; GFX11-FAKE16-NEXT: v_or_b32_e32 v5, v5, v55
-; GFX11-FAKE16-NEXT: v_or_b32_e32 v53, v65, v53
-; GFX11-FAKE16-NEXT: v_lshlrev_b16 v54, 8, v42
-; GFX11-FAKE16-NEXT: v_and_b32_e32 v55, 0xff, v147
-; GFX11-FAKE16-NEXT: v_lshlrev_b16 v65, 8, v41
-; GFX11-FAKE16-NEXT: v_or_b32_e32 v4, v4, v39
+; GFX11-FAKE16-NEXT: v_lshlrev_b32_e32 v34, 16, v37
+; GFX11-FAKE16-NEXT: v_or_b32_e32 v5, v5, v51
+; GFX11-FAKE16-NEXT: v_or_b32_e32 v37, v55, v66
+; GFX11-FAKE16-NEXT: v_lshlrev_b16 v51, 8, v56
+; GFX11-FAKE16-NEXT: v_and_b32_e32 v55, 0xff, v161
+; GFX11-FAKE16-NEXT: v_lshlrev_b16 v66, 8, v47
+; GFX11-FAKE16-NEXT: v_or_b32_e32 v4, v4, v34
; GFX11-FAKE16-NEXT: v_and_b32_e32 v5, 0xffff, v5
-; GFX11-FAKE16-NEXT: v_lshlrev_b32_e32 v39, 16, v53
-; GFX11-FAKE16-NEXT: v_or_b32_e32 v6, v6, v54
-; GFX11-FAKE16-NEXT: v_or_b32_e32 v53, v55, v65
-; GFX11-FAKE16-NEXT: v_lshlrev_b16 v54, 8, v183
-; GFX11-FAKE16-NEXT: v_and_b32_e32 v55, 0xff, v181
-; GFX11-FAKE16-NEXT: v_or_b32_e32 v5, v5, v39
+; GFX11-FAKE16-NEXT: v_lshlrev_b32_e32 v34, 16, v37
+; GFX11-FAKE16-NEXT: v_or_b32_e32 v6, v6, v51
+; GFX11-FAKE16-NEXT: v_or_b32_e32 v37, v55, v66
+; GFX11-FAKE16-NEXT: v_lshlrev_b16 v51, 8, v45
+; GFX11-FAKE16-NEXT: v_and_b32_e32 v55, 0xff, v43
+; GFX11-FAKE16-NEXT: v_lshlrev_b16 v66, 8, v85
+; GFX11-FAKE16-NEXT: v_or_b32_e32 v5, v5, v34
; GFX11-FAKE16-NEXT: v_and_b32_e32 v6, 0xffff, v6
-; GFX11-FAKE16-NEXT: v_lshlrev_b32_e32 v39, 16, v53
-; GFX11-FAKE16-NEXT: v_or_b32_e32 v7, v7, v54
-; GFX11-FAKE16-NEXT: v_or_b32_e32 v52, v55, v52
-; GFX11-FAKE16-NEXT: v_lshlrev_b16 v53, 8, v180
-; GFX11-FAKE16-NEXT: v_and_b32_e32 v54, 0xff, v132
-; GFX11-FAKE16-NEXT: v_lshlrev_b16 v55, 8, v179
-; GFX11-FAKE16-NEXT: v_or_b32_e32 v6, v6, v39
-; GFX11-FAKE16-NEXT: v_lshlrev_b32_e32 v39, 16, v52
-; GFX11-FAKE16-NEXT: v_or_b32_e32 v8, v8, v53
+; GFX11-FAKE16-NEXT: v_lshlrev_b32_e32 v34, 16, v37
+; GFX11-FAKE16-NEXT: v_or_b32_e32 v7, v7, v51
+; GFX11-FAKE16-NEXT: v_or_b32_e32 v37, v55, v66
+; GFX11-FAKE16-NEXT: v_lshlrev_b16 v51, 8, v42
+; GFX11-FAKE16-NEXT: v_and_b32_e32 v55, 0xff, v146
+; GFX11-FAKE16-NEXT: v_lshlrev_b16 v66, 8, v41
+; GFX11-FAKE16-NEXT: v_or_b32_e32 v6, v6, v34
+; GFX11-FAKE16-NEXT: v_lshlrev_b32_e32 v34, 16, v37
+; GFX11-FAKE16-NEXT: v_or_b32_e32 v8, v8, v51
+; GFX11-FAKE16-NEXT: v_lshlrev_b16 v51, 8, v183
+; GFX11-FAKE16-NEXT: v_or_b32_e32 v37, v55, v66
+; GFX11-FAKE16-NEXT: v_and_b32_e32 v55, 0xff, v181
+; GFX11-FAKE16-NEXT: v_lshlrev_b16 v66, 8, v81
; GFX11-FAKE16-NEXT: v_and_b32_e32 v7, 0xffff, v7
-; GFX11-FAKE16-NEXT: v_or_b32_e32 v52, v54, v55
-; GFX11-FAKE16-NEXT: v_and_b32_e32 v54, 0xff, v167
-; GFX11-FAKE16-NEXT: v_and_b32_e32 v9, 0xff, v9
-; GFX11-FAKE16-NEXT: v_lshlrev_b16 v53, 8, v177
-; GFX11-FAKE16-NEXT: v_and_b32_e32 v10, 0xff, v10
-; GFX11-FAKE16-NEXT: v_lshlrev_b16 v55, 8, v165
+; GFX11-FAKE16-NEXT: v_lshlrev_b16 v68, 8, v179
; GFX11-FAKE16-NEXT: v_and_b32_e32 v8, 0xffff, v8
-; GFX11-FAKE16-NEXT: v_lshlrev_b32_e32 v52, 16, v52
-; GFX11-FAKE16-NEXT: v_or_b32_e32 v51, v54, v51
-; GFX11-FAKE16-NEXT: v_or_b32_e32 v9, v9, v53
-; GFX11-FAKE16-NEXT: v_or_b32_e32 v10, v10, v55
-; GFX11-FAKE16-NEXT: v_or_b32_e32 v7, v7, v39
-; GFX11-FAKE16-NEXT: v_or_b32_e32 v8, v8, v52
-; GFX11-FAKE16-NEXT: v_lshlrev_b32_e32 v39, 16, v51
-; GFX11-FAKE16-NEXT: v_and_b32_e32 v51, 0xff, v119
-; GFX11-FAKE16-NEXT: v_lshlrev_b16 v52, 8, v164
+; GFX11-FAKE16-NEXT: v_lshlrev_b32_e32 v37, 16, v37
+; GFX11-FAKE16-NEXT: v_or_b32_e32 v9, v9, v51
+; GFX11-FAKE16-NEXT: v_or_b32_e32 v51, v55, v66
+; GFX11-FAKE16-NEXT: v_or_b32_e32 v10, v10, v68
+; GFX11-FAKE16-NEXT: v_or_b32_e32 v7, v7, v34
+; GFX11-FAKE16-NEXT: v_or_b32_e32 v8, v8, v37
+; GFX11-FAKE16-NEXT: v_and_b32_e32 v37, 0xff, v133
+; GFX11-FAKE16-NEXT: v_lshlrev_b32_e32 v34, 16, v51
+; GFX11-FAKE16-NEXT: v_lshlrev_b16 v51, 8, v178
; GFX11-FAKE16-NEXT: v_and_b32_e32 v11, 0xff, v11
-; GFX11-FAKE16-NEXT: v_lshlrev_b16 v53, 8, v162
-; GFX11-FAKE16-NEXT: v_and_b32_e32 v54, 0xff, v161
-; GFX11-FAKE16-NEXT: v_lshlrev_b16 v50, 8, v50
+; GFX11-FAKE16-NEXT: v_lshlrev_b16 v55, 8, v176
+; GFX11-FAKE16-NEXT: v_and_b32_e32 v66, 0xff, v167
+; GFX11-FAKE16-NEXT: v_lshlrev_b16 v68, 8, v80
; GFX11-FAKE16-NEXT: v_and_b32_e32 v12, 0xff, v12
-; GFX11-FAKE16-NEXT: v_lshlrev_b16 v55, 8, v150
-; GFX11-FAKE16-NEXT: v_and_b32_e32 v65, 0xff, v118
-; GFX11-FAKE16-NEXT: v_lshlrev_b16 v66, 8, v149
-; GFX11-FAKE16-NEXT: v_or_b32_e32 v51, v51, v52
-; GFX11-FAKE16-NEXT: v_or_b32_e32 v11, v11, v53
-; GFX11-FAKE16-NEXT: v_or_b32_e32 v50, v54, v50
-; GFX11-FAKE16-NEXT: v_or_b32_e32 v12, v12, v55
-; GFX11-FAKE16-NEXT: v_or_b32_e32 v52, v65, v66
+; GFX11-FAKE16-NEXT: v_lshlrev_b16 v80, 8, v164
+; GFX11-FAKE16-NEXT: v_and_b32_e32 v81, 0xff, v131
+; GFX11-FAKE16-NEXT: v_lshlrev_b16 v82, 8, v163
+; GFX11-FAKE16-NEXT: v_or_b32_e32 v37, v37, v51
+; GFX11-FAKE16-NEXT: v_or_b32_e32 v11, v11, v55
+; GFX11-FAKE16-NEXT: v_or_b32_e32 v51, v66, v68
+; GFX11-FAKE16-NEXT: v_or_b32_e32 v12, v12, v80
+; GFX11-FAKE16-NEXT: v_or_b32_e32 v55, v81, v82
; GFX11-FAKE16-NEXT: v_and_b32_e32 v9, 0xffff, v9
; GFX11-FAKE16-NEXT: v_and_b32_e32 v10, 0xffff, v10
-; GFX11-FAKE16-NEXT: v_lshlrev_b32_e32 v51, 16, v51
+; GFX11-FAKE16-NEXT: v_lshlrev_b32_e32 v37, 16, v37
; GFX11-FAKE16-NEXT: v_and_b32_e32 v11, 0xffff, v11
-; GFX11-FAKE16-NEXT: v_lshlrev_b32_e32 v50, 16, v50
+; GFX11-FAKE16-NEXT: v_lshlrev_b32_e32 v51, 16, v51
; GFX11-FAKE16-NEXT: v_and_b32_e32 v12, 0xffff, v12
-; GFX11-FAKE16-NEXT: v_lshlrev_b32_e32 v52, 16, v52
+; GFX11-FAKE16-NEXT: v_lshlrev_b32_e32 v55, 16, v55
; GFX11-FAKE16-NEXT: s_clause 0x1
; GFX11-FAKE16-NEXT: scratch_store_b128 v0, v[1:4], off
; GFX11-FAKE16-NEXT: scratch_store_b128 v0, v[5:8], off offset:16
-; GFX11-FAKE16-NEXT: v_or_b32_e32 v1, v9, v39
-; GFX11-FAKE16-NEXT: v_or_b32_e32 v2, v10, v51
-; GFX11-FAKE16-NEXT: v_or_b32_e32 v3, v11, v50
-; GFX11-FAKE16-NEXT: v_or_b32_e32 v4, v12, v52
+; GFX11-FAKE16-NEXT: v_or_b32_e32 v1, v9, v34
+; GFX11-FAKE16-NEXT: v_or_b32_e32 v2, v10, v37
+; GFX11-FAKE16-NEXT: v_or_b32_e32 v3, v11, v51
+; GFX11-FAKE16-NEXT: v_or_b32_e32 v4, v12, v55
; GFX11-FAKE16-NEXT: v_and_b32_e32 v5, 0xff, v13
-; GFX11-FAKE16-NEXT: v_lshlrev_b16 v6, 8, v148
-; GFX11-FAKE16-NEXT: v_and_b32_e32 v7, 0xff, v145
-; GFX11-FAKE16-NEXT: v_lshlrev_b16 v8, 8, v49
+; GFX11-FAKE16-NEXT: v_lshlrev_b16 v6, 8, v162
+; GFX11-FAKE16-NEXT: v_and_b32_e32 v7, 0xff, v151
+; GFX11-FAKE16-NEXT: v_lshlrev_b16 v8, 8, v67
; GFX11-FAKE16-NEXT: v_and_b32_e32 v9, 0xff, v14
-; GFX11-FAKE16-NEXT: v_lshlrev_b16 v10, 8, v144
-; GFX11-FAKE16-NEXT: v_and_b32_e32 v11, 0xff, v116
-; GFX11-FAKE16-NEXT: v_lshlrev_b16 v12, 8, v135
+; GFX11-FAKE16-NEXT: v_lshlrev_b16 v10, 8, v150
+; GFX11-FAKE16-NEXT: v_and_b32_e32 v11, 0xff, v129
+; GFX11-FAKE16-NEXT: v_lshlrev_b16 v12, 8, v149
; GFX11-FAKE16-NEXT: v_and_b32_e32 v13, 0xff, v15
-; GFX11-FAKE16-NEXT: v_lshlrev_b16 v14, 8, v134
+; GFX11-FAKE16-NEXT: v_lshlrev_b16 v14, 8, v148
; GFX11-FAKE16-NEXT: v_or_b32_e32 v5, v5, v6
; GFX11-FAKE16-NEXT: v_or_b32_e32 v6, v7, v8
; GFX11-FAKE16-NEXT: v_or_b32_e32 v7, v9, v10
; GFX11-FAKE16-NEXT: v_or_b32_e32 v8, v11, v12
; GFX11-FAKE16-NEXT: v_or_b32_e32 v9, v13, v14
-; GFX11-FAKE16-NEXT: v_and_b32_e32 v10, 0xff, v131
-; GFX11-FAKE16-NEXT: v_lshlrev_b16 v11, 8, v48
+; GFX11-FAKE16-NEXT: v_and_b32_e32 v10, 0xff, v145
+; GFX11-FAKE16-NEXT: v_lshlrev_b16 v11, 8, v64
; GFX11-FAKE16-NEXT: v_and_b32_e32 v12, 0xff, v16
-; GFX11-FAKE16-NEXT: v_lshlrev_b16 v13, 8, v130
-; GFX11-FAKE16-NEXT: v_and_b32_e32 v14, 0xff, v114
-; GFX11-FAKE16-NEXT: v_lshlrev_b16 v15, 8, v129
+; GFX11-FAKE16-NEXT: v_lshlrev_b16 v13, 8, v144
+; GFX11-FAKE16-NEXT: v_and_b32_e32 v14, 0xff, v103
+; GFX11-FAKE16-NEXT: v_lshlrev_b16 v15, 8, v135
; GFX11-FAKE16-NEXT: v_and_b32_e32 v16, 0xff, v17
-; GFX11-FAKE16-NEXT: v_lshlrev_b16 v17, 8, v74
-; GFX11-FAKE16-NEXT: v_and_b32_e32 v39, 0xff, v72
-; GFX11-FAKE16-NEXT: v_lshlrev_b16 v48, 8, v67
+; GFX11-FAKE16-NEXT: v_lshlrev_b16 v17, 8, v79
+; GFX11-FAKE16-NEXT: v_and_b32_e32 v34, 0xff, v77
+; GFX11-FAKE16-NEXT: v_lshlrev_b16 v33, 8, v33
; GFX11-FAKE16-NEXT: v_or_b32_e32 v10, v10, v11
; GFX11-FAKE16-NEXT: v_or_b32_e32 v11, v12, v13
; GFX11-FAKE16-NEXT: v_or_b32_e32 v12, v14, v15
; GFX11-FAKE16-NEXT: v_or_b32_e32 v13, v16, v17
-; GFX11-FAKE16-NEXT: v_or_b32_e32 v14, v39, v48
+; GFX11-FAKE16-NEXT: v_or_b32_e32 v14, v34, v33
; GFX11-FAKE16-NEXT: v_and_b32_e32 v5, 0xffff, v5
; GFX11-FAKE16-NEXT: v_lshlrev_b32_e32 v6, 16, v6
; GFX11-FAKE16-NEXT: v_and_b32_e32 v7, 0xffff, v7
@@ -169981,35 +170068,35 @@ define <128 x i8> @bitcast_v64bf16_to_v128i8(<64 x bfloat> %a, i32 %b) {
; GFX11-FAKE16-NEXT: v_or_b32_e32 v8, v11, v12
; GFX11-FAKE16-NEXT: v_or_b32_e32 v9, v13, v14
; GFX11-FAKE16-NEXT: v_and_b32_e32 v10, 0xff, v18
-; GFX11-FAKE16-NEXT: v_lshlrev_b16 v11, 8, v61
-; GFX11-FAKE16-NEXT: v_and_b32_e32 v12, 0xff, v69
-; GFX11-FAKE16-NEXT: v_lshlrev_b16 v13, 8, v59
+; GFX11-FAKE16-NEXT: v_lshlrev_b16 v11, 8, v74
+; GFX11-FAKE16-NEXT: v_and_b32_e32 v12, 0xff, v132
+; GFX11-FAKE16-NEXT: v_lshlrev_b16 v13, 8, v38
; GFX11-FAKE16-NEXT: v_and_b32_e32 v14, 0xff, v19
-; GFX11-FAKE16-NEXT: v_lshlrev_b16 v15, 8, v57
-; GFX11-FAKE16-NEXT: v_and_b32_e32 v16, 0xff, v47
-; GFX11-FAKE16-NEXT: v_lshlrev_b16 v17, 8, v64
+; GFX11-FAKE16-NEXT: v_lshlrev_b16 v15, 8, v63
+; GFX11-FAKE16-NEXT: v_and_b32_e32 v16, 0xff, v61
+; GFX11-FAKE16-NEXT: v_lshlrev_b16 v17, 8, v36
; GFX11-FAKE16-NEXT: v_and_b32_e32 v18, 0xff, v20
-; GFX11-FAKE16-NEXT: v_lshlrev_b16 v19, 8, v46
+; GFX11-FAKE16-NEXT: v_lshlrev_b16 v19, 8, v60
; GFX11-FAKE16-NEXT: v_or_b32_e32 v10, v10, v11
; GFX11-FAKE16-NEXT: v_or_b32_e32 v11, v12, v13
; GFX11-FAKE16-NEXT: v_or_b32_e32 v12, v14, v15
; GFX11-FAKE16-NEXT: v_or_b32_e32 v13, v16, v17
; GFX11-FAKE16-NEXT: v_or_b32_e32 v14, v18, v19
-; GFX11-FAKE16-NEXT: v_and_b32_e32 v15, 0xff, v146
-; GFX11-FAKE16-NEXT: v_lshlrev_b16 v16, 8, v43
+; GFX11-FAKE16-NEXT: v_and_b32_e32 v15, 0xff, v160
+; GFX11-FAKE16-NEXT: v_lshlrev_b16 v16, 8, v57
; GFX11-FAKE16-NEXT: v_and_b32_e32 v17, 0xff, v21
-; GFX11-FAKE16-NEXT: v_lshlrev_b16 v18, 8, v70
-; GFX11-FAKE16-NEXT: v_and_b32_e32 v19, 0xff, v40
-; GFX11-FAKE16-NEXT: v_lshlrev_b16 v20, 8, v38
+; GFX11-FAKE16-NEXT: v_lshlrev_b16 v18, 8, v52
+; GFX11-FAKE16-NEXT: v_and_b32_e32 v19, 0xff, v46
+; GFX11-FAKE16-NEXT: v_lshlrev_b16 v20, 8, v65
; GFX11-FAKE16-NEXT: v_and_b32_e32 v21, 0xff, v22
-; GFX11-FAKE16-NEXT: v_lshlrev_b16 v22, 8, v71
-; GFX11-FAKE16-NEXT: v_and_b32_e32 v38, 0xff, v133
-; GFX11-FAKE16-NEXT: v_lshlrev_b16 v39, 8, v182
+; GFX11-FAKE16-NEXT: v_lshlrev_b16 v22, 8, v53
+; GFX11-FAKE16-NEXT: v_and_b32_e32 v33, 0xff, v147
+; GFX11-FAKE16-NEXT: v_lshlrev_b16 v34, 8, v44
; GFX11-FAKE16-NEXT: v_or_b32_e32 v15, v15, v16
; GFX11-FAKE16-NEXT: v_or_b32_e32 v16, v17, v18
; GFX11-FAKE16-NEXT: v_or_b32_e32 v17, v19, v20
; GFX11-FAKE16-NEXT: v_or_b32_e32 v18, v21, v22
-; GFX11-FAKE16-NEXT: v_or_b32_e32 v19, v38, v39
+; GFX11-FAKE16-NEXT: v_or_b32_e32 v19, v33, v34
; GFX11-FAKE16-NEXT: v_and_b32_e32 v10, 0xffff, v10
; GFX11-FAKE16-NEXT: v_lshlrev_b32_e32 v11, 16, v11
; GFX11-FAKE16-NEXT: v_and_b32_e32 v12, 0xffff, v12
@@ -170026,35 +170113,35 @@ define <128 x i8> @bitcast_v64bf16_to_v128i8(<64 x bfloat> %a, i32 %b) {
; GFX11-FAKE16-NEXT: v_or_b32_e32 v13, v16, v17
; GFX11-FAKE16-NEXT: v_or_b32_e32 v14, v18, v19
; GFX11-FAKE16-NEXT: v_and_b32_e32 v15, 0xff, v23
-; GFX11-FAKE16-NEXT: v_lshlrev_b16 v16, 8, v80
-; GFX11-FAKE16-NEXT: v_and_b32_e32 v17, 0xff, v178
-; GFX11-FAKE16-NEXT: v_lshlrev_b16 v18, 8, v37
+; GFX11-FAKE16-NEXT: v_lshlrev_b16 v16, 8, v69
+; GFX11-FAKE16-NEXT: v_and_b32_e32 v17, 0xff, v40
+; GFX11-FAKE16-NEXT: v_lshlrev_b16 v18, 8, v54
; GFX11-FAKE16-NEXT: v_and_b32_e32 v19, 0xff, v24
-; GFX11-FAKE16-NEXT: v_lshlrev_b16 v20, 8, v81
-; GFX11-FAKE16-NEXT: v_and_b32_e32 v21, 0xff, v128
-; GFX11-FAKE16-NEXT: v_lshlrev_b16 v22, 8, v176
+; GFX11-FAKE16-NEXT: v_lshlrev_b16 v20, 8, v70
+; GFX11-FAKE16-NEXT: v_and_b32_e32 v21, 0xff, v134
+; GFX11-FAKE16-NEXT: v_lshlrev_b16 v22, 8, v182
; GFX11-FAKE16-NEXT: v_and_b32_e32 v23, 0xff, v25
-; GFX11-FAKE16-NEXT: v_lshlrev_b16 v24, 8, v82
+; GFX11-FAKE16-NEXT: v_lshlrev_b16 v24, 8, v83
; GFX11-FAKE16-NEXT: v_or_b32_e32 v15, v15, v16
; GFX11-FAKE16-NEXT: v_or_b32_e32 v16, v17, v18
; GFX11-FAKE16-NEXT: v_or_b32_e32 v17, v19, v20
; GFX11-FAKE16-NEXT: v_or_b32_e32 v18, v21, v22
; GFX11-FAKE16-NEXT: v_or_b32_e32 v19, v23, v24
-; GFX11-FAKE16-NEXT: v_and_b32_e32 v20, 0xff, v163
-; GFX11-FAKE16-NEXT: v_lshlrev_b16 v21, 8, v36
+; GFX11-FAKE16-NEXT: v_and_b32_e32 v20, 0xff, v177
+; GFX11-FAKE16-NEXT: v_lshlrev_b16 v21, 8, v50
; GFX11-FAKE16-NEXT: v_and_b32_e32 v22, 0xff, v26
-; GFX11-FAKE16-NEXT: v_lshlrev_b16 v23, 8, v83
-; GFX11-FAKE16-NEXT: v_and_b32_e32 v24, 0xff, v117
-; GFX11-FAKE16-NEXT: v_lshlrev_b16 v25, 8, v151
+; GFX11-FAKE16-NEXT: v_lshlrev_b16 v23, 8, v84
+; GFX11-FAKE16-NEXT: v_and_b32_e32 v24, 0xff, v130
+; GFX11-FAKE16-NEXT: v_lshlrev_b16 v25, 8, v165
; GFX11-FAKE16-NEXT: v_and_b32_e32 v26, 0xff, v27
-; GFX11-FAKE16-NEXT: v_lshlrev_b16 v27, 8, v84
-; GFX11-FAKE16-NEXT: v_and_b32_e32 v36, 0xff, v103
-; GFX11-FAKE16-NEXT: v_lshlrev_b16 v35, 8, v35
+; GFX11-FAKE16-NEXT: v_lshlrev_b16 v27, 8, v96
+; GFX11-FAKE16-NEXT: v_and_b32_e32 v33, 0xff, v119
+; GFX11-FAKE16-NEXT: v_lshlrev_b16 v34, 8, v49
; GFX11-FAKE16-NEXT: v_or_b32_e32 v20, v20, v21
; GFX11-FAKE16-NEXT: v_or_b32_e32 v21, v22, v23
; GFX11-FAKE16-NEXT: v_or_b32_e32 v22, v24, v25
; GFX11-FAKE16-NEXT: v_or_b32_e32 v23, v26, v27
-; GFX11-FAKE16-NEXT: v_or_b32_e32 v24, v36, v35
+; GFX11-FAKE16-NEXT: v_or_b32_e32 v24, v33, v34
; GFX11-FAKE16-NEXT: v_and_b32_e32 v15, 0xffff, v15
; GFX11-FAKE16-NEXT: v_lshlrev_b32_e32 v16, 16, v16
; GFX11-FAKE16-NEXT: v_and_b32_e32 v17, 0xffff, v17
@@ -170071,31 +170158,31 @@ define <128 x i8> @bitcast_v64bf16_to_v128i8(<64 x bfloat> %a, i32 %b) {
; GFX11-FAKE16-NEXT: v_or_b32_e32 v18, v21, v22
; GFX11-FAKE16-NEXT: v_or_b32_e32 v19, v23, v24
; GFX11-FAKE16-NEXT: v_and_b32_e32 v20, 0xff, v28
-; GFX11-FAKE16-NEXT: v_lshlrev_b16 v21, 8, v85
-; GFX11-FAKE16-NEXT: v_and_b32_e32 v22, 0xff, v115
-; GFX11-FAKE16-NEXT: v_lshlrev_b16 v23, 8, v102
+; GFX11-FAKE16-NEXT: v_lshlrev_b16 v21, 8, v97
+; GFX11-FAKE16-NEXT: v_and_b32_e32 v22, 0xff, v128
+; GFX11-FAKE16-NEXT: v_lshlrev_b16 v23, 8, v118
; GFX11-FAKE16-NEXT: v_and_b32_e32 v24, 0xff, v29
-; GFX11-FAKE16-NEXT: v_lshlrev_b16 v25, 8, v86
-; GFX11-FAKE16-NEXT: v_and_b32_e32 v26, 0xff, v101
-; GFX11-FAKE16-NEXT: v_lshlrev_b16 v27, 8, v34
+; GFX11-FAKE16-NEXT: v_lshlrev_b16 v25, 8, v101
+; GFX11-FAKE16-NEXT: v_and_b32_e32 v26, 0xff, v117
+; GFX11-FAKE16-NEXT: v_lshlrev_b16 v27, 8, v48
; GFX11-FAKE16-NEXT: v_and_b32_e32 v28, 0xff, v30
-; GFX11-FAKE16-NEXT: v_lshlrev_b16 v29, 8, v87
+; GFX11-FAKE16-NEXT: v_lshlrev_b16 v29, 8, v102
; GFX11-FAKE16-NEXT: v_or_b32_e32 v20, v20, v21
; GFX11-FAKE16-NEXT: v_or_b32_e32 v21, v22, v23
; GFX11-FAKE16-NEXT: v_or_b32_e32 v22, v24, v25
; GFX11-FAKE16-NEXT: v_or_b32_e32 v23, v26, v27
; GFX11-FAKE16-NEXT: v_or_b32_e32 v24, v28, v29
-; GFX11-FAKE16-NEXT: v_and_b32_e32 v25, 0xff, v113
-; GFX11-FAKE16-NEXT: v_lshlrev_b16 v26, 8, v100
+; GFX11-FAKE16-NEXT: v_and_b32_e32 v25, 0xff, v71
+; GFX11-FAKE16-NEXT: v_lshlrev_b16 v26, 8, v116
; GFX11-FAKE16-NEXT: s_waitcnt vmcnt(0)
; GFX11-FAKE16-NEXT: v_and_b32_e32 v27, 0xff, v31
-; GFX11-FAKE16-NEXT: v_lshlrev_b16 v28, 8, v96
-; GFX11-FAKE16-NEXT: v_and_b32_e32 v29, 0xff, v99
-; GFX11-FAKE16-NEXT: v_lshlrev_b16 v30, 8, v33
+; GFX11-FAKE16-NEXT: v_lshlrev_b16 v28, 8, v112
+; GFX11-FAKE16-NEXT: v_and_b32_e32 v29, 0xff, v115
+; GFX11-FAKE16-NEXT: v_lshlrev_b16 v30, 8, v35
; GFX11-FAKE16-NEXT: v_and_b32_e32 v31, 0xff, v32
-; GFX11-FAKE16-NEXT: v_lshlrev_b16 v32, 8, v97
-; GFX11-FAKE16-NEXT: v_and_b32_e32 v33, 0xff, v112
-; GFX11-FAKE16-NEXT: v_lshlrev_b16 v34, 8, v98
+; GFX11-FAKE16-NEXT: v_lshlrev_b16 v32, 8, v113
+; GFX11-FAKE16-NEXT: v_and_b32_e32 v33, 0xff, v39
+; GFX11-FAKE16-NEXT: v_lshlrev_b16 v34, 8, v114
; GFX11-FAKE16-NEXT: v_or_b32_e32 v25, v25, v26
; GFX11-FAKE16-NEXT: v_or_b32_e32 v26, v27, v28
; GFX11-FAKE16-NEXT: v_or_b32_e32 v27, v29, v30
@@ -170123,29 +170210,33 @@ define <128 x i8> @bitcast_v64bf16_to_v128i8(<64 x bfloat> %a, i32 %b) {
; GFX11-FAKE16-NEXT: scratch_store_b128 v0, v[13:16], off offset:80
; GFX11-FAKE16-NEXT: scratch_store_b128 v0, v[17:20], off offset:96
; GFX11-FAKE16-NEXT: scratch_store_b128 v0, v[21:24], off offset:112
-; GFX11-FAKE16-NEXT: s_clause 0x15
-; GFX11-FAKE16-NEXT: scratch_load_b32 v77, off, s32 offset:12
-; GFX11-FAKE16-NEXT: scratch_load_b32 v76, off, s32 offset:16
-; GFX11-FAKE16-NEXT: scratch_load_b32 v75, off, s32 offset:20
-; GFX11-FAKE16-NEXT: scratch_load_b32 v74, off, s32 offset:24
-; GFX11-FAKE16-NEXT: scratch_load_b32 v73, off, s32 offset:28
-; GFX11-FAKE16-NEXT: scratch_load_b32 v72, off, s32 offset:32
-; GFX11-FAKE16-NEXT: scratch_load_b32 v63, off, s32 offset:36
-; GFX11-FAKE16-NEXT: scratch_load_b32 v62, off, s32 offset:40
-; GFX11-FAKE16-NEXT: scratch_load_b32 v61, off, s32 offset:44
-; GFX11-FAKE16-NEXT: scratch_load_b32 v60, off, s32 offset:48
-; GFX11-FAKE16-NEXT: scratch_load_b32 v59, off, s32 offset:52
-; GFX11-FAKE16-NEXT: scratch_load_b32 v58, off, s32 offset:56
-; GFX11-FAKE16-NEXT: scratch_load_b32 v57, off, s32 offset:60
-; GFX11-FAKE16-NEXT: scratch_load_b32 v56, off, s32 offset:64
-; GFX11-FAKE16-NEXT: scratch_load_b32 v47, off, s32 offset:68
-; GFX11-FAKE16-NEXT: scratch_load_b32 v46, off, s32 offset:72
-; GFX11-FAKE16-NEXT: scratch_load_b32 v45, off, s32 offset:76
-; GFX11-FAKE16-NEXT: scratch_load_b32 v44, off, s32 offset:80
-; GFX11-FAKE16-NEXT: scratch_load_b32 v43, off, s32 offset:84
-; GFX11-FAKE16-NEXT: scratch_load_b32 v42, off, s32 offset:88
-; GFX11-FAKE16-NEXT: scratch_load_b32 v41, off, s32 offset:92
-; GFX11-FAKE16-NEXT: scratch_load_b32 v40, off, s32 offset:96
+; GFX11-FAKE16-NEXT: s_clause 0x19
+; GFX11-FAKE16-NEXT: scratch_load_b32 v89, off, s32 offset:12
+; GFX11-FAKE16-NEXT: scratch_load_b32 v88, off, s32 offset:16
+; GFX11-FAKE16-NEXT: scratch_load_b32 v79, off, s32 offset:20
+; GFX11-FAKE16-NEXT: scratch_load_b32 v78, off, s32 offset:24
+; GFX11-FAKE16-NEXT: scratch_load_b32 v77, off, s32 offset:28
+; GFX11-FAKE16-NEXT: scratch_load_b32 v76, off, s32 offset:32
+; GFX11-FAKE16-NEXT: scratch_load_b32 v75, off, s32 offset:36
+; GFX11-FAKE16-NEXT: scratch_load_b32 v74, off, s32 offset:40
+; GFX11-FAKE16-NEXT: scratch_load_b32 v73, off, s32 offset:44
+; GFX11-FAKE16-NEXT: scratch_load_b32 v72, off, s32 offset:48
+; GFX11-FAKE16-NEXT: scratch_load_b32 v63, off, s32 offset:52
+; GFX11-FAKE16-NEXT: scratch_load_b32 v62, off, s32 offset:56
+; GFX11-FAKE16-NEXT: scratch_load_b32 v61, off, s32 offset:60
+; GFX11-FAKE16-NEXT: scratch_load_b32 v60, off, s32 offset:64
+; GFX11-FAKE16-NEXT: scratch_load_b32 v59, off, s32 offset:68
+; GFX11-FAKE16-NEXT: scratch_load_b32 v58, off, s32 offset:72
+; GFX11-FAKE16-NEXT: scratch_load_b32 v57, off, s32 offset:76
+; GFX11-FAKE16-NEXT: scratch_load_b32 v56, off, s32 offset:80
+; GFX11-FAKE16-NEXT: scratch_load_b32 v47, off, s32 offset:84
+; GFX11-FAKE16-NEXT: scratch_load_b32 v46, off, s32 offset:88
+; GFX11-FAKE16-NEXT: scratch_load_b32 v45, off, s32 offset:92
+; GFX11-FAKE16-NEXT: scratch_load_b32 v44, off, s32 offset:96
+; GFX11-FAKE16-NEXT: scratch_load_b32 v43, off, s32 offset:100
+; GFX11-FAKE16-NEXT: scratch_load_b32 v42, off, s32 offset:104
+; GFX11-FAKE16-NEXT: scratch_load_b32 v41, off, s32 offset:108
+; GFX11-FAKE16-NEXT: scratch_load_b32 v40, off, s32 offset:112
; GFX11-FAKE16-NEXT: s_waitcnt vmcnt(0)
; GFX11-FAKE16-NEXT: s_setpc_b64 s[30:31]
%cmp = icmp eq i32 %b, 0
@@ -170206,464 +170297,484 @@ define inreg <128 x i8> @bitcast_v64bf16_to_v128i8_scalar(<64 x bfloat> inreg %a
; SI-NEXT: buffer_load_dword v41, off, s[0:3], s32 offset:64
; SI-NEXT: buffer_load_dword v42, off, s[0:3], s32 offset:76
; SI-NEXT: buffer_load_dword v43, off, s[0:3], s32 offset:72
-; SI-NEXT: s_waitcnt expcnt(4)
-; SI-NEXT: v_mul_f32_e32 v59, 1.0, v1
+; SI-NEXT: v_mul_f32_e32 v46, 1.0, v1
; SI-NEXT: v_mul_f32_e32 v1, 1.0, v6
-; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:156 ; 4-byte Folded Spill
-; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_mul_f32_e32 v1, 1.0, v14
-; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:160 ; 4-byte Folded Spill
-; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_mul_f32_e32 v1, 1.0, v18
; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:164 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
+; SI-NEXT: v_mul_f32_e32 v1, 1.0, v13
+; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:168 ; 4-byte Folded Spill
+; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_mul_f32_e32 v1, 1.0, v22
-; SI-NEXT: v_mul_f32_e32 v46, 1.0, v2
+; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:172 ; 4-byte Folded Spill
+; SI-NEXT: s_waitcnt expcnt(0)
+; SI-NEXT: v_mul_f32_e32 v1, 1.0, v30
+; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:480 ; 4-byte Folded Spill
+; SI-NEXT: v_mul_f32_e32 v59, 1.0, v2
; SI-NEXT: v_mul_f32_e32 v4, 1.0, v4
-; SI-NEXT: v_mul_f32_e32 v3, 1.0, v3
+; SI-NEXT: v_mul_f32_e32 v44, 1.0, v3
; SI-NEXT: v_mul_f32_e32 v61, 1.0, v5
-; SI-NEXT: v_mul_f32_e32 v44, 1.0, v8
+; SI-NEXT: v_mul_f32_e32 v47, 1.0, v8
; SI-NEXT: v_mul_f32_e32 v7, 1.0, v7
-; SI-NEXT: v_mul_f32_e32 v56, 1.0, v10
+; SI-NEXT: v_mul_f32_e32 v60, 1.0, v10
; SI-NEXT: v_mul_f32_e32 v63, 1.0, v9
-; SI-NEXT: v_mul_f32_e32 v47, 1.0, v12
-; SI-NEXT: v_mul_f32_e32 v57, 1.0, v11
-; SI-NEXT: v_mul_f32_e32 v45, 1.0, v13
-; SI-NEXT: v_mul_f32_e32 v58, 1.0, v15
-; SI-NEXT: v_mul_f32_e32 v18, 1.0, v17
-; SI-NEXT: v_mul_f32_e32 v62, 1.0, v20
-; SI-NEXT: v_mul_f32_e32 v60, 1.0, v19
-; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:468 ; 4-byte Folded Spill
+; SI-NEXT: v_mul_f32_e32 v57, 1.0, v12
+; SI-NEXT: v_mul_f32_e32 v58, 1.0, v11
+; SI-NEXT: v_mul_f32_e32 v45, 1.0, v14
+; SI-NEXT: v_mul_f32_e32 v62, 1.0, v16
+; SI-NEXT: v_mul_f32_e32 v56, 1.0, v15
+; SI-NEXT: v_mul_f32_e32 v11, 1.0, v18
+; SI-NEXT: v_mul_f32_e32 v17, 1.0, v17
+; SI-NEXT: v_mul_f32_e32 v2, 1.0, v19
; SI-NEXT: v_mul_f32_e32 v21, 1.0, v21
-; SI-NEXT: v_mul_f32_e32 v19, 1.0, v24
-; SI-NEXT: v_mul_f32_e32 v22, 1.0, v23
-; SI-NEXT: v_mul_f32_e32 v23, 1.0, v26
-; SI-NEXT: v_mul_f32_e32 v24, 1.0, v25
-; SI-NEXT: v_mul_f32_e32 v25, 1.0, v28
+; SI-NEXT: v_mul_f32_e32 v22, 1.0, v24
+; SI-NEXT: v_mul_f32_e32 v23, 1.0, v23
+; SI-NEXT: v_mul_f32_e32 v24, 1.0, v26
+; SI-NEXT: v_mul_f32_e32 v26, 1.0, v25
+; SI-NEXT: v_mul_f32_e32 v15, 1.0, v28
; SI-NEXT: v_mul_f32_e32 v27, 1.0, v27
-; SI-NEXT: v_mul_f32_e32 v20, 1.0, v30
-; SI-NEXT: v_mul_f32_e32 v26, 1.0, v29
-; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_mul_f32_e64 v1, 1.0, s17
-; SI-NEXT: v_mul_f32_e64 v2, 1.0, s16
-; SI-NEXT: v_mul_f32_e64 v11, 1.0, s19
+; SI-NEXT: v_mul_f32_e32 v18, 1.0, v29
+; SI-NEXT: v_mul_f32_e64 v3, 1.0, s16
; SI-NEXT: v_mul_f32_e64 v8, 1.0, s18
; SI-NEXT: v_mul_f32_e64 v5, 1.0, s21
; SI-NEXT: v_mul_f32_e64 v6, 1.0, s20
+; SI-NEXT: v_mul_f32_e64 v14, 1.0, s23
; SI-NEXT: v_mul_f32_e64 v9, 1.0, s22
; SI-NEXT: v_mul_f32_e64 v10, 1.0, s25
; SI-NEXT: v_mul_f32_e64 v13, 1.0, s24
; SI-NEXT: v_mul_f32_e64 v12, 1.0, s26
-; SI-NEXT: v_mul_f32_e64 v17, 1.0, s28
+; SI-NEXT: v_mul_f32_e64 v16, 1.0, s29
+; SI-NEXT: v_mul_f32_e64 v19, 1.0, s28
; SI-NEXT: s_waitcnt vmcnt(14)
; SI-NEXT: v_cmp_ne_u32_e32 vcc, 0, v31
-; SI-NEXT: v_mul_f32_e32 v31, 1.0, v16
-; SI-NEXT: v_mul_f32_e32 v29, 1.0, v32
-; SI-NEXT: v_mul_f32_e32 v30, 1.0, v33
+; SI-NEXT: v_mul_f32_e32 v31, 1.0, v20
+; SI-NEXT: v_mul_f32_e32 v30, 1.0, v32
+; SI-NEXT: v_mul_f32_e32 v32, 1.0, v33
; SI-NEXT: s_and_b64 s[4:5], vcc, exec
-; SI-NEXT: v_mul_f32_e32 v32, 1.0, v34
-; SI-NEXT: v_mul_f32_e32 v14, 1.0, v35
+; SI-NEXT: v_mul_f32_e32 v25, 1.0, v34
+; SI-NEXT: v_mul_f32_e32 v29, 1.0, v35
+; SI-NEXT: s_waitcnt expcnt(0)
+; SI-NEXT: v_mul_f32_e32 v1, 1.0, v38
+; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:488 ; 4-byte Folded Spill
; SI-NEXT: v_mul_f32_e32 v35, 1.0, v36
; SI-NEXT: v_mul_f32_e32 v37, 1.0, v37
-; SI-NEXT: v_mul_f32_e32 v34, 1.0, v38
-; SI-NEXT: v_mul_f32_e32 v15, 1.0, v39
+; SI-NEXT: v_mul_f32_e32 v34, 1.0, v39
; SI-NEXT: v_mul_f32_e32 v38, 1.0, v48
; SI-NEXT: v_mul_f32_e32 v48, 1.0, v49
-; SI-NEXT: s_waitcnt vmcnt(13)
+; SI-NEXT: s_waitcnt vmcnt(14)
; SI-NEXT: v_mul_f32_e32 v28, 1.0, v50
-; SI-NEXT: s_waitcnt vmcnt(12)
-; SI-NEXT: v_mul_f32_e32 v51, 1.0, v51
-; SI-NEXT: s_waitcnt vmcnt(11)
+; SI-NEXT: s_waitcnt vmcnt(10) expcnt(0)
+; SI-NEXT: v_mul_f32_e32 v1, 1.0, v54
+; SI-NEXT: v_mul_f32_e32 v49, 1.0, v51
; SI-NEXT: v_mul_f32_e32 v50, 1.0, v52
+; SI-NEXT: v_mul_f32_e32 v51, 1.0, v53
+; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:492 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt vmcnt(10)
-; SI-NEXT: v_mul_f32_e32 v52, 1.0, v53
+; SI-NEXT: v_mul_f32_e32 v54, 1.0, v55
; SI-NEXT: s_waitcnt vmcnt(9)
-; SI-NEXT: v_mul_f32_e32 v33, 1.0, v54
+; SI-NEXT: v_mul_f32_e32 v52, 1.0, v40
; SI-NEXT: s_waitcnt vmcnt(8)
-; SI-NEXT: v_mul_f32_e32 v36, 1.0, v55
+; SI-NEXT: v_mul_f32_e32 v55, 1.0, v41
; SI-NEXT: s_waitcnt vmcnt(7)
-; SI-NEXT: v_mul_f32_e32 v55, 1.0, v40
+; SI-NEXT: v_mul_f32_e32 v39, 1.0, v42
; SI-NEXT: s_waitcnt vmcnt(6)
-; SI-NEXT: v_mul_f32_e32 v41, 1.0, v41
-; SI-NEXT: s_waitcnt vmcnt(5)
-; SI-NEXT: v_mul_f32_e32 v42, 1.0, v42
-; SI-NEXT: s_waitcnt vmcnt(4)
-; SI-NEXT: v_mul_f32_e32 v54, 1.0, v43
-; SI-NEXT: v_mul_f32_e64 v39, 1.0, s23
-; SI-NEXT: v_mul_f32_e64 v49, 1.0, s27
-; SI-NEXT: v_mul_f32_e64 v16, 1.0, s29
-; SI-NEXT: buffer_store_dword v46, off, s[0:3], s32 offset:484 ; 4-byte Folded Spill
-; SI-NEXT: buffer_store_dword v59, off, s[0:3], s32 offset:488 ; 4-byte Folded Spill
-; SI-NEXT: buffer_store_dword v61, off, s[0:3], s32 offset:492 ; 4-byte Folded Spill
-; SI-NEXT: buffer_store_dword v11, off, s[0:3], s32 offset:496 ; 4-byte Folded Spill
-; SI-NEXT: buffer_store_dword v8, off, s[0:3], s32 offset:500 ; 4-byte Folded Spill
-; SI-NEXT: buffer_store_dword v39, off, s[0:3], s32 offset:504 ; 4-byte Folded Spill
-; SI-NEXT: buffer_store_dword v9, off, s[0:3], s32 offset:508 ; 4-byte Folded Spill
-; SI-NEXT: buffer_store_dword v49, off, s[0:3], s32 offset:512 ; 4-byte Folded Spill
-; SI-NEXT: buffer_store_dword v12, off, s[0:3], s32 offset:516 ; 4-byte Folded Spill
-; SI-NEXT: buffer_store_dword v56, off, s[0:3], s32 offset:520 ; 4-byte Folded Spill
-; SI-NEXT: buffer_store_dword v45, off, s[0:3], s32 offset:524 ; 4-byte Folded Spill
-; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:528 ; 4-byte Folded Spill
-; SI-NEXT: buffer_store_dword v2, off, s[0:3], s32 offset:532 ; 4-byte Folded Spill
-; SI-NEXT: buffer_store_dword v4, off, s[0:3], s32 offset:536 ; 4-byte Folded Spill
-; SI-NEXT: buffer_store_dword v3, off, s[0:3], s32 offset:540 ; 4-byte Folded Spill
-; SI-NEXT: buffer_store_dword v18, off, s[0:3], s32 offset:544 ; 4-byte Folded Spill
-; SI-NEXT: buffer_store_dword v5, off, s[0:3], s32 offset:548 ; 4-byte Folded Spill
-; SI-NEXT: buffer_store_dword v6, off, s[0:3], s32 offset:552 ; 4-byte Folded Spill
-; SI-NEXT: buffer_store_dword v44, off, s[0:3], s32 offset:556 ; 4-byte Folded Spill
-; SI-NEXT: buffer_store_dword v7, off, s[0:3], s32 offset:560 ; 4-byte Folded Spill
-; SI-NEXT: buffer_store_dword v21, off, s[0:3], s32 offset:564 ; 4-byte Folded Spill
-; SI-NEXT: buffer_store_dword v10, off, s[0:3], s32 offset:568 ; 4-byte Folded Spill
-; SI-NEXT: buffer_store_dword v13, off, s[0:3], s32 offset:572 ; 4-byte Folded Spill
-; SI-NEXT: buffer_store_dword v47, off, s[0:3], s32 offset:576 ; 4-byte Folded Spill
-; SI-NEXT: buffer_store_dword v57, off, s[0:3], s32 offset:580 ; 4-byte Folded Spill
-; SI-NEXT: buffer_store_dword v23, off, s[0:3], s32 offset:584 ; 4-byte Folded Spill
-; SI-NEXT: buffer_store_dword v24, off, s[0:3], s32 offset:588 ; 4-byte Folded Spill
-; SI-NEXT: buffer_store_dword v16, off, s[0:3], s32 offset:592 ; 4-byte Folded Spill
-; SI-NEXT: buffer_store_dword v17, off, s[0:3], s32 offset:596 ; 4-byte Folded Spill
-; SI-NEXT: buffer_store_dword v31, off, s[0:3], s32 offset:600 ; 4-byte Folded Spill
+; SI-NEXT: v_mul_f32_e32 v43, 1.0, v43
+; SI-NEXT: s_waitcnt expcnt(0)
+; SI-NEXT: v_mul_f32_e64 v1, 1.0, s17
+; SI-NEXT: v_mul_f32_e64 v42, 1.0, s19
+; SI-NEXT: v_mul_f32_e64 v20, 1.0, s27
+; SI-NEXT: buffer_store_dword v59, off, s[0:3], s32 offset:500 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v46, off, s[0:3], s32 offset:504 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v61, off, s[0:3], s32 offset:508 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v42, off, s[0:3], s32 offset:512 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v8, off, s[0:3], s32 offset:516 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v14, off, s[0:3], s32 offset:520 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v9, off, s[0:3], s32 offset:524 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v20, off, s[0:3], s32 offset:528 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v12, off, s[0:3], s32 offset:532 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v60, off, s[0:3], s32 offset:536 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v63, off, s[0:3], s32 offset:540 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v45, off, s[0:3], s32 offset:544 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:548 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v3, off, s[0:3], s32 offset:552 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v4, off, s[0:3], s32 offset:556 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v44, off, s[0:3], s32 offset:560 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v11, off, s[0:3], s32 offset:564 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v17, off, s[0:3], s32 offset:568 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v5, off, s[0:3], s32 offset:572 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v6, off, s[0:3], s32 offset:576 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v47, off, s[0:3], s32 offset:580 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v7, off, s[0:3], s32 offset:584 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v21, off, s[0:3], s32 offset:588 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v10, off, s[0:3], s32 offset:592 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v13, off, s[0:3], s32 offset:596 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v57, off, s[0:3], s32 offset:600 ; 4-byte Folded Spill
; SI-NEXT: buffer_store_dword v58, off, s[0:3], s32 offset:604 ; 4-byte Folded Spill
-; SI-NEXT: buffer_store_dword v26, off, s[0:3], s32 offset:608 ; 4-byte Folded Spill
-; SI-NEXT: buffer_store_dword v62, off, s[0:3], s32 offset:612 ; 4-byte Folded Spill
-; SI-NEXT: buffer_store_dword v60, off, s[0:3], s32 offset:616 ; 4-byte Folded Spill
-; SI-NEXT: buffer_store_dword v20, off, s[0:3], s32 offset:620 ; 4-byte Folded Spill
-; SI-NEXT: buffer_store_dword v32, off, s[0:3], s32 offset:624 ; 4-byte Folded Spill
-; SI-NEXT: buffer_store_dword v14, off, s[0:3], s32 offset:628 ; 4-byte Folded Spill
-; SI-NEXT: buffer_store_dword v19, off, s[0:3], s32 offset:632 ; 4-byte Folded Spill
-; SI-NEXT: buffer_store_dword v22, off, s[0:3], s32 offset:636 ; 4-byte Folded Spill
-; SI-NEXT: buffer_store_dword v15, off, s[0:3], s32 offset:640 ; 4-byte Folded Spill
-; SI-NEXT: buffer_store_dword v25, off, s[0:3], s32 offset:644 ; 4-byte Folded Spill
-; SI-NEXT: buffer_store_dword v27, off, s[0:3], s32 offset:648 ; 4-byte Folded Spill
-; SI-NEXT: buffer_store_dword v51, off, s[0:3], s32 offset:456 ; 4-byte Folded Spill
-; SI-NEXT: buffer_store_dword v29, off, s[0:3], s32 offset:652 ; 4-byte Folded Spill
-; SI-NEXT: buffer_store_dword v30, off, s[0:3], s32 offset:656 ; 4-byte Folded Spill
-; SI-NEXT: buffer_store_dword v36, off, s[0:3], s32 offset:460 ; 4-byte Folded Spill
-; SI-NEXT: buffer_store_dword v35, off, s[0:3], s32 offset:660 ; 4-byte Folded Spill
-; SI-NEXT: buffer_store_dword v37, off, s[0:3], s32 offset:464 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v24, off, s[0:3], s32 offset:608 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v26, off, s[0:3], s32 offset:612 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v16, off, s[0:3], s32 offset:616 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v19, off, s[0:3], s32 offset:620 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v62, off, s[0:3], s32 offset:624 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v18, off, s[0:3], s32 offset:628 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v31, off, s[0:3], s32 offset:632 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v2, off, s[0:3], s32 offset:636 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v27, off, s[0:3], s32 offset:640 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v23, off, s[0:3], s32 offset:472 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v30, off, s[0:3], s32 offset:644 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v32, off, s[0:3], s32 offset:648 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v54, off, s[0:3], s32 offset:652 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v35, off, s[0:3], s32 offset:656 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v37, off, s[0:3], s32 offset:660 ; 4-byte Folded Spill
; SI-NEXT: buffer_store_dword v38, off, s[0:3], s32 offset:664 ; 4-byte Folded Spill
; SI-NEXT: buffer_store_dword v48, off, s[0:3], s32 offset:668 ; 4-byte Folded Spill
; SI-NEXT: buffer_store_dword v50, off, s[0:3], s32 offset:672 ; 4-byte Folded Spill
-; SI-NEXT: buffer_store_dword v52, off, s[0:3], s32 offset:676 ; 4-byte Folded Spill
-; SI-NEXT: buffer_store_dword v55, off, s[0:3], s32 offset:680 ; 4-byte Folded Spill
-; SI-NEXT: buffer_store_dword v41, off, s[0:3], s32 offset:684 ; 4-byte Folded Spill
-; SI-NEXT: buffer_store_dword v54, off, s[0:3], s32 offset:688 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v51, off, s[0:3], s32 offset:676 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v52, off, s[0:3], s32 offset:680 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v55, off, s[0:3], s32 offset:684 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v28, off, s[0:3], s32 offset:688 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v22, off, s[0:3], s32 offset:476 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v43, off, s[0:3], s32 offset:692 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v56, off, s[0:3], s32 offset:484 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v15, off, s[0:3], s32 offset:496 ; 4-byte Folded Spill
; SI-NEXT: s_cbranch_scc0 .LBB91_2
; SI-NEXT: ; %bb.1: ; %cmp.false
; SI-NEXT: v_lshrrev_b32_e32 v1, 16, v1
-; SI-NEXT: v_mov_b32_e32 v43, v36
-; SI-NEXT: v_alignbit_b32 v36, v1, v2, 16
+; SI-NEXT: v_alignbit_b32 v36, v1, v3, 16
; SI-NEXT: v_lshrrev_b32_e32 v1, 16, v5
-; SI-NEXT: v_alignbit_b32 v6, v1, v6, 16
+; SI-NEXT: v_alignbit_b32 v33, v1, v6, 16
; SI-NEXT: v_lshrrev_b32_e32 v1, 16, v10
-; SI-NEXT: v_alignbit_b32 v2, v1, v13, 16
+; SI-NEXT: v_mov_b32_e32 v41, v28
+; SI-NEXT: s_waitcnt expcnt(4)
+; SI-NEXT: v_alignbit_b32 v28, v1, v13, 16
; SI-NEXT: v_lshrrev_b32_e32 v1, 16, v16
-; SI-NEXT: v_alignbit_b32 v5, v1, v17, 16
+; SI-NEXT: v_mov_b32_e32 v40, v25
+; SI-NEXT: v_alignbit_b32 v25, v1, v19, 16
; SI-NEXT: v_lshrrev_b32_e32 v1, 16, v4
-; SI-NEXT: v_alignbit_b32 v4, v1, v3, 16
-; SI-NEXT: v_lshrrev_b32_e32 v1, 16, v44
-; SI-NEXT: v_alignbit_b32 v3, v1, v7, 16
+; SI-NEXT: v_alignbit_b32 v53, v1, v44, 16
; SI-NEXT: v_lshrrev_b32_e32 v1, 16, v47
-; SI-NEXT: v_alignbit_b32 v16, v1, v57, 16
-; SI-NEXT: v_lshrrev_b32_e32 v1, 16, v31
-; SI-NEXT: buffer_load_dword v31, off, s[0:3], s32 offset:156 ; 4-byte Folded Reload
-; SI-NEXT: v_alignbit_b32 v13, v1, v58, 16
+; SI-NEXT: v_alignbit_b32 v19, v1, v7, 16
+; SI-NEXT: v_lshrrev_b32_e32 v1, 16, v57
+; SI-NEXT: v_alignbit_b32 v57, v1, v58, 16
; SI-NEXT: v_lshrrev_b32_e32 v1, 16, v62
-; SI-NEXT: v_alignbit_b32 v10, v1, v60, 16
-; SI-NEXT: v_lshrrev_b32_e32 v1, 16, v19
-; SI-NEXT: v_lshrrev_b32_e32 v19, 16, v11
-; SI-NEXT: v_alignbit_b32 v44, v19, v8, 16
-; SI-NEXT: v_alignbit_b32 v7, v1, v22, 16
-; SI-NEXT: v_lshrrev_b32_e32 v1, 16, v25
-; SI-NEXT: v_alignbit_b32 v8, v44, v36, 24
-; SI-NEXT: v_alignbit_b32 v60, v1, v27, 16
-; SI-NEXT: v_lshrrev_b32_e32 v1, 16, v29
-; SI-NEXT: buffer_store_dword v8, off, s[0:3], s32 offset:396 ; 4-byte Folded Spill
+; SI-NEXT: v_alignbit_b32 v13, v1, v56, 16
+; SI-NEXT: v_lshrrev_b32_e32 v1, 16, v31
+; SI-NEXT: v_alignbit_b32 v10, v1, v2, 16
+; SI-NEXT: v_lshrrev_b32_e32 v1, 16, v22
+; SI-NEXT: v_alignbit_b32 v7, v1, v23, 16
+; SI-NEXT: v_lshrrev_b32_e32 v1, 16, v15
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_alignbit_b32 v8, v44, v36, 16
-; SI-NEXT: v_lshrrev_b32_e32 v22, 16, v39
-; SI-NEXT: v_alignbit_b32 v57, v1, v30, 16
-; SI-NEXT: v_lshrrev_b32_e32 v1, 16, v35
-; SI-NEXT: buffer_store_dword v8, off, s[0:3], s32 offset:408 ; 4-byte Folded Spill
+; SI-NEXT: v_mov_b32_e32 v15, v11
+; SI-NEXT: buffer_load_dword v11, off, s[0:3], s32 offset:164 ; 4-byte Folded Reload
+; SI-NEXT: v_lshrrev_b32_e32 v2, 16, v42
+; SI-NEXT: v_alignbit_b32 v44, v2, v8, 16
+; SI-NEXT: buffer_store_dword v2, off, s[0:3], s32 offset:696 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_alignbit_b32 v8, v44, v36, 8
-; SI-NEXT: v_alignbit_b32 v58, v22, v9, 16
-; SI-NEXT: v_alignbit_b32 v40, v1, v37, 16
-; SI-NEXT: buffer_store_dword v8, off, s[0:3], s32 offset:420 ; 4-byte Folded Spill
+; SI-NEXT: v_alignbit_b32 v2, v44, v36, 24
+; SI-NEXT: buffer_store_dword v2, off, s[0:3], s32 offset:436 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_alignbit_b32 v8, v58, v6, 24
-; SI-NEXT: buffer_load_dword v37, off, s[0:3], s32 offset:160 ; 4-byte Folded Reload
-; SI-NEXT: buffer_load_dword v9, off, s[0:3], s32 offset:164 ; 4-byte Folded Reload
-; SI-NEXT: buffer_store_dword v8, off, s[0:3], s32 offset:384 ; 4-byte Folded Spill
+; SI-NEXT: v_alignbit_b32 v2, v44, v36, 16
+; SI-NEXT: buffer_store_dword v2, off, s[0:3], s32 offset:444 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_alignbit_b32 v8, v58, v6, 16
-; SI-NEXT: v_lshrrev_b32_e32 v25, 16, v49
-; SI-NEXT: buffer_store_dword v8, off, s[0:3], s32 offset:392 ; 4-byte Folded Spill
+; SI-NEXT: v_lshrrev_b32_e32 v2, 16, v14
+; SI-NEXT: v_alignbit_b32 v58, v2, v9, 16
+; SI-NEXT: buffer_store_dword v2, off, s[0:3], s32 offset:700 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_alignbit_b32 v8, v58, v6, 8
-; SI-NEXT: v_alignbit_b32 v47, v25, v12, 16
-; SI-NEXT: buffer_store_dword v8, off, s[0:3], s32 offset:400 ; 4-byte Folded Spill
+; SI-NEXT: v_alignbit_b32 v2, v58, v33, 24
+; SI-NEXT: buffer_store_dword v2, off, s[0:3], s32 offset:412 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_alignbit_b32 v8, v47, v2, 24
-; SI-NEXT: buffer_store_dword v8, off, s[0:3], s32 offset:368 ; 4-byte Folded Spill
+; SI-NEXT: v_alignbit_b32 v2, v58, v33, 16
+; SI-NEXT: v_lshrrev_b32_e32 v23, 16, v20
+; SI-NEXT: buffer_store_dword v2, off, s[0:3], s32 offset:420 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_alignbit_b32 v8, v47, v2, 16
-; SI-NEXT: buffer_store_dword v8, off, s[0:3], s32 offset:372 ; 4-byte Folded Spill
+; SI-NEXT: v_alignbit_b32 v2, v58, v33, 8
+; SI-NEXT: v_alignbit_b32 v47, v23, v12, 16
+; SI-NEXT: buffer_store_dword v2, off, s[0:3], s32 offset:432 ; 4-byte Folded Spill
+; SI-NEXT: s_waitcnt expcnt(0)
+; SI-NEXT: v_alignbit_b32 v2, v47, v28, 24
+; SI-NEXT: v_alignbit_b32 v6, v1, v27, 16
+; SI-NEXT: v_lshrrev_b32_e32 v1, 16, v30
+; SI-NEXT: buffer_store_dword v2, off, s[0:3], s32 offset:396 ; 4-byte Folded Spill
+; SI-NEXT: s_waitcnt expcnt(0)
+; SI-NEXT: v_alignbit_b32 v2, v47, v28, 16
+; SI-NEXT: v_alignbit_b32 v5, v1, v32, 16
+; SI-NEXT: v_lshrrev_b32_e32 v1, 16, v35
+; SI-NEXT: buffer_store_dword v2, off, s[0:3], s32 offset:400 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_alignbit_b32 v8, v47, v2, 8
+; SI-NEXT: v_alignbit_b32 v2, v47, v28, 8
+; SI-NEXT: v_alignbit_b32 v4, v1, v37, 16
; SI-NEXT: v_lshrrev_b32_e32 v1, 16, v38
-; SI-NEXT: buffer_store_dword v8, off, s[0:3], s32 offset:380 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v2, off, s[0:3], s32 offset:408 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_lshrrev_b32_e32 v8, 16, v46
-; SI-NEXT: v_alignbit_b32 v53, v1, v48, 16
+; SI-NEXT: v_lshrrev_b32_e32 v2, 16, v59
+; SI-NEXT: v_alignbit_b32 v3, v1, v48, 16
; SI-NEXT: v_lshrrev_b32_e32 v1, 16, v50
-; SI-NEXT: v_alignbit_b32 v50, v8, v59, 16
-; SI-NEXT: buffer_load_dword v59, off, s[0:3], s32 offset:468 ; 4-byte Folded Reload
-; SI-NEXT: buffer_store_dword v8, off, s[0:3], s32 offset:476 ; 4-byte Folded Spill
+; SI-NEXT: v_alignbit_b32 v50, v2, v46, 16
+; SI-NEXT: buffer_store_dword v2, off, s[0:3], s32 offset:468 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_alignbit_b32 v8, v50, v5, 24
-; SI-NEXT: buffer_store_dword v8, off, s[0:3], s32 offset:344 ; 4-byte Folded Spill
+; SI-NEXT: v_alignbit_b32 v2, v50, v25, 24
+; SI-NEXT: buffer_store_dword v2, off, s[0:3], s32 offset:372 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_alignbit_b32 v8, v50, v5, 16
-; SI-NEXT: buffer_store_dword v8, off, s[0:3], s32 offset:356 ; 4-byte Folded Spill
+; SI-NEXT: v_alignbit_b32 v2, v50, v25, 16
+; SI-NEXT: buffer_store_dword v2, off, s[0:3], s32 offset:380 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_alignbit_b32 v8, v50, v5, 8
-; SI-NEXT: buffer_store_dword v8, off, s[0:3], s32 offset:364 ; 4-byte Folded Spill
-; SI-NEXT: v_alignbit_b32 v52, v1, v52, 16
-; SI-NEXT: v_lshrrev_b32_e32 v1, 16, v55
-; SI-NEXT: v_mov_b32_e32 v17, v63
-; SI-NEXT: v_alignbit_b32 v1, v1, v41, 16
+; SI-NEXT: v_alignbit_b32 v2, v50, v25, 8
+; SI-NEXT: buffer_store_dword v2, off, s[0:3], s32 offset:392 ; 4-byte Folded Spill
+; SI-NEXT: buffer_load_dword v9, off, s[0:3], s32 offset:172 ; 4-byte Folded Reload
+; SI-NEXT: v_alignbit_b32 v56, v1, v51, 16
+; SI-NEXT: v_lshrrev_b32_e32 v1, 16, v52
+; SI-NEXT: v_alignbit_b32 v1, v1, v55, 16
+; SI-NEXT: v_lshrrev_b32_e32 v51, 16, v45
+; SI-NEXT: v_lshrrev_b32_e32 v22, 16, v15
+; SI-NEXT: v_alignbit_b32 v35, v22, v17, 16
+; SI-NEXT: v_alignbit_b32 v8, v35, v13, 24
+; SI-NEXT: buffer_store_dword v8, off, s[0:3], s32 offset:300 ; 4-byte Folded Spill
+; SI-NEXT: s_waitcnt expcnt(0)
+; SI-NEXT: v_alignbit_b32 v8, v35, v13, 16
+; SI-NEXT: buffer_store_dword v8, off, s[0:3], s32 offset:304 ; 4-byte Folded Spill
+; SI-NEXT: v_mov_b32_e32 v17, v40
+; SI-NEXT: v_mov_b32_e32 v40, v29
+; SI-NEXT: v_mov_b32_e32 v32, v49
; SI-NEXT: s_mov_b64 s[4:5], 0
-; SI-NEXT: s_waitcnt vmcnt(14) expcnt(0)
-; SI-NEXT: v_lshrrev_b32_e32 v8, 16, v31
-; SI-NEXT: v_alignbit_b32 v62, v8, v61, 16
-; SI-NEXT: buffer_store_dword v8, off, s[0:3], s32 offset:416 ; 4-byte Folded Spill
+; SI-NEXT: v_alignbit_b32 v52, v44, v36, 8
+; SI-NEXT: v_alignbit_b32 v46, v35, v13, 8
+; SI-NEXT: v_mov_b32_e32 v48, v34
+; SI-NEXT: buffer_load_dword v31, off, s[0:3], s32 offset:492 ; 4-byte Folded Reload
+; SI-NEXT: s_waitcnt vmcnt(14)
+; SI-NEXT: v_lshrrev_b32_e32 v2, 16, v11
+; SI-NEXT: v_alignbit_b32 v62, v2, v61, 16
+; SI-NEXT: buffer_store_dword v2, off, s[0:3], s32 offset:452 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_alignbit_b32 v8, v62, v4, 24
-; SI-NEXT: buffer_store_dword v8, off, s[0:3], s32 offset:328 ; 4-byte Folded Spill
+; SI-NEXT: v_alignbit_b32 v2, v62, v53, 24
+; SI-NEXT: buffer_store_dword v2, off, s[0:3], s32 offset:356 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_alignbit_b32 v8, v62, v4, 16
-; SI-NEXT: buffer_store_dword v8, off, s[0:3], s32 offset:336 ; 4-byte Folded Spill
+; SI-NEXT: v_alignbit_b32 v2, v62, v53, 16
+; SI-NEXT: buffer_store_dword v2, off, s[0:3], s32 offset:364 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_lshrrev_b32_e32 v8, 16, v56
-; SI-NEXT: v_alignbit_b32 v55, v8, v63, 16
-; SI-NEXT: buffer_store_dword v8, off, s[0:3], s32 offset:472 ; 4-byte Folded Spill
+; SI-NEXT: v_alignbit_b32 v2, v62, v53, 8
+; SI-NEXT: buffer_store_dword v2, off, s[0:3], s32 offset:368 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_alignbit_b32 v8, v55, v3, 24
-; SI-NEXT: buffer_store_dword v8, off, s[0:3], s32 offset:312 ; 4-byte Folded Spill
+; SI-NEXT: v_lshrrev_b32_e32 v2, 16, v60
+; SI-NEXT: v_alignbit_b32 v55, v2, v63, 16
+; SI-NEXT: buffer_store_dword v2, off, s[0:3], s32 offset:320 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_alignbit_b32 v8, v55, v3, 16
-; SI-NEXT: buffer_store_dword v8, off, s[0:3], s32 offset:316 ; 4-byte Folded Spill
+; SI-NEXT: v_alignbit_b32 v2, v55, v19, 24
+; SI-NEXT: buffer_store_dword v2, off, s[0:3], s32 offset:340 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_alignbit_b32 v8, v55, v3, 8
-; SI-NEXT: buffer_store_dword v8, off, s[0:3], s32 offset:324 ; 4-byte Folded Spill
-; SI-NEXT: v_alignbit_b32 v48, v62, v4, 8
-; SI-NEXT: v_lshrrev_b32_e32 v31, 24, v31
-; SI-NEXT: s_waitcnt vmcnt(14) expcnt(0)
-; SI-NEXT: v_lshrrev_b32_e32 v8, 16, v37
-; SI-NEXT: v_alignbit_b32 v38, v8, v45, 16
-; SI-NEXT: buffer_store_dword v8, off, s[0:3], s32 offset:412 ; 4-byte Folded Spill
+; SI-NEXT: v_alignbit_b32 v2, v55, v19, 16
+; SI-NEXT: buffer_store_dword v2, off, s[0:3], s32 offset:344 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_alignbit_b32 v8, v38, v16, 8
-; SI-NEXT: buffer_store_dword v8, off, s[0:3], s32 offset:480 ; 4-byte Folded Spill
+; SI-NEXT: v_alignbit_b32 v2, v55, v19, 8
+; SI-NEXT: buffer_store_dword v2, off, s[0:3], s32 offset:352 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_lshrrev_b32_e32 v8, 16, v9
-; SI-NEXT: v_alignbit_b32 v35, v8, v18, 16
-; SI-NEXT: v_mov_b32_e32 v45, v8
-; SI-NEXT: v_alignbit_b32 v8, v35, v13, 24
-; SI-NEXT: buffer_store_dword v8, off, s[0:3], s32 offset:292 ; 4-byte Folded Spill
+; SI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:168 ; 4-byte Folded Reload
+; SI-NEXT: v_lshrrev_b32_e32 v11, 24, v11
+; SI-NEXT: buffer_store_dword v11, off, s[0:3], s32 offset:428 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_alignbit_b32 v8, v35, v13, 16
-; SI-NEXT: buffer_store_dword v8, off, s[0:3], s32 offset:296 ; 4-byte Folded Spill
-; SI-NEXT: v_alignbit_b32 v29, v35, v13, 8
-; SI-NEXT: v_alignbit_b32 v61, v38, v16, 24
-; SI-NEXT: v_alignbit_b32 v41, v38, v16, 16
-; SI-NEXT: s_waitcnt vmcnt(14) expcnt(0)
-; SI-NEXT: v_lshrrev_b32_e32 v8, 16, v59
+; SI-NEXT: v_lshrrev_b32_e32 v11, 8, v62
+; SI-NEXT: s_waitcnt vmcnt(13)
+; SI-NEXT: v_lshrrev_b32_e32 v8, 16, v9
; SI-NEXT: v_alignbit_b32 v30, v8, v21, 16
-; SI-NEXT: buffer_store_dword v8, off, s[0:3], s32 offset:340 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v8, off, s[0:3], s32 offset:440 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_alignbit_b32 v8, v30, v10, 24
-; SI-NEXT: buffer_store_dword v8, off, s[0:3], s32 offset:276 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v8, off, s[0:3], s32 offset:284 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_alignbit_b32 v8, v30, v10, 16
-; SI-NEXT: buffer_store_dword v8, off, s[0:3], s32 offset:284 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v8, off, s[0:3], s32 offset:292 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_alignbit_b32 v8, v30, v10, 8
-; SI-NEXT: buffer_store_dword v8, off, s[0:3], s32 offset:288 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v8, off, s[0:3], s32 offset:296 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_lshrrev_b32_e32 v8, 16, v23
-; SI-NEXT: v_alignbit_b32 v27, v8, v24, 16
-; SI-NEXT: buffer_store_dword v8, off, s[0:3], s32 offset:352 ; 4-byte Folded Spill
+; SI-NEXT: v_lshrrev_b32_e32 v8, 16, v24
+; SI-NEXT: v_alignbit_b32 v27, v8, v26, 16
+; SI-NEXT: buffer_store_dword v8, off, s[0:3], s32 offset:388 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_alignbit_b32 v8, v27, v7, 24
-; SI-NEXT: buffer_store_dword v8, off, s[0:3], s32 offset:256 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v8, off, s[0:3], s32 offset:264 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_alignbit_b32 v8, v27, v7, 16
-; SI-NEXT: buffer_store_dword v8, off, s[0:3], s32 offset:264 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v8, off, s[0:3], s32 offset:272 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_alignbit_b32 v8, v27, v7, 8
-; SI-NEXT: buffer_store_dword v8, off, s[0:3], s32 offset:268 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v8, off, s[0:3], s32 offset:276 ; 4-byte Folded Spill
+; SI-NEXT: v_lshrrev_b32_e32 v26, 8, v44
+; SI-NEXT: s_waitcnt vmcnt(9)
+; SI-NEXT: v_alignbit_b32 v38, v51, v2, 16
+; SI-NEXT: v_alignbit_b32 v2, v38, v57, 24
+; SI-NEXT: buffer_store_dword v2, off, s[0:3], s32 offset:324 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_lshrrev_b32_e32 v8, 16, v20
-; SI-NEXT: v_alignbit_b32 v24, v8, v26, 16
-; SI-NEXT: buffer_store_dword v8, off, s[0:3], s32 offset:252 ; 4-byte Folded Spill
+; SI-NEXT: v_alignbit_b32 v2, v38, v57, 16
+; SI-NEXT: buffer_store_dword v2, off, s[0:3], s32 offset:328 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_alignbit_b32 v8, v24, v60, 24
-; SI-NEXT: buffer_store_dword v8, off, s[0:3], s32 offset:236 ; 4-byte Folded Spill
+; SI-NEXT: v_alignbit_b32 v2, v38, v57, 8
+; SI-NEXT: buffer_store_dword v2, off, s[0:3], s32 offset:332 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_alignbit_b32 v8, v24, v60, 8
-; SI-NEXT: buffer_store_dword v8, off, s[0:3], s32 offset:260 ; 4-byte Folded Spill
+; SI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:480 ; 4-byte Folded Reload
+; SI-NEXT: s_waitcnt vmcnt(0)
+; SI-NEXT: v_lshrrev_b32_e32 v8, 16, v2
+; SI-NEXT: v_alignbit_b32 v37, v8, v18, 16
+; SI-NEXT: buffer_store_dword v8, off, s[0:3], s32 offset:316 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_lshrrev_b32_e32 v8, 16, v32
-; SI-NEXT: v_alignbit_b32 v21, v8, v14, 16
+; SI-NEXT: v_alignbit_b32 v8, v37, v6, 24
; SI-NEXT: buffer_store_dword v8, off, s[0:3], s32 offset:248 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_alignbit_b32 v8, v21, v57, 24
-; SI-NEXT: buffer_store_dword v8, off, s[0:3], s32 offset:220 ; 4-byte Folded Spill
+; SI-NEXT: v_alignbit_b32 v8, v37, v6, 16
+; SI-NEXT: buffer_store_dword v8, off, s[0:3], s32 offset:256 ; 4-byte Folded Spill
+; SI-NEXT: s_waitcnt expcnt(0)
+; SI-NEXT: v_alignbit_b32 v8, v37, v6, 8
+; SI-NEXT: buffer_store_dword v8, off, s[0:3], s32 offset:268 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_alignbit_b32 v8, v21, v57, 16
+; SI-NEXT: v_lshrrev_b32_e32 v8, 16, v17
+; SI-NEXT: v_alignbit_b32 v21, v8, v29, 16
+; SI-NEXT: buffer_load_dword v29, off, s[0:3], s32 offset:488 ; 4-byte Folded Reload
+; SI-NEXT: buffer_store_dword v8, off, s[0:3], s32 offset:160 ; 4-byte Folded Spill
+; SI-NEXT: s_waitcnt expcnt(0)
+; SI-NEXT: v_alignbit_b32 v8, v21, v5, 24
; SI-NEXT: buffer_store_dword v8, off, s[0:3], s32 offset:232 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_alignbit_b32 v8, v21, v57, 8
-; SI-NEXT: buffer_store_dword v8, off, s[0:3], s32 offset:240 ; 4-byte Folded Spill
+; SI-NEXT: v_alignbit_b32 v8, v21, v5, 16
+; SI-NEXT: buffer_store_dword v8, off, s[0:3], s32 offset:244 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_lshrrev_b32_e32 v8, 16, v34
-; SI-NEXT: v_alignbit_b32 v18, v8, v15, 16
-; SI-NEXT: buffer_store_dword v8, off, s[0:3], s32 offset:188 ; 4-byte Folded Spill
+; SI-NEXT: v_alignbit_b32 v8, v21, v5, 8
+; SI-NEXT: buffer_store_dword v8, off, s[0:3], s32 offset:252 ; 4-byte Folded Spill
+; SI-NEXT: v_mov_b32_e32 v18, v9
+; SI-NEXT: s_waitcnt vmcnt(4) expcnt(0)
+; SI-NEXT: v_lshrrev_b32_e32 v8, 16, v29
+; SI-NEXT: v_alignbit_b32 v61, v8, v34, 16
+; SI-NEXT: buffer_store_dword v8, off, s[0:3], s32 offset:156 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_alignbit_b32 v8, v18, v40, 24
-; SI-NEXT: buffer_store_dword v8, off, s[0:3], s32 offset:208 ; 4-byte Folded Spill
+; SI-NEXT: v_alignbit_b32 v8, v61, v4, 24
+; SI-NEXT: buffer_store_dword v8, off, s[0:3], s32 offset:220 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_alignbit_b32 v8, v18, v40, 16
-; SI-NEXT: buffer_store_dword v8, off, s[0:3], s32 offset:216 ; 4-byte Folded Spill
-; SI-NEXT: buffer_store_dword v40, off, s[0:3], s32 offset:440 ; 4-byte Folded Spill
-; SI-NEXT: s_waitcnt expcnt(1)
-; SI-NEXT: v_alignbit_b32 v8, v18, v40, 8
+; SI-NEXT: v_alignbit_b32 v8, v61, v4, 16
; SI-NEXT: buffer_store_dword v8, off, s[0:3], s32 offset:228 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_lshrrev_b32_e32 v8, 16, v28
-; SI-NEXT: v_alignbit_b32 v63, v8, v51, 16
-; SI-NEXT: buffer_store_dword v8, off, s[0:3], s32 offset:152 ; 4-byte Folded Spill
+; SI-NEXT: v_alignbit_b32 v8, v61, v4, 8
+; SI-NEXT: buffer_store_dword v8, off, s[0:3], s32 offset:240 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_alignbit_b32 v8, v63, v53, 24
-; SI-NEXT: buffer_store_dword v8, off, s[0:3], s32 offset:192 ; 4-byte Folded Spill
+; SI-NEXT: v_lshrrev_b32_e32 v8, 16, v41
+; SI-NEXT: v_alignbit_b32 v63, v8, v49, 16
+; SI-NEXT: buffer_store_dword v8, off, s[0:3], s32 offset:152 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_alignbit_b32 v8, v63, v53, 16
-; SI-NEXT: v_lshrrev_b32_e32 v40, 16, v33
+; SI-NEXT: v_alignbit_b32 v8, v63, v3, 24
; SI-NEXT: buffer_store_dword v8, off, s[0:3], s32 offset:204 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_alignbit_b32 v8, v63, v53, 8
-; SI-NEXT: v_alignbit_b32 v12, v40, v43, 16
-; SI-NEXT: buffer_store_dword v8, off, s[0:3], s32 offset:212 ; 4-byte Folded Spill
+; SI-NEXT: v_alignbit_b32 v8, v63, v3, 16
+; SI-NEXT: v_lshrrev_b32_e32 v49, 16, v31
+; SI-NEXT: buffer_store_dword v8, off, s[0:3], s32 offset:216 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_alignbit_b32 v8, v12, v52, 24
-; SI-NEXT: buffer_store_dword v8, off, s[0:3], s32 offset:244 ; 4-byte Folded Spill
+; SI-NEXT: v_alignbit_b32 v8, v63, v3, 8
+; SI-NEXT: v_alignbit_b32 v12, v49, v54, 16
+; SI-NEXT: buffer_store_dword v8, off, s[0:3], s32 offset:224 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_alignbit_b32 v8, v12, v52, 16
-; SI-NEXT: buffer_store_dword v8, off, s[0:3], s32 offset:436 ; 4-byte Folded Spill
+; SI-NEXT: v_alignbit_b32 v8, v12, v56, 24
+; SI-NEXT: buffer_store_dword v8, off, s[0:3], s32 offset:192 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_alignbit_b32 v8, v12, v52, 8
+; SI-NEXT: v_alignbit_b32 v8, v12, v56, 16
; SI-NEXT: buffer_store_dword v8, off, s[0:3], s32 offset:200 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_lshrrev_b32_e32 v8, 16, v42
-; SI-NEXT: v_mov_b32_e32 v15, v9
-; SI-NEXT: v_alignbit_b32 v9, v8, v54, 16
-; SI-NEXT: buffer_store_dword v8, off, s[0:3], s32 offset:148 ; 4-byte Folded Spill
+; SI-NEXT: v_alignbit_b32 v8, v12, v56, 8
+; SI-NEXT: buffer_store_dword v8, off, s[0:3], s32 offset:212 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_alignbit_b32 v8, v9, v1, 24
-; SI-NEXT: buffer_store_dword v8, off, s[0:3], s32 offset:176 ; 4-byte Folded Spill
+; SI-NEXT: v_lshrrev_b32_e32 v8, 16, v39
+; SI-NEXT: v_alignbit_b32 v9, v8, v43, 16
+; SI-NEXT: v_mov_b32_e32 v43, v17
+; SI-NEXT: v_lshrrev_b32_e32 v17, 24, v24
+; SI-NEXT: buffer_store_dword v17, off, s[0:3], s32 offset:280 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_alignbit_b32 v8, v9, v1, 16
-; SI-NEXT: buffer_store_dword v8, off, s[0:3], s32 offset:184 ; 4-byte Folded Spill
+; SI-NEXT: v_lshrrev_b32_e32 v17, 24, v2
+; SI-NEXT: buffer_store_dword v17, off, s[0:3], s32 offset:260 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_alignbit_b32 v8, v9, v1, 8
-; SI-NEXT: buffer_store_dword v8, off, s[0:3], s32 offset:308 ; 4-byte Folded Spill
+; SI-NEXT: v_lshrrev_b32_e32 v17, 8, v37
+; SI-NEXT: buffer_store_dword v17, off, s[0:3], s32 offset:416 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_mov_b32_e32 v8, v37
-; SI-NEXT: v_lshrrev_b32_e32 v37, 24, v49
-; SI-NEXT: buffer_store_dword v37, off, s[0:3], s32 offset:432 ; 4-byte Folded Spill
+; SI-NEXT: v_lshrrev_b32_e32 v17, 24, v43
+; SI-NEXT: buffer_store_dword v8, off, s[0:3], s32 offset:148 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_lshrrev_b32_e32 v37, 24, v46
-; SI-NEXT: v_lshrrev_b32_e32 v46, 24, v56
-; SI-NEXT: v_lshrrev_b32_e32 v56, 24, v32
-; SI-NEXT: v_lshrrev_b32_e32 v8, 24, v8
-; SI-NEXT: buffer_store_dword v56, off, s[0:3], s32 offset:224 ; 4-byte Folded Spill
+; SI-NEXT: v_alignbit_b32 v8, v9, v1, 24
+; SI-NEXT: buffer_store_dword v17, off, s[0:3], s32 offset:236 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_lshrrev_b32_e32 v56, 24, v34
-; SI-NEXT: buffer_store_dword v8, off, s[0:3], s32 offset:348 ; 4-byte Folded Spill
+; SI-NEXT: v_lshrrev_b32_e32 v17, 8, v21
+; SI-NEXT: buffer_store_dword v8, off, s[0:3], s32 offset:180 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_lshrrev_b32_e32 v8, 8, v38
-; SI-NEXT: buffer_store_dword v56, off, s[0:3], s32 offset:196 ; 4-byte Folded Spill
+; SI-NEXT: v_alignbit_b32 v8, v9, v1, 16
+; SI-NEXT: buffer_store_dword v17, off, s[0:3], s32 offset:384 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_lshrrev_b32_e32 v56, 24, v28
-; SI-NEXT: buffer_store_dword v8, off, s[0:3], s32 offset:444 ; 4-byte Folded Spill
+; SI-NEXT: v_lshrrev_b32_e32 v17, 24, v29
+; SI-NEXT: buffer_store_dword v8, off, s[0:3], s32 offset:188 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_lshrrev_b32_e32 v8, 24, v15
-; SI-NEXT: buffer_store_dword v56, off, s[0:3], s32 offset:180 ; 4-byte Folded Spill
+; SI-NEXT: v_alignbit_b32 v8, v9, v1, 8
+; SI-NEXT: buffer_store_dword v17, off, s[0:3], s32 offset:208 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_lshrrev_b32_e32 v56, 24, v33
-; SI-NEXT: buffer_store_dword v8, off, s[0:3], s32 offset:320 ; 4-byte Folded Spill
+; SI-NEXT: v_lshrrev_b32_e32 v17, 8, v61
+; SI-NEXT: buffer_store_dword v8, off, s[0:3], s32 offset:196 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_lshrrev_b32_e32 v8, 24, v59
-; SI-NEXT: v_lshrrev_b32_e32 v20, 24, v20
-; SI-NEXT: buffer_store_dword v56, off, s[0:3], s32 offset:172 ; 4-byte Folded Spill
+; SI-NEXT: v_lshrrev_b32_e32 v8, 24, v42
+; SI-NEXT: buffer_store_dword v17, off, s[0:3], s32 offset:360 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_lshrrev_b32_e32 v56, 8, v12
-; SI-NEXT: buffer_store_dword v8, off, s[0:3], s32 offset:304 ; 4-byte Folded Spill
+; SI-NEXT: v_lshrrev_b32_e32 v17, 24, v41
+; SI-NEXT: buffer_store_dword v8, off, s[0:3], s32 offset:464 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_lshrrev_b32_e32 v8, 8, v30
-; SI-NEXT: buffer_store_dword v20, off, s[0:3], s32 offset:452 ; 4-byte Folded Spill
+; SI-NEXT: v_lshrrev_b32_e32 v8, 24, v14
+; SI-NEXT: v_lshrrev_b32_e32 v14, 24, v59
+; SI-NEXT: buffer_store_dword v17, off, s[0:3], s32 offset:184 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_lshrrev_b32_e32 v20, 8, v24
-; SI-NEXT: buffer_store_dword v56, off, s[0:3], s32 offset:300 ; 4-byte Folded Spill
+; SI-NEXT: v_lshrrev_b32_e32 v17, 8, v63
+; SI-NEXT: buffer_store_dword v14, off, s[0:3], s32 offset:448 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_lshrrev_b32_e32 v56, 24, v42
-; SI-NEXT: v_lshrrev_b32_e32 v14, 24, v11
-; SI-NEXT: v_lshrrev_b32_e32 v11, 24, v39
-; SI-NEXT: buffer_store_dword v8, off, s[0:3], s32 offset:428 ; 4-byte Folded Spill
+; SI-NEXT: v_lshrrev_b32_e32 v14, 24, v60
+; SI-NEXT: buffer_store_dword v17, off, s[0:3], s32 offset:336 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_lshrrev_b32_e32 v8, 24, v23
-; SI-NEXT: buffer_store_dword v20, off, s[0:3], s32 offset:388 ; 4-byte Folded Spill
+; SI-NEXT: v_lshrrev_b32_e32 v17, 24, v31
+; SI-NEXT: buffer_store_dword v14, off, s[0:3], s32 offset:404 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_mov_b32_e32 v20, v29
-; SI-NEXT: v_lshrrev_b32_e32 v29, 8, v21
-; SI-NEXT: v_lshrrev_b32_e32 v32, 8, v18
-; SI-NEXT: buffer_store_dword v56, off, s[0:3], s32 offset:168 ; 4-byte Folded Spill
+; SI-NEXT: v_lshrrev_b32_e32 v14, 24, v45
+; SI-NEXT: buffer_store_dword v17, off, s[0:3], s32 offset:176 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_lshrrev_b32_e32 v56, 8, v9
-; SI-NEXT: v_alignbit_b32 v26, v24, v60, 16
-; SI-NEXT: v_lshrrev_b32_e32 v51, 8, v44
-; SI-NEXT: buffer_store_dword v11, off, s[0:3], s32 offset:448 ; 4-byte Folded Spill
+; SI-NEXT: v_lshrrev_b32_e32 v17, 8, v12
+; SI-NEXT: buffer_store_dword v14, off, s[0:3], s32 offset:376 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_lshrrev_b32_e32 v11, 8, v58
-; SI-NEXT: v_lshrrev_b32_e32 v49, 8, v47
-; SI-NEXT: buffer_store_dword v37, off, s[0:3], s32 offset:424 ; 4-byte Folded Spill
-; SI-NEXT: v_lshrrev_b32_e32 v39, 8, v50
-; SI-NEXT: buffer_store_dword v31, off, s[0:3], s32 offset:404 ; 4-byte Folded Spill
+; SI-NEXT: v_lshrrev_b32_e32 v14, 24, v15
+; SI-NEXT: buffer_store_dword v17, off, s[0:3], s32 offset:308 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_lshrrev_b32_e32 v31, 8, v62
-; SI-NEXT: buffer_store_dword v46, off, s[0:3], s32 offset:376 ; 4-byte Folded Spill
+; SI-NEXT: v_lshrrev_b32_e32 v17, 24, v39
+; SI-NEXT: buffer_store_dword v8, off, s[0:3], s32 offset:460 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_lshrrev_b32_e32 v46, 8, v55
-; SI-NEXT: v_lshrrev_b32_e32 v15, 8, v35
-; SI-NEXT: buffer_store_dword v8, off, s[0:3], s32 offset:272 ; 4-byte Folded Spill
+; SI-NEXT: v_lshrrev_b32_e32 v8, 24, v20
+; SI-NEXT: buffer_store_dword v14, off, s[0:3], s32 offset:348 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_lshrrev_b32_e32 v8, 8, v27
-; SI-NEXT: buffer_store_dword v29, off, s[0:3], s32 offset:360 ; 4-byte Folded Spill
-; SI-NEXT: v_mov_b32_e32 v37, v34
-; SI-NEXT: buffer_store_dword v32, off, s[0:3], s32 offset:332 ; 4-byte Folded Spill
-; SI-NEXT: s_waitcnt expcnt(1)
-; SI-NEXT: v_mov_b32_e32 v29, v28
-; SI-NEXT: v_mov_b32_e32 v23, v48
+; SI-NEXT: v_lshrrev_b32_e32 v14, 24, v18
+; SI-NEXT: buffer_store_dword v17, off, s[0:3], s32 offset:424 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_lshrrev_b32_e32 v32, 8, v63
-; SI-NEXT: v_mov_b32_e32 v48, v33
-; SI-NEXT: v_mov_b32_e32 v34, v53
-; SI-NEXT: v_mov_b32_e32 v53, v42
-; SI-NEXT: buffer_store_dword v56, off, s[0:3], s32 offset:280 ; 4-byte Folded Spill
+; SI-NEXT: v_lshrrev_b32_e32 v17, 8, v9
+; SI-NEXT: buffer_store_dword v8, off, s[0:3], s32 offset:456 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v14, off, s[0:3], s32 offset:312 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v17, off, s[0:3], s32 offset:288 ; 4-byte Folded Spill
+; SI-NEXT: v_mov_b32_e32 v24, v51
+; SI-NEXT: v_mov_b32_e32 v51, v2
+; SI-NEXT: v_mov_b32_e32 v2, v22
+; SI-NEXT: v_mov_b32_e32 v22, v23
+; SI-NEXT: buffer_load_dword v23, off, s[0:3], s32 offset:700 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v16, off, s[0:3], s32 offset:696 ; 4-byte Folded Reload
+; SI-NEXT: v_lshrrev_b32_e32 v34, 8, v58
+; SI-NEXT: s_waitcnt expcnt(2)
+; SI-NEXT: v_lshrrev_b32_e32 v8, 8, v47
+; SI-NEXT: v_lshrrev_b32_e32 v59, 8, v50
+; SI-NEXT: v_lshrrev_b32_e32 v20, 8, v55
+; SI-NEXT: v_lshrrev_b32_e32 v60, 8, v38
+; SI-NEXT: v_lshrrev_b32_e32 v45, 8, v35
+; SI-NEXT: s_waitcnt expcnt(1)
+; SI-NEXT: v_lshrrev_b32_e32 v14, 8, v30
+; SI-NEXT: v_lshrrev_b32_e32 v18, 8, v27
+; SI-NEXT: v_mov_b32_e32 v15, v43
+; SI-NEXT: v_mov_b32_e32 v43, v39
; SI-NEXT: s_branch .LBB91_3
; SI-NEXT: .LBB91_2:
-; SI-NEXT: buffer_load_dword v59, off, s[0:3], s32 offset:468 ; 4-byte Folded Reload
+; SI-NEXT: v_mov_b32_e32 v40, v29
+; SI-NEXT: buffer_load_dword v51, off, s[0:3], s32 offset:480 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v29, off, s[0:3], s32 offset:488 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v31, off, s[0:3], s32 offset:492 ; 4-byte Folded Reload
+; SI-NEXT: ; implicit-def: $vgpr1
+; SI-NEXT: ; kill: killed $vgpr1
; SI-NEXT: ; implicit-def: $vgpr1
; SI-NEXT: ; kill: killed $vgpr1
; SI-NEXT: ; implicit-def: $vgpr1
@@ -170806,139 +170917,116 @@ define inreg <128 x i8> @bitcast_v64bf16_to_v128i8_scalar(<64 x bfloat> inreg %a
; SI-NEXT: ; implicit-def: $vgpr1
; SI-NEXT: ; kill: killed $vgpr9
; SI-NEXT: ; implicit-def: $vgpr9
-; SI-NEXT: ; implicit-def: $vgpr40
-; SI-NEXT: v_mov_b32_e32 v53, v42
-; SI-NEXT: s_waitcnt expcnt(5)
-; SI-NEXT: v_mov_b32_e32 v48, v33
-; SI-NEXT: v_mov_b32_e32 v29, v28
-; SI-NEXT: v_mov_b32_e32 v37, v34
-; SI-NEXT: v_mov_b32_e32 v17, v63
+; SI-NEXT: s_waitcnt expcnt(2)
+; SI-NEXT: v_mov_b32_e32 v43, v39
+; SI-NEXT: v_mov_b32_e32 v32, v49
+; SI-NEXT: v_mov_b32_e32 v48, v34
+; SI-NEXT: s_waitcnt expcnt(0)
+; SI-NEXT: v_mov_b32_e32 v15, v25
; SI-NEXT: s_mov_b64 s[4:5], -1
; SI-NEXT: ; kill: killed $vgpr1
; SI-NEXT: ; implicit-def: $vgpr1
; SI-NEXT: ; kill: killed $vgpr9
; SI-NEXT: ; implicit-def: $vgpr9
-; SI-NEXT: ; kill: killed $vgpr40
-; SI-NEXT: ; implicit-def: $vgpr40
+; SI-NEXT: ; implicit-def: $vgpr17
+; SI-NEXT: ; implicit-def: $vgpr49
; SI-NEXT: ; implicit-def: $vgpr36
+; SI-NEXT: ; implicit-def: $vgpr52
; SI-NEXT: ; implicit-def: $vgpr44
-; SI-NEXT: ; implicit-def: $vgpr51
-; SI-NEXT: ; implicit-def: $vgpr19
-; SI-NEXT: ; implicit-def: $vgpr14
-; SI-NEXT: ; implicit-def: $vgpr6
+; SI-NEXT: ; implicit-def: $vgpr26
+; SI-NEXT: ; implicit-def: $vgpr16
+; SI-NEXT: ; implicit-def: $vgpr33
; SI-NEXT: ; implicit-def: $vgpr58
-; SI-NEXT: ; implicit-def: $vgpr11
-; SI-NEXT: ; implicit-def: $vgpr22
-; SI-NEXT: ; implicit-def: $vgpr2
+; SI-NEXT: ; implicit-def: $vgpr34
+; SI-NEXT: ; implicit-def: $vgpr23
+; SI-NEXT: ; implicit-def: $vgpr28
; SI-NEXT: ; implicit-def: $vgpr47
-; SI-NEXT: ; implicit-def: $vgpr49
+; SI-NEXT: ; implicit-def: $vgpr8
+; SI-NEXT: ; implicit-def: $vgpr22
; SI-NEXT: ; implicit-def: $vgpr25
-; SI-NEXT: ; implicit-def: $vgpr5
; SI-NEXT: ; implicit-def: $vgpr50
-; SI-NEXT: ; implicit-def: $vgpr39
-; SI-NEXT: ; implicit-def: $vgpr4
-; SI-NEXT: ; implicit-def: $vgpr23
+; SI-NEXT: ; implicit-def: $vgpr59
+; SI-NEXT: ; implicit-def: $vgpr53
; SI-NEXT: ; implicit-def: $vgpr62
-; SI-NEXT: ; implicit-def: $vgpr31
-; SI-NEXT: ; implicit-def: $vgpr3
+; SI-NEXT: ; implicit-def: $vgpr11
+; SI-NEXT: ; implicit-def: $vgpr19
; SI-NEXT: ; implicit-def: $vgpr55
-; SI-NEXT: ; implicit-def: $vgpr46
-; SI-NEXT: ; implicit-def: $vgpr16
-; SI-NEXT: ; implicit-def: $vgpr41
-; SI-NEXT: ; implicit-def: $vgpr61
+; SI-NEXT: ; implicit-def: $vgpr20
+; SI-NEXT: ; implicit-def: $vgpr57
; SI-NEXT: ; implicit-def: $vgpr38
+; SI-NEXT: ; implicit-def: $vgpr60
+; SI-NEXT: ; implicit-def: $vgpr24
; SI-NEXT: ; implicit-def: $vgpr13
-; SI-NEXT: ; implicit-def: $vgpr20
+; SI-NEXT: ; implicit-def: $vgpr46
; SI-NEXT: ; implicit-def: $vgpr35
-; SI-NEXT: ; implicit-def: $vgpr15
; SI-NEXT: ; implicit-def: $vgpr45
+; SI-NEXT: ; implicit-def: $vgpr2
; SI-NEXT: ; implicit-def: $vgpr10
; SI-NEXT: ; implicit-def: $vgpr30
+; SI-NEXT: ; implicit-def: $vgpr14
; SI-NEXT: ; implicit-def: $vgpr7
; SI-NEXT: ; implicit-def: $vgpr27
-; SI-NEXT: ; implicit-def: $vgpr8
-; SI-NEXT: ; implicit-def: $vgpr60
-; SI-NEXT: ; implicit-def: $vgpr26
-; SI-NEXT: ; implicit-def: $vgpr24
-; SI-NEXT: ; implicit-def: $vgpr57
-; SI-NEXT: ; implicit-def: $vgpr21
; SI-NEXT: ; implicit-def: $vgpr18
-; SI-NEXT: ; implicit-def: $vgpr34
+; SI-NEXT: ; implicit-def: $vgpr6
+; SI-NEXT: ; implicit-def: $vgpr37
+; SI-NEXT: ; implicit-def: $vgpr5
+; SI-NEXT: ; implicit-def: $vgpr21
+; SI-NEXT: ; implicit-def: $vgpr4
+; SI-NEXT: ; implicit-def: $vgpr61
+; SI-NEXT: ; implicit-def: $vgpr3
; SI-NEXT: ; implicit-def: $vgpr63
-; SI-NEXT: ; implicit-def: $vgpr32
-; SI-NEXT: ; implicit-def: $vgpr52
+; SI-NEXT: ; implicit-def: $vgpr56
; SI-NEXT: ; implicit-def: $vgpr12
; SI-NEXT: ; kill: killed $vgpr1
; SI-NEXT: ; implicit-def: $vgpr1
; SI-NEXT: ; kill: killed $vgpr9
; SI-NEXT: ; implicit-def: $vgpr9
-; SI-NEXT: ; kill: killed $vgpr40
-; SI-NEXT: ; implicit-def: $vgpr40
-; SI-NEXT: ; implicit-def: $vgpr56
-; SI-NEXT: ; kill: killed $vgpr56
+; SI-NEXT: ; kill: killed $vgpr17
+; SI-NEXT: ; kill: killed $vgpr49
+; SI-NEXT: ; implicit-def: $vgpr49
+; SI-NEXT: ; implicit-def: $vgpr17
+; SI-NEXT: ; kill: killed $vgpr17
; SI-NEXT: .LBB91_3: ; %Flow
-; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_mov_b32_e32 v56, v17
-; SI-NEXT: v_mov_b32_e32 v54, v61
-; SI-NEXT: buffer_load_dword v43, off, s[0:3], s32 offset:480 ; 4-byte Folded Reload
-; SI-NEXT: v_mov_b32_e32 v42, v32
-; SI-NEXT: buffer_load_dword v28, off, s[0:3], s32 offset:456 ; 4-byte Folded Reload
-; SI-NEXT: buffer_load_dword v32, off, s[0:3], s32 offset:460 ; 4-byte Folded Reload
-; SI-NEXT: buffer_load_dword v61, off, s[0:3], s32 offset:472 ; 4-byte Folded Reload
-; SI-NEXT: buffer_load_dword v33, off, s[0:3], s32 offset:464 ; 4-byte Folded Reload
-; SI-NEXT: buffer_load_dword v17, off, s[0:3], s32 offset:476 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v41, off, s[0:3], s32 offset:484 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v54, off, s[0:3], s32 offset:476 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v39, off, s[0:3], s32 offset:472 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v42, off, s[0:3], s32 offset:496 ; 4-byte Folded Reload
; SI-NEXT: s_andn2_b64 vcc, exec, s[4:5]
; SI-NEXT: s_cbranch_vccnz .LBB91_5
; SI-NEXT: ; %bb.4: ; %cmp.true
; SI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:684 ; 4-byte Folded Reload
; SI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:680 ; 4-byte Folded Reload
+; SI-NEXT: s_waitcnt vmcnt(2)
+; SI-NEXT: v_and_b32_e32 v10, 0xffff0000, v42
+; SI-NEXT: v_add_f32_e32 v10, 0x40c00000, v10
+; SI-NEXT: v_lshrrev_b32_e32 v10, 16, v10
+; SI-NEXT: v_and_b32_e32 v13, 0xffff0000, v54
+; SI-NEXT: v_add_f32_e32 v13, 0x40c00000, v13
+; SI-NEXT: v_lshrrev_b32_e32 v13, 16, v13
; SI-NEXT: buffer_load_dword v3, off, s[0:3], s32 offset:672 ; 4-byte Folded Reload
-; SI-NEXT: buffer_load_dword v8, off, s[0:3], s32 offset:644 ; 4-byte Folded Reload
-; SI-NEXT: buffer_load_dword v13, off, s[0:3], s32 offset:632 ; 4-byte Folded Reload
-; SI-NEXT: buffer_load_dword v14, off, s[0:3], s32 offset:612 ; 4-byte Folded Reload
-; SI-NEXT: buffer_load_dword v15, off, s[0:3], s32 offset:600 ; 4-byte Folded Reload
-; SI-NEXT: buffer_load_dword v17, off, s[0:3], s32 offset:576 ; 4-byte Folded Reload
-; SI-NEXT: buffer_load_dword v44, off, s[0:3], s32 offset:496 ; 4-byte Folded Reload
-; SI-NEXT: s_waitcnt vmcnt(8)
+; SI-NEXT: buffer_load_dword v18, off, s[0:3], s32 offset:596 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v44, off, s[0:3], s32 offset:512 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v23, off, s[0:3], s32 offset:168 ; 4-byte Folded Reload
+; SI-NEXT: s_waitcnt vmcnt(5)
; SI-NEXT: v_and_b32_e32 v1, 0xffff0000, v1
-; SI-NEXT: s_waitcnt vmcnt(7)
+; SI-NEXT: s_waitcnt vmcnt(4)
; SI-NEXT: v_and_b32_e32 v2, 0xffff0000, v2
; SI-NEXT: v_add_f32_e32 v2, 0x40c00000, v2
; SI-NEXT: v_add_f32_e32 v1, 0x40c00000, v1
; SI-NEXT: v_lshrrev_b32_e32 v2, 16, v2
; SI-NEXT: v_alignbit_b32 v1, v2, v1, 16
; SI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:676 ; 4-byte Folded Reload
-; SI-NEXT: s_waitcnt vmcnt(7)
+; SI-NEXT: s_waitcnt vmcnt(4)
; SI-NEXT: v_and_b32_e32 v3, 0xffff0000, v3
; SI-NEXT: v_add_f32_e32 v3, 0x40c00000, v3
; SI-NEXT: v_lshrrev_b32_e32 v3, 16, v3
-; SI-NEXT: s_waitcnt vmcnt(6)
-; SI-NEXT: v_and_b32_e32 v10, 0xffff0000, v8
-; SI-NEXT: v_add_f32_e32 v10, 0x40c00000, v10
-; SI-NEXT: v_lshrrev_b32_e32 v10, 16, v10
-; SI-NEXT: s_waitcnt vmcnt(5)
-; SI-NEXT: v_and_b32_e32 v13, 0xffff0000, v13
-; SI-NEXT: v_add_f32_e32 v13, 0x40c00000, v13
-; SI-NEXT: v_lshrrev_b32_e32 v13, 16, v13
-; SI-NEXT: s_waitcnt vmcnt(4)
-; SI-NEXT: v_and_b32_e32 v14, 0xffff0000, v14
-; SI-NEXT: v_add_f32_e32 v14, 0x40c00000, v14
-; SI-NEXT: v_lshrrev_b32_e32 v14, 16, v14
-; SI-NEXT: s_waitcnt vmcnt(3)
-; SI-NEXT: v_and_b32_e32 v16, 0xffff0000, v15
-; SI-NEXT: v_add_f32_e32 v16, 0x40c00000, v16
-; SI-NEXT: v_lshrrev_b32_e32 v16, 16, v16
-; SI-NEXT: s_waitcnt vmcnt(2)
-; SI-NEXT: v_and_b32_e32 v19, 0xffff0000, v17
-; SI-NEXT: buffer_load_dword v17, off, s[0:3], s32 offset:584 ; 4-byte Folded Reload
-; SI-NEXT: v_add_f32_e32 v19, 0x40c00000, v19
-; SI-NEXT: v_lshrrev_b32_e32 v19, 16, v19
; SI-NEXT: s_waitcnt vmcnt(2)
; SI-NEXT: v_and_b32_e32 v44, 0xffff0000, v44
-; SI-NEXT: s_waitcnt vmcnt(1)
+; SI-NEXT: v_add_f32_e32 v46, 0x40c00000, v44
+; SI-NEXT: s_waitcnt vmcnt(0)
; SI-NEXT: v_and_b32_e32 v2, 0xffff0000, v2
; SI-NEXT: v_add_f32_e32 v2, 0x40c00000, v2
-; SI-NEXT: v_alignbit_b32 v52, v3, v2, 16
+; SI-NEXT: v_alignbit_b32 v56, v3, v2, 16
; SI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:668 ; 4-byte Folded Reload
; SI-NEXT: s_waitcnt vmcnt(0)
; SI-NEXT: v_and_b32_e32 v3, 0xffff0000, v2
@@ -170949,941 +171037,960 @@ define inreg <128 x i8> @bitcast_v64bf16_to_v128i8_scalar(<64 x bfloat> inreg %a
; SI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:660 ; 4-byte Folded Reload
; SI-NEXT: v_add_f32_e32 v4, 0x40c00000, v4
; SI-NEXT: v_lshrrev_b32_e32 v4, 16, v4
-; SI-NEXT: v_alignbit_b32 v34, v4, v3, 16
-; SI-NEXT: v_and_b32_e32 v4, 0xffff0000, v33
+; SI-NEXT: v_alignbit_b32 v3, v4, v3, 16
+; SI-NEXT: s_waitcnt vmcnt(0)
+; SI-NEXT: v_and_b32_e32 v4, 0xffff0000, v2
+; SI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:656 ; 4-byte Folded Reload
; SI-NEXT: v_add_f32_e32 v4, 0x40c00000, v4
; SI-NEXT: s_waitcnt vmcnt(0)
; SI-NEXT: v_and_b32_e32 v5, 0xffff0000, v2
-; SI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:688 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:692 ; 4-byte Folded Reload
; SI-NEXT: v_add_f32_e32 v5, 0x40c00000, v5
; SI-NEXT: v_lshrrev_b32_e32 v5, 16, v5
-; SI-NEXT: v_alignbit_b32 v51, v5, v4, 16
+; SI-NEXT: v_alignbit_b32 v4, v5, v4, 16
; SI-NEXT: s_waitcnt vmcnt(0)
; SI-NEXT: v_and_b32_e32 v5, 0xffff0000, v2
-; SI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:656 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:648 ; 4-byte Folded Reload
; SI-NEXT: v_add_f32_e32 v6, 0x40c00000, v5
-; SI-NEXT: v_and_b32_e32 v5, 0xffff0000, v53
-; SI-NEXT: v_add_f32_e32 v26, 0x40c00000, v5
+; SI-NEXT: v_and_b32_e32 v5, 0xffff0000, v43
+; SI-NEXT: v_add_f32_e32 v8, 0x40c00000, v5
; SI-NEXT: s_waitcnt vmcnt(0)
; SI-NEXT: v_and_b32_e32 v5, 0xffff0000, v2
-; SI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:652 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:644 ; 4-byte Folded Reload
; SI-NEXT: v_add_f32_e32 v5, 0x40c00000, v5
; SI-NEXT: s_waitcnt vmcnt(0)
; SI-NEXT: v_and_b32_e32 v7, 0xffff0000, v2
+; SI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:652 ; 4-byte Folded Reload
; SI-NEXT: v_add_f32_e32 v7, 0x40c00000, v7
; SI-NEXT: v_lshrrev_b32_e32 v7, 16, v7
-; SI-NEXT: v_alignbit_b32 v57, v7, v5, 16
-; SI-NEXT: v_lshrrev_b32_e32 v7, 16, v26
+; SI-NEXT: v_alignbit_b32 v5, v7, v5, 16
+; SI-NEXT: v_lshrrev_b32_e32 v7, 16, v8
; SI-NEXT: v_alignbit_b32 v9, v7, v6, 16
-; SI-NEXT: v_and_b32_e32 v6, 0xffff0000, v32
; SI-NEXT: buffer_store_dword v7, off, s[0:3], s32 offset:148 ; 4-byte Folded Spill
-; SI-NEXT: s_waitcnt expcnt(0)
+; SI-NEXT: v_lshrrev_b32_e32 v8, 24, v8
+; SI-NEXT: buffer_store_dword v8, off, s[0:3], s32 offset:424 ; 4-byte Folded Spill
+; SI-NEXT: s_waitcnt vmcnt(2)
+; SI-NEXT: v_and_b32_e32 v6, 0xffff0000, v2
+; SI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:640 ; 4-byte Folded Reload
+; SI-NEXT: s_waitcnt expcnt(1)
; SI-NEXT: v_add_f32_e32 v7, 0x40c00000, v6
-; SI-NEXT: v_and_b32_e32 v6, 0xffff0000, v48
+; SI-NEXT: v_and_b32_e32 v6, 0xffff0000, v31
; SI-NEXT: v_add_f32_e32 v11, 0x40c00000, v6
-; SI-NEXT: buffer_load_dword v6, off, s[0:3], s32 offset:648 ; 4-byte Folded Reload
-; SI-NEXT: v_lshrrev_b32_e32 v8, 16, v11
-; SI-NEXT: v_alignbit_b32 v12, v8, v7, 16
-; SI-NEXT: v_and_b32_e32 v7, 0xffff0000, v28
-; SI-NEXT: s_waitcnt vmcnt(0)
-; SI-NEXT: v_and_b32_e32 v6, 0xffff0000, v6
+; SI-NEXT: v_lshrrev_b32_e32 v59, 16, v11
+; SI-NEXT: v_alignbit_b32 v12, v59, v7, 16
+; SI-NEXT: v_and_b32_e32 v7, 0xffff0000, v32
+; SI-NEXT: v_lshrrev_b32_e32 v11, 24, v11
+; SI-NEXT: buffer_store_dword v11, off, s[0:3], s32 offset:176 ; 4-byte Folded Spill
+; SI-NEXT: s_waitcnt vmcnt(1)
+; SI-NEXT: v_and_b32_e32 v6, 0xffff0000, v2
+; SI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:688 ; 4-byte Folded Reload
; SI-NEXT: v_add_f32_e32 v6, 0x40c00000, v6
-; SI-NEXT: v_alignbit_b32 v60, v10, v6, 16
+; SI-NEXT: v_alignbit_b32 v6, v10, v6, 16
; SI-NEXT: v_add_f32_e32 v10, 0x40c00000, v7
-; SI-NEXT: v_and_b32_e32 v7, 0xffff0000, v29
-; SI-NEXT: v_add_f32_e32 v31, 0x40c00000, v7
-; SI-NEXT: buffer_load_dword v7, off, s[0:3], s32 offset:636 ; 4-byte Folded Reload
; SI-NEXT: s_waitcnt vmcnt(0)
-; SI-NEXT: v_and_b32_e32 v7, 0xffff0000, v7
+; SI-NEXT: v_and_b32_e32 v7, 0xffff0000, v2
+; SI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:636 ; 4-byte Folded Reload
+; SI-NEXT: v_add_f32_e32 v14, 0x40c00000, v7
+; SI-NEXT: v_and_b32_e32 v7, 0xffff0000, v39
; SI-NEXT: v_add_f32_e32 v7, 0x40c00000, v7
; SI-NEXT: v_alignbit_b32 v7, v13, v7, 16
-; SI-NEXT: v_lshrrev_b32_e32 v13, 16, v31
+; SI-NEXT: v_lshrrev_b32_e32 v13, 16, v14
; SI-NEXT: v_alignbit_b32 v63, v13, v10, 16
-; SI-NEXT: buffer_load_dword v10, off, s[0:3], s32 offset:640 ; 4-byte Folded Reload
+; SI-NEXT: v_and_b32_e32 v10, 0xffff0000, v48
; SI-NEXT: buffer_store_dword v13, off, s[0:3], s32 offset:152 ; 4-byte Folded Spill
-; SI-NEXT: s_waitcnt vmcnt(1)
-; SI-NEXT: v_and_b32_e32 v10, 0xffff0000, v10
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_add_f32_e32 v13, 0x40c00000, v10
-; SI-NEXT: v_and_b32_e32 v10, 0xffff0000, v37
-; SI-NEXT: v_add_f32_e32 v32, 0x40c00000, v10
-; SI-NEXT: buffer_load_dword v10, off, s[0:3], s32 offset:616 ; 4-byte Folded Reload
-; SI-NEXT: s_waitcnt vmcnt(0)
-; SI-NEXT: v_and_b32_e32 v10, 0xffff0000, v10
+; SI-NEXT: v_and_b32_e32 v10, 0xffff0000, v29
+; SI-NEXT: v_add_f32_e32 v17, 0x40c00000, v10
+; SI-NEXT: v_lshrrev_b32_e32 v14, 24, v14
+; SI-NEXT: buffer_store_dword v14, off, s[0:3], s32 offset:184 ; 4-byte Folded Spill
+; SI-NEXT: s_waitcnt vmcnt(2)
+; SI-NEXT: v_and_b32_e32 v10, 0xffff0000, v2
+; SI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:632 ; 4-byte Folded Reload
; SI-NEXT: v_add_f32_e32 v10, 0x40c00000, v10
-; SI-NEXT: v_alignbit_b32 v10, v14, v10, 16
-; SI-NEXT: v_lshrrev_b32_e32 v14, 16, v32
-; SI-NEXT: v_alignbit_b32 v18, v14, v13, 16
-; SI-NEXT: buffer_load_dword v13, off, s[0:3], s32 offset:628 ; 4-byte Folded Reload
-; SI-NEXT: buffer_store_dword v14, off, s[0:3], s32 offset:188 ; 4-byte Folded Spill
-; SI-NEXT: s_waitcnt vmcnt(1)
-; SI-NEXT: v_and_b32_e32 v13, 0xffff0000, v13
-; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_add_f32_e32 v14, 0x40c00000, v13
-; SI-NEXT: buffer_load_dword v13, off, s[0:3], s32 offset:624 ; 4-byte Folded Reload
; SI-NEXT: s_waitcnt vmcnt(0)
-; SI-NEXT: v_and_b32_e32 v13, 0xffff0000, v13
-; SI-NEXT: v_add_f32_e32 v49, 0x40c00000, v13
-; SI-NEXT: buffer_load_dword v13, off, s[0:3], s32 offset:604 ; 4-byte Folded Reload
-; SI-NEXT: v_lshrrev_b32_e32 v15, 16, v49
-; SI-NEXT: buffer_store_dword v15, off, s[0:3], s32 offset:248 ; 4-byte Folded Spill
-; SI-NEXT: v_alignbit_b32 v21, v15, v14, 16
-; SI-NEXT: buffer_load_dword v14, off, s[0:3], s32 offset:608 ; 4-byte Folded Reload
-; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: buffer_load_dword v15, off, s[0:3], s32 offset:620 ; 4-byte Folded Reload
-; SI-NEXT: s_waitcnt vmcnt(3)
-; SI-NEXT: v_and_b32_e32 v13, 0xffff0000, v13
+; SI-NEXT: v_and_b32_e32 v16, 0xffff0000, v2
+; SI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:624 ; 4-byte Folded Reload
+; SI-NEXT: v_add_f32_e32 v16, 0x40c00000, v16
+; SI-NEXT: v_lshrrev_b32_e32 v16, 16, v16
+; SI-NEXT: v_alignbit_b32 v10, v16, v10, 16
+; SI-NEXT: v_lshrrev_b32_e32 v16, 16, v17
+; SI-NEXT: v_alignbit_b32 v61, v16, v13, 16
+; SI-NEXT: v_and_b32_e32 v13, 0xffff0000, v40
+; SI-NEXT: buffer_store_dword v16, off, s[0:3], s32 offset:156 ; 4-byte Folded Spill
+; SI-NEXT: s_waitcnt expcnt(0)
+; SI-NEXT: v_add_f32_e32 v16, 0x40c00000, v13
+; SI-NEXT: v_and_b32_e32 v13, 0xffff0000, v15
+; SI-NEXT: v_add_f32_e32 v20, 0x40c00000, v13
+; SI-NEXT: v_and_b32_e32 v13, 0xffff0000, v41
; SI-NEXT: v_add_f32_e32 v13, 0x40c00000, v13
-; SI-NEXT: v_alignbit_b32 v13, v16, v13, 16
-; SI-NEXT: s_waitcnt vmcnt(0)
-; SI-NEXT: v_and_b32_e32 v16, 0xffff0000, v15
-; SI-NEXT: v_add_f32_e32 v23, 0x40c00000, v16
-; SI-NEXT: buffer_load_dword v16, off, s[0:3], s32 offset:580 ; 4-byte Folded Reload
-; SI-NEXT: v_lshrrev_b32_e32 v15, 16, v23
-; SI-NEXT: buffer_store_dword v15, off, s[0:3], s32 offset:252 ; 4-byte Folded Spill
+; SI-NEXT: v_lshrrev_b32_e32 v17, 24, v17
+; SI-NEXT: buffer_store_dword v17, off, s[0:3], s32 offset:208 ; 4-byte Folded Spill
+; SI-NEXT: s_waitcnt vmcnt(2)
+; SI-NEXT: v_and_b32_e32 v19, 0xffff0000, v2
+; SI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:628 ; 4-byte Folded Reload
+; SI-NEXT: v_add_f32_e32 v19, 0x40c00000, v19
+; SI-NEXT: v_lshrrev_b32_e32 v19, 16, v19
+; SI-NEXT: v_alignbit_b32 v13, v19, v13, 16
+; SI-NEXT: v_lshrrev_b32_e32 v19, 16, v20
+; SI-NEXT: v_alignbit_b32 v21, v19, v16, 16
+; SI-NEXT: buffer_store_dword v19, off, s[0:3], s32 offset:160 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt vmcnt(1)
-; SI-NEXT: v_and_b32_e32 v16, 0xffff0000, v16
+; SI-NEXT: v_and_b32_e32 v16, 0xffff0000, v2
+; SI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:604 ; 4-byte Folded Reload
+; SI-NEXT: s_waitcnt expcnt(0)
+; SI-NEXT: v_add_f32_e32 v19, 0x40c00000, v16
+; SI-NEXT: v_and_b32_e32 v16, 0xffff0000, v51
+; SI-NEXT: v_add_f32_e32 v31, 0x40c00000, v16
+; SI-NEXT: s_waitcnt vmcnt(0)
+; SI-NEXT: v_and_b32_e32 v16, 0xffff0000, v2
+; SI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:600 ; 4-byte Folded Reload
; SI-NEXT: v_add_f32_e32 v16, 0x40c00000, v16
-; SI-NEXT: v_alignbit_b32 v16, v19, v16, 16
-; SI-NEXT: v_and_b32_e32 v19, 0xffff0000, v17
-; SI-NEXT: buffer_load_dword v17, off, s[0:3], s32 offset:560 ; 4-byte Folded Reload
-; SI-NEXT: v_add_f32_e32 v20, 0x40c00000, v19
; SI-NEXT: s_waitcnt vmcnt(0)
-; SI-NEXT: v_and_b32_e32 v19, 0xffff0000, v17
-; SI-NEXT: buffer_load_dword v17, off, s[0:3], s32 offset:556 ; 4-byte Folded Reload
-; SI-NEXT: v_and_b32_e32 v14, 0xffff0000, v14
-; SI-NEXT: v_add_f32_e32 v14, 0x40c00000, v14
-; SI-NEXT: v_alignbit_b32 v24, v15, v14, 16
-; SI-NEXT: buffer_load_dword v14, off, s[0:3], s32 offset:588 ; 4-byte Folded Reload
+; SI-NEXT: v_and_b32_e32 v22, 0xffff0000, v2
+; SI-NEXT: v_lshrrev_b32_e32 v2, 16, v31
+; SI-NEXT: buffer_store_dword v2, off, s[0:3], s32 offset:316 ; 4-byte Folded Spill
+; SI-NEXT: v_alignbit_b32 v37, v2, v19, 16
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_lshrrev_b32_e32 v15, 16, v20
-; SI-NEXT: v_add_f32_e32 v19, 0x40c00000, v19
-; SI-NEXT: buffer_store_dword v15, off, s[0:3], s32 offset:352 ; 4-byte Folded Spill
-; SI-NEXT: v_lshrrev_b32_e32 v29, 8, v24
-; SI-NEXT: s_waitcnt vmcnt(2)
-; SI-NEXT: v_and_b32_e32 v22, 0xffff0000, v17
-; SI-NEXT: buffer_load_dword v17, off, s[0:3], s32 offset:540 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:612 ; 4-byte Folded Reload
; SI-NEXT: v_add_f32_e32 v22, 0x40c00000, v22
; SI-NEXT: v_lshrrev_b32_e32 v22, 16, v22
-; SI-NEXT: s_waitcnt vmcnt(2)
-; SI-NEXT: v_and_b32_e32 v14, 0xffff0000, v14
-; SI-NEXT: v_add_f32_e32 v14, 0x40c00000, v14
-; SI-NEXT: v_alignbit_b32 v27, v15, v14, 16
-; SI-NEXT: buffer_load_dword v14, off, s[0:3], s32 offset:564 ; 4-byte Folded Reload
-; SI-NEXT: v_alignbit_b32 v3, v22, v19, 16
-; SI-NEXT: v_and_b32_e32 v22, 0xffff0000, v59
-; SI-NEXT: v_add_f32_e32 v54, 0x40c00000, v22
+; SI-NEXT: v_alignbit_b32 v57, v22, v16, 16
+; SI-NEXT: v_lshrrev_b32_e32 v17, 8, v37
+; SI-NEXT: buffer_store_dword v17, off, s[0:3], s32 offset:416 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_lshrrev_b32_e32 v15, 16, v54
-; SI-NEXT: buffer_store_dword v15, off, s[0:3], s32 offset:340 ; 4-byte Folded Spill
-; SI-NEXT: v_add_f32_e32 v59, 0x40c00000, v44
-; SI-NEXT: v_lshrrev_b32_e32 v19, 16, v59
-; SI-NEXT: s_waitcnt vmcnt(2)
-; SI-NEXT: v_and_b32_e32 v22, 0xffff0000, v17
-; SI-NEXT: buffer_load_dword v17, off, s[0:3], s32 offset:536 ; 4-byte Folded Reload
-; SI-NEXT: v_add_f32_e32 v22, 0x40c00000, v22
-; SI-NEXT: s_waitcnt vmcnt(2)
-; SI-NEXT: v_and_b32_e32 v14, 0xffff0000, v14
-; SI-NEXT: v_add_f32_e32 v14, 0x40c00000, v14
-; SI-NEXT: v_alignbit_b32 v30, v15, v14, 16
-; SI-NEXT: buffer_load_dword v14, off, s[0:3], s32 offset:544 ; 4-byte Folded Reload
+; SI-NEXT: v_lshrrev_b32_e32 v17, 8, v21
+; SI-NEXT: buffer_store_dword v17, off, s[0:3], s32 offset:384 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: buffer_load_dword v15, off, s[0:3], s32 offset:596 ; 4-byte Folded Reload
-; SI-NEXT: s_waitcnt vmcnt(2)
-; SI-NEXT: v_and_b32_e32 v25, 0xffff0000, v17
-; SI-NEXT: buffer_load_dword v17, off, s[0:3], s32 offset:164 ; 4-byte Folded Reload
+; SI-NEXT: v_lshrrev_b32_e32 v17, 8, v61
+; SI-NEXT: buffer_store_dword v17, off, s[0:3], s32 offset:360 ; 4-byte Folded Spill
+; SI-NEXT: s_waitcnt expcnt(0)
+; SI-NEXT: v_lshrrev_b32_e32 v17, 8, v63
+; SI-NEXT: buffer_store_dword v17, off, s[0:3], s32 offset:336 ; 4-byte Folded Spill
+; SI-NEXT: s_waitcnt expcnt(0)
+; SI-NEXT: v_lshrrev_b32_e32 v17, 8, v12
+; SI-NEXT: buffer_store_dword v17, off, s[0:3], s32 offset:308 ; 4-byte Folded Spill
+; SI-NEXT: s_waitcnt expcnt(0)
+; SI-NEXT: v_lshrrev_b32_e32 v17, 8, v9
+; SI-NEXT: buffer_store_dword v17, off, s[0:3], s32 offset:288 ; 4-byte Folded Spill
+; SI-NEXT: v_lshrrev_b32_e32 v16, 16, v46
+; SI-NEXT: s_waitcnt vmcnt(6)
+; SI-NEXT: v_and_b32_e32 v19, 0xffff0000, v2
+; SI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:608 ; 4-byte Folded Reload
+; SI-NEXT: v_add_f32_e32 v22, 0x40c00000, v19
+; SI-NEXT: s_waitcnt vmcnt(0)
+; SI-NEXT: v_and_b32_e32 v19, 0xffff0000, v2
+; SI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:584 ; 4-byte Folded Reload
+; SI-NEXT: v_add_f32_e32 v39, 0x40c00000, v19
+; SI-NEXT: s_waitcnt vmcnt(0)
+; SI-NEXT: v_and_b32_e32 v19, 0xffff0000, v2
+; SI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:580 ; 4-byte Folded Reload
+; SI-NEXT: v_add_f32_e32 v19, 0x40c00000, v19
+; SI-NEXT: s_waitcnt vmcnt(0)
+; SI-NEXT: v_and_b32_e32 v25, 0xffff0000, v2
+; SI-NEXT: v_lshrrev_b32_e32 v2, 16, v39
+; SI-NEXT: buffer_store_dword v2, off, s[0:3], s32 offset:388 ; 4-byte Folded Spill
+; SI-NEXT: v_alignbit_b32 v27, v2, v22, 16
+; SI-NEXT: s_waitcnt expcnt(0)
+; SI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:588 ; 4-byte Folded Reload
; SI-NEXT: v_add_f32_e32 v25, 0x40c00000, v25
; SI-NEXT: v_lshrrev_b32_e32 v25, 16, v25
-; SI-NEXT: v_alignbit_b32 v4, v25, v22, 16
-; SI-NEXT: s_waitcnt vmcnt(2)
-; SI-NEXT: v_and_b32_e32 v14, 0xffff0000, v14
-; SI-NEXT: v_add_f32_e32 v14, 0x40c00000, v14
+; SI-NEXT: v_alignbit_b32 v19, v25, v19, 16
; SI-NEXT: s_waitcnt vmcnt(0)
-; SI-NEXT: v_and_b32_e32 v25, 0xffff0000, v17
-; SI-NEXT: v_add_f32_e32 v40, 0x40c00000, v25
-; SI-NEXT: v_and_b32_e32 v25, 0xffff0000, v15
-; SI-NEXT: buffer_load_dword v15, off, s[0:3], s32 offset:592 ; 4-byte Folded Reload
-; SI-NEXT: buffer_load_dword v17, off, s[0:3], s32 offset:160 ; 4-byte Folded Reload
-; SI-NEXT: v_lshrrev_b32_e32 v45, 16, v40
-; SI-NEXT: v_alignbit_b32 v35, v45, v14, 16
-; SI-NEXT: buffer_load_dword v14, off, s[0:3], s32 offset:524 ; 4-byte Folded Reload
-; SI-NEXT: v_add_f32_e32 v25, 0x40c00000, v25
-; SI-NEXT: s_waitcnt vmcnt(2)
-; SI-NEXT: v_and_b32_e32 v28, 0xffff0000, v15
+; SI-NEXT: v_and_b32_e32 v22, 0xffff0000, v2
+; SI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:172 ; 4-byte Folded Reload
+; SI-NEXT: v_add_f32_e32 v25, 0x40c00000, v22
+; SI-NEXT: s_waitcnt vmcnt(0)
+; SI-NEXT: v_and_b32_e32 v22, 0xffff0000, v2
+; SI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:560 ; 4-byte Folded Reload
+; SI-NEXT: v_add_f32_e32 v49, 0x40c00000, v22
+; SI-NEXT: s_waitcnt vmcnt(0)
+; SI-NEXT: v_and_b32_e32 v22, 0xffff0000, v2
+; SI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:556 ; 4-byte Folded Reload
+; SI-NEXT: v_add_f32_e32 v22, 0x40c00000, v22
+; SI-NEXT: s_waitcnt vmcnt(0)
+; SI-NEXT: v_and_b32_e32 v28, 0xffff0000, v2
+; SI-NEXT: v_lshrrev_b32_e32 v2, 16, v49
+; SI-NEXT: buffer_store_dword v2, off, s[0:3], s32 offset:440 ; 4-byte Folded Spill
+; SI-NEXT: v_alignbit_b32 v30, v2, v25, 16
+; SI-NEXT: s_waitcnt expcnt(0)
+; SI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:568 ; 4-byte Folded Reload
; SI-NEXT: v_add_f32_e32 v28, 0x40c00000, v28
; SI-NEXT: v_lshrrev_b32_e32 v28, 16, v28
-; SI-NEXT: v_alignbit_b32 v5, v28, v25, 16
-; SI-NEXT: s_waitcnt vmcnt(1)
-; SI-NEXT: v_and_b32_e32 v28, 0xffff0000, v17
-; SI-NEXT: buffer_load_dword v17, off, s[0:3], s32 offset:572 ; 4-byte Folded Reload
-; SI-NEXT: v_add_f32_e32 v41, 0x40c00000, v28
+; SI-NEXT: v_alignbit_b32 v53, v28, v22, 16
+; SI-NEXT: v_lshrrev_b32_e32 v14, 8, v30
+; SI-NEXT: s_waitcnt vmcnt(0)
+; SI-NEXT: v_and_b32_e32 v25, 0xffff0000, v2
+; SI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:564 ; 4-byte Folded Reload
+; SI-NEXT: v_add_f32_e32 v28, 0x40c00000, v25
+; SI-NEXT: s_waitcnt vmcnt(0)
+; SI-NEXT: v_and_b32_e32 v25, 0xffff0000, v2
+; SI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:620 ; 4-byte Folded Reload
+; SI-NEXT: v_add_f32_e32 v51, 0x40c00000, v25
+; SI-NEXT: s_waitcnt vmcnt(0)
+; SI-NEXT: v_and_b32_e32 v25, 0xffff0000, v2
+; SI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:616 ; 4-byte Folded Reload
+; SI-NEXT: v_add_f32_e32 v25, 0x40c00000, v25
+; SI-NEXT: s_waitcnt vmcnt(0)
+; SI-NEXT: v_and_b32_e32 v32, 0xffff0000, v2
+; SI-NEXT: v_lshrrev_b32_e32 v2, 16, v51
+; SI-NEXT: v_alignbit_b32 v35, v2, v28, 16
+; SI-NEXT: v_and_b32_e32 v28, 0xffff0000, v23
+; SI-NEXT: buffer_load_dword v23, off, s[0:3], s32 offset:544 ; 4-byte Folded Reload
+; SI-NEXT: v_add_f32_e32 v32, 0x40c00000, v32
+; SI-NEXT: v_lshrrev_b32_e32 v32, 16, v32
+; SI-NEXT: v_alignbit_b32 v25, v32, v25, 16
+; SI-NEXT: v_add_f32_e32 v32, 0x40c00000, v28
+; SI-NEXT: v_lshrrev_b32_e32 v45, 8, v35
+; SI-NEXT: s_waitcnt vmcnt(0)
+; SI-NEXT: v_and_b32_e32 v28, 0xffff0000, v23
+; SI-NEXT: buffer_load_dword v23, off, s[0:3], s32 offset:540 ; 4-byte Folded Reload
+; SI-NEXT: v_add_f32_e32 v52, 0x40c00000, v28
+; SI-NEXT: v_and_b32_e32 v28, 0xffff0000, v18
+; SI-NEXT: buffer_load_dword v18, off, s[0:3], s32 offset:592 ; 4-byte Folded Reload
+; SI-NEXT: v_lshrrev_b32_e32 v24, 16, v52
+; SI-NEXT: v_alignbit_b32 v38, v24, v32, 16
+; SI-NEXT: v_add_f32_e32 v28, 0x40c00000, v28
+; SI-NEXT: v_lshrrev_b32_e32 v60, 8, v38
; SI-NEXT: s_waitcnt vmcnt(1)
-; SI-NEXT: v_and_b32_e32 v14, 0xffff0000, v14
-; SI-NEXT: v_add_f32_e32 v14, 0x40c00000, v14
-; SI-NEXT: v_lshrrev_b32_e32 v15, 16, v41
-; SI-NEXT: v_alignbit_b32 v38, v15, v14, 16
-; SI-NEXT: v_and_b32_e32 v14, 0xffff0000, v56
-; SI-NEXT: buffer_store_dword v15, off, s[0:3], s32 offset:412 ; 4-byte Folded Spill
-; SI-NEXT: v_add_f32_e32 v14, 0x40c00000, v14
+; SI-NEXT: v_and_b32_e32 v32, 0xffff0000, v23
+; SI-NEXT: buffer_load_dword v23, off, s[0:3], s32 offset:536 ; 4-byte Folded Reload
+; SI-NEXT: v_add_f32_e32 v32, 0x40c00000, v32
; SI-NEXT: s_waitcnt vmcnt(1)
-; SI-NEXT: v_and_b32_e32 v28, 0xffff0000, v17
-; SI-NEXT: buffer_load_dword v17, off, s[0:3], s32 offset:568 ; 4-byte Folded Reload
-; SI-NEXT: v_add_f32_e32 v28, 0x40c00000, v28
-; SI-NEXT: s_waitcnt vmcnt(0)
-; SI-NEXT: v_and_b32_e32 v33, 0xffff0000, v17
+; SI-NEXT: v_and_b32_e32 v33, 0xffff0000, v18
; SI-NEXT: v_add_f32_e32 v33, 0x40c00000, v33
; SI-NEXT: v_lshrrev_b32_e32 v33, 16, v33
-; SI-NEXT: v_alignbit_b32 v2, v33, v28, 16
-; SI-NEXT: buffer_load_dword v33, off, s[0:3], s32 offset:520 ; 4-byte Folded Reload
-; SI-NEXT: buffer_load_dword v17, off, s[0:3], s32 offset:552 ; 4-byte Folded Reload
-; SI-NEXT: s_waitcnt vmcnt(1)
-; SI-NEXT: v_and_b32_e32 v33, 0xffff0000, v33
-; SI-NEXT: v_add_f32_e32 v43, 0x40c00000, v33
-; SI-NEXT: s_waitcnt vmcnt(0)
-; SI-NEXT: v_and_b32_e32 v33, 0xffff0000, v17
-; SI-NEXT: buffer_load_dword v17, off, s[0:3], s32 offset:548 ; 4-byte Folded Reload
+; SI-NEXT: v_alignbit_b32 v28, v33, v28, 16
+; SI-NEXT: s_waitcnt vmcnt(0)
+; SI-NEXT: v_and_b32_e32 v33, 0xffff0000, v23
+; SI-NEXT: buffer_load_dword v23, off, s[0:3], s32 offset:576 ; 4-byte Folded Reload
+; SI-NEXT: v_add_f32_e32 v54, 0x40c00000, v33
+; SI-NEXT: v_lshrrev_b32_e32 v18, 16, v54
+; SI-NEXT: v_alignbit_b32 v55, v18, v32, 16
+; SI-NEXT: buffer_load_dword v32, off, s[0:3], s32 offset:508 ; 4-byte Folded Reload
+; SI-NEXT: buffer_store_dword v18, off, s[0:3], s32 offset:320 ; 4-byte Folded Spill
+; SI-NEXT: s_waitcnt vmcnt(2)
+; SI-NEXT: v_and_b32_e32 v33, 0xffff0000, v23
+; SI-NEXT: buffer_load_dword v23, off, s[0:3], s32 offset:572 ; 4-byte Folded Reload
; SI-NEXT: v_add_f32_e32 v33, 0x40c00000, v33
-; SI-NEXT: v_lshrrev_b32_e32 v61, 16, v43
-; SI-NEXT: v_alignbit_b32 v55, v61, v14, 16
-; SI-NEXT: buffer_load_dword v14, off, s[0:3], s32 offset:492 ; 4-byte Folded Reload
-; SI-NEXT: s_waitcnt vmcnt(1)
-; SI-NEXT: v_and_b32_e32 v36, 0xffff0000, v17
-; SI-NEXT: buffer_load_dword v17, off, s[0:3], s32 offset:532 ; 4-byte Folded Reload
+; SI-NEXT: s_waitcnt vmcnt(0)
+; SI-NEXT: v_and_b32_e32 v36, 0xffff0000, v23
; SI-NEXT: v_add_f32_e32 v36, 0x40c00000, v36
; SI-NEXT: v_lshrrev_b32_e32 v36, 16, v36
-; SI-NEXT: v_alignbit_b32 v6, v36, v33, 16
-; SI-NEXT: buffer_load_dword v36, off, s[0:3], s32 offset:156 ; 4-byte Folded Reload
-; SI-NEXT: s_waitcnt vmcnt(2)
-; SI-NEXT: v_and_b32_e32 v14, 0xffff0000, v14
-; SI-NEXT: v_add_f32_e32 v14, 0x40c00000, v14
+; SI-NEXT: v_alignbit_b32 v33, v36, v33, 16
+; SI-NEXT: buffer_load_dword v23, off, s[0:3], s32 offset:552 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v36, off, s[0:3], s32 offset:164 ; 4-byte Folded Reload
+; SI-NEXT: v_and_b32_e32 v32, 0xffff0000, v32
+; SI-NEXT: v_add_f32_e32 v32, 0x40c00000, v32
; SI-NEXT: s_waitcnt vmcnt(0)
; SI-NEXT: v_and_b32_e32 v36, 0xffff0000, v36
-; SI-NEXT: v_add_f32_e32 v46, 0x40c00000, v36
-; SI-NEXT: v_and_b32_e32 v36, 0xffff0000, v17
-; SI-NEXT: buffer_load_dword v17, off, s[0:3], s32 offset:528 ; 4-byte Folded Reload
-; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_lshrrev_b32_e32 v15, 16, v46
+; SI-NEXT: v_add_f32_e32 v41, 0x40c00000, v36
+; SI-NEXT: v_and_b32_e32 v36, 0xffff0000, v23
+; SI-NEXT: buffer_load_dword v23, off, s[0:3], s32 offset:548 ; 4-byte Folded Reload
+; SI-NEXT: v_lshrrev_b32_e32 v15, 16, v41
+; SI-NEXT: v_alignbit_b32 v62, v15, v32, 16
+; SI-NEXT: buffer_load_dword v32, off, s[0:3], s32 offset:504 ; 4-byte Folded Reload
; SI-NEXT: v_add_f32_e32 v36, 0x40c00000, v36
-; SI-NEXT: v_alignbit_b32 v62, v15, v14, 16
-; SI-NEXT: buffer_load_dword v14, off, s[0:3], s32 offset:488 ; 4-byte Folded Reload
-; SI-NEXT: buffer_store_dword v15, off, s[0:3], s32 offset:416 ; 4-byte Folded Spill
-; SI-NEXT: s_waitcnt vmcnt(2)
-; SI-NEXT: v_and_b32_e32 v39, 0xffff0000, v17
-; SI-NEXT: v_add_f32_e32 v39, 0x40c00000, v39
-; SI-NEXT: v_lshrrev_b32_e32 v39, 16, v39
-; SI-NEXT: v_alignbit_b32 v36, v39, v36, 16
-; SI-NEXT: buffer_load_dword v39, off, s[0:3], s32 offset:484 ; 4-byte Folded Reload
-; SI-NEXT: s_waitcnt vmcnt(2)
-; SI-NEXT: v_and_b32_e32 v14, 0xffff0000, v14
-; SI-NEXT: v_add_f32_e32 v14, 0x40c00000, v14
-; SI-NEXT: s_waitcnt vmcnt(0)
-; SI-NEXT: v_and_b32_e32 v39, 0xffff0000, v39
-; SI-NEXT: v_add_f32_e32 v42, 0x40c00000, v39
-; SI-NEXT: v_lshrrev_b32_e32 v17, 16, v42
-; SI-NEXT: v_alignbit_b32 v50, v17, v14, 16
-; SI-NEXT: buffer_load_dword v14, off, s[0:3], s32 offset:516 ; 4-byte Folded Reload
-; SI-NEXT: buffer_load_dword v39, off, s[0:3], s32 offset:512 ; 4-byte Folded Reload
-; SI-NEXT: s_waitcnt vmcnt(1)
-; SI-NEXT: v_and_b32_e32 v14, 0xffff0000, v14
-; SI-NEXT: s_waitcnt vmcnt(0)
-; SI-NEXT: v_and_b32_e32 v39, 0xffff0000, v39
-; SI-NEXT: v_add_f32_e32 v56, 0x40c00000, v39
-; SI-NEXT: v_add_f32_e32 v14, 0x40c00000, v14
-; SI-NEXT: v_lshrrev_b32_e32 v25, 16, v56
-; SI-NEXT: v_alignbit_b32 v47, v25, v14, 16
-; SI-NEXT: buffer_load_dword v14, off, s[0:3], s32 offset:508 ; 4-byte Folded Reload
-; SI-NEXT: buffer_load_dword v39, off, s[0:3], s32 offset:504 ; 4-byte Folded Reload
-; SI-NEXT: s_waitcnt vmcnt(1)
-; SI-NEXT: v_and_b32_e32 v14, 0xffff0000, v14
-; SI-NEXT: s_waitcnt vmcnt(0)
-; SI-NEXT: v_and_b32_e32 v39, 0xffff0000, v39
-; SI-NEXT: v_add_f32_e32 v39, 0x40c00000, v39
-; SI-NEXT: v_add_f32_e32 v14, 0x40c00000, v14
-; SI-NEXT: v_lshrrev_b32_e32 v22, 16, v39
-; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_lshrrev_b32_e32 v15, 24, v39
-; SI-NEXT: v_lshrrev_b32_e32 v39, 8, v50
-; SI-NEXT: v_alignbit_b32 v58, v22, v14, 16
-; SI-NEXT: buffer_load_dword v14, off, s[0:3], s32 offset:500 ; 4-byte Folded Reload
-; SI-NEXT: buffer_store_dword v15, off, s[0:3], s32 offset:448 ; 4-byte Folded Spill
-; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_lshrrev_b32_e32 v15, 24, v56
-; SI-NEXT: buffer_store_dword v15, off, s[0:3], s32 offset:432 ; 4-byte Folded Spill
-; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_lshrrev_b32_e32 v15, 24, v42
-; SI-NEXT: v_lshrrev_b32_e32 v42, 8, v63
-; SI-NEXT: buffer_store_dword v15, off, s[0:3], s32 offset:424 ; 4-byte Folded Spill
-; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_lshrrev_b32_e32 v15, 24, v46
-; SI-NEXT: v_lshrrev_b32_e32 v46, 8, v55
-; SI-NEXT: buffer_store_dword v15, off, s[0:3], s32 offset:404 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v15, off, s[0:3], s32 offset:452 ; 4-byte Folded Spill
+; SI-NEXT: s_waitcnt expcnt(1)
+; SI-NEXT: v_lshrrev_b32_e32 v18, 24, v41
+; SI-NEXT: buffer_store_dword v18, off, s[0:3], s32 offset:428 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_lshrrev_b32_e32 v15, 24, v43
-; SI-NEXT: v_alignbit_b32 v43, v38, v16, 8
-; SI-NEXT: buffer_store_dword v15, off, s[0:3], s32 offset:376 ; 4-byte Folded Spill
+; SI-NEXT: v_lshrrev_b32_e32 v18, 24, v54
+; SI-NEXT: buffer_store_dword v18, off, s[0:3], s32 offset:404 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_lshrrev_b32_e32 v15, 24, v41
-; SI-NEXT: v_alignbit_b32 v41, v38, v16, 16
-; SI-NEXT: buffer_store_dword v15, off, s[0:3], s32 offset:348 ; 4-byte Folded Spill
+; SI-NEXT: v_lshrrev_b32_e32 v18, 24, v52
+; SI-NEXT: buffer_store_dword v18, off, s[0:3], s32 offset:376 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_lshrrev_b32_e32 v15, 24, v40
-; SI-NEXT: v_mov_b32_e32 v40, v8
-; SI-NEXT: v_lshrrev_b32_e32 v8, 24, v54
-; SI-NEXT: v_alignbit_b32 v54, v38, v16, 24
-; SI-NEXT: buffer_store_dword v8, off, s[0:3], s32 offset:304 ; 4-byte Folded Spill
+; SI-NEXT: v_lshrrev_b32_e32 v18, 24, v51
+; SI-NEXT: buffer_store_dword v18, off, s[0:3], s32 offset:348 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_lshrrev_b32_e32 v8, 24, v20
-; SI-NEXT: buffer_store_dword v15, off, s[0:3], s32 offset:320 ; 4-byte Folded Spill
-; SI-NEXT: v_alignbit_b32 v20, v35, v13, 8
+; SI-NEXT: v_lshrrev_b32_e32 v18, 24, v49
+; SI-NEXT: buffer_store_dword v18, off, s[0:3], s32 offset:312 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_lshrrev_b32_e32 v15, 8, v35
-; SI-NEXT: buffer_store_dword v8, off, s[0:3], s32 offset:272 ; 4-byte Folded Spill
+; SI-NEXT: v_lshrrev_b32_e32 v18, 24, v39
+; SI-NEXT: buffer_store_dword v18, off, s[0:3], s32 offset:280 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_lshrrev_b32_e32 v8, 24, v23
-; SI-NEXT: v_alignbit_b32 v23, v62, v4, 8
-; SI-NEXT: buffer_store_dword v8, off, s[0:3], s32 offset:452 ; 4-byte Folded Spill
+; SI-NEXT: v_lshrrev_b32_e32 v18, 24, v31
+; SI-NEXT: buffer_store_dword v18, off, s[0:3], s32 offset:260 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_lshrrev_b32_e32 v8, 24, v49
-; SI-NEXT: v_lshrrev_b32_e32 v49, 8, v47
-; SI-NEXT: buffer_store_dword v8, off, s[0:3], s32 offset:224 ; 4-byte Folded Spill
+; SI-NEXT: v_lshrrev_b32_e32 v18, 24, v20
+; SI-NEXT: v_mov_b32_e32 v49, v59
+; SI-NEXT: buffer_store_dword v18, off, s[0:3], s32 offset:236 ; 4-byte Folded Spill
+; SI-NEXT: v_lshrrev_b32_e32 v11, 8, v62
+; SI-NEXT: v_lshrrev_b32_e32 v20, 8, v55
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_lshrrev_b32_e32 v8, 24, v32
-; SI-NEXT: v_lshrrev_b32_e32 v32, 8, v18
-; SI-NEXT: buffer_store_dword v8, off, s[0:3], s32 offset:196 ; 4-byte Folded Spill
+; SI-NEXT: v_lshrrev_b32_e32 v18, 8, v27
+; SI-NEXT: s_waitcnt vmcnt(10)
+; SI-NEXT: v_and_b32_e32 v48, 0xffff0000, v23
+; SI-NEXT: v_add_f32_e32 v48, 0x40c00000, v48
+; SI-NEXT: v_lshrrev_b32_e32 v48, 16, v48
+; SI-NEXT: v_alignbit_b32 v36, v48, v36, 16
+; SI-NEXT: buffer_load_dword v48, off, s[0:3], s32 offset:500 ; 4-byte Folded Reload
+; SI-NEXT: s_waitcnt vmcnt(10)
+; SI-NEXT: v_and_b32_e32 v32, 0xffff0000, v32
+; SI-NEXT: v_add_f32_e32 v32, 0x40c00000, v32
+; SI-NEXT: s_waitcnt vmcnt(0)
+; SI-NEXT: v_and_b32_e32 v48, 0xffff0000, v48
+; SI-NEXT: v_add_f32_e32 v40, 0x40c00000, v48
+; SI-NEXT: v_lshrrev_b32_e32 v15, 16, v40
+; SI-NEXT: v_alignbit_b32 v50, v15, v32, 16
+; SI-NEXT: buffer_load_dword v32, off, s[0:3], s32 offset:532 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v48, off, s[0:3], s32 offset:528 ; 4-byte Folded Reload
+; SI-NEXT: buffer_store_dword v15, off, s[0:3], s32 offset:468 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_lshrrev_b32_e32 v8, 24, v31
-; SI-NEXT: v_lshrrev_b32_e32 v31, 8, v62
-; SI-NEXT: buffer_store_dword v8, off, s[0:3], s32 offset:180 ; 4-byte Folded Spill
+; SI-NEXT: v_lshrrev_b32_e32 v15, 24, v46
+; SI-NEXT: buffer_store_dword v15, off, s[0:3], s32 offset:464 ; 4-byte Folded Spill
+; SI-NEXT: v_alignbit_b32 v46, v35, v13, 8
+; SI-NEXT: v_lshrrev_b32_e32 v59, 8, v50
+; SI-NEXT: s_waitcnt vmcnt(3)
+; SI-NEXT: v_and_b32_e32 v32, 0xffff0000, v32
+; SI-NEXT: s_waitcnt vmcnt(2)
+; SI-NEXT: v_and_b32_e32 v48, 0xffff0000, v48
+; SI-NEXT: v_add_f32_e32 v42, 0x40c00000, v48
+; SI-NEXT: v_add_f32_e32 v32, 0x40c00000, v32
+; SI-NEXT: v_lshrrev_b32_e32 v22, 16, v42
+; SI-NEXT: v_alignbit_b32 v47, v22, v32, 16
+; SI-NEXT: buffer_load_dword v32, off, s[0:3], s32 offset:524 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v48, off, s[0:3], s32 offset:520 ; 4-byte Folded Reload
+; SI-NEXT: s_waitcnt vmcnt(1)
+; SI-NEXT: v_and_b32_e32 v32, 0xffff0000, v32
+; SI-NEXT: s_waitcnt vmcnt(0)
+; SI-NEXT: v_and_b32_e32 v48, 0xffff0000, v48
+; SI-NEXT: v_add_f32_e32 v43, 0x40c00000, v48
+; SI-NEXT: v_add_f32_e32 v32, 0x40c00000, v32
+; SI-NEXT: v_lshrrev_b32_e32 v23, 16, v43
+; SI-NEXT: v_alignbit_b32 v58, v23, v32, 16
+; SI-NEXT: buffer_load_dword v32, off, s[0:3], s32 offset:516 ; 4-byte Folded Reload
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_lshrrev_b32_e32 v8, 24, v11
-; SI-NEXT: v_lshrrev_b32_e32 v11, 8, v58
-; SI-NEXT: buffer_store_dword v8, off, s[0:3], s32 offset:172 ; 4-byte Folded Spill
+; SI-NEXT: v_lshrrev_b32_e32 v15, 24, v43
+; SI-NEXT: buffer_store_dword v15, off, s[0:3], s32 offset:460 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_lshrrev_b32_e32 v8, 24, v26
-; SI-NEXT: v_alignbit_b32 v26, v24, v60, 16
-; SI-NEXT: buffer_store_dword v8, off, s[0:3], s32 offset:168 ; 4-byte Folded Spill
-; SI-NEXT: s_waitcnt vmcnt(14)
-; SI-NEXT: v_and_b32_e32 v14, 0xffff0000, v14
-; SI-NEXT: v_add_f32_e32 v14, 0x40c00000, v14
-; SI-NEXT: v_alignbit_b32 v44, v19, v14, 16
-; SI-NEXT: v_lshrrev_b32_e32 v14, 24, v59
+; SI-NEXT: v_lshrrev_b32_e32 v15, 24, v42
+; SI-NEXT: buffer_store_dword v15, off, s[0:3], s32 offset:456 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
+; SI-NEXT: v_lshrrev_b32_e32 v15, 24, v40
+; SI-NEXT: v_lshrrev_b32_e32 v34, 8, v58
+; SI-NEXT: buffer_store_dword v15, off, s[0:3], s32 offset:448 ; 4-byte Folded Spill
+; SI-NEXT: s_waitcnt vmcnt(3)
+; SI-NEXT: v_and_b32_e32 v32, 0xffff0000, v32
+; SI-NEXT: v_add_f32_e32 v32, 0x40c00000, v32
+; SI-NEXT: v_alignbit_b32 v44, v16, v32, 16
; SI-NEXT: v_alignbit_b32 v8, v44, v36, 24
-; SI-NEXT: buffer_store_dword v8, off, s[0:3], s32 offset:396 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v8, off, s[0:3], s32 offset:436 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_alignbit_b32 v8, v44, v36, 16
-; SI-NEXT: buffer_store_dword v8, off, s[0:3], s32 offset:408 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v8, off, s[0:3], s32 offset:444 ; 4-byte Folded Spill
+; SI-NEXT: s_waitcnt expcnt(0)
+; SI-NEXT: v_alignbit_b32 v8, v58, v33, 24
+; SI-NEXT: buffer_store_dword v8, off, s[0:3], s32 offset:412 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_alignbit_b32 v8, v44, v36, 8
+; SI-NEXT: v_alignbit_b32 v8, v58, v33, 16
; SI-NEXT: buffer_store_dword v8, off, s[0:3], s32 offset:420 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_alignbit_b32 v8, v58, v6, 24
-; SI-NEXT: buffer_store_dword v8, off, s[0:3], s32 offset:384 ; 4-byte Folded Spill
+; SI-NEXT: v_alignbit_b32 v8, v58, v33, 8
+; SI-NEXT: buffer_store_dword v8, off, s[0:3], s32 offset:432 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_alignbit_b32 v8, v58, v6, 16
-; SI-NEXT: buffer_store_dword v8, off, s[0:3], s32 offset:392 ; 4-byte Folded Spill
+; SI-NEXT: v_alignbit_b32 v8, v47, v28, 24
+; SI-NEXT: buffer_store_dword v8, off, s[0:3], s32 offset:396 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_alignbit_b32 v8, v58, v6, 8
+; SI-NEXT: v_alignbit_b32 v8, v47, v28, 16
; SI-NEXT: buffer_store_dword v8, off, s[0:3], s32 offset:400 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_alignbit_b32 v8, v47, v2, 24
-; SI-NEXT: buffer_store_dword v8, off, s[0:3], s32 offset:368 ; 4-byte Folded Spill
+; SI-NEXT: v_alignbit_b32 v8, v47, v28, 8
+; SI-NEXT: buffer_store_dword v8, off, s[0:3], s32 offset:408 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_alignbit_b32 v8, v47, v2, 16
+; SI-NEXT: v_alignbit_b32 v8, v50, v25, 24
; SI-NEXT: buffer_store_dword v8, off, s[0:3], s32 offset:372 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_alignbit_b32 v8, v47, v2, 8
+; SI-NEXT: v_alignbit_b32 v8, v50, v25, 16
; SI-NEXT: buffer_store_dword v8, off, s[0:3], s32 offset:380 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_alignbit_b32 v8, v50, v5, 24
-; SI-NEXT: buffer_store_dword v8, off, s[0:3], s32 offset:344 ; 4-byte Folded Spill
+; SI-NEXT: v_alignbit_b32 v8, v50, v25, 8
+; SI-NEXT: buffer_store_dword v8, off, s[0:3], s32 offset:392 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_alignbit_b32 v8, v50, v5, 16
+; SI-NEXT: v_alignbit_b32 v8, v62, v53, 24
; SI-NEXT: buffer_store_dword v8, off, s[0:3], s32 offset:356 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_alignbit_b32 v8, v50, v5, 8
+; SI-NEXT: v_alignbit_b32 v8, v62, v53, 16
; SI-NEXT: buffer_store_dword v8, off, s[0:3], s32 offset:364 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_alignbit_b32 v8, v62, v4, 24
-; SI-NEXT: buffer_store_dword v8, off, s[0:3], s32 offset:328 ; 4-byte Folded Spill
+; SI-NEXT: v_alignbit_b32 v8, v62, v53, 8
+; SI-NEXT: buffer_store_dword v8, off, s[0:3], s32 offset:368 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_alignbit_b32 v8, v62, v4, 16
-; SI-NEXT: buffer_store_dword v8, off, s[0:3], s32 offset:336 ; 4-byte Folded Spill
+; SI-NEXT: v_alignbit_b32 v8, v55, v19, 24
+; SI-NEXT: buffer_store_dword v8, off, s[0:3], s32 offset:340 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_alignbit_b32 v8, v55, v3, 24
-; SI-NEXT: buffer_store_dword v8, off, s[0:3], s32 offset:312 ; 4-byte Folded Spill
+; SI-NEXT: v_alignbit_b32 v8, v55, v19, 16
+; SI-NEXT: buffer_store_dword v8, off, s[0:3], s32 offset:344 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_alignbit_b32 v8, v55, v3, 16
-; SI-NEXT: buffer_store_dword v8, off, s[0:3], s32 offset:316 ; 4-byte Folded Spill
+; SI-NEXT: v_alignbit_b32 v8, v55, v19, 8
+; SI-NEXT: buffer_store_dword v8, off, s[0:3], s32 offset:352 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_alignbit_b32 v8, v55, v3, 8
+; SI-NEXT: v_alignbit_b32 v8, v38, v57, 24
; SI-NEXT: buffer_store_dword v8, off, s[0:3], s32 offset:324 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
+; SI-NEXT: v_alignbit_b32 v8, v38, v57, 16
+; SI-NEXT: buffer_store_dword v8, off, s[0:3], s32 offset:328 ; 4-byte Folded Spill
+; SI-NEXT: s_waitcnt expcnt(0)
+; SI-NEXT: v_alignbit_b32 v8, v38, v57, 8
+; SI-NEXT: buffer_store_dword v8, off, s[0:3], s32 offset:332 ; 4-byte Folded Spill
+; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_alignbit_b32 v8, v35, v13, 24
-; SI-NEXT: buffer_store_dword v8, off, s[0:3], s32 offset:292 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v8, off, s[0:3], s32 offset:300 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_alignbit_b32 v8, v35, v13, 16
-; SI-NEXT: buffer_store_dword v8, off, s[0:3], s32 offset:296 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v8, off, s[0:3], s32 offset:304 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_alignbit_b32 v8, v30, v10, 24
-; SI-NEXT: buffer_store_dword v8, off, s[0:3], s32 offset:276 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v8, off, s[0:3], s32 offset:284 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_alignbit_b32 v8, v30, v10, 16
-; SI-NEXT: buffer_store_dword v8, off, s[0:3], s32 offset:284 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v8, off, s[0:3], s32 offset:292 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_alignbit_b32 v8, v30, v10, 8
-; SI-NEXT: buffer_store_dword v8, off, s[0:3], s32 offset:288 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v8, off, s[0:3], s32 offset:296 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_alignbit_b32 v8, v27, v7, 24
-; SI-NEXT: buffer_store_dword v8, off, s[0:3], s32 offset:256 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v8, off, s[0:3], s32 offset:264 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_alignbit_b32 v8, v27, v7, 16
-; SI-NEXT: buffer_store_dword v8, off, s[0:3], s32 offset:264 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v8, off, s[0:3], s32 offset:272 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_alignbit_b32 v8, v27, v7, 8
-; SI-NEXT: buffer_store_dword v8, off, s[0:3], s32 offset:268 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v8, off, s[0:3], s32 offset:276 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_alignbit_b32 v8, v24, v60, 24
-; SI-NEXT: buffer_store_dword v8, off, s[0:3], s32 offset:236 ; 4-byte Folded Spill
+; SI-NEXT: v_alignbit_b32 v8, v37, v6, 24
+; SI-NEXT: buffer_store_dword v8, off, s[0:3], s32 offset:248 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_alignbit_b32 v8, v24, v60, 8
-; SI-NEXT: buffer_store_dword v8, off, s[0:3], s32 offset:260 ; 4-byte Folded Spill
+; SI-NEXT: v_alignbit_b32 v8, v37, v6, 16
+; SI-NEXT: buffer_store_dword v8, off, s[0:3], s32 offset:256 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_alignbit_b32 v8, v21, v57, 24
-; SI-NEXT: buffer_store_dword v8, off, s[0:3], s32 offset:220 ; 4-byte Folded Spill
+; SI-NEXT: v_alignbit_b32 v8, v37, v6, 8
+; SI-NEXT: buffer_store_dword v8, off, s[0:3], s32 offset:268 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_alignbit_b32 v8, v21, v57, 16
+; SI-NEXT: v_alignbit_b32 v8, v21, v5, 24
; SI-NEXT: buffer_store_dword v8, off, s[0:3], s32 offset:232 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_alignbit_b32 v8, v21, v57, 8
-; SI-NEXT: buffer_store_dword v8, off, s[0:3], s32 offset:240 ; 4-byte Folded Spill
-; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_alignbit_b32 v8, v18, v51, 24
-; SI-NEXT: buffer_store_dword v8, off, s[0:3], s32 offset:208 ; 4-byte Folded Spill
+; SI-NEXT: v_alignbit_b32 v8, v21, v5, 16
+; SI-NEXT: buffer_store_dword v8, off, s[0:3], s32 offset:244 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_alignbit_b32 v8, v18, v51, 16
-; SI-NEXT: buffer_store_dword v8, off, s[0:3], s32 offset:216 ; 4-byte Folded Spill
-; SI-NEXT: buffer_store_dword v51, off, s[0:3], s32 offset:440 ; 4-byte Folded Spill
-; SI-NEXT: s_waitcnt expcnt(1)
-; SI-NEXT: v_alignbit_b32 v8, v18, v51, 8
-; SI-NEXT: buffer_store_dword v32, off, s[0:3], s32 offset:332 ; 4-byte Folded Spill
+; SI-NEXT: v_alignbit_b32 v8, v21, v5, 8
+; SI-NEXT: buffer_store_dword v8, off, s[0:3], s32 offset:252 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_lshrrev_b32_e32 v32, 8, v12
-; SI-NEXT: v_lshrrev_b32_e32 v51, 8, v44
-; SI-NEXT: buffer_store_dword v29, off, s[0:3], s32 offset:388 ; 4-byte Folded Spill
+; SI-NEXT: v_alignbit_b32 v8, v61, v4, 24
+; SI-NEXT: buffer_store_dword v8, off, s[0:3], s32 offset:220 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_lshrrev_b32_e32 v29, 8, v21
+; SI-NEXT: v_alignbit_b32 v8, v61, v4, 16
; SI-NEXT: buffer_store_dword v8, off, s[0:3], s32 offset:228 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_alignbit_b32 v8, v63, v34, 24
-; SI-NEXT: buffer_store_dword v32, off, s[0:3], s32 offset:300 ; 4-byte Folded Spill
-; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_lshrrev_b32_e32 v32, 8, v9
-; SI-NEXT: buffer_store_dword v29, off, s[0:3], s32 offset:360 ; 4-byte Folded Spill
-; SI-NEXT: buffer_store_dword v8, off, s[0:3], s32 offset:192 ; 4-byte Folded Spill
+; SI-NEXT: v_alignbit_b32 v8, v61, v4, 8
+; SI-NEXT: buffer_store_dword v8, off, s[0:3], s32 offset:240 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_alignbit_b32 v8, v63, v34, 16
-; SI-NEXT: buffer_store_dword v32, off, s[0:3], s32 offset:280 ; 4-byte Folded Spill
+; SI-NEXT: v_alignbit_b32 v8, v63, v3, 24
; SI-NEXT: buffer_store_dword v8, off, s[0:3], s32 offset:204 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_alignbit_b32 v8, v63, v34, 8
-; SI-NEXT: buffer_store_dword v8, off, s[0:3], s32 offset:212 ; 4-byte Folded Spill
+; SI-NEXT: v_alignbit_b32 v8, v63, v3, 16
+; SI-NEXT: buffer_store_dword v8, off, s[0:3], s32 offset:216 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_alignbit_b32 v8, v12, v52, 24
-; SI-NEXT: buffer_store_dword v8, off, s[0:3], s32 offset:244 ; 4-byte Folded Spill
+; SI-NEXT: v_alignbit_b32 v8, v63, v3, 8
+; SI-NEXT: buffer_store_dword v8, off, s[0:3], s32 offset:224 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_alignbit_b32 v8, v12, v52, 16
-; SI-NEXT: buffer_store_dword v8, off, s[0:3], s32 offset:436 ; 4-byte Folded Spill
+; SI-NEXT: v_alignbit_b32 v8, v12, v56, 24
+; SI-NEXT: buffer_store_dword v8, off, s[0:3], s32 offset:192 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_alignbit_b32 v8, v12, v52, 8
+; SI-NEXT: v_alignbit_b32 v8, v12, v56, 16
; SI-NEXT: buffer_store_dword v8, off, s[0:3], s32 offset:200 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
+; SI-NEXT: v_alignbit_b32 v8, v12, v56, 8
+; SI-NEXT: buffer_store_dword v8, off, s[0:3], s32 offset:212 ; 4-byte Folded Spill
+; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_alignbit_b32 v8, v9, v1, 24
-; SI-NEXT: buffer_store_dword v8, off, s[0:3], s32 offset:176 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v8, off, s[0:3], s32 offset:180 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_alignbit_b32 v8, v9, v1, 16
-; SI-NEXT: buffer_store_dword v8, off, s[0:3], s32 offset:184 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v8, off, s[0:3], s32 offset:188 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_alignbit_b32 v8, v9, v1, 8
-; SI-NEXT: buffer_store_dword v8, off, s[0:3], s32 offset:308 ; 4-byte Folded Spill
-; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_lshrrev_b32_e32 v8, 8, v38
-; SI-NEXT: buffer_store_dword v8, off, s[0:3], s32 offset:444 ; 4-byte Folded Spill
-; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_lshrrev_b32_e32 v8, 8, v30
-; SI-NEXT: buffer_store_dword v8, off, s[0:3], s32 offset:428 ; 4-byte Folded Spill
+; SI-NEXT: v_alignbit_b32 v52, v44, v36, 8
+; SI-NEXT: buffer_store_dword v8, off, s[0:3], s32 offset:196 ; 4-byte Folded Spill
+; SI-NEXT: v_lshrrev_b32_e32 v26, 8, v44
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_lshrrev_b32_e32 v8, 8, v27
+; SI-NEXT: v_lshrrev_b32_e32 v8, 8, v47
; SI-NEXT: .LBB91_5: ; %end
-; SI-NEXT: buffer_load_dword v28, off, s[0:3], s32 offset:420 ; 4-byte Folded Reload
-; SI-NEXT: buffer_load_dword v29, off, s[0:3], s32 offset:408 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v15, off, s[0:3], s32 offset:444 ; 4-byte Folded Reload
+; SI-NEXT: s_waitcnt expcnt(0)
+; SI-NEXT: buffer_load_dword v17, off, s[0:3], s32 offset:436 ; 4-byte Folded Reload
; SI-NEXT: v_and_b32_e32 v36, 0xff, v36
-; SI-NEXT: v_lshlrev_b32_e32 v14, 24, v14
+; SI-NEXT: v_lshlrev_b32_e32 v52, 8, v52
+; SI-NEXT: v_or_b32_e32 v36, v36, v52
+; SI-NEXT: v_and_b32_e32 v36, 0xffff, v36
+; SI-NEXT: s_waitcnt vmcnt(3)
+; SI-NEXT: v_lshlrev_b32_e32 v39, 8, v26
+; SI-NEXT: v_and_b32_e32 v33, 0xff, v33
+; SI-NEXT: v_lshlrev_b32_e32 v31, 8, v34
+; SI-NEXT: v_lshlrev_b32_e32 v8, 8, v8
; SI-NEXT: v_lshlrev_b32_e32 v11, 8, v11
-; SI-NEXT: v_and_b32_e32 v10, 0xff, v10
; SI-NEXT: v_and_b32_e32 v7, 0xff, v7
-; SI-NEXT: v_lshlrev_b32_e32 v8, 8, v8
+; SI-NEXT: v_and_b32_e32 v6, 0xff, v6
+; SI-NEXT: v_and_b32_e32 v5, 0xff, v5
+; SI-NEXT: v_and_b32_e32 v4, 0xff, v4
+; SI-NEXT: v_and_b32_e32 v3, 0xff, v3
; SI-NEXT: v_and_b32_e32 v1, 0xff, v1
; SI-NEXT: s_waitcnt vmcnt(1)
-; SI-NEXT: v_lshlrev_b32_e32 v32, 8, v28
-; SI-NEXT: v_or_b32_e32 v32, v36, v32
-; SI-NEXT: s_waitcnt vmcnt(0)
-; SI-NEXT: v_and_b32_e32 v36, 0xff, v29
-; SI-NEXT: buffer_load_dword v29, off, s[0:3], s32 offset:396 ; 4-byte Folded Reload
-; SI-NEXT: v_lshlrev_b32_e32 v36, 16, v36
-; SI-NEXT: v_and_b32_e32 v32, 0xffff, v32
+; SI-NEXT: v_and_b32_e32 v52, 0xff, v15
+; SI-NEXT: v_lshlrev_b32_e32 v52, 16, v52
; SI-NEXT: s_waitcnt vmcnt(0)
-; SI-NEXT: v_lshlrev_b32_e32 v56, 24, v29
-; SI-NEXT: v_or_b32_e32 v36, v56, v36
-; SI-NEXT: v_or_b32_e32 v32, v32, v36
-; SI-NEXT: buffer_store_dword v32, v0, s[0:3], 0 offen
-; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_and_b32_e32 v32, 0xff, v44
-; SI-NEXT: v_lshlrev_b32_e32 v36, 8, v51
-; SI-NEXT: v_or_b32_e32 v32, v32, v36
-; SI-NEXT: v_and_b32_e32 v36, 0xff, v19
-; SI-NEXT: v_lshlrev_b32_e32 v36, 16, v36
-; SI-NEXT: v_or_b32_e32 v14, v14, v36
-; SI-NEXT: v_and_b32_e32 v32, 0xffff, v32
-; SI-NEXT: v_or_b32_e32 v14, v32, v14
-; SI-NEXT: v_add_i32_e32 v32, vcc, 4, v0
-; SI-NEXT: buffer_store_dword v14, v32, s[0:3], 0 offen
-; SI-NEXT: buffer_load_dword v19, off, s[0:3], s32 offset:400 ; 4-byte Folded Reload
+; SI-NEXT: v_lshlrev_b32_e32 v51, 24, v17
+; SI-NEXT: v_or_b32_e32 v51, v51, v52
+; SI-NEXT: v_or_b32_e32 v36, v36, v51
+; SI-NEXT: buffer_store_dword v36, v0, s[0:3], 0 offen
+; SI-NEXT: buffer_load_dword v15, off, s[0:3], s32 offset:464 ; 4-byte Folded Reload
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_and_b32_e32 v14, 0xff, v6
-; SI-NEXT: s_waitcnt vmcnt(0)
-; SI-NEXT: v_lshlrev_b32_e32 v32, 8, v19
-; SI-NEXT: buffer_load_dword v19, off, s[0:3], s32 offset:392 ; 4-byte Folded Reload
-; SI-NEXT: v_or_b32_e32 v14, v14, v32
-; SI-NEXT: v_and_b32_e32 v14, 0xffff, v14
+; SI-NEXT: v_and_b32_e32 v36, 0xff, v44
+; SI-NEXT: v_or_b32_e32 v36, v36, v39
+; SI-NEXT: v_and_b32_e32 v39, 0xff, v16
+; SI-NEXT: v_lshlrev_b32_e32 v39, 16, v39
+; SI-NEXT: v_and_b32_e32 v36, 0xffff, v36
; SI-NEXT: s_waitcnt vmcnt(0)
-; SI-NEXT: v_and_b32_e32 v32, 0xff, v19
-; SI-NEXT: buffer_load_dword v19, off, s[0:3], s32 offset:384 ; 4-byte Folded Reload
-; SI-NEXT: v_lshlrev_b32_e32 v32, 16, v32
+; SI-NEXT: v_lshlrev_b32_e32 v51, 24, v15
+; SI-NEXT: v_or_b32_e32 v39, v51, v39
+; SI-NEXT: v_or_b32_e32 v36, v36, v39
+; SI-NEXT: v_add_i32_e32 v39, vcc, 4, v0
+; SI-NEXT: buffer_store_dword v36, v39, s[0:3], 0 offen
+; SI-NEXT: buffer_load_dword v17, off, s[0:3], s32 offset:432 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v26, off, s[0:3], s32 offset:420 ; 4-byte Folded Reload
+; SI-NEXT: s_waitcnt vmcnt(1)
+; SI-NEXT: v_lshlrev_b32_e32 v17, 8, v17
+; SI-NEXT: v_or_b32_e32 v17, v33, v17
; SI-NEXT: s_waitcnt vmcnt(0)
-; SI-NEXT: v_lshlrev_b32_e32 v33, 24, v19
-; SI-NEXT: v_or_b32_e32 v32, v33, v32
-; SI-NEXT: v_or_b32_e32 v14, v14, v32
-; SI-NEXT: v_add_i32_e32 v32, vcc, 8, v0
-; SI-NEXT: buffer_store_dword v14, v32, s[0:3], 0 offen
-; SI-NEXT: buffer_load_dword v6, off, s[0:3], s32 offset:448 ; 4-byte Folded Reload
-; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_and_b32_e32 v14, 0xff, v58
-; SI-NEXT: v_or_b32_e32 v11, v14, v11
-; SI-NEXT: v_and_b32_e32 v14, 0xff, v22
-; SI-NEXT: v_lshlrev_b32_e32 v14, 16, v14
-; SI-NEXT: v_and_b32_e32 v11, 0xffff, v11
+; SI-NEXT: v_and_b32_e32 v33, 0xff, v26
+; SI-NEXT: buffer_load_dword v26, off, s[0:3], s32 offset:412 ; 4-byte Folded Reload
+; SI-NEXT: v_lshlrev_b32_e32 v33, 16, v33
+; SI-NEXT: v_and_b32_e32 v17, 0xffff, v17
+; SI-NEXT: s_waitcnt vmcnt(0) expcnt(0)
+; SI-NEXT: v_lshlrev_b32_e32 v36, 24, v26
+; SI-NEXT: v_or_b32_e32 v33, v36, v33
+; SI-NEXT: v_or_b32_e32 v17, v17, v33
+; SI-NEXT: v_add_i32_e32 v33, vcc, 8, v0
+; SI-NEXT: buffer_store_dword v17, v33, s[0:3], 0 offen
+; SI-NEXT: buffer_load_dword v15, off, s[0:3], s32 offset:460 ; 4-byte Folded Reload
+; SI-NEXT: s_waitcnt expcnt(0)
+; SI-NEXT: v_and_b32_e32 v17, 0xff, v58
+; SI-NEXT: v_or_b32_e32 v17, v17, v31
+; SI-NEXT: v_and_b32_e32 v31, 0xff, v23
+; SI-NEXT: v_lshlrev_b32_e32 v31, 16, v31
+; SI-NEXT: v_and_b32_e32 v17, 0xffff, v17
; SI-NEXT: s_waitcnt vmcnt(0)
-; SI-NEXT: v_lshlrev_b32_e32 v32, 24, v6
-; SI-NEXT: v_or_b32_e32 v14, v32, v14
-; SI-NEXT: v_or_b32_e32 v11, v11, v14
-; SI-NEXT: v_add_i32_e32 v14, vcc, 12, v0
-; SI-NEXT: buffer_store_dword v11, v14, s[0:3], 0 offen
-; SI-NEXT: buffer_load_dword v14, off, s[0:3], s32 offset:380 ; 4-byte Folded Reload
+; SI-NEXT: v_lshlrev_b32_e32 v32, 24, v15
+; SI-NEXT: v_or_b32_e32 v31, v32, v31
+; SI-NEXT: v_or_b32_e32 v17, v17, v31
+; SI-NEXT: v_add_i32_e32 v31, vcc, 12, v0
+; SI-NEXT: buffer_store_dword v17, v31, s[0:3], 0 offen
+; SI-NEXT: buffer_load_dword v23, off, s[0:3], s32 offset:408 ; 4-byte Folded Reload
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_and_b32_e32 v11, 0xff, v2
-; SI-NEXT: v_and_b32_e32 v6, 0xff, v60
-; SI-NEXT: buffer_load_dword v19, off, s[0:3], s32 offset:368 ; 4-byte Folded Reload
-; SI-NEXT: s_waitcnt vmcnt(1)
-; SI-NEXT: v_lshlrev_b32_e32 v14, 8, v14
-; SI-NEXT: v_or_b32_e32 v11, v11, v14
-; SI-NEXT: buffer_load_dword v14, off, s[0:3], s32 offset:372 ; 4-byte Folded Reload
-; SI-NEXT: v_and_b32_e32 v11, 0xffff, v11
-; SI-NEXT: s_waitcnt vmcnt(1)
-; SI-NEXT: v_lshlrev_b32_e32 v28, 24, v19
+; SI-NEXT: v_and_b32_e32 v17, 0xff, v28
; SI-NEXT: s_waitcnt vmcnt(0)
-; SI-NEXT: v_and_b32_e32 v14, 0xff, v14
-; SI-NEXT: v_lshlrev_b32_e32 v14, 16, v14
-; SI-NEXT: v_or_b32_e32 v14, v28, v14
-; SI-NEXT: v_or_b32_e32 v11, v11, v14
-; SI-NEXT: v_add_i32_e32 v14, vcc, 16, v0
-; SI-NEXT: buffer_store_dword v11, v14, s[0:3], 0 offen
-; SI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:432 ; 4-byte Folded Reload
+; SI-NEXT: v_lshlrev_b32_e32 v28, 8, v23
+; SI-NEXT: buffer_load_dword v23, off, s[0:3], s32 offset:400 ; 4-byte Folded Reload
+; SI-NEXT: v_or_b32_e32 v17, v17, v28
+; SI-NEXT: v_and_b32_e32 v17, 0xffff, v17
+; SI-NEXT: s_waitcnt vmcnt(0)
+; SI-NEXT: v_and_b32_e32 v28, 0xff, v23
+; SI-NEXT: buffer_load_dword v23, off, s[0:3], s32 offset:396 ; 4-byte Folded Reload
+; SI-NEXT: v_lshlrev_b32_e32 v28, 16, v28
+; SI-NEXT: s_waitcnt vmcnt(0)
+; SI-NEXT: v_lshlrev_b32_e32 v31, 24, v23
+; SI-NEXT: v_or_b32_e32 v28, v31, v28
+; SI-NEXT: v_or_b32_e32 v17, v17, v28
+; SI-NEXT: v_add_i32_e32 v28, vcc, 16, v0
+; SI-NEXT: buffer_store_dword v17, v28, s[0:3], 0 offen
+; SI-NEXT: buffer_load_dword v15, off, s[0:3], s32 offset:456 ; 4-byte Folded Reload
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_and_b32_e32 v11, 0xff, v47
-; SI-NEXT: v_lshlrev_b32_e32 v14, 8, v49
-; SI-NEXT: v_or_b32_e32 v11, v11, v14
-; SI-NEXT: v_and_b32_e32 v14, 0xff, v25
-; SI-NEXT: v_lshlrev_b32_e32 v14, 16, v14
-; SI-NEXT: v_and_b32_e32 v11, 0xffff, v11
+; SI-NEXT: v_and_b32_e32 v17, 0xff, v47
+; SI-NEXT: v_or_b32_e32 v8, v17, v8
+; SI-NEXT: v_and_b32_e32 v17, 0xff, v22
+; SI-NEXT: v_lshlrev_b32_e32 v17, 16, v17
+; SI-NEXT: v_and_b32_e32 v8, 0xffff, v8
; SI-NEXT: s_waitcnt vmcnt(0)
-; SI-NEXT: v_lshlrev_b32_e32 v28, 24, v2
-; SI-NEXT: v_or_b32_e32 v14, v28, v14
-; SI-NEXT: v_or_b32_e32 v11, v11, v14
-; SI-NEXT: v_add_i32_e32 v14, vcc, 20, v0
-; SI-NEXT: buffer_store_dword v11, v14, s[0:3], 0 offen
-; SI-NEXT: buffer_load_dword v14, off, s[0:3], s32 offset:364 ; 4-byte Folded Reload
+; SI-NEXT: v_lshlrev_b32_e32 v28, 24, v15
+; SI-NEXT: v_or_b32_e32 v17, v28, v17
+; SI-NEXT: v_or_b32_e32 v8, v8, v17
+; SI-NEXT: v_add_i32_e32 v17, vcc, 20, v0
+; SI-NEXT: buffer_store_dword v8, v17, s[0:3], 0 offen
+; SI-NEXT: buffer_load_dword v17, off, s[0:3], s32 offset:392 ; 4-byte Folded Reload
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_and_b32_e32 v11, 0xff, v5
-; SI-NEXT: v_and_b32_e32 v5, 0xff, v57
-; SI-NEXT: buffer_load_dword v19, off, s[0:3], s32 offset:344 ; 4-byte Folded Reload
+; SI-NEXT: v_and_b32_e32 v8, 0xff, v25
+; SI-NEXT: buffer_load_dword v25, off, s[0:3], s32 offset:372 ; 4-byte Folded Reload
; SI-NEXT: s_waitcnt vmcnt(1)
-; SI-NEXT: v_lshlrev_b32_e32 v14, 8, v14
-; SI-NEXT: v_or_b32_e32 v11, v11, v14
-; SI-NEXT: buffer_load_dword v14, off, s[0:3], s32 offset:356 ; 4-byte Folded Reload
-; SI-NEXT: v_and_b32_e32 v11, 0xffff, v11
+; SI-NEXT: v_lshlrev_b32_e32 v17, 8, v17
+; SI-NEXT: v_or_b32_e32 v8, v8, v17
+; SI-NEXT: buffer_load_dword v17, off, s[0:3], s32 offset:380 ; 4-byte Folded Reload
+; SI-NEXT: v_and_b32_e32 v8, 0xffff, v8
; SI-NEXT: s_waitcnt vmcnt(1)
-; SI-NEXT: v_lshlrev_b32_e32 v25, 24, v19
+; SI-NEXT: v_lshlrev_b32_e32 v25, 24, v25
; SI-NEXT: s_waitcnt vmcnt(0)
-; SI-NEXT: v_and_b32_e32 v14, 0xff, v14
-; SI-NEXT: v_lshlrev_b32_e32 v14, 16, v14
-; SI-NEXT: v_or_b32_e32 v14, v25, v14
-; SI-NEXT: v_or_b32_e32 v11, v11, v14
-; SI-NEXT: v_add_i32_e32 v14, vcc, 24, v0
-; SI-NEXT: buffer_store_dword v11, v14, s[0:3], 0 offen
-; SI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:424 ; 4-byte Folded Reload
+; SI-NEXT: v_and_b32_e32 v17, 0xff, v17
+; SI-NEXT: v_lshlrev_b32_e32 v17, 16, v17
+; SI-NEXT: v_or_b32_e32 v17, v25, v17
+; SI-NEXT: v_or_b32_e32 v8, v8, v17
+; SI-NEXT: v_add_i32_e32 v17, vcc, 24, v0
+; SI-NEXT: buffer_store_dword v8, v17, s[0:3], 0 offen
+; SI-NEXT: buffer_load_dword v15, off, s[0:3], s32 offset:468 ; 4-byte Folded Reload
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_and_b32_e32 v11, 0xff, v50
-; SI-NEXT: v_lshlrev_b32_e32 v14, 8, v39
-; SI-NEXT: v_or_b32_e32 v11, v11, v14
-; SI-NEXT: v_and_b32_e32 v14, 0xff, v17
-; SI-NEXT: v_lshlrev_b32_e32 v14, 16, v14
-; SI-NEXT: v_and_b32_e32 v11, 0xffff, v11
+; SI-NEXT: v_and_b32_e32 v8, 0xff, v50
+; SI-NEXT: v_lshlrev_b32_e32 v17, 8, v59
+; SI-NEXT: v_or_b32_e32 v8, v8, v17
+; SI-NEXT: v_and_b32_e32 v8, 0xffff, v8
; SI-NEXT: s_waitcnt vmcnt(0)
-; SI-NEXT: v_lshlrev_b32_e32 v25, 24, v2
-; SI-NEXT: v_or_b32_e32 v14, v25, v14
-; SI-NEXT: v_or_b32_e32 v11, v11, v14
-; SI-NEXT: v_add_i32_e32 v14, vcc, 28, v0
-; SI-NEXT: buffer_store_dword v11, v14, s[0:3], 0 offen
-; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_and_b32_e32 v11, 0xff, v4
-; SI-NEXT: v_lshlrev_b32_e32 v14, 8, v23
-; SI-NEXT: v_or_b32_e32 v11, v11, v14
-; SI-NEXT: buffer_load_dword v14, off, s[0:3], s32 offset:336 ; 4-byte Folded Reload
-; SI-NEXT: buffer_load_dword v17, off, s[0:3], s32 offset:328 ; 4-byte Folded Reload
-; SI-NEXT: v_and_b32_e32 v11, 0xffff, v11
-; SI-NEXT: s_waitcnt vmcnt(1)
-; SI-NEXT: v_and_b32_e32 v14, 0xff, v14
-; SI-NEXT: v_lshlrev_b32_e32 v14, 16, v14
+; SI-NEXT: v_and_b32_e32 v17, 0xff, v15
+; SI-NEXT: buffer_load_dword v15, off, s[0:3], s32 offset:448 ; 4-byte Folded Reload
+; SI-NEXT: v_lshlrev_b32_e32 v17, 16, v17
; SI-NEXT: s_waitcnt vmcnt(0)
-; SI-NEXT: v_lshlrev_b32_e32 v22, 24, v17
-; SI-NEXT: v_or_b32_e32 v14, v22, v14
-; SI-NEXT: v_or_b32_e32 v11, v11, v14
-; SI-NEXT: v_add_i32_e32 v14, vcc, 32, v0
-; SI-NEXT: buffer_store_dword v11, v14, s[0:3], 0 offen
-; SI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:416 ; 4-byte Folded Reload
-; SI-NEXT: buffer_load_dword v17, off, s[0:3], s32 offset:404 ; 4-byte Folded Reload
+; SI-NEXT: v_lshlrev_b32_e32 v25, 24, v15
+; SI-NEXT: v_or_b32_e32 v17, v25, v17
+; SI-NEXT: v_or_b32_e32 v8, v8, v17
+; SI-NEXT: v_add_i32_e32 v17, vcc, 28, v0
+; SI-NEXT: buffer_store_dword v8, v17, s[0:3], 0 offen
+; SI-NEXT: buffer_load_dword v17, off, s[0:3], s32 offset:368 ; 4-byte Folded Reload
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_and_b32_e32 v11, 0xff, v62
-; SI-NEXT: v_lshlrev_b32_e32 v14, 8, v31
-; SI-NEXT: v_or_b32_e32 v11, v11, v14
-; SI-NEXT: v_and_b32_e32 v11, 0xffff, v11
+; SI-NEXT: v_and_b32_e32 v8, 0xff, v53
+; SI-NEXT: buffer_load_dword v22, off, s[0:3], s32 offset:356 ; 4-byte Folded Reload
; SI-NEXT: s_waitcnt vmcnt(1)
-; SI-NEXT: v_and_b32_e32 v14, 0xff, v2
-; SI-NEXT: v_lshlrev_b32_e32 v14, 16, v14
+; SI-NEXT: v_lshlrev_b32_e32 v17, 8, v17
+; SI-NEXT: v_or_b32_e32 v8, v8, v17
+; SI-NEXT: buffer_load_dword v17, off, s[0:3], s32 offset:364 ; 4-byte Folded Reload
+; SI-NEXT: v_and_b32_e32 v8, 0xffff, v8
+; SI-NEXT: s_waitcnt vmcnt(1)
+; SI-NEXT: v_lshlrev_b32_e32 v22, 24, v22
; SI-NEXT: s_waitcnt vmcnt(0)
-; SI-NEXT: v_lshlrev_b32_e32 v22, 24, v17
-; SI-NEXT: v_or_b32_e32 v14, v22, v14
-; SI-NEXT: v_or_b32_e32 v11, v11, v14
-; SI-NEXT: v_add_i32_e32 v14, vcc, 36, v0
-; SI-NEXT: buffer_store_dword v11, v14, s[0:3], 0 offen
-; SI-NEXT: buffer_load_dword v14, off, s[0:3], s32 offset:324 ; 4-byte Folded Reload
+; SI-NEXT: v_and_b32_e32 v17, 0xff, v17
+; SI-NEXT: v_lshlrev_b32_e32 v17, 16, v17
+; SI-NEXT: v_or_b32_e32 v17, v22, v17
+; SI-NEXT: v_or_b32_e32 v8, v8, v17
+; SI-NEXT: v_add_i32_e32 v17, vcc, 32, v0
+; SI-NEXT: buffer_store_dword v8, v17, s[0:3], 0 offen
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_and_b32_e32 v11, 0xff, v3
-; SI-NEXT: v_and_b32_e32 v3, 0xff, v34
-; SI-NEXT: buffer_load_dword v17, off, s[0:3], s32 offset:312 ; 4-byte Folded Reload
-; SI-NEXT: s_waitcnt vmcnt(1)
-; SI-NEXT: v_lshlrev_b32_e32 v14, 8, v14
-; SI-NEXT: v_or_b32_e32 v11, v11, v14
-; SI-NEXT: buffer_load_dword v14, off, s[0:3], s32 offset:316 ; 4-byte Folded Reload
-; SI-NEXT: v_and_b32_e32 v11, 0xffff, v11
+; SI-NEXT: v_and_b32_e32 v8, 0xff, v62
+; SI-NEXT: v_or_b32_e32 v8, v8, v11
+; SI-NEXT: buffer_load_dword v11, off, s[0:3], s32 offset:452 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v17, off, s[0:3], s32 offset:428 ; 4-byte Folded Reload
+; SI-NEXT: v_and_b32_e32 v8, 0xffff, v8
; SI-NEXT: s_waitcnt vmcnt(1)
-; SI-NEXT: v_lshlrev_b32_e32 v19, 24, v17
-; SI-NEXT: s_waitcnt vmcnt(0)
-; SI-NEXT: v_and_b32_e32 v14, 0xff, v14
-; SI-NEXT: v_lshlrev_b32_e32 v14, 16, v14
-; SI-NEXT: v_or_b32_e32 v14, v19, v14
-; SI-NEXT: v_or_b32_e32 v11, v11, v14
-; SI-NEXT: v_add_i32_e32 v14, vcc, 40, v0
-; SI-NEXT: buffer_store_dword v11, v14, s[0:3], 0 offen
-; SI-NEXT: buffer_load_dword v17, off, s[0:3], s32 offset:376 ; 4-byte Folded Reload
-; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_and_b32_e32 v11, 0xff, v55
-; SI-NEXT: v_lshlrev_b32_e32 v14, 8, v46
-; SI-NEXT: v_or_b32_e32 v11, v11, v14
-; SI-NEXT: v_and_b32_e32 v14, 0xff, v61
-; SI-NEXT: v_lshlrev_b32_e32 v14, 16, v14
-; SI-NEXT: v_and_b32_e32 v11, 0xffff, v11
+; SI-NEXT: v_and_b32_e32 v11, 0xff, v11
+; SI-NEXT: v_lshlrev_b32_e32 v11, 16, v11
; SI-NEXT: s_waitcnt vmcnt(0)
-; SI-NEXT: v_lshlrev_b32_e32 v19, 24, v17
-; SI-NEXT: v_or_b32_e32 v14, v19, v14
-; SI-NEXT: v_or_b32_e32 v11, v11, v14
-; SI-NEXT: v_add_i32_e32 v14, vcc, 44, v0
-; SI-NEXT: buffer_store_dword v11, v14, s[0:3], 0 offen
-; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_and_b32_e32 v11, 0xff, v16
-; SI-NEXT: v_lshlrev_b32_e32 v14, 8, v43
-; SI-NEXT: v_or_b32_e32 v11, v11, v14
-; SI-NEXT: v_and_b32_e32 v14, 0xff, v41
-; SI-NEXT: v_lshlrev_b32_e32 v14, 16, v14
-; SI-NEXT: v_lshlrev_b32_e32 v16, 24, v54
-; SI-NEXT: v_or_b32_e32 v14, v16, v14
-; SI-NEXT: v_and_b32_e32 v11, 0xffff, v11
-; SI-NEXT: v_or_b32_e32 v11, v11, v14
-; SI-NEXT: v_add_i32_e32 v14, vcc, 48, v0
-; SI-NEXT: buffer_store_dword v11, v14, s[0:3], 0 offen
-; SI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:444 ; 4-byte Folded Reload
+; SI-NEXT: v_lshlrev_b32_e32 v17, 24, v17
+; SI-NEXT: v_or_b32_e32 v11, v17, v11
+; SI-NEXT: v_or_b32_e32 v8, v8, v11
+; SI-NEXT: v_add_i32_e32 v11, vcc, 36, v0
+; SI-NEXT: buffer_store_dword v8, v11, s[0:3], 0 offen
+; SI-NEXT: buffer_load_dword v11, off, s[0:3], s32 offset:352 ; 4-byte Folded Reload
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_and_b32_e32 v11, 0xff, v38
-; SI-NEXT: buffer_load_dword v16, off, s[0:3], s32 offset:348 ; 4-byte Folded Reload
+; SI-NEXT: v_and_b32_e32 v8, 0xff, v19
+; SI-NEXT: buffer_load_dword v17, off, s[0:3], s32 offset:340 ; 4-byte Folded Reload
; SI-NEXT: s_waitcnt vmcnt(1)
-; SI-NEXT: v_lshlrev_b32_e32 v14, 8, v2
-; SI-NEXT: v_or_b32_e32 v11, v11, v14
-; SI-NEXT: buffer_load_dword v14, off, s[0:3], s32 offset:412 ; 4-byte Folded Reload
-; SI-NEXT: v_and_b32_e32 v11, 0xffff, v11
+; SI-NEXT: v_lshlrev_b32_e32 v11, 8, v11
+; SI-NEXT: v_or_b32_e32 v8, v8, v11
+; SI-NEXT: buffer_load_dword v11, off, s[0:3], s32 offset:344 ; 4-byte Folded Reload
+; SI-NEXT: v_and_b32_e32 v8, 0xffff, v8
; SI-NEXT: s_waitcnt vmcnt(1)
-; SI-NEXT: v_lshlrev_b32_e32 v16, 24, v16
+; SI-NEXT: v_lshlrev_b32_e32 v17, 24, v17
; SI-NEXT: s_waitcnt vmcnt(0)
-; SI-NEXT: v_and_b32_e32 v14, 0xff, v14
-; SI-NEXT: v_lshlrev_b32_e32 v14, 16, v14
-; SI-NEXT: v_or_b32_e32 v14, v16, v14
-; SI-NEXT: v_or_b32_e32 v11, v11, v14
-; SI-NEXT: v_add_i32_e32 v14, vcc, 52, v0
-; SI-NEXT: buffer_store_dword v11, v14, s[0:3], 0 offen
-; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_and_b32_e32 v11, 0xff, v13
-; SI-NEXT: v_lshlrev_b32_e32 v13, 8, v20
-; SI-NEXT: v_or_b32_e32 v11, v11, v13
-; SI-NEXT: buffer_load_dword v13, off, s[0:3], s32 offset:296 ; 4-byte Folded Reload
-; SI-NEXT: buffer_load_dword v14, off, s[0:3], s32 offset:292 ; 4-byte Folded Reload
-; SI-NEXT: v_and_b32_e32 v11, 0xffff, v11
+; SI-NEXT: v_and_b32_e32 v11, 0xff, v11
+; SI-NEXT: v_lshlrev_b32_e32 v11, 16, v11
+; SI-NEXT: v_or_b32_e32 v11, v17, v11
+; SI-NEXT: v_or_b32_e32 v8, v8, v11
+; SI-NEXT: v_add_i32_e32 v11, vcc, 40, v0
+; SI-NEXT: buffer_store_dword v8, v11, s[0:3], 0 offen
+; SI-NEXT: s_waitcnt expcnt(0)
+; SI-NEXT: v_and_b32_e32 v8, 0xff, v55
+; SI-NEXT: v_lshlrev_b32_e32 v11, 8, v20
+; SI-NEXT: v_or_b32_e32 v8, v8, v11
+; SI-NEXT: buffer_load_dword v11, off, s[0:3], s32 offset:320 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v17, off, s[0:3], s32 offset:404 ; 4-byte Folded Reload
+; SI-NEXT: v_and_b32_e32 v8, 0xffff, v8
; SI-NEXT: s_waitcnt vmcnt(1)
-; SI-NEXT: v_and_b32_e32 v13, 0xff, v13
-; SI-NEXT: v_lshlrev_b32_e32 v13, 16, v13
-; SI-NEXT: s_waitcnt vmcnt(0)
-; SI-NEXT: v_lshlrev_b32_e32 v14, 24, v14
-; SI-NEXT: v_or_b32_e32 v13, v14, v13
-; SI-NEXT: v_or_b32_e32 v11, v11, v13
-; SI-NEXT: v_add_i32_e32 v13, vcc, 56, v0
-; SI-NEXT: buffer_store_dword v11, v13, s[0:3], 0 offen
-; SI-NEXT: buffer_load_dword v14, off, s[0:3], s32 offset:320 ; 4-byte Folded Reload
-; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_and_b32_e32 v11, 0xff, v35
-; SI-NEXT: v_lshlrev_b32_e32 v13, 8, v15
-; SI-NEXT: v_or_b32_e32 v11, v11, v13
-; SI-NEXT: v_and_b32_e32 v13, 0xff, v45
-; SI-NEXT: v_lshlrev_b32_e32 v13, 16, v13
-; SI-NEXT: v_and_b32_e32 v11, 0xffff, v11
+; SI-NEXT: v_and_b32_e32 v11, 0xff, v11
+; SI-NEXT: v_lshlrev_b32_e32 v11, 16, v11
; SI-NEXT: s_waitcnt vmcnt(0)
-; SI-NEXT: v_lshlrev_b32_e32 v14, 24, v14
-; SI-NEXT: v_or_b32_e32 v13, v14, v13
-; SI-NEXT: v_or_b32_e32 v11, v11, v13
-; SI-NEXT: v_add_i32_e32 v13, vcc, 60, v0
-; SI-NEXT: buffer_store_dword v11, v13, s[0:3], 0 offen
+; SI-NEXT: v_lshlrev_b32_e32 v17, 24, v17
+; SI-NEXT: v_or_b32_e32 v11, v17, v11
+; SI-NEXT: v_or_b32_e32 v8, v8, v11
+; SI-NEXT: v_add_i32_e32 v11, vcc, 44, v0
+; SI-NEXT: buffer_store_dword v8, v11, s[0:3], 0 offen
+; SI-NEXT: buffer_load_dword v11, off, s[0:3], s32 offset:332 ; 4-byte Folded Reload
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: buffer_load_dword v11, off, s[0:3], s32 offset:288 ; 4-byte Folded Reload
-; SI-NEXT: buffer_load_dword v13, off, s[0:3], s32 offset:276 ; 4-byte Folded Reload
+; SI-NEXT: v_and_b32_e32 v8, 0xff, v57
+; SI-NEXT: buffer_load_dword v16, off, s[0:3], s32 offset:324 ; 4-byte Folded Reload
; SI-NEXT: s_waitcnt vmcnt(1)
; SI-NEXT: v_lshlrev_b32_e32 v11, 8, v11
-; SI-NEXT: v_or_b32_e32 v10, v10, v11
-; SI-NEXT: buffer_load_dword v11, off, s[0:3], s32 offset:284 ; 4-byte Folded Reload
-; SI-NEXT: v_and_b32_e32 v10, 0xffff, v10
+; SI-NEXT: v_or_b32_e32 v8, v8, v11
+; SI-NEXT: buffer_load_dword v11, off, s[0:3], s32 offset:328 ; 4-byte Folded Reload
+; SI-NEXT: v_and_b32_e32 v8, 0xffff, v8
; SI-NEXT: s_waitcnt vmcnt(1)
-; SI-NEXT: v_lshlrev_b32_e32 v13, 24, v13
+; SI-NEXT: v_lshlrev_b32_e32 v16, 24, v16
; SI-NEXT: s_waitcnt vmcnt(0)
; SI-NEXT: v_and_b32_e32 v11, 0xff, v11
; SI-NEXT: v_lshlrev_b32_e32 v11, 16, v11
-; SI-NEXT: v_or_b32_e32 v11, v13, v11
-; SI-NEXT: v_or_b32_e32 v10, v10, v11
-; SI-NEXT: v_add_i32_e32 v11, vcc, 64, v0
-; SI-NEXT: buffer_store_dword v10, v11, s[0:3], 0 offen
-; SI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:428 ; 4-byte Folded Reload
+; SI-NEXT: v_or_b32_e32 v11, v16, v11
+; SI-NEXT: v_or_b32_e32 v8, v8, v11
+; SI-NEXT: v_add_i32_e32 v11, vcc, 48, v0
+; SI-NEXT: buffer_store_dword v8, v11, s[0:3], 0 offen
+; SI-NEXT: buffer_load_dword v16, off, s[0:3], s32 offset:376 ; 4-byte Folded Reload
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_and_b32_e32 v10, 0xff, v30
-; SI-NEXT: buffer_load_dword v13, off, s[0:3], s32 offset:304 ; 4-byte Folded Reload
-; SI-NEXT: s_waitcnt vmcnt(1)
-; SI-NEXT: v_lshlrev_b32_e32 v11, 8, v2
-; SI-NEXT: v_or_b32_e32 v10, v10, v11
-; SI-NEXT: buffer_load_dword v11, off, s[0:3], s32 offset:340 ; 4-byte Folded Reload
-; SI-NEXT: v_and_b32_e32 v10, 0xffff, v10
-; SI-NEXT: s_waitcnt vmcnt(1)
-; SI-NEXT: v_lshlrev_b32_e32 v13, 24, v13
+; SI-NEXT: v_and_b32_e32 v8, 0xff, v38
+; SI-NEXT: v_lshlrev_b32_e32 v11, 8, v60
+; SI-NEXT: v_or_b32_e32 v8, v8, v11
+; SI-NEXT: v_and_b32_e32 v11, 0xff, v24
+; SI-NEXT: v_lshlrev_b32_e32 v11, 16, v11
+; SI-NEXT: v_and_b32_e32 v8, 0xffff, v8
; SI-NEXT: s_waitcnt vmcnt(0)
+; SI-NEXT: v_lshlrev_b32_e32 v16, 24, v16
+; SI-NEXT: v_or_b32_e32 v11, v16, v11
+; SI-NEXT: v_or_b32_e32 v8, v8, v11
+; SI-NEXT: v_add_i32_e32 v11, vcc, 52, v0
+; SI-NEXT: buffer_store_dword v8, v11, s[0:3], 0 offen
+; SI-NEXT: s_waitcnt expcnt(0)
+; SI-NEXT: v_and_b32_e32 v8, 0xff, v13
+; SI-NEXT: v_lshlrev_b32_e32 v11, 8, v46
+; SI-NEXT: v_or_b32_e32 v8, v8, v11
+; SI-NEXT: buffer_load_dword v11, off, s[0:3], s32 offset:304 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v13, off, s[0:3], s32 offset:300 ; 4-byte Folded Reload
+; SI-NEXT: v_and_b32_e32 v8, 0xffff, v8
+; SI-NEXT: s_waitcnt vmcnt(1)
; SI-NEXT: v_and_b32_e32 v11, 0xff, v11
; SI-NEXT: v_lshlrev_b32_e32 v11, 16, v11
+; SI-NEXT: s_waitcnt vmcnt(0)
+; SI-NEXT: v_lshlrev_b32_e32 v13, 24, v13
; SI-NEXT: v_or_b32_e32 v11, v13, v11
-; SI-NEXT: v_or_b32_e32 v10, v10, v11
-; SI-NEXT: v_add_i32_e32 v11, vcc, 0x44, v0
-; SI-NEXT: buffer_store_dword v10, v11, s[0:3], 0 offen
+; SI-NEXT: v_or_b32_e32 v8, v8, v11
+; SI-NEXT: v_add_i32_e32 v11, vcc, 56, v0
+; SI-NEXT: buffer_store_dword v8, v11, s[0:3], 0 offen
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: buffer_load_dword v10, off, s[0:3], s32 offset:268 ; 4-byte Folded Reload
-; SI-NEXT: buffer_load_dword v11, off, s[0:3], s32 offset:256 ; 4-byte Folded Reload
-; SI-NEXT: s_waitcnt vmcnt(1)
-; SI-NEXT: v_lshlrev_b32_e32 v10, 8, v10
-; SI-NEXT: v_or_b32_e32 v7, v7, v10
-; SI-NEXT: buffer_load_dword v10, off, s[0:3], s32 offset:264 ; 4-byte Folded Reload
-; SI-NEXT: v_and_b32_e32 v7, 0xffff, v7
-; SI-NEXT: s_waitcnt vmcnt(1)
-; SI-NEXT: v_lshlrev_b32_e32 v11, 24, v11
+; SI-NEXT: v_and_b32_e32 v8, 0xff, v35
+; SI-NEXT: v_lshlrev_b32_e32 v11, 8, v45
+; SI-NEXT: v_or_b32_e32 v8, v8, v11
+; SI-NEXT: v_and_b32_e32 v11, 0xff, v2
+; SI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:348 ; 4-byte Folded Reload
+; SI-NEXT: v_lshlrev_b32_e32 v11, 16, v11
+; SI-NEXT: v_and_b32_e32 v8, 0xffff, v8
; SI-NEXT: s_waitcnt vmcnt(0)
-; SI-NEXT: v_and_b32_e32 v10, 0xff, v10
+; SI-NEXT: v_lshlrev_b32_e32 v13, 24, v2
+; SI-NEXT: v_or_b32_e32 v11, v13, v11
+; SI-NEXT: v_or_b32_e32 v8, v8, v11
+; SI-NEXT: v_add_i32_e32 v11, vcc, 60, v0
+; SI-NEXT: buffer_store_dword v8, v11, s[0:3], 0 offen
+; SI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:296 ; 4-byte Folded Reload
+; SI-NEXT: s_waitcnt expcnt(0)
+; SI-NEXT: v_and_b32_e32 v8, 0xff, v10
+; SI-NEXT: s_waitcnt vmcnt(0)
+; SI-NEXT: v_lshlrev_b32_e32 v10, 8, v2
+; SI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:292 ; 4-byte Folded Reload
+; SI-NEXT: v_or_b32_e32 v8, v8, v10
+; SI-NEXT: v_and_b32_e32 v8, 0xffff, v8
+; SI-NEXT: s_waitcnt vmcnt(0)
+; SI-NEXT: v_and_b32_e32 v10, 0xff, v2
+; SI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:284 ; 4-byte Folded Reload
+; SI-NEXT: v_lshlrev_b32_e32 v10, 16, v10
+; SI-NEXT: s_waitcnt vmcnt(0)
+; SI-NEXT: v_lshlrev_b32_e32 v11, 24, v2
+; SI-NEXT: v_or_b32_e32 v10, v11, v10
+; SI-NEXT: v_or_b32_e32 v8, v8, v10
+; SI-NEXT: v_add_i32_e32 v10, vcc, 64, v0
+; SI-NEXT: buffer_store_dword v8, v10, s[0:3], 0 offen
+; SI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:440 ; 4-byte Folded Reload
+; SI-NEXT: s_waitcnt expcnt(0)
+; SI-NEXT: v_and_b32_e32 v8, 0xff, v30
+; SI-NEXT: v_lshlrev_b32_e32 v10, 8, v14
+; SI-NEXT: v_or_b32_e32 v8, v8, v10
+; SI-NEXT: v_and_b32_e32 v8, 0xffff, v8
+; SI-NEXT: s_waitcnt vmcnt(0)
+; SI-NEXT: v_and_b32_e32 v10, 0xff, v2
+; SI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:312 ; 4-byte Folded Reload
; SI-NEXT: v_lshlrev_b32_e32 v10, 16, v10
+; SI-NEXT: s_waitcnt vmcnt(0)
+; SI-NEXT: v_lshlrev_b32_e32 v11, 24, v2
; SI-NEXT: v_or_b32_e32 v10, v11, v10
-; SI-NEXT: v_or_b32_e32 v7, v7, v10
-; SI-NEXT: v_add_i32_e32 v10, vcc, 0x48, v0
-; SI-NEXT: buffer_store_dword v7, v10, s[0:3], 0 offen
+; SI-NEXT: v_or_b32_e32 v8, v8, v10
+; SI-NEXT: v_add_i32_e32 v10, vcc, 0x44, v0
+; SI-NEXT: buffer_store_dword v8, v10, s[0:3], 0 offen
+; SI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:276 ; 4-byte Folded Reload
+; SI-NEXT: s_waitcnt vmcnt(0) expcnt(0)
+; SI-NEXT: v_lshlrev_b32_e32 v8, 8, v2
+; SI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:272 ; 4-byte Folded Reload
+; SI-NEXT: v_or_b32_e32 v7, v7, v8
+; SI-NEXT: v_and_b32_e32 v7, 0xffff, v7
+; SI-NEXT: s_waitcnt vmcnt(0)
+; SI-NEXT: v_and_b32_e32 v8, 0xff, v2
+; SI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:264 ; 4-byte Folded Reload
+; SI-NEXT: v_lshlrev_b32_e32 v8, 16, v8
+; SI-NEXT: s_waitcnt vmcnt(0)
+; SI-NEXT: v_lshlrev_b32_e32 v10, 24, v2
+; SI-NEXT: v_or_b32_e32 v8, v10, v8
+; SI-NEXT: v_or_b32_e32 v7, v7, v8
+; SI-NEXT: v_add_i32_e32 v8, vcc, 0x48, v0
+; SI-NEXT: buffer_store_dword v7, v8, s[0:3], 0 offen
+; SI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:388 ; 4-byte Folded Reload
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_and_b32_e32 v7, 0xff, v27
+; SI-NEXT: v_lshlrev_b32_e32 v8, 8, v18
; SI-NEXT: v_or_b32_e32 v7, v7, v8
-; SI-NEXT: buffer_load_dword v8, off, s[0:3], s32 offset:352 ; 4-byte Folded Reload
-; SI-NEXT: buffer_load_dword v10, off, s[0:3], s32 offset:272 ; 4-byte Folded Reload
; SI-NEXT: v_and_b32_e32 v7, 0xffff, v7
-; SI-NEXT: s_waitcnt vmcnt(1)
-; SI-NEXT: v_and_b32_e32 v8, 0xff, v8
+; SI-NEXT: s_waitcnt vmcnt(0)
+; SI-NEXT: v_and_b32_e32 v8, 0xff, v2
+; SI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:280 ; 4-byte Folded Reload
; SI-NEXT: v_lshlrev_b32_e32 v8, 16, v8
; SI-NEXT: s_waitcnt vmcnt(0)
-; SI-NEXT: v_lshlrev_b32_e32 v10, 24, v10
+; SI-NEXT: v_lshlrev_b32_e32 v10, 24, v2
; SI-NEXT: v_or_b32_e32 v8, v10, v8
; SI-NEXT: v_or_b32_e32 v7, v7, v8
; SI-NEXT: v_add_i32_e32 v8, vcc, 0x4c, v0
; SI-NEXT: buffer_store_dword v7, v8, s[0:3], 0 offen
-; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: buffer_load_dword v7, off, s[0:3], s32 offset:260 ; 4-byte Folded Reload
-; SI-NEXT: buffer_load_dword v8, off, s[0:3], s32 offset:236 ; 4-byte Folded Reload
-; SI-NEXT: s_waitcnt vmcnt(1)
-; SI-NEXT: v_lshlrev_b32_e32 v7, 8, v7
+; SI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:268 ; 4-byte Folded Reload
+; SI-NEXT: s_waitcnt vmcnt(0) expcnt(0)
+; SI-NEXT: v_lshlrev_b32_e32 v7, 8, v2
+; SI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:256 ; 4-byte Folded Reload
; SI-NEXT: v_or_b32_e32 v6, v6, v7
-; SI-NEXT: v_and_b32_e32 v7, 0xff, v26
+; SI-NEXT: v_and_b32_e32 v6, 0xffff, v6
+; SI-NEXT: s_waitcnt vmcnt(0)
+; SI-NEXT: v_and_b32_e32 v7, 0xff, v2
+; SI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:248 ; 4-byte Folded Reload
; SI-NEXT: v_lshlrev_b32_e32 v7, 16, v7
; SI-NEXT: s_waitcnt vmcnt(0)
-; SI-NEXT: v_lshlrev_b32_e32 v8, 24, v8
-; SI-NEXT: v_and_b32_e32 v6, 0xffff, v6
+; SI-NEXT: v_lshlrev_b32_e32 v8, 24, v2
; SI-NEXT: v_or_b32_e32 v7, v8, v7
; SI-NEXT: v_or_b32_e32 v6, v6, v7
; SI-NEXT: v_add_i32_e32 v7, vcc, 0x50, v0
; SI-NEXT: buffer_store_dword v6, v7, s[0:3], 0 offen
-; SI-NEXT: buffer_load_dword v7, off, s[0:3], s32 offset:388 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:416 ; 4-byte Folded Reload
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_and_b32_e32 v6, 0xff, v24
-; SI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:452 ; 4-byte Folded Reload
-; SI-NEXT: s_waitcnt vmcnt(1)
-; SI-NEXT: v_lshlrev_b32_e32 v7, 8, v7
+; SI-NEXT: v_and_b32_e32 v6, 0xff, v37
+; SI-NEXT: s_waitcnt vmcnt(0)
+; SI-NEXT: v_lshlrev_b32_e32 v7, 8, v2
+; SI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:316 ; 4-byte Folded Reload
; SI-NEXT: v_or_b32_e32 v6, v6, v7
-; SI-NEXT: buffer_load_dword v7, off, s[0:3], s32 offset:252 ; 4-byte Folded Reload
; SI-NEXT: v_and_b32_e32 v6, 0xffff, v6
-; SI-NEXT: s_waitcnt vmcnt(1)
-; SI-NEXT: v_lshlrev_b32_e32 v8, 24, v2
; SI-NEXT: s_waitcnt vmcnt(0)
-; SI-NEXT: v_and_b32_e32 v7, 0xff, v7
+; SI-NEXT: v_and_b32_e32 v7, 0xff, v2
+; SI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:260 ; 4-byte Folded Reload
; SI-NEXT: v_lshlrev_b32_e32 v7, 16, v7
+; SI-NEXT: s_waitcnt vmcnt(0)
+; SI-NEXT: v_lshlrev_b32_e32 v8, 24, v2
; SI-NEXT: v_or_b32_e32 v7, v8, v7
; SI-NEXT: v_or_b32_e32 v6, v6, v7
; SI-NEXT: v_add_i32_e32 v7, vcc, 0x54, v0
; SI-NEXT: buffer_store_dword v6, v7, s[0:3], 0 offen
-; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: buffer_load_dword v6, off, s[0:3], s32 offset:240 ; 4-byte Folded Reload
-; SI-NEXT: buffer_load_dword v7, off, s[0:3], s32 offset:220 ; 4-byte Folded Reload
-; SI-NEXT: s_waitcnt vmcnt(1)
-; SI-NEXT: v_lshlrev_b32_e32 v6, 8, v6
+; SI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:252 ; 4-byte Folded Reload
+; SI-NEXT: s_waitcnt vmcnt(0) expcnt(0)
+; SI-NEXT: v_lshlrev_b32_e32 v6, 8, v2
+; SI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:244 ; 4-byte Folded Reload
; SI-NEXT: v_or_b32_e32 v5, v5, v6
-; SI-NEXT: buffer_load_dword v6, off, s[0:3], s32 offset:232 ; 4-byte Folded Reload
-; SI-NEXT: s_waitcnt vmcnt(1)
-; SI-NEXT: v_lshlrev_b32_e32 v7, 24, v7
; SI-NEXT: v_and_b32_e32 v5, 0xffff, v5
; SI-NEXT: s_waitcnt vmcnt(0)
-; SI-NEXT: v_and_b32_e32 v6, 0xff, v6
+; SI-NEXT: v_and_b32_e32 v6, 0xff, v2
+; SI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:232 ; 4-byte Folded Reload
; SI-NEXT: v_lshlrev_b32_e32 v6, 16, v6
+; SI-NEXT: s_waitcnt vmcnt(0)
+; SI-NEXT: v_lshlrev_b32_e32 v7, 24, v2
; SI-NEXT: v_or_b32_e32 v6, v7, v6
; SI-NEXT: v_or_b32_e32 v5, v5, v6
; SI-NEXT: v_add_i32_e32 v6, vcc, 0x58, v0
; SI-NEXT: buffer_store_dword v5, v6, s[0:3], 0 offen
-; SI-NEXT: buffer_load_dword v6, off, s[0:3], s32 offset:360 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:384 ; 4-byte Folded Reload
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_and_b32_e32 v5, 0xff, v21
-; SI-NEXT: buffer_load_dword v7, off, s[0:3], s32 offset:224 ; 4-byte Folded Reload
-; SI-NEXT: s_waitcnt vmcnt(1)
-; SI-NEXT: v_lshlrev_b32_e32 v6, 8, v6
+; SI-NEXT: s_waitcnt vmcnt(0)
+; SI-NEXT: v_lshlrev_b32_e32 v6, 8, v2
; SI-NEXT: v_or_b32_e32 v5, v5, v6
-; SI-NEXT: buffer_load_dword v6, off, s[0:3], s32 offset:248 ; 4-byte Folded Reload
-; SI-NEXT: s_waitcnt vmcnt(1)
-; SI-NEXT: v_lshlrev_b32_e32 v7, 24, v7
+; SI-NEXT: buffer_load_dword v6, off, s[0:3], s32 offset:160 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:236 ; 4-byte Folded Reload
; SI-NEXT: v_and_b32_e32 v5, 0xffff, v5
-; SI-NEXT: s_waitcnt vmcnt(0)
+; SI-NEXT: s_waitcnt vmcnt(1)
; SI-NEXT: v_and_b32_e32 v6, 0xff, v6
; SI-NEXT: v_lshlrev_b32_e32 v6, 16, v6
+; SI-NEXT: s_waitcnt vmcnt(0)
+; SI-NEXT: v_lshlrev_b32_e32 v7, 24, v2
; SI-NEXT: v_or_b32_e32 v6, v7, v6
; SI-NEXT: v_or_b32_e32 v5, v5, v6
; SI-NEXT: v_add_i32_e32 v6, vcc, 0x5c, v0
; SI-NEXT: buffer_store_dword v5, v6, s[0:3], 0 offen
-; SI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:440 ; 4-byte Folded Reload
-; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: buffer_load_dword v5, off, s[0:3], s32 offset:228 ; 4-byte Folded Reload
-; SI-NEXT: buffer_load_dword v6, off, s[0:3], s32 offset:208 ; 4-byte Folded Reload
-; SI-NEXT: s_waitcnt vmcnt(2)
-; SI-NEXT: v_and_b32_e32 v4, 0xff, v2
-; SI-NEXT: s_waitcnt vmcnt(1)
-; SI-NEXT: v_lshlrev_b32_e32 v5, 8, v5
+; SI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:240 ; 4-byte Folded Reload
+; SI-NEXT: s_waitcnt vmcnt(0) expcnt(0)
+; SI-NEXT: v_lshlrev_b32_e32 v5, 8, v2
+; SI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:228 ; 4-byte Folded Reload
; SI-NEXT: v_or_b32_e32 v4, v4, v5
-; SI-NEXT: buffer_load_dword v5, off, s[0:3], s32 offset:216 ; 4-byte Folded Reload
-; SI-NEXT: s_waitcnt vmcnt(1)
-; SI-NEXT: v_lshlrev_b32_e32 v6, 24, v6
; SI-NEXT: v_and_b32_e32 v4, 0xffff, v4
-; SI-NEXT: v_and_b32_e32 v2, 0xff, v52
; SI-NEXT: s_waitcnt vmcnt(0)
-; SI-NEXT: v_and_b32_e32 v5, 0xff, v5
+; SI-NEXT: v_and_b32_e32 v5, 0xff, v2
+; SI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:220 ; 4-byte Folded Reload
; SI-NEXT: v_lshlrev_b32_e32 v5, 16, v5
+; SI-NEXT: s_waitcnt vmcnt(0)
+; SI-NEXT: v_lshlrev_b32_e32 v6, 24, v2
; SI-NEXT: v_or_b32_e32 v5, v6, v5
; SI-NEXT: v_or_b32_e32 v4, v4, v5
; SI-NEXT: v_add_i32_e32 v5, vcc, 0x60, v0
; SI-NEXT: buffer_store_dword v4, v5, s[0:3], 0 offen
-; SI-NEXT: buffer_load_dword v5, off, s[0:3], s32 offset:332 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:360 ; 4-byte Folded Reload
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_and_b32_e32 v4, 0xff, v18
-; SI-NEXT: buffer_load_dword v6, off, s[0:3], s32 offset:196 ; 4-byte Folded Reload
-; SI-NEXT: s_waitcnt vmcnt(1)
-; SI-NEXT: v_lshlrev_b32_e32 v5, 8, v5
+; SI-NEXT: v_and_b32_e32 v4, 0xff, v61
+; SI-NEXT: s_waitcnt vmcnt(0)
+; SI-NEXT: v_lshlrev_b32_e32 v5, 8, v2
; SI-NEXT: v_or_b32_e32 v4, v4, v5
-; SI-NEXT: buffer_load_dword v5, off, s[0:3], s32 offset:188 ; 4-byte Folded Reload
-; SI-NEXT: s_waitcnt vmcnt(1)
-; SI-NEXT: v_lshlrev_b32_e32 v6, 24, v6
+; SI-NEXT: buffer_load_dword v5, off, s[0:3], s32 offset:156 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:208 ; 4-byte Folded Reload
; SI-NEXT: v_and_b32_e32 v4, 0xffff, v4
-; SI-NEXT: s_waitcnt vmcnt(0)
+; SI-NEXT: s_waitcnt vmcnt(1)
; SI-NEXT: v_and_b32_e32 v5, 0xff, v5
; SI-NEXT: v_lshlrev_b32_e32 v5, 16, v5
+; SI-NEXT: s_waitcnt vmcnt(0)
+; SI-NEXT: v_lshlrev_b32_e32 v6, 24, v2
; SI-NEXT: v_or_b32_e32 v5, v6, v5
; SI-NEXT: v_or_b32_e32 v4, v4, v5
; SI-NEXT: v_add_i32_e32 v5, vcc, 0x64, v0
; SI-NEXT: buffer_store_dword v4, v5, s[0:3], 0 offen
-; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: buffer_load_dword v4, off, s[0:3], s32 offset:212 ; 4-byte Folded Reload
-; SI-NEXT: buffer_load_dword v5, off, s[0:3], s32 offset:192 ; 4-byte Folded Reload
-; SI-NEXT: s_waitcnt vmcnt(1)
-; SI-NEXT: v_lshlrev_b32_e32 v4, 8, v4
+; SI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:224 ; 4-byte Folded Reload
+; SI-NEXT: s_waitcnt vmcnt(0) expcnt(0)
+; SI-NEXT: v_lshlrev_b32_e32 v4, 8, v2
+; SI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:216 ; 4-byte Folded Reload
; SI-NEXT: v_or_b32_e32 v3, v3, v4
-; SI-NEXT: buffer_load_dword v4, off, s[0:3], s32 offset:204 ; 4-byte Folded Reload
-; SI-NEXT: s_waitcnt vmcnt(1)
-; SI-NEXT: v_lshlrev_b32_e32 v5, 24, v5
; SI-NEXT: v_and_b32_e32 v3, 0xffff, v3
; SI-NEXT: s_waitcnt vmcnt(0)
-; SI-NEXT: v_and_b32_e32 v4, 0xff, v4
+; SI-NEXT: v_and_b32_e32 v4, 0xff, v2
+; SI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:204 ; 4-byte Folded Reload
; SI-NEXT: v_lshlrev_b32_e32 v4, 16, v4
+; SI-NEXT: s_waitcnt vmcnt(0)
+; SI-NEXT: v_lshlrev_b32_e32 v5, 24, v2
; SI-NEXT: v_or_b32_e32 v4, v5, v4
; SI-NEXT: v_or_b32_e32 v3, v3, v4
; SI-NEXT: v_add_i32_e32 v4, vcc, 0x68, v0
; SI-NEXT: buffer_store_dword v3, v4, s[0:3], 0 offen
+; SI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:336 ; 4-byte Folded Reload
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_and_b32_e32 v3, 0xff, v63
-; SI-NEXT: v_lshlrev_b32_e32 v4, 8, v42
+; SI-NEXT: s_waitcnt vmcnt(0)
+; SI-NEXT: v_lshlrev_b32_e32 v4, 8, v2
; SI-NEXT: v_or_b32_e32 v3, v3, v4
; SI-NEXT: buffer_load_dword v4, off, s[0:3], s32 offset:152 ; 4-byte Folded Reload
-; SI-NEXT: buffer_load_dword v5, off, s[0:3], s32 offset:180 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:184 ; 4-byte Folded Reload
; SI-NEXT: v_and_b32_e32 v3, 0xffff, v3
; SI-NEXT: s_waitcnt vmcnt(1)
; SI-NEXT: v_and_b32_e32 v4, 0xff, v4
; SI-NEXT: v_lshlrev_b32_e32 v4, 16, v4
; SI-NEXT: s_waitcnt vmcnt(0)
-; SI-NEXT: v_lshlrev_b32_e32 v5, 24, v5
+; SI-NEXT: v_lshlrev_b32_e32 v5, 24, v2
; SI-NEXT: v_or_b32_e32 v4, v5, v4
; SI-NEXT: v_or_b32_e32 v3, v3, v4
; SI-NEXT: v_add_i32_e32 v4, vcc, 0x6c, v0
; SI-NEXT: buffer_store_dword v3, v4, s[0:3], 0 offen
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: buffer_load_dword v3, off, s[0:3], s32 offset:200 ; 4-byte Folded Reload
-; SI-NEXT: buffer_load_dword v4, off, s[0:3], s32 offset:244 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v3, off, s[0:3], s32 offset:212 ; 4-byte Folded Reload
+; SI-NEXT: v_and_b32_e32 v2, 0xff, v56
+; SI-NEXT: buffer_load_dword v4, off, s[0:3], s32 offset:192 ; 4-byte Folded Reload
; SI-NEXT: s_waitcnt vmcnt(1)
; SI-NEXT: v_lshlrev_b32_e32 v3, 8, v3
; SI-NEXT: v_or_b32_e32 v2, v2, v3
-; SI-NEXT: buffer_load_dword v3, off, s[0:3], s32 offset:436 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v3, off, s[0:3], s32 offset:200 ; 4-byte Folded Reload
; SI-NEXT: s_waitcnt vmcnt(1)
; SI-NEXT: v_lshlrev_b32_e32 v4, 24, v4
; SI-NEXT: v_and_b32_e32 v2, 0xffff, v2
@@ -171894,14 +172001,14 @@ define inreg <128 x i8> @bitcast_v64bf16_to_v128i8_scalar(<64 x bfloat> inreg %a
; SI-NEXT: v_or_b32_e32 v2, v2, v3
; SI-NEXT: v_add_i32_e32 v3, vcc, 0x70, v0
; SI-NEXT: buffer_store_dword v2, v3, s[0:3], 0 offen
-; SI-NEXT: buffer_load_dword v3, off, s[0:3], s32 offset:300 ; 4-byte Folded Reload
-; SI-NEXT: buffer_load_dword v4, off, s[0:3], s32 offset:172 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v3, off, s[0:3], s32 offset:308 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v4, off, s[0:3], s32 offset:176 ; 4-byte Folded Reload
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_and_b32_e32 v2, 0xff, v12
; SI-NEXT: s_waitcnt vmcnt(1)
; SI-NEXT: v_lshlrev_b32_e32 v3, 8, v3
; SI-NEXT: v_or_b32_e32 v2, v2, v3
-; SI-NEXT: v_and_b32_e32 v3, 0xff, v40
+; SI-NEXT: v_and_b32_e32 v3, 0xff, v49
; SI-NEXT: v_lshlrev_b32_e32 v3, 16, v3
; SI-NEXT: s_waitcnt vmcnt(0)
; SI-NEXT: v_lshlrev_b32_e32 v4, 24, v4
@@ -171911,12 +172018,12 @@ define inreg <128 x i8> @bitcast_v64bf16_to_v128i8_scalar(<64 x bfloat> inreg %a
; SI-NEXT: v_add_i32_e32 v3, vcc, 0x74, v0
; SI-NEXT: buffer_store_dword v2, v3, s[0:3], 0 offen
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:308 ; 4-byte Folded Reload
-; SI-NEXT: buffer_load_dword v3, off, s[0:3], s32 offset:176 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:196 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v3, off, s[0:3], s32 offset:180 ; 4-byte Folded Reload
; SI-NEXT: s_waitcnt vmcnt(1)
; SI-NEXT: v_lshlrev_b32_e32 v2, 8, v2
; SI-NEXT: v_or_b32_e32 v1, v1, v2
-; SI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:184 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:188 ; 4-byte Folded Reload
; SI-NEXT: s_waitcnt vmcnt(1)
; SI-NEXT: v_lshlrev_b32_e32 v3, 24, v3
; SI-NEXT: v_and_b32_e32 v1, 0xffff, v1
@@ -171927,10 +172034,10 @@ define inreg <128 x i8> @bitcast_v64bf16_to_v128i8_scalar(<64 x bfloat> inreg %a
; SI-NEXT: v_or_b32_e32 v1, v1, v2
; SI-NEXT: v_add_i32_e32 v2, vcc, 0x78, v0
; SI-NEXT: buffer_store_dword v1, v2, s[0:3], 0 offen
-; SI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:280 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:288 ; 4-byte Folded Reload
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_and_b32_e32 v1, 0xff, v9
-; SI-NEXT: buffer_load_dword v3, off, s[0:3], s32 offset:168 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v3, off, s[0:3], s32 offset:424 ; 4-byte Folded Reload
; SI-NEXT: v_add_i32_e32 v0, vcc, 0x7c, v0
; SI-NEXT: s_waitcnt vmcnt(1)
; SI-NEXT: v_lshlrev_b32_e32 v2, 8, v2
@@ -171968,8 +172075,8 @@ define inreg <128 x i8> @bitcast_v64bf16_to_v128i8_scalar(<64 x bfloat> inreg %a
; VI: ; %bb.0:
; VI-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; VI-NEXT: s_or_saveexec_b64 s[4:5], -1
-; VI-NEXT: buffer_store_dword v63, off, s[0:3], s32 offset:392 ; 4-byte Folded Spill
-; VI-NEXT: buffer_store_dword v62, off, s[0:3], s32 offset:396 ; 4-byte Folded Spill
+; VI-NEXT: buffer_store_dword v63, off, s[0:3], s32 offset:396 ; 4-byte Folded Spill
+; VI-NEXT: buffer_store_dword v62, off, s[0:3], s32 offset:400 ; 4-byte Folded Spill
; VI-NEXT: s_mov_b64 exec, s[4:5]
; VI-NEXT: v_writelane_b32 v63, s30, 0
; VI-NEXT: v_writelane_b32 v63, s31, 1
@@ -172774,26 +172881,26 @@ define inreg <128 x i8> @bitcast_v64bf16_to_v128i8_scalar(<64 x bfloat> inreg %a
; VI-NEXT: v_lshrrev_b32_e32 v31, 16, v31
; VI-NEXT: v_alignbit_b32 v31, v31, v33, 16
; VI-NEXT: v_lshrrev_b64 v[33:34], 24, v[31:32]
-; VI-NEXT: buffer_store_dword v33, off, s[0:3], s32 offset:168 ; 4-byte Folded Spill
-; VI-NEXT: buffer_store_dword v34, off, s[0:3], s32 offset:172 ; 4-byte Folded Spill
+; VI-NEXT: buffer_store_dword v33, off, s[0:3], s32 offset:176 ; 4-byte Folded Spill
+; VI-NEXT: buffer_store_dword v34, off, s[0:3], s32 offset:180 ; 4-byte Folded Spill
; VI-NEXT: v_lshrrev_b64 v[33:34], 24, v[29:30]
-; VI-NEXT: buffer_store_dword v33, off, s[0:3], s32 offset:196 ; 4-byte Folded Spill
-; VI-NEXT: buffer_store_dword v34, off, s[0:3], s32 offset:200 ; 4-byte Folded Spill
-; VI-NEXT: v_lshrrev_b64 v[33:34], 24, v[27:28]
; VI-NEXT: buffer_store_dword v33, off, s[0:3], s32 offset:204 ; 4-byte Folded Spill
; VI-NEXT: buffer_store_dword v34, off, s[0:3], s32 offset:208 ; 4-byte Folded Spill
+; VI-NEXT: v_lshrrev_b64 v[33:34], 24, v[27:28]
+; VI-NEXT: buffer_store_dword v33, off, s[0:3], s32 offset:212 ; 4-byte Folded Spill
+; VI-NEXT: buffer_store_dword v34, off, s[0:3], s32 offset:216 ; 4-byte Folded Spill
; VI-NEXT: v_lshrrev_b64 v[33:34], 24, v[25:26]
; VI-NEXT: buffer_store_dword v33, off, s[0:3], s32 offset:56 ; 4-byte Folded Spill
; VI-NEXT: buffer_store_dword v34, off, s[0:3], s32 offset:60 ; 4-byte Folded Spill
; VI-NEXT: v_lshrrev_b64 v[33:34], 24, v[21:22]
-; VI-NEXT: buffer_store_dword v33, off, s[0:3], s32 offset:252 ; 4-byte Folded Spill
-; VI-NEXT: buffer_store_dword v34, off, s[0:3], s32 offset:256 ; 4-byte Folded Spill
+; VI-NEXT: buffer_store_dword v33, off, s[0:3], s32 offset:260 ; 4-byte Folded Spill
+; VI-NEXT: buffer_store_dword v34, off, s[0:3], s32 offset:264 ; 4-byte Folded Spill
; VI-NEXT: v_lshrrev_b64 v[33:34], 24, v[19:20]
-; VI-NEXT: buffer_store_dword v33, off, s[0:3], s32 offset:292 ; 4-byte Folded Spill
-; VI-NEXT: buffer_store_dword v34, off, s[0:3], s32 offset:296 ; 4-byte Folded Spill
+; VI-NEXT: buffer_store_dword v33, off, s[0:3], s32 offset:300 ; 4-byte Folded Spill
+; VI-NEXT: buffer_store_dword v34, off, s[0:3], s32 offset:304 ; 4-byte Folded Spill
; VI-NEXT: v_lshrrev_b64 v[33:34], 24, v[17:18]
-; VI-NEXT: buffer_store_dword v33, off, s[0:3], s32 offset:324 ; 4-byte Folded Spill
-; VI-NEXT: buffer_store_dword v34, off, s[0:3], s32 offset:328 ; 4-byte Folded Spill
+; VI-NEXT: buffer_store_dword v33, off, s[0:3], s32 offset:332 ; 4-byte Folded Spill
+; VI-NEXT: buffer_store_dword v34, off, s[0:3], s32 offset:336 ; 4-byte Folded Spill
; VI-NEXT: v_lshrrev_b64 v[33:34], 24, v[15:16]
; VI-NEXT: buffer_store_dword v33, off, s[0:3], s32 offset:64 ; 4-byte Folded Spill
; VI-NEXT: buffer_store_dword v34, off, s[0:3], s32 offset:68 ; 4-byte Folded Spill
@@ -172804,126 +172911,129 @@ define inreg <128 x i8> @bitcast_v64bf16_to_v128i8_scalar(<64 x bfloat> inreg %a
; VI-NEXT: buffer_store_dword v33, off, s[0:3], s32 offset:96 ; 4-byte Folded Spill
; VI-NEXT: buffer_store_dword v34, off, s[0:3], s32 offset:100 ; 4-byte Folded Spill
; VI-NEXT: v_lshrrev_b64 v[33:34], 24, v[9:10]
-; VI-NEXT: buffer_store_dword v33, off, s[0:3], s32 offset:116 ; 4-byte Folded Spill
-; VI-NEXT: buffer_store_dword v34, off, s[0:3], s32 offset:120 ; 4-byte Folded Spill
+; VI-NEXT: buffer_store_dword v33, off, s[0:3], s32 offset:112 ; 4-byte Folded Spill
+; VI-NEXT: buffer_store_dword v34, off, s[0:3], s32 offset:116 ; 4-byte Folded Spill
; VI-NEXT: v_lshrrev_b64 v[33:34], 24, v[7:8]
-; VI-NEXT: buffer_store_dword v33, off, s[0:3], s32 offset:132 ; 4-byte Folded Spill
-; VI-NEXT: buffer_store_dword v34, off, s[0:3], s32 offset:136 ; 4-byte Folded Spill
+; VI-NEXT: buffer_store_dword v33, off, s[0:3], s32 offset:124 ; 4-byte Folded Spill
+; VI-NEXT: buffer_store_dword v34, off, s[0:3], s32 offset:128 ; 4-byte Folded Spill
; VI-NEXT: v_lshrrev_b64 v[33:34], 24, v[5:6]
-; VI-NEXT: buffer_store_dword v33, off, s[0:3], s32 offset:140 ; 4-byte Folded Spill
-; VI-NEXT: buffer_store_dword v34, off, s[0:3], s32 offset:144 ; 4-byte Folded Spill
+; VI-NEXT: buffer_store_dword v33, off, s[0:3], s32 offset:136 ; 4-byte Folded Spill
+; VI-NEXT: buffer_store_dword v34, off, s[0:3], s32 offset:140 ; 4-byte Folded Spill
; VI-NEXT: v_lshrrev_b64 v[33:34], 24, v[3:4]
-; VI-NEXT: buffer_store_dword v33, off, s[0:3], s32 offset:152 ; 4-byte Folded Spill
-; VI-NEXT: buffer_store_dword v34, off, s[0:3], s32 offset:156 ; 4-byte Folded Spill
+; VI-NEXT: buffer_store_dword v33, off, s[0:3], s32 offset:160 ; 4-byte Folded Spill
+; VI-NEXT: buffer_store_dword v34, off, s[0:3], s32 offset:164 ; 4-byte Folded Spill
; VI-NEXT: v_lshrrev_b64 v[33:34], 24, v[1:2]
-; VI-NEXT: buffer_store_dword v33, off, s[0:3], s32 offset:188 ; 4-byte Folded Spill
-; VI-NEXT: buffer_store_dword v34, off, s[0:3], s32 offset:192 ; 4-byte Folded Spill
+; VI-NEXT: buffer_store_dword v33, off, s[0:3], s32 offset:196 ; 4-byte Folded Spill
+; VI-NEXT: buffer_store_dword v34, off, s[0:3], s32 offset:200 ; 4-byte Folded Spill
; VI-NEXT: v_lshrrev_b32_e32 v33, 24, v32
-; VI-NEXT: buffer_store_dword v33, off, s[0:3], s32 offset:220 ; 4-byte Folded Spill
+; VI-NEXT: buffer_store_dword v33, off, s[0:3], s32 offset:228 ; 4-byte Folded Spill
; VI-NEXT: v_lshrrev_b32_e32 v33, 16, v32
-; VI-NEXT: buffer_store_dword v33, off, s[0:3], s32 offset:212 ; 4-byte Folded Spill
+; VI-NEXT: buffer_store_dword v33, off, s[0:3], s32 offset:220 ; 4-byte Folded Spill
; VI-NEXT: v_lshrrev_b32_e32 v33, 8, v32
-; VI-NEXT: buffer_store_dword v33, off, s[0:3], s32 offset:216 ; 4-byte Folded Spill
+; VI-NEXT: buffer_store_dword v33, off, s[0:3], s32 offset:224 ; 4-byte Folded Spill
; VI-NEXT: v_lshrrev_b32_e32 v33, 16, v31
-; VI-NEXT: buffer_store_dword v33, off, s[0:3], s32 offset:228 ; 4-byte Folded Spill
+; VI-NEXT: buffer_store_dword v33, off, s[0:3], s32 offset:236 ; 4-byte Folded Spill
; VI-NEXT: v_lshrrev_b32_e32 v33, 8, v31
-; VI-NEXT: buffer_store_dword v33, off, s[0:3], s32 offset:224 ; 4-byte Folded Spill
+; VI-NEXT: buffer_store_dword v33, off, s[0:3], s32 offset:232 ; 4-byte Folded Spill
; VI-NEXT: v_lshrrev_b32_e32 v33, 24, v30
-; VI-NEXT: buffer_store_dword v33, off, s[0:3], s32 offset:240 ; 4-byte Folded Spill
+; VI-NEXT: buffer_store_dword v33, off, s[0:3], s32 offset:248 ; 4-byte Folded Spill
; VI-NEXT: v_lshrrev_b32_e32 v33, 16, v30
-; VI-NEXT: buffer_store_dword v33, off, s[0:3], s32 offset:232 ; 4-byte Folded Spill
+; VI-NEXT: buffer_store_dword v33, off, s[0:3], s32 offset:240 ; 4-byte Folded Spill
; VI-NEXT: v_lshrrev_b32_e32 v33, 8, v30
-; VI-NEXT: buffer_store_dword v33, off, s[0:3], s32 offset:236 ; 4-byte Folded Spill
+; VI-NEXT: buffer_store_dword v33, off, s[0:3], s32 offset:244 ; 4-byte Folded Spill
; VI-NEXT: v_lshrrev_b32_e32 v33, 16, v29
-; VI-NEXT: buffer_store_dword v33, off, s[0:3], s32 offset:248 ; 4-byte Folded Spill
+; VI-NEXT: buffer_store_dword v33, off, s[0:3], s32 offset:256 ; 4-byte Folded Spill
; VI-NEXT: v_lshrrev_b32_e32 v33, 8, v29
-; VI-NEXT: buffer_store_dword v33, off, s[0:3], s32 offset:244 ; 4-byte Folded Spill
+; VI-NEXT: buffer_store_dword v33, off, s[0:3], s32 offset:252 ; 4-byte Folded Spill
; VI-NEXT: v_lshrrev_b32_e32 v33, 24, v28
-; VI-NEXT: buffer_store_dword v33, off, s[0:3], s32 offset:268 ; 4-byte Folded Spill
+; VI-NEXT: buffer_store_dword v33, off, s[0:3], s32 offset:276 ; 4-byte Folded Spill
; VI-NEXT: v_lshrrev_b32_e32 v33, 16, v28
-; VI-NEXT: buffer_store_dword v33, off, s[0:3], s32 offset:260 ; 4-byte Folded Spill
+; VI-NEXT: buffer_store_dword v33, off, s[0:3], s32 offset:268 ; 4-byte Folded Spill
; VI-NEXT: v_lshrrev_b32_e32 v33, 8, v28
-; VI-NEXT: buffer_store_dword v33, off, s[0:3], s32 offset:264 ; 4-byte Folded Spill
+; VI-NEXT: buffer_store_dword v33, off, s[0:3], s32 offset:272 ; 4-byte Folded Spill
; VI-NEXT: v_lshrrev_b32_e32 v33, 16, v27
-; VI-NEXT: buffer_store_dword v33, off, s[0:3], s32 offset:276 ; 4-byte Folded Spill
+; VI-NEXT: buffer_store_dword v33, off, s[0:3], s32 offset:284 ; 4-byte Folded Spill
; VI-NEXT: v_lshrrev_b32_e32 v33, 8, v27
-; VI-NEXT: buffer_store_dword v33, off, s[0:3], s32 offset:272 ; 4-byte Folded Spill
+; VI-NEXT: buffer_store_dword v33, off, s[0:3], s32 offset:280 ; 4-byte Folded Spill
; VI-NEXT: v_lshrrev_b32_e32 v33, 24, v26
-; VI-NEXT: buffer_store_dword v33, off, s[0:3], s32 offset:288 ; 4-byte Folded Spill
+; VI-NEXT: buffer_store_dword v33, off, s[0:3], s32 offset:296 ; 4-byte Folded Spill
; VI-NEXT: v_lshrrev_b32_e32 v33, 16, v26
-; VI-NEXT: buffer_store_dword v33, off, s[0:3], s32 offset:280 ; 4-byte Folded Spill
+; VI-NEXT: buffer_store_dword v33, off, s[0:3], s32 offset:288 ; 4-byte Folded Spill
; VI-NEXT: v_lshrrev_b32_e32 v33, 8, v26
-; VI-NEXT: buffer_store_dword v33, off, s[0:3], s32 offset:284 ; 4-byte Folded Spill
+; VI-NEXT: buffer_store_dword v33, off, s[0:3], s32 offset:292 ; 4-byte Folded Spill
; VI-NEXT: v_lshrrev_b32_e32 v33, 16, v25
-; VI-NEXT: buffer_store_dword v33, off, s[0:3], s32 offset:304 ; 4-byte Folded Spill
+; VI-NEXT: buffer_store_dword v33, off, s[0:3], s32 offset:312 ; 4-byte Folded Spill
; VI-NEXT: v_lshrrev_b32_e32 v33, 8, v25
-; VI-NEXT: buffer_store_dword v33, off, s[0:3], s32 offset:300 ; 4-byte Folded Spill
+; VI-NEXT: buffer_store_dword v33, off, s[0:3], s32 offset:308 ; 4-byte Folded Spill
; VI-NEXT: v_lshrrev_b32_e32 v33, 24, v24
-; VI-NEXT: buffer_store_dword v33, off, s[0:3], s32 offset:316 ; 4-byte Folded Spill
+; VI-NEXT: buffer_store_dword v33, off, s[0:3], s32 offset:324 ; 4-byte Folded Spill
; VI-NEXT: v_lshrrev_b32_e32 v33, 16, v24
-; VI-NEXT: buffer_store_dword v33, off, s[0:3], s32 offset:308 ; 4-byte Folded Spill
+; VI-NEXT: buffer_store_dword v33, off, s[0:3], s32 offset:316 ; 4-byte Folded Spill
; VI-NEXT: v_lshrrev_b32_e32 v33, 8, v24
-; VI-NEXT: buffer_store_dword v33, off, s[0:3], s32 offset:312 ; 4-byte Folded Spill
+; VI-NEXT: buffer_store_dword v33, off, s[0:3], s32 offset:320 ; 4-byte Folded Spill
; VI-NEXT: v_lshrrev_b32_e32 v33, 16, v23
-; VI-NEXT: buffer_store_dword v33, off, s[0:3], s32 offset:332 ; 4-byte Folded Spill
+; VI-NEXT: buffer_store_dword v33, off, s[0:3], s32 offset:348 ; 4-byte Folded Spill
; VI-NEXT: v_lshrrev_b32_e32 v33, 8, v23
-; VI-NEXT: buffer_store_dword v33, off, s[0:3], s32 offset:320 ; 4-byte Folded Spill
+; VI-NEXT: buffer_store_dword v33, off, s[0:3], s32 offset:328 ; 4-byte Folded Spill
; VI-NEXT: v_lshrrev_b32_e32 v33, 24, v22
-; VI-NEXT: buffer_store_dword v33, off, s[0:3], s32 offset:344 ; 4-byte Folded Spill
+; VI-NEXT: buffer_store_dword v33, off, s[0:3], s32 offset:360 ; 4-byte Folded Spill
; VI-NEXT: v_lshrrev_b32_e32 v33, 16, v22
-; VI-NEXT: buffer_store_dword v33, off, s[0:3], s32 offset:336 ; 4-byte Folded Spill
+; VI-NEXT: v_lshrrev_b32_e32 v34, 16, v16
+; VI-NEXT: buffer_store_dword v33, off, s[0:3], s32 offset:352 ; 4-byte Folded Spill
; VI-NEXT: v_lshrrev_b32_e32 v33, 8, v22
-; VI-NEXT: buffer_store_dword v33, off, s[0:3], s32 offset:340 ; 4-byte Folded Spill
+; VI-NEXT: buffer_store_dword v34, off, s[0:3], s32 offset:340 ; 4-byte Folded Spill
+; VI-NEXT: v_lshrrev_b32_e32 v34, 16, v15
+; VI-NEXT: buffer_store_dword v33, off, s[0:3], s32 offset:356 ; 4-byte Folded Spill
; VI-NEXT: v_lshrrev_b32_e32 v33, 16, v21
-; VI-NEXT: buffer_store_dword v33, off, s[0:3], s32 offset:352 ; 4-byte Folded Spill
+; VI-NEXT: buffer_store_dword v34, off, s[0:3], s32 offset:344 ; 4-byte Folded Spill
+; VI-NEXT: v_lshrrev_b32_e32 v34, 24, v4
+; VI-NEXT: buffer_store_dword v33, off, s[0:3], s32 offset:368 ; 4-byte Folded Spill
; VI-NEXT: v_lshrrev_b32_e32 v33, 8, v21
-; VI-NEXT: buffer_store_dword v33, off, s[0:3], s32 offset:348 ; 4-byte Folded Spill
-; VI-NEXT: v_lshrrev_b32_e32 v33, 24, v20
+; VI-NEXT: buffer_store_dword v34, off, s[0:3], s32 offset:144 ; 4-byte Folded Spill
+; VI-NEXT: v_lshrrev_b32_e32 v34, 8, v4
; VI-NEXT: buffer_store_dword v33, off, s[0:3], s32 offset:364 ; 4-byte Folded Spill
+; VI-NEXT: v_lshrrev_b32_e32 v33, 24, v20
+; VI-NEXT: buffer_store_dword v34, off, s[0:3], s32 offset:148 ; 4-byte Folded Spill
+; VI-NEXT: v_lshrrev_b32_e32 v34, 16, v3
+; VI-NEXT: buffer_store_dword v33, off, s[0:3], s32 offset:380 ; 4-byte Folded Spill
; VI-NEXT: v_lshrrev_b32_e32 v33, 16, v20
-; VI-NEXT: buffer_store_dword v33, off, s[0:3], s32 offset:356 ; 4-byte Folded Spill
-; VI-NEXT: v_lshrrev_b32_e32 v33, 8, v20
-; VI-NEXT: buffer_store_dword v33, off, s[0:3], s32 offset:360 ; 4-byte Folded Spill
-; VI-NEXT: v_lshrrev_b32_e32 v33, 16, v19
+; VI-NEXT: buffer_store_dword v34, off, s[0:3], s32 offset:156 ; 4-byte Folded Spill
+; VI-NEXT: v_lshrrev_b32_e32 v34, 8, v3
; VI-NEXT: buffer_store_dword v33, off, s[0:3], s32 offset:372 ; 4-byte Folded Spill
-; VI-NEXT: v_lshrrev_b32_e32 v33, 8, v19
-; VI-NEXT: buffer_store_dword v33, off, s[0:3], s32 offset:368 ; 4-byte Folded Spill
-; VI-NEXT: v_lshrrev_b32_e32 v33, 24, v18
+; VI-NEXT: v_lshrrev_b32_e32 v33, 8, v20
+; VI-NEXT: buffer_store_dword v34, off, s[0:3], s32 offset:152 ; 4-byte Folded Spill
; VI-NEXT: v_lshrrev_b32_e32 v34, 24, v2
-; VI-NEXT: buffer_store_dword v33, off, s[0:3], s32 offset:384 ; 4-byte Folded Spill
-; VI-NEXT: v_lshrrev_b32_e32 v33, 16, v18
-; VI-NEXT: v_lshrrev_b32_e32 v35, 16, v16
-; VI-NEXT: buffer_store_dword v34, off, s[0:3], s32 offset:164 ; 4-byte Folded Spill
-; VI-NEXT: v_lshrrev_b32_e32 v34, 16, v2
; VI-NEXT: buffer_store_dword v33, off, s[0:3], s32 offset:376 ; 4-byte Folded Spill
-; VI-NEXT: v_lshrrev_b32_e32 v33, 8, v18
-; VI-NEXT: buffer_store_dword v35, off, s[0:3], s32 offset:148 ; 4-byte Folded Spill
-; VI-NEXT: v_lshrrev_b32_e32 v51, 24, v12
-; VI-NEXT: v_lshrrev_b32_e32 v35, 24, v8
-; VI-NEXT: buffer_store_dword v34, off, s[0:3], s32 offset:160 ; 4-byte Folded Spill
-; VI-NEXT: v_lshrrev_b32_e32 v34, 8, v2
+; VI-NEXT: v_lshrrev_b32_e32 v33, 16, v19
+; VI-NEXT: buffer_store_dword v34, off, s[0:3], s32 offset:172 ; 4-byte Folded Spill
+; VI-NEXT: v_lshrrev_b32_e32 v34, 16, v2
; VI-NEXT: v_lshrrev_b64 v[41:42], 24, v[23:24]
; VI-NEXT: buffer_store_dword v33, off, s[0:3], s32 offset:388 ; 4-byte Folded Spill
-; VI-NEXT: v_lshrrev_b32_e32 v33, 16, v17
-; VI-NEXT: v_lshrrev_b32_e32 v45, 24, v16
+; VI-NEXT: v_lshrrev_b32_e32 v33, 8, v19
+; VI-NEXT: v_lshrrev_b32_e32 v51, 24, v12
+; VI-NEXT: buffer_store_dword v34, off, s[0:3], s32 offset:168 ; 4-byte Folded Spill
+; VI-NEXT: v_lshrrev_b32_e32 v34, 8, v2
+; VI-NEXT: buffer_store_dword v33, off, s[0:3], s32 offset:384 ; 4-byte Folded Spill
+; VI-NEXT: v_lshrrev_b32_e32 v33, 16, v18
; VI-NEXT: v_lshrrev_b32_e32 v55, 8, v16
; VI-NEXT: v_lshrrev_b32_e32 v56, 8, v13
; VI-NEXT: buffer_store_dword v51, off, s[0:3], s32 offset:104 ; 4-byte Folded Spill
; VI-NEXT: v_lshrrev_b32_e32 v51, 8, v12
; VI-NEXT: v_lshrrev_b32_e32 v57, 16, v9
-; VI-NEXT: buffer_store_dword v35, off, s[0:3], s32 offset:124 ; 4-byte Folded Spill
-; VI-NEXT: v_lshrrev_b32_e32 v35, 16, v8
-; VI-NEXT: v_lshrrev_b32_e32 v58, 8, v7
-; VI-NEXT: buffer_store_dword v34, off, s[0:3], s32 offset:176 ; 4-byte Folded Spill
+; VI-NEXT: v_lshrrev_b32_e32 v60, 8, v7
+; VI-NEXT: v_lshrrev_b32_e32 v42, 8, v5
+; VI-NEXT: buffer_store_dword v34, off, s[0:3], s32 offset:184 ; 4-byte Folded Spill
; VI-NEXT: v_lshrrev_b32_e32 v34, 16, v1
-; VI-NEXT: buffer_store_dword v33, off, s[0:3], s32 offset:380 ; 4-byte Folded Spill
-; VI-NEXT: v_lshrrev_b32_e32 v33, 8, v17
-; VI-NEXT: buffer_store_dword v45, off, s[0:3], s32 offset:72 ; 4-byte Folded Spill
+; VI-NEXT: v_lshrrev_b32_e32 v49, 24, v18
+; VI-NEXT: buffer_store_dword v33, off, s[0:3], s32 offset:392 ; 4-byte Folded Spill
+; VI-NEXT: v_lshrrev_b32_e32 v52, 8, v18
+; VI-NEXT: v_lshrrev_b32_e32 v33, 16, v17
+; VI-NEXT: v_lshrrev_b32_e32 v36, 8, v17
+; VI-NEXT: v_lshrrev_b32_e32 v45, 24, v16
; VI-NEXT: buffer_store_dword v55, off, s[0:3], s32 offset:76 ; 4-byte Folded Spill
-; VI-NEXT: v_lshrrev_b32_e32 v55, 16, v15
; VI-NEXT: v_lshrrev_b32_e32 v50, 8, v15
; VI-NEXT: v_lshrrev_b32_e32 v43, 24, v14
-; VI-NEXT: v_lshrrev_b32_e32 v46, 16, v14
+; VI-NEXT: v_lshrrev_b32_e32 v55, 16, v14
; VI-NEXT: v_lshrrev_b32_e32 v48, 8, v14
; VI-NEXT: v_lshrrev_b32_e32 v47, 16, v13
; VI-NEXT: buffer_store_dword v56, off, s[0:3], s32 offset:88 ; 4-byte Folded Spill
@@ -172936,23 +173046,21 @@ define inreg <128 x i8> @bitcast_v64bf16_to_v128i8_scalar(<64 x bfloat> inreg %a
; VI-NEXT: v_lshrrev_b32_e32 v40, 8, v10
; VI-NEXT: buffer_store_dword v57, off, s[0:3], s32 offset:108 ; 4-byte Folded Spill
; VI-NEXT: v_lshrrev_b32_e32 v57, 8, v9
-; VI-NEXT: buffer_store_dword v35, off, s[0:3], s32 offset:112 ; 4-byte Folded Spill
-; VI-NEXT: v_lshrrev_b32_e32 v35, 8, v8
-; VI-NEXT: v_lshrrev_b32_e32 v37, 16, v7
-; VI-NEXT: buffer_store_dword v58, off, s[0:3], s32 offset:128 ; 4-byte Folded Spill
-; VI-NEXT: v_lshrrev_b32_e32 v59, 24, v6
-; VI-NEXT: v_lshrrev_b32_e32 v58, 16, v6
-; VI-NEXT: v_lshrrev_b32_e32 v60, 8, v6
-; VI-NEXT: v_lshrrev_b32_e32 v38, 16, v5
-; VI-NEXT: v_lshrrev_b32_e32 v45, 8, v5
-; VI-NEXT: v_lshrrev_b32_e32 v42, 24, v4
-; VI-NEXT: v_lshrrev_b32_e32 v39, 16, v4
-; VI-NEXT: v_lshrrev_b32_e32 v61, 8, v4
-; VI-NEXT: v_lshrrev_b32_e32 v52, 16, v3
-; VI-NEXT: v_lshrrev_b32_e32 v49, 8, v3
-; VI-NEXT: buffer_store_dword v34, off, s[0:3], s32 offset:184 ; 4-byte Folded Spill
+; VI-NEXT: v_lshrrev_b32_e32 v37, 24, v8
+; VI-NEXT: v_lshrrev_b32_e32 v35, 16, v8
+; VI-NEXT: v_lshrrev_b32_e32 v58, 8, v8
+; VI-NEXT: v_lshrrev_b32_e32 v59, 16, v7
+; VI-NEXT: buffer_store_dword v60, off, s[0:3], s32 offset:120 ; 4-byte Folded Spill
+; VI-NEXT: v_lshrrev_b32_e32 v38, 24, v6
+; VI-NEXT: v_lshrrev_b32_e32 v60, 16, v6
+; VI-NEXT: v_lshrrev_b32_e32 v61, 8, v6
+; VI-NEXT: v_lshrrev_b32_e32 v39, 16, v5
+; VI-NEXT: buffer_store_dword v42, off, s[0:3], s32 offset:132 ; 4-byte Folded Spill
+; VI-NEXT: v_lshrrev_b32_e32 v42, 16, v4
+; VI-NEXT: buffer_store_dword v34, off, s[0:3], s32 offset:192 ; 4-byte Folded Spill
; VI-NEXT: v_lshrrev_b32_e32 v34, 8, v1
-; VI-NEXT: buffer_store_dword v34, off, s[0:3], s32 offset:180 ; 4-byte Folded Spill
+; VI-NEXT: buffer_store_dword v45, off, s[0:3], s32 offset:72 ; 4-byte Folded Spill
+; VI-NEXT: buffer_store_dword v34, off, s[0:3], s32 offset:188 ; 4-byte Folded Spill
; VI-NEXT: s_branch .LBB91_5
; VI-NEXT: .LBB91_3:
; VI-NEXT: ; implicit-def: $sgpr46
@@ -173112,23 +173220,29 @@ define inreg <128 x i8> @bitcast_v64bf16_to_v128i8_scalar(<64 x bfloat> inreg %a
; VI-NEXT: s_branch .LBB91_2
; VI-NEXT: .LBB91_4:
; VI-NEXT: v_mov_b32_e32 v33, s71
-; VI-NEXT: buffer_store_dword v33, off, s[0:3], s32 offset:180 ; 4-byte Folded Spill
+; VI-NEXT: buffer_store_dword v33, off, s[0:3], s32 offset:188 ; 4-byte Folded Spill
; VI-NEXT: v_mov_b32_e32 v33, s69
-; VI-NEXT: buffer_store_dword v33, off, s[0:3], s32 offset:184 ; 4-byte Folded Spill
+; VI-NEXT: buffer_store_dword v33, off, s[0:3], s32 offset:192 ; 4-byte Folded Spill
; VI-NEXT: v_mov_b32_e32 v33, s70
-; VI-NEXT: buffer_store_dword v33, off, s[0:3], s32 offset:176 ; 4-byte Folded Spill
+; VI-NEXT: buffer_store_dword v33, off, s[0:3], s32 offset:184 ; 4-byte Folded Spill
; VI-NEXT: v_mov_b32_e32 v33, s68
-; VI-NEXT: buffer_store_dword v33, off, s[0:3], s32 offset:160 ; 4-byte Folded Spill
+; VI-NEXT: buffer_store_dword v33, off, s[0:3], s32 offset:168 ; 4-byte Folded Spill
; VI-NEXT: v_mov_b32_e32 v33, s67
-; VI-NEXT: buffer_store_dword v33, off, s[0:3], s32 offset:164 ; 4-byte Folded Spill
-; VI-NEXT: v_mov_b32_e32 v33, s86
-; VI-NEXT: buffer_store_dword v33, off, s[0:3], s32 offset:128 ; 4-byte Folded Spill
-; VI-NEXT: v_mov_b32_e32 v33, s83
+; VI-NEXT: buffer_store_dword v33, off, s[0:3], s32 offset:172 ; 4-byte Folded Spill
+; VI-NEXT: v_mov_b32_e32 v33, s66
+; VI-NEXT: buffer_store_dword v33, off, s[0:3], s32 offset:152 ; 4-byte Folded Spill
+; VI-NEXT: v_mov_b32_e32 v33, s64
+; VI-NEXT: buffer_store_dword v33, off, s[0:3], s32 offset:156 ; 4-byte Folded Spill
+; VI-NEXT: v_mov_b32_e32 v33, s65
+; VI-NEXT: buffer_store_dword v33, off, s[0:3], s32 offset:148 ; 4-byte Folded Spill
+; VI-NEXT: v_mov_b32_e32 v33, s54
+; VI-NEXT: buffer_store_dword v33, off, s[0:3], s32 offset:144 ; 4-byte Folded Spill
+; VI-NEXT: v_mov_b32_e32 v33, s53
; VI-NEXT: v_mov_b32_e32 v31, s4
-; VI-NEXT: buffer_store_dword v33, off, s[0:3], s32 offset:112 ; 4-byte Folded Spill
-; VI-NEXT: v_mov_b32_e32 v33, s82
+; VI-NEXT: buffer_store_dword v33, off, s[0:3], s32 offset:132 ; 4-byte Folded Spill
+; VI-NEXT: v_mov_b32_e32 v33, s86
; VI-NEXT: v_readlane_b32 s4, v62, 0
-; VI-NEXT: buffer_store_dword v33, off, s[0:3], s32 offset:124 ; 4-byte Folded Spill
+; VI-NEXT: buffer_store_dword v33, off, s[0:3], s32 offset:120 ; 4-byte Folded Spill
; VI-NEXT: v_mov_b32_e32 v33, s4
; VI-NEXT: v_readlane_b32 s4, v62, 1
; VI-NEXT: v_mov_b32_e32 v40, s4
@@ -173160,171 +173274,170 @@ define inreg <128 x i8> @bitcast_v64bf16_to_v128i8_scalar(<64 x bfloat> inreg %a
; VI-NEXT: v_readlane_b32 s4, v62, 13
; VI-NEXT: v_mov_b32_e32 v46, s4
; VI-NEXT: v_readlane_b32 s4, v62, 14
-; VI-NEXT: v_mov_b32_e32 v50, s4
-; VI-NEXT: v_readlane_b32 s4, v62, 15
; VI-NEXT: buffer_store_dword v33, off, s[0:3], s32 offset:88 ; 4-byte Folded Spill
; VI-NEXT: v_mov_b32_e32 v33, s4
+; VI-NEXT: v_readlane_b32 s4, v62, 15
+; VI-NEXT: buffer_store_dword v33, off, s[0:3], s32 offset:344 ; 4-byte Folded Spill
+; VI-NEXT: v_mov_b32_e32 v33, s4
; VI-NEXT: v_readlane_b32 s4, v62, 16
; VI-NEXT: buffer_store_dword v33, off, s[0:3], s32 offset:76 ; 4-byte Folded Spill
; VI-NEXT: v_mov_b32_e32 v33, s4
; VI-NEXT: v_readlane_b32 s4, v62, 17
-; VI-NEXT: buffer_store_dword v33, off, s[0:3], s32 offset:148 ; 4-byte Folded Spill
+; VI-NEXT: buffer_store_dword v33, off, s[0:3], s32 offset:340 ; 4-byte Folded Spill
; VI-NEXT: v_mov_b32_e32 v33, s4
; VI-NEXT: v_readlane_b32 s4, v62, 18
+; VI-NEXT: v_mov_b32_e32 v36, s4
+; VI-NEXT: v_readlane_b32 s4, v62, 19
; VI-NEXT: buffer_store_dword v33, off, s[0:3], s32 offset:72 ; 4-byte Folded Spill
; VI-NEXT: v_mov_b32_e32 v33, s4
-; VI-NEXT: v_readlane_b32 s4, v62, 19
-; VI-NEXT: v_mov_b32_e32 v55, s4
; VI-NEXT: v_readlane_b32 s4, v62, 20
-; VI-NEXT: buffer_store_dword v55, off, s[0:3], s32 offset:380 ; 4-byte Folded Spill
-; VI-NEXT: v_mov_b32_e32 v55, s4
+; VI-NEXT: v_mov_b32_e32 v52, s4
; VI-NEXT: v_readlane_b32 s4, v62, 21
-; VI-NEXT: buffer_store_dword v55, off, s[0:3], s32 offset:388 ; 4-byte Folded Spill
-; VI-NEXT: v_mov_b32_e32 v55, s4
+; VI-NEXT: v_mov_b32_e32 v49, s4
; VI-NEXT: v_readlane_b32 s4, v62, 22
-; VI-NEXT: buffer_store_dword v55, off, s[0:3], s32 offset:376 ; 4-byte Folded Spill
-; VI-NEXT: v_mov_b32_e32 v55, s4
+; VI-NEXT: buffer_store_dword v49, off, s[0:3], s32 offset:392 ; 4-byte Folded Spill
+; VI-NEXT: v_mov_b32_e32 v49, s4
; VI-NEXT: v_readlane_b32 s4, v62, 23
-; VI-NEXT: buffer_store_dword v55, off, s[0:3], s32 offset:384 ; 4-byte Folded Spill
-; VI-NEXT: v_mov_b32_e32 v55, s4
+; VI-NEXT: v_mov_b32_e32 v50, s4
; VI-NEXT: v_readlane_b32 s4, v62, 24
-; VI-NEXT: buffer_store_dword v55, off, s[0:3], s32 offset:368 ; 4-byte Folded Spill
-; VI-NEXT: v_mov_b32_e32 v55, s4
+; VI-NEXT: buffer_store_dword v50, off, s[0:3], s32 offset:384 ; 4-byte Folded Spill
+; VI-NEXT: v_mov_b32_e32 v50, s4
; VI-NEXT: v_readlane_b32 s4, v62, 25
-; VI-NEXT: buffer_store_dword v55, off, s[0:3], s32 offset:372 ; 4-byte Folded Spill
-; VI-NEXT: v_mov_b32_e32 v55, s4
+; VI-NEXT: buffer_store_dword v50, off, s[0:3], s32 offset:388 ; 4-byte Folded Spill
+; VI-NEXT: v_mov_b32_e32 v50, s4
; VI-NEXT: v_readlane_b32 s4, v62, 26
-; VI-NEXT: buffer_store_dword v55, off, s[0:3], s32 offset:360 ; 4-byte Folded Spill
-; VI-NEXT: v_mov_b32_e32 v55, s4
+; VI-NEXT: buffer_store_dword v50, off, s[0:3], s32 offset:376 ; 4-byte Folded Spill
+; VI-NEXT: v_mov_b32_e32 v50, s4
; VI-NEXT: v_readlane_b32 s4, v62, 27
-; VI-NEXT: buffer_store_dword v55, off, s[0:3], s32 offset:356 ; 4-byte Folded Spill
-; VI-NEXT: v_mov_b32_e32 v55, s4
+; VI-NEXT: buffer_store_dword v50, off, s[0:3], s32 offset:372 ; 4-byte Folded Spill
+; VI-NEXT: v_mov_b32_e32 v50, s4
; VI-NEXT: v_readlane_b32 s4, v62, 28
-; VI-NEXT: buffer_store_dword v55, off, s[0:3], s32 offset:364 ; 4-byte Folded Spill
-; VI-NEXT: v_mov_b32_e32 v55, s4
+; VI-NEXT: buffer_store_dword v50, off, s[0:3], s32 offset:380 ; 4-byte Folded Spill
+; VI-NEXT: v_mov_b32_e32 v50, s4
; VI-NEXT: v_readlane_b32 s4, v62, 29
-; VI-NEXT: buffer_store_dword v55, off, s[0:3], s32 offset:348 ; 4-byte Folded Spill
-; VI-NEXT: v_mov_b32_e32 v55, s4
+; VI-NEXT: buffer_store_dword v50, off, s[0:3], s32 offset:364 ; 4-byte Folded Spill
+; VI-NEXT: v_mov_b32_e32 v50, s4
; VI-NEXT: v_readlane_b32 s4, v62, 30
-; VI-NEXT: buffer_store_dword v55, off, s[0:3], s32 offset:352 ; 4-byte Folded Spill
-; VI-NEXT: v_mov_b32_e32 v55, s4
+; VI-NEXT: buffer_store_dword v50, off, s[0:3], s32 offset:368 ; 4-byte Folded Spill
+; VI-NEXT: v_mov_b32_e32 v50, s4
; VI-NEXT: v_readlane_b32 s4, v62, 31
-; VI-NEXT: buffer_store_dword v55, off, s[0:3], s32 offset:340 ; 4-byte Folded Spill
-; VI-NEXT: v_mov_b32_e32 v55, s4
+; VI-NEXT: buffer_store_dword v50, off, s[0:3], s32 offset:356 ; 4-byte Folded Spill
+; VI-NEXT: v_mov_b32_e32 v50, s4
; VI-NEXT: v_readlane_b32 s4, v62, 32
-; VI-NEXT: buffer_store_dword v55, off, s[0:3], s32 offset:336 ; 4-byte Folded Spill
-; VI-NEXT: v_mov_b32_e32 v55, s4
+; VI-NEXT: buffer_store_dword v50, off, s[0:3], s32 offset:352 ; 4-byte Folded Spill
+; VI-NEXT: v_mov_b32_e32 v50, s4
; VI-NEXT: v_readlane_b32 s4, v62, 33
-; VI-NEXT: buffer_store_dword v55, off, s[0:3], s32 offset:344 ; 4-byte Folded Spill
-; VI-NEXT: v_mov_b32_e32 v55, s4
+; VI-NEXT: v_mov_b32_e32 v41, s4
; VI-NEXT: v_readlane_b32 s4, v62, 34
-; VI-NEXT: buffer_store_dword v55, off, s[0:3], s32 offset:320 ; 4-byte Folded Spill
-; VI-NEXT: v_mov_b32_e32 v55, s4
+; VI-NEXT: buffer_store_dword v50, off, s[0:3], s32 offset:360 ; 4-byte Folded Spill
+; VI-NEXT: v_mov_b32_e32 v50, s4
; VI-NEXT: v_readlane_b32 s4, v62, 35
-; VI-NEXT: buffer_store_dword v55, off, s[0:3], s32 offset:332 ; 4-byte Folded Spill
-; VI-NEXT: v_mov_b32_e32 v55, s4
+; VI-NEXT: buffer_store_dword v41, off, s[0:3], s32 offset:328 ; 4-byte Folded Spill
+; VI-NEXT: v_mov_b32_e32 v41, s4
; VI-NEXT: v_readlane_b32 s4, v62, 36
-; VI-NEXT: buffer_store_dword v55, off, s[0:3], s32 offset:312 ; 4-byte Folded Spill
-; VI-NEXT: v_mov_b32_e32 v55, s4
+; VI-NEXT: buffer_store_dword v41, off, s[0:3], s32 offset:320 ; 4-byte Folded Spill
+; VI-NEXT: v_mov_b32_e32 v41, s4
; VI-NEXT: v_readlane_b32 s4, v62, 37
-; VI-NEXT: buffer_store_dword v55, off, s[0:3], s32 offset:308 ; 4-byte Folded Spill
-; VI-NEXT: v_mov_b32_e32 v55, s4
+; VI-NEXT: buffer_store_dword v41, off, s[0:3], s32 offset:316 ; 4-byte Folded Spill
+; VI-NEXT: v_mov_b32_e32 v41, s4
; VI-NEXT: v_readlane_b32 s4, v62, 38
-; VI-NEXT: buffer_store_dword v55, off, s[0:3], s32 offset:316 ; 4-byte Folded Spill
-; VI-NEXT: v_mov_b32_e32 v55, s4
+; VI-NEXT: buffer_store_dword v41, off, s[0:3], s32 offset:324 ; 4-byte Folded Spill
+; VI-NEXT: v_mov_b32_e32 v41, s4
; VI-NEXT: v_readlane_b32 s4, v62, 39
-; VI-NEXT: buffer_store_dword v55, off, s[0:3], s32 offset:300 ; 4-byte Folded Spill
-; VI-NEXT: v_mov_b32_e32 v55, s4
+; VI-NEXT: buffer_store_dword v41, off, s[0:3], s32 offset:308 ; 4-byte Folded Spill
+; VI-NEXT: v_mov_b32_e32 v41, s4
; VI-NEXT: v_readlane_b32 s4, v62, 40
-; VI-NEXT: buffer_store_dword v55, off, s[0:3], s32 offset:304 ; 4-byte Folded Spill
-; VI-NEXT: v_mov_b32_e32 v55, s4
+; VI-NEXT: buffer_store_dword v41, off, s[0:3], s32 offset:312 ; 4-byte Folded Spill
+; VI-NEXT: v_mov_b32_e32 v41, s4
; VI-NEXT: v_readlane_b32 s4, v62, 41
-; VI-NEXT: buffer_store_dword v55, off, s[0:3], s32 offset:284 ; 4-byte Folded Spill
-; VI-NEXT: v_mov_b32_e32 v55, s4
+; VI-NEXT: buffer_store_dword v41, off, s[0:3], s32 offset:292 ; 4-byte Folded Spill
+; VI-NEXT: v_mov_b32_e32 v41, s4
; VI-NEXT: v_readlane_b32 s4, v62, 42
-; VI-NEXT: buffer_store_dword v55, off, s[0:3], s32 offset:280 ; 4-byte Folded Spill
-; VI-NEXT: v_mov_b32_e32 v55, s4
+; VI-NEXT: buffer_store_dword v41, off, s[0:3], s32 offset:288 ; 4-byte Folded Spill
+; VI-NEXT: v_mov_b32_e32 v41, s4
; VI-NEXT: v_readlane_b32 s4, v62, 43
-; VI-NEXT: buffer_store_dword v55, off, s[0:3], s32 offset:288 ; 4-byte Folded Spill
-; VI-NEXT: v_mov_b32_e32 v55, s4
+; VI-NEXT: buffer_store_dword v41, off, s[0:3], s32 offset:296 ; 4-byte Folded Spill
+; VI-NEXT: v_mov_b32_e32 v41, s4
; VI-NEXT: v_readlane_b32 s4, v62, 44
-; VI-NEXT: buffer_store_dword v55, off, s[0:3], s32 offset:272 ; 4-byte Folded Spill
-; VI-NEXT: v_mov_b32_e32 v55, s4
+; VI-NEXT: buffer_store_dword v41, off, s[0:3], s32 offset:280 ; 4-byte Folded Spill
+; VI-NEXT: v_mov_b32_e32 v41, s4
; VI-NEXT: v_readlane_b32 s4, v62, 45
-; VI-NEXT: buffer_store_dword v55, off, s[0:3], s32 offset:276 ; 4-byte Folded Spill
-; VI-NEXT: v_mov_b32_e32 v55, s4
+; VI-NEXT: buffer_store_dword v41, off, s[0:3], s32 offset:284 ; 4-byte Folded Spill
+; VI-NEXT: v_mov_b32_e32 v41, s4
; VI-NEXT: v_readlane_b32 s4, v62, 46
-; VI-NEXT: buffer_store_dword v55, off, s[0:3], s32 offset:264 ; 4-byte Folded Spill
-; VI-NEXT: v_mov_b32_e32 v55, s4
+; VI-NEXT: buffer_store_dword v41, off, s[0:3], s32 offset:272 ; 4-byte Folded Spill
+; VI-NEXT: v_mov_b32_e32 v41, s4
; VI-NEXT: v_readlane_b32 s4, v62, 47
-; VI-NEXT: buffer_store_dword v55, off, s[0:3], s32 offset:260 ; 4-byte Folded Spill
-; VI-NEXT: v_mov_b32_e32 v55, s4
+; VI-NEXT: buffer_store_dword v41, off, s[0:3], s32 offset:268 ; 4-byte Folded Spill
+; VI-NEXT: v_mov_b32_e32 v41, s4
; VI-NEXT: v_readlane_b32 s4, v62, 48
-; VI-NEXT: buffer_store_dword v55, off, s[0:3], s32 offset:268 ; 4-byte Folded Spill
-; VI-NEXT: v_mov_b32_e32 v55, s4
+; VI-NEXT: buffer_store_dword v41, off, s[0:3], s32 offset:276 ; 4-byte Folded Spill
+; VI-NEXT: v_mov_b32_e32 v41, s4
; VI-NEXT: v_readlane_b32 s4, v62, 49
-; VI-NEXT: buffer_store_dword v55, off, s[0:3], s32 offset:244 ; 4-byte Folded Spill
-; VI-NEXT: v_mov_b32_e32 v55, s4
+; VI-NEXT: buffer_store_dword v41, off, s[0:3], s32 offset:252 ; 4-byte Folded Spill
+; VI-NEXT: v_mov_b32_e32 v41, s4
; VI-NEXT: v_readlane_b32 s4, v62, 50
-; VI-NEXT: buffer_store_dword v55, off, s[0:3], s32 offset:248 ; 4-byte Folded Spill
-; VI-NEXT: v_mov_b32_e32 v55, s4
+; VI-NEXT: buffer_store_dword v41, off, s[0:3], s32 offset:256 ; 4-byte Folded Spill
+; VI-NEXT: v_mov_b32_e32 v41, s4
; VI-NEXT: v_readlane_b32 s4, v62, 51
-; VI-NEXT: buffer_store_dword v55, off, s[0:3], s32 offset:236 ; 4-byte Folded Spill
-; VI-NEXT: v_mov_b32_e32 v55, s4
+; VI-NEXT: buffer_store_dword v41, off, s[0:3], s32 offset:244 ; 4-byte Folded Spill
+; VI-NEXT: v_mov_b32_e32 v41, s4
; VI-NEXT: v_readlane_b32 s4, v62, 52
-; VI-NEXT: buffer_store_dword v55, off, s[0:3], s32 offset:232 ; 4-byte Folded Spill
-; VI-NEXT: v_mov_b32_e32 v55, s4
+; VI-NEXT: buffer_store_dword v41, off, s[0:3], s32 offset:240 ; 4-byte Folded Spill
+; VI-NEXT: v_mov_b32_e32 v41, s4
; VI-NEXT: v_readlane_b32 s4, v62, 53
-; VI-NEXT: buffer_store_dword v55, off, s[0:3], s32 offset:240 ; 4-byte Folded Spill
-; VI-NEXT: v_mov_b32_e32 v55, s4
+; VI-NEXT: buffer_store_dword v41, off, s[0:3], s32 offset:248 ; 4-byte Folded Spill
+; VI-NEXT: v_mov_b32_e32 v41, s4
; VI-NEXT: v_readlane_b32 s4, v62, 54
-; VI-NEXT: buffer_store_dword v55, off, s[0:3], s32 offset:224 ; 4-byte Folded Spill
-; VI-NEXT: v_mov_b32_e32 v55, s4
+; VI-NEXT: buffer_store_dword v41, off, s[0:3], s32 offset:232 ; 4-byte Folded Spill
+; VI-NEXT: v_mov_b32_e32 v41, s4
; VI-NEXT: v_readlane_b32 s4, v62, 55
-; VI-NEXT: buffer_store_dword v55, off, s[0:3], s32 offset:228 ; 4-byte Folded Spill
-; VI-NEXT: v_mov_b32_e32 v55, s4
+; VI-NEXT: buffer_store_dword v41, off, s[0:3], s32 offset:236 ; 4-byte Folded Spill
+; VI-NEXT: v_mov_b32_e32 v41, s4
; VI-NEXT: v_readlane_b32 s4, v62, 56
-; VI-NEXT: buffer_store_dword v55, off, s[0:3], s32 offset:216 ; 4-byte Folded Spill
-; VI-NEXT: v_mov_b32_e32 v55, s4
+; VI-NEXT: buffer_store_dword v41, off, s[0:3], s32 offset:224 ; 4-byte Folded Spill
+; VI-NEXT: v_mov_b32_e32 v41, s4
; VI-NEXT: v_readlane_b32 s4, v62, 57
-; VI-NEXT: v_mov_b32_e32 v42, s54
-; VI-NEXT: buffer_store_dword v55, off, s[0:3], s32 offset:212 ; 4-byte Folded Spill
-; VI-NEXT: v_mov_b32_e32 v55, s4
+; VI-NEXT: buffer_store_dword v41, off, s[0:3], s32 offset:220 ; 4-byte Folded Spill
+; VI-NEXT: v_mov_b32_e32 v41, s4
+; VI-NEXT: v_mov_b32_e32 v42, s51
+; VI-NEXT: buffer_store_dword v41, off, s[0:3], s32 offset:228 ; 4-byte Folded Spill
; VI-NEXT: v_mov_b32_e32 v41, s46
-; VI-NEXT: buffer_store_dword v55, off, s[0:3], s32 offset:220 ; 4-byte Folded Spill
-; VI-NEXT: buffer_store_dword v41, off, s[0:3], s32 offset:168 ; 4-byte Folded Spill
-; VI-NEXT: buffer_store_dword v42, off, s[0:3], s32 offset:172 ; 4-byte Folded Spill
+; VI-NEXT: buffer_store_dword v50, off, s[0:3], s32 offset:348 ; 4-byte Folded Spill
+; VI-NEXT: buffer_store_dword v41, off, s[0:3], s32 offset:176 ; 4-byte Folded Spill
+; VI-NEXT: buffer_store_dword v42, off, s[0:3], s32 offset:180 ; 4-byte Folded Spill
; VI-NEXT: v_mov_b32_e32 v41, s56
-; VI-NEXT: buffer_store_dword v41, off, s[0:3], s32 offset:196 ; 4-byte Folded Spill
-; VI-NEXT: buffer_store_dword v42, off, s[0:3], s32 offset:200 ; 4-byte Folded Spill
-; VI-NEXT: v_mov_b32_e32 v41, s58
; VI-NEXT: buffer_store_dword v41, off, s[0:3], s32 offset:204 ; 4-byte Folded Spill
; VI-NEXT: buffer_store_dword v42, off, s[0:3], s32 offset:208 ; 4-byte Folded Spill
+; VI-NEXT: v_mov_b32_e32 v41, s58
+; VI-NEXT: buffer_store_dword v41, off, s[0:3], s32 offset:212 ; 4-byte Folded Spill
+; VI-NEXT: buffer_store_dword v42, off, s[0:3], s32 offset:216 ; 4-byte Folded Spill
; VI-NEXT: v_mov_b32_e32 v41, s60
; VI-NEXT: v_mov_b32_e32 v45, s72
; VI-NEXT: buffer_store_dword v41, off, s[0:3], s32 offset:56 ; 4-byte Folded Spill
; VI-NEXT: buffer_store_dword v42, off, s[0:3], s32 offset:60 ; 4-byte Folded Spill
-; VI-NEXT: buffer_store_dword v45, off, s[0:3], s32 offset:252 ; 4-byte Folded Spill
-; VI-NEXT: buffer_store_dword v46, off, s[0:3], s32 offset:256 ; 4-byte Folded Spill
+; VI-NEXT: buffer_store_dword v45, off, s[0:3], s32 offset:260 ; 4-byte Folded Spill
+; VI-NEXT: buffer_store_dword v46, off, s[0:3], s32 offset:264 ; 4-byte Folded Spill
; VI-NEXT: v_mov_b32_e32 v45, s74
-; VI-NEXT: buffer_store_dword v45, off, s[0:3], s32 offset:292 ; 4-byte Folded Spill
-; VI-NEXT: buffer_store_dword v46, off, s[0:3], s32 offset:296 ; 4-byte Folded Spill
+; VI-NEXT: buffer_store_dword v45, off, s[0:3], s32 offset:300 ; 4-byte Folded Spill
+; VI-NEXT: buffer_store_dword v46, off, s[0:3], s32 offset:304 ; 4-byte Folded Spill
; VI-NEXT: v_mov_b32_e32 v45, s76
-; VI-NEXT: buffer_store_dword v45, off, s[0:3], s32 offset:324 ; 4-byte Folded Spill
-; VI-NEXT: buffer_store_dword v46, off, s[0:3], s32 offset:328 ; 4-byte Folded Spill
+; VI-NEXT: buffer_store_dword v45, off, s[0:3], s32 offset:332 ; 4-byte Folded Spill
+; VI-NEXT: buffer_store_dword v46, off, s[0:3], s32 offset:336 ; 4-byte Folded Spill
; VI-NEXT: v_mov_b32_e32 v45, s78
; VI-NEXT: v_mov_b32_e32 v55, s88
; VI-NEXT: buffer_store_dword v45, off, s[0:3], s32 offset:64 ; 4-byte Folded Spill
; VI-NEXT: buffer_store_dword v46, off, s[0:3], s32 offset:68 ; 4-byte Folded Spill
; VI-NEXT: buffer_store_dword v55, off, s[0:3], s32 offset:80 ; 4-byte Folded Spill
; VI-NEXT: buffer_store_dword v56, off, s[0:3], s32 offset:84 ; 4-byte Folded Spill
-; VI-NEXT: v_mov_b32_e32 v36, s66
-; VI-NEXT: v_mov_b32_e32 v52, s64
-; VI-NEXT: v_mov_b32_e32 v55, v50
-; VI-NEXT: v_mov_b32_e32 v35, s30
-; VI-NEXT: v_mov_b32_e32 v59, s87
-; VI-NEXT: v_mov_b32_e32 v58, s34
-; VI-NEXT: v_mov_b32_e32 v45, s36
+; VI-NEXT: v_mov_b32_e32 v61, s50
+; VI-NEXT: v_mov_b32_e32 v58, s83
+; VI-NEXT: v_mov_b32_e32 v55, v48
+; VI-NEXT: v_mov_b32_e32 v48, v47
+; VI-NEXT: v_mov_b32_e32 v57, s30
+; VI-NEXT: v_mov_b32_e32 v35, s83
+; VI-NEXT: v_mov_b32_e32 v60, s34
+; VI-NEXT: v_mov_b32_e32 v42, s36
; VI-NEXT: v_mov_b32_e32 v34, s38
; VI-NEXT: v_mov_b32_e32 v1, s44
; VI-NEXT: v_mov_b32_e32 v2, s45
@@ -173357,44 +173470,44 @@ define inreg <128 x i8> @bitcast_v64bf16_to_v128i8_scalar(<64 x bfloat> inreg %a
; VI-NEXT: v_mov_b32_e32 v29, s28
; VI-NEXT: v_mov_b32_e32 v30, s29
; VI-NEXT: v_mov_b32_e32 v32, s5
+; VI-NEXT: v_mov_b32_e32 v38, s87
+; VI-NEXT: v_mov_b32_e32 v37, s82
; VI-NEXT: v_mov_b32_e32 v41, s62
-; VI-NEXT: v_mov_b32_e32 v57, s81
-; VI-NEXT: v_mov_b32_e32 v37, s84
-; VI-NEXT: v_mov_b32_e32 v60, s52
-; VI-NEXT: v_mov_b32_e32 v38, s51
-; VI-NEXT: v_mov_b32_e32 v61, s65
-; VI-NEXT: v_mov_b32_e32 v49, s66
-; VI-NEXT: v_mov_b32_e32 v39, s55
+; VI-NEXT: v_mov_b32_e32 v59, s84
+; VI-NEXT: v_mov_b32_e32 v39, s51
; VI-NEXT: v_mov_b32_e32 v50, v46
-; VI-NEXT: v_mov_b32_e32 v46, v48
-; VI-NEXT: v_mov_b32_e32 v48, v47
; VI-NEXT: v_mov_b32_e32 v47, v56
; VI-NEXT: v_mov_b32_e32 v56, v51
; VI-NEXT: v_mov_b32_e32 v51, s90
; VI-NEXT: buffer_store_dword v51, off, s[0:3], s32 offset:96 ; 4-byte Folded Spill
; VI-NEXT: buffer_store_dword v52, off, s[0:3], s32 offset:100 ; 4-byte Folded Spill
-; VI-NEXT: buffer_store_dword v35, off, s[0:3], s32 offset:116 ; 4-byte Folded Spill
-; VI-NEXT: buffer_store_dword v36, off, s[0:3], s32 offset:120 ; 4-byte Folded Spill
-; VI-NEXT: v_mov_b32_e32 v35, s85
-; VI-NEXT: buffer_store_dword v58, off, s[0:3], s32 offset:132 ; 4-byte Folded Spill
-; VI-NEXT: buffer_store_dword v59, off, s[0:3], s32 offset:136 ; 4-byte Folded Spill
-; VI-NEXT: buffer_store_dword v45, off, s[0:3], s32 offset:140 ; 4-byte Folded Spill
-; VI-NEXT: buffer_store_dword v46, off, s[0:3], s32 offset:144 ; 4-byte Folded Spill
-; VI-NEXT: buffer_store_dword v34, off, s[0:3], s32 offset:152 ; 4-byte Folded Spill
-; VI-NEXT: buffer_store_dword v35, off, s[0:3], s32 offset:156 ; 4-byte Folded Spill
+; VI-NEXT: buffer_store_dword v57, off, s[0:3], s32 offset:112 ; 4-byte Folded Spill
+; VI-NEXT: buffer_store_dword v58, off, s[0:3], s32 offset:116 ; 4-byte Folded Spill
+; VI-NEXT: buffer_store_dword v60, off, s[0:3], s32 offset:124 ; 4-byte Folded Spill
+; VI-NEXT: buffer_store_dword v61, off, s[0:3], s32 offset:128 ; 4-byte Folded Spill
+; VI-NEXT: buffer_store_dword v42, off, s[0:3], s32 offset:136 ; 4-byte Folded Spill
+; VI-NEXT: buffer_store_dword v43, off, s[0:3], s32 offset:140 ; 4-byte Folded Spill
+; VI-NEXT: buffer_store_dword v34, off, s[0:3], s32 offset:160 ; 4-byte Folded Spill
+; VI-NEXT: buffer_store_dword v35, off, s[0:3], s32 offset:164 ; 4-byte Folded Spill
; VI-NEXT: v_mov_b32_e32 v34, s48
; VI-NEXT: v_mov_b32_e32 v51, v53
; VI-NEXT: v_mov_b32_e32 v53, v54
; VI-NEXT: v_mov_b32_e32 v54, v40
; VI-NEXT: v_mov_b32_e32 v40, s80
-; VI-NEXT: v_mov_b32_e32 v58, s50
-; VI-NEXT: v_mov_b32_e32 v45, s53
-; VI-NEXT: buffer_store_dword v34, off, s[0:3], s32 offset:188 ; 4-byte Folded Spill
-; VI-NEXT: buffer_store_dword v35, off, s[0:3], s32 offset:192 ; 4-byte Folded Spill
+; VI-NEXT: v_mov_b32_e32 v57, s81
+; VI-NEXT: v_mov_b32_e32 v58, s85
+; VI-NEXT: v_mov_b32_e32 v60, s50
+; VI-NEXT: v_mov_b32_e32 v61, s52
+; VI-NEXT: v_mov_b32_e32 v42, s55
+; VI-NEXT: buffer_store_dword v34, off, s[0:3], s32 offset:196 ; 4-byte Folded Spill
+; VI-NEXT: buffer_store_dword v35, off, s[0:3], s32 offset:200 ; 4-byte Folded Spill
; VI-NEXT: .LBB91_5: ; %end
-; VI-NEXT: v_lshlrev_b32_e32 v36, 8, v33
-; VI-NEXT: buffer_load_dword v33, off, s[0:3], s32 offset:388 ; 4-byte Folded Reload
+; VI-NEXT: buffer_load_dword v45, off, s[0:3], s32 offset:332 ; 4-byte Folded Reload
+; VI-NEXT: buffer_load_dword v46, off, s[0:3], s32 offset:336 ; 4-byte Folded Reload
+; VI-NEXT: v_lshlrev_b32_e32 v36, 8, v36
; VI-NEXT: v_or_b32_sdwa v17, v17, v36 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; VI-NEXT: v_lshlrev_b32_e32 v36, 8, v52
+; VI-NEXT: v_or_b32_sdwa v18, v18, v36 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
; VI-NEXT: v_readlane_b32 s87, v63, 31
; VI-NEXT: v_readlane_b32 s86, v63, 30
; VI-NEXT: v_readlane_b32 s85, v63, 29
@@ -173427,44 +173540,35 @@ define inreg <128 x i8> @bitcast_v64bf16_to_v128i8_scalar(<64 x bfloat> inreg %a
; VI-NEXT: v_readlane_b32 s34, v63, 2
; VI-NEXT: v_readlane_b32 s31, v63, 1
; VI-NEXT: v_readlane_b32 s30, v63, 0
-; VI-NEXT: s_waitcnt vmcnt(0)
-; VI-NEXT: v_lshlrev_b32_e32 v36, 8, v33
-; VI-NEXT: buffer_load_dword v33, off, s[0:3], s32 offset:324 ; 4-byte Folded Reload
-; VI-NEXT: buffer_load_dword v34, off, s[0:3], s32 offset:328 ; 4-byte Folded Reload
-; VI-NEXT: v_or_b32_sdwa v18, v18, v36 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
; VI-NEXT: s_waitcnt vmcnt(1)
-; VI-NEXT: v_lshlrev_b32_e32 v36, 8, v33
-; VI-NEXT: buffer_load_dword v33, off, s[0:3], s32 offset:380 ; 4-byte Folded Reload
-; VI-NEXT: s_waitcnt vmcnt(0)
+; VI-NEXT: v_lshlrev_b32_e32 v36, 8, v45
; VI-NEXT: v_or_b32_sdwa v34, v33, v36 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
; VI-NEXT: v_or_b32_sdwa v17, v17, v34 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_0 src1_sel:DWORD
; VI-NEXT: buffer_store_dword v17, v0, s[0:3], 0 offen
-; VI-NEXT: buffer_load_dword v17, off, s[0:3], s32 offset:384 ; 4-byte Folded Reload
-; VI-NEXT: buffer_load_dword v33, off, s[0:3], s32 offset:376 ; 4-byte Folded Reload
-; VI-NEXT: s_waitcnt vmcnt(1)
-; VI-NEXT: v_lshlrev_b32_e32 v17, 8, v17
+; VI-NEXT: buffer_load_dword v33, off, s[0:3], s32 offset:392 ; 4-byte Folded Reload
+; VI-NEXT: v_lshlrev_b32_e32 v17, 8, v49
; VI-NEXT: s_waitcnt vmcnt(0)
; VI-NEXT: v_or_b32_sdwa v17, v33, v17 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
; VI-NEXT: v_or_b32_sdwa v17, v18, v17 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_0 src1_sel:DWORD
; VI-NEXT: v_add_u32_e32 v18, vcc, 4, v0
; VI-NEXT: buffer_store_dword v17, v18, s[0:3], 0 offen
-; VI-NEXT: buffer_load_dword v17, off, s[0:3], s32 offset:292 ; 4-byte Folded Reload
-; VI-NEXT: buffer_load_dword v18, off, s[0:3], s32 offset:296 ; 4-byte Folded Reload
-; VI-NEXT: buffer_load_dword v18, off, s[0:3], s32 offset:372 ; 4-byte Folded Reload
+; VI-NEXT: buffer_load_dword v17, off, s[0:3], s32 offset:300 ; 4-byte Folded Reload
+; VI-NEXT: buffer_load_dword v18, off, s[0:3], s32 offset:304 ; 4-byte Folded Reload
+; VI-NEXT: buffer_load_dword v18, off, s[0:3], s32 offset:388 ; 4-byte Folded Reload
; VI-NEXT: s_waitcnt vmcnt(2)
; VI-NEXT: v_lshlrev_b32_e32 v17, 8, v17
; VI-NEXT: s_waitcnt vmcnt(0)
; VI-NEXT: v_or_b32_sdwa v17, v18, v17 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
-; VI-NEXT: buffer_load_dword v18, off, s[0:3], s32 offset:368 ; 4-byte Folded Reload
+; VI-NEXT: buffer_load_dword v18, off, s[0:3], s32 offset:384 ; 4-byte Folded Reload
; VI-NEXT: s_waitcnt vmcnt(0)
; VI-NEXT: v_lshlrev_b32_e32 v18, 8, v18
; VI-NEXT: v_or_b32_sdwa v18, v19, v18 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
; VI-NEXT: v_or_b32_sdwa v17, v18, v17 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_0 src1_sel:DWORD
; VI-NEXT: v_add_u32_e32 v18, vcc, 8, v0
; VI-NEXT: buffer_store_dword v17, v18, s[0:3], 0 offen
-; VI-NEXT: buffer_load_dword v17, off, s[0:3], s32 offset:360 ; 4-byte Folded Reload
-; VI-NEXT: buffer_load_dword v18, off, s[0:3], s32 offset:364 ; 4-byte Folded Reload
-; VI-NEXT: buffer_load_dword v19, off, s[0:3], s32 offset:356 ; 4-byte Folded Reload
+; VI-NEXT: buffer_load_dword v17, off, s[0:3], s32 offset:376 ; 4-byte Folded Reload
+; VI-NEXT: buffer_load_dword v18, off, s[0:3], s32 offset:380 ; 4-byte Folded Reload
+; VI-NEXT: buffer_load_dword v19, off, s[0:3], s32 offset:372 ; 4-byte Folded Reload
; VI-NEXT: s_waitcnt vmcnt(2)
; VI-NEXT: v_lshlrev_b32_e32 v17, 8, v17
; VI-NEXT: s_waitcnt vmcnt(1)
@@ -173475,23 +173579,23 @@ define inreg <128 x i8> @bitcast_v64bf16_to_v128i8_scalar(<64 x bfloat> inreg %a
; VI-NEXT: v_or_b32_sdwa v17, v17, v18 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_0 src1_sel:DWORD
; VI-NEXT: v_add_u32_e32 v18, vcc, 12, v0
; VI-NEXT: buffer_store_dword v17, v18, s[0:3], 0 offen
-; VI-NEXT: buffer_load_dword v17, off, s[0:3], s32 offset:252 ; 4-byte Folded Reload
-; VI-NEXT: buffer_load_dword v18, off, s[0:3], s32 offset:256 ; 4-byte Folded Reload
-; VI-NEXT: buffer_load_dword v18, off, s[0:3], s32 offset:352 ; 4-byte Folded Reload
+; VI-NEXT: buffer_load_dword v17, off, s[0:3], s32 offset:260 ; 4-byte Folded Reload
+; VI-NEXT: buffer_load_dword v18, off, s[0:3], s32 offset:264 ; 4-byte Folded Reload
+; VI-NEXT: buffer_load_dword v18, off, s[0:3], s32 offset:368 ; 4-byte Folded Reload
; VI-NEXT: s_waitcnt vmcnt(2)
; VI-NEXT: v_lshlrev_b32_e32 v17, 8, v17
; VI-NEXT: s_waitcnt vmcnt(0)
; VI-NEXT: v_or_b32_sdwa v17, v18, v17 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
-; VI-NEXT: buffer_load_dword v18, off, s[0:3], s32 offset:348 ; 4-byte Folded Reload
+; VI-NEXT: buffer_load_dword v18, off, s[0:3], s32 offset:364 ; 4-byte Folded Reload
; VI-NEXT: s_waitcnt vmcnt(0)
; VI-NEXT: v_lshlrev_b32_e32 v18, 8, v18
; VI-NEXT: v_or_b32_sdwa v18, v21, v18 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
; VI-NEXT: v_or_b32_sdwa v17, v18, v17 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_0 src1_sel:DWORD
; VI-NEXT: v_add_u32_e32 v18, vcc, 16, v0
; VI-NEXT: buffer_store_dword v17, v18, s[0:3], 0 offen
-; VI-NEXT: buffer_load_dword v17, off, s[0:3], s32 offset:340 ; 4-byte Folded Reload
-; VI-NEXT: buffer_load_dword v18, off, s[0:3], s32 offset:344 ; 4-byte Folded Reload
-; VI-NEXT: buffer_load_dword v19, off, s[0:3], s32 offset:336 ; 4-byte Folded Reload
+; VI-NEXT: buffer_load_dword v17, off, s[0:3], s32 offset:356 ; 4-byte Folded Reload
+; VI-NEXT: buffer_load_dword v18, off, s[0:3], s32 offset:360 ; 4-byte Folded Reload
+; VI-NEXT: buffer_load_dword v19, off, s[0:3], s32 offset:352 ; 4-byte Folded Reload
; VI-NEXT: s_waitcnt vmcnt(2)
; VI-NEXT: v_lshlrev_b32_e32 v17, 8, v17
; VI-NEXT: s_waitcnt vmcnt(1)
@@ -173502,20 +173606,20 @@ define inreg <128 x i8> @bitcast_v64bf16_to_v128i8_scalar(<64 x bfloat> inreg %a
; VI-NEXT: v_or_b32_sdwa v17, v17, v18 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_0 src1_sel:DWORD
; VI-NEXT: v_add_u32_e32 v18, vcc, 20, v0
; VI-NEXT: buffer_store_dword v17, v18, s[0:3], 0 offen
-; VI-NEXT: buffer_load_dword v18, off, s[0:3], s32 offset:332 ; 4-byte Folded Reload
+; VI-NEXT: buffer_load_dword v18, off, s[0:3], s32 offset:348 ; 4-byte Folded Reload
; VI-NEXT: v_lshlrev_b32_e32 v17, 8, v41
; VI-NEXT: s_waitcnt vmcnt(0)
; VI-NEXT: v_or_b32_sdwa v17, v18, v17 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
-; VI-NEXT: buffer_load_dword v18, off, s[0:3], s32 offset:320 ; 4-byte Folded Reload
+; VI-NEXT: buffer_load_dword v18, off, s[0:3], s32 offset:328 ; 4-byte Folded Reload
; VI-NEXT: s_waitcnt vmcnt(0)
; VI-NEXT: v_lshlrev_b32_e32 v18, 8, v18
; VI-NEXT: v_or_b32_sdwa v18, v23, v18 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
; VI-NEXT: v_or_b32_sdwa v17, v18, v17 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_0 src1_sel:DWORD
; VI-NEXT: v_add_u32_e32 v18, vcc, 24, v0
; VI-NEXT: buffer_store_dword v17, v18, s[0:3], 0 offen
-; VI-NEXT: buffer_load_dword v17, off, s[0:3], s32 offset:312 ; 4-byte Folded Reload
-; VI-NEXT: buffer_load_dword v18, off, s[0:3], s32 offset:316 ; 4-byte Folded Reload
-; VI-NEXT: buffer_load_dword v19, off, s[0:3], s32 offset:308 ; 4-byte Folded Reload
+; VI-NEXT: buffer_load_dword v17, off, s[0:3], s32 offset:320 ; 4-byte Folded Reload
+; VI-NEXT: buffer_load_dword v18, off, s[0:3], s32 offset:324 ; 4-byte Folded Reload
+; VI-NEXT: buffer_load_dword v19, off, s[0:3], s32 offset:316 ; 4-byte Folded Reload
; VI-NEXT: s_waitcnt vmcnt(2)
; VI-NEXT: v_lshlrev_b32_e32 v17, 8, v17
; VI-NEXT: s_waitcnt vmcnt(1)
@@ -173528,21 +173632,21 @@ define inreg <128 x i8> @bitcast_v64bf16_to_v128i8_scalar(<64 x bfloat> inreg %a
; VI-NEXT: buffer_store_dword v17, v18, s[0:3], 0 offen
; VI-NEXT: buffer_load_dword v17, off, s[0:3], s32 offset:56 ; 4-byte Folded Reload
; VI-NEXT: buffer_load_dword v18, off, s[0:3], s32 offset:60 ; 4-byte Folded Reload
-; VI-NEXT: buffer_load_dword v18, off, s[0:3], s32 offset:304 ; 4-byte Folded Reload
+; VI-NEXT: buffer_load_dword v18, off, s[0:3], s32 offset:312 ; 4-byte Folded Reload
; VI-NEXT: s_waitcnt vmcnt(2)
; VI-NEXT: v_lshlrev_b32_e32 v17, 8, v17
; VI-NEXT: s_waitcnt vmcnt(0)
; VI-NEXT: v_or_b32_sdwa v17, v18, v17 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
-; VI-NEXT: buffer_load_dword v18, off, s[0:3], s32 offset:300 ; 4-byte Folded Reload
+; VI-NEXT: buffer_load_dword v18, off, s[0:3], s32 offset:308 ; 4-byte Folded Reload
; VI-NEXT: s_waitcnt vmcnt(0)
; VI-NEXT: v_lshlrev_b32_e32 v18, 8, v18
; VI-NEXT: v_or_b32_sdwa v18, v25, v18 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
; VI-NEXT: v_or_b32_sdwa v17, v18, v17 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_0 src1_sel:DWORD
; VI-NEXT: v_add_u32_e32 v18, vcc, 32, v0
; VI-NEXT: buffer_store_dword v17, v18, s[0:3], 0 offen
-; VI-NEXT: buffer_load_dword v17, off, s[0:3], s32 offset:284 ; 4-byte Folded Reload
-; VI-NEXT: buffer_load_dword v18, off, s[0:3], s32 offset:288 ; 4-byte Folded Reload
-; VI-NEXT: buffer_load_dword v19, off, s[0:3], s32 offset:280 ; 4-byte Folded Reload
+; VI-NEXT: buffer_load_dword v17, off, s[0:3], s32 offset:292 ; 4-byte Folded Reload
+; VI-NEXT: buffer_load_dword v18, off, s[0:3], s32 offset:296 ; 4-byte Folded Reload
+; VI-NEXT: buffer_load_dword v19, off, s[0:3], s32 offset:288 ; 4-byte Folded Reload
; VI-NEXT: s_waitcnt vmcnt(2)
; VI-NEXT: v_lshlrev_b32_e32 v17, 8, v17
; VI-NEXT: s_waitcnt vmcnt(1)
@@ -173553,23 +173657,23 @@ define inreg <128 x i8> @bitcast_v64bf16_to_v128i8_scalar(<64 x bfloat> inreg %a
; VI-NEXT: v_or_b32_sdwa v17, v17, v18 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_0 src1_sel:DWORD
; VI-NEXT: v_add_u32_e32 v18, vcc, 36, v0
; VI-NEXT: buffer_store_dword v17, v18, s[0:3], 0 offen
-; VI-NEXT: buffer_load_dword v17, off, s[0:3], s32 offset:204 ; 4-byte Folded Reload
-; VI-NEXT: buffer_load_dword v18, off, s[0:3], s32 offset:208 ; 4-byte Folded Reload
-; VI-NEXT: buffer_load_dword v18, off, s[0:3], s32 offset:276 ; 4-byte Folded Reload
+; VI-NEXT: buffer_load_dword v17, off, s[0:3], s32 offset:212 ; 4-byte Folded Reload
+; VI-NEXT: buffer_load_dword v18, off, s[0:3], s32 offset:216 ; 4-byte Folded Reload
+; VI-NEXT: buffer_load_dword v18, off, s[0:3], s32 offset:284 ; 4-byte Folded Reload
; VI-NEXT: s_waitcnt vmcnt(2)
; VI-NEXT: v_lshlrev_b32_e32 v17, 8, v17
; VI-NEXT: s_waitcnt vmcnt(0)
; VI-NEXT: v_or_b32_sdwa v17, v18, v17 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
-; VI-NEXT: buffer_load_dword v18, off, s[0:3], s32 offset:272 ; 4-byte Folded Reload
+; VI-NEXT: buffer_load_dword v18, off, s[0:3], s32 offset:280 ; 4-byte Folded Reload
; VI-NEXT: s_waitcnt vmcnt(0)
; VI-NEXT: v_lshlrev_b32_e32 v18, 8, v18
; VI-NEXT: v_or_b32_sdwa v18, v27, v18 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
; VI-NEXT: v_or_b32_sdwa v17, v18, v17 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_0 src1_sel:DWORD
; VI-NEXT: v_add_u32_e32 v18, vcc, 40, v0
; VI-NEXT: buffer_store_dword v17, v18, s[0:3], 0 offen
-; VI-NEXT: buffer_load_dword v17, off, s[0:3], s32 offset:264 ; 4-byte Folded Reload
-; VI-NEXT: buffer_load_dword v18, off, s[0:3], s32 offset:268 ; 4-byte Folded Reload
-; VI-NEXT: buffer_load_dword v19, off, s[0:3], s32 offset:260 ; 4-byte Folded Reload
+; VI-NEXT: buffer_load_dword v17, off, s[0:3], s32 offset:272 ; 4-byte Folded Reload
+; VI-NEXT: buffer_load_dword v18, off, s[0:3], s32 offset:276 ; 4-byte Folded Reload
+; VI-NEXT: buffer_load_dword v19, off, s[0:3], s32 offset:268 ; 4-byte Folded Reload
; VI-NEXT: s_waitcnt vmcnt(2)
; VI-NEXT: v_lshlrev_b32_e32 v17, 8, v17
; VI-NEXT: s_waitcnt vmcnt(1)
@@ -173580,23 +173684,23 @@ define inreg <128 x i8> @bitcast_v64bf16_to_v128i8_scalar(<64 x bfloat> inreg %a
; VI-NEXT: v_or_b32_sdwa v17, v17, v18 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_0 src1_sel:DWORD
; VI-NEXT: v_add_u32_e32 v18, vcc, 44, v0
; VI-NEXT: buffer_store_dword v17, v18, s[0:3], 0 offen
-; VI-NEXT: buffer_load_dword v17, off, s[0:3], s32 offset:196 ; 4-byte Folded Reload
-; VI-NEXT: buffer_load_dword v18, off, s[0:3], s32 offset:200 ; 4-byte Folded Reload
-; VI-NEXT: buffer_load_dword v18, off, s[0:3], s32 offset:248 ; 4-byte Folded Reload
+; VI-NEXT: buffer_load_dword v17, off, s[0:3], s32 offset:204 ; 4-byte Folded Reload
+; VI-NEXT: buffer_load_dword v18, off, s[0:3], s32 offset:208 ; 4-byte Folded Reload
+; VI-NEXT: buffer_load_dword v18, off, s[0:3], s32 offset:256 ; 4-byte Folded Reload
; VI-NEXT: s_waitcnt vmcnt(2)
; VI-NEXT: v_lshlrev_b32_e32 v17, 8, v17
; VI-NEXT: s_waitcnt vmcnt(0)
; VI-NEXT: v_or_b32_sdwa v17, v18, v17 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
-; VI-NEXT: buffer_load_dword v18, off, s[0:3], s32 offset:244 ; 4-byte Folded Reload
+; VI-NEXT: buffer_load_dword v18, off, s[0:3], s32 offset:252 ; 4-byte Folded Reload
; VI-NEXT: s_waitcnt vmcnt(0)
; VI-NEXT: v_lshlrev_b32_e32 v18, 8, v18
; VI-NEXT: v_or_b32_sdwa v18, v29, v18 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
; VI-NEXT: v_or_b32_sdwa v17, v18, v17 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_0 src1_sel:DWORD
; VI-NEXT: v_add_u32_e32 v18, vcc, 48, v0
; VI-NEXT: buffer_store_dword v17, v18, s[0:3], 0 offen
-; VI-NEXT: buffer_load_dword v17, off, s[0:3], s32 offset:236 ; 4-byte Folded Reload
-; VI-NEXT: buffer_load_dword v18, off, s[0:3], s32 offset:240 ; 4-byte Folded Reload
-; VI-NEXT: buffer_load_dword v19, off, s[0:3], s32 offset:232 ; 4-byte Folded Reload
+; VI-NEXT: buffer_load_dword v17, off, s[0:3], s32 offset:244 ; 4-byte Folded Reload
+; VI-NEXT: buffer_load_dword v18, off, s[0:3], s32 offset:248 ; 4-byte Folded Reload
+; VI-NEXT: buffer_load_dword v19, off, s[0:3], s32 offset:240 ; 4-byte Folded Reload
; VI-NEXT: s_waitcnt vmcnt(2)
; VI-NEXT: v_lshlrev_b32_e32 v17, 8, v17
; VI-NEXT: s_waitcnt vmcnt(1)
@@ -173607,23 +173711,23 @@ define inreg <128 x i8> @bitcast_v64bf16_to_v128i8_scalar(<64 x bfloat> inreg %a
; VI-NEXT: v_or_b32_sdwa v17, v17, v18 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_0 src1_sel:DWORD
; VI-NEXT: v_add_u32_e32 v18, vcc, 52, v0
; VI-NEXT: buffer_store_dword v17, v18, s[0:3], 0 offen
-; VI-NEXT: buffer_load_dword v17, off, s[0:3], s32 offset:168 ; 4-byte Folded Reload
-; VI-NEXT: buffer_load_dword v18, off, s[0:3], s32 offset:172 ; 4-byte Folded Reload
-; VI-NEXT: buffer_load_dword v18, off, s[0:3], s32 offset:228 ; 4-byte Folded Reload
+; VI-NEXT: buffer_load_dword v17, off, s[0:3], s32 offset:176 ; 4-byte Folded Reload
+; VI-NEXT: buffer_load_dword v18, off, s[0:3], s32 offset:180 ; 4-byte Folded Reload
+; VI-NEXT: buffer_load_dword v18, off, s[0:3], s32 offset:236 ; 4-byte Folded Reload
; VI-NEXT: s_waitcnt vmcnt(2)
; VI-NEXT: v_lshlrev_b32_e32 v17, 8, v17
; VI-NEXT: s_waitcnt vmcnt(0)
; VI-NEXT: v_or_b32_sdwa v17, v18, v17 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
-; VI-NEXT: buffer_load_dword v18, off, s[0:3], s32 offset:224 ; 4-byte Folded Reload
+; VI-NEXT: buffer_load_dword v18, off, s[0:3], s32 offset:232 ; 4-byte Folded Reload
; VI-NEXT: s_waitcnt vmcnt(0)
; VI-NEXT: v_lshlrev_b32_e32 v18, 8, v18
; VI-NEXT: v_or_b32_sdwa v18, v31, v18 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
; VI-NEXT: v_or_b32_sdwa v17, v18, v17 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_0 src1_sel:DWORD
; VI-NEXT: v_add_u32_e32 v18, vcc, 56, v0
; VI-NEXT: buffer_store_dword v17, v18, s[0:3], 0 offen
-; VI-NEXT: buffer_load_dword v17, off, s[0:3], s32 offset:216 ; 4-byte Folded Reload
-; VI-NEXT: buffer_load_dword v18, off, s[0:3], s32 offset:220 ; 4-byte Folded Reload
-; VI-NEXT: buffer_load_dword v19, off, s[0:3], s32 offset:212 ; 4-byte Folded Reload
+; VI-NEXT: buffer_load_dword v17, off, s[0:3], s32 offset:224 ; 4-byte Folded Reload
+; VI-NEXT: buffer_load_dword v18, off, s[0:3], s32 offset:228 ; 4-byte Folded Reload
+; VI-NEXT: buffer_load_dword v19, off, s[0:3], s32 offset:220 ; 4-byte Folded Reload
; VI-NEXT: s_waitcnt vmcnt(2)
; VI-NEXT: v_lshlrev_b32_e32 v17, 8, v17
; VI-NEXT: s_waitcnt vmcnt(1)
@@ -173634,90 +173738,95 @@ define inreg <128 x i8> @bitcast_v64bf16_to_v128i8_scalar(<64 x bfloat> inreg %a
; VI-NEXT: v_or_b32_sdwa v17, v17, v18 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_0 src1_sel:DWORD
; VI-NEXT: v_add_u32_e32 v18, vcc, 60, v0
; VI-NEXT: buffer_store_dword v17, v18, s[0:3], 0 offen
-; VI-NEXT: buffer_load_dword v17, off, s[0:3], s32 offset:188 ; 4-byte Folded Reload
+; VI-NEXT: buffer_load_dword v17, off, s[0:3], s32 offset:196 ; 4-byte Folded Reload
+; VI-NEXT: buffer_load_dword v18, off, s[0:3], s32 offset:200 ; 4-byte Folded Reload
; VI-NEXT: buffer_load_dword v18, off, s[0:3], s32 offset:192 ; 4-byte Folded Reload
-; VI-NEXT: buffer_load_dword v18, off, s[0:3], s32 offset:184 ; 4-byte Folded Reload
; VI-NEXT: s_waitcnt vmcnt(2)
; VI-NEXT: v_lshlrev_b32_e32 v17, 8, v17
; VI-NEXT: s_waitcnt vmcnt(0)
; VI-NEXT: v_or_b32_sdwa v17, v18, v17 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
-; VI-NEXT: buffer_load_dword v18, off, s[0:3], s32 offset:180 ; 4-byte Folded Reload
+; VI-NEXT: buffer_load_dword v18, off, s[0:3], s32 offset:188 ; 4-byte Folded Reload
; VI-NEXT: s_waitcnt vmcnt(0)
; VI-NEXT: v_lshlrev_b32_e32 v18, 8, v18
; VI-NEXT: v_or_b32_sdwa v1, v1, v18 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
; VI-NEXT: v_or_b32_sdwa v1, v1, v17 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_0 src1_sel:DWORD
; VI-NEXT: v_add_u32_e32 v17, vcc, 64, v0
; VI-NEXT: buffer_store_dword v1, v17, s[0:3], 0 offen
-; VI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:176 ; 4-byte Folded Reload
-; VI-NEXT: buffer_load_dword v17, off, s[0:3], s32 offset:160 ; 4-byte Folded Reload
+; VI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:184 ; 4-byte Folded Reload
+; VI-NEXT: buffer_load_dword v17, off, s[0:3], s32 offset:168 ; 4-byte Folded Reload
; VI-NEXT: s_waitcnt vmcnt(1)
; VI-NEXT: v_lshlrev_b32_e32 v1, 8, v1
; VI-NEXT: v_or_b32_sdwa v1, v2, v1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
-; VI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:164 ; 4-byte Folded Reload
+; VI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:172 ; 4-byte Folded Reload
; VI-NEXT: s_waitcnt vmcnt(0)
; VI-NEXT: v_lshlrev_b32_e32 v2, 8, v2
; VI-NEXT: v_or_b32_sdwa v2, v17, v2 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
; VI-NEXT: v_or_b32_sdwa v1, v1, v2 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_0 src1_sel:DWORD
; VI-NEXT: v_add_u32_e32 v2, vcc, 0x44, v0
; VI-NEXT: buffer_store_dword v1, v2, s[0:3], 0 offen
-; VI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:152 ; 4-byte Folded Reload
+; VI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:160 ; 4-byte Folded Reload
+; VI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:164 ; 4-byte Folded Reload
; VI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:156 ; 4-byte Folded Reload
+; VI-NEXT: s_waitcnt vmcnt(2)
+; VI-NEXT: v_lshlrev_b32_e32 v1, 8, v1
; VI-NEXT: s_waitcnt vmcnt(0)
-; VI-NEXT: v_lshlrev_b32_e32 v2, 8, v49
+; VI-NEXT: v_or_b32_sdwa v1, v2, v1 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; VI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:152 ; 4-byte Folded Reload
+; VI-NEXT: s_waitcnt vmcnt(0)
+; VI-NEXT: v_lshlrev_b32_e32 v2, 8, v2
; VI-NEXT: v_or_b32_sdwa v2, v3, v2 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
-; VI-NEXT: v_lshlrev_b32_e32 v1, 8, v1
-; VI-NEXT: v_or_b32_sdwa v1, v52, v1 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
; VI-NEXT: v_or_b32_sdwa v1, v2, v1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_0 src1_sel:DWORD
; VI-NEXT: v_add_u32_e32 v2, vcc, 0x48, v0
; VI-NEXT: buffer_store_dword v1, v2, s[0:3], 0 offen
-; VI-NEXT: v_lshlrev_b32_e32 v1, 8, v61
-; VI-NEXT: v_lshlrev_b32_e32 v2, 8, v42
+; VI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:148 ; 4-byte Folded Reload
+; VI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:144 ; 4-byte Folded Reload
+; VI-NEXT: s_waitcnt vmcnt(1)
+; VI-NEXT: v_lshlrev_b32_e32 v1, 8, v1
+; VI-NEXT: s_waitcnt vmcnt(0)
+; VI-NEXT: v_lshlrev_b32_e32 v2, 8, v2
; VI-NEXT: v_or_b32_sdwa v1, v4, v1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
-; VI-NEXT: v_or_b32_sdwa v2, v39, v2 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; VI-NEXT: v_or_b32_sdwa v2, v42, v2 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
; VI-NEXT: v_or_b32_sdwa v1, v1, v2 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_0 src1_sel:DWORD
; VI-NEXT: v_add_u32_e32 v2, vcc, 0x4c, v0
; VI-NEXT: buffer_store_dword v1, v2, s[0:3], 0 offen
-; VI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:140 ; 4-byte Folded Reload
-; VI-NEXT: buffer_load_dword v3, off, s[0:3], s32 offset:144 ; 4-byte Folded Reload
-; VI-NEXT: v_lshlrev_b32_e32 v1, 8, v45
-; VI-NEXT: v_or_b32_sdwa v1, v5, v1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; VI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:132 ; 4-byte Folded Reload
+; VI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:136 ; 4-byte Folded Reload
+; VI-NEXT: buffer_load_dword v3, off, s[0:3], s32 offset:140 ; 4-byte Folded Reload
; VI-NEXT: s_waitcnt vmcnt(1)
; VI-NEXT: v_lshlrev_b32_e32 v2, 8, v2
-; VI-NEXT: v_or_b32_sdwa v2, v38, v2 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; VI-NEXT: v_lshlrev_b32_e32 v1, 8, v1
+; VI-NEXT: v_or_b32_sdwa v1, v5, v1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; VI-NEXT: v_or_b32_sdwa v2, v39, v2 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
; VI-NEXT: v_or_b32_sdwa v1, v1, v2 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_0 src1_sel:DWORD
; VI-NEXT: v_add_u32_e32 v2, vcc, 0x50, v0
; VI-NEXT: buffer_store_dword v1, v2, s[0:3], 0 offen
-; VI-NEXT: v_lshlrev_b32_e32 v1, 8, v60
-; VI-NEXT: v_lshlrev_b32_e32 v2, 8, v59
+; VI-NEXT: v_lshlrev_b32_e32 v1, 8, v61
+; VI-NEXT: v_lshlrev_b32_e32 v2, 8, v38
; VI-NEXT: v_or_b32_sdwa v1, v6, v1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
-; VI-NEXT: v_or_b32_sdwa v2, v58, v2 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; VI-NEXT: v_or_b32_sdwa v2, v60, v2 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
; VI-NEXT: v_or_b32_sdwa v1, v1, v2 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_0 src1_sel:DWORD
; VI-NEXT: v_add_u32_e32 v2, vcc, 0x54, v0
; VI-NEXT: buffer_store_dword v1, v2, s[0:3], 0 offen
-; VI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:128 ; 4-byte Folded Reload
-; VI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:132 ; 4-byte Folded Reload
-; VI-NEXT: buffer_load_dword v3, off, s[0:3], s32 offset:136 ; 4-byte Folded Reload
+; VI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:120 ; 4-byte Folded Reload
+; VI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:124 ; 4-byte Folded Reload
+; VI-NEXT: buffer_load_dword v3, off, s[0:3], s32 offset:128 ; 4-byte Folded Reload
; VI-NEXT: s_waitcnt vmcnt(1)
; VI-NEXT: v_lshlrev_b32_e32 v2, 8, v2
; VI-NEXT: v_lshlrev_b32_e32 v1, 8, v1
; VI-NEXT: v_or_b32_sdwa v1, v7, v1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
-; VI-NEXT: v_or_b32_sdwa v2, v37, v2 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; VI-NEXT: v_or_b32_sdwa v2, v59, v2 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
; VI-NEXT: v_or_b32_sdwa v1, v1, v2 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_0 src1_sel:DWORD
; VI-NEXT: v_add_u32_e32 v2, vcc, 0x58, v0
; VI-NEXT: buffer_store_dword v1, v2, s[0:3], 0 offen
-; VI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:124 ; 4-byte Folded Reload
-; VI-NEXT: buffer_load_dword v3, off, s[0:3], s32 offset:112 ; 4-byte Folded Reload
-; VI-NEXT: v_lshlrev_b32_e32 v1, 8, v35
+; VI-NEXT: v_lshlrev_b32_e32 v1, 8, v58
+; VI-NEXT: v_lshlrev_b32_e32 v2, 8, v37
; VI-NEXT: v_or_b32_sdwa v1, v8, v1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
-; VI-NEXT: s_waitcnt vmcnt(1)
-; VI-NEXT: v_lshlrev_b32_e32 v2, 8, v2
-; VI-NEXT: s_waitcnt vmcnt(0)
-; VI-NEXT: v_or_b32_sdwa v2, v3, v2 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; VI-NEXT: v_or_b32_sdwa v2, v35, v2 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
; VI-NEXT: v_or_b32_sdwa v1, v1, v2 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_0 src1_sel:DWORD
; VI-NEXT: v_add_u32_e32 v2, vcc, 0x5c, v0
; VI-NEXT: buffer_store_dword v1, v2, s[0:3], 0 offen
-; VI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:116 ; 4-byte Folded Reload
-; VI-NEXT: buffer_load_dword v3, off, s[0:3], s32 offset:120 ; 4-byte Folded Reload
+; VI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:112 ; 4-byte Folded Reload
+; VI-NEXT: buffer_load_dword v3, off, s[0:3], s32 offset:116 ; 4-byte Folded Reload
; VI-NEXT: buffer_load_dword v3, off, s[0:3], s32 offset:108 ; 4-byte Folded Reload
; VI-NEXT: v_lshlrev_b32_e32 v1, 8, v57
; VI-NEXT: v_or_b32_sdwa v1, v9, v1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
@@ -173771,22 +173880,24 @@ define inreg <128 x i8> @bitcast_v64bf16_to_v128i8_scalar(<64 x bfloat> inreg %a
; VI-NEXT: v_lshlrev_b32_e32 v1, 8, v48
; VI-NEXT: v_lshlrev_b32_e32 v2, 8, v43
; VI-NEXT: v_or_b32_sdwa v1, v14, v1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
-; VI-NEXT: v_or_b32_sdwa v2, v46, v2 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; VI-NEXT: v_or_b32_sdwa v2, v55, v2 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
; VI-NEXT: v_or_b32_sdwa v1, v1, v2 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_0 src1_sel:DWORD
; VI-NEXT: v_add_u32_e32 v2, vcc, 0x74, v0
; VI-NEXT: buffer_store_dword v1, v2, s[0:3], 0 offen
; VI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:64 ; 4-byte Folded Reload
; VI-NEXT: buffer_load_dword v3, off, s[0:3], s32 offset:68 ; 4-byte Folded Reload
+; VI-NEXT: buffer_load_dword v3, off, s[0:3], s32 offset:344 ; 4-byte Folded Reload
; VI-NEXT: v_lshlrev_b32_e32 v1, 8, v50
; VI-NEXT: v_or_b32_sdwa v1, v15, v1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
-; VI-NEXT: s_waitcnt vmcnt(1)
+; VI-NEXT: s_waitcnt vmcnt(2)
; VI-NEXT: v_lshlrev_b32_e32 v2, 8, v2
-; VI-NEXT: v_or_b32_sdwa v2, v55, v2 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; VI-NEXT: s_waitcnt vmcnt(0)
+; VI-NEXT: v_or_b32_sdwa v2, v3, v2 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
; VI-NEXT: v_or_b32_sdwa v1, v1, v2 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_0 src1_sel:DWORD
; VI-NEXT: v_add_u32_e32 v2, vcc, 0x78, v0
; VI-NEXT: buffer_store_dword v1, v2, s[0:3], 0 offen
; VI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:76 ; 4-byte Folded Reload
-; VI-NEXT: buffer_load_dword v3, off, s[0:3], s32 offset:148 ; 4-byte Folded Reload
+; VI-NEXT: buffer_load_dword v3, off, s[0:3], s32 offset:340 ; 4-byte Folded Reload
; VI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:72 ; 4-byte Folded Reload
; VI-NEXT: v_add_u32_e32 v0, vcc, 0x7c, v0
; VI-NEXT: s_waitcnt vmcnt(2)
@@ -173812,8 +173923,8 @@ define inreg <128 x i8> @bitcast_v64bf16_to_v128i8_scalar(<64 x bfloat> inreg %a
; VI-NEXT: buffer_load_dword v41, off, s[0:3], s32 offset:48 ; 4-byte Folded Reload
; VI-NEXT: buffer_load_dword v40, off, s[0:3], s32 offset:52 ; 4-byte Folded Reload
; VI-NEXT: s_or_saveexec_b64 s[4:5], -1
-; VI-NEXT: buffer_load_dword v63, off, s[0:3], s32 offset:392 ; 4-byte Folded Reload
-; VI-NEXT: buffer_load_dword v62, off, s[0:3], s32 offset:396 ; 4-byte Folded Reload
+; VI-NEXT: buffer_load_dword v63, off, s[0:3], s32 offset:396 ; 4-byte Folded Reload
+; VI-NEXT: buffer_load_dword v62, off, s[0:3], s32 offset:400 ; 4-byte Folded Reload
; VI-NEXT: s_mov_b64 exec, s[4:5]
; VI-NEXT: s_waitcnt vmcnt(0)
; VI-NEXT: s_setpc_b64 s[30:31]
@@ -173876,8 +173987,8 @@ define inreg <128 x i8> @bitcast_v64bf16_to_v128i8_scalar(<64 x bfloat> inreg %a
; GFX9-NEXT: v_readfirstlane_b32 s59, v14
; GFX9-NEXT: v_readfirstlane_b32 s56, v15
; GFX9-NEXT: v_readfirstlane_b32 s57, v16
-; GFX9-NEXT: v_readfirstlane_b32 s46, v17
-; GFX9-NEXT: v_readfirstlane_b32 s47, v18
+; GFX9-NEXT: v_readfirstlane_b32 s44, v17
+; GFX9-NEXT: v_readfirstlane_b32 s45, v18
; GFX9-NEXT: v_readfirstlane_b32 s4, v1
; GFX9-NEXT: s_and_b64 s[6:7], vcc, exec
; GFX9-NEXT: v_readfirstlane_b32 s5, v2
@@ -173992,51 +174103,51 @@ define inreg <128 x i8> @bitcast_v64bf16_to_v128i8_scalar(<64 x bfloat> inreg %a
; GFX9-NEXT: v_writelane_b32 v62, s41, 3
; GFX9-NEXT: s_lshr_b64 s[40:41], s[22:23], 24
; GFX9-NEXT: v_writelane_b32 v62, s40, 0
-; GFX9-NEXT: s_lshr_b32 s70, s47, 24
-; GFX9-NEXT: s_lshr_b32 s15, s47, 16
-; GFX9-NEXT: s_lshr_b32 s7, s47, 8
-; GFX9-NEXT: s_lshr_b32 s53, s46, 16
-; GFX9-NEXT: s_lshr_b32 s52, s46, 8
-; GFX9-NEXT: s_lshr_b32 s67, s57, 24
+; GFX9-NEXT: s_lshr_b32 s53, s45, 24
+; GFX9-NEXT: s_lshr_b32 s15, s45, 16
+; GFX9-NEXT: s_lshr_b32 s70, s45, 8
+; GFX9-NEXT: s_lshr_b32 s7, s44, 16
+; GFX9-NEXT: s_lshr_b32 s6, s44, 8
+; GFX9-NEXT: s_lshr_b32 s65, s57, 24
; GFX9-NEXT: s_lshr_b32 s14, s57, 16
-; GFX9-NEXT: s_lshr_b32 s69, s57, 8
-; GFX9-NEXT: s_lshr_b32 s6, s56, 16
-; GFX9-NEXT: s_lshr_b32 s71, s56, 8
-; GFX9-NEXT: s_lshr_b32 s64, s59, 24
+; GFX9-NEXT: s_lshr_b32 s68, s57, 8
+; GFX9-NEXT: s_lshr_b32 s69, s56, 16
+; GFX9-NEXT: s_lshr_b32 s38, s56, 8
+; GFX9-NEXT: s_lshr_b32 s54, s59, 24
; GFX9-NEXT: s_lshr_b32 s13, s59, 16
-; GFX9-NEXT: s_lshr_b32 s66, s59, 8
-; GFX9-NEXT: s_lshr_b32 s51, s58, 16
-; GFX9-NEXT: s_lshr_b32 s68, s58, 8
-; GFX9-NEXT: s_lshr_b32 s99, s61, 24
+; GFX9-NEXT: s_lshr_b32 s52, s59, 8
+; GFX9-NEXT: s_lshr_b32 s67, s58, 16
+; GFX9-NEXT: s_lshr_b32 s66, s58, 8
+; GFX9-NEXT: s_lshr_b32 s97, s61, 24
; GFX9-NEXT: s_lshr_b32 s12, s61, 16
-; GFX9-NEXT: s_lshr_b32 s55, s61, 8
-; GFX9-NEXT: s_lshr_b32 s50, s60, 16
-; GFX9-NEXT: s_lshr_b32 s65, s60, 8
-; GFX9-NEXT: s_lshr_b32 s96, s63, 24
+; GFX9-NEXT: s_lshr_b32 s51, s61, 8
+; GFX9-NEXT: s_lshr_b32 s64, s60, 16
+; GFX9-NEXT: s_lshr_b32 s55, s60, 8
+; GFX9-NEXT: s_lshr_b32 s86, s63, 24
; GFX9-NEXT: s_lshr_b32 s11, s63, 16
-; GFX9-NEXT: s_lshr_b32 s98, s63, 8
-; GFX9-NEXT: s_lshr_b32 s49, s62, 16
-; GFX9-NEXT: s_lshr_b32 s54, s62, 8
-; GFX9-NEXT: s_lshr_b32 s85, s73, 24
+; GFX9-NEXT: s_lshr_b32 s50, s63, 8
+; GFX9-NEXT: s_lshr_b32 s99, s62, 16
+; GFX9-NEXT: s_lshr_b32 s98, s62, 8
+; GFX9-NEXT: s_lshr_b32 s83, s73, 24
; GFX9-NEXT: s_lshr_b32 s10, s73, 16
-; GFX9-NEXT: s_lshr_b32 s87, s73, 8
-; GFX9-NEXT: s_lshr_b32 s48, s72, 16
-; GFX9-NEXT: s_lshr_b32 s97, s72, 8
-; GFX9-NEXT: s_lshr_b32 s82, s75, 24
+; GFX9-NEXT: s_lshr_b32 s49, s73, 8
+; GFX9-NEXT: s_lshr_b32 s96, s72, 16
+; GFX9-NEXT: s_lshr_b32 s87, s72, 8
+; GFX9-NEXT: s_lshr_b32 s80, s75, 24
; GFX9-NEXT: s_lshr_b32 s9, s75, 16
-; GFX9-NEXT: s_lshr_b32 s84, s75, 8
-; GFX9-NEXT: s_lshr_b32 s39, s74, 16
-; GFX9-NEXT: s_lshr_b32 s86, s74, 8
-; GFX9-NEXT: s_lshr_b32 s80, s77, 24
+; GFX9-NEXT: s_lshr_b32 s48, s75, 8
+; GFX9-NEXT: s_lshr_b32 s85, s74, 16
+; GFX9-NEXT: s_lshr_b32 s84, s74, 8
+; GFX9-NEXT: s_lshr_b32 s71, s77, 24
; GFX9-NEXT: s_lshr_b32 s8, s77, 16
-; GFX9-NEXT: s_lshr_b32 s81, s77, 8
-; GFX9-NEXT: s_lshr_b32 s38, s76, 16
-; GFX9-NEXT: s_lshr_b32 s83, s76, 8
+; GFX9-NEXT: s_lshr_b32 s39, s77, 8
+; GFX9-NEXT: s_lshr_b32 s82, s76, 16
+; GFX9-NEXT: s_lshr_b32 s81, s76, 8
; GFX9-NEXT: v_writelane_b32 v62, s41, 1
; GFX9-NEXT: s_lshr_b64 s[40:41], s[20:21], 24
; GFX9-NEXT: s_lshr_b64 s[42:43], s[18:19], 24
-; GFX9-NEXT: s_lshr_b64 s[44:45], s[16:17], 24
-; GFX9-NEXT: s_lshr_b64 s[78:79], s[46:47], 24
+; GFX9-NEXT: s_lshr_b64 s[46:47], s[16:17], 24
+; GFX9-NEXT: s_lshr_b64 s[78:79], s[44:45], 24
; GFX9-NEXT: s_lshr_b64 s[88:89], s[56:57], 24
; GFX9-NEXT: s_lshr_b64 s[90:91], s[58:59], 24
; GFX9-NEXT: s_lshr_b64 s[92:93], s[60:61], 24
@@ -174047,698 +174158,697 @@ define inreg <128 x i8> @bitcast_v64bf16_to_v128i8_scalar(<64 x bfloat> inreg %a
; GFX9-NEXT: s_cbranch_execnz .LBB91_4
; GFX9-NEXT: .LBB91_2: ; %cmp.true
; GFX9-NEXT: s_and_b32 s6, s77, 0xffff0000
-; GFX9-NEXT: v_mov_b32_e32 v1, 0x40c00000
-; GFX9-NEXT: v_add_f32_e32 v2, s6, v1
-; GFX9-NEXT: v_bfe_u32 v3, v2, 16, 1
-; GFX9-NEXT: v_add_u32_e32 v3, v3, v2
-; GFX9-NEXT: v_add_u32_e32 v3, 0x7fff, v3
-; GFX9-NEXT: v_or_b32_e32 v4, 0x400000, v2
-; GFX9-NEXT: v_cmp_u_f32_e32 vcc, v2, v2
-; GFX9-NEXT: v_cndmask_b32_e32 v2, v3, v4, vcc
+; GFX9-NEXT: v_mov_b32_e32 v5, 0x40c00000
+; GFX9-NEXT: v_add_f32_e32 v1, s6, v5
+; GFX9-NEXT: v_bfe_u32 v2, v1, 16, 1
+; GFX9-NEXT: v_add_u32_e32 v2, v2, v1
+; GFX9-NEXT: v_add_u32_e32 v2, 0x7fff, v2
+; GFX9-NEXT: v_or_b32_e32 v3, 0x400000, v1
+; GFX9-NEXT: v_cmp_u_f32_e32 vcc, v1, v1
+; GFX9-NEXT: v_cndmask_b32_e32 v1, v2, v3, vcc
; GFX9-NEXT: s_lshl_b32 s6, s77, 16
-; GFX9-NEXT: v_lshrrev_b32_e32 v5, 16, v2
-; GFX9-NEXT: v_add_f32_e32 v2, s6, v1
-; GFX9-NEXT: v_bfe_u32 v3, v2, 16, 1
-; GFX9-NEXT: v_add_u32_e32 v3, v3, v2
-; GFX9-NEXT: v_add_u32_e32 v3, 0x7fff, v3
-; GFX9-NEXT: v_or_b32_e32 v4, 0x400000, v2
-; GFX9-NEXT: v_cmp_u_f32_e32 vcc, v2, v2
-; GFX9-NEXT: v_cndmask_b32_e32 v2, v3, v4, vcc
-; GFX9-NEXT: v_lshrrev_b32_e32 v2, 16, v2
-; GFX9-NEXT: buffer_store_dword v2, off, s[0:3], s32 offset:60 ; 4-byte Folded Spill
-; GFX9-NEXT: v_and_b32_e32 v2, 0xffff, v2
+; GFX9-NEXT: v_lshrrev_b32_e32 v4, 16, v1
+; GFX9-NEXT: v_add_f32_e32 v1, s6, v5
+; GFX9-NEXT: v_bfe_u32 v2, v1, 16, 1
+; GFX9-NEXT: v_add_u32_e32 v2, v2, v1
+; GFX9-NEXT: v_add_u32_e32 v2, 0x7fff, v2
+; GFX9-NEXT: v_or_b32_e32 v3, 0x400000, v1
+; GFX9-NEXT: v_cmp_u_f32_e32 vcc, v1, v1
+; GFX9-NEXT: v_cndmask_b32_e32 v1, v2, v3, vcc
+; GFX9-NEXT: v_lshrrev_b32_e32 v1, 16, v1
+; GFX9-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:60 ; 4-byte Folded Spill
+; GFX9-NEXT: v_and_b32_e32 v1, 0xffff, v1
; GFX9-NEXT: s_and_b32 s6, s76, 0xffff0000
-; GFX9-NEXT: v_lshl_or_b32 v8, v5, 16, v2
-; GFX9-NEXT: v_add_f32_e32 v2, s6, v1
-; GFX9-NEXT: v_bfe_u32 v3, v2, 16, 1
-; GFX9-NEXT: v_add_u32_e32 v3, v3, v2
+; GFX9-NEXT: v_lshl_or_b32 v2, v4, 16, v1
+; GFX9-NEXT: v_add_f32_e32 v1, s6, v5
+; GFX9-NEXT: v_bfe_u32 v3, v1, 16, 1
+; GFX9-NEXT: v_add_u32_e32 v3, v3, v1
+; GFX9-NEXT: buffer_store_dword v4, off, s[0:3], s32 offset:56 ; 4-byte Folded Spill
; GFX9-NEXT: v_add_u32_e32 v3, 0x7fff, v3
-; GFX9-NEXT: v_or_b32_e32 v4, 0x400000, v2
-; GFX9-NEXT: v_cmp_u_f32_e32 vcc, v2, v2
+; GFX9-NEXT: v_or_b32_e32 v4, 0x400000, v1
+; GFX9-NEXT: v_cmp_u_f32_e32 vcc, v1, v1
; GFX9-NEXT: s_lshl_b32 s6, s76, 16
-; GFX9-NEXT: v_cndmask_b32_e32 v2, v3, v4, vcc
-; GFX9-NEXT: v_add_f32_e32 v3, s6, v1
+; GFX9-NEXT: v_cndmask_b32_e32 v1, v3, v4, vcc
+; GFX9-NEXT: v_add_f32_e32 v3, s6, v5
; GFX9-NEXT: v_bfe_u32 v4, v3, 16, 1
; GFX9-NEXT: v_add_u32_e32 v4, v4, v3
-; GFX9-NEXT: buffer_store_dword v5, off, s[0:3], s32 offset:56 ; 4-byte Folded Spill
; GFX9-NEXT: v_add_u32_e32 v4, 0x7fff, v4
-; GFX9-NEXT: v_or_b32_e32 v5, 0x400000, v3
+; GFX9-NEXT: v_or_b32_e32 v6, 0x400000, v3
; GFX9-NEXT: v_cmp_u_f32_e32 vcc, v3, v3
-; GFX9-NEXT: v_cndmask_b32_e32 v3, v4, v5, vcc
+; GFX9-NEXT: v_cndmask_b32_e32 v3, v4, v6, vcc
; GFX9-NEXT: v_lshrrev_b32_e32 v3, 16, v3
-; GFX9-NEXT: v_lshrrev_b32_e32 v2, 16, v2
+; GFX9-NEXT: v_lshrrev_b32_e32 v1, 16, v1
; GFX9-NEXT: buffer_store_dword v3, off, s[0:3], s32 offset:68 ; 4-byte Folded Spill
; GFX9-NEXT: v_and_b32_e32 v3, 0xffff, v3
; GFX9-NEXT: s_and_b32 s6, s75, 0xffff0000
-; GFX9-NEXT: v_lshl_or_b32 v7, v2, 16, v3
-; GFX9-NEXT: v_add_f32_e32 v2, s6, v1
-; GFX9-NEXT: v_bfe_u32 v3, v2, 16, 1
-; GFX9-NEXT: v_add_u32_e32 v3, v3, v2
-; GFX9-NEXT: v_add_u32_e32 v3, 0x7fff, v3
-; GFX9-NEXT: v_or_b32_e32 v4, 0x400000, v2
-; GFX9-NEXT: v_cmp_u_f32_e32 vcc, v2, v2
-; GFX9-NEXT: v_cndmask_b32_e32 v2, v3, v4, vcc
+; GFX9-NEXT: v_lshl_or_b32 v1, v1, 16, v3
+; GFX9-NEXT: v_add_f32_e32 v3, s6, v5
+; GFX9-NEXT: v_bfe_u32 v4, v3, 16, 1
+; GFX9-NEXT: v_add_u32_e32 v4, v4, v3
+; GFX9-NEXT: v_add_u32_e32 v4, 0x7fff, v4
+; GFX9-NEXT: v_or_b32_e32 v6, 0x400000, v3
+; GFX9-NEXT: v_cmp_u_f32_e32 vcc, v3, v3
+; GFX9-NEXT: v_cndmask_b32_e32 v3, v4, v6, vcc
; GFX9-NEXT: s_lshl_b32 s6, s75, 16
-; GFX9-NEXT: v_lshrrev_b32_e32 v5, 16, v2
-; GFX9-NEXT: v_add_f32_e32 v2, s6, v1
-; GFX9-NEXT: v_bfe_u32 v3, v2, 16, 1
-; GFX9-NEXT: v_add_u32_e32 v3, v3, v2
-; GFX9-NEXT: v_add_u32_e32 v3, 0x7fff, v3
-; GFX9-NEXT: v_or_b32_e32 v4, 0x400000, v2
-; GFX9-NEXT: v_cmp_u_f32_e32 vcc, v2, v2
-; GFX9-NEXT: v_cndmask_b32_e32 v2, v3, v4, vcc
-; GFX9-NEXT: v_lshrrev_b32_e32 v31, 16, v2
-; GFX9-NEXT: v_and_b32_e32 v2, 0xffff, v31
-; GFX9-NEXT: s_and_b32 s6, s74, 0xffff0000
-; GFX9-NEXT: v_lshl_or_b32 v14, v5, 16, v2
-; GFX9-NEXT: v_add_f32_e32 v2, s6, v1
-; GFX9-NEXT: v_bfe_u32 v3, v2, 16, 1
-; GFX9-NEXT: v_add_u32_e32 v3, v3, v2
-; GFX9-NEXT: v_add_u32_e32 v3, 0x7fff, v3
-; GFX9-NEXT: v_or_b32_e32 v4, 0x400000, v2
-; GFX9-NEXT: v_cmp_u_f32_e32 vcc, v2, v2
-; GFX9-NEXT: s_lshl_b32 s6, s74, 16
-; GFX9-NEXT: v_cndmask_b32_e32 v2, v3, v4, vcc
-; GFX9-NEXT: v_add_f32_e32 v3, s6, v1
+; GFX9-NEXT: v_lshrrev_b32_e32 v7, 16, v3
+; GFX9-NEXT: v_add_f32_e32 v3, s6, v5
; GFX9-NEXT: v_bfe_u32 v4, v3, 16, 1
; GFX9-NEXT: v_add_u32_e32 v4, v4, v3
-; GFX9-NEXT: buffer_store_dword v5, off, s[0:3], s32 offset:64 ; 4-byte Folded Spill
; GFX9-NEXT: v_add_u32_e32 v4, 0x7fff, v4
-; GFX9-NEXT: v_or_b32_e32 v5, 0x400000, v3
+; GFX9-NEXT: v_or_b32_e32 v6, 0x400000, v3
; GFX9-NEXT: v_cmp_u_f32_e32 vcc, v3, v3
-; GFX9-NEXT: v_cndmask_b32_e32 v3, v4, v5, vcc
-; GFX9-NEXT: v_lshrrev_b32_e32 v33, 16, v3
-; GFX9-NEXT: v_lshrrev_b32_e32 v2, 16, v2
-; GFX9-NEXT: v_and_b32_e32 v3, 0xffff, v33
+; GFX9-NEXT: v_cndmask_b32_e32 v3, v4, v6, vcc
+; GFX9-NEXT: v_lshrrev_b32_e32 v32, 16, v3
+; GFX9-NEXT: v_and_b32_e32 v3, 0xffff, v32
+; GFX9-NEXT: s_and_b32 s6, s74, 0xffff0000
+; GFX9-NEXT: v_lshl_or_b32 v4, v7, 16, v3
+; GFX9-NEXT: v_add_f32_e32 v3, s6, v5
+; GFX9-NEXT: v_bfe_u32 v6, v3, 16, 1
+; GFX9-NEXT: v_add_u32_e32 v6, v6, v3
+; GFX9-NEXT: buffer_store_dword v7, off, s[0:3], s32 offset:64 ; 4-byte Folded Spill
+; GFX9-NEXT: v_add_u32_e32 v6, 0x7fff, v6
+; GFX9-NEXT: v_or_b32_e32 v7, 0x400000, v3
+; GFX9-NEXT: v_cmp_u_f32_e32 vcc, v3, v3
+; GFX9-NEXT: s_lshl_b32 s6, s74, 16
+; GFX9-NEXT: v_cndmask_b32_e32 v3, v6, v7, vcc
+; GFX9-NEXT: v_add_f32_e32 v6, s6, v5
+; GFX9-NEXT: v_bfe_u32 v7, v6, 16, 1
+; GFX9-NEXT: v_add_u32_e32 v7, v7, v6
+; GFX9-NEXT: v_add_u32_e32 v7, 0x7fff, v7
+; GFX9-NEXT: v_or_b32_e32 v8, 0x400000, v6
+; GFX9-NEXT: v_cmp_u_f32_e32 vcc, v6, v6
+; GFX9-NEXT: v_cndmask_b32_e32 v6, v7, v8, vcc
+; GFX9-NEXT: v_lshrrev_b32_e32 v34, 16, v6
+; GFX9-NEXT: v_lshrrev_b32_e32 v3, 16, v3
+; GFX9-NEXT: v_and_b32_e32 v6, 0xffff, v34
; GFX9-NEXT: s_and_b32 s6, s73, 0xffff0000
-; GFX9-NEXT: v_lshl_or_b32 v13, v2, 16, v3
-; GFX9-NEXT: v_add_f32_e32 v2, s6, v1
-; GFX9-NEXT: v_bfe_u32 v3, v2, 16, 1
-; GFX9-NEXT: v_add_u32_e32 v3, v3, v2
-; GFX9-NEXT: v_add_u32_e32 v3, 0x7fff, v3
-; GFX9-NEXT: v_or_b32_e32 v4, 0x400000, v2
-; GFX9-NEXT: v_cmp_u_f32_e32 vcc, v2, v2
-; GFX9-NEXT: v_cndmask_b32_e32 v2, v3, v4, vcc
+; GFX9-NEXT: v_lshl_or_b32 v3, v3, 16, v6
+; GFX9-NEXT: v_add_f32_e32 v6, s6, v5
+; GFX9-NEXT: v_bfe_u32 v7, v6, 16, 1
+; GFX9-NEXT: v_add_u32_e32 v7, v7, v6
+; GFX9-NEXT: v_add_u32_e32 v7, 0x7fff, v7
+; GFX9-NEXT: v_or_b32_e32 v8, 0x400000, v6
+; GFX9-NEXT: v_cmp_u_f32_e32 vcc, v6, v6
+; GFX9-NEXT: v_cndmask_b32_e32 v6, v7, v8, vcc
; GFX9-NEXT: s_lshl_b32 s6, s73, 16
-; GFX9-NEXT: v_lshrrev_b32_e32 v32, 16, v2
-; GFX9-NEXT: v_add_f32_e32 v2, s6, v1
-; GFX9-NEXT: v_bfe_u32 v3, v2, 16, 1
-; GFX9-NEXT: v_add_u32_e32 v3, v3, v2
-; GFX9-NEXT: v_add_u32_e32 v3, 0x7fff, v3
-; GFX9-NEXT: v_or_b32_e32 v4, 0x400000, v2
-; GFX9-NEXT: v_cmp_u_f32_e32 vcc, v2, v2
-; GFX9-NEXT: v_cndmask_b32_e32 v2, v3, v4, vcc
-; GFX9-NEXT: v_lshrrev_b32_e32 v34, 16, v2
-; GFX9-NEXT: v_and_b32_e32 v2, 0xffff, v34
+; GFX9-NEXT: v_lshrrev_b32_e32 v33, 16, v6
+; GFX9-NEXT: v_add_f32_e32 v6, s6, v5
+; GFX9-NEXT: v_bfe_u32 v7, v6, 16, 1
+; GFX9-NEXT: v_add_u32_e32 v7, v7, v6
+; GFX9-NEXT: v_add_u32_e32 v7, 0x7fff, v7
+; GFX9-NEXT: v_or_b32_e32 v8, 0x400000, v6
+; GFX9-NEXT: v_cmp_u_f32_e32 vcc, v6, v6
+; GFX9-NEXT: v_cndmask_b32_e32 v6, v7, v8, vcc
+; GFX9-NEXT: v_lshrrev_b32_e32 v35, 16, v6
+; GFX9-NEXT: v_and_b32_e32 v6, 0xffff, v35
; GFX9-NEXT: s_and_b32 s6, s72, 0xffff0000
-; GFX9-NEXT: v_lshl_or_b32 v16, v32, 16, v2
-; GFX9-NEXT: v_add_f32_e32 v2, s6, v1
-; GFX9-NEXT: v_bfe_u32 v3, v2, 16, 1
-; GFX9-NEXT: v_add_u32_e32 v3, v3, v2
-; GFX9-NEXT: v_add_u32_e32 v3, 0x7fff, v3
-; GFX9-NEXT: v_or_b32_e32 v4, 0x400000, v2
-; GFX9-NEXT: v_cmp_u_f32_e32 vcc, v2, v2
+; GFX9-NEXT: v_lshl_or_b32 v10, v33, 16, v6
+; GFX9-NEXT: v_add_f32_e32 v6, s6, v5
+; GFX9-NEXT: v_bfe_u32 v7, v6, 16, 1
+; GFX9-NEXT: v_add_u32_e32 v7, v7, v6
+; GFX9-NEXT: v_add_u32_e32 v7, 0x7fff, v7
+; GFX9-NEXT: v_or_b32_e32 v8, 0x400000, v6
+; GFX9-NEXT: v_cmp_u_f32_e32 vcc, v6, v6
; GFX9-NEXT: s_lshl_b32 s6, s72, 16
-; GFX9-NEXT: v_cndmask_b32_e32 v2, v3, v4, vcc
-; GFX9-NEXT: v_add_f32_e32 v3, s6, v1
-; GFX9-NEXT: v_bfe_u32 v4, v3, 16, 1
-; GFX9-NEXT: v_add_u32_e32 v4, v4, v3
-; GFX9-NEXT: v_add_u32_e32 v4, 0x7fff, v4
-; GFX9-NEXT: v_or_b32_e32 v5, 0x400000, v3
-; GFX9-NEXT: v_cmp_u_f32_e32 vcc, v3, v3
-; GFX9-NEXT: v_cndmask_b32_e32 v3, v4, v5, vcc
-; GFX9-NEXT: v_lshrrev_b32_e32 v36, 16, v3
-; GFX9-NEXT: v_lshrrev_b32_e32 v2, 16, v2
-; GFX9-NEXT: v_and_b32_e32 v3, 0xffff, v36
+; GFX9-NEXT: v_cndmask_b32_e32 v6, v7, v8, vcc
+; GFX9-NEXT: v_add_f32_e32 v7, s6, v5
+; GFX9-NEXT: v_bfe_u32 v8, v7, 16, 1
+; GFX9-NEXT: v_add_u32_e32 v8, v8, v7
+; GFX9-NEXT: v_add_u32_e32 v8, 0x7fff, v8
+; GFX9-NEXT: v_or_b32_e32 v9, 0x400000, v7
+; GFX9-NEXT: v_cmp_u_f32_e32 vcc, v7, v7
+; GFX9-NEXT: v_cndmask_b32_e32 v7, v8, v9, vcc
+; GFX9-NEXT: v_lshrrev_b32_e32 v37, 16, v7
+; GFX9-NEXT: v_lshrrev_b32_e32 v6, 16, v6
+; GFX9-NEXT: v_and_b32_e32 v7, 0xffff, v37
; GFX9-NEXT: s_and_b32 s6, s63, 0xffff0000
-; GFX9-NEXT: v_lshl_or_b32 v15, v2, 16, v3
-; GFX9-NEXT: v_add_f32_e32 v2, s6, v1
-; GFX9-NEXT: v_bfe_u32 v3, v2, 16, 1
-; GFX9-NEXT: v_add_u32_e32 v3, v3, v2
-; GFX9-NEXT: v_add_u32_e32 v3, 0x7fff, v3
-; GFX9-NEXT: v_or_b32_e32 v4, 0x400000, v2
-; GFX9-NEXT: v_cmp_u_f32_e32 vcc, v2, v2
-; GFX9-NEXT: v_cndmask_b32_e32 v2, v3, v4, vcc
+; GFX9-NEXT: v_lshl_or_b32 v9, v6, 16, v7
+; GFX9-NEXT: v_add_f32_e32 v6, s6, v5
+; GFX9-NEXT: v_bfe_u32 v7, v6, 16, 1
+; GFX9-NEXT: v_add_u32_e32 v7, v7, v6
+; GFX9-NEXT: v_add_u32_e32 v7, 0x7fff, v7
+; GFX9-NEXT: v_or_b32_e32 v8, 0x400000, v6
+; GFX9-NEXT: v_cmp_u_f32_e32 vcc, v6, v6
+; GFX9-NEXT: v_cndmask_b32_e32 v6, v7, v8, vcc
; GFX9-NEXT: s_lshl_b32 s6, s63, 16
-; GFX9-NEXT: v_lshrrev_b32_e32 v35, 16, v2
-; GFX9-NEXT: v_add_f32_e32 v2, s6, v1
-; GFX9-NEXT: v_bfe_u32 v3, v2, 16, 1
-; GFX9-NEXT: v_add_u32_e32 v3, v3, v2
-; GFX9-NEXT: v_add_u32_e32 v3, 0x7fff, v3
-; GFX9-NEXT: v_or_b32_e32 v4, 0x400000, v2
-; GFX9-NEXT: v_cmp_u_f32_e32 vcc, v2, v2
-; GFX9-NEXT: v_cndmask_b32_e32 v2, v3, v4, vcc
-; GFX9-NEXT: v_lshrrev_b32_e32 v37, 16, v2
-; GFX9-NEXT: v_and_b32_e32 v2, 0xffff, v37
+; GFX9-NEXT: v_lshrrev_b32_e32 v36, 16, v6
+; GFX9-NEXT: v_add_f32_e32 v6, s6, v5
+; GFX9-NEXT: v_bfe_u32 v7, v6, 16, 1
+; GFX9-NEXT: v_add_u32_e32 v7, v7, v6
+; GFX9-NEXT: v_add_u32_e32 v7, 0x7fff, v7
+; GFX9-NEXT: v_or_b32_e32 v8, 0x400000, v6
+; GFX9-NEXT: v_cmp_u_f32_e32 vcc, v6, v6
+; GFX9-NEXT: v_cndmask_b32_e32 v6, v7, v8, vcc
+; GFX9-NEXT: v_lshrrev_b32_e32 v38, 16, v6
+; GFX9-NEXT: v_and_b32_e32 v6, 0xffff, v38
; GFX9-NEXT: s_and_b32 s6, s62, 0xffff0000
-; GFX9-NEXT: v_lshl_or_b32 v18, v35, 16, v2
-; GFX9-NEXT: v_add_f32_e32 v2, s6, v1
-; GFX9-NEXT: v_bfe_u32 v3, v2, 16, 1
-; GFX9-NEXT: v_add_u32_e32 v3, v3, v2
-; GFX9-NEXT: v_add_u32_e32 v3, 0x7fff, v3
-; GFX9-NEXT: v_or_b32_e32 v4, 0x400000, v2
-; GFX9-NEXT: v_cmp_u_f32_e32 vcc, v2, v2
+; GFX9-NEXT: v_lshl_or_b32 v16, v36, 16, v6
+; GFX9-NEXT: v_add_f32_e32 v6, s6, v5
+; GFX9-NEXT: v_bfe_u32 v7, v6, 16, 1
+; GFX9-NEXT: v_add_u32_e32 v7, v7, v6
+; GFX9-NEXT: v_add_u32_e32 v7, 0x7fff, v7
+; GFX9-NEXT: v_or_b32_e32 v8, 0x400000, v6
+; GFX9-NEXT: v_cmp_u_f32_e32 vcc, v6, v6
; GFX9-NEXT: s_lshl_b32 s6, s62, 16
-; GFX9-NEXT: v_cndmask_b32_e32 v2, v3, v4, vcc
-; GFX9-NEXT: v_add_f32_e32 v3, s6, v1
-; GFX9-NEXT: v_bfe_u32 v4, v3, 16, 1
-; GFX9-NEXT: v_add_u32_e32 v4, v4, v3
-; GFX9-NEXT: v_add_u32_e32 v4, 0x7fff, v4
-; GFX9-NEXT: v_or_b32_e32 v5, 0x400000, v3
-; GFX9-NEXT: v_cmp_u_f32_e32 vcc, v3, v3
-; GFX9-NEXT: v_cndmask_b32_e32 v3, v4, v5, vcc
-; GFX9-NEXT: v_lshrrev_b32_e32 v39, 16, v3
-; GFX9-NEXT: v_lshrrev_b32_e32 v2, 16, v2
-; GFX9-NEXT: v_and_b32_e32 v3, 0xffff, v39
+; GFX9-NEXT: v_cndmask_b32_e32 v6, v7, v8, vcc
+; GFX9-NEXT: v_add_f32_e32 v7, s6, v5
+; GFX9-NEXT: v_bfe_u32 v8, v7, 16, 1
+; GFX9-NEXT: v_add_u32_e32 v8, v8, v7
+; GFX9-NEXT: v_add_u32_e32 v8, 0x7fff, v8
+; GFX9-NEXT: v_or_b32_e32 v11, 0x400000, v7
+; GFX9-NEXT: v_cmp_u_f32_e32 vcc, v7, v7
+; GFX9-NEXT: v_cndmask_b32_e32 v7, v8, v11, vcc
+; GFX9-NEXT: v_lshrrev_b32_e32 v48, 16, v7
+; GFX9-NEXT: v_lshrrev_b32_e32 v6, 16, v6
+; GFX9-NEXT: v_and_b32_e32 v7, 0xffff, v48
; GFX9-NEXT: s_and_b32 s6, s61, 0xffff0000
-; GFX9-NEXT: v_lshl_or_b32 v17, v2, 16, v3
-; GFX9-NEXT: v_add_f32_e32 v2, s6, v1
-; GFX9-NEXT: v_bfe_u32 v3, v2, 16, 1
-; GFX9-NEXT: v_add_u32_e32 v3, v3, v2
-; GFX9-NEXT: v_add_u32_e32 v3, 0x7fff, v3
-; GFX9-NEXT: v_or_b32_e32 v4, 0x400000, v2
-; GFX9-NEXT: v_cmp_u_f32_e32 vcc, v2, v2
-; GFX9-NEXT: v_cndmask_b32_e32 v2, v3, v4, vcc
+; GFX9-NEXT: v_lshl_or_b32 v15, v6, 16, v7
+; GFX9-NEXT: v_add_f32_e32 v6, s6, v5
+; GFX9-NEXT: v_bfe_u32 v7, v6, 16, 1
+; GFX9-NEXT: v_add_u32_e32 v7, v7, v6
+; GFX9-NEXT: v_add_u32_e32 v7, 0x7fff, v7
+; GFX9-NEXT: v_or_b32_e32 v8, 0x400000, v6
+; GFX9-NEXT: v_cmp_u_f32_e32 vcc, v6, v6
+; GFX9-NEXT: v_cndmask_b32_e32 v6, v7, v8, vcc
; GFX9-NEXT: s_lshl_b32 s6, s61, 16
-; GFX9-NEXT: v_lshrrev_b32_e32 v38, 16, v2
-; GFX9-NEXT: v_add_f32_e32 v2, s6, v1
-; GFX9-NEXT: v_bfe_u32 v3, v2, 16, 1
-; GFX9-NEXT: v_add_u32_e32 v3, v3, v2
-; GFX9-NEXT: v_add_u32_e32 v3, 0x7fff, v3
-; GFX9-NEXT: v_or_b32_e32 v4, 0x400000, v2
-; GFX9-NEXT: v_cmp_u_f32_e32 vcc, v2, v2
-; GFX9-NEXT: v_cndmask_b32_e32 v2, v3, v4, vcc
-; GFX9-NEXT: v_lshrrev_b32_e32 v48, 16, v2
-; GFX9-NEXT: v_and_b32_e32 v2, 0xffff, v48
+; GFX9-NEXT: v_lshrrev_b32_e32 v39, 16, v6
+; GFX9-NEXT: v_add_f32_e32 v6, s6, v5
+; GFX9-NEXT: v_bfe_u32 v7, v6, 16, 1
+; GFX9-NEXT: v_add_u32_e32 v7, v7, v6
+; GFX9-NEXT: v_add_u32_e32 v7, 0x7fff, v7
+; GFX9-NEXT: v_or_b32_e32 v8, 0x400000, v6
+; GFX9-NEXT: v_cmp_u_f32_e32 vcc, v6, v6
+; GFX9-NEXT: v_cndmask_b32_e32 v6, v7, v8, vcc
+; GFX9-NEXT: v_lshrrev_b32_e32 v49, 16, v6
+; GFX9-NEXT: v_and_b32_e32 v6, 0xffff, v49
; GFX9-NEXT: s_and_b32 s6, s60, 0xffff0000
-; GFX9-NEXT: v_lshl_or_b32 v20, v38, 16, v2
-; GFX9-NEXT: v_add_f32_e32 v2, s6, v1
-; GFX9-NEXT: v_bfe_u32 v3, v2, 16, 1
-; GFX9-NEXT: v_add_u32_e32 v3, v3, v2
-; GFX9-NEXT: v_add_u32_e32 v3, 0x7fff, v3
-; GFX9-NEXT: v_or_b32_e32 v4, 0x400000, v2
-; GFX9-NEXT: v_cmp_u_f32_e32 vcc, v2, v2
+; GFX9-NEXT: v_lshl_or_b32 v21, v39, 16, v6
+; GFX9-NEXT: v_add_f32_e32 v6, s6, v5
+; GFX9-NEXT: v_bfe_u32 v7, v6, 16, 1
+; GFX9-NEXT: v_add_u32_e32 v7, v7, v6
+; GFX9-NEXT: v_add_u32_e32 v7, 0x7fff, v7
+; GFX9-NEXT: v_or_b32_e32 v8, 0x400000, v6
+; GFX9-NEXT: v_cmp_u_f32_e32 vcc, v6, v6
; GFX9-NEXT: s_lshl_b32 s6, s60, 16
-; GFX9-NEXT: v_cndmask_b32_e32 v2, v3, v4, vcc
-; GFX9-NEXT: v_add_f32_e32 v3, s6, v1
-; GFX9-NEXT: v_bfe_u32 v4, v3, 16, 1
-; GFX9-NEXT: v_add_u32_e32 v4, v4, v3
-; GFX9-NEXT: v_add_u32_e32 v4, 0x7fff, v4
-; GFX9-NEXT: v_or_b32_e32 v5, 0x400000, v3
-; GFX9-NEXT: v_cmp_u_f32_e32 vcc, v3, v3
-; GFX9-NEXT: v_cndmask_b32_e32 v3, v4, v5, vcc
-; GFX9-NEXT: v_lshrrev_b32_e32 v50, 16, v3
-; GFX9-NEXT: v_lshrrev_b32_e32 v2, 16, v2
-; GFX9-NEXT: v_and_b32_e32 v3, 0xffff, v50
+; GFX9-NEXT: v_cndmask_b32_e32 v6, v7, v8, vcc
+; GFX9-NEXT: v_add_f32_e32 v7, s6, v5
+; GFX9-NEXT: v_bfe_u32 v8, v7, 16, 1
+; GFX9-NEXT: v_add_u32_e32 v8, v8, v7
+; GFX9-NEXT: v_add_u32_e32 v8, 0x7fff, v8
+; GFX9-NEXT: v_or_b32_e32 v11, 0x400000, v7
+; GFX9-NEXT: v_cmp_u_f32_e32 vcc, v7, v7
+; GFX9-NEXT: v_cndmask_b32_e32 v7, v8, v11, vcc
+; GFX9-NEXT: v_lshrrev_b32_e32 v51, 16, v7
+; GFX9-NEXT: v_lshrrev_b32_e32 v6, 16, v6
+; GFX9-NEXT: v_and_b32_e32 v7, 0xffff, v51
; GFX9-NEXT: s_and_b32 s6, s59, 0xffff0000
-; GFX9-NEXT: v_lshl_or_b32 v19, v2, 16, v3
-; GFX9-NEXT: v_add_f32_e32 v2, s6, v1
-; GFX9-NEXT: v_bfe_u32 v3, v2, 16, 1
-; GFX9-NEXT: v_add_u32_e32 v3, v3, v2
-; GFX9-NEXT: v_add_u32_e32 v3, 0x7fff, v3
-; GFX9-NEXT: v_or_b32_e32 v4, 0x400000, v2
-; GFX9-NEXT: v_cmp_u_f32_e32 vcc, v2, v2
-; GFX9-NEXT: v_cndmask_b32_e32 v2, v3, v4, vcc
+; GFX9-NEXT: v_lshl_or_b32 v20, v6, 16, v7
+; GFX9-NEXT: v_add_f32_e32 v6, s6, v5
+; GFX9-NEXT: v_bfe_u32 v7, v6, 16, 1
+; GFX9-NEXT: v_add_u32_e32 v7, v7, v6
+; GFX9-NEXT: v_add_u32_e32 v7, 0x7fff, v7
+; GFX9-NEXT: v_or_b32_e32 v8, 0x400000, v6
+; GFX9-NEXT: v_cmp_u_f32_e32 vcc, v6, v6
+; GFX9-NEXT: v_cndmask_b32_e32 v6, v7, v8, vcc
; GFX9-NEXT: s_lshl_b32 s6, s59, 16
-; GFX9-NEXT: v_lshrrev_b32_e32 v49, 16, v2
-; GFX9-NEXT: v_add_f32_e32 v2, s6, v1
-; GFX9-NEXT: v_bfe_u32 v3, v2, 16, 1
-; GFX9-NEXT: v_add_u32_e32 v3, v3, v2
-; GFX9-NEXT: v_add_u32_e32 v3, 0x7fff, v3
-; GFX9-NEXT: v_or_b32_e32 v4, 0x400000, v2
-; GFX9-NEXT: v_cmp_u_f32_e32 vcc, v2, v2
-; GFX9-NEXT: v_cndmask_b32_e32 v2, v3, v4, vcc
-; GFX9-NEXT: v_lshrrev_b32_e32 v51, 16, v2
-; GFX9-NEXT: v_and_b32_e32 v2, 0xffff, v51
+; GFX9-NEXT: v_lshrrev_b32_e32 v50, 16, v6
+; GFX9-NEXT: v_add_f32_e32 v6, s6, v5
+; GFX9-NEXT: v_bfe_u32 v7, v6, 16, 1
+; GFX9-NEXT: v_add_u32_e32 v7, v7, v6
+; GFX9-NEXT: v_add_u32_e32 v7, 0x7fff, v7
+; GFX9-NEXT: v_or_b32_e32 v8, 0x400000, v6
+; GFX9-NEXT: v_cmp_u_f32_e32 vcc, v6, v6
+; GFX9-NEXT: v_cndmask_b32_e32 v6, v7, v8, vcc
+; GFX9-NEXT: v_lshrrev_b32_e32 v52, 16, v6
+; GFX9-NEXT: v_and_b32_e32 v6, 0xffff, v52
; GFX9-NEXT: s_and_b32 s6, s58, 0xffff0000
-; GFX9-NEXT: v_lshl_or_b32 v22, v49, 16, v2
-; GFX9-NEXT: v_add_f32_e32 v2, s6, v1
-; GFX9-NEXT: v_bfe_u32 v3, v2, 16, 1
-; GFX9-NEXT: v_add_u32_e32 v3, v3, v2
-; GFX9-NEXT: v_add_u32_e32 v3, 0x7fff, v3
-; GFX9-NEXT: v_or_b32_e32 v4, 0x400000, v2
-; GFX9-NEXT: v_cmp_u_f32_e32 vcc, v2, v2
+; GFX9-NEXT: v_lshl_or_b32 v23, v50, 16, v6
+; GFX9-NEXT: v_add_f32_e32 v6, s6, v5
+; GFX9-NEXT: v_bfe_u32 v7, v6, 16, 1
+; GFX9-NEXT: v_add_u32_e32 v7, v7, v6
+; GFX9-NEXT: v_add_u32_e32 v7, 0x7fff, v7
+; GFX9-NEXT: v_or_b32_e32 v8, 0x400000, v6
+; GFX9-NEXT: v_cmp_u_f32_e32 vcc, v6, v6
; GFX9-NEXT: s_lshl_b32 s6, s58, 16
-; GFX9-NEXT: v_cndmask_b32_e32 v2, v3, v4, vcc
-; GFX9-NEXT: v_add_f32_e32 v3, s6, v1
-; GFX9-NEXT: v_bfe_u32 v4, v3, 16, 1
-; GFX9-NEXT: v_add_u32_e32 v4, v4, v3
-; GFX9-NEXT: v_add_u32_e32 v4, 0x7fff, v4
-; GFX9-NEXT: v_or_b32_e32 v5, 0x400000, v3
-; GFX9-NEXT: v_cmp_u_f32_e32 vcc, v3, v3
-; GFX9-NEXT: v_cndmask_b32_e32 v3, v4, v5, vcc
-; GFX9-NEXT: v_lshrrev_b32_e32 v53, 16, v3
-; GFX9-NEXT: v_lshrrev_b32_e32 v2, 16, v2
-; GFX9-NEXT: v_and_b32_e32 v3, 0xffff, v53
+; GFX9-NEXT: v_cndmask_b32_e32 v6, v7, v8, vcc
+; GFX9-NEXT: v_add_f32_e32 v7, s6, v5
+; GFX9-NEXT: v_bfe_u32 v8, v7, 16, 1
+; GFX9-NEXT: v_add_u32_e32 v8, v8, v7
+; GFX9-NEXT: v_add_u32_e32 v8, 0x7fff, v8
+; GFX9-NEXT: v_or_b32_e32 v11, 0x400000, v7
+; GFX9-NEXT: v_cmp_u_f32_e32 vcc, v7, v7
+; GFX9-NEXT: v_cndmask_b32_e32 v7, v8, v11, vcc
+; GFX9-NEXT: v_lshrrev_b32_e32 v54, 16, v7
+; GFX9-NEXT: v_lshrrev_b32_e32 v6, 16, v6
+; GFX9-NEXT: v_and_b32_e32 v7, 0xffff, v54
; GFX9-NEXT: s_and_b32 s6, s57, 0xffff0000
-; GFX9-NEXT: v_lshl_or_b32 v21, v2, 16, v3
-; GFX9-NEXT: v_add_f32_e32 v2, s6, v1
-; GFX9-NEXT: v_bfe_u32 v3, v2, 16, 1
-; GFX9-NEXT: v_add_u32_e32 v3, v3, v2
-; GFX9-NEXT: v_add_u32_e32 v3, 0x7fff, v3
-; GFX9-NEXT: v_or_b32_e32 v4, 0x400000, v2
-; GFX9-NEXT: v_cmp_u_f32_e32 vcc, v2, v2
-; GFX9-NEXT: v_cndmask_b32_e32 v2, v3, v4, vcc
+; GFX9-NEXT: v_lshl_or_b32 v22, v6, 16, v7
+; GFX9-NEXT: v_add_f32_e32 v6, s6, v5
+; GFX9-NEXT: v_bfe_u32 v7, v6, 16, 1
+; GFX9-NEXT: v_add_u32_e32 v7, v7, v6
+; GFX9-NEXT: v_add_u32_e32 v7, 0x7fff, v7
+; GFX9-NEXT: v_or_b32_e32 v8, 0x400000, v6
+; GFX9-NEXT: v_cmp_u_f32_e32 vcc, v6, v6
+; GFX9-NEXT: v_cndmask_b32_e32 v6, v7, v8, vcc
; GFX9-NEXT: s_lshl_b32 s6, s57, 16
-; GFX9-NEXT: v_lshrrev_b32_e32 v52, 16, v2
-; GFX9-NEXT: v_add_f32_e32 v2, s6, v1
-; GFX9-NEXT: v_bfe_u32 v3, v2, 16, 1
-; GFX9-NEXT: v_add_u32_e32 v3, v3, v2
-; GFX9-NEXT: v_add_u32_e32 v3, 0x7fff, v3
-; GFX9-NEXT: v_or_b32_e32 v4, 0x400000, v2
-; GFX9-NEXT: v_cmp_u_f32_e32 vcc, v2, v2
-; GFX9-NEXT: v_cndmask_b32_e32 v2, v3, v4, vcc
-; GFX9-NEXT: v_lshrrev_b32_e32 v54, 16, v2
-; GFX9-NEXT: v_and_b32_e32 v2, 0xffff, v54
+; GFX9-NEXT: v_lshrrev_b32_e32 v53, 16, v6
+; GFX9-NEXT: v_add_f32_e32 v6, s6, v5
+; GFX9-NEXT: v_bfe_u32 v7, v6, 16, 1
+; GFX9-NEXT: v_add_u32_e32 v7, v7, v6
+; GFX9-NEXT: v_add_u32_e32 v7, 0x7fff, v7
+; GFX9-NEXT: v_or_b32_e32 v8, 0x400000, v6
+; GFX9-NEXT: v_cmp_u_f32_e32 vcc, v6, v6
+; GFX9-NEXT: v_cndmask_b32_e32 v6, v7, v8, vcc
+; GFX9-NEXT: v_lshrrev_b32_e32 v55, 16, v6
+; GFX9-NEXT: v_and_b32_e32 v6, 0xffff, v55
; GFX9-NEXT: s_and_b32 s6, s56, 0xffff0000
-; GFX9-NEXT: v_lshl_or_b32 v24, v52, 16, v2
-; GFX9-NEXT: v_add_f32_e32 v2, s6, v1
-; GFX9-NEXT: v_bfe_u32 v3, v2, 16, 1
-; GFX9-NEXT: v_add_u32_e32 v3, v3, v2
-; GFX9-NEXT: v_add_u32_e32 v3, 0x7fff, v3
-; GFX9-NEXT: v_or_b32_e32 v4, 0x400000, v2
-; GFX9-NEXT: v_cmp_u_f32_e32 vcc, v2, v2
+; GFX9-NEXT: v_lshl_or_b32 v25, v53, 16, v6
+; GFX9-NEXT: v_add_f32_e32 v6, s6, v5
+; GFX9-NEXT: v_bfe_u32 v7, v6, 16, 1
+; GFX9-NEXT: v_add_u32_e32 v7, v7, v6
+; GFX9-NEXT: v_add_u32_e32 v7, 0x7fff, v7
+; GFX9-NEXT: v_or_b32_e32 v8, 0x400000, v6
+; GFX9-NEXT: v_cmp_u_f32_e32 vcc, v6, v6
; GFX9-NEXT: s_lshl_b32 s6, s56, 16
-; GFX9-NEXT: v_cndmask_b32_e32 v2, v3, v4, vcc
-; GFX9-NEXT: v_add_f32_e32 v3, s6, v1
-; GFX9-NEXT: v_bfe_u32 v4, v3, 16, 1
-; GFX9-NEXT: v_add_u32_e32 v4, v4, v3
-; GFX9-NEXT: v_add_u32_e32 v4, 0x7fff, v4
-; GFX9-NEXT: v_or_b32_e32 v5, 0x400000, v3
-; GFX9-NEXT: v_cmp_u_f32_e32 vcc, v3, v3
-; GFX9-NEXT: v_cndmask_b32_e32 v3, v4, v5, vcc
-; GFX9-NEXT: v_lshrrev_b32_e32 v40, 16, v3
-; GFX9-NEXT: v_lshrrev_b32_e32 v2, 16, v2
-; GFX9-NEXT: v_and_b32_e32 v3, 0xffff, v40
-; GFX9-NEXT: s_and_b32 s6, s47, 0xffff0000
-; GFX9-NEXT: v_lshl_or_b32 v23, v2, 16, v3
-; GFX9-NEXT: v_add_f32_e32 v2, s6, v1
-; GFX9-NEXT: v_bfe_u32 v3, v2, 16, 1
-; GFX9-NEXT: v_add_u32_e32 v3, v3, v2
-; GFX9-NEXT: v_add_u32_e32 v3, 0x7fff, v3
-; GFX9-NEXT: v_or_b32_e32 v4, 0x400000, v2
-; GFX9-NEXT: v_cmp_u_f32_e32 vcc, v2, v2
-; GFX9-NEXT: v_cndmask_b32_e32 v2, v3, v4, vcc
-; GFX9-NEXT: s_lshl_b32 s6, s47, 16
-; GFX9-NEXT: v_lshrrev_b32_e32 v55, 16, v2
-; GFX9-NEXT: v_add_f32_e32 v2, s6, v1
-; GFX9-NEXT: v_bfe_u32 v3, v2, 16, 1
-; GFX9-NEXT: v_add_u32_e32 v3, v3, v2
-; GFX9-NEXT: v_add_u32_e32 v3, 0x7fff, v3
-; GFX9-NEXT: v_or_b32_e32 v4, 0x400000, v2
-; GFX9-NEXT: v_cmp_u_f32_e32 vcc, v2, v2
-; GFX9-NEXT: v_cndmask_b32_e32 v2, v3, v4, vcc
-; GFX9-NEXT: v_lshrrev_b32_e32 v41, 16, v2
-; GFX9-NEXT: v_and_b32_e32 v2, 0xffff, v41
-; GFX9-NEXT: s_and_b32 s6, s46, 0xffff0000
-; GFX9-NEXT: v_lshl_or_b32 v26, v55, 16, v2
-; GFX9-NEXT: v_add_f32_e32 v2, s6, v1
-; GFX9-NEXT: v_bfe_u32 v3, v2, 16, 1
-; GFX9-NEXT: v_add_u32_e32 v3, v3, v2
-; GFX9-NEXT: v_add_u32_e32 v3, 0x7fff, v3
-; GFX9-NEXT: v_or_b32_e32 v4, 0x400000, v2
-; GFX9-NEXT: v_cmp_u_f32_e32 vcc, v2, v2
-; GFX9-NEXT: s_lshl_b32 s6, s46, 16
-; GFX9-NEXT: v_cndmask_b32_e32 v2, v3, v4, vcc
-; GFX9-NEXT: v_add_f32_e32 v3, s6, v1
-; GFX9-NEXT: v_bfe_u32 v4, v3, 16, 1
-; GFX9-NEXT: v_add_u32_e32 v4, v4, v3
-; GFX9-NEXT: v_add_u32_e32 v4, 0x7fff, v4
-; GFX9-NEXT: v_or_b32_e32 v5, 0x400000, v3
-; GFX9-NEXT: v_cmp_u_f32_e32 vcc, v3, v3
-; GFX9-NEXT: v_cndmask_b32_e32 v3, v4, v5, vcc
-; GFX9-NEXT: v_lshrrev_b32_e32 v42, 16, v3
-; GFX9-NEXT: v_lshrrev_b32_e32 v2, 16, v2
-; GFX9-NEXT: v_and_b32_e32 v3, 0xffff, v42
+; GFX9-NEXT: v_cndmask_b32_e32 v6, v7, v8, vcc
+; GFX9-NEXT: v_add_f32_e32 v7, s6, v5
+; GFX9-NEXT: v_bfe_u32 v8, v7, 16, 1
+; GFX9-NEXT: v_add_u32_e32 v8, v8, v7
+; GFX9-NEXT: v_add_u32_e32 v8, 0x7fff, v8
+; GFX9-NEXT: v_or_b32_e32 v11, 0x400000, v7
+; GFX9-NEXT: v_cmp_u_f32_e32 vcc, v7, v7
+; GFX9-NEXT: v_cndmask_b32_e32 v7, v8, v11, vcc
+; GFX9-NEXT: v_lshrrev_b32_e32 v41, 16, v7
+; GFX9-NEXT: v_lshrrev_b32_e32 v6, 16, v6
+; GFX9-NEXT: v_and_b32_e32 v7, 0xffff, v41
+; GFX9-NEXT: s_and_b32 s6, s45, 0xffff0000
+; GFX9-NEXT: v_lshl_or_b32 v24, v6, 16, v7
+; GFX9-NEXT: v_add_f32_e32 v6, s6, v5
+; GFX9-NEXT: v_bfe_u32 v7, v6, 16, 1
+; GFX9-NEXT: v_add_u32_e32 v7, v7, v6
+; GFX9-NEXT: v_add_u32_e32 v7, 0x7fff, v7
+; GFX9-NEXT: v_or_b32_e32 v8, 0x400000, v6
+; GFX9-NEXT: v_cmp_u_f32_e32 vcc, v6, v6
+; GFX9-NEXT: v_cndmask_b32_e32 v6, v7, v8, vcc
+; GFX9-NEXT: s_lshl_b32 s6, s45, 16
+; GFX9-NEXT: v_lshrrev_b32_e32 v40, 16, v6
+; GFX9-NEXT: v_add_f32_e32 v6, s6, v5
+; GFX9-NEXT: v_bfe_u32 v7, v6, 16, 1
+; GFX9-NEXT: v_add_u32_e32 v7, v7, v6
+; GFX9-NEXT: v_add_u32_e32 v7, 0x7fff, v7
+; GFX9-NEXT: v_or_b32_e32 v8, 0x400000, v6
+; GFX9-NEXT: v_cmp_u_f32_e32 vcc, v6, v6
+; GFX9-NEXT: v_cndmask_b32_e32 v6, v7, v8, vcc
+; GFX9-NEXT: v_lshrrev_b32_e32 v42, 16, v6
+; GFX9-NEXT: v_and_b32_e32 v6, 0xffff, v42
+; GFX9-NEXT: s_and_b32 s6, s44, 0xffff0000
+; GFX9-NEXT: v_lshl_or_b32 v27, v40, 16, v6
+; GFX9-NEXT: v_add_f32_e32 v6, s6, v5
+; GFX9-NEXT: v_bfe_u32 v7, v6, 16, 1
+; GFX9-NEXT: v_add_u32_e32 v7, v7, v6
+; GFX9-NEXT: v_add_u32_e32 v7, 0x7fff, v7
+; GFX9-NEXT: v_or_b32_e32 v8, 0x400000, v6
+; GFX9-NEXT: v_cmp_u_f32_e32 vcc, v6, v6
+; GFX9-NEXT: s_lshl_b32 s6, s44, 16
+; GFX9-NEXT: v_cndmask_b32_e32 v6, v7, v8, vcc
+; GFX9-NEXT: v_add_f32_e32 v7, s6, v5
+; GFX9-NEXT: v_bfe_u32 v8, v7, 16, 1
+; GFX9-NEXT: v_add_u32_e32 v8, v8, v7
+; GFX9-NEXT: v_add_u32_e32 v8, 0x7fff, v8
+; GFX9-NEXT: v_or_b32_e32 v11, 0x400000, v7
+; GFX9-NEXT: v_cmp_u_f32_e32 vcc, v7, v7
+; GFX9-NEXT: v_cndmask_b32_e32 v7, v8, v11, vcc
+; GFX9-NEXT: v_lshrrev_b32_e32 v43, 16, v7
+; GFX9-NEXT: v_lshrrev_b32_e32 v6, 16, v6
+; GFX9-NEXT: v_and_b32_e32 v7, 0xffff, v43
; GFX9-NEXT: s_and_b32 s6, s17, 0xffff0000
-; GFX9-NEXT: v_lshl_or_b32 v25, v2, 16, v3
-; GFX9-NEXT: v_add_f32_e32 v2, s6, v1
-; GFX9-NEXT: v_readfirstlane_b32 s6, v2
+; GFX9-NEXT: v_lshl_or_b32 v26, v6, 16, v7
+; GFX9-NEXT: v_add_f32_e32 v6, s6, v5
+; GFX9-NEXT: v_readfirstlane_b32 s6, v6
; GFX9-NEXT: s_bfe_u32 s7, s6, 0x10010
; GFX9-NEXT: s_add_i32 s7, s7, s6
; GFX9-NEXT: s_add_i32 s8, s7, 0x7fff
; GFX9-NEXT: s_or_b32 s9, s6, 0x400000
-; GFX9-NEXT: v_cmp_u_f32_e32 vcc, v2, v2
+; GFX9-NEXT: v_cmp_u_f32_e32 vcc, v6, v6
; GFX9-NEXT: s_and_b64 s[6:7], vcc, exec
; GFX9-NEXT: s_cselect_b32 s6, s9, s8
; GFX9-NEXT: s_lshr_b32 s11, s6, 16
; GFX9-NEXT: s_lshl_b32 s6, s17, 16
-; GFX9-NEXT: v_add_f32_e32 v2, s6, v1
-; GFX9-NEXT: v_readfirstlane_b32 s6, v2
+; GFX9-NEXT: v_add_f32_e32 v6, s6, v5
+; GFX9-NEXT: v_readfirstlane_b32 s6, v6
; GFX9-NEXT: s_bfe_u32 s7, s6, 0x10010
; GFX9-NEXT: s_add_i32 s7, s7, s6
; GFX9-NEXT: s_add_i32 s8, s7, 0x7fff
; GFX9-NEXT: s_or_b32 s9, s6, 0x400000
-; GFX9-NEXT: v_cmp_u_f32_e32 vcc, v2, v2
+; GFX9-NEXT: v_cmp_u_f32_e32 vcc, v6, v6
; GFX9-NEXT: s_and_b64 s[6:7], vcc, exec
; GFX9-NEXT: s_cselect_b32 s6, s9, s8
; GFX9-NEXT: s_lshr_b32 s17, s6, 16
; GFX9-NEXT: s_and_b32 s6, s16, 0xffff0000
-; GFX9-NEXT: v_add_f32_e32 v2, s6, v1
-; GFX9-NEXT: v_readfirstlane_b32 s6, v2
+; GFX9-NEXT: v_add_f32_e32 v6, s6, v5
+; GFX9-NEXT: v_readfirstlane_b32 s6, v6
; GFX9-NEXT: s_bfe_u32 s7, s6, 0x10010
; GFX9-NEXT: s_add_i32 s7, s7, s6
; GFX9-NEXT: s_add_i32 s8, s7, 0x7fff
; GFX9-NEXT: s_or_b32 s9, s6, 0x400000
-; GFX9-NEXT: v_cmp_u_f32_e32 vcc, v2, v2
+; GFX9-NEXT: v_cmp_u_f32_e32 vcc, v6, v6
; GFX9-NEXT: s_and_b64 s[6:7], vcc, exec
; GFX9-NEXT: s_cselect_b32 s6, s9, s8
; GFX9-NEXT: s_lshr_b32 s8, s6, 16
; GFX9-NEXT: s_lshl_b32 s6, s16, 16
-; GFX9-NEXT: v_add_f32_e32 v2, s6, v1
-; GFX9-NEXT: v_readfirstlane_b32 s6, v2
+; GFX9-NEXT: v_add_f32_e32 v6, s6, v5
+; GFX9-NEXT: v_readfirstlane_b32 s6, v6
; GFX9-NEXT: s_bfe_u32 s7, s6, 0x10010
; GFX9-NEXT: s_add_i32 s7, s7, s6
; GFX9-NEXT: s_add_i32 s9, s7, 0x7fff
; GFX9-NEXT: s_or_b32 s10, s6, 0x400000
-; GFX9-NEXT: v_cmp_u_f32_e32 vcc, v2, v2
+; GFX9-NEXT: v_cmp_u_f32_e32 vcc, v6, v6
; GFX9-NEXT: s_and_b64 s[6:7], vcc, exec
; GFX9-NEXT: s_cselect_b32 s6, s10, s9
; GFX9-NEXT: s_lshr_b32 s16, s6, 16
; GFX9-NEXT: s_and_b32 s6, s19, 0xffff0000
-; GFX9-NEXT: v_add_f32_e32 v2, s6, v1
-; GFX9-NEXT: v_readfirstlane_b32 s6, v2
+; GFX9-NEXT: v_add_f32_e32 v6, s6, v5
+; GFX9-NEXT: v_readfirstlane_b32 s6, v6
; GFX9-NEXT: s_bfe_u32 s7, s6, 0x10010
; GFX9-NEXT: s_add_i32 s7, s7, s6
-; GFX9-NEXT: s_pack_ll_b32_b16 s46, s16, s8
+; GFX9-NEXT: s_pack_ll_b32_b16 s44, s16, s8
; GFX9-NEXT: s_add_i32 s8, s7, 0x7fff
; GFX9-NEXT: s_or_b32 s9, s6, 0x400000
-; GFX9-NEXT: v_cmp_u_f32_e32 vcc, v2, v2
+; GFX9-NEXT: v_cmp_u_f32_e32 vcc, v6, v6
; GFX9-NEXT: s_and_b64 s[6:7], vcc, exec
; GFX9-NEXT: s_cselect_b32 s6, s9, s8
; GFX9-NEXT: s_lshr_b32 s12, s6, 16
; GFX9-NEXT: s_lshl_b32 s6, s19, 16
-; GFX9-NEXT: v_add_f32_e32 v2, s6, v1
-; GFX9-NEXT: v_readfirstlane_b32 s6, v2
+; GFX9-NEXT: v_add_f32_e32 v6, s6, v5
+; GFX9-NEXT: v_readfirstlane_b32 s6, v6
; GFX9-NEXT: s_bfe_u32 s7, s6, 0x10010
; GFX9-NEXT: s_add_i32 s7, s7, s6
; GFX9-NEXT: s_add_i32 s8, s7, 0x7fff
; GFX9-NEXT: s_or_b32 s9, s6, 0x400000
-; GFX9-NEXT: v_cmp_u_f32_e32 vcc, v2, v2
+; GFX9-NEXT: v_cmp_u_f32_e32 vcc, v6, v6
; GFX9-NEXT: s_and_b64 s[6:7], vcc, exec
; GFX9-NEXT: s_cselect_b32 s6, s9, s8
; GFX9-NEXT: s_lshr_b32 s19, s6, 16
; GFX9-NEXT: s_and_b32 s6, s18, 0xffff0000
-; GFX9-NEXT: v_add_f32_e32 v2, s6, v1
-; GFX9-NEXT: v_readfirstlane_b32 s6, v2
+; GFX9-NEXT: v_add_f32_e32 v6, s6, v5
+; GFX9-NEXT: v_readfirstlane_b32 s6, v6
; GFX9-NEXT: s_bfe_u32 s7, s6, 0x10010
; GFX9-NEXT: s_add_i32 s7, s7, s6
; GFX9-NEXT: s_add_i32 s8, s7, 0x7fff
; GFX9-NEXT: s_or_b32 s9, s6, 0x400000
-; GFX9-NEXT: v_cmp_u_f32_e32 vcc, v2, v2
+; GFX9-NEXT: v_cmp_u_f32_e32 vcc, v6, v6
; GFX9-NEXT: s_and_b64 s[6:7], vcc, exec
; GFX9-NEXT: s_cselect_b32 s6, s9, s8
; GFX9-NEXT: s_lshr_b32 s8, s6, 16
; GFX9-NEXT: s_lshl_b32 s6, s18, 16
-; GFX9-NEXT: v_add_f32_e32 v2, s6, v1
-; GFX9-NEXT: v_readfirstlane_b32 s6, v2
+; GFX9-NEXT: v_add_f32_e32 v6, s6, v5
+; GFX9-NEXT: v_readfirstlane_b32 s6, v6
; GFX9-NEXT: s_bfe_u32 s7, s6, 0x10010
; GFX9-NEXT: s_add_i32 s7, s7, s6
; GFX9-NEXT: s_add_i32 s9, s7, 0x7fff
; GFX9-NEXT: s_or_b32 s10, s6, 0x400000
-; GFX9-NEXT: v_cmp_u_f32_e32 vcc, v2, v2
+; GFX9-NEXT: v_cmp_u_f32_e32 vcc, v6, v6
; GFX9-NEXT: s_and_b64 s[6:7], vcc, exec
; GFX9-NEXT: s_cselect_b32 s6, s10, s9
; GFX9-NEXT: s_lshr_b32 s18, s6, 16
; GFX9-NEXT: s_and_b32 s6, s21, 0xffff0000
-; GFX9-NEXT: v_add_f32_e32 v2, s6, v1
-; GFX9-NEXT: v_readfirstlane_b32 s6, v2
+; GFX9-NEXT: v_add_f32_e32 v6, s6, v5
+; GFX9-NEXT: v_readfirstlane_b32 s6, v6
; GFX9-NEXT: s_bfe_u32 s7, s6, 0x10010
; GFX9-NEXT: s_add_i32 s7, s7, s6
; GFX9-NEXT: s_pack_ll_b32_b16 s56, s18, s8
; GFX9-NEXT: s_add_i32 s8, s7, 0x7fff
; GFX9-NEXT: s_or_b32 s9, s6, 0x400000
-; GFX9-NEXT: v_cmp_u_f32_e32 vcc, v2, v2
+; GFX9-NEXT: v_cmp_u_f32_e32 vcc, v6, v6
; GFX9-NEXT: s_and_b64 s[6:7], vcc, exec
; GFX9-NEXT: s_cselect_b32 s6, s9, s8
; GFX9-NEXT: s_lshr_b32 s13, s6, 16
; GFX9-NEXT: s_lshl_b32 s6, s21, 16
-; GFX9-NEXT: v_add_f32_e32 v2, s6, v1
-; GFX9-NEXT: v_readfirstlane_b32 s6, v2
+; GFX9-NEXT: v_add_f32_e32 v6, s6, v5
+; GFX9-NEXT: v_readfirstlane_b32 s6, v6
; GFX9-NEXT: s_bfe_u32 s7, s6, 0x10010
; GFX9-NEXT: s_add_i32 s7, s7, s6
; GFX9-NEXT: s_add_i32 s8, s7, 0x7fff
; GFX9-NEXT: s_or_b32 s9, s6, 0x400000
-; GFX9-NEXT: v_cmp_u_f32_e32 vcc, v2, v2
+; GFX9-NEXT: v_cmp_u_f32_e32 vcc, v6, v6
; GFX9-NEXT: s_and_b64 s[6:7], vcc, exec
; GFX9-NEXT: s_cselect_b32 s6, s9, s8
; GFX9-NEXT: s_lshr_b32 s21, s6, 16
; GFX9-NEXT: s_and_b32 s6, s20, 0xffff0000
-; GFX9-NEXT: v_add_f32_e32 v2, s6, v1
-; GFX9-NEXT: v_readfirstlane_b32 s6, v2
+; GFX9-NEXT: v_add_f32_e32 v6, s6, v5
+; GFX9-NEXT: v_readfirstlane_b32 s6, v6
; GFX9-NEXT: s_bfe_u32 s7, s6, 0x10010
; GFX9-NEXT: s_add_i32 s7, s7, s6
; GFX9-NEXT: s_add_i32 s8, s7, 0x7fff
; GFX9-NEXT: s_or_b32 s9, s6, 0x400000
-; GFX9-NEXT: v_cmp_u_f32_e32 vcc, v2, v2
+; GFX9-NEXT: v_cmp_u_f32_e32 vcc, v6, v6
; GFX9-NEXT: s_and_b64 s[6:7], vcc, exec
; GFX9-NEXT: s_cselect_b32 s6, s9, s8
; GFX9-NEXT: s_lshr_b32 s8, s6, 16
; GFX9-NEXT: s_lshl_b32 s6, s20, 16
-; GFX9-NEXT: v_add_f32_e32 v2, s6, v1
-; GFX9-NEXT: v_readfirstlane_b32 s6, v2
+; GFX9-NEXT: v_add_f32_e32 v6, s6, v5
+; GFX9-NEXT: v_readfirstlane_b32 s6, v6
; GFX9-NEXT: s_bfe_u32 s7, s6, 0x10010
; GFX9-NEXT: s_add_i32 s7, s7, s6
; GFX9-NEXT: s_add_i32 s9, s7, 0x7fff
; GFX9-NEXT: s_or_b32 s10, s6, 0x400000
-; GFX9-NEXT: v_cmp_u_f32_e32 vcc, v2, v2
+; GFX9-NEXT: v_cmp_u_f32_e32 vcc, v6, v6
; GFX9-NEXT: s_and_b64 s[6:7], vcc, exec
; GFX9-NEXT: s_cselect_b32 s6, s10, s9
; GFX9-NEXT: s_lshr_b32 s20, s6, 16
; GFX9-NEXT: s_and_b32 s6, s23, 0xffff0000
-; GFX9-NEXT: v_add_f32_e32 v2, s6, v1
-; GFX9-NEXT: v_readfirstlane_b32 s6, v2
+; GFX9-NEXT: v_add_f32_e32 v6, s6, v5
+; GFX9-NEXT: v_readfirstlane_b32 s6, v6
; GFX9-NEXT: s_bfe_u32 s7, s6, 0x10010
; GFX9-NEXT: s_add_i32 s7, s7, s6
; GFX9-NEXT: s_pack_ll_b32_b16 s58, s20, s8
; GFX9-NEXT: s_add_i32 s8, s7, 0x7fff
; GFX9-NEXT: s_or_b32 s9, s6, 0x400000
-; GFX9-NEXT: v_cmp_u_f32_e32 vcc, v2, v2
+; GFX9-NEXT: v_cmp_u_f32_e32 vcc, v6, v6
; GFX9-NEXT: s_and_b64 s[6:7], vcc, exec
; GFX9-NEXT: s_cselect_b32 s6, s9, s8
; GFX9-NEXT: s_lshr_b32 s14, s6, 16
; GFX9-NEXT: s_lshl_b32 s6, s23, 16
-; GFX9-NEXT: v_add_f32_e32 v2, s6, v1
-; GFX9-NEXT: v_readfirstlane_b32 s6, v2
+; GFX9-NEXT: v_add_f32_e32 v6, s6, v5
+; GFX9-NEXT: v_readfirstlane_b32 s6, v6
; GFX9-NEXT: s_bfe_u32 s7, s6, 0x10010
; GFX9-NEXT: s_add_i32 s7, s7, s6
; GFX9-NEXT: s_add_i32 s8, s7, 0x7fff
; GFX9-NEXT: s_or_b32 s9, s6, 0x400000
-; GFX9-NEXT: v_cmp_u_f32_e32 vcc, v2, v2
+; GFX9-NEXT: v_cmp_u_f32_e32 vcc, v6, v6
; GFX9-NEXT: s_and_b64 s[6:7], vcc, exec
; GFX9-NEXT: s_cselect_b32 s6, s9, s8
; GFX9-NEXT: s_lshr_b32 s23, s6, 16
; GFX9-NEXT: s_and_b32 s6, s22, 0xffff0000
-; GFX9-NEXT: v_add_f32_e32 v2, s6, v1
-; GFX9-NEXT: v_readfirstlane_b32 s6, v2
+; GFX9-NEXT: v_add_f32_e32 v6, s6, v5
+; GFX9-NEXT: v_readfirstlane_b32 s6, v6
; GFX9-NEXT: s_bfe_u32 s7, s6, 0x10010
; GFX9-NEXT: s_add_i32 s7, s7, s6
; GFX9-NEXT: s_add_i32 s8, s7, 0x7fff
; GFX9-NEXT: s_or_b32 s9, s6, 0x400000
-; GFX9-NEXT: v_cmp_u_f32_e32 vcc, v2, v2
+; GFX9-NEXT: v_cmp_u_f32_e32 vcc, v6, v6
; GFX9-NEXT: s_and_b64 s[6:7], vcc, exec
; GFX9-NEXT: s_cselect_b32 s6, s9, s8
; GFX9-NEXT: s_lshr_b32 s8, s6, 16
; GFX9-NEXT: s_lshl_b32 s6, s22, 16
-; GFX9-NEXT: v_add_f32_e32 v2, s6, v1
-; GFX9-NEXT: v_readfirstlane_b32 s6, v2
+; GFX9-NEXT: v_add_f32_e32 v6, s6, v5
+; GFX9-NEXT: v_readfirstlane_b32 s6, v6
; GFX9-NEXT: s_bfe_u32 s7, s6, 0x10010
; GFX9-NEXT: s_add_i32 s7, s7, s6
; GFX9-NEXT: s_add_i32 s9, s7, 0x7fff
; GFX9-NEXT: s_or_b32 s10, s6, 0x400000
-; GFX9-NEXT: v_cmp_u_f32_e32 vcc, v2, v2
+; GFX9-NEXT: v_cmp_u_f32_e32 vcc, v6, v6
; GFX9-NEXT: s_and_b64 s[6:7], vcc, exec
; GFX9-NEXT: s_cselect_b32 s6, s10, s9
; GFX9-NEXT: s_lshr_b32 s22, s6, 16
; GFX9-NEXT: s_and_b32 s6, s25, 0xffff0000
-; GFX9-NEXT: v_add_f32_e32 v2, s6, v1
-; GFX9-NEXT: v_readfirstlane_b32 s6, v2
+; GFX9-NEXT: v_add_f32_e32 v6, s6, v5
+; GFX9-NEXT: v_readfirstlane_b32 s6, v6
; GFX9-NEXT: s_bfe_u32 s7, s6, 0x10010
; GFX9-NEXT: s_add_i32 s7, s7, s6
; GFX9-NEXT: s_pack_ll_b32_b16 s60, s22, s8
; GFX9-NEXT: s_add_i32 s8, s7, 0x7fff
; GFX9-NEXT: s_or_b32 s9, s6, 0x400000
-; GFX9-NEXT: v_cmp_u_f32_e32 vcc, v2, v2
+; GFX9-NEXT: v_cmp_u_f32_e32 vcc, v6, v6
; GFX9-NEXT: s_and_b64 s[6:7], vcc, exec
; GFX9-NEXT: s_cselect_b32 s6, s9, s8
; GFX9-NEXT: s_lshr_b32 s15, s6, 16
; GFX9-NEXT: s_lshl_b32 s6, s25, 16
-; GFX9-NEXT: v_add_f32_e32 v2, s6, v1
-; GFX9-NEXT: v_readfirstlane_b32 s6, v2
+; GFX9-NEXT: v_add_f32_e32 v6, s6, v5
+; GFX9-NEXT: v_readfirstlane_b32 s6, v6
; GFX9-NEXT: s_bfe_u32 s7, s6, 0x10010
; GFX9-NEXT: s_add_i32 s7, s7, s6
; GFX9-NEXT: s_add_i32 s8, s7, 0x7fff
; GFX9-NEXT: s_or_b32 s9, s6, 0x400000
-; GFX9-NEXT: v_cmp_u_f32_e32 vcc, v2, v2
+; GFX9-NEXT: v_cmp_u_f32_e32 vcc, v6, v6
; GFX9-NEXT: s_and_b64 s[6:7], vcc, exec
; GFX9-NEXT: s_cselect_b32 s6, s9, s8
; GFX9-NEXT: s_lshr_b32 s25, s6, 16
; GFX9-NEXT: s_and_b32 s6, s24, 0xffff0000
-; GFX9-NEXT: v_add_f32_e32 v2, s6, v1
-; GFX9-NEXT: v_readfirstlane_b32 s6, v2
+; GFX9-NEXT: v_add_f32_e32 v6, s6, v5
+; GFX9-NEXT: v_readfirstlane_b32 s6, v6
; GFX9-NEXT: s_bfe_u32 s7, s6, 0x10010
; GFX9-NEXT: s_add_i32 s7, s7, s6
; GFX9-NEXT: s_add_i32 s8, s7, 0x7fff
; GFX9-NEXT: s_or_b32 s9, s6, 0x400000
-; GFX9-NEXT: v_cmp_u_f32_e32 vcc, v2, v2
+; GFX9-NEXT: v_cmp_u_f32_e32 vcc, v6, v6
; GFX9-NEXT: s_and_b64 s[6:7], vcc, exec
; GFX9-NEXT: s_cselect_b32 s6, s9, s8
; GFX9-NEXT: s_lshr_b32 s8, s6, 16
; GFX9-NEXT: s_lshl_b32 s6, s24, 16
-; GFX9-NEXT: v_add_f32_e32 v2, s6, v1
-; GFX9-NEXT: v_readfirstlane_b32 s6, v2
+; GFX9-NEXT: v_add_f32_e32 v6, s6, v5
+; GFX9-NEXT: v_readfirstlane_b32 s6, v6
; GFX9-NEXT: s_bfe_u32 s7, s6, 0x10010
; GFX9-NEXT: s_add_i32 s7, s7, s6
; GFX9-NEXT: s_add_i32 s9, s7, 0x7fff
; GFX9-NEXT: s_or_b32 s10, s6, 0x400000
-; GFX9-NEXT: v_cmp_u_f32_e32 vcc, v2, v2
+; GFX9-NEXT: v_cmp_u_f32_e32 vcc, v6, v6
; GFX9-NEXT: s_and_b64 s[6:7], vcc, exec
; GFX9-NEXT: s_cselect_b32 s6, s10, s9
; GFX9-NEXT: s_lshr_b32 s24, s6, 16
; GFX9-NEXT: s_and_b32 s6, s27, 0xffff0000
-; GFX9-NEXT: v_add_f32_e32 v2, s6, v1
-; GFX9-NEXT: v_readfirstlane_b32 s6, v2
+; GFX9-NEXT: v_add_f32_e32 v6, s6, v5
+; GFX9-NEXT: v_readfirstlane_b32 s6, v6
; GFX9-NEXT: s_bfe_u32 s7, s6, 0x10010
; GFX9-NEXT: s_add_i32 s7, s7, s6
; GFX9-NEXT: s_pack_ll_b32_b16 s62, s24, s8
; GFX9-NEXT: s_add_i32 s8, s7, 0x7fff
; GFX9-NEXT: s_or_b32 s9, s6, 0x400000
-; GFX9-NEXT: v_cmp_u_f32_e32 vcc, v2, v2
+; GFX9-NEXT: v_cmp_u_f32_e32 vcc, v6, v6
; GFX9-NEXT: s_and_b64 s[6:7], vcc, exec
; GFX9-NEXT: s_cselect_b32 s6, s9, s8
; GFX9-NEXT: s_lshr_b32 s76, s6, 16
; GFX9-NEXT: s_lshl_b32 s6, s27, 16
-; GFX9-NEXT: v_add_f32_e32 v2, s6, v1
-; GFX9-NEXT: v_readfirstlane_b32 s6, v2
+; GFX9-NEXT: v_add_f32_e32 v6, s6, v5
+; GFX9-NEXT: v_readfirstlane_b32 s6, v6
; GFX9-NEXT: s_bfe_u32 s7, s6, 0x10010
; GFX9-NEXT: s_add_i32 s7, s7, s6
; GFX9-NEXT: s_add_i32 s8, s7, 0x7fff
; GFX9-NEXT: s_or_b32 s9, s6, 0x400000
-; GFX9-NEXT: v_cmp_u_f32_e32 vcc, v2, v2
+; GFX9-NEXT: v_cmp_u_f32_e32 vcc, v6, v6
; GFX9-NEXT: s_and_b64 s[6:7], vcc, exec
; GFX9-NEXT: s_cselect_b32 s6, s9, s8
; GFX9-NEXT: s_lshr_b32 s27, s6, 16
; GFX9-NEXT: s_and_b32 s6, s26, 0xffff0000
-; GFX9-NEXT: v_add_f32_e32 v2, s6, v1
-; GFX9-NEXT: v_readfirstlane_b32 s6, v2
+; GFX9-NEXT: v_add_f32_e32 v6, s6, v5
+; GFX9-NEXT: v_readfirstlane_b32 s6, v6
; GFX9-NEXT: s_bfe_u32 s7, s6, 0x10010
; GFX9-NEXT: s_add_i32 s7, s7, s6
; GFX9-NEXT: s_add_i32 s8, s7, 0x7fff
; GFX9-NEXT: s_or_b32 s9, s6, 0x400000
-; GFX9-NEXT: v_cmp_u_f32_e32 vcc, v2, v2
+; GFX9-NEXT: v_cmp_u_f32_e32 vcc, v6, v6
; GFX9-NEXT: s_and_b64 s[6:7], vcc, exec
; GFX9-NEXT: s_cselect_b32 s6, s9, s8
; GFX9-NEXT: s_lshr_b32 s8, s6, 16
; GFX9-NEXT: s_lshl_b32 s6, s26, 16
-; GFX9-NEXT: v_add_f32_e32 v2, s6, v1
-; GFX9-NEXT: v_readfirstlane_b32 s6, v2
+; GFX9-NEXT: v_add_f32_e32 v6, s6, v5
+; GFX9-NEXT: v_readfirstlane_b32 s6, v6
; GFX9-NEXT: s_bfe_u32 s7, s6, 0x10010
; GFX9-NEXT: s_add_i32 s7, s7, s6
; GFX9-NEXT: s_add_i32 s9, s7, 0x7fff
; GFX9-NEXT: s_or_b32 s10, s6, 0x400000
-; GFX9-NEXT: v_cmp_u_f32_e32 vcc, v2, v2
+; GFX9-NEXT: v_cmp_u_f32_e32 vcc, v6, v6
; GFX9-NEXT: s_and_b64 s[6:7], vcc, exec
; GFX9-NEXT: s_cselect_b32 s6, s10, s9
; GFX9-NEXT: s_lshr_b32 s26, s6, 16
; GFX9-NEXT: s_and_b32 s6, s29, 0xffff0000
-; GFX9-NEXT: v_add_f32_e32 v2, s6, v1
-; GFX9-NEXT: v_readfirstlane_b32 s6, v2
+; GFX9-NEXT: v_add_f32_e32 v6, s6, v5
+; GFX9-NEXT: v_readfirstlane_b32 s6, v6
; GFX9-NEXT: s_bfe_u32 s7, s6, 0x10010
; GFX9-NEXT: s_add_i32 s7, s7, s6
; GFX9-NEXT: s_pack_ll_b32_b16 s72, s26, s8
; GFX9-NEXT: s_add_i32 s8, s7, 0x7fff
; GFX9-NEXT: s_or_b32 s9, s6, 0x400000
-; GFX9-NEXT: v_cmp_u_f32_e32 vcc, v2, v2
+; GFX9-NEXT: v_cmp_u_f32_e32 vcc, v6, v6
; GFX9-NEXT: s_and_b64 s[6:7], vcc, exec
; GFX9-NEXT: s_cselect_b32 s6, s9, s8
; GFX9-NEXT: s_lshr_b32 s77, s6, 16
; GFX9-NEXT: s_lshl_b32 s6, s29, 16
-; GFX9-NEXT: v_add_f32_e32 v2, s6, v1
-; GFX9-NEXT: v_readfirstlane_b32 s6, v2
+; GFX9-NEXT: v_add_f32_e32 v6, s6, v5
+; GFX9-NEXT: v_readfirstlane_b32 s6, v6
; GFX9-NEXT: s_bfe_u32 s7, s6, 0x10010
; GFX9-NEXT: s_add_i32 s7, s7, s6
; GFX9-NEXT: s_add_i32 s8, s7, 0x7fff
; GFX9-NEXT: s_or_b32 s9, s6, 0x400000
-; GFX9-NEXT: v_cmp_u_f32_e32 vcc, v2, v2
+; GFX9-NEXT: v_cmp_u_f32_e32 vcc, v6, v6
; GFX9-NEXT: s_and_b64 s[6:7], vcc, exec
; GFX9-NEXT: s_cselect_b32 s6, s9, s8
; GFX9-NEXT: s_lshr_b32 s29, s6, 16
; GFX9-NEXT: s_and_b32 s6, s28, 0xffff0000
-; GFX9-NEXT: v_add_f32_e32 v2, s6, v1
-; GFX9-NEXT: v_readfirstlane_b32 s6, v2
+; GFX9-NEXT: v_add_f32_e32 v6, s6, v5
+; GFX9-NEXT: v_readfirstlane_b32 s6, v6
; GFX9-NEXT: s_bfe_u32 s7, s6, 0x10010
; GFX9-NEXT: s_add_i32 s7, s7, s6
; GFX9-NEXT: s_add_i32 s8, s7, 0x7fff
; GFX9-NEXT: s_or_b32 s9, s6, 0x400000
-; GFX9-NEXT: v_cmp_u_f32_e32 vcc, v2, v2
+; GFX9-NEXT: v_cmp_u_f32_e32 vcc, v6, v6
; GFX9-NEXT: s_and_b64 s[6:7], vcc, exec
; GFX9-NEXT: s_cselect_b32 s6, s9, s8
; GFX9-NEXT: s_lshr_b32 s8, s6, 16
; GFX9-NEXT: s_lshl_b32 s6, s28, 16
-; GFX9-NEXT: v_add_f32_e32 v2, s6, v1
-; GFX9-NEXT: v_readfirstlane_b32 s6, v2
+; GFX9-NEXT: v_add_f32_e32 v6, s6, v5
+; GFX9-NEXT: v_readfirstlane_b32 s6, v6
; GFX9-NEXT: s_bfe_u32 s7, s6, 0x10010
; GFX9-NEXT: s_add_i32 s7, s7, s6
; GFX9-NEXT: s_add_i32 s9, s7, 0x7fff
; GFX9-NEXT: s_or_b32 s10, s6, 0x400000
-; GFX9-NEXT: v_cmp_u_f32_e32 vcc, v2, v2
+; GFX9-NEXT: v_cmp_u_f32_e32 vcc, v6, v6
; GFX9-NEXT: s_and_b64 s[6:7], vcc, exec
; GFX9-NEXT: s_cselect_b32 s6, s10, s9
; GFX9-NEXT: s_lshr_b32 s28, s6, 16
; GFX9-NEXT: s_and_b32 s6, s5, 0xffff0000
-; GFX9-NEXT: v_add_f32_e32 v2, s6, v1
-; GFX9-NEXT: v_readfirstlane_b32 s6, v2
+; GFX9-NEXT: v_add_f32_e32 v6, s6, v5
+; GFX9-NEXT: v_readfirstlane_b32 s6, v6
; GFX9-NEXT: s_bfe_u32 s7, s6, 0x10010
; GFX9-NEXT: s_add_i32 s7, s7, s6
; GFX9-NEXT: s_pack_ll_b32_b16 s74, s28, s8
; GFX9-NEXT: s_add_i32 s8, s7, 0x7fff
; GFX9-NEXT: s_or_b32 s9, s6, 0x400000
-; GFX9-NEXT: v_cmp_u_f32_e32 vcc, v2, v2
+; GFX9-NEXT: v_cmp_u_f32_e32 vcc, v6, v6
; GFX9-NEXT: s_and_b64 s[6:7], vcc, exec
; GFX9-NEXT: s_cselect_b32 s6, s9, s8
; GFX9-NEXT: s_lshl_b32 s5, s5, 16
-; GFX9-NEXT: v_add_f32_e32 v2, s5, v1
-; GFX9-NEXT: v_readfirstlane_b32 s5, v2
+; GFX9-NEXT: v_add_f32_e32 v6, s5, v5
+; GFX9-NEXT: v_readfirstlane_b32 s5, v6
; GFX9-NEXT: s_lshr_b32 s78, s6, 16
; GFX9-NEXT: s_bfe_u32 s6, s5, 0x10010
; GFX9-NEXT: s_add_i32 s6, s6, s5
; GFX9-NEXT: s_add_i32 s8, s6, 0x7fff
; GFX9-NEXT: s_bitset1_b32 s5, 22
-; GFX9-NEXT: v_cmp_u_f32_e32 vcc, v2, v2
+; GFX9-NEXT: v_cmp_u_f32_e32 vcc, v6, v6
; GFX9-NEXT: s_and_b64 s[6:7], vcc, exec
; GFX9-NEXT: s_cselect_b32 s5, s5, s8
; GFX9-NEXT: s_and_b32 s6, s4, 0xffff0000
-; GFX9-NEXT: v_add_f32_e32 v2, s6, v1
-; GFX9-NEXT: v_readfirstlane_b32 s6, v2
+; GFX9-NEXT: v_add_f32_e32 v6, s6, v5
+; GFX9-NEXT: v_readfirstlane_b32 s6, v6
; GFX9-NEXT: s_bfe_u32 s8, s6, 0x10010
; GFX9-NEXT: s_add_i32 s8, s8, s6
; GFX9-NEXT: s_lshr_b32 s5, s5, 16
; GFX9-NEXT: s_add_i32 s10, s8, 0x7fff
; GFX9-NEXT: s_bitset1_b32 s6, 22
-; GFX9-NEXT: v_cmp_u_f32_e32 vcc, v2, v2
+; GFX9-NEXT: v_cmp_u_f32_e32 vcc, v6, v6
; GFX9-NEXT: s_and_b64 s[8:9], vcc, exec
; GFX9-NEXT: s_cselect_b32 s6, s6, s10
; GFX9-NEXT: s_lshl_b32 s4, s4, 16
-; GFX9-NEXT: v_add_f32_e32 v1, s4, v1
-; GFX9-NEXT: v_readfirstlane_b32 s4, v1
+; GFX9-NEXT: v_add_f32_e32 v5, s4, v5
+; GFX9-NEXT: v_readfirstlane_b32 s4, v5
; GFX9-NEXT: s_bfe_u32 s8, s4, 0x10010
; GFX9-NEXT: s_add_i32 s8, s8, s4
; GFX9-NEXT: s_lshr_b32 s6, s6, 16
; GFX9-NEXT: s_add_i32 s10, s8, 0x7fff
; GFX9-NEXT: s_bitset1_b32 s4, 22
-; GFX9-NEXT: v_cmp_u_f32_e32 vcc, v1, v1
-; GFX9-NEXT: v_lshrrev_b64 v[1:2], 24, v[25:26]
+; GFX9-NEXT: v_cmp_u_f32_e32 vcc, v5, v5
; GFX9-NEXT: s_and_b64 s[8:9], vcc, exec
-; GFX9-NEXT: v_lshrrev_b64 v[2:3], 24, v[23:24]
; GFX9-NEXT: s_cselect_b32 s4, s4, s10
-; GFX9-NEXT: v_lshrrev_b64 v[3:4], 24, v[21:22]
-; GFX9-NEXT: v_lshrrev_b64 v[9:10], 24, v[15:16]
-; GFX9-NEXT: s_pack_ll_b32_b16 s47, s17, s11
+; GFX9-NEXT: v_lshrrev_b64 v[5:6], 24, v[26:27]
+; GFX9-NEXT: v_lshrrev_b64 v[11:12], 24, v[20:21]
+; GFX9-NEXT: s_pack_ll_b32_b16 s45, s17, s11
; GFX9-NEXT: s_pack_ll_b32_b16 s57, s19, s12
; GFX9-NEXT: s_pack_ll_b32_b16 s59, s21, s13
; GFX9-NEXT: s_lshr_b32 s4, s4, 16
-; GFX9-NEXT: v_lshrrev_b64 v[4:5], 24, v[19:20]
-; GFX9-NEXT: v_lshrrev_b64 v[10:11], 24, v[13:14]
+; GFX9-NEXT: v_lshrrev_b64 v[6:7], 24, v[24:25]
+; GFX9-NEXT: v_lshrrev_b64 v[12:13], 24, v[15:16]
+; GFX9-NEXT: v_lshrrev_b64 v[17:18], 24, v[3:4]
; GFX9-NEXT: s_pack_ll_b32_b16 s61, s23, s14
; GFX9-NEXT: s_pack_ll_b32_b16 s63, s25, s15
; GFX9-NEXT: s_pack_ll_b32_b16 s73, s27, s76
@@ -174747,9 +174857,10 @@ define inreg <128 x i8> @bitcast_v64bf16_to_v128i8_scalar(<64 x bfloat> inreg %a
; GFX9-NEXT: s_pack_ll_b32_b16 s6, s4, s6
; GFX9-NEXT: s_lshr_b64 s[40:41], s[58:59], 24
; GFX9-NEXT: s_lshr_b64 s[42:43], s[56:57], 24
-; GFX9-NEXT: s_lshr_b64 s[44:45], s[46:47], 24
-; GFX9-NEXT: v_lshrrev_b64 v[5:6], 24, v[17:18]
-; GFX9-NEXT: v_lshrrev_b64 v[11:12], 24, v[7:8]
+; GFX9-NEXT: s_lshr_b64 s[46:47], s[44:45], 24
+; GFX9-NEXT: v_lshrrev_b64 v[7:8], 24, v[22:23]
+; GFX9-NEXT: v_lshrrev_b64 v[13:14], 24, v[9:10]
+; GFX9-NEXT: v_lshrrev_b64 v[18:19], 24, v[1:2]
; GFX9-NEXT: s_lshr_b64 s[34:35], s[6:7], 24
; GFX9-NEXT: s_lshr_b64 s[36:37], s[74:75], 24
; GFX9-NEXT: s_lshr_b64 s[38:39], s[72:73], 24
@@ -174759,7 +174870,7 @@ define inreg <128 x i8> @bitcast_v64bf16_to_v128i8_scalar(<64 x bfloat> inreg %a
; GFX9-NEXT: s_lshr_b32 s10, s7, 8
; GFX9-NEXT: s_lshr_b32 s41, s6, 16
; GFX9-NEXT: s_lshr_b32 s43, s6, 8
-; GFX9-NEXT: s_lshr_b32 s45, s75, 24
+; GFX9-NEXT: s_lshr_b32 s47, s75, 24
; GFX9-NEXT: s_lshr_b32 s75, s75, 8
; GFX9-NEXT: s_lshr_b32 s79, s74, 16
; GFX9-NEXT: s_lshr_b32 s74, s74, 8
@@ -174783,42 +174894,42 @@ define inreg <128 x i8> @bitcast_v64bf16_to_v128i8_scalar(<64 x bfloat> inreg %a
; GFX9-NEXT: s_lshr_b32 s57, s57, 8
; GFX9-NEXT: s_lshr_b32 vcc_hi, s56, 16
; GFX9-NEXT: s_lshr_b32 s56, s56, 8
-; GFX9-NEXT: s_lshr_b32 s30, s47, 24
-; GFX9-NEXT: s_lshr_b32 s47, s47, 8
-; GFX9-NEXT: s_lshr_b32 s8, s46, 16
-; GFX9-NEXT: s_lshr_b32 s7, s46, 8
-; GFX9-NEXT: v_lshrrev_b32_e32 v6, 24, v26
-; GFX9-NEXT: v_lshrrev_b32_e32 v12, 8, v26
-; GFX9-NEXT: v_lshrrev_b32_e32 v26, 16, v25
+; GFX9-NEXT: s_lshr_b32 s30, s45, 24
+; GFX9-NEXT: s_lshr_b32 s45, s45, 8
+; GFX9-NEXT: s_lshr_b32 s8, s44, 16
+; GFX9-NEXT: s_lshr_b32 s7, s44, 8
+; GFX9-NEXT: v_lshrrev_b32_e32 v8, 24, v27
+; GFX9-NEXT: v_lshrrev_b32_e32 v14, 8, v27
+; GFX9-NEXT: v_lshrrev_b32_e32 v19, 16, v26
+; GFX9-NEXT: v_lshrrev_b32_e32 v26, 8, v26
+; GFX9-NEXT: v_lshrrev_b32_e32 v27, 24, v25
; GFX9-NEXT: v_lshrrev_b32_e32 v25, 8, v25
-; GFX9-NEXT: v_lshrrev_b32_e32 v43, 24, v24
+; GFX9-NEXT: v_lshrrev_b32_e32 v44, 16, v24
; GFX9-NEXT: v_lshrrev_b32_e32 v24, 8, v24
-; GFX9-NEXT: v_lshrrev_b32_e32 v44, 16, v23
+; GFX9-NEXT: v_lshrrev_b32_e32 v45, 24, v23
; GFX9-NEXT: v_lshrrev_b32_e32 v23, 8, v23
-; GFX9-NEXT: v_lshrrev_b32_e32 v45, 24, v22
+; GFX9-NEXT: v_lshrrev_b32_e32 v46, 16, v22
; GFX9-NEXT: v_lshrrev_b32_e32 v22, 8, v22
-; GFX9-NEXT: v_lshrrev_b32_e32 v46, 16, v21
+; GFX9-NEXT: v_lshrrev_b32_e32 v47, 24, v21
; GFX9-NEXT: v_lshrrev_b32_e32 v21, 8, v21
-; GFX9-NEXT: v_lshrrev_b32_e32 v47, 24, v20
+; GFX9-NEXT: v_lshrrev_b32_e32 v56, 16, v20
; GFX9-NEXT: v_lshrrev_b32_e32 v20, 8, v20
-; GFX9-NEXT: v_lshrrev_b32_e32 v56, 16, v19
-; GFX9-NEXT: v_lshrrev_b32_e32 v19, 8, v19
-; GFX9-NEXT: v_lshrrev_b32_e32 v57, 24, v18
-; GFX9-NEXT: v_lshrrev_b32_e32 v18, 8, v18
-; GFX9-NEXT: v_lshrrev_b32_e32 v58, 16, v17
-; GFX9-NEXT: v_lshrrev_b32_e32 v17, 8, v17
-; GFX9-NEXT: v_lshrrev_b32_e32 v59, 24, v16
+; GFX9-NEXT: v_lshrrev_b32_e32 v57, 24, v16
; GFX9-NEXT: v_lshrrev_b32_e32 v16, 8, v16
-; GFX9-NEXT: v_lshrrev_b32_e32 v60, 16, v15
+; GFX9-NEXT: v_lshrrev_b32_e32 v58, 16, v15
; GFX9-NEXT: v_lshrrev_b32_e32 v15, 8, v15
-; GFX9-NEXT: v_lshrrev_b32_e32 v61, 24, v14
-; GFX9-NEXT: v_lshrrev_b32_e32 v14, 8, v14
-; GFX9-NEXT: v_lshrrev_b32_e32 v27, 16, v13
-; GFX9-NEXT: v_lshrrev_b32_e32 v13, 8, v13
-; GFX9-NEXT: v_lshrrev_b32_e32 v28, 24, v8
-; GFX9-NEXT: v_lshrrev_b32_e32 v8, 8, v8
-; GFX9-NEXT: v_lshrrev_b32_e32 v29, 16, v7
-; GFX9-NEXT: v_lshrrev_b32_e32 v7, 8, v7
+; GFX9-NEXT: v_lshrrev_b32_e32 v59, 24, v10
+; GFX9-NEXT: v_lshrrev_b32_e32 v10, 8, v10
+; GFX9-NEXT: v_lshrrev_b32_e32 v60, 16, v9
+; GFX9-NEXT: v_lshrrev_b32_e32 v9, 8, v9
+; GFX9-NEXT: v_lshrrev_b32_e32 v61, 24, v4
+; GFX9-NEXT: v_lshrrev_b32_e32 v4, 8, v4
+; GFX9-NEXT: v_lshrrev_b32_e32 v28, 16, v3
+; GFX9-NEXT: v_lshrrev_b32_e32 v3, 8, v3
+; GFX9-NEXT: v_lshrrev_b32_e32 v29, 24, v2
+; GFX9-NEXT: v_lshrrev_b32_e32 v2, 8, v2
+; GFX9-NEXT: v_lshrrev_b32_e32 v30, 16, v1
+; GFX9-NEXT: v_lshrrev_b32_e32 v1, 8, v1
; GFX9-NEXT: s_branch .LBB91_5
; GFX9-NEXT: .LBB91_3:
; GFX9-NEXT: ; implicit-def: $sgpr6
@@ -174833,46 +174944,46 @@ define inreg <128 x i8> @bitcast_v64bf16_to_v128i8_scalar(<64 x bfloat> inreg %a
; GFX9-NEXT: ; implicit-def: $sgpr6
; GFX9-NEXT: ; kill: killed $sgpr6
; GFX9-NEXT: ; implicit-def: $sgpr78
-; GFX9-NEXT: ; implicit-def: $sgpr83
-; GFX9-NEXT: ; implicit-def: $sgpr38
; GFX9-NEXT: ; implicit-def: $sgpr81
-; GFX9-NEXT: ; implicit-def: $sgpr8
-; GFX9-NEXT: ; implicit-def: $sgpr80
-; GFX9-NEXT: ; implicit-def: $sgpr86
+; GFX9-NEXT: ; implicit-def: $sgpr82
; GFX9-NEXT: ; implicit-def: $sgpr39
+; GFX9-NEXT: ; implicit-def: $sgpr8
+; GFX9-NEXT: ; implicit-def: $sgpr71
; GFX9-NEXT: ; implicit-def: $sgpr84
-; GFX9-NEXT: ; implicit-def: $sgpr9
-; GFX9-NEXT: ; implicit-def: $sgpr82
-; GFX9-NEXT: ; implicit-def: $sgpr97
+; GFX9-NEXT: ; implicit-def: $sgpr85
; GFX9-NEXT: ; implicit-def: $sgpr48
+; GFX9-NEXT: ; implicit-def: $sgpr9
+; GFX9-NEXT: ; implicit-def: $sgpr80
; GFX9-NEXT: ; implicit-def: $sgpr87
-; GFX9-NEXT: ; implicit-def: $sgpr10
-; GFX9-NEXT: ; implicit-def: $sgpr85
-; GFX9-NEXT: ; implicit-def: $sgpr54
+; GFX9-NEXT: ; implicit-def: $sgpr96
; GFX9-NEXT: ; implicit-def: $sgpr49
+; GFX9-NEXT: ; implicit-def: $sgpr10
+; GFX9-NEXT: ; implicit-def: $sgpr83
; GFX9-NEXT: ; implicit-def: $sgpr98
-; GFX9-NEXT: ; implicit-def: $sgpr11
-; GFX9-NEXT: ; implicit-def: $sgpr96
-; GFX9-NEXT: ; implicit-def: $sgpr65
+; GFX9-NEXT: ; implicit-def: $sgpr99
; GFX9-NEXT: ; implicit-def: $sgpr50
+; GFX9-NEXT: ; implicit-def: $sgpr11
+; GFX9-NEXT: ; implicit-def: $sgpr86
; GFX9-NEXT: ; implicit-def: $sgpr55
-; GFX9-NEXT: ; implicit-def: $sgpr12
-; GFX9-NEXT: ; implicit-def: $sgpr99
-; GFX9-NEXT: ; implicit-def: $sgpr68
+; GFX9-NEXT: ; implicit-def: $sgpr64
; GFX9-NEXT: ; implicit-def: $sgpr51
+; GFX9-NEXT: ; implicit-def: $sgpr12
+; GFX9-NEXT: ; implicit-def: $sgpr97
; GFX9-NEXT: ; implicit-def: $sgpr66
+; GFX9-NEXT: ; implicit-def: $sgpr67
+; GFX9-NEXT: ; implicit-def: $sgpr52
; GFX9-NEXT: ; implicit-def: $sgpr13
-; GFX9-NEXT: ; implicit-def: $sgpr64
-; GFX9-NEXT: ; implicit-def: $sgpr71
+; GFX9-NEXT: ; implicit-def: $sgpr54
+; GFX9-NEXT: ; implicit-def: $sgpr38
; GFX9-NEXT: ; implicit-def: $sgpr69
+; GFX9-NEXT: ; implicit-def: $sgpr68
; GFX9-NEXT: ; implicit-def: $sgpr14
-; GFX9-NEXT: ; implicit-def: $sgpr67
-; GFX9-NEXT: ; implicit-def: $sgpr52
-; GFX9-NEXT: ; implicit-def: $sgpr53
+; GFX9-NEXT: ; implicit-def: $sgpr65
; GFX9-NEXT: ; implicit-def: $sgpr7
-; GFX9-NEXT: ; implicit-def: $sgpr15
; GFX9-NEXT: ; implicit-def: $sgpr70
-; GFX9-NEXT: ; implicit-def: $sgpr44
+; GFX9-NEXT: ; implicit-def: $sgpr15
+; GFX9-NEXT: ; implicit-def: $sgpr53
+; GFX9-NEXT: ; implicit-def: $sgpr46
; GFX9-NEXT: ; implicit-def: $sgpr42
; GFX9-NEXT: ; implicit-def: $sgpr40
; GFX9-NEXT: ; implicit-def: $sgpr36
@@ -174976,72 +175087,72 @@ define inreg <128 x i8> @bitcast_v64bf16_to_v128i8_scalar(<64 x bfloat> inreg %a
; GFX9-NEXT: v_mov_b32_e32 v1, s77
; GFX9-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:60 ; 4-byte Folded Spill
; GFX9-NEXT: v_mov_b32_e32 v1, s8
-; GFX9-NEXT: v_mov_b32_e32 v46, s51
-; GFX9-NEXT: v_mov_b32_e32 v56, s50
-; GFX9-NEXT: v_mov_b32_e32 v58, s49
-; GFX9-NEXT: v_mov_b32_e32 v60, s48
-; GFX9-NEXT: v_mov_b32_e32 v27, s39
-; GFX9-NEXT: v_mov_b32_e32 v29, s38
-; GFX9-NEXT: v_mov_b32_e32 v10, s34
-; GFX9-NEXT: v_mov_b32_e32 v11, s36
+; GFX9-NEXT: v_mov_b32_e32 v24, s38
+; GFX9-NEXT: v_mov_b32_e32 v21, s51
+; GFX9-NEXT: v_mov_b32_e32 v16, s50
+; GFX9-NEXT: v_mov_b32_e32 v10, s49
+; GFX9-NEXT: v_mov_b32_e32 v4, s48
+; GFX9-NEXT: v_mov_b32_e32 v2, s39
+; GFX9-NEXT: v_mov_b32_e32 v17, s34
+; GFX9-NEXT: v_mov_b32_e32 v18, s36
; GFX9-NEXT: v_readlane_b32 s34, v62, 8
; GFX9-NEXT: v_readlane_b32 s36, v62, 6
; GFX9-NEXT: v_readlane_b32 s38, v62, 4
; GFX9-NEXT: v_readlane_b32 s48, v62, 2
; GFX9-NEXT: v_readlane_b32 s50, v62, 0
-; GFX9-NEXT: v_mov_b32_e32 v42, s46
-; GFX9-NEXT: v_mov_b32_e32 v41, s47
-; GFX9-NEXT: v_mov_b32_e32 v55, s15
-; GFX9-NEXT: v_mov_b32_e32 v40, s56
-; GFX9-NEXT: v_mov_b32_e32 v54, s57
-; GFX9-NEXT: v_mov_b32_e32 v52, s14
-; GFX9-NEXT: v_mov_b32_e32 v53, s58
-; GFX9-NEXT: v_mov_b32_e32 v51, s59
-; GFX9-NEXT: v_mov_b32_e32 v49, s13
-; GFX9-NEXT: v_mov_b32_e32 v50, s60
-; GFX9-NEXT: v_mov_b32_e32 v48, s61
-; GFX9-NEXT: v_mov_b32_e32 v38, s12
-; GFX9-NEXT: v_mov_b32_e32 v39, s62
-; GFX9-NEXT: v_mov_b32_e32 v37, s63
-; GFX9-NEXT: v_mov_b32_e32 v35, s11
-; GFX9-NEXT: v_mov_b32_e32 v36, s72
-; GFX9-NEXT: v_mov_b32_e32 v34, s73
-; GFX9-NEXT: v_mov_b32_e32 v32, s10
-; GFX9-NEXT: v_mov_b32_e32 v33, s74
-; GFX9-NEXT: v_mov_b32_e32 v31, s75
+; GFX9-NEXT: v_mov_b32_e32 v43, s44
+; GFX9-NEXT: v_mov_b32_e32 v42, s45
+; GFX9-NEXT: v_mov_b32_e32 v40, s15
+; GFX9-NEXT: v_mov_b32_e32 v41, s56
+; GFX9-NEXT: v_mov_b32_e32 v55, s57
+; GFX9-NEXT: v_mov_b32_e32 v53, s14
+; GFX9-NEXT: v_mov_b32_e32 v54, s58
+; GFX9-NEXT: v_mov_b32_e32 v52, s59
+; GFX9-NEXT: v_mov_b32_e32 v50, s13
+; GFX9-NEXT: v_mov_b32_e32 v51, s60
+; GFX9-NEXT: v_mov_b32_e32 v49, s61
+; GFX9-NEXT: v_mov_b32_e32 v39, s12
+; GFX9-NEXT: v_mov_b32_e32 v48, s62
+; GFX9-NEXT: v_mov_b32_e32 v38, s63
+; GFX9-NEXT: v_mov_b32_e32 v36, s11
+; GFX9-NEXT: v_mov_b32_e32 v37, s72
+; GFX9-NEXT: v_mov_b32_e32 v35, s73
+; GFX9-NEXT: v_mov_b32_e32 v33, s10
+; GFX9-NEXT: v_mov_b32_e32 v34, s74
+; GFX9-NEXT: v_mov_b32_e32 v32, s75
; GFX9-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:56 ; 4-byte Folded Spill
-; GFX9-NEXT: v_mov_b32_e32 v26, s53
-; GFX9-NEXT: v_mov_b32_e32 v25, s52
-; GFX9-NEXT: v_mov_b32_e32 v6, s70
-; GFX9-NEXT: v_mov_b32_e32 v12, s7
-; GFX9-NEXT: v_mov_b32_e32 v44, s6
-; GFX9-NEXT: v_mov_b32_e32 v23, s71
-; GFX9-NEXT: v_mov_b32_e32 v43, s67
-; GFX9-NEXT: v_mov_b32_e32 v24, s69
-; GFX9-NEXT: v_mov_b32_e32 v21, s68
-; GFX9-NEXT: v_mov_b32_e32 v45, s64
+; GFX9-NEXT: v_mov_b32_e32 v19, s7
+; GFX9-NEXT: v_mov_b32_e32 v26, s6
+; GFX9-NEXT: v_mov_b32_e32 v8, s53
+; GFX9-NEXT: v_mov_b32_e32 v14, s70
+; GFX9-NEXT: v_mov_b32_e32 v44, s69
+; GFX9-NEXT: v_mov_b32_e32 v27, s65
+; GFX9-NEXT: v_mov_b32_e32 v25, s68
+; GFX9-NEXT: v_mov_b32_e32 v46, s67
; GFX9-NEXT: v_mov_b32_e32 v22, s66
-; GFX9-NEXT: v_mov_b32_e32 v19, s65
-; GFX9-NEXT: v_mov_b32_e32 v47, s99
+; GFX9-NEXT: v_mov_b32_e32 v45, s54
+; GFX9-NEXT: v_mov_b32_e32 v23, s52
+; GFX9-NEXT: v_mov_b32_e32 v56, s64
; GFX9-NEXT: v_mov_b32_e32 v20, s55
-; GFX9-NEXT: v_mov_b32_e32 v17, s54
-; GFX9-NEXT: v_mov_b32_e32 v57, s96
-; GFX9-NEXT: v_mov_b32_e32 v18, s98
-; GFX9-NEXT: v_mov_b32_e32 v15, s97
-; GFX9-NEXT: v_mov_b32_e32 v59, s85
-; GFX9-NEXT: v_mov_b32_e32 v16, s87
-; GFX9-NEXT: v_mov_b32_e32 v13, s86
-; GFX9-NEXT: v_mov_b32_e32 v61, s82
-; GFX9-NEXT: v_mov_b32_e32 v14, s84
-; GFX9-NEXT: v_mov_b32_e32 v7, s83
-; GFX9-NEXT: v_mov_b32_e32 v28, s80
-; GFX9-NEXT: v_mov_b32_e32 v8, s81
-; GFX9-NEXT: v_mov_b32_e32 v1, s78
-; GFX9-NEXT: v_mov_b32_e32 v2, s88
-; GFX9-NEXT: v_mov_b32_e32 v3, s90
-; GFX9-NEXT: v_mov_b32_e32 v4, s92
-; GFX9-NEXT: v_mov_b32_e32 v5, s94
-; GFX9-NEXT: v_mov_b32_e32 v9, s30
+; GFX9-NEXT: v_mov_b32_e32 v47, s97
+; GFX9-NEXT: v_mov_b32_e32 v58, s99
+; GFX9-NEXT: v_mov_b32_e32 v15, s98
+; GFX9-NEXT: v_mov_b32_e32 v57, s86
+; GFX9-NEXT: v_mov_b32_e32 v60, s96
+; GFX9-NEXT: v_mov_b32_e32 v9, s87
+; GFX9-NEXT: v_mov_b32_e32 v59, s83
+; GFX9-NEXT: v_mov_b32_e32 v28, s85
+; GFX9-NEXT: v_mov_b32_e32 v3, s84
+; GFX9-NEXT: v_mov_b32_e32 v61, s80
+; GFX9-NEXT: v_mov_b32_e32 v30, s82
+; GFX9-NEXT: v_mov_b32_e32 v1, s81
+; GFX9-NEXT: v_mov_b32_e32 v29, s71
+; GFX9-NEXT: v_mov_b32_e32 v5, s78
+; GFX9-NEXT: v_mov_b32_e32 v6, s88
+; GFX9-NEXT: v_mov_b32_e32 v7, s90
+; GFX9-NEXT: v_mov_b32_e32 v11, s92
+; GFX9-NEXT: v_mov_b32_e32 v12, s94
+; GFX9-NEXT: v_mov_b32_e32 v13, s30
; GFX9-NEXT: v_readlane_b32 s11, v62, 10
; GFX9-NEXT: v_readlane_b32 s12, v62, 11
; GFX9-NEXT: v_readlane_b32 s13, v62, 12
@@ -175054,7 +175165,7 @@ define inreg <128 x i8> @bitcast_v64bf16_to_v128i8_scalar(<64 x bfloat> inreg %a
; GFX9-NEXT: v_readlane_b32 s10, v62, 19
; GFX9-NEXT: v_readlane_b32 s41, v62, 20
; GFX9-NEXT: v_readlane_b32 s43, v62, 21
-; GFX9-NEXT: v_readlane_b32 s45, v62, 22
+; GFX9-NEXT: v_readlane_b32 s47, v62, 22
; GFX9-NEXT: v_readlane_b32 s75, v62, 23
; GFX9-NEXT: v_readlane_b32 s79, v62, 24
; GFX9-NEXT: v_readlane_b32 s74, v62, 25
@@ -175079,7 +175190,7 @@ define inreg <128 x i8> @bitcast_v64bf16_to_v128i8_scalar(<64 x bfloat> inreg %a
; GFX9-NEXT: v_readlane_b32 vcc_hi, v62, 44
; GFX9-NEXT: v_readlane_b32 s56, v62, 45
; GFX9-NEXT: v_readlane_b32 s30, v62, 46
-; GFX9-NEXT: v_readlane_b32 s47, v62, 47
+; GFX9-NEXT: v_readlane_b32 s45, v62, 47
; GFX9-NEXT: v_readlane_b32 s8, v62, 48
; GFX9-NEXT: v_readlane_b32 s7, v62, 49
; GFX9-NEXT: v_readlane_b32 s35, v62, 9
@@ -175092,14 +175203,14 @@ define inreg <128 x i8> @bitcast_v64bf16_to_v128i8_scalar(<64 x bfloat> inreg %a
; GFX9-NEXT: s_lshl_b32 s7, s7, 8
; GFX9-NEXT: s_or_b32 s6, s6, s7
; GFX9-NEXT: s_and_b32 s7, s8, 0xff
-; GFX9-NEXT: s_lshl_b32 s8, s44, 8
+; GFX9-NEXT: s_lshl_b32 s8, s46, 8
; GFX9-NEXT: s_or_b32 s7, s7, s8
; GFX9-NEXT: s_and_b32 s6, s6, 0xffff
; GFX9-NEXT: s_lshl_b32 s7, s7, 16
; GFX9-NEXT: s_or_b32 s6, s6, s7
-; GFX9-NEXT: v_mov_b32_e32 v30, s6
+; GFX9-NEXT: v_mov_b32_e32 v31, s6
; GFX9-NEXT: s_and_b32 s6, s17, 0xff
-; GFX9-NEXT: s_lshl_b32 s7, s47, 8
+; GFX9-NEXT: s_lshl_b32 s7, s45, 8
; GFX9-NEXT: s_or_b32 s6, s6, s7
; GFX9-NEXT: s_and_b32 s7, s11, 0xff
; GFX9-NEXT: s_lshl_b32 s8, s30, 8
@@ -175107,8 +175218,8 @@ define inreg <128 x i8> @bitcast_v64bf16_to_v128i8_scalar(<64 x bfloat> inreg %a
; GFX9-NEXT: s_and_b32 s6, s6, 0xffff
; GFX9-NEXT: s_lshl_b32 s7, s7, 16
; GFX9-NEXT: s_or_b32 s6, s6, s7
-; GFX9-NEXT: buffer_store_dword v30, v0, s[0:3], 0 offen
-; GFX9-NEXT: v_mov_b32_e32 v30, s6
+; GFX9-NEXT: buffer_store_dword v31, v0, s[0:3], 0 offen
+; GFX9-NEXT: v_mov_b32_e32 v31, s6
; GFX9-NEXT: s_and_b32 s6, s18, 0xff
; GFX9-NEXT: s_lshl_b32 s7, s56, 8
; GFX9-NEXT: s_or_b32 s6, s6, s7
@@ -175118,8 +175229,8 @@ define inreg <128 x i8> @bitcast_v64bf16_to_v128i8_scalar(<64 x bfloat> inreg %a
; GFX9-NEXT: s_and_b32 s6, s6, 0xffff
; GFX9-NEXT: s_lshl_b32 s7, s7, 16
; GFX9-NEXT: s_or_b32 s6, s6, s7
-; GFX9-NEXT: buffer_store_dword v30, v0, s[0:3], 0 offen offset:4
-; GFX9-NEXT: v_mov_b32_e32 v30, s6
+; GFX9-NEXT: buffer_store_dword v31, v0, s[0:3], 0 offen offset:4
+; GFX9-NEXT: v_mov_b32_e32 v31, s6
; GFX9-NEXT: s_and_b32 s6, s19, 0xff
; GFX9-NEXT: s_lshl_b32 s7, s57, 8
; GFX9-NEXT: s_or_b32 s6, s6, s7
@@ -175129,8 +175240,8 @@ define inreg <128 x i8> @bitcast_v64bf16_to_v128i8_scalar(<64 x bfloat> inreg %a
; GFX9-NEXT: s_and_b32 s6, s6, 0xffff
; GFX9-NEXT: s_lshl_b32 s7, s7, 16
; GFX9-NEXT: s_or_b32 s6, s6, s7
-; GFX9-NEXT: buffer_store_dword v30, v0, s[0:3], 0 offen offset:8
-; GFX9-NEXT: v_mov_b32_e32 v30, s6
+; GFX9-NEXT: buffer_store_dword v31, v0, s[0:3], 0 offen offset:8
+; GFX9-NEXT: v_mov_b32_e32 v31, s6
; GFX9-NEXT: s_and_b32 s6, s20, 0xff
; GFX9-NEXT: s_lshl_b32 s7, s58, 8
; GFX9-NEXT: s_or_b32 s6, s6, s7
@@ -175140,8 +175251,8 @@ define inreg <128 x i8> @bitcast_v64bf16_to_v128i8_scalar(<64 x bfloat> inreg %a
; GFX9-NEXT: s_and_b32 s6, s6, 0xffff
; GFX9-NEXT: s_lshl_b32 s7, s7, 16
; GFX9-NEXT: s_or_b32 s6, s6, s7
-; GFX9-NEXT: buffer_store_dword v30, v0, s[0:3], 0 offen offset:12
-; GFX9-NEXT: v_mov_b32_e32 v30, s6
+; GFX9-NEXT: buffer_store_dword v31, v0, s[0:3], 0 offen offset:12
+; GFX9-NEXT: v_mov_b32_e32 v31, s6
; GFX9-NEXT: s_and_b32 s6, s21, 0xff
; GFX9-NEXT: s_lshl_b32 s7, s59, 8
; GFX9-NEXT: s_or_b32 s6, s6, s7
@@ -175151,8 +175262,8 @@ define inreg <128 x i8> @bitcast_v64bf16_to_v128i8_scalar(<64 x bfloat> inreg %a
; GFX9-NEXT: s_and_b32 s6, s6, 0xffff
; GFX9-NEXT: s_lshl_b32 s7, s7, 16
; GFX9-NEXT: s_or_b32 s6, s6, s7
-; GFX9-NEXT: buffer_store_dword v30, v0, s[0:3], 0 offen offset:16
-; GFX9-NEXT: v_mov_b32_e32 v30, s6
+; GFX9-NEXT: buffer_store_dword v31, v0, s[0:3], 0 offen offset:16
+; GFX9-NEXT: v_mov_b32_e32 v31, s6
; GFX9-NEXT: s_and_b32 s6, s22, 0xff
; GFX9-NEXT: s_lshl_b32 s7, s60, 8
; GFX9-NEXT: s_or_b32 s6, s6, s7
@@ -175162,8 +175273,8 @@ define inreg <128 x i8> @bitcast_v64bf16_to_v128i8_scalar(<64 x bfloat> inreg %a
; GFX9-NEXT: s_and_b32 s6, s6, 0xffff
; GFX9-NEXT: s_lshl_b32 s7, s7, 16
; GFX9-NEXT: s_or_b32 s6, s6, s7
-; GFX9-NEXT: buffer_store_dword v30, v0, s[0:3], 0 offen offset:20
-; GFX9-NEXT: v_mov_b32_e32 v30, s6
+; GFX9-NEXT: buffer_store_dword v31, v0, s[0:3], 0 offen offset:20
+; GFX9-NEXT: v_mov_b32_e32 v31, s6
; GFX9-NEXT: s_and_b32 s6, s23, 0xff
; GFX9-NEXT: s_lshl_b32 s7, s61, 8
; GFX9-NEXT: s_or_b32 s6, s6, s7
@@ -175173,8 +175284,8 @@ define inreg <128 x i8> @bitcast_v64bf16_to_v128i8_scalar(<64 x bfloat> inreg %a
; GFX9-NEXT: s_and_b32 s6, s6, 0xffff
; GFX9-NEXT: s_lshl_b32 s7, s7, 16
; GFX9-NEXT: s_or_b32 s6, s6, s7
-; GFX9-NEXT: buffer_store_dword v30, v0, s[0:3], 0 offen offset:24
-; GFX9-NEXT: v_mov_b32_e32 v30, s6
+; GFX9-NEXT: buffer_store_dword v31, v0, s[0:3], 0 offen offset:24
+; GFX9-NEXT: v_mov_b32_e32 v31, s6
; GFX9-NEXT: s_and_b32 s6, s24, 0xff
; GFX9-NEXT: s_lshl_b32 s7, s62, 8
; GFX9-NEXT: s_or_b32 s6, s6, s7
@@ -175184,8 +175295,8 @@ define inreg <128 x i8> @bitcast_v64bf16_to_v128i8_scalar(<64 x bfloat> inreg %a
; GFX9-NEXT: s_and_b32 s6, s6, 0xffff
; GFX9-NEXT: s_lshl_b32 s7, s7, 16
; GFX9-NEXT: s_or_b32 s6, s6, s7
-; GFX9-NEXT: buffer_store_dword v30, v0, s[0:3], 0 offen offset:28
-; GFX9-NEXT: v_mov_b32_e32 v30, s6
+; GFX9-NEXT: buffer_store_dword v31, v0, s[0:3], 0 offen offset:28
+; GFX9-NEXT: v_mov_b32_e32 v31, s6
; GFX9-NEXT: s_and_b32 s6, s25, 0xff
; GFX9-NEXT: s_lshl_b32 s7, s63, 8
; GFX9-NEXT: s_or_b32 s6, s6, s7
@@ -175195,8 +175306,8 @@ define inreg <128 x i8> @bitcast_v64bf16_to_v128i8_scalar(<64 x bfloat> inreg %a
; GFX9-NEXT: s_and_b32 s6, s6, 0xffff
; GFX9-NEXT: s_lshl_b32 s7, s7, 16
; GFX9-NEXT: s_or_b32 s6, s6, s7
-; GFX9-NEXT: buffer_store_dword v30, v0, s[0:3], 0 offen offset:32
-; GFX9-NEXT: v_mov_b32_e32 v30, s6
+; GFX9-NEXT: buffer_store_dword v31, v0, s[0:3], 0 offen offset:32
+; GFX9-NEXT: v_mov_b32_e32 v31, s6
; GFX9-NEXT: s_and_b32 s6, s26, 0xff
; GFX9-NEXT: s_lshl_b32 s7, s72, 8
; GFX9-NEXT: s_or_b32 s6, s6, s7
@@ -175206,8 +175317,8 @@ define inreg <128 x i8> @bitcast_v64bf16_to_v128i8_scalar(<64 x bfloat> inreg %a
; GFX9-NEXT: s_and_b32 s6, s6, 0xffff
; GFX9-NEXT: s_lshl_b32 s7, s7, 16
; GFX9-NEXT: s_or_b32 s6, s6, s7
-; GFX9-NEXT: buffer_store_dword v30, v0, s[0:3], 0 offen offset:36
-; GFX9-NEXT: v_mov_b32_e32 v30, s6
+; GFX9-NEXT: buffer_store_dword v31, v0, s[0:3], 0 offen offset:36
+; GFX9-NEXT: v_mov_b32_e32 v31, s6
; GFX9-NEXT: s_and_b32 s6, s27, 0xff
; GFX9-NEXT: s_lshl_b32 s7, s73, 8
; GFX9-NEXT: s_or_b32 s6, s6, s7
@@ -175217,8 +175328,8 @@ define inreg <128 x i8> @bitcast_v64bf16_to_v128i8_scalar(<64 x bfloat> inreg %a
; GFX9-NEXT: s_and_b32 s6, s6, 0xffff
; GFX9-NEXT: s_lshl_b32 s7, s7, 16
; GFX9-NEXT: s_or_b32 s6, s6, s7
-; GFX9-NEXT: buffer_store_dword v30, v0, s[0:3], 0 offen offset:40
-; GFX9-NEXT: v_mov_b32_e32 v30, s6
+; GFX9-NEXT: buffer_store_dword v31, v0, s[0:3], 0 offen offset:40
+; GFX9-NEXT: v_mov_b32_e32 v31, s6
; GFX9-NEXT: s_and_b32 s6, s28, 0xff
; GFX9-NEXT: s_lshl_b32 s7, s74, 8
; GFX9-NEXT: s_or_b32 s6, s6, s7
@@ -175228,19 +175339,19 @@ define inreg <128 x i8> @bitcast_v64bf16_to_v128i8_scalar(<64 x bfloat> inreg %a
; GFX9-NEXT: s_and_b32 s6, s6, 0xffff
; GFX9-NEXT: s_lshl_b32 s7, s7, 16
; GFX9-NEXT: s_or_b32 s6, s6, s7
-; GFX9-NEXT: buffer_store_dword v30, v0, s[0:3], 0 offen offset:44
-; GFX9-NEXT: v_mov_b32_e32 v30, s6
+; GFX9-NEXT: buffer_store_dword v31, v0, s[0:3], 0 offen offset:44
+; GFX9-NEXT: v_mov_b32_e32 v31, s6
; GFX9-NEXT: s_and_b32 s6, s29, 0xff
; GFX9-NEXT: s_lshl_b32 s7, s75, 8
; GFX9-NEXT: s_or_b32 s6, s6, s7
; GFX9-NEXT: s_and_b32 s7, s77, 0xff
-; GFX9-NEXT: s_lshl_b32 s8, s45, 8
+; GFX9-NEXT: s_lshl_b32 s8, s47, 8
; GFX9-NEXT: s_or_b32 s7, s7, s8
; GFX9-NEXT: s_and_b32 s6, s6, 0xffff
; GFX9-NEXT: s_lshl_b32 s7, s7, 16
; GFX9-NEXT: s_or_b32 s6, s6, s7
-; GFX9-NEXT: buffer_store_dword v30, v0, s[0:3], 0 offen offset:48
-; GFX9-NEXT: v_mov_b32_e32 v30, s6
+; GFX9-NEXT: buffer_store_dword v31, v0, s[0:3], 0 offen offset:48
+; GFX9-NEXT: v_mov_b32_e32 v31, s6
; GFX9-NEXT: s_and_b32 s4, s4, 0xff
; GFX9-NEXT: s_lshl_b32 s6, s43, 8
; GFX9-NEXT: s_or_b32 s4, s4, s6
@@ -175250,8 +175361,8 @@ define inreg <128 x i8> @bitcast_v64bf16_to_v128i8_scalar(<64 x bfloat> inreg %a
; GFX9-NEXT: s_and_b32 s4, s4, 0xffff
; GFX9-NEXT: s_lshl_b32 s6, s6, 16
; GFX9-NEXT: s_or_b32 s4, s4, s6
-; GFX9-NEXT: buffer_store_dword v30, v0, s[0:3], 0 offen offset:52
-; GFX9-NEXT: v_mov_b32_e32 v30, s4
+; GFX9-NEXT: buffer_store_dword v31, v0, s[0:3], 0 offen offset:52
+; GFX9-NEXT: v_mov_b32_e32 v31, s4
; GFX9-NEXT: s_and_b32 s4, s5, 0xff
; GFX9-NEXT: s_lshl_b32 s5, s10, 8
; GFX9-NEXT: s_or_b32 s4, s4, s5
@@ -175261,23 +175372,13 @@ define inreg <128 x i8> @bitcast_v64bf16_to_v128i8_scalar(<64 x bfloat> inreg %a
; GFX9-NEXT: s_and_b32 s4, s4, 0xffff
; GFX9-NEXT: s_lshl_b32 s5, s5, 16
; GFX9-NEXT: s_or_b32 s4, s4, s5
-; GFX9-NEXT: buffer_store_dword v30, v0, s[0:3], 0 offen offset:56
-; GFX9-NEXT: v_mov_b32_e32 v30, s4
-; GFX9-NEXT: buffer_store_dword v30, v0, s[0:3], 0 offen offset:60
-; GFX9-NEXT: buffer_load_dword v30, off, s[0:3], s32 offset:68 ; 4-byte Folded Reload
-; GFX9-NEXT: v_lshlrev_b32_e32 v7, 8, v7
-; GFX9-NEXT: v_lshlrev_b32_e32 v11, 8, v11
-; GFX9-NEXT: v_or_b32_sdwa v11, v29, v11 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
-; GFX9-NEXT: v_lshlrev_b32_e32 v5, 8, v5
-; GFX9-NEXT: v_or_b32_sdwa v5, v58, v5 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
-; GFX9-NEXT: v_lshlrev_b32_e32 v4, 8, v4
-; GFX9-NEXT: v_or_b32_sdwa v4, v56, v4 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
-; GFX9-NEXT: v_lshlrev_b32_e32 v3, 8, v3
-; GFX9-NEXT: v_or_b32_sdwa v3, v46, v3 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
-; GFX9-NEXT: v_lshlrev_b32_e32 v2, 8, v2
-; GFX9-NEXT: v_or_b32_sdwa v2, v44, v2 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; GFX9-NEXT: buffer_store_dword v31, v0, s[0:3], 0 offen offset:56
+; GFX9-NEXT: v_mov_b32_e32 v31, s4
+; GFX9-NEXT: buffer_store_dword v31, v0, s[0:3], 0 offen offset:60
+; GFX9-NEXT: buffer_load_dword v31, off, s[0:3], s32 offset:68 ; 4-byte Folded Reload
; GFX9-NEXT: v_lshlrev_b32_e32 v1, 8, v1
-; GFX9-NEXT: v_or_b32_sdwa v1, v26, v1 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; GFX9-NEXT: v_lshlrev_b32_e32 v18, 8, v18
+; GFX9-NEXT: v_or_b32_sdwa v18, v30, v18 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
; GFX9-NEXT: v_readlane_b32 s99, v63, 35
; GFX9-NEXT: v_readlane_b32 s98, v63, 34
; GFX9-NEXT: v_readlane_b32 s97, v63, 33
@@ -175315,93 +175416,103 @@ define inreg <128 x i8> @bitcast_v64bf16_to_v128i8_scalar(<64 x bfloat> inreg %a
; GFX9-NEXT: v_readlane_b32 s31, v63, 1
; GFX9-NEXT: v_readlane_b32 s30, v63, 0
; GFX9-NEXT: s_waitcnt vmcnt(0)
-; GFX9-NEXT: v_or_b32_sdwa v7, v30, v7 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
-; GFX9-NEXT: v_or_b32_sdwa v7, v7, v11 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_0 src1_sel:DWORD
-; GFX9-NEXT: buffer_store_dword v7, v0, s[0:3], 0 offen offset:64
-; GFX9-NEXT: v_lshlrev_b32_e32 v7, 8, v8
-; GFX9-NEXT: buffer_load_dword v8, off, s[0:3], s32 offset:60 ; 4-byte Folded Reload
-; GFX9-NEXT: buffer_load_dword v11, off, s[0:3], s32 offset:56 ; 4-byte Folded Reload
+; GFX9-NEXT: v_or_b32_sdwa v1, v31, v1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; GFX9-NEXT: v_or_b32_sdwa v1, v1, v18 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_0 src1_sel:DWORD
+; GFX9-NEXT: buffer_store_dword v1, v0, s[0:3], 0 offen offset:64
+; GFX9-NEXT: v_lshlrev_b32_e32 v1, 8, v2
+; GFX9-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:60 ; 4-byte Folded Reload
+; GFX9-NEXT: buffer_load_dword v18, off, s[0:3], s32 offset:56 ; 4-byte Folded Reload
; GFX9-NEXT: s_waitcnt vmcnt(1)
-; GFX9-NEXT: v_or_b32_sdwa v7, v8, v7 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
-; GFX9-NEXT: v_lshlrev_b32_e32 v8, 8, v28
+; GFX9-NEXT: v_or_b32_sdwa v1, v2, v1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; GFX9-NEXT: v_lshlrev_b32_e32 v2, 8, v29
; GFX9-NEXT: s_waitcnt vmcnt(0)
-; GFX9-NEXT: v_or_b32_sdwa v8, v11, v8 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
-; GFX9-NEXT: v_or_b32_sdwa v7, v7, v8 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_0 src1_sel:DWORD
-; GFX9-NEXT: buffer_store_dword v7, v0, s[0:3], 0 offen offset:68
-; GFX9-NEXT: v_lshlrev_b32_e32 v7, 8, v13
-; GFX9-NEXT: v_lshlrev_b32_e32 v8, 8, v10
-; GFX9-NEXT: v_or_b32_sdwa v7, v33, v7 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
-; GFX9-NEXT: v_or_b32_sdwa v8, v27, v8 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
-; GFX9-NEXT: v_or_b32_sdwa v7, v7, v8 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_0 src1_sel:DWORD
-; GFX9-NEXT: buffer_store_dword v7, v0, s[0:3], 0 offen offset:72
-; GFX9-NEXT: buffer_load_dword v10, off, s[0:3], s32 offset:64 ; 4-byte Folded Reload
-; GFX9-NEXT: v_lshlrev_b32_e32 v7, 8, v14
-; GFX9-NEXT: v_lshlrev_b32_e32 v8, 8, v61
-; GFX9-NEXT: v_or_b32_sdwa v7, v31, v7 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; GFX9-NEXT: v_or_b32_sdwa v2, v18, v2 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; GFX9-NEXT: v_or_b32_sdwa v1, v1, v2 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_0 src1_sel:DWORD
+; GFX9-NEXT: buffer_store_dword v1, v0, s[0:3], 0 offen offset:68
+; GFX9-NEXT: v_lshlrev_b32_e32 v1, 8, v3
+; GFX9-NEXT: v_lshlrev_b32_e32 v2, 8, v17
+; GFX9-NEXT: v_or_b32_sdwa v1, v34, v1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; GFX9-NEXT: v_or_b32_sdwa v2, v28, v2 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; GFX9-NEXT: v_or_b32_sdwa v1, v1, v2 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_0 src1_sel:DWORD
+; GFX9-NEXT: buffer_store_dword v1, v0, s[0:3], 0 offen offset:72
+; GFX9-NEXT: buffer_load_dword v3, off, s[0:3], s32 offset:64 ; 4-byte Folded Reload
+; GFX9-NEXT: v_lshlrev_b32_e32 v1, 8, v4
+; GFX9-NEXT: v_lshlrev_b32_e32 v2, 8, v61
+; GFX9-NEXT: v_or_b32_sdwa v1, v32, v1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
; GFX9-NEXT: s_waitcnt vmcnt(0)
-; GFX9-NEXT: v_or_b32_sdwa v8, v10, v8 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
-; GFX9-NEXT: v_or_b32_sdwa v7, v7, v8 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_0 src1_sel:DWORD
-; GFX9-NEXT: buffer_store_dword v7, v0, s[0:3], 0 offen offset:76
-; GFX9-NEXT: v_lshlrev_b32_e32 v7, 8, v15
-; GFX9-NEXT: v_lshlrev_b32_e32 v8, 8, v9
-; GFX9-NEXT: v_or_b32_sdwa v7, v36, v7 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
-; GFX9-NEXT: v_or_b32_sdwa v8, v60, v8 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
-; GFX9-NEXT: v_or_b32_sdwa v7, v7, v8 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_0 src1_sel:DWORD
-; GFX9-NEXT: buffer_store_dword v7, v0, s[0:3], 0 offen offset:80
-; GFX9-NEXT: v_lshlrev_b32_e32 v7, 8, v16
-; GFX9-NEXT: v_lshlrev_b32_e32 v8, 8, v59
-; GFX9-NEXT: v_or_b32_sdwa v7, v34, v7 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
-; GFX9-NEXT: v_or_b32_sdwa v8, v32, v8 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
-; GFX9-NEXT: v_or_b32_sdwa v7, v7, v8 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_0 src1_sel:DWORD
-; GFX9-NEXT: buffer_store_dword v7, v0, s[0:3], 0 offen offset:84
-; GFX9-NEXT: v_lshlrev_b32_e32 v7, 8, v17
-; GFX9-NEXT: v_or_b32_sdwa v7, v39, v7 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
-; GFX9-NEXT: v_or_b32_sdwa v5, v7, v5 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_0 src1_sel:DWORD
-; GFX9-NEXT: buffer_store_dword v5, v0, s[0:3], 0 offen offset:88
-; GFX9-NEXT: v_lshlrev_b32_e32 v5, 8, v18
-; GFX9-NEXT: v_lshlrev_b32_e32 v7, 8, v57
-; GFX9-NEXT: v_or_b32_sdwa v5, v37, v5 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
-; GFX9-NEXT: v_or_b32_sdwa v7, v35, v7 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
-; GFX9-NEXT: v_or_b32_sdwa v5, v5, v7 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_0 src1_sel:DWORD
-; GFX9-NEXT: buffer_store_dword v5, v0, s[0:3], 0 offen offset:92
-; GFX9-NEXT: v_lshlrev_b32_e32 v5, 8, v19
-; GFX9-NEXT: v_or_b32_sdwa v5, v50, v5 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
-; GFX9-NEXT: v_or_b32_sdwa v4, v5, v4 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_0 src1_sel:DWORD
-; GFX9-NEXT: buffer_store_dword v4, v0, s[0:3], 0 offen offset:96
-; GFX9-NEXT: v_lshlrev_b32_e32 v4, 8, v20
-; GFX9-NEXT: v_lshlrev_b32_e32 v5, 8, v47
-; GFX9-NEXT: v_or_b32_sdwa v4, v48, v4 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
-; GFX9-NEXT: v_or_b32_sdwa v5, v38, v5 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
-; GFX9-NEXT: v_or_b32_sdwa v4, v4, v5 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_0 src1_sel:DWORD
-; GFX9-NEXT: buffer_store_dword v4, v0, s[0:3], 0 offen offset:100
-; GFX9-NEXT: v_lshlrev_b32_e32 v4, 8, v21
-; GFX9-NEXT: v_or_b32_sdwa v4, v53, v4 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
-; GFX9-NEXT: v_or_b32_sdwa v3, v4, v3 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_0 src1_sel:DWORD
-; GFX9-NEXT: buffer_store_dword v3, v0, s[0:3], 0 offen offset:104
-; GFX9-NEXT: v_lshlrev_b32_e32 v3, 8, v22
-; GFX9-NEXT: v_lshlrev_b32_e32 v4, 8, v45
-; GFX9-NEXT: v_or_b32_sdwa v3, v51, v3 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
-; GFX9-NEXT: v_or_b32_sdwa v4, v49, v4 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
-; GFX9-NEXT: v_or_b32_sdwa v3, v3, v4 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_0 src1_sel:DWORD
-; GFX9-NEXT: buffer_store_dword v3, v0, s[0:3], 0 offen offset:108
-; GFX9-NEXT: v_lshlrev_b32_e32 v3, 8, v23
-; GFX9-NEXT: v_or_b32_sdwa v3, v40, v3 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
-; GFX9-NEXT: v_or_b32_sdwa v2, v3, v2 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_0 src1_sel:DWORD
-; GFX9-NEXT: buffer_store_dword v2, v0, s[0:3], 0 offen offset:112
-; GFX9-NEXT: v_lshlrev_b32_e32 v2, 8, v24
-; GFX9-NEXT: v_lshlrev_b32_e32 v3, 8, v43
-; GFX9-NEXT: v_or_b32_sdwa v2, v54, v2 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
-; GFX9-NEXT: v_or_b32_sdwa v3, v52, v3 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
-; GFX9-NEXT: v_or_b32_sdwa v2, v2, v3 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_0 src1_sel:DWORD
-; GFX9-NEXT: buffer_store_dword v2, v0, s[0:3], 0 offen offset:116
-; GFX9-NEXT: v_lshlrev_b32_e32 v2, 8, v25
-; GFX9-NEXT: v_or_b32_sdwa v2, v42, v2 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
-; GFX9-NEXT: v_or_b32_sdwa v1, v2, v1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_0 src1_sel:DWORD
-; GFX9-NEXT: buffer_store_dword v1, v0, s[0:3], 0 offen offset:120
-; GFX9-NEXT: v_lshlrev_b32_e32 v1, 8, v12
+; GFX9-NEXT: v_or_b32_sdwa v2, v3, v2 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; GFX9-NEXT: v_or_b32_sdwa v1, v1, v2 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_0 src1_sel:DWORD
+; GFX9-NEXT: buffer_store_dword v1, v0, s[0:3], 0 offen offset:76
+; GFX9-NEXT: v_lshlrev_b32_e32 v1, 8, v9
+; GFX9-NEXT: v_lshlrev_b32_e32 v2, 8, v13
+; GFX9-NEXT: v_or_b32_sdwa v1, v37, v1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; GFX9-NEXT: v_or_b32_sdwa v2, v60, v2 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; GFX9-NEXT: v_or_b32_sdwa v1, v1, v2 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_0 src1_sel:DWORD
+; GFX9-NEXT: buffer_store_dword v1, v0, s[0:3], 0 offen offset:80
+; GFX9-NEXT: v_lshlrev_b32_e32 v1, 8, v10
+; GFX9-NEXT: v_lshlrev_b32_e32 v2, 8, v59
+; GFX9-NEXT: v_or_b32_sdwa v1, v35, v1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; GFX9-NEXT: v_or_b32_sdwa v2, v33, v2 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; GFX9-NEXT: v_or_b32_sdwa v1, v1, v2 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_0 src1_sel:DWORD
+; GFX9-NEXT: buffer_store_dword v1, v0, s[0:3], 0 offen offset:84
+; GFX9-NEXT: v_lshlrev_b32_e32 v1, 8, v15
+; GFX9-NEXT: v_lshlrev_b32_e32 v2, 8, v12
+; GFX9-NEXT: v_or_b32_sdwa v1, v48, v1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; GFX9-NEXT: v_or_b32_sdwa v2, v58, v2 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; GFX9-NEXT: v_or_b32_sdwa v1, v1, v2 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_0 src1_sel:DWORD
+; GFX9-NEXT: buffer_store_dword v1, v0, s[0:3], 0 offen offset:88
+; GFX9-NEXT: v_lshlrev_b32_e32 v1, 8, v16
+; GFX9-NEXT: v_lshlrev_b32_e32 v2, 8, v57
+; GFX9-NEXT: v_or_b32_sdwa v1, v38, v1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; GFX9-NEXT: v_or_b32_sdwa v2, v36, v2 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; GFX9-NEXT: v_or_b32_sdwa v1, v1, v2 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_0 src1_sel:DWORD
+; GFX9-NEXT: buffer_store_dword v1, v0, s[0:3], 0 offen offset:92
+; GFX9-NEXT: v_lshlrev_b32_e32 v1, 8, v20
+; GFX9-NEXT: v_lshlrev_b32_e32 v2, 8, v11
+; GFX9-NEXT: v_or_b32_sdwa v1, v51, v1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; GFX9-NEXT: v_or_b32_sdwa v2, v56, v2 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; GFX9-NEXT: v_or_b32_sdwa v1, v1, v2 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_0 src1_sel:DWORD
+; GFX9-NEXT: buffer_store_dword v1, v0, s[0:3], 0 offen offset:96
+; GFX9-NEXT: v_lshlrev_b32_e32 v1, 8, v21
+; GFX9-NEXT: v_lshlrev_b32_e32 v2, 8, v47
+; GFX9-NEXT: v_or_b32_sdwa v1, v49, v1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; GFX9-NEXT: v_or_b32_sdwa v2, v39, v2 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; GFX9-NEXT: v_or_b32_sdwa v1, v1, v2 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_0 src1_sel:DWORD
+; GFX9-NEXT: buffer_store_dword v1, v0, s[0:3], 0 offen offset:100
+; GFX9-NEXT: v_lshlrev_b32_e32 v1, 8, v22
+; GFX9-NEXT: v_lshlrev_b32_e32 v2, 8, v7
+; GFX9-NEXT: v_or_b32_sdwa v1, v54, v1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; GFX9-NEXT: v_or_b32_sdwa v2, v46, v2 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; GFX9-NEXT: v_or_b32_sdwa v1, v1, v2 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_0 src1_sel:DWORD
+; GFX9-NEXT: buffer_store_dword v1, v0, s[0:3], 0 offen offset:104
+; GFX9-NEXT: v_lshlrev_b32_e32 v1, 8, v23
+; GFX9-NEXT: v_lshlrev_b32_e32 v2, 8, v45
+; GFX9-NEXT: v_or_b32_sdwa v1, v52, v1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; GFX9-NEXT: v_or_b32_sdwa v2, v50, v2 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; GFX9-NEXT: v_or_b32_sdwa v1, v1, v2 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_0 src1_sel:DWORD
+; GFX9-NEXT: buffer_store_dword v1, v0, s[0:3], 0 offen offset:108
+; GFX9-NEXT: v_lshlrev_b32_e32 v1, 8, v24
; GFX9-NEXT: v_lshlrev_b32_e32 v2, 8, v6
; GFX9-NEXT: v_or_b32_sdwa v1, v41, v1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
-; GFX9-NEXT: v_or_b32_sdwa v2, v55, v2 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; GFX9-NEXT: v_or_b32_sdwa v2, v44, v2 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; GFX9-NEXT: v_or_b32_sdwa v1, v1, v2 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_0 src1_sel:DWORD
+; GFX9-NEXT: buffer_store_dword v1, v0, s[0:3], 0 offen offset:112
+; GFX9-NEXT: v_lshlrev_b32_e32 v1, 8, v25
+; GFX9-NEXT: v_lshlrev_b32_e32 v2, 8, v27
+; GFX9-NEXT: v_or_b32_sdwa v1, v55, v1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; GFX9-NEXT: v_or_b32_sdwa v2, v53, v2 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; GFX9-NEXT: v_or_b32_sdwa v1, v1, v2 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_0 src1_sel:DWORD
+; GFX9-NEXT: buffer_store_dword v1, v0, s[0:3], 0 offen offset:116
+; GFX9-NEXT: v_lshlrev_b32_e32 v1, 8, v26
+; GFX9-NEXT: v_lshlrev_b32_e32 v2, 8, v5
+; GFX9-NEXT: v_or_b32_sdwa v1, v43, v1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; GFX9-NEXT: v_or_b32_sdwa v2, v19, v2 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; GFX9-NEXT: v_or_b32_sdwa v1, v1, v2 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_0 src1_sel:DWORD
+; GFX9-NEXT: buffer_store_dword v1, v0, s[0:3], 0 offen offset:120
+; GFX9-NEXT: v_lshlrev_b32_e32 v1, 8, v14
+; GFX9-NEXT: v_lshlrev_b32_e32 v2, 8, v8
+; GFX9-NEXT: v_or_b32_sdwa v1, v42, v1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; GFX9-NEXT: v_or_b32_sdwa v2, v40, v2 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
; GFX9-NEXT: v_or_b32_sdwa v1, v1, v2 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_0 src1_sel:DWORD
; GFX9-NEXT: buffer_store_dword v1, v0, s[0:3], 0 offen offset:124
; GFX9-NEXT: buffer_load_dword v61, off, s[0:3], s32 ; 4-byte Folded Reload
@@ -175444,26 +175555,26 @@ define inreg <128 x i8> @bitcast_v64bf16_to_v128i8_scalar(<64 x bfloat> inreg %a
; GFX11-TRUE16-NEXT: v_writelane_b32 v41, s97, 1
; GFX11-TRUE16-NEXT: v_readfirstlane_b32 s62, v3
; GFX11-TRUE16-NEXT: v_readfirstlane_b32 s63, v4
-; GFX11-TRUE16-NEXT: v_readfirstlane_b32 s60, v5
+; GFX11-TRUE16-NEXT: v_readfirstlane_b32 s58, v5
; GFX11-TRUE16-NEXT: v_writelane_b32 v40, s34, 2
; GFX11-TRUE16-NEXT: v_writelane_b32 v41, s98, 2
-; GFX11-TRUE16-NEXT: v_readfirstlane_b32 s61, v6
-; GFX11-TRUE16-NEXT: v_readfirstlane_b32 s58, v7
-; GFX11-TRUE16-NEXT: v_readfirstlane_b32 s59, v8
+; GFX11-TRUE16-NEXT: v_readfirstlane_b32 s59, v6
+; GFX11-TRUE16-NEXT: v_readfirstlane_b32 s44, v7
+; GFX11-TRUE16-NEXT: v_readfirstlane_b32 s45, v8
; GFX11-TRUE16-NEXT: v_writelane_b32 v40, s35, 3
; GFX11-TRUE16-NEXT: v_writelane_b32 v41, s99, 3
-; GFX11-TRUE16-NEXT: v_readfirstlane_b32 s44, v9
-; GFX11-TRUE16-NEXT: v_readfirstlane_b32 s45, v10
-; GFX11-TRUE16-NEXT: v_readfirstlane_b32 s42, v11
+; GFX11-TRUE16-NEXT: v_readfirstlane_b32 s8, v9
+; GFX11-TRUE16-NEXT: v_readfirstlane_b32 s9, v10
+; GFX11-TRUE16-NEXT: v_readfirstlane_b32 s6, v11
; GFX11-TRUE16-NEXT: v_writelane_b32 v40, s36, 4
; GFX11-TRUE16-NEXT: v_writelane_b32 v41, s100, 4
-; GFX11-TRUE16-NEXT: v_readfirstlane_b32 s43, v12
-; GFX11-TRUE16-NEXT: v_readfirstlane_b32 s40, v13
-; GFX11-TRUE16-NEXT: v_readfirstlane_b32 s41, v14
+; GFX11-TRUE16-NEXT: v_readfirstlane_b32 s7, v12
+; GFX11-TRUE16-NEXT: v_readfirstlane_b32 s4, v13
+; GFX11-TRUE16-NEXT: v_readfirstlane_b32 s5, v14
; GFX11-TRUE16-NEXT: v_writelane_b32 v40, s37, 5
; GFX11-TRUE16-NEXT: v_writelane_b32 v41, s101, 5
; GFX11-TRUE16-NEXT: s_mov_b32 vcc_hi, 0
-; GFX11-TRUE16-NEXT: s_and_b32 s4, vcc_lo, exec_lo
+; GFX11-TRUE16-NEXT: s_and_b32 s10, vcc_lo, exec_lo
; GFX11-TRUE16-NEXT: ; implicit-def: $vgpr43 : SGPR spill to VGPR lane
; GFX11-TRUE16-NEXT: ; implicit-def: $vgpr42 : SGPR spill to VGPR lane
; GFX11-TRUE16-NEXT: v_writelane_b32 v40, s38, 6
@@ -175497,333 +175608,332 @@ define inreg <128 x i8> @bitcast_v64bf16_to_v128i8_scalar(<64 x bfloat> inreg %a
; GFX11-TRUE16-NEXT: v_writelane_b32 v40, s87, 31
; GFX11-TRUE16-NEXT: s_cbranch_scc0 .LBB91_3
; GFX11-TRUE16-NEXT: ; %bb.1: ; %cmp.false
-; GFX11-TRUE16-NEXT: s_lshr_b32 s4, s27, 24
-; GFX11-TRUE16-NEXT: s_lshr_b64 s[14:15], s[24:25], 24
-; GFX11-TRUE16-NEXT: v_writelane_b32 v42, s4, 7
-; GFX11-TRUE16-NEXT: s_lshr_b32 s4, s27, 16
+; GFX11-TRUE16-NEXT: s_lshr_b32 s10, s27, 24
+; GFX11-TRUE16-NEXT: s_lshr_b64 s[46:47], s[24:25], 24
+; GFX11-TRUE16-NEXT: v_writelane_b32 v42, s10, 7
+; GFX11-TRUE16-NEXT: s_lshr_b32 s10, s27, 16
; GFX11-TRUE16-NEXT: s_lshr_b32 s99, s2, 16
-; GFX11-TRUE16-NEXT: s_lshr_b32 s12, s2, 8
+; GFX11-TRUE16-NEXT: s_lshr_b32 s42, s2, 8
; GFX11-TRUE16-NEXT: s_lshr_b32 s102, s1, 24
-; GFX11-TRUE16-NEXT: v_writelane_b32 v42, s4, 6
-; GFX11-TRUE16-NEXT: s_lshr_b32 s4, s27, 8
+; GFX11-TRUE16-NEXT: v_writelane_b32 v42, s10, 6
+; GFX11-TRUE16-NEXT: s_lshr_b32 s10, s27, 8
; GFX11-TRUE16-NEXT: s_lshr_b32 s103, s1, 8
; GFX11-TRUE16-NEXT: s_lshr_b32 s104, s0, 16
; GFX11-TRUE16-NEXT: s_lshr_b32 s34, s0, 8
-; GFX11-TRUE16-NEXT: v_writelane_b32 v42, s4, 5
-; GFX11-TRUE16-NEXT: s_lshr_b32 s4, s26, 16
-; GFX11-TRUE16-NEXT: s_lshr_b32 s35, s41, 24
-; GFX11-TRUE16-NEXT: s_lshr_b32 s87, s41, 16
-; GFX11-TRUE16-NEXT: s_lshr_b32 s55, s41, 8
-; GFX11-TRUE16-NEXT: v_writelane_b32 v42, s4, 4
-; GFX11-TRUE16-NEXT: s_lshr_b32 s4, s26, 8
-; GFX11-TRUE16-NEXT: s_lshr_b32 s69, s40, 16
-; GFX11-TRUE16-NEXT: s_lshr_b32 s70, s40, 8
-; GFX11-TRUE16-NEXT: s_lshr_b32 s36, s43, 24
-; GFX11-TRUE16-NEXT: v_writelane_b32 v42, s4, 3
-; GFX11-TRUE16-NEXT: s_lshr_b32 s4, s25, 24
-; GFX11-TRUE16-NEXT: s_lshr_b32 s96, s43, 16
-; GFX11-TRUE16-NEXT: s_lshr_b32 s38, s43, 8
-; GFX11-TRUE16-NEXT: s_lshr_b32 s71, s42, 16
-; GFX11-TRUE16-NEXT: v_writelane_b32 v42, s4, 2
-; GFX11-TRUE16-NEXT: s_lshr_b32 s4, s25, 16
-; GFX11-TRUE16-NEXT: s_lshr_b32 s48, s42, 8
+; GFX11-TRUE16-NEXT: v_writelane_b32 v42, s10, 5
+; GFX11-TRUE16-NEXT: s_lshr_b32 s10, s26, 16
+; GFX11-TRUE16-NEXT: s_lshr_b32 s35, s5, 24
+; GFX11-TRUE16-NEXT: s_lshr_b32 s87, s5, 16
+; GFX11-TRUE16-NEXT: s_lshr_b32 s54, s5, 8
+; GFX11-TRUE16-NEXT: v_writelane_b32 v42, s10, 4
+; GFX11-TRUE16-NEXT: s_lshr_b32 s10, s26, 8
+; GFX11-TRUE16-NEXT: s_lshr_b32 s68, s4, 16
+; GFX11-TRUE16-NEXT: s_lshr_b32 s69, s4, 8
+; GFX11-TRUE16-NEXT: s_lshr_b32 s36, s7, 24
+; GFX11-TRUE16-NEXT: v_writelane_b32 v42, s10, 3
+; GFX11-TRUE16-NEXT: s_lshr_b32 s10, s25, 24
+; GFX11-TRUE16-NEXT: s_lshr_b32 s96, s7, 16
+; GFX11-TRUE16-NEXT: s_lshr_b32 s39, s7, 8
+; GFX11-TRUE16-NEXT: s_lshr_b32 s70, s6, 16
+; GFX11-TRUE16-NEXT: v_writelane_b32 v42, s10, 2
+; GFX11-TRUE16-NEXT: s_lshr_b32 s10, s25, 16
+; GFX11-TRUE16-NEXT: s_lshr_b32 s49, s6, 8
+; GFX11-TRUE16-NEXT: s_lshr_b32 s37, s9, 24
+; GFX11-TRUE16-NEXT: s_lshr_b32 s14, s9, 16
+; GFX11-TRUE16-NEXT: v_writelane_b32 v42, s10, 1
+; GFX11-TRUE16-NEXT: s_lshr_b32 s10, s25, 8
+; GFX11-TRUE16-NEXT: s_lshr_b32 s55, s9, 8
+; GFX11-TRUE16-NEXT: s_lshr_b32 s71, s8, 16
+; GFX11-TRUE16-NEXT: s_lshr_b32 s80, s8, 8
+; GFX11-TRUE16-NEXT: v_writelane_b32 v42, s10, 0
+; GFX11-TRUE16-NEXT: s_lshr_b32 s10, s24, 16
; GFX11-TRUE16-NEXT: s_lshr_b32 s50, s45, 24
-; GFX11-TRUE16-NEXT: s_lshr_b32 s8, s45, 16
-; GFX11-TRUE16-NEXT: v_writelane_b32 v42, s4, 1
-; GFX11-TRUE16-NEXT: s_lshr_b32 s4, s25, 8
+; GFX11-TRUE16-NEXT: v_writelane_b32 v43, s10, 31
+; GFX11-TRUE16-NEXT: s_lshr_b32 s10, s24, 8
+; GFX11-TRUE16-NEXT: v_writelane_b32 v42, s46, 8
+; GFX11-TRUE16-NEXT: s_lshr_b32 s15, s45, 16
; GFX11-TRUE16-NEXT: s_lshr_b32 s64, s45, 8
-; GFX11-TRUE16-NEXT: s_lshr_b32 s80, s44, 16
-; GFX11-TRUE16-NEXT: s_lshr_b32 s81, s44, 8
-; GFX11-TRUE16-NEXT: v_writelane_b32 v42, s4, 0
-; GFX11-TRUE16-NEXT: s_lshr_b32 s4, s24, 16
-; GFX11-TRUE16-NEXT: s_lshr_b32 s51, s59, 24
-; GFX11-TRUE16-NEXT: v_writelane_b32 v43, s4, 31
-; GFX11-TRUE16-NEXT: s_lshr_b32 s4, s24, 8
-; GFX11-TRUE16-NEXT: v_writelane_b32 v42, s14, 8
-; GFX11-TRUE16-NEXT: s_lshr_b32 s9, s59, 16
+; GFX11-TRUE16-NEXT: v_writelane_b32 v43, s10, 30
+; GFX11-TRUE16-NEXT: s_lshr_b32 s10, s23, 24
+; GFX11-TRUE16-NEXT: v_writelane_b32 v42, s47, 9
+; GFX11-TRUE16-NEXT: s_lshr_b64 s[46:47], s[22:23], 24
+; GFX11-TRUE16-NEXT: s_lshr_b32 s81, s44, 16
+; GFX11-TRUE16-NEXT: v_writelane_b32 v43, s10, 29
+; GFX11-TRUE16-NEXT: s_lshr_b32 s10, s23, 16
+; GFX11-TRUE16-NEXT: s_lshr_b32 s82, s44, 8
+; GFX11-TRUE16-NEXT: s_lshr_b32 s38, s59, 24
+; GFX11-TRUE16-NEXT: s_lshr_b32 s40, s59, 16
+; GFX11-TRUE16-NEXT: v_writelane_b32 v43, s10, 28
+; GFX11-TRUE16-NEXT: s_lshr_b32 s10, s23, 8
; GFX11-TRUE16-NEXT: s_lshr_b32 s65, s59, 8
-; GFX11-TRUE16-NEXT: v_writelane_b32 v43, s4, 30
-; GFX11-TRUE16-NEXT: s_lshr_b32 s4, s23, 24
-; GFX11-TRUE16-NEXT: v_writelane_b32 v42, s15, 9
-; GFX11-TRUE16-NEXT: s_lshr_b64 s[14:15], s[22:23], 24
-; GFX11-TRUE16-NEXT: s_lshr_b32 s82, s58, 16
-; GFX11-TRUE16-NEXT: v_writelane_b32 v43, s4, 29
-; GFX11-TRUE16-NEXT: s_lshr_b32 s4, s23, 16
-; GFX11-TRUE16-NEXT: s_lshr_b32 s49, s58, 8
-; GFX11-TRUE16-NEXT: s_lshr_b32 s37, s61, 24
-; GFX11-TRUE16-NEXT: s_lshr_b32 s10, s61, 16
-; GFX11-TRUE16-NEXT: v_writelane_b32 v43, s4, 28
-; GFX11-TRUE16-NEXT: s_lshr_b32 s4, s23, 8
-; GFX11-TRUE16-NEXT: s_lshr_b32 s66, s61, 8
-; GFX11-TRUE16-NEXT: s_lshr_b32 s83, s60, 16
-; GFX11-TRUE16-NEXT: s_lshr_b32 s84, s60, 8
-; GFX11-TRUE16-NEXT: v_writelane_b32 v43, s4, 27
-; GFX11-TRUE16-NEXT: s_lshr_b32 s4, s22, 16
-; GFX11-TRUE16-NEXT: s_lshr_b32 s52, s63, 24
-; GFX11-TRUE16-NEXT: s_lshr_b32 s11, s63, 16
-; GFX11-TRUE16-NEXT: s_lshr_b32 s67, s63, 8
-; GFX11-TRUE16-NEXT: v_writelane_b32 v43, s4, 24
-; GFX11-TRUE16-NEXT: s_lshr_b32 s4, s22, 8
+; GFX11-TRUE16-NEXT: s_lshr_b32 s83, s58, 16
+; GFX11-TRUE16-NEXT: s_lshr_b32 s84, s58, 8
+; GFX11-TRUE16-NEXT: v_writelane_b32 v43, s10, 27
+; GFX11-TRUE16-NEXT: s_lshr_b32 s10, s22, 16
+; GFX11-TRUE16-NEXT: s_lshr_b32 s51, s63, 24
+; GFX11-TRUE16-NEXT: s_lshr_b32 s41, s63, 16
+; GFX11-TRUE16-NEXT: s_lshr_b32 s66, s63, 8
+; GFX11-TRUE16-NEXT: v_writelane_b32 v43, s10, 24
+; GFX11-TRUE16-NEXT: s_lshr_b32 s10, s22, 8
; GFX11-TRUE16-NEXT: s_lshr_b32 s85, s62, 16
-; GFX11-TRUE16-NEXT: s_lshr_b32 s53, s73, 24
+; GFX11-TRUE16-NEXT: s_lshr_b32 s52, s73, 24
; GFX11-TRUE16-NEXT: s_lshr_b32 s97, s73, 16
-; GFX11-TRUE16-NEXT: v_writelane_b32 v43, s4, 23
-; GFX11-TRUE16-NEXT: s_lshr_b32 s4, s21, 24
-; GFX11-TRUE16-NEXT: s_lshr_b32 s39, s73, 8
-; GFX11-TRUE16-NEXT: s_lshr_b32 s5, s72, 16
-; GFX11-TRUE16-NEXT: s_lshr_b32 s6, s72, 8
-; GFX11-TRUE16-NEXT: v_writelane_b32 v43, s4, 22
-; GFX11-TRUE16-NEXT: s_lshr_b32 s4, s21, 16
-; GFX11-TRUE16-NEXT: s_lshr_b32 s54, s29, 24
+; GFX11-TRUE16-NEXT: v_writelane_b32 v43, s10, 23
+; GFX11-TRUE16-NEXT: s_lshr_b32 s10, s21, 24
+; GFX11-TRUE16-NEXT: s_lshr_b32 s48, s73, 8
+; GFX11-TRUE16-NEXT: s_lshr_b32 s11, s72, 16
+; GFX11-TRUE16-NEXT: s_lshr_b32 s12, s72, 8
+; GFX11-TRUE16-NEXT: v_writelane_b32 v43, s10, 22
+; GFX11-TRUE16-NEXT: s_lshr_b32 s10, s21, 16
+; GFX11-TRUE16-NEXT: s_lshr_b32 s53, s29, 24
; GFX11-TRUE16-NEXT: s_lshr_b32 s98, s29, 16
-; GFX11-TRUE16-NEXT: s_lshr_b32 s68, s29, 8
-; GFX11-TRUE16-NEXT: v_writelane_b32 v43, s4, 21
-; GFX11-TRUE16-NEXT: s_lshr_b32 s4, s21, 8
-; GFX11-TRUE16-NEXT: s_lshr_b32 s7, s28, 16
+; GFX11-TRUE16-NEXT: s_lshr_b32 s67, s29, 8
+; GFX11-TRUE16-NEXT: v_writelane_b32 v43, s10, 21
+; GFX11-TRUE16-NEXT: s_lshr_b32 s10, s21, 8
+; GFX11-TRUE16-NEXT: s_lshr_b32 s13, s28, 16
; GFX11-TRUE16-NEXT: s_lshr_b32 s86, s28, 8
; GFX11-TRUE16-NEXT: s_lshr_b64 s[100:101], s[26:27], 24
-; GFX11-TRUE16-NEXT: v_writelane_b32 v43, s4, 20
-; GFX11-TRUE16-NEXT: s_lshr_b32 s4, s20, 16
-; GFX11-TRUE16-NEXT: s_lshr_b64 s[46:47], s[2:3], 24
-; GFX11-TRUE16-NEXT: s_lshr_b64 s[56:57], s[0:1], 24
-; GFX11-TRUE16-NEXT: s_lshr_b64 s[74:75], s[40:41], 24
-; GFX11-TRUE16-NEXT: v_writelane_b32 v43, s4, 17
-; GFX11-TRUE16-NEXT: s_lshr_b32 s4, s20, 8
-; GFX11-TRUE16-NEXT: s_lshr_b64 s[76:77], s[42:43], 24
-; GFX11-TRUE16-NEXT: s_lshr_b64 s[78:79], s[44:45], 24
-; GFX11-TRUE16-NEXT: s_lshr_b64 s[88:89], s[58:59], 24
-; GFX11-TRUE16-NEXT: v_writelane_b32 v43, s4, 16
-; GFX11-TRUE16-NEXT: s_lshr_b32 s4, s19, 24
-; GFX11-TRUE16-NEXT: s_lshr_b64 s[90:91], s[60:61], 24
+; GFX11-TRUE16-NEXT: v_writelane_b32 v43, s10, 20
+; GFX11-TRUE16-NEXT: s_lshr_b32 s10, s20, 16
+; GFX11-TRUE16-NEXT: s_lshr_b64 s[56:57], s[2:3], 24
+; GFX11-TRUE16-NEXT: s_lshr_b64 s[60:61], s[0:1], 24
+; GFX11-TRUE16-NEXT: s_lshr_b64 s[74:75], s[4:5], 24
+; GFX11-TRUE16-NEXT: v_writelane_b32 v43, s10, 17
+; GFX11-TRUE16-NEXT: s_lshr_b32 s10, s20, 8
+; GFX11-TRUE16-NEXT: s_lshr_b64 s[76:77], s[6:7], 24
+; GFX11-TRUE16-NEXT: s_lshr_b64 s[78:79], s[8:9], 24
+; GFX11-TRUE16-NEXT: s_lshr_b64 s[88:89], s[44:45], 24
+; GFX11-TRUE16-NEXT: v_writelane_b32 v43, s10, 16
+; GFX11-TRUE16-NEXT: s_lshr_b32 s10, s19, 24
+; GFX11-TRUE16-NEXT: s_lshr_b64 s[90:91], s[58:59], 24
; GFX11-TRUE16-NEXT: s_lshr_b64 s[92:93], s[62:63], 24
; GFX11-TRUE16-NEXT: s_lshr_b64 s[94:95], s[72:73], 24
-; GFX11-TRUE16-NEXT: v_writelane_b32 v43, s4, 15
-; GFX11-TRUE16-NEXT: s_lshr_b32 s4, s19, 16
+; GFX11-TRUE16-NEXT: v_writelane_b32 v43, s10, 15
+; GFX11-TRUE16-NEXT: s_lshr_b32 s10, s19, 16
; GFX11-TRUE16-NEXT: s_lshr_b64 s[30:31], s[28:29], 24
-; GFX11-TRUE16-NEXT: v_writelane_b32 v43, s4, 14
-; GFX11-TRUE16-NEXT: s_lshr_b32 s4, s19, 8
+; GFX11-TRUE16-NEXT: v_writelane_b32 v43, s10, 14
+; GFX11-TRUE16-NEXT: s_lshr_b32 s10, s19, 8
; GFX11-TRUE16-NEXT: s_delay_alu instid0(SALU_CYCLE_1) | instskip(SKIP_1) | instid1(SALU_CYCLE_1)
-; GFX11-TRUE16-NEXT: v_writelane_b32 v43, s4, 13
-; GFX11-TRUE16-NEXT: s_lshr_b32 s4, s18, 16
-; GFX11-TRUE16-NEXT: v_writelane_b32 v43, s4, 10
-; GFX11-TRUE16-NEXT: s_lshr_b32 s4, s18, 8
+; GFX11-TRUE16-NEXT: v_writelane_b32 v43, s10, 13
+; GFX11-TRUE16-NEXT: s_lshr_b32 s10, s18, 16
+; GFX11-TRUE16-NEXT: v_writelane_b32 v43, s10, 10
+; GFX11-TRUE16-NEXT: s_lshr_b32 s10, s18, 8
; GFX11-TRUE16-NEXT: s_delay_alu instid0(SALU_CYCLE_1) | instskip(SKIP_1) | instid1(SALU_CYCLE_1)
-; GFX11-TRUE16-NEXT: v_writelane_b32 v43, s4, 9
-; GFX11-TRUE16-NEXT: s_lshr_b32 s4, s17, 24
-; GFX11-TRUE16-NEXT: v_writelane_b32 v43, s4, 8
-; GFX11-TRUE16-NEXT: s_lshr_b32 s4, s17, 16
+; GFX11-TRUE16-NEXT: v_writelane_b32 v43, s10, 9
+; GFX11-TRUE16-NEXT: s_lshr_b32 s10, s17, 24
+; GFX11-TRUE16-NEXT: v_writelane_b32 v43, s10, 8
+; GFX11-TRUE16-NEXT: s_lshr_b32 s10, s17, 16
; GFX11-TRUE16-NEXT: s_delay_alu instid0(SALU_CYCLE_1) | instskip(SKIP_1) | instid1(SALU_CYCLE_1)
-; GFX11-TRUE16-NEXT: v_writelane_b32 v43, s4, 7
-; GFX11-TRUE16-NEXT: s_lshr_b32 s4, s17, 8
-; GFX11-TRUE16-NEXT: v_writelane_b32 v43, s4, 6
-; GFX11-TRUE16-NEXT: s_lshr_b32 s4, s16, 16
+; GFX11-TRUE16-NEXT: v_writelane_b32 v43, s10, 7
+; GFX11-TRUE16-NEXT: s_lshr_b32 s10, s17, 8
+; GFX11-TRUE16-NEXT: v_writelane_b32 v43, s10, 6
+; GFX11-TRUE16-NEXT: s_lshr_b32 s10, s16, 16
; GFX11-TRUE16-NEXT: s_delay_alu instid0(SALU_CYCLE_1) | instskip(SKIP_1) | instid1(SALU_CYCLE_1)
-; GFX11-TRUE16-NEXT: v_writelane_b32 v43, s4, 5
-; GFX11-TRUE16-NEXT: s_lshr_b32 s4, s16, 8
-; GFX11-TRUE16-NEXT: v_writelane_b32 v43, s4, 4
-; GFX11-TRUE16-NEXT: s_lshr_b32 s4, s3, 24
+; GFX11-TRUE16-NEXT: v_writelane_b32 v43, s10, 5
+; GFX11-TRUE16-NEXT: s_lshr_b32 s10, s16, 8
+; GFX11-TRUE16-NEXT: v_writelane_b32 v43, s10, 4
+; GFX11-TRUE16-NEXT: s_lshr_b32 s10, s3, 24
; GFX11-TRUE16-NEXT: s_delay_alu instid0(SALU_CYCLE_1) | instskip(SKIP_1) | instid1(SALU_CYCLE_1)
-; GFX11-TRUE16-NEXT: v_writelane_b32 v43, s4, 3
-; GFX11-TRUE16-NEXT: s_lshr_b32 s4, s3, 16
-; GFX11-TRUE16-NEXT: v_writelane_b32 v43, s4, 2
-; GFX11-TRUE16-NEXT: s_lshr_b32 s4, s3, 8
+; GFX11-TRUE16-NEXT: v_writelane_b32 v43, s10, 3
+; GFX11-TRUE16-NEXT: s_lshr_b32 s10, s3, 16
+; GFX11-TRUE16-NEXT: v_writelane_b32 v43, s10, 2
+; GFX11-TRUE16-NEXT: s_lshr_b32 s10, s3, 8
; GFX11-TRUE16-NEXT: s_delay_alu instid0(SALU_CYCLE_1) | instskip(SKIP_1) | instid1(SALU_CYCLE_1)
-; GFX11-TRUE16-NEXT: v_writelane_b32 v43, s4, 1
-; GFX11-TRUE16-NEXT: s_lshr_b32 s4, s1, 16
-; GFX11-TRUE16-NEXT: v_writelane_b32 v43, s4, 0
-; GFX11-TRUE16-NEXT: s_lshr_b32 s4, s62, 8
-; GFX11-TRUE16-NEXT: v_writelane_b32 v43, s14, 25
-; GFX11-TRUE16-NEXT: v_writelane_b32 v43, s15, 26
-; GFX11-TRUE16-NEXT: s_lshr_b64 s[14:15], s[20:21], 24
+; GFX11-TRUE16-NEXT: v_writelane_b32 v43, s10, 1
+; GFX11-TRUE16-NEXT: s_lshr_b32 s10, s1, 16
+; GFX11-TRUE16-NEXT: v_writelane_b32 v43, s10, 0
+; GFX11-TRUE16-NEXT: s_lshr_b32 s10, s62, 8
+; GFX11-TRUE16-NEXT: v_writelane_b32 v43, s46, 25
+; GFX11-TRUE16-NEXT: v_writelane_b32 v43, s47, 26
+; GFX11-TRUE16-NEXT: s_lshr_b64 s[46:47], s[20:21], 24
; GFX11-TRUE16-NEXT: s_delay_alu instid0(SALU_CYCLE_1) | instskip(SKIP_2) | instid1(SALU_CYCLE_1)
-; GFX11-TRUE16-NEXT: v_writelane_b32 v43, s14, 18
-; GFX11-TRUE16-NEXT: v_writelane_b32 v43, s15, 19
-; GFX11-TRUE16-NEXT: s_lshr_b64 s[14:15], s[18:19], 24
-; GFX11-TRUE16-NEXT: v_writelane_b32 v43, s14, 11
-; GFX11-TRUE16-NEXT: v_writelane_b32 v43, s15, 12
-; GFX11-TRUE16-NEXT: s_lshr_b64 s[14:15], s[16:17], 24
+; GFX11-TRUE16-NEXT: v_writelane_b32 v43, s46, 18
+; GFX11-TRUE16-NEXT: v_writelane_b32 v43, s47, 19
+; GFX11-TRUE16-NEXT: s_lshr_b64 s[46:47], s[18:19], 24
+; GFX11-TRUE16-NEXT: v_writelane_b32 v43, s46, 11
+; GFX11-TRUE16-NEXT: v_writelane_b32 v43, s47, 12
+; GFX11-TRUE16-NEXT: s_lshr_b64 s[46:47], s[16:17], 24
; GFX11-TRUE16-NEXT: s_and_not1_b32 vcc_lo, exec_lo, vcc_hi
; GFX11-TRUE16-NEXT: s_cbranch_vccnz .LBB91_4
; GFX11-TRUE16-NEXT: .LBB91_2: ; %cmp.true
-; GFX11-TRUE16-NEXT: s_and_b32 s4, s29, 0xffff0000
-; GFX11-TRUE16-NEXT: s_and_b32 s14, s58, 0xffff0000
-; GFX11-TRUE16-NEXT: v_add_f32_e64 v1, 0x40c00000, s4
-; GFX11-TRUE16-NEXT: s_and_b32 s4, s1, 0xffff0000
-; GFX11-TRUE16-NEXT: s_lshl_b32 s15, s58, 16
-; GFX11-TRUE16-NEXT: v_add_f32_e64 v6, 0x40c00000, s4
-; GFX11-TRUE16-NEXT: s_lshl_b32 s6, s29, 16
+; GFX11-TRUE16-NEXT: s_and_b32 s10, s29, 0xffff0000
+; GFX11-TRUE16-NEXT: s_and_b32 s42, s58, 0xffff0000
+; GFX11-TRUE16-NEXT: v_add_f32_e64 v1, 0x40c00000, s10
+; GFX11-TRUE16-NEXT: s_and_b32 s10, s1, 0xffff0000
+; GFX11-TRUE16-NEXT: s_lshl_b32 s43, s58, 16
+; GFX11-TRUE16-NEXT: v_add_f32_e64 v6, 0x40c00000, s10
+; GFX11-TRUE16-NEXT: s_lshl_b32 s11, s29, 16
; GFX11-TRUE16-NEXT: v_bfe_u32 v3, v1, 16, 1
-; GFX11-TRUE16-NEXT: v_add_f32_e64 v2, 0x40c00000, s6
-; GFX11-TRUE16-NEXT: s_and_b32 s8, s41, 0xffff0000
+; GFX11-TRUE16-NEXT: v_add_f32_e64 v2, 0x40c00000, s11
+; GFX11-TRUE16-NEXT: s_and_b32 s13, s6, 0xffff0000
; GFX11-TRUE16-NEXT: v_readfirstlane_b32 s58, v6
-; GFX11-TRUE16-NEXT: v_cmp_u_f32_e32 vcc_lo, v6, v6
+; GFX11-TRUE16-NEXT: s_lshl_b32 s10, s6, 16
; GFX11-TRUE16-NEXT: v_add_nc_u32_e32 v3, v3, v1
-; GFX11-TRUE16-NEXT: s_lshl_b32 s7, s41, 16
-; GFX11-TRUE16-NEXT: s_and_b32 s77, s28, 0xffff0000
+; GFX11-TRUE16-NEXT: v_cmp_u_f32_e32 vcc_lo, v6, v6
+; GFX11-TRUE16-NEXT: s_and_b32 s47, s59, 0xffff0000
; GFX11-TRUE16-NEXT: s_bfe_u32 s6, s58, 0x10010
-; GFX11-TRUE16-NEXT: s_lshl_b32 s78, s28, 16
-; GFX11-TRUE16-NEXT: s_add_i32 s41, s6, s58
-; GFX11-TRUE16-NEXT: s_and_b32 s5, s73, 0xffff0000
-; GFX11-TRUE16-NEXT: s_lshl_b32 s76, s73, 16
-; GFX11-TRUE16-NEXT: s_and_b32 s74, s72, 0xffff0000
-; GFX11-TRUE16-NEXT: s_lshl_b32 s75, s72, 16
+; GFX11-TRUE16-NEXT: s_lshl_b32 s46, s59, 16
+; GFX11-TRUE16-NEXT: s_add_i32 s59, s6, s58
+; GFX11-TRUE16-NEXT: s_and_b32 s75, s28, 0xffff0000
+; GFX11-TRUE16-NEXT: s_lshl_b32 s76, s28, 16
+; GFX11-TRUE16-NEXT: s_and_b32 s74, s73, 0xffff0000
+; GFX11-TRUE16-NEXT: s_lshl_b32 s73, s73, 16
+; GFX11-TRUE16-NEXT: s_and_b32 s61, s72, 0xffff0000
+; GFX11-TRUE16-NEXT: s_lshl_b32 s72, s72, 16
; GFX11-TRUE16-NEXT: s_and_b32 s12, s63, 0xffff0000
-; GFX11-TRUE16-NEXT: s_lshl_b32 s73, s63, 16
-; GFX11-TRUE16-NEXT: s_and_b32 s63, s62, 0xffff0000
-; GFX11-TRUE16-NEXT: s_lshl_b32 s72, s62, 16
-; GFX11-TRUE16-NEXT: s_and_b32 s62, s61, 0xffff0000
-; GFX11-TRUE16-NEXT: s_lshl_b32 s57, s61, 16
-; GFX11-TRUE16-NEXT: s_and_b32 s47, s60, 0xffff0000
-; GFX11-TRUE16-NEXT: s_lshl_b32 s56, s60, 16
-; GFX11-TRUE16-NEXT: s_and_b32 s46, s59, 0xffff0000
-; GFX11-TRUE16-NEXT: s_lshl_b32 s28, s59, 16
-; GFX11-TRUE16-NEXT: s_and_b32 s11, s45, 0xffff0000
-; GFX11-TRUE16-NEXT: s_lshl_b32 s9, s45, 16
-; GFX11-TRUE16-NEXT: s_and_b32 s45, s44, 0xffff0000
-; GFX11-TRUE16-NEXT: s_lshl_b32 s44, s44, 16
-; GFX11-TRUE16-NEXT: s_and_b32 s29, s43, 0xffff0000
-; GFX11-TRUE16-NEXT: s_lshl_b32 s43, s43, 16
-; GFX11-TRUE16-NEXT: s_and_b32 s13, s42, 0xffff0000
-; GFX11-TRUE16-NEXT: s_lshl_b32 s10, s42, 16
-; GFX11-TRUE16-NEXT: s_and_b32 s4, s40, 0xffff0000
-; GFX11-TRUE16-NEXT: s_lshl_b32 s6, s40, 16
-; GFX11-TRUE16-NEXT: s_addk_i32 s41, 0x7fff
+; GFX11-TRUE16-NEXT: s_lshl_b32 s60, s63, 16
+; GFX11-TRUE16-NEXT: s_and_b32 s56, s62, 0xffff0000
+; GFX11-TRUE16-NEXT: s_lshl_b32 s57, s62, 16
+; GFX11-TRUE16-NEXT: s_and_b32 s41, s45, 0xffff0000
+; GFX11-TRUE16-NEXT: s_lshl_b32 s28, s45, 16
+; GFX11-TRUE16-NEXT: s_and_b32 s14, s44, 0xffff0000
+; GFX11-TRUE16-NEXT: s_lshl_b32 s15, s44, 16
+; GFX11-TRUE16-NEXT: s_and_b32 s11, s9, 0xffff0000
+; GFX11-TRUE16-NEXT: s_lshl_b32 s9, s9, 16
+; GFX11-TRUE16-NEXT: s_and_b32 s44, s8, 0xffff0000
+; GFX11-TRUE16-NEXT: s_lshl_b32 s45, s8, 16
+; GFX11-TRUE16-NEXT: s_and_b32 s29, s7, 0xffff0000
+; GFX11-TRUE16-NEXT: s_lshl_b32 s40, s7, 16
+; GFX11-TRUE16-NEXT: s_and_b32 s8, s5, 0xffff0000
+; GFX11-TRUE16-NEXT: s_lshl_b32 s7, s5, 16
+; GFX11-TRUE16-NEXT: s_and_b32 s5, s4, 0xffff0000
+; GFX11-TRUE16-NEXT: s_lshl_b32 s6, s4, 16
+; GFX11-TRUE16-NEXT: s_addk_i32 s59, 0x7fff
; GFX11-TRUE16-NEXT: s_bitset1_b32 s58, 22
; GFX11-TRUE16-NEXT: v_bfe_u32 v4, v2, 16, 1
; GFX11-TRUE16-NEXT: v_or_b32_e32 v5, 0x400000, v1
; GFX11-TRUE16-NEXT: v_add_nc_u32_e32 v3, 0x7fff, v3
-; GFX11-TRUE16-NEXT: s_and_b32 s40, vcc_lo, exec_lo
+; GFX11-TRUE16-NEXT: s_and_b32 s4, vcc_lo, exec_lo
; GFX11-TRUE16-NEXT: v_cmp_u_f32_e32 vcc_lo, v1, v1
-; GFX11-TRUE16-NEXT: s_cselect_b32 s40, s58, s41
+; GFX11-TRUE16-NEXT: s_cselect_b32 s4, s58, s59
; GFX11-TRUE16-NEXT: s_lshl_b32 s1, s1, 16
; GFX11-TRUE16-NEXT: v_add_nc_u32_e32 v4, v4, v2
-; GFX11-TRUE16-NEXT: s_lshr_b32 s58, s40, 16
+; GFX11-TRUE16-NEXT: s_lshr_b32 s63, s4, 16
; GFX11-TRUE16-NEXT: v_cndmask_b32_e32 v1, v3, v5, vcc_lo
; GFX11-TRUE16-NEXT: v_add_f32_e64 v3, 0x40c00000, s1
; GFX11-TRUE16-NEXT: v_or_b32_e32 v5, 0x400000, v2
; GFX11-TRUE16-NEXT: v_add_nc_u32_e32 v4, 0x7fff, v4
; GFX11-TRUE16-NEXT: v_cmp_u_f32_e32 vcc_lo, v2, v2
-; GFX11-TRUE16-NEXT: v_add_f32_e64 v6, 0x40c00000, s77
+; GFX11-TRUE16-NEXT: v_add_f32_e64 v6, 0x40c00000, s75
; GFX11-TRUE16-NEXT: v_readfirstlane_b32 s1, v3
-; GFX11-TRUE16-NEXT: v_add_f32_e64 v7, 0x40c00000, s78
+; GFX11-TRUE16-NEXT: v_add_f32_e64 v7, 0x40c00000, s76
; GFX11-TRUE16-NEXT: v_lshrrev_b32_e32 v20, 16, v1
; GFX11-TRUE16-NEXT: v_cndmask_b32_e32 v2, v4, v5, vcc_lo
; GFX11-TRUE16-NEXT: v_cmp_u_f32_e32 vcc_lo, v3, v3
-; GFX11-TRUE16-NEXT: s_bfe_u32 s41, s1, 0x10010
+; GFX11-TRUE16-NEXT: s_bfe_u32 s58, s1, 0x10010
; GFX11-TRUE16-NEXT: v_bfe_u32 v4, v6, 16, 1
-; GFX11-TRUE16-NEXT: s_add_i32 s41, s41, s1
+; GFX11-TRUE16-NEXT: s_add_i32 s58, s58, s1
; GFX11-TRUE16-NEXT: s_bitset1_b32 s1, 22
-; GFX11-TRUE16-NEXT: s_addk_i32 s41, 0x7fff
-; GFX11-TRUE16-NEXT: s_and_b32 s40, vcc_lo, exec_lo
-; GFX11-TRUE16-NEXT: s_cselect_b32 s1, s1, s41
-; GFX11-TRUE16-NEXT: s_and_b32 s40, s0, 0xffff0000
+; GFX11-TRUE16-NEXT: s_addk_i32 s58, 0x7fff
+; GFX11-TRUE16-NEXT: s_and_b32 s4, vcc_lo, exec_lo
+; GFX11-TRUE16-NEXT: s_cselect_b32 s1, s1, s58
+; GFX11-TRUE16-NEXT: s_and_b32 s4, s0, 0xffff0000
; GFX11-TRUE16-NEXT: v_lshrrev_b32_e32 v21, 16, v2
-; GFX11-TRUE16-NEXT: v_add_f32_e64 v2, 0x40c00000, s40
+; GFX11-TRUE16-NEXT: v_add_f32_e64 v2, 0x40c00000, s4
; GFX11-TRUE16-NEXT: v_bfe_u32 v5, v7, 16, 1
; GFX11-TRUE16-NEXT: v_add_nc_u32_e32 v3, v4, v6
; GFX11-TRUE16-NEXT: s_lshr_b32 s1, s1, 16
; GFX11-TRUE16-NEXT: v_or_b32_e32 v8, 0x400000, v7
-; GFX11-TRUE16-NEXT: v_readfirstlane_b32 s40, v2
+; GFX11-TRUE16-NEXT: v_readfirstlane_b32 s4, v2
; GFX11-TRUE16-NEXT: v_cmp_u_f32_e32 vcc_lo, v2, v2
; GFX11-TRUE16-NEXT: v_add_nc_u32_e32 v3, 0x7fff, v3
; GFX11-TRUE16-NEXT: v_add_f32_e64 v51, 0x40c00000, s6
-; GFX11-TRUE16-NEXT: s_bfe_u32 s41, s40, 0x10010
+; GFX11-TRUE16-NEXT: s_bfe_u32 s58, s4, 0x10010
; GFX11-TRUE16-NEXT: s_delay_alu instid0(SALU_CYCLE_1)
-; GFX11-TRUE16-NEXT: s_add_i32 s41, s41, s40
-; GFX11-TRUE16-NEXT: s_bitset1_b32 s40, 22
-; GFX11-TRUE16-NEXT: s_addk_i32 s41, 0x7fff
-; GFX11-TRUE16-NEXT: s_and_b32 s42, vcc_lo, exec_lo
+; GFX11-TRUE16-NEXT: s_add_i32 s58, s58, s4
+; GFX11-TRUE16-NEXT: s_bitset1_b32 s4, 22
+; GFX11-TRUE16-NEXT: s_addk_i32 s58, 0x7fff
+; GFX11-TRUE16-NEXT: s_and_b32 s59, vcc_lo, exec_lo
; GFX11-TRUE16-NEXT: v_cmp_u_f32_e32 vcc_lo, v6, v6
; GFX11-TRUE16-NEXT: v_add_nc_u32_e32 v1, v5, v7
; GFX11-TRUE16-NEXT: v_or_b32_e32 v5, 0x400000, v6
-; GFX11-TRUE16-NEXT: s_cselect_b32 s40, s40, s41
+; GFX11-TRUE16-NEXT: s_cselect_b32 s4, s4, s58
; GFX11-TRUE16-NEXT: s_lshl_b32 s0, s0, 16
-; GFX11-TRUE16-NEXT: v_add_f32_e64 v6, 0x40c00000, s76
-; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_2)
+; GFX11-TRUE16-NEXT: s_lshr_b32 s4, s4, 16
+; GFX11-TRUE16-NEXT: v_add_f32_e64 v6, 0x40c00000, s73
; GFX11-TRUE16-NEXT: v_cndmask_b32_e32 v2, v3, v5, vcc_lo
; GFX11-TRUE16-NEXT: v_add_f32_e64 v3, 0x40c00000, s0
; GFX11-TRUE16-NEXT: v_add_nc_u32_e32 v1, 0x7fff, v1
; GFX11-TRUE16-NEXT: v_cmp_u_f32_e32 vcc_lo, v7, v7
; GFX11-TRUE16-NEXT: v_and_b32_e32 v4, 0xffff, v21
-; GFX11-TRUE16-NEXT: v_add_f32_e64 v5, 0x40c00000, s5
+; GFX11-TRUE16-NEXT: v_add_f32_e64 v5, 0x40c00000, s74
; GFX11-TRUE16-NEXT: v_readfirstlane_b32 s0, v3
; GFX11-TRUE16-NEXT: v_lshrrev_b32_e32 v7, 16, v2
; GFX11-TRUE16-NEXT: v_cndmask_b32_e32 v1, v1, v8, vcc_lo
; GFX11-TRUE16-NEXT: v_cmp_u_f32_e32 vcc_lo, v3, v3
; GFX11-TRUE16-NEXT: v_bfe_u32 v8, v6, 16, 1
-; GFX11-TRUE16-NEXT: s_bfe_u32 s5, s0, 0x10010
+; GFX11-TRUE16-NEXT: s_bfe_u32 s58, s0, 0x10010
; GFX11-TRUE16-NEXT: v_lshl_or_b32 v2, v20, 16, v4
-; GFX11-TRUE16-NEXT: s_add_i32 s41, s5, s0
-; GFX11-TRUE16-NEXT: s_lshr_b32 s5, s40, 16
-; GFX11-TRUE16-NEXT: s_addk_i32 s41, 0x7fff
+; GFX11-TRUE16-NEXT: s_add_i32 s58, s58, s0
; GFX11-TRUE16-NEXT: s_bitset1_b32 s0, 22
-; GFX11-TRUE16-NEXT: s_and_b32 s40, vcc_lo, exec_lo
-; GFX11-TRUE16-NEXT: s_cselect_b32 s0, s0, s41
-; GFX11-TRUE16-NEXT: s_and_b32 s40, s3, 0xffff0000
+; GFX11-TRUE16-NEXT: s_addk_i32 s58, 0x7fff
+; GFX11-TRUE16-NEXT: s_and_b32 s59, vcc_lo, exec_lo
+; GFX11-TRUE16-NEXT: s_cselect_b32 s0, s0, s58
+; GFX11-TRUE16-NEXT: s_and_b32 s58, s3, 0xffff0000
; GFX11-TRUE16-NEXT: v_lshrrev_b32_e32 v23, 16, v1
-; GFX11-TRUE16-NEXT: v_add_f32_e64 v9, 0x40c00000, s40
+; GFX11-TRUE16-NEXT: v_add_f32_e64 v9, 0x40c00000, s58
; GFX11-TRUE16-NEXT: v_bfe_u32 v1, v5, 16, 1
; GFX11-TRUE16-NEXT: v_add_nc_u32_e32 v4, v8, v6
; GFX11-TRUE16-NEXT: s_lshr_b32 s0, s0, 16
; GFX11-TRUE16-NEXT: v_and_b32_e32 v3, 0xffff, v23
-; GFX11-TRUE16-NEXT: v_readfirstlane_b32 s40, v9
+; GFX11-TRUE16-NEXT: v_readfirstlane_b32 s58, v9
; GFX11-TRUE16-NEXT: v_cmp_u_f32_e32 vcc_lo, v9, v9
; GFX11-TRUE16-NEXT: v_add_nc_u32_e32 v10, v1, v5
; GFX11-TRUE16-NEXT: v_add_nc_u32_e32 v4, 0x7fff, v4
; GFX11-TRUE16-NEXT: v_lshl_or_b32 v1, v7, 16, v3
-; GFX11-TRUE16-NEXT: s_bfe_u32 s41, s40, 0x10010
+; GFX11-TRUE16-NEXT: s_bfe_u32 s59, s58, 0x10010
; GFX11-TRUE16-NEXT: v_or_b32_e32 v8, 0x400000, v6
-; GFX11-TRUE16-NEXT: s_add_i32 s41, s41, s40
-; GFX11-TRUE16-NEXT: s_bitset1_b32 s40, 22
-; GFX11-TRUE16-NEXT: s_addk_i32 s41, 0x7fff
-; GFX11-TRUE16-NEXT: s_and_b32 s42, vcc_lo, exec_lo
-; GFX11-TRUE16-NEXT: s_cselect_b32 s40, s40, s41
+; GFX11-TRUE16-NEXT: s_add_i32 s59, s59, s58
+; GFX11-TRUE16-NEXT: s_bitset1_b32 s58, 22
+; GFX11-TRUE16-NEXT: s_addk_i32 s59, 0x7fff
+; GFX11-TRUE16-NEXT: s_and_b32 s62, vcc_lo, exec_lo
+; GFX11-TRUE16-NEXT: s_cselect_b32 s58, s58, s59
; GFX11-TRUE16-NEXT: s_lshl_b32 s3, s3, 16
; GFX11-TRUE16-NEXT: v_add_nc_u32_e32 v3, 0x7fff, v10
; GFX11-TRUE16-NEXT: v_add_f32_e64 v10, 0x40c00000, s3
; GFX11-TRUE16-NEXT: v_cmp_u_f32_e32 vcc_lo, v6, v6
; GFX11-TRUE16-NEXT: v_or_b32_e32 v7, 0x400000, v5
-; GFX11-TRUE16-NEXT: v_add_f32_e64 v9, 0x40c00000, s75
-; GFX11-TRUE16-NEXT: s_lshr_b32 s59, s40, 16
+; GFX11-TRUE16-NEXT: v_add_f32_e64 v9, 0x40c00000, s72
+; GFX11-TRUE16-NEXT: s_lshr_b32 s62, s58, 16
; GFX11-TRUE16-NEXT: v_readfirstlane_b32 s3, v10
; GFX11-TRUE16-NEXT: v_cndmask_b32_e32 v4, v4, v8, vcc_lo
; GFX11-TRUE16-NEXT: v_cmp_u_f32_e32 vcc_lo, v5, v5
; GFX11-TRUE16-NEXT: v_bfe_u32 v8, v9, 16, 1
-; GFX11-TRUE16-NEXT: v_add_f32_e64 v6, 0x40c00000, s74
-; GFX11-TRUE16-NEXT: s_bfe_u32 s41, s3, 0x10010
+; GFX11-TRUE16-NEXT: v_add_f32_e64 v6, 0x40c00000, s61
+; GFX11-TRUE16-NEXT: s_bfe_u32 s59, s3, 0x10010
; GFX11-TRUE16-NEXT: v_lshrrev_b32_e32 v22, 16, v4
; GFX11-TRUE16-NEXT: v_cndmask_b32_e32 v3, v3, v7, vcc_lo
; GFX11-TRUE16-NEXT: v_cmp_u_f32_e32 vcc_lo, v10, v10
-; GFX11-TRUE16-NEXT: s_add_i32 s41, s41, s3
+; GFX11-TRUE16-NEXT: s_add_i32 s59, s59, s3
; GFX11-TRUE16-NEXT: s_bitset1_b32 s3, 22
-; GFX11-TRUE16-NEXT: s_addk_i32 s41, 0x7fff
+; GFX11-TRUE16-NEXT: s_addk_i32 s59, 0x7fff
; GFX11-TRUE16-NEXT: v_add_nc_u32_e32 v5, v8, v9
-; GFX11-TRUE16-NEXT: s_and_b32 s40, vcc_lo, exec_lo
-; GFX11-TRUE16-NEXT: s_cselect_b32 s3, s3, s41
-; GFX11-TRUE16-NEXT: s_and_b32 s40, s2, 0xffff0000
+; GFX11-TRUE16-NEXT: s_and_b32 s58, vcc_lo, exec_lo
+; GFX11-TRUE16-NEXT: s_cselect_b32 s3, s3, s59
+; GFX11-TRUE16-NEXT: s_and_b32 s58, s2, 0xffff0000
; GFX11-TRUE16-NEXT: v_lshrrev_b32_e32 v24, 16, v3
-; GFX11-TRUE16-NEXT: v_add_f32_e64 v3, 0x40c00000, s40
+; GFX11-TRUE16-NEXT: v_add_f32_e64 v3, 0x40c00000, s58
; GFX11-TRUE16-NEXT: v_add_nc_u32_e32 v5, 0x7fff, v5
; GFX11-TRUE16-NEXT: v_or_b32_e32 v8, 0x400000, v9
; GFX11-TRUE16-NEXT: v_cmp_u_f32_e32 vcc_lo, v9, v9
; GFX11-TRUE16-NEXT: v_bfe_u32 v4, v6, 16, 1
-; GFX11-TRUE16-NEXT: v_readfirstlane_b32 s40, v3
+; GFX11-TRUE16-NEXT: v_readfirstlane_b32 s58, v3
; GFX11-TRUE16-NEXT: s_lshr_b32 s3, s3, 16
; GFX11-TRUE16-NEXT: v_or_b32_e32 v10, 0x400000, v6
; GFX11-TRUE16-NEXT: v_cndmask_b32_e32 v5, v5, v8, vcc_lo
; GFX11-TRUE16-NEXT: v_cmp_u_f32_e32 vcc_lo, v3, v3
-; GFX11-TRUE16-NEXT: s_bfe_u32 s41, s40, 0x10010
+; GFX11-TRUE16-NEXT: s_bfe_u32 s59, s58, 0x10010
; GFX11-TRUE16-NEXT: v_add_nc_u32_e32 v4, v4, v6
-; GFX11-TRUE16-NEXT: s_add_i32 s41, s41, s40
-; GFX11-TRUE16-NEXT: s_bitset1_b32 s40, 22
-; GFX11-TRUE16-NEXT: s_addk_i32 s41, 0x7fff
-; GFX11-TRUE16-NEXT: s_and_b32 s42, vcc_lo, exec_lo
-; GFX11-TRUE16-NEXT: s_cselect_b32 s40, s40, s41
+; GFX11-TRUE16-NEXT: s_add_i32 s59, s59, s58
+; GFX11-TRUE16-NEXT: s_bitset1_b32 s58, 22
+; GFX11-TRUE16-NEXT: s_addk_i32 s59, 0x7fff
+; GFX11-TRUE16-NEXT: s_and_b32 s61, vcc_lo, exec_lo
+; GFX11-TRUE16-NEXT: s_cselect_b32 s58, s58, s59
; GFX11-TRUE16-NEXT: s_lshl_b32 s2, s2, 16
; GFX11-TRUE16-NEXT: v_add_nc_u32_e32 v4, 0x7fff, v4
; GFX11-TRUE16-NEXT: v_add_f32_e64 v3, 0x40c00000, s2
; GFX11-TRUE16-NEXT: v_cmp_u_f32_e32 vcc_lo, v6, v6
-; GFX11-TRUE16-NEXT: v_add_f32_e64 v8, 0x40c00000, s73
+; GFX11-TRUE16-NEXT: v_add_f32_e64 v8, 0x40c00000, s60
; GFX11-TRUE16-NEXT: v_lshrrev_b32_e32 v27, 16, v5
; GFX11-TRUE16-NEXT: v_add_f32_e64 v5, 0x40c00000, s12
; GFX11-TRUE16-NEXT: v_readfirstlane_b32 s2, v3
@@ -175833,210 +175943,214 @@ define inreg <128 x i8> @bitcast_v64bf16_to_v128i8_scalar(<64 x bfloat> inreg %a
; GFX11-TRUE16-NEXT: v_bfe_u32 v6, v8, 16, 1
; GFX11-TRUE16-NEXT: s_bfe_u32 s12, s2, 0x10010
; GFX11-TRUE16-NEXT: v_lshrrev_b32_e32 v9, 16, v4
-; GFX11-TRUE16-NEXT: s_add_i32 s41, s12, s2
-; GFX11-TRUE16-NEXT: s_lshr_b32 s12, s40, 16
-; GFX11-TRUE16-NEXT: s_addk_i32 s41, 0x7fff
+; GFX11-TRUE16-NEXT: s_add_i32 s59, s12, s2
+; GFX11-TRUE16-NEXT: s_lshr_b32 s12, s58, 16
+; GFX11-TRUE16-NEXT: s_addk_i32 s59, 0x7fff
; GFX11-TRUE16-NEXT: s_bitset1_b32 s2, 22
-; GFX11-TRUE16-NEXT: s_and_b32 s40, vcc_lo, exec_lo
-; GFX11-TRUE16-NEXT: s_cselect_b32 s2, s2, s41
-; GFX11-TRUE16-NEXT: s_and_b32 s40, s17, 0xffff0000
+; GFX11-TRUE16-NEXT: s_and_b32 s58, vcc_lo, exec_lo
+; GFX11-TRUE16-NEXT: s_cselect_b32 s2, s2, s59
+; GFX11-TRUE16-NEXT: s_and_b32 s58, s17, 0xffff0000
; GFX11-TRUE16-NEXT: v_and_b32_e32 v10, 0xffff, v27
; GFX11-TRUE16-NEXT: v_add_nc_u32_e32 v6, v6, v8
; GFX11-TRUE16-NEXT: v_lshl_or_b32 v4, v24, 16, v7
-; GFX11-TRUE16-NEXT: v_add_f32_e64 v7, 0x40c00000, s40
+; GFX11-TRUE16-NEXT: v_add_f32_e64 v7, 0x40c00000, s58
; GFX11-TRUE16-NEXT: v_cmp_u_f32_e32 vcc_lo, v8, v8
; GFX11-TRUE16-NEXT: v_lshl_or_b32 v3, v9, 16, v10
; GFX11-TRUE16-NEXT: v_add_nc_u32_e32 v6, 0x7fff, v6
; GFX11-TRUE16-NEXT: v_or_b32_e32 v9, 0x400000, v8
-; GFX11-TRUE16-NEXT: v_readfirstlane_b32 s40, v7
+; GFX11-TRUE16-NEXT: v_readfirstlane_b32 s58, v7
; GFX11-TRUE16-NEXT: v_bfe_u32 v11, v5, 16, 1
; GFX11-TRUE16-NEXT: s_lshr_b32 s2, s2, 16
-; GFX11-TRUE16-NEXT: v_lshrrev_b32_e32 v85, 24, v4
+; GFX11-TRUE16-NEXT: s_pack_ll_b32_b16 s4, s0, s4
; GFX11-TRUE16-NEXT: v_cndmask_b32_e32 v6, v6, v9, vcc_lo
-; GFX11-TRUE16-NEXT: s_bfe_u32 s41, s40, 0x10010
+; GFX11-TRUE16-NEXT: s_bfe_u32 s59, s58, 0x10010
; GFX11-TRUE16-NEXT: v_cmp_u_f32_e32 vcc_lo, v7, v7
-; GFX11-TRUE16-NEXT: s_add_i32 s41, s41, s40
-; GFX11-TRUE16-NEXT: s_bitset1_b32 s40, 22
-; GFX11-TRUE16-NEXT: s_addk_i32 s41, 0x7fff
+; GFX11-TRUE16-NEXT: s_add_i32 s59, s59, s58
+; GFX11-TRUE16-NEXT: s_bitset1_b32 s58, 22
+; GFX11-TRUE16-NEXT: s_addk_i32 s59, 0x7fff
; GFX11-TRUE16-NEXT: v_add_nc_u32_e32 v11, v11, v5
-; GFX11-TRUE16-NEXT: s_and_b32 s42, vcc_lo, exec_lo
-; GFX11-TRUE16-NEXT: s_cselect_b32 s40, s40, s41
+; GFX11-TRUE16-NEXT: s_and_b32 s60, vcc_lo, exec_lo
+; GFX11-TRUE16-NEXT: s_cselect_b32 s58, s58, s59
; GFX11-TRUE16-NEXT: s_lshl_b32 s17, s17, 16
; GFX11-TRUE16-NEXT: v_cmp_u_f32_e32 vcc_lo, v5, v5
; GFX11-TRUE16-NEXT: v_add_f32_e64 v8, 0x40c00000, s17
; GFX11-TRUE16-NEXT: v_add_nc_u32_e32 v10, 0x7fff, v11
; GFX11-TRUE16-NEXT: v_or_b32_e32 v11, 0x400000, v5
-; GFX11-TRUE16-NEXT: v_add_f32_e64 v7, 0x40c00000, s72
+; GFX11-TRUE16-NEXT: v_add_f32_e64 v7, 0x40c00000, s57
; GFX11-TRUE16-NEXT: v_lshrrev_b32_e32 v26, 16, v6
; GFX11-TRUE16-NEXT: v_readfirstlane_b32 s17, v8
-; GFX11-TRUE16-NEXT: v_add_f32_e64 v6, 0x40c00000, s63
+; GFX11-TRUE16-NEXT: v_add_f32_e64 v6, 0x40c00000, s56
; GFX11-TRUE16-NEXT: v_cndmask_b32_e32 v5, v10, v11, vcc_lo
; GFX11-TRUE16-NEXT: v_cmp_u_f32_e32 vcc_lo, v8, v8
; GFX11-TRUE16-NEXT: v_bfe_u32 v9, v7, 16, 1
-; GFX11-TRUE16-NEXT: s_bfe_u32 s41, s17, 0x10010
+; GFX11-TRUE16-NEXT: s_bfe_u32 s56, s17, 0x10010
; GFX11-TRUE16-NEXT: v_bfe_u32 v10, v6, 16, 1
-; GFX11-TRUE16-NEXT: s_add_i32 s41, s41, s17
-; GFX11-TRUE16-NEXT: s_lshr_b32 s60, s40, 16
-; GFX11-TRUE16-NEXT: s_addk_i32 s41, 0x7fff
+; GFX11-TRUE16-NEXT: s_add_i32 s56, s56, s17
+; GFX11-TRUE16-NEXT: s_lshr_b32 s58, s58, 16
+; GFX11-TRUE16-NEXT: s_addk_i32 s56, 0x7fff
; GFX11-TRUE16-NEXT: s_bitset1_b32 s17, 22
-; GFX11-TRUE16-NEXT: s_and_b32 s40, vcc_lo, exec_lo
-; GFX11-TRUE16-NEXT: s_cselect_b32 s17, s17, s41
-; GFX11-TRUE16-NEXT: s_and_b32 s40, s16, 0xffff0000
+; GFX11-TRUE16-NEXT: s_and_b32 s57, vcc_lo, exec_lo
+; GFX11-TRUE16-NEXT: s_cselect_b32 s17, s17, s56
+; GFX11-TRUE16-NEXT: s_and_b32 s56, s16, 0xffff0000
; GFX11-TRUE16-NEXT: v_lshrrev_b32_e32 v25, 16, v5
; GFX11-TRUE16-NEXT: v_and_b32_e32 v5, 0xffff, v26
; GFX11-TRUE16-NEXT: v_add_nc_u32_e32 v9, v9, v7
; GFX11-TRUE16-NEXT: v_add_nc_u32_e32 v8, v10, v6
-; GFX11-TRUE16-NEXT: v_add_f32_e64 v10, 0x40c00000, s40
+; GFX11-TRUE16-NEXT: v_add_f32_e64 v10, 0x40c00000, s56
; GFX11-TRUE16-NEXT: v_cmp_u_f32_e32 vcc_lo, v7, v7
; GFX11-TRUE16-NEXT: v_lshl_or_b32 v15, v25, 16, v5
; GFX11-TRUE16-NEXT: v_add_nc_u32_e32 v5, 0x7fff, v9
; GFX11-TRUE16-NEXT: v_or_b32_e32 v9, 0x400000, v7
-; GFX11-TRUE16-NEXT: v_readfirstlane_b32 s40, v10
+; GFX11-TRUE16-NEXT: v_readfirstlane_b32 s56, v10
+; GFX11-TRUE16-NEXT: v_add_f32_e64 v7, 0x40c00000, s47
; GFX11-TRUE16-NEXT: s_lshr_b32 s17, s17, 16
-; GFX11-TRUE16-NEXT: v_add_nc_u32_e32 v8, 0x7fff, v8
-; GFX11-TRUE16-NEXT: v_or_b32_e32 v11, 0x400000, v6
-; GFX11-TRUE16-NEXT: v_cndmask_b32_e32 v5, v5, v9, vcc_lo
-; GFX11-TRUE16-NEXT: s_bfe_u32 s41, s40, 0x10010
+; GFX11-TRUE16-NEXT: v_dual_cndmask_b32 v5, v5, v9 :: v_dual_add_nc_u32 v8, 0x7fff, v8
+; GFX11-TRUE16-NEXT: s_bfe_u32 s47, s56, 0x10010
; GFX11-TRUE16-NEXT: v_cmp_u_f32_e32 vcc_lo, v10, v10
-; GFX11-TRUE16-NEXT: s_add_i32 s41, s41, s40
-; GFX11-TRUE16-NEXT: s_bitset1_b32 s40, 22
-; GFX11-TRUE16-NEXT: s_addk_i32 s41, 0x7fff
-; GFX11-TRUE16-NEXT: v_add_f32_e64 v7, 0x40c00000, s62
-; GFX11-TRUE16-NEXT: s_and_b32 s42, vcc_lo, exec_lo
-; GFX11-TRUE16-NEXT: s_cselect_b32 s40, s40, s41
+; GFX11-TRUE16-NEXT: s_add_i32 s47, s47, s56
+; GFX11-TRUE16-NEXT: s_bitset1_b32 s56, 22
+; GFX11-TRUE16-NEXT: s_addk_i32 s47, 0x7fff
+; GFX11-TRUE16-NEXT: v_or_b32_e32 v11, 0x400000, v6
+; GFX11-TRUE16-NEXT: s_and_b32 s57, vcc_lo, exec_lo
+; GFX11-TRUE16-NEXT: s_cselect_b32 s47, s56, s47
; GFX11-TRUE16-NEXT: s_lshl_b32 s16, s16, 16
; GFX11-TRUE16-NEXT: v_cmp_u_f32_e32 vcc_lo, v6, v6
; GFX11-TRUE16-NEXT: v_add_f32_e64 v10, 0x40c00000, s16
; GFX11-TRUE16-NEXT: v_bfe_u32 v9, v7, 16, 1
-; GFX11-TRUE16-NEXT: s_lshr_b32 s42, s40, 16
; GFX11-TRUE16-NEXT: v_lshrrev_b32_e32 v29, 16, v5
+; GFX11-TRUE16-NEXT: v_lshrrev_b32_e32 v83, 24, v15
; GFX11-TRUE16-NEXT: v_cndmask_b32_e32 v6, v8, v11, vcc_lo
; GFX11-TRUE16-NEXT: v_readfirstlane_b32 s16, v10
+; GFX11-TRUE16-NEXT: v_add_f32_e64 v8, 0x40c00000, s46
; GFX11-TRUE16-NEXT: v_cmp_u_f32_e32 vcc_lo, v10, v10
; GFX11-TRUE16-NEXT: v_add_nc_u32_e32 v5, v9, v7
-; GFX11-TRUE16-NEXT: v_or_b32_e32 v10, 0x400000, v7
-; GFX11-TRUE16-NEXT: v_add_f32_e64 v8, 0x40c00000, s57
-; GFX11-TRUE16-NEXT: s_bfe_u32 s41, s16, 0x10010
; GFX11-TRUE16-NEXT: v_lshrrev_b32_e32 v6, 16, v6
-; GFX11-TRUE16-NEXT: s_add_i32 s41, s41, s16
+; GFX11-TRUE16-NEXT: s_bfe_u32 s46, s16, 0x10010
+; GFX11-TRUE16-NEXT: v_and_b32_e32 v9, 0xffff, v29
+; GFX11-TRUE16-NEXT: s_add_i32 s56, s46, s16
+; GFX11-TRUE16-NEXT: s_lshr_b32 s46, s47, 16
+; GFX11-TRUE16-NEXT: s_addk_i32 s56, 0x7fff
; GFX11-TRUE16-NEXT: s_bitset1_b32 s16, 22
-; GFX11-TRUE16-NEXT: s_addk_i32 s41, 0x7fff
-; GFX11-TRUE16-NEXT: s_and_b32 s40, vcc_lo, exec_lo
-; GFX11-TRUE16-NEXT: s_cselect_b32 s16, s16, s41
-; GFX11-TRUE16-NEXT: s_and_b32 s40, s19, 0xffff0000
+; GFX11-TRUE16-NEXT: s_and_b32 s47, vcc_lo, exec_lo
+; GFX11-TRUE16-NEXT: s_cselect_b32 s16, s16, s56
+; GFX11-TRUE16-NEXT: s_and_b32 s47, s19, 0xffff0000
; GFX11-TRUE16-NEXT: v_add_nc_u32_e32 v5, 0x7fff, v5
-; GFX11-TRUE16-NEXT: v_add_f32_e64 v12, 0x40c00000, s40
+; GFX11-TRUE16-NEXT: v_add_f32_e64 v12, 0x40c00000, s47
+; GFX11-TRUE16-NEXT: v_or_b32_e32 v10, 0x400000, v7
; GFX11-TRUE16-NEXT: v_cmp_u_f32_e32 vcc_lo, v7, v7
-; GFX11-TRUE16-NEXT: v_and_b32_e32 v9, 0xffff, v29
+; GFX11-TRUE16-NEXT: v_lshl_or_b32 v14, v6, 16, v9
+; GFX11-TRUE16-NEXT: v_add_f32_e64 v9, 0x40c00000, s42
+; GFX11-TRUE16-NEXT: v_readfirstlane_b32 s47, v12
; GFX11-TRUE16-NEXT: v_bfe_u32 v11, v8, 16, 1
-; GFX11-TRUE16-NEXT: s_lshr_b32 s16, s16, 16
-; GFX11-TRUE16-NEXT: v_readfirstlane_b32 s40, v12
; GFX11-TRUE16-NEXT: v_cndmask_b32_e32 v5, v5, v10, vcc_lo
; GFX11-TRUE16-NEXT: v_cmp_u_f32_e32 vcc_lo, v12, v12
-; GFX11-TRUE16-NEXT: v_lshl_or_b32 v14, v6, 16, v9
-; GFX11-TRUE16-NEXT: v_add_f32_e64 v9, 0x40c00000, s47
-; GFX11-TRUE16-NEXT: s_bfe_u32 s41, s40, 0x10010
-; GFX11-TRUE16-NEXT: v_add_nc_u32_e32 v11, v11, v8
-; GFX11-TRUE16-NEXT: s_add_i32 s41, s41, s40
-; GFX11-TRUE16-NEXT: s_bitset1_b32 s40, 22
-; GFX11-TRUE16-NEXT: s_addk_i32 s41, 0x7fff
-; GFX11-TRUE16-NEXT: s_and_b32 s47, vcc_lo, exec_lo
-; GFX11-TRUE16-NEXT: s_cselect_b32 s40, s40, s41
+; GFX11-TRUE16-NEXT: s_lshr_b32 s16, s16, 16
+; GFX11-TRUE16-NEXT: s_bfe_u32 s42, s47, 0x10010
+; GFX11-TRUE16-NEXT: v_add_f32_e64 v10, 0x40c00000, s43
+; GFX11-TRUE16-NEXT: s_add_i32 s42, s42, s47
+; GFX11-TRUE16-NEXT: s_bitset1_b32 s47, 22
+; GFX11-TRUE16-NEXT: s_addk_i32 s42, 0x7fff
+; GFX11-TRUE16-NEXT: s_and_b32 s43, vcc_lo, exec_lo
+; GFX11-TRUE16-NEXT: s_cselect_b32 s42, s47, s42
; GFX11-TRUE16-NEXT: s_lshl_b32 s19, s19, 16
-; GFX11-TRUE16-NEXT: v_add_f32_e64 v10, 0x40c00000, s56
+; GFX11-TRUE16-NEXT: v_add_nc_u32_e32 v11, v11, v8
; GFX11-TRUE16-NEXT: v_add_f32_e64 v12, 0x40c00000, s19
-; GFX11-TRUE16-NEXT: v_add_nc_u32_e32 v6, 0x7fff, v11
; GFX11-TRUE16-NEXT: v_or_b32_e32 v7, 0x400000, v8
; GFX11-TRUE16-NEXT: v_cmp_u_f32_e32 vcc_lo, v8, v8
-; GFX11-TRUE16-NEXT: v_bfe_u32 v11, v10, 16, 1
+; GFX11-TRUE16-NEXT: s_lshr_b32 s59, s42, 16
+; GFX11-TRUE16-NEXT: v_add_nc_u32_e32 v6, 0x7fff, v11
; GFX11-TRUE16-NEXT: v_readfirstlane_b32 s19, v12
-; GFX11-TRUE16-NEXT: s_lshr_b32 s61, s40, 16
+; GFX11-TRUE16-NEXT: v_bfe_u32 v11, v10, 16, 1
; GFX11-TRUE16-NEXT: v_lshrrev_b32_e32 v28, 16, v5
+; GFX11-TRUE16-NEXT: v_lshrrev_b32_e32 v84, 16, v14
; GFX11-TRUE16-NEXT: v_cndmask_b32_e32 v6, v6, v7, vcc_lo
+; GFX11-TRUE16-NEXT: s_bfe_u32 s43, s19, 0x10010
; GFX11-TRUE16-NEXT: v_cmp_u_f32_e32 vcc_lo, v12, v12
-; GFX11-TRUE16-NEXT: s_bfe_u32 s41, s19, 0x10010
+; GFX11-TRUE16-NEXT: s_add_i32 s43, s43, s19
; GFX11-TRUE16-NEXT: v_bfe_u32 v7, v9, 16, 1
-; GFX11-TRUE16-NEXT: s_add_i32 s41, s41, s19
; GFX11-TRUE16-NEXT: v_add_nc_u32_e32 v8, v11, v10
-; GFX11-TRUE16-NEXT: s_addk_i32 s41, 0x7fff
+; GFX11-TRUE16-NEXT: s_addk_i32 s43, 0x7fff
; GFX11-TRUE16-NEXT: s_bitset1_b32 s19, 22
-; GFX11-TRUE16-NEXT: s_and_b32 s40, vcc_lo, exec_lo
-; GFX11-TRUE16-NEXT: s_cselect_b32 s19, s19, s41
-; GFX11-TRUE16-NEXT: s_and_b32 s40, s18, 0xffff0000
+; GFX11-TRUE16-NEXT: s_and_b32 s42, vcc_lo, exec_lo
+; GFX11-TRUE16-NEXT: s_cselect_b32 s19, s19, s43
+; GFX11-TRUE16-NEXT: s_and_b32 s42, s18, 0xffff0000
; GFX11-TRUE16-NEXT: v_lshrrev_b32_e32 v30, 16, v6
; GFX11-TRUE16-NEXT: v_add_nc_u32_e32 v5, v7, v9
; GFX11-TRUE16-NEXT: v_add_nc_u32_e32 v6, 0x7fff, v8
; GFX11-TRUE16-NEXT: v_or_b32_e32 v7, 0x400000, v10
-; GFX11-TRUE16-NEXT: v_add_f32_e64 v8, 0x40c00000, s40
+; GFX11-TRUE16-NEXT: v_add_f32_e64 v8, 0x40c00000, s42
; GFX11-TRUE16-NEXT: v_cmp_u_f32_e32 vcc_lo, v10, v10
; GFX11-TRUE16-NEXT: v_add_nc_u32_e32 v5, 0x7fff, v5
; GFX11-TRUE16-NEXT: v_or_b32_e32 v11, 0x400000, v9
; GFX11-TRUE16-NEXT: s_lshr_b32 s19, s19, 16
-; GFX11-TRUE16-NEXT: v_readfirstlane_b32 s40, v8
+; GFX11-TRUE16-NEXT: v_add_f32_e64 v12, 0x40c00000, s15
; GFX11-TRUE16-NEXT: v_cndmask_b32_e32 v6, v6, v7, vcc_lo
+; GFX11-TRUE16-NEXT: v_add_f32_e64 v7, 0x40c00000, s41
+; GFX11-TRUE16-NEXT: v_readfirstlane_b32 s41, v8
; GFX11-TRUE16-NEXT: v_cmp_u_f32_e32 vcc_lo, v9, v9
-; GFX11-TRUE16-NEXT: v_add_f32_e64 v7, 0x40c00000, s46
-; GFX11-TRUE16-NEXT: v_add_f32_e64 v12, 0x40c00000, s15
-; GFX11-TRUE16-NEXT: s_bfe_u32 s41, s40, 0x10010
+; GFX11-TRUE16-NEXT: v_lshrrev_b32_e32 v85, 24, v4
; GFX11-TRUE16-NEXT: v_lshrrev_b32_e32 v31, 16, v6
+; GFX11-TRUE16-NEXT: v_bfe_u32 v6, v7, 16, 1
+; GFX11-TRUE16-NEXT: s_bfe_u32 s42, s41, 0x10010
; GFX11-TRUE16-NEXT: v_cndmask_b32_e32 v5, v5, v11, vcc_lo
; GFX11-TRUE16-NEXT: v_cmp_u_f32_e32 vcc_lo, v8, v8
-; GFX11-TRUE16-NEXT: s_add_i32 s41, s41, s40
-; GFX11-TRUE16-NEXT: s_bitset1_b32 s40, 22
-; GFX11-TRUE16-NEXT: s_addk_i32 s41, 0x7fff
+; GFX11-TRUE16-NEXT: s_add_i32 s42, s42, s41
+; GFX11-TRUE16-NEXT: s_bitset1_b32 s41, 22
+; GFX11-TRUE16-NEXT: s_addk_i32 s42, 0x7fff
; GFX11-TRUE16-NEXT: v_add_f32_e64 v11, 0x40c00000, s28
-; GFX11-TRUE16-NEXT: s_and_b32 s46, vcc_lo, exec_lo
-; GFX11-TRUE16-NEXT: s_cselect_b32 s40, s40, s41
+; GFX11-TRUE16-NEXT: s_and_b32 s43, vcc_lo, exec_lo
+; GFX11-TRUE16-NEXT: s_cselect_b32 s41, s41, s42
; GFX11-TRUE16-NEXT: s_lshl_b32 s18, s18, 16
-; GFX11-TRUE16-NEXT: v_bfe_u32 v6, v7, 16, 1
-; GFX11-TRUE16-NEXT: v_add_f32_e64 v9, 0x40c00000, s18
-; GFX11-TRUE16-NEXT: s_lshr_b32 s46, s40, 16
; GFX11-TRUE16-NEXT: v_and_b32_e32 v8, 0xffff, v30
+; GFX11-TRUE16-NEXT: v_add_f32_e64 v9, 0x40c00000, s18
; GFX11-TRUE16-NEXT: v_lshrrev_b32_e32 v5, 16, v5
; GFX11-TRUE16-NEXT: v_and_b32_e32 v10, 0xffff, v31
-; GFX11-TRUE16-NEXT: v_readfirstlane_b32 s18, v9
-; GFX11-TRUE16-NEXT: v_cmp_u_f32_e32 vcc_lo, v9, v9
; GFX11-TRUE16-NEXT: v_add_nc_u32_e32 v6, v6, v7
; GFX11-TRUE16-NEXT: v_lshl_or_b32 v17, v28, 16, v8
+; GFX11-TRUE16-NEXT: v_readfirstlane_b32 s18, v9
+; GFX11-TRUE16-NEXT: v_cmp_u_f32_e32 vcc_lo, v9, v9
; GFX11-TRUE16-NEXT: v_lshl_or_b32 v16, v5, 16, v10
-; GFX11-TRUE16-NEXT: s_bfe_u32 s28, s18, 0x10010
-; GFX11-TRUE16-NEXT: v_or_b32_e32 v8, 0x400000, v7
-; GFX11-TRUE16-NEXT: s_add_i32 s28, s28, s18
-; GFX11-TRUE16-NEXT: s_bitset1_b32 s18, 22
-; GFX11-TRUE16-NEXT: s_addk_i32 s28, 0x7fff
-; GFX11-TRUE16-NEXT: s_and_b32 s40, vcc_lo, exec_lo
-; GFX11-TRUE16-NEXT: s_cselect_b32 s18, s18, s28
-; GFX11-TRUE16-NEXT: s_and_b32 s28, s21, 0xffff0000
; GFX11-TRUE16-NEXT: v_add_nc_u32_e32 v5, 0x7fff, v6
-; GFX11-TRUE16-NEXT: v_add_f32_e64 v9, 0x40c00000, s28
+; GFX11-TRUE16-NEXT: v_or_b32_e32 v8, 0x400000, v7
+; GFX11-TRUE16-NEXT: s_bfe_u32 s28, s18, 0x10010
; GFX11-TRUE16-NEXT: v_add_f32_e64 v10, 0x40c00000, s14
+; GFX11-TRUE16-NEXT: s_add_i32 s42, s28, s18
+; GFX11-TRUE16-NEXT: s_lshr_b32 s28, s41, 16
+; GFX11-TRUE16-NEXT: s_addk_i32 s42, 0x7fff
+; GFX11-TRUE16-NEXT: s_bitset1_b32 s18, 22
+; GFX11-TRUE16-NEXT: s_and_b32 s41, vcc_lo, exec_lo
+; GFX11-TRUE16-NEXT: s_cselect_b32 s18, s18, s42
+; GFX11-TRUE16-NEXT: s_and_b32 s41, s21, 0xffff0000
; GFX11-TRUE16-NEXT: v_cmp_u_f32_e32 vcc_lo, v7, v7
+; GFX11-TRUE16-NEXT: v_add_f32_e64 v9, 0x40c00000, s41
; GFX11-TRUE16-NEXT: v_bfe_u32 v6, v11, 16, 1
; GFX11-TRUE16-NEXT: s_lshr_b32 s18, s18, 16
-; GFX11-TRUE16-NEXT: v_readfirstlane_b32 s14, v9
; GFX11-TRUE16-NEXT: v_or_b32_e32 v7, 0x400000, v11
; GFX11-TRUE16-NEXT: v_cndmask_b32_e32 v5, v5, v8, vcc_lo
+; GFX11-TRUE16-NEXT: v_readfirstlane_b32 s14, v9
; GFX11-TRUE16-NEXT: v_cmp_u_f32_e32 vcc_lo, v9, v9
; GFX11-TRUE16-NEXT: v_add_nc_u32_e32 v6, v6, v11
-; GFX11-TRUE16-NEXT: s_bfe_u32 s15, s14, 0x10010
; GFX11-TRUE16-NEXT: v_bfe_u32 v8, v10, 16, 1
+; GFX11-TRUE16-NEXT: v_bfe_u32 v9, v12, 16, 1
+; GFX11-TRUE16-NEXT: s_bfe_u32 s15, s14, 0x10010
+; GFX11-TRUE16-NEXT: v_lshrrev_b32_e32 v32, 16, v5
; GFX11-TRUE16-NEXT: s_add_i32 s15, s15, s14
; GFX11-TRUE16-NEXT: s_bitset1_b32 s14, 22
; GFX11-TRUE16-NEXT: s_addk_i32 s15, 0x7fff
-; GFX11-TRUE16-NEXT: s_and_b32 s28, vcc_lo, exec_lo
+; GFX11-TRUE16-NEXT: s_and_b32 s41, vcc_lo, exec_lo
; GFX11-TRUE16-NEXT: s_cselect_b32 s14, s14, s15
; GFX11-TRUE16-NEXT: s_lshl_b32 s15, s21, 16
; GFX11-TRUE16-NEXT: v_add_nc_u32_e32 v6, 0x7fff, v6
; GFX11-TRUE16-NEXT: v_add_f32_e64 v13, 0x40c00000, s15
; GFX11-TRUE16-NEXT: v_cmp_u_f32_e32 vcc_lo, v11, v11
-; GFX11-TRUE16-NEXT: v_bfe_u32 v9, v12, 16, 1
-; GFX11-TRUE16-NEXT: s_lshr_b32 s62, s14, 16
-; GFX11-TRUE16-NEXT: v_lshrrev_b32_e32 v32, 16, v5
+; GFX11-TRUE16-NEXT: s_lshr_b32 s72, s14, 16
+; GFX11-TRUE16-NEXT: v_add_nc_u32_e32 v5, v9, v12
+; GFX11-TRUE16-NEXT: v_lshrrev_b32_e32 v81, 24, v17
; GFX11-TRUE16-NEXT: v_readfirstlane_b32 s15, v13
; GFX11-TRUE16-NEXT: v_cndmask_b32_e32 v6, v6, v7, vcc_lo
; GFX11-TRUE16-NEXT: v_cmp_u_f32_e32 vcc_lo, v13, v13
; GFX11-TRUE16-NEXT: v_add_nc_u32_e32 v7, v8, v10
-; GFX11-TRUE16-NEXT: v_add_nc_u32_e32 v5, v9, v12
+; GFX11-TRUE16-NEXT: v_add_nc_u32_e32 v5, 0x7fff, v5
; GFX11-TRUE16-NEXT: s_bfe_u32 s21, s15, 0x10010
; GFX11-TRUE16-NEXT: v_lshrrev_b32_e32 v33, 16, v6
; GFX11-TRUE16-NEXT: s_add_i32 s21, s21, s15
@@ -176049,20 +176163,20 @@ define inreg <128 x i8> @bitcast_v64bf16_to_v128i8_scalar(<64 x bfloat> inreg %a
; GFX11-TRUE16-NEXT: v_or_b32_e32 v7, 0x400000, v10
; GFX11-TRUE16-NEXT: v_add_f32_e64 v9, 0x40c00000, s15
; GFX11-TRUE16-NEXT: v_cmp_u_f32_e32 vcc_lo, v10, v10
-; GFX11-TRUE16-NEXT: v_add_nc_u32_e32 v5, 0x7fff, v5
; GFX11-TRUE16-NEXT: v_or_b32_e32 v8, 0x400000, v12
; GFX11-TRUE16-NEXT: s_lshr_b32 s21, s14, 16
-; GFX11-TRUE16-NEXT: s_pack_ll_b32_b16 s28, s0, s5
+; GFX11-TRUE16-NEXT: v_lshrrev_b32_e32 v82, 16, v16
+; GFX11-TRUE16-NEXT: v_lshrrev_b32_e32 v86, 8, v4
; GFX11-TRUE16-NEXT: v_cndmask_b32_e32 v6, v6, v7, vcc_lo
; GFX11-TRUE16-NEXT: v_cmp_u_f32_e32 vcc_lo, v12, v12
; GFX11-TRUE16-NEXT: v_add_f32_e64 v7, 0x40c00000, s11
; GFX11-TRUE16-NEXT: v_readfirstlane_b32 s11, v9
-; GFX11-TRUE16-NEXT: s_pack_ll_b32_b16 s40, s2, s12
+; GFX11-TRUE16-NEXT: v_lshrrev_b32_e32 v87, 16, v3
; GFX11-TRUE16-NEXT: v_lshrrev_b32_e32 v6, 16, v6
; GFX11-TRUE16-NEXT: v_cndmask_b32_e32 v5, v5, v8, vcc_lo
; GFX11-TRUE16-NEXT: v_cmp_u_f32_e32 vcc_lo, v9, v9
; GFX11-TRUE16-NEXT: s_bfe_u32 s15, s11, 0x10010
-; GFX11-TRUE16-NEXT: v_and_b32_e32 v8, 0xffff, v33
+; GFX11-TRUE16-NEXT: v_add_f32_e64 v9, 0x40c00000, s9
; GFX11-TRUE16-NEXT: s_add_i32 s15, s15, s11
; GFX11-TRUE16-NEXT: s_bitset1_b32 s11, 22
; GFX11-TRUE16-NEXT: s_addk_i32 s15, 0x7fff
@@ -176072,35 +176186,35 @@ define inreg <128 x i8> @bitcast_v64bf16_to_v128i8_scalar(<64 x bfloat> inreg %a
; GFX11-TRUE16-NEXT: v_lshrrev_b32_e32 v34, 16, v5
; GFX11-TRUE16-NEXT: v_add_f32_e64 v10, 0x40c00000, s14
; GFX11-TRUE16-NEXT: v_bfe_u32 v5, v7, 16, 1
-; GFX11-TRUE16-NEXT: v_add_f32_e64 v9, 0x40c00000, s9
-; GFX11-TRUE16-NEXT: v_lshl_or_b32 v19, v32, 16, v8
+; GFX11-TRUE16-NEXT: v_and_b32_e32 v8, 0xffff, v33
+; GFX11-TRUE16-NEXT: s_lshr_b32 s11, s11, 16
; GFX11-TRUE16-NEXT: v_and_b32_e32 v11, 0xffff, v34
-; GFX11-TRUE16-NEXT: v_readfirstlane_b32 s14, v10
+; GFX11-TRUE16-NEXT: v_readfirstlane_b32 s9, v10
; GFX11-TRUE16-NEXT: v_add_nc_u32_e32 v5, v5, v7
; GFX11-TRUE16-NEXT: v_cmp_u_f32_e32 vcc_lo, v10, v10
+; GFX11-TRUE16-NEXT: v_lshl_or_b32 v19, v32, 16, v8
; GFX11-TRUE16-NEXT: v_bfe_u32 v8, v9, 16, 1
+; GFX11-TRUE16-NEXT: s_bfe_u32 s14, s9, 0x10010
; GFX11-TRUE16-NEXT: v_lshl_or_b32 v18, v6, 16, v11
-; GFX11-TRUE16-NEXT: s_bfe_u32 s9, s14, 0x10010
+; GFX11-TRUE16-NEXT: s_add_i32 s14, s14, s9
+; GFX11-TRUE16-NEXT: s_bitset1_b32 s9, 22
+; GFX11-TRUE16-NEXT: s_addk_i32 s14, 0x7fff
; GFX11-TRUE16-NEXT: v_add_nc_u32_e32 v5, 0x7fff, v5
-; GFX11-TRUE16-NEXT: s_add_i32 s15, s9, s14
-; GFX11-TRUE16-NEXT: s_lshr_b32 s9, s11, 16
-; GFX11-TRUE16-NEXT: s_addk_i32 s15, 0x7fff
-; GFX11-TRUE16-NEXT: s_bitset1_b32 s14, 22
; GFX11-TRUE16-NEXT: v_or_b32_e32 v6, 0x400000, v7
-; GFX11-TRUE16-NEXT: s_and_b32 s11, vcc_lo, exec_lo
+; GFX11-TRUE16-NEXT: s_and_b32 s15, vcc_lo, exec_lo
; GFX11-TRUE16-NEXT: v_cmp_u_f32_e32 vcc_lo, v7, v7
-; GFX11-TRUE16-NEXT: s_cselect_b32 s11, s14, s15
+; GFX11-TRUE16-NEXT: s_cselect_b32 s9, s9, s14
; GFX11-TRUE16-NEXT: s_and_b32 s14, s23, 0xffff0000
; GFX11-TRUE16-NEXT: v_add_nc_u32_e32 v8, v8, v9
-; GFX11-TRUE16-NEXT: s_lshr_b32 s20, s11, 16
+; GFX11-TRUE16-NEXT: s_lshr_b32 s20, s9, 16
; GFX11-TRUE16-NEXT: v_cndmask_b32_e32 v5, v5, v6, vcc_lo
; GFX11-TRUE16-NEXT: v_add_f32_e64 v6, 0x40c00000, s14
; GFX11-TRUE16-NEXT: v_cmp_u_f32_e32 vcc_lo, v9, v9
; GFX11-TRUE16-NEXT: v_add_nc_u32_e32 v7, 0x7fff, v8
; GFX11-TRUE16-NEXT: v_or_b32_e32 v8, 0x400000, v9
-; GFX11-TRUE16-NEXT: v_add_f32_e64 v10, 0x40c00000, s45
+; GFX11-TRUE16-NEXT: v_add_f32_e64 v10, 0x40c00000, s44
; GFX11-TRUE16-NEXT: v_readfirstlane_b32 s14, v6
-; GFX11-TRUE16-NEXT: v_add_f32_e64 v11, 0x40c00000, s44
+; GFX11-TRUE16-NEXT: v_add_f32_e64 v11, 0x40c00000, s45
; GFX11-TRUE16-NEXT: v_lshrrev_b32_e32 v35, 16, v5
; GFX11-TRUE16-NEXT: v_cndmask_b32_e32 v7, v7, v8, vcc_lo
; GFX11-TRUE16-NEXT: v_cmp_u_f32_e32 vcc_lo, v6, v6
@@ -176109,14 +176223,14 @@ define inreg <128 x i8> @bitcast_v64bf16_to_v128i8_scalar(<64 x bfloat> inreg %a
; GFX11-TRUE16-NEXT: s_add_i32 s15, s15, s14
; GFX11-TRUE16-NEXT: s_bitset1_b32 s14, 22
; GFX11-TRUE16-NEXT: s_addk_i32 s15, 0x7fff
-; GFX11-TRUE16-NEXT: s_and_b32 s11, vcc_lo, exec_lo
-; GFX11-TRUE16-NEXT: s_cselect_b32 s11, s14, s15
+; GFX11-TRUE16-NEXT: s_and_b32 s9, vcc_lo, exec_lo
+; GFX11-TRUE16-NEXT: s_cselect_b32 s9, s14, s15
; GFX11-TRUE16-NEXT: s_lshl_b32 s14, s23, 16
; GFX11-TRUE16-NEXT: v_bfe_u32 v9, v11, 16, 1
; GFX11-TRUE16-NEXT: v_add_f32_e64 v6, 0x40c00000, s14
; GFX11-TRUE16-NEXT: v_lshrrev_b32_e32 v36, 16, v7
; GFX11-TRUE16-NEXT: v_add_nc_u32_e32 v7, v8, v10
-; GFX11-TRUE16-NEXT: s_lshr_b32 s63, s11, 16
+; GFX11-TRUE16-NEXT: s_lshr_b32 s44, s9, 16
; GFX11-TRUE16-NEXT: v_add_nc_u32_e32 v5, v9, v11
; GFX11-TRUE16-NEXT: v_readfirstlane_b32 s14, v6
; GFX11-TRUE16-NEXT: v_cmp_u_f32_e32 vcc_lo, v6, v6
@@ -176128,12 +176242,12 @@ define inreg <128 x i8> @bitcast_v64bf16_to_v128i8_scalar(<64 x bfloat> inreg %a
; GFX11-TRUE16-NEXT: s_add_i32 s15, s15, s14
; GFX11-TRUE16-NEXT: s_bitset1_b32 s14, 22
; GFX11-TRUE16-NEXT: s_addk_i32 s15, 0x7fff
-; GFX11-TRUE16-NEXT: s_and_b32 s11, vcc_lo, exec_lo
+; GFX11-TRUE16-NEXT: s_and_b32 s9, vcc_lo, exec_lo
; GFX11-TRUE16-NEXT: v_cmp_u_f32_e32 vcc_lo, v10, v10
-; GFX11-TRUE16-NEXT: s_cselect_b32 s11, s14, s15
+; GFX11-TRUE16-NEXT: s_cselect_b32 s9, s14, s15
; GFX11-TRUE16-NEXT: s_and_b32 s14, s22, 0xffff0000
-; GFX11-TRUE16-NEXT: s_lshr_b32 s23, s11, 16
-; GFX11-TRUE16-NEXT: v_add_f32_e64 v10, 0x40c00000, s43
+; GFX11-TRUE16-NEXT: s_lshr_b32 s23, s9, 16
+; GFX11-TRUE16-NEXT: v_add_f32_e64 v10, 0x40c00000, s40
; GFX11-TRUE16-NEXT: v_cndmask_b32_e32 v6, v7, v9, vcc_lo
; GFX11-TRUE16-NEXT: v_add_f32_e64 v7, 0x40c00000, s14
; GFX11-TRUE16-NEXT: v_cmp_u_f32_e32 vcc_lo, v11, v11
@@ -176150,15 +176264,15 @@ define inreg <128 x i8> @bitcast_v64bf16_to_v128i8_scalar(<64 x bfloat> inreg %a
; GFX11-TRUE16-NEXT: s_add_i32 s15, s15, s14
; GFX11-TRUE16-NEXT: s_bitset1_b32 s14, 22
; GFX11-TRUE16-NEXT: s_addk_i32 s15, 0x7fff
-; GFX11-TRUE16-NEXT: s_and_b32 s11, vcc_lo, exec_lo
-; GFX11-TRUE16-NEXT: s_cselect_b32 s11, s14, s15
+; GFX11-TRUE16-NEXT: s_and_b32 s9, vcc_lo, exec_lo
+; GFX11-TRUE16-NEXT: s_cselect_b32 s9, s14, s15
; GFX11-TRUE16-NEXT: s_lshl_b32 s14, s22, 16
; GFX11-TRUE16-NEXT: v_lshrrev_b32_e32 v6, 16, v6
; GFX11-TRUE16-NEXT: v_add_f32_e64 v12, 0x40c00000, s14
; GFX11-TRUE16-NEXT: v_bfe_u32 v5, v9, 16, 1
; GFX11-TRUE16-NEXT: v_and_b32_e32 v7, 0xffff, v39
; GFX11-TRUE16-NEXT: v_add_nc_u32_e32 v11, v11, v10
-; GFX11-TRUE16-NEXT: s_lshr_b32 s11, s11, 16
+; GFX11-TRUE16-NEXT: s_lshr_b32 s29, s9, 16
; GFX11-TRUE16-NEXT: v_readfirstlane_b32 s14, v12
; GFX11-TRUE16-NEXT: v_cmp_u_f32_e32 vcc_lo, v12, v12
; GFX11-TRUE16-NEXT: v_add_nc_u32_e32 v5, v5, v9
@@ -176169,34 +176283,34 @@ define inreg <128 x i8> @bitcast_v64bf16_to_v128i8_scalar(<64 x bfloat> inreg %a
; GFX11-TRUE16-NEXT: s_add_i32 s13, s13, s14
; GFX11-TRUE16-NEXT: s_bitset1_b32 s14, 22
; GFX11-TRUE16-NEXT: s_addk_i32 s13, 0x7fff
-; GFX11-TRUE16-NEXT: s_and_b32 s15, vcc_lo, exec_lo
-; GFX11-TRUE16-NEXT: s_cselect_b32 s13, s14, s13
-; GFX11-TRUE16-NEXT: s_and_b32 s14, s25, 0xffff0000
+; GFX11-TRUE16-NEXT: s_and_b32 s9, vcc_lo, exec_lo
+; GFX11-TRUE16-NEXT: s_cselect_b32 s9, s14, s13
+; GFX11-TRUE16-NEXT: s_and_b32 s13, s25, 0xffff0000
; GFX11-TRUE16-NEXT: v_cmp_u_f32_e32 vcc_lo, v10, v10
-; GFX11-TRUE16-NEXT: v_add_f32_e64 v12, 0x40c00000, s14
+; GFX11-TRUE16-NEXT: v_add_f32_e64 v12, 0x40c00000, s13
; GFX11-TRUE16-NEXT: v_add_nc_u32_e32 v5, 0x7fff, v5
; GFX11-TRUE16-NEXT: v_or_b32_e32 v11, 0x400000, v9
; GFX11-TRUE16-NEXT: v_add_f32_e64 v10, 0x40c00000, s10
; GFX11-TRUE16-NEXT: v_cndmask_b32_e32 v6, v6, v7, vcc_lo
; GFX11-TRUE16-NEXT: v_readfirstlane_b32 s10, v12
; GFX11-TRUE16-NEXT: v_cmp_u_f32_e32 vcc_lo, v9, v9
-; GFX11-TRUE16-NEXT: s_lshr_b32 s22, s13, 16
+; GFX11-TRUE16-NEXT: s_lshr_b32 s22, s9, 16
; GFX11-TRUE16-NEXT: v_bfe_u32 v7, v8, 16, 1
; GFX11-TRUE16-NEXT: v_lshrrev_b32_e32 v37, 16, v6
-; GFX11-TRUE16-NEXT: s_bfe_u32 s14, s10, 0x10010
+; GFX11-TRUE16-NEXT: s_bfe_u32 s13, s10, 0x10010
; GFX11-TRUE16-NEXT: v_cndmask_b32_e32 v5, v5, v11, vcc_lo
; GFX11-TRUE16-NEXT: v_cmp_u_f32_e32 vcc_lo, v12, v12
-; GFX11-TRUE16-NEXT: s_add_i32 s14, s14, s10
+; GFX11-TRUE16-NEXT: s_add_i32 s13, s13, s10
; GFX11-TRUE16-NEXT: s_bitset1_b32 s10, 22
-; GFX11-TRUE16-NEXT: s_addk_i32 s14, 0x7fff
+; GFX11-TRUE16-NEXT: s_addk_i32 s13, 0x7fff
; GFX11-TRUE16-NEXT: v_lshrrev_b32_e32 v38, 16, v5
-; GFX11-TRUE16-NEXT: s_and_b32 s13, vcc_lo, exec_lo
-; GFX11-TRUE16-NEXT: s_cselect_b32 s10, s10, s14
-; GFX11-TRUE16-NEXT: s_lshl_b32 s13, s25, 16
+; GFX11-TRUE16-NEXT: s_and_b32 s9, vcc_lo, exec_lo
+; GFX11-TRUE16-NEXT: s_cselect_b32 s9, s10, s13
+; GFX11-TRUE16-NEXT: s_lshl_b32 s10, s25, 16
; GFX11-TRUE16-NEXT: v_add_f32_e64 v11, 0x40c00000, s8
-; GFX11-TRUE16-NEXT: v_add_f32_e64 v5, 0x40c00000, s13
+; GFX11-TRUE16-NEXT: v_add_f32_e64 v5, 0x40c00000, s10
; GFX11-TRUE16-NEXT: v_add_nc_u32_e32 v6, v7, v8
-; GFX11-TRUE16-NEXT: s_lshr_b32 s72, s10, 16
+; GFX11-TRUE16-NEXT: s_lshr_b32 s45, s9, 16
; GFX11-TRUE16-NEXT: v_bfe_u32 v7, v10, 16, 1
; GFX11-TRUE16-NEXT: v_or_b32_e32 v12, 0x400000, v8
; GFX11-TRUE16-NEXT: v_readfirstlane_b32 s8, v5
@@ -176204,19 +176318,19 @@ define inreg <128 x i8> @bitcast_v64bf16_to_v128i8_scalar(<64 x bfloat> inreg %a
; GFX11-TRUE16-NEXT: v_add_nc_u32_e32 v6, 0x7fff, v6
; GFX11-TRUE16-NEXT: v_add_nc_u32_e32 v7, v7, v10
; GFX11-TRUE16-NEXT: v_or_b32_e32 v13, 0x400000, v10
-; GFX11-TRUE16-NEXT: s_bfe_u32 s13, s8, 0x10010
+; GFX11-TRUE16-NEXT: s_bfe_u32 s10, s8, 0x10010
; GFX11-TRUE16-NEXT: v_bfe_u32 v48, v11, 16, 1
-; GFX11-TRUE16-NEXT: s_add_i32 s13, s13, s8
+; GFX11-TRUE16-NEXT: s_add_i32 s10, s10, s8
; GFX11-TRUE16-NEXT: s_bitset1_b32 s8, 22
-; GFX11-TRUE16-NEXT: s_addk_i32 s13, 0x7fff
-; GFX11-TRUE16-NEXT: s_and_b32 s10, vcc_lo, exec_lo
+; GFX11-TRUE16-NEXT: s_addk_i32 s10, 0x7fff
+; GFX11-TRUE16-NEXT: s_and_b32 s9, vcc_lo, exec_lo
; GFX11-TRUE16-NEXT: v_cmp_u_f32_e32 vcc_lo, v8, v8
-; GFX11-TRUE16-NEXT: s_cselect_b32 s8, s8, s13
-; GFX11-TRUE16-NEXT: s_and_b32 s10, s24, 0xffff0000
+; GFX11-TRUE16-NEXT: s_cselect_b32 s8, s8, s10
+; GFX11-TRUE16-NEXT: s_and_b32 s9, s24, 0xffff0000
; GFX11-TRUE16-NEXT: v_add_nc_u32_e32 v7, 0x7fff, v7
; GFX11-TRUE16-NEXT: s_lshr_b32 s25, s8, 16
; GFX11-TRUE16-NEXT: v_cndmask_b32_e32 v5, v6, v12, vcc_lo
-; GFX11-TRUE16-NEXT: v_add_f32_e64 v6, 0x40c00000, s10
+; GFX11-TRUE16-NEXT: v_add_f32_e64 v6, 0x40c00000, s9
; GFX11-TRUE16-NEXT: v_cmp_u_f32_e32 vcc_lo, v10, v10
; GFX11-TRUE16-NEXT: v_add_f32_e64 v10, 0x40c00000, s7
; GFX11-TRUE16-NEXT: v_add_nc_u32_e32 v8, v48, v11
@@ -176226,39 +176340,39 @@ define inreg <128 x i8> @bitcast_v64bf16_to_v128i8_scalar(<64 x bfloat> inreg %a
; GFX11-TRUE16-NEXT: v_cmp_u_f32_e32 vcc_lo, v6, v6
; GFX11-TRUE16-NEXT: v_or_b32_e32 v6, 0x400000, v11
; GFX11-TRUE16-NEXT: v_and_b32_e32 v9, 0xffff, v37
-; GFX11-TRUE16-NEXT: s_bfe_u32 s10, s7, 0x10010
+; GFX11-TRUE16-NEXT: s_bfe_u32 s9, s7, 0x10010
; GFX11-TRUE16-NEXT: v_lshrrev_b32_e32 v50, 16, v7
-; GFX11-TRUE16-NEXT: s_add_i32 s10, s10, s7
+; GFX11-TRUE16-NEXT: s_add_i32 s9, s9, s7
; GFX11-TRUE16-NEXT: s_bitset1_b32 s7, 22
-; GFX11-TRUE16-NEXT: s_addk_i32 s10, 0x7fff
+; GFX11-TRUE16-NEXT: s_addk_i32 s9, 0x7fff
; GFX11-TRUE16-NEXT: s_and_b32 s8, vcc_lo, exec_lo
-; GFX11-TRUE16-NEXT: s_cselect_b32 s7, s7, s10
+; GFX11-TRUE16-NEXT: s_cselect_b32 s7, s7, s9
; GFX11-TRUE16-NEXT: s_lshl_b32 s8, s24, 16
; GFX11-TRUE16-NEXT: v_add_nc_u32_e32 v7, 0x7fff, v8
; GFX11-TRUE16-NEXT: v_add_f32_e64 v12, 0x40c00000, s8
; GFX11-TRUE16-NEXT: v_cmp_u_f32_e32 vcc_lo, v11, v11
-; GFX11-TRUE16-NEXT: v_add_f32_e64 v11, 0x40c00000, s4
+; GFX11-TRUE16-NEXT: v_add_f32_e64 v11, 0x40c00000, s5
; GFX11-TRUE16-NEXT: v_bfe_u32 v8, v10, 16, 1
-; GFX11-TRUE16-NEXT: s_lshr_b32 s10, s7, 16
+; GFX11-TRUE16-NEXT: s_lshr_b32 s42, s7, 16
; GFX11-TRUE16-NEXT: v_readfirstlane_b32 s8, v12
; GFX11-TRUE16-NEXT: v_cndmask_b32_e32 v6, v7, v6, vcc_lo
; GFX11-TRUE16-NEXT: v_cmp_u_f32_e32 vcc_lo, v12, v12
; GFX11-TRUE16-NEXT: v_add_nc_u32_e32 v8, v8, v10
; GFX11-TRUE16-NEXT: v_bfe_u32 v12, v51, 16, 1
-; GFX11-TRUE16-NEXT: s_bfe_u32 s4, s8, 0x10010
+; GFX11-TRUE16-NEXT: s_bfe_u32 s5, s8, 0x10010
; GFX11-TRUE16-NEXT: v_lshrrev_b32_e32 v48, 16, v6
-; GFX11-TRUE16-NEXT: s_add_i32 s4, s4, s8
+; GFX11-TRUE16-NEXT: s_add_i32 s5, s5, s8
; GFX11-TRUE16-NEXT: s_bitset1_b32 s8, 22
-; GFX11-TRUE16-NEXT: s_addk_i32 s4, 0x7fff
+; GFX11-TRUE16-NEXT: s_addk_i32 s5, 0x7fff
; GFX11-TRUE16-NEXT: s_and_b32 s6, vcc_lo, exec_lo
-; GFX11-TRUE16-NEXT: s_cselect_b32 s4, s8, s4
+; GFX11-TRUE16-NEXT: s_cselect_b32 s5, s8, s5
; GFX11-TRUE16-NEXT: s_and_b32 s6, s27, 0xffff0000
; GFX11-TRUE16-NEXT: v_add_nc_u32_e32 v7, 0x7fff, v8
; GFX11-TRUE16-NEXT: v_add_f32_e64 v52, 0x40c00000, s6
; GFX11-TRUE16-NEXT: v_or_b32_e32 v8, 0x400000, v10
; GFX11-TRUE16-NEXT: v_cmp_u_f32_e32 vcc_lo, v10, v10
; GFX11-TRUE16-NEXT: v_add_nc_u32_e32 v10, v12, v51
-; GFX11-TRUE16-NEXT: s_lshr_b32 s24, s4, 16
+; GFX11-TRUE16-NEXT: s_lshr_b32 s24, s5, 16
; GFX11-TRUE16-NEXT: v_readfirstlane_b32 s6, v52
; GFX11-TRUE16-NEXT: v_or_b32_e32 v12, 0x400000, v11
; GFX11-TRUE16-NEXT: v_cndmask_b32_e32 v7, v7, v8, vcc_lo
@@ -176269,8 +176383,8 @@ define inreg <128 x i8> @bitcast_v64bf16_to_v128i8_scalar(<64 x bfloat> inreg %a
; GFX11-TRUE16-NEXT: s_add_i32 s7, s7, s6
; GFX11-TRUE16-NEXT: s_bitset1_b32 s6, 22
; GFX11-TRUE16-NEXT: s_addk_i32 s7, 0x7fff
-; GFX11-TRUE16-NEXT: s_and_b32 s4, vcc_lo, exec_lo
-; GFX11-TRUE16-NEXT: s_cselect_b32 s4, s6, s7
+; GFX11-TRUE16-NEXT: s_and_b32 s5, vcc_lo, exec_lo
+; GFX11-TRUE16-NEXT: s_cselect_b32 s5, s6, s7
; GFX11-TRUE16-NEXT: s_lshl_b32 s6, s27, 16
; GFX11-TRUE16-NEXT: v_lshrrev_b32_e32 v49, 16, v7
; GFX11-TRUE16-NEXT: v_add_nc_u32_e32 v6, v8, v11
@@ -176279,7 +176393,7 @@ define inreg <128 x i8> @bitcast_v64bf16_to_v128i8_scalar(<64 x bfloat> inreg %a
; GFX11-TRUE16-NEXT: v_add_f32_e64 v10, 0x40c00000, s6
; GFX11-TRUE16-NEXT: v_cmp_u_f32_e32 vcc_lo, v51, v51
; GFX11-TRUE16-NEXT: v_add_nc_u32_e32 v6, 0x7fff, v6
-; GFX11-TRUE16-NEXT: s_lshr_b32 s73, s4, 16
+; GFX11-TRUE16-NEXT: s_lshr_b32 s73, s5, 16
; GFX11-TRUE16-NEXT: v_lshl_or_b32 v66, v5, 16, v13
; GFX11-TRUE16-NEXT: v_readfirstlane_b32 s6, v10
; GFX11-TRUE16-NEXT: v_cndmask_b32_e32 v7, v7, v8, vcc_lo
@@ -176294,52 +176408,56 @@ define inreg <128 x i8> @bitcast_v64bf16_to_v128i8_scalar(<64 x bfloat> inreg %a
; GFX11-TRUE16-NEXT: s_bitset1_b32 s6, 22
; GFX11-TRUE16-NEXT: s_addk_i32 s7, 0x7fff
; GFX11-TRUE16-NEXT: v_lshrrev_b32_e32 v6, 16, v6
-; GFX11-TRUE16-NEXT: s_and_b32 s4, vcc_lo, exec_lo
-; GFX11-TRUE16-NEXT: s_cselect_b32 s4, s6, s7
-; GFX11-TRUE16-NEXT: s_and_b32 s6, s26, 0xffff0000
-; GFX11-TRUE16-NEXT: s_lshr_b32 s27, s4, 16
-; GFX11-TRUE16-NEXT: v_add_f32_e64 v7, 0x40c00000, s6
+; GFX11-TRUE16-NEXT: s_and_b32 s5, vcc_lo, exec_lo
+; GFX11-TRUE16-NEXT: s_cselect_b32 s6, s6, s7
+; GFX11-TRUE16-NEXT: s_and_b32 s5, s26, 0xffff0000
+; GFX11-TRUE16-NEXT: s_lshr_b32 s27, s6, 16
+; GFX11-TRUE16-NEXT: v_add_f32_e64 v7, 0x40c00000, s5
; GFX11-TRUE16-NEXT: v_and_b32_e32 v10, 0xffff, v51
; GFX11-TRUE16-NEXT: v_lshl_or_b32 v55, v48, 16, v8
-; GFX11-TRUE16-NEXT: s_pack_ll_b32_b16 s8, s22, s11
-; GFX11-TRUE16-NEXT: s_pack_ll_b32_b16 s29, s1, s58
-; GFX11-TRUE16-NEXT: v_readfirstlane_b32 s6, v7
+; GFX11-TRUE16-NEXT: s_pack_ll_b32_b16 s10, s22, s29
+; GFX11-TRUE16-NEXT: s_pack_ll_b32_b16 s5, s1, s63
+; GFX11-TRUE16-NEXT: v_readfirstlane_b32 s8, v7
; GFX11-TRUE16-NEXT: v_cmp_u_f32_e32 vcc_lo, v7, v7
; GFX11-TRUE16-NEXT: v_lshl_or_b32 v54, v6, 16, v10
-; GFX11-TRUE16-NEXT: s_pack_ll_b32_b16 s41, s3, s59
-; GFX11-TRUE16-NEXT: s_pack_ll_b32_b16 s43, s17, s60
-; GFX11-TRUE16-NEXT: s_bfe_u32 s5, s6, 0x10010
-; GFX11-TRUE16-NEXT: s_pack_ll_b32_b16 s42, s16, s42
-; GFX11-TRUE16-NEXT: s_add_i32 s5, s5, s6
-; GFX11-TRUE16-NEXT: s_bitset1_b32 s6, 22
-; GFX11-TRUE16-NEXT: s_addk_i32 s5, 0x7fff
-; GFX11-TRUE16-NEXT: s_and_b32 s4, vcc_lo, exec_lo
-; GFX11-TRUE16-NEXT: s_cselect_b32 s13, s6, s5
-; GFX11-TRUE16-NEXT: s_lshl_b32 s4, s26, 16
-; GFX11-TRUE16-NEXT: s_lshr_b32 s13, s13, 16
-; GFX11-TRUE16-NEXT: v_add_f32_e64 v5, 0x40c00000, s4
-; GFX11-TRUE16-NEXT: s_pack_ll_b32_b16 s7, s21, s62
-; GFX11-TRUE16-NEXT: s_pack_ll_b32_b16 s6, s20, s9
-; GFX11-TRUE16-NEXT: s_pack_ll_b32_b16 s9, s23, s63
-; GFX11-TRUE16-NEXT: s_pack_ll_b32_b16 s45, s27, s73
-; GFX11-TRUE16-NEXT: v_readfirstlane_b32 s12, v5
+; GFX11-TRUE16-NEXT: s_pack_ll_b32_b16 s7, s3, s62
+; GFX11-TRUE16-NEXT: s_pack_ll_b32_b16 s15, s21, s72
+; GFX11-TRUE16-NEXT: s_bfe_u32 s9, s8, 0x10010
+; GFX11-TRUE16-NEXT: s_pack_ll_b32_b16 s14, s20, s11
+; GFX11-TRUE16-NEXT: s_add_i32 s9, s9, s8
+; GFX11-TRUE16-NEXT: s_bitset1_b32 s8, 22
+; GFX11-TRUE16-NEXT: s_addk_i32 s9, 0x7fff
+; GFX11-TRUE16-NEXT: s_and_b32 s6, vcc_lo, exec_lo
+; GFX11-TRUE16-NEXT: s_cselect_b32 s40, s8, s9
+; GFX11-TRUE16-NEXT: s_lshl_b32 s8, s26, 16
+; GFX11-TRUE16-NEXT: s_pack_ll_b32_b16 s6, s2, s12
+; GFX11-TRUE16-NEXT: v_add_f32_e64 v5, 0x40c00000, s8
+; GFX11-TRUE16-NEXT: s_pack_ll_b32_b16 s12, s18, s28
+; GFX11-TRUE16-NEXT: s_lshr_b32 s40, s40, 16
+; GFX11-TRUE16-NEXT: s_pack_ll_b32_b16 s9, s17, s58
+; GFX11-TRUE16-NEXT: s_pack_ll_b32_b16 s8, s16, s46
+; GFX11-TRUE16-NEXT: v_readfirstlane_b32 s26, v5
; GFX11-TRUE16-NEXT: v_cmp_u_f32_e32 vcc_lo, v5, v5
; GFX11-TRUE16-NEXT: v_lshrrev_b64 v[5:6], 24, v[54:55]
; GFX11-TRUE16-NEXT: v_lshrrev_b64 v[6:7], 24, v[66:67]
; GFX11-TRUE16-NEXT: v_lshrrev_b64 v[7:8], 24, v[70:71]
-; GFX11-TRUE16-NEXT: s_bfe_u32 s11, s12, 0x10010
+; GFX11-TRUE16-NEXT: s_bfe_u32 s28, s26, 0x10010
; GFX11-TRUE16-NEXT: v_lshrrev_b64 v[8:9], 24, v[18:19]
-; GFX11-TRUE16-NEXT: s_add_i32 s11, s11, s12
-; GFX11-TRUE16-NEXT: s_bitset1_b32 s12, 22
-; GFX11-TRUE16-NEXT: s_addk_i32 s11, 0x7fff
-; GFX11-TRUE16-NEXT: s_and_b32 s14, vcc_lo, exec_lo
-; GFX11-TRUE16-NEXT: s_cselect_b32 s12, s12, s11
+; GFX11-TRUE16-NEXT: s_add_i32 s28, s28, s26
+; GFX11-TRUE16-NEXT: s_bitset1_b32 s26, 22
+; GFX11-TRUE16-NEXT: s_addk_i32 s28, 0x7fff
+; GFX11-TRUE16-NEXT: s_and_b32 s29, vcc_lo, exec_lo
+; GFX11-TRUE16-NEXT: s_cselect_b32 s26, s26, s28
; GFX11-TRUE16-NEXT: v_lshrrev_b64 v[9:10], 24, v[16:17]
; GFX11-TRUE16-NEXT: v_lshrrev_b64 v[10:11], 24, v[14:15]
-; GFX11-TRUE16-NEXT: s_lshr_b32 s26, s12, 16
+; GFX11-TRUE16-NEXT: s_lshr_b32 s26, s26, 16
; GFX11-TRUE16-NEXT: v_lshrrev_b64 v[11:12], 24, v[3:4]
; GFX11-TRUE16-NEXT: v_lshrrev_b64 v[12:13], 24, v[1:2]
-; GFX11-TRUE16-NEXT: s_pack_ll_b32_b16 s44, s26, s13
+; GFX11-TRUE16-NEXT: s_pack_ll_b32_b16 s11, s23, s44
+; GFX11-TRUE16-NEXT: s_pack_ll_b32_b16 s29, s25, s45
+; GFX11-TRUE16-NEXT: s_pack_ll_b32_b16 s41, s27, s73
+; GFX11-TRUE16-NEXT: s_pack_ll_b32_b16 s40, s26, s40
+; GFX11-TRUE16-NEXT: s_pack_ll_b32_b16 s28, s24, s42
; GFX11-TRUE16-NEXT: v_lshrrev_b32_e32 v13, 24, v55
; GFX11-TRUE16-NEXT: v_lshrrev_b32_e32 v52, 8, v55
; GFX11-TRUE16-NEXT: v_lshrrev_b32_e32 v53, 16, v54
@@ -176356,116 +176474,107 @@ define inreg <128 x i8> @bitcast_v64bf16_to_v128i8_scalar(<64 x bfloat> inreg %a
; GFX11-TRUE16-NEXT: v_lshrrev_b32_e32 v19, 8, v19
; GFX11-TRUE16-NEXT: v_lshrrev_b32_e32 v80, 16, v18
; GFX11-TRUE16-NEXT: v_lshrrev_b32_e32 v18, 8, v18
-; GFX11-TRUE16-NEXT: v_lshrrev_b32_e32 v81, 24, v17
; GFX11-TRUE16-NEXT: v_lshrrev_b32_e32 v17, 8, v17
-; GFX11-TRUE16-NEXT: v_lshrrev_b32_e32 v82, 16, v16
; GFX11-TRUE16-NEXT: v_lshrrev_b32_e32 v16, 8, v16
-; GFX11-TRUE16-NEXT: v_lshrrev_b32_e32 v83, 24, v15
; GFX11-TRUE16-NEXT: v_lshrrev_b32_e32 v15, 8, v15
-; GFX11-TRUE16-NEXT: v_lshrrev_b32_e32 v84, 16, v14
; GFX11-TRUE16-NEXT: v_lshrrev_b32_e32 v14, 8, v14
-; GFX11-TRUE16-NEXT: v_lshrrev_b32_e32 v86, 8, v4
-; GFX11-TRUE16-NEXT: v_lshrrev_b32_e32 v87, 16, v3
; GFX11-TRUE16-NEXT: v_lshrrev_b32_e32 v96, 8, v3
; GFX11-TRUE16-NEXT: v_lshrrev_b32_e32 v97, 24, v2
; GFX11-TRUE16-NEXT: v_lshrrev_b32_e32 v98, 8, v2
; GFX11-TRUE16-NEXT: v_lshrrev_b32_e32 v99, 16, v1
; GFX11-TRUE16-NEXT: v_lshrrev_b32_e32 v100, 8, v1
-; GFX11-TRUE16-NEXT: s_pack_ll_b32_b16 s5, s19, s61
-; GFX11-TRUE16-NEXT: s_pack_ll_b32_b16 s4, s18, s46
-; GFX11-TRUE16-NEXT: s_pack_ll_b32_b16 s11, s25, s72
-; GFX11-TRUE16-NEXT: s_pack_ll_b32_b16 s10, s24, s10
-; GFX11-TRUE16-NEXT: s_lshr_b64 s[92:93], s[44:45], 24
-; GFX11-TRUE16-NEXT: s_lshr_b64 s[94:95], s[8:9], 24
-; GFX11-TRUE16-NEXT: s_lshr_b64 s[30:31], s[6:7], 24
-; GFX11-TRUE16-NEXT: s_lshr_b64 s[14:15], s[42:43], 24
-; GFX11-TRUE16-NEXT: s_lshr_b64 s[46:47], s[40:41], 24
-; GFX11-TRUE16-NEXT: s_lshr_b64 s[56:57], s[28:29], 24
-; GFX11-TRUE16-NEXT: s_lshr_b64 s[36:37], s[10:11], 24
-; GFX11-TRUE16-NEXT: s_lshr_b64 s[38:39], s[4:5], 24
-; GFX11-TRUE16-NEXT: s_lshr_b32 s13, s45, 24
-; GFX11-TRUE16-NEXT: s_lshr_b32 s15, s45, 8
-; GFX11-TRUE16-NEXT: s_lshr_b32 s45, s44, 16
-; GFX11-TRUE16-NEXT: s_lshr_b32 s44, s44, 8
-; GFX11-TRUE16-NEXT: s_lshr_b32 s47, s11, 24
+; GFX11-TRUE16-NEXT: s_pack_ll_b32_b16 s13, s19, s59
+; GFX11-TRUE16-NEXT: s_lshr_b64 s[88:89], s[40:41], 24
+; GFX11-TRUE16-NEXT: s_lshr_b64 s[94:95], s[28:29], 24
+; GFX11-TRUE16-NEXT: s_lshr_b64 s[90:91], s[10:11], 24
+; GFX11-TRUE16-NEXT: s_lshr_b64 s[92:93], s[14:15], 24
+; GFX11-TRUE16-NEXT: s_lshr_b64 s[46:47], s[8:9], 24
+; GFX11-TRUE16-NEXT: s_lshr_b64 s[56:57], s[6:7], 24
+; GFX11-TRUE16-NEXT: s_lshr_b64 s[60:61], s[4:5], 24
+; GFX11-TRUE16-NEXT: s_lshr_b64 s[30:31], s[12:13], 24
+; GFX11-TRUE16-NEXT: s_lshr_b32 s43, s41, 24
+; GFX11-TRUE16-NEXT: s_lshr_b32 s41, s41, 8
+; GFX11-TRUE16-NEXT: s_lshr_b32 s47, s40, 16
+; GFX11-TRUE16-NEXT: s_lshr_b32 s40, s40, 8
+; GFX11-TRUE16-NEXT: s_lshr_b32 s57, s29, 24
+; GFX11-TRUE16-NEXT: s_lshr_b32 s29, s29, 8
+; GFX11-TRUE16-NEXT: s_lshr_b32 s61, s28, 16
+; GFX11-TRUE16-NEXT: s_lshr_b32 s28, s28, 8
+; GFX11-TRUE16-NEXT: s_lshr_b32 s74, s11, 24
; GFX11-TRUE16-NEXT: s_lshr_b32 s11, s11, 8
-; GFX11-TRUE16-NEXT: s_lshr_b32 s57, s10, 16
+; GFX11-TRUE16-NEXT: s_lshr_b32 s75, s10, 16
; GFX11-TRUE16-NEXT: s_lshr_b32 s10, s10, 8
-; GFX11-TRUE16-NEXT: s_lshr_b32 s74, s9, 24
+; GFX11-TRUE16-NEXT: s_lshr_b32 s76, s15, 24
+; GFX11-TRUE16-NEXT: s_lshr_b32 s15, s15, 8
+; GFX11-TRUE16-NEXT: s_lshr_b32 s77, s14, 16
+; GFX11-TRUE16-NEXT: s_lshr_b32 s14, s14, 8
+; GFX11-TRUE16-NEXT: s_lshr_b32 s78, s13, 24
+; GFX11-TRUE16-NEXT: s_lshr_b32 s13, s13, 8
+; GFX11-TRUE16-NEXT: s_lshr_b32 s79, s12, 16
+; GFX11-TRUE16-NEXT: s_lshr_b32 s12, s12, 8
+; GFX11-TRUE16-NEXT: s_lshr_b32 s89, s9, 24
; GFX11-TRUE16-NEXT: s_lshr_b32 s9, s9, 8
-; GFX11-TRUE16-NEXT: s_lshr_b32 s75, s8, 16
+; GFX11-TRUE16-NEXT: s_lshr_b32 s91, s8, 16
; GFX11-TRUE16-NEXT: s_lshr_b32 s8, s8, 8
-; GFX11-TRUE16-NEXT: s_lshr_b32 s76, s7, 24
-; GFX11-TRUE16-NEXT: s_lshr_b32 s77, s7, 8
-; GFX11-TRUE16-NEXT: s_lshr_b32 s78, s6, 16
-; GFX11-TRUE16-NEXT: s_lshr_b32 s79, s6, 8
-; GFX11-TRUE16-NEXT: s_lshr_b32 s88, s5, 24
-; GFX11-TRUE16-NEXT: s_lshr_b32 s89, s5, 8
-; GFX11-TRUE16-NEXT: s_lshr_b32 s90, s4, 16
-; GFX11-TRUE16-NEXT: s_lshr_b32 s91, s4, 8
-; GFX11-TRUE16-NEXT: s_lshr_b32 s93, s43, 24
-; GFX11-TRUE16-NEXT: s_lshr_b32 s43, s43, 8
-; GFX11-TRUE16-NEXT: s_lshr_b32 s95, s42, 16
-; GFX11-TRUE16-NEXT: s_lshr_b32 s42, s42, 8
-; GFX11-TRUE16-NEXT: s_lshr_b32 s31, s41, 24
-; GFX11-TRUE16-NEXT: s_lshr_b32 s41, s41, 8
-; GFX11-TRUE16-NEXT: s_lshr_b32 s99, s40, 16
-; GFX11-TRUE16-NEXT: s_lshr_b32 s12, s40, 8
-; GFX11-TRUE16-NEXT: s_lshr_b32 s102, s29, 24
-; GFX11-TRUE16-NEXT: s_lshr_b32 s103, s29, 8
-; GFX11-TRUE16-NEXT: s_lshr_b32 s104, s28, 16
-; GFX11-TRUE16-NEXT: s_lshr_b32 s34, s28, 8
+; GFX11-TRUE16-NEXT: s_lshr_b32 s95, s7, 24
+; GFX11-TRUE16-NEXT: s_lshr_b32 s93, s7, 8
+; GFX11-TRUE16-NEXT: s_lshr_b32 s99, s6, 16
+; GFX11-TRUE16-NEXT: s_lshr_b32 s42, s6, 8
+; GFX11-TRUE16-NEXT: s_lshr_b32 s102, s5, 24
+; GFX11-TRUE16-NEXT: s_lshr_b32 s103, s5, 8
+; GFX11-TRUE16-NEXT: s_lshr_b32 s104, s4, 16
+; GFX11-TRUE16-NEXT: s_lshr_b32 s34, s4, 8
; GFX11-TRUE16-NEXT: s_branch .LBB91_5
; GFX11-TRUE16-NEXT: .LBB91_3:
-; GFX11-TRUE16-NEXT: ; implicit-def: $sgpr4_lo16
-; GFX11-TRUE16-NEXT: v_writelane_b32 v43, s4, 0
-; GFX11-TRUE16-NEXT: ; implicit-def: $sgpr4_lo16
+; GFX11-TRUE16-NEXT: ; implicit-def: $sgpr10_lo16
+; GFX11-TRUE16-NEXT: v_writelane_b32 v43, s10, 0
+; GFX11-TRUE16-NEXT: ; implicit-def: $sgpr10_lo16
; GFX11-TRUE16-NEXT: ; implicit-def: $sgpr74_lo16
; GFX11-TRUE16-NEXT: ; implicit-def: $sgpr34_lo16
; GFX11-TRUE16-NEXT: ; implicit-def: $sgpr104_lo16
-; GFX11-TRUE16-NEXT: ; implicit-def: $sgpr56_lo16
+; GFX11-TRUE16-NEXT: ; implicit-def: $sgpr60_lo16
; GFX11-TRUE16-NEXT: ; implicit-def: $sgpr103_lo16
; GFX11-TRUE16-NEXT: ; implicit-def: $sgpr102_lo16
-; GFX11-TRUE16-NEXT: ; implicit-def: $sgpr12_lo16
+; GFX11-TRUE16-NEXT: ; implicit-def: $sgpr42_lo16
; GFX11-TRUE16-NEXT: ; implicit-def: $sgpr99_lo16
+; GFX11-TRUE16-NEXT: ; implicit-def: $sgpr56_lo16
; GFX11-TRUE16-NEXT: ; implicit-def: $sgpr46_lo16
-; GFX11-TRUE16-NEXT: ; implicit-def: $sgpr14_lo16
; GFX11-TRUE16-NEXT: ; implicit-def: $sgpr86_lo16
-; GFX11-TRUE16-NEXT: ; implicit-def: $sgpr7_lo16
-; GFX11-TRUE16-NEXT: ; implicit-def: $sgpr68_lo16
+; GFX11-TRUE16-NEXT: ; implicit-def: $sgpr13_lo16
+; GFX11-TRUE16-NEXT: ; implicit-def: $sgpr67_lo16
; GFX11-TRUE16-NEXT: ; implicit-def: $sgpr98_lo16
-; GFX11-TRUE16-NEXT: ; implicit-def: $sgpr54_lo16
-; GFX11-TRUE16-NEXT: ; implicit-def: $sgpr6_lo16
-; GFX11-TRUE16-NEXT: ; implicit-def: $sgpr39_lo16
-; GFX11-TRUE16-NEXT: ; implicit-def: $sgpr97_lo16
; GFX11-TRUE16-NEXT: ; implicit-def: $sgpr53_lo16
-; GFX11-TRUE16-NEXT: ; implicit-def: $sgpr85_lo16
-; GFX11-TRUE16-NEXT: ; implicit-def: $sgpr67_lo16
-; GFX11-TRUE16-NEXT: ; implicit-def: $sgpr11_lo16
+; GFX11-TRUE16-NEXT: ; implicit-def: $sgpr12_lo16
+; GFX11-TRUE16-NEXT: ; implicit-def: $sgpr48_lo16
+; GFX11-TRUE16-NEXT: ; implicit-def: $sgpr97_lo16
; GFX11-TRUE16-NEXT: ; implicit-def: $sgpr52_lo16
+; GFX11-TRUE16-NEXT: ; implicit-def: $sgpr85_lo16
+; GFX11-TRUE16-NEXT: ; implicit-def: $sgpr66_lo16
+; GFX11-TRUE16-NEXT: ; implicit-def: $sgpr41_lo16
+; GFX11-TRUE16-NEXT: ; implicit-def: $sgpr51_lo16
; GFX11-TRUE16-NEXT: ; implicit-def: $sgpr84_lo16
; GFX11-TRUE16-NEXT: ; implicit-def: $sgpr83_lo16
-; GFX11-TRUE16-NEXT: ; implicit-def: $sgpr66_lo16
-; GFX11-TRUE16-NEXT: ; implicit-def: $sgpr10_lo16
-; GFX11-TRUE16-NEXT: ; implicit-def: $sgpr37_lo16
-; GFX11-TRUE16-NEXT: ; implicit-def: $sgpr49_lo16
-; GFX11-TRUE16-NEXT: ; implicit-def: $sgpr82_lo16
; GFX11-TRUE16-NEXT: ; implicit-def: $sgpr65_lo16
-; GFX11-TRUE16-NEXT: ; implicit-def: $sgpr9_lo16
-; GFX11-TRUE16-NEXT: ; implicit-def: $sgpr51_lo16
+; GFX11-TRUE16-NEXT: ; implicit-def: $sgpr40_lo16
+; GFX11-TRUE16-NEXT: ; implicit-def: $sgpr38_lo16
+; GFX11-TRUE16-NEXT: ; implicit-def: $sgpr82_lo16
; GFX11-TRUE16-NEXT: ; implicit-def: $sgpr81_lo16
-; GFX11-TRUE16-NEXT: ; implicit-def: $sgpr80_lo16
; GFX11-TRUE16-NEXT: ; implicit-def: $sgpr64_lo16
-; GFX11-TRUE16-NEXT: ; implicit-def: $sgpr8_lo16
+; GFX11-TRUE16-NEXT: ; implicit-def: $sgpr15_lo16
; GFX11-TRUE16-NEXT: ; implicit-def: $sgpr50_lo16
-; GFX11-TRUE16-NEXT: ; implicit-def: $sgpr48_lo16
+; GFX11-TRUE16-NEXT: ; implicit-def: $sgpr80_lo16
; GFX11-TRUE16-NEXT: ; implicit-def: $sgpr71_lo16
-; GFX11-TRUE16-NEXT: ; implicit-def: $sgpr38_lo16
+; GFX11-TRUE16-NEXT: ; implicit-def: $sgpr55_lo16
+; GFX11-TRUE16-NEXT: ; implicit-def: $sgpr14_lo16
+; GFX11-TRUE16-NEXT: ; implicit-def: $sgpr37_lo16
+; GFX11-TRUE16-NEXT: ; implicit-def: $sgpr49_lo16
+; GFX11-TRUE16-NEXT: ; implicit-def: $sgpr70_lo16
+; GFX11-TRUE16-NEXT: ; implicit-def: $sgpr39_lo16
; GFX11-TRUE16-NEXT: ; implicit-def: $sgpr96_lo16
; GFX11-TRUE16-NEXT: ; implicit-def: $sgpr36_lo16
-; GFX11-TRUE16-NEXT: ; implicit-def: $sgpr70_lo16
; GFX11-TRUE16-NEXT: ; implicit-def: $sgpr69_lo16
-; GFX11-TRUE16-NEXT: ; implicit-def: $sgpr55_lo16
+; GFX11-TRUE16-NEXT: ; implicit-def: $sgpr68_lo16
+; GFX11-TRUE16-NEXT: ; implicit-def: $sgpr54_lo16
; GFX11-TRUE16-NEXT: ; implicit-def: $sgpr87_lo16
; GFX11-TRUE16-NEXT: ; implicit-def: $sgpr35_lo16
; GFX11-TRUE16-NEXT: ; implicit-def: $sgpr100_lo16
@@ -176476,198 +176585,198 @@ define inreg <128 x i8> @bitcast_v64bf16_to_v128i8_scalar(<64 x bfloat> inreg %a
; GFX11-TRUE16-NEXT: ; implicit-def: $sgpr88_lo16
; GFX11-TRUE16-NEXT: ; implicit-def: $sgpr78_lo16
; GFX11-TRUE16-NEXT: ; implicit-def: $sgpr76_lo16
-; GFX11-TRUE16-NEXT: v_writelane_b32 v43, s4, 1
-; GFX11-TRUE16-NEXT: ; implicit-def: $sgpr4_lo16
-; GFX11-TRUE16-NEXT: v_writelane_b32 v43, s4, 2
-; GFX11-TRUE16-NEXT: ; implicit-def: $sgpr4_lo16
-; GFX11-TRUE16-NEXT: v_writelane_b32 v43, s4, 3
-; GFX11-TRUE16-NEXT: ; implicit-def: $sgpr4_lo16
-; GFX11-TRUE16-NEXT: v_writelane_b32 v43, s4, 4
-; GFX11-TRUE16-NEXT: ; implicit-def: $sgpr4_lo16
-; GFX11-TRUE16-NEXT: v_writelane_b32 v43, s4, 5
-; GFX11-TRUE16-NEXT: ; implicit-def: $sgpr4_lo16
-; GFX11-TRUE16-NEXT: v_writelane_b32 v43, s4, 6
-; GFX11-TRUE16-NEXT: ; implicit-def: $sgpr4_lo16
-; GFX11-TRUE16-NEXT: v_writelane_b32 v43, s4, 7
-; GFX11-TRUE16-NEXT: ; implicit-def: $sgpr4_lo16
-; GFX11-TRUE16-NEXT: v_writelane_b32 v43, s4, 8
-; GFX11-TRUE16-NEXT: ; implicit-def: $sgpr4_lo16
-; GFX11-TRUE16-NEXT: v_writelane_b32 v43, s4, 9
-; GFX11-TRUE16-NEXT: ; implicit-def: $sgpr4_lo16
-; GFX11-TRUE16-NEXT: v_writelane_b32 v43, s4, 10
-; GFX11-TRUE16-NEXT: ; implicit-def: $sgpr4_lo16
-; GFX11-TRUE16-NEXT: v_writelane_b32 v43, s4, 11
-; GFX11-TRUE16-NEXT: v_writelane_b32 v43, s5, 12
-; GFX11-TRUE16-NEXT: ; implicit-def: $sgpr4_lo16
-; GFX11-TRUE16-NEXT: v_writelane_b32 v43, s4, 13
-; GFX11-TRUE16-NEXT: ; implicit-def: $sgpr4_lo16
-; GFX11-TRUE16-NEXT: v_writelane_b32 v43, s4, 14
-; GFX11-TRUE16-NEXT: ; implicit-def: $sgpr4_lo16
-; GFX11-TRUE16-NEXT: v_writelane_b32 v43, s4, 15
-; GFX11-TRUE16-NEXT: ; implicit-def: $sgpr4_lo16
-; GFX11-TRUE16-NEXT: v_writelane_b32 v43, s4, 16
-; GFX11-TRUE16-NEXT: ; implicit-def: $sgpr4_lo16
-; GFX11-TRUE16-NEXT: v_writelane_b32 v43, s4, 17
-; GFX11-TRUE16-NEXT: ; implicit-def: $sgpr4_lo16
-; GFX11-TRUE16-NEXT: v_writelane_b32 v43, s4, 18
-; GFX11-TRUE16-NEXT: v_writelane_b32 v43, s5, 19
-; GFX11-TRUE16-NEXT: ; implicit-def: $sgpr4_lo16
-; GFX11-TRUE16-NEXT: v_writelane_b32 v43, s4, 20
-; GFX11-TRUE16-NEXT: ; implicit-def: $sgpr4_lo16
-; GFX11-TRUE16-NEXT: v_writelane_b32 v43, s4, 21
-; GFX11-TRUE16-NEXT: ; implicit-def: $sgpr4_lo16
-; GFX11-TRUE16-NEXT: v_writelane_b32 v43, s4, 22
-; GFX11-TRUE16-NEXT: ; implicit-def: $sgpr4_lo16
-; GFX11-TRUE16-NEXT: v_writelane_b32 v43, s4, 23
-; GFX11-TRUE16-NEXT: ; implicit-def: $sgpr4_lo16
-; GFX11-TRUE16-NEXT: v_writelane_b32 v43, s4, 24
-; GFX11-TRUE16-NEXT: ; implicit-def: $sgpr4_lo16
-; GFX11-TRUE16-NEXT: v_writelane_b32 v43, s4, 25
-; GFX11-TRUE16-NEXT: v_writelane_b32 v43, s5, 26
-; GFX11-TRUE16-NEXT: ; implicit-def: $sgpr4_lo16
-; GFX11-TRUE16-NEXT: ; implicit-def: $sgpr5_lo16
-; GFX11-TRUE16-NEXT: v_writelane_b32 v43, s4, 27
-; GFX11-TRUE16-NEXT: ; implicit-def: $sgpr4_lo16
-; GFX11-TRUE16-NEXT: v_writelane_b32 v43, s4, 28
-; GFX11-TRUE16-NEXT: ; implicit-def: $sgpr4_lo16
-; GFX11-TRUE16-NEXT: v_writelane_b32 v43, s4, 29
-; GFX11-TRUE16-NEXT: ; implicit-def: $sgpr4_lo16
-; GFX11-TRUE16-NEXT: v_writelane_b32 v43, s4, 30
-; GFX11-TRUE16-NEXT: ; implicit-def: $sgpr4_lo16
-; GFX11-TRUE16-NEXT: v_writelane_b32 v43, s4, 31
-; GFX11-TRUE16-NEXT: ; implicit-def: $sgpr4_lo16
-; GFX11-TRUE16-NEXT: v_writelane_b32 v42, s4, 0
-; GFX11-TRUE16-NEXT: ; implicit-def: $sgpr4_lo16
-; GFX11-TRUE16-NEXT: v_writelane_b32 v42, s4, 1
-; GFX11-TRUE16-NEXT: ; implicit-def: $sgpr4_lo16
-; GFX11-TRUE16-NEXT: v_writelane_b32 v42, s4, 2
-; GFX11-TRUE16-NEXT: ; implicit-def: $sgpr4_lo16
-; GFX11-TRUE16-NEXT: v_writelane_b32 v42, s4, 3
-; GFX11-TRUE16-NEXT: ; implicit-def: $sgpr4_lo16
-; GFX11-TRUE16-NEXT: v_writelane_b32 v42, s4, 4
-; GFX11-TRUE16-NEXT: ; implicit-def: $sgpr4_lo16
-; GFX11-TRUE16-NEXT: v_writelane_b32 v42, s4, 5
-; GFX11-TRUE16-NEXT: ; implicit-def: $sgpr4_lo16
-; GFX11-TRUE16-NEXT: v_writelane_b32 v42, s4, 6
-; GFX11-TRUE16-NEXT: ; implicit-def: $sgpr4_lo16
-; GFX11-TRUE16-NEXT: v_writelane_b32 v42, s4, 7
-; GFX11-TRUE16-NEXT: ; implicit-def: $sgpr4_lo16
+; GFX11-TRUE16-NEXT: v_writelane_b32 v43, s10, 1
+; GFX11-TRUE16-NEXT: ; implicit-def: $sgpr10_lo16
+; GFX11-TRUE16-NEXT: v_writelane_b32 v43, s10, 2
+; GFX11-TRUE16-NEXT: ; implicit-def: $sgpr10_lo16
+; GFX11-TRUE16-NEXT: v_writelane_b32 v43, s10, 3
+; GFX11-TRUE16-NEXT: ; implicit-def: $sgpr10_lo16
+; GFX11-TRUE16-NEXT: v_writelane_b32 v43, s10, 4
+; GFX11-TRUE16-NEXT: ; implicit-def: $sgpr10_lo16
+; GFX11-TRUE16-NEXT: v_writelane_b32 v43, s10, 5
+; GFX11-TRUE16-NEXT: ; implicit-def: $sgpr10_lo16
+; GFX11-TRUE16-NEXT: v_writelane_b32 v43, s10, 6
+; GFX11-TRUE16-NEXT: ; implicit-def: $sgpr10_lo16
+; GFX11-TRUE16-NEXT: v_writelane_b32 v43, s10, 7
+; GFX11-TRUE16-NEXT: ; implicit-def: $sgpr10_lo16
+; GFX11-TRUE16-NEXT: v_writelane_b32 v43, s10, 8
+; GFX11-TRUE16-NEXT: ; implicit-def: $sgpr10_lo16
+; GFX11-TRUE16-NEXT: v_writelane_b32 v43, s10, 9
+; GFX11-TRUE16-NEXT: ; implicit-def: $sgpr10_lo16
+; GFX11-TRUE16-NEXT: v_writelane_b32 v43, s10, 10
+; GFX11-TRUE16-NEXT: ; implicit-def: $sgpr10_lo16
+; GFX11-TRUE16-NEXT: v_writelane_b32 v43, s10, 11
+; GFX11-TRUE16-NEXT: v_writelane_b32 v43, s11, 12
+; GFX11-TRUE16-NEXT: ; implicit-def: $sgpr10_lo16
+; GFX11-TRUE16-NEXT: v_writelane_b32 v43, s10, 13
+; GFX11-TRUE16-NEXT: ; implicit-def: $sgpr10_lo16
+; GFX11-TRUE16-NEXT: v_writelane_b32 v43, s10, 14
+; GFX11-TRUE16-NEXT: ; implicit-def: $sgpr10_lo16
+; GFX11-TRUE16-NEXT: v_writelane_b32 v43, s10, 15
+; GFX11-TRUE16-NEXT: ; implicit-def: $sgpr10_lo16
+; GFX11-TRUE16-NEXT: v_writelane_b32 v43, s10, 16
+; GFX11-TRUE16-NEXT: ; implicit-def: $sgpr10_lo16
+; GFX11-TRUE16-NEXT: v_writelane_b32 v43, s10, 17
+; GFX11-TRUE16-NEXT: ; implicit-def: $sgpr10_lo16
+; GFX11-TRUE16-NEXT: v_writelane_b32 v43, s10, 18
+; GFX11-TRUE16-NEXT: v_writelane_b32 v43, s11, 19
+; GFX11-TRUE16-NEXT: ; implicit-def: $sgpr10_lo16
+; GFX11-TRUE16-NEXT: v_writelane_b32 v43, s10, 20
+; GFX11-TRUE16-NEXT: ; implicit-def: $sgpr10_lo16
+; GFX11-TRUE16-NEXT: v_writelane_b32 v43, s10, 21
+; GFX11-TRUE16-NEXT: ; implicit-def: $sgpr10_lo16
+; GFX11-TRUE16-NEXT: v_writelane_b32 v43, s10, 22
+; GFX11-TRUE16-NEXT: ; implicit-def: $sgpr10_lo16
+; GFX11-TRUE16-NEXT: v_writelane_b32 v43, s10, 23
+; GFX11-TRUE16-NEXT: ; implicit-def: $sgpr10_lo16
+; GFX11-TRUE16-NEXT: v_writelane_b32 v43, s10, 24
+; GFX11-TRUE16-NEXT: ; implicit-def: $sgpr10_lo16
+; GFX11-TRUE16-NEXT: v_writelane_b32 v43, s10, 25
+; GFX11-TRUE16-NEXT: v_writelane_b32 v43, s11, 26
+; GFX11-TRUE16-NEXT: ; implicit-def: $sgpr10_lo16
+; GFX11-TRUE16-NEXT: ; implicit-def: $sgpr11_lo16
+; GFX11-TRUE16-NEXT: v_writelane_b32 v43, s10, 27
+; GFX11-TRUE16-NEXT: ; implicit-def: $sgpr10_lo16
+; GFX11-TRUE16-NEXT: v_writelane_b32 v43, s10, 28
+; GFX11-TRUE16-NEXT: ; implicit-def: $sgpr10_lo16
+; GFX11-TRUE16-NEXT: v_writelane_b32 v43, s10, 29
+; GFX11-TRUE16-NEXT: ; implicit-def: $sgpr10_lo16
+; GFX11-TRUE16-NEXT: v_writelane_b32 v43, s10, 30
+; GFX11-TRUE16-NEXT: ; implicit-def: $sgpr10_lo16
+; GFX11-TRUE16-NEXT: v_writelane_b32 v43, s10, 31
+; GFX11-TRUE16-NEXT: ; implicit-def: $sgpr10_lo16
+; GFX11-TRUE16-NEXT: v_writelane_b32 v42, s10, 0
+; GFX11-TRUE16-NEXT: ; implicit-def: $sgpr10_lo16
+; GFX11-TRUE16-NEXT: v_writelane_b32 v42, s10, 1
+; GFX11-TRUE16-NEXT: ; implicit-def: $sgpr10_lo16
+; GFX11-TRUE16-NEXT: v_writelane_b32 v42, s10, 2
+; GFX11-TRUE16-NEXT: ; implicit-def: $sgpr10_lo16
+; GFX11-TRUE16-NEXT: v_writelane_b32 v42, s10, 3
+; GFX11-TRUE16-NEXT: ; implicit-def: $sgpr10_lo16
+; GFX11-TRUE16-NEXT: v_writelane_b32 v42, s10, 4
+; GFX11-TRUE16-NEXT: ; implicit-def: $sgpr10_lo16
+; GFX11-TRUE16-NEXT: v_writelane_b32 v42, s10, 5
+; GFX11-TRUE16-NEXT: ; implicit-def: $sgpr10_lo16
+; GFX11-TRUE16-NEXT: v_writelane_b32 v42, s10, 6
+; GFX11-TRUE16-NEXT: ; implicit-def: $sgpr10_lo16
+; GFX11-TRUE16-NEXT: v_writelane_b32 v42, s10, 7
+; GFX11-TRUE16-NEXT: ; implicit-def: $sgpr10_lo16
; GFX11-TRUE16-NEXT: v_writelane_b32 v42, s74, 8
; GFX11-TRUE16-NEXT: v_writelane_b32 v42, s75, 9
; GFX11-TRUE16-NEXT: ; implicit-def: $sgpr74_lo16
; GFX11-TRUE16-NEXT: s_branch .LBB91_2
; GFX11-TRUE16-NEXT: .LBB91_4:
+; GFX11-TRUE16-NEXT: v_mov_b16_e32 v9.l, s90
+; GFX11-TRUE16-NEXT: v_mov_b16_e32 v10.l, s92
; GFX11-TRUE16-NEXT: v_mov_b16_e32 v11.l, s94
+; GFX11-TRUE16-NEXT: v_readlane_b32 s90, v43, 25
+; GFX11-TRUE16-NEXT: v_readlane_b32 s92, v43, 18
+; GFX11-TRUE16-NEXT: v_readlane_b32 s94, v42, 8
; GFX11-TRUE16-NEXT: v_mov_b16_e32 v12.l, s30
-; GFX11-TRUE16-NEXT: v_readlane_b32 s94, v43, 25
-; GFX11-TRUE16-NEXT: v_readlane_b32 s30, v43, 18
-; GFX11-TRUE16-NEXT: v_mov_b16_e32 v55.l, s36
-; GFX11-TRUE16-NEXT: v_mov_b16_e32 v64.l, s38
-; GFX11-TRUE16-NEXT: v_mov_b16_e32 v81.l, s37
-; GFX11-TRUE16-NEXT: v_mov_b16_e32 v86.l, s39
-; GFX11-TRUE16-NEXT: v_readlane_b32 s95, v43, 26
-; GFX11-TRUE16-NEXT: v_readlane_b32 s31, v43, 19
-; GFX11-TRUE16-NEXT: v_readlane_b32 s36, v42, 8
-; GFX11-TRUE16-NEXT: v_readlane_b32 s38, v43, 11
-; GFX11-TRUE16-NEXT: v_mov_b16_e32 v51.l, s40
-; GFX11-TRUE16-NEXT: v_mov_b16_e32 v49.l, s41
+; GFX11-TRUE16-NEXT: v_readlane_b32 s91, v43, 26
+; GFX11-TRUE16-NEXT: v_readlane_b32 s93, v43, 19
+; GFX11-TRUE16-NEXT: v_readlane_b32 s95, v42, 9
+; GFX11-TRUE16-NEXT: v_readlane_b32 s30, v43, 11
+; GFX11-TRUE16-NEXT: v_mov_b16_e32 v51.l, s4
+; GFX11-TRUE16-NEXT: v_mov_b16_e32 v49.l, s5
; GFX11-TRUE16-NEXT: v_mov_b16_e32 v48.l, s87
-; GFX11-TRUE16-NEXT: v_mov_b16_e32 v50.l, s42
-; GFX11-TRUE16-NEXT: v_mov_b16_e32 v37.l, s43
+; GFX11-TRUE16-NEXT: v_mov_b16_e32 v50.l, s6
+; GFX11-TRUE16-NEXT: v_mov_b16_e32 v37.l, s7
; GFX11-TRUE16-NEXT: v_mov_b16_e32 v38.l, s96
-; GFX11-TRUE16-NEXT: v_mov_b16_e32 v39.l, s44
-; GFX11-TRUE16-NEXT: v_mov_b16_e32 v36.l, s45
-; GFX11-TRUE16-NEXT: v_mov_b16_e32 v35.l, s8
-; GFX11-TRUE16-NEXT: v_mov_b16_e32 v34.l, s58
-; GFX11-TRUE16-NEXT: v_mov_b16_e32 v33.l, s59
-; GFX11-TRUE16-NEXT: v_mov_b16_e32 v32.l, s9
-; GFX11-TRUE16-NEXT: v_mov_b16_e32 v31.l, s60
-; GFX11-TRUE16-NEXT: v_mov_b16_e32 v30.l, s61
-; GFX11-TRUE16-NEXT: v_mov_b16_e32 v28.l, s10
+; GFX11-TRUE16-NEXT: v_mov_b16_e32 v39.l, s8
+; GFX11-TRUE16-NEXT: v_mov_b16_e32 v36.l, s9
+; GFX11-TRUE16-NEXT: v_mov_b16_e32 v35.l, s14
+; GFX11-TRUE16-NEXT: v_mov_b16_e32 v34.l, s44
+; GFX11-TRUE16-NEXT: v_mov_b16_e32 v33.l, s45
+; GFX11-TRUE16-NEXT: v_mov_b16_e32 v32.l, s15
+; GFX11-TRUE16-NEXT: v_mov_b16_e32 v31.l, s58
+; GFX11-TRUE16-NEXT: v_mov_b16_e32 v30.l, s59
+; GFX11-TRUE16-NEXT: v_mov_b16_e32 v28.l, s40
; GFX11-TRUE16-NEXT: v_mov_b16_e32 v29.l, s62
; GFX11-TRUE16-NEXT: v_mov_b16_e32 v26.l, s63
-; GFX11-TRUE16-NEXT: v_mov_b16_e32 v25.l, s11
+; GFX11-TRUE16-NEXT: v_mov_b16_e32 v25.l, s41
; GFX11-TRUE16-NEXT: v_mov_b16_e32 v27.l, s72
; GFX11-TRUE16-NEXT: v_mov_b16_e32 v22.l, s73
; GFX11-TRUE16-NEXT: v_mov_b16_e32 v24.l, s97
; GFX11-TRUE16-NEXT: v_mov_b16_e32 v23.l, s28
; GFX11-TRUE16-NEXT: v_mov_b16_e32 v21.l, s29
; GFX11-TRUE16-NEXT: v_mov_b16_e32 v20.l, s98
-; GFX11-TRUE16-NEXT: v_mov_b16_e32 v53.l, s69
-; GFX11-TRUE16-NEXT: v_mov_b16_e32 v54.l, s70
+; GFX11-TRUE16-NEXT: v_mov_b16_e32 v53.l, s68
+; GFX11-TRUE16-NEXT: v_mov_b16_e32 v54.l, s69
; GFX11-TRUE16-NEXT: v_mov_b16_e32 v13.l, s35
-; GFX11-TRUE16-NEXT: v_mov_b16_e32 v52.l, s55
-; GFX11-TRUE16-NEXT: v_mov_b16_e32 v65.l, s71
-; GFX11-TRUE16-NEXT: v_mov_b16_e32 v66.l, s48
-; GFX11-TRUE16-NEXT: v_mov_b16_e32 v69.l, s80
-; GFX11-TRUE16-NEXT: v_mov_b16_e32 v70.l, s81
-; GFX11-TRUE16-NEXT: v_mov_b16_e32 v67.l, s50
-; GFX11-TRUE16-NEXT: v_mov_b16_e32 v68.l, s64
-; GFX11-TRUE16-NEXT: v_mov_b16_e32 v80.l, s82
-; GFX11-TRUE16-NEXT: v_mov_b16_e32 v18.l, s49
-; GFX11-TRUE16-NEXT: v_mov_b16_e32 v71.l, s51
-; GFX11-TRUE16-NEXT: v_mov_b16_e32 v19.l, s65
+; GFX11-TRUE16-NEXT: v_mov_b16_e32 v52.l, s54
+; GFX11-TRUE16-NEXT: v_mov_b16_e32 v65.l, s70
+; GFX11-TRUE16-NEXT: v_mov_b16_e32 v66.l, s49
+; GFX11-TRUE16-NEXT: v_mov_b16_e32 v55.l, s36
+; GFX11-TRUE16-NEXT: v_mov_b16_e32 v64.l, s39
+; GFX11-TRUE16-NEXT: v_mov_b16_e32 v69.l, s71
+; GFX11-TRUE16-NEXT: v_mov_b16_e32 v70.l, s80
+; GFX11-TRUE16-NEXT: v_mov_b16_e32 v67.l, s37
+; GFX11-TRUE16-NEXT: v_mov_b16_e32 v68.l, s55
+; GFX11-TRUE16-NEXT: v_mov_b16_e32 v80.l, s81
+; GFX11-TRUE16-NEXT: v_mov_b16_e32 v18.l, s82
+; GFX11-TRUE16-NEXT: v_mov_b16_e32 v71.l, s50
+; GFX11-TRUE16-NEXT: v_mov_b16_e32 v19.l, s64
; GFX11-TRUE16-NEXT: v_mov_b16_e32 v82.l, s83
; GFX11-TRUE16-NEXT: v_mov_b16_e32 v16.l, s84
-; GFX11-TRUE16-NEXT: v_mov_b16_e32 v17.l, s66
+; GFX11-TRUE16-NEXT: v_mov_b16_e32 v81.l, s38
+; GFX11-TRUE16-NEXT: v_mov_b16_e32 v17.l, s65
; GFX11-TRUE16-NEXT: v_mov_b16_e32 v84.l, s85
-; GFX11-TRUE16-NEXT: v_mov_b16_e32 v14.l, s4
-; GFX11-TRUE16-NEXT: v_mov_b16_e32 v83.l, s52
-; GFX11-TRUE16-NEXT: v_mov_b16_e32 v15.l, s67
-; GFX11-TRUE16-NEXT: v_mov_b16_e32 v87.l, s5
-; GFX11-TRUE16-NEXT: v_mov_b16_e32 v96.l, s6
-; GFX11-TRUE16-NEXT: v_mov_b16_e32 v85.l, s53
-; GFX11-TRUE16-NEXT: v_mov_b16_e32 v99.l, s7
+; GFX11-TRUE16-NEXT: v_mov_b16_e32 v14.l, s10
+; GFX11-TRUE16-NEXT: v_mov_b16_e32 v83.l, s51
+; GFX11-TRUE16-NEXT: v_mov_b16_e32 v15.l, s66
+; GFX11-TRUE16-NEXT: v_mov_b16_e32 v87.l, s11
+; GFX11-TRUE16-NEXT: v_mov_b16_e32 v96.l, s12
+; GFX11-TRUE16-NEXT: v_mov_b16_e32 v85.l, s52
+; GFX11-TRUE16-NEXT: v_mov_b16_e32 v86.l, s48
+; GFX11-TRUE16-NEXT: v_mov_b16_e32 v99.l, s13
; GFX11-TRUE16-NEXT: v_mov_b16_e32 v100.l, s86
-; GFX11-TRUE16-NEXT: v_mov_b16_e32 v97.l, s54
-; GFX11-TRUE16-NEXT: v_mov_b16_e32 v98.l, s68
+; GFX11-TRUE16-NEXT: v_mov_b16_e32 v97.l, s53
+; GFX11-TRUE16-NEXT: v_mov_b16_e32 v98.l, s67
; GFX11-TRUE16-NEXT: v_mov_b16_e32 v5.l, s74
; GFX11-TRUE16-NEXT: v_mov_b16_e32 v6.l, s76
; GFX11-TRUE16-NEXT: v_mov_b16_e32 v7.l, s78
; GFX11-TRUE16-NEXT: v_mov_b16_e32 v8.l, s88
-; GFX11-TRUE16-NEXT: v_mov_b16_e32 v9.l, s90
-; GFX11-TRUE16-NEXT: v_mov_b16_e32 v10.l, s92
-; GFX11-TRUE16-NEXT: v_readlane_b32 s58, v43, 0
-; GFX11-TRUE16-NEXT: v_readlane_b32 s59, v43, 2
-; GFX11-TRUE16-NEXT: v_readlane_b32 s60, v43, 7
-; GFX11-TRUE16-NEXT: v_readlane_b32 s61, v43, 14
-; GFX11-TRUE16-NEXT: v_readlane_b32 s62, v43, 21
-; GFX11-TRUE16-NEXT: v_readlane_b32 s63, v43, 28
-; GFX11-TRUE16-NEXT: v_readlane_b32 s72, v42, 1
+; GFX11-TRUE16-NEXT: v_readlane_b32 s63, v43, 0
+; GFX11-TRUE16-NEXT: v_readlane_b32 s62, v43, 2
+; GFX11-TRUE16-NEXT: v_readlane_b32 s58, v43, 7
+; GFX11-TRUE16-NEXT: v_readlane_b32 s59, v43, 14
+; GFX11-TRUE16-NEXT: v_readlane_b32 s72, v43, 21
+; GFX11-TRUE16-NEXT: v_readlane_b32 s44, v43, 28
+; GFX11-TRUE16-NEXT: v_readlane_b32 s45, v42, 1
; GFX11-TRUE16-NEXT: v_readlane_b32 s73, v42, 6
-; GFX11-TRUE16-NEXT: v_readlane_b32 s13, v42, 7
-; GFX11-TRUE16-NEXT: v_readlane_b32 s15, v42, 5
-; GFX11-TRUE16-NEXT: v_readlane_b32 s45, v42, 4
-; GFX11-TRUE16-NEXT: v_readlane_b32 s44, v42, 3
-; GFX11-TRUE16-NEXT: v_readlane_b32 s47, v42, 2
-; GFX11-TRUE16-NEXT: v_readlane_b32 s11, v42, 0
-; GFX11-TRUE16-NEXT: v_readlane_b32 s57, v43, 31
-; GFX11-TRUE16-NEXT: v_readlane_b32 s10, v43, 30
+; GFX11-TRUE16-NEXT: v_readlane_b32 s43, v42, 7
+; GFX11-TRUE16-NEXT: v_readlane_b32 s41, v42, 5
+; GFX11-TRUE16-NEXT: v_readlane_b32 s47, v42, 4
+; GFX11-TRUE16-NEXT: v_readlane_b32 s40, v42, 3
+; GFX11-TRUE16-NEXT: v_readlane_b32 s57, v42, 2
+; GFX11-TRUE16-NEXT: v_readlane_b32 s29, v42, 0
+; GFX11-TRUE16-NEXT: v_readlane_b32 s61, v43, 31
+; GFX11-TRUE16-NEXT: v_readlane_b32 s28, v43, 30
; GFX11-TRUE16-NEXT: v_readlane_b32 s74, v43, 29
-; GFX11-TRUE16-NEXT: v_readlane_b32 s9, v43, 27
+; GFX11-TRUE16-NEXT: v_readlane_b32 s11, v43, 27
; GFX11-TRUE16-NEXT: v_readlane_b32 s75, v43, 24
-; GFX11-TRUE16-NEXT: v_readlane_b32 s8, v43, 23
+; GFX11-TRUE16-NEXT: v_readlane_b32 s10, v43, 23
; GFX11-TRUE16-NEXT: v_readlane_b32 s76, v43, 22
-; GFX11-TRUE16-NEXT: v_readlane_b32 s77, v43, 20
-; GFX11-TRUE16-NEXT: v_readlane_b32 s78, v43, 17
-; GFX11-TRUE16-NEXT: v_readlane_b32 s79, v43, 16
-; GFX11-TRUE16-NEXT: v_readlane_b32 s88, v43, 15
-; GFX11-TRUE16-NEXT: v_readlane_b32 s89, v43, 13
-; GFX11-TRUE16-NEXT: v_readlane_b32 s90, v43, 10
-; GFX11-TRUE16-NEXT: v_readlane_b32 s91, v43, 9
-; GFX11-TRUE16-NEXT: s_mov_b32 s92, s100
-; GFX11-TRUE16-NEXT: v_readlane_b32 s93, v43, 8
-; GFX11-TRUE16-NEXT: v_readlane_b32 s43, v43, 6
-; GFX11-TRUE16-NEXT: v_readlane_b32 s95, v43, 5
-; GFX11-TRUE16-NEXT: v_readlane_b32 s37, v42, 9
-; GFX11-TRUE16-NEXT: v_readlane_b32 s42, v43, 4
-; GFX11-TRUE16-NEXT: v_readlane_b32 s41, v43, 1
-; GFX11-TRUE16-NEXT: v_readlane_b32 s39, v43, 12
-; GFX11-TRUE16-NEXT: v_readlane_b32 s31, v43, 3
+; GFX11-TRUE16-NEXT: v_readlane_b32 s15, v43, 20
+; GFX11-TRUE16-NEXT: v_readlane_b32 s77, v43, 17
+; GFX11-TRUE16-NEXT: v_readlane_b32 s14, v43, 16
+; GFX11-TRUE16-NEXT: v_readlane_b32 s78, v43, 15
+; GFX11-TRUE16-NEXT: v_readlane_b32 s13, v43, 13
+; GFX11-TRUE16-NEXT: v_readlane_b32 s79, v43, 10
+; GFX11-TRUE16-NEXT: v_readlane_b32 s12, v43, 9
+; GFX11-TRUE16-NEXT: s_mov_b32 s88, s100
+; GFX11-TRUE16-NEXT: v_readlane_b32 s89, v43, 8
+; GFX11-TRUE16-NEXT: v_readlane_b32 s9, v43, 6
+; GFX11-TRUE16-NEXT: v_readlane_b32 s91, v43, 5
+; GFX11-TRUE16-NEXT: v_readlane_b32 s8, v43, 4
+; GFX11-TRUE16-NEXT: v_readlane_b32 s93, v43, 1
+; GFX11-TRUE16-NEXT: v_readlane_b32 s31, v43, 12
+; GFX11-TRUE16-NEXT: v_readlane_b32 s95, v43, 3
; GFX11-TRUE16-NEXT: .LBB91_5: ; %end
; GFX11-TRUE16-NEXT: s_mov_b32 s4, s34
; GFX11-TRUE16-NEXT: s_mov_b32 s5, s104
-; GFX11-TRUE16-NEXT: s_mov_b32 s6, s56
+; GFX11-TRUE16-NEXT: s_mov_b32 s6, s60
; GFX11-TRUE16-NEXT: s_and_b32 s0, s0, 0xff
; GFX11-TRUE16-NEXT: s_lshl_b32 s4, s4, 8
; GFX11-TRUE16-NEXT: s_and_b32 s5, s5, 0xff
@@ -176675,7 +176784,7 @@ define inreg <128 x i8> @bitcast_v64bf16_to_v128i8_scalar(<64 x bfloat> inreg %a
; GFX11-TRUE16-NEXT: s_or_b32 s0, s0, s4
; GFX11-TRUE16-NEXT: s_or_b32 s4, s5, s6
; GFX11-TRUE16-NEXT: s_mov_b32 s5, s103
-; GFX11-TRUE16-NEXT: s_mov_b32 s6, s58
+; GFX11-TRUE16-NEXT: s_mov_b32 s6, s63
; GFX11-TRUE16-NEXT: s_mov_b32 s7, s102
; GFX11-TRUE16-NEXT: s_and_b32 s1, s1, 0xff
; GFX11-TRUE16-NEXT: s_lshl_b32 s5, s5, 8
@@ -176689,18 +176798,18 @@ define inreg <128 x i8> @bitcast_v64bf16_to_v128i8_scalar(<64 x bfloat> inreg %a
; GFX11-TRUE16-NEXT: s_lshl_b32 s5, s5, 16
; GFX11-TRUE16-NEXT: s_or_b32 s0, s0, s4
; GFX11-TRUE16-NEXT: s_or_b32 s1, s1, s5
-; GFX11-TRUE16-NEXT: s_mov_b32 s4, s12
+; GFX11-TRUE16-NEXT: s_mov_b32 s4, s42
; GFX11-TRUE16-NEXT: s_mov_b32 s5, s99
-; GFX11-TRUE16-NEXT: s_mov_b32 s6, s46
+; GFX11-TRUE16-NEXT: s_mov_b32 s6, s56
; GFX11-TRUE16-NEXT: s_and_b32 s2, s2, 0xff
; GFX11-TRUE16-NEXT: s_lshl_b32 s4, s4, 8
; GFX11-TRUE16-NEXT: s_and_b32 s5, s5, 0xff
; GFX11-TRUE16-NEXT: s_lshl_b32 s6, s6, 8
; GFX11-TRUE16-NEXT: s_or_b32 s2, s2, s4
; GFX11-TRUE16-NEXT: s_or_b32 s4, s5, s6
-; GFX11-TRUE16-NEXT: s_mov_b32 s5, s41
-; GFX11-TRUE16-NEXT: s_mov_b32 s6, s59
-; GFX11-TRUE16-NEXT: s_mov_b32 s7, s31
+; GFX11-TRUE16-NEXT: s_mov_b32 s5, s93
+; GFX11-TRUE16-NEXT: s_mov_b32 s6, s62
+; GFX11-TRUE16-NEXT: s_mov_b32 s7, s95
; GFX11-TRUE16-NEXT: s_and_b32 s3, s3, 0xff
; GFX11-TRUE16-NEXT: s_lshl_b32 s5, s5, 8
; GFX11-TRUE16-NEXT: s_and_b32 s6, s6, 0xff
@@ -176716,9 +176825,9 @@ define inreg <128 x i8> @bitcast_v64bf16_to_v128i8_scalar(<64 x bfloat> inreg %a
; GFX11-TRUE16-NEXT: v_dual_mov_b32 v1, s0 :: v_dual_mov_b32 v2, s1
; GFX11-TRUE16-NEXT: v_dual_mov_b32 v3, s2 :: v_dual_mov_b32 v4, s3
; GFX11-TRUE16-NEXT: s_mov_b32 s0, s16
-; GFX11-TRUE16-NEXT: s_mov_b32 s1, s42
-; GFX11-TRUE16-NEXT: s_mov_b32 s2, s95
-; GFX11-TRUE16-NEXT: s_mov_b32 s3, s14
+; GFX11-TRUE16-NEXT: s_mov_b32 s1, s8
+; GFX11-TRUE16-NEXT: s_mov_b32 s2, s91
+; GFX11-TRUE16-NEXT: s_mov_b32 s3, s46
; GFX11-TRUE16-NEXT: s_and_b32 s0, s0, 0xff
; GFX11-TRUE16-NEXT: s_lshl_b32 s1, s1, 8
; GFX11-TRUE16-NEXT: s_and_b32 s2, s2, 0xff
@@ -176726,9 +176835,9 @@ define inreg <128 x i8> @bitcast_v64bf16_to_v128i8_scalar(<64 x bfloat> inreg %a
; GFX11-TRUE16-NEXT: s_or_b32 s0, s0, s1
; GFX11-TRUE16-NEXT: s_or_b32 s1, s2, s3
; GFX11-TRUE16-NEXT: s_mov_b32 s2, s17
-; GFX11-TRUE16-NEXT: s_mov_b32 s3, s43
-; GFX11-TRUE16-NEXT: s_mov_b32 s4, s60
-; GFX11-TRUE16-NEXT: s_mov_b32 s5, s93
+; GFX11-TRUE16-NEXT: s_mov_b32 s3, s9
+; GFX11-TRUE16-NEXT: s_mov_b32 s4, s58
+; GFX11-TRUE16-NEXT: s_mov_b32 s5, s89
; GFX11-TRUE16-NEXT: s_and_b32 s2, s2, 0xff
; GFX11-TRUE16-NEXT: s_lshl_b32 s3, s3, 8
; GFX11-TRUE16-NEXT: s_and_b32 s4, s4, 0xff
@@ -176742,9 +176851,9 @@ define inreg <128 x i8> @bitcast_v64bf16_to_v128i8_scalar(<64 x bfloat> inreg %a
; GFX11-TRUE16-NEXT: s_or_b32 s0, s0, s1
; GFX11-TRUE16-NEXT: s_or_b32 s1, s2, s3
; GFX11-TRUE16-NEXT: s_mov_b32 s2, s18
-; GFX11-TRUE16-NEXT: s_mov_b32 s3, s91
-; GFX11-TRUE16-NEXT: s_mov_b32 s4, s90
-; GFX11-TRUE16-NEXT: s_mov_b32 s5, s38
+; GFX11-TRUE16-NEXT: s_mov_b32 s3, s12
+; GFX11-TRUE16-NEXT: s_mov_b32 s4, s79
+; GFX11-TRUE16-NEXT: s_mov_b32 s5, s30
; GFX11-TRUE16-NEXT: s_and_b32 s2, s2, 0xff
; GFX11-TRUE16-NEXT: s_lshl_b32 s3, s3, 8
; GFX11-TRUE16-NEXT: s_and_b32 s4, s4, 0xff
@@ -176752,9 +176861,9 @@ define inreg <128 x i8> @bitcast_v64bf16_to_v128i8_scalar(<64 x bfloat> inreg %a
; GFX11-TRUE16-NEXT: s_or_b32 s2, s2, s3
; GFX11-TRUE16-NEXT: s_or_b32 s3, s4, s5
; GFX11-TRUE16-NEXT: s_mov_b32 s4, s19
-; GFX11-TRUE16-NEXT: s_mov_b32 s5, s89
-; GFX11-TRUE16-NEXT: s_mov_b32 s6, s61
-; GFX11-TRUE16-NEXT: s_mov_b32 s7, s88
+; GFX11-TRUE16-NEXT: s_mov_b32 s5, s13
+; GFX11-TRUE16-NEXT: s_mov_b32 s6, s59
+; GFX11-TRUE16-NEXT: s_mov_b32 s7, s78
; GFX11-TRUE16-NEXT: s_and_b32 s4, s4, 0xff
; GFX11-TRUE16-NEXT: s_lshl_b32 s5, s5, 8
; GFX11-TRUE16-NEXT: s_and_b32 s6, s6, 0xff
@@ -176770,9 +176879,9 @@ define inreg <128 x i8> @bitcast_v64bf16_to_v128i8_scalar(<64 x bfloat> inreg %a
; GFX11-TRUE16-NEXT: v_dual_mov_b32 v112, s0 :: v_dual_mov_b32 v113, s1
; GFX11-TRUE16-NEXT: v_dual_mov_b32 v114, s2 :: v_dual_mov_b32 v115, s3
; GFX11-TRUE16-NEXT: s_mov_b32 s0, s20
-; GFX11-TRUE16-NEXT: s_mov_b32 s1, s79
-; GFX11-TRUE16-NEXT: s_mov_b32 s2, s78
-; GFX11-TRUE16-NEXT: s_mov_b32 s3, s30
+; GFX11-TRUE16-NEXT: s_mov_b32 s1, s14
+; GFX11-TRUE16-NEXT: s_mov_b32 s2, s77
+; GFX11-TRUE16-NEXT: s_mov_b32 s3, s92
; GFX11-TRUE16-NEXT: s_and_b32 s0, s0, 0xff
; GFX11-TRUE16-NEXT: s_lshl_b32 s1, s1, 8
; GFX11-TRUE16-NEXT: s_and_b32 s2, s2, 0xff
@@ -176780,8 +176889,8 @@ define inreg <128 x i8> @bitcast_v64bf16_to_v128i8_scalar(<64 x bfloat> inreg %a
; GFX11-TRUE16-NEXT: s_or_b32 s0, s0, s1
; GFX11-TRUE16-NEXT: s_or_b32 s1, s2, s3
; GFX11-TRUE16-NEXT: s_mov_b32 s2, s21
-; GFX11-TRUE16-NEXT: s_mov_b32 s3, s77
-; GFX11-TRUE16-NEXT: s_mov_b32 s4, s62
+; GFX11-TRUE16-NEXT: s_mov_b32 s3, s15
+; GFX11-TRUE16-NEXT: s_mov_b32 s4, s72
; GFX11-TRUE16-NEXT: s_mov_b32 s5, s76
; GFX11-TRUE16-NEXT: s_and_b32 s2, s2, 0xff
; GFX11-TRUE16-NEXT: s_lshl_b32 s3, s3, 8
@@ -176796,9 +176905,9 @@ define inreg <128 x i8> @bitcast_v64bf16_to_v128i8_scalar(<64 x bfloat> inreg %a
; GFX11-TRUE16-NEXT: s_or_b32 s0, s0, s1
; GFX11-TRUE16-NEXT: s_or_b32 s1, s2, s3
; GFX11-TRUE16-NEXT: s_mov_b32 s2, s22
-; GFX11-TRUE16-NEXT: s_mov_b32 s3, s8
+; GFX11-TRUE16-NEXT: s_mov_b32 s3, s10
; GFX11-TRUE16-NEXT: s_mov_b32 s4, s75
-; GFX11-TRUE16-NEXT: s_mov_b32 s5, s94
+; GFX11-TRUE16-NEXT: s_mov_b32 s5, s90
; GFX11-TRUE16-NEXT: s_and_b32 s2, s2, 0xff
; GFX11-TRUE16-NEXT: s_lshl_b32 s3, s3, 8
; GFX11-TRUE16-NEXT: s_and_b32 s4, s4, 0xff
@@ -176806,8 +176915,8 @@ define inreg <128 x i8> @bitcast_v64bf16_to_v128i8_scalar(<64 x bfloat> inreg %a
; GFX11-TRUE16-NEXT: s_or_b32 s2, s2, s3
; GFX11-TRUE16-NEXT: s_or_b32 s3, s4, s5
; GFX11-TRUE16-NEXT: s_mov_b32 s4, s23
-; GFX11-TRUE16-NEXT: s_mov_b32 s5, s9
-; GFX11-TRUE16-NEXT: s_mov_b32 s6, s63
+; GFX11-TRUE16-NEXT: s_mov_b32 s5, s11
+; GFX11-TRUE16-NEXT: s_mov_b32 s6, s44
; GFX11-TRUE16-NEXT: s_mov_b32 s7, s74
; GFX11-TRUE16-NEXT: s_and_b32 s4, s4, 0xff
; GFX11-TRUE16-NEXT: s_lshl_b32 s5, s5, 8
@@ -176827,9 +176936,9 @@ define inreg <128 x i8> @bitcast_v64bf16_to_v128i8_scalar(<64 x bfloat> inreg %a
; GFX11-TRUE16-NEXT: v_dual_mov_b32 v1, s0 :: v_dual_mov_b32 v2, s1
; GFX11-TRUE16-NEXT: v_dual_mov_b32 v3, s2 :: v_dual_mov_b32 v4, s3
; GFX11-TRUE16-NEXT: s_mov_b32 s0, s24
-; GFX11-TRUE16-NEXT: s_mov_b32 s1, s10
-; GFX11-TRUE16-NEXT: s_mov_b32 s2, s57
-; GFX11-TRUE16-NEXT: s_mov_b32 s3, s36
+; GFX11-TRUE16-NEXT: s_mov_b32 s1, s28
+; GFX11-TRUE16-NEXT: s_mov_b32 s2, s61
+; GFX11-TRUE16-NEXT: s_mov_b32 s3, s94
; GFX11-TRUE16-NEXT: s_and_b32 s0, s0, 0xff
; GFX11-TRUE16-NEXT: s_lshl_b32 s1, s1, 8
; GFX11-TRUE16-NEXT: s_and_b32 s2, s2, 0xff
@@ -176837,9 +176946,9 @@ define inreg <128 x i8> @bitcast_v64bf16_to_v128i8_scalar(<64 x bfloat> inreg %a
; GFX11-TRUE16-NEXT: s_or_b32 s0, s0, s1
; GFX11-TRUE16-NEXT: s_or_b32 s1, s2, s3
; GFX11-TRUE16-NEXT: s_mov_b32 s2, s25
-; GFX11-TRUE16-NEXT: s_mov_b32 s3, s11
-; GFX11-TRUE16-NEXT: s_mov_b32 s4, s72
-; GFX11-TRUE16-NEXT: s_mov_b32 s5, s47
+; GFX11-TRUE16-NEXT: s_mov_b32 s3, s29
+; GFX11-TRUE16-NEXT: s_mov_b32 s4, s45
+; GFX11-TRUE16-NEXT: s_mov_b32 s5, s57
; GFX11-TRUE16-NEXT: s_and_b32 s2, s2, 0xff
; GFX11-TRUE16-NEXT: s_lshl_b32 s3, s3, 8
; GFX11-TRUE16-NEXT: s_and_b32 s4, s4, 0xff
@@ -176853,9 +176962,9 @@ define inreg <128 x i8> @bitcast_v64bf16_to_v128i8_scalar(<64 x bfloat> inreg %a
; GFX11-TRUE16-NEXT: s_or_b32 s0, s0, s1
; GFX11-TRUE16-NEXT: s_or_b32 s1, s2, s3
; GFX11-TRUE16-NEXT: s_mov_b32 s2, s26
-; GFX11-TRUE16-NEXT: s_mov_b32 s3, s44
-; GFX11-TRUE16-NEXT: s_mov_b32 s4, s45
-; GFX11-TRUE16-NEXT: s_mov_b32 s5, s92
+; GFX11-TRUE16-NEXT: s_mov_b32 s3, s40
+; GFX11-TRUE16-NEXT: s_mov_b32 s4, s47
+; GFX11-TRUE16-NEXT: s_mov_b32 s5, s88
; GFX11-TRUE16-NEXT: s_and_b32 s2, s2, 0xff
; GFX11-TRUE16-NEXT: s_lshl_b32 s3, s3, 8
; GFX11-TRUE16-NEXT: s_and_b32 s4, s4, 0xff
@@ -176863,13 +176972,13 @@ define inreg <128 x i8> @bitcast_v64bf16_to_v128i8_scalar(<64 x bfloat> inreg %a
; GFX11-TRUE16-NEXT: s_or_b32 s2, s2, s3
; GFX11-TRUE16-NEXT: s_or_b32 s3, s4, s5
; GFX11-TRUE16-NEXT: s_mov_b32 s4, s27
-; GFX11-TRUE16-NEXT: s_mov_b32 s5, s15
+; GFX11-TRUE16-NEXT: s_mov_b32 s5, s41
; GFX11-TRUE16-NEXT: s_and_b32 s4, s4, 0xff
; GFX11-TRUE16-NEXT: s_lshl_b32 s5, s5, 8
; GFX11-TRUE16-NEXT: s_mov_b32 s6, s73
; GFX11-TRUE16-NEXT: s_or_b32 s4, s4, s5
; GFX11-TRUE16-NEXT: s_and_b32 s5, s6, 0xff
-; GFX11-TRUE16-NEXT: s_mov_b32 s6, s13
+; GFX11-TRUE16-NEXT: s_mov_b32 s6, s43
; GFX11-TRUE16-NEXT: s_and_b32 s2, s2, 0xffff
; GFX11-TRUE16-NEXT: s_lshl_b32 s6, s6, 8
; GFX11-TRUE16-NEXT: s_lshl_b32 s3, s3, 16
@@ -177110,20 +177219,20 @@ define inreg <128 x i8> @bitcast_v64bf16_to_v128i8_scalar(<64 x bfloat> inreg %a
; GFX11-FAKE16-NEXT: v_readfirstlane_b32 s59, v8
; GFX11-FAKE16-NEXT: v_writelane_b32 v40, s35, 3
; GFX11-FAKE16-NEXT: v_writelane_b32 v41, s99, 3
-; GFX11-FAKE16-NEXT: v_readfirstlane_b32 s56, v9
-; GFX11-FAKE16-NEXT: v_readfirstlane_b32 s57, v10
-; GFX11-FAKE16-NEXT: v_readfirstlane_b32 s46, v11
+; GFX11-FAKE16-NEXT: v_readfirstlane_b32 s8, v9
+; GFX11-FAKE16-NEXT: v_readfirstlane_b32 s9, v10
+; GFX11-FAKE16-NEXT: v_readfirstlane_b32 s6, v11
; GFX11-FAKE16-NEXT: v_writelane_b32 v40, s36, 4
; GFX11-FAKE16-NEXT: v_writelane_b32 v41, s100, 4
-; GFX11-FAKE16-NEXT: v_readfirstlane_b32 s47, v12
-; GFX11-FAKE16-NEXT: v_readfirstlane_b32 s44, v13
-; GFX11-FAKE16-NEXT: v_readfirstlane_b32 s45, v14
+; GFX11-FAKE16-NEXT: v_readfirstlane_b32 s7, v12
+; GFX11-FAKE16-NEXT: v_readfirstlane_b32 s4, v13
+; GFX11-FAKE16-NEXT: v_readfirstlane_b32 s5, v14
; GFX11-FAKE16-NEXT: v_writelane_b32 v40, s37, 5
; GFX11-FAKE16-NEXT: v_writelane_b32 v41, s101, 5
; GFX11-FAKE16-NEXT: s_mov_b32 vcc_hi, 0
-; GFX11-FAKE16-NEXT: s_and_b32 s4, vcc_lo, exec_lo
-; GFX11-FAKE16-NEXT: ; implicit-def: $vgpr43 : SGPR spill to VGPR lane
+; GFX11-FAKE16-NEXT: s_and_b32 s10, vcc_lo, exec_lo
; GFX11-FAKE16-NEXT: ; implicit-def: $vgpr42 : SGPR spill to VGPR lane
+; GFX11-FAKE16-NEXT: ; implicit-def: $vgpr43 : SGPR spill to VGPR lane
; GFX11-FAKE16-NEXT: v_writelane_b32 v40, s38, 6
; GFX11-FAKE16-NEXT: v_writelane_b32 v41, s102, 6
; GFX11-FAKE16-NEXT: v_writelane_b32 v40, s39, 7
@@ -177155,522 +177264,511 @@ define inreg <128 x i8> @bitcast_v64bf16_to_v128i8_scalar(<64 x bfloat> inreg %a
; GFX11-FAKE16-NEXT: v_writelane_b32 v40, s87, 31
; GFX11-FAKE16-NEXT: s_cbranch_scc0 .LBB91_3
; GFX11-FAKE16-NEXT: ; %bb.1: ; %cmp.false
-; GFX11-FAKE16-NEXT: s_lshr_b32 s4, s27, 24
-; GFX11-FAKE16-NEXT: s_lshr_b64 s[12:13], s[26:27], 24
-; GFX11-FAKE16-NEXT: v_writelane_b32 v43, s4, 15
-; GFX11-FAKE16-NEXT: s_lshr_b32 s4, s27, 16
-; GFX11-FAKE16-NEXT: s_lshr_b32 s99, s2, 16
-; GFX11-FAKE16-NEXT: s_lshr_b32 s100, s2, 8
-; GFX11-FAKE16-NEXT: s_lshr_b32 s101, s1, 24
-; GFX11-FAKE16-NEXT: v_writelane_b32 v43, s4, 14
-; GFX11-FAKE16-NEXT: s_lshr_b32 s4, s27, 8
-; GFX11-FAKE16-NEXT: s_lshr_b32 s11, s1, 16
-; GFX11-FAKE16-NEXT: s_lshr_b32 s102, s1, 8
-; GFX11-FAKE16-NEXT: s_lshr_b32 s103, s0, 16
-; GFX11-FAKE16-NEXT: v_writelane_b32 v43, s4, 16
-; GFX11-FAKE16-NEXT: s_lshr_b32 s4, s26, 16
+; GFX11-FAKE16-NEXT: s_lshr_b32 s10, s27, 24
+; GFX11-FAKE16-NEXT: s_lshr_b64 s[42:43], s[22:23], 24
+; GFX11-FAKE16-NEXT: v_writelane_b32 v42, s10, 13
+; GFX11-FAKE16-NEXT: s_lshr_b32 s10, s27, 16
+; GFX11-FAKE16-NEXT: s_lshr_b64 s[46:47], s[16:17], 24
+; GFX11-FAKE16-NEXT: s_lshr_b32 s41, s1, 16
+; GFX11-FAKE16-NEXT: s_lshr_b32 s99, s1, 8
+; GFX11-FAKE16-NEXT: v_writelane_b32 v42, s10, 12
+; GFX11-FAKE16-NEXT: s_lshr_b32 s10, s27, 8
+; GFX11-FAKE16-NEXT: s_lshr_b32 s44, s0, 16
; GFX11-FAKE16-NEXT: s_lshr_b32 s104, s0, 8
-; GFX11-FAKE16-NEXT: s_lshr_b32 s85, s45, 24
-; GFX11-FAKE16-NEXT: s_lshr_b32 s10, s45, 16
-; GFX11-FAKE16-NEXT: v_writelane_b32 v43, s4, 17
-; GFX11-FAKE16-NEXT: s_lshr_b32 s4, s26, 8
-; GFX11-FAKE16-NEXT: s_lshr_b32 s5, s45, 8
-; GFX11-FAKE16-NEXT: s_lshr_b32 s87, s44, 16
-; GFX11-FAKE16-NEXT: s_lshr_b32 s86, s44, 8
-; GFX11-FAKE16-NEXT: v_writelane_b32 v43, s4, 18
-; GFX11-FAKE16-NEXT: s_lshr_b32 s4, s25, 24
-; GFX11-FAKE16-NEXT: s_lshr_b32 s81, s47, 24
-; GFX11-FAKE16-NEXT: s_lshr_b32 s98, s47, 16
-; GFX11-FAKE16-NEXT: s_lshr_b32 s84, s47, 8
-; GFX11-FAKE16-NEXT: v_writelane_b32 v43, s4, 19
-; GFX11-FAKE16-NEXT: s_lshr_b32 s4, s25, 16
-; GFX11-FAKE16-NEXT: s_lshr_b32 s48, s46, 8
-; GFX11-FAKE16-NEXT: s_lshr_b32 s70, s57, 24
-; GFX11-FAKE16-NEXT: s_lshr_b32 s97, s57, 16
-; GFX11-FAKE16-NEXT: v_writelane_b32 v43, s4, 13
-; GFX11-FAKE16-NEXT: s_lshr_b32 s4, s25, 8
-; GFX11-FAKE16-NEXT: s_lshr_b32 s80, s57, 8
-; GFX11-FAKE16-NEXT: s_lshr_b32 s83, s56, 16
-; GFX11-FAKE16-NEXT: s_lshr_b32 s82, s56, 8
-; GFX11-FAKE16-NEXT: v_writelane_b32 v43, s4, 20
-; GFX11-FAKE16-NEXT: s_lshr_b32 s4, s24, 16
-; GFX11-FAKE16-NEXT: s_lshr_b32 s66, s59, 24
-; GFX11-FAKE16-NEXT: s_lshr_b32 s9, s59, 16
-; GFX11-FAKE16-NEXT: s_lshr_b32 s69, s59, 8
-; GFX11-FAKE16-NEXT: v_writelane_b32 v43, s4, 21
-; GFX11-FAKE16-NEXT: s_lshr_b32 s4, s24, 8
+; GFX11-FAKE16-NEXT: s_lshr_b32 s84, s5, 24
+; GFX11-FAKE16-NEXT: v_writelane_b32 v42, s10, 14
+; GFX11-FAKE16-NEXT: s_lshr_b32 s10, s26, 16
+; GFX11-FAKE16-NEXT: s_lshr_b32 s40, s5, 16
+; GFX11-FAKE16-NEXT: s_lshr_b32 s11, s5, 8
+; GFX11-FAKE16-NEXT: s_lshr_b32 s87, s4, 16
+; GFX11-FAKE16-NEXT: v_writelane_b32 v42, s10, 15
+; GFX11-FAKE16-NEXT: s_lshr_b32 s10, s26, 8
+; GFX11-FAKE16-NEXT: s_lshr_b32 s86, s4, 8
+; GFX11-FAKE16-NEXT: s_lshr_b32 s48, s7, 24
+; GFX11-FAKE16-NEXT: s_lshr_b32 s98, s7, 16
+; GFX11-FAKE16-NEXT: v_writelane_b32 v42, s10, 16
+; GFX11-FAKE16-NEXT: s_lshr_b32 s10, s25, 24
+; GFX11-FAKE16-NEXT: s_lshr_b32 s83, s7, 8
+; GFX11-FAKE16-NEXT: s_lshr_b32 s85, s6, 8
+; GFX11-FAKE16-NEXT: s_lshr_b32 s69, s9, 24
+; GFX11-FAKE16-NEXT: v_writelane_b32 v42, s10, 17
+; GFX11-FAKE16-NEXT: s_lshr_b32 s10, s25, 16
+; GFX11-FAKE16-NEXT: s_lshr_b32 s97, s9, 16
+; GFX11-FAKE16-NEXT: s_lshr_b32 s80, s9, 8
+; GFX11-FAKE16-NEXT: s_lshr_b32 s82, s8, 16
+; GFX11-FAKE16-NEXT: v_writelane_b32 v42, s10, 11
+; GFX11-FAKE16-NEXT: s_lshr_b32 s10, s25, 8
+; GFX11-FAKE16-NEXT: s_lshr_b32 s81, s8, 8
+; GFX11-FAKE16-NEXT: s_lshr_b32 s65, s59, 24
+; GFX11-FAKE16-NEXT: s_lshr_b32 s15, s59, 16
+; GFX11-FAKE16-NEXT: v_writelane_b32 v42, s10, 18
+; GFX11-FAKE16-NEXT: s_lshr_b32 s10, s24, 16
+; GFX11-FAKE16-NEXT: s_lshr_b32 s68, s59, 8
; GFX11-FAKE16-NEXT: s_lshr_b32 s71, s58, 16
-; GFX11-FAKE16-NEXT: s_lshr_b32 s39, s58, 8
-; GFX11-FAKE16-NEXT: s_lshr_b32 s55, s61, 24
-; GFX11-FAKE16-NEXT: v_writelane_b32 v43, s4, 22
-; GFX11-FAKE16-NEXT: s_lshr_b32 s4, s23, 24
-; GFX11-FAKE16-NEXT: s_lshr_b32 s8, s61, 16
-; GFX11-FAKE16-NEXT: s_lshr_b32 s65, s61, 8
-; GFX11-FAKE16-NEXT: s_lshr_b32 s68, s60, 16
-; GFX11-FAKE16-NEXT: v_writelane_b32 v43, s4, 23
-; GFX11-FAKE16-NEXT: s_lshr_b32 s4, s23, 16
-; GFX11-FAKE16-NEXT: s_lshr_b32 s67, s60, 8
-; GFX11-FAKE16-NEXT: s_lshr_b32 s51, s63, 24
+; GFX11-FAKE16-NEXT: s_lshr_b32 s70, s58, 8
+; GFX11-FAKE16-NEXT: v_writelane_b32 v42, s10, 19
+; GFX11-FAKE16-NEXT: s_lshr_b32 s10, s24, 8
+; GFX11-FAKE16-NEXT: s_lshr_b32 s54, s61, 24
+; GFX11-FAKE16-NEXT: s_lshr_b32 s14, s61, 16
+; GFX11-FAKE16-NEXT: s_lshr_b32 s39, s61, 8
+; GFX11-FAKE16-NEXT: v_writelane_b32 v42, s10, 20
+; GFX11-FAKE16-NEXT: s_lshr_b32 s10, s23, 24
+; GFX11-FAKE16-NEXT: s_lshr_b32 s67, s60, 16
+; GFX11-FAKE16-NEXT: s_lshr_b32 s66, s60, 8
+; GFX11-FAKE16-NEXT: s_lshr_b32 s38, s63, 24
+; GFX11-FAKE16-NEXT: v_writelane_b32 v42, s10, 21
+; GFX11-FAKE16-NEXT: s_lshr_b32 s10, s23, 16
; GFX11-FAKE16-NEXT: s_lshr_b32 s96, s63, 16
-; GFX11-FAKE16-NEXT: v_writelane_b32 v43, s4, 12
-; GFX11-FAKE16-NEXT: s_lshr_b32 s4, s23, 8
-; GFX11-FAKE16-NEXT: s_lshr_b32 s54, s63, 8
-; GFX11-FAKE16-NEXT: s_lshr_b32 s38, s62, 16
-; GFX11-FAKE16-NEXT: s_lshr_b32 s64, s62, 8
-; GFX11-FAKE16-NEXT: v_writelane_b32 v43, s4, 24
-; GFX11-FAKE16-NEXT: s_lshr_b32 s4, s22, 16
+; GFX11-FAKE16-NEXT: s_lshr_b32 s53, s63, 8
+; GFX11-FAKE16-NEXT: s_lshr_b32 s64, s62, 16
+; GFX11-FAKE16-NEXT: v_writelane_b32 v42, s10, 10
+; GFX11-FAKE16-NEXT: s_lshr_b32 s10, s23, 8
+; GFX11-FAKE16-NEXT: s_lshr_b32 s55, s62, 8
; GFX11-FAKE16-NEXT: s_lshr_b32 s36, s73, 24
-; GFX11-FAKE16-NEXT: s_lshr_b32 s7, s73, 16
+; GFX11-FAKE16-NEXT: s_lshr_b32 s13, s73, 16
+; GFX11-FAKE16-NEXT: v_writelane_b32 v42, s10, 22
+; GFX11-FAKE16-NEXT: s_lshr_b32 s10, s22, 16
; GFX11-FAKE16-NEXT: s_lshr_b32 s50, s73, 8
-; GFX11-FAKE16-NEXT: v_writelane_b32 v43, s4, 25
-; GFX11-FAKE16-NEXT: s_lshr_b32 s4, s22, 8
-; GFX11-FAKE16-NEXT: s_lshr_b32 s53, s72, 16
-; GFX11-FAKE16-NEXT: s_lshr_b32 s52, s72, 8
+; GFX11-FAKE16-NEXT: s_lshr_b32 s52, s72, 16
+; GFX11-FAKE16-NEXT: s_lshr_b32 s51, s72, 8
+; GFX11-FAKE16-NEXT: v_writelane_b32 v42, s10, 23
+; GFX11-FAKE16-NEXT: s_lshr_b32 s10, s22, 8
; GFX11-FAKE16-NEXT: s_lshr_b32 s34, s29, 24
-; GFX11-FAKE16-NEXT: v_writelane_b32 v43, s4, 26
-; GFX11-FAKE16-NEXT: s_lshr_b32 s4, s21, 24
-; GFX11-FAKE16-NEXT: s_lshr_b32 s6, s29, 16
+; GFX11-FAKE16-NEXT: s_lshr_b32 s12, s29, 16
; GFX11-FAKE16-NEXT: s_lshr_b32 s35, s29, 8
-; GFX11-FAKE16-NEXT: s_lshr_b32 s37, s28, 16
-; GFX11-FAKE16-NEXT: v_writelane_b32 v43, s4, 27
-; GFX11-FAKE16-NEXT: s_lshr_b32 s4, s21, 16
-; GFX11-FAKE16-NEXT: s_lshr_b32 s49, s28, 8
-; GFX11-FAKE16-NEXT: s_lshr_b64 s[14:15], s[16:17], 24
-; GFX11-FAKE16-NEXT: s_lshr_b64 s[40:41], s[2:3], 24
-; GFX11-FAKE16-NEXT: v_writelane_b32 v43, s4, 11
-; GFX11-FAKE16-NEXT: s_lshr_b32 s4, s21, 8
-; GFX11-FAKE16-NEXT: s_lshr_b64 s[42:43], s[0:1], 24
-; GFX11-FAKE16-NEXT: s_lshr_b64 s[74:75], s[44:45], 24
-; GFX11-FAKE16-NEXT: s_lshr_b64 s[76:77], s[46:47], 24
-; GFX11-FAKE16-NEXT: v_writelane_b32 v43, s4, 28
-; GFX11-FAKE16-NEXT: s_lshr_b32 s4, s20, 16
-; GFX11-FAKE16-NEXT: s_lshr_b64 s[78:79], s[56:57], 24
+; GFX11-FAKE16-NEXT: v_writelane_b32 v42, s10, 24
+; GFX11-FAKE16-NEXT: s_lshr_b32 s10, s21, 24
+; GFX11-FAKE16-NEXT: s_lshr_b32 s49, s28, 16
+; GFX11-FAKE16-NEXT: s_lshr_b32 s37, s28, 8
+; GFX11-FAKE16-NEXT: s_lshr_b64 s[100:101], s[26:27], 24
+; GFX11-FAKE16-NEXT: v_writelane_b32 v42, s10, 25
+; GFX11-FAKE16-NEXT: s_lshr_b32 s10, s21, 16
+; GFX11-FAKE16-NEXT: s_lshr_b64 s[102:103], s[24:25], 24
+; GFX11-FAKE16-NEXT: s_lshr_b64 s[56:57], s[0:1], 24
+; GFX11-FAKE16-NEXT: s_lshr_b64 s[74:75], s[4:5], 24
+; GFX11-FAKE16-NEXT: v_writelane_b32 v42, s10, 9
+; GFX11-FAKE16-NEXT: s_lshr_b32 s10, s21, 8
+; GFX11-FAKE16-NEXT: s_lshr_b64 s[76:77], s[6:7], 24
+; GFX11-FAKE16-NEXT: s_lshr_b64 s[78:79], s[8:9], 24
; GFX11-FAKE16-NEXT: s_lshr_b64 s[88:89], s[58:59], 24
+; GFX11-FAKE16-NEXT: v_writelane_b32 v42, s10, 26
+; GFX11-FAKE16-NEXT: s_lshr_b32 s10, s20, 16
; GFX11-FAKE16-NEXT: s_lshr_b64 s[90:91], s[60:61], 24
-; GFX11-FAKE16-NEXT: v_writelane_b32 v43, s4, 29
-; GFX11-FAKE16-NEXT: s_lshr_b32 s4, s20, 8
; GFX11-FAKE16-NEXT: s_lshr_b64 s[92:93], s[62:63], 24
; GFX11-FAKE16-NEXT: s_lshr_b64 s[94:95], s[72:73], 24
+; GFX11-FAKE16-NEXT: v_writelane_b32 v42, s10, 27
+; GFX11-FAKE16-NEXT: s_lshr_b32 s10, s20, 8
; GFX11-FAKE16-NEXT: s_lshr_b64 s[30:31], s[28:29], 24
-; GFX11-FAKE16-NEXT: v_writelane_b32 v43, s4, 30
-; GFX11-FAKE16-NEXT: s_lshr_b32 s4, s19, 24
+; GFX11-FAKE16-NEXT: v_writelane_b32 v42, s10, 28
+; GFX11-FAKE16-NEXT: s_lshr_b32 s10, s19, 24
; GFX11-FAKE16-NEXT: s_delay_alu instid0(SALU_CYCLE_1) | instskip(SKIP_1) | instid1(SALU_CYCLE_1)
-; GFX11-FAKE16-NEXT: v_writelane_b32 v43, s4, 31
-; GFX11-FAKE16-NEXT: s_lshr_b32 s4, s19, 16
-; GFX11-FAKE16-NEXT: v_writelane_b32 v43, s4, 10
-; GFX11-FAKE16-NEXT: s_lshr_b32 s4, s19, 8
+; GFX11-FAKE16-NEXT: v_writelane_b32 v42, s10, 29
+; GFX11-FAKE16-NEXT: s_lshr_b32 s10, s19, 16
+; GFX11-FAKE16-NEXT: v_writelane_b32 v42, s10, 8
+; GFX11-FAKE16-NEXT: s_lshr_b32 s10, s19, 8
; GFX11-FAKE16-NEXT: s_delay_alu instid0(SALU_CYCLE_1) | instskip(SKIP_1) | instid1(SALU_CYCLE_1)
-; GFX11-FAKE16-NEXT: v_writelane_b32 v42, s4, 0
-; GFX11-FAKE16-NEXT: s_lshr_b32 s4, s18, 16
-; GFX11-FAKE16-NEXT: v_writelane_b32 v42, s4, 1
-; GFX11-FAKE16-NEXT: s_lshr_b32 s4, s18, 8
+; GFX11-FAKE16-NEXT: v_writelane_b32 v42, s10, 30
+; GFX11-FAKE16-NEXT: s_lshr_b32 s10, s18, 16
+; GFX11-FAKE16-NEXT: v_writelane_b32 v42, s10, 31
+; GFX11-FAKE16-NEXT: s_lshr_b32 s10, s18, 8
; GFX11-FAKE16-NEXT: s_delay_alu instid0(SALU_CYCLE_1) | instskip(SKIP_1) | instid1(SALU_CYCLE_1)
-; GFX11-FAKE16-NEXT: v_writelane_b32 v42, s4, 2
-; GFX11-FAKE16-NEXT: s_lshr_b32 s4, s17, 24
-; GFX11-FAKE16-NEXT: v_writelane_b32 v42, s4, 3
-; GFX11-FAKE16-NEXT: s_lshr_b32 s4, s17, 16
+; GFX11-FAKE16-NEXT: v_writelane_b32 v43, s10, 0
+; GFX11-FAKE16-NEXT: s_lshr_b32 s10, s17, 24
+; GFX11-FAKE16-NEXT: v_writelane_b32 v43, s10, 1
+; GFX11-FAKE16-NEXT: s_lshr_b32 s10, s17, 16
; GFX11-FAKE16-NEXT: s_delay_alu instid0(SALU_CYCLE_1) | instskip(SKIP_1) | instid1(SALU_CYCLE_1)
-; GFX11-FAKE16-NEXT: v_writelane_b32 v43, s4, 9
-; GFX11-FAKE16-NEXT: s_lshr_b32 s4, s17, 8
-; GFX11-FAKE16-NEXT: v_writelane_b32 v42, s4, 4
-; GFX11-FAKE16-NEXT: s_lshr_b32 s4, s16, 16
+; GFX11-FAKE16-NEXT: v_writelane_b32 v42, s10, 7
+; GFX11-FAKE16-NEXT: s_lshr_b32 s10, s17, 8
+; GFX11-FAKE16-NEXT: v_writelane_b32 v43, s10, 2
+; GFX11-FAKE16-NEXT: s_lshr_b32 s10, s16, 16
; GFX11-FAKE16-NEXT: s_delay_alu instid0(SALU_CYCLE_1) | instskip(SKIP_1) | instid1(SALU_CYCLE_1)
-; GFX11-FAKE16-NEXT: v_writelane_b32 v42, s4, 5
-; GFX11-FAKE16-NEXT: s_lshr_b32 s4, s16, 8
-; GFX11-FAKE16-NEXT: v_writelane_b32 v42, s4, 6
-; GFX11-FAKE16-NEXT: s_lshr_b32 s4, s3, 24
+; GFX11-FAKE16-NEXT: v_writelane_b32 v43, s10, 3
+; GFX11-FAKE16-NEXT: s_lshr_b32 s10, s16, 8
+; GFX11-FAKE16-NEXT: v_writelane_b32 v43, s10, 4
+; GFX11-FAKE16-NEXT: s_lshr_b32 s10, s3, 24
; GFX11-FAKE16-NEXT: s_delay_alu instid0(SALU_CYCLE_1) | instskip(SKIP_1) | instid1(SALU_CYCLE_1)
-; GFX11-FAKE16-NEXT: v_writelane_b32 v42, s4, 7
-; GFX11-FAKE16-NEXT: s_lshr_b32 s4, s3, 16
-; GFX11-FAKE16-NEXT: v_writelane_b32 v43, s4, 8
-; GFX11-FAKE16-NEXT: s_lshr_b32 s4, s3, 8
-; GFX11-FAKE16-NEXT: s_delay_alu instid0(SALU_CYCLE_1) | instskip(SKIP_4) | instid1(SALU_CYCLE_1)
-; GFX11-FAKE16-NEXT: v_writelane_b32 v42, s4, 8
-; GFX11-FAKE16-NEXT: s_lshr_b32 s4, s46, 16
-; GFX11-FAKE16-NEXT: v_writelane_b32 v43, s12, 6
-; GFX11-FAKE16-NEXT: v_writelane_b32 v43, s13, 7
-; GFX11-FAKE16-NEXT: s_lshr_b64 s[12:13], s[24:25], 24
-; GFX11-FAKE16-NEXT: v_writelane_b32 v43, s12, 4
-; GFX11-FAKE16-NEXT: v_writelane_b32 v43, s13, 5
-; GFX11-FAKE16-NEXT: s_lshr_b64 s[12:13], s[22:23], 24
-; GFX11-FAKE16-NEXT: s_delay_alu instid0(SALU_CYCLE_1) | instskip(SKIP_2) | instid1(SALU_CYCLE_1)
-; GFX11-FAKE16-NEXT: v_writelane_b32 v43, s12, 2
-; GFX11-FAKE16-NEXT: v_writelane_b32 v43, s13, 3
-; GFX11-FAKE16-NEXT: s_lshr_b64 s[12:13], s[20:21], 24
-; GFX11-FAKE16-NEXT: v_writelane_b32 v43, s12, 0
-; GFX11-FAKE16-NEXT: v_writelane_b32 v43, s13, 1
-; GFX11-FAKE16-NEXT: s_lshr_b64 s[12:13], s[18:19], 24
+; GFX11-FAKE16-NEXT: v_writelane_b32 v43, s10, 5
+; GFX11-FAKE16-NEXT: s_lshr_b32 s10, s3, 16
+; GFX11-FAKE16-NEXT: v_writelane_b32 v42, s10, 6
+; GFX11-FAKE16-NEXT: s_lshr_b32 s10, s3, 8
+; GFX11-FAKE16-NEXT: s_delay_alu instid0(SALU_CYCLE_1)
+; GFX11-FAKE16-NEXT: v_writelane_b32 v43, s10, 7
+; GFX11-FAKE16-NEXT: s_lshr_b32 s10, s2, 16
+; GFX11-FAKE16-NEXT: v_writelane_b32 v42, s42, 4
+; GFX11-FAKE16-NEXT: v_writelane_b32 v43, s10, 6
+; GFX11-FAKE16-NEXT: s_lshr_b32 s10, s2, 8
+; GFX11-FAKE16-NEXT: v_writelane_b32 v42, s43, 5
+; GFX11-FAKE16-NEXT: s_lshr_b64 s[42:43], s[20:21], 24
+; GFX11-FAKE16-NEXT: v_writelane_b32 v43, s10, 8
+; GFX11-FAKE16-NEXT: s_lshr_b32 s10, s1, 24
+; GFX11-FAKE16-NEXT: v_writelane_b32 v42, s42, 2
+; GFX11-FAKE16-NEXT: v_writelane_b32 v43, s10, 9
+; GFX11-FAKE16-NEXT: s_lshr_b32 s10, s6, 16
+; GFX11-FAKE16-NEXT: v_writelane_b32 v42, s43, 3
+; GFX11-FAKE16-NEXT: s_lshr_b64 s[42:43], s[18:19], 24
+; GFX11-FAKE16-NEXT: v_writelane_b32 v42, s46, 0
+; GFX11-FAKE16-NEXT: v_writelane_b32 v42, s47, 1
+; GFX11-FAKE16-NEXT: s_lshr_b64 s[46:47], s[2:3], 24
; GFX11-FAKE16-NEXT: s_and_not1_b32 vcc_lo, exec_lo, vcc_hi
; GFX11-FAKE16-NEXT: s_cbranch_vccnz .LBB91_4
; GFX11-FAKE16-NEXT: .LBB91_2: ; %cmp.true
-; GFX11-FAKE16-NEXT: s_and_b32 s4, s29, 0xffff0000
-; GFX11-FAKE16-NEXT: s_and_b32 s14, s47, 0xffff0000
-; GFX11-FAKE16-NEXT: v_add_f32_e64 v1, 0x40c00000, s4
-; GFX11-FAKE16-NEXT: s_and_b32 s4, s1, 0xffff0000
-; GFX11-FAKE16-NEXT: s_lshl_b32 s15, s47, 16
-; GFX11-FAKE16-NEXT: v_add_f32_e64 v6, 0x40c00000, s4
-; GFX11-FAKE16-NEXT: s_lshl_b32 s6, s29, 16
+; GFX11-FAKE16-NEXT: s_and_b32 s10, s29, 0xffff0000
+; GFX11-FAKE16-NEXT: s_and_b32 s13, s9, 0xffff0000
+; GFX11-FAKE16-NEXT: v_add_f32_e64 v1, 0x40c00000, s10
+; GFX11-FAKE16-NEXT: s_lshl_b32 s10, s9, 16
+; GFX11-FAKE16-NEXT: s_and_b32 s9, s1, 0xffff0000
+; GFX11-FAKE16-NEXT: s_lshl_b32 s11, s29, 16
+; GFX11-FAKE16-NEXT: v_add_f32_e64 v6, 0x40c00000, s9
+; GFX11-FAKE16-NEXT: s_and_b32 s76, s28, 0xffff0000
+; GFX11-FAKE16-NEXT: s_lshl_b32 s77, s28, 16
; GFX11-FAKE16-NEXT: v_bfe_u32 v3, v1, 16, 1
-; GFX11-FAKE16-NEXT: v_add_f32_e64 v2, 0x40c00000, s6
-; GFX11-FAKE16-NEXT: s_and_b32 s8, s45, 0xffff0000
-; GFX11-FAKE16-NEXT: v_readfirstlane_b32 s47, v6
-; GFX11-FAKE16-NEXT: v_cmp_u_f32_e32 vcc_lo, v6, v6
+; GFX11-FAKE16-NEXT: s_and_b32 s28, s58, 0xffff0000
+; GFX11-FAKE16-NEXT: s_lshl_b32 s29, s58, 16
+; GFX11-FAKE16-NEXT: v_readfirstlane_b32 s58, v6
+; GFX11-FAKE16-NEXT: v_add_f32_e64 v2, 0x40c00000, s11
; GFX11-FAKE16-NEXT: v_add_nc_u32_e32 v3, v3, v1
-; GFX11-FAKE16-NEXT: s_lshl_b32 s7, s45, 16
-; GFX11-FAKE16-NEXT: s_and_b32 s78, s28, 0xffff0000
-; GFX11-FAKE16-NEXT: s_bfe_u32 s6, s47, 0x10010
-; GFX11-FAKE16-NEXT: s_lshl_b32 s79, s28, 16
-; GFX11-FAKE16-NEXT: s_add_i32 s45, s6, s47
-; GFX11-FAKE16-NEXT: s_and_b32 s5, s73, 0xffff0000
-; GFX11-FAKE16-NEXT: s_lshl_b32 s77, s73, 16
-; GFX11-FAKE16-NEXT: s_and_b32 s75, s72, 0xffff0000
-; GFX11-FAKE16-NEXT: s_lshl_b32 s76, s72, 16
-; GFX11-FAKE16-NEXT: s_and_b32 s11, s63, 0xffff0000
-; GFX11-FAKE16-NEXT: s_lshl_b32 s74, s63, 16
-; GFX11-FAKE16-NEXT: s_and_b32 s72, s62, 0xffff0000
-; GFX11-FAKE16-NEXT: s_lshl_b32 s73, s62, 16
-; GFX11-FAKE16-NEXT: s_and_b32 s63, s61, 0xffff0000
-; GFX11-FAKE16-NEXT: s_lshl_b32 s62, s61, 16
-; GFX11-FAKE16-NEXT: s_and_b32 s61, s60, 0xffff0000
-; GFX11-FAKE16-NEXT: s_lshl_b32 s60, s60, 16
+; GFX11-FAKE16-NEXT: s_and_b32 s12, s6, 0xffff0000
+; GFX11-FAKE16-NEXT: s_lshl_b32 s9, s6, 16
+; GFX11-FAKE16-NEXT: s_bfe_u32 s6, s58, 0x10010
+; GFX11-FAKE16-NEXT: v_cmp_u_f32_e32 vcc_lo, v6, v6
; GFX11-FAKE16-NEXT: s_and_b32 s41, s59, 0xffff0000
; GFX11-FAKE16-NEXT: s_lshl_b32 s40, s59, 16
-; GFX11-FAKE16-NEXT: s_and_b32 s28, s58, 0xffff0000
-; GFX11-FAKE16-NEXT: s_lshl_b32 s29, s58, 16
-; GFX11-FAKE16-NEXT: s_and_b32 s13, s57, 0xffff0000
-; GFX11-FAKE16-NEXT: s_lshl_b32 s10, s57, 16
-; GFX11-FAKE16-NEXT: s_and_b32 s42, s56, 0xffff0000
-; GFX11-FAKE16-NEXT: s_lshl_b32 s43, s56, 16
-; GFX11-FAKE16-NEXT: s_and_b32 s12, s46, 0xffff0000
-; GFX11-FAKE16-NEXT: s_lshl_b32 s9, s46, 16
-; GFX11-FAKE16-NEXT: s_and_b32 s4, s44, 0xffff0000
-; GFX11-FAKE16-NEXT: s_lshl_b32 s6, s44, 16
-; GFX11-FAKE16-NEXT: s_addk_i32 s45, 0x7fff
-; GFX11-FAKE16-NEXT: s_bitset1_b32 s47, 22
+; GFX11-FAKE16-NEXT: s_add_i32 s59, s6, s58
+; GFX11-FAKE16-NEXT: s_and_b32 s74, s73, 0xffff0000
+; GFX11-FAKE16-NEXT: s_lshl_b32 s75, s73, 16
+; GFX11-FAKE16-NEXT: s_and_b32 s73, s72, 0xffff0000
+; GFX11-FAKE16-NEXT: s_lshl_b32 s72, s72, 16
+; GFX11-FAKE16-NEXT: s_and_b32 s11, s63, 0xffff0000
+; GFX11-FAKE16-NEXT: s_lshl_b32 s63, s63, 16
+; GFX11-FAKE16-NEXT: s_and_b32 s56, s62, 0xffff0000
+; GFX11-FAKE16-NEXT: s_lshl_b32 s57, s62, 16
+; GFX11-FAKE16-NEXT: s_and_b32 s47, s61, 0xffff0000
+; GFX11-FAKE16-NEXT: s_lshl_b32 s45, s61, 16
+; GFX11-FAKE16-NEXT: s_and_b32 s44, s60, 0xffff0000
+; GFX11-FAKE16-NEXT: s_lshl_b32 s46, s60, 16
+; GFX11-FAKE16-NEXT: s_and_b32 s42, s8, 0xffff0000
+; GFX11-FAKE16-NEXT: s_lshl_b32 s43, s8, 16
+; GFX11-FAKE16-NEXT: s_and_b32 s14, s7, 0xffff0000
+; GFX11-FAKE16-NEXT: s_lshl_b32 s15, s7, 16
+; GFX11-FAKE16-NEXT: s_and_b32 s8, s5, 0xffff0000
+; GFX11-FAKE16-NEXT: s_lshl_b32 s7, s5, 16
+; GFX11-FAKE16-NEXT: s_and_b32 s5, s4, 0xffff0000
+; GFX11-FAKE16-NEXT: s_lshl_b32 s6, s4, 16
+; GFX11-FAKE16-NEXT: s_addk_i32 s59, 0x7fff
+; GFX11-FAKE16-NEXT: s_bitset1_b32 s58, 22
; GFX11-FAKE16-NEXT: v_bfe_u32 v4, v2, 16, 1
; GFX11-FAKE16-NEXT: v_or_b32_e32 v5, 0x400000, v1
; GFX11-FAKE16-NEXT: v_add_nc_u32_e32 v3, 0x7fff, v3
-; GFX11-FAKE16-NEXT: s_and_b32 s44, vcc_lo, exec_lo
+; GFX11-FAKE16-NEXT: s_and_b32 s4, vcc_lo, exec_lo
; GFX11-FAKE16-NEXT: v_cmp_u_f32_e32 vcc_lo, v1, v1
-; GFX11-FAKE16-NEXT: s_cselect_b32 s44, s47, s45
+; GFX11-FAKE16-NEXT: s_cselect_b32 s4, s58, s59
; GFX11-FAKE16-NEXT: s_lshl_b32 s1, s1, 16
; GFX11-FAKE16-NEXT: v_add_nc_u32_e32 v4, v4, v2
-; GFX11-FAKE16-NEXT: s_lshr_b32 s58, s44, 16
+; GFX11-FAKE16-NEXT: s_lshr_b32 s61, s4, 16
; GFX11-FAKE16-NEXT: v_cndmask_b32_e32 v1, v3, v5, vcc_lo
; GFX11-FAKE16-NEXT: v_add_f32_e64 v3, 0x40c00000, s1
; GFX11-FAKE16-NEXT: v_or_b32_e32 v5, 0x400000, v2
; GFX11-FAKE16-NEXT: v_add_nc_u32_e32 v4, 0x7fff, v4
; GFX11-FAKE16-NEXT: v_cmp_u_f32_e32 vcc_lo, v2, v2
-; GFX11-FAKE16-NEXT: v_add_f32_e64 v6, 0x40c00000, s78
+; GFX11-FAKE16-NEXT: v_add_f32_e64 v6, 0x40c00000, s76
; GFX11-FAKE16-NEXT: v_readfirstlane_b32 s1, v3
-; GFX11-FAKE16-NEXT: v_add_f32_e64 v7, 0x40c00000, s79
-; GFX11-FAKE16-NEXT: v_lshrrev_b32_e32 v22, 16, v1
+; GFX11-FAKE16-NEXT: v_add_f32_e64 v7, 0x40c00000, s77
+; GFX11-FAKE16-NEXT: v_lshrrev_b32_e32 v23, 16, v1
; GFX11-FAKE16-NEXT: v_cndmask_b32_e32 v2, v4, v5, vcc_lo
; GFX11-FAKE16-NEXT: v_cmp_u_f32_e32 vcc_lo, v3, v3
-; GFX11-FAKE16-NEXT: s_bfe_u32 s45, s1, 0x10010
+; GFX11-FAKE16-NEXT: s_bfe_u32 s58, s1, 0x10010
; GFX11-FAKE16-NEXT: v_bfe_u32 v4, v6, 16, 1
-; GFX11-FAKE16-NEXT: s_add_i32 s45, s45, s1
+; GFX11-FAKE16-NEXT: s_add_i32 s58, s58, s1
; GFX11-FAKE16-NEXT: s_bitset1_b32 s1, 22
-; GFX11-FAKE16-NEXT: s_addk_i32 s45, 0x7fff
-; GFX11-FAKE16-NEXT: s_and_b32 s44, vcc_lo, exec_lo
-; GFX11-FAKE16-NEXT: s_cselect_b32 s1, s1, s45
-; GFX11-FAKE16-NEXT: s_and_b32 s44, s0, 0xffff0000
-; GFX11-FAKE16-NEXT: v_lshrrev_b32_e32 v21, 16, v2
-; GFX11-FAKE16-NEXT: v_add_f32_e64 v2, 0x40c00000, s44
+; GFX11-FAKE16-NEXT: s_addk_i32 s58, 0x7fff
+; GFX11-FAKE16-NEXT: s_and_b32 s4, vcc_lo, exec_lo
+; GFX11-FAKE16-NEXT: s_cselect_b32 s1, s1, s58
+; GFX11-FAKE16-NEXT: s_and_b32 s4, s0, 0xffff0000
+; GFX11-FAKE16-NEXT: v_lshrrev_b32_e32 v22, 16, v2
+; GFX11-FAKE16-NEXT: v_add_f32_e64 v2, 0x40c00000, s4
; GFX11-FAKE16-NEXT: v_bfe_u32 v5, v7, 16, 1
; GFX11-FAKE16-NEXT: v_add_nc_u32_e32 v3, v4, v6
; GFX11-FAKE16-NEXT: s_lshr_b32 s1, s1, 16
; GFX11-FAKE16-NEXT: v_or_b32_e32 v8, 0x400000, v7
-; GFX11-FAKE16-NEXT: v_readfirstlane_b32 s44, v2
+; GFX11-FAKE16-NEXT: v_readfirstlane_b32 s4, v2
; GFX11-FAKE16-NEXT: v_cmp_u_f32_e32 vcc_lo, v2, v2
; GFX11-FAKE16-NEXT: v_add_nc_u32_e32 v3, 0x7fff, v3
-; GFX11-FAKE16-NEXT: s_bfe_u32 s45, s44, 0x10010
+; GFX11-FAKE16-NEXT: s_bfe_u32 s58, s4, 0x10010
; GFX11-FAKE16-NEXT: s_delay_alu instid0(SALU_CYCLE_1)
-; GFX11-FAKE16-NEXT: s_add_i32 s45, s45, s44
-; GFX11-FAKE16-NEXT: s_bitset1_b32 s44, 22
-; GFX11-FAKE16-NEXT: s_addk_i32 s45, 0x7fff
-; GFX11-FAKE16-NEXT: s_and_b32 s46, vcc_lo, exec_lo
+; GFX11-FAKE16-NEXT: s_add_i32 s58, s58, s4
+; GFX11-FAKE16-NEXT: s_bitset1_b32 s4, 22
+; GFX11-FAKE16-NEXT: s_addk_i32 s58, 0x7fff
+; GFX11-FAKE16-NEXT: s_and_b32 s59, vcc_lo, exec_lo
; GFX11-FAKE16-NEXT: v_cmp_u_f32_e32 vcc_lo, v6, v6
; GFX11-FAKE16-NEXT: v_add_nc_u32_e32 v1, v5, v7
; GFX11-FAKE16-NEXT: v_or_b32_e32 v5, 0x400000, v6
-; GFX11-FAKE16-NEXT: s_cselect_b32 s44, s44, s45
+; GFX11-FAKE16-NEXT: s_cselect_b32 s4, s4, s58
; GFX11-FAKE16-NEXT: s_lshl_b32 s0, s0, 16
-; GFX11-FAKE16-NEXT: s_delay_alu instid0(VALU_DEP_1)
+; GFX11-FAKE16-NEXT: s_lshr_b32 s4, s4, 16
+; GFX11-FAKE16-NEXT: v_add_f32_e64 v6, 0x40c00000, s75
; GFX11-FAKE16-NEXT: v_cndmask_b32_e32 v2, v3, v5, vcc_lo
; GFX11-FAKE16-NEXT: v_add_f32_e64 v3, 0x40c00000, s0
; GFX11-FAKE16-NEXT: v_add_nc_u32_e32 v1, 0x7fff, v1
; GFX11-FAKE16-NEXT: v_cmp_u_f32_e32 vcc_lo, v7, v7
-; GFX11-FAKE16-NEXT: v_and_b32_e32 v4, 0xffff, v21
-; GFX11-FAKE16-NEXT: v_add_f32_e64 v5, 0x40c00000, s5
+; GFX11-FAKE16-NEXT: v_and_b32_e32 v4, 0xffff, v22
+; GFX11-FAKE16-NEXT: v_add_f32_e64 v5, 0x40c00000, s74
; GFX11-FAKE16-NEXT: v_readfirstlane_b32 s0, v3
-; GFX11-FAKE16-NEXT: v_lshrrev_b32_e32 v2, 16, v2
+; GFX11-FAKE16-NEXT: v_lshrrev_b32_e32 v7, 16, v2
; GFX11-FAKE16-NEXT: v_cndmask_b32_e32 v1, v1, v8, vcc_lo
; GFX11-FAKE16-NEXT: v_cmp_u_f32_e32 vcc_lo, v3, v3
-; GFX11-FAKE16-NEXT: v_add_f32_e64 v8, 0x40c00000, s77
-; GFX11-FAKE16-NEXT: s_bfe_u32 s5, s0, 0x10010
-; GFX11-FAKE16-NEXT: v_lshl_or_b32 v7, v22, 16, v4
-; GFX11-FAKE16-NEXT: s_add_i32 s45, s5, s0
-; GFX11-FAKE16-NEXT: s_lshr_b32 s5, s44, 16
-; GFX11-FAKE16-NEXT: s_addk_i32 s45, 0x7fff
+; GFX11-FAKE16-NEXT: v_bfe_u32 v8, v6, 16, 1
+; GFX11-FAKE16-NEXT: s_bfe_u32 s58, s0, 0x10010
+; GFX11-FAKE16-NEXT: v_lshl_or_b32 v2, v23, 16, v4
+; GFX11-FAKE16-NEXT: s_add_i32 s58, s58, s0
; GFX11-FAKE16-NEXT: s_bitset1_b32 s0, 22
-; GFX11-FAKE16-NEXT: s_and_b32 s44, vcc_lo, exec_lo
-; GFX11-FAKE16-NEXT: s_cselect_b32 s0, s0, s45
-; GFX11-FAKE16-NEXT: s_and_b32 s44, s3, 0xffff0000
-; GFX11-FAKE16-NEXT: v_lshrrev_b32_e32 v23, 16, v1
-; GFX11-FAKE16-NEXT: v_add_f32_e64 v9, 0x40c00000, s44
-; GFX11-FAKE16-NEXT: v_bfe_u32 v6, v8, 16, 1
+; GFX11-FAKE16-NEXT: s_addk_i32 s58, 0x7fff
+; GFX11-FAKE16-NEXT: s_and_b32 s59, vcc_lo, exec_lo
+; GFX11-FAKE16-NEXT: s_cselect_b32 s0, s0, s58
+; GFX11-FAKE16-NEXT: s_and_b32 s58, s3, 0xffff0000
+; GFX11-FAKE16-NEXT: v_lshrrev_b32_e32 v24, 16, v1
+; GFX11-FAKE16-NEXT: v_add_f32_e64 v9, 0x40c00000, s58
; GFX11-FAKE16-NEXT: v_bfe_u32 v1, v5, 16, 1
+; GFX11-FAKE16-NEXT: v_add_nc_u32_e32 v4, v8, v6
; GFX11-FAKE16-NEXT: s_lshr_b32 s0, s0, 16
-; GFX11-FAKE16-NEXT: v_and_b32_e32 v3, 0xffff, v23
-; GFX11-FAKE16-NEXT: v_readfirstlane_b32 s44, v9
+; GFX11-FAKE16-NEXT: v_and_b32_e32 v3, 0xffff, v24
+; GFX11-FAKE16-NEXT: v_readfirstlane_b32 s58, v9
; GFX11-FAKE16-NEXT: v_cmp_u_f32_e32 vcc_lo, v9, v9
-; GFX11-FAKE16-NEXT: v_add_nc_u32_e32 v4, v6, v8
-; GFX11-FAKE16-NEXT: v_add_nc_u32_e32 v1, v1, v5
-; GFX11-FAKE16-NEXT: v_lshl_or_b32 v6, v2, 16, v3
-; GFX11-FAKE16-NEXT: s_bfe_u32 s45, s44, 0x10010
-; GFX11-FAKE16-NEXT: v_or_b32_e32 v2, 0x400000, v5
-; GFX11-FAKE16-NEXT: s_add_i32 s45, s45, s44
-; GFX11-FAKE16-NEXT: s_bitset1_b32 s44, 22
-; GFX11-FAKE16-NEXT: s_addk_i32 s45, 0x7fff
-; GFX11-FAKE16-NEXT: s_and_b32 s46, vcc_lo, exec_lo
-; GFX11-FAKE16-NEXT: v_add_nc_u32_e32 v3, 0x7fff, v4
-; GFX11-FAKE16-NEXT: v_or_b32_e32 v4, 0x400000, v8
-; GFX11-FAKE16-NEXT: s_cselect_b32 s44, s44, s45
+; GFX11-FAKE16-NEXT: v_add_nc_u32_e32 v10, v1, v5
+; GFX11-FAKE16-NEXT: v_add_nc_u32_e32 v4, 0x7fff, v4
+; GFX11-FAKE16-NEXT: v_lshl_or_b32 v1, v7, 16, v3
+; GFX11-FAKE16-NEXT: s_bfe_u32 s59, s58, 0x10010
+; GFX11-FAKE16-NEXT: v_or_b32_e32 v8, 0x400000, v6
+; GFX11-FAKE16-NEXT: s_add_i32 s59, s59, s58
+; GFX11-FAKE16-NEXT: s_bitset1_b32 s58, 22
+; GFX11-FAKE16-NEXT: s_addk_i32 s59, 0x7fff
+; GFX11-FAKE16-NEXT: s_and_b32 s60, vcc_lo, exec_lo
+; GFX11-FAKE16-NEXT: s_cselect_b32 s58, s58, s59
; GFX11-FAKE16-NEXT: s_lshl_b32 s3, s3, 16
-; GFX11-FAKE16-NEXT: v_cmp_u_f32_e32 vcc_lo, v8, v8
+; GFX11-FAKE16-NEXT: v_add_nc_u32_e32 v3, 0x7fff, v10
; GFX11-FAKE16-NEXT: v_add_f32_e64 v10, 0x40c00000, s3
-; GFX11-FAKE16-NEXT: v_add_nc_u32_e32 v1, 0x7fff, v1
-; GFX11-FAKE16-NEXT: v_add_f32_e64 v9, 0x40c00000, s76
-; GFX11-FAKE16-NEXT: s_lshr_b32 s59, s44, 16
-; GFX11-FAKE16-NEXT: v_cndmask_b32_e32 v3, v3, v4, vcc_lo
-; GFX11-FAKE16-NEXT: v_add_f32_e64 v4, 0x40c00000, s75
-; GFX11-FAKE16-NEXT: v_cmp_u_f32_e32 vcc_lo, v5, v5
+; GFX11-FAKE16-NEXT: v_cmp_u_f32_e32 vcc_lo, v6, v6
+; GFX11-FAKE16-NEXT: v_or_b32_e32 v7, 0x400000, v5
+; GFX11-FAKE16-NEXT: v_add_f32_e64 v9, 0x40c00000, s72
+; GFX11-FAKE16-NEXT: s_lshr_b32 s62, s58, 16
; GFX11-FAKE16-NEXT: v_readfirstlane_b32 s3, v10
+; GFX11-FAKE16-NEXT: v_cndmask_b32_e32 v4, v4, v8, vcc_lo
+; GFX11-FAKE16-NEXT: v_cmp_u_f32_e32 vcc_lo, v5, v5
; GFX11-FAKE16-NEXT: v_bfe_u32 v8, v9, 16, 1
-; GFX11-FAKE16-NEXT: v_lshrrev_b32_e32 v87, 24, v7
-; GFX11-FAKE16-NEXT: v_lshrrev_b32_e32 v96, 16, v6
-; GFX11-FAKE16-NEXT: v_cndmask_b32_e32 v1, v1, v2, vcc_lo
-; GFX11-FAKE16-NEXT: v_bfe_u32 v2, v4, 16, 1
-; GFX11-FAKE16-NEXT: s_bfe_u32 s45, s3, 0x10010
+; GFX11-FAKE16-NEXT: v_add_f32_e64 v6, 0x40c00000, s73
+; GFX11-FAKE16-NEXT: s_bfe_u32 s59, s3, 0x10010
+; GFX11-FAKE16-NEXT: v_lshrrev_b32_e32 v25, 16, v4
+; GFX11-FAKE16-NEXT: v_cndmask_b32_e32 v3, v3, v7, vcc_lo
; GFX11-FAKE16-NEXT: v_cmp_u_f32_e32 vcc_lo, v10, v10
-; GFX11-FAKE16-NEXT: s_add_i32 s45, s45, s3
+; GFX11-FAKE16-NEXT: s_add_i32 s59, s59, s3
; GFX11-FAKE16-NEXT: s_bitset1_b32 s3, 22
-; GFX11-FAKE16-NEXT: v_add_nc_u32_e32 v2, v2, v4
-; GFX11-FAKE16-NEXT: s_addk_i32 s45, 0x7fff
-; GFX11-FAKE16-NEXT: s_and_b32 s44, vcc_lo, exec_lo
-; GFX11-FAKE16-NEXT: s_cselect_b32 s3, s3, s45
-; GFX11-FAKE16-NEXT: s_and_b32 s44, s2, 0xffff0000
-; GFX11-FAKE16-NEXT: v_lshrrev_b32_e32 v25, 16, v1
-; GFX11-FAKE16-NEXT: v_add_f32_e64 v1, 0x40c00000, s44
-; GFX11-FAKE16-NEXT: v_add_nc_u32_e32 v2, 0x7fff, v2
-; GFX11-FAKE16-NEXT: v_lshrrev_b32_e32 v24, 16, v3
-; GFX11-FAKE16-NEXT: v_add_nc_u32_e32 v3, v8, v9
+; GFX11-FAKE16-NEXT: s_addk_i32 s59, 0x7fff
+; GFX11-FAKE16-NEXT: v_add_nc_u32_e32 v5, v8, v9
+; GFX11-FAKE16-NEXT: s_and_b32 s58, vcc_lo, exec_lo
+; GFX11-FAKE16-NEXT: s_cselect_b32 s3, s3, s59
+; GFX11-FAKE16-NEXT: s_and_b32 s58, s2, 0xffff0000
+; GFX11-FAKE16-NEXT: v_lshrrev_b32_e32 v26, 16, v3
+; GFX11-FAKE16-NEXT: v_add_f32_e64 v3, 0x40c00000, s58
+; GFX11-FAKE16-NEXT: v_add_nc_u32_e32 v5, 0x7fff, v5
; GFX11-FAKE16-NEXT: v_or_b32_e32 v8, 0x400000, v9
-; GFX11-FAKE16-NEXT: v_readfirstlane_b32 s44, v1
; GFX11-FAKE16-NEXT: v_cmp_u_f32_e32 vcc_lo, v9, v9
+; GFX11-FAKE16-NEXT: v_bfe_u32 v4, v6, 16, 1
+; GFX11-FAKE16-NEXT: v_readfirstlane_b32 s58, v3
; GFX11-FAKE16-NEXT: s_lshr_b32 s3, s3, 16
-; GFX11-FAKE16-NEXT: v_add_nc_u32_e32 v3, 0x7fff, v3
-; GFX11-FAKE16-NEXT: v_or_b32_e32 v10, 0x400000, v4
-; GFX11-FAKE16-NEXT: s_bfe_u32 s45, s44, 0x10010
-; GFX11-FAKE16-NEXT: v_and_b32_e32 v5, 0xffff, v24
-; GFX11-FAKE16-NEXT: s_add_i32 s45, s45, s44
-; GFX11-FAKE16-NEXT: v_cndmask_b32_e32 v3, v3, v8, vcc_lo
-; GFX11-FAKE16-NEXT: v_cmp_u_f32_e32 vcc_lo, v1, v1
-; GFX11-FAKE16-NEXT: s_addk_i32 s45, 0x7fff
-; GFX11-FAKE16-NEXT: s_bitset1_b32 s44, 22
-; GFX11-FAKE16-NEXT: v_add_f32_e64 v1, 0x40c00000, s74
-; GFX11-FAKE16-NEXT: v_lshl_or_b32 v14, v25, 16, v5
-; GFX11-FAKE16-NEXT: s_and_b32 s46, vcc_lo, exec_lo
-; GFX11-FAKE16-NEXT: s_cselect_b32 s44, s44, s45
+; GFX11-FAKE16-NEXT: v_or_b32_e32 v10, 0x400000, v6
+; GFX11-FAKE16-NEXT: v_cndmask_b32_e32 v5, v5, v8, vcc_lo
+; GFX11-FAKE16-NEXT: v_cmp_u_f32_e32 vcc_lo, v3, v3
+; GFX11-FAKE16-NEXT: s_bfe_u32 s59, s58, 0x10010
+; GFX11-FAKE16-NEXT: v_add_nc_u32_e32 v4, v4, v6
+; GFX11-FAKE16-NEXT: s_add_i32 s59, s59, s58
+; GFX11-FAKE16-NEXT: s_bitset1_b32 s58, 22
+; GFX11-FAKE16-NEXT: s_addk_i32 s59, 0x7fff
+; GFX11-FAKE16-NEXT: s_and_b32 s60, vcc_lo, exec_lo
+; GFX11-FAKE16-NEXT: s_cselect_b32 s58, s58, s59
; GFX11-FAKE16-NEXT: s_lshl_b32 s2, s2, 16
-; GFX11-FAKE16-NEXT: v_cmp_u_f32_e32 vcc_lo, v4, v4
+; GFX11-FAKE16-NEXT: v_add_nc_u32_e32 v4, 0x7fff, v4
; GFX11-FAKE16-NEXT: v_add_f32_e64 v8, 0x40c00000, s2
-; GFX11-FAKE16-NEXT: v_bfe_u32 v4, v1, 16, 1
-; GFX11-FAKE16-NEXT: v_lshrrev_b32_e32 v85, 24, v14
-; GFX11-FAKE16-NEXT: v_cndmask_b32_e32 v2, v2, v10, vcc_lo
-; GFX11-FAKE16-NEXT: s_delay_alu instid0(VALU_DEP_4) | instskip(NEXT) | instid1(VALU_DEP_4)
+; GFX11-FAKE16-NEXT: v_cmp_u_f32_e32 vcc_lo, v6, v6
+; GFX11-FAKE16-NEXT: v_add_f32_e64 v3, 0x40c00000, s63
+; GFX11-FAKE16-NEXT: v_lshrrev_b32_e32 v27, 16, v5
+; GFX11-FAKE16-NEXT: v_add_f32_e64 v5, 0x40c00000, s11
; GFX11-FAKE16-NEXT: v_readfirstlane_b32 s2, v8
-; GFX11-FAKE16-NEXT: v_add_nc_u32_e32 v4, v4, v1
-; GFX11-FAKE16-NEXT: v_lshrrev_b32_e32 v26, 16, v3
-; GFX11-FAKE16-NEXT: v_add_f32_e64 v3, 0x40c00000, s11
+; GFX11-FAKE16-NEXT: v_cndmask_b32_e32 v4, v4, v10, vcc_lo
; GFX11-FAKE16-NEXT: v_cmp_u_f32_e32 vcc_lo, v8, v8
+; GFX11-FAKE16-NEXT: v_and_b32_e32 v7, 0xffff, v25
+; GFX11-FAKE16-NEXT: v_bfe_u32 v6, v3, 16, 1
; GFX11-FAKE16-NEXT: s_bfe_u32 s11, s2, 0x10010
-; GFX11-FAKE16-NEXT: v_lshrrev_b32_e32 v2, 16, v2
-; GFX11-FAKE16-NEXT: s_add_i32 s45, s11, s2
-; GFX11-FAKE16-NEXT: s_lshr_b32 s11, s44, 16
-; GFX11-FAKE16-NEXT: s_addk_i32 s45, 0x7fff
+; GFX11-FAKE16-NEXT: v_lshrrev_b32_e32 v4, 16, v4
+; GFX11-FAKE16-NEXT: s_add_i32 s59, s11, s2
+; GFX11-FAKE16-NEXT: s_lshr_b32 s11, s58, 16
+; GFX11-FAKE16-NEXT: s_addk_i32 s59, 0x7fff
; GFX11-FAKE16-NEXT: s_bitset1_b32 s2, 22
-; GFX11-FAKE16-NEXT: s_and_b32 s44, vcc_lo, exec_lo
-; GFX11-FAKE16-NEXT: s_cselect_b32 s2, s2, s45
-; GFX11-FAKE16-NEXT: s_and_b32 s44, s17, 0xffff0000
-; GFX11-FAKE16-NEXT: v_and_b32_e32 v9, 0xffff, v26
-; GFX11-FAKE16-NEXT: v_add_f32_e64 v5, 0x40c00000, s44
-; GFX11-FAKE16-NEXT: v_cmp_u_f32_e32 vcc_lo, v1, v1
-; GFX11-FAKE16-NEXT: v_bfe_u32 v10, v3, 16, 1
+; GFX11-FAKE16-NEXT: s_and_b32 s58, vcc_lo, exec_lo
+; GFX11-FAKE16-NEXT: s_cselect_b32 s2, s2, s59
+; GFX11-FAKE16-NEXT: s_and_b32 s58, s17, 0xffff0000
+; GFX11-FAKE16-NEXT: v_and_b32_e32 v10, 0xffff, v27
+; GFX11-FAKE16-NEXT: v_add_nc_u32_e32 v6, v6, v3
+; GFX11-FAKE16-NEXT: v_lshl_or_b32 v9, v26, 16, v7
+; GFX11-FAKE16-NEXT: v_add_f32_e64 v7, 0x40c00000, s58
+; GFX11-FAKE16-NEXT: v_cmp_u_f32_e32 vcc_lo, v3, v3
+; GFX11-FAKE16-NEXT: v_lshl_or_b32 v8, v4, 16, v10
+; GFX11-FAKE16-NEXT: v_add_nc_u32_e32 v4, 0x7fff, v6
+; GFX11-FAKE16-NEXT: v_or_b32_e32 v6, 0x400000, v3
+; GFX11-FAKE16-NEXT: v_readfirstlane_b32 s58, v7
+; GFX11-FAKE16-NEXT: v_bfe_u32 v11, v5, 16, 1
; GFX11-FAKE16-NEXT: s_lshr_b32 s2, s2, 16
-; GFX11-FAKE16-NEXT: v_lshl_or_b32 v13, v2, 16, v9
-; GFX11-FAKE16-NEXT: v_add_nc_u32_e32 v2, 0x7fff, v4
-; GFX11-FAKE16-NEXT: v_or_b32_e32 v4, 0x400000, v1
-; GFX11-FAKE16-NEXT: v_readfirstlane_b32 s44, v5
-; GFX11-FAKE16-NEXT: v_add_nc_u32_e32 v8, v10, v3
-; GFX11-FAKE16-NEXT: v_or_b32_e32 v9, 0x400000, v3
-; GFX11-FAKE16-NEXT: v_lshrrev_b32_e32 v86, 16, v13
-; GFX11-FAKE16-NEXT: v_cndmask_b32_e32 v1, v2, v4, vcc_lo
-; GFX11-FAKE16-NEXT: s_bfe_u32 s45, s44, 0x10010
-; GFX11-FAKE16-NEXT: v_cmp_u_f32_e32 vcc_lo, v5, v5
-; GFX11-FAKE16-NEXT: s_add_i32 s45, s45, s44
-; GFX11-FAKE16-NEXT: s_bitset1_b32 s44, 22
-; GFX11-FAKE16-NEXT: s_addk_i32 s45, 0x7fff
-; GFX11-FAKE16-NEXT: v_add_nc_u32_e32 v8, 0x7fff, v8
-; GFX11-FAKE16-NEXT: s_and_b32 s46, vcc_lo, exec_lo
-; GFX11-FAKE16-NEXT: s_cselect_b32 s44, s44, s45
+; GFX11-FAKE16-NEXT: s_pack_ll_b32_b16 s4, s0, s4
+; GFX11-FAKE16-NEXT: v_cndmask_b32_e32 v3, v4, v6, vcc_lo
+; GFX11-FAKE16-NEXT: s_bfe_u32 s59, s58, 0x10010
+; GFX11-FAKE16-NEXT: v_cmp_u_f32_e32 vcc_lo, v7, v7
+; GFX11-FAKE16-NEXT: s_add_i32 s59, s59, s58
+; GFX11-FAKE16-NEXT: s_bitset1_b32 s58, 22
+; GFX11-FAKE16-NEXT: s_addk_i32 s59, 0x7fff
+; GFX11-FAKE16-NEXT: v_add_nc_u32_e32 v11, v11, v5
+; GFX11-FAKE16-NEXT: s_and_b32 s60, vcc_lo, exec_lo
+; GFX11-FAKE16-NEXT: s_cselect_b32 s58, s58, s59
; GFX11-FAKE16-NEXT: s_lshl_b32 s17, s17, 16
-; GFX11-FAKE16-NEXT: v_add_f32_e64 v2, 0x40c00000, s73
-; GFX11-FAKE16-NEXT: v_add_f32_e64 v4, 0x40c00000, s17
-; GFX11-FAKE16-NEXT: v_cmp_u_f32_e32 vcc_lo, v3, v3
-; GFX11-FAKE16-NEXT: v_lshrrev_b32_e32 v27, 16, v1
-; GFX11-FAKE16-NEXT: v_add_f32_e64 v1, 0x40c00000, s72
-; GFX11-FAKE16-NEXT: v_bfe_u32 v5, v2, 16, 1
-; GFX11-FAKE16-NEXT: v_readfirstlane_b32 s17, v4
-; GFX11-FAKE16-NEXT: v_cndmask_b32_e32 v3, v8, v9, vcc_lo
-; GFX11-FAKE16-NEXT: v_cmp_u_f32_e32 vcc_lo, v4, v4
-; GFX11-FAKE16-NEXT: s_lshr_b32 s72, s44, 16
-; GFX11-FAKE16-NEXT: v_add_nc_u32_e32 v5, v5, v2
-; GFX11-FAKE16-NEXT: s_bfe_u32 s45, s17, 0x10010
+; GFX11-FAKE16-NEXT: v_cmp_u_f32_e32 vcc_lo, v5, v5
+; GFX11-FAKE16-NEXT: v_add_f32_e64 v6, 0x40c00000, s17
+; GFX11-FAKE16-NEXT: v_add_nc_u32_e32 v10, 0x7fff, v11
+; GFX11-FAKE16-NEXT: v_or_b32_e32 v11, 0x400000, v5
+; GFX11-FAKE16-NEXT: v_add_f32_e64 v4, 0x40c00000, s57
; GFX11-FAKE16-NEXT: v_lshrrev_b32_e32 v28, 16, v3
-; GFX11-FAKE16-NEXT: v_and_b32_e32 v3, 0xffff, v27
-; GFX11-FAKE16-NEXT: s_add_i32 s45, s45, s17
+; GFX11-FAKE16-NEXT: v_readfirstlane_b32 s17, v6
+; GFX11-FAKE16-NEXT: v_add_f32_e64 v3, 0x40c00000, s56
+; GFX11-FAKE16-NEXT: v_cndmask_b32_e32 v5, v10, v11, vcc_lo
+; GFX11-FAKE16-NEXT: v_cmp_u_f32_e32 vcc_lo, v6, v6
+; GFX11-FAKE16-NEXT: v_bfe_u32 v7, v4, 16, 1
+; GFX11-FAKE16-NEXT: s_bfe_u32 s56, s17, 0x10010
+; GFX11-FAKE16-NEXT: v_bfe_u32 v10, v3, 16, 1
+; GFX11-FAKE16-NEXT: s_add_i32 s56, s56, s17
+; GFX11-FAKE16-NEXT: s_lshr_b32 s58, s58, 16
+; GFX11-FAKE16-NEXT: s_addk_i32 s56, 0x7fff
; GFX11-FAKE16-NEXT: s_bitset1_b32 s17, 22
-; GFX11-FAKE16-NEXT: s_addk_i32 s45, 0x7fff
-; GFX11-FAKE16-NEXT: s_and_b32 s44, vcc_lo, exec_lo
-; GFX11-FAKE16-NEXT: v_lshl_or_b32 v16, v28, 16, v3
-; GFX11-FAKE16-NEXT: v_add_nc_u32_e32 v3, 0x7fff, v5
-; GFX11-FAKE16-NEXT: v_or_b32_e32 v5, 0x400000, v2
-; GFX11-FAKE16-NEXT: v_cmp_u_f32_e32 vcc_lo, v2, v2
-; GFX11-FAKE16-NEXT: v_bfe_u32 v8, v1, 16, 1
-; GFX11-FAKE16-NEXT: s_cselect_b32 s17, s17, s45
-; GFX11-FAKE16-NEXT: s_and_b32 s44, s16, 0xffff0000
+; GFX11-FAKE16-NEXT: s_and_b32 s57, vcc_lo, exec_lo
+; GFX11-FAKE16-NEXT: s_cselect_b32 s17, s17, s56
+; GFX11-FAKE16-NEXT: s_and_b32 s56, s16, 0xffff0000
+; GFX11-FAKE16-NEXT: v_lshrrev_b32_e32 v29, 16, v5
+; GFX11-FAKE16-NEXT: v_and_b32_e32 v5, 0xffff, v28
+; GFX11-FAKE16-NEXT: v_add_nc_u32_e32 v7, v7, v4
+; GFX11-FAKE16-NEXT: v_add_nc_u32_e32 v6, v10, v3
+; GFX11-FAKE16-NEXT: v_add_f32_e64 v10, 0x40c00000, s56
+; GFX11-FAKE16-NEXT: v_cmp_u_f32_e32 vcc_lo, v4, v4
+; GFX11-FAKE16-NEXT: v_lshl_or_b32 v13, v29, 16, v5
+; GFX11-FAKE16-NEXT: v_add_nc_u32_e32 v5, 0x7fff, v7
+; GFX11-FAKE16-NEXT: v_or_b32_e32 v7, 0x400000, v4
+; GFX11-FAKE16-NEXT: v_readfirstlane_b32 s56, v10
; GFX11-FAKE16-NEXT: s_lshr_b32 s17, s17, 16
-; GFX11-FAKE16-NEXT: v_cndmask_b32_e32 v2, v3, v5, vcc_lo
-; GFX11-FAKE16-NEXT: v_add_f32_e64 v3, 0x40c00000, s63
-; GFX11-FAKE16-NEXT: v_or_b32_e32 v9, 0x400000, v1
-; GFX11-FAKE16-NEXT: v_lshrrev_b32_e32 v83, 24, v16
-; GFX11-FAKE16-NEXT: s_delay_alu instid0(VALU_DEP_4) | instskip(NEXT) | instid1(VALU_DEP_4)
-; GFX11-FAKE16-NEXT: v_lshrrev_b32_e32 v29, 16, v2
-; GFX11-FAKE16-NEXT: v_bfe_u32 v5, v3, 16, 1
-; GFX11-FAKE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_3)
-; GFX11-FAKE16-NEXT: v_add_nc_u32_e32 v2, v5, v3
-; GFX11-FAKE16-NEXT: v_and_b32_e32 v5, 0xffff, v29
-; GFX11-FAKE16-NEXT: s_delay_alu instid0(VALU_DEP_2) | instskip(SKIP_2) | instid1(VALU_DEP_2)
-; GFX11-FAKE16-NEXT: v_add_nc_u32_e32 v2, 0x7fff, v2
-; GFX11-FAKE16-NEXT: v_add_nc_u32_e32 v4, v8, v1
-; GFX11-FAKE16-NEXT: v_add_f32_e64 v8, 0x40c00000, s44
-; GFX11-FAKE16-NEXT: v_add_nc_u32_e32 v4, 0x7fff, v4
-; GFX11-FAKE16-NEXT: s_delay_alu instid0(VALU_DEP_2) | instskip(SKIP_2) | instid1(SALU_CYCLE_1)
-; GFX11-FAKE16-NEXT: v_readfirstlane_b32 s44, v8
-; GFX11-FAKE16-NEXT: v_cmp_u_f32_e32 vcc_lo, v8, v8
-; GFX11-FAKE16-NEXT: s_bfe_u32 s45, s44, 0x10010
-; GFX11-FAKE16-NEXT: s_add_i32 s45, s45, s44
-; GFX11-FAKE16-NEXT: s_bitset1_b32 s44, 22
-; GFX11-FAKE16-NEXT: s_addk_i32 s45, 0x7fff
-; GFX11-FAKE16-NEXT: s_and_b32 s46, vcc_lo, exec_lo
-; GFX11-FAKE16-NEXT: s_cselect_b32 s44, s44, s45
+; GFX11-FAKE16-NEXT: v_add_nc_u32_e32 v6, 0x7fff, v6
+; GFX11-FAKE16-NEXT: v_or_b32_e32 v11, 0x400000, v3
+; GFX11-FAKE16-NEXT: v_cndmask_b32_e32 v4, v5, v7, vcc_lo
+; GFX11-FAKE16-NEXT: v_add_f32_e64 v5, 0x40c00000, s47
+; GFX11-FAKE16-NEXT: s_bfe_u32 s47, s56, 0x10010
+; GFX11-FAKE16-NEXT: v_cmp_u_f32_e32 vcc_lo, v10, v10
+; GFX11-FAKE16-NEXT: s_add_i32 s47, s47, s56
+; GFX11-FAKE16-NEXT: s_bitset1_b32 s56, 22
+; GFX11-FAKE16-NEXT: s_addk_i32 s47, 0x7fff
+; GFX11-FAKE16-NEXT: v_bfe_u32 v7, v5, 16, 1
+; GFX11-FAKE16-NEXT: s_and_b32 s57, vcc_lo, exec_lo
+; GFX11-FAKE16-NEXT: s_cselect_b32 s47, s56, s47
; GFX11-FAKE16-NEXT: s_lshl_b32 s16, s16, 16
-; GFX11-FAKE16-NEXT: v_cmp_u_f32_e32 vcc_lo, v1, v1
-; GFX11-FAKE16-NEXT: v_add_f32_e64 v8, 0x40c00000, s16
-; GFX11-FAKE16-NEXT: s_lshr_b32 s46, s44, 16
-; GFX11-FAKE16-NEXT: v_cndmask_b32_e32 v1, v4, v9, vcc_lo
-; GFX11-FAKE16-NEXT: s_delay_alu instid0(VALU_DEP_2)
-; GFX11-FAKE16-NEXT: v_readfirstlane_b32 s16, v8
-; GFX11-FAKE16-NEXT: v_cmp_u_f32_e32 vcc_lo, v8, v8
-; GFX11-FAKE16-NEXT: v_or_b32_e32 v8, 0x400000, v3
-; GFX11-FAKE16-NEXT: v_add_f32_e64 v4, 0x40c00000, s62
-; GFX11-FAKE16-NEXT: v_lshrrev_b32_e32 v1, 16, v1
-; GFX11-FAKE16-NEXT: s_bfe_u32 s45, s16, 0x10010
-; GFX11-FAKE16-NEXT: s_delay_alu instid0(SALU_CYCLE_1)
-; GFX11-FAKE16-NEXT: s_add_i32 s45, s45, s16
-; GFX11-FAKE16-NEXT: s_bitset1_b32 s16, 22
-; GFX11-FAKE16-NEXT: s_addk_i32 s45, 0x7fff
-; GFX11-FAKE16-NEXT: s_and_b32 s44, vcc_lo, exec_lo
-; GFX11-FAKE16-NEXT: s_cselect_b32 s16, s16, s45
-; GFX11-FAKE16-NEXT: s_and_b32 s44, s19, 0xffff0000
; GFX11-FAKE16-NEXT: v_cmp_u_f32_e32 vcc_lo, v3, v3
-; GFX11-FAKE16-NEXT: v_add_f32_e64 v10, 0x40c00000, s44
-; GFX11-FAKE16-NEXT: v_lshl_or_b32 v15, v1, 16, v5
-; GFX11-FAKE16-NEXT: v_bfe_u32 v9, v4, 16, 1
-; GFX11-FAKE16-NEXT: s_lshr_b32 s16, s16, 16
-; GFX11-FAKE16-NEXT: v_cndmask_b32_e32 v1, v2, v8, vcc_lo
-; GFX11-FAKE16-NEXT: v_readfirstlane_b32 s44, v10
+; GFX11-FAKE16-NEXT: v_add_f32_e64 v10, 0x40c00000, s16
+; GFX11-FAKE16-NEXT: v_lshrrev_b32_e32 v30, 16, v4
+; GFX11-FAKE16-NEXT: v_add_nc_u32_e32 v4, v7, v5
+; GFX11-FAKE16-NEXT: v_lshrrev_b32_e32 v83, 24, v13
+; GFX11-FAKE16-NEXT: v_cndmask_b32_e32 v3, v6, v11, vcc_lo
+; GFX11-FAKE16-NEXT: v_readfirstlane_b32 s16, v10
+; GFX11-FAKE16-NEXT: v_add_f32_e64 v6, 0x40c00000, s45
; GFX11-FAKE16-NEXT: v_cmp_u_f32_e32 vcc_lo, v10, v10
-; GFX11-FAKE16-NEXT: v_add_nc_u32_e32 v9, v9, v4
-; GFX11-FAKE16-NEXT: v_add_f32_e64 v8, 0x40c00000, s60
-; GFX11-FAKE16-NEXT: v_or_b32_e32 v3, 0x400000, v4
-; GFX11-FAKE16-NEXT: s_bfe_u32 s45, s44, 0x10010
-; GFX11-FAKE16-NEXT: v_add_f32_e64 v5, 0x40c00000, s61
-; GFX11-FAKE16-NEXT: s_add_i32 s45, s45, s44
-; GFX11-FAKE16-NEXT: s_bitset1_b32 s44, 22
-; GFX11-FAKE16-NEXT: s_addk_i32 s45, 0x7fff
+; GFX11-FAKE16-NEXT: v_and_b32_e32 v7, 0xffff, v30
+; GFX11-FAKE16-NEXT: v_lshrrev_b32_e32 v3, 16, v3
+; GFX11-FAKE16-NEXT: s_bfe_u32 s45, s16, 0x10010
+; GFX11-FAKE16-NEXT: v_add_nc_u32_e32 v4, 0x7fff, v4
+; GFX11-FAKE16-NEXT: s_add_i32 s56, s45, s16
+; GFX11-FAKE16-NEXT: s_lshr_b32 s45, s47, 16
+; GFX11-FAKE16-NEXT: s_addk_i32 s56, 0x7fff
+; GFX11-FAKE16-NEXT: s_bitset1_b32 s16, 22
; GFX11-FAKE16-NEXT: s_and_b32 s47, vcc_lo, exec_lo
-; GFX11-FAKE16-NEXT: s_cselect_b32 s44, s44, s45
+; GFX11-FAKE16-NEXT: s_cselect_b32 s16, s16, s56
+; GFX11-FAKE16-NEXT: s_and_b32 s47, s19, 0xffff0000
+; GFX11-FAKE16-NEXT: v_or_b32_e32 v10, 0x400000, v5
+; GFX11-FAKE16-NEXT: v_add_f32_e64 v14, 0x40c00000, s47
+; GFX11-FAKE16-NEXT: v_cmp_u_f32_e32 vcc_lo, v5, v5
+; GFX11-FAKE16-NEXT: v_lshl_or_b32 v12, v3, 16, v7
+; GFX11-FAKE16-NEXT: v_add_f32_e64 v7, 0x40c00000, s44
+; GFX11-FAKE16-NEXT: v_bfe_u32 v11, v6, 16, 1
+; GFX11-FAKE16-NEXT: v_readfirstlane_b32 s47, v14
+; GFX11-FAKE16-NEXT: v_cndmask_b32_e32 v3, v4, v10, vcc_lo
+; GFX11-FAKE16-NEXT: v_cmp_u_f32_e32 vcc_lo, v14, v14
+; GFX11-FAKE16-NEXT: s_lshr_b32 s16, s16, 16
+; GFX11-FAKE16-NEXT: v_add_f32_e64 v10, 0x40c00000, s46
+; GFX11-FAKE16-NEXT: s_bfe_u32 s44, s47, 0x10010
+; GFX11-FAKE16-NEXT: v_add_nc_u32_e32 v11, v11, v6
+; GFX11-FAKE16-NEXT: s_add_i32 s44, s44, s47
+; GFX11-FAKE16-NEXT: s_bitset1_b32 s47, 22
+; GFX11-FAKE16-NEXT: s_addk_i32 s44, 0x7fff
+; GFX11-FAKE16-NEXT: s_and_b32 s46, vcc_lo, exec_lo
+; GFX11-FAKE16-NEXT: s_cselect_b32 s44, s47, s44
; GFX11-FAKE16-NEXT: s_lshl_b32 s19, s19, 16
-; GFX11-FAKE16-NEXT: v_add_nc_u32_e32 v2, 0x7fff, v9
-; GFX11-FAKE16-NEXT: v_add_f32_e64 v10, 0x40c00000, s19
-; GFX11-FAKE16-NEXT: v_cmp_u_f32_e32 vcc_lo, v4, v4
-; GFX11-FAKE16-NEXT: v_bfe_u32 v9, v8, 16, 1
-; GFX11-FAKE16-NEXT: s_lshr_b32 s60, s44, 16
-; GFX11-FAKE16-NEXT: v_lshrrev_b32_e32 v30, 16, v1
-; GFX11-FAKE16-NEXT: v_readfirstlane_b32 s19, v10
-; GFX11-FAKE16-NEXT: v_cndmask_b32_e32 v2, v2, v3, vcc_lo
-; GFX11-FAKE16-NEXT: v_cmp_u_f32_e32 vcc_lo, v10, v10
-; GFX11-FAKE16-NEXT: v_bfe_u32 v3, v5, 16, 1
-; GFX11-FAKE16-NEXT: v_add_nc_u32_e32 v4, v9, v8
-; GFX11-FAKE16-NEXT: s_bfe_u32 s45, s19, 0x10010
-; GFX11-FAKE16-NEXT: v_lshrrev_b32_e32 v31, 16, v2
-; GFX11-FAKE16-NEXT: s_add_i32 s45, s45, s19
+; GFX11-FAKE16-NEXT: v_add_nc_u32_e32 v4, 0x7fff, v11
+; GFX11-FAKE16-NEXT: v_add_f32_e64 v14, 0x40c00000, s19
+; GFX11-FAKE16-NEXT: v_or_b32_e32 v5, 0x400000, v6
+; GFX11-FAKE16-NEXT: v_cmp_u_f32_e32 vcc_lo, v6, v6
+; GFX11-FAKE16-NEXT: v_bfe_u32 v11, v10, 16, 1
+; GFX11-FAKE16-NEXT: s_lshr_b32 s59, s44, 16
+; GFX11-FAKE16-NEXT: v_readfirstlane_b32 s19, v14
+; GFX11-FAKE16-NEXT: v_lshrrev_b32_e32 v31, 16, v3
+; GFX11-FAKE16-NEXT: v_cndmask_b32_e32 v4, v4, v5, vcc_lo
+; GFX11-FAKE16-NEXT: v_cmp_u_f32_e32 vcc_lo, v14, v14
+; GFX11-FAKE16-NEXT: v_bfe_u32 v5, v7, 16, 1
+; GFX11-FAKE16-NEXT: s_bfe_u32 s46, s19, 0x10010
+; GFX11-FAKE16-NEXT: v_add_nc_u32_e32 v6, v11, v10
+; GFX11-FAKE16-NEXT: s_add_i32 s46, s46, s19
; GFX11-FAKE16-NEXT: s_bitset1_b32 s19, 22
-; GFX11-FAKE16-NEXT: s_addk_i32 s45, 0x7fff
+; GFX11-FAKE16-NEXT: s_addk_i32 s46, 0x7fff
; GFX11-FAKE16-NEXT: s_and_b32 s44, vcc_lo, exec_lo
-; GFX11-FAKE16-NEXT: s_cselect_b32 s19, s19, s45
+; GFX11-FAKE16-NEXT: s_cselect_b32 s19, s19, s46
; GFX11-FAKE16-NEXT: s_and_b32 s44, s18, 0xffff0000
-; GFX11-FAKE16-NEXT: v_add_nc_u32_e32 v1, v3, v5
-; GFX11-FAKE16-NEXT: v_add_nc_u32_e32 v2, 0x7fff, v4
-; GFX11-FAKE16-NEXT: v_or_b32_e32 v3, 0x400000, v8
-; GFX11-FAKE16-NEXT: v_add_f32_e64 v4, 0x40c00000, s44
-; GFX11-FAKE16-NEXT: v_cmp_u_f32_e32 vcc_lo, v8, v8
-; GFX11-FAKE16-NEXT: v_add_nc_u32_e32 v1, 0x7fff, v1
-; GFX11-FAKE16-NEXT: v_or_b32_e32 v9, 0x400000, v5
+; GFX11-FAKE16-NEXT: v_lshrrev_b32_e32 v32, 16, v4
+; GFX11-FAKE16-NEXT: v_add_nc_u32_e32 v3, v5, v7
+; GFX11-FAKE16-NEXT: v_add_nc_u32_e32 v4, 0x7fff, v6
+; GFX11-FAKE16-NEXT: v_or_b32_e32 v5, 0x400000, v10
+; GFX11-FAKE16-NEXT: v_add_f32_e64 v6, 0x40c00000, s44
+; GFX11-FAKE16-NEXT: v_cmp_u_f32_e32 vcc_lo, v10, v10
+; GFX11-FAKE16-NEXT: v_add_nc_u32_e32 v3, 0x7fff, v3
+; GFX11-FAKE16-NEXT: v_or_b32_e32 v11, 0x400000, v7
; GFX11-FAKE16-NEXT: s_lshr_b32 s19, s19, 16
-; GFX11-FAKE16-NEXT: v_add_f32_e64 v10, 0x40c00000, s29
-; GFX11-FAKE16-NEXT: v_cndmask_b32_e32 v2, v2, v3, vcc_lo
-; GFX11-FAKE16-NEXT: v_add_f32_e64 v3, 0x40c00000, s41
-; GFX11-FAKE16-NEXT: v_readfirstlane_b32 s41, v4
-; GFX11-FAKE16-NEXT: v_cmp_u_f32_e32 vcc_lo, v5, v5
-; GFX11-FAKE16-NEXT: s_pack_ll_b32_b16 s47, s17, s72
-; GFX11-FAKE16-NEXT: v_lshrrev_b32_e32 v32, 16, v2
-; GFX11-FAKE16-NEXT: v_bfe_u32 v2, v3, 16, 1
+; GFX11-FAKE16-NEXT: v_add_f32_e64 v14, 0x40c00000, s29
+; GFX11-FAKE16-NEXT: v_cndmask_b32_e32 v4, v4, v5, vcc_lo
+; GFX11-FAKE16-NEXT: v_add_f32_e64 v5, 0x40c00000, s41
+; GFX11-FAKE16-NEXT: v_readfirstlane_b32 s41, v6
+; GFX11-FAKE16-NEXT: v_cmp_u_f32_e32 vcc_lo, v7, v7
+; GFX11-FAKE16-NEXT: v_lshrrev_b32_e32 v84, 16, v12
+; GFX11-FAKE16-NEXT: v_lshrrev_b32_e32 v33, 16, v4
+; GFX11-FAKE16-NEXT: v_bfe_u32 v4, v5, 16, 1
; GFX11-FAKE16-NEXT: s_bfe_u32 s44, s41, 0x10010
-; GFX11-FAKE16-NEXT: v_cndmask_b32_e32 v1, v1, v9, vcc_lo
-; GFX11-FAKE16-NEXT: v_cmp_u_f32_e32 vcc_lo, v4, v4
+; GFX11-FAKE16-NEXT: v_cndmask_b32_e32 v3, v3, v11, vcc_lo
+; GFX11-FAKE16-NEXT: v_cmp_u_f32_e32 vcc_lo, v6, v6
; GFX11-FAKE16-NEXT: s_add_i32 s44, s44, s41
; GFX11-FAKE16-NEXT: s_bitset1_b32 s41, 22
; GFX11-FAKE16-NEXT: s_addk_i32 s44, 0x7fff
-; GFX11-FAKE16-NEXT: v_add_f32_e64 v9, 0x40c00000, s40
-; GFX11-FAKE16-NEXT: s_and_b32 s45, vcc_lo, exec_lo
+; GFX11-FAKE16-NEXT: v_add_f32_e64 v11, 0x40c00000, s40
+; GFX11-FAKE16-NEXT: s_and_b32 s46, vcc_lo, exec_lo
; GFX11-FAKE16-NEXT: s_cselect_b32 s41, s41, s44
; GFX11-FAKE16-NEXT: s_lshl_b32 s18, s18, 16
-; GFX11-FAKE16-NEXT: v_and_b32_e32 v4, 0xffff, v31
-; GFX11-FAKE16-NEXT: v_add_f32_e64 v5, 0x40c00000, s18
-; GFX11-FAKE16-NEXT: v_lshrrev_b32_e32 v1, 16, v1
-; GFX11-FAKE16-NEXT: v_and_b32_e32 v8, 0xffff, v32
-; GFX11-FAKE16-NEXT: v_add_nc_u32_e32 v2, v2, v3
-; GFX11-FAKE16-NEXT: v_lshl_or_b32 v18, v30, 16, v4
-; GFX11-FAKE16-NEXT: v_readfirstlane_b32 s18, v5
-; GFX11-FAKE16-NEXT: v_cmp_u_f32_e32 vcc_lo, v5, v5
-; GFX11-FAKE16-NEXT: v_lshl_or_b32 v17, v1, 16, v8
-; GFX11-FAKE16-NEXT: v_add_nc_u32_e32 v1, 0x7fff, v2
-; GFX11-FAKE16-NEXT: v_or_b32_e32 v4, 0x400000, v3
+; GFX11-FAKE16-NEXT: v_and_b32_e32 v6, 0xffff, v32
+; GFX11-FAKE16-NEXT: v_add_f32_e64 v7, 0x40c00000, s18
+; GFX11-FAKE16-NEXT: v_lshrrev_b32_e32 v3, 16, v3
+; GFX11-FAKE16-NEXT: v_and_b32_e32 v10, 0xffff, v33
+; GFX11-FAKE16-NEXT: v_add_nc_u32_e32 v4, v4, v5
+; GFX11-FAKE16-NEXT: v_lshl_or_b32 v19, v31, 16, v6
+; GFX11-FAKE16-NEXT: v_readfirstlane_b32 s18, v7
+; GFX11-FAKE16-NEXT: v_cmp_u_f32_e32 vcc_lo, v7, v7
+; GFX11-FAKE16-NEXT: v_lshl_or_b32 v18, v3, 16, v10
+; GFX11-FAKE16-NEXT: v_add_nc_u32_e32 v3, 0x7fff, v4
+; GFX11-FAKE16-NEXT: v_or_b32_e32 v6, 0x400000, v5
; GFX11-FAKE16-NEXT: s_bfe_u32 s40, s18, 0x10010
-; GFX11-FAKE16-NEXT: v_add_f32_e64 v8, 0x40c00000, s28
+; GFX11-FAKE16-NEXT: v_add_f32_e64 v10, 0x40c00000, s28
; GFX11-FAKE16-NEXT: s_add_i32 s44, s40, s18
; GFX11-FAKE16-NEXT: s_lshr_b32 s40, s41, 16
; GFX11-FAKE16-NEXT: s_addk_i32 s44, 0x7fff
@@ -177678,105 +177776,105 @@ define inreg <128 x i8> @bitcast_v64bf16_to_v128i8_scalar(<64 x bfloat> inreg %a
; GFX11-FAKE16-NEXT: s_and_b32 s41, vcc_lo, exec_lo
; GFX11-FAKE16-NEXT: s_cselect_b32 s18, s18, s44
; GFX11-FAKE16-NEXT: s_and_b32 s41, s21, 0xffff0000
-; GFX11-FAKE16-NEXT: v_cmp_u_f32_e32 vcc_lo, v3, v3
-; GFX11-FAKE16-NEXT: v_add_f32_e64 v5, 0x40c00000, s41
-; GFX11-FAKE16-NEXT: v_bfe_u32 v2, v9, 16, 1
-; GFX11-FAKE16-NEXT: s_lshr_b32 s18, s18, 16
-; GFX11-FAKE16-NEXT: v_or_b32_e32 v3, 0x400000, v9
-; GFX11-FAKE16-NEXT: v_cndmask_b32_e32 v1, v1, v4, vcc_lo
-; GFX11-FAKE16-NEXT: v_readfirstlane_b32 s28, v5
; GFX11-FAKE16-NEXT: v_cmp_u_f32_e32 vcc_lo, v5, v5
-; GFX11-FAKE16-NEXT: v_add_nc_u32_e32 v2, v2, v9
-; GFX11-FAKE16-NEXT: v_bfe_u32 v4, v8, 16, 1
-; GFX11-FAKE16-NEXT: v_bfe_u32 v5, v10, 16, 1
+; GFX11-FAKE16-NEXT: v_add_f32_e64 v7, 0x40c00000, s41
+; GFX11-FAKE16-NEXT: v_bfe_u32 v4, v11, 16, 1
+; GFX11-FAKE16-NEXT: s_lshr_b32 s18, s18, 16
+; GFX11-FAKE16-NEXT: v_or_b32_e32 v5, 0x400000, v11
+; GFX11-FAKE16-NEXT: v_cndmask_b32_e32 v3, v3, v6, vcc_lo
+; GFX11-FAKE16-NEXT: v_readfirstlane_b32 s28, v7
+; GFX11-FAKE16-NEXT: v_cmp_u_f32_e32 vcc_lo, v7, v7
+; GFX11-FAKE16-NEXT: v_add_nc_u32_e32 v4, v4, v11
+; GFX11-FAKE16-NEXT: v_bfe_u32 v6, v10, 16, 1
+; GFX11-FAKE16-NEXT: v_bfe_u32 v7, v14, 16, 1
; GFX11-FAKE16-NEXT: s_bfe_u32 s29, s28, 0x10010
-; GFX11-FAKE16-NEXT: v_lshrrev_b32_e32 v33, 16, v1
+; GFX11-FAKE16-NEXT: v_lshrrev_b32_e32 v34, 16, v3
; GFX11-FAKE16-NEXT: s_add_i32 s29, s29, s28
; GFX11-FAKE16-NEXT: s_bitset1_b32 s28, 22
; GFX11-FAKE16-NEXT: s_addk_i32 s29, 0x7fff
; GFX11-FAKE16-NEXT: s_and_b32 s41, vcc_lo, exec_lo
; GFX11-FAKE16-NEXT: s_cselect_b32 s28, s28, s29
; GFX11-FAKE16-NEXT: s_lshl_b32 s21, s21, 16
-; GFX11-FAKE16-NEXT: v_add_nc_u32_e32 v2, 0x7fff, v2
-; GFX11-FAKE16-NEXT: v_add_f32_e64 v11, 0x40c00000, s21
-; GFX11-FAKE16-NEXT: v_cmp_u_f32_e32 vcc_lo, v9, v9
-; GFX11-FAKE16-NEXT: s_lshr_b32 s61, s28, 16
-; GFX11-FAKE16-NEXT: v_add_nc_u32_e32 v1, v5, v10
-; GFX11-FAKE16-NEXT: s_pack_ll_b32_b16 s44, s2, s11
-; GFX11-FAKE16-NEXT: v_readfirstlane_b32 s21, v11
-; GFX11-FAKE16-NEXT: v_cndmask_b32_e32 v2, v2, v3, vcc_lo
+; GFX11-FAKE16-NEXT: v_add_nc_u32_e32 v4, 0x7fff, v4
+; GFX11-FAKE16-NEXT: v_add_f32_e64 v15, 0x40c00000, s21
; GFX11-FAKE16-NEXT: v_cmp_u_f32_e32 vcc_lo, v11, v11
-; GFX11-FAKE16-NEXT: v_add_nc_u32_e32 v3, v4, v8
-; GFX11-FAKE16-NEXT: v_add_nc_u32_e32 v1, 0x7fff, v1
+; GFX11-FAKE16-NEXT: s_lshr_b32 s60, s28, 16
+; GFX11-FAKE16-NEXT: v_add_nc_u32_e32 v3, v7, v14
+; GFX11-FAKE16-NEXT: v_lshrrev_b32_e32 v81, 24, v19
+; GFX11-FAKE16-NEXT: v_readfirstlane_b32 s21, v15
+; GFX11-FAKE16-NEXT: v_cndmask_b32_e32 v4, v4, v5, vcc_lo
+; GFX11-FAKE16-NEXT: v_cmp_u_f32_e32 vcc_lo, v15, v15
+; GFX11-FAKE16-NEXT: v_add_nc_u32_e32 v5, v6, v10
+; GFX11-FAKE16-NEXT: v_add_nc_u32_e32 v3, 0x7fff, v3
; GFX11-FAKE16-NEXT: s_bfe_u32 s29, s21, 0x10010
-; GFX11-FAKE16-NEXT: v_lshrrev_b32_e32 v34, 16, v2
+; GFX11-FAKE16-NEXT: v_lshrrev_b32_e32 v35, 16, v4
; GFX11-FAKE16-NEXT: s_add_i32 s29, s29, s21
; GFX11-FAKE16-NEXT: s_bitset1_b32 s21, 22
; GFX11-FAKE16-NEXT: s_addk_i32 s29, 0x7fff
; GFX11-FAKE16-NEXT: s_and_b32 s28, vcc_lo, exec_lo
; GFX11-FAKE16-NEXT: s_cselect_b32 s21, s21, s29
; GFX11-FAKE16-NEXT: s_and_b32 s28, s20, 0xffff0000
-; GFX11-FAKE16-NEXT: v_add_nc_u32_e32 v2, 0x7fff, v3
-; GFX11-FAKE16-NEXT: v_or_b32_e32 v3, 0x400000, v8
-; GFX11-FAKE16-NEXT: v_add_f32_e64 v5, 0x40c00000, s28
-; GFX11-FAKE16-NEXT: v_cmp_u_f32_e32 vcc_lo, v8, v8
-; GFX11-FAKE16-NEXT: v_or_b32_e32 v4, 0x400000, v10
-; GFX11-FAKE16-NEXT: s_lshr_b32 s21, s21, 16
-; GFX11-FAKE16-NEXT: s_pack_ll_b32_b16 s45, s3, s59
-; GFX11-FAKE16-NEXT: s_pack_ll_b32_b16 s46, s16, s46
-; GFX11-FAKE16-NEXT: v_cndmask_b32_e32 v2, v2, v3, vcc_lo
+; GFX11-FAKE16-NEXT: v_add_nc_u32_e32 v4, 0x7fff, v5
+; GFX11-FAKE16-NEXT: v_or_b32_e32 v5, 0x400000, v10
+; GFX11-FAKE16-NEXT: v_add_f32_e64 v7, 0x40c00000, s28
; GFX11-FAKE16-NEXT: v_cmp_u_f32_e32 vcc_lo, v10, v10
-; GFX11-FAKE16-NEXT: v_add_f32_e64 v3, 0x40c00000, s13
-; GFX11-FAKE16-NEXT: v_readfirstlane_b32 s13, v5
-; GFX11-FAKE16-NEXT: v_lshrrev_b32_e32 v81, 24, v18
-; GFX11-FAKE16-NEXT: v_lshrrev_b32_e32 v2, 16, v2
-; GFX11-FAKE16-NEXT: v_cndmask_b32_e32 v1, v1, v4, vcc_lo
-; GFX11-FAKE16-NEXT: v_cmp_u_f32_e32 vcc_lo, v5, v5
+; GFX11-FAKE16-NEXT: v_or_b32_e32 v6, 0x400000, v14
+; GFX11-FAKE16-NEXT: s_lshr_b32 s21, s21, 16
+; GFX11-FAKE16-NEXT: v_lshrrev_b32_e32 v82, 16, v18
+; GFX11-FAKE16-NEXT: v_lshrrev_b32_e32 v85, 24, v9
+; GFX11-FAKE16-NEXT: v_cndmask_b32_e32 v4, v4, v5, vcc_lo
+; GFX11-FAKE16-NEXT: v_cmp_u_f32_e32 vcc_lo, v14, v14
+; GFX11-FAKE16-NEXT: v_add_f32_e64 v5, 0x40c00000, s13
+; GFX11-FAKE16-NEXT: v_readfirstlane_b32 s13, v7
+; GFX11-FAKE16-NEXT: v_lshrrev_b32_e32 v86, 16, v8
+; GFX11-FAKE16-NEXT: v_lshrrev_b32_e32 v4, 16, v4
+; GFX11-FAKE16-NEXT: v_cndmask_b32_e32 v3, v3, v6, vcc_lo
+; GFX11-FAKE16-NEXT: v_cmp_u_f32_e32 vcc_lo, v7, v7
; GFX11-FAKE16-NEXT: s_bfe_u32 s28, s13, 0x10010
-; GFX11-FAKE16-NEXT: v_and_b32_e32 v4, 0xffff, v34
+; GFX11-FAKE16-NEXT: v_and_b32_e32 v6, 0xffff, v35
; GFX11-FAKE16-NEXT: s_add_i32 s28, s28, s13
; GFX11-FAKE16-NEXT: s_bitset1_b32 s13, 22
; GFX11-FAKE16-NEXT: s_addk_i32 s28, 0x7fff
; GFX11-FAKE16-NEXT: s_and_b32 s29, vcc_lo, exec_lo
; GFX11-FAKE16-NEXT: s_cselect_b32 s13, s13, s28
; GFX11-FAKE16-NEXT: s_lshl_b32 s20, s20, 16
-; GFX11-FAKE16-NEXT: v_lshrrev_b32_e32 v35, 16, v1
-; GFX11-FAKE16-NEXT: v_add_f32_e64 v8, 0x40c00000, s20
-; GFX11-FAKE16-NEXT: v_bfe_u32 v1, v3, 16, 1
-; GFX11-FAKE16-NEXT: v_add_f32_e64 v5, 0x40c00000, s10
-; GFX11-FAKE16-NEXT: v_lshl_or_b32 v20, v33, 16, v4
-; GFX11-FAKE16-NEXT: v_and_b32_e32 v9, 0xffff, v35
-; GFX11-FAKE16-NEXT: v_readfirstlane_b32 s20, v8
-; GFX11-FAKE16-NEXT: v_add_nc_u32_e32 v1, v1, v3
-; GFX11-FAKE16-NEXT: v_cmp_u_f32_e32 vcc_lo, v8, v8
-; GFX11-FAKE16-NEXT: v_bfe_u32 v4, v5, 16, 1
-; GFX11-FAKE16-NEXT: v_lshl_or_b32 v19, v2, 16, v9
+; GFX11-FAKE16-NEXT: v_lshrrev_b32_e32 v36, 16, v3
+; GFX11-FAKE16-NEXT: v_add_f32_e64 v10, 0x40c00000, s20
+; GFX11-FAKE16-NEXT: v_bfe_u32 v3, v5, 16, 1
+; GFX11-FAKE16-NEXT: v_add_f32_e64 v7, 0x40c00000, s10
+; GFX11-FAKE16-NEXT: v_lshl_or_b32 v21, v34, 16, v6
+; GFX11-FAKE16-NEXT: v_and_b32_e32 v11, 0xffff, v36
+; GFX11-FAKE16-NEXT: v_readfirstlane_b32 s20, v10
+; GFX11-FAKE16-NEXT: v_add_nc_u32_e32 v3, v3, v5
+; GFX11-FAKE16-NEXT: v_cmp_u_f32_e32 vcc_lo, v10, v10
+; GFX11-FAKE16-NEXT: v_bfe_u32 v6, v7, 16, 1
+; GFX11-FAKE16-NEXT: v_lshl_or_b32 v20, v4, 16, v11
; GFX11-FAKE16-NEXT: s_bfe_u32 s10, s20, 0x10010
-; GFX11-FAKE16-NEXT: v_add_nc_u32_e32 v1, 0x7fff, v1
+; GFX11-FAKE16-NEXT: v_add_nc_u32_e32 v3, 0x7fff, v3
; GFX11-FAKE16-NEXT: s_add_i32 s28, s10, s20
; GFX11-FAKE16-NEXT: s_lshr_b32 s10, s13, 16
; GFX11-FAKE16-NEXT: s_addk_i32 s28, 0x7fff
; GFX11-FAKE16-NEXT: s_bitset1_b32 s20, 22
-; GFX11-FAKE16-NEXT: v_or_b32_e32 v2, 0x400000, v3
+; GFX11-FAKE16-NEXT: v_or_b32_e32 v4, 0x400000, v5
; GFX11-FAKE16-NEXT: s_and_b32 s13, vcc_lo, exec_lo
-; GFX11-FAKE16-NEXT: v_cmp_u_f32_e32 vcc_lo, v3, v3
+; GFX11-FAKE16-NEXT: v_cmp_u_f32_e32 vcc_lo, v5, v5
; GFX11-FAKE16-NEXT: s_cselect_b32 s13, s20, s28
; GFX11-FAKE16-NEXT: s_and_b32 s20, s23, 0xffff0000
-; GFX11-FAKE16-NEXT: v_add_nc_u32_e32 v4, v4, v5
-; GFX11-FAKE16-NEXT: v_add_f32_e64 v8, 0x40c00000, s42
-; GFX11-FAKE16-NEXT: v_cndmask_b32_e32 v1, v1, v2, vcc_lo
-; GFX11-FAKE16-NEXT: v_add_f32_e64 v2, 0x40c00000, s20
-; GFX11-FAKE16-NEXT: v_cmp_u_f32_e32 vcc_lo, v5, v5
-; GFX11-FAKE16-NEXT: v_add_nc_u32_e32 v3, 0x7fff, v4
-; GFX11-FAKE16-NEXT: v_or_b32_e32 v4, 0x400000, v5
-; GFX11-FAKE16-NEXT: v_add_f32_e64 v9, 0x40c00000, s43
-; GFX11-FAKE16-NEXT: v_readfirstlane_b32 s28, v2
-; GFX11-FAKE16-NEXT: v_lshrrev_b32_e32 v37, 16, v1
-; GFX11-FAKE16-NEXT: v_lshrrev_b32_e32 v80, 16, v19
+; GFX11-FAKE16-NEXT: v_add_nc_u32_e32 v6, v6, v7
+; GFX11-FAKE16-NEXT: v_add_f32_e64 v10, 0x40c00000, s42
; GFX11-FAKE16-NEXT: v_cndmask_b32_e32 v3, v3, v4, vcc_lo
-; GFX11-FAKE16-NEXT: v_cmp_u_f32_e32 vcc_lo, v2, v2
+; GFX11-FAKE16-NEXT: v_add_f32_e64 v4, 0x40c00000, s20
+; GFX11-FAKE16-NEXT: v_cmp_u_f32_e32 vcc_lo, v7, v7
+; GFX11-FAKE16-NEXT: v_add_nc_u32_e32 v5, 0x7fff, v6
+; GFX11-FAKE16-NEXT: v_or_b32_e32 v6, 0x400000, v7
+; GFX11-FAKE16-NEXT: v_add_f32_e64 v11, 0x40c00000, s43
+; GFX11-FAKE16-NEXT: v_readfirstlane_b32 s28, v4
+; GFX11-FAKE16-NEXT: v_lshrrev_b32_e32 v38, 16, v3
+; GFX11-FAKE16-NEXT: v_lshrrev_b32_e32 v80, 16, v20
+; GFX11-FAKE16-NEXT: v_cndmask_b32_e32 v5, v5, v6, vcc_lo
+; GFX11-FAKE16-NEXT: v_cmp_u_f32_e32 vcc_lo, v4, v4
; GFX11-FAKE16-NEXT: s_bfe_u32 s20, s28, 0x10010
-; GFX11-FAKE16-NEXT: v_bfe_u32 v4, v8, 16, 1
+; GFX11-FAKE16-NEXT: v_bfe_u32 v6, v10, 16, 1
; GFX11-FAKE16-NEXT: s_add_i32 s29, s20, s28
; GFX11-FAKE16-NEXT: s_lshr_b32 s20, s13, 16
; GFX11-FAKE16-NEXT: s_addk_i32 s29, 0x7fff
@@ -177784,235 +177882,241 @@ define inreg <128 x i8> @bitcast_v64bf16_to_v128i8_scalar(<64 x bfloat> inreg %a
; GFX11-FAKE16-NEXT: s_and_b32 s13, vcc_lo, exec_lo
; GFX11-FAKE16-NEXT: s_cselect_b32 s13, s28, s29
; GFX11-FAKE16-NEXT: s_lshl_b32 s23, s23, 16
-; GFX11-FAKE16-NEXT: v_bfe_u32 v5, v9, 16, 1
-; GFX11-FAKE16-NEXT: v_add_f32_e64 v2, 0x40c00000, s23
-; GFX11-FAKE16-NEXT: v_lshrrev_b32_e32 v36, 16, v3
-; GFX11-FAKE16-NEXT: v_add_nc_u32_e32 v3, v4, v8
-; GFX11-FAKE16-NEXT: s_lshr_b32 s62, s13, 16
-; GFX11-FAKE16-NEXT: v_add_nc_u32_e32 v1, v5, v9
-; GFX11-FAKE16-NEXT: v_readfirstlane_b32 s23, v2
-; GFX11-FAKE16-NEXT: v_cmp_u_f32_e32 vcc_lo, v2, v2
+; GFX11-FAKE16-NEXT: v_bfe_u32 v7, v11, 16, 1
+; GFX11-FAKE16-NEXT: v_add_f32_e64 v4, 0x40c00000, s23
+; GFX11-FAKE16-NEXT: v_lshrrev_b32_e32 v37, 16, v5
+; GFX11-FAKE16-NEXT: v_add_nc_u32_e32 v5, v6, v10
+; GFX11-FAKE16-NEXT: s_lshr_b32 s63, s13, 16
+; GFX11-FAKE16-NEXT: v_add_nc_u32_e32 v3, v7, v11
+; GFX11-FAKE16-NEXT: v_readfirstlane_b32 s23, v4
+; GFX11-FAKE16-NEXT: v_cmp_u_f32_e32 vcc_lo, v4, v4
+; GFX11-FAKE16-NEXT: v_add_nc_u32_e32 v5, 0x7fff, v5
+; GFX11-FAKE16-NEXT: v_or_b32_e32 v7, 0x400000, v10
; GFX11-FAKE16-NEXT: v_add_nc_u32_e32 v3, 0x7fff, v3
-; GFX11-FAKE16-NEXT: v_or_b32_e32 v5, 0x400000, v8
-; GFX11-FAKE16-NEXT: v_add_nc_u32_e32 v1, 0x7fff, v1
; GFX11-FAKE16-NEXT: s_bfe_u32 s28, s23, 0x10010
-; GFX11-FAKE16-NEXT: v_or_b32_e32 v10, 0x400000, v9
+; GFX11-FAKE16-NEXT: v_or_b32_e32 v14, 0x400000, v11
; GFX11-FAKE16-NEXT: s_add_i32 s28, s28, s23
; GFX11-FAKE16-NEXT: s_bitset1_b32 s23, 22
; GFX11-FAKE16-NEXT: s_addk_i32 s28, 0x7fff
; GFX11-FAKE16-NEXT: s_and_b32 s13, vcc_lo, exec_lo
-; GFX11-FAKE16-NEXT: v_cmp_u_f32_e32 vcc_lo, v8, v8
+; GFX11-FAKE16-NEXT: v_cmp_u_f32_e32 vcc_lo, v10, v10
; GFX11-FAKE16-NEXT: s_cselect_b32 s13, s23, s28
; GFX11-FAKE16-NEXT: s_and_b32 s23, s22, 0xffff0000
-; GFX11-FAKE16-NEXT: v_add_f32_e64 v8, 0x40c00000, s15
-; GFX11-FAKE16-NEXT: v_and_b32_e32 v4, 0xffff, v36
-; GFX11-FAKE16-NEXT: v_cndmask_b32_e32 v2, v3, v5, vcc_lo
-; GFX11-FAKE16-NEXT: v_add_f32_e64 v3, 0x40c00000, s23
-; GFX11-FAKE16-NEXT: v_cmp_u_f32_e32 vcc_lo, v9, v9
-; GFX11-FAKE16-NEXT: v_add_f32_e64 v5, 0x40c00000, s14
+; GFX11-FAKE16-NEXT: v_add_f32_e64 v10, 0x40c00000, s15
+; GFX11-FAKE16-NEXT: v_and_b32_e32 v6, 0xffff, v37
+; GFX11-FAKE16-NEXT: v_cndmask_b32_e32 v4, v5, v7, vcc_lo
+; GFX11-FAKE16-NEXT: v_add_f32_e64 v5, 0x40c00000, s23
+; GFX11-FAKE16-NEXT: v_cmp_u_f32_e32 vcc_lo, v11, v11
+; GFX11-FAKE16-NEXT: v_add_f32_e64 v7, 0x40c00000, s14
; GFX11-FAKE16-NEXT: s_lshr_b32 s23, s13, 16
-; GFX11-FAKE16-NEXT: v_bfe_u32 v9, v8, 16, 1
-; GFX11-FAKE16-NEXT: v_readfirstlane_b32 s14, v3
-; GFX11-FAKE16-NEXT: v_cndmask_b32_e32 v1, v1, v10, vcc_lo
-; GFX11-FAKE16-NEXT: v_cmp_u_f32_e32 vcc_lo, v3, v3
-; GFX11-FAKE16-NEXT: v_lshl_or_b32 v71, v37, 16, v4
-; GFX11-FAKE16-NEXT: v_add_f32_e64 v4, 0x40c00000, s12
+; GFX11-FAKE16-NEXT: v_bfe_u32 v11, v10, 16, 1
+; GFX11-FAKE16-NEXT: v_readfirstlane_b32 s14, v5
+; GFX11-FAKE16-NEXT: v_cndmask_b32_e32 v3, v3, v14, vcc_lo
+; GFX11-FAKE16-NEXT: v_cmp_u_f32_e32 vcc_lo, v5, v5
+; GFX11-FAKE16-NEXT: v_lshl_or_b32 v71, v38, 16, v6
+; GFX11-FAKE16-NEXT: v_add_f32_e64 v6, 0x40c00000, s12
; GFX11-FAKE16-NEXT: s_bfe_u32 s15, s14, 0x10010
-; GFX11-FAKE16-NEXT: v_lshrrev_b32_e32 v38, 16, v1
+; GFX11-FAKE16-NEXT: v_lshrrev_b32_e32 v39, 16, v3
; GFX11-FAKE16-NEXT: s_add_i32 s15, s15, s14
; GFX11-FAKE16-NEXT: s_bitset1_b32 s14, 22
; GFX11-FAKE16-NEXT: s_addk_i32 s15, 0x7fff
; GFX11-FAKE16-NEXT: s_and_b32 s13, vcc_lo, exec_lo
; GFX11-FAKE16-NEXT: s_cselect_b32 s13, s14, s15
; GFX11-FAKE16-NEXT: s_lshl_b32 s14, s22, 16
-; GFX11-FAKE16-NEXT: v_lshrrev_b32_e32 v2, 16, v2
-; GFX11-FAKE16-NEXT: v_add_f32_e64 v10, 0x40c00000, s14
-; GFX11-FAKE16-NEXT: v_bfe_u32 v1, v5, 16, 1
-; GFX11-FAKE16-NEXT: v_and_b32_e32 v3, 0xffff, v38
-; GFX11-FAKE16-NEXT: v_add_nc_u32_e32 v9, v9, v8
-; GFX11-FAKE16-NEXT: s_lshr_b32 s13, s13, 16
-; GFX11-FAKE16-NEXT: v_readfirstlane_b32 s14, v10
-; GFX11-FAKE16-NEXT: v_cmp_u_f32_e32 vcc_lo, v10, v10
-; GFX11-FAKE16-NEXT: v_add_nc_u32_e32 v1, v1, v5
-; GFX11-FAKE16-NEXT: v_lshl_or_b32 v70, v2, 16, v3
-; GFX11-FAKE16-NEXT: v_add_nc_u32_e32 v2, 0x7fff, v9
+; GFX11-FAKE16-NEXT: v_lshrrev_b32_e32 v4, 16, v4
+; GFX11-FAKE16-NEXT: v_add_f32_e64 v14, 0x40c00000, s14
+; GFX11-FAKE16-NEXT: v_bfe_u32 v3, v7, 16, 1
+; GFX11-FAKE16-NEXT: v_and_b32_e32 v5, 0xffff, v39
+; GFX11-FAKE16-NEXT: v_add_nc_u32_e32 v11, v11, v10
+; GFX11-FAKE16-NEXT: s_lshr_b32 s28, s13, 16
+; GFX11-FAKE16-NEXT: v_readfirstlane_b32 s14, v14
+; GFX11-FAKE16-NEXT: v_cmp_u_f32_e32 vcc_lo, v14, v14
+; GFX11-FAKE16-NEXT: v_add_nc_u32_e32 v3, v3, v7
+; GFX11-FAKE16-NEXT: v_lshl_or_b32 v70, v4, 16, v5
+; GFX11-FAKE16-NEXT: v_add_nc_u32_e32 v4, 0x7fff, v11
; GFX11-FAKE16-NEXT: s_bfe_u32 s12, s14, 0x10010
-; GFX11-FAKE16-NEXT: v_or_b32_e32 v3, 0x400000, v8
+; GFX11-FAKE16-NEXT: v_or_b32_e32 v5, 0x400000, v10
; GFX11-FAKE16-NEXT: s_add_i32 s12, s12, s14
; GFX11-FAKE16-NEXT: s_bitset1_b32 s14, 22
; GFX11-FAKE16-NEXT: s_addk_i32 s12, 0x7fff
-; GFX11-FAKE16-NEXT: s_and_b32 s15, vcc_lo, exec_lo
+; GFX11-FAKE16-NEXT: s_and_b32 s13, vcc_lo, exec_lo
; GFX11-FAKE16-NEXT: s_cselect_b32 s12, s14, s12
-; GFX11-FAKE16-NEXT: s_and_b32 s14, s25, 0xffff0000
-; GFX11-FAKE16-NEXT: v_cmp_u_f32_e32 vcc_lo, v8, v8
-; GFX11-FAKE16-NEXT: v_add_f32_e64 v10, 0x40c00000, s14
-; GFX11-FAKE16-NEXT: v_add_nc_u32_e32 v1, 0x7fff, v1
-; GFX11-FAKE16-NEXT: v_or_b32_e32 v9, 0x400000, v5
-; GFX11-FAKE16-NEXT: v_add_f32_e64 v8, 0x40c00000, s9
-; GFX11-FAKE16-NEXT: v_cndmask_b32_e32 v2, v2, v3, vcc_lo
-; GFX11-FAKE16-NEXT: v_readfirstlane_b32 s9, v10
-; GFX11-FAKE16-NEXT: v_cmp_u_f32_e32 vcc_lo, v5, v5
-; GFX11-FAKE16-NEXT: s_lshr_b32 s22, s12, 16
-; GFX11-FAKE16-NEXT: v_bfe_u32 v3, v4, 16, 1
-; GFX11-FAKE16-NEXT: v_lshrrev_b32_e32 v39, 16, v2
-; GFX11-FAKE16-NEXT: s_bfe_u32 s14, s9, 0x10010
-; GFX11-FAKE16-NEXT: v_cndmask_b32_e32 v1, v1, v9, vcc_lo
+; GFX11-FAKE16-NEXT: s_and_b32 s13, s25, 0xffff0000
; GFX11-FAKE16-NEXT: v_cmp_u_f32_e32 vcc_lo, v10, v10
-; GFX11-FAKE16-NEXT: s_add_i32 s14, s14, s9
+; GFX11-FAKE16-NEXT: v_add_f32_e64 v14, 0x40c00000, s13
+; GFX11-FAKE16-NEXT: v_add_nc_u32_e32 v3, 0x7fff, v3
+; GFX11-FAKE16-NEXT: v_or_b32_e32 v11, 0x400000, v7
+; GFX11-FAKE16-NEXT: v_add_f32_e64 v10, 0x40c00000, s9
+; GFX11-FAKE16-NEXT: v_cndmask_b32_e32 v4, v4, v5, vcc_lo
+; GFX11-FAKE16-NEXT: v_readfirstlane_b32 s9, v14
+; GFX11-FAKE16-NEXT: v_cmp_u_f32_e32 vcc_lo, v7, v7
+; GFX11-FAKE16-NEXT: s_lshr_b32 s22, s12, 16
+; GFX11-FAKE16-NEXT: v_bfe_u32 v5, v6, 16, 1
+; GFX11-FAKE16-NEXT: v_lshrrev_b32_e32 v48, 16, v4
+; GFX11-FAKE16-NEXT: s_bfe_u32 s13, s9, 0x10010
+; GFX11-FAKE16-NEXT: v_cndmask_b32_e32 v3, v3, v11, vcc_lo
+; GFX11-FAKE16-NEXT: v_cmp_u_f32_e32 vcc_lo, v14, v14
+; GFX11-FAKE16-NEXT: s_add_i32 s13, s13, s9
; GFX11-FAKE16-NEXT: s_bitset1_b32 s9, 22
-; GFX11-FAKE16-NEXT: s_addk_i32 s14, 0x7fff
-; GFX11-FAKE16-NEXT: v_lshrrev_b32_e32 v48, 16, v1
+; GFX11-FAKE16-NEXT: s_addk_i32 s13, 0x7fff
+; GFX11-FAKE16-NEXT: v_lshrrev_b32_e32 v49, 16, v3
; GFX11-FAKE16-NEXT: s_and_b32 s12, vcc_lo, exec_lo
-; GFX11-FAKE16-NEXT: s_cselect_b32 s9, s9, s14
+; GFX11-FAKE16-NEXT: s_cselect_b32 s9, s9, s13
; GFX11-FAKE16-NEXT: s_lshl_b32 s12, s25, 16
-; GFX11-FAKE16-NEXT: v_add_f32_e64 v9, 0x40c00000, s8
-; GFX11-FAKE16-NEXT: v_add_f32_e64 v1, 0x40c00000, s12
-; GFX11-FAKE16-NEXT: v_add_nc_u32_e32 v2, v3, v4
-; GFX11-FAKE16-NEXT: s_lshr_b32 s63, s9, 16
-; GFX11-FAKE16-NEXT: v_bfe_u32 v3, v8, 16, 1
-; GFX11-FAKE16-NEXT: v_or_b32_e32 v10, 0x400000, v4
-; GFX11-FAKE16-NEXT: v_readfirstlane_b32 s8, v1
-; GFX11-FAKE16-NEXT: v_cmp_u_f32_e32 vcc_lo, v1, v1
-; GFX11-FAKE16-NEXT: v_add_nc_u32_e32 v2, 0x7fff, v2
-; GFX11-FAKE16-NEXT: v_add_nc_u32_e32 v3, v3, v8
-; GFX11-FAKE16-NEXT: v_or_b32_e32 v11, 0x400000, v8
+; GFX11-FAKE16-NEXT: v_add_f32_e64 v11, 0x40c00000, s8
+; GFX11-FAKE16-NEXT: v_add_f32_e64 v3, 0x40c00000, s12
+; GFX11-FAKE16-NEXT: v_add_nc_u32_e32 v4, v5, v6
+; GFX11-FAKE16-NEXT: s_lshr_b32 s72, s9, 16
+; GFX11-FAKE16-NEXT: v_bfe_u32 v5, v10, 16, 1
+; GFX11-FAKE16-NEXT: v_or_b32_e32 v14, 0x400000, v6
+; GFX11-FAKE16-NEXT: v_readfirstlane_b32 s8, v3
+; GFX11-FAKE16-NEXT: v_cmp_u_f32_e32 vcc_lo, v3, v3
+; GFX11-FAKE16-NEXT: v_add_nc_u32_e32 v4, 0x7fff, v4
+; GFX11-FAKE16-NEXT: v_add_nc_u32_e32 v5, v5, v10
+; GFX11-FAKE16-NEXT: v_or_b32_e32 v15, 0x400000, v10
; GFX11-FAKE16-NEXT: s_bfe_u32 s12, s8, 0x10010
-; GFX11-FAKE16-NEXT: v_bfe_u32 v12, v9, 16, 1
+; GFX11-FAKE16-NEXT: v_bfe_u32 v16, v11, 16, 1
; GFX11-FAKE16-NEXT: s_add_i32 s12, s12, s8
; GFX11-FAKE16-NEXT: s_bitset1_b32 s8, 22
; GFX11-FAKE16-NEXT: s_addk_i32 s12, 0x7fff
; GFX11-FAKE16-NEXT: s_and_b32 s9, vcc_lo, exec_lo
-; GFX11-FAKE16-NEXT: v_cmp_u_f32_e32 vcc_lo, v4, v4
+; GFX11-FAKE16-NEXT: v_cmp_u_f32_e32 vcc_lo, v6, v6
; GFX11-FAKE16-NEXT: s_cselect_b32 s8, s8, s12
; GFX11-FAKE16-NEXT: s_and_b32 s9, s24, 0xffff0000
-; GFX11-FAKE16-NEXT: v_add_nc_u32_e32 v3, 0x7fff, v3
+; GFX11-FAKE16-NEXT: v_add_nc_u32_e32 v5, 0x7fff, v5
; GFX11-FAKE16-NEXT: s_lshr_b32 s25, s8, 16
-; GFX11-FAKE16-NEXT: v_cndmask_b32_e32 v1, v2, v10, vcc_lo
-; GFX11-FAKE16-NEXT: v_add_f32_e64 v2, 0x40c00000, s9
-; GFX11-FAKE16-NEXT: v_cmp_u_f32_e32 vcc_lo, v8, v8
-; GFX11-FAKE16-NEXT: v_add_f32_e64 v8, 0x40c00000, s7
-; GFX11-FAKE16-NEXT: v_add_nc_u32_e32 v4, v12, v9
-; GFX11-FAKE16-NEXT: v_add_f32_e64 v12, 0x40c00000, s6
-; GFX11-FAKE16-NEXT: v_readfirstlane_b32 s7, v2
-; GFX11-FAKE16-NEXT: v_cndmask_b32_e32 v3, v3, v11, vcc_lo
-; GFX11-FAKE16-NEXT: v_cmp_u_f32_e32 vcc_lo, v2, v2
-; GFX11-FAKE16-NEXT: v_or_b32_e32 v2, 0x400000, v9
-; GFX11-FAKE16-NEXT: s_pack_ll_b32_b16 s28, s0, s5
+; GFX11-FAKE16-NEXT: v_cndmask_b32_e32 v3, v4, v14, vcc_lo
+; GFX11-FAKE16-NEXT: v_add_f32_e64 v4, 0x40c00000, s9
+; GFX11-FAKE16-NEXT: v_cmp_u_f32_e32 vcc_lo, v10, v10
+; GFX11-FAKE16-NEXT: v_add_f32_e64 v10, 0x40c00000, s7
+; GFX11-FAKE16-NEXT: v_add_nc_u32_e32 v6, v16, v11
+; GFX11-FAKE16-NEXT: v_add_f32_e64 v16, 0x40c00000, s6
+; GFX11-FAKE16-NEXT: v_readfirstlane_b32 s7, v4
+; GFX11-FAKE16-NEXT: v_cndmask_b32_e32 v5, v5, v15, vcc_lo
+; GFX11-FAKE16-NEXT: v_cmp_u_f32_e32 vcc_lo, v4, v4
+; GFX11-FAKE16-NEXT: v_or_b32_e32 v4, 0x400000, v11
+; GFX11-FAKE16-NEXT: v_lshrrev_b32_e32 v3, 16, v3
; GFX11-FAKE16-NEXT: s_bfe_u32 s9, s7, 0x10010
-; GFX11-FAKE16-NEXT: v_lshrrev_b32_e32 v49, 16, v3
+; GFX11-FAKE16-NEXT: v_lshrrev_b32_e32 v50, 16, v5
; GFX11-FAKE16-NEXT: s_add_i32 s9, s9, s7
; GFX11-FAKE16-NEXT: s_bitset1_b32 s7, 22
; GFX11-FAKE16-NEXT: s_addk_i32 s9, 0x7fff
; GFX11-FAKE16-NEXT: s_and_b32 s8, vcc_lo, exec_lo
; GFX11-FAKE16-NEXT: s_cselect_b32 s7, s7, s9
; GFX11-FAKE16-NEXT: s_lshl_b32 s8, s24, 16
-; GFX11-FAKE16-NEXT: v_add_nc_u32_e32 v3, 0x7fff, v4
-; GFX11-FAKE16-NEXT: v_add_f32_e64 v10, 0x40c00000, s8
-; GFX11-FAKE16-NEXT: v_cmp_u_f32_e32 vcc_lo, v9, v9
-; GFX11-FAKE16-NEXT: v_add_f32_e64 v9, 0x40c00000, s4
-; GFX11-FAKE16-NEXT: v_bfe_u32 v4, v8, 16, 1
-; GFX11-FAKE16-NEXT: s_lshr_b32 s12, s7, 16
-; GFX11-FAKE16-NEXT: v_readfirstlane_b32 s8, v10
-; GFX11-FAKE16-NEXT: v_cndmask_b32_e32 v2, v3, v2, vcc_lo
-; GFX11-FAKE16-NEXT: v_cmp_u_f32_e32 vcc_lo, v10, v10
-; GFX11-FAKE16-NEXT: v_add_nc_u32_e32 v4, v4, v8
-; GFX11-FAKE16-NEXT: v_bfe_u32 v10, v12, 16, 1
-; GFX11-FAKE16-NEXT: s_bfe_u32 s4, s8, 0x10010
-; GFX11-FAKE16-NEXT: v_lshrrev_b32_e32 v50, 16, v2
-; GFX11-FAKE16-NEXT: s_add_i32 s4, s4, s8
+; GFX11-FAKE16-NEXT: v_add_nc_u32_e32 v5, 0x7fff, v6
+; GFX11-FAKE16-NEXT: v_add_f32_e64 v14, 0x40c00000, s8
+; GFX11-FAKE16-NEXT: v_cmp_u_f32_e32 vcc_lo, v11, v11
+; GFX11-FAKE16-NEXT: v_add_f32_e64 v11, 0x40c00000, s5
+; GFX11-FAKE16-NEXT: v_bfe_u32 v6, v10, 16, 1
+; GFX11-FAKE16-NEXT: s_lshr_b32 s42, s7, 16
+; GFX11-FAKE16-NEXT: v_readfirstlane_b32 s8, v14
+; GFX11-FAKE16-NEXT: v_cndmask_b32_e32 v4, v5, v4, vcc_lo
+; GFX11-FAKE16-NEXT: v_cmp_u_f32_e32 vcc_lo, v14, v14
+; GFX11-FAKE16-NEXT: v_add_nc_u32_e32 v6, v6, v10
+; GFX11-FAKE16-NEXT: v_bfe_u32 v14, v16, 16, 1
+; GFX11-FAKE16-NEXT: s_bfe_u32 s5, s8, 0x10010
+; GFX11-FAKE16-NEXT: v_lshrrev_b32_e32 v51, 16, v4
+; GFX11-FAKE16-NEXT: s_add_i32 s5, s5, s8
; GFX11-FAKE16-NEXT: s_bitset1_b32 s8, 22
-; GFX11-FAKE16-NEXT: s_addk_i32 s4, 0x7fff
+; GFX11-FAKE16-NEXT: s_addk_i32 s5, 0x7fff
; GFX11-FAKE16-NEXT: s_and_b32 s6, vcc_lo, exec_lo
-; GFX11-FAKE16-NEXT: s_cselect_b32 s4, s8, s4
+; GFX11-FAKE16-NEXT: s_cselect_b32 s5, s8, s5
; GFX11-FAKE16-NEXT: s_and_b32 s6, s27, 0xffff0000
-; GFX11-FAKE16-NEXT: v_add_nc_u32_e32 v3, 0x7fff, v4
-; GFX11-FAKE16-NEXT: v_add_f32_e64 v52, 0x40c00000, s6
-; GFX11-FAKE16-NEXT: v_or_b32_e32 v4, 0x400000, v8
-; GFX11-FAKE16-NEXT: v_cmp_u_f32_e32 vcc_lo, v8, v8
-; GFX11-FAKE16-NEXT: v_add_nc_u32_e32 v8, v10, v12
-; GFX11-FAKE16-NEXT: s_lshr_b32 s24, s4, 16
-; GFX11-FAKE16-NEXT: v_readfirstlane_b32 s6, v52
-; GFX11-FAKE16-NEXT: v_or_b32_e32 v10, 0x400000, v9
-; GFX11-FAKE16-NEXT: v_cndmask_b32_e32 v3, v3, v4, vcc_lo
-; GFX11-FAKE16-NEXT: v_cmp_u_f32_e32 vcc_lo, v52, v52
-; GFX11-FAKE16-NEXT: v_bfe_u32 v4, v9, 16, 1
+; GFX11-FAKE16-NEXT: v_add_nc_u32_e32 v5, 0x7fff, v6
+; GFX11-FAKE16-NEXT: v_add_f32_e64 v17, 0x40c00000, s6
+; GFX11-FAKE16-NEXT: v_or_b32_e32 v6, 0x400000, v10
+; GFX11-FAKE16-NEXT: v_cmp_u_f32_e32 vcc_lo, v10, v10
+; GFX11-FAKE16-NEXT: v_add_nc_u32_e32 v10, v14, v16
+; GFX11-FAKE16-NEXT: s_lshr_b32 s24, s5, 16
+; GFX11-FAKE16-NEXT: v_readfirstlane_b32 s6, v17
+; GFX11-FAKE16-NEXT: v_or_b32_e32 v14, 0x400000, v11
+; GFX11-FAKE16-NEXT: v_cndmask_b32_e32 v5, v5, v6, vcc_lo
+; GFX11-FAKE16-NEXT: v_cmp_u_f32_e32 vcc_lo, v17, v17
+; GFX11-FAKE16-NEXT: v_bfe_u32 v6, v11, 16, 1
; GFX11-FAKE16-NEXT: s_bfe_u32 s7, s6, 0x10010
-; GFX11-FAKE16-NEXT: v_lshrrev_b32_e32 v1, 16, v1
+; GFX11-FAKE16-NEXT: v_and_b32_e32 v15, 0xffff, v50
; GFX11-FAKE16-NEXT: s_add_i32 s7, s7, s6
; GFX11-FAKE16-NEXT: s_bitset1_b32 s6, 22
; GFX11-FAKE16-NEXT: s_addk_i32 s7, 0x7fff
-; GFX11-FAKE16-NEXT: s_and_b32 s4, vcc_lo, exec_lo
-; GFX11-FAKE16-NEXT: s_cselect_b32 s4, s6, s7
+; GFX11-FAKE16-NEXT: s_and_b32 s5, vcc_lo, exec_lo
+; GFX11-FAKE16-NEXT: s_cselect_b32 s5, s6, s7
; GFX11-FAKE16-NEXT: s_lshl_b32 s6, s27, 16
-; GFX11-FAKE16-NEXT: v_lshrrev_b32_e32 v51, 16, v3
-; GFX11-FAKE16-NEXT: v_add_nc_u32_e32 v2, v4, v9
-; GFX11-FAKE16-NEXT: v_add_nc_u32_e32 v3, 0x7fff, v8
-; GFX11-FAKE16-NEXT: v_or_b32_e32 v4, 0x400000, v12
-; GFX11-FAKE16-NEXT: v_add_f32_e64 v8, 0x40c00000, s6
-; GFX11-FAKE16-NEXT: v_cmp_u_f32_e32 vcc_lo, v12, v12
-; GFX11-FAKE16-NEXT: v_add_nc_u32_e32 v2, 0x7fff, v2
-; GFX11-FAKE16-NEXT: s_lshr_b32 s73, s4, 16
-; GFX11-FAKE16-NEXT: v_and_b32_e32 v11, 0xffff, v49
-; GFX11-FAKE16-NEXT: v_readfirstlane_b32 s6, v8
-; GFX11-FAKE16-NEXT: v_cndmask_b32_e32 v3, v3, v4, vcc_lo
-; GFX11-FAKE16-NEXT: v_cmp_u_f32_e32 vcc_lo, v9, v9
-; GFX11-FAKE16-NEXT: v_and_b32_e32 v4, 0xffff, v51
-; GFX11-FAKE16-NEXT: v_lshl_or_b32 v66, v1, 16, v11
+; GFX11-FAKE16-NEXT: v_lshrrev_b32_e32 v52, 16, v5
+; GFX11-FAKE16-NEXT: v_add_nc_u32_e32 v4, v6, v11
+; GFX11-FAKE16-NEXT: v_add_nc_u32_e32 v5, 0x7fff, v10
+; GFX11-FAKE16-NEXT: v_or_b32_e32 v6, 0x400000, v16
+; GFX11-FAKE16-NEXT: v_add_f32_e64 v10, 0x40c00000, s6
+; GFX11-FAKE16-NEXT: v_cmp_u_f32_e32 vcc_lo, v16, v16
+; GFX11-FAKE16-NEXT: v_add_nc_u32_e32 v4, 0x7fff, v4
+; GFX11-FAKE16-NEXT: s_lshr_b32 s73, s5, 16
+; GFX11-FAKE16-NEXT: v_lshl_or_b32 v66, v3, 16, v15
+; GFX11-FAKE16-NEXT: v_readfirstlane_b32 s6, v10
+; GFX11-FAKE16-NEXT: v_cndmask_b32_e32 v5, v5, v6, vcc_lo
+; GFX11-FAKE16-NEXT: v_cmp_u_f32_e32 vcc_lo, v11, v11
+; GFX11-FAKE16-NEXT: v_and_b32_e32 v6, 0xffff, v52
+; GFX11-FAKE16-NEXT: v_and_b32_e32 v7, 0xffff, v48
; GFX11-FAKE16-NEXT: s_bfe_u32 s7, s6, 0x10010
-; GFX11-FAKE16-NEXT: v_lshrrev_b32_e32 v52, 16, v3
-; GFX11-FAKE16-NEXT: v_cndmask_b32_e32 v2, v2, v10, vcc_lo
-; GFX11-FAKE16-NEXT: v_cmp_u_f32_e32 vcc_lo, v8, v8
+; GFX11-FAKE16-NEXT: v_lshrrev_b32_e32 v53, 16, v5
+; GFX11-FAKE16-NEXT: v_cndmask_b32_e32 v4, v4, v14, vcc_lo
+; GFX11-FAKE16-NEXT: v_cmp_u_f32_e32 vcc_lo, v10, v10
; GFX11-FAKE16-NEXT: s_add_i32 s7, s7, s6
; GFX11-FAKE16-NEXT: s_bitset1_b32 s6, 22
; GFX11-FAKE16-NEXT: s_addk_i32 s7, 0x7fff
-; GFX11-FAKE16-NEXT: v_lshrrev_b32_e32 v2, 16, v2
-; GFX11-FAKE16-NEXT: s_and_b32 s4, vcc_lo, exec_lo
-; GFX11-FAKE16-NEXT: s_cselect_b32 s4, s6, s7
-; GFX11-FAKE16-NEXT: s_and_b32 s6, s26, 0xffff0000
-; GFX11-FAKE16-NEXT: s_lshr_b32 s27, s4, 16
-; GFX11-FAKE16-NEXT: v_add_f32_e64 v3, 0x40c00000, s6
-; GFX11-FAKE16-NEXT: v_and_b32_e32 v8, 0xffff, v52
-; GFX11-FAKE16-NEXT: v_and_b32_e32 v5, 0xffff, v39
-; GFX11-FAKE16-NEXT: v_lshl_or_b32 v55, v50, 16, v4
-; GFX11-FAKE16-NEXT: s_pack_ll_b32_b16 s8, s22, s13
-; GFX11-FAKE16-NEXT: v_readfirstlane_b32 s6, v3
+; GFX11-FAKE16-NEXT: v_lshrrev_b32_e32 v4, 16, v4
+; GFX11-FAKE16-NEXT: s_and_b32 s5, vcc_lo, exec_lo
+; GFX11-FAKE16-NEXT: s_cselect_b32 s6, s6, s7
+; GFX11-FAKE16-NEXT: s_and_b32 s5, s26, 0xffff0000
+; GFX11-FAKE16-NEXT: s_lshr_b32 s27, s6, 16
+; GFX11-FAKE16-NEXT: v_add_f32_e64 v5, 0x40c00000, s5
+; GFX11-FAKE16-NEXT: v_and_b32_e32 v10, 0xffff, v53
+; GFX11-FAKE16-NEXT: s_pack_ll_b32_b16 s14, s22, s28
+; GFX11-FAKE16-NEXT: v_lshl_or_b32 v55, v51, 16, v6
+; GFX11-FAKE16-NEXT: v_lshl_or_b32 v67, v49, 16, v7
+; GFX11-FAKE16-NEXT: v_readfirstlane_b32 s8, v5
+; GFX11-FAKE16-NEXT: v_cmp_u_f32_e32 vcc_lo, v5, v5
+; GFX11-FAKE16-NEXT: v_lshl_or_b32 v54, v4, 16, v10
+; GFX11-FAKE16-NEXT: s_pack_ll_b32_b16 s12, s18, s40
+; GFX11-FAKE16-NEXT: v_lshrrev_b64 v[14:15], 24, v[12:13]
+; GFX11-FAKE16-NEXT: s_bfe_u32 s9, s8, 0x10010
+; GFX11-FAKE16-NEXT: v_lshrrev_b64 v[15:16], 24, v[8:9]
+; GFX11-FAKE16-NEXT: s_add_i32 s9, s9, s8
+; GFX11-FAKE16-NEXT: s_bitset1_b32 s8, 22
+; GFX11-FAKE16-NEXT: s_addk_i32 s9, 0x7fff
+; GFX11-FAKE16-NEXT: s_and_b32 s6, vcc_lo, exec_lo
+; GFX11-FAKE16-NEXT: s_cselect_b32 s29, s8, s9
+; GFX11-FAKE16-NEXT: s_lshl_b32 s8, s26, 16
+; GFX11-FAKE16-NEXT: s_lshr_b32 s40, s29, 16
+; GFX11-FAKE16-NEXT: v_add_f32_e64 v3, 0x40c00000, s8
+; GFX11-FAKE16-NEXT: v_lshrrev_b64 v[10:11], 24, v[18:19]
+; GFX11-FAKE16-NEXT: v_lshrrev_b64 v[16:17], 24, v[1:2]
+; GFX11-FAKE16-NEXT: s_pack_ll_b32_b16 s5, s1, s61
+; GFX11-FAKE16-NEXT: s_pack_ll_b32_b16 s7, s3, s62
+; GFX11-FAKE16-NEXT: v_readfirstlane_b32 s26, v3
; GFX11-FAKE16-NEXT: v_cmp_u_f32_e32 vcc_lo, v3, v3
-; GFX11-FAKE16-NEXT: v_lshl_or_b32 v54, v2, 16, v8
-; GFX11-FAKE16-NEXT: v_lshl_or_b32 v67, v48, 16, v5
-; GFX11-FAKE16-NEXT: v_lshrrev_b64 v[8:9], 24, v[17:18]
-; GFX11-FAKE16-NEXT: s_bfe_u32 s5, s6, 0x10010
-; GFX11-FAKE16-NEXT: v_lshrrev_b64 v[9:10], 24, v[15:16]
-; GFX11-FAKE16-NEXT: s_add_i32 s5, s5, s6
-; GFX11-FAKE16-NEXT: s_bitset1_b32 s6, 22
-; GFX11-FAKE16-NEXT: s_addk_i32 s5, 0x7fff
-; GFX11-FAKE16-NEXT: s_and_b32 s4, vcc_lo, exec_lo
-; GFX11-FAKE16-NEXT: s_cselect_b32 s14, s6, s5
-; GFX11-FAKE16-NEXT: s_lshl_b32 s4, s26, 16
-; GFX11-FAKE16-NEXT: s_pack_ll_b32_b16 s6, s20, s10
-; GFX11-FAKE16-NEXT: v_add_f32_e64 v1, 0x40c00000, s4
-; GFX11-FAKE16-NEXT: s_lshr_b32 s13, s14, 16
-; GFX11-FAKE16-NEXT: v_lshrrev_b64 v[10:11], 24, v[13:14]
-; GFX11-FAKE16-NEXT: v_lshrrev_b64 v[11:12], 24, v[6:7]
-; GFX11-FAKE16-NEXT: s_pack_ll_b32_b16 s29, s1, s58
-; GFX11-FAKE16-NEXT: v_readfirstlane_b32 s11, v1
-; GFX11-FAKE16-NEXT: v_cmp_u_f32_e32 vcc_lo, v1, v1
-; GFX11-FAKE16-NEXT: v_lshrrev_b64 v[1:2], 24, v[54:55]
-; GFX11-FAKE16-NEXT: v_lshrrev_b64 v[2:3], 24, v[66:67]
-; GFX11-FAKE16-NEXT: v_lshrrev_b64 v[3:4], 24, v[70:71]
-; GFX11-FAKE16-NEXT: s_bfe_u32 s10, s11, 0x10010
-; GFX11-FAKE16-NEXT: v_lshrrev_b64 v[4:5], 24, v[19:20]
-; GFX11-FAKE16-NEXT: s_add_i32 s10, s10, s11
-; GFX11-FAKE16-NEXT: s_bitset1_b32 s11, 22
-; GFX11-FAKE16-NEXT: s_addk_i32 s10, 0x7fff
-; GFX11-FAKE16-NEXT: s_and_b32 s14, vcc_lo, exec_lo
-; GFX11-FAKE16-NEXT: s_cselect_b32 s10, s11, s10
-; GFX11-FAKE16-NEXT: s_pack_ll_b32_b16 s5, s19, s60
-; GFX11-FAKE16-NEXT: s_lshr_b32 s26, s10, 16
-; GFX11-FAKE16-NEXT: s_pack_ll_b32_b16 s4, s18, s40
-; GFX11-FAKE16-NEXT: s_pack_ll_b32_b16 s9, s23, s62
-; GFX11-FAKE16-NEXT: v_lshrrev_b32_e32 v5, 24, v55
-; GFX11-FAKE16-NEXT: v_lshrrev_b32_e32 v12, 8, v55
-; GFX11-FAKE16-NEXT: v_lshrrev_b32_e32 v53, 16, v54
+; GFX11-FAKE16-NEXT: v_lshrrev_b64 v[3:4], 24, v[54:55]
+; GFX11-FAKE16-NEXT: v_lshrrev_b64 v[4:5], 24, v[66:67]
+; GFX11-FAKE16-NEXT: v_lshrrev_b64 v[5:6], 24, v[70:71]
+; GFX11-FAKE16-NEXT: s_bfe_u32 s28, s26, 0x10010
+; GFX11-FAKE16-NEXT: v_lshrrev_b64 v[6:7], 24, v[20:21]
+; GFX11-FAKE16-NEXT: s_add_i32 s28, s28, s26
+; GFX11-FAKE16-NEXT: s_bitset1_b32 s26, 22
+; GFX11-FAKE16-NEXT: s_addk_i32 s28, 0x7fff
+; GFX11-FAKE16-NEXT: s_and_b32 s29, vcc_lo, exec_lo
+; GFX11-FAKE16-NEXT: s_cselect_b32 s26, s26, s28
+; GFX11-FAKE16-NEXT: s_pack_ll_b32_b16 s6, s2, s11
+; GFX11-FAKE16-NEXT: s_lshr_b32 s26, s26, 16
+; GFX11-FAKE16-NEXT: s_pack_ll_b32_b16 s13, s19, s59
+; GFX11-FAKE16-NEXT: s_pack_ll_b32_b16 s11, s21, s60
+; GFX11-FAKE16-NEXT: s_pack_ll_b32_b16 s10, s20, s10
+; GFX11-FAKE16-NEXT: s_pack_ll_b32_b16 s15, s23, s63
+; GFX11-FAKE16-NEXT: s_pack_ll_b32_b16 s29, s25, s72
+; GFX11-FAKE16-NEXT: s_pack_ll_b32_b16 s41, s27, s73
+; GFX11-FAKE16-NEXT: s_pack_ll_b32_b16 s40, s26, s40
+; GFX11-FAKE16-NEXT: s_pack_ll_b32_b16 s28, s24, s42
+; GFX11-FAKE16-NEXT: v_lshrrev_b32_e32 v7, 24, v55
+; GFX11-FAKE16-NEXT: v_lshrrev_b32_e32 v11, 8, v55
+; GFX11-FAKE16-NEXT: v_lshrrev_b32_e32 v17, 16, v54
; GFX11-FAKE16-NEXT: v_lshrrev_b32_e32 v54, 8, v54
; GFX11-FAKE16-NEXT: v_lshrrev_b32_e32 v55, 24, v67
; GFX11-FAKE16-NEXT: v_lshrrev_b32_e32 v64, 8, v67
@@ -178022,120 +178126,114 @@ define inreg <128 x i8> @bitcast_v64bf16_to_v128i8_scalar(<64 x bfloat> inreg %a
; GFX11-FAKE16-NEXT: v_lshrrev_b32_e32 v68, 8, v71
; GFX11-FAKE16-NEXT: v_lshrrev_b32_e32 v69, 16, v70
; GFX11-FAKE16-NEXT: v_lshrrev_b32_e32 v70, 8, v70
-; GFX11-FAKE16-NEXT: v_lshrrev_b32_e32 v71, 24, v20
+; GFX11-FAKE16-NEXT: v_lshrrev_b32_e32 v71, 24, v21
+; GFX11-FAKE16-NEXT: v_lshrrev_b32_e32 v21, 8, v21
; GFX11-FAKE16-NEXT: v_lshrrev_b32_e32 v20, 8, v20
; GFX11-FAKE16-NEXT: v_lshrrev_b32_e32 v19, 8, v19
; GFX11-FAKE16-NEXT: v_lshrrev_b32_e32 v18, 8, v18
-; GFX11-FAKE16-NEXT: v_lshrrev_b32_e32 v82, 16, v17
-; GFX11-FAKE16-NEXT: v_lshrrev_b32_e32 v17, 8, v17
-; GFX11-FAKE16-NEXT: v_lshrrev_b32_e32 v16, 8, v16
-; GFX11-FAKE16-NEXT: v_lshrrev_b32_e32 v84, 16, v15
-; GFX11-FAKE16-NEXT: v_lshrrev_b32_e32 v15, 8, v15
-; GFX11-FAKE16-NEXT: v_lshrrev_b32_e32 v14, 8, v14
; GFX11-FAKE16-NEXT: v_lshrrev_b32_e32 v13, 8, v13
-; GFX11-FAKE16-NEXT: v_lshrrev_b32_e32 v7, 8, v7
-; GFX11-FAKE16-NEXT: v_lshrrev_b32_e32 v6, 8, v6
-; GFX11-FAKE16-NEXT: s_pack_ll_b32_b16 s7, s21, s61
-; GFX11-FAKE16-NEXT: s_pack_ll_b32_b16 s11, s25, s63
-; GFX11-FAKE16-NEXT: s_pack_ll_b32_b16 s57, s27, s73
-; GFX11-FAKE16-NEXT: s_pack_ll_b32_b16 s56, s26, s13
-; GFX11-FAKE16-NEXT: s_pack_ll_b32_b16 s10, s24, s12
-; GFX11-FAKE16-NEXT: s_lshr_b64 s[94:95], s[8:9], 24
-; GFX11-FAKE16-NEXT: s_lshr_b64 s[12:13], s[4:5], 24
-; GFX11-FAKE16-NEXT: s_lshr_b64 s[14:15], s[46:47], 24
-; GFX11-FAKE16-NEXT: s_lshr_b64 s[40:41], s[44:45], 24
-; GFX11-FAKE16-NEXT: s_lshr_b64 s[42:43], s[28:29], 24
-; GFX11-FAKE16-NEXT: s_lshr_b64 vcc, s[56:57], 24
-; GFX11-FAKE16-NEXT: s_lshr_b64 s[34:35], s[10:11], 24
-; GFX11-FAKE16-NEXT: s_lshr_b64 s[30:31], s[6:7], 24
-; GFX11-FAKE16-NEXT: s_lshr_b32 s13, s57, 24
-; GFX11-FAKE16-NEXT: s_lshr_b32 s15, s57, 8
-; GFX11-FAKE16-NEXT: s_lshr_b32 s41, s56, 16
-; GFX11-FAKE16-NEXT: s_lshr_b32 s43, s56, 8
-; GFX11-FAKE16-NEXT: s_lshr_b32 s56, s11, 24
+; GFX11-FAKE16-NEXT: v_lshrrev_b32_e32 v12, 8, v12
+; GFX11-FAKE16-NEXT: v_lshrrev_b32_e32 v9, 8, v9
+; GFX11-FAKE16-NEXT: v_lshrrev_b32_e32 v8, 8, v8
+; GFX11-FAKE16-NEXT: v_lshrrev_b32_e32 v87, 24, v2
+; GFX11-FAKE16-NEXT: v_lshrrev_b32_e32 v2, 8, v2
+; GFX11-FAKE16-NEXT: v_lshrrev_b32_e32 v96, 16, v1
+; GFX11-FAKE16-NEXT: v_lshrrev_b32_e32 v1, 8, v1
+; GFX11-FAKE16-NEXT: s_pack_ll_b32_b16 s9, s17, s58
+; GFX11-FAKE16-NEXT: s_pack_ll_b32_b16 s8, s16, s45
+; GFX11-FAKE16-NEXT: s_lshr_b64 s[92:93], s[40:41], 24
+; GFX11-FAKE16-NEXT: s_lshr_b64 vcc, s[28:29], 24
+; GFX11-FAKE16-NEXT: s_lshr_b64 s[90:91], s[14:15], 24
+; GFX11-FAKE16-NEXT: s_lshr_b64 s[94:95], s[10:11], 24
+; GFX11-FAKE16-NEXT: s_lshr_b64 s[42:43], s[12:13], 24
+; GFX11-FAKE16-NEXT: s_lshr_b64 s[46:47], s[6:7], 24
+; GFX11-FAKE16-NEXT: s_lshr_b64 s[56:57], s[4:5], 24
+; GFX11-FAKE16-NEXT: s_lshr_b64 s[30:31], s[8:9], 24
+; GFX11-FAKE16-NEXT: s_lshr_b32 s43, s41, 24
+; GFX11-FAKE16-NEXT: s_lshr_b32 s41, s41, 8
+; GFX11-FAKE16-NEXT: s_lshr_b32 s45, s40, 16
+; GFX11-FAKE16-NEXT: s_lshr_b32 s40, s40, 8
+; GFX11-FAKE16-NEXT: s_lshr_b32 s47, s29, 24
+; GFX11-FAKE16-NEXT: s_lshr_b32 s29, s29, 8
+; GFX11-FAKE16-NEXT: s_lshr_b32 s57, s28, 16
+; GFX11-FAKE16-NEXT: s_lshr_b32 s28, s28, 8
+; GFX11-FAKE16-NEXT: s_lshr_b32 s74, s15, 24
+; GFX11-FAKE16-NEXT: s_lshr_b32 s15, s15, 8
+; GFX11-FAKE16-NEXT: s_lshr_b32 s75, s14, 16
+; GFX11-FAKE16-NEXT: s_lshr_b32 s14, s14, 8
+; GFX11-FAKE16-NEXT: s_lshr_b32 s76, s11, 24
; GFX11-FAKE16-NEXT: s_lshr_b32 s11, s11, 8
-; GFX11-FAKE16-NEXT: s_lshr_b32 s57, s10, 16
+; GFX11-FAKE16-NEXT: s_lshr_b32 s77, s10, 16
; GFX11-FAKE16-NEXT: s_lshr_b32 s10, s10, 8
-; GFX11-FAKE16-NEXT: s_lshr_b32 s74, s9, 24
+; GFX11-FAKE16-NEXT: s_lshr_b32 s78, s13, 24
+; GFX11-FAKE16-NEXT: s_lshr_b32 s13, s13, 8
+; GFX11-FAKE16-NEXT: s_lshr_b32 s79, s12, 16
+; GFX11-FAKE16-NEXT: s_lshr_b32 s12, s12, 8
+; GFX11-FAKE16-NEXT: s_lshr_b32 s88, s9, 24
; GFX11-FAKE16-NEXT: s_lshr_b32 s9, s9, 8
-; GFX11-FAKE16-NEXT: s_lshr_b32 s75, s8, 16
+; GFX11-FAKE16-NEXT: s_lshr_b32 s89, s8, 16
; GFX11-FAKE16-NEXT: s_lshr_b32 s8, s8, 8
-; GFX11-FAKE16-NEXT: s_lshr_b32 s76, s7, 24
-; GFX11-FAKE16-NEXT: s_lshr_b32 s77, s7, 8
-; GFX11-FAKE16-NEXT: s_lshr_b32 s78, s6, 16
-; GFX11-FAKE16-NEXT: s_lshr_b32 s79, s6, 8
-; GFX11-FAKE16-NEXT: s_lshr_b32 s88, s5, 24
-; GFX11-FAKE16-NEXT: s_lshr_b32 s89, s5, 8
-; GFX11-FAKE16-NEXT: s_lshr_b32 s90, s4, 16
-; GFX11-FAKE16-NEXT: s_lshr_b32 s91, s4, 8
-; GFX11-FAKE16-NEXT: s_lshr_b32 s92, s47, 24
-; GFX11-FAKE16-NEXT: s_lshr_b32 s47, s47, 8
-; GFX11-FAKE16-NEXT: s_lshr_b32 s93, s46, 16
-; GFX11-FAKE16-NEXT: s_lshr_b32 s46, s46, 8
-; GFX11-FAKE16-NEXT: s_lshr_b32 s95, s45, 24
-; GFX11-FAKE16-NEXT: s_lshr_b32 s45, s45, 8
-; GFX11-FAKE16-NEXT: s_lshr_b32 s99, s44, 16
-; GFX11-FAKE16-NEXT: s_lshr_b32 s100, s44, 8
-; GFX11-FAKE16-NEXT: s_lshr_b32 s101, s29, 24
-; GFX11-FAKE16-NEXT: s_lshr_b32 s102, s29, 8
-; GFX11-FAKE16-NEXT: s_lshr_b32 s103, s28, 16
-; GFX11-FAKE16-NEXT: s_lshr_b32 s104, s28, 8
+; GFX11-FAKE16-NEXT: s_lshr_b32 s91, s7, 24
+; GFX11-FAKE16-NEXT: s_lshr_b32 s95, s7, 8
+; GFX11-FAKE16-NEXT: s_lshr_b32 s93, s6, 16
+; GFX11-FAKE16-NEXT: s_lshr_b32 vcc_hi, s6, 8
+; GFX11-FAKE16-NEXT: s_lshr_b32 s7, s5, 24
+; GFX11-FAKE16-NEXT: s_lshr_b32 s99, s5, 8
+; GFX11-FAKE16-NEXT: s_lshr_b32 s44, s4, 16
+; GFX11-FAKE16-NEXT: s_lshr_b32 s104, s4, 8
; GFX11-FAKE16-NEXT: s_branch .LBB91_5
; GFX11-FAKE16-NEXT: .LBB91_3:
-; GFX11-FAKE16-NEXT: ; implicit-def: $sgpr4
-; GFX11-FAKE16-NEXT: ; kill: killed $sgpr4
-; GFX11-FAKE16-NEXT: ; implicit-def: $sgpr74
-; GFX11-FAKE16-NEXT: ; implicit-def: $sgpr4
-; GFX11-FAKE16-NEXT: ; kill: killed $sgpr4
+; GFX11-FAKE16-NEXT: ; implicit-def: $sgpr10
+; GFX11-FAKE16-NEXT: ; kill: killed $sgpr10
+; GFX11-FAKE16-NEXT: ; implicit-def: $sgpr10
+; GFX11-FAKE16-NEXT: ; kill: killed $sgpr10
; GFX11-FAKE16-NEXT: ; implicit-def: $sgpr104
-; GFX11-FAKE16-NEXT: ; implicit-def: $sgpr103
-; GFX11-FAKE16-NEXT: ; implicit-def: $sgpr42
-; GFX11-FAKE16-NEXT: ; implicit-def: $sgpr102
-; GFX11-FAKE16-NEXT: ; implicit-def: $sgpr11
-; GFX11-FAKE16-NEXT: ; implicit-def: $sgpr101
-; GFX11-FAKE16-NEXT: ; implicit-def: $sgpr100
+; GFX11-FAKE16-NEXT: ; implicit-def: $sgpr44
+; GFX11-FAKE16-NEXT: ; implicit-def: $sgpr56
; GFX11-FAKE16-NEXT: ; implicit-def: $sgpr99
-; GFX11-FAKE16-NEXT: ; implicit-def: $sgpr40
-; GFX11-FAKE16-NEXT: ; implicit-def: $sgpr14
-; GFX11-FAKE16-NEXT: ; implicit-def: $sgpr12
-; GFX11-FAKE16-NEXT: ; implicit-def: $sgpr49
+; GFX11-FAKE16-NEXT: ; implicit-def: $sgpr41
+; GFX11-FAKE16-NEXT: ; implicit-def: $sgpr46
+; GFX11-FAKE16-NEXT: ; implicit-def: $sgpr42
; GFX11-FAKE16-NEXT: ; implicit-def: $sgpr37
+; GFX11-FAKE16-NEXT: ; implicit-def: $sgpr49
; GFX11-FAKE16-NEXT: ; implicit-def: $sgpr35
-; GFX11-FAKE16-NEXT: ; implicit-def: $sgpr6
+; GFX11-FAKE16-NEXT: ; implicit-def: $sgpr12
; GFX11-FAKE16-NEXT: ; implicit-def: $sgpr34
+; GFX11-FAKE16-NEXT: ; implicit-def: $sgpr51
; GFX11-FAKE16-NEXT: ; implicit-def: $sgpr52
-; GFX11-FAKE16-NEXT: ; implicit-def: $sgpr53
; GFX11-FAKE16-NEXT: ; implicit-def: $sgpr50
-; GFX11-FAKE16-NEXT: ; implicit-def: $sgpr7
+; GFX11-FAKE16-NEXT: ; implicit-def: $sgpr13
; GFX11-FAKE16-NEXT: ; implicit-def: $sgpr36
+; GFX11-FAKE16-NEXT: ; implicit-def: $sgpr55
; GFX11-FAKE16-NEXT: ; implicit-def: $sgpr64
-; GFX11-FAKE16-NEXT: ; implicit-def: $sgpr38
-; GFX11-FAKE16-NEXT: ; implicit-def: $sgpr54
+; GFX11-FAKE16-NEXT: ; implicit-def: $sgpr53
; GFX11-FAKE16-NEXT: ; implicit-def: $sgpr96
-; GFX11-FAKE16-NEXT: ; implicit-def: $sgpr51
+; GFX11-FAKE16-NEXT: ; implicit-def: $sgpr38
+; GFX11-FAKE16-NEXT: ; implicit-def: $sgpr66
; GFX11-FAKE16-NEXT: ; implicit-def: $sgpr67
-; GFX11-FAKE16-NEXT: ; implicit-def: $sgpr68
-; GFX11-FAKE16-NEXT: ; implicit-def: $sgpr65
-; GFX11-FAKE16-NEXT: ; implicit-def: $sgpr8
-; GFX11-FAKE16-NEXT: ; implicit-def: $sgpr55
; GFX11-FAKE16-NEXT: ; implicit-def: $sgpr39
+; GFX11-FAKE16-NEXT: ; implicit-def: $sgpr14
+; GFX11-FAKE16-NEXT: ; implicit-def: $sgpr54
+; GFX11-FAKE16-NEXT: ; implicit-def: $sgpr70
; GFX11-FAKE16-NEXT: ; implicit-def: $sgpr71
-; GFX11-FAKE16-NEXT: ; implicit-def: $sgpr69
-; GFX11-FAKE16-NEXT: ; implicit-def: $sgpr9
-; GFX11-FAKE16-NEXT: ; implicit-def: $sgpr66
+; GFX11-FAKE16-NEXT: ; implicit-def: $sgpr68
+; GFX11-FAKE16-NEXT: ; implicit-def: $sgpr15
+; GFX11-FAKE16-NEXT: ; implicit-def: $sgpr65
+; GFX11-FAKE16-NEXT: ; implicit-def: $sgpr81
; GFX11-FAKE16-NEXT: ; implicit-def: $sgpr82
-; GFX11-FAKE16-NEXT: ; implicit-def: $sgpr83
; GFX11-FAKE16-NEXT: ; implicit-def: $sgpr80
; GFX11-FAKE16-NEXT: ; implicit-def: $sgpr97
-; GFX11-FAKE16-NEXT: ; implicit-def: $sgpr70
-; GFX11-FAKE16-NEXT: ; implicit-def: $sgpr48
-; GFX11-FAKE16-NEXT: ; implicit-def: $sgpr84
+; GFX11-FAKE16-NEXT: ; implicit-def: $sgpr69
+; GFX11-FAKE16-NEXT: ; implicit-def: $sgpr85
+; GFX11-FAKE16-NEXT: ; implicit-def: $sgpr83
; GFX11-FAKE16-NEXT: ; implicit-def: $sgpr98
-; GFX11-FAKE16-NEXT: ; implicit-def: $sgpr81
+; GFX11-FAKE16-NEXT: ; implicit-def: $sgpr48
; GFX11-FAKE16-NEXT: ; implicit-def: $sgpr86
; GFX11-FAKE16-NEXT: ; implicit-def: $sgpr87
-; GFX11-FAKE16-NEXT: ; implicit-def: $sgpr10
-; GFX11-FAKE16-NEXT: ; implicit-def: $sgpr85
+; GFX11-FAKE16-NEXT: ; implicit-def: $sgpr40
+; GFX11-FAKE16-NEXT: ; implicit-def: $sgpr84
+; GFX11-FAKE16-NEXT: ; implicit-def: $sgpr102
+; GFX11-FAKE16-NEXT: ; implicit-def: $sgpr100
; GFX11-FAKE16-NEXT: ; implicit-def: $sgpr30
; GFX11-FAKE16-NEXT: ; implicit-def: $sgpr94
; GFX11-FAKE16-NEXT: ; implicit-def: $sgpr92
@@ -178143,169 +178241,176 @@ define inreg <128 x i8> @bitcast_v64bf16_to_v128i8_scalar(<64 x bfloat> inreg %a
; GFX11-FAKE16-NEXT: ; implicit-def: $sgpr88
; GFX11-FAKE16-NEXT: ; implicit-def: $sgpr78
; GFX11-FAKE16-NEXT: ; implicit-def: $sgpr76
-; GFX11-FAKE16-NEXT: ; implicit-def: $sgpr4
-; GFX11-FAKE16-NEXT: ; kill: killed $sgpr4
-; GFX11-FAKE16-NEXT: ; implicit-def: $sgpr4
-; GFX11-FAKE16-NEXT: ; kill: killed $sgpr4
-; GFX11-FAKE16-NEXT: ; implicit-def: $sgpr4
-; GFX11-FAKE16-NEXT: ; kill: killed $sgpr4
-; GFX11-FAKE16-NEXT: ; implicit-def: $sgpr4
-; GFX11-FAKE16-NEXT: ; kill: killed $sgpr4
-; GFX11-FAKE16-NEXT: ; implicit-def: $sgpr4
-; GFX11-FAKE16-NEXT: ; kill: killed $sgpr4
-; GFX11-FAKE16-NEXT: ; implicit-def: $sgpr4
-; GFX11-FAKE16-NEXT: ; kill: killed $sgpr4
-; GFX11-FAKE16-NEXT: ; implicit-def: $sgpr4
-; GFX11-FAKE16-NEXT: ; kill: killed $sgpr4
-; GFX11-FAKE16-NEXT: ; implicit-def: $sgpr4
-; GFX11-FAKE16-NEXT: ; kill: killed $sgpr4
-; GFX11-FAKE16-NEXT: ; implicit-def: $sgpr4
-; GFX11-FAKE16-NEXT: ; kill: killed $sgpr4
-; GFX11-FAKE16-NEXT: ; implicit-def: $sgpr4
-; GFX11-FAKE16-NEXT: ; kill: killed $sgpr4
-; GFX11-FAKE16-NEXT: ; implicit-def: $sgpr4
-; GFX11-FAKE16-NEXT: ; kill: killed $sgpr4
-; GFX11-FAKE16-NEXT: ; implicit-def: $sgpr4
-; GFX11-FAKE16-NEXT: ; kill: killed $sgpr4
-; GFX11-FAKE16-NEXT: ; implicit-def: $sgpr4
-; GFX11-FAKE16-NEXT: ; kill: killed $sgpr4
-; GFX11-FAKE16-NEXT: ; implicit-def: $sgpr4
-; GFX11-FAKE16-NEXT: v_writelane_b32 v43, s4, 0
-; GFX11-FAKE16-NEXT: v_writelane_b32 v43, s5, 1
-; GFX11-FAKE16-NEXT: ; implicit-def: $sgpr4
-; GFX11-FAKE16-NEXT: ; kill: killed $sgpr4
-; GFX11-FAKE16-NEXT: ; implicit-def: $sgpr4
-; GFX11-FAKE16-NEXT: ; kill: killed $sgpr4
-; GFX11-FAKE16-NEXT: ; implicit-def: $sgpr4
-; GFX11-FAKE16-NEXT: ; kill: killed $sgpr4
-; GFX11-FAKE16-NEXT: ; implicit-def: $sgpr4
-; GFX11-FAKE16-NEXT: ; kill: killed $sgpr4
-; GFX11-FAKE16-NEXT: ; implicit-def: $sgpr4
-; GFX11-FAKE16-NEXT: ; kill: killed $sgpr4
-; GFX11-FAKE16-NEXT: ; implicit-def: $sgpr4
-; GFX11-FAKE16-NEXT: v_writelane_b32 v43, s4, 2
-; GFX11-FAKE16-NEXT: v_writelane_b32 v43, s5, 3
-; GFX11-FAKE16-NEXT: ; implicit-def: $sgpr4
-; GFX11-FAKE16-NEXT: ; kill: killed $sgpr4
-; GFX11-FAKE16-NEXT: ; implicit-def: $sgpr4
-; GFX11-FAKE16-NEXT: ; kill: killed $sgpr4
-; GFX11-FAKE16-NEXT: ; implicit-def: $sgpr4
-; GFX11-FAKE16-NEXT: ; kill: killed $sgpr4
-; GFX11-FAKE16-NEXT: ; implicit-def: $sgpr4
-; GFX11-FAKE16-NEXT: ; kill: killed $sgpr4
-; GFX11-FAKE16-NEXT: ; implicit-def: $sgpr4
-; GFX11-FAKE16-NEXT: v_writelane_b32 v43, s74, 4
-; GFX11-FAKE16-NEXT: ; kill: killed $sgpr4
-; GFX11-FAKE16-NEXT: ; implicit-def: $sgpr4
-; GFX11-FAKE16-NEXT: ; kill: killed $sgpr4
-; GFX11-FAKE16-NEXT: ; implicit-def: $sgpr4
-; GFX11-FAKE16-NEXT: ; kill: killed $sgpr4
-; GFX11-FAKE16-NEXT: ; implicit-def: $sgpr4
-; GFX11-FAKE16-NEXT: ; kill: killed $sgpr4
-; GFX11-FAKE16-NEXT: ; implicit-def: $sgpr4
-; GFX11-FAKE16-NEXT: ; kill: killed $sgpr4
-; GFX11-FAKE16-NEXT: ; implicit-def: $sgpr4
-; GFX11-FAKE16-NEXT: v_writelane_b32 v43, s75, 5
-; GFX11-FAKE16-NEXT: ; implicit-def: $sgpr74
-; GFX11-FAKE16-NEXT: ; kill: killed $sgpr4
-; GFX11-FAKE16-NEXT: ; implicit-def: $sgpr4
-; GFX11-FAKE16-NEXT: ; kill: killed $sgpr4
-; GFX11-FAKE16-NEXT: ; implicit-def: $sgpr4
-; GFX11-FAKE16-NEXT: ; kill: killed $sgpr4
-; GFX11-FAKE16-NEXT: ; implicit-def: $sgpr4
-; GFX11-FAKE16-NEXT: ; kill: killed $sgpr4
-; GFX11-FAKE16-NEXT: ; implicit-def: $sgpr4
-; GFX11-FAKE16-NEXT: ; implicit-def: $sgpr5
-; GFX11-FAKE16-NEXT: v_writelane_b32 v43, s74, 6
-; GFX11-FAKE16-NEXT: v_writelane_b32 v43, s75, 7
; GFX11-FAKE16-NEXT: ; implicit-def: $sgpr74
+; GFX11-FAKE16-NEXT: ; implicit-def: $sgpr10
+; GFX11-FAKE16-NEXT: ; kill: killed $sgpr10
+; GFX11-FAKE16-NEXT: ; implicit-def: $sgpr10
+; GFX11-FAKE16-NEXT: ; kill: killed $sgpr10
+; GFX11-FAKE16-NEXT: ; implicit-def: $sgpr10
+; GFX11-FAKE16-NEXT: ; kill: killed $sgpr10
+; GFX11-FAKE16-NEXT: ; implicit-def: $sgpr10
+; GFX11-FAKE16-NEXT: ; kill: killed $sgpr10
+; GFX11-FAKE16-NEXT: ; implicit-def: $sgpr10
+; GFX11-FAKE16-NEXT: ; kill: killed $sgpr10
+; GFX11-FAKE16-NEXT: ; implicit-def: $sgpr10
+; GFX11-FAKE16-NEXT: ; kill: killed $sgpr10
+; GFX11-FAKE16-NEXT: ; implicit-def: $sgpr10
+; GFX11-FAKE16-NEXT: v_writelane_b32 v42, s10, 0
+; GFX11-FAKE16-NEXT: v_writelane_b32 v42, s11, 1
+; GFX11-FAKE16-NEXT: ; implicit-def: $sgpr10
+; GFX11-FAKE16-NEXT: ; kill: killed $sgpr10
+; GFX11-FAKE16-NEXT: ; implicit-def: $sgpr10
+; GFX11-FAKE16-NEXT: ; kill: killed $sgpr10
+; GFX11-FAKE16-NEXT: ; implicit-def: $sgpr10
+; GFX11-FAKE16-NEXT: ; kill: killed $sgpr10
+; GFX11-FAKE16-NEXT: ; implicit-def: $sgpr10
+; GFX11-FAKE16-NEXT: ; kill: killed $sgpr10
+; GFX11-FAKE16-NEXT: ; implicit-def: $sgpr10
+; GFX11-FAKE16-NEXT: ; kill: killed $sgpr10
+; GFX11-FAKE16-NEXT: ; implicit-def: $sgpr10
+; GFX11-FAKE16-NEXT: ; kill: killed $sgpr10
+; GFX11-FAKE16-NEXT: ; implicit-def: $sgpr10
+; GFX11-FAKE16-NEXT: ; kill: killed $sgpr10
+; GFX11-FAKE16-NEXT: ; implicit-def: $sgpr10
+; GFX11-FAKE16-NEXT: ; kill: killed $sgpr10
+; GFX11-FAKE16-NEXT: ; implicit-def: $sgpr10
+; GFX11-FAKE16-NEXT: ; kill: killed $sgpr10
+; GFX11-FAKE16-NEXT: ; implicit-def: $sgpr10
+; GFX11-FAKE16-NEXT: ; kill: killed $sgpr10
+; GFX11-FAKE16-NEXT: ; implicit-def: $sgpr10
+; GFX11-FAKE16-NEXT: v_writelane_b32 v42, s10, 2
+; GFX11-FAKE16-NEXT: v_writelane_b32 v42, s11, 3
+; GFX11-FAKE16-NEXT: ; implicit-def: $sgpr10
+; GFX11-FAKE16-NEXT: ; kill: killed $sgpr10
+; GFX11-FAKE16-NEXT: ; implicit-def: $sgpr10
+; GFX11-FAKE16-NEXT: ; kill: killed $sgpr10
+; GFX11-FAKE16-NEXT: ; implicit-def: $sgpr10
+; GFX11-FAKE16-NEXT: ; kill: killed $sgpr10
+; GFX11-FAKE16-NEXT: ; implicit-def: $sgpr10
+; GFX11-FAKE16-NEXT: ; kill: killed $sgpr10
+; GFX11-FAKE16-NEXT: ; implicit-def: $sgpr10
+; GFX11-FAKE16-NEXT: ; kill: killed $sgpr10
+; GFX11-FAKE16-NEXT: ; implicit-def: $sgpr10
+; GFX11-FAKE16-NEXT: v_writelane_b32 v42, s10, 4
+; GFX11-FAKE16-NEXT: v_writelane_b32 v42, s11, 5
+; GFX11-FAKE16-NEXT: ; implicit-def: $sgpr10
+; GFX11-FAKE16-NEXT: ; kill: killed $sgpr10
+; GFX11-FAKE16-NEXT: ; implicit-def: $sgpr10
+; GFX11-FAKE16-NEXT: ; kill: killed $sgpr10
+; GFX11-FAKE16-NEXT: ; implicit-def: $sgpr10
+; GFX11-FAKE16-NEXT: ; kill: killed $sgpr10
+; GFX11-FAKE16-NEXT: ; implicit-def: $sgpr10
+; GFX11-FAKE16-NEXT: ; kill: killed $sgpr10
+; GFX11-FAKE16-NEXT: ; implicit-def: $sgpr10
+; GFX11-FAKE16-NEXT: ; kill: killed $sgpr10
+; GFX11-FAKE16-NEXT: ; implicit-def: $sgpr10
+; GFX11-FAKE16-NEXT: ; kill: killed $sgpr10
+; GFX11-FAKE16-NEXT: ; implicit-def: $sgpr10
+; GFX11-FAKE16-NEXT: ; kill: killed $sgpr10
+; GFX11-FAKE16-NEXT: ; implicit-def: $sgpr10
+; GFX11-FAKE16-NEXT: ; kill: killed $sgpr10
+; GFX11-FAKE16-NEXT: ; implicit-def: $sgpr10
+; GFX11-FAKE16-NEXT: ; kill: killed $sgpr10
+; GFX11-FAKE16-NEXT: ; implicit-def: $sgpr10
+; GFX11-FAKE16-NEXT: ; kill: killed $sgpr10
+; GFX11-FAKE16-NEXT: ; implicit-def: $sgpr10
+; GFX11-FAKE16-NEXT: ; kill: killed $sgpr10
+; GFX11-FAKE16-NEXT: ; implicit-def: $sgpr10
+; GFX11-FAKE16-NEXT: ; kill: killed $sgpr10
+; GFX11-FAKE16-NEXT: ; implicit-def: $sgpr10
+; GFX11-FAKE16-NEXT: ; kill: killed $sgpr10
+; GFX11-FAKE16-NEXT: ; implicit-def: $sgpr10
+; GFX11-FAKE16-NEXT: ; implicit-def: $sgpr11
; GFX11-FAKE16-NEXT: s_branch .LBB91_2
; GFX11-FAKE16-NEXT: .LBB91_4:
-; GFX11-FAKE16-NEXT: v_dual_mov_b32 v10, s94 :: v_dual_mov_b32 v11, s30
-; GFX11-FAKE16-NEXT: v_readlane_b32 s94, v43, 2
-; GFX11-FAKE16-NEXT: v_dual_mov_b32 v96, s37 :: v_dual_mov_b32 v87, s34
-; GFX11-FAKE16-NEXT: v_dual_mov_b32 v6, s49 :: v_dual_mov_b32 v7, s35
-; GFX11-FAKE16-NEXT: v_readlane_b32 s95, v43, 3
-; GFX11-FAKE16-NEXT: v_readlane_b32 vcc_lo, v43, 6
-; GFX11-FAKE16-NEXT: v_readlane_b32 s30, v43, 0
-; GFX11-FAKE16-NEXT: v_readlane_b32 s34, v43, 4
-; GFX11-FAKE16-NEXT: v_dual_mov_b32 v52, s44 :: v_dual_mov_b32 v51, s45
-; GFX11-FAKE16-NEXT: v_dual_mov_b32 v50, s10 :: v_dual_mov_b32 v49, s46
-; GFX11-FAKE16-NEXT: v_dual_mov_b32 v39, s47 :: v_dual_mov_b32 v48, s98
-; GFX11-FAKE16-NEXT: v_dual_mov_b32 v38, s56 :: v_dual_mov_b32 v37, s97
-; GFX11-FAKE16-NEXT: v_dual_mov_b32 v36, s57 :: v_dual_mov_b32 v35, s58
-; GFX11-FAKE16-NEXT: v_dual_mov_b32 v34, s59 :: v_dual_mov_b32 v33, s9
-; GFX11-FAKE16-NEXT: v_dual_mov_b32 v32, s60 :: v_dual_mov_b32 v31, s61
-; GFX11-FAKE16-NEXT: v_dual_mov_b32 v30, s8 :: v_dual_mov_b32 v29, s62
-; GFX11-FAKE16-NEXT: v_dual_mov_b32 v27, s63 :: v_dual_mov_b32 v28, s96
-; GFX11-FAKE16-NEXT: v_dual_mov_b32 v26, s72 :: v_dual_mov_b32 v25, s7
-; GFX11-FAKE16-NEXT: v_dual_mov_b32 v24, s73 :: v_dual_mov_b32 v23, s28
-; GFX11-FAKE16-NEXT: v_dual_mov_b32 v21, s29 :: v_dual_mov_b32 v22, s6
-; GFX11-FAKE16-NEXT: v_dual_mov_b32 v53, s87 :: v_dual_mov_b32 v54, s86
-; GFX11-FAKE16-NEXT: v_dual_mov_b32 v5, s85 :: v_dual_mov_b32 v12, s5
-; GFX11-FAKE16-NEXT: v_dual_mov_b32 v65, s4 :: v_dual_mov_b32 v66, s48
-; GFX11-FAKE16-NEXT: v_dual_mov_b32 v55, s81 :: v_dual_mov_b32 v64, s84
-; GFX11-FAKE16-NEXT: v_dual_mov_b32 v69, s83 :: v_dual_mov_b32 v70, s82
-; GFX11-FAKE16-NEXT: v_dual_mov_b32 v67, s70 :: v_dual_mov_b32 v68, s80
-; GFX11-FAKE16-NEXT: v_dual_mov_b32 v80, s71 :: v_dual_mov_b32 v19, s39
-; GFX11-FAKE16-NEXT: v_dual_mov_b32 v71, s66 :: v_dual_mov_b32 v20, s69
-; GFX11-FAKE16-NEXT: v_dual_mov_b32 v82, s68 :: v_dual_mov_b32 v17, s67
-; GFX11-FAKE16-NEXT: v_dual_mov_b32 v81, s55 :: v_dual_mov_b32 v18, s65
-; GFX11-FAKE16-NEXT: v_dual_mov_b32 v84, s38 :: v_dual_mov_b32 v15, s64
-; GFX11-FAKE16-NEXT: v_dual_mov_b32 v83, s51 :: v_dual_mov_b32 v16, s54
-; GFX11-FAKE16-NEXT: v_dual_mov_b32 v86, s53 :: v_dual_mov_b32 v13, s52
-; GFX11-FAKE16-NEXT: v_dual_mov_b32 v85, s36 :: v_dual_mov_b32 v14, s50
-; GFX11-FAKE16-NEXT: v_dual_mov_b32 v1, s74 :: v_dual_mov_b32 v2, s76
-; GFX11-FAKE16-NEXT: v_dual_mov_b32 v3, s78 :: v_dual_mov_b32 v4, s88
-; GFX11-FAKE16-NEXT: v_dual_mov_b32 v8, s90 :: v_dual_mov_b32 v9, s92
-; GFX11-FAKE16-NEXT: s_mov_b32 s58, s11
-; GFX11-FAKE16-NEXT: v_readlane_b32 s59, v43, 8
-; GFX11-FAKE16-NEXT: v_readlane_b32 s72, v43, 9
-; GFX11-FAKE16-NEXT: v_readlane_b32 s60, v43, 10
-; GFX11-FAKE16-NEXT: v_readlane_b32 s61, v43, 11
-; GFX11-FAKE16-NEXT: v_readlane_b32 s62, v43, 12
-; GFX11-FAKE16-NEXT: v_readlane_b32 s63, v43, 13
-; GFX11-FAKE16-NEXT: v_readlane_b32 s73, v43, 14
-; GFX11-FAKE16-NEXT: v_readlane_b32 s13, v43, 15
-; GFX11-FAKE16-NEXT: v_readlane_b32 s15, v43, 16
-; GFX11-FAKE16-NEXT: v_readlane_b32 s41, v43, 17
-; GFX11-FAKE16-NEXT: v_readlane_b32 s43, v43, 18
-; GFX11-FAKE16-NEXT: v_readlane_b32 s56, v43, 19
-; GFX11-FAKE16-NEXT: v_readlane_b32 s11, v43, 20
-; GFX11-FAKE16-NEXT: v_readlane_b32 s57, v43, 21
-; GFX11-FAKE16-NEXT: v_readlane_b32 s10, v43, 22
-; GFX11-FAKE16-NEXT: v_readlane_b32 s74, v43, 23
-; GFX11-FAKE16-NEXT: v_readlane_b32 s9, v43, 24
-; GFX11-FAKE16-NEXT: v_readlane_b32 s75, v43, 25
-; GFX11-FAKE16-NEXT: v_readlane_b32 s8, v43, 26
-; GFX11-FAKE16-NEXT: v_readlane_b32 s76, v43, 27
-; GFX11-FAKE16-NEXT: v_readlane_b32 s77, v43, 28
-; GFX11-FAKE16-NEXT: v_readlane_b32 s78, v43, 29
-; GFX11-FAKE16-NEXT: v_readlane_b32 s79, v43, 30
-; GFX11-FAKE16-NEXT: v_readlane_b32 s88, v43, 31
-; GFX11-FAKE16-NEXT: v_readlane_b32 s89, v42, 0
-; GFX11-FAKE16-NEXT: v_readlane_b32 s90, v42, 1
-; GFX11-FAKE16-NEXT: v_readlane_b32 s91, v42, 2
-; GFX11-FAKE16-NEXT: v_readlane_b32 s92, v42, 3
-; GFX11-FAKE16-NEXT: v_readlane_b32 s47, v42, 4
-; GFX11-FAKE16-NEXT: v_readlane_b32 s93, v42, 5
-; GFX11-FAKE16-NEXT: v_readlane_b32 vcc_hi, v43, 7
-; GFX11-FAKE16-NEXT: v_readlane_b32 s46, v42, 6
-; GFX11-FAKE16-NEXT: v_readlane_b32 s31, v43, 1
-; GFX11-FAKE16-NEXT: v_readlane_b32 s95, v42, 7
-; GFX11-FAKE16-NEXT: v_readlane_b32 s45, v42, 8
-; GFX11-FAKE16-NEXT: v_readlane_b32 s35, v43, 5
+; GFX11-FAKE16-NEXT: v_dual_mov_b32 v3, s74 :: v_dual_mov_b32 v10, s90
+; GFX11-FAKE16-NEXT: v_dual_mov_b32 v15, s94 :: v_dual_mov_b32 v16, s30
+; GFX11-FAKE16-NEXT: v_readlane_b32 s90, v42, 4
+; GFX11-FAKE16-NEXT: v_readlane_b32 s94, v42, 2
+; GFX11-FAKE16-NEXT: v_readlane_b32 s91, v42, 5
+; GFX11-FAKE16-NEXT: v_readlane_b32 s95, v42, 3
+; GFX11-FAKE16-NEXT: v_readlane_b32 s30, v42, 0
+; GFX11-FAKE16-NEXT: v_dual_mov_b32 v53, s4 :: v_dual_mov_b32 v52, s5
+; GFX11-FAKE16-NEXT: v_dual_mov_b32 v51, s40 :: v_dual_mov_b32 v50, s6
+; GFX11-FAKE16-NEXT: v_dual_mov_b32 v48, s7 :: v_dual_mov_b32 v49, s98
+; GFX11-FAKE16-NEXT: v_dual_mov_b32 v39, s8 :: v_dual_mov_b32 v38, s97
+; GFX11-FAKE16-NEXT: v_dual_mov_b32 v37, s9 :: v_dual_mov_b32 v36, s58
+; GFX11-FAKE16-NEXT: v_dual_mov_b32 v35, s59 :: v_dual_mov_b32 v34, s15
+; GFX11-FAKE16-NEXT: v_dual_mov_b32 v33, s60 :: v_dual_mov_b32 v32, s61
+; GFX11-FAKE16-NEXT: v_dual_mov_b32 v31, s14 :: v_dual_mov_b32 v30, s62
+; GFX11-FAKE16-NEXT: v_dual_mov_b32 v28, s63 :: v_dual_mov_b32 v29, s96
+; GFX11-FAKE16-NEXT: v_dual_mov_b32 v27, s72 :: v_dual_mov_b32 v26, s13
+; GFX11-FAKE16-NEXT: v_dual_mov_b32 v25, s73 :: v_dual_mov_b32 v24, s28
+; GFX11-FAKE16-NEXT: v_dual_mov_b32 v22, s29 :: v_dual_mov_b32 v23, s12
+; GFX11-FAKE16-NEXT: v_dual_mov_b32 v17, s87 :: v_dual_mov_b32 v54, s86
+; GFX11-FAKE16-NEXT: v_dual_mov_b32 v7, s84 :: v_dual_mov_b32 v66, s85
+; GFX11-FAKE16-NEXT: v_dual_mov_b32 v11, s11 :: v_dual_mov_b32 v64, s83
+; GFX11-FAKE16-NEXT: v_dual_mov_b32 v65, s10 :: v_dual_mov_b32 v70, s81
+; GFX11-FAKE16-NEXT: v_dual_mov_b32 v55, s48 :: v_dual_mov_b32 v68, s80
+; GFX11-FAKE16-NEXT: v_dual_mov_b32 v69, s82 :: v_dual_mov_b32 v80, s71
+; GFX11-FAKE16-NEXT: v_dual_mov_b32 v67, s69 :: v_dual_mov_b32 v20, s70
+; GFX11-FAKE16-NEXT: v_dual_mov_b32 v71, s65 :: v_dual_mov_b32 v82, s67
+; GFX11-FAKE16-NEXT: v_dual_mov_b32 v21, s68 :: v_dual_mov_b32 v18, s66
+; GFX11-FAKE16-NEXT: v_dual_mov_b32 v81, s54 :: v_dual_mov_b32 v84, s64
+; GFX11-FAKE16-NEXT: v_dual_mov_b32 v19, s39 :: v_dual_mov_b32 v12, s55
+; GFX11-FAKE16-NEXT: v_dual_mov_b32 v83, s38 :: v_dual_mov_b32 v86, s52
+; GFX11-FAKE16-NEXT: v_dual_mov_b32 v13, s53 :: v_dual_mov_b32 v8, s51
+; GFX11-FAKE16-NEXT: v_dual_mov_b32 v85, s36 :: v_dual_mov_b32 v96, s49
+; GFX11-FAKE16-NEXT: v_dual_mov_b32 v9, s50 :: v_dual_mov_b32 v2, s35
+; GFX11-FAKE16-NEXT: v_dual_mov_b32 v1, s37 :: v_dual_mov_b32 v4, s76
+; GFX11-FAKE16-NEXT: v_dual_mov_b32 v87, s34 :: v_dual_mov_b32 v6, s88
+; GFX11-FAKE16-NEXT: v_dual_mov_b32 v5, s78 :: v_dual_mov_b32 v14, s92
+; GFX11-FAKE16-NEXT: s_mov_b32 s61, s41
+; GFX11-FAKE16-NEXT: v_readlane_b32 s62, v42, 6
+; GFX11-FAKE16-NEXT: v_readlane_b32 s58, v42, 7
+; GFX11-FAKE16-NEXT: v_readlane_b32 s59, v42, 8
+; GFX11-FAKE16-NEXT: v_readlane_b32 s60, v42, 9
+; GFX11-FAKE16-NEXT: v_readlane_b32 s63, v42, 10
+; GFX11-FAKE16-NEXT: v_readlane_b32 s72, v42, 11
+; GFX11-FAKE16-NEXT: v_readlane_b32 s73, v42, 12
+; GFX11-FAKE16-NEXT: v_readlane_b32 s43, v42, 13
+; GFX11-FAKE16-NEXT: v_readlane_b32 s41, v42, 14
+; GFX11-FAKE16-NEXT: v_readlane_b32 s45, v42, 15
+; GFX11-FAKE16-NEXT: v_readlane_b32 s40, v42, 16
+; GFX11-FAKE16-NEXT: v_readlane_b32 s47, v42, 17
+; GFX11-FAKE16-NEXT: v_readlane_b32 s29, v42, 18
+; GFX11-FAKE16-NEXT: v_readlane_b32 s57, v42, 19
+; GFX11-FAKE16-NEXT: v_readlane_b32 s28, v42, 20
+; GFX11-FAKE16-NEXT: v_readlane_b32 s74, v42, 21
+; GFX11-FAKE16-NEXT: v_readlane_b32 s15, v42, 22
+; GFX11-FAKE16-NEXT: v_readlane_b32 s75, v42, 23
+; GFX11-FAKE16-NEXT: v_readlane_b32 s14, v42, 24
+; GFX11-FAKE16-NEXT: v_readlane_b32 s76, v42, 25
+; GFX11-FAKE16-NEXT: v_readlane_b32 s11, v42, 26
+; GFX11-FAKE16-NEXT: v_readlane_b32 s77, v42, 27
+; GFX11-FAKE16-NEXT: v_readlane_b32 s10, v42, 28
+; GFX11-FAKE16-NEXT: v_readlane_b32 s78, v42, 29
+; GFX11-FAKE16-NEXT: v_readlane_b32 s13, v42, 30
+; GFX11-FAKE16-NEXT: v_readlane_b32 s79, v42, 31
+; GFX11-FAKE16-NEXT: v_readlane_b32 s12, v43, 0
+; GFX11-FAKE16-NEXT: v_readlane_b32 s88, v43, 1
+; GFX11-FAKE16-NEXT: v_readlane_b32 s9, v43, 2
+; GFX11-FAKE16-NEXT: v_readlane_b32 s89, v43, 3
+; GFX11-FAKE16-NEXT: s_mov_b32 s92, s100
+; GFX11-FAKE16-NEXT: v_readlane_b32 s8, v43, 4
+; GFX11-FAKE16-NEXT: v_readlane_b32 s91, v43, 5
+; GFX11-FAKE16-NEXT: v_readlane_b32 s93, v43, 6
+; GFX11-FAKE16-NEXT: v_readlane_b32 s95, v43, 7
+; GFX11-FAKE16-NEXT: s_mov_b32 vcc_lo, s102
+; GFX11-FAKE16-NEXT: v_readlane_b32 vcc_hi, v43, 8
+; GFX11-FAKE16-NEXT: v_readlane_b32 s31, v42, 1
+; GFX11-FAKE16-NEXT: v_readlane_b32 s7, v43, 9
; GFX11-FAKE16-NEXT: .LBB91_5: ; %end
; GFX11-FAKE16-NEXT: s_and_b32 s0, s0, 0xff
; GFX11-FAKE16-NEXT: s_lshl_b32 s4, s104, 8
-; GFX11-FAKE16-NEXT: s_and_b32 s5, s103, 0xff
-; GFX11-FAKE16-NEXT: s_lshl_b32 s6, s42, 8
+; GFX11-FAKE16-NEXT: s_and_b32 s5, s44, 0xff
+; GFX11-FAKE16-NEXT: s_lshl_b32 s6, s56, 8
; GFX11-FAKE16-NEXT: s_or_b32 s0, s0, s4
; GFX11-FAKE16-NEXT: s_or_b32 s4, s5, s6
; GFX11-FAKE16-NEXT: s_and_b32 s1, s1, 0xff
-; GFX11-FAKE16-NEXT: s_lshl_b32 s5, s102, 8
-; GFX11-FAKE16-NEXT: s_and_b32 s6, s58, 0xff
-; GFX11-FAKE16-NEXT: s_lshl_b32 s7, s101, 8
+; GFX11-FAKE16-NEXT: s_lshl_b32 s5, s99, 8
+; GFX11-FAKE16-NEXT: s_and_b32 s6, s61, 0xff
+; GFX11-FAKE16-NEXT: s_lshl_b32 s7, s7, 8
; GFX11-FAKE16-NEXT: s_or_b32 s1, s1, s5
; GFX11-FAKE16-NEXT: s_or_b32 s5, s6, s7
; GFX11-FAKE16-NEXT: s_and_b32 s0, s0, 0xffff
@@ -178315,15 +178420,15 @@ define inreg <128 x i8> @bitcast_v64bf16_to_v128i8_scalar(<64 x bfloat> inreg %a
; GFX11-FAKE16-NEXT: s_or_b32 s0, s0, s4
; GFX11-FAKE16-NEXT: s_or_b32 s1, s1, s5
; GFX11-FAKE16-NEXT: s_and_b32 s2, s2, 0xff
-; GFX11-FAKE16-NEXT: s_lshl_b32 s4, s100, 8
-; GFX11-FAKE16-NEXT: s_and_b32 s5, s99, 0xff
-; GFX11-FAKE16-NEXT: s_lshl_b32 s6, s40, 8
+; GFX11-FAKE16-NEXT: s_lshl_b32 s4, vcc_hi, 8
+; GFX11-FAKE16-NEXT: s_and_b32 s5, s93, 0xff
+; GFX11-FAKE16-NEXT: s_lshl_b32 s6, s46, 8
; GFX11-FAKE16-NEXT: s_or_b32 s2, s2, s4
; GFX11-FAKE16-NEXT: s_or_b32 s4, s5, s6
; GFX11-FAKE16-NEXT: s_and_b32 s3, s3, 0xff
-; GFX11-FAKE16-NEXT: s_lshl_b32 s5, s45, 8
-; GFX11-FAKE16-NEXT: s_and_b32 s6, s59, 0xff
-; GFX11-FAKE16-NEXT: s_lshl_b32 s7, s95, 8
+; GFX11-FAKE16-NEXT: s_lshl_b32 s5, s95, 8
+; GFX11-FAKE16-NEXT: s_and_b32 s6, s62, 0xff
+; GFX11-FAKE16-NEXT: s_lshl_b32 s7, s91, 8
; GFX11-FAKE16-NEXT: s_or_b32 s3, s3, s5
; GFX11-FAKE16-NEXT: s_or_b32 s5, s6, s7
; GFX11-FAKE16-NEXT: s_and_b32 s2, s2, 0xffff
@@ -178335,15 +178440,15 @@ define inreg <128 x i8> @bitcast_v64bf16_to_v128i8_scalar(<64 x bfloat> inreg %a
; GFX11-FAKE16-NEXT: v_dual_mov_b32 v97, s0 :: v_dual_mov_b32 v98, s1
; GFX11-FAKE16-NEXT: v_dual_mov_b32 v99, s2 :: v_dual_mov_b32 v100, s3
; GFX11-FAKE16-NEXT: s_and_b32 s0, s16, 0xff
-; GFX11-FAKE16-NEXT: s_lshl_b32 s1, s46, 8
-; GFX11-FAKE16-NEXT: s_and_b32 s2, s93, 0xff
-; GFX11-FAKE16-NEXT: s_lshl_b32 s3, s14, 8
+; GFX11-FAKE16-NEXT: s_lshl_b32 s1, s8, 8
+; GFX11-FAKE16-NEXT: s_and_b32 s2, s89, 0xff
+; GFX11-FAKE16-NEXT: s_lshl_b32 s3, s30, 8
; GFX11-FAKE16-NEXT: s_or_b32 s0, s0, s1
; GFX11-FAKE16-NEXT: s_or_b32 s1, s2, s3
; GFX11-FAKE16-NEXT: s_and_b32 s2, s17, 0xff
-; GFX11-FAKE16-NEXT: s_lshl_b32 s3, s47, 8
-; GFX11-FAKE16-NEXT: s_and_b32 s4, s72, 0xff
-; GFX11-FAKE16-NEXT: s_lshl_b32 s5, s92, 8
+; GFX11-FAKE16-NEXT: s_lshl_b32 s3, s9, 8
+; GFX11-FAKE16-NEXT: s_and_b32 s4, s58, 0xff
+; GFX11-FAKE16-NEXT: s_lshl_b32 s5, s88, 8
; GFX11-FAKE16-NEXT: s_or_b32 s2, s2, s3
; GFX11-FAKE16-NEXT: s_or_b32 s3, s4, s5
; GFX11-FAKE16-NEXT: s_and_b32 s0, s0, 0xffff
@@ -178353,15 +178458,15 @@ define inreg <128 x i8> @bitcast_v64bf16_to_v128i8_scalar(<64 x bfloat> inreg %a
; GFX11-FAKE16-NEXT: s_or_b32 s0, s0, s1
; GFX11-FAKE16-NEXT: s_or_b32 s1, s2, s3
; GFX11-FAKE16-NEXT: s_and_b32 s2, s18, 0xff
-; GFX11-FAKE16-NEXT: s_lshl_b32 s3, s91, 8
-; GFX11-FAKE16-NEXT: s_and_b32 s4, s90, 0xff
-; GFX11-FAKE16-NEXT: s_lshl_b32 s5, s12, 8
+; GFX11-FAKE16-NEXT: s_lshl_b32 s3, s12, 8
+; GFX11-FAKE16-NEXT: s_and_b32 s4, s79, 0xff
+; GFX11-FAKE16-NEXT: s_lshl_b32 s5, s42, 8
; GFX11-FAKE16-NEXT: s_or_b32 s2, s2, s3
; GFX11-FAKE16-NEXT: s_or_b32 s3, s4, s5
; GFX11-FAKE16-NEXT: s_and_b32 s4, s19, 0xff
-; GFX11-FAKE16-NEXT: s_lshl_b32 s5, s89, 8
-; GFX11-FAKE16-NEXT: s_and_b32 s6, s60, 0xff
-; GFX11-FAKE16-NEXT: s_lshl_b32 s7, s88, 8
+; GFX11-FAKE16-NEXT: s_lshl_b32 s5, s13, 8
+; GFX11-FAKE16-NEXT: s_and_b32 s6, s59, 0xff
+; GFX11-FAKE16-NEXT: s_lshl_b32 s7, s78, 8
; GFX11-FAKE16-NEXT: s_or_b32 s4, s4, s5
; GFX11-FAKE16-NEXT: s_or_b32 s5, s6, s7
; GFX11-FAKE16-NEXT: s_and_b32 s2, s2, 0xffff
@@ -178373,14 +178478,14 @@ define inreg <128 x i8> @bitcast_v64bf16_to_v128i8_scalar(<64 x bfloat> inreg %a
; GFX11-FAKE16-NEXT: v_dual_mov_b32 v112, s0 :: v_dual_mov_b32 v113, s1
; GFX11-FAKE16-NEXT: v_dual_mov_b32 v114, s2 :: v_dual_mov_b32 v115, s3
; GFX11-FAKE16-NEXT: s_and_b32 s0, s20, 0xff
-; GFX11-FAKE16-NEXT: s_lshl_b32 s1, s79, 8
-; GFX11-FAKE16-NEXT: s_and_b32 s2, s78, 0xff
-; GFX11-FAKE16-NEXT: s_lshl_b32 s3, s30, 8
+; GFX11-FAKE16-NEXT: s_lshl_b32 s1, s10, 8
+; GFX11-FAKE16-NEXT: s_and_b32 s2, s77, 0xff
+; GFX11-FAKE16-NEXT: s_lshl_b32 s3, s94, 8
; GFX11-FAKE16-NEXT: s_or_b32 s0, s0, s1
; GFX11-FAKE16-NEXT: s_or_b32 s1, s2, s3
; GFX11-FAKE16-NEXT: s_and_b32 s2, s21, 0xff
-; GFX11-FAKE16-NEXT: s_lshl_b32 s3, s77, 8
-; GFX11-FAKE16-NEXT: s_and_b32 s4, s61, 0xff
+; GFX11-FAKE16-NEXT: s_lshl_b32 s3, s11, 8
+; GFX11-FAKE16-NEXT: s_and_b32 s4, s60, 0xff
; GFX11-FAKE16-NEXT: s_lshl_b32 s5, s76, 8
; GFX11-FAKE16-NEXT: s_or_b32 s2, s2, s3
; GFX11-FAKE16-NEXT: s_or_b32 s3, s4, s5
@@ -178391,14 +178496,14 @@ define inreg <128 x i8> @bitcast_v64bf16_to_v128i8_scalar(<64 x bfloat> inreg %a
; GFX11-FAKE16-NEXT: s_or_b32 s0, s0, s1
; GFX11-FAKE16-NEXT: s_or_b32 s1, s2, s3
; GFX11-FAKE16-NEXT: s_and_b32 s2, s22, 0xff
-; GFX11-FAKE16-NEXT: s_lshl_b32 s3, s8, 8
+; GFX11-FAKE16-NEXT: s_lshl_b32 s3, s14, 8
; GFX11-FAKE16-NEXT: s_and_b32 s4, s75, 0xff
-; GFX11-FAKE16-NEXT: s_lshl_b32 s5, s94, 8
+; GFX11-FAKE16-NEXT: s_lshl_b32 s5, s90, 8
; GFX11-FAKE16-NEXT: s_or_b32 s2, s2, s3
; GFX11-FAKE16-NEXT: s_or_b32 s3, s4, s5
; GFX11-FAKE16-NEXT: s_and_b32 s4, s23, 0xff
-; GFX11-FAKE16-NEXT: s_lshl_b32 s5, s9, 8
-; GFX11-FAKE16-NEXT: s_and_b32 s6, s62, 0xff
+; GFX11-FAKE16-NEXT: s_lshl_b32 s5, s15, 8
+; GFX11-FAKE16-NEXT: s_and_b32 s6, s63, 0xff
; GFX11-FAKE16-NEXT: s_lshl_b32 s7, s74, 8
; GFX11-FAKE16-NEXT: s_and_b32 s2, s2, 0xffff
; GFX11-FAKE16-NEXT: s_lshl_b32 s3, s3, 16
@@ -178414,32 +178519,32 @@ define inreg <128 x i8> @bitcast_v64bf16_to_v128i8_scalar(<64 x bfloat> inreg %a
; GFX11-FAKE16-NEXT: v_dual_mov_b32 v97, s0 :: v_dual_mov_b32 v98, s1
; GFX11-FAKE16-NEXT: v_dual_mov_b32 v99, s2 :: v_dual_mov_b32 v100, s3
; GFX11-FAKE16-NEXT: s_and_b32 s0, s24, 0xff
-; GFX11-FAKE16-NEXT: s_lshl_b32 s1, s10, 8
+; GFX11-FAKE16-NEXT: s_lshl_b32 s1, s28, 8
; GFX11-FAKE16-NEXT: s_and_b32 s2, s57, 0xff
-; GFX11-FAKE16-NEXT: s_lshl_b32 s4, s34, 8
+; GFX11-FAKE16-NEXT: s_lshl_b32 s4, vcc_lo, 8
; GFX11-FAKE16-NEXT: s_or_b32 s0, s0, s1
; GFX11-FAKE16-NEXT: s_or_b32 s1, s2, s4
; GFX11-FAKE16-NEXT: s_and_b32 s0, s0, 0xffff
; GFX11-FAKE16-NEXT: s_lshl_b32 s1, s1, 16
-; GFX11-FAKE16-NEXT: s_lshl_b32 s2, s11, 8
+; GFX11-FAKE16-NEXT: s_lshl_b32 s2, s29, 8
; GFX11-FAKE16-NEXT: s_or_b32 s0, s0, s1
; GFX11-FAKE16-NEXT: s_and_b32 s1, s25, 0xff
-; GFX11-FAKE16-NEXT: s_and_b32 s3, s63, 0xff
-; GFX11-FAKE16-NEXT: s_lshl_b32 s4, s56, 8
+; GFX11-FAKE16-NEXT: s_and_b32 s3, s72, 0xff
+; GFX11-FAKE16-NEXT: s_lshl_b32 s4, s47, 8
; GFX11-FAKE16-NEXT: s_or_b32 s1, s1, s2
; GFX11-FAKE16-NEXT: s_or_b32 s2, s3, s4
; GFX11-FAKE16-NEXT: s_and_b32 s1, s1, 0xffff
; GFX11-FAKE16-NEXT: s_lshl_b32 s2, s2, 16
; GFX11-FAKE16-NEXT: s_and_b32 s3, s26, 0xff
-; GFX11-FAKE16-NEXT: s_lshl_b32 s4, s43, 8
+; GFX11-FAKE16-NEXT: s_lshl_b32 s4, s40, 8
; GFX11-FAKE16-NEXT: s_or_b32 s1, s1, s2
; GFX11-FAKE16-NEXT: s_or_b32 s2, s3, s4
-; GFX11-FAKE16-NEXT: s_and_b32 s3, s41, 0xff
-; GFX11-FAKE16-NEXT: s_lshl_b32 s4, vcc_lo, 8
-; GFX11-FAKE16-NEXT: s_lshl_b32 s5, s15, 8
+; GFX11-FAKE16-NEXT: s_and_b32 s3, s45, 0xff
+; GFX11-FAKE16-NEXT: s_lshl_b32 s4, s92, 8
+; GFX11-FAKE16-NEXT: s_lshl_b32 s5, s41, 8
; GFX11-FAKE16-NEXT: s_or_b32 s3, s3, s4
; GFX11-FAKE16-NEXT: s_and_b32 s4, s27, 0xff
-; GFX11-FAKE16-NEXT: s_lshl_b32 s6, s13, 8
+; GFX11-FAKE16-NEXT: s_lshl_b32 s6, s43, 8
; GFX11-FAKE16-NEXT: s_or_b32 s4, s4, s5
; GFX11-FAKE16-NEXT: s_and_b32 s5, s73, 0xff
; GFX11-FAKE16-NEXT: s_and_b32 s2, s2, 0xffff
@@ -178447,160 +178552,160 @@ define inreg <128 x i8> @bitcast_v64bf16_to_v128i8_scalar(<64 x bfloat> inreg %a
; GFX11-FAKE16-NEXT: s_lshl_b32 s3, s3, 16
; GFX11-FAKE16-NEXT: s_and_b32 s4, s4, 0xffff
; GFX11-FAKE16-NEXT: s_lshl_b32 s5, s5, 16
-; GFX11-FAKE16-NEXT: v_dual_mov_b32 v112, s0 :: v_dual_and_b32 v23, 0xff, v23
-; GFX11-FAKE16-NEXT: v_dual_mov_b32 v113, s1 :: v_dual_lshlrev_b32 v6, 8, v6
+; GFX11-FAKE16-NEXT: v_dual_mov_b32 v113, s1 :: v_dual_and_b32 v24, 0xff, v24
+; GFX11-FAKE16-NEXT: v_dual_mov_b32 v112, s0 :: v_dual_lshlrev_b32 v1, 8, v1
+; GFX11-FAKE16-NEXT: v_lshlrev_b32_e32 v16, 8, v16
; GFX11-FAKE16-NEXT: s_or_b32 s2, s2, s3
-; GFX11-FAKE16-NEXT: s_delay_alu instid0(SALU_CYCLE_1) | instskip(SKIP_1) | instid1(SALU_CYCLE_1)
-; GFX11-FAKE16-NEXT: v_dual_mov_b32 v114, s2 :: v_dual_lshlrev_b32 v11, 8, v11
; GFX11-FAKE16-NEXT: s_or_b32 s3, s4, s5
+; GFX11-FAKE16-NEXT: s_delay_alu instid0(SALU_CYCLE_1) | instskip(SKIP_3) | instid1(VALU_DEP_4)
; GFX11-FAKE16-NEXT: v_dual_mov_b32 v115, s3 :: v_dual_and_b32 v96, 0xff, v96
-; GFX11-FAKE16-NEXT: v_or_b32_e32 v6, v23, v6
-; GFX11-FAKE16-NEXT: v_lshlrev_b32_e32 v7, 8, v7
-; GFX11-FAKE16-NEXT: v_lshlrev_b32_e32 v13, 8, v13
-; GFX11-FAKE16-NEXT: s_delay_alu instid0(VALU_DEP_4)
-; GFX11-FAKE16-NEXT: v_or_b32_e32 v11, v96, v11
-; GFX11-FAKE16-NEXT: v_lshlrev_b32_e32 v10, 8, v10
-; GFX11-FAKE16-NEXT: v_and_b32_e32 v6, 0xffff, v6
-; GFX11-FAKE16-NEXT: v_and_b32_e32 v24, 0xff, v24
-; GFX11-FAKE16-NEXT: v_lshlrev_b32_e32 v14, 8, v14
-; GFX11-FAKE16-NEXT: v_lshlrev_b32_e32 v11, 16, v11
-; GFX11-FAKE16-NEXT: v_lshlrev_b32_e32 v15, 8, v15
-; GFX11-FAKE16-NEXT: v_lshlrev_b32_e32 v9, 8, v9
-; GFX11-FAKE16-NEXT: v_lshlrev_b32_e32 v16, 8, v16
+; GFX11-FAKE16-NEXT: v_or_b32_e32 v1, v24, v1
+; GFX11-FAKE16-NEXT: v_mov_b32_e32 v114, s2
+; GFX11-FAKE16-NEXT: v_lshlrev_b32_e32 v2, 8, v2
+; GFX11-FAKE16-NEXT: v_or_b32_e32 v16, v96, v16
; GFX11-FAKE16-NEXT: v_lshlrev_b32_e32 v8, 8, v8
-; GFX11-FAKE16-NEXT: v_or_b32_e32 v23, v6, v11
-; GFX11-FAKE16-NEXT: v_and_b32_e32 v6, 0xff, v21
-; GFX11-FAKE16-NEXT: v_and_b32_e32 v11, 0xff, v22
-; GFX11-FAKE16-NEXT: v_lshlrev_b32_e32 v21, 8, v87
-; GFX11-FAKE16-NEXT: v_and_b32_e32 v22, 0xff, v26
-; GFX11-FAKE16-NEXT: v_and_b32_e32 v26, 0xff, v86
-; GFX11-FAKE16-NEXT: v_or_b32_e32 v6, v6, v7
-; GFX11-FAKE16-NEXT: v_lshlrev_b32_e32 v4, 8, v4
-; GFX11-FAKE16-NEXT: v_or_b32_e32 v7, v11, v21
-; GFX11-FAKE16-NEXT: v_or_b32_e32 v11, v22, v13
-; GFX11-FAKE16-NEXT: v_or_b32_e32 v10, v26, v10
-; GFX11-FAKE16-NEXT: v_or_b32_e32 v13, v24, v14
-; GFX11-FAKE16-NEXT: v_and_b32_e32 v14, 0xff, v25
-; GFX11-FAKE16-NEXT: v_lshlrev_b32_e32 v21, 8, v85
-; GFX11-FAKE16-NEXT: v_and_b32_e32 v22, 0xff, v29
-; GFX11-FAKE16-NEXT: v_and_b32_e32 v24, 0xff, v84
-; GFX11-FAKE16-NEXT: v_and_b32_e32 v25, 0xff, v27
+; GFX11-FAKE16-NEXT: v_and_b32_e32 v1, 0xffff, v1
+; GFX11-FAKE16-NEXT: v_and_b32_e32 v25, 0xff, v25
+; GFX11-FAKE16-NEXT: v_lshlrev_b32_e32 v9, 8, v9
+; GFX11-FAKE16-NEXT: v_lshlrev_b32_e32 v16, 16, v16
+; GFX11-FAKE16-NEXT: v_lshlrev_b32_e32 v15, 8, v15
+; GFX11-FAKE16-NEXT: v_lshlrev_b32_e32 v12, 8, v12
+; GFX11-FAKE16-NEXT: v_lshlrev_b32_e32 v14, 8, v14
+; GFX11-FAKE16-NEXT: v_or_b32_e32 v9, v25, v9
+; GFX11-FAKE16-NEXT: v_or_b32_e32 v24, v1, v16
+; GFX11-FAKE16-NEXT: v_and_b32_e32 v1, 0xff, v22
+; GFX11-FAKE16-NEXT: v_and_b32_e32 v16, 0xff, v23
+; GFX11-FAKE16-NEXT: v_lshlrev_b32_e32 v22, 8, v87
+; GFX11-FAKE16-NEXT: v_and_b32_e32 v23, 0xff, v27
+; GFX11-FAKE16-NEXT: v_and_b32_e32 v27, 0xff, v86
+; GFX11-FAKE16-NEXT: v_or_b32_e32 v1, v1, v2
+; GFX11-FAKE16-NEXT: v_and_b32_e32 v25, 0xff, v84
+; GFX11-FAKE16-NEXT: v_or_b32_e32 v2, v16, v22
+; GFX11-FAKE16-NEXT: v_or_b32_e32 v8, v23, v8
+; GFX11-FAKE16-NEXT: v_and_b32_e32 v16, 0xff, v26
+; GFX11-FAKE16-NEXT: v_lshlrev_b32_e32 v22, 8, v85
+; GFX11-FAKE16-NEXT: v_and_b32_e32 v23, 0xff, v30
+; GFX11-FAKE16-NEXT: v_or_b32_e32 v15, v27, v15
; GFX11-FAKE16-NEXT: v_and_b32_e32 v26, 0xff, v28
-; GFX11-FAKE16-NEXT: v_lshlrev_b32_e32 v27, 8, v83
-; GFX11-FAKE16-NEXT: v_or_b32_e32 v14, v14, v21
-; GFX11-FAKE16-NEXT: v_or_b32_e32 v15, v22, v15
-; GFX11-FAKE16-NEXT: v_or_b32_e32 v9, v24, v9
-; GFX11-FAKE16-NEXT: v_or_b32_e32 v16, v25, v16
-; GFX11-FAKE16-NEXT: v_or_b32_e32 v21, v26, v27
-; GFX11-FAKE16-NEXT: v_and_b32_e32 v6, 0xffff, v6
-; GFX11-FAKE16-NEXT: v_lshlrev_b32_e32 v7, 16, v7
-; GFX11-FAKE16-NEXT: v_and_b32_e32 v11, 0xffff, v11
-; GFX11-FAKE16-NEXT: v_lshlrev_b32_e32 v10, 16, v10
-; GFX11-FAKE16-NEXT: v_and_b32_e32 v13, 0xffff, v13
+; GFX11-FAKE16-NEXT: v_lshlrev_b32_e32 v13, 8, v13
+; GFX11-FAKE16-NEXT: v_and_b32_e32 v27, 0xff, v29
+; GFX11-FAKE16-NEXT: v_lshlrev_b32_e32 v28, 8, v83
+; GFX11-FAKE16-NEXT: v_or_b32_e32 v16, v16, v22
+; GFX11-FAKE16-NEXT: v_or_b32_e32 v12, v23, v12
+; GFX11-FAKE16-NEXT: v_or_b32_e32 v14, v25, v14
+; GFX11-FAKE16-NEXT: v_and_b32_e32 v1, 0xffff, v1
+; GFX11-FAKE16-NEXT: v_lshlrev_b32_e32 v2, 16, v2
+; GFX11-FAKE16-NEXT: v_and_b32_e32 v8, 0xffff, v8
+; GFX11-FAKE16-NEXT: v_lshlrev_b32_e32 v15, 16, v15
+; GFX11-FAKE16-NEXT: v_and_b32_e32 v9, 0xffff, v9
+; GFX11-FAKE16-NEXT: v_or_b32_e32 v13, v26, v13
+; GFX11-FAKE16-NEXT: v_or_b32_e32 v22, v27, v28
+; GFX11-FAKE16-NEXT: v_lshlrev_b32_e32 v16, 16, v16
+; GFX11-FAKE16-NEXT: v_and_b32_e32 v12, 0xffff, v12
; GFX11-FAKE16-NEXT: v_lshlrev_b32_e32 v14, 16, v14
-; GFX11-FAKE16-NEXT: v_and_b32_e32 v15, 0xffff, v15
-; GFX11-FAKE16-NEXT: v_lshlrev_b32_e32 v9, 16, v9
-; GFX11-FAKE16-NEXT: v_and_b32_e32 v16, 0xffff, v16
-; GFX11-FAKE16-NEXT: v_lshlrev_b32_e32 v21, 16, v21
-; GFX11-FAKE16-NEXT: v_or_b32_e32 v24, v6, v7
-; GFX11-FAKE16-NEXT: v_or_b32_e32 v25, v11, v10
-; GFX11-FAKE16-NEXT: v_or_b32_e32 v26, v13, v14
-; GFX11-FAKE16-NEXT: v_or_b32_e32 v6, v15, v9
-; GFX11-FAKE16-NEXT: v_or_b32_e32 v7, v16, v21
-; GFX11-FAKE16-NEXT: v_and_b32_e32 v9, 0xff, v32
-; GFX11-FAKE16-NEXT: v_lshlrev_b32_e32 v10, 8, v17
-; GFX11-FAKE16-NEXT: v_and_b32_e32 v11, 0xff, v82
-; GFX11-FAKE16-NEXT: v_and_b32_e32 v13, 0xff, v31
-; GFX11-FAKE16-NEXT: v_lshlrev_b32_e32 v14, 8, v18
-; GFX11-FAKE16-NEXT: v_and_b32_e32 v15, 0xff, v30
+; GFX11-FAKE16-NEXT: v_and_b32_e32 v13, 0xffff, v13
+; GFX11-FAKE16-NEXT: v_lshlrev_b32_e32 v22, 16, v22
+; GFX11-FAKE16-NEXT: v_or_b32_e32 v25, v1, v2
+; GFX11-FAKE16-NEXT: v_or_b32_e32 v26, v8, v15
+; GFX11-FAKE16-NEXT: v_or_b32_e32 v27, v9, v16
+; GFX11-FAKE16-NEXT: v_or_b32_e32 v12, v12, v14
+; GFX11-FAKE16-NEXT: v_and_b32_e32 v1, 0xff, v33
+; GFX11-FAKE16-NEXT: v_lshlrev_b32_e32 v2, 8, v18
+; GFX11-FAKE16-NEXT: v_and_b32_e32 v8, 0xff, v82
+; GFX11-FAKE16-NEXT: v_lshlrev_b32_e32 v9, 8, v10
+; GFX11-FAKE16-NEXT: v_and_b32_e32 v10, 0xff, v32
+; GFX11-FAKE16-NEXT: v_lshlrev_b32_e32 v14, 8, v19
+; GFX11-FAKE16-NEXT: v_and_b32_e32 v15, 0xff, v31
; GFX11-FAKE16-NEXT: v_lshlrev_b32_e32 v16, 8, v81
-; GFX11-FAKE16-NEXT: v_and_b32_e32 v17, 0xff, v35
-; GFX11-FAKE16-NEXT: v_lshlrev_b32_e32 v18, 8, v19
-; GFX11-FAKE16-NEXT: v_or_b32_e32 v9, v9, v10
-; GFX11-FAKE16-NEXT: v_or_b32_e32 v8, v11, v8
-; GFX11-FAKE16-NEXT: v_or_b32_e32 v10, v13, v14
-; GFX11-FAKE16-NEXT: v_or_b32_e32 v11, v15, v16
-; GFX11-FAKE16-NEXT: v_or_b32_e32 v13, v17, v18
+; GFX11-FAKE16-NEXT: v_and_b32_e32 v18, 0xff, v36
+; GFX11-FAKE16-NEXT: v_lshlrev_b32_e32 v19, 8, v20
+; GFX11-FAKE16-NEXT: v_or_b32_e32 v13, v13, v22
+; GFX11-FAKE16-NEXT: v_or_b32_e32 v1, v1, v2
+; GFX11-FAKE16-NEXT: v_or_b32_e32 v2, v8, v9
+; GFX11-FAKE16-NEXT: v_or_b32_e32 v8, v10, v14
+; GFX11-FAKE16-NEXT: v_or_b32_e32 v9, v15, v16
+; GFX11-FAKE16-NEXT: v_or_b32_e32 v10, v18, v19
; GFX11-FAKE16-NEXT: v_and_b32_e32 v14, 0xff, v80
-; GFX11-FAKE16-NEXT: v_and_b32_e32 v15, 0xff, v34
-; GFX11-FAKE16-NEXT: v_lshlrev_b32_e32 v16, 8, v20
-; GFX11-FAKE16-NEXT: v_and_b32_e32 v17, 0xff, v33
-; GFX11-FAKE16-NEXT: v_lshlrev_b32_e32 v18, 8, v71
-; GFX11-FAKE16-NEXT: v_and_b32_e32 v19, 0xff, v38
-; GFX11-FAKE16-NEXT: v_lshlrev_b32_e32 v20, 8, v70
-; GFX11-FAKE16-NEXT: v_and_b32_e32 v21, 0xff, v69
-; GFX11-FAKE16-NEXT: v_lshlrev_b32_e32 v3, 8, v3
-; GFX11-FAKE16-NEXT: v_or_b32_e32 v4, v14, v4
+; GFX11-FAKE16-NEXT: v_lshlrev_b32_e32 v6, 8, v6
+; GFX11-FAKE16-NEXT: v_and_b32_e32 v15, 0xff, v35
+; GFX11-FAKE16-NEXT: v_lshlrev_b32_e32 v16, 8, v21
+; GFX11-FAKE16-NEXT: v_and_b32_e32 v18, 0xff, v34
+; GFX11-FAKE16-NEXT: v_lshlrev_b32_e32 v19, 8, v71
+; GFX11-FAKE16-NEXT: v_and_b32_e32 v20, 0xff, v39
+; GFX11-FAKE16-NEXT: v_lshlrev_b32_e32 v21, 8, v70
+; GFX11-FAKE16-NEXT: v_and_b32_e32 v22, 0xff, v69
+; GFX11-FAKE16-NEXT: v_lshlrev_b32_e32 v5, 8, v5
+; GFX11-FAKE16-NEXT: v_or_b32_e32 v6, v14, v6
; GFX11-FAKE16-NEXT: v_or_b32_e32 v14, v15, v16
-; GFX11-FAKE16-NEXT: v_or_b32_e32 v15, v17, v18
-; GFX11-FAKE16-NEXT: v_or_b32_e32 v16, v19, v20
-; GFX11-FAKE16-NEXT: v_or_b32_e32 v3, v21, v3
-; GFX11-FAKE16-NEXT: v_and_b32_e32 v9, 0xffff, v9
-; GFX11-FAKE16-NEXT: v_lshlrev_b32_e32 v8, 16, v8
+; GFX11-FAKE16-NEXT: v_or_b32_e32 v15, v18, v19
+; GFX11-FAKE16-NEXT: v_or_b32_e32 v16, v20, v21
+; GFX11-FAKE16-NEXT: v_or_b32_e32 v5, v22, v5
+; GFX11-FAKE16-NEXT: v_and_b32_e32 v1, 0xffff, v1
+; GFX11-FAKE16-NEXT: v_lshlrev_b32_e32 v2, 16, v2
+; GFX11-FAKE16-NEXT: v_and_b32_e32 v8, 0xffff, v8
+; GFX11-FAKE16-NEXT: v_lshlrev_b32_e32 v9, 16, v9
; GFX11-FAKE16-NEXT: v_and_b32_e32 v10, 0xffff, v10
-; GFX11-FAKE16-NEXT: v_lshlrev_b32_e32 v11, 16, v11
-; GFX11-FAKE16-NEXT: v_and_b32_e32 v13, 0xffff, v13
-; GFX11-FAKE16-NEXT: v_lshlrev_b32_e32 v4, 16, v4
-; GFX11-FAKE16-NEXT: v_and_b32_e32 v14, 0xffff, v14
-; GFX11-FAKE16-NEXT: v_lshlrev_b32_e32 v15, 16, v15
+; GFX11-FAKE16-NEXT: v_lshlrev_b32_e32 v6, 16, v6
+; GFX11-FAKE16-NEXT: v_and_b32_e32 v18, 0xffff, v14
+; GFX11-FAKE16-NEXT: v_lshlrev_b32_e32 v19, 16, v15
; GFX11-FAKE16-NEXT: v_and_b32_e32 v16, 0xffff, v16
-; GFX11-FAKE16-NEXT: v_lshlrev_b32_e32 v3, 16, v3
-; GFX11-FAKE16-NEXT: v_or_b32_e32 v8, v9, v8
-; GFX11-FAKE16-NEXT: v_or_b32_e32 v9, v10, v11
-; GFX11-FAKE16-NEXT: v_or_b32_e32 v13, v13, v4
-; GFX11-FAKE16-NEXT: v_or_b32_e32 v14, v14, v15
-; GFX11-FAKE16-NEXT: v_or_b32_e32 v15, v16, v3
-; GFX11-FAKE16-NEXT: v_and_b32_e32 v3, 0xff, v36
-; GFX11-FAKE16-NEXT: v_lshlrev_b32_e32 v4, 8, v68
-; GFX11-FAKE16-NEXT: v_and_b32_e32 v10, 0xff, v37
-; GFX11-FAKE16-NEXT: v_lshlrev_b32_e32 v11, 8, v67
+; GFX11-FAKE16-NEXT: v_lshlrev_b32_e32 v5, 16, v5
+; GFX11-FAKE16-NEXT: v_or_b32_e32 v14, v1, v2
+; GFX11-FAKE16-NEXT: v_or_b32_e32 v15, v8, v9
+; GFX11-FAKE16-NEXT: v_or_b32_e32 v8, v10, v6
+; GFX11-FAKE16-NEXT: v_or_b32_e32 v9, v18, v19
+; GFX11-FAKE16-NEXT: v_or_b32_e32 v10, v16, v5
+; GFX11-FAKE16-NEXT: v_and_b32_e32 v1, 0xff, v37
+; GFX11-FAKE16-NEXT: v_lshlrev_b32_e32 v2, 8, v68
+; GFX11-FAKE16-NEXT: v_and_b32_e32 v5, 0xff, v38
+; GFX11-FAKE16-NEXT: v_lshlrev_b32_e32 v6, 8, v67
+; GFX11-FAKE16-NEXT: v_and_b32_e32 v16, 0xff, v50
+; GFX11-FAKE16-NEXT: v_lshlrev_b32_e32 v18, 8, v66
+; GFX11-FAKE16-NEXT: v_and_b32_e32 v19, 0xff, v65
+; GFX11-FAKE16-NEXT: v_lshlrev_b32_e32 v4, 8, v4
+; GFX11-FAKE16-NEXT: v_and_b32_e32 v20, 0xff, v48
+; GFX11-FAKE16-NEXT: v_lshlrev_b32_e32 v21, 8, v64
+; GFX11-FAKE16-NEXT: v_or_b32_e32 v1, v1, v2
+; GFX11-FAKE16-NEXT: v_or_b32_e32 v2, v5, v6
+; GFX11-FAKE16-NEXT: v_or_b32_e32 v5, v16, v18
+; GFX11-FAKE16-NEXT: v_or_b32_e32 v4, v19, v4
+; GFX11-FAKE16-NEXT: v_or_b32_e32 v6, v20, v21
; GFX11-FAKE16-NEXT: v_and_b32_e32 v16, 0xff, v49
-; GFX11-FAKE16-NEXT: v_lshlrev_b32_e32 v17, 8, v66
-; GFX11-FAKE16-NEXT: v_and_b32_e32 v18, 0xff, v65
-; GFX11-FAKE16-NEXT: v_lshlrev_b32_e32 v2, 8, v2
-; GFX11-FAKE16-NEXT: v_and_b32_e32 v19, 0xff, v39
-; GFX11-FAKE16-NEXT: v_lshlrev_b32_e32 v20, 8, v64
-; GFX11-FAKE16-NEXT: v_or_b32_e32 v3, v3, v4
-; GFX11-FAKE16-NEXT: v_or_b32_e32 v4, v10, v11
-; GFX11-FAKE16-NEXT: v_or_b32_e32 v10, v16, v17
-; GFX11-FAKE16-NEXT: v_or_b32_e32 v2, v18, v2
-; GFX11-FAKE16-NEXT: v_or_b32_e32 v11, v19, v20
-; GFX11-FAKE16-NEXT: v_and_b32_e32 v16, 0xff, v48
-; GFX11-FAKE16-NEXT: v_lshlrev_b32_e32 v17, 8, v55
-; GFX11-FAKE16-NEXT: v_and_b32_e32 v18, 0xff, v52
-; GFX11-FAKE16-NEXT: v_lshlrev_b32_e32 v19, 8, v54
-; GFX11-FAKE16-NEXT: v_and_b32_e32 v20, 0xff, v53
-; GFX11-FAKE16-NEXT: v_lshlrev_b32_e32 v1, 8, v1
-; GFX11-FAKE16-NEXT: v_and_b32_e32 v21, 0xff, v51
-; GFX11-FAKE16-NEXT: v_lshlrev_b32_e32 v12, 8, v12
-; GFX11-FAKE16-NEXT: v_and_b32_e32 v22, 0xff, v50
-; GFX11-FAKE16-NEXT: v_lshlrev_b32_e32 v5, 8, v5
-; GFX11-FAKE16-NEXT: v_or_b32_e32 v16, v16, v17
-; GFX11-FAKE16-NEXT: v_or_b32_e32 v17, v18, v19
-; GFX11-FAKE16-NEXT: v_or_b32_e32 v1, v20, v1
-; GFX11-FAKE16-NEXT: v_or_b32_e32 v12, v21, v12
-; GFX11-FAKE16-NEXT: v_or_b32_e32 v5, v22, v5
-; GFX11-FAKE16-NEXT: v_and_b32_e32 v3, 0xffff, v3
-; GFX11-FAKE16-NEXT: v_lshlrev_b32_e32 v4, 16, v4
-; GFX11-FAKE16-NEXT: v_and_b32_e32 v10, 0xffff, v10
+; GFX11-FAKE16-NEXT: v_lshlrev_b32_e32 v18, 8, v55
+; GFX11-FAKE16-NEXT: v_and_b32_e32 v19, 0xff, v53
+; GFX11-FAKE16-NEXT: v_lshlrev_b32_e32 v20, 8, v54
+; GFX11-FAKE16-NEXT: v_and_b32_e32 v17, 0xff, v17
+; GFX11-FAKE16-NEXT: v_lshlrev_b32_e32 v3, 8, v3
+; GFX11-FAKE16-NEXT: v_and_b32_e32 v21, 0xff, v52
+; GFX11-FAKE16-NEXT: v_lshlrev_b32_e32 v11, 8, v11
+; GFX11-FAKE16-NEXT: v_and_b32_e32 v22, 0xff, v51
+; GFX11-FAKE16-NEXT: v_lshlrev_b32_e32 v7, 8, v7
+; GFX11-FAKE16-NEXT: v_or_b32_e32 v16, v16, v18
+; GFX11-FAKE16-NEXT: v_or_b32_e32 v18, v19, v20
+; GFX11-FAKE16-NEXT: v_or_b32_e32 v3, v17, v3
+; GFX11-FAKE16-NEXT: v_or_b32_e32 v11, v21, v11
+; GFX11-FAKE16-NEXT: v_or_b32_e32 v7, v22, v7
+; GFX11-FAKE16-NEXT: v_and_b32_e32 v1, 0xffff, v1
; GFX11-FAKE16-NEXT: v_lshlrev_b32_e32 v2, 16, v2
-; GFX11-FAKE16-NEXT: v_and_b32_e32 v11, 0xffff, v11
-; GFX11-FAKE16-NEXT: v_lshlrev_b32_e32 v18, 16, v16
-; GFX11-FAKE16-NEXT: v_and_b32_e32 v17, 0xffff, v17
-; GFX11-FAKE16-NEXT: v_lshlrev_b32_e32 v19, 16, v1
-; GFX11-FAKE16-NEXT: v_and_b32_e32 v12, 0xffff, v12
-; GFX11-FAKE16-NEXT: v_lshlrev_b32_e32 v5, 16, v5
-; GFX11-FAKE16-NEXT: v_or_b32_e32 v16, v3, v4
-; GFX11-FAKE16-NEXT: v_or_b32_e32 v1, v10, v2
-; GFX11-FAKE16-NEXT: v_or_b32_e32 v2, v11, v18
-; GFX11-FAKE16-NEXT: v_or_b32_e32 v3, v17, v19
-; GFX11-FAKE16-NEXT: v_or_b32_e32 v4, v12, v5
+; GFX11-FAKE16-NEXT: v_and_b32_e32 v5, 0xffff, v5
+; GFX11-FAKE16-NEXT: v_lshlrev_b32_e32 v4, 16, v4
+; GFX11-FAKE16-NEXT: v_and_b32_e32 v6, 0xffff, v6
+; GFX11-FAKE16-NEXT: v_lshlrev_b32_e32 v16, 16, v16
+; GFX11-FAKE16-NEXT: v_and_b32_e32 v17, 0xffff, v18
+; GFX11-FAKE16-NEXT: v_lshlrev_b32_e32 v3, 16, v3
+; GFX11-FAKE16-NEXT: v_and_b32_e32 v18, 0xffff, v11
+; GFX11-FAKE16-NEXT: v_lshlrev_b32_e32 v7, 16, v7
+; GFX11-FAKE16-NEXT: v_or_b32_e32 v11, v1, v2
+; GFX11-FAKE16-NEXT: v_or_b32_e32 v1, v5, v4
+; GFX11-FAKE16-NEXT: v_or_b32_e32 v2, v6, v16
+; GFX11-FAKE16-NEXT: v_or_b32_e32 v3, v17, v3
+; GFX11-FAKE16-NEXT: v_or_b32_e32 v4, v18, v7
; GFX11-FAKE16-NEXT: s_clause 0x5
; GFX11-FAKE16-NEXT: scratch_store_b128 v0, v[97:100], off offset:32
; GFX11-FAKE16-NEXT: scratch_store_b128 v0, v[112:115], off offset:48
-; GFX11-FAKE16-NEXT: scratch_store_b128 v0, v[23:26], off offset:64
-; GFX11-FAKE16-NEXT: scratch_store_b128 v0, v[6:9], off offset:80
-; GFX11-FAKE16-NEXT: scratch_store_b128 v0, v[13:16], off offset:96
+; GFX11-FAKE16-NEXT: scratch_store_b128 v0, v[24:27], off offset:64
+; GFX11-FAKE16-NEXT: scratch_store_b128 v0, v[12:15], off offset:80
+; GFX11-FAKE16-NEXT: scratch_store_b128 v0, v[8:11], off offset:96
; GFX11-FAKE16-NEXT: scratch_store_b128 v0, v[1:4], off offset:112
; GFX11-FAKE16-NEXT: v_readlane_b32 s104, v41, 8
; GFX11-FAKE16-NEXT: v_readlane_b32 s103, v41, 7
@@ -178689,39 +178794,40 @@ define <64 x half> @bitcast_v128i8_to_v64f16(<128 x i8> %a, i32 %b) {
; SI-NEXT: buffer_store_dword v61, off, s[0:3], s32 offset:404 ; 4-byte Folded Spill
; SI-NEXT: buffer_store_dword v62, off, s[0:3], s32 offset:400 ; 4-byte Folded Spill
; SI-NEXT: buffer_store_dword v63, off, s[0:3], s32 offset:396 ; 4-byte Folded Spill
-; SI-NEXT: buffer_store_dword v29, off, s[0:3], s32 offset:472 ; 4-byte Folded Spill
-; SI-NEXT: buffer_store_dword v27, off, s[0:3], s32 offset:468 ; 4-byte Folded Spill
-; SI-NEXT: buffer_store_dword v25, off, s[0:3], s32 offset:464 ; 4-byte Folded Spill
-; SI-NEXT: buffer_store_dword v23, off, s[0:3], s32 offset:460 ; 4-byte Folded Spill
-; SI-NEXT: v_mov_b32_e32 v44, v19
-; SI-NEXT: v_mov_b32_e32 v43, v17
-; SI-NEXT: v_mov_b32_e32 v32, v14
-; SI-NEXT: v_mov_b32_e32 v14, v12
-; SI-NEXT: v_mov_b32_e32 v12, v10
-; SI-NEXT: v_mov_b32_e32 v41, v7
-; SI-NEXT: v_mov_b32_e32 v55, v5
-; SI-NEXT: v_mov_b32_e32 v54, v3
-; SI-NEXT: v_mov_b32_e32 v51, v1
-; SI-NEXT: v_mov_b32_e32 v10, v0
-; SI-NEXT: s_waitcnt expcnt(1)
-; SI-NEXT: buffer_load_dword v25, off, s[0:3], s32 offset:392
+; SI-NEXT: buffer_store_dword v29, off, s[0:3], s32 offset:508 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v27, off, s[0:3], s32 offset:504 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v25, off, s[0:3], s32 offset:500 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v23, off, s[0:3], s32 offset:496 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v21, off, s[0:3], s32 offset:492 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v19, off, s[0:3], s32 offset:488 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v17, off, s[0:3], s32 offset:484 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v15, off, s[0:3], s32 offset:480 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v13, off, s[0:3], s32 offset:476 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v11, off, s[0:3], s32 offset:472 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v9, off, s[0:3], s32 offset:468 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v7, off, s[0:3], s32 offset:464 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v5, off, s[0:3], s32 offset:460 ; 4-byte Folded Spill
+; SI-NEXT: v_mov_b32_e32 v51, v3
+; SI-NEXT: v_mov_b32_e32 v49, v1
+; SI-NEXT: s_waitcnt expcnt(0)
+; SI-NEXT: v_mov_b32_e32 v5, v0
+; SI-NEXT: buffer_load_dword v27, off, s[0:3], s32 offset:392
; SI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:4
; SI-NEXT: buffer_load_dword v3, off, s[0:3], s32 offset:12
; SI-NEXT: buffer_load_dword v7, off, s[0:3], s32 offset:20
-; SI-NEXT: buffer_load_dword v49, off, s[0:3], s32 offset:28
-; SI-NEXT: buffer_load_dword v50, off, s[0:3], s32 offset:36
-; SI-NEXT: buffer_load_dword v52, off, s[0:3], s32 offset:44
-; SI-NEXT: buffer_load_dword v53, off, s[0:3], s32 offset:52
-; SI-NEXT: buffer_load_dword v40, off, s[0:3], s32 offset:60
-; SI-NEXT: buffer_load_dword v42, off, s[0:3], s32 offset:68
-; SI-NEXT: buffer_load_dword v17, off, s[0:3], s32 offset:76
-; SI-NEXT: s_waitcnt expcnt(0)
+; SI-NEXT: buffer_load_dword v9, off, s[0:3], s32 offset:28
+; SI-NEXT: buffer_load_dword v11, off, s[0:3], s32 offset:36
+; SI-NEXT: buffer_load_dword v13, off, s[0:3], s32 offset:44
+; SI-NEXT: buffer_load_dword v15, off, s[0:3], s32 offset:52
+; SI-NEXT: buffer_load_dword v17, off, s[0:3], s32 offset:60
+; SI-NEXT: buffer_load_dword v19, off, s[0:3], s32 offset:68
+; SI-NEXT: buffer_load_dword v21, off, s[0:3], s32 offset:76
; SI-NEXT: buffer_load_dword v23, off, s[0:3], s32 offset:84
; SI-NEXT: buffer_load_dword v31, off, s[0:3], s32 offset:92
; SI-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:100
-; SI-NEXT: buffer_load_dword v27, off, s[0:3], s32 offset:108
+; SI-NEXT: buffer_load_dword v25, off, s[0:3], s32 offset:108
; SI-NEXT: buffer_load_dword v29, off, s[0:3], s32 offset:116
-; SI-NEXT: buffer_load_dword v5, off, s[0:3], s32 offset:124
+; SI-NEXT: buffer_load_dword v32, off, s[0:3], s32 offset:124
; SI-NEXT: buffer_load_dword v33, off, s[0:3], s32 offset:132
; SI-NEXT: buffer_load_dword v34, off, s[0:3], s32 offset:140
; SI-NEXT: buffer_load_dword v35, off, s[0:3], s32 offset:148
@@ -178730,129 +178836,135 @@ define <64 x half> @bitcast_v128i8_to_v64f16(<128 x i8> %a, i32 %b) {
; SI-NEXT: buffer_load_dword v38, off, s[0:3], s32 offset:172
; SI-NEXT: buffer_load_dword v39, off, s[0:3], s32 offset:180
; SI-NEXT: buffer_load_dword v48, off, s[0:3], s32 offset:188
-; SI-NEXT: buffer_load_dword v19, off, s[0:3], s32 offset:128
-; SI-NEXT: buffer_load_dword v45, off, s[0:3], s32 offset:136
+; SI-NEXT: buffer_load_dword v50, off, s[0:3], s32 offset:128
; SI-NEXT: v_lshlrev_b32_e32 v2, 8, v2
-; SI-NEXT: buffer_store_dword v2, off, s[0:3], s32 offset:476 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v2, off, s[0:3], s32 offset:516 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_lshlrev_b32_e32 v2, 8, v4
-; SI-NEXT: buffer_store_dword v2, off, s[0:3], s32 offset:484 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v2, off, s[0:3], s32 offset:524 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_lshlrev_b32_e32 v2, 8, v16
-; SI-NEXT: buffer_store_dword v2, off, s[0:3], s32 offset:492 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v2, off, s[0:3], s32 offset:532 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_lshlrev_b32_e32 v2, 8, v18
-; SI-NEXT: buffer_store_dword v2, off, s[0:3], s32 offset:496 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v2, off, s[0:3], s32 offset:536 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_lshlrev_b32_e32 v2, 8, v20
-; SI-NEXT: buffer_store_dword v2, off, s[0:3], s32 offset:500 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v2, off, s[0:3], s32 offset:540 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_lshlrev_b32_e32 v2, 8, v22
-; SI-NEXT: buffer_store_dword v2, off, s[0:3], s32 offset:504 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v2, off, s[0:3], s32 offset:544 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_lshlrev_b32_e32 v2, 8, v24
-; SI-NEXT: buffer_store_dword v2, off, s[0:3], s32 offset:508 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v2, off, s[0:3], s32 offset:548 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_lshlrev_b32_e32 v2, 8, v26
-; SI-NEXT: buffer_store_dword v2, off, s[0:3], s32 offset:512 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v2, off, s[0:3], s32 offset:552 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_lshlrev_b32_e32 v2, 8, v28
-; SI-NEXT: buffer_store_dword v2, off, s[0:3], s32 offset:516 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v2, off, s[0:3], s32 offset:556 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_lshlrev_b32_e32 v2, 8, v30
-; SI-NEXT: buffer_store_dword v2, off, s[0:3], s32 offset:520 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v2, off, s[0:3], s32 offset:560 ; 4-byte Folded Spill
; SI-NEXT: v_lshlrev_b32_e32 v6, 8, v6
; SI-NEXT: v_lshlrev_b32_e32 v8, 8, v8
+; SI-NEXT: v_lshlrev_b32_e32 v10, 8, v10
; SI-NEXT: v_lshlrev_b32_e32 v12, 8, v12
; SI-NEXT: v_lshlrev_b32_e32 v14, 8, v14
-; SI-NEXT: v_lshlrev_b32_e32 v32, 8, v32
; SI-NEXT: s_waitcnt vmcnt(14)
-; SI-NEXT: v_cmp_ne_u32_e32 vcc, 0, v25
+; SI-NEXT: v_cmp_ne_u32_e32 vcc, 0, v27
; SI-NEXT: v_lshlrev_b32_e32 v1, 8, v1
-; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:524 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:564 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_lshlrev_b32_e32 v1, 8, v3
-; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:528 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:568 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_lshlrev_b32_e32 v1, 8, v49
-; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:532 ; 4-byte Folded Spill
+; SI-NEXT: v_lshlrev_b32_e32 v1, 8, v9
+; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:572 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_lshlrev_b32_e32 v1, 8, v50
-; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:536 ; 4-byte Folded Spill
+; SI-NEXT: v_lshlrev_b32_e32 v1, 8, v11
+; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:576 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_lshlrev_b32_e32 v1, 8, v52
-; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:540 ; 4-byte Folded Spill
+; SI-NEXT: v_lshlrev_b32_e32 v1, 8, v13
+; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:580 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_lshlrev_b32_e32 v1, 8, v53
-; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:544 ; 4-byte Folded Spill
+; SI-NEXT: v_lshlrev_b32_e32 v1, 8, v15
+; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:584 ; 4-byte Folded Spill
; SI-NEXT: v_lshlrev_b32_e32 v0, 8, v0
-; SI-NEXT: buffer_store_dword v0, off, s[0:3], s32 offset:632 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v0, off, s[0:3], s32 offset:664 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_lshlrev_b32_e32 v0, 8, v27
-; SI-NEXT: buffer_store_dword v0, off, s[0:3], s32 offset:648 ; 4-byte Folded Spill
+; SI-NEXT: v_lshlrev_b32_e32 v0, 8, v25
+; SI-NEXT: buffer_store_dword v0, off, s[0:3], s32 offset:680 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_lshlrev_b32_e32 v0, 8, v29
-; SI-NEXT: v_lshlrev_b32_e32 v1, 8, v40
-; SI-NEXT: buffer_store_dword v0, off, s[0:3], s32 offset:664 ; 4-byte Folded Spill
+; SI-NEXT: v_lshlrev_b32_e32 v1, 8, v17
+; SI-NEXT: buffer_store_dword v0, off, s[0:3], s32 offset:696 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_lshlrev_b32_e32 v0, 8, v5
-; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:548 ; 4-byte Folded Spill
+; SI-NEXT: v_lshlrev_b32_e32 v0, 8, v32
+; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:588 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_lshlrev_b32_e32 v1, 8, v42
-; SI-NEXT: buffer_store_dword v0, off, s[0:3], s32 offset:680 ; 4-byte Folded Spill
+; SI-NEXT: v_lshlrev_b32_e32 v1, 8, v19
+; SI-NEXT: buffer_store_dword v0, off, s[0:3], s32 offset:712 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_lshlrev_b32_e32 v0, 8, v33
-; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:556 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt vmcnt(14)
-; SI-NEXT: buffer_store_dword v45, off, s[0:3], s32 offset:480 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v50, off, s[0:3], s32 offset:512 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: buffer_load_dword v45, off, s[0:3], s32 offset:144
-; SI-NEXT: v_lshlrev_b32_e32 v1, 8, v17
-; SI-NEXT: buffer_store_dword v0, off, s[0:3], s32 offset:696 ; 4-byte Folded Spill
+; SI-NEXT: buffer_load_dword v50, off, s[0:3], s32 offset:136
+; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:592 ; 4-byte Folded Spill
+; SI-NEXT: s_waitcnt expcnt(0)
+; SI-NEXT: v_lshlrev_b32_e32 v1, 8, v21
+; SI-NEXT: buffer_store_dword v0, off, s[0:3], s32 offset:728 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_lshlrev_b32_e32 v0, 8, v34
-; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:572 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:604 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_lshlrev_b32_e32 v1, 8, v23
-; SI-NEXT: buffer_store_dword v0, off, s[0:3], s32 offset:712 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v0, off, s[0:3], s32 offset:744 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_lshlrev_b32_e32 v0, 8, v35
-; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:592 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:624 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_lshlrev_b32_e32 v1, 8, v31
-; SI-NEXT: buffer_store_dword v0, off, s[0:3], s32 offset:728 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v0, off, s[0:3], s32 offset:760 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_lshlrev_b32_e32 v0, 8, v36
-; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:612 ; 4-byte Folded Spill
-; SI-NEXT: buffer_store_dword v0, off, s[0:3], s32 offset:744 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:644 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v0, off, s[0:3], s32 offset:776 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_lshlrev_b32_e32 v0, 8, v37
; SI-NEXT: v_lshlrev_b32_e32 v7, 8, v7
-; SI-NEXT: ; implicit-def: $vgpr36
-; SI-NEXT: ; implicit-def: $vgpr33
+; SI-NEXT: ; implicit-def: $vgpr35
+; SI-NEXT: ; implicit-def: $vgpr32
; SI-NEXT: ; implicit-def: $vgpr29
; SI-NEXT: ; implicit-def: $vgpr37
; SI-NEXT: ; implicit-def: $vgpr23
-; SI-NEXT: ; implicit-def: $vgpr35
-; SI-NEXT: ; implicit-def: $vgpr17
+; SI-NEXT: ; implicit-def: $vgpr33
+; SI-NEXT: ; implicit-def: $vgpr13
; SI-NEXT: ; implicit-def: $vgpr31
-; SI-NEXT: ; implicit-def: $vgpr5
-; SI-NEXT: ; implicit-def: $vgpr27
-; SI-NEXT: s_waitcnt vmcnt(7)
-; SI-NEXT: buffer_store_dword v45, off, s[0:3], s32 offset:488 ; 4-byte Folded Spill
-; SI-NEXT: buffer_load_dword v50, off, s[0:3], s32 offset:160
-; SI-NEXT: buffer_load_dword v53, off, s[0:3], s32 offset:168
-; SI-NEXT: buffer_load_dword v40, off, s[0:3], s32 offset:176
-; SI-NEXT: buffer_store_dword v0, off, s[0:3], s32 offset:772 ; 4-byte Folded Spill
+; SI-NEXT: ; implicit-def: $vgpr11
+; SI-NEXT: ; implicit-def: $vgpr25
+; SI-NEXT: ; implicit-def: $vgpr9
+; SI-NEXT: ; implicit-def: $vgpr21
+; SI-NEXT: s_waitcnt vmcnt(8)
+; SI-NEXT: buffer_store_dword v50, off, s[0:3], s32 offset:520 ; 4-byte Folded Spill
+; SI-NEXT: s_waitcnt expcnt(0)
+; SI-NEXT: buffer_load_dword v50, off, s[0:3], s32 offset:144
+; SI-NEXT: s_waitcnt vmcnt(0)
+; SI-NEXT: buffer_store_dword v50, off, s[0:3], s32 offset:528 ; 4-byte Folded Spill
+; SI-NEXT: buffer_load_dword v15, off, s[0:3], s32 offset:160
+; SI-NEXT: buffer_load_dword v17, off, s[0:3], s32 offset:168
+; SI-NEXT: buffer_load_dword v19, off, s[0:3], s32 offset:176
+; SI-NEXT: buffer_store_dword v0, off, s[0:3], s32 offset:804 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_lshlrev_b32_e32 v0, 8, v38
-; SI-NEXT: buffer_store_dword v0, off, s[0:3], s32 offset:788 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v0, off, s[0:3], s32 offset:820 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_lshlrev_b32_e32 v0, 8, v39
-; SI-NEXT: buffer_store_dword v0, off, s[0:3], s32 offset:796 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v0, off, s[0:3], s32 offset:828 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_lshlrev_b32_e32 v0, 8, v48
-; SI-NEXT: buffer_store_dword v0, off, s[0:3], s32 offset:812 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v0, off, s[0:3], s32 offset:844 ; 4-byte Folded Spill
; SI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:196
; SI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:204
; SI-NEXT: buffer_load_dword v3, off, s[0:3], s32 offset:212
@@ -178863,27 +178975,27 @@ define <64 x half> @bitcast_v128i8_to_v64f16(<128 x i8> %a, i32 %b) {
; SI-NEXT: ; implicit-def: $vgpr38
; SI-NEXT: ; implicit-def: $vgpr39
; SI-NEXT: s_waitcnt vmcnt(0)
-; SI-NEXT: buffer_store_dword v0, off, s[0:3], s32 offset:584 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v0, off, s[0:3], s32 offset:616 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:200
; SI-NEXT: s_waitcnt vmcnt(0)
-; SI-NEXT: buffer_store_dword v0, off, s[0:3], s32 offset:604 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v0, off, s[0:3], s32 offset:636 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:208
; SI-NEXT: s_waitcnt vmcnt(0)
-; SI-NEXT: buffer_store_dword v0, off, s[0:3], s32 offset:624 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v0, off, s[0:3], s32 offset:656 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_lshlrev_b32_e32 v0, 8, v1
-; SI-NEXT: buffer_store_dword v0, off, s[0:3], s32 offset:852 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v0, off, s[0:3], s32 offset:884 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_lshlrev_b32_e32 v0, 8, v2
-; SI-NEXT: buffer_store_dword v0, off, s[0:3], s32 offset:860 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v0, off, s[0:3], s32 offset:892 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_lshlrev_b32_e32 v0, 8, v3
-; SI-NEXT: buffer_store_dword v0, off, s[0:3], s32 offset:864 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v0, off, s[0:3], s32 offset:896 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_lshlrev_b32_e32 v0, 8, v4
-; SI-NEXT: buffer_store_dword v0, off, s[0:3], s32 offset:868 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v0, off, s[0:3], s32 offset:900 ; 4-byte Folded Spill
; SI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:228
; SI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:236
; SI-NEXT: buffer_load_dword v3, off, s[0:3], s32 offset:244
@@ -178891,82 +179003,78 @@ define <64 x half> @bitcast_v128i8_to_v64f16(<128 x i8> %a, i32 %b) {
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:224
; SI-NEXT: s_waitcnt vmcnt(0)
-; SI-NEXT: buffer_store_dword v0, off, s[0:3], s32 offset:752 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v0, off, s[0:3], s32 offset:784 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:232
; SI-NEXT: s_waitcnt vmcnt(0)
-; SI-NEXT: buffer_store_dword v0, off, s[0:3], s32 offset:768 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v0, off, s[0:3], s32 offset:800 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:240
; SI-NEXT: s_waitcnt vmcnt(0)
-; SI-NEXT: buffer_store_dword v0, off, s[0:3], s32 offset:780 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v0, off, s[0:3], s32 offset:812 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_lshlrev_b32_e32 v0, 8, v1
-; SI-NEXT: buffer_store_dword v0, off, s[0:3], s32 offset:904 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v0, off, s[0:3], s32 offset:936 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_lshlrev_b32_e32 v0, 8, v2
-; SI-NEXT: buffer_store_dword v0, off, s[0:3], s32 offset:912 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v0, off, s[0:3], s32 offset:944 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_lshlrev_b32_e32 v0, 8, v3
-; SI-NEXT: buffer_store_dword v0, off, s[0:3], s32 offset:916 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v0, off, s[0:3], s32 offset:948 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_lshlrev_b32_e32 v0, 8, v4
-; SI-NEXT: buffer_store_dword v0, off, s[0:3], s32 offset:920 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v0, off, s[0:3], s32 offset:952 ; 4-byte Folded Spill
; SI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:260
; SI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:268
; SI-NEXT: buffer_load_dword v3, off, s[0:3], s32 offset:276
; SI-NEXT: buffer_load_dword v4, off, s[0:3], s32 offset:284
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:256
+; SI-NEXT: s_waitcnt vmcnt(1)
+; SI-NEXT: v_lshlrev_b32_e32 v54, 8, v4
; SI-NEXT: s_waitcnt vmcnt(0)
-; SI-NEXT: buffer_store_dword v0, off, s[0:3], s32 offset:844 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v0, off, s[0:3], s32 offset:876 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:264
; SI-NEXT: s_waitcnt vmcnt(0)
-; SI-NEXT: buffer_store_dword v0, off, s[0:3], s32 offset:848 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v0, off, s[0:3], s32 offset:880 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:272
; SI-NEXT: s_waitcnt vmcnt(0)
-; SI-NEXT: buffer_store_dword v0, off, s[0:3], s32 offset:856 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v0, off, s[0:3], s32 offset:888 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_lshlrev_b32_e32 v0, 8, v1
-; SI-NEXT: buffer_store_dword v0, off, s[0:3], s32 offset:936 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v0, off, s[0:3], s32 offset:964 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_lshlrev_b32_e32 v0, 8, v2
-; SI-NEXT: buffer_store_dword v0, off, s[0:3], s32 offset:944 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v0, off, s[0:3], s32 offset:972 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_lshlrev_b32_e32 v0, 8, v3
-; SI-NEXT: buffer_store_dword v0, off, s[0:3], s32 offset:948 ; 4-byte Folded Spill
-; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_lshlrev_b32_e32 v0, 8, v4
-; SI-NEXT: buffer_store_dword v0, off, s[0:3], s32 offset:952 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v0, off, s[0:3], s32 offset:976 ; 4-byte Folded Spill
; SI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:292
; SI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:300
; SI-NEXT: buffer_load_dword v3, off, s[0:3], s32 offset:308
; SI-NEXT: buffer_load_dword v4, off, s[0:3], s32 offset:316
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:288
+; SI-NEXT: s_waitcnt vmcnt(4)
+; SI-NEXT: v_lshlrev_b32_e32 v40, 8, v1
+; SI-NEXT: s_waitcnt vmcnt(3)
+; SI-NEXT: v_lshlrev_b32_e32 v43, 8, v2
+; SI-NEXT: s_waitcnt vmcnt(2)
+; SI-NEXT: v_lshlrev_b32_e32 v44, 8, v3
; SI-NEXT: s_waitcnt vmcnt(1)
; SI-NEXT: v_lshlrev_b32_e32 v45, 8, v4
; SI-NEXT: s_waitcnt vmcnt(0)
-; SI-NEXT: buffer_store_dword v0, off, s[0:3], s32 offset:896 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v0, off, s[0:3], s32 offset:928 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:296
; SI-NEXT: s_waitcnt vmcnt(0)
-; SI-NEXT: buffer_store_dword v0, off, s[0:3], s32 offset:900 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v0, off, s[0:3], s32 offset:932 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:304
; SI-NEXT: s_waitcnt vmcnt(0)
-; SI-NEXT: buffer_store_dword v0, off, s[0:3], s32 offset:908 ; 4-byte Folded Spill
-; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_lshlrev_b32_e32 v0, 8, v1
-; SI-NEXT: buffer_store_dword v0, off, s[0:3], s32 offset:968 ; 4-byte Folded Spill
-; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_lshlrev_b32_e32 v0, 8, v2
-; SI-NEXT: buffer_store_dword v0, off, s[0:3], s32 offset:976 ; 4-byte Folded Spill
-; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_lshlrev_b32_e32 v0, 8, v3
-; SI-NEXT: buffer_store_dword v0, off, s[0:3], s32 offset:980 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v0, off, s[0:3], s32 offset:940 ; 4-byte Folded Spill
; SI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:324
; SI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:332
; SI-NEXT: buffer_load_dword v3, off, s[0:3], s32 offset:340
@@ -178982,79 +179090,64 @@ define <64 x half> @bitcast_v128i8_to_v64f16(<128 x i8> %a, i32 %b) {
; SI-NEXT: s_waitcnt vmcnt(1)
; SI-NEXT: v_lshlrev_b32_e32 v61, 8, v4
; SI-NEXT: s_waitcnt vmcnt(0)
-; SI-NEXT: buffer_store_dword v0, off, s[0:3], s32 offset:924 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v0, off, s[0:3], s32 offset:956 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:328
; SI-NEXT: s_waitcnt vmcnt(0)
-; SI-NEXT: buffer_store_dword v0, off, s[0:3], s32 offset:932 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v0, off, s[0:3], s32 offset:960 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:336
; SI-NEXT: s_waitcnt vmcnt(0)
-; SI-NEXT: buffer_store_dword v0, off, s[0:3], s32 offset:940 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v0, off, s[0:3], s32 offset:968 ; 4-byte Folded Spill
; SI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:356
; SI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:364
; SI-NEXT: buffer_load_dword v3, off, s[0:3], s32 offset:372
; SI-NEXT: buffer_load_dword v4, off, s[0:3], s32 offset:380
+; SI-NEXT: buffer_load_dword v55, off, s[0:3], s32 offset:352
+; SI-NEXT: buffer_load_dword v41, off, s[0:3], s32 offset:360
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:352
-; SI-NEXT: s_waitcnt vmcnt(4)
+; SI-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:368
+; SI-NEXT: s_waitcnt vmcnt(6)
; SI-NEXT: v_lshlrev_b32_e32 v62, 8, v1
-; SI-NEXT: s_waitcnt vmcnt(3)
-; SI-NEXT: v_lshlrev_b32_e32 v25, 8, v2
-; SI-NEXT: s_waitcnt vmcnt(2)
+; SI-NEXT: s_waitcnt vmcnt(5)
+; SI-NEXT: v_lshlrev_b32_e32 v27, 8, v2
+; SI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:388
+; SI-NEXT: buffer_load_dword v47, off, s[0:3], s32 offset:384
+; SI-NEXT: s_waitcnt vmcnt(6)
; SI-NEXT: v_lshlrev_b32_e32 v63, 8, v3
-; SI-NEXT: s_waitcnt vmcnt(1)
+; SI-NEXT: s_waitcnt vmcnt(5)
; SI-NEXT: v_lshlrev_b32_e32 v1, 8, v4
+; SI-NEXT: s_waitcnt vmcnt(1)
+; SI-NEXT: v_lshlrev_b32_e32 v3, 8, v2
+; SI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:48
; SI-NEXT: s_waitcnt vmcnt(0)
-; SI-NEXT: buffer_store_dword v0, off, s[0:3], s32 offset:960 ; 4-byte Folded Spill
-; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:360
-; SI-NEXT: s_waitcnt vmcnt(0)
-; SI-NEXT: buffer_store_dword v0, off, s[0:3], s32 offset:964 ; 4-byte Folded Spill
-; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:368
-; SI-NEXT: s_waitcnt vmcnt(0)
-; SI-NEXT: buffer_store_dword v0, off, s[0:3], s32 offset:972 ; 4-byte Folded Spill
-; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:48
-; SI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:388
-; SI-NEXT: buffer_load_dword v47, off, s[0:3], s32 offset:384
-; SI-NEXT: s_waitcnt vmcnt(2)
-; SI-NEXT: buffer_store_dword v0, off, s[0:3], s32 offset:892 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v2, off, s[0:3], s32 offset:924 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:40
-; SI-NEXT: s_waitcnt vmcnt(3)
-; SI-NEXT: v_lshlrev_b32_e32 v3, 8, v2
+; SI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:40
; SI-NEXT: s_waitcnt vmcnt(0)
-; SI-NEXT: buffer_store_dword v0, off, s[0:3], s32 offset:888 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v2, off, s[0:3], s32 offset:920 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:32
+; SI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:32
; SI-NEXT: s_waitcnt vmcnt(0)
-; SI-NEXT: buffer_store_dword v0, off, s[0:3], s32 offset:884 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v2, off, s[0:3], s32 offset:916 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:24
+; SI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:24
; SI-NEXT: s_waitcnt vmcnt(0)
-; SI-NEXT: buffer_store_dword v0, off, s[0:3], s32 offset:880 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v2, off, s[0:3], s32 offset:912 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:16
+; SI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:16
; SI-NEXT: s_waitcnt vmcnt(0)
-; SI-NEXT: buffer_store_dword v0, off, s[0:3], s32 offset:876 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v2, off, s[0:3], s32 offset:908 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:8
+; SI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:8
; SI-NEXT: s_waitcnt vmcnt(0)
-; SI-NEXT: buffer_store_dword v0, off, s[0:3], s32 offset:872 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v2, off, s[0:3], s32 offset:904 ; 4-byte Folded Spill
; SI-NEXT: buffer_load_dword v16, off, s[0:3], s32 offset:112
; SI-NEXT: buffer_load_dword v20, off, s[0:3], s32 offset:120
-; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:152
-; SI-NEXT: s_waitcnt vmcnt(0)
-; SI-NEXT: buffer_store_dword v0, off, s[0:3], s32 offset:928 ; 4-byte Folded Spill
-; SI-NEXT: buffer_load_dword v49, off, s[0:3], s32 offset:184
+; SI-NEXT: buffer_load_dword v36, off, s[0:3], s32 offset:152
+; SI-NEXT: buffer_load_dword v50, off, s[0:3], s32 offset:184
; SI-NEXT: buffer_load_dword v52, off, s[0:3], s32 offset:216
-; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:248
-; SI-NEXT: s_waitcnt vmcnt(0)
-; SI-NEXT: buffer_store_dword v0, off, s[0:3], s32 offset:956 ; 4-byte Folded Spill
+; SI-NEXT: buffer_load_dword v53, off, s[0:3], s32 offset:248
; SI-NEXT: buffer_load_dword v42, off, s[0:3], s32 offset:280
; SI-NEXT: buffer_load_dword v46, off, s[0:3], s32 offset:312
; SI-NEXT: buffer_load_dword v56, off, s[0:3], s32 offset:344
@@ -179067,6 +179160,265 @@ define <64 x half> @bitcast_v128i8_to_v64f16(<128 x i8> %a, i32 %b) {
; SI-NEXT: buffer_load_dword v24, off, s[0:3], s32 offset:72
; SI-NEXT: buffer_load_dword v22, off, s[0:3], s32 offset:64
; SI-NEXT: buffer_load_dword v18, off, s[0:3], s32 offset:56
+; SI-NEXT: ; implicit-def: $vgpr2
+; SI-NEXT: ; kill: killed $vgpr2
+; SI-NEXT: ; implicit-def: $vgpr2
+; SI-NEXT: ; kill: killed $vgpr2
+; SI-NEXT: ; implicit-def: $vgpr2
+; SI-NEXT: ; kill: killed $vgpr2
+; SI-NEXT: ; implicit-def: $vgpr2
+; SI-NEXT: ; kill: killed $vgpr2
+; SI-NEXT: ; implicit-def: $vgpr2
+; SI-NEXT: ; kill: killed $vgpr2
+; SI-NEXT: ; implicit-def: $vgpr2
+; SI-NEXT: ; kill: killed $vgpr2
+; SI-NEXT: ; implicit-def: $vgpr2
+; SI-NEXT: ; kill: killed $vgpr2
+; SI-NEXT: ; implicit-def: $vgpr2
+; SI-NEXT: ; kill: killed $vgpr2
+; SI-NEXT: ; implicit-def: $vgpr2
+; SI-NEXT: ; kill: killed $vgpr2
+; SI-NEXT: ; implicit-def: $vgpr2
+; SI-NEXT: ; kill: killed $vgpr2
+; SI-NEXT: ; implicit-def: $vgpr2
+; SI-NEXT: ; kill: killed $vgpr2
+; SI-NEXT: ; implicit-def: $vgpr2
+; SI-NEXT: ; kill: killed $vgpr2
+; SI-NEXT: ; implicit-def: $vgpr2
+; SI-NEXT: ; kill: killed $vgpr2
+; SI-NEXT: ; implicit-def: $vgpr2
+; SI-NEXT: ; kill: killed $vgpr2
+; SI-NEXT: ; implicit-def: $vgpr2
+; SI-NEXT: ; kill: killed $vgpr2
+; SI-NEXT: ; implicit-def: $vgpr2
+; SI-NEXT: ; kill: killed $vgpr2
+; SI-NEXT: ; implicit-def: $vgpr2
+; SI-NEXT: ; kill: killed $vgpr2
+; SI-NEXT: ; implicit-def: $vgpr2
+; SI-NEXT: ; kill: killed $vgpr2
+; SI-NEXT: ; implicit-def: $vgpr2
+; SI-NEXT: ; kill: killed $vgpr2
+; SI-NEXT: ; implicit-def: $vgpr2
+; SI-NEXT: ; kill: killed $vgpr2
+; SI-NEXT: ; implicit-def: $vgpr2
+; SI-NEXT: ; kill: killed $vgpr2
+; SI-NEXT: ; implicit-def: $vgpr2
+; SI-NEXT: ; kill: killed $vgpr2
+; SI-NEXT: ; implicit-def: $vgpr2
+; SI-NEXT: ; kill: killed $vgpr2
+; SI-NEXT: ; implicit-def: $vgpr2
+; SI-NEXT: ; kill: killed $vgpr2
+; SI-NEXT: ; implicit-def: $vgpr2
+; SI-NEXT: ; kill: killed $vgpr2
+; SI-NEXT: ; implicit-def: $vgpr2
+; SI-NEXT: ; kill: killed $vgpr2
+; SI-NEXT: ; implicit-def: $vgpr2
+; SI-NEXT: ; kill: killed $vgpr2
+; SI-NEXT: ; implicit-def: $vgpr2
+; SI-NEXT: ; kill: killed $vgpr2
+; SI-NEXT: ; implicit-def: $vgpr2
+; SI-NEXT: ; kill: killed $vgpr2
+; SI-NEXT: ; implicit-def: $vgpr2
+; SI-NEXT: ; kill: killed $vgpr2
+; SI-NEXT: ; implicit-def: $vgpr2
+; SI-NEXT: ; kill: killed $vgpr2
+; SI-NEXT: ; implicit-def: $vgpr2
+; SI-NEXT: ; kill: killed $vgpr2
+; SI-NEXT: ; implicit-def: $vgpr2
+; SI-NEXT: ; kill: killed $vgpr2
+; SI-NEXT: ; implicit-def: $vgpr2
+; SI-NEXT: ; kill: killed $vgpr2
+; SI-NEXT: ; implicit-def: $vgpr2
+; SI-NEXT: ; kill: killed $vgpr2
+; SI-NEXT: ; implicit-def: $vgpr2
+; SI-NEXT: ; kill: killed $vgpr2
+; SI-NEXT: ; implicit-def: $vgpr2
+; SI-NEXT: ; kill: killed $vgpr2
+; SI-NEXT: ; implicit-def: $vgpr2
+; SI-NEXT: ; kill: killed $vgpr2
+; SI-NEXT: ; implicit-def: $vgpr2
+; SI-NEXT: ; kill: killed $vgpr2
+; SI-NEXT: ; implicit-def: $vgpr2
+; SI-NEXT: ; kill: killed $vgpr2
+; SI-NEXT: ; implicit-def: $vgpr2
+; SI-NEXT: ; kill: killed $vgpr2
+; SI-NEXT: ; implicit-def: $vgpr2
+; SI-NEXT: ; kill: killed $vgpr2
+; SI-NEXT: ; implicit-def: $vgpr2
+; SI-NEXT: ; kill: killed $vgpr2
+; SI-NEXT: ; implicit-def: $vgpr2
+; SI-NEXT: ; kill: killed $vgpr2
+; SI-NEXT: ; implicit-def: $vgpr2
+; SI-NEXT: ; kill: killed $vgpr2
+; SI-NEXT: ; implicit-def: $vgpr2
+; SI-NEXT: ; kill: killed $vgpr2
+; SI-NEXT: ; implicit-def: $vgpr2
+; SI-NEXT: ; kill: killed $vgpr2
+; SI-NEXT: ; implicit-def: $vgpr2
+; SI-NEXT: ; kill: killed $vgpr2
+; SI-NEXT: ; implicit-def: $vgpr2
+; SI-NEXT: ; kill: killed $vgpr2
+; SI-NEXT: s_and_saveexec_b64 s[4:5], vcc
+; SI-NEXT: s_xor_b64 s[4:5], exec, s[4:5]
+; SI-NEXT: s_cbranch_execz .LBB92_2
+; SI-NEXT: ; %bb.1: ; %cmp.false
+; SI-NEXT: v_mov_b32_e32 v11, v7
+; SI-NEXT: v_mov_b32_e32 v7, v6
+; SI-NEXT: buffer_load_dword v6, off, s[0:3], s32 offset:516 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v9, off, s[0:3], s32 offset:976 ; 4-byte Folded Reload
+; SI-NEXT: s_waitcnt expcnt(0)
+; SI-NEXT: v_and_b32_e32 v2, 0xff, v49
+; SI-NEXT: s_waitcnt vmcnt(2)
+; SI-NEXT: v_and_b32_e32 v18, 0xff, v18
+; SI-NEXT: v_and_b32_e32 v22, 0xff, v22
+; SI-NEXT: v_and_b32_e32 v24, 0xff, v24
+; SI-NEXT: v_and_b32_e32 v26, 0xff, v26
+; SI-NEXT: v_and_b32_e32 v28, 0xff, v28
+; SI-NEXT: v_and_b32_e32 v30, 0xff, v30
+; SI-NEXT: v_and_b32_e32 v34, 0xff, v34
+; SI-NEXT: v_and_b32_e32 v16, 0xff, v16
+; SI-NEXT: v_and_b32_e32 v20, 0xff, v20
+; SI-NEXT: v_and_b32_e32 v36, 0xff, v36
+; SI-NEXT: v_and_b32_e32 v50, 0xff, v50
+; SI-NEXT: v_and_b32_e32 v52, 0xff, v52
+; SI-NEXT: v_and_b32_e32 v42, 0xff, v42
+; SI-NEXT: v_or_b32_e32 v54, v42, v54
+; SI-NEXT: v_and_b32_e32 v53, 0xff, v53
+; SI-NEXT: v_and_b32_e32 v41, 0xff, v41
+; SI-NEXT: v_or_b32_e32 v27, v41, v27
+; SI-NEXT: v_and_b32_e32 v41, 0xff, v57
+; SI-NEXT: v_or_b32_e32 v1, v41, v1
+; SI-NEXT: v_and_b32_e32 v41, 0xff, v47
+; SI-NEXT: v_or_b32_e32 v3, v41, v3
+; SI-NEXT: v_and_b32_e32 v0, 0xff, v0
+; SI-NEXT: v_or_b32_e32 v0, v0, v63
+; SI-NEXT: v_and_b32_e32 v56, 0xff, v56
+; SI-NEXT: v_and_b32_e32 v55, 0xff, v55
+; SI-NEXT: v_or_b32_e32 v56, v56, v61
+; SI-NEXT: v_or_b32_e32 v55, v55, v62
+; SI-NEXT: ; implicit-def: $vgpr57
+; SI-NEXT: ; implicit-def: $vgpr47
+; SI-NEXT: ; implicit-def: $vgpr61
+; SI-NEXT: ; implicit-def: $vgpr62
+; SI-NEXT: ; implicit-def: $vgpr63
+; SI-NEXT: buffer_load_dword v21, off, s[0:3], s32 offset:892 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v48, off, s[0:3], s32 offset:964 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v42, off, s[0:3], s32 offset:928 ; 4-byte Folded Reload
+; SI-NEXT: s_waitcnt vmcnt(4)
+; SI-NEXT: v_or_b32_e32 v2, v2, v6
+; SI-NEXT: buffer_load_dword v6, off, s[0:3], s32 offset:524 ; 4-byte Folded Reload
+; SI-NEXT: buffer_store_dword v2, off, s[0:3], s32 offset:516 ; 4-byte Folded Spill
+; SI-NEXT: s_waitcnt expcnt(0)
+; SI-NEXT: v_and_b32_e32 v2, 0xff, v51
+; SI-NEXT: buffer_load_dword v51, off, s[0:3], s32 offset:972 ; 4-byte Folded Reload
+; SI-NEXT: s_waitcnt vmcnt(2)
+; SI-NEXT: v_or_b32_e32 v2, v2, v6
+; SI-NEXT: buffer_store_dword v2, off, s[0:3], s32 offset:524 ; 4-byte Folded Spill
+; SI-NEXT: s_waitcnt expcnt(0)
+; SI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:460 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v6, off, s[0:3], s32 offset:532 ; 4-byte Folded Reload
+; SI-NEXT: s_waitcnt vmcnt(1)
+; SI-NEXT: v_and_b32_e32 v2, 0xff, v2
+; SI-NEXT: v_or_b32_e32 v2, v2, v7
+; SI-NEXT: buffer_store_dword v2, off, s[0:3], s32 offset:460 ; 4-byte Folded Spill
+; SI-NEXT: s_waitcnt expcnt(0)
+; SI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:464 ; 4-byte Folded Reload
+; SI-NEXT: s_waitcnt vmcnt(0)
+; SI-NEXT: v_and_b32_e32 v2, 0xff, v2
+; SI-NEXT: v_or_b32_e32 v2, v2, v8
+; SI-NEXT: buffer_store_dword v2, off, s[0:3], s32 offset:464 ; 4-byte Folded Spill
+; SI-NEXT: s_waitcnt expcnt(0)
+; SI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:468 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v8, off, s[0:3], s32 offset:572 ; 4-byte Folded Reload
+; SI-NEXT: s_waitcnt vmcnt(1)
+; SI-NEXT: v_and_b32_e32 v2, 0xff, v2
+; SI-NEXT: v_or_b32_e32 v2, v2, v10
+; SI-NEXT: buffer_store_dword v2, off, s[0:3], s32 offset:468 ; 4-byte Folded Spill
+; SI-NEXT: s_waitcnt expcnt(0)
+; SI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:472 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v10, off, s[0:3], s32 offset:744 ; 4-byte Folded Reload
+; SI-NEXT: s_waitcnt vmcnt(1)
+; SI-NEXT: v_and_b32_e32 v2, 0xff, v2
+; SI-NEXT: v_or_b32_e32 v2, v2, v12
+; SI-NEXT: buffer_store_dword v2, off, s[0:3], s32 offset:472 ; 4-byte Folded Spill
+; SI-NEXT: s_waitcnt expcnt(0)
+; SI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:476 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v12, off, s[0:3], s32 offset:760 ; 4-byte Folded Reload
+; SI-NEXT: s_waitcnt vmcnt(1)
+; SI-NEXT: v_and_b32_e32 v2, 0xff, v2
+; SI-NEXT: v_or_b32_e32 v2, v2, v14
+; SI-NEXT: buffer_store_dword v2, off, s[0:3], s32 offset:476 ; 4-byte Folded Spill
+; SI-NEXT: s_waitcnt expcnt(0)
+; SI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:480 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v14, off, s[0:3], s32 offset:804 ; 4-byte Folded Reload
+; SI-NEXT: s_waitcnt vmcnt(1)
+; SI-NEXT: v_and_b32_e32 v2, 0xff, v2
+; SI-NEXT: v_or_b32_e32 v2, v2, v6
+; SI-NEXT: buffer_store_dword v2, off, s[0:3], s32 offset:480 ; 4-byte Folded Spill
+; SI-NEXT: s_waitcnt expcnt(0)
+; SI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:484 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v6, off, s[0:3], s32 offset:536 ; 4-byte Folded Reload
+; SI-NEXT: s_waitcnt vmcnt(1)
+; SI-NEXT: v_and_b32_e32 v2, 0xff, v2
+; SI-NEXT: s_waitcnt vmcnt(0)
+; SI-NEXT: v_or_b32_e32 v2, v2, v6
+; SI-NEXT: buffer_store_dword v2, off, s[0:3], s32 offset:484 ; 4-byte Folded Spill
+; SI-NEXT: s_waitcnt expcnt(0)
+; SI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:488 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v6, off, s[0:3], s32 offset:540 ; 4-byte Folded Reload
+; SI-NEXT: s_waitcnt vmcnt(1)
+; SI-NEXT: v_and_b32_e32 v2, 0xff, v2
+; SI-NEXT: s_waitcnt vmcnt(0)
+; SI-NEXT: v_or_b32_e32 v2, v2, v6
+; SI-NEXT: buffer_store_dword v2, off, s[0:3], s32 offset:488 ; 4-byte Folded Spill
+; SI-NEXT: s_waitcnt expcnt(0)
+; SI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:492 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v6, off, s[0:3], s32 offset:544 ; 4-byte Folded Reload
+; SI-NEXT: s_waitcnt vmcnt(1)
+; SI-NEXT: v_and_b32_e32 v2, 0xff, v2
+; SI-NEXT: s_waitcnt vmcnt(0)
+; SI-NEXT: v_or_b32_e32 v23, v2, v6
+; SI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:496 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v6, off, s[0:3], s32 offset:548 ; 4-byte Folded Reload
+; SI-NEXT: s_waitcnt vmcnt(1)
+; SI-NEXT: v_and_b32_e32 v2, 0xff, v2
+; SI-NEXT: s_waitcnt vmcnt(0)
+; SI-NEXT: v_or_b32_e32 v7, v2, v6
+; SI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:500 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v6, off, s[0:3], s32 offset:552 ; 4-byte Folded Reload
+; SI-NEXT: v_cvt_f32_f16_e32 v7, v7
+; SI-NEXT: buffer_store_dword v7, off, s[0:3], s32 offset:832 ; 4-byte Folded Spill
+; SI-NEXT: s_waitcnt vmcnt(2)
+; SI-NEXT: v_and_b32_e32 v2, 0xff, v2
+; SI-NEXT: s_waitcnt vmcnt(1)
+; SI-NEXT: v_or_b32_e32 v31, v2, v6
+; SI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:504 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v6, off, s[0:3], s32 offset:556 ; 4-byte Folded Reload
+; SI-NEXT: s_waitcnt expcnt(0)
+; SI-NEXT: v_cvt_f32_f16_e32 v7, v31
+; SI-NEXT: v_cvt_f32_f16_e32 v31, v55
+; SI-NEXT: ; implicit-def: $vgpr55
+; SI-NEXT: buffer_store_dword v7, off, s[0:3], s32 offset:792 ; 4-byte Folded Spill
+; SI-NEXT: s_waitcnt vmcnt(2)
+; SI-NEXT: v_and_b32_e32 v2, 0xff, v2
+; SI-NEXT: s_waitcnt vmcnt(1)
+; SI-NEXT: v_or_b32_e32 v49, v2, v6
+; SI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:508 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v6, off, s[0:3], s32 offset:560 ; 4-byte Folded Reload
+; SI-NEXT: s_waitcnt expcnt(0)
+; SI-NEXT: v_cvt_f32_f16_e32 v7, v49
+; SI-NEXT: ; implicit-def: $vgpr49
+; SI-NEXT: buffer_store_dword v7, off, s[0:3], s32 offset:816 ; 4-byte Folded Spill
+; SI-NEXT: s_waitcnt vmcnt(2)
+; SI-NEXT: v_and_b32_e32 v2, 0xff, v2
+; SI-NEXT: s_waitcnt vmcnt(1)
+; SI-NEXT: v_or_b32_e32 v25, v2, v6
+; SI-NEXT: v_and_b32_e32 v2, 0xff, v4
+; SI-NEXT: buffer_load_dword v4, off, s[0:3], s32 offset:564 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v6, off, s[0:3], s32 offset:568 ; 4-byte Folded Reload
+; SI-NEXT: s_waitcnt expcnt(0)
+; SI-NEXT: v_cvt_f32_f16_e32 v7, v25
+; SI-NEXT: v_cvt_f32_f16_e32 v25, v0
; SI-NEXT: ; implicit-def: $vgpr0
; SI-NEXT: ; kill: killed $vgpr0
; SI-NEXT: ; implicit-def: $vgpr0
@@ -179140,748 +179492,483 @@ define <64 x half> @bitcast_v128i8_to_v64f16(<128 x i8> %a, i32 %b) {
; SI-NEXT: ; implicit-def: $vgpr0
; SI-NEXT: ; kill: killed $vgpr0
; SI-NEXT: ; implicit-def: $vgpr0
+; SI-NEXT: buffer_store_dword v7, off, s[0:3], s32 offset:772 ; 4-byte Folded Spill
; SI-NEXT: ; kill: killed $vgpr0
; SI-NEXT: ; implicit-def: $vgpr0
-; SI-NEXT: ; kill: killed $vgpr0
-; SI-NEXT: ; implicit-def: $vgpr0
-; SI-NEXT: ; kill: killed $vgpr0
-; SI-NEXT: ; implicit-def: $vgpr0
-; SI-NEXT: ; kill: killed $vgpr0
-; SI-NEXT: ; implicit-def: $vgpr0
-; SI-NEXT: ; kill: killed $vgpr0
-; SI-NEXT: ; implicit-def: $vgpr0
-; SI-NEXT: ; kill: killed $vgpr0
-; SI-NEXT: ; implicit-def: $vgpr0
-; SI-NEXT: ; kill: killed $vgpr0
-; SI-NEXT: ; implicit-def: $vgpr0
-; SI-NEXT: ; kill: killed $vgpr0
-; SI-NEXT: ; implicit-def: $vgpr0
-; SI-NEXT: ; kill: killed $vgpr0
-; SI-NEXT: ; implicit-def: $vgpr0
-; SI-NEXT: ; kill: killed $vgpr0
-; SI-NEXT: ; implicit-def: $vgpr0
-; SI-NEXT: ; kill: killed $vgpr0
-; SI-NEXT: ; implicit-def: $vgpr0
-; SI-NEXT: ; kill: killed $vgpr0
-; SI-NEXT: ; implicit-def: $vgpr0
-; SI-NEXT: ; kill: killed $vgpr0
-; SI-NEXT: ; implicit-def: $vgpr0
-; SI-NEXT: ; kill: killed $vgpr0
-; SI-NEXT: ; implicit-def: $vgpr0
-; SI-NEXT: ; kill: killed $vgpr0
-; SI-NEXT: s_and_saveexec_b64 s[4:5], vcc
-; SI-NEXT: s_xor_b64 s[4:5], exec, s[4:5]
-; SI-NEXT: s_cbranch_execz .LBB92_2
-; SI-NEXT: ; %bb.1: ; %cmp.false
-; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:476 ; 4-byte Folded Reload
-; SI-NEXT: v_and_b32_e32 v2, 0xff, v51
-; SI-NEXT: s_waitcnt vmcnt(8)
-; SI-NEXT: v_and_b32_e32 v4, 0xff, v4
-; SI-NEXT: s_waitcnt vmcnt(1)
-; SI-NEXT: v_and_b32_e32 v18, 0xff, v18
-; SI-NEXT: v_and_b32_e32 v22, 0xff, v22
-; SI-NEXT: v_and_b32_e32 v24, 0xff, v24
-; SI-NEXT: v_and_b32_e32 v26, 0xff, v26
-; SI-NEXT: v_and_b32_e32 v28, 0xff, v28
-; SI-NEXT: v_and_b32_e32 v30, 0xff, v30
-; SI-NEXT: v_and_b32_e32 v34, 0xff, v34
-; SI-NEXT: v_and_b32_e32 v16, 0xff, v16
-; SI-NEXT: v_and_b32_e32 v20, 0xff, v20
-; SI-NEXT: v_and_b32_e32 v39, 0xff, v50
-; SI-NEXT: v_and_b32_e32 v48, 0xff, v40
-; SI-NEXT: v_and_b32_e32 v49, 0xff, v49
-; SI-NEXT: v_and_b32_e32 v52, 0xff, v52
-; SI-NEXT: v_and_b32_e32 v42, 0xff, v42
-; SI-NEXT: v_and_b32_e32 v46, 0xff, v46
-; SI-NEXT: v_or_b32_e32 v45, v46, v45
-; SI-NEXT: v_and_b32_e32 v56, 0xff, v56
-; SI-NEXT: v_or_b32_e32 v56, v56, v61
-; SI-NEXT: v_and_b32_e32 v57, 0xff, v57
-; SI-NEXT: v_and_b32_e32 v47, 0xff, v47
-; SI-NEXT: v_or_b32_e32 v1, v57, v1
-; SI-NEXT: v_or_b32_e32 v3, v47, v3
-; SI-NEXT: ; implicit-def: $vgpr57
-; SI-NEXT: ; implicit-def: $vgpr47
-; SI-NEXT: s_waitcnt vmcnt(0)
-; SI-NEXT: v_or_b32_e32 v0, v2, v0
-; SI-NEXT: buffer_store_dword v0, off, s[0:3], s32 offset:476 ; 4-byte Folded Spill
-; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:484 ; 4-byte Folded Reload
-; SI-NEXT: v_and_b32_e32 v2, 0xff, v54
-; SI-NEXT: s_waitcnt vmcnt(0)
-; SI-NEXT: v_or_b32_e32 v0, v2, v0
-; SI-NEXT: v_and_b32_e32 v2, 0xff, v55
-; SI-NEXT: buffer_store_dword v0, off, s[0:3], s32 offset:484 ; 4-byte Folded Spill
-; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_or_b32_e32 v0, v2, v6
-; SI-NEXT: v_and_b32_e32 v2, 0xff, v41
-; SI-NEXT: buffer_store_dword v0, off, s[0:3], s32 offset:552 ; 4-byte Folded Spill
-; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_or_b32_e32 v0, v2, v8
-; SI-NEXT: v_and_b32_e32 v2, 0xff, v9
-; SI-NEXT: buffer_store_dword v0, off, s[0:3], s32 offset:560 ; 4-byte Folded Spill
-; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_or_b32_e32 v0, v2, v12
-; SI-NEXT: v_and_b32_e32 v2, 0xff, v11
-; SI-NEXT: buffer_store_dword v0, off, s[0:3], s32 offset:564 ; 4-byte Folded Spill
-; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_or_b32_e32 v0, v2, v14
-; SI-NEXT: buffer_store_dword v0, off, s[0:3], s32 offset:568 ; 4-byte Folded Spill
-; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:492 ; 4-byte Folded Reload
-; SI-NEXT: v_and_b32_e32 v2, 0xff, v13
-; SI-NEXT: v_mov_b32_e32 v8, v7
-; SI-NEXT: v_mov_b32_e32 v7, v19
-; SI-NEXT: v_or_b32_e32 v19, v2, v32
-; SI-NEXT: v_and_b32_e32 v2, 0xff, v15
-; SI-NEXT: v_and_b32_e32 v35, 0xff, v7
-; SI-NEXT: v_and_b32_e32 v7, 0xff, v53
-; SI-NEXT: v_cvt_f32_f16_e32 v19, v19
-; SI-NEXT: buffer_load_dword v15, off, s[0:3], s32 offset:948 ; 4-byte Folded Reload
-; SI-NEXT: buffer_store_dword v19, off, s[0:3], s32 offset:804 ; 4-byte Folded Spill
-; SI-NEXT: ; implicit-def: $vgpr19
-; SI-NEXT: s_waitcnt vmcnt(2)
-; SI-NEXT: v_or_b32_e32 v17, v2, v0
-; SI-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:496 ; 4-byte Folded Reload
-; SI-NEXT: v_and_b32_e32 v2, 0xff, v43
-; SI-NEXT: v_cvt_f32_f16_e32 v17, v17
-; SI-NEXT: buffer_store_dword v17, off, s[0:3], s32 offset:828 ; 4-byte Folded Spill
-; SI-NEXT: s_waitcnt vmcnt(1)
-; SI-NEXT: v_or_b32_e32 v23, v2, v0
-; SI-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:500 ; 4-byte Folded Reload
-; SI-NEXT: v_and_b32_e32 v2, 0xff, v44
-; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_cvt_f32_f16_e32 v17, v23
-; SI-NEXT: v_cvt_f32_f16_e32 v23, v56
-; SI-NEXT: ; implicit-def: $vgpr56
-; SI-NEXT: buffer_store_dword v17, off, s[0:3], s32 offset:792 ; 4-byte Folded Spill
-; SI-NEXT: s_waitcnt vmcnt(1)
-; SI-NEXT: v_or_b32_e32 v31, v2, v0
-; SI-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:504 ; 4-byte Folded Reload
-; SI-NEXT: v_and_b32_e32 v2, 0xff, v21
-; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_cvt_f32_f16_e32 v17, v31
-; SI-NEXT: buffer_load_dword v21, off, s[0:3], s32 offset:976 ; 4-byte Folded Reload
-; SI-NEXT: buffer_store_dword v17, off, s[0:3], s32 offset:808 ; 4-byte Folded Spill
+; SI-NEXT: ; implicit-def: $vgpr7
; SI-NEXT: s_waitcnt vmcnt(2)
-; SI-NEXT: v_or_b32_e32 v51, v2, v0
-; SI-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:460 ; 4-byte Folded Reload
-; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_cvt_f32_f16_e32 v17, v51
-; SI-NEXT: ; implicit-def: $vgpr51
-; SI-NEXT: buffer_store_dword v17, off, s[0:3], s32 offset:776 ; 4-byte Folded Spill
-; SI-NEXT: s_waitcnt vmcnt(1)
-; SI-NEXT: v_and_b32_e32 v2, 0xff, v0
-; SI-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:508 ; 4-byte Folded Reload
-; SI-NEXT: s_waitcnt vmcnt(0)
-; SI-NEXT: v_or_b32_e32 v27, v2, v0
-; SI-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:464 ; 4-byte Folded Reload
-; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_cvt_f32_f16_e32 v17, v27
-; SI-NEXT: v_cvt_f32_f16_e32 v27, v3
-; SI-NEXT: ; implicit-def: $vgpr3
-; SI-NEXT: buffer_store_dword v17, off, s[0:3], s32 offset:800 ; 4-byte Folded Spill
+; SI-NEXT: v_or_b32_e32 v4, v2, v4
+; SI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:904 ; 4-byte Folded Reload
+; SI-NEXT: v_cvt_f32_f16_e32 v4, v4
+; SI-NEXT: buffer_store_dword v4, off, s[0:3], s32 offset:796 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt vmcnt(1)
-; SI-NEXT: v_and_b32_e32 v2, 0xff, v0
-; SI-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:512 ; 4-byte Folded Reload
-; SI-NEXT: s_waitcnt vmcnt(0)
-; SI-NEXT: v_or_b32_e32 v29, v2, v0
-; SI-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:468 ; 4-byte Folded Reload
+; SI-NEXT: v_and_b32_e32 v2, 0xff, v2
+; SI-NEXT: v_or_b32_e32 v29, v2, v6
+; SI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:908 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v6, off, s[0:3], s32 offset:912 ; 4-byte Folded Reload
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_cvt_f32_f16_e32 v17, v29
-; SI-NEXT: buffer_store_dword v17, off, s[0:3], s32 offset:760 ; 4-byte Folded Spill
+; SI-NEXT: v_cvt_f32_f16_e32 v4, v29
+; SI-NEXT: buffer_store_dword v4, off, s[0:3], s32 offset:764 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt vmcnt(1)
-; SI-NEXT: v_and_b32_e32 v2, 0xff, v0
-; SI-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:516 ; 4-byte Folded Reload
-; SI-NEXT: s_waitcnt vmcnt(0)
-; SI-NEXT: v_or_b32_e32 v2, v2, v0
-; SI-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:472 ; 4-byte Folded Reload
+; SI-NEXT: v_and_b32_e32 v6, 0xff, v6
+; SI-NEXT: v_or_b32_e32 v32, v6, v8
+; SI-NEXT: buffer_load_dword v6, off, s[0:3], s32 offset:916 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v8, off, s[0:3], s32 offset:576 ; 4-byte Folded Reload
+; SI-NEXT: v_and_b32_e32 v2, 0xff, v2
+; SI-NEXT: v_or_b32_e32 v2, v2, v11
; SI-NEXT: v_cvt_f32_f16_e32 v2, v2
-; SI-NEXT: buffer_store_dword v2, off, s[0:3], s32 offset:784 ; 4-byte Folded Spill
-; SI-NEXT: s_waitcnt vmcnt(1)
-; SI-NEXT: v_and_b32_e32 v5, 0xff, v0
-; SI-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:520 ; 4-byte Folded Reload
-; SI-NEXT: s_waitcnt vmcnt(0)
-; SI-NEXT: v_or_b32_e32 v5, v5, v0
-; SI-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:524 ; 4-byte Folded Reload
+; SI-NEXT: buffer_store_dword v2, off, s[0:3], s32 offset:788 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_cvt_f32_f16_e32 v2, v5
-; SI-NEXT: v_cvt_f32_f16_e32 v5, v1
-; SI-NEXT: ; implicit-def: $vgpr1
-; SI-NEXT: buffer_store_dword v2, off, s[0:3], s32 offset:740 ; 4-byte Folded Spill
-; SI-NEXT: s_waitcnt vmcnt(1)
-; SI-NEXT: v_or_b32_e32 v4, v4, v0
-; SI-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:872 ; 4-byte Folded Reload
+; SI-NEXT: v_cvt_f32_f16_e32 v2, v32
+; SI-NEXT: buffer_store_dword v2, off, s[0:3], s32 offset:752 ; 4-byte Folded Spill
+; SI-NEXT: s_waitcnt vmcnt(3)
+; SI-NEXT: v_and_b32_e32 v6, 0xff, v6
+; SI-NEXT: s_waitcnt vmcnt(2)
+; SI-NEXT: v_or_b32_e32 v11, v6, v8
+; SI-NEXT: buffer_load_dword v6, off, s[0:3], s32 offset:920 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v8, off, s[0:3], s32 offset:580 ; 4-byte Folded Reload
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_cvt_f32_f16_e32 v2, v4
-; SI-NEXT: ; implicit-def: $vgpr4
-; SI-NEXT: buffer_store_dword v2, off, s[0:3], s32 offset:764 ; 4-byte Folded Spill
+; SI-NEXT: v_cvt_f32_f16_e32 v2, v11
+; SI-NEXT: v_cvt_f32_f16_e32 v11, v27
+; SI-NEXT: ; implicit-def: $vgpr27
+; SI-NEXT: buffer_store_dword v2, off, s[0:3], s32 offset:780 ; 4-byte Folded Spill
+; SI-NEXT: s_waitcnt vmcnt(2)
+; SI-NEXT: v_and_b32_e32 v6, 0xff, v6
; SI-NEXT: s_waitcnt vmcnt(1)
-; SI-NEXT: v_and_b32_e32 v6, 0xff, v0
-; SI-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:528 ; 4-byte Folded Reload
-; SI-NEXT: s_waitcnt vmcnt(0)
-; SI-NEXT: v_or_b32_e32 v33, v6, v0
-; SI-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:876 ; 4-byte Folded Reload
+; SI-NEXT: v_or_b32_e32 v33, v6, v8
+; SI-NEXT: buffer_load_dword v6, off, s[0:3], s32 offset:924 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v8, off, s[0:3], s32 offset:584 ; 4-byte Folded Reload
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_cvt_f32_f16_e32 v2, v33
-; SI-NEXT: v_cvt_f32_f16_e32 v33, v45
-; SI-NEXT: ; implicit-def: $vgpr45
-; SI-NEXT: buffer_store_dword v2, off, s[0:3], s32 offset:732 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v2, off, s[0:3], s32 offset:740 ; 4-byte Folded Spill
+; SI-NEXT: s_waitcnt vmcnt(2)
+; SI-NEXT: v_and_b32_e32 v6, 0xff, v6
; SI-NEXT: s_waitcnt vmcnt(1)
-; SI-NEXT: v_and_b32_e32 v6, 0xff, v0
-; SI-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:880 ; 4-byte Folded Reload
; SI-NEXT: v_or_b32_e32 v6, v6, v8
+; SI-NEXT: buffer_load_dword v8, off, s[0:3], s32 offset:588 ; 4-byte Folded Reload
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_cvt_f32_f16_e32 v2, v6
; SI-NEXT: ; implicit-def: $vgpr6
-; SI-NEXT: buffer_store_dword v2, off, s[0:3], s32 offset:756 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v2, off, s[0:3], s32 offset:768 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt vmcnt(1)
-; SI-NEXT: v_and_b32_e32 v8, 0xff, v0
-; SI-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:532 ; 4-byte Folded Reload
-; SI-NEXT: s_waitcnt vmcnt(0)
-; SI-NEXT: v_or_b32_e32 v8, v8, v0
-; SI-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:884 ; 4-byte Folded Reload
-; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_cvt_f32_f16_e32 v2, v8
-; SI-NEXT: ; implicit-def: $vgpr8
-; SI-NEXT: buffer_store_dword v2, off, s[0:3], s32 offset:720 ; 4-byte Folded Spill
-; SI-NEXT: s_waitcnt vmcnt(1)
-; SI-NEXT: v_and_b32_e32 v12, 0xff, v0
-; SI-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:536 ; 4-byte Folded Reload
-; SI-NEXT: s_waitcnt vmcnt(0)
-; SI-NEXT: v_or_b32_e32 v12, v12, v0
-; SI-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:888 ; 4-byte Folded Reload
-; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_cvt_f32_f16_e32 v2, v12
-; SI-NEXT: ; implicit-def: $vgpr12
-; SI-NEXT: buffer_store_dword v2, off, s[0:3], s32 offset:748 ; 4-byte Folded Spill
-; SI-NEXT: s_waitcnt vmcnt(1)
-; SI-NEXT: v_and_b32_e32 v14, 0xff, v0
-; SI-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:540 ; 4-byte Folded Reload
-; SI-NEXT: s_waitcnt vmcnt(0)
-; SI-NEXT: v_or_b32_e32 v14, v14, v0
-; SI-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:892 ; 4-byte Folded Reload
-; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_cvt_f32_f16_e32 v2, v14
-; SI-NEXT: ; implicit-def: $vgpr14
-; SI-NEXT: buffer_store_dword v2, off, s[0:3], s32 offset:708 ; 4-byte Folded Spill
-; SI-NEXT: s_waitcnt vmcnt(1)
-; SI-NEXT: v_and_b32_e32 v32, 0xff, v0
-; SI-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:544 ; 4-byte Folded Reload
-; SI-NEXT: s_waitcnt vmcnt(0)
-; SI-NEXT: v_or_b32_e32 v32, v32, v0
-; SI-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:548 ; 4-byte Folded Reload
-; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_cvt_f32_f16_e32 v2, v32
-; SI-NEXT: ; implicit-def: $vgpr32
-; SI-NEXT: buffer_store_dword v2, off, s[0:3], s32 offset:736 ; 4-byte Folded Spill
-; SI-NEXT: s_waitcnt vmcnt(1)
-; SI-NEXT: v_or_b32_e32 v18, v18, v0
-; SI-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:556 ; 4-byte Folded Reload
+; SI-NEXT: v_or_b32_e32 v18, v18, v8
+; SI-NEXT: buffer_load_dword v8, off, s[0:3], s32 offset:592 ; 4-byte Folded Reload
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_cvt_f32_f16_e32 v2, v18
; SI-NEXT: ; implicit-def: $vgpr18
-; SI-NEXT: buffer_store_dword v2, off, s[0:3], s32 offset:700 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v2, off, s[0:3], s32 offset:732 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt vmcnt(1)
-; SI-NEXT: v_or_b32_e32 v22, v22, v0
-; SI-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:572 ; 4-byte Folded Reload
+; SI-NEXT: v_or_b32_e32 v22, v22, v8
+; SI-NEXT: buffer_load_dword v8, off, s[0:3], s32 offset:604 ; 4-byte Folded Reload
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_cvt_f32_f16_e32 v2, v22
; SI-NEXT: ; implicit-def: $vgpr22
-; SI-NEXT: buffer_store_dword v2, off, s[0:3], s32 offset:724 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v2, off, s[0:3], s32 offset:756 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt vmcnt(1)
-; SI-NEXT: v_or_b32_e32 v24, v24, v0
-; SI-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:592 ; 4-byte Folded Reload
+; SI-NEXT: v_or_b32_e32 v24, v24, v8
+; SI-NEXT: buffer_load_dword v8, off, s[0:3], s32 offset:624 ; 4-byte Folded Reload
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_cvt_f32_f16_e32 v2, v24
; SI-NEXT: ; implicit-def: $vgpr24
-; SI-NEXT: buffer_store_dword v2, off, s[0:3], s32 offset:688 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v2, off, s[0:3], s32 offset:720 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt vmcnt(1)
-; SI-NEXT: v_or_b32_e32 v26, v26, v0
-; SI-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:612 ; 4-byte Folded Reload
+; SI-NEXT: v_or_b32_e32 v26, v26, v8
+; SI-NEXT: buffer_load_dword v8, off, s[0:3], s32 offset:644 ; 4-byte Folded Reload
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_cvt_f32_f16_e32 v2, v26
; SI-NEXT: ; implicit-def: $vgpr26
-; SI-NEXT: buffer_store_dword v2, off, s[0:3], s32 offset:716 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v2, off, s[0:3], s32 offset:748 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt vmcnt(1)
-; SI-NEXT: v_or_b32_e32 v28, v28, v0
-; SI-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:632 ; 4-byte Folded Reload
+; SI-NEXT: v_or_b32_e32 v28, v28, v8
+; SI-NEXT: buffer_load_dword v8, off, s[0:3], s32 offset:664 ; 4-byte Folded Reload
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_cvt_f32_f16_e32 v2, v28
; SI-NEXT: ; implicit-def: $vgpr28
-; SI-NEXT: buffer_store_dword v2, off, s[0:3], s32 offset:676 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v2, off, s[0:3], s32 offset:708 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt vmcnt(1)
-; SI-NEXT: v_or_b32_e32 v30, v30, v0
-; SI-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:648 ; 4-byte Folded Reload
+; SI-NEXT: v_or_b32_e32 v30, v30, v8
+; SI-NEXT: buffer_load_dword v8, off, s[0:3], s32 offset:680 ; 4-byte Folded Reload
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_cvt_f32_f16_e32 v2, v30
; SI-NEXT: ; implicit-def: $vgpr30
-; SI-NEXT: buffer_store_dword v2, off, s[0:3], s32 offset:704 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v2, off, s[0:3], s32 offset:736 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt vmcnt(1)
-; SI-NEXT: v_or_b32_e32 v34, v34, v0
-; SI-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:664 ; 4-byte Folded Reload
+; SI-NEXT: v_or_b32_e32 v34, v34, v8
+; SI-NEXT: buffer_load_dword v8, off, s[0:3], s32 offset:696 ; 4-byte Folded Reload
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_cvt_f32_f16_e32 v2, v34
; SI-NEXT: ; implicit-def: $vgpr34
-; SI-NEXT: buffer_store_dword v2, off, s[0:3], s32 offset:668 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v2, off, s[0:3], s32 offset:700 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt vmcnt(1)
-; SI-NEXT: v_or_b32_e32 v16, v16, v0
-; SI-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:680 ; 4-byte Folded Reload
+; SI-NEXT: v_or_b32_e32 v16, v16, v8
+; SI-NEXT: buffer_load_dword v8, off, s[0:3], s32 offset:712 ; 4-byte Folded Reload
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_cvt_f32_f16_e32 v2, v16
; SI-NEXT: ; implicit-def: $vgpr16
-; SI-NEXT: buffer_store_dword v2, off, s[0:3], s32 offset:692 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v2, off, s[0:3], s32 offset:724 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt vmcnt(1)
-; SI-NEXT: v_or_b32_e32 v20, v20, v0
-; SI-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:696 ; 4-byte Folded Reload
+; SI-NEXT: v_or_b32_e32 v20, v20, v8
+; SI-NEXT: buffer_load_dword v8, off, s[0:3], s32 offset:512 ; 4-byte Folded Reload
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_cvt_f32_f16_e32 v2, v20
; SI-NEXT: ; implicit-def: $vgpr20
-; SI-NEXT: buffer_store_dword v2, off, s[0:3], s32 offset:656 ; 4-byte Folded Spill
-; SI-NEXT: s_waitcnt vmcnt(1)
-; SI-NEXT: v_or_b32_e32 v35, v35, v0
-; SI-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:480 ; 4-byte Folded Reload
-; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_cvt_f32_f16_e32 v2, v35
-; SI-NEXT: buffer_store_dword v2, off, s[0:3], s32 offset:684 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v2, off, s[0:3], s32 offset:688 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt vmcnt(1)
-; SI-NEXT: v_and_b32_e32 v36, 0xff, v0
-; SI-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:712 ; 4-byte Folded Reload
+; SI-NEXT: v_and_b32_e32 v13, 0xff, v8
+; SI-NEXT: buffer_load_dword v8, off, s[0:3], s32 offset:728 ; 4-byte Folded Reload
; SI-NEXT: s_waitcnt vmcnt(0)
-; SI-NEXT: v_or_b32_e32 v36, v36, v0
-; SI-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:488 ; 4-byte Folded Reload
+; SI-NEXT: v_or_b32_e32 v13, v13, v8
+; SI-NEXT: buffer_load_dword v8, off, s[0:3], s32 offset:520 ; 4-byte Folded Reload
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_cvt_f32_f16_e32 v2, v36
-; SI-NEXT: buffer_store_dword v2, off, s[0:3], s32 offset:644 ; 4-byte Folded Spill
+; SI-NEXT: v_cvt_f32_f16_e32 v2, v13
+; SI-NEXT: v_cvt_f32_f16_e32 v13, v56
+; SI-NEXT: ; implicit-def: $vgpr56
+; SI-NEXT: buffer_store_dword v2, off, s[0:3], s32 offset:716 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt vmcnt(1)
-; SI-NEXT: v_and_b32_e32 v37, 0xff, v0
-; SI-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:728 ; 4-byte Folded Reload
-; SI-NEXT: s_waitcnt vmcnt(0)
-; SI-NEXT: v_or_b32_e32 v37, v37, v0
-; SI-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:928 ; 4-byte Folded Reload
+; SI-NEXT: v_and_b32_e32 v8, 0xff, v8
+; SI-NEXT: v_or_b32_e32 v8, v8, v10
+; SI-NEXT: buffer_load_dword v10, off, s[0:3], s32 offset:528 ; 4-byte Folded Reload
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_cvt_f32_f16_e32 v2, v37
-; SI-NEXT: buffer_store_dword v2, off, s[0:3], s32 offset:672 ; 4-byte Folded Spill
+; SI-NEXT: v_cvt_f32_f16_e32 v2, v8
+; SI-NEXT: ; implicit-def: $vgpr8
+; SI-NEXT: buffer_store_dword v2, off, s[0:3], s32 offset:676 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt vmcnt(1)
-; SI-NEXT: v_and_b32_e32 v38, 0xff, v0
-; SI-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:744 ; 4-byte Folded Reload
-; SI-NEXT: s_waitcnt vmcnt(0)
-; SI-NEXT: v_or_b32_e32 v38, v38, v0
-; SI-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:772 ; 4-byte Folded Reload
+; SI-NEXT: v_and_b32_e32 v10, 0xff, v10
+; SI-NEXT: v_or_b32_e32 v10, v10, v12
+; SI-NEXT: buffer_load_dword v12, off, s[0:3], s32 offset:776 ; 4-byte Folded Reload
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_cvt_f32_f16_e32 v2, v38
-; SI-NEXT: buffer_store_dword v2, off, s[0:3], s32 offset:636 ; 4-byte Folded Spill
+; SI-NEXT: v_cvt_f32_f16_e32 v2, v10
+; SI-NEXT: ; implicit-def: $vgpr10
+; SI-NEXT: buffer_store_dword v2, off, s[0:3], s32 offset:704 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt vmcnt(1)
-; SI-NEXT: v_or_b32_e32 v39, v39, v0
-; SI-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:788 ; 4-byte Folded Reload
+; SI-NEXT: v_or_b32_e32 v36, v36, v12
+; SI-NEXT: v_and_b32_e32 v12, 0xff, v15
+; SI-NEXT: v_or_b32_e32 v12, v12, v14
+; SI-NEXT: buffer_load_dword v14, off, s[0:3], s32 offset:820 ; 4-byte Folded Reload
+; SI-NEXT: v_and_b32_e32 v15, 0xff, v17
+; SI-NEXT: buffer_load_dword v17, off, s[0:3], s32 offset:828 ; 4-byte Folded Reload
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_cvt_f32_f16_e32 v2, v39
-; SI-NEXT: buffer_store_dword v2, off, s[0:3], s32 offset:660 ; 4-byte Folded Spill
-; SI-NEXT: s_waitcnt vmcnt(1)
-; SI-NEXT: v_or_b32_e32 v7, v7, v0
-; SI-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:796 ; 4-byte Folded Reload
+; SI-NEXT: v_cvt_f32_f16_e32 v2, v36
+; SI-NEXT: ; implicit-def: $vgpr36
+; SI-NEXT: buffer_store_dword v2, off, s[0:3], s32 offset:668 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_cvt_f32_f16_e32 v2, v7
-; SI-NEXT: ; implicit-def: $vgpr7
-; SI-NEXT: buffer_store_dword v2, off, s[0:3], s32 offset:620 ; 4-byte Folded Spill
-; SI-NEXT: s_waitcnt vmcnt(1)
-; SI-NEXT: v_or_b32_e32 v48, v48, v0
-; SI-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:812 ; 4-byte Folded Reload
+; SI-NEXT: v_cvt_f32_f16_e32 v2, v12
+; SI-NEXT: ; implicit-def: $vgpr12
+; SI-NEXT: buffer_store_dword v2, off, s[0:3], s32 offset:692 ; 4-byte Folded Spill
+; SI-NEXT: s_waitcnt vmcnt(3)
+; SI-NEXT: v_or_b32_e32 v15, v15, v14
+; SI-NEXT: v_and_b32_e32 v14, 0xff, v19
+; SI-NEXT: s_waitcnt vmcnt(2)
+; SI-NEXT: v_or_b32_e32 v14, v14, v17
+; SI-NEXT: buffer_load_dword v17, off, s[0:3], s32 offset:844 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v19, off, s[0:3], s32 offset:884 ; 4-byte Folded Reload
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_cvt_f32_f16_e32 v2, v48
+; SI-NEXT: v_cvt_f32_f16_e32 v2, v15
+; SI-NEXT: ; implicit-def: $vgpr15
; SI-NEXT: buffer_store_dword v2, off, s[0:3], s32 offset:652 ; 4-byte Folded Spill
-; SI-NEXT: s_waitcnt vmcnt(1)
-; SI-NEXT: v_or_b32_e32 v49, v49, v0
-; SI-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:584 ; 4-byte Folded Reload
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_cvt_f32_f16_e32 v2, v49
-; SI-NEXT: ; implicit-def: $vgpr49
-; SI-NEXT: buffer_store_dword v2, off, s[0:3], s32 offset:608 ; 4-byte Folded Spill
-; SI-NEXT: s_waitcnt vmcnt(1)
-; SI-NEXT: v_and_b32_e32 v50, 0xff, v0
-; SI-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:852 ; 4-byte Folded Reload
-; SI-NEXT: s_waitcnt vmcnt(0)
-; SI-NEXT: v_or_b32_e32 v50, v50, v0
-; SI-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:604 ; 4-byte Folded Reload
+; SI-NEXT: v_cvt_f32_f16_e32 v2, v14
+; SI-NEXT: ; implicit-def: $vgpr14
+; SI-NEXT: buffer_store_dword v2, off, s[0:3], s32 offset:684 ; 4-byte Folded Spill
+; SI-NEXT: s_waitcnt vmcnt(3)
+; SI-NEXT: v_or_b32_e32 v50, v50, v17
+; SI-NEXT: buffer_load_dword v17, off, s[0:3], s32 offset:616 ; 4-byte Folded Reload
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_cvt_f32_f16_e32 v2, v50
; SI-NEXT: ; implicit-def: $vgpr50
; SI-NEXT: buffer_store_dword v2, off, s[0:3], s32 offset:640 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt vmcnt(1)
-; SI-NEXT: v_and_b32_e32 v9, 0xff, v0
-; SI-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:860 ; 4-byte Folded Reload
-; SI-NEXT: s_waitcnt vmcnt(0)
-; SI-NEXT: v_or_b32_e32 v9, v9, v0
-; SI-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:624 ; 4-byte Folded Reload
+; SI-NEXT: v_and_b32_e32 v17, 0xff, v17
+; SI-NEXT: v_or_b32_e32 v17, v17, v19
+; SI-NEXT: buffer_load_dword v19, off, s[0:3], s32 offset:636 ; 4-byte Folded Reload
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_cvt_f32_f16_e32 v2, v9
-; SI-NEXT: ; implicit-def: $vgpr9
-; SI-NEXT: buffer_store_dword v2, off, s[0:3], s32 offset:596 ; 4-byte Folded Spill
+; SI-NEXT: v_cvt_f32_f16_e32 v2, v17
+; SI-NEXT: ; implicit-def: $vgpr17
+; SI-NEXT: buffer_store_dword v2, off, s[0:3], s32 offset:672 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt vmcnt(1)
-; SI-NEXT: v_and_b32_e32 v54, 0xff, v0
-; SI-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:864 ; 4-byte Folded Reload
+; SI-NEXT: v_and_b32_e32 v19, 0xff, v19
+; SI-NEXT: v_or_b32_e32 v19, v19, v21
+; SI-NEXT: buffer_load_dword v21, off, s[0:3], s32 offset:656 ; 4-byte Folded Reload
+; SI-NEXT: s_waitcnt expcnt(0)
+; SI-NEXT: v_cvt_f32_f16_e32 v2, v19
+; SI-NEXT: ; implicit-def: $vgpr19
+; SI-NEXT: buffer_store_dword v2, off, s[0:3], s32 offset:628 ; 4-byte Folded Spill
+; SI-NEXT: s_waitcnt vmcnt(1)
+; SI-NEXT: v_and_b32_e32 v35, 0xff, v21
+; SI-NEXT: buffer_load_dword v21, off, s[0:3], s32 offset:896 ; 4-byte Folded Reload
; SI-NEXT: s_waitcnt vmcnt(0)
-; SI-NEXT: v_or_b32_e32 v54, v54, v0
-; SI-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:868 ; 4-byte Folded Reload
+; SI-NEXT: v_or_b32_e32 v35, v35, v21
+; SI-NEXT: buffer_load_dword v21, off, s[0:3], s32 offset:900 ; 4-byte Folded Reload
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_cvt_f32_f16_e32 v2, v54
+; SI-NEXT: v_cvt_f32_f16_e32 v2, v35
+; SI-NEXT: v_cvt_f32_f16_e32 v35, v54
; SI-NEXT: ; implicit-def: $vgpr54
-; SI-NEXT: buffer_store_dword v2, off, s[0:3], s32 offset:628 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v2, off, s[0:3], s32 offset:660 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt vmcnt(1)
-; SI-NEXT: v_or_b32_e32 v52, v52, v0
-; SI-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:752 ; 4-byte Folded Reload
+; SI-NEXT: v_or_b32_e32 v52, v52, v21
+; SI-NEXT: buffer_load_dword v21, off, s[0:3], s32 offset:784 ; 4-byte Folded Reload
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_cvt_f32_f16_e32 v2, v52
; SI-NEXT: ; implicit-def: $vgpr52
-; SI-NEXT: buffer_store_dword v2, off, s[0:3], s32 offset:580 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v2, off, s[0:3], s32 offset:612 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt vmcnt(1)
-; SI-NEXT: v_and_b32_e32 v11, 0xff, v0
-; SI-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:904 ; 4-byte Folded Reload
+; SI-NEXT: v_and_b32_e32 v37, 0xff, v21
+; SI-NEXT: buffer_load_dword v21, off, s[0:3], s32 offset:936 ; 4-byte Folded Reload
; SI-NEXT: s_waitcnt vmcnt(0)
-; SI-NEXT: v_or_b32_e32 v11, v11, v0
-; SI-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:768 ; 4-byte Folded Reload
+; SI-NEXT: v_or_b32_e32 v37, v37, v21
+; SI-NEXT: buffer_load_dword v21, off, s[0:3], s32 offset:800 ; 4-byte Folded Reload
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_cvt_f32_f16_e32 v2, v11
-; SI-NEXT: ; implicit-def: $vgpr11
-; SI-NEXT: buffer_store_dword v2, off, s[0:3], s32 offset:616 ; 4-byte Folded Spill
+; SI-NEXT: v_cvt_f32_f16_e32 v2, v37
+; SI-NEXT: buffer_store_dword v2, off, s[0:3], s32 offset:648 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt vmcnt(1)
-; SI-NEXT: v_and_b32_e32 v53, 0xff, v0
-; SI-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:912 ; 4-byte Folded Reload
+; SI-NEXT: v_and_b32_e32 v38, 0xff, v21
+; SI-NEXT: buffer_load_dword v21, off, s[0:3], s32 offset:944 ; 4-byte Folded Reload
; SI-NEXT: s_waitcnt vmcnt(0)
-; SI-NEXT: v_or_b32_e32 v53, v53, v0
-; SI-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:780 ; 4-byte Folded Reload
+; SI-NEXT: v_or_b32_e32 v38, v38, v21
+; SI-NEXT: buffer_load_dword v21, off, s[0:3], s32 offset:812 ; 4-byte Folded Reload
+; SI-NEXT: s_waitcnt expcnt(0)
+; SI-NEXT: v_cvt_f32_f16_e32 v2, v38
+; SI-NEXT: buffer_store_dword v2, off, s[0:3], s32 offset:600 ; 4-byte Folded Spill
+; SI-NEXT: s_waitcnt vmcnt(1)
+; SI-NEXT: v_and_b32_e32 v39, 0xff, v21
+; SI-NEXT: buffer_load_dword v21, off, s[0:3], s32 offset:948 ; 4-byte Folded Reload
+; SI-NEXT: s_waitcnt vmcnt(0)
+; SI-NEXT: v_or_b32_e32 v39, v39, v21
+; SI-NEXT: buffer_load_dword v21, off, s[0:3], s32 offset:952 ; 4-byte Folded Reload
+; SI-NEXT: s_waitcnt expcnt(0)
+; SI-NEXT: v_cvt_f32_f16_e32 v2, v39
+; SI-NEXT: buffer_store_dword v2, off, s[0:3], s32 offset:632 ; 4-byte Folded Spill
+; SI-NEXT: s_waitcnt vmcnt(1)
+; SI-NEXT: v_or_b32_e32 v53, v53, v21
+; SI-NEXT: buffer_load_dword v21, off, s[0:3], s32 offset:876 ; 4-byte Folded Reload
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_cvt_f32_f16_e32 v2, v53
; SI-NEXT: ; implicit-def: $vgpr53
; SI-NEXT: s_waitcnt vmcnt(0)
-; SI-NEXT: v_and_b32_e32 v55, 0xff, v0
-; SI-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:916 ; 4-byte Folded Reload
-; SI-NEXT: s_waitcnt vmcnt(0)
-; SI-NEXT: v_or_b32_e32 v55, v55, v0
-; SI-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:956 ; 4-byte Folded Reload
-; SI-NEXT: s_waitcnt vmcnt(0)
-; SI-NEXT: v_and_b32_e32 v13, 0xff, v0
-; SI-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:920 ; 4-byte Folded Reload
-; SI-NEXT: s_waitcnt vmcnt(0)
-; SI-NEXT: v_or_b32_e32 v13, v13, v0
-; SI-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:844 ; 4-byte Folded Reload
-; SI-NEXT: s_waitcnt vmcnt(0)
-; SI-NEXT: v_and_b32_e32 v41, 0xff, v0
-; SI-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:936 ; 4-byte Folded Reload
-; SI-NEXT: s_waitcnt vmcnt(0)
-; SI-NEXT: v_or_b32_e32 v41, v41, v0
-; SI-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:848 ; 4-byte Folded Reload
-; SI-NEXT: s_waitcnt vmcnt(0)
-; SI-NEXT: v_and_b32_e32 v40, 0xff, v0
-; SI-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:944 ; 4-byte Folded Reload
-; SI-NEXT: s_waitcnt vmcnt(0)
-; SI-NEXT: v_or_b32_e32 v40, v40, v0
-; SI-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:856 ; 4-byte Folded Reload
-; SI-NEXT: v_cvt_f32_f16_e32 v48, v40
-; SI-NEXT: ; implicit-def: $vgpr40
+; SI-NEXT: v_and_b32_e32 v21, 0xff, v21
+; SI-NEXT: v_or_b32_e32 v21, v21, v48
+; SI-NEXT: buffer_load_dword v48, off, s[0:3], s32 offset:880 ; 4-byte Folded Reload
+; SI-NEXT: v_cvt_f32_f16_e32 v4, v21
+; SI-NEXT: v_cvt_f32_f16_e32 v21, v3
+; SI-NEXT: ; implicit-def: $vgpr3
+; SI-NEXT: buffer_store_dword v4, off, s[0:3], s32 offset:620 ; 4-byte Folded Spill
+; SI-NEXT: ; implicit-def: $vgpr4
+; SI-NEXT: s_waitcnt vmcnt(1)
+; SI-NEXT: v_and_b32_e32 v48, 0xff, v48
+; SI-NEXT: v_or_b32_e32 v48, v48, v51
+; SI-NEXT: buffer_load_dword v51, off, s[0:3], s32 offset:888 ; 4-byte Folded Reload
+; SI-NEXT: v_cvt_f32_f16_e32 v38, v48
+; SI-NEXT: v_mov_b32_e32 v48, v2
; SI-NEXT: s_waitcnt vmcnt(0)
-; SI-NEXT: v_and_b32_e32 v0, 0xff, v0
-; SI-NEXT: v_or_b32_e32 v0, v0, v15
-; SI-NEXT: buffer_load_dword v15, off, s[0:3], s32 offset:952 ; 4-byte Folded Reload
-; SI-NEXT: v_cvt_f32_f16_e32 v0, v0
-; SI-NEXT: buffer_store_dword v0, off, s[0:3], s32 offset:576 ; 4-byte Folded Spill
+; SI-NEXT: v_and_b32_e32 v51, 0xff, v51
+; SI-NEXT: v_or_b32_e32 v51, v51, v9
+; SI-NEXT: buffer_load_dword v9, off, s[0:3], s32 offset:516 ; 4-byte Folded Reload
+; SI-NEXT: v_cvt_f32_f16_e32 v2, v51
+; SI-NEXT: ; implicit-def: $vgpr51
+; SI-NEXT: buffer_store_dword v2, off, s[0:3], s32 offset:608 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt vmcnt(1)
-; SI-NEXT: v_or_b32_e32 v42, v42, v15
-; SI-NEXT: buffer_load_dword v15, off, s[0:3], s32 offset:896 ; 4-byte Folded Reload
-; SI-NEXT: v_cvt_f32_f16_e32 v38, v42
+; SI-NEXT: v_cvt_f32_f16_e32 v41, v9
+; SI-NEXT: buffer_load_dword v9, off, s[0:3], s32 offset:524 ; 4-byte Folded Reload
+; SI-NEXT: buffer_store_dword v41, off, s[0:3], s32 offset:856 ; 4-byte Folded Spill
+; SI-NEXT: s_waitcnt vmcnt(1) expcnt(0)
+; SI-NEXT: v_cvt_f32_f16_e32 v41, v9
+; SI-NEXT: buffer_load_dword v9, off, s[0:3], s32 offset:460 ; 4-byte Folded Reload
+; SI-NEXT: v_and_b32_e32 v42, 0xff, v42
+; SI-NEXT: v_or_b32_e32 v40, v42, v40
+; SI-NEXT: buffer_load_dword v42, off, s[0:3], s32 offset:932 ; 4-byte Folded Reload
+; SI-NEXT: buffer_store_dword v41, off, s[0:3], s32 offset:872 ; 4-byte Folded Spill
+; SI-NEXT: v_cvt_f32_f16_e32 v2, v40
+; SI-NEXT: ; implicit-def: $vgpr40
+; SI-NEXT: buffer_store_dword v2, off, s[0:3], s32 offset:596 ; 4-byte Folded Spill
+; SI-NEXT: s_waitcnt vmcnt(3) expcnt(1)
+; SI-NEXT: v_cvt_f32_f16_e32 v41, v9
+; SI-NEXT: buffer_load_dword v9, off, s[0:3], s32 offset:464 ; 4-byte Folded Reload
+; SI-NEXT: s_waitcnt vmcnt(3)
+; SI-NEXT: v_and_b32_e32 v42, 0xff, v42
+; SI-NEXT: v_or_b32_e32 v42, v42, v43
+; SI-NEXT: buffer_load_dword v43, off, s[0:3], s32 offset:940 ; 4-byte Folded Reload
+; SI-NEXT: buffer_store_dword v41, off, s[0:3], s32 offset:852 ; 4-byte Folded Spill
+; SI-NEXT: v_cvt_f32_f16_e32 v32, v42
; SI-NEXT: ; implicit-def: $vgpr42
-; SI-NEXT: s_waitcnt vmcnt(0)
-; SI-NEXT: v_and_b32_e32 v43, 0xff, v15
-; SI-NEXT: buffer_load_dword v15, off, s[0:3], s32 offset:968 ; 4-byte Folded Reload
-; SI-NEXT: s_waitcnt vmcnt(0)
-; SI-NEXT: v_or_b32_e32 v43, v43, v15
-; SI-NEXT: buffer_load_dword v15, off, s[0:3], s32 offset:900 ; 4-byte Folded Reload
-; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_cvt_f32_f16_e32 v0, v43
+; SI-NEXT: s_waitcnt vmcnt(2) expcnt(0)
+; SI-NEXT: v_cvt_f32_f16_e32 v41, v9
+; SI-NEXT: buffer_load_dword v9, off, s[0:3], s32 offset:468 ; 4-byte Folded Reload
+; SI-NEXT: s_waitcnt vmcnt(2)
+; SI-NEXT: v_and_b32_e32 v43, 0xff, v43
+; SI-NEXT: v_or_b32_e32 v43, v43, v44
+; SI-NEXT: v_and_b32_e32 v44, 0xff, v46
+; SI-NEXT: v_or_b32_e32 v44, v44, v45
+; SI-NEXT: buffer_load_dword v45, off, s[0:3], s32 offset:956 ; 4-byte Folded Reload
+; SI-NEXT: buffer_store_dword v41, off, s[0:3], s32 offset:868 ; 4-byte Folded Spill
+; SI-NEXT: buffer_load_dword v46, off, s[0:3], s32 offset:960 ; 4-byte Folded Reload
+; SI-NEXT: v_cvt_f32_f16_e32 v39, v43
+; SI-NEXT: v_cvt_f32_f16_e32 v29, v44
; SI-NEXT: ; implicit-def: $vgpr43
-; SI-NEXT: s_waitcnt vmcnt(0)
-; SI-NEXT: v_and_b32_e32 v15, 0xff, v15
-; SI-NEXT: v_or_b32_e32 v15, v15, v21
-; SI-NEXT: buffer_load_dword v21, off, s[0:3], s32 offset:908 ; 4-byte Folded Reload
-; SI-NEXT: v_cvt_f32_f16_e32 v36, v15
-; SI-NEXT: ; implicit-def: $vgpr15
-; SI-NEXT: s_waitcnt vmcnt(0)
-; SI-NEXT: v_and_b32_e32 v44, 0xff, v21
-; SI-NEXT: buffer_load_dword v21, off, s[0:3], s32 offset:980 ; 4-byte Folded Reload
-; SI-NEXT: s_waitcnt vmcnt(0)
-; SI-NEXT: v_or_b32_e32 v44, v44, v21
-; SI-NEXT: buffer_load_dword v21, off, s[0:3], s32 offset:924 ; 4-byte Folded Reload
-; SI-NEXT: s_waitcnt vmcnt(0)
-; SI-NEXT: v_and_b32_e32 v46, 0xff, v21
-; SI-NEXT: buffer_load_dword v21, off, s[0:3], s32 offset:932 ; 4-byte Folded Reload
-; SI-NEXT: v_or_b32_e32 v46, v46, v58
-; SI-NEXT: v_cvt_f32_f16_e32 v39, v46
-; SI-NEXT: ; implicit-def: $vgpr46
-; SI-NEXT: s_waitcnt vmcnt(0)
-; SI-NEXT: v_and_b32_e32 v58, 0xff, v21
-; SI-NEXT: buffer_load_dword v21, off, s[0:3], s32 offset:940 ; 4-byte Folded Reload
-; SI-NEXT: v_or_b32_e32 v58, v58, v59
-; SI-NEXT: v_cvt_f32_f16_e32 v29, v58
-; SI-NEXT: ; implicit-def: $vgpr58
-; SI-NEXT: s_waitcnt vmcnt(0)
-; SI-NEXT: v_and_b32_e32 v59, 0xff, v21
-; SI-NEXT: buffer_load_dword v21, off, s[0:3], s32 offset:960 ; 4-byte Folded Reload
-; SI-NEXT: v_or_b32_e32 v59, v59, v60
-; SI-NEXT: v_cvt_f32_f16_e32 v37, v59
+; SI-NEXT: ; implicit-def: $vgpr44
+; SI-NEXT: s_waitcnt vmcnt(3) expcnt(0)
+; SI-NEXT: v_cvt_f32_f16_e32 v41, v9
+; SI-NEXT: buffer_load_dword v9, off, s[0:3], s32 offset:472 ; 4-byte Folded Reload
+; SI-NEXT: buffer_store_dword v41, off, s[0:3], s32 offset:848 ; 4-byte Folded Spill
+; SI-NEXT: s_waitcnt vmcnt(4)
+; SI-NEXT: v_and_b32_e32 v45, 0xff, v45
+; SI-NEXT: v_or_b32_e32 v45, v45, v58
+; SI-NEXT: buffer_load_dword v58, off, s[0:3], s32 offset:968 ; 4-byte Folded Reload
+; SI-NEXT: v_cvt_f32_f16_e32 v37, v45
+; SI-NEXT: ; implicit-def: $vgpr45
+; SI-NEXT: s_waitcnt vmcnt(2) expcnt(0)
+; SI-NEXT: v_cvt_f32_f16_e32 v41, v9
+; SI-NEXT: buffer_load_dword v9, off, s[0:3], s32 offset:476 ; 4-byte Folded Reload
+; SI-NEXT: v_and_b32_e32 v46, 0xff, v46
+; SI-NEXT: v_or_b32_e32 v46, v46, v59
+; SI-NEXT: buffer_store_dword v41, off, s[0:3], s32 offset:864 ; 4-byte Folded Spill
; SI-NEXT: ; implicit-def: $vgpr59
-; SI-NEXT: s_waitcnt vmcnt(0)
-; SI-NEXT: v_and_b32_e32 v60, 0xff, v21
-; SI-NEXT: buffer_load_dword v21, off, s[0:3], s32 offset:964 ; 4-byte Folded Reload
-; SI-NEXT: v_or_b32_e32 v60, v60, v62
-; SI-NEXT: v_cvt_f32_f16_e32 v35, v60
+; SI-NEXT: s_waitcnt vmcnt(1) expcnt(0)
+; SI-NEXT: v_cvt_f32_f16_e32 v41, v9
+; SI-NEXT: buffer_load_dword v9, off, s[0:3], s32 offset:480 ; 4-byte Folded Reload
+; SI-NEXT: v_and_b32_e32 v58, 0xff, v58
+; SI-NEXT: v_or_b32_e32 v58, v58, v60
+; SI-NEXT: buffer_store_dword v41, off, s[0:3], s32 offset:836 ; 4-byte Folded Spill
+; SI-NEXT: v_cvt_f32_f16_e32 v33, v58
+; SI-NEXT: ; implicit-def: $vgpr58
; SI-NEXT: ; implicit-def: $vgpr60
-; SI-NEXT: ; implicit-def: $vgpr62
-; SI-NEXT: s_waitcnt vmcnt(0)
-; SI-NEXT: v_and_b32_e32 v61, 0xff, v21
-; SI-NEXT: buffer_load_dword v21, off, s[0:3], s32 offset:972 ; 4-byte Folded Reload
-; SI-NEXT: v_or_b32_e32 v25, v61, v25
-; SI-NEXT: v_cvt_f32_f16_e32 v17, v25
-; SI-NEXT: ; implicit-def: $vgpr25
-; SI-NEXT: s_waitcnt vmcnt(0)
-; SI-NEXT: v_and_b32_e32 v61, 0xff, v21
-; SI-NEXT: buffer_load_dword v21, off, s[0:3], s32 offset:476 ; 4-byte Folded Reload
-; SI-NEXT: v_or_b32_e32 v61, v61, v63
-; SI-NEXT: v_cvt_f32_f16_e32 v31, v61
-; SI-NEXT: ; implicit-def: $vgpr61
-; SI-NEXT: ; implicit-def: $vgpr63
-; SI-NEXT: s_waitcnt vmcnt(0)
-; SI-NEXT: v_cvt_f32_f16_e32 v21, v21
-; SI-NEXT: buffer_store_dword v21, off, s[0:3], s32 offset:824 ; 4-byte Folded Spill
-; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: buffer_load_dword v21, off, s[0:3], s32 offset:484 ; 4-byte Folded Reload
-; SI-NEXT: s_waitcnt vmcnt(0)
-; SI-NEXT: v_cvt_f32_f16_e32 v21, v21
-; SI-NEXT: buffer_store_dword v21, off, s[0:3], s32 offset:840 ; 4-byte Folded Spill
-; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: buffer_load_dword v21, off, s[0:3], s32 offset:552 ; 4-byte Folded Reload
-; SI-NEXT: s_waitcnt vmcnt(0)
-; SI-NEXT: v_cvt_f32_f16_e32 v21, v21
-; SI-NEXT: buffer_store_dword v21, off, s[0:3], s32 offset:820 ; 4-byte Folded Spill
-; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: buffer_load_dword v21, off, s[0:3], s32 offset:560 ; 4-byte Folded Reload
-; SI-NEXT: s_waitcnt vmcnt(0)
-; SI-NEXT: v_cvt_f32_f16_e32 v21, v21
-; SI-NEXT: buffer_store_dword v21, off, s[0:3], s32 offset:836 ; 4-byte Folded Spill
-; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: buffer_load_dword v21, off, s[0:3], s32 offset:564 ; 4-byte Folded Reload
-; SI-NEXT: buffer_store_dword v0, off, s[0:3], s32 offset:564 ; 4-byte Folded Spill
-; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_cvt_f32_f16_e32 v0, v44
-; SI-NEXT: ; implicit-def: $vgpr44
-; SI-NEXT: buffer_store_dword v0, off, s[0:3], s32 offset:552 ; 4-byte Folded Spill
-; SI-NEXT: ; implicit-def: $vgpr0
-; SI-NEXT: ; kill: killed $vgpr0
-; SI-NEXT: ; implicit-def: $vgpr0
-; SI-NEXT: ; kill: killed $vgpr0
-; SI-NEXT: ; implicit-def: $vgpr0
-; SI-NEXT: ; kill: killed $vgpr0
-; SI-NEXT: ; implicit-def: $vgpr0
-; SI-NEXT: ; kill: killed $vgpr0
-; SI-NEXT: ; implicit-def: $vgpr0
-; SI-NEXT: ; kill: killed $vgpr0
-; SI-NEXT: ; implicit-def: $vgpr0
-; SI-NEXT: ; kill: killed $vgpr0
-; SI-NEXT: ; implicit-def: $vgpr0
-; SI-NEXT: ; kill: killed $vgpr0
-; SI-NEXT: ; implicit-def: $vgpr0
-; SI-NEXT: ; kill: killed $vgpr0
-; SI-NEXT: ; implicit-def: $vgpr0
-; SI-NEXT: ; kill: killed $vgpr0
-; SI-NEXT: ; implicit-def: $vgpr0
-; SI-NEXT: ; kill: killed $vgpr0
-; SI-NEXT: ; implicit-def: $vgpr0
-; SI-NEXT: ; kill: killed $vgpr0
-; SI-NEXT: ; implicit-def: $vgpr0
-; SI-NEXT: ; kill: killed $vgpr0
-; SI-NEXT: ; implicit-def: $vgpr0
-; SI-NEXT: ; kill: killed $vgpr0
-; SI-NEXT: ; implicit-def: $vgpr0
-; SI-NEXT: ; kill: killed $vgpr0
-; SI-NEXT: ; implicit-def: $vgpr0
-; SI-NEXT: ; kill: killed $vgpr0
-; SI-NEXT: ; implicit-def: $vgpr0
-; SI-NEXT: ; kill: killed $vgpr0
-; SI-NEXT: ; implicit-def: $vgpr0
-; SI-NEXT: ; kill: killed $vgpr0
-; SI-NEXT: ; implicit-def: $vgpr0
-; SI-NEXT: ; kill: killed $vgpr0
-; SI-NEXT: ; implicit-def: $vgpr0
-; SI-NEXT: ; kill: killed $vgpr0
-; SI-NEXT: ; implicit-def: $vgpr0
-; SI-NEXT: ; kill: killed $vgpr0
-; SI-NEXT: ; implicit-def: $vgpr0
-; SI-NEXT: ; kill: killed $vgpr0
-; SI-NEXT: ; implicit-def: $vgpr0
-; SI-NEXT: ; kill: killed $vgpr0
-; SI-NEXT: ; implicit-def: $vgpr0
-; SI-NEXT: ; kill: killed $vgpr0
-; SI-NEXT: ; implicit-def: $vgpr0
-; SI-NEXT: ; kill: killed $vgpr0
-; SI-NEXT: ; implicit-def: $vgpr0
-; SI-NEXT: ; kill: killed $vgpr0
-; SI-NEXT: ; implicit-def: $vgpr0
-; SI-NEXT: ; kill: killed $vgpr0
-; SI-NEXT: ; implicit-def: $vgpr0
-; SI-NEXT: ; kill: killed $vgpr0
-; SI-NEXT: ; implicit-def: $vgpr0
-; SI-NEXT: ; kill: killed $vgpr0
-; SI-NEXT: ; implicit-def: $vgpr0
-; SI-NEXT: ; kill: killed $vgpr0
-; SI-NEXT: ; implicit-def: $vgpr0
-; SI-NEXT: ; kill: killed $vgpr0
-; SI-NEXT: ; implicit-def: $vgpr0
-; SI-NEXT: ; kill: killed $vgpr0
-; SI-NEXT: ; implicit-def: $vgpr0
-; SI-NEXT: ; kill: killed $vgpr0
-; SI-NEXT: ; implicit-def: $vgpr0
-; SI-NEXT: ; kill: killed $vgpr0
-; SI-NEXT: ; implicit-def: $vgpr0
-; SI-NEXT: ; kill: killed $vgpr0
-; SI-NEXT: ; implicit-def: $vgpr0
-; SI-NEXT: ; kill: killed $vgpr0
-; SI-NEXT: ; implicit-def: $vgpr0
-; SI-NEXT: ; kill: killed $vgpr0
-; SI-NEXT: ; implicit-def: $vgpr0
-; SI-NEXT: ; kill: killed $vgpr0
-; SI-NEXT: ; implicit-def: $vgpr0
-; SI-NEXT: ; kill: killed $vgpr0
-; SI-NEXT: ; implicit-def: $vgpr0
-; SI-NEXT: ; kill: killed $vgpr0
-; SI-NEXT: ; implicit-def: $vgpr0
-; SI-NEXT: ; kill: killed $vgpr0
-; SI-NEXT: ; implicit-def: $vgpr0
-; SI-NEXT: ; kill: killed $vgpr0
-; SI-NEXT: ; implicit-def: $vgpr0
-; SI-NEXT: ; kill: killed $vgpr0
-; SI-NEXT: ; implicit-def: $vgpr0
-; SI-NEXT: ; kill: killed $vgpr0
-; SI-NEXT: ; implicit-def: $vgpr0
-; SI-NEXT: ; kill: killed $vgpr0
-; SI-NEXT: ; implicit-def: $vgpr0
-; SI-NEXT: ; kill: killed $vgpr0
-; SI-NEXT: ; implicit-def: $vgpr0
-; SI-NEXT: ; kill: killed $vgpr0
-; SI-NEXT: ; implicit-def: $vgpr0
-; SI-NEXT: ; kill: killed $vgpr0
-; SI-NEXT: ; implicit-def: $vgpr0
-; SI-NEXT: ; kill: killed $vgpr0
-; SI-NEXT: ; implicit-def: $vgpr0
-; SI-NEXT: ; kill: killed $vgpr0
-; SI-NEXT: ; implicit-def: $vgpr0
-; SI-NEXT: ; kill: killed $vgpr0
-; SI-NEXT: ; implicit-def: $vgpr0
-; SI-NEXT: ; kill: killed $vgpr0
-; SI-NEXT: ; implicit-def: $vgpr0
-; SI-NEXT: ; kill: killed $vgpr0
-; SI-NEXT: ; implicit-def: $vgpr0
-; SI-NEXT: ; kill: killed $vgpr0
-; SI-NEXT: ; implicit-def: $vgpr0
-; SI-NEXT: ; kill: killed $vgpr0
-; SI-NEXT: ; implicit-def: $vgpr0
-; SI-NEXT: ; kill: killed $vgpr0
-; SI-NEXT: ; implicit-def: $vgpr0
-; SI-NEXT: ; kill: killed $vgpr0
-; SI-NEXT: ; implicit-def: $vgpr0
-; SI-NEXT: ; kill: killed $vgpr0
-; SI-NEXT: ; implicit-def: $vgpr0
-; SI-NEXT: ; kill: killed $vgpr0
-; SI-NEXT: ; implicit-def: $vgpr0
-; SI-NEXT: ; kill: killed $vgpr0
-; SI-NEXT: ; implicit-def: $vgpr0
-; SI-NEXT: ; kill: killed $vgpr0
-; SI-NEXT: ; implicit-def: $vgpr0
-; SI-NEXT: ; kill: killed $vgpr0
-; SI-NEXT: ; implicit-def: $vgpr0
-; SI-NEXT: ; kill: killed $vgpr0
-; SI-NEXT: ; implicit-def: $vgpr0
-; SI-NEXT: ; kill: killed $vgpr0
-; SI-NEXT: ; implicit-def: $vgpr0
-; SI-NEXT: ; kill: killed $vgpr0
-; SI-NEXT: ; implicit-def: $vgpr0
-; SI-NEXT: ; kill: killed $vgpr0
-; SI-NEXT: ; implicit-def: $vgpr0
-; SI-NEXT: ; kill: killed $vgpr0
-; SI-NEXT: ; implicit-def: $vgpr0
-; SI-NEXT: ; kill: killed $vgpr0
-; SI-NEXT: ; implicit-def: $vgpr0
-; SI-NEXT: ; kill: killed $vgpr0
-; SI-NEXT: ; implicit-def: $vgpr0
-; SI-NEXT: ; kill: killed $vgpr0
-; SI-NEXT: ; implicit-def: $vgpr0
-; SI-NEXT: ; kill: killed $vgpr0
-; SI-NEXT: ; implicit-def: $vgpr0
-; SI-NEXT: ; kill: killed $vgpr0
-; SI-NEXT: ; implicit-def: $vgpr0
-; SI-NEXT: ; kill: killed $vgpr0
-; SI-NEXT: ; implicit-def: $vgpr0
-; SI-NEXT: ; kill: killed $vgpr0
-; SI-NEXT: ; implicit-def: $vgpr0
-; SI-NEXT: s_waitcnt vmcnt(2)
-; SI-NEXT: v_cvt_f32_f16_e32 v21, v21
-; SI-NEXT: ; kill: killed $vgpr0
-; SI-NEXT: ; implicit-def: $vgpr0
-; SI-NEXT: ; kill: killed $vgpr0
-; SI-NEXT: ; implicit-def: $vgpr0
-; SI-NEXT: ; kill: killed $vgpr0
-; SI-NEXT: ; implicit-def: $vgpr0
-; SI-NEXT: ; kill: killed $vgpr0
-; SI-NEXT: ; implicit-def: $vgpr0
-; SI-NEXT: buffer_store_dword v21, off, s[0:3], s32 offset:816 ; 4-byte Folded Spill
-; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: buffer_load_dword v21, off, s[0:3], s32 offset:568 ; 4-byte Folded Reload
-; SI-NEXT: buffer_store_dword v2, off, s[0:3], s32 offset:568 ; 4-byte Folded Spill
-; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_cvt_f32_f16_e32 v2, v55
-; SI-NEXT: ; kill: killed $vgpr0
-; SI-NEXT: ; implicit-def: $vgpr0
-; SI-NEXT: ; implicit-def: $vgpr55
-; SI-NEXT: ; kill: killed $vgpr0
-; SI-NEXT: ; implicit-def: $vgpr0
-; SI-NEXT: ; kill: killed $vgpr0
-; SI-NEXT: buffer_store_dword v2, off, s[0:3], s32 offset:600 ; 4-byte Folded Spill
+; SI-NEXT: s_waitcnt vmcnt(1) expcnt(0)
+; SI-NEXT: v_cvt_f32_f16_e32 v41, v9
+; SI-NEXT: buffer_load_dword v9, off, s[0:3], s32 offset:484 ; 4-byte Folded Reload
+; SI-NEXT: buffer_store_dword v41, off, s[0:3], s32 offset:860 ; 4-byte Folded Spill
+; SI-NEXT: s_waitcnt vmcnt(1) expcnt(0)
+; SI-NEXT: v_cvt_f32_f16_e32 v41, v9
+; SI-NEXT: buffer_load_dword v9, off, s[0:3], s32 offset:488 ; 4-byte Folded Reload
+; SI-NEXT: buffer_store_dword v41, off, s[0:3], s32 offset:824 ; 4-byte Folded Spill
+; SI-NEXT: ; implicit-def: $vgpr41
+; SI-NEXT: s_waitcnt vmcnt(1)
+; SI-NEXT: v_cvt_f32_f16_e32 v9, v9
+; SI-NEXT: buffer_store_dword v9, off, s[0:3], s32 offset:840 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_cvt_f32_f16_e32 v2, v13
-; SI-NEXT: ; implicit-def: $vgpr13
-; SI-NEXT: buffer_store_dword v2, off, s[0:3], s32 offset:560 ; 4-byte Folded Spill
+; SI-NEXT: v_cvt_f32_f16_e32 v9, v23
+; SI-NEXT: v_cvt_f32_f16_e32 v23, v46
+; SI-NEXT: ; implicit-def: $vgpr46
+; SI-NEXT: buffer_store_dword v9, off, s[0:3], s32 offset:808 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_cvt_f32_f16_e32 v2, v41
-; SI-NEXT: ; implicit-def: $vgpr41
-; SI-NEXT: buffer_store_dword v2, off, s[0:3], s32 offset:588 ; 4-byte Folded Spill
-; SI-NEXT: s_waitcnt vmcnt(4)
-; SI-NEXT: v_cvt_f32_f16_e32 v21, v21
-; SI-NEXT: buffer_store_dword v21, off, s[0:3], s32 offset:832 ; 4-byte Folded Spill
-; SI-NEXT: ; implicit-def: $vgpr21
+; SI-NEXT: v_cvt_f32_f16_e32 v9, v1
+; SI-NEXT: ; implicit-def: $vgpr1
+; SI-NEXT: ; kill: killed $vgpr1
+; SI-NEXT: ; implicit-def: $vgpr1
+; SI-NEXT: ; kill: killed $vgpr1
+; SI-NEXT: ; implicit-def: $vgpr1
+; SI-NEXT: ; kill: killed $vgpr1
+; SI-NEXT: ; implicit-def: $vgpr1
+; SI-NEXT: ; kill: killed $vgpr1
+; SI-NEXT: ; implicit-def: $vgpr1
+; SI-NEXT: ; kill: killed $vgpr1
+; SI-NEXT: ; implicit-def: $vgpr1
+; SI-NEXT: ; kill: killed $vgpr1
+; SI-NEXT: ; implicit-def: $vgpr1
+; SI-NEXT: ; kill: killed $vgpr1
+; SI-NEXT: ; implicit-def: $vgpr1
+; SI-NEXT: ; kill: killed $vgpr1
+; SI-NEXT: ; implicit-def: $vgpr1
+; SI-NEXT: ; kill: killed $vgpr1
+; SI-NEXT: ; implicit-def: $vgpr1
+; SI-NEXT: ; kill: killed $vgpr1
+; SI-NEXT: ; implicit-def: $vgpr1
+; SI-NEXT: ; kill: killed $vgpr1
+; SI-NEXT: ; implicit-def: $vgpr1
+; SI-NEXT: ; kill: killed $vgpr1
+; SI-NEXT: ; implicit-def: $vgpr1
+; SI-NEXT: ; kill: killed $vgpr1
+; SI-NEXT: ; implicit-def: $vgpr1
+; SI-NEXT: ; kill: killed $vgpr1
+; SI-NEXT: ; implicit-def: $vgpr1
+; SI-NEXT: ; kill: killed $vgpr1
+; SI-NEXT: ; implicit-def: $vgpr1
+; SI-NEXT: ; kill: killed $vgpr1
+; SI-NEXT: ; implicit-def: $vgpr1
+; SI-NEXT: ; kill: killed $vgpr1
+; SI-NEXT: ; implicit-def: $vgpr1
+; SI-NEXT: ; kill: killed $vgpr1
+; SI-NEXT: ; implicit-def: $vgpr1
+; SI-NEXT: ; kill: killed $vgpr1
+; SI-NEXT: ; implicit-def: $vgpr1
+; SI-NEXT: ; kill: killed $vgpr1
+; SI-NEXT: ; implicit-def: $vgpr1
+; SI-NEXT: ; kill: killed $vgpr1
+; SI-NEXT: ; implicit-def: $vgpr1
+; SI-NEXT: ; kill: killed $vgpr1
+; SI-NEXT: ; implicit-def: $vgpr1
+; SI-NEXT: ; kill: killed $vgpr1
+; SI-NEXT: ; implicit-def: $vgpr1
+; SI-NEXT: ; kill: killed $vgpr1
+; SI-NEXT: ; implicit-def: $vgpr1
+; SI-NEXT: ; kill: killed $vgpr1
+; SI-NEXT: ; implicit-def: $vgpr1
+; SI-NEXT: ; kill: killed $vgpr1
+; SI-NEXT: ; implicit-def: $vgpr1
+; SI-NEXT: ; kill: killed $vgpr1
+; SI-NEXT: ; implicit-def: $vgpr1
+; SI-NEXT: ; kill: killed $vgpr1
+; SI-NEXT: ; implicit-def: $vgpr1
+; SI-NEXT: ; kill: killed $vgpr1
+; SI-NEXT: ; implicit-def: $vgpr1
+; SI-NEXT: ; kill: killed $vgpr1
+; SI-NEXT: ; implicit-def: $vgpr1
+; SI-NEXT: ; kill: killed $vgpr1
+; SI-NEXT: ; implicit-def: $vgpr1
+; SI-NEXT: ; kill: killed $vgpr1
+; SI-NEXT: ; implicit-def: $vgpr1
+; SI-NEXT: ; kill: killed $vgpr1
+; SI-NEXT: ; implicit-def: $vgpr1
+; SI-NEXT: ; kill: killed $vgpr1
+; SI-NEXT: ; implicit-def: $vgpr1
+; SI-NEXT: ; kill: killed $vgpr1
+; SI-NEXT: ; implicit-def: $vgpr1
+; SI-NEXT: ; kill: killed $vgpr1
+; SI-NEXT: ; implicit-def: $vgpr1
+; SI-NEXT: ; kill: killed $vgpr1
+; SI-NEXT: ; implicit-def: $vgpr1
+; SI-NEXT: ; kill: killed $vgpr1
+; SI-NEXT: ; implicit-def: $vgpr1
+; SI-NEXT: ; kill: killed $vgpr1
+; SI-NEXT: ; implicit-def: $vgpr1
+; SI-NEXT: ; kill: killed $vgpr1
+; SI-NEXT: ; implicit-def: $vgpr1
+; SI-NEXT: ; kill: killed $vgpr1
+; SI-NEXT: ; implicit-def: $vgpr1
+; SI-NEXT: ; kill: killed $vgpr1
+; SI-NEXT: ; implicit-def: $vgpr1
+; SI-NEXT: ; kill: killed $vgpr1
+; SI-NEXT: ; implicit-def: $vgpr1
+; SI-NEXT: ; kill: killed $vgpr1
+; SI-NEXT: ; implicit-def: $vgpr1
; SI-NEXT: .LBB92_2: ; %Flow
; SI-NEXT: s_andn2_saveexec_b64 s[4:5], s[4:5]
; SI-NEXT: s_cbranch_execz .LBB92_4
; SI-NEXT: ; %bb.3: ; %cmp.true
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:972 ; 4-byte Folded Reload
; SI-NEXT: v_add_i32_e32 v2, vcc, 3, v47
; SI-NEXT: v_and_b32_e32 v2, 0xff, v2
-; SI-NEXT: v_or_b32_e32 v5, v3, v2
-; SI-NEXT: s_waitcnt vmcnt(9)
+; SI-NEXT: v_or_b32_e32 v9, v3, v2
+; SI-NEXT: s_waitcnt vmcnt(8)
; SI-NEXT: v_add_i32_e32 v2, vcc, 3, v57
; SI-NEXT: v_and_b32_e32 v2, 0xff, v2
; SI-NEXT: v_or_b32_e32 v1, v1, v2
-; SI-NEXT: v_add_i32_e32 v29, vcc, 3, v19
-; SI-NEXT: v_and_b32_e32 v29, 0xff, v29
-; SI-NEXT: v_add_i32_e32 v20, vcc, 3, v20
-; SI-NEXT: v_and_b32_e32 v20, 0xff, v20
+; SI-NEXT: v_add_i32_e32 v2, vcc, 3, v0
+; SI-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:968 ; 4-byte Folded Reload
+; SI-NEXT: v_and_b32_e32 v2, 0xff, v2
+; SI-NEXT: v_or_b32_e32 v11, v63, v2
+; SI-NEXT: v_add_i32_e32 v2, vcc, 3, v41
+; SI-NEXT: v_and_b32_e32 v2, 0xff, v2
+; SI-NEXT: v_or_b32_e32 v13, v27, v2
+; SI-NEXT: v_add_i32_e32 v2, vcc, 3, v55
+; SI-NEXT: v_and_b32_e32 v2, 0xff, v2
+; SI-NEXT: v_or_b32_e32 v21, v62, v2
+; SI-NEXT: v_add_i32_e32 v2, vcc, 3, v56
+; SI-NEXT: v_and_b32_e32 v2, 0xff, v2
+; SI-NEXT: v_or_b32_e32 v23, v61, v2
; SI-NEXT: v_add_i32_e32 v16, vcc, 3, v16
; SI-NEXT: v_and_b32_e32 v16, 0xff, v16
-; SI-NEXT: s_waitcnt vmcnt(7)
-; SI-NEXT: v_add_i32_e32 v34, vcc, 3, v34
-; SI-NEXT: v_and_b32_e32 v34, 0xff, v34
; SI-NEXT: s_waitcnt vmcnt(6)
-; SI-NEXT: v_add_i32_e32 v30, vcc, 3, v30
-; SI-NEXT: v_and_b32_e32 v30, 0xff, v30
+; SI-NEXT: v_add_i32_e32 v29, vcc, 3, v30
+; SI-NEXT: v_and_b32_e32 v29, 0xff, v29
; SI-NEXT: s_waitcnt vmcnt(5)
; SI-NEXT: v_add_i32_e32 v28, vcc, 3, v28
; SI-NEXT: v_and_b32_e32 v28, 0xff, v28
@@ -179899,791 +179986,737 @@ define <64 x half> @bitcast_v128i8_to_v64f16(<128 x i8> %a, i32 %b) {
; SI-NEXT: v_and_b32_e32 v18, 0xff, v18
; SI-NEXT: v_add_i32_e32 v4, vcc, 3, v4
; SI-NEXT: v_and_b32_e32 v4, 0xff, v4
-; SI-NEXT: v_mov_b32_e32 v17, v43
-; SI-NEXT: v_mov_b32_e32 v19, v44
-; SI-NEXT: v_add_i32_e32 v47, vcc, 3, v19
-; SI-NEXT: v_and_b32_e32 v47, 0xff, v47
-; SI-NEXT: v_add_i32_e32 v57, vcc, 3, v15
-; SI-NEXT: v_and_b32_e32 v57, 0xff, v57
-; SI-NEXT: v_add_i32_e32 v3, vcc, 3, v51
-; SI-NEXT: v_and_b32_e32 v3, 0xff, v3
; SI-NEXT: s_movk_i32 s6, 0x300
+; SI-NEXT: v_add_i32_e32 v63, vcc, 3, v51
+; SI-NEXT: v_add_i32_e32 v3, vcc, 3, v49
+; SI-NEXT: v_and_b32_e32 v63, 0xff, v63
+; SI-NEXT: v_and_b32_e32 v3, 0xff, v3
; SI-NEXT: s_waitcnt vmcnt(0)
; SI-NEXT: v_add_i32_e32 v2, vcc, 3, v0
-; SI-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:964 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:960 ; 4-byte Folded Reload
; SI-NEXT: v_and_b32_e32 v2, 0xff, v2
-; SI-NEXT: v_or_b32_e32 v23, v63, v2
-; SI-NEXT: v_add_i32_e32 v63, vcc, 3, v54
-; SI-NEXT: v_and_b32_e32 v63, 0xff, v63
+; SI-NEXT: v_or_b32_e32 v25, v60, v2
; SI-NEXT: s_waitcnt vmcnt(0)
; SI-NEXT: v_add_i32_e32 v2, vcc, 3, v0
-; SI-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:960 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:956 ; 4-byte Folded Reload
; SI-NEXT: v_and_b32_e32 v2, 0xff, v2
-; SI-NEXT: v_or_b32_e32 v25, v25, v2
+; SI-NEXT: v_or_b32_e32 v27, v59, v2
; SI-NEXT: s_waitcnt vmcnt(0)
; SI-NEXT: v_add_i32_e32 v2, vcc, 3, v0
; SI-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:940 ; 4-byte Folded Reload
; SI-NEXT: v_and_b32_e32 v2, 0xff, v2
-; SI-NEXT: v_or_b32_e32 v31, v62, v2
-; SI-NEXT: v_add_i32_e32 v2, vcc, 3, v56
+; SI-NEXT: v_or_b32_e32 v31, v58, v2
+; SI-NEXT: v_add_i32_e32 v2, vcc, 3, v46
; SI-NEXT: v_and_b32_e32 v2, 0xff, v2
-; SI-NEXT: v_or_b32_e32 v33, v61, v2
-; SI-NEXT: v_add_i32_e32 v56, vcc, 3, v17
-; SI-NEXT: v_and_b32_e32 v56, 0xff, v56
-; SI-NEXT: v_add_i32_e32 v61, vcc, 3, v41
-; SI-NEXT: v_and_b32_e32 v61, 0xff, v61
-; SI-NEXT: v_add_i32_e32 v62, vcc, 3, v55
-; SI-NEXT: v_and_b32_e32 v62, 0xff, v62
+; SI-NEXT: v_or_b32_e32 v32, v45, v2
; SI-NEXT: s_waitcnt vmcnt(0)
; SI-NEXT: v_add_i32_e32 v2, vcc, 3, v0
; SI-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:932 ; 4-byte Folded Reload
; SI-NEXT: v_and_b32_e32 v2, 0xff, v2
-; SI-NEXT: v_or_b32_e32 v35, v60, v2
-; SI-NEXT: v_add_i32_e32 v60, vcc, 3, v9
-; SI-NEXT: v_and_b32_e32 v60, 0xff, v60
-; SI-NEXT: v_or_b32_e32 v12, v12, v60
-; SI-NEXT: v_add_i32_e32 v60, vcc, s6, v25
+; SI-NEXT: v_or_b32_e32 v33, v44, v2
; SI-NEXT: s_waitcnt vmcnt(0)
; SI-NEXT: v_add_i32_e32 v2, vcc, 3, v0
-; SI-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:924 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:928 ; 4-byte Folded Reload
; SI-NEXT: v_and_b32_e32 v2, 0xff, v2
-; SI-NEXT: v_or_b32_e32 v37, v59, v2
-; SI-NEXT: v_add_i32_e32 v59, vcc, 3, v11
-; SI-NEXT: v_and_b32_e32 v59, 0xff, v59
-; SI-NEXT: v_or_b32_e32 v14, v14, v59
-; SI-NEXT: v_add_i32_e32 v59, vcc, s6, v31
+; SI-NEXT: v_or_b32_e32 v35, v43, v2
; SI-NEXT: s_waitcnt vmcnt(0)
; SI-NEXT: v_add_i32_e32 v2, vcc, 3, v0
-; SI-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:908 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:888 ; 4-byte Folded Reload
; SI-NEXT: v_and_b32_e32 v2, 0xff, v2
-; SI-NEXT: v_or_b32_e32 v38, v58, v2
-; SI-NEXT: v_add_i32_e32 v2, vcc, 3, v46
+; SI-NEXT: v_or_b32_e32 v37, v40, v2
+; SI-NEXT: v_add_i32_e32 v2, vcc, 3, v42
; SI-NEXT: v_and_b32_e32 v2, 0xff, v2
-; SI-NEXT: v_or_b32_e32 v39, v45, v2
-; SI-NEXT: v_add_i32_e32 v46, vcc, 3, v21
-; SI-NEXT: v_and_b32_e32 v46, 0xff, v46
-; SI-NEXT: v_add_i32_e32 v58, vcc, 3, v13
-; SI-NEXT: v_and_b32_e32 v58, 0xff, v58
-; SI-NEXT: v_or_b32_e32 v32, v32, v58
-; SI-NEXT: v_add_i32_e32 v58, vcc, s6, v33
-; SI-NEXT: v_add_i32_e32 v9, vcc, s6, v32
+; SI-NEXT: v_or_b32_e32 v38, v54, v2
; SI-NEXT: s_waitcnt vmcnt(0)
; SI-NEXT: v_add_i32_e32 v2, vcc, 3, v0
-; SI-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:980 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:976 ; 4-byte Folded Reload
; SI-NEXT: v_and_b32_e32 v2, 0xff, v2
; SI-NEXT: s_waitcnt vmcnt(0)
-; SI-NEXT: v_or_b32_e32 v48, v0, v2
-; SI-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:900 ; 4-byte Folded Reload
+; SI-NEXT: v_or_b32_e32 v39, v0, v2
+; SI-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:880 ; 4-byte Folded Reload
; SI-NEXT: s_waitcnt vmcnt(0)
; SI-NEXT: v_add_i32_e32 v2, vcc, 3, v0
-; SI-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:976 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:972 ; 4-byte Folded Reload
; SI-NEXT: v_and_b32_e32 v2, 0xff, v2
; SI-NEXT: s_waitcnt vmcnt(0)
-; SI-NEXT: v_or_b32_e32 v0, v0, v2
-; SI-NEXT: buffer_store_dword v0, off, s[0:3], s32 offset:552 ; 4-byte Folded Spill
-; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:896 ; 4-byte Folded Reload
+; SI-NEXT: v_or_b32_e32 v48, v0, v2
+; SI-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:876 ; 4-byte Folded Reload
; SI-NEXT: s_waitcnt vmcnt(0)
; SI-NEXT: v_add_i32_e32 v2, vcc, 3, v0
-; SI-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:968 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:964 ; 4-byte Folded Reload
; SI-NEXT: v_and_b32_e32 v2, 0xff, v2
; SI-NEXT: s_waitcnt vmcnt(0)
-; SI-NEXT: v_or_b32_e32 v0, v0, v2
-; SI-NEXT: buffer_store_dword v0, off, s[0:3], s32 offset:560 ; 4-byte Folded Spill
-; SI-NEXT: s_waitcnt expcnt(0)
+; SI-NEXT: v_or_b32_e32 v54, v0, v2
; SI-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:952 ; 4-byte Folded Reload
-; SI-NEXT: v_add_i32_e32 v2, vcc, 3, v42
+; SI-NEXT: v_add_i32_e32 v2, vcc, 3, v53
; SI-NEXT: v_and_b32_e32 v2, 0xff, v2
; SI-NEXT: s_waitcnt vmcnt(0)
-; SI-NEXT: v_or_b32_e32 v0, v0, v2
-; SI-NEXT: buffer_store_dword v0, off, s[0:3], s32 offset:564 ; 4-byte Folded Spill
-; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:856 ; 4-byte Folded Reload
+; SI-NEXT: v_or_b32_e32 v55, v0, v2
+; SI-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:812 ; 4-byte Folded Reload
; SI-NEXT: s_waitcnt vmcnt(0)
; SI-NEXT: v_add_i32_e32 v2, vcc, 3, v0
; SI-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:948 ; 4-byte Folded Reload
; SI-NEXT: v_and_b32_e32 v2, 0xff, v2
; SI-NEXT: s_waitcnt vmcnt(0)
-; SI-NEXT: v_or_b32_e32 v0, v0, v2
-; SI-NEXT: buffer_store_dword v0, off, s[0:3], s32 offset:568 ; 4-byte Folded Spill
-; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:848 ; 4-byte Folded Reload
+; SI-NEXT: v_or_b32_e32 v41, v0, v2
+; SI-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:800 ; 4-byte Folded Reload
; SI-NEXT: s_waitcnt vmcnt(0)
; SI-NEXT: v_add_i32_e32 v2, vcc, 3, v0
; SI-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:944 ; 4-byte Folded Reload
; SI-NEXT: v_and_b32_e32 v2, 0xff, v2
; SI-NEXT: s_waitcnt vmcnt(0)
; SI-NEXT: v_or_b32_e32 v0, v0, v2
-; SI-NEXT: buffer_store_dword v0, off, s[0:3], s32 offset:576 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v0, off, s[0:3], s32 offset:596 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:844 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:784 ; 4-byte Folded Reload
; SI-NEXT: s_waitcnt vmcnt(0)
; SI-NEXT: v_add_i32_e32 v2, vcc, 3, v0
; SI-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:936 ; 4-byte Folded Reload
; SI-NEXT: v_and_b32_e32 v2, 0xff, v2
; SI-NEXT: s_waitcnt vmcnt(0)
; SI-NEXT: v_or_b32_e32 v0, v0, v2
-; SI-NEXT: buffer_store_dword v0, off, s[0:3], s32 offset:580 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v0, off, s[0:3], s32 offset:600 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:956 ; 4-byte Folded Reload
-; SI-NEXT: s_waitcnt vmcnt(0)
-; SI-NEXT: v_add_i32_e32 v2, vcc, 3, v0
-; SI-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:920 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:900 ; 4-byte Folded Reload
+; SI-NEXT: v_add_i32_e32 v2, vcc, 3, v52
; SI-NEXT: v_and_b32_e32 v2, 0xff, v2
; SI-NEXT: s_waitcnt vmcnt(0)
; SI-NEXT: v_or_b32_e32 v0, v0, v2
-; SI-NEXT: buffer_store_dword v0, off, s[0:3], s32 offset:588 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v0, off, s[0:3], s32 offset:608 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:780 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:656 ; 4-byte Folded Reload
; SI-NEXT: s_waitcnt vmcnt(0)
; SI-NEXT: v_add_i32_e32 v2, vcc, 3, v0
-; SI-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:916 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:896 ; 4-byte Folded Reload
; SI-NEXT: v_and_b32_e32 v2, 0xff, v2
; SI-NEXT: s_waitcnt vmcnt(0)
; SI-NEXT: v_or_b32_e32 v0, v0, v2
-; SI-NEXT: buffer_store_dword v0, off, s[0:3], s32 offset:596 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v0, off, s[0:3], s32 offset:612 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:768 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:636 ; 4-byte Folded Reload
; SI-NEXT: s_waitcnt vmcnt(0)
; SI-NEXT: v_add_i32_e32 v2, vcc, 3, v0
-; SI-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:912 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:892 ; 4-byte Folded Reload
; SI-NEXT: v_and_b32_e32 v2, 0xff, v2
; SI-NEXT: s_waitcnt vmcnt(0)
; SI-NEXT: v_or_b32_e32 v0, v0, v2
-; SI-NEXT: buffer_store_dword v0, off, s[0:3], s32 offset:600 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v0, off, s[0:3], s32 offset:620 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:752 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:616 ; 4-byte Folded Reload
; SI-NEXT: s_waitcnt vmcnt(0)
; SI-NEXT: v_add_i32_e32 v2, vcc, 3, v0
-; SI-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:904 ; 4-byte Folded Reload
-; SI-NEXT: v_and_b32_e32 v2, 0xff, v2
-; SI-NEXT: s_waitcnt vmcnt(0)
-; SI-NEXT: v_or_b32_e32 v0, v0, v2
-; SI-NEXT: buffer_store_dword v0, off, s[0:3], s32 offset:608 ; 4-byte Folded Spill
-; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:868 ; 4-byte Folded Reload
-; SI-NEXT: v_add_i32_e32 v2, vcc, 3, v52
+; SI-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:884 ; 4-byte Folded Reload
; SI-NEXT: v_and_b32_e32 v2, 0xff, v2
; SI-NEXT: s_waitcnt vmcnt(0)
; SI-NEXT: v_or_b32_e32 v0, v0, v2
; SI-NEXT: buffer_store_dword v0, off, s[0:3], s32 offset:616 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:624 ; 4-byte Folded Reload
-; SI-NEXT: s_waitcnt vmcnt(0)
-; SI-NEXT: v_add_i32_e32 v2, vcc, 3, v0
-; SI-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:864 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:844 ; 4-byte Folded Reload
+; SI-NEXT: v_add_i32_e32 v2, vcc, 3, v50
; SI-NEXT: v_and_b32_e32 v2, 0xff, v2
; SI-NEXT: s_waitcnt vmcnt(0)
-; SI-NEXT: v_or_b32_e32 v0, v0, v2
-; SI-NEXT: buffer_store_dword v0, off, s[0:3], s32 offset:620 ; 4-byte Folded Spill
-; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:604 ; 4-byte Folded Reload
-; SI-NEXT: s_waitcnt vmcnt(0)
-; SI-NEXT: v_add_i32_e32 v2, vcc, 3, v0
-; SI-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:860 ; 4-byte Folded Reload
+; SI-NEXT: v_or_b32_e32 v50, v0, v2
+; SI-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:828 ; 4-byte Folded Reload
+; SI-NEXT: v_add_i32_e32 v2, vcc, 3, v19
; SI-NEXT: v_and_b32_e32 v2, 0xff, v2
; SI-NEXT: s_waitcnt vmcnt(0)
-; SI-NEXT: v_or_b32_e32 v0, v0, v2
-; SI-NEXT: buffer_store_dword v0, off, s[0:3], s32 offset:604 ; 4-byte Folded Spill
-; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:584 ; 4-byte Folded Reload
-; SI-NEXT: s_waitcnt vmcnt(0)
-; SI-NEXT: v_add_i32_e32 v2, vcc, 3, v0
-; SI-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:852 ; 4-byte Folded Reload
+; SI-NEXT: v_or_b32_e32 v19, v0, v2
+; SI-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:820 ; 4-byte Folded Reload
+; SI-NEXT: v_add_i32_e32 v2, vcc, 3, v17
; SI-NEXT: v_and_b32_e32 v2, 0xff, v2
; SI-NEXT: s_waitcnt vmcnt(0)
-; SI-NEXT: v_or_b32_e32 v0, v0, v2
-; SI-NEXT: buffer_store_dword v0, off, s[0:3], s32 offset:584 ; 4-byte Folded Spill
-; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:812 ; 4-byte Folded Reload
-; SI-NEXT: v_add_i32_e32 v2, vcc, 3, v49
+; SI-NEXT: v_or_b32_e32 v17, v0, v2
+; SI-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:804 ; 4-byte Folded Reload
+; SI-NEXT: v_add_i32_e32 v2, vcc, 3, v15
; SI-NEXT: v_and_b32_e32 v2, 0xff, v2
; SI-NEXT: s_waitcnt vmcnt(0)
-; SI-NEXT: v_or_b32_e32 v0, v0, v2
-; SI-NEXT: buffer_store_dword v0, off, s[0:3], s32 offset:624 ; 4-byte Folded Spill
-; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:796 ; 4-byte Folded Reload
-; SI-NEXT: v_add_i32_e32 v2, vcc, 3, v40
+; SI-NEXT: v_or_b32_e32 v15, v0, v2
+; SI-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:776 ; 4-byte Folded Reload
+; SI-NEXT: v_add_i32_e32 v2, vcc, 3, v36
; SI-NEXT: v_and_b32_e32 v2, 0xff, v2
; SI-NEXT: s_waitcnt vmcnt(0)
; SI-NEXT: v_or_b32_e32 v0, v0, v2
; SI-NEXT: buffer_store_dword v0, off, s[0:3], s32 offset:628 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:788 ; 4-byte Folded Reload
-; SI-NEXT: v_add_i32_e32 v2, vcc, 3, v53
-; SI-NEXT: v_and_b32_e32 v2, 0xff, v2
+; SI-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:528 ; 4-byte Folded Reload
; SI-NEXT: s_waitcnt vmcnt(0)
-; SI-NEXT: v_or_b32_e32 v0, v0, v2
-; SI-NEXT: buffer_store_dword v0, off, s[0:3], s32 offset:636 ; 4-byte Folded Spill
-; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:772 ; 4-byte Folded Reload
-; SI-NEXT: v_add_i32_e32 v2, vcc, 3, v50
+; SI-NEXT: v_add_i32_e32 v2, vcc, 3, v0
+; SI-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:760 ; 4-byte Folded Reload
; SI-NEXT: v_and_b32_e32 v2, 0xff, v2
; SI-NEXT: s_waitcnt vmcnt(0)
; SI-NEXT: v_or_b32_e32 v0, v0, v2
-; SI-NEXT: buffer_store_dword v0, off, s[0:3], s32 offset:640 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v0, off, s[0:3], s32 offset:528 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:928 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:520 ; 4-byte Folded Reload
; SI-NEXT: s_waitcnt vmcnt(0)
; SI-NEXT: v_add_i32_e32 v2, vcc, 3, v0
; SI-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:744 ; 4-byte Folded Reload
; SI-NEXT: v_and_b32_e32 v2, 0xff, v2
; SI-NEXT: s_waitcnt vmcnt(0)
; SI-NEXT: v_or_b32_e32 v0, v0, v2
-; SI-NEXT: buffer_store_dword v0, off, s[0:3], s32 offset:644 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v0, off, s[0:3], s32 offset:520 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:488 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:512 ; 4-byte Folded Reload
; SI-NEXT: s_waitcnt vmcnt(0)
; SI-NEXT: v_add_i32_e32 v2, vcc, 3, v0
; SI-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:728 ; 4-byte Folded Reload
; SI-NEXT: v_and_b32_e32 v2, 0xff, v2
; SI-NEXT: s_waitcnt vmcnt(0)
-; SI-NEXT: v_or_b32_e32 v2, v0, v2
-; SI-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:480 ; 4-byte Folded Reload
-; SI-NEXT: s_waitcnt vmcnt(0)
-; SI-NEXT: v_add_i32_e32 v27, vcc, 3, v0
+; SI-NEXT: v_or_b32_e32 v0, v0, v2
+; SI-NEXT: buffer_store_dword v0, off, s[0:3], s32 offset:512 ; 4-byte Folded Spill
+; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:712 ; 4-byte Folded Reload
-; SI-NEXT: v_and_b32_e32 v27, 0xff, v27
+; SI-NEXT: v_add_i32_e32 v2, vcc, 3, v20
+; SI-NEXT: v_and_b32_e32 v2, 0xff, v2
+; SI-NEXT: v_add_i32_e32 v20, vcc, 3, v34
+; SI-NEXT: v_and_b32_e32 v20, 0xff, v20
; SI-NEXT: s_waitcnt vmcnt(0)
-; SI-NEXT: v_or_b32_e32 v0, v0, v27
-; SI-NEXT: buffer_store_dword v0, off, s[0:3], s32 offset:480 ; 4-byte Folded Spill
-; SI-NEXT: s_waitcnt expcnt(0)
+; SI-NEXT: v_or_b32_e32 v2, v0, v2
; SI-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:696 ; 4-byte Folded Reload
-; SI-NEXT: v_add_i32_e32 v27, vcc, s6, v2
-; SI-NEXT: s_waitcnt vmcnt(0)
-; SI-NEXT: v_or_b32_e32 v0, v0, v29
-; SI-NEXT: buffer_store_dword v0, off, s[0:3], s32 offset:488 ; 4-byte Folded Spill
-; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:680 ; 4-byte Folded Reload
-; SI-NEXT: s_waitcnt vmcnt(0)
-; SI-NEXT: v_or_b32_e32 v20, v0, v20
-; SI-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:664 ; 4-byte Folded Reload
; SI-NEXT: s_waitcnt vmcnt(0)
; SI-NEXT: v_or_b32_e32 v16, v0, v16
-; SI-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:648 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:680 ; 4-byte Folded Reload
+; SI-NEXT: v_add_i32_e32 v16, vcc, s6, v16
; SI-NEXT: s_waitcnt vmcnt(0)
-; SI-NEXT: v_or_b32_e32 v0, v0, v34
-; SI-NEXT: buffer_store_dword v0, off, s[0:3], s32 offset:648 ; 4-byte Folded Spill
+; SI-NEXT: v_or_b32_e32 v0, v0, v20
+; SI-NEXT: buffer_store_dword v0, off, s[0:3], s32 offset:632 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:632 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:664 ; 4-byte Folded Reload
; SI-NEXT: s_waitcnt vmcnt(0)
-; SI-NEXT: v_or_b32_e32 v0, v0, v30
-; SI-NEXT: buffer_store_dword v0, off, s[0:3], s32 offset:632 ; 4-byte Folded Spill
+; SI-NEXT: v_or_b32_e32 v0, v0, v29
+; SI-NEXT: buffer_store_dword v0, off, s[0:3], s32 offset:636 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:612 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:644 ; 4-byte Folded Reload
+; SI-NEXT: v_add_i32_e32 v29, vcc, s6, v17
; SI-NEXT: s_waitcnt vmcnt(0)
; SI-NEXT: v_or_b32_e32 v0, v0, v28
-; SI-NEXT: buffer_store_dword v0, off, s[0:3], s32 offset:612 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v0, off, s[0:3], s32 offset:640 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:592 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:624 ; 4-byte Folded Reload
+; SI-NEXT: v_add_i32_e32 v28, vcc, s6, v15
; SI-NEXT: s_waitcnt vmcnt(0)
; SI-NEXT: v_or_b32_e32 v0, v0, v26
-; SI-NEXT: buffer_store_dword v0, off, s[0:3], s32 offset:592 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v0, off, s[0:3], s32 offset:624 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:572 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:604 ; 4-byte Folded Reload
; SI-NEXT: s_waitcnt vmcnt(0)
; SI-NEXT: v_or_b32_e32 v0, v0, v24
-; SI-NEXT: buffer_store_dword v0, off, s[0:3], s32 offset:572 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v0, off, s[0:3], s32 offset:604 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:556 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:592 ; 4-byte Folded Reload
; SI-NEXT: s_waitcnt vmcnt(0)
; SI-NEXT: v_or_b32_e32 v0, v0, v22
-; SI-NEXT: buffer_store_dword v0, off, s[0:3], s32 offset:556 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v0, off, s[0:3], s32 offset:592 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:548 ; 4-byte Folded Reload
-; SI-NEXT: v_add_i32_e32 v22, vcc, s6, v20
-; SI-NEXT: v_add_i32_e32 v20, vcc, s6, v16
+; SI-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:588 ; 4-byte Folded Reload
; SI-NEXT: s_waitcnt vmcnt(0)
; SI-NEXT: v_or_b32_e32 v0, v0, v18
-; SI-NEXT: buffer_store_dword v0, off, s[0:3], s32 offset:548 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v0, off, s[0:3], s32 offset:588 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:892 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:924 ; 4-byte Folded Reload
+; SI-NEXT: v_add_i32_e32 v18, vcc, s6, v2
; SI-NEXT: s_waitcnt vmcnt(0)
-; SI-NEXT: v_add_i32_e32 v36, vcc, 3, v0
-; SI-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:544 ; 4-byte Folded Reload
-; SI-NEXT: v_and_b32_e32 v36, 0xff, v36
+; SI-NEXT: v_add_i32_e32 v30, vcc, 3, v0
+; SI-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:584 ; 4-byte Folded Reload
+; SI-NEXT: v_and_b32_e32 v30, 0xff, v30
; SI-NEXT: s_waitcnt vmcnt(0)
-; SI-NEXT: v_or_b32_e32 v0, v0, v36
-; SI-NEXT: buffer_store_dword v0, off, s[0:3], s32 offset:544 ; 4-byte Folded Spill
+; SI-NEXT: v_or_b32_e32 v0, v0, v30
+; SI-NEXT: buffer_store_dword v0, off, s[0:3], s32 offset:584 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:888 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:920 ; 4-byte Folded Reload
+; SI-NEXT: v_add_i32_e32 v30, vcc, s6, v19
; SI-NEXT: s_waitcnt vmcnt(0)
-; SI-NEXT: v_add_i32_e32 v49, vcc, 3, v0
-; SI-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:540 ; 4-byte Folded Reload
-; SI-NEXT: v_and_b32_e32 v49, 0xff, v49
+; SI-NEXT: v_add_i32_e32 v34, vcc, 3, v0
+; SI-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:580 ; 4-byte Folded Reload
+; SI-NEXT: v_and_b32_e32 v34, 0xff, v34
; SI-NEXT: s_waitcnt vmcnt(0)
-; SI-NEXT: v_or_b32_e32 v0, v0, v49
-; SI-NEXT: buffer_store_dword v0, off, s[0:3], s32 offset:540 ; 4-byte Folded Spill
+; SI-NEXT: v_or_b32_e32 v0, v0, v34
+; SI-NEXT: buffer_store_dword v0, off, s[0:3], s32 offset:580 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:884 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:916 ; 4-byte Folded Reload
; SI-NEXT: s_waitcnt vmcnt(0)
-; SI-NEXT: v_add_i32_e32 v50, vcc, 3, v0
-; SI-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:536 ; 4-byte Folded Reload
-; SI-NEXT: v_and_b32_e32 v50, 0xff, v50
+; SI-NEXT: v_add_i32_e32 v36, vcc, 3, v0
+; SI-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:576 ; 4-byte Folded Reload
+; SI-NEXT: v_and_b32_e32 v36, 0xff, v36
; SI-NEXT: s_waitcnt vmcnt(0)
-; SI-NEXT: v_or_b32_e32 v0, v0, v50
-; SI-NEXT: buffer_store_dword v0, off, s[0:3], s32 offset:536 ; 4-byte Folded Spill
+; SI-NEXT: v_or_b32_e32 v0, v0, v36
+; SI-NEXT: buffer_store_dword v0, off, s[0:3], s32 offset:576 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:880 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:912 ; 4-byte Folded Reload
; SI-NEXT: s_waitcnt vmcnt(0)
; SI-NEXT: v_add_i32_e32 v52, vcc, 3, v0
-; SI-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:532 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:572 ; 4-byte Folded Reload
; SI-NEXT: v_and_b32_e32 v52, 0xff, v52
; SI-NEXT: s_waitcnt vmcnt(0)
; SI-NEXT: v_or_b32_e32 v0, v0, v52
-; SI-NEXT: buffer_store_dword v0, off, s[0:3], s32 offset:532 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v0, off, s[0:3], s32 offset:572 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:876 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:908 ; 4-byte Folded Reload
+; SI-NEXT: v_add_i32_e32 v52, vcc, s6, v54
; SI-NEXT: s_waitcnt vmcnt(0)
; SI-NEXT: v_add_i32_e32 v53, vcc, 3, v0
-; SI-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:872 ; 4-byte Folded Reload
; SI-NEXT: v_and_b32_e32 v53, 0xff, v53
-; SI-NEXT: v_or_b32_e32 v7, v7, v53
-; SI-NEXT: v_add_i32_e32 v54, vcc, s6, v7
-; SI-NEXT: v_add_i32_e32 v7, vcc, s6, v14
-; SI-NEXT: s_waitcnt vmcnt(0)
+; SI-NEXT: v_or_b32_e32 v0, v7, v53
+; SI-NEXT: buffer_store_dword v0, off, s[0:3], s32 offset:644 ; 4-byte Folded Spill
+; SI-NEXT: s_waitcnt expcnt(0)
+; SI-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:904 ; 4-byte Folded Reload
+; SI-NEXT: v_add_i32_e32 v53, vcc, s6, v48
+; SI-NEXT: v_add_i32_e32 v48, vcc, s6, v55
+; SI-NEXT: v_cvt_f32_f16_e32 v48, v48
+; SI-NEXT: buffer_load_dword v7, off, s[0:3], s32 offset:572 ; 4-byte Folded Reload
+; SI-NEXT: s_waitcnt vmcnt(1)
; SI-NEXT: v_add_i32_e32 v40, vcc, 3, v0
-; SI-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:528 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:568 ; 4-byte Folded Reload
; SI-NEXT: v_and_b32_e32 v40, 0xff, v40
; SI-NEXT: s_waitcnt vmcnt(0)
; SI-NEXT: v_or_b32_e32 v0, v0, v40
-; SI-NEXT: buffer_store_dword v0, off, s[0:3], s32 offset:528 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v0, off, s[0:3], s32 offset:568 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:524 ; 4-byte Folded Reload
-; SI-NEXT: s_waitcnt vmcnt(0)
+; SI-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:564 ; 4-byte Folded Reload
+; SI-NEXT: v_add_i32_e32 v40, vcc, s6, v39
+; SI-NEXT: v_add_i32_e32 v39, vcc, s6, v41
+; SI-NEXT: v_add_i32_e32 v41, vcc, s6, v7
+; SI-NEXT: buffer_load_dword v7, off, s[0:3], s32 offset:644 ; 4-byte Folded Reload
+; SI-NEXT: s_waitcnt vmcnt(1)
; SI-NEXT: v_or_b32_e32 v0, v0, v4
-; SI-NEXT: buffer_store_dword v0, off, s[0:3], s32 offset:524 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v0, off, s[0:3], s32 offset:564 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:472 ; 4-byte Folded Reload
-; SI-NEXT: s_waitcnt vmcnt(0)
+; SI-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:508 ; 4-byte Folded Reload
+; SI-NEXT: s_waitcnt vmcnt(2)
+; SI-NEXT: v_add_i32_e32 v55, vcc, s6, v7
+; SI-NEXT: buffer_load_dword v7, off, s[0:3], s32 offset:568 ; 4-byte Folded Reload
+; SI-NEXT: s_waitcnt vmcnt(1)
; SI-NEXT: v_add_i32_e32 v42, vcc, 3, v0
-; SI-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:520 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:560 ; 4-byte Folded Reload
; SI-NEXT: v_and_b32_e32 v42, 0xff, v42
-; SI-NEXT: s_waitcnt vmcnt(0)
+; SI-NEXT: s_waitcnt vmcnt(1)
+; SI-NEXT: v_add_i32_e32 v54, vcc, s6, v7
+; SI-NEXT: buffer_load_dword v7, off, s[0:3], s32 offset:564 ; 4-byte Folded Reload
+; SI-NEXT: s_waitcnt vmcnt(1)
; SI-NEXT: v_or_b32_e32 v0, v0, v42
-; SI-NEXT: buffer_store_dword v0, off, s[0:3], s32 offset:472 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v0, off, s[0:3], s32 offset:508 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:468 ; 4-byte Folded Reload
-; SI-NEXT: s_waitcnt vmcnt(0)
+; SI-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:504 ; 4-byte Folded Reload
+; SI-NEXT: v_add_i32_e32 v42, vcc, s6, v38
+; SI-NEXT: s_waitcnt vmcnt(2)
+; SI-NEXT: v_add_i32_e32 v51, vcc, s6, v7
+; SI-NEXT: buffer_load_dword v7, off, s[0:3], s32 offset:508 ; 4-byte Folded Reload
+; SI-NEXT: s_waitcnt vmcnt(1)
; SI-NEXT: v_add_i32_e32 v43, vcc, 3, v0
-; SI-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:516 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:556 ; 4-byte Folded Reload
; SI-NEXT: v_and_b32_e32 v43, 0xff, v43
; SI-NEXT: s_waitcnt vmcnt(0)
; SI-NEXT: v_or_b32_e32 v0, v0, v43
-; SI-NEXT: buffer_store_dword v0, off, s[0:3], s32 offset:468 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v0, off, s[0:3], s32 offset:504 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:464 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:500 ; 4-byte Folded Reload
+; SI-NEXT: v_add_i32_e32 v43, vcc, s6, v37
; SI-NEXT: s_waitcnt vmcnt(0)
; SI-NEXT: v_add_i32_e32 v44, vcc, 3, v0
-; SI-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:512 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:552 ; 4-byte Folded Reload
; SI-NEXT: v_and_b32_e32 v44, 0xff, v44
; SI-NEXT: s_waitcnt vmcnt(0)
; SI-NEXT: v_or_b32_e32 v0, v0, v44
-; SI-NEXT: buffer_store_dword v0, off, s[0:3], s32 offset:464 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v0, off, s[0:3], s32 offset:500 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:460 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:496 ; 4-byte Folded Reload
+; SI-NEXT: v_add_i32_e32 v44, vcc, s6, v35
; SI-NEXT: s_waitcnt vmcnt(0)
; SI-NEXT: v_add_i32_e32 v45, vcc, 3, v0
-; SI-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:508 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:548 ; 4-byte Folded Reload
; SI-NEXT: v_and_b32_e32 v45, 0xff, v45
; SI-NEXT: s_waitcnt vmcnt(0)
; SI-NEXT: v_or_b32_e32 v0, v0, v45
-; SI-NEXT: buffer_store_dword v0, off, s[0:3], s32 offset:508 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v0, off, s[0:3], s32 offset:496 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:504 ; 4-byte Folded Reload
-; SI-NEXT: v_add_i32_e32 v45, vcc, s6, v48
+; SI-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:492 ; 4-byte Folded Reload
+; SI-NEXT: v_add_i32_e32 v45, vcc, s6, v33
+; SI-NEXT: s_waitcnt vmcnt(0)
+; SI-NEXT: v_add_i32_e32 v46, vcc, 3, v0
+; SI-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:544 ; 4-byte Folded Reload
+; SI-NEXT: v_and_b32_e32 v46, 0xff, v46
; SI-NEXT: s_waitcnt vmcnt(0)
; SI-NEXT: v_or_b32_e32 v0, v0, v46
-; SI-NEXT: buffer_store_dword v0, off, s[0:3], s32 offset:504 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v0, off, s[0:3], s32 offset:492 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:500 ; 4-byte Folded Reload
-; SI-NEXT: v_add_i32_e32 v46, vcc, s6, v39
-; SI-NEXT: s_waitcnt vmcnt(0)
+; SI-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:488 ; 4-byte Folded Reload
+; SI-NEXT: v_add_i32_e32 v46, vcc, s6, v32
+; SI-NEXT: v_add_i32_e32 v32, vcc, s6, v50
+; SI-NEXT: v_add_i32_e32 v50, vcc, s6, v7
+; SI-NEXT: buffer_load_dword v7, off, s[0:3], s32 offset:504 ; 4-byte Folded Reload
+; SI-NEXT: s_waitcnt vmcnt(1)
+; SI-NEXT: v_add_i32_e32 v47, vcc, 3, v0
+; SI-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:540 ; 4-byte Folded Reload
+; SI-NEXT: v_and_b32_e32 v47, 0xff, v47
+; SI-NEXT: s_waitcnt vmcnt(1)
+; SI-NEXT: v_add_i32_e32 v49, vcc, s6, v7
+; SI-NEXT: buffer_load_dword v7, off, s[0:3], s32 offset:500 ; 4-byte Folded Reload
+; SI-NEXT: s_waitcnt vmcnt(1)
; SI-NEXT: v_or_b32_e32 v0, v0, v47
-; SI-NEXT: buffer_store_dword v0, off, s[0:3], s32 offset:500 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v0, off, s[0:3], s32 offset:488 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:496 ; 4-byte Folded Reload
-; SI-NEXT: v_add_i32_e32 v47, vcc, s6, v38
+; SI-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:484 ; 4-byte Folded Reload
+; SI-NEXT: v_add_i32_e32 v47, vcc, s6, v31
+; SI-NEXT: s_waitcnt vmcnt(0)
+; SI-NEXT: v_add_i32_e32 v56, vcc, 3, v0
+; SI-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:536 ; 4-byte Folded Reload
+; SI-NEXT: v_and_b32_e32 v56, 0xff, v56
; SI-NEXT: s_waitcnt vmcnt(0)
; SI-NEXT: v_or_b32_e32 v0, v0, v56
-; SI-NEXT: buffer_store_dword v0, off, s[0:3], s32 offset:496 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v0, off, s[0:3], s32 offset:484 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:492 ; 4-byte Folded Reload
-; SI-NEXT: v_add_i32_e32 v56, vcc, s6, v37
+; SI-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:480 ; 4-byte Folded Reload
+; SI-NEXT: v_add_i32_e32 v56, vcc, s6, v27
+; SI-NEXT: s_waitcnt vmcnt(0)
+; SI-NEXT: v_add_i32_e32 v57, vcc, 3, v0
+; SI-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:532 ; 4-byte Folded Reload
+; SI-NEXT: v_and_b32_e32 v57, 0xff, v57
; SI-NEXT: s_waitcnt vmcnt(0)
; SI-NEXT: v_or_b32_e32 v0, v0, v57
-; SI-NEXT: buffer_store_dword v0, off, s[0:3], s32 offset:492 ; 4-byte Folded Spill
-; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_or_b32_e32 v0, v8, v61
-; SI-NEXT: buffer_store_dword v0, off, s[0:3], s32 offset:512 ; 4-byte Folded Spill
-; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_or_b32_e32 v0, v6, v62
-; SI-NEXT: buffer_store_dword v0, off, s[0:3], s32 offset:516 ; 4-byte Folded Spill
+; SI-NEXT: v_add_i32_e32 v57, vcc, s6, v25
+; SI-NEXT: v_add_i32_e32 v25, vcc, s6, v7
+; SI-NEXT: buffer_load_dword v7, off, s[0:3], s32 offset:496 ; 4-byte Folded Reload
+; SI-NEXT: buffer_store_dword v0, off, s[0:3], s32 offset:480 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:484 ; 4-byte Folded Reload
-; SI-NEXT: v_add_i32_e32 v57, vcc, s6, v35
-; SI-NEXT: v_add_i32_e32 v61, vcc, s6, v23
-; SI-NEXT: v_add_i32_e32 v62, vcc, s6, v1
-; SI-NEXT: s_waitcnt vmcnt(0)
-; SI-NEXT: v_or_b32_e32 v63, v0, v63
; SI-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:476 ; 4-byte Folded Reload
+; SI-NEXT: s_waitcnt vmcnt(2)
+; SI-NEXT: v_add_i32_e32 v27, vcc, s6, v7
+; SI-NEXT: buffer_load_dword v7, off, s[0:3], s32 offset:492 ; 4-byte Folded Reload
; SI-NEXT: s_waitcnt vmcnt(0)
-; SI-NEXT: v_or_b32_e32 v0, v0, v3
+; SI-NEXT: v_add_i32_e32 v31, vcc, s6, v7
+; SI-NEXT: buffer_load_dword v7, off, s[0:3], s32 offset:488 ; 4-byte Folded Reload
+; SI-NEXT: v_add_i32_e32 v58, vcc, 3, v0
+; SI-NEXT: v_and_b32_e32 v58, 0xff, v58
+; SI-NEXT: v_or_b32_e32 v0, v14, v58
; SI-NEXT: buffer_store_dword v0, off, s[0:3], s32 offset:476 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_add_i32_e32 v0, vcc, 0x300, v5
-; SI-NEXT: buffer_load_dword v5, off, s[0:3], s32 offset:536 ; 4-byte Folded Reload
-; SI-NEXT: buffer_store_dword v0, off, s[0:3], s32 offset:460 ; 4-byte Folded Spill
+; SI-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:472 ; 4-byte Folded Reload
+; SI-NEXT: v_add_i32_e32 v58, vcc, s6, v23
+; SI-NEXT: s_waitcnt vmcnt(2)
+; SI-NEXT: v_add_i32_e32 v23, vcc, s6, v7
+; SI-NEXT: buffer_load_dword v7, off, s[0:3], s32 offset:484 ; 4-byte Folded Reload
+; SI-NEXT: s_waitcnt vmcnt(1)
+; SI-NEXT: v_add_i32_e32 v59, vcc, 3, v0
+; SI-NEXT: v_and_b32_e32 v59, 0xff, v59
+; SI-NEXT: v_or_b32_e32 v0, v12, v59
+; SI-NEXT: v_add_i32_e32 v59, vcc, s6, v21
+; SI-NEXT: buffer_store_dword v0, off, s[0:3], s32 offset:472 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:552 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:468 ; 4-byte Folded Reload
; SI-NEXT: s_waitcnt vmcnt(2)
-; SI-NEXT: v_add_i32_e32 v41, vcc, s6, v5
-; SI-NEXT: buffer_load_dword v5, off, s[0:3], s32 offset:532 ; 4-byte Folded Reload
-; SI-NEXT: s_waitcnt vmcnt(0)
-; SI-NEXT: v_add_i32_e32 v55, vcc, s6, v5
-; SI-NEXT: buffer_load_dword v5, off, s[0:3], s32 offset:528 ; 4-byte Folded Reload
+; SI-NEXT: v_add_i32_e32 v21, vcc, s6, v7
+; SI-NEXT: buffer_load_dword v7, off, s[0:3], s32 offset:480 ; 4-byte Folded Reload
; SI-NEXT: s_waitcnt vmcnt(0)
-; SI-NEXT: v_add_i32_e32 v51, vcc, s6, v5
-; SI-NEXT: buffer_load_dword v5, off, s[0:3], s32 offset:524 ; 4-byte Folded Reload
+; SI-NEXT: v_add_i32_e32 v19, vcc, s6, v7
+; SI-NEXT: buffer_load_dword v7, off, s[0:3], s32 offset:476 ; 4-byte Folded Reload
; SI-NEXT: s_waitcnt vmcnt(0)
-; SI-NEXT: v_add_i32_e32 v25, vcc, s6, v5
-; SI-NEXT: buffer_load_dword v5, off, s[0:3], s32 offset:472 ; 4-byte Folded Reload
-; SI-NEXT: s_waitcnt vmcnt(0)
-; SI-NEXT: v_add_i32_e32 v31, vcc, s6, v5
-; SI-NEXT: buffer_load_dword v5, off, s[0:3], s32 offset:468 ; 4-byte Folded Reload
-; SI-NEXT: v_add_i32_e32 v44, vcc, s6, v0
-; SI-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:560 ; 4-byte Folded Reload
+; SI-NEXT: v_add_i32_e32 v17, vcc, s6, v7
+; SI-NEXT: buffer_load_dword v7, off, s[0:3], s32 offset:472 ; 4-byte Folded Reload
+; SI-NEXT: v_add_i32_e32 v60, vcc, 3, v0
+; SI-NEXT: v_and_b32_e32 v60, 0xff, v60
+; SI-NEXT: v_or_b32_e32 v0, v10, v60
+; SI-NEXT: buffer_store_dword v0, off, s[0:3], s32 offset:468 ; 4-byte Folded Spill
+; SI-NEXT: s_waitcnt expcnt(0)
+; SI-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:464 ; 4-byte Folded Reload
+; SI-NEXT: v_add_i32_e32 v60, vcc, s6, v13
+; SI-NEXT: s_waitcnt vmcnt(2)
+; SI-NEXT: v_add_i32_e32 v15, vcc, s6, v7
+; SI-NEXT: buffer_load_dword v7, off, s[0:3], s32 offset:468 ; 4-byte Folded Reload
; SI-NEXT: s_waitcnt vmcnt(1)
-; SI-NEXT: v_add_i32_e32 v23, vcc, s6, v5
-; SI-NEXT: buffer_load_dword v5, off, s[0:3], s32 offset:464 ; 4-byte Folded Reload
+; SI-NEXT: v_add_i32_e32 v61, vcc, 3, v0
+; SI-NEXT: v_and_b32_e32 v61, 0xff, v61
+; SI-NEXT: v_or_b32_e32 v0, v8, v61
+; SI-NEXT: buffer_store_dword v0, off, s[0:3], s32 offset:464 ; 4-byte Folded Spill
+; SI-NEXT: s_waitcnt expcnt(0)
+; SI-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:460 ; 4-byte Folded Reload
+; SI-NEXT: v_add_i32_e32 v61, vcc, s6, v11
+; SI-NEXT: s_waitcnt vmcnt(2)
+; SI-NEXT: v_add_i32_e32 v13, vcc, s6, v7
+; SI-NEXT: buffer_load_dword v7, off, s[0:3], s32 offset:464 ; 4-byte Folded Reload
; SI-NEXT: s_waitcnt vmcnt(1)
-; SI-NEXT: v_add_i32_e32 v43, vcc, s6, v0
-; SI-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:564 ; 4-byte Folded Reload
-; SI-NEXT: s_waitcnt vmcnt(0)
-; SI-NEXT: v_add_i32_e32 v42, vcc, s6, v0
-; SI-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:568 ; 4-byte Folded Reload
-; SI-NEXT: s_waitcnt vmcnt(0)
-; SI-NEXT: v_add_i32_e32 v40, vcc, s6, v0
-; SI-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:576 ; 4-byte Folded Reload
-; SI-NEXT: s_waitcnt vmcnt(0)
-; SI-NEXT: v_add_i32_e32 v53, vcc, s6, v0
-; SI-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:580 ; 4-byte Folded Reload
-; SI-NEXT: s_waitcnt vmcnt(0)
-; SI-NEXT: v_add_i32_e32 v52, vcc, s6, v0
-; SI-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:588 ; 4-byte Folded Reload
+; SI-NEXT: v_add_i32_e32 v62, vcc, 3, v0
+; SI-NEXT: v_and_b32_e32 v62, 0xff, v62
+; SI-NEXT: v_or_b32_e32 v0, v6, v62
+; SI-NEXT: buffer_store_dword v0, off, s[0:3], s32 offset:532 ; 4-byte Folded Spill
+; SI-NEXT: s_waitcnt expcnt(0)
+; SI-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:524 ; 4-byte Folded Reload
+; SI-NEXT: v_add_i32_e32 v62, vcc, s6, v1
+; SI-NEXT: s_waitcnt vmcnt(2)
+; SI-NEXT: v_add_i32_e32 v11, vcc, s6, v7
+; SI-NEXT: buffer_load_dword v7, off, s[0:3], s32 offset:532 ; 4-byte Folded Reload
+; SI-NEXT: s_waitcnt vmcnt(1)
+; SI-NEXT: v_or_b32_e32 v63, v0, v63
+; SI-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:516 ; 4-byte Folded Reload
; SI-NEXT: s_waitcnt vmcnt(0)
-; SI-NEXT: v_add_i32_e32 v50, vcc, s6, v0
+; SI-NEXT: v_or_b32_e32 v0, v0, v3
+; SI-NEXT: buffer_store_dword v0, off, s[0:3], s32 offset:516 ; 4-byte Folded Spill
+; SI-NEXT: s_waitcnt expcnt(0)
+; SI-NEXT: v_add_i32_e32 v0, vcc, 0x300, v9
+; SI-NEXT: buffer_store_dword v0, off, s[0:3], s32 offset:460 ; 4-byte Folded Spill
+; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:596 ; 4-byte Folded Reload
-; SI-NEXT: s_waitcnt vmcnt(0)
-; SI-NEXT: v_add_i32_e32 v49, vcc, s6, v0
-; SI-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:600 ; 4-byte Folded Reload
-; SI-NEXT: s_waitcnt vmcnt(0)
-; SI-NEXT: v_add_i32_e32 v48, vcc, s6, v0
-; SI-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:608 ; 4-byte Folded Reload
-; SI-NEXT: s_waitcnt vmcnt(0)
-; SI-NEXT: v_add_i32_e32 v39, vcc, s6, v0
-; SI-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:616 ; 4-byte Folded Reload
-; SI-NEXT: s_waitcnt vmcnt(0)
+; SI-NEXT: v_add_i32_e32 v9, vcc, s6, v7
+; SI-NEXT: v_add_i32_e32 v7, vcc, s6, v63
+; SI-NEXT: v_cvt_f32_f16_e32 v7, v7
+; SI-NEXT: buffer_load_dword v63, off, s[0:3], s32 offset:516 ; 4-byte Folded Reload
+; SI-NEXT: buffer_store_dword v7, off, s[0:3], s32 offset:872 ; 4-byte Folded Spill
+; SI-NEXT: s_waitcnt expcnt(0)
+; SI-NEXT: v_cvt_f32_f16_e32 v7, v9
+; SI-NEXT: v_cvt_f32_f16_e32 v9, v62
+; SI-NEXT: buffer_store_dword v7, off, s[0:3], s32 offset:852 ; 4-byte Folded Spill
+; SI-NEXT: s_waitcnt expcnt(0)
+; SI-NEXT: v_cvt_f32_f16_e32 v7, v11
+; SI-NEXT: v_cvt_f32_f16_e32 v11, v60
+; SI-NEXT: buffer_store_dword v7, off, s[0:3], s32 offset:868 ; 4-byte Folded Spill
+; SI-NEXT: s_waitcnt expcnt(0)
+; SI-NEXT: v_cvt_f32_f16_e32 v7, v13
+; SI-NEXT: v_cvt_f32_f16_e32 v13, v58
+; SI-NEXT: buffer_store_dword v7, off, s[0:3], s32 offset:848 ; 4-byte Folded Spill
+; SI-NEXT: s_waitcnt expcnt(0)
+; SI-NEXT: v_cvt_f32_f16_e32 v7, v15
+; SI-NEXT: buffer_store_dword v7, off, s[0:3], s32 offset:864 ; 4-byte Folded Spill
+; SI-NEXT: s_waitcnt expcnt(0)
+; SI-NEXT: v_cvt_f32_f16_e32 v7, v17
+; SI-NEXT: buffer_store_dword v7, off, s[0:3], s32 offset:836 ; 4-byte Folded Spill
+; SI-NEXT: s_waitcnt expcnt(0)
+; SI-NEXT: v_cvt_f32_f16_e32 v7, v19
+; SI-NEXT: buffer_store_dword v7, off, s[0:3], s32 offset:860 ; 4-byte Folded Spill
+; SI-NEXT: s_waitcnt expcnt(0)
+; SI-NEXT: v_cvt_f32_f16_e32 v7, v21
+; SI-NEXT: buffer_store_dword v7, off, s[0:3], s32 offset:824 ; 4-byte Folded Spill
+; SI-NEXT: s_waitcnt expcnt(0)
+; SI-NEXT: v_cvt_f32_f16_e32 v7, v23
+; SI-NEXT: v_cvt_f32_f16_e32 v23, v56
+; SI-NEXT: buffer_store_dword v7, off, s[0:3], s32 offset:840 ; 4-byte Folded Spill
+; SI-NEXT: s_waitcnt expcnt(0)
+; SI-NEXT: v_cvt_f32_f16_e32 v7, v31
+; SI-NEXT: v_cvt_f32_f16_e32 v31, v59
+; SI-NEXT: buffer_store_dword v7, off, s[0:3], s32 offset:808 ; 4-byte Folded Spill
+; SI-NEXT: s_waitcnt expcnt(0)
+; SI-NEXT: v_cvt_f32_f16_e32 v7, v27
+; SI-NEXT: buffer_store_dword v7, off, s[0:3], s32 offset:832 ; 4-byte Folded Spill
+; SI-NEXT: s_waitcnt expcnt(0)
+; SI-NEXT: v_cvt_f32_f16_e32 v7, v25
+; SI-NEXT: v_cvt_f32_f16_e32 v25, v61
+; SI-NEXT: buffer_store_dword v7, off, s[0:3], s32 offset:792 ; 4-byte Folded Spill
+; SI-NEXT: s_waitcnt expcnt(0)
+; SI-NEXT: v_cvt_f32_f16_e32 v7, v49
+; SI-NEXT: buffer_store_dword v7, off, s[0:3], s32 offset:816 ; 4-byte Folded Spill
+; SI-NEXT: s_waitcnt expcnt(0)
+; SI-NEXT: v_cvt_f32_f16_e32 v7, v50
+; SI-NEXT: buffer_store_dword v7, off, s[0:3], s32 offset:772 ; 4-byte Folded Spill
+; SI-NEXT: s_waitcnt expcnt(0)
+; SI-NEXT: v_cvt_f32_f16_e32 v7, v51
+; SI-NEXT: buffer_store_dword v7, off, s[0:3], s32 offset:796 ; 4-byte Folded Spill
+; SI-NEXT: s_waitcnt vmcnt(14)
; SI-NEXT: v_add_i32_e32 v38, vcc, s6, v0
-; SI-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:620 ; 4-byte Folded Reload
-; SI-NEXT: s_waitcnt vmcnt(0)
+; SI-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:600 ; 4-byte Folded Reload
+; SI-NEXT: s_waitcnt expcnt(0)
+; SI-NEXT: v_cvt_f32_f16_e32 v7, v54
+; SI-NEXT: buffer_store_dword v7, off, s[0:3], s32 offset:764 ; 4-byte Folded Spill
+; SI-NEXT: s_waitcnt expcnt(0)
+; SI-NEXT: v_cvt_f32_f16_e32 v7, v55
+; SI-NEXT: buffer_store_dword v7, off, s[0:3], s32 offset:788 ; 4-byte Folded Spill
+; SI-NEXT: s_waitcnt expcnt(0)
+; SI-NEXT: v_cvt_f32_f16_e32 v7, v41
+; SI-NEXT: buffer_store_dword v7, off, s[0:3], s32 offset:752 ; 4-byte Folded Spill
+; SI-NEXT: s_waitcnt vmcnt(3)
; SI-NEXT: v_add_i32_e32 v37, vcc, s6, v0
-; SI-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:604 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:608 ; 4-byte Folded Reload
; SI-NEXT: s_waitcnt vmcnt(0)
; SI-NEXT: v_add_i32_e32 v36, vcc, s6, v0
-; SI-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:584 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:612 ; 4-byte Folded Reload
; SI-NEXT: s_waitcnt vmcnt(0)
; SI-NEXT: v_add_i32_e32 v35, vcc, s6, v0
-; SI-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:624 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:620 ; 4-byte Folded Reload
; SI-NEXT: s_waitcnt vmcnt(0)
; SI-NEXT: v_add_i32_e32 v34, vcc, s6, v0
-; SI-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:628 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:616 ; 4-byte Folded Reload
; SI-NEXT: s_waitcnt vmcnt(0)
; SI-NEXT: v_add_i32_e32 v33, vcc, s6, v0
-; SI-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:636 ; 4-byte Folded Reload
-; SI-NEXT: s_waitcnt vmcnt(0)
-; SI-NEXT: v_add_i32_e32 v30, vcc, s6, v0
-; SI-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:640 ; 4-byte Folded Reload
-; SI-NEXT: s_waitcnt vmcnt(0)
-; SI-NEXT: v_add_i32_e32 v29, vcc, s6, v0
-; SI-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:644 ; 4-byte Folded Reload
-; SI-NEXT: s_waitcnt vmcnt(0)
-; SI-NEXT: v_add_i32_e32 v28, vcc, s6, v0
-; SI-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:480 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:628 ; 4-byte Folded Reload
; SI-NEXT: s_waitcnt vmcnt(0)
; SI-NEXT: v_add_i32_e32 v26, vcc, s6, v0
-; SI-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:488 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:528 ; 4-byte Folded Reload
; SI-NEXT: s_waitcnt vmcnt(0)
; SI-NEXT: v_add_i32_e32 v24, vcc, s6, v0
-; SI-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:648 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:520 ; 4-byte Folded Reload
+; SI-NEXT: s_waitcnt vmcnt(0)
+; SI-NEXT: v_add_i32_e32 v22, vcc, s6, v0
+; SI-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:512 ; 4-byte Folded Reload
; SI-NEXT: s_waitcnt vmcnt(0)
-; SI-NEXT: v_add_i32_e32 v18, vcc, s6, v0
+; SI-NEXT: v_add_i32_e32 v20, vcc, s6, v0
; SI-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:632 ; 4-byte Folded Reload
; SI-NEXT: s_waitcnt vmcnt(0)
-; SI-NEXT: v_add_i32_e32 v16, vcc, s6, v0
-; SI-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:612 ; 4-byte Folded Reload
+; SI-NEXT: v_add_i32_e32 v14, vcc, s6, v0
+; SI-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:636 ; 4-byte Folded Reload
+; SI-NEXT: s_waitcnt vmcnt(0)
+; SI-NEXT: v_add_i32_e32 v12, vcc, s6, v0
+; SI-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:640 ; 4-byte Folded Reload
+; SI-NEXT: s_waitcnt vmcnt(0)
+; SI-NEXT: v_add_i32_e32 v10, vcc, s6, v0
+; SI-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:624 ; 4-byte Folded Reload
; SI-NEXT: s_waitcnt vmcnt(0)
; SI-NEXT: v_add_i32_e32 v8, vcc, s6, v0
-; SI-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:592 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:604 ; 4-byte Folded Reload
; SI-NEXT: s_waitcnt vmcnt(0)
; SI-NEXT: v_add_i32_e32 v6, vcc, s6, v0
-; SI-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:572 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:592 ; 4-byte Folded Reload
; SI-NEXT: s_waitcnt vmcnt(0)
; SI-NEXT: v_add_i32_e32 v4, vcc, s6, v0
-; SI-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:556 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:588 ; 4-byte Folded Reload
; SI-NEXT: s_waitcnt vmcnt(0)
; SI-NEXT: v_add_i32_e32 v3, vcc, s6, v0
-; SI-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:548 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:584 ; 4-byte Folded Reload
; SI-NEXT: s_waitcnt vmcnt(0)
; SI-NEXT: v_add_i32_e32 v2, vcc, s6, v0
-; SI-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:544 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:580 ; 4-byte Folded Reload
; SI-NEXT: s_waitcnt vmcnt(0)
; SI-NEXT: v_add_i32_e32 v1, vcc, s6, v0
-; SI-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:540 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:576 ; 4-byte Folded Reload
; SI-NEXT: s_waitcnt vmcnt(0)
; SI-NEXT: v_add_i32_e32 v0, vcc, s6, v0
; SI-NEXT: v_cvt_f32_f16_e32 v0, v0
-; SI-NEXT: buffer_store_dword v0, off, s[0:3], s32 offset:708 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v0, off, s[0:3], s32 offset:780 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_cvt_f32_f16_e32 v0, v1
-; SI-NEXT: buffer_store_dword v0, off, s[0:3], s32 offset:736 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v0, off, s[0:3], s32 offset:740 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_cvt_f32_f16_e32 v0, v2
-; SI-NEXT: buffer_store_dword v0, off, s[0:3], s32 offset:700 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v0, off, s[0:3], s32 offset:768 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_cvt_f32_f16_e32 v0, v3
-; SI-NEXT: buffer_store_dword v0, off, s[0:3], s32 offset:724 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v0, off, s[0:3], s32 offset:732 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_cvt_f32_f16_e32 v0, v4
-; SI-NEXT: buffer_store_dword v0, off, s[0:3], s32 offset:688 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v0, off, s[0:3], s32 offset:756 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_cvt_f32_f16_e32 v0, v6
-; SI-NEXT: buffer_store_dword v0, off, s[0:3], s32 offset:716 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v0, off, s[0:3], s32 offset:720 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_cvt_f32_f16_e32 v0, v8
-; SI-NEXT: buffer_store_dword v0, off, s[0:3], s32 offset:676 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v0, off, s[0:3], s32 offset:748 ; 4-byte Folded Spill
+; SI-NEXT: s_waitcnt expcnt(0)
+; SI-NEXT: v_cvt_f32_f16_e32 v0, v10
+; SI-NEXT: buffer_store_dword v0, off, s[0:3], s32 offset:708 ; 4-byte Folded Spill
+; SI-NEXT: s_waitcnt expcnt(0)
+; SI-NEXT: v_cvt_f32_f16_e32 v0, v12
+; SI-NEXT: buffer_store_dword v0, off, s[0:3], s32 offset:736 ; 4-byte Folded Spill
+; SI-NEXT: s_waitcnt expcnt(0)
+; SI-NEXT: v_cvt_f32_f16_e32 v0, v14
+; SI-NEXT: buffer_store_dword v0, off, s[0:3], s32 offset:700 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_cvt_f32_f16_e32 v0, v16
-; SI-NEXT: buffer_store_dword v0, off, s[0:3], s32 offset:704 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v0, off, s[0:3], s32 offset:724 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_cvt_f32_f16_e32 v0, v18
-; SI-NEXT: buffer_store_dword v0, off, s[0:3], s32 offset:668 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v0, off, s[0:3], s32 offset:688 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_cvt_f32_f16_e32 v0, v20
-; SI-NEXT: buffer_store_dword v0, off, s[0:3], s32 offset:692 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v0, off, s[0:3], s32 offset:716 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_cvt_f32_f16_e32 v0, v22
-; SI-NEXT: buffer_store_dword v0, off, s[0:3], s32 offset:656 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v0, off, s[0:3], s32 offset:676 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_cvt_f32_f16_e32 v0, v24
-; SI-NEXT: buffer_store_dword v0, off, s[0:3], s32 offset:684 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v0, off, s[0:3], s32 offset:704 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_cvt_f32_f16_e32 v0, v26
-; SI-NEXT: buffer_store_dword v0, off, s[0:3], s32 offset:644 ; 4-byte Folded Spill
-; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_cvt_f32_f16_e32 v0, v27
-; SI-NEXT: buffer_store_dword v0, off, s[0:3], s32 offset:672 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v0, off, s[0:3], s32 offset:668 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_cvt_f32_f16_e32 v0, v28
-; SI-NEXT: buffer_store_dword v0, off, s[0:3], s32 offset:636 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v0, off, s[0:3], s32 offset:692 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_cvt_f32_f16_e32 v0, v29
-; SI-NEXT: v_cvt_f32_f16_e32 v29, v56
-; SI-NEXT: buffer_store_dword v0, off, s[0:3], s32 offset:660 ; 4-byte Folded Spill
+; SI-NEXT: v_cvt_f32_f16_e32 v29, v46
+; SI-NEXT: buffer_store_dword v0, off, s[0:3], s32 offset:652 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_cvt_f32_f16_e32 v0, v30
-; SI-NEXT: buffer_store_dword v0, off, s[0:3], s32 offset:620 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v0, off, s[0:3], s32 offset:684 ; 4-byte Folded Spill
+; SI-NEXT: s_waitcnt expcnt(0)
+; SI-NEXT: v_cvt_f32_f16_e32 v0, v32
+; SI-NEXT: v_cvt_f32_f16_e32 v32, v44
+; SI-NEXT: buffer_store_dword v0, off, s[0:3], s32 offset:640 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_cvt_f32_f16_e32 v0, v33
-; SI-NEXT: v_cvt_f32_f16_e32 v33, v46
-; SI-NEXT: buffer_store_dword v0, off, s[0:3], s32 offset:652 ; 4-byte Folded Spill
+; SI-NEXT: v_cvt_f32_f16_e32 v33, v57
+; SI-NEXT: buffer_store_dword v0, off, s[0:3], s32 offset:672 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_cvt_f32_f16_e32 v0, v34
-; SI-NEXT: buffer_store_dword v0, off, s[0:3], s32 offset:608 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v0, off, s[0:3], s32 offset:628 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_cvt_f32_f16_e32 v0, v35
-; SI-NEXT: v_cvt_f32_f16_e32 v35, v59
-; SI-NEXT: buffer_store_dword v0, off, s[0:3], s32 offset:640 ; 4-byte Folded Spill
+; SI-NEXT: v_cvt_f32_f16_e32 v35, v42
+; SI-NEXT: buffer_store_dword v0, off, s[0:3], s32 offset:660 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_cvt_f32_f16_e32 v0, v36
-; SI-NEXT: v_cvt_f32_f16_e32 v36, v44
-; SI-NEXT: buffer_store_dword v0, off, s[0:3], s32 offset:596 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v0, off, s[0:3], s32 offset:612 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_cvt_f32_f16_e32 v0, v37
-; SI-NEXT: v_cvt_f32_f16_e32 v37, v57
-; SI-NEXT: buffer_store_dword v0, off, s[0:3], s32 offset:628 ; 4-byte Folded Spill
+; SI-NEXT: v_cvt_f32_f16_e32 v37, v47
+; SI-NEXT: buffer_store_dword v0, off, s[0:3], s32 offset:648 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_cvt_f32_f16_e32 v0, v38
-; SI-NEXT: v_cvt_f32_f16_e32 v38, v42
-; SI-NEXT: buffer_store_dword v0, off, s[0:3], s32 offset:580 ; 4-byte Folded Spill
-; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_cvt_f32_f16_e32 v0, v39
-; SI-NEXT: v_cvt_f32_f16_e32 v39, v47
-; SI-NEXT: buffer_store_dword v0, off, s[0:3], s32 offset:616 ; 4-byte Folded Spill
-; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_cvt_f32_f16_e32 v0, v48
-; SI-NEXT: v_cvt_f32_f16_e32 v48, v53
-; SI-NEXT: buffer_store_dword v0, off, s[0:3], s32 offset:568 ; 4-byte Folded Spill
-; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_cvt_f32_f16_e32 v0, v49
+; SI-NEXT: v_cvt_f32_f16_e32 v38, v53
; SI-NEXT: buffer_store_dword v0, off, s[0:3], s32 offset:600 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_cvt_f32_f16_e32 v0, v50
-; SI-NEXT: buffer_store_dword v0, off, s[0:3], s32 offset:560 ; 4-byte Folded Spill
+; SI-NEXT: v_cvt_f32_f16_e32 v0, v39
+; SI-NEXT: v_cvt_f32_f16_e32 v39, v45
+; SI-NEXT: buffer_store_dword v0, off, s[0:3], s32 offset:632 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_cvt_f32_f16_e32 v0, v52
-; SI-NEXT: buffer_store_dword v0, off, s[0:3], s32 offset:588 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v0, off, s[0:3], s32 offset:620 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_cvt_f32_f16_e32 v0, v40
-; SI-NEXT: buffer_store_dword v0, off, s[0:3], s32 offset:576 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v0, off, s[0:3], s32 offset:608 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_cvt_f32_f16_e32 v0, v43
-; SI-NEXT: buffer_store_dword v0, off, s[0:3], s32 offset:564 ; 4-byte Folded Spill
-; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_cvt_f32_f16_e32 v0, v45
-; SI-NEXT: buffer_store_dword v0, off, s[0:3], s32 offset:552 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v0, off, s[0:3], s32 offset:596 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:460 ; 4-byte Folded Reload
-; SI-NEXT: v_add_i32_e32 v21, vcc, s6, v5
-; SI-NEXT: buffer_load_dword v5, off, s[0:3], s32 offset:508 ; 4-byte Folded Reload
-; SI-NEXT: s_waitcnt vmcnt(1)
-; SI-NEXT: v_cvt_f32_f16_e32 v27, v0
-; SI-NEXT: s_waitcnt vmcnt(0)
-; SI-NEXT: v_add_i32_e32 v19, vcc, s6, v5
-; SI-NEXT: buffer_load_dword v5, off, s[0:3], s32 offset:504 ; 4-byte Folded Reload
-; SI-NEXT: s_waitcnt vmcnt(0)
-; SI-NEXT: v_add_i32_e32 v17, vcc, s6, v5
-; SI-NEXT: buffer_load_dword v5, off, s[0:3], s32 offset:500 ; 4-byte Folded Reload
-; SI-NEXT: s_waitcnt vmcnt(0)
-; SI-NEXT: v_add_i32_e32 v15, vcc, s6, v5
-; SI-NEXT: buffer_load_dword v5, off, s[0:3], s32 offset:496 ; 4-byte Folded Reload
-; SI-NEXT: s_waitcnt vmcnt(0)
-; SI-NEXT: v_add_i32_e32 v13, vcc, s6, v5
-; SI-NEXT: buffer_load_dword v5, off, s[0:3], s32 offset:492 ; 4-byte Folded Reload
-; SI-NEXT: s_waitcnt vmcnt(0)
-; SI-NEXT: v_add_i32_e32 v11, vcc, s6, v5
-; SI-NEXT: v_add_i32_e32 v5, vcc, s6, v12
-; SI-NEXT: buffer_load_dword v12, off, s[0:3], s32 offset:512 ; 4-byte Folded Reload
-; SI-NEXT: v_cvt_f32_f16_e32 v5, v5
-; SI-NEXT: buffer_store_dword v5, off, s[0:3], s32 offset:816 ; 4-byte Folded Spill
-; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_cvt_f32_f16_e32 v5, v7
-; SI-NEXT: buffer_store_dword v5, off, s[0:3], s32 offset:832 ; 4-byte Folded Spill
-; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_cvt_f32_f16_e32 v5, v9
-; SI-NEXT: buffer_store_dword v5, off, s[0:3], s32 offset:804 ; 4-byte Folded Spill
-; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_cvt_f32_f16_e32 v5, v11
-; SI-NEXT: buffer_store_dword v5, off, s[0:3], s32 offset:828 ; 4-byte Folded Spill
-; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_cvt_f32_f16_e32 v5, v13
-; SI-NEXT: buffer_store_dword v5, off, s[0:3], s32 offset:792 ; 4-byte Folded Spill
-; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_cvt_f32_f16_e32 v5, v15
-; SI-NEXT: buffer_store_dword v5, off, s[0:3], s32 offset:808 ; 4-byte Folded Spill
-; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_cvt_f32_f16_e32 v5, v17
-; SI-NEXT: v_cvt_f32_f16_e32 v17, v60
-; SI-NEXT: buffer_store_dword v5, off, s[0:3], s32 offset:776 ; 4-byte Folded Spill
-; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_cvt_f32_f16_e32 v5, v19
-; SI-NEXT: buffer_store_dword v5, off, s[0:3], s32 offset:800 ; 4-byte Folded Spill
-; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_cvt_f32_f16_e32 v5, v21
-; SI-NEXT: buffer_store_dword v5, off, s[0:3], s32 offset:760 ; 4-byte Folded Spill
-; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_cvt_f32_f16_e32 v5, v23
-; SI-NEXT: v_cvt_f32_f16_e32 v23, v58
-; SI-NEXT: buffer_store_dword v5, off, s[0:3], s32 offset:784 ; 4-byte Folded Spill
-; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_cvt_f32_f16_e32 v5, v31
-; SI-NEXT: v_cvt_f32_f16_e32 v31, v61
-; SI-NEXT: buffer_store_dword v5, off, s[0:3], s32 offset:740 ; 4-byte Folded Spill
-; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_cvt_f32_f16_e32 v5, v25
-; SI-NEXT: buffer_store_dword v5, off, s[0:3], s32 offset:764 ; 4-byte Folded Spill
-; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_cvt_f32_f16_e32 v5, v51
-; SI-NEXT: buffer_store_dword v5, off, s[0:3], s32 offset:732 ; 4-byte Folded Spill
-; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_cvt_f32_f16_e32 v5, v54
-; SI-NEXT: buffer_store_dword v5, off, s[0:3], s32 offset:756 ; 4-byte Folded Spill
-; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_cvt_f32_f16_e32 v5, v55
-; SI-NEXT: buffer_store_dword v5, off, s[0:3], s32 offset:720 ; 4-byte Folded Spill
-; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_cvt_f32_f16_e32 v5, v41
-; SI-NEXT: s_waitcnt vmcnt(14)
-; SI-NEXT: v_add_i32_e32 v32, vcc, s6, v12
-; SI-NEXT: buffer_load_dword v12, off, s[0:3], s32 offset:516 ; 4-byte Folded Reload
-; SI-NEXT: buffer_store_dword v5, off, s[0:3], s32 offset:748 ; 4-byte Folded Spill
-; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_cvt_f32_f16_e32 v5, v62
-; SI-NEXT: s_waitcnt vmcnt(1)
-; SI-NEXT: v_add_i32_e32 v14, vcc, s6, v12
-; SI-NEXT: v_add_i32_e32 v12, vcc, s6, v63
-; SI-NEXT: buffer_load_dword v63, off, s[0:3], s32 offset:476 ; 4-byte Folded Reload
-; SI-NEXT: v_cvt_f32_f16_e32 v12, v12
-; SI-NEXT: buffer_store_dword v12, off, s[0:3], s32 offset:840 ; 4-byte Folded Spill
-; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_cvt_f32_f16_e32 v12, v14
-; SI-NEXT: buffer_store_dword v12, off, s[0:3], s32 offset:820 ; 4-byte Folded Spill
-; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_cvt_f32_f16_e32 v12, v32
-; SI-NEXT: buffer_store_dword v12, off, s[0:3], s32 offset:836 ; 4-byte Folded Spill
-; SI-NEXT: s_waitcnt vmcnt(3)
; SI-NEXT: v_add_i32_e32 v63, vcc, s6, v63
; SI-NEXT: v_cvt_f32_f16_e32 v63, v63
-; SI-NEXT: buffer_store_dword v63, off, s[0:3], s32 offset:824 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v63, off, s[0:3], s32 offset:856 ; 4-byte Folded Spill
+; SI-NEXT: s_waitcnt vmcnt(1)
+; SI-NEXT: v_cvt_f32_f16_e32 v21, v0
; SI-NEXT: .LBB92_4: ; %end
; SI-NEXT: s_or_b64 exec, exec, s[4:5]
+; SI-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:872 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:856 ; 4-byte Folded Reload
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:840 ; 4-byte Folded Reload
-; SI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:824 ; 4-byte Folded Reload
-; SI-NEXT: v_add_i32_e32 v2, vcc, 12, v10
+; SI-NEXT: v_add_i32_e32 v2, vcc, 12, v5
; SI-NEXT: s_waitcnt vmcnt(1)
; SI-NEXT: v_cvt_f16_f32_e32 v0, v0
; SI-NEXT: s_waitcnt vmcnt(0)
; SI-NEXT: v_cvt_f16_f32_e32 v1, v1
; SI-NEXT: v_lshlrev_b32_e32 v0, 16, v0
; SI-NEXT: v_or_b32_e32 v0, v1, v0
-; SI-NEXT: buffer_store_dword v0, v10, s[0:3], 0 offen
+; SI-NEXT: buffer_store_dword v0, v5, s[0:3], 0 offen
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:836 ; 4-byte Folded Reload
-; SI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:820 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:868 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:852 ; 4-byte Folded Reload
; SI-NEXT: s_waitcnt vmcnt(1)
; SI-NEXT: v_cvt_f16_f32_e32 v0, v0
; SI-NEXT: s_waitcnt vmcnt(0)
; SI-NEXT: v_cvt_f16_f32_e32 v1, v1
; SI-NEXT: v_lshlrev_b32_e32 v0, 16, v0
; SI-NEXT: v_or_b32_e32 v0, v1, v0
-; SI-NEXT: v_add_i32_e32 v1, vcc, 4, v10
+; SI-NEXT: v_add_i32_e32 v1, vcc, 4, v5
; SI-NEXT: buffer_store_dword v0, v1, s[0:3], 0 offen
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:832 ; 4-byte Folded Reload
-; SI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:816 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:864 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:848 ; 4-byte Folded Reload
; SI-NEXT: s_waitcnt vmcnt(1)
; SI-NEXT: v_cvt_f16_f32_e32 v0, v0
; SI-NEXT: s_waitcnt vmcnt(0)
; SI-NEXT: v_cvt_f16_f32_e32 v1, v1
; SI-NEXT: v_lshlrev_b32_e32 v0, 16, v0
; SI-NEXT: v_or_b32_e32 v0, v1, v0
-; SI-NEXT: v_add_i32_e32 v1, vcc, 8, v10
+; SI-NEXT: v_add_i32_e32 v1, vcc, 8, v5
; SI-NEXT: buffer_store_dword v0, v1, s[0:3], 0 offen
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:828 ; 4-byte Folded Reload
-; SI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:804 ; 4-byte Folded Reload
-; SI-NEXT: s_waitcnt vmcnt(1)
-; SI-NEXT: v_cvt_f16_f32_e32 v0, v0
-; SI-NEXT: s_waitcnt vmcnt(0)
-; SI-NEXT: v_cvt_f16_f32_e32 v1, v1
-; SI-NEXT: v_lshlrev_b32_e32 v0, 16, v0
-; SI-NEXT: v_or_b32_e32 v0, v1, v0
-; SI-NEXT: buffer_store_dword v0, v2, s[0:3], 0 offen
-; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:808 ; 4-byte Folded Reload
-; SI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:792 ; 4-byte Folded Reload
-; SI-NEXT: v_add_i32_e32 v2, vcc, 16, v10
+; SI-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:860 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:836 ; 4-byte Folded Reload
; SI-NEXT: s_waitcnt vmcnt(1)
; SI-NEXT: v_cvt_f16_f32_e32 v0, v0
; SI-NEXT: s_waitcnt vmcnt(0)
@@ -180692,9 +180725,9 @@ define <64 x half> @bitcast_v128i8_to_v64f16(<128 x i8> %a, i32 %b) {
; SI-NEXT: v_or_b32_e32 v0, v1, v0
; SI-NEXT: buffer_store_dword v0, v2, s[0:3], 0 offen
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:800 ; 4-byte Folded Reload
-; SI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:776 ; 4-byte Folded Reload
-; SI-NEXT: v_add_i32_e32 v2, vcc, 20, v10
+; SI-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:840 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:824 ; 4-byte Folded Reload
+; SI-NEXT: v_add_i32_e32 v2, vcc, 16, v5
; SI-NEXT: s_waitcnt vmcnt(1)
; SI-NEXT: v_cvt_f16_f32_e32 v0, v0
; SI-NEXT: s_waitcnt vmcnt(0)
@@ -180703,9 +180736,9 @@ define <64 x half> @bitcast_v128i8_to_v64f16(<128 x i8> %a, i32 %b) {
; SI-NEXT: v_or_b32_e32 v0, v1, v0
; SI-NEXT: buffer_store_dword v0, v2, s[0:3], 0 offen
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:784 ; 4-byte Folded Reload
-; SI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:760 ; 4-byte Folded Reload
-; SI-NEXT: v_add_i32_e32 v2, vcc, 24, v10
+; SI-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:832 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:808 ; 4-byte Folded Reload
+; SI-NEXT: v_add_i32_e32 v2, vcc, 20, v5
; SI-NEXT: s_waitcnt vmcnt(1)
; SI-NEXT: v_cvt_f16_f32_e32 v0, v0
; SI-NEXT: s_waitcnt vmcnt(0)
@@ -180714,9 +180747,9 @@ define <64 x half> @bitcast_v128i8_to_v64f16(<128 x i8> %a, i32 %b) {
; SI-NEXT: v_or_b32_e32 v0, v1, v0
; SI-NEXT: buffer_store_dword v0, v2, s[0:3], 0 offen
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:764 ; 4-byte Folded Reload
-; SI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:740 ; 4-byte Folded Reload
-; SI-NEXT: v_add_i32_e32 v2, vcc, 28, v10
+; SI-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:816 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:792 ; 4-byte Folded Reload
+; SI-NEXT: v_add_i32_e32 v2, vcc, 24, v5
; SI-NEXT: s_waitcnt vmcnt(1)
; SI-NEXT: v_cvt_f16_f32_e32 v0, v0
; SI-NEXT: s_waitcnt vmcnt(0)
@@ -180725,9 +180758,9 @@ define <64 x half> @bitcast_v128i8_to_v64f16(<128 x i8> %a, i32 %b) {
; SI-NEXT: v_or_b32_e32 v0, v1, v0
; SI-NEXT: buffer_store_dword v0, v2, s[0:3], 0 offen
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:756 ; 4-byte Folded Reload
-; SI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:732 ; 4-byte Folded Reload
-; SI-NEXT: v_add_i32_e32 v2, vcc, 32, v10
+; SI-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:796 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:772 ; 4-byte Folded Reload
+; SI-NEXT: v_add_i32_e32 v2, vcc, 28, v5
; SI-NEXT: s_waitcnt vmcnt(1)
; SI-NEXT: v_cvt_f16_f32_e32 v0, v0
; SI-NEXT: s_waitcnt vmcnt(0)
@@ -180736,9 +180769,9 @@ define <64 x half> @bitcast_v128i8_to_v64f16(<128 x i8> %a, i32 %b) {
; SI-NEXT: v_or_b32_e32 v0, v1, v0
; SI-NEXT: buffer_store_dword v0, v2, s[0:3], 0 offen
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:748 ; 4-byte Folded Reload
-; SI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:720 ; 4-byte Folded Reload
-; SI-NEXT: v_add_i32_e32 v2, vcc, 36, v10
+; SI-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:788 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:764 ; 4-byte Folded Reload
+; SI-NEXT: v_add_i32_e32 v2, vcc, 32, v5
; SI-NEXT: s_waitcnt vmcnt(1)
; SI-NEXT: v_cvt_f16_f32_e32 v0, v0
; SI-NEXT: s_waitcnt vmcnt(0)
@@ -180747,9 +180780,9 @@ define <64 x half> @bitcast_v128i8_to_v64f16(<128 x i8> %a, i32 %b) {
; SI-NEXT: v_or_b32_e32 v0, v1, v0
; SI-NEXT: buffer_store_dword v0, v2, s[0:3], 0 offen
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:736 ; 4-byte Folded Reload
-; SI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:708 ; 4-byte Folded Reload
-; SI-NEXT: v_add_i32_e32 v2, vcc, 40, v10
+; SI-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:780 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:752 ; 4-byte Folded Reload
+; SI-NEXT: v_add_i32_e32 v2, vcc, 36, v5
; SI-NEXT: s_waitcnt vmcnt(1)
; SI-NEXT: v_cvt_f16_f32_e32 v0, v0
; SI-NEXT: s_waitcnt vmcnt(0)
@@ -180758,9 +180791,9 @@ define <64 x half> @bitcast_v128i8_to_v64f16(<128 x i8> %a, i32 %b) {
; SI-NEXT: v_or_b32_e32 v0, v1, v0
; SI-NEXT: buffer_store_dword v0, v2, s[0:3], 0 offen
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:724 ; 4-byte Folded Reload
-; SI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:700 ; 4-byte Folded Reload
-; SI-NEXT: v_add_i32_e32 v2, vcc, 44, v10
+; SI-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:768 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:740 ; 4-byte Folded Reload
+; SI-NEXT: v_add_i32_e32 v2, vcc, 40, v5
; SI-NEXT: s_waitcnt vmcnt(1)
; SI-NEXT: v_cvt_f16_f32_e32 v0, v0
; SI-NEXT: s_waitcnt vmcnt(0)
@@ -180769,9 +180802,9 @@ define <64 x half> @bitcast_v128i8_to_v64f16(<128 x i8> %a, i32 %b) {
; SI-NEXT: v_or_b32_e32 v0, v1, v0
; SI-NEXT: buffer_store_dword v0, v2, s[0:3], 0 offen
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:716 ; 4-byte Folded Reload
-; SI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:688 ; 4-byte Folded Reload
-; SI-NEXT: v_add_i32_e32 v2, vcc, 48, v10
+; SI-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:756 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:732 ; 4-byte Folded Reload
+; SI-NEXT: v_add_i32_e32 v2, vcc, 44, v5
; SI-NEXT: s_waitcnt vmcnt(1)
; SI-NEXT: v_cvt_f16_f32_e32 v0, v0
; SI-NEXT: s_waitcnt vmcnt(0)
@@ -180780,9 +180813,9 @@ define <64 x half> @bitcast_v128i8_to_v64f16(<128 x i8> %a, i32 %b) {
; SI-NEXT: v_or_b32_e32 v0, v1, v0
; SI-NEXT: buffer_store_dword v0, v2, s[0:3], 0 offen
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:704 ; 4-byte Folded Reload
-; SI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:676 ; 4-byte Folded Reload
-; SI-NEXT: v_add_i32_e32 v2, vcc, 52, v10
+; SI-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:748 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:720 ; 4-byte Folded Reload
+; SI-NEXT: v_add_i32_e32 v2, vcc, 48, v5
; SI-NEXT: s_waitcnt vmcnt(1)
; SI-NEXT: v_cvt_f16_f32_e32 v0, v0
; SI-NEXT: s_waitcnt vmcnt(0)
@@ -180791,9 +180824,9 @@ define <64 x half> @bitcast_v128i8_to_v64f16(<128 x i8> %a, i32 %b) {
; SI-NEXT: v_or_b32_e32 v0, v1, v0
; SI-NEXT: buffer_store_dword v0, v2, s[0:3], 0 offen
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:692 ; 4-byte Folded Reload
-; SI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:668 ; 4-byte Folded Reload
-; SI-NEXT: v_add_i32_e32 v2, vcc, 56, v10
+; SI-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:736 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:708 ; 4-byte Folded Reload
+; SI-NEXT: v_add_i32_e32 v2, vcc, 52, v5
; SI-NEXT: s_waitcnt vmcnt(1)
; SI-NEXT: v_cvt_f16_f32_e32 v0, v0
; SI-NEXT: s_waitcnt vmcnt(0)
@@ -180802,9 +180835,9 @@ define <64 x half> @bitcast_v128i8_to_v64f16(<128 x i8> %a, i32 %b) {
; SI-NEXT: v_or_b32_e32 v0, v1, v0
; SI-NEXT: buffer_store_dword v0, v2, s[0:3], 0 offen
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:684 ; 4-byte Folded Reload
-; SI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:656 ; 4-byte Folded Reload
-; SI-NEXT: v_add_i32_e32 v2, vcc, 60, v10
+; SI-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:724 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:700 ; 4-byte Folded Reload
+; SI-NEXT: v_add_i32_e32 v2, vcc, 56, v5
; SI-NEXT: s_waitcnt vmcnt(1)
; SI-NEXT: v_cvt_f16_f32_e32 v0, v0
; SI-NEXT: s_waitcnt vmcnt(0)
@@ -180813,9 +180846,9 @@ define <64 x half> @bitcast_v128i8_to_v64f16(<128 x i8> %a, i32 %b) {
; SI-NEXT: v_or_b32_e32 v0, v1, v0
; SI-NEXT: buffer_store_dword v0, v2, s[0:3], 0 offen
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:672 ; 4-byte Folded Reload
-; SI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:644 ; 4-byte Folded Reload
-; SI-NEXT: v_add_i32_e32 v2, vcc, 64, v10
+; SI-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:716 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:688 ; 4-byte Folded Reload
+; SI-NEXT: v_add_i32_e32 v2, vcc, 60, v5
; SI-NEXT: s_waitcnt vmcnt(1)
; SI-NEXT: v_cvt_f16_f32_e32 v0, v0
; SI-NEXT: s_waitcnt vmcnt(0)
@@ -180824,9 +180857,9 @@ define <64 x half> @bitcast_v128i8_to_v64f16(<128 x i8> %a, i32 %b) {
; SI-NEXT: v_or_b32_e32 v0, v1, v0
; SI-NEXT: buffer_store_dword v0, v2, s[0:3], 0 offen
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:660 ; 4-byte Folded Reload
-; SI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:636 ; 4-byte Folded Reload
-; SI-NEXT: v_add_i32_e32 v2, vcc, 0x44, v10
+; SI-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:704 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:676 ; 4-byte Folded Reload
+; SI-NEXT: v_add_i32_e32 v2, vcc, 64, v5
; SI-NEXT: s_waitcnt vmcnt(1)
; SI-NEXT: v_cvt_f16_f32_e32 v0, v0
; SI-NEXT: s_waitcnt vmcnt(0)
@@ -180835,9 +180868,9 @@ define <64 x half> @bitcast_v128i8_to_v64f16(<128 x i8> %a, i32 %b) {
; SI-NEXT: v_or_b32_e32 v0, v1, v0
; SI-NEXT: buffer_store_dword v0, v2, s[0:3], 0 offen
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:652 ; 4-byte Folded Reload
-; SI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:620 ; 4-byte Folded Reload
-; SI-NEXT: v_add_i32_e32 v2, vcc, 0x48, v10
+; SI-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:692 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:668 ; 4-byte Folded Reload
+; SI-NEXT: v_add_i32_e32 v2, vcc, 0x44, v5
; SI-NEXT: s_waitcnt vmcnt(1)
; SI-NEXT: v_cvt_f16_f32_e32 v0, v0
; SI-NEXT: s_waitcnt vmcnt(0)
@@ -180846,9 +180879,9 @@ define <64 x half> @bitcast_v128i8_to_v64f16(<128 x i8> %a, i32 %b) {
; SI-NEXT: v_or_b32_e32 v0, v1, v0
; SI-NEXT: buffer_store_dword v0, v2, s[0:3], 0 offen
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:640 ; 4-byte Folded Reload
-; SI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:608 ; 4-byte Folded Reload
-; SI-NEXT: v_add_i32_e32 v2, vcc, 0x4c, v10
+; SI-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:684 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:652 ; 4-byte Folded Reload
+; SI-NEXT: v_add_i32_e32 v2, vcc, 0x48, v5
; SI-NEXT: s_waitcnt vmcnt(1)
; SI-NEXT: v_cvt_f16_f32_e32 v0, v0
; SI-NEXT: s_waitcnt vmcnt(0)
@@ -180857,9 +180890,9 @@ define <64 x half> @bitcast_v128i8_to_v64f16(<128 x i8> %a, i32 %b) {
; SI-NEXT: v_or_b32_e32 v0, v1, v0
; SI-NEXT: buffer_store_dword v0, v2, s[0:3], 0 offen
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:628 ; 4-byte Folded Reload
-; SI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:596 ; 4-byte Folded Reload
-; SI-NEXT: v_add_i32_e32 v2, vcc, 0x50, v10
+; SI-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:672 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:640 ; 4-byte Folded Reload
+; SI-NEXT: v_add_i32_e32 v2, vcc, 0x4c, v5
; SI-NEXT: s_waitcnt vmcnt(1)
; SI-NEXT: v_cvt_f16_f32_e32 v0, v0
; SI-NEXT: s_waitcnt vmcnt(0)
@@ -180868,9 +180901,9 @@ define <64 x half> @bitcast_v128i8_to_v64f16(<128 x i8> %a, i32 %b) {
; SI-NEXT: v_or_b32_e32 v0, v1, v0
; SI-NEXT: buffer_store_dword v0, v2, s[0:3], 0 offen
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:616 ; 4-byte Folded Reload
-; SI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:580 ; 4-byte Folded Reload
-; SI-NEXT: v_add_i32_e32 v2, vcc, 0x54, v10
+; SI-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:660 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:628 ; 4-byte Folded Reload
+; SI-NEXT: v_add_i32_e32 v2, vcc, 0x50, v5
; SI-NEXT: s_waitcnt vmcnt(1)
; SI-NEXT: v_cvt_f16_f32_e32 v0, v0
; SI-NEXT: s_waitcnt vmcnt(0)
@@ -180879,9 +180912,9 @@ define <64 x half> @bitcast_v128i8_to_v64f16(<128 x i8> %a, i32 %b) {
; SI-NEXT: v_or_b32_e32 v0, v1, v0
; SI-NEXT: buffer_store_dword v0, v2, s[0:3], 0 offen
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:600 ; 4-byte Folded Reload
-; SI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:568 ; 4-byte Folded Reload
-; SI-NEXT: v_add_i32_e32 v2, vcc, 0x58, v10
+; SI-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:648 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:612 ; 4-byte Folded Reload
+; SI-NEXT: v_add_i32_e32 v2, vcc, 0x54, v5
; SI-NEXT: s_waitcnt vmcnt(1)
; SI-NEXT: v_cvt_f16_f32_e32 v0, v0
; SI-NEXT: s_waitcnt vmcnt(0)
@@ -180890,9 +180923,9 @@ define <64 x half> @bitcast_v128i8_to_v64f16(<128 x i8> %a, i32 %b) {
; SI-NEXT: v_or_b32_e32 v0, v1, v0
; SI-NEXT: buffer_store_dword v0, v2, s[0:3], 0 offen
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:588 ; 4-byte Folded Reload
-; SI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:560 ; 4-byte Folded Reload
-; SI-NEXT: v_add_i32_e32 v2, vcc, 0x5c, v10
+; SI-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:632 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:600 ; 4-byte Folded Reload
+; SI-NEXT: v_add_i32_e32 v2, vcc, 0x58, v5
; SI-NEXT: s_waitcnt vmcnt(1)
; SI-NEXT: v_cvt_f16_f32_e32 v0, v0
; SI-NEXT: s_waitcnt vmcnt(0)
@@ -180901,27 +180934,27 @@ define <64 x half> @bitcast_v128i8_to_v64f16(<128 x i8> %a, i32 %b) {
; SI-NEXT: v_or_b32_e32 v0, v1, v0
; SI-NEXT: buffer_store_dword v0, v2, s[0:3], 0 offen
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:576 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:620 ; 4-byte Folded Reload
; SI-NEXT: v_cvt_f16_f32_e32 v1, v48
-; SI-NEXT: v_add_i32_e32 v2, vcc, 0x60, v10
+; SI-NEXT: v_add_i32_e32 v2, vcc, 0x5c, v5
; SI-NEXT: s_waitcnt vmcnt(0)
; SI-NEXT: v_cvt_f16_f32_e32 v0, v0
; SI-NEXT: v_lshlrev_b32_e32 v0, 16, v0
; SI-NEXT: v_or_b32_e32 v0, v1, v0
; SI-NEXT: buffer_store_dword v0, v2, s[0:3], 0 offen
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:564 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:608 ; 4-byte Folded Reload
; SI-NEXT: v_cvt_f16_f32_e32 v1, v38
-; SI-NEXT: v_add_i32_e32 v2, vcc, 0x64, v10
+; SI-NEXT: v_add_i32_e32 v2, vcc, 0x60, v5
; SI-NEXT: s_waitcnt vmcnt(0)
; SI-NEXT: v_cvt_f16_f32_e32 v0, v0
; SI-NEXT: v_lshlrev_b32_e32 v0, 16, v0
; SI-NEXT: v_or_b32_e32 v0, v1, v0
; SI-NEXT: buffer_store_dword v0, v2, s[0:3], 0 offen
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:552 ; 4-byte Folded Reload
-; SI-NEXT: v_cvt_f16_f32_e32 v1, v36
-; SI-NEXT: v_add_i32_e32 v2, vcc, 0x68, v10
+; SI-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:596 ; 4-byte Folded Reload
+; SI-NEXT: v_cvt_f16_f32_e32 v1, v35
+; SI-NEXT: v_add_i32_e32 v2, vcc, 0x64, v5
; SI-NEXT: s_waitcnt vmcnt(0)
; SI-NEXT: v_cvt_f16_f32_e32 v0, v0
; SI-NEXT: v_lshlrev_b32_e32 v0, 16, v0
@@ -180929,36 +180962,43 @@ define <64 x half> @bitcast_v128i8_to_v64f16(<128 x i8> %a, i32 %b) {
; SI-NEXT: buffer_store_dword v0, v2, s[0:3], 0 offen
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_cvt_f16_f32_e32 v0, v39
-; SI-NEXT: v_cvt_f16_f32_e32 v1, v33
-; SI-NEXT: v_add_i32_e32 v2, vcc, 0x6c, v10
+; SI-NEXT: v_cvt_f16_f32_e32 v1, v32
+; SI-NEXT: v_add_i32_e32 v2, vcc, 0x68, v5
; SI-NEXT: v_lshlrev_b32_e32 v0, 16, v0
; SI-NEXT: v_or_b32_e32 v0, v1, v0
; SI-NEXT: buffer_store_dword v0, v2, s[0:3], 0 offen
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_cvt_f16_f32_e32 v0, v37
; SI-NEXT: v_cvt_f16_f32_e32 v1, v29
-; SI-NEXT: v_add_i32_e32 v2, vcc, 0x70, v10
+; SI-NEXT: v_add_i32_e32 v2, vcc, 0x6c, v5
; SI-NEXT: v_lshlrev_b32_e32 v0, 16, v0
; SI-NEXT: v_or_b32_e32 v0, v1, v0
; SI-NEXT: buffer_store_dword v0, v2, s[0:3], 0 offen
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_cvt_f16_f32_e32 v0, v35
+; SI-NEXT: v_cvt_f16_f32_e32 v0, v33
; SI-NEXT: v_cvt_f16_f32_e32 v1, v23
-; SI-NEXT: v_add_i32_e32 v2, vcc, 0x74, v10
+; SI-NEXT: v_add_i32_e32 v2, vcc, 0x70, v5
; SI-NEXT: v_lshlrev_b32_e32 v0, 16, v0
; SI-NEXT: v_or_b32_e32 v0, v1, v0
; SI-NEXT: buffer_store_dword v0, v2, s[0:3], 0 offen
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_cvt_f16_f32_e32 v0, v31
-; SI-NEXT: v_cvt_f16_f32_e32 v1, v17
-; SI-NEXT: v_add_i32_e32 v2, vcc, 0x78, v10
+; SI-NEXT: v_cvt_f16_f32_e32 v1, v13
+; SI-NEXT: v_add_i32_e32 v2, vcc, 0x74, v5
; SI-NEXT: v_lshlrev_b32_e32 v0, 16, v0
; SI-NEXT: v_or_b32_e32 v0, v1, v0
; SI-NEXT: buffer_store_dword v0, v2, s[0:3], 0 offen
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_cvt_f16_f32_e32 v0, v27
-; SI-NEXT: v_cvt_f16_f32_e32 v1, v5
-; SI-NEXT: v_add_i32_e32 v2, vcc, 0x7c, v10
+; SI-NEXT: v_cvt_f16_f32_e32 v0, v25
+; SI-NEXT: v_cvt_f16_f32_e32 v1, v11
+; SI-NEXT: v_add_i32_e32 v2, vcc, 0x78, v5
+; SI-NEXT: v_lshlrev_b32_e32 v0, 16, v0
+; SI-NEXT: v_or_b32_e32 v0, v1, v0
+; SI-NEXT: buffer_store_dword v0, v2, s[0:3], 0 offen
+; SI-NEXT: s_waitcnt expcnt(0)
+; SI-NEXT: v_cvt_f16_f32_e32 v0, v21
+; SI-NEXT: v_cvt_f16_f32_e32 v1, v9
+; SI-NEXT: v_add_i32_e32 v2, vcc, 0x7c, v5
; SI-NEXT: v_lshlrev_b32_e32 v0, 16, v0
; SI-NEXT: v_or_b32_e32 v0, v1, v0
; SI-NEXT: buffer_store_dword v0, v2, s[0:3], 0 offen
@@ -185174,11 +185214,7 @@ define inreg <64 x half> @bitcast_v128i8_to_v64f16_scalar(<128 x i8> inreg %a, i
; SI-NEXT: buffer_load_dword v31, off, s[0:3], s32 offset:332
; SI-NEXT: ; implicit-def: $vgpr61 : SGPR spill to VGPR lane
; SI-NEXT: s_mov_b32 s10, s16
-; SI-NEXT: s_waitcnt expcnt(1)
-; SI-NEXT: v_writelane_b32 v61, s29, 0
-; SI-NEXT: v_writelane_b32 v61, s28, 1
-; SI-NEXT: v_writelane_b32 v61, s27, 2
-; SI-NEXT: s_mov_b32 s61, s21
+; SI-NEXT: s_waitcnt expcnt(2)
; SI-NEXT: v_writelane_b32 v63, s30, 0
; SI-NEXT: v_writelane_b32 v63, s31, 1
; SI-NEXT: v_writelane_b32 v63, s34, 2
@@ -185213,59 +185249,59 @@ define inreg <64 x half> @bitcast_v128i8_to_v64f16_scalar(<128 x i8> inreg %a, i
; SI-NEXT: v_writelane_b32 v63, s87, 31
; SI-NEXT: v_writelane_b32 v63, s96, 32
; SI-NEXT: v_writelane_b32 v63, s97, 33
-; SI-NEXT: s_mov_b32 s67, s19
-; SI-NEXT: s_mov_b32 s54, s17
-; SI-NEXT: s_mov_b32 s35, s23
-; SI-NEXT: s_mov_b32 s39, s26
-; SI-NEXT: s_mov_b32 s62, s25
+; SI-NEXT: s_mov_b32 s54, s27
+; SI-NEXT: s_mov_b32 s79, s29
+; SI-NEXT: s_mov_b32 s66, s26
+; SI-NEXT: s_mov_b32 s64, s23
+; SI-NEXT: s_mov_b32 s65, s19
+; SI-NEXT: s_mov_b32 s67, s17
; SI-NEXT: v_writelane_b32 v63, s98, 34
; SI-NEXT: v_writelane_b32 v63, s99, 35
-; SI-NEXT: v_readfirstlane_b32 s99, v1
-; SI-NEXT: v_readfirstlane_b32 s74, v24
+; SI-NEXT: s_mov_b32 s92, s24
+; SI-NEXT: v_readfirstlane_b32 s31, v1
+; SI-NEXT: v_readfirstlane_b32 s81, v23
; SI-NEXT: ; implicit-def: $vgpr62 : SGPR spill to VGPR lane
-; SI-NEXT: v_readfirstlane_b32 s6, v23
-; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_writelane_b32 v62, s74, 0
; SI-NEXT: v_readfirstlane_b32 s12, v26
-; SI-NEXT: v_writelane_b32 v62, s6, 1
+; SI-NEXT: s_waitcnt expcnt(0)
+; SI-NEXT: v_writelane_b32 v62, s81, 0
; SI-NEXT: v_readfirstlane_b32 s14, v25
-; SI-NEXT: v_writelane_b32 v62, s12, 2
+; SI-NEXT: v_writelane_b32 v62, s12, 1
; SI-NEXT: v_readfirstlane_b32 s46, v28
-; SI-NEXT: v_writelane_b32 v62, s14, 3
-; SI-NEXT: v_readfirstlane_b32 s56, v27
-; SI-NEXT: v_writelane_b32 v62, s46, 4
-; SI-NEXT: v_readfirstlane_b32 s57, v30
-; SI-NEXT: v_writelane_b32 v62, s56, 5
-; SI-NEXT: v_readfirstlane_b32 s59, v29
-; SI-NEXT: v_writelane_b32 v62, s57, 6
-; SI-NEXT: v_writelane_b32 v62, s59, 7
-; SI-NEXT: s_mov_b32 s60, s20
-; SI-NEXT: s_mov_b32 s63, s24
-; SI-NEXT: v_readfirstlane_b32 s95, v3
-; SI-NEXT: v_readfirstlane_b32 s31, v5
-; SI-NEXT: v_readfirstlane_b32 s24, v9
-; SI-NEXT: v_readfirstlane_b32 s38, v12
+; SI-NEXT: v_writelane_b32 v62, s14, 2
+; SI-NEXT: v_readfirstlane_b32 s57, v27
+; SI-NEXT: v_writelane_b32 v62, s46, 3
+; SI-NEXT: v_readfirstlane_b32 s58, v30
+; SI-NEXT: v_writelane_b32 v62, s57, 4
+; SI-NEXT: s_mov_b32 s77, s25
+; SI-NEXT: v_readfirstlane_b32 s25, v29
+; SI-NEXT: v_writelane_b32 v62, s58, 5
+; SI-NEXT: v_writelane_b32 v62, s25, 6
+; SI-NEXT: v_readfirstlane_b32 s55, v3
+; SI-NEXT: v_readfirstlane_b32 s80, v5
+; SI-NEXT: v_readfirstlane_b32 s51, v7
; SI-NEXT: v_readfirstlane_b32 s36, v11
-; SI-NEXT: v_readfirstlane_b32 s8, v14
-; SI-NEXT: v_readfirstlane_b32 s27, v13
-; SI-NEXT: v_readfirstlane_b32 s9, v16
-; SI-NEXT: v_readfirstlane_b32 s79, v15
+; SI-NEXT: v_readfirstlane_b32 s87, v16
+; SI-NEXT: v_readfirstlane_b32 s84, v15
; SI-NEXT: v_readfirstlane_b32 s13, v18
; SI-NEXT: v_readfirstlane_b32 s15, v17
; SI-NEXT: v_readfirstlane_b32 s42, v20
; SI-NEXT: v_readfirstlane_b32 s43, v19
; SI-NEXT: v_readfirstlane_b32 s44, v22
+; SI-NEXT: v_readfirstlane_b32 s73, v21
+; SI-NEXT: v_readfirstlane_b32 s74, v24
+; SI-NEXT: v_readfirstlane_b32 s62, v14
+; SI-NEXT: v_readfirstlane_b32 s9, v13
+; SI-NEXT: v_readfirstlane_b32 s63, v12
+; SI-NEXT: v_readfirstlane_b32 s61, v10
+; SI-NEXT: v_readfirstlane_b32 s94, v9
+; SI-NEXT: v_readfirstlane_b32 s60, v8
; SI-NEXT: s_waitcnt vmcnt(0)
; SI-NEXT: v_readfirstlane_b32 s4, v31
; SI-NEXT: buffer_load_dword v31, off, s[0:3], s32 offset:328
-; SI-NEXT: v_writelane_b32 v61, s4, 3
-; SI-NEXT: v_readfirstlane_b32 s45, v21
-; SI-NEXT: v_readfirstlane_b32 s98, v10
-; SI-NEXT: v_readfirstlane_b32 s90, v8
-; SI-NEXT: v_readfirstlane_b32 s88, v7
-; SI-NEXT: v_readfirstlane_b32 s91, v6
-; SI-NEXT: v_readfirstlane_b32 s93, v4
-; SI-NEXT: v_readfirstlane_b32 s55, v2
+; SI-NEXT: v_writelane_b32 v61, s4, 0
+; SI-NEXT: v_readfirstlane_b32 s35, v6
+; SI-NEXT: v_readfirstlane_b32 s91, v4
+; SI-NEXT: v_readfirstlane_b32 s95, v2
; SI-NEXT: buffer_store_dword v40, off, s[0:3], s32 offset:388 ; 4-byte Folded Spill
; SI-NEXT: buffer_store_dword v41, off, s[0:3], s32 offset:384 ; 4-byte Folded Spill
; SI-NEXT: buffer_store_dword v42, off, s[0:3], s32 offset:380 ; 4-byte Folded Spill
@@ -185283,142 +185319,142 @@ define inreg <64 x half> @bitcast_v128i8_to_v64f16_scalar(<128 x i8> inreg %a, i
; SI-NEXT: s_waitcnt vmcnt(14)
; SI-NEXT: v_readfirstlane_b32 s4, v31
; SI-NEXT: buffer_load_dword v31, off, s[0:3], s32 offset:324
-; SI-NEXT: v_writelane_b32 v61, s4, 4
+; SI-NEXT: v_writelane_b32 v61, s4, 1
; SI-NEXT: s_waitcnt vmcnt(1)
; SI-NEXT: v_cmp_ne_u32_e32 vcc, 0, v1
; SI-NEXT: s_waitcnt vmcnt(0)
; SI-NEXT: v_readfirstlane_b32 s4, v31
; SI-NEXT: buffer_load_dword v31, off, s[0:3], s32 offset:320
-; SI-NEXT: v_writelane_b32 v61, s4, 5
+; SI-NEXT: v_writelane_b32 v61, s4, 2
; SI-NEXT: s_waitcnt vmcnt(0)
; SI-NEXT: v_readfirstlane_b32 s4, v31
; SI-NEXT: buffer_load_dword v31, off, s[0:3], s32 offset:316
-; SI-NEXT: v_writelane_b32 v61, s4, 6
+; SI-NEXT: v_writelane_b32 v61, s4, 3
; SI-NEXT: s_waitcnt vmcnt(0)
; SI-NEXT: v_readfirstlane_b32 s4, v31
; SI-NEXT: buffer_load_dword v31, off, s[0:3], s32 offset:312
-; SI-NEXT: v_writelane_b32 v61, s4, 7
+; SI-NEXT: v_writelane_b32 v61, s4, 4
; SI-NEXT: s_waitcnt vmcnt(0)
; SI-NEXT: v_readfirstlane_b32 s4, v31
; SI-NEXT: buffer_load_dword v31, off, s[0:3], s32 offset:308
-; SI-NEXT: v_writelane_b32 v61, s4, 8
+; SI-NEXT: v_writelane_b32 v61, s4, 5
; SI-NEXT: s_waitcnt vmcnt(0)
; SI-NEXT: v_readfirstlane_b32 s4, v31
; SI-NEXT: buffer_load_dword v31, off, s[0:3], s32 offset:304
-; SI-NEXT: v_writelane_b32 v61, s4, 9
+; SI-NEXT: v_writelane_b32 v61, s4, 6
; SI-NEXT: s_waitcnt vmcnt(0)
; SI-NEXT: v_readfirstlane_b32 s4, v31
; SI-NEXT: buffer_load_dword v31, off, s[0:3], s32 offset:300
-; SI-NEXT: v_writelane_b32 v61, s4, 10
+; SI-NEXT: v_writelane_b32 v61, s4, 7
; SI-NEXT: s_waitcnt vmcnt(0)
; SI-NEXT: v_readfirstlane_b32 s4, v31
; SI-NEXT: buffer_load_dword v31, off, s[0:3], s32 offset:296
-; SI-NEXT: v_writelane_b32 v61, s4, 11
+; SI-NEXT: v_writelane_b32 v61, s4, 8
; SI-NEXT: s_waitcnt vmcnt(0)
; SI-NEXT: v_readfirstlane_b32 s4, v31
; SI-NEXT: buffer_load_dword v31, off, s[0:3], s32 offset:292
-; SI-NEXT: v_writelane_b32 v61, s4, 12
+; SI-NEXT: v_writelane_b32 v61, s4, 9
; SI-NEXT: s_waitcnt vmcnt(0)
; SI-NEXT: v_readfirstlane_b32 s4, v31
; SI-NEXT: buffer_load_dword v31, off, s[0:3], s32 offset:288
-; SI-NEXT: v_writelane_b32 v61, s4, 13
+; SI-NEXT: v_writelane_b32 v61, s4, 10
; SI-NEXT: s_waitcnt vmcnt(0)
; SI-NEXT: v_readfirstlane_b32 s4, v31
; SI-NEXT: buffer_load_dword v31, off, s[0:3], s32 offset:284
-; SI-NEXT: v_writelane_b32 v61, s4, 14
+; SI-NEXT: v_writelane_b32 v61, s4, 11
; SI-NEXT: s_waitcnt vmcnt(0)
; SI-NEXT: v_readfirstlane_b32 s4, v31
; SI-NEXT: buffer_load_dword v31, off, s[0:3], s32 offset:280
-; SI-NEXT: v_writelane_b32 v61, s4, 15
+; SI-NEXT: v_writelane_b32 v61, s4, 12
; SI-NEXT: s_waitcnt vmcnt(0)
; SI-NEXT: v_readfirstlane_b32 s4, v31
; SI-NEXT: buffer_load_dword v31, off, s[0:3], s32 offset:276
-; SI-NEXT: v_writelane_b32 v61, s4, 16
+; SI-NEXT: v_writelane_b32 v61, s4, 13
; SI-NEXT: s_waitcnt vmcnt(0)
; SI-NEXT: v_readfirstlane_b32 s4, v31
; SI-NEXT: buffer_load_dword v31, off, s[0:3], s32 offset:272
-; SI-NEXT: v_writelane_b32 v61, s4, 17
+; SI-NEXT: v_writelane_b32 v61, s4, 14
; SI-NEXT: s_waitcnt vmcnt(0)
; SI-NEXT: v_readfirstlane_b32 s4, v31
; SI-NEXT: buffer_load_dword v31, off, s[0:3], s32 offset:268
-; SI-NEXT: v_writelane_b32 v61, s4, 18
+; SI-NEXT: v_writelane_b32 v61, s4, 15
; SI-NEXT: s_waitcnt vmcnt(0)
; SI-NEXT: v_readfirstlane_b32 s4, v31
; SI-NEXT: buffer_load_dword v31, off, s[0:3], s32 offset:264
-; SI-NEXT: v_writelane_b32 v61, s4, 19
+; SI-NEXT: v_writelane_b32 v61, s4, 16
; SI-NEXT: s_waitcnt vmcnt(0)
; SI-NEXT: v_readfirstlane_b32 s4, v31
; SI-NEXT: buffer_load_dword v31, off, s[0:3], s32 offset:260
-; SI-NEXT: v_writelane_b32 v61, s4, 20
+; SI-NEXT: v_writelane_b32 v61, s4, 17
; SI-NEXT: s_waitcnt vmcnt(0)
; SI-NEXT: v_readfirstlane_b32 s4, v31
; SI-NEXT: buffer_load_dword v31, off, s[0:3], s32 offset:256
-; SI-NEXT: v_writelane_b32 v61, s4, 21
+; SI-NEXT: v_writelane_b32 v61, s4, 18
; SI-NEXT: s_waitcnt vmcnt(0)
; SI-NEXT: v_readfirstlane_b32 s4, v31
; SI-NEXT: buffer_load_dword v31, off, s[0:3], s32 offset:252
-; SI-NEXT: v_writelane_b32 v61, s4, 22
+; SI-NEXT: v_writelane_b32 v61, s4, 19
; SI-NEXT: s_waitcnt vmcnt(0)
; SI-NEXT: v_readfirstlane_b32 s4, v31
; SI-NEXT: buffer_load_dword v31, off, s[0:3], s32 offset:248
-; SI-NEXT: v_writelane_b32 v61, s4, 23
+; SI-NEXT: v_writelane_b32 v61, s4, 20
; SI-NEXT: s_waitcnt vmcnt(0)
; SI-NEXT: v_readfirstlane_b32 s4, v31
; SI-NEXT: buffer_load_dword v31, off, s[0:3], s32 offset:244
-; SI-NEXT: v_writelane_b32 v61, s4, 24
+; SI-NEXT: v_writelane_b32 v61, s4, 21
; SI-NEXT: s_waitcnt vmcnt(0)
; SI-NEXT: v_readfirstlane_b32 s4, v31
; SI-NEXT: buffer_load_dword v31, off, s[0:3], s32 offset:240
-; SI-NEXT: v_writelane_b32 v61, s4, 25
+; SI-NEXT: v_writelane_b32 v61, s4, 22
; SI-NEXT: s_waitcnt vmcnt(0)
; SI-NEXT: v_readfirstlane_b32 s4, v31
; SI-NEXT: buffer_load_dword v31, off, s[0:3], s32 offset:236
-; SI-NEXT: v_writelane_b32 v61, s4, 26
+; SI-NEXT: v_writelane_b32 v61, s4, 23
; SI-NEXT: s_waitcnt vmcnt(0)
; SI-NEXT: v_readfirstlane_b32 s4, v31
; SI-NEXT: buffer_load_dword v31, off, s[0:3], s32 offset:232
-; SI-NEXT: v_writelane_b32 v61, s4, 27
+; SI-NEXT: v_writelane_b32 v61, s4, 24
; SI-NEXT: s_waitcnt vmcnt(0)
; SI-NEXT: v_readfirstlane_b32 s4, v31
; SI-NEXT: buffer_load_dword v31, off, s[0:3], s32 offset:228
-; SI-NEXT: v_writelane_b32 v61, s4, 28
+; SI-NEXT: v_writelane_b32 v61, s4, 25
; SI-NEXT: s_waitcnt vmcnt(0)
; SI-NEXT: v_readfirstlane_b32 s4, v31
; SI-NEXT: buffer_load_dword v31, off, s[0:3], s32 offset:224
-; SI-NEXT: v_writelane_b32 v61, s4, 29
+; SI-NEXT: v_writelane_b32 v61, s4, 26
; SI-NEXT: s_waitcnt vmcnt(0)
; SI-NEXT: v_readfirstlane_b32 s4, v31
; SI-NEXT: buffer_load_dword v31, off, s[0:3], s32 offset:220
-; SI-NEXT: v_writelane_b32 v61, s4, 30
+; SI-NEXT: v_writelane_b32 v61, s4, 27
; SI-NEXT: s_waitcnt vmcnt(0)
-; SI-NEXT: v_readfirstlane_b32 s4, v31
+; SI-NEXT: v_readfirstlane_b32 s16, v31
; SI-NEXT: buffer_load_dword v31, off, s[0:3], s32 offset:216
-; SI-NEXT: v_writelane_b32 v61, s4, 31
; SI-NEXT: s_waitcnt vmcnt(0)
; SI-NEXT: v_readfirstlane_b32 s4, v31
; SI-NEXT: buffer_load_dword v31, off, s[0:3], s32 offset:212
-; SI-NEXT: v_writelane_b32 v61, s4, 32
+; SI-NEXT: v_writelane_b32 v61, s4, 28
; SI-NEXT: s_waitcnt vmcnt(0)
-; SI-NEXT: v_readfirstlane_b32 s16, v31
+; SI-NEXT: v_readfirstlane_b32 s4, v31
; SI-NEXT: buffer_load_dword v31, off, s[0:3], s32 offset:208
+; SI-NEXT: v_writelane_b32 v61, s4, 29
; SI-NEXT: s_waitcnt vmcnt(0)
; SI-NEXT: v_readfirstlane_b32 s4, v31
; SI-NEXT: buffer_load_dword v31, off, s[0:3], s32 offset:204
-; SI-NEXT: v_writelane_b32 v61, s4, 33
+; SI-NEXT: v_writelane_b32 v61, s4, 30
; SI-NEXT: s_waitcnt vmcnt(0)
; SI-NEXT: v_readfirstlane_b32 s89, v31
; SI-NEXT: buffer_load_dword v31, off, s[0:3], s32 offset:200
; SI-NEXT: s_waitcnt vmcnt(0)
; SI-NEXT: v_readfirstlane_b32 s4, v31
; SI-NEXT: buffer_load_dword v31, off, s[0:3], s32 offset:196
-; SI-NEXT: v_writelane_b32 v61, s4, 34
+; SI-NEXT: v_writelane_b32 v61, s4, 31
; SI-NEXT: s_waitcnt vmcnt(0)
-; SI-NEXT: v_readfirstlane_b32 s73, v31
+; SI-NEXT: v_readfirstlane_b32 s93, v31
; SI-NEXT: buffer_load_dword v31, off, s[0:3], s32 offset:192
; SI-NEXT: s_waitcnt vmcnt(0)
; SI-NEXT: v_readfirstlane_b32 s4, v31
; SI-NEXT: buffer_load_dword v31, off, s[0:3], s32 offset:188
-; SI-NEXT: v_writelane_b32 v61, s4, 35
+; SI-NEXT: v_writelane_b32 v61, s4, 32
; SI-NEXT: s_waitcnt vmcnt(0)
; SI-NEXT: v_readfirstlane_b32 s72, v31
; SI-NEXT: buffer_load_dword v31, off, s[0:3], s32 offset:184
@@ -185426,270 +185462,265 @@ define inreg <64 x half> @bitcast_v128i8_to_v64f16_scalar(<128 x i8> inreg %a, i
; SI-NEXT: v_readfirstlane_b32 s40, v31
; SI-NEXT: buffer_load_dword v31, off, s[0:3], s32 offset:180
; SI-NEXT: s_waitcnt vmcnt(0)
-; SI-NEXT: v_readfirstlane_b32 s21, v31
+; SI-NEXT: v_readfirstlane_b32 s97, v31
; SI-NEXT: buffer_load_dword v31, off, s[0:3], s32 offset:176
; SI-NEXT: s_waitcnt vmcnt(0)
-; SI-NEXT: v_readfirstlane_b32 s85, v31
+; SI-NEXT: v_readfirstlane_b32 s45, v31
; SI-NEXT: buffer_load_dword v31, off, s[0:3], s32 offset:172
; SI-NEXT: s_waitcnt vmcnt(0)
-; SI-NEXT: v_readfirstlane_b32 s81, v31
+; SI-NEXT: v_readfirstlane_b32 s85, v31
; SI-NEXT: buffer_load_dword v31, off, s[0:3], s32 offset:168
; SI-NEXT: s_waitcnt vmcnt(0)
-; SI-NEXT: v_readfirstlane_b32 s97, v31
+; SI-NEXT: v_readfirstlane_b32 s11, v31
; SI-NEXT: buffer_load_dword v31, off, s[0:3], s32 offset:164
; SI-NEXT: s_waitcnt vmcnt(0)
-; SI-NEXT: v_readfirstlane_b32 s7, v31
+; SI-NEXT: v_readfirstlane_b32 s4, v31
; SI-NEXT: buffer_load_dword v31, off, s[0:3], s32 offset:160
+; SI-NEXT: v_writelane_b32 v61, s4, 33
; SI-NEXT: s_waitcnt vmcnt(0)
-; SI-NEXT: v_readfirstlane_b32 s11, v31
+; SI-NEXT: v_readfirstlane_b32 s7, v31
; SI-NEXT: buffer_load_dword v31, off, s[0:3], s32 offset:156
; SI-NEXT: s_waitcnt vmcnt(0)
-; SI-NEXT: v_readfirstlane_b32 s41, v31
+; SI-NEXT: v_readfirstlane_b32 s47, v31
; SI-NEXT: buffer_load_dword v31, off, s[0:3], s32 offset:152
; SI-NEXT: s_waitcnt vmcnt(0)
-; SI-NEXT: v_readfirstlane_b32 s47, v31
+; SI-NEXT: v_readfirstlane_b32 s41, v31
; SI-NEXT: buffer_load_dword v31, off, s[0:3], s32 offset:148
; SI-NEXT: s_waitcnt vmcnt(0)
-; SI-NEXT: v_readfirstlane_b32 s58, v31
+; SI-NEXT: v_readfirstlane_b32 s59, v31
; SI-NEXT: buffer_load_dword v31, off, s[0:3], s32 offset:144
; SI-NEXT: s_waitcnt vmcnt(0)
-; SI-NEXT: v_readfirstlane_b32 s76, v31
+; SI-NEXT: v_readfirstlane_b32 s56, v31
; SI-NEXT: buffer_load_dword v31, off, s[0:3], s32 offset:140
; SI-NEXT: s_waitcnt vmcnt(0)
-; SI-NEXT: v_readfirstlane_b32 s29, v31
+; SI-NEXT: v_readfirstlane_b32 s78, v31
; SI-NEXT: buffer_load_dword v31, off, s[0:3], s32 offset:136
; SI-NEXT: s_waitcnt vmcnt(0)
-; SI-NEXT: v_readfirstlane_b32 s4, v31
+; SI-NEXT: v_readfirstlane_b32 s27, v31
; SI-NEXT: buffer_load_dword v31, off, s[0:3], s32 offset:132
-; SI-NEXT: v_writelane_b32 v61, s4, 36
-; SI-NEXT: v_writelane_b32 v61, s54, 37
-; SI-NEXT: v_writelane_b32 v61, s10, 38
-; SI-NEXT: v_writelane_b32 v61, s67, 39
-; SI-NEXT: v_writelane_b32 v61, s18, 40
-; SI-NEXT: v_writelane_b32 v61, s61, 41
-; SI-NEXT: v_writelane_b32 v61, s60, 42
-; SI-NEXT: v_writelane_b32 v61, s35, 43
-; SI-NEXT: v_writelane_b32 v61, s22, 44
-; SI-NEXT: v_writelane_b32 v61, s62, 45
-; SI-NEXT: v_writelane_b32 v61, s63, 46
-; SI-NEXT: v_writelane_b32 v61, s39, 47
-; SI-NEXT: v_writelane_b32 v61, s99, 48
-; SI-NEXT: v_writelane_b32 v61, s95, 49
-; SI-NEXT: v_writelane_b32 v61, s31, 50
-; SI-NEXT: v_writelane_b32 v61, s24, 51
-; SI-NEXT: v_writelane_b32 v61, s38, 52
-; SI-NEXT: v_writelane_b32 v61, s36, 53
-; SI-NEXT: v_writelane_b32 v61, s8, 54
-; SI-NEXT: v_writelane_b32 v61, s27, 55
-; SI-NEXT: v_writelane_b32 v61, s9, 56
-; SI-NEXT: v_writelane_b32 v61, s79, 57
-; SI-NEXT: v_writelane_b32 v61, s13, 58
-; SI-NEXT: v_writelane_b32 v61, s15, 59
-; SI-NEXT: v_writelane_b32 v61, s42, 60
-; SI-NEXT: v_writelane_b32 v61, s43, 61
-; SI-NEXT: v_writelane_b32 v61, s44, 62
-; SI-NEXT: v_writelane_b32 v61, s45, 63
-; SI-NEXT: s_and_b64 s[4:5], vcc, exec
; SI-NEXT: s_waitcnt vmcnt(0)
-; SI-NEXT: v_readfirstlane_b32 s37, v31
+; SI-NEXT: v_readfirstlane_b32 s39, v31
; SI-NEXT: buffer_load_dword v31, off, s[0:3], s32 offset:128
; SI-NEXT: s_waitcnt vmcnt(0)
-; SI-NEXT: v_readfirstlane_b32 s50, v31
+; SI-NEXT: v_readfirstlane_b32 s53, v31
; SI-NEXT: buffer_load_dword v31, off, s[0:3], s32 offset:124
; SI-NEXT: s_waitcnt vmcnt(0)
-; SI-NEXT: v_readfirstlane_b32 s48, v31
+; SI-NEXT: v_readfirstlane_b32 s50, v31
; SI-NEXT: buffer_load_dword v31, off, s[0:3], s32 offset:120
; SI-NEXT: s_waitcnt vmcnt(0)
-; SI-NEXT: v_readfirstlane_b32 s19, v31
+; SI-NEXT: v_readfirstlane_b32 s4, v31
; SI-NEXT: buffer_load_dword v31, off, s[0:3], s32 offset:116
+; SI-NEXT: v_writelane_b32 v61, s4, 34
; SI-NEXT: s_waitcnt vmcnt(0)
-; SI-NEXT: v_readfirstlane_b32 s64, v31
+; SI-NEXT: v_readfirstlane_b32 s29, v31
; SI-NEXT: buffer_load_dword v31, off, s[0:3], s32 offset:112
; SI-NEXT: s_waitcnt vmcnt(0)
-; SI-NEXT: v_readfirstlane_b32 s17, v31
+; SI-NEXT: v_readfirstlane_b32 s26, v31
; SI-NEXT: buffer_load_dword v31, off, s[0:3], s32 offset:108
; SI-NEXT: s_waitcnt vmcnt(0)
-; SI-NEXT: v_readfirstlane_b32 s65, v31
+; SI-NEXT: v_readfirstlane_b32 s23, v31
; SI-NEXT: buffer_load_dword v31, off, s[0:3], s32 offset:104
; SI-NEXT: s_waitcnt vmcnt(0)
-; SI-NEXT: v_readfirstlane_b32 s71, v31
+; SI-NEXT: v_readfirstlane_b32 s4, v31
; SI-NEXT: buffer_load_dword v31, off, s[0:3], s32 offset:100
+; SI-NEXT: v_writelane_b32 v61, s4, 35
+; SI-NEXT: v_writelane_b32 v61, s67, 36
+; SI-NEXT: v_writelane_b32 v61, s10, 37
+; SI-NEXT: v_writelane_b32 v61, s65, 38
+; SI-NEXT: v_writelane_b32 v61, s18, 39
+; SI-NEXT: v_writelane_b32 v61, s21, 40
+; SI-NEXT: v_writelane_b32 v61, s20, 41
+; SI-NEXT: v_writelane_b32 v61, s64, 42
+; SI-NEXT: v_writelane_b32 v61, s22, 43
+; SI-NEXT: v_writelane_b32 v61, s77, 44
+; SI-NEXT: v_writelane_b32 v61, s92, 45
+; SI-NEXT: v_writelane_b32 v61, s54, 46
+; SI-NEXT: v_writelane_b32 v61, s66, 47
+; SI-NEXT: v_writelane_b32 v61, s79, 48
+; SI-NEXT: v_writelane_b32 v61, s31, 49
+; SI-NEXT: v_writelane_b32 v61, s28, 50
+; SI-NEXT: v_writelane_b32 v61, s55, 51
+; SI-NEXT: v_writelane_b32 v61, s80, 52
+; SI-NEXT: v_writelane_b32 v61, s51, 53
+; SI-NEXT: v_writelane_b32 v61, s36, 54
+; SI-NEXT: v_writelane_b32 v61, s87, 55
+; SI-NEXT: v_writelane_b32 v61, s84, 56
+; SI-NEXT: v_writelane_b32 v61, s13, 57
+; SI-NEXT: v_writelane_b32 v61, s15, 58
+; SI-NEXT: v_writelane_b32 v61, s42, 59
+; SI-NEXT: v_writelane_b32 v61, s43, 60
+; SI-NEXT: v_writelane_b32 v61, s44, 61
+; SI-NEXT: v_writelane_b32 v61, s73, 62
+; SI-NEXT: v_writelane_b32 v61, s74, 63
+; SI-NEXT: s_and_b64 s[4:5], vcc, exec
; SI-NEXT: s_waitcnt vmcnt(0)
-; SI-NEXT: v_readfirstlane_b32 s70, v31
+; SI-NEXT: v_readfirstlane_b32 s19, v31
; SI-NEXT: buffer_load_dword v31, off, s[0:3], s32 offset:96
; SI-NEXT: s_waitcnt vmcnt(0)
-; SI-NEXT: v_readfirstlane_b32 s83, v31
+; SI-NEXT: v_readfirstlane_b32 s71, v31
; SI-NEXT: buffer_load_dword v31, off, s[0:3], s32 offset:92
; SI-NEXT: s_waitcnt vmcnt(0)
-; SI-NEXT: v_readfirstlane_b32 s49, v31
+; SI-NEXT: v_readfirstlane_b32 s17, v31
; SI-NEXT: buffer_load_dword v31, off, s[0:3], s32 offset:88
; SI-NEXT: s_waitcnt vmcnt(0)
-; SI-NEXT: v_readfirstlane_b32 s80, v31
+; SI-NEXT: v_readfirstlane_b32 s70, v31
; SI-NEXT: buffer_load_dword v31, off, s[0:3], s32 offset:84
; SI-NEXT: s_waitcnt vmcnt(0)
-; SI-NEXT: v_readfirstlane_b32 s82, v31
+; SI-NEXT: v_readfirstlane_b32 s37, v31
; SI-NEXT: buffer_load_dword v31, off, s[0:3], s32 offset:80
; SI-NEXT: s_waitcnt vmcnt(0)
-; SI-NEXT: v_readfirstlane_b32 s87, v31
+; SI-NEXT: v_readfirstlane_b32 s82, v31
; SI-NEXT: buffer_load_dword v31, off, s[0:3], s32 offset:76
; SI-NEXT: s_waitcnt vmcnt(0)
-; SI-NEXT: v_readfirstlane_b32 s84, v31
+; SI-NEXT: v_readfirstlane_b32 s83, v31
; SI-NEXT: buffer_load_dword v31, off, s[0:3], s32 offset:72
; SI-NEXT: s_waitcnt vmcnt(0)
-; SI-NEXT: v_readfirstlane_b32 s51, v31
+; SI-NEXT: v_readfirstlane_b32 s86, v31
; SI-NEXT: buffer_load_dword v31, off, s[0:3], s32 offset:68
; SI-NEXT: s_waitcnt vmcnt(0)
-; SI-NEXT: v_readfirstlane_b32 s86, v31
+; SI-NEXT: v_readfirstlane_b32 s30, v31
; SI-NEXT: buffer_load_dword v31, off, s[0:3], s32 offset:64
; SI-NEXT: s_waitcnt vmcnt(0)
-; SI-NEXT: v_readfirstlane_b32 s94, v31
+; SI-NEXT: v_readfirstlane_b32 s96, v31
; SI-NEXT: buffer_load_dword v31, off, s[0:3], s32 offset:60
; SI-NEXT: s_waitcnt vmcnt(0)
-; SI-NEXT: v_readfirstlane_b32 s96, v31
+; SI-NEXT: v_readfirstlane_b32 s48, v31
; SI-NEXT: buffer_load_dword v31, off, s[0:3], s32 offset:56
; SI-NEXT: s_waitcnt vmcnt(0)
-; SI-NEXT: v_readfirstlane_b32 s68, v31
+; SI-NEXT: v_readfirstlane_b32 s98, v31
; SI-NEXT: buffer_load_dword v31, off, s[0:3], s32 offset:52
; SI-NEXT: s_waitcnt vmcnt(0)
-; SI-NEXT: v_readfirstlane_b32 s34, v31
+; SI-NEXT: v_readfirstlane_b32 s38, v31
; SI-NEXT: buffer_load_dword v31, off, s[0:3], s32 offset:48
; SI-NEXT: s_waitcnt vmcnt(0)
-; SI-NEXT: v_readfirstlane_b32 s77, v31
+; SI-NEXT: v_readfirstlane_b32 s68, v31
; SI-NEXT: buffer_load_dword v31, off, s[0:3], s32 offset:44
; SI-NEXT: s_waitcnt vmcnt(0)
-; SI-NEXT: v_readfirstlane_b32 s66, v31
+; SI-NEXT: v_readfirstlane_b32 s99, v31
; SI-NEXT: buffer_load_dword v31, off, s[0:3], s32 offset:40
; SI-NEXT: s_waitcnt vmcnt(0)
-; SI-NEXT: v_readfirstlane_b32 s78, v31
+; SI-NEXT: v_readfirstlane_b32 s69, v31
; SI-NEXT: buffer_load_dword v31, off, s[0:3], s32 offset:36
; SI-NEXT: s_waitcnt vmcnt(0)
-; SI-NEXT: v_readfirstlane_b32 s53, v31
+; SI-NEXT: v_readfirstlane_b32 s49, v31
; SI-NEXT: buffer_load_dword v31, off, s[0:3], s32 offset:32
; SI-NEXT: s_waitcnt vmcnt(0)
-; SI-NEXT: v_readfirstlane_b32 s69, v31
+; SI-NEXT: v_readfirstlane_b32 s6, v31
; SI-NEXT: buffer_load_dword v31, off, s[0:3], s32 offset:28
; SI-NEXT: s_waitcnt vmcnt(0)
-; SI-NEXT: v_readfirstlane_b32 s30, v31
+; SI-NEXT: v_readfirstlane_b32 s90, v31
; SI-NEXT: buffer_load_dword v31, off, s[0:3], s32 offset:24
; SI-NEXT: s_waitcnt vmcnt(0)
-; SI-NEXT: v_readfirstlane_b32 s52, v31
+; SI-NEXT: v_readfirstlane_b32 s34, v31
; SI-NEXT: buffer_load_dword v31, off, s[0:3], s32 offset:20
; SI-NEXT: s_waitcnt vmcnt(0)
-; SI-NEXT: v_readfirstlane_b32 s75, v31
+; SI-NEXT: v_readfirstlane_b32 s52, v31
; SI-NEXT: buffer_load_dword v31, off, s[0:3], s32 offset:16
; SI-NEXT: s_waitcnt vmcnt(0)
-; SI-NEXT: v_readfirstlane_b32 s23, v31
+; SI-NEXT: v_readfirstlane_b32 s88, v31
; SI-NEXT: buffer_load_dword v31, off, s[0:3], s32 offset:12
; SI-NEXT: s_waitcnt vmcnt(0)
-; SI-NEXT: v_readfirstlane_b32 s28, v31
+; SI-NEXT: v_readfirstlane_b32 s8, v31
; SI-NEXT: buffer_load_dword v31, off, s[0:3], s32 offset:8
; SI-NEXT: s_waitcnt vmcnt(0)
-; SI-NEXT: v_readfirstlane_b32 s26, v31
+; SI-NEXT: v_readfirstlane_b32 s24, v31
; SI-NEXT: buffer_load_dword v31, off, s[0:3], s32 offset:4
; SI-NEXT: s_waitcnt vmcnt(0)
-; SI-NEXT: v_readfirstlane_b32 s25, v31
+; SI-NEXT: v_readfirstlane_b32 s76, v31
; SI-NEXT: buffer_load_dword v31, off, s[0:3], s32
-; SI-NEXT: v_writelane_b32 v62, s25, 8
-; SI-NEXT: v_writelane_b32 v62, s28, 9
+; SI-NEXT: v_writelane_b32 v62, s76, 7
+; SI-NEXT: v_writelane_b32 v62, s8, 8
; SI-NEXT: s_waitcnt vmcnt(0)
-; SI-NEXT: v_readfirstlane_b32 s92, v31
-; SI-NEXT: v_writelane_b32 v62, s92, 10
-; SI-NEXT: v_writelane_b32 v62, s75, 11
-; SI-NEXT: v_writelane_b32 v62, s26, 12
-; SI-NEXT: v_writelane_b32 v62, s30, 13
-; SI-NEXT: v_writelane_b32 v62, s23, 14
-; SI-NEXT: v_writelane_b32 v62, s52, 15
-; SI-NEXT: v_writelane_b32 v62, s64, 16
-; SI-NEXT: v_writelane_b32 v62, s17, 17
-; SI-NEXT: v_writelane_b32 v62, s65, 18
-; SI-NEXT: v_writelane_b32 v62, s70, 19
-; SI-NEXT: v_writelane_b32 v62, s71, 20
-; SI-NEXT: v_writelane_b32 v62, s49, 21
-; SI-NEXT: v_writelane_b32 v62, s83, 22
-; SI-NEXT: v_writelane_b32 v62, s80, 23
-; SI-NEXT: v_writelane_b32 v62, s82, 24
-; SI-NEXT: v_writelane_b32 v62, s84, 25
-; SI-NEXT: v_writelane_b32 v62, s87, 26
-; SI-NEXT: v_writelane_b32 v62, s86, 27
-; SI-NEXT: v_writelane_b32 v62, s51, 28
-; SI-NEXT: v_writelane_b32 v62, s96, 29
-; SI-NEXT: v_writelane_b32 v62, s34, 30
-; SI-NEXT: v_writelane_b32 v62, s94, 31
-; SI-NEXT: v_writelane_b32 v62, s53, 32
-; SI-NEXT: v_writelane_b32 v62, s66, 33
-; SI-NEXT: v_writelane_b32 v62, s68, 34
-; SI-NEXT: v_writelane_b32 v62, s69, 35
-; SI-NEXT: v_writelane_b32 v62, s77, 36
-; SI-NEXT: v_writelane_b32 v62, s78, 37
-; SI-NEXT: s_cbranch_scc0 .LBB93_4
+; SI-NEXT: v_readfirstlane_b32 s75, v31
+; SI-NEXT: v_writelane_b32 v62, s75, 9
+; SI-NEXT: v_writelane_b32 v62, s52, 10
+; SI-NEXT: v_writelane_b32 v62, s24, 11
+; SI-NEXT: v_writelane_b32 v62, s90, 12
+; SI-NEXT: v_writelane_b32 v62, s88, 13
+; SI-NEXT: v_writelane_b32 v62, s34, 14
+; SI-NEXT: v_writelane_b32 v62, s17, 15
+; SI-NEXT: v_writelane_b32 v62, s71, 16
+; SI-NEXT: v_writelane_b32 v62, s70, 17
+; SI-NEXT: v_writelane_b32 v62, s37, 18
+; SI-NEXT: v_writelane_b32 v62, s83, 19
+; SI-NEXT: v_writelane_b32 v62, s82, 20
+; SI-NEXT: v_writelane_b32 v62, s30, 21
+; SI-NEXT: v_writelane_b32 v62, s86, 22
+; SI-NEXT: v_writelane_b32 v62, s48, 23
+; SI-NEXT: v_writelane_b32 v62, s38, 24
+; SI-NEXT: v_writelane_b32 v62, s96, 25
+; SI-NEXT: v_writelane_b32 v62, s49, 26
+; SI-NEXT: v_writelane_b32 v62, s99, 27
+; SI-NEXT: v_writelane_b32 v62, s98, 28
+; SI-NEXT: v_writelane_b32 v62, s6, 29
+; SI-NEXT: v_writelane_b32 v62, s68, 30
+; SI-NEXT: v_writelane_b32 v62, s69, 31
+; SI-NEXT: s_cbranch_scc0 .LBB93_2
; SI-NEXT: ; %bb.1: ; %cmp.false
; SI-NEXT: s_and_b32 s4, s10, 0xff
-; SI-NEXT: s_lshl_b32 s5, s54, 8
+; SI-NEXT: s_lshl_b32 s5, s67, 8
; SI-NEXT: s_or_b32 s4, s4, s5
; SI-NEXT: v_cvt_f32_f16_e32 v6, s4
; SI-NEXT: s_and_b32 s4, s18, 0xff
-; SI-NEXT: s_lshl_b32 s5, s67, 8
+; SI-NEXT: s_lshl_b32 s5, s65, 8
; SI-NEXT: s_or_b32 s4, s4, s5
; SI-NEXT: v_cvt_f32_f16_e32 v5, s4
-; SI-NEXT: s_and_b32 s4, s60, 0xff
-; SI-NEXT: s_lshl_b32 s5, s61, 8
+; SI-NEXT: s_and_b32 s4, s20, 0xff
+; SI-NEXT: s_lshl_b32 s5, s21, 8
; SI-NEXT: s_or_b32 s4, s4, s5
; SI-NEXT: v_cvt_f32_f16_e32 v1, s4
; SI-NEXT: s_and_b32 s4, s22, 0xff
-; SI-NEXT: s_lshl_b32 s5, s35, 8
+; SI-NEXT: s_lshl_b32 s5, s64, 8
; SI-NEXT: s_or_b32 s4, s4, s5
; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:392 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_cvt_f32_f16_e32 v1, s4
-; SI-NEXT: s_and_b32 s4, s63, 0xff
-; SI-NEXT: s_lshl_b32 s5, s62, 8
+; SI-NEXT: s_and_b32 s4, s92, 0xff
+; SI-NEXT: s_lshl_b32 s5, s77, 8
; SI-NEXT: s_or_b32 s4, s4, s5
; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:400 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_cvt_f32_f16_e32 v1, s4
-; SI-NEXT: v_readlane_b32 s5, v61, 2
-; SI-NEXT: s_and_b32 s4, s39, 0xff
-; SI-NEXT: s_lshl_b32 s5, s5, 8
+; SI-NEXT: s_and_b32 s4, s66, 0xff
+; SI-NEXT: s_lshl_b32 s5, s54, 8
+; SI-NEXT: s_or_b32 s4, s4, s5
+; SI-NEXT: v_cvt_f32_f16_e32 v7, s4
+; SI-NEXT: s_and_b32 s4, s28, 0xff
+; SI-NEXT: s_lshl_b32 s5, s79, 8
; SI-NEXT: s_or_b32 s4, s4, s5
; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:396 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_cvt_f32_f16_e32 v1, s4
-; SI-NEXT: v_readlane_b32 s4, v61, 1
-; SI-NEXT: v_readlane_b32 s5, v61, 0
-; SI-NEXT: s_and_b32 s4, s4, 0xff
-; SI-NEXT: s_lshl_b32 s5, s5, 8
-; SI-NEXT: s_or_b32 s4, s4, s5
-; SI-NEXT: v_cvt_f32_f16_e32 v7, s4
-; SI-NEXT: s_and_b32 s4, s99, 0xff
-; SI-NEXT: s_lshl_b32 s5, s55, 8
+; SI-NEXT: s_and_b32 s4, s31, 0xff
+; SI-NEXT: s_lshl_b32 s5, s95, 8
; SI-NEXT: s_or_b32 s4, s4, s5
; SI-NEXT: v_cvt_f32_f16_e32 v10, s4
-; SI-NEXT: s_and_b32 s4, s95, 0xff
-; SI-NEXT: s_lshl_b32 s5, s93, 8
+; SI-NEXT: s_and_b32 s4, s55, 0xff
+; SI-NEXT: s_lshl_b32 s5, s91, 8
; SI-NEXT: s_or_b32 s4, s4, s5
; SI-NEXT: v_cvt_f32_f16_e32 v8, s4
-; SI-NEXT: s_and_b32 s4, s31, 0xff
-; SI-NEXT: s_lshl_b32 s5, s91, 8
+; SI-NEXT: s_and_b32 s4, s80, 0xff
+; SI-NEXT: s_lshl_b32 s5, s35, 8
; SI-NEXT: s_or_b32 s4, s4, s5
-; SI-NEXT: v_cvt_f32_f16_e32 v11, s4
-; SI-NEXT: s_and_b32 s4, s88, 0xff
-; SI-NEXT: s_lshl_b32 s5, s90, 8
+; SI-NEXT: v_cvt_f32_f16_e32 v12, s4
+; SI-NEXT: s_and_b32 s4, s51, 0xff
+; SI-NEXT: s_lshl_b32 s5, s60, 8
; SI-NEXT: s_or_b32 s4, s4, s5
; SI-NEXT: v_cvt_f32_f16_e32 v9, s4
-; SI-NEXT: s_and_b32 s4, s24, 0xff
-; SI-NEXT: s_lshl_b32 s5, s98, 8
+; SI-NEXT: s_and_b32 s4, s94, 0xff
+; SI-NEXT: s_lshl_b32 s5, s61, 8
; SI-NEXT: s_or_b32 s4, s4, s5
; SI-NEXT: v_cvt_f32_f16_e32 v13, s4
; SI-NEXT: s_and_b32 s4, s36, 0xff
-; SI-NEXT: s_lshl_b32 s5, s38, 8
+; SI-NEXT: s_lshl_b32 s5, s63, 8
; SI-NEXT: s_or_b32 s4, s4, s5
-; SI-NEXT: v_cvt_f32_f16_e32 v12, s4
-; SI-NEXT: s_and_b32 s4, s27, 0xff
-; SI-NEXT: s_lshl_b32 s5, s8, 8
+; SI-NEXT: v_cvt_f32_f16_e32 v11, s4
+; SI-NEXT: s_and_b32 s4, s9, 0xff
+; SI-NEXT: s_lshl_b32 s5, s62, 8
; SI-NEXT: s_or_b32 s4, s4, s5
; SI-NEXT: v_cvt_f32_f16_e32 v15, s4
-; SI-NEXT: s_and_b32 s4, s79, 0xff
-; SI-NEXT: s_lshl_b32 s5, s9, 8
+; SI-NEXT: s_and_b32 s4, s84, 0xff
+; SI-NEXT: s_lshl_b32 s5, s87, 8
; SI-NEXT: s_or_b32 s4, s4, s5
; SI-NEXT: v_cvt_f32_f16_e32 v14, s4
; SI-NEXT: s_and_b32 s4, s15, 0xff
@@ -185700,11 +185731,11 @@ define inreg <64 x half> @bitcast_v128i8_to_v64f16_scalar(<128 x i8> inreg %a, i
; SI-NEXT: s_lshl_b32 s5, s42, 8
; SI-NEXT: s_or_b32 s4, s4, s5
; SI-NEXT: v_cvt_f32_f16_e32 v16, s4
-; SI-NEXT: s_and_b32 s4, s45, 0xff
+; SI-NEXT: s_and_b32 s4, s73, 0xff
; SI-NEXT: s_lshl_b32 s5, s44, 8
; SI-NEXT: s_or_b32 s4, s4, s5
; SI-NEXT: v_cvt_f32_f16_e32 v19, s4
-; SI-NEXT: s_and_b32 s4, s6, 0xff
+; SI-NEXT: s_and_b32 s4, s81, 0xff
; SI-NEXT: s_lshl_b32 s5, s74, 8
; SI-NEXT: s_or_b32 s4, s4, s5
; SI-NEXT: v_cvt_f32_f16_e32 v18, s4
@@ -185712,244 +185743,393 @@ define inreg <64 x half> @bitcast_v128i8_to_v64f16_scalar(<128 x i8> inreg %a, i
; SI-NEXT: s_lshl_b32 s5, s12, 8
; SI-NEXT: s_or_b32 s4, s4, s5
; SI-NEXT: v_cvt_f32_f16_e32 v21, s4
-; SI-NEXT: s_and_b32 s4, s56, 0xff
+; SI-NEXT: s_and_b32 s4, s57, 0xff
; SI-NEXT: s_lshl_b32 s5, s46, 8
; SI-NEXT: s_or_b32 s4, s4, s5
; SI-NEXT: v_cvt_f32_f16_e32 v20, s4
-; SI-NEXT: s_and_b32 s4, s59, 0xff
-; SI-NEXT: s_lshl_b32 s5, s57, 8
+; SI-NEXT: s_and_b32 s4, s25, 0xff
+; SI-NEXT: s_lshl_b32 s5, s58, 8
; SI-NEXT: s_or_b32 s4, s4, s5
; SI-NEXT: v_cvt_f32_f16_e32 v23, s4
-; SI-NEXT: s_and_b32 s4, s92, 0xff
-; SI-NEXT: s_lshl_b32 s5, s25, 8
+; SI-NEXT: s_and_b32 s4, s75, 0xff
+; SI-NEXT: s_lshl_b32 s5, s76, 8
; SI-NEXT: s_or_b32 s4, s4, s5
; SI-NEXT: v_cvt_f32_f16_e32 v22, s4
-; SI-NEXT: s_and_b32 s4, s26, 0xff
-; SI-NEXT: s_lshl_b32 s5, s28, 8
+; SI-NEXT: s_and_b32 s4, s24, 0xff
+; SI-NEXT: s_lshl_b32 s5, s8, 8
; SI-NEXT: s_or_b32 s4, s4, s5
; SI-NEXT: v_cvt_f32_f16_e32 v25, s4
-; SI-NEXT: s_and_b32 s4, s23, 0xff
-; SI-NEXT: s_lshl_b32 s5, s75, 8
+; SI-NEXT: s_and_b32 s4, s88, 0xff
+; SI-NEXT: s_lshl_b32 s5, s52, 8
; SI-NEXT: s_or_b32 s4, s4, s5
; SI-NEXT: v_cvt_f32_f16_e32 v24, s4
-; SI-NEXT: s_and_b32 s4, s52, 0xff
-; SI-NEXT: s_lshl_b32 s5, s30, 8
+; SI-NEXT: s_and_b32 s4, s34, 0xff
+; SI-NEXT: s_lshl_b32 s5, s90, 8
; SI-NEXT: s_or_b32 s4, s4, s5
; SI-NEXT: v_cvt_f32_f16_e32 v27, s4
-; SI-NEXT: s_and_b32 s4, s69, 0xff
-; SI-NEXT: s_lshl_b32 s5, s53, 8
+; SI-NEXT: s_and_b32 s4, s6, 0xff
+; SI-NEXT: s_lshl_b32 s5, s49, 8
; SI-NEXT: s_or_b32 s4, s4, s5
; SI-NEXT: v_cvt_f32_f16_e32 v26, s4
-; SI-NEXT: s_and_b32 s4, s78, 0xff
-; SI-NEXT: s_lshl_b32 s5, s66, 8
+; SI-NEXT: s_and_b32 s4, s69, 0xff
+; SI-NEXT: s_lshl_b32 s5, s99, 8
; SI-NEXT: s_or_b32 s4, s4, s5
; SI-NEXT: v_cvt_f32_f16_e32 v29, s4
-; SI-NEXT: s_and_b32 s4, s77, 0xff
-; SI-NEXT: s_lshl_b32 s5, s34, 8
+; SI-NEXT: s_and_b32 s4, s68, 0xff
+; SI-NEXT: s_lshl_b32 s5, s38, 8
; SI-NEXT: s_or_b32 s4, s4, s5
; SI-NEXT: v_cvt_f32_f16_e32 v28, s4
-; SI-NEXT: s_and_b32 s4, s68, 0xff
-; SI-NEXT: s_lshl_b32 s5, s96, 8
+; SI-NEXT: s_and_b32 s4, s98, 0xff
+; SI-NEXT: s_lshl_b32 s5, s48, 8
; SI-NEXT: s_or_b32 s4, s4, s5
; SI-NEXT: v_cvt_f32_f16_e32 v31, s4
-; SI-NEXT: s_and_b32 s4, s94, 0xff
-; SI-NEXT: s_lshl_b32 s5, s86, 8
+; SI-NEXT: s_and_b32 s4, s96, 0xff
+; SI-NEXT: s_lshl_b32 s5, s30, 8
; SI-NEXT: s_or_b32 s4, s4, s5
; SI-NEXT: v_cvt_f32_f16_e32 v30, s4
-; SI-NEXT: s_and_b32 s4, s51, 0xff
-; SI-NEXT: s_lshl_b32 s5, s84, 8
+; SI-NEXT: s_and_b32 s4, s86, 0xff
+; SI-NEXT: s_lshl_b32 s5, s83, 8
; SI-NEXT: s_or_b32 s4, s4, s5
; SI-NEXT: v_cvt_f32_f16_e32 v33, s4
-; SI-NEXT: s_and_b32 s4, s87, 0xff
-; SI-NEXT: s_lshl_b32 s5, s82, 8
+; SI-NEXT: s_and_b32 s4, s82, 0xff
+; SI-NEXT: s_lshl_b32 s5, s37, 8
; SI-NEXT: s_or_b32 s4, s4, s5
; SI-NEXT: v_cvt_f32_f16_e32 v32, s4
-; SI-NEXT: s_and_b32 s4, s80, 0xff
-; SI-NEXT: s_lshl_b32 s5, s49, 8
+; SI-NEXT: s_and_b32 s4, s70, 0xff
+; SI-NEXT: s_lshl_b32 s5, s17, 8
; SI-NEXT: s_or_b32 s4, s4, s5
; SI-NEXT: v_cvt_f32_f16_e32 v35, s4
-; SI-NEXT: s_and_b32 s4, s83, 0xff
-; SI-NEXT: s_lshl_b32 s5, s70, 8
+; SI-NEXT: s_and_b32 s4, s71, 0xff
+; SI-NEXT: s_lshl_b32 s5, s19, 8
; SI-NEXT: s_or_b32 s4, s4, s5
+; SI-NEXT: v_readlane_b32 s13, v61, 35
; SI-NEXT: v_cvt_f32_f16_e32 v34, s4
-; SI-NEXT: s_and_b32 s4, s71, 0xff
-; SI-NEXT: s_lshl_b32 s5, s65, 8
+; SI-NEXT: s_and_b32 s4, s13, 0xff
+; SI-NEXT: s_lshl_b32 s5, s23, 8
; SI-NEXT: s_or_b32 s4, s4, s5
; SI-NEXT: v_cvt_f32_f16_e32 v37, s4
-; SI-NEXT: s_and_b32 s4, s17, 0xff
-; SI-NEXT: s_lshl_b32 s5, s64, 8
+; SI-NEXT: s_and_b32 s4, s26, 0xff
+; SI-NEXT: s_lshl_b32 s5, s29, 8
+; SI-NEXT: s_mov_b32 s79, s9
; SI-NEXT: s_or_b32 s4, s4, s5
+; SI-NEXT: v_readlane_b32 s9, v61, 34
; SI-NEXT: v_cvt_f32_f16_e32 v36, s4
-; SI-NEXT: s_and_b32 s4, s19, 0xff
-; SI-NEXT: s_lshl_b32 s5, s48, 8
+; SI-NEXT: s_and_b32 s4, s9, 0xff
+; SI-NEXT: s_lshl_b32 s5, s50, 8
; SI-NEXT: s_or_b32 s4, s4, s5
; SI-NEXT: v_cvt_f32_f16_e32 v39, s4
-; SI-NEXT: s_and_b32 s4, s50, 0xff
-; SI-NEXT: s_lshl_b32 s5, s37, 8
+; SI-NEXT: s_and_b32 s4, s53, 0xff
+; SI-NEXT: s_lshl_b32 s5, s39, 8
; SI-NEXT: s_or_b32 s4, s4, s5
-; SI-NEXT: v_readlane_b32 s8, v61, 36
; SI-NEXT: v_cvt_f32_f16_e32 v38, s4
-; SI-NEXT: s_and_b32 s4, s8, 0xff
-; SI-NEXT: s_lshl_b32 s5, s29, 8
+; SI-NEXT: s_and_b32 s4, s27, 0xff
+; SI-NEXT: s_lshl_b32 s5, s78, 8
; SI-NEXT: s_or_b32 s4, s4, s5
; SI-NEXT: v_cvt_f32_f16_e32 v49, s4
-; SI-NEXT: s_and_b32 s4, s76, 0xff
-; SI-NEXT: s_lshl_b32 s5, s58, 8
+; SI-NEXT: s_and_b32 s4, s56, 0xff
+; SI-NEXT: s_lshl_b32 s5, s59, 8
; SI-NEXT: s_or_b32 s4, s4, s5
; SI-NEXT: v_cvt_f32_f16_e32 v48, s4
-; SI-NEXT: s_and_b32 s4, s47, 0xff
-; SI-NEXT: s_lshl_b32 s5, s41, 8
+; SI-NEXT: s_and_b32 s4, s41, 0xff
+; SI-NEXT: s_lshl_b32 s5, s47, 8
; SI-NEXT: s_or_b32 s4, s4, s5
+; SI-NEXT: v_readlane_b32 s21, v61, 33
; SI-NEXT: v_cvt_f32_f16_e32 v51, s4
-; SI-NEXT: s_and_b32 s4, s11, 0xff
-; SI-NEXT: s_lshl_b32 s5, s7, 8
+; SI-NEXT: s_and_b32 s4, s7, 0xff
+; SI-NEXT: s_lshl_b32 s5, s21, 8
; SI-NEXT: s_or_b32 s4, s4, s5
; SI-NEXT: v_cvt_f32_f16_e32 v50, s4
-; SI-NEXT: s_and_b32 s4, s97, 0xff
-; SI-NEXT: s_lshl_b32 s5, s81, 8
+; SI-NEXT: s_and_b32 s4, s11, 0xff
+; SI-NEXT: s_lshl_b32 s5, s85, 8
; SI-NEXT: s_or_b32 s4, s4, s5
; SI-NEXT: v_cvt_f32_f16_e32 v53, s4
-; SI-NEXT: s_and_b32 s4, s85, 0xff
-; SI-NEXT: s_lshl_b32 s5, s21, 8
+; SI-NEXT: s_and_b32 s4, s45, 0xff
+; SI-NEXT: s_lshl_b32 s5, s97, 8
; SI-NEXT: s_or_b32 s4, s4, s5
; SI-NEXT: v_cvt_f32_f16_e32 v52, s4
; SI-NEXT: s_and_b32 s4, s40, 0xff
; SI-NEXT: s_lshl_b32 s5, s72, 8
; SI-NEXT: s_or_b32 s4, s4, s5
-; SI-NEXT: v_readlane_b32 s69, v61, 35
+; SI-NEXT: v_readlane_b32 s69, v61, 32
; SI-NEXT: v_cvt_f32_f16_e32 v55, s4
; SI-NEXT: s_and_b32 s4, s69, 0xff
-; SI-NEXT: s_lshl_b32 s5, s73, 8
+; SI-NEXT: s_lshl_b32 s5, s93, 8
; SI-NEXT: s_or_b32 s4, s4, s5
-; SI-NEXT: v_readlane_b32 s68, v61, 34
+; SI-NEXT: v_readlane_b32 s68, v61, 31
; SI-NEXT: v_cvt_f32_f16_e32 v54, s4
; SI-NEXT: s_and_b32 s4, s68, 0xff
; SI-NEXT: s_lshl_b32 s5, s89, 8
; SI-NEXT: s_or_b32 s4, s4, s5
-; SI-NEXT: v_readlane_b32 s66, v61, 33
+; SI-NEXT: v_readlane_b32 s66, v61, 30
+; SI-NEXT: v_readlane_b32 s20, v61, 29
; SI-NEXT: v_cvt_f32_f16_e32 v41, s4
; SI-NEXT: s_and_b32 s4, s66, 0xff
-; SI-NEXT: s_lshl_b32 s5, s16, 8
+; SI-NEXT: s_lshl_b32 s5, s20, 8
+; SI-NEXT: s_mov_b32 s17, s19
+; SI-NEXT: s_mov_b32 s19, s23
+; SI-NEXT: s_mov_b32 s23, s26
+; SI-NEXT: s_mov_b32 s26, s29
+; SI-NEXT: s_mov_b32 s29, s53
; SI-NEXT: s_or_b32 s4, s4, s5
-; SI-NEXT: v_readlane_b32 s53, v61, 32
-; SI-NEXT: v_readlane_b32 s94, v61, 31
+; SI-NEXT: v_readlane_b32 s53, v61, 28
; SI-NEXT: v_cvt_f32_f16_e32 v40, s4
; SI-NEXT: s_and_b32 s4, s53, 0xff
-; SI-NEXT: s_lshl_b32 s5, s94, 8
+; SI-NEXT: s_lshl_b32 s5, s16, 8
; SI-NEXT: s_or_b32 s4, s4, s5
-; SI-NEXT: v_readlane_b32 s34, v61, 30
-; SI-NEXT: v_readlane_b32 s96, v61, 29
+; SI-NEXT: v_readlane_b32 s34, v61, 27
+; SI-NEXT: v_readlane_b32 s6, v61, 26
; SI-NEXT: v_cvt_f32_f16_e32 v43, s4
; SI-NEXT: s_and_b32 s4, s34, 0xff
-; SI-NEXT: s_lshl_b32 s5, s96, 8
+; SI-NEXT: s_lshl_b32 s5, s6, 8
; SI-NEXT: s_or_b32 s4, s4, s5
-; SI-NEXT: v_readlane_b32 s51, v61, 28
-; SI-NEXT: v_readlane_b32 s86, v61, 27
+; SI-NEXT: v_readlane_b32 s98, v61, 25
+; SI-NEXT: v_readlane_b32 s99, v61, 24
; SI-NEXT: v_cvt_f32_f16_e32 v42, s4
-; SI-NEXT: s_and_b32 s4, s51, 0xff
-; SI-NEXT: s_lshl_b32 s5, s86, 8
+; SI-NEXT: s_and_b32 s4, s98, 0xff
+; SI-NEXT: s_lshl_b32 s5, s99, 8
; SI-NEXT: s_or_b32 s4, s4, s5
-; SI-NEXT: v_readlane_b32 s87, v61, 26
-; SI-NEXT: v_readlane_b32 s84, v61, 25
+; SI-NEXT: v_readlane_b32 s49, v61, 23
+; SI-NEXT: v_readlane_b32 s96, v61, 22
; SI-NEXT: v_cvt_f32_f16_e32 v45, s4
-; SI-NEXT: s_and_b32 s4, s87, 0xff
-; SI-NEXT: s_lshl_b32 s5, s84, 8
+; SI-NEXT: s_and_b32 s4, s49, 0xff
+; SI-NEXT: s_lshl_b32 s5, s96, 8
; SI-NEXT: s_or_b32 s4, s4, s5
-; SI-NEXT: v_readlane_b32 s82, v61, 24
-; SI-NEXT: v_readlane_b32 s80, v61, 23
+; SI-NEXT: v_readlane_b32 s38, v61, 21
+; SI-NEXT: v_readlane_b32 s48, v61, 20
; SI-NEXT: v_cvt_f32_f16_e32 v44, s4
-; SI-NEXT: s_and_b32 s4, s82, 0xff
-; SI-NEXT: s_lshl_b32 s5, s80, 8
+; SI-NEXT: s_and_b32 s4, s38, 0xff
+; SI-NEXT: s_lshl_b32 s5, s48, 8
; SI-NEXT: s_or_b32 s4, s4, s5
-; SI-NEXT: v_readlane_b32 s83, v61, 22
-; SI-NEXT: v_readlane_b32 s49, v61, 21
+; SI-NEXT: v_readlane_b32 s86, v61, 19
+; SI-NEXT: v_readlane_b32 s30, v61, 18
; SI-NEXT: v_cvt_f32_f16_e32 v47, s4
-; SI-NEXT: s_and_b32 s4, s83, 0xff
-; SI-NEXT: s_lshl_b32 s5, s49, 8
+; SI-NEXT: s_and_b32 s4, s86, 0xff
+; SI-NEXT: s_lshl_b32 s5, s30, 8
; SI-NEXT: s_or_b32 s4, s4, s5
-; SI-NEXT: v_readlane_b32 s71, v61, 20
-; SI-NEXT: v_readlane_b32 s70, v61, 19
+; SI-NEXT: v_readlane_b32 s82, v61, 17
+; SI-NEXT: v_readlane_b32 s83, v61, 16
; SI-NEXT: v_cvt_f32_f16_e32 v46, s4
-; SI-NEXT: s_and_b32 s4, s71, 0xff
-; SI-NEXT: s_lshl_b32 s5, s70, 8
+; SI-NEXT: s_and_b32 s4, s82, 0xff
+; SI-NEXT: s_lshl_b32 s5, s83, 8
; SI-NEXT: s_or_b32 s4, s4, s5
-; SI-NEXT: v_readlane_b32 s65, v61, 18
-; SI-NEXT: v_readlane_b32 s54, v61, 17
+; SI-NEXT: v_readlane_b32 s37, v61, 15
+; SI-NEXT: v_readlane_b32 s70, v61, 14
; SI-NEXT: v_cvt_f32_f16_e32 v57, s4
-; SI-NEXT: s_and_b32 s4, s65, 0xff
-; SI-NEXT: s_lshl_b32 s5, s54, 8
-; SI-NEXT: s_mov_b32 s17, s19
-; SI-NEXT: s_mov_b32 s19, s50
+; SI-NEXT: s_and_b32 s4, s37, 0xff
+; SI-NEXT: s_lshl_b32 s5, s70, 8
; SI-NEXT: s_or_b32 s4, s4, s5
-; SI-NEXT: v_readlane_b32 s67, v61, 16
-; SI-NEXT: v_readlane_b32 s50, v61, 15
+; SI-NEXT: v_readlane_b32 s71, v61, 13
+; SI-NEXT: v_readlane_b32 s67, v61, 12
; SI-NEXT: v_cvt_f32_f16_e32 v56, s4
-; SI-NEXT: s_and_b32 s4, s67, 0xff
-; SI-NEXT: s_lshl_b32 s5, s50, 8
+; SI-NEXT: s_and_b32 s4, s71, 0xff
+; SI-NEXT: s_lshl_b32 s5, s67, 8
; SI-NEXT: s_or_b32 s4, s4, s5
-; SI-NEXT: v_readlane_b32 s64, v61, 14
-; SI-NEXT: v_readlane_b32 s52, v61, 13
+; SI-NEXT: v_readlane_b32 s65, v61, 11
+; SI-NEXT: v_readlane_b32 s64, v61, 10
; SI-NEXT: v_cvt_f32_f16_e32 v59, s4
-; SI-NEXT: s_and_b32 s4, s64, 0xff
-; SI-NEXT: s_lshl_b32 s5, s52, 8
-; SI-NEXT: s_mov_b32 s23, s48
+; SI-NEXT: s_and_b32 s4, s65, 0xff
+; SI-NEXT: s_lshl_b32 s5, s64, 8
+; SI-NEXT: s_mov_b32 s88, s50
; SI-NEXT: s_or_b32 s4, s4, s5
-; SI-NEXT: v_readlane_b32 s35, v61, 12
-; SI-NEXT: v_readlane_b32 s48, v61, 11
+; SI-NEXT: v_readlane_b32 s54, v61, 9
+; SI-NEXT: v_readlane_b32 s50, v61, 8
; SI-NEXT: v_cvt_f32_f16_e32 v58, s4
-; SI-NEXT: s_and_b32 s4, s35, 0xff
-; SI-NEXT: s_lshl_b32 s5, s48, 8
+; SI-NEXT: s_and_b32 s4, s54, 0xff
+; SI-NEXT: s_lshl_b32 s5, s50, 8
+; SI-NEXT: s_mov_b32 s24, s39
; SI-NEXT: s_or_b32 s4, s4, s5
-; SI-NEXT: v_readlane_b32 s30, v61, 10
-; SI-NEXT: v_readlane_b32 s39, v61, 9
+; SI-NEXT: v_readlane_b32 s90, v61, 7
+; SI-NEXT: v_readlane_b32 s39, v61, 6
; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:404 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_cvt_f32_f16_e32 v1, s4
-; SI-NEXT: s_and_b32 s4, s30, 0xff
+; SI-NEXT: s_and_b32 s4, s90, 0xff
; SI-NEXT: s_lshl_b32 s5, s39, 8
-; SI-NEXT: s_mov_b32 s26, s37
; SI-NEXT: s_or_b32 s4, s4, s5
-; SI-NEXT: v_readlane_b32 s37, v61, 8
-; SI-NEXT: v_readlane_b32 s75, v61, 7
+; SI-NEXT: v_readlane_b32 s52, v61, 5
+; SI-NEXT: v_readlane_b32 s75, v61, 4
; SI-NEXT: v_cvt_f32_f16_e32 v60, s4
-; SI-NEXT: s_and_b32 s4, s37, 0xff
+; SI-NEXT: s_and_b32 s4, s52, 0xff
; SI-NEXT: s_lshl_b32 s5, s75, 8
; SI-NEXT: s_or_b32 s4, s4, s5
-; SI-NEXT: v_readlane_b32 s92, v61, 6
-; SI-NEXT: v_readlane_b32 s77, v61, 5
+; SI-NEXT: v_readlane_b32 s92, v61, 3
+; SI-NEXT: v_readlane_b32 s77, v61, 2
; SI-NEXT: v_cvt_f32_f16_e32 v2, s4
; SI-NEXT: s_and_b32 s4, s92, 0xff
; SI-NEXT: s_lshl_b32 s5, s77, 8
-; SI-NEXT: s_mov_b32 s28, s29
-; SI-NEXT: s_mov_b32 s29, s76
+; SI-NEXT: s_mov_b32 s8, s27
+; SI-NEXT: s_mov_b32 s27, s78
; SI-NEXT: s_or_b32 s4, s4, s5
-; SI-NEXT: v_readlane_b32 s78, v61, 4
-; SI-NEXT: v_readlane_b32 s76, v61, 3
+; SI-NEXT: v_readlane_b32 s78, v61, 1
+; SI-NEXT: v_readlane_b32 s76, v61, 0
; SI-NEXT: v_cvt_f32_f16_e32 v3, s4
; SI-NEXT: s_and_b32 s4, s78, 0xff
; SI-NEXT: s_lshl_b32 s5, s76, 8
; SI-NEXT: s_or_b32 s4, s4, s5
-; SI-NEXT: s_mov_b32 s99, s55
-; SI-NEXT: s_mov_b32 s20, s88
-; SI-NEXT: s_mov_b32 s24, s98
-; SI-NEXT: s_mov_b32 s59, s58
-; SI-NEXT: s_mov_b32 s56, s47
-; SI-NEXT: s_mov_b32 s46, s41
-; SI-NEXT: s_mov_b32 s12, s11
-; SI-NEXT: s_mov_b32 s11, s7
-; SI-NEXT: s_mov_b32 s7, s97
-; SI-NEXT: s_mov_b32 s97, s81
-; SI-NEXT: s_mov_b32 s81, s85
-; SI-NEXT: s_mov_b32 s6, s40
+; SI-NEXT: s_mov_b32 s31, s95
+; SI-NEXT: s_mov_b32 s57, s56
+; SI-NEXT: s_mov_b32 s25, s59
+; SI-NEXT: s_mov_b32 s14, s41
+; SI-NEXT: s_mov_b32 s46, s47
+; SI-NEXT: s_mov_b32 s12, s7
+; SI-NEXT: s_mov_b32 s7, s11
+; SI-NEXT: s_mov_b32 s11, s85
+; SI-NEXT: s_mov_b32 s81, s45
+; SI-NEXT: s_mov_b32 s85, s97
+; SI-NEXT: s_mov_b32 s45, s40
; SI-NEXT: s_mov_b32 s40, s72
-; SI-NEXT: s_mov_b32 s45, s73
+; SI-NEXT: s_mov_b32 s44, s93
; SI-NEXT: s_mov_b32 s15, s89
; SI-NEXT: v_cvt_f32_f16_e32 v4, s4
-; SI-NEXT: s_mov_b32 s55, s93
+; SI-NEXT: s_mov_b64 s[4:5], 0
+; SI-NEXT: s_mov_b32 s95, s91
+; SI-NEXT: s_mov_b32 s55, s35
+; SI-NEXT: s_mov_b32 s80, s60
+; SI-NEXT: s_mov_b32 s91, s61
+; SI-NEXT: s_mov_b32 s51, s63
+; SI-NEXT: s_mov_b32 s36, s62
+; SI-NEXT: s_branch .LBB93_3
+; SI-NEXT: .LBB93_2:
+; SI-NEXT: ; implicit-def: $vgpr1
+; SI-NEXT: ; kill: killed $vgpr1
+; SI-NEXT: s_mov_b32 s17, s19
+; SI-NEXT: ; implicit-def: $vgpr1
+; SI-NEXT: ; kill: killed $vgpr1
+; SI-NEXT: s_mov_b32 s19, s23
+; SI-NEXT: ; implicit-def: $vgpr1
+; SI-NEXT: s_mov_b32 s23, s26
+; SI-NEXT: s_mov_b32 s26, s29
+; SI-NEXT: s_mov_b32 s29, s53
+; SI-NEXT: s_mov_b32 s88, s50
+; SI-NEXT: s_mov_b32 s24, s39
+; SI-NEXT: s_mov_b32 s8, s27
+; SI-NEXT: s_mov_b32 s27, s78
+; SI-NEXT: s_mov_b32 s25, s59
+; SI-NEXT: s_mov_b32 s57, s56
+; SI-NEXT: s_mov_b32 s46, s47
+; SI-NEXT: s_mov_b32 s14, s41
+; SI-NEXT: s_mov_b32 s12, s7
+; SI-NEXT: s_mov_b32 s7, s11
+; SI-NEXT: s_mov_b32 s11, s85
+; SI-NEXT: s_mov_b32 s85, s97
+; SI-NEXT: s_mov_b32 s81, s45
+; SI-NEXT: s_mov_b32 s45, s40
+; SI-NEXT: s_mov_b32 s40, s72
+; SI-NEXT: s_mov_b32 s44, s93
+; SI-NEXT: s_mov_b32 s15, s89
+; SI-NEXT: s_mov_b32 s79, s9
+; SI-NEXT: s_mov_b32 s31, s95
+; SI-NEXT: s_mov_b64 s[4:5], -1
+; SI-NEXT: ; kill: killed $vgpr1
+; SI-NEXT: ; implicit-def: $vgpr1
+; SI-NEXT: v_readlane_b32 s75, v61, 4
+; SI-NEXT: v_readlane_b32 s76, v61, 0
+; SI-NEXT: v_readlane_b32 s77, v61, 2
+; SI-NEXT: v_readlane_b32 s78, v61, 1
+; SI-NEXT: v_readlane_b32 s92, v61, 3
+; SI-NEXT: v_readlane_b32 s39, v61, 6
+; SI-NEXT: v_readlane_b32 s52, v61, 5
+; SI-NEXT: v_readlane_b32 s90, v61, 7
+; SI-NEXT: v_readlane_b32 s50, v61, 8
+; SI-NEXT: v_readlane_b32 s64, v61, 10
+; SI-NEXT: v_readlane_b32 s54, v61, 9
+; SI-NEXT: v_readlane_b32 s67, v61, 12
+; SI-NEXT: v_readlane_b32 s65, v61, 11
+; SI-NEXT: v_readlane_b32 s70, v61, 14
+; SI-NEXT: v_readlane_b32 s71, v61, 13
+; SI-NEXT: v_readlane_b32 s37, v61, 15
+; SI-NEXT: v_readlane_b32 s83, v61, 16
+; SI-NEXT: v_readlane_b32 s30, v61, 18
+; SI-NEXT: v_readlane_b32 s82, v61, 17
+; SI-NEXT: v_readlane_b32 s48, v61, 20
+; SI-NEXT: v_readlane_b32 s86, v61, 19
+; SI-NEXT: v_readlane_b32 s96, v61, 22
+; SI-NEXT: v_readlane_b32 s38, v61, 21
+; SI-NEXT: v_readlane_b32 s49, v61, 23
+; SI-NEXT: v_readlane_b32 s99, v61, 24
+; SI-NEXT: v_readlane_b32 s6, v61, 26
+; SI-NEXT: v_readlane_b32 s98, v61, 25
; SI-NEXT: s_mov_b32 s95, s91
-; SI-NEXT: s_mov_b32 s31, s90
-; SI-NEXT: s_cbranch_execnz .LBB93_3
-; SI-NEXT: .LBB93_2: ; %cmp.true
+; SI-NEXT: s_mov_b32 s55, s35
+; SI-NEXT: s_mov_b32 s80, s60
+; SI-NEXT: v_readlane_b32 s20, v61, 29
+; SI-NEXT: s_mov_b32 s91, s61
+; SI-NEXT: s_mov_b32 s51, s63
+; SI-NEXT: s_mov_b32 s36, s62
+; SI-NEXT: v_readlane_b32 s34, v61, 27
+; SI-NEXT: v_readlane_b32 s53, v61, 28
+; SI-NEXT: v_readlane_b32 s66, v61, 30
+; SI-NEXT: v_readlane_b32 s68, v61, 31
+; SI-NEXT: v_readlane_b32 s69, v61, 32
+; SI-NEXT: v_readlane_b32 s21, v61, 33
+; SI-NEXT: v_readlane_b32 s9, v61, 34
+; SI-NEXT: v_readlane_b32 s13, v61, 35
+; SI-NEXT: ; implicit-def: $vgpr6
+; SI-NEXT: ; implicit-def: $vgpr5
+; SI-NEXT: ; implicit-def: $vgpr7
+; SI-NEXT: ; kill: killed $vgpr1
+; SI-NEXT: ; implicit-def: $vgpr10
+; SI-NEXT: ; implicit-def: $vgpr8
+; SI-NEXT: ; implicit-def: $vgpr12
+; SI-NEXT: ; implicit-def: $vgpr9
+; SI-NEXT: ; implicit-def: $vgpr13
+; SI-NEXT: ; implicit-def: $vgpr11
+; SI-NEXT: ; implicit-def: $vgpr15
+; SI-NEXT: ; implicit-def: $vgpr14
+; SI-NEXT: ; implicit-def: $vgpr17
+; SI-NEXT: ; implicit-def: $vgpr16
+; SI-NEXT: ; implicit-def: $vgpr19
+; SI-NEXT: ; implicit-def: $vgpr18
+; SI-NEXT: ; implicit-def: $vgpr21
+; SI-NEXT: ; implicit-def: $vgpr20
+; SI-NEXT: ; implicit-def: $vgpr23
+; SI-NEXT: ; implicit-def: $vgpr22
+; SI-NEXT: ; implicit-def: $vgpr25
+; SI-NEXT: ; implicit-def: $vgpr24
+; SI-NEXT: ; implicit-def: $vgpr27
+; SI-NEXT: ; implicit-def: $vgpr26
+; SI-NEXT: ; implicit-def: $vgpr29
+; SI-NEXT: ; implicit-def: $vgpr28
+; SI-NEXT: ; implicit-def: $vgpr31
+; SI-NEXT: ; implicit-def: $vgpr30
+; SI-NEXT: ; implicit-def: $vgpr33
+; SI-NEXT: ; implicit-def: $vgpr32
+; SI-NEXT: ; implicit-def: $vgpr35
+; SI-NEXT: ; implicit-def: $vgpr34
+; SI-NEXT: ; implicit-def: $vgpr37
+; SI-NEXT: ; implicit-def: $vgpr36
+; SI-NEXT: ; implicit-def: $vgpr39
+; SI-NEXT: ; implicit-def: $vgpr38
+; SI-NEXT: ; implicit-def: $vgpr49
+; SI-NEXT: ; implicit-def: $vgpr48
+; SI-NEXT: ; implicit-def: $vgpr51
+; SI-NEXT: ; implicit-def: $vgpr50
+; SI-NEXT: ; implicit-def: $vgpr53
+; SI-NEXT: ; implicit-def: $vgpr52
+; SI-NEXT: ; implicit-def: $vgpr55
+; SI-NEXT: ; implicit-def: $vgpr54
+; SI-NEXT: ; implicit-def: $vgpr41
+; SI-NEXT: ; implicit-def: $vgpr40
+; SI-NEXT: ; implicit-def: $vgpr43
+; SI-NEXT: ; implicit-def: $vgpr42
+; SI-NEXT: ; implicit-def: $vgpr45
+; SI-NEXT: ; implicit-def: $vgpr44
+; SI-NEXT: ; implicit-def: $vgpr47
+; SI-NEXT: ; implicit-def: $vgpr46
+; SI-NEXT: ; implicit-def: $vgpr57
+; SI-NEXT: ; implicit-def: $vgpr56
+; SI-NEXT: ; implicit-def: $vgpr59
+; SI-NEXT: ; implicit-def: $vgpr58
+; SI-NEXT: ; implicit-def: $vgpr1
+; SI-NEXT: ; implicit-def: $vgpr60
+; SI-NEXT: ; implicit-def: $vgpr2
+; SI-NEXT: ; implicit-def: $vgpr3
+; SI-NEXT: ; implicit-def: $vgpr4
+; SI-NEXT: .LBB93_3: ; %Flow
+; SI-NEXT: s_andn2_b64 vcc, exec, s[4:5]
+; SI-NEXT: s_mov_b32 s35, s79
+; SI-NEXT: s_cbranch_vccnz .LBB93_5
+; SI-NEXT: ; %bb.4: ; %cmp.true
; SI-NEXT: s_add_i32 s4, s78, 3
; SI-NEXT: s_and_b32 s4, s4, 0xff
; SI-NEXT: s_lshl_b32 s5, s76, 8
@@ -185958,60 +186138,60 @@ define inreg <64 x half> @bitcast_v128i8_to_v64f16_scalar(<128 x i8> inreg %a, i
; SI-NEXT: s_and_b32 s5, s5, 0xff
; SI-NEXT: s_lshl_b32 vcc_lo, s77, 8
; SI-NEXT: s_or_b32 s5, vcc_lo, s5
-; SI-NEXT: s_add_i32 vcc_lo, s37, 3
+; SI-NEXT: s_add_i32 vcc_lo, s52, 3
; SI-NEXT: s_and_b32 vcc_lo, vcc_lo, 0xff
; SI-NEXT: s_lshl_b32 vcc_hi, s75, 8
; SI-NEXT: s_or_b32 vcc_lo, vcc_hi, vcc_lo
-; SI-NEXT: s_add_i32 vcc_hi, s30, 3
+; SI-NEXT: s_add_i32 vcc_hi, s90, 3
; SI-NEXT: s_and_b32 vcc_hi, vcc_hi, 0xff
; SI-NEXT: s_lshl_b32 s60, s39, 8
; SI-NEXT: s_or_b32 s60, s60, vcc_hi
-; SI-NEXT: s_add_i32 vcc_hi, s35, 3
+; SI-NEXT: s_add_i32 vcc_hi, s54, 3
; SI-NEXT: s_and_b32 vcc_hi, vcc_hi, 0xff
-; SI-NEXT: s_lshl_b32 s61, s48, 8
+; SI-NEXT: s_lshl_b32 s61, s50, 8
; SI-NEXT: s_or_b32 s61, s61, vcc_hi
-; SI-NEXT: s_add_i32 vcc_hi, s64, 3
+; SI-NEXT: s_add_i32 vcc_hi, s65, 3
; SI-NEXT: s_and_b32 vcc_hi, vcc_hi, 0xff
-; SI-NEXT: s_lshl_b32 s62, s52, 8
+; SI-NEXT: s_lshl_b32 s62, s64, 8
; SI-NEXT: s_or_b32 s62, s62, vcc_hi
-; SI-NEXT: s_add_i32 vcc_hi, s67, 3
+; SI-NEXT: s_add_i32 vcc_hi, s71, 3
; SI-NEXT: s_and_b32 vcc_hi, vcc_hi, 0xff
-; SI-NEXT: s_lshl_b32 s63, s50, 8
+; SI-NEXT: s_lshl_b32 s63, s67, 8
; SI-NEXT: s_or_b32 s10, s63, vcc_hi
-; SI-NEXT: s_add_i32 vcc_hi, s65, 3
+; SI-NEXT: s_add_i32 vcc_hi, s37, 3
; SI-NEXT: s_and_b32 vcc_hi, vcc_hi, 0xff
-; SI-NEXT: s_lshl_b32 s72, s54, 8
+; SI-NEXT: s_lshl_b32 s72, s70, 8
; SI-NEXT: s_or_b32 s72, s72, vcc_hi
-; SI-NEXT: s_add_i32 vcc_hi, s71, 3
+; SI-NEXT: s_add_i32 vcc_hi, s82, 3
; SI-NEXT: s_and_b32 vcc_hi, vcc_hi, 0xff
-; SI-NEXT: s_lshl_b32 s73, s70, 8
+; SI-NEXT: s_lshl_b32 s73, s83, 8
; SI-NEXT: s_or_b32 s73, s73, vcc_hi
-; SI-NEXT: s_add_i32 vcc_hi, s83, 3
+; SI-NEXT: s_add_i32 vcc_hi, s86, 3
; SI-NEXT: s_and_b32 vcc_hi, vcc_hi, 0xff
-; SI-NEXT: s_lshl_b32 s74, s49, 8
+; SI-NEXT: s_lshl_b32 s74, s30, 8
; SI-NEXT: s_or_b32 s74, s74, vcc_hi
-; SI-NEXT: s_add_i32 vcc_hi, s82, 3
+; SI-NEXT: s_add_i32 vcc_hi, s38, 3
; SI-NEXT: s_and_b32 vcc_hi, vcc_hi, 0xff
-; SI-NEXT: s_lshl_b32 s75, s80, 8
+; SI-NEXT: s_lshl_b32 s75, s48, 8
; SI-NEXT: s_or_b32 s75, s75, vcc_hi
-; SI-NEXT: s_add_i32 vcc_hi, s87, 3
+; SI-NEXT: s_add_i32 vcc_hi, s49, 3
; SI-NEXT: s_and_b32 vcc_hi, vcc_hi, 0xff
-; SI-NEXT: s_lshl_b32 s76, s84, 8
+; SI-NEXT: s_lshl_b32 s76, s96, 8
; SI-NEXT: s_or_b32 s76, s76, vcc_hi
-; SI-NEXT: s_add_i32 vcc_hi, s51, 3
+; SI-NEXT: s_add_i32 vcc_hi, s98, 3
; SI-NEXT: s_add_i32 s93, s53, 3
; SI-NEXT: s_and_b32 vcc_hi, vcc_hi, 0xff
-; SI-NEXT: s_lshl_b32 s77, s86, 8
+; SI-NEXT: s_lshl_b32 s77, s99, 8
; SI-NEXT: s_add_i32 s89, s34, 3
; SI-NEXT: s_and_b32 s93, s93, 0xff
-; SI-NEXT: s_lshl_b32 s78, s94, 8
+; SI-NEXT: s_lshl_b32 s78, s16, 8
; SI-NEXT: s_add_i32 s34, s66, 3
; SI-NEXT: s_or_b32 s77, s77, vcc_hi
; SI-NEXT: s_and_b32 s89, s89, 0xff
-; SI-NEXT: s_lshl_b32 vcc_hi, s96, 8
+; SI-NEXT: s_lshl_b32 vcc_hi, s6, 8
; SI-NEXT: s_or_b32 s22, s78, s93
; SI-NEXT: s_and_b32 s93, s34, 0xff
-; SI-NEXT: s_lshl_b32 s92, s16, 8
+; SI-NEXT: s_lshl_b32 s92, s20, 8
; SI-NEXT: s_add_i32 s53, s68, 3
; SI-NEXT: s_or_b32 s89, vcc_hi, s89
; SI-NEXT: s_or_b32 s92, s92, s93
@@ -186020,261 +186200,251 @@ define inreg <64 x half> @bitcast_v128i8_to_v64f16_scalar(<128 x i8> inreg %a, i
; SI-NEXT: s_add_i32 s66, s69, 3
; SI-NEXT: s_or_b32 s93, vcc_hi, s93
; SI-NEXT: s_and_b32 vcc_hi, s66, 0xff
-; SI-NEXT: s_lshl_b32 s34, s45, 8
-; SI-NEXT: s_add_i32 s68, s6, 3
+; SI-NEXT: s_lshl_b32 s34, s44, 8
+; SI-NEXT: s_add_i32 s68, s45, 3
; SI-NEXT: s_or_b32 vcc_hi, s34, vcc_hi
; SI-NEXT: s_and_b32 s34, s68, 0xff
; SI-NEXT: s_lshl_b32 s39, s40, 8
; SI-NEXT: s_add_i32 s69, s81, 3
; SI-NEXT: s_or_b32 s34, s39, s34
; SI-NEXT: s_and_b32 s39, s69, 0xff
-; SI-NEXT: s_lshl_b32 s52, s21, 8
+; SI-NEXT: s_lshl_b32 s52, s85, 8
; SI-NEXT: s_add_i32 s81, s7, 3
; SI-NEXT: s_or_b32 s39, s52, s39
; SI-NEXT: s_and_b32 s52, s81, 0xff
-; SI-NEXT: s_lshl_b32 s53, s97, 8
+; SI-NEXT: s_lshl_b32 s53, s11, 8
; SI-NEXT: s_add_i32 s85, s12, 3
; SI-NEXT: s_or_b32 s52, s53, s52
; SI-NEXT: s_and_b32 s53, s85, 0xff
-; SI-NEXT: s_lshl_b32 s64, s11, 8
-; SI-NEXT: s_add_i32 s97, s56, 3
+; SI-NEXT: s_lshl_b32 s64, s21, 8
+; SI-NEXT: s_add_i32 s97, s14, 3
; SI-NEXT: s_or_b32 s53, s64, s53
; SI-NEXT: s_and_b32 s64, s97, 0xff
; SI-NEXT: s_lshl_b32 s66, s46, 8
-; SI-NEXT: s_add_i32 s21, s29, 3
+; SI-NEXT: s_add_i32 s21, s57, 3
; SI-NEXT: s_or_b32 s64, s66, s64
; SI-NEXT: s_and_b32 s21, s21, 0xff
-; SI-NEXT: s_lshl_b32 s66, s59, 8
+; SI-NEXT: s_lshl_b32 s66, s25, 8
; SI-NEXT: s_add_i32 s25, s8, 3
; SI-NEXT: s_or_b32 s66, s66, s21
; SI-NEXT: s_and_b32 s21, s25, 0xff
-; SI-NEXT: s_lshl_b32 s6, s28, 8
-; SI-NEXT: s_add_i32 s29, s19, 3
+; SI-NEXT: s_lshl_b32 s6, s27, 8
+; SI-NEXT: s_add_i32 s29, s29, 3
+; SI-NEXT: v_readlane_b32 s16, v62, 14
; SI-NEXT: s_or_b32 s67, s6, s21
; SI-NEXT: s_and_b32 s6, s29, 0xff
-; SI-NEXT: s_lshl_b32 s18, s26, 8
-; SI-NEXT: s_add_i32 s28, s17, 3
+; SI-NEXT: s_lshl_b32 s18, s24, 8
+; SI-NEXT: s_add_i32 s28, s9, 3
+; SI-NEXT: s_add_i32 s27, s16, 3
+; SI-NEXT: v_readlane_b32 s16, v62, 12
; SI-NEXT: s_or_b32 s68, s18, s6
; SI-NEXT: s_and_b32 s6, s28, 0xff
-; SI-NEXT: s_lshl_b32 s18, s23, 8
+; SI-NEXT: s_lshl_b32 s18, s88, 8
+; SI-NEXT: s_add_i32 s7, s23, 3
+; SI-NEXT: s_lshl_b32 s23, s16, 8
+; SI-NEXT: v_readlane_b32 s16, v62, 13
; SI-NEXT: s_or_b32 s69, s18, s6
-; SI-NEXT: v_readlane_b32 s6, v62, 17
-; SI-NEXT: s_add_i32 s7, s6, 3
-; SI-NEXT: v_readlane_b32 s16, v62, 15
; SI-NEXT: s_and_b32 s6, s7, 0xff
-; SI-NEXT: v_readlane_b32 s7, v62, 16
-; SI-NEXT: s_add_i32 s27, s16, 3
-; SI-NEXT: v_readlane_b32 s16, v62, 13
-; SI-NEXT: s_lshl_b32 s7, s7, 8
-; SI-NEXT: s_lshl_b32 s23, s16, 8
-; SI-NEXT: v_readlane_b32 s16, v62, 14
-; SI-NEXT: s_mov_b32 s91, s24
-; SI-NEXT: s_or_b32 s70, s7, s6
-; SI-NEXT: v_readlane_b32 s6, v62, 20
+; SI-NEXT: s_lshl_b32 s7, s26, 8
+; SI-NEXT: s_add_i32 s11, s13, 3
; SI-NEXT: s_add_i32 s24, s16, 3
-; SI-NEXT: v_readlane_b32 s16, v62, 11
-; SI-NEXT: s_add_i32 s11, s6, 3
-; SI-NEXT: v_readlane_b32 s7, v62, 18
-; SI-NEXT: s_lshl_b32 s19, s16, 8
-; SI-NEXT: v_readlane_b32 s16, v62, 12
-; SI-NEXT: s_mov_b32 s90, s20
+; SI-NEXT: v_readlane_b32 s16, v62, 10
+; SI-NEXT: s_or_b32 s70, s7, s6
; SI-NEXT: s_and_b32 s6, s11, 0xff
-; SI-NEXT: s_lshl_b32 s7, s7, 8
-; SI-NEXT: s_add_i32 s20, s16, 3
-; SI-NEXT: v_readlane_b32 s16, v62, 9
+; SI-NEXT: s_lshl_b32 s7, s19, 8
+; SI-NEXT: s_lshl_b32 s19, s16, 8
+; SI-NEXT: v_readlane_b32 s16, v62, 11
; SI-NEXT: s_or_b32 s71, s7, s6
-; SI-NEXT: v_readlane_b32 s6, v62, 22
+; SI-NEXT: v_readlane_b32 s6, v62, 16
+; SI-NEXT: s_add_i32 s20, s16, 3
+; SI-NEXT: v_readlane_b32 s16, v62, 8
+; SI-NEXT: s_add_i32 s12, s6, 3
+; SI-NEXT: s_lshl_b32 s7, s17, 8
; SI-NEXT: s_and_b32 s20, s20, 0xff
; SI-NEXT: s_lshl_b32 s17, s16, 8
-; SI-NEXT: v_readlane_b32 s16, v62, 10
-; SI-NEXT: s_add_i32 s12, s6, 3
-; SI-NEXT: v_readlane_b32 s7, v62, 19
+; SI-NEXT: v_readlane_b32 s16, v62, 9
+; SI-NEXT: s_and_b32 s6, s12, 0xff
; SI-NEXT: s_or_b32 s17, s17, s20
; SI-NEXT: s_add_i32 s16, s16, 3
-; SI-NEXT: v_readlane_b32 s20, v62, 8
-; SI-NEXT: s_and_b32 s6, s12, 0xff
-; SI-NEXT: s_lshl_b32 s7, s7, 8
+; SI-NEXT: v_readlane_b32 s20, v62, 7
+; SI-NEXT: s_or_b32 s81, s7, s6
+; SI-NEXT: v_readlane_b32 s6, v62, 17
; SI-NEXT: s_and_b32 s16, s16, 0xff
; SI-NEXT: s_lshl_b32 s20, s20, 8
-; SI-NEXT: s_or_b32 s81, s7, s6
-; SI-NEXT: v_readlane_b32 s6, v62, 23
+; SI-NEXT: s_add_i32 s14, s6, 3
+; SI-NEXT: v_readlane_b32 s7, v62, 15
; SI-NEXT: s_and_b32 s24, s24, 0xff
; SI-NEXT: s_or_b32 s16, s20, s16
-; SI-NEXT: v_readlane_b32 s20, v62, 7
-; SI-NEXT: s_add_i32 s14, s6, 3
-; SI-NEXT: v_readlane_b32 s7, v62, 21
-; SI-NEXT: s_or_b32 s19, s19, s24
-; SI-NEXT: s_add_i32 s98, s20, 3
-; SI-NEXT: v_readlane_b32 s24, v62, 6
+; SI-NEXT: v_readlane_b32 s20, v62, 6
; SI-NEXT: s_and_b32 s6, s14, 0xff
; SI-NEXT: s_lshl_b32 s7, s7, 8
+; SI-NEXT: s_or_b32 s19, s19, s24
+; SI-NEXT: s_add_i32 s98, s20, 3
+; SI-NEXT: v_readlane_b32 s24, v62, 5
+; SI-NEXT: s_or_b32 s83, s7, s6
+; SI-NEXT: v_readlane_b32 s6, v62, 20
; SI-NEXT: s_and_b32 s20, s98, 0xff
; SI-NEXT: s_lshl_b32 s24, s24, 8
-; SI-NEXT: s_or_b32 s83, s7, s6
-; SI-NEXT: v_readlane_b32 s6, v62, 26
+; SI-NEXT: s_add_i32 s41, s6, 3
+; SI-NEXT: v_readlane_b32 s7, v62, 18
; SI-NEXT: s_and_b32 s27, s27, 0xff
; SI-NEXT: s_or_b32 s20, s24, s20
-; SI-NEXT: v_readlane_b32 s24, v62, 5
-; SI-NEXT: s_add_i32 s41, s6, 3
-; SI-NEXT: v_readlane_b32 s7, v62, 24
-; SI-NEXT: s_or_b32 s23, s23, s27
-; SI-NEXT: s_add_i32 s86, s24, 3
-; SI-NEXT: v_readlane_b32 s27, v62, 4
+; SI-NEXT: v_readlane_b32 s24, v62, 4
; SI-NEXT: s_and_b32 s6, s41, 0xff
; SI-NEXT: s_lshl_b32 s7, s7, 8
+; SI-NEXT: s_or_b32 s23, s23, s27
+; SI-NEXT: s_add_i32 s86, s24, 3
+; SI-NEXT: v_readlane_b32 s27, v62, 3
+; SI-NEXT: s_or_b32 s85, s7, s6
+; SI-NEXT: v_readlane_b32 s6, v62, 22
; SI-NEXT: s_and_b32 s24, s86, 0xff
; SI-NEXT: s_lshl_b32 s27, s27, 8
-; SI-NEXT: s_or_b32 s85, s7, s6
-; SI-NEXT: v_readlane_b32 s6, v62, 28
-; SI-NEXT: s_or_b32 s24, s27, s24
-; SI-NEXT: v_readlane_b32 s27, v62, 3
; SI-NEXT: s_add_i32 s46, s6, 3
-; SI-NEXT: v_readlane_b32 s7, v62, 25
-; SI-NEXT: s_add_i32 s12, s73, 0x300
-; SI-NEXT: s_add_i32 s82, s27, 3
-; SI-NEXT: v_readlane_b32 s73, v62, 2
+; SI-NEXT: v_readlane_b32 s7, v62, 19
+; SI-NEXT: s_or_b32 s24, s27, s24
+; SI-NEXT: v_readlane_b32 s27, v62, 2
; SI-NEXT: s_and_b32 s6, s46, 0xff
; SI-NEXT: s_lshl_b32 s7, s7, 8
+; SI-NEXT: s_add_i32 s12, s73, 0x300
+; SI-NEXT: s_add_i32 s82, s27, 3
+; SI-NEXT: v_readlane_b32 s73, v62, 1
+; SI-NEXT: s_or_b32 s96, s7, s6
+; SI-NEXT: v_readlane_b32 s6, v62, 25
; SI-NEXT: s_and_b32 s27, s82, 0xff
; SI-NEXT: s_lshl_b32 s73, s73, 8
-; SI-NEXT: s_or_b32 s96, s7, s6
-; SI-NEXT: v_readlane_b32 s6, v62, 31
-; SI-NEXT: s_or_b32 s27, s73, s27
-; SI-NEXT: v_readlane_b32 s73, v62, 1
; SI-NEXT: s_add_i32 s47, s6, 3
-; SI-NEXT: v_readlane_b32 s7, v62, 27
-; SI-NEXT: s_add_i32 s13, s74, 0x300
-; SI-NEXT: s_add_i32 s65, s73, 3
-; SI-NEXT: v_readlane_b32 s74, v62, 0
+; SI-NEXT: v_readlane_b32 s7, v62, 21
+; SI-NEXT: s_or_b32 s27, s73, s27
+; SI-NEXT: v_readlane_b32 s73, v62, 0
; SI-NEXT: s_and_b32 s6, s47, 0xff
; SI-NEXT: s_lshl_b32 s7, s7, 8
+; SI-NEXT: s_add_i32 s13, s74, 0x300
+; SI-NEXT: s_add_i32 s65, s73, 3
+; SI-NEXT: v_readlane_b32 s74, v61, 63
+; SI-NEXT: s_or_b32 s97, s7, s6
+; SI-NEXT: v_readlane_b32 s6, v62, 28
; SI-NEXT: s_and_b32 s73, s65, 0xff
; SI-NEXT: s_lshl_b32 s74, s74, 8
-; SI-NEXT: s_or_b32 s97, s7, s6
-; SI-NEXT: v_readlane_b32 s6, v62, 34
-; SI-NEXT: s_or_b32 s73, s74, s73
-; SI-NEXT: v_readlane_b32 s74, v61, 63
; SI-NEXT: s_add_i32 s56, s6, 3
-; SI-NEXT: v_readlane_b32 s7, v62, 29
-; SI-NEXT: s_add_i32 s14, s75, 0x300
-; SI-NEXT: s_add_i32 s54, s74, 3
-; SI-NEXT: v_readlane_b32 s75, v61, 62
+; SI-NEXT: v_readlane_b32 s7, v62, 23
+; SI-NEXT: s_or_b32 s73, s74, s73
+; SI-NEXT: v_readlane_b32 s74, v61, 62
; SI-NEXT: s_and_b32 s6, s56, 0xff
; SI-NEXT: s_lshl_b32 s7, s7, 8
+; SI-NEXT: s_add_i32 s14, s75, 0x300
+; SI-NEXT: s_add_i32 s54, s74, 3
+; SI-NEXT: v_readlane_b32 s75, v61, 61
+; SI-NEXT: s_or_b32 s63, s7, s6
+; SI-NEXT: v_readlane_b32 s6, v62, 30
; SI-NEXT: s_and_b32 s74, s54, 0xff
; SI-NEXT: s_lshl_b32 s75, s75, 8
-; SI-NEXT: s_or_b32 s63, s7, s6
-; SI-NEXT: v_readlane_b32 s6, v62, 36
-; SI-NEXT: s_or_b32 s74, s75, s74
-; SI-NEXT: v_readlane_b32 s75, v61, 61
; SI-NEXT: s_add_i32 s58, s6, 3
-; SI-NEXT: v_readlane_b32 s7, v62, 30
-; SI-NEXT: s_add_i32 s15, s76, 0x300
-; SI-NEXT: s_add_i32 s50, s75, 3
-; SI-NEXT: v_readlane_b32 s76, v61, 60
+; SI-NEXT: v_readlane_b32 s7, v62, 24
+; SI-NEXT: s_or_b32 s74, s75, s74
+; SI-NEXT: v_readlane_b32 s75, v61, 60
; SI-NEXT: s_and_b32 s6, s58, 0xff
; SI-NEXT: s_lshl_b32 s7, s7, 8
+; SI-NEXT: s_add_i32 s15, s76, 0x300
+; SI-NEXT: s_add_i32 s50, s75, 3
+; SI-NEXT: v_readlane_b32 s76, v61, 59
+; SI-NEXT: s_or_b32 s79, s7, s6
+; SI-NEXT: v_readlane_b32 s6, v62, 31
; SI-NEXT: s_and_b32 s75, s50, 0xff
; SI-NEXT: s_lshl_b32 s76, s76, 8
-; SI-NEXT: s_or_b32 s79, s7, s6
-; SI-NEXT: v_readlane_b32 s6, v62, 37
-; SI-NEXT: s_or_b32 s75, s76, s75
-; SI-NEXT: v_readlane_b32 s76, v61, 59
; SI-NEXT: s_add_i32 s59, s6, 3
-; SI-NEXT: v_readlane_b32 s7, v62, 33
-; SI-NEXT: s_add_i32 s18, s77, 0x300
-; SI-NEXT: s_add_i32 s48, s76, 3
-; SI-NEXT: v_readlane_b32 s77, v61, 58
+; SI-NEXT: v_readlane_b32 s7, v62, 27
+; SI-NEXT: s_or_b32 s75, s76, s75
+; SI-NEXT: v_readlane_b32 s76, v61, 58
; SI-NEXT: s_and_b32 s6, s59, 0xff
; SI-NEXT: s_lshl_b32 s7, s7, 8
+; SI-NEXT: s_add_i32 s18, s77, 0x300
+; SI-NEXT: s_add_i32 s48, s76, 3
+; SI-NEXT: v_readlane_b32 s77, v61, 57
+; SI-NEXT: s_or_b32 s78, s7, s6
+; SI-NEXT: v_readlane_b32 s6, v62, 29
; SI-NEXT: s_and_b32 s76, s48, 0xff
; SI-NEXT: s_lshl_b32 s77, s77, 8
-; SI-NEXT: s_or_b32 s78, s7, s6
-; SI-NEXT: v_readlane_b32 s6, v62, 35
-; SI-NEXT: s_or_b32 s76, s77, s76
-; SI-NEXT: v_readlane_b32 s77, v61, 57
; SI-NEXT: s_add_i32 s57, s6, 3
-; SI-NEXT: v_readlane_b32 s7, v62, 32
+; SI-NEXT: v_readlane_b32 s7, v62, 26
+; SI-NEXT: s_or_b32 s76, s77, s76
+; SI-NEXT: v_readlane_b32 s77, v61, 56
+; SI-NEXT: s_and_b32 s6, s57, 0xff
+; SI-NEXT: s_lshl_b32 s7, s7, 8
; SI-NEXT: s_add_i32 s11, s72, 0x300
; SI-NEXT: s_add_i32 s72, s79, 0x300
; SI-NEXT: s_add_i32 s37, s77, 3
-; SI-NEXT: v_readlane_b32 s79, v61, 56
-; SI-NEXT: s_and_b32 s6, s57, 0xff
-; SI-NEXT: s_lshl_b32 s7, s7, 8
+; SI-NEXT: v_readlane_b32 s79, v61, 55
+; SI-NEXT: s_or_b32 s88, s7, s6
; SI-NEXT: s_and_b32 s77, s37, 0xff
; SI-NEXT: s_lshl_b32 s79, s79, 8
-; SI-NEXT: s_or_b32 s88, s7, s6
-; SI-NEXT: s_or_b32 s77, s79, s77
-; SI-NEXT: v_readlane_b32 s79, v61, 55
+; SI-NEXT: s_add_i32 s35, s35, 3
; SI-NEXT: s_add_i32 s21, s89, 0x300
; SI-NEXT: s_add_i32 s89, s88, 0x300
-; SI-NEXT: s_add_i32 s35, s79, 3
-; SI-NEXT: v_readlane_b32 s88, v61, 54
+; SI-NEXT: s_or_b32 s77, s79, s77
; SI-NEXT: s_and_b32 s79, s35, 0xff
-; SI-NEXT: s_lshl_b32 s88, s88, 8
+; SI-NEXT: s_lshl_b32 s88, s36, 8
; SI-NEXT: s_or_b32 s79, s88, s79
-; SI-NEXT: v_readlane_b32 s88, v61, 53
-; SI-NEXT: s_add_i32 s25, s92, 0x300
+; SI-NEXT: v_readlane_b32 s88, v61, 54
; SI-NEXT: s_add_i32 s30, s88, 3
-; SI-NEXT: v_readlane_b32 s92, v61, 52
+; SI-NEXT: s_add_i32 s25, s92, 0x300
; SI-NEXT: s_and_b32 s88, s30, 0xff
-; SI-NEXT: s_lshl_b32 s92, s92, 8
+; SI-NEXT: s_lshl_b32 s92, s51, 8
+; SI-NEXT: s_add_i32 s94, s94, 3
+; SI-NEXT: v_readlane_b32 s90, v61, 53
; SI-NEXT: s_or_b32 s88, s92, s88
-; SI-NEXT: v_readlane_b32 s92, v61, 51
-; SI-NEXT: s_add_i32 s94, s92, 3
; SI-NEXT: s_and_b32 s92, s94, 0xff
; SI-NEXT: s_lshl_b32 s91, s91, 8
; SI-NEXT: s_add_i32 s90, s90, 3
; SI-NEXT: s_or_b32 s91, s91, s92
; SI-NEXT: s_and_b32 s90, s90, 0xff
-; SI-NEXT: s_lshl_b32 s92, s31, 8
+; SI-NEXT: s_lshl_b32 s92, s80, 8
; SI-NEXT: s_or_b32 s90, s92, s90
-; SI-NEXT: v_readlane_b32 s92, v61, 50
+; SI-NEXT: v_readlane_b32 s92, v61, 52
; SI-NEXT: s_add_i32 s92, s92, 3
; SI-NEXT: s_add_i32 s26, s93, 0x300
; SI-NEXT: s_and_b32 s92, s92, 0xff
-; SI-NEXT: s_lshl_b32 s93, s95, 8
+; SI-NEXT: s_lshl_b32 s93, s55, 8
; SI-NEXT: s_or_b32 s92, s93, s92
-; SI-NEXT: v_readlane_b32 s93, v61, 49
+; SI-NEXT: v_readlane_b32 s93, v61, 51
; SI-NEXT: s_add_i32 s93, s93, 3
; SI-NEXT: s_and_b32 s93, s93, 0xff
-; SI-NEXT: s_lshl_b32 s94, s55, 8
+; SI-NEXT: s_lshl_b32 s94, s95, 8
; SI-NEXT: s_or_b32 s93, s94, s93
-; SI-NEXT: v_readlane_b32 s94, v61, 48
+; SI-NEXT: v_readlane_b32 s94, v61, 49
; SI-NEXT: s_add_i32 s94, s94, 3
; SI-NEXT: s_and_b32 s94, s94, 0xff
-; SI-NEXT: s_lshl_b32 s95, s99, 8
+; SI-NEXT: s_lshl_b32 s95, s31, 8
; SI-NEXT: s_or_b32 s94, s95, s94
-; SI-NEXT: v_readlane_b32 s95, v61, 1
+; SI-NEXT: v_readlane_b32 s95, v61, 50
; SI-NEXT: s_add_i32 s95, s95, 3
-; SI-NEXT: v_readlane_b32 s30, v61, 0
+; SI-NEXT: v_readlane_b32 s30, v61, 48
; SI-NEXT: s_add_i32 s6, vcc_lo, 0x300
; SI-NEXT: s_and_b32 s95, s95, 0xff
; SI-NEXT: s_lshl_b32 vcc_lo, s30, 8
; SI-NEXT: v_readlane_b32 s30, v61, 47
; SI-NEXT: s_or_b32 s95, vcc_lo, s95
; SI-NEXT: s_add_i32 vcc_lo, s30, 3
-; SI-NEXT: v_readlane_b32 s30, v61, 2
+; SI-NEXT: v_readlane_b32 s30, v61, 46
; SI-NEXT: s_add_i32 s28, vcc_hi, 0x300
; SI-NEXT: s_and_b32 vcc_lo, vcc_lo, 0xff
; SI-NEXT: s_lshl_b32 vcc_hi, s30, 8
-; SI-NEXT: v_readlane_b32 s30, v61, 46
+; SI-NEXT: v_readlane_b32 s30, v61, 45
; SI-NEXT: s_or_b32 vcc_lo, vcc_hi, vcc_lo
; SI-NEXT: s_add_i32 vcc_hi, s30, 3
-; SI-NEXT: v_readlane_b32 s30, v61, 45
+; SI-NEXT: v_readlane_b32 s30, v61, 44
; SI-NEXT: s_and_b32 vcc_hi, vcc_hi, 0xff
; SI-NEXT: s_lshl_b32 s30, s30, 8
; SI-NEXT: s_or_b32 vcc_hi, s30, vcc_hi
-; SI-NEXT: v_readlane_b32 s30, v61, 44
+; SI-NEXT: v_readlane_b32 s30, v61, 43
; SI-NEXT: s_add_i32 s30, s30, 3
-; SI-NEXT: v_readlane_b32 s31, v61, 43
+; SI-NEXT: v_readlane_b32 s31, v61, 42
; SI-NEXT: s_and_b32 s30, s30, 0xff
; SI-NEXT: s_lshl_b32 s31, s31, 8
; SI-NEXT: s_or_b32 s30, s31, s30
-; SI-NEXT: v_readlane_b32 s31, v61, 42
+; SI-NEXT: v_readlane_b32 s31, v61, 41
; SI-NEXT: s_add_i32 s29, s34, 0x300
; SI-NEXT: s_add_i32 s31, s31, 3
-; SI-NEXT: v_readlane_b32 s34, v61, 41
+; SI-NEXT: v_readlane_b32 s34, v61, 40
; SI-NEXT: s_and_b32 s31, s31, 0xff
; SI-NEXT: s_lshl_b32 s34, s34, 8
; SI-NEXT: s_or_b32 s31, s34, s31
@@ -186282,25 +186452,25 @@ define inreg <64 x half> @bitcast_v128i8_to_v64f16_scalar(<128 x i8> inreg %a, i
; SI-NEXT: v_cvt_f32_f16_e32 v1, s31
; SI-NEXT: s_addk_i32 s30, 0x300
; SI-NEXT: s_addk_i32 vcc_hi, 0x300
-; SI-NEXT: v_readlane_b32 s34, v61, 40
+; SI-NEXT: v_readlane_b32 s34, v61, 39
; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:392 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_cvt_f32_f16_e32 v1, s30
; SI-NEXT: s_add_i32 s34, s34, 3
-; SI-NEXT: v_readlane_b32 s35, v61, 39
+; SI-NEXT: v_readlane_b32 s35, v61, 38
; SI-NEXT: s_and_b32 s34, s34, 0xff
; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:400 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_cvt_f32_f16_e32 v1, vcc_hi
; SI-NEXT: s_lshl_b32 s35, s35, 8
-; SI-NEXT: s_addk_i32 vcc_lo, 0x300
+; SI-NEXT: s_addk_i32 s95, 0x300
; SI-NEXT: s_or_b32 s34, s35, s34
-; SI-NEXT: v_readlane_b32 s35, v61, 38
+; SI-NEXT: v_readlane_b32 s35, v61, 37
; SI-NEXT: s_add_i32 s35, s35, 3
-; SI-NEXT: v_readlane_b32 s36, v61, 37
+; SI-NEXT: v_readlane_b32 s36, v61, 36
; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:396 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_cvt_f32_f16_e32 v1, vcc_lo
+; SI-NEXT: v_cvt_f32_f16_e32 v1, s95
; SI-NEXT: s_and_b32 s35, s35, 0xff
; SI-NEXT: s_lshl_b32 s36, s36, 8
; SI-NEXT: s_or_b32 s35, s36, s35
@@ -186347,19 +186517,19 @@ define inreg <64 x half> @bitcast_v128i8_to_v64f16_scalar(<128 x i8> inreg %a, i
; SI-NEXT: s_addk_i32 s92, 0x300
; SI-NEXT: s_addk_i32 s93, 0x300
; SI-NEXT: s_addk_i32 s94, 0x300
-; SI-NEXT: s_addk_i32 s95, 0x300
+; SI-NEXT: s_addk_i32 vcc_lo, 0x300
; SI-NEXT: s_addk_i32 s34, 0x300
; SI-NEXT: s_addk_i32 s35, 0x300
; SI-NEXT: v_cvt_f32_f16_e32 v6, s35
; SI-NEXT: v_cvt_f32_f16_e32 v5, s34
+; SI-NEXT: v_cvt_f32_f16_e32 v7, vcc_lo
; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:404 ; 4-byte Folded Spill
-; SI-NEXT: v_cvt_f32_f16_e32 v7, s95
; SI-NEXT: v_cvt_f32_f16_e32 v10, s94
; SI-NEXT: v_cvt_f32_f16_e32 v8, s93
-; SI-NEXT: v_cvt_f32_f16_e32 v11, s92
+; SI-NEXT: v_cvt_f32_f16_e32 v12, s92
; SI-NEXT: v_cvt_f32_f16_e32 v9, s90
; SI-NEXT: v_cvt_f32_f16_e32 v13, s91
-; SI-NEXT: v_cvt_f32_f16_e32 v12, s88
+; SI-NEXT: v_cvt_f32_f16_e32 v11, s88
; SI-NEXT: v_cvt_f32_f16_e32 v15, s79
; SI-NEXT: v_cvt_f32_f16_e32 v14, s77
; SI-NEXT: v_cvt_f32_f16_e32 v17, s76
@@ -186412,7 +186582,7 @@ define inreg <64 x half> @bitcast_v128i8_to_v64f16_scalar(<128 x i8> inreg %a, i
; SI-NEXT: v_cvt_f32_f16_e32 v2, s6
; SI-NEXT: v_cvt_f32_f16_e32 v3, s5
; SI-NEXT: v_cvt_f32_f16_e32 v4, s4
-; SI-NEXT: .LBB93_3: ; %end
+; SI-NEXT: .LBB93_5: ; %end
; SI-NEXT: v_cvt_f16_f32_e32 v5, v5
; SI-NEXT: v_cvt_f16_f32_e32 v6, v6
; SI-NEXT: v_cvt_f16_f32_e32 v1, v1
@@ -186467,26 +186637,26 @@ define inreg <64 x half> @bitcast_v128i8_to_v64f16_scalar(<128 x i8> inreg %a, i
; SI-NEXT: v_or_b32_e32 v5, v6, v5
; SI-NEXT: v_add_i32_e32 v6, vcc, 4, v0
; SI-NEXT: buffer_store_dword v5, v6, s[0:3], 0 offen
-; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: buffer_load_dword v5, off, s[0:3], s32 offset:404 ; 4-byte Folded Reload
; SI-NEXT: buffer_load_dword v6, off, s[0:3], s32 offset:396 ; 4-byte Folded Reload
-; SI-NEXT: s_waitcnt vmcnt(1)
-; SI-NEXT: v_cvt_f16_f32_e32 v5, v5
+; SI-NEXT: s_waitcnt expcnt(0)
+; SI-NEXT: v_cvt_f16_f32_e32 v5, v7
+; SI-NEXT: v_add_i32_e32 v7, vcc, 12, v0
+; SI-NEXT: v_lshlrev_b32_e32 v5, 16, v5
; SI-NEXT: s_waitcnt vmcnt(0)
; SI-NEXT: v_cvt_f16_f32_e32 v6, v6
-; SI-NEXT: v_lshlrev_b32_e32 v5, 16, v5
; SI-NEXT: v_or_b32_e32 v5, v6, v5
; SI-NEXT: v_add_i32_e32 v6, vcc, 8, v0
; SI-NEXT: buffer_store_dword v5, v6, s[0:3], 0 offen
+; SI-NEXT: buffer_load_dword v6, off, s[0:3], s32 offset:404 ; 4-byte Folded Reload
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_cvt_f16_f32_e32 v5, v10
-; SI-NEXT: v_cvt_f16_f32_e32 v6, v7
-; SI-NEXT: v_add_i32_e32 v7, vcc, 12, v0
; SI-NEXT: v_lshlrev_b32_e32 v5, 16, v5
+; SI-NEXT: s_waitcnt vmcnt(0)
+; SI-NEXT: v_cvt_f16_f32_e32 v6, v6
; SI-NEXT: v_or_b32_e32 v5, v6, v5
; SI-NEXT: buffer_store_dword v5, v7, s[0:3], 0 offen
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_cvt_f16_f32_e32 v5, v11
+; SI-NEXT: v_cvt_f16_f32_e32 v5, v12
; SI-NEXT: v_cvt_f16_f32_e32 v6, v8
; SI-NEXT: v_add_i32_e32 v7, vcc, 16, v0
; SI-NEXT: v_lshlrev_b32_e32 v5, 16, v5
@@ -186501,7 +186671,7 @@ define inreg <64 x half> @bitcast_v128i8_to_v64f16_scalar(<128 x i8> inreg %a, i
; SI-NEXT: buffer_store_dword v5, v7, s[0:3], 0 offen
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_cvt_f16_f32_e32 v5, v15
-; SI-NEXT: v_cvt_f16_f32_e32 v6, v12
+; SI-NEXT: v_cvt_f16_f32_e32 v6, v11
; SI-NEXT: v_add_i32_e32 v7, vcc, 24, v0
; SI-NEXT: v_lshlrev_b32_e32 v5, 16, v5
; SI-NEXT: v_or_b32_e32 v5, v6, v5
@@ -186699,134 +186869,6 @@ define inreg <64 x half> @bitcast_v128i8_to_v64f16_scalar(<128 x i8> inreg %a, i
; SI-NEXT: s_mov_b64 exec, s[4:5]
; SI-NEXT: s_waitcnt vmcnt(0) expcnt(0)
; SI-NEXT: s_setpc_b64 s[30:31]
-; SI-NEXT: .LBB93_4:
-; SI-NEXT: ; implicit-def: $vgpr1
-; SI-NEXT: ; kill: killed $vgpr1
-; SI-NEXT: s_mov_b32 s17, s19
-; SI-NEXT: ; implicit-def: $vgpr1
-; SI-NEXT: ; kill: killed $vgpr1
-; SI-NEXT: s_mov_b32 s19, s50
-; SI-NEXT: ; implicit-def: $vgpr1
-; SI-NEXT: s_mov_b32 s23, s48
-; SI-NEXT: s_mov_b32 s26, s37
-; SI-NEXT: s_mov_b32 s28, s29
-; SI-NEXT: s_mov_b32 s29, s76
-; SI-NEXT: s_mov_b32 s59, s58
-; SI-NEXT: s_mov_b32 s56, s47
-; SI-NEXT: s_mov_b32 s46, s41
-; SI-NEXT: s_mov_b32 s12, s11
-; SI-NEXT: s_mov_b32 s11, s7
-; SI-NEXT: s_mov_b32 s7, s97
-; SI-NEXT: s_mov_b32 s97, s81
-; SI-NEXT: s_mov_b32 s81, s85
-; SI-NEXT: s_mov_b32 s6, s40
-; SI-NEXT: s_mov_b32 s40, s72
-; SI-NEXT: s_mov_b32 s45, s73
-; SI-NEXT: s_mov_b32 s15, s89
-; SI-NEXT: s_mov_b32 s24, s98
-; SI-NEXT: s_mov_b32 s20, s88
-; SI-NEXT: s_mov_b32 s99, s55
-; SI-NEXT: ; kill: killed $vgpr1
-; SI-NEXT: ; implicit-def: $vgpr1
-; SI-NEXT: v_readlane_b32 s75, v61, 7
-; SI-NEXT: v_readlane_b32 s76, v61, 3
-; SI-NEXT: v_readlane_b32 s77, v61, 5
-; SI-NEXT: v_readlane_b32 s78, v61, 4
-; SI-NEXT: v_readlane_b32 s92, v61, 6
-; SI-NEXT: v_readlane_b32 s39, v61, 9
-; SI-NEXT: v_readlane_b32 s37, v61, 8
-; SI-NEXT: v_readlane_b32 s30, v61, 10
-; SI-NEXT: v_readlane_b32 s48, v61, 11
-; SI-NEXT: v_readlane_b32 s52, v61, 13
-; SI-NEXT: v_readlane_b32 s35, v61, 12
-; SI-NEXT: v_readlane_b32 s50, v61, 15
-; SI-NEXT: v_readlane_b32 s64, v61, 14
-; SI-NEXT: v_readlane_b32 s54, v61, 17
-; SI-NEXT: v_readlane_b32 s67, v61, 16
-; SI-NEXT: v_readlane_b32 s65, v61, 18
-; SI-NEXT: v_readlane_b32 s70, v61, 19
-; SI-NEXT: v_readlane_b32 s49, v61, 21
-; SI-NEXT: v_readlane_b32 s71, v61, 20
-; SI-NEXT: v_readlane_b32 s80, v61, 23
-; SI-NEXT: v_readlane_b32 s83, v61, 22
-; SI-NEXT: v_readlane_b32 s84, v61, 25
-; SI-NEXT: v_readlane_b32 s82, v61, 24
-; SI-NEXT: v_readlane_b32 s87, v61, 26
-; SI-NEXT: v_readlane_b32 s86, v61, 27
-; SI-NEXT: v_readlane_b32 s96, v61, 29
-; SI-NEXT: v_readlane_b32 s51, v61, 28
-; SI-NEXT: s_mov_b32 s55, s93
-; SI-NEXT: s_mov_b32 s95, s91
-; SI-NEXT: v_readlane_b32 s94, v61, 31
-; SI-NEXT: s_mov_b32 s31, s90
-; SI-NEXT: v_readlane_b32 s34, v61, 30
-; SI-NEXT: v_readlane_b32 s53, v61, 32
-; SI-NEXT: v_readlane_b32 s66, v61, 33
-; SI-NEXT: v_readlane_b32 s68, v61, 34
-; SI-NEXT: v_readlane_b32 s69, v61, 35
-; SI-NEXT: v_readlane_b32 s8, v61, 36
-; SI-NEXT: ; implicit-def: $vgpr6
-; SI-NEXT: ; implicit-def: $vgpr5
-; SI-NEXT: ; kill: killed $vgpr1
-; SI-NEXT: ; implicit-def: $vgpr7
-; SI-NEXT: ; implicit-def: $vgpr10
-; SI-NEXT: ; implicit-def: $vgpr8
-; SI-NEXT: ; implicit-def: $vgpr11
-; SI-NEXT: ; implicit-def: $vgpr9
-; SI-NEXT: ; implicit-def: $vgpr13
-; SI-NEXT: ; implicit-def: $vgpr12
-; SI-NEXT: ; implicit-def: $vgpr15
-; SI-NEXT: ; implicit-def: $vgpr14
-; SI-NEXT: ; implicit-def: $vgpr17
-; SI-NEXT: ; implicit-def: $vgpr16
-; SI-NEXT: ; implicit-def: $vgpr19
-; SI-NEXT: ; implicit-def: $vgpr18
-; SI-NEXT: ; implicit-def: $vgpr21
-; SI-NEXT: ; implicit-def: $vgpr20
-; SI-NEXT: ; implicit-def: $vgpr23
-; SI-NEXT: ; implicit-def: $vgpr22
-; SI-NEXT: ; implicit-def: $vgpr25
-; SI-NEXT: ; implicit-def: $vgpr24
-; SI-NEXT: ; implicit-def: $vgpr27
-; SI-NEXT: ; implicit-def: $vgpr26
-; SI-NEXT: ; implicit-def: $vgpr29
-; SI-NEXT: ; implicit-def: $vgpr28
-; SI-NEXT: ; implicit-def: $vgpr31
-; SI-NEXT: ; implicit-def: $vgpr30
-; SI-NEXT: ; implicit-def: $vgpr33
-; SI-NEXT: ; implicit-def: $vgpr32
-; SI-NEXT: ; implicit-def: $vgpr35
-; SI-NEXT: ; implicit-def: $vgpr34
-; SI-NEXT: ; implicit-def: $vgpr37
-; SI-NEXT: ; implicit-def: $vgpr36
-; SI-NEXT: ; implicit-def: $vgpr39
-; SI-NEXT: ; implicit-def: $vgpr38
-; SI-NEXT: ; implicit-def: $vgpr49
-; SI-NEXT: ; implicit-def: $vgpr48
-; SI-NEXT: ; implicit-def: $vgpr51
-; SI-NEXT: ; implicit-def: $vgpr50
-; SI-NEXT: ; implicit-def: $vgpr53
-; SI-NEXT: ; implicit-def: $vgpr52
-; SI-NEXT: ; implicit-def: $vgpr55
-; SI-NEXT: ; implicit-def: $vgpr54
-; SI-NEXT: ; implicit-def: $vgpr41
-; SI-NEXT: ; implicit-def: $vgpr40
-; SI-NEXT: ; implicit-def: $vgpr43
-; SI-NEXT: ; implicit-def: $vgpr42
-; SI-NEXT: ; implicit-def: $vgpr45
-; SI-NEXT: ; implicit-def: $vgpr44
-; SI-NEXT: ; implicit-def: $vgpr47
-; SI-NEXT: ; implicit-def: $vgpr46
-; SI-NEXT: ; implicit-def: $vgpr57
-; SI-NEXT: ; implicit-def: $vgpr56
-; SI-NEXT: ; implicit-def: $vgpr59
-; SI-NEXT: ; implicit-def: $vgpr58
-; SI-NEXT: ; implicit-def: $vgpr1
-; SI-NEXT: ; implicit-def: $vgpr60
-; SI-NEXT: ; implicit-def: $vgpr2
-; SI-NEXT: ; implicit-def: $vgpr3
-; SI-NEXT: ; implicit-def: $vgpr4
-; SI-NEXT: s_branch .LBB93_2
;
; VI-LABEL: bitcast_v128i8_to_v64f16_scalar:
; VI: ; %bb.0:
@@ -186888,13 +186930,14 @@ define inreg <64 x half> @bitcast_v128i8_to_v64f16_scalar(<128 x i8> inreg %a, i
; VI-NEXT: buffer_load_ushort v37, off, s[0:3], s32 offset:168
; VI-NEXT: buffer_load_ushort v38, off, s[0:3], s32 offset:176
; VI-NEXT: v_lshlrev_b32_e32 v1, 8, v1
-; VI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:508 ; 4-byte Folded Spill
+; VI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:504 ; 4-byte Folded Spill
; VI-NEXT: v_lshlrev_b32_e32 v1, 8, v3
; VI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:464 ; 4-byte Folded Spill
; VI-NEXT: v_lshlrev_b32_e32 v1, 8, v5
; VI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:468 ; 4-byte Folded Spill
; VI-NEXT: v_lshlrev_b32_e32 v1, 8, v7
; VI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:472 ; 4-byte Folded Spill
+; VI-NEXT: v_lshlrev_b32_e32 v40, 8, v27
; VI-NEXT: v_lshlrev_b32_e32 v9, 8, v9
; VI-NEXT: v_lshlrev_b32_e32 v11, 8, v11
; VI-NEXT: v_lshlrev_b32_e32 v13, 8, v13
@@ -186906,46 +186949,42 @@ define inreg <64 x half> @bitcast_v128i8_to_v64f16_scalar(<128 x i8> inreg %a, i
; VI-NEXT: v_lshlrev_b32_e32 v25, 8, v25
; VI-NEXT: s_waitcnt vmcnt(14)
; VI-NEXT: v_cmp_ne_u32_e32 vcc, 0, v0
-; VI-NEXT: v_lshlrev_b32_e32 v0, 8, v27
-; VI-NEXT: buffer_store_dword v0, off, s[0:3], s32 offset:476 ; 4-byte Folded Spill
; VI-NEXT: v_lshlrev_b32_e32 v0, 8, v29
-; VI-NEXT: buffer_store_dword v0, off, s[0:3], s32 offset:480 ; 4-byte Folded Spill
+; VI-NEXT: buffer_store_dword v0, off, s[0:3], s32 offset:476 ; 4-byte Folded Spill
; VI-NEXT: v_lshlrev_b32_e32 v0, 8, v2
-; VI-NEXT: buffer_store_dword v0, off, s[0:3], s32 offset:484 ; 4-byte Folded Spill
+; VI-NEXT: buffer_store_dword v0, off, s[0:3], s32 offset:480 ; 4-byte Folded Spill
; VI-NEXT: v_lshlrev_b32_e32 v0, 8, v4
-; VI-NEXT: buffer_store_dword v0, off, s[0:3], s32 offset:488 ; 4-byte Folded Spill
+; VI-NEXT: buffer_store_dword v0, off, s[0:3], s32 offset:484 ; 4-byte Folded Spill
; VI-NEXT: v_lshlrev_b32_e32 v0, 8, v6
-; VI-NEXT: buffer_store_dword v0, off, s[0:3], s32 offset:492 ; 4-byte Folded Spill
+; VI-NEXT: buffer_store_dword v0, off, s[0:3], s32 offset:488 ; 4-byte Folded Spill
; VI-NEXT: v_lshlrev_b32_e32 v0, 8, v8
-; VI-NEXT: buffer_store_dword v0, off, s[0:3], s32 offset:496 ; 4-byte Folded Spill
+; VI-NEXT: buffer_store_dword v0, off, s[0:3], s32 offset:492 ; 4-byte Folded Spill
; VI-NEXT: v_lshlrev_b32_e32 v0, 8, v10
-; VI-NEXT: buffer_store_dword v0, off, s[0:3], s32 offset:500 ; 4-byte Folded Spill
+; VI-NEXT: buffer_store_dword v0, off, s[0:3], s32 offset:496 ; 4-byte Folded Spill
; VI-NEXT: v_lshlrev_b32_e32 v0, 8, v12
-; VI-NEXT: buffer_store_dword v0, off, s[0:3], s32 offset:504 ; 4-byte Folded Spill
-; VI-NEXT: v_lshlrev_b32_e32 v0, 8, v26
-; VI-NEXT: buffer_store_dword v0, off, s[0:3], s32 offset:512 ; 4-byte Folded Spill
+; VI-NEXT: buffer_store_dword v0, off, s[0:3], s32 offset:500 ; 4-byte Folded Spill
+; VI-NEXT: v_lshlrev_b32_e32 v0, 8, v24
+; VI-NEXT: buffer_store_dword v0, off, s[0:3], s32 offset:508 ; 4-byte Folded Spill
; VI-NEXT: s_waitcnt vmcnt(14)
-; VI-NEXT: v_lshlrev_b32_e32 v0, 8, v28
-; VI-NEXT: buffer_store_dword v0, off, s[0:3], s32 offset:516 ; 4-byte Folded Spill
; VI-NEXT: v_lshlrev_b32_e32 v0, 8, v30
-; VI-NEXT: buffer_store_dword v0, off, s[0:3], s32 offset:520 ; 4-byte Folded Spill
+; VI-NEXT: buffer_store_dword v0, off, s[0:3], s32 offset:512 ; 4-byte Folded Spill
; VI-NEXT: v_lshlrev_b32_e32 v0, 8, v31
-; VI-NEXT: buffer_store_dword v0, off, s[0:3], s32 offset:560 ; 4-byte Folded Spill
+; VI-NEXT: buffer_store_dword v0, off, s[0:3], s32 offset:556 ; 4-byte Folded Spill
; VI-NEXT: v_lshlrev_b32_e32 v0, 8, v32
-; VI-NEXT: buffer_store_dword v0, off, s[0:3], s32 offset:564 ; 4-byte Folded Spill
-; VI-NEXT: v_lshlrev_b32_e32 v0, 8, v33
-; VI-NEXT: buffer_store_dword v0, off, s[0:3], s32 offset:592 ; 4-byte Folded Spill
-; VI-NEXT: v_lshlrev_b32_e32 v0, 8, v34
; VI-NEXT: buffer_store_dword v0, off, s[0:3], s32 offset:588 ; 4-byte Folded Spill
-; VI-NEXT: v_lshlrev_b32_e32 v0, 8, v35
+; VI-NEXT: v_lshlrev_b32_e32 v0, 8, v33
; VI-NEXT: buffer_store_dword v0, off, s[0:3], s32 offset:584 ; 4-byte Folded Spill
-; VI-NEXT: v_lshlrev_b32_e32 v0, 8, v36
+; VI-NEXT: v_lshlrev_b32_e32 v0, 8, v34
; VI-NEXT: buffer_store_dword v0, off, s[0:3], s32 offset:576 ; 4-byte Folded Spill
-; VI-NEXT: v_lshlrev_b32_e32 v0, 8, v37
+; VI-NEXT: v_lshlrev_b32_e32 v0, 8, v35
; VI-NEXT: buffer_store_dword v0, off, s[0:3], s32 offset:572 ; 4-byte Folded Spill
+; VI-NEXT: v_lshlrev_b32_e32 v0, 8, v36
+; VI-NEXT: buffer_store_dword v0, off, s[0:3], s32 offset:564 ; 4-byte Folded Spill
; VI-NEXT: s_waitcnt vmcnt(14)
+; VI-NEXT: v_lshlrev_b32_e32 v0, 8, v37
+; VI-NEXT: buffer_store_dword v0, off, s[0:3], s32 offset:560 ; 4-byte Folded Spill
; VI-NEXT: v_lshlrev_b32_e32 v0, 8, v38
-; VI-NEXT: buffer_store_dword v0, off, s[0:3], s32 offset:580 ; 4-byte Folded Spill
+; VI-NEXT: buffer_store_dword v0, off, s[0:3], s32 offset:568 ; 4-byte Folded Spill
; VI-NEXT: buffer_load_ushort v0, off, s[0:3], s32 offset:184
; VI-NEXT: buffer_load_ushort v1, off, s[0:3], s32 offset:192
; VI-NEXT: buffer_load_ushort v2, off, s[0:3], s32 offset:200
@@ -186954,34 +186993,37 @@ define inreg <64 x half> @bitcast_v128i8_to_v64f16_scalar(<128 x i8> inreg %a, i
; VI-NEXT: buffer_load_ushort v5, off, s[0:3], s32 offset:224
; VI-NEXT: buffer_load_ushort v6, off, s[0:3], s32 offset:232
; VI-NEXT: buffer_load_ushort v7, off, s[0:3], s32 offset:240
-; VI-NEXT: v_lshlrev_b32_e32 v45, 8, v22
-; VI-NEXT: v_lshlrev_b32_e32 v8, 8, v24
+; VI-NEXT: v_lshlrev_b32_e32 v8, 8, v26
+; VI-NEXT: v_lshlrev_b32_e32 v10, 8, v28
; VI-NEXT: s_and_b64 s[4:5], vcc, exec
; VI-NEXT: v_lshlrev_b32_e32 v14, 8, v14
; VI-NEXT: v_lshlrev_b32_e32 v16, 8, v16
; VI-NEXT: v_lshlrev_b32_e32 v18, 8, v18
; VI-NEXT: v_lshlrev_b32_e32 v20, 8, v20
+; VI-NEXT: v_lshlrev_b32_e32 v22, 8, v22
; VI-NEXT: s_waitcnt vmcnt(7)
-; VI-NEXT: v_lshlrev_b32_e32 v22, 8, v0
+; VI-NEXT: v_lshlrev_b32_e32 v24, 8, v0
; VI-NEXT: s_waitcnt vmcnt(6)
; VI-NEXT: v_lshlrev_b32_e32 v0, 8, v1
-; VI-NEXT: buffer_store_dword v0, off, s[0:3], s32 offset:568 ; 4-byte Folded Spill
-; VI-NEXT: s_waitcnt vmcnt(5)
+; VI-NEXT: buffer_store_dword v0, off, s[0:3], s32 offset:592 ; 4-byte Folded Spill
+; VI-NEXT: s_waitcnt vmcnt(6)
+; VI-NEXT: v_lshlrev_b32_e32 v0, 8, v2
+; VI-NEXT: buffer_store_dword v0, off, s[0:3], s32 offset:580 ; 4-byte Folded Spill
+; VI-NEXT: s_waitcnt vmcnt(6)
; VI-NEXT: v_lshlrev_b32_e32 v0, 8, v3
; VI-NEXT: buffer_store_dword v0, off, s[0:3], s32 offset:600 ; 4-byte Folded Spill
-; VI-NEXT: s_waitcnt vmcnt(5)
+; VI-NEXT: s_waitcnt vmcnt(6)
; VI-NEXT: v_lshlrev_b32_e32 v0, 8, v4
; VI-NEXT: buffer_store_dword v0, off, s[0:3], s32 offset:596 ; 4-byte Folded Spill
-; VI-NEXT: s_waitcnt vmcnt(5)
+; VI-NEXT: s_waitcnt vmcnt(6)
; VI-NEXT: v_lshlrev_b32_e32 v0, 8, v5
; VI-NEXT: buffer_store_dword v0, off, s[0:3], s32 offset:608 ; 4-byte Folded Spill
-; VI-NEXT: s_waitcnt vmcnt(5)
+; VI-NEXT: s_waitcnt vmcnt(6)
; VI-NEXT: v_lshlrev_b32_e32 v0, 8, v6
; VI-NEXT: buffer_store_dword v0, off, s[0:3], s32 offset:612 ; 4-byte Folded Spill
-; VI-NEXT: s_waitcnt vmcnt(5)
+; VI-NEXT: s_waitcnt vmcnt(6)
; VI-NEXT: v_lshlrev_b32_e32 v0, 8, v7
; VI-NEXT: buffer_store_dword v0, off, s[0:3], s32 offset:604 ; 4-byte Folded Spill
-; VI-NEXT: v_lshlrev_b32_e32 v24, 8, v2
; VI-NEXT: buffer_load_ushort v0, off, s[0:3], s32 offset:248
; VI-NEXT: buffer_load_ushort v1, off, s[0:3], s32 offset:256
; VI-NEXT: buffer_load_ushort v2, off, s[0:3], s32 offset:264
@@ -187000,6 +187042,11 @@ define inreg <64 x half> @bitcast_v128i8_to_v64f16_scalar(<128 x i8> inreg %a, i
; VI-NEXT: s_waitcnt vmcnt(4)
; VI-NEXT: v_lshlrev_b32_e32 v0, 8, v4
; VI-NEXT: buffer_store_dword v0, off, s[0:3], s32 offset:620 ; 4-byte Folded Spill
+; VI-NEXT: s_waitcnt vmcnt(3)
+; VI-NEXT: v_lshlrev_b32_e32 v4, 8, v6
+; VI-NEXT: s_waitcnt vmcnt(2)
+; VI-NEXT: v_lshlrev_b32_e32 v0, 8, v7
+; VI-NEXT: buffer_store_dword v0, off, s[0:3], s32 offset:544 ; 4-byte Folded Spill
; VI-NEXT: buffer_load_ushort v0, off, s[0:3], s32 offset:312
; VI-NEXT: buffer_load_ushort v1, off, s[0:3], s32 offset:320
; VI-NEXT: buffer_load_ushort v2, off, s[0:3], s32 offset:328
@@ -187008,12 +187055,8 @@ define inreg <64 x half> @bitcast_v128i8_to_v64f16_scalar(<128 x i8> inreg %a, i
; VI-NEXT: buffer_load_ushort v33, off, s[0:3], s32 offset:20
; VI-NEXT: buffer_load_ushort v34, off, s[0:3], s32 offset:28
; VI-NEXT: buffer_load_ushort v35, off, s[0:3], s32 offset:36
-; VI-NEXT: s_waitcnt vmcnt(11)
-; VI-NEXT: v_lshlrev_b32_e32 v4, 8, v6
; VI-NEXT: v_lshlrev_b32_e32 v28, 8, v3
; VI-NEXT: v_lshlrev_b32_e32 v3, 8, v5
-; VI-NEXT: s_waitcnt vmcnt(10)
-; VI-NEXT: v_lshlrev_b32_e32 v5, 8, v7
; VI-NEXT: s_waitcnt vmcnt(7)
; VI-NEXT: v_lshlrev_b32_e32 v6, 8, v0
; VI-NEXT: buffer_load_ushort v36, off, s[0:3], s32 offset:44
@@ -187022,47 +187065,45 @@ define inreg <64 x half> @bitcast_v128i8_to_v64f16_scalar(<128 x i8> inreg %a, i
; VI-NEXT: buffer_load_ushort v39, off, s[0:3], s32 offset:68
; VI-NEXT: buffer_load_ushort v48, off, s[0:3], s32 offset:76
; VI-NEXT: buffer_load_ushort v49, off, s[0:3], s32 offset:84
-; VI-NEXT: buffer_load_ushort v50, off, s[0:3], s32 offset:92
-; VI-NEXT: buffer_load_ushort v51, off, s[0:3], s32 offset:100
-; VI-NEXT: buffer_load_ushort v54, off, s[0:3], s32 offset:108
+; VI-NEXT: buffer_load_ushort v51, off, s[0:3], s32 offset:92
+; VI-NEXT: buffer_load_ushort v53, off, s[0:3], s32 offset:100
+; VI-NEXT: buffer_load_ushort v55, off, s[0:3], s32 offset:108
; VI-NEXT: buffer_load_ushort v41, off, s[0:3], s32 offset:116
; VI-NEXT: buffer_load_ushort v56, off, s[0:3], s32 offset:124
-; VI-NEXT: buffer_load_ushort v55, off, s[0:3], s32 offset:132
-; VI-NEXT: buffer_load_ushort v52, off, s[0:3], s32 offset:140
-; VI-NEXT: buffer_load_ushort v43, off, s[0:3], s32 offset:148
-; VI-NEXT: buffer_load_ushort v53, off, s[0:3], s32 offset:156
-; VI-NEXT: buffer_load_ushort v46, off, s[0:3], s32 offset:164
-; VI-NEXT: buffer_load_ushort v58, off, s[0:3], s32 offset:172
+; VI-NEXT: buffer_load_ushort v50, off, s[0:3], s32 offset:132
+; VI-NEXT: buffer_load_ushort v43, off, s[0:3], s32 offset:140
+; VI-NEXT: buffer_load_ushort v58, off, s[0:3], s32 offset:148
+; VI-NEXT: buffer_load_ushort v52, off, s[0:3], s32 offset:156
+; VI-NEXT: buffer_load_ushort v60, off, s[0:3], s32 offset:164
+; VI-NEXT: buffer_load_ushort v54, off, s[0:3], s32 offset:172
; VI-NEXT: buffer_load_ushort v47, off, s[0:3], s32 offset:180
; VI-NEXT: buffer_load_ushort v63, off, s[0:3], s32 offset:188
-; VI-NEXT: buffer_load_ushort v57, off, s[0:3], s32 offset:196
+; VI-NEXT: buffer_load_ushort v46, off, s[0:3], s32 offset:196
; VI-NEXT: buffer_load_ushort v61, off, s[0:3], s32 offset:204
-; VI-NEXT: buffer_load_ushort v60, off, s[0:3], s32 offset:212
+; VI-NEXT: buffer_load_ushort v57, off, s[0:3], s32 offset:212
; VI-NEXT: buffer_load_ushort v59, off, s[0:3], s32 offset:220
-; VI-NEXT: buffer_load_ushort v40, off, s[0:3], s32 offset:228
-; VI-NEXT: buffer_load_ushort v42, off, s[0:3], s32 offset:236
-; VI-NEXT: buffer_load_ushort v62, off, s[0:3], s32 offset:244
-; VI-NEXT: buffer_load_ushort v44, off, s[0:3], s32 offset:252
+; VI-NEXT: buffer_load_ushort v42, off, s[0:3], s32 offset:228
+; VI-NEXT: buffer_load_ushort v44, off, s[0:3], s32 offset:236
+; VI-NEXT: buffer_load_ushort v45, off, s[0:3], s32 offset:244
+; VI-NEXT: buffer_load_ushort v62, off, s[0:3], s32 offset:252
; VI-NEXT: buffer_load_ushort v0, off, s[0:3], s32 offset:260
; VI-NEXT: s_waitcnt vmcnt(14)
-; VI-NEXT: v_lshlrev_b32_e32 v1, 8, v1
-; VI-NEXT: v_lshlrev_b32_e32 v2, 8, v2
-; VI-NEXT: s_waitcnt vmcnt(0)
-; VI-NEXT: buffer_store_dword v0, off, s[0:3], s32 offset:552 ; 4-byte Folded Spill
-; VI-NEXT: buffer_load_ushort v29, off, s[0:3], s32 offset:268
-; VI-NEXT: buffer_load_ushort v0, off, s[0:3], s32 offset:276
+; VI-NEXT: v_lshlrev_b32_e32 v5, 8, v1
+; VI-NEXT: v_lshlrev_b32_e32 v1, 8, v2
; VI-NEXT: s_waitcnt vmcnt(0)
-; VI-NEXT: buffer_store_dword v0, off, s[0:3], s32 offset:556 ; 4-byte Folded Spill
-; VI-NEXT: buffer_load_ushort v30, off, s[0:3], s32 offset:284
-; VI-NEXT: buffer_load_ushort v31, off, s[0:3], s32 offset:292
-; VI-NEXT: buffer_load_ushort v0, off, s[0:3], s32 offset:300
-; VI-NEXT: buffer_load_ushort v7, off, s[0:3], s32 offset:308
+; VI-NEXT: buffer_store_dword v0, off, s[0:3], s32 offset:548 ; 4-byte Folded Spill
+; VI-NEXT: buffer_load_ushort v0, off, s[0:3], s32 offset:268
; VI-NEXT: s_waitcnt vmcnt(0)
-; VI-NEXT: buffer_store_dword v7, off, s[0:3], s32 offset:528 ; 4-byte Folded Spill
-; VI-NEXT: buffer_load_ushort v7, off, s[0:3], s32 offset:316
-; VI-NEXT: buffer_load_ushort v10, off, s[0:3], s32 offset:324
+; VI-NEXT: buffer_store_dword v0, off, s[0:3], s32 offset:552 ; 4-byte Folded Spill
+; VI-NEXT: buffer_load_ushort v29, off, s[0:3], s32 offset:276
+; VI-NEXT: buffer_load_ushort v31, off, s[0:3], s32 offset:284
+; VI-NEXT: buffer_load_ushort v30, off, s[0:3], s32 offset:292
+; VI-NEXT: buffer_load_ushort v7, off, s[0:3], s32 offset:300
+; VI-NEXT: buffer_load_ushort v0, off, s[0:3], s32 offset:308
; VI-NEXT: s_waitcnt vmcnt(0)
-; VI-NEXT: buffer_store_dword v10, off, s[0:3], s32 offset:540 ; 4-byte Folded Spill
+; VI-NEXT: buffer_store_dword v0, off, s[0:3], s32 offset:520 ; 4-byte Folded Spill
+; VI-NEXT: buffer_load_ushort v0, off, s[0:3], s32 offset:316
+; VI-NEXT: buffer_load_ushort v2, off, s[0:3], s32 offset:324
; VI-NEXT: buffer_store_dword v9, off, s[0:3], s32 offset:624 ; 4-byte Folded Spill
; VI-NEXT: buffer_store_dword v11, off, s[0:3], s32 offset:628 ; 4-byte Folded Spill
; VI-NEXT: buffer_store_dword v13, off, s[0:3], s32 offset:632 ; 4-byte Folded Spill
@@ -187072,46 +187113,50 @@ define inreg <64 x half> @bitcast_v128i8_to_v64f16_scalar(<128 x i8> inreg %a, i
; VI-NEXT: buffer_store_dword v21, off, s[0:3], s32 offset:648 ; 4-byte Folded Spill
; VI-NEXT: buffer_store_dword v23, off, s[0:3], s32 offset:652 ; 4-byte Folded Spill
; VI-NEXT: buffer_store_dword v25, off, s[0:3], s32 offset:656 ; 4-byte Folded Spill
-; VI-NEXT: buffer_store_dword v14, off, s[0:3], s32 offset:660 ; 4-byte Folded Spill
-; VI-NEXT: buffer_store_dword v16, off, s[0:3], s32 offset:664 ; 4-byte Folded Spill
-; VI-NEXT: buffer_store_dword v18, off, s[0:3], s32 offset:668 ; 4-byte Folded Spill
-; VI-NEXT: buffer_store_dword v20, off, s[0:3], s32 offset:672 ; 4-byte Folded Spill
-; VI-NEXT: buffer_store_dword v45, off, s[0:3], s32 offset:676 ; 4-byte Folded Spill
-; VI-NEXT: buffer_store_dword v8, off, s[0:3], s32 offset:680 ; 4-byte Folded Spill
+; VI-NEXT: buffer_store_dword v40, off, s[0:3], s32 offset:660 ; 4-byte Folded Spill
+; VI-NEXT: buffer_store_dword v14, off, s[0:3], s32 offset:664 ; 4-byte Folded Spill
+; VI-NEXT: buffer_store_dword v16, off, s[0:3], s32 offset:668 ; 4-byte Folded Spill
+; VI-NEXT: buffer_store_dword v18, off, s[0:3], s32 offset:672 ; 4-byte Folded Spill
+; VI-NEXT: buffer_store_dword v20, off, s[0:3], s32 offset:676 ; 4-byte Folded Spill
+; VI-NEXT: buffer_store_dword v22, off, s[0:3], s32 offset:680 ; 4-byte Folded Spill
; VI-NEXT: buffer_store_dword v12, off, s[0:3], s32 offset:684 ; 4-byte Folded Spill
; VI-NEXT: buffer_store_dword v32, off, s[0:3], s32 offset:688 ; 4-byte Folded Spill
; VI-NEXT: buffer_store_dword v33, off, s[0:3], s32 offset:692 ; 4-byte Folded Spill
; VI-NEXT: buffer_store_dword v34, off, s[0:3], s32 offset:696 ; 4-byte Folded Spill
; VI-NEXT: buffer_store_dword v35, off, s[0:3], s32 offset:700 ; 4-byte Folded Spill
-; VI-NEXT: buffer_store_dword v36, off, s[0:3], s32 offset:704 ; 4-byte Folded Spill
-; VI-NEXT: buffer_store_dword v37, off, s[0:3], s32 offset:708 ; 4-byte Folded Spill
-; VI-NEXT: buffer_store_dword v38, off, s[0:3], s32 offset:712 ; 4-byte Folded Spill
-; VI-NEXT: buffer_store_dword v39, off, s[0:3], s32 offset:716 ; 4-byte Folded Spill
-; VI-NEXT: buffer_store_dword v48, off, s[0:3], s32 offset:720 ; 4-byte Folded Spill
-; VI-NEXT: buffer_store_dword v49, off, s[0:3], s32 offset:724 ; 4-byte Folded Spill
-; VI-NEXT: buffer_store_dword v50, off, s[0:3], s32 offset:728 ; 4-byte Folded Spill
-; VI-NEXT: buffer_store_dword v51, off, s[0:3], s32 offset:732 ; 4-byte Folded Spill
-; VI-NEXT: buffer_store_dword v54, off, s[0:3], s32 offset:736 ; 4-byte Folded Spill
-; VI-NEXT: buffer_store_dword v41, off, s[0:3], s32 offset:740 ; 4-byte Folded Spill
-; VI-NEXT: buffer_store_dword v56, off, s[0:3], s32 offset:744 ; 4-byte Folded Spill
-; VI-NEXT: buffer_store_dword v55, off, s[0:3], s32 offset:748 ; 4-byte Folded Spill
-; VI-NEXT: buffer_store_dword v52, off, s[0:3], s32 offset:752 ; 4-byte Folded Spill
-; VI-NEXT: buffer_store_dword v43, off, s[0:3], s32 offset:756 ; 4-byte Folded Spill
-; VI-NEXT: buffer_store_dword v53, off, s[0:3], s32 offset:760 ; 4-byte Folded Spill
-; VI-NEXT: buffer_store_dword v46, off, s[0:3], s32 offset:764 ; 4-byte Folded Spill
-; VI-NEXT: buffer_store_dword v58, off, s[0:3], s32 offset:768 ; 4-byte Folded Spill
-; VI-NEXT: buffer_store_dword v47, off, s[0:3], s32 offset:772 ; 4-byte Folded Spill
-; VI-NEXT: buffer_store_dword v63, off, s[0:3], s32 offset:776 ; 4-byte Folded Spill
-; VI-NEXT: buffer_store_dword v57, off, s[0:3], s32 offset:780 ; 4-byte Folded Spill
-; VI-NEXT: buffer_store_dword v61, off, s[0:3], s32 offset:784 ; 4-byte Folded Spill
-; VI-NEXT: buffer_store_dword v29, off, s[0:3], s32 offset:788 ; 4-byte Folded Spill
-; VI-NEXT: buffer_store_dword v30, off, s[0:3], s32 offset:792 ; 4-byte Folded Spill
-; VI-NEXT: buffer_store_dword v31, off, s[0:3], s32 offset:796 ; 4-byte Folded Spill
+; VI-NEXT: buffer_store_dword v8, off, s[0:3], s32 offset:704 ; 4-byte Folded Spill
+; VI-NEXT: buffer_store_dword v36, off, s[0:3], s32 offset:708 ; 4-byte Folded Spill
+; VI-NEXT: buffer_store_dword v37, off, s[0:3], s32 offset:712 ; 4-byte Folded Spill
+; VI-NEXT: buffer_store_dword v38, off, s[0:3], s32 offset:716 ; 4-byte Folded Spill
+; VI-NEXT: buffer_store_dword v39, off, s[0:3], s32 offset:720 ; 4-byte Folded Spill
+; VI-NEXT: buffer_store_dword v10, off, s[0:3], s32 offset:724 ; 4-byte Folded Spill
+; VI-NEXT: buffer_store_dword v48, off, s[0:3], s32 offset:728 ; 4-byte Folded Spill
+; VI-NEXT: buffer_store_dword v49, off, s[0:3], s32 offset:732 ; 4-byte Folded Spill
+; VI-NEXT: buffer_store_dword v51, off, s[0:3], s32 offset:736 ; 4-byte Folded Spill
+; VI-NEXT: buffer_store_dword v53, off, s[0:3], s32 offset:740 ; 4-byte Folded Spill
+; VI-NEXT: buffer_store_dword v55, off, s[0:3], s32 offset:744 ; 4-byte Folded Spill
+; VI-NEXT: buffer_store_dword v41, off, s[0:3], s32 offset:748 ; 4-byte Folded Spill
+; VI-NEXT: buffer_store_dword v56, off, s[0:3], s32 offset:752 ; 4-byte Folded Spill
+; VI-NEXT: buffer_store_dword v50, off, s[0:3], s32 offset:756 ; 4-byte Folded Spill
+; VI-NEXT: buffer_store_dword v43, off, s[0:3], s32 offset:760 ; 4-byte Folded Spill
+; VI-NEXT: buffer_store_dword v58, off, s[0:3], s32 offset:764 ; 4-byte Folded Spill
+; VI-NEXT: buffer_store_dword v52, off, s[0:3], s32 offset:768 ; 4-byte Folded Spill
+; VI-NEXT: buffer_store_dword v60, off, s[0:3], s32 offset:772 ; 4-byte Folded Spill
+; VI-NEXT: buffer_store_dword v54, off, s[0:3], s32 offset:776 ; 4-byte Folded Spill
+; VI-NEXT: buffer_store_dword v47, off, s[0:3], s32 offset:780 ; 4-byte Folded Spill
+; VI-NEXT: buffer_store_dword v63, off, s[0:3], s32 offset:784 ; 4-byte Folded Spill
+; VI-NEXT: buffer_store_dword v46, off, s[0:3], s32 offset:788 ; 4-byte Folded Spill
+; VI-NEXT: buffer_store_dword v61, off, s[0:3], s32 offset:792 ; 4-byte Folded Spill
+; VI-NEXT: buffer_store_dword v29, off, s[0:3], s32 offset:796 ; 4-byte Folded Spill
+; VI-NEXT: buffer_store_dword v31, off, s[0:3], s32 offset:800 ; 4-byte Folded Spill
+; VI-NEXT: buffer_store_dword v30, off, s[0:3], s32 offset:804 ; 4-byte Folded Spill
+; VI-NEXT: buffer_store_dword v7, off, s[0:3], s32 offset:516 ; 4-byte Folded Spill
+; VI-NEXT: s_waitcnt vmcnt(14)
; VI-NEXT: buffer_store_dword v0, off, s[0:3], s32 offset:524 ; 4-byte Folded Spill
+; VI-NEXT: buffer_store_dword v2, off, s[0:3], s32 offset:528 ; 4-byte Folded Spill
; VI-NEXT: buffer_store_dword v6, off, s[0:3], s32 offset:532 ; 4-byte Folded Spill
-; VI-NEXT: buffer_store_dword v7, off, s[0:3], s32 offset:536 ; 4-byte Folded Spill
-; VI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:544 ; 4-byte Folded Spill
-; VI-NEXT: buffer_store_dword v2, off, s[0:3], s32 offset:548 ; 4-byte Folded Spill
+; VI-NEXT: buffer_store_dword v5, off, s[0:3], s32 offset:536 ; 4-byte Folded Spill
+; VI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:540 ; 4-byte Folded Spill
; VI-NEXT: s_cbranch_scc0 .LBB93_4
; VI-NEXT: ; %bb.1: ; %cmp.false
; VI-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:400 ; 4-byte Folded Reload
@@ -187128,11 +187173,10 @@ define inreg <64 x half> @bitcast_v128i8_to_v64f16_scalar(<128 x i8> inreg %a, i
; VI-NEXT: v_or_b32_sdwa v0, v0, v1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
; VI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:404 ; 4-byte Folded Reload
; VI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:468 ; 4-byte Folded Reload
-; VI-NEXT: buffer_store_dword v3, off, s[0:3], s32 offset:800 ; 4-byte Folded Spill
+; VI-NEXT: buffer_store_dword v3, off, s[0:3], s32 offset:808 ; 4-byte Folded Spill
; VI-NEXT: buffer_load_dword v3, off, s[0:3], s32 offset:472 ; 4-byte Folded Reload
-; VI-NEXT: buffer_store_dword v4, off, s[0:3], s32 offset:808 ; 4-byte Folded Spill
-; VI-NEXT: buffer_store_dword v5, off, s[0:3], s32 offset:804 ; 4-byte Folded Spill
-; VI-NEXT: s_waitcnt vmcnt(4)
+; VI-NEXT: buffer_store_dword v4, off, s[0:3], s32 offset:812 ; 4-byte Folded Spill
+; VI-NEXT: s_waitcnt vmcnt(3)
; VI-NEXT: v_or_b32_sdwa v1, v1, v2 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
; VI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:408 ; 4-byte Folded Reload
; VI-NEXT: v_or_b32_sdwa v4, v0, v1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_0 src1_sel:DWORD
@@ -187156,6 +187200,7 @@ define inreg <64 x half> @bitcast_v128i8_to_v64f16_scalar(<128 x i8> inreg %a, i
; VI-NEXT: v_or_b32_sdwa v0, v0, v15 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
; VI-NEXT: s_waitcnt vmcnt(0)
; VI-NEXT: v_or_b32_sdwa v1, v1, v17 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; VI-NEXT: v_mov_b32_e32 v17, v10
; VI-NEXT: v_or_b32_sdwa v7, v0, v1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_0 src1_sel:DWORD
; VI-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:432 ; 4-byte Folded Reload
; VI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:436 ; 4-byte Folded Reload
@@ -187172,38 +187217,43 @@ define inreg <64 x half> @bitcast_v128i8_to_v64f16_scalar(<128 x i8> inreg %a, i
; VI-NEXT: v_or_b32_sdwa v1, v1, v25 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
; VI-NEXT: v_or_b32_sdwa v9, v0, v1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_0 src1_sel:DWORD
; VI-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:448 ; 4-byte Folded Reload
-; VI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:476 ; 4-byte Folded Reload
-; VI-NEXT: s_waitcnt vmcnt(0)
-; VI-NEXT: v_or_b32_sdwa v0, v0, v1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
; VI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:452 ; 4-byte Folded Reload
-; VI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:480 ; 4-byte Folded Reload
+; VI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:476 ; 4-byte Folded Reload
+; VI-NEXT: s_waitcnt vmcnt(2)
+; VI-NEXT: v_or_b32_sdwa v0, v0, v40 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
; VI-NEXT: s_waitcnt vmcnt(0)
; VI-NEXT: v_or_b32_sdwa v1, v1, v2 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; VI-NEXT: v_mov_b32_e32 v40, v42
+; VI-NEXT: v_mov_b32_e32 v42, v44
+; VI-NEXT: v_mov_b32_e32 v44, v45
+; VI-NEXT: v_mov_b32_e32 v45, v62
+; VI-NEXT: v_or_b32_sdwa v2, v51, v3 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; VI-NEXT: v_or_b32_sdwa v3, v53, v17 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
; VI-NEXT: v_or_b32_sdwa v10, v0, v1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_0 src1_sel:DWORD
; VI-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:456 ; 4-byte Folded Reload
-; VI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:484 ; 4-byte Folded Reload
-; VI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:512 ; 4-byte Folded Reload
+; VI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:480 ; 4-byte Folded Reload
+; VI-NEXT: v_or_b32_sdwa v17, v2, v3 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_0 src1_sel:DWORD
+; VI-NEXT: buffer_load_dword v62, off, s[0:3], s32 offset:548 ; 4-byte Folded Reload
; VI-NEXT: s_waitcnt vmcnt(1)
; VI-NEXT: v_or_b32_sdwa v0, v0, v1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
-; VI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:488 ; 4-byte Folded Reload
-; VI-NEXT: s_waitcnt vmcnt(1)
-; VI-NEXT: v_or_b32_sdwa v2, v50, v2 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; VI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:484 ; 4-byte Folded Reload
; VI-NEXT: s_waitcnt vmcnt(0)
; VI-NEXT: v_or_b32_sdwa v1, v12, v1 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
; VI-NEXT: v_or_b32_sdwa v11, v0, v1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_0 src1_sel:DWORD
-; VI-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:492 ; 4-byte Folded Reload
-; VI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:496 ; 4-byte Folded Reload
+; VI-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:488 ; 4-byte Folded Reload
+; VI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:492 ; 4-byte Folded Reload
; VI-NEXT: s_waitcnt vmcnt(1)
; VI-NEXT: v_or_b32_sdwa v0, v32, v0 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
; VI-NEXT: s_waitcnt vmcnt(0)
; VI-NEXT: v_or_b32_sdwa v1, v33, v1 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
; VI-NEXT: v_or_b32_sdwa v12, v0, v1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_0 src1_sel:DWORD
-; VI-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:500 ; 4-byte Folded Reload
-; VI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:504 ; 4-byte Folded Reload
+; VI-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:496 ; 4-byte Folded Reload
+; VI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:500 ; 4-byte Folded Reload
; VI-NEXT: s_waitcnt vmcnt(1)
; VI-NEXT: v_or_b32_sdwa v0, v34, v0 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
; VI-NEXT: s_waitcnt vmcnt(0)
; VI-NEXT: v_or_b32_sdwa v1, v35, v1 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; VI-NEXT: v_mov_b32_e32 v34, v24
; VI-NEXT: v_or_b32_sdwa v13, v0, v1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_0 src1_sel:DWORD
; VI-NEXT: v_or_b32_sdwa v0, v36, v14 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
; VI-NEXT: v_or_b32_sdwa v1, v37, v16 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
@@ -187211,77 +187261,74 @@ define inreg <64 x half> @bitcast_v128i8_to_v64f16_scalar(<128 x i8> inreg %a, i
; VI-NEXT: v_or_b32_sdwa v0, v38, v18 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
; VI-NEXT: v_or_b32_sdwa v1, v39, v20 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
; VI-NEXT: v_or_b32_sdwa v15, v0, v1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_0 src1_sel:DWORD
-; VI-NEXT: v_or_b32_sdwa v0, v48, v45 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
-; VI-NEXT: v_or_b32_sdwa v1, v49, v3 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
-; VI-NEXT: buffer_load_dword v3, off, s[0:3], s32 offset:516 ; 4-byte Folded Reload
-; VI-NEXT: v_mov_b32_e32 v45, v62
+; VI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:508 ; 4-byte Folded Reload
+; VI-NEXT: v_or_b32_sdwa v0, v48, v22 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; VI-NEXT: s_waitcnt vmcnt(0)
+; VI-NEXT: v_or_b32_sdwa v1, v49, v1 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
; VI-NEXT: v_or_b32_sdwa v16, v0, v1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_0 src1_sel:DWORD
-; VI-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:520 ; 4-byte Folded Reload
-; VI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:560 ; 4-byte Folded Reload
-; VI-NEXT: buffer_load_dword v62, off, s[0:3], s32 offset:552 ; 4-byte Folded Reload
-; VI-NEXT: s_waitcnt vmcnt(3)
-; VI-NEXT: v_or_b32_sdwa v3, v51, v3 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
-; VI-NEXT: v_or_b32_sdwa v17, v2, v3 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_0 src1_sel:DWORD
-; VI-NEXT: s_waitcnt vmcnt(2)
-; VI-NEXT: v_or_b32_sdwa v0, v54, v0 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; VI-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:512 ; 4-byte Folded Reload
+; VI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:556 ; 4-byte Folded Reload
; VI-NEXT: s_waitcnt vmcnt(1)
+; VI-NEXT: v_or_b32_sdwa v0, v55, v0 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; VI-NEXT: s_waitcnt vmcnt(0)
; VI-NEXT: v_mov_b32_e32 v32, v1
; VI-NEXT: v_or_b32_sdwa v1, v41, v1 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
-; VI-NEXT: v_mov_b32_e32 v54, v22
-; VI-NEXT: v_mov_b32_e32 v41, v24
; VI-NEXT: v_or_b32_sdwa v18, v0, v1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_0 src1_sel:DWORD
-; VI-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:564 ; 4-byte Folded Reload
-; VI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:592 ; 4-byte Folded Reload
+; VI-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:588 ; 4-byte Folded Reload
+; VI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:584 ; 4-byte Folded Reload
; VI-NEXT: s_waitcnt vmcnt(1)
-; VI-NEXT: v_mov_b32_e32 v34, v0
+; VI-NEXT: v_mov_b32_e32 v33, v0
; VI-NEXT: v_or_b32_sdwa v0, v56, v0 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
; VI-NEXT: s_waitcnt vmcnt(0)
; VI-NEXT: v_mov_b32_e32 v37, v1
-; VI-NEXT: v_or_b32_sdwa v1, v55, v1 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
-; VI-NEXT: v_mov_b32_e32 v55, v26
+; VI-NEXT: v_or_b32_sdwa v1, v50, v1 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; VI-NEXT: v_mov_b32_e32 v50, v26
; VI-NEXT: v_or_b32_sdwa v19, v0, v1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_0 src1_sel:DWORD
-; VI-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:588 ; 4-byte Folded Reload
-; VI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:584 ; 4-byte Folded Reload
+; VI-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:576 ; 4-byte Folded Reload
+; VI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:572 ; 4-byte Folded Reload
; VI-NEXT: s_waitcnt vmcnt(1)
; VI-NEXT: v_mov_b32_e32 v39, v0
-; VI-NEXT: v_or_b32_sdwa v0, v52, v0 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; VI-NEXT: v_or_b32_sdwa v0, v43, v0 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
; VI-NEXT: s_waitcnt vmcnt(0)
; VI-NEXT: v_mov_b32_e32 v49, v1
-; VI-NEXT: v_or_b32_sdwa v1, v43, v1 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
-; VI-NEXT: v_mov_b32_e32 v43, v27
+; VI-NEXT: v_or_b32_sdwa v1, v58, v1 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
; VI-NEXT: v_or_b32_sdwa v20, v0, v1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_0 src1_sel:DWORD
-; VI-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:576 ; 4-byte Folded Reload
-; VI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:572 ; 4-byte Folded Reload
+; VI-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:564 ; 4-byte Folded Reload
+; VI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:560 ; 4-byte Folded Reload
; VI-NEXT: s_waitcnt vmcnt(1)
; VI-NEXT: v_mov_b32_e32 v51, v0
-; VI-NEXT: v_or_b32_sdwa v0, v53, v0 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; VI-NEXT: v_or_b32_sdwa v0, v52, v0 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
; VI-NEXT: s_waitcnt vmcnt(0)
-; VI-NEXT: v_mov_b32_e32 v35, v1
-; VI-NEXT: v_or_b32_sdwa v1, v46, v1 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
-; VI-NEXT: v_mov_b32_e32 v53, v28
+; VI-NEXT: v_mov_b32_e32 v53, v1
+; VI-NEXT: v_or_b32_sdwa v1, v60, v1 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; VI-NEXT: v_mov_b32_e32 v52, v28
; VI-NEXT: v_or_b32_sdwa v21, v0, v1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_0 src1_sel:DWORD
-; VI-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:580 ; 4-byte Folded Reload
-; VI-NEXT: v_or_b32_sdwa v1, v47, v22 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
-; VI-NEXT: s_waitcnt vmcnt(0)
-; VI-NEXT: v_mov_b32_e32 v33, v0
-; VI-NEXT: v_or_b32_sdwa v0, v58, v0 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
-; VI-NEXT: v_or_b32_sdwa v22, v0, v1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_0 src1_sel:DWORD
; VI-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:568 ; 4-byte Folded Reload
-; VI-NEXT: v_or_b32_sdwa v1, v57, v24 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; VI-NEXT: v_or_b32_sdwa v1, v47, v24 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
; VI-NEXT: s_waitcnt vmcnt(0)
-; VI-NEXT: v_mov_b32_e32 v36, v0
+; VI-NEXT: v_mov_b32_e32 v55, v0
+; VI-NEXT: v_or_b32_sdwa v0, v54, v0 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; VI-NEXT: v_or_b32_sdwa v22, v0, v1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_0 src1_sel:DWORD
+; VI-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:592 ; 4-byte Folded Reload
+; VI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:580 ; 4-byte Folded Reload
+; VI-NEXT: s_waitcnt vmcnt(1)
+; VI-NEXT: v_mov_b32_e32 v35, v0
; VI-NEXT: v_or_b32_sdwa v0, v63, v0 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; VI-NEXT: s_waitcnt vmcnt(0)
+; VI-NEXT: v_mov_b32_e32 v41, v1
+; VI-NEXT: v_or_b32_sdwa v1, v46, v1 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; VI-NEXT: v_mov_b32_e32 v63, v27
+; VI-NEXT: v_mov_b32_e32 v46, v57
; VI-NEXT: v_or_b32_sdwa v23, v0, v1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_0 src1_sel:DWORD
; VI-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:600 ; 4-byte Folded Reload
; VI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:596 ; 4-byte Folded Reload
; VI-NEXT: s_waitcnt vmcnt(1)
-; VI-NEXT: v_mov_b32_e32 v56, v0
+; VI-NEXT: v_mov_b32_e32 v36, v0
; VI-NEXT: v_or_b32_sdwa v0, v61, v0 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
; VI-NEXT: s_waitcnt vmcnt(0)
-; VI-NEXT: v_mov_b32_e32 v58, v1
-; VI-NEXT: v_or_b32_sdwa v1, v60, v1 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
-; VI-NEXT: v_mov_b32_e32 v61, v60
-; VI-NEXT: v_mov_b32_e32 v60, v59
+; VI-NEXT: v_mov_b32_e32 v56, v1
+; VI-NEXT: v_or_b32_sdwa v1, v57, v1 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; VI-NEXT: v_mov_b32_e32 v61, v59
; VI-NEXT: v_or_b32_sdwa v24, v0, v1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_0 src1_sel:DWORD
; VI-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:608 ; 4-byte Folded Reload
; VI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:612 ; 4-byte Folded Reload
@@ -187293,55 +187340,53 @@ define inreg <64 x half> @bitcast_v128i8_to_v64f16_scalar(<128 x i8> inreg %a, i
; VI-NEXT: v_or_b32_sdwa v1, v40, v1 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
; VI-NEXT: v_or_b32_sdwa v25, v0, v1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_0 src1_sel:DWORD
; VI-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:604 ; 4-byte Folded Reload
-; VI-NEXT: buffer_load_dword v59, off, s[0:3], s32 offset:556 ; 4-byte Folded Reload
-; VI-NEXT: v_or_b32_sdwa v1, v45, v26 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; VI-NEXT: buffer_load_dword v59, off, s[0:3], s32 offset:552 ; 4-byte Folded Reload
+; VI-NEXT: v_or_b32_sdwa v1, v44, v26 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
; VI-NEXT: s_waitcnt vmcnt(1)
-; VI-NEXT: v_mov_b32_e32 v50, v0
+; VI-NEXT: v_mov_b32_e32 v58, v0
; VI-NEXT: v_or_b32_sdwa v0, v42, v0 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
; VI-NEXT: v_or_b32_sdwa v26, v0, v1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_0 src1_sel:DWORD
; VI-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:616 ; 4-byte Folded Reload
; VI-NEXT: v_or_b32_sdwa v1, v62, v27 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
; VI-NEXT: s_waitcnt vmcnt(0)
-; VI-NEXT: v_mov_b32_e32 v52, v0
-; VI-NEXT: v_or_b32_sdwa v0, v44, v0 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; VI-NEXT: v_mov_b32_e32 v43, v0
+; VI-NEXT: v_or_b32_sdwa v0, v45, v0 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
; VI-NEXT: v_or_b32_sdwa v27, v0, v1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_0 src1_sel:DWORD
; VI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:620 ; 4-byte Folded Reload
-; VI-NEXT: v_or_b32_sdwa v0, v29, v28 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; VI-NEXT: v_or_b32_sdwa v0, v59, v28 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
; VI-NEXT: s_waitcnt vmcnt(0)
-; VI-NEXT: v_mov_b32_e32 v46, v1
-; VI-NEXT: v_or_b32_sdwa v1, v59, v1 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; VI-NEXT: v_mov_b32_e32 v60, v1
+; VI-NEXT: v_or_b32_sdwa v1, v29, v1 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
; VI-NEXT: v_or_b32_sdwa v28, v0, v1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_0 src1_sel:DWORD
-; VI-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:800 ; 4-byte Folded Reload
-; VI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:808 ; 4-byte Folded Reload
+; VI-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:808 ; 4-byte Folded Reload
+; VI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:812 ; 4-byte Folded Reload
; VI-NEXT: s_waitcnt vmcnt(1)
-; VI-NEXT: v_mov_b32_e32 v63, v0
-; VI-NEXT: v_or_b32_sdwa v0, v30, v0 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; VI-NEXT: v_mov_b32_e32 v54, v0
+; VI-NEXT: v_or_b32_sdwa v0, v31, v0 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
; VI-NEXT: s_waitcnt vmcnt(0)
; VI-NEXT: v_mov_b32_e32 v47, v1
-; VI-NEXT: v_or_b32_sdwa v1, v31, v1 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; VI-NEXT: v_or_b32_sdwa v1, v30, v1 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
; VI-NEXT: v_or_b32_sdwa v29, v0, v1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_0 src1_sel:DWORD
-; VI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:804 ; 4-byte Folded Reload
-; VI-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:524 ; 4-byte Folded Reload
-; VI-NEXT: s_waitcnt vmcnt(1)
-; VI-NEXT: v_mov_b32_e32 v57, v1
-; VI-NEXT: s_waitcnt vmcnt(0)
-; VI-NEXT: v_or_b32_sdwa v0, v0, v1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
-; VI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:528 ; 4-byte Folded Reload
+; VI-NEXT: buffer_load_dword v57, off, s[0:3], s32 offset:544 ; 4-byte Folded Reload
+; VI-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:516 ; 4-byte Folded Reload
+; VI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:520 ; 4-byte Folded Reload
; VI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:532 ; 4-byte Folded Reload
+; VI-NEXT: s_waitcnt vmcnt(2)
+; VI-NEXT: v_or_b32_sdwa v0, v0, v57 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
; VI-NEXT: s_waitcnt vmcnt(0)
; VI-NEXT: v_or_b32_sdwa v1, v1, v2 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
; VI-NEXT: v_or_b32_sdwa v30, v0, v1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_0 src1_sel:DWORD
-; VI-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:544 ; 4-byte Folded Reload
+; VI-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:524 ; 4-byte Folded Reload
; VI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:536 ; 4-byte Folded Reload
; VI-NEXT: s_waitcnt vmcnt(0)
-; VI-NEXT: v_or_b32_sdwa v0, v1, v0 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
-; VI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:548 ; 4-byte Folded Reload
-; VI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:540 ; 4-byte Folded Reload
+; VI-NEXT: v_or_b32_sdwa v0, v0, v1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; VI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:540 ; 4-byte Folded Reload
+; VI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:528 ; 4-byte Folded Reload
; VI-NEXT: s_waitcnt vmcnt(0)
; VI-NEXT: v_or_b32_sdwa v1, v2, v1 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
; VI-NEXT: v_or_b32_sdwa v31, v0, v1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_0 src1_sel:DWORD
; VI-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:460 ; 4-byte Folded Reload
-; VI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:508 ; 4-byte Folded Reload
+; VI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:504 ; 4-byte Folded Reload
; VI-NEXT: s_waitcnt vmcnt(0)
; VI-NEXT: v_or_b32_sdwa v0, v0, v1 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
; VI-NEXT: v_or_b32_e32 v3, s4, v0
@@ -187373,12 +187418,10 @@ define inreg <64 x half> @bitcast_v128i8_to_v64f16_scalar(<128 x i8> inreg %a, i
; VI-NEXT: v_mov_b32_e32 v2, s6
; VI-NEXT: s_cbranch_execnz .LBB93_3
; VI-NEXT: .LBB93_2: ; %cmp.true
-; VI-NEXT: s_waitcnt vmcnt(0)
-; VI-NEXT: v_add_u32_e32 v3, vcc, 3, v59
-; VI-NEXT: v_or_b32_sdwa v29, v46, v3 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:BYTE_0
-; VI-NEXT: buffer_load_dword v3, off, s[0:3], s32 offset:788 ; 4-byte Folded Reload
-; VI-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:540 ; 4-byte Folded Reload
-; VI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:548 ; 4-byte Folded Reload
+; VI-NEXT: buffer_load_dword v3, off, s[0:3], s32 offset:796 ; 4-byte Folded Reload
+; VI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:540 ; 4-byte Folded Reload
+; VI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:532 ; 4-byte Folded Reload
+; VI-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:528 ; 4-byte Folded Reload
; VI-NEXT: s_add_i32 s28, s28, 3
; VI-NEXT: s_and_b32 s4, s28, 0xff
; VI-NEXT: s_lshl_b32 s5, s29, 8
@@ -187397,165 +187440,147 @@ define inreg <64 x half> @bitcast_v128i8_to_v64f16_scalar(<128 x i8> inreg %a, i
; VI-NEXT: s_lshl_b32 s9, s19, 8
; VI-NEXT: s_add_i32 s16, s16, 3
; VI-NEXT: s_lshl_b32 s10, s17, 8
-; VI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:532 ; 4-byte Folded Reload
-; VI-NEXT: s_waitcnt vmcnt(3)
+; VI-NEXT: buffer_load_dword v4, off, s[0:3], s32 offset:512 ; 4-byte Folded Reload
+; VI-NEXT: s_waitcnt vmcnt(4)
; VI-NEXT: v_add_u32_e32 v3, vcc, 3, v3
-; VI-NEXT: v_or_b32_sdwa v26, v53, v3 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:BYTE_0
+; VI-NEXT: v_or_b32_sdwa v28, v60, v3 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:BYTE_0
+; VI-NEXT: v_add_u32_e32 v3, vcc, 3, v59
+; VI-NEXT: v_or_b32_sdwa v25, v52, v3 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:BYTE_0
; VI-NEXT: v_add_u32_e32 v3, vcc, 3, v62
-; VI-NEXT: v_or_b32_sdwa v28, v43, v3 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:BYTE_0
-; VI-NEXT: v_add_u32_e32 v3, vcc, 3, v44
-; VI-NEXT: v_or_b32_sdwa v53, v52, v3 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:BYTE_0
+; VI-NEXT: v_or_b32_sdwa v27, v63, v3 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:BYTE_0
; VI-NEXT: v_add_u32_e32 v3, vcc, 3, v45
-; VI-NEXT: v_or_b32_sdwa v27, v55, v3 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:BYTE_0
+; VI-NEXT: v_or_b32_sdwa v52, v43, v3 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:BYTE_0
+; VI-NEXT: v_add_u32_e32 v3, vcc, 3, v44
+; VI-NEXT: v_or_b32_sdwa v26, v50, v3 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:BYTE_0
; VI-NEXT: v_add_u32_e32 v3, vcc, 3, v42
-; VI-NEXT: v_or_b32_sdwa v52, v50, v3 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:BYTE_0
+; VI-NEXT: v_or_b32_sdwa v63, v58, v3 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:BYTE_0
; VI-NEXT: v_add_u32_e32 v3, vcc, 3, v40
-; VI-NEXT: v_or_b32_sdwa v25, v48, v3 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:BYTE_0
-; VI-NEXT: v_add_u32_e32 v3, vcc, 3, v60
-; VI-NEXT: v_or_b32_sdwa v59, v38, v3 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:BYTE_0
+; VI-NEXT: v_or_b32_sdwa v43, v48, v3 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:BYTE_0
; VI-NEXT: v_add_u32_e32 v3, vcc, 3, v61
-; VI-NEXT: v_or_b32_sdwa v24, v58, v3 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:BYTE_0
-; VI-NEXT: buffer_load_dword v3, off, s[0:3], s32 offset:784 ; 4-byte Folded Reload
+; VI-NEXT: v_or_b32_sdwa v59, v38, v3 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:BYTE_0
+; VI-NEXT: v_add_u32_e32 v3, vcc, 3, v46
+; VI-NEXT: v_or_b32_sdwa v24, v56, v3 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:BYTE_0
+; VI-NEXT: buffer_load_dword v3, off, s[0:3], s32 offset:792 ; 4-byte Folded Reload
; VI-NEXT: s_waitcnt vmcnt(0)
; VI-NEXT: v_add_u32_e32 v3, vcc, 3, v3
-; VI-NEXT: v_or_b32_sdwa v48, v56, v3 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:BYTE_0
-; VI-NEXT: buffer_load_dword v3, off, s[0:3], s32 offset:780 ; 4-byte Folded Reload
+; VI-NEXT: v_or_b32_sdwa v48, v36, v3 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:BYTE_0
+; VI-NEXT: buffer_load_dword v3, off, s[0:3], s32 offset:788 ; 4-byte Folded Reload
; VI-NEXT: v_add_u32_e32 v48, vcc, 0x300, v48
; VI-NEXT: v_or_b32_sdwa v24, v24, v48 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:WORD_0
; VI-NEXT: v_add_u32_e32 v24, vcc, 0x3000000, v24
; VI-NEXT: s_waitcnt vmcnt(0)
; VI-NEXT: v_add_u32_e32 v3, vcc, 3, v3
; VI-NEXT: v_or_b32_sdwa v23, v41, v3 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:BYTE_0
-; VI-NEXT: buffer_load_dword v3, off, s[0:3], s32 offset:776 ; 4-byte Folded Reload
+; VI-NEXT: buffer_load_dword v3, off, s[0:3], s32 offset:784 ; 4-byte Folded Reload
; VI-NEXT: s_waitcnt vmcnt(0)
; VI-NEXT: v_add_u32_e32 v3, vcc, 3, v3
-; VI-NEXT: v_or_b32_sdwa v38, v36, v3 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:BYTE_0
-; VI-NEXT: buffer_load_dword v3, off, s[0:3], s32 offset:772 ; 4-byte Folded Reload
+; VI-NEXT: v_or_b32_sdwa v38, v35, v3 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:BYTE_0
+; VI-NEXT: buffer_load_dword v3, off, s[0:3], s32 offset:780 ; 4-byte Folded Reload
; VI-NEXT: v_add_u32_e32 v38, vcc, 0x300, v38
; VI-NEXT: v_or_b32_sdwa v23, v23, v38 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:WORD_0
; VI-NEXT: v_add_u32_e32 v23, vcc, 0x3000000, v23
; VI-NEXT: s_waitcnt vmcnt(0)
; VI-NEXT: v_add_u32_e32 v3, vcc, 3, v3
-; VI-NEXT: v_or_b32_sdwa v22, v54, v3 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:BYTE_0
-; VI-NEXT: buffer_load_dword v3, off, s[0:3], s32 offset:768 ; 4-byte Folded Reload
+; VI-NEXT: v_or_b32_sdwa v22, v34, v3 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:BYTE_0
+; VI-NEXT: buffer_load_dword v3, off, s[0:3], s32 offset:776 ; 4-byte Folded Reload
; VI-NEXT: s_waitcnt vmcnt(0)
; VI-NEXT: v_add_u32_e32 v3, vcc, 3, v3
-; VI-NEXT: v_or_b32_sdwa v50, v33, v3 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:BYTE_0
-; VI-NEXT: buffer_load_dword v3, off, s[0:3], s32 offset:764 ; 4-byte Folded Reload
-; VI-NEXT: v_add_u32_e32 v50, vcc, 0x300, v50
+; VI-NEXT: v_or_b32_sdwa v36, v55, v3 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:BYTE_0
+; VI-NEXT: buffer_load_dword v3, off, s[0:3], s32 offset:772 ; 4-byte Folded Reload
+; VI-NEXT: v_add_u32_e32 v50, vcc, 0x300, v36
; VI-NEXT: v_or_b32_sdwa v22, v22, v50 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:WORD_0
; VI-NEXT: v_add_u32_e32 v22, vcc, 0x3000000, v22
; VI-NEXT: s_waitcnt vmcnt(0)
; VI-NEXT: v_add_u32_e32 v3, vcc, 3, v3
-; VI-NEXT: v_or_b32_sdwa v21, v35, v3 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:BYTE_0
-; VI-NEXT: buffer_load_dword v3, off, s[0:3], s32 offset:760 ; 4-byte Folded Reload
+; VI-NEXT: v_or_b32_sdwa v21, v53, v3 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:BYTE_0
+; VI-NEXT: buffer_load_dword v3, off, s[0:3], s32 offset:768 ; 4-byte Folded Reload
; VI-NEXT: s_waitcnt vmcnt(0)
; VI-NEXT: v_add_u32_e32 v3, vcc, 3, v3
-; VI-NEXT: v_or_b32_sdwa v54, v51, v3 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:BYTE_0
-; VI-NEXT: buffer_load_dword v3, off, s[0:3], s32 offset:756 ; 4-byte Folded Reload
+; VI-NEXT: v_or_b32_sdwa v53, v51, v3 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:BYTE_0
+; VI-NEXT: buffer_load_dword v3, off, s[0:3], s32 offset:764 ; 4-byte Folded Reload
; VI-NEXT: s_waitcnt vmcnt(0)
; VI-NEXT: v_add_u32_e32 v3, vcc, 3, v3
; VI-NEXT: v_or_b32_sdwa v20, v49, v3 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:BYTE_0
-; VI-NEXT: buffer_load_dword v3, off, s[0:3], s32 offset:752 ; 4-byte Folded Reload
+; VI-NEXT: buffer_load_dword v3, off, s[0:3], s32 offset:760 ; 4-byte Folded Reload
; VI-NEXT: s_waitcnt vmcnt(0)
; VI-NEXT: v_add_u32_e32 v3, vcc, 3, v3
; VI-NEXT: v_or_b32_sdwa v49, v39, v3 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:BYTE_0
-; VI-NEXT: buffer_load_dword v3, off, s[0:3], s32 offset:748 ; 4-byte Folded Reload
-; VI-NEXT: v_add_u32_e32 v0, vcc, 3, v0
-; VI-NEXT: v_or_b32_sdwa v0, v1, v0 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:BYTE_0
-; VI-NEXT: buffer_store_dword v0, off, s[0:3], s32 offset:540 ; 4-byte Folded Spill
-; VI-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:536 ; 4-byte Folded Reload
-; VI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:544 ; 4-byte Folded Reload
+; VI-NEXT: buffer_load_dword v3, off, s[0:3], s32 offset:756 ; 4-byte Folded Reload
; VI-NEXT: v_add_u32_e32 v49, vcc, 0x300, v49
; VI-NEXT: v_or_b32_sdwa v20, v20, v49 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:WORD_0
; VI-NEXT: v_add_u32_e32 v20, vcc, 0x3000000, v20
-; VI-NEXT: s_waitcnt vmcnt(3)
+; VI-NEXT: s_waitcnt vmcnt(0)
; VI-NEXT: v_add_u32_e32 v3, vcc, 3, v3
; VI-NEXT: v_or_b32_sdwa v19, v37, v3 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:BYTE_0
-; VI-NEXT: buffer_load_dword v3, off, s[0:3], s32 offset:744 ; 4-byte Folded Reload
-; VI-NEXT: s_waitcnt vmcnt(2)
-; VI-NEXT: v_add_u32_e32 v0, vcc, 3, v0
-; VI-NEXT: s_waitcnt vmcnt(1)
-; VI-NEXT: v_or_b32_sdwa v0, v1, v0 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:BYTE_0
-; VI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:528 ; 4-byte Folded Reload
-; VI-NEXT: s_waitcnt vmcnt(1)
+; VI-NEXT: buffer_load_dword v3, off, s[0:3], s32 offset:752 ; 4-byte Folded Reload
+; VI-NEXT: s_waitcnt vmcnt(0)
; VI-NEXT: v_add_u32_e32 v3, vcc, 3, v3
-; VI-NEXT: v_or_b32_sdwa v37, v34, v3 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:BYTE_0
-; VI-NEXT: buffer_load_dword v3, off, s[0:3], s32 offset:740 ; 4-byte Folded Reload
+; VI-NEXT: v_or_b32_sdwa v37, v33, v3 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:BYTE_0
+; VI-NEXT: buffer_load_dword v3, off, s[0:3], s32 offset:748 ; 4-byte Folded Reload
; VI-NEXT: v_add_u32_e32 v37, vcc, 0x300, v37
-; VI-NEXT: s_waitcnt vmcnt(1)
-; VI-NEXT: v_add_u32_e32 v1, vcc, 3, v1
-; VI-NEXT: v_or_b32_sdwa v31, v2, v1 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:BYTE_0
-; VI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:524 ; 4-byte Folded Reload
-; VI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:796 ; 4-byte Folded Reload
; VI-NEXT: v_or_b32_sdwa v19, v19, v37 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:WORD_0
; VI-NEXT: v_add_u32_e32 v19, vcc, 0x3000000, v19
-; VI-NEXT: s_waitcnt vmcnt(2)
+; VI-NEXT: s_waitcnt vmcnt(0)
; VI-NEXT: v_add_u32_e32 v3, vcc, 3, v3
; VI-NEXT: v_or_b32_sdwa v18, v32, v3 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:BYTE_0
-; VI-NEXT: buffer_load_dword v3, off, s[0:3], s32 offset:736 ; 4-byte Folded Reload
-; VI-NEXT: buffer_load_dword v4, off, s[0:3], s32 offset:520 ; 4-byte Folded Reload
+; VI-NEXT: buffer_load_dword v3, off, s[0:3], s32 offset:744 ; 4-byte Folded Reload
; VI-NEXT: buffer_load_dword v32, off, s[0:3], s32 offset:460 ; 4-byte Folded Reload
-; VI-NEXT: s_waitcnt vmcnt(4)
-; VI-NEXT: v_add_u32_e32 v1, vcc, 3, v1
-; VI-NEXT: v_or_b32_sdwa v1, v57, v1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:BYTE_0
-; VI-NEXT: s_waitcnt vmcnt(2)
-; VI-NEXT: v_add_u32_e32 v3, vcc, 3, v3
; VI-NEXT: s_waitcnt vmcnt(1)
-; VI-NEXT: v_or_b32_sdwa v57, v4, v3 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:BYTE_0
-; VI-NEXT: buffer_load_dword v3, off, s[0:3], s32 offset:732 ; 4-byte Folded Reload
-; VI-NEXT: buffer_load_dword v4, off, s[0:3], s32 offset:516 ; 4-byte Folded Reload
-; VI-NEXT: v_add_u32_e32 v35, vcc, 0x300, v57
-; VI-NEXT: v_or_b32_sdwa v18, v18, v35 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:WORD_0
+; VI-NEXT: v_add_u32_e32 v3, vcc, 3, v3
+; VI-NEXT: v_or_b32_sdwa v58, v4, v3 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:BYTE_0
+; VI-NEXT: buffer_load_dword v3, off, s[0:3], s32 offset:740 ; 4-byte Folded Reload
+; VI-NEXT: buffer_load_dword v4, off, s[0:3], s32 offset:724 ; 4-byte Folded Reload
+; VI-NEXT: v_add_u32_e32 v36, vcc, 0x300, v58
+; VI-NEXT: v_or_b32_sdwa v18, v18, v36 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:WORD_0
; VI-NEXT: v_add_u32_e32 v18, vcc, 0x3000000, v18
; VI-NEXT: s_waitcnt vmcnt(1)
; VI-NEXT: v_add_u32_e32 v3, vcc, 3, v3
; VI-NEXT: s_waitcnt vmcnt(0)
; VI-NEXT: v_or_b32_sdwa v16, v4, v3 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:BYTE_0
-; VI-NEXT: buffer_load_dword v3, off, s[0:3], s32 offset:728 ; 4-byte Folded Reload
-; VI-NEXT: buffer_load_dword v4, off, s[0:3], s32 offset:512 ; 4-byte Folded Reload
+; VI-NEXT: buffer_load_dword v3, off, s[0:3], s32 offset:736 ; 4-byte Folded Reload
+; VI-NEXT: buffer_load_dword v4, off, s[0:3], s32 offset:704 ; 4-byte Folded Reload
; VI-NEXT: s_waitcnt vmcnt(1)
; VI-NEXT: v_add_u32_e32 v3, vcc, 3, v3
; VI-NEXT: s_waitcnt vmcnt(0)
; VI-NEXT: v_or_b32_sdwa v10, v4, v3 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:BYTE_0
-; VI-NEXT: buffer_load_dword v3, off, s[0:3], s32 offset:724 ; 4-byte Folded Reload
-; VI-NEXT: buffer_load_dword v4, off, s[0:3], s32 offset:680 ; 4-byte Folded Reload
+; VI-NEXT: buffer_load_dword v3, off, s[0:3], s32 offset:732 ; 4-byte Folded Reload
+; VI-NEXT: buffer_load_dword v4, off, s[0:3], s32 offset:508 ; 4-byte Folded Reload
; VI-NEXT: s_waitcnt vmcnt(1)
; VI-NEXT: v_add_u32_e32 v3, vcc, 3, v3
; VI-NEXT: s_waitcnt vmcnt(0)
; VI-NEXT: v_or_b32_sdwa v17, v4, v3 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:BYTE_0
+; VI-NEXT: buffer_load_dword v3, off, s[0:3], s32 offset:728 ; 4-byte Folded Reload
+; VI-NEXT: buffer_load_dword v4, off, s[0:3], s32 offset:680 ; 4-byte Folded Reload
+; VI-NEXT: s_waitcnt vmcnt(1)
+; VI-NEXT: v_add_u32_e32 v3, vcc, 3, v3
+; VI-NEXT: s_waitcnt vmcnt(0)
+; VI-NEXT: v_or_b32_sdwa v11, v4, v3 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:BYTE_0
; VI-NEXT: buffer_load_dword v3, off, s[0:3], s32 offset:720 ; 4-byte Folded Reload
; VI-NEXT: buffer_load_dword v4, off, s[0:3], s32 offset:676 ; 4-byte Folded Reload
; VI-NEXT: s_waitcnt vmcnt(1)
; VI-NEXT: v_add_u32_e32 v3, vcc, 3, v3
; VI-NEXT: s_waitcnt vmcnt(0)
-; VI-NEXT: v_or_b32_sdwa v11, v4, v3 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:BYTE_0
+; VI-NEXT: v_or_b32_sdwa v15, v4, v3 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:BYTE_0
; VI-NEXT: buffer_load_dword v3, off, s[0:3], s32 offset:716 ; 4-byte Folded Reload
; VI-NEXT: buffer_load_dword v4, off, s[0:3], s32 offset:672 ; 4-byte Folded Reload
; VI-NEXT: s_waitcnt vmcnt(1)
; VI-NEXT: v_add_u32_e32 v3, vcc, 3, v3
; VI-NEXT: s_waitcnt vmcnt(0)
-; VI-NEXT: v_or_b32_sdwa v15, v4, v3 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:BYTE_0
+; VI-NEXT: v_or_b32_sdwa v56, v4, v3 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:BYTE_0
; VI-NEXT: buffer_load_dword v3, off, s[0:3], s32 offset:712 ; 4-byte Folded Reload
; VI-NEXT: buffer_load_dword v4, off, s[0:3], s32 offset:668 ; 4-byte Folded Reload
; VI-NEXT: s_waitcnt vmcnt(1)
; VI-NEXT: v_add_u32_e32 v3, vcc, 3, v3
; VI-NEXT: s_waitcnt vmcnt(0)
-; VI-NEXT: v_or_b32_sdwa v56, v4, v3 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:BYTE_0
+; VI-NEXT: v_or_b32_sdwa v14, v4, v3 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:BYTE_0
; VI-NEXT: buffer_load_dword v3, off, s[0:3], s32 offset:708 ; 4-byte Folded Reload
; VI-NEXT: buffer_load_dword v4, off, s[0:3], s32 offset:664 ; 4-byte Folded Reload
; VI-NEXT: s_waitcnt vmcnt(1)
; VI-NEXT: v_add_u32_e32 v3, vcc, 3, v3
; VI-NEXT: s_waitcnt vmcnt(0)
-; VI-NEXT: v_or_b32_sdwa v14, v4, v3 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:BYTE_0
-; VI-NEXT: buffer_load_dword v3, off, s[0:3], s32 offset:704 ; 4-byte Folded Reload
-; VI-NEXT: buffer_load_dword v4, off, s[0:3], s32 offset:660 ; 4-byte Folded Reload
-; VI-NEXT: s_waitcnt vmcnt(1)
-; VI-NEXT: v_add_u32_e32 v3, vcc, 3, v3
-; VI-NEXT: s_waitcnt vmcnt(0)
; VI-NEXT: v_or_b32_sdwa v34, v4, v3 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:BYTE_0
; VI-NEXT: buffer_load_dword v3, off, s[0:3], s32 offset:700 ; 4-byte Folded Reload
-; VI-NEXT: buffer_load_dword v4, off, s[0:3], s32 offset:504 ; 4-byte Folded Reload
+; VI-NEXT: buffer_load_dword v4, off, s[0:3], s32 offset:500 ; 4-byte Folded Reload
; VI-NEXT: v_add_u32_e32 v34, vcc, 0x300, v34
; VI-NEXT: v_or_b32_sdwa v14, v14, v34 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:WORD_0
; VI-NEXT: v_add_u32_e32 v14, vcc, 0x3000000, v14
@@ -187564,67 +187589,78 @@ define inreg <64 x half> @bitcast_v128i8_to_v64f16_scalar(<128 x i8> inreg %a, i
; VI-NEXT: s_waitcnt vmcnt(0)
; VI-NEXT: v_or_b32_sdwa v13, v4, v3 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:BYTE_0
; VI-NEXT: buffer_load_dword v3, off, s[0:3], s32 offset:696 ; 4-byte Folded Reload
-; VI-NEXT: buffer_load_dword v4, off, s[0:3], s32 offset:500 ; 4-byte Folded Reload
-; VI-NEXT: s_waitcnt vmcnt(1)
+; VI-NEXT: buffer_load_dword v4, off, s[0:3], s32 offset:496 ; 4-byte Folded Reload
+; VI-NEXT: v_add_u32_e32 v0, vcc, 3, v0
+; VI-NEXT: v_or_b32_sdwa v31, v1, v0 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:BYTE_0
+; VI-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:524 ; 4-byte Folded Reload
+; VI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:536 ; 4-byte Folded Reload
+; VI-NEXT: s_waitcnt vmcnt(3)
; VI-NEXT: v_add_u32_e32 v3, vcc, 3, v3
-; VI-NEXT: s_waitcnt vmcnt(0)
-; VI-NEXT: v_or_b32_sdwa v36, v4, v3 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:BYTE_0
+; VI-NEXT: s_waitcnt vmcnt(2)
+; VI-NEXT: v_or_b32_sdwa v35, v4, v3 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:BYTE_0
; VI-NEXT: buffer_load_dword v3, off, s[0:3], s32 offset:692 ; 4-byte Folded Reload
-; VI-NEXT: buffer_load_dword v4, off, s[0:3], s32 offset:496 ; 4-byte Folded Reload
-; VI-NEXT: v_add_u32_e32 v36, vcc, 0x300, v36
-; VI-NEXT: v_or_b32_sdwa v13, v13, v36 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:WORD_0
-; VI-NEXT: v_add_u32_e32 v36, vcc, 0x300, v26
-; VI-NEXT: v_add_u32_e32 v26, vcc, 0x300, v52
-; VI-NEXT: v_or_b32_sdwa v26, v27, v26 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:WORD_0
-; VI-NEXT: v_add_u32_e32 v52, vcc, 0x300, v54
-; VI-NEXT: v_or_b32_sdwa v21, v21, v52 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:WORD_0
+; VI-NEXT: buffer_load_dword v4, off, s[0:3], s32 offset:492 ; 4-byte Folded Reload
+; VI-NEXT: s_waitcnt vmcnt(3)
+; VI-NEXT: v_add_u32_e32 v0, vcc, 3, v0
+; VI-NEXT: s_waitcnt vmcnt(2)
+; VI-NEXT: v_or_b32_sdwa v0, v1, v0 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:BYTE_0
+; VI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:520 ; 4-byte Folded Reload
+; VI-NEXT: v_add_u32_e32 v35, vcc, 0x300, v35
+; VI-NEXT: v_or_b32_sdwa v13, v13, v35 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:WORD_0
+; VI-NEXT: v_add_u32_e32 v35, vcc, 0x300, v25
+; VI-NEXT: v_add_u32_e32 v25, vcc, 0x300, v59
+; VI-NEXT: v_or_b32_sdwa v25, v43, v25 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:WORD_0
+; VI-NEXT: v_or_b32_sdwa v28, v28, v35 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:WORD_0
; VI-NEXT: v_add_u32_e32 v13, vcc, 0x3000000, v13
-; VI-NEXT: v_add_u32_e32 v21, vcc, 0x3000000, v21
-; VI-NEXT: v_add_u32_e32 v26, vcc, 0x3000000, v26
-; VI-NEXT: s_waitcnt vmcnt(1)
+; VI-NEXT: v_add_u32_e32 v25, vcc, 0x3000000, v25
+; VI-NEXT: v_add_u32_e32 v28, vcc, 0x3000000, v28
+; VI-NEXT: s_waitcnt vmcnt(2)
; VI-NEXT: v_add_u32_e32 v3, vcc, 3, v3
-; VI-NEXT: s_waitcnt vmcnt(0)
+; VI-NEXT: s_waitcnt vmcnt(1)
; VI-NEXT: v_or_b32_sdwa v12, v4, v3 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:BYTE_0
; VI-NEXT: buffer_load_dword v3, off, s[0:3], s32 offset:688 ; 4-byte Folded Reload
-; VI-NEXT: buffer_load_dword v4, off, s[0:3], s32 offset:492 ; 4-byte Folded Reload
-; VI-NEXT: s_waitcnt vmcnt(1)
+; VI-NEXT: buffer_load_dword v4, off, s[0:3], s32 offset:488 ; 4-byte Folded Reload
+; VI-NEXT: s_waitcnt vmcnt(2)
+; VI-NEXT: v_add_u32_e32 v1, vcc, 3, v1
+; VI-NEXT: v_or_b32_sdwa v30, v2, v1 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:BYTE_0
+; VI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:516 ; 4-byte Folded Reload
+; VI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:804 ; 4-byte Folded Reload
+; VI-NEXT: s_waitcnt vmcnt(3)
; VI-NEXT: v_add_u32_e32 v3, vcc, 3, v3
-; VI-NEXT: s_waitcnt vmcnt(0)
+; VI-NEXT: s_waitcnt vmcnt(2)
; VI-NEXT: v_or_b32_sdwa v51, v4, v3 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:BYTE_0
; VI-NEXT: buffer_load_dword v3, off, s[0:3], s32 offset:684 ; 4-byte Folded Reload
-; VI-NEXT: buffer_load_dword v4, off, s[0:3], s32 offset:488 ; 4-byte Folded Reload
+; VI-NEXT: buffer_load_dword v4, off, s[0:3], s32 offset:484 ; 4-byte Folded Reload
+; VI-NEXT: s_waitcnt vmcnt(3)
+; VI-NEXT: v_add_u32_e32 v1, vcc, 3, v1
+; VI-NEXT: v_or_b32_sdwa v1, v57, v1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:BYTE_0
; VI-NEXT: v_add_u32_e32 v51, vcc, 0x300, v51
; VI-NEXT: v_or_b32_sdwa v12, v12, v51 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:WORD_0
-; VI-NEXT: v_add_u32_e32 v51, vcc, 0x300, v59
-; VI-NEXT: v_or_b32_sdwa v25, v25, v51 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:WORD_0
+; VI-NEXT: v_add_u32_e32 v33, vcc, 0x300, v1
+; VI-NEXT: v_add_u32_e32 v51, vcc, 0x300, v63
+; VI-NEXT: v_or_b32_sdwa v26, v26, v51 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:WORD_0
+; VI-NEXT: v_or_b32_sdwa v30, v30, v33 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:WORD_0
; VI-NEXT: v_add_u32_e32 v12, vcc, 0x3000000, v12
-; VI-NEXT: v_add_u32_e32 v25, vcc, 0x3000000, v25
+; VI-NEXT: v_add_u32_e32 v26, vcc, 0x3000000, v26
+; VI-NEXT: v_add_u32_e32 v30, vcc, 0x3000000, v30
; VI-NEXT: s_waitcnt vmcnt(1)
; VI-NEXT: v_add_u32_e32 v3, vcc, 3, v3
; VI-NEXT: s_waitcnt vmcnt(0)
-; VI-NEXT: v_or_b32_sdwa v33, v4, v3 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:BYTE_0
+; VI-NEXT: v_or_b32_sdwa v57, v4, v3 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:BYTE_0
; VI-NEXT: buffer_load_dword v3, off, s[0:3], s32 offset:456 ; 4-byte Folded Reload
-; VI-NEXT: buffer_load_dword v4, off, s[0:3], s32 offset:484 ; 4-byte Folded Reload
+; VI-NEXT: buffer_load_dword v4, off, s[0:3], s32 offset:480 ; 4-byte Folded Reload
; VI-NEXT: s_waitcnt vmcnt(1)
; VI-NEXT: v_add_u32_e32 v3, vcc, 3, v3
; VI-NEXT: s_waitcnt vmcnt(0)
; VI-NEXT: v_or_b32_sdwa v40, v4, v3 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:BYTE_0
; VI-NEXT: buffer_load_dword v3, off, s[0:3], s32 offset:452 ; 4-byte Folded Reload
-; VI-NEXT: buffer_load_dword v4, off, s[0:3], s32 offset:480 ; 4-byte Folded Reload
-; VI-NEXT: v_add_u32_e32 v2, vcc, 3, v2
-; VI-NEXT: v_or_b32_sdwa v30, v47, v2 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:BYTE_0
-; VI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:792 ; 4-byte Folded Reload
-; VI-NEXT: buffer_load_dword v47, off, s[0:3], s32 offset:464 ; 4-byte Folded Reload
-; VI-NEXT: s_waitcnt vmcnt(3)
+; VI-NEXT: buffer_load_dword v4, off, s[0:3], s32 offset:476 ; 4-byte Folded Reload
+; VI-NEXT: s_waitcnt vmcnt(1)
; VI-NEXT: v_add_u32_e32 v3, vcc, 3, v3
-; VI-NEXT: s_waitcnt vmcnt(2)
+; VI-NEXT: s_waitcnt vmcnt(0)
; VI-NEXT: v_or_b32_sdwa v39, v4, v3 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:BYTE_0
; VI-NEXT: buffer_load_dword v3, off, s[0:3], s32 offset:448 ; 4-byte Folded Reload
-; VI-NEXT: buffer_load_dword v4, off, s[0:3], s32 offset:476 ; 4-byte Folded Reload
-; VI-NEXT: s_waitcnt vmcnt(3)
-; VI-NEXT: v_add_u32_e32 v2, vcc, 3, v2
-; VI-NEXT: v_or_b32_sdwa v2, v63, v2 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:BYTE_0
-; VI-NEXT: v_add_u32_e32 v34, vcc, 0x300, v2
+; VI-NEXT: buffer_load_dword v4, off, s[0:3], s32 offset:660 ; 4-byte Folded Reload
; VI-NEXT: s_waitcnt vmcnt(1)
; VI-NEXT: v_add_u32_e32 v3, vcc, 3, v3
; VI-NEXT: s_waitcnt vmcnt(0)
@@ -187648,15 +187684,14 @@ define inreg <64 x half> @bitcast_v128i8_to_v64f16_scalar(<128 x i8> inreg %a, i
; VI-NEXT: v_add_u32_e32 v41, vcc, 0x300, v10
; VI-NEXT: v_add_u32_e32 v10, vcc, 0x300, v55
; VI-NEXT: v_or_b32_sdwa v10, v39, v10 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:WORD_0
-; VI-NEXT: v_add_u32_e32 v39, vcc, 0x300, v53
-; VI-NEXT: v_or_b32_sdwa v27, v28, v39 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:WORD_0
-; VI-NEXT: v_or_b32_sdwa v28, v29, v36 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:WORD_0
-; VI-NEXT: v_or_b32_sdwa v29, v30, v34 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:WORD_0
+; VI-NEXT: v_add_u32_e32 v39, vcc, 0x300, v52
+; VI-NEXT: v_add_u32_e32 v52, vcc, 0x300, v53
+; VI-NEXT: v_or_b32_sdwa v21, v21, v52 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:WORD_0
+; VI-NEXT: v_or_b32_sdwa v27, v27, v39 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:WORD_0
; VI-NEXT: v_add_u32_e32 v9, vcc, 0x3000000, v9
; VI-NEXT: v_add_u32_e32 v10, vcc, 0x3000000, v10
+; VI-NEXT: v_add_u32_e32 v21, vcc, 0x3000000, v21
; VI-NEXT: v_add_u32_e32 v27, vcc, 0x3000000, v27
-; VI-NEXT: v_add_u32_e32 v28, vcc, 0x3000000, v28
-; VI-NEXT: v_add_u32_e32 v29, vcc, 0x3000000, v29
; VI-NEXT: s_waitcnt vmcnt(1)
; VI-NEXT: v_add_u32_e32 v3, vcc, 3, v3
; VI-NEXT: s_waitcnt vmcnt(0)
@@ -187672,18 +187707,14 @@ define inreg <64 x half> @bitcast_v128i8_to_v64f16_scalar(<128 x i8> inreg %a, i
; VI-NEXT: v_add_u32_e32 v42, vcc, 0x300, v42
; VI-NEXT: v_or_b32_sdwa v8, v8, v42 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:WORD_0
; VI-NEXT: v_add_u32_e32 v42, vcc, 0x300, v11
-; VI-NEXT: v_add_u32_e32 v11, vcc, 0x300, v40
-; VI-NEXT: v_or_b32_sdwa v11, v33, v11 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:WORD_0
-; VI-NEXT: v_add_u32_e32 v33, vcc, 0x300, v1
-; VI-NEXT: v_or_b32_sdwa v30, v31, v33 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:WORD_0
-; VI-NEXT: buffer_load_dword v31, off, s[0:3], s32 offset:540 ; 4-byte Folded Reload
; VI-NEXT: v_or_b32_sdwa v17, v17, v42 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:WORD_0
+; VI-NEXT: v_add_u32_e32 v11, vcc, 0x300, v40
+; VI-NEXT: v_or_b32_sdwa v11, v57, v11 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:WORD_0
; VI-NEXT: v_add_u32_e32 v8, vcc, 0x3000000, v8
; VI-NEXT: v_add_u32_e32 v11, vcc, 0x3000000, v11
-; VI-NEXT: v_add_u32_e32 v30, vcc, 0x3000000, v30
-; VI-NEXT: s_waitcnt vmcnt(2)
-; VI-NEXT: v_add_u32_e32 v3, vcc, 3, v3
; VI-NEXT: s_waitcnt vmcnt(1)
+; VI-NEXT: v_add_u32_e32 v3, vcc, 3, v3
+; VI-NEXT: s_waitcnt vmcnt(0)
; VI-NEXT: v_or_b32_sdwa v7, v4, v3 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:BYTE_0
; VI-NEXT: buffer_load_dword v3, off, s[0:3], s32 offset:424 ; 4-byte Folded Reload
; VI-NEXT: buffer_load_dword v4, off, s[0:3], s32 offset:636 ; 4-byte Folded Reload
@@ -187723,19 +187754,29 @@ define inreg <64 x half> @bitcast_v128i8_to_v64f16_scalar(<128 x i8> inreg %a, i
; VI-NEXT: v_or_b32_sdwa v46, v4, v3 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:BYTE_0
; VI-NEXT: buffer_load_dword v3, off, s[0:3], s32 offset:404 ; 4-byte Folded Reload
; VI-NEXT: buffer_load_dword v4, off, s[0:3], s32 offset:468 ; 4-byte Folded Reload
+; VI-NEXT: v_add_u32_e32 v2, vcc, 3, v2
+; VI-NEXT: v_or_b32_sdwa v29, v47, v2 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:BYTE_0
+; VI-NEXT: buffer_load_dword v47, off, s[0:3], s32 offset:464 ; 4-byte Folded Reload
+; VI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:800 ; 4-byte Folded Reload
; VI-NEXT: v_add_u32_e32 v46, vcc, 0x300, v46
; VI-NEXT: v_or_b32_sdwa v5, v5, v46 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:WORD_0
; VI-NEXT: v_add_u32_e32 v5, vcc, 0x3000000, v5
-; VI-NEXT: s_waitcnt vmcnt(1)
+; VI-NEXT: s_waitcnt vmcnt(3)
; VI-NEXT: v_add_u32_e32 v3, vcc, 3, v3
-; VI-NEXT: s_waitcnt vmcnt(0)
+; VI-NEXT: s_waitcnt vmcnt(2)
; VI-NEXT: v_or_b32_sdwa v3, v4, v3 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:BYTE_0
; VI-NEXT: buffer_load_dword v4, off, s[0:3], s32 offset:400 ; 4-byte Folded Reload
+; VI-NEXT: s_waitcnt vmcnt(1)
+; VI-NEXT: v_add_u32_e32 v2, vcc, 3, v2
+; VI-NEXT: v_or_b32_sdwa v2, v54, v2 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:BYTE_0
+; VI-NEXT: v_add_u32_e32 v34, vcc, 0x300, v2
+; VI-NEXT: v_or_b32_sdwa v29, v29, v34 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:WORD_0
+; VI-NEXT: v_add_u32_e32 v29, vcc, 0x3000000, v29
; VI-NEXT: s_waitcnt vmcnt(0)
; VI-NEXT: v_add_u32_e32 v4, vcc, 3, v4
; VI-NEXT: v_or_b32_sdwa v4, v47, v4 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:BYTE_0
; VI-NEXT: v_add_u32_e32 v47, vcc, 3, v32
-; VI-NEXT: buffer_load_dword v32, off, s[0:3], s32 offset:508 ; 4-byte Folded Reload
+; VI-NEXT: buffer_load_dword v32, off, s[0:3], s32 offset:504 ; 4-byte Folded Reload
; VI-NEXT: v_add_u32_e32 v4, vcc, 0x300, v4
; VI-NEXT: v_or_b32_sdwa v4, v3, v4 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:WORD_0
; VI-NEXT: v_add_u32_e32 v4, vcc, 0x3000000, v4
@@ -187802,35 +187843,38 @@ define inreg <64 x half> @bitcast_v128i8_to_v64f16_scalar(<128 x i8> inreg %a, i
; VI-NEXT: s_waitcnt vmcnt(0)
; VI-NEXT: s_setpc_b64 s[30:31]
; VI-NEXT: .LBB93_4:
-; VI-NEXT: buffer_load_dword v46, off, s[0:3], s32 offset:620 ; 4-byte Folded Reload
-; VI-NEXT: buffer_load_dword v52, off, s[0:3], s32 offset:616 ; 4-byte Folded Reload
-; VI-NEXT: buffer_load_dword v50, off, s[0:3], s32 offset:604 ; 4-byte Folded Reload
+; VI-NEXT: buffer_load_dword v60, off, s[0:3], s32 offset:620 ; 4-byte Folded Reload
+; VI-NEXT: buffer_load_dword v43, off, s[0:3], s32 offset:616 ; 4-byte Folded Reload
+; VI-NEXT: buffer_load_dword v58, off, s[0:3], s32 offset:604 ; 4-byte Folded Reload
; VI-NEXT: buffer_load_dword v48, off, s[0:3], s32 offset:612 ; 4-byte Folded Reload
; VI-NEXT: buffer_load_dword v38, off, s[0:3], s32 offset:608 ; 4-byte Folded Reload
-; VI-NEXT: buffer_load_dword v58, off, s[0:3], s32 offset:596 ; 4-byte Folded Reload
-; VI-NEXT: buffer_load_dword v56, off, s[0:3], s32 offset:600 ; 4-byte Folded Reload
-; VI-NEXT: buffer_load_dword v36, off, s[0:3], s32 offset:568 ; 4-byte Folded Reload
-; VI-NEXT: buffer_load_dword v33, off, s[0:3], s32 offset:580 ; 4-byte Folded Reload
-; VI-NEXT: buffer_load_dword v35, off, s[0:3], s32 offset:572 ; 4-byte Folded Reload
-; VI-NEXT: buffer_load_dword v51, off, s[0:3], s32 offset:576 ; 4-byte Folded Reload
-; VI-NEXT: buffer_load_dword v49, off, s[0:3], s32 offset:584 ; 4-byte Folded Reload
-; VI-NEXT: buffer_load_dword v39, off, s[0:3], s32 offset:588 ; 4-byte Folded Reload
-; VI-NEXT: buffer_load_dword v37, off, s[0:3], s32 offset:592 ; 4-byte Folded Reload
-; VI-NEXT: buffer_load_dword v34, off, s[0:3], s32 offset:564 ; 4-byte Folded Reload
-; VI-NEXT: buffer_load_dword v32, off, s[0:3], s32 offset:560 ; 4-byte Folded Reload
-; VI-NEXT: v_mov_b32_e32 v61, v60
-; VI-NEXT: v_mov_b32_e32 v60, v59
+; VI-NEXT: buffer_load_dword v56, off, s[0:3], s32 offset:596 ; 4-byte Folded Reload
+; VI-NEXT: buffer_load_dword v36, off, s[0:3], s32 offset:600 ; 4-byte Folded Reload
+; VI-NEXT: buffer_load_dword v41, off, s[0:3], s32 offset:580 ; 4-byte Folded Reload
+; VI-NEXT: buffer_load_dword v35, off, s[0:3], s32 offset:592 ; 4-byte Folded Reload
+; VI-NEXT: buffer_load_dword v55, off, s[0:3], s32 offset:568 ; 4-byte Folded Reload
+; VI-NEXT: buffer_load_dword v53, off, s[0:3], s32 offset:560 ; 4-byte Folded Reload
+; VI-NEXT: buffer_load_dword v51, off, s[0:3], s32 offset:564 ; 4-byte Folded Reload
+; VI-NEXT: buffer_load_dword v49, off, s[0:3], s32 offset:572 ; 4-byte Folded Reload
+; VI-NEXT: buffer_load_dword v39, off, s[0:3], s32 offset:576 ; 4-byte Folded Reload
+; VI-NEXT: buffer_load_dword v37, off, s[0:3], s32 offset:584 ; 4-byte Folded Reload
+; VI-NEXT: buffer_load_dword v33, off, s[0:3], s32 offset:588 ; 4-byte Folded Reload
+; VI-NEXT: buffer_load_dword v32, off, s[0:3], s32 offset:556 ; 4-byte Folded Reload
+; VI-NEXT: v_mov_b32_e32 v61, v59
+; VI-NEXT: v_mov_b32_e32 v46, v57
+; VI-NEXT: buffer_load_dword v57, off, s[0:3], s32 offset:544 ; 4-byte Folded Reload
+; VI-NEXT: v_mov_b32_e32 v40, v42
+; VI-NEXT: v_mov_b32_e32 v42, v44
+; VI-NEXT: v_mov_b32_e32 v44, v45
; VI-NEXT: v_mov_b32_e32 v45, v62
-; VI-NEXT: buffer_load_dword v62, off, s[0:3], s32 offset:552 ; 4-byte Folded Reload
-; VI-NEXT: buffer_load_dword v59, off, s[0:3], s32 offset:556 ; 4-byte Folded Reload
-; VI-NEXT: v_mov_b32_e32 v57, v5
+; VI-NEXT: buffer_load_dword v62, off, s[0:3], s32 offset:548 ; 4-byte Folded Reload
+; VI-NEXT: buffer_load_dword v59, off, s[0:3], s32 offset:552 ; 4-byte Folded Reload
; VI-NEXT: v_mov_b32_e32 v47, v4
-; VI-NEXT: v_mov_b32_e32 v63, v3
-; VI-NEXT: v_mov_b32_e32 v53, v28
-; VI-NEXT: v_mov_b32_e32 v43, v27
-; VI-NEXT: v_mov_b32_e32 v55, v26
-; VI-NEXT: v_mov_b32_e32 v41, v24
-; VI-NEXT: v_mov_b32_e32 v54, v22
+; VI-NEXT: v_mov_b32_e32 v54, v3
+; VI-NEXT: v_mov_b32_e32 v52, v28
+; VI-NEXT: v_mov_b32_e32 v63, v27
+; VI-NEXT: v_mov_b32_e32 v50, v26
+; VI-NEXT: v_mov_b32_e32 v34, v24
; VI-NEXT: ; implicit-def: $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7_vgpr8_vgpr9_vgpr10_vgpr11_vgpr12_vgpr13_vgpr14_vgpr15
; VI-NEXT: ; implicit-def: $vgpr16_vgpr17_vgpr18_vgpr19_vgpr20_vgpr21_vgpr22_vgpr23_vgpr24_vgpr25_vgpr26_vgpr27_vgpr28_vgpr29_vgpr30_vgpr31
; VI-NEXT: s_branch .LBB93_2
@@ -187892,18 +187936,18 @@ define inreg <64 x half> @bitcast_v128i8_to_v64f16_scalar(<128 x i8> inreg %a, i
; GFX9-NEXT: v_lshlrev_b32_e32 v1, 8, v29
; GFX9-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:400 ; 4-byte Folded Spill
; GFX9-NEXT: v_lshlrev_b32_e32 v14, 8, v3
-; GFX9-NEXT: v_lshlrev_b32_e32 v46, 8, v5
+; GFX9-NEXT: v_lshlrev_b32_e32 v18, 8, v5
; GFX9-NEXT: v_lshlrev_b32_e32 v22, 8, v7
; GFX9-NEXT: v_lshlrev_b32_e32 v24, 8, v9
; GFX9-NEXT: v_lshlrev_b32_e32 v26, 8, v11
; GFX9-NEXT: v_lshlrev_b32_e32 v20, 8, v13
-; GFX9-NEXT: v_lshlrev_b32_e32 v28, 8, v15
-; GFX9-NEXT: v_lshlrev_b32_e32 v18, 8, v17
-; GFX9-NEXT: v_lshlrev_b32_e32 v17, 8, v25
-; GFX9-NEXT: v_lshlrev_b32_e32 v12, 8, v27
+; GFX9-NEXT: v_lshlrev_b32_e32 v12, 8, v15
+; GFX9-NEXT: v_lshlrev_b32_e32 v17, 8, v17
; GFX9-NEXT: v_lshlrev_b32_e32 v19, 8, v19
; GFX9-NEXT: v_lshlrev_b32_e32 v21, 8, v21
; GFX9-NEXT: v_lshlrev_b32_e32 v23, 8, v23
+; GFX9-NEXT: v_lshlrev_b32_e32 v25, 8, v25
+; GFX9-NEXT: v_lshlrev_b32_e32 v27, 8, v27
; GFX9-NEXT: s_waitcnt vmcnt(24)
; GFX9-NEXT: v_cmp_ne_u32_e32 vcc, 0, v43
; GFX9-NEXT: s_waitcnt vmcnt(23)
@@ -187932,10 +187976,10 @@ define inreg <64 x half> @bitcast_v128i8_to_v64f16_scalar(<128 x i8> inreg %a, i
; GFX9-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:608 ; 4-byte Folded Spill
; GFX9-NEXT: s_waitcnt vmcnt(23)
; GFX9-NEXT: v_lshlrev_b32_e32 v1, 8, v52
-; GFX9-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:600 ; 4-byte Folded Spill
+; GFX9-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:604 ; 4-byte Folded Spill
; GFX9-NEXT: s_waitcnt vmcnt(23)
; GFX9-NEXT: v_lshlrev_b32_e32 v1, 8, v51
-; GFX9-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:604 ; 4-byte Folded Spill
+; GFX9-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:600 ; 4-byte Folded Spill
; GFX9-NEXT: s_waitcnt vmcnt(23)
; GFX9-NEXT: v_lshlrev_b32_e32 v1, 8, v50
; GFX9-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:496 ; 4-byte Folded Spill
@@ -187947,7 +187991,7 @@ define inreg <64 x half> @bitcast_v128i8_to_v64f16_scalar(<128 x i8> inreg %a, i
; GFX9-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:596 ; 4-byte Folded Spill
; GFX9-NEXT: s_waitcnt vmcnt(23)
; GFX9-NEXT: v_lshlrev_b32_e32 v1, 8, v39
-; GFX9-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:592 ; 4-byte Folded Spill
+; GFX9-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:588 ; 4-byte Folded Spill
; GFX9-NEXT: s_waitcnt vmcnt(23)
; GFX9-NEXT: v_lshlrev_b32_e32 v1, 8, v30
; GFX9-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:448 ; 4-byte Folded Spill
@@ -187995,7 +188039,7 @@ define inreg <64 x half> @bitcast_v128i8_to_v64f16_scalar(<128 x i8> inreg %a, i
; GFX9-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:584 ; 4-byte Folded Spill
; GFX9-NEXT: s_waitcnt vmcnt(6)
; GFX9-NEXT: v_lshlrev_b32_e32 v1, 8, v3
-; GFX9-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:588 ; 4-byte Folded Spill
+; GFX9-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:592 ; 4-byte Folded Spill
; GFX9-NEXT: s_waitcnt vmcnt(6)
; GFX9-NEXT: v_lshlrev_b32_e32 v1, 8, v13
; GFX9-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:536 ; 4-byte Folded Spill
@@ -188022,23 +188066,23 @@ define inreg <64 x half> @bitcast_v128i8_to_v64f16_scalar(<128 x i8> inreg %a, i
; GFX9-NEXT: v_lshlrev_b32_e32 v11, 8, v11
; GFX9-NEXT: s_waitcnt vmcnt(6)
; GFX9-NEXT: v_lshlrev_b32_e32 v1, 8, v1
-; GFX9-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:556 ; 4-byte Folded Spill
+; GFX9-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:572 ; 4-byte Folded Spill
; GFX9-NEXT: s_waitcnt vmcnt(6)
; GFX9-NEXT: v_lshlrev_b32_e32 v1, 8, v15
-; GFX9-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:572 ; 4-byte Folded Spill
+; GFX9-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:568 ; 4-byte Folded Spill
; GFX9-NEXT: s_waitcnt vmcnt(6)
; GFX9-NEXT: v_lshlrev_b32_e32 v1, 8, v3
; GFX9-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:552 ; 4-byte Folded Spill
; GFX9-NEXT: s_waitcnt vmcnt(6)
; GFX9-NEXT: v_lshlrev_b32_e32 v1, 8, v13
-; GFX9-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:568 ; 4-byte Folded Spill
+; GFX9-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:564 ; 4-byte Folded Spill
; GFX9-NEXT: s_waitcnt vmcnt(6)
; GFX9-NEXT: v_lshlrev_b32_e32 v1, 8, v5
; GFX9-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:548 ; 4-byte Folded Spill
; GFX9-NEXT: s_waitcnt vmcnt(6)
; GFX9-NEXT: v_lshlrev_b32_e32 v1, 8, v9
-; GFX9-NEXT: buffer_store_dword v11, off, s[0:3], s32 offset:564 ; 4-byte Folded Spill
-; GFX9-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:560 ; 4-byte Folded Spill
+; GFX9-NEXT: buffer_store_dword v11, off, s[0:3], s32 offset:560 ; 4-byte Folded Spill
+; GFX9-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:556 ; 4-byte Folded Spill
; GFX9-NEXT: buffer_load_ushort v3, off, s[0:3], s32 offset:312
; GFX9-NEXT: s_nop 0
; GFX9-NEXT: buffer_load_ushort v1, off, s[0:3], s32 offset:320
@@ -188051,48 +188095,49 @@ define inreg <64 x half> @bitcast_v128i8_to_v64f16_scalar(<128 x i8> inreg %a, i
; GFX9-NEXT: s_waitcnt vmcnt(15)
; GFX9-NEXT: v_lshlrev_b32_e32 v7, 8, v7
; GFX9-NEXT: s_waitcnt vmcnt(7)
-; GFX9-NEXT: v_lshlrev_b32_e32 v3, 8, v3
+; GFX9-NEXT: v_lshlrev_b32_e32 v9, 8, v3
; GFX9-NEXT: s_waitcnt vmcnt(6)
-; GFX9-NEXT: v_lshlrev_b32_e32 v9, 8, v1
+; GFX9-NEXT: v_lshlrev_b32_e32 v1, 8, v1
+; GFX9-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:520 ; 4-byte Folded Spill
; GFX9-NEXT: buffer_load_ushort v52, off, s[0:3], s32 offset:44
; GFX9-NEXT: buffer_load_ushort v50, off, s[0:3], s32 offset:52
; GFX9-NEXT: buffer_load_ushort v49, off, s[0:3], s32 offset:60
; GFX9-NEXT: buffer_load_ushort v15, off, s[0:3], s32 offset:68
; GFX9-NEXT: buffer_load_ushort v48, off, s[0:3], s32 offset:76
; GFX9-NEXT: buffer_load_ushort v39, off, s[0:3], s32 offset:84
-; GFX9-NEXT: buffer_load_ushort v25, off, s[0:3], s32 offset:92
+; GFX9-NEXT: buffer_load_ushort v40, off, s[0:3], s32 offset:92
; GFX9-NEXT: buffer_load_ushort v34, off, s[0:3], s32 offset:100
; GFX9-NEXT: buffer_load_ushort v33, off, s[0:3], s32 offset:108
-; GFX9-NEXT: buffer_load_ushort v45, off, s[0:3], s32 offset:116
+; GFX9-NEXT: buffer_load_ushort v46, off, s[0:3], s32 offset:116
; GFX9-NEXT: buffer_load_ushort v43, off, s[0:3], s32 offset:124
; GFX9-NEXT: buffer_load_ushort v38, off, s[0:3], s32 offset:132
-; GFX9-NEXT: buffer_load_ushort v56, off, s[0:3], s32 offset:140
+; GFX9-NEXT: buffer_load_ushort v61, off, s[0:3], s32 offset:140
; GFX9-NEXT: buffer_load_ushort v47, off, s[0:3], s32 offset:148
-; GFX9-NEXT: buffer_load_ushort v61, off, s[0:3], s32 offset:156
+; GFX9-NEXT: buffer_load_ushort v62, off, s[0:3], s32 offset:156
; GFX9-NEXT: buffer_load_ushort v1, off, s[0:3], s32 offset:164
-; GFX9-NEXT: s_waitcnt vmcnt(21)
-; GFX9-NEXT: v_lshlrev_b32_e32 v5, 8, v5
+; GFX9-NEXT: s_waitcnt vmcnt(22)
+; GFX9-NEXT: v_lshlrev_b32_e32 v3, 8, v5
; GFX9-NEXT: s_waitcnt vmcnt(0)
; GFX9-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:576 ; 4-byte Folded Spill
-; GFX9-NEXT: buffer_load_ushort v31, off, s[0:3], s32 offset:172
-; GFX9-NEXT: buffer_load_ushort v32, off, s[0:3], s32 offset:180
-; GFX9-NEXT: buffer_load_ushort v51, off, s[0:3], s32 offset:188
-; GFX9-NEXT: buffer_load_ushort v30, off, s[0:3], s32 offset:196
-; GFX9-NEXT: buffer_load_ushort v44, off, s[0:3], s32 offset:204
+; GFX9-NEXT: buffer_load_ushort v57, off, s[0:3], s32 offset:172
+; GFX9-NEXT: buffer_load_ushort v31, off, s[0:3], s32 offset:180
+; GFX9-NEXT: buffer_load_ushort v36, off, s[0:3], s32 offset:188
+; GFX9-NEXT: buffer_load_ushort v37, off, s[0:3], s32 offset:196
+; GFX9-NEXT: buffer_load_ushort v30, off, s[0:3], s32 offset:204
; GFX9-NEXT: buffer_load_ushort v35, off, s[0:3], s32 offset:212
-; GFX9-NEXT: buffer_load_ushort v36, off, s[0:3], s32 offset:220
-; GFX9-NEXT: buffer_load_ushort v37, off, s[0:3], s32 offset:228
+; GFX9-NEXT: buffer_load_ushort v28, off, s[0:3], s32 offset:220
+; GFX9-NEXT: buffer_load_ushort v51, off, s[0:3], s32 offset:228
; GFX9-NEXT: buffer_load_ushort v41, off, s[0:3], s32 offset:236
; GFX9-NEXT: buffer_load_ushort v54, off, s[0:3], s32 offset:244
-; GFX9-NEXT: buffer_load_ushort v29, off, s[0:3], s32 offset:252
-; GFX9-NEXT: buffer_load_ushort v27, off, s[0:3], s32 offset:260
+; GFX9-NEXT: buffer_load_ushort v44, off, s[0:3], s32 offset:252
+; GFX9-NEXT: buffer_load_ushort v45, off, s[0:3], s32 offset:260
; GFX9-NEXT: buffer_load_ushort v60, off, s[0:3], s32 offset:268
-; GFX9-NEXT: buffer_load_ushort v57, off, s[0:3], s32 offset:276
+; GFX9-NEXT: buffer_load_ushort v29, off, s[0:3], s32 offset:276
; GFX9-NEXT: buffer_load_ushort v59, off, s[0:3], s32 offset:284
; GFX9-NEXT: buffer_load_ushort v63, off, s[0:3], s32 offset:292
-; GFX9-NEXT: buffer_load_ushort v62, off, s[0:3], s32 offset:300
-; GFX9-NEXT: buffer_load_ushort v40, off, s[0:3], s32 offset:308
-; GFX9-NEXT: buffer_load_ushort v58, off, s[0:3], s32 offset:316
+; GFX9-NEXT: buffer_load_ushort v58, off, s[0:3], s32 offset:300
+; GFX9-NEXT: buffer_load_ushort v32, off, s[0:3], s32 offset:308
+; GFX9-NEXT: buffer_load_ushort v56, off, s[0:3], s32 offset:316
; GFX9-NEXT: buffer_load_ushort v1, off, s[0:3], s32 offset:324
; GFX9-NEXT: s_nop 0
; GFX9-NEXT: buffer_store_dword v10, off, s[0:3], s32 offset:620 ; 4-byte Folded Spill
@@ -188103,55 +188148,54 @@ define inreg <64 x half> @bitcast_v128i8_to_v64f16_scalar(<128 x i8> inreg %a, i
; GFX9-NEXT: buffer_store_dword v0, off, s[0:3], s32 offset:640 ; 4-byte Folded Spill
; GFX9-NEXT: buffer_store_dword v2, off, s[0:3], s32 offset:644 ; 4-byte Folded Spill
; GFX9-NEXT: buffer_store_dword v4, off, s[0:3], s32 offset:648 ; 4-byte Folded Spill
-; GFX9-NEXT: buffer_store_dword v12, off, s[0:3], s32 offset:652 ; 4-byte Folded Spill
+; GFX9-NEXT: buffer_store_dword v27, off, s[0:3], s32 offset:652 ; 4-byte Folded Spill
; GFX9-NEXT: buffer_store_dword v7, off, s[0:3], s32 offset:508 ; 4-byte Folded Spill
; GFX9-NEXT: s_waitcnt vmcnt(28)
-; GFX9-NEXT: buffer_store_dword v32, off, s[0:3], s32 offset:656 ; 4-byte Folded Spill
-; GFX9-NEXT: buffer_store_dword v31, off, s[0:3], s32 offset:660 ; 4-byte Folded Spill
-; GFX9-NEXT: buffer_store_dword v17, off, s[0:3], s32 offset:664 ; 4-byte Folded Spill
+; GFX9-NEXT: buffer_store_dword v31, off, s[0:3], s32 offset:656 ; 4-byte Folded Spill
+; GFX9-NEXT: buffer_store_dword v57, off, s[0:3], s32 offset:660 ; 4-byte Folded Spill
+; GFX9-NEXT: buffer_store_dword v25, off, s[0:3], s32 offset:664 ; 4-byte Folded Spill
; GFX9-NEXT: buffer_store_dword v23, off, s[0:3], s32 offset:668 ; 4-byte Folded Spill
; GFX9-NEXT: s_waitcnt vmcnt(30)
-; GFX9-NEXT: buffer_store_dword v30, off, s[0:3], s32 offset:672 ; 4-byte Folded Spill
-; GFX9-NEXT: buffer_store_dword v51, off, s[0:3], s32 offset:676 ; 4-byte Folded Spill
+; GFX9-NEXT: buffer_store_dword v37, off, s[0:3], s32 offset:672 ; 4-byte Folded Spill
+; GFX9-NEXT: buffer_store_dword v36, off, s[0:3], s32 offset:676 ; 4-byte Folded Spill
; GFX9-NEXT: buffer_store_dword v21, off, s[0:3], s32 offset:680 ; 4-byte Folded Spill
; GFX9-NEXT: buffer_store_dword v19, off, s[0:3], s32 offset:684 ; 4-byte Folded Spill
-; GFX9-NEXT: buffer_store_dword v3, off, s[0:3], s32 offset:512 ; 4-byte Folded Spill
+; GFX9-NEXT: buffer_store_dword v9, off, s[0:3], s32 offset:512 ; 4-byte Folded Spill
; GFX9-NEXT: s_waitcnt vmcnt(33)
; GFX9-NEXT: buffer_store_dword v35, off, s[0:3], s32 offset:688 ; 4-byte Folded Spill
-; GFX9-NEXT: buffer_store_dword v44, off, s[0:3], s32 offset:692 ; 4-byte Folded Spill
-; GFX9-NEXT: buffer_store_dword v18, off, s[0:3], s32 offset:696 ; 4-byte Folded Spill
-; GFX9-NEXT: buffer_store_dword v28, off, s[0:3], s32 offset:700 ; 4-byte Folded Spill
-; GFX9-NEXT: buffer_store_dword v5, off, s[0:3], s32 offset:516 ; 4-byte Folded Spill
+; GFX9-NEXT: buffer_store_dword v30, off, s[0:3], s32 offset:692 ; 4-byte Folded Spill
+; GFX9-NEXT: buffer_store_dword v17, off, s[0:3], s32 offset:696 ; 4-byte Folded Spill
+; GFX9-NEXT: buffer_store_dword v12, off, s[0:3], s32 offset:700 ; 4-byte Folded Spill
+; GFX9-NEXT: buffer_store_dword v3, off, s[0:3], s32 offset:516 ; 4-byte Folded Spill
; GFX9-NEXT: s_waitcnt vmcnt(36)
-; GFX9-NEXT: buffer_store_dword v37, off, s[0:3], s32 offset:704 ; 4-byte Folded Spill
-; GFX9-NEXT: buffer_store_dword v9, off, s[0:3], s32 offset:520 ; 4-byte Folded Spill
-; GFX9-NEXT: buffer_store_dword v36, off, s[0:3], s32 offset:708 ; 4-byte Folded Spill
+; GFX9-NEXT: buffer_store_dword v51, off, s[0:3], s32 offset:704 ; 4-byte Folded Spill
+; GFX9-NEXT: buffer_store_dword v28, off, s[0:3], s32 offset:708 ; 4-byte Folded Spill
; GFX9-NEXT: buffer_store_dword v20, off, s[0:3], s32 offset:712 ; 4-byte Folded Spill
; GFX9-NEXT: buffer_store_dword v26, off, s[0:3], s32 offset:716 ; 4-byte Folded Spill
-; GFX9-NEXT: s_waitcnt vmcnt(39)
+; GFX9-NEXT: s_waitcnt vmcnt(38)
; GFX9-NEXT: buffer_store_dword v54, off, s[0:3], s32 offset:720 ; 4-byte Folded Spill
; GFX9-NEXT: buffer_store_dword v41, off, s[0:3], s32 offset:724 ; 4-byte Folded Spill
; GFX9-NEXT: buffer_store_dword v24, off, s[0:3], s32 offset:728 ; 4-byte Folded Spill
; GFX9-NEXT: buffer_store_dword v22, off, s[0:3], s32 offset:732 ; 4-byte Folded Spill
-; GFX9-NEXT: s_waitcnt vmcnt(41)
-; GFX9-NEXT: buffer_store_dword v27, off, s[0:3], s32 offset:736 ; 4-byte Folded Spill
-; GFX9-NEXT: buffer_store_dword v29, off, s[0:3], s32 offset:740 ; 4-byte Folded Spill
-; GFX9-NEXT: s_waitcnt vmcnt(41)
-; GFX9-NEXT: buffer_store_dword v57, off, s[0:3], s32 offset:744 ; 4-byte Folded Spill
+; GFX9-NEXT: s_waitcnt vmcnt(40)
+; GFX9-NEXT: buffer_store_dword v45, off, s[0:3], s32 offset:736 ; 4-byte Folded Spill
+; GFX9-NEXT: buffer_store_dword v44, off, s[0:3], s32 offset:740 ; 4-byte Folded Spill
+; GFX9-NEXT: s_waitcnt vmcnt(40)
+; GFX9-NEXT: buffer_store_dword v29, off, s[0:3], s32 offset:744 ; 4-byte Folded Spill
; GFX9-NEXT: buffer_store_dword v60, off, s[0:3], s32 offset:748 ; 4-byte Folded Spill
-; GFX9-NEXT: s_waitcnt vmcnt(41)
+; GFX9-NEXT: s_waitcnt vmcnt(40)
; GFX9-NEXT: buffer_store_dword v63, off, s[0:3], s32 offset:752 ; 4-byte Folded Spill
; GFX9-NEXT: buffer_store_dword v59, off, s[0:3], s32 offset:756 ; 4-byte Folded Spill
-; GFX9-NEXT: s_waitcnt vmcnt(41)
-; GFX9-NEXT: buffer_store_dword v40, off, s[0:3], s32 offset:760 ; 4-byte Folded Spill
-; GFX9-NEXT: buffer_store_dword v62, off, s[0:3], s32 offset:764 ; 4-byte Folded Spill
-; GFX9-NEXT: s_waitcnt vmcnt(41)
+; GFX9-NEXT: s_waitcnt vmcnt(40)
+; GFX9-NEXT: buffer_store_dword v32, off, s[0:3], s32 offset:760 ; 4-byte Folded Spill
+; GFX9-NEXT: buffer_store_dword v58, off, s[0:3], s32 offset:764 ; 4-byte Folded Spill
+; GFX9-NEXT: s_waitcnt vmcnt(40)
; GFX9-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:524 ; 4-byte Folded Spill
-; GFX9-NEXT: buffer_store_dword v58, off, s[0:3], s32 offset:768 ; 4-byte Folded Spill
+; GFX9-NEXT: buffer_store_dword v56, off, s[0:3], s32 offset:768 ; 4-byte Folded Spill
; GFX9-NEXT: buffer_store_dword v39, off, s[0:3], s32 offset:772 ; 4-byte Folded Spill
; GFX9-NEXT: buffer_store_dword v48, off, s[0:3], s32 offset:776 ; 4-byte Folded Spill
; GFX9-NEXT: buffer_store_dword v34, off, s[0:3], s32 offset:780 ; 4-byte Folded Spill
-; GFX9-NEXT: buffer_store_dword v25, off, s[0:3], s32 offset:784 ; 4-byte Folded Spill
+; GFX9-NEXT: buffer_store_dword v40, off, s[0:3], s32 offset:784 ; 4-byte Folded Spill
; GFX9-NEXT: buffer_store_dword v15, off, s[0:3], s32 offset:788 ; 4-byte Folded Spill
; GFX9-NEXT: buffer_store_dword v49, off, s[0:3], s32 offset:792 ; 4-byte Folded Spill
; GFX9-NEXT: buffer_store_dword v50, off, s[0:3], s32 offset:796 ; 4-byte Folded Spill
@@ -188161,7 +188205,7 @@ define inreg <64 x half> @bitcast_v128i8_to_v64f16_scalar(<128 x i8> inreg %a, i
; GFX9-NEXT: buffer_store_dword v53, off, s[0:3], s32 offset:812 ; 4-byte Folded Spill
; GFX9-NEXT: buffer_store_dword v55, off, s[0:3], s32 offset:816 ; 4-byte Folded Spill
; GFX9-NEXT: buffer_store_dword v42, off, s[0:3], s32 offset:820 ; 4-byte Folded Spill
-; GFX9-NEXT: buffer_store_dword v46, off, s[0:3], s32 offset:824 ; 4-byte Folded Spill
+; GFX9-NEXT: buffer_store_dword v18, off, s[0:3], s32 offset:824 ; 4-byte Folded Spill
; GFX9-NEXT: buffer_store_dword v14, off, s[0:3], s32 offset:828 ; 4-byte Folded Spill
; GFX9-NEXT: buffer_store_dword v11, off, s[0:3], s32 offset:832 ; 4-byte Folded Spill
; GFX9-NEXT: s_cbranch_scc0 .LBB93_2
@@ -188174,7 +188218,7 @@ define inreg <64 x half> @bitcast_v128i8_to_v64f16_scalar(<128 x i8> inreg %a, i
; GFX9-NEXT: v_or_b32_sdwa v1, v2, v14 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
; GFX9-NEXT: v_and_b32_e32 v1, 0xffff, v1
; GFX9-NEXT: v_or_b32_sdwa v2, v0, v16 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
-; GFX9-NEXT: v_or_b32_sdwa v4, v4, v46 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; GFX9-NEXT: v_or_b32_sdwa v4, v4, v18 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
; GFX9-NEXT: v_or_b32_sdwa v0, v6, v22 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
; GFX9-NEXT: v_lshl_or_b32 v4, v4, 16, v1
; GFX9-NEXT: v_and_b32_e32 v0, 0xffff, v0
@@ -188211,10 +188255,10 @@ define inreg <64 x half> @bitcast_v128i8_to_v64f16_scalar(<128 x i8> inreg %a, i
; GFX9-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:452 ; 4-byte Folded Reload
; GFX9-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:456 ; 4-byte Folded Reload
; GFX9-NEXT: s_waitcnt vmcnt(1)
-; GFX9-NEXT: v_or_b32_sdwa v0, v0, v28 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; GFX9-NEXT: v_or_b32_sdwa v0, v0, v12 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
; GFX9-NEXT: v_and_b32_e32 v0, 0xffff, v0
; GFX9-NEXT: s_waitcnt vmcnt(0)
-; GFX9-NEXT: v_or_b32_sdwa v1, v1, v18 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; GFX9-NEXT: v_or_b32_sdwa v1, v1, v17 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
; GFX9-NEXT: v_lshl_or_b32 v7, v1, 16, v0
; GFX9-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:436 ; 4-byte Folded Reload
; GFX9-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:440 ; 4-byte Folded Reload
@@ -188230,13 +188274,13 @@ define inreg <64 x half> @bitcast_v128i8_to_v64f16_scalar(<128 x i8> inreg %a, i
; GFX9-NEXT: v_or_b32_sdwa v0, v0, v23 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
; GFX9-NEXT: v_and_b32_e32 v0, 0xffff, v0
; GFX9-NEXT: s_waitcnt vmcnt(0)
-; GFX9-NEXT: v_or_b32_sdwa v1, v1, v17 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; GFX9-NEXT: v_or_b32_sdwa v1, v1, v25 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
; GFX9-NEXT: v_lshl_or_b32 v9, v1, 16, v0
; GFX9-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:420 ; 4-byte Folded Reload
; GFX9-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:424 ; 4-byte Folded Reload
; GFX9-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:400 ; 4-byte Folded Reload
; GFX9-NEXT: s_waitcnt vmcnt(2)
-; GFX9-NEXT: v_or_b32_sdwa v0, v0, v12 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; GFX9-NEXT: v_or_b32_sdwa v0, v0, v27 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
; GFX9-NEXT: v_and_b32_e32 v0, 0xffff, v0
; GFX9-NEXT: s_waitcnt vmcnt(0)
; GFX9-NEXT: v_or_b32_sdwa v1, v1, v2 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
@@ -188244,7 +188288,7 @@ define inreg <64 x half> @bitcast_v128i8_to_v64f16_scalar(<128 x i8> inreg %a, i
; GFX9-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:416 ; 4-byte Folded Reload
; GFX9-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:504 ; 4-byte Folded Reload
; GFX9-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:492 ; 4-byte Folded Reload
-; GFX9-NEXT: buffer_load_dword v16, off, s[0:3], s32 offset:592 ; 4-byte Folded Reload
+; GFX9-NEXT: buffer_load_dword v16, off, s[0:3], s32 offset:588 ; 4-byte Folded Reload
; GFX9-NEXT: s_waitcnt vmcnt(2)
; GFX9-NEXT: v_or_b32_sdwa v0, v0, v1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
; GFX9-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:404 ; 4-byte Folded Reload
@@ -188285,8 +188329,8 @@ define inreg <64 x half> @bitcast_v128i8_to_v64f16_scalar(<128 x i8> inreg %a, i
; GFX9-NEXT: v_mov_b32_e32 v52, v1
; GFX9-NEXT: v_or_b32_sdwa v1, v50, v1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
; GFX9-NEXT: v_lshl_or_b32 v14, v1, 16, v0
-; GFX9-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:600 ; 4-byte Folded Reload
-; GFX9-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:604 ; 4-byte Folded Reload
+; GFX9-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:604 ; 4-byte Folded Reload
+; GFX9-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:600 ; 4-byte Folded Reload
; GFX9-NEXT: s_waitcnt vmcnt(1)
; GFX9-NEXT: v_mov_b32_e32 v50, v0
; GFX9-NEXT: v_or_b32_sdwa v0, v49, v0 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
@@ -188304,16 +188348,16 @@ define inreg <64 x half> @bitcast_v128i8_to_v64f16_scalar(<128 x i8> inreg %a, i
; GFX9-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:448 ; 4-byte Folded Reload
; GFX9-NEXT: s_waitcnt vmcnt(1)
; GFX9-NEXT: v_mov_b32_e32 v48, v1
-; GFX9-NEXT: v_or_b32_sdwa v1, v25, v1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; GFX9-NEXT: v_or_b32_sdwa v1, v40, v1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
; GFX9-NEXT: v_and_b32_e32 v1, 0xffff, v1
; GFX9-NEXT: v_lshl_or_b32 v17, v17, 16, v1
; GFX9-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:444 ; 4-byte Folded Reload
; GFX9-NEXT: s_waitcnt vmcnt(1)
; GFX9-NEXT: v_or_b32_sdwa v0, v33, v0 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
; GFX9-NEXT: v_and_b32_e32 v0, 0xffff, v0
-; GFX9-NEXT: v_mov_b32_e32 v33, v45
+; GFX9-NEXT: v_mov_b32_e32 v33, v46
; GFX9-NEXT: s_waitcnt vmcnt(0)
-; GFX9-NEXT: v_or_b32_sdwa v1, v45, v1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; GFX9-NEXT: v_or_b32_sdwa v1, v46, v1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
; GFX9-NEXT: v_lshl_or_b32 v18, v1, 16, v0
; GFX9-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:464 ; 4-byte Folded Reload
; GFX9-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:460 ; 4-byte Folded Reload
@@ -188326,7 +188370,7 @@ define inreg <64 x half> @bitcast_v128i8_to_v64f16_scalar(<128 x i8> inreg %a, i
; GFX9-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:472 ; 4-byte Folded Reload
; GFX9-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:468 ; 4-byte Folded Reload
; GFX9-NEXT: s_waitcnt vmcnt(1)
-; GFX9-NEXT: v_or_b32_sdwa v0, v56, v0 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; GFX9-NEXT: v_or_b32_sdwa v0, v42, v0 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
; GFX9-NEXT: v_and_b32_e32 v0, 0xffff, v0
; GFX9-NEXT: s_waitcnt vmcnt(0)
; GFX9-NEXT: v_or_b32_sdwa v1, v47, v1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
@@ -188335,7 +188379,7 @@ define inreg <64 x half> @bitcast_v128i8_to_v64f16_scalar(<128 x i8> inreg %a, i
; GFX9-NEXT: buffer_load_dword v61, off, s[0:3], s32 offset:576 ; 4-byte Folded Reload
; GFX9-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:476 ; 4-byte Folded Reload
; GFX9-NEXT: s_waitcnt vmcnt(2)
-; GFX9-NEXT: v_or_b32_sdwa v0, v42, v0 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; GFX9-NEXT: v_or_b32_sdwa v0, v62, v0 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
; GFX9-NEXT: v_and_b32_e32 v0, 0xffff, v0
; GFX9-NEXT: s_waitcnt vmcnt(0)
; GFX9-NEXT: v_or_b32_sdwa v1, v61, v1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
@@ -188343,121 +188387,122 @@ define inreg <64 x half> @bitcast_v128i8_to_v64f16_scalar(<128 x i8> inreg %a, i
; GFX9-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:528 ; 4-byte Folded Reload
; GFX9-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:532 ; 4-byte Folded Reload
; GFX9-NEXT: s_waitcnt vmcnt(1)
-; GFX9-NEXT: v_or_b32_sdwa v0, v31, v0 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; GFX9-NEXT: v_or_b32_sdwa v0, v57, v0 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
; GFX9-NEXT: v_and_b32_e32 v0, 0xffff, v0
; GFX9-NEXT: s_waitcnt vmcnt(0)
-; GFX9-NEXT: v_or_b32_sdwa v1, v32, v1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; GFX9-NEXT: v_or_b32_sdwa v1, v31, v1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
; GFX9-NEXT: v_lshl_or_b32 v22, v1, 16, v0
; GFX9-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:488 ; 4-byte Folded Reload
; GFX9-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:584 ; 4-byte Folded Reload
; GFX9-NEXT: s_waitcnt vmcnt(1)
-; GFX9-NEXT: v_or_b32_sdwa v0, v51, v0 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; GFX9-NEXT: v_or_b32_sdwa v0, v36, v0 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
; GFX9-NEXT: v_and_b32_e32 v0, 0xffff, v0
; GFX9-NEXT: s_waitcnt vmcnt(0)
; GFX9-NEXT: v_mov_b32_e32 v34, v1
-; GFX9-NEXT: v_or_b32_sdwa v1, v30, v1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; GFX9-NEXT: v_or_b32_sdwa v1, v37, v1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
; GFX9-NEXT: v_lshl_or_b32 v23, v1, 16, v0
-; GFX9-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:588 ; 4-byte Folded Reload
-; GFX9-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:572 ; 4-byte Folded Reload
-; GFX9-NEXT: buffer_load_dword v45, off, s[0:3], s32 offset:536 ; 4-byte Folded Reload
-; GFX9-NEXT: buffer_load_dword v51, off, s[0:3], s32 offset:564 ; 4-byte Folded Reload
-; GFX9-NEXT: buffer_load_dword v32, off, s[0:3], s32 offset:552 ; 4-byte Folded Reload
+; GFX9-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:592 ; 4-byte Folded Reload
+; GFX9-NEXT: buffer_load_dword v57, off, s[0:3], s32 offset:572 ; 4-byte Folded Reload
+; GFX9-NEXT: buffer_load_dword v46, off, s[0:3], s32 offset:536 ; 4-byte Folded Reload
+; GFX9-NEXT: buffer_load_dword v36, off, s[0:3], s32 offset:544 ; 4-byte Folded Reload
+; GFX9-NEXT: buffer_load_dword v37, off, s[0:3], s32 offset:540 ; 4-byte Folded Reload
; GFX9-NEXT: s_waitcnt vmcnt(4)
-; GFX9-NEXT: v_mov_b32_e32 v46, v0
-; GFX9-NEXT: v_or_b32_sdwa v0, v44, v0 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; GFX9-NEXT: v_mov_b32_e32 v40, v0
+; GFX9-NEXT: v_or_b32_sdwa v0, v30, v0 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
; GFX9-NEXT: v_and_b32_e32 v0, 0xffff, v0
; GFX9-NEXT: s_waitcnt vmcnt(2)
-; GFX9-NEXT: v_or_b32_sdwa v1, v35, v45 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; GFX9-NEXT: v_or_b32_sdwa v1, v35, v46 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
; GFX9-NEXT: v_lshl_or_b32 v24, v1, 16, v0
; GFX9-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:580 ; 4-byte Folded Reload
-; GFX9-NEXT: buffer_load_dword v44, off, s[0:3], s32 offset:568 ; 4-byte Folded Reload
-; GFX9-NEXT: v_mov_b32_e32 v35, v45
-; GFX9-NEXT: v_mov_b32_e32 v45, v61
-; GFX9-NEXT: v_mov_b32_e32 v61, v42
+; GFX9-NEXT: s_waitcnt vmcnt(2)
+; GFX9-NEXT: v_or_b32_sdwa v1, v51, v36 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; GFX9-NEXT: buffer_load_dword v51, off, s[0:3], s32 offset:560 ; 4-byte Folded Reload
; GFX9-NEXT: s_waitcnt vmcnt(1)
; GFX9-NEXT: v_mov_b32_e32 v38, v0
-; GFX9-NEXT: v_or_b32_sdwa v0, v36, v0 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
-; GFX9-NEXT: buffer_load_dword v36, off, s[0:3], s32 offset:544 ; 4-byte Folded Reload
+; GFX9-NEXT: v_or_b32_sdwa v0, v28, v0 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
; GFX9-NEXT: v_and_b32_e32 v0, 0xffff, v0
-; GFX9-NEXT: s_waitcnt vmcnt(0)
-; GFX9-NEXT: v_or_b32_sdwa v1, v37, v36 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
-; GFX9-NEXT: buffer_load_dword v37, off, s[0:3], s32 offset:540 ; 4-byte Folded Reload
; GFX9-NEXT: v_lshl_or_b32 v25, v1, 16, v0
-; GFX9-NEXT: v_or_b32_sdwa v1, v54, v51 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
-; GFX9-NEXT: v_mov_b32_e32 v54, v2
-; GFX9-NEXT: s_waitcnt vmcnt(0)
; GFX9-NEXT: v_or_b32_sdwa v0, v41, v37 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
-; GFX9-NEXT: buffer_load_dword v41, off, s[0:3], s32 offset:556 ; 4-byte Folded Reload
; GFX9-NEXT: v_and_b32_e32 v0, 0xffff, v0
-; GFX9-NEXT: v_lshl_or_b32 v26, v1, 16, v0
-; GFX9-NEXT: v_or_b32_sdwa v1, v27, v2 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; GFX9-NEXT: v_mov_b32_e32 v41, v57
; GFX9-NEXT: s_waitcnt vmcnt(0)
-; GFX9-NEXT: v_or_b32_sdwa v0, v29, v41 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; GFX9-NEXT: v_or_b32_sdwa v1, v54, v51 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; GFX9-NEXT: buffer_load_dword v54, off, s[0:3], s32 offset:568 ; 4-byte Folded Reload
+; GFX9-NEXT: v_lshl_or_b32 v26, v1, 16, v0
+; GFX9-NEXT: v_or_b32_sdwa v0, v44, v57 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; GFX9-NEXT: buffer_load_dword v44, off, s[0:3], s32 offset:564 ; 4-byte Folded Reload
; GFX9-NEXT: v_and_b32_e32 v0, 0xffff, v0
+; GFX9-NEXT: s_waitcnt vmcnt(1)
+; GFX9-NEXT: v_or_b32_sdwa v1, v45, v54 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; GFX9-NEXT: buffer_load_dword v45, off, s[0:3], s32 offset:552 ; 4-byte Folded Reload
; GFX9-NEXT: v_lshl_or_b32 v27, v1, 16, v0
-; GFX9-NEXT: v_or_b32_sdwa v0, v60, v32 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; GFX9-NEXT: s_waitcnt vmcnt(0)
+; GFX9-NEXT: v_or_b32_sdwa v0, v60, v45 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
; GFX9-NEXT: buffer_load_dword v60, off, s[0:3], s32 offset:548 ; 4-byte Folded Reload
-; GFX9-NEXT: v_or_b32_sdwa v1, v57, v44 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
-; GFX9-NEXT: buffer_load_dword v57, off, s[0:3], s32 offset:560 ; 4-byte Folded Reload
+; GFX9-NEXT: v_or_b32_sdwa v1, v29, v44 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
; GFX9-NEXT: v_and_b32_e32 v0, 0xffff, v0
; GFX9-NEXT: v_lshl_or_b32 v28, v1, 16, v0
-; GFX9-NEXT: s_waitcnt vmcnt(1)
+; GFX9-NEXT: s_waitcnt vmcnt(0)
; GFX9-NEXT: v_or_b32_sdwa v0, v59, v60 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; GFX9-NEXT: buffer_load_dword v59, off, s[0:3], s32 offset:556 ; 4-byte Folded Reload
; GFX9-NEXT: v_and_b32_e32 v0, 0xffff, v0
; GFX9-NEXT: s_waitcnt vmcnt(0)
-; GFX9-NEXT: v_or_b32_sdwa v1, v63, v57 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; GFX9-NEXT: v_or_b32_sdwa v1, v63, v59 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; GFX9-NEXT: v_mov_b32_e32 v57, v59
; GFX9-NEXT: v_lshl_or_b32 v29, v1, 16, v0
; GFX9-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:508 ; 4-byte Folded Reload
; GFX9-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:512 ; 4-byte Folded Reload
; GFX9-NEXT: s_waitcnt vmcnt(1)
-; GFX9-NEXT: v_or_b32_sdwa v0, v62, v0 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
-; GFX9-NEXT: v_and_b32_e32 v0, 0xffff, v0
+; GFX9-NEXT: v_or_b32_sdwa v0, v58, v0 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
; GFX9-NEXT: s_waitcnt vmcnt(0)
-; GFX9-NEXT: v_or_b32_sdwa v1, v40, v1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; GFX9-NEXT: v_or_b32_sdwa v1, v32, v1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; GFX9-NEXT: v_and_b32_e32 v0, 0xffff, v0
; GFX9-NEXT: v_lshl_or_b32 v30, v1, 16, v0
; GFX9-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:520 ; 4-byte Folded Reload
; GFX9-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:524 ; 4-byte Folded Reload
; GFX9-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:516 ; 4-byte Folded Reload
; GFX9-NEXT: s_waitcnt vmcnt(2)
-; GFX9-NEXT: v_or_b32_sdwa v0, v58, v0 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
-; GFX9-NEXT: v_and_b32_e32 v0, 0xffff, v0
+; GFX9-NEXT: v_or_b32_sdwa v0, v56, v0 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
; GFX9-NEXT: s_waitcnt vmcnt(0)
; GFX9-NEXT: v_or_b32_sdwa v1, v1, v2 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; GFX9-NEXT: v_mov_b32_e32 v2, s6
+; GFX9-NEXT: v_mov_b32_e32 v56, v42
+; GFX9-NEXT: v_and_b32_e32 v0, 0xffff, v0
; GFX9-NEXT: v_lshl_or_b32 v31, v1, 16, v0
; GFX9-NEXT: v_mov_b32_e32 v0, s4
; GFX9-NEXT: v_mov_b32_e32 v1, s5
-; GFX9-NEXT: v_mov_b32_e32 v2, s6
; GFX9-NEXT: s_mov_b64 s[4:5], 0
; GFX9-NEXT: s_branch .LBB93_3
; GFX9-NEXT: .LBB93_2:
; GFX9-NEXT: buffer_load_dword v55, off, s[0:3], s32 offset:616 ; 4-byte Folded Reload
; GFX9-NEXT: buffer_load_dword v53, off, s[0:3], s32 offset:612 ; 4-byte Folded Reload
; GFX9-NEXT: buffer_load_dword v52, off, s[0:3], s32 offset:608 ; 4-byte Folded Reload
-; GFX9-NEXT: buffer_load_dword v50, off, s[0:3], s32 offset:600 ; 4-byte Folded Reload
-; GFX9-NEXT: buffer_load_dword v49, off, s[0:3], s32 offset:604 ; 4-byte Folded Reload
+; GFX9-NEXT: buffer_load_dword v50, off, s[0:3], s32 offset:604 ; 4-byte Folded Reload
+; GFX9-NEXT: buffer_load_dword v49, off, s[0:3], s32 offset:600 ; 4-byte Folded Reload
; GFX9-NEXT: buffer_load_dword v48, off, s[0:3], s32 offset:596 ; 4-byte Folded Reload
-; GFX9-NEXT: buffer_load_dword v39, off, s[0:3], s32 offset:592 ; 4-byte Folded Reload
+; GFX9-NEXT: buffer_load_dword v39, off, s[0:3], s32 offset:588 ; 4-byte Folded Reload
; GFX9-NEXT: buffer_load_dword v38, off, s[0:3], s32 offset:580 ; 4-byte Folded Reload
-; GFX9-NEXT: buffer_load_dword v46, off, s[0:3], s32 offset:588 ; 4-byte Folded Reload
+; GFX9-NEXT: buffer_load_dword v40, off, s[0:3], s32 offset:592 ; 4-byte Folded Reload
; GFX9-NEXT: buffer_load_dword v34, off, s[0:3], s32 offset:584 ; 4-byte Folded Reload
-; GFX9-NEXT: v_mov_b32_e32 v33, v45
-; GFX9-NEXT: buffer_load_dword v35, off, s[0:3], s32 offset:536 ; 4-byte Folded Reload
+; GFX9-NEXT: v_mov_b32_e32 v33, v46
+; GFX9-NEXT: buffer_load_dword v46, off, s[0:3], s32 offset:536 ; 4-byte Folded Reload
; GFX9-NEXT: buffer_load_dword v36, off, s[0:3], s32 offset:544 ; 4-byte Folded Reload
; GFX9-NEXT: buffer_load_dword v37, off, s[0:3], s32 offset:540 ; 4-byte Folded Reload
-; GFX9-NEXT: buffer_load_dword v51, off, s[0:3], s32 offset:564 ; 4-byte Folded Reload
-; GFX9-NEXT: buffer_load_dword v54, off, s[0:3], s32 offset:572 ; 4-byte Folded Reload
-; GFX9-NEXT: buffer_load_dword v41, off, s[0:3], s32 offset:556 ; 4-byte Folded Reload
-; GFX9-NEXT: buffer_load_dword v44, off, s[0:3], s32 offset:568 ; 4-byte Folded Reload
-; GFX9-NEXT: buffer_load_dword v32, off, s[0:3], s32 offset:552 ; 4-byte Folded Reload
-; GFX9-NEXT: buffer_load_dword v57, off, s[0:3], s32 offset:560 ; 4-byte Folded Reload
+; GFX9-NEXT: buffer_load_dword v51, off, s[0:3], s32 offset:560 ; 4-byte Folded Reload
+; GFX9-NEXT: buffer_load_dword v54, off, s[0:3], s32 offset:568 ; 4-byte Folded Reload
+; GFX9-NEXT: buffer_load_dword v41, off, s[0:3], s32 offset:572 ; 4-byte Folded Reload
+; GFX9-NEXT: buffer_load_dword v44, off, s[0:3], s32 offset:564 ; 4-byte Folded Reload
+; GFX9-NEXT: buffer_load_dword v45, off, s[0:3], s32 offset:552 ; 4-byte Folded Reload
+; GFX9-NEXT: v_mov_b32_e32 v56, v61
+; GFX9-NEXT: buffer_load_dword v57, off, s[0:3], s32 offset:556 ; 4-byte Folded Reload
; GFX9-NEXT: buffer_load_dword v60, off, s[0:3], s32 offset:548 ; 4-byte Folded Reload
-; GFX9-NEXT: buffer_load_dword v45, off, s[0:3], s32 offset:576 ; 4-byte Folded Reload
+; GFX9-NEXT: buffer_load_dword v61, off, s[0:3], s32 offset:576 ; 4-byte Folded Reload
; GFX9-NEXT: s_mov_b64 s[4:5], -1
; GFX9-NEXT: ; implicit-def: $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7_vgpr8_vgpr9_vgpr10_vgpr11_vgpr12_vgpr13_vgpr14_vgpr15
; GFX9-NEXT: ; implicit-def: $vgpr16_vgpr17_vgpr18_vgpr19_vgpr20_vgpr21_vgpr22_vgpr23_vgpr24_vgpr25_vgpr26_vgpr27_vgpr28_vgpr29_vgpr30_vgpr31
; GFX9-NEXT: .LBB93_3: ; %Flow
; GFX9-NEXT: buffer_load_dword v42, off, s[0:3], s32 offset:412 ; 4-byte Folded Reload
-; GFX9-NEXT: buffer_load_dword v40, off, s[0:3], s32 offset:408 ; 4-byte Folded Reload
+; GFX9-NEXT: buffer_load_dword v32, off, s[0:3], s32 offset:408 ; 4-byte Folded Reload
; GFX9-NEXT: buffer_load_dword v58, off, s[0:3], s32 offset:404 ; 4-byte Folded Reload
; GFX9-NEXT: buffer_load_dword v63, off, s[0:3], s32 offset:400 ; 4-byte Folded Reload
; GFX9-NEXT: s_andn2_b64 vcc, exec, s[4:5]
@@ -188660,7 +188705,7 @@ define inreg <64 x half> @bitcast_v128i8_to_v64f16_scalar(<128 x i8> inreg %a, i
; GFX9-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:816 ; 4-byte Folded Reload
; GFX9-NEXT: s_waitcnt vmcnt(1)
; GFX9-NEXT: v_add_u32_e32 v0, 3, v0
-; GFX9-NEXT: v_or_b32_sdwa v0, v40, v0 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:BYTE_0
+; GFX9-NEXT: v_or_b32_sdwa v0, v32, v0 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:BYTE_0
; GFX9-NEXT: s_waitcnt vmcnt(0)
; GFX9-NEXT: v_add_u32_e32 v1, 3, v1
; GFX9-NEXT: v_or_b32_sdwa v1, v2, v1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:BYTE_0
@@ -188720,11 +188765,11 @@ define inreg <64 x half> @bitcast_v128i8_to_v64f16_scalar(<128 x i8> inreg %a, i
; GFX9-NEXT: v_and_b32_e32 v0, 0xffff, v0
; GFX9-NEXT: s_waitcnt vmcnt(0)
; GFX9-NEXT: v_add_u32_e32 v26, 3, v26
-; GFX9-NEXT: v_or_b32_sdwa v48, v46, v26 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:BYTE_0
+; GFX9-NEXT: v_or_b32_sdwa v48, v40, v26 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:BYTE_0
; GFX9-NEXT: buffer_load_dword v26, off, s[0:3], s32 offset:688 ; 4-byte Folded Reload
; GFX9-NEXT: s_waitcnt vmcnt(0)
; GFX9-NEXT: v_add_u32_e32 v26, 3, v26
-; GFX9-NEXT: v_or_b32_sdwa v49, v35, v26 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:BYTE_0
+; GFX9-NEXT: v_or_b32_sdwa v49, v46, v26 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:BYTE_0
; GFX9-NEXT: buffer_load_dword v26, off, s[0:3], s32 offset:676 ; 4-byte Folded Reload
; GFX9-NEXT: s_waitcnt vmcnt(0)
; GFX9-NEXT: v_add_u32_e32 v26, 3, v26
@@ -188759,7 +188804,7 @@ define inreg <64 x half> @bitcast_v128i8_to_v64f16_scalar(<128 x i8> inreg %a, i
; GFX9-NEXT: v_or_b32_sdwa v53, v27, v26 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:BYTE_0
; GFX9-NEXT: buffer_load_dword v27, off, s[0:3], s32 offset:480 ; 4-byte Folded Reload
; GFX9-NEXT: v_add_u32_e32 v24, 3, v24
-; GFX9-NEXT: v_add_u32_e32 v26, 3, v61
+; GFX9-NEXT: v_add_u32_e32 v26, 3, v62
; GFX9-NEXT: v_or_b32_sdwa v24, v54, v24 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:BYTE_0
; GFX9-NEXT: v_add_u32_e32 v36, 0x300, v24
; GFX9-NEXT: v_add_u32_e32 v24, 0x300, v48
@@ -188768,7 +188813,7 @@ define inreg <64 x half> @bitcast_v128i8_to_v64f16_scalar(<128 x i8> inreg %a, i
; GFX9-NEXT: s_waitcnt vmcnt(0)
; GFX9-NEXT: v_or_b32_sdwa v54, v27, v26 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:BYTE_0
; GFX9-NEXT: buffer_load_dword v27, off, s[0:3], s32 offset:476 ; 4-byte Folded Reload
-; GFX9-NEXT: v_add_u32_e32 v26, 3, v45
+; GFX9-NEXT: v_add_u32_e32 v26, 3, v61
; GFX9-NEXT: v_add_u32_e32 v20, 3, v20
; GFX9-NEXT: v_or_b32_sdwa v20, v57, v20 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:BYTE_0
; GFX9-NEXT: v_add_u32_e32 v34, 0x300, v20
@@ -188777,7 +188822,7 @@ define inreg <64 x half> @bitcast_v128i8_to_v64f16_scalar(<128 x i8> inreg %a, i
; GFX9-NEXT: buffer_load_dword v27, off, s[0:3], s32 offset:472 ; 4-byte Folded Reload
; GFX9-NEXT: v_add_u32_e32 v26, 3, v56
; GFX9-NEXT: v_add_u32_e32 v21, 3, v21
-; GFX9-NEXT: v_or_b32_sdwa v21, v32, v21 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:BYTE_0
+; GFX9-NEXT: v_or_b32_sdwa v21, v45, v21 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:BYTE_0
; GFX9-NEXT: v_add_u32_e32 v28, 0x300, v21
; GFX9-NEXT: v_add_u32_e32 v21, 0x300, v54
; GFX9-NEXT: v_and_b32_e32 v21, 0xffff, v21
@@ -190633,212 +190678,207 @@ define <128 x i8> @bitcast_v64f16_to_v128i8(<64 x half> %a, i32 %b) {
; SI-NEXT: buffer_load_dword v52, off, s[0:3], s32 offset:32
; SI-NEXT: buffer_load_dword v53, off, s[0:3], s32 offset:44
; SI-NEXT: buffer_load_dword v54, off, s[0:3], s32 offset:40
-; SI-NEXT: buffer_load_dword v55, off, s[0:3], s32 offset:52
-; SI-NEXT: buffer_load_dword v40, off, s[0:3], s32 offset:48
-; SI-NEXT: buffer_load_dword v41, off, s[0:3], s32 offset:60
-; SI-NEXT: buffer_load_dword v42, off, s[0:3], s32 offset:56
-; SI-NEXT: buffer_load_dword v43, off, s[0:3], s32 offset:68
-; SI-NEXT: buffer_load_dword v44, off, s[0:3], s32 offset:64
+; SI-NEXT: buffer_load_dword v41, off, s[0:3], s32 offset:52
+; SI-NEXT: buffer_load_dword v44, off, s[0:3], s32 offset:48
+; SI-NEXT: s_waitcnt expcnt(6)
+; SI-NEXT: buffer_load_dword v57, off, s[0:3], s32 offset:60
+; SI-NEXT: s_waitcnt expcnt(5)
+; SI-NEXT: buffer_load_dword v58, off, s[0:3], s32 offset:56
; SI-NEXT: s_waitcnt expcnt(3)
-; SI-NEXT: buffer_load_dword v60, off, s[0:3], s32 offset:76
+; SI-NEXT: buffer_load_dword v60, off, s[0:3], s32 offset:68
; SI-NEXT: s_waitcnt expcnt(2)
-; SI-NEXT: buffer_load_dword v61, off, s[0:3], s32 offset:72
+; SI-NEXT: buffer_load_dword v61, off, s[0:3], s32 offset:64
+; SI-NEXT: s_waitcnt expcnt(1)
+; SI-NEXT: buffer_load_dword v62, off, s[0:3], s32 offset:76
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: buffer_load_dword v63, off, s[0:3], s32 offset:84
-; SI-NEXT: buffer_load_dword v45, off, s[0:3], s32 offset:80
-; SI-NEXT: buffer_load_dword v46, off, s[0:3], s32 offset:92
+; SI-NEXT: buffer_load_dword v63, off, s[0:3], s32 offset:72
+; SI-NEXT: buffer_load_dword v55, off, s[0:3], s32 offset:84
+; SI-NEXT: buffer_load_dword v40, off, s[0:3], s32 offset:80
+; SI-NEXT: buffer_load_dword v42, off, s[0:3], s32 offset:92
; SI-NEXT: v_cvt_f16_f32_e32 v2, v2
-; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:340 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:356 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_cvt_f16_f32_e32 v1, v3
-; SI-NEXT: v_cvt_f16_f32_e32 v33, v4
-; SI-NEXT: buffer_store_dword v2, off, s[0:3], s32 offset:360 ; 4-byte Folded Spill
-; SI-NEXT: v_cvt_f16_f32_e32 v31, v12
-; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:308 ; 4-byte Folded Spill
+; SI-NEXT: v_cvt_f16_f32_e32 v33, v8
+; SI-NEXT: buffer_store_dword v2, off, s[0:3], s32 offset:376 ; 4-byte Folded Spill
+; SI-NEXT: v_cvt_f16_f32_e32 v46, v4
+; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:328 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_cvt_f16_f32_e32 v1, v6
-; SI-NEXT: v_cvt_f16_f32_e32 v12, v16
-; SI-NEXT: v_cvt_f16_f32_e32 v32, v8
-; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:440 ; 4-byte Folded Spill
+; SI-NEXT: v_cvt_f16_f32_e32 v31, v12
+; SI-NEXT: v_cvt_f16_f32_e32 v32, v24
+; SI-NEXT: v_cvt_f16_f32_e32 v12, v28
+; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:456 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_cvt_f16_f32_e32 v1, v5
-; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:416 ; 4-byte Folded Spill
+; SI-NEXT: ; implicit-def: $vgpr28
+; SI-NEXT: ; implicit-def: $vgpr24
+; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:432 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_cvt_f16_f32_e32 v1, v7
-; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:388 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:404 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_cvt_f16_f32_e32 v1, v10
-; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:492 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:508 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_cvt_f16_f32_e32 v1, v9
-; SI-NEXT: v_cvt_f16_f32_e32 v9, v20
-; SI-NEXT: ; implicit-def: $vgpr20
-; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:476 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:492 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_cvt_f16_f32_e32 v1, v11
-; SI-NEXT: v_cvt_f16_f32_e32 v11, v28
-; SI-NEXT: ; implicit-def: $vgpr28
-; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:464 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:480 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_cvt_f16_f32_e32 v1, v14
-; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:512 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:528 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_cvt_f16_f32_e32 v1, v13
-; SI-NEXT: v_cvt_f16_f32_e32 v13, v24
-; SI-NEXT: ; implicit-def: $vgpr24
-; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:508 ; 4-byte Folded Spill
+; SI-NEXT: v_cvt_f16_f32_e32 v13, v16
+; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:524 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_cvt_f16_f32_e32 v1, v15
-; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:496 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:512 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_cvt_f16_f32_e32 v1, v18
; SI-NEXT: ; implicit-def: $vgpr18
-; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:548 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:564 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_cvt_f16_f32_e32 v1, v17
; SI-NEXT: s_waitcnt vmcnt(14)
-; SI-NEXT: v_cvt_f16_f32_e32 v10, v35
+; SI-NEXT: v_cvt_f16_f32_e32 v11, v35
+; SI-NEXT: v_cvt_f16_f32_e32 v9, v37
; SI-NEXT: v_cmp_ne_u32_e32 vcc, 0, v34
+; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:552 ; 4-byte Folded Spill
+; SI-NEXT: s_waitcnt expcnt(0)
+; SI-NEXT: v_cvt_f16_f32_e32 v1, v20
+; SI-NEXT: v_cvt_f16_f32_e32 v45, v49
+; SI-NEXT: v_cvt_f16_f32_e32 v43, v50
+; SI-NEXT: ; implicit-def: $vgpr49
+; SI-NEXT: ; implicit-def: $vgpr50
+; SI-NEXT: ; implicit-def: $vgpr37
; SI-NEXT: ; implicit-def: $vgpr35
; SI-NEXT: ; implicit-def: $vgpr34
-; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:536 ; 4-byte Folded Spill
+; SI-NEXT: ; implicit-def: $vgpr20
+; SI-NEXT: ; implicit-def: $vgpr17
+; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:228 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_cvt_f16_f32_e32 v1, v19
-; SI-NEXT: v_cvt_f16_f32_e32 v8, v39
-; SI-NEXT: ; implicit-def: $vgpr39
-; SI-NEXT: ; implicit-def: $vgpr19
-; SI-NEXT: v_cvt_f16_f32_e32 v7, v51
-; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:524 ; 4-byte Folded Spill
-; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_cvt_f16_f32_e32 v1, v22
; SI-NEXT: v_cvt_f16_f32_e32 v56, v53
-; SI-NEXT: v_cvt_f16_f32_e32 v5, v55
; SI-NEXT: v_cvt_f16_f32_e32 v47, v54
-; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:600 ; 4-byte Folded Spill
+; SI-NEXT: v_cvt_f16_f32_e32 v10, v44
+; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:540 ; 4-byte Folded Spill
+; SI-NEXT: s_waitcnt expcnt(0)
+; SI-NEXT: v_cvt_f16_f32_e32 v1, v22
+; SI-NEXT: v_cvt_f16_f32_e32 v59, v57
+; SI-NEXT: v_cvt_f16_f32_e32 v58, v58
+; SI-NEXT: v_cvt_f16_f32_e32 v2, v62
+; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:616 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_cvt_f16_f32_e32 v1, v21
-; SI-NEXT: v_cvt_f16_f32_e32 v59, v41
-; SI-NEXT: v_cvt_f16_f32_e32 v58, v42
-; SI-NEXT: v_cvt_f16_f32_e32 v57, v44
-; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:596 ; 4-byte Folded Spill
+; SI-NEXT: v_cvt_f16_f32_e32 v3, v55
+; SI-NEXT: v_cvt_f16_f32_e32 v57, v61
+; SI-NEXT: s_waitcnt vmcnt(14)
+; SI-NEXT: v_cvt_f16_f32_e32 v61, v42
+; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:612 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_cvt_f16_f32_e32 v1, v23
-; SI-NEXT: v_cvt_f16_f32_e32 v62, v60
-; SI-NEXT: v_cvt_f16_f32_e32 v61, v61
-; SI-NEXT: v_cvt_f16_f32_e32 v60, v45
-; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:584 ; 4-byte Folded Spill
-; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_cvt_f16_f32_e32 v1, v26
-; SI-NEXT: ; implicit-def: $vgpr45
-; SI-NEXT: ; kill: killed $vgpr45
+; SI-NEXT: buffer_store_dword v3, off, s[0:3], s32 offset:224 ; 4-byte Folded Spill
+; SI-NEXT: ; implicit-def: $vgpr42
; SI-NEXT: ; implicit-def: $vgpr44
-; SI-NEXT: ; implicit-def: $vgpr41
; SI-NEXT: ; implicit-def: $vgpr54
; SI-NEXT: ; implicit-def: $vgpr53
-; SI-NEXT: ; implicit-def: $vgpr51
-; SI-NEXT: ; implicit-def: $vgpr26
; SI-NEXT: ; implicit-def: $vgpr22
; SI-NEXT: ; implicit-def: $vgpr23
; SI-NEXT: ; implicit-def: $vgpr21
-; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:612 ; 4-byte Folded Spill
+; SI-NEXT: ; implicit-def: $vgpr19
+; SI-NEXT: ; kill: killed $vgpr42
+; SI-NEXT: ; implicit-def: $vgpr42
+; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:600 ; 4-byte Folded Spill
+; SI-NEXT: s_waitcnt expcnt(0)
+; SI-NEXT: v_cvt_f16_f32_e32 v1, v26
+; SI-NEXT: ; implicit-def: $vgpr26
+; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:628 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_cvt_f16_f32_e32 v1, v25
; SI-NEXT: ; implicit-def: $vgpr25
-; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:608 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:624 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_cvt_f16_f32_e32 v1, v27
; SI-NEXT: ; implicit-def: $vgpr27
-; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:604 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:620 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_cvt_f16_f32_e32 v1, v30
; SI-NEXT: ; implicit-def: $vgpr30
-; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:624 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:640 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_cvt_f16_f32_e32 v1, v29
; SI-NEXT: ; implicit-def: $vgpr29
-; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:620 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:636 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_cvt_f16_f32_e32 v1, v36
; SI-NEXT: ; implicit-def: $vgpr36
-; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:616 ; 4-byte Folded Spill
-; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_cvt_f16_f32_e32 v1, v37
-; SI-NEXT: ; implicit-def: $vgpr37
-; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:636 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:632 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_cvt_f16_f32_e32 v1, v38
; SI-NEXT: ; implicit-def: $vgpr38
-; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:632 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:648 ; 4-byte Folded Spill
+; SI-NEXT: s_waitcnt expcnt(0)
+; SI-NEXT: v_cvt_f16_f32_e32 v1, v39
+; SI-NEXT: ; implicit-def: $vgpr39
+; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:220 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_cvt_f16_f32_e32 v1, v48
; SI-NEXT: ; implicit-def: $vgpr48
-; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:628 ; 4-byte Folded Spill
-; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_cvt_f16_f32_e32 v1, v49
-; SI-NEXT: ; implicit-def: $vgpr49
-; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:648 ; 4-byte Folded Spill
-; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_cvt_f16_f32_e32 v1, v50
-; SI-NEXT: ; implicit-def: $vgpr50
; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:644 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
+; SI-NEXT: v_cvt_f16_f32_e32 v1, v51
+; SI-NEXT: ; implicit-def: $vgpr51
+; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:216 ; 4-byte Folded Spill
+; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_cvt_f16_f32_e32 v1, v52
; SI-NEXT: ; implicit-def: $vgpr52
-; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:640 ; 4-byte Folded Spill
-; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_cvt_f16_f32_e32 v1, v40
; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:652 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_cvt_f16_f32_e32 v1, v43
+; SI-NEXT: v_cvt_f16_f32_e32 v1, v41
+; SI-NEXT: ; implicit-def: $vgpr41
; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:212 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_cvt_f16_f32_e32 v1, v63
-; SI-NEXT: s_waitcnt vmcnt(14)
-; SI-NEXT: v_cvt_f16_f32_e32 v63, v46
-; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:204 ; 4-byte Folded Spill
+; SI-NEXT: v_cvt_f16_f32_e32 v1, v60
+; SI-NEXT: v_cvt_f16_f32_e32 v60, v40
+; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:208 ; 4-byte Folded Spill
+; SI-NEXT: buffer_load_dword v3, off, s[0:3], s32 offset:88
+; SI-NEXT: buffer_load_dword v5, off, s[0:3], s32 offset:100
+; SI-NEXT: buffer_load_dword v6, off, s[0:3], s32 offset:96
+; SI-NEXT: buffer_load_dword v7, off, s[0:3], s32 offset:108
+; SI-NEXT: buffer_load_dword v8, off, s[0:3], s32 offset:104
+; SI-NEXT: buffer_load_dword v14, off, s[0:3], s32 offset:116
+; SI-NEXT: buffer_load_dword v15, off, s[0:3], s32 offset:112
+; SI-NEXT: buffer_load_dword v16, off, s[0:3], s32 offset:124
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:88
-; SI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:100
-; SI-NEXT: buffer_load_dword v3, off, s[0:3], s32 offset:96
-; SI-NEXT: buffer_load_dword v4, off, s[0:3], s32 offset:108
-; SI-NEXT: buffer_load_dword v14, off, s[0:3], s32 offset:104
-; SI-NEXT: buffer_load_dword v15, off, s[0:3], s32 offset:116
-; SI-NEXT: buffer_load_dword v16, off, s[0:3], s32 offset:112
-; SI-NEXT: buffer_load_dword v17, off, s[0:3], s32 offset:124
+; SI-NEXT: v_cvt_f16_f32_e32 v1, v63
; SI-NEXT: s_waitcnt vmcnt(7)
-; SI-NEXT: v_cvt_f16_f32_e32 v40, v1
+; SI-NEXT: v_cvt_f16_f32_e32 v4, v3
; SI-NEXT: s_waitcnt vmcnt(6)
-; SI-NEXT: v_cvt_f16_f32_e32 v1, v2
-; SI-NEXT: s_waitcnt vmcnt(5)
-; SI-NEXT: v_cvt_f16_f32_e32 v55, v3
+; SI-NEXT: v_cvt_f16_f32_e32 v3, v5
; SI-NEXT: s_waitcnt vmcnt(4)
-; SI-NEXT: v_cvt_f16_f32_e32 v43, v4
-; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:208 ; 4-byte Folded Spill
-; SI-NEXT: s_waitcnt vmcnt(4) expcnt(0)
-; SI-NEXT: v_cvt_f16_f32_e32 v1, v14
+; SI-NEXT: v_cvt_f16_f32_e32 v5, v7
; SI-NEXT: s_waitcnt vmcnt(3)
-; SI-NEXT: v_cvt_f16_f32_e32 v6, v15
-; SI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:120
+; SI-NEXT: v_cvt_f16_f32_e32 v63, v8
+; SI-NEXT: buffer_store_dword v3, off, s[0:3], s32 offset:204 ; 4-byte Folded Spill
+; SI-NEXT: s_waitcnt expcnt(0)
+; SI-NEXT: v_cvt_f16_f32_e32 v3, v6
+; SI-NEXT: s_waitcnt vmcnt(3)
+; SI-NEXT: v_cvt_f16_f32_e32 v40, v14
+; SI-NEXT: s_waitcnt vmcnt(2)
+; SI-NEXT: v_cvt_f16_f32_e32 v62, v15
+; SI-NEXT: buffer_load_dword v6, off, s[0:3], s32 offset:120
; SI-NEXT: buffer_load_dword v14, off, s[0:3], s32 offset:132
; SI-NEXT: buffer_load_dword v15, off, s[0:3], s32 offset:128
-; SI-NEXT: s_waitcnt vmcnt(5)
-; SI-NEXT: v_cvt_f16_f32_e32 v42, v16
; SI-NEXT: s_waitcnt vmcnt(4)
-; SI-NEXT: v_cvt_f16_f32_e32 v3, v17
+; SI-NEXT: v_cvt_f16_f32_e32 v7, v16
; SI-NEXT: ; implicit-def: $vgpr16
-; SI-NEXT: ; implicit-def: $vgpr17
; SI-NEXT: s_waitcnt vmcnt(2)
-; SI-NEXT: v_cvt_f16_f32_e32 v4, v2
+; SI-NEXT: v_cvt_f16_f32_e32 v8, v6
; SI-NEXT: s_waitcnt vmcnt(1)
-; SI-NEXT: v_cvt_f16_f32_e32 v46, v14
-; SI-NEXT: ; implicit-def: $vgpr14
-; SI-NEXT: ; kill: killed $vgpr14
-; SI-NEXT: ; implicit-def: $vgpr14
-; SI-NEXT: ; kill: killed $vgpr14
-; SI-NEXT: ; implicit-def: $vgpr14
-; SI-NEXT: ; kill: killed $vgpr14
-; SI-NEXT: ; implicit-def: $vgpr14
-; SI-NEXT: ; kill: killed $vgpr14
+; SI-NEXT: v_cvt_f16_f32_e32 v55, v14
; SI-NEXT: ; implicit-def: $vgpr14
; SI-NEXT: ; kill: killed $vgpr14
; SI-NEXT: ; implicit-def: $vgpr14
@@ -190962,20 +191002,30 @@ define <128 x i8> @bitcast_v64f16_to_v128i8(<64 x half> %a, i32 %b) {
; SI-NEXT: ; implicit-def: $vgpr14
; SI-NEXT: ; kill: killed $vgpr14
; SI-NEXT: ; implicit-def: $vgpr14
+; SI-NEXT: s_waitcnt vmcnt(0)
+; SI-NEXT: v_cvt_f16_f32_e32 v6, v15
; SI-NEXT: ; kill: killed $vgpr14
; SI-NEXT: ; implicit-def: $vgpr14
+; SI-NEXT: ; implicit-def: $vgpr15
; SI-NEXT: ; kill: killed $vgpr14
; SI-NEXT: ; implicit-def: $vgpr14
+; SI-NEXT: ; kill: killed $vgpr15
+; SI-NEXT: ; implicit-def: $vgpr15
; SI-NEXT: ; kill: killed $vgpr14
; SI-NEXT: ; implicit-def: $vgpr14
+; SI-NEXT: ; kill: killed $vgpr15
+; SI-NEXT: ; implicit-def: $vgpr15
; SI-NEXT: ; kill: killed $vgpr14
; SI-NEXT: ; implicit-def: $vgpr14
-; SI-NEXT: s_waitcnt vmcnt(0)
-; SI-NEXT: v_cvt_f16_f32_e32 v2, v15
+; SI-NEXT: ; kill: killed $vgpr15
+; SI-NEXT: ; implicit-def: $vgpr15
; SI-NEXT: ; kill: killed $vgpr14
; SI-NEXT: ; implicit-def: $vgpr14
+; SI-NEXT: ; kill: killed $vgpr15
+; SI-NEXT: ; implicit-def: $vgpr15
; SI-NEXT: ; kill: killed $vgpr14
; SI-NEXT: ; implicit-def: $vgpr14
+; SI-NEXT: ; kill: killed $vgpr15
; SI-NEXT: ; implicit-def: $vgpr15
; SI-NEXT: ; kill: killed $vgpr14
; SI-NEXT: ; implicit-def: $vgpr14
@@ -190985,407 +191035,400 @@ define <128 x i8> @bitcast_v64f16_to_v128i8(<64 x half> %a, i32 %b) {
; SI-NEXT: ; implicit-def: $vgpr14
; SI-NEXT: ; kill: killed $vgpr15
; SI-NEXT: ; implicit-def: $vgpr15
-; SI-NEXT: v_mov_b32_e32 v45, v46
-; SI-NEXT: v_mov_b32_e32 v46, v6
-; SI-NEXT: v_mov_b32_e32 v6, v5
-; SI-NEXT: ; implicit-def: $vgpr5
; SI-NEXT: ; kill: killed $vgpr14
; SI-NEXT: ; implicit-def: $vgpr14
; SI-NEXT: ; kill: killed $vgpr15
; SI-NEXT: ; implicit-def: $vgpr15
-; SI-NEXT: ; kill: killed $vgpr5
-; SI-NEXT: ; implicit-def: $vgpr5
-; SI-NEXT: ; kill: killed $vgpr5
; SI-NEXT: s_and_saveexec_b64 s[4:5], vcc
; SI-NEXT: s_xor_b64 s[4:5], exec, s[4:5]
; SI-NEXT: s_cbranch_execz .LBB94_2
; SI-NEXT: ; %bb.1: ; %cmp.false
-; SI-NEXT: buffer_load_dword v5, off, s[0:3], s32 offset:360 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v14, off, s[0:3], s32 offset:376 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v15, off, s[0:3], s32 offset:356 ; 4-byte Folded Reload
+; SI-NEXT: v_mov_b32_e32 v16, v46
+; SI-NEXT: v_mov_b32_e32 v42, v40
+; SI-NEXT: v_mov_b32_e32 v40, v55
+; SI-NEXT: v_lshlrev_b32_e32 v9, 16, v9
+; SI-NEXT: v_lshlrev_b32_e32 v2, 16, v2
+; SI-NEXT: v_or_b32_e32 v20, v1, v2
+; SI-NEXT: buffer_load_dword v55, off, s[0:3], s32 offset:220 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:204 ; 4-byte Folded Reload
+; SI-NEXT: s_waitcnt vmcnt(3)
+; SI-NEXT: v_lshlrev_b32_e32 v14, 16, v14
+; SI-NEXT: s_waitcnt vmcnt(2)
+; SI-NEXT: v_or_b32_e32 v44, v15, v14
+; SI-NEXT: buffer_load_dword v15, off, s[0:3], s32 offset:328 ; 4-byte Folded Reload
+; SI-NEXT: v_lshlrev_b32_e32 v14, 16, v46
+; SI-NEXT: v_mov_b32_e32 v46, v33
; SI-NEXT: s_waitcnt vmcnt(0)
-; SI-NEXT: v_lshlrev_b32_e32 v14, 16, v5
-; SI-NEXT: buffer_load_dword v5, off, s[0:3], s32 offset:340 ; 4-byte Folded Reload
+; SI-NEXT: v_or_b32_e32 v41, v15, v14
+; SI-NEXT: buffer_load_dword v14, off, s[0:3], s32 offset:456 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v15, off, s[0:3], s32 offset:432 ; 4-byte Folded Reload
+; SI-NEXT: s_waitcnt vmcnt(1)
+; SI-NEXT: v_lshlrev_b32_e32 v14, 16, v14
; SI-NEXT: s_waitcnt vmcnt(0)
-; SI-NEXT: v_or_b32_e32 v44, v5, v14
-; SI-NEXT: buffer_load_dword v5, off, s[0:3], s32 offset:308 ; 4-byte Folded Reload
+; SI-NEXT: v_or_b32_e32 v54, v15, v14
+; SI-NEXT: buffer_load_dword v15, off, s[0:3], s32 offset:404 ; 4-byte Folded Reload
; SI-NEXT: v_lshlrev_b32_e32 v14, 16, v33
+; SI-NEXT: v_mov_b32_e32 v33, v32
+; SI-NEXT: v_mov_b32_e32 v32, v31
; SI-NEXT: s_waitcnt vmcnt(0)
-; SI-NEXT: v_or_b32_e32 v41, v5, v14
-; SI-NEXT: buffer_load_dword v5, off, s[0:3], s32 offset:440 ; 4-byte Folded Reload
-; SI-NEXT: s_waitcnt vmcnt(0)
-; SI-NEXT: v_lshlrev_b32_e32 v14, 16, v5
-; SI-NEXT: buffer_load_dword v5, off, s[0:3], s32 offset:416 ; 4-byte Folded Reload
-; SI-NEXT: s_waitcnt vmcnt(0)
-; SI-NEXT: v_or_b32_e32 v54, v5, v14
-; SI-NEXT: buffer_load_dword v5, off, s[0:3], s32 offset:388 ; 4-byte Folded Reload
-; SI-NEXT: v_lshlrev_b32_e32 v14, 16, v32
-; SI-NEXT: s_waitcnt vmcnt(0)
-; SI-NEXT: v_or_b32_e32 v53, v5, v14
-; SI-NEXT: buffer_load_dword v5, off, s[0:3], s32 offset:492 ; 4-byte Folded Reload
-; SI-NEXT: s_waitcnt vmcnt(0)
-; SI-NEXT: v_lshlrev_b32_e32 v14, 16, v5
-; SI-NEXT: buffer_load_dword v5, off, s[0:3], s32 offset:476 ; 4-byte Folded Reload
+; SI-NEXT: v_or_b32_e32 v53, v15, v14
+; SI-NEXT: buffer_load_dword v14, off, s[0:3], s32 offset:508 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v15, off, s[0:3], s32 offset:492 ; 4-byte Folded Reload
+; SI-NEXT: s_waitcnt vmcnt(1)
+; SI-NEXT: v_lshlrev_b32_e32 v14, 16, v14
; SI-NEXT: s_waitcnt vmcnt(0)
-; SI-NEXT: v_or_b32_e32 v51, v5, v14
-; SI-NEXT: buffer_load_dword v5, off, s[0:3], s32 offset:464 ; 4-byte Folded Reload
+; SI-NEXT: v_or_b32_e32 v51, v15, v14
+; SI-NEXT: buffer_load_dword v15, off, s[0:3], s32 offset:480 ; 4-byte Folded Reload
; SI-NEXT: v_lshlrev_b32_e32 v14, 16, v31
+; SI-NEXT: v_mov_b32_e32 v31, v13
; SI-NEXT: s_waitcnt vmcnt(0)
-; SI-NEXT: v_or_b32_e32 v52, v5, v14
-; SI-NEXT: buffer_load_dword v5, off, s[0:3], s32 offset:512 ; 4-byte Folded Reload
-; SI-NEXT: s_waitcnt vmcnt(0)
-; SI-NEXT: v_lshlrev_b32_e32 v14, 16, v5
-; SI-NEXT: buffer_load_dword v5, off, s[0:3], s32 offset:508 ; 4-byte Folded Reload
-; SI-NEXT: s_waitcnt vmcnt(0)
-; SI-NEXT: v_or_b32_e32 v49, v5, v14
-; SI-NEXT: buffer_load_dword v5, off, s[0:3], s32 offset:496 ; 4-byte Folded Reload
-; SI-NEXT: v_lshlrev_b32_e32 v14, 16, v12
-; SI-NEXT: s_waitcnt vmcnt(0)
-; SI-NEXT: v_or_b32_e32 v50, v5, v14
-; SI-NEXT: buffer_load_dword v5, off, s[0:3], s32 offset:548 ; 4-byte Folded Reload
-; SI-NEXT: s_waitcnt vmcnt(0)
-; SI-NEXT: v_lshlrev_b32_e32 v14, 16, v5
-; SI-NEXT: buffer_load_dword v5, off, s[0:3], s32 offset:536 ; 4-byte Folded Reload
-; SI-NEXT: s_waitcnt vmcnt(0)
-; SI-NEXT: v_or_b32_e32 v48, v5, v14
-; SI-NEXT: buffer_load_dword v5, off, s[0:3], s32 offset:524 ; 4-byte Folded Reload
-; SI-NEXT: v_lshlrev_b32_e32 v14, 16, v9
-; SI-NEXT: s_waitcnt vmcnt(0)
-; SI-NEXT: v_or_b32_e32 v39, v5, v14
-; SI-NEXT: v_alignbit_b32 v5, v41, v44, 24
-; SI-NEXT: buffer_store_dword v5, off, s[0:3], s32 offset:280 ; 4-byte Folded Spill
-; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: buffer_load_dword v5, off, s[0:3], s32 offset:600 ; 4-byte Folded Reload
-; SI-NEXT: s_waitcnt vmcnt(0)
-; SI-NEXT: v_lshlrev_b32_e32 v14, 16, v5
-; SI-NEXT: buffer_load_dword v5, off, s[0:3], s32 offset:596 ; 4-byte Folded Reload
+; SI-NEXT: v_or_b32_e32 v52, v15, v14
+; SI-NEXT: buffer_load_dword v14, off, s[0:3], s32 offset:528 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v15, off, s[0:3], s32 offset:524 ; 4-byte Folded Reload
+; SI-NEXT: s_waitcnt vmcnt(1)
+; SI-NEXT: v_lshlrev_b32_e32 v14, 16, v14
; SI-NEXT: s_waitcnt vmcnt(0)
-; SI-NEXT: v_or_b32_e32 v37, v5, v14
-; SI-NEXT: buffer_load_dword v5, off, s[0:3], s32 offset:584 ; 4-byte Folded Reload
+; SI-NEXT: v_or_b32_e32 v49, v15, v14
+; SI-NEXT: buffer_load_dword v15, off, s[0:3], s32 offset:512 ; 4-byte Folded Reload
; SI-NEXT: v_lshlrev_b32_e32 v14, 16, v13
+; SI-NEXT: v_mov_b32_e32 v13, v12
+; SI-NEXT: v_mov_b32_e32 v12, v11
+; SI-NEXT: buffer_load_dword v11, off, s[0:3], s32 offset:228 ; 4-byte Folded Reload
+; SI-NEXT: s_waitcnt vmcnt(1)
+; SI-NEXT: v_or_b32_e32 v50, v15, v14
+; SI-NEXT: buffer_load_dword v14, off, s[0:3], s32 offset:564 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v15, off, s[0:3], s32 offset:552 ; 4-byte Folded Reload
+; SI-NEXT: s_waitcnt vmcnt(1)
+; SI-NEXT: v_lshlrev_b32_e32 v14, 16, v14
; SI-NEXT: s_waitcnt vmcnt(0)
-; SI-NEXT: v_or_b32_e32 v38, v5, v14
-; SI-NEXT: v_alignbit_b32 v5, v41, v44, 16
-; SI-NEXT: buffer_store_dword v5, off, s[0:3], s32 offset:300 ; 4-byte Folded Spill
-; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: buffer_load_dword v5, off, s[0:3], s32 offset:612 ; 4-byte Folded Reload
-; SI-NEXT: s_waitcnt vmcnt(0)
-; SI-NEXT: v_lshlrev_b32_e32 v14, 16, v5
-; SI-NEXT: buffer_load_dword v5, off, s[0:3], s32 offset:608 ; 4-byte Folded Reload
-; SI-NEXT: s_waitcnt vmcnt(0)
-; SI-NEXT: v_or_b32_e32 v35, v5, v14
-; SI-NEXT: buffer_load_dword v5, off, s[0:3], s32 offset:604 ; 4-byte Folded Reload
+; SI-NEXT: v_or_b32_e32 v48, v15, v14
+; SI-NEXT: buffer_load_dword v15, off, s[0:3], s32 offset:540 ; 4-byte Folded Reload
; SI-NEXT: v_lshlrev_b32_e32 v14, 16, v11
; SI-NEXT: s_waitcnt vmcnt(0)
-; SI-NEXT: v_or_b32_e32 v36, v5, v14
-; SI-NEXT: v_alignbit_b32 v5, v41, v44, 8
-; SI-NEXT: buffer_store_dword v5, off, s[0:3], s32 offset:656 ; 4-byte Folded Spill
+; SI-NEXT: v_or_b32_e32 v39, v15, v14
+; SI-NEXT: v_alignbit_b32 v14, v41, v44, 24
+; SI-NEXT: buffer_store_dword v14, off, s[0:3], s32 offset:300 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: buffer_load_dword v5, off, s[0:3], s32 offset:624 ; 4-byte Folded Reload
-; SI-NEXT: s_waitcnt vmcnt(0)
-; SI-NEXT: v_lshlrev_b32_e32 v14, 16, v5
-; SI-NEXT: buffer_load_dword v5, off, s[0:3], s32 offset:620 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v14, off, s[0:3], s32 offset:616 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v15, off, s[0:3], s32 offset:612 ; 4-byte Folded Reload
+; SI-NEXT: s_waitcnt vmcnt(1)
+; SI-NEXT: v_lshlrev_b32_e32 v14, 16, v14
; SI-NEXT: s_waitcnt vmcnt(0)
-; SI-NEXT: v_or_b32_e32 v30, v5, v14
-; SI-NEXT: buffer_load_dword v5, off, s[0:3], s32 offset:616 ; 4-byte Folded Reload
-; SI-NEXT: v_lshlrev_b32_e32 v14, 16, v10
+; SI-NEXT: v_or_b32_e32 v37, v15, v14
+; SI-NEXT: buffer_load_dword v15, off, s[0:3], s32 offset:600 ; 4-byte Folded Reload
+; SI-NEXT: v_lshlrev_b32_e32 v14, 16, v33
; SI-NEXT: s_waitcnt vmcnt(0)
-; SI-NEXT: v_or_b32_e32 v34, v5, v14
-; SI-NEXT: v_alignbit_b32 v5, v53, v54, 24
-; SI-NEXT: buffer_store_dword v5, off, s[0:3], s32 offset:312 ; 4-byte Folded Spill
+; SI-NEXT: v_or_b32_e32 v38, v15, v14
+; SI-NEXT: v_alignbit_b32 v14, v41, v44, 16
+; SI-NEXT: buffer_store_dword v14, off, s[0:3], s32 offset:320 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: buffer_load_dword v5, off, s[0:3], s32 offset:636 ; 4-byte Folded Reload
-; SI-NEXT: s_waitcnt vmcnt(0)
-; SI-NEXT: v_lshlrev_b32_e32 v14, 16, v5
-; SI-NEXT: buffer_load_dword v5, off, s[0:3], s32 offset:632 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v14, off, s[0:3], s32 offset:628 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v15, off, s[0:3], s32 offset:624 ; 4-byte Folded Reload
+; SI-NEXT: s_waitcnt vmcnt(1)
+; SI-NEXT: v_lshlrev_b32_e32 v14, 16, v14
; SI-NEXT: s_waitcnt vmcnt(0)
-; SI-NEXT: v_or_b32_e32 v28, v5, v14
-; SI-NEXT: buffer_load_dword v5, off, s[0:3], s32 offset:628 ; 4-byte Folded Reload
-; SI-NEXT: v_lshlrev_b32_e32 v14, 16, v8
+; SI-NEXT: v_or_b32_e32 v35, v15, v14
+; SI-NEXT: buffer_load_dword v15, off, s[0:3], s32 offset:620 ; 4-byte Folded Reload
+; SI-NEXT: v_lshlrev_b32_e32 v14, 16, v13
; SI-NEXT: s_waitcnt vmcnt(0)
-; SI-NEXT: v_or_b32_e32 v29, v5, v14
-; SI-NEXT: v_alignbit_b32 v5, v53, v54, 16
-; SI-NEXT: buffer_store_dword v5, off, s[0:3], s32 offset:344 ; 4-byte Folded Spill
+; SI-NEXT: v_or_b32_e32 v36, v15, v14
+; SI-NEXT: v_alignbit_b32 v14, v41, v44, 8
+; SI-NEXT: buffer_store_dword v14, off, s[0:3], s32 offset:348 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: buffer_load_dword v5, off, s[0:3], s32 offset:648 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v14, off, s[0:3], s32 offset:640 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v15, off, s[0:3], s32 offset:636 ; 4-byte Folded Reload
+; SI-NEXT: s_waitcnt vmcnt(1)
+; SI-NEXT: v_lshlrev_b32_e32 v14, 16, v14
; SI-NEXT: s_waitcnt vmcnt(0)
-; SI-NEXT: v_lshlrev_b32_e32 v14, 16, v5
-; SI-NEXT: buffer_load_dword v5, off, s[0:3], s32 offset:644 ; 4-byte Folded Reload
+; SI-NEXT: v_or_b32_e32 v30, v15, v14
+; SI-NEXT: buffer_load_dword v15, off, s[0:3], s32 offset:632 ; 4-byte Folded Reload
+; SI-NEXT: v_lshlrev_b32_e32 v14, 16, v12
; SI-NEXT: s_waitcnt vmcnt(0)
-; SI-NEXT: v_or_b32_e32 v26, v5, v14
-; SI-NEXT: buffer_load_dword v5, off, s[0:3], s32 offset:640 ; 4-byte Folded Reload
-; SI-NEXT: v_lshlrev_b32_e32 v14, 16, v7
+; SI-NEXT: v_or_b32_e32 v34, v15, v14
+; SI-NEXT: v_alignbit_b32 v14, v53, v54, 24
+; SI-NEXT: buffer_store_dword v14, off, s[0:3], s32 offset:332 ; 4-byte Folded Spill
+; SI-NEXT: s_waitcnt expcnt(0)
+; SI-NEXT: buffer_load_dword v14, off, s[0:3], s32 offset:648 ; 4-byte Folded Reload
; SI-NEXT: s_waitcnt vmcnt(0)
-; SI-NEXT: v_or_b32_e32 v27, v5, v14
-; SI-NEXT: v_alignbit_b32 v5, v53, v54, 8
-; SI-NEXT: buffer_store_dword v5, off, s[0:3], s32 offset:380 ; 4-byte Folded Spill
+; SI-NEXT: v_or_b32_e32 v28, v14, v9
+; SI-NEXT: buffer_load_dword v14, off, s[0:3], s32 offset:644 ; 4-byte Folded Reload
+; SI-NEXT: v_lshlrev_b32_e32 v9, 16, v55
+; SI-NEXT: s_waitcnt vmcnt(0)
+; SI-NEXT: v_or_b32_e32 v29, v14, v9
+; SI-NEXT: v_alignbit_b32 v9, v53, v54, 16
+; SI-NEXT: buffer_store_dword v9, off, s[0:3], s32 offset:360 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: buffer_load_dword v5, off, s[0:3], s32 offset:652 ; 4-byte Folded Reload
-; SI-NEXT: v_lshlrev_b32_e32 v14, 16, v56
-; SI-NEXT: buffer_load_dword v56, off, s[0:3], s32 offset:212 ; 4-byte Folded Reload
-; SI-NEXT: v_or_b32_e32 v24, v47, v14
-; SI-NEXT: buffer_load_dword v47, off, s[0:3], s32 offset:204 ; 4-byte Folded Reload
-; SI-NEXT: v_lshlrev_b32_e32 v14, 16, v6
+; SI-NEXT: v_lshlrev_b32_e32 v9, 16, v45
+; SI-NEXT: buffer_load_dword v45, off, s[0:3], s32 offset:216 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v14, off, s[0:3], s32 offset:652 ; 4-byte Folded Reload
+; SI-NEXT: v_or_b32_e32 v26, v43, v9
+; SI-NEXT: buffer_load_dword v43, off, s[0:3], s32 offset:212 ; 4-byte Folded Reload
; SI-NEXT: s_waitcnt vmcnt(2)
-; SI-NEXT: v_or_b32_e32 v25, v5, v14
-; SI-NEXT: v_alignbit_b32 v5, v52, v51, 24
-; SI-NEXT: buffer_store_dword v5, off, s[0:3], s32 offset:356 ; 4-byte Folded Spill
+; SI-NEXT: v_lshlrev_b32_e32 v9, 16, v45
+; SI-NEXT: s_waitcnt vmcnt(1)
+; SI-NEXT: v_or_b32_e32 v27, v14, v9
+; SI-NEXT: v_alignbit_b32 v9, v53, v54, 8
+; SI-NEXT: buffer_store_dword v9, off, s[0:3], s32 offset:396 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_lshlrev_b32_e32 v5, 16, v59
-; SI-NEXT: v_or_b32_e32 v22, v58, v5
-; SI-NEXT: s_waitcnt vmcnt(2)
-; SI-NEXT: v_lshlrev_b32_e32 v5, 16, v56
-; SI-NEXT: v_or_b32_e32 v23, v57, v5
-; SI-NEXT: v_alignbit_b32 v5, v52, v51, 16
-; SI-NEXT: buffer_store_dword v5, off, s[0:3], s32 offset:392 ; 4-byte Folded Spill
+; SI-NEXT: v_lshlrev_b32_e32 v9, 16, v56
+; SI-NEXT: v_or_b32_e32 v24, v47, v9
+; SI-NEXT: s_waitcnt vmcnt(1)
+; SI-NEXT: v_lshlrev_b32_e32 v9, 16, v43
+; SI-NEXT: v_or_b32_e32 v25, v10, v9
+; SI-NEXT: buffer_load_dword v10, off, s[0:3], s32 offset:208 ; 4-byte Folded Reload
+; SI-NEXT: v_alignbit_b32 v9, v52, v51, 24
+; SI-NEXT: buffer_store_dword v9, off, s[0:3], s32 offset:372 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_lshlrev_b32_e32 v5, 16, v62
-; SI-NEXT: v_or_b32_e32 v20, v61, v5
-; SI-NEXT: s_waitcnt vmcnt(2)
-; SI-NEXT: v_lshlrev_b32_e32 v5, 16, v47
-; SI-NEXT: v_or_b32_e32 v21, v60, v5
-; SI-NEXT: v_alignbit_b32 v5, v52, v51, 8
-; SI-NEXT: buffer_store_dword v5, off, s[0:3], s32 offset:428 ; 4-byte Folded Spill
-; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_lshlrev_b32_e32 v5, 16, v63
-; SI-NEXT: v_or_b32_e32 v18, v40, v5
-; SI-NEXT: buffer_load_dword v40, off, s[0:3], s32 offset:208 ; 4-byte Folded Reload
+; SI-NEXT: v_lshlrev_b32_e32 v9, 16, v59
+; SI-NEXT: v_or_b32_e32 v22, v58, v9
+; SI-NEXT: ; implicit-def: $vgpr56
+; SI-NEXT: ; implicit-def: $vgpr47
; SI-NEXT: ; implicit-def: $vgpr59
; SI-NEXT: ; implicit-def: $vgpr58
+; SI-NEXT: s_waitcnt vmcnt(1)
+; SI-NEXT: v_lshlrev_b32_e32 v9, 16, v10
+; SI-NEXT: v_or_b32_e32 v23, v57, v9
+; SI-NEXT: v_alignbit_b32 v9, v52, v51, 16
+; SI-NEXT: buffer_store_dword v9, off, s[0:3], s32 offset:408 ; 4-byte Folded Spill
+; SI-NEXT: s_waitcnt expcnt(0)
+; SI-NEXT: buffer_load_dword v9, off, s[0:3], s32 offset:224 ; 4-byte Folded Reload
; SI-NEXT: ; implicit-def: $vgpr57
-; SI-NEXT: ; implicit-def: $vgpr62
-; SI-NEXT: ; implicit-def: $vgpr61
-; SI-NEXT: ; implicit-def: $vgpr60
-; SI-NEXT: ; implicit-def: $vgpr63
; SI-NEXT: s_waitcnt vmcnt(0)
-; SI-NEXT: v_lshlrev_b32_e32 v5, 16, v40
-; SI-NEXT: v_or_b32_e32 v19, v55, v5
-; SI-NEXT: v_alignbit_b32 v5, v50, v49, 24
-; SI-NEXT: buffer_store_dword v5, off, s[0:3], s32 offset:408 ; 4-byte Folded Spill
+; SI-NEXT: v_lshlrev_b32_e32 v1, 16, v9
+; SI-NEXT: v_or_b32_e32 v21, v60, v1
+; SI-NEXT: v_alignbit_b32 v1, v52, v51, 8
+; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:444 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_lshlrev_b32_e32 v5, 16, v43
-; SI-NEXT: v_or_b32_e32 v16, v1, v5
-; SI-NEXT: v_lshlrev_b32_e32 v1, 16, v46
-; SI-NEXT: v_or_b32_e32 v17, v42, v1
+; SI-NEXT: v_lshlrev_b32_e32 v1, 16, v61
+; SI-NEXT: v_or_b32_e32 v18, v4, v1
+; SI-NEXT: v_lshlrev_b32_e32 v1, 16, v2
+; SI-NEXT: v_or_b32_e32 v19, v3, v1
+; SI-NEXT: v_alignbit_b32 v1, v50, v49, 24
+; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:424 ; 4-byte Folded Spill
+; SI-NEXT: s_waitcnt expcnt(0)
+; SI-NEXT: v_lshlrev_b32_e32 v1, 16, v5
+; SI-NEXT: v_or_b32_e32 v14, v63, v1
+; SI-NEXT: v_lshlrev_b32_e32 v1, 16, v42
+; SI-NEXT: v_or_b32_e32 v17, v62, v1
; SI-NEXT: v_alignbit_b32 v1, v50, v49, 16
-; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:444 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:460 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_lshlrev_b32_e32 v1, 16, v3
-; SI-NEXT: v_or_b32_e32 v14, v4, v1
-; SI-NEXT: v_lshlrev_b32_e32 v1, 16, v45
-; SI-NEXT: v_or_b32_e32 v15, v2, v1
+; SI-NEXT: v_lshlrev_b32_e32 v1, 16, v7
+; SI-NEXT: v_or_b32_e32 v3, v8, v1
+; SI-NEXT: v_lshlrev_b32_e32 v1, 16, v40
+; SI-NEXT: v_or_b32_e32 v15, v6, v1
; SI-NEXT: v_alignbit_b32 v1, v50, v49, 8
-; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:480 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:496 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_alignbit_b32 v1, v39, v48, 24
-; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:452 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:468 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_alignbit_b32 v1, v39, v48, 16
-; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:456 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:472 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_alignbit_b32 v1, v39, v48, 8
-; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:468 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:484 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_alignbit_b32 v1, v38, v37, 24
-; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:420 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:436 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_alignbit_b32 v1, v38, v37, 16
-; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:432 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:448 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_alignbit_b32 v1, v38, v37, 8
-; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:448 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:464 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_alignbit_b32 v1, v36, v35, 24
-; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:396 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:412 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_alignbit_b32 v1, v36, v35, 16
-; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:400 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:416 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_alignbit_b32 v1, v36, v35, 8
-; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:412 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:428 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_alignbit_b32 v1, v34, v30, 24
-; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:364 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:380 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_alignbit_b32 v1, v34, v30, 16
-; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:372 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:388 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_alignbit_b32 v1, v34, v30, 8
-; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:376 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:392 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_alignbit_b32 v1, v29, v28, 24
-; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:324 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:344 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_alignbit_b32 v1, v29, v28, 16
-; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:336 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:352 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_alignbit_b32 v1, v29, v28, 8
-; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:348 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:364 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_alignbit_b32 v1, v27, v26, 24
-; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:296 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:316 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_alignbit_b32 v1, v27, v26, 16
-; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:304 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:324 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_alignbit_b32 v1, v27, v26, 8
-; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:320 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:340 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_alignbit_b32 v1, v25, v24, 24
-; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:276 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:296 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_alignbit_b32 v1, v25, v24, 16
-; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:284 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:304 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_alignbit_b32 v1, v25, v24, 8
-; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:292 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:312 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_alignbit_b32 v1, v23, v22, 24
-; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:264 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:280 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_alignbit_b32 v1, v23, v22, 16
-; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:332 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:288 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_alignbit_b32 v1, v23, v22, 8
-; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:272 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:292 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_alignbit_b32 v1, v21, v20, 24
-; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:252 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:268 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_alignbit_b32 v1, v21, v20, 16
-; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:256 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:272 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_alignbit_b32 v1, v21, v20, 8
-; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:260 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:276 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_alignbit_b32 v1, v19, v18, 24
-; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:240 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:256 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_alignbit_b32 v1, v19, v18, 16
-; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:244 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:260 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_alignbit_b32 v1, v19, v18, 8
+; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:264 ; 4-byte Folded Spill
+; SI-NEXT: s_waitcnt expcnt(0)
+; SI-NEXT: v_alignbit_b32 v1, v17, v14, 24
+; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:244 ; 4-byte Folded Spill
+; SI-NEXT: s_waitcnt expcnt(0)
+; SI-NEXT: v_alignbit_b32 v1, v17, v14, 16
; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:248 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_alignbit_b32 v1, v17, v16, 24
-; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:228 ; 4-byte Folded Spill
+; SI-NEXT: v_alignbit_b32 v1, v17, v14, 8
+; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:252 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_alignbit_b32 v1, v17, v16, 16
+; SI-NEXT: v_alignbit_b32 v1, v15, v3, 24
; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:232 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_alignbit_b32 v1, v17, v16, 8
+; SI-NEXT: v_alignbit_b32 v1, v15, v3, 16
; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:236 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_alignbit_b32 v1, v15, v14, 24
-; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:216 ; 4-byte Folded Spill
-; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_alignbit_b32 v1, v15, v14, 16
-; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:220 ; 4-byte Folded Spill
-; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_alignbit_b32 v1, v15, v14, 8
-; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:224 ; 4-byte Folded Spill
+; SI-NEXT: v_alignbit_b32 v1, v15, v3, 8
+; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:240 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_lshrrev_b32_e32 v1, 8, v41
-; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:576 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:592 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_lshrrev_b32_e32 v1, 8, v53
-; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:568 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:584 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_lshrrev_b32_e32 v1, 8, v52
-; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:560 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:576 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_lshrrev_b32_e32 v1, 8, v50
-; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:552 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:568 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_lshrrev_b32_e32 v1, 8, v39
-; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:540 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:556 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_lshrrev_b32_e32 v1, 8, v38
-; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:528 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:544 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_lshrrev_b32_e32 v1, 8, v36
-; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:516 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:532 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_lshrrev_b32_e32 v1, 8, v34
-; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:504 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:520 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_lshrrev_b32_e32 v1, 8, v29
-; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:484 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:500 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_lshrrev_b32_e32 v1, 8, v27
-; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:460 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:476 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_lshrrev_b32_e32 v1, 8, v25
-; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:424 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:440 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_lshrrev_b32_e32 v1, 8, v23
-; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:384 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:400 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_lshrrev_b32_e32 v1, 8, v21
-; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:352 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:368 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_lshrrev_b32_e32 v1, 8, v19
-; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:316 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:336 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_lshrrev_b32_e32 v1, 8, v17
-; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:288 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:308 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_lshrrev_b32_e32 v1, 8, v15
-; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:268 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:284 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_bfe_u32 v1, v33, 8, 8
-; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:592 ; 4-byte Folded Spill
+; SI-NEXT: v_bfe_u32 v1, v16, 8, 8
+; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:608 ; 4-byte Folded Spill
+; SI-NEXT: s_waitcnt expcnt(0)
+; SI-NEXT: v_bfe_u32 v1, v46, 8, 8
+; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:604 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_bfe_u32 v1, v32, 8, 8
-; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:588 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:596 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_bfe_u32 v1, v31, 8, 8
+; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:588 ; 4-byte Folded Spill
+; SI-NEXT: s_waitcnt expcnt(0)
+; SI-NEXT: v_bfe_u32 v1, v11, 8, 8
+; SI-NEXT: v_mov_b32_e32 v11, v12
+; SI-NEXT: v_mov_b32_e32 v12, v13
+; SI-NEXT: v_mov_b32_e32 v13, v31
+; SI-NEXT: v_mov_b32_e32 v31, v32
+; SI-NEXT: v_mov_b32_e32 v32, v33
; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:580 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_bfe_u32 v1, v12, 8, 8
+; SI-NEXT: v_bfe_u32 v1, v32, 8, 8
; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:572 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_bfe_u32 v1, v9, 8, 8
-; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:564 ; 4-byte Folded Spill
-; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_bfe_u32 v1, v13, 8, 8
-; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:556 ; 4-byte Folded Spill
+; SI-NEXT: v_bfe_u32 v1, v12, 8, 8
+; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:560 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_bfe_u32 v1, v11, 8, 8
-; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:544 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:548 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_bfe_u32 v1, v10, 8, 8
-; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:532 ; 4-byte Folded Spill
+; SI-NEXT: v_bfe_u32 v1, v55, 8, 8
+; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:536 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_bfe_u32 v1, v8, 8, 8
-; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:520 ; 4-byte Folded Spill
+; SI-NEXT: v_bfe_u32 v1, v45, 8, 8
+; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:516 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_bfe_u32 v1, v7, 8, 8
-; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:500 ; 4-byte Folded Spill
+; SI-NEXT: v_bfe_u32 v1, v43, 8, 8
+; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:504 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_bfe_u32 v1, v6, 8, 8
+; SI-NEXT: v_bfe_u32 v1, v10, 8, 8
; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:488 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_bfe_u32 v1, v56, 8, 8
-; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:472 ; 4-byte Folded Spill
+; SI-NEXT: v_bfe_u32 v1, v9, 8, 8
+; SI-NEXT: v_mov_b32_e32 v55, v40
+; SI-NEXT: v_mov_b32_e32 v40, v42
+; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:452 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_bfe_u32 v1, v47, 8, 8
-; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:436 ; 4-byte Folded Spill
+; SI-NEXT: v_bfe_u32 v1, v2, 8, 8
+; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:420 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_bfe_u32 v1, v40, 8, 8
-; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:404 ; 4-byte Folded Spill
-; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_bfe_u32 v1, v46, 8, 8
-; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:368 ; 4-byte Folded Spill
-; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_bfe_u32 v1, v45, 8, 8
-; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:328 ; 4-byte Folded Spill
-; SI-NEXT: ; implicit-def: $vgpr1
-; SI-NEXT: ; kill: killed $vgpr1
-; SI-NEXT: ; implicit-def: $vgpr1
-; SI-NEXT: ; kill: killed $vgpr1
-; SI-NEXT: ; implicit-def: $vgpr1
-; SI-NEXT: ; kill: killed $vgpr1
-; SI-NEXT: ; implicit-def: $vgpr1
-; SI-NEXT: ; kill: killed $vgpr1
+; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:384 ; 4-byte Folded Spill
; SI-NEXT: ; implicit-def: $vgpr1
; SI-NEXT: ; kill: killed $vgpr1
; SI-NEXT: ; implicit-def: $vgpr1
@@ -191439,202 +191482,209 @@ define <128 x i8> @bitcast_v64f16_to_v128i8(<64 x half> %a, i32 %b) {
; SI-NEXT: ; implicit-def: $vgpr1
; SI-NEXT: ; kill: killed $vgpr1
; SI-NEXT: ; implicit-def: $vgpr1
-; SI-NEXT: ; implicit-def: $vgpr56
-; SI-NEXT: ; implicit-def: $vgpr47
-; SI-NEXT: ; kill: killed $vgpr1
-; SI-NEXT: ; implicit-def: $vgpr40
-; SI-NEXT: ; implicit-def: $vgpr55
+; SI-NEXT: v_mov_b32_e32 v33, v46
+; SI-NEXT: v_mov_b32_e32 v46, v16
+; SI-NEXT: v_mov_b32_e32 v16, v3
+; SI-NEXT: v_bfe_u32 v42, v55, 8, 8
+; SI-NEXT: ; implicit-def: $vgpr9
+; SI-NEXT: ; implicit-def: $vgpr45
; SI-NEXT: ; implicit-def: $vgpr43
+; SI-NEXT: ; kill: killed $vgpr1
+; SI-NEXT: ; implicit-def: $vgpr10
+; SI-NEXT: ; implicit-def: $vgpr2
; SI-NEXT: ; implicit-def: $vgpr1
-; SI-NEXT: ; implicit-def: $vgpr42
-; SI-NEXT: ; implicit-def: $vgpr3
+; SI-NEXT: ; implicit-def: $vgpr60
+; SI-NEXT: ; implicit-def: $vgpr61
; SI-NEXT: ; implicit-def: $vgpr4
-; SI-NEXT: ; implicit-def: $vgpr2
+; SI-NEXT: ; implicit-def: $vgpr3
+; SI-NEXT: ; implicit-def: $vgpr5
+; SI-NEXT: ; implicit-def: $vgpr63
+; SI-NEXT: ; implicit-def: $vgpr62
+; SI-NEXT: ; implicit-def: $vgpr7
+; SI-NEXT: ; implicit-def: $vgpr8
+; SI-NEXT: ; implicit-def: $vgpr6
; SI-NEXT: .LBB94_2: ; %Flow
-; SI-NEXT: s_or_saveexec_b64 s[4:5], s[4:5]
-; SI-NEXT: buffer_load_dword v5, off, s[0:3], s32 offset:656 ; 4-byte Folded Reload
-; SI-NEXT: s_xor_b64 exec, exec, s[4:5]
+; SI-NEXT: s_andn2_saveexec_b64 s[4:5], s[4:5]
; SI-NEXT: s_cbranch_execz .LBB94_4
; SI-NEXT: ; %bb.3: ; %cmp.true
-; SI-NEXT: v_cvt_f32_f16_e32 v3, v3
+; SI-NEXT: v_cvt_f32_f16_e32 v7, v7
+; SI-NEXT: v_cvt_f32_f16_e32 v8, v8
+; SI-NEXT: v_cvt_f32_f16_e32 v6, v6
+; SI-NEXT: v_cvt_f32_f16_e32 v5, v5
+; SI-NEXT: v_add_f32_e32 v7, 0x38000000, v7
+; SI-NEXT: v_add_f32_e32 v8, 0x38000000, v8
+; SI-NEXT: v_cvt_f16_f32_e32 v7, v7
+; SI-NEXT: v_cvt_f16_f32_e32 v8, v8
+; SI-NEXT: v_add_f32_e32 v6, 0x38000000, v6
+; SI-NEXT: v_cvt_f16_f32_e32 v6, v6
+; SI-NEXT: v_lshlrev_b32_e32 v7, 16, v7
+; SI-NEXT: v_or_b32_e32 v16, v8, v7
+; SI-NEXT: v_cvt_f32_f16_e32 v7, v55
+; SI-NEXT: v_add_f32_e32 v5, 0x38000000, v5
+; SI-NEXT: v_cvt_f16_f32_e32 v5, v5
; SI-NEXT: v_cvt_f32_f16_e32 v4, v4
+; SI-NEXT: v_add_f32_e32 v7, 0x38000000, v7
+; SI-NEXT: v_cvt_f16_f32_e32 v55, v7
+; SI-NEXT: v_lshlrev_b32_e32 v5, 16, v5
+; SI-NEXT: v_add_f32_e32 v4, 0x38000000, v4
+; SI-NEXT: v_cvt_f16_f32_e32 v4, v4
+; SI-NEXT: v_lshlrev_b32_e32 v7, 16, v55
+; SI-NEXT: v_or_b32_e32 v15, v6, v7
+; SI-NEXT: v_cvt_f32_f16_e32 v6, v63
; SI-NEXT: v_cvt_f32_f16_e32 v2, v2
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_cvt_f32_f16_e32 v1, v1
-; SI-NEXT: v_add_f32_e32 v3, 0x38000000, v3
-; SI-NEXT: v_add_f32_e32 v4, 0x38000000, v4
-; SI-NEXT: v_cvt_f16_f32_e32 v3, v3
-; SI-NEXT: v_cvt_f16_f32_e32 v4, v4
+; SI-NEXT: v_cvt_f32_f16_e32 v3, v3
+; SI-NEXT: v_add_f32_e32 v6, 0x38000000, v6
+; SI-NEXT: v_cvt_f16_f32_e32 v6, v6
; SI-NEXT: v_add_f32_e32 v2, 0x38000000, v2
; SI-NEXT: v_cvt_f16_f32_e32 v2, v2
-; SI-NEXT: v_lshlrev_b32_e32 v3, 16, v3
-; SI-NEXT: v_or_b32_e32 v14, v4, v3
-; SI-NEXT: v_cvt_f32_f16_e32 v3, v45
; SI-NEXT: v_add_f32_e32 v1, 0x38000000, v1
+; SI-NEXT: v_or_b32_e32 v14, v6, v5
+; SI-NEXT: v_cvt_f32_f16_e32 v5, v40
+; SI-NEXT: v_cvt_f32_f16_e32 v6, v62
; SI-NEXT: v_cvt_f16_f32_e32 v1, v1
-; SI-NEXT: v_cvt_f32_f16_e32 v4, v47
-; SI-NEXT: v_add_f32_e32 v3, 0x38000000, v3
-; SI-NEXT: v_cvt_f16_f32_e32 v45, v3
-; SI-NEXT: v_lshlrev_b32_e32 v3, 16, v45
-; SI-NEXT: v_or_b32_e32 v15, v2, v3
-; SI-NEXT: v_cvt_f32_f16_e32 v2, v43
-; SI-NEXT: buffer_load_dword v3, off, s[0:3], s32 offset:212 ; 4-byte Folded Reload
-; SI-NEXT: v_add_f32_e32 v2, 0x38000000, v2
-; SI-NEXT: v_cvt_f16_f32_e32 v2, v2
; SI-NEXT: v_lshlrev_b32_e32 v2, 16, v2
-; SI-NEXT: v_or_b32_e32 v16, v1, v2
-; SI-NEXT: v_cvt_f32_f16_e32 v1, v46
-; SI-NEXT: v_cvt_f32_f16_e32 v2, v42
-; SI-NEXT: v_add_f32_e32 v1, 0x38000000, v1
-; SI-NEXT: v_cvt_f16_f32_e32 v46, v1
-; SI-NEXT: v_add_f32_e32 v1, 0x38000000, v2
-; SI-NEXT: v_cvt_f16_f32_e32 v1, v1
-; SI-NEXT: v_lshlrev_b32_e32 v2, 16, v46
-; SI-NEXT: v_or_b32_e32 v17, v1, v2
-; SI-NEXT: v_cvt_f32_f16_e32 v1, v63
-; SI-NEXT: v_cvt_f32_f16_e32 v2, v40
-; SI-NEXT: v_add_f32_e32 v1, 0x38000000, v1
-; SI-NEXT: v_cvt_f16_f32_e32 v1, v1
-; SI-NEXT: v_add_f32_e32 v2, 0x38000000, v2
-; SI-NEXT: v_cvt_f16_f32_e32 v2, v2
-; SI-NEXT: v_lshlrev_b32_e32 v1, 16, v1
-; SI-NEXT: v_or_b32_e32 v18, v2, v1
-; SI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:208 ; 4-byte Folded Reload
-; SI-NEXT: v_cvt_f32_f16_e32 v2, v55
+; SI-NEXT: v_add_f32_e32 v5, 0x38000000, v5
+; SI-NEXT: v_cvt_f16_f32_e32 v40, v5
+; SI-NEXT: v_add_f32_e32 v5, 0x38000000, v6
+; SI-NEXT: v_cvt_f16_f32_e32 v5, v5
+; SI-NEXT: v_or_b32_e32 v20, v1, v2
+; SI-NEXT: v_lshlrev_b32_e32 v6, 16, v40
+; SI-NEXT: v_add_f32_e32 v3, 0x38000000, v3
+; SI-NEXT: v_or_b32_e32 v17, v5, v6
+; SI-NEXT: v_cvt_f32_f16_e32 v5, v61
+; SI-NEXT: v_cvt_f16_f32_e32 v3, v3
+; SI-NEXT: v_cvt_f32_f16_e32 v2, v60
+; SI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:224 ; 4-byte Folded Reload
+; SI-NEXT: v_add_f32_e32 v5, 0x38000000, v5
+; SI-NEXT: v_cvt_f16_f32_e32 v5, v5
+; SI-NEXT: v_lshlrev_b32_e32 v5, 16, v5
+; SI-NEXT: v_or_b32_e32 v18, v4, v5
+; SI-NEXT: buffer_load_dword v4, off, s[0:3], s32 offset:204 ; 4-byte Folded Reload
; SI-NEXT: s_waitcnt vmcnt(1)
-; SI-NEXT: v_cvt_f32_f16_e32 v3, v3
-; SI-NEXT: s_waitcnt vmcnt(0)
; SI-NEXT: v_cvt_f32_f16_e32 v1, v1
; SI-NEXT: v_add_f32_e32 v1, 0x38000000, v1
-; SI-NEXT: v_cvt_f16_f32_e32 v40, v1
+; SI-NEXT: v_cvt_f16_f32_e32 v42, v1
; SI-NEXT: v_add_f32_e32 v1, 0x38000000, v2
; SI-NEXT: v_cvt_f16_f32_e32 v1, v1
-; SI-NEXT: v_lshlrev_b32_e32 v2, 16, v40
-; SI-NEXT: v_or_b32_e32 v19, v1, v2
-; SI-NEXT: v_cvt_f32_f16_e32 v1, v62
-; SI-NEXT: v_cvt_f32_f16_e32 v2, v61
-; SI-NEXT: v_add_f32_e32 v1, 0x38000000, v1
-; SI-NEXT: v_cvt_f16_f32_e32 v1, v1
-; SI-NEXT: v_add_f32_e32 v2, 0x38000000, v2
-; SI-NEXT: v_cvt_f16_f32_e32 v2, v2
-; SI-NEXT: v_lshlrev_b32_e32 v1, 16, v1
-; SI-NEXT: v_or_b32_e32 v20, v2, v1
-; SI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:204 ; 4-byte Folded Reload
-; SI-NEXT: v_cvt_f32_f16_e32 v2, v60
; SI-NEXT: s_waitcnt vmcnt(0)
-; SI-NEXT: v_cvt_f32_f16_e32 v1, v1
-; SI-NEXT: v_add_f32_e32 v1, 0x38000000, v1
-; SI-NEXT: v_cvt_f16_f32_e32 v55, v1
-; SI-NEXT: v_add_f32_e32 v1, 0x38000000, v2
-; SI-NEXT: v_cvt_f16_f32_e32 v1, v1
-; SI-NEXT: v_lshlrev_b32_e32 v2, 16, v55
+; SI-NEXT: v_cvt_f32_f16_e32 v4, v4
+; SI-NEXT: v_lshlrev_b32_e32 v2, 16, v42
; SI-NEXT: v_or_b32_e32 v21, v1, v2
+; SI-NEXT: v_add_f32_e32 v4, 0x38000000, v4
+; SI-NEXT: v_cvt_f16_f32_e32 v5, v4
; SI-NEXT: v_cvt_f32_f16_e32 v1, v59
; SI-NEXT: v_cvt_f32_f16_e32 v2, v58
+; SI-NEXT: v_lshlrev_b32_e32 v4, 16, v5
+; SI-NEXT: v_or_b32_e32 v19, v3, v4
+; SI-NEXT: buffer_load_dword v3, off, s[0:3], s32 offset:208 ; 4-byte Folded Reload
; SI-NEXT: v_add_f32_e32 v1, 0x38000000, v1
; SI-NEXT: v_cvt_f16_f32_e32 v1, v1
; SI-NEXT: v_add_f32_e32 v2, 0x38000000, v2
; SI-NEXT: v_cvt_f16_f32_e32 v2, v2
+; SI-NEXT: v_cvt_f32_f16_e32 v4, v47
; SI-NEXT: v_lshlrev_b32_e32 v1, 16, v1
; SI-NEXT: v_or_b32_e32 v22, v2, v1
; SI-NEXT: v_cvt_f32_f16_e32 v1, v57
-; SI-NEXT: v_add_f32_e32 v2, 0x38000000, v3
-; SI-NEXT: v_cvt_f32_f16_e32 v3, v56
-; SI-NEXT: v_cvt_f16_f32_e32 v42, v2
; SI-NEXT: v_add_f32_e32 v1, 0x38000000, v1
; SI-NEXT: v_cvt_f16_f32_e32 v1, v1
+; SI-NEXT: s_waitcnt vmcnt(0)
+; SI-NEXT: v_cvt_f32_f16_e32 v3, v3
+; SI-NEXT: v_add_f32_e32 v2, 0x38000000, v3
+; SI-NEXT: v_cvt_f32_f16_e32 v3, v56
+; SI-NEXT: v_cvt_f16_f32_e32 v6, v2
; SI-NEXT: v_add_f32_e32 v2, 0x38000000, v3
-; SI-NEXT: v_lshlrev_b32_e32 v3, 16, v42
; SI-NEXT: v_cvt_f16_f32_e32 v2, v2
+; SI-NEXT: v_lshlrev_b32_e32 v3, 16, v6
; SI-NEXT: v_or_b32_e32 v23, v1, v3
; SI-NEXT: v_add_f32_e32 v3, 0x38000000, v4
-; SI-NEXT: buffer_load_dword v4, off, s[0:3], s32 offset:652 ; 4-byte Folded Reload
-; SI-NEXT: v_cvt_f16_f32_e32 v3, v3
; SI-NEXT: v_lshlrev_b32_e32 v1, 16, v2
-; SI-NEXT: v_cvt_f32_f16_e32 v2, v6
+; SI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:212 ; 4-byte Folded Reload
+; SI-NEXT: v_cvt_f32_f16_e32 v4, v10
+; SI-NEXT: v_cvt_f16_f32_e32 v3, v3
; SI-NEXT: v_or_b32_e32 v24, v3, v1
-; SI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:648 ; 4-byte Folded Reload
-; SI-NEXT: v_add_f32_e32 v2, 0x38000000, v2
-; SI-NEXT: v_cvt_f16_f32_e32 v6, v2
-; SI-NEXT: v_lshlrev_b32_e32 v3, 16, v6
-; SI-NEXT: s_waitcnt vmcnt(1)
-; SI-NEXT: v_cvt_f32_f16_e32 v4, v4
-; SI-NEXT: v_add_f32_e32 v2, 0x38000000, v4
-; SI-NEXT: v_cvt_f16_f32_e32 v2, v2
-; SI-NEXT: buffer_load_dword v4, off, s[0:3], s32 offset:632 ; 4-byte Folded Reload
-; SI-NEXT: v_or_b32_e32 v25, v2, v3
-; SI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:644 ; 4-byte Folded Reload
-; SI-NEXT: v_cvt_f32_f16_e32 v3, v7
-; SI-NEXT: s_waitcnt vmcnt(2)
-; SI-NEXT: v_cvt_f32_f16_e32 v1, v1
-; SI-NEXT: v_add_f32_e32 v3, 0x38000000, v3
-; SI-NEXT: v_cvt_f16_f32_e32 v7, v3
-; SI-NEXT: buffer_load_dword v3, off, s[0:3], s32 offset:640 ; 4-byte Folded Reload
+; SI-NEXT: v_cvt_f32_f16_e32 v1, v45
; SI-NEXT: v_add_f32_e32 v1, 0x38000000, v1
; SI-NEXT: v_cvt_f16_f32_e32 v1, v1
; SI-NEXT: v_lshlrev_b32_e32 v1, 16, v1
-; SI-NEXT: s_waitcnt vmcnt(2)
-; SI-NEXT: v_cvt_f32_f16_e32 v4, v4
-; SI-NEXT: s_waitcnt vmcnt(1)
+; SI-NEXT: s_waitcnt vmcnt(0)
; SI-NEXT: v_cvt_f32_f16_e32 v2, v2
-; SI-NEXT: v_add_f32_e32 v4, 0x38000000, v4
-; SI-NEXT: v_cvt_f16_f32_e32 v4, v4
+; SI-NEXT: v_add_f32_e32 v2, 0x38000000, v2
+; SI-NEXT: v_cvt_f16_f32_e32 v7, v2
+; SI-NEXT: v_add_f32_e32 v2, 0x38000000, v4
+; SI-NEXT: buffer_load_dword v4, off, s[0:3], s32 offset:648 ; 4-byte Folded Reload
+; SI-NEXT: v_cvt_f16_f32_e32 v2, v2
+; SI-NEXT: v_lshlrev_b32_e32 v3, 16, v7
+; SI-NEXT: v_or_b32_e32 v25, v2, v3
+; SI-NEXT: buffer_load_dword v3, off, s[0:3], s32 offset:216 ; 4-byte Folded Reload
+; SI-NEXT: v_cvt_f32_f16_e32 v2, v43
; SI-NEXT: v_add_f32_e32 v2, 0x38000000, v2
; SI-NEXT: v_cvt_f16_f32_e32 v2, v2
; SI-NEXT: v_or_b32_e32 v26, v2, v1
-; SI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:636 ; 4-byte Folded Reload
-; SI-NEXT: v_lshlrev_b32_e32 v1, 16, v7
-; SI-NEXT: s_waitcnt vmcnt(0)
-; SI-NEXT: v_cvt_f32_f16_e32 v2, v2
+; SI-NEXT: v_cvt_f32_f16_e32 v2, v9
; SI-NEXT: v_add_f32_e32 v2, 0x38000000, v2
; SI-NEXT: v_cvt_f16_f32_e32 v2, v2
; SI-NEXT: v_lshlrev_b32_e32 v2, 16, v2
+; SI-NEXT: s_waitcnt vmcnt(1)
+; SI-NEXT: v_cvt_f32_f16_e32 v4, v4
+; SI-NEXT: v_add_f32_e32 v4, 0x38000000, v4
+; SI-NEXT: v_cvt_f16_f32_e32 v4, v4
; SI-NEXT: v_or_b32_e32 v28, v4, v2
-; SI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:628 ; 4-byte Folded Reload
-; SI-NEXT: buffer_load_dword v4, off, s[0:3], s32 offset:620 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:644 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v4, off, s[0:3], s32 offset:636 ; 4-byte Folded Reload
+; SI-NEXT: s_waitcnt vmcnt(2)
; SI-NEXT: v_cvt_f32_f16_e32 v3, v3
; SI-NEXT: v_add_f32_e32 v3, 0x38000000, v3
-; SI-NEXT: v_cvt_f16_f32_e32 v3, v3
-; SI-NEXT: v_or_b32_e32 v27, v3, v1
-; SI-NEXT: buffer_load_dword v3, off, s[0:3], s32 offset:624 ; 4-byte Folded Reload
-; SI-NEXT: v_cvt_f32_f16_e32 v1, v8
-; SI-NEXT: v_add_f32_e32 v1, 0x38000000, v1
-; SI-NEXT: v_cvt_f16_f32_e32 v8, v1
+; SI-NEXT: v_cvt_f16_f32_e32 v8, v3
+; SI-NEXT: buffer_load_dword v3, off, s[0:3], s32 offset:652 ; 4-byte Folded Reload
+; SI-NEXT: v_lshlrev_b32_e32 v1, 16, v8
; SI-NEXT: s_waitcnt vmcnt(2)
; SI-NEXT: v_cvt_f32_f16_e32 v2, v2
; SI-NEXT: s_waitcnt vmcnt(1)
-; SI-NEXT: v_cvt_f32_f16_e32 v4, v4
-; SI-NEXT: v_add_f32_e32 v1, 0x38000000, v2
-; SI-NEXT: v_cvt_f16_f32_e32 v1, v1
+; SI-NEXT: v_cvt_f32_f16_e32 v4, v4
+; SI-NEXT: s_waitcnt vmcnt(0)
+; SI-NEXT: v_cvt_f32_f16_e32 v3, v3
+; SI-NEXT: v_add_f32_e32 v3, 0x38000000, v3
+; SI-NEXT: v_cvt_f16_f32_e32 v3, v3
+; SI-NEXT: v_or_b32_e32 v27, v3, v1
+; SI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:220 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v3, off, s[0:3], s32 offset:640 ; 4-byte Folded Reload
+; SI-NEXT: s_waitcnt vmcnt(1)
+; SI-NEXT: v_cvt_f32_f16_e32 v1, v1
; SI-NEXT: s_waitcnt vmcnt(0)
; SI-NEXT: v_cvt_f32_f16_e32 v3, v3
+; SI-NEXT: v_add_f32_e32 v1, 0x38000000, v1
+; SI-NEXT: v_cvt_f16_f32_e32 v9, v1
+; SI-NEXT: v_add_f32_e32 v1, 0x38000000, v2
+; SI-NEXT: v_cvt_f16_f32_e32 v1, v1
; SI-NEXT: v_add_f32_e32 v2, 0x38000000, v3
-; SI-NEXT: v_lshlrev_b32_e32 v3, 16, v8
+; SI-NEXT: v_lshlrev_b32_e32 v3, 16, v9
+; SI-NEXT: v_cvt_f16_f32_e32 v2, v2
; SI-NEXT: v_or_b32_e32 v29, v1, v3
; SI-NEXT: v_add_f32_e32 v3, 0x38000000, v4
-; SI-NEXT: buffer_load_dword v4, off, s[0:3], s32 offset:616 ; 4-byte Folded Reload
-; SI-NEXT: v_cvt_f16_f32_e32 v2, v2
+; SI-NEXT: buffer_load_dword v4, off, s[0:3], s32 offset:632 ; 4-byte Folded Reload
; SI-NEXT: v_cvt_f16_f32_e32 v3, v3
; SI-NEXT: v_lshlrev_b32_e32 v1, 16, v2
-; SI-NEXT: v_cvt_f32_f16_e32 v2, v10
+; SI-NEXT: v_cvt_f32_f16_e32 v2, v11
; SI-NEXT: v_or_b32_e32 v30, v3, v1
-; SI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:612 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:628 ; 4-byte Folded Reload
; SI-NEXT: v_add_f32_e32 v2, 0x38000000, v2
-; SI-NEXT: v_cvt_f16_f32_e32 v10, v2
-; SI-NEXT: v_lshlrev_b32_e32 v3, 16, v10
+; SI-NEXT: v_cvt_f16_f32_e32 v11, v2
+; SI-NEXT: v_lshlrev_b32_e32 v3, 16, v11
; SI-NEXT: s_waitcnt vmcnt(1)
; SI-NEXT: v_cvt_f32_f16_e32 v4, v4
; SI-NEXT: v_add_f32_e32 v2, 0x38000000, v4
; SI-NEXT: v_cvt_f16_f32_e32 v2, v2
-; SI-NEXT: buffer_load_dword v4, off, s[0:3], s32 offset:596 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v4, off, s[0:3], s32 offset:612 ; 4-byte Folded Reload
; SI-NEXT: v_or_b32_e32 v34, v2, v3
-; SI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:608 ; 4-byte Folded Reload
-; SI-NEXT: v_cvt_f32_f16_e32 v3, v11
+; SI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:624 ; 4-byte Folded Reload
+; SI-NEXT: v_cvt_f32_f16_e32 v3, v12
; SI-NEXT: s_waitcnt vmcnt(2)
; SI-NEXT: v_cvt_f32_f16_e32 v1, v1
; SI-NEXT: v_add_f32_e32 v3, 0x38000000, v3
-; SI-NEXT: v_cvt_f16_f32_e32 v11, v3
-; SI-NEXT: buffer_load_dword v3, off, s[0:3], s32 offset:604 ; 4-byte Folded Reload
+; SI-NEXT: v_cvt_f16_f32_e32 v12, v3
+; SI-NEXT: buffer_load_dword v3, off, s[0:3], s32 offset:620 ; 4-byte Folded Reload
; SI-NEXT: v_add_f32_e32 v1, 0x38000000, v1
; SI-NEXT: v_cvt_f16_f32_e32 v1, v1
; SI-NEXT: v_lshlrev_b32_e32 v1, 16, v1
@@ -191647,24 +191697,24 @@ define <128 x i8> @bitcast_v64f16_to_v128i8(<64 x half> %a, i32 %b) {
; SI-NEXT: v_add_f32_e32 v2, 0x38000000, v2
; SI-NEXT: v_cvt_f16_f32_e32 v2, v2
; SI-NEXT: v_or_b32_e32 v35, v2, v1
-; SI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:600 ; 4-byte Folded Reload
-; SI-NEXT: v_lshlrev_b32_e32 v1, 16, v11
+; SI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:616 ; 4-byte Folded Reload
+; SI-NEXT: v_lshlrev_b32_e32 v1, 16, v12
; SI-NEXT: s_waitcnt vmcnt(0)
; SI-NEXT: v_cvt_f32_f16_e32 v2, v2
; SI-NEXT: v_add_f32_e32 v2, 0x38000000, v2
; SI-NEXT: v_cvt_f16_f32_e32 v2, v2
; SI-NEXT: v_lshlrev_b32_e32 v2, 16, v2
; SI-NEXT: v_or_b32_e32 v37, v4, v2
-; SI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:584 ; 4-byte Folded Reload
-; SI-NEXT: buffer_load_dword v4, off, s[0:3], s32 offset:536 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:600 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v4, off, s[0:3], s32 offset:552 ; 4-byte Folded Reload
; SI-NEXT: v_cvt_f32_f16_e32 v3, v3
; SI-NEXT: v_add_f32_e32 v3, 0x38000000, v3
; SI-NEXT: v_cvt_f16_f32_e32 v3, v3
; SI-NEXT: v_or_b32_e32 v36, v3, v1
-; SI-NEXT: buffer_load_dword v3, off, s[0:3], s32 offset:548 ; 4-byte Folded Reload
-; SI-NEXT: v_cvt_f32_f16_e32 v1, v13
+; SI-NEXT: buffer_load_dword v3, off, s[0:3], s32 offset:564 ; 4-byte Folded Reload
+; SI-NEXT: v_cvt_f32_f16_e32 v1, v32
; SI-NEXT: v_add_f32_e32 v1, 0x38000000, v1
-; SI-NEXT: v_cvt_f16_f32_e32 v13, v1
+; SI-NEXT: v_cvt_f16_f32_e32 v32, v1
; SI-NEXT: s_waitcnt vmcnt(2)
; SI-NEXT: v_cvt_f32_f16_e32 v2, v2
; SI-NEXT: s_waitcnt vmcnt(1)
@@ -191674,34 +191724,36 @@ define <128 x i8> @bitcast_v64f16_to_v128i8(<64 x half> %a, i32 %b) {
; SI-NEXT: s_waitcnt vmcnt(0)
; SI-NEXT: v_cvt_f32_f16_e32 v3, v3
; SI-NEXT: v_add_f32_e32 v2, 0x38000000, v3
-; SI-NEXT: v_lshlrev_b32_e32 v3, 16, v13
+; SI-NEXT: v_cvt_f16_f32_e32 v2, v2
+; SI-NEXT: v_lshlrev_b32_e32 v3, 16, v32
; SI-NEXT: v_or_b32_e32 v38, v1, v3
; SI-NEXT: v_add_f32_e32 v3, 0x38000000, v4
-; SI-NEXT: buffer_load_dword v4, off, s[0:3], s32 offset:524 ; 4-byte Folded Reload
-; SI-NEXT: v_cvt_f16_f32_e32 v2, v2
-; SI-NEXT: v_cvt_f16_f32_e32 v3, v3
; SI-NEXT: v_lshlrev_b32_e32 v1, 16, v2
-; SI-NEXT: v_cvt_f32_f16_e32 v2, v9
+; SI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:228 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v4, off, s[0:3], s32 offset:540 ; 4-byte Folded Reload
+; SI-NEXT: v_cvt_f16_f32_e32 v3, v3
; SI-NEXT: v_or_b32_e32 v48, v3, v1
-; SI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:512 ; 4-byte Folded Reload
-; SI-NEXT: v_add_f32_e32 v2, 0x38000000, v2
-; SI-NEXT: v_cvt_f16_f32_e32 v9, v2
-; SI-NEXT: v_lshlrev_b32_e32 v3, 16, v9
+; SI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:528 ; 4-byte Folded Reload
+; SI-NEXT: s_waitcnt vmcnt(2)
+; SI-NEXT: v_cvt_f32_f16_e32 v2, v2
; SI-NEXT: s_waitcnt vmcnt(1)
; SI-NEXT: v_cvt_f32_f16_e32 v4, v4
+; SI-NEXT: v_add_f32_e32 v2, 0x38000000, v2
+; SI-NEXT: v_cvt_f16_f32_e32 v10, v2
; SI-NEXT: v_add_f32_e32 v2, 0x38000000, v4
; SI-NEXT: v_cvt_f16_f32_e32 v2, v2
-; SI-NEXT: buffer_load_dword v4, off, s[0:3], s32 offset:476 ; 4-byte Folded Reload
-; SI-NEXT: v_or_b32_e32 v39, v2, v3
-; SI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:508 ; 4-byte Folded Reload
-; SI-NEXT: v_cvt_f32_f16_e32 v3, v12
-; SI-NEXT: s_waitcnt vmcnt(2)
+; SI-NEXT: s_waitcnt vmcnt(0)
; SI-NEXT: v_cvt_f32_f16_e32 v1, v1
-; SI-NEXT: v_add_f32_e32 v3, 0x38000000, v3
-; SI-NEXT: v_cvt_f16_f32_e32 v12, v3
-; SI-NEXT: buffer_load_dword v3, off, s[0:3], s32 offset:496 ; 4-byte Folded Reload
+; SI-NEXT: v_lshlrev_b32_e32 v3, 16, v10
+; SI-NEXT: buffer_load_dword v4, off, s[0:3], s32 offset:492 ; 4-byte Folded Reload
+; SI-NEXT: v_or_b32_e32 v39, v2, v3
+; SI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:524 ; 4-byte Folded Reload
+; SI-NEXT: v_cvt_f32_f16_e32 v3, v13
; SI-NEXT: v_add_f32_e32 v1, 0x38000000, v1
; SI-NEXT: v_cvt_f16_f32_e32 v1, v1
+; SI-NEXT: v_add_f32_e32 v3, 0x38000000, v3
+; SI-NEXT: v_cvt_f16_f32_e32 v13, v3
+; SI-NEXT: buffer_load_dword v3, off, s[0:3], s32 offset:512 ; 4-byte Folded Reload
; SI-NEXT: v_lshlrev_b32_e32 v1, 16, v1
; SI-NEXT: s_waitcnt vmcnt(2)
; SI-NEXT: v_cvt_f32_f16_e32 v4, v4
@@ -191712,21 +191764,21 @@ define <128 x i8> @bitcast_v64f16_to_v128i8(<64 x half> %a, i32 %b) {
; SI-NEXT: v_add_f32_e32 v2, 0x38000000, v2
; SI-NEXT: v_cvt_f16_f32_e32 v2, v2
; SI-NEXT: v_or_b32_e32 v49, v2, v1
-; SI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:492 ; 4-byte Folded Reload
-; SI-NEXT: v_lshlrev_b32_e32 v1, 16, v12
+; SI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:508 ; 4-byte Folded Reload
+; SI-NEXT: v_lshlrev_b32_e32 v1, 16, v13
; SI-NEXT: s_waitcnt vmcnt(0)
; SI-NEXT: v_cvt_f32_f16_e32 v2, v2
; SI-NEXT: v_add_f32_e32 v2, 0x38000000, v2
; SI-NEXT: v_cvt_f16_f32_e32 v2, v2
; SI-NEXT: v_lshlrev_b32_e32 v2, 16, v2
; SI-NEXT: v_or_b32_e32 v51, v4, v2
-; SI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:464 ; 4-byte Folded Reload
-; SI-NEXT: buffer_load_dword v4, off, s[0:3], s32 offset:416 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:480 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v4, off, s[0:3], s32 offset:432 ; 4-byte Folded Reload
; SI-NEXT: v_cvt_f32_f16_e32 v3, v3
; SI-NEXT: v_add_f32_e32 v3, 0x38000000, v3
; SI-NEXT: v_cvt_f16_f32_e32 v3, v3
; SI-NEXT: v_or_b32_e32 v50, v3, v1
-; SI-NEXT: buffer_load_dword v3, off, s[0:3], s32 offset:440 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v3, off, s[0:3], s32 offset:456 ; 4-byte Folded Reload
; SI-NEXT: v_cvt_f32_f16_e32 v1, v31
; SI-NEXT: v_add_f32_e32 v1, 0x38000000, v1
; SI-NEXT: v_cvt_f16_f32_e32 v31, v1
@@ -191742,30 +191794,30 @@ define <128 x i8> @bitcast_v64f16_to_v128i8(<64 x half> %a, i32 %b) {
; SI-NEXT: v_lshlrev_b32_e32 v3, 16, v31
; SI-NEXT: v_or_b32_e32 v52, v1, v3
; SI-NEXT: v_add_f32_e32 v3, 0x38000000, v4
-; SI-NEXT: buffer_load_dword v4, off, s[0:3], s32 offset:388 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v4, off, s[0:3], s32 offset:404 ; 4-byte Folded Reload
; SI-NEXT: v_cvt_f16_f32_e32 v2, v2
; SI-NEXT: v_cvt_f16_f32_e32 v3, v3
; SI-NEXT: v_lshlrev_b32_e32 v1, 16, v2
-; SI-NEXT: v_cvt_f32_f16_e32 v2, v32
+; SI-NEXT: v_cvt_f32_f16_e32 v2, v33
; SI-NEXT: v_or_b32_e32 v54, v3, v1
-; SI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:360 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:376 ; 4-byte Folded Reload
; SI-NEXT: v_add_f32_e32 v2, 0x38000000, v2
-; SI-NEXT: v_cvt_f16_f32_e32 v32, v2
-; SI-NEXT: v_lshlrev_b32_e32 v3, 16, v32
+; SI-NEXT: v_cvt_f16_f32_e32 v33, v2
+; SI-NEXT: v_lshlrev_b32_e32 v3, 16, v33
; SI-NEXT: s_waitcnt vmcnt(1)
; SI-NEXT: v_cvt_f32_f16_e32 v4, v4
; SI-NEXT: v_add_f32_e32 v2, 0x38000000, v4
; SI-NEXT: v_cvt_f16_f32_e32 v2, v2
-; SI-NEXT: buffer_load_dword v4, off, s[0:3], s32 offset:308 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v4, off, s[0:3], s32 offset:328 ; 4-byte Folded Reload
; SI-NEXT: v_or_b32_e32 v53, v2, v3
-; SI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:340 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:356 ; 4-byte Folded Reload
; SI-NEXT: s_waitcnt vmcnt(2)
; SI-NEXT: v_cvt_f32_f16_e32 v1, v1
-; SI-NEXT: v_cvt_f32_f16_e32 v3, v33
+; SI-NEXT: v_cvt_f32_f16_e32 v3, v46
; SI-NEXT: v_add_f32_e32 v1, 0x38000000, v1
; SI-NEXT: v_cvt_f16_f32_e32 v1, v1
; SI-NEXT: v_add_f32_e32 v3, 0x38000000, v3
-; SI-NEXT: v_cvt_f16_f32_e32 v33, v3
+; SI-NEXT: v_cvt_f16_f32_e32 v46, v3
; SI-NEXT: v_lshlrev_b32_e32 v1, 16, v1
; SI-NEXT: s_waitcnt vmcnt(1)
; SI-NEXT: v_cvt_f32_f16_e32 v4, v4
@@ -191776,274 +191828,279 @@ define <128 x i8> @bitcast_v64f16_to_v128i8(<64 x half> %a, i32 %b) {
; SI-NEXT: v_add_f32_e32 v2, 0x38000000, v2
; SI-NEXT: v_cvt_f16_f32_e32 v2, v2
; SI-NEXT: v_or_b32_e32 v44, v2, v1
-; SI-NEXT: v_lshlrev_b32_e32 v1, 16, v33
+; SI-NEXT: v_lshlrev_b32_e32 v1, 16, v46
; SI-NEXT: v_or_b32_e32 v41, v3, v1
; SI-NEXT: v_alignbit_b32 v1, v41, v44, 24
-; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:280 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:300 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_alignbit_b32 v1, v41, v44, 16
-; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:300 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:320 ; 4-byte Folded Spill
+; SI-NEXT: s_waitcnt expcnt(0)
+; SI-NEXT: v_alignbit_b32 v1, v41, v44, 8
+; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:348 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_alignbit_b32 v1, v53, v54, 24
-; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:312 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:332 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_alignbit_b32 v1, v53, v54, 16
-; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:344 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:360 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_alignbit_b32 v1, v53, v54, 8
-; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:380 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:396 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_alignbit_b32 v1, v52, v51, 24
-; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:356 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:372 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_alignbit_b32 v1, v52, v51, 16
-; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:392 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:408 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_alignbit_b32 v1, v52, v51, 8
-; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:428 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:444 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_alignbit_b32 v1, v50, v49, 24
-; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:408 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:424 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_alignbit_b32 v1, v50, v49, 16
-; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:444 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:460 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_alignbit_b32 v1, v50, v49, 8
-; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:480 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:496 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_alignbit_b32 v1, v39, v48, 24
-; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:452 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:468 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_alignbit_b32 v1, v39, v48, 16
-; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:456 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:472 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_alignbit_b32 v1, v39, v48, 8
-; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:468 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:484 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_alignbit_b32 v1, v38, v37, 24
-; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:420 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:436 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_alignbit_b32 v1, v38, v37, 16
-; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:432 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:448 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_alignbit_b32 v1, v38, v37, 8
-; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:448 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:464 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_alignbit_b32 v1, v36, v35, 24
-; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:396 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:412 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_alignbit_b32 v1, v36, v35, 16
-; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:400 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:416 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_alignbit_b32 v1, v36, v35, 8
-; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:412 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:428 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_alignbit_b32 v1, v34, v30, 24
-; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:364 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:380 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_alignbit_b32 v1, v34, v30, 16
-; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:372 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:388 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_alignbit_b32 v1, v34, v30, 8
-; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:376 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:392 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_alignbit_b32 v1, v29, v28, 24
-; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:324 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:344 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_alignbit_b32 v1, v29, v28, 16
-; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:336 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:352 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_alignbit_b32 v1, v29, v28, 8
-; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:348 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:364 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_alignbit_b32 v1, v27, v26, 24
-; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:296 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:316 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_alignbit_b32 v1, v27, v26, 16
-; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:304 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:324 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_alignbit_b32 v1, v27, v26, 8
-; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:320 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:340 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_alignbit_b32 v1, v25, v24, 24
-; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:276 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:296 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_alignbit_b32 v1, v25, v24, 16
-; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:284 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:304 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_alignbit_b32 v1, v25, v24, 8
-; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:292 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:312 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_alignbit_b32 v1, v23, v22, 24
-; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:264 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:280 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_alignbit_b32 v1, v23, v22, 16
-; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:332 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:288 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_alignbit_b32 v1, v23, v22, 8
-; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:272 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:292 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_alignbit_b32 v1, v21, v20, 24
-; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:252 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:268 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_alignbit_b32 v1, v21, v20, 16
-; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:256 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:272 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_alignbit_b32 v1, v21, v20, 8
-; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:260 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:276 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_alignbit_b32 v1, v19, v18, 24
-; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:240 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:256 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_alignbit_b32 v1, v19, v18, 16
-; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:244 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:260 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_alignbit_b32 v1, v19, v18, 8
+; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:264 ; 4-byte Folded Spill
+; SI-NEXT: s_waitcnt expcnt(0)
+; SI-NEXT: v_alignbit_b32 v1, v17, v14, 24
+; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:244 ; 4-byte Folded Spill
+; SI-NEXT: s_waitcnt expcnt(0)
+; SI-NEXT: v_alignbit_b32 v1, v17, v14, 16
; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:248 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_alignbit_b32 v1, v17, v16, 24
-; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:228 ; 4-byte Folded Spill
+; SI-NEXT: v_alignbit_b32 v1, v17, v14, 8
+; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:252 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_alignbit_b32 v1, v17, v16, 16
+; SI-NEXT: v_alignbit_b32 v1, v15, v16, 24
; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:232 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_alignbit_b32 v1, v17, v16, 8
+; SI-NEXT: v_alignbit_b32 v1, v15, v16, 16
; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:236 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_alignbit_b32 v1, v15, v14, 24
-; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:216 ; 4-byte Folded Spill
-; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_alignbit_b32 v1, v15, v14, 16
-; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:220 ; 4-byte Folded Spill
-; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_alignbit_b32 v1, v15, v14, 8
-; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:224 ; 4-byte Folded Spill
+; SI-NEXT: v_alignbit_b32 v1, v15, v16, 8
+; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:240 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_lshrrev_b32_e32 v1, 8, v41
-; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:576 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:592 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_lshrrev_b32_e32 v1, 8, v53
-; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:568 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:584 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_lshrrev_b32_e32 v1, 8, v52
-; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:560 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:576 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_lshrrev_b32_e32 v1, 8, v50
-; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:552 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:568 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_lshrrev_b32_e32 v1, 8, v39
-; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:540 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:556 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_lshrrev_b32_e32 v1, 8, v38
-; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:528 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:544 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_lshrrev_b32_e32 v1, 8, v36
-; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:516 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:532 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_lshrrev_b32_e32 v1, 8, v34
-; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:504 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:520 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_lshrrev_b32_e32 v1, 8, v29
-; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:484 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:500 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_lshrrev_b32_e32 v1, 8, v27
-; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:460 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:476 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_lshrrev_b32_e32 v1, 8, v25
-; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:424 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:440 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_lshrrev_b32_e32 v1, 8, v23
-; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:384 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:400 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_lshrrev_b32_e32 v1, 8, v21
-; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:352 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:368 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_lshrrev_b32_e32 v1, 8, v19
-; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:316 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:336 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_lshrrev_b32_e32 v1, 8, v17
-; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:288 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:308 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_lshrrev_b32_e32 v1, 8, v15
-; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:268 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:284 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_bfe_u32 v1, v33, 8, 8
-; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:592 ; 4-byte Folded Spill
+; SI-NEXT: v_bfe_u32 v1, v46, 8, 8
+; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:608 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_bfe_u32 v1, v32, 8, 8
-; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:588 ; 4-byte Folded Spill
+; SI-NEXT: v_bfe_u32 v1, v33, 8, 8
+; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:604 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_bfe_u32 v1, v31, 8, 8
+; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:596 ; 4-byte Folded Spill
+; SI-NEXT: s_waitcnt expcnt(0)
+; SI-NEXT: v_bfe_u32 v1, v13, 8, 8
+; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:588 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v10, off, s[0:3], s32 offset:228 ; 4-byte Folded Spill
+; SI-NEXT: s_waitcnt expcnt(1)
+; SI-NEXT: v_bfe_u32 v1, v10, 8, 8
; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:580 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_bfe_u32 v1, v12, 8, 8
+; SI-NEXT: v_bfe_u32 v1, v32, 8, 8
; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:572 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_bfe_u32 v1, v9, 8, 8
-; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:564 ; 4-byte Folded Spill
-; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_bfe_u32 v1, v13, 8, 8
-; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:556 ; 4-byte Folded Spill
+; SI-NEXT: v_bfe_u32 v1, v12, 8, 8
+; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:560 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_bfe_u32 v1, v11, 8, 8
-; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:544 ; 4-byte Folded Spill
-; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_bfe_u32 v1, v10, 8, 8
-; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:532 ; 4-byte Folded Spill
-; SI-NEXT: s_waitcnt expcnt(0)
+; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:548 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v9, off, s[0:3], s32 offset:220 ; 4-byte Folded Spill
+; SI-NEXT: s_waitcnt expcnt(1)
+; SI-NEXT: v_bfe_u32 v1, v9, 8, 8
+; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:536 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v8, off, s[0:3], s32 offset:216 ; 4-byte Folded Spill
+; SI-NEXT: s_waitcnt expcnt(1)
; SI-NEXT: v_bfe_u32 v1, v8, 8, 8
-; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:520 ; 4-byte Folded Spill
-; SI-NEXT: s_waitcnt expcnt(0)
+; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:516 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v7, off, s[0:3], s32 offset:212 ; 4-byte Folded Spill
+; SI-NEXT: s_waitcnt expcnt(1)
; SI-NEXT: v_bfe_u32 v1, v7, 8, 8
-; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:500 ; 4-byte Folded Spill
-; SI-NEXT: s_waitcnt expcnt(0)
+; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:504 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v6, off, s[0:3], s32 offset:208 ; 4-byte Folded Spill
+; SI-NEXT: s_waitcnt expcnt(1)
; SI-NEXT: v_bfe_u32 v1, v6, 8, 8
; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:488 ; 4-byte Folded Spill
-; SI-NEXT: buffer_store_dword v42, off, s[0:3], s32 offset:212 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v42, off, s[0:3], s32 offset:224 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(1)
; SI-NEXT: v_bfe_u32 v1, v42, 8, 8
-; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:472 ; 4-byte Folded Spill
-; SI-NEXT: buffer_store_dword v55, off, s[0:3], s32 offset:204 ; 4-byte Folded Spill
-; SI-NEXT: s_waitcnt expcnt(1)
-; SI-NEXT: v_bfe_u32 v1, v55, 8, 8
-; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:436 ; 4-byte Folded Spill
-; SI-NEXT: buffer_store_dword v40, off, s[0:3], s32 offset:208 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:452 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v5, off, s[0:3], s32 offset:204 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(1)
-; SI-NEXT: v_bfe_u32 v1, v40, 8, 8
-; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:404 ; 4-byte Folded Spill
-; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_bfe_u32 v1, v46, 8, 8
-; SI-NEXT: v_alignbit_b32 v5, v41, v44, 8
-; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:368 ; 4-byte Folded Spill
+; SI-NEXT: v_bfe_u32 v1, v5, 8, 8
+; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:420 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_bfe_u32 v1, v45, 8, 8
-; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:328 ; 4-byte Folded Spill
+; SI-NEXT: v_bfe_u32 v1, v40, 8, 8
+; SI-NEXT: v_bfe_u32 v42, v55, 8, 8
+; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:384 ; 4-byte Folded Spill
; SI-NEXT: .LBB94_4: ; %end
; SI-NEXT: s_or_b64 exec, exec, s[4:5]
+; SI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:348 ; 4-byte Folded Reload
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_and_b32_e32 v1, 0xff, v44
-; SI-NEXT: s_waitcnt vmcnt(0)
-; SI-NEXT: v_lshlrev_b32_e32 v2, 8, v5
+; SI-NEXT: buffer_load_dword v3, off, s[0:3], s32 offset:300 ; 4-byte Folded Reload
+; SI-NEXT: s_waitcnt vmcnt(1)
+; SI-NEXT: v_lshlrev_b32_e32 v2, 8, v2
; SI-NEXT: v_or_b32_e32 v1, v1, v2
-; SI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:300 ; 4-byte Folded Reload
-; SI-NEXT: buffer_load_dword v3, off, s[0:3], s32 offset:280 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:320 ; 4-byte Folded Reload
; SI-NEXT: v_and_b32_e32 v1, 0xffff, v1
; SI-NEXT: s_waitcnt vmcnt(1)
+; SI-NEXT: v_lshlrev_b32_e32 v3, 24, v3
+; SI-NEXT: s_waitcnt vmcnt(0)
; SI-NEXT: v_and_b32_e32 v2, 0xff, v2
; SI-NEXT: v_lshlrev_b32_e32 v2, 16, v2
-; SI-NEXT: s_waitcnt vmcnt(0)
-; SI-NEXT: v_lshlrev_b32_e32 v3, 24, v3
; SI-NEXT: v_or_b32_e32 v2, v3, v2
; SI-NEXT: v_or_b32_e32 v1, v1, v2
; SI-NEXT: buffer_store_dword v1, v0, s[0:3], 0 offen
-; SI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:576 ; 4-byte Folded Reload
-; SI-NEXT: buffer_load_dword v3, off, s[0:3], s32 offset:592 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:592 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v3, off, s[0:3], s32 offset:608 ; 4-byte Folded Reload
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_and_b32_e32 v1, 0xff, v41
; SI-NEXT: s_waitcnt vmcnt(1)
; SI-NEXT: v_lshlrev_b32_e32 v2, 8, v2
; SI-NEXT: v_or_b32_e32 v1, v1, v2
-; SI-NEXT: v_and_b32_e32 v2, 0xff, v33
+; SI-NEXT: v_and_b32_e32 v2, 0xff, v46
; SI-NEXT: v_lshlrev_b32_e32 v2, 16, v2
; SI-NEXT: s_waitcnt vmcnt(0)
; SI-NEXT: v_lshlrev_b32_e32 v3, 24, v3
@@ -192052,14 +192109,14 @@ define <128 x i8> @bitcast_v64f16_to_v128i8(<64 x half> %a, i32 %b) {
; SI-NEXT: v_or_b32_e32 v1, v1, v2
; SI-NEXT: v_add_i32_e32 v2, vcc, 4, v0
; SI-NEXT: buffer_store_dword v1, v2, s[0:3], 0 offen
-; SI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:380 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:396 ; 4-byte Folded Reload
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_and_b32_e32 v1, 0xff, v54
-; SI-NEXT: buffer_load_dword v3, off, s[0:3], s32 offset:312 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v3, off, s[0:3], s32 offset:332 ; 4-byte Folded Reload
; SI-NEXT: s_waitcnt vmcnt(1)
; SI-NEXT: v_lshlrev_b32_e32 v2, 8, v2
; SI-NEXT: v_or_b32_e32 v1, v1, v2
-; SI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:344 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:360 ; 4-byte Folded Reload
; SI-NEXT: v_and_b32_e32 v1, 0xffff, v1
; SI-NEXT: s_waitcnt vmcnt(1)
; SI-NEXT: v_lshlrev_b32_e32 v3, 24, v3
@@ -192070,14 +192127,14 @@ define <128 x i8> @bitcast_v64f16_to_v128i8(<64 x half> %a, i32 %b) {
; SI-NEXT: v_or_b32_e32 v1, v1, v2
; SI-NEXT: v_add_i32_e32 v2, vcc, 8, v0
; SI-NEXT: buffer_store_dword v1, v2, s[0:3], 0 offen
-; SI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:568 ; 4-byte Folded Reload
-; SI-NEXT: buffer_load_dword v3, off, s[0:3], s32 offset:588 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:584 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v3, off, s[0:3], s32 offset:604 ; 4-byte Folded Reload
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_and_b32_e32 v1, 0xff, v53
; SI-NEXT: s_waitcnt vmcnt(1)
; SI-NEXT: v_lshlrev_b32_e32 v2, 8, v2
; SI-NEXT: v_or_b32_e32 v1, v1, v2
-; SI-NEXT: v_and_b32_e32 v2, 0xff, v32
+; SI-NEXT: v_and_b32_e32 v2, 0xff, v33
; SI-NEXT: v_lshlrev_b32_e32 v2, 16, v2
; SI-NEXT: s_waitcnt vmcnt(0)
; SI-NEXT: v_lshlrev_b32_e32 v3, 24, v3
@@ -192086,14 +192143,14 @@ define <128 x i8> @bitcast_v64f16_to_v128i8(<64 x half> %a, i32 %b) {
; SI-NEXT: v_or_b32_e32 v1, v1, v2
; SI-NEXT: v_add_i32_e32 v2, vcc, 12, v0
; SI-NEXT: buffer_store_dword v1, v2, s[0:3], 0 offen
-; SI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:428 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:444 ; 4-byte Folded Reload
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_and_b32_e32 v1, 0xff, v51
-; SI-NEXT: buffer_load_dword v3, off, s[0:3], s32 offset:356 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v3, off, s[0:3], s32 offset:372 ; 4-byte Folded Reload
; SI-NEXT: s_waitcnt vmcnt(1)
; SI-NEXT: v_lshlrev_b32_e32 v2, 8, v2
; SI-NEXT: v_or_b32_e32 v1, v1, v2
-; SI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:392 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:408 ; 4-byte Folded Reload
; SI-NEXT: v_and_b32_e32 v1, 0xffff, v1
; SI-NEXT: s_waitcnt vmcnt(1)
; SI-NEXT: v_lshlrev_b32_e32 v3, 24, v3
@@ -192104,8 +192161,8 @@ define <128 x i8> @bitcast_v64f16_to_v128i8(<64 x half> %a, i32 %b) {
; SI-NEXT: v_or_b32_e32 v1, v1, v2
; SI-NEXT: v_add_i32_e32 v2, vcc, 16, v0
; SI-NEXT: buffer_store_dword v1, v2, s[0:3], 0 offen
-; SI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:560 ; 4-byte Folded Reload
-; SI-NEXT: buffer_load_dword v3, off, s[0:3], s32 offset:580 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:576 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v3, off, s[0:3], s32 offset:596 ; 4-byte Folded Reload
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_and_b32_e32 v1, 0xff, v52
; SI-NEXT: s_waitcnt vmcnt(1)
@@ -192120,14 +192177,14 @@ define <128 x i8> @bitcast_v64f16_to_v128i8(<64 x half> %a, i32 %b) {
; SI-NEXT: v_or_b32_e32 v1, v1, v2
; SI-NEXT: v_add_i32_e32 v2, vcc, 20, v0
; SI-NEXT: buffer_store_dword v1, v2, s[0:3], 0 offen
-; SI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:480 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:496 ; 4-byte Folded Reload
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_and_b32_e32 v1, 0xff, v49
-; SI-NEXT: buffer_load_dword v3, off, s[0:3], s32 offset:408 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v3, off, s[0:3], s32 offset:424 ; 4-byte Folded Reload
; SI-NEXT: s_waitcnt vmcnt(1)
; SI-NEXT: v_lshlrev_b32_e32 v2, 8, v2
; SI-NEXT: v_or_b32_e32 v1, v1, v2
-; SI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:444 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:460 ; 4-byte Folded Reload
; SI-NEXT: v_and_b32_e32 v1, 0xffff, v1
; SI-NEXT: s_waitcnt vmcnt(1)
; SI-NEXT: v_lshlrev_b32_e32 v3, 24, v3
@@ -192138,14 +192195,14 @@ define <128 x i8> @bitcast_v64f16_to_v128i8(<64 x half> %a, i32 %b) {
; SI-NEXT: v_or_b32_e32 v1, v1, v2
; SI-NEXT: v_add_i32_e32 v2, vcc, 24, v0
; SI-NEXT: buffer_store_dword v1, v2, s[0:3], 0 offen
-; SI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:552 ; 4-byte Folded Reload
-; SI-NEXT: buffer_load_dword v3, off, s[0:3], s32 offset:572 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:568 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v3, off, s[0:3], s32 offset:588 ; 4-byte Folded Reload
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_and_b32_e32 v1, 0xff, v50
; SI-NEXT: s_waitcnt vmcnt(1)
; SI-NEXT: v_lshlrev_b32_e32 v2, 8, v2
; SI-NEXT: v_or_b32_e32 v1, v1, v2
-; SI-NEXT: v_and_b32_e32 v2, 0xff, v12
+; SI-NEXT: v_and_b32_e32 v2, 0xff, v13
; SI-NEXT: v_lshlrev_b32_e32 v2, 16, v2
; SI-NEXT: s_waitcnt vmcnt(0)
; SI-NEXT: v_lshlrev_b32_e32 v3, 24, v3
@@ -192154,14 +192211,14 @@ define <128 x i8> @bitcast_v64f16_to_v128i8(<64 x half> %a, i32 %b) {
; SI-NEXT: v_or_b32_e32 v1, v1, v2
; SI-NEXT: v_add_i32_e32 v2, vcc, 28, v0
; SI-NEXT: buffer_store_dword v1, v2, s[0:3], 0 offen
-; SI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:468 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:484 ; 4-byte Folded Reload
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_and_b32_e32 v1, 0xff, v48
-; SI-NEXT: buffer_load_dword v3, off, s[0:3], s32 offset:452 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v3, off, s[0:3], s32 offset:468 ; 4-byte Folded Reload
; SI-NEXT: s_waitcnt vmcnt(1)
; SI-NEXT: v_lshlrev_b32_e32 v2, 8, v2
; SI-NEXT: v_or_b32_e32 v1, v1, v2
-; SI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:456 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:472 ; 4-byte Folded Reload
; SI-NEXT: v_and_b32_e32 v1, 0xffff, v1
; SI-NEXT: s_waitcnt vmcnt(1)
; SI-NEXT: v_lshlrev_b32_e32 v3, 24, v3
@@ -192172,30 +192229,32 @@ define <128 x i8> @bitcast_v64f16_to_v128i8(<64 x half> %a, i32 %b) {
; SI-NEXT: v_or_b32_e32 v1, v1, v2
; SI-NEXT: v_add_i32_e32 v2, vcc, 32, v0
; SI-NEXT: buffer_store_dword v1, v2, s[0:3], 0 offen
-; SI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:540 ; 4-byte Folded Reload
-; SI-NEXT: buffer_load_dword v3, off, s[0:3], s32 offset:564 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:556 ; 4-byte Folded Reload
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_and_b32_e32 v1, 0xff, v39
+; SI-NEXT: buffer_load_dword v3, off, s[0:3], s32 offset:580 ; 4-byte Folded Reload
; SI-NEXT: s_waitcnt vmcnt(1)
; SI-NEXT: v_lshlrev_b32_e32 v2, 8, v2
; SI-NEXT: v_or_b32_e32 v1, v1, v2
-; SI-NEXT: v_and_b32_e32 v2, 0xff, v9
-; SI-NEXT: v_lshlrev_b32_e32 v2, 16, v2
-; SI-NEXT: s_waitcnt vmcnt(0)
+; SI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:228 ; 4-byte Folded Reload
+; SI-NEXT: v_and_b32_e32 v1, 0xffff, v1
+; SI-NEXT: s_waitcnt vmcnt(1)
; SI-NEXT: v_lshlrev_b32_e32 v3, 24, v3
+; SI-NEXT: s_waitcnt vmcnt(0)
+; SI-NEXT: v_and_b32_e32 v2, 0xff, v2
+; SI-NEXT: v_lshlrev_b32_e32 v2, 16, v2
; SI-NEXT: v_or_b32_e32 v2, v3, v2
-; SI-NEXT: v_and_b32_e32 v1, 0xffff, v1
; SI-NEXT: v_or_b32_e32 v1, v1, v2
; SI-NEXT: v_add_i32_e32 v2, vcc, 36, v0
; SI-NEXT: buffer_store_dword v1, v2, s[0:3], 0 offen
-; SI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:448 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:464 ; 4-byte Folded Reload
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_and_b32_e32 v1, 0xff, v37
-; SI-NEXT: buffer_load_dword v3, off, s[0:3], s32 offset:420 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v3, off, s[0:3], s32 offset:436 ; 4-byte Folded Reload
; SI-NEXT: s_waitcnt vmcnt(1)
; SI-NEXT: v_lshlrev_b32_e32 v2, 8, v2
; SI-NEXT: v_or_b32_e32 v1, v1, v2
-; SI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:432 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:448 ; 4-byte Folded Reload
; SI-NEXT: v_and_b32_e32 v1, 0xffff, v1
; SI-NEXT: s_waitcnt vmcnt(1)
; SI-NEXT: v_lshlrev_b32_e32 v3, 24, v3
@@ -192206,14 +192265,14 @@ define <128 x i8> @bitcast_v64f16_to_v128i8(<64 x half> %a, i32 %b) {
; SI-NEXT: v_or_b32_e32 v1, v1, v2
; SI-NEXT: v_add_i32_e32 v2, vcc, 40, v0
; SI-NEXT: buffer_store_dword v1, v2, s[0:3], 0 offen
-; SI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:528 ; 4-byte Folded Reload
-; SI-NEXT: buffer_load_dword v3, off, s[0:3], s32 offset:556 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:544 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v3, off, s[0:3], s32 offset:572 ; 4-byte Folded Reload
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_and_b32_e32 v1, 0xff, v38
; SI-NEXT: s_waitcnt vmcnt(1)
; SI-NEXT: v_lshlrev_b32_e32 v2, 8, v2
; SI-NEXT: v_or_b32_e32 v1, v1, v2
-; SI-NEXT: v_and_b32_e32 v2, 0xff, v13
+; SI-NEXT: v_and_b32_e32 v2, 0xff, v32
; SI-NEXT: v_lshlrev_b32_e32 v2, 16, v2
; SI-NEXT: s_waitcnt vmcnt(0)
; SI-NEXT: v_lshlrev_b32_e32 v3, 24, v3
@@ -192222,14 +192281,14 @@ define <128 x i8> @bitcast_v64f16_to_v128i8(<64 x half> %a, i32 %b) {
; SI-NEXT: v_or_b32_e32 v1, v1, v2
; SI-NEXT: v_add_i32_e32 v2, vcc, 44, v0
; SI-NEXT: buffer_store_dword v1, v2, s[0:3], 0 offen
-; SI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:412 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:428 ; 4-byte Folded Reload
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_and_b32_e32 v1, 0xff, v35
-; SI-NEXT: buffer_load_dword v3, off, s[0:3], s32 offset:396 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v3, off, s[0:3], s32 offset:412 ; 4-byte Folded Reload
; SI-NEXT: s_waitcnt vmcnt(1)
; SI-NEXT: v_lshlrev_b32_e32 v2, 8, v2
; SI-NEXT: v_or_b32_e32 v1, v1, v2
-; SI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:400 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:416 ; 4-byte Folded Reload
; SI-NEXT: v_and_b32_e32 v1, 0xffff, v1
; SI-NEXT: s_waitcnt vmcnt(1)
; SI-NEXT: v_lshlrev_b32_e32 v3, 24, v3
@@ -192240,14 +192299,14 @@ define <128 x i8> @bitcast_v64f16_to_v128i8(<64 x half> %a, i32 %b) {
; SI-NEXT: v_or_b32_e32 v1, v1, v2
; SI-NEXT: v_add_i32_e32 v2, vcc, 48, v0
; SI-NEXT: buffer_store_dword v1, v2, s[0:3], 0 offen
-; SI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:516 ; 4-byte Folded Reload
-; SI-NEXT: buffer_load_dword v3, off, s[0:3], s32 offset:544 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:532 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v3, off, s[0:3], s32 offset:560 ; 4-byte Folded Reload
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_and_b32_e32 v1, 0xff, v36
; SI-NEXT: s_waitcnt vmcnt(1)
; SI-NEXT: v_lshlrev_b32_e32 v2, 8, v2
; SI-NEXT: v_or_b32_e32 v1, v1, v2
-; SI-NEXT: v_and_b32_e32 v2, 0xff, v11
+; SI-NEXT: v_and_b32_e32 v2, 0xff, v12
; SI-NEXT: v_lshlrev_b32_e32 v2, 16, v2
; SI-NEXT: s_waitcnt vmcnt(0)
; SI-NEXT: v_lshlrev_b32_e32 v3, 24, v3
@@ -192256,14 +192315,14 @@ define <128 x i8> @bitcast_v64f16_to_v128i8(<64 x half> %a, i32 %b) {
; SI-NEXT: v_or_b32_e32 v1, v1, v2
; SI-NEXT: v_add_i32_e32 v2, vcc, 52, v0
; SI-NEXT: buffer_store_dword v1, v2, s[0:3], 0 offen
-; SI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:376 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:392 ; 4-byte Folded Reload
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_and_b32_e32 v1, 0xff, v30
-; SI-NEXT: buffer_load_dword v3, off, s[0:3], s32 offset:364 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v3, off, s[0:3], s32 offset:380 ; 4-byte Folded Reload
; SI-NEXT: s_waitcnt vmcnt(1)
; SI-NEXT: v_lshlrev_b32_e32 v2, 8, v2
; SI-NEXT: v_or_b32_e32 v1, v1, v2
-; SI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:372 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:388 ; 4-byte Folded Reload
; SI-NEXT: v_and_b32_e32 v1, 0xffff, v1
; SI-NEXT: s_waitcnt vmcnt(1)
; SI-NEXT: v_lshlrev_b32_e32 v3, 24, v3
@@ -192274,14 +192333,14 @@ define <128 x i8> @bitcast_v64f16_to_v128i8(<64 x half> %a, i32 %b) {
; SI-NEXT: v_or_b32_e32 v1, v1, v2
; SI-NEXT: v_add_i32_e32 v2, vcc, 56, v0
; SI-NEXT: buffer_store_dword v1, v2, s[0:3], 0 offen
-; SI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:504 ; 4-byte Folded Reload
-; SI-NEXT: buffer_load_dword v3, off, s[0:3], s32 offset:532 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:520 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v3, off, s[0:3], s32 offset:548 ; 4-byte Folded Reload
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_and_b32_e32 v1, 0xff, v34
; SI-NEXT: s_waitcnt vmcnt(1)
; SI-NEXT: v_lshlrev_b32_e32 v2, 8, v2
; SI-NEXT: v_or_b32_e32 v1, v1, v2
-; SI-NEXT: v_and_b32_e32 v2, 0xff, v10
+; SI-NEXT: v_and_b32_e32 v2, 0xff, v11
; SI-NEXT: v_lshlrev_b32_e32 v2, 16, v2
; SI-NEXT: s_waitcnt vmcnt(0)
; SI-NEXT: v_lshlrev_b32_e32 v3, 24, v3
@@ -192290,14 +192349,14 @@ define <128 x i8> @bitcast_v64f16_to_v128i8(<64 x half> %a, i32 %b) {
; SI-NEXT: v_or_b32_e32 v1, v1, v2
; SI-NEXT: v_add_i32_e32 v2, vcc, 60, v0
; SI-NEXT: buffer_store_dword v1, v2, s[0:3], 0 offen
-; SI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:348 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:364 ; 4-byte Folded Reload
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_and_b32_e32 v1, 0xff, v28
-; SI-NEXT: buffer_load_dword v3, off, s[0:3], s32 offset:324 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v3, off, s[0:3], s32 offset:344 ; 4-byte Folded Reload
; SI-NEXT: s_waitcnt vmcnt(1)
; SI-NEXT: v_lshlrev_b32_e32 v2, 8, v2
; SI-NEXT: v_or_b32_e32 v1, v1, v2
-; SI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:336 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:352 ; 4-byte Folded Reload
; SI-NEXT: v_and_b32_e32 v1, 0xffff, v1
; SI-NEXT: s_waitcnt vmcnt(1)
; SI-NEXT: v_lshlrev_b32_e32 v3, 24, v3
@@ -192308,30 +192367,32 @@ define <128 x i8> @bitcast_v64f16_to_v128i8(<64 x half> %a, i32 %b) {
; SI-NEXT: v_or_b32_e32 v1, v1, v2
; SI-NEXT: v_add_i32_e32 v2, vcc, 64, v0
; SI-NEXT: buffer_store_dword v1, v2, s[0:3], 0 offen
-; SI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:484 ; 4-byte Folded Reload
-; SI-NEXT: buffer_load_dword v3, off, s[0:3], s32 offset:520 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:500 ; 4-byte Folded Reload
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_and_b32_e32 v1, 0xff, v29
+; SI-NEXT: buffer_load_dword v3, off, s[0:3], s32 offset:536 ; 4-byte Folded Reload
; SI-NEXT: s_waitcnt vmcnt(1)
; SI-NEXT: v_lshlrev_b32_e32 v2, 8, v2
; SI-NEXT: v_or_b32_e32 v1, v1, v2
-; SI-NEXT: v_and_b32_e32 v2, 0xff, v8
-; SI-NEXT: v_lshlrev_b32_e32 v2, 16, v2
-; SI-NEXT: s_waitcnt vmcnt(0)
-; SI-NEXT: v_lshlrev_b32_e32 v3, 24, v3
+; SI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:220 ; 4-byte Folded Reload
; SI-NEXT: v_and_b32_e32 v1, 0xffff, v1
+; SI-NEXT: s_waitcnt vmcnt(1)
+; SI-NEXT: v_lshlrev_b32_e32 v3, 24, v3
+; SI-NEXT: s_waitcnt vmcnt(0)
+; SI-NEXT: v_and_b32_e32 v2, 0xff, v2
+; SI-NEXT: v_lshlrev_b32_e32 v2, 16, v2
; SI-NEXT: v_or_b32_e32 v2, v3, v2
; SI-NEXT: v_or_b32_e32 v1, v1, v2
; SI-NEXT: v_add_i32_e32 v2, vcc, 0x44, v0
; SI-NEXT: buffer_store_dword v1, v2, s[0:3], 0 offen
-; SI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:320 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:340 ; 4-byte Folded Reload
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_and_b32_e32 v1, 0xff, v26
-; SI-NEXT: buffer_load_dword v3, off, s[0:3], s32 offset:296 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v3, off, s[0:3], s32 offset:316 ; 4-byte Folded Reload
; SI-NEXT: s_waitcnt vmcnt(1)
; SI-NEXT: v_lshlrev_b32_e32 v2, 8, v2
; SI-NEXT: v_or_b32_e32 v1, v1, v2
-; SI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:304 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:324 ; 4-byte Folded Reload
; SI-NEXT: v_and_b32_e32 v1, 0xffff, v1
; SI-NEXT: s_waitcnt vmcnt(1)
; SI-NEXT: v_lshlrev_b32_e32 v3, 24, v3
@@ -192342,33 +192403,35 @@ define <128 x i8> @bitcast_v64f16_to_v128i8(<64 x half> %a, i32 %b) {
; SI-NEXT: v_or_b32_e32 v1, v1, v2
; SI-NEXT: v_add_i32_e32 v2, vcc, 0x48, v0
; SI-NEXT: buffer_store_dword v1, v2, s[0:3], 0 offen
-; SI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:460 ; 4-byte Folded Reload
-; SI-NEXT: buffer_load_dword v3, off, s[0:3], s32 offset:500 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:476 ; 4-byte Folded Reload
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_and_b32_e32 v1, 0xff, v27
+; SI-NEXT: buffer_load_dword v3, off, s[0:3], s32 offset:516 ; 4-byte Folded Reload
; SI-NEXT: s_waitcnt vmcnt(1)
; SI-NEXT: v_lshlrev_b32_e32 v2, 8, v2
; SI-NEXT: v_or_b32_e32 v1, v1, v2
-; SI-NEXT: v_and_b32_e32 v2, 0xff, v7
-; SI-NEXT: v_lshlrev_b32_e32 v2, 16, v2
-; SI-NEXT: s_waitcnt vmcnt(0)
-; SI-NEXT: v_lshlrev_b32_e32 v3, 24, v3
+; SI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:216 ; 4-byte Folded Reload
; SI-NEXT: v_and_b32_e32 v1, 0xffff, v1
+; SI-NEXT: s_waitcnt vmcnt(1)
+; SI-NEXT: v_lshlrev_b32_e32 v3, 24, v3
+; SI-NEXT: s_waitcnt vmcnt(0)
+; SI-NEXT: v_and_b32_e32 v2, 0xff, v2
+; SI-NEXT: v_lshlrev_b32_e32 v2, 16, v2
; SI-NEXT: v_or_b32_e32 v2, v3, v2
; SI-NEXT: v_or_b32_e32 v1, v1, v2
; SI-NEXT: v_add_i32_e32 v2, vcc, 0x4c, v0
; SI-NEXT: buffer_store_dword v1, v2, s[0:3], 0 offen
-; SI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:292 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:312 ; 4-byte Folded Reload
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_and_b32_e32 v1, 0xff, v24
-; SI-NEXT: buffer_load_dword v3, off, s[0:3], s32 offset:276 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v3, off, s[0:3], s32 offset:296 ; 4-byte Folded Reload
; SI-NEXT: s_waitcnt vmcnt(1)
; SI-NEXT: v_lshlrev_b32_e32 v2, 8, v2
; SI-NEXT: v_or_b32_e32 v1, v1, v2
-; SI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:284 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:304 ; 4-byte Folded Reload
+; SI-NEXT: v_and_b32_e32 v1, 0xffff, v1
; SI-NEXT: s_waitcnt vmcnt(1)
; SI-NEXT: v_lshlrev_b32_e32 v3, 24, v3
-; SI-NEXT: v_and_b32_e32 v1, 0xffff, v1
; SI-NEXT: s_waitcnt vmcnt(0)
; SI-NEXT: v_and_b32_e32 v2, 0xff, v2
; SI-NEXT: v_lshlrev_b32_e32 v2, 16, v2
@@ -192376,30 +192439,32 @@ define <128 x i8> @bitcast_v64f16_to_v128i8(<64 x half> %a, i32 %b) {
; SI-NEXT: v_or_b32_e32 v1, v1, v2
; SI-NEXT: v_add_i32_e32 v2, vcc, 0x50, v0
; SI-NEXT: buffer_store_dword v1, v2, s[0:3], 0 offen
-; SI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:424 ; 4-byte Folded Reload
-; SI-NEXT: buffer_load_dword v3, off, s[0:3], s32 offset:488 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:440 ; 4-byte Folded Reload
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_and_b32_e32 v1, 0xff, v25
+; SI-NEXT: buffer_load_dword v3, off, s[0:3], s32 offset:504 ; 4-byte Folded Reload
; SI-NEXT: s_waitcnt vmcnt(1)
; SI-NEXT: v_lshlrev_b32_e32 v2, 8, v2
; SI-NEXT: v_or_b32_e32 v1, v1, v2
-; SI-NEXT: v_and_b32_e32 v2, 0xff, v6
-; SI-NEXT: v_lshlrev_b32_e32 v2, 16, v2
-; SI-NEXT: s_waitcnt vmcnt(0)
-; SI-NEXT: v_lshlrev_b32_e32 v3, 24, v3
+; SI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:212 ; 4-byte Folded Reload
; SI-NEXT: v_and_b32_e32 v1, 0xffff, v1
+; SI-NEXT: s_waitcnt vmcnt(1)
+; SI-NEXT: v_lshlrev_b32_e32 v3, 24, v3
+; SI-NEXT: s_waitcnt vmcnt(0)
+; SI-NEXT: v_and_b32_e32 v2, 0xff, v2
+; SI-NEXT: v_lshlrev_b32_e32 v2, 16, v2
; SI-NEXT: v_or_b32_e32 v2, v3, v2
; SI-NEXT: v_or_b32_e32 v1, v1, v2
; SI-NEXT: v_add_i32_e32 v2, vcc, 0x54, v0
; SI-NEXT: buffer_store_dword v1, v2, s[0:3], 0 offen
-; SI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:272 ; 4-byte Folded Reload
-; SI-NEXT: buffer_load_dword v3, off, s[0:3], s32 offset:264 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:292 ; 4-byte Folded Reload
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_and_b32_e32 v1, 0xff, v22
+; SI-NEXT: buffer_load_dword v3, off, s[0:3], s32 offset:280 ; 4-byte Folded Reload
; SI-NEXT: s_waitcnt vmcnt(1)
; SI-NEXT: v_lshlrev_b32_e32 v2, 8, v2
; SI-NEXT: v_or_b32_e32 v1, v1, v2
-; SI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:332 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:288 ; 4-byte Folded Reload
; SI-NEXT: s_waitcnt vmcnt(1)
; SI-NEXT: v_lshlrev_b32_e32 v3, 24, v3
; SI-NEXT: v_and_b32_e32 v1, 0xffff, v1
@@ -192410,14 +192475,14 @@ define <128 x i8> @bitcast_v64f16_to_v128i8(<64 x half> %a, i32 %b) {
; SI-NEXT: v_or_b32_e32 v1, v1, v2
; SI-NEXT: v_add_i32_e32 v2, vcc, 0x58, v0
; SI-NEXT: buffer_store_dword v1, v2, s[0:3], 0 offen
-; SI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:384 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:400 ; 4-byte Folded Reload
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_and_b32_e32 v1, 0xff, v23
-; SI-NEXT: buffer_load_dword v3, off, s[0:3], s32 offset:472 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v3, off, s[0:3], s32 offset:488 ; 4-byte Folded Reload
; SI-NEXT: s_waitcnt vmcnt(1)
; SI-NEXT: v_lshlrev_b32_e32 v2, 8, v2
; SI-NEXT: v_or_b32_e32 v1, v1, v2
-; SI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:212 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:208 ; 4-byte Folded Reload
; SI-NEXT: s_waitcnt vmcnt(1)
; SI-NEXT: v_lshlrev_b32_e32 v3, 24, v3
; SI-NEXT: v_and_b32_e32 v1, 0xffff, v1
@@ -192428,14 +192493,14 @@ define <128 x i8> @bitcast_v64f16_to_v128i8(<64 x half> %a, i32 %b) {
; SI-NEXT: v_or_b32_e32 v1, v1, v2
; SI-NEXT: v_add_i32_e32 v2, vcc, 0x5c, v0
; SI-NEXT: buffer_store_dword v1, v2, s[0:3], 0 offen
-; SI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:260 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:276 ; 4-byte Folded Reload
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_and_b32_e32 v1, 0xff, v20
-; SI-NEXT: buffer_load_dword v3, off, s[0:3], s32 offset:252 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v3, off, s[0:3], s32 offset:268 ; 4-byte Folded Reload
; SI-NEXT: s_waitcnt vmcnt(1)
; SI-NEXT: v_lshlrev_b32_e32 v2, 8, v2
; SI-NEXT: v_or_b32_e32 v1, v1, v2
-; SI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:256 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:272 ; 4-byte Folded Reload
; SI-NEXT: s_waitcnt vmcnt(1)
; SI-NEXT: v_lshlrev_b32_e32 v3, 24, v3
; SI-NEXT: v_and_b32_e32 v1, 0xffff, v1
@@ -192446,14 +192511,14 @@ define <128 x i8> @bitcast_v64f16_to_v128i8(<64 x half> %a, i32 %b) {
; SI-NEXT: v_or_b32_e32 v1, v1, v2
; SI-NEXT: v_add_i32_e32 v2, vcc, 0x60, v0
; SI-NEXT: buffer_store_dword v1, v2, s[0:3], 0 offen
-; SI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:352 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:368 ; 4-byte Folded Reload
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_and_b32_e32 v1, 0xff, v21
-; SI-NEXT: buffer_load_dword v3, off, s[0:3], s32 offset:436 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v3, off, s[0:3], s32 offset:452 ; 4-byte Folded Reload
; SI-NEXT: s_waitcnt vmcnt(1)
; SI-NEXT: v_lshlrev_b32_e32 v2, 8, v2
; SI-NEXT: v_or_b32_e32 v1, v1, v2
-; SI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:204 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:224 ; 4-byte Folded Reload
; SI-NEXT: s_waitcnt vmcnt(1)
; SI-NEXT: v_lshlrev_b32_e32 v3, 24, v3
; SI-NEXT: v_and_b32_e32 v1, 0xffff, v1
@@ -192464,14 +192529,14 @@ define <128 x i8> @bitcast_v64f16_to_v128i8(<64 x half> %a, i32 %b) {
; SI-NEXT: v_or_b32_e32 v1, v1, v2
; SI-NEXT: v_add_i32_e32 v2, vcc, 0x64, v0
; SI-NEXT: buffer_store_dword v1, v2, s[0:3], 0 offen
-; SI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:248 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:264 ; 4-byte Folded Reload
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_and_b32_e32 v1, 0xff, v18
-; SI-NEXT: buffer_load_dword v3, off, s[0:3], s32 offset:240 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v3, off, s[0:3], s32 offset:256 ; 4-byte Folded Reload
; SI-NEXT: s_waitcnt vmcnt(1)
; SI-NEXT: v_lshlrev_b32_e32 v2, 8, v2
; SI-NEXT: v_or_b32_e32 v1, v1, v2
-; SI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:244 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:260 ; 4-byte Folded Reload
; SI-NEXT: s_waitcnt vmcnt(1)
; SI-NEXT: v_lshlrev_b32_e32 v3, 24, v3
; SI-NEXT: v_and_b32_e32 v1, 0xffff, v1
@@ -192482,14 +192547,14 @@ define <128 x i8> @bitcast_v64f16_to_v128i8(<64 x half> %a, i32 %b) {
; SI-NEXT: v_or_b32_e32 v1, v1, v2
; SI-NEXT: v_add_i32_e32 v2, vcc, 0x68, v0
; SI-NEXT: buffer_store_dword v1, v2, s[0:3], 0 offen
-; SI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:316 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:336 ; 4-byte Folded Reload
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_and_b32_e32 v1, 0xff, v19
-; SI-NEXT: buffer_load_dword v3, off, s[0:3], s32 offset:404 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v3, off, s[0:3], s32 offset:420 ; 4-byte Folded Reload
; SI-NEXT: s_waitcnt vmcnt(1)
; SI-NEXT: v_lshlrev_b32_e32 v2, 8, v2
; SI-NEXT: v_or_b32_e32 v1, v1, v2
-; SI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:208 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:204 ; 4-byte Folded Reload
; SI-NEXT: s_waitcnt vmcnt(1)
; SI-NEXT: v_lshlrev_b32_e32 v3, 24, v3
; SI-NEXT: v_and_b32_e32 v1, 0xffff, v1
@@ -192500,14 +192565,14 @@ define <128 x i8> @bitcast_v64f16_to_v128i8(<64 x half> %a, i32 %b) {
; SI-NEXT: v_or_b32_e32 v1, v1, v2
; SI-NEXT: v_add_i32_e32 v2, vcc, 0x6c, v0
; SI-NEXT: buffer_store_dword v1, v2, s[0:3], 0 offen
-; SI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:236 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:252 ; 4-byte Folded Reload
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_and_b32_e32 v1, 0xff, v16
-; SI-NEXT: buffer_load_dword v3, off, s[0:3], s32 offset:228 ; 4-byte Folded Reload
+; SI-NEXT: v_and_b32_e32 v1, 0xff, v14
+; SI-NEXT: buffer_load_dword v3, off, s[0:3], s32 offset:244 ; 4-byte Folded Reload
; SI-NEXT: s_waitcnt vmcnt(1)
; SI-NEXT: v_lshlrev_b32_e32 v2, 8, v2
; SI-NEXT: v_or_b32_e32 v1, v1, v2
-; SI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:232 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:248 ; 4-byte Folded Reload
; SI-NEXT: s_waitcnt vmcnt(1)
; SI-NEXT: v_lshlrev_b32_e32 v3, 24, v3
; SI-NEXT: v_and_b32_e32 v1, 0xffff, v1
@@ -192518,14 +192583,14 @@ define <128 x i8> @bitcast_v64f16_to_v128i8(<64 x half> %a, i32 %b) {
; SI-NEXT: v_or_b32_e32 v1, v1, v2
; SI-NEXT: v_add_i32_e32 v2, vcc, 0x70, v0
; SI-NEXT: buffer_store_dword v1, v2, s[0:3], 0 offen
-; SI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:288 ; 4-byte Folded Reload
-; SI-NEXT: buffer_load_dword v3, off, s[0:3], s32 offset:368 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:308 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v3, off, s[0:3], s32 offset:384 ; 4-byte Folded Reload
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_and_b32_e32 v1, 0xff, v17
; SI-NEXT: s_waitcnt vmcnt(1)
; SI-NEXT: v_lshlrev_b32_e32 v2, 8, v2
; SI-NEXT: v_or_b32_e32 v1, v1, v2
-; SI-NEXT: v_and_b32_e32 v2, 0xff, v46
+; SI-NEXT: v_and_b32_e32 v2, 0xff, v40
; SI-NEXT: v_lshlrev_b32_e32 v2, 16, v2
; SI-NEXT: s_waitcnt vmcnt(0)
; SI-NEXT: v_lshlrev_b32_e32 v3, 24, v3
@@ -192534,14 +192599,14 @@ define <128 x i8> @bitcast_v64f16_to_v128i8(<64 x half> %a, i32 %b) {
; SI-NEXT: v_or_b32_e32 v1, v1, v2
; SI-NEXT: v_add_i32_e32 v2, vcc, 0x74, v0
; SI-NEXT: buffer_store_dword v1, v2, s[0:3], 0 offen
-; SI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:224 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:240 ; 4-byte Folded Reload
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_and_b32_e32 v1, 0xff, v14
-; SI-NEXT: buffer_load_dword v3, off, s[0:3], s32 offset:216 ; 4-byte Folded Reload
+; SI-NEXT: v_and_b32_e32 v1, 0xff, v16
+; SI-NEXT: buffer_load_dword v3, off, s[0:3], s32 offset:232 ; 4-byte Folded Reload
; SI-NEXT: s_waitcnt vmcnt(1)
; SI-NEXT: v_lshlrev_b32_e32 v2, 8, v2
; SI-NEXT: v_or_b32_e32 v1, v1, v2
-; SI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:220 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:236 ; 4-byte Folded Reload
; SI-NEXT: s_waitcnt vmcnt(1)
; SI-NEXT: v_lshlrev_b32_e32 v3, 24, v3
; SI-NEXT: v_and_b32_e32 v1, 0xffff, v1
@@ -192552,18 +192617,16 @@ define <128 x i8> @bitcast_v64f16_to_v128i8(<64 x half> %a, i32 %b) {
; SI-NEXT: v_or_b32_e32 v1, v1, v2
; SI-NEXT: v_add_i32_e32 v2, vcc, 0x78, v0
; SI-NEXT: buffer_store_dword v1, v2, s[0:3], 0 offen
-; SI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:268 ; 4-byte Folded Reload
-; SI-NEXT: buffer_load_dword v3, off, s[0:3], s32 offset:328 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:284 ; 4-byte Folded Reload
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_and_b32_e32 v1, 0xff, v15
+; SI-NEXT: v_lshlrev_b32_e32 v3, 24, v42
; SI-NEXT: v_add_i32_e32 v0, vcc, 0x7c, v0
-; SI-NEXT: s_waitcnt vmcnt(1)
+; SI-NEXT: s_waitcnt vmcnt(0)
; SI-NEXT: v_lshlrev_b32_e32 v2, 8, v2
; SI-NEXT: v_or_b32_e32 v1, v1, v2
-; SI-NEXT: v_and_b32_e32 v2, 0xff, v45
+; SI-NEXT: v_and_b32_e32 v2, 0xff, v55
; SI-NEXT: v_lshlrev_b32_e32 v2, 16, v2
-; SI-NEXT: s_waitcnt vmcnt(0)
-; SI-NEXT: v_lshlrev_b32_e32 v3, 24, v3
; SI-NEXT: v_and_b32_e32 v1, 0xffff, v1
; SI-NEXT: v_or_b32_e32 v2, v3, v2
; SI-NEXT: v_or_b32_e32 v1, v1, v2
@@ -192607,54 +192670,56 @@ define <128 x i8> @bitcast_v64f16_to_v128i8(<64 x half> %a, i32 %b) {
; VI-NEXT: buffer_store_dword v62, off, s[0:3], s32 offset:16 ; 4-byte Folded Spill
; VI-NEXT: buffer_store_dword v63, off, s[0:3], s32 offset:12 ; 4-byte Folded Spill
; VI-NEXT: buffer_load_dword v31, off, s[0:3], s32 offset:8
-; VI-NEXT: buffer_load_dword v61, off, s[0:3], s32 offset:4
-; VI-NEXT: buffer_load_dword v60, off, s[0:3], s32
+; VI-NEXT: buffer_load_dword v62, off, s[0:3], s32 offset:4
+; VI-NEXT: buffer_load_dword v61, off, s[0:3], s32
+; VI-NEXT: v_lshrrev_b32_e32 v32, 16, v16
+; VI-NEXT: buffer_store_dword v32, off, s[0:3], s32 offset:168 ; 4-byte Folded Spill
+; VI-NEXT: v_lshrrev_b32_e32 v32, 16, v15
+; VI-NEXT: buffer_store_dword v32, off, s[0:3], s32 offset:176 ; 4-byte Folded Spill
; VI-NEXT: v_lshrrev_b32_e32 v32, 16, v13
-; VI-NEXT: buffer_store_dword v32, off, s[0:3], s32 offset:160 ; 4-byte Folded Spill
-; VI-NEXT: v_lshrrev_b32_e32 v32, 16, v9
-; VI-NEXT: buffer_store_dword v32, off, s[0:3], s32 offset:164 ; 4-byte Folded Spill
+; VI-NEXT: buffer_store_dword v32, off, s[0:3], s32 offset:184 ; 4-byte Folded Spill
+; VI-NEXT: v_lshrrev_b32_e32 v32, 16, v10
+; VI-NEXT: buffer_store_dword v32, off, s[0:3], s32 offset:196 ; 4-byte Folded Spill
; VI-NEXT: v_lshrrev_b32_e32 v32, 16, v7
-; VI-NEXT: buffer_store_dword v32, off, s[0:3], s32 offset:168 ; 4-byte Folded Spill
-; VI-NEXT: v_lshrrev_b32_e32 v32, 16, v5
-; VI-NEXT: buffer_store_dword v32, off, s[0:3], s32 offset:172 ; 4-byte Folded Spill
-; VI-NEXT: v_lshrrev_b32_e32 v32, 16, v3
; VI-NEXT: buffer_store_dword v32, off, s[0:3], s32 offset:180 ; 4-byte Folded Spill
+; VI-NEXT: v_lshrrev_b32_e32 v32, 16, v3
+; VI-NEXT: buffer_store_dword v32, off, s[0:3], s32 offset:192 ; 4-byte Folded Spill
; VI-NEXT: v_lshrrev_b32_e32 v32, 16, v1
-; VI-NEXT: buffer_store_dword v32, off, s[0:3], s32 offset:176 ; 4-byte Folded Spill
+; VI-NEXT: buffer_store_dword v32, off, s[0:3], s32 offset:188 ; 4-byte Folded Spill
; VI-NEXT: v_lshrrev_b32_e32 v32, 16, v30
; VI-NEXT: buffer_store_dword v32, off, s[0:3], s32 offset:76 ; 4-byte Folded Spill
; VI-NEXT: v_lshrrev_b32_e32 v32, 16, v29
-; VI-NEXT: buffer_store_dword v32, off, s[0:3], s32 offset:152 ; 4-byte Folded Spill
+; VI-NEXT: buffer_store_dword v32, off, s[0:3], s32 offset:156 ; 4-byte Folded Spill
+; VI-NEXT: v_lshrrev_b32_e32 v32, 16, v28
+; VI-NEXT: v_lshrrev_b32_e32 v48, 16, v14
+; VI-NEXT: v_lshrrev_b32_e32 v41, 16, v12
+; VI-NEXT: v_lshrrev_b32_e32 v57, 16, v11
+; VI-NEXT: v_lshrrev_b32_e32 v63, 16, v9
+; VI-NEXT: v_lshrrev_b32_e32 v45, 16, v8
+; VI-NEXT: v_lshrrev_b32_e32 v44, 16, v6
+; VI-NEXT: v_lshrrev_b32_e32 v47, 16, v5
+; VI-NEXT: v_lshrrev_b32_e32 v37, 16, v4
+; VI-NEXT: v_lshrrev_b32_e32 v49, 16, v2
+; VI-NEXT: buffer_store_dword v32, off, s[0:3], s32 offset:112 ; 4-byte Folded Spill
; VI-NEXT: v_lshrrev_b32_e32 v32, 16, v27
-; VI-NEXT: buffer_store_dword v32, off, s[0:3], s32 offset:148 ; 4-byte Folded Spill
-; VI-NEXT: v_lshrrev_b32_e32 v32, 16, v25
-; VI-NEXT: v_lshrrev_b32_e32 v48, 16, v16
-; VI-NEXT: v_lshrrev_b32_e32 v39, 16, v15
-; VI-NEXT: v_lshrrev_b32_e32 v33, 16, v14
-; VI-NEXT: v_lshrrev_b32_e32 v54, 16, v12
-; VI-NEXT: v_lshrrev_b32_e32 v62, 16, v11
-; VI-NEXT: v_lshrrev_b32_e32 v63, 16, v10
-; VI-NEXT: v_lshrrev_b32_e32 v37, 16, v8
-; VI-NEXT: v_lshrrev_b32_e32 v47, 16, v6
-; VI-NEXT: v_lshrrev_b32_e32 v44, 16, v4
-; VI-NEXT: v_lshrrev_b32_e32 v55, 16, v2
-; VI-NEXT: v_lshrrev_b32_e32 v56, 16, v28
; VI-NEXT: v_lshrrev_b32_e32 v58, 16, v26
-; VI-NEXT: buffer_store_dword v32, off, s[0:3], s32 offset:156 ; 4-byte Folded Spill
-; VI-NEXT: v_lshrrev_b32_e32 v57, 16, v24
-; VI-NEXT: v_lshrrev_b32_e32 v32, 16, v23
-; VI-NEXT: v_lshrrev_b32_e32 v59, 16, v22
-; VI-NEXT: v_lshrrev_b32_e32 v38, 16, v21
-; VI-NEXT: v_lshrrev_b32_e32 v53, 16, v20
-; VI-NEXT: v_lshrrev_b32_e32 v43, 16, v19
-; VI-NEXT: v_lshrrev_b32_e32 v41, 16, v18
-; VI-NEXT: v_lshrrev_b32_e32 v46, 16, v17
-; VI-NEXT: buffer_store_dword v32, off, s[0:3], s32 offset:144 ; 4-byte Folded Spill
-; VI-NEXT: ; implicit-def: $vgpr52
-; VI-NEXT: ; implicit-def: $vgpr45
-; VI-NEXT: ; implicit-def: $vgpr42
+; VI-NEXT: v_lshrrev_b32_e32 v35, 16, v25
+; VI-NEXT: v_lshrrev_b32_e32 v55, 16, v24
+; VI-NEXT: v_lshrrev_b32_e32 v38, 16, v23
+; VI-NEXT: v_lshrrev_b32_e32 v34, 16, v22
+; VI-NEXT: v_lshrrev_b32_e32 v53, 16, v21
+; VI-NEXT: v_lshrrev_b32_e32 v46, 16, v20
+; VI-NEXT: v_lshrrev_b32_e32 v39, 16, v19
+; VI-NEXT: v_lshrrev_b32_e32 v54, 16, v18
+; VI-NEXT: buffer_store_dword v32, off, s[0:3], s32 offset:160 ; 4-byte Folded Spill
; VI-NEXT: s_waitcnt vmcnt(13)
; VI-NEXT: v_cmp_ne_u32_e32 vcc, 0, v31
+; VI-NEXT: s_waitcnt vmcnt(12)
+; VI-NEXT: v_lshrrev_b32_e32 v51, 16, v62
+; VI-NEXT: s_waitcnt vmcnt(11)
+; VI-NEXT: v_lshrrev_b32_e32 v43, 16, v61
+; VI-NEXT: v_lshrrev_b32_e32 v31, 16, v17
+; VI-NEXT: buffer_store_dword v31, off, s[0:3], s32 offset:164 ; 4-byte Folded Spill
; VI-NEXT: ; implicit-def: $vgpr31
; VI-NEXT: ; kill: killed $vgpr31
; VI-NEXT: ; implicit-def: $vgpr31
@@ -192707,6 +192772,7 @@ define <128 x i8> @bitcast_v64f16_to_v128i8(<64 x half> %a, i32 %b) {
; VI-NEXT: ; kill: killed $vgpr31
; VI-NEXT: ; implicit-def: $vgpr31
; VI-NEXT: ; kill: killed $vgpr31
+; VI-NEXT: ; implicit-def: $vgpr59
; VI-NEXT: ; implicit-def: $vgpr31
; VI-NEXT: ; kill: killed $vgpr31
; VI-NEXT: ; implicit-def: $vgpr31
@@ -192750,31 +192816,30 @@ define <128 x i8> @bitcast_v64f16_to_v128i8(<64 x half> %a, i32 %b) {
; VI-NEXT: ; implicit-def: $vgpr31
; VI-NEXT: ; kill: killed $vgpr31
; VI-NEXT: ; implicit-def: $vgpr31
-; VI-NEXT: s_waitcnt vmcnt(12)
-; VI-NEXT: v_lshrrev_b32_e32 v36, 16, v61
-; VI-NEXT: s_waitcnt vmcnt(11)
-; VI-NEXT: v_lshrrev_b32_e32 v51, 16, v60
-; VI-NEXT: ; kill: killed $vgpr31
-; VI-NEXT: ; implicit-def: $vgpr31
-; VI-NEXT: buffer_store_dword v31, off, s[0:3], s32 offset:136 ; 4-byte Folded Spill
+; VI-NEXT: ; implicit-def: $vgpr56
+; VI-NEXT: ; implicit-def: $vgpr42
+; VI-NEXT: ; implicit-def: $vgpr32
+; VI-NEXT: buffer_store_dword v32, off, s[0:3], s32 offset:148 ; 4-byte Folded Spill
+; VI-NEXT: buffer_store_dword v33, off, s[0:3], s32 offset:152 ; 4-byte Folded Spill
+; VI-NEXT: ; implicit-def: $vgpr32
; VI-NEXT: buffer_store_dword v32, off, s[0:3], s32 offset:140 ; 4-byte Folded Spill
-; VI-NEXT: ; implicit-def: $vgpr31
-; VI-NEXT: buffer_store_dword v31, off, s[0:3], s32 offset:128 ; 4-byte Folded Spill
+; VI-NEXT: buffer_store_dword v33, off, s[0:3], s32 offset:144 ; 4-byte Folded Spill
+; VI-NEXT: ; implicit-def: $vgpr32
; VI-NEXT: buffer_store_dword v32, off, s[0:3], s32 offset:132 ; 4-byte Folded Spill
-; VI-NEXT: ; implicit-def: $vgpr31
-; VI-NEXT: buffer_store_dword v31, off, s[0:3], s32 offset:120 ; 4-byte Folded Spill
-; VI-NEXT: buffer_store_dword v32, off, s[0:3], s32 offset:124 ; 4-byte Folded Spill
-; VI-NEXT: ; implicit-def: $vgpr31
-; VI-NEXT: buffer_store_dword v31, off, s[0:3], s32 offset:112 ; 4-byte Folded Spill
+; VI-NEXT: buffer_store_dword v33, off, s[0:3], s32 offset:136 ; 4-byte Folded Spill
+; VI-NEXT: ; implicit-def: $vgpr32
; VI-NEXT: buffer_store_dword v32, off, s[0:3], s32 offset:116 ; 4-byte Folded Spill
-; VI-NEXT: ; implicit-def: $vgpr31
-; VI-NEXT: buffer_store_dword v31, off, s[0:3], s32 offset:96 ; 4-byte Folded Spill
-; VI-NEXT: buffer_store_dword v32, off, s[0:3], s32 offset:100 ; 4-byte Folded Spill
+; VI-NEXT: buffer_store_dword v33, off, s[0:3], s32 offset:120 ; 4-byte Folded Spill
+; VI-NEXT: ; implicit-def: $vgpr32
+; VI-NEXT: buffer_store_dword v32, off, s[0:3], s32 offset:96 ; 4-byte Folded Spill
+; VI-NEXT: buffer_store_dword v33, off, s[0:3], s32 offset:100 ; 4-byte Folded Spill
; VI-NEXT: ; implicit-def: $vgpr50
-; VI-NEXT: ; implicit-def: $vgpr31
+; VI-NEXT: ; implicit-def: $vgpr33
+; VI-NEXT: ; implicit-def: $vgpr52
; VI-NEXT: ; implicit-def: $vgpr40
-; VI-NEXT: ; implicit-def: $vgpr49
-; VI-NEXT: ; implicit-def: $vgpr35
+; VI-NEXT: ; implicit-def: $vgpr32
+; VI-NEXT: buffer_store_dword v32, off, s[0:3], s32 offset:124 ; 4-byte Folded Spill
+; VI-NEXT: buffer_store_dword v33, off, s[0:3], s32 offset:128 ; 4-byte Folded Spill
; VI-NEXT: ; implicit-def: $vgpr32
; VI-NEXT: buffer_store_dword v32, off, s[0:3], s32 offset:104 ; 4-byte Folded Spill
; VI-NEXT: buffer_store_dword v33, off, s[0:3], s32 offset:108 ; 4-byte Folded Spill
@@ -192789,770 +192854,773 @@ define <128 x i8> @bitcast_v64f16_to_v128i8(<64 x half> %a, i32 %b) {
; VI-NEXT: s_cbranch_execz .LBB94_2
; VI-NEXT: ; %bb.1: ; %cmp.false
; VI-NEXT: v_lshrrev_b32_e32 v31, 24, v16
-; VI-NEXT: buffer_store_dword v31, off, s[0:3], s32 offset:232 ; 4-byte Folded Spill
+; VI-NEXT: buffer_store_dword v31, off, s[0:3], s32 offset:252 ; 4-byte Folded Spill
; VI-NEXT: v_lshrrev_b32_e32 v31, 8, v16
-; VI-NEXT: buffer_store_dword v31, off, s[0:3], s32 offset:184 ; 4-byte Folded Spill
+; VI-NEXT: buffer_store_dword v31, off, s[0:3], s32 offset:204 ; 4-byte Folded Spill
; VI-NEXT: v_lshrrev_b32_e32 v31, 8, v15
-; VI-NEXT: buffer_store_dword v31, off, s[0:3], s32 offset:188 ; 4-byte Folded Spill
+; VI-NEXT: buffer_store_dword v31, off, s[0:3], s32 offset:208 ; 4-byte Folded Spill
; VI-NEXT: v_lshrrev_b32_e32 v31, 24, v14
-; VI-NEXT: buffer_store_dword v31, off, s[0:3], s32 offset:256 ; 4-byte Folded Spill
+; VI-NEXT: buffer_store_dword v31, off, s[0:3], s32 offset:276 ; 4-byte Folded Spill
; VI-NEXT: v_lshrrev_b32_e32 v31, 8, v14
-; VI-NEXT: buffer_store_dword v31, off, s[0:3], s32 offset:192 ; 4-byte Folded Spill
+; VI-NEXT: buffer_store_dword v31, off, s[0:3], s32 offset:212 ; 4-byte Folded Spill
; VI-NEXT: v_lshrrev_b32_e32 v31, 8, v13
-; VI-NEXT: buffer_store_dword v31, off, s[0:3], s32 offset:196 ; 4-byte Folded Spill
+; VI-NEXT: buffer_store_dword v31, off, s[0:3], s32 offset:216 ; 4-byte Folded Spill
; VI-NEXT: v_lshrrev_b32_e32 v31, 24, v12
-; VI-NEXT: buffer_store_dword v31, off, s[0:3], s32 offset:284 ; 4-byte Folded Spill
+; VI-NEXT: buffer_store_dword v31, off, s[0:3], s32 offset:300 ; 4-byte Folded Spill
; VI-NEXT: v_lshrrev_b32_e32 v31, 8, v12
-; VI-NEXT: buffer_store_dword v31, off, s[0:3], s32 offset:200 ; 4-byte Folded Spill
+; VI-NEXT: buffer_store_dword v31, off, s[0:3], s32 offset:220 ; 4-byte Folded Spill
; VI-NEXT: v_lshrrev_b32_e32 v31, 8, v11
-; VI-NEXT: buffer_store_dword v31, off, s[0:3], s32 offset:204 ; 4-byte Folded Spill
+; VI-NEXT: buffer_store_dword v31, off, s[0:3], s32 offset:224 ; 4-byte Folded Spill
; VI-NEXT: v_lshrrev_b32_e32 v31, 24, v10
-; VI-NEXT: buffer_store_dword v31, off, s[0:3], s32 offset:308 ; 4-byte Folded Spill
+; VI-NEXT: buffer_store_dword v31, off, s[0:3], s32 offset:320 ; 4-byte Folded Spill
; VI-NEXT: v_lshrrev_b32_e32 v31, 8, v10
-; VI-NEXT: buffer_store_dword v31, off, s[0:3], s32 offset:216 ; 4-byte Folded Spill
+; VI-NEXT: buffer_store_dword v31, off, s[0:3], s32 offset:236 ; 4-byte Folded Spill
; VI-NEXT: v_lshrrev_b32_e32 v31, 8, v9
-; VI-NEXT: buffer_store_dword v31, off, s[0:3], s32 offset:224 ; 4-byte Folded Spill
+; VI-NEXT: buffer_store_dword v31, off, s[0:3], s32 offset:244 ; 4-byte Folded Spill
; VI-NEXT: v_lshrrev_b32_e32 v31, 24, v8
-; VI-NEXT: buffer_store_dword v31, off, s[0:3], s32 offset:328 ; 4-byte Folded Spill
+; VI-NEXT: buffer_store_dword v31, off, s[0:3], s32 offset:340 ; 4-byte Folded Spill
; VI-NEXT: v_lshrrev_b32_e32 v31, 8, v8
-; VI-NEXT: buffer_store_dword v31, off, s[0:3], s32 offset:240 ; 4-byte Folded Spill
+; VI-NEXT: buffer_store_dword v31, off, s[0:3], s32 offset:260 ; 4-byte Folded Spill
; VI-NEXT: v_lshrrev_b32_e32 v31, 8, v7
-; VI-NEXT: buffer_store_dword v31, off, s[0:3], s32 offset:248 ; 4-byte Folded Spill
+; VI-NEXT: buffer_store_dword v31, off, s[0:3], s32 offset:268 ; 4-byte Folded Spill
; VI-NEXT: v_lshrrev_b32_e32 v31, 24, v6
-; VI-NEXT: buffer_store_dword v31, off, s[0:3], s32 offset:356 ; 4-byte Folded Spill
+; VI-NEXT: buffer_store_dword v31, off, s[0:3], s32 offset:368 ; 4-byte Folded Spill
; VI-NEXT: v_lshrrev_b32_e32 v31, 8, v6
-; VI-NEXT: buffer_store_dword v31, off, s[0:3], s32 offset:272 ; 4-byte Folded Spill
+; VI-NEXT: buffer_store_dword v31, off, s[0:3], s32 offset:288 ; 4-byte Folded Spill
; VI-NEXT: v_lshrrev_b32_e32 v31, 8, v5
-; VI-NEXT: buffer_store_dword v31, off, s[0:3], s32 offset:280 ; 4-byte Folded Spill
+; VI-NEXT: buffer_store_dword v31, off, s[0:3], s32 offset:296 ; 4-byte Folded Spill
; VI-NEXT: v_lshrrev_b32_e32 v31, 24, v4
-; VI-NEXT: buffer_store_dword v31, off, s[0:3], s32 offset:368 ; 4-byte Folded Spill
+; VI-NEXT: buffer_store_dword v31, off, s[0:3], s32 offset:380 ; 4-byte Folded Spill
; VI-NEXT: v_lshrrev_b32_e32 v31, 8, v4
-; VI-NEXT: buffer_store_dword v31, off, s[0:3], s32 offset:304 ; 4-byte Folded Spill
+; VI-NEXT: buffer_store_dword v31, off, s[0:3], s32 offset:316 ; 4-byte Folded Spill
; VI-NEXT: v_lshrrev_b32_e32 v31, 8, v3
-; VI-NEXT: buffer_store_dword v31, off, s[0:3], s32 offset:312 ; 4-byte Folded Spill
+; VI-NEXT: buffer_store_dword v31, off, s[0:3], s32 offset:324 ; 4-byte Folded Spill
; VI-NEXT: v_lshrrev_b32_e32 v31, 24, v2
-; VI-NEXT: buffer_store_dword v31, off, s[0:3], s32 offset:384 ; 4-byte Folded Spill
+; VI-NEXT: buffer_store_dword v31, off, s[0:3], s32 offset:396 ; 4-byte Folded Spill
; VI-NEXT: v_lshrrev_b32_e32 v31, 8, v2
-; VI-NEXT: buffer_store_dword v31, off, s[0:3], s32 offset:336 ; 4-byte Folded Spill
+; VI-NEXT: buffer_store_dword v31, off, s[0:3], s32 offset:348 ; 4-byte Folded Spill
; VI-NEXT: v_lshrrev_b32_e32 v31, 8, v1
-; VI-NEXT: buffer_store_dword v31, off, s[0:3], s32 offset:344 ; 4-byte Folded Spill
-; VI-NEXT: v_lshrrev_b32_e32 v31, 24, v61
-; VI-NEXT: buffer_store_dword v31, off, s[0:3], s32 offset:264 ; 4-byte Folded Spill
+; VI-NEXT: buffer_store_dword v31, off, s[0:3], s32 offset:356 ; 4-byte Folded Spill
+; VI-NEXT: v_lshrrev_b32_e32 v31, 24, v62
+; VI-NEXT: buffer_store_dword v31, off, s[0:3], s32 offset:284 ; 4-byte Folded Spill
+; VI-NEXT: v_lshrrev_b32_e32 v31, 8, v62
+; VI-NEXT: buffer_store_dword v31, off, s[0:3], s32 offset:228 ; 4-byte Folded Spill
; VI-NEXT: v_lshrrev_b32_e32 v31, 8, v61
-; VI-NEXT: buffer_store_dword v31, off, s[0:3], s32 offset:208 ; 4-byte Folded Spill
-; VI-NEXT: v_lshrrev_b32_e32 v31, 8, v60
-; VI-NEXT: buffer_store_dword v31, off, s[0:3], s32 offset:212 ; 4-byte Folded Spill
+; VI-NEXT: buffer_store_dword v31, off, s[0:3], s32 offset:232 ; 4-byte Folded Spill
; VI-NEXT: v_lshrrev_b32_e32 v31, 24, v30
-; VI-NEXT: buffer_store_dword v31, off, s[0:3], s32 offset:296 ; 4-byte Folded Spill
+; VI-NEXT: buffer_store_dword v31, off, s[0:3], s32 offset:308 ; 4-byte Folded Spill
; VI-NEXT: v_lshrrev_b32_e32 v31, 8, v30
-; VI-NEXT: buffer_store_dword v31, off, s[0:3], s32 offset:220 ; 4-byte Folded Spill
+; VI-NEXT: buffer_store_dword v31, off, s[0:3], s32 offset:240 ; 4-byte Folded Spill
; VI-NEXT: v_lshrrev_b32_e32 v31, 8, v29
-; VI-NEXT: buffer_store_dword v31, off, s[0:3], s32 offset:228 ; 4-byte Folded Spill
+; VI-NEXT: buffer_store_dword v31, off, s[0:3], s32 offset:248 ; 4-byte Folded Spill
; VI-NEXT: v_lshrrev_b32_e32 v31, 24, v28
-; VI-NEXT: buffer_store_dword v31, off, s[0:3], s32 offset:320 ; 4-byte Folded Spill
+; VI-NEXT: buffer_store_dword v31, off, s[0:3], s32 offset:332 ; 4-byte Folded Spill
; VI-NEXT: v_lshrrev_b32_e32 v31, 8, v28
-; VI-NEXT: buffer_store_dword v31, off, s[0:3], s32 offset:244 ; 4-byte Folded Spill
+; VI-NEXT: buffer_store_dword v31, off, s[0:3], s32 offset:264 ; 4-byte Folded Spill
; VI-NEXT: v_lshrrev_b32_e32 v31, 8, v27
-; VI-NEXT: buffer_store_dword v31, off, s[0:3], s32 offset:252 ; 4-byte Folded Spill
+; VI-NEXT: buffer_store_dword v31, off, s[0:3], s32 offset:272 ; 4-byte Folded Spill
; VI-NEXT: v_lshrrev_b32_e32 v31, 24, v26
-; VI-NEXT: buffer_store_dword v31, off, s[0:3], s32 offset:332 ; 4-byte Folded Spill
+; VI-NEXT: buffer_store_dword v31, off, s[0:3], s32 offset:344 ; 4-byte Folded Spill
; VI-NEXT: v_lshrrev_b32_e32 v31, 8, v26
-; VI-NEXT: buffer_store_dword v31, off, s[0:3], s32 offset:276 ; 4-byte Folded Spill
-; VI-NEXT: v_lshrrev_b32_e32 v31, 8, v25
; VI-NEXT: buffer_store_dword v31, off, s[0:3], s32 offset:292 ; 4-byte Folded Spill
+; VI-NEXT: v_lshrrev_b32_e32 v31, 8, v25
+; VI-NEXT: buffer_store_dword v31, off, s[0:3], s32 offset:304 ; 4-byte Folded Spill
; VI-NEXT: v_lshrrev_b32_e32 v31, 24, v24
-; VI-NEXT: buffer_store_dword v31, off, s[0:3], s32 offset:348 ; 4-byte Folded Spill
+; VI-NEXT: buffer_store_dword v31, off, s[0:3], s32 offset:360 ; 4-byte Folded Spill
; VI-NEXT: v_lshrrev_b32_e32 v31, 8, v24
-; VI-NEXT: buffer_store_dword v31, off, s[0:3], s32 offset:300 ; 4-byte Folded Spill
+; VI-NEXT: buffer_store_dword v31, off, s[0:3], s32 offset:312 ; 4-byte Folded Spill
; VI-NEXT: v_lshrrev_b32_e32 v31, 8, v23
-; VI-NEXT: buffer_store_dword v31, off, s[0:3], s32 offset:316 ; 4-byte Folded Spill
+; VI-NEXT: buffer_store_dword v31, off, s[0:3], s32 offset:328 ; 4-byte Folded Spill
; VI-NEXT: v_lshrrev_b32_e32 v31, 24, v22
-; VI-NEXT: buffer_store_dword v31, off, s[0:3], s32 offset:364 ; 4-byte Folded Spill
+; VI-NEXT: buffer_store_dword v31, off, s[0:3], s32 offset:376 ; 4-byte Folded Spill
; VI-NEXT: v_lshrrev_b32_e32 v31, 8, v22
-; VI-NEXT: buffer_store_dword v31, off, s[0:3], s32 offset:324 ; 4-byte Folded Spill
+; VI-NEXT: buffer_store_dword v31, off, s[0:3], s32 offset:336 ; 4-byte Folded Spill
; VI-NEXT: v_lshrrev_b32_e32 v31, 8, v21
-; VI-NEXT: buffer_store_dword v31, off, s[0:3], s32 offset:340 ; 4-byte Folded Spill
+; VI-NEXT: buffer_store_dword v31, off, s[0:3], s32 offset:352 ; 4-byte Folded Spill
; VI-NEXT: v_lshrrev_b32_e32 v31, 24, v20
-; VI-NEXT: buffer_store_dword v31, off, s[0:3], s32 offset:380 ; 4-byte Folded Spill
+; VI-NEXT: buffer_store_dword v31, off, s[0:3], s32 offset:392 ; 4-byte Folded Spill
; VI-NEXT: v_lshrrev_b32_e32 v31, 8, v20
-; VI-NEXT: buffer_store_dword v31, off, s[0:3], s32 offset:352 ; 4-byte Folded Spill
+; VI-NEXT: buffer_store_dword v31, off, s[0:3], s32 offset:364 ; 4-byte Folded Spill
; VI-NEXT: v_lshrrev_b32_e32 v31, 8, v19
-; VI-NEXT: buffer_store_dword v31, off, s[0:3], s32 offset:360 ; 4-byte Folded Spill
-; VI-NEXT: v_lshrrev_b32_e32 v31, 24, v18
-; VI-NEXT: buffer_store_dword v31, off, s[0:3], s32 offset:388 ; 4-byte Folded Spill
-; VI-NEXT: v_lshrrev_b32_e32 v31, 8, v18
; VI-NEXT: buffer_store_dword v31, off, s[0:3], s32 offset:372 ; 4-byte Folded Spill
+; VI-NEXT: v_lshrrev_b32_e32 v31, 8, v18
+; VI-NEXT: buffer_store_dword v31, off, s[0:3], s32 offset:384 ; 4-byte Folded Spill
; VI-NEXT: v_lshrrev_b32_e32 v31, 8, v17
-; VI-NEXT: buffer_store_dword v31, off, s[0:3], s32 offset:376 ; 4-byte Folded Spill
+; VI-NEXT: buffer_store_dword v31, off, s[0:3], s32 offset:388 ; 4-byte Folded Spill
; VI-NEXT: v_lshrrev_b64 v[31:32], 24, v[15:16]
; VI-NEXT: buffer_store_dword v31, off, s[0:3], s32 offset:96 ; 4-byte Folded Spill
; VI-NEXT: buffer_store_dword v32, off, s[0:3], s32 offset:100 ; 4-byte Folded Spill
; VI-NEXT: v_lshrrev_b64 v[31:32], 24, v[13:14]
-; VI-NEXT: buffer_store_dword v31, off, s[0:3], s32 offset:112 ; 4-byte Folded Spill
-; VI-NEXT: buffer_store_dword v32, off, s[0:3], s32 offset:116 ; 4-byte Folded Spill
+; VI-NEXT: buffer_store_dword v31, off, s[0:3], s32 offset:116 ; 4-byte Folded Spill
+; VI-NEXT: buffer_store_dword v32, off, s[0:3], s32 offset:120 ; 4-byte Folded Spill
; VI-NEXT: v_lshrrev_b64 v[31:32], 24, v[11:12]
-; VI-NEXT: buffer_store_dword v31, off, s[0:3], s32 offset:120 ; 4-byte Folded Spill
-; VI-NEXT: buffer_store_dword v32, off, s[0:3], s32 offset:124 ; 4-byte Folded Spill
+; VI-NEXT: buffer_store_dword v31, off, s[0:3], s32 offset:132 ; 4-byte Folded Spill
+; VI-NEXT: buffer_store_dword v32, off, s[0:3], s32 offset:136 ; 4-byte Folded Spill
; VI-NEXT: v_lshrrev_b64 v[31:32], 24, v[9:10]
-; VI-NEXT: buffer_store_dword v31, off, s[0:3], s32 offset:128 ; 4-byte Folded Spill
-; VI-NEXT: buffer_store_dword v32, off, s[0:3], s32 offset:132 ; 4-byte Folded Spill
+; VI-NEXT: buffer_store_dword v31, off, s[0:3], s32 offset:140 ; 4-byte Folded Spill
+; VI-NEXT: buffer_store_dword v32, off, s[0:3], s32 offset:144 ; 4-byte Folded Spill
; VI-NEXT: v_lshrrev_b64 v[31:32], 24, v[7:8]
-; VI-NEXT: buffer_store_dword v31, off, s[0:3], s32 offset:136 ; 4-byte Folded Spill
-; VI-NEXT: buffer_store_dword v32, off, s[0:3], s32 offset:140 ; 4-byte Folded Spill
-; VI-NEXT: v_mov_b32_e32 v31, v33
-; VI-NEXT: v_mov_b32_e32 v33, v43
+; VI-NEXT: buffer_store_dword v31, off, s[0:3], s32 offset:148 ; 4-byte Folded Spill
+; VI-NEXT: buffer_store_dword v32, off, s[0:3], s32 offset:152 ; 4-byte Folded Spill
+; VI-NEXT: v_mov_b32_e32 v31, v43
; VI-NEXT: v_lshrrev_b64 v[42:43], 24, v[5:6]
-; VI-NEXT: v_mov_b32_e32 v43, v33
-; VI-NEXT: v_mov_b32_e32 v33, v46
-; VI-NEXT: v_lshrrev_b64 v[45:46], 24, v[3:4]
-; VI-NEXT: v_mov_b32_e32 v46, v33
-; VI-NEXT: v_mov_b32_e32 v33, v53
-; VI-NEXT: v_lshrrev_b64 v[52:53], 24, v[1:2]
-; VI-NEXT: v_mov_b32_e32 v53, v33
-; VI-NEXT: v_lshrrev_b64 v[32:33], 24, v[60:61]
+; VI-NEXT: v_mov_b32_e32 v32, v57
+; VI-NEXT: v_lshrrev_b64 v[56:57], 24, v[3:4]
+; VI-NEXT: v_mov_b32_e32 v43, v31
+; VI-NEXT: v_mov_b32_e32 v57, v32
+; VI-NEXT: v_lshrrev_b64 v[31:32], 24, v[1:2]
+; VI-NEXT: v_lshrrev_b64 v[32:33], 24, v[61:62]
; VI-NEXT: buffer_store_dword v32, off, s[0:3], s32 offset:80 ; 4-byte Folded Spill
; VI-NEXT: buffer_store_dword v33, off, s[0:3], s32 offset:84 ; 4-byte Folded Spill
; VI-NEXT: v_lshrrev_b64 v[32:33], 24, v[29:30]
-; VI-NEXT: v_lshrrev_b64 v[34:35], 24, v[27:28]
; VI-NEXT: buffer_store_dword v32, off, s[0:3], s32 offset:88 ; 4-byte Folded Spill
; VI-NEXT: buffer_store_dword v33, off, s[0:3], s32 offset:92 ; 4-byte Folded Spill
-; VI-NEXT: buffer_store_dword v34, off, s[0:3], s32 offset:104 ; 4-byte Folded Spill
-; VI-NEXT: buffer_store_dword v35, off, s[0:3], s32 offset:108 ; 4-byte Folded Spill
-; VI-NEXT: v_mov_b32_e32 v33, v36
+; VI-NEXT: v_lshrrev_b64 v[59:60], 24, v[27:28]
+; VI-NEXT: v_mov_b32_e32 v32, v35
; VI-NEXT: v_lshrrev_b64 v[35:36], 24, v[25:26]
-; VI-NEXT: v_lshrrev_b64 v[49:50], 24, v[23:24]
-; VI-NEXT: v_mov_b32_e32 v36, v33
-; VI-NEXT: v_mov_b32_e32 v33, v41
-; VI-NEXT: v_lshrrev_b64 v[40:41], 24, v[21:22]
-; VI-NEXT: v_mov_b32_e32 v34, v51
+; VI-NEXT: buffer_store_dword v59, off, s[0:3], s32 offset:104 ; 4-byte Folded Spill
+; VI-NEXT: buffer_store_dword v60, off, s[0:3], s32 offset:108 ; 4-byte Folded Spill
+; VI-NEXT: buffer_store_dword v35, off, s[0:3], s32 offset:124 ; 4-byte Folded Spill
+; VI-NEXT: buffer_store_dword v36, off, s[0:3], s32 offset:128 ; 4-byte Folded Spill
+; VI-NEXT: v_mov_b32_e32 v35, v32
+; VI-NEXT: v_mov_b32_e32 v32, v41
+; VI-NEXT: v_lshrrev_b64 v[40:41], 24, v[23:24]
+; VI-NEXT: v_mov_b32_e32 v41, v32
+; VI-NEXT: v_mov_b32_e32 v32, v53
+; VI-NEXT: v_lshrrev_b64 v[52:53], 24, v[21:22]
+; VI-NEXT: v_lshrrev_b32_e32 v50, 24, v18
+; VI-NEXT: v_mov_b32_e32 v53, v32
+; VI-NEXT: v_mov_b32_e32 v32, v34
+; VI-NEXT: v_lshrrev_b64 v[33:34], 24, v[19:20]
+; VI-NEXT: v_mov_b32_e32 v59, v50
+; VI-NEXT: v_mov_b32_e32 v34, v32
+; VI-NEXT: v_mov_b32_e32 v32, v51
; VI-NEXT: v_lshrrev_b64 v[50:51], 24, v[17:18]
-; VI-NEXT: v_mov_b32_e32 v41, v33
-; VI-NEXT: v_mov_b32_e32 v33, v31
-; VI-NEXT: v_lshrrev_b64 v[31:32], 24, v[19:20]
-; VI-NEXT: v_mov_b32_e32 v51, v34
+; VI-NEXT: v_mov_b32_e32 v51, v32
; VI-NEXT: .LBB94_2: ; %Flow
; VI-NEXT: s_andn2_saveexec_b64 s[4:5], s[4:5]
; VI-NEXT: s_cbranch_execz .LBB94_4
; VI-NEXT: ; %bb.3: ; %cmp.true
-; VI-NEXT: v_mov_b32_e32 v63, 0x200
-; VI-NEXT: v_add_f16_sdwa v31, v18, v63 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_1 src1_sel:DWORD
-; VI-NEXT: buffer_store_dword v31, off, s[0:3], s32 offset:392 ; 4-byte Folded Spill
+; VI-NEXT: v_mov_b32_e32 v55, 0x200
+; VI-NEXT: v_add_f16_sdwa v31, v18, v55 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_1 src1_sel:DWORD
+; VI-NEXT: buffer_store_dword v31, off, s[0:3], s32 offset:404 ; 4-byte Folded Spill
; VI-NEXT: v_lshlrev_b32_e32 v31, 16, v31
; VI-NEXT: v_add_f16_e32 v18, 0x200, v18
; VI-NEXT: v_or_b32_e32 v32, v18, v31
-; VI-NEXT: v_add_f16_sdwa v31, v17, v63 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_1 src1_sel:DWORD
-; VI-NEXT: buffer_store_dword v31, off, s[0:3], s32 offset:400 ; 4-byte Folded Spill
+; VI-NEXT: v_add_f16_sdwa v31, v17, v55 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_1 src1_sel:DWORD
+; VI-NEXT: buffer_store_dword v31, off, s[0:3], s32 offset:164 ; 4-byte Folded Spill
; VI-NEXT: v_lshlrev_b32_e32 v31, 16, v31
; VI-NEXT: v_add_f16_e32 v17, 0x200, v17
; VI-NEXT: v_or_b32_e32 v31, v17, v31
-; VI-NEXT: buffer_store_dword v31, off, s[0:3], s32 offset:232 ; 4-byte Folded Spill
-; VI-NEXT: buffer_store_dword v32, off, s[0:3], s32 offset:236 ; 4-byte Folded Spill
-; VI-NEXT: v_add_f16_sdwa v31, v20, v63 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_1 src1_sel:DWORD
-; VI-NEXT: buffer_store_dword v31, off, s[0:3], s32 offset:404 ; 4-byte Folded Spill
+; VI-NEXT: buffer_store_dword v31, off, s[0:3], s32 offset:168 ; 4-byte Folded Spill
+; VI-NEXT: buffer_store_dword v32, off, s[0:3], s32 offset:172 ; 4-byte Folded Spill
+; VI-NEXT: v_add_f16_sdwa v31, v20, v55 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_1 src1_sel:DWORD
+; VI-NEXT: buffer_store_dword v31, off, s[0:3], s32 offset:416 ; 4-byte Folded Spill
; VI-NEXT: v_lshlrev_b32_e32 v31, 16, v31
; VI-NEXT: v_add_f16_e32 v20, 0x200, v20
; VI-NEXT: v_or_b32_e32 v32, v20, v31
-; VI-NEXT: v_add_f16_sdwa v31, v19, v63 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_1 src1_sel:DWORD
-; VI-NEXT: buffer_store_dword v31, off, s[0:3], s32 offset:408 ; 4-byte Folded Spill
+; VI-NEXT: v_add_f16_sdwa v31, v19, v55 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_1 src1_sel:DWORD
+; VI-NEXT: buffer_store_dword v31, off, s[0:3], s32 offset:400 ; 4-byte Folded Spill
; VI-NEXT: v_lshlrev_b32_e32 v31, 16, v31
; VI-NEXT: v_add_f16_e32 v19, 0x200, v19
; VI-NEXT: v_or_b32_e32 v31, v19, v31
-; VI-NEXT: buffer_store_dword v31, off, s[0:3], s32 offset:256 ; 4-byte Folded Spill
-; VI-NEXT: buffer_store_dword v32, off, s[0:3], s32 offset:260 ; 4-byte Folded Spill
-; VI-NEXT: v_add_f16_sdwa v34, v22, v63 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_1 src1_sel:DWORD
-; VI-NEXT: v_lshlrev_b32_e32 v31, 16, v34
+; VI-NEXT: buffer_store_dword v31, off, s[0:3], s32 offset:196 ; 4-byte Folded Spill
+; VI-NEXT: buffer_store_dword v32, off, s[0:3], s32 offset:200 ; 4-byte Folded Spill
+; VI-NEXT: v_add_f16_sdwa v31, v22, v55 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_1 src1_sel:DWORD
+; VI-NEXT: buffer_store_dword v31, off, s[0:3], s32 offset:408 ; 4-byte Folded Spill
+; VI-NEXT: v_lshlrev_b32_e32 v31, 16, v31
; VI-NEXT: v_add_f16_e32 v22, 0x200, v22
; VI-NEXT: v_or_b32_e32 v32, v22, v31
-; VI-NEXT: v_add_f16_sdwa v31, v21, v63 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_1 src1_sel:DWORD
-; VI-NEXT: buffer_store_dword v31, off, s[0:3], s32 offset:396 ; 4-byte Folded Spill
+; VI-NEXT: v_add_f16_sdwa v31, v21, v55 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_1 src1_sel:DWORD
+; VI-NEXT: buffer_store_dword v31, off, s[0:3], s32 offset:412 ; 4-byte Folded Spill
; VI-NEXT: v_lshlrev_b32_e32 v31, 16, v31
; VI-NEXT: v_add_f16_e32 v21, 0x200, v21
; VI-NEXT: v_or_b32_e32 v31, v21, v31
-; VI-NEXT: buffer_store_dword v31, off, s[0:3], s32 offset:264 ; 4-byte Folded Spill
-; VI-NEXT: buffer_store_dword v32, off, s[0:3], s32 offset:268 ; 4-byte Folded Spill
-; VI-NEXT: v_add_f16_sdwa v31, v24, v63 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_1 src1_sel:DWORD
-; VI-NEXT: buffer_store_dword v31, off, s[0:3], s32 offset:420 ; 4-byte Folded Spill
+; VI-NEXT: buffer_store_dword v31, off, s[0:3], s32 offset:252 ; 4-byte Folded Spill
+; VI-NEXT: buffer_store_dword v32, off, s[0:3], s32 offset:256 ; 4-byte Folded Spill
+; VI-NEXT: v_add_f16_sdwa v31, v24, v55 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_1 src1_sel:DWORD
+; VI-NEXT: buffer_store_dword v31, off, s[0:3], s32 offset:432 ; 4-byte Folded Spill
; VI-NEXT: v_lshlrev_b32_e32 v31, 16, v31
; VI-NEXT: v_add_f16_e32 v24, 0x200, v24
; VI-NEXT: v_or_b32_e32 v32, v24, v31
-; VI-NEXT: v_add_f16_sdwa v31, v23, v63 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_1 src1_sel:DWORD
-; VI-NEXT: buffer_store_dword v31, off, s[0:3], s32 offset:144 ; 4-byte Folded Spill
+; VI-NEXT: v_add_f16_sdwa v31, v23, v55 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_1 src1_sel:DWORD
+; VI-NEXT: buffer_store_dword v31, off, s[0:3], s32 offset:420 ; 4-byte Folded Spill
; VI-NEXT: v_lshlrev_b32_e32 v31, 16, v31
; VI-NEXT: v_add_f16_e32 v23, 0x200, v23
; VI-NEXT: v_or_b32_e32 v31, v23, v31
-; VI-NEXT: buffer_store_dword v31, off, s[0:3], s32 offset:284 ; 4-byte Folded Spill
-; VI-NEXT: buffer_store_dword v32, off, s[0:3], s32 offset:288 ; 4-byte Folded Spill
-; VI-NEXT: v_add_f16_sdwa v31, v26, v63 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_1 src1_sel:DWORD
-; VI-NEXT: buffer_store_dword v31, off, s[0:3], s32 offset:432 ; 4-byte Folded Spill
+; VI-NEXT: buffer_store_dword v31, off, s[0:3], s32 offset:276 ; 4-byte Folded Spill
+; VI-NEXT: buffer_store_dword v32, off, s[0:3], s32 offset:280 ; 4-byte Folded Spill
+; VI-NEXT: v_add_f16_sdwa v31, v26, v55 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_1 src1_sel:DWORD
+; VI-NEXT: buffer_store_dword v31, off, s[0:3], s32 offset:440 ; 4-byte Folded Spill
; VI-NEXT: v_lshlrev_b32_e32 v31, 16, v31
; VI-NEXT: v_add_f16_e32 v26, 0x200, v26
-; VI-NEXT: v_or_b32_e32 v36, v26, v31
-; VI-NEXT: v_add_f16_sdwa v31, v25, v63 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_1 src1_sel:DWORD
-; VI-NEXT: buffer_store_dword v31, off, s[0:3], s32 offset:156 ; 4-byte Folded Spill
+; VI-NEXT: v_or_b32_e32 v32, v26, v31
+; VI-NEXT: v_add_f16_sdwa v31, v25, v55 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_1 src1_sel:DWORD
+; VI-NEXT: buffer_store_dword v31, off, s[0:3], s32 offset:424 ; 4-byte Folded Spill
; VI-NEXT: v_lshlrev_b32_e32 v31, 16, v31
; VI-NEXT: v_add_f16_e32 v25, 0x200, v25
-; VI-NEXT: v_or_b32_e32 v35, v25, v31
-; VI-NEXT: v_add_f16_sdwa v31, v28, v63 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_1 src1_sel:DWORD
-; VI-NEXT: buffer_store_dword v31, off, s[0:3], s32 offset:440 ; 4-byte Folded Spill
+; VI-NEXT: v_or_b32_e32 v31, v25, v31
+; VI-NEXT: buffer_store_dword v31, off, s[0:3], s32 offset:124 ; 4-byte Folded Spill
+; VI-NEXT: buffer_store_dword v32, off, s[0:3], s32 offset:128 ; 4-byte Folded Spill
+; VI-NEXT: v_add_f16_sdwa v31, v28, v55 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_1 src1_sel:DWORD
+; VI-NEXT: buffer_store_dword v31, off, s[0:3], s32 offset:112 ; 4-byte Folded Spill
; VI-NEXT: v_lshlrev_b32_e32 v31, 16, v31
; VI-NEXT: v_add_f16_e32 v28, 0x200, v28
; VI-NEXT: v_or_b32_e32 v38, v28, v31
-; VI-NEXT: v_add_f16_sdwa v31, v27, v63 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_1 src1_sel:DWORD
-; VI-NEXT: buffer_store_dword v31, off, s[0:3], s32 offset:148 ; 4-byte Folded Spill
+; VI-NEXT: v_add_f16_sdwa v31, v27, v55 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_1 src1_sel:DWORD
+; VI-NEXT: buffer_store_dword v31, off, s[0:3], s32 offset:160 ; 4-byte Folded Spill
; VI-NEXT: v_lshlrev_b32_e32 v31, 16, v31
; VI-NEXT: v_add_f16_e32 v27, 0x200, v27
; VI-NEXT: v_or_b32_e32 v37, v27, v31
-; VI-NEXT: v_add_f16_sdwa v31, v30, v63 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_1 src1_sel:DWORD
+; VI-NEXT: v_add_f16_sdwa v31, v30, v55 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_1 src1_sel:DWORD
; VI-NEXT: buffer_store_dword v31, off, s[0:3], s32 offset:76 ; 4-byte Folded Spill
; VI-NEXT: v_lshlrev_b32_e32 v31, 16, v31
; VI-NEXT: v_add_f16_e32 v30, 0x200, v30
-; VI-NEXT: v_add_f16_sdwa v32, v29, v63 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_1 src1_sel:DWORD
+; VI-NEXT: v_add_f16_sdwa v32, v29, v55 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_1 src1_sel:DWORD
; VI-NEXT: v_add_f16_e32 v29, 0x200, v29
-; VI-NEXT: v_or_b32_e32 v49, v30, v31
+; VI-NEXT: v_or_b32_e32 v34, v30, v31
; VI-NEXT: v_lshlrev_b32_e32 v31, 16, v32
-; VI-NEXT: v_or_b32_e32 v48, v29, v31
-; VI-NEXT: v_add_f16_sdwa v31, v61, v63 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_1 src1_sel:DWORD
-; VI-NEXT: buffer_store_dword v32, off, s[0:3], s32 offset:152 ; 4-byte Folded Spill
+; VI-NEXT: v_or_b32_e32 v33, v29, v31
+; VI-NEXT: v_add_f16_sdwa v31, v62, v55 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_1 src1_sel:DWORD
+; VI-NEXT: buffer_store_dword v32, off, s[0:3], s32 offset:156 ; 4-byte Folded Spill
; VI-NEXT: buffer_store_dword v31, off, s[0:3], s32 offset:428 ; 4-byte Folded Spill
-; VI-NEXT: v_add_f16_e32 v61, 0x200, v61
-; VI-NEXT: v_add_f16_sdwa v32, v60, v63 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_1 src1_sel:DWORD
+; VI-NEXT: v_add_f16_e32 v62, 0x200, v62
+; VI-NEXT: v_add_f16_sdwa v32, v61, v55 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_1 src1_sel:DWORD
; VI-NEXT: v_lshlrev_b32_e32 v31, 16, v31
-; VI-NEXT: v_add_f16_e32 v60, 0x200, v60
-; VI-NEXT: v_or_b32_e32 v51, v61, v31
+; VI-NEXT: v_add_f16_e32 v61, 0x200, v61
+; VI-NEXT: v_or_b32_e32 v51, v62, v31
; VI-NEXT: v_lshlrev_b32_e32 v31, 16, v32
-; VI-NEXT: v_or_b32_e32 v50, v60, v31
-; VI-NEXT: v_add_f16_sdwa v31, v2, v63 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_1 src1_sel:DWORD
-; VI-NEXT: buffer_store_dword v32, off, s[0:3], s32 offset:416 ; 4-byte Folded Spill
-; VI-NEXT: buffer_store_dword v31, off, s[0:3], s32 offset:412 ; 4-byte Folded Spill
+; VI-NEXT: v_add_f16_sdwa v47, v2, v55 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_1 src1_sel:DWORD
+; VI-NEXT: buffer_store_dword v32, off, s[0:3], s32 offset:448 ; 4-byte Folded Spill
+; VI-NEXT: v_or_b32_e32 v50, v61, v31
; VI-NEXT: v_add_f16_e32 v2, 0x200, v2
-; VI-NEXT: v_add_f16_sdwa v32, v1, v63 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_1 src1_sel:DWORD
-; VI-NEXT: v_lshlrev_b32_e32 v31, 16, v31
+; VI-NEXT: v_add_f16_sdwa v32, v1, v55 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_1 src1_sel:DWORD
+; VI-NEXT: v_lshlrev_b32_e32 v31, 16, v47
; VI-NEXT: v_add_f16_e32 v1, 0x200, v1
; VI-NEXT: v_or_b32_e32 v53, v2, v31
; VI-NEXT: v_lshlrev_b32_e32 v31, 16, v32
; VI-NEXT: v_or_b32_e32 v52, v1, v31
-; VI-NEXT: v_add_f16_sdwa v31, v4, v63 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_1 src1_sel:DWORD
-; VI-NEXT: buffer_store_dword v32, off, s[0:3], s32 offset:176 ; 4-byte Folded Spill
-; VI-NEXT: buffer_store_dword v31, off, s[0:3], s32 offset:424 ; 4-byte Folded Spill
+; VI-NEXT: v_add_f16_sdwa v31, v4, v55 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_1 src1_sel:DWORD
+; VI-NEXT: buffer_store_dword v32, off, s[0:3], s32 offset:188 ; 4-byte Folded Spill
+; VI-NEXT: buffer_store_dword v31, off, s[0:3], s32 offset:436 ; 4-byte Folded Spill
; VI-NEXT: v_add_f16_e32 v4, 0x200, v4
-; VI-NEXT: v_add_f16_sdwa v32, v3, v63 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_1 src1_sel:DWORD
+; VI-NEXT: v_add_f16_sdwa v32, v3, v55 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_1 src1_sel:DWORD
; VI-NEXT: v_lshlrev_b32_e32 v31, 16, v31
; VI-NEXT: v_add_f16_e32 v3, 0x200, v3
; VI-NEXT: v_or_b32_e32 v46, v4, v31
; VI-NEXT: v_lshlrev_b32_e32 v31, 16, v32
+; VI-NEXT: v_add_f16_sdwa v36, v6, v55 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_1 src1_sel:DWORD
+; VI-NEXT: buffer_store_dword v32, off, s[0:3], s32 offset:192 ; 4-byte Folded Spill
; VI-NEXT: v_or_b32_e32 v45, v3, v31
-; VI-NEXT: v_add_f16_sdwa v31, v6, v63 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_1 src1_sel:DWORD
-; VI-NEXT: buffer_store_dword v32, off, s[0:3], s32 offset:180 ; 4-byte Folded Spill
-; VI-NEXT: buffer_store_dword v31, off, s[0:3], s32 offset:436 ; 4-byte Folded Spill
; VI-NEXT: v_add_f16_e32 v6, 0x200, v6
-; VI-NEXT: v_add_f16_sdwa v32, v5, v63 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_1 src1_sel:DWORD
-; VI-NEXT: v_lshlrev_b32_e32 v31, 16, v31
+; VI-NEXT: v_add_f16_sdwa v32, v5, v55 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_1 src1_sel:DWORD
+; VI-NEXT: v_lshlrev_b32_e32 v31, 16, v36
; VI-NEXT: v_add_f16_e32 v5, 0x200, v5
; VI-NEXT: v_or_b32_e32 v43, v6, v31
; VI-NEXT: v_lshlrev_b32_e32 v31, 16, v32
-; VI-NEXT: v_add_f16_sdwa v44, v8, v63 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_1 src1_sel:DWORD
-; VI-NEXT: buffer_store_dword v32, off, s[0:3], s32 offset:172 ; 4-byte Folded Spill
+; VI-NEXT: v_add_f16_sdwa v44, v8, v55 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_1 src1_sel:DWORD
+; VI-NEXT: buffer_store_dword v32, off, s[0:3], s32 offset:444 ; 4-byte Folded Spill
; VI-NEXT: v_or_b32_e32 v42, v5, v31
; VI-NEXT: v_add_f16_e32 v8, 0x200, v8
-; VI-NEXT: v_add_f16_sdwa v32, v7, v63 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_1 src1_sel:DWORD
+; VI-NEXT: v_add_f16_sdwa v32, v7, v55 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_1 src1_sel:DWORD
; VI-NEXT: v_lshlrev_b32_e32 v31, 16, v44
; VI-NEXT: v_add_f16_e32 v7, 0x200, v7
; VI-NEXT: v_or_b32_e32 v41, v8, v31
; VI-NEXT: v_lshlrev_b32_e32 v31, 16, v32
+; VI-NEXT: v_add_f16_sdwa v49, v10, v55 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_1 src1_sel:DWORD
; VI-NEXT: v_or_b32_e32 v40, v7, v31
-; VI-NEXT: v_add_f16_sdwa v31, v10, v63 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_1 src1_sel:DWORD
-; VI-NEXT: buffer_store_dword v32, off, s[0:3], s32 offset:168 ; 4-byte Folded Spill
-; VI-NEXT: buffer_store_dword v31, off, s[0:3], s32 offset:444 ; 4-byte Folded Spill
; VI-NEXT: v_add_f16_e32 v10, 0x200, v10
-; VI-NEXT: v_add_f16_sdwa v32, v9, v63 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_1 src1_sel:DWORD
-; VI-NEXT: v_lshlrev_b32_e32 v31, 16, v31
-; VI-NEXT: v_add_f16_e32 v9, 0x200, v9
-; VI-NEXT: v_or_b32_e32 v55, v10, v31
-; VI-NEXT: v_lshlrev_b32_e32 v31, 16, v32
-; VI-NEXT: v_add_f16_sdwa v39, v12, v63 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_1 src1_sel:DWORD
-; VI-NEXT: v_or_b32_e32 v54, v9, v31
+; VI-NEXT: v_add_f16_sdwa v35, v9, v55 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_1 src1_sel:DWORD
+; VI-NEXT: v_lshlrev_b32_e32 v31, 16, v49
+; VI-NEXT: v_add_f16_sdwa v39, v12, v55 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_1 src1_sel:DWORD
+; VI-NEXT: buffer_store_dword v32, off, s[0:3], s32 offset:180 ; 4-byte Folded Spill
+; VI-NEXT: buffer_store_dword v35, off, s[0:3], s32 offset:452 ; 4-byte Folded Spill
+; VI-NEXT: v_or_b32_e32 v32, v10, v31
+; VI-NEXT: v_lshlrev_b32_e32 v31, 16, v35
; VI-NEXT: v_add_f16_e32 v12, 0x200, v12
-; VI-NEXT: v_add_f16_sdwa v33, v11, v63 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_1 src1_sel:DWORD
-; VI-NEXT: v_lshlrev_b32_e32 v31, 16, v39
-; VI-NEXT: v_add_f16_sdwa v47, v14, v63 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_1 src1_sel:DWORD
-; VI-NEXT: buffer_store_dword v32, off, s[0:3], s32 offset:164 ; 4-byte Folded Spill
-; VI-NEXT: buffer_store_dword v33, off, s[0:3], s32 offset:448 ; 4-byte Folded Spill
-; VI-NEXT: v_or_b32_e32 v32, v12, v31
-; VI-NEXT: v_lshlrev_b32_e32 v31, 16, v33
+; VI-NEXT: v_add_f16_sdwa v35, v11, v55 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_1 src1_sel:DWORD
+; VI-NEXT: v_lshlrev_b32_e32 v54, 16, v39
+; VI-NEXT: v_add_f16_e32 v11, 0x200, v11
+; VI-NEXT: v_or_b32_e32 v57, v12, v54
+; VI-NEXT: v_lshlrev_b32_e32 v54, 16, v35
+; VI-NEXT: v_add_f16_sdwa v63, v14, v55 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_1 src1_sel:DWORD
+; VI-NEXT: v_or_b32_e32 v56, v11, v54
; VI-NEXT: v_add_f16_e32 v14, 0x200, v14
-; VI-NEXT: v_add_f16_sdwa v33, v13, v63 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_1 src1_sel:DWORD
-; VI-NEXT: v_lshlrev_b32_e32 v56, 16, v47
-; VI-NEXT: v_or_b32_e32 v57, v14, v56
-; VI-NEXT: buffer_store_dword v33, off, s[0:3], s32 offset:160 ; 4-byte Folded Spill
-; VI-NEXT: v_lshlrev_b32_e32 v56, 16, v33
-; VI-NEXT: v_add_f16_sdwa v33, v16, v63 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_1 src1_sel:DWORD
+; VI-NEXT: v_add_f16_sdwa v48, v13, v55 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_1 src1_sel:DWORD
+; VI-NEXT: v_lshlrev_b32_e32 v54, 16, v63
+; VI-NEXT: buffer_store_dword v48, off, s[0:3], s32 offset:184 ; 4-byte Folded Spill
+; VI-NEXT: v_add_f16_e32 v13, 0x200, v13
+; VI-NEXT: v_or_b32_e32 v59, v14, v54
+; VI-NEXT: v_lshlrev_b32_e32 v54, 16, v48
+; VI-NEXT: v_add_f16_sdwa v48, v16, v55 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_1 src1_sel:DWORD
+; VI-NEXT: v_or_b32_e32 v58, v13, v54
; VI-NEXT: v_add_f16_e32 v16, 0x200, v16
-; VI-NEXT: v_add_f16_sdwa v63, v15, v63 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_1 src1_sel:DWORD
-; VI-NEXT: v_lshlrev_b32_e32 v58, 16, v33
+; VI-NEXT: v_add_f16_sdwa v60, v15, v55 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_1 src1_sel:DWORD
+; VI-NEXT: v_lshlrev_b32_e32 v54, 16, v48
; VI-NEXT: v_add_f16_e32 v15, 0x200, v15
-; VI-NEXT: v_or_b32_e32 v59, v16, v58
-; VI-NEXT: v_lshlrev_b32_e32 v58, 16, v63
-; VI-NEXT: v_or_b32_e32 v58, v15, v58
-; VI-NEXT: v_lshrrev_b32_e32 v62, 8, v59
-; VI-NEXT: buffer_store_dword v62, off, s[0:3], s32 offset:184 ; 4-byte Folded Spill
-; VI-NEXT: v_lshrrev_b32_e32 v62, 8, v58
+; VI-NEXT: v_or_b32_e32 v55, v16, v54
+; VI-NEXT: v_lshlrev_b32_e32 v54, 16, v60
+; VI-NEXT: buffer_store_dword v60, off, s[0:3], s32 offset:176 ; 4-byte Folded Spill
+; VI-NEXT: v_or_b32_e32 v54, v15, v54
+; VI-NEXT: v_lshrrev_b32_e32 v60, 8, v55
+; VI-NEXT: buffer_store_dword v60, off, s[0:3], s32 offset:204 ; 4-byte Folded Spill
+; VI-NEXT: v_lshrrev_b32_e32 v60, 8, v54
+; VI-NEXT: v_lshrrev_b64 v[54:55], 24, v[54:55]
+; VI-NEXT: buffer_store_dword v60, off, s[0:3], s32 offset:208 ; 4-byte Folded Spill
+; VI-NEXT: buffer_store_dword v54, off, s[0:3], s32 offset:96 ; 4-byte Folded Spill
+; VI-NEXT: buffer_store_dword v55, off, s[0:3], s32 offset:100 ; 4-byte Folded Spill
+; VI-NEXT: v_lshrrev_b32_e32 v55, 8, v59
+; VI-NEXT: buffer_store_dword v55, off, s[0:3], s32 offset:212 ; 4-byte Folded Spill
+; VI-NEXT: v_lshrrev_b32_e32 v55, 8, v58
; VI-NEXT: v_lshrrev_b64 v[58:59], 24, v[58:59]
-; VI-NEXT: v_add_f16_e32 v13, 0x200, v13
-; VI-NEXT: buffer_store_dword v62, off, s[0:3], s32 offset:188 ; 4-byte Folded Spill
-; VI-NEXT: buffer_store_dword v58, off, s[0:3], s32 offset:96 ; 4-byte Folded Spill
-; VI-NEXT: buffer_store_dword v59, off, s[0:3], s32 offset:100 ; 4-byte Folded Spill
-; VI-NEXT: v_or_b32_e32 v56, v13, v56
-; VI-NEXT: v_lshrrev_b32_e32 v58, 8, v57
-; VI-NEXT: buffer_store_dword v58, off, s[0:3], s32 offset:192 ; 4-byte Folded Spill
-; VI-NEXT: v_lshrrev_b32_e32 v58, 8, v56
+; VI-NEXT: buffer_store_dword v55, off, s[0:3], s32 offset:216 ; 4-byte Folded Spill
+; VI-NEXT: buffer_store_dword v58, off, s[0:3], s32 offset:116 ; 4-byte Folded Spill
+; VI-NEXT: buffer_store_dword v59, off, s[0:3], s32 offset:120 ; 4-byte Folded Spill
+; VI-NEXT: v_lshrrev_b32_e32 v60, 8, v57
+; VI-NEXT: buffer_load_dword v58, off, s[0:3], s32 offset:440 ; 4-byte Folded Reload
+; VI-NEXT: buffer_load_dword v55, off, s[0:3], s32 offset:432 ; 4-byte Folded Reload
+; VI-NEXT: buffer_store_dword v60, off, s[0:3], s32 offset:220 ; 4-byte Folded Spill
+; VI-NEXT: v_lshrrev_b32_e32 v60, 8, v56
; VI-NEXT: v_lshrrev_b64 v[56:57], 24, v[56:57]
-; VI-NEXT: v_add_f16_e32 v11, 0x200, v11
-; VI-NEXT: buffer_store_dword v58, off, s[0:3], s32 offset:196 ; 4-byte Folded Spill
-; VI-NEXT: buffer_store_dword v56, off, s[0:3], s32 offset:112 ; 4-byte Folded Spill
-; VI-NEXT: buffer_store_dword v57, off, s[0:3], s32 offset:116 ; 4-byte Folded Spill
-; VI-NEXT: v_or_b32_e32 v31, v11, v31
+; VI-NEXT: v_add_f16_e32 v9, 0x200, v9
+; VI-NEXT: buffer_store_dword v60, off, s[0:3], s32 offset:224 ; 4-byte Folded Spill
+; VI-NEXT: buffer_store_dword v56, off, s[0:3], s32 offset:132 ; 4-byte Folded Spill
+; VI-NEXT: buffer_store_dword v57, off, s[0:3], s32 offset:136 ; 4-byte Folded Spill
+; VI-NEXT: v_or_b32_e32 v31, v9, v31
; VI-NEXT: v_lshrrev_b32_e32 v56, 8, v32
-; VI-NEXT: buffer_load_dword v59, off, s[0:3], s32 offset:420 ; 4-byte Folded Reload
-; VI-NEXT: buffer_load_dword v58, off, s[0:3], s32 offset:440 ; 4-byte Folded Reload
-; VI-NEXT: buffer_load_dword v57, off, s[0:3], s32 offset:432 ; 4-byte Folded Reload
-; VI-NEXT: buffer_store_dword v56, off, s[0:3], s32 offset:200 ; 4-byte Folded Spill
+; VI-NEXT: buffer_store_dword v56, off, s[0:3], s32 offset:236 ; 4-byte Folded Spill
; VI-NEXT: v_lshrrev_b32_e32 v56, 8, v31
; VI-NEXT: v_lshrrev_b64 v[31:32], 24, v[31:32]
-; VI-NEXT: buffer_store_dword v56, off, s[0:3], s32 offset:204 ; 4-byte Folded Spill
-; VI-NEXT: buffer_store_dword v31, off, s[0:3], s32 offset:120 ; 4-byte Folded Spill
-; VI-NEXT: buffer_store_dword v32, off, s[0:3], s32 offset:124 ; 4-byte Folded Spill
-; VI-NEXT: v_lshrrev_b32_e32 v31, 8, v55
-; VI-NEXT: buffer_store_dword v31, off, s[0:3], s32 offset:216 ; 4-byte Folded Spill
-; VI-NEXT: v_lshrrev_b32_e32 v31, 8, v54
-; VI-NEXT: buffer_store_dword v31, off, s[0:3], s32 offset:224 ; 4-byte Folded Spill
-; VI-NEXT: v_lshrrev_b64 v[31:32], 24, v[54:55]
-; VI-NEXT: buffer_store_dword v31, off, s[0:3], s32 offset:128 ; 4-byte Folded Spill
-; VI-NEXT: buffer_store_dword v32, off, s[0:3], s32 offset:132 ; 4-byte Folded Spill
+; VI-NEXT: buffer_store_dword v56, off, s[0:3], s32 offset:244 ; 4-byte Folded Spill
+; VI-NEXT: buffer_store_dword v31, off, s[0:3], s32 offset:140 ; 4-byte Folded Spill
+; VI-NEXT: buffer_store_dword v32, off, s[0:3], s32 offset:144 ; 4-byte Folded Spill
; VI-NEXT: v_lshrrev_b32_e32 v31, 8, v41
-; VI-NEXT: buffer_store_dword v31, off, s[0:3], s32 offset:240 ; 4-byte Folded Spill
+; VI-NEXT: buffer_store_dword v31, off, s[0:3], s32 offset:260 ; 4-byte Folded Spill
; VI-NEXT: v_lshrrev_b32_e32 v31, 8, v40
-; VI-NEXT: buffer_store_dword v31, off, s[0:3], s32 offset:248 ; 4-byte Folded Spill
+; VI-NEXT: buffer_store_dword v31, off, s[0:3], s32 offset:268 ; 4-byte Folded Spill
; VI-NEXT: v_lshrrev_b64 v[31:32], 24, v[40:41]
-; VI-NEXT: buffer_store_dword v31, off, s[0:3], s32 offset:136 ; 4-byte Folded Spill
-; VI-NEXT: buffer_store_dword v32, off, s[0:3], s32 offset:140 ; 4-byte Folded Spill
+; VI-NEXT: buffer_store_dword v31, off, s[0:3], s32 offset:148 ; 4-byte Folded Spill
+; VI-NEXT: buffer_store_dword v32, off, s[0:3], s32 offset:152 ; 4-byte Folded Spill
; VI-NEXT: v_lshrrev_b32_e32 v31, 8, v43
-; VI-NEXT: buffer_store_dword v31, off, s[0:3], s32 offset:272 ; 4-byte Folded Spill
+; VI-NEXT: buffer_store_dword v31, off, s[0:3], s32 offset:288 ; 4-byte Folded Spill
; VI-NEXT: v_lshrrev_b32_e32 v31, 8, v42
-; VI-NEXT: buffer_store_dword v31, off, s[0:3], s32 offset:280 ; 4-byte Folded Spill
+; VI-NEXT: buffer_store_dword v31, off, s[0:3], s32 offset:296 ; 4-byte Folded Spill
; VI-NEXT: v_lshrrev_b32_e32 v31, 8, v46
-; VI-NEXT: buffer_store_dword v31, off, s[0:3], s32 offset:304 ; 4-byte Folded Spill
+; VI-NEXT: buffer_store_dword v31, off, s[0:3], s32 offset:316 ; 4-byte Folded Spill
; VI-NEXT: v_lshrrev_b32_e32 v31, 8, v45
-; VI-NEXT: buffer_store_dword v31, off, s[0:3], s32 offset:312 ; 4-byte Folded Spill
+; VI-NEXT: buffer_store_dword v31, off, s[0:3], s32 offset:324 ; 4-byte Folded Spill
; VI-NEXT: v_lshrrev_b32_e32 v31, 8, v53
-; VI-NEXT: buffer_store_dword v31, off, s[0:3], s32 offset:336 ; 4-byte Folded Spill
+; VI-NEXT: buffer_store_dword v31, off, s[0:3], s32 offset:348 ; 4-byte Folded Spill
; VI-NEXT: v_lshrrev_b32_e32 v31, 8, v52
-; VI-NEXT: buffer_store_dword v31, off, s[0:3], s32 offset:344 ; 4-byte Folded Spill
-; VI-NEXT: v_lshrrev_b32_e32 v31, 8, v51
-; VI-NEXT: buffer_store_dword v31, off, s[0:3], s32 offset:208 ; 4-byte Folded Spill
-; VI-NEXT: v_lshrrev_b32_e32 v31, 8, v50
-; VI-NEXT: buffer_store_dword v31, off, s[0:3], s32 offset:212 ; 4-byte Folded Spill
-; VI-NEXT: v_lshrrev_b64 v[31:32], 24, v[50:51]
+; VI-NEXT: buffer_store_dword v31, off, s[0:3], s32 offset:356 ; 4-byte Folded Spill
+; VI-NEXT: v_lshrrev_b64 v[31:32], 24, v[52:53]
+; VI-NEXT: v_lshrrev_b32_e32 v32, 8, v51
+; VI-NEXT: buffer_store_dword v32, off, s[0:3], s32 offset:228 ; 4-byte Folded Spill
+; VI-NEXT: v_lshrrev_b32_e32 v32, 8, v50
+; VI-NEXT: v_lshrrev_b64 v[50:51], 24, v[50:51]
; VI-NEXT: v_lshrrev_b64 v[42:43], 24, v[42:43]
-; VI-NEXT: v_lshrrev_b64 v[45:46], 24, v[45:46]
-; VI-NEXT: v_lshrrev_b64 v[52:53], 24, v[52:53]
-; VI-NEXT: buffer_load_dword v43, off, s[0:3], s32 offset:408 ; 4-byte Folded Reload
-; VI-NEXT: buffer_load_dword v46, off, s[0:3], s32 offset:400 ; 4-byte Folded Reload
-; VI-NEXT: buffer_load_dword v53, off, s[0:3], s32 offset:404 ; 4-byte Folded Reload
-; VI-NEXT: buffer_store_dword v31, off, s[0:3], s32 offset:80 ; 4-byte Folded Spill
-; VI-NEXT: buffer_store_dword v32, off, s[0:3], s32 offset:84 ; 4-byte Folded Spill
-; VI-NEXT: v_lshrrev_b32_e32 v31, 8, v49
-; VI-NEXT: buffer_store_dword v31, off, s[0:3], s32 offset:220 ; 4-byte Folded Spill
-; VI-NEXT: v_lshrrev_b32_e32 v31, 8, v48
-; VI-NEXT: buffer_store_dword v31, off, s[0:3], s32 offset:228 ; 4-byte Folded Spill
-; VI-NEXT: v_lshrrev_b64 v[31:32], 24, v[48:49]
-; VI-NEXT: buffer_store_dword v31, off, s[0:3], s32 offset:88 ; 4-byte Folded Spill
-; VI-NEXT: buffer_store_dword v32, off, s[0:3], s32 offset:92 ; 4-byte Folded Spill
-; VI-NEXT: v_lshrrev_b32_e32 v31, 8, v38
-; VI-NEXT: buffer_store_dword v31, off, s[0:3], s32 offset:244 ; 4-byte Folded Spill
-; VI-NEXT: v_lshrrev_b32_e32 v31, 8, v37
-; VI-NEXT: buffer_store_dword v31, off, s[0:3], s32 offset:252 ; 4-byte Folded Spill
-; VI-NEXT: v_lshrrev_b64 v[31:32], 24, v[37:38]
-; VI-NEXT: buffer_store_dword v31, off, s[0:3], s32 offset:104 ; 4-byte Folded Spill
-; VI-NEXT: buffer_store_dword v32, off, s[0:3], s32 offset:108 ; 4-byte Folded Spill
-; VI-NEXT: v_lshrrev_b32_e32 v31, 8, v36
-; VI-NEXT: buffer_store_dword v31, off, s[0:3], s32 offset:276 ; 4-byte Folded Spill
-; VI-NEXT: v_lshrrev_b32_e32 v31, 8, v35
-; VI-NEXT: buffer_store_dword v31, off, s[0:3], s32 offset:292 ; 4-byte Folded Spill
-; VI-NEXT: v_lshrrev_b64 v[35:36], 24, v[35:36]
-; VI-NEXT: buffer_load_dword v38, off, s[0:3], s32 offset:396 ; 4-byte Folded Reload
-; VI-NEXT: buffer_load_dword v36, off, s[0:3], s32 offset:428 ; 4-byte Folded Reload
-; VI-NEXT: buffer_load_dword v48, off, s[0:3], s32 offset:284 ; 4-byte Folded Reload
-; VI-NEXT: buffer_load_dword v49, off, s[0:3], s32 offset:288 ; 4-byte Folded Reload
-; VI-NEXT: v_mov_b32_e32 v54, v39
-; VI-NEXT: v_mov_b32_e32 v37, v44
-; VI-NEXT: s_waitcnt vmcnt(14)
-; VI-NEXT: v_mov_b32_e32 v56, v58
+; VI-NEXT: buffer_store_dword v32, off, s[0:3], s32 offset:232 ; 4-byte Folded Spill
+; VI-NEXT: v_lshrrev_b32_e32 v32, 8, v34
+; VI-NEXT: buffer_load_dword v43, off, s[0:3], s32 offset:448 ; 4-byte Folded Reload
+; VI-NEXT: v_lshrrev_b64 v[56:57], 24, v[45:46]
+; VI-NEXT: buffer_load_dword v46, off, s[0:3], s32 offset:416 ; 4-byte Folded Reload
+; VI-NEXT: buffer_store_dword v50, off, s[0:3], s32 offset:80 ; 4-byte Folded Spill
+; VI-NEXT: buffer_store_dword v51, off, s[0:3], s32 offset:84 ; 4-byte Folded Spill
+; VI-NEXT: buffer_store_dword v32, off, s[0:3], s32 offset:240 ; 4-byte Folded Spill
+; VI-NEXT: v_lshrrev_b32_e32 v32, 8, v33
+; VI-NEXT: buffer_store_dword v32, off, s[0:3], s32 offset:248 ; 4-byte Folded Spill
+; VI-NEXT: v_lshrrev_b64 v[32:33], 24, v[33:34]
+; VI-NEXT: buffer_store_dword v32, off, s[0:3], s32 offset:88 ; 4-byte Folded Spill
+; VI-NEXT: buffer_store_dword v33, off, s[0:3], s32 offset:92 ; 4-byte Folded Spill
+; VI-NEXT: v_lshrrev_b32_e32 v32, 8, v38
+; VI-NEXT: buffer_store_dword v32, off, s[0:3], s32 offset:264 ; 4-byte Folded Spill
+; VI-NEXT: v_lshrrev_b32_e32 v32, 8, v37
+; VI-NEXT: buffer_store_dword v32, off, s[0:3], s32 offset:272 ; 4-byte Folded Spill
+; VI-NEXT: v_lshrrev_b64 v[32:33], 24, v[37:38]
+; VI-NEXT: buffer_store_dword v32, off, s[0:3], s32 offset:104 ; 4-byte Folded Spill
+; VI-NEXT: buffer_store_dword v33, off, s[0:3], s32 offset:108 ; 4-byte Folded Spill
+; VI-NEXT: buffer_load_dword v38, off, s[0:3], s32 offset:420 ; 4-byte Folded Reload
+; VI-NEXT: buffer_load_dword v33, off, s[0:3], s32 offset:124 ; 4-byte Folded Reload
+; VI-NEXT: buffer_load_dword v34, off, s[0:3], s32 offset:128 ; 4-byte Folded Reload
+; VI-NEXT: v_mov_b32_e32 v57, v35
+; VI-NEXT: v_mov_b32_e32 v45, v44
+; VI-NEXT: s_waitcnt vmcnt(0)
+; VI-NEXT: v_lshrrev_b32_e32 v32, 8, v34
+; VI-NEXT: buffer_store_dword v32, off, s[0:3], s32 offset:292 ; 4-byte Folded Spill
+; VI-NEXT: v_lshrrev_b32_e32 v32, 8, v33
+; VI-NEXT: buffer_store_dword v32, off, s[0:3], s32 offset:304 ; 4-byte Folded Spill
+; VI-NEXT: v_lshrrev_b64 v[32:33], 24, v[33:34]
+; VI-NEXT: buffer_store_dword v32, off, s[0:3], s32 offset:124 ; 4-byte Folded Spill
+; VI-NEXT: buffer_store_dword v33, off, s[0:3], s32 offset:128 ; 4-byte Folded Spill
+; VI-NEXT: buffer_load_dword v35, off, s[0:3], s32 offset:424 ; 4-byte Folded Reload
+; VI-NEXT: buffer_load_dword v33, off, s[0:3], s32 offset:276 ; 4-byte Folded Reload
+; VI-NEXT: buffer_load_dword v34, off, s[0:3], s32 offset:280 ; 4-byte Folded Reload
; VI-NEXT: s_waitcnt vmcnt(0)
-; VI-NEXT: v_lshrrev_b32_e32 v31, 8, v49
-; VI-NEXT: buffer_store_dword v31, off, s[0:3], s32 offset:300 ; 4-byte Folded Spill
-; VI-NEXT: v_lshrrev_b32_e32 v31, 8, v48
-; VI-NEXT: buffer_store_dword v31, off, s[0:3], s32 offset:316 ; 4-byte Folded Spill
-; VI-NEXT: v_lshrrev_b64 v[49:50], 24, v[48:49]
-; VI-NEXT: buffer_load_dword v50, off, s[0:3], s32 offset:264 ; 4-byte Folded Reload
-; VI-NEXT: buffer_load_dword v51, off, s[0:3], s32 offset:268 ; 4-byte Folded Reload
-; VI-NEXT: v_mov_b32_e32 v48, v33
+; VI-NEXT: v_lshrrev_b32_e32 v32, 8, v34
+; VI-NEXT: buffer_store_dword v32, off, s[0:3], s32 offset:312 ; 4-byte Folded Spill
+; VI-NEXT: v_lshrrev_b32_e32 v32, 8, v33
+; VI-NEXT: buffer_store_dword v32, off, s[0:3], s32 offset:328 ; 4-byte Folded Spill
+; VI-NEXT: v_lshrrev_b64 v[40:41], 24, v[33:34]
+; VI-NEXT: buffer_load_dword v33, off, s[0:3], s32 offset:252 ; 4-byte Folded Reload
+; VI-NEXT: buffer_load_dword v34, off, s[0:3], s32 offset:256 ; 4-byte Folded Reload
+; VI-NEXT: v_mov_b32_e32 v41, v39
; VI-NEXT: s_waitcnt vmcnt(0)
-; VI-NEXT: v_lshrrev_b32_e32 v31, 8, v51
-; VI-NEXT: buffer_store_dword v31, off, s[0:3], s32 offset:324 ; 4-byte Folded Spill
-; VI-NEXT: v_lshrrev_b32_e32 v31, 8, v50
-; VI-NEXT: buffer_store_dword v31, off, s[0:3], s32 offset:340 ; 4-byte Folded Spill
-; VI-NEXT: v_lshrrev_b64 v[40:41], 24, v[50:51]
-; VI-NEXT: buffer_load_dword v41, off, s[0:3], s32 offset:392 ; 4-byte Folded Reload
-; VI-NEXT: buffer_load_dword v50, off, s[0:3], s32 offset:256 ; 4-byte Folded Reload
-; VI-NEXT: buffer_load_dword v51, off, s[0:3], s32 offset:260 ; 4-byte Folded Reload
+; VI-NEXT: v_lshrrev_b32_e32 v32, 8, v34
+; VI-NEXT: buffer_store_dword v32, off, s[0:3], s32 offset:336 ; 4-byte Folded Spill
+; VI-NEXT: v_lshrrev_b32_e32 v32, 8, v33
+; VI-NEXT: buffer_store_dword v32, off, s[0:3], s32 offset:352 ; 4-byte Folded Spill
+; VI-NEXT: v_lshrrev_b64 v[52:53], 24, v[33:34]
+; VI-NEXT: buffer_load_dword v53, off, s[0:3], s32 offset:412 ; 4-byte Folded Reload
+; VI-NEXT: buffer_load_dword v33, off, s[0:3], s32 offset:196 ; 4-byte Folded Reload
+; VI-NEXT: buffer_load_dword v34, off, s[0:3], s32 offset:200 ; 4-byte Folded Reload
; VI-NEXT: s_waitcnt vmcnt(0)
-; VI-NEXT: v_lshrrev_b32_e32 v31, 8, v51
-; VI-NEXT: buffer_store_dword v31, off, s[0:3], s32 offset:352 ; 4-byte Folded Spill
-; VI-NEXT: v_lshrrev_b32_e32 v31, 8, v50
-; VI-NEXT: buffer_store_dword v31, off, s[0:3], s32 offset:360 ; 4-byte Folded Spill
-; VI-NEXT: v_lshrrev_b64 v[31:32], 24, v[50:51]
-; VI-NEXT: buffer_load_dword v50, off, s[0:3], s32 offset:232 ; 4-byte Folded Reload
-; VI-NEXT: buffer_load_dword v51, off, s[0:3], s32 offset:236 ; 4-byte Folded Reload
-; VI-NEXT: buffer_load_dword v62, off, s[0:3], s32 offset:448 ; 4-byte Folded Reload
+; VI-NEXT: v_lshrrev_b32_e32 v32, 8, v34
+; VI-NEXT: buffer_store_dword v32, off, s[0:3], s32 offset:364 ; 4-byte Folded Spill
+; VI-NEXT: v_lshrrev_b32_e32 v32, 8, v33
+; VI-NEXT: buffer_store_dword v32, off, s[0:3], s32 offset:372 ; 4-byte Folded Spill
+; VI-NEXT: v_lshrrev_b64 v[33:34], 24, v[33:34]
+; VI-NEXT: buffer_load_dword v34, off, s[0:3], s32 offset:408 ; 4-byte Folded Reload
+; VI-NEXT: buffer_load_dword v50, off, s[0:3], s32 offset:168 ; 4-byte Folded Reload
+; VI-NEXT: buffer_load_dword v51, off, s[0:3], s32 offset:172 ; 4-byte Folded Reload
+; VI-NEXT: buffer_store_dword v48, off, s[0:3], s32 offset:168 ; 4-byte Folded Spill
; VI-NEXT: s_waitcnt vmcnt(1)
; VI-NEXT: v_lshrrev_b32_e32 v32, 8, v51
-; VI-NEXT: buffer_store_dword v32, off, s[0:3], s32 offset:372 ; 4-byte Folded Spill
+; VI-NEXT: buffer_store_dword v32, off, s[0:3], s32 offset:384 ; 4-byte Folded Spill
; VI-NEXT: v_lshrrev_b32_e32 v32, 8, v50
-; VI-NEXT: buffer_store_dword v32, off, s[0:3], s32 offset:376 ; 4-byte Folded Spill
-; VI-NEXT: v_bfe_u32 v32, v33, 8, 8
-; VI-NEXT: v_mov_b32_e32 v33, v47
-; VI-NEXT: buffer_store_dword v32, off, s[0:3], s32 offset:232 ; 4-byte Folded Spill
-; VI-NEXT: v_bfe_u32 v32, v33, 8, 8
-; VI-NEXT: buffer_store_dword v32, off, s[0:3], s32 offset:256 ; 4-byte Folded Spill
-; VI-NEXT: v_bfe_u32 v32, v39, 8, 8
-; VI-NEXT: v_mov_b32_e32 v39, v63
-; VI-NEXT: buffer_load_dword v63, off, s[0:3], s32 offset:444 ; 4-byte Folded Reload
-; VI-NEXT: buffer_load_dword v47, off, s[0:3], s32 offset:436 ; 4-byte Folded Reload
-; VI-NEXT: buffer_store_dword v32, off, s[0:3], s32 offset:284 ; 4-byte Folded Spill
+; VI-NEXT: buffer_store_dword v32, off, s[0:3], s32 offset:388 ; 4-byte Folded Spill
+; VI-NEXT: v_bfe_u32 v32, v48, 8, 8
+; VI-NEXT: v_mov_b32_e32 v48, v63
+; VI-NEXT: buffer_store_dword v32, off, s[0:3], s32 offset:252 ; 4-byte Folded Spill
+; VI-NEXT: v_bfe_u32 v32, v48, 8, 8
; VI-NEXT: v_lshrrev_b64 v[50:51], 24, v[50:51]
-; VI-NEXT: buffer_load_dword v51, off, s[0:3], s32 offset:416 ; 4-byte Folded Reload
-; VI-NEXT: s_waitcnt vmcnt(3)
-; VI-NEXT: v_bfe_u32 v32, v63, 8, 8
-; VI-NEXT: buffer_store_dword v32, off, s[0:3], s32 offset:308 ; 4-byte Folded Spill
+; VI-NEXT: buffer_store_dword v32, off, s[0:3], s32 offset:276 ; 4-byte Folded Spill
+; VI-NEXT: v_bfe_u32 v32, v41, 8, 8
+; VI-NEXT: buffer_load_dword v51, off, s[0:3], s32 offset:428 ; 4-byte Folded Reload
+; VI-NEXT: buffer_store_dword v32, off, s[0:3], s32 offset:300 ; 4-byte Folded Spill
+; VI-NEXT: buffer_store_dword v49, off, s[0:3], s32 offset:196 ; 4-byte Folded Spill
+; VI-NEXT: v_bfe_u32 v32, v49, 8, 8
+; VI-NEXT: buffer_store_dword v32, off, s[0:3], s32 offset:320 ; 4-byte Folded Spill
; VI-NEXT: v_bfe_u32 v32, v44, 8, 8
-; VI-NEXT: buffer_store_dword v32, off, s[0:3], s32 offset:328 ; 4-byte Folded Spill
-; VI-NEXT: s_waitcnt vmcnt(4)
-; VI-NEXT: v_bfe_u32 v32, v47, 8, 8
-; VI-NEXT: buffer_store_dword v32, off, s[0:3], s32 offset:356 ; 4-byte Folded Spill
-; VI-NEXT: buffer_load_dword v32, off, s[0:3], s32 offset:424 ; 4-byte Folded Reload
-; VI-NEXT: s_waitcnt vmcnt(0)
-; VI-NEXT: v_mov_b32_e32 v44, v32
-; VI-NEXT: v_bfe_u32 v32, v32, 8, 8
+; VI-NEXT: buffer_store_dword v32, off, s[0:3], s32 offset:340 ; 4-byte Folded Spill
+; VI-NEXT: v_bfe_u32 v32, v36, 8, 8
; VI-NEXT: buffer_store_dword v32, off, s[0:3], s32 offset:368 ; 4-byte Folded Spill
-; VI-NEXT: buffer_load_dword v32, off, s[0:3], s32 offset:412 ; 4-byte Folded Reload
-; VI-NEXT: s_waitcnt vmcnt(0)
-; VI-NEXT: v_mov_b32_e32 v55, v32
+; VI-NEXT: buffer_load_dword v63, off, s[0:3], s32 offset:452 ; 4-byte Folded Reload
+; VI-NEXT: buffer_load_dword v32, off, s[0:3], s32 offset:436 ; 4-byte Folded Reload
+; VI-NEXT: buffer_load_dword v39, off, s[0:3], s32 offset:400 ; 4-byte Folded Reload
+; VI-NEXT: v_mov_b32_e32 v49, v47
+; VI-NEXT: v_mov_b32_e32 v44, v36
+; VI-NEXT: s_waitcnt vmcnt(1)
+; VI-NEXT: v_mov_b32_e32 v37, v32
; VI-NEXT: v_bfe_u32 v32, v32, 8, 8
-; VI-NEXT: buffer_store_dword v32, off, s[0:3], s32 offset:384 ; 4-byte Folded Spill
-; VI-NEXT: v_bfe_u32 v32, v36, 8, 8
-; VI-NEXT: buffer_store_dword v32, off, s[0:3], s32 offset:264 ; 4-byte Folded Spill
+; VI-NEXT: buffer_store_dword v32, off, s[0:3], s32 offset:380 ; 4-byte Folded Spill
+; VI-NEXT: v_bfe_u32 v32, v47, 8, 8
+; VI-NEXT: buffer_store_dword v32, off, s[0:3], s32 offset:396 ; 4-byte Folded Spill
+; VI-NEXT: v_bfe_u32 v32, v51, 8, 8
+; VI-NEXT: buffer_store_dword v32, off, s[0:3], s32 offset:284 ; 4-byte Folded Spill
; VI-NEXT: buffer_load_dword v32, off, s[0:3], s32 offset:76 ; 4-byte Folded Reload
+; VI-NEXT: buffer_load_dword v47, off, s[0:3], s32 offset:444 ; 4-byte Folded Reload
+; VI-NEXT: s_waitcnt vmcnt(1)
+; VI-NEXT: v_bfe_u32 v32, v32, 8, 8
+; VI-NEXT: buffer_store_dword v32, off, s[0:3], s32 offset:308 ; 4-byte Folded Spill
+; VI-NEXT: buffer_load_dword v32, off, s[0:3], s32 offset:112 ; 4-byte Folded Reload
; VI-NEXT: s_waitcnt vmcnt(0)
; VI-NEXT: v_bfe_u32 v32, v32, 8, 8
-; VI-NEXT: buffer_store_dword v32, off, s[0:3], s32 offset:296 ; 4-byte Folded Spill
-; VI-NEXT: v_bfe_u32 v32, v58, 8, 8
-; VI-NEXT: buffer_store_dword v32, off, s[0:3], s32 offset:320 ; 4-byte Folded Spill
-; VI-NEXT: v_bfe_u32 v32, v57, 8, 8
; VI-NEXT: buffer_store_dword v32, off, s[0:3], s32 offset:332 ; 4-byte Folded Spill
-; VI-NEXT: v_bfe_u32 v32, v59, 8, 8
-; VI-NEXT: buffer_store_dword v32, off, s[0:3], s32 offset:348 ; 4-byte Folded Spill
+; VI-NEXT: v_bfe_u32 v32, v58, 8, 8
+; VI-NEXT: buffer_store_dword v32, off, s[0:3], s32 offset:344 ; 4-byte Folded Spill
+; VI-NEXT: v_bfe_u32 v32, v55, 8, 8
+; VI-NEXT: buffer_store_dword v32, off, s[0:3], s32 offset:360 ; 4-byte Folded Spill
; VI-NEXT: v_bfe_u32 v32, v34, 8, 8
-; VI-NEXT: buffer_store_dword v32, off, s[0:3], s32 offset:364 ; 4-byte Folded Spill
-; VI-NEXT: v_bfe_u32 v32, v53, 8, 8
-; VI-NEXT: v_mov_b32_e32 v58, v57
-; VI-NEXT: v_mov_b32_e32 v57, v59
-; VI-NEXT: v_mov_b32_e32 v59, v34
-; VI-NEXT: buffer_store_dword v32, off, s[0:3], s32 offset:380 ; 4-byte Folded Spill
-; VI-NEXT: v_bfe_u32 v32, v41, 8, 8
-; VI-NEXT: buffer_store_dword v32, off, s[0:3], s32 offset:388 ; 4-byte Folded Spill
+; VI-NEXT: buffer_store_dword v32, off, s[0:3], s32 offset:376 ; 4-byte Folded Spill
+; VI-NEXT: v_bfe_u32 v32, v46, 8, 8
+; VI-NEXT: buffer_store_dword v32, off, s[0:3], s32 offset:392 ; 4-byte Folded Spill
+; VI-NEXT: buffer_load_dword v32, off, s[0:3], s32 offset:404 ; 4-byte Folded Reload
+; VI-NEXT: s_waitcnt vmcnt(0)
+; VI-NEXT: v_mov_b32_e32 v54, v32
+; VI-NEXT: v_bfe_u32 v59, v32, 8, 8
; VI-NEXT: .LBB94_4: ; %end
; VI-NEXT: s_or_b64 exec, exec, s[4:5]
-; VI-NEXT: buffer_load_dword v32, off, s[0:3], s32 offset:344 ; 4-byte Folded Reload
+; VI-NEXT: buffer_load_dword v32, off, s[0:3], s32 offset:356 ; 4-byte Folded Reload
+; VI-NEXT: v_lshlrev_b16_e32 v31, 8, v31
; VI-NEXT: s_waitcnt vmcnt(0)
-; VI-NEXT: v_lshlrev_b16_e32 v34, 8, v32
-; VI-NEXT: buffer_load_dword v32, off, s[0:3], s32 offset:336 ; 4-byte Folded Reload
-; VI-NEXT: v_or_b32_sdwa v1, v1, v34 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; VI-NEXT: v_lshlrev_b16_e32 v32, 8, v32
+; VI-NEXT: v_or_b32_sdwa v1, v1, v32 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; VI-NEXT: buffer_load_dword v32, off, s[0:3], s32 offset:348 ; 4-byte Folded Reload
; VI-NEXT: s_waitcnt vmcnt(0)
-; VI-NEXT: v_lshlrev_b16_e32 v34, 8, v32
-; VI-NEXT: buffer_load_dword v32, off, s[0:3], s32 offset:176 ; 4-byte Folded Reload
-; VI-NEXT: v_or_b32_sdwa v2, v2, v34 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
-; VI-NEXT: v_lshlrev_b16_e32 v34, 8, v52
+; VI-NEXT: v_lshlrev_b16_e32 v32, 8, v32
+; VI-NEXT: v_or_b32_sdwa v2, v2, v32 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; VI-NEXT: buffer_load_dword v32, off, s[0:3], s32 offset:188 ; 4-byte Folded Reload
; VI-NEXT: s_waitcnt vmcnt(0)
-; VI-NEXT: v_or_b32_sdwa v34, v32, v34 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
-; VI-NEXT: v_or_b32_sdwa v1, v1, v34 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_0 src1_sel:DWORD
+; VI-NEXT: v_or_b32_sdwa v31, v32, v31 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; VI-NEXT: v_or_b32_sdwa v1, v1, v31 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_0 src1_sel:DWORD
; VI-NEXT: buffer_store_dword v1, v0, s[0:3], 0 offen
-; VI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:384 ; 4-byte Folded Reload
+; VI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:396 ; 4-byte Folded Reload
; VI-NEXT: s_waitcnt vmcnt(0)
; VI-NEXT: v_lshlrev_b16_e32 v1, 8, v1
-; VI-NEXT: v_or_b32_sdwa v1, v55, v1 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; VI-NEXT: v_or_b32_sdwa v1, v49, v1 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
; VI-NEXT: v_or_b32_sdwa v1, v2, v1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_0 src1_sel:DWORD
; VI-NEXT: v_add_u32_e32 v2, vcc, 4, v0
; VI-NEXT: buffer_store_dword v1, v2, s[0:3], 0 offen
-; VI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:180 ; 4-byte Folded Reload
-; VI-NEXT: v_lshlrev_b16_e32 v1, 8, v45
+; VI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:192 ; 4-byte Folded Reload
+; VI-NEXT: v_lshlrev_b16_e32 v1, 8, v56
; VI-NEXT: s_waitcnt vmcnt(0)
; VI-NEXT: v_or_b32_sdwa v1, v2, v1 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
-; VI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:312 ; 4-byte Folded Reload
+; VI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:324 ; 4-byte Folded Reload
; VI-NEXT: s_waitcnt vmcnt(0)
; VI-NEXT: v_lshlrev_b16_e32 v2, 8, v2
; VI-NEXT: v_or_b32_sdwa v2, v3, v2 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
; VI-NEXT: v_or_b32_sdwa v1, v2, v1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_0 src1_sel:DWORD
; VI-NEXT: v_add_u32_e32 v2, vcc, 8, v0
; VI-NEXT: buffer_store_dword v1, v2, s[0:3], 0 offen
-; VI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:304 ; 4-byte Folded Reload
-; VI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:368 ; 4-byte Folded Reload
+; VI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:316 ; 4-byte Folded Reload
+; VI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:380 ; 4-byte Folded Reload
; VI-NEXT: s_waitcnt vmcnt(1)
; VI-NEXT: v_lshlrev_b16_e32 v1, 8, v1
; VI-NEXT: s_waitcnt vmcnt(0)
; VI-NEXT: v_lshlrev_b16_e32 v2, 8, v2
; VI-NEXT: v_or_b32_sdwa v1, v4, v1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
-; VI-NEXT: v_or_b32_sdwa v2, v44, v2 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; VI-NEXT: v_or_b32_sdwa v2, v37, v2 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
; VI-NEXT: v_or_b32_sdwa v1, v1, v2 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_0 src1_sel:DWORD
; VI-NEXT: v_add_u32_e32 v2, vcc, 12, v0
; VI-NEXT: buffer_store_dword v1, v2, s[0:3], 0 offen
-; VI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:172 ; 4-byte Folded Reload
+; VI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:296 ; 4-byte Folded Reload
; VI-NEXT: v_lshlrev_b16_e32 v1, 8, v42
-; VI-NEXT: s_waitcnt vmcnt(0)
-; VI-NEXT: v_or_b32_sdwa v1, v2, v1 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
-; VI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:280 ; 4-byte Folded Reload
+; VI-NEXT: v_or_b32_sdwa v1, v47, v1 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
; VI-NEXT: s_waitcnt vmcnt(0)
; VI-NEXT: v_lshlrev_b16_e32 v2, 8, v2
; VI-NEXT: v_or_b32_sdwa v2, v5, v2 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
; VI-NEXT: v_or_b32_sdwa v1, v2, v1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_0 src1_sel:DWORD
; VI-NEXT: v_add_u32_e32 v2, vcc, 16, v0
; VI-NEXT: buffer_store_dword v1, v2, s[0:3], 0 offen
-; VI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:272 ; 4-byte Folded Reload
-; VI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:356 ; 4-byte Folded Reload
+; VI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:288 ; 4-byte Folded Reload
+; VI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:368 ; 4-byte Folded Reload
; VI-NEXT: s_waitcnt vmcnt(1)
; VI-NEXT: v_lshlrev_b16_e32 v1, 8, v1
; VI-NEXT: s_waitcnt vmcnt(0)
; VI-NEXT: v_lshlrev_b16_e32 v2, 8, v2
; VI-NEXT: v_or_b32_sdwa v1, v6, v1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
-; VI-NEXT: v_or_b32_sdwa v2, v47, v2 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; VI-NEXT: v_or_b32_sdwa v2, v44, v2 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
; VI-NEXT: v_or_b32_sdwa v1, v1, v2 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_0 src1_sel:DWORD
; VI-NEXT: v_add_u32_e32 v2, vcc, 20, v0
; VI-NEXT: buffer_store_dword v1, v2, s[0:3], 0 offen
-; VI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:136 ; 4-byte Folded Reload
-; VI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:140 ; 4-byte Folded Reload
-; VI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:168 ; 4-byte Folded Reload
+; VI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:148 ; 4-byte Folded Reload
+; VI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:152 ; 4-byte Folded Reload
+; VI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:180 ; 4-byte Folded Reload
; VI-NEXT: s_waitcnt vmcnt(2)
; VI-NEXT: v_lshlrev_b16_e32 v1, 8, v1
; VI-NEXT: s_waitcnt vmcnt(0)
; VI-NEXT: v_or_b32_sdwa v1, v2, v1 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
-; VI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:248 ; 4-byte Folded Reload
+; VI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:268 ; 4-byte Folded Reload
; VI-NEXT: s_waitcnt vmcnt(0)
; VI-NEXT: v_lshlrev_b16_e32 v2, 8, v2
; VI-NEXT: v_or_b32_sdwa v2, v7, v2 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
; VI-NEXT: v_or_b32_sdwa v1, v2, v1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_0 src1_sel:DWORD
; VI-NEXT: v_add_u32_e32 v2, vcc, 24, v0
; VI-NEXT: buffer_store_dword v1, v2, s[0:3], 0 offen
-; VI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:240 ; 4-byte Folded Reload
-; VI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:328 ; 4-byte Folded Reload
+; VI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:260 ; 4-byte Folded Reload
+; VI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:340 ; 4-byte Folded Reload
; VI-NEXT: s_waitcnt vmcnt(1)
; VI-NEXT: v_lshlrev_b16_e32 v1, 8, v1
; VI-NEXT: s_waitcnt vmcnt(0)
; VI-NEXT: v_lshlrev_b16_e32 v2, 8, v2
; VI-NEXT: v_or_b32_sdwa v1, v8, v1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
-; VI-NEXT: v_or_b32_sdwa v2, v37, v2 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; VI-NEXT: v_or_b32_sdwa v2, v45, v2 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
; VI-NEXT: v_or_b32_sdwa v1, v1, v2 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_0 src1_sel:DWORD
; VI-NEXT: v_add_u32_e32 v2, vcc, 28, v0
; VI-NEXT: buffer_store_dword v1, v2, s[0:3], 0 offen
-; VI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:128 ; 4-byte Folded Reload
-; VI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:132 ; 4-byte Folded Reload
-; VI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:164 ; 4-byte Folded Reload
+; VI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:140 ; 4-byte Folded Reload
+; VI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:144 ; 4-byte Folded Reload
+; VI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:244 ; 4-byte Folded Reload
; VI-NEXT: s_waitcnt vmcnt(2)
; VI-NEXT: v_lshlrev_b16_e32 v1, 8, v1
-; VI-NEXT: s_waitcnt vmcnt(0)
-; VI-NEXT: v_or_b32_sdwa v1, v2, v1 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
-; VI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:224 ; 4-byte Folded Reload
+; VI-NEXT: v_or_b32_sdwa v1, v63, v1 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
; VI-NEXT: s_waitcnt vmcnt(0)
; VI-NEXT: v_lshlrev_b16_e32 v2, 8, v2
; VI-NEXT: v_or_b32_sdwa v2, v9, v2 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
; VI-NEXT: v_or_b32_sdwa v1, v2, v1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_0 src1_sel:DWORD
; VI-NEXT: v_add_u32_e32 v2, vcc, 32, v0
; VI-NEXT: buffer_store_dword v1, v2, s[0:3], 0 offen
-; VI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:216 ; 4-byte Folded Reload
-; VI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:308 ; 4-byte Folded Reload
-; VI-NEXT: s_waitcnt vmcnt(1)
+; VI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:236 ; 4-byte Folded Reload
+; VI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:320 ; 4-byte Folded Reload
+; VI-NEXT: buffer_load_dword v3, off, s[0:3], s32 offset:196 ; 4-byte Folded Reload
+; VI-NEXT: s_waitcnt vmcnt(2)
; VI-NEXT: v_lshlrev_b16_e32 v1, 8, v1
-; VI-NEXT: s_waitcnt vmcnt(0)
+; VI-NEXT: s_waitcnt vmcnt(1)
; VI-NEXT: v_lshlrev_b16_e32 v2, 8, v2
; VI-NEXT: v_or_b32_sdwa v1, v10, v1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
-; VI-NEXT: v_or_b32_sdwa v2, v63, v2 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; VI-NEXT: s_waitcnt vmcnt(0)
+; VI-NEXT: v_or_b32_sdwa v2, v3, v2 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
; VI-NEXT: v_or_b32_sdwa v1, v1, v2 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_0 src1_sel:DWORD
; VI-NEXT: v_add_u32_e32 v2, vcc, 36, v0
; VI-NEXT: buffer_store_dword v1, v2, s[0:3], 0 offen
-; VI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:120 ; 4-byte Folded Reload
-; VI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:124 ; 4-byte Folded Reload
-; VI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:204 ; 4-byte Folded Reload
+; VI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:132 ; 4-byte Folded Reload
+; VI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:136 ; 4-byte Folded Reload
+; VI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:224 ; 4-byte Folded Reload
; VI-NEXT: s_waitcnt vmcnt(2)
; VI-NEXT: v_lshlrev_b16_e32 v1, 8, v1
-; VI-NEXT: v_or_b32_sdwa v1, v62, v1 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; VI-NEXT: v_or_b32_sdwa v1, v57, v1 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
; VI-NEXT: s_waitcnt vmcnt(0)
; VI-NEXT: v_lshlrev_b16_e32 v2, 8, v2
; VI-NEXT: v_or_b32_sdwa v2, v11, v2 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
; VI-NEXT: v_or_b32_sdwa v1, v2, v1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_0 src1_sel:DWORD
; VI-NEXT: v_add_u32_e32 v2, vcc, 40, v0
; VI-NEXT: buffer_store_dword v1, v2, s[0:3], 0 offen
-; VI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:200 ; 4-byte Folded Reload
-; VI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:284 ; 4-byte Folded Reload
+; VI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:220 ; 4-byte Folded Reload
+; VI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:300 ; 4-byte Folded Reload
; VI-NEXT: s_waitcnt vmcnt(1)
; VI-NEXT: v_lshlrev_b16_e32 v1, 8, v1
; VI-NEXT: s_waitcnt vmcnt(0)
; VI-NEXT: v_lshlrev_b16_e32 v2, 8, v2
; VI-NEXT: v_or_b32_sdwa v1, v12, v1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
-; VI-NEXT: v_or_b32_sdwa v2, v54, v2 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; VI-NEXT: v_or_b32_sdwa v2, v41, v2 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
; VI-NEXT: v_or_b32_sdwa v1, v1, v2 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_0 src1_sel:DWORD
; VI-NEXT: v_add_u32_e32 v2, vcc, 44, v0
; VI-NEXT: buffer_store_dword v1, v2, s[0:3], 0 offen
-; VI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:112 ; 4-byte Folded Reload
-; VI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:116 ; 4-byte Folded Reload
-; VI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:160 ; 4-byte Folded Reload
+; VI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:116 ; 4-byte Folded Reload
+; VI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:120 ; 4-byte Folded Reload
+; VI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:184 ; 4-byte Folded Reload
; VI-NEXT: s_waitcnt vmcnt(2)
; VI-NEXT: v_lshlrev_b16_e32 v1, 8, v1
; VI-NEXT: s_waitcnt vmcnt(0)
; VI-NEXT: v_or_b32_sdwa v1, v2, v1 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
-; VI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:196 ; 4-byte Folded Reload
+; VI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:216 ; 4-byte Folded Reload
; VI-NEXT: s_waitcnt vmcnt(0)
; VI-NEXT: v_lshlrev_b16_e32 v2, 8, v2
; VI-NEXT: v_or_b32_sdwa v2, v13, v2 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
; VI-NEXT: v_or_b32_sdwa v1, v2, v1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_0 src1_sel:DWORD
; VI-NEXT: v_add_u32_e32 v2, vcc, 48, v0
; VI-NEXT: buffer_store_dword v1, v2, s[0:3], 0 offen
-; VI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:192 ; 4-byte Folded Reload
-; VI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:256 ; 4-byte Folded Reload
+; VI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:212 ; 4-byte Folded Reload
+; VI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:276 ; 4-byte Folded Reload
; VI-NEXT: s_waitcnt vmcnt(1)
; VI-NEXT: v_lshlrev_b16_e32 v1, 8, v1
; VI-NEXT: s_waitcnt vmcnt(0)
; VI-NEXT: v_lshlrev_b16_e32 v2, 8, v2
; VI-NEXT: v_or_b32_sdwa v1, v14, v1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
-; VI-NEXT: v_or_b32_sdwa v2, v33, v2 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; VI-NEXT: v_or_b32_sdwa v2, v48, v2 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
; VI-NEXT: v_or_b32_sdwa v1, v1, v2 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_0 src1_sel:DWORD
; VI-NEXT: v_add_u32_e32 v2, vcc, 52, v0
; VI-NEXT: buffer_store_dword v1, v2, s[0:3], 0 offen
; VI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:96 ; 4-byte Folded Reload
; VI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:100 ; 4-byte Folded Reload
-; VI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:188 ; 4-byte Folded Reload
+; VI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:176 ; 4-byte Folded Reload
; VI-NEXT: s_waitcnt vmcnt(2)
; VI-NEXT: v_lshlrev_b16_e32 v1, 8, v1
-; VI-NEXT: v_or_b32_sdwa v1, v39, v1 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; VI-NEXT: s_waitcnt vmcnt(0)
+; VI-NEXT: v_or_b32_sdwa v1, v2, v1 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; VI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:208 ; 4-byte Folded Reload
; VI-NEXT: s_waitcnt vmcnt(0)
; VI-NEXT: v_lshlrev_b16_e32 v2, 8, v2
; VI-NEXT: v_or_b32_sdwa v2, v15, v2 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
; VI-NEXT: v_or_b32_sdwa v1, v2, v1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_0 src1_sel:DWORD
; VI-NEXT: v_add_u32_e32 v2, vcc, 56, v0
; VI-NEXT: buffer_store_dword v1, v2, s[0:3], 0 offen
-; VI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:184 ; 4-byte Folded Reload
-; VI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:232 ; 4-byte Folded Reload
-; VI-NEXT: s_waitcnt vmcnt(1)
+; VI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:204 ; 4-byte Folded Reload
+; VI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:252 ; 4-byte Folded Reload
+; VI-NEXT: buffer_load_dword v3, off, s[0:3], s32 offset:168 ; 4-byte Folded Reload
+; VI-NEXT: s_waitcnt vmcnt(2)
; VI-NEXT: v_lshlrev_b16_e32 v1, 8, v1
-; VI-NEXT: s_waitcnt vmcnt(0)
+; VI-NEXT: s_waitcnt vmcnt(1)
; VI-NEXT: v_lshlrev_b16_e32 v2, 8, v2
; VI-NEXT: v_or_b32_sdwa v1, v16, v1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
-; VI-NEXT: v_or_b32_sdwa v2, v48, v2 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; VI-NEXT: s_waitcnt vmcnt(0)
+; VI-NEXT: v_or_b32_sdwa v2, v3, v2 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
; VI-NEXT: v_or_b32_sdwa v1, v1, v2 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_0 src1_sel:DWORD
; VI-NEXT: v_add_u32_e32 v2, vcc, 60, v0
; VI-NEXT: buffer_store_dword v1, v2, s[0:3], 0 offen
-; VI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:376 ; 4-byte Folded Reload
+; VI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:164 ; 4-byte Folded Reload
; VI-NEXT: v_lshlrev_b16_e32 v1, 8, v50
-; VI-NEXT: v_or_b32_sdwa v1, v46, v1 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; VI-NEXT: s_waitcnt vmcnt(0)
+; VI-NEXT: v_or_b32_sdwa v1, v2, v1 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; VI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:388 ; 4-byte Folded Reload
; VI-NEXT: s_waitcnt vmcnt(0)
; VI-NEXT: v_lshlrev_b16_e32 v2, 8, v2
; VI-NEXT: v_or_b32_sdwa v2, v17, v2 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
; VI-NEXT: v_or_b32_sdwa v1, v2, v1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_0 src1_sel:DWORD
; VI-NEXT: v_add_u32_e32 v2, vcc, 64, v0
; VI-NEXT: buffer_store_dword v1, v2, s[0:3], 0 offen
-; VI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:372 ; 4-byte Folded Reload
-; VI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:388 ; 4-byte Folded Reload
-; VI-NEXT: s_waitcnt vmcnt(1)
-; VI-NEXT: v_lshlrev_b16_e32 v1, 8, v1
+; VI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:384 ; 4-byte Folded Reload
+; VI-NEXT: v_lshlrev_b16_e32 v2, 8, v59
+; VI-NEXT: v_or_b32_sdwa v2, v54, v2 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
; VI-NEXT: s_waitcnt vmcnt(0)
-; VI-NEXT: v_lshlrev_b16_e32 v2, 8, v2
+; VI-NEXT: v_lshlrev_b16_e32 v1, 8, v1
; VI-NEXT: v_or_b32_sdwa v1, v18, v1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
-; VI-NEXT: v_or_b32_sdwa v2, v41, v2 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
; VI-NEXT: v_or_b32_sdwa v1, v1, v2 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_0 src1_sel:DWORD
; VI-NEXT: v_add_u32_e32 v2, vcc, 0x44, v0
; VI-NEXT: buffer_store_dword v1, v2, s[0:3], 0 offen
-; VI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:360 ; 4-byte Folded Reload
-; VI-NEXT: v_lshlrev_b16_e32 v1, 8, v31
-; VI-NEXT: v_or_b32_sdwa v1, v43, v1 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; VI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:372 ; 4-byte Folded Reload
+; VI-NEXT: v_lshlrev_b16_e32 v1, 8, v33
+; VI-NEXT: v_or_b32_sdwa v1, v39, v1 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
; VI-NEXT: s_waitcnt vmcnt(0)
; VI-NEXT: v_lshlrev_b16_e32 v2, 8, v2
; VI-NEXT: v_or_b32_sdwa v2, v19, v2 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
; VI-NEXT: v_or_b32_sdwa v1, v2, v1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_0 src1_sel:DWORD
; VI-NEXT: v_add_u32_e32 v2, vcc, 0x48, v0
; VI-NEXT: buffer_store_dword v1, v2, s[0:3], 0 offen
-; VI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:352 ; 4-byte Folded Reload
-; VI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:380 ; 4-byte Folded Reload
+; VI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:364 ; 4-byte Folded Reload
+; VI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:392 ; 4-byte Folded Reload
; VI-NEXT: s_waitcnt vmcnt(1)
; VI-NEXT: v_lshlrev_b16_e32 v1, 8, v1
; VI-NEXT: s_waitcnt vmcnt(0)
; VI-NEXT: v_lshlrev_b16_e32 v2, 8, v2
; VI-NEXT: v_or_b32_sdwa v1, v20, v1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
-; VI-NEXT: v_or_b32_sdwa v2, v53, v2 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; VI-NEXT: v_or_b32_sdwa v2, v46, v2 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
; VI-NEXT: v_or_b32_sdwa v1, v1, v2 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_0 src1_sel:DWORD
; VI-NEXT: v_add_u32_e32 v2, vcc, 0x4c, v0
; VI-NEXT: buffer_store_dword v1, v2, s[0:3], 0 offen
-; VI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:340 ; 4-byte Folded Reload
-; VI-NEXT: v_lshlrev_b16_e32 v2, 8, v40
-; VI-NEXT: v_or_b32_sdwa v2, v38, v2 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; VI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:352 ; 4-byte Folded Reload
+; VI-NEXT: v_lshlrev_b16_e32 v2, 8, v52
+; VI-NEXT: v_or_b32_sdwa v2, v53, v2 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
; VI-NEXT: s_waitcnt vmcnt(0)
; VI-NEXT: v_lshlrev_b16_e32 v1, 8, v1
; VI-NEXT: v_or_b32_sdwa v1, v21, v1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
; VI-NEXT: v_or_b32_sdwa v1, v1, v2 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_0 src1_sel:DWORD
; VI-NEXT: v_add_u32_e32 v2, vcc, 0x50, v0
; VI-NEXT: buffer_store_dword v1, v2, s[0:3], 0 offen
-; VI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:324 ; 4-byte Folded Reload
-; VI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:364 ; 4-byte Folded Reload
+; VI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:336 ; 4-byte Folded Reload
+; VI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:376 ; 4-byte Folded Reload
; VI-NEXT: s_waitcnt vmcnt(1)
; VI-NEXT: v_lshlrev_b16_e32 v1, 8, v1
; VI-NEXT: s_waitcnt vmcnt(0)
; VI-NEXT: v_lshlrev_b16_e32 v2, 8, v2
; VI-NEXT: v_or_b32_sdwa v1, v22, v1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
-; VI-NEXT: v_or_b32_sdwa v2, v59, v2 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; VI-NEXT: v_or_b32_sdwa v2, v34, v2 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
; VI-NEXT: v_or_b32_sdwa v1, v1, v2 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_0 src1_sel:DWORD
; VI-NEXT: v_add_u32_e32 v2, vcc, 0x54, v0
; VI-NEXT: buffer_store_dword v1, v2, s[0:3], 0 offen
-; VI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:316 ; 4-byte Folded Reload
-; VI-NEXT: buffer_load_dword v3, off, s[0:3], s32 offset:144 ; 4-byte Folded Reload
-; VI-NEXT: v_lshlrev_b16_e32 v2, 8, v49
-; VI-NEXT: s_waitcnt vmcnt(1)
+; VI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:328 ; 4-byte Folded Reload
+; VI-NEXT: v_lshlrev_b16_e32 v2, 8, v40
+; VI-NEXT: v_or_b32_sdwa v2, v38, v2 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; VI-NEXT: s_waitcnt vmcnt(0)
; VI-NEXT: v_lshlrev_b16_e32 v1, 8, v1
; VI-NEXT: v_or_b32_sdwa v1, v23, v1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
-; VI-NEXT: s_waitcnt vmcnt(0)
-; VI-NEXT: v_or_b32_sdwa v2, v3, v2 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
; VI-NEXT: v_or_b32_sdwa v1, v1, v2 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_0 src1_sel:DWORD
; VI-NEXT: v_add_u32_e32 v2, vcc, 0x58, v0
; VI-NEXT: buffer_store_dword v1, v2, s[0:3], 0 offen
-; VI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:300 ; 4-byte Folded Reload
-; VI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:348 ; 4-byte Folded Reload
+; VI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:312 ; 4-byte Folded Reload
+; VI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:360 ; 4-byte Folded Reload
; VI-NEXT: s_waitcnt vmcnt(1)
; VI-NEXT: v_lshlrev_b16_e32 v1, 8, v1
; VI-NEXT: s_waitcnt vmcnt(0)
; VI-NEXT: v_lshlrev_b16_e32 v2, 8, v2
; VI-NEXT: v_or_b32_sdwa v1, v24, v1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
-; VI-NEXT: v_or_b32_sdwa v2, v57, v2 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; VI-NEXT: v_or_b32_sdwa v2, v55, v2 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
; VI-NEXT: v_or_b32_sdwa v1, v1, v2 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_0 src1_sel:DWORD
; VI-NEXT: v_add_u32_e32 v2, vcc, 0x5c, v0
; VI-NEXT: buffer_store_dword v1, v2, s[0:3], 0 offen
-; VI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:292 ; 4-byte Folded Reload
-; VI-NEXT: buffer_load_dword v3, off, s[0:3], s32 offset:156 ; 4-byte Folded Reload
-; VI-NEXT: v_lshlrev_b16_e32 v2, 8, v35
+; VI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:304 ; 4-byte Folded Reload
+; VI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:124 ; 4-byte Folded Reload
+; VI-NEXT: buffer_load_dword v3, off, s[0:3], s32 offset:128 ; 4-byte Folded Reload
; VI-NEXT: s_waitcnt vmcnt(1)
+; VI-NEXT: v_lshlrev_b16_e32 v2, 8, v2
; VI-NEXT: v_lshlrev_b16_e32 v1, 8, v1
; VI-NEXT: v_or_b32_sdwa v1, v25, v1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
-; VI-NEXT: s_waitcnt vmcnt(0)
-; VI-NEXT: v_or_b32_sdwa v2, v3, v2 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; VI-NEXT: v_or_b32_sdwa v2, v35, v2 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
; VI-NEXT: v_or_b32_sdwa v1, v1, v2 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_0 src1_sel:DWORD
; VI-NEXT: v_add_u32_e32 v2, vcc, 0x60, v0
; VI-NEXT: buffer_store_dword v1, v2, s[0:3], 0 offen
-; VI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:276 ; 4-byte Folded Reload
-; VI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:332 ; 4-byte Folded Reload
+; VI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:292 ; 4-byte Folded Reload
+; VI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:344 ; 4-byte Folded Reload
; VI-NEXT: s_waitcnt vmcnt(1)
; VI-NEXT: v_lshlrev_b16_e32 v1, 8, v1
; VI-NEXT: s_waitcnt vmcnt(0)
@@ -193562,10 +193630,10 @@ define <128 x i8> @bitcast_v64f16_to_v128i8(<64 x half> %a, i32 %b) {
; VI-NEXT: v_or_b32_sdwa v1, v1, v2 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_0 src1_sel:DWORD
; VI-NEXT: v_add_u32_e32 v2, vcc, 0x64, v0
; VI-NEXT: buffer_store_dword v1, v2, s[0:3], 0 offen
-; VI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:252 ; 4-byte Folded Reload
+; VI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:272 ; 4-byte Folded Reload
; VI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:104 ; 4-byte Folded Reload
; VI-NEXT: buffer_load_dword v3, off, s[0:3], s32 offset:108 ; 4-byte Folded Reload
-; VI-NEXT: buffer_load_dword v3, off, s[0:3], s32 offset:148 ; 4-byte Folded Reload
+; VI-NEXT: buffer_load_dword v3, off, s[0:3], s32 offset:160 ; 4-byte Folded Reload
; VI-NEXT: s_waitcnt vmcnt(3)
; VI-NEXT: v_lshlrev_b16_e32 v1, 8, v1
; VI-NEXT: s_waitcnt vmcnt(2)
@@ -193576,21 +193644,23 @@ define <128 x i8> @bitcast_v64f16_to_v128i8(<64 x half> %a, i32 %b) {
; VI-NEXT: v_or_b32_sdwa v1, v1, v2 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_0 src1_sel:DWORD
; VI-NEXT: v_add_u32_e32 v2, vcc, 0x68, v0
; VI-NEXT: buffer_store_dword v1, v2, s[0:3], 0 offen
-; VI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:244 ; 4-byte Folded Reload
-; VI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:320 ; 4-byte Folded Reload
-; VI-NEXT: s_waitcnt vmcnt(1)
+; VI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:264 ; 4-byte Folded Reload
+; VI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:332 ; 4-byte Folded Reload
+; VI-NEXT: buffer_load_dword v3, off, s[0:3], s32 offset:112 ; 4-byte Folded Reload
+; VI-NEXT: s_waitcnt vmcnt(2)
; VI-NEXT: v_lshlrev_b16_e32 v1, 8, v1
-; VI-NEXT: s_waitcnt vmcnt(0)
+; VI-NEXT: s_waitcnt vmcnt(1)
; VI-NEXT: v_lshlrev_b16_e32 v2, 8, v2
; VI-NEXT: v_or_b32_sdwa v1, v28, v1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
-; VI-NEXT: v_or_b32_sdwa v2, v56, v2 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; VI-NEXT: s_waitcnt vmcnt(0)
+; VI-NEXT: v_or_b32_sdwa v2, v3, v2 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
; VI-NEXT: v_or_b32_sdwa v1, v1, v2 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_0 src1_sel:DWORD
; VI-NEXT: v_add_u32_e32 v2, vcc, 0x6c, v0
; VI-NEXT: buffer_store_dword v1, v2, s[0:3], 0 offen
-; VI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:228 ; 4-byte Folded Reload
+; VI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:248 ; 4-byte Folded Reload
; VI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:88 ; 4-byte Folded Reload
; VI-NEXT: buffer_load_dword v3, off, s[0:3], s32 offset:92 ; 4-byte Folded Reload
-; VI-NEXT: buffer_load_dword v3, off, s[0:3], s32 offset:152 ; 4-byte Folded Reload
+; VI-NEXT: buffer_load_dword v3, off, s[0:3], s32 offset:156 ; 4-byte Folded Reload
; VI-NEXT: s_waitcnt vmcnt(3)
; VI-NEXT: v_lshlrev_b16_e32 v1, 8, v1
; VI-NEXT: s_waitcnt vmcnt(2)
@@ -193601,8 +193671,8 @@ define <128 x i8> @bitcast_v64f16_to_v128i8(<64 x half> %a, i32 %b) {
; VI-NEXT: v_or_b32_sdwa v1, v1, v2 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_0 src1_sel:DWORD
; VI-NEXT: v_add_u32_e32 v2, vcc, 0x70, v0
; VI-NEXT: buffer_store_dword v1, v2, s[0:3], 0 offen
-; VI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:220 ; 4-byte Folded Reload
-; VI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:296 ; 4-byte Folded Reload
+; VI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:240 ; 4-byte Folded Reload
+; VI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:308 ; 4-byte Folded Reload
; VI-NEXT: buffer_load_dword v3, off, s[0:3], s32 offset:76 ; 4-byte Folded Reload
; VI-NEXT: s_waitcnt vmcnt(2)
; VI-NEXT: v_lshlrev_b16_e32 v1, 8, v1
@@ -193614,27 +193684,27 @@ define <128 x i8> @bitcast_v64f16_to_v128i8(<64 x half> %a, i32 %b) {
; VI-NEXT: v_or_b32_sdwa v1, v1, v2 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_0 src1_sel:DWORD
; VI-NEXT: v_add_u32_e32 v2, vcc, 0x74, v0
; VI-NEXT: buffer_store_dword v1, v2, s[0:3], 0 offen
-; VI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:212 ; 4-byte Folded Reload
+; VI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:232 ; 4-byte Folded Reload
; VI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:80 ; 4-byte Folded Reload
; VI-NEXT: buffer_load_dword v3, off, s[0:3], s32 offset:84 ; 4-byte Folded Reload
; VI-NEXT: s_waitcnt vmcnt(2)
; VI-NEXT: v_lshlrev_b16_e32 v1, 8, v1
; VI-NEXT: s_waitcnt vmcnt(1)
; VI-NEXT: v_lshlrev_b16_e32 v2, 8, v2
-; VI-NEXT: v_or_b32_sdwa v1, v60, v1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
-; VI-NEXT: v_or_b32_sdwa v2, v51, v2 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; VI-NEXT: v_or_b32_sdwa v1, v61, v1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; VI-NEXT: v_or_b32_sdwa v2, v43, v2 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
; VI-NEXT: v_or_b32_sdwa v1, v1, v2 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_0 src1_sel:DWORD
; VI-NEXT: v_add_u32_e32 v2, vcc, 0x78, v0
; VI-NEXT: buffer_store_dword v1, v2, s[0:3], 0 offen
-; VI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:208 ; 4-byte Folded Reload
-; VI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:264 ; 4-byte Folded Reload
+; VI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:228 ; 4-byte Folded Reload
+; VI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:284 ; 4-byte Folded Reload
; VI-NEXT: v_add_u32_e32 v0, vcc, 0x7c, v0
; VI-NEXT: s_waitcnt vmcnt(1)
; VI-NEXT: v_lshlrev_b16_e32 v1, 8, v1
; VI-NEXT: s_waitcnt vmcnt(0)
; VI-NEXT: v_lshlrev_b16_e32 v2, 8, v2
-; VI-NEXT: v_or_b32_sdwa v1, v61, v1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
-; VI-NEXT: v_or_b32_sdwa v2, v36, v2 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; VI-NEXT: v_or_b32_sdwa v1, v62, v1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; VI-NEXT: v_or_b32_sdwa v2, v51, v2 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
; VI-NEXT: v_or_b32_sdwa v1, v1, v2 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_0 src1_sel:DWORD
; VI-NEXT: buffer_store_dword v1, v0, s[0:3], 0 offen
; VI-NEXT: buffer_load_dword v63, off, s[0:3], s32 offset:12 ; 4-byte Folded Reload
@@ -195917,7 +195987,11 @@ define inreg <128 x i8> @bitcast_v64f16_to_v128i8_scalar(<64 x half> inreg %a, i
; SI-NEXT: buffer_store_dword v61, off, s[0:3], s32 offset:92 ; 4-byte Folded Spill
; SI-NEXT: buffer_store_dword v62, off, s[0:3], s32 offset:88 ; 4-byte Folded Spill
; SI-NEXT: buffer_store_dword v63, off, s[0:3], s32 offset:84 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v30, off, s[0:3], s32 offset:152 ; 4-byte Folded Spill
; SI-NEXT: buffer_store_dword v29, off, s[0:3], s32 offset:148 ; 4-byte Folded Spill
+; SI-NEXT: s_waitcnt expcnt(5)
+; SI-NEXT: v_cvt_f16_f32_e32 v60, v1
+; SI-NEXT: v_cvt_f16_f32_e32 v1, v9
; SI-NEXT: buffer_load_dword v34, off, s[0:3], s32 offset:80
; SI-NEXT: buffer_load_dword v35, off, s[0:3], s32 offset:4
; SI-NEXT: buffer_load_dword v36, off, s[0:3], s32
@@ -195939,459 +196013,489 @@ define inreg <128 x i8> @bitcast_v64f16_to_v128i8_scalar(<64 x half> inreg %a, i
; SI-NEXT: buffer_load_dword v46, off, s[0:3], s32 offset:64
; SI-NEXT: buffer_load_dword v47, off, s[0:3], s32 offset:76
; SI-NEXT: buffer_load_dword v57, off, s[0:3], s32 offset:72
-; SI-NEXT: s_waitcnt expcnt(1)
-; SI-NEXT: v_cvt_f16_f32_e32 v63, v5
-; SI-NEXT: buffer_load_dword v5, off, s[0:3], s32 offset:148 ; 4-byte Folded Reload
+; SI-NEXT: v_mov_b32_e32 v33, v28
+; SI-NEXT: v_mov_b32_e32 v28, v27
+; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:440 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_cvt_f16_f32_e32 v29, v2
-; SI-NEXT: v_cvt_f16_f32_e32 v60, v1
-; SI-NEXT: v_cvt_f16_f32_e32 v1, v9
-; SI-NEXT: v_cvt_f16_f32_e32 v2, v11
-; SI-NEXT: v_mov_b32_e32 v59, v18
+; SI-NEXT: v_cvt_f16_f32_e32 v1, v26
+; SI-NEXT: v_cvt_f16_f32_e32 v26, v28
+; SI-NEXT: v_cvt_f16_f32_e32 v32, v2
; SI-NEXT: v_cvt_f16_f32_e32 v43, v4
-; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:444 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:472 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_cvt_f16_f32_e32 v1, v13
+; SI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:152 ; 4-byte Folded Reload
; SI-NEXT: v_cvt_f16_f32_e32 v42, v3
-; SI-NEXT: v_cvt_f16_f32_e32 v33, v6
+; SI-NEXT: v_cvt_f16_f32_e32 v6, v6
+; SI-NEXT: v_cvt_f16_f32_e32 v63, v5
; SI-NEXT: v_cvt_f16_f32_e32 v61, v8
-; SI-NEXT: v_cvt_f16_f32_e32 v58, v7
-; SI-NEXT: v_cvt_f16_f32_e32 v32, v10
-; SI-NEXT: v_cvt_f16_f32_e32 v3, v12
-; SI-NEXT: v_cvt_f16_f32_e32 v31, v14
-; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:448 ; 4-byte Folded Spill
-; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_cvt_f16_f32_e32 v1, v16
-; SI-NEXT: v_cvt_f16_f32_e32 v4, v15
-; SI-NEXT: v_cvt_f16_f32_e32 v10, v59
-; SI-NEXT: v_cvt_f16_f32_e32 v6, v20
-; SI-NEXT: v_cvt_f16_f32_e32 v59, v19
-; SI-NEXT: v_cvt_f16_f32_e32 v9, v22
-; SI-NEXT: v_cvt_f16_f32_e32 v56, v21
-; SI-NEXT: v_cvt_f16_f32_e32 v18, v24
-; SI-NEXT: v_cvt_f16_f32_e32 v62, v23
-; SI-NEXT: v_cvt_f16_f32_e32 v7, v26
+; SI-NEXT: v_cvt_f16_f32_e32 v3, v7
+; SI-NEXT: v_cvt_f16_f32_e32 v31, v10
+; SI-NEXT: v_cvt_f16_f32_e32 v7, v12
+; SI-NEXT: v_cvt_f16_f32_e32 v59, v11
+; SI-NEXT: v_cvt_f16_f32_e32 v10, v14
+; SI-NEXT: v_cvt_f16_f32_e32 v5, v13
+; SI-NEXT: v_cvt_f16_f32_e32 v58, v16
+; SI-NEXT: v_cvt_f16_f32_e32 v15, v15
+; SI-NEXT: v_cvt_f16_f32_e32 v9, v18
+; SI-NEXT: v_cvt_f16_f32_e32 v29, v20
+; SI-NEXT: v_cvt_f16_f32_e32 v56, v19
+; SI-NEXT: v_cvt_f16_f32_e32 v8, v22
+; SI-NEXT: v_cvt_f16_f32_e32 v62, v21
+; SI-NEXT: v_cvt_f16_f32_e32 v27, v24
+; SI-NEXT: v_cvt_f16_f32_e32 v30, v23
; SI-NEXT: v_cvt_f16_f32_e32 v25, v25
-; SI-NEXT: v_cvt_f16_f32_e32 v26, v28
-; SI-NEXT: v_cvt_f16_f32_e32 v27, v27
-; SI-NEXT: v_cvt_f16_f32_e32 v8, v30
+; SI-NEXT: v_cvt_f16_f32_e32 v33, v33
+; SI-NEXT: v_cvt_f16_f32_e32 v11, s17
; SI-NEXT: v_cvt_f16_f32_e32 v13, s16
+; SI-NEXT: v_cvt_f16_f32_e32 v20, s19
; SI-NEXT: v_cvt_f16_f32_e32 v12, s18
; SI-NEXT: v_cvt_f16_f32_e32 v16, s20
-; SI-NEXT: v_cvt_f16_f32_e32 v20, s23
+; SI-NEXT: v_cvt_f16_f32_e32 v18, s23
; SI-NEXT: v_cvt_f16_f32_e32 v14, s22
; SI-NEXT: v_cvt_f16_f32_e32 v22, s25
; SI-NEXT: v_cvt_f16_f32_e32 v21, s24
-; SI-NEXT: v_cvt_f16_f32_e32 v15, s27
; SI-NEXT: v_cvt_f16_f32_e32 v19, s26
; SI-NEXT: v_cvt_f16_f32_e32 v24, s29
; SI-NEXT: v_cvt_f16_f32_e32 v23, s28
; SI-NEXT: s_waitcnt vmcnt(14)
; SI-NEXT: v_cmp_ne_u32_e32 vcc, 0, v34
; SI-NEXT: v_cvt_f16_f32_e32 v34, v17
-; SI-NEXT: v_cvt_f16_f32_e32 v30, v35
+; SI-NEXT: v_cvt_f16_f32_e32 v35, v35
; SI-NEXT: v_cvt_f16_f32_e32 v36, v36
-; SI-NEXT: v_cvt_f16_f32_e32 v35, v38
-; SI-NEXT: v_cvt_f16_f32_e32 v38, v39
-; SI-NEXT: v_cvt_f16_f32_e32 v39, v48
-; SI-NEXT: v_cvt_f16_f32_e32 v11, v49
+; SI-NEXT: v_cvt_f16_f32_e32 v38, v38
+; SI-NEXT: v_cvt_f16_f32_e32 v39, v39
+; SI-NEXT: v_cvt_f16_f32_e32 v17, s21
+; SI-NEXT: v_cvt_f16_f32_e32 v4, v49
; SI-NEXT: v_cvt_f16_f32_e32 v50, v50
; SI-NEXT: v_cvt_f16_f32_e32 v49, v51
; SI-NEXT: s_waitcnt vmcnt(13)
-; SI-NEXT: v_cvt_f16_f32_e32 v48, v52
-; SI-NEXT: buffer_store_dword v11, off, s[0:3], s32 offset:148 ; 4-byte Folded Spill
-; SI-NEXT: s_waitcnt vmcnt(13) expcnt(0)
-; SI-NEXT: v_cvt_f16_f32_e32 v11, v53
+; SI-NEXT: v_cvt_f16_f32_e32 v52, v52
; SI-NEXT: s_waitcnt vmcnt(12)
-; SI-NEXT: v_cvt_f16_f32_e32 v54, v54
+; SI-NEXT: v_cvt_f16_f32_e32 v2, v53
; SI-NEXT: s_waitcnt vmcnt(11)
-; SI-NEXT: v_cvt_f16_f32_e32 v51, v55
+; SI-NEXT: v_cvt_f16_f32_e32 v51, v54
; SI-NEXT: s_waitcnt vmcnt(10)
-; SI-NEXT: v_cvt_f16_f32_e32 v53, v40
-; SI-NEXT: buffer_store_dword v11, off, s[0:3], s32 offset:492 ; 4-byte Folded Spill
+; SI-NEXT: v_cvt_f16_f32_e32 v53, v55
; SI-NEXT: s_waitcnt vmcnt(9)
-; SI-NEXT: v_cvt_f16_f32_e32 v40, v44
-; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_cvt_f16_f32_e32 v11, s17
+; SI-NEXT: v_cvt_f16_f32_e32 v54, v40
+; SI-NEXT: s_and_b64 s[4:5], vcc, exec
+; SI-NEXT: s_waitcnt vmcnt(7)
+; SI-NEXT: v_cvt_f16_f32_e32 v44, v44
; SI-NEXT: s_waitcnt vmcnt(6)
-; SI-NEXT: v_cvt_f16_f32_e32 v52, v47
-; SI-NEXT: v_cvt_f16_f32_e32 v44, s19
-; SI-NEXT: v_cvt_f16_f32_e32 v17, s21
-; SI-NEXT: s_waitcnt vmcnt(4)
-; SI-NEXT: v_cvt_f16_f32_e32 v28, v5
-; SI-NEXT: v_cvt_f16_f32_e32 v5, v37
-; SI-NEXT: v_cvt_f16_f32_e32 v37, v41
-; SI-NEXT: v_cvt_f16_f32_e32 v41, v45
+; SI-NEXT: v_cvt_f16_f32_e32 v40, v45
+; SI-NEXT: s_waitcnt vmcnt(5)
; SI-NEXT: v_cvt_f16_f32_e32 v45, v46
-; SI-NEXT: v_cvt_f16_f32_e32 v46, v57
-; SI-NEXT: s_and_b64 s[4:5], vcc, exec
-; SI-NEXT: buffer_store_dword v60, off, s[0:3], s32 offset:504 ; 4-byte Folded Spill
-; SI-NEXT: buffer_store_dword v12, off, s[0:3], s32 offset:508 ; 4-byte Folded Spill
-; SI-NEXT: buffer_store_dword v63, off, s[0:3], s32 offset:512 ; 4-byte Folded Spill
-; SI-NEXT: buffer_store_dword v13, off, s[0:3], s32 offset:516 ; 4-byte Folded Spill
-; SI-NEXT: buffer_store_dword v11, off, s[0:3], s32 offset:520 ; 4-byte Folded Spill
-; SI-NEXT: buffer_store_dword v42, off, s[0:3], s32 offset:524 ; 4-byte Folded Spill
-; SI-NEXT: buffer_store_dword v14, off, s[0:3], s32 offset:528 ; 4-byte Folded Spill
-; SI-NEXT: buffer_store_dword v16, off, s[0:3], s32 offset:532 ; 4-byte Folded Spill
-; SI-NEXT: buffer_store_dword v17, off, s[0:3], s32 offset:536 ; 4-byte Folded Spill
-; SI-NEXT: buffer_store_dword v19, off, s[0:3], s32 offset:540 ; 4-byte Folded Spill
-; SI-NEXT: buffer_store_dword v58, off, s[0:3], s32 offset:544 ; 4-byte Folded Spill
-; SI-NEXT: buffer_store_dword v21, off, s[0:3], s32 offset:548 ; 4-byte Folded Spill
-; SI-NEXT: buffer_store_dword v22, off, s[0:3], s32 offset:552 ; 4-byte Folded Spill
-; SI-NEXT: buffer_store_dword v23, off, s[0:3], s32 offset:556 ; 4-byte Folded Spill
-; SI-NEXT: buffer_store_dword v24, off, s[0:3], s32 offset:560 ; 4-byte Folded Spill
-; SI-NEXT: buffer_store_dword v34, off, s[0:3], s32 offset:564 ; 4-byte Folded Spill
-; SI-NEXT: buffer_store_dword v4, off, s[0:3], s32 offset:452 ; 4-byte Folded Spill
-; SI-NEXT: buffer_store_dword v15, off, s[0:3], s32 offset:456 ; 4-byte Folded Spill
-; SI-NEXT: buffer_store_dword v20, off, s[0:3], s32 offset:460 ; 4-byte Folded Spill
-; SI-NEXT: buffer_store_dword v44, off, s[0:3], s32 offset:464 ; 4-byte Folded Spill
+; SI-NEXT: s_waitcnt vmcnt(3)
+; SI-NEXT: v_cvt_f16_f32_e32 v57, v57
+; SI-NEXT: s_waitcnt vmcnt(0)
+; SI-NEXT: v_cvt_f16_f32_e32 v1, v1
+; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:476 ; 4-byte Folded Spill
+; SI-NEXT: s_waitcnt expcnt(0)
+; SI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:148 ; 4-byte Folded Reload
+; SI-NEXT: s_waitcnt vmcnt(0)
+; SI-NEXT: v_cvt_f16_f32_e32 v28, v1
+; SI-NEXT: v_cvt_f16_f32_e32 v1, v37
+; SI-NEXT: v_cvt_f16_f32_e32 v37, v48
+; SI-NEXT: v_cvt_f16_f32_e32 v48, s27
; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:468 ; 4-byte Folded Spill
-; SI-NEXT: buffer_store_dword v56, off, s[0:3], s32 offset:472 ; 4-byte Folded Spill
-; SI-NEXT: buffer_store_dword v59, off, s[0:3], s32 offset:568 ; 4-byte Folded Spill
-; SI-NEXT: buffer_store_dword v6, off, s[0:3], s32 offset:572 ; 4-byte Folded Spill
-; SI-NEXT: buffer_store_dword v25, off, s[0:3], s32 offset:576 ; 4-byte Folded Spill
-; SI-NEXT: buffer_store_dword v62, off, s[0:3], s32 offset:580 ; 4-byte Folded Spill
-; SI-NEXT: buffer_store_dword v18, off, s[0:3], s32 offset:584 ; 4-byte Folded Spill
-; SI-NEXT: buffer_store_dword v28, off, s[0:3], s32 offset:588 ; 4-byte Folded Spill
-; SI-NEXT: buffer_store_dword v27, off, s[0:3], s32 offset:592 ; 4-byte Folded Spill
-; SI-NEXT: buffer_store_dword v26, off, s[0:3], s32 offset:596 ; 4-byte Folded Spill
+; SI-NEXT: s_waitcnt expcnt(0)
+; SI-NEXT: v_cvt_f16_f32_e32 v1, v41
+; SI-NEXT: v_cvt_f16_f32_e32 v41, v47
+; SI-NEXT: buffer_store_dword v60, off, s[0:3], s32 offset:488 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v12, off, s[0:3], s32 offset:492 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v63, off, s[0:3], s32 offset:496 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v13, off, s[0:3], s32 offset:500 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v11, off, s[0:3], s32 offset:504 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v42, off, s[0:3], s32 offset:508 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v14, off, s[0:3], s32 offset:512 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v16, off, s[0:3], s32 offset:516 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v17, off, s[0:3], s32 offset:520 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v19, off, s[0:3], s32 offset:524 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v21, off, s[0:3], s32 offset:528 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v22, off, s[0:3], s32 offset:532 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v61, off, s[0:3], s32 offset:536 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v3, off, s[0:3], s32 offset:444 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v23, off, s[0:3], s32 offset:540 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v59, off, s[0:3], s32 offset:448 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v24, off, s[0:3], s32 offset:544 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v7, off, s[0:3], s32 offset:452 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v34, off, s[0:3], s32 offset:548 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v15, off, s[0:3], s32 offset:456 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v58, off, s[0:3], s32 offset:460 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v62, off, s[0:3], s32 offset:552 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v5, off, s[0:3], s32 offset:464 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v56, off, s[0:3], s32 offset:556 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v29, off, s[0:3], s32 offset:560 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v25, off, s[0:3], s32 offset:564 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v30, off, s[0:3], s32 offset:568 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v27, off, s[0:3], s32 offset:572 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v43, off, s[0:3], s32 offset:576 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v28, off, s[0:3], s32 offset:580 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v26, off, s[0:3], s32 offset:584 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v33, off, s[0:3], s32 offset:588 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v38, off, s[0:3], s32 offset:592 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v36, off, s[0:3], s32 offset:596 ; 4-byte Folded Spill
; SI-NEXT: buffer_store_dword v35, off, s[0:3], s32 offset:600 ; 4-byte Folded Spill
-; SI-NEXT: buffer_store_dword v36, off, s[0:3], s32 offset:604 ; 4-byte Folded Spill
-; SI-NEXT: buffer_store_dword v30, off, s[0:3], s32 offset:608 ; 4-byte Folded Spill
-; SI-NEXT: buffer_store_dword v50, off, s[0:3], s32 offset:612 ; 4-byte Folded Spill
-; SI-NEXT: buffer_store_dword v39, off, s[0:3], s32 offset:616 ; 4-byte Folded Spill
-; SI-NEXT: buffer_store_dword v38, off, s[0:3], s32 offset:620 ; 4-byte Folded Spill
-; SI-NEXT: buffer_store_dword v54, off, s[0:3], s32 offset:624 ; 4-byte Folded Spill
-; SI-NEXT: buffer_store_dword v48, off, s[0:3], s32 offset:628 ; 4-byte Folded Spill
-; SI-NEXT: buffer_store_dword v49, off, s[0:3], s32 offset:632 ; 4-byte Folded Spill
-; SI-NEXT: buffer_store_dword v40, off, s[0:3], s32 offset:636 ; 4-byte Folded Spill
-; SI-NEXT: buffer_store_dword v53, off, s[0:3], s32 offset:640 ; 4-byte Folded Spill
-; SI-NEXT: buffer_store_dword v3, off, s[0:3], s32 offset:476 ; 4-byte Folded Spill
-; SI-NEXT: buffer_store_dword v51, off, s[0:3], s32 offset:644 ; 4-byte Folded Spill
-; SI-NEXT: buffer_store_dword v45, off, s[0:3], s32 offset:648 ; 4-byte Folded Spill
-; SI-NEXT: buffer_store_dword v43, off, s[0:3], s32 offset:480 ; 4-byte Folded Spill
-; SI-NEXT: buffer_store_dword v41, off, s[0:3], s32 offset:652 ; 4-byte Folded Spill
-; SI-NEXT: buffer_store_dword v61, off, s[0:3], s32 offset:484 ; 4-byte Folded Spill
-; SI-NEXT: buffer_store_dword v46, off, s[0:3], s32 offset:656 ; 4-byte Folded Spill
-; SI-NEXT: buffer_store_dword v2, off, s[0:3], s32 offset:488 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v50, off, s[0:3], s32 offset:604 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v37, off, s[0:3], s32 offset:608 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v39, off, s[0:3], s32 offset:612 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v51, off, s[0:3], s32 offset:616 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v52, off, s[0:3], s32 offset:620 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v49, off, s[0:3], s32 offset:624 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v44, off, s[0:3], s32 offset:628 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v54, off, s[0:3], s32 offset:632 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v53, off, s[0:3], s32 offset:636 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v45, off, s[0:3], s32 offset:640 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v40, off, s[0:3], s32 offset:644 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v57, off, s[0:3], s32 offset:648 ; 4-byte Folded Spill
; SI-NEXT: s_cbranch_scc0 .LBB95_2
; SI-NEXT: ; %bb.1: ; %cmp.false
; SI-NEXT: v_lshlrev_b32_e32 v11, 16, v11
-; SI-NEXT: v_or_b32_e32 v13, v13, v11
-; SI-NEXT: v_lshlrev_b32_e32 v11, 16, v44
+; SI-NEXT: v_or_b32_e32 v47, v13, v11
+; SI-NEXT: v_lshlrev_b32_e32 v11, 16, v20
; SI-NEXT: v_or_b32_e32 v55, v12, v11
-; SI-NEXT: buffer_load_dword v12, off, s[0:3], s32 offset:444 ; 4-byte Folded Reload
; SI-NEXT: v_lshlrev_b32_e32 v11, 16, v17
-; SI-NEXT: v_or_b32_e32 v57, v16, v11
-; SI-NEXT: v_lshlrev_b32_e32 v11, 16, v20
-; SI-NEXT: v_or_b32_e32 v17, v14, v11
+; SI-NEXT: v_or_b32_e32 v13, v16, v11
+; SI-NEXT: v_lshlrev_b32_e32 v11, 16, v18
+; SI-NEXT: v_or_b32_e32 v46, v14, v11
; SI-NEXT: v_lshlrev_b32_e32 v11, 16, v22
-; SI-NEXT: v_or_b32_e32 v21, v21, v11
-; SI-NEXT: v_lshlrev_b32_e32 v11, 16, v15
+; SI-NEXT: v_or_b32_e32 v17, v21, v11
+; SI-NEXT: v_lshlrev_b32_e32 v11, 16, v48
; SI-NEXT: v_or_b32_e32 v16, v19, v11
; SI-NEXT: v_lshlrev_b32_e32 v11, 16, v24
-; SI-NEXT: v_or_b32_e32 v19, v23, v11
-; SI-NEXT: v_lshlrev_b32_e32 v11, 16, v29
-; SI-NEXT: v_or_b32_e32 v47, v60, v11
+; SI-NEXT: v_mov_b32_e32 v24, v32
+; SI-NEXT: v_or_b32_e32 v22, v23, v11
+; SI-NEXT: v_lshlrev_b32_e32 v11, 16, v24
+; SI-NEXT: v_or_b32_e32 v14, v60, v11
; SI-NEXT: v_lshlrev_b32_e32 v11, 16, v43
-; SI-NEXT: s_waitcnt expcnt(4)
-; SI-NEXT: v_or_b32_e32 v43, v42, v11
-; SI-NEXT: v_lshlrev_b32_e32 v11, 16, v33
-; SI-NEXT: v_or_b32_e32 v14, v63, v11
+; SI-NEXT: v_mov_b32_e32 v21, v6
+; SI-NEXT: v_or_b32_e32 v23, v42, v11
+; SI-NEXT: v_lshlrev_b32_e32 v11, 16, v21
+; SI-NEXT: v_or_b32_e32 v60, v63, v11
; SI-NEXT: v_lshlrev_b32_e32 v11, 16, v61
-; SI-NEXT: v_or_b32_e32 v42, v58, v11
+; SI-NEXT: v_or_b32_e32 v43, v3, v11
+; SI-NEXT: buffer_load_dword v3, off, s[0:3], s32 offset:440 ; 4-byte Folded Reload
+; SI-NEXT: v_mov_b32_e32 v32, v31
; SI-NEXT: v_lshlrev_b32_e32 v11, 16, v32
+; SI-NEXT: v_mov_b32_e32 v31, v10
+; SI-NEXT: v_mov_b32_e32 v10, v9
+; SI-NEXT: v_mov_b32_e32 v9, v8
+; SI-NEXT: buffer_load_dword v8, off, s[0:3], s32 offset:472 ; 4-byte Folded Reload
; SI-NEXT: s_mov_b64 s[4:5], 0
-; SI-NEXT: s_waitcnt vmcnt(0)
-; SI-NEXT: v_or_b32_e32 v60, v12, v11
-; SI-NEXT: v_lshlrev_b32_e32 v11, 16, v3
-; SI-NEXT: v_or_b32_e32 v22, v2, v11
-; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:448 ; 4-byte Folded Reload
+; SI-NEXT: s_waitcnt vmcnt(1)
+; SI-NEXT: v_or_b32_e32 v6, v3, v11
+; SI-NEXT: v_lshlrev_b32_e32 v11, 16, v7
+; SI-NEXT: v_or_b32_e32 v12, v59, v11
; SI-NEXT: v_lshlrev_b32_e32 v11, 16, v31
-; SI-NEXT: v_lshlrev_b32_e32 v12, 16, v52
-; SI-NEXT: v_or_b32_e32 v12, v46, v12
-; SI-NEXT: s_waitcnt vmcnt(0)
-; SI-NEXT: v_or_b32_e32 v24, v2, v11
-; SI-NEXT: v_lshlrev_b32_e32 v11, 16, v1
-; SI-NEXT: v_or_b32_e32 v4, v4, v11
+; SI-NEXT: v_or_b32_e32 v63, v5, v11
+; SI-NEXT: v_lshlrev_b32_e32 v11, 16, v58
+; SI-NEXT: v_or_b32_e32 v42, v15, v11
; SI-NEXT: v_lshlrev_b32_e32 v11, 16, v10
; SI-NEXT: v_or_b32_e32 v34, v34, v11
-; SI-NEXT: v_lshlrev_b32_e32 v11, 16, v6
-; SI-NEXT: v_or_b32_e32 v3, v59, v11
+; SI-NEXT: v_lshlrev_b32_e32 v11, 16, v29
+; SI-NEXT: v_or_b32_e32 v5, v56, v11
; SI-NEXT: v_lshlrev_b32_e32 v11, 16, v9
-; SI-NEXT: v_or_b32_e32 v59, v56, v11
-; SI-NEXT: v_lshlrev_b32_e32 v11, 16, v18
-; SI-NEXT: v_or_b32_e32 v6, v62, v11
-; SI-NEXT: v_lshlrev_b32_e32 v11, 16, v7
-; SI-NEXT: v_or_b32_e32 v62, v25, v11
-; SI-NEXT: v_lshlrev_b32_e32 v11, 16, v26
-; SI-NEXT: v_or_b32_e32 v2, v27, v11
+; SI-NEXT: v_or_b32_e32 v56, v62, v11
+; SI-NEXT: v_lshlrev_b32_e32 v11, 16, v27
+; SI-NEXT: v_or_b32_e32 v3, v30, v11
+; SI-NEXT: buffer_load_dword v30, off, s[0:3], s32 offset:476 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v62, off, s[0:3], s32 offset:468 ; 4-byte Folded Reload
+; SI-NEXT: s_waitcnt vmcnt(2)
; SI-NEXT: v_lshlrev_b32_e32 v11, 16, v8
-; SI-NEXT: v_or_b32_e32 v25, v28, v11
+; SI-NEXT: v_or_b32_e32 v58, v25, v11
+; SI-NEXT: v_lshlrev_b32_e32 v11, 16, v33
+; SI-NEXT: v_or_b32_e32 v26, v26, v11
+; SI-NEXT: buffer_store_dword v26, off, s[0:3], s32 offset:264 ; 4-byte Folded Spill
+; SI-NEXT: v_mov_b32_e32 v33, v32
+; SI-NEXT: s_waitcnt vmcnt(2)
; SI-NEXT: v_lshlrev_b32_e32 v11, 16, v30
-; SI-NEXT: v_or_b32_e32 v1, v36, v11
-; SI-NEXT: v_lshlrev_b32_e32 v11, 16, v5
-; SI-NEXT: v_or_b32_e32 v23, v35, v11
-; SI-NEXT: v_lshlrev_b32_e32 v11, 16, v38
-; SI-NEXT: buffer_load_dword v38, off, s[0:3], s32 offset:148 ; 4-byte Folded Reload
-; SI-NEXT: v_or_b32_e32 v18, v39, v11
-; SI-NEXT: v_mov_b32_e32 v36, v2
-; SI-NEXT: v_mov_b32_e32 v35, v1
-; SI-NEXT: v_alignbit_b32 v1, v55, v13, 24
-; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:500 ; 4-byte Folded Spill
+; SI-NEXT: v_or_b32_e32 v25, v28, v11
+; SI-NEXT: v_lshlrev_b32_e32 v11, 16, v35
+; SI-NEXT: v_or_b32_e32 v29, v36, v11
+; SI-NEXT: s_waitcnt vmcnt(1)
+; SI-NEXT: v_lshlrev_b32_e32 v11, 16, v62
+; SI-NEXT: v_or_b32_e32 v27, v38, v11
+; SI-NEXT: v_lshlrev_b32_e32 v11, 16, v39
+; SI-NEXT: v_or_b32_e32 v19, v37, v11
+; SI-NEXT: v_lshlrev_b32_e32 v11, 16, v4
+; SI-NEXT: v_or_b32_e32 v7, v50, v11
+; SI-NEXT: v_lshlrev_b32_e32 v11, 16, v49
+; SI-NEXT: v_mov_b32_e32 v36, v5
+; SI-NEXT: v_or_b32_e32 v5, v52, v11
+; SI-NEXT: v_lshlrev_b32_e32 v11, 16, v2
+; SI-NEXT: v_or_b32_e32 v39, v51, v11
+; SI-NEXT: v_lshlrev_b32_e32 v11, 16, v53
+; SI-NEXT: v_mov_b32_e32 v38, v3
+; SI-NEXT: v_or_b32_e32 v3, v54, v11
+; SI-NEXT: s_waitcnt expcnt(5)
+; SI-NEXT: v_mov_b32_e32 v54, v7
+; SI-NEXT: v_alignbit_b32 v7, v55, v47, 24
+; SI-NEXT: buffer_store_dword v7, off, s[0:3], s32 offset:364 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_alignbit_b32 v1, v55, v13, 16
-; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:368 ; 4-byte Folded Spill
+; SI-NEXT: v_alignbit_b32 v7, v55, v47, 16
+; SI-NEXT: buffer_store_dword v7, off, s[0:3], s32 offset:368 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_alignbit_b32 v1, v55, v13, 8
-; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:372 ; 4-byte Folded Spill
+; SI-NEXT: v_alignbit_b32 v7, v55, v47, 8
+; SI-NEXT: buffer_store_dword v7, off, s[0:3], s32 offset:372 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_alignbit_b32 v1, v17, v57, 24
-; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:348 ; 4-byte Folded Spill
+; SI-NEXT: v_alignbit_b32 v7, v46, v13, 24
+; SI-NEXT: buffer_store_dword v7, off, s[0:3], s32 offset:344 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_alignbit_b32 v1, v17, v57, 16
-; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:360 ; 4-byte Folded Spill
+; SI-NEXT: v_alignbit_b32 v7, v46, v13, 16
+; SI-NEXT: buffer_store_dword v7, off, s[0:3], s32 offset:352 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_alignbit_b32 v1, v17, v57, 8
-; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:364 ; 4-byte Folded Spill
+; SI-NEXT: v_alignbit_b32 v7, v46, v13, 8
+; SI-NEXT: buffer_store_dword v7, off, s[0:3], s32 offset:360 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_alignbit_b32 v1, v16, v21, 24
-; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:332 ; 4-byte Folded Spill
+; SI-NEXT: v_alignbit_b32 v7, v16, v17, 24
+; SI-NEXT: buffer_store_dword v7, off, s[0:3], s32 offset:328 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_alignbit_b32 v1, v16, v21, 16
-; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:340 ; 4-byte Folded Spill
+; SI-NEXT: v_alignbit_b32 v7, v16, v17, 16
+; SI-NEXT: buffer_store_dword v7, off, s[0:3], s32 offset:336 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_alignbit_b32 v1, v16, v21, 8
-; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:344 ; 4-byte Folded Spill
+; SI-NEXT: v_alignbit_b32 v7, v16, v17, 8
+; SI-NEXT: buffer_store_dword v7, off, s[0:3], s32 offset:340 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_alignbit_b32 v1, v47, v19, 24
-; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:316 ; 4-byte Folded Spill
+; SI-NEXT: v_alignbit_b32 v7, v14, v22, 24
+; SI-NEXT: buffer_store_dword v7, off, s[0:3], s32 offset:312 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_alignbit_b32 v1, v47, v19, 16
-; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:320 ; 4-byte Folded Spill
+; SI-NEXT: v_alignbit_b32 v7, v14, v22, 16
+; SI-NEXT: buffer_store_dword v7, off, s[0:3], s32 offset:316 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_alignbit_b32 v1, v47, v19, 8
-; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:324 ; 4-byte Folded Spill
+; SI-NEXT: v_alignbit_b32 v7, v14, v22, 8
+; SI-NEXT: buffer_store_dword v7, off, s[0:3], s32 offset:320 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_alignbit_b32 v1, v14, v43, 24
-; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:296 ; 4-byte Folded Spill
+; SI-NEXT: v_alignbit_b32 v7, v60, v23, 24
+; SI-NEXT: v_lshlrev_b32_e32 v11, 16, v1
+; SI-NEXT: buffer_store_dword v7, off, s[0:3], s32 offset:292 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_alignbit_b32 v1, v14, v43, 16
-; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:300 ; 4-byte Folded Spill
+; SI-NEXT: v_alignbit_b32 v7, v60, v23, 16
+; SI-NEXT: v_or_b32_e32 v50, v44, v11
+; SI-NEXT: v_lshlrev_b32_e32 v11, 16, v40
+; SI-NEXT: v_mov_b32_e32 v40, v43
+; SI-NEXT: buffer_store_dword v7, off, s[0:3], s32 offset:296 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_alignbit_b32 v1, v14, v43, 8
-; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:496 ; 4-byte Folded Spill
+; SI-NEXT: v_alignbit_b32 v7, v60, v23, 8
+; SI-NEXT: buffer_store_dword v7, off, s[0:3], s32 offset:304 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_alignbit_b32 v1, v60, v42, 24
-; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:276 ; 4-byte Folded Spill
+; SI-NEXT: v_alignbit_b32 v7, v6, v40, 24
+; SI-NEXT: buffer_store_dword v7, off, s[0:3], s32 offset:272 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_alignbit_b32 v1, v60, v42, 16
-; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:284 ; 4-byte Folded Spill
+; SI-NEXT: v_alignbit_b32 v7, v6, v40, 16
+; SI-NEXT: v_mov_b32_e32 v53, v12
+; SI-NEXT: buffer_store_dword v7, off, s[0:3], s32 offset:280 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_alignbit_b32 v1, v60, v42, 8
-; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:292 ; 4-byte Folded Spill
+; SI-NEXT: v_alignbit_b32 v7, v6, v40, 8
+; SI-NEXT: buffer_store_dword v7, off, s[0:3], s32 offset:288 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_alignbit_b32 v1, v24, v22, 24
-; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:260 ; 4-byte Folded Spill
+; SI-NEXT: v_alignbit_b32 v7, v63, v53, 24
+; SI-NEXT: buffer_store_dword v7, off, s[0:3], s32 offset:260 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_alignbit_b32 v1, v24, v22, 16
-; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:264 ; 4-byte Folded Spill
+; SI-NEXT: v_alignbit_b32 v7, v63, v53, 8
+; SI-NEXT: buffer_store_dword v7, off, s[0:3], s32 offset:480 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_alignbit_b32 v1, v24, v22, 8
-; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:272 ; 4-byte Folded Spill
+; SI-NEXT: v_alignbit_b32 v7, v34, v42, 24
+; SI-NEXT: buffer_store_dword v7, off, s[0:3], s32 offset:240 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_alignbit_b32 v1, v34, v4, 24
-; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:240 ; 4-byte Folded Spill
+; SI-NEXT: v_alignbit_b32 v7, v34, v42, 16
+; SI-NEXT: buffer_store_dword v29, off, s[0:3], s32 offset:268 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v7, off, s[0:3], s32 offset:248 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v42, off, s[0:3], s32 offset:484 ; 4-byte Folded Spill
+; SI-NEXT: s_waitcnt expcnt(1)
+; SI-NEXT: v_alignbit_b32 v7, v34, v42, 8
+; SI-NEXT: buffer_store_dword v7, off, s[0:3], s32 offset:252 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_alignbit_b32 v1, v34, v4, 16
-; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:248 ; 4-byte Folded Spill
+; SI-NEXT: v_alignbit_b32 v7, v56, v36, 24
+; SI-NEXT: buffer_store_dword v7, off, s[0:3], s32 offset:224 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_alignbit_b32 v1, v34, v4, 8
-; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:252 ; 4-byte Folded Spill
+; SI-NEXT: v_alignbit_b32 v7, v56, v36, 16
+; SI-NEXT: buffer_store_dword v7, off, s[0:3], s32 offset:228 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_alignbit_b32 v1, v59, v3, 24
-; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:224 ; 4-byte Folded Spill
+; SI-NEXT: v_alignbit_b32 v7, v56, v36, 8
+; SI-NEXT: buffer_store_dword v7, off, s[0:3], s32 offset:232 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_alignbit_b32 v1, v59, v3, 16
-; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:228 ; 4-byte Folded Spill
+; SI-NEXT: v_alignbit_b32 v7, v58, v38, 24
+; SI-NEXT: buffer_store_dword v7, off, s[0:3], s32 offset:204 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_alignbit_b32 v1, v59, v3, 8
-; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:232 ; 4-byte Folded Spill
-; SI-NEXT: v_lshrrev_b32_e32 v26, 8, v34
-; SI-NEXT: s_waitcnt vmcnt(14)
-; SI-NEXT: v_lshlrev_b32_e32 v11, 16, v38
-; SI-NEXT: v_or_b32_e32 v61, v50, v11
-; SI-NEXT: v_lshlrev_b32_e32 v11, 16, v49
-; SI-NEXT: v_or_b32_e32 v2, v48, v11
-; SI-NEXT: buffer_load_dword v48, off, s[0:3], s32 offset:492 ; 4-byte Folded Reload
-; SI-NEXT: v_mov_b32_e32 v49, v6
+; SI-NEXT: v_alignbit_b32 v7, v58, v38, 16
+; SI-NEXT: buffer_store_dword v7, off, s[0:3], s32 offset:208 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_alignbit_b32 v1, v62, v49, 24
-; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:204 ; 4-byte Folded Spill
+; SI-NEXT: v_alignbit_b32 v7, v58, v38, 8
+; SI-NEXT: buffer_store_dword v7, off, s[0:3], s32 offset:216 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_alignbit_b32 v1, v62, v49, 16
-; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:208 ; 4-byte Folded Spill
+; SI-NEXT: v_alignbit_b32 v7, v25, v26, 24
+; SI-NEXT: buffer_store_dword v7, off, s[0:3], s32 offset:188 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_alignbit_b32 v1, v62, v49, 8
-; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:216 ; 4-byte Folded Spill
+; SI-NEXT: v_alignbit_b32 v7, v25, v26, 16
+; SI-NEXT: v_or_b32_e32 v15, v45, v11
+; SI-NEXT: v_mov_b32_e32 v45, v27
+; SI-NEXT: buffer_store_dword v7, off, s[0:3], s32 offset:192 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_alignbit_b32 v1, v25, v36, 24
-; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:188 ; 4-byte Folded Spill
+; SI-NEXT: v_alignbit_b32 v7, v25, v26, 8
+; SI-NEXT: buffer_store_dword v7, off, s[0:3], s32 offset:200 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_alignbit_b32 v1, v25, v36, 16
-; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:192 ; 4-byte Folded Spill
+; SI-NEXT: v_alignbit_b32 v7, v45, v29, 24
+; SI-NEXT: buffer_store_dword v7, off, s[0:3], s32 offset:172 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_alignbit_b32 v1, v25, v36, 8
-; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:200 ; 4-byte Folded Spill
+; SI-NEXT: v_alignbit_b32 v7, v45, v29, 16
+; SI-NEXT: buffer_store_dword v7, off, s[0:3], s32 offset:180 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_alignbit_b32 v1, v23, v35, 24
-; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:172 ; 4-byte Folded Spill
+; SI-NEXT: v_alignbit_b32 v7, v45, v29, 8
+; SI-NEXT: buffer_store_dword v7, off, s[0:3], s32 offset:184 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_alignbit_b32 v1, v23, v35, 16
-; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:180 ; 4-byte Folded Spill
+; SI-NEXT: v_alignbit_b32 v7, v54, v19, 24
+; SI-NEXT: v_lshlrev_b32_e32 v12, 16, v41
+; SI-NEXT: buffer_store_dword v7, off, s[0:3], s32 offset:160 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_alignbit_b32 v1, v23, v35, 8
-; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:184 ; 4-byte Folded Spill
+; SI-NEXT: v_alignbit_b32 v7, v54, v19, 16
+; SI-NEXT: v_alignbit_b32 v11, v54, v19, 8
+; SI-NEXT: v_or_b32_e32 v12, v57, v12
+; SI-NEXT: buffer_store_dword v7, off, s[0:3], s32 offset:164 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_alignbit_b32 v1, v61, v18, 24
-; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:160 ; 4-byte Folded Spill
+; SI-NEXT: v_mov_b32_e32 v7, v19
+; SI-NEXT: buffer_store_dword v11, off, s[0:3], s32 offset:168 ; 4-byte Folded Spill
+; SI-NEXT: v_alignbit_b32 v29, v39, v5, 24
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_alignbit_b32 v1, v61, v18, 16
-; SI-NEXT: s_waitcnt vmcnt(10)
-; SI-NEXT: v_lshlrev_b32_e32 v11, 16, v48
-; SI-NEXT: v_or_b32_e32 v58, v54, v11
-; SI-NEXT: v_lshlrev_b32_e32 v11, 16, v51
-; SI-NEXT: v_or_b32_e32 v6, v53, v11
-; SI-NEXT: buffer_store_dword v6, off, s[0:3], s32 offset:356 ; 4-byte Folded Spill
-; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:164 ; 4-byte Folded Spill
-; SI-NEXT: buffer_store_dword v18, off, s[0:3], s32 offset:352 ; 4-byte Folded Spill
-; SI-NEXT: s_waitcnt expcnt(1)
-; SI-NEXT: v_alignbit_b32 v1, v61, v18, 8
-; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:168 ; 4-byte Folded Spill
+; SI-NEXT: v_alignbit_b32 v11, v39, v5, 16
+; SI-NEXT: v_mov_b32_e32 v19, v5
+; SI-NEXT: v_alignbit_b32 v5, v39, v5, 8
+; SI-NEXT: buffer_store_dword v5, off, s[0:3], s32 offset:156 ; 4-byte Folded Spill
+; SI-NEXT: v_alignbit_b32 v42, v50, v3, 24
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_alignbit_b32 v1, v58, v2, 16
-; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:152 ; 4-byte Folded Spill
-; SI-NEXT: buffer_store_dword v2, off, s[0:3], s32 offset:308 ; 4-byte Folded Spill
-; SI-NEXT: s_waitcnt expcnt(1)
-; SI-NEXT: v_alignbit_b32 v1, v58, v2, 8
-; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:156 ; 4-byte Folded Spill
+; SI-NEXT: v_alignbit_b32 v5, v50, v3, 16
+; SI-NEXT: v_mov_b32_e32 v59, v3
+; SI-NEXT: v_alignbit_b32 v57, v50, v3, 8
+; SI-NEXT: v_alignbit_b32 v3, v12, v15, 24
+; SI-NEXT: buffer_store_dword v3, off, s[0:3], s32 offset:148 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_lshrrev_b32_e32 v1, 8, v55
-; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:432 ; 4-byte Folded Spill
+; SI-NEXT: v_lshrrev_b32_e32 v3, 8, v55
+; SI-NEXT: buffer_store_dword v3, off, s[0:3], s32 offset:428 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_lshrrev_b32_e32 v1, 8, v17
-; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:424 ; 4-byte Folded Spill
+; SI-NEXT: v_lshrrev_b32_e32 v3, 8, v46
+; SI-NEXT: buffer_store_dword v3, off, s[0:3], s32 offset:416 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_lshrrev_b32_e32 v1, 8, v16
-; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:412 ; 4-byte Folded Spill
+; SI-NEXT: v_lshrrev_b32_e32 v3, 8, v16
+; SI-NEXT: buffer_store_dword v3, off, s[0:3], s32 offset:408 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_lshrrev_b32_e32 v1, 8, v47
-; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:404 ; 4-byte Folded Spill
+; SI-NEXT: v_lshrrev_b32_e32 v3, 8, v14
+; SI-NEXT: buffer_store_dword v3, off, s[0:3], s32 offset:400 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_lshrrev_b32_e32 v1, 8, v14
-; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:396 ; 4-byte Folded Spill
+; SI-NEXT: v_lshrrev_b32_e32 v3, 8, v60
+; SI-NEXT: buffer_store_dword v3, off, s[0:3], s32 offset:392 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_lshrrev_b32_e32 v1, 8, v60
-; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:388 ; 4-byte Folded Spill
+; SI-NEXT: v_lshrrev_b32_e32 v3, 8, v6
+; SI-NEXT: buffer_store_dword v3, off, s[0:3], s32 offset:384 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_lshrrev_b32_e32 v1, 8, v24
-; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:376 ; 4-byte Folded Spill
+; SI-NEXT: v_lshrrev_b32_e32 v3, 8, v63
+; SI-NEXT: buffer_store_dword v3, off, s[0:3], s32 offset:376 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_lshrrev_b32_e32 v1, 8, v59
-; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:328 ; 4-byte Folded Spill
+; SI-NEXT: v_lshrrev_b32_e32 v3, 8, v34
+; SI-NEXT: buffer_store_dword v3, off, s[0:3], s32 offset:356 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_lshrrev_b32_e32 v1, 8, v62
-; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:304 ; 4-byte Folded Spill
+; SI-NEXT: v_lshrrev_b32_e32 v3, 8, v56
+; SI-NEXT: buffer_store_dword v3, off, s[0:3], s32 offset:324 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_lshrrev_b32_e32 v1, 8, v25
-; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:280 ; 4-byte Folded Spill
+; SI-NEXT: v_lshrrev_b32_e32 v3, 8, v58
+; SI-NEXT: buffer_store_dword v3, off, s[0:3], s32 offset:300 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_lshrrev_b32_e32 v1, 8, v23
-; SI-NEXT: v_lshlrev_b32_e32 v11, 16, v37
-; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:256 ; 4-byte Folded Spill
+; SI-NEXT: v_lshrrev_b32_e32 v3, 8, v25
+; SI-NEXT: buffer_store_dword v3, off, s[0:3], s32 offset:276 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_lshrrev_b32_e32 v1, 8, v61
-; SI-NEXT: v_or_b32_e32 v54, v40, v11
-; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:236 ; 4-byte Folded Spill
+; SI-NEXT: v_lshrrev_b32_e32 v3, 8, v45
+; SI-NEXT: buffer_store_dword v3, off, s[0:3], s32 offset:256 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_lshrrev_b32_e32 v1, 8, v58
-; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:212 ; 4-byte Folded Spill
+; SI-NEXT: v_lshrrev_b32_e32 v3, 8, v54
+; SI-NEXT: buffer_store_dword v3, off, s[0:3], s32 offset:236 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_lshrrev_b32_e32 v1, 8, v54
-; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:196 ; 4-byte Folded Spill
+; SI-NEXT: v_lshrrev_b32_e32 v3, 8, v39
+; SI-NEXT: buffer_store_dword v3, off, s[0:3], s32 offset:212 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_lshrrev_b32_e32 v1, 8, v12
-; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:176 ; 4-byte Folded Spill
+; SI-NEXT: v_lshrrev_b32_e32 v3, 8, v50
+; SI-NEXT: buffer_store_dword v3, off, s[0:3], s32 offset:196 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_bfe_u32 v1, v44, 8, 8
-; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:440 ; 4-byte Folded Spill
+; SI-NEXT: v_lshrrev_b32_e32 v3, 8, v12
+; SI-NEXT: buffer_store_dword v3, off, s[0:3], s32 offset:176 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_bfe_u32 v1, v20, 8, 8
-; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:436 ; 4-byte Folded Spill
+; SI-NEXT: v_bfe_u32 v3, v20, 8, 8
+; SI-NEXT: buffer_store_dword v3, off, s[0:3], s32 offset:436 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_bfe_u32 v1, v15, 8, 8
-; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:428 ; 4-byte Folded Spill
+; SI-NEXT: v_bfe_u32 v3, v18, 8, 8
+; SI-NEXT: buffer_store_dword v3, off, s[0:3], s32 offset:432 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_bfe_u32 v1, v29, 8, 8
-; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:420 ; 4-byte Folded Spill
+; SI-NEXT: v_bfe_u32 v3, v48, 8, 8
+; SI-NEXT: buffer_store_dword v3, off, s[0:3], s32 offset:424 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_bfe_u32 v1, v33, 8, 8
-; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:416 ; 4-byte Folded Spill
+; SI-NEXT: v_bfe_u32 v3, v24, 8, 8
+; SI-NEXT: buffer_store_dword v3, off, s[0:3], s32 offset:420 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_bfe_u32 v1, v32, 8, 8
-; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:408 ; 4-byte Folded Spill
+; SI-NEXT: v_bfe_u32 v3, v21, 8, 8
+; SI-NEXT: buffer_store_dword v3, off, s[0:3], s32 offset:412 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_bfe_u32 v1, v31, 8, 8
-; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:400 ; 4-byte Folded Spill
+; SI-NEXT: v_bfe_u32 v3, v32, 8, 8
+; SI-NEXT: buffer_store_dword v3, off, s[0:3], s32 offset:404 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_bfe_u32 v1, v10, 8, 8
-; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:392 ; 4-byte Folded Spill
+; SI-NEXT: v_bfe_u32 v3, v31, 8, 8
+; SI-NEXT: buffer_store_dword v3, off, s[0:3], s32 offset:396 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_bfe_u32 v1, v9, 8, 8
-; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:380 ; 4-byte Folded Spill
+; SI-NEXT: v_bfe_u32 v3, v10, 8, 8
+; SI-NEXT: buffer_store_dword v3, off, s[0:3], s32 offset:388 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_bfe_u32 v1, v8, 8, 8
-; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:336 ; 4-byte Folded Spill
+; SI-NEXT: v_bfe_u32 v3, v9, 8, 8
+; SI-NEXT: buffer_store_dword v3, off, s[0:3], s32 offset:380 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_bfe_u32 v1, v5, 8, 8
-; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:312 ; 4-byte Folded Spill
+; SI-NEXT: v_bfe_u32 v3, v8, 8, 8
+; SI-NEXT: buffer_store_dword v3, off, s[0:3], s32 offset:348 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_bfe_u32 v1, v38, 8, 8
-; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:288 ; 4-byte Folded Spill
+; SI-NEXT: v_bfe_u32 v3, v30, 8, 8
+; SI-NEXT: buffer_store_dword v3, off, s[0:3], s32 offset:332 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_bfe_u32 v1, v48, 8, 8
-; SI-NEXT: v_lshlrev_b32_e32 v11, 16, v41
-; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:268 ; 4-byte Folded Spill
+; SI-NEXT: v_bfe_u32 v3, v62, 8, 8
+; SI-NEXT: v_mov_b32_e32 v44, v1
+; SI-NEXT: v_bfe_u32 v1, v1, 8, 8
+; SI-NEXT: buffer_store_dword v3, off, s[0:3], s32 offset:308 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_bfe_u32 v1, v37, 8, 8
-; SI-NEXT: v_or_b32_e32 v11, v45, v11
+; SI-NEXT: v_bfe_u32 v3, v4, 8, 8
; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:244 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_bfe_u32 v1, v52, 8, 8
-; SI-NEXT: v_alignbit_b32 v28, v58, v2, 24
-; SI-NEXT: v_alignbit_b32 v2, v54, v6, 24
-; SI-NEXT: v_alignbit_b32 v39, v54, v6, 16
-; SI-NEXT: v_alignbit_b32 v40, v54, v6, 8
-; SI-NEXT: v_alignbit_b32 v27, v12, v11, 24
-; SI-NEXT: v_alignbit_b32 v56, v12, v11, 16
-; SI-NEXT: buffer_store_dword v11, off, s[0:3], s32 offset:384 ; 4-byte Folded Spill
-; SI-NEXT: v_alignbit_b32 v30, v12, v11, 8
-; SI-NEXT: v_mov_b32_e32 v20, v29
-; SI-NEXT: v_mov_b32_e32 v15, v33
-; SI-NEXT: v_mov_b32_e32 v33, v32
+; SI-NEXT: v_bfe_u32 v1, v41, 8, 8
+; SI-NEXT: v_alignbit_b32 v27, v63, v53, 16
+; SI-NEXT: buffer_store_dword v11, off, s[0:3], s32 offset:152 ; 4-byte Folded Spill
+; SI-NEXT: v_alignbit_b32 v26, v12, v15, 16
+; SI-NEXT: s_waitcnt expcnt(0)
+; SI-NEXT: v_mov_b32_e32 v11, v15
+; SI-NEXT: v_alignbit_b32 v43, v12, v15, 8
+; SI-NEXT: v_mov_b32_e32 v52, v20
+; SI-NEXT: v_mov_b32_e32 v20, v18
+; SI-NEXT: v_mov_b32_e32 v18, v48
+; SI-NEXT: v_mov_b32_e32 v15, v24
+; SI-NEXT: v_mov_b32_e32 v28, v21
+; SI-NEXT: v_mov_b32_e32 v21, v23
+; SI-NEXT: v_mov_b32_e32 v23, v22
+; SI-NEXT: v_mov_b32_e32 v22, v17
+; SI-NEXT: v_mov_b32_e32 v17, v13
+; SI-NEXT: v_mov_b32_e32 v13, v47
; SI-NEXT: v_mov_b32_e32 v32, v31
; SI-NEXT: v_mov_b32_e32 v31, v10
; SI-NEXT: v_mov_b32_e32 v10, v9
-; SI-NEXT: v_mov_b32_e32 v9, v7
-; SI-NEXT: v_bfe_u32 v29, v7, 8, 8
-; SI-NEXT: v_mov_b32_e32 v7, v8
-; SI-NEXT: v_mov_b32_e32 v8, v5
-; SI-NEXT: v_mov_b32_e32 v44, v37
+; SI-NEXT: v_mov_b32_e32 v51, v8
+; SI-NEXT: v_mov_b32_e32 v48, v30
+; SI-NEXT: v_mov_b32_e32 v8, v4
+; SI-NEXT: buffer_store_dword v3, off, s[0:3], s32 offset:284 ; 4-byte Folded Spill
+; SI-NEXT: s_waitcnt expcnt(0)
+; SI-NEXT: v_mov_b32_e32 v3, v2
+; SI-NEXT: v_bfe_u32 v30, v2, 8, 8
; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:220 ; 4-byte Folded Spill
+; SI-NEXT: v_mov_b32_e32 v37, v62
; SI-NEXT: s_branch .LBB95_3
; SI-NEXT: .LBB95_2:
-; SI-NEXT: buffer_load_dword v48, off, s[0:3], s32 offset:492 ; 4-byte Folded Reload
+; SI-NEXT: v_mov_b32_e32 v52, v20
+; SI-NEXT: v_mov_b32_e32 v20, v18
+; SI-NEXT: v_mov_b32_e32 v18, v48
+; SI-NEXT: buffer_load_dword v51, off, s[0:3], s32 offset:472 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v48, off, s[0:3], s32 offset:476 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v37, off, s[0:3], s32 offset:468 ; 4-byte Folded Reload
+; SI-NEXT: s_waitcnt expcnt(5)
+; SI-NEXT: v_mov_b32_e32 v44, v1
; SI-NEXT: ; implicit-def: $vgpr1
; SI-NEXT: ; kill: killed $vgpr1
; SI-NEXT: ; implicit-def: $vgpr1
@@ -196537,391 +196641,390 @@ define inreg <128 x i8> @bitcast_v64f16_to_v128i8_scalar(<64 x half> inreg %a, i
; SI-NEXT: ; implicit-def: $vgpr1
; SI-NEXT: ; kill: killed $vgpr1
; SI-NEXT: ; implicit-def: $vgpr1
-; SI-NEXT: v_mov_b32_e32 v20, v29
-; SI-NEXT: v_mov_b32_e32 v15, v33
-; SI-NEXT: v_mov_b32_e32 v33, v32
-; SI-NEXT: v_mov_b32_e32 v32, v31
-; SI-NEXT: v_mov_b32_e32 v31, v10
-; SI-NEXT: v_mov_b32_e32 v10, v9
-; SI-NEXT: v_mov_b32_e32 v9, v7
-; SI-NEXT: v_mov_b32_e32 v7, v8
-; SI-NEXT: v_mov_b32_e32 v8, v5
-; SI-NEXT: v_mov_b32_e32 v44, v37
+; SI-NEXT: v_mov_b32_e32 v15, v32
+; SI-NEXT: v_mov_b32_e32 v28, v6
+; SI-NEXT: v_mov_b32_e32 v33, v31
+; SI-NEXT: v_mov_b32_e32 v32, v10
+; SI-NEXT: v_mov_b32_e32 v31, v9
+; SI-NEXT: v_mov_b32_e32 v10, v8
+; SI-NEXT: v_mov_b32_e32 v8, v4
+; SI-NEXT: v_mov_b32_e32 v3, v2
; SI-NEXT: s_mov_b64 s[4:5], -1
; SI-NEXT: ; kill: killed $vgpr1
; SI-NEXT: ; implicit-def: $vgpr1
; SI-NEXT: ; implicit-def: $vgpr13
; SI-NEXT: ; implicit-def: $vgpr55
-; SI-NEXT: ; implicit-def: $vgpr57
; SI-NEXT: ; implicit-def: $vgpr17
-; SI-NEXT: ; implicit-def: $vgpr21
+; SI-NEXT: ; implicit-def: $vgpr46
+; SI-NEXT: ; implicit-def: $vgpr22
; SI-NEXT: ; implicit-def: $vgpr16
-; SI-NEXT: ; implicit-def: $vgpr19
-; SI-NEXT: ; implicit-def: $vgpr47
-; SI-NEXT: ; implicit-def: $vgpr43
+; SI-NEXT: ; implicit-def: $vgpr23
; SI-NEXT: ; implicit-def: $vgpr14
-; SI-NEXT: ; implicit-def: $vgpr42
+; SI-NEXT: ; implicit-def: $vgpr21
; SI-NEXT: ; implicit-def: $vgpr60
-; SI-NEXT: ; implicit-def: $vgpr22
-; SI-NEXT: ; implicit-def: $vgpr24
-; SI-NEXT: ; implicit-def: $vgpr4
+; SI-NEXT: ; implicit-def: $vgpr40
+; SI-NEXT: ; implicit-def: $vgpr6
+; SI-NEXT: ; implicit-def: $vgpr53
+; SI-NEXT: ; implicit-def: $vgpr27
+; SI-NEXT: ; implicit-def: $vgpr63
; SI-NEXT: ; implicit-def: $vgpr34
-; SI-NEXT: ; implicit-def: $vgpr26
-; SI-NEXT: ; implicit-def: $vgpr3
-; SI-NEXT: ; implicit-def: $vgpr59
-; SI-NEXT: ; implicit-def: $vgpr49
-; SI-NEXT: ; implicit-def: $vgpr62
-; SI-NEXT: ; implicit-def: $vgpr29
; SI-NEXT: ; implicit-def: $vgpr36
-; SI-NEXT: ; implicit-def: $vgpr25
-; SI-NEXT: ; implicit-def: $vgpr35
-; SI-NEXT: ; implicit-def: $vgpr23
-; SI-NEXT: ; implicit-def: $vgpr61
-; SI-NEXT: ; implicit-def: $vgpr28
+; SI-NEXT: ; implicit-def: $vgpr56
+; SI-NEXT: ; implicit-def: $vgpr38
; SI-NEXT: ; implicit-def: $vgpr58
-; SI-NEXT: ; implicit-def: $vgpr40
-; SI-NEXT: ; implicit-def: $vgpr39
-; SI-NEXT: ; implicit-def: $vgpr2
+; SI-NEXT: ; implicit-def: $vgpr25
+; SI-NEXT: ; implicit-def: $vgpr45
+; SI-NEXT: ; implicit-def: $vgpr7
; SI-NEXT: ; implicit-def: $vgpr54
+; SI-NEXT: ; implicit-def: $vgpr19
+; SI-NEXT: ; implicit-def: $vgpr29
+; SI-NEXT: ; implicit-def: $vgpr39
; SI-NEXT: ; implicit-def: $vgpr30
-; SI-NEXT: ; implicit-def: $vgpr56
-; SI-NEXT: ; implicit-def: $vgpr27
+; SI-NEXT: ; implicit-def: $vgpr59
+; SI-NEXT: ; implicit-def: $vgpr57
+; SI-NEXT: ; implicit-def: $vgpr5
+; SI-NEXT: ; implicit-def: $vgpr42
+; SI-NEXT: ; implicit-def: $vgpr50
+; SI-NEXT: ; implicit-def: $vgpr11
+; SI-NEXT: ; implicit-def: $vgpr43
+; SI-NEXT: ; implicit-def: $vgpr26
; SI-NEXT: ; implicit-def: $vgpr12
; SI-NEXT: ; kill: killed $vgpr1
; SI-NEXT: ; implicit-def: $vgpr1
; SI-NEXT: ; kill: killed $vgpr1
; SI-NEXT: .LBB95_3: ; %Flow
-; SI-NEXT: s_waitcnt expcnt(5)
-; SI-NEXT: buffer_load_dword v45, off, s[0:3], s32 offset:480 ; 4-byte Folded Reload
-; SI-NEXT: s_waitcnt expcnt(3)
-; SI-NEXT: buffer_load_dword v41, off, s[0:3], s32 offset:484 ; 4-byte Folded Reload
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_mov_b32_e32 v1, v44
-; SI-NEXT: buffer_load_dword v46, off, s[0:3], s32 offset:488 ; 4-byte Folded Reload
-; SI-NEXT: buffer_load_dword v51, off, s[0:3], s32 offset:476 ; 4-byte Folded Reload
-; SI-NEXT: buffer_load_dword v37, off, s[0:3], s32 offset:452 ; 4-byte Folded Reload
-; SI-NEXT: v_mov_b32_e32 v5, v8
-; SI-NEXT: v_mov_b32_e32 v6, v7
-; SI-NEXT: v_mov_b32_e32 v7, v9
-; SI-NEXT: v_mov_b32_e32 v8, v10
+; SI-NEXT: buffer_load_dword v44, off, s[0:3], s32 offset:444 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v49, off, s[0:3], s32 offset:464 ; 4-byte Folded Reload
+; SI-NEXT: s_waitcnt vmcnt(3)
+; SI-NEXT: v_mov_b32_e32 v47, v48
+; SI-NEXT: v_mov_b32_e32 v62, v51
+; SI-NEXT: buffer_load_dword v61, off, s[0:3], s32 offset:448 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v48, off, s[0:3], s32 offset:452 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v51, off, s[0:3], s32 offset:456 ; 4-byte Folded Reload
+; SI-NEXT: v_mov_b32_e32 v2, v3
; SI-NEXT: v_mov_b32_e32 v9, v31
; SI-NEXT: v_mov_b32_e32 v31, v33
-; SI-NEXT: v_mov_b32_e32 v44, v15
-; SI-NEXT: v_mov_b32_e32 v33, v20
-; SI-NEXT: buffer_load_dword v15, off, s[0:3], s32 offset:456 ; 4-byte Folded Reload
-; SI-NEXT: buffer_load_dword v18, off, s[0:3], s32 offset:460 ; 4-byte Folded Reload
-; SI-NEXT: buffer_load_dword v20, off, s[0:3], s32 offset:464 ; 4-byte Folded Reload
-; SI-NEXT: buffer_load_dword v10, off, s[0:3], s32 offset:468 ; 4-byte Folded Reload
-; SI-NEXT: buffer_load_dword v63, off, s[0:3], s32 offset:472 ; 4-byte Folded Reload
-; SI-NEXT: v_mov_b32_e32 v50, v2
-; SI-NEXT: v_mov_b32_e32 v53, v40
-; SI-NEXT: v_mov_b32_e32 v40, v28
-; SI-NEXT: s_waitcnt vmcnt(10)
-; SI-NEXT: v_mov_b32_e32 v2, v48
-; SI-NEXT: buffer_load_dword v28, off, s[0:3], s32 offset:496 ; 4-byte Folded Reload
-; SI-NEXT: buffer_load_dword v48, off, s[0:3], s32 offset:500 ; 4-byte Folded Reload
+; SI-NEXT: v_mov_b32_e32 v33, v15
+; SI-NEXT: v_mov_b32_e32 v15, v18
+; SI-NEXT: v_mov_b32_e32 v18, v20
+; SI-NEXT: v_mov_b32_e32 v20, v52
+; SI-NEXT: buffer_load_dword v52, off, s[0:3], s32 offset:460 ; 4-byte Folded Reload
+; SI-NEXT: v_mov_b32_e32 v3, v41
+; SI-NEXT: v_mov_b32_e32 v35, v30
+; SI-NEXT: buffer_load_dword v30, off, s[0:3], s32 offset:480 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v41, off, s[0:3], s32 offset:484 ; 4-byte Folded Reload
; SI-NEXT: s_andn2_b64 vcc, exec, s[4:5]
-; SI-NEXT: v_mov_b32_e32 v11, v27
-; SI-NEXT: v_mov_b32_e32 v38, v30
-; SI-NEXT: v_mov_b32_e32 v27, v52
-; SI-NEXT: v_mov_b32_e32 v30, v29
-; SI-NEXT: v_mov_b32_e32 v29, v26
+; SI-NEXT: v_mov_b32_e32 v4, v8
+; SI-NEXT: v_mov_b32_e32 v8, v10
+; SI-NEXT: v_mov_b32_e32 v10, v32
+; SI-NEXT: v_mov_b32_e32 v32, v28
+; SI-NEXT: v_mov_b32_e32 v28, v43
+; SI-NEXT: v_mov_b32_e32 v43, v42
+; SI-NEXT: v_mov_b32_e32 v42, v5
+; SI-NEXT: v_mov_b32_e32 v24, v29
+; SI-NEXT: v_mov_b32_e32 v29, v1
+; SI-NEXT: v_mov_b32_e32 v1, v19
+; SI-NEXT: v_mov_b32_e32 v19, v7
+; SI-NEXT: s_waitcnt vmcnt(8)
+; SI-NEXT: v_mov_b32_e32 v5, v37
; SI-NEXT: s_cbranch_vccnz .LBB95_5
; SI-NEXT: ; %bb.4: ; %cmp.true
-; SI-NEXT: buffer_load_dword v3, off, s[0:3], s32 offset:652 ; 4-byte Folded Reload
-; SI-NEXT: buffer_load_dword v4, off, s[0:3], s32 offset:656 ; 4-byte Folded Reload
-; SI-NEXT: v_cvt_f32_f16_e32 v1, v1
+; SI-NEXT: buffer_load_dword v7, off, s[0:3], s32 offset:644 ; 4-byte Folded Reload
+; SI-NEXT: v_cvt_f32_f16_e32 v3, v3
+; SI-NEXT: v_cvt_f32_f16_e32 v1, v29
; SI-NEXT: v_cvt_f32_f16_e32 v2, v2
-; SI-NEXT: v_cvt_f32_f16_e32 v5, v5
-; SI-NEXT: v_cvt_f32_f16_e32 v6, v6
+; SI-NEXT: v_cvt_f32_f16_e32 v4, v4
+; SI-NEXT: v_add_f32_e32 v3, 0x38000000, v3
+; SI-NEXT: v_cvt_f16_f32_e32 v3, v3
; SI-NEXT: v_add_f32_e32 v1, 0x38000000, v1
; SI-NEXT: v_cvt_f16_f32_e32 v1, v1
; SI-NEXT: v_add_f32_e32 v2, 0x38000000, v2
+; SI-NEXT: v_lshlrev_b32_e32 v13, 16, v3
; SI-NEXT: v_cvt_f16_f32_e32 v2, v2
-; SI-NEXT: v_add_f32_e32 v5, 0x38000000, v5
; SI-NEXT: v_lshlrev_b32_e32 v16, 16, v1
-; SI-NEXT: v_cvt_f16_f32_e32 v5, v5
+; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:468 ; 4-byte Folded Spill
+; SI-NEXT: v_add_f32_e32 v4, 0x38000000, v4
; SI-NEXT: v_lshlrev_b32_e32 v19, 16, v2
-; SI-NEXT: v_add_f32_e32 v6, 0x38000000, v6
-; SI-NEXT: v_cvt_f16_f32_e32 v6, v6
-; SI-NEXT: v_lshlrev_b32_e32 v25, 16, v5
-; SI-NEXT: v_cvt_f32_f16_e32 v7, v7
+; SI-NEXT: v_cvt_f16_f32_e32 v4, v4
+; SI-NEXT: v_cvt_f32_f16_e32 v5, v5
+; SI-NEXT: v_cvt_f32_f16_e32 v6, v47
; SI-NEXT: v_cvt_f32_f16_e32 v8, v8
-; SI-NEXT: s_waitcnt vmcnt(9)
-; SI-NEXT: v_cvt_f32_f16_e32 v35, v37
-; SI-NEXT: v_cvt_f32_f16_e32 v9, v9
-; SI-NEXT: v_add_f32_e32 v7, 0x38000000, v7
-; SI-NEXT: v_cvt_f16_f32_e32 v7, v7
+; SI-NEXT: v_lshlrev_b32_e32 v22, 16, v4
+; SI-NEXT: v_add_f32_e32 v5, 0x38000000, v5
+; SI-NEXT: v_cvt_f16_f32_e32 v5, v5
+; SI-NEXT: v_add_f32_e32 v6, 0x38000000, v6
+; SI-NEXT: v_cvt_f16_f32_e32 v47, v6
; SI-NEXT: v_add_f32_e32 v8, 0x38000000, v8
+; SI-NEXT: v_lshlrev_b32_e32 v25, 16, v5
; SI-NEXT: v_cvt_f16_f32_e32 v8, v8
+; SI-NEXT: s_waitcnt vmcnt(5)
+; SI-NEXT: v_cvt_f32_f16_e32 v35, v51
+; SI-NEXT: v_cvt_f32_f16_e32 v9, v9
+; SI-NEXT: v_cvt_f32_f16_e32 v10, v10
+; SI-NEXT: v_lshlrev_b32_e32 v34, 16, v8
+; SI-NEXT: v_cvt_f32_f16_e32 v37, v49
; SI-NEXT: v_add_f32_e32 v9, 0x38000000, v9
; SI-NEXT: v_cvt_f16_f32_e32 v9, v9
-; SI-NEXT: v_cvt_f32_f16_e32 v38, v41
+; SI-NEXT: v_add_f32_e32 v10, 0x38000000, v10
+; SI-NEXT: v_cvt_f16_f32_e32 v10, v10
+; SI-NEXT: v_add_f32_e32 v37, 0x38000000, v37
+; SI-NEXT: v_cvt_f16_f32_e32 v37, v37
+; SI-NEXT: v_cvt_f32_f16_e32 v39, v44
; SI-NEXT: v_cvt_f32_f16_e32 v31, v31
+; SI-NEXT: v_cvt_f32_f16_e32 v32, v32
; SI-NEXT: v_cvt_f32_f16_e32 v33, v33
-; SI-NEXT: s_waitcnt vmcnt(8)
-; SI-NEXT: v_cvt_f32_f16_e32 v15, v15
-; SI-NEXT: v_add_f32_e32 v38, 0x38000000, v38
-; SI-NEXT: v_cvt_f16_f32_e32 v38, v38
+; SI-NEXT: v_add_f32_e32 v39, 0x38000000, v39
+; SI-NEXT: v_cvt_f16_f32_e32 v39, v39
; SI-NEXT: v_add_f32_e32 v31, 0x38000000, v31
; SI-NEXT: v_cvt_f16_f32_e32 v31, v31
+; SI-NEXT: v_add_f32_e32 v32, 0x38000000, v32
+; SI-NEXT: v_cvt_f16_f32_e32 v32, v32
; SI-NEXT: v_add_f32_e32 v33, 0x38000000, v33
; SI-NEXT: v_cvt_f16_f32_e32 v33, v33
-; SI-NEXT: v_add_f32_e32 v15, 0x38000000, v15
-; SI-NEXT: v_cvt_f16_f32_e32 v15, v15
-; SI-NEXT: s_waitcnt vmcnt(7)
+; SI-NEXT: v_cvt_f32_f16_e32 v15, v15
; SI-NEXT: v_cvt_f32_f16_e32 v18, v18
-; SI-NEXT: s_waitcnt vmcnt(6)
; SI-NEXT: v_cvt_f32_f16_e32 v20, v20
-; SI-NEXT: buffer_load_dword v49, off, s[0:3], s32 offset:524 ; 4-byte Folded Reload
-; SI-NEXT: buffer_load_dword v39, off, s[0:3], s32 offset:544 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v38, off, s[0:3], s32 offset:536 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v49, off, s[0:3], s32 offset:508 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v51, off, s[0:3], s32 offset:492 ; 4-byte Folded Reload
+; SI-NEXT: v_add_f32_e32 v15, 0x38000000, v15
+; SI-NEXT: v_cvt_f16_f32_e32 v15, v15
; SI-NEXT: v_add_f32_e32 v18, 0x38000000, v18
; SI-NEXT: v_cvt_f16_f32_e32 v18, v18
; SI-NEXT: v_add_f32_e32 v20, 0x38000000, v20
; SI-NEXT: v_cvt_f16_f32_e32 v20, v20
-; SI-NEXT: buffer_load_dword v50, off, s[0:3], s32 offset:548 ; 4-byte Folded Reload
-; SI-NEXT: s_waitcnt vmcnt(4)
-; SI-NEXT: v_cvt_f32_f16_e32 v11, v3
-; SI-NEXT: buffer_load_dword v3, off, s[0:3], s32 offset:648 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v50, off, s[0:3], s32 offset:528 ; 4-byte Folded Reload
+; SI-NEXT: s_waitcnt vmcnt(5)
+; SI-NEXT: v_cvt_f32_f16_e32 v11, v7
+; SI-NEXT: buffer_load_dword v7, off, s[0:3], s32 offset:640 ; 4-byte Folded Reload
; SI-NEXT: v_add_f32_e32 v11, 0x38000000, v11
; SI-NEXT: v_cvt_f16_f32_e32 v11, v11
; SI-NEXT: v_lshlrev_b32_e32 v11, 16, v11
+; SI-NEXT: s_waitcnt vmcnt(4)
+; SI-NEXT: v_cvt_f32_f16_e32 v38, v38
; SI-NEXT: s_waitcnt vmcnt(3)
; SI-NEXT: v_cvt_f32_f16_e32 v49, v49
; SI-NEXT: s_waitcnt vmcnt(2)
-; SI-NEXT: v_cvt_f32_f16_e32 v39, v39
-; SI-NEXT: v_add_f32_e32 v39, 0x38000000, v39
-; SI-NEXT: v_cvt_f16_f32_e32 v39, v39
+; SI-NEXT: v_cvt_f32_f16_e32 v51, v51
+; SI-NEXT: v_add_f32_e32 v38, 0x38000000, v38
+; SI-NEXT: v_cvt_f16_f32_e32 v38, v38
+; SI-NEXT: v_add_f32_e32 v51, 0x38000000, v51
+; SI-NEXT: v_cvt_f16_f32_e32 v51, v51
; SI-NEXT: s_waitcnt vmcnt(1)
; SI-NEXT: v_cvt_f32_f16_e32 v50, v50
; SI-NEXT: v_add_f32_e32 v50, 0x38000000, v50
; SI-NEXT: v_cvt_f16_f32_e32 v50, v50
; SI-NEXT: s_waitcnt vmcnt(0)
-; SI-NEXT: v_cvt_f32_f16_e32 v12, v3
-; SI-NEXT: v_cvt_f32_f16_e32 v3, v27
+; SI-NEXT: v_cvt_f32_f16_e32 v12, v7
+; SI-NEXT: buffer_load_dword v7, off, s[0:3], s32 offset:648 ; 4-byte Folded Reload
; SI-NEXT: v_add_f32_e32 v12, 0x38000000, v12
-; SI-NEXT: v_add_f32_e32 v3, 0x38000000, v3
-; SI-NEXT: v_cvt_f16_f32_e32 v34, v3
-; SI-NEXT: buffer_load_dword v3, off, s[0:3], s32 offset:644 ; 4-byte Folded Reload
; SI-NEXT: v_cvt_f16_f32_e32 v12, v12
-; SI-NEXT: v_lshlrev_b32_e32 v13, 16, v34
-; SI-NEXT: v_or_b32_e32 v56, v12, v11
-; SI-NEXT: v_cvt_f32_f16_e32 v12, v4
+; SI-NEXT: v_or_b32_e32 v11, v12, v11
+; SI-NEXT: s_waitcnt vmcnt(0)
+; SI-NEXT: v_cvt_f32_f16_e32 v12, v7
+; SI-NEXT: buffer_load_dword v7, off, s[0:3], s32 offset:636 ; 4-byte Folded Reload
; SI-NEXT: v_add_f32_e32 v12, 0x38000000, v12
; SI-NEXT: v_cvt_f16_f32_e32 v12, v12
; SI-NEXT: v_or_b32_e32 v12, v12, v13
; SI-NEXT: s_waitcnt vmcnt(0)
-; SI-NEXT: v_cvt_f32_f16_e32 v13, v3
-; SI-NEXT: buffer_load_dword v3, off, s[0:3], s32 offset:640 ; 4-byte Folded Reload
+; SI-NEXT: v_cvt_f32_f16_e32 v13, v7
+; SI-NEXT: buffer_load_dword v7, off, s[0:3], s32 offset:632 ; 4-byte Folded Reload
; SI-NEXT: v_add_f32_e32 v13, 0x38000000, v13
; SI-NEXT: v_cvt_f16_f32_e32 v13, v13
; SI-NEXT: v_lshlrev_b32_e32 v13, 16, v13
; SI-NEXT: s_waitcnt vmcnt(0)
-; SI-NEXT: v_cvt_f32_f16_e32 v14, v3
-; SI-NEXT: buffer_load_dword v3, off, s[0:3], s32 offset:636 ; 4-byte Folded Reload
+; SI-NEXT: v_cvt_f32_f16_e32 v14, v7
+; SI-NEXT: buffer_load_dword v7, off, s[0:3], s32 offset:628 ; 4-byte Folded Reload
; SI-NEXT: v_add_f32_e32 v14, 0x38000000, v14
; SI-NEXT: v_cvt_f16_f32_e32 v14, v14
-; SI-NEXT: v_or_b32_e32 v36, v14, v13
+; SI-NEXT: v_or_b32_e32 v59, v14, v13
; SI-NEXT: s_waitcnt vmcnt(0)
-; SI-NEXT: v_cvt_f32_f16_e32 v14, v3
-; SI-NEXT: buffer_load_dword v3, off, s[0:3], s32 offset:632 ; 4-byte Folded Reload
+; SI-NEXT: v_cvt_f32_f16_e32 v14, v7
+; SI-NEXT: buffer_load_dword v7, off, s[0:3], s32 offset:624 ; 4-byte Folded Reload
; SI-NEXT: v_add_f32_e32 v14, 0x38000000, v14
; SI-NEXT: v_cvt_f16_f32_e32 v14, v14
-; SI-NEXT: v_or_b32_e32 v54, v14, v16
+; SI-NEXT: v_or_b32_e32 v55, v14, v16
; SI-NEXT: s_waitcnt vmcnt(0)
-; SI-NEXT: v_cvt_f32_f16_e32 v16, v3
-; SI-NEXT: buffer_load_dword v3, off, s[0:3], s32 offset:628 ; 4-byte Folded Reload
+; SI-NEXT: v_cvt_f32_f16_e32 v16, v7
+; SI-NEXT: buffer_load_dword v7, off, s[0:3], s32 offset:620 ; 4-byte Folded Reload
; SI-NEXT: v_add_f32_e32 v16, 0x38000000, v16
; SI-NEXT: v_cvt_f16_f32_e32 v16, v16
; SI-NEXT: v_lshlrev_b32_e32 v16, 16, v16
; SI-NEXT: s_waitcnt vmcnt(0)
-; SI-NEXT: v_cvt_f32_f16_e32 v17, v3
-; SI-NEXT: buffer_load_dword v3, off, s[0:3], s32 offset:624 ; 4-byte Folded Reload
+; SI-NEXT: v_cvt_f32_f16_e32 v17, v7
+; SI-NEXT: buffer_load_dword v7, off, s[0:3], s32 offset:616 ; 4-byte Folded Reload
; SI-NEXT: v_add_f32_e32 v17, 0x38000000, v17
; SI-NEXT: v_cvt_f16_f32_e32 v17, v17
-; SI-NEXT: v_or_b32_e32 v52, v17, v16
+; SI-NEXT: s_waitcnt expcnt(0)
+; SI-NEXT: v_or_b32_e32 v1, v17, v16
; SI-NEXT: s_waitcnt vmcnt(0)
-; SI-NEXT: v_cvt_f32_f16_e32 v17, v3
-; SI-NEXT: buffer_load_dword v3, off, s[0:3], s32 offset:620 ; 4-byte Folded Reload
+; SI-NEXT: v_cvt_f32_f16_e32 v17, v7
+; SI-NEXT: buffer_load_dword v7, off, s[0:3], s32 offset:612 ; 4-byte Folded Reload
; SI-NEXT: v_add_f32_e32 v17, 0x38000000, v17
; SI-NEXT: v_cvt_f16_f32_e32 v17, v17
-; SI-NEXT: v_or_b32_e32 v58, v17, v19
-; SI-NEXT: v_alignbit_b32 v40, v58, v52, 24
+; SI-NEXT: v_or_b32_e32 v17, v17, v19
; SI-NEXT: s_waitcnt vmcnt(0)
-; SI-NEXT: v_cvt_f32_f16_e32 v19, v3
-; SI-NEXT: buffer_load_dword v3, off, s[0:3], s32 offset:616 ; 4-byte Folded Reload
+; SI-NEXT: v_cvt_f32_f16_e32 v19, v7
+; SI-NEXT: buffer_load_dword v7, off, s[0:3], s32 offset:608 ; 4-byte Folded Reload
; SI-NEXT: v_add_f32_e32 v19, 0x38000000, v19
; SI-NEXT: v_cvt_f16_f32_e32 v19, v19
; SI-NEXT: v_lshlrev_b32_e32 v19, 16, v19
; SI-NEXT: s_waitcnt vmcnt(0)
-; SI-NEXT: v_cvt_f32_f16_e32 v21, v3
-; SI-NEXT: buffer_load_dword v3, off, s[0:3], s32 offset:148 ; 4-byte Folded Reload
+; SI-NEXT: v_cvt_f32_f16_e32 v21, v7
+; SI-NEXT: buffer_load_dword v7, off, s[0:3], s32 offset:604 ; 4-byte Folded Reload
; SI-NEXT: v_add_f32_e32 v21, 0x38000000, v21
; SI-NEXT: v_cvt_f16_f32_e32 v21, v21
-; SI-NEXT: v_or_b32_e32 v11, v21, v19
+; SI-NEXT: v_or_b32_e32 v19, v21, v19
; SI-NEXT: s_waitcnt vmcnt(0)
-; SI-NEXT: v_cvt_f32_f16_e32 v4, v3
-; SI-NEXT: buffer_load_dword v3, off, s[0:3], s32 offset:612 ; 4-byte Folded Reload
-; SI-NEXT: v_add_f32_e32 v4, 0x38000000, v4
-; SI-NEXT: v_cvt_f16_f32_e32 v13, v4
-; SI-NEXT: buffer_load_dword v4, off, s[0:3], s32 offset:584 ; 4-byte Folded Reload
-; SI-NEXT: v_lshlrev_b32_e32 v22, 16, v13
-; SI-NEXT: s_waitcnt vmcnt(1)
-; SI-NEXT: v_cvt_f32_f16_e32 v21, v3
-; SI-NEXT: buffer_load_dword v3, off, s[0:3], s32 offset:608 ; 4-byte Folded Reload
+; SI-NEXT: v_cvt_f32_f16_e32 v21, v7
+; SI-NEXT: buffer_load_dword v7, off, s[0:3], s32 offset:600 ; 4-byte Folded Reload
; SI-NEXT: v_add_f32_e32 v21, 0x38000000, v21
; SI-NEXT: v_cvt_f16_f32_e32 v21, v21
-; SI-NEXT: s_waitcnt vmcnt(1)
-; SI-NEXT: v_cvt_f32_f16_e32 v27, v4
-; SI-NEXT: buffer_load_dword v4, off, s[0:3], s32 offset:580 ; 4-byte Folded Reload
-; SI-NEXT: v_or_b32_e32 v61, v21, v22
-; SI-NEXT: v_add_f32_e32 v27, 0x38000000, v27
-; SI-NEXT: v_cvt_f16_f32_e32 v27, v27
-; SI-NEXT: v_lshlrev_b32_e32 v27, 16, v27
-; SI-NEXT: s_waitcnt vmcnt(1)
-; SI-NEXT: v_cvt_f32_f16_e32 v22, v3
-; SI-NEXT: buffer_load_dword v3, off, s[0:3], s32 offset:604 ; 4-byte Folded Reload
+; SI-NEXT: v_or_b32_e32 v54, v21, v22
+; SI-NEXT: s_waitcnt vmcnt(0)
+; SI-NEXT: v_cvt_f32_f16_e32 v22, v7
+; SI-NEXT: buffer_load_dword v7, off, s[0:3], s32 offset:596 ; 4-byte Folded Reload
; SI-NEXT: v_add_f32_e32 v22, 0x38000000, v22
; SI-NEXT: v_cvt_f16_f32_e32 v22, v22
; SI-NEXT: v_lshlrev_b32_e32 v22, 16, v22
; SI-NEXT: s_waitcnt vmcnt(0)
-; SI-NEXT: v_cvt_f32_f16_e32 v23, v3
-; SI-NEXT: buffer_load_dword v3, off, s[0:3], s32 offset:600 ; 4-byte Folded Reload
+; SI-NEXT: v_cvt_f32_f16_e32 v23, v7
+; SI-NEXT: buffer_load_dword v7, off, s[0:3], s32 offset:592 ; 4-byte Folded Reload
; SI-NEXT: v_add_f32_e32 v23, 0x38000000, v23
; SI-NEXT: v_cvt_f16_f32_e32 v23, v23
-; SI-NEXT: v_or_b32_e32 v16, v23, v22
+; SI-NEXT: v_or_b32_e32 v36, v23, v22
; SI-NEXT: s_waitcnt vmcnt(0)
-; SI-NEXT: v_cvt_f32_f16_e32 v23, v3
-; SI-NEXT: buffer_load_dword v3, off, s[0:3], s32 offset:596 ; 4-byte Folded Reload
+; SI-NEXT: v_cvt_f32_f16_e32 v23, v7
+; SI-NEXT: buffer_load_dword v7, off, s[0:3], s32 offset:588 ; 4-byte Folded Reload
; SI-NEXT: v_add_f32_e32 v23, 0x38000000, v23
; SI-NEXT: v_cvt_f16_f32_e32 v23, v23
-; SI-NEXT: v_or_b32_e32 v23, v23, v25
+; SI-NEXT: v_or_b32_e32 v45, v23, v25
; SI-NEXT: s_waitcnt vmcnt(0)
-; SI-NEXT: v_cvt_f32_f16_e32 v24, v3
-; SI-NEXT: buffer_load_dword v3, off, s[0:3], s32 offset:592 ; 4-byte Folded Reload
+; SI-NEXT: v_cvt_f32_f16_e32 v24, v7
+; SI-NEXT: buffer_load_dword v7, off, s[0:3], s32 offset:584 ; 4-byte Folded Reload
; SI-NEXT: v_add_f32_e32 v24, 0x38000000, v24
; SI-NEXT: v_cvt_f16_f32_e32 v24, v24
; SI-NEXT: v_lshlrev_b32_e32 v24, 16, v24
; SI-NEXT: s_waitcnt vmcnt(0)
-; SI-NEXT: v_cvt_f32_f16_e32 v26, v3
-; SI-NEXT: buffer_load_dword v3, off, s[0:3], s32 offset:588 ; 4-byte Folded Reload
+; SI-NEXT: v_cvt_f32_f16_e32 v26, v7
+; SI-NEXT: buffer_load_dword v7, off, s[0:3], s32 offset:580 ; 4-byte Folded Reload
; SI-NEXT: v_add_f32_e32 v25, 0x38000000, v26
; SI-NEXT: v_cvt_f16_f32_e32 v25, v25
-; SI-NEXT: v_or_b32_e32 v48, v25, v24
-; SI-NEXT: v_lshlrev_b32_e32 v25, 16, v6
+; SI-NEXT: v_or_b32_e32 v13, v25, v24
+; SI-NEXT: v_lshlrev_b32_e32 v25, 16, v47
; SI-NEXT: s_waitcnt vmcnt(0)
-; SI-NEXT: v_cvt_f32_f16_e32 v26, v3
+; SI-NEXT: v_cvt_f32_f16_e32 v26, v7
+; SI-NEXT: buffer_load_dword v7, off, s[0:3], s32 offset:572 ; 4-byte Folded Reload
; SI-NEXT: v_add_f32_e32 v26, 0x38000000, v26
; SI-NEXT: v_cvt_f16_f32_e32 v26, v26
; SI-NEXT: v_or_b32_e32 v25, v26, v25
-; SI-NEXT: v_cvt_f32_f16_e32 v26, v4
-; SI-NEXT: buffer_load_dword v4, off, s[0:3], s32 offset:576 ; 4-byte Folded Reload
+; SI-NEXT: s_waitcnt vmcnt(0)
+; SI-NEXT: v_cvt_f32_f16_e32 v27, v7
+; SI-NEXT: buffer_load_dword v7, off, s[0:3], s32 offset:568 ; 4-byte Folded Reload
+; SI-NEXT: v_add_f32_e32 v27, 0x38000000, v27
+; SI-NEXT: v_cvt_f16_f32_e32 v27, v27
+; SI-NEXT: v_lshlrev_b32_e32 v27, 16, v27
+; SI-NEXT: s_waitcnt vmcnt(0)
+; SI-NEXT: v_cvt_f32_f16_e32 v26, v7
+; SI-NEXT: v_cvt_f32_f16_e32 v7, v62
; SI-NEXT: v_add_f32_e32 v26, 0x38000000, v26
+; SI-NEXT: v_add_f32_e32 v7, 0x38000000, v7
+; SI-NEXT: v_cvt_f16_f32_e32 v16, v7
+; SI-NEXT: buffer_load_dword v7, off, s[0:3], s32 offset:564 ; 4-byte Folded Reload
; SI-NEXT: v_cvt_f16_f32_e32 v26, v26
-; SI-NEXT: v_or_b32_e32 v53, v26, v27
+; SI-NEXT: v_or_b32_e32 v14, v26, v27
; SI-NEXT: v_mov_b32_e32 v26, v13
-; SI-NEXT: buffer_load_dword v13, off, s[0:3], s32 offset:572 ; 4-byte Folded Reload
-; SI-NEXT: v_lshlrev_b32_e32 v27, 16, v7
+; SI-NEXT: buffer_load_dword v13, off, s[0:3], s32 offset:552 ; 4-byte Folded Reload
+; SI-NEXT: v_lshlrev_b32_e32 v27, 16, v16
; SI-NEXT: s_waitcnt vmcnt(1)
-; SI-NEXT: v_cvt_f32_f16_e32 v28, v4
+; SI-NEXT: v_cvt_f32_f16_e32 v28, v7
+; SI-NEXT: buffer_load_dword v7, off, s[0:3], s32 offset:560 ; 4-byte Folded Reload
; SI-NEXT: v_add_f32_e32 v28, 0x38000000, v28
; SI-NEXT: v_cvt_f16_f32_e32 v28, v28
-; SI-NEXT: v_or_b32_e32 v62, v28, v27
+; SI-NEXT: v_or_b32_e32 v58, v28, v27
; SI-NEXT: s_waitcnt vmcnt(0)
-; SI-NEXT: v_cvt_f32_f16_e32 v29, v13
-; SI-NEXT: buffer_load_dword v13, off, s[0:3], s32 offset:568 ; 4-byte Folded Reload
-; SI-NEXT: v_mov_b32_e32 v27, v34
-; SI-NEXT: v_lshlrev_b32_e32 v34, 16, v8
+; SI-NEXT: v_cvt_f32_f16_e32 v29, v7
+; SI-NEXT: buffer_load_dword v7, off, s[0:3], s32 offset:556 ; 4-byte Folded Reload
; SI-NEXT: v_add_f32_e32 v29, 0x38000000, v29
; SI-NEXT: v_cvt_f16_f32_e32 v29, v29
; SI-NEXT: v_lshlrev_b32_e32 v28, 16, v29
-; SI-NEXT: v_cvt_f32_f16_e32 v29, v63
+; SI-NEXT: v_cvt_f32_f16_e32 v29, v13
+; SI-NEXT: buffer_load_dword v13, off, s[0:3], s32 offset:548 ; 4-byte Folded Reload
; SI-NEXT: v_add_f32_e32 v29, 0x38000000, v29
; SI-NEXT: v_cvt_f16_f32_e32 v29, v29
-; SI-NEXT: v_or_b32_e32 v59, v29, v34
+; SI-NEXT: v_or_b32_e32 v56, v29, v34
; SI-NEXT: v_add_f32_e32 v34, 0x38000000, v35
-; SI-NEXT: buffer_load_dword v35, off, s[0:3], s32 offset:564 ; 4-byte Folded Reload
-; SI-NEXT: v_mov_b32_e32 v29, v36
-; SI-NEXT: v_cvt_f32_f16_e32 v36, v51
-; SI-NEXT: buffer_load_dword v51, off, s[0:3], s32 offset:508 ; 4-byte Folded Reload
; SI-NEXT: v_cvt_f16_f32_e32 v34, v34
-; SI-NEXT: v_add_f32_e32 v36, 0x38000000, v36
-; SI-NEXT: v_cvt_f16_f32_e32 v36, v36
-; SI-NEXT: v_lshlrev_b32_e32 v36, 16, v36
-; SI-NEXT: s_waitcnt vmcnt(2)
-; SI-NEXT: v_cvt_f32_f16_e32 v30, v13
-; SI-NEXT: buffer_load_dword v13, off, s[0:3], s32 offset:448 ; 4-byte Folded Reload
+; SI-NEXT: v_mov_b32_e32 v29, v16
+; SI-NEXT: v_mov_b32_e32 v62, v29
+; SI-NEXT: s_waitcnt vmcnt(1)
+; SI-NEXT: v_cvt_f32_f16_e32 v30, v7
; SI-NEXT: v_add_f32_e32 v30, 0x38000000, v30
; SI-NEXT: v_cvt_f16_f32_e32 v30, v30
-; SI-NEXT: v_or_b32_e32 v3, v30, v28
-; SI-NEXT: v_cvt_f32_f16_e32 v30, v10
-; SI-NEXT: v_cvt_f32_f16_e32 v10, v32
-; SI-NEXT: v_cvt_f32_f16_e32 v32, v44
+; SI-NEXT: v_or_b32_e32 v7, v30, v28
+; SI-NEXT: v_cvt_f32_f16_e32 v30, v52
+; SI-NEXT: s_waitcnt vmcnt(0)
+; SI-NEXT: v_cvt_f32_f16_e32 v35, v13
+; SI-NEXT: v_mov_b32_e32 v28, v36
+; SI-NEXT: v_cvt_f32_f16_e32 v36, v48
; SI-NEXT: v_add_f32_e32 v30, 0x38000000, v30
; SI-NEXT: v_cvt_f16_f32_e32 v30, v30
-; SI-NEXT: v_add_f32_e32 v10, 0x38000000, v10
-; SI-NEXT: v_cvt_f16_f32_e32 v10, v10
-; SI-NEXT: v_add_f32_e32 v32, 0x38000000, v32
-; SI-NEXT: v_lshlrev_b32_e32 v30, 16, v30
-; SI-NEXT: s_waitcnt vmcnt(2)
-; SI-NEXT: v_cvt_f32_f16_e32 v35, v35
-; SI-NEXT: v_or_b32_e32 v4, v34, v30
-; SI-NEXT: v_lshlrev_b32_e32 v34, 16, v9
-; SI-NEXT: v_cvt_f16_f32_e32 v44, v32
; SI-NEXT: v_add_f32_e32 v35, 0x38000000, v35
; SI-NEXT: v_cvt_f16_f32_e32 v35, v35
-; SI-NEXT: s_waitcnt vmcnt(1)
-; SI-NEXT: v_cvt_f32_f16_e32 v51, v51
-; SI-NEXT: v_mov_b32_e32 v30, v10
-; SI-NEXT: v_mov_b32_e32 v32, v30
+; SI-NEXT: v_add_f32_e32 v36, 0x38000000, v36
+; SI-NEXT: v_lshlrev_b32_e32 v30, 16, v30
+; SI-NEXT: v_or_b32_e32 v41, v34, v30
+; SI-NEXT: v_lshlrev_b32_e32 v34, 16, v9
; SI-NEXT: v_or_b32_e32 v34, v35, v34
-; SI-NEXT: v_cvt_f32_f16_e32 v35, v46
-; SI-NEXT: v_add_f32_e32 v51, 0x38000000, v51
-; SI-NEXT: v_cvt_f16_f32_e32 v51, v51
+; SI-NEXT: v_cvt_f32_f16_e32 v35, v61
+; SI-NEXT: v_cvt_f16_f32_e32 v36, v36
+; SI-NEXT: buffer_load_dword v13, off, s[0:3], s32 offset:544 ; 4-byte Folded Reload
; SI-NEXT: v_add_f32_e32 v35, 0x38000000, v35
; SI-NEXT: v_cvt_f16_f32_e32 v35, v35
-; SI-NEXT: v_or_b32_e32 v22, v35, v36
+; SI-NEXT: v_lshlrev_b32_e32 v36, 16, v36
+; SI-NEXT: v_or_b32_e32 v53, v35, v36
; SI-NEXT: v_lshlrev_b32_e32 v36, 16, v10
-; SI-NEXT: v_mov_b32_e32 v35, v16
-; SI-NEXT: s_waitcnt vmcnt(0)
-; SI-NEXT: v_cvt_f32_f16_e32 v37, v13
-; SI-NEXT: buffer_load_dword v13, off, s[0:3], s32 offset:444 ; 4-byte Folded Reload
+; SI-NEXT: v_or_b32_e32 v63, v37, v36
+; SI-NEXT: v_mov_b32_e32 v36, v7
+; SI-NEXT: buffer_load_dword v7, off, s[0:3], s32 offset:440 ; 4-byte Folded Reload
+; SI-NEXT: v_lshlrev_b32_e32 v37, 16, v38
+; SI-NEXT: v_or_b32_e32 v40, v39, v37
+; SI-NEXT: v_mov_b32_e32 v38, v14
+; SI-NEXT: v_alignbit_b32 v27, v63, v53, 16
+; SI-NEXT: v_alignbit_b32 v30, v63, v53, 8
+; SI-NEXT: v_bfe_u32 v35, v2, 8, 8
+; SI-NEXT: s_waitcnt vmcnt(0)
+; SI-NEXT: v_cvt_f32_f16_e32 v37, v7
+; SI-NEXT: buffer_load_dword v7, off, s[0:3], s32 offset:576 ; 4-byte Folded Reload
; SI-NEXT: v_add_f32_e32 v37, 0x38000000, v37
; SI-NEXT: v_cvt_f16_f32_e32 v37, v37
-; SI-NEXT: v_or_b32_e32 v24, v37, v36
-; SI-NEXT: v_lshlrev_b32_e32 v37, 16, v38
-; SI-NEXT: v_or_b32_e32 v42, v39, v37
-; SI-NEXT: v_cvt_f32_f16_e32 v39, v45
-; SI-NEXT: v_mov_b32_e32 v36, v48
+; SI-NEXT: s_waitcnt vmcnt(0)
+; SI-NEXT: v_cvt_f32_f16_e32 v39, v7
; SI-NEXT: v_add_f32_e32 v39, 0x38000000, v39
; SI-NEXT: v_cvt_f16_f32_e32 v48, v39
; SI-NEXT: v_lshlrev_b32_e32 v39, 16, v31
-; SI-NEXT: s_waitcnt vmcnt(0)
-; SI-NEXT: v_cvt_f32_f16_e32 v37, v13
-; SI-NEXT: v_add_f32_e32 v37, 0x38000000, v37
-; SI-NEXT: v_cvt_f16_f32_e32 v37, v37
-; SI-NEXT: v_or_b32_e32 v60, v37, v39
+; SI-NEXT: v_or_b32_e32 v6, v37, v39
+; SI-NEXT: v_mov_b32_e32 v39, v17
; SI-NEXT: v_lshlrev_b32_e32 v37, 16, v48
; SI-NEXT: v_add_f32_e32 v48, 0x38000000, v49
-; SI-NEXT: buffer_load_dword v49, off, s[0:3], s32 offset:512 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v49, off, s[0:3], s32 offset:496 ; 4-byte Folded Reload
; SI-NEXT: v_cvt_f16_f32_e32 v48, v48
-; SI-NEXT: v_alignbit_b32 v39, v54, v29, 16
-; SI-NEXT: v_or_b32_e32 v43, v48, v37
-; SI-NEXT: buffer_load_dword v37, off, s[0:3], s32 offset:560 ; 4-byte Folded Reload
-; SI-NEXT: v_lshlrev_b32_e32 v48, 16, v44
-; SI-NEXT: s_waitcnt vmcnt(1)
-; SI-NEXT: v_cvt_f32_f16_e32 v49, v49
-; SI-NEXT: v_add_f32_e32 v49, 0x38000000, v49
-; SI-NEXT: v_cvt_f16_f32_e32 v49, v49
-; SI-NEXT: s_waitcnt vmcnt(0)
-; SI-NEXT: v_cvt_f32_f16_e32 v37, v37
-; SI-NEXT: v_or_b32_e32 v14, v49, v48
-; SI-NEXT: buffer_load_dword v48, off, s[0:3], s32 offset:556 ; 4-byte Folded Reload
+; SI-NEXT: v_alignbit_b32 v24, v39, v1, 24
+; SI-NEXT: v_or_b32_e32 v21, v48, v37
+; SI-NEXT: v_lshlrev_b32_e32 v48, 16, v32
+; SI-NEXT: v_cvt_f32_f16_e32 v37, v13
; SI-NEXT: v_add_f32_e32 v37, 0x38000000, v37
; SI-NEXT: v_cvt_f16_f32_e32 v37, v37
-; SI-NEXT: buffer_load_dword v49, off, s[0:3], s32 offset:504 ; 4-byte Folded Reload
-; SI-NEXT: v_alignbit_b32 v28, v14, v43, 8
; SI-NEXT: v_lshlrev_b32_e32 v37, 16, v37
+; SI-NEXT: s_waitcnt vmcnt(0)
+; SI-NEXT: v_cvt_f32_f16_e32 v49, v49
+; SI-NEXT: v_add_f32_e32 v49, 0x38000000, v49
+; SI-NEXT: v_cvt_f16_f32_e32 v49, v49
+; SI-NEXT: v_or_b32_e32 v60, v49, v48
+; SI-NEXT: buffer_load_dword v48, off, s[0:3], s32 offset:540 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v49, off, s[0:3], s32 offset:488 ; 4-byte Folded Reload
; SI-NEXT: s_waitcnt vmcnt(1)
; SI-NEXT: v_cvt_f32_f16_e32 v48, v48
; SI-NEXT: s_waitcnt vmcnt(0)
@@ -196930,20 +197033,20 @@ define inreg <128 x i8> @bitcast_v64f16_to_v128i8_scalar(<64 x half> inreg %a, i
; SI-NEXT: v_cvt_f16_f32_e32 v48, v48
; SI-NEXT: v_add_f32_e32 v49, 0x38000000, v49
; SI-NEXT: v_cvt_f16_f32_e32 v49, v49
-; SI-NEXT: v_or_b32_e32 v19, v48, v37
-; SI-NEXT: buffer_load_dword v48, off, s[0:3], s32 offset:552 ; 4-byte Folded Reload
+; SI-NEXT: v_or_b32_e32 v23, v48, v37
+; SI-NEXT: buffer_load_dword v48, off, s[0:3], s32 offset:532 ; 4-byte Folded Reload
; SI-NEXT: v_lshlrev_b32_e32 v37, 16, v33
-; SI-NEXT: v_or_b32_e32 v47, v49, v37
+; SI-NEXT: v_or_b32_e32 v14, v49, v37
; SI-NEXT: v_lshlrev_b32_e32 v49, 16, v15
; SI-NEXT: s_waitcnt vmcnt(0)
; SI-NEXT: v_cvt_f32_f16_e32 v48, v48
; SI-NEXT: v_add_f32_e32 v48, 0x38000000, v48
; SI-NEXT: v_cvt_f16_f32_e32 v48, v48
; SI-NEXT: v_lshlrev_b32_e32 v37, 16, v48
-; SI-NEXT: v_or_b32_e32 v21, v50, v37
-; SI-NEXT: buffer_load_dword v37, off, s[0:3], s32 offset:540 ; 4-byte Folded Reload
-; SI-NEXT: buffer_load_dword v48, off, s[0:3], s32 offset:536 ; 4-byte Folded Reload
-; SI-NEXT: buffer_load_dword v50, off, s[0:3], s32 offset:532 ; 4-byte Folded Reload
+; SI-NEXT: v_or_b32_e32 v22, v50, v37
+; SI-NEXT: buffer_load_dword v37, off, s[0:3], s32 offset:524 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v48, off, s[0:3], s32 offset:520 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v50, off, s[0:3], s32 offset:516 ; 4-byte Folded Reload
; SI-NEXT: s_waitcnt vmcnt(2)
; SI-NEXT: v_cvt_f32_f16_e32 v37, v37
; SI-NEXT: s_waitcnt vmcnt(1)
@@ -196953,26 +197056,27 @@ define inreg <128 x i8> @bitcast_v64f16_to_v128i8_scalar(<64 x half> inreg %a, i
; SI-NEXT: v_add_f32_e32 v48, 0x38000000, v48
; SI-NEXT: v_cvt_f16_f32_e32 v48, v48
; SI-NEXT: v_or_b32_e32 v16, v37, v49
-; SI-NEXT: buffer_load_dword v49, off, s[0:3], s32 offset:528 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v49, off, s[0:3], s32 offset:512 ; 4-byte Folded Reload
; SI-NEXT: s_waitcnt vmcnt(1)
; SI-NEXT: v_cvt_f32_f16_e32 v50, v50
; SI-NEXT: v_lshlrev_b32_e32 v37, 16, v48
; SI-NEXT: v_add_f32_e32 v48, 0x38000000, v50
; SI-NEXT: v_cvt_f16_f32_e32 v48, v48
-; SI-NEXT: v_alignbit_b32 v50, v54, v29, 24
-; SI-NEXT: v_or_b32_e32 v57, v48, v37
-; SI-NEXT: buffer_load_dword v37, off, s[0:3], s32 offset:520 ; 4-byte Folded Reload
+; SI-NEXT: v_mov_b32_e32 v50, v55
+; SI-NEXT: v_alignbit_b32 v43, v50, v59, 24
+; SI-NEXT: v_alignbit_b32 v42, v50, v59, 16
+; SI-NEXT: v_or_b32_e32 v17, v48, v37
+; SI-NEXT: buffer_load_dword v37, off, s[0:3], s32 offset:504 ; 4-byte Folded Reload
; SI-NEXT: v_lshlrev_b32_e32 v48, 16, v18
+; SI-NEXT: v_alignbit_b32 v57, v50, v59, 8
; SI-NEXT: s_waitcnt vmcnt(1)
; SI-NEXT: v_cvt_f32_f16_e32 v49, v49
; SI-NEXT: v_add_f32_e32 v49, 0x38000000, v49
; SI-NEXT: v_cvt_f16_f32_e32 v49, v49
-; SI-NEXT: v_or_b32_e32 v17, v49, v48
-; SI-NEXT: buffer_load_dword v48, off, s[0:3], s32 offset:516 ; 4-byte Folded Reload
+; SI-NEXT: v_or_b32_e32 v46, v49, v48
+; SI-NEXT: buffer_load_dword v48, off, s[0:3], s32 offset:500 ; 4-byte Folded Reload
; SI-NEXT: s_waitcnt vmcnt(1)
; SI-NEXT: v_cvt_f32_f16_e32 v37, v37
-; SI-NEXT: v_mov_b32_e32 v49, v53
-; SI-NEXT: v_alignbit_b32 v53, v54, v29, 8
; SI-NEXT: v_add_f32_e32 v37, 0x38000000, v37
; SI-NEXT: v_cvt_f16_f32_e32 v37, v37
; SI-NEXT: v_lshlrev_b32_e32 v37, 16, v37
@@ -196983,573 +197087,575 @@ define inreg <128 x i8> @bitcast_v64f16_to_v128i8_scalar(<64 x half> inreg %a, i
; SI-NEXT: v_or_b32_e32 v13, v48, v37
; SI-NEXT: v_lshlrev_b32_e32 v37, 16, v20
; SI-NEXT: v_or_b32_e32 v55, v51, v37
-; SI-NEXT: v_alignbit_b32 v10, v55, v13, 16
-; SI-NEXT: buffer_store_dword v10, off, s[0:3], s32 offset:368 ; 4-byte Folded Spill
+; SI-NEXT: v_alignbit_b32 v7, v55, v13, 24
+; SI-NEXT: buffer_store_dword v7, off, s[0:3], s32 offset:364 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_alignbit_b32 v10, v55, v13, 8
-; SI-NEXT: buffer_store_dword v10, off, s[0:3], s32 offset:372 ; 4-byte Folded Spill
+; SI-NEXT: v_alignbit_b32 v7, v55, v13, 16
+; SI-NEXT: buffer_store_dword v7, off, s[0:3], s32 offset:368 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_alignbit_b32 v10, v17, v57, 24
-; SI-NEXT: buffer_store_dword v10, off, s[0:3], s32 offset:348 ; 4-byte Folded Spill
+; SI-NEXT: v_alignbit_b32 v7, v55, v13, 8
+; SI-NEXT: buffer_store_dword v7, off, s[0:3], s32 offset:372 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_alignbit_b32 v10, v17, v57, 16
-; SI-NEXT: buffer_store_dword v10, off, s[0:3], s32 offset:360 ; 4-byte Folded Spill
+; SI-NEXT: v_alignbit_b32 v7, v46, v17, 24
+; SI-NEXT: buffer_store_dword v7, off, s[0:3], s32 offset:344 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_alignbit_b32 v10, v17, v57, 8
-; SI-NEXT: buffer_store_dword v10, off, s[0:3], s32 offset:364 ; 4-byte Folded Spill
+; SI-NEXT: v_alignbit_b32 v7, v46, v17, 16
+; SI-NEXT: buffer_store_dword v7, off, s[0:3], s32 offset:352 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_alignbit_b32 v10, v16, v21, 24
-; SI-NEXT: buffer_store_dword v10, off, s[0:3], s32 offset:332 ; 4-byte Folded Spill
+; SI-NEXT: v_alignbit_b32 v7, v46, v17, 8
+; SI-NEXT: buffer_store_dword v7, off, s[0:3], s32 offset:360 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_alignbit_b32 v10, v16, v21, 16
-; SI-NEXT: buffer_store_dword v10, off, s[0:3], s32 offset:340 ; 4-byte Folded Spill
+; SI-NEXT: v_alignbit_b32 v7, v16, v22, 24
+; SI-NEXT: buffer_store_dword v7, off, s[0:3], s32 offset:328 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_alignbit_b32 v10, v16, v21, 8
-; SI-NEXT: buffer_store_dword v10, off, s[0:3], s32 offset:344 ; 4-byte Folded Spill
+; SI-NEXT: v_alignbit_b32 v7, v16, v22, 16
+; SI-NEXT: buffer_store_dword v7, off, s[0:3], s32 offset:336 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_alignbit_b32 v10, v47, v19, 24
-; SI-NEXT: buffer_store_dword v10, off, s[0:3], s32 offset:316 ; 4-byte Folded Spill
+; SI-NEXT: v_alignbit_b32 v7, v16, v22, 8
+; SI-NEXT: buffer_store_dword v7, off, s[0:3], s32 offset:340 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_alignbit_b32 v10, v47, v19, 16
-; SI-NEXT: buffer_store_dword v10, off, s[0:3], s32 offset:320 ; 4-byte Folded Spill
+; SI-NEXT: v_alignbit_b32 v7, v14, v23, 24
+; SI-NEXT: buffer_store_dword v7, off, s[0:3], s32 offset:312 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_alignbit_b32 v10, v47, v19, 8
-; SI-NEXT: buffer_store_dword v10, off, s[0:3], s32 offset:324 ; 4-byte Folded Spill
+; SI-NEXT: v_alignbit_b32 v7, v14, v23, 16
+; SI-NEXT: buffer_store_dword v7, off, s[0:3], s32 offset:316 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_alignbit_b32 v10, v14, v43, 24
-; SI-NEXT: buffer_store_dword v10, off, s[0:3], s32 offset:296 ; 4-byte Folded Spill
+; SI-NEXT: v_alignbit_b32 v7, v14, v23, 8
+; SI-NEXT: buffer_store_dword v7, off, s[0:3], s32 offset:320 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_alignbit_b32 v10, v14, v43, 16
-; SI-NEXT: buffer_store_dword v10, off, s[0:3], s32 offset:300 ; 4-byte Folded Spill
+; SI-NEXT: v_alignbit_b32 v7, v60, v21, 24
+; SI-NEXT: buffer_store_dword v7, off, s[0:3], s32 offset:292 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_alignbit_b32 v10, v60, v42, 24
-; SI-NEXT: buffer_store_dword v10, off, s[0:3], s32 offset:276 ; 4-byte Folded Spill
+; SI-NEXT: v_alignbit_b32 v7, v60, v21, 16
+; SI-NEXT: buffer_store_dword v7, off, s[0:3], s32 offset:296 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_alignbit_b32 v10, v60, v42, 16
-; SI-NEXT: buffer_store_dword v10, off, s[0:3], s32 offset:284 ; 4-byte Folded Spill
+; SI-NEXT: v_alignbit_b32 v7, v60, v21, 8
+; SI-NEXT: buffer_store_dword v7, off, s[0:3], s32 offset:304 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_alignbit_b32 v10, v60, v42, 8
-; SI-NEXT: buffer_store_dword v10, off, s[0:3], s32 offset:292 ; 4-byte Folded Spill
+; SI-NEXT: v_alignbit_b32 v7, v6, v40, 24
+; SI-NEXT: buffer_store_dword v7, off, s[0:3], s32 offset:272 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_alignbit_b32 v10, v24, v22, 24
-; SI-NEXT: buffer_store_dword v10, off, s[0:3], s32 offset:260 ; 4-byte Folded Spill
+; SI-NEXT: v_alignbit_b32 v7, v6, v40, 16
+; SI-NEXT: buffer_store_dword v7, off, s[0:3], s32 offset:280 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_alignbit_b32 v10, v24, v22, 16
-; SI-NEXT: buffer_store_dword v10, off, s[0:3], s32 offset:264 ; 4-byte Folded Spill
+; SI-NEXT: v_alignbit_b32 v7, v6, v40, 8
+; SI-NEXT: buffer_store_dword v7, off, s[0:3], s32 offset:288 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_alignbit_b32 v10, v24, v22, 8
-; SI-NEXT: buffer_store_dword v10, off, s[0:3], s32 offset:272 ; 4-byte Folded Spill
+; SI-NEXT: v_alignbit_b32 v7, v63, v53, 24
+; SI-NEXT: buffer_store_dword v7, off, s[0:3], s32 offset:260 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_alignbit_b32 v10, v34, v4, 24
-; SI-NEXT: buffer_store_dword v10, off, s[0:3], s32 offset:240 ; 4-byte Folded Spill
+; SI-NEXT: v_alignbit_b32 v7, v34, v41, 24
+; SI-NEXT: buffer_store_dword v7, off, s[0:3], s32 offset:240 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_alignbit_b32 v10, v34, v4, 16
-; SI-NEXT: buffer_store_dword v10, off, s[0:3], s32 offset:248 ; 4-byte Folded Spill
+; SI-NEXT: v_alignbit_b32 v7, v34, v41, 16
+; SI-NEXT: buffer_store_dword v7, off, s[0:3], s32 offset:248 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_alignbit_b32 v10, v34, v4, 8
-; SI-NEXT: buffer_store_dword v10, off, s[0:3], s32 offset:252 ; 4-byte Folded Spill
+; SI-NEXT: v_alignbit_b32 v7, v34, v41, 8
+; SI-NEXT: buffer_store_dword v7, off, s[0:3], s32 offset:252 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_alignbit_b32 v10, v59, v3, 24
-; SI-NEXT: buffer_store_dword v10, off, s[0:3], s32 offset:224 ; 4-byte Folded Spill
+; SI-NEXT: v_alignbit_b32 v7, v56, v36, 24
+; SI-NEXT: buffer_store_dword v7, off, s[0:3], s32 offset:224 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_alignbit_b32 v10, v59, v3, 16
-; SI-NEXT: buffer_store_dword v10, off, s[0:3], s32 offset:228 ; 4-byte Folded Spill
+; SI-NEXT: v_alignbit_b32 v7, v56, v36, 16
+; SI-NEXT: buffer_store_dword v7, off, s[0:3], s32 offset:228 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_alignbit_b32 v10, v59, v3, 8
-; SI-NEXT: buffer_store_dword v10, off, s[0:3], s32 offset:232 ; 4-byte Folded Spill
+; SI-NEXT: v_alignbit_b32 v7, v56, v36, 8
+; SI-NEXT: buffer_store_dword v7, off, s[0:3], s32 offset:232 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_alignbit_b32 v10, v62, v49, 24
-; SI-NEXT: buffer_store_dword v10, off, s[0:3], s32 offset:204 ; 4-byte Folded Spill
+; SI-NEXT: v_alignbit_b32 v7, v58, v38, 24
+; SI-NEXT: buffer_store_dword v7, off, s[0:3], s32 offset:204 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_alignbit_b32 v10, v62, v49, 16
-; SI-NEXT: buffer_store_dword v10, off, s[0:3], s32 offset:208 ; 4-byte Folded Spill
+; SI-NEXT: v_alignbit_b32 v7, v58, v38, 16
+; SI-NEXT: buffer_store_dword v7, off, s[0:3], s32 offset:208 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_alignbit_b32 v10, v62, v49, 8
-; SI-NEXT: buffer_store_dword v10, off, s[0:3], s32 offset:216 ; 4-byte Folded Spill
+; SI-NEXT: v_alignbit_b32 v7, v58, v38, 8
+; SI-NEXT: buffer_store_dword v7, off, s[0:3], s32 offset:216 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_alignbit_b32 v10, v25, v36, 24
-; SI-NEXT: buffer_store_dword v10, off, s[0:3], s32 offset:188 ; 4-byte Folded Spill
+; SI-NEXT: v_alignbit_b32 v7, v25, v26, 24
+; SI-NEXT: buffer_store_dword v7, off, s[0:3], s32 offset:188 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_alignbit_b32 v10, v25, v36, 16
-; SI-NEXT: buffer_store_dword v10, off, s[0:3], s32 offset:192 ; 4-byte Folded Spill
+; SI-NEXT: v_alignbit_b32 v7, v25, v26, 16
+; SI-NEXT: buffer_store_dword v7, off, s[0:3], s32 offset:192 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v26, off, s[0:3], s32 offset:264 ; 4-byte Folded Spill
+; SI-NEXT: s_waitcnt expcnt(1)
+; SI-NEXT: v_alignbit_b32 v7, v25, v26, 8
+; SI-NEXT: buffer_store_dword v7, off, s[0:3], s32 offset:200 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_alignbit_b32 v10, v25, v36, 8
-; SI-NEXT: buffer_store_dword v10, off, s[0:3], s32 offset:200 ; 4-byte Folded Spill
+; SI-NEXT: v_alignbit_b32 v7, v45, v28, 24
+; SI-NEXT: buffer_store_dword v7, off, s[0:3], s32 offset:172 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_alignbit_b32 v10, v23, v35, 24
-; SI-NEXT: buffer_store_dword v10, off, s[0:3], s32 offset:172 ; 4-byte Folded Spill
+; SI-NEXT: v_alignbit_b32 v7, v45, v28, 16
+; SI-NEXT: buffer_store_dword v7, off, s[0:3], s32 offset:180 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v28, off, s[0:3], s32 offset:268 ; 4-byte Folded Spill
+; SI-NEXT: s_waitcnt expcnt(1)
+; SI-NEXT: v_alignbit_b32 v7, v45, v28, 8
+; SI-NEXT: buffer_store_dword v7, off, s[0:3], s32 offset:184 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_alignbit_b32 v10, v23, v35, 16
-; SI-NEXT: buffer_store_dword v10, off, s[0:3], s32 offset:180 ; 4-byte Folded Spill
+; SI-NEXT: v_alignbit_b32 v7, v54, v19, 24
+; SI-NEXT: buffer_store_dword v7, off, s[0:3], s32 offset:160 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_alignbit_b32 v10, v23, v35, 8
-; SI-NEXT: buffer_store_dword v10, off, s[0:3], s32 offset:184 ; 4-byte Folded Spill
+; SI-NEXT: v_alignbit_b32 v7, v54, v19, 16
+; SI-NEXT: buffer_store_dword v7, off, s[0:3], s32 offset:164 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_alignbit_b32 v10, v61, v11, 24
-; SI-NEXT: buffer_store_dword v10, off, s[0:3], s32 offset:160 ; 4-byte Folded Spill
+; SI-NEXT: v_alignbit_b32 v7, v54, v19, 8
+; SI-NEXT: buffer_store_dword v7, off, s[0:3], s32 offset:168 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_alignbit_b32 v10, v61, v11, 16
-; SI-NEXT: buffer_store_dword v10, off, s[0:3], s32 offset:164 ; 4-byte Folded Spill
-; SI-NEXT: buffer_store_dword v11, off, s[0:3], s32 offset:352 ; 4-byte Folded Spill
-; SI-NEXT: s_waitcnt expcnt(1)
-; SI-NEXT: v_alignbit_b32 v10, v61, v11, 8
-; SI-NEXT: buffer_store_dword v10, off, s[0:3], s32 offset:168 ; 4-byte Folded Spill
+; SI-NEXT: v_alignbit_b32 v7, v39, v1, 16
+; SI-NEXT: buffer_store_dword v7, off, s[0:3], s32 offset:152 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_alignbit_b32 v10, v58, v52, 16
-; SI-NEXT: buffer_store_dword v10, off, s[0:3], s32 offset:152 ; 4-byte Folded Spill
-; SI-NEXT: buffer_store_dword v52, off, s[0:3], s32 offset:308 ; 4-byte Folded Spill
-; SI-NEXT: s_waitcnt expcnt(1)
-; SI-NEXT: v_alignbit_b32 v10, v58, v52, 8
-; SI-NEXT: buffer_store_dword v10, off, s[0:3], s32 offset:156 ; 4-byte Folded Spill
+; SI-NEXT: v_alignbit_b32 v7, v39, v1, 8
+; SI-NEXT: buffer_store_dword v7, off, s[0:3], s32 offset:156 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_mov_b32_e32 v10, v56
-; SI-NEXT: v_alignbit_b32 v11, v12, v10, 24
-; SI-NEXT: v_alignbit_b32 v56, v12, v10, 16
-; SI-NEXT: buffer_store_dword v10, off, s[0:3], s32 offset:384 ; 4-byte Folded Spill
-; SI-NEXT: v_alignbit_b32 v38, v12, v10, 8
+; SI-NEXT: v_lshrrev_b32_e32 v7, 8, v55
+; SI-NEXT: buffer_store_dword v7, off, s[0:3], s32 offset:428 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_lshrrev_b32_e32 v10, 8, v55
-; SI-NEXT: buffer_store_dword v10, off, s[0:3], s32 offset:432 ; 4-byte Folded Spill
+; SI-NEXT: v_lshrrev_b32_e32 v7, 8, v46
+; SI-NEXT: buffer_store_dword v7, off, s[0:3], s32 offset:416 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_lshrrev_b32_e32 v10, 8, v17
-; SI-NEXT: buffer_store_dword v10, off, s[0:3], s32 offset:424 ; 4-byte Folded Spill
+; SI-NEXT: v_lshrrev_b32_e32 v7, 8, v16
+; SI-NEXT: buffer_store_dword v7, off, s[0:3], s32 offset:408 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_lshrrev_b32_e32 v10, 8, v16
-; SI-NEXT: buffer_store_dword v10, off, s[0:3], s32 offset:412 ; 4-byte Folded Spill
+; SI-NEXT: v_lshrrev_b32_e32 v7, 8, v14
+; SI-NEXT: buffer_store_dword v7, off, s[0:3], s32 offset:400 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_lshrrev_b32_e32 v10, 8, v47
-; SI-NEXT: buffer_store_dword v10, off, s[0:3], s32 offset:404 ; 4-byte Folded Spill
+; SI-NEXT: v_lshrrev_b32_e32 v7, 8, v60
+; SI-NEXT: buffer_store_dword v7, off, s[0:3], s32 offset:392 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_lshrrev_b32_e32 v10, 8, v14
-; SI-NEXT: buffer_store_dword v10, off, s[0:3], s32 offset:396 ; 4-byte Folded Spill
+; SI-NEXT: v_lshrrev_b32_e32 v7, 8, v6
+; SI-NEXT: buffer_store_dword v7, off, s[0:3], s32 offset:384 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_lshrrev_b32_e32 v10, 8, v60
-; SI-NEXT: buffer_store_dword v10, off, s[0:3], s32 offset:388 ; 4-byte Folded Spill
+; SI-NEXT: v_lshrrev_b32_e32 v7, 8, v63
+; SI-NEXT: buffer_store_dword v7, off, s[0:3], s32 offset:376 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_lshrrev_b32_e32 v10, 8, v24
-; SI-NEXT: buffer_store_dword v10, off, s[0:3], s32 offset:376 ; 4-byte Folded Spill
+; SI-NEXT: v_lshrrev_b32_e32 v7, 8, v34
+; SI-NEXT: buffer_store_dword v7, off, s[0:3], s32 offset:356 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_lshrrev_b32_e32 v10, 8, v59
-; SI-NEXT: buffer_store_dword v10, off, s[0:3], s32 offset:328 ; 4-byte Folded Spill
+; SI-NEXT: v_lshrrev_b32_e32 v7, 8, v56
+; SI-NEXT: buffer_store_dword v7, off, s[0:3], s32 offset:324 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_lshrrev_b32_e32 v10, 8, v62
-; SI-NEXT: buffer_store_dword v10, off, s[0:3], s32 offset:304 ; 4-byte Folded Spill
+; SI-NEXT: v_lshrrev_b32_e32 v7, 8, v58
+; SI-NEXT: buffer_store_dword v7, off, s[0:3], s32 offset:300 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_lshrrev_b32_e32 v10, 8, v25
-; SI-NEXT: buffer_store_dword v10, off, s[0:3], s32 offset:280 ; 4-byte Folded Spill
+; SI-NEXT: v_lshrrev_b32_e32 v7, 8, v25
+; SI-NEXT: buffer_store_dword v7, off, s[0:3], s32 offset:276 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_lshrrev_b32_e32 v10, 8, v23
-; SI-NEXT: buffer_store_dword v10, off, s[0:3], s32 offset:256 ; 4-byte Folded Spill
+; SI-NEXT: v_lshrrev_b32_e32 v7, 8, v45
+; SI-NEXT: buffer_store_dword v7, off, s[0:3], s32 offset:256 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_lshrrev_b32_e32 v10, 8, v61
-; SI-NEXT: buffer_store_dword v10, off, s[0:3], s32 offset:236 ; 4-byte Folded Spill
+; SI-NEXT: v_lshrrev_b32_e32 v7, 8, v54
+; SI-NEXT: buffer_store_dword v7, off, s[0:3], s32 offset:236 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_lshrrev_b32_e32 v10, 8, v58
-; SI-NEXT: buffer_store_dword v10, off, s[0:3], s32 offset:212 ; 4-byte Folded Spill
+; SI-NEXT: v_lshrrev_b32_e32 v7, 8, v39
+; SI-NEXT: buffer_store_dword v7, off, s[0:3], s32 offset:212 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_lshrrev_b32_e32 v10, 8, v54
-; SI-NEXT: buffer_store_dword v10, off, s[0:3], s32 offset:196 ; 4-byte Folded Spill
+; SI-NEXT: v_lshrrev_b32_e32 v7, 8, v50
+; SI-NEXT: buffer_store_dword v7, off, s[0:3], s32 offset:196 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_lshrrev_b32_e32 v10, 8, v12
-; SI-NEXT: buffer_store_dword v10, off, s[0:3], s32 offset:176 ; 4-byte Folded Spill
+; SI-NEXT: v_lshrrev_b32_e32 v7, 8, v12
+; SI-NEXT: buffer_store_dword v7, off, s[0:3], s32 offset:176 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_bfe_u32 v10, v20, 8, 8
-; SI-NEXT: buffer_store_dword v10, off, s[0:3], s32 offset:440 ; 4-byte Folded Spill
+; SI-NEXT: v_bfe_u32 v7, v20, 8, 8
+; SI-NEXT: buffer_store_dword v7, off, s[0:3], s32 offset:436 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_bfe_u32 v10, v18, 8, 8
-; SI-NEXT: buffer_store_dword v10, off, s[0:3], s32 offset:436 ; 4-byte Folded Spill
+; SI-NEXT: v_bfe_u32 v7, v18, 8, 8
+; SI-NEXT: buffer_store_dword v7, off, s[0:3], s32 offset:432 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_bfe_u32 v10, v15, 8, 8
-; SI-NEXT: buffer_store_dword v10, off, s[0:3], s32 offset:428 ; 4-byte Folded Spill
+; SI-NEXT: v_bfe_u32 v7, v15, 8, 8
+; SI-NEXT: buffer_store_dword v7, off, s[0:3], s32 offset:424 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_bfe_u32 v10, v33, 8, 8
-; SI-NEXT: buffer_store_dword v10, off, s[0:3], s32 offset:420 ; 4-byte Folded Spill
+; SI-NEXT: v_bfe_u32 v7, v33, 8, 8
+; SI-NEXT: buffer_store_dword v7, off, s[0:3], s32 offset:420 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_bfe_u32 v10, v44, 8, 8
-; SI-NEXT: buffer_store_dword v10, off, s[0:3], s32 offset:416 ; 4-byte Folded Spill
+; SI-NEXT: v_bfe_u32 v7, v32, 8, 8
+; SI-NEXT: buffer_store_dword v7, off, s[0:3], s32 offset:412 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_bfe_u32 v10, v31, 8, 8
-; SI-NEXT: buffer_store_dword v10, off, s[0:3], s32 offset:408 ; 4-byte Folded Spill
+; SI-NEXT: v_bfe_u32 v7, v31, 8, 8
+; SI-NEXT: buffer_store_dword v7, off, s[0:3], s32 offset:404 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_bfe_u32 v10, v30, 8, 8
-; SI-NEXT: buffer_store_dword v10, off, s[0:3], s32 offset:400 ; 4-byte Folded Spill
+; SI-NEXT: v_bfe_u32 v7, v10, 8, 8
+; SI-NEXT: buffer_store_dword v7, off, s[0:3], s32 offset:396 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_bfe_u32 v10, v9, 8, 8
-; SI-NEXT: buffer_store_dword v10, off, s[0:3], s32 offset:392 ; 4-byte Folded Spill
+; SI-NEXT: v_bfe_u32 v7, v9, 8, 8
+; SI-NEXT: buffer_store_dword v7, off, s[0:3], s32 offset:388 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_bfe_u32 v10, v8, 8, 8
-; SI-NEXT: buffer_store_dword v10, off, s[0:3], s32 offset:380 ; 4-byte Folded Spill
+; SI-NEXT: v_bfe_u32 v7, v8, 8, 8
+; SI-NEXT: buffer_store_dword v7, off, s[0:3], s32 offset:380 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_bfe_u32 v10, v6, 8, 8
-; SI-NEXT: buffer_store_dword v10, off, s[0:3], s32 offset:336 ; 4-byte Folded Spill
+; SI-NEXT: v_bfe_u32 v7, v29, 8, 8
+; SI-NEXT: buffer_load_dword v29, off, s[0:3], s32 offset:468 ; 4-byte Folded Reload
+; SI-NEXT: buffer_store_dword v7, off, s[0:3], s32 offset:348 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_bfe_u32 v10, v5, 8, 8
-; SI-NEXT: buffer_store_dword v29, off, s[0:3], s32 offset:356 ; 4-byte Folded Spill
-; SI-NEXT: buffer_store_dword v10, off, s[0:3], s32 offset:312 ; 4-byte Folded Spill
-; SI-NEXT: buffer_store_dword v26, off, s[0:3], s32 offset:148 ; 4-byte Folded Spill
-; SI-NEXT: s_waitcnt expcnt(1)
-; SI-NEXT: v_bfe_u32 v10, v26, 8, 8
-; SI-NEXT: buffer_store_dword v10, off, s[0:3], s32 offset:288 ; 4-byte Folded Spill
+; SI-NEXT: v_bfe_u32 v7, v47, 8, 8
+; SI-NEXT: buffer_store_dword v7, off, s[0:3], s32 offset:332 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_bfe_u32 v10, v2, 8, 8
-; SI-NEXT: buffer_store_dword v10, off, s[0:3], s32 offset:268 ; 4-byte Folded Spill
+; SI-NEXT: v_bfe_u32 v7, v5, 8, 8
+; SI-NEXT: buffer_store_dword v7, off, s[0:3], s32 offset:308 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_bfe_u32 v10, v1, 8, 8
-; SI-NEXT: v_alignbit_b32 v48, v55, v13, 24
-; SI-NEXT: v_lshrrev_b32_e32 v29, 8, v34
-; SI-NEXT: v_bfe_u32 v30, v7, 8, 8
-; SI-NEXT: buffer_store_dword v10, off, s[0:3], s32 offset:244 ; 4-byte Folded Spill
+; SI-NEXT: v_bfe_u32 v7, v4, 8, 8
+; SI-NEXT: v_alignbit_b32 v26, v12, v11, 24
+; SI-NEXT: buffer_store_dword v7, off, s[0:3], s32 offset:284 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v26, off, s[0:3], s32 offset:148 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_bfe_u32 v10, v27, 8, 8
-; SI-NEXT: buffer_store_dword v10, off, s[0:3], s32 offset:220 ; 4-byte Folded Spill
+; SI-NEXT: v_alignbit_b32 v26, v12, v11, 16
+; SI-NEXT: v_alignbit_b32 v28, v12, v11, 8
+; SI-NEXT: s_waitcnt vmcnt(5)
+; SI-NEXT: v_bfe_u32 v7, v29, 8, 8
+; SI-NEXT: buffer_store_dword v7, off, s[0:3], s32 offset:244 ; 4-byte Folded Spill
+; SI-NEXT: s_waitcnt expcnt(0)
+; SI-NEXT: v_bfe_u32 v7, v3, 8, 8
+; SI-NEXT: buffer_store_dword v7, off, s[0:3], s32 offset:220 ; 4-byte Folded Spill
; SI-NEXT: .LBB95_5: ; %end
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: buffer_load_dword v10, off, s[0:3], s32 offset:372 ; 4-byte Folded Reload
-; SI-NEXT: s_waitcnt vmcnt(8)
+; SI-NEXT: buffer_load_dword v7, off, s[0:3], s32 offset:372 ; 4-byte Folded Reload
; SI-NEXT: v_and_b32_e32 v37, 0xff, v13
-; SI-NEXT: s_waitcnt vmcnt(1)
-; SI-NEXT: v_lshlrev_b32_e32 v52, 24, v48
; SI-NEXT: v_and_b32_e32 v20, 0xff, v20
; SI-NEXT: v_lshlrev_b32_e32 v20, 16, v20
; SI-NEXT: v_and_b32_e32 v18, 0xff, v18
; SI-NEXT: v_lshlrev_b32_e32 v18, 16, v18
; SI-NEXT: v_and_b32_e32 v15, 0xff, v15
; SI-NEXT: v_lshlrev_b32_e32 v15, 16, v15
+; SI-NEXT: v_and_b32_e32 v10, 0xff, v10
+; SI-NEXT: v_lshlrev_b32_e32 v10, 16, v10
; SI-NEXT: v_and_b32_e32 v9, 0xff, v9
; SI-NEXT: v_lshlrev_b32_e32 v9, 16, v9
; SI-NEXT: v_and_b32_e32 v8, 0xff, v8
; SI-NEXT: v_lshlrev_b32_e32 v8, 16, v8
-; SI-NEXT: v_and_b32_e32 v7, 0xff, v7
-; SI-NEXT: v_lshlrev_b32_e32 v7, 16, v7
-; SI-NEXT: v_and_b32_e32 v6, 0xff, v6
-; SI-NEXT: v_lshlrev_b32_e32 v6, 16, v6
; SI-NEXT: v_and_b32_e32 v5, 0xff, v5
; SI-NEXT: v_lshlrev_b32_e32 v5, 16, v5
+; SI-NEXT: v_and_b32_e32 v4, 0xff, v4
+; SI-NEXT: v_lshlrev_b32_e32 v4, 16, v4
; SI-NEXT: v_and_b32_e32 v2, 0xff, v2
; SI-NEXT: v_lshlrev_b32_e32 v2, 16, v2
-; SI-NEXT: v_and_b32_e32 v1, 0xff, v1
-; SI-NEXT: v_lshlrev_b32_e32 v1, 16, v1
; SI-NEXT: s_waitcnt vmcnt(0)
-; SI-NEXT: v_lshlrev_b32_e32 v51, 8, v10
-; SI-NEXT: buffer_load_dword v10, off, s[0:3], s32 offset:368 ; 4-byte Folded Reload
+; SI-NEXT: v_lshlrev_b32_e32 v51, 8, v7
+; SI-NEXT: buffer_load_dword v7, off, s[0:3], s32 offset:368 ; 4-byte Folded Reload
; SI-NEXT: v_or_b32_e32 v37, v37, v51
; SI-NEXT: v_and_b32_e32 v37, 0xffff, v37
; SI-NEXT: s_waitcnt vmcnt(0)
-; SI-NEXT: v_and_b32_e32 v51, 0xff, v10
+; SI-NEXT: v_and_b32_e32 v51, 0xff, v7
+; SI-NEXT: buffer_load_dword v7, off, s[0:3], s32 offset:364 ; 4-byte Folded Reload
; SI-NEXT: v_lshlrev_b32_e32 v51, 16, v51
+; SI-NEXT: s_waitcnt vmcnt(0)
+; SI-NEXT: v_lshlrev_b32_e32 v52, 24, v7
; SI-NEXT: v_or_b32_e32 v51, v52, v51
; SI-NEXT: v_or_b32_e32 v37, v37, v51
; SI-NEXT: buffer_store_dword v37, v0, s[0:3], 0 offen
-; SI-NEXT: buffer_load_dword v10, off, s[0:3], s32 offset:432 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v7, off, s[0:3], s32 offset:428 ; 4-byte Folded Reload
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_and_b32_e32 v37, 0xff, v55
; SI-NEXT: s_waitcnt vmcnt(0)
-; SI-NEXT: v_lshlrev_b32_e32 v51, 8, v10
-; SI-NEXT: buffer_load_dword v10, off, s[0:3], s32 offset:440 ; 4-byte Folded Reload
+; SI-NEXT: v_lshlrev_b32_e32 v51, 8, v7
+; SI-NEXT: buffer_load_dword v7, off, s[0:3], s32 offset:436 ; 4-byte Folded Reload
; SI-NEXT: v_or_b32_e32 v37, v37, v51
; SI-NEXT: v_and_b32_e32 v37, 0xffff, v37
; SI-NEXT: s_waitcnt vmcnt(0)
-; SI-NEXT: v_lshlrev_b32_e32 v48, 24, v10
+; SI-NEXT: v_lshlrev_b32_e32 v48, 24, v7
; SI-NEXT: v_or_b32_e32 v20, v48, v20
; SI-NEXT: v_or_b32_e32 v20, v37, v20
; SI-NEXT: v_add_i32_e32 v37, vcc, 4, v0
; SI-NEXT: buffer_store_dword v20, v37, s[0:3], 0 offen
-; SI-NEXT: buffer_load_dword v10, off, s[0:3], s32 offset:364 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v7, off, s[0:3], s32 offset:360 ; 4-byte Folded Reload
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_and_b32_e32 v20, 0xff, v57
+; SI-NEXT: v_and_b32_e32 v20, 0xff, v17
; SI-NEXT: s_waitcnt vmcnt(0)
-; SI-NEXT: v_lshlrev_b32_e32 v37, 8, v10
-; SI-NEXT: buffer_load_dword v10, off, s[0:3], s32 offset:360 ; 4-byte Folded Reload
+; SI-NEXT: v_lshlrev_b32_e32 v37, 8, v7
+; SI-NEXT: buffer_load_dword v7, off, s[0:3], s32 offset:352 ; 4-byte Folded Reload
; SI-NEXT: v_or_b32_e32 v20, v20, v37
; SI-NEXT: v_and_b32_e32 v20, 0xffff, v20
; SI-NEXT: s_waitcnt vmcnt(0)
-; SI-NEXT: v_and_b32_e32 v37, 0xff, v10
-; SI-NEXT: buffer_load_dword v10, off, s[0:3], s32 offset:348 ; 4-byte Folded Reload
+; SI-NEXT: v_and_b32_e32 v37, 0xff, v7
+; SI-NEXT: buffer_load_dword v7, off, s[0:3], s32 offset:344 ; 4-byte Folded Reload
; SI-NEXT: v_lshlrev_b32_e32 v37, 16, v37
; SI-NEXT: s_waitcnt vmcnt(0)
-; SI-NEXT: v_lshlrev_b32_e32 v48, 24, v10
+; SI-NEXT: v_lshlrev_b32_e32 v48, 24, v7
; SI-NEXT: v_or_b32_e32 v37, v48, v37
; SI-NEXT: v_or_b32_e32 v20, v20, v37
; SI-NEXT: v_add_i32_e32 v37, vcc, 8, v0
; SI-NEXT: buffer_store_dword v20, v37, s[0:3], 0 offen
-; SI-NEXT: buffer_load_dword v10, off, s[0:3], s32 offset:424 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v7, off, s[0:3], s32 offset:416 ; 4-byte Folded Reload
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_and_b32_e32 v20, 0xff, v17
+; SI-NEXT: v_and_b32_e32 v20, 0xff, v46
; SI-NEXT: s_waitcnt vmcnt(0)
-; SI-NEXT: v_lshlrev_b32_e32 v37, 8, v10
-; SI-NEXT: buffer_load_dword v10, off, s[0:3], s32 offset:436 ; 4-byte Folded Reload
+; SI-NEXT: v_lshlrev_b32_e32 v37, 8, v7
+; SI-NEXT: buffer_load_dword v7, off, s[0:3], s32 offset:432 ; 4-byte Folded Reload
; SI-NEXT: v_or_b32_e32 v20, v20, v37
; SI-NEXT: v_and_b32_e32 v20, 0xffff, v20
; SI-NEXT: s_waitcnt vmcnt(0)
-; SI-NEXT: v_lshlrev_b32_e32 v37, 24, v10
+; SI-NEXT: v_lshlrev_b32_e32 v37, 24, v7
; SI-NEXT: v_or_b32_e32 v18, v37, v18
; SI-NEXT: v_or_b32_e32 v18, v20, v18
; SI-NEXT: v_add_i32_e32 v20, vcc, 12, v0
; SI-NEXT: buffer_store_dword v18, v20, s[0:3], 0 offen
-; SI-NEXT: buffer_load_dword v10, off, s[0:3], s32 offset:344 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v7, off, s[0:3], s32 offset:340 ; 4-byte Folded Reload
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_and_b32_e32 v18, 0xff, v21
+; SI-NEXT: v_and_b32_e32 v18, 0xff, v22
; SI-NEXT: s_waitcnt vmcnt(0)
-; SI-NEXT: v_lshlrev_b32_e32 v20, 8, v10
-; SI-NEXT: buffer_load_dword v10, off, s[0:3], s32 offset:340 ; 4-byte Folded Reload
+; SI-NEXT: v_lshlrev_b32_e32 v20, 8, v7
+; SI-NEXT: buffer_load_dword v7, off, s[0:3], s32 offset:336 ; 4-byte Folded Reload
; SI-NEXT: v_or_b32_e32 v18, v18, v20
; SI-NEXT: v_and_b32_e32 v18, 0xffff, v18
; SI-NEXT: s_waitcnt vmcnt(0)
-; SI-NEXT: v_and_b32_e32 v20, 0xff, v10
-; SI-NEXT: buffer_load_dword v10, off, s[0:3], s32 offset:332 ; 4-byte Folded Reload
+; SI-NEXT: v_and_b32_e32 v20, 0xff, v7
+; SI-NEXT: buffer_load_dword v7, off, s[0:3], s32 offset:328 ; 4-byte Folded Reload
; SI-NEXT: v_lshlrev_b32_e32 v20, 16, v20
; SI-NEXT: s_waitcnt vmcnt(0)
-; SI-NEXT: v_lshlrev_b32_e32 v37, 24, v10
+; SI-NEXT: v_lshlrev_b32_e32 v37, 24, v7
; SI-NEXT: v_or_b32_e32 v20, v37, v20
; SI-NEXT: v_or_b32_e32 v18, v18, v20
; SI-NEXT: v_add_i32_e32 v20, vcc, 16, v0
; SI-NEXT: buffer_store_dword v18, v20, s[0:3], 0 offen
-; SI-NEXT: buffer_load_dword v10, off, s[0:3], s32 offset:412 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v7, off, s[0:3], s32 offset:408 ; 4-byte Folded Reload
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_and_b32_e32 v18, 0xff, v16
; SI-NEXT: s_waitcnt vmcnt(0)
-; SI-NEXT: v_lshlrev_b32_e32 v20, 8, v10
-; SI-NEXT: buffer_load_dword v10, off, s[0:3], s32 offset:428 ; 4-byte Folded Reload
+; SI-NEXT: v_lshlrev_b32_e32 v20, 8, v7
+; SI-NEXT: buffer_load_dword v7, off, s[0:3], s32 offset:424 ; 4-byte Folded Reload
; SI-NEXT: v_or_b32_e32 v18, v18, v20
; SI-NEXT: v_and_b32_e32 v18, 0xffff, v18
; SI-NEXT: s_waitcnt vmcnt(0)
-; SI-NEXT: v_lshlrev_b32_e32 v20, 24, v10
+; SI-NEXT: v_lshlrev_b32_e32 v20, 24, v7
; SI-NEXT: v_or_b32_e32 v15, v20, v15
; SI-NEXT: v_or_b32_e32 v15, v18, v15
; SI-NEXT: v_add_i32_e32 v18, vcc, 20, v0
; SI-NEXT: buffer_store_dword v15, v18, s[0:3], 0 offen
-; SI-NEXT: buffer_load_dword v10, off, s[0:3], s32 offset:324 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v7, off, s[0:3], s32 offset:320 ; 4-byte Folded Reload
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_and_b32_e32 v15, 0xff, v19
+; SI-NEXT: v_and_b32_e32 v15, 0xff, v23
; SI-NEXT: s_waitcnt vmcnt(0)
-; SI-NEXT: v_lshlrev_b32_e32 v18, 8, v10
-; SI-NEXT: buffer_load_dword v10, off, s[0:3], s32 offset:320 ; 4-byte Folded Reload
+; SI-NEXT: v_lshlrev_b32_e32 v18, 8, v7
+; SI-NEXT: buffer_load_dword v7, off, s[0:3], s32 offset:316 ; 4-byte Folded Reload
; SI-NEXT: v_or_b32_e32 v15, v15, v18
; SI-NEXT: v_and_b32_e32 v15, 0xffff, v15
; SI-NEXT: s_waitcnt vmcnt(0)
-; SI-NEXT: v_and_b32_e32 v18, 0xff, v10
-; SI-NEXT: buffer_load_dword v10, off, s[0:3], s32 offset:316 ; 4-byte Folded Reload
+; SI-NEXT: v_and_b32_e32 v18, 0xff, v7
+; SI-NEXT: buffer_load_dword v7, off, s[0:3], s32 offset:312 ; 4-byte Folded Reload
; SI-NEXT: v_lshlrev_b32_e32 v18, 16, v18
; SI-NEXT: s_waitcnt vmcnt(0)
-; SI-NEXT: v_lshlrev_b32_e32 v20, 24, v10
+; SI-NEXT: v_lshlrev_b32_e32 v20, 24, v7
; SI-NEXT: v_or_b32_e32 v18, v20, v18
; SI-NEXT: v_or_b32_e32 v15, v15, v18
; SI-NEXT: v_add_i32_e32 v18, vcc, 24, v0
; SI-NEXT: buffer_store_dword v15, v18, s[0:3], 0 offen
-; SI-NEXT: buffer_load_dword v10, off, s[0:3], s32 offset:404 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v7, off, s[0:3], s32 offset:400 ; 4-byte Folded Reload
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_and_b32_e32 v15, 0xff, v47
+; SI-NEXT: v_and_b32_e32 v15, 0xff, v14
; SI-NEXT: s_waitcnt vmcnt(0)
-; SI-NEXT: v_lshlrev_b32_e32 v18, 8, v10
-; SI-NEXT: buffer_load_dword v10, off, s[0:3], s32 offset:420 ; 4-byte Folded Reload
+; SI-NEXT: v_lshlrev_b32_e32 v18, 8, v7
+; SI-NEXT: buffer_load_dword v7, off, s[0:3], s32 offset:420 ; 4-byte Folded Reload
; SI-NEXT: v_or_b32_e32 v15, v15, v18
; SI-NEXT: v_and_b32_e32 v18, 0xff, v33
; SI-NEXT: v_lshlrev_b32_e32 v18, 16, v18
; SI-NEXT: v_and_b32_e32 v15, 0xffff, v15
; SI-NEXT: s_waitcnt vmcnt(0)
-; SI-NEXT: v_lshlrev_b32_e32 v20, 24, v10
+; SI-NEXT: v_lshlrev_b32_e32 v20, 24, v7
; SI-NEXT: v_or_b32_e32 v18, v20, v18
; SI-NEXT: v_or_b32_e32 v15, v15, v18
; SI-NEXT: v_add_i32_e32 v18, vcc, 28, v0
; SI-NEXT: buffer_store_dword v15, v18, s[0:3], 0 offen
-; SI-NEXT: buffer_load_dword v10, off, s[0:3], s32 offset:300 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v7, off, s[0:3], s32 offset:304 ; 4-byte Folded Reload
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_and_b32_e32 v15, 0xff, v43
-; SI-NEXT: v_lshlrev_b32_e32 v18, 8, v28
+; SI-NEXT: v_and_b32_e32 v15, 0xff, v21
+; SI-NEXT: s_waitcnt vmcnt(0)
+; SI-NEXT: v_lshlrev_b32_e32 v18, 8, v7
+; SI-NEXT: buffer_load_dword v7, off, s[0:3], s32 offset:296 ; 4-byte Folded Reload
; SI-NEXT: v_or_b32_e32 v15, v15, v18
; SI-NEXT: v_and_b32_e32 v15, 0xffff, v15
; SI-NEXT: s_waitcnt vmcnt(0)
-; SI-NEXT: v_and_b32_e32 v18, 0xff, v10
-; SI-NEXT: buffer_load_dword v10, off, s[0:3], s32 offset:296 ; 4-byte Folded Reload
+; SI-NEXT: v_and_b32_e32 v18, 0xff, v7
+; SI-NEXT: buffer_load_dword v7, off, s[0:3], s32 offset:292 ; 4-byte Folded Reload
; SI-NEXT: v_lshlrev_b32_e32 v18, 16, v18
; SI-NEXT: s_waitcnt vmcnt(0)
-; SI-NEXT: v_lshlrev_b32_e32 v20, 24, v10
+; SI-NEXT: v_lshlrev_b32_e32 v20, 24, v7
; SI-NEXT: v_or_b32_e32 v18, v20, v18
; SI-NEXT: v_or_b32_e32 v15, v15, v18
; SI-NEXT: v_add_i32_e32 v18, vcc, 32, v0
; SI-NEXT: buffer_store_dword v15, v18, s[0:3], 0 offen
-; SI-NEXT: buffer_load_dword v10, off, s[0:3], s32 offset:396 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v7, off, s[0:3], s32 offset:392 ; 4-byte Folded Reload
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_and_b32_e32 v15, 0xff, v14
+; SI-NEXT: v_and_b32_e32 v15, 0xff, v60
; SI-NEXT: s_waitcnt vmcnt(0)
-; SI-NEXT: v_lshlrev_b32_e32 v18, 8, v10
-; SI-NEXT: buffer_load_dword v10, off, s[0:3], s32 offset:416 ; 4-byte Folded Reload
+; SI-NEXT: v_lshlrev_b32_e32 v18, 8, v7
+; SI-NEXT: buffer_load_dword v7, off, s[0:3], s32 offset:412 ; 4-byte Folded Reload
; SI-NEXT: v_or_b32_e32 v15, v15, v18
-; SI-NEXT: v_and_b32_e32 v18, 0xff, v44
+; SI-NEXT: v_and_b32_e32 v18, 0xff, v32
; SI-NEXT: v_lshlrev_b32_e32 v18, 16, v18
; SI-NEXT: v_and_b32_e32 v15, 0xffff, v15
; SI-NEXT: s_waitcnt vmcnt(0)
-; SI-NEXT: v_lshlrev_b32_e32 v20, 24, v10
+; SI-NEXT: v_lshlrev_b32_e32 v20, 24, v7
; SI-NEXT: v_or_b32_e32 v18, v20, v18
; SI-NEXT: v_or_b32_e32 v15, v15, v18
; SI-NEXT: v_add_i32_e32 v18, vcc, 36, v0
; SI-NEXT: buffer_store_dword v15, v18, s[0:3], 0 offen
-; SI-NEXT: buffer_load_dword v10, off, s[0:3], s32 offset:292 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v7, off, s[0:3], s32 offset:288 ; 4-byte Folded Reload
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_and_b32_e32 v15, 0xff, v42
+; SI-NEXT: v_and_b32_e32 v15, 0xff, v40
; SI-NEXT: s_waitcnt vmcnt(0)
-; SI-NEXT: v_lshlrev_b32_e32 v18, 8, v10
-; SI-NEXT: buffer_load_dword v10, off, s[0:3], s32 offset:284 ; 4-byte Folded Reload
+; SI-NEXT: v_lshlrev_b32_e32 v18, 8, v7
+; SI-NEXT: buffer_load_dword v7, off, s[0:3], s32 offset:280 ; 4-byte Folded Reload
; SI-NEXT: v_or_b32_e32 v15, v15, v18
; SI-NEXT: v_and_b32_e32 v15, 0xffff, v15
; SI-NEXT: s_waitcnt vmcnt(0)
-; SI-NEXT: v_and_b32_e32 v18, 0xff, v10
-; SI-NEXT: buffer_load_dword v10, off, s[0:3], s32 offset:276 ; 4-byte Folded Reload
+; SI-NEXT: v_and_b32_e32 v18, 0xff, v7
+; SI-NEXT: buffer_load_dword v7, off, s[0:3], s32 offset:272 ; 4-byte Folded Reload
; SI-NEXT: v_lshlrev_b32_e32 v18, 16, v18
; SI-NEXT: s_waitcnt vmcnt(0)
-; SI-NEXT: v_lshlrev_b32_e32 v20, 24, v10
+; SI-NEXT: v_lshlrev_b32_e32 v20, 24, v7
; SI-NEXT: v_or_b32_e32 v18, v20, v18
; SI-NEXT: v_or_b32_e32 v15, v15, v18
; SI-NEXT: v_add_i32_e32 v18, vcc, 40, v0
; SI-NEXT: buffer_store_dword v15, v18, s[0:3], 0 offen
-; SI-NEXT: buffer_load_dword v10, off, s[0:3], s32 offset:388 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v7, off, s[0:3], s32 offset:384 ; 4-byte Folded Reload
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_and_b32_e32 v15, 0xff, v60
+; SI-NEXT: v_and_b32_e32 v15, 0xff, v6
+; SI-NEXT: v_and_b32_e32 v6, 0xff, v47
+; SI-NEXT: v_lshlrev_b32_e32 v6, 16, v6
; SI-NEXT: s_waitcnt vmcnt(0)
-; SI-NEXT: v_lshlrev_b32_e32 v18, 8, v10
-; SI-NEXT: buffer_load_dword v10, off, s[0:3], s32 offset:408 ; 4-byte Folded Reload
+; SI-NEXT: v_lshlrev_b32_e32 v18, 8, v7
+; SI-NEXT: buffer_load_dword v7, off, s[0:3], s32 offset:404 ; 4-byte Folded Reload
; SI-NEXT: v_or_b32_e32 v15, v15, v18
; SI-NEXT: v_and_b32_e32 v18, 0xff, v31
; SI-NEXT: v_lshlrev_b32_e32 v18, 16, v18
; SI-NEXT: v_and_b32_e32 v15, 0xffff, v15
; SI-NEXT: s_waitcnt vmcnt(0)
-; SI-NEXT: v_lshlrev_b32_e32 v20, 24, v10
+; SI-NEXT: v_lshlrev_b32_e32 v20, 24, v7
; SI-NEXT: v_or_b32_e32 v18, v20, v18
; SI-NEXT: v_or_b32_e32 v15, v15, v18
; SI-NEXT: v_add_i32_e32 v18, vcc, 44, v0
; SI-NEXT: buffer_store_dword v15, v18, s[0:3], 0 offen
-; SI-NEXT: buffer_load_dword v10, off, s[0:3], s32 offset:272 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v7, off, s[0:3], s32 offset:260 ; 4-byte Folded Reload
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_and_b32_e32 v15, 0xff, v22
-; SI-NEXT: s_waitcnt vmcnt(0)
-; SI-NEXT: v_lshlrev_b32_e32 v18, 8, v10
-; SI-NEXT: buffer_load_dword v10, off, s[0:3], s32 offset:264 ; 4-byte Folded Reload
+; SI-NEXT: v_and_b32_e32 v15, 0xff, v53
+; SI-NEXT: v_lshlrev_b32_e32 v18, 8, v30
; SI-NEXT: v_or_b32_e32 v15, v15, v18
-; SI-NEXT: v_and_b32_e32 v15, 0xffff, v15
-; SI-NEXT: s_waitcnt vmcnt(0)
-; SI-NEXT: v_and_b32_e32 v18, 0xff, v10
-; SI-NEXT: buffer_load_dword v10, off, s[0:3], s32 offset:260 ; 4-byte Folded Reload
+; SI-NEXT: v_and_b32_e32 v18, 0xff, v27
; SI-NEXT: v_lshlrev_b32_e32 v18, 16, v18
+; SI-NEXT: v_and_b32_e32 v15, 0xffff, v15
; SI-NEXT: s_waitcnt vmcnt(0)
-; SI-NEXT: v_lshlrev_b32_e32 v20, 24, v10
+; SI-NEXT: v_lshlrev_b32_e32 v20, 24, v7
; SI-NEXT: v_or_b32_e32 v18, v20, v18
; SI-NEXT: v_or_b32_e32 v15, v15, v18
; SI-NEXT: v_add_i32_e32 v18, vcc, 48, v0
; SI-NEXT: buffer_store_dword v15, v18, s[0:3], 0 offen
-; SI-NEXT: buffer_load_dword v10, off, s[0:3], s32 offset:376 ; 4-byte Folded Reload
-; SI-NEXT: buffer_load_dword v13, off, s[0:3], s32 offset:400 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v7, off, s[0:3], s32 offset:376 ; 4-byte Folded Reload
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_and_b32_e32 v15, 0xff, v24
-; SI-NEXT: s_waitcnt vmcnt(1)
-; SI-NEXT: v_lshlrev_b32_e32 v18, 8, v10
-; SI-NEXT: v_and_b32_e32 v10, 0xff, v32
+; SI-NEXT: v_and_b32_e32 v15, 0xff, v63
+; SI-NEXT: s_waitcnt vmcnt(0)
+; SI-NEXT: v_lshlrev_b32_e32 v18, 8, v7
+; SI-NEXT: buffer_load_dword v7, off, s[0:3], s32 offset:396 ; 4-byte Folded Reload
; SI-NEXT: v_or_b32_e32 v15, v15, v18
-; SI-NEXT: v_lshlrev_b32_e32 v10, 16, v10
+; SI-NEXT: v_and_b32_e32 v15, 0xffff, v15
; SI-NEXT: s_waitcnt vmcnt(0)
-; SI-NEXT: v_lshlrev_b32_e32 v18, 24, v13
+; SI-NEXT: v_lshlrev_b32_e32 v18, 24, v7
; SI-NEXT: v_or_b32_e32 v10, v18, v10
-; SI-NEXT: v_and_b32_e32 v15, 0xffff, v15
; SI-NEXT: v_or_b32_e32 v10, v15, v10
; SI-NEXT: v_add_i32_e32 v15, vcc, 52, v0
; SI-NEXT: buffer_store_dword v10, v15, s[0:3], 0 offen
-; SI-NEXT: buffer_load_dword v13, off, s[0:3], s32 offset:252 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v7, off, s[0:3], s32 offset:252 ; 4-byte Folded Reload
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_and_b32_e32 v10, 0xff, v4
+; SI-NEXT: v_and_b32_e32 v10, 0xff, v41
; SI-NEXT: s_waitcnt vmcnt(0)
-; SI-NEXT: v_lshlrev_b32_e32 v15, 8, v13
-; SI-NEXT: buffer_load_dword v13, off, s[0:3], s32 offset:248 ; 4-byte Folded Reload
+; SI-NEXT: v_lshlrev_b32_e32 v15, 8, v7
+; SI-NEXT: buffer_load_dword v7, off, s[0:3], s32 offset:248 ; 4-byte Folded Reload
; SI-NEXT: v_or_b32_e32 v10, v10, v15
; SI-NEXT: v_and_b32_e32 v10, 0xffff, v10
; SI-NEXT: s_waitcnt vmcnt(0)
-; SI-NEXT: v_and_b32_e32 v15, 0xff, v13
-; SI-NEXT: buffer_load_dword v13, off, s[0:3], s32 offset:240 ; 4-byte Folded Reload
+; SI-NEXT: v_and_b32_e32 v15, 0xff, v7
+; SI-NEXT: buffer_load_dword v7, off, s[0:3], s32 offset:240 ; 4-byte Folded Reload
; SI-NEXT: v_lshlrev_b32_e32 v15, 16, v15
; SI-NEXT: s_waitcnt vmcnt(0)
-; SI-NEXT: v_lshlrev_b32_e32 v18, 24, v13
+; SI-NEXT: v_lshlrev_b32_e32 v18, 24, v7
; SI-NEXT: v_or_b32_e32 v15, v18, v15
; SI-NEXT: v_or_b32_e32 v10, v10, v15
; SI-NEXT: v_add_i32_e32 v15, vcc, 56, v0
; SI-NEXT: buffer_store_dword v10, v15, s[0:3], 0 offen
-; SI-NEXT: buffer_load_dword v13, off, s[0:3], s32 offset:392 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v7, off, s[0:3], s32 offset:356 ; 4-byte Folded Reload
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_and_b32_e32 v10, 0xff, v34
-; SI-NEXT: v_lshlrev_b32_e32 v15, 8, v29
+; SI-NEXT: s_waitcnt vmcnt(0)
+; SI-NEXT: v_lshlrev_b32_e32 v15, 8, v7
+; SI-NEXT: buffer_load_dword v7, off, s[0:3], s32 offset:388 ; 4-byte Folded Reload
; SI-NEXT: v_or_b32_e32 v10, v10, v15
; SI-NEXT: v_and_b32_e32 v10, 0xffff, v10
; SI-NEXT: s_waitcnt vmcnt(0)
-; SI-NEXT: v_lshlrev_b32_e32 v15, 24, v13
+; SI-NEXT: v_lshlrev_b32_e32 v15, 24, v7
; SI-NEXT: v_or_b32_e32 v9, v15, v9
; SI-NEXT: v_or_b32_e32 v9, v10, v9
; SI-NEXT: v_add_i32_e32 v10, vcc, 60, v0
; SI-NEXT: buffer_store_dword v9, v10, s[0:3], 0 offen
-; SI-NEXT: buffer_load_dword v10, off, s[0:3], s32 offset:232 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v7, off, s[0:3], s32 offset:232 ; 4-byte Folded Reload
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_and_b32_e32 v9, 0xff, v3
-; SI-NEXT: buffer_load_dword v13, off, s[0:3], s32 offset:224 ; 4-byte Folded Reload
-; SI-NEXT: s_waitcnt vmcnt(1)
-; SI-NEXT: v_lshlrev_b32_e32 v10, 8, v10
+; SI-NEXT: v_and_b32_e32 v9, 0xff, v36
+; SI-NEXT: s_waitcnt vmcnt(0)
+; SI-NEXT: v_lshlrev_b32_e32 v10, 8, v7
+; SI-NEXT: buffer_load_dword v7, off, s[0:3], s32 offset:228 ; 4-byte Folded Reload
; SI-NEXT: v_or_b32_e32 v9, v9, v10
-; SI-NEXT: buffer_load_dword v10, off, s[0:3], s32 offset:228 ; 4-byte Folded Reload
; SI-NEXT: v_and_b32_e32 v9, 0xffff, v9
-; SI-NEXT: s_waitcnt vmcnt(1)
-; SI-NEXT: v_lshlrev_b32_e32 v15, 24, v13
; SI-NEXT: s_waitcnt vmcnt(0)
-; SI-NEXT: v_and_b32_e32 v10, 0xff, v10
+; SI-NEXT: v_and_b32_e32 v10, 0xff, v7
+; SI-NEXT: buffer_load_dword v7, off, s[0:3], s32 offset:224 ; 4-byte Folded Reload
; SI-NEXT: v_lshlrev_b32_e32 v10, 16, v10
+; SI-NEXT: s_waitcnt vmcnt(0)
+; SI-NEXT: v_lshlrev_b32_e32 v15, 24, v7
; SI-NEXT: v_or_b32_e32 v10, v15, v10
; SI-NEXT: v_or_b32_e32 v9, v9, v10
; SI-NEXT: v_add_i32_e32 v10, vcc, 64, v0
; SI-NEXT: buffer_store_dword v9, v10, s[0:3], 0 offen
-; SI-NEXT: buffer_load_dword v3, off, s[0:3], s32 offset:328 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v7, off, s[0:3], s32 offset:324 ; 4-byte Folded Reload
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_and_b32_e32 v9, 0xff, v59
+; SI-NEXT: v_and_b32_e32 v9, 0xff, v56
; SI-NEXT: s_waitcnt vmcnt(0)
-; SI-NEXT: v_lshlrev_b32_e32 v10, 8, v3
-; SI-NEXT: buffer_load_dword v3, off, s[0:3], s32 offset:380 ; 4-byte Folded Reload
+; SI-NEXT: v_lshlrev_b32_e32 v10, 8, v7
+; SI-NEXT: buffer_load_dword v7, off, s[0:3], s32 offset:380 ; 4-byte Folded Reload
; SI-NEXT: v_or_b32_e32 v9, v9, v10
; SI-NEXT: v_and_b32_e32 v9, 0xffff, v9
; SI-NEXT: s_waitcnt vmcnt(0)
-; SI-NEXT: v_lshlrev_b32_e32 v10, 24, v3
+; SI-NEXT: v_lshlrev_b32_e32 v10, 24, v7
; SI-NEXT: v_or_b32_e32 v8, v10, v8
; SI-NEXT: v_or_b32_e32 v8, v9, v8
; SI-NEXT: v_add_i32_e32 v9, vcc, 0x44, v0
; SI-NEXT: buffer_store_dword v8, v9, s[0:3], 0 offen
-; SI-NEXT: buffer_load_dword v4, off, s[0:3], s32 offset:216 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v7, off, s[0:3], s32 offset:216 ; 4-byte Folded Reload
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_and_b32_e32 v8, 0xff, v49
+; SI-NEXT: v_and_b32_e32 v8, 0xff, v38
; SI-NEXT: s_waitcnt vmcnt(0)
-; SI-NEXT: v_lshlrev_b32_e32 v9, 8, v4
-; SI-NEXT: buffer_load_dword v4, off, s[0:3], s32 offset:208 ; 4-byte Folded Reload
+; SI-NEXT: v_lshlrev_b32_e32 v9, 8, v7
+; SI-NEXT: buffer_load_dword v7, off, s[0:3], s32 offset:208 ; 4-byte Folded Reload
; SI-NEXT: v_or_b32_e32 v8, v8, v9
; SI-NEXT: v_and_b32_e32 v8, 0xffff, v8
; SI-NEXT: s_waitcnt vmcnt(0)
-; SI-NEXT: v_and_b32_e32 v9, 0xff, v4
-; SI-NEXT: buffer_load_dword v4, off, s[0:3], s32 offset:204 ; 4-byte Folded Reload
+; SI-NEXT: v_and_b32_e32 v9, 0xff, v7
+; SI-NEXT: buffer_load_dword v7, off, s[0:3], s32 offset:204 ; 4-byte Folded Reload
; SI-NEXT: v_lshlrev_b32_e32 v9, 16, v9
; SI-NEXT: s_waitcnt vmcnt(0)
-; SI-NEXT: v_lshlrev_b32_e32 v10, 24, v4
+; SI-NEXT: v_lshlrev_b32_e32 v10, 24, v7
; SI-NEXT: v_or_b32_e32 v9, v10, v9
; SI-NEXT: v_or_b32_e32 v8, v8, v9
; SI-NEXT: v_add_i32_e32 v9, vcc, 0x48, v0
; SI-NEXT: buffer_store_dword v8, v9, s[0:3], 0 offen
-; SI-NEXT: buffer_load_dword v4, off, s[0:3], s32 offset:304 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v7, off, s[0:3], s32 offset:300 ; 4-byte Folded Reload
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_and_b32_e32 v8, 0xff, v62
+; SI-NEXT: v_and_b32_e32 v8, 0xff, v58
; SI-NEXT: s_waitcnt vmcnt(0)
-; SI-NEXT: v_lshlrev_b32_e32 v9, 8, v4
+; SI-NEXT: v_lshlrev_b32_e32 v9, 8, v7
; SI-NEXT: v_or_b32_e32 v8, v8, v9
-; SI-NEXT: v_lshlrev_b32_e32 v9, 24, v30
+; SI-NEXT: buffer_load_dword v9, off, s[0:3], s32 offset:348 ; 4-byte Folded Reload
+; SI-NEXT: v_and_b32_e32 v7, 0xff, v62
+; SI-NEXT: v_lshlrev_b32_e32 v7, 16, v7
; SI-NEXT: v_and_b32_e32 v8, 0xffff, v8
+; SI-NEXT: s_waitcnt vmcnt(0)
+; SI-NEXT: v_lshlrev_b32_e32 v9, 24, v9
; SI-NEXT: v_or_b32_e32 v7, v9, v7
; SI-NEXT: v_or_b32_e32 v7, v8, v7
; SI-NEXT: v_add_i32_e32 v8, vcc, 0x4c, v0
; SI-NEXT: buffer_store_dword v7, v8, s[0:3], 0 offen
-; SI-NEXT: buffer_load_dword v4, off, s[0:3], s32 offset:200 ; 4-byte Folded Reload
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_and_b32_e32 v7, 0xff, v36
+; SI-NEXT: buffer_load_dword v7, off, s[0:3], s32 offset:264 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v8, off, s[0:3], s32 offset:200 ; 4-byte Folded Reload
; SI-NEXT: buffer_load_dword v9, off, s[0:3], s32 offset:188 ; 4-byte Folded Reload
+; SI-NEXT: s_waitcnt vmcnt(2)
+; SI-NEXT: v_and_b32_e32 v7, 0xff, v7
; SI-NEXT: s_waitcnt vmcnt(1)
-; SI-NEXT: v_lshlrev_b32_e32 v8, 8, v4
+; SI-NEXT: v_lshlrev_b32_e32 v8, 8, v8
; SI-NEXT: v_or_b32_e32 v7, v7, v8
; SI-NEXT: buffer_load_dword v8, off, s[0:3], s32 offset:192 ; 4-byte Folded Reload
; SI-NEXT: v_and_b32_e32 v7, 0xffff, v7
@@ -197562,24 +197668,26 @@ define inreg <128 x i8> @bitcast_v64f16_to_v128i8_scalar(<64 x half> inreg %a, i
; SI-NEXT: v_or_b32_e32 v7, v7, v8
; SI-NEXT: v_add_i32_e32 v8, vcc, 0x50, v0
; SI-NEXT: buffer_store_dword v7, v8, s[0:3], 0 offen
-; SI-NEXT: buffer_load_dword v4, off, s[0:3], s32 offset:280 ; 4-byte Folded Reload
-; SI-NEXT: buffer_load_dword v3, off, s[0:3], s32 offset:336 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v8, off, s[0:3], s32 offset:276 ; 4-byte Folded Reload
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_and_b32_e32 v7, 0xff, v25
-; SI-NEXT: s_waitcnt vmcnt(1)
-; SI-NEXT: v_lshlrev_b32_e32 v8, 8, v4
-; SI-NEXT: v_or_b32_e32 v7, v7, v8
; SI-NEXT: s_waitcnt vmcnt(0)
-; SI-NEXT: v_lshlrev_b32_e32 v8, 24, v3
+; SI-NEXT: v_lshlrev_b32_e32 v8, 8, v8
+; SI-NEXT: v_or_b32_e32 v7, v7, v8
+; SI-NEXT: buffer_load_dword v8, off, s[0:3], s32 offset:332 ; 4-byte Folded Reload
; SI-NEXT: v_and_b32_e32 v7, 0xffff, v7
+; SI-NEXT: s_waitcnt vmcnt(0)
+; SI-NEXT: v_lshlrev_b32_e32 v8, 24, v8
; SI-NEXT: v_or_b32_e32 v6, v8, v6
; SI-NEXT: v_or_b32_e32 v6, v7, v6
; SI-NEXT: v_add_i32_e32 v7, vcc, 0x54, v0
; SI-NEXT: buffer_store_dword v6, v7, s[0:3], 0 offen
-; SI-NEXT: buffer_load_dword v7, off, s[0:3], s32 offset:184 ; 4-byte Folded Reload
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_and_b32_e32 v6, 0xff, v35
+; SI-NEXT: buffer_load_dword v6, off, s[0:3], s32 offset:268 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v7, off, s[0:3], s32 offset:184 ; 4-byte Folded Reload
; SI-NEXT: buffer_load_dword v8, off, s[0:3], s32 offset:172 ; 4-byte Folded Reload
+; SI-NEXT: s_waitcnt vmcnt(2)
+; SI-NEXT: v_and_b32_e32 v6, 0xff, v6
; SI-NEXT: s_waitcnt vmcnt(1)
; SI-NEXT: v_lshlrev_b32_e32 v7, 8, v7
; SI-NEXT: v_or_b32_e32 v6, v6, v7
@@ -197594,25 +197702,24 @@ define inreg <128 x i8> @bitcast_v64f16_to_v128i8_scalar(<64 x half> inreg %a, i
; SI-NEXT: v_or_b32_e32 v6, v6, v7
; SI-NEXT: v_add_i32_e32 v7, vcc, 0x58, v0
; SI-NEXT: buffer_store_dword v6, v7, s[0:3], 0 offen
-; SI-NEXT: buffer_load_dword v4, off, s[0:3], s32 offset:256 ; 4-byte Folded Reload
-; SI-NEXT: buffer_load_dword v3, off, s[0:3], s32 offset:312 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v7, off, s[0:3], s32 offset:256 ; 4-byte Folded Reload
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_and_b32_e32 v6, 0xff, v23
-; SI-NEXT: s_waitcnt vmcnt(1)
-; SI-NEXT: v_lshlrev_b32_e32 v7, 8, v4
-; SI-NEXT: v_or_b32_e32 v6, v6, v7
+; SI-NEXT: v_and_b32_e32 v6, 0xff, v45
; SI-NEXT: s_waitcnt vmcnt(0)
-; SI-NEXT: v_lshlrev_b32_e32 v7, 24, v3
+; SI-NEXT: v_lshlrev_b32_e32 v7, 8, v7
+; SI-NEXT: v_or_b32_e32 v6, v6, v7
+; SI-NEXT: buffer_load_dword v7, off, s[0:3], s32 offset:308 ; 4-byte Folded Reload
; SI-NEXT: v_and_b32_e32 v6, 0xffff, v6
+; SI-NEXT: s_waitcnt vmcnt(0)
+; SI-NEXT: v_lshlrev_b32_e32 v7, 24, v7
; SI-NEXT: v_or_b32_e32 v5, v7, v5
; SI-NEXT: v_or_b32_e32 v5, v6, v5
; SI-NEXT: v_add_i32_e32 v6, vcc, 0x5c, v0
; SI-NEXT: buffer_store_dword v5, v6, s[0:3], 0 offen
-; SI-NEXT: buffer_load_dword v3, off, s[0:3], s32 offset:352 ; 4-byte Folded Reload
; SI-NEXT: buffer_load_dword v6, off, s[0:3], s32 offset:168 ; 4-byte Folded Reload
+; SI-NEXT: s_waitcnt expcnt(0)
+; SI-NEXT: v_and_b32_e32 v5, 0xff, v19
; SI-NEXT: buffer_load_dword v7, off, s[0:3], s32 offset:160 ; 4-byte Folded Reload
-; SI-NEXT: s_waitcnt vmcnt(2) expcnt(0)
-; SI-NEXT: v_and_b32_e32 v5, 0xff, v3
; SI-NEXT: s_waitcnt vmcnt(1)
; SI-NEXT: v_lshlrev_b32_e32 v6, 8, v6
; SI-NEXT: v_or_b32_e32 v5, v5, v6
@@ -197627,17 +197734,13 @@ define inreg <128 x i8> @bitcast_v64f16_to_v128i8_scalar(<64 x half> inreg %a, i
; SI-NEXT: v_or_b32_e32 v5, v5, v6
; SI-NEXT: v_add_i32_e32 v6, vcc, 0x60, v0
; SI-NEXT: buffer_store_dword v5, v6, s[0:3], 0 offen
-; SI-NEXT: buffer_load_dword v4, off, s[0:3], s32 offset:236 ; 4-byte Folded Reload
-; SI-NEXT: buffer_load_dword v3, off, s[0:3], s32 offset:148 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v6, off, s[0:3], s32 offset:236 ; 4-byte Folded Reload
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_and_b32_e32 v5, 0xff, v61
-; SI-NEXT: s_waitcnt vmcnt(1)
-; SI-NEXT: v_lshlrev_b32_e32 v6, 8, v4
+; SI-NEXT: v_and_b32_e32 v5, 0xff, v54
+; SI-NEXT: s_waitcnt vmcnt(0)
+; SI-NEXT: v_lshlrev_b32_e32 v6, 8, v6
; SI-NEXT: v_or_b32_e32 v5, v5, v6
-; SI-NEXT: buffer_load_dword v6, off, s[0:3], s32 offset:288 ; 4-byte Folded Reload
-; SI-NEXT: s_waitcnt vmcnt(1)
-; SI-NEXT: v_and_b32_e32 v4, 0xff, v3
-; SI-NEXT: v_lshlrev_b32_e32 v4, 16, v4
+; SI-NEXT: buffer_load_dword v6, off, s[0:3], s32 offset:284 ; 4-byte Folded Reload
; SI-NEXT: v_and_b32_e32 v5, 0xffff, v5
; SI-NEXT: s_waitcnt vmcnt(0)
; SI-NEXT: v_lshlrev_b32_e32 v6, 24, v6
@@ -197645,11 +197748,12 @@ define inreg <128 x i8> @bitcast_v64f16_to_v128i8_scalar(<64 x half> inreg %a, i
; SI-NEXT: v_or_b32_e32 v4, v5, v4
; SI-NEXT: v_add_i32_e32 v5, vcc, 0x64, v0
; SI-NEXT: buffer_store_dword v4, v5, s[0:3], 0 offen
-; SI-NEXT: buffer_load_dword v3, off, s[0:3], s32 offset:308 ; 4-byte Folded Reload
; SI-NEXT: buffer_load_dword v5, off, s[0:3], s32 offset:156 ; 4-byte Folded Reload
-; SI-NEXT: v_lshlrev_b32_e32 v6, 24, v40
-; SI-NEXT: s_waitcnt vmcnt(1) expcnt(0)
-; SI-NEXT: v_and_b32_e32 v4, 0xff, v3
+; SI-NEXT: s_waitcnt expcnt(0)
+; SI-NEXT: v_and_b32_e32 v4, 0xff, v1
+; SI-NEXT: v_lshlrev_b32_e32 v6, 24, v24
+; SI-NEXT: v_and_b32_e32 v1, 0xff, v29
+; SI-NEXT: v_lshlrev_b32_e32 v1, 16, v1
; SI-NEXT: s_waitcnt vmcnt(0)
; SI-NEXT: v_lshlrev_b32_e32 v5, 8, v5
; SI-NEXT: v_or_b32_e32 v4, v4, v5
@@ -197664,27 +197768,23 @@ define inreg <128 x i8> @bitcast_v64f16_to_v128i8_scalar(<64 x half> inreg %a, i
; SI-NEXT: buffer_store_dword v4, v5, s[0:3], 0 offen
; SI-NEXT: buffer_load_dword v5, off, s[0:3], s32 offset:212 ; 4-byte Folded Reload
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_and_b32_e32 v4, 0xff, v58
+; SI-NEXT: v_and_b32_e32 v4, 0xff, v39
; SI-NEXT: s_waitcnt vmcnt(0)
; SI-NEXT: v_lshlrev_b32_e32 v5, 8, v5
; SI-NEXT: v_or_b32_e32 v4, v4, v5
-; SI-NEXT: buffer_load_dword v5, off, s[0:3], s32 offset:268 ; 4-byte Folded Reload
+; SI-NEXT: v_lshlrev_b32_e32 v5, 24, v35
; SI-NEXT: v_and_b32_e32 v4, 0xffff, v4
-; SI-NEXT: s_waitcnt vmcnt(0)
-; SI-NEXT: v_lshlrev_b32_e32 v5, 24, v5
; SI-NEXT: v_or_b32_e32 v2, v5, v2
; SI-NEXT: v_or_b32_e32 v2, v4, v2
; SI-NEXT: v_add_i32_e32 v4, vcc, 0x6c, v0
; SI-NEXT: buffer_store_dword v2, v4, s[0:3], 0 offen
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:356 ; 4-byte Folded Reload
-; SI-NEXT: v_lshlrev_b32_e32 v4, 8, v53
-; SI-NEXT: v_lshlrev_b32_e32 v5, 24, v50
-; SI-NEXT: s_waitcnt vmcnt(0)
-; SI-NEXT: v_and_b32_e32 v2, 0xff, v2
+; SI-NEXT: v_and_b32_e32 v2, 0xff, v59
+; SI-NEXT: v_lshlrev_b32_e32 v4, 8, v57
; SI-NEXT: v_or_b32_e32 v2, v2, v4
-; SI-NEXT: v_and_b32_e32 v4, 0xff, v39
+; SI-NEXT: v_and_b32_e32 v4, 0xff, v42
; SI-NEXT: v_lshlrev_b32_e32 v4, 16, v4
+; SI-NEXT: v_lshlrev_b32_e32 v5, 24, v43
; SI-NEXT: v_and_b32_e32 v2, 0xffff, v2
; SI-NEXT: v_or_b32_e32 v4, v5, v4
; SI-NEXT: v_or_b32_e32 v2, v2, v4
@@ -197692,7 +197792,7 @@ define inreg <128 x i8> @bitcast_v64f16_to_v128i8_scalar(<64 x half> inreg %a, i
; SI-NEXT: buffer_store_dword v2, v4, s[0:3], 0 offen
; SI-NEXT: buffer_load_dword v4, off, s[0:3], s32 offset:196 ; 4-byte Folded Reload
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_and_b32_e32 v2, 0xff, v54
+; SI-NEXT: v_and_b32_e32 v2, 0xff, v50
; SI-NEXT: s_waitcnt vmcnt(0)
; SI-NEXT: v_lshlrev_b32_e32 v4, 8, v4
; SI-NEXT: v_or_b32_e32 v2, v2, v4
@@ -197704,33 +197804,33 @@ define inreg <128 x i8> @bitcast_v64f16_to_v128i8_scalar(<64 x half> inreg %a, i
; SI-NEXT: v_or_b32_e32 v1, v2, v1
; SI-NEXT: v_add_i32_e32 v2, vcc, 0x74, v0
; SI-NEXT: buffer_store_dword v1, v2, s[0:3], 0 offen
+; SI-NEXT: buffer_load_dword v4, off, s[0:3], s32 offset:148 ; 4-byte Folded Reload
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:384 ; 4-byte Folded Reload
-; SI-NEXT: v_lshlrev_b32_e32 v2, 8, v38
-; SI-NEXT: v_lshlrev_b32_e32 v4, 24, v11
-; SI-NEXT: s_waitcnt vmcnt(0)
-; SI-NEXT: v_and_b32_e32 v1, 0xff, v1
+; SI-NEXT: v_and_b32_e32 v1, 0xff, v11
+; SI-NEXT: v_lshlrev_b32_e32 v2, 8, v28
; SI-NEXT: v_or_b32_e32 v1, v1, v2
-; SI-NEXT: v_and_b32_e32 v2, 0xff, v56
+; SI-NEXT: v_and_b32_e32 v2, 0xff, v26
; SI-NEXT: v_lshlrev_b32_e32 v2, 16, v2
; SI-NEXT: v_and_b32_e32 v1, 0xffff, v1
+; SI-NEXT: s_waitcnt vmcnt(0)
+; SI-NEXT: v_lshlrev_b32_e32 v4, 24, v4
; SI-NEXT: v_or_b32_e32 v2, v4, v2
; SI-NEXT: v_or_b32_e32 v1, v1, v2
; SI-NEXT: v_add_i32_e32 v2, vcc, 0x78, v0
; SI-NEXT: buffer_store_dword v1, v2, s[0:3], 0 offen
; SI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:176 ; 4-byte Folded Reload
-; SI-NEXT: buffer_load_dword v3, off, s[0:3], s32 offset:220 ; 4-byte Folded Reload
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_and_b32_e32 v1, 0xff, v12
; SI-NEXT: v_add_i32_e32 v0, vcc, 0x7c, v0
-; SI-NEXT: s_waitcnt vmcnt(1)
+; SI-NEXT: s_waitcnt vmcnt(0)
; SI-NEXT: v_lshlrev_b32_e32 v2, 8, v2
; SI-NEXT: v_or_b32_e32 v1, v1, v2
-; SI-NEXT: v_and_b32_e32 v2, 0xff, v27
+; SI-NEXT: v_and_b32_e32 v2, 0xff, v3
+; SI-NEXT: buffer_load_dword v3, off, s[0:3], s32 offset:220 ; 4-byte Folded Reload
; SI-NEXT: v_lshlrev_b32_e32 v2, 16, v2
+; SI-NEXT: v_and_b32_e32 v1, 0xffff, v1
; SI-NEXT: s_waitcnt vmcnt(0)
; SI-NEXT: v_lshlrev_b32_e32 v3, 24, v3
-; SI-NEXT: v_and_b32_e32 v1, 0xffff, v1
; SI-NEXT: v_or_b32_e32 v2, v3, v2
; SI-NEXT: v_or_b32_e32 v1, v1, v2
; SI-NEXT: buffer_store_dword v1, v0, s[0:3], 0 offen
@@ -197757,8 +197857,8 @@ define inreg <128 x i8> @bitcast_v64f16_to_v128i8_scalar(<64 x half> inreg %a, i
; VI: ; %bb.0:
; VI-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; VI-NEXT: s_or_saveexec_b64 s[4:5], -1
-; VI-NEXT: buffer_store_dword v63, off, s[0:3], s32 offset:340 ; 4-byte Folded Spill
-; VI-NEXT: buffer_store_dword v62, off, s[0:3], s32 offset:344 ; 4-byte Folded Spill
+; VI-NEXT: buffer_store_dword v63, off, s[0:3], s32 offset:336 ; 4-byte Folded Spill
+; VI-NEXT: buffer_store_dword v62, off, s[0:3], s32 offset:340 ; 4-byte Folded Spill
; VI-NEXT: s_mov_b64 exec, s[4:5]
; VI-NEXT: v_writelane_b32 v63, s30, 0
; VI-NEXT: v_writelane_b32 v63, s31, 1
@@ -197988,112 +198088,117 @@ define inreg <128 x i8> @bitcast_v64f16_to_v128i8_scalar(<64 x half> inreg %a, i
; VI-NEXT: s_lshr_b32 s46, s45, 16
; VI-NEXT: v_mov_b32_e32 v7, 0x200
; VI-NEXT: v_add_f16_e32 v1, s46, v7
-; VI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:336 ; 4-byte Folded Spill
+; VI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:60 ; 4-byte Folded Spill
; VI-NEXT: v_lshlrev_b32_e32 v1, 16, v1
; VI-NEXT: v_add_f16_e32 v2, s45, v7
; VI-NEXT: s_lshr_b32 s45, s44, 16
; VI-NEXT: v_or_b32_e32 v23, v2, v1
; VI-NEXT: v_add_f16_e32 v1, s45, v7
-; VI-NEXT: buffer_store_dword v2, off, s[0:3], s32 offset:64 ; 4-byte Folded Spill
-; VI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:176 ; 4-byte Folded Spill
+; VI-NEXT: buffer_store_dword v2, off, s[0:3], s32 offset:88 ; 4-byte Folded Spill
+; VI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:152 ; 4-byte Folded Spill
; VI-NEXT: v_lshlrev_b32_e32 v1, 16, v1
; VI-NEXT: v_add_f16_e32 v2, s44, v7
; VI-NEXT: s_lshr_b32 s44, s43, 16
; VI-NEXT: v_or_b32_e32 v22, v2, v1
; VI-NEXT: v_add_f16_e32 v1, s44, v7
-; VI-NEXT: buffer_store_dword v2, off, s[0:3], s32 offset:96 ; 4-byte Folded Spill
-; VI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:56 ; 4-byte Folded Spill
+; VI-NEXT: buffer_store_dword v2, off, s[0:3], s32 offset:120 ; 4-byte Folded Spill
+; VI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:64 ; 4-byte Folded Spill
; VI-NEXT: v_lshlrev_b32_e32 v1, 16, v1
; VI-NEXT: v_add_f16_e32 v2, s43, v7
; VI-NEXT: s_lshr_b32 s43, s42, 16
; VI-NEXT: v_or_b32_e32 v25, v2, v1
; VI-NEXT: v_add_f16_e32 v1, s43, v7
-; VI-NEXT: buffer_store_dword v2, off, s[0:3], s32 offset:68 ; 4-byte Folded Spill
-; VI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:136 ; 4-byte Folded Spill
+; VI-NEXT: buffer_store_dword v2, off, s[0:3], s32 offset:92 ; 4-byte Folded Spill
+; VI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:156 ; 4-byte Folded Spill
; VI-NEXT: v_lshlrev_b32_e32 v1, 16, v1
; VI-NEXT: v_add_f16_e32 v2, s42, v7
; VI-NEXT: s_lshr_b32 s42, s41, 16
; VI-NEXT: v_or_b32_e32 v24, v2, v1
; VI-NEXT: v_add_f16_e32 v1, s42, v7
-; VI-NEXT: buffer_store_dword v2, off, s[0:3], s32 offset:100 ; 4-byte Folded Spill
-; VI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:60 ; 4-byte Folded Spill
+; VI-NEXT: buffer_store_dword v2, off, s[0:3], s32 offset:124 ; 4-byte Folded Spill
+; VI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:68 ; 4-byte Folded Spill
; VI-NEXT: v_lshlrev_b32_e32 v1, 16, v1
; VI-NEXT: v_add_f16_e32 v2, s41, v7
; VI-NEXT: s_lshr_b32 s41, s40, 16
-; VI-NEXT: buffer_store_dword v2, off, s[0:3], s32 offset:72 ; 4-byte Folded Spill
; VI-NEXT: v_or_b32_e32 v27, v2, v1
; VI-NEXT: v_add_f16_e32 v1, s41, v7
+; VI-NEXT: buffer_store_dword v2, off, s[0:3], s32 offset:96 ; 4-byte Folded Spill
+; VI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:160 ; 4-byte Folded Spill
+; VI-NEXT: v_lshlrev_b32_e32 v1, 16, v1
; VI-NEXT: v_add_f16_e32 v2, s40, v7
; VI-NEXT: s_lshr_b32 s40, s15, 16
-; VI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:140 ; 4-byte Folded Spill
-; VI-NEXT: v_lshlrev_b32_e32 v1, 16, v1
-; VI-NEXT: v_add_f16_e32 v53, s40, v7
-; VI-NEXT: buffer_store_dword v2, off, s[0:3], s32 offset:104 ; 4-byte Folded Spill
; VI-NEXT: v_or_b32_e32 v26, v2, v1
-; VI-NEXT: v_lshlrev_b32_e32 v1, 16, v53
+; VI-NEXT: v_add_f16_e32 v1, s40, v7
+; VI-NEXT: buffer_store_dword v2, off, s[0:3], s32 offset:128 ; 4-byte Folded Spill
+; VI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:72 ; 4-byte Folded Spill
+; VI-NEXT: v_lshlrev_b32_e32 v1, 16, v1
; VI-NEXT: v_add_f16_e32 v2, s15, v7
; VI-NEXT: s_lshr_b32 s15, s14, 16
-; VI-NEXT: buffer_store_dword v2, off, s[0:3], s32 offset:76 ; 4-byte Folded Spill
; VI-NEXT: v_or_b32_e32 v29, v2, v1
; VI-NEXT: v_add_f16_e32 v1, s15, v7
+; VI-NEXT: buffer_store_dword v2, off, s[0:3], s32 offset:100 ; 4-byte Folded Spill
+; VI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:164 ; 4-byte Folded Spill
+; VI-NEXT: v_lshlrev_b32_e32 v1, 16, v1
; VI-NEXT: v_add_f16_e32 v2, s14, v7
; VI-NEXT: s_lshr_b32 s14, s13, 16
-; VI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:144 ; 4-byte Folded Spill
-; VI-NEXT: v_lshlrev_b32_e32 v1, 16, v1
-; VI-NEXT: v_add_f16_e32 v43, s14, v7
-; VI-NEXT: buffer_store_dword v2, off, s[0:3], s32 offset:108 ; 4-byte Folded Spill
; VI-NEXT: v_or_b32_e32 v28, v2, v1
-; VI-NEXT: v_lshlrev_b32_e32 v1, 16, v43
+; VI-NEXT: v_add_f16_e32 v1, s14, v7
+; VI-NEXT: buffer_store_dword v2, off, s[0:3], s32 offset:132 ; 4-byte Folded Spill
+; VI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:76 ; 4-byte Folded Spill
+; VI-NEXT: v_lshlrev_b32_e32 v1, 16, v1
; VI-NEXT: v_add_f16_e32 v2, s13, v7
; VI-NEXT: s_lshr_b32 s13, s12, 16
-; VI-NEXT: buffer_store_dword v2, off, s[0:3], s32 offset:80 ; 4-byte Folded Spill
; VI-NEXT: v_or_b32_e32 v6, v2, v1
; VI-NEXT: v_add_f16_e32 v1, s13, v7
+; VI-NEXT: buffer_store_dword v2, off, s[0:3], s32 offset:104 ; 4-byte Folded Spill
+; VI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:168 ; 4-byte Folded Spill
+; VI-NEXT: v_lshlrev_b32_e32 v1, 16, v1
; VI-NEXT: v_add_f16_e32 v2, s12, v7
; VI-NEXT: s_lshr_b32 s12, s11, 16
-; VI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:148 ; 4-byte Folded Spill
-; VI-NEXT: v_lshlrev_b32_e32 v1, 16, v1
-; VI-NEXT: v_add_f16_e32 v37, s12, v7
-; VI-NEXT: buffer_store_dword v2, off, s[0:3], s32 offset:112 ; 4-byte Folded Spill
; VI-NEXT: v_or_b32_e32 v5, v2, v1
-; VI-NEXT: v_lshlrev_b32_e32 v1, 16, v37
+; VI-NEXT: v_add_f16_e32 v1, s12, v7
+; VI-NEXT: buffer_store_dword v2, off, s[0:3], s32 offset:136 ; 4-byte Folded Spill
+; VI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:80 ; 4-byte Folded Spill
+; VI-NEXT: v_lshlrev_b32_e32 v1, 16, v1
; VI-NEXT: v_add_f16_e32 v2, s11, v7
; VI-NEXT: s_lshr_b32 s11, s10, 16
-; VI-NEXT: buffer_store_dword v2, off, s[0:3], s32 offset:84 ; 4-byte Folded Spill
; VI-NEXT: v_or_b32_e32 v31, v2, v1
; VI-NEXT: v_add_f16_e32 v1, s11, v7
+; VI-NEXT: buffer_store_dword v2, off, s[0:3], s32 offset:108 ; 4-byte Folded Spill
+; VI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:172 ; 4-byte Folded Spill
+; VI-NEXT: v_lshlrev_b32_e32 v1, 16, v1
; VI-NEXT: v_add_f16_e32 v2, s10, v7
; VI-NEXT: s_lshr_b32 s10, s9, 16
-; VI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:152 ; 4-byte Folded Spill
-; VI-NEXT: v_lshlrev_b32_e32 v1, 16, v1
-; VI-NEXT: v_add_f16_e32 v52, s10, v7
-; VI-NEXT: buffer_store_dword v2, off, s[0:3], s32 offset:116 ; 4-byte Folded Spill
; VI-NEXT: v_or_b32_e32 v30, v2, v1
-; VI-NEXT: v_lshlrev_b32_e32 v1, 16, v52
+; VI-NEXT: v_add_f16_e32 v1, s10, v7
+; VI-NEXT: buffer_store_dword v2, off, s[0:3], s32 offset:140 ; 4-byte Folded Spill
+; VI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:84 ; 4-byte Folded Spill
+; VI-NEXT: v_lshlrev_b32_e32 v1, 16, v1
; VI-NEXT: v_add_f16_e32 v2, s9, v7
; VI-NEXT: s_lshr_b32 s9, s8, 16
-; VI-NEXT: buffer_store_dword v2, off, s[0:3], s32 offset:88 ; 4-byte Folded Spill
; VI-NEXT: v_or_b32_e32 v4, v2, v1
; VI-NEXT: v_add_f16_e32 v1, s9, v7
+; VI-NEXT: buffer_store_dword v2, off, s[0:3], s32 offset:112 ; 4-byte Folded Spill
+; VI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:176 ; 4-byte Folded Spill
+; VI-NEXT: v_lshlrev_b32_e32 v1, 16, v1
; VI-NEXT: v_add_f16_e32 v2, s8, v7
; VI-NEXT: s_lshr_b32 s8, s7, 16
-; VI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:156 ; 4-byte Folded Spill
-; VI-NEXT: v_lshlrev_b32_e32 v1, 16, v1
-; VI-NEXT: v_add_f16_e32 v50, s8, v7
-; VI-NEXT: buffer_store_dword v2, off, s[0:3], s32 offset:120 ; 4-byte Folded Spill
; VI-NEXT: v_or_b32_e32 v3, v2, v1
-; VI-NEXT: v_lshlrev_b32_e32 v1, 16, v50
+; VI-NEXT: v_add_f16_e32 v1, s8, v7
+; VI-NEXT: buffer_store_dword v2, off, s[0:3], s32 offset:144 ; 4-byte Folded Spill
+; VI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:56 ; 4-byte Folded Spill
+; VI-NEXT: v_lshlrev_b32_e32 v1, 16, v1
; VI-NEXT: v_add_f16_e32 v2, s7, v7
; VI-NEXT: s_lshr_b32 s7, s6, 16
-; VI-NEXT: buffer_store_dword v2, off, s[0:3], s32 offset:92 ; 4-byte Folded Spill
+; VI-NEXT: buffer_store_dword v2, off, s[0:3], s32 offset:116 ; 4-byte Folded Spill
; VI-NEXT: v_or_b32_e32 v2, v2, v1
; VI-NEXT: v_add_f16_e32 v1, s7, v7
; VI-NEXT: v_add_f16_e32 v8, s6, v7
; VI-NEXT: s_lshr_b32 s6, s17, 16
-; VI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:160 ; 4-byte Folded Spill
+; VI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:180 ; 4-byte Folded Spill
; VI-NEXT: v_lshlrev_b32_e32 v1, 16, v1
; VI-NEXT: v_add_f16_e32 v36, s6, v7
-; VI-NEXT: buffer_store_dword v8, off, s[0:3], s32 offset:124 ; 4-byte Folded Spill
+; VI-NEXT: buffer_store_dword v8, off, s[0:3], s32 offset:148 ; 4-byte Folded Spill
; VI-NEXT: v_or_b32_e32 v1, v8, v1
; VI-NEXT: v_lshlrev_b32_e32 v8, 16, v36
; VI-NEXT: v_add_f16_e32 v9, s17, v7
@@ -198101,12 +198206,12 @@ define inreg <128 x i8> @bitcast_v64f16_to_v128i8_scalar(<64 x half> inreg %a, i
; VI-NEXT: v_or_b32_e32 v33, v9, v8
; VI-NEXT: v_add_f16_e32 v8, s6, v7
; VI-NEXT: s_lshr_b32 s6, s19, 16
-; VI-NEXT: buffer_store_dword v9, off, s[0:3], s32 offset:164 ; 4-byte Folded Spill
-; VI-NEXT: buffer_store_dword v8, off, s[0:3], s32 offset:220 ; 4-byte Folded Spill
+; VI-NEXT: buffer_store_dword v9, off, s[0:3], s32 offset:184 ; 4-byte Folded Spill
+; VI-NEXT: buffer_store_dword v8, off, s[0:3], s32 offset:236 ; 4-byte Folded Spill
; VI-NEXT: v_lshlrev_b32_e32 v8, 16, v8
; VI-NEXT: v_add_f16_e32 v9, s16, v7
; VI-NEXT: v_add_f16_e32 v38, s6, v7
-; VI-NEXT: buffer_store_dword v9, off, s[0:3], s32 offset:172 ; 4-byte Folded Spill
+; VI-NEXT: buffer_store_dword v9, off, s[0:3], s32 offset:192 ; 4-byte Folded Spill
; VI-NEXT: v_or_b32_e32 v32, v9, v8
; VI-NEXT: v_lshlrev_b32_e32 v8, 16, v38
; VI-NEXT: v_add_f16_e32 v9, s19, v7
@@ -198114,12 +198219,12 @@ define inreg <128 x i8> @bitcast_v64f16_to_v128i8_scalar(<64 x half> inreg %a, i
; VI-NEXT: v_or_b32_e32 v21, v9, v8
; VI-NEXT: v_add_f16_e32 v8, s6, v7
; VI-NEXT: s_lshr_b32 s6, s21, 16
-; VI-NEXT: buffer_store_dword v9, off, s[0:3], s32 offset:168 ; 4-byte Folded Spill
-; VI-NEXT: buffer_store_dword v8, off, s[0:3], s32 offset:228 ; 4-byte Folded Spill
+; VI-NEXT: buffer_store_dword v9, off, s[0:3], s32 offset:188 ; 4-byte Folded Spill
+; VI-NEXT: buffer_store_dword v8, off, s[0:3], s32 offset:244 ; 4-byte Folded Spill
; VI-NEXT: v_lshlrev_b32_e32 v8, 16, v8
; VI-NEXT: v_add_f16_e32 v9, s18, v7
; VI-NEXT: v_add_f16_e32 v61, s6, v7
-; VI-NEXT: buffer_store_dword v9, off, s[0:3], s32 offset:188 ; 4-byte Folded Spill
+; VI-NEXT: buffer_store_dword v9, off, s[0:3], s32 offset:204 ; 4-byte Folded Spill
; VI-NEXT: v_or_b32_e32 v20, v9, v8
; VI-NEXT: s_lshr_b32 s7, s20, 16
; VI-NEXT: v_lshlrev_b32_e32 v8, 16, v61
@@ -198127,12 +198232,12 @@ define inreg <128 x i8> @bitcast_v64f16_to_v128i8_scalar(<64 x half> inreg %a, i
; VI-NEXT: v_or_b32_e32 v35, v9, v8
; VI-NEXT: v_add_f16_e32 v8, s7, v7
; VI-NEXT: s_lshr_b32 s6, s23, 16
-; VI-NEXT: buffer_store_dword v9, off, s[0:3], s32 offset:180 ; 4-byte Folded Spill
-; VI-NEXT: buffer_store_dword v8, off, s[0:3], s32 offset:232 ; 4-byte Folded Spill
+; VI-NEXT: buffer_store_dword v9, off, s[0:3], s32 offset:196 ; 4-byte Folded Spill
+; VI-NEXT: buffer_store_dword v8, off, s[0:3], s32 offset:248 ; 4-byte Folded Spill
; VI-NEXT: v_lshlrev_b32_e32 v8, 16, v8
; VI-NEXT: v_add_f16_e32 v9, s20, v7
; VI-NEXT: v_add_f16_e32 v45, s6, v7
-; VI-NEXT: buffer_store_dword v9, off, s[0:3], s32 offset:196 ; 4-byte Folded Spill
+; VI-NEXT: buffer_store_dword v9, off, s[0:3], s32 offset:212 ; 4-byte Folded Spill
; VI-NEXT: v_or_b32_e32 v34, v9, v8
; VI-NEXT: s_lshr_b32 s7, s22, 16
; VI-NEXT: v_lshlrev_b32_e32 v8, 16, v45
@@ -198140,12 +198245,12 @@ define inreg <128 x i8> @bitcast_v64f16_to_v128i8_scalar(<64 x half> inreg %a, i
; VI-NEXT: v_or_b32_e32 v19, v9, v8
; VI-NEXT: v_add_f16_e32 v8, s7, v7
; VI-NEXT: s_lshr_b32 s6, s25, 16
-; VI-NEXT: buffer_store_dword v9, off, s[0:3], s32 offset:184 ; 4-byte Folded Spill
-; VI-NEXT: buffer_store_dword v8, off, s[0:3], s32 offset:236 ; 4-byte Folded Spill
+; VI-NEXT: buffer_store_dword v9, off, s[0:3], s32 offset:200 ; 4-byte Folded Spill
+; VI-NEXT: buffer_store_dword v8, off, s[0:3], s32 offset:252 ; 4-byte Folded Spill
; VI-NEXT: v_lshlrev_b32_e32 v8, 16, v8
; VI-NEXT: v_add_f16_e32 v9, s22, v7
; VI-NEXT: v_add_f16_e32 v47, s6, v7
-; VI-NEXT: buffer_store_dword v9, off, s[0:3], s32 offset:204 ; 4-byte Folded Spill
+; VI-NEXT: buffer_store_dword v9, off, s[0:3], s32 offset:220 ; 4-byte Folded Spill
; VI-NEXT: v_or_b32_e32 v18, v9, v8
; VI-NEXT: s_lshr_b32 s7, s24, 16
; VI-NEXT: v_lshlrev_b32_e32 v8, 16, v47
@@ -198153,12 +198258,12 @@ define inreg <128 x i8> @bitcast_v64f16_to_v128i8_scalar(<64 x half> inreg %a, i
; VI-NEXT: v_or_b32_e32 v16, v9, v8
; VI-NEXT: v_add_f16_e32 v8, s7, v7
; VI-NEXT: s_lshr_b32 s6, s27, 16
-; VI-NEXT: buffer_store_dword v9, off, s[0:3], s32 offset:192 ; 4-byte Folded Spill
-; VI-NEXT: buffer_store_dword v8, off, s[0:3], s32 offset:240 ; 4-byte Folded Spill
+; VI-NEXT: buffer_store_dword v9, off, s[0:3], s32 offset:208 ; 4-byte Folded Spill
+; VI-NEXT: buffer_store_dword v8, off, s[0:3], s32 offset:256 ; 4-byte Folded Spill
; VI-NEXT: v_lshlrev_b32_e32 v8, 16, v8
; VI-NEXT: v_add_f16_e32 v9, s24, v7
; VI-NEXT: v_add_f16_e32 v57, s6, v7
-; VI-NEXT: buffer_store_dword v9, off, s[0:3], s32 offset:212 ; 4-byte Folded Spill
+; VI-NEXT: buffer_store_dword v9, off, s[0:3], s32 offset:228 ; 4-byte Folded Spill
; VI-NEXT: v_or_b32_e32 v15, v9, v8
; VI-NEXT: s_lshr_b32 s7, s26, 16
; VI-NEXT: v_lshlrev_b32_e32 v8, 16, v57
@@ -198166,12 +198271,12 @@ define inreg <128 x i8> @bitcast_v64f16_to_v128i8_scalar(<64 x half> inreg %a, i
; VI-NEXT: v_or_b32_e32 v13, v9, v8
; VI-NEXT: v_add_f16_e32 v8, s7, v7
; VI-NEXT: s_lshr_b32 s6, s29, 16
-; VI-NEXT: buffer_store_dword v9, off, s[0:3], s32 offset:200 ; 4-byte Folded Spill
-; VI-NEXT: buffer_store_dword v8, off, s[0:3], s32 offset:244 ; 4-byte Folded Spill
+; VI-NEXT: buffer_store_dword v9, off, s[0:3], s32 offset:216 ; 4-byte Folded Spill
+; VI-NEXT: buffer_store_dword v8, off, s[0:3], s32 offset:260 ; 4-byte Folded Spill
; VI-NEXT: v_lshlrev_b32_e32 v8, 16, v8
; VI-NEXT: v_add_f16_e32 v9, s26, v7
; VI-NEXT: v_add_f16_e32 v59, s6, v7
-; VI-NEXT: buffer_store_dword v9, off, s[0:3], s32 offset:216 ; 4-byte Folded Spill
+; VI-NEXT: buffer_store_dword v9, off, s[0:3], s32 offset:232 ; 4-byte Folded Spill
; VI-NEXT: v_or_b32_e32 v12, v9, v8
; VI-NEXT: s_lshr_b32 s7, s28, 16
; VI-NEXT: v_lshlrev_b32_e32 v8, 16, v59
@@ -198181,96 +198286,80 @@ define inreg <128 x i8> @bitcast_v64f16_to_v128i8_scalar(<64 x half> inreg %a, i
; VI-NEXT: v_add_f16_e32 v8, s7, v7
; VI-NEXT: s_lshr_b32 s7, s4, 16
; VI-NEXT: v_add_f16_e32 v51, s6, v7
-; VI-NEXT: buffer_store_dword v9, off, s[0:3], s32 offset:208 ; 4-byte Folded Spill
-; VI-NEXT: buffer_store_dword v8, off, s[0:3], s32 offset:248 ; 4-byte Folded Spill
+; VI-NEXT: buffer_store_dword v9, off, s[0:3], s32 offset:224 ; 4-byte Folded Spill
+; VI-NEXT: buffer_store_dword v8, off, s[0:3], s32 offset:264 ; 4-byte Folded Spill
; VI-NEXT: v_lshlrev_b32_e32 v8, 16, v8
; VI-NEXT: v_add_f16_e32 v9, s28, v7
; VI-NEXT: v_add_f16_e32 v54, s5, v7
-; VI-NEXT: v_add_f16_e32 v11, s7, v7
+; VI-NEXT: v_add_f16_e32 v53, s7, v7
; VI-NEXT: v_add_f16_e32 v55, s4, v7
; VI-NEXT: v_lshlrev_b32_e32 v7, 16, v51
-; VI-NEXT: buffer_store_dword v9, off, s[0:3], s32 offset:224 ; 4-byte Folded Spill
+; VI-NEXT: buffer_store_dword v9, off, s[0:3], s32 offset:240 ; 4-byte Folded Spill
; VI-NEXT: v_or_b32_e32 v9, v9, v8
; VI-NEXT: v_or_b32_e32 v8, v54, v7
-; VI-NEXT: v_lshlrev_b32_e32 v7, 16, v11
-; VI-NEXT: buffer_store_dword v11, off, s[0:3], s32 offset:252 ; 4-byte Folded Spill
+; VI-NEXT: v_lshlrev_b32_e32 v7, 16, v53
; VI-NEXT: v_or_b32_e32 v7, v55, v7
; VI-NEXT: v_lshrrev_b32_e32 v11, 8, v8
-; VI-NEXT: buffer_store_dword v11, off, s[0:3], s32 offset:316 ; 4-byte Folded Spill
-; VI-NEXT: v_lshrrev_b32_e32 v11, 8, v7
+; VI-NEXT: v_lshrrev_b32_e32 v41, 8, v7
; VI-NEXT: v_lshrrev_b64 v[7:8], 24, v[7:8]
-; VI-NEXT: v_lshrrev_b32_e32 v8, 8, v10
-; VI-NEXT: buffer_store_dword v8, off, s[0:3], s32 offset:324 ; 4-byte Folded Spill
-; VI-NEXT: v_lshrrev_b32_e32 v8, 8, v9
-; VI-NEXT: buffer_store_dword v8, off, s[0:3], s32 offset:328 ; 4-byte Folded Spill
-; VI-NEXT: v_lshrrev_b32_e32 v8, 8, v13
-; VI-NEXT: buffer_store_dword v8, off, s[0:3], s32 offset:332 ; 4-byte Folded Spill
; VI-NEXT: v_lshrrev_b32_e32 v8, 8, v2
-; VI-NEXT: buffer_store_dword v8, off, s[0:3], s32 offset:260 ; 4-byte Folded Spill
+; VI-NEXT: buffer_store_dword v8, off, s[0:3], s32 offset:272 ; 4-byte Folded Spill
; VI-NEXT: v_lshrrev_b32_e32 v8, 8, v1
; VI-NEXT: v_lshrrev_b64 v[1:2], 24, v[1:2]
-; VI-NEXT: buffer_store_dword v11, off, s[0:3], s32 offset:320 ; 4-byte Folded Spill
-; VI-NEXT: buffer_store_dword v8, off, s[0:3], s32 offset:268 ; 4-byte Folded Spill
-; VI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:292 ; 4-byte Folded Spill
-; VI-NEXT: buffer_store_dword v2, off, s[0:3], s32 offset:296 ; 4-byte Folded Spill
+; VI-NEXT: buffer_store_dword v11, off, s[0:3], s32 offset:332 ; 4-byte Folded Spill
+; VI-NEXT: buffer_store_dword v8, off, s[0:3], s32 offset:280 ; 4-byte Folded Spill
+; VI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:304 ; 4-byte Folded Spill
+; VI-NEXT: buffer_store_dword v2, off, s[0:3], s32 offset:308 ; 4-byte Folded Spill
; VI-NEXT: v_lshrrev_b32_e32 v1, 8, v4
+; VI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:288 ; 4-byte Folded Spill
+; VI-NEXT: v_lshrrev_b32_e32 v1, 8, v3
+; VI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:292 ; 4-byte Folded Spill
+; VI-NEXT: v_lshrrev_b64 v[1:2], 24, v[3:4]
; VI-NEXT: v_lshrrev_b32_e32 v39, 8, v16
; VI-NEXT: v_lshrrev_b64 v[16:17], 24, v[15:16]
-; VI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:276 ; 4-byte Folded Spill
-; VI-NEXT: v_lshrrev_b32_e32 v1, 8, v3
+; VI-NEXT: v_lshrrev_b32_e32 v37, 8, v13
; VI-NEXT: v_lshrrev_b64 v[13:14], 24, v[12:13]
-; VI-NEXT: v_lshrrev_b32_e32 v44, 8, v18
+; VI-NEXT: v_lshrrev_b32_e32 v46, 8, v18
; VI-NEXT: v_lshrrev_b64 v[17:18], 24, v[18:19]
-; VI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:280 ; 4-byte Folded Spill
-; VI-NEXT: v_lshrrev_b64 v[1:2], 24, v[3:4]
-; VI-NEXT: v_lshrrev_b32_e32 v49, 8, v15
-; VI-NEXT: v_lshrrev_b32_e32 v40, 8, v19
-; VI-NEXT: v_lshrrev_b32_e32 v60, 8, v35
+; VI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:320 ; 4-byte Folded Spill
+; VI-NEXT: buffer_store_dword v2, off, s[0:3], s32 offset:324 ; 4-byte Folded Spill
+; VI-NEXT: v_lshrrev_b32_e32 v60, 8, v15
+; VI-NEXT: v_lshrrev_b32_e32 v49, 8, v19
+; VI-NEXT: v_lshrrev_b32_e32 v44, 8, v35
; VI-NEXT: v_lshrrev_b64 v[18:19], 24, v[34:35]
; VI-NEXT: v_lshrrev_b32_e32 v35, 8, v23
; VI-NEXT: v_lshrrev_b64 v[14:15], 24, v[22:23]
-; VI-NEXT: v_bfe_u32 v23, v50, 8, 8
-; VI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:308 ; 4-byte Folded Spill
-; VI-NEXT: buffer_store_dword v2, off, s[0:3], s32 offset:312 ; 4-byte Folded Spill
-; VI-NEXT: buffer_store_dword v23, off, s[0:3], s32 offset:256 ; 4-byte Folded Spill
-; VI-NEXT: v_bfe_u32 v23, v52, 8, 8
+; VI-NEXT: buffer_load_dword v23, off, s[0:3], s32 offset:56 ; 4-byte Folded Reload
; VI-NEXT: v_lshrrev_b32_e32 v1, 8, v31
-; VI-NEXT: buffer_store_dword v23, off, s[0:3], s32 offset:264 ; 4-byte Folded Spill
-; VI-NEXT: v_bfe_u32 v23, v37, 8, 8
-; VI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:288 ; 4-byte Folded Spill
-; VI-NEXT: v_lshrrev_b32_e32 v1, 8, v30
-; VI-NEXT: buffer_store_dword v23, off, s[0:3], s32 offset:272 ; 4-byte Folded Spill
-; VI-NEXT: v_bfe_u32 v23, v43, 8, 8
-; VI-NEXT: v_lshrrev_b64 v[10:11], 24, v[9:10]
-; VI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:300 ; 4-byte Folded Spill
-; VI-NEXT: buffer_store_dword v50, off, s[0:3], s32 offset:128 ; 4-byte Folded Spill
-; VI-NEXT: buffer_load_dword v50, off, s[0:3], s32 offset:336 ; 4-byte Folded Reload
-; VI-NEXT: buffer_store_dword v23, off, s[0:3], s32 offset:284 ; 4-byte Folded Spill
-; VI-NEXT: buffer_store_dword v53, off, s[0:3], s32 offset:132 ; 4-byte Folded Spill
-; VI-NEXT: v_bfe_u32 v23, v53, 8, 8
-; VI-NEXT: v_lshrrev_b32_e32 v48, 8, v12
-; VI-NEXT: v_lshrrev_b32_e32 v42, 8, v34
-; VI-NEXT: v_lshrrev_b32_e32 v34, 8, v24
-; VI-NEXT: v_lshrrev_b64 v[11:12], 24, v[24:25]
-; VI-NEXT: buffer_store_dword v23, off, s[0:3], s32 offset:304 ; 4-byte Folded Spill
-; VI-NEXT: buffer_load_dword v23, off, s[0:3], s32 offset:60 ; 4-byte Folded Reload
-; VI-NEXT: buffer_load_dword v24, off, s[0:3], s32 offset:56 ; 4-byte Folded Reload
; VI-NEXT: v_lshrrev_b64 v[3:4], 24, v[30:31]
-; VI-NEXT: v_lshrrev_b32_e32 v56, 8, v20
+; VI-NEXT: v_lshrrev_b32_e32 v42, 8, v10
+; VI-NEXT: v_lshrrev_b64 v[10:11], 24, v[9:10]
+; VI-NEXT: v_lshrrev_b32_e32 v40, 8, v20
; VI-NEXT: v_lshrrev_b64 v[19:20], 24, v[20:21]
+; VI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:300 ; 4-byte Folded Spill
+; VI-NEXT: v_lshrrev_b32_e32 v1, 8, v30
; VI-NEXT: v_lshrrev_b32_e32 v31, 8, v5
; VI-NEXT: v_lshrrev_b64 v[4:5], 24, v[5:6]
-; VI-NEXT: v_lshrrev_b32_e32 v41, 8, v21
+; VI-NEXT: v_lshrrev_b32_e32 v43, 8, v9
+; VI-NEXT: v_lshrrev_b32_e32 v48, 8, v12
+; VI-NEXT: v_lshrrev_b32_e32 v50, 8, v34
+; VI-NEXT: v_lshrrev_b32_e32 v56, 8, v21
+; VI-NEXT: v_lshrrev_b32_e32 v52, 8, v32
; VI-NEXT: v_lshrrev_b64 v[20:21], 24, v[32:33]
-; VI-NEXT: v_lshrrev_b32_e32 v30, 8, v6
+; VI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:312 ; 4-byte Folded Spill
+; VI-NEXT: v_lshrrev_b32_e32 v1, 8, v6
+; VI-NEXT: v_lshrrev_b32_e32 v32, 8, v29
; VI-NEXT: v_lshrrev_b64 v[5:6], 24, v[28:29]
+; VI-NEXT: v_lshrrev_b32_e32 v29, 8, v26
; VI-NEXT: v_lshrrev_b64 v[8:9], 24, v[26:27]
-; VI-NEXT: v_lshrrev_b32_e32 v46, 8, v33
-; VI-NEXT: v_lshrrev_b32_e32 v58, 8, v32
-; VI-NEXT: v_lshrrev_b32_e32 v32, 8, v29
+; VI-NEXT: v_lshrrev_b32_e32 v34, 8, v24
+; VI-NEXT: v_lshrrev_b64 v[11:12], 24, v[24:25]
+; VI-NEXT: buffer_load_dword v24, off, s[0:3], s32 offset:64 ; 4-byte Folded Reload
+; VI-NEXT: buffer_load_dword v26, off, s[0:3], s32 offset:60 ; 4-byte Folded Reload
+; VI-NEXT: v_lshrrev_b32_e32 v58, 8, v33
+; VI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:328 ; 4-byte Folded Spill
; VI-NEXT: v_lshrrev_b32_e32 v33, 8, v28
; VI-NEXT: v_lshrrev_b32_e32 v28, 8, v27
-; VI-NEXT: v_lshrrev_b32_e32 v29, 8, v26
; VI-NEXT: v_lshrrev_b32_e32 v9, 8, v25
; VI-NEXT: v_lshrrev_b32_e32 v21, 8, v22
; VI-NEXT: v_bfe_u32 v25, v51, 8, 8
@@ -198282,11 +198371,31 @@ define inreg <128 x i8> @bitcast_v64f16_to_v128i8_scalar(<64 x half> inreg %a, i
; VI-NEXT: v_bfe_u32 v22, v38, 8, 8
; VI-NEXT: v_bfe_u32 v2, v36, 8, 8
; VI-NEXT: s_waitcnt vmcnt(5)
-; VI-NEXT: v_bfe_u32 v26, v50, 8, 8
-; VI-NEXT: s_waitcnt vmcnt(1)
; VI-NEXT: v_bfe_u32 v23, v23, 8, 8
-; VI-NEXT: s_waitcnt vmcnt(0)
+; VI-NEXT: buffer_store_dword v23, off, s[0:3], s32 offset:268 ; 4-byte Folded Spill
+; VI-NEXT: buffer_load_dword v23, off, s[0:3], s32 offset:84 ; 4-byte Folded Reload
+; VI-NEXT: s_waitcnt vmcnt(4)
; VI-NEXT: v_bfe_u32 v24, v24, 8, 8
+; VI-NEXT: s_waitcnt vmcnt(3)
+; VI-NEXT: v_bfe_u32 v26, v26, 8, 8
+; VI-NEXT: s_waitcnt vmcnt(0)
+; VI-NEXT: v_bfe_u32 v23, v23, 8, 8
+; VI-NEXT: buffer_store_dword v23, off, s[0:3], s32 offset:276 ; 4-byte Folded Spill
+; VI-NEXT: buffer_load_dword v23, off, s[0:3], s32 offset:80 ; 4-byte Folded Reload
+; VI-NEXT: s_waitcnt vmcnt(0)
+; VI-NEXT: v_bfe_u32 v23, v23, 8, 8
+; VI-NEXT: buffer_store_dword v23, off, s[0:3], s32 offset:284 ; 4-byte Folded Spill
+; VI-NEXT: buffer_load_dword v23, off, s[0:3], s32 offset:76 ; 4-byte Folded Reload
+; VI-NEXT: s_waitcnt vmcnt(0)
+; VI-NEXT: v_bfe_u32 v23, v23, 8, 8
+; VI-NEXT: buffer_store_dword v23, off, s[0:3], s32 offset:296 ; 4-byte Folded Spill
+; VI-NEXT: buffer_load_dword v23, off, s[0:3], s32 offset:72 ; 4-byte Folded Reload
+; VI-NEXT: s_waitcnt vmcnt(0)
+; VI-NEXT: v_bfe_u32 v23, v23, 8, 8
+; VI-NEXT: buffer_store_dword v23, off, s[0:3], s32 offset:316 ; 4-byte Folded Spill
+; VI-NEXT: buffer_load_dword v23, off, s[0:3], s32 offset:68 ; 4-byte Folded Reload
+; VI-NEXT: s_waitcnt vmcnt(0)
+; VI-NEXT: v_bfe_u32 v23, v23, 8, 8
; VI-NEXT: s_branch .LBB95_5
; VI-NEXT: .LBB95_3:
; VI-NEXT: ; implicit-def: $sgpr46
@@ -198446,133 +198555,136 @@ define inreg <128 x i8> @bitcast_v64f16_to_v128i8_scalar(<64 x half> inreg %a, i
; VI-NEXT: s_branch .LBB95_2
; VI-NEXT: .LBB95_4:
; VI-NEXT: v_mov_b32_e32 v1, s44
-; VI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:96 ; 4-byte Folded Spill
+; VI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:120 ; 4-byte Folded Spill
; VI-NEXT: v_mov_b32_e32 v1, s45
-; VI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:64 ; 4-byte Folded Spill
+; VI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:88 ; 4-byte Folded Spill
; VI-NEXT: v_mov_b32_e32 v1, s42
-; VI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:100 ; 4-byte Folded Spill
+; VI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:124 ; 4-byte Folded Spill
; VI-NEXT: v_mov_b32_e32 v1, s43
-; VI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:68 ; 4-byte Folded Spill
+; VI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:92 ; 4-byte Folded Spill
; VI-NEXT: v_mov_b32_e32 v1, s40
-; VI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:104 ; 4-byte Folded Spill
+; VI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:128 ; 4-byte Folded Spill
; VI-NEXT: v_mov_b32_e32 v1, s41
-; VI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:72 ; 4-byte Folded Spill
+; VI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:96 ; 4-byte Folded Spill
; VI-NEXT: v_mov_b32_e32 v1, s14
-; VI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:108 ; 4-byte Folded Spill
+; VI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:132 ; 4-byte Folded Spill
; VI-NEXT: v_mov_b32_e32 v1, s15
-; VI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:76 ; 4-byte Folded Spill
+; VI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:100 ; 4-byte Folded Spill
; VI-NEXT: v_mov_b32_e32 v1, s12
-; VI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:112 ; 4-byte Folded Spill
+; VI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:136 ; 4-byte Folded Spill
; VI-NEXT: v_mov_b32_e32 v1, s13
-; VI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:80 ; 4-byte Folded Spill
+; VI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:104 ; 4-byte Folded Spill
; VI-NEXT: v_mov_b32_e32 v1, s10
-; VI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:116 ; 4-byte Folded Spill
+; VI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:140 ; 4-byte Folded Spill
; VI-NEXT: v_mov_b32_e32 v1, s11
-; VI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:84 ; 4-byte Folded Spill
+; VI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:108 ; 4-byte Folded Spill
; VI-NEXT: v_mov_b32_e32 v1, s8
-; VI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:120 ; 4-byte Folded Spill
+; VI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:144 ; 4-byte Folded Spill
; VI-NEXT: v_mov_b32_e32 v1, s9
-; VI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:88 ; 4-byte Folded Spill
+; VI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:112 ; 4-byte Folded Spill
; VI-NEXT: v_mov_b32_e32 v1, s6
-; VI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:124 ; 4-byte Folded Spill
+; VI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:148 ; 4-byte Folded Spill
; VI-NEXT: v_mov_b32_e32 v1, s7
-; VI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:92 ; 4-byte Folded Spill
+; VI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:116 ; 4-byte Folded Spill
; VI-NEXT: v_mov_b32_e32 v1, s16
-; VI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:172 ; 4-byte Folded Spill
+; VI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:192 ; 4-byte Folded Spill
; VI-NEXT: v_mov_b32_e32 v1, s17
-; VI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:164 ; 4-byte Folded Spill
+; VI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:184 ; 4-byte Folded Spill
; VI-NEXT: v_mov_b32_e32 v1, s18
-; VI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:188 ; 4-byte Folded Spill
+; VI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:204 ; 4-byte Folded Spill
; VI-NEXT: v_mov_b32_e32 v1, s19
-; VI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:168 ; 4-byte Folded Spill
+; VI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:188 ; 4-byte Folded Spill
; VI-NEXT: v_mov_b32_e32 v1, s20
-; VI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:196 ; 4-byte Folded Spill
+; VI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:212 ; 4-byte Folded Spill
; VI-NEXT: v_mov_b32_e32 v1, s21
-; VI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:180 ; 4-byte Folded Spill
+; VI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:196 ; 4-byte Folded Spill
; VI-NEXT: v_mov_b32_e32 v1, s22
-; VI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:204 ; 4-byte Folded Spill
+; VI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:220 ; 4-byte Folded Spill
; VI-NEXT: v_mov_b32_e32 v1, s23
-; VI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:184 ; 4-byte Folded Spill
+; VI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:200 ; 4-byte Folded Spill
; VI-NEXT: v_mov_b32_e32 v1, s24
-; VI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:212 ; 4-byte Folded Spill
+; VI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:228 ; 4-byte Folded Spill
; VI-NEXT: v_mov_b32_e32 v1, s25
-; VI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:192 ; 4-byte Folded Spill
+; VI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:208 ; 4-byte Folded Spill
; VI-NEXT: v_mov_b32_e32 v1, s26
-; VI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:216 ; 4-byte Folded Spill
+; VI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:232 ; 4-byte Folded Spill
; VI-NEXT: v_mov_b32_e32 v1, s27
-; VI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:200 ; 4-byte Folded Spill
+; VI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:216 ; 4-byte Folded Spill
; VI-NEXT: v_mov_b32_e32 v1, s28
-; VI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:224 ; 4-byte Folded Spill
+; VI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:240 ; 4-byte Folded Spill
; VI-NEXT: v_mov_b32_e32 v1, s29
-; VI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:208 ; 4-byte Folded Spill
+; VI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:224 ; 4-byte Folded Spill
; VI-NEXT: v_mov_b32_e32 v1, s71
-; VI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:176 ; 4-byte Folded Spill
+; VI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:152 ; 4-byte Folded Spill
+; VI-NEXT: v_mov_b32_e32 v1, s70
+; VI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:60 ; 4-byte Folded Spill
; VI-NEXT: v_mov_b32_e32 v1, s69
-; VI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:136 ; 4-byte Folded Spill
+; VI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:156 ; 4-byte Folded Spill
; VI-NEXT: v_mov_b32_e32 v1, s68
-; VI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:56 ; 4-byte Folded Spill
+; VI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:64 ; 4-byte Folded Spill
; VI-NEXT: v_mov_b32_e32 v1, s67
-; VI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:140 ; 4-byte Folded Spill
+; VI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:160 ; 4-byte Folded Spill
; VI-NEXT: v_mov_b32_e32 v1, s66
-; VI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:60 ; 4-byte Folded Spill
+; VI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:68 ; 4-byte Folded Spill
; VI-NEXT: v_mov_b32_e32 v1, s65
-; VI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:144 ; 4-byte Folded Spill
+; VI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:164 ; 4-byte Folded Spill
; VI-NEXT: v_mov_b32_e32 v1, s64
-; VI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:132 ; 4-byte Folded Spill
+; VI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:72 ; 4-byte Folded Spill
; VI-NEXT: v_mov_b32_e32 v1, s55
-; VI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:148 ; 4-byte Folded Spill
+; VI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:168 ; 4-byte Folded Spill
+; VI-NEXT: v_mov_b32_e32 v1, s54
+; VI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:76 ; 4-byte Folded Spill
; VI-NEXT: v_mov_b32_e32 v1, s87
-; VI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:152 ; 4-byte Folded Spill
+; VI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:172 ; 4-byte Folded Spill
+; VI-NEXT: v_mov_b32_e32 v1, s86
+; VI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:80 ; 4-byte Folded Spill
; VI-NEXT: v_mov_b32_e32 v1, s85
-; VI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:156 ; 4-byte Folded Spill
+; VI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:176 ; 4-byte Folded Spill
+; VI-NEXT: v_mov_b32_e32 v1, s84
+; VI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:84 ; 4-byte Folded Spill
; VI-NEXT: v_mov_b32_e32 v1, s53
-; VI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:160 ; 4-byte Folded Spill
+; VI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:180 ; 4-byte Folded Spill
; VI-NEXT: v_mov_b32_e32 v1, s52
; VI-NEXT: v_readlane_b32 s6, v62, 0
-; VI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:128 ; 4-byte Folded Spill
+; VI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:56 ; 4-byte Folded Spill
; VI-NEXT: v_mov_b32_e32 v1, s6
; VI-NEXT: v_readlane_b32 s6, v62, 1
; VI-NEXT: v_mov_b32_e32 v36, s6
; VI-NEXT: v_readlane_b32 s6, v62, 2
-; VI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:220 ; 4-byte Folded Spill
+; VI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:236 ; 4-byte Folded Spill
; VI-NEXT: v_mov_b32_e32 v1, s6
; VI-NEXT: v_readlane_b32 s6, v62, 3
; VI-NEXT: v_mov_b32_e32 v38, s6
; VI-NEXT: v_readlane_b32 s6, v62, 4
-; VI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:228 ; 4-byte Folded Spill
+; VI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:244 ; 4-byte Folded Spill
; VI-NEXT: v_mov_b32_e32 v1, s6
; VI-NEXT: v_readlane_b32 s6, v62, 5
; VI-NEXT: v_mov_b32_e32 v61, s6
; VI-NEXT: v_readlane_b32 s6, v62, 6
-; VI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:232 ; 4-byte Folded Spill
+; VI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:248 ; 4-byte Folded Spill
; VI-NEXT: v_mov_b32_e32 v1, s6
; VI-NEXT: v_readlane_b32 s6, v62, 7
; VI-NEXT: v_mov_b32_e32 v45, s6
; VI-NEXT: v_readlane_b32 s6, v62, 8
-; VI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:236 ; 4-byte Folded Spill
+; VI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:252 ; 4-byte Folded Spill
; VI-NEXT: v_mov_b32_e32 v1, s6
; VI-NEXT: v_readlane_b32 s6, v62, 9
; VI-NEXT: v_mov_b32_e32 v47, s6
; VI-NEXT: v_readlane_b32 s6, v62, 10
-; VI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:240 ; 4-byte Folded Spill
+; VI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:256 ; 4-byte Folded Spill
; VI-NEXT: v_mov_b32_e32 v1, s6
; VI-NEXT: v_readlane_b32 s6, v62, 11
-; VI-NEXT: v_mov_b32_e32 v57, s6
-; VI-NEXT: v_readlane_b32 s6, v62, 12
-; VI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:244 ; 4-byte Folded Spill
-; VI-NEXT: v_mov_b32_e32 v1, s6
-; VI-NEXT: v_readlane_b32 s6, v62, 13
; VI-NEXT: v_mov_b32_e32 v55, s4
; VI-NEXT: v_readlane_b32 s4, v62, 16
-; VI-NEXT: v_mov_b32_e32 v59, s6
-; VI-NEXT: v_readlane_b32 s6, v62, 14
+; VI-NEXT: v_mov_b32_e32 v57, s6
+; VI-NEXT: v_readlane_b32 s6, v62, 12
; VI-NEXT: v_mov_b32_e32 v2, s4
; VI-NEXT: v_readlane_b32 s4, v62, 17
-; VI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:248 ; 4-byte Folded Spill
+; VI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:260 ; 4-byte Folded Spill
; VI-NEXT: v_mov_b32_e32 v1, s6
; VI-NEXT: v_mov_b32_e32 v22, s4
; VI-NEXT: v_readlane_b32 s4, v62, 18
-; VI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:252 ; 4-byte Folded Spill
+; VI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:264 ; 4-byte Folded Spill
; VI-NEXT: v_mov_b32_e32 v1, s4
; VI-NEXT: v_readlane_b32 s4, v62, 19
; VI-NEXT: v_mov_b32_e32 v15, s4
@@ -198587,77 +198699,74 @@ define inreg <128 x i8> @bitcast_v64f16_to_v128i8_scalar(<64 x half> inreg %a, i
; VI-NEXT: v_readlane_b32 s4, v62, 24
; VI-NEXT: v_mov_b32_e32 v3, s4
; VI-NEXT: v_readlane_b32 s4, v62, 25
-; VI-NEXT: buffer_store_dword v3, off, s[0:3], s32 offset:256 ; 4-byte Folded Spill
+; VI-NEXT: buffer_store_dword v3, off, s[0:3], s32 offset:268 ; 4-byte Folded Spill
; VI-NEXT: v_mov_b32_e32 v3, s4
; VI-NEXT: v_readlane_b32 s4, v62, 26
-; VI-NEXT: v_mov_b32_e32 v46, s4
+; VI-NEXT: v_mov_b32_e32 v58, s4
; VI-NEXT: v_readlane_b32 s4, v62, 27
-; VI-NEXT: v_mov_b32_e32 v41, s4
+; VI-NEXT: v_mov_b32_e32 v56, s4
; VI-NEXT: v_readlane_b32 s4, v62, 28
-; VI-NEXT: buffer_store_dword v3, off, s[0:3], s32 offset:264 ; 4-byte Folded Spill
-; VI-NEXT: v_mov_b32_e32 v3, s80
-; VI-NEXT: v_mov_b32_e32 v60, s4
+; VI-NEXT: v_mov_b32_e32 v44, s4
; VI-NEXT: v_readlane_b32 s4, v62, 29
-; VI-NEXT: buffer_store_dword v3, off, s[0:3], s32 offset:272 ; 4-byte Folded Spill
-; VI-NEXT: v_mov_b32_e32 v3, s81
-; VI-NEXT: v_mov_b32_e32 v40, s4
+; VI-NEXT: v_mov_b32_e32 v49, s4
; VI-NEXT: v_readlane_b32 s4, v62, 30
-; VI-NEXT: buffer_store_dword v3, off, s[0:3], s32 offset:284 ; 4-byte Folded Spill
-; VI-NEXT: v_mov_b32_e32 v3, s82
+; VI-NEXT: buffer_store_dword v3, off, s[0:3], s32 offset:276 ; 4-byte Folded Spill
+; VI-NEXT: v_mov_b32_e32 v3, s80
; VI-NEXT: v_mov_b32_e32 v39, s4
; VI-NEXT: v_readlane_b32 s4, v62, 31
-; VI-NEXT: buffer_store_dword v3, off, s[0:3], s32 offset:304 ; 4-byte Folded Spill
-; VI-NEXT: v_mov_b32_e32 v3, s4
+; VI-NEXT: buffer_store_dword v3, off, s[0:3], s32 offset:284 ; 4-byte Folded Spill
+; VI-NEXT: v_mov_b32_e32 v3, s81
+; VI-NEXT: v_mov_b32_e32 v37, s4
; VI-NEXT: v_readlane_b32 s4, v62, 32
-; VI-NEXT: buffer_store_dword v3, off, s[0:3], s32 offset:332 ; 4-byte Folded Spill
-; VI-NEXT: v_mov_b32_e32 v3, s4
+; VI-NEXT: buffer_store_dword v3, off, s[0:3], s32 offset:296 ; 4-byte Folded Spill
+; VI-NEXT: v_mov_b32_e32 v3, s82
+; VI-NEXT: v_mov_b32_e32 v42, s4
; VI-NEXT: v_readlane_b32 s4, v62, 33
-; VI-NEXT: buffer_store_dword v3, off, s[0:3], s32 offset:324 ; 4-byte Folded Spill
-; VI-NEXT: v_mov_b32_e32 v3, s4
-; VI-NEXT: v_readlane_b32 s4, v62, 34
; VI-NEXT: buffer_store_dword v3, off, s[0:3], s32 offset:316 ; 4-byte Folded Spill
; VI-NEXT: v_mov_b32_e32 v3, s4
+; VI-NEXT: v_readlane_b32 s4, v62, 34
+; VI-NEXT: v_mov_b32_e32 v41, s4
; VI-NEXT: v_readlane_b32 s4, v62, 35
-; VI-NEXT: buffer_store_dword v3, off, s[0:3], s32 offset:320 ; 4-byte Folded Spill
-; VI-NEXT: v_mov_b32_e32 v3, s4
+; VI-NEXT: v_mov_b32_e32 v43, s4
; VI-NEXT: v_readlane_b32 s4, v62, 36
; VI-NEXT: v_mov_b32_e32 v48, s4
; VI-NEXT: v_readlane_b32 s4, v62, 37
-; VI-NEXT: v_mov_b32_e32 v49, s4
+; VI-NEXT: v_mov_b32_e32 v60, s4
; VI-NEXT: v_readlane_b32 s4, v62, 38
-; VI-NEXT: v_mov_b32_e32 v44, s4
+; VI-NEXT: v_mov_b32_e32 v46, s4
; VI-NEXT: v_readlane_b32 s4, v62, 39
-; VI-NEXT: v_mov_b32_e32 v42, s4
+; VI-NEXT: v_mov_b32_e32 v50, s4
; VI-NEXT: v_readlane_b32 s4, v62, 40
-; VI-NEXT: v_mov_b32_e32 v56, s4
+; VI-NEXT: v_mov_b32_e32 v40, s4
; VI-NEXT: v_readlane_b32 s4, v62, 41
-; VI-NEXT: v_mov_b32_e32 v58, s4
+; VI-NEXT: v_mov_b32_e32 v52, s4
; VI-NEXT: v_readlane_b32 s4, v62, 42
-; VI-NEXT: buffer_store_dword v3, off, s[0:3], s32 offset:328 ; 4-byte Folded Spill
+; VI-NEXT: buffer_store_dword v3, off, s[0:3], s32 offset:332 ; 4-byte Folded Spill
; VI-NEXT: v_mov_b32_e32 v3, s4
; VI-NEXT: v_readlane_b32 s4, v62, 43
-; VI-NEXT: buffer_store_dword v3, off, s[0:3], s32 offset:268 ; 4-byte Folded Spill
+; VI-NEXT: buffer_store_dword v3, off, s[0:3], s32 offset:280 ; 4-byte Folded Spill
; VI-NEXT: v_mov_b32_e32 v3, s4
; VI-NEXT: v_readlane_b32 s4, v62, 44
-; VI-NEXT: buffer_store_dword v3, off, s[0:3], s32 offset:260 ; 4-byte Folded Spill
+; VI-NEXT: buffer_store_dword v3, off, s[0:3], s32 offset:272 ; 4-byte Folded Spill
; VI-NEXT: v_mov_b32_e32 v3, s4
; VI-NEXT: v_readlane_b32 s4, v62, 45
-; VI-NEXT: buffer_store_dword v3, off, s[0:3], s32 offset:280 ; 4-byte Folded Spill
+; VI-NEXT: buffer_store_dword v3, off, s[0:3], s32 offset:292 ; 4-byte Folded Spill
; VI-NEXT: v_mov_b32_e32 v3, s4
; VI-NEXT: v_readlane_b32 s4, v62, 46
-; VI-NEXT: buffer_store_dword v3, off, s[0:3], s32 offset:276 ; 4-byte Folded Spill
+; VI-NEXT: buffer_store_dword v3, off, s[0:3], s32 offset:288 ; 4-byte Folded Spill
; VI-NEXT: v_mov_b32_e32 v3, s4
; VI-NEXT: v_readlane_b32 s4, v62, 47
-; VI-NEXT: buffer_store_dword v3, off, s[0:3], s32 offset:300 ; 4-byte Folded Spill
+; VI-NEXT: buffer_store_dword v3, off, s[0:3], s32 offset:312 ; 4-byte Folded Spill
; VI-NEXT: v_mov_b32_e32 v3, s4
-; VI-NEXT: buffer_store_dword v3, off, s[0:3], s32 offset:288 ; 4-byte Folded Spill
-; VI-NEXT: v_mov_b32_e32 v3, s78
-; VI-NEXT: buffer_store_dword v3, off, s[0:3], s32 offset:292 ; 4-byte Folded Spill
-; VI-NEXT: buffer_store_dword v4, off, s[0:3], s32 offset:296 ; 4-byte Folded Spill
; VI-NEXT: v_readlane_b32 s4, v62, 48
; VI-NEXT: v_mov_b32_e32 v31, s4
; VI-NEXT: v_readlane_b32 s4, v62, 49
-; VI-NEXT: v_mov_b32_e32 v30, s4
+; VI-NEXT: buffer_store_dword v3, off, s[0:3], s32 offset:300 ; 4-byte Folded Spill
+; VI-NEXT: v_mov_b32_e32 v3, s4
+; VI-NEXT: buffer_store_dword v3, off, s[0:3], s32 offset:328 ; 4-byte Folded Spill
+; VI-NEXT: v_mov_b32_e32 v3, s78
+; VI-NEXT: buffer_store_dword v3, off, s[0:3], s32 offset:304 ; 4-byte Folded Spill
+; VI-NEXT: buffer_store_dword v4, off, s[0:3], s32 offset:308 ; 4-byte Folded Spill
; VI-NEXT: v_readlane_b32 s4, v62, 50
; VI-NEXT: v_mov_b32_e32 v33, s4
; VI-NEXT: v_readlane_b32 s4, v62, 51
@@ -198667,20 +198776,20 @@ define inreg <128 x i8> @bitcast_v64f16_to_v128i8_scalar(<64 x half> inreg %a, i
; VI-NEXT: v_readlane_b32 s4, v62, 53
; VI-NEXT: v_mov_b32_e32 v28, s4
; VI-NEXT: v_readlane_b32 s4, v62, 54
+; VI-NEXT: v_readlane_b32 s6, v62, 13
; VI-NEXT: v_mov_b32_e32 v34, s4
; VI-NEXT: v_readlane_b32 s4, v62, 55
+; VI-NEXT: v_mov_b32_e32 v59, s6
+; VI-NEXT: v_readlane_b32 s6, v62, 14
; VI-NEXT: v_mov_b32_e32 v9, s4
; VI-NEXT: v_readlane_b32 s4, v62, 56
; VI-NEXT: v_mov_b32_e32 v3, s88
+; VI-NEXT: v_mov_b32_e32 v53, s6
; VI-NEXT: v_readlane_b32 s6, v62, 15
; VI-NEXT: v_mov_b32_e32 v21, s4
; VI-NEXT: v_readlane_b32 s4, v62, 57
-; VI-NEXT: buffer_store_dword v3, off, s[0:3], s32 offset:308 ; 4-byte Folded Spill
-; VI-NEXT: buffer_store_dword v4, off, s[0:3], s32 offset:312 ; 4-byte Folded Spill
-; VI-NEXT: v_mov_b32_e32 v50, s70
-; VI-NEXT: v_mov_b32_e32 v43, s54
-; VI-NEXT: v_mov_b32_e32 v37, s86
-; VI-NEXT: v_mov_b32_e32 v52, s84
+; VI-NEXT: buffer_store_dword v3, off, s[0:3], s32 offset:320 ; 4-byte Folded Spill
+; VI-NEXT: buffer_store_dword v4, off, s[0:3], s32 offset:324 ; 4-byte Folded Spill
; VI-NEXT: v_mov_b32_e32 v51, s6
; VI-NEXT: v_mov_b32_e32 v54, s5
; VI-NEXT: v_mov_b32_e32 v23, s83
@@ -198702,8 +198811,8 @@ define inreg <128 x i8> @bitcast_v64f16_to_v128i8_scalar(<64 x half> inreg %a, i
; VI-NEXT: v_mov_b32_e32 v11, s38
; VI-NEXT: v_mov_b32_e32 v14, s48
; VI-NEXT: .LBB95_5: ; %end
-; VI-NEXT: buffer_load_dword v53, off, s[0:3], s32 offset:172 ; 4-byte Folded Reload
-; VI-NEXT: v_lshlrev_b32_e32 v58, 8, v58
+; VI-NEXT: buffer_load_dword v30, off, s[0:3], s32 offset:192 ; 4-byte Folded Reload
+; VI-NEXT: v_lshlrev_b32_e32 v52, 8, v52
; VI-NEXT: v_lshlrev_b32_e32 v20, 8, v20
; VI-NEXT: v_lshlrev_b32_e32 v2, 8, v2
; VI-NEXT: v_or_b32_sdwa v2, v36, v2 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
@@ -198744,31 +198853,31 @@ define inreg <128 x i8> @bitcast_v64f16_to_v128i8_scalar(<64 x half> inreg %a, i
; VI-NEXT: v_readlane_b32 s31, v63, 1
; VI-NEXT: v_readlane_b32 s30, v63, 0
; VI-NEXT: s_waitcnt vmcnt(0)
-; VI-NEXT: v_or_b32_sdwa v58, v53, v58 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
-; VI-NEXT: buffer_load_dword v53, off, s[0:3], s32 offset:220 ; 4-byte Folded Reload
+; VI-NEXT: v_or_b32_sdwa v52, v30, v52 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; VI-NEXT: buffer_load_dword v30, off, s[0:3], s32 offset:236 ; 4-byte Folded Reload
; VI-NEXT: s_waitcnt vmcnt(0)
-; VI-NEXT: v_or_b32_sdwa v20, v53, v20 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
-; VI-NEXT: v_or_b32_sdwa v20, v58, v20 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_0 src1_sel:DWORD
+; VI-NEXT: v_or_b32_sdwa v20, v30, v20 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; VI-NEXT: v_or_b32_sdwa v20, v52, v20 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_0 src1_sel:DWORD
; VI-NEXT: buffer_store_dword v20, v0, s[0:3], 0 offen
-; VI-NEXT: v_lshlrev_b32_e32 v20, 8, v46
-; VI-NEXT: buffer_load_dword v46, off, s[0:3], s32 offset:164 ; 4-byte Folded Reload
+; VI-NEXT: buffer_load_dword v52, off, s[0:3], s32 offset:184 ; 4-byte Folded Reload
+; VI-NEXT: v_lshlrev_b32_e32 v20, 8, v58
; VI-NEXT: s_waitcnt vmcnt(0)
-; VI-NEXT: v_or_b32_sdwa v20, v46, v20 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; VI-NEXT: v_or_b32_sdwa v20, v52, v20 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
; VI-NEXT: v_or_b32_sdwa v2, v20, v2 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_0 src1_sel:DWORD
; VI-NEXT: v_add_u32_e32 v20, vcc, 4, v0
; VI-NEXT: buffer_store_dword v2, v20, s[0:3], 0 offen
-; VI-NEXT: buffer_load_dword v20, off, s[0:3], s32 offset:188 ; 4-byte Folded Reload
-; VI-NEXT: v_lshlrev_b32_e32 v2, 8, v56
+; VI-NEXT: buffer_load_dword v20, off, s[0:3], s32 offset:204 ; 4-byte Folded Reload
+; VI-NEXT: v_lshlrev_b32_e32 v2, 8, v40
; VI-NEXT: s_waitcnt vmcnt(0)
; VI-NEXT: v_or_b32_sdwa v2, v20, v2 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
-; VI-NEXT: buffer_load_dword v20, off, s[0:3], s32 offset:228 ; 4-byte Folded Reload
+; VI-NEXT: buffer_load_dword v20, off, s[0:3], s32 offset:244 ; 4-byte Folded Reload
; VI-NEXT: s_waitcnt vmcnt(0)
; VI-NEXT: v_or_b32_sdwa v19, v20, v19 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
; VI-NEXT: v_or_b32_sdwa v2, v2, v19 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_0 src1_sel:DWORD
; VI-NEXT: v_add_u32_e32 v19, vcc, 8, v0
; VI-NEXT: buffer_store_dword v2, v19, s[0:3], 0 offen
-; VI-NEXT: buffer_load_dword v19, off, s[0:3], s32 offset:168 ; 4-byte Folded Reload
-; VI-NEXT: v_lshlrev_b32_e32 v2, 8, v41
+; VI-NEXT: buffer_load_dword v19, off, s[0:3], s32 offset:188 ; 4-byte Folded Reload
+; VI-NEXT: v_lshlrev_b32_e32 v2, 8, v56
; VI-NEXT: s_waitcnt vmcnt(0)
; VI-NEXT: v_or_b32_sdwa v2, v19, v2 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
; VI-NEXT: v_lshlrev_b32_e32 v19, 8, v22
@@ -198776,36 +198885,36 @@ define inreg <128 x i8> @bitcast_v64f16_to_v128i8_scalar(<64 x half> inreg %a, i
; VI-NEXT: v_or_b32_sdwa v2, v2, v19 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_0 src1_sel:DWORD
; VI-NEXT: v_add_u32_e32 v19, vcc, 12, v0
; VI-NEXT: buffer_store_dword v2, v19, s[0:3], 0 offen
-; VI-NEXT: buffer_load_dword v19, off, s[0:3], s32 offset:196 ; 4-byte Folded Reload
-; VI-NEXT: v_lshlrev_b32_e32 v2, 8, v42
+; VI-NEXT: buffer_load_dword v19, off, s[0:3], s32 offset:212 ; 4-byte Folded Reload
+; VI-NEXT: v_lshlrev_b32_e32 v2, 8, v50
; VI-NEXT: s_waitcnt vmcnt(0)
; VI-NEXT: v_or_b32_sdwa v2, v19, v2 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
-; VI-NEXT: buffer_load_dword v19, off, s[0:3], s32 offset:232 ; 4-byte Folded Reload
+; VI-NEXT: buffer_load_dword v19, off, s[0:3], s32 offset:248 ; 4-byte Folded Reload
; VI-NEXT: s_waitcnt vmcnt(0)
; VI-NEXT: v_or_b32_sdwa v18, v19, v18 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
; VI-NEXT: v_or_b32_sdwa v2, v2, v18 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_0 src1_sel:DWORD
; VI-NEXT: v_add_u32_e32 v18, vcc, 16, v0
; VI-NEXT: buffer_store_dword v2, v18, s[0:3], 0 offen
-; VI-NEXT: buffer_load_dword v18, off, s[0:3], s32 offset:180 ; 4-byte Folded Reload
-; VI-NEXT: v_lshlrev_b32_e32 v2, 8, v60
+; VI-NEXT: buffer_load_dword v18, off, s[0:3], s32 offset:196 ; 4-byte Folded Reload
+; VI-NEXT: v_lshlrev_b32_e32 v2, 8, v44
; VI-NEXT: s_waitcnt vmcnt(0)
; VI-NEXT: v_or_b32_sdwa v2, v18, v2 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
; VI-NEXT: v_or_b32_sdwa v1, v2, v1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_0 src1_sel:DWORD
; VI-NEXT: v_add_u32_e32 v2, vcc, 20, v0
; VI-NEXT: buffer_store_dword v1, v2, s[0:3], 0 offen
-; VI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:204 ; 4-byte Folded Reload
-; VI-NEXT: v_lshlrev_b32_e32 v1, 8, v44
+; VI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:220 ; 4-byte Folded Reload
+; VI-NEXT: v_lshlrev_b32_e32 v1, 8, v46
; VI-NEXT: s_waitcnt vmcnt(0)
; VI-NEXT: v_or_b32_sdwa v1, v2, v1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
; VI-NEXT: v_lshlrev_b32_e32 v2, 8, v17
-; VI-NEXT: buffer_load_dword v17, off, s[0:3], s32 offset:236 ; 4-byte Folded Reload
+; VI-NEXT: buffer_load_dword v17, off, s[0:3], s32 offset:252 ; 4-byte Folded Reload
; VI-NEXT: s_waitcnt vmcnt(0)
; VI-NEXT: v_or_b32_sdwa v2, v17, v2 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
; VI-NEXT: v_or_b32_sdwa v1, v1, v2 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_0 src1_sel:DWORD
; VI-NEXT: v_add_u32_e32 v2, vcc, 24, v0
; VI-NEXT: buffer_store_dword v1, v2, s[0:3], 0 offen
-; VI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:184 ; 4-byte Folded Reload
-; VI-NEXT: v_lshlrev_b32_e32 v1, 8, v40
+; VI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:200 ; 4-byte Folded Reload
+; VI-NEXT: v_lshlrev_b32_e32 v1, 8, v49
; VI-NEXT: s_waitcnt vmcnt(0)
; VI-NEXT: v_or_b32_sdwa v1, v2, v1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
; VI-NEXT: v_lshlrev_b32_e32 v2, 8, v15
@@ -198813,9 +198922,9 @@ define inreg <128 x i8> @bitcast_v64f16_to_v128i8_scalar(<64 x half> inreg %a, i
; VI-NEXT: v_or_b32_sdwa v1, v1, v2 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_0 src1_sel:DWORD
; VI-NEXT: v_add_u32_e32 v2, vcc, 28, v0
; VI-NEXT: buffer_store_dword v1, v2, s[0:3], 0 offen
-; VI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:212 ; 4-byte Folded Reload
-; VI-NEXT: buffer_load_dword v15, off, s[0:3], s32 offset:240 ; 4-byte Folded Reload
-; VI-NEXT: v_lshlrev_b32_e32 v1, 8, v49
+; VI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:228 ; 4-byte Folded Reload
+; VI-NEXT: buffer_load_dword v15, off, s[0:3], s32 offset:256 ; 4-byte Folded Reload
+; VI-NEXT: v_lshlrev_b32_e32 v1, 8, v60
; VI-NEXT: s_waitcnt vmcnt(1)
; VI-NEXT: v_or_b32_sdwa v1, v2, v1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
; VI-NEXT: v_lshlrev_b32_e32 v2, 8, v16
@@ -198824,7 +198933,7 @@ define inreg <128 x i8> @bitcast_v64f16_to_v128i8_scalar(<64 x half> inreg %a, i
; VI-NEXT: v_or_b32_sdwa v1, v1, v2 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_0 src1_sel:DWORD
; VI-NEXT: v_add_u32_e32 v2, vcc, 32, v0
; VI-NEXT: buffer_store_dword v1, v2, s[0:3], 0 offen
-; VI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:192 ; 4-byte Folded Reload
+; VI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:208 ; 4-byte Folded Reload
; VI-NEXT: v_lshlrev_b32_e32 v1, 8, v39
; VI-NEXT: s_waitcnt vmcnt(0)
; VI-NEXT: v_or_b32_sdwa v1, v2, v1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
@@ -198833,8 +198942,8 @@ define inreg <128 x i8> @bitcast_v64f16_to_v128i8_scalar(<64 x half> inreg %a, i
; VI-NEXT: v_or_b32_sdwa v1, v1, v2 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_0 src1_sel:DWORD
; VI-NEXT: v_add_u32_e32 v2, vcc, 36, v0
; VI-NEXT: buffer_store_dword v1, v2, s[0:3], 0 offen
-; VI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:216 ; 4-byte Folded Reload
-; VI-NEXT: buffer_load_dword v12, off, s[0:3], s32 offset:244 ; 4-byte Folded Reload
+; VI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:232 ; 4-byte Folded Reload
+; VI-NEXT: buffer_load_dword v12, off, s[0:3], s32 offset:260 ; 4-byte Folded Reload
; VI-NEXT: v_lshlrev_b32_e32 v1, 8, v48
; VI-NEXT: s_waitcnt vmcnt(1)
; VI-NEXT: v_or_b32_sdwa v1, v2, v1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
@@ -198844,10 +198953,8 @@ define inreg <128 x i8> @bitcast_v64f16_to_v128i8_scalar(<64 x half> inreg %a, i
; VI-NEXT: v_or_b32_sdwa v1, v1, v2 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_0 src1_sel:DWORD
; VI-NEXT: v_add_u32_e32 v2, vcc, 40, v0
; VI-NEXT: buffer_store_dword v1, v2, s[0:3], 0 offen
-; VI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:332 ; 4-byte Folded Reload
-; VI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:200 ; 4-byte Folded Reload
-; VI-NEXT: s_waitcnt vmcnt(1)
-; VI-NEXT: v_lshlrev_b32_e32 v1, 8, v1
+; VI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:216 ; 4-byte Folded Reload
+; VI-NEXT: v_lshlrev_b32_e32 v1, 8, v37
; VI-NEXT: s_waitcnt vmcnt(0)
; VI-NEXT: v_or_b32_sdwa v1, v2, v1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
; VI-NEXT: v_lshlrev_b32_e32 v2, 8, v6
@@ -198855,11 +198962,9 @@ define inreg <128 x i8> @bitcast_v64f16_to_v128i8_scalar(<64 x half> inreg %a, i
; VI-NEXT: v_or_b32_sdwa v1, v1, v2 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_0 src1_sel:DWORD
; VI-NEXT: v_add_u32_e32 v2, vcc, 44, v0
; VI-NEXT: buffer_store_dword v1, v2, s[0:3], 0 offen
-; VI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:328 ; 4-byte Folded Reload
-; VI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:224 ; 4-byte Folded Reload
-; VI-NEXT: buffer_load_dword v6, off, s[0:3], s32 offset:248 ; 4-byte Folded Reload
-; VI-NEXT: s_waitcnt vmcnt(2)
-; VI-NEXT: v_lshlrev_b32_e32 v1, 8, v1
+; VI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:240 ; 4-byte Folded Reload
+; VI-NEXT: buffer_load_dword v6, off, s[0:3], s32 offset:264 ; 4-byte Folded Reload
+; VI-NEXT: v_lshlrev_b32_e32 v1, 8, v43
; VI-NEXT: s_waitcnt vmcnt(1)
; VI-NEXT: v_or_b32_sdwa v1, v2, v1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
; VI-NEXT: v_lshlrev_b32_e32 v2, 8, v10
@@ -198868,10 +198973,8 @@ define inreg <128 x i8> @bitcast_v64f16_to_v128i8_scalar(<64 x half> inreg %a, i
; VI-NEXT: v_or_b32_sdwa v1, v1, v2 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_0 src1_sel:DWORD
; VI-NEXT: v_add_u32_e32 v2, vcc, 48, v0
; VI-NEXT: buffer_store_dword v1, v2, s[0:3], 0 offen
-; VI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:324 ; 4-byte Folded Reload
-; VI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:208 ; 4-byte Folded Reload
-; VI-NEXT: s_waitcnt vmcnt(1)
-; VI-NEXT: v_lshlrev_b32_e32 v1, 8, v1
+; VI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:224 ; 4-byte Folded Reload
+; VI-NEXT: v_lshlrev_b32_e32 v1, 8, v42
; VI-NEXT: s_waitcnt vmcnt(0)
; VI-NEXT: v_or_b32_sdwa v1, v2, v1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
; VI-NEXT: v_lshlrev_b32_e32 v2, 8, v27
@@ -198879,18 +198982,14 @@ define inreg <128 x i8> @bitcast_v64f16_to_v128i8_scalar(<64 x half> inreg %a, i
; VI-NEXT: v_or_b32_sdwa v1, v1, v2 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_0 src1_sel:DWORD
; VI-NEXT: v_add_u32_e32 v2, vcc, 52, v0
; VI-NEXT: buffer_store_dword v1, v2, s[0:3], 0 offen
-; VI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:320 ; 4-byte Folded Reload
-; VI-NEXT: buffer_load_dword v6, off, s[0:3], s32 offset:252 ; 4-byte Folded Reload
+; VI-NEXT: v_lshlrev_b32_e32 v1, 8, v41
; VI-NEXT: v_lshlrev_b32_e32 v2, 8, v7
-; VI-NEXT: s_waitcnt vmcnt(1)
-; VI-NEXT: v_lshlrev_b32_e32 v1, 8, v1
; VI-NEXT: v_or_b32_sdwa v1, v55, v1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
-; VI-NEXT: s_waitcnt vmcnt(0)
-; VI-NEXT: v_or_b32_sdwa v2, v6, v2 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; VI-NEXT: v_or_b32_sdwa v2, v53, v2 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
; VI-NEXT: v_or_b32_sdwa v1, v1, v2 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_0 src1_sel:DWORD
; VI-NEXT: v_add_u32_e32 v2, vcc, 56, v0
; VI-NEXT: buffer_store_dword v1, v2, s[0:3], 0 offen
-; VI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:316 ; 4-byte Folded Reload
+; VI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:332 ; 4-byte Folded Reload
; VI-NEXT: v_lshlrev_b32_e32 v2, 8, v25
; VI-NEXT: v_or_b32_sdwa v2, v51, v2 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
; VI-NEXT: s_waitcnt vmcnt(0)
@@ -198899,8 +198998,8 @@ define inreg <128 x i8> @bitcast_v64f16_to_v128i8_scalar(<64 x half> inreg %a, i
; VI-NEXT: v_or_b32_sdwa v1, v1, v2 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_0 src1_sel:DWORD
; VI-NEXT: v_add_u32_e32 v2, vcc, 60, v0
; VI-NEXT: buffer_store_dword v1, v2, s[0:3], 0 offen
-; VI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:96 ; 4-byte Folded Reload
-; VI-NEXT: buffer_load_dword v6, off, s[0:3], s32 offset:176 ; 4-byte Folded Reload
+; VI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:120 ; 4-byte Folded Reload
+; VI-NEXT: buffer_load_dword v6, off, s[0:3], s32 offset:152 ; 4-byte Folded Reload
; VI-NEXT: v_lshlrev_b32_e32 v1, 8, v21
; VI-NEXT: s_waitcnt vmcnt(1)
; VI-NEXT: v_or_b32_sdwa v1, v2, v1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
@@ -198910,17 +199009,19 @@ define inreg <128 x i8> @bitcast_v64f16_to_v128i8_scalar(<64 x half> inreg %a, i
; VI-NEXT: v_or_b32_sdwa v1, v1, v2 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_0 src1_sel:DWORD
; VI-NEXT: v_add_u32_e32 v2, vcc, 64, v0
; VI-NEXT: buffer_store_dword v1, v2, s[0:3], 0 offen
-; VI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:64 ; 4-byte Folded Reload
+; VI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:88 ; 4-byte Folded Reload
+; VI-NEXT: buffer_load_dword v6, off, s[0:3], s32 offset:60 ; 4-byte Folded Reload
; VI-NEXT: v_lshlrev_b32_e32 v1, 8, v35
-; VI-NEXT: s_waitcnt vmcnt(0)
+; VI-NEXT: s_waitcnt vmcnt(1)
; VI-NEXT: v_or_b32_sdwa v1, v2, v1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
; VI-NEXT: v_lshlrev_b32_e32 v2, 8, v26
-; VI-NEXT: v_or_b32_sdwa v2, v50, v2 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; VI-NEXT: s_waitcnt vmcnt(0)
+; VI-NEXT: v_or_b32_sdwa v2, v6, v2 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
; VI-NEXT: v_or_b32_sdwa v1, v1, v2 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_0 src1_sel:DWORD
; VI-NEXT: v_add_u32_e32 v2, vcc, 0x44, v0
; VI-NEXT: buffer_store_dword v1, v2, s[0:3], 0 offen
-; VI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:100 ; 4-byte Folded Reload
-; VI-NEXT: buffer_load_dword v6, off, s[0:3], s32 offset:136 ; 4-byte Folded Reload
+; VI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:124 ; 4-byte Folded Reload
+; VI-NEXT: buffer_load_dword v6, off, s[0:3], s32 offset:156 ; 4-byte Folded Reload
; VI-NEXT: v_lshlrev_b32_e32 v1, 8, v34
; VI-NEXT: s_waitcnt vmcnt(1)
; VI-NEXT: v_or_b32_sdwa v1, v2, v1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
@@ -198930,8 +199031,8 @@ define inreg <128 x i8> @bitcast_v64f16_to_v128i8_scalar(<64 x half> inreg %a, i
; VI-NEXT: v_or_b32_sdwa v1, v1, v2 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_0 src1_sel:DWORD
; VI-NEXT: v_add_u32_e32 v2, vcc, 0x48, v0
; VI-NEXT: buffer_store_dword v1, v2, s[0:3], 0 offen
-; VI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:68 ; 4-byte Folded Reload
-; VI-NEXT: buffer_load_dword v6, off, s[0:3], s32 offset:56 ; 4-byte Folded Reload
+; VI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:92 ; 4-byte Folded Reload
+; VI-NEXT: buffer_load_dword v6, off, s[0:3], s32 offset:64 ; 4-byte Folded Reload
; VI-NEXT: v_lshlrev_b32_e32 v1, 8, v9
; VI-NEXT: s_waitcnt vmcnt(1)
; VI-NEXT: v_or_b32_sdwa v1, v2, v1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
@@ -198941,8 +199042,8 @@ define inreg <128 x i8> @bitcast_v64f16_to_v128i8_scalar(<64 x half> inreg %a, i
; VI-NEXT: v_or_b32_sdwa v1, v1, v2 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_0 src1_sel:DWORD
; VI-NEXT: v_add_u32_e32 v2, vcc, 0x4c, v0
; VI-NEXT: buffer_store_dword v1, v2, s[0:3], 0 offen
-; VI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:104 ; 4-byte Folded Reload
-; VI-NEXT: buffer_load_dword v6, off, s[0:3], s32 offset:140 ; 4-byte Folded Reload
+; VI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:128 ; 4-byte Folded Reload
+; VI-NEXT: buffer_load_dword v6, off, s[0:3], s32 offset:160 ; 4-byte Folded Reload
; VI-NEXT: v_lshlrev_b32_e32 v1, 8, v29
; VI-NEXT: s_waitcnt vmcnt(1)
; VI-NEXT: v_or_b32_sdwa v1, v2, v1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
@@ -198952,8 +199053,8 @@ define inreg <128 x i8> @bitcast_v64f16_to_v128i8_scalar(<64 x half> inreg %a, i
; VI-NEXT: v_or_b32_sdwa v1, v1, v2 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_0 src1_sel:DWORD
; VI-NEXT: v_add_u32_e32 v2, vcc, 0x50, v0
; VI-NEXT: buffer_store_dword v1, v2, s[0:3], 0 offen
-; VI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:72 ; 4-byte Folded Reload
-; VI-NEXT: buffer_load_dword v6, off, s[0:3], s32 offset:60 ; 4-byte Folded Reload
+; VI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:96 ; 4-byte Folded Reload
+; VI-NEXT: buffer_load_dword v6, off, s[0:3], s32 offset:68 ; 4-byte Folded Reload
; VI-NEXT: v_lshlrev_b32_e32 v1, 8, v28
; VI-NEXT: s_waitcnt vmcnt(1)
; VI-NEXT: v_or_b32_sdwa v1, v2, v1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
@@ -198963,86 +199064,90 @@ define inreg <128 x i8> @bitcast_v64f16_to_v128i8_scalar(<64 x half> inreg %a, i
; VI-NEXT: v_or_b32_sdwa v1, v1, v2 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_0 src1_sel:DWORD
; VI-NEXT: v_add_u32_e32 v2, vcc, 0x54, v0
; VI-NEXT: buffer_store_dword v1, v2, s[0:3], 0 offen
-; VI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:108 ; 4-byte Folded Reload
+; VI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:132 ; 4-byte Folded Reload
; VI-NEXT: v_lshlrev_b32_e32 v1, 8, v33
; VI-NEXT: s_waitcnt vmcnt(0)
; VI-NEXT: v_or_b32_sdwa v1, v2, v1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
; VI-NEXT: v_lshlrev_b32_e32 v2, 8, v5
-; VI-NEXT: buffer_load_dword v5, off, s[0:3], s32 offset:144 ; 4-byte Folded Reload
+; VI-NEXT: buffer_load_dword v5, off, s[0:3], s32 offset:164 ; 4-byte Folded Reload
; VI-NEXT: s_waitcnt vmcnt(0)
; VI-NEXT: v_or_b32_sdwa v2, v5, v2 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
; VI-NEXT: v_or_b32_sdwa v1, v1, v2 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_0 src1_sel:DWORD
; VI-NEXT: v_add_u32_e32 v2, vcc, 0x58, v0
; VI-NEXT: buffer_store_dword v1, v2, s[0:3], 0 offen
-; VI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:76 ; 4-byte Folded Reload
-; VI-NEXT: buffer_load_dword v5, off, s[0:3], s32 offset:132 ; 4-byte Folded Reload
+; VI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:100 ; 4-byte Folded Reload
+; VI-NEXT: buffer_load_dword v5, off, s[0:3], s32 offset:72 ; 4-byte Folded Reload
; VI-NEXT: v_lshlrev_b32_e32 v1, 8, v32
; VI-NEXT: s_waitcnt vmcnt(1)
; VI-NEXT: v_or_b32_sdwa v1, v2, v1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
-; VI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:304 ; 4-byte Folded Reload
+; VI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:316 ; 4-byte Folded Reload
; VI-NEXT: s_waitcnt vmcnt(0)
; VI-NEXT: v_lshlrev_b32_e32 v2, 8, v2
; VI-NEXT: v_or_b32_sdwa v2, v5, v2 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
; VI-NEXT: v_or_b32_sdwa v1, v1, v2 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_0 src1_sel:DWORD
; VI-NEXT: v_add_u32_e32 v2, vcc, 0x5c, v0
; VI-NEXT: buffer_store_dword v1, v2, s[0:3], 0 offen
-; VI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:112 ; 4-byte Folded Reload
+; VI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:136 ; 4-byte Folded Reload
; VI-NEXT: v_lshlrev_b32_e32 v1, 8, v31
; VI-NEXT: s_waitcnt vmcnt(0)
; VI-NEXT: v_or_b32_sdwa v1, v2, v1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
; VI-NEXT: v_lshlrev_b32_e32 v2, 8, v4
-; VI-NEXT: buffer_load_dword v4, off, s[0:3], s32 offset:148 ; 4-byte Folded Reload
+; VI-NEXT: buffer_load_dword v4, off, s[0:3], s32 offset:168 ; 4-byte Folded Reload
; VI-NEXT: s_waitcnt vmcnt(0)
; VI-NEXT: v_or_b32_sdwa v2, v4, v2 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
; VI-NEXT: v_or_b32_sdwa v1, v1, v2 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_0 src1_sel:DWORD
; VI-NEXT: v_add_u32_e32 v2, vcc, 0x60, v0
; VI-NEXT: buffer_store_dword v1, v2, s[0:3], 0 offen
-; VI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:80 ; 4-byte Folded Reload
-; VI-NEXT: v_lshlrev_b32_e32 v1, 8, v30
-; VI-NEXT: s_waitcnt vmcnt(0)
+; VI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:328 ; 4-byte Folded Reload
+; VI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:104 ; 4-byte Folded Reload
+; VI-NEXT: buffer_load_dword v4, off, s[0:3], s32 offset:76 ; 4-byte Folded Reload
+; VI-NEXT: s_waitcnt vmcnt(2)
+; VI-NEXT: v_lshlrev_b32_e32 v1, 8, v1
+; VI-NEXT: s_waitcnt vmcnt(1)
; VI-NEXT: v_or_b32_sdwa v1, v2, v1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
-; VI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:284 ; 4-byte Folded Reload
+; VI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:296 ; 4-byte Folded Reload
; VI-NEXT: s_waitcnt vmcnt(0)
; VI-NEXT: v_lshlrev_b32_e32 v2, 8, v2
-; VI-NEXT: v_or_b32_sdwa v2, v43, v2 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; VI-NEXT: v_or_b32_sdwa v2, v4, v2 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
; VI-NEXT: v_or_b32_sdwa v1, v1, v2 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_0 src1_sel:DWORD
; VI-NEXT: v_add_u32_e32 v2, vcc, 0x64, v0
; VI-NEXT: buffer_store_dword v1, v2, s[0:3], 0 offen
-; VI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:300 ; 4-byte Folded Reload
-; VI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:116 ; 4-byte Folded Reload
+; VI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:312 ; 4-byte Folded Reload
+; VI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:140 ; 4-byte Folded Reload
; VI-NEXT: s_waitcnt vmcnt(1)
; VI-NEXT: v_lshlrev_b32_e32 v1, 8, v1
; VI-NEXT: s_waitcnt vmcnt(0)
; VI-NEXT: v_or_b32_sdwa v1, v2, v1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
; VI-NEXT: v_lshlrev_b32_e32 v2, 8, v3
-; VI-NEXT: buffer_load_dword v3, off, s[0:3], s32 offset:152 ; 4-byte Folded Reload
+; VI-NEXT: buffer_load_dword v3, off, s[0:3], s32 offset:172 ; 4-byte Folded Reload
; VI-NEXT: s_waitcnt vmcnt(0)
; VI-NEXT: v_or_b32_sdwa v2, v3, v2 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
; VI-NEXT: v_or_b32_sdwa v1, v1, v2 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_0 src1_sel:DWORD
; VI-NEXT: v_add_u32_e32 v2, vcc, 0x68, v0
; VI-NEXT: buffer_store_dword v1, v2, s[0:3], 0 offen
-; VI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:288 ; 4-byte Folded Reload
-; VI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:84 ; 4-byte Folded Reload
-; VI-NEXT: s_waitcnt vmcnt(1)
+; VI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:300 ; 4-byte Folded Reload
+; VI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:108 ; 4-byte Folded Reload
+; VI-NEXT: buffer_load_dword v3, off, s[0:3], s32 offset:80 ; 4-byte Folded Reload
+; VI-NEXT: s_waitcnt vmcnt(2)
; VI-NEXT: v_lshlrev_b32_e32 v1, 8, v1
-; VI-NEXT: s_waitcnt vmcnt(0)
+; VI-NEXT: s_waitcnt vmcnt(1)
; VI-NEXT: v_or_b32_sdwa v1, v2, v1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
-; VI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:272 ; 4-byte Folded Reload
+; VI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:284 ; 4-byte Folded Reload
; VI-NEXT: s_waitcnt vmcnt(0)
; VI-NEXT: v_lshlrev_b32_e32 v2, 8, v2
-; VI-NEXT: v_or_b32_sdwa v2, v37, v2 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; VI-NEXT: v_or_b32_sdwa v2, v3, v2 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
; VI-NEXT: v_or_b32_sdwa v1, v1, v2 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_0 src1_sel:DWORD
; VI-NEXT: v_add_u32_e32 v2, vcc, 0x6c, v0
; VI-NEXT: buffer_store_dword v1, v2, s[0:3], 0 offen
-; VI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:280 ; 4-byte Folded Reload
-; VI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:120 ; 4-byte Folded Reload
+; VI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:292 ; 4-byte Folded Reload
+; VI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:144 ; 4-byte Folded Reload
; VI-NEXT: s_waitcnt vmcnt(1)
; VI-NEXT: v_lshlrev_b32_e32 v1, 8, v1
; VI-NEXT: s_waitcnt vmcnt(0)
; VI-NEXT: v_or_b32_sdwa v1, v2, v1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
-; VI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:308 ; 4-byte Folded Reload
-; VI-NEXT: buffer_load_dword v3, off, s[0:3], s32 offset:312 ; 4-byte Folded Reload
-; VI-NEXT: buffer_load_dword v3, off, s[0:3], s32 offset:156 ; 4-byte Folded Reload
+; VI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:320 ; 4-byte Folded Reload
+; VI-NEXT: buffer_load_dword v3, off, s[0:3], s32 offset:324 ; 4-byte Folded Reload
+; VI-NEXT: buffer_load_dword v3, off, s[0:3], s32 offset:176 ; 4-byte Folded Reload
; VI-NEXT: s_waitcnt vmcnt(2)
; VI-NEXT: v_lshlrev_b32_e32 v2, 8, v2
; VI-NEXT: s_waitcnt vmcnt(0)
@@ -199050,28 +199155,29 @@ define inreg <128 x i8> @bitcast_v64f16_to_v128i8_scalar(<64 x half> inreg %a, i
; VI-NEXT: v_or_b32_sdwa v1, v1, v2 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_0 src1_sel:DWORD
; VI-NEXT: v_add_u32_e32 v2, vcc, 0x70, v0
; VI-NEXT: buffer_store_dword v1, v2, s[0:3], 0 offen
-; VI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:276 ; 4-byte Folded Reload
-; VI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:88 ; 4-byte Folded Reload
-; VI-NEXT: s_waitcnt vmcnt(1)
+; VI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:288 ; 4-byte Folded Reload
+; VI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:112 ; 4-byte Folded Reload
+; VI-NEXT: buffer_load_dword v3, off, s[0:3], s32 offset:84 ; 4-byte Folded Reload
+; VI-NEXT: s_waitcnt vmcnt(2)
; VI-NEXT: v_lshlrev_b32_e32 v1, 8, v1
-; VI-NEXT: s_waitcnt vmcnt(0)
+; VI-NEXT: s_waitcnt vmcnt(1)
; VI-NEXT: v_or_b32_sdwa v1, v2, v1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
-; VI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:264 ; 4-byte Folded Reload
+; VI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:276 ; 4-byte Folded Reload
; VI-NEXT: s_waitcnt vmcnt(0)
; VI-NEXT: v_lshlrev_b32_e32 v2, 8, v2
-; VI-NEXT: v_or_b32_sdwa v2, v52, v2 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; VI-NEXT: v_or_b32_sdwa v2, v3, v2 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
; VI-NEXT: v_or_b32_sdwa v1, v1, v2 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_0 src1_sel:DWORD
; VI-NEXT: v_add_u32_e32 v2, vcc, 0x74, v0
; VI-NEXT: buffer_store_dword v1, v2, s[0:3], 0 offen
-; VI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:268 ; 4-byte Folded Reload
-; VI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:124 ; 4-byte Folded Reload
+; VI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:280 ; 4-byte Folded Reload
+; VI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:148 ; 4-byte Folded Reload
; VI-NEXT: s_waitcnt vmcnt(1)
; VI-NEXT: v_lshlrev_b32_e32 v1, 8, v1
; VI-NEXT: s_waitcnt vmcnt(0)
; VI-NEXT: v_or_b32_sdwa v1, v2, v1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
-; VI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:292 ; 4-byte Folded Reload
-; VI-NEXT: buffer_load_dword v3, off, s[0:3], s32 offset:296 ; 4-byte Folded Reload
-; VI-NEXT: buffer_load_dword v3, off, s[0:3], s32 offset:160 ; 4-byte Folded Reload
+; VI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:304 ; 4-byte Folded Reload
+; VI-NEXT: buffer_load_dword v3, off, s[0:3], s32 offset:308 ; 4-byte Folded Reload
+; VI-NEXT: buffer_load_dword v3, off, s[0:3], s32 offset:180 ; 4-byte Folded Reload
; VI-NEXT: s_waitcnt vmcnt(2)
; VI-NEXT: v_lshlrev_b32_e32 v2, 8, v2
; VI-NEXT: s_waitcnt vmcnt(0)
@@ -199079,15 +199185,15 @@ define inreg <128 x i8> @bitcast_v64f16_to_v128i8_scalar(<64 x half> inreg %a, i
; VI-NEXT: v_or_b32_sdwa v1, v1, v2 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_0 src1_sel:DWORD
; VI-NEXT: v_add_u32_e32 v2, vcc, 0x78, v0
; VI-NEXT: buffer_store_dword v1, v2, s[0:3], 0 offen
-; VI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:260 ; 4-byte Folded Reload
-; VI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:92 ; 4-byte Folded Reload
-; VI-NEXT: buffer_load_dword v3, off, s[0:3], s32 offset:128 ; 4-byte Folded Reload
+; VI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:272 ; 4-byte Folded Reload
+; VI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:116 ; 4-byte Folded Reload
+; VI-NEXT: buffer_load_dword v3, off, s[0:3], s32 offset:56 ; 4-byte Folded Reload
; VI-NEXT: v_add_u32_e32 v0, vcc, 0x7c, v0
; VI-NEXT: s_waitcnt vmcnt(2)
; VI-NEXT: v_lshlrev_b32_e32 v1, 8, v1
; VI-NEXT: s_waitcnt vmcnt(1)
; VI-NEXT: v_or_b32_sdwa v1, v2, v1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
-; VI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:256 ; 4-byte Folded Reload
+; VI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:268 ; 4-byte Folded Reload
; VI-NEXT: s_waitcnt vmcnt(0)
; VI-NEXT: v_lshlrev_b32_e32 v2, 8, v2
; VI-NEXT: v_or_b32_sdwa v2, v3, v2 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
@@ -199108,8 +199214,8 @@ define inreg <128 x i8> @bitcast_v64f16_to_v128i8_scalar(<64 x half> inreg %a, i
; VI-NEXT: buffer_load_dword v41, off, s[0:3], s32 offset:48 ; 4-byte Folded Reload
; VI-NEXT: buffer_load_dword v40, off, s[0:3], s32 offset:52 ; 4-byte Folded Reload
; VI-NEXT: s_or_saveexec_b64 s[4:5], -1
-; VI-NEXT: buffer_load_dword v63, off, s[0:3], s32 offset:340 ; 4-byte Folded Reload
-; VI-NEXT: buffer_load_dword v62, off, s[0:3], s32 offset:344 ; 4-byte Folded Reload
+; VI-NEXT: buffer_load_dword v63, off, s[0:3], s32 offset:336 ; 4-byte Folded Reload
+; VI-NEXT: buffer_load_dword v62, off, s[0:3], s32 offset:340 ; 4-byte Folded Reload
; VI-NEXT: s_mov_b64 exec, s[4:5]
; VI-NEXT: s_waitcnt vmcnt(0)
; VI-NEXT: s_setpc_b64 s[30:31]
@@ -199525,7 +199631,7 @@ define inreg <128 x i8> @bitcast_v64f16_to_v128i8_scalar(<64 x half> inreg %a, i
; GFX9-NEXT: v_lshrrev_b32_e32 v15, 24, v36
; GFX9-NEXT: buffer_store_dword v19, off, s[0:3], s32 offset:252 ; 4-byte Folded Spill
; GFX9-NEXT: v_lshrrev_b32_e32 v19, 8, v13
-; GFX9-NEXT: buffer_store_dword v15, off, s[0:3], s32 offset:376 ; 4-byte Folded Spill
+; GFX9-NEXT: buffer_store_dword v15, off, s[0:3], s32 offset:380 ; 4-byte Folded Spill
; GFX9-NEXT: v_lshrrev_b32_e32 v15, 16, v36
; GFX9-NEXT: buffer_store_dword v19, off, s[0:3], s32 offset:224 ; 4-byte Folded Spill
; GFX9-NEXT: v_lshrrev_b32_e32 v19, 24, v22
@@ -199533,7 +199639,7 @@ define inreg <128 x i8> @bitcast_v64f16_to_v128i8_scalar(<64 x half> inreg %a, i
; GFX9-NEXT: v_lshrrev_b32_e32 v15, 16, v35
; GFX9-NEXT: buffer_store_dword v19, off, s[0:3], s32 offset:260 ; 4-byte Folded Spill
; GFX9-NEXT: v_lshrrev_b32_e32 v19, 16, v22
-; GFX9-NEXT: buffer_store_dword v15, off, s[0:3], s32 offset:380 ; 4-byte Folded Spill
+; GFX9-NEXT: buffer_store_dword v15, off, s[0:3], s32 offset:376 ; 4-byte Folded Spill
; GFX9-NEXT: v_lshrrev_b32_e32 v15, 16, v38
; GFX9-NEXT: buffer_store_dword v19, off, s[0:3], s32 offset:256 ; 4-byte Folded Spill
; GFX9-NEXT: v_lshrrev_b32_e32 v19, 8, v22
@@ -199832,7 +199938,7 @@ define inreg <128 x i8> @bitcast_v64f16_to_v128i8_scalar(<64 x half> inreg %a, i
; GFX9-NEXT: v_readlane_b32 s4, v62, 22
; GFX9-NEXT: v_mov_b32_e32 v60, s4
; GFX9-NEXT: v_readlane_b32 s4, v62, 23
-; GFX9-NEXT: buffer_store_dword v17, off, s[0:3], s32 offset:380 ; 4-byte Folded Spill
+; GFX9-NEXT: buffer_store_dword v17, off, s[0:3], s32 offset:376 ; 4-byte Folded Spill
; GFX9-NEXT: v_mov_b32_e32 v17, s4
; GFX9-NEXT: v_readlane_b32 s4, v62, 24
; GFX9-NEXT: buffer_store_dword v17, off, s[0:3], s32 offset:372 ; 4-byte Folded Spill
@@ -199840,7 +199946,7 @@ define inreg <128 x i8> @bitcast_v64f16_to_v128i8_scalar(<64 x half> inreg %a, i
; GFX9-NEXT: v_readlane_b32 s4, v62, 25
; GFX9-NEXT: v_mov_b32_e32 v23, s4
; GFX9-NEXT: v_readlane_b32 s4, v62, 26
-; GFX9-NEXT: buffer_store_dword v17, off, s[0:3], s32 offset:376 ; 4-byte Folded Spill
+; GFX9-NEXT: buffer_store_dword v17, off, s[0:3], s32 offset:380 ; 4-byte Folded Spill
; GFX9-NEXT: v_mov_b32_e32 v17, s4
; GFX9-NEXT: v_readlane_b32 s4, v62, 27
; GFX9-NEXT: v_mov_b32_e32 v59, s4
@@ -200110,14 +200216,14 @@ define inreg <128 x i8> @bitcast_v64f16_to_v128i8_scalar(<64 x half> inreg %a, i
; GFX9-NEXT: buffer_load_dword v15, off, s[0:3], s32 offset:284 ; 4-byte Folded Reload
; GFX9-NEXT: s_nop 0
; GFX9-NEXT: buffer_load_dword v16, off, s[0:3], s32 offset:288 ; 4-byte Folded Reload
-; GFX9-NEXT: buffer_load_dword v16, off, s[0:3], s32 offset:380 ; 4-byte Folded Reload
+; GFX9-NEXT: buffer_load_dword v16, off, s[0:3], s32 offset:376 ; 4-byte Folded Reload
; GFX9-NEXT: s_waitcnt vmcnt(2)
; GFX9-NEXT: v_lshlrev_b32_e32 v15, 8, v15
; GFX9-NEXT: s_waitcnt vmcnt(0)
; GFX9-NEXT: v_or_b32_sdwa v15, v16, v15 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
; GFX9-NEXT: v_or_b32_sdwa v15, v35, v15 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_0 src1_sel:DWORD
; GFX9-NEXT: buffer_store_dword v15, v0, s[0:3], 0 offen offset:16
-; GFX9-NEXT: buffer_load_dword v15, off, s[0:3], s32 offset:376 ; 4-byte Folded Reload
+; GFX9-NEXT: buffer_load_dword v15, off, s[0:3], s32 offset:380 ; 4-byte Folded Reload
; GFX9-NEXT: s_nop 0
; GFX9-NEXT: buffer_load_dword v16, off, s[0:3], s32 offset:372 ; 4-byte Folded Reload
; GFX9-NEXT: s_waitcnt vmcnt(1)
@@ -202550,19 +202656,19 @@ define <64 x i16> @bitcast_v128i8_to_v64i16(<128 x i8> %a, i32 %b) {
; SI-NEXT: buffer_store_dword v63, off, s[0:3], s32 offset:396 ; 4-byte Folded Spill
; SI-NEXT: buffer_store_dword v29, off, s[0:3], s32 offset:492 ; 4-byte Folded Spill
; SI-NEXT: buffer_store_dword v27, off, s[0:3], s32 offset:496 ; 4-byte Folded Spill
-; SI-NEXT: buffer_store_dword v25, off, s[0:3], s32 offset:656 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v25, off, s[0:3], s32 offset:660 ; 4-byte Folded Spill
; SI-NEXT: buffer_store_dword v23, off, s[0:3], s32 offset:472 ; 4-byte Folded Spill
; SI-NEXT: buffer_store_dword v21, off, s[0:3], s32 offset:488 ; 4-byte Folded Spill
; SI-NEXT: buffer_store_dword v19, off, s[0:3], s32 offset:484 ; 4-byte Folded Spill
-; SI-NEXT: buffer_store_dword v17, off, s[0:3], s32 offset:652 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v17, off, s[0:3], s32 offset:648 ; 4-byte Folded Spill
; SI-NEXT: buffer_store_dword v15, off, s[0:3], s32 offset:460 ; 4-byte Folded Spill
; SI-NEXT: buffer_store_dword v13, off, s[0:3], s32 offset:480 ; 4-byte Folded Spill
; SI-NEXT: buffer_store_dword v11, off, s[0:3], s32 offset:476 ; 4-byte Folded Spill
-; SI-NEXT: buffer_store_dword v9, off, s[0:3], s32 offset:640 ; 4-byte Folded Spill
-; SI-NEXT: buffer_store_dword v7, off, s[0:3], s32 offset:944 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v9, off, s[0:3], s32 offset:636 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v7, off, s[0:3], s32 offset:980 ; 4-byte Folded Spill
; SI-NEXT: buffer_store_dword v5, off, s[0:3], s32 offset:468 ; 4-byte Folded Spill
; SI-NEXT: buffer_store_dword v3, off, s[0:3], s32 offset:464 ; 4-byte Folded Spill
-; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:628 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:632 ; 4-byte Folded Spill
; SI-NEXT: buffer_load_dword v25, off, s[0:3], s32 offset:392
; SI-NEXT: s_waitcnt expcnt(1)
; SI-NEXT: buffer_load_dword v3, off, s[0:3], s32 offset:4
@@ -202588,431 +202694,439 @@ define <64 x i16> @bitcast_v128i8_to_v64i16(<128 x i8> %a, i32 %b) {
; SI-NEXT: buffer_store_dword v4, off, s[0:3], s32 offset:512 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_lshlrev_b32_e32 v4, 8, v14
+; SI-NEXT: buffer_store_dword v4, off, s[0:3], s32 offset:520 ; 4-byte Folded Spill
+; SI-NEXT: s_waitcnt expcnt(0)
+; SI-NEXT: v_lshlrev_b32_e32 v4, 24, v16
; SI-NEXT: buffer_store_dword v4, off, s[0:3], s32 offset:516 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_lshlrev_b32_e32 v4, 24, v12
-; SI-NEXT: buffer_store_dword v4, off, s[0:3], s32 offset:520 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v4, off, s[0:3], s32 offset:524 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_lshlrev_b32_e32 v4, 8, v22
-; SI-NEXT: buffer_store_dword v4, off, s[0:3], s32 offset:524 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v4, off, s[0:3], s32 offset:528 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_lshlrev_b32_e32 v4, 24, v20
-; SI-NEXT: v_lshlrev_b32_e32 v6, 8, v6
-; SI-NEXT: buffer_store_dword v4, off, s[0:3], s32 offset:528 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v4, off, s[0:3], s32 offset:532 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_lshlrev_b32_e32 v4, 8, v30
+; SI-NEXT: buffer_store_dword v4, off, s[0:3], s32 offset:540 ; 4-byte Folded Spill
+; SI-NEXT: s_waitcnt expcnt(0)
+; SI-NEXT: v_lshlrev_b32_e32 v4, 24, v28
+; SI-NEXT: buffer_store_dword v4, off, s[0:3], s32 offset:544 ; 4-byte Folded Spill
+; SI-NEXT: v_lshlrev_b32_e32 v6, 8, v6
; SI-NEXT: buffer_store_dword v6, off, s[0:3], s32 offset:508 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_lshlrev_b32_e32 v6, 24, v8
-; SI-NEXT: buffer_store_dword v4, off, s[0:3], s32 offset:532 ; 4-byte Folded Spill
-; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_lshlrev_b32_e32 v4, 24, v28
; SI-NEXT: buffer_store_dword v6, off, s[0:3], s32 offset:504 ; 4-byte Folded Spill
-; SI-NEXT: buffer_store_dword v4, off, s[0:3], s32 offset:536 ; 4-byte Folded Spill
-; SI-NEXT: v_lshlrev_b32_e32 v2, 8, v2
-; SI-NEXT: v_lshlrev_b32_e32 v16, 24, v16
; SI-NEXT: v_lshlrev_b32_e32 v24, 24, v24
-; SI-NEXT: ; implicit-def: $vgpr44
-; SI-NEXT: ; implicit-def: $vgpr36
-; SI-NEXT: ; implicit-def: $vgpr43
; SI-NEXT: ; implicit-def: $vgpr22
-; SI-NEXT: ; implicit-def: $vgpr58
+; SI-NEXT: ; implicit-def: $vgpr57
+; SI-NEXT: ; implicit-def: $vgpr43
; SI-NEXT: ; implicit-def: $vgpr40
; SI-NEXT: ; implicit-def: $vgpr42
-; SI-NEXT: ; implicit-def: $vgpr55
+; SI-NEXT: ; implicit-def: $vgpr36
; SI-NEXT: ; implicit-def: $vgpr39
+; SI-NEXT: ; implicit-def: $vgpr55
+; SI-NEXT: ; implicit-def: $vgpr53
; SI-NEXT: ; implicit-def: $vgpr20
; SI-NEXT: ; implicit-def: $vgpr28
-; SI-NEXT: ; implicit-def: $vgpr52
; SI-NEXT: ; implicit-def: $vgpr38
-; SI-NEXT: ; implicit-def: $vgpr46
; SI-NEXT: ; implicit-def: $vgpr62
-; SI-NEXT: ; implicit-def: $vgpr50
+; SI-NEXT: ; implicit-def: $vgpr56
; SI-NEXT: ; implicit-def: $vgpr59
; SI-NEXT: ; implicit-def: $vgpr30
-; SI-NEXT: ; implicit-def: $vgpr49
+; SI-NEXT: ; implicit-def: $vgpr50
; SI-NEXT: ; implicit-def: $vgpr47
-; SI-NEXT: ; implicit-def: $vgpr48
+; SI-NEXT: ; implicit-def: $vgpr49
; SI-NEXT: ; implicit-def: $vgpr45
-; SI-NEXT: ; kill: killed $vgpr44
+; SI-NEXT: ; implicit-def: $vgpr48
; SI-NEXT: ; implicit-def: $vgpr44
-; SI-NEXT: ; kill: killed $vgpr36
-; SI-NEXT: ; implicit-def: $vgpr36
-; SI-NEXT: buffer_load_dword v61, off, s[0:3], s32 offset:360
+; SI-NEXT: buffer_load_dword v16, off, s[0:3], s32 offset:332
+; SI-NEXT: buffer_load_dword v52, off, s[0:3], s32 offset:360
; SI-NEXT: s_waitcnt vmcnt(14)
; SI-NEXT: v_cmp_ne_u32_e32 vcc, 0, v25
-; SI-NEXT: v_lshlrev_b32_e32 v25, 24, v3
+; SI-NEXT: v_lshlrev_b32_e32 v3, 24, v3
+; SI-NEXT: buffer_store_dword v3, off, s[0:3], s32 offset:536 ; 4-byte Folded Spill
+; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_lshlrev_b32_e32 v3, 8, v7
-; SI-NEXT: buffer_store_dword v3, off, s[0:3], s32 offset:548 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v3, off, s[0:3], s32 offset:552 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_lshlrev_b32_e32 v3, 24, v9
-; SI-NEXT: buffer_store_dword v3, off, s[0:3], s32 offset:540 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v3, off, s[0:3], s32 offset:548 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_lshlrev_b32_e32 v3, 24, v5
; SI-NEXT: buffer_store_dword v3, off, s[0:3], s32 offset:576 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_lshlrev_b32_e32 v3, 8, v13
-; SI-NEXT: buffer_store_dword v3, off, s[0:3], s32 offset:588 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v3, off, s[0:3], s32 offset:592 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_lshlrev_b32_e32 v3, 24, v15
-; SI-NEXT: buffer_store_dword v3, off, s[0:3], s32 offset:584 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v3, off, s[0:3], s32 offset:588 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_lshlrev_b32_e32 v3, 24, v11
-; SI-NEXT: buffer_store_dword v3, off, s[0:3], s32 offset:592 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v3, off, s[0:3], s32 offset:596 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_lshlrev_b32_e32 v3, 8, v19
-; SI-NEXT: buffer_store_dword v3, off, s[0:3], s32 offset:620 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v3, off, s[0:3], s32 offset:624 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_lshlrev_b32_e32 v3, 24, v21
-; SI-NEXT: buffer_store_dword v3, off, s[0:3], s32 offset:612 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt vmcnt(14)
; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:500 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:96
+; SI-NEXT: v_lshlrev_b32_e32 v4, 24, v27
+; SI-NEXT: buffer_store_dword v3, off, s[0:3], s32 offset:616 ; 4-byte Folded Spill
+; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_lshlrev_b32_e32 v3, 24, v17
-; SI-NEXT: buffer_store_dword v3, off, s[0:3], s32 offset:624 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v4, off, s[0:3], s32 offset:644 ; 4-byte Folded Spill
+; SI-NEXT: s_waitcnt expcnt(0)
+; SI-NEXT: v_lshlrev_b32_e32 v4, 24, v29
+; SI-NEXT: buffer_store_dword v3, off, s[0:3], s32 offset:628 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_lshlrev_b32_e32 v3, 8, v23
-; SI-NEXT: buffer_store_dword v3, off, s[0:3], s32 offset:648 ; 4-byte Folded Spill
-; SI-NEXT: v_lshlrev_b32_e32 v4, 24, v27
-; SI-NEXT: v_lshlrev_b32_e32 v6, 24, v32
+; SI-NEXT: buffer_store_dword v4, off, s[0:3], s32 offset:672 ; 4-byte Folded Spill
+; SI-NEXT: s_waitcnt expcnt(0)
+; SI-NEXT: v_lshlrev_b32_e32 v4, 8, v31
+; SI-NEXT: buffer_store_dword v3, off, s[0:3], s32 offset:656 ; 4-byte Folded Spill
+; SI-NEXT: s_waitcnt expcnt(0)
+; SI-NEXT: buffer_load_dword v3, off, s[0:3], s32 offset:128
+; SI-NEXT: buffer_store_dword v4, off, s[0:3], s32 offset:704 ; 4-byte Folded Spill
+; SI-NEXT: s_waitcnt expcnt(0)
+; SI-NEXT: buffer_load_dword v4, off, s[0:3], s32 offset:160
+; SI-NEXT: v_lshlrev_b32_e32 v5, 24, v32
+; SI-NEXT: buffer_load_dword v15, off, s[0:3], s32 offset:300
+; SI-NEXT: v_lshlrev_b32_e32 v32, 8, v2
+; SI-NEXT: v_lshlrev_b32_e32 v2, 8, v18
; SI-NEXT: ; implicit-def: $vgpr21
; SI-NEXT: ; implicit-def: $vgpr27
+; SI-NEXT: ; implicit-def: $vgpr29
; SI-NEXT: ; implicit-def: $vgpr19
-; SI-NEXT: ; implicit-def: $vgpr17
-; SI-NEXT: s_waitcnt vmcnt(2)
-; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:552 ; 4-byte Folded Spill
+; SI-NEXT: s_waitcnt vmcnt(9)
+; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:556 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:88
-; SI-NEXT: s_waitcnt vmcnt(0)
-; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:560 ; 4-byte Folded Spill
+; SI-NEXT: s_waitcnt vmcnt(3)
+; SI-NEXT: buffer_store_dword v4, off, s[0:3], s32 offset:600 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:80
-; SI-NEXT: s_waitcnt vmcnt(0)
-; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:564 ; 4-byte Folded Spill
+; SI-NEXT: buffer_load_dword v4, off, s[0:3], s32 offset:184
+; SI-NEXT: buffer_store_dword v3, off, s[0:3], s32 offset:580 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:112
-; SI-NEXT: s_waitcnt vmcnt(0)
-; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:572 ; 4-byte Folded Spill
-; SI-NEXT: buffer_load_dword v23, off, s[0:3], s32 offset:128
; SI-NEXT: buffer_load_dword v3, off, s[0:3], s32 offset:152
-; SI-NEXT: buffer_store_dword v4, off, s[0:3], s32 offset:636 ; 4-byte Folded Spill
-; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_lshlrev_b32_e32 v4, 24, v29
-; SI-NEXT: buffer_store_dword v4, off, s[0:3], s32 offset:664 ; 4-byte Folded Spill
-; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_lshlrev_b32_e32 v4, 8, v31
-; SI-NEXT: buffer_store_dword v4, off, s[0:3], s32 offset:692 ; 4-byte Folded Spill
-; SI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:108
-; SI-NEXT: ; implicit-def: $vgpr29
+; SI-NEXT: s_waitcnt vmcnt(4)
+; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:560 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: buffer_load_dword v4, off, s[0:3], s32 offset:160
-; SI-NEXT: s_waitcnt vmcnt(5)
-; SI-NEXT: buffer_store_dword v3, off, s[0:3], s32 offset:600 ; 4-byte Folded Spill
+; SI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:80
+; SI-NEXT: s_waitcnt vmcnt(4)
+; SI-NEXT: buffer_store_dword v4, off, s[0:3], s32 offset:640 ; 4-byte Folded Spill
+; SI-NEXT: s_waitcnt vmcnt(3)
+; SI-NEXT: buffer_store_dword v3, off, s[0:3], s32 offset:604 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: buffer_load_dword v3, off, s[0:3], s32 offset:144
-; SI-NEXT: s_waitcnt vmcnt(2)
-; SI-NEXT: buffer_store_dword v4, off, s[0:3], s32 offset:596 ; 4-byte Folded Spill
+; SI-NEXT: s_waitcnt vmcnt(3)
+; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:564 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: buffer_load_dword v4, off, s[0:3], s32 offset:184
+; SI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:112
; SI-NEXT: s_waitcnt vmcnt(2)
-; SI-NEXT: buffer_store_dword v3, off, s[0:3], s32 offset:608 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v3, off, s[0:3], s32 offset:612 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: buffer_load_dword v3, off, s[0:3], s32 offset:140
; SI-NEXT: s_waitcnt vmcnt(2)
-; SI-NEXT: buffer_store_dword v4, off, s[0:3], s32 offset:632 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:572 ; 4-byte Folded Spill
+; SI-NEXT: buffer_load_dword v6, off, s[0:3], s32 offset:180
+; SI-NEXT: buffer_load_dword v4, off, s[0:3], s32 offset:176
+; SI-NEXT: buffer_store_dword v5, off, s[0:3], s32 offset:700 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: buffer_load_dword v4, off, s[0:3], s32 offset:180
-; SI-NEXT: buffer_load_dword v5, off, s[0:3], s32 offset:176
-; SI-NEXT: buffer_store_dword v6, off, s[0:3], s32 offset:688 ; 4-byte Folded Spill
+; SI-NEXT: v_lshlrev_b32_e32 v5, 24, v33
+; SI-NEXT: buffer_store_dword v5, off, s[0:3], s32 offset:712 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_lshlrev_b32_e32 v6, 24, v33
+; SI-NEXT: v_lshlrev_b32_e32 v5, 8, v34
+; SI-NEXT: buffer_store_dword v5, off, s[0:3], s32 offset:728 ; 4-byte Folded Spill
+; SI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:108
+; SI-NEXT: v_lshlrev_b32_e32 v34, 8, v10
+; SI-NEXT: ; implicit-def: $vgpr10
; SI-NEXT: ; implicit-def: $vgpr33
-; SI-NEXT: buffer_store_dword v6, off, s[0:3], s32 offset:700 ; 4-byte Folded Spill
-; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_lshlrev_b32_e32 v6, 8, v34
-; SI-NEXT: buffer_store_dword v6, off, s[0:3], s32 offset:720 ; 4-byte Folded Spill
+; SI-NEXT: s_waitcnt vmcnt(5)
+; SI-NEXT: v_lshlrev_b32_e32 v6, 24, v6
; SI-NEXT: s_waitcnt vmcnt(4)
-; SI-NEXT: v_lshlrev_b32_e32 v4, 24, v4
-; SI-NEXT: s_waitcnt vmcnt(3)
-; SI-NEXT: buffer_store_dword v5, off, s[0:3], s32 offset:644 ; 4-byte Folded Spill
-; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: buffer_load_dword v5, off, s[0:3], s32 offset:172
+; SI-NEXT: buffer_store_dword v4, off, s[0:3], s32 offset:652 ; 4-byte Folded Spill
; SI-NEXT: buffer_load_dword v7, off, s[0:3], s32 offset:196
; SI-NEXT: buffer_load_dword v8, off, s[0:3], s32 offset:220
-; SI-NEXT: buffer_load_dword v6, off, s[0:3], s32 offset:192
-; SI-NEXT: s_waitcnt vmcnt(2)
+; SI-NEXT: s_waitcnt expcnt(1)
+; SI-NEXT: buffer_load_dword v5, off, s[0:3], s32 offset:192
+; SI-NEXT: s_waitcnt expcnt(0)
+; SI-NEXT: buffer_load_dword v4, off, s[0:3], s32 offset:172
+; SI-NEXT: s_waitcnt vmcnt(3)
; SI-NEXT: v_lshlrev_b32_e32 v7, 24, v7
-; SI-NEXT: s_waitcnt vmcnt(0)
-; SI-NEXT: buffer_store_dword v6, off, s[0:3], s32 offset:672 ; 4-byte Folded Spill
+; SI-NEXT: s_waitcnt vmcnt(1)
+; SI-NEXT: buffer_store_dword v5, off, s[0:3], s32 offset:680 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: buffer_load_dword v6, off, s[0:3], s32 offset:216
+; SI-NEXT: buffer_load_dword v5, off, s[0:3], s32 offset:216
; SI-NEXT: s_waitcnt vmcnt(0)
-; SI-NEXT: buffer_store_dword v6, off, s[0:3], s32 offset:704 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v5, off, s[0:3], s32 offset:716 ; 4-byte Folded Spill
; SI-NEXT: buffer_load_dword v9, off, s[0:3], s32 offset:212
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: buffer_load_dword v6, off, s[0:3], s32 offset:208
-; SI-NEXT: buffer_store_dword v4, off, s[0:3], s32 offset:736 ; 4-byte Folded Spill
+; SI-NEXT: buffer_load_dword v5, off, s[0:3], s32 offset:208
+; SI-NEXT: buffer_store_dword v6, off, s[0:3], s32 offset:744 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_lshlrev_b32_e32 v4, 8, v8
-; SI-NEXT: buffer_store_dword v7, off, s[0:3], s32 offset:728 ; 4-byte Folded Spill
-; SI-NEXT: buffer_store_dword v4, off, s[0:3], s32 offset:752 ; 4-byte Folded Spill
+; SI-NEXT: v_lshlrev_b32_e32 v6, 8, v8
+; SI-NEXT: buffer_store_dword v7, off, s[0:3], s32 offset:736 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v6, off, s[0:3], s32 offset:760 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt vmcnt(3)
-; SI-NEXT: buffer_store_dword v6, off, s[0:3], s32 offset:708 ; 4-byte Folded Spill
-; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: buffer_load_dword v6, off, s[0:3], s32 offset:204
-; SI-NEXT: buffer_load_dword v4, off, s[0:3], s32 offset:228
+; SI-NEXT: buffer_store_dword v5, off, s[0:3], s32 offset:720 ; 4-byte Folded Spill
+; SI-NEXT: s_waitcnt expcnt(2)
+; SI-NEXT: buffer_load_dword v7, off, s[0:3], s32 offset:228
; SI-NEXT: buffer_load_dword v8, off, s[0:3], s32 offset:252
-; SI-NEXT: buffer_load_dword v7, off, s[0:3], s32 offset:224
-; SI-NEXT: s_waitcnt vmcnt(2)
-; SI-NEXT: v_lshlrev_b32_e32 v4, 24, v4
-; SI-NEXT: s_waitcnt vmcnt(0)
-; SI-NEXT: buffer_store_dword v7, off, s[0:3], s32 offset:744 ; 4-byte Folded Spill
+; SI-NEXT: s_waitcnt expcnt(1)
+; SI-NEXT: buffer_load_dword v6, off, s[0:3], s32 offset:224
+; SI-NEXT: s_waitcnt expcnt(0)
+; SI-NEXT: buffer_load_dword v5, off, s[0:3], s32 offset:204
+; SI-NEXT: s_waitcnt vmcnt(3)
+; SI-NEXT: v_lshlrev_b32_e32 v7, 24, v7
+; SI-NEXT: s_waitcnt vmcnt(1)
+; SI-NEXT: buffer_store_dword v6, off, s[0:3], s32 offset:752 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: buffer_load_dword v7, off, s[0:3], s32 offset:248
+; SI-NEXT: buffer_load_dword v6, off, s[0:3], s32 offset:248
; SI-NEXT: s_waitcnt vmcnt(0)
-; SI-NEXT: buffer_store_dword v7, off, s[0:3], s32 offset:764 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v6, off, s[0:3], s32 offset:772 ; 4-byte Folded Spill
; SI-NEXT: buffer_load_dword v11, off, s[0:3], s32 offset:244
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: buffer_load_dword v7, off, s[0:3], s32 offset:240
-; SI-NEXT: buffer_store_dword v4, off, s[0:3], s32 offset:796 ; 4-byte Folded Spill
+; SI-NEXT: buffer_load_dword v6, off, s[0:3], s32 offset:240
+; SI-NEXT: buffer_store_dword v7, off, s[0:3], s32 offset:804 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_lshlrev_b32_e32 v4, 24, v9
-; SI-NEXT: buffer_store_dword v4, off, s[0:3], s32 offset:808 ; 4-byte Folded Spill
+; SI-NEXT: v_lshlrev_b32_e32 v7, 24, v9
+; SI-NEXT: buffer_store_dword v7, off, s[0:3], s32 offset:816 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_lshlrev_b32_e32 v4, 8, v8
-; SI-NEXT: buffer_store_dword v4, off, s[0:3], s32 offset:832 ; 4-byte Folded Spill
+; SI-NEXT: v_lshlrev_b32_e32 v7, 8, v8
+; SI-NEXT: buffer_store_dword v7, off, s[0:3], s32 offset:840 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt vmcnt(3)
-; SI-NEXT: buffer_store_dword v7, off, s[0:3], s32 offset:768 ; 4-byte Folded Spill
-; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: buffer_load_dword v7, off, s[0:3], s32 offset:236
-; SI-NEXT: buffer_load_dword v4, off, s[0:3], s32 offset:260
+; SI-NEXT: buffer_store_dword v6, off, s[0:3], s32 offset:776 ; 4-byte Folded Spill
+; SI-NEXT: s_waitcnt expcnt(1)
+; SI-NEXT: buffer_load_dword v7, off, s[0:3], s32 offset:260
; SI-NEXT: buffer_load_dword v8, off, s[0:3], s32 offset:284
; SI-NEXT: buffer_load_dword v9, off, s[0:3], s32 offset:256
-; SI-NEXT: s_waitcnt vmcnt(2)
-; SI-NEXT: v_lshlrev_b32_e32 v4, 24, v4
-; SI-NEXT: s_waitcnt vmcnt(0)
-; SI-NEXT: buffer_store_dword v9, off, s[0:3], s32 offset:788 ; 4-byte Folded Spill
+; SI-NEXT: s_waitcnt expcnt(0)
+; SI-NEXT: buffer_load_dword v6, off, s[0:3], s32 offset:236
+; SI-NEXT: s_waitcnt vmcnt(3)
+; SI-NEXT: v_lshlrev_b32_e32 v7, 24, v7
+; SI-NEXT: s_waitcnt vmcnt(1)
+; SI-NEXT: buffer_store_dword v9, off, s[0:3], s32 offset:796 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: buffer_load_dword v9, off, s[0:3], s32 offset:280
; SI-NEXT: s_waitcnt vmcnt(0)
-; SI-NEXT: buffer_store_dword v9, off, s[0:3], s32 offset:812 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v9, off, s[0:3], s32 offset:820 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: buffer_load_dword v9, off, s[0:3], s32 offset:276
; SI-NEXT: buffer_load_dword v12, off, s[0:3], s32 offset:272
-; SI-NEXT: buffer_store_dword v4, off, s[0:3], s32 offset:836 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v7, off, s[0:3], s32 offset:844 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_lshlrev_b32_e32 v4, 24, v11
-; SI-NEXT: buffer_store_dword v4, off, s[0:3], s32 offset:844 ; 4-byte Folded Spill
+; SI-NEXT: v_lshlrev_b32_e32 v7, 24, v11
+; SI-NEXT: buffer_store_dword v7, off, s[0:3], s32 offset:852 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_lshlrev_b32_e32 v4, 8, v8
-; SI-NEXT: buffer_store_dword v4, off, s[0:3], s32 offset:932 ; 4-byte Folded Spill
+; SI-NEXT: v_lshlrev_b32_e32 v7, 8, v8
+; SI-NEXT: buffer_store_dword v7, off, s[0:3], s32 offset:940 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt vmcnt(3)
-; SI-NEXT: buffer_store_dword v12, off, s[0:3], s32 offset:816 ; 4-byte Folded Spill
-; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: buffer_load_dword v12, off, s[0:3], s32 offset:268
-; SI-NEXT: buffer_load_dword v4, off, s[0:3], s32 offset:292
+; SI-NEXT: buffer_store_dword v12, off, s[0:3], s32 offset:824 ; 4-byte Folded Spill
+; SI-NEXT: s_waitcnt expcnt(1)
+; SI-NEXT: buffer_load_dword v7, off, s[0:3], s32 offset:292
; SI-NEXT: buffer_load_dword v8, off, s[0:3], s32 offset:316
; SI-NEXT: buffer_load_dword v11, off, s[0:3], s32 offset:288
+; SI-NEXT: s_waitcnt expcnt(0)
+; SI-NEXT: buffer_load_dword v12, off, s[0:3], s32 offset:268
+; SI-NEXT: s_waitcnt vmcnt(3)
+; SI-NEXT: v_lshlrev_b32_e32 v7, 24, v7
; SI-NEXT: s_waitcnt vmcnt(2)
-; SI-NEXT: v_lshlrev_b32_e32 v4, 24, v4
-; SI-NEXT: s_waitcnt vmcnt(1)
; SI-NEXT: v_lshlrev_b32_e32 v35, 8, v8
-; SI-NEXT: s_waitcnt vmcnt(0)
-; SI-NEXT: buffer_store_dword v11, off, s[0:3], s32 offset:824 ; 4-byte Folded Spill
+; SI-NEXT: s_waitcnt vmcnt(1)
+; SI-NEXT: buffer_store_dword v11, off, s[0:3], s32 offset:832 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: buffer_load_dword v11, off, s[0:3], s32 offset:312
; SI-NEXT: s_waitcnt vmcnt(0)
-; SI-NEXT: buffer_store_dword v11, off, s[0:3], s32 offset:848 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v11, off, s[0:3], s32 offset:856 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: buffer_load_dword v11, off, s[0:3], s32 offset:308
; SI-NEXT: buffer_load_dword v13, off, s[0:3], s32 offset:304
-; SI-NEXT: buffer_store_dword v4, off, s[0:3], s32 offset:936 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v7, off, s[0:3], s32 offset:944 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_lshlrev_b32_e32 v4, 24, v9
-; SI-NEXT: buffer_store_dword v4, off, s[0:3], s32 offset:952 ; 4-byte Folded Spill
+; SI-NEXT: v_lshlrev_b32_e32 v7, 24, v9
+; SI-NEXT: buffer_store_dword v7, off, s[0:3], s32 offset:956 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt vmcnt(3)
; SI-NEXT: v_lshlrev_b32_e32 v51, 24, v11
; SI-NEXT: s_waitcnt vmcnt(2)
-; SI-NEXT: buffer_store_dword v13, off, s[0:3], s32 offset:916 ; 4-byte Folded Spill
-; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: buffer_load_dword v13, off, s[0:3], s32 offset:300
-; SI-NEXT: buffer_load_dword v4, off, s[0:3], s32 offset:324
+; SI-NEXT: buffer_store_dword v13, off, s[0:3], s32 offset:924 ; 4-byte Folded Spill
+; SI-NEXT: s_waitcnt expcnt(1)
+; SI-NEXT: buffer_load_dword v7, off, s[0:3], s32 offset:324
; SI-NEXT: buffer_load_dword v8, off, s[0:3], s32 offset:348
; SI-NEXT: buffer_load_dword v9, off, s[0:3], s32 offset:320
; SI-NEXT: s_waitcnt vmcnt(2)
-; SI-NEXT: v_lshlrev_b32_e32 v37, 24, v4
+; SI-NEXT: v_lshlrev_b32_e32 v37, 24, v7
; SI-NEXT: s_waitcnt vmcnt(1)
-; SI-NEXT: v_lshlrev_b32_e32 v63, 8, v8
+; SI-NEXT: v_lshlrev_b32_e32 v60, 8, v8
; SI-NEXT: s_waitcnt vmcnt(0)
-; SI-NEXT: buffer_store_dword v9, off, s[0:3], s32 offset:924 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v9, off, s[0:3], s32 offset:932 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: buffer_load_dword v9, off, s[0:3], s32 offset:344
; SI-NEXT: s_waitcnt vmcnt(0)
-; SI-NEXT: buffer_store_dword v9, off, s[0:3], s32 offset:956 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v9, off, s[0:3], s32 offset:960 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: buffer_load_dword v9, off, s[0:3], s32 offset:340
-; SI-NEXT: buffer_load_dword v14, off, s[0:3], s32 offset:336
+; SI-NEXT: buffer_load_dword v13, off, s[0:3], s32 offset:336
; SI-NEXT: s_waitcnt vmcnt(1)
-; SI-NEXT: v_lshlrev_b32_e32 v60, 24, v9
+; SI-NEXT: v_lshlrev_b32_e32 v61, 24, v9
; SI-NEXT: s_waitcnt vmcnt(0)
-; SI-NEXT: buffer_store_dword v14, off, s[0:3], s32 offset:964 ; 4-byte Folded Spill
-; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: buffer_load_dword v14, off, s[0:3], s32 offset:332
-; SI-NEXT: buffer_load_dword v4, off, s[0:3], s32 offset:356
+; SI-NEXT: buffer_store_dword v13, off, s[0:3], s32 offset:968 ; 4-byte Folded Spill
+; SI-NEXT: buffer_load_dword v7, off, s[0:3], s32 offset:356
; SI-NEXT: buffer_load_dword v8, off, s[0:3], s32 offset:380
-; SI-NEXT: buffer_load_dword v34, off, s[0:3], s32 offset:352
+; SI-NEXT: buffer_load_dword v14, off, s[0:3], s32 offset:352
; SI-NEXT: buffer_load_dword v54, off, s[0:3], s32 offset:376
; SI-NEXT: buffer_load_dword v11, off, s[0:3], s32 offset:372
; SI-NEXT: buffer_load_dword v41, off, s[0:3], s32 offset:368
-; SI-NEXT: buffer_load_dword v15, off, s[0:3], s32 offset:364
+; SI-NEXT: buffer_load_dword v17, off, s[0:3], s32 offset:364
+; SI-NEXT: v_lshlrev_b32_e32 v9, 8, v16
+; SI-NEXT: ; implicit-def: $vgpr16
; SI-NEXT: s_waitcnt vmcnt(6)
-; SI-NEXT: v_lshlrev_b32_e32 v56, 24, v4
-; SI-NEXT: s_waitcnt vmcnt(5)
-; SI-NEXT: v_lshlrev_b32_e32 v4, 8, v8
-; SI-NEXT: buffer_load_dword v8, off, s[0:3], s32 offset:388
-; SI-NEXT: buffer_load_dword v57, off, s[0:3], s32 offset:384
-; SI-NEXT: buffer_store_dword v2, off, s[0:3], s32 offset:544 ; 4-byte Folded Spill
-; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_lshlrev_b32_e32 v2, 8, v10
-; SI-NEXT: ; implicit-def: $vgpr10
-; SI-NEXT: buffer_store_dword v2, off, s[0:3], s32 offset:556 ; 4-byte Folded Spill
-; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_lshlrev_b32_e32 v2, 8, v18
-; SI-NEXT: ; implicit-def: $vgpr18
+; SI-NEXT: v_lshlrev_b32_e32 v31, 24, v7
+; SI-NEXT: buffer_load_dword v7, off, s[0:3], s32 offset:388
+; SI-NEXT: buffer_load_dword v46, off, s[0:3], s32 offset:384
; SI-NEXT: buffer_store_dword v2, off, s[0:3], s32 offset:568 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_lshlrev_b32_e32 v2, 8, v26
-; SI-NEXT: ; implicit-def: $vgpr26
-; SI-NEXT: buffer_store_dword v2, off, s[0:3], s32 offset:604 ; 4-byte Folded Spill
-; SI-NEXT: s_waitcnt vmcnt(8)
-; SI-NEXT: v_lshlrev_b32_e32 v9, 24, v11
-; SI-NEXT: ; implicit-def: $vgpr11
+; SI-NEXT: buffer_store_dword v2, off, s[0:3], s32 offset:608 ; 4-byte Folded Spill
+; SI-NEXT: s_waitcnt vmcnt(9)
+; SI-NEXT: v_lshlrev_b32_e32 v23, 8, v8
; SI-NEXT: s_waitcnt vmcnt(6)
-; SI-NEXT: v_lshlrev_b32_e32 v31, 8, v15
-; SI-NEXT: ; implicit-def: $vgpr15
-; SI-NEXT: s_waitcnt vmcnt(5)
-; SI-NEXT: v_lshlrev_b32_e32 v32, 24, v8
+; SI-NEXT: v_lshlrev_b32_e32 v13, 24, v11
+; SI-NEXT: s_waitcnt vmcnt(4)
+; SI-NEXT: v_lshlrev_b32_e32 v63, 8, v17
+; SI-NEXT: ; implicit-def: $vgpr11
+; SI-NEXT: ; implicit-def: $vgpr26
+; SI-NEXT: ; implicit-def: $vgpr17
+; SI-NEXT: ; implicit-def: $vgpr8
+; SI-NEXT: s_waitcnt vmcnt(3)
+; SI-NEXT: v_lshlrev_b32_e32 v25, 24, v7
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:12
-; SI-NEXT: buffer_load_dword v8, off, s[0:3], s32 offset:8
+; SI-NEXT: buffer_load_dword v7, off, s[0:3], s32 offset:8
; SI-NEXT: s_waitcnt vmcnt(1)
; SI-NEXT: v_lshlrev_b32_e32 v2, 8, v2
; SI-NEXT: s_waitcnt vmcnt(0)
-; SI-NEXT: buffer_store_dword v8, off, s[0:3], s32 offset:580 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v7, off, s[0:3], s32 offset:584 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: buffer_load_dword v8, off, s[0:3], s32 offset:120
-; SI-NEXT: buffer_store_dword v2, off, s[0:3], s32 offset:712 ; 4-byte Folded Spill
+; SI-NEXT: buffer_load_dword v7, off, s[0:3], s32 offset:120
+; SI-NEXT: buffer_store_dword v2, off, s[0:3], s32 offset:724 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt vmcnt(1)
-; SI-NEXT: buffer_store_dword v8, off, s[0:3], s32 offset:800 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v7, off, s[0:3], s32 offset:808 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(1)
; SI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:44
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: buffer_load_dword v8, off, s[0:3], s32 offset:40
+; SI-NEXT: buffer_load_dword v7, off, s[0:3], s32 offset:40
; SI-NEXT: s_waitcnt vmcnt(1)
; SI-NEXT: v_lshlrev_b32_e32 v2, 8, v2
; SI-NEXT: s_waitcnt vmcnt(0)
-; SI-NEXT: buffer_store_dword v8, off, s[0:3], s32 offset:616 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v7, off, s[0:3], s32 offset:620 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: buffer_load_dword v8, off, s[0:3], s32 offset:32
-; SI-NEXT: buffer_store_dword v2, off, s[0:3], s32 offset:740 ; 4-byte Folded Spill
+; SI-NEXT: buffer_load_dword v7, off, s[0:3], s32 offset:32
+; SI-NEXT: buffer_store_dword v2, off, s[0:3], s32 offset:748 ; 4-byte Folded Spill
; SI-NEXT: v_lshlrev_b32_e32 v1, 8, v1
; SI-NEXT: s_waitcnt vmcnt(1)
-; SI-NEXT: buffer_store_dword v8, off, s[0:3], s32 offset:696 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v7, off, s[0:3], s32 offset:708 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(1)
; SI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:76
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: buffer_load_dword v8, off, s[0:3], s32 offset:72
-; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:792 ; 4-byte Folded Spill
+; SI-NEXT: buffer_load_dword v7, off, s[0:3], s32 offset:72
+; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:800 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt vmcnt(2)
; SI-NEXT: v_lshlrev_b32_e32 v2, 8, v2
; SI-NEXT: s_waitcnt vmcnt(1)
-; SI-NEXT: buffer_store_dword v8, off, s[0:3], s32 offset:668 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v7, off, s[0:3], s32 offset:676 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: buffer_load_dword v8, off, s[0:3], s32 offset:64
-; SI-NEXT: buffer_store_dword v2, off, s[0:3], s32 offset:756 ; 4-byte Folded Spill
+; SI-NEXT: buffer_load_dword v7, off, s[0:3], s32 offset:64
+; SI-NEXT: buffer_store_dword v2, off, s[0:3], s32 offset:764 ; 4-byte Folded Spill
; SI-NEXT: ; implicit-def: $vgpr2
; SI-NEXT: s_waitcnt vmcnt(1)
-; SI-NEXT: buffer_store_dword v8, off, s[0:3], s32 offset:760 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v7, off, s[0:3], s32 offset:768 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: buffer_load_dword v8, off, s[0:3], s32 offset:56
+; SI-NEXT: buffer_load_dword v7, off, s[0:3], s32 offset:56
; SI-NEXT: v_lshlrev_b32_e32 v1, 8, v3
-; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:820 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:828 ; 4-byte Folded Spill
+; SI-NEXT: s_waitcnt expcnt(0)
+; SI-NEXT: v_lshlrev_b32_e32 v1, 8, v4
+; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:848 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_lshlrev_b32_e32 v1, 8, v5
-; SI-NEXT: ; implicit-def: $vgpr5
-; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:840 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:936 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_lshlrev_b32_e32 v1, 8, v6
-; SI-NEXT: ; implicit-def: $vgpr6
-; SI-NEXT: ; kill: killed $vgpr6
-; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:928 ; 4-byte Folded Spill
-; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_lshlrev_b32_e32 v1, 8, v7
-; SI-NEXT: ; implicit-def: $vgpr6
-; SI-NEXT: ; kill: killed $vgpr6
-; SI-NEXT: ; implicit-def: $vgpr7
-; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:960 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:964 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_lshlrev_b32_e32 v1, 8, v12
-; SI-NEXT: ; implicit-def: $vgpr6
-; SI-NEXT: ; implicit-def: $vgpr12
-; SI-NEXT: ; kill: killed $vgpr6
-; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:972 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:976 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:24
+; SI-NEXT: ; implicit-def: $vgpr3
+; SI-NEXT: ; kill: killed $vgpr3
+; SI-NEXT: ; implicit-def: $vgpr3
+; SI-NEXT: ; kill: killed $vgpr3
+; SI-NEXT: ; implicit-def: $vgpr3
+; SI-NEXT: ; kill: killed $vgpr3
+; SI-NEXT: ; implicit-def: $vgpr3
+; SI-NEXT: ; kill: killed $vgpr3
+; SI-NEXT: ; implicit-def: $vgpr3
+; SI-NEXT: ; kill: killed $vgpr3
+; SI-NEXT: ; implicit-def: $vgpr3
+; SI-NEXT: ; implicit-def: $vgpr12
+; SI-NEXT: ; implicit-def: $vgpr5
; SI-NEXT: ; implicit-def: $vgpr6
+; SI-NEXT: ; implicit-def: $vgpr4
+; SI-NEXT: ; kill: killed $vgpr3
+; SI-NEXT: ; implicit-def: $vgpr3
+; SI-NEXT: ; kill: killed $vgpr3
; SI-NEXT: s_waitcnt vmcnt(6)
-; SI-NEXT: buffer_store_dword v8, off, s[0:3], s32 offset:772 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v7, off, s[0:3], s32 offset:780 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: buffer_load_dword v8, off, s[0:3], s32 offset:48
-; SI-NEXT: v_lshlrev_b32_e32 v53, 8, v13
-; SI-NEXT: ; implicit-def: $vgpr13
+; SI-NEXT: buffer_load_dword v7, off, s[0:3], s32 offset:48
+; SI-NEXT: v_lshlrev_b32_e32 v58, 8, v15
+; SI-NEXT: ; implicit-def: $vgpr15
; SI-NEXT: s_waitcnt vmcnt(2)
-; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:732 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:740 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:16
; SI-NEXT: s_waitcnt vmcnt(2)
-; SI-NEXT: buffer_store_dword v8, off, s[0:3], s32 offset:780 ; 4-byte Folded Spill
-; SI-NEXT: ; implicit-def: $vgpr8
+; SI-NEXT: buffer_store_dword v7, off, s[0:3], s32 offset:788 ; 4-byte Folded Spill
+; SI-NEXT: ; implicit-def: $vgpr7
; SI-NEXT: s_waitcnt vmcnt(1)
-; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:784 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:792 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:104
-; SI-NEXT: v_lshlrev_b32_e32 v3, 8, v14
-; SI-NEXT: ; implicit-def: $vgpr14
-; SI-NEXT: ; kill: killed $vgpr14
-; SI-NEXT: ; implicit-def: $vgpr14
; SI-NEXT: s_waitcnt vmcnt(0)
-; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:724 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:732 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:136
; SI-NEXT: s_waitcnt vmcnt(0)
-; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:748 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:756 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:168
; SI-NEXT: s_waitcnt vmcnt(0)
-; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:776 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:784 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:200
; SI-NEXT: s_waitcnt vmcnt(0)
-; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:804 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:812 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:232
; SI-NEXT: s_waitcnt vmcnt(0)
-; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:828 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:836 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:264
; SI-NEXT: s_waitcnt vmcnt(0)
-; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:920 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:928 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:296
; SI-NEXT: s_waitcnt vmcnt(0)
-; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:948 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:952 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:328
; SI-NEXT: s_waitcnt vmcnt(0)
-; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:968 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:972 ; 4-byte Folded Spill
; SI-NEXT: ; implicit-def: $vgpr1
; SI-NEXT: ; kill: killed $vgpr1
; SI-NEXT: ; implicit-def: $vgpr1
@@ -203053,19 +203167,26 @@ define <64 x i16> @bitcast_v128i8_to_v64i16(<128 x i8> %a, i32 %b) {
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:468 ; 4-byte Folded Reload
; SI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:508 ; 4-byte Folded Reload
-; SI-NEXT: buffer_load_dword v20, off, s[0:3], s32 offset:748 ; 4-byte Folded Reload
-; SI-NEXT: s_waitcnt vmcnt(2)
+; SI-NEXT: buffer_load_dword v49, off, s[0:3], s32 offset:936 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v48, off, s[0:3], s32 offset:848 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v10, off, s[0:3], s32 offset:944 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v50, off, s[0:3], s32 offset:976 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v22, off, s[0:3], s32 offset:956 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v3, off, s[0:3], s32 offset:712 ; 4-byte Folded Reload
+; SI-NEXT: s_waitcnt vmcnt(7)
; SI-NEXT: v_and_b32_e32 v1, 0xff, v1
-; SI-NEXT: s_waitcnt vmcnt(1)
-; SI-NEXT: v_or_b32_e32 v12, v1, v2
-; SI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:944 ; 4-byte Folded Reload
+; SI-NEXT: s_waitcnt vmcnt(6)
+; SI-NEXT: v_or_b32_e32 v7, v1, v2
+; SI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:980 ; 4-byte Folded Reload
; SI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:504 ; 4-byte Folded Reload
-; SI-NEXT: v_and_b32_e32 v12, 0xffff, v12
+; SI-NEXT: v_and_b32_e32 v7, 0xffff, v7
; SI-NEXT: s_waitcnt vmcnt(1)
; SI-NEXT: v_and_b32_e32 v1, 0xff, v1
; SI-NEXT: v_lshlrev_b32_e32 v1, 16, v1
; SI-NEXT: s_waitcnt vmcnt(0)
-; SI-NEXT: v_or_b32_e32 v42, v2, v1
+; SI-NEXT: v_or_b32_e32 v1, v2, v1
+; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:468 ; 4-byte Folded Spill
+; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:464 ; 4-byte Folded Reload
; SI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:512 ; 4-byte Folded Reload
; SI-NEXT: s_waitcnt vmcnt(1)
@@ -203074,7 +203195,7 @@ define <64 x i16> @bitcast_v128i8_to_v64i16(<128 x i8> %a, i32 %b) {
; SI-NEXT: s_waitcnt vmcnt(0)
; SI-NEXT: v_or_b32_e32 v5, v2, v1
; SI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:480 ; 4-byte Folded Reload
-; SI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:516 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:520 ; 4-byte Folded Reload
; SI-NEXT: s_waitcnt vmcnt(1)
; SI-NEXT: v_and_b32_e32 v1, 0xff, v1
; SI-NEXT: s_waitcnt vmcnt(0)
@@ -203082,488 +203203,444 @@ define <64 x i16> @bitcast_v128i8_to_v64i16(<128 x i8> %a, i32 %b) {
; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:464 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:460 ; 4-byte Folded Reload
-; SI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:520 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:516 ; 4-byte Folded Reload
; SI-NEXT: s_waitcnt vmcnt(1)
; SI-NEXT: v_and_b32_e32 v1, 0xff, v1
; SI-NEXT: v_lshlrev_b32_e32 v1, 16, v1
-; SI-NEXT: v_or_b32_e32 v7, v16, v1
-; SI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:476 ; 4-byte Folded Reload
; SI-NEXT: s_waitcnt vmcnt(0)
+; SI-NEXT: v_or_b32_e32 v26, v2, v1
+; SI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:476 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:524 ; 4-byte Folded Reload
+; SI-NEXT: s_waitcnt vmcnt(1)
; SI-NEXT: v_and_b32_e32 v1, 0xff, v1
; SI-NEXT: v_lshlrev_b32_e32 v1, 16, v1
+; SI-NEXT: s_waitcnt vmcnt(0)
; SI-NEXT: v_or_b32_e32 v11, v2, v1
; SI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:488 ; 4-byte Folded Reload
-; SI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:524 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:528 ; 4-byte Folded Reload
; SI-NEXT: s_waitcnt vmcnt(1)
; SI-NEXT: v_and_b32_e32 v1, 0xff, v1
; SI-NEXT: s_waitcnt vmcnt(0)
-; SI-NEXT: v_or_b32_e32 v40, v1, v2
+; SI-NEXT: v_or_b32_e32 v42, v1, v2
; SI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:472 ; 4-byte Folded Reload
-; SI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:528 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:532 ; 4-byte Folded Reload
; SI-NEXT: s_waitcnt vmcnt(1)
; SI-NEXT: v_and_b32_e32 v1, 0xff, v1
; SI-NEXT: v_lshlrev_b32_e32 v1, 16, v1
-; SI-NEXT: v_or_b32_e32 v10, v24, v1
+; SI-NEXT: v_or_b32_e32 v39, v24, v1
; SI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:484 ; 4-byte Folded Reload
; SI-NEXT: s_waitcnt vmcnt(0)
; SI-NEXT: v_and_b32_e32 v1, 0xff, v1
; SI-NEXT: v_lshlrev_b32_e32 v1, 16, v1
-; SI-NEXT: v_or_b32_e32 v15, v2, v1
+; SI-NEXT: v_or_b32_e32 v17, v2, v1
; SI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:492 ; 4-byte Folded Reload
-; SI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:532 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:540 ; 4-byte Folded Reload
; SI-NEXT: s_waitcnt vmcnt(1)
; SI-NEXT: v_and_b32_e32 v1, 0xff, v1
; SI-NEXT: s_waitcnt vmcnt(0)
-; SI-NEXT: v_or_b32_e32 v26, v1, v2
+; SI-NEXT: v_or_b32_e32 v55, v1, v2
; SI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:500 ; 4-byte Folded Reload
; SI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:536 ; 4-byte Folded Reload
; SI-NEXT: s_waitcnt vmcnt(1)
; SI-NEXT: v_and_b32_e32 v1, 0xff, v1
; SI-NEXT: v_lshlrev_b32_e32 v1, 16, v1
-; SI-NEXT: v_or_b32_e32 v18, v25, v1
-; SI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:496 ; 4-byte Folded Reload
; SI-NEXT: s_waitcnt vmcnt(0)
+; SI-NEXT: v_or_b32_e32 v18, v2, v1
+; SI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:496 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:544 ; 4-byte Folded Reload
+; SI-NEXT: s_waitcnt vmcnt(1)
; SI-NEXT: v_and_b32_e32 v1, 0xff, v1
; SI-NEXT: v_lshlrev_b32_e32 v1, 16, v1
-; SI-NEXT: v_or_b32_e32 v19, v2, v1
-; SI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:732 ; 4-byte Folded Reload
-; SI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:548 ; 4-byte Folded Reload
+; SI-NEXT: s_waitcnt vmcnt(0)
+; SI-NEXT: v_or_b32_e32 v20, v2, v1
+; SI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:740 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:552 ; 4-byte Folded Reload
; SI-NEXT: s_waitcnt vmcnt(1)
; SI-NEXT: v_and_b32_e32 v1, 0xff, v1
; SI-NEXT: s_waitcnt vmcnt(0)
; SI-NEXT: v_or_b32_e32 v1, v1, v2
; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:460 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:696 ; 4-byte Folded Reload
-; SI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:540 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:708 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:548 ; 4-byte Folded Reload
; SI-NEXT: s_waitcnt vmcnt(1)
; SI-NEXT: v_and_b32_e32 v1, 0xff, v1
; SI-NEXT: v_lshlrev_b32_e32 v1, 16, v1
; SI-NEXT: s_waitcnt vmcnt(0)
-; SI-NEXT: v_or_b32_e32 v30, v2, v1
-; SI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:784 ; 4-byte Folded Reload
+; SI-NEXT: v_or_b32_e32 v28, v2, v1
+; SI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:792 ; 4-byte Folded Reload
; SI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:576 ; 4-byte Folded Reload
; SI-NEXT: s_waitcnt vmcnt(1)
; SI-NEXT: v_and_b32_e32 v1, 0xff, v1
; SI-NEXT: v_lshlrev_b32_e32 v1, 16, v1
; SI-NEXT: s_waitcnt vmcnt(0)
-; SI-NEXT: v_or_b32_e32 v21, v2, v1
-; SI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:772 ; 4-byte Folded Reload
-; SI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:588 ; 4-byte Folded Reload
+; SI-NEXT: v_or_b32_e32 v27, v2, v1
+; SI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:780 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:592 ; 4-byte Folded Reload
; SI-NEXT: s_waitcnt vmcnt(1)
; SI-NEXT: v_and_b32_e32 v1, 0xff, v1
; SI-NEXT: s_waitcnt vmcnt(0)
; SI-NEXT: v_or_b32_e32 v29, v1, v2
-; SI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:760 ; 4-byte Folded Reload
-; SI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:584 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:768 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:588 ; 4-byte Folded Reload
; SI-NEXT: s_waitcnt vmcnt(1)
; SI-NEXT: v_and_b32_e32 v1, 0xff, v1
; SI-NEXT: v_lshlrev_b32_e32 v1, 16, v1
; SI-NEXT: s_waitcnt vmcnt(0)
-; SI-NEXT: v_or_b32_e32 v33, v2, v1
-; SI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:780 ; 4-byte Folded Reload
-; SI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:592 ; 4-byte Folded Reload
+; SI-NEXT: v_or_b32_e32 v30, v2, v1
+; SI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:788 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:596 ; 4-byte Folded Reload
; SI-NEXT: s_waitcnt vmcnt(1)
; SI-NEXT: v_and_b32_e32 v1, 0xff, v1
; SI-NEXT: v_lshlrev_b32_e32 v1, 16, v1
; SI-NEXT: s_waitcnt vmcnt(0)
-; SI-NEXT: v_or_b32_e32 v27, v2, v1
+; SI-NEXT: v_or_b32_e32 v53, v2, v1
; SI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:560 ; 4-byte Folded Reload
-; SI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:620 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:624 ; 4-byte Folded Reload
; SI-NEXT: s_waitcnt vmcnt(1)
; SI-NEXT: v_and_b32_e32 v1, 0xff, v1
; SI-NEXT: s_waitcnt vmcnt(0)
-; SI-NEXT: v_or_b32_e32 v1, v1, v2
-; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:468 ; 4-byte Folded Spill
-; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:552 ; 4-byte Folded Reload
-; SI-NEXT: s_waitcnt vmcnt(0)
-; SI-NEXT: v_and_b32_e32 v2, 0xff, v1
-; SI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:612 ; 4-byte Folded Reload
-; SI-NEXT: v_lshlrev_b32_e32 v2, 16, v2
+; SI-NEXT: v_or_b32_e32 v19, v1, v2
+; SI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:556 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:616 ; 4-byte Folded Reload
+; SI-NEXT: s_waitcnt vmcnt(1)
+; SI-NEXT: v_and_b32_e32 v1, 0xff, v1
+; SI-NEXT: v_lshlrev_b32_e32 v1, 16, v1
; SI-NEXT: s_waitcnt vmcnt(0)
-; SI-NEXT: v_or_b32_e32 v38, v1, v2
+; SI-NEXT: v_or_b32_e32 v33, v2, v1
; SI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:564 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:628 ; 4-byte Folded Reload
+; SI-NEXT: s_waitcnt vmcnt(1)
+; SI-NEXT: v_and_b32_e32 v1, 0xff, v1
+; SI-NEXT: v_lshlrev_b32_e32 v1, 16, v1
; SI-NEXT: s_waitcnt vmcnt(0)
-; SI-NEXT: v_and_b32_e32 v2, 0xff, v1
-; SI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:624 ; 4-byte Folded Reload
-; SI-NEXT: v_lshlrev_b32_e32 v2, 16, v2
-; SI-NEXT: s_waitcnt vmcnt(0)
-; SI-NEXT: v_or_b32_e32 v52, v1, v2
-; SI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:800 ; 4-byte Folded Reload
-; SI-NEXT: s_waitcnt vmcnt(0)
-; SI-NEXT: v_and_b32_e32 v2, 0xff, v1
-; SI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:648 ; 4-byte Folded Reload
+; SI-NEXT: v_or_b32_e32 v43, v2, v1
+; SI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:808 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:656 ; 4-byte Folded Reload
+; SI-NEXT: s_waitcnt vmcnt(1)
+; SI-NEXT: v_and_b32_e32 v1, 0xff, v1
; SI-NEXT: s_waitcnt vmcnt(0)
-; SI-NEXT: v_or_b32_e32 v1, v2, v1
-; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:472 ; 4-byte Folded Spill
-; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:636 ; 4-byte Folded Reload
-; SI-NEXT: v_and_b32_e32 v2, 0xff, v23
-; SI-NEXT: v_lshlrev_b32_e32 v2, 16, v2
+; SI-NEXT: v_or_b32_e32 v15, v1, v2
+; SI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:580 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:644 ; 4-byte Folded Reload
+; SI-NEXT: s_waitcnt vmcnt(1)
+; SI-NEXT: v_and_b32_e32 v1, 0xff, v1
+; SI-NEXT: v_lshlrev_b32_e32 v1, 16, v1
; SI-NEXT: s_waitcnt vmcnt(0)
-; SI-NEXT: v_or_b32_e32 v2, v1, v2
+; SI-NEXT: v_or_b32_e32 v38, v2, v1
; SI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:572 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:672 ; 4-byte Folded Reload
+; SI-NEXT: s_waitcnt vmcnt(1)
+; SI-NEXT: v_and_b32_e32 v1, 0xff, v1
+; SI-NEXT: v_lshlrev_b32_e32 v1, 16, v1
; SI-NEXT: s_waitcnt vmcnt(0)
-; SI-NEXT: v_and_b32_e32 v6, 0xff, v1
-; SI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:664 ; 4-byte Folded Reload
-; SI-NEXT: v_lshlrev_b32_e32 v6, 16, v6
+; SI-NEXT: v_or_b32_e32 v45, v2, v1
+; SI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:604 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:704 ; 4-byte Folded Reload
+; SI-NEXT: s_waitcnt vmcnt(1)
+; SI-NEXT: v_and_b32_e32 v1, 0xff, v1
; SI-NEXT: s_waitcnt vmcnt(0)
-; SI-NEXT: v_or_b32_e32 v55, v1, v6
+; SI-NEXT: v_or_b32_e32 v47, v1, v2
; SI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:600 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:700 ; 4-byte Folded Reload
+; SI-NEXT: s_waitcnt vmcnt(1)
+; SI-NEXT: v_and_b32_e32 v1, 0xff, v1
+; SI-NEXT: v_lshlrev_b32_e32 v1, 16, v1
; SI-NEXT: s_waitcnt vmcnt(0)
-; SI-NEXT: v_and_b32_e32 v6, 0xff, v1
-; SI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:692 ; 4-byte Folded Reload
-; SI-NEXT: s_waitcnt vmcnt(0)
-; SI-NEXT: v_or_b32_e32 v17, v6, v1
-; SI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:596 ; 4-byte Folded Reload
-; SI-NEXT: s_waitcnt vmcnt(0)
-; SI-NEXT: v_and_b32_e32 v6, 0xff, v1
-; SI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:688 ; 4-byte Folded Reload
-; SI-NEXT: v_lshlrev_b32_e32 v6, 16, v6
-; SI-NEXT: s_waitcnt vmcnt(0)
-; SI-NEXT: v_or_b32_e32 v14, v1, v6
-; SI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:608 ; 4-byte Folded Reload
-; SI-NEXT: s_waitcnt vmcnt(0)
-; SI-NEXT: v_and_b32_e32 v6, 0xff, v1
-; SI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:700 ; 4-byte Folded Reload
-; SI-NEXT: v_lshlrev_b32_e32 v6, 16, v6
-; SI-NEXT: s_waitcnt vmcnt(0)
-; SI-NEXT: v_or_b32_e32 v44, v1, v6
-; SI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:632 ; 4-byte Folded Reload
-; SI-NEXT: s_waitcnt vmcnt(0)
-; SI-NEXT: v_and_b32_e32 v6, 0xff, v1
-; SI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:720 ; 4-byte Folded Reload
-; SI-NEXT: s_waitcnt vmcnt(0)
-; SI-NEXT: v_or_b32_e32 v45, v6, v1
-; SI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:672 ; 4-byte Folded Reload
+; SI-NEXT: v_or_b32_e32 v2, v2, v1
+; SI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:612 ; 4-byte Folded Reload
; SI-NEXT: s_waitcnt vmcnt(0)
-; SI-NEXT: v_and_b32_e32 v6, 0xff, v1
-; SI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:728 ; 4-byte Folded Reload
-; SI-NEXT: v_lshlrev_b32_e32 v6, 16, v6
+; SI-NEXT: v_and_b32_e32 v1, 0xff, v1
+; SI-NEXT: v_lshlrev_b32_e32 v1, 16, v1
+; SI-NEXT: v_or_b32_e32 v57, v3, v1
+; SI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:640 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v3, off, s[0:3], s32 offset:728 ; 4-byte Folded Reload
+; SI-NEXT: s_waitcnt vmcnt(1)
+; SI-NEXT: v_and_b32_e32 v1, 0xff, v1
; SI-NEXT: s_waitcnt vmcnt(0)
-; SI-NEXT: v_or_b32_e32 v25, v1, v6
-; SI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:644 ; 4-byte Folded Reload
+; SI-NEXT: v_or_b32_e32 v59, v1, v3
+; SI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:680 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v3, off, s[0:3], s32 offset:736 ; 4-byte Folded Reload
+; SI-NEXT: s_waitcnt vmcnt(1)
+; SI-NEXT: v_and_b32_e32 v1, 0xff, v1
+; SI-NEXT: v_lshlrev_b32_e32 v1, 16, v1
; SI-NEXT: s_waitcnt vmcnt(0)
-; SI-NEXT: v_and_b32_e32 v6, 0xff, v1
-; SI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:736 ; 4-byte Folded Reload
-; SI-NEXT: v_lshlrev_b32_e32 v6, 16, v6
+; SI-NEXT: v_or_b32_e32 v4, v3, v1
+; SI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:652 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v3, off, s[0:3], s32 offset:744 ; 4-byte Folded Reload
+; SI-NEXT: s_waitcnt vmcnt(1)
+; SI-NEXT: v_and_b32_e32 v1, 0xff, v1
+; SI-NEXT: v_lshlrev_b32_e32 v1, 16, v1
; SI-NEXT: s_waitcnt vmcnt(0)
-; SI-NEXT: v_or_b32_e32 v47, v1, v6
-; SI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:704 ; 4-byte Folded Reload
+; SI-NEXT: v_or_b32_e32 v62, v3, v1
+; SI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:716 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v3, off, s[0:3], s32 offset:760 ; 4-byte Folded Reload
+; SI-NEXT: s_waitcnt vmcnt(1)
+; SI-NEXT: v_and_b32_e32 v1, 0xff, v1
; SI-NEXT: s_waitcnt vmcnt(0)
-; SI-NEXT: v_and_b32_e32 v6, 0xff, v1
+; SI-NEXT: v_or_b32_e32 v44, v1, v3
; SI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:752 ; 4-byte Folded Reload
; SI-NEXT: s_waitcnt vmcnt(0)
-; SI-NEXT: v_or_b32_e32 v58, v6, v1
-; SI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:744 ; 4-byte Folded Reload
-; SI-NEXT: s_waitcnt vmcnt(0)
-; SI-NEXT: v_and_b32_e32 v6, 0xff, v1
-; SI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:796 ; 4-byte Folded Reload
-; SI-NEXT: v_lshlrev_b32_e32 v6, 16, v6
-; SI-NEXT: s_waitcnt vmcnt(0)
-; SI-NEXT: v_or_b32_e32 v16, v1, v6
-; SI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:708 ; 4-byte Folded Reload
-; SI-NEXT: s_waitcnt vmcnt(0)
-; SI-NEXT: v_and_b32_e32 v6, 0xff, v1
-; SI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:808 ; 4-byte Folded Reload
-; SI-NEXT: v_lshlrev_b32_e32 v6, 16, v6
+; SI-NEXT: v_and_b32_e32 v3, 0xff, v1
+; SI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:804 ; 4-byte Folded Reload
+; SI-NEXT: v_lshlrev_b32_e32 v3, 16, v3
; SI-NEXT: s_waitcnt vmcnt(0)
-; SI-NEXT: v_or_b32_e32 v59, v1, v6
-; SI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:764 ; 4-byte Folded Reload
+; SI-NEXT: v_or_b32_e32 v8, v1, v3
+; SI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:720 ; 4-byte Folded Reload
; SI-NEXT: s_waitcnt vmcnt(0)
-; SI-NEXT: v_and_b32_e32 v6, 0xff, v1
-; SI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:832 ; 4-byte Folded Reload
+; SI-NEXT: v_and_b32_e32 v3, 0xff, v1
+; SI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:816 ; 4-byte Folded Reload
+; SI-NEXT: v_lshlrev_b32_e32 v3, 16, v3
; SI-NEXT: s_waitcnt vmcnt(0)
-; SI-NEXT: v_or_b32_e32 v62, v6, v1
-; SI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:788 ; 4-byte Folded Reload
+; SI-NEXT: v_or_b32_e32 v21, v1, v3
+; SI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:772 ; 4-byte Folded Reload
; SI-NEXT: s_waitcnt vmcnt(0)
-; SI-NEXT: v_and_b32_e32 v6, 0xff, v1
-; SI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:836 ; 4-byte Folded Reload
-; SI-NEXT: v_lshlrev_b32_e32 v6, 16, v6
+; SI-NEXT: v_and_b32_e32 v3, 0xff, v1
+; SI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:840 ; 4-byte Folded Reload
; SI-NEXT: s_waitcnt vmcnt(0)
-; SI-NEXT: v_or_b32_e32 v24, v1, v6
-; SI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:768 ; 4-byte Folded Reload
+; SI-NEXT: v_or_b32_e32 v6, v3, v1
+; SI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:796 ; 4-byte Folded Reload
+; SI-NEXT: v_and_b32_e32 v6, 0xffff, v6
; SI-NEXT: s_waitcnt vmcnt(0)
-; SI-NEXT: v_and_b32_e32 v6, 0xff, v1
+; SI-NEXT: v_and_b32_e32 v3, 0xff, v1
; SI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:844 ; 4-byte Folded Reload
-; SI-NEXT: v_lshlrev_b32_e32 v6, 16, v6
-; SI-NEXT: s_waitcnt vmcnt(0)
-; SI-NEXT: v_or_b32_e32 v13, v1, v6
-; SI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:812 ; 4-byte Folded Reload
+; SI-NEXT: v_lshlrev_b32_e32 v3, 16, v3
; SI-NEXT: s_waitcnt vmcnt(0)
-; SI-NEXT: v_and_b32_e32 v6, 0xff, v1
-; SI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:932 ; 4-byte Folded Reload
+; SI-NEXT: v_or_b32_e32 v3, v1, v3
+; SI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:776 ; 4-byte Folded Reload
; SI-NEXT: s_waitcnt vmcnt(0)
-; SI-NEXT: v_or_b32_e32 v6, v6, v1
-; SI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:824 ; 4-byte Folded Reload
-; SI-NEXT: v_and_b32_e32 v6, 0xffff, v6
+; SI-NEXT: v_and_b32_e32 v12, 0xff, v1
+; SI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:852 ; 4-byte Folded Reload
+; SI-NEXT: v_lshlrev_b32_e32 v12, 16, v12
; SI-NEXT: s_waitcnt vmcnt(0)
-; SI-NEXT: v_and_b32_e32 v8, 0xff, v1
-; SI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:936 ; 4-byte Folded Reload
-; SI-NEXT: v_lshlrev_b32_e32 v8, 16, v8
+; SI-NEXT: v_or_b32_e32 v16, v1, v12
+; SI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:820 ; 4-byte Folded Reload
; SI-NEXT: s_waitcnt vmcnt(0)
-; SI-NEXT: v_or_b32_e32 v23, v1, v8
-; SI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:816 ; 4-byte Folded Reload
+; SI-NEXT: v_and_b32_e32 v12, 0xff, v1
+; SI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:940 ; 4-byte Folded Reload
; SI-NEXT: s_waitcnt vmcnt(0)
-; SI-NEXT: v_and_b32_e32 v8, 0xff, v1
-; SI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:952 ; 4-byte Folded Reload
-; SI-NEXT: v_lshlrev_b32_e32 v8, 16, v8
+; SI-NEXT: v_or_b32_e32 v24, v12, v1
+; SI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:832 ; 4-byte Folded Reload
; SI-NEXT: s_waitcnt vmcnt(0)
-; SI-NEXT: v_or_b32_e32 v36, v1, v8
-; SI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:848 ; 4-byte Folded Reload
+; SI-NEXT: v_and_b32_e32 v12, 0xff, v1
+; SI-NEXT: v_lshlrev_b32_e32 v12, 16, v12
+; SI-NEXT: v_mov_b32_e32 v1, v32
+; SI-NEXT: v_or_b32_e32 v32, v10, v12
+; SI-NEXT: buffer_load_dword v10, off, s[0:3], s32 offset:824 ; 4-byte Folded Reload
; SI-NEXT: s_waitcnt vmcnt(0)
-; SI-NEXT: v_and_b32_e32 v8, 0xff, v1
-; SI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:924 ; 4-byte Folded Reload
-; SI-NEXT: v_or_b32_e32 v8, v8, v35
+; SI-NEXT: v_and_b32_e32 v12, 0xff, v10
+; SI-NEXT: v_lshlrev_b32_e32 v12, 16, v12
+; SI-NEXT: v_mov_b32_e32 v10, v34
+; SI-NEXT: v_or_b32_e32 v34, v22, v12
+; SI-NEXT: buffer_load_dword v12, off, s[0:3], s32 offset:856 ; 4-byte Folded Reload
; SI-NEXT: s_waitcnt vmcnt(0)
-; SI-NEXT: v_and_b32_e32 v22, 0xff, v1
-; SI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:916 ; 4-byte Folded Reload
-; SI-NEXT: v_lshlrev_b32_e32 v22, 16, v22
-; SI-NEXT: v_or_b32_e32 v35, v37, v22
+; SI-NEXT: v_and_b32_e32 v12, 0xff, v12
+; SI-NEXT: v_or_b32_e32 v36, v12, v35
+; SI-NEXT: buffer_load_dword v12, off, s[0:3], s32 offset:932 ; 4-byte Folded Reload
; SI-NEXT: s_waitcnt vmcnt(0)
-; SI-NEXT: v_and_b32_e32 v22, 0xff, v1
-; SI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:956 ; 4-byte Folded Reload
-; SI-NEXT: v_lshlrev_b32_e32 v22, 16, v22
-; SI-NEXT: v_or_b32_e32 v37, v51, v22
+; SI-NEXT: v_and_b32_e32 v12, 0xff, v12
+; SI-NEXT: v_lshlrev_b32_e32 v12, 16, v12
+; SI-NEXT: v_or_b32_e32 v35, v37, v12
+; SI-NEXT: buffer_load_dword v12, off, s[0:3], s32 offset:924 ; 4-byte Folded Reload
; SI-NEXT: s_waitcnt vmcnt(0)
-; SI-NEXT: v_and_b32_e32 v22, 0xff, v1
-; SI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:964 ; 4-byte Folded Reload
-; SI-NEXT: v_or_b32_e32 v51, v22, v63
-; SI-NEXT: v_and_b32_e32 v22, 0xff, v34
-; SI-NEXT: v_lshlrev_b32_e32 v22, 16, v22
-; SI-NEXT: v_or_b32_e32 v34, v56, v22
+; SI-NEXT: v_and_b32_e32 v12, 0xff, v12
+; SI-NEXT: v_lshlrev_b32_e32 v12, 16, v12
+; SI-NEXT: v_or_b32_e32 v37, v51, v12
+; SI-NEXT: buffer_load_dword v12, off, s[0:3], s32 offset:960 ; 4-byte Folded Reload
; SI-NEXT: s_waitcnt vmcnt(0)
-; SI-NEXT: v_and_b32_e32 v22, 0xff, v1
-; SI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:628 ; 4-byte Folded Reload
-; SI-NEXT: v_lshlrev_b32_e32 v22, 16, v22
-; SI-NEXT: v_or_b32_e32 v56, v60, v22
-; SI-NEXT: v_and_b32_e32 v22, 0xff, v54
-; SI-NEXT: v_or_b32_e32 v54, v22, v4
-; SI-NEXT: v_and_b32_e32 v22, 0xff, v41
-; SI-NEXT: v_lshlrev_b32_e32 v22, 16, v22
-; SI-NEXT: v_or_b32_e32 v9, v9, v22
-; SI-NEXT: v_and_b32_e32 v4, 0xff, v57
-; SI-NEXT: v_lshlrev_b32_e32 v4, 16, v4
-; SI-NEXT: v_or_b32_e32 v4, v32, v4
+; SI-NEXT: v_and_b32_e32 v12, 0xff, v12
+; SI-NEXT: v_or_b32_e32 v51, v12, v60
+; SI-NEXT: v_and_b32_e32 v12, 0xff, v14
+; SI-NEXT: v_lshlrev_b32_e32 v12, 16, v12
+; SI-NEXT: v_or_b32_e32 v14, v31, v12
+; SI-NEXT: buffer_load_dword v12, off, s[0:3], s32 offset:968 ; 4-byte Folded Reload
; SI-NEXT: s_waitcnt vmcnt(0)
-; SI-NEXT: v_and_b32_e32 v22, 0xff, v1
-; SI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:544 ; 4-byte Folded Reload
+; SI-NEXT: v_and_b32_e32 v12, 0xff, v12
+; SI-NEXT: v_lshlrev_b32_e32 v12, 16, v12
+; SI-NEXT: v_or_b32_e32 v31, v61, v12
+; SI-NEXT: v_and_b32_e32 v12, 0xff, v54
+; SI-NEXT: v_or_b32_e32 v54, v12, v23
+; SI-NEXT: v_and_b32_e32 v12, 0xff, v46
+; SI-NEXT: v_lshlrev_b32_e32 v12, 16, v12
+; SI-NEXT: v_or_b32_e32 v23, v25, v12
+; SI-NEXT: v_and_b32_e32 v12, 0xff, v41
+; SI-NEXT: v_lshlrev_b32_e32 v12, 16, v12
+; SI-NEXT: v_or_b32_e32 v13, v13, v12
+; SI-NEXT: buffer_load_dword v12, off, s[0:3], s32 offset:632 ; 4-byte Folded Reload
; SI-NEXT: s_waitcnt vmcnt(0)
-; SI-NEXT: v_or_b32_e32 v28, v22, v1
-; SI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:640 ; 4-byte Folded Reload
+; SI-NEXT: v_and_b32_e32 v12, 0xff, v12
+; SI-NEXT: v_or_b32_e32 v12, v12, v1
+; SI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:636 ; 4-byte Folded Reload
; SI-NEXT: s_waitcnt vmcnt(0)
; SI-NEXT: v_and_b32_e32 v22, 0xff, v1
-; SI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:556 ; 4-byte Folded Reload
-; SI-NEXT: s_waitcnt vmcnt(0)
-; SI-NEXT: v_or_b32_e32 v32, v22, v1
-; SI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:652 ; 4-byte Folded Reload
-; SI-NEXT: s_waitcnt vmcnt(0)
+; SI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:648 ; 4-byte Folded Reload
+; SI-NEXT: v_or_b32_e32 v25, v22, v10
+; SI-NEXT: buffer_load_dword v10, off, s[0:3], s32 offset:756 ; 4-byte Folded Reload
+; SI-NEXT: s_waitcnt vmcnt(1)
; SI-NEXT: v_and_b32_e32 v22, 0xff, v1
; SI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:568 ; 4-byte Folded Reload
; SI-NEXT: s_waitcnt vmcnt(0)
-; SI-NEXT: v_or_b32_e32 v39, v22, v1
-; SI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:656 ; 4-byte Folded Reload
+; SI-NEXT: v_or_b32_e32 v40, v22, v1
+; SI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:660 ; 4-byte Folded Reload
; SI-NEXT: s_waitcnt vmcnt(0)
; SI-NEXT: v_and_b32_e32 v22, 0xff, v1
-; SI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:604 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:608 ; 4-byte Folded Reload
; SI-NEXT: s_waitcnt vmcnt(0)
; SI-NEXT: v_or_b32_e32 v41, v22, v1
-; SI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:580 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:584 ; 4-byte Folded Reload
; SI-NEXT: s_waitcnt vmcnt(0)
; SI-NEXT: v_and_b32_e32 v22, 0xff, v1
-; SI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:712 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:724 ; 4-byte Folded Reload
; SI-NEXT: s_waitcnt vmcnt(0)
-; SI-NEXT: v_or_b32_e32 v57, v22, v1
-; SI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:616 ; 4-byte Folded Reload
+; SI-NEXT: v_or_b32_e32 v46, v22, v1
+; SI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:620 ; 4-byte Folded Reload
; SI-NEXT: s_waitcnt vmcnt(0)
; SI-NEXT: v_and_b32_e32 v22, 0xff, v1
-; SI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:740 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:748 ; 4-byte Folded Reload
; SI-NEXT: s_waitcnt vmcnt(0)
; SI-NEXT: v_or_b32_e32 v60, v22, v1
-; SI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:668 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:676 ; 4-byte Folded Reload
; SI-NEXT: s_waitcnt vmcnt(0)
; SI-NEXT: v_and_b32_e32 v22, 0xff, v1
-; SI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:756 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:764 ; 4-byte Folded Reload
; SI-NEXT: s_waitcnt vmcnt(0)
-; SI-NEXT: v_or_b32_e32 v63, v22, v1
-; SI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:724 ; 4-byte Folded Reload
+; SI-NEXT: v_or_b32_e32 v61, v22, v1
+; SI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:732 ; 4-byte Folded Reload
; SI-NEXT: s_waitcnt vmcnt(0)
; SI-NEXT: v_and_b32_e32 v22, 0xff, v1
-; SI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:792 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:800 ; 4-byte Folded Reload
; SI-NEXT: s_waitcnt vmcnt(0)
; SI-NEXT: v_or_b32_e32 v1, v22, v1
-; SI-NEXT: v_and_b32_e32 v22, 0xff, v20
-; SI-NEXT: buffer_load_dword v20, off, s[0:3], s32 offset:820 ; 4-byte Folded Reload
+; SI-NEXT: v_and_b32_e32 v22, 0xff, v10
+; SI-NEXT: buffer_load_dword v10, off, s[0:3], s32 offset:828 ; 4-byte Folded Reload
; SI-NEXT: v_and_b32_e32 v1, 0xffff, v1
-; SI-NEXT: v_or_b32_e32 v1, v1, v55
-; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:684 ; 4-byte Folded Spill
+; SI-NEXT: v_or_b32_e32 v1, v1, v45
+; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:692 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt vmcnt(1)
-; SI-NEXT: v_or_b32_e32 v48, v22, v20
-; SI-NEXT: buffer_load_dword v20, off, s[0:3], s32 offset:776 ; 4-byte Folded Reload
-; SI-NEXT: s_waitcnt vmcnt(0)
-; SI-NEXT: v_and_b32_e32 v22, 0xff, v20
-; SI-NEXT: buffer_load_dword v20, off, s[0:3], s32 offset:840 ; 4-byte Folded Reload
+; SI-NEXT: v_or_b32_e32 v10, v22, v10
+; SI-NEXT: buffer_load_dword v22, off, s[0:3], s32 offset:784 ; 4-byte Folded Reload
; SI-NEXT: s_waitcnt vmcnt(0)
-; SI-NEXT: v_or_b32_e32 v46, v22, v20
-; SI-NEXT: buffer_load_dword v20, off, s[0:3], s32 offset:804 ; 4-byte Folded Reload
-; SI-NEXT: s_waitcnt vmcnt(0)
-; SI-NEXT: v_and_b32_e32 v22, 0xff, v20
-; SI-NEXT: buffer_load_dword v20, off, s[0:3], s32 offset:928 ; 4-byte Folded Reload
-; SI-NEXT: s_waitcnt vmcnt(0)
-; SI-NEXT: v_or_b32_e32 v49, v22, v20
-; SI-NEXT: buffer_load_dword v20, off, s[0:3], s32 offset:828 ; 4-byte Folded Reload
-; SI-NEXT: s_waitcnt vmcnt(0)
-; SI-NEXT: v_and_b32_e32 v22, 0xff, v20
-; SI-NEXT: buffer_load_dword v20, off, s[0:3], s32 offset:960 ; 4-byte Folded Reload
+; SI-NEXT: v_and_b32_e32 v22, 0xff, v22
+; SI-NEXT: v_or_b32_e32 v48, v22, v48
+; SI-NEXT: buffer_load_dword v22, off, s[0:3], s32 offset:812 ; 4-byte Folded Reload
; SI-NEXT: s_waitcnt vmcnt(0)
-; SI-NEXT: v_or_b32_e32 v50, v22, v20
-; SI-NEXT: buffer_load_dword v20, off, s[0:3], s32 offset:920 ; 4-byte Folded Reload
+; SI-NEXT: v_and_b32_e32 v22, 0xff, v22
+; SI-NEXT: v_or_b32_e32 v56, v22, v49
+; SI-NEXT: buffer_load_dword v22, off, s[0:3], s32 offset:836 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v49, off, s[0:3], s32 offset:964 ; 4-byte Folded Reload
+; SI-NEXT: s_waitcnt vmcnt(1)
+; SI-NEXT: v_and_b32_e32 v22, 0xff, v22
; SI-NEXT: s_waitcnt vmcnt(0)
-; SI-NEXT: v_and_b32_e32 v22, 0xff, v20
-; SI-NEXT: buffer_load_dword v20, off, s[0:3], s32 offset:972 ; 4-byte Folded Reload
+; SI-NEXT: v_or_b32_e32 v49, v22, v49
+; SI-NEXT: buffer_load_dword v22, off, s[0:3], s32 offset:928 ; 4-byte Folded Reload
; SI-NEXT: s_waitcnt vmcnt(0)
-; SI-NEXT: v_or_b32_e32 v20, v22, v20
-; SI-NEXT: buffer_load_dword v22, off, s[0:3], s32 offset:948 ; 4-byte Folded Reload
+; SI-NEXT: v_and_b32_e32 v22, 0xff, v22
+; SI-NEXT: v_or_b32_e32 v50, v22, v50
+; SI-NEXT: buffer_load_dword v22, off, s[0:3], s32 offset:952 ; 4-byte Folded Reload
; SI-NEXT: s_waitcnt vmcnt(0)
; SI-NEXT: v_and_b32_e32 v22, 0xff, v22
-; SI-NEXT: v_or_b32_e32 v53, v22, v53
-; SI-NEXT: buffer_load_dword v22, off, s[0:3], s32 offset:968 ; 4-byte Folded Reload
+; SI-NEXT: v_or_b32_e32 v58, v22, v58
+; SI-NEXT: buffer_load_dword v22, off, s[0:3], s32 offset:972 ; 4-byte Folded Reload
; SI-NEXT: s_waitcnt vmcnt(0)
; SI-NEXT: v_and_b32_e32 v22, 0xff, v22
-; SI-NEXT: v_or_b32_e32 v3, v22, v3
-; SI-NEXT: v_and_b32_e32 v22, 0xff, v61
-; SI-NEXT: v_mov_b32_e32 v61, v42
-; SI-NEXT: v_or_b32_e32 v31, v22, v31
-; SI-NEXT: v_or_b32_e32 v22, v12, v61
-; SI-NEXT: v_and_b32_e32 v12, 0xffff, v28
-; SI-NEXT: v_or_b32_e32 v43, v12, v5
+; SI-NEXT: v_or_b32_e32 v9, v22, v9
+; SI-NEXT: v_and_b32_e32 v22, 0xff, v52
+; SI-NEXT: v_or_b32_e32 v52, v22, v63
+; SI-NEXT: buffer_load_dword v63, off, s[0:3], s32 offset:468 ; 4-byte Folded Reload
+; SI-NEXT: s_waitcnt vmcnt(0)
+; SI-NEXT: v_or_b32_e32 v22, v7, v63
+; SI-NEXT: v_and_b32_e32 v7, 0xffff, v12
+; SI-NEXT: v_or_b32_e32 v12, v7, v5
; SI-NEXT: v_alignbit_b32 v5, v22, v5, 16
-; SI-NEXT: buffer_store_dword v5, off, s[0:3], s32 offset:852 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v5, off, s[0:3], s32 offset:860 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: buffer_load_dword v5, off, s[0:3], s32 offset:464 ; 4-byte Folded Reload
-; SI-NEXT: v_mov_b32_e32 v12, v7
-; SI-NEXT: v_and_b32_e32 v7, 0xffff, v32
+; SI-NEXT: v_and_b32_e32 v7, 0xffff, v25
; SI-NEXT: v_or_b32_e32 v7, v7, v11
-; SI-NEXT: v_and_b32_e32 v32, 0xffff, v49
-; SI-NEXT: v_or_b32_e32 v32, v32, v59
-; SI-NEXT: buffer_store_dword v32, off, s[0:3], s32 offset:660 ; 4-byte Folded Spill
-; SI-NEXT: v_and_b32_e32 v3, 0xffff, v3
-; SI-NEXT: s_waitcnt vmcnt(1)
+; SI-NEXT: v_mov_b32_e32 v25, v39
+; SI-NEXT: s_waitcnt vmcnt(0)
; SI-NEXT: v_and_b32_e32 v5, 0xffff, v5
-; SI-NEXT: v_or_b32_e32 v5, v5, v12
+; SI-NEXT: v_or_b32_e32 v5, v5, v26
; SI-NEXT: v_alignbit_b32 v11, v5, v11, 16
-; SI-NEXT: buffer_store_dword v11, off, s[0:3], s32 offset:856 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v11, off, s[0:3], s32 offset:864 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
+; SI-NEXT: v_and_b32_e32 v11, 0xffff, v42
+; SI-NEXT: v_or_b32_e32 v42, v11, v25
; SI-NEXT: v_and_b32_e32 v11, 0xffff, v40
-; SI-NEXT: v_or_b32_e32 v42, v11, v10
-; SI-NEXT: v_and_b32_e32 v11, 0xffff, v39
-; SI-NEXT: v_or_b32_e32 v40, v11, v15
-; SI-NEXT: v_alignbit_b32 v11, v42, v15, 16
-; SI-NEXT: buffer_store_dword v11, off, s[0:3], s32 offset:860 ; 4-byte Folded Spill
-; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_and_b32_e32 v11, 0xffff, v26
-; SI-NEXT: v_or_b32_e32 v26, v11, v18
+; SI-NEXT: v_or_b32_e32 v40, v11, v17
+; SI-NEXT: v_alignbit_b32 v11, v42, v17, 16
+; SI-NEXT: buffer_store_dword v11, off, s[0:3], s32 offset:868 ; 4-byte Folded Spill
+; SI-NEXT: s_waitcnt expcnt(0)
+; SI-NEXT: v_and_b32_e32 v11, 0xffff, v55
+; SI-NEXT: v_or_b32_e32 v55, v11, v18
; SI-NEXT: v_and_b32_e32 v11, 0xffff, v41
-; SI-NEXT: v_or_b32_e32 v39, v11, v19
-; SI-NEXT: v_alignbit_b32 v11, v26, v19, 16
-; SI-NEXT: buffer_store_dword v11, off, s[0:3], s32 offset:864 ; 4-byte Folded Spill
+; SI-NEXT: v_or_b32_e32 v39, v11, v20
+; SI-NEXT: v_alignbit_b32 v11, v55, v20, 16
+; SI-NEXT: buffer_store_dword v11, off, s[0:3], s32 offset:872 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: buffer_load_dword v11, off, s[0:3], s32 offset:460 ; 4-byte Folded Reload
-; SI-NEXT: v_lshrrev_b32_e32 v12, 16, v12
-; SI-NEXT: ; implicit-def: $vgpr41
+; SI-NEXT: v_mov_b32_e32 v41, v28
; SI-NEXT: s_waitcnt vmcnt(0)
; SI-NEXT: v_and_b32_e32 v11, 0xffff, v11
-; SI-NEXT: v_or_b32_e32 v28, v11, v30
-; SI-NEXT: v_and_b32_e32 v11, 0xffff, v57
-; SI-NEXT: v_or_b32_e32 v11, v11, v21
-; SI-NEXT: buffer_store_dword v11, off, s[0:3], s32 offset:716 ; 4-byte Folded Spill
-; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_alignbit_b32 v11, v28, v21, 16
-; SI-NEXT: buffer_store_dword v11, off, s[0:3], s32 offset:868 ; 4-byte Folded Spill
+; SI-NEXT: v_or_b32_e32 v28, v11, v41
+; SI-NEXT: v_and_b32_e32 v11, 0xffff, v46
+; SI-NEXT: v_or_b32_e32 v20, v11, v27
+; SI-NEXT: v_alignbit_b32 v11, v28, v27, 16
+; SI-NEXT: buffer_store_dword v11, off, s[0:3], s32 offset:876 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_and_b32_e32 v11, 0xffff, v29
-; SI-NEXT: v_or_b32_e32 v29, v11, v33
+; SI-NEXT: v_or_b32_e32 v29, v11, v30
; SI-NEXT: v_and_b32_e32 v11, 0xffff, v60
-; SI-NEXT: v_or_b32_e32 v21, v11, v27
-; SI-NEXT: v_alignbit_b32 v11, v29, v27, 16
-; SI-NEXT: buffer_store_dword v11, off, s[0:3], s32 offset:872 ; 4-byte Folded Spill
+; SI-NEXT: v_or_b32_e32 v27, v11, v53
+; SI-NEXT: v_alignbit_b32 v11, v29, v53, 16
+; SI-NEXT: buffer_store_dword v11, off, s[0:3], s32 offset:880 ; 4-byte Folded Spill
+; SI-NEXT: s_waitcnt expcnt(0)
+; SI-NEXT: v_and_b32_e32 v11, 0xffff, v19
+; SI-NEXT: v_or_b32_e32 v19, v11, v33
+; SI-NEXT: v_and_b32_e32 v11, 0xffff, v61
+; SI-NEXT: v_or_b32_e32 v11, v11, v43
+; SI-NEXT: buffer_store_dword v11, off, s[0:3], s32 offset:696 ; 4-byte Folded Spill
+; SI-NEXT: s_waitcnt expcnt(0)
+; SI-NEXT: v_alignbit_b32 v11, v19, v43, 16
+; SI-NEXT: buffer_store_dword v11, off, s[0:3], s32 offset:884 ; 4-byte Folded Spill
+; SI-NEXT: s_waitcnt expcnt(0)
+; SI-NEXT: v_and_b32_e32 v11, 0xffff, v15
+; SI-NEXT: v_or_b32_e32 v11, v11, v38
+; SI-NEXT: v_alignbit_b32 v1, v11, v45, 16
+; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:888 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: buffer_load_dword v11, off, s[0:3], s32 offset:468 ; 4-byte Folded Reload
-; SI-NEXT: ; implicit-def: $vgpr57
-; SI-NEXT: ; implicit-def: $vgpr60
-; SI-NEXT: s_waitcnt vmcnt(0)
-; SI-NEXT: v_and_b32_e32 v11, 0xffff, v11
-; SI-NEXT: v_or_b32_e32 v19, v11, v38
-; SI-NEXT: v_and_b32_e32 v11, 0xffff, v63
-; SI-NEXT: v_or_b32_e32 v27, v11, v52
-; SI-NEXT: v_alignbit_b32 v11, v19, v52, 16
-; SI-NEXT: buffer_store_dword v11, off, s[0:3], s32 offset:876 ; 4-byte Folded Spill
+; SI-NEXT: v_and_b32_e32 v1, 0xffff, v47
+; SI-NEXT: v_or_b32_e32 v15, v1, v2
+; SI-NEXT: v_and_b32_e32 v1, 0xffff, v10
+; SI-NEXT: v_or_b32_e32 v1, v1, v57
+; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:688 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: buffer_load_dword v11, off, s[0:3], s32 offset:472 ; 4-byte Folded Reload
-; SI-NEXT: v_lshrrev_b32_e32 v52, 16, v30
-; SI-NEXT: v_lshrrev_b32_e32 v30, 16, v23
-; SI-NEXT: ; implicit-def: $vgpr63
-; SI-NEXT: s_waitcnt vmcnt(0)
-; SI-NEXT: v_and_b32_e32 v11, 0xffff, v11
-; SI-NEXT: v_or_b32_e32 v11, v11, v2
-; SI-NEXT: v_alignbit_b32 v1, v11, v55, 16
-; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:880 ; 4-byte Folded Spill
+; SI-NEXT: v_alignbit_b32 v1, v15, v57, 16
+; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:892 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_and_b32_e32 v1, 0xffff, v17
-; SI-NEXT: v_or_b32_e32 v15, v1, v14
+; SI-NEXT: v_and_b32_e32 v1, 0xffff, v59
+; SI-NEXT: v_or_b32_e32 v17, v1, v4
; SI-NEXT: v_and_b32_e32 v1, 0xffff, v48
-; SI-NEXT: v_or_b32_e32 v1, v1, v44
-; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:680 ; 4-byte Folded Spill
-; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_alignbit_b32 v1, v15, v44, 16
-; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:884 ; 4-byte Folded Spill
+; SI-NEXT: v_or_b32_e32 v1, v1, v62
+; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:684 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_and_b32_e32 v1, 0xffff, v45
-; SI-NEXT: v_or_b32_e32 v17, v1, v25
-; SI-NEXT: v_and_b32_e32 v1, 0xffff, v46
-; SI-NEXT: v_or_b32_e32 v1, v1, v47
-; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:676 ; 4-byte Folded Spill
+; SI-NEXT: v_alignbit_b32 v1, v17, v62, 16
+; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:896 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_alignbit_b32 v1, v17, v47, 16
-; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:888 ; 4-byte Folded Spill
-; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_and_b32_e32 v1, 0xffff, v58
-; SI-NEXT: v_or_b32_e32 v1, v1, v16
-; SI-NEXT: v_alignbit_b32 v32, v1, v59, 16
-; SI-NEXT: buffer_store_dword v32, off, s[0:3], s32 offset:892 ; 4-byte Folded Spill
-; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_and_b32_e32 v32, 0xffff, v62
-; SI-NEXT: v_or_b32_e32 v59, v6, v23
-; SI-NEXT: v_and_b32_e32 v6, 0xffff, v20
-; SI-NEXT: v_or_b32_e32 v62, v32, v24
-; SI-NEXT: v_and_b32_e32 v32, 0xffff, v50
-; SI-NEXT: v_or_b32_e32 v50, v6, v36
-; SI-NEXT: v_alignbit_b32 v6, v59, v36, 16
-; SI-NEXT: buffer_store_dword v6, off, s[0:3], s32 offset:900 ; 4-byte Folded Spill
-; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_and_b32_e32 v6, 0xffff, v8
-; SI-NEXT: v_or_b32_e32 v47, v6, v35
-; SI-NEXT: v_and_b32_e32 v6, 0xffff, v53
-; SI-NEXT: v_or_b32_e32 v49, v6, v37
-; SI-NEXT: v_alignbit_b32 v6, v47, v37, 16
-; SI-NEXT: buffer_store_dword v6, off, s[0:3], s32 offset:904 ; 4-byte Folded Spill
-; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_and_b32_e32 v6, 0xffff, v51
-; SI-NEXT: v_or_b32_e32 v45, v6, v34
-; SI-NEXT: v_or_b32_e32 v48, v3, v56
-; SI-NEXT: v_alignbit_b32 v3, v45, v56, 16
-; SI-NEXT: buffer_store_dword v3, off, s[0:3], s32 offset:908 ; 4-byte Folded Spill
-; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_and_b32_e32 v3, 0xffff, v54
-; SI-NEXT: v_or_b32_e32 v44, v3, v4
-; SI-NEXT: v_and_b32_e32 v3, 0xffff, v31
-; SI-NEXT: v_or_b32_e32 v3, v3, v9
-; SI-NEXT: v_lshrrev_b32_e32 v20, 16, v18
-; SI-NEXT: v_lshrrev_b32_e32 v18, 16, v14
-; SI-NEXT: v_mov_b32_e32 v14, v3
-; SI-NEXT: v_lshrrev_b32_e32 v3, 16, v4
-; SI-NEXT: buffer_store_dword v3, off, s[0:3], s32 offset:940 ; 4-byte Folded Spill
+; SI-NEXT: v_and_b32_e32 v1, 0xffff, v44
+; SI-NEXT: v_and_b32_e32 v10, 0xffff, v56
+; SI-NEXT: v_or_b32_e32 v1, v1, v8
+; SI-NEXT: v_or_b32_e32 v10, v10, v21
+; SI-NEXT: buffer_store_dword v10, off, s[0:3], s32 offset:668 ; 4-byte Folded Spill
+; SI-NEXT: s_waitcnt expcnt(0)
+; SI-NEXT: v_alignbit_b32 v10, v1, v21, 16
+; SI-NEXT: buffer_store_dword v10, off, s[0:3], s32 offset:900 ; 4-byte Folded Spill
+; SI-NEXT: v_or_b32_e32 v62, v6, v3
+; SI-NEXT: s_waitcnt expcnt(0)
+; SI-NEXT: v_lshrrev_b32_e32 v10, 16, v33
+; SI-NEXT: v_lshrrev_b32_e32 v33, 16, v3
+; SI-NEXT: v_lshrrev_b32_e32 v3, 16, v23
+; SI-NEXT: buffer_store_dword v3, off, s[0:3], s32 offset:948 ; 4-byte Folded Spill
; SI-NEXT: ; implicit-def: $vgpr3
; SI-NEXT: ; kill: killed $vgpr3
; SI-NEXT: ; implicit-def: $vgpr3
@@ -203747,133 +203824,175 @@ define <64 x i16> @bitcast_v128i8_to_v64i16(<128 x i8> %a, i32 %b) {
; SI-NEXT: ; implicit-def: $vgpr3
; SI-NEXT: ; kill: killed $vgpr3
; SI-NEXT: ; implicit-def: $vgpr3
+; SI-NEXT: v_and_b32_e32 v6, 0xffff, v49
; SI-NEXT: ; kill: killed $vgpr3
; SI-NEXT: ; implicit-def: $vgpr3
+; SI-NEXT: v_or_b32_e32 v6, v6, v16
; SI-NEXT: ; kill: killed $vgpr3
; SI-NEXT: ; implicit-def: $vgpr3
+; SI-NEXT: buffer_store_dword v6, off, s[0:3], s32 offset:664 ; 4-byte Folded Spill
+; SI-NEXT: s_waitcnt expcnt(0)
+; SI-NEXT: v_alignbit_b32 v6, v62, v16, 16
; SI-NEXT: ; kill: killed $vgpr3
; SI-NEXT: ; implicit-def: $vgpr3
+; SI-NEXT: buffer_store_dword v6, off, s[0:3], s32 offset:904 ; 4-byte Folded Spill
+; SI-NEXT: s_waitcnt expcnt(0)
+; SI-NEXT: v_and_b32_e32 v6, 0xffff, v24
; SI-NEXT: ; kill: killed $vgpr3
; SI-NEXT: ; implicit-def: $vgpr3
+; SI-NEXT: v_or_b32_e32 v59, v6, v32
+; SI-NEXT: v_and_b32_e32 v6, 0xffff, v50
; SI-NEXT: ; kill: killed $vgpr3
; SI-NEXT: ; implicit-def: $vgpr3
+; SI-NEXT: v_or_b32_e32 v56, v6, v34
+; SI-NEXT: v_alignbit_b32 v6, v59, v34, 16
; SI-NEXT: ; kill: killed $vgpr3
; SI-NEXT: ; implicit-def: $vgpr3
+; SI-NEXT: buffer_store_dword v6, off, s[0:3], s32 offset:908 ; 4-byte Folded Spill
+; SI-NEXT: s_waitcnt expcnt(0)
+; SI-NEXT: v_and_b32_e32 v6, 0xffff, v36
; SI-NEXT: ; kill: killed $vgpr3
; SI-NEXT: ; implicit-def: $vgpr3
+; SI-NEXT: v_or_b32_e32 v47, v6, v35
+; SI-NEXT: v_and_b32_e32 v6, 0xffff, v58
; SI-NEXT: ; kill: killed $vgpr3
; SI-NEXT: ; implicit-def: $vgpr3
+; SI-NEXT: v_or_b32_e32 v50, v6, v37
+; SI-NEXT: v_alignbit_b32 v6, v47, v37, 16
; SI-NEXT: ; kill: killed $vgpr3
; SI-NEXT: ; implicit-def: $vgpr3
+; SI-NEXT: buffer_store_dword v6, off, s[0:3], s32 offset:912 ; 4-byte Folded Spill
+; SI-NEXT: s_waitcnt expcnt(0)
+; SI-NEXT: v_and_b32_e32 v6, 0xffff, v51
; SI-NEXT: ; kill: killed $vgpr3
; SI-NEXT: ; implicit-def: $vgpr3
+; SI-NEXT: v_or_b32_e32 v45, v6, v14
+; SI-NEXT: v_and_b32_e32 v6, 0xffff, v9
; SI-NEXT: ; kill: killed $vgpr3
; SI-NEXT: ; implicit-def: $vgpr3
+; SI-NEXT: v_or_b32_e32 v49, v6, v31
+; SI-NEXT: v_alignbit_b32 v6, v45, v31, 16
; SI-NEXT: ; kill: killed $vgpr3
; SI-NEXT: ; implicit-def: $vgpr3
+; SI-NEXT: buffer_store_dword v6, off, s[0:3], s32 offset:916 ; 4-byte Folded Spill
+; SI-NEXT: s_waitcnt expcnt(0)
+; SI-NEXT: v_and_b32_e32 v6, 0xffff, v54
; SI-NEXT: ; kill: killed $vgpr3
; SI-NEXT: ; implicit-def: $vgpr3
+; SI-NEXT: v_or_b32_e32 v44, v6, v23
+; SI-NEXT: v_and_b32_e32 v6, 0xffff, v52
; SI-NEXT: ; kill: killed $vgpr3
; SI-NEXT: ; implicit-def: $vgpr3
-; SI-NEXT: v_or_b32_e32 v46, v32, v13
-; SI-NEXT: v_alignbit_b32 v13, v62, v13, 16
-; SI-NEXT: v_alignbit_b32 v6, v44, v9, 16
+; SI-NEXT: v_or_b32_e32 v48, v6, v13
+; SI-NEXT: v_alignbit_b32 v6, v44, v13, 16
; SI-NEXT: ; kill: killed $vgpr3
; SI-NEXT: ; implicit-def: $vgpr3
-; SI-NEXT: buffer_store_dword v13, off, s[0:3], s32 offset:896 ; 4-byte Folded Spill
-; SI-NEXT: buffer_store_dword v6, off, s[0:3], s32 offset:912 ; 4-byte Folded Spill
-; SI-NEXT: v_lshrrev_b32_e32 v58, 16, v61
-; SI-NEXT: v_lshrrev_b32_e32 v55, 16, v10
-; SI-NEXT: s_waitcnt expcnt(1)
-; SI-NEXT: v_lshrrev_b32_e32 v13, 16, v33
-; SI-NEXT: v_lshrrev_b32_e32 v8, 16, v38
-; SI-NEXT: v_lshrrev_b32_e32 v38, 16, v2
-; SI-NEXT: v_lshrrev_b32_e32 v10, 16, v25
-; SI-NEXT: v_lshrrev_b32_e32 v2, 16, v16
-; SI-NEXT: v_lshrrev_b32_e32 v33, 16, v24
+; SI-NEXT: buffer_store_dword v6, off, s[0:3], s32 offset:920 ; 4-byte Folded Spill
+; SI-NEXT: v_lshrrev_b32_e32 v57, 16, v63
+; SI-NEXT: v_lshrrev_b32_e32 v43, 16, v26
+; SI-NEXT: v_lshrrev_b32_e32 v36, 16, v25
+; SI-NEXT: v_lshrrev_b32_e32 v53, 16, v18
+; SI-NEXT: v_lshrrev_b32_e32 v21, 16, v41
+; SI-NEXT: v_lshrrev_b32_e32 v16, 16, v30
+; SI-NEXT: v_lshrrev_b32_e32 v38, 16, v38
+; SI-NEXT: v_lshrrev_b32_e32 v26, 16, v2
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_lshrrev_b32_e32 v6, 16, v35
-; SI-NEXT: v_lshrrev_b32_e32 v36, 16, v34
-; SI-NEXT: ; implicit-def: $vgpr23
-; SI-NEXT: ; implicit-def: $vgpr34
-; SI-NEXT: ; implicit-def: $vgpr61
+; SI-NEXT: v_lshrrev_b32_e32 v6, 16, v4
+; SI-NEXT: v_lshrrev_b32_e32 v2, 16, v8
+; SI-NEXT: v_lshrrev_b32_e32 v30, 16, v32
+; SI-NEXT: v_lshrrev_b32_e32 v8, 16, v35
+; SI-NEXT: v_lshrrev_b32_e32 v4, 16, v14
+; SI-NEXT: ; implicit-def: $vgpr14
+; SI-NEXT: ; implicit-def: $vgpr52
+; SI-NEXT: ; implicit-def: $vgpr41
; SI-NEXT: ; implicit-def: $vgpr54
-; SI-NEXT: ; implicit-def: $vgpr16
+; SI-NEXT: ; implicit-def: $vgpr46
; SI-NEXT: ; implicit-def: $vgpr24
-; SI-NEXT: ; implicit-def: $vgpr25
; SI-NEXT: ; implicit-def: $vgpr35
; SI-NEXT: ; implicit-def: $vgpr37
; SI-NEXT: ; implicit-def: $vgpr51
-; SI-NEXT: ; implicit-def: $vgpr56
-; SI-NEXT: ; implicit-def: $vgpr4
+; SI-NEXT: ; implicit-def: $vgpr60
+; SI-NEXT: ; implicit-def: $vgpr31
+; SI-NEXT: ; implicit-def: $vgpr61
+; SI-NEXT: ; implicit-def: $vgpr23
+; SI-NEXT: ; implicit-def: $vgpr25
+; SI-NEXT: ; implicit-def: $vgpr13
; SI-NEXT: ; implicit-def: $vgpr32
-; SI-NEXT: ; implicit-def: $vgpr9
+; SI-NEXT: ; implicit-def: $vgpr34
; SI-NEXT: ; kill: killed $vgpr3
-; SI-NEXT: ; implicit-def: $vgpr53
; SI-NEXT: ; implicit-def: $vgpr3
-; SI-NEXT: ; implicit-def: $vgpr31
+; SI-NEXT: ; kill: killed $vgpr3
+; SI-NEXT: ; implicit-def: $vgpr58
+; SI-NEXT: ; implicit-def: $vgpr9
+; SI-NEXT: ; implicit-def: $vgpr63
; SI-NEXT: .LBB96_2: ; %Flow
; SI-NEXT: s_andn2_saveexec_b64 s[4:5], s[4:5]
; SI-NEXT: s_cbranch_execz .LBB96_4
; SI-NEXT: ; %bb.3: ; %cmp.true
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_add_i32_e32 v1, vcc, 3, v61
+; SI-NEXT: v_add_i32_e32 v1, vcc, 3, v52
; SI-NEXT: v_and_b32_e32 v1, 0xff, v1
; SI-NEXT: v_add_i32_e32 v2, vcc, 3, v41
-; SI-NEXT: v_or_b32_e32 v1, v31, v1
+; SI-NEXT: v_or_b32_e32 v1, v63, v1
; SI-NEXT: v_and_b32_e32 v2, 0xff, v2
; SI-NEXT: v_lshlrev_b32_e32 v2, 16, v2
; SI-NEXT: v_add_i32_e32 v1, vcc, 0x300, v1
-; SI-NEXT: v_or_b32_e32 v2, v9, v2
+; SI-NEXT: v_or_b32_e32 v2, v13, v2
; SI-NEXT: v_and_b32_e32 v1, 0xffff, v1
; SI-NEXT: v_or_b32_e32 v1, v2, v1
; SI-NEXT: v_add_i32_e32 v2, vcc, 3, v54
; SI-NEXT: v_and_b32_e32 v2, 0xff, v2
-; SI-NEXT: v_or_b32_e32 v2, v4, v2
-; SI-NEXT: v_add_i32_e32 v4, vcc, 3, v57
-; SI-NEXT: v_and_b32_e32 v4, 0xff, v4
+; SI-NEXT: v_add_i32_e32 v3, vcc, 3, v46
+; SI-NEXT: v_or_b32_e32 v2, v23, v2
+; SI-NEXT: v_and_b32_e32 v3, 0xff, v3
; SI-NEXT: s_movk_i32 s6, 0x300
-; SI-NEXT: v_lshlrev_b32_e32 v4, 16, v4
+; SI-NEXT: v_lshlrev_b32_e32 v3, 16, v3
; SI-NEXT: v_add_i32_e32 v2, vcc, s6, v2
-; SI-NEXT: v_or_b32_e32 v4, v32, v4
+; SI-NEXT: v_or_b32_e32 v3, v25, v3
; SI-NEXT: v_and_b32_e32 v2, 0xffff, v2
-; SI-NEXT: v_or_b32_e32 v2, v4, v2
+; SI-NEXT: v_or_b32_e32 v2, v3, v2
+; SI-NEXT: buffer_load_dword v3, off, s[0:3], s32 offset:972 ; 4-byte Folded Reload
; SI-NEXT: buffer_load_dword v4, off, s[0:3], s32 offset:968 ; 4-byte Folded Reload
-; SI-NEXT: v_add_i32_e32 v5, vcc, 3, v34
+; SI-NEXT: v_add_i32_e32 v5, vcc, 3, v14
; SI-NEXT: v_and_b32_e32 v5, 0xff, v5
; SI-NEXT: v_lshlrev_b32_e32 v5, 16, v5
-; SI-NEXT: v_or_b32_e32 v5, v56, v5
-; SI-NEXT: v_mov_b32_e32 v30, v16
+; SI-NEXT: v_or_b32_e32 v5, v31, v5
; SI-NEXT: s_mov_b32 s7, 0x3000000
-; SI-NEXT: v_mov_b32_e32 v31, v24
+; SI-NEXT: v_mov_b32_e32 v30, v24
; SI-NEXT: v_add_i32_e32 v44, vcc, s7, v2
-; SI-NEXT: v_add_i32_e32 v27, vcc, s7, v1
-; SI-NEXT: buffer_load_dword v6, off, s[0:3], s32 offset:916 ; 4-byte Folded Reload
-; SI-NEXT: buffer_load_dword v7, off, s[0:3], s32 offset:924 ; 4-byte Folded Reload
-; SI-NEXT: buffer_load_dword v8, off, s[0:3], s32 offset:972 ; 4-byte Folded Reload
-; SI-NEXT: buffer_load_dword v9, off, s[0:3], s32 offset:952 ; 4-byte Folded Reload
-; SI-NEXT: buffer_load_dword v10, off, s[0:3], s32 offset:936 ; 4-byte Folded Reload
-; SI-NEXT: buffer_load_dword v11, off, s[0:3], s32 offset:844 ; 4-byte Folded Reload
-; SI-NEXT: buffer_load_dword v12, off, s[0:3], s32 offset:836 ; 4-byte Folded Reload
-; SI-NEXT: buffer_load_dword v13, off, s[0:3], s32 offset:808 ; 4-byte Folded Reload
-; SI-NEXT: buffer_load_dword v14, off, s[0:3], s32 offset:796 ; 4-byte Folded Reload
-; SI-NEXT: buffer_load_dword v15, off, s[0:3], s32 offset:736 ; 4-byte Folded Reload
-; SI-NEXT: buffer_load_dword v17, off, s[0:3], s32 offset:728 ; 4-byte Folded Reload
-; SI-NEXT: buffer_load_dword v18, off, s[0:3], s32 offset:688 ; 4-byte Folded Reload
-; SI-NEXT: buffer_load_dword v16, off, s[0:3], s32 offset:608 ; 4-byte Folded Reload
-; SI-NEXT: buffer_load_dword v19, off, s[0:3], s32 offset:664 ; 4-byte Folded Reload
-; SI-NEXT: buffer_load_dword v20, off, s[0:3], s32 offset:636 ; 4-byte Folded Reload
-; SI-NEXT: buffer_load_dword v21, off, s[0:3], s32 offset:624 ; 4-byte Folded Reload
-; SI-NEXT: buffer_load_dword v22, off, s[0:3], s32 offset:612 ; 4-byte Folded Reload
-; SI-NEXT: buffer_load_dword v26, off, s[0:3], s32 offset:584 ; 4-byte Folded Reload
-; SI-NEXT: buffer_load_dword v24, off, s[0:3], s32 offset:712 ; 4-byte Folded Reload
+; SI-NEXT: v_add_i32_e32 v31, vcc, s7, v1
+; SI-NEXT: v_mov_b32_e32 v48, v31
+; SI-NEXT: buffer_load_dword v6, off, s[0:3], s32 offset:924 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v7, off, s[0:3], s32 offset:932 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v8, off, s[0:3], s32 offset:976 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v10, off, s[0:3], s32 offset:944 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v11, off, s[0:3], s32 offset:852 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v12, off, s[0:3], s32 offset:844 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v13, off, s[0:3], s32 offset:816 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v14, off, s[0:3], s32 offset:804 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v15, off, s[0:3], s32 offset:744 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v16, off, s[0:3], s32 offset:736 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v17, off, s[0:3], s32 offset:712 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v18, off, s[0:3], s32 offset:700 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v19, off, s[0:3], s32 offset:672 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v21, off, s[0:3], s32 offset:628 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v20, off, s[0:3], s32 offset:644 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v22, off, s[0:3], s32 offset:616 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v23, off, s[0:3], s32 offset:596 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v25, off, s[0:3], s32 offset:588 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v24, off, s[0:3], s32 offset:724 ; 4-byte Folded Reload
; SI-NEXT: s_waitcnt vmcnt(14)
+; SI-NEXT: v_add_i32_e32 v3, vcc, 3, v3
+; SI-NEXT: v_and_b32_e32 v3, 0xff, v3
; SI-NEXT: v_add_i32_e32 v4, vcc, 3, v4
+; SI-NEXT: v_or_b32_e32 v3, v9, v3
; SI-NEXT: v_and_b32_e32 v4, 0xff, v4
-; SI-NEXT: v_or_b32_e32 v3, v3, v4
-; SI-NEXT: buffer_load_dword v4, off, s[0:3], s32 offset:964 ; 4-byte Folded Reload
+; SI-NEXT: v_lshlrev_b32_e32 v4, 16, v4
; SI-NEXT: v_add_i32_e32 v3, vcc, s6, v3
+; SI-NEXT: v_or_b32_e32 v4, v61, v4
; SI-NEXT: v_and_b32_e32 v3, 0xffff, v3
+; SI-NEXT: v_or_b32_e32 v4, v4, v3
+; SI-NEXT: buffer_load_dword v3, off, s[0:3], s32 offset:960 ; 4-byte Folded Reload
; SI-NEXT: v_add_i32_e32 v6, vcc, 3, v6
; SI-NEXT: v_and_b32_e32 v6, 0xff, v6
; SI-NEXT: v_lshlrev_b32_e32 v6, 16, v6
@@ -203882,41 +204001,27 @@ define <64 x i16> @bitcast_v128i8_to_v64i16(<128 x i8> %a, i32 %b) {
; SI-NEXT: v_and_b32_e32 v7, 0xff, v7
; SI-NEXT: v_lshlrev_b32_e32 v7, 16, v7
; SI-NEXT: v_or_b32_e32 v7, v37, v7
-; SI-NEXT: s_waitcnt vmcnt(7)
-; SI-NEXT: v_add_i32_e32 v16, vcc, 3, v16
-; SI-NEXT: v_and_b32_e32 v16, 0xff, v16
-; SI-NEXT: v_lshlrev_b32_e32 v16, 16, v16
-; SI-NEXT: s_waitcnt vmcnt(0)
-; SI-NEXT: v_add_i32_e32 v4, vcc, 3, v4
-; SI-NEXT: v_and_b32_e32 v4, 0xff, v4
-; SI-NEXT: v_lshlrev_b32_e32 v4, 16, v4
-; SI-NEXT: v_or_b32_e32 v4, v60, v4
-; SI-NEXT: v_or_b32_e32 v3, v4, v3
-; SI-NEXT: buffer_load_dword v4, off, s[0:3], s32 offset:956 ; 4-byte Folded Reload
-; SI-NEXT: v_add_i32_e32 v32, vcc, s7, v3
-; SI-NEXT: buffer_load_dword v3, off, s[0:3], s32 offset:548 ; 4-byte Folded Reload
-; SI-NEXT: v_mov_b32_e32 v48, v32
-; SI-NEXT: s_waitcnt vmcnt(1)
-; SI-NEXT: v_add_i32_e32 v4, vcc, 3, v4
-; SI-NEXT: v_and_b32_e32 v4, 0xff, v4
-; SI-NEXT: v_or_b32_e32 v4, v63, v4
-; SI-NEXT: v_add_i32_e32 v4, vcc, s6, v4
-; SI-NEXT: v_and_b32_e32 v4, 0xffff, v4
-; SI-NEXT: v_or_b32_e32 v4, v5, v4
-; SI-NEXT: buffer_load_dword v5, off, s[0:3], s32 offset:948 ; 4-byte Folded Reload
-; SI-NEXT: v_add_i32_e32 v45, vcc, s7, v4
-; SI-NEXT: buffer_load_dword v4, off, s[0:3], s32 offset:540 ; 4-byte Folded Reload
-; SI-NEXT: v_lshrrev_b32_e32 v36, 16, v45
-; SI-NEXT: s_waitcnt vmcnt(1)
-; SI-NEXT: v_add_i32_e32 v5, vcc, 3, v5
-; SI-NEXT: v_and_b32_e32 v5, 0xff, v5
-; SI-NEXT: v_or_b32_e32 v5, v53, v5
-; SI-NEXT: v_add_i32_e32 v5, vcc, s6, v5
-; SI-NEXT: v_and_b32_e32 v5, 0xffff, v5
-; SI-NEXT: v_or_b32_e32 v5, v6, v5
-; SI-NEXT: buffer_load_dword v6, off, s[0:3], s32 offset:848 ; 4-byte Folded Reload
-; SI-NEXT: v_add_i32_e32 v33, vcc, s7, v5
+; SI-NEXT: buffer_load_dword v9, off, s[0:3], s32 offset:956 ; 4-byte Folded Reload
+; SI-NEXT: v_add_i32_e32 v33, vcc, s7, v4
+; SI-NEXT: buffer_load_dword v4, off, s[0:3], s32 offset:548 ; 4-byte Folded Reload
; SI-NEXT: v_mov_b32_e32 v49, v33
+; SI-NEXT: s_waitcnt vmcnt(2)
+; SI-NEXT: v_add_i32_e32 v3, vcc, 3, v3
+; SI-NEXT: v_and_b32_e32 v3, 0xff, v3
+; SI-NEXT: v_or_b32_e32 v3, v60, v3
+; SI-NEXT: v_add_i32_e32 v3, vcc, s6, v3
+; SI-NEXT: v_and_b32_e32 v3, 0xffff, v3
+; SI-NEXT: v_or_b32_e32 v5, v5, v3
+; SI-NEXT: buffer_load_dword v3, off, s[0:3], s32 offset:952 ; 4-byte Folded Reload
+; SI-NEXT: v_add_i32_e32 v45, vcc, s7, v5
+; SI-NEXT: s_waitcnt vmcnt(0)
+; SI-NEXT: v_add_i32_e32 v3, vcc, 3, v3
+; SI-NEXT: v_and_b32_e32 v3, 0xff, v3
+; SI-NEXT: v_or_b32_e32 v3, v58, v3
+; SI-NEXT: v_add_i32_e32 v3, vcc, s6, v3
+; SI-NEXT: v_and_b32_e32 v3, 0xffff, v3
+; SI-NEXT: v_or_b32_e32 v3, v6, v3
+; SI-NEXT: buffer_load_dword v6, off, s[0:3], s32 offset:856 ; 4-byte Folded Reload
; SI-NEXT: s_waitcnt vmcnt(0)
; SI-NEXT: v_add_i32_e32 v6, vcc, 3, v6
; SI-NEXT: v_and_b32_e32 v6, 0xff, v6
@@ -203924,13 +204029,16 @@ define <64 x i16> @bitcast_v128i8_to_v64i16(<128 x i8> %a, i32 %b) {
; SI-NEXT: v_add_i32_e32 v6, vcc, s6, v6
; SI-NEXT: v_and_b32_e32 v6, 0xffff, v6
; SI-NEXT: v_or_b32_e32 v6, v7, v6
-; SI-NEXT: buffer_load_dword v7, off, s[0:3], s32 offset:920 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v7, off, s[0:3], s32 offset:928 ; 4-byte Folded Reload
+; SI-NEXT: v_add_i32_e32 v35, vcc, s7, v3
+; SI-NEXT: buffer_load_dword v3, off, s[0:3], s32 offset:552 ; 4-byte Folded Reload
; SI-NEXT: v_add_i32_e32 v47, vcc, s7, v6
-; SI-NEXT: s_waitcnt vmcnt(0)
+; SI-NEXT: v_mov_b32_e32 v50, v35
+; SI-NEXT: s_waitcnt vmcnt(1)
; SI-NEXT: v_add_i32_e32 v7, vcc, 3, v7
; SI-NEXT: v_and_b32_e32 v7, 0xff, v7
; SI-NEXT: v_or_b32_e32 v7, v8, v7
-; SI-NEXT: buffer_load_dword v8, off, s[0:3], s32 offset:816 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v8, off, s[0:3], s32 offset:824 ; 4-byte Folded Reload
; SI-NEXT: v_add_i32_e32 v7, vcc, s6, v7
; SI-NEXT: v_and_b32_e32 v7, 0xffff, v7
; SI-NEXT: s_waitcnt vmcnt(0)
@@ -203939,16 +204047,16 @@ define <64 x i16> @bitcast_v128i8_to_v64i16(<128 x i8> %a, i32 %b) {
; SI-NEXT: v_lshlrev_b32_e32 v8, 16, v8
; SI-NEXT: v_or_b32_e32 v8, v9, v8
; SI-NEXT: v_or_b32_e32 v7, v8, v7
-; SI-NEXT: buffer_load_dword v8, off, s[0:3], s32 offset:812 ; 4-byte Folded Reload
-; SI-NEXT: buffer_load_dword v9, off, s[0:3], s32 offset:932 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v8, off, s[0:3], s32 offset:820 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v9, off, s[0:3], s32 offset:940 ; 4-byte Folded Reload
; SI-NEXT: v_add_i32_e32 v6, vcc, s7, v7
-; SI-NEXT: v_mov_b32_e32 v50, v6
+; SI-NEXT: v_mov_b32_e32 v56, v6
; SI-NEXT: s_waitcnt vmcnt(1)
; SI-NEXT: v_add_i32_e32 v8, vcc, 3, v8
; SI-NEXT: v_and_b32_e32 v8, 0xff, v8
; SI-NEXT: s_waitcnt vmcnt(0)
; SI-NEXT: v_or_b32_e32 v8, v9, v8
-; SI-NEXT: buffer_load_dword v9, off, s[0:3], s32 offset:824 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v9, off, s[0:3], s32 offset:832 ; 4-byte Folded Reload
; SI-NEXT: v_add_i32_e32 v8, vcc, s6, v8
; SI-NEXT: v_and_b32_e32 v8, 0xffff, v8
; SI-NEXT: s_waitcnt vmcnt(0)
@@ -203957,15 +204065,15 @@ define <64 x i16> @bitcast_v128i8_to_v64i16(<128 x i8> %a, i32 %b) {
; SI-NEXT: v_lshlrev_b32_e32 v9, 16, v9
; SI-NEXT: v_or_b32_e32 v9, v10, v9
; SI-NEXT: v_or_b32_e32 v8, v9, v8
-; SI-NEXT: buffer_load_dword v9, off, s[0:3], s32 offset:828 ; 4-byte Folded Reload
-; SI-NEXT: buffer_load_dword v10, off, s[0:3], s32 offset:960 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v9, off, s[0:3], s32 offset:836 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v10, off, s[0:3], s32 offset:964 ; 4-byte Folded Reload
; SI-NEXT: v_add_i32_e32 v59, vcc, s7, v8
; SI-NEXT: s_waitcnt vmcnt(1)
; SI-NEXT: v_add_i32_e32 v9, vcc, 3, v9
; SI-NEXT: v_and_b32_e32 v9, 0xff, v9
; SI-NEXT: s_waitcnt vmcnt(0)
; SI-NEXT: v_or_b32_e32 v9, v10, v9
-; SI-NEXT: buffer_load_dword v10, off, s[0:3], s32 offset:768 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v10, off, s[0:3], s32 offset:776 ; 4-byte Folded Reload
; SI-NEXT: v_add_i32_e32 v9, vcc, s6, v9
; SI-NEXT: v_and_b32_e32 v9, 0xffff, v9
; SI-NEXT: s_waitcnt vmcnt(0)
@@ -203974,16 +204082,15 @@ define <64 x i16> @bitcast_v128i8_to_v64i16(<128 x i8> %a, i32 %b) {
; SI-NEXT: v_lshlrev_b32_e32 v10, 16, v10
; SI-NEXT: v_or_b32_e32 v10, v11, v10
; SI-NEXT: v_or_b32_e32 v9, v10, v9
-; SI-NEXT: buffer_load_dword v10, off, s[0:3], s32 offset:764 ; 4-byte Folded Reload
-; SI-NEXT: buffer_load_dword v11, off, s[0:3], s32 offset:832 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v10, off, s[0:3], s32 offset:772 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v11, off, s[0:3], s32 offset:840 ; 4-byte Folded Reload
; SI-NEXT: v_add_i32_e32 v8, vcc, s7, v9
-; SI-NEXT: v_mov_b32_e32 v46, v8
; SI-NEXT: s_waitcnt vmcnt(1)
; SI-NEXT: v_add_i32_e32 v10, vcc, 3, v10
; SI-NEXT: v_and_b32_e32 v10, 0xff, v10
; SI-NEXT: s_waitcnt vmcnt(0)
; SI-NEXT: v_or_b32_e32 v10, v11, v10
-; SI-NEXT: buffer_load_dword v11, off, s[0:3], s32 offset:788 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v11, off, s[0:3], s32 offset:796 ; 4-byte Folded Reload
; SI-NEXT: v_add_i32_e32 v10, vcc, s6, v10
; SI-NEXT: v_and_b32_e32 v10, 0xffff, v10
; SI-NEXT: s_waitcnt vmcnt(0)
@@ -203992,15 +204099,15 @@ define <64 x i16> @bitcast_v128i8_to_v64i16(<128 x i8> %a, i32 %b) {
; SI-NEXT: v_lshlrev_b32_e32 v11, 16, v11
; SI-NEXT: v_or_b32_e32 v11, v12, v11
; SI-NEXT: v_or_b32_e32 v10, v11, v10
-; SI-NEXT: buffer_load_dword v11, off, s[0:3], s32 offset:804 ; 4-byte Folded Reload
-; SI-NEXT: buffer_load_dword v12, off, s[0:3], s32 offset:928 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v11, off, s[0:3], s32 offset:812 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v12, off, s[0:3], s32 offset:936 ; 4-byte Folded Reload
; SI-NEXT: v_add_i32_e32 v62, vcc, s7, v10
; SI-NEXT: s_waitcnt vmcnt(1)
; SI-NEXT: v_add_i32_e32 v11, vcc, 3, v11
; SI-NEXT: v_and_b32_e32 v11, 0xff, v11
; SI-NEXT: s_waitcnt vmcnt(0)
; SI-NEXT: v_or_b32_e32 v11, v12, v11
-; SI-NEXT: buffer_load_dword v12, off, s[0:3], s32 offset:708 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v12, off, s[0:3], s32 offset:720 ; 4-byte Folded Reload
; SI-NEXT: v_add_i32_e32 v11, vcc, s6, v11
; SI-NEXT: v_and_b32_e32 v11, 0xffff, v11
; SI-NEXT: s_waitcnt vmcnt(0)
@@ -204009,15 +204116,15 @@ define <64 x i16> @bitcast_v128i8_to_v64i16(<128 x i8> %a, i32 %b) {
; SI-NEXT: v_lshlrev_b32_e32 v12, 16, v12
; SI-NEXT: v_or_b32_e32 v12, v13, v12
; SI-NEXT: v_or_b32_e32 v11, v12, v11
-; SI-NEXT: buffer_load_dword v12, off, s[0:3], s32 offset:704 ; 4-byte Folded Reload
-; SI-NEXT: buffer_load_dword v13, off, s[0:3], s32 offset:752 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v12, off, s[0:3], s32 offset:716 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v13, off, s[0:3], s32 offset:760 ; 4-byte Folded Reload
; SI-NEXT: v_add_i32_e32 v9, vcc, s7, v11
; SI-NEXT: s_waitcnt vmcnt(1)
; SI-NEXT: v_add_i32_e32 v12, vcc, 3, v12
; SI-NEXT: v_and_b32_e32 v12, 0xff, v12
; SI-NEXT: s_waitcnt vmcnt(0)
; SI-NEXT: v_or_b32_e32 v12, v13, v12
-; SI-NEXT: buffer_load_dword v13, off, s[0:3], s32 offset:744 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v13, off, s[0:3], s32 offset:752 ; 4-byte Folded Reload
; SI-NEXT: v_add_i32_e32 v12, vcc, s6, v12
; SI-NEXT: v_and_b32_e32 v12, 0xffff, v12
; SI-NEXT: s_waitcnt vmcnt(0)
@@ -204026,15 +204133,15 @@ define <64 x i16> @bitcast_v128i8_to_v64i16(<128 x i8> %a, i32 %b) {
; SI-NEXT: v_lshlrev_b32_e32 v13, 16, v13
; SI-NEXT: v_or_b32_e32 v13, v14, v13
; SI-NEXT: v_or_b32_e32 v12, v13, v12
-; SI-NEXT: buffer_load_dword v13, off, s[0:3], s32 offset:776 ; 4-byte Folded Reload
-; SI-NEXT: buffer_load_dword v14, off, s[0:3], s32 offset:840 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v13, off, s[0:3], s32 offset:784 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v14, off, s[0:3], s32 offset:848 ; 4-byte Folded Reload
; SI-NEXT: v_add_i32_e32 v1, vcc, s7, v12
; SI-NEXT: s_waitcnt vmcnt(1)
; SI-NEXT: v_add_i32_e32 v13, vcc, 3, v13
; SI-NEXT: v_and_b32_e32 v13, 0xff, v13
; SI-NEXT: s_waitcnt vmcnt(0)
; SI-NEXT: v_or_b32_e32 v13, v14, v13
-; SI-NEXT: buffer_load_dword v14, off, s[0:3], s32 offset:644 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v14, off, s[0:3], s32 offset:652 ; 4-byte Folded Reload
; SI-NEXT: v_add_i32_e32 v13, vcc, s6, v13
; SI-NEXT: v_and_b32_e32 v13, 0xffff, v13
; SI-NEXT: s_waitcnt vmcnt(0)
@@ -204043,45 +204150,48 @@ define <64 x i16> @bitcast_v128i8_to_v64i16(<128 x i8> %a, i32 %b) {
; SI-NEXT: v_lshlrev_b32_e32 v14, 16, v14
; SI-NEXT: v_or_b32_e32 v14, v15, v14
; SI-NEXT: v_or_b32_e32 v13, v14, v13
-; SI-NEXT: buffer_load_dword v14, off, s[0:3], s32 offset:632 ; 4-byte Folded Reload
-; SI-NEXT: buffer_load_dword v15, off, s[0:3], s32 offset:720 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v14, off, s[0:3], s32 offset:640 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v15, off, s[0:3], s32 offset:728 ; 4-byte Folded Reload
; SI-NEXT: v_add_i32_e32 v10, vcc, s7, v13
; SI-NEXT: s_waitcnt vmcnt(1)
; SI-NEXT: v_add_i32_e32 v14, vcc, 3, v14
; SI-NEXT: v_and_b32_e32 v14, 0xff, v14
; SI-NEXT: s_waitcnt vmcnt(0)
; SI-NEXT: v_or_b32_e32 v14, v15, v14
-; SI-NEXT: buffer_load_dword v15, off, s[0:3], s32 offset:672 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v15, off, s[0:3], s32 offset:680 ; 4-byte Folded Reload
; SI-NEXT: v_add_i32_e32 v14, vcc, s6, v14
; SI-NEXT: v_and_b32_e32 v14, 0xffff, v14
; SI-NEXT: s_waitcnt vmcnt(0)
; SI-NEXT: v_add_i32_e32 v15, vcc, 3, v15
; SI-NEXT: v_and_b32_e32 v15, 0xff, v15
; SI-NEXT: v_lshlrev_b32_e32 v15, 16, v15
-; SI-NEXT: v_or_b32_e32 v15, v17, v15
+; SI-NEXT: v_or_b32_e32 v15, v16, v15
; SI-NEXT: v_or_b32_e32 v14, v15, v14
-; SI-NEXT: buffer_load_dword v15, off, s[0:3], s32 offset:748 ; 4-byte Folded Reload
-; SI-NEXT: buffer_load_dword v17, off, s[0:3], s32 offset:820 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v15, off, s[0:3], s32 offset:756 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v16, off, s[0:3], s32 offset:828 ; 4-byte Folded Reload
; SI-NEXT: s_waitcnt vmcnt(1)
; SI-NEXT: v_add_i32_e32 v15, vcc, 3, v15
; SI-NEXT: v_and_b32_e32 v15, 0xff, v15
; SI-NEXT: s_waitcnt vmcnt(0)
-; SI-NEXT: v_or_b32_e32 v15, v17, v15
-; SI-NEXT: buffer_load_dword v17, off, s[0:3], s32 offset:700 ; 4-byte Folded Reload
+; SI-NEXT: v_or_b32_e32 v15, v16, v15
+; SI-NEXT: buffer_load_dword v16, off, s[0:3], s32 offset:612 ; 4-byte Folded Reload
; SI-NEXT: v_add_i32_e32 v15, vcc, s6, v15
; SI-NEXT: v_and_b32_e32 v15, 0xffff, v15
; SI-NEXT: s_waitcnt vmcnt(0)
+; SI-NEXT: v_add_i32_e32 v16, vcc, 3, v16
+; SI-NEXT: v_and_b32_e32 v16, 0xff, v16
+; SI-NEXT: v_lshlrev_b32_e32 v16, 16, v16
; SI-NEXT: v_or_b32_e32 v16, v17, v16
; SI-NEXT: v_or_b32_e32 v15, v16, v15
-; SI-NEXT: buffer_load_dword v16, off, s[0:3], s32 offset:600 ; 4-byte Folded Reload
-; SI-NEXT: buffer_load_dword v17, off, s[0:3], s32 offset:692 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v16, off, s[0:3], s32 offset:604 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v17, off, s[0:3], s32 offset:704 ; 4-byte Folded Reload
; SI-NEXT: v_add_i32_e32 v13, vcc, s7, v15
; SI-NEXT: s_waitcnt vmcnt(1)
; SI-NEXT: v_add_i32_e32 v16, vcc, 3, v16
; SI-NEXT: v_and_b32_e32 v16, 0xff, v16
; SI-NEXT: s_waitcnt vmcnt(0)
; SI-NEXT: v_or_b32_e32 v16, v17, v16
-; SI-NEXT: buffer_load_dword v17, off, s[0:3], s32 offset:596 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v17, off, s[0:3], s32 offset:600 ; 4-byte Folded Reload
; SI-NEXT: v_add_i32_e32 v16, vcc, s6, v16
; SI-NEXT: v_and_b32_e32 v16, 0xffff, v16
; SI-NEXT: s_waitcnt vmcnt(0)
@@ -204090,9 +204200,10 @@ define <64 x i16> @bitcast_v128i8_to_v64i16(<128 x i8> %a, i32 %b) {
; SI-NEXT: v_lshlrev_b32_e32 v17, 16, v17
; SI-NEXT: v_or_b32_e32 v17, v18, v17
; SI-NEXT: v_or_b32_e32 v16, v17, v16
-; SI-NEXT: buffer_load_dword v17, off, s[0:3], s32 offset:724 ; 4-byte Folded Reload
-; SI-NEXT: buffer_load_dword v18, off, s[0:3], s32 offset:792 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v17, off, s[0:3], s32 offset:732 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v18, off, s[0:3], s32 offset:800 ; 4-byte Folded Reload
; SI-NEXT: v_add_i32_e32 v15, vcc, s7, v16
+; SI-NEXT: v_lshrrev_b32_e32 v26, 16, v15
; SI-NEXT: s_waitcnt vmcnt(1)
; SI-NEXT: v_add_i32_e32 v17, vcc, 3, v17
; SI-NEXT: v_and_b32_e32 v17, 0xff, v17
@@ -204107,30 +204218,30 @@ define <64 x i16> @bitcast_v128i8_to_v64i16(<128 x i8> %a, i32 %b) {
; SI-NEXT: v_lshlrev_b32_e32 v18, 16, v18
; SI-NEXT: v_or_b32_e32 v18, v19, v18
; SI-NEXT: v_or_b32_e32 v18, v18, v17
-; SI-NEXT: buffer_load_dword v17, off, s[0:3], s32 offset:800 ; 4-byte Folded Reload
-; SI-NEXT: buffer_load_dword v19, off, s[0:3], s32 offset:648 ; 4-byte Folded Reload
-; SI-NEXT: v_add_i32_e32 v16, vcc, s7, v18
+; SI-NEXT: buffer_load_dword v17, off, s[0:3], s32 offset:808 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v19, off, s[0:3], s32 offset:656 ; 4-byte Folded Reload
; SI-NEXT: s_waitcnt vmcnt(1)
; SI-NEXT: v_add_i32_e32 v17, vcc, 3, v17
; SI-NEXT: v_and_b32_e32 v17, 0xff, v17
; SI-NEXT: s_waitcnt vmcnt(0)
; SI-NEXT: v_or_b32_e32 v17, v19, v17
-; SI-NEXT: v_add_i32_e32 v19, vcc, 3, v23
+; SI-NEXT: buffer_load_dword v19, off, s[0:3], s32 offset:580 ; 4-byte Folded Reload
+; SI-NEXT: v_add_i32_e32 v17, vcc, s6, v17
+; SI-NEXT: v_and_b32_e32 v17, 0xffff, v17
+; SI-NEXT: s_waitcnt vmcnt(0)
+; SI-NEXT: v_add_i32_e32 v19, vcc, 3, v19
; SI-NEXT: v_and_b32_e32 v19, 0xff, v19
; SI-NEXT: v_lshlrev_b32_e32 v19, 16, v19
-; SI-NEXT: v_add_i32_e32 v17, vcc, s6, v17
; SI-NEXT: v_or_b32_e32 v19, v20, v19
-; SI-NEXT: v_and_b32_e32 v17, 0xffff, v17
; SI-NEXT: v_or_b32_e32 v19, v19, v17
-; SI-NEXT: buffer_load_dword v17, off, s[0:3], s32 offset:668 ; 4-byte Folded Reload
-; SI-NEXT: buffer_load_dword v20, off, s[0:3], s32 offset:756 ; 4-byte Folded Reload
-; SI-NEXT: buffer_load_dword v23, off, s[0:3], s32 offset:592 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v17, off, s[0:3], s32 offset:676 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v20, off, s[0:3], s32 offset:764 ; 4-byte Folded Reload
; SI-NEXT: v_add_i32_e32 v11, vcc, s7, v19
; SI-NEXT: v_lshrrev_b32_e32 v38, 16, v11
-; SI-NEXT: s_waitcnt vmcnt(2)
+; SI-NEXT: s_waitcnt vmcnt(1)
; SI-NEXT: v_add_i32_e32 v17, vcc, 3, v17
; SI-NEXT: v_and_b32_e32 v17, 0xff, v17
-; SI-NEXT: s_waitcnt vmcnt(1)
+; SI-NEXT: s_waitcnt vmcnt(0)
; SI-NEXT: v_or_b32_e32 v17, v20, v17
; SI-NEXT: buffer_load_dword v20, off, s[0:3], s32 offset:564 ; 4-byte Folded Reload
; SI-NEXT: v_add_i32_e32 v17, vcc, s6, v17
@@ -204142,13 +204253,14 @@ define <64 x i16> @bitcast_v128i8_to_v64i16(<128 x i8> %a, i32 %b) {
; SI-NEXT: v_or_b32_e32 v20, v21, v20
; SI-NEXT: v_or_b32_e32 v20, v20, v17
; SI-NEXT: buffer_load_dword v17, off, s[0:3], s32 offset:560 ; 4-byte Folded Reload
-; SI-NEXT: buffer_load_dword v21, off, s[0:3], s32 offset:620 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v21, off, s[0:3], s32 offset:624 ; 4-byte Folded Reload
+; SI-NEXT: v_add_i32_e32 v16, vcc, s7, v20
; SI-NEXT: s_waitcnt vmcnt(1)
; SI-NEXT: v_add_i32_e32 v17, vcc, 3, v17
; SI-NEXT: v_and_b32_e32 v17, 0xff, v17
; SI-NEXT: s_waitcnt vmcnt(0)
; SI-NEXT: v_or_b32_e32 v17, v21, v17
-; SI-NEXT: buffer_load_dword v21, off, s[0:3], s32 offset:552 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v21, off, s[0:3], s32 offset:556 ; 4-byte Folded Reload
; SI-NEXT: v_add_i32_e32 v17, vcc, s6, v17
; SI-NEXT: v_and_b32_e32 v17, 0xffff, v17
; SI-NEXT: s_waitcnt vmcnt(0)
@@ -204157,15 +204269,15 @@ define <64 x i16> @bitcast_v128i8_to_v64i16(<128 x i8> %a, i32 %b) {
; SI-NEXT: v_lshlrev_b32_e32 v21, 16, v21
; SI-NEXT: v_or_b32_e32 v21, v22, v21
; SI-NEXT: v_or_b32_e32 v21, v21, v17
-; SI-NEXT: buffer_load_dword v17, off, s[0:3], s32 offset:616 ; 4-byte Folded Reload
-; SI-NEXT: buffer_load_dword v22, off, s[0:3], s32 offset:740 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v17, off, s[0:3], s32 offset:620 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v22, off, s[0:3], s32 offset:748 ; 4-byte Folded Reload
; SI-NEXT: v_add_i32_e32 v19, vcc, s7, v21
; SI-NEXT: s_waitcnt vmcnt(1)
; SI-NEXT: v_add_i32_e32 v17, vcc, 3, v17
; SI-NEXT: v_and_b32_e32 v17, 0xff, v17
; SI-NEXT: s_waitcnt vmcnt(0)
; SI-NEXT: v_or_b32_e32 v17, v22, v17
-; SI-NEXT: buffer_load_dword v22, off, s[0:3], s32 offset:780 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v22, off, s[0:3], s32 offset:788 ; 4-byte Folded Reload
; SI-NEXT: v_add_i32_e32 v17, vcc, s6, v17
; SI-NEXT: v_and_b32_e32 v17, 0xffff, v17
; SI-NEXT: s_waitcnt vmcnt(0)
@@ -204174,32 +204286,31 @@ define <64 x i16> @bitcast_v128i8_to_v64i16(<128 x i8> %a, i32 %b) {
; SI-NEXT: v_lshlrev_b32_e32 v22, 16, v22
; SI-NEXT: v_or_b32_e32 v22, v23, v22
; SI-NEXT: v_or_b32_e32 v22, v22, v17
-; SI-NEXT: buffer_load_dword v17, off, s[0:3], s32 offset:772 ; 4-byte Folded Reload
-; SI-NEXT: buffer_load_dword v23, off, s[0:3], s32 offset:588 ; 4-byte Folded Reload
-; SI-NEXT: v_add_i32_e32 v21, vcc, s7, v22
+; SI-NEXT: buffer_load_dword v17, off, s[0:3], s32 offset:780 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v23, off, s[0:3], s32 offset:592 ; 4-byte Folded Reload
+; SI-NEXT: v_add_i32_e32 v27, vcc, s7, v22
; SI-NEXT: s_waitcnt vmcnt(1)
; SI-NEXT: v_add_i32_e32 v17, vcc, 3, v17
; SI-NEXT: v_and_b32_e32 v17, 0xff, v17
; SI-NEXT: s_waitcnt vmcnt(0)
; SI-NEXT: v_or_b32_e32 v17, v23, v17
-; SI-NEXT: buffer_load_dword v23, off, s[0:3], s32 offset:760 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v23, off, s[0:3], s32 offset:768 ; 4-byte Folded Reload
; SI-NEXT: v_add_i32_e32 v17, vcc, s6, v17
; SI-NEXT: v_and_b32_e32 v17, 0xffff, v17
; SI-NEXT: s_waitcnt vmcnt(0)
; SI-NEXT: v_add_i32_e32 v23, vcc, 3, v23
; SI-NEXT: v_and_b32_e32 v23, 0xff, v23
; SI-NEXT: v_lshlrev_b32_e32 v23, 16, v23
-; SI-NEXT: v_or_b32_e32 v23, v26, v23
+; SI-NEXT: v_or_b32_e32 v23, v25, v23
; SI-NEXT: v_or_b32_e32 v23, v23, v17
-; SI-NEXT: buffer_load_dword v17, off, s[0:3], s32 offset:580 ; 4-byte Folded Reload
-; SI-NEXT: v_mov_b32_e32 v26, v25
+; SI-NEXT: buffer_load_dword v17, off, s[0:3], s32 offset:584 ; 4-byte Folded Reload
; SI-NEXT: buffer_load_dword v25, off, s[0:3], s32 offset:576 ; 4-byte Folded Reload
; SI-NEXT: v_add_i32_e32 v29, vcc, s7, v23
; SI-NEXT: s_waitcnt vmcnt(1)
; SI-NEXT: v_add_i32_e32 v17, vcc, 3, v17
; SI-NEXT: v_and_b32_e32 v17, 0xff, v17
; SI-NEXT: v_or_b32_e32 v24, v24, v17
-; SI-NEXT: buffer_load_dword v17, off, s[0:3], s32 offset:784 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v17, off, s[0:3], s32 offset:792 ; 4-byte Folded Reload
; SI-NEXT: v_add_i32_e32 v2, vcc, s6, v24
; SI-NEXT: v_and_b32_e32 v2, 0xffff, v2
; SI-NEXT: s_waitcnt vmcnt(0)
@@ -204208,16 +204319,15 @@ define <64 x i16> @bitcast_v128i8_to_v64i16(<128 x i8> %a, i32 %b) {
; SI-NEXT: v_lshlrev_b32_e32 v17, 16, v17
; SI-NEXT: v_or_b32_e32 v25, v25, v17
; SI-NEXT: v_or_b32_e32 v2, v25, v2
-; SI-NEXT: v_add_i32_e32 v18, vcc, s7, v2
-; SI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:732 ; 4-byte Folded Reload
+; SI-NEXT: v_add_i32_e32 v20, vcc, s7, v2
+; SI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:740 ; 4-byte Folded Reload
; SI-NEXT: v_add_i32_e32 v17, vcc, s7, v14
-; SI-NEXT: v_mov_b32_e32 v14, v27
-; SI-NEXT: v_add_i32_e32 v27, vcc, s7, v20
+; SI-NEXT: v_add_i32_e32 v14, vcc, s7, v18
; SI-NEXT: s_waitcnt vmcnt(0)
; SI-NEXT: v_add_i32_e32 v2, vcc, 3, v2
; SI-NEXT: v_and_b32_e32 v2, 0xff, v2
; SI-NEXT: v_or_b32_e32 v2, v3, v2
-; SI-NEXT: buffer_load_dword v3, off, s[0:3], s32 offset:696 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v3, off, s[0:3], s32 offset:708 ; 4-byte Folded Reload
; SI-NEXT: v_add_i32_e32 v2, vcc, s6, v2
; SI-NEXT: v_and_b32_e32 v2, 0xffff, v2
; SI-NEXT: s_waitcnt vmcnt(0)
@@ -204227,10 +204337,10 @@ define <64 x i16> @bitcast_v128i8_to_v64i16(<128 x i8> %a, i32 %b) {
; SI-NEXT: v_or_b32_e32 v3, v4, v3
; SI-NEXT: v_or_b32_e32 v2, v3, v2
; SI-NEXT: v_add_i32_e32 v28, vcc, s7, v2
-; SI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:656 ; 4-byte Folded Reload
-; SI-NEXT: buffer_load_dword v3, off, s[0:3], s32 offset:604 ; 4-byte Folded Reload
-; SI-NEXT: buffer_load_dword v4, off, s[0:3], s32 offset:536 ; 4-byte Folded Reload
-; SI-NEXT: v_lshrrev_b32_e32 v52, 16, v28
+; SI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:660 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v3, off, s[0:3], s32 offset:608 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v4, off, s[0:3], s32 offset:544 ; 4-byte Folded Reload
+; SI-NEXT: v_lshrrev_b32_e32 v21, 16, v28
; SI-NEXT: s_waitcnt vmcnt(2)
; SI-NEXT: v_add_i32_e32 v2, vcc, 3, v2
; SI-NEXT: v_and_b32_e32 v2, 0xff, v2
@@ -204247,8 +204357,8 @@ define <64 x i16> @bitcast_v128i8_to_v64i16(<128 x i8> %a, i32 %b) {
; SI-NEXT: v_or_b32_e32 v2, v3, v2
; SI-NEXT: v_add_i32_e32 v39, vcc, s7, v2
; SI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:492 ; 4-byte Folded Reload
-; SI-NEXT: buffer_load_dword v3, off, s[0:3], s32 offset:532 ; 4-byte Folded Reload
-; SI-NEXT: buffer_load_dword v4, off, s[0:3], s32 offset:528 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v3, off, s[0:3], s32 offset:540 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v4, off, s[0:3], s32 offset:536 ; 4-byte Folded Reload
; SI-NEXT: s_waitcnt vmcnt(2)
; SI-NEXT: v_add_i32_e32 v2, vcc, 3, v2
; SI-NEXT: v_and_b32_e32 v2, 0xff, v2
@@ -204261,16 +204371,17 @@ define <64 x i16> @bitcast_v128i8_to_v64i16(<128 x i8> %a, i32 %b) {
; SI-NEXT: v_add_i32_e32 v3, vcc, 3, v3
; SI-NEXT: v_and_b32_e32 v3, 0xff, v3
; SI-NEXT: v_lshlrev_b32_e32 v3, 16, v3
-; SI-NEXT: v_or_b32_e32 v3, v26, v3
+; SI-NEXT: v_or_b32_e32 v3, v4, v3
; SI-NEXT: v_or_b32_e32 v2, v3, v2
-; SI-NEXT: v_add_i32_e32 v26, vcc, s7, v2
-; SI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:652 ; 4-byte Folded Reload
+; SI-NEXT: v_add_i32_e32 v55, vcc, s7, v2
+; SI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:648 ; 4-byte Folded Reload
; SI-NEXT: buffer_load_dword v3, off, s[0:3], s32 offset:568 ; 4-byte Folded Reload
-; SI-NEXT: v_lshrrev_b32_e32 v20, 16, v26
-; SI-NEXT: s_waitcnt vmcnt(1)
+; SI-NEXT: buffer_load_dword v4, off, s[0:3], s32 offset:532 ; 4-byte Folded Reload
+; SI-NEXT: v_lshrrev_b32_e32 v53, 16, v55
+; SI-NEXT: s_waitcnt vmcnt(2)
; SI-NEXT: v_add_i32_e32 v2, vcc, 3, v2
; SI-NEXT: v_and_b32_e32 v2, 0xff, v2
-; SI-NEXT: s_waitcnt vmcnt(0)
+; SI-NEXT: s_waitcnt vmcnt(1)
; SI-NEXT: v_or_b32_e32 v2, v3, v2
; SI-NEXT: buffer_load_dword v3, off, s[0:3], s32 offset:484 ; 4-byte Folded Reload
; SI-NEXT: v_add_i32_e32 v2, vcc, s6, v2
@@ -204283,8 +204394,8 @@ define <64 x i16> @bitcast_v128i8_to_v64i16(<128 x i8> %a, i32 %b) {
; SI-NEXT: v_or_b32_e32 v2, v3, v2
; SI-NEXT: v_add_i32_e32 v40, vcc, s7, v2
; SI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:488 ; 4-byte Folded Reload
-; SI-NEXT: buffer_load_dword v3, off, s[0:3], s32 offset:524 ; 4-byte Folded Reload
-; SI-NEXT: buffer_load_dword v4, off, s[0:3], s32 offset:520 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v3, off, s[0:3], s32 offset:528 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v4, off, s[0:3], s32 offset:524 ; 4-byte Folded Reload
; SI-NEXT: s_waitcnt vmcnt(2)
; SI-NEXT: v_add_i32_e32 v2, vcc, 3, v2
; SI-NEXT: v_and_b32_e32 v2, 0xff, v2
@@ -204297,30 +204408,29 @@ define <64 x i16> @bitcast_v128i8_to_v64i16(<128 x i8> %a, i32 %b) {
; SI-NEXT: v_add_i32_e32 v3, vcc, 3, v3
; SI-NEXT: v_and_b32_e32 v3, 0xff, v3
; SI-NEXT: v_lshlrev_b32_e32 v3, 16, v3
-; SI-NEXT: v_or_b32_e32 v3, v31, v3
+; SI-NEXT: v_or_b32_e32 v3, v30, v3
; SI-NEXT: v_or_b32_e32 v2, v3, v2
; SI-NEXT: v_add_i32_e32 v42, vcc, s7, v2
-; SI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:640 ; 4-byte Folded Reload
-; SI-NEXT: buffer_load_dword v3, off, s[0:3], s32 offset:556 ; 4-byte Folded Reload
-; SI-NEXT: v_lshrrev_b32_e32 v55, 16, v42
+; SI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:636 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v3, off, s[0:3], s32 offset:476 ; 4-byte Folded Reload
+; SI-NEXT: v_lshrrev_b32_e32 v36, 16, v42
+; SI-NEXT: v_lshrrev_b32_e32 v30, 16, v59
; SI-NEXT: s_waitcnt vmcnt(1)
; SI-NEXT: v_add_i32_e32 v2, vcc, 3, v2
; SI-NEXT: v_and_b32_e32 v2, 0xff, v2
; SI-NEXT: s_waitcnt vmcnt(0)
-; SI-NEXT: v_or_b32_e32 v2, v3, v2
-; SI-NEXT: buffer_load_dword v3, off, s[0:3], s32 offset:476 ; 4-byte Folded Reload
-; SI-NEXT: v_add_i32_e32 v2, vcc, s6, v2
-; SI-NEXT: v_and_b32_e32 v2, 0xffff, v2
-; SI-NEXT: s_waitcnt vmcnt(0)
; SI-NEXT: v_add_i32_e32 v3, vcc, 3, v3
+; SI-NEXT: v_or_b32_e32 v2, v34, v2
; SI-NEXT: v_and_b32_e32 v3, 0xff, v3
+; SI-NEXT: v_add_i32_e32 v2, vcc, s6, v2
; SI-NEXT: v_lshlrev_b32_e32 v3, 16, v3
+; SI-NEXT: v_and_b32_e32 v2, 0xffff, v2
; SI-NEXT: v_or_b32_e32 v3, v4, v3
; SI-NEXT: v_or_b32_e32 v2, v3, v2
; SI-NEXT: v_add_i32_e32 v7, vcc, s7, v2
; SI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:480 ; 4-byte Folded Reload
-; SI-NEXT: buffer_load_dword v3, off, s[0:3], s32 offset:516 ; 4-byte Folded Reload
-; SI-NEXT: buffer_load_dword v4, off, s[0:3], s32 offset:512 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v3, off, s[0:3], s32 offset:520 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v4, off, s[0:3], s32 offset:516 ; 4-byte Folded Reload
; SI-NEXT: s_waitcnt vmcnt(2)
; SI-NEXT: v_add_i32_e32 v2, vcc, 3, v2
; SI-NEXT: v_and_b32_e32 v2, 0xff, v2
@@ -204333,28 +204443,27 @@ define <64 x i16> @bitcast_v128i8_to_v64i16(<128 x i8> %a, i32 %b) {
; SI-NEXT: v_add_i32_e32 v3, vcc, 3, v3
; SI-NEXT: v_and_b32_e32 v3, 0xff, v3
; SI-NEXT: v_lshlrev_b32_e32 v3, 16, v3
-; SI-NEXT: v_or_b32_e32 v3, v30, v3
+; SI-NEXT: v_or_b32_e32 v3, v4, v3
; SI-NEXT: v_or_b32_e32 v2, v3, v2
; SI-NEXT: v_add_i32_e32 v5, vcc, s7, v2
-; SI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:628 ; 4-byte Folded Reload
-; SI-NEXT: buffer_load_dword v3, off, s[0:3], s32 offset:544 ; 4-byte Folded Reload
-; SI-NEXT: v_lshrrev_b32_e32 v12, 16, v5
-; SI-NEXT: v_lshrrev_b32_e32 v30, 16, v59
-; SI-NEXT: s_waitcnt vmcnt(1)
+; SI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:632 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v3, off, s[0:3], s32 offset:464 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v4, off, s[0:3], s32 offset:512 ; 4-byte Folded Reload
+; SI-NEXT: v_lshrrev_b32_e32 v43, 16, v5
+; SI-NEXT: s_waitcnt vmcnt(2)
; SI-NEXT: v_add_i32_e32 v2, vcc, 3, v2
; SI-NEXT: v_and_b32_e32 v2, 0xff, v2
-; SI-NEXT: s_waitcnt vmcnt(0)
-; SI-NEXT: v_or_b32_e32 v2, v3, v2
-; SI-NEXT: buffer_load_dword v3, off, s[0:3], s32 offset:464 ; 4-byte Folded Reload
-; SI-NEXT: v_add_i32_e32 v2, vcc, s6, v2
-; SI-NEXT: v_and_b32_e32 v2, 0xffff, v2
-; SI-NEXT: s_waitcnt vmcnt(0)
+; SI-NEXT: s_waitcnt vmcnt(1)
; SI-NEXT: v_add_i32_e32 v3, vcc, 3, v3
+; SI-NEXT: v_or_b32_e32 v2, v32, v2
; SI-NEXT: v_and_b32_e32 v3, 0xff, v3
+; SI-NEXT: v_add_i32_e32 v2, vcc, s6, v2
; SI-NEXT: v_lshlrev_b32_e32 v3, 16, v3
+; SI-NEXT: v_and_b32_e32 v2, 0xffff, v2
+; SI-NEXT: s_waitcnt vmcnt(0)
; SI-NEXT: v_or_b32_e32 v3, v4, v3
; SI-NEXT: v_or_b32_e32 v2, v3, v2
-; SI-NEXT: v_add_i32_e32 v43, vcc, s7, v2
+; SI-NEXT: v_add_i32_e32 v12, vcc, s7, v2
; SI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:468 ; 4-byte Folded Reload
; SI-NEXT: buffer_load_dword v3, off, s[0:3], s32 offset:508 ; 4-byte Folded Reload
; SI-NEXT: buffer_load_dword v4, off, s[0:3], s32 offset:504 ; 4-byte Folded Reload
@@ -204363,7 +204472,7 @@ define <64 x i16> @bitcast_v128i8_to_v64i16(<128 x i8> %a, i32 %b) {
; SI-NEXT: v_and_b32_e32 v2, 0xff, v2
; SI-NEXT: s_waitcnt vmcnt(1)
; SI-NEXT: v_or_b32_e32 v2, v3, v2
-; SI-NEXT: buffer_load_dword v3, off, s[0:3], s32 offset:944 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v3, off, s[0:3], s32 offset:980 ; 4-byte Folded Reload
; SI-NEXT: v_add_i32_e32 v2, vcc, s6, v2
; SI-NEXT: v_and_b32_e32 v2, 0xffff, v2
; SI-NEXT: s_waitcnt vmcnt(0)
@@ -204373,236 +204482,239 @@ define <64 x i16> @bitcast_v128i8_to_v64i16(<128 x i8> %a, i32 %b) {
; SI-NEXT: v_or_b32_e32 v3, v4, v3
; SI-NEXT: v_or_b32_e32 v2, v3, v2
; SI-NEXT: v_add_i32_e32 v22, vcc, s7, v2
-; SI-NEXT: v_alignbit_b32 v2, v22, v43, 16
-; SI-NEXT: buffer_store_dword v2, off, s[0:3], s32 offset:852 ; 4-byte Folded Spill
+; SI-NEXT: v_alignbit_b32 v2, v22, v12, 16
+; SI-NEXT: buffer_store_dword v2, off, s[0:3], s32 offset:860 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_alignbit_b32 v2, v5, v7, 16
-; SI-NEXT: buffer_store_dword v2, off, s[0:3], s32 offset:856 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v2, off, s[0:3], s32 offset:864 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_alignbit_b32 v2, v42, v40, 16
-; SI-NEXT: buffer_store_dword v2, off, s[0:3], s32 offset:860 ; 4-byte Folded Spill
-; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_alignbit_b32 v2, v26, v39, 16
-; SI-NEXT: buffer_store_dword v2, off, s[0:3], s32 offset:864 ; 4-byte Folded Spill
-; SI-NEXT: buffer_store_dword v18, off, s[0:3], s32 offset:716 ; 4-byte Folded Spill
-; SI-NEXT: s_waitcnt expcnt(1)
-; SI-NEXT: v_alignbit_b32 v2, v28, v18, 16
; SI-NEXT: buffer_store_dword v2, off, s[0:3], s32 offset:868 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_alignbit_b32 v2, v29, v21, 16
+; SI-NEXT: v_alignbit_b32 v2, v55, v39, 16
; SI-NEXT: buffer_store_dword v2, off, s[0:3], s32 offset:872 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_alignbit_b32 v2, v19, v27, 16
+; SI-NEXT: v_alignbit_b32 v2, v28, v20, 16
; SI-NEXT: buffer_store_dword v2, off, s[0:3], s32 offset:876 ; 4-byte Folded Spill
-; SI-NEXT: buffer_store_dword v16, off, s[0:3], s32 offset:684 ; 4-byte Folded Spill
-; SI-NEXT: s_waitcnt expcnt(1)
-; SI-NEXT: v_alignbit_b32 v2, v11, v16, 16
+; SI-NEXT: s_waitcnt expcnt(0)
+; SI-NEXT: v_alignbit_b32 v2, v29, v27, 16
; SI-NEXT: buffer_store_dword v2, off, s[0:3], s32 offset:880 ; 4-byte Folded Spill
-; SI-NEXT: buffer_store_dword v13, off, s[0:3], s32 offset:680 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v16, off, s[0:3], s32 offset:696 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(1)
-; SI-NEXT: v_alignbit_b32 v2, v15, v13, 16
+; SI-NEXT: v_alignbit_b32 v2, v19, v16, 16
; SI-NEXT: buffer_store_dword v2, off, s[0:3], s32 offset:884 ; 4-byte Folded Spill
-; SI-NEXT: buffer_store_dword v10, off, s[0:3], s32 offset:676 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v14, off, s[0:3], s32 offset:692 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(1)
-; SI-NEXT: v_alignbit_b32 v2, v17, v10, 16
+; SI-NEXT: v_alignbit_b32 v2, v11, v14, 16
; SI-NEXT: buffer_store_dword v2, off, s[0:3], s32 offset:888 ; 4-byte Folded Spill
-; SI-NEXT: buffer_store_dword v9, off, s[0:3], s32 offset:660 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v13, off, s[0:3], s32 offset:688 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(1)
-; SI-NEXT: v_alignbit_b32 v2, v1, v9, 16
+; SI-NEXT: v_alignbit_b32 v2, v15, v13, 16
; SI-NEXT: buffer_store_dword v2, off, s[0:3], s32 offset:892 ; 4-byte Folded Spill
-; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_alignbit_b32 v2, v62, v8, 16
+; SI-NEXT: buffer_store_dword v10, off, s[0:3], s32 offset:684 ; 4-byte Folded Spill
+; SI-NEXT: s_waitcnt expcnt(1)
+; SI-NEXT: v_alignbit_b32 v2, v17, v10, 16
; SI-NEXT: buffer_store_dword v2, off, s[0:3], s32 offset:896 ; 4-byte Folded Spill
-; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_alignbit_b32 v2, v59, v6, 16
+; SI-NEXT: buffer_store_dword v9, off, s[0:3], s32 offset:668 ; 4-byte Folded Spill
+; SI-NEXT: s_waitcnt expcnt(1)
+; SI-NEXT: v_alignbit_b32 v2, v1, v9, 16
; SI-NEXT: buffer_store_dword v2, off, s[0:3], s32 offset:900 ; 4-byte Folded Spill
-; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_alignbit_b32 v2, v47, v33, 16
+; SI-NEXT: buffer_store_dword v8, off, s[0:3], s32 offset:664 ; 4-byte Folded Spill
+; SI-NEXT: s_waitcnt expcnt(1)
+; SI-NEXT: v_alignbit_b32 v2, v62, v8, 16
; SI-NEXT: buffer_store_dword v2, off, s[0:3], s32 offset:904 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_alignbit_b32 v2, v45, v32, 16
+; SI-NEXT: v_alignbit_b32 v2, v59, v6, 16
; SI-NEXT: buffer_store_dword v2, off, s[0:3], s32 offset:908 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_alignbit_b32 v2, v44, v14, 16
+; SI-NEXT: v_alignbit_b32 v2, v47, v35, 16
; SI-NEXT: buffer_store_dword v2, off, s[0:3], s32 offset:912 ; 4-byte Folded Spill
-; SI-NEXT: v_lshrrev_b32_e32 v58, 16, v22
-; SI-NEXT: v_lshrrev_b32_e32 v13, 16, v29
-; SI-NEXT: v_lshrrev_b32_e32 v8, 16, v19
-; SI-NEXT: v_lshrrev_b32_e32 v18, 16, v15
-; SI-NEXT: v_lshrrev_b32_e32 v10, 16, v17
+; SI-NEXT: s_waitcnt expcnt(0)
+; SI-NEXT: v_alignbit_b32 v2, v45, v33, 16
+; SI-NEXT: buffer_store_dword v2, off, s[0:3], s32 offset:916 ; 4-byte Folded Spill
+; SI-NEXT: s_waitcnt expcnt(0)
+; SI-NEXT: v_alignbit_b32 v2, v44, v31, 16
+; SI-NEXT: buffer_store_dword v2, off, s[0:3], s32 offset:920 ; 4-byte Folded Spill
+; SI-NEXT: v_lshrrev_b32_e32 v57, 16, v22
+; SI-NEXT: v_lshrrev_b32_e32 v16, 16, v29
+; SI-NEXT: v_lshrrev_b32_e32 v10, 16, v19
+; SI-NEXT: v_lshrrev_b32_e32 v6, 16, v17
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_lshrrev_b32_e32 v2, 16, v1
; SI-NEXT: v_lshrrev_b32_e32 v33, 16, v62
-; SI-NEXT: v_lshrrev_b32_e32 v6, 16, v47
+; SI-NEXT: v_lshrrev_b32_e32 v8, 16, v47
+; SI-NEXT: v_lshrrev_b32_e32 v4, 16, v45
; SI-NEXT: v_lshrrev_b32_e32 v3, 16, v44
-; SI-NEXT: buffer_store_dword v3, off, s[0:3], s32 offset:940 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v3, off, s[0:3], s32 offset:948 ; 4-byte Folded Spill
; SI-NEXT: .LBB96_4: ; %end
; SI-NEXT: s_or_b64 exec, exec, s[4:5]
-; SI-NEXT: buffer_load_dword v4, off, s[0:3], s32 offset:852 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v9, off, s[0:3], s32 offset:860 ; 4-byte Folded Reload
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_and_b32_e32 v3, 0xffff, v43
+; SI-NEXT: v_and_b32_e32 v3, 0xffff, v12
; SI-NEXT: v_and_b32_e32 v1, 0xffff, v1
; SI-NEXT: v_lshlrev_b32_e32 v2, 16, v2
; SI-NEXT: v_or_b32_e32 v1, v1, v2
; SI-NEXT: v_add_i32_e32 v2, vcc, 0x54, v0
; SI-NEXT: s_waitcnt vmcnt(0)
-; SI-NEXT: v_lshlrev_b32_e32 v4, 16, v4
-; SI-NEXT: v_or_b32_e32 v3, v3, v4
+; SI-NEXT: v_lshlrev_b32_e32 v9, 16, v9
+; SI-NEXT: v_or_b32_e32 v3, v3, v9
; SI-NEXT: buffer_store_dword v3, v0, s[0:3], 0 offen
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_and_b32_e32 v3, 0xffff, v22
-; SI-NEXT: v_lshlrev_b32_e32 v4, 16, v58
-; SI-NEXT: v_or_b32_e32 v3, v3, v4
-; SI-NEXT: v_add_i32_e32 v4, vcc, 4, v0
-; SI-NEXT: buffer_store_dword v3, v4, s[0:3], 0 offen
-; SI-NEXT: buffer_load_dword v4, off, s[0:3], s32 offset:856 ; 4-byte Folded Reload
+; SI-NEXT: v_lshlrev_b32_e32 v9, 16, v57
+; SI-NEXT: v_or_b32_e32 v3, v3, v9
+; SI-NEXT: v_add_i32_e32 v9, vcc, 4, v0
+; SI-NEXT: buffer_store_dword v3, v9, s[0:3], 0 offen
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_and_b32_e32 v3, 0xffff, v7
+; SI-NEXT: buffer_load_dword v7, off, s[0:3], s32 offset:864 ; 4-byte Folded Reload
; SI-NEXT: s_waitcnt vmcnt(0)
-; SI-NEXT: v_lshlrev_b32_e32 v4, 16, v4
-; SI-NEXT: v_or_b32_e32 v3, v3, v4
-; SI-NEXT: v_add_i32_e32 v4, vcc, 8, v0
-; SI-NEXT: buffer_store_dword v3, v4, s[0:3], 0 offen
+; SI-NEXT: v_lshlrev_b32_e32 v7, 16, v7
+; SI-NEXT: v_or_b32_e32 v3, v3, v7
+; SI-NEXT: v_add_i32_e32 v7, vcc, 8, v0
+; SI-NEXT: buffer_store_dword v3, v7, s[0:3], 0 offen
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_and_b32_e32 v3, 0xffff, v5
-; SI-NEXT: v_lshlrev_b32_e32 v4, 16, v12
-; SI-NEXT: v_or_b32_e32 v3, v3, v4
-; SI-NEXT: v_add_i32_e32 v4, vcc, 12, v0
-; SI-NEXT: buffer_store_dword v3, v4, s[0:3], 0 offen
-; SI-NEXT: buffer_load_dword v4, off, s[0:3], s32 offset:860 ; 4-byte Folded Reload
+; SI-NEXT: v_lshlrev_b32_e32 v5, 16, v43
+; SI-NEXT: v_or_b32_e32 v3, v3, v5
+; SI-NEXT: v_add_i32_e32 v5, vcc, 12, v0
+; SI-NEXT: buffer_store_dword v3, v5, s[0:3], 0 offen
+; SI-NEXT: buffer_load_dword v5, off, s[0:3], s32 offset:868 ; 4-byte Folded Reload
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_and_b32_e32 v3, 0xffff, v40
; SI-NEXT: s_waitcnt vmcnt(0)
-; SI-NEXT: v_lshlrev_b32_e32 v4, 16, v4
-; SI-NEXT: v_or_b32_e32 v3, v3, v4
-; SI-NEXT: v_add_i32_e32 v4, vcc, 16, v0
-; SI-NEXT: buffer_store_dword v3, v4, s[0:3], 0 offen
+; SI-NEXT: v_lshlrev_b32_e32 v5, 16, v5
+; SI-NEXT: v_or_b32_e32 v3, v3, v5
+; SI-NEXT: v_add_i32_e32 v5, vcc, 16, v0
+; SI-NEXT: buffer_store_dword v3, v5, s[0:3], 0 offen
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_and_b32_e32 v3, 0xffff, v42
-; SI-NEXT: v_lshlrev_b32_e32 v4, 16, v55
-; SI-NEXT: v_or_b32_e32 v3, v3, v4
-; SI-NEXT: v_add_i32_e32 v4, vcc, 20, v0
-; SI-NEXT: buffer_store_dword v3, v4, s[0:3], 0 offen
-; SI-NEXT: buffer_load_dword v4, off, s[0:3], s32 offset:864 ; 4-byte Folded Reload
+; SI-NEXT: v_lshlrev_b32_e32 v5, 16, v36
+; SI-NEXT: v_or_b32_e32 v3, v3, v5
+; SI-NEXT: v_add_i32_e32 v5, vcc, 20, v0
+; SI-NEXT: buffer_store_dword v3, v5, s[0:3], 0 offen
+; SI-NEXT: buffer_load_dword v5, off, s[0:3], s32 offset:872 ; 4-byte Folded Reload
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_and_b32_e32 v3, 0xffff, v39
; SI-NEXT: s_waitcnt vmcnt(0)
-; SI-NEXT: v_lshlrev_b32_e32 v4, 16, v4
-; SI-NEXT: v_or_b32_e32 v3, v3, v4
-; SI-NEXT: v_add_i32_e32 v4, vcc, 24, v0
-; SI-NEXT: buffer_store_dword v3, v4, s[0:3], 0 offen
+; SI-NEXT: v_lshlrev_b32_e32 v5, 16, v5
+; SI-NEXT: v_or_b32_e32 v3, v3, v5
+; SI-NEXT: v_add_i32_e32 v5, vcc, 24, v0
+; SI-NEXT: buffer_store_dword v3, v5, s[0:3], 0 offen
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_and_b32_e32 v3, 0xffff, v26
-; SI-NEXT: v_lshlrev_b32_e32 v4, 16, v20
-; SI-NEXT: v_or_b32_e32 v3, v3, v4
-; SI-NEXT: v_add_i32_e32 v4, vcc, 28, v0
-; SI-NEXT: buffer_store_dword v3, v4, s[0:3], 0 offen
+; SI-NEXT: v_and_b32_e32 v3, 0xffff, v55
+; SI-NEXT: v_lshlrev_b32_e32 v5, 16, v53
+; SI-NEXT: v_or_b32_e32 v3, v3, v5
+; SI-NEXT: v_add_i32_e32 v5, vcc, 28, v0
+; SI-NEXT: buffer_store_dword v3, v5, s[0:3], 0 offen
+; SI-NEXT: buffer_load_dword v5, off, s[0:3], s32 offset:876 ; 4-byte Folded Reload
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: buffer_load_dword v3, off, s[0:3], s32 offset:716 ; 4-byte Folded Reload
-; SI-NEXT: buffer_load_dword v4, off, s[0:3], s32 offset:868 ; 4-byte Folded Reload
-; SI-NEXT: s_waitcnt vmcnt(1)
-; SI-NEXT: v_and_b32_e32 v3, 0xffff, v3
+; SI-NEXT: v_and_b32_e32 v3, 0xffff, v20
; SI-NEXT: s_waitcnt vmcnt(0)
-; SI-NEXT: v_lshlrev_b32_e32 v4, 16, v4
-; SI-NEXT: v_or_b32_e32 v3, v3, v4
-; SI-NEXT: v_add_i32_e32 v4, vcc, 32, v0
-; SI-NEXT: buffer_store_dword v3, v4, s[0:3], 0 offen
+; SI-NEXT: v_lshlrev_b32_e32 v5, 16, v5
+; SI-NEXT: v_or_b32_e32 v3, v3, v5
+; SI-NEXT: v_add_i32_e32 v5, vcc, 32, v0
+; SI-NEXT: buffer_store_dword v3, v5, s[0:3], 0 offen
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_and_b32_e32 v3, 0xffff, v28
-; SI-NEXT: v_lshlrev_b32_e32 v4, 16, v52
-; SI-NEXT: v_or_b32_e32 v3, v3, v4
-; SI-NEXT: v_add_i32_e32 v4, vcc, 36, v0
-; SI-NEXT: buffer_store_dword v3, v4, s[0:3], 0 offen
-; SI-NEXT: buffer_load_dword v4, off, s[0:3], s32 offset:872 ; 4-byte Folded Reload
+; SI-NEXT: v_lshlrev_b32_e32 v5, 16, v21
+; SI-NEXT: v_or_b32_e32 v3, v3, v5
+; SI-NEXT: v_add_i32_e32 v5, vcc, 36, v0
+; SI-NEXT: buffer_store_dword v3, v5, s[0:3], 0 offen
+; SI-NEXT: buffer_load_dword v5, off, s[0:3], s32 offset:880 ; 4-byte Folded Reload
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_and_b32_e32 v3, 0xffff, v21
+; SI-NEXT: v_and_b32_e32 v3, 0xffff, v27
; SI-NEXT: s_waitcnt vmcnt(0)
-; SI-NEXT: v_lshlrev_b32_e32 v4, 16, v4
-; SI-NEXT: v_or_b32_e32 v3, v3, v4
-; SI-NEXT: v_add_i32_e32 v4, vcc, 40, v0
-; SI-NEXT: buffer_store_dword v3, v4, s[0:3], 0 offen
+; SI-NEXT: v_lshlrev_b32_e32 v5, 16, v5
+; SI-NEXT: v_or_b32_e32 v3, v3, v5
+; SI-NEXT: v_add_i32_e32 v5, vcc, 40, v0
+; SI-NEXT: buffer_store_dword v3, v5, s[0:3], 0 offen
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_and_b32_e32 v3, 0xffff, v29
-; SI-NEXT: v_lshlrev_b32_e32 v4, 16, v13
-; SI-NEXT: v_or_b32_e32 v3, v3, v4
-; SI-NEXT: v_add_i32_e32 v4, vcc, 44, v0
-; SI-NEXT: buffer_store_dword v3, v4, s[0:3], 0 offen
-; SI-NEXT: buffer_load_dword v4, off, s[0:3], s32 offset:876 ; 4-byte Folded Reload
+; SI-NEXT: v_lshlrev_b32_e32 v5, 16, v16
+; SI-NEXT: v_or_b32_e32 v3, v3, v5
+; SI-NEXT: v_add_i32_e32 v5, vcc, 44, v0
+; SI-NEXT: buffer_store_dword v3, v5, s[0:3], 0 offen
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_and_b32_e32 v3, 0xffff, v27
+; SI-NEXT: buffer_load_dword v3, off, s[0:3], s32 offset:696 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v5, off, s[0:3], s32 offset:884 ; 4-byte Folded Reload
+; SI-NEXT: s_waitcnt vmcnt(1)
+; SI-NEXT: v_and_b32_e32 v3, 0xffff, v3
; SI-NEXT: s_waitcnt vmcnt(0)
-; SI-NEXT: v_lshlrev_b32_e32 v4, 16, v4
-; SI-NEXT: v_or_b32_e32 v3, v3, v4
-; SI-NEXT: v_add_i32_e32 v4, vcc, 48, v0
-; SI-NEXT: buffer_store_dword v3, v4, s[0:3], 0 offen
+; SI-NEXT: v_lshlrev_b32_e32 v5, 16, v5
+; SI-NEXT: v_or_b32_e32 v3, v3, v5
+; SI-NEXT: v_add_i32_e32 v5, vcc, 48, v0
+; SI-NEXT: buffer_store_dword v3, v5, s[0:3], 0 offen
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_and_b32_e32 v3, 0xffff, v19
-; SI-NEXT: v_lshlrev_b32_e32 v4, 16, v8
-; SI-NEXT: v_or_b32_e32 v3, v3, v4
-; SI-NEXT: v_add_i32_e32 v4, vcc, 52, v0
-; SI-NEXT: buffer_store_dword v3, v4, s[0:3], 0 offen
+; SI-NEXT: v_lshlrev_b32_e32 v5, 16, v10
+; SI-NEXT: v_or_b32_e32 v3, v3, v5
+; SI-NEXT: v_add_i32_e32 v5, vcc, 52, v0
+; SI-NEXT: buffer_store_dword v3, v5, s[0:3], 0 offen
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: buffer_load_dword v3, off, s[0:3], s32 offset:684 ; 4-byte Folded Reload
-; SI-NEXT: buffer_load_dword v4, off, s[0:3], s32 offset:880 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v3, off, s[0:3], s32 offset:692 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v5, off, s[0:3], s32 offset:888 ; 4-byte Folded Reload
; SI-NEXT: s_waitcnt vmcnt(1)
; SI-NEXT: v_and_b32_e32 v3, 0xffff, v3
; SI-NEXT: s_waitcnt vmcnt(0)
-; SI-NEXT: v_lshlrev_b32_e32 v4, 16, v4
-; SI-NEXT: v_or_b32_e32 v3, v3, v4
-; SI-NEXT: v_add_i32_e32 v4, vcc, 56, v0
-; SI-NEXT: buffer_store_dword v3, v4, s[0:3], 0 offen
+; SI-NEXT: v_lshlrev_b32_e32 v5, 16, v5
+; SI-NEXT: v_or_b32_e32 v3, v3, v5
+; SI-NEXT: v_add_i32_e32 v5, vcc, 56, v0
+; SI-NEXT: buffer_store_dword v3, v5, s[0:3], 0 offen
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_and_b32_e32 v3, 0xffff, v11
-; SI-NEXT: v_lshlrev_b32_e32 v4, 16, v38
-; SI-NEXT: v_or_b32_e32 v3, v3, v4
-; SI-NEXT: v_add_i32_e32 v4, vcc, 60, v0
-; SI-NEXT: buffer_store_dword v3, v4, s[0:3], 0 offen
+; SI-NEXT: v_lshlrev_b32_e32 v5, 16, v38
+; SI-NEXT: v_or_b32_e32 v3, v3, v5
+; SI-NEXT: v_add_i32_e32 v5, vcc, 60, v0
+; SI-NEXT: buffer_store_dword v3, v5, s[0:3], 0 offen
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: buffer_load_dword v3, off, s[0:3], s32 offset:680 ; 4-byte Folded Reload
-; SI-NEXT: buffer_load_dword v4, off, s[0:3], s32 offset:884 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v3, off, s[0:3], s32 offset:688 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v5, off, s[0:3], s32 offset:892 ; 4-byte Folded Reload
; SI-NEXT: s_waitcnt vmcnt(1)
; SI-NEXT: v_and_b32_e32 v3, 0xffff, v3
; SI-NEXT: s_waitcnt vmcnt(0)
-; SI-NEXT: v_lshlrev_b32_e32 v4, 16, v4
-; SI-NEXT: v_or_b32_e32 v3, v3, v4
-; SI-NEXT: v_add_i32_e32 v4, vcc, 64, v0
-; SI-NEXT: buffer_store_dword v3, v4, s[0:3], 0 offen
+; SI-NEXT: v_lshlrev_b32_e32 v5, 16, v5
+; SI-NEXT: v_or_b32_e32 v3, v3, v5
+; SI-NEXT: v_add_i32_e32 v5, vcc, 64, v0
+; SI-NEXT: buffer_store_dword v3, v5, s[0:3], 0 offen
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_and_b32_e32 v3, 0xffff, v15
-; SI-NEXT: v_lshlrev_b32_e32 v4, 16, v18
-; SI-NEXT: v_or_b32_e32 v3, v3, v4
-; SI-NEXT: v_add_i32_e32 v4, vcc, 0x44, v0
-; SI-NEXT: buffer_store_dword v3, v4, s[0:3], 0 offen
+; SI-NEXT: v_lshlrev_b32_e32 v5, 16, v26
+; SI-NEXT: v_or_b32_e32 v3, v3, v5
+; SI-NEXT: v_add_i32_e32 v5, vcc, 0x44, v0
+; SI-NEXT: buffer_store_dword v3, v5, s[0:3], 0 offen
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: buffer_load_dword v3, off, s[0:3], s32 offset:676 ; 4-byte Folded Reload
-; SI-NEXT: buffer_load_dword v4, off, s[0:3], s32 offset:888 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v3, off, s[0:3], s32 offset:684 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v5, off, s[0:3], s32 offset:896 ; 4-byte Folded Reload
; SI-NEXT: s_waitcnt vmcnt(1)
; SI-NEXT: v_and_b32_e32 v3, 0xffff, v3
; SI-NEXT: s_waitcnt vmcnt(0)
-; SI-NEXT: v_lshlrev_b32_e32 v4, 16, v4
-; SI-NEXT: v_or_b32_e32 v3, v3, v4
-; SI-NEXT: v_add_i32_e32 v4, vcc, 0x48, v0
-; SI-NEXT: buffer_store_dword v3, v4, s[0:3], 0 offen
+; SI-NEXT: v_lshlrev_b32_e32 v5, 16, v5
+; SI-NEXT: v_or_b32_e32 v3, v3, v5
+; SI-NEXT: v_add_i32_e32 v5, vcc, 0x48, v0
+; SI-NEXT: buffer_store_dword v3, v5, s[0:3], 0 offen
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_and_b32_e32 v3, 0xffff, v17
-; SI-NEXT: v_lshlrev_b32_e32 v4, 16, v10
-; SI-NEXT: v_or_b32_e32 v3, v3, v4
-; SI-NEXT: v_add_i32_e32 v4, vcc, 0x4c, v0
-; SI-NEXT: buffer_store_dword v3, v4, s[0:3], 0 offen
+; SI-NEXT: v_lshlrev_b32_e32 v5, 16, v6
+; SI-NEXT: v_or_b32_e32 v3, v3, v5
+; SI-NEXT: v_add_i32_e32 v5, vcc, 0x4c, v0
+; SI-NEXT: buffer_store_dword v3, v5, s[0:3], 0 offen
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: buffer_load_dword v3, off, s[0:3], s32 offset:660 ; 4-byte Folded Reload
-; SI-NEXT: buffer_load_dword v4, off, s[0:3], s32 offset:892 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v3, off, s[0:3], s32 offset:668 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v5, off, s[0:3], s32 offset:900 ; 4-byte Folded Reload
; SI-NEXT: s_waitcnt vmcnt(1)
; SI-NEXT: v_and_b32_e32 v3, 0xffff, v3
; SI-NEXT: s_waitcnt vmcnt(0)
-; SI-NEXT: v_lshlrev_b32_e32 v4, 16, v4
-; SI-NEXT: v_or_b32_e32 v3, v3, v4
-; SI-NEXT: v_add_i32_e32 v4, vcc, 0x50, v0
-; SI-NEXT: buffer_store_dword v3, v4, s[0:3], 0 offen
+; SI-NEXT: v_lshlrev_b32_e32 v5, 16, v5
+; SI-NEXT: v_or_b32_e32 v3, v3, v5
+; SI-NEXT: v_add_i32_e32 v5, vcc, 0x50, v0
+; SI-NEXT: buffer_store_dword v3, v5, s[0:3], 0 offen
; SI-NEXT: buffer_store_dword v1, v2, s[0:3], 0 offen
-; SI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:896 ; 4-byte Folded Reload
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_and_b32_e32 v1, 0xffff, v46
+; SI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:664 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:904 ; 4-byte Folded Reload
+; SI-NEXT: s_waitcnt vmcnt(1)
+; SI-NEXT: v_and_b32_e32 v1, 0xffff, v1
; SI-NEXT: s_waitcnt vmcnt(0)
; SI-NEXT: v_lshlrev_b32_e32 v2, 16, v2
; SI-NEXT: v_or_b32_e32 v1, v1, v2
@@ -204614,9 +204726,9 @@ define <64 x i16> @bitcast_v128i8_to_v64i16(<128 x i8> %a, i32 %b) {
; SI-NEXT: v_or_b32_e32 v1, v1, v2
; SI-NEXT: v_add_i32_e32 v2, vcc, 0x5c, v0
; SI-NEXT: buffer_store_dword v1, v2, s[0:3], 0 offen
-; SI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:900 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:908 ; 4-byte Folded Reload
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_and_b32_e32 v1, 0xffff, v50
+; SI-NEXT: v_and_b32_e32 v1, 0xffff, v56
; SI-NEXT: s_waitcnt vmcnt(0)
; SI-NEXT: v_lshlrev_b32_e32 v2, 16, v2
; SI-NEXT: v_or_b32_e32 v1, v1, v2
@@ -204628,9 +204740,9 @@ define <64 x i16> @bitcast_v128i8_to_v64i16(<128 x i8> %a, i32 %b) {
; SI-NEXT: v_or_b32_e32 v1, v1, v2
; SI-NEXT: v_add_i32_e32 v2, vcc, 0x64, v0
; SI-NEXT: buffer_store_dword v1, v2, s[0:3], 0 offen
-; SI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:904 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:912 ; 4-byte Folded Reload
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_and_b32_e32 v1, 0xffff, v49
+; SI-NEXT: v_and_b32_e32 v1, 0xffff, v50
; SI-NEXT: s_waitcnt vmcnt(0)
; SI-NEXT: v_lshlrev_b32_e32 v2, 16, v2
; SI-NEXT: v_or_b32_e32 v1, v1, v2
@@ -204638,13 +204750,13 @@ define <64 x i16> @bitcast_v128i8_to_v64i16(<128 x i8> %a, i32 %b) {
; SI-NEXT: buffer_store_dword v1, v2, s[0:3], 0 offen
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_and_b32_e32 v1, 0xffff, v47
-; SI-NEXT: v_lshlrev_b32_e32 v2, 16, v6
+; SI-NEXT: v_lshlrev_b32_e32 v2, 16, v8
; SI-NEXT: v_or_b32_e32 v1, v1, v2
; SI-NEXT: v_add_i32_e32 v2, vcc, 0x6c, v0
; SI-NEXT: buffer_store_dword v1, v2, s[0:3], 0 offen
-; SI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:908 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:916 ; 4-byte Folded Reload
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_and_b32_e32 v1, 0xffff, v48
+; SI-NEXT: v_and_b32_e32 v1, 0xffff, v49
; SI-NEXT: s_waitcnt vmcnt(0)
; SI-NEXT: v_lshlrev_b32_e32 v2, 16, v2
; SI-NEXT: v_or_b32_e32 v1, v1, v2
@@ -204652,19 +204764,19 @@ define <64 x i16> @bitcast_v128i8_to_v64i16(<128 x i8> %a, i32 %b) {
; SI-NEXT: buffer_store_dword v1, v2, s[0:3], 0 offen
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_and_b32_e32 v1, 0xffff, v45
-; SI-NEXT: v_lshlrev_b32_e32 v2, 16, v36
+; SI-NEXT: v_lshlrev_b32_e32 v2, 16, v4
; SI-NEXT: v_or_b32_e32 v1, v1, v2
; SI-NEXT: v_add_i32_e32 v2, vcc, 0x74, v0
; SI-NEXT: buffer_store_dword v1, v2, s[0:3], 0 offen
-; SI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:912 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:920 ; 4-byte Folded Reload
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_and_b32_e32 v1, 0xffff, v14
+; SI-NEXT: v_and_b32_e32 v1, 0xffff, v48
; SI-NEXT: s_waitcnt vmcnt(0)
; SI-NEXT: v_lshlrev_b32_e32 v2, 16, v2
; SI-NEXT: v_or_b32_e32 v1, v1, v2
; SI-NEXT: v_add_i32_e32 v2, vcc, 0x78, v0
; SI-NEXT: buffer_store_dword v1, v2, s[0:3], 0 offen
-; SI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:940 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:948 ; 4-byte Folded Reload
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_and_b32_e32 v1, 0xffff, v44
; SI-NEXT: v_add_i32_e32 v0, vcc, 0x7c, v0
@@ -208877,8 +208989,8 @@ define inreg <64 x i16> @bitcast_v128i8_to_v64i16_scalar(<128 x i8> inreg %a, i3
; SI: ; %bb.0:
; SI-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; SI-NEXT: s_or_saveexec_b64 s[4:5], -1
-; SI-NEXT: buffer_store_dword v63, off, s[0:3], s32 offset:428 ; 4-byte Folded Spill
-; SI-NEXT: buffer_store_dword v62, off, s[0:3], s32 offset:432 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v63, off, s[0:3], s32 offset:424 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v62, off, s[0:3], s32 offset:428 ; 4-byte Folded Spill
; SI-NEXT: s_mov_b64 exec, s[4:5]
; SI-NEXT: buffer_store_dword v40, off, s[0:3], s32 offset:392 ; 4-byte Folded Spill
; SI-NEXT: buffer_store_dword v41, off, s[0:3], s32 offset:388 ; 4-byte Folded Spill
@@ -208896,21 +209008,21 @@ define inreg <64 x i16> @bitcast_v128i8_to_v64i16_scalar(<128 x i8> inreg %a, i3
; SI-NEXT: buffer_store_dword v61, off, s[0:3], s32 offset:340 ; 4-byte Folded Spill
; SI-NEXT: buffer_store_dword v29, off, s[0:3], s32 offset:396 ; 4-byte Folded Spill
; SI-NEXT: buffer_store_dword v21, off, s[0:3], s32 offset:400 ; 4-byte Folded Spill
-; SI-NEXT: buffer_load_dword v34, off, s[0:3], s32 offset:332
-; SI-NEXT: buffer_load_dword v38, off, s[0:3], s32 offset:328
-; SI-NEXT: buffer_load_dword v49, off, s[0:3], s32 offset:324
-; SI-NEXT: buffer_load_dword v50, off, s[0:3], s32 offset:320
-; SI-NEXT: buffer_load_dword v52, off, s[0:3], s32 offset:308
-; SI-NEXT: buffer_load_dword v54, off, s[0:3], s32 offset:304
-; SI-NEXT: buffer_load_dword v55, off, s[0:3], s32 offset:300
-; SI-NEXT: buffer_load_dword v41, off, s[0:3], s32 offset:296
-; SI-NEXT: buffer_load_dword v42, off, s[0:3], s32 offset:292
-; SI-NEXT: buffer_load_dword v43, off, s[0:3], s32 offset:288
+; SI-NEXT: buffer_load_dword v33, off, s[0:3], s32 offset:332
+; SI-NEXT: buffer_load_dword v34, off, s[0:3], s32 offset:328
+; SI-NEXT: buffer_load_dword v37, off, s[0:3], s32 offset:324
+; SI-NEXT: buffer_load_dword v48, off, s[0:3], s32 offset:320
+; SI-NEXT: buffer_load_dword v50, off, s[0:3], s32 offset:308
+; SI-NEXT: buffer_load_dword v52, off, s[0:3], s32 offset:304
+; SI-NEXT: buffer_load_dword v53, off, s[0:3], s32 offset:300
+; SI-NEXT: buffer_load_dword v55, off, s[0:3], s32 offset:296
+; SI-NEXT: buffer_load_dword v41, off, s[0:3], s32 offset:292
+; SI-NEXT: buffer_load_dword v42, off, s[0:3], s32 offset:288
; SI-NEXT: buffer_load_dword v31, off, s[0:3], s32 offset:276
; SI-NEXT: buffer_load_dword v32, off, s[0:3], s32 offset:272
-; SI-NEXT: buffer_load_dword v33, off, s[0:3], s32 offset:268
+; SI-NEXT: buffer_load_dword v44, off, s[0:3], s32 offset:268
; SI-NEXT: buffer_load_dword v45, off, s[0:3], s32 offset:264
-; SI-NEXT: buffer_load_dword v46, off, s[0:3], s32 offset:260
+; SI-NEXT: buffer_load_dword v47, off, s[0:3], s32 offset:260
; SI-NEXT: buffer_load_dword v56, off, s[0:3], s32 offset:256
; SI-NEXT: s_waitcnt expcnt(6)
; SI-NEXT: buffer_load_dword v57, off, s[0:3], s32 offset:244
@@ -208920,22 +209032,13 @@ define inreg <64 x i16> @bitcast_v128i8_to_v64i16_scalar(<128 x i8> inreg %a, i3
; SI-NEXT: buffer_load_dword v59, off, s[0:3], s32 offset:236
; SI-NEXT: buffer_load_dword v35, off, s[0:3], s32 offset:232
; SI-NEXT: buffer_load_dword v36, off, s[0:3], s32 offset:228
-; SI-NEXT: buffer_load_dword v37, off, s[0:3], s32 offset:224
+; SI-NEXT: buffer_load_dword v38, off, s[0:3], s32 offset:224
; SI-NEXT: buffer_load_dword v39, off, s[0:3], s32 offset:212
; SI-NEXT: buffer_load_dword v40, off, s[0:3], s32 offset:208
-; SI-NEXT: buffer_load_dword v53, off, s[0:3], s32 offset:204
+; SI-NEXT: buffer_load_dword v54, off, s[0:3], s32 offset:204
; SI-NEXT: buffer_load_dword v51, off, s[0:3], s32 offset:200
-; SI-NEXT: buffer_load_dword v48, off, s[0:3], s32 offset:196
-; SI-NEXT: ; implicit-def: $vgpr62 : SGPR spill to VGPR lane
+; SI-NEXT: buffer_load_dword v49, off, s[0:3], s32 offset:196
; SI-NEXT: v_writelane_b32 v63, s30, 0
-; SI-NEXT: v_writelane_b32 v62, s28, 0
-; SI-NEXT: v_writelane_b32 v62, s25, 1
-; SI-NEXT: v_writelane_b32 v62, s24, 2
-; SI-NEXT: v_writelane_b32 v62, s23, 3
-; SI-NEXT: v_writelane_b32 v62, s22, 4
-; SI-NEXT: v_writelane_b32 v62, s21, 5
-; SI-NEXT: v_writelane_b32 v62, s18, 6
-; SI-NEXT: v_writelane_b32 v62, s16, 7
; SI-NEXT: v_writelane_b32 v63, s31, 1
; SI-NEXT: v_writelane_b32 v63, s34, 2
; SI-NEXT: v_writelane_b32 v63, s35, 3
@@ -208948,17 +209051,29 @@ define inreg <64 x i16> @bitcast_v128i8_to_v64i16_scalar(<128 x i8> inreg %a, i3
; SI-NEXT: v_writelane_b32 v63, s50, 10
; SI-NEXT: v_writelane_b32 v63, s51, 11
; SI-NEXT: v_writelane_b32 v63, s52, 12
+; SI-NEXT: ; implicit-def: $vgpr62 : SGPR spill to VGPR lane
; SI-NEXT: v_writelane_b32 v63, s53, 13
+; SI-NEXT: v_writelane_b32 v62, s28, 0
; SI-NEXT: v_writelane_b32 v63, s54, 14
+; SI-NEXT: v_writelane_b32 v62, s27, 1
; SI-NEXT: v_writelane_b32 v63, s55, 15
+; SI-NEXT: v_writelane_b32 v62, s26, 2
; SI-NEXT: v_writelane_b32 v63, s64, 16
+; SI-NEXT: v_writelane_b32 v62, s25, 3
; SI-NEXT: v_writelane_b32 v63, s65, 17
+; SI-NEXT: v_writelane_b32 v62, s24, 4
; SI-NEXT: v_writelane_b32 v63, s66, 18
+; SI-NEXT: v_writelane_b32 v62, s23, 5
; SI-NEXT: v_writelane_b32 v63, s67, 19
+; SI-NEXT: v_writelane_b32 v62, s22, 6
; SI-NEXT: v_writelane_b32 v63, s68, 20
+; SI-NEXT: v_writelane_b32 v62, s21, 7
; SI-NEXT: v_writelane_b32 v63, s69, 21
+; SI-NEXT: v_writelane_b32 v62, s20, 8
; SI-NEXT: v_writelane_b32 v63, s70, 22
+; SI-NEXT: v_writelane_b32 v62, s18, 9
; SI-NEXT: v_writelane_b32 v63, s71, 23
+; SI-NEXT: v_writelane_b32 v62, s16, 10
; SI-NEXT: v_writelane_b32 v63, s80, 24
; SI-NEXT: v_writelane_b32 v63, s81, 25
; SI-NEXT: v_writelane_b32 v63, s82, 26
@@ -208970,254 +209085,252 @@ define inreg <64 x i16> @bitcast_v128i8_to_v64i16_scalar(<128 x i8> inreg %a, i3
; SI-NEXT: v_writelane_b32 v63, s96, 32
; SI-NEXT: v_writelane_b32 v63, s97, 33
; SI-NEXT: v_writelane_b32 v63, s98, 34
+; SI-NEXT: s_waitcnt expcnt(0)
+; SI-NEXT: v_mov_b32_e32 v21, v5
; SI-NEXT: v_writelane_b32 v63, s99, 35
-; SI-NEXT: s_waitcnt expcnt(1)
-; SI-NEXT: v_mov_b32_e32 v29, v26
; SI-NEXT: v_readfirstlane_b32 s15, v16
-; SI-NEXT: v_readfirstlane_b32 s18, v25
+; SI-NEXT: v_readfirstlane_b32 s21, v25
; SI-NEXT: v_readfirstlane_b32 s43, v15
; SI-NEXT: v_readfirstlane_b32 s42, v24
; SI-NEXT: v_readfirstlane_b32 s44, v23
-; SI-NEXT: v_readfirstlane_b32 s49, v12
-; SI-NEXT: v_readfirstlane_b32 s8, v11
-; SI-NEXT: v_readfirstlane_b32 s53, v20
; SI-NEXT: s_waitcnt vmcnt(14)
-; SI-NEXT: v_readfirstlane_b32 s4, v34
-; SI-NEXT: v_writelane_b32 v62, s4, 8
-; SI-NEXT: v_readfirstlane_b32 s4, v38
-; SI-NEXT: v_writelane_b32 v62, s4, 9
-; SI-NEXT: v_readfirstlane_b32 s4, v49
-; SI-NEXT: v_writelane_b32 v62, s4, 10
-; SI-NEXT: v_readfirstlane_b32 s4, v50
+; SI-NEXT: v_readfirstlane_b32 s4, v33
; SI-NEXT: v_writelane_b32 v62, s4, 11
-; SI-NEXT: v_readfirstlane_b32 s79, v52
-; SI-NEXT: v_readfirstlane_b32 s88, v54
-; SI-NEXT: v_readfirstlane_b32 s4, v55
-; SI-NEXT: buffer_load_dword v38, off, s[0:3], s32 offset:192
-; SI-NEXT: buffer_load_dword v49, off, s[0:3], s32 offset:180
-; SI-NEXT: buffer_load_dword v54, off, s[0:3], s32 offset:176
-; SI-NEXT: buffer_load_dword v44, off, s[0:3], s32 offset:172
-; SI-NEXT: buffer_load_dword v47, off, s[0:3], s32 offset:168
+; SI-NEXT: v_readfirstlane_b32 s4, v34
+; SI-NEXT: v_writelane_b32 v62, s4, 12
+; SI-NEXT: v_readfirstlane_b32 s52, v37
+; SI-NEXT: v_readfirstlane_b32 s82, v48
+; SI-NEXT: v_readfirstlane_b32 s4, v53
+; SI-NEXT: v_readfirstlane_b32 s79, v50
+; SI-NEXT: v_readfirstlane_b32 s88, v52
+; SI-NEXT: v_writelane_b32 v62, s4, 13
+; SI-NEXT: v_readfirstlane_b32 s77, v55
+; SI-NEXT: v_readfirstlane_b32 s4, v41
+; SI-NEXT: v_readfirstlane_b32 s35, v42
+; SI-NEXT: buffer_load_dword v37, off, s[0:3], s32 offset:192
+; SI-NEXT: buffer_load_dword v50, off, s[0:3], s32 offset:180
+; SI-NEXT: buffer_load_dword v53, off, s[0:3], s32 offset:176
+; SI-NEXT: buffer_load_dword v43, off, s[0:3], s32 offset:172
+; SI-NEXT: buffer_load_dword v46, off, s[0:3], s32 offset:168
; SI-NEXT: buffer_load_dword v60, off, s[0:3], s32 offset:164
; SI-NEXT: buffer_load_dword v61, off, s[0:3], s32 offset:160
-; SI-NEXT: buffer_load_dword v34, off, s[0:3], s32 offset:148
-; SI-NEXT: buffer_load_dword v50, off, s[0:3], s32 offset:144
-; SI-NEXT: buffer_load_dword v52, off, s[0:3], s32 offset:140
-; SI-NEXT: buffer_load_dword v55, off, s[0:3], s32 offset:136
-; SI-NEXT: v_writelane_b32 v62, s4, 12
-; SI-NEXT: v_readfirstlane_b32 s77, v41
-; SI-NEXT: v_readfirstlane_b32 s4, v42
-; SI-NEXT: v_readfirstlane_b32 s94, v31
-; SI-NEXT: v_readfirstlane_b32 s70, v32
-; SI-NEXT: v_readfirstlane_b32 s51, v33
+; SI-NEXT: buffer_load_dword v33, off, s[0:3], s32 offset:148
+; SI-NEXT: buffer_load_dword v34, off, s[0:3], s32 offset:144
+; SI-NEXT: buffer_load_dword v48, off, s[0:3], s32 offset:140
+; SI-NEXT: buffer_load_dword v52, off, s[0:3], s32 offset:136
+; SI-NEXT: v_readfirstlane_b32 s16, v31
+; SI-NEXT: v_readfirstlane_b32 s26, v32
; SI-NEXT: s_waitcnt vmcnt(14)
-; SI-NEXT: v_readfirstlane_b32 s37, v45
-; SI-NEXT: v_readfirstlane_b32 s24, v56
+; SI-NEXT: v_readfirstlane_b32 s76, v45
+; SI-NEXT: v_readfirstlane_b32 s66, v56
; SI-NEXT: v_readfirstlane_b32 s7, v57
; SI-NEXT: v_readfirstlane_b32 s92, v58
-; SI-NEXT: v_readfirstlane_b32 s28, v59
+; SI-NEXT: v_readfirstlane_b32 s27, v59
; SI-NEXT: buffer_load_dword v31, off, s[0:3], s32 offset:132
; SI-NEXT: buffer_load_dword v32, off, s[0:3], s32 offset:128
-; SI-NEXT: buffer_load_dword v33, off, s[0:3], s32 offset:116
-; SI-NEXT: buffer_load_dword v56, off, s[0:3], s32 offset:112
-; SI-NEXT: buffer_load_dword v57, off, s[0:3], s32 offset:108
-; SI-NEXT: buffer_load_dword v58, off, s[0:3], s32 offset:104
-; SI-NEXT: buffer_load_dword v59, off, s[0:3], s32 offset:100
+; SI-NEXT: buffer_load_dword v56, off, s[0:3], s32 offset:116
+; SI-NEXT: buffer_load_dword v57, off, s[0:3], s32 offset:112
+; SI-NEXT: buffer_load_dword v58, off, s[0:3], s32 offset:108
+; SI-NEXT: buffer_load_dword v59, off, s[0:3], s32 offset:104
+; SI-NEXT: buffer_load_dword v55, off, s[0:3], s32 offset:100
; SI-NEXT: buffer_load_dword v41, off, s[0:3], s32 offset:96
; SI-NEXT: buffer_load_dword v42, off, s[0:3], s32 offset:84
; SI-NEXT: buffer_load_dword v45, off, s[0:3], s32 offset:80
-; SI-NEXT: v_readfirstlane_b32 s35, v43
-; SI-NEXT: v_readfirstlane_b32 s55, v46
-; SI-NEXT: v_readfirstlane_b32 s68, v35
-; SI-NEXT: v_readfirstlane_b32 s87, v37
+; SI-NEXT: v_readfirstlane_b32 s51, v44
+; SI-NEXT: v_readfirstlane_b32 s55, v47
+; SI-NEXT: v_readfirstlane_b32 s6, v35
+; SI-NEXT: v_readfirstlane_b32 s98, v36
+; SI-NEXT: v_readfirstlane_b32 s18, v38
; SI-NEXT: v_readfirstlane_b32 s67, v39
; SI-NEXT: buffer_load_dword v35, off, s[0:3], s32 offset:76
+; SI-NEXT: buffer_load_dword v5, off, s[0:3], s32 offset:316
; SI-NEXT: s_waitcnt vmcnt(14)
-; SI-NEXT: v_readfirstlane_b32 s74, v53
-; SI-NEXT: buffer_load_dword v37, off, s[0:3], s32 offset:68
-; SI-NEXT: buffer_load_dword v39, off, s[0:3], s32 offset:64
-; SI-NEXT: v_readfirstlane_b32 s85, v48
-; SI-NEXT: buffer_load_dword v48, off, s[0:3], s32 offset:52
-; SI-NEXT: buffer_load_dword v53, off, s[0:3], s32 offset:48
-; SI-NEXT: buffer_load_dword v43, off, s[0:3], s32 offset:44
-; SI-NEXT: buffer_load_dword v46, off, s[0:3], s32 offset:40
-; SI-NEXT: v_writelane_b32 v62, s4, 13
-; SI-NEXT: v_readfirstlane_b32 s98, v40
+; SI-NEXT: v_readfirstlane_b32 s34, v54
; SI-NEXT: v_readfirstlane_b32 s69, v51
-; SI-NEXT: v_readfirstlane_b32 s21, v36
-; SI-NEXT: v_readfirstlane_b32 s40, v19
-; SI-NEXT: v_readfirstlane_b32 s23, v28
-; SI-NEXT: v_readfirstlane_b32 s34, v27
-; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_mov_b32_e32 v21, v13
-; SI-NEXT: v_mov_b32_e32 v13, v5
-; SI-NEXT: v_readfirstlane_b32 s97, v29
+; SI-NEXT: buffer_load_dword v38, off, s[0:3], s32 offset:68
+; SI-NEXT: buffer_load_dword v39, off, s[0:3], s32 offset:64
+; SI-NEXT: buffer_load_dword v51, off, s[0:3], s32 offset:52
+; SI-NEXT: buffer_load_dword v54, off, s[0:3], s32 offset:48
+; SI-NEXT: buffer_load_dword v44, off, s[0:3], s32 offset:44
+; SI-NEXT: buffer_load_dword v47, off, s[0:3], s32 offset:40
+; SI-NEXT: v_readfirstlane_b32 s87, v40
+; SI-NEXT: v_readfirstlane_b32 s86, v49
+; SI-NEXT: v_writelane_b32 v62, s4, 14
+; SI-NEXT: v_writelane_b32 v62, s17, 15
+; SI-NEXT: v_writelane_b32 v62, s15, 16
+; SI-NEXT: v_writelane_b32 v62, s21, 17
+; SI-NEXT: v_writelane_b32 v62, s43, 18
+; SI-NEXT: v_writelane_b32 v62, s42, 19
+; SI-NEXT: v_writelane_b32 v62, s44, 20
+; SI-NEXT: v_readfirstlane_b32 s53, v12
+; SI-NEXT: v_readfirstlane_b32 s23, v11
+; SI-NEXT: v_readfirstlane_b32 s8, v20
+; SI-NEXT: v_readfirstlane_b32 s48, v19
+; SI-NEXT: v_readfirstlane_b32 s63, v28
+; SI-NEXT: v_readfirstlane_b32 s95, v27
+; SI-NEXT: v_mov_b32_e32 v29, v13
+; SI-NEXT: v_readfirstlane_b32 s97, v26
; SI-NEXT: v_readfirstlane_b32 s80, v18
-; SI-NEXT: v_lshlrev_b32_e32 v6, 24, v6
-; SI-NEXT: v_lshlrev_b32_e32 v14, 24, v14
; SI-NEXT: v_lshlrev_b32_e32 v18, 24, v22
-; SI-NEXT: v_lshlrev_b32_e32 v19, 24, v30
; SI-NEXT: v_readfirstlane_b32 s96, v17
+; SI-NEXT: v_readfirstlane_b32 s65, v10
; SI-NEXT: v_readfirstlane_b32 s64, v9
-; SI-NEXT: v_readfirstlane_b32 s25, v8
+; SI-NEXT: v_readfirstlane_b32 s68, v8
; SI-NEXT: v_readfirstlane_b32 s83, v7
; SI-NEXT: v_readfirstlane_b32 s84, v4
; SI-NEXT: v_readfirstlane_b32 s93, v3
-; SI-NEXT: v_readfirstlane_b32 s76, v1
-; SI-NEXT: v_readfirstlane_b32 s58, v38
-; SI-NEXT: v_readfirstlane_b32 s65, v49
-; SI-NEXT: v_readfirstlane_b32 s62, v54
-; SI-NEXT: v_readfirstlane_b32 s81, v44
-; SI-NEXT: v_readfirstlane_b32 s71, v47
-; SI-NEXT: v_readfirstlane_b32 s38, v60
-; SI-NEXT: v_readfirstlane_b32 s86, v61
-; SI-NEXT: buffer_load_dword v60, off, s[0:3], s32 offset:156
-; SI-NEXT: buffer_load_dword v61, off, s[0:3], s32 offset:220
+; SI-NEXT: v_readfirstlane_b32 s90, v2
+; SI-NEXT: v_readfirstlane_b32 s11, v1
+; SI-NEXT: v_readfirstlane_b32 s59, v37
+; SI-NEXT: v_readfirstlane_b32 s94, v50
+; SI-NEXT: v_readfirstlane_b32 s39, v53
+; SI-NEXT: v_readfirstlane_b32 s81, v43
+; SI-NEXT: v_readfirstlane_b32 s71, v46
+; SI-NEXT: v_readfirstlane_b32 s85, v60
+; SI-NEXT: v_readfirstlane_b32 s89, v61
+; SI-NEXT: v_readfirstlane_b32 s49, v33
+; SI-NEXT: v_readfirstlane_b32 s70, v34
; SI-NEXT: s_waitcnt vmcnt(14)
-; SI-NEXT: v_readfirstlane_b32 s90, v50
-; SI-NEXT: v_readfirstlane_b32 s31, v52
-; SI-NEXT: v_readfirstlane_b32 s4, v55
+; SI-NEXT: v_readfirstlane_b32 s74, v31
+; SI-NEXT: buffer_load_dword v31, off, s[0:3], s32 offset:220
+; SI-NEXT: v_readfirstlane_b32 s91, v52
; SI-NEXT: buffer_load_dword v49, off, s[0:3], s32 offset:36
-; SI-NEXT: buffer_load_dword v51, off, s[0:3], s32 offset:32
-; SI-NEXT: buffer_load_dword v54, off, s[0:3], s32 offset:20
+; SI-NEXT: buffer_load_dword v50, off, s[0:3], s32 offset:32
+; SI-NEXT: buffer_load_dword v53, off, s[0:3], s32 offset:20
; SI-NEXT: buffer_load_dword v40, off, s[0:3], s32 offset:16
-; SI-NEXT: buffer_load_dword v44, off, s[0:3], s32 offset:12
-; SI-NEXT: buffer_load_dword v50, off, s[0:3], s32 offset:336
-; SI-NEXT: buffer_load_dword v52, off, s[0:3], s32 offset:8
-; SI-NEXT: buffer_load_dword v55, off, s[0:3], s32 offset:4
-; SI-NEXT: buffer_load_dword v47, off, s[0:3], s32
-; SI-NEXT: v_readfirstlane_b32 s72, v31
-; SI-NEXT: buffer_load_dword v31, off, s[0:3], s32 offset:316
-; SI-NEXT: buffer_load_dword v36, off, s[0:3], s32 offset:72
-; SI-NEXT: v_readfirstlane_b32 s82, v56
-; SI-NEXT: v_readfirstlane_b32 s95, v57
+; SI-NEXT: buffer_load_dword v43, off, s[0:3], s32 offset:12
+; SI-NEXT: buffer_load_dword v52, off, s[0:3], s32 offset:336
+; SI-NEXT: buffer_load_dword v46, off, s[0:3], s32 offset:8
+; SI-NEXT: buffer_load_dword v60, off, s[0:3], s32 offset:4
+; SI-NEXT: buffer_load_dword v61, off, s[0:3], s32
+; SI-NEXT: v_readfirstlane_b32 s37, v56
; SI-NEXT: s_waitcnt vmcnt(14)
-; SI-NEXT: v_readfirstlane_b32 s39, v58
-; SI-NEXT: v_readfirstlane_b32 s56, v59
-; SI-NEXT: v_readfirstlane_b32 s57, v41
+; SI-NEXT: v_readfirstlane_b32 s38, v59
+; SI-NEXT: buffer_load_dword v59, off, s[0:3], s32 offset:156
+; SI-NEXT: buffer_load_dword v36, off, s[0:3], s32 offset:72
+; SI-NEXT: v_readfirstlane_b32 s25, v57
+; SI-NEXT: v_readfirstlane_b32 s56, v58
+; SI-NEXT: v_readfirstlane_b32 s57, v55
+; SI-NEXT: v_readfirstlane_b32 s58, v41
; SI-NEXT: v_readfirstlane_b32 s36, v42
-; SI-NEXT: v_readfirstlane_b32 s73, v45
-; SI-NEXT: buffer_load_dword v41, off, s[0:3], s32 offset:284
-; SI-NEXT: buffer_load_dword v42, off, s[0:3], s32 offset:252
-; SI-NEXT: buffer_load_dword v45, off, s[0:3], s32 offset:188
-; SI-NEXT: buffer_load_dword v56, off, s[0:3], s32 offset:124
-; SI-NEXT: buffer_load_dword v57, off, s[0:3], s32 offset:28
-; SI-NEXT: buffer_load_dword v58, off, s[0:3], s32 offset:60
-; SI-NEXT: buffer_load_dword v59, off, s[0:3], s32 offset:92
-; SI-NEXT: v_readfirstlane_b32 s16, v34
-; SI-NEXT: v_readfirstlane_b32 s48, v32
-; SI-NEXT: v_readfirstlane_b32 s52, v33
-; SI-NEXT: v_writelane_b32 v62, s4, 14
+; SI-NEXT: v_readfirstlane_b32 s40, v45
+; SI-NEXT: buffer_load_dword v55, off, s[0:3], s32 offset:284
+; SI-NEXT: buffer_load_dword v41, off, s[0:3], s32 offset:252
+; SI-NEXT: buffer_load_dword v42, off, s[0:3], s32 offset:188
+; SI-NEXT: buffer_load_dword v45, off, s[0:3], s32 offset:124
+; SI-NEXT: buffer_load_dword v56, off, s[0:3], s32 offset:28
+; SI-NEXT: buffer_load_dword v57, off, s[0:3], s32 offset:60
+; SI-NEXT: buffer_load_dword v58, off, s[0:3], s32 offset:92
+; SI-NEXT: v_readfirstlane_b32 s75, v32
; SI-NEXT: v_readfirstlane_b32 s47, v35
-; SI-NEXT: v_readfirstlane_b32 s60, v37
-; SI-NEXT: v_readfirstlane_b32 s61, v39
-; SI-NEXT: v_readfirstlane_b32 s89, v43
+; SI-NEXT: v_writelane_b32 v62, s56, 21
+; SI-NEXT: v_writelane_b32 v62, s49, 22
+; SI-NEXT: v_readfirstlane_b32 s72, v38
+; SI-NEXT: v_readfirstlane_b32 s73, v39
; SI-NEXT: s_waitcnt vmcnt(14)
-; SI-NEXT: v_readfirstlane_b32 s99, v46
-; SI-NEXT: buffer_load_dword v34, off, s[0:3], s32 offset:312
+; SI-NEXT: v_readfirstlane_b32 s22, v44
+; SI-NEXT: v_readfirstlane_b32 s99, v47
+; SI-NEXT: v_writelane_b32 v62, s53, 23
+; SI-NEXT: v_writelane_b32 v62, s70, 24
+; SI-NEXT: v_writelane_b32 v62, s23, 25
+; SI-NEXT: v_writelane_b32 v62, s57, 26
+; SI-NEXT: v_readfirstlane_b32 s54, v51
+; SI-NEXT: v_readfirstlane_b32 s50, v54
+; SI-NEXT: v_readfirstlane_b32 s31, v48
+; SI-NEXT: v_readfirstlane_b32 s78, v49
+; SI-NEXT: v_readfirstlane_b32 s30, v50
+; SI-NEXT: v_readfirstlane_b32 s24, v53
+; SI-NEXT: v_readfirstlane_b32 s28, v40
+; SI-NEXT: s_waitcnt vmcnt(13)
+; SI-NEXT: v_readfirstlane_b32 s20, v43
+; SI-NEXT: s_waitcnt vmcnt(12)
+; SI-NEXT: v_cmp_ne_u32_e32 vcc, 0, v52
+; SI-NEXT: s_waitcnt vmcnt(11)
+; SI-NEXT: v_readfirstlane_b32 s45, v46
+; SI-NEXT: v_writelane_b32 v62, s45, 27
+; SI-NEXT: v_writelane_b32 v62, s8, 28
+; SI-NEXT: v_writelane_b32 v62, s58, 29
+; SI-NEXT: v_writelane_b32 v62, s59, 30
+; SI-NEXT: v_writelane_b32 v62, s47, 31
+; SI-NEXT: s_waitcnt vmcnt(7)
+; SI-NEXT: v_readfirstlane_b32 s60, v36
+; SI-NEXT: buffer_load_dword v37, off, s[0:3], s32 offset:312
; SI-NEXT: buffer_load_dword v39, off, s[0:3], s32 offset:280
; SI-NEXT: buffer_load_dword v33, off, s[0:3], s32 offset:248
; SI-NEXT: buffer_load_dword v38, off, s[0:3], s32 offset:216
-; SI-NEXT: buffer_load_dword v37, off, s[0:3], s32 offset:184
-; SI-NEXT: buffer_load_dword v35, off, s[0:3], s32 offset:152
-; SI-NEXT: buffer_load_dword v26, off, s[0:3], s32 offset:120
-; SI-NEXT: buffer_load_dword v46, off, s[0:3], s32 offset:88
-; SI-NEXT: buffer_load_dword v32, off, s[0:3], s32 offset:56
-; SI-NEXT: buffer_load_dword v43, off, s[0:3], s32 offset:24
-; SI-NEXT: v_readfirstlane_b32 s54, v48
-; SI-NEXT: v_readfirstlane_b32 s50, v53
-; SI-NEXT: v_readfirstlane_b32 s78, v49
-; SI-NEXT: v_readfirstlane_b32 s30, v51
-; SI-NEXT: v_readfirstlane_b32 s66, v54
-; SI-NEXT: v_readfirstlane_b32 s91, v40
-; SI-NEXT: s_waitcnt vmcnt(14)
-; SI-NEXT: v_readfirstlane_b32 s6, v44
-; SI-NEXT: v_cmp_ne_u32_e32 vcc, 0, v50
-; SI-NEXT: s_and_b64 s[4:5], vcc, exec
-; SI-NEXT: v_readfirstlane_b32 s4, v10
-; SI-NEXT: v_writelane_b32 v62, s4, 15
-; SI-NEXT: v_readfirstlane_b32 s4, v2
-; SI-NEXT: v_writelane_b32 v62, s4, 16
-; SI-NEXT: v_writelane_b32 v62, s17, 17
-; SI-NEXT: v_writelane_b32 v62, s15, 18
-; SI-NEXT: v_writelane_b32 v62, s18, 19
-; SI-NEXT: v_writelane_b32 v62, s43, 20
-; SI-NEXT: v_writelane_b32 v62, s42, 21
-; SI-NEXT: v_writelane_b32 v62, s44, 22
-; SI-NEXT: v_writelane_b32 v62, s16, 23
-; SI-NEXT: v_writelane_b32 v62, s49, 24
-; SI-NEXT: v_writelane_b32 v62, s8, 25
-; SI-NEXT: v_writelane_b32 v62, s6, 26
-; SI-NEXT: v_readfirstlane_b32 s45, v52
-; SI-NEXT: v_writelane_b32 v62, s56, 27
-; SI-NEXT: v_writelane_b32 v62, s45, 28
-; SI-NEXT: v_writelane_b32 v62, s53, 29
-; SI-NEXT: v_writelane_b32 v62, s94, 30
-; SI-NEXT: v_writelane_b32 v62, s57, 31
-; SI-NEXT: v_writelane_b32 v62, s58, 32
-; SI-NEXT: v_writelane_b32 v62, s47, 33
-; SI-NEXT: v_readfirstlane_b32 s46, v55
-; SI-NEXT: v_writelane_b32 v62, s40, 34
-; SI-NEXT: v_readfirstlane_b32 s59, v47
+; SI-NEXT: buffer_load_dword v32, off, s[0:3], s32 offset:184
+; SI-NEXT: buffer_load_dword v36, off, s[0:3], s32 offset:152
+; SI-NEXT: buffer_load_dword v35, off, s[0:3], s32 offset:120
+; SI-NEXT: buffer_load_dword v34, off, s[0:3], s32 offset:88
+; SI-NEXT: buffer_load_dword v47, off, s[0:3], s32 offset:56
+; SI-NEXT: buffer_load_dword v44, off, s[0:3], s32 offset:24
+; SI-NEXT: v_writelane_b32 v62, s48, 32
+; SI-NEXT: v_writelane_b32 v62, s26, 33
+; SI-NEXT: v_readfirstlane_b32 s46, v60
+; SI-NEXT: v_writelane_b32 v62, s60, 34
+; SI-NEXT: v_readfirstlane_b32 s61, v61
; SI-NEXT: v_writelane_b32 v62, s46, 35
-; SI-NEXT: v_writelane_b32 v62, s59, 36
-; SI-NEXT: v_writelane_b32 v62, s60, 37
+; SI-NEXT: v_writelane_b32 v62, s61, 36
+; SI-NEXT: v_writelane_b32 v62, s72, 37
; SI-NEXT: v_writelane_b32 v62, s36, 38
-; SI-NEXT: v_writelane_b32 v62, s65, 39
-; SI-NEXT: v_writelane_b32 v62, s61, 40
-; SI-NEXT: v_writelane_b32 v62, s73, 41
-; SI-NEXT: v_writelane_b32 v62, s62, 42
-; SI-NEXT: v_writelane_b32 v62, s72, 43
-; SI-NEXT: v_writelane_b32 v62, s23, 44
-; SI-NEXT: v_writelane_b32 v62, s48, 45
-; SI-NEXT: v_writelane_b32 v62, s34, 46
+; SI-NEXT: v_writelane_b32 v62, s94, 39
+; SI-NEXT: v_writelane_b32 v62, s73, 40
+; SI-NEXT: v_writelane_b32 v62, s40, 41
+; SI-NEXT: v_writelane_b32 v62, s39, 42
+; SI-NEXT: v_writelane_b32 v62, s74, 43
+; SI-NEXT: v_writelane_b32 v62, s63, 44
+; SI-NEXT: v_writelane_b32 v62, s75, 45
+; SI-NEXT: v_writelane_b32 v62, s95, 46
; SI-NEXT: v_writelane_b32 v62, s78, 47
; SI-NEXT: v_writelane_b32 v62, s30, 48
; SI-NEXT: v_writelane_b32 v62, s54, 49
; SI-NEXT: v_writelane_b32 v62, s50, 50
-; SI-NEXT: v_writelane_b32 v62, s52, 51
-; SI-NEXT: v_writelane_b32 v62, s82, 52
-; SI-NEXT: v_writelane_b32 v62, s66, 53
-; SI-NEXT: v_readfirstlane_b32 s22, v36
+; SI-NEXT: v_writelane_b32 v62, s25, 51
+; SI-NEXT: v_writelane_b32 v62, s24, 52
+; SI-NEXT: s_and_b64 s[4:5], vcc, exec
+; SI-NEXT: v_lshlrev_b32_e32 v60, 24, v6
+; SI-NEXT: v_lshlrev_b32_e32 v61, 24, v14
+; SI-NEXT: v_lshlrev_b32_e32 v6, 24, v30
; SI-NEXT: s_waitcnt vmcnt(12)
-; SI-NEXT: v_lshlrev_b32_e32 v22, 24, v57
+; SI-NEXT: v_lshlrev_b32_e32 v20, 24, v56
; SI-NEXT: s_waitcnt vmcnt(11)
-; SI-NEXT: v_lshlrev_b32_e32 v23, 24, v58
+; SI-NEXT: v_lshlrev_b32_e32 v22, 24, v57
; SI-NEXT: s_waitcnt vmcnt(10)
-; SI-NEXT: v_lshlrev_b32_e32 v59, 24, v59
-; SI-NEXT: v_lshlrev_b32_e32 v24, 24, v56
-; SI-NEXT: v_lshlrev_b32_e32 v44, 24, v60
-; SI-NEXT: v_lshlrev_b32_e32 v45, 24, v45
-; SI-NEXT: v_lshlrev_b32_e32 v47, 24, v61
-; SI-NEXT: v_lshlrev_b32_e32 v56, 24, v42
-; SI-NEXT: v_lshlrev_b32_e32 v57, 24, v41
-; SI-NEXT: v_lshlrev_b32_e32 v58, 24, v31
-; SI-NEXT: v_writelane_b32 v62, s91, 54
+; SI-NEXT: v_lshlrev_b32_e32 v23, 24, v58
+; SI-NEXT: v_lshlrev_b32_e32 v14, 24, v45
+; SI-NEXT: v_lshlrev_b32_e32 v43, 24, v59
+; SI-NEXT: v_lshlrev_b32_e32 v45, 24, v42
+; SI-NEXT: v_lshlrev_b32_e32 v46, 24, v31
+; SI-NEXT: v_lshlrev_b32_e32 v56, 24, v41
+; SI-NEXT: v_lshlrev_b32_e32 v57, 24, v55
+; SI-NEXT: v_lshlrev_b32_e32 v58, 24, v5
+; SI-NEXT: v_writelane_b32 v62, s28, 53
; SI-NEXT: buffer_store_dword v6, off, s[0:3], s32 offset:404 ; 4-byte Folded Spill
-; SI-NEXT: buffer_store_dword v22, off, s[0:3], s32 offset:408 ; 4-byte Folded Spill
-; SI-NEXT: s_waitcnt vmcnt(2)
-; SI-NEXT: buffer_store_dword v43, off, s[0:3], s32 offset:412 ; 4-byte Folded Spill
-; SI-NEXT: buffer_store_dword v32, off, s[0:3], s32 offset:416 ; 4-byte Folded Spill
-; SI-NEXT: buffer_store_dword v46, off, s[0:3], s32 offset:420 ; 4-byte Folded Spill
-; SI-NEXT: buffer_store_dword v44, off, s[0:3], s32 offset:424 ; 4-byte Folded Spill
+; SI-NEXT: s_waitcnt vmcnt(1)
+; SI-NEXT: buffer_store_dword v44, off, s[0:3], s32 offset:408 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v47, off, s[0:3], s32 offset:412 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v14, off, s[0:3], s32 offset:416 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v43, off, s[0:3], s32 offset:420 ; 4-byte Folded Spill
; SI-NEXT: s_cbranch_scc0 .LBB97_4
; SI-NEXT: ; %bb.1: ; %cmp.false
-; SI-NEXT: buffer_load_dword v29, off, s[0:3], s32 offset:400 ; 4-byte Folded Reload
-; SI-NEXT: v_mov_b32_e32 v5, v13
-; SI-NEXT: v_and_b32_e32 v1, 0xff, v13
-; SI-NEXT: buffer_load_dword v13, off, s[0:3], s32 offset:396 ; 4-byte Folded Reload
-; SI-NEXT: v_readlane_b32 s5, v62, 5
-; SI-NEXT: s_and_b32 s4, s20, 0xff
+; SI-NEXT: buffer_load_dword v19, off, s[0:3], s32 offset:400 ; 4-byte Folded Reload
+; SI-NEXT: v_readlane_b32 s4, v62, 8
+; SI-NEXT: v_readlane_b32 s5, v62, 7
+; SI-NEXT: v_mov_b32_e32 v13, v21
+; SI-NEXT: v_and_b32_e32 v1, 0xff, v21
+; SI-NEXT: buffer_load_dword v21, off, s[0:3], s32 offset:396 ; 4-byte Folded Reload
+; SI-NEXT: s_and_b32 s4, s4, 0xff
; SI-NEXT: s_lshl_b32 s5, s5, 8
; SI-NEXT: s_or_b32 s4, s4, s5
-; SI-NEXT: v_writelane_b32 v62, s4, 55
-; SI-NEXT: v_readlane_b32 s4, v62, 4
+; SI-NEXT: v_writelane_b32 v62, s4, 54
+; SI-NEXT: v_readlane_b32 s4, v62, 6
; SI-NEXT: s_and_b32 s4, s4, 0xff
-; SI-NEXT: v_readlane_b32 s5, v62, 3
+; SI-NEXT: v_readlane_b32 s5, v62, 5
; SI-NEXT: s_lshl_b32 s4, s4, 16
; SI-NEXT: s_lshl_b32 s5, s5, 24
-; SI-NEXT: s_or_b32 s63, s5, s4
-; SI-NEXT: v_readlane_b32 s4, v62, 6
+; SI-NEXT: s_or_b32 s4, s5, s4
+; SI-NEXT: v_writelane_b32 v62, s4, 55
+; SI-NEXT: v_readlane_b32 s4, v62, 9
; SI-NEXT: s_and_b32 s5, s4, 0xff
; SI-NEXT: s_lshl_b32 s5, s5, 16
; SI-NEXT: s_lshl_b32 s9, s19, 24
@@ -209227,27 +209340,29 @@ define inreg <64 x i16> @bitcast_v128i8_to_v64i16_scalar(<128 x i8> inreg %a, i3
; SI-NEXT: s_lshl_b32 s10, s29, 8
; SI-NEXT: s_or_b32 s4, s5, s10
; SI-NEXT: v_writelane_b32 v62, s4, 56
-; SI-NEXT: s_and_b32 s5, s76, 0xff
-; SI-NEXT: v_readlane_b32 s10, v62, 16
+; SI-NEXT: v_writelane_b32 v62, s37, 57
+; SI-NEXT: s_and_b32 s5, s11, 0xff
+; SI-NEXT: s_mov_b32 s37, s11
; SI-NEXT: s_lshl_b32 s5, s5, 16
-; SI-NEXT: s_lshl_b32 s11, s10, 24
+; SI-NEXT: s_lshl_b32 s11, s90, 24
+; SI-NEXT: v_readlane_b32 s4, v62, 2
; SI-NEXT: s_or_b32 s5, s11, s5
-; SI-NEXT: s_and_b32 s11, s26, 0xff
+; SI-NEXT: s_and_b32 s11, s4, 0xff
+; SI-NEXT: v_readlane_b32 s4, v62, 1
; SI-NEXT: s_lshl_b32 s11, s11, 16
-; SI-NEXT: s_lshl_b32 s12, s27, 24
+; SI-NEXT: s_lshl_b32 s12, s4, 24
; SI-NEXT: s_or_b32 s14, s12, s11
; SI-NEXT: s_and_b32 s11, s83, 0xff
-; SI-NEXT: s_lshl_b32 s12, s25, 8
-; SI-NEXT: s_or_b32 s10, s11, s12
-; SI-NEXT: v_writelane_b32 v62, s10, 57
+; SI-NEXT: s_lshl_b32 s12, s68, 8
+; SI-NEXT: s_or_b32 s4, s11, s12
; SI-NEXT: s_and_b32 s11, s64, 0xff
-; SI-NEXT: v_readlane_b32 s10, v62, 15
; SI-NEXT: s_lshl_b32 s11, s11, 16
-; SI-NEXT: s_lshl_b32 s13, s10, 24
+; SI-NEXT: s_lshl_b32 s13, s65, 24
; SI-NEXT: s_or_b32 s41, s13, s11
; SI-NEXT: s_and_b32 s11, s43, 0xff
; SI-NEXT: s_lshl_b32 s13, s15, 8
-; SI-NEXT: s_or_b32 s10, s11, s13
+; SI-NEXT: v_writelane_b32 v62, s4, 58
+; SI-NEXT: s_or_b32 s4, s11, s13
; SI-NEXT: s_and_b32 s11, s96, 0xff
; SI-NEXT: s_lshl_b32 s11, s11, 16
; SI-NEXT: s_lshl_b32 s15, s80, 24
@@ -209255,266 +209370,269 @@ define inreg <64 x i16> @bitcast_v128i8_to_v64i16_scalar(<128 x i8> inreg %a, i3
; SI-NEXT: s_and_b32 s11, s44, 0xff
; SI-NEXT: s_lshl_b32 s15, s42, 8
; SI-NEXT: s_or_b32 s13, s11, s15
-; SI-NEXT: s_and_b32 s11, s18, 0xff
+; SI-NEXT: s_and_b32 s11, s21, 0xff
; SI-NEXT: s_lshl_b32 s11, s11, 16
; SI-NEXT: s_lshl_b32 s15, s97, 24
; SI-NEXT: s_or_b32 s44, s15, s11
-; SI-NEXT: s_and_b32 s11, s59, 0xff
+; SI-NEXT: s_and_b32 s11, s61, 0xff
; SI-NEXT: s_lshl_b32 s15, s46, 8
; SI-NEXT: s_or_b32 s12, s11, s15
; SI-NEXT: s_and_b32 s11, s45, 0xff
; SI-NEXT: s_lshl_b32 s11, s11, 16
-; SI-NEXT: s_lshl_b32 s15, s6, 24
+; SI-NEXT: s_lshl_b32 s15, s20, 24
; SI-NEXT: s_or_b32 s45, s15, s11
; SI-NEXT: s_and_b32 s11, s30, 0xff
; SI-NEXT: s_lshl_b32 s15, s78, 8
-; SI-NEXT: v_writelane_b32 v62, s10, 58
; SI-NEXT: s_or_b32 s10, s11, s15
; SI-NEXT: s_and_b32 s11, s99, 0xff
; SI-NEXT: s_lshl_b32 s11, s11, 16
-; SI-NEXT: s_lshl_b32 s15, s89, 24
+; SI-NEXT: s_lshl_b32 s15, s22, 24
; SI-NEXT: s_or_b32 s46, s15, s11
-; SI-NEXT: s_and_b32 s11, s61, 0xff
-; SI-NEXT: s_lshl_b32 s15, s60, 8
-; SI-NEXT: s_or_b32 s6, s11, s15
-; SI-NEXT: s_and_b32 s11, s22, 0xff
+; SI-NEXT: s_and_b32 s11, s73, 0xff
+; SI-NEXT: s_lshl_b32 s15, s72, 8
+; SI-NEXT: v_writelane_b32 v62, s4, 59
+; SI-NEXT: s_or_b32 s4, s11, s15
+; SI-NEXT: s_and_b32 s11, s60, 0xff
; SI-NEXT: s_lshl_b32 s11, s11, 16
; SI-NEXT: s_lshl_b32 s15, s47, 24
; SI-NEXT: s_or_b32 s47, s15, s11
-; SI-NEXT: s_and_b32 s11, s57, 0xff
-; SI-NEXT: s_lshl_b32 s15, s56, 8
-; SI-NEXT: v_writelane_b32 v62, s6, 59
-; SI-NEXT: s_or_b32 s6, s11, s15
-; SI-NEXT: s_and_b32 s11, s39, 0xff
-; SI-NEXT: v_writelane_b32 v62, s6, 60
+; SI-NEXT: s_and_b32 s11, s58, 0xff
+; SI-NEXT: s_lshl_b32 s15, s57, 8
+; SI-NEXT: s_mov_b32 s62, s16
+; SI-NEXT: s_or_b32 s16, s11, s15
+; SI-NEXT: s_and_b32 s11, s38, 0xff
; SI-NEXT: s_lshl_b32 s11, s11, 16
-; SI-NEXT: s_lshl_b32 s15, s95, 24
+; SI-NEXT: s_lshl_b32 s15, s56, 24
; SI-NEXT: s_or_b32 s56, s15, s11
-; SI-NEXT: s_and_b32 s11, s48, 0xff
-; SI-NEXT: s_lshl_b32 s15, s72, 8
-; SI-NEXT: v_readlane_b32 s6, v62, 14
-; SI-NEXT: s_or_b32 s48, s11, s15
-; SI-NEXT: s_and_b32 s11, s6, 0xff
+; SI-NEXT: s_and_b32 s11, s75, 0xff
+; SI-NEXT: s_lshl_b32 s15, s74, 8
+; SI-NEXT: s_or_b32 s73, s11, s15
+; SI-NEXT: s_and_b32 s11, s91, 0xff
; SI-NEXT: s_lshl_b32 s11, s11, 16
; SI-NEXT: s_lshl_b32 s15, s31, 24
; SI-NEXT: s_or_b32 vcc_lo, s15, s11
-; SI-NEXT: s_and_b32 s11, s86, 0xff
-; SI-NEXT: s_lshl_b32 s15, s38, 8
+; SI-NEXT: s_and_b32 s11, s89, 0xff
+; SI-NEXT: s_lshl_b32 s15, s85, 8
; SI-NEXT: s_or_b32 s72, s11, s15
; SI-NEXT: s_and_b32 s11, s71, 0xff
; SI-NEXT: s_lshl_b32 s11, s11, 16
; SI-NEXT: s_lshl_b32 s15, s81, 24
; SI-NEXT: s_or_b32 vcc_hi, s15, s11
-; SI-NEXT: s_and_b32 s11, s58, 0xff
-; SI-NEXT: s_lshl_b32 s15, s85, 8
+; SI-NEXT: s_and_b32 s11, s59, 0xff
+; SI-NEXT: s_lshl_b32 s15, s86, 8
; SI-NEXT: s_or_b32 s57, s11, s15
; SI-NEXT: s_and_b32 s11, s69, 0xff
; SI-NEXT: s_lshl_b32 s11, s11, 16
-; SI-NEXT: s_lshl_b32 s15, s74, 24
-; SI-NEXT: v_writelane_b32 v62, s74, 61
+; SI-NEXT: s_lshl_b32 s15, s34, 24
; SI-NEXT: s_or_b32 s74, s15, s11
-; SI-NEXT: s_and_b32 s11, s87, 0xff
-; SI-NEXT: s_lshl_b32 s15, s21, 8
+; SI-NEXT: s_and_b32 s11, s18, 0xff
+; SI-NEXT: s_lshl_b32 s15, s98, 8
; SI-NEXT: s_or_b32 s58, s11, s15
-; SI-NEXT: s_and_b32 s11, s68, 0xff
+; SI-NEXT: s_and_b32 s11, s6, 0xff
+; SI-NEXT: v_writelane_b32 v62, s4, 60
; SI-NEXT: s_lshl_b32 s11, s11, 16
-; SI-NEXT: s_lshl_b32 s15, s28, 24
+; SI-NEXT: s_lshl_b32 s15, s27, 24
+; SI-NEXT: v_writelane_b32 v62, s34, 61
; SI-NEXT: s_or_b32 s75, s15, s11
-; SI-NEXT: s_and_b32 s11, s24, 0xff
+; SI-NEXT: s_and_b32 s11, s66, 0xff
; SI-NEXT: s_lshl_b32 s15, s55, 8
-; SI-NEXT: v_writelane_b32 v62, s25, 62
+; SI-NEXT: v_writelane_b32 v62, s6, 62
; SI-NEXT: s_or_b32 s59, s11, s15
-; SI-NEXT: s_and_b32 s11, s37, 0xff
+; SI-NEXT: s_and_b32 s11, s76, 0xff
; SI-NEXT: s_lshl_b32 s11, s11, 16
; SI-NEXT: s_lshl_b32 s15, s51, 24
-; SI-NEXT: v_readlane_b32 s4, v62, 13
-; SI-NEXT: s_mov_b32 s18, s21
-; SI-NEXT: s_mov_b32 s21, s97
-; SI-NEXT: s_mov_b32 s97, s37
-; SI-NEXT: s_mov_b32 s37, s76
+; SI-NEXT: v_writelane_b32 v62, s85, 63
+; SI-NEXT: s_mov_b32 s4, s97
+; SI-NEXT: s_mov_b32 s97, s76
; SI-NEXT: s_or_b32 s76, s15, s11
+; SI-NEXT: v_readlane_b32 s15, v62, 14
; SI-NEXT: s_and_b32 s11, s35, 0xff
-; SI-NEXT: s_lshl_b32 s15, s4, 8
+; SI-NEXT: s_lshl_b32 s15, s15, 8
; SI-NEXT: s_or_b32 s60, s11, s15
; SI-NEXT: s_and_b32 s11, s77, 0xff
-; SI-NEXT: v_readlane_b32 s4, v62, 12
+; SI-NEXT: v_readlane_b32 s15, v62, 13
; SI-NEXT: s_lshl_b32 s11, s11, 16
-; SI-NEXT: s_lshl_b32 s15, s4, 24
-; SI-NEXT: v_readlane_b32 s4, v62, 11
-; SI-NEXT: s_mov_b32 s6, s95
-; SI-NEXT: s_mov_b32 s95, s39
-; SI-NEXT: s_mov_b32 s39, s89
-; SI-NEXT: s_mov_b32 s89, s99
+; SI-NEXT: s_lshl_b32 s15, s15, 24
+; SI-NEXT: s_mov_b32 s21, s20
+; SI-NEXT: s_mov_b32 s20, s38
+; SI-NEXT: s_mov_b32 s38, s99
; SI-NEXT: s_mov_b32 s99, s83
; SI-NEXT: s_mov_b32 s83, s55
; SI-NEXT: s_mov_b32 s55, s64
; SI-NEXT: s_mov_b32 s64, s35
; SI-NEXT: s_mov_b32 s35, s77
; SI-NEXT: s_or_b32 s77, s15, s11
-; SI-NEXT: s_and_b32 s11, s4, 0xff
-; SI-NEXT: v_readlane_b32 s4, v62, 10
-; SI-NEXT: s_lshl_b32 s15, s4, 8
-; SI-NEXT: v_readlane_b32 s4, v62, 9
+; SI-NEXT: s_and_b32 s11, s82, 0xff
+; SI-NEXT: s_lshl_b32 s15, s52, 8
; SI-NEXT: s_or_b32 s61, s11, s15
-; SI-NEXT: s_and_b32 s11, s4, 0xff
-; SI-NEXT: v_readlane_b32 s4, v62, 8
+; SI-NEXT: v_readlane_b32 s11, v62, 12
+; SI-NEXT: s_and_b32 s11, s11, 0xff
+; SI-NEXT: v_readlane_b32 s15, v62, 11
; SI-NEXT: s_lshl_b32 s11, s11, 16
-; SI-NEXT: s_lshl_b32 s15, s4, 24
+; SI-NEXT: s_lshl_b32 s15, s15, 24
; SI-NEXT: s_or_b32 s78, s15, s11
-; SI-NEXT: v_readlane_b32 s11, v62, 7
+; SI-NEXT: v_readlane_b32 s11, v62, 10
; SI-NEXT: s_and_b32 s11, s11, 0xff
; SI-NEXT: s_lshl_b32 s15, s17, 8
; SI-NEXT: s_or_b32 s11, s11, s15
; SI-NEXT: s_and_b32 s11, s11, 0xffff
+; SI-NEXT: s_mov_b32 s30, s18
+; SI-NEXT: s_mov_b32 s18, s89
+; SI-NEXT: s_mov_b32 s89, s98
+; SI-NEXT: s_mov_b32 s98, s96
+; SI-NEXT: s_mov_b32 s96, s66
+; SI-NEXT: s_mov_b32 s66, s82
+; SI-NEXT: s_mov_b32 s82, s52
; SI-NEXT: v_mov_b32_e32 v51, s9
-; SI-NEXT: s_or_b32 s17, s11, s9
-; SI-NEXT: v_readlane_b32 s9, v62, 2
-; SI-NEXT: v_readlane_b32 s11, v62, 1
+; SI-NEXT: s_or_b32 s52, s11, s9
+; SI-NEXT: v_readlane_b32 s9, v62, 4
+; SI-NEXT: v_readlane_b32 s11, v62, 3
; SI-NEXT: s_and_b32 s9, s9, 0xff
; SI-NEXT: s_lshl_b32 s15, s11, 8
; SI-NEXT: s_or_b32 s9, s9, s15
; SI-NEXT: s_and_b32 s9, s9, 0xffff
-; SI-NEXT: s_mov_b32 s4, s96
-; SI-NEXT: s_mov_b32 s96, s24
; SI-NEXT: v_mov_b32_e32 v52, s14
-; SI-NEXT: s_or_b32 s24, s9, s14
+; SI-NEXT: s_or_b32 s17, s9, s14
; SI-NEXT: s_and_b32 s14, s93, 0xff
; SI-NEXT: s_lshl_b32 s15, s84, 8
; SI-NEXT: s_or_b32 s14, s14, s15
; SI-NEXT: v_lshlrev_b32_e32 v1, 16, v1
-; SI-NEXT: v_or_b32_e32 v53, v6, v1
+; SI-NEXT: v_or_b32_e32 v53, v60, v1
; SI-NEXT: s_and_b32 s14, s14, 0xffff
; SI-NEXT: v_or_b32_e32 v50, s14, v53
-; SI-NEXT: s_and_b32 s14, s8, 0xff
-; SI-NEXT: s_lshl_b32 s15, s49, 8
-; SI-NEXT: v_and_b32_e32 v1, 0xff, v21
+; SI-NEXT: s_and_b32 s14, s23, 0xff
+; SI-NEXT: s_lshl_b32 s15, s53, 8
+; SI-NEXT: v_and_b32_e32 v1, 0xff, v29
; SI-NEXT: s_or_b32 s14, s14, s15
; SI-NEXT: v_lshlrev_b32_e32 v1, 16, v1
-; SI-NEXT: v_or_b32_e32 v54, v14, v1
+; SI-NEXT: v_or_b32_e32 v54, v61, v1
; SI-NEXT: s_and_b32 s14, s14, 0xffff
; SI-NEXT: v_or_b32_e32 v17, s14, v54
-; SI-NEXT: s_and_b32 s14, s40, 0xff
-; SI-NEXT: s_lshl_b32 s15, s53, 8
+; SI-NEXT: s_and_b32 s14, s48, 0xff
+; SI-NEXT: s_lshl_b32 s15, s8, 8
; SI-NEXT: s_waitcnt vmcnt(1)
-; SI-NEXT: v_and_b32_e32 v1, 0xff, v29
+; SI-NEXT: v_and_b32_e32 v1, 0xff, v19
; SI-NEXT: s_or_b32 s14, s14, s15
; SI-NEXT: v_lshlrev_b32_e32 v1, 16, v1
; SI-NEXT: v_or_b32_e32 v55, v18, v1
; SI-NEXT: s_and_b32 s14, s14, 0xffff
; SI-NEXT: v_or_b32_e32 v16, s14, v55
-; SI-NEXT: s_and_b32 s14, s34, 0xff
-; SI-NEXT: s_lshl_b32 s15, s23, 8
+; SI-NEXT: s_and_b32 s14, s95, 0xff
+; SI-NEXT: s_lshl_b32 s15, s63, 8
; SI-NEXT: s_waitcnt vmcnt(0)
-; SI-NEXT: v_and_b32_e32 v1, 0xff, v13
+; SI-NEXT: v_and_b32_e32 v1, 0xff, v21
; SI-NEXT: s_or_b32 s14, s14, s15
; SI-NEXT: v_lshlrev_b32_e32 v1, 16, v1
-; SI-NEXT: v_or_b32_e32 v40, v19, v1
+; SI-NEXT: v_or_b32_e32 v40, v6, v1
; SI-NEXT: s_and_b32 s14, s14, 0xffff
; SI-NEXT: v_or_b32_e32 v15, s14, v40
-; SI-NEXT: s_and_b32 s14, s91, 0xff
-; SI-NEXT: s_lshl_b32 s15, s66, 8
-; SI-NEXT: v_and_b32_e32 v1, 0xff, v43
+; SI-NEXT: s_and_b32 s14, s28, 0xff
+; SI-NEXT: s_lshl_b32 s15, s24, 8
+; SI-NEXT: v_and_b32_e32 v1, 0xff, v44
; SI-NEXT: s_or_b32 s14, s14, s15
; SI-NEXT: v_lshlrev_b32_e32 v1, 16, v1
-; SI-NEXT: v_or_b32_e32 v41, v22, v1
+; SI-NEXT: v_or_b32_e32 v41, v20, v1
; SI-NEXT: s_and_b32 s14, s14, 0xffff
; SI-NEXT: v_or_b32_e32 v12, s14, v41
; SI-NEXT: s_and_b32 s14, s50, 0xff
; SI-NEXT: s_lshl_b32 s15, s54, 8
-; SI-NEXT: v_and_b32_e32 v1, 0xff, v32
+; SI-NEXT: v_and_b32_e32 v1, 0xff, v47
; SI-NEXT: s_or_b32 s14, s14, s15
; SI-NEXT: v_lshlrev_b32_e32 v1, 16, v1
-; SI-NEXT: v_or_b32_e32 v42, v23, v1
+; SI-NEXT: v_or_b32_e32 v42, v22, v1
; SI-NEXT: s_and_b32 s14, s14, 0xffff
; SI-NEXT: v_or_b32_e32 v11, s14, v42
-; SI-NEXT: s_and_b32 s14, s73, 0xff
+; SI-NEXT: s_and_b32 s14, s40, 0xff
; SI-NEXT: s_lshl_b32 s15, s36, 8
-; SI-NEXT: v_and_b32_e32 v1, 0xff, v46
+; SI-NEXT: v_and_b32_e32 v1, 0xff, v34
; SI-NEXT: s_or_b32 s14, s14, s15
; SI-NEXT: v_lshlrev_b32_e32 v1, 16, v1
-; SI-NEXT: v_or_b32_e32 v28, v59, v1
+; SI-NEXT: v_or_b32_e32 v59, v23, v1
; SI-NEXT: s_and_b32 s14, s14, 0xffff
-; SI-NEXT: v_or_b32_e32 v10, s14, v28
-; SI-NEXT: s_and_b32 s14, s82, 0xff
-; SI-NEXT: s_lshl_b32 s15, s52, 8
-; SI-NEXT: v_and_b32_e32 v1, 0xff, v26
+; SI-NEXT: v_readlane_b32 s8, v62, 57
+; SI-NEXT: v_or_b32_e32 v10, s14, v59
+; SI-NEXT: s_and_b32 s14, s25, 0xff
+; SI-NEXT: s_lshl_b32 s15, s8, 8
+; SI-NEXT: v_and_b32_e32 v1, 0xff, v35
; SI-NEXT: s_or_b32 s14, s14, s15
; SI-NEXT: v_lshlrev_b32_e32 v1, 16, v1
-; SI-NEXT: v_or_b32_e32 v60, v24, v1
+; SI-NEXT: v_mov_b32_e32 v25, v20
+; SI-NEXT: v_mov_b32_e32 v20, v60
+; SI-NEXT: v_or_b32_e32 v60, v14, v1
; SI-NEXT: s_and_b32 s14, s14, 0xffff
; SI-NEXT: v_or_b32_e32 v9, s14, v60
-; SI-NEXT: s_and_b32 s14, s90, 0xff
-; SI-NEXT: s_lshl_b32 s15, s16, 8
-; SI-NEXT: v_and_b32_e32 v1, 0xff, v35
+; SI-NEXT: s_and_b32 s14, s70, 0xff
+; SI-NEXT: s_lshl_b32 s15, s49, 8
+; SI-NEXT: v_and_b32_e32 v1, 0xff, v36
; SI-NEXT: s_or_b32 s14, s14, s15
; SI-NEXT: v_lshlrev_b32_e32 v1, 16, v1
-; SI-NEXT: v_or_b32_e32 v31, v44, v1
+; SI-NEXT: v_or_b32_e32 v5, v43, v1
; SI-NEXT: s_and_b32 s14, s14, 0xffff
-; SI-NEXT: v_or_b32_e32 v8, s14, v31
-; SI-NEXT: s_and_b32 s14, s62, 0xff
-; SI-NEXT: s_lshl_b32 s15, s65, 8
-; SI-NEXT: v_and_b32_e32 v1, 0xff, v37
+; SI-NEXT: v_or_b32_e32 v8, s14, v5
+; SI-NEXT: s_and_b32 s14, s39, 0xff
+; SI-NEXT: s_lshl_b32 s15, s94, 8
+; SI-NEXT: v_and_b32_e32 v1, 0xff, v32
; SI-NEXT: s_or_b32 s14, s14, s15
; SI-NEXT: v_lshlrev_b32_e32 v1, 16, v1
-; SI-NEXT: v_or_b32_e32 v61, v45, v1
+; SI-NEXT: v_or_b32_e32 v31, v45, v1
; SI-NEXT: s_and_b32 s14, s14, 0xffff
-; SI-NEXT: v_or_b32_e32 v7, s14, v61
-; SI-NEXT: s_and_b32 s14, s98, 0xff
+; SI-NEXT: v_or_b32_e32 v7, s14, v31
+; SI-NEXT: s_and_b32 s14, s87, 0xff
; SI-NEXT: s_lshl_b32 s15, s67, 8
; SI-NEXT: v_and_b32_e32 v1, 0xff, v38
; SI-NEXT: s_or_b32 s14, s14, s15
; SI-NEXT: v_lshlrev_b32_e32 v1, 16, v1
-; SI-NEXT: s_waitcnt expcnt(5)
-; SI-NEXT: v_or_b32_e32 v6, v47, v1
+; SI-NEXT: v_mov_b32_e32 v26, v34
+; SI-NEXT: v_mov_b32_e32 v34, v22
+; SI-NEXT: v_mov_b32_e32 v22, v61
+; SI-NEXT: v_or_b32_e32 v61, v46, v1
; SI-NEXT: s_and_b32 s14, s14, 0xffff
-; SI-NEXT: v_or_b32_e32 v4, s14, v6
+; SI-NEXT: v_or_b32_e32 v4, s14, v61
; SI-NEXT: s_and_b32 s14, s92, 0xff
; SI-NEXT: s_lshl_b32 s15, s7, 8
; SI-NEXT: v_and_b32_e32 v1, 0xff, v33
; SI-NEXT: s_or_b32 s14, s14, s15
; SI-NEXT: v_lshlrev_b32_e32 v1, 16, v1
-; SI-NEXT: v_readlane_b32 s8, v62, 55
+; SI-NEXT: v_readlane_b32 s8, v62, 54
; SI-NEXT: s_waitcnt expcnt(4)
-; SI-NEXT: v_mov_b32_e32 v22, v14
-; SI-NEXT: v_or_b32_e32 v14, v56, v1
+; SI-NEXT: v_or_b32_e32 v6, v56, v1
; SI-NEXT: s_and_b32 s14, s14, 0xffff
; SI-NEXT: s_and_b32 s8, s8, 0xffff
-; SI-NEXT: v_or_b32_e32 v2, s14, v14
-; SI-NEXT: s_and_b32 s14, s70, 0xff
-; SI-NEXT: s_lshl_b32 s15, s94, 8
+; SI-NEXT: v_readlane_b32 s11, v62, 55
+; SI-NEXT: v_mov_b32_e32 v28, v36
+; SI-NEXT: v_or_b32_e32 v36, s14, v6
+; SI-NEXT: s_and_b32 s14, s26, 0xff
+; SI-NEXT: s_lshl_b32 s15, s62, 8
; SI-NEXT: v_and_b32_e32 v1, 0xff, v39
-; SI-NEXT: s_or_b32 s42, s8, s63
+; SI-NEXT: s_or_b32 s42, s8, s11
; SI-NEXT: v_readlane_b32 s8, v62, 56
; SI-NEXT: s_or_b32 s14, s14, s15
; SI-NEXT: v_lshlrev_b32_e32 v1, 16, v1
; SI-NEXT: s_and_b32 s8, s8, 0xffff
-; SI-NEXT: s_waitcnt expcnt(2)
-; SI-NEXT: v_mov_b32_e32 v32, v23
-; SI-NEXT: v_mov_b32_e32 v23, v18
-; SI-NEXT: v_or_b32_e32 v18, v57, v1
+; SI-NEXT: s_waitcnt expcnt(1)
+; SI-NEXT: v_or_b32_e32 v14, v57, v1
; SI-NEXT: s_and_b32 s14, s14, 0xffff
; SI-NEXT: s_or_b32 s40, s8, s5
-; SI-NEXT: v_readlane_b32 s8, v62, 57
-; SI-NEXT: v_or_b32_e32 v1, s14, v18
+; SI-NEXT: v_readlane_b32 s8, v62, 58
+; SI-NEXT: v_or_b32_e32 v1, s14, v14
; SI-NEXT: s_and_b32 s14, s88, 0xff
; SI-NEXT: s_lshl_b32 s15, s79, 8
-; SI-NEXT: v_and_b32_e32 v3, 0xff, v34
+; SI-NEXT: v_and_b32_e32 v3, 0xff, v37
; SI-NEXT: s_and_b32 s8, s8, 0xffff
-; SI-NEXT: v_readlane_b32 s9, v62, 60
; SI-NEXT: s_or_b32 s14, s14, s15
; SI-NEXT: v_lshlrev_b32_e32 v3, 16, v3
; SI-NEXT: s_or_b32 s15, s8, s41
-; SI-NEXT: v_readlane_b32 s8, v62, 58
-; SI-NEXT: s_and_b32 s16, s9, 0xffff
-; SI-NEXT: v_mov_b32_e32 v27, v26
-; SI-NEXT: v_mov_b32_e32 v26, v24
-; SI-NEXT: v_mov_b32_e32 v24, v19
-; SI-NEXT: v_or_b32_e32 v19, v58, v3
+; SI-NEXT: v_readlane_b32 s8, v62, 59
+; SI-NEXT: s_and_b32 s16, s16, 0xffff
+; SI-NEXT: v_mov_b32_e32 v27, v35
+; SI-NEXT: v_mov_b32_e32 v35, v23
+; SI-NEXT: v_mov_b32_e32 v23, v18
+; SI-NEXT: v_or_b32_e32 v18, v58, v3
; SI-NEXT: s_and_b32 s14, s14, 0xffff
; SI-NEXT: s_and_b32 s8, s8, 0xffff
; SI-NEXT: s_or_b32 s36, s16, s56
-; SI-NEXT: s_and_b32 s16, s48, 0xffff
-; SI-NEXT: v_or_b32_e32 v3, s14, v19
+; SI-NEXT: s_and_b32 s16, s73, 0xffff
+; SI-NEXT: v_or_b32_e32 v3, s14, v18
; SI-NEXT: s_or_b32 s14, s8, s43
; SI-NEXT: s_and_b32 s8, s13, 0xffff
; SI-NEXT: s_or_b32 s53, s16, vcc_lo
@@ -209528,49 +209646,46 @@ define inreg <64 x i16> @bitcast_v128i8_to_v64i16_scalar(<128 x i8> inreg %a, i3
; SI-NEXT: s_or_b32 s49, s16, s74
; SI-NEXT: s_and_b32 s16, s58, 0xffff
; SI-NEXT: s_or_b32 s10, s8, s46
-; SI-NEXT: v_readlane_b32 s8, v62, 59
+; SI-NEXT: v_readlane_b32 s8, v62, 60
; SI-NEXT: s_or_b32 s48, s16, s75
; SI-NEXT: s_and_b32 s16, s59, 0xffff
; SI-NEXT: s_and_b32 s8, s8, 0xffff
-; SI-NEXT: s_or_b32 s11, s16, s76
+; SI-NEXT: s_or_b32 s39, s16, s76
; SI-NEXT: s_and_b32 s16, s60, 0xffff
; SI-NEXT: s_and_b32 s23, s61, 0xffff
-; SI-NEXT: s_mov_b32 s30, s87
-; SI-NEXT: s_mov_b32 s87, s85
; SI-NEXT: s_or_b32 s8, s8, s47
-; SI-NEXT: s_or_b32 s9, s16, s77
-; SI-NEXT: s_or_b32 s16, s23, s78
-; SI-NEXT: v_mov_b32_e32 v36, v35
-; SI-NEXT: v_mov_b32_e32 v30, v37
-; SI-NEXT: v_mov_b32_e32 v35, v45
-; SI-NEXT: v_mov_b32_e32 v20, v47
-; SI-NEXT: v_mov_b32_e32 v49, v56
-; SI-NEXT: v_mov_b32_e32 v48, v39
-; SI-NEXT: v_mov_b32_e32 v39, v57
-; SI-NEXT: v_mov_b32_e32 v25, v58
+; SI-NEXT: s_or_b32 s70, s16, s77
+; SI-NEXT: s_or_b32 s9, s23, s78
+; SI-NEXT: v_mov_b32_e32 v24, v45
+; SI-NEXT: v_mov_b32_e32 v48, v46
+; SI-NEXT: v_mov_b32_e32 v30, v56
+; SI-NEXT: v_mov_b32_e32 v49, v57
+; SI-NEXT: v_mov_b32_e32 v2, v58
; SI-NEXT: v_alignbit_b32 v57, s42, v51, 16
; SI-NEXT: v_alignbit_b32 v58, s40, v52, 16
; SI-NEXT: v_alignbit_b32 v56, s15, v53, 16
; SI-NEXT: v_alignbit_b32 v47, s14, v54, 16
-; SI-NEXT: s_waitcnt expcnt(1)
; SI-NEXT: v_alignbit_b32 v46, s13, v55, 16
; SI-NEXT: v_alignbit_b32 v45, s12, v40, 16
-; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_alignbit_b32 v44, s10, v41, 16
+; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_alignbit_b32 v43, s8, v42, 16
-; SI-NEXT: v_alignbit_b32 v42, s36, v28, 16
+; SI-NEXT: v_alignbit_b32 v42, s36, v59, 16
; SI-NEXT: v_alignbit_b32 v41, s53, v60, 16
-; SI-NEXT: v_alignbit_b32 v40, s94, v31, 16
-; SI-NEXT: v_alignbit_b32 v55, s49, v61, 16
-; SI-NEXT: v_alignbit_b32 v54, s48, v6, 16
-; SI-NEXT: v_alignbit_b32 v53, s11, v14, 16
-; SI-NEXT: v_mov_b32_e32 v14, v22
-; SI-NEXT: v_alignbit_b32 v52, s9, v18, 16
+; SI-NEXT: v_mov_b32_e32 v60, v20
+; SI-NEXT: v_mov_b32_e32 v20, v25
+; SI-NEXT: v_alignbit_b32 v40, s94, v5, 16
+; SI-NEXT: v_alignbit_b32 v55, s49, v31, 16
+; SI-NEXT: v_alignbit_b32 v54, s48, v61, 16
+; SI-NEXT: v_mov_b32_e32 v61, v22
+; SI-NEXT: v_mov_b32_e32 v22, v34
+; SI-NEXT: v_alignbit_b32 v53, s39, v6, 16
+; SI-NEXT: s_mov_b32 s16, s62
+; SI-NEXT: v_alignbit_b32 v52, s70, v14, 16
+; SI-NEXT: v_alignbit_b32 v51, s9, v18, 16
; SI-NEXT: v_mov_b32_e32 v18, v23
-; SI-NEXT: v_alignbit_b32 v51, s16, v19, 16
-; SI-NEXT: v_mov_b32_e32 v19, v24
-; SI-NEXT: v_mov_b32_e32 v24, v26
-; SI-NEXT: s_lshr_b32 s73, s63, 16
+; SI-NEXT: v_mov_b32_e32 v23, v35
+; SI-NEXT: s_lshr_b32 s73, s11, 16
; SI-NEXT: s_lshr_b32 s72, s5, 16
; SI-NEXT: s_lshr_b32 s63, s41, 16
; SI-NEXT: s_lshr_b32 s62, s43, 16
@@ -209582,87 +209697,85 @@ define inreg <64 x i16> @bitcast_v128i8_to_v64i16_scalar(<128 x i8> inreg %a, i3
; SI-NEXT: s_lshr_b32 s56, vcc_lo, 16
; SI-NEXT: s_lshr_b32 s47, vcc_hi, 16
; SI-NEXT: s_lshr_b32 s46, s74, 16
-; SI-NEXT: v_readlane_b32 s25, v62, 62
-; SI-NEXT: v_readlane_b32 s74, v62, 61
+; SI-NEXT: v_readlane_b32 s34, v62, 61
; SI-NEXT: s_lshr_b32 s45, s75, 16
+; SI-NEXT: v_readlane_b32 s6, v62, 62
; SI-NEXT: s_lshr_b32 s44, s76, 16
-; SI-NEXT: s_mov_b32 s76, s37
-; SI-NEXT: s_mov_b32 s37, s97
-; SI-NEXT: s_mov_b32 s97, s21
-; SI-NEXT: s_mov_b32 s21, s18
-; SI-NEXT: s_mov_b32 s18, s17
-; SI-NEXT: s_mov_b32 s85, s87
-; SI-NEXT: s_mov_b32 s87, s30
-; SI-NEXT: s_mov_b32 s17, s24
+; SI-NEXT: s_mov_b32 s11, s37
+; SI-NEXT: v_readlane_b32 s37, v62, 57
+; SI-NEXT: s_mov_b32 s76, s97
+; SI-NEXT: s_mov_b32 s97, s4
+; SI-NEXT: v_readlane_b32 s85, v62, 63
; SI-NEXT: s_lshr_b32 s43, s77, 16
; SI-NEXT: s_mov_b32 s77, s35
; SI-NEXT: s_mov_b32 s35, s64
; SI-NEXT: s_mov_b32 s64, s55
; SI-NEXT: s_mov_b32 s55, s83
; SI-NEXT: s_mov_b32 s83, s99
-; SI-NEXT: s_mov_b32 s99, s89
-; SI-NEXT: s_mov_b32 s89, s39
-; SI-NEXT: s_mov_b32 s39, s95
-; SI-NEXT: s_mov_b32 s95, s6
+; SI-NEXT: s_mov_b32 s99, s38
+; SI-NEXT: s_mov_b32 s38, s20
+; SI-NEXT: s_mov_b32 s20, s21
+; SI-NEXT: s_mov_b32 s21, s52
; SI-NEXT: s_lshr_b32 s41, s78, 16
-; SI-NEXT: s_mov_b32 s24, s96
-; SI-NEXT: s_mov_b32 s96, s4
+; SI-NEXT: s_mov_b32 s52, s82
+; SI-NEXT: s_mov_b32 s82, s66
+; SI-NEXT: s_mov_b32 s66, s96
+; SI-NEXT: s_mov_b32 s96, s98
+; SI-NEXT: s_mov_b32 s98, s89
+; SI-NEXT: s_mov_b32 s89, s18
+; SI-NEXT: s_mov_b32 s18, s30
+; SI-NEXT: v_mov_b32_e32 v6, v19
+; SI-NEXT: v_mov_b32_e32 v14, v21
; SI-NEXT: s_cbranch_execnz .LBB97_3
; SI-NEXT: .LBB97_2: ; %cmp.true
-; SI-NEXT: buffer_load_dword v9, off, s[0:3], s32 offset:424 ; 4-byte Folded Reload
-; SI-NEXT: v_add_i32_e32 v8, vcc, 3, v36
-; SI-NEXT: v_and_b32_e32 v8, 0xff, v8
-; SI-NEXT: v_mov_b32_e32 v6, v5
-; SI-NEXT: v_mov_b32_e32 v5, v27
-; SI-NEXT: v_lshlrev_b32_e32 v8, 16, v8
+; SI-NEXT: buffer_load_dword v8, off, s[0:3], s32 offset:420 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v9, off, s[0:3], s32 offset:416 ; 4-byte Folded Reload
; SI-NEXT: s_add_i32 s4, s88, 3
; SI-NEXT: s_and_b32 s4, s4, 0xff
; SI-NEXT: s_lshl_b32 s5, s79, 8
-; SI-NEXT: v_add_i32_e32 v1, vcc, 3, v34
+; SI-NEXT: v_add_i32_e32 v1, vcc, 3, v37
; SI-NEXT: s_or_b32 s4, s5, s4
; SI-NEXT: v_and_b32_e32 v1, 0xff, v1
; SI-NEXT: s_addk_i32 s4, 0x300
; SI-NEXT: v_lshlrev_b32_e32 v1, 16, v1
; SI-NEXT: s_and_b32 s4, s4, 0xffff
-; SI-NEXT: v_or_b32_e32 v1, v25, v1
+; SI-NEXT: v_or_b32_e32 v1, v2, v1
; SI-NEXT: v_or_b32_e32 v1, s4, v1
-; SI-NEXT: v_readlane_b32 s4, v62, 11
-; SI-NEXT: s_add_i32 s4, s4, 3
-; SI-NEXT: v_readlane_b32 s5, v62, 10
-; SI-NEXT: v_readlane_b32 s6, v62, 9
+; SI-NEXT: s_add_i32 s4, s82, 3
+; SI-NEXT: v_readlane_b32 s8, v62, 12
; SI-NEXT: s_and_b32 s4, s4, 0xff
-; SI-NEXT: s_lshl_b32 s5, s5, 8
-; SI-NEXT: s_add_i32 s8, s6, 3
+; SI-NEXT: s_lshl_b32 s5, s52, 8
+; SI-NEXT: s_add_i32 s8, s8, 3
; SI-NEXT: s_or_b32 s4, s5, s4
-; SI-NEXT: v_readlane_b32 s5, v62, 8
+; SI-NEXT: v_readlane_b32 s5, v62, 11
; SI-NEXT: s_and_b32 s8, s8, 0xff
; SI-NEXT: s_lshl_b32 s5, s5, 24
; SI-NEXT: s_lshl_b32 s8, s8, 16
; SI-NEXT: s_addk_i32 s4, 0x300
; SI-NEXT: s_or_b32 s5, s5, s8
; SI-NEXT: s_and_b32 s4, s4, 0xffff
-; SI-NEXT: s_add_i32 s70, s70, 3
-; SI-NEXT: v_readlane_b32 s6, v62, 30
; SI-NEXT: s_or_b32 s4, s5, s4
+; SI-NEXT: v_readlane_b32 s5, v62, 33
+; SI-NEXT: s_add_i32 s70, s5, 3
; SI-NEXT: s_and_b32 s5, s70, 0xff
-; SI-NEXT: s_lshl_b32 s8, s6, 8
-; SI-NEXT: v_add_i32_e32 v2, vcc, 3, v48
+; SI-NEXT: s_lshl_b32 s8, s16, 8
+; SI-NEXT: v_add_i32_e32 v2, vcc, 3, v39
; SI-NEXT: s_or_b32 s5, s8, s5
; SI-NEXT: v_and_b32_e32 v2, 0xff, v2
; SI-NEXT: s_addk_i32 s5, 0x300
; SI-NEXT: v_lshlrev_b32_e32 v2, 16, v2
; SI-NEXT: s_and_b32 s5, s5, 0xffff
-; SI-NEXT: v_or_b32_e32 v2, v39, v2
+; SI-NEXT: v_or_b32_e32 v2, v49, v2
; SI-NEXT: v_or_b32_e32 v2, s5, v2
; SI-NEXT: s_add_i32 s5, s35, 3
-; SI-NEXT: v_readlane_b32 s6, v62, 13
+; SI-NEXT: v_readlane_b32 s8, v62, 14
; SI-NEXT: s_and_b32 s5, s5, 0xff
-; SI-NEXT: s_lshl_b32 s8, s6, 8
+; SI-NEXT: s_lshl_b32 s8, s8, 8
; SI-NEXT: s_add_i32 s9, s77, 3
; SI-NEXT: s_or_b32 s5, s8, s5
-; SI-NEXT: v_readlane_b32 s6, v62, 12
+; SI-NEXT: v_readlane_b32 s8, v62, 13
; SI-NEXT: s_and_b32 s9, s9, 0xff
-; SI-NEXT: s_lshl_b32 s8, s6, 24
+; SI-NEXT: s_lshl_b32 s8, s8, 24
; SI-NEXT: s_lshl_b32 s9, s9, 16
; SI-NEXT: s_addk_i32 s5, 0x300
; SI-NEXT: s_or_b32 s8, s8, s9
@@ -209670,9 +209783,9 @@ define inreg <64 x i16> @bitcast_v128i8_to_v64i16_scalar(<128 x i8> inreg %a, i3
; SI-NEXT: s_or_b32 s5, s8, s5
; SI-NEXT: s_add_i32 s79, s92, 3
; SI-NEXT: v_add_i32_e32 v3, vcc, 0x3000000, v1
-; SI-NEXT: s_add_i32 s16, s4, 0x3000000
+; SI-NEXT: s_add_i32 s9, s4, 0x3000000
; SI-NEXT: v_add_i32_e32 v1, vcc, 0x3000000, v2
-; SI-NEXT: s_add_i32 s9, s5, 0x3000000
+; SI-NEXT: s_add_i32 s70, s5, 0x3000000
; SI-NEXT: s_and_b32 s4, s79, 0xff
; SI-NEXT: s_lshl_b32 s5, s7, 8
; SI-NEXT: v_add_i32_e32 v2, vcc, 3, v33
@@ -209681,16 +209794,12 @@ define inreg <64 x i16> @bitcast_v128i8_to_v64i16_scalar(<128 x i8> inreg %a, i3
; SI-NEXT: s_addk_i32 s4, 0x300
; SI-NEXT: v_lshlrev_b32_e32 v2, 16, v2
; SI-NEXT: s_and_b32 s4, s4, 0xffff
-; SI-NEXT: v_or_b32_e32 v2, v49, v2
+; SI-NEXT: v_or_b32_e32 v2, v30, v2
; SI-NEXT: v_or_b32_e32 v2, s4, v2
-; SI-NEXT: s_waitcnt vmcnt(0)
-; SI-NEXT: v_or_b32_e32 v8, v9, v8
-; SI-NEXT: v_add_i32_e32 v9, vcc, 3, v5
-; SI-NEXT: buffer_load_dword v5, off, s[0:3], s32 offset:420 ; 4-byte Folded Reload
-; SI-NEXT: s_add_i32 s4, s24, 3
+; SI-NEXT: s_add_i32 s4, s66, 3
; SI-NEXT: s_and_b32 s4, s4, 0xff
; SI-NEXT: s_lshl_b32 s5, s55, 8
-; SI-NEXT: s_add_i32 s8, s37, 3
+; SI-NEXT: s_add_i32 s8, s76, 3
; SI-NEXT: s_or_b32 s4, s5, s4
; SI-NEXT: s_and_b32 s8, s8, 0xff
; SI-NEXT: s_addk_i32 s4, 0x300
@@ -209699,8 +209808,8 @@ define inreg <64 x i16> @bitcast_v128i8_to_v64i16_scalar(<128 x i8> inreg %a, i3
; SI-NEXT: s_and_b32 s4, s4, 0xffff
; SI-NEXT: s_or_b32 s5, s5, s8
; SI-NEXT: s_or_b32 s4, s5, s4
-; SI-NEXT: s_add_i32 s52, s98, 3
-; SI-NEXT: s_add_i32 s11, s4, 0x3000000
+; SI-NEXT: s_add_i32 s52, s87, 3
+; SI-NEXT: s_add_i32 s39, s4, 0x3000000
; SI-NEXT: s_and_b32 s4, s52, 0xff
; SI-NEXT: s_lshl_b32 s5, s67, 8
; SI-NEXT: v_add_i32_e32 v4, vcc, 3, v38
@@ -209709,64 +209818,68 @@ define inreg <64 x i16> @bitcast_v128i8_to_v64i16_scalar(<128 x i8> inreg %a, i3
; SI-NEXT: s_addk_i32 s4, 0x300
; SI-NEXT: v_lshlrev_b32_e32 v4, 16, v4
; SI-NEXT: s_and_b32 s4, s4, 0xffff
-; SI-NEXT: v_or_b32_e32 v4, v20, v4
-; SI-NEXT: s_add_i32 s30, s87, 3
+; SI-NEXT: v_or_b32_e32 v4, v48, v4
+; SI-NEXT: s_add_i32 s30, s18, 3
; SI-NEXT: v_or_b32_e32 v4, s4, v4
; SI-NEXT: s_and_b32 s4, s30, 0xff
-; SI-NEXT: s_lshl_b32 s5, s21, 8
-; SI-NEXT: s_add_i32 s8, s68, 3
+; SI-NEXT: s_lshl_b32 s5, s98, 8
+; SI-NEXT: s_add_i32 s8, s6, 3
; SI-NEXT: s_or_b32 s4, s5, s4
; SI-NEXT: s_and_b32 s8, s8, 0xff
; SI-NEXT: s_addk_i32 s4, 0x300
-; SI-NEXT: s_lshl_b32 s5, s28, 24
+; SI-NEXT: s_lshl_b32 s5, s27, 24
; SI-NEXT: s_lshl_b32 s8, s8, 16
; SI-NEXT: s_and_b32 s4, s4, 0xffff
; SI-NEXT: s_or_b32 s5, s5, s8
; SI-NEXT: s_or_b32 s4, s5, s4
; SI-NEXT: s_add_i32 s48, s4, 0x3000000
; SI-NEXT: v_readlane_b32 s4, v62, 42
-; SI-NEXT: v_mov_b32_e32 v22, v30
; SI-NEXT: s_add_i32 s87, s4, 3
; SI-NEXT: v_readlane_b32 s5, v62, 39
; SI-NEXT: s_and_b32 s4, s87, 0xff
; SI-NEXT: s_lshl_b32 s5, s5, 8
-; SI-NEXT: v_add_i32_e32 v7, vcc, 3, v22
+; SI-NEXT: v_add_i32_e32 v5, vcc, 3, v32
; SI-NEXT: s_or_b32 s4, s5, s4
-; SI-NEXT: v_and_b32_e32 v7, 0xff, v7
+; SI-NEXT: v_and_b32_e32 v5, 0xff, v5
; SI-NEXT: s_addk_i32 s4, 0x300
-; SI-NEXT: v_lshlrev_b32_e32 v7, 16, v7
+; SI-NEXT: v_lshlrev_b32_e32 v5, 16, v5
; SI-NEXT: s_and_b32 s4, s4, 0xffff
-; SI-NEXT: v_or_b32_e32 v7, v35, v7
-; SI-NEXT: v_or_b32_e32 v7, s4, v7
-; SI-NEXT: v_readlane_b32 s4, v62, 32
+; SI-NEXT: v_or_b32_e32 v5, v24, v5
+; SI-NEXT: v_or_b32_e32 v5, s4, v5
+; SI-NEXT: v_readlane_b32 s4, v62, 30
; SI-NEXT: s_add_i32 s67, s4, 3
; SI-NEXT: s_and_b32 s4, s67, 0xff
-; SI-NEXT: s_lshl_b32 s5, s85, 8
+; SI-NEXT: s_lshl_b32 s5, s86, 8
; SI-NEXT: s_add_i32 s8, s69, 3
; SI-NEXT: s_or_b32 s4, s5, s4
; SI-NEXT: s_and_b32 s8, s8, 0xff
; SI-NEXT: s_addk_i32 s4, 0x300
-; SI-NEXT: s_lshl_b32 s5, s74, 24
+; SI-NEXT: s_lshl_b32 s5, s34, 24
; SI-NEXT: s_lshl_b32 s8, s8, 16
; SI-NEXT: s_and_b32 s4, s4, 0xffff
; SI-NEXT: s_or_b32 s5, s5, s8
; SI-NEXT: s_or_b32 s4, s5, s4
-; SI-NEXT: s_add_i32 s50, s90, 3
-; SI-NEXT: v_readlane_b32 s5, v62, 23
; SI-NEXT: s_add_i32 s49, s4, 0x3000000
+; SI-NEXT: v_readlane_b32 s4, v62, 24
+; SI-NEXT: v_mov_b32_e32 v25, v28
+; SI-NEXT: s_add_i32 s50, s4, 3
+; SI-NEXT: v_readlane_b32 s5, v62, 22
+; SI-NEXT: v_add_i32_e32 v7, vcc, 0x3000000, v5
; SI-NEXT: s_and_b32 s4, s50, 0xff
; SI-NEXT: s_lshl_b32 s5, s5, 8
+; SI-NEXT: v_add_i32_e32 v5, vcc, 3, v25
; SI-NEXT: s_or_b32 s4, s5, s4
+; SI-NEXT: v_and_b32_e32 v5, 0xff, v5
; SI-NEXT: s_addk_i32 s4, 0x300
+; SI-NEXT: v_lshlrev_b32_e32 v5, 16, v5
; SI-NEXT: s_and_b32 s4, s4, 0xffff
-; SI-NEXT: s_add_i32 s94, s86, 3
-; SI-NEXT: v_or_b32_e32 v8, s4, v8
+; SI-NEXT: s_waitcnt vmcnt(1)
+; SI-NEXT: v_or_b32_e32 v5, v8, v5
+; SI-NEXT: s_add_i32 s94, s89, 3
+; SI-NEXT: v_or_b32_e32 v5, s4, v5
; SI-NEXT: s_and_b32 s4, s94, 0xff
-; SI-NEXT: s_lshl_b32 s5, s38, 8
+; SI-NEXT: s_lshl_b32 s5, s85, 8
; SI-NEXT: s_add_i32 s8, s71, 3
-; SI-NEXT: s_waitcnt vmcnt(0)
-; SI-NEXT: v_add_i32_e32 v10, vcc, 3, v5
-; SI-NEXT: buffer_load_dword v5, off, s[0:3], s32 offset:416 ; 4-byte Folded Reload
; SI-NEXT: s_or_b32 s4, s5, s4
; SI-NEXT: s_and_b32 s8, s8, 0xff
; SI-NEXT: s_addk_i32 s4, 0x300
@@ -209776,25 +209889,27 @@ define inreg <64 x i16> @bitcast_v128i8_to_v64i16_scalar(<128 x i8> inreg %a, i3
; SI-NEXT: s_or_b32 s5, s5, s8
; SI-NEXT: s_or_b32 s4, s5, s4
; SI-NEXT: s_add_i32 s94, s4, 0x3000000
-; SI-NEXT: v_readlane_b32 s4, v62, 52
+; SI-NEXT: v_readlane_b32 s4, v62, 51
+; SI-NEXT: v_mov_b32_e32 v21, v27
; SI-NEXT: s_add_i32 s18, s4, 3
-; SI-NEXT: v_readlane_b32 s5, v62, 51
+; SI-NEXT: v_add_i32_e32 v8, vcc, 0x3000000, v5
; SI-NEXT: s_and_b32 s4, s18, 0xff
-; SI-NEXT: s_lshl_b32 s5, s5, 8
+; SI-NEXT: s_lshl_b32 s5, s37, 8
+; SI-NEXT: v_add_i32_e32 v5, vcc, 3, v21
; SI-NEXT: s_or_b32 s4, s5, s4
-; SI-NEXT: v_and_b32_e32 v9, 0xff, v9
+; SI-NEXT: v_and_b32_e32 v5, 0xff, v5
; SI-NEXT: s_addk_i32 s4, 0x300
-; SI-NEXT: v_lshlrev_b32_e32 v9, 16, v9
+; SI-NEXT: v_lshlrev_b32_e32 v5, 16, v5
; SI-NEXT: s_and_b32 s4, s4, 0xffff
-; SI-NEXT: v_or_b32_e32 v9, v24, v9
-; SI-NEXT: v_or_b32_e32 v9, s4, v9
+; SI-NEXT: s_waitcnt vmcnt(0)
+; SI-NEXT: v_or_b32_e32 v5, v9, v5
+; SI-NEXT: v_or_b32_e32 v5, s4, v5
; SI-NEXT: v_readlane_b32 s4, v62, 45
; SI-NEXT: s_add_i32 s98, s4, 3
; SI-NEXT: v_readlane_b32 s5, v62, 43
-; SI-NEXT: v_readlane_b32 s6, v62, 14
; SI-NEXT: s_and_b32 s4, s98, 0xff
; SI-NEXT: s_lshl_b32 s5, s5, 8
-; SI-NEXT: s_add_i32 s8, s6, 3
+; SI-NEXT: s_add_i32 s8, s91, 3
; SI-NEXT: s_or_b32 s4, s5, s4
; SI-NEXT: s_and_b32 s8, s8, 0xff
; SI-NEXT: s_addk_i32 s4, 0x300
@@ -209805,27 +209920,33 @@ define inreg <64 x i16> @bitcast_v128i8_to_v64i16_scalar(<128 x i8> inreg %a, i3
; SI-NEXT: s_or_b32 s4, s5, s4
; SI-NEXT: s_add_i32 s53, s4, 0x3000000
; SI-NEXT: v_readlane_b32 s4, v62, 41
+; SI-NEXT: v_mov_b32_e32 v19, v26
; SI-NEXT: s_add_i32 s86, s4, 3
; SI-NEXT: v_readlane_b32 s5, v62, 38
+; SI-NEXT: v_add_i32_e32 v9, vcc, 0x3000000, v5
; SI-NEXT: s_and_b32 s4, s86, 0xff
; SI-NEXT: s_lshl_b32 s5, s5, 8
+; SI-NEXT: v_add_i32_e32 v5, vcc, 3, v19
; SI-NEXT: s_or_b32 s4, s5, s4
-; SI-NEXT: v_and_b32_e32 v10, 0xff, v10
+; SI-NEXT: v_and_b32_e32 v5, 0xff, v5
; SI-NEXT: s_addk_i32 s4, 0x300
-; SI-NEXT: v_lshlrev_b32_e32 v10, 16, v10
+; SI-NEXT: v_lshlrev_b32_e32 v5, 16, v5
; SI-NEXT: s_and_b32 s4, s4, 0xffff
-; SI-NEXT: v_or_b32_e32 v10, v59, v10
-; SI-NEXT: v_or_b32_e32 v10, s4, v10
-; SI-NEXT: v_readlane_b32 s4, v62, 31
+; SI-NEXT: v_or_b32_e32 v5, v23, v5
+; SI-NEXT: v_or_b32_e32 v5, s4, v5
+; SI-NEXT: v_add_i32_e32 v10, vcc, 0x3000000, v5
+; SI-NEXT: buffer_load_dword v5, off, s[0:3], s32 offset:412 ; 4-byte Folded Reload
+; SI-NEXT: v_readlane_b32 s4, v62, 29
; SI-NEXT: s_add_i32 s66, s4, 3
-; SI-NEXT: v_readlane_b32 s5, v62, 27
+; SI-NEXT: v_readlane_b32 s5, v62, 26
; SI-NEXT: s_and_b32 s4, s66, 0xff
; SI-NEXT: s_lshl_b32 s5, s5, 8
-; SI-NEXT: s_add_i32 s37, s39, 3
+; SI-NEXT: s_add_i32 s37, s38, 3
; SI-NEXT: s_or_b32 s4, s5, s4
+; SI-NEXT: v_readlane_b32 s5, v62, 21
; SI-NEXT: s_and_b32 s8, s37, 0xff
; SI-NEXT: s_addk_i32 s4, 0x300
-; SI-NEXT: s_lshl_b32 s5, s95, 24
+; SI-NEXT: s_lshl_b32 s5, s5, 24
; SI-NEXT: s_lshl_b32 s8, s8, 16
; SI-NEXT: s_and_b32 s4, s4, 0xffff
; SI-NEXT: s_or_b32 s5, s5, s8
@@ -209840,33 +209961,65 @@ define inreg <64 x i16> @bitcast_v128i8_to_v64i16_scalar(<128 x i8> inreg %a, i3
; SI-NEXT: s_addk_i32 s4, 0x300
; SI-NEXT: s_and_b32 s4, s4, 0xffff
; SI-NEXT: v_readlane_b32 s5, v62, 37
+; SI-NEXT: v_readlane_b32 s6, v62, 34
; SI-NEXT: s_lshl_b32 s5, s5, 8
-; SI-NEXT: buffer_load_dword v12, off, s[0:3], s32 offset:408 ; 4-byte Folded Reload
-; SI-NEXT: s_add_i32 s71, s22, 3
+; SI-NEXT: s_add_i32 s71, s6, 3
; SI-NEXT: s_and_b32 s8, s71, 0xff
; SI-NEXT: s_lshl_b32 s8, s8, 16
; SI-NEXT: s_add_i32 s35, s99, 3
-; SI-NEXT: s_waitcnt vmcnt(1)
-; SI-NEXT: v_add_i32_e32 v11, vcc, 3, v5
-; SI-NEXT: buffer_load_dword v5, off, s[0:3], s32 offset:412 ; 4-byte Folded Reload
-; SI-NEXT: v_and_b32_e32 v11, 0xff, v11
-; SI-NEXT: v_lshlrev_b32_e32 v11, 16, v11
-; SI-NEXT: v_or_b32_e32 v11, v32, v11
-; SI-NEXT: v_or_b32_e32 v11, s4, v11
+; SI-NEXT: s_and_b32 s6, s35, 0xff
+; SI-NEXT: s_lshl_b32 s6, s6, 16
+; SI-NEXT: v_add_i32_e32 v13, vcc, 3, v13
+; SI-NEXT: v_and_b32_e32 v13, 0xff, v13
+; SI-NEXT: v_lshlrev_b32_e32 v13, 16, v13
+; SI-NEXT: v_mov_b32_e32 v31, s9
+; SI-NEXT: v_mov_b32_e32 v39, s70
+; SI-NEXT: v_add_i32_e32 v36, vcc, 0x3000000, v2
+; SI-NEXT: v_mov_b32_e32 v28, s39
+; SI-NEXT: v_add_i32_e32 v4, vcc, 0x3000000, v4
+; SI-NEXT: v_mov_b32_e32 v27, s48
+; SI-NEXT: v_mov_b32_e32 v26, s49
+; SI-NEXT: v_mov_b32_e32 v25, s94
+; SI-NEXT: v_mov_b32_e32 v24, s53
+; SI-NEXT: v_mov_b32_e32 v23, s36
+; SI-NEXT: v_alignbit_b32 v42, v23, v10, 16
+; SI-NEXT: v_alignbit_b32 v41, v24, v9, 16
+; SI-NEXT: v_alignbit_b32 v40, v25, v8, 16
+; SI-NEXT: v_alignbit_b32 v55, v26, v7, 16
+; SI-NEXT: v_alignbit_b32 v54, v27, v4, 16
+; SI-NEXT: v_alignbit_b32 v53, v28, v36, 16
+; SI-NEXT: v_alignbit_b32 v52, v39, v1, 16
+; SI-NEXT: v_alignbit_b32 v51, v31, v3, 16
+; SI-NEXT: s_lshr_b32 s57, s36, 16
+; SI-NEXT: s_lshr_b32 s56, s53, 16
+; SI-NEXT: s_lshr_b32 s47, s94, 16
+; SI-NEXT: s_lshr_b32 s46, s49, 16
+; SI-NEXT: s_lshr_b32 s45, s48, 16
+; SI-NEXT: s_lshr_b32 s44, s39, 16
+; SI-NEXT: s_lshr_b32 s43, s70, 16
+; SI-NEXT: s_lshr_b32 s41, s9, 16
+; SI-NEXT: s_waitcnt vmcnt(0)
+; SI-NEXT: v_add_i32_e32 v5, vcc, 3, v5
+; SI-NEXT: v_and_b32_e32 v5, 0xff, v5
+; SI-NEXT: v_lshlrev_b32_e32 v5, 16, v5
+; SI-NEXT: v_or_b32_e32 v5, v22, v5
+; SI-NEXT: v_or_b32_e32 v5, s4, v5
+; SI-NEXT: v_add_i32_e32 v11, vcc, 0x3000000, v5
+; SI-NEXT: buffer_load_dword v5, off, s[0:3], s32 offset:408 ; 4-byte Folded Reload
; SI-NEXT: v_readlane_b32 s4, v62, 40
; SI-NEXT: s_add_i32 s85, s4, 3
; SI-NEXT: s_and_b32 s4, s85, 0xff
; SI-NEXT: s_or_b32 s4, s5, s4
-; SI-NEXT: v_readlane_b32 s5, v62, 33
+; SI-NEXT: v_readlane_b32 s5, v62, 31
; SI-NEXT: s_addk_i32 s4, 0x300
; SI-NEXT: s_lshl_b32 s5, s5, 24
; SI-NEXT: s_and_b32 s4, s4, 0xffff
; SI-NEXT: s_or_b32 s5, s5, s8
; SI-NEXT: s_or_b32 s4, s5, s4
; SI-NEXT: s_add_i32 s8, s4, 0x3000000
-; SI-NEXT: v_readlane_b32 s4, v62, 54
+; SI-NEXT: v_readlane_b32 s4, v62, 53
; SI-NEXT: s_add_i32 s17, s4, 3
-; SI-NEXT: v_readlane_b32 s5, v62, 53
+; SI-NEXT: v_readlane_b32 s5, v62, 52
; SI-NEXT: s_and_b32 s4, s17, 0xff
; SI-NEXT: s_lshl_b32 s5, s5, 8
; SI-NEXT: s_or_b32 s4, s5, s4
@@ -209874,58 +210027,25 @@ define inreg <64 x i16> @bitcast_v128i8_to_v64i16_scalar(<128 x i8> inreg %a, i3
; SI-NEXT: s_and_b32 s4, s4, 0xffff
; SI-NEXT: v_readlane_b32 s5, v62, 47
; SI-NEXT: s_lshl_b32 s5, s5, 8
-; SI-NEXT: s_and_b32 s6, s35, 0xff
-; SI-NEXT: s_lshl_b32 s6, s6, 16
-; SI-NEXT: v_mov_b32_e32 v30, s16
-; SI-NEXT: v_mov_b32_e32 v39, s9
-; SI-NEXT: v_add_i32_e32 v2, vcc, 0x3000000, v2
-; SI-NEXT: v_mov_b32_e32 v28, s11
-; SI-NEXT: v_add_i32_e32 v4, vcc, 0x3000000, v4
-; SI-NEXT: v_mov_b32_e32 v27, s48
-; SI-NEXT: v_add_i32_e32 v7, vcc, 0x3000000, v7
-; SI-NEXT: v_mov_b32_e32 v26, s49
-; SI-NEXT: v_add_i32_e32 v8, vcc, 0x3000000, v8
-; SI-NEXT: v_mov_b32_e32 v25, s94
-; SI-NEXT: v_add_i32_e32 v9, vcc, 0x3000000, v9
-; SI-NEXT: v_mov_b32_e32 v24, s53
-; SI-NEXT: v_add_i32_e32 v10, vcc, 0x3000000, v10
-; SI-NEXT: v_mov_b32_e32 v23, s36
-; SI-NEXT: v_add_i32_e32 v11, vcc, 0x3000000, v11
; SI-NEXT: v_mov_b32_e32 v22, s8
+; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_alignbit_b32 v43, v22, v11, 16
-; SI-NEXT: v_alignbit_b32 v42, v23, v10, 16
-; SI-NEXT: v_alignbit_b32 v41, v24, v9, 16
-; SI-NEXT: v_alignbit_b32 v40, v25, v8, 16
-; SI-NEXT: v_alignbit_b32 v55, v26, v7, 16
-; SI-NEXT: v_alignbit_b32 v54, v27, v4, 16
-; SI-NEXT: v_alignbit_b32 v53, v28, v2, 16
-; SI-NEXT: v_alignbit_b32 v52, v39, v1, 16
-; SI-NEXT: v_alignbit_b32 v51, v30, v3, 16
; SI-NEXT: s_lshr_b32 s58, s8, 16
-; SI-NEXT: s_lshr_b32 s57, s36, 16
-; SI-NEXT: s_lshr_b32 s56, s53, 16
-; SI-NEXT: s_lshr_b32 s47, s94, 16
-; SI-NEXT: s_lshr_b32 s46, s49, 16
-; SI-NEXT: s_lshr_b32 s45, s48, 16
-; SI-NEXT: s_lshr_b32 s44, s11, 16
-; SI-NEXT: s_lshr_b32 s43, s9, 16
-; SI-NEXT: s_lshr_b32 s41, s16, 16
; SI-NEXT: s_waitcnt vmcnt(0)
; SI-NEXT: v_add_i32_e32 v5, vcc, 3, v5
; SI-NEXT: v_and_b32_e32 v5, 0xff, v5
; SI-NEXT: v_lshlrev_b32_e32 v5, 16, v5
-; SI-NEXT: v_or_b32_e32 v5, v12, v5
+; SI-NEXT: v_or_b32_e32 v5, v20, v5
; SI-NEXT: v_or_b32_e32 v5, s4, v5
; SI-NEXT: v_add_i32_e32 v12, vcc, 0x3000000, v5
-; SI-NEXT: v_add_i32_e32 v5, vcc, 3, v13
-; SI-NEXT: v_add_i32_e32 v13, vcc, 3, v6
-; SI-NEXT: buffer_load_dword v6, off, s[0:3], s32 offset:404 ; 4-byte Folded Reload
+; SI-NEXT: v_add_i32_e32 v5, vcc, 3, v14
+; SI-NEXT: buffer_load_dword v14, off, s[0:3], s32 offset:404 ; 4-byte Folded Reload
; SI-NEXT: v_readlane_b32 s4, v62, 48
; SI-NEXT: s_add_i32 s7, s4, 3
; SI-NEXT: s_and_b32 s4, s7, 0xff
; SI-NEXT: s_or_b32 s4, s5, s4
; SI-NEXT: s_addk_i32 s4, 0x300
-; SI-NEXT: s_lshl_b32 s5, s89, 24
+; SI-NEXT: s_lshl_b32 s5, s22, 24
; SI-NEXT: s_and_b32 s4, s4, 0xffff
; SI-NEXT: s_or_b32 s5, s5, s6
; SI-NEXT: s_or_b32 s4, s5, s4
@@ -209940,32 +210060,35 @@ define inreg <64 x i16> @bitcast_v128i8_to_v64i16_scalar(<128 x i8> inreg %a, i3
; SI-NEXT: s_addk_i32 s4, 0x300
; SI-NEXT: v_lshlrev_b32_e32 v5, 16, v5
; SI-NEXT: s_and_b32 s4, s4, 0xffff
-; SI-NEXT: v_or_b32_e32 v5, v19, v5
+; SI-NEXT: v_readlane_b32 s5, v62, 35
+; SI-NEXT: v_readlane_b32 s6, v62, 27
+; SI-NEXT: s_lshl_b32 s5, s5, 8
+; SI-NEXT: s_add_i32 s55, s6, 3
+; SI-NEXT: s_and_b32 s6, s55, 0xff
+; SI-NEXT: s_lshl_b32 s6, s6, 16
+; SI-NEXT: v_mov_b32_e32 v20, s10
+; SI-NEXT: v_alignbit_b32 v44, v20, v12, 16
+; SI-NEXT: s_lshr_b32 s59, s10, 16
+; SI-NEXT: s_waitcnt vmcnt(0)
+; SI-NEXT: v_or_b32_e32 v5, v14, v5
; SI-NEXT: v_or_b32_e32 v5, s4, v5
; SI-NEXT: v_readlane_b32 s4, v62, 36
; SI-NEXT: s_add_i32 s81, s4, 3
-; SI-NEXT: v_readlane_b32 s5, v62, 35
-; SI-NEXT: v_readlane_b32 s6, v62, 28
; SI-NEXT: s_and_b32 s4, s81, 0xff
-; SI-NEXT: s_lshl_b32 s5, s5, 8
-; SI-NEXT: s_add_i32 s55, s6, 3
; SI-NEXT: s_or_b32 s4, s5, s4
-; SI-NEXT: v_readlane_b32 s5, v62, 26
-; SI-NEXT: s_and_b32 s6, s55, 0xff
; SI-NEXT: s_addk_i32 s4, 0x300
-; SI-NEXT: s_lshl_b32 s5, s5, 24
-; SI-NEXT: s_lshl_b32 s6, s6, 16
+; SI-NEXT: s_lshl_b32 s5, s20, 24
; SI-NEXT: s_and_b32 s4, s4, 0xffff
; SI-NEXT: s_or_b32 s5, s5, s6
; SI-NEXT: s_or_b32 s4, s5, s4
; SI-NEXT: s_add_i32 s12, s4, 0x3000000
-; SI-NEXT: v_readlane_b32 s4, v62, 34
+; SI-NEXT: v_readlane_b32 s4, v62, 32
; SI-NEXT: s_add_i32 s69, s4, 3
-; SI-NEXT: v_readlane_b32 s5, v62, 29
+; SI-NEXT: v_readlane_b32 s5, v62, 28
; SI-NEXT: v_add_i32_e32 v15, vcc, 0x3000000, v5
; SI-NEXT: s_and_b32 s4, s69, 0xff
; SI-NEXT: s_lshl_b32 s5, s5, 8
-; SI-NEXT: v_add_i32_e32 v5, vcc, 3, v29
+; SI-NEXT: v_add_i32_e32 v5, vcc, 3, v6
; SI-NEXT: s_or_b32 s4, s5, s4
; SI-NEXT: v_and_b32_e32 v5, 0xff, v5
; SI-NEXT: s_addk_i32 s4, 0x300
@@ -209973,10 +210096,10 @@ define inreg <64 x i16> @bitcast_v128i8_to_v64i16_scalar(<128 x i8> inreg %a, i3
; SI-NEXT: s_and_b32 s4, s4, 0xffff
; SI-NEXT: v_or_b32_e32 v5, v18, v5
; SI-NEXT: v_or_b32_e32 v5, s4, v5
-; SI-NEXT: v_readlane_b32 s4, v62, 22
+; SI-NEXT: v_readlane_b32 s4, v62, 20
; SI-NEXT: s_add_i32 s34, s4, 3
-; SI-NEXT: v_readlane_b32 s5, v62, 21
-; SI-NEXT: v_readlane_b32 s6, v62, 19
+; SI-NEXT: v_readlane_b32 s5, v62, 19
+; SI-NEXT: v_readlane_b32 s6, v62, 17
; SI-NEXT: s_and_b32 s4, s34, 0xff
; SI-NEXT: s_lshl_b32 s5, s5, 8
; SI-NEXT: s_add_i32 s92, s6, 3
@@ -209991,21 +210114,21 @@ define inreg <64 x i16> @bitcast_v128i8_to_v64i16_scalar(<128 x i8> inreg %a, i3
; SI-NEXT: s_add_i32 s13, s4, 0x3000000
; SI-NEXT: v_readlane_b32 s4, v62, 25
; SI-NEXT: s_add_i32 s51, s4, 3
-; SI-NEXT: v_readlane_b32 s5, v62, 24
+; SI-NEXT: v_readlane_b32 s5, v62, 23
; SI-NEXT: v_add_i32_e32 v16, vcc, 0x3000000, v5
; SI-NEXT: s_and_b32 s4, s51, 0xff
; SI-NEXT: s_lshl_b32 s5, s5, 8
-; SI-NEXT: v_add_i32_e32 v5, vcc, 3, v21
+; SI-NEXT: v_add_i32_e32 v5, vcc, 3, v29
; SI-NEXT: s_or_b32 s4, s5, s4
; SI-NEXT: v_and_b32_e32 v5, 0xff, v5
; SI-NEXT: s_addk_i32 s4, 0x300
; SI-NEXT: v_lshlrev_b32_e32 v5, 16, v5
; SI-NEXT: s_and_b32 s4, s4, 0xffff
-; SI-NEXT: v_or_b32_e32 v5, v14, v5
+; SI-NEXT: v_or_b32_e32 v5, v61, v5
; SI-NEXT: v_or_b32_e32 v5, s4, v5
-; SI-NEXT: v_readlane_b32 s4, v62, 20
+; SI-NEXT: v_readlane_b32 s4, v62, 18
; SI-NEXT: s_add_i32 s95, s4, 3
-; SI-NEXT: v_readlane_b32 s5, v62, 18
+; SI-NEXT: v_readlane_b32 s5, v62, 16
; SI-NEXT: s_and_b32 s4, s95, 0xff
; SI-NEXT: s_lshl_b32 s5, s5, 8
; SI-NEXT: s_add_i32 s6, s96, 3
@@ -210022,37 +210145,35 @@ define inreg <64 x i16> @bitcast_v128i8_to_v64i16_scalar(<128 x i8> inreg %a, i3
; SI-NEXT: s_and_b32 s4, s4, 0xff
; SI-NEXT: s_lshl_b32 s5, s84, 8
; SI-NEXT: s_or_b32 s4, s5, s4
-; SI-NEXT: v_and_b32_e32 v13, 0xff, v13
; SI-NEXT: s_addk_i32 s4, 0x300
-; SI-NEXT: v_lshlrev_b32_e32 v13, 16, v13
; SI-NEXT: s_and_b32 s4, s4, 0xffff
-; SI-NEXT: s_waitcnt vmcnt(0)
-; SI-NEXT: v_or_b32_e32 v6, v6, v13
+; SI-NEXT: v_or_b32_e32 v6, v60, v13
; SI-NEXT: v_or_b32_e32 v6, s4, v6
; SI-NEXT: s_add_i32 s4, s83, 3
; SI-NEXT: s_and_b32 s4, s4, 0xff
-; SI-NEXT: s_lshl_b32 s5, s25, 8
+; SI-NEXT: s_lshl_b32 s5, s68, 8
; SI-NEXT: s_add_i32 s6, s64, 3
; SI-NEXT: s_or_b32 s4, s5, s4
-; SI-NEXT: v_readlane_b32 s5, v62, 15
; SI-NEXT: s_and_b32 s6, s6, 0xff
; SI-NEXT: s_addk_i32 s4, 0x300
-; SI-NEXT: s_lshl_b32 s5, s5, 24
+; SI-NEXT: s_lshl_b32 s5, s65, 24
; SI-NEXT: s_lshl_b32 s6, s6, 16
; SI-NEXT: s_and_b32 s4, s4, 0xffff
; SI-NEXT: s_or_b32 s5, s5, s6
; SI-NEXT: s_or_b32 s4, s5, s4
; SI-NEXT: s_add_i32 s15, s4, 0x3000000
-; SI-NEXT: v_readlane_b32 s4, v62, 2
+; SI-NEXT: v_readlane_b32 s4, v62, 4
; SI-NEXT: s_add_i32 s4, s4, 3
-; SI-NEXT: v_readlane_b32 s5, v62, 1
+; SI-NEXT: v_readlane_b32 s5, v62, 3
+; SI-NEXT: v_readlane_b32 s6, v62, 2
; SI-NEXT: s_and_b32 s4, s4, 0xff
; SI-NEXT: s_lshl_b32 s5, s5, 8
-; SI-NEXT: s_add_i32 s6, s26, 3
+; SI-NEXT: s_add_i32 s6, s6, 3
; SI-NEXT: s_or_b32 s4, s5, s4
+; SI-NEXT: v_readlane_b32 s5, v62, 1
; SI-NEXT: s_and_b32 s6, s6, 0xff
; SI-NEXT: s_addk_i32 s4, 0x300
-; SI-NEXT: s_lshl_b32 s5, s27, 24
+; SI-NEXT: s_lshl_b32 s5, s5, 24
; SI-NEXT: s_lshl_b32 s6, s6, 16
; SI-NEXT: s_and_b32 s4, s4, 0xffff
; SI-NEXT: s_or_b32 s5, s5, s6
@@ -210062,21 +210183,20 @@ define inreg <64 x i16> @bitcast_v128i8_to_v64i16_scalar(<128 x i8> inreg %a, i3
; SI-NEXT: s_add_i32 s4, s4, 3
; SI-NEXT: s_and_b32 s4, s4, 0xff
; SI-NEXT: s_lshl_b32 s5, s29, 8
-; SI-NEXT: s_add_i32 s6, s76, 3
+; SI-NEXT: s_add_i32 s6, s11, 3
; SI-NEXT: s_or_b32 s4, s5, s4
-; SI-NEXT: v_readlane_b32 s5, v62, 16
; SI-NEXT: s_and_b32 s6, s6, 0xff
; SI-NEXT: s_addk_i32 s4, 0x300
-; SI-NEXT: s_lshl_b32 s5, s5, 24
+; SI-NEXT: s_lshl_b32 s5, s90, 24
; SI-NEXT: s_lshl_b32 s6, s6, 16
; SI-NEXT: s_and_b32 s4, s4, 0xffff
; SI-NEXT: s_or_b32 s5, s5, s6
; SI-NEXT: s_or_b32 s4, s5, s4
; SI-NEXT: s_add_i32 s40, s4, 0x3000000
-; SI-NEXT: v_readlane_b32 s4, v62, 7
+; SI-NEXT: v_readlane_b32 s4, v62, 10
; SI-NEXT: s_add_i32 s4, s4, 3
-; SI-NEXT: v_readlane_b32 s5, v62, 17
-; SI-NEXT: v_readlane_b32 s6, v62, 6
+; SI-NEXT: v_readlane_b32 s5, v62, 15
+; SI-NEXT: v_readlane_b32 s6, v62, 9
; SI-NEXT: s_and_b32 s4, s4, 0xff
; SI-NEXT: s_lshl_b32 s5, s5, 8
; SI-NEXT: s_add_i32 s6, s6, 3
@@ -210088,15 +210208,16 @@ define inreg <64 x i16> @bitcast_v128i8_to_v64i16_scalar(<128 x i8> inreg %a, i3
; SI-NEXT: s_and_b32 s4, s4, 0xffff
; SI-NEXT: s_or_b32 s5, s5, s6
; SI-NEXT: s_or_b32 s4, s5, s4
-; SI-NEXT: s_add_i32 s18, s4, 0x3000000
-; SI-NEXT: s_add_i32 s4, s20, 3
-; SI-NEXT: v_readlane_b32 s5, v62, 5
-; SI-NEXT: v_readlane_b32 s6, v62, 4
+; SI-NEXT: s_add_i32 s21, s4, 0x3000000
+; SI-NEXT: v_readlane_b32 s4, v62, 8
+; SI-NEXT: s_add_i32 s4, s4, 3
+; SI-NEXT: v_readlane_b32 s5, v62, 7
+; SI-NEXT: v_readlane_b32 s6, v62, 6
; SI-NEXT: s_and_b32 s4, s4, 0xff
; SI-NEXT: s_lshl_b32 s5, s5, 8
; SI-NEXT: s_add_i32 s6, s6, 3
; SI-NEXT: s_or_b32 s4, s5, s4
-; SI-NEXT: v_readlane_b32 s5, v62, 3
+; SI-NEXT: v_readlane_b32 s5, v62, 5
; SI-NEXT: s_and_b32 s6, s6, 0xff
; SI-NEXT: s_addk_i32 s4, 0x300
; SI-NEXT: s_lshl_b32 s5, s5, 24
@@ -210105,8 +210226,7 @@ define inreg <64 x i16> @bitcast_v128i8_to_v64i16_scalar(<128 x i8> inreg %a, i3
; SI-NEXT: s_or_b32 s5, s5, s6
; SI-NEXT: s_or_b32 s4, s5, s4
; SI-NEXT: s_add_i32 s42, s4, 0x3000000
-; SI-NEXT: v_mov_b32_e32 v13, s18
-; SI-NEXT: v_mov_b32_e32 v20, s10
+; SI-NEXT: v_mov_b32_e32 v13, s21
; SI-NEXT: v_mov_b32_e32 v19, s12
; SI-NEXT: v_mov_b32_e32 v18, s13
; SI-NEXT: v_add_i32_e32 v17, vcc, 0x3000000, v5
@@ -210118,20 +210238,16 @@ define inreg <64 x i16> @bitcast_v128i8_to_v64i16_scalar(<128 x i8> inreg %a, i3
; SI-NEXT: v_alignbit_b32 v58, s40, v13, 16
; SI-NEXT: v_alignbit_b32 v56, v6, v50, 16
; SI-NEXT: v_alignbit_b32 v47, v5, v17, 16
-; SI-NEXT: s_waitcnt expcnt(1)
; SI-NEXT: v_alignbit_b32 v46, v18, v16, 16
; SI-NEXT: v_alignbit_b32 v45, v19, v15, 16
-; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_alignbit_b32 v44, v20, v12, 16
; SI-NEXT: s_lshr_b32 s73, s42, 16
; SI-NEXT: s_lshr_b32 s72, s40, 16
; SI-NEXT: s_lshr_b32 s63, s15, 16
; SI-NEXT: s_lshr_b32 s62, s14, 16
; SI-NEXT: s_lshr_b32 s61, s13, 16
; SI-NEXT: s_lshr_b32 s60, s12, 16
-; SI-NEXT: s_lshr_b32 s59, s10, 16
; SI-NEXT: .LBB97_3: ; %end
-; SI-NEXT: s_and_b32 s4, s18, 0xffff
+; SI-NEXT: s_and_b32 s4, s21, 0xffff
; SI-NEXT: v_lshlrev_b32_e32 v5, 16, v57
; SI-NEXT: v_or_b32_e32 v5, s4, v5
; SI-NEXT: s_and_b32 s4, s42, 0xffff
@@ -210140,6 +210256,7 @@ define inreg <64 x i16> @bitcast_v128i8_to_v64i16_scalar(<128 x i8> inreg %a, i3
; SI-NEXT: buffer_store_dword v5, v0, s[0:3], 0 offen
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_add_i32_e32 v5, vcc, 4, v0
+; SI-NEXT: s_waitcnt vmcnt(2)
; SI-NEXT: v_mov_b32_e32 v6, s4
; SI-NEXT: buffer_store_dword v6, v5, s[0:3], 0 offen
; SI-NEXT: s_and_b32 s4, s17, 0xffff
@@ -210297,9 +210414,9 @@ define inreg <64 x i16> @bitcast_v128i8_to_v64i16_scalar(<128 x i8> inreg %a, i3
; SI-NEXT: v_add_i32_e32 v4, vcc, 0x64, v0
; SI-NEXT: v_mov_b32_e32 v5, s4
; SI-NEXT: buffer_store_dword v5, v4, s[0:3], 0 offen
-; SI-NEXT: v_and_b32_e32 v2, 0xffff, v2
+; SI-NEXT: v_and_b32_e32 v2, 0xffff, v36
; SI-NEXT: v_lshlrev_b32_e32 v4, 16, v53
-; SI-NEXT: s_and_b32 s4, s11, 0xffff
+; SI-NEXT: s_and_b32 s4, s39, 0xffff
; SI-NEXT: s_lshl_b32 s5, s44, 16
; SI-NEXT: v_or_b32_e32 v2, v2, v4
; SI-NEXT: v_add_i32_e32 v4, vcc, 0x68, v0
@@ -210311,7 +210428,7 @@ define inreg <64 x i16> @bitcast_v128i8_to_v64i16_scalar(<128 x i8> inreg %a, i3
; SI-NEXT: buffer_store_dword v4, v2, s[0:3], 0 offen
; SI-NEXT: v_and_b32_e32 v1, 0xffff, v1
; SI-NEXT: v_lshlrev_b32_e32 v2, 16, v52
-; SI-NEXT: s_and_b32 s4, s9, 0xffff
+; SI-NEXT: s_and_b32 s4, s70, 0xffff
; SI-NEXT: s_lshl_b32 s5, s43, 16
; SI-NEXT: v_or_b32_e32 v1, v1, v2
; SI-NEXT: v_add_i32_e32 v2, vcc, 0x70, v0
@@ -210324,7 +210441,7 @@ define inreg <64 x i16> @bitcast_v128i8_to_v64i16_scalar(<128 x i8> inreg %a, i3
; SI-NEXT: v_and_b32_e32 v1, 0xffff, v3
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_lshlrev_b32_e32 v2, 16, v51
-; SI-NEXT: s_and_b32 s4, s16, 0xffff
+; SI-NEXT: s_and_b32 s4, s9, 0xffff
; SI-NEXT: s_lshl_b32 s5, s41, 16
; SI-NEXT: v_or_b32_e32 v1, v1, v2
; SI-NEXT: v_add_i32_e32 v2, vcc, 0x78, v0
@@ -210385,27 +210502,26 @@ define inreg <64 x i16> @bitcast_v128i8_to_v64i16_scalar(<128 x i8> inreg %a, i3
; SI-NEXT: v_readlane_b32 s31, v63, 1
; SI-NEXT: v_readlane_b32 s30, v63, 0
; SI-NEXT: s_or_saveexec_b64 s[4:5], -1
-; SI-NEXT: buffer_load_dword v63, off, s[0:3], s32 offset:428 ; 4-byte Folded Reload
-; SI-NEXT: buffer_load_dword v62, off, s[0:3], s32 offset:432 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v63, off, s[0:3], s32 offset:424 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v62, off, s[0:3], s32 offset:428 ; 4-byte Folded Reload
; SI-NEXT: s_mov_b64 exec, s[4:5]
; SI-NEXT: s_waitcnt vmcnt(0) expcnt(0)
; SI-NEXT: s_setpc_b64 s[30:31]
; SI-NEXT: .LBB97_4:
-; SI-NEXT: v_mov_b32_e32 v5, v13
-; SI-NEXT: buffer_load_dword v29, off, s[0:3], s32 offset:400 ; 4-byte Folded Reload
-; SI-NEXT: buffer_load_dword v13, off, s[0:3], s32 offset:396 ; 4-byte Folded Reload
-; SI-NEXT: v_mov_b32_e32 v25, v58
-; SI-NEXT: v_mov_b32_e32 v48, v39
-; SI-NEXT: v_mov_b32_e32 v39, v57
-; SI-NEXT: v_mov_b32_e32 v49, v56
-; SI-NEXT: v_mov_b32_e32 v20, v47
-; SI-NEXT: v_mov_b32_e32 v30, v37
-; SI-NEXT: v_mov_b32_e32 v36, v35
-; SI-NEXT: v_mov_b32_e32 v35, v45
-; SI-NEXT: v_mov_b32_e32 v27, v26
-; SI-NEXT: s_waitcnt expcnt(2)
-; SI-NEXT: v_mov_b32_e32 v32, v23
-; SI-NEXT: ; implicit-def: $sgpr18
+; SI-NEXT: s_waitcnt expcnt(4)
+; SI-NEXT: buffer_load_dword v6, off, s[0:3], s32 offset:400 ; 4-byte Folded Reload
+; SI-NEXT: s_waitcnt expcnt(1)
+; SI-NEXT: buffer_load_dword v14, off, s[0:3], s32 offset:396 ; 4-byte Folded Reload
+; SI-NEXT: v_mov_b32_e32 v2, v58
+; SI-NEXT: v_mov_b32_e32 v49, v57
+; SI-NEXT: v_mov_b32_e32 v30, v56
+; SI-NEXT: v_mov_b32_e32 v48, v46
+; SI-NEXT: v_mov_b32_e32 v24, v45
+; SI-NEXT: v_mov_b32_e32 v28, v36
+; SI-NEXT: v_mov_b32_e32 v27, v35
+; SI-NEXT: v_mov_b32_e32 v26, v34
+; SI-NEXT: v_mov_b32_e32 v13, v21
+; SI-NEXT: ; implicit-def: $sgpr21
; SI-NEXT: ; implicit-def: $vgpr57
; SI-NEXT: ; implicit-def: $sgpr42
; SI-NEXT: ; implicit-def: $sgpr73
@@ -210457,17 +210573,17 @@ define inreg <64 x i16> @bitcast_v128i8_to_v64i16_scalar(<128 x i8> inreg %a, i3
; SI-NEXT: ; implicit-def: $vgpr54
; SI-NEXT: ; implicit-def: $sgpr48
; SI-NEXT: ; implicit-def: $sgpr45
-; SI-NEXT: ; implicit-def: $vgpr2
+; SI-NEXT: ; implicit-def: $vgpr36
; SI-NEXT: ; implicit-def: $vgpr53
-; SI-NEXT: ; implicit-def: $sgpr11
+; SI-NEXT: ; implicit-def: $sgpr39
; SI-NEXT: ; implicit-def: $sgpr44
; SI-NEXT: ; implicit-def: $vgpr1
; SI-NEXT: ; implicit-def: $vgpr52
-; SI-NEXT: ; implicit-def: $sgpr9
+; SI-NEXT: ; implicit-def: $sgpr70
; SI-NEXT: ; implicit-def: $sgpr43
; SI-NEXT: ; implicit-def: $vgpr3
; SI-NEXT: ; implicit-def: $vgpr51
-; SI-NEXT: ; implicit-def: $sgpr16
+; SI-NEXT: ; implicit-def: $sgpr9
; SI-NEXT: ; implicit-def: $sgpr41
; SI-NEXT: s_branch .LBB97_2
;
@@ -210531,13 +210647,14 @@ define inreg <64 x i16> @bitcast_v128i8_to_v64i16_scalar(<128 x i8> inreg %a, i3
; VI-NEXT: buffer_load_ushort v37, off, s[0:3], s32 offset:168
; VI-NEXT: buffer_load_ushort v38, off, s[0:3], s32 offset:176
; VI-NEXT: v_lshlrev_b32_e32 v1, 8, v1
-; VI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:508 ; 4-byte Folded Spill
+; VI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:504 ; 4-byte Folded Spill
; VI-NEXT: v_lshlrev_b32_e32 v1, 8, v3
; VI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:464 ; 4-byte Folded Spill
; VI-NEXT: v_lshlrev_b32_e32 v1, 8, v5
; VI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:468 ; 4-byte Folded Spill
; VI-NEXT: v_lshlrev_b32_e32 v1, 8, v7
; VI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:472 ; 4-byte Folded Spill
+; VI-NEXT: v_lshlrev_b32_e32 v40, 8, v27
; VI-NEXT: v_lshlrev_b32_e32 v9, 8, v9
; VI-NEXT: v_lshlrev_b32_e32 v11, 8, v11
; VI-NEXT: v_lshlrev_b32_e32 v13, 8, v13
@@ -210549,46 +210666,42 @@ define inreg <64 x i16> @bitcast_v128i8_to_v64i16_scalar(<128 x i8> inreg %a, i3
; VI-NEXT: v_lshlrev_b32_e32 v25, 8, v25
; VI-NEXT: s_waitcnt vmcnt(14)
; VI-NEXT: v_cmp_ne_u32_e32 vcc, 0, v0
-; VI-NEXT: v_lshlrev_b32_e32 v0, 8, v27
-; VI-NEXT: buffer_store_dword v0, off, s[0:3], s32 offset:476 ; 4-byte Folded Spill
; VI-NEXT: v_lshlrev_b32_e32 v0, 8, v29
-; VI-NEXT: buffer_store_dword v0, off, s[0:3], s32 offset:480 ; 4-byte Folded Spill
+; VI-NEXT: buffer_store_dword v0, off, s[0:3], s32 offset:476 ; 4-byte Folded Spill
; VI-NEXT: v_lshlrev_b32_e32 v0, 8, v2
-; VI-NEXT: buffer_store_dword v0, off, s[0:3], s32 offset:484 ; 4-byte Folded Spill
+; VI-NEXT: buffer_store_dword v0, off, s[0:3], s32 offset:480 ; 4-byte Folded Spill
; VI-NEXT: v_lshlrev_b32_e32 v0, 8, v4
-; VI-NEXT: buffer_store_dword v0, off, s[0:3], s32 offset:488 ; 4-byte Folded Spill
+; VI-NEXT: buffer_store_dword v0, off, s[0:3], s32 offset:484 ; 4-byte Folded Spill
; VI-NEXT: v_lshlrev_b32_e32 v0, 8, v6
-; VI-NEXT: buffer_store_dword v0, off, s[0:3], s32 offset:492 ; 4-byte Folded Spill
+; VI-NEXT: buffer_store_dword v0, off, s[0:3], s32 offset:488 ; 4-byte Folded Spill
; VI-NEXT: v_lshlrev_b32_e32 v0, 8, v8
-; VI-NEXT: buffer_store_dword v0, off, s[0:3], s32 offset:496 ; 4-byte Folded Spill
+; VI-NEXT: buffer_store_dword v0, off, s[0:3], s32 offset:492 ; 4-byte Folded Spill
; VI-NEXT: v_lshlrev_b32_e32 v0, 8, v10
-; VI-NEXT: buffer_store_dword v0, off, s[0:3], s32 offset:500 ; 4-byte Folded Spill
+; VI-NEXT: buffer_store_dword v0, off, s[0:3], s32 offset:496 ; 4-byte Folded Spill
; VI-NEXT: v_lshlrev_b32_e32 v0, 8, v12
-; VI-NEXT: buffer_store_dword v0, off, s[0:3], s32 offset:504 ; 4-byte Folded Spill
-; VI-NEXT: v_lshlrev_b32_e32 v0, 8, v26
-; VI-NEXT: buffer_store_dword v0, off, s[0:3], s32 offset:512 ; 4-byte Folded Spill
+; VI-NEXT: buffer_store_dword v0, off, s[0:3], s32 offset:500 ; 4-byte Folded Spill
+; VI-NEXT: v_lshlrev_b32_e32 v0, 8, v24
+; VI-NEXT: buffer_store_dword v0, off, s[0:3], s32 offset:508 ; 4-byte Folded Spill
; VI-NEXT: s_waitcnt vmcnt(14)
-; VI-NEXT: v_lshlrev_b32_e32 v0, 8, v28
-; VI-NEXT: buffer_store_dword v0, off, s[0:3], s32 offset:516 ; 4-byte Folded Spill
; VI-NEXT: v_lshlrev_b32_e32 v0, 8, v30
-; VI-NEXT: buffer_store_dword v0, off, s[0:3], s32 offset:520 ; 4-byte Folded Spill
+; VI-NEXT: buffer_store_dword v0, off, s[0:3], s32 offset:512 ; 4-byte Folded Spill
; VI-NEXT: v_lshlrev_b32_e32 v0, 8, v31
-; VI-NEXT: buffer_store_dword v0, off, s[0:3], s32 offset:560 ; 4-byte Folded Spill
+; VI-NEXT: buffer_store_dword v0, off, s[0:3], s32 offset:556 ; 4-byte Folded Spill
; VI-NEXT: v_lshlrev_b32_e32 v0, 8, v32
-; VI-NEXT: buffer_store_dword v0, off, s[0:3], s32 offset:564 ; 4-byte Folded Spill
-; VI-NEXT: v_lshlrev_b32_e32 v0, 8, v33
-; VI-NEXT: buffer_store_dword v0, off, s[0:3], s32 offset:592 ; 4-byte Folded Spill
-; VI-NEXT: v_lshlrev_b32_e32 v0, 8, v34
; VI-NEXT: buffer_store_dword v0, off, s[0:3], s32 offset:588 ; 4-byte Folded Spill
-; VI-NEXT: v_lshlrev_b32_e32 v0, 8, v35
+; VI-NEXT: v_lshlrev_b32_e32 v0, 8, v33
; VI-NEXT: buffer_store_dword v0, off, s[0:3], s32 offset:584 ; 4-byte Folded Spill
-; VI-NEXT: v_lshlrev_b32_e32 v0, 8, v36
+; VI-NEXT: v_lshlrev_b32_e32 v0, 8, v34
; VI-NEXT: buffer_store_dword v0, off, s[0:3], s32 offset:576 ; 4-byte Folded Spill
-; VI-NEXT: v_lshlrev_b32_e32 v0, 8, v37
+; VI-NEXT: v_lshlrev_b32_e32 v0, 8, v35
; VI-NEXT: buffer_store_dword v0, off, s[0:3], s32 offset:572 ; 4-byte Folded Spill
+; VI-NEXT: v_lshlrev_b32_e32 v0, 8, v36
+; VI-NEXT: buffer_store_dword v0, off, s[0:3], s32 offset:564 ; 4-byte Folded Spill
; VI-NEXT: s_waitcnt vmcnt(14)
+; VI-NEXT: v_lshlrev_b32_e32 v0, 8, v37
+; VI-NEXT: buffer_store_dword v0, off, s[0:3], s32 offset:560 ; 4-byte Folded Spill
; VI-NEXT: v_lshlrev_b32_e32 v0, 8, v38
-; VI-NEXT: buffer_store_dword v0, off, s[0:3], s32 offset:580 ; 4-byte Folded Spill
+; VI-NEXT: buffer_store_dword v0, off, s[0:3], s32 offset:568 ; 4-byte Folded Spill
; VI-NEXT: buffer_load_ushort v0, off, s[0:3], s32 offset:184
; VI-NEXT: buffer_load_ushort v1, off, s[0:3], s32 offset:192
; VI-NEXT: buffer_load_ushort v2, off, s[0:3], s32 offset:200
@@ -210597,34 +210710,37 @@ define inreg <64 x i16> @bitcast_v128i8_to_v64i16_scalar(<128 x i8> inreg %a, i3
; VI-NEXT: buffer_load_ushort v5, off, s[0:3], s32 offset:224
; VI-NEXT: buffer_load_ushort v6, off, s[0:3], s32 offset:232
; VI-NEXT: buffer_load_ushort v7, off, s[0:3], s32 offset:240
-; VI-NEXT: v_lshlrev_b32_e32 v45, 8, v22
-; VI-NEXT: v_lshlrev_b32_e32 v8, 8, v24
+; VI-NEXT: v_lshlrev_b32_e32 v8, 8, v26
+; VI-NEXT: v_lshlrev_b32_e32 v10, 8, v28
; VI-NEXT: s_and_b64 s[4:5], vcc, exec
; VI-NEXT: v_lshlrev_b32_e32 v14, 8, v14
; VI-NEXT: v_lshlrev_b32_e32 v16, 8, v16
; VI-NEXT: v_lshlrev_b32_e32 v18, 8, v18
; VI-NEXT: v_lshlrev_b32_e32 v20, 8, v20
+; VI-NEXT: v_lshlrev_b32_e32 v22, 8, v22
; VI-NEXT: s_waitcnt vmcnt(7)
-; VI-NEXT: v_lshlrev_b32_e32 v22, 8, v0
+; VI-NEXT: v_lshlrev_b32_e32 v24, 8, v0
; VI-NEXT: s_waitcnt vmcnt(6)
; VI-NEXT: v_lshlrev_b32_e32 v0, 8, v1
-; VI-NEXT: buffer_store_dword v0, off, s[0:3], s32 offset:568 ; 4-byte Folded Spill
-; VI-NEXT: s_waitcnt vmcnt(5)
+; VI-NEXT: buffer_store_dword v0, off, s[0:3], s32 offset:592 ; 4-byte Folded Spill
+; VI-NEXT: s_waitcnt vmcnt(6)
+; VI-NEXT: v_lshlrev_b32_e32 v0, 8, v2
+; VI-NEXT: buffer_store_dword v0, off, s[0:3], s32 offset:580 ; 4-byte Folded Spill
+; VI-NEXT: s_waitcnt vmcnt(6)
; VI-NEXT: v_lshlrev_b32_e32 v0, 8, v3
; VI-NEXT: buffer_store_dword v0, off, s[0:3], s32 offset:600 ; 4-byte Folded Spill
-; VI-NEXT: s_waitcnt vmcnt(5)
+; VI-NEXT: s_waitcnt vmcnt(6)
; VI-NEXT: v_lshlrev_b32_e32 v0, 8, v4
; VI-NEXT: buffer_store_dword v0, off, s[0:3], s32 offset:596 ; 4-byte Folded Spill
-; VI-NEXT: s_waitcnt vmcnt(5)
+; VI-NEXT: s_waitcnt vmcnt(6)
; VI-NEXT: v_lshlrev_b32_e32 v0, 8, v5
; VI-NEXT: buffer_store_dword v0, off, s[0:3], s32 offset:608 ; 4-byte Folded Spill
-; VI-NEXT: s_waitcnt vmcnt(5)
+; VI-NEXT: s_waitcnt vmcnt(6)
; VI-NEXT: v_lshlrev_b32_e32 v0, 8, v6
; VI-NEXT: buffer_store_dword v0, off, s[0:3], s32 offset:612 ; 4-byte Folded Spill
-; VI-NEXT: s_waitcnt vmcnt(5)
+; VI-NEXT: s_waitcnt vmcnt(6)
; VI-NEXT: v_lshlrev_b32_e32 v0, 8, v7
; VI-NEXT: buffer_store_dword v0, off, s[0:3], s32 offset:604 ; 4-byte Folded Spill
-; VI-NEXT: v_lshlrev_b32_e32 v24, 8, v2
; VI-NEXT: buffer_load_ushort v0, off, s[0:3], s32 offset:248
; VI-NEXT: buffer_load_ushort v1, off, s[0:3], s32 offset:256
; VI-NEXT: buffer_load_ushort v2, off, s[0:3], s32 offset:264
@@ -210643,6 +210759,11 @@ define inreg <64 x i16> @bitcast_v128i8_to_v64i16_scalar(<128 x i8> inreg %a, i3
; VI-NEXT: s_waitcnt vmcnt(4)
; VI-NEXT: v_lshlrev_b32_e32 v0, 8, v4
; VI-NEXT: buffer_store_dword v0, off, s[0:3], s32 offset:620 ; 4-byte Folded Spill
+; VI-NEXT: s_waitcnt vmcnt(3)
+; VI-NEXT: v_lshlrev_b32_e32 v4, 8, v6
+; VI-NEXT: s_waitcnt vmcnt(2)
+; VI-NEXT: v_lshlrev_b32_e32 v0, 8, v7
+; VI-NEXT: buffer_store_dword v0, off, s[0:3], s32 offset:544 ; 4-byte Folded Spill
; VI-NEXT: buffer_load_ushort v0, off, s[0:3], s32 offset:312
; VI-NEXT: buffer_load_ushort v1, off, s[0:3], s32 offset:320
; VI-NEXT: buffer_load_ushort v2, off, s[0:3], s32 offset:328
@@ -210651,12 +210772,8 @@ define inreg <64 x i16> @bitcast_v128i8_to_v64i16_scalar(<128 x i8> inreg %a, i3
; VI-NEXT: buffer_load_ushort v33, off, s[0:3], s32 offset:20
; VI-NEXT: buffer_load_ushort v34, off, s[0:3], s32 offset:28
; VI-NEXT: buffer_load_ushort v35, off, s[0:3], s32 offset:36
-; VI-NEXT: s_waitcnt vmcnt(11)
-; VI-NEXT: v_lshlrev_b32_e32 v4, 8, v6
; VI-NEXT: v_lshlrev_b32_e32 v28, 8, v3
; VI-NEXT: v_lshlrev_b32_e32 v3, 8, v5
-; VI-NEXT: s_waitcnt vmcnt(10)
-; VI-NEXT: v_lshlrev_b32_e32 v5, 8, v7
; VI-NEXT: s_waitcnt vmcnt(7)
; VI-NEXT: v_lshlrev_b32_e32 v6, 8, v0
; VI-NEXT: buffer_load_ushort v36, off, s[0:3], s32 offset:44
@@ -210665,47 +210782,45 @@ define inreg <64 x i16> @bitcast_v128i8_to_v64i16_scalar(<128 x i8> inreg %a, i3
; VI-NEXT: buffer_load_ushort v39, off, s[0:3], s32 offset:68
; VI-NEXT: buffer_load_ushort v48, off, s[0:3], s32 offset:76
; VI-NEXT: buffer_load_ushort v49, off, s[0:3], s32 offset:84
-; VI-NEXT: buffer_load_ushort v50, off, s[0:3], s32 offset:92
-; VI-NEXT: buffer_load_ushort v51, off, s[0:3], s32 offset:100
-; VI-NEXT: buffer_load_ushort v54, off, s[0:3], s32 offset:108
+; VI-NEXT: buffer_load_ushort v51, off, s[0:3], s32 offset:92
+; VI-NEXT: buffer_load_ushort v53, off, s[0:3], s32 offset:100
+; VI-NEXT: buffer_load_ushort v55, off, s[0:3], s32 offset:108
; VI-NEXT: buffer_load_ushort v41, off, s[0:3], s32 offset:116
; VI-NEXT: buffer_load_ushort v56, off, s[0:3], s32 offset:124
-; VI-NEXT: buffer_load_ushort v55, off, s[0:3], s32 offset:132
-; VI-NEXT: buffer_load_ushort v52, off, s[0:3], s32 offset:140
-; VI-NEXT: buffer_load_ushort v43, off, s[0:3], s32 offset:148
-; VI-NEXT: buffer_load_ushort v53, off, s[0:3], s32 offset:156
-; VI-NEXT: buffer_load_ushort v46, off, s[0:3], s32 offset:164
-; VI-NEXT: buffer_load_ushort v58, off, s[0:3], s32 offset:172
+; VI-NEXT: buffer_load_ushort v50, off, s[0:3], s32 offset:132
+; VI-NEXT: buffer_load_ushort v43, off, s[0:3], s32 offset:140
+; VI-NEXT: buffer_load_ushort v58, off, s[0:3], s32 offset:148
+; VI-NEXT: buffer_load_ushort v52, off, s[0:3], s32 offset:156
+; VI-NEXT: buffer_load_ushort v60, off, s[0:3], s32 offset:164
+; VI-NEXT: buffer_load_ushort v54, off, s[0:3], s32 offset:172
; VI-NEXT: buffer_load_ushort v47, off, s[0:3], s32 offset:180
; VI-NEXT: buffer_load_ushort v63, off, s[0:3], s32 offset:188
-; VI-NEXT: buffer_load_ushort v57, off, s[0:3], s32 offset:196
+; VI-NEXT: buffer_load_ushort v46, off, s[0:3], s32 offset:196
; VI-NEXT: buffer_load_ushort v61, off, s[0:3], s32 offset:204
-; VI-NEXT: buffer_load_ushort v60, off, s[0:3], s32 offset:212
+; VI-NEXT: buffer_load_ushort v57, off, s[0:3], s32 offset:212
; VI-NEXT: buffer_load_ushort v59, off, s[0:3], s32 offset:220
-; VI-NEXT: buffer_load_ushort v40, off, s[0:3], s32 offset:228
-; VI-NEXT: buffer_load_ushort v42, off, s[0:3], s32 offset:236
-; VI-NEXT: buffer_load_ushort v62, off, s[0:3], s32 offset:244
-; VI-NEXT: buffer_load_ushort v44, off, s[0:3], s32 offset:252
+; VI-NEXT: buffer_load_ushort v42, off, s[0:3], s32 offset:228
+; VI-NEXT: buffer_load_ushort v44, off, s[0:3], s32 offset:236
+; VI-NEXT: buffer_load_ushort v45, off, s[0:3], s32 offset:244
+; VI-NEXT: buffer_load_ushort v62, off, s[0:3], s32 offset:252
; VI-NEXT: buffer_load_ushort v0, off, s[0:3], s32 offset:260
; VI-NEXT: s_waitcnt vmcnt(14)
-; VI-NEXT: v_lshlrev_b32_e32 v1, 8, v1
-; VI-NEXT: v_lshlrev_b32_e32 v2, 8, v2
+; VI-NEXT: v_lshlrev_b32_e32 v5, 8, v1
+; VI-NEXT: v_lshlrev_b32_e32 v1, 8, v2
; VI-NEXT: s_waitcnt vmcnt(0)
-; VI-NEXT: buffer_store_dword v0, off, s[0:3], s32 offset:552 ; 4-byte Folded Spill
-; VI-NEXT: buffer_load_ushort v29, off, s[0:3], s32 offset:268
-; VI-NEXT: buffer_load_ushort v0, off, s[0:3], s32 offset:276
+; VI-NEXT: buffer_store_dword v0, off, s[0:3], s32 offset:548 ; 4-byte Folded Spill
+; VI-NEXT: buffer_load_ushort v0, off, s[0:3], s32 offset:268
; VI-NEXT: s_waitcnt vmcnt(0)
-; VI-NEXT: buffer_store_dword v0, off, s[0:3], s32 offset:556 ; 4-byte Folded Spill
-; VI-NEXT: buffer_load_ushort v30, off, s[0:3], s32 offset:284
-; VI-NEXT: buffer_load_ushort v31, off, s[0:3], s32 offset:292
-; VI-NEXT: buffer_load_ushort v0, off, s[0:3], s32 offset:300
-; VI-NEXT: buffer_load_ushort v7, off, s[0:3], s32 offset:308
-; VI-NEXT: s_waitcnt vmcnt(0)
-; VI-NEXT: buffer_store_dword v7, off, s[0:3], s32 offset:528 ; 4-byte Folded Spill
-; VI-NEXT: buffer_load_ushort v7, off, s[0:3], s32 offset:316
-; VI-NEXT: buffer_load_ushort v10, off, s[0:3], s32 offset:324
+; VI-NEXT: buffer_store_dword v0, off, s[0:3], s32 offset:552 ; 4-byte Folded Spill
+; VI-NEXT: buffer_load_ushort v29, off, s[0:3], s32 offset:276
+; VI-NEXT: buffer_load_ushort v31, off, s[0:3], s32 offset:284
+; VI-NEXT: buffer_load_ushort v30, off, s[0:3], s32 offset:292
+; VI-NEXT: buffer_load_ushort v7, off, s[0:3], s32 offset:300
+; VI-NEXT: buffer_load_ushort v0, off, s[0:3], s32 offset:308
; VI-NEXT: s_waitcnt vmcnt(0)
-; VI-NEXT: buffer_store_dword v10, off, s[0:3], s32 offset:540 ; 4-byte Folded Spill
+; VI-NEXT: buffer_store_dword v0, off, s[0:3], s32 offset:520 ; 4-byte Folded Spill
+; VI-NEXT: buffer_load_ushort v0, off, s[0:3], s32 offset:316
+; VI-NEXT: buffer_load_ushort v2, off, s[0:3], s32 offset:324
; VI-NEXT: buffer_store_dword v9, off, s[0:3], s32 offset:624 ; 4-byte Folded Spill
; VI-NEXT: buffer_store_dword v11, off, s[0:3], s32 offset:628 ; 4-byte Folded Spill
; VI-NEXT: buffer_store_dword v13, off, s[0:3], s32 offset:632 ; 4-byte Folded Spill
@@ -210715,46 +210830,50 @@ define inreg <64 x i16> @bitcast_v128i8_to_v64i16_scalar(<128 x i8> inreg %a, i3
; VI-NEXT: buffer_store_dword v21, off, s[0:3], s32 offset:648 ; 4-byte Folded Spill
; VI-NEXT: buffer_store_dword v23, off, s[0:3], s32 offset:652 ; 4-byte Folded Spill
; VI-NEXT: buffer_store_dword v25, off, s[0:3], s32 offset:656 ; 4-byte Folded Spill
-; VI-NEXT: buffer_store_dword v14, off, s[0:3], s32 offset:660 ; 4-byte Folded Spill
-; VI-NEXT: buffer_store_dword v16, off, s[0:3], s32 offset:664 ; 4-byte Folded Spill
-; VI-NEXT: buffer_store_dword v18, off, s[0:3], s32 offset:668 ; 4-byte Folded Spill
-; VI-NEXT: buffer_store_dword v20, off, s[0:3], s32 offset:672 ; 4-byte Folded Spill
-; VI-NEXT: buffer_store_dword v45, off, s[0:3], s32 offset:676 ; 4-byte Folded Spill
-; VI-NEXT: buffer_store_dword v8, off, s[0:3], s32 offset:680 ; 4-byte Folded Spill
+; VI-NEXT: buffer_store_dword v40, off, s[0:3], s32 offset:660 ; 4-byte Folded Spill
+; VI-NEXT: buffer_store_dword v14, off, s[0:3], s32 offset:664 ; 4-byte Folded Spill
+; VI-NEXT: buffer_store_dword v16, off, s[0:3], s32 offset:668 ; 4-byte Folded Spill
+; VI-NEXT: buffer_store_dword v18, off, s[0:3], s32 offset:672 ; 4-byte Folded Spill
+; VI-NEXT: buffer_store_dword v20, off, s[0:3], s32 offset:676 ; 4-byte Folded Spill
+; VI-NEXT: buffer_store_dword v22, off, s[0:3], s32 offset:680 ; 4-byte Folded Spill
; VI-NEXT: buffer_store_dword v12, off, s[0:3], s32 offset:684 ; 4-byte Folded Spill
; VI-NEXT: buffer_store_dword v32, off, s[0:3], s32 offset:688 ; 4-byte Folded Spill
; VI-NEXT: buffer_store_dword v33, off, s[0:3], s32 offset:692 ; 4-byte Folded Spill
; VI-NEXT: buffer_store_dword v34, off, s[0:3], s32 offset:696 ; 4-byte Folded Spill
; VI-NEXT: buffer_store_dword v35, off, s[0:3], s32 offset:700 ; 4-byte Folded Spill
-; VI-NEXT: buffer_store_dword v36, off, s[0:3], s32 offset:704 ; 4-byte Folded Spill
-; VI-NEXT: buffer_store_dword v37, off, s[0:3], s32 offset:708 ; 4-byte Folded Spill
-; VI-NEXT: buffer_store_dword v38, off, s[0:3], s32 offset:712 ; 4-byte Folded Spill
-; VI-NEXT: buffer_store_dword v39, off, s[0:3], s32 offset:716 ; 4-byte Folded Spill
-; VI-NEXT: buffer_store_dword v48, off, s[0:3], s32 offset:720 ; 4-byte Folded Spill
-; VI-NEXT: buffer_store_dword v49, off, s[0:3], s32 offset:724 ; 4-byte Folded Spill
-; VI-NEXT: buffer_store_dword v50, off, s[0:3], s32 offset:728 ; 4-byte Folded Spill
-; VI-NEXT: buffer_store_dword v51, off, s[0:3], s32 offset:732 ; 4-byte Folded Spill
-; VI-NEXT: buffer_store_dword v54, off, s[0:3], s32 offset:736 ; 4-byte Folded Spill
-; VI-NEXT: buffer_store_dword v41, off, s[0:3], s32 offset:740 ; 4-byte Folded Spill
-; VI-NEXT: buffer_store_dword v56, off, s[0:3], s32 offset:744 ; 4-byte Folded Spill
-; VI-NEXT: buffer_store_dword v55, off, s[0:3], s32 offset:748 ; 4-byte Folded Spill
-; VI-NEXT: buffer_store_dword v52, off, s[0:3], s32 offset:752 ; 4-byte Folded Spill
-; VI-NEXT: buffer_store_dword v43, off, s[0:3], s32 offset:756 ; 4-byte Folded Spill
-; VI-NEXT: buffer_store_dword v53, off, s[0:3], s32 offset:760 ; 4-byte Folded Spill
-; VI-NEXT: buffer_store_dword v46, off, s[0:3], s32 offset:764 ; 4-byte Folded Spill
-; VI-NEXT: buffer_store_dword v58, off, s[0:3], s32 offset:768 ; 4-byte Folded Spill
-; VI-NEXT: buffer_store_dword v47, off, s[0:3], s32 offset:772 ; 4-byte Folded Spill
-; VI-NEXT: buffer_store_dword v63, off, s[0:3], s32 offset:776 ; 4-byte Folded Spill
-; VI-NEXT: buffer_store_dword v57, off, s[0:3], s32 offset:780 ; 4-byte Folded Spill
-; VI-NEXT: buffer_store_dword v61, off, s[0:3], s32 offset:784 ; 4-byte Folded Spill
-; VI-NEXT: buffer_store_dword v29, off, s[0:3], s32 offset:788 ; 4-byte Folded Spill
-; VI-NEXT: buffer_store_dword v30, off, s[0:3], s32 offset:792 ; 4-byte Folded Spill
-; VI-NEXT: buffer_store_dword v31, off, s[0:3], s32 offset:796 ; 4-byte Folded Spill
+; VI-NEXT: buffer_store_dword v8, off, s[0:3], s32 offset:704 ; 4-byte Folded Spill
+; VI-NEXT: buffer_store_dword v36, off, s[0:3], s32 offset:708 ; 4-byte Folded Spill
+; VI-NEXT: buffer_store_dword v37, off, s[0:3], s32 offset:712 ; 4-byte Folded Spill
+; VI-NEXT: buffer_store_dword v38, off, s[0:3], s32 offset:716 ; 4-byte Folded Spill
+; VI-NEXT: buffer_store_dword v39, off, s[0:3], s32 offset:720 ; 4-byte Folded Spill
+; VI-NEXT: buffer_store_dword v10, off, s[0:3], s32 offset:724 ; 4-byte Folded Spill
+; VI-NEXT: buffer_store_dword v48, off, s[0:3], s32 offset:728 ; 4-byte Folded Spill
+; VI-NEXT: buffer_store_dword v49, off, s[0:3], s32 offset:732 ; 4-byte Folded Spill
+; VI-NEXT: buffer_store_dword v51, off, s[0:3], s32 offset:736 ; 4-byte Folded Spill
+; VI-NEXT: buffer_store_dword v53, off, s[0:3], s32 offset:740 ; 4-byte Folded Spill
+; VI-NEXT: buffer_store_dword v55, off, s[0:3], s32 offset:744 ; 4-byte Folded Spill
+; VI-NEXT: buffer_store_dword v41, off, s[0:3], s32 offset:748 ; 4-byte Folded Spill
+; VI-NEXT: buffer_store_dword v56, off, s[0:3], s32 offset:752 ; 4-byte Folded Spill
+; VI-NEXT: buffer_store_dword v50, off, s[0:3], s32 offset:756 ; 4-byte Folded Spill
+; VI-NEXT: buffer_store_dword v43, off, s[0:3], s32 offset:760 ; 4-byte Folded Spill
+; VI-NEXT: buffer_store_dword v58, off, s[0:3], s32 offset:764 ; 4-byte Folded Spill
+; VI-NEXT: buffer_store_dword v52, off, s[0:3], s32 offset:768 ; 4-byte Folded Spill
+; VI-NEXT: buffer_store_dword v60, off, s[0:3], s32 offset:772 ; 4-byte Folded Spill
+; VI-NEXT: buffer_store_dword v54, off, s[0:3], s32 offset:776 ; 4-byte Folded Spill
+; VI-NEXT: buffer_store_dword v47, off, s[0:3], s32 offset:780 ; 4-byte Folded Spill
+; VI-NEXT: buffer_store_dword v63, off, s[0:3], s32 offset:784 ; 4-byte Folded Spill
+; VI-NEXT: buffer_store_dword v46, off, s[0:3], s32 offset:788 ; 4-byte Folded Spill
+; VI-NEXT: buffer_store_dword v61, off, s[0:3], s32 offset:792 ; 4-byte Folded Spill
+; VI-NEXT: buffer_store_dword v29, off, s[0:3], s32 offset:796 ; 4-byte Folded Spill
+; VI-NEXT: buffer_store_dword v31, off, s[0:3], s32 offset:800 ; 4-byte Folded Spill
+; VI-NEXT: buffer_store_dword v30, off, s[0:3], s32 offset:804 ; 4-byte Folded Spill
+; VI-NEXT: buffer_store_dword v7, off, s[0:3], s32 offset:516 ; 4-byte Folded Spill
+; VI-NEXT: s_waitcnt vmcnt(14)
; VI-NEXT: buffer_store_dword v0, off, s[0:3], s32 offset:524 ; 4-byte Folded Spill
+; VI-NEXT: buffer_store_dword v2, off, s[0:3], s32 offset:528 ; 4-byte Folded Spill
; VI-NEXT: buffer_store_dword v6, off, s[0:3], s32 offset:532 ; 4-byte Folded Spill
-; VI-NEXT: buffer_store_dword v7, off, s[0:3], s32 offset:536 ; 4-byte Folded Spill
-; VI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:544 ; 4-byte Folded Spill
-; VI-NEXT: buffer_store_dword v2, off, s[0:3], s32 offset:548 ; 4-byte Folded Spill
+; VI-NEXT: buffer_store_dword v5, off, s[0:3], s32 offset:536 ; 4-byte Folded Spill
+; VI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:540 ; 4-byte Folded Spill
; VI-NEXT: s_cbranch_scc0 .LBB97_4
; VI-NEXT: ; %bb.1: ; %cmp.false
; VI-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:400 ; 4-byte Folded Reload
@@ -210771,11 +210890,10 @@ define inreg <64 x i16> @bitcast_v128i8_to_v64i16_scalar(<128 x i8> inreg %a, i3
; VI-NEXT: v_or_b32_sdwa v0, v0, v1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
; VI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:404 ; 4-byte Folded Reload
; VI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:468 ; 4-byte Folded Reload
-; VI-NEXT: buffer_store_dword v3, off, s[0:3], s32 offset:800 ; 4-byte Folded Spill
+; VI-NEXT: buffer_store_dword v3, off, s[0:3], s32 offset:808 ; 4-byte Folded Spill
; VI-NEXT: buffer_load_dword v3, off, s[0:3], s32 offset:472 ; 4-byte Folded Reload
-; VI-NEXT: buffer_store_dword v4, off, s[0:3], s32 offset:808 ; 4-byte Folded Spill
-; VI-NEXT: buffer_store_dword v5, off, s[0:3], s32 offset:804 ; 4-byte Folded Spill
-; VI-NEXT: s_waitcnt vmcnt(4)
+; VI-NEXT: buffer_store_dword v4, off, s[0:3], s32 offset:812 ; 4-byte Folded Spill
+; VI-NEXT: s_waitcnt vmcnt(3)
; VI-NEXT: v_or_b32_sdwa v1, v1, v2 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
; VI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:408 ; 4-byte Folded Reload
; VI-NEXT: v_or_b32_sdwa v4, v0, v1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_0 src1_sel:DWORD
@@ -210799,6 +210917,7 @@ define inreg <64 x i16> @bitcast_v128i8_to_v64i16_scalar(<128 x i8> inreg %a, i3
; VI-NEXT: v_or_b32_sdwa v0, v0, v15 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
; VI-NEXT: s_waitcnt vmcnt(0)
; VI-NEXT: v_or_b32_sdwa v1, v1, v17 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; VI-NEXT: v_mov_b32_e32 v17, v10
; VI-NEXT: v_or_b32_sdwa v7, v0, v1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_0 src1_sel:DWORD
; VI-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:432 ; 4-byte Folded Reload
; VI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:436 ; 4-byte Folded Reload
@@ -210815,38 +210934,43 @@ define inreg <64 x i16> @bitcast_v128i8_to_v64i16_scalar(<128 x i8> inreg %a, i3
; VI-NEXT: v_or_b32_sdwa v1, v1, v25 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
; VI-NEXT: v_or_b32_sdwa v9, v0, v1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_0 src1_sel:DWORD
; VI-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:448 ; 4-byte Folded Reload
-; VI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:476 ; 4-byte Folded Reload
-; VI-NEXT: s_waitcnt vmcnt(0)
-; VI-NEXT: v_or_b32_sdwa v0, v0, v1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
; VI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:452 ; 4-byte Folded Reload
-; VI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:480 ; 4-byte Folded Reload
+; VI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:476 ; 4-byte Folded Reload
+; VI-NEXT: s_waitcnt vmcnt(2)
+; VI-NEXT: v_or_b32_sdwa v0, v0, v40 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
; VI-NEXT: s_waitcnt vmcnt(0)
; VI-NEXT: v_or_b32_sdwa v1, v1, v2 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; VI-NEXT: v_mov_b32_e32 v40, v42
+; VI-NEXT: v_mov_b32_e32 v42, v44
+; VI-NEXT: v_mov_b32_e32 v44, v45
+; VI-NEXT: v_mov_b32_e32 v45, v62
+; VI-NEXT: v_or_b32_sdwa v2, v51, v3 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; VI-NEXT: v_or_b32_sdwa v3, v53, v17 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
; VI-NEXT: v_or_b32_sdwa v10, v0, v1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_0 src1_sel:DWORD
; VI-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:456 ; 4-byte Folded Reload
-; VI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:484 ; 4-byte Folded Reload
-; VI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:512 ; 4-byte Folded Reload
+; VI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:480 ; 4-byte Folded Reload
+; VI-NEXT: v_or_b32_sdwa v17, v2, v3 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_0 src1_sel:DWORD
+; VI-NEXT: buffer_load_dword v62, off, s[0:3], s32 offset:548 ; 4-byte Folded Reload
; VI-NEXT: s_waitcnt vmcnt(1)
; VI-NEXT: v_or_b32_sdwa v0, v0, v1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
-; VI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:488 ; 4-byte Folded Reload
-; VI-NEXT: s_waitcnt vmcnt(1)
-; VI-NEXT: v_or_b32_sdwa v2, v50, v2 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; VI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:484 ; 4-byte Folded Reload
; VI-NEXT: s_waitcnt vmcnt(0)
; VI-NEXT: v_or_b32_sdwa v1, v12, v1 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
; VI-NEXT: v_or_b32_sdwa v11, v0, v1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_0 src1_sel:DWORD
-; VI-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:492 ; 4-byte Folded Reload
-; VI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:496 ; 4-byte Folded Reload
+; VI-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:488 ; 4-byte Folded Reload
+; VI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:492 ; 4-byte Folded Reload
; VI-NEXT: s_waitcnt vmcnt(1)
; VI-NEXT: v_or_b32_sdwa v0, v32, v0 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
; VI-NEXT: s_waitcnt vmcnt(0)
; VI-NEXT: v_or_b32_sdwa v1, v33, v1 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
; VI-NEXT: v_or_b32_sdwa v12, v0, v1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_0 src1_sel:DWORD
-; VI-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:500 ; 4-byte Folded Reload
-; VI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:504 ; 4-byte Folded Reload
+; VI-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:496 ; 4-byte Folded Reload
+; VI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:500 ; 4-byte Folded Reload
; VI-NEXT: s_waitcnt vmcnt(1)
; VI-NEXT: v_or_b32_sdwa v0, v34, v0 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
; VI-NEXT: s_waitcnt vmcnt(0)
; VI-NEXT: v_or_b32_sdwa v1, v35, v1 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; VI-NEXT: v_mov_b32_e32 v34, v24
; VI-NEXT: v_or_b32_sdwa v13, v0, v1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_0 src1_sel:DWORD
; VI-NEXT: v_or_b32_sdwa v0, v36, v14 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
; VI-NEXT: v_or_b32_sdwa v1, v37, v16 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
@@ -210854,77 +210978,74 @@ define inreg <64 x i16> @bitcast_v128i8_to_v64i16_scalar(<128 x i8> inreg %a, i3
; VI-NEXT: v_or_b32_sdwa v0, v38, v18 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
; VI-NEXT: v_or_b32_sdwa v1, v39, v20 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
; VI-NEXT: v_or_b32_sdwa v15, v0, v1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_0 src1_sel:DWORD
-; VI-NEXT: v_or_b32_sdwa v0, v48, v45 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
-; VI-NEXT: v_or_b32_sdwa v1, v49, v3 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
-; VI-NEXT: buffer_load_dword v3, off, s[0:3], s32 offset:516 ; 4-byte Folded Reload
-; VI-NEXT: v_mov_b32_e32 v45, v62
+; VI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:508 ; 4-byte Folded Reload
+; VI-NEXT: v_or_b32_sdwa v0, v48, v22 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; VI-NEXT: s_waitcnt vmcnt(0)
+; VI-NEXT: v_or_b32_sdwa v1, v49, v1 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
; VI-NEXT: v_or_b32_sdwa v16, v0, v1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_0 src1_sel:DWORD
-; VI-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:520 ; 4-byte Folded Reload
-; VI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:560 ; 4-byte Folded Reload
-; VI-NEXT: buffer_load_dword v62, off, s[0:3], s32 offset:552 ; 4-byte Folded Reload
-; VI-NEXT: s_waitcnt vmcnt(3)
-; VI-NEXT: v_or_b32_sdwa v3, v51, v3 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
-; VI-NEXT: v_or_b32_sdwa v17, v2, v3 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_0 src1_sel:DWORD
-; VI-NEXT: s_waitcnt vmcnt(2)
-; VI-NEXT: v_or_b32_sdwa v0, v54, v0 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; VI-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:512 ; 4-byte Folded Reload
+; VI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:556 ; 4-byte Folded Reload
; VI-NEXT: s_waitcnt vmcnt(1)
+; VI-NEXT: v_or_b32_sdwa v0, v55, v0 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; VI-NEXT: s_waitcnt vmcnt(0)
; VI-NEXT: v_mov_b32_e32 v32, v1
; VI-NEXT: v_or_b32_sdwa v1, v41, v1 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
-; VI-NEXT: v_mov_b32_e32 v54, v22
-; VI-NEXT: v_mov_b32_e32 v41, v24
; VI-NEXT: v_or_b32_sdwa v18, v0, v1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_0 src1_sel:DWORD
-; VI-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:564 ; 4-byte Folded Reload
-; VI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:592 ; 4-byte Folded Reload
+; VI-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:588 ; 4-byte Folded Reload
+; VI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:584 ; 4-byte Folded Reload
; VI-NEXT: s_waitcnt vmcnt(1)
-; VI-NEXT: v_mov_b32_e32 v34, v0
+; VI-NEXT: v_mov_b32_e32 v33, v0
; VI-NEXT: v_or_b32_sdwa v0, v56, v0 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
; VI-NEXT: s_waitcnt vmcnt(0)
; VI-NEXT: v_mov_b32_e32 v37, v1
-; VI-NEXT: v_or_b32_sdwa v1, v55, v1 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
-; VI-NEXT: v_mov_b32_e32 v55, v26
+; VI-NEXT: v_or_b32_sdwa v1, v50, v1 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; VI-NEXT: v_mov_b32_e32 v50, v26
; VI-NEXT: v_or_b32_sdwa v19, v0, v1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_0 src1_sel:DWORD
-; VI-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:588 ; 4-byte Folded Reload
-; VI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:584 ; 4-byte Folded Reload
+; VI-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:576 ; 4-byte Folded Reload
+; VI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:572 ; 4-byte Folded Reload
; VI-NEXT: s_waitcnt vmcnt(1)
; VI-NEXT: v_mov_b32_e32 v39, v0
-; VI-NEXT: v_or_b32_sdwa v0, v52, v0 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; VI-NEXT: v_or_b32_sdwa v0, v43, v0 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
; VI-NEXT: s_waitcnt vmcnt(0)
; VI-NEXT: v_mov_b32_e32 v49, v1
-; VI-NEXT: v_or_b32_sdwa v1, v43, v1 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
-; VI-NEXT: v_mov_b32_e32 v43, v27
+; VI-NEXT: v_or_b32_sdwa v1, v58, v1 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
; VI-NEXT: v_or_b32_sdwa v20, v0, v1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_0 src1_sel:DWORD
-; VI-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:576 ; 4-byte Folded Reload
-; VI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:572 ; 4-byte Folded Reload
+; VI-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:564 ; 4-byte Folded Reload
+; VI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:560 ; 4-byte Folded Reload
; VI-NEXT: s_waitcnt vmcnt(1)
; VI-NEXT: v_mov_b32_e32 v51, v0
-; VI-NEXT: v_or_b32_sdwa v0, v53, v0 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; VI-NEXT: v_or_b32_sdwa v0, v52, v0 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
; VI-NEXT: s_waitcnt vmcnt(0)
-; VI-NEXT: v_mov_b32_e32 v35, v1
-; VI-NEXT: v_or_b32_sdwa v1, v46, v1 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
-; VI-NEXT: v_mov_b32_e32 v53, v28
+; VI-NEXT: v_mov_b32_e32 v53, v1
+; VI-NEXT: v_or_b32_sdwa v1, v60, v1 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; VI-NEXT: v_mov_b32_e32 v52, v28
; VI-NEXT: v_or_b32_sdwa v21, v0, v1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_0 src1_sel:DWORD
-; VI-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:580 ; 4-byte Folded Reload
-; VI-NEXT: v_or_b32_sdwa v1, v47, v22 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
-; VI-NEXT: s_waitcnt vmcnt(0)
-; VI-NEXT: v_mov_b32_e32 v33, v0
-; VI-NEXT: v_or_b32_sdwa v0, v58, v0 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
-; VI-NEXT: v_or_b32_sdwa v22, v0, v1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_0 src1_sel:DWORD
; VI-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:568 ; 4-byte Folded Reload
-; VI-NEXT: v_or_b32_sdwa v1, v57, v24 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; VI-NEXT: v_or_b32_sdwa v1, v47, v24 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
; VI-NEXT: s_waitcnt vmcnt(0)
-; VI-NEXT: v_mov_b32_e32 v36, v0
+; VI-NEXT: v_mov_b32_e32 v55, v0
+; VI-NEXT: v_or_b32_sdwa v0, v54, v0 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; VI-NEXT: v_or_b32_sdwa v22, v0, v1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_0 src1_sel:DWORD
+; VI-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:592 ; 4-byte Folded Reload
+; VI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:580 ; 4-byte Folded Reload
+; VI-NEXT: s_waitcnt vmcnt(1)
+; VI-NEXT: v_mov_b32_e32 v35, v0
; VI-NEXT: v_or_b32_sdwa v0, v63, v0 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; VI-NEXT: s_waitcnt vmcnt(0)
+; VI-NEXT: v_mov_b32_e32 v41, v1
+; VI-NEXT: v_or_b32_sdwa v1, v46, v1 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; VI-NEXT: v_mov_b32_e32 v63, v27
+; VI-NEXT: v_mov_b32_e32 v46, v57
; VI-NEXT: v_or_b32_sdwa v23, v0, v1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_0 src1_sel:DWORD
; VI-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:600 ; 4-byte Folded Reload
; VI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:596 ; 4-byte Folded Reload
; VI-NEXT: s_waitcnt vmcnt(1)
-; VI-NEXT: v_mov_b32_e32 v56, v0
+; VI-NEXT: v_mov_b32_e32 v36, v0
; VI-NEXT: v_or_b32_sdwa v0, v61, v0 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
; VI-NEXT: s_waitcnt vmcnt(0)
-; VI-NEXT: v_mov_b32_e32 v58, v1
-; VI-NEXT: v_or_b32_sdwa v1, v60, v1 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
-; VI-NEXT: v_mov_b32_e32 v61, v60
-; VI-NEXT: v_mov_b32_e32 v60, v59
+; VI-NEXT: v_mov_b32_e32 v56, v1
+; VI-NEXT: v_or_b32_sdwa v1, v57, v1 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; VI-NEXT: v_mov_b32_e32 v61, v59
; VI-NEXT: v_or_b32_sdwa v24, v0, v1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_0 src1_sel:DWORD
; VI-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:608 ; 4-byte Folded Reload
; VI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:612 ; 4-byte Folded Reload
@@ -210936,55 +211057,53 @@ define inreg <64 x i16> @bitcast_v128i8_to_v64i16_scalar(<128 x i8> inreg %a, i3
; VI-NEXT: v_or_b32_sdwa v1, v40, v1 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
; VI-NEXT: v_or_b32_sdwa v25, v0, v1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_0 src1_sel:DWORD
; VI-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:604 ; 4-byte Folded Reload
-; VI-NEXT: buffer_load_dword v59, off, s[0:3], s32 offset:556 ; 4-byte Folded Reload
-; VI-NEXT: v_or_b32_sdwa v1, v45, v26 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; VI-NEXT: buffer_load_dword v59, off, s[0:3], s32 offset:552 ; 4-byte Folded Reload
+; VI-NEXT: v_or_b32_sdwa v1, v44, v26 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
; VI-NEXT: s_waitcnt vmcnt(1)
-; VI-NEXT: v_mov_b32_e32 v50, v0
+; VI-NEXT: v_mov_b32_e32 v58, v0
; VI-NEXT: v_or_b32_sdwa v0, v42, v0 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
; VI-NEXT: v_or_b32_sdwa v26, v0, v1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_0 src1_sel:DWORD
; VI-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:616 ; 4-byte Folded Reload
; VI-NEXT: v_or_b32_sdwa v1, v62, v27 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
; VI-NEXT: s_waitcnt vmcnt(0)
-; VI-NEXT: v_mov_b32_e32 v52, v0
-; VI-NEXT: v_or_b32_sdwa v0, v44, v0 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; VI-NEXT: v_mov_b32_e32 v43, v0
+; VI-NEXT: v_or_b32_sdwa v0, v45, v0 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
; VI-NEXT: v_or_b32_sdwa v27, v0, v1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_0 src1_sel:DWORD
; VI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:620 ; 4-byte Folded Reload
-; VI-NEXT: v_or_b32_sdwa v0, v29, v28 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; VI-NEXT: v_or_b32_sdwa v0, v59, v28 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
; VI-NEXT: s_waitcnt vmcnt(0)
-; VI-NEXT: v_mov_b32_e32 v46, v1
-; VI-NEXT: v_or_b32_sdwa v1, v59, v1 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; VI-NEXT: v_mov_b32_e32 v60, v1
+; VI-NEXT: v_or_b32_sdwa v1, v29, v1 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
; VI-NEXT: v_or_b32_sdwa v28, v0, v1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_0 src1_sel:DWORD
-; VI-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:800 ; 4-byte Folded Reload
-; VI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:808 ; 4-byte Folded Reload
+; VI-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:808 ; 4-byte Folded Reload
+; VI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:812 ; 4-byte Folded Reload
; VI-NEXT: s_waitcnt vmcnt(1)
-; VI-NEXT: v_mov_b32_e32 v63, v0
-; VI-NEXT: v_or_b32_sdwa v0, v30, v0 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; VI-NEXT: v_mov_b32_e32 v54, v0
+; VI-NEXT: v_or_b32_sdwa v0, v31, v0 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
; VI-NEXT: s_waitcnt vmcnt(0)
; VI-NEXT: v_mov_b32_e32 v47, v1
-; VI-NEXT: v_or_b32_sdwa v1, v31, v1 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; VI-NEXT: v_or_b32_sdwa v1, v30, v1 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
; VI-NEXT: v_or_b32_sdwa v29, v0, v1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_0 src1_sel:DWORD
-; VI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:804 ; 4-byte Folded Reload
-; VI-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:524 ; 4-byte Folded Reload
-; VI-NEXT: s_waitcnt vmcnt(1)
-; VI-NEXT: v_mov_b32_e32 v57, v1
-; VI-NEXT: s_waitcnt vmcnt(0)
-; VI-NEXT: v_or_b32_sdwa v0, v0, v1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
-; VI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:528 ; 4-byte Folded Reload
+; VI-NEXT: buffer_load_dword v57, off, s[0:3], s32 offset:544 ; 4-byte Folded Reload
+; VI-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:516 ; 4-byte Folded Reload
+; VI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:520 ; 4-byte Folded Reload
; VI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:532 ; 4-byte Folded Reload
+; VI-NEXT: s_waitcnt vmcnt(2)
+; VI-NEXT: v_or_b32_sdwa v0, v0, v57 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
; VI-NEXT: s_waitcnt vmcnt(0)
; VI-NEXT: v_or_b32_sdwa v1, v1, v2 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
; VI-NEXT: v_or_b32_sdwa v30, v0, v1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_0 src1_sel:DWORD
-; VI-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:544 ; 4-byte Folded Reload
+; VI-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:524 ; 4-byte Folded Reload
; VI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:536 ; 4-byte Folded Reload
; VI-NEXT: s_waitcnt vmcnt(0)
-; VI-NEXT: v_or_b32_sdwa v0, v1, v0 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
-; VI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:548 ; 4-byte Folded Reload
-; VI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:540 ; 4-byte Folded Reload
+; VI-NEXT: v_or_b32_sdwa v0, v0, v1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; VI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:540 ; 4-byte Folded Reload
+; VI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:528 ; 4-byte Folded Reload
; VI-NEXT: s_waitcnt vmcnt(0)
; VI-NEXT: v_or_b32_sdwa v1, v2, v1 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
; VI-NEXT: v_or_b32_sdwa v31, v0, v1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_0 src1_sel:DWORD
; VI-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:460 ; 4-byte Folded Reload
-; VI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:508 ; 4-byte Folded Reload
+; VI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:504 ; 4-byte Folded Reload
; VI-NEXT: s_waitcnt vmcnt(0)
; VI-NEXT: v_or_b32_sdwa v0, v0, v1 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
; VI-NEXT: v_or_b32_e32 v3, s4, v0
@@ -211016,12 +211135,10 @@ define inreg <64 x i16> @bitcast_v128i8_to_v64i16_scalar(<128 x i8> inreg %a, i3
; VI-NEXT: v_mov_b32_e32 v2, s6
; VI-NEXT: s_cbranch_execnz .LBB97_3
; VI-NEXT: .LBB97_2: ; %cmp.true
-; VI-NEXT: s_waitcnt vmcnt(0)
-; VI-NEXT: v_add_u32_e32 v3, vcc, 3, v59
-; VI-NEXT: v_or_b32_sdwa v29, v46, v3 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:BYTE_0
-; VI-NEXT: buffer_load_dword v3, off, s[0:3], s32 offset:788 ; 4-byte Folded Reload
-; VI-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:540 ; 4-byte Folded Reload
-; VI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:548 ; 4-byte Folded Reload
+; VI-NEXT: buffer_load_dword v3, off, s[0:3], s32 offset:796 ; 4-byte Folded Reload
+; VI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:540 ; 4-byte Folded Reload
+; VI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:532 ; 4-byte Folded Reload
+; VI-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:528 ; 4-byte Folded Reload
; VI-NEXT: s_add_i32 s28, s28, 3
; VI-NEXT: s_and_b32 s4, s28, 0xff
; VI-NEXT: s_lshl_b32 s5, s29, 8
@@ -211040,165 +211157,147 @@ define inreg <64 x i16> @bitcast_v128i8_to_v64i16_scalar(<128 x i8> inreg %a, i3
; VI-NEXT: s_lshl_b32 s9, s19, 8
; VI-NEXT: s_add_i32 s16, s16, 3
; VI-NEXT: s_lshl_b32 s10, s17, 8
-; VI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:532 ; 4-byte Folded Reload
-; VI-NEXT: s_waitcnt vmcnt(3)
+; VI-NEXT: buffer_load_dword v4, off, s[0:3], s32 offset:512 ; 4-byte Folded Reload
+; VI-NEXT: s_waitcnt vmcnt(4)
; VI-NEXT: v_add_u32_e32 v3, vcc, 3, v3
-; VI-NEXT: v_or_b32_sdwa v26, v53, v3 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:BYTE_0
+; VI-NEXT: v_or_b32_sdwa v28, v60, v3 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:BYTE_0
+; VI-NEXT: v_add_u32_e32 v3, vcc, 3, v59
+; VI-NEXT: v_or_b32_sdwa v25, v52, v3 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:BYTE_0
; VI-NEXT: v_add_u32_e32 v3, vcc, 3, v62
-; VI-NEXT: v_or_b32_sdwa v28, v43, v3 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:BYTE_0
-; VI-NEXT: v_add_u32_e32 v3, vcc, 3, v44
-; VI-NEXT: v_or_b32_sdwa v53, v52, v3 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:BYTE_0
+; VI-NEXT: v_or_b32_sdwa v27, v63, v3 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:BYTE_0
; VI-NEXT: v_add_u32_e32 v3, vcc, 3, v45
-; VI-NEXT: v_or_b32_sdwa v27, v55, v3 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:BYTE_0
+; VI-NEXT: v_or_b32_sdwa v52, v43, v3 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:BYTE_0
+; VI-NEXT: v_add_u32_e32 v3, vcc, 3, v44
+; VI-NEXT: v_or_b32_sdwa v26, v50, v3 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:BYTE_0
; VI-NEXT: v_add_u32_e32 v3, vcc, 3, v42
-; VI-NEXT: v_or_b32_sdwa v52, v50, v3 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:BYTE_0
+; VI-NEXT: v_or_b32_sdwa v63, v58, v3 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:BYTE_0
; VI-NEXT: v_add_u32_e32 v3, vcc, 3, v40
-; VI-NEXT: v_or_b32_sdwa v25, v48, v3 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:BYTE_0
-; VI-NEXT: v_add_u32_e32 v3, vcc, 3, v60
-; VI-NEXT: v_or_b32_sdwa v59, v38, v3 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:BYTE_0
+; VI-NEXT: v_or_b32_sdwa v43, v48, v3 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:BYTE_0
; VI-NEXT: v_add_u32_e32 v3, vcc, 3, v61
-; VI-NEXT: v_or_b32_sdwa v24, v58, v3 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:BYTE_0
-; VI-NEXT: buffer_load_dword v3, off, s[0:3], s32 offset:784 ; 4-byte Folded Reload
+; VI-NEXT: v_or_b32_sdwa v59, v38, v3 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:BYTE_0
+; VI-NEXT: v_add_u32_e32 v3, vcc, 3, v46
+; VI-NEXT: v_or_b32_sdwa v24, v56, v3 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:BYTE_0
+; VI-NEXT: buffer_load_dword v3, off, s[0:3], s32 offset:792 ; 4-byte Folded Reload
; VI-NEXT: s_waitcnt vmcnt(0)
; VI-NEXT: v_add_u32_e32 v3, vcc, 3, v3
-; VI-NEXT: v_or_b32_sdwa v48, v56, v3 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:BYTE_0
-; VI-NEXT: buffer_load_dword v3, off, s[0:3], s32 offset:780 ; 4-byte Folded Reload
+; VI-NEXT: v_or_b32_sdwa v48, v36, v3 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:BYTE_0
+; VI-NEXT: buffer_load_dword v3, off, s[0:3], s32 offset:788 ; 4-byte Folded Reload
; VI-NEXT: v_add_u32_e32 v48, vcc, 0x300, v48
; VI-NEXT: v_or_b32_sdwa v24, v24, v48 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:WORD_0
; VI-NEXT: v_add_u32_e32 v24, vcc, 0x3000000, v24
; VI-NEXT: s_waitcnt vmcnt(0)
; VI-NEXT: v_add_u32_e32 v3, vcc, 3, v3
; VI-NEXT: v_or_b32_sdwa v23, v41, v3 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:BYTE_0
-; VI-NEXT: buffer_load_dword v3, off, s[0:3], s32 offset:776 ; 4-byte Folded Reload
+; VI-NEXT: buffer_load_dword v3, off, s[0:3], s32 offset:784 ; 4-byte Folded Reload
; VI-NEXT: s_waitcnt vmcnt(0)
; VI-NEXT: v_add_u32_e32 v3, vcc, 3, v3
-; VI-NEXT: v_or_b32_sdwa v38, v36, v3 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:BYTE_0
-; VI-NEXT: buffer_load_dword v3, off, s[0:3], s32 offset:772 ; 4-byte Folded Reload
+; VI-NEXT: v_or_b32_sdwa v38, v35, v3 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:BYTE_0
+; VI-NEXT: buffer_load_dword v3, off, s[0:3], s32 offset:780 ; 4-byte Folded Reload
; VI-NEXT: v_add_u32_e32 v38, vcc, 0x300, v38
; VI-NEXT: v_or_b32_sdwa v23, v23, v38 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:WORD_0
; VI-NEXT: v_add_u32_e32 v23, vcc, 0x3000000, v23
; VI-NEXT: s_waitcnt vmcnt(0)
; VI-NEXT: v_add_u32_e32 v3, vcc, 3, v3
-; VI-NEXT: v_or_b32_sdwa v22, v54, v3 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:BYTE_0
-; VI-NEXT: buffer_load_dword v3, off, s[0:3], s32 offset:768 ; 4-byte Folded Reload
+; VI-NEXT: v_or_b32_sdwa v22, v34, v3 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:BYTE_0
+; VI-NEXT: buffer_load_dword v3, off, s[0:3], s32 offset:776 ; 4-byte Folded Reload
; VI-NEXT: s_waitcnt vmcnt(0)
; VI-NEXT: v_add_u32_e32 v3, vcc, 3, v3
-; VI-NEXT: v_or_b32_sdwa v50, v33, v3 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:BYTE_0
-; VI-NEXT: buffer_load_dword v3, off, s[0:3], s32 offset:764 ; 4-byte Folded Reload
-; VI-NEXT: v_add_u32_e32 v50, vcc, 0x300, v50
+; VI-NEXT: v_or_b32_sdwa v36, v55, v3 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:BYTE_0
+; VI-NEXT: buffer_load_dword v3, off, s[0:3], s32 offset:772 ; 4-byte Folded Reload
+; VI-NEXT: v_add_u32_e32 v50, vcc, 0x300, v36
; VI-NEXT: v_or_b32_sdwa v22, v22, v50 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:WORD_0
; VI-NEXT: v_add_u32_e32 v22, vcc, 0x3000000, v22
; VI-NEXT: s_waitcnt vmcnt(0)
; VI-NEXT: v_add_u32_e32 v3, vcc, 3, v3
-; VI-NEXT: v_or_b32_sdwa v21, v35, v3 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:BYTE_0
-; VI-NEXT: buffer_load_dword v3, off, s[0:3], s32 offset:760 ; 4-byte Folded Reload
+; VI-NEXT: v_or_b32_sdwa v21, v53, v3 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:BYTE_0
+; VI-NEXT: buffer_load_dword v3, off, s[0:3], s32 offset:768 ; 4-byte Folded Reload
; VI-NEXT: s_waitcnt vmcnt(0)
; VI-NEXT: v_add_u32_e32 v3, vcc, 3, v3
-; VI-NEXT: v_or_b32_sdwa v54, v51, v3 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:BYTE_0
-; VI-NEXT: buffer_load_dword v3, off, s[0:3], s32 offset:756 ; 4-byte Folded Reload
+; VI-NEXT: v_or_b32_sdwa v53, v51, v3 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:BYTE_0
+; VI-NEXT: buffer_load_dword v3, off, s[0:3], s32 offset:764 ; 4-byte Folded Reload
; VI-NEXT: s_waitcnt vmcnt(0)
; VI-NEXT: v_add_u32_e32 v3, vcc, 3, v3
; VI-NEXT: v_or_b32_sdwa v20, v49, v3 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:BYTE_0
-; VI-NEXT: buffer_load_dword v3, off, s[0:3], s32 offset:752 ; 4-byte Folded Reload
+; VI-NEXT: buffer_load_dword v3, off, s[0:3], s32 offset:760 ; 4-byte Folded Reload
; VI-NEXT: s_waitcnt vmcnt(0)
; VI-NEXT: v_add_u32_e32 v3, vcc, 3, v3
; VI-NEXT: v_or_b32_sdwa v49, v39, v3 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:BYTE_0
-; VI-NEXT: buffer_load_dword v3, off, s[0:3], s32 offset:748 ; 4-byte Folded Reload
-; VI-NEXT: v_add_u32_e32 v0, vcc, 3, v0
-; VI-NEXT: v_or_b32_sdwa v0, v1, v0 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:BYTE_0
-; VI-NEXT: buffer_store_dword v0, off, s[0:3], s32 offset:540 ; 4-byte Folded Spill
-; VI-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:536 ; 4-byte Folded Reload
-; VI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:544 ; 4-byte Folded Reload
+; VI-NEXT: buffer_load_dword v3, off, s[0:3], s32 offset:756 ; 4-byte Folded Reload
; VI-NEXT: v_add_u32_e32 v49, vcc, 0x300, v49
; VI-NEXT: v_or_b32_sdwa v20, v20, v49 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:WORD_0
; VI-NEXT: v_add_u32_e32 v20, vcc, 0x3000000, v20
-; VI-NEXT: s_waitcnt vmcnt(3)
+; VI-NEXT: s_waitcnt vmcnt(0)
; VI-NEXT: v_add_u32_e32 v3, vcc, 3, v3
; VI-NEXT: v_or_b32_sdwa v19, v37, v3 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:BYTE_0
-; VI-NEXT: buffer_load_dword v3, off, s[0:3], s32 offset:744 ; 4-byte Folded Reload
-; VI-NEXT: s_waitcnt vmcnt(2)
-; VI-NEXT: v_add_u32_e32 v0, vcc, 3, v0
-; VI-NEXT: s_waitcnt vmcnt(1)
-; VI-NEXT: v_or_b32_sdwa v0, v1, v0 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:BYTE_0
-; VI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:528 ; 4-byte Folded Reload
-; VI-NEXT: s_waitcnt vmcnt(1)
+; VI-NEXT: buffer_load_dword v3, off, s[0:3], s32 offset:752 ; 4-byte Folded Reload
+; VI-NEXT: s_waitcnt vmcnt(0)
; VI-NEXT: v_add_u32_e32 v3, vcc, 3, v3
-; VI-NEXT: v_or_b32_sdwa v37, v34, v3 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:BYTE_0
-; VI-NEXT: buffer_load_dword v3, off, s[0:3], s32 offset:740 ; 4-byte Folded Reload
+; VI-NEXT: v_or_b32_sdwa v37, v33, v3 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:BYTE_0
+; VI-NEXT: buffer_load_dword v3, off, s[0:3], s32 offset:748 ; 4-byte Folded Reload
; VI-NEXT: v_add_u32_e32 v37, vcc, 0x300, v37
-; VI-NEXT: s_waitcnt vmcnt(1)
-; VI-NEXT: v_add_u32_e32 v1, vcc, 3, v1
-; VI-NEXT: v_or_b32_sdwa v31, v2, v1 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:BYTE_0
-; VI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:524 ; 4-byte Folded Reload
-; VI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:796 ; 4-byte Folded Reload
; VI-NEXT: v_or_b32_sdwa v19, v19, v37 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:WORD_0
; VI-NEXT: v_add_u32_e32 v19, vcc, 0x3000000, v19
-; VI-NEXT: s_waitcnt vmcnt(2)
+; VI-NEXT: s_waitcnt vmcnt(0)
; VI-NEXT: v_add_u32_e32 v3, vcc, 3, v3
; VI-NEXT: v_or_b32_sdwa v18, v32, v3 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:BYTE_0
-; VI-NEXT: buffer_load_dword v3, off, s[0:3], s32 offset:736 ; 4-byte Folded Reload
-; VI-NEXT: buffer_load_dword v4, off, s[0:3], s32 offset:520 ; 4-byte Folded Reload
+; VI-NEXT: buffer_load_dword v3, off, s[0:3], s32 offset:744 ; 4-byte Folded Reload
; VI-NEXT: buffer_load_dword v32, off, s[0:3], s32 offset:460 ; 4-byte Folded Reload
-; VI-NEXT: s_waitcnt vmcnt(4)
-; VI-NEXT: v_add_u32_e32 v1, vcc, 3, v1
-; VI-NEXT: v_or_b32_sdwa v1, v57, v1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:BYTE_0
-; VI-NEXT: s_waitcnt vmcnt(2)
-; VI-NEXT: v_add_u32_e32 v3, vcc, 3, v3
; VI-NEXT: s_waitcnt vmcnt(1)
-; VI-NEXT: v_or_b32_sdwa v57, v4, v3 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:BYTE_0
-; VI-NEXT: buffer_load_dword v3, off, s[0:3], s32 offset:732 ; 4-byte Folded Reload
-; VI-NEXT: buffer_load_dword v4, off, s[0:3], s32 offset:516 ; 4-byte Folded Reload
-; VI-NEXT: v_add_u32_e32 v35, vcc, 0x300, v57
-; VI-NEXT: v_or_b32_sdwa v18, v18, v35 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:WORD_0
+; VI-NEXT: v_add_u32_e32 v3, vcc, 3, v3
+; VI-NEXT: v_or_b32_sdwa v58, v4, v3 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:BYTE_0
+; VI-NEXT: buffer_load_dword v3, off, s[0:3], s32 offset:740 ; 4-byte Folded Reload
+; VI-NEXT: buffer_load_dword v4, off, s[0:3], s32 offset:724 ; 4-byte Folded Reload
+; VI-NEXT: v_add_u32_e32 v36, vcc, 0x300, v58
+; VI-NEXT: v_or_b32_sdwa v18, v18, v36 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:WORD_0
; VI-NEXT: v_add_u32_e32 v18, vcc, 0x3000000, v18
; VI-NEXT: s_waitcnt vmcnt(1)
; VI-NEXT: v_add_u32_e32 v3, vcc, 3, v3
; VI-NEXT: s_waitcnt vmcnt(0)
; VI-NEXT: v_or_b32_sdwa v16, v4, v3 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:BYTE_0
-; VI-NEXT: buffer_load_dword v3, off, s[0:3], s32 offset:728 ; 4-byte Folded Reload
-; VI-NEXT: buffer_load_dword v4, off, s[0:3], s32 offset:512 ; 4-byte Folded Reload
+; VI-NEXT: buffer_load_dword v3, off, s[0:3], s32 offset:736 ; 4-byte Folded Reload
+; VI-NEXT: buffer_load_dword v4, off, s[0:3], s32 offset:704 ; 4-byte Folded Reload
; VI-NEXT: s_waitcnt vmcnt(1)
; VI-NEXT: v_add_u32_e32 v3, vcc, 3, v3
; VI-NEXT: s_waitcnt vmcnt(0)
; VI-NEXT: v_or_b32_sdwa v10, v4, v3 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:BYTE_0
-; VI-NEXT: buffer_load_dword v3, off, s[0:3], s32 offset:724 ; 4-byte Folded Reload
-; VI-NEXT: buffer_load_dword v4, off, s[0:3], s32 offset:680 ; 4-byte Folded Reload
+; VI-NEXT: buffer_load_dword v3, off, s[0:3], s32 offset:732 ; 4-byte Folded Reload
+; VI-NEXT: buffer_load_dword v4, off, s[0:3], s32 offset:508 ; 4-byte Folded Reload
; VI-NEXT: s_waitcnt vmcnt(1)
; VI-NEXT: v_add_u32_e32 v3, vcc, 3, v3
; VI-NEXT: s_waitcnt vmcnt(0)
; VI-NEXT: v_or_b32_sdwa v17, v4, v3 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:BYTE_0
+; VI-NEXT: buffer_load_dword v3, off, s[0:3], s32 offset:728 ; 4-byte Folded Reload
+; VI-NEXT: buffer_load_dword v4, off, s[0:3], s32 offset:680 ; 4-byte Folded Reload
+; VI-NEXT: s_waitcnt vmcnt(1)
+; VI-NEXT: v_add_u32_e32 v3, vcc, 3, v3
+; VI-NEXT: s_waitcnt vmcnt(0)
+; VI-NEXT: v_or_b32_sdwa v11, v4, v3 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:BYTE_0
; VI-NEXT: buffer_load_dword v3, off, s[0:3], s32 offset:720 ; 4-byte Folded Reload
; VI-NEXT: buffer_load_dword v4, off, s[0:3], s32 offset:676 ; 4-byte Folded Reload
; VI-NEXT: s_waitcnt vmcnt(1)
; VI-NEXT: v_add_u32_e32 v3, vcc, 3, v3
; VI-NEXT: s_waitcnt vmcnt(0)
-; VI-NEXT: v_or_b32_sdwa v11, v4, v3 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:BYTE_0
+; VI-NEXT: v_or_b32_sdwa v15, v4, v3 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:BYTE_0
; VI-NEXT: buffer_load_dword v3, off, s[0:3], s32 offset:716 ; 4-byte Folded Reload
; VI-NEXT: buffer_load_dword v4, off, s[0:3], s32 offset:672 ; 4-byte Folded Reload
; VI-NEXT: s_waitcnt vmcnt(1)
; VI-NEXT: v_add_u32_e32 v3, vcc, 3, v3
; VI-NEXT: s_waitcnt vmcnt(0)
-; VI-NEXT: v_or_b32_sdwa v15, v4, v3 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:BYTE_0
+; VI-NEXT: v_or_b32_sdwa v56, v4, v3 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:BYTE_0
; VI-NEXT: buffer_load_dword v3, off, s[0:3], s32 offset:712 ; 4-byte Folded Reload
; VI-NEXT: buffer_load_dword v4, off, s[0:3], s32 offset:668 ; 4-byte Folded Reload
; VI-NEXT: s_waitcnt vmcnt(1)
; VI-NEXT: v_add_u32_e32 v3, vcc, 3, v3
; VI-NEXT: s_waitcnt vmcnt(0)
-; VI-NEXT: v_or_b32_sdwa v56, v4, v3 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:BYTE_0
+; VI-NEXT: v_or_b32_sdwa v14, v4, v3 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:BYTE_0
; VI-NEXT: buffer_load_dword v3, off, s[0:3], s32 offset:708 ; 4-byte Folded Reload
; VI-NEXT: buffer_load_dword v4, off, s[0:3], s32 offset:664 ; 4-byte Folded Reload
; VI-NEXT: s_waitcnt vmcnt(1)
; VI-NEXT: v_add_u32_e32 v3, vcc, 3, v3
; VI-NEXT: s_waitcnt vmcnt(0)
-; VI-NEXT: v_or_b32_sdwa v14, v4, v3 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:BYTE_0
-; VI-NEXT: buffer_load_dword v3, off, s[0:3], s32 offset:704 ; 4-byte Folded Reload
-; VI-NEXT: buffer_load_dword v4, off, s[0:3], s32 offset:660 ; 4-byte Folded Reload
-; VI-NEXT: s_waitcnt vmcnt(1)
-; VI-NEXT: v_add_u32_e32 v3, vcc, 3, v3
-; VI-NEXT: s_waitcnt vmcnt(0)
; VI-NEXT: v_or_b32_sdwa v34, v4, v3 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:BYTE_0
; VI-NEXT: buffer_load_dword v3, off, s[0:3], s32 offset:700 ; 4-byte Folded Reload
-; VI-NEXT: buffer_load_dword v4, off, s[0:3], s32 offset:504 ; 4-byte Folded Reload
+; VI-NEXT: buffer_load_dword v4, off, s[0:3], s32 offset:500 ; 4-byte Folded Reload
; VI-NEXT: v_add_u32_e32 v34, vcc, 0x300, v34
; VI-NEXT: v_or_b32_sdwa v14, v14, v34 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:WORD_0
; VI-NEXT: v_add_u32_e32 v14, vcc, 0x3000000, v14
@@ -211207,67 +211306,78 @@ define inreg <64 x i16> @bitcast_v128i8_to_v64i16_scalar(<128 x i8> inreg %a, i3
; VI-NEXT: s_waitcnt vmcnt(0)
; VI-NEXT: v_or_b32_sdwa v13, v4, v3 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:BYTE_0
; VI-NEXT: buffer_load_dword v3, off, s[0:3], s32 offset:696 ; 4-byte Folded Reload
-; VI-NEXT: buffer_load_dword v4, off, s[0:3], s32 offset:500 ; 4-byte Folded Reload
-; VI-NEXT: s_waitcnt vmcnt(1)
+; VI-NEXT: buffer_load_dword v4, off, s[0:3], s32 offset:496 ; 4-byte Folded Reload
+; VI-NEXT: v_add_u32_e32 v0, vcc, 3, v0
+; VI-NEXT: v_or_b32_sdwa v31, v1, v0 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:BYTE_0
+; VI-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:524 ; 4-byte Folded Reload
+; VI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:536 ; 4-byte Folded Reload
+; VI-NEXT: s_waitcnt vmcnt(3)
; VI-NEXT: v_add_u32_e32 v3, vcc, 3, v3
-; VI-NEXT: s_waitcnt vmcnt(0)
-; VI-NEXT: v_or_b32_sdwa v36, v4, v3 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:BYTE_0
+; VI-NEXT: s_waitcnt vmcnt(2)
+; VI-NEXT: v_or_b32_sdwa v35, v4, v3 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:BYTE_0
; VI-NEXT: buffer_load_dword v3, off, s[0:3], s32 offset:692 ; 4-byte Folded Reload
-; VI-NEXT: buffer_load_dword v4, off, s[0:3], s32 offset:496 ; 4-byte Folded Reload
-; VI-NEXT: v_add_u32_e32 v36, vcc, 0x300, v36
-; VI-NEXT: v_or_b32_sdwa v13, v13, v36 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:WORD_0
-; VI-NEXT: v_add_u32_e32 v36, vcc, 0x300, v26
-; VI-NEXT: v_add_u32_e32 v26, vcc, 0x300, v52
-; VI-NEXT: v_or_b32_sdwa v26, v27, v26 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:WORD_0
-; VI-NEXT: v_add_u32_e32 v52, vcc, 0x300, v54
-; VI-NEXT: v_or_b32_sdwa v21, v21, v52 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:WORD_0
+; VI-NEXT: buffer_load_dword v4, off, s[0:3], s32 offset:492 ; 4-byte Folded Reload
+; VI-NEXT: s_waitcnt vmcnt(3)
+; VI-NEXT: v_add_u32_e32 v0, vcc, 3, v0
+; VI-NEXT: s_waitcnt vmcnt(2)
+; VI-NEXT: v_or_b32_sdwa v0, v1, v0 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:BYTE_0
+; VI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:520 ; 4-byte Folded Reload
+; VI-NEXT: v_add_u32_e32 v35, vcc, 0x300, v35
+; VI-NEXT: v_or_b32_sdwa v13, v13, v35 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:WORD_0
+; VI-NEXT: v_add_u32_e32 v35, vcc, 0x300, v25
+; VI-NEXT: v_add_u32_e32 v25, vcc, 0x300, v59
+; VI-NEXT: v_or_b32_sdwa v25, v43, v25 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:WORD_0
+; VI-NEXT: v_or_b32_sdwa v28, v28, v35 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:WORD_0
; VI-NEXT: v_add_u32_e32 v13, vcc, 0x3000000, v13
-; VI-NEXT: v_add_u32_e32 v21, vcc, 0x3000000, v21
-; VI-NEXT: v_add_u32_e32 v26, vcc, 0x3000000, v26
-; VI-NEXT: s_waitcnt vmcnt(1)
+; VI-NEXT: v_add_u32_e32 v25, vcc, 0x3000000, v25
+; VI-NEXT: v_add_u32_e32 v28, vcc, 0x3000000, v28
+; VI-NEXT: s_waitcnt vmcnt(2)
; VI-NEXT: v_add_u32_e32 v3, vcc, 3, v3
-; VI-NEXT: s_waitcnt vmcnt(0)
+; VI-NEXT: s_waitcnt vmcnt(1)
; VI-NEXT: v_or_b32_sdwa v12, v4, v3 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:BYTE_0
; VI-NEXT: buffer_load_dword v3, off, s[0:3], s32 offset:688 ; 4-byte Folded Reload
-; VI-NEXT: buffer_load_dword v4, off, s[0:3], s32 offset:492 ; 4-byte Folded Reload
-; VI-NEXT: s_waitcnt vmcnt(1)
+; VI-NEXT: buffer_load_dword v4, off, s[0:3], s32 offset:488 ; 4-byte Folded Reload
+; VI-NEXT: s_waitcnt vmcnt(2)
+; VI-NEXT: v_add_u32_e32 v1, vcc, 3, v1
+; VI-NEXT: v_or_b32_sdwa v30, v2, v1 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:BYTE_0
+; VI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:516 ; 4-byte Folded Reload
+; VI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:804 ; 4-byte Folded Reload
+; VI-NEXT: s_waitcnt vmcnt(3)
; VI-NEXT: v_add_u32_e32 v3, vcc, 3, v3
-; VI-NEXT: s_waitcnt vmcnt(0)
+; VI-NEXT: s_waitcnt vmcnt(2)
; VI-NEXT: v_or_b32_sdwa v51, v4, v3 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:BYTE_0
; VI-NEXT: buffer_load_dword v3, off, s[0:3], s32 offset:684 ; 4-byte Folded Reload
-; VI-NEXT: buffer_load_dword v4, off, s[0:3], s32 offset:488 ; 4-byte Folded Reload
+; VI-NEXT: buffer_load_dword v4, off, s[0:3], s32 offset:484 ; 4-byte Folded Reload
+; VI-NEXT: s_waitcnt vmcnt(3)
+; VI-NEXT: v_add_u32_e32 v1, vcc, 3, v1
+; VI-NEXT: v_or_b32_sdwa v1, v57, v1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:BYTE_0
; VI-NEXT: v_add_u32_e32 v51, vcc, 0x300, v51
; VI-NEXT: v_or_b32_sdwa v12, v12, v51 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:WORD_0
-; VI-NEXT: v_add_u32_e32 v51, vcc, 0x300, v59
-; VI-NEXT: v_or_b32_sdwa v25, v25, v51 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:WORD_0
+; VI-NEXT: v_add_u32_e32 v33, vcc, 0x300, v1
+; VI-NEXT: v_add_u32_e32 v51, vcc, 0x300, v63
+; VI-NEXT: v_or_b32_sdwa v26, v26, v51 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:WORD_0
+; VI-NEXT: v_or_b32_sdwa v30, v30, v33 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:WORD_0
; VI-NEXT: v_add_u32_e32 v12, vcc, 0x3000000, v12
-; VI-NEXT: v_add_u32_e32 v25, vcc, 0x3000000, v25
+; VI-NEXT: v_add_u32_e32 v26, vcc, 0x3000000, v26
+; VI-NEXT: v_add_u32_e32 v30, vcc, 0x3000000, v30
; VI-NEXT: s_waitcnt vmcnt(1)
; VI-NEXT: v_add_u32_e32 v3, vcc, 3, v3
; VI-NEXT: s_waitcnt vmcnt(0)
-; VI-NEXT: v_or_b32_sdwa v33, v4, v3 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:BYTE_0
+; VI-NEXT: v_or_b32_sdwa v57, v4, v3 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:BYTE_0
; VI-NEXT: buffer_load_dword v3, off, s[0:3], s32 offset:456 ; 4-byte Folded Reload
-; VI-NEXT: buffer_load_dword v4, off, s[0:3], s32 offset:484 ; 4-byte Folded Reload
+; VI-NEXT: buffer_load_dword v4, off, s[0:3], s32 offset:480 ; 4-byte Folded Reload
; VI-NEXT: s_waitcnt vmcnt(1)
; VI-NEXT: v_add_u32_e32 v3, vcc, 3, v3
; VI-NEXT: s_waitcnt vmcnt(0)
; VI-NEXT: v_or_b32_sdwa v40, v4, v3 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:BYTE_0
; VI-NEXT: buffer_load_dword v3, off, s[0:3], s32 offset:452 ; 4-byte Folded Reload
-; VI-NEXT: buffer_load_dword v4, off, s[0:3], s32 offset:480 ; 4-byte Folded Reload
-; VI-NEXT: v_add_u32_e32 v2, vcc, 3, v2
-; VI-NEXT: v_or_b32_sdwa v30, v47, v2 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:BYTE_0
-; VI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:792 ; 4-byte Folded Reload
-; VI-NEXT: buffer_load_dword v47, off, s[0:3], s32 offset:464 ; 4-byte Folded Reload
-; VI-NEXT: s_waitcnt vmcnt(3)
+; VI-NEXT: buffer_load_dword v4, off, s[0:3], s32 offset:476 ; 4-byte Folded Reload
+; VI-NEXT: s_waitcnt vmcnt(1)
; VI-NEXT: v_add_u32_e32 v3, vcc, 3, v3
-; VI-NEXT: s_waitcnt vmcnt(2)
+; VI-NEXT: s_waitcnt vmcnt(0)
; VI-NEXT: v_or_b32_sdwa v39, v4, v3 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:BYTE_0
; VI-NEXT: buffer_load_dword v3, off, s[0:3], s32 offset:448 ; 4-byte Folded Reload
-; VI-NEXT: buffer_load_dword v4, off, s[0:3], s32 offset:476 ; 4-byte Folded Reload
-; VI-NEXT: s_waitcnt vmcnt(3)
-; VI-NEXT: v_add_u32_e32 v2, vcc, 3, v2
-; VI-NEXT: v_or_b32_sdwa v2, v63, v2 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:BYTE_0
-; VI-NEXT: v_add_u32_e32 v34, vcc, 0x300, v2
+; VI-NEXT: buffer_load_dword v4, off, s[0:3], s32 offset:660 ; 4-byte Folded Reload
; VI-NEXT: s_waitcnt vmcnt(1)
; VI-NEXT: v_add_u32_e32 v3, vcc, 3, v3
; VI-NEXT: s_waitcnt vmcnt(0)
@@ -211291,15 +211401,14 @@ define inreg <64 x i16> @bitcast_v128i8_to_v64i16_scalar(<128 x i8> inreg %a, i3
; VI-NEXT: v_add_u32_e32 v41, vcc, 0x300, v10
; VI-NEXT: v_add_u32_e32 v10, vcc, 0x300, v55
; VI-NEXT: v_or_b32_sdwa v10, v39, v10 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:WORD_0
-; VI-NEXT: v_add_u32_e32 v39, vcc, 0x300, v53
-; VI-NEXT: v_or_b32_sdwa v27, v28, v39 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:WORD_0
-; VI-NEXT: v_or_b32_sdwa v28, v29, v36 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:WORD_0
-; VI-NEXT: v_or_b32_sdwa v29, v30, v34 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:WORD_0
+; VI-NEXT: v_add_u32_e32 v39, vcc, 0x300, v52
+; VI-NEXT: v_add_u32_e32 v52, vcc, 0x300, v53
+; VI-NEXT: v_or_b32_sdwa v21, v21, v52 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:WORD_0
+; VI-NEXT: v_or_b32_sdwa v27, v27, v39 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:WORD_0
; VI-NEXT: v_add_u32_e32 v9, vcc, 0x3000000, v9
; VI-NEXT: v_add_u32_e32 v10, vcc, 0x3000000, v10
+; VI-NEXT: v_add_u32_e32 v21, vcc, 0x3000000, v21
; VI-NEXT: v_add_u32_e32 v27, vcc, 0x3000000, v27
-; VI-NEXT: v_add_u32_e32 v28, vcc, 0x3000000, v28
-; VI-NEXT: v_add_u32_e32 v29, vcc, 0x3000000, v29
; VI-NEXT: s_waitcnt vmcnt(1)
; VI-NEXT: v_add_u32_e32 v3, vcc, 3, v3
; VI-NEXT: s_waitcnt vmcnt(0)
@@ -211315,18 +211424,14 @@ define inreg <64 x i16> @bitcast_v128i8_to_v64i16_scalar(<128 x i8> inreg %a, i3
; VI-NEXT: v_add_u32_e32 v42, vcc, 0x300, v42
; VI-NEXT: v_or_b32_sdwa v8, v8, v42 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:WORD_0
; VI-NEXT: v_add_u32_e32 v42, vcc, 0x300, v11
-; VI-NEXT: v_add_u32_e32 v11, vcc, 0x300, v40
-; VI-NEXT: v_or_b32_sdwa v11, v33, v11 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:WORD_0
-; VI-NEXT: v_add_u32_e32 v33, vcc, 0x300, v1
-; VI-NEXT: v_or_b32_sdwa v30, v31, v33 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:WORD_0
-; VI-NEXT: buffer_load_dword v31, off, s[0:3], s32 offset:540 ; 4-byte Folded Reload
; VI-NEXT: v_or_b32_sdwa v17, v17, v42 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:WORD_0
+; VI-NEXT: v_add_u32_e32 v11, vcc, 0x300, v40
+; VI-NEXT: v_or_b32_sdwa v11, v57, v11 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:WORD_0
; VI-NEXT: v_add_u32_e32 v8, vcc, 0x3000000, v8
; VI-NEXT: v_add_u32_e32 v11, vcc, 0x3000000, v11
-; VI-NEXT: v_add_u32_e32 v30, vcc, 0x3000000, v30
-; VI-NEXT: s_waitcnt vmcnt(2)
-; VI-NEXT: v_add_u32_e32 v3, vcc, 3, v3
; VI-NEXT: s_waitcnt vmcnt(1)
+; VI-NEXT: v_add_u32_e32 v3, vcc, 3, v3
+; VI-NEXT: s_waitcnt vmcnt(0)
; VI-NEXT: v_or_b32_sdwa v7, v4, v3 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:BYTE_0
; VI-NEXT: buffer_load_dword v3, off, s[0:3], s32 offset:424 ; 4-byte Folded Reload
; VI-NEXT: buffer_load_dword v4, off, s[0:3], s32 offset:636 ; 4-byte Folded Reload
@@ -211366,19 +211471,29 @@ define inreg <64 x i16> @bitcast_v128i8_to_v64i16_scalar(<128 x i8> inreg %a, i3
; VI-NEXT: v_or_b32_sdwa v46, v4, v3 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:BYTE_0
; VI-NEXT: buffer_load_dword v3, off, s[0:3], s32 offset:404 ; 4-byte Folded Reload
; VI-NEXT: buffer_load_dword v4, off, s[0:3], s32 offset:468 ; 4-byte Folded Reload
+; VI-NEXT: v_add_u32_e32 v2, vcc, 3, v2
+; VI-NEXT: v_or_b32_sdwa v29, v47, v2 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:BYTE_0
+; VI-NEXT: buffer_load_dword v47, off, s[0:3], s32 offset:464 ; 4-byte Folded Reload
+; VI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:800 ; 4-byte Folded Reload
; VI-NEXT: v_add_u32_e32 v46, vcc, 0x300, v46
; VI-NEXT: v_or_b32_sdwa v5, v5, v46 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:WORD_0
; VI-NEXT: v_add_u32_e32 v5, vcc, 0x3000000, v5
-; VI-NEXT: s_waitcnt vmcnt(1)
+; VI-NEXT: s_waitcnt vmcnt(3)
; VI-NEXT: v_add_u32_e32 v3, vcc, 3, v3
-; VI-NEXT: s_waitcnt vmcnt(0)
+; VI-NEXT: s_waitcnt vmcnt(2)
; VI-NEXT: v_or_b32_sdwa v3, v4, v3 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:BYTE_0
; VI-NEXT: buffer_load_dword v4, off, s[0:3], s32 offset:400 ; 4-byte Folded Reload
+; VI-NEXT: s_waitcnt vmcnt(1)
+; VI-NEXT: v_add_u32_e32 v2, vcc, 3, v2
+; VI-NEXT: v_or_b32_sdwa v2, v54, v2 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:BYTE_0
+; VI-NEXT: v_add_u32_e32 v34, vcc, 0x300, v2
+; VI-NEXT: v_or_b32_sdwa v29, v29, v34 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:WORD_0
+; VI-NEXT: v_add_u32_e32 v29, vcc, 0x3000000, v29
; VI-NEXT: s_waitcnt vmcnt(0)
; VI-NEXT: v_add_u32_e32 v4, vcc, 3, v4
; VI-NEXT: v_or_b32_sdwa v4, v47, v4 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:BYTE_0
; VI-NEXT: v_add_u32_e32 v47, vcc, 3, v32
-; VI-NEXT: buffer_load_dword v32, off, s[0:3], s32 offset:508 ; 4-byte Folded Reload
+; VI-NEXT: buffer_load_dword v32, off, s[0:3], s32 offset:504 ; 4-byte Folded Reload
; VI-NEXT: v_add_u32_e32 v4, vcc, 0x300, v4
; VI-NEXT: v_or_b32_sdwa v4, v3, v4 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:WORD_0
; VI-NEXT: v_add_u32_e32 v4, vcc, 0x3000000, v4
@@ -211445,35 +211560,38 @@ define inreg <64 x i16> @bitcast_v128i8_to_v64i16_scalar(<128 x i8> inreg %a, i3
; VI-NEXT: s_waitcnt vmcnt(0)
; VI-NEXT: s_setpc_b64 s[30:31]
; VI-NEXT: .LBB97_4:
-; VI-NEXT: buffer_load_dword v46, off, s[0:3], s32 offset:620 ; 4-byte Folded Reload
-; VI-NEXT: buffer_load_dword v52, off, s[0:3], s32 offset:616 ; 4-byte Folded Reload
-; VI-NEXT: buffer_load_dword v50, off, s[0:3], s32 offset:604 ; 4-byte Folded Reload
+; VI-NEXT: buffer_load_dword v60, off, s[0:3], s32 offset:620 ; 4-byte Folded Reload
+; VI-NEXT: buffer_load_dword v43, off, s[0:3], s32 offset:616 ; 4-byte Folded Reload
+; VI-NEXT: buffer_load_dword v58, off, s[0:3], s32 offset:604 ; 4-byte Folded Reload
; VI-NEXT: buffer_load_dword v48, off, s[0:3], s32 offset:612 ; 4-byte Folded Reload
; VI-NEXT: buffer_load_dword v38, off, s[0:3], s32 offset:608 ; 4-byte Folded Reload
-; VI-NEXT: buffer_load_dword v58, off, s[0:3], s32 offset:596 ; 4-byte Folded Reload
-; VI-NEXT: buffer_load_dword v56, off, s[0:3], s32 offset:600 ; 4-byte Folded Reload
-; VI-NEXT: buffer_load_dword v36, off, s[0:3], s32 offset:568 ; 4-byte Folded Reload
-; VI-NEXT: buffer_load_dword v33, off, s[0:3], s32 offset:580 ; 4-byte Folded Reload
-; VI-NEXT: buffer_load_dword v35, off, s[0:3], s32 offset:572 ; 4-byte Folded Reload
-; VI-NEXT: buffer_load_dword v51, off, s[0:3], s32 offset:576 ; 4-byte Folded Reload
-; VI-NEXT: buffer_load_dword v49, off, s[0:3], s32 offset:584 ; 4-byte Folded Reload
-; VI-NEXT: buffer_load_dword v39, off, s[0:3], s32 offset:588 ; 4-byte Folded Reload
-; VI-NEXT: buffer_load_dword v37, off, s[0:3], s32 offset:592 ; 4-byte Folded Reload
-; VI-NEXT: buffer_load_dword v34, off, s[0:3], s32 offset:564 ; 4-byte Folded Reload
-; VI-NEXT: buffer_load_dword v32, off, s[0:3], s32 offset:560 ; 4-byte Folded Reload
-; VI-NEXT: v_mov_b32_e32 v61, v60
-; VI-NEXT: v_mov_b32_e32 v60, v59
+; VI-NEXT: buffer_load_dword v56, off, s[0:3], s32 offset:596 ; 4-byte Folded Reload
+; VI-NEXT: buffer_load_dword v36, off, s[0:3], s32 offset:600 ; 4-byte Folded Reload
+; VI-NEXT: buffer_load_dword v41, off, s[0:3], s32 offset:580 ; 4-byte Folded Reload
+; VI-NEXT: buffer_load_dword v35, off, s[0:3], s32 offset:592 ; 4-byte Folded Reload
+; VI-NEXT: buffer_load_dword v55, off, s[0:3], s32 offset:568 ; 4-byte Folded Reload
+; VI-NEXT: buffer_load_dword v53, off, s[0:3], s32 offset:560 ; 4-byte Folded Reload
+; VI-NEXT: buffer_load_dword v51, off, s[0:3], s32 offset:564 ; 4-byte Folded Reload
+; VI-NEXT: buffer_load_dword v49, off, s[0:3], s32 offset:572 ; 4-byte Folded Reload
+; VI-NEXT: buffer_load_dword v39, off, s[0:3], s32 offset:576 ; 4-byte Folded Reload
+; VI-NEXT: buffer_load_dword v37, off, s[0:3], s32 offset:584 ; 4-byte Folded Reload
+; VI-NEXT: buffer_load_dword v33, off, s[0:3], s32 offset:588 ; 4-byte Folded Reload
+; VI-NEXT: buffer_load_dword v32, off, s[0:3], s32 offset:556 ; 4-byte Folded Reload
+; VI-NEXT: v_mov_b32_e32 v61, v59
+; VI-NEXT: v_mov_b32_e32 v46, v57
+; VI-NEXT: buffer_load_dword v57, off, s[0:3], s32 offset:544 ; 4-byte Folded Reload
+; VI-NEXT: v_mov_b32_e32 v40, v42
+; VI-NEXT: v_mov_b32_e32 v42, v44
+; VI-NEXT: v_mov_b32_e32 v44, v45
; VI-NEXT: v_mov_b32_e32 v45, v62
-; VI-NEXT: buffer_load_dword v62, off, s[0:3], s32 offset:552 ; 4-byte Folded Reload
-; VI-NEXT: buffer_load_dword v59, off, s[0:3], s32 offset:556 ; 4-byte Folded Reload
-; VI-NEXT: v_mov_b32_e32 v57, v5
+; VI-NEXT: buffer_load_dword v62, off, s[0:3], s32 offset:548 ; 4-byte Folded Reload
+; VI-NEXT: buffer_load_dword v59, off, s[0:3], s32 offset:552 ; 4-byte Folded Reload
; VI-NEXT: v_mov_b32_e32 v47, v4
-; VI-NEXT: v_mov_b32_e32 v63, v3
-; VI-NEXT: v_mov_b32_e32 v53, v28
-; VI-NEXT: v_mov_b32_e32 v43, v27
-; VI-NEXT: v_mov_b32_e32 v55, v26
-; VI-NEXT: v_mov_b32_e32 v41, v24
-; VI-NEXT: v_mov_b32_e32 v54, v22
+; VI-NEXT: v_mov_b32_e32 v54, v3
+; VI-NEXT: v_mov_b32_e32 v52, v28
+; VI-NEXT: v_mov_b32_e32 v63, v27
+; VI-NEXT: v_mov_b32_e32 v50, v26
+; VI-NEXT: v_mov_b32_e32 v34, v24
; VI-NEXT: ; implicit-def: $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7_vgpr8_vgpr9_vgpr10_vgpr11_vgpr12_vgpr13_vgpr14_vgpr15
; VI-NEXT: ; implicit-def: $vgpr16_vgpr17_vgpr18_vgpr19_vgpr20_vgpr21_vgpr22_vgpr23_vgpr24_vgpr25_vgpr26_vgpr27_vgpr28_vgpr29_vgpr30_vgpr31
; VI-NEXT: s_branch .LBB97_2
@@ -211535,18 +211653,18 @@ define inreg <64 x i16> @bitcast_v128i8_to_v64i16_scalar(<128 x i8> inreg %a, i3
; GFX9-NEXT: v_lshlrev_b32_e32 v1, 8, v29
; GFX9-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:400 ; 4-byte Folded Spill
; GFX9-NEXT: v_lshlrev_b32_e32 v14, 8, v3
-; GFX9-NEXT: v_lshlrev_b32_e32 v46, 8, v5
+; GFX9-NEXT: v_lshlrev_b32_e32 v18, 8, v5
; GFX9-NEXT: v_lshlrev_b32_e32 v22, 8, v7
; GFX9-NEXT: v_lshlrev_b32_e32 v24, 8, v9
; GFX9-NEXT: v_lshlrev_b32_e32 v26, 8, v11
; GFX9-NEXT: v_lshlrev_b32_e32 v20, 8, v13
-; GFX9-NEXT: v_lshlrev_b32_e32 v28, 8, v15
-; GFX9-NEXT: v_lshlrev_b32_e32 v18, 8, v17
-; GFX9-NEXT: v_lshlrev_b32_e32 v17, 8, v25
-; GFX9-NEXT: v_lshlrev_b32_e32 v12, 8, v27
+; GFX9-NEXT: v_lshlrev_b32_e32 v12, 8, v15
+; GFX9-NEXT: v_lshlrev_b32_e32 v17, 8, v17
; GFX9-NEXT: v_lshlrev_b32_e32 v19, 8, v19
; GFX9-NEXT: v_lshlrev_b32_e32 v21, 8, v21
; GFX9-NEXT: v_lshlrev_b32_e32 v23, 8, v23
+; GFX9-NEXT: v_lshlrev_b32_e32 v25, 8, v25
+; GFX9-NEXT: v_lshlrev_b32_e32 v27, 8, v27
; GFX9-NEXT: s_waitcnt vmcnt(24)
; GFX9-NEXT: v_cmp_ne_u32_e32 vcc, 0, v43
; GFX9-NEXT: s_waitcnt vmcnt(23)
@@ -211575,10 +211693,10 @@ define inreg <64 x i16> @bitcast_v128i8_to_v64i16_scalar(<128 x i8> inreg %a, i3
; GFX9-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:608 ; 4-byte Folded Spill
; GFX9-NEXT: s_waitcnt vmcnt(23)
; GFX9-NEXT: v_lshlrev_b32_e32 v1, 8, v52
-; GFX9-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:600 ; 4-byte Folded Spill
+; GFX9-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:604 ; 4-byte Folded Spill
; GFX9-NEXT: s_waitcnt vmcnt(23)
; GFX9-NEXT: v_lshlrev_b32_e32 v1, 8, v51
-; GFX9-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:604 ; 4-byte Folded Spill
+; GFX9-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:600 ; 4-byte Folded Spill
; GFX9-NEXT: s_waitcnt vmcnt(23)
; GFX9-NEXT: v_lshlrev_b32_e32 v1, 8, v50
; GFX9-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:496 ; 4-byte Folded Spill
@@ -211590,7 +211708,7 @@ define inreg <64 x i16> @bitcast_v128i8_to_v64i16_scalar(<128 x i8> inreg %a, i3
; GFX9-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:596 ; 4-byte Folded Spill
; GFX9-NEXT: s_waitcnt vmcnt(23)
; GFX9-NEXT: v_lshlrev_b32_e32 v1, 8, v39
-; GFX9-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:592 ; 4-byte Folded Spill
+; GFX9-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:588 ; 4-byte Folded Spill
; GFX9-NEXT: s_waitcnt vmcnt(23)
; GFX9-NEXT: v_lshlrev_b32_e32 v1, 8, v30
; GFX9-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:448 ; 4-byte Folded Spill
@@ -211638,7 +211756,7 @@ define inreg <64 x i16> @bitcast_v128i8_to_v64i16_scalar(<128 x i8> inreg %a, i3
; GFX9-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:584 ; 4-byte Folded Spill
; GFX9-NEXT: s_waitcnt vmcnt(6)
; GFX9-NEXT: v_lshlrev_b32_e32 v1, 8, v3
-; GFX9-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:588 ; 4-byte Folded Spill
+; GFX9-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:592 ; 4-byte Folded Spill
; GFX9-NEXT: s_waitcnt vmcnt(6)
; GFX9-NEXT: v_lshlrev_b32_e32 v1, 8, v13
; GFX9-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:536 ; 4-byte Folded Spill
@@ -211665,23 +211783,23 @@ define inreg <64 x i16> @bitcast_v128i8_to_v64i16_scalar(<128 x i8> inreg %a, i3
; GFX9-NEXT: v_lshlrev_b32_e32 v11, 8, v11
; GFX9-NEXT: s_waitcnt vmcnt(6)
; GFX9-NEXT: v_lshlrev_b32_e32 v1, 8, v1
-; GFX9-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:556 ; 4-byte Folded Spill
+; GFX9-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:572 ; 4-byte Folded Spill
; GFX9-NEXT: s_waitcnt vmcnt(6)
; GFX9-NEXT: v_lshlrev_b32_e32 v1, 8, v15
-; GFX9-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:572 ; 4-byte Folded Spill
+; GFX9-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:568 ; 4-byte Folded Spill
; GFX9-NEXT: s_waitcnt vmcnt(6)
; GFX9-NEXT: v_lshlrev_b32_e32 v1, 8, v3
; GFX9-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:552 ; 4-byte Folded Spill
; GFX9-NEXT: s_waitcnt vmcnt(6)
; GFX9-NEXT: v_lshlrev_b32_e32 v1, 8, v13
-; GFX9-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:568 ; 4-byte Folded Spill
+; GFX9-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:564 ; 4-byte Folded Spill
; GFX9-NEXT: s_waitcnt vmcnt(6)
; GFX9-NEXT: v_lshlrev_b32_e32 v1, 8, v5
; GFX9-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:548 ; 4-byte Folded Spill
; GFX9-NEXT: s_waitcnt vmcnt(6)
; GFX9-NEXT: v_lshlrev_b32_e32 v1, 8, v9
-; GFX9-NEXT: buffer_store_dword v11, off, s[0:3], s32 offset:564 ; 4-byte Folded Spill
-; GFX9-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:560 ; 4-byte Folded Spill
+; GFX9-NEXT: buffer_store_dword v11, off, s[0:3], s32 offset:560 ; 4-byte Folded Spill
+; GFX9-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:556 ; 4-byte Folded Spill
; GFX9-NEXT: buffer_load_ushort v3, off, s[0:3], s32 offset:312
; GFX9-NEXT: s_nop 0
; GFX9-NEXT: buffer_load_ushort v1, off, s[0:3], s32 offset:320
@@ -211694,48 +211812,49 @@ define inreg <64 x i16> @bitcast_v128i8_to_v64i16_scalar(<128 x i8> inreg %a, i3
; GFX9-NEXT: s_waitcnt vmcnt(15)
; GFX9-NEXT: v_lshlrev_b32_e32 v7, 8, v7
; GFX9-NEXT: s_waitcnt vmcnt(7)
-; GFX9-NEXT: v_lshlrev_b32_e32 v3, 8, v3
+; GFX9-NEXT: v_lshlrev_b32_e32 v9, 8, v3
; GFX9-NEXT: s_waitcnt vmcnt(6)
-; GFX9-NEXT: v_lshlrev_b32_e32 v9, 8, v1
+; GFX9-NEXT: v_lshlrev_b32_e32 v1, 8, v1
+; GFX9-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:520 ; 4-byte Folded Spill
; GFX9-NEXT: buffer_load_ushort v52, off, s[0:3], s32 offset:44
; GFX9-NEXT: buffer_load_ushort v50, off, s[0:3], s32 offset:52
; GFX9-NEXT: buffer_load_ushort v49, off, s[0:3], s32 offset:60
; GFX9-NEXT: buffer_load_ushort v15, off, s[0:3], s32 offset:68
; GFX9-NEXT: buffer_load_ushort v48, off, s[0:3], s32 offset:76
; GFX9-NEXT: buffer_load_ushort v39, off, s[0:3], s32 offset:84
-; GFX9-NEXT: buffer_load_ushort v25, off, s[0:3], s32 offset:92
+; GFX9-NEXT: buffer_load_ushort v40, off, s[0:3], s32 offset:92
; GFX9-NEXT: buffer_load_ushort v34, off, s[0:3], s32 offset:100
; GFX9-NEXT: buffer_load_ushort v33, off, s[0:3], s32 offset:108
-; GFX9-NEXT: buffer_load_ushort v45, off, s[0:3], s32 offset:116
+; GFX9-NEXT: buffer_load_ushort v46, off, s[0:3], s32 offset:116
; GFX9-NEXT: buffer_load_ushort v43, off, s[0:3], s32 offset:124
; GFX9-NEXT: buffer_load_ushort v38, off, s[0:3], s32 offset:132
-; GFX9-NEXT: buffer_load_ushort v56, off, s[0:3], s32 offset:140
+; GFX9-NEXT: buffer_load_ushort v61, off, s[0:3], s32 offset:140
; GFX9-NEXT: buffer_load_ushort v47, off, s[0:3], s32 offset:148
-; GFX9-NEXT: buffer_load_ushort v61, off, s[0:3], s32 offset:156
+; GFX9-NEXT: buffer_load_ushort v62, off, s[0:3], s32 offset:156
; GFX9-NEXT: buffer_load_ushort v1, off, s[0:3], s32 offset:164
-; GFX9-NEXT: s_waitcnt vmcnt(21)
-; GFX9-NEXT: v_lshlrev_b32_e32 v5, 8, v5
+; GFX9-NEXT: s_waitcnt vmcnt(22)
+; GFX9-NEXT: v_lshlrev_b32_e32 v3, 8, v5
; GFX9-NEXT: s_waitcnt vmcnt(0)
; GFX9-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:576 ; 4-byte Folded Spill
-; GFX9-NEXT: buffer_load_ushort v31, off, s[0:3], s32 offset:172
-; GFX9-NEXT: buffer_load_ushort v32, off, s[0:3], s32 offset:180
-; GFX9-NEXT: buffer_load_ushort v51, off, s[0:3], s32 offset:188
-; GFX9-NEXT: buffer_load_ushort v30, off, s[0:3], s32 offset:196
-; GFX9-NEXT: buffer_load_ushort v44, off, s[0:3], s32 offset:204
+; GFX9-NEXT: buffer_load_ushort v57, off, s[0:3], s32 offset:172
+; GFX9-NEXT: buffer_load_ushort v31, off, s[0:3], s32 offset:180
+; GFX9-NEXT: buffer_load_ushort v36, off, s[0:3], s32 offset:188
+; GFX9-NEXT: buffer_load_ushort v37, off, s[0:3], s32 offset:196
+; GFX9-NEXT: buffer_load_ushort v30, off, s[0:3], s32 offset:204
; GFX9-NEXT: buffer_load_ushort v35, off, s[0:3], s32 offset:212
-; GFX9-NEXT: buffer_load_ushort v36, off, s[0:3], s32 offset:220
-; GFX9-NEXT: buffer_load_ushort v37, off, s[0:3], s32 offset:228
+; GFX9-NEXT: buffer_load_ushort v28, off, s[0:3], s32 offset:220
+; GFX9-NEXT: buffer_load_ushort v51, off, s[0:3], s32 offset:228
; GFX9-NEXT: buffer_load_ushort v41, off, s[0:3], s32 offset:236
; GFX9-NEXT: buffer_load_ushort v54, off, s[0:3], s32 offset:244
-; GFX9-NEXT: buffer_load_ushort v29, off, s[0:3], s32 offset:252
-; GFX9-NEXT: buffer_load_ushort v27, off, s[0:3], s32 offset:260
+; GFX9-NEXT: buffer_load_ushort v44, off, s[0:3], s32 offset:252
+; GFX9-NEXT: buffer_load_ushort v45, off, s[0:3], s32 offset:260
; GFX9-NEXT: buffer_load_ushort v60, off, s[0:3], s32 offset:268
-; GFX9-NEXT: buffer_load_ushort v57, off, s[0:3], s32 offset:276
+; GFX9-NEXT: buffer_load_ushort v29, off, s[0:3], s32 offset:276
; GFX9-NEXT: buffer_load_ushort v59, off, s[0:3], s32 offset:284
; GFX9-NEXT: buffer_load_ushort v63, off, s[0:3], s32 offset:292
-; GFX9-NEXT: buffer_load_ushort v62, off, s[0:3], s32 offset:300
-; GFX9-NEXT: buffer_load_ushort v40, off, s[0:3], s32 offset:308
-; GFX9-NEXT: buffer_load_ushort v58, off, s[0:3], s32 offset:316
+; GFX9-NEXT: buffer_load_ushort v58, off, s[0:3], s32 offset:300
+; GFX9-NEXT: buffer_load_ushort v32, off, s[0:3], s32 offset:308
+; GFX9-NEXT: buffer_load_ushort v56, off, s[0:3], s32 offset:316
; GFX9-NEXT: buffer_load_ushort v1, off, s[0:3], s32 offset:324
; GFX9-NEXT: s_nop 0
; GFX9-NEXT: buffer_store_dword v10, off, s[0:3], s32 offset:620 ; 4-byte Folded Spill
@@ -211746,55 +211865,54 @@ define inreg <64 x i16> @bitcast_v128i8_to_v64i16_scalar(<128 x i8> inreg %a, i3
; GFX9-NEXT: buffer_store_dword v0, off, s[0:3], s32 offset:640 ; 4-byte Folded Spill
; GFX9-NEXT: buffer_store_dword v2, off, s[0:3], s32 offset:644 ; 4-byte Folded Spill
; GFX9-NEXT: buffer_store_dword v4, off, s[0:3], s32 offset:648 ; 4-byte Folded Spill
-; GFX9-NEXT: buffer_store_dword v12, off, s[0:3], s32 offset:652 ; 4-byte Folded Spill
+; GFX9-NEXT: buffer_store_dword v27, off, s[0:3], s32 offset:652 ; 4-byte Folded Spill
; GFX9-NEXT: buffer_store_dword v7, off, s[0:3], s32 offset:508 ; 4-byte Folded Spill
; GFX9-NEXT: s_waitcnt vmcnt(28)
-; GFX9-NEXT: buffer_store_dword v32, off, s[0:3], s32 offset:656 ; 4-byte Folded Spill
-; GFX9-NEXT: buffer_store_dword v31, off, s[0:3], s32 offset:660 ; 4-byte Folded Spill
-; GFX9-NEXT: buffer_store_dword v17, off, s[0:3], s32 offset:664 ; 4-byte Folded Spill
+; GFX9-NEXT: buffer_store_dword v31, off, s[0:3], s32 offset:656 ; 4-byte Folded Spill
+; GFX9-NEXT: buffer_store_dword v57, off, s[0:3], s32 offset:660 ; 4-byte Folded Spill
+; GFX9-NEXT: buffer_store_dword v25, off, s[0:3], s32 offset:664 ; 4-byte Folded Spill
; GFX9-NEXT: buffer_store_dword v23, off, s[0:3], s32 offset:668 ; 4-byte Folded Spill
; GFX9-NEXT: s_waitcnt vmcnt(30)
-; GFX9-NEXT: buffer_store_dword v30, off, s[0:3], s32 offset:672 ; 4-byte Folded Spill
-; GFX9-NEXT: buffer_store_dword v51, off, s[0:3], s32 offset:676 ; 4-byte Folded Spill
+; GFX9-NEXT: buffer_store_dword v37, off, s[0:3], s32 offset:672 ; 4-byte Folded Spill
+; GFX9-NEXT: buffer_store_dword v36, off, s[0:3], s32 offset:676 ; 4-byte Folded Spill
; GFX9-NEXT: buffer_store_dword v21, off, s[0:3], s32 offset:680 ; 4-byte Folded Spill
; GFX9-NEXT: buffer_store_dword v19, off, s[0:3], s32 offset:684 ; 4-byte Folded Spill
-; GFX9-NEXT: buffer_store_dword v3, off, s[0:3], s32 offset:512 ; 4-byte Folded Spill
+; GFX9-NEXT: buffer_store_dword v9, off, s[0:3], s32 offset:512 ; 4-byte Folded Spill
; GFX9-NEXT: s_waitcnt vmcnt(33)
; GFX9-NEXT: buffer_store_dword v35, off, s[0:3], s32 offset:688 ; 4-byte Folded Spill
-; GFX9-NEXT: buffer_store_dword v44, off, s[0:3], s32 offset:692 ; 4-byte Folded Spill
-; GFX9-NEXT: buffer_store_dword v18, off, s[0:3], s32 offset:696 ; 4-byte Folded Spill
-; GFX9-NEXT: buffer_store_dword v28, off, s[0:3], s32 offset:700 ; 4-byte Folded Spill
-; GFX9-NEXT: buffer_store_dword v5, off, s[0:3], s32 offset:516 ; 4-byte Folded Spill
+; GFX9-NEXT: buffer_store_dword v30, off, s[0:3], s32 offset:692 ; 4-byte Folded Spill
+; GFX9-NEXT: buffer_store_dword v17, off, s[0:3], s32 offset:696 ; 4-byte Folded Spill
+; GFX9-NEXT: buffer_store_dword v12, off, s[0:3], s32 offset:700 ; 4-byte Folded Spill
+; GFX9-NEXT: buffer_store_dword v3, off, s[0:3], s32 offset:516 ; 4-byte Folded Spill
; GFX9-NEXT: s_waitcnt vmcnt(36)
-; GFX9-NEXT: buffer_store_dword v37, off, s[0:3], s32 offset:704 ; 4-byte Folded Spill
-; GFX9-NEXT: buffer_store_dword v9, off, s[0:3], s32 offset:520 ; 4-byte Folded Spill
-; GFX9-NEXT: buffer_store_dword v36, off, s[0:3], s32 offset:708 ; 4-byte Folded Spill
+; GFX9-NEXT: buffer_store_dword v51, off, s[0:3], s32 offset:704 ; 4-byte Folded Spill
+; GFX9-NEXT: buffer_store_dword v28, off, s[0:3], s32 offset:708 ; 4-byte Folded Spill
; GFX9-NEXT: buffer_store_dword v20, off, s[0:3], s32 offset:712 ; 4-byte Folded Spill
; GFX9-NEXT: buffer_store_dword v26, off, s[0:3], s32 offset:716 ; 4-byte Folded Spill
-; GFX9-NEXT: s_waitcnt vmcnt(39)
+; GFX9-NEXT: s_waitcnt vmcnt(38)
; GFX9-NEXT: buffer_store_dword v54, off, s[0:3], s32 offset:720 ; 4-byte Folded Spill
; GFX9-NEXT: buffer_store_dword v41, off, s[0:3], s32 offset:724 ; 4-byte Folded Spill
; GFX9-NEXT: buffer_store_dword v24, off, s[0:3], s32 offset:728 ; 4-byte Folded Spill
; GFX9-NEXT: buffer_store_dword v22, off, s[0:3], s32 offset:732 ; 4-byte Folded Spill
-; GFX9-NEXT: s_waitcnt vmcnt(41)
-; GFX9-NEXT: buffer_store_dword v27, off, s[0:3], s32 offset:736 ; 4-byte Folded Spill
-; GFX9-NEXT: buffer_store_dword v29, off, s[0:3], s32 offset:740 ; 4-byte Folded Spill
-; GFX9-NEXT: s_waitcnt vmcnt(41)
-; GFX9-NEXT: buffer_store_dword v57, off, s[0:3], s32 offset:744 ; 4-byte Folded Spill
+; GFX9-NEXT: s_waitcnt vmcnt(40)
+; GFX9-NEXT: buffer_store_dword v45, off, s[0:3], s32 offset:736 ; 4-byte Folded Spill
+; GFX9-NEXT: buffer_store_dword v44, off, s[0:3], s32 offset:740 ; 4-byte Folded Spill
+; GFX9-NEXT: s_waitcnt vmcnt(40)
+; GFX9-NEXT: buffer_store_dword v29, off, s[0:3], s32 offset:744 ; 4-byte Folded Spill
; GFX9-NEXT: buffer_store_dword v60, off, s[0:3], s32 offset:748 ; 4-byte Folded Spill
-; GFX9-NEXT: s_waitcnt vmcnt(41)
+; GFX9-NEXT: s_waitcnt vmcnt(40)
; GFX9-NEXT: buffer_store_dword v63, off, s[0:3], s32 offset:752 ; 4-byte Folded Spill
; GFX9-NEXT: buffer_store_dword v59, off, s[0:3], s32 offset:756 ; 4-byte Folded Spill
-; GFX9-NEXT: s_waitcnt vmcnt(41)
-; GFX9-NEXT: buffer_store_dword v40, off, s[0:3], s32 offset:760 ; 4-byte Folded Spill
-; GFX9-NEXT: buffer_store_dword v62, off, s[0:3], s32 offset:764 ; 4-byte Folded Spill
-; GFX9-NEXT: s_waitcnt vmcnt(41)
+; GFX9-NEXT: s_waitcnt vmcnt(40)
+; GFX9-NEXT: buffer_store_dword v32, off, s[0:3], s32 offset:760 ; 4-byte Folded Spill
+; GFX9-NEXT: buffer_store_dword v58, off, s[0:3], s32 offset:764 ; 4-byte Folded Spill
+; GFX9-NEXT: s_waitcnt vmcnt(40)
; GFX9-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:524 ; 4-byte Folded Spill
-; GFX9-NEXT: buffer_store_dword v58, off, s[0:3], s32 offset:768 ; 4-byte Folded Spill
+; GFX9-NEXT: buffer_store_dword v56, off, s[0:3], s32 offset:768 ; 4-byte Folded Spill
; GFX9-NEXT: buffer_store_dword v39, off, s[0:3], s32 offset:772 ; 4-byte Folded Spill
; GFX9-NEXT: buffer_store_dword v48, off, s[0:3], s32 offset:776 ; 4-byte Folded Spill
; GFX9-NEXT: buffer_store_dword v34, off, s[0:3], s32 offset:780 ; 4-byte Folded Spill
-; GFX9-NEXT: buffer_store_dword v25, off, s[0:3], s32 offset:784 ; 4-byte Folded Spill
+; GFX9-NEXT: buffer_store_dword v40, off, s[0:3], s32 offset:784 ; 4-byte Folded Spill
; GFX9-NEXT: buffer_store_dword v15, off, s[0:3], s32 offset:788 ; 4-byte Folded Spill
; GFX9-NEXT: buffer_store_dword v49, off, s[0:3], s32 offset:792 ; 4-byte Folded Spill
; GFX9-NEXT: buffer_store_dword v50, off, s[0:3], s32 offset:796 ; 4-byte Folded Spill
@@ -211804,7 +211922,7 @@ define inreg <64 x i16> @bitcast_v128i8_to_v64i16_scalar(<128 x i8> inreg %a, i3
; GFX9-NEXT: buffer_store_dword v53, off, s[0:3], s32 offset:812 ; 4-byte Folded Spill
; GFX9-NEXT: buffer_store_dword v55, off, s[0:3], s32 offset:816 ; 4-byte Folded Spill
; GFX9-NEXT: buffer_store_dword v42, off, s[0:3], s32 offset:820 ; 4-byte Folded Spill
-; GFX9-NEXT: buffer_store_dword v46, off, s[0:3], s32 offset:824 ; 4-byte Folded Spill
+; GFX9-NEXT: buffer_store_dword v18, off, s[0:3], s32 offset:824 ; 4-byte Folded Spill
; GFX9-NEXT: buffer_store_dword v14, off, s[0:3], s32 offset:828 ; 4-byte Folded Spill
; GFX9-NEXT: buffer_store_dword v11, off, s[0:3], s32 offset:832 ; 4-byte Folded Spill
; GFX9-NEXT: s_cbranch_scc0 .LBB97_2
@@ -211817,7 +211935,7 @@ define inreg <64 x i16> @bitcast_v128i8_to_v64i16_scalar(<128 x i8> inreg %a, i3
; GFX9-NEXT: v_or_b32_sdwa v1, v2, v14 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
; GFX9-NEXT: v_and_b32_e32 v1, 0xffff, v1
; GFX9-NEXT: v_or_b32_sdwa v2, v0, v16 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
-; GFX9-NEXT: v_or_b32_sdwa v4, v4, v46 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; GFX9-NEXT: v_or_b32_sdwa v4, v4, v18 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
; GFX9-NEXT: v_or_b32_sdwa v0, v6, v22 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
; GFX9-NEXT: v_lshl_or_b32 v4, v4, 16, v1
; GFX9-NEXT: v_and_b32_e32 v0, 0xffff, v0
@@ -211854,10 +211972,10 @@ define inreg <64 x i16> @bitcast_v128i8_to_v64i16_scalar(<128 x i8> inreg %a, i3
; GFX9-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:452 ; 4-byte Folded Reload
; GFX9-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:456 ; 4-byte Folded Reload
; GFX9-NEXT: s_waitcnt vmcnt(1)
-; GFX9-NEXT: v_or_b32_sdwa v0, v0, v28 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; GFX9-NEXT: v_or_b32_sdwa v0, v0, v12 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
; GFX9-NEXT: v_and_b32_e32 v0, 0xffff, v0
; GFX9-NEXT: s_waitcnt vmcnt(0)
-; GFX9-NEXT: v_or_b32_sdwa v1, v1, v18 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; GFX9-NEXT: v_or_b32_sdwa v1, v1, v17 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
; GFX9-NEXT: v_lshl_or_b32 v7, v1, 16, v0
; GFX9-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:436 ; 4-byte Folded Reload
; GFX9-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:440 ; 4-byte Folded Reload
@@ -211873,13 +211991,13 @@ define inreg <64 x i16> @bitcast_v128i8_to_v64i16_scalar(<128 x i8> inreg %a, i3
; GFX9-NEXT: v_or_b32_sdwa v0, v0, v23 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
; GFX9-NEXT: v_and_b32_e32 v0, 0xffff, v0
; GFX9-NEXT: s_waitcnt vmcnt(0)
-; GFX9-NEXT: v_or_b32_sdwa v1, v1, v17 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; GFX9-NEXT: v_or_b32_sdwa v1, v1, v25 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
; GFX9-NEXT: v_lshl_or_b32 v9, v1, 16, v0
; GFX9-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:420 ; 4-byte Folded Reload
; GFX9-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:424 ; 4-byte Folded Reload
; GFX9-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:400 ; 4-byte Folded Reload
; GFX9-NEXT: s_waitcnt vmcnt(2)
-; GFX9-NEXT: v_or_b32_sdwa v0, v0, v12 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; GFX9-NEXT: v_or_b32_sdwa v0, v0, v27 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
; GFX9-NEXT: v_and_b32_e32 v0, 0xffff, v0
; GFX9-NEXT: s_waitcnt vmcnt(0)
; GFX9-NEXT: v_or_b32_sdwa v1, v1, v2 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
@@ -211887,7 +212005,7 @@ define inreg <64 x i16> @bitcast_v128i8_to_v64i16_scalar(<128 x i8> inreg %a, i3
; GFX9-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:416 ; 4-byte Folded Reload
; GFX9-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:504 ; 4-byte Folded Reload
; GFX9-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:492 ; 4-byte Folded Reload
-; GFX9-NEXT: buffer_load_dword v16, off, s[0:3], s32 offset:592 ; 4-byte Folded Reload
+; GFX9-NEXT: buffer_load_dword v16, off, s[0:3], s32 offset:588 ; 4-byte Folded Reload
; GFX9-NEXT: s_waitcnt vmcnt(2)
; GFX9-NEXT: v_or_b32_sdwa v0, v0, v1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
; GFX9-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:404 ; 4-byte Folded Reload
@@ -211928,8 +212046,8 @@ define inreg <64 x i16> @bitcast_v128i8_to_v64i16_scalar(<128 x i8> inreg %a, i3
; GFX9-NEXT: v_mov_b32_e32 v52, v1
; GFX9-NEXT: v_or_b32_sdwa v1, v50, v1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
; GFX9-NEXT: v_lshl_or_b32 v14, v1, 16, v0
-; GFX9-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:600 ; 4-byte Folded Reload
-; GFX9-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:604 ; 4-byte Folded Reload
+; GFX9-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:604 ; 4-byte Folded Reload
+; GFX9-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:600 ; 4-byte Folded Reload
; GFX9-NEXT: s_waitcnt vmcnt(1)
; GFX9-NEXT: v_mov_b32_e32 v50, v0
; GFX9-NEXT: v_or_b32_sdwa v0, v49, v0 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
@@ -211947,16 +212065,16 @@ define inreg <64 x i16> @bitcast_v128i8_to_v64i16_scalar(<128 x i8> inreg %a, i3
; GFX9-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:448 ; 4-byte Folded Reload
; GFX9-NEXT: s_waitcnt vmcnt(1)
; GFX9-NEXT: v_mov_b32_e32 v48, v1
-; GFX9-NEXT: v_or_b32_sdwa v1, v25, v1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; GFX9-NEXT: v_or_b32_sdwa v1, v40, v1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
; GFX9-NEXT: v_and_b32_e32 v1, 0xffff, v1
; GFX9-NEXT: v_lshl_or_b32 v17, v17, 16, v1
; GFX9-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:444 ; 4-byte Folded Reload
; GFX9-NEXT: s_waitcnt vmcnt(1)
; GFX9-NEXT: v_or_b32_sdwa v0, v33, v0 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
; GFX9-NEXT: v_and_b32_e32 v0, 0xffff, v0
-; GFX9-NEXT: v_mov_b32_e32 v33, v45
+; GFX9-NEXT: v_mov_b32_e32 v33, v46
; GFX9-NEXT: s_waitcnt vmcnt(0)
-; GFX9-NEXT: v_or_b32_sdwa v1, v45, v1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; GFX9-NEXT: v_or_b32_sdwa v1, v46, v1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
; GFX9-NEXT: v_lshl_or_b32 v18, v1, 16, v0
; GFX9-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:464 ; 4-byte Folded Reload
; GFX9-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:460 ; 4-byte Folded Reload
@@ -211969,7 +212087,7 @@ define inreg <64 x i16> @bitcast_v128i8_to_v64i16_scalar(<128 x i8> inreg %a, i3
; GFX9-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:472 ; 4-byte Folded Reload
; GFX9-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:468 ; 4-byte Folded Reload
; GFX9-NEXT: s_waitcnt vmcnt(1)
-; GFX9-NEXT: v_or_b32_sdwa v0, v56, v0 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; GFX9-NEXT: v_or_b32_sdwa v0, v42, v0 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
; GFX9-NEXT: v_and_b32_e32 v0, 0xffff, v0
; GFX9-NEXT: s_waitcnt vmcnt(0)
; GFX9-NEXT: v_or_b32_sdwa v1, v47, v1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
@@ -211978,7 +212096,7 @@ define inreg <64 x i16> @bitcast_v128i8_to_v64i16_scalar(<128 x i8> inreg %a, i3
; GFX9-NEXT: buffer_load_dword v61, off, s[0:3], s32 offset:576 ; 4-byte Folded Reload
; GFX9-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:476 ; 4-byte Folded Reload
; GFX9-NEXT: s_waitcnt vmcnt(2)
-; GFX9-NEXT: v_or_b32_sdwa v0, v42, v0 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; GFX9-NEXT: v_or_b32_sdwa v0, v62, v0 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
; GFX9-NEXT: v_and_b32_e32 v0, 0xffff, v0
; GFX9-NEXT: s_waitcnt vmcnt(0)
; GFX9-NEXT: v_or_b32_sdwa v1, v61, v1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
@@ -211986,121 +212104,122 @@ define inreg <64 x i16> @bitcast_v128i8_to_v64i16_scalar(<128 x i8> inreg %a, i3
; GFX9-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:528 ; 4-byte Folded Reload
; GFX9-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:532 ; 4-byte Folded Reload
; GFX9-NEXT: s_waitcnt vmcnt(1)
-; GFX9-NEXT: v_or_b32_sdwa v0, v31, v0 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; GFX9-NEXT: v_or_b32_sdwa v0, v57, v0 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
; GFX9-NEXT: v_and_b32_e32 v0, 0xffff, v0
; GFX9-NEXT: s_waitcnt vmcnt(0)
-; GFX9-NEXT: v_or_b32_sdwa v1, v32, v1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; GFX9-NEXT: v_or_b32_sdwa v1, v31, v1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
; GFX9-NEXT: v_lshl_or_b32 v22, v1, 16, v0
; GFX9-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:488 ; 4-byte Folded Reload
; GFX9-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:584 ; 4-byte Folded Reload
; GFX9-NEXT: s_waitcnt vmcnt(1)
-; GFX9-NEXT: v_or_b32_sdwa v0, v51, v0 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; GFX9-NEXT: v_or_b32_sdwa v0, v36, v0 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
; GFX9-NEXT: v_and_b32_e32 v0, 0xffff, v0
; GFX9-NEXT: s_waitcnt vmcnt(0)
; GFX9-NEXT: v_mov_b32_e32 v34, v1
-; GFX9-NEXT: v_or_b32_sdwa v1, v30, v1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; GFX9-NEXT: v_or_b32_sdwa v1, v37, v1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
; GFX9-NEXT: v_lshl_or_b32 v23, v1, 16, v0
-; GFX9-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:588 ; 4-byte Folded Reload
-; GFX9-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:572 ; 4-byte Folded Reload
-; GFX9-NEXT: buffer_load_dword v45, off, s[0:3], s32 offset:536 ; 4-byte Folded Reload
-; GFX9-NEXT: buffer_load_dword v51, off, s[0:3], s32 offset:564 ; 4-byte Folded Reload
-; GFX9-NEXT: buffer_load_dword v32, off, s[0:3], s32 offset:552 ; 4-byte Folded Reload
+; GFX9-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:592 ; 4-byte Folded Reload
+; GFX9-NEXT: buffer_load_dword v57, off, s[0:3], s32 offset:572 ; 4-byte Folded Reload
+; GFX9-NEXT: buffer_load_dword v46, off, s[0:3], s32 offset:536 ; 4-byte Folded Reload
+; GFX9-NEXT: buffer_load_dword v36, off, s[0:3], s32 offset:544 ; 4-byte Folded Reload
+; GFX9-NEXT: buffer_load_dword v37, off, s[0:3], s32 offset:540 ; 4-byte Folded Reload
; GFX9-NEXT: s_waitcnt vmcnt(4)
-; GFX9-NEXT: v_mov_b32_e32 v46, v0
-; GFX9-NEXT: v_or_b32_sdwa v0, v44, v0 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; GFX9-NEXT: v_mov_b32_e32 v40, v0
+; GFX9-NEXT: v_or_b32_sdwa v0, v30, v0 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
; GFX9-NEXT: v_and_b32_e32 v0, 0xffff, v0
; GFX9-NEXT: s_waitcnt vmcnt(2)
-; GFX9-NEXT: v_or_b32_sdwa v1, v35, v45 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; GFX9-NEXT: v_or_b32_sdwa v1, v35, v46 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
; GFX9-NEXT: v_lshl_or_b32 v24, v1, 16, v0
; GFX9-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:580 ; 4-byte Folded Reload
-; GFX9-NEXT: buffer_load_dword v44, off, s[0:3], s32 offset:568 ; 4-byte Folded Reload
-; GFX9-NEXT: v_mov_b32_e32 v35, v45
-; GFX9-NEXT: v_mov_b32_e32 v45, v61
-; GFX9-NEXT: v_mov_b32_e32 v61, v42
+; GFX9-NEXT: s_waitcnt vmcnt(2)
+; GFX9-NEXT: v_or_b32_sdwa v1, v51, v36 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; GFX9-NEXT: buffer_load_dword v51, off, s[0:3], s32 offset:560 ; 4-byte Folded Reload
; GFX9-NEXT: s_waitcnt vmcnt(1)
; GFX9-NEXT: v_mov_b32_e32 v38, v0
-; GFX9-NEXT: v_or_b32_sdwa v0, v36, v0 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
-; GFX9-NEXT: buffer_load_dword v36, off, s[0:3], s32 offset:544 ; 4-byte Folded Reload
+; GFX9-NEXT: v_or_b32_sdwa v0, v28, v0 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
; GFX9-NEXT: v_and_b32_e32 v0, 0xffff, v0
-; GFX9-NEXT: s_waitcnt vmcnt(0)
-; GFX9-NEXT: v_or_b32_sdwa v1, v37, v36 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
-; GFX9-NEXT: buffer_load_dword v37, off, s[0:3], s32 offset:540 ; 4-byte Folded Reload
; GFX9-NEXT: v_lshl_or_b32 v25, v1, 16, v0
-; GFX9-NEXT: v_or_b32_sdwa v1, v54, v51 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
-; GFX9-NEXT: v_mov_b32_e32 v54, v2
-; GFX9-NEXT: s_waitcnt vmcnt(0)
; GFX9-NEXT: v_or_b32_sdwa v0, v41, v37 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
-; GFX9-NEXT: buffer_load_dword v41, off, s[0:3], s32 offset:556 ; 4-byte Folded Reload
; GFX9-NEXT: v_and_b32_e32 v0, 0xffff, v0
-; GFX9-NEXT: v_lshl_or_b32 v26, v1, 16, v0
-; GFX9-NEXT: v_or_b32_sdwa v1, v27, v2 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; GFX9-NEXT: v_mov_b32_e32 v41, v57
; GFX9-NEXT: s_waitcnt vmcnt(0)
-; GFX9-NEXT: v_or_b32_sdwa v0, v29, v41 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; GFX9-NEXT: v_or_b32_sdwa v1, v54, v51 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; GFX9-NEXT: buffer_load_dword v54, off, s[0:3], s32 offset:568 ; 4-byte Folded Reload
+; GFX9-NEXT: v_lshl_or_b32 v26, v1, 16, v0
+; GFX9-NEXT: v_or_b32_sdwa v0, v44, v57 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; GFX9-NEXT: buffer_load_dword v44, off, s[0:3], s32 offset:564 ; 4-byte Folded Reload
; GFX9-NEXT: v_and_b32_e32 v0, 0xffff, v0
+; GFX9-NEXT: s_waitcnt vmcnt(1)
+; GFX9-NEXT: v_or_b32_sdwa v1, v45, v54 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; GFX9-NEXT: buffer_load_dword v45, off, s[0:3], s32 offset:552 ; 4-byte Folded Reload
; GFX9-NEXT: v_lshl_or_b32 v27, v1, 16, v0
-; GFX9-NEXT: v_or_b32_sdwa v0, v60, v32 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; GFX9-NEXT: s_waitcnt vmcnt(0)
+; GFX9-NEXT: v_or_b32_sdwa v0, v60, v45 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
; GFX9-NEXT: buffer_load_dword v60, off, s[0:3], s32 offset:548 ; 4-byte Folded Reload
-; GFX9-NEXT: v_or_b32_sdwa v1, v57, v44 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
-; GFX9-NEXT: buffer_load_dword v57, off, s[0:3], s32 offset:560 ; 4-byte Folded Reload
+; GFX9-NEXT: v_or_b32_sdwa v1, v29, v44 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
; GFX9-NEXT: v_and_b32_e32 v0, 0xffff, v0
; GFX9-NEXT: v_lshl_or_b32 v28, v1, 16, v0
-; GFX9-NEXT: s_waitcnt vmcnt(1)
+; GFX9-NEXT: s_waitcnt vmcnt(0)
; GFX9-NEXT: v_or_b32_sdwa v0, v59, v60 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; GFX9-NEXT: buffer_load_dword v59, off, s[0:3], s32 offset:556 ; 4-byte Folded Reload
; GFX9-NEXT: v_and_b32_e32 v0, 0xffff, v0
; GFX9-NEXT: s_waitcnt vmcnt(0)
-; GFX9-NEXT: v_or_b32_sdwa v1, v63, v57 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; GFX9-NEXT: v_or_b32_sdwa v1, v63, v59 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; GFX9-NEXT: v_mov_b32_e32 v57, v59
; GFX9-NEXT: v_lshl_or_b32 v29, v1, 16, v0
; GFX9-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:508 ; 4-byte Folded Reload
; GFX9-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:512 ; 4-byte Folded Reload
; GFX9-NEXT: s_waitcnt vmcnt(1)
-; GFX9-NEXT: v_or_b32_sdwa v0, v62, v0 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
-; GFX9-NEXT: v_and_b32_e32 v0, 0xffff, v0
+; GFX9-NEXT: v_or_b32_sdwa v0, v58, v0 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
; GFX9-NEXT: s_waitcnt vmcnt(0)
-; GFX9-NEXT: v_or_b32_sdwa v1, v40, v1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; GFX9-NEXT: v_or_b32_sdwa v1, v32, v1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; GFX9-NEXT: v_and_b32_e32 v0, 0xffff, v0
; GFX9-NEXT: v_lshl_or_b32 v30, v1, 16, v0
; GFX9-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:520 ; 4-byte Folded Reload
; GFX9-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:524 ; 4-byte Folded Reload
; GFX9-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:516 ; 4-byte Folded Reload
; GFX9-NEXT: s_waitcnt vmcnt(2)
-; GFX9-NEXT: v_or_b32_sdwa v0, v58, v0 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
-; GFX9-NEXT: v_and_b32_e32 v0, 0xffff, v0
+; GFX9-NEXT: v_or_b32_sdwa v0, v56, v0 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
; GFX9-NEXT: s_waitcnt vmcnt(0)
; GFX9-NEXT: v_or_b32_sdwa v1, v1, v2 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; GFX9-NEXT: v_mov_b32_e32 v2, s6
+; GFX9-NEXT: v_mov_b32_e32 v56, v42
+; GFX9-NEXT: v_and_b32_e32 v0, 0xffff, v0
; GFX9-NEXT: v_lshl_or_b32 v31, v1, 16, v0
; GFX9-NEXT: v_mov_b32_e32 v0, s4
; GFX9-NEXT: v_mov_b32_e32 v1, s5
-; GFX9-NEXT: v_mov_b32_e32 v2, s6
; GFX9-NEXT: s_mov_b64 s[4:5], 0
; GFX9-NEXT: s_branch .LBB97_3
; GFX9-NEXT: .LBB97_2:
; GFX9-NEXT: buffer_load_dword v55, off, s[0:3], s32 offset:616 ; 4-byte Folded Reload
; GFX9-NEXT: buffer_load_dword v53, off, s[0:3], s32 offset:612 ; 4-byte Folded Reload
; GFX9-NEXT: buffer_load_dword v52, off, s[0:3], s32 offset:608 ; 4-byte Folded Reload
-; GFX9-NEXT: buffer_load_dword v50, off, s[0:3], s32 offset:600 ; 4-byte Folded Reload
-; GFX9-NEXT: buffer_load_dword v49, off, s[0:3], s32 offset:604 ; 4-byte Folded Reload
+; GFX9-NEXT: buffer_load_dword v50, off, s[0:3], s32 offset:604 ; 4-byte Folded Reload
+; GFX9-NEXT: buffer_load_dword v49, off, s[0:3], s32 offset:600 ; 4-byte Folded Reload
; GFX9-NEXT: buffer_load_dword v48, off, s[0:3], s32 offset:596 ; 4-byte Folded Reload
-; GFX9-NEXT: buffer_load_dword v39, off, s[0:3], s32 offset:592 ; 4-byte Folded Reload
+; GFX9-NEXT: buffer_load_dword v39, off, s[0:3], s32 offset:588 ; 4-byte Folded Reload
; GFX9-NEXT: buffer_load_dword v38, off, s[0:3], s32 offset:580 ; 4-byte Folded Reload
-; GFX9-NEXT: buffer_load_dword v46, off, s[0:3], s32 offset:588 ; 4-byte Folded Reload
+; GFX9-NEXT: buffer_load_dword v40, off, s[0:3], s32 offset:592 ; 4-byte Folded Reload
; GFX9-NEXT: buffer_load_dword v34, off, s[0:3], s32 offset:584 ; 4-byte Folded Reload
-; GFX9-NEXT: v_mov_b32_e32 v33, v45
-; GFX9-NEXT: buffer_load_dword v35, off, s[0:3], s32 offset:536 ; 4-byte Folded Reload
+; GFX9-NEXT: v_mov_b32_e32 v33, v46
+; GFX9-NEXT: buffer_load_dword v46, off, s[0:3], s32 offset:536 ; 4-byte Folded Reload
; GFX9-NEXT: buffer_load_dword v36, off, s[0:3], s32 offset:544 ; 4-byte Folded Reload
; GFX9-NEXT: buffer_load_dword v37, off, s[0:3], s32 offset:540 ; 4-byte Folded Reload
-; GFX9-NEXT: buffer_load_dword v51, off, s[0:3], s32 offset:564 ; 4-byte Folded Reload
-; GFX9-NEXT: buffer_load_dword v54, off, s[0:3], s32 offset:572 ; 4-byte Folded Reload
-; GFX9-NEXT: buffer_load_dword v41, off, s[0:3], s32 offset:556 ; 4-byte Folded Reload
-; GFX9-NEXT: buffer_load_dword v44, off, s[0:3], s32 offset:568 ; 4-byte Folded Reload
-; GFX9-NEXT: buffer_load_dword v32, off, s[0:3], s32 offset:552 ; 4-byte Folded Reload
-; GFX9-NEXT: buffer_load_dword v57, off, s[0:3], s32 offset:560 ; 4-byte Folded Reload
+; GFX9-NEXT: buffer_load_dword v51, off, s[0:3], s32 offset:560 ; 4-byte Folded Reload
+; GFX9-NEXT: buffer_load_dword v54, off, s[0:3], s32 offset:568 ; 4-byte Folded Reload
+; GFX9-NEXT: buffer_load_dword v41, off, s[0:3], s32 offset:572 ; 4-byte Folded Reload
+; GFX9-NEXT: buffer_load_dword v44, off, s[0:3], s32 offset:564 ; 4-byte Folded Reload
+; GFX9-NEXT: buffer_load_dword v45, off, s[0:3], s32 offset:552 ; 4-byte Folded Reload
+; GFX9-NEXT: v_mov_b32_e32 v56, v61
+; GFX9-NEXT: buffer_load_dword v57, off, s[0:3], s32 offset:556 ; 4-byte Folded Reload
; GFX9-NEXT: buffer_load_dword v60, off, s[0:3], s32 offset:548 ; 4-byte Folded Reload
-; GFX9-NEXT: buffer_load_dword v45, off, s[0:3], s32 offset:576 ; 4-byte Folded Reload
+; GFX9-NEXT: buffer_load_dword v61, off, s[0:3], s32 offset:576 ; 4-byte Folded Reload
; GFX9-NEXT: s_mov_b64 s[4:5], -1
; GFX9-NEXT: ; implicit-def: $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7_vgpr8_vgpr9_vgpr10_vgpr11_vgpr12_vgpr13_vgpr14_vgpr15
; GFX9-NEXT: ; implicit-def: $vgpr16_vgpr17_vgpr18_vgpr19_vgpr20_vgpr21_vgpr22_vgpr23_vgpr24_vgpr25_vgpr26_vgpr27_vgpr28_vgpr29_vgpr30_vgpr31
; GFX9-NEXT: .LBB97_3: ; %Flow
; GFX9-NEXT: buffer_load_dword v42, off, s[0:3], s32 offset:412 ; 4-byte Folded Reload
-; GFX9-NEXT: buffer_load_dword v40, off, s[0:3], s32 offset:408 ; 4-byte Folded Reload
+; GFX9-NEXT: buffer_load_dword v32, off, s[0:3], s32 offset:408 ; 4-byte Folded Reload
; GFX9-NEXT: buffer_load_dword v58, off, s[0:3], s32 offset:404 ; 4-byte Folded Reload
; GFX9-NEXT: buffer_load_dword v63, off, s[0:3], s32 offset:400 ; 4-byte Folded Reload
; GFX9-NEXT: s_andn2_b64 vcc, exec, s[4:5]
@@ -212303,7 +212422,7 @@ define inreg <64 x i16> @bitcast_v128i8_to_v64i16_scalar(<128 x i8> inreg %a, i3
; GFX9-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:816 ; 4-byte Folded Reload
; GFX9-NEXT: s_waitcnt vmcnt(1)
; GFX9-NEXT: v_add_u32_e32 v0, 3, v0
-; GFX9-NEXT: v_or_b32_sdwa v0, v40, v0 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:BYTE_0
+; GFX9-NEXT: v_or_b32_sdwa v0, v32, v0 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:BYTE_0
; GFX9-NEXT: s_waitcnt vmcnt(0)
; GFX9-NEXT: v_add_u32_e32 v1, 3, v1
; GFX9-NEXT: v_or_b32_sdwa v1, v2, v1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:BYTE_0
@@ -212363,11 +212482,11 @@ define inreg <64 x i16> @bitcast_v128i8_to_v64i16_scalar(<128 x i8> inreg %a, i3
; GFX9-NEXT: v_and_b32_e32 v0, 0xffff, v0
; GFX9-NEXT: s_waitcnt vmcnt(0)
; GFX9-NEXT: v_add_u32_e32 v26, 3, v26
-; GFX9-NEXT: v_or_b32_sdwa v48, v46, v26 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:BYTE_0
+; GFX9-NEXT: v_or_b32_sdwa v48, v40, v26 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:BYTE_0
; GFX9-NEXT: buffer_load_dword v26, off, s[0:3], s32 offset:688 ; 4-byte Folded Reload
; GFX9-NEXT: s_waitcnt vmcnt(0)
; GFX9-NEXT: v_add_u32_e32 v26, 3, v26
-; GFX9-NEXT: v_or_b32_sdwa v49, v35, v26 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:BYTE_0
+; GFX9-NEXT: v_or_b32_sdwa v49, v46, v26 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:BYTE_0
; GFX9-NEXT: buffer_load_dword v26, off, s[0:3], s32 offset:676 ; 4-byte Folded Reload
; GFX9-NEXT: s_waitcnt vmcnt(0)
; GFX9-NEXT: v_add_u32_e32 v26, 3, v26
@@ -212402,7 +212521,7 @@ define inreg <64 x i16> @bitcast_v128i8_to_v64i16_scalar(<128 x i8> inreg %a, i3
; GFX9-NEXT: v_or_b32_sdwa v53, v27, v26 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:BYTE_0
; GFX9-NEXT: buffer_load_dword v27, off, s[0:3], s32 offset:480 ; 4-byte Folded Reload
; GFX9-NEXT: v_add_u32_e32 v24, 3, v24
-; GFX9-NEXT: v_add_u32_e32 v26, 3, v61
+; GFX9-NEXT: v_add_u32_e32 v26, 3, v62
; GFX9-NEXT: v_or_b32_sdwa v24, v54, v24 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:BYTE_0
; GFX9-NEXT: v_add_u32_e32 v36, 0x300, v24
; GFX9-NEXT: v_add_u32_e32 v24, 0x300, v48
@@ -212411,7 +212530,7 @@ define inreg <64 x i16> @bitcast_v128i8_to_v64i16_scalar(<128 x i8> inreg %a, i3
; GFX9-NEXT: s_waitcnt vmcnt(0)
; GFX9-NEXT: v_or_b32_sdwa v54, v27, v26 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:BYTE_0
; GFX9-NEXT: buffer_load_dword v27, off, s[0:3], s32 offset:476 ; 4-byte Folded Reload
-; GFX9-NEXT: v_add_u32_e32 v26, 3, v45
+; GFX9-NEXT: v_add_u32_e32 v26, 3, v61
; GFX9-NEXT: v_add_u32_e32 v20, 3, v20
; GFX9-NEXT: v_or_b32_sdwa v20, v57, v20 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:BYTE_0
; GFX9-NEXT: v_add_u32_e32 v34, 0x300, v20
@@ -212420,7 +212539,7 @@ define inreg <64 x i16> @bitcast_v128i8_to_v64i16_scalar(<128 x i8> inreg %a, i3
; GFX9-NEXT: buffer_load_dword v27, off, s[0:3], s32 offset:472 ; 4-byte Folded Reload
; GFX9-NEXT: v_add_u32_e32 v26, 3, v56
; GFX9-NEXT: v_add_u32_e32 v21, 3, v21
-; GFX9-NEXT: v_or_b32_sdwa v21, v32, v21 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:BYTE_0
+; GFX9-NEXT: v_or_b32_sdwa v21, v45, v21 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:BYTE_0
; GFX9-NEXT: v_add_u32_e32 v28, 0x300, v21
; GFX9-NEXT: v_add_u32_e32 v21, 0x300, v54
; GFX9-NEXT: v_and_b32_e32 v21, 0xffff, v21
@@ -216299,1192 +216418,1031 @@ define <128 x i8> @bitcast_v64i16_to_v128i8(<64 x i16> %a, i32 %b) {
; VI-NEXT: buffer_store_dword v61, off, s[0:3], s32 offset:20 ; 4-byte Folded Spill
; VI-NEXT: buffer_store_dword v62, off, s[0:3], s32 offset:16 ; 4-byte Folded Spill
; VI-NEXT: buffer_store_dword v63, off, s[0:3], s32 offset:12 ; 4-byte Folded Spill
-; VI-NEXT: buffer_load_dword v31, off, s[0:3], s32 offset:8
-; VI-NEXT: buffer_load_dword v37, off, s[0:3], s32 offset:4
-; VI-NEXT: buffer_load_dword v36, off, s[0:3], s32
-; VI-NEXT: v_lshrrev_b32_e32 v32, 16, v15
-; VI-NEXT: buffer_store_dword v32, off, s[0:3], s32 offset:192 ; 4-byte Folded Spill
-; VI-NEXT: v_lshrrev_b32_e32 v32, 16, v13
-; VI-NEXT: buffer_store_dword v32, off, s[0:3], s32 offset:196 ; 4-byte Folded Spill
-; VI-NEXT: v_lshrrev_b32_e32 v32, 16, v11
-; VI-NEXT: buffer_store_dword v32, off, s[0:3], s32 offset:200 ; 4-byte Folded Spill
-; VI-NEXT: v_lshrrev_b32_e32 v32, 16, v9
-; VI-NEXT: buffer_store_dword v32, off, s[0:3], s32 offset:204 ; 4-byte Folded Spill
-; VI-NEXT: v_lshrrev_b32_e32 v32, 16, v7
-; VI-NEXT: buffer_store_dword v32, off, s[0:3], s32 offset:208 ; 4-byte Folded Spill
-; VI-NEXT: v_lshrrev_b32_e32 v32, 16, v5
-; VI-NEXT: buffer_store_dword v32, off, s[0:3], s32 offset:212 ; 4-byte Folded Spill
-; VI-NEXT: v_lshrrev_b32_e32 v32, 16, v3
-; VI-NEXT: buffer_store_dword v32, off, s[0:3], s32 offset:216 ; 4-byte Folded Spill
-; VI-NEXT: v_lshrrev_b32_e32 v32, 16, v1
-; VI-NEXT: buffer_store_dword v32, off, s[0:3], s32 offset:220 ; 4-byte Folded Spill
-; VI-NEXT: v_lshrrev_b32_e32 v32, 16, v29
-; VI-NEXT: buffer_store_dword v32, off, s[0:3], s32 offset:168 ; 4-byte Folded Spill
-; VI-NEXT: v_lshrrev_b32_e32 v32, 16, v28
-; VI-NEXT: buffer_store_dword v32, off, s[0:3], s32 offset:156 ; 4-byte Folded Spill
-; VI-NEXT: v_lshrrev_b32_e32 v32, 16, v27
-; VI-NEXT: buffer_store_dword v32, off, s[0:3], s32 offset:172 ; 4-byte Folded Spill
-; VI-NEXT: v_lshrrev_b32_e32 v32, 16, v26
-; VI-NEXT: buffer_store_dword v32, off, s[0:3], s32 offset:160 ; 4-byte Folded Spill
-; VI-NEXT: v_lshrrev_b32_e32 v32, 16, v25
-; VI-NEXT: buffer_store_dword v32, off, s[0:3], s32 offset:180 ; 4-byte Folded Spill
-; VI-NEXT: v_lshrrev_b32_e32 v32, 16, v24
-; VI-NEXT: buffer_store_dword v32, off, s[0:3], s32 offset:164 ; 4-byte Folded Spill
-; VI-NEXT: v_lshrrev_b32_e32 v32, 16, v23
-; VI-NEXT: v_lshrrev_b32_e32 v60, 16, v16
-; VI-NEXT: v_lshrrev_b32_e32 v33, 16, v14
-; VI-NEXT: v_lshrrev_b32_e32 v63, 16, v12
-; VI-NEXT: v_lshrrev_b32_e32 v59, 16, v10
-; VI-NEXT: v_lshrrev_b32_e32 v34, 16, v8
-; VI-NEXT: v_lshrrev_b32_e32 v47, 16, v6
-; VI-NEXT: v_lshrrev_b32_e32 v56, 16, v4
-; VI-NEXT: v_lshrrev_b32_e32 v57, 16, v2
-; VI-NEXT: v_lshrrev_b32_e32 v46, 16, v30
-; VI-NEXT: buffer_store_dword v32, off, s[0:3], s32 offset:184 ; 4-byte Folded Spill
-; VI-NEXT: v_lshrrev_b32_e32 v32, 16, v22
-; VI-NEXT: v_lshrrev_b32_e32 v49, 16, v21
-; VI-NEXT: v_lshrrev_b32_e32 v53, 16, v20
-; VI-NEXT: v_lshrrev_b32_e32 v51, 16, v19
-; VI-NEXT: v_lshrrev_b32_e32 v52, 16, v18
-; VI-NEXT: buffer_store_dword v32, off, s[0:3], s32 offset:176 ; 4-byte Folded Spill
+; VI-NEXT: buffer_load_dword v33, off, s[0:3], s32 offset:8
+; VI-NEXT: buffer_load_dword v32, off, s[0:3], s32 offset:4
+; VI-NEXT: buffer_load_dword v31, off, s[0:3], s32
+; VI-NEXT: v_lshrrev_b32_e32 v34, 16, v16
+; VI-NEXT: buffer_store_dword v34, off, s[0:3], s32 offset:152 ; 4-byte Folded Spill
+; VI-NEXT: v_lshrrev_b32_e32 v34, 16, v15
+; VI-NEXT: buffer_store_dword v34, off, s[0:3], s32 offset:164 ; 4-byte Folded Spill
+; VI-NEXT: v_lshrrev_b32_e32 v34, 16, v13
+; VI-NEXT: buffer_store_dword v34, off, s[0:3], s32 offset:160 ; 4-byte Folded Spill
+; VI-NEXT: v_lshrrev_b32_e32 v34, 16, v11
+; VI-NEXT: buffer_store_dword v34, off, s[0:3], s32 offset:168 ; 4-byte Folded Spill
+; VI-NEXT: v_lshrrev_b32_e32 v34, 16, v9
+; VI-NEXT: buffer_store_dword v34, off, s[0:3], s32 offset:172 ; 4-byte Folded Spill
+; VI-NEXT: v_lshrrev_b32_e32 v34, 16, v7
+; VI-NEXT: buffer_store_dword v34, off, s[0:3], s32 offset:176 ; 4-byte Folded Spill
+; VI-NEXT: v_lshrrev_b32_e32 v34, 16, v5
+; VI-NEXT: buffer_store_dword v34, off, s[0:3], s32 offset:180 ; 4-byte Folded Spill
+; VI-NEXT: v_lshrrev_b32_e32 v34, 16, v3
+; VI-NEXT: buffer_store_dword v34, off, s[0:3], s32 offset:184 ; 4-byte Folded Spill
+; VI-NEXT: v_lshrrev_b32_e32 v34, 16, v1
+; VI-NEXT: buffer_store_dword v34, off, s[0:3], s32 offset:188 ; 4-byte Folded Spill
+; VI-NEXT: v_lshrrev_b32_e32 v34, 16, v25
+; VI-NEXT: buffer_store_dword v34, off, s[0:3], s32 offset:128 ; 4-byte Folded Spill
+; VI-NEXT: v_lshrrev_b32_e32 v34, 16, v23
+; VI-NEXT: buffer_store_dword v34, off, s[0:3], s32 offset:132 ; 4-byte Folded Spill
+; VI-NEXT: v_lshrrev_b32_e32 v34, 16, v22
+; VI-NEXT: v_lshrrev_b32_e32 v59, 16, v14
+; VI-NEXT: v_lshrrev_b32_e32 v57, 16, v12
+; VI-NEXT: v_lshrrev_b32_e32 v62, 16, v10
+; VI-NEXT: v_lshrrev_b32_e32 v52, 16, v8
+; VI-NEXT: v_lshrrev_b32_e32 v54, 16, v6
+; VI-NEXT: v_lshrrev_b32_e32 v51, 16, v4
+; VI-NEXT: v_lshrrev_b32_e32 v56, 16, v2
+; VI-NEXT: v_lshrrev_b32_e32 v36, 16, v30
+; VI-NEXT: v_lshrrev_b32_e32 v43, 16, v29
+; VI-NEXT: v_lshrrev_b32_e32 v50, 16, v28
+; VI-NEXT: v_lshrrev_b32_e32 v47, 16, v27
+; VI-NEXT: v_lshrrev_b32_e32 v53, 16, v26
+; VI-NEXT: v_lshrrev_b32_e32 v40, 16, v24
+; VI-NEXT: buffer_store_dword v34, off, s[0:3], s32 offset:76 ; 4-byte Folded Spill
+; VI-NEXT: v_lshrrev_b32_e32 v34, 16, v21
+; VI-NEXT: v_lshrrev_b32_e32 v63, 16, v18
+; VI-NEXT: buffer_store_dword v34, off, s[0:3], s32 offset:136 ; 4-byte Folded Spill
; VI-NEXT: s_waitcnt vmcnt(14)
-; VI-NEXT: v_cmp_ne_u32_e32 vcc, 0, v31
-; VI-NEXT: v_lshrrev_b32_e32 v31, 16, v37
-; VI-NEXT: buffer_store_dword v31, off, s[0:3], s32 offset:224 ; 4-byte Folded Spill
-; VI-NEXT: v_lshrrev_b32_e32 v50, 16, v36
-; VI-NEXT: v_lshrrev_b32_e32 v31, 16, v17
-; VI-NEXT: buffer_store_dword v31, off, s[0:3], s32 offset:188 ; 4-byte Folded Spill
-; VI-NEXT: ; implicit-def: $vgpr31
-; VI-NEXT: ; kill: killed $vgpr31
-; VI-NEXT: ; implicit-def: $vgpr31
-; VI-NEXT: ; kill: killed $vgpr31
-; VI-NEXT: ; implicit-def: $vgpr31
-; VI-NEXT: ; kill: killed $vgpr31
-; VI-NEXT: ; implicit-def: $vgpr31
-; VI-NEXT: ; kill: killed $vgpr31
-; VI-NEXT: ; implicit-def: $vgpr31
-; VI-NEXT: ; kill: killed $vgpr31
-; VI-NEXT: ; implicit-def: $vgpr31
-; VI-NEXT: ; kill: killed $vgpr31
-; VI-NEXT: ; implicit-def: $vgpr31
-; VI-NEXT: ; kill: killed $vgpr31
-; VI-NEXT: ; implicit-def: $vgpr31
-; VI-NEXT: ; kill: killed $vgpr31
-; VI-NEXT: ; implicit-def: $vgpr31
-; VI-NEXT: ; kill: killed $vgpr31
-; VI-NEXT: ; implicit-def: $vgpr31
-; VI-NEXT: ; kill: killed $vgpr31
-; VI-NEXT: ; implicit-def: $vgpr31
-; VI-NEXT: ; kill: killed $vgpr31
-; VI-NEXT: ; implicit-def: $vgpr31
-; VI-NEXT: ; kill: killed $vgpr31
-; VI-NEXT: ; implicit-def: $vgpr31
-; VI-NEXT: ; kill: killed $vgpr31
-; VI-NEXT: ; implicit-def: $vgpr31
-; VI-NEXT: ; kill: killed $vgpr31
-; VI-NEXT: ; implicit-def: $vgpr31
-; VI-NEXT: ; kill: killed $vgpr31
-; VI-NEXT: ; implicit-def: $vgpr31
-; VI-NEXT: ; kill: killed $vgpr31
-; VI-NEXT: ; implicit-def: $vgpr31
-; VI-NEXT: ; kill: killed $vgpr31
-; VI-NEXT: ; implicit-def: $vgpr31
-; VI-NEXT: ; kill: killed $vgpr31
-; VI-NEXT: ; implicit-def: $vgpr31
-; VI-NEXT: ; kill: killed $vgpr31
-; VI-NEXT: ; implicit-def: $vgpr31
-; VI-NEXT: ; kill: killed $vgpr31
-; VI-NEXT: ; implicit-def: $vgpr31
-; VI-NEXT: ; kill: killed $vgpr31
-; VI-NEXT: ; implicit-def: $vgpr31
-; VI-NEXT: ; kill: killed $vgpr31
-; VI-NEXT: ; implicit-def: $vgpr31
-; VI-NEXT: ; kill: killed $vgpr31
-; VI-NEXT: ; implicit-def: $vgpr31
-; VI-NEXT: ; kill: killed $vgpr31
-; VI-NEXT: ; implicit-def: $vgpr31
-; VI-NEXT: ; kill: killed $vgpr31
-; VI-NEXT: ; implicit-def: $vgpr31
-; VI-NEXT: ; kill: killed $vgpr31
-; VI-NEXT: ; implicit-def: $vgpr31
-; VI-NEXT: ; kill: killed $vgpr31
-; VI-NEXT: ; implicit-def: $vgpr31
-; VI-NEXT: ; kill: killed $vgpr31
-; VI-NEXT: ; implicit-def: $vgpr31
-; VI-NEXT: ; kill: killed $vgpr31
-; VI-NEXT: ; implicit-def: $vgpr31
-; VI-NEXT: ; kill: killed $vgpr31
-; VI-NEXT: ; implicit-def: $vgpr31
-; VI-NEXT: ; kill: killed $vgpr31
-; VI-NEXT: ; implicit-def: $vgpr31
-; VI-NEXT: ; kill: killed $vgpr31
-; VI-NEXT: ; implicit-def: $vgpr31
-; VI-NEXT: ; kill: killed $vgpr31
-; VI-NEXT: ; implicit-def: $vgpr31
-; VI-NEXT: ; kill: killed $vgpr31
-; VI-NEXT: ; implicit-def: $vgpr31
-; VI-NEXT: ; kill: killed $vgpr31
-; VI-NEXT: ; implicit-def: $vgpr32
-; VI-NEXT: ; implicit-def: $vgpr31
-; VI-NEXT: ; kill: killed $vgpr31
-; VI-NEXT: ; implicit-def: $vgpr31
-; VI-NEXT: ; kill: killed $vgpr31
-; VI-NEXT: ; implicit-def: $vgpr31
-; VI-NEXT: ; kill: killed $vgpr31
-; VI-NEXT: ; implicit-def: $vgpr31
-; VI-NEXT: ; kill: killed $vgpr31
-; VI-NEXT: ; implicit-def: $vgpr31
-; VI-NEXT: ; kill: killed $vgpr31
-; VI-NEXT: ; implicit-def: $vgpr55
-; VI-NEXT: ; implicit-def: $vgpr31
-; VI-NEXT: ; kill: killed $vgpr31
-; VI-NEXT: ; implicit-def: $vgpr62
-; VI-NEXT: ; implicit-def: $vgpr31
-; VI-NEXT: ; implicit-def: $vgpr35
-; VI-NEXT: ; kill: killed $vgpr35
-; VI-NEXT: ; implicit-def: $vgpr38
-; VI-NEXT: ; implicit-def: $vgpr35
-; VI-NEXT: ; kill: killed $vgpr35
-; VI-NEXT: ; implicit-def: $vgpr42
-; VI-NEXT: ; implicit-def: $vgpr61
-; VI-NEXT: ; implicit-def: $vgpr35
-; VI-NEXT: ; kill: killed $vgpr35
-; VI-NEXT: ; implicit-def: $vgpr35
-; VI-NEXT: ; kill: killed $vgpr35
-; VI-NEXT: ; implicit-def: $vgpr35
-; VI-NEXT: ; kill: killed $vgpr35
-; VI-NEXT: ; implicit-def: $vgpr35
-; VI-NEXT: ; kill: killed $vgpr35
+; VI-NEXT: v_cmp_ne_u32_e32 vcc, 0, v33
+; VI-NEXT: v_lshrrev_b32_e32 v33, 16, v20
+; VI-NEXT: buffer_store_dword v33, off, s[0:3], s32 offset:84 ; 4-byte Folded Spill
+; VI-NEXT: v_lshrrev_b32_e32 v33, 16, v19
+; VI-NEXT: buffer_store_dword v33, off, s[0:3], s32 offset:144 ; 4-byte Folded Spill
+; VI-NEXT: v_lshrrev_b32_e32 v33, 16, v32
+; VI-NEXT: buffer_store_dword v33, off, s[0:3], s32 offset:80 ; 4-byte Folded Spill
+; VI-NEXT: s_waitcnt vmcnt(14)
+; VI-NEXT: v_lshrrev_b32_e32 v33, 16, v31
+; VI-NEXT: buffer_store_dword v33, off, s[0:3], s32 offset:140 ; 4-byte Folded Spill
+; VI-NEXT: v_lshrrev_b32_e32 v33, 16, v17
+; VI-NEXT: buffer_store_dword v33, off, s[0:3], s32 offset:148 ; 4-byte Folded Spill
+; VI-NEXT: ; implicit-def: $vgpr33
+; VI-NEXT: ; kill: killed $vgpr33
+; VI-NEXT: ; implicit-def: $vgpr33
+; VI-NEXT: ; kill: killed $vgpr33
+; VI-NEXT: ; implicit-def: $vgpr60
+; VI-NEXT: ; implicit-def: $vgpr33
+; VI-NEXT: ; kill: killed $vgpr33
+; VI-NEXT: ; implicit-def: $vgpr33
+; VI-NEXT: ; kill: killed $vgpr33
+; VI-NEXT: ; implicit-def: $vgpr33
+; VI-NEXT: ; kill: killed $vgpr33
+; VI-NEXT: ; implicit-def: $vgpr33
+; VI-NEXT: ; kill: killed $vgpr33
+; VI-NEXT: ; implicit-def: $vgpr33
+; VI-NEXT: ; kill: killed $vgpr33
+; VI-NEXT: ; implicit-def: $vgpr33
+; VI-NEXT: ; kill: killed $vgpr33
+; VI-NEXT: ; implicit-def: $vgpr33
+; VI-NEXT: ; kill: killed $vgpr33
+; VI-NEXT: ; implicit-def: $vgpr33
+; VI-NEXT: ; kill: killed $vgpr33
+; VI-NEXT: ; implicit-def: $vgpr33
+; VI-NEXT: ; kill: killed $vgpr33
+; VI-NEXT: ; implicit-def: $vgpr33
+; VI-NEXT: ; kill: killed $vgpr33
+; VI-NEXT: ; implicit-def: $vgpr33
+; VI-NEXT: ; kill: killed $vgpr33
+; VI-NEXT: ; implicit-def: $vgpr33
+; VI-NEXT: ; kill: killed $vgpr33
+; VI-NEXT: ; implicit-def: $vgpr33
+; VI-NEXT: ; kill: killed $vgpr33
+; VI-NEXT: ; implicit-def: $vgpr33
+; VI-NEXT: ; kill: killed $vgpr33
+; VI-NEXT: ; implicit-def: $vgpr33
+; VI-NEXT: ; kill: killed $vgpr33
+; VI-NEXT: ; implicit-def: $vgpr33
+; VI-NEXT: ; kill: killed $vgpr33
+; VI-NEXT: ; implicit-def: $vgpr33
+; VI-NEXT: ; kill: killed $vgpr33
+; VI-NEXT: ; implicit-def: $vgpr33
+; VI-NEXT: ; kill: killed $vgpr33
+; VI-NEXT: ; implicit-def: $vgpr33
+; VI-NEXT: ; kill: killed $vgpr33
+; VI-NEXT: ; implicit-def: $vgpr33
+; VI-NEXT: ; kill: killed $vgpr33
+; VI-NEXT: ; implicit-def: $vgpr33
+; VI-NEXT: ; kill: killed $vgpr33
; VI-NEXT: ; implicit-def: $vgpr58
-; VI-NEXT: ; implicit-def: $vgpr35
-; VI-NEXT: ; kill: killed $vgpr35
-; VI-NEXT: ; implicit-def: $vgpr35
-; VI-NEXT: ; kill: killed $vgpr35
-; VI-NEXT: ; implicit-def: $vgpr35
-; VI-NEXT: ; kill: killed $vgpr35
-; VI-NEXT: ; implicit-def: $vgpr35
-; VI-NEXT: ; kill: killed $vgpr35
-; VI-NEXT: ; implicit-def: $vgpr39
-; VI-NEXT: ; implicit-def: $vgpr35
-; VI-NEXT: ; kill: killed $vgpr35
-; VI-NEXT: ; implicit-def: $vgpr35
-; VI-NEXT: ; kill: killed $vgpr35
-; VI-NEXT: ; implicit-def: $vgpr35
-; VI-NEXT: ; kill: killed $vgpr35
-; VI-NEXT: ; implicit-def: $vgpr35
-; VI-NEXT: ; kill: killed $vgpr35
-; VI-NEXT: ; implicit-def: $vgpr35
-; VI-NEXT: ; implicit-def: $vgpr54
-; VI-NEXT: ; kill: killed $vgpr54
-; VI-NEXT: ; implicit-def: $vgpr48
-; VI-NEXT: ; kill: killed $vgpr48
-; VI-NEXT: ; implicit-def: $vgpr54
-; VI-NEXT: ; kill: killed $vgpr54
-; VI-NEXT: ; implicit-def: $vgpr48
-; VI-NEXT: ; kill: killed $vgpr48
-; VI-NEXT: ; implicit-def: $vgpr48
-; VI-NEXT: ; kill: killed $vgpr48
-; VI-NEXT: ; implicit-def: $vgpr54
-; VI-NEXT: ; kill: killed $vgpr54
-; VI-NEXT: ; implicit-def: $vgpr54
-; VI-NEXT: ; kill: killed $vgpr54
-; VI-NEXT: ; implicit-def: $vgpr54
-; VI-NEXT: ; kill: killed $vgpr54
-; VI-NEXT: ; implicit-def: $vgpr54
-; VI-NEXT: ; kill: killed $vgpr54
-; VI-NEXT: ; implicit-def: $vgpr48
-; VI-NEXT: ; kill: killed $vgpr48
-; VI-NEXT: ; implicit-def: $vgpr54
-; VI-NEXT: ; kill: killed $vgpr54
-; VI-NEXT: ; implicit-def: $vgpr54
-; VI-NEXT: ; kill: killed $vgpr54
-; VI-NEXT: ; implicit-def: $vgpr54
-; VI-NEXT: ; kill: killed $vgpr54
-; VI-NEXT: ; implicit-def: $vgpr54
-; VI-NEXT: ; kill: killed $vgpr54
-; VI-NEXT: ; implicit-def: $vgpr54
-; VI-NEXT: ; kill: killed $vgpr54
-; VI-NEXT: ; implicit-def: $vgpr43
-; VI-NEXT: ; implicit-def: $vgpr40
-; VI-NEXT: ; implicit-def: $vgpr48
-; VI-NEXT: buffer_store_dword v48, off, s[0:3], s32 offset:148 ; 4-byte Folded Spill
-; VI-NEXT: buffer_store_dword v49, off, s[0:3], s32 offset:152 ; 4-byte Folded Spill
+; VI-NEXT: ; implicit-def: $vgpr37
+; VI-NEXT: ; implicit-def: $vgpr55
+; VI-NEXT: ; implicit-def: $vgpr33
+; VI-NEXT: ; kill: killed $vgpr33
+; VI-NEXT: ; implicit-def: $vgpr33
+; VI-NEXT: ; kill: killed $vgpr33
; VI-NEXT: ; implicit-def: $vgpr48
-; VI-NEXT: buffer_store_dword v48, off, s[0:3], s32 offset:132 ; 4-byte Folded Spill
-; VI-NEXT: buffer_store_dword v49, off, s[0:3], s32 offset:136 ; 4-byte Folded Spill
-; VI-NEXT: ; implicit-def: $vgpr54
-; VI-NEXT: buffer_store_dword v54, off, s[0:3], s32 offset:100 ; 4-byte Folded Spill
-; VI-NEXT: buffer_store_dword v55, off, s[0:3], s32 offset:104 ; 4-byte Folded Spill
-; VI-NEXT: ; implicit-def: $vgpr54
-; VI-NEXT: buffer_store_dword v54, off, s[0:3], s32 offset:92 ; 4-byte Folded Spill
-; VI-NEXT: buffer_store_dword v55, off, s[0:3], s32 offset:96 ; 4-byte Folded Spill
-; VI-NEXT: ; implicit-def: $vgpr54
-; VI-NEXT: buffer_store_dword v54, off, s[0:3], s32 offset:84 ; 4-byte Folded Spill
-; VI-NEXT: buffer_store_dword v55, off, s[0:3], s32 offset:88 ; 4-byte Folded Spill
-; VI-NEXT: ; implicit-def: $vgpr54
-; VI-NEXT: buffer_store_dword v54, off, s[0:3], s32 offset:76 ; 4-byte Folded Spill
-; VI-NEXT: buffer_store_dword v55, off, s[0:3], s32 offset:80 ; 4-byte Folded Spill
+; VI-NEXT: ; implicit-def: $vgpr33
+; VI-NEXT: ; kill: killed $vgpr33
+; VI-NEXT: ; implicit-def: $vgpr33
+; VI-NEXT: ; kill: killed $vgpr33
+; VI-NEXT: ; implicit-def: $vgpr33
+; VI-NEXT: ; kill: killed $vgpr33
+; VI-NEXT: ; implicit-def: $vgpr33
+; VI-NEXT: ; kill: killed $vgpr33
+; VI-NEXT: ; implicit-def: $vgpr33
+; VI-NEXT: ; kill: killed $vgpr33
+; VI-NEXT: ; implicit-def: $vgpr33
+; VI-NEXT: ; kill: killed $vgpr33
+; VI-NEXT: ; implicit-def: $vgpr33
+; VI-NEXT: ; kill: killed $vgpr33
+; VI-NEXT: ; implicit-def: $vgpr33
+; VI-NEXT: ; kill: killed $vgpr33
+; VI-NEXT: ; implicit-def: $vgpr33
+; VI-NEXT: ; kill: killed $vgpr33
+; VI-NEXT: ; implicit-def: $vgpr33
+; VI-NEXT: ; kill: killed $vgpr33
+; VI-NEXT: ; implicit-def: $vgpr33
+; VI-NEXT: ; kill: killed $vgpr33
+; VI-NEXT: ; implicit-def: $vgpr33
+; VI-NEXT: ; kill: killed $vgpr33
+; VI-NEXT: ; implicit-def: $vgpr33
+; VI-NEXT: ; kill: killed $vgpr33
+; VI-NEXT: ; implicit-def: $vgpr33
+; VI-NEXT: ; kill: killed $vgpr33
+; VI-NEXT: ; implicit-def: $vgpr33
+; VI-NEXT: ; kill: killed $vgpr33
+; VI-NEXT: ; implicit-def: $vgpr33
+; VI-NEXT: ; kill: killed $vgpr33
+; VI-NEXT: ; implicit-def: $vgpr33
+; VI-NEXT: ; kill: killed $vgpr33
+; VI-NEXT: ; implicit-def: $vgpr33
+; VI-NEXT: ; kill: killed $vgpr33
+; VI-NEXT: ; implicit-def: $vgpr46
; VI-NEXT: ; implicit-def: $vgpr45
; VI-NEXT: ; implicit-def: $vgpr44
; VI-NEXT: ; implicit-def: $vgpr41
-; VI-NEXT: ; implicit-def: $vgpr54
-; VI-NEXT: ; implicit-def: $vgpr48
-; VI-NEXT: buffer_store_dword v48, off, s[0:3], s32 offset:140 ; 4-byte Folded Spill
-; VI-NEXT: buffer_store_dword v49, off, s[0:3], s32 offset:144 ; 4-byte Folded Spill
-; VI-NEXT: ; implicit-def: $vgpr48
-; VI-NEXT: buffer_store_dword v48, off, s[0:3], s32 offset:124 ; 4-byte Folded Spill
-; VI-NEXT: buffer_store_dword v49, off, s[0:3], s32 offset:128 ; 4-byte Folded Spill
-; VI-NEXT: ; implicit-def: $vgpr48
-; VI-NEXT: buffer_store_dword v48, off, s[0:3], s32 offset:116 ; 4-byte Folded Spill
-; VI-NEXT: buffer_store_dword v49, off, s[0:3], s32 offset:120 ; 4-byte Folded Spill
-; VI-NEXT: ; implicit-def: $vgpr48
-; VI-NEXT: buffer_store_dword v48, off, s[0:3], s32 offset:108 ; 4-byte Folded Spill
-; VI-NEXT: buffer_store_dword v49, off, s[0:3], s32 offset:112 ; 4-byte Folded Spill
+; VI-NEXT: ; implicit-def: $vgpr33
+; VI-NEXT: ; implicit-def: $vgpr34
+; VI-NEXT: buffer_store_dword v34, off, s[0:3], s32 offset:120 ; 4-byte Folded Spill
+; VI-NEXT: buffer_store_dword v35, off, s[0:3], s32 offset:124 ; 4-byte Folded Spill
+; VI-NEXT: ; implicit-def: $vgpr34
+; VI-NEXT: buffer_store_dword v34, off, s[0:3], s32 offset:112 ; 4-byte Folded Spill
+; VI-NEXT: buffer_store_dword v35, off, s[0:3], s32 offset:116 ; 4-byte Folded Spill
+; VI-NEXT: ; implicit-def: $vgpr34
+; VI-NEXT: buffer_store_dword v34, off, s[0:3], s32 offset:104 ; 4-byte Folded Spill
+; VI-NEXT: buffer_store_dword v35, off, s[0:3], s32 offset:108 ; 4-byte Folded Spill
+; VI-NEXT: ; implicit-def: $vgpr42
+; VI-NEXT: ; implicit-def: $vgpr39
+; VI-NEXT: ; implicit-def: $vgpr61
+; VI-NEXT: ; implicit-def: $vgpr49
+; VI-NEXT: ; implicit-def: $vgpr38
+; VI-NEXT: ; implicit-def: $vgpr35
+; VI-NEXT: ; implicit-def: $vgpr34
+; VI-NEXT: buffer_store_dword v34, off, s[0:3], s32 offset:96 ; 4-byte Folded Spill
+; VI-NEXT: buffer_store_dword v35, off, s[0:3], s32 offset:100 ; 4-byte Folded Spill
+; VI-NEXT: ; implicit-def: $vgpr34
+; VI-NEXT: buffer_store_dword v34, off, s[0:3], s32 offset:88 ; 4-byte Folded Spill
+; VI-NEXT: buffer_store_dword v35, off, s[0:3], s32 offset:92 ; 4-byte Folded Spill
; VI-NEXT: s_and_saveexec_b64 s[4:5], vcc
; VI-NEXT: s_xor_b64 s[4:5], exec, s[4:5]
; VI-NEXT: s_cbranch_execz .LBB98_2
; VI-NEXT: ; %bb.1: ; %cmp.false
-; VI-NEXT: v_lshrrev_b32_e32 v31, 24, v16
-; VI-NEXT: buffer_store_dword v31, off, s[0:3], s32 offset:268 ; 4-byte Folded Spill
-; VI-NEXT: v_lshrrev_b32_e32 v31, 8, v16
-; VI-NEXT: buffer_store_dword v31, off, s[0:3], s32 offset:228 ; 4-byte Folded Spill
-; VI-NEXT: v_lshrrev_b32_e32 v31, 8, v15
-; VI-NEXT: buffer_store_dword v31, off, s[0:3], s32 offset:232 ; 4-byte Folded Spill
-; VI-NEXT: v_lshrrev_b32_e32 v31, 24, v14
-; VI-NEXT: buffer_store_dword v31, off, s[0:3], s32 offset:280 ; 4-byte Folded Spill
-; VI-NEXT: v_lshrrev_b32_e32 v31, 8, v14
-; VI-NEXT: buffer_store_dword v31, off, s[0:3], s32 offset:236 ; 4-byte Folded Spill
-; VI-NEXT: v_lshrrev_b32_e32 v31, 8, v13
-; VI-NEXT: buffer_store_dword v31, off, s[0:3], s32 offset:240 ; 4-byte Folded Spill
-; VI-NEXT: v_lshrrev_b32_e32 v31, 24, v12
-; VI-NEXT: buffer_store_dword v31, off, s[0:3], s32 offset:288 ; 4-byte Folded Spill
-; VI-NEXT: v_lshrrev_b32_e32 v31, 8, v12
-; VI-NEXT: buffer_store_dword v31, off, s[0:3], s32 offset:244 ; 4-byte Folded Spill
-; VI-NEXT: v_lshrrev_b32_e32 v31, 8, v11
-; VI-NEXT: buffer_store_dword v31, off, s[0:3], s32 offset:248 ; 4-byte Folded Spill
-; VI-NEXT: v_lshrrev_b32_e32 v31, 24, v10
-; VI-NEXT: buffer_store_dword v31, off, s[0:3], s32 offset:292 ; 4-byte Folded Spill
-; VI-NEXT: v_lshrrev_b32_e32 v31, 8, v10
-; VI-NEXT: buffer_store_dword v31, off, s[0:3], s32 offset:252 ; 4-byte Folded Spill
-; VI-NEXT: v_lshrrev_b32_e32 v31, 8, v9
-; VI-NEXT: buffer_store_dword v31, off, s[0:3], s32 offset:256 ; 4-byte Folded Spill
-; VI-NEXT: v_lshrrev_b32_e32 v31, 24, v8
-; VI-NEXT: buffer_store_dword v31, off, s[0:3], s32 offset:296 ; 4-byte Folded Spill
-; VI-NEXT: v_lshrrev_b32_e32 v31, 8, v8
-; VI-NEXT: buffer_store_dword v31, off, s[0:3], s32 offset:260 ; 4-byte Folded Spill
-; VI-NEXT: v_lshrrev_b32_e32 v31, 8, v7
-; VI-NEXT: buffer_store_dword v31, off, s[0:3], s32 offset:272 ; 4-byte Folded Spill
-; VI-NEXT: v_lshrrev_b32_e32 v31, 24, v6
-; VI-NEXT: buffer_store_dword v31, off, s[0:3], s32 offset:300 ; 4-byte Folded Spill
-; VI-NEXT: v_lshrrev_b32_e32 v31, 8, v6
-; VI-NEXT: buffer_store_dword v31, off, s[0:3], s32 offset:276 ; 4-byte Folded Spill
-; VI-NEXT: v_lshrrev_b32_e32 v31, 8, v5
-; VI-NEXT: buffer_store_dword v31, off, s[0:3], s32 offset:284 ; 4-byte Folded Spill
-; VI-NEXT: v_lshrrev_b64 v[31:32], 24, v[15:16]
-; VI-NEXT: buffer_store_dword v31, off, s[0:3], s32 offset:76 ; 4-byte Folded Spill
-; VI-NEXT: buffer_store_dword v32, off, s[0:3], s32 offset:80 ; 4-byte Folded Spill
-; VI-NEXT: v_lshrrev_b64 v[31:32], 24, v[13:14]
-; VI-NEXT: buffer_store_dword v31, off, s[0:3], s32 offset:84 ; 4-byte Folded Spill
-; VI-NEXT: buffer_store_dword v32, off, s[0:3], s32 offset:88 ; 4-byte Folded Spill
-; VI-NEXT: v_lshrrev_b64 v[31:32], 24, v[11:12]
-; VI-NEXT: buffer_store_dword v31, off, s[0:3], s32 offset:92 ; 4-byte Folded Spill
-; VI-NEXT: buffer_store_dword v32, off, s[0:3], s32 offset:96 ; 4-byte Folded Spill
-; VI-NEXT: v_lshrrev_b64 v[31:32], 24, v[9:10]
-; VI-NEXT: buffer_store_dword v31, off, s[0:3], s32 offset:100 ; 4-byte Folded Spill
-; VI-NEXT: buffer_store_dword v32, off, s[0:3], s32 offset:104 ; 4-byte Folded Spill
-; VI-NEXT: v_mov_b32_e32 v31, v7
-; VI-NEXT: buffer_store_dword v31, off, s[0:3], s32 offset:264 ; 4-byte Folded Spill
-; VI-NEXT: buffer_store_dword v9, off, s[0:3], s32 offset:368 ; 4-byte Folded Spill
-; VI-NEXT: v_mov_b32_e32 v9, v10
-; VI-NEXT: buffer_store_dword v9, off, s[0:3], s32 offset:344 ; 4-byte Folded Spill
-; VI-NEXT: v_mov_b32_e32 v9, v11
-; VI-NEXT: buffer_store_dword v9, off, s[0:3], s32 offset:380 ; 4-byte Folded Spill
-; VI-NEXT: v_mov_b32_e32 v9, v12
-; VI-NEXT: buffer_store_dword v9, off, s[0:3], s32 offset:376 ; 4-byte Folded Spill
-; VI-NEXT: v_mov_b32_e32 v9, v13
-; VI-NEXT: buffer_store_dword v9, off, s[0:3], s32 offset:416 ; 4-byte Folded Spill
-; VI-NEXT: v_mov_b32_e32 v9, v14
-; VI-NEXT: buffer_store_dword v9, off, s[0:3], s32 offset:400 ; 4-byte Folded Spill
-; VI-NEXT: v_mov_b32_e32 v9, v16
-; VI-NEXT: buffer_store_dword v9, off, s[0:3], s32 offset:420 ; 4-byte Folded Spill
-; VI-NEXT: v_mov_b32_e32 v9, v8
-; VI-NEXT: v_lshrrev_b64 v[7:8], 24, v[7:8]
-; VI-NEXT: buffer_store_dword v9, off, s[0:3], s32 offset:396 ; 4-byte Folded Spill
-; VI-NEXT: buffer_store_dword v7, off, s[0:3], s32 offset:132 ; 4-byte Folded Spill
-; VI-NEXT: buffer_store_dword v8, off, s[0:3], s32 offset:136 ; 4-byte Folded Spill
-; VI-NEXT: v_mov_b32_e32 v7, v5
-; VI-NEXT: buffer_store_dword v7, off, s[0:3], s32 offset:412 ; 4-byte Folded Spill
-; VI-NEXT: v_mov_b32_e32 v7, v6
-; VI-NEXT: v_lshrrev_b64 v[5:6], 24, v[5:6]
-; VI-NEXT: buffer_store_dword v7, off, s[0:3], s32 offset:392 ; 4-byte Folded Spill
-; VI-NEXT: buffer_store_dword v5, off, s[0:3], s32 offset:148 ; 4-byte Folded Spill
-; VI-NEXT: buffer_store_dword v6, off, s[0:3], s32 offset:152 ; 4-byte Folded Spill
-; VI-NEXT: v_lshrrev_b32_e32 v5, 24, v4
-; VI-NEXT: buffer_store_dword v5, off, s[0:3], s32 offset:476 ; 4-byte Folded Spill
-; VI-NEXT: v_lshrrev_b32_e32 v5, 8, v4
-; VI-NEXT: buffer_store_dword v5, off, s[0:3], s32 offset:436 ; 4-byte Folded Spill
-; VI-NEXT: v_lshrrev_b32_e32 v5, 8, v3
-; VI-NEXT: buffer_store_dword v5, off, s[0:3], s32 offset:440 ; 4-byte Folded Spill
-; VI-NEXT: v_mov_b32_e32 v5, v3
-; VI-NEXT: v_lshrrev_b64 v[40:41], 24, v[3:4]
-; VI-NEXT: v_lshrrev_b32_e32 v3, 24, v2
-; VI-NEXT: buffer_store_dword v3, off, s[0:3], s32 offset:492 ; 4-byte Folded Spill
-; VI-NEXT: v_lshrrev_b32_e32 v3, 8, v2
-; VI-NEXT: buffer_store_dword v3, off, s[0:3], s32 offset:444 ; 4-byte Folded Spill
-; VI-NEXT: v_lshrrev_b32_e32 v3, 8, v1
-; VI-NEXT: buffer_store_dword v3, off, s[0:3], s32 offset:452 ; 4-byte Folded Spill
-; VI-NEXT: v_mov_b32_e32 v3, v1
-; VI-NEXT: v_lshrrev_b64 v[43:44], 24, v[1:2]
-; VI-NEXT: v_lshrrev_b32_e32 v1, 24, v37
-; VI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:432 ; 4-byte Folded Spill
-; VI-NEXT: v_lshrrev_b32_e32 v1, 8, v37
-; VI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:424 ; 4-byte Folded Spill
-; VI-NEXT: v_lshrrev_b32_e32 v1, 8, v36
-; VI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:428 ; 4-byte Folded Spill
-; VI-NEXT: v_mov_b32_e32 v1, v36
-; VI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:328 ; 4-byte Folded Spill
-; VI-NEXT: v_mov_b32_e32 v1, v37
-; VI-NEXT: buffer_store_dword v3, off, s[0:3], s32 offset:404 ; 4-byte Folded Spill
-; VI-NEXT: v_mov_b32_e32 v3, v2
-; VI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:320 ; 4-byte Folded Spill
-; VI-NEXT: v_lshrrev_b64 v[1:2], 24, v[36:37]
-; VI-NEXT: buffer_store_dword v5, off, s[0:3], s32 offset:408 ; 4-byte Folded Spill
-; VI-NEXT: v_mov_b32_e32 v5, v4
-; VI-NEXT: buffer_store_dword v5, off, s[0:3], s32 offset:388 ; 4-byte Folded Spill
-; VI-NEXT: buffer_store_dword v3, off, s[0:3], s32 offset:384 ; 4-byte Folded Spill
-; VI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:108 ; 4-byte Folded Spill
-; VI-NEXT: buffer_store_dword v2, off, s[0:3], s32 offset:112 ; 4-byte Folded Spill
-; VI-NEXT: v_lshrrev_b32_e32 v1, 24, v30
-; VI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:484 ; 4-byte Folded Spill
-; VI-NEXT: v_lshrrev_b32_e32 v1, 8, v30
-; VI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:448 ; 4-byte Folded Spill
-; VI-NEXT: v_lshrrev_b32_e32 v1, 8, v29
-; VI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:456 ; 4-byte Folded Spill
-; VI-NEXT: v_mov_b32_e32 v1, v29
-; VI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:356 ; 4-byte Folded Spill
-; VI-NEXT: v_mov_b32_e32 v1, v30
-; VI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:316 ; 4-byte Folded Spill
-; VI-NEXT: v_lshrrev_b64 v[1:2], 24, v[29:30]
-; VI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:116 ; 4-byte Folded Spill
-; VI-NEXT: buffer_store_dword v2, off, s[0:3], s32 offset:120 ; 4-byte Folded Spill
-; VI-NEXT: v_lshrrev_b32_e32 v1, 24, v28
-; VI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:496 ; 4-byte Folded Spill
-; VI-NEXT: v_lshrrev_b32_e32 v1, 8, v28
-; VI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:460 ; 4-byte Folded Spill
-; VI-NEXT: v_lshrrev_b32_e32 v1, 8, v27
-; VI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:464 ; 4-byte Folded Spill
-; VI-NEXT: v_mov_b32_e32 v1, v27
-; VI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:352 ; 4-byte Folded Spill
-; VI-NEXT: v_mov_b32_e32 v1, v28
-; VI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:308 ; 4-byte Folded Spill
-; VI-NEXT: v_lshrrev_b64 v[1:2], 24, v[27:28]
-; VI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:124 ; 4-byte Folded Spill
-; VI-NEXT: buffer_store_dword v2, off, s[0:3], s32 offset:128 ; 4-byte Folded Spill
-; VI-NEXT: v_lshrrev_b32_e32 v1, 8, v26
-; VI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:468 ; 4-byte Folded Spill
-; VI-NEXT: v_lshrrev_b32_e32 v1, 8, v25
-; VI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:472 ; 4-byte Folded Spill
-; VI-NEXT: v_mov_b32_e32 v1, v25
-; VI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:340 ; 4-byte Folded Spill
-; VI-NEXT: v_mov_b32_e32 v1, v26
-; VI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:304 ; 4-byte Folded Spill
-; VI-NEXT: v_lshrrev_b64 v[1:2], 24, v[25:26]
-; VI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:140 ; 4-byte Folded Spill
-; VI-NEXT: buffer_store_dword v2, off, s[0:3], s32 offset:144 ; 4-byte Folded Spill
-; VI-NEXT: v_lshrrev_b32_e32 v1, 8, v24
-; VI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:480 ; 4-byte Folded Spill
-; VI-NEXT: v_lshrrev_b32_e32 v1, 8, v23
-; VI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:488 ; 4-byte Folded Spill
-; VI-NEXT: v_mov_b32_e32 v1, v23
-; VI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:348 ; 4-byte Folded Spill
-; VI-NEXT: v_mov_b32_e32 v1, v24
-; VI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:312 ; 4-byte Folded Spill
-; VI-NEXT: v_lshrrev_b32_e32 v1, 8, v22
-; VI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:500 ; 4-byte Folded Spill
-; VI-NEXT: v_lshrrev_b32_e32 v1, 8, v21
-; VI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:504 ; 4-byte Folded Spill
-; VI-NEXT: v_mov_b32_e32 v1, v21
-; VI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:360 ; 4-byte Folded Spill
-; VI-NEXT: v_mov_b32_e32 v1, v22
-; VI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:324 ; 4-byte Folded Spill
-; VI-NEXT: v_mov_b32_e32 v1, v19
-; VI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:364 ; 4-byte Folded Spill
-; VI-NEXT: v_mov_b32_e32 v1, v20
-; VI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:332 ; 4-byte Folded Spill
-; VI-NEXT: v_mov_b32_e32 v1, v17
-; VI-NEXT: v_lshrrev_b64 v[44:45], 24, v[19:20]
-; VI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:372 ; 4-byte Folded Spill
-; VI-NEXT: v_mov_b32_e32 v1, v18
-; VI-NEXT: v_lshrrev_b64 v[54:55], 24, v[23:24]
-; VI-NEXT: v_lshrrev_b64 v[41:42], 24, v[21:22]
-; VI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:336 ; 4-byte Folded Spill
-; VI-NEXT: v_mov_b32_e32 v1, v46
-; VI-NEXT: v_lshrrev_b64 v[45:46], 24, v[17:18]
-; VI-NEXT: v_mov_b32_e32 v32, v15
-; VI-NEXT: v_lshrrev_b32_e32 v35, 24, v26
-; VI-NEXT: v_lshrrev_b32_e32 v39, 24, v24
-; VI-NEXT: v_lshrrev_b32_e32 v58, 24, v22
-; VI-NEXT: v_lshrrev_b32_e32 v61, 24, v20
-; VI-NEXT: v_lshrrev_b32_e32 v42, 8, v20
-; VI-NEXT: v_lshrrev_b32_e32 v38, 8, v19
-; VI-NEXT: v_lshrrev_b32_e32 v31, 24, v18
-; VI-NEXT: v_lshrrev_b32_e32 v62, 8, v18
-; VI-NEXT: v_lshrrev_b32_e32 v55, 8, v17
-; VI-NEXT: v_mov_b32_e32 v46, v1
-; VI-NEXT: ; implicit-def: $vgpr1
-; VI-NEXT: ; implicit-def: $vgpr3
-; VI-NEXT: ; implicit-def: $vgpr5
-; VI-NEXT: ; implicit-def: $vgpr7
-; VI-NEXT: ; implicit-def: $vgpr9
-; VI-NEXT: ; implicit-def: $vgpr11
-; VI-NEXT: ; implicit-def: $vgpr13
-; VI-NEXT: ; implicit-def: $vgpr15
-; VI-NEXT: ; implicit-def: $vgpr17
-; VI-NEXT: ; implicit-def: $vgpr19
-; VI-NEXT: ; implicit-def: $vgpr21
-; VI-NEXT: ; implicit-def: $vgpr23
-; VI-NEXT: ; implicit-def: $vgpr25
-; VI-NEXT: ; implicit-def: $vgpr27
-; VI-NEXT: ; implicit-def: $vgpr29
-; VI-NEXT: ; implicit-def: $vgpr37
+; VI-NEXT: v_lshrrev_b32_e32 v33, 24, v16
+; VI-NEXT: buffer_store_dword v33, off, s[0:3], s32 offset:224 ; 4-byte Folded Spill
+; VI-NEXT: v_lshrrev_b32_e32 v33, 8, v16
+; VI-NEXT: buffer_store_dword v33, off, s[0:3], s32 offset:192 ; 4-byte Folded Spill
+; VI-NEXT: v_lshrrev_b32_e32 v33, 8, v15
+; VI-NEXT: buffer_store_dword v33, off, s[0:3], s32 offset:196 ; 4-byte Folded Spill
+; VI-NEXT: v_lshrrev_b32_e32 v33, 24, v14
+; VI-NEXT: buffer_store_dword v33, off, s[0:3], s32 offset:252 ; 4-byte Folded Spill
+; VI-NEXT: v_lshrrev_b32_e32 v33, 8, v14
+; VI-NEXT: buffer_store_dword v33, off, s[0:3], s32 offset:200 ; 4-byte Folded Spill
+; VI-NEXT: v_lshrrev_b32_e32 v33, 8, v13
+; VI-NEXT: buffer_store_dword v33, off, s[0:3], s32 offset:204 ; 4-byte Folded Spill
+; VI-NEXT: v_lshrrev_b32_e32 v33, 24, v12
+; VI-NEXT: buffer_store_dword v33, off, s[0:3], s32 offset:272 ; 4-byte Folded Spill
+; VI-NEXT: v_lshrrev_b32_e32 v33, 8, v12
+; VI-NEXT: buffer_store_dword v33, off, s[0:3], s32 offset:212 ; 4-byte Folded Spill
+; VI-NEXT: v_lshrrev_b32_e32 v33, 8, v11
+; VI-NEXT: buffer_store_dword v33, off, s[0:3], s32 offset:220 ; 4-byte Folded Spill
+; VI-NEXT: v_lshrrev_b32_e32 v33, 24, v10
+; VI-NEXT: buffer_store_dword v33, off, s[0:3], s32 offset:296 ; 4-byte Folded Spill
+; VI-NEXT: v_lshrrev_b32_e32 v33, 8, v10
+; VI-NEXT: buffer_store_dword v33, off, s[0:3], s32 offset:240 ; 4-byte Folded Spill
+; VI-NEXT: v_lshrrev_b32_e32 v33, 8, v9
+; VI-NEXT: buffer_store_dword v33, off, s[0:3], s32 offset:248 ; 4-byte Folded Spill
+; VI-NEXT: v_lshrrev_b32_e32 v33, 24, v8
+; VI-NEXT: buffer_store_dword v33, off, s[0:3], s32 offset:316 ; 4-byte Folded Spill
+; VI-NEXT: v_lshrrev_b32_e32 v33, 8, v8
+; VI-NEXT: buffer_store_dword v33, off, s[0:3], s32 offset:264 ; 4-byte Folded Spill
+; VI-NEXT: v_lshrrev_b32_e32 v33, 8, v7
+; VI-NEXT: buffer_store_dword v33, off, s[0:3], s32 offset:276 ; 4-byte Folded Spill
+; VI-NEXT: v_lshrrev_b32_e32 v33, 24, v6
+; VI-NEXT: buffer_store_dword v33, off, s[0:3], s32 offset:340 ; 4-byte Folded Spill
+; VI-NEXT: v_lshrrev_b32_e32 v33, 8, v6
+; VI-NEXT: buffer_store_dword v33, off, s[0:3], s32 offset:292 ; 4-byte Folded Spill
+; VI-NEXT: v_lshrrev_b32_e32 v33, 8, v5
+; VI-NEXT: buffer_store_dword v33, off, s[0:3], s32 offset:300 ; 4-byte Folded Spill
+; VI-NEXT: v_lshrrev_b32_e32 v33, 24, v4
+; VI-NEXT: buffer_store_dword v33, off, s[0:3], s32 offset:364 ; 4-byte Folded Spill
+; VI-NEXT: v_lshrrev_b32_e32 v33, 8, v4
+; VI-NEXT: buffer_store_dword v33, off, s[0:3], s32 offset:320 ; 4-byte Folded Spill
+; VI-NEXT: v_lshrrev_b32_e32 v33, 8, v3
+; VI-NEXT: buffer_store_dword v33, off, s[0:3], s32 offset:328 ; 4-byte Folded Spill
+; VI-NEXT: v_lshrrev_b32_e32 v33, 8, v2
+; VI-NEXT: buffer_store_dword v33, off, s[0:3], s32 offset:348 ; 4-byte Folded Spill
+; VI-NEXT: v_lshrrev_b32_e32 v33, 8, v1
+; VI-NEXT: buffer_store_dword v33, off, s[0:3], s32 offset:356 ; 4-byte Folded Spill
+; VI-NEXT: v_lshrrev_b32_e32 v33, 24, v32
+; VI-NEXT: buffer_store_dword v33, off, s[0:3], s32 offset:244 ; 4-byte Folded Spill
+; VI-NEXT: v_lshrrev_b32_e32 v33, 8, v32
+; VI-NEXT: buffer_store_dword v33, off, s[0:3], s32 offset:208 ; 4-byte Folded Spill
+; VI-NEXT: v_lshrrev_b32_e32 v33, 8, v31
+; VI-NEXT: buffer_store_dword v33, off, s[0:3], s32 offset:216 ; 4-byte Folded Spill
+; VI-NEXT: v_lshrrev_b32_e32 v33, 24, v30
+; VI-NEXT: buffer_store_dword v33, off, s[0:3], s32 offset:268 ; 4-byte Folded Spill
+; VI-NEXT: v_lshrrev_b32_e32 v33, 8, v30
+; VI-NEXT: buffer_store_dword v33, off, s[0:3], s32 offset:232 ; 4-byte Folded Spill
+; VI-NEXT: v_lshrrev_b32_e32 v33, 8, v29
+; VI-NEXT: buffer_store_dword v33, off, s[0:3], s32 offset:236 ; 4-byte Folded Spill
+; VI-NEXT: v_lshrrev_b32_e32 v33, 24, v28
+; VI-NEXT: buffer_store_dword v33, off, s[0:3], s32 offset:288 ; 4-byte Folded Spill
+; VI-NEXT: v_lshrrev_b32_e32 v33, 8, v28
+; VI-NEXT: buffer_store_dword v33, off, s[0:3], s32 offset:256 ; 4-byte Folded Spill
+; VI-NEXT: v_lshrrev_b32_e32 v33, 8, v27
+; VI-NEXT: buffer_store_dword v33, off, s[0:3], s32 offset:260 ; 4-byte Folded Spill
+; VI-NEXT: v_lshrrev_b32_e32 v33, 24, v26
+; VI-NEXT: buffer_store_dword v33, off, s[0:3], s32 offset:308 ; 4-byte Folded Spill
+; VI-NEXT: v_lshrrev_b32_e32 v33, 8, v26
+; VI-NEXT: buffer_store_dword v33, off, s[0:3], s32 offset:280 ; 4-byte Folded Spill
+; VI-NEXT: v_lshrrev_b32_e32 v33, 8, v25
+; VI-NEXT: buffer_store_dword v33, off, s[0:3], s32 offset:284 ; 4-byte Folded Spill
+; VI-NEXT: v_lshrrev_b32_e32 v33, 24, v24
+; VI-NEXT: buffer_store_dword v33, off, s[0:3], s32 offset:332 ; 4-byte Folded Spill
+; VI-NEXT: v_lshrrev_b32_e32 v33, 8, v24
+; VI-NEXT: buffer_store_dword v33, off, s[0:3], s32 offset:304 ; 4-byte Folded Spill
+; VI-NEXT: v_lshrrev_b32_e32 v33, 8, v23
+; VI-NEXT: buffer_store_dword v33, off, s[0:3], s32 offset:312 ; 4-byte Folded Spill
+; VI-NEXT: v_lshrrev_b32_e32 v33, 24, v22
+; VI-NEXT: buffer_store_dword v33, off, s[0:3], s32 offset:344 ; 4-byte Folded Spill
+; VI-NEXT: v_lshrrev_b32_e32 v33, 8, v22
+; VI-NEXT: buffer_store_dword v33, off, s[0:3], s32 offset:324 ; 4-byte Folded Spill
+; VI-NEXT: v_lshrrev_b32_e32 v33, 8, v21
+; VI-NEXT: buffer_store_dword v33, off, s[0:3], s32 offset:336 ; 4-byte Folded Spill
+; VI-NEXT: v_lshrrev_b32_e32 v33, 8, v20
+; VI-NEXT: buffer_store_dword v33, off, s[0:3], s32 offset:352 ; 4-byte Folded Spill
+; VI-NEXT: v_lshrrev_b32_e32 v33, 8, v19
+; VI-NEXT: buffer_store_dword v33, off, s[0:3], s32 offset:360 ; 4-byte Folded Spill
+; VI-NEXT: v_lshrrev_b64 v[33:34], 24, v[15:16]
+; VI-NEXT: buffer_store_dword v33, off, s[0:3], s32 offset:104 ; 4-byte Folded Spill
+; VI-NEXT: buffer_store_dword v34, off, s[0:3], s32 offset:108 ; 4-byte Folded Spill
+; VI-NEXT: v_lshrrev_b64 v[33:34], 24, v[13:14]
+; VI-NEXT: buffer_store_dword v33, off, s[0:3], s32 offset:112 ; 4-byte Folded Spill
+; VI-NEXT: buffer_store_dword v34, off, s[0:3], s32 offset:116 ; 4-byte Folded Spill
+; VI-NEXT: v_lshrrev_b64 v[33:34], 24, v[11:12]
+; VI-NEXT: buffer_store_dword v33, off, s[0:3], s32 offset:120 ; 4-byte Folded Spill
+; VI-NEXT: buffer_store_dword v34, off, s[0:3], s32 offset:124 ; 4-byte Folded Spill
+; VI-NEXT: v_lshrrev_b64 v[44:45], 24, v[5:6]
+; VI-NEXT: v_lshrrev_b64 v[33:34], 24, v[9:10]
+; VI-NEXT: v_lshrrev_b64 v[45:46], 24, v[3:4]
+; VI-NEXT: v_mov_b32_e32 v34, v47
+; VI-NEXT: v_lshrrev_b64 v[46:47], 24, v[1:2]
+; VI-NEXT: v_mov_b32_e32 v47, v34
+; VI-NEXT: v_lshrrev_b64 v[34:35], 24, v[31:32]
+; VI-NEXT: buffer_store_dword v34, off, s[0:3], s32 offset:88 ; 4-byte Folded Spill
+; VI-NEXT: buffer_store_dword v35, off, s[0:3], s32 offset:92 ; 4-byte Folded Spill
+; VI-NEXT: v_lshrrev_b64 v[34:35], 24, v[29:30]
+; VI-NEXT: buffer_store_dword v34, off, s[0:3], s32 offset:96 ; 4-byte Folded Spill
+; VI-NEXT: buffer_store_dword v35, off, s[0:3], s32 offset:100 ; 4-byte Folded Spill
+; VI-NEXT: v_mov_b32_e32 v34, v36
+; VI-NEXT: v_lshrrev_b64 v[35:36], 24, v[27:28]
+; VI-NEXT: v_lshrrev_b64 v[38:39], 24, v[25:26]
+; VI-NEXT: v_mov_b32_e32 v36, v34
+; VI-NEXT: v_mov_b32_e32 v34, v50
+; VI-NEXT: v_lshrrev_b64 v[49:50], 24, v[23:24]
+; VI-NEXT: v_mov_b32_e32 v39, v62
+; VI-NEXT: v_lshrrev_b64 v[61:62], 24, v[21:22]
+; VI-NEXT: v_lshrrev_b64 v[41:42], 24, v[7:8]
+; VI-NEXT: v_mov_b32_e32 v50, v34
+; VI-NEXT: v_mov_b32_e32 v62, v39
+; VI-NEXT: v_mov_b32_e32 v34, v40
+; VI-NEXT: v_lshrrev_b64 v[39:40], 24, v[19:20]
+; VI-NEXT: v_mov_b32_e32 v40, v34
+; VI-NEXT: v_mov_b32_e32 v34, v43
+; VI-NEXT: v_lshrrev_b64 v[42:43], 24, v[17:18]
+; VI-NEXT: v_lshrrev_b32_e32 v60, 24, v2
+; VI-NEXT: v_lshrrev_b32_e32 v48, 24, v20
+; VI-NEXT: v_lshrrev_b32_e32 v55, 24, v18
+; VI-NEXT: v_lshrrev_b32_e32 v37, 8, v18
+; VI-NEXT: v_lshrrev_b32_e32 v58, 8, v17
+; VI-NEXT: v_mov_b32_e32 v43, v34
; VI-NEXT: .LBB98_2: ; %Flow
; VI-NEXT: s_andn2_saveexec_b64 s[4:5], s[4:5]
; VI-NEXT: s_cbranch_execz .LBB98_4
; VI-NEXT: ; %bb.3: ; %cmp.true
-; VI-NEXT: v_mov_b32_e32 v31, 3
-; VI-NEXT: v_add_u16_sdwa v51, v18, v31 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_1 src1_sel:DWORD
-; VI-NEXT: v_add_u16_e32 v32, 3, v18
-; VI-NEXT: v_lshlrev_b32_e32 v18, 16, v51
-; VI-NEXT: buffer_store_dword v32, off, s[0:3], s32 offset:336 ; 4-byte Folded Spill
-; VI-NEXT: v_or_b32_e32 v18, v32, v18
-; VI-NEXT: v_add_u16_e32 v32, 3, v17
-; VI-NEXT: v_add_u16_sdwa v17, v17, v31 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_1 src1_sel:DWORD
-; VI-NEXT: buffer_store_dword v17, off, s[0:3], s32 offset:188 ; 4-byte Folded Spill
+; VI-NEXT: v_mov_b32_e32 v33, 3
+; VI-NEXT: v_add_u16_e32 v34, 3, v18
+; VI-NEXT: v_add_u16_sdwa v18, v18, v33 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_1 src1_sel:DWORD
+; VI-NEXT: buffer_store_dword v18, off, s[0:3], s32 offset:368 ; 4-byte Folded Spill
+; VI-NEXT: v_lshlrev_b32_e32 v18, 16, v18
+; VI-NEXT: v_or_b32_e32 v35, v34, v18
+; VI-NEXT: v_add_u16_e32 v18, 3, v17
+; VI-NEXT: v_add_u16_sdwa v17, v17, v33 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_1 src1_sel:DWORD
+; VI-NEXT: buffer_store_dword v17, off, s[0:3], s32 offset:148 ; 4-byte Folded Spill
; VI-NEXT: v_lshlrev_b32_e32 v17, 16, v17
-; VI-NEXT: buffer_store_dword v32, off, s[0:3], s32 offset:372 ; 4-byte Folded Spill
-; VI-NEXT: v_or_b32_e32 v17, v32, v17
-; VI-NEXT: v_add_u16_e32 v32, 3, v20
-; VI-NEXT: v_add_u16_sdwa v20, v20, v31 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_1 src1_sel:DWORD
-; VI-NEXT: buffer_store_dword v20, off, s[0:3], s32 offset:512 ; 4-byte Folded Spill
-; VI-NEXT: v_lshlrev_b32_e32 v20, 16, v20
-; VI-NEXT: buffer_store_dword v32, off, s[0:3], s32 offset:332 ; 4-byte Folded Spill
-; VI-NEXT: v_or_b32_e32 v20, v32, v20
-; VI-NEXT: v_add_u16_e32 v32, 3, v19
-; VI-NEXT: v_add_u16_sdwa v19, v19, v31 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_1 src1_sel:DWORD
-; VI-NEXT: buffer_store_dword v19, off, s[0:3], s32 offset:508 ; 4-byte Folded Spill
-; VI-NEXT: v_lshlrev_b32_e32 v19, 16, v19
-; VI-NEXT: v_add_u16_sdwa v48, v22, v31 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_1 src1_sel:DWORD
-; VI-NEXT: buffer_store_dword v32, off, s[0:3], s32 offset:364 ; 4-byte Folded Spill
-; VI-NEXT: v_or_b32_e32 v19, v32, v19
-; VI-NEXT: v_add_u16_e32 v32, 3, v22
-; VI-NEXT: v_lshlrev_b32_e32 v22, 16, v48
-; VI-NEXT: v_add_u16_sdwa v53, v21, v31 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_1 src1_sel:DWORD
-; VI-NEXT: buffer_store_dword v32, off, s[0:3], s32 offset:324 ; 4-byte Folded Spill
-; VI-NEXT: v_or_b32_e32 v22, v32, v22
-; VI-NEXT: v_add_u16_e32 v32, 3, v21
-; VI-NEXT: v_lshlrev_b32_e32 v21, 16, v53
-; VI-NEXT: v_add_u16_sdwa v61, v24, v31 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_1 src1_sel:DWORD
-; VI-NEXT: buffer_store_dword v32, off, s[0:3], s32 offset:360 ; 4-byte Folded Spill
-; VI-NEXT: v_or_b32_e32 v21, v32, v21
-; VI-NEXT: v_add_u16_e32 v32, 3, v24
-; VI-NEXT: v_lshlrev_b32_e32 v24, 16, v61
-; VI-NEXT: buffer_store_dword v32, off, s[0:3], s32 offset:312 ; 4-byte Folded Spill
-; VI-NEXT: v_or_b32_e32 v24, v32, v24
-; VI-NEXT: v_add_u16_e32 v32, 3, v23
-; VI-NEXT: v_add_u16_sdwa v23, v23, v31 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_1 src1_sel:DWORD
-; VI-NEXT: buffer_store_dword v23, off, s[0:3], s32 offset:184 ; 4-byte Folded Spill
-; VI-NEXT: v_lshlrev_b32_e32 v23, 16, v23
-; VI-NEXT: v_add_u16_sdwa v58, v26, v31 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_1 src1_sel:DWORD
-; VI-NEXT: buffer_store_dword v32, off, s[0:3], s32 offset:348 ; 4-byte Folded Spill
-; VI-NEXT: v_or_b32_e32 v23, v32, v23
-; VI-NEXT: v_add_u16_e32 v32, 3, v26
-; VI-NEXT: v_lshlrev_b32_e32 v26, 16, v58
-; VI-NEXT: buffer_store_dword v32, off, s[0:3], s32 offset:304 ; 4-byte Folded Spill
-; VI-NEXT: v_or_b32_e32 v26, v32, v26
-; VI-NEXT: v_add_u16_e32 v32, 3, v25
-; VI-NEXT: v_add_u16_sdwa v25, v25, v31 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_1 src1_sel:DWORD
-; VI-NEXT: buffer_store_dword v25, off, s[0:3], s32 offset:180 ; 4-byte Folded Spill
-; VI-NEXT: v_lshlrev_b32_e32 v25, 16, v25
-; VI-NEXT: v_add_u16_sdwa v39, v28, v31 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_1 src1_sel:DWORD
-; VI-NEXT: buffer_store_dword v32, off, s[0:3], s32 offset:340 ; 4-byte Folded Spill
-; VI-NEXT: v_or_b32_e32 v25, v32, v25
-; VI-NEXT: v_add_u16_e32 v32, 3, v28
-; VI-NEXT: v_lshlrev_b32_e32 v28, 16, v39
-; VI-NEXT: buffer_store_dword v32, off, s[0:3], s32 offset:308 ; 4-byte Folded Spill
-; VI-NEXT: v_or_b32_e32 v28, v32, v28
-; VI-NEXT: v_add_u16_e32 v32, 3, v27
-; VI-NEXT: v_add_u16_sdwa v27, v27, v31 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_1 src1_sel:DWORD
-; VI-NEXT: buffer_store_dword v27, off, s[0:3], s32 offset:172 ; 4-byte Folded Spill
-; VI-NEXT: v_lshlrev_b32_e32 v27, 16, v27
-; VI-NEXT: v_add_u16_sdwa v35, v30, v31 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_1 src1_sel:DWORD
-; VI-NEXT: buffer_store_dword v32, off, s[0:3], s32 offset:352 ; 4-byte Folded Spill
-; VI-NEXT: v_or_b32_e32 v27, v32, v27
-; VI-NEXT: v_add_u16_e32 v33, 3, v30
-; VI-NEXT: v_add_u16_e32 v34, 3, v29
-; VI-NEXT: v_add_u16_sdwa v32, v29, v31 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_1 src1_sel:DWORD
-; VI-NEXT: v_lshlrev_b32_e32 v29, 16, v35
-; VI-NEXT: v_add_u16_sdwa v52, v37, v31 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_1 src1_sel:DWORD
-; VI-NEXT: buffer_store_dword v33, off, s[0:3], s32 offset:316 ; 4-byte Folded Spill
-; VI-NEXT: v_or_b32_e32 v30, v33, v29
-; VI-NEXT: buffer_store_dword v32, off, s[0:3], s32 offset:168 ; 4-byte Folded Spill
-; VI-NEXT: v_lshlrev_b32_e32 v29, 16, v32
-; VI-NEXT: v_add_u16_e32 v33, 3, v37
-; VI-NEXT: v_add_u16_sdwa v50, v36, v31 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_1 src1_sel:DWORD
-; VI-NEXT: v_lshlrev_b32_e32 v32, 16, v52
-; VI-NEXT: buffer_store_dword v34, off, s[0:3], s32 offset:356 ; 4-byte Folded Spill
-; VI-NEXT: v_or_b32_e32 v29, v34, v29
-; VI-NEXT: v_add_u16_e32 v34, 3, v36
-; VI-NEXT: v_or_b32_e32 v37, v33, v32
-; VI-NEXT: v_lshlrev_b32_e32 v32, 16, v50
-; VI-NEXT: v_add_u16_sdwa v57, v2, v31 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_1 src1_sel:DWORD
-; VI-NEXT: buffer_store_dword v33, off, s[0:3], s32 offset:320 ; 4-byte Folded Spill
-; VI-NEXT: buffer_store_dword v34, off, s[0:3], s32 offset:328 ; 4-byte Folded Spill
-; VI-NEXT: v_or_b32_e32 v36, v34, v32
-; VI-NEXT: v_add_u16_e32 v33, 3, v2
-; VI-NEXT: v_add_u16_e32 v34, 3, v1
-; VI-NEXT: v_add_u16_sdwa v32, v1, v31 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_1 src1_sel:DWORD
-; VI-NEXT: v_lshlrev_b32_e32 v1, 16, v57
-; VI-NEXT: v_or_b32_e32 v2, v33, v1
-; VI-NEXT: v_lshlrev_b32_e32 v1, 16, v32
-; VI-NEXT: v_add_u16_sdwa v56, v4, v31 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_1 src1_sel:DWORD
-; VI-NEXT: buffer_store_dword v33, off, s[0:3], s32 offset:384 ; 4-byte Folded Spill
-; VI-NEXT: buffer_store_dword v32, off, s[0:3], s32 offset:220 ; 4-byte Folded Spill
-; VI-NEXT: buffer_store_dword v34, off, s[0:3], s32 offset:404 ; 4-byte Folded Spill
-; VI-NEXT: v_or_b32_e32 v1, v34, v1
-; VI-NEXT: v_add_u16_e32 v33, 3, v4
-; VI-NEXT: v_add_u16_e32 v34, 3, v3
-; VI-NEXT: v_add_u16_sdwa v32, v3, v31 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_1 src1_sel:DWORD
-; VI-NEXT: v_lshlrev_b32_e32 v3, 16, v56
-; VI-NEXT: v_or_b32_e32 v4, v33, v3
-; VI-NEXT: v_lshlrev_b32_e32 v3, 16, v32
-; VI-NEXT: v_add_u16_sdwa v47, v6, v31 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_1 src1_sel:DWORD
-; VI-NEXT: buffer_store_dword v33, off, s[0:3], s32 offset:388 ; 4-byte Folded Spill
-; VI-NEXT: buffer_store_dword v32, off, s[0:3], s32 offset:216 ; 4-byte Folded Spill
-; VI-NEXT: buffer_store_dword v34, off, s[0:3], s32 offset:408 ; 4-byte Folded Spill
-; VI-NEXT: v_or_b32_e32 v3, v34, v3
-; VI-NEXT: v_add_u16_e32 v33, 3, v6
-; VI-NEXT: v_add_u16_e32 v34, 3, v5
-; VI-NEXT: v_add_u16_sdwa v32, v5, v31 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_1 src1_sel:DWORD
-; VI-NEXT: v_lshlrev_b32_e32 v5, 16, v47
-; VI-NEXT: v_or_b32_e32 v6, v33, v5
-; VI-NEXT: v_lshlrev_b32_e32 v5, 16, v32
-; VI-NEXT: buffer_store_dword v34, off, s[0:3], s32 offset:412 ; 4-byte Folded Spill
-; VI-NEXT: v_or_b32_e32 v5, v34, v5
-; VI-NEXT: v_add_u16_sdwa v34, v8, v31 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_1 src1_sel:DWORD
-; VI-NEXT: buffer_store_dword v33, off, s[0:3], s32 offset:392 ; 4-byte Folded Spill
-; VI-NEXT: buffer_store_dword v32, off, s[0:3], s32 offset:212 ; 4-byte Folded Spill
-; VI-NEXT: v_add_u16_e32 v38, 3, v8
-; VI-NEXT: v_add_u16_e32 v33, 3, v7
-; VI-NEXT: v_add_u16_sdwa v32, v7, v31 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_1 src1_sel:DWORD
-; VI-NEXT: v_lshlrev_b32_e32 v7, 16, v34
-; VI-NEXT: v_or_b32_e32 v8, v38, v7
-; VI-NEXT: v_lshlrev_b32_e32 v7, 16, v32
-; VI-NEXT: v_add_u16_sdwa v59, v10, v31 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_1 src1_sel:DWORD
-; VI-NEXT: buffer_store_dword v38, off, s[0:3], s32 offset:396 ; 4-byte Folded Spill
-; VI-NEXT: buffer_store_dword v32, off, s[0:3], s32 offset:208 ; 4-byte Folded Spill
-; VI-NEXT: buffer_store_dword v33, off, s[0:3], s32 offset:264 ; 4-byte Folded Spill
-; VI-NEXT: v_or_b32_e32 v7, v33, v7
-; VI-NEXT: v_add_u16_e32 v33, 3, v10
-; VI-NEXT: v_add_u16_e32 v38, 3, v9
-; VI-NEXT: v_add_u16_sdwa v32, v9, v31 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_1 src1_sel:DWORD
-; VI-NEXT: v_lshlrev_b32_e32 v9, 16, v59
-; VI-NEXT: v_or_b32_e32 v10, v33, v9
-; VI-NEXT: v_lshlrev_b32_e32 v9, 16, v32
-; VI-NEXT: v_add_u16_sdwa v63, v12, v31 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_1 src1_sel:DWORD
-; VI-NEXT: buffer_store_dword v33, off, s[0:3], s32 offset:344 ; 4-byte Folded Spill
-; VI-NEXT: buffer_store_dword v32, off, s[0:3], s32 offset:204 ; 4-byte Folded Spill
-; VI-NEXT: buffer_store_dword v38, off, s[0:3], s32 offset:368 ; 4-byte Folded Spill
-; VI-NEXT: v_or_b32_e32 v9, v38, v9
-; VI-NEXT: v_add_u16_e32 v33, 3, v12
-; VI-NEXT: v_add_u16_e32 v38, 3, v11
-; VI-NEXT: v_add_u16_sdwa v32, v11, v31 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_1 src1_sel:DWORD
-; VI-NEXT: v_lshlrev_b32_e32 v11, 16, v63
-; VI-NEXT: buffer_store_dword v33, off, s[0:3], s32 offset:376 ; 4-byte Folded Spill
-; VI-NEXT: v_or_b32_e32 v12, v33, v11
-; VI-NEXT: v_lshlrev_b32_e32 v11, 16, v32
-; VI-NEXT: v_add_u16_sdwa v33, v14, v31 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_1 src1_sel:DWORD
-; VI-NEXT: buffer_store_dword v32, off, s[0:3], s32 offset:200 ; 4-byte Folded Spill
-; VI-NEXT: buffer_store_dword v38, off, s[0:3], s32 offset:380 ; 4-byte Folded Spill
-; VI-NEXT: v_or_b32_e32 v11, v38, v11
-; VI-NEXT: v_add_u16_e32 v38, 3, v14
-; VI-NEXT: v_add_u16_e32 v49, 3, v13
-; VI-NEXT: v_add_u16_sdwa v32, v13, v31 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_1 src1_sel:DWORD
-; VI-NEXT: v_lshlrev_b32_e32 v13, 16, v33
-; VI-NEXT: v_add_u16_sdwa v60, v16, v31 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_1 src1_sel:DWORD
-; VI-NEXT: v_or_b32_e32 v14, v38, v13
-; VI-NEXT: buffer_store_dword v32, off, s[0:3], s32 offset:196 ; 4-byte Folded Spill
-; VI-NEXT: v_lshlrev_b32_e32 v13, 16, v32
-; VI-NEXT: v_add_u16_sdwa v31, v15, v31 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_1 src1_sel:DWORD
+; VI-NEXT: buffer_store_dword v34, off, s[0:3], s32 offset:88 ; 4-byte Folded Spill
+; VI-NEXT: v_or_b32_e32 v34, v18, v17
+; VI-NEXT: buffer_store_dword v34, off, s[0:3], s32 offset:152 ; 4-byte Folded Spill
+; VI-NEXT: buffer_store_dword v35, off, s[0:3], s32 offset:156 ; 4-byte Folded Spill
+; VI-NEXT: v_add_u16_sdwa v17, v20, v33 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_1 src1_sel:DWORD
+; VI-NEXT: v_add_u16_e32 v34, 3, v20
+; VI-NEXT: buffer_store_dword v17, off, s[0:3], s32 offset:84 ; 4-byte Folded Spill
+; VI-NEXT: v_lshlrev_b32_e32 v17, 16, v17
+; VI-NEXT: v_or_b32_e32 v35, v34, v17
+; VI-NEXT: v_add_u16_sdwa v17, v19, v33 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_1 src1_sel:DWORD
+; VI-NEXT: v_add_u16_e32 v20, 3, v19
+; VI-NEXT: buffer_store_dword v17, off, s[0:3], s32 offset:144 ; 4-byte Folded Spill
+; VI-NEXT: v_lshlrev_b32_e32 v17, 16, v17
+; VI-NEXT: buffer_store_dword v34, off, s[0:3], s32 offset:96 ; 4-byte Folded Spill
+; VI-NEXT: v_or_b32_e32 v34, v20, v17
+; VI-NEXT: v_add_u16_sdwa v17, v22, v33 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_1 src1_sel:DWORD
+; VI-NEXT: buffer_store_dword v34, off, s[0:3], s32 offset:224 ; 4-byte Folded Spill
+; VI-NEXT: buffer_store_dword v35, off, s[0:3], s32 offset:228 ; 4-byte Folded Spill
+; VI-NEXT: v_add_u16_e32 v19, 3, v22
+; VI-NEXT: buffer_store_dword v17, off, s[0:3], s32 offset:76 ; 4-byte Folded Spill
+; VI-NEXT: v_lshlrev_b32_e32 v17, 16, v17
+; VI-NEXT: v_or_b32_e32 v61, v19, v17
+; VI-NEXT: v_add_u16_sdwa v17, v21, v33 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_1 src1_sel:DWORD
+; VI-NEXT: v_add_u16_e32 v22, 3, v21
+; VI-NEXT: buffer_store_dword v17, off, s[0:3], s32 offset:136 ; 4-byte Folded Spill
+; VI-NEXT: v_lshlrev_b32_e32 v17, 16, v17
+; VI-NEXT: v_or_b32_e32 v60, v22, v17
+; VI-NEXT: v_add_u16_sdwa v17, v24, v33 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_1 src1_sel:DWORD
+; VI-NEXT: buffer_store_dword v19, off, s[0:3], s32 offset:104 ; 4-byte Folded Spill
+; VI-NEXT: v_add_u16_e32 v19, 3, v24
+; VI-NEXT: buffer_store_dword v17, off, s[0:3], s32 offset:372 ; 4-byte Folded Spill
+; VI-NEXT: v_lshlrev_b32_e32 v17, 16, v17
+; VI-NEXT: v_or_b32_e32 v63, v19, v17
+; VI-NEXT: v_add_u16_sdwa v17, v23, v33 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_1 src1_sel:DWORD
+; VI-NEXT: v_add_u16_e32 v24, 3, v23
+; VI-NEXT: buffer_store_dword v17, off, s[0:3], s32 offset:132 ; 4-byte Folded Spill
+; VI-NEXT: v_lshlrev_b32_e32 v17, 16, v17
+; VI-NEXT: v_or_b32_e32 v62, v24, v17
+; VI-NEXT: v_add_u16_sdwa v17, v26, v33 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_1 src1_sel:DWORD
+; VI-NEXT: buffer_store_dword v19, off, s[0:3], s32 offset:112 ; 4-byte Folded Spill
+; VI-NEXT: v_add_u16_e32 v19, 3, v26
+; VI-NEXT: buffer_store_dword v17, off, s[0:3], s32 offset:376 ; 4-byte Folded Spill
+; VI-NEXT: v_lshlrev_b32_e32 v17, 16, v17
+; VI-NEXT: v_or_b32_e32 v58, v19, v17
+; VI-NEXT: v_add_u16_sdwa v17, v25, v33 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_1 src1_sel:DWORD
+; VI-NEXT: v_add_u16_e32 v26, 3, v25
+; VI-NEXT: buffer_store_dword v17, off, s[0:3], s32 offset:128 ; 4-byte Folded Spill
+; VI-NEXT: v_lshlrev_b32_e32 v17, 16, v17
+; VI-NEXT: v_or_b32_e32 v57, v26, v17
+; VI-NEXT: v_add_u16_sdwa v17, v28, v33 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_1 src1_sel:DWORD
+; VI-NEXT: buffer_store_dword v19, off, s[0:3], s32 offset:120 ; 4-byte Folded Spill
+; VI-NEXT: v_add_u16_e32 v19, 3, v28
+; VI-NEXT: buffer_store_dword v17, off, s[0:3], s32 offset:380 ; 4-byte Folded Spill
+; VI-NEXT: v_lshlrev_b32_e32 v17, 16, v17
+; VI-NEXT: v_or_b32_e32 v56, v19, v17
+; VI-NEXT: v_add_u16_sdwa v17, v27, v33 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_1 src1_sel:DWORD
+; VI-NEXT: v_add_u16_e32 v28, 3, v27
+; VI-NEXT: buffer_store_dword v17, off, s[0:3], s32 offset:388 ; 4-byte Folded Spill
+; VI-NEXT: v_lshlrev_b32_e32 v17, 16, v17
+; VI-NEXT: v_or_b32_e32 v55, v28, v17
+; VI-NEXT: v_add_u16_sdwa v17, v30, v33 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_1 src1_sel:DWORD
+; VI-NEXT: buffer_store_dword v19, off, s[0:3], s32 offset:192 ; 4-byte Folded Spill
+; VI-NEXT: v_add_u16_e32 v21, 3, v30
+; VI-NEXT: buffer_store_dword v17, off, s[0:3], s32 offset:392 ; 4-byte Folded Spill
+; VI-NEXT: v_add_u16_sdwa v19, v29, v33 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_1 src1_sel:DWORD
+; VI-NEXT: v_lshlrev_b32_e32 v17, 16, v17
+; VI-NEXT: v_add_u16_e32 v30, 3, v29
+; VI-NEXT: v_or_b32_e32 v40, v21, v17
+; VI-NEXT: v_lshlrev_b32_e32 v17, 16, v19
+; VI-NEXT: v_or_b32_e32 v39, v30, v17
+; VI-NEXT: v_add_u16_sdwa v17, v32, v33 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_1 src1_sel:DWORD
+; VI-NEXT: buffer_store_dword v19, off, s[0:3], s32 offset:384 ; 4-byte Folded Spill
+; VI-NEXT: v_add_u16_e32 v45, 3, v32
+; VI-NEXT: buffer_store_dword v17, off, s[0:3], s32 offset:80 ; 4-byte Folded Spill
+; VI-NEXT: v_add_u16_sdwa v19, v31, v33 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_1 src1_sel:DWORD
+; VI-NEXT: v_lshlrev_b32_e32 v17, 16, v17
+; VI-NEXT: buffer_store_dword v21, off, s[0:3], s32 offset:196 ; 4-byte Folded Spill
+; VI-NEXT: v_add_u16_e32 v32, 3, v31
+; VI-NEXT: v_or_b32_e32 v38, v45, v17
+; VI-NEXT: v_lshlrev_b32_e32 v17, 16, v19
+; VI-NEXT: v_add_u16_sdwa v21, v2, v33 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_1 src1_sel:DWORD
+; VI-NEXT: buffer_store_dword v19, off, s[0:3], s32 offset:140 ; 4-byte Folded Spill
+; VI-NEXT: v_or_b32_e32 v37, v32, v17
+; VI-NEXT: v_add_u16_e32 v17, 3, v2
+; VI-NEXT: v_add_u16_e32 v2, 3, v1
+; VI-NEXT: v_add_u16_sdwa v19, v1, v33 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_1 src1_sel:DWORD
+; VI-NEXT: v_lshlrev_b32_e32 v1, 16, v21
+; VI-NEXT: v_or_b32_e32 v49, v17, v1
+; VI-NEXT: v_lshlrev_b32_e32 v1, 16, v19
+; VI-NEXT: v_or_b32_e32 v48, v2, v1
+; VI-NEXT: v_add_u16_sdwa v1, v4, v33 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_1 src1_sel:DWORD
+; VI-NEXT: buffer_store_dword v19, off, s[0:3], s32 offset:188 ; 4-byte Folded Spill
+; VI-NEXT: v_add_u16_e32 v19, 3, v4
+; VI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:400 ; 4-byte Folded Spill
+; VI-NEXT: v_add_u16_e32 v4, 3, v3
+; VI-NEXT: v_add_u16_sdwa v3, v3, v33 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_1 src1_sel:DWORD
+; VI-NEXT: v_lshlrev_b32_e32 v1, 16, v1
+; VI-NEXT: v_or_b32_e32 v47, v19, v1
+; VI-NEXT: v_lshlrev_b32_e32 v1, 16, v3
+; VI-NEXT: v_or_b32_e32 v46, v4, v1
+; VI-NEXT: v_add_u16_sdwa v1, v6, v33 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_1 src1_sel:DWORD
+; VI-NEXT: buffer_store_dword v21, off, s[0:3], s32 offset:396 ; 4-byte Folded Spill
+; VI-NEXT: buffer_store_dword v3, off, s[0:3], s32 offset:184 ; 4-byte Folded Spill
+; VI-NEXT: v_add_u16_e32 v21, 3, v6
+; VI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:404 ; 4-byte Folded Spill
+; VI-NEXT: v_add_u16_sdwa v3, v5, v33 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_1 src1_sel:DWORD
+; VI-NEXT: v_lshlrev_b32_e32 v1, 16, v1
+; VI-NEXT: v_add_u16_e32 v6, 3, v5
+; VI-NEXT: v_or_b32_e32 v44, v21, v1
+; VI-NEXT: v_lshlrev_b32_e32 v1, 16, v3
+; VI-NEXT: v_or_b32_e32 v43, v6, v1
+; VI-NEXT: v_add_u16_sdwa v1, v8, v33 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_1 src1_sel:DWORD
+; VI-NEXT: buffer_store_dword v3, off, s[0:3], s32 offset:180 ; 4-byte Folded Spill
+; VI-NEXT: v_add_u16_e32 v23, 3, v8
+; VI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:408 ; 4-byte Folded Spill
+; VI-NEXT: v_add_u16_sdwa v3, v7, v33 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_1 src1_sel:DWORD
+; VI-NEXT: v_lshlrev_b32_e32 v1, 16, v1
+; VI-NEXT: v_add_u16_e32 v8, 3, v7
+; VI-NEXT: v_or_b32_e32 v42, v23, v1
+; VI-NEXT: v_lshlrev_b32_e32 v1, 16, v3
+; VI-NEXT: v_or_b32_e32 v41, v8, v1
+; VI-NEXT: v_add_u16_sdwa v1, v10, v33 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_1 src1_sel:DWORD
+; VI-NEXT: buffer_store_dword v3, off, s[0:3], s32 offset:176 ; 4-byte Folded Spill
+; VI-NEXT: v_add_u16_e32 v25, 3, v10
+; VI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:412 ; 4-byte Folded Spill
+; VI-NEXT: v_add_u16_sdwa v3, v9, v33 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_1 src1_sel:DWORD
+; VI-NEXT: v_lshlrev_b32_e32 v1, 16, v1
+; VI-NEXT: v_add_u16_e32 v10, 3, v9
+; VI-NEXT: v_or_b32_e32 v54, v25, v1
+; VI-NEXT: v_lshlrev_b32_e32 v1, 16, v3
+; VI-NEXT: v_add_u16_sdwa v52, v12, v33 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_1 src1_sel:DWORD
+; VI-NEXT: buffer_store_dword v3, off, s[0:3], s32 offset:172 ; 4-byte Folded Spill
+; VI-NEXT: v_or_b32_e32 v53, v10, v1
+; VI-NEXT: v_add_u16_e32 v27, 3, v12
+; VI-NEXT: v_add_u16_sdwa v3, v11, v33 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_1 src1_sel:DWORD
+; VI-NEXT: v_lshlrev_b32_e32 v1, 16, v52
+; VI-NEXT: v_add_u16_e32 v12, 3, v11
+; VI-NEXT: v_or_b32_e32 v51, v27, v1
+; VI-NEXT: v_lshlrev_b32_e32 v1, 16, v3
+; VI-NEXT: v_or_b32_e32 v50, v12, v1
+; VI-NEXT: v_add_u16_sdwa v1, v14, v33 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_1 src1_sel:DWORD
+; VI-NEXT: buffer_store_dword v3, off, s[0:3], s32 offset:168 ; 4-byte Folded Spill
+; VI-NEXT: v_add_u16_e32 v29, 3, v14
+; VI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:416 ; 4-byte Folded Spill
+; VI-NEXT: v_add_u16_sdwa v3, v13, v33 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_1 src1_sel:DWORD
+; VI-NEXT: v_lshlrev_b32_e32 v1, 16, v1
+; VI-NEXT: v_add_u16_e32 v14, 3, v13
+; VI-NEXT: v_or_b32_e32 v36, v29, v1
+; VI-NEXT: v_lshlrev_b32_e32 v1, 16, v3
+; VI-NEXT: v_add_u16_sdwa v59, v16, v33 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_1 src1_sel:DWORD
+; VI-NEXT: buffer_store_dword v3, off, s[0:3], s32 offset:160 ; 4-byte Folded Spill
+; VI-NEXT: v_or_b32_e32 v35, v14, v1
+; VI-NEXT: v_add_u16_sdwa v3, v15, v33 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_1 src1_sel:DWORD
; VI-NEXT: v_add_u16_e32 v16, 3, v16
-; VI-NEXT: v_add_u16_e32 v32, 3, v15
-; VI-NEXT: v_lshlrev_b32_e32 v15, 16, v60
-; VI-NEXT: buffer_store_dword v16, off, s[0:3], s32 offset:420 ; 4-byte Folded Spill
-; VI-NEXT: v_or_b32_e32 v16, v16, v15
-; VI-NEXT: v_lshlrev_b32_e32 v15, 16, v31
-; VI-NEXT: buffer_store_dword v31, off, s[0:3], s32 offset:192 ; 4-byte Folded Spill
-; VI-NEXT: v_or_b32_e32 v15, v32, v15
-; VI-NEXT: v_lshrrev_b32_e32 v31, 8, v16
-; VI-NEXT: buffer_store_dword v31, off, s[0:3], s32 offset:228 ; 4-byte Folded Spill
-; VI-NEXT: v_lshrrev_b32_e32 v31, 8, v15
-; VI-NEXT: v_lshrrev_b64 v[15:16], 24, v[15:16]
-; VI-NEXT: buffer_store_dword v52, off, s[0:3], s32 offset:224 ; 4-byte Folded Spill
-; VI-NEXT: buffer_store_dword v38, off, s[0:3], s32 offset:400 ; 4-byte Folded Spill
-; VI-NEXT: buffer_store_dword v49, off, s[0:3], s32 offset:416 ; 4-byte Folded Spill
-; VI-NEXT: buffer_store_dword v31, off, s[0:3], s32 offset:232 ; 4-byte Folded Spill
-; VI-NEXT: buffer_store_dword v15, off, s[0:3], s32 offset:76 ; 4-byte Folded Spill
-; VI-NEXT: buffer_store_dword v16, off, s[0:3], s32 offset:80 ; 4-byte Folded Spill
-; VI-NEXT: v_or_b32_e32 v13, v49, v13
-; VI-NEXT: v_lshrrev_b32_e32 v15, 8, v14
-; VI-NEXT: buffer_store_dword v15, off, s[0:3], s32 offset:236 ; 4-byte Folded Spill
-; VI-NEXT: v_lshrrev_b32_e32 v15, 8, v13
-; VI-NEXT: v_lshrrev_b64 v[13:14], 24, v[13:14]
-; VI-NEXT: buffer_store_dword v15, off, s[0:3], s32 offset:240 ; 4-byte Folded Spill
-; VI-NEXT: buffer_store_dword v13, off, s[0:3], s32 offset:84 ; 4-byte Folded Spill
-; VI-NEXT: buffer_store_dword v14, off, s[0:3], s32 offset:88 ; 4-byte Folded Spill
-; VI-NEXT: v_lshrrev_b32_e32 v13, 8, v12
-; VI-NEXT: buffer_store_dword v13, off, s[0:3], s32 offset:244 ; 4-byte Folded Spill
-; VI-NEXT: v_lshrrev_b32_e32 v13, 8, v11
-; VI-NEXT: v_lshrrev_b64 v[11:12], 24, v[11:12]
-; VI-NEXT: buffer_store_dword v13, off, s[0:3], s32 offset:248 ; 4-byte Folded Spill
-; VI-NEXT: buffer_store_dword v11, off, s[0:3], s32 offset:92 ; 4-byte Folded Spill
-; VI-NEXT: buffer_store_dword v12, off, s[0:3], s32 offset:96 ; 4-byte Folded Spill
-; VI-NEXT: v_lshrrev_b32_e32 v11, 8, v10
-; VI-NEXT: buffer_store_dword v11, off, s[0:3], s32 offset:252 ; 4-byte Folded Spill
-; VI-NEXT: v_lshrrev_b32_e32 v11, 8, v9
-; VI-NEXT: v_lshrrev_b64 v[9:10], 24, v[9:10]
-; VI-NEXT: buffer_store_dword v11, off, s[0:3], s32 offset:256 ; 4-byte Folded Spill
-; VI-NEXT: buffer_store_dword v9, off, s[0:3], s32 offset:100 ; 4-byte Folded Spill
-; VI-NEXT: buffer_store_dword v10, off, s[0:3], s32 offset:104 ; 4-byte Folded Spill
-; VI-NEXT: v_lshrrev_b32_e32 v9, 8, v8
-; VI-NEXT: buffer_store_dword v9, off, s[0:3], s32 offset:260 ; 4-byte Folded Spill
-; VI-NEXT: v_lshrrev_b32_e32 v9, 8, v7
-; VI-NEXT: v_lshrrev_b64 v[7:8], 24, v[7:8]
-; VI-NEXT: buffer_store_dword v9, off, s[0:3], s32 offset:272 ; 4-byte Folded Spill
-; VI-NEXT: buffer_store_dword v7, off, s[0:3], s32 offset:132 ; 4-byte Folded Spill
-; VI-NEXT: buffer_store_dword v8, off, s[0:3], s32 offset:136 ; 4-byte Folded Spill
-; VI-NEXT: v_lshrrev_b32_e32 v7, 8, v6
-; VI-NEXT: buffer_store_dword v7, off, s[0:3], s32 offset:276 ; 4-byte Folded Spill
-; VI-NEXT: v_lshrrev_b32_e32 v7, 8, v5
-; VI-NEXT: v_lshrrev_b64 v[5:6], 24, v[5:6]
-; VI-NEXT: buffer_store_dword v7, off, s[0:3], s32 offset:284 ; 4-byte Folded Spill
-; VI-NEXT: buffer_store_dword v5, off, s[0:3], s32 offset:148 ; 4-byte Folded Spill
-; VI-NEXT: buffer_store_dword v6, off, s[0:3], s32 offset:152 ; 4-byte Folded Spill
-; VI-NEXT: v_lshrrev_b32_e32 v5, 8, v4
-; VI-NEXT: buffer_store_dword v5, off, s[0:3], s32 offset:436 ; 4-byte Folded Spill
-; VI-NEXT: v_lshrrev_b32_e32 v5, 8, v3
-; VI-NEXT: v_lshrrev_b64 v[40:41], 24, v[3:4]
-; VI-NEXT: v_lshrrev_b32_e32 v3, 8, v2
-; VI-NEXT: buffer_store_dword v3, off, s[0:3], s32 offset:444 ; 4-byte Folded Spill
-; VI-NEXT: v_lshrrev_b32_e32 v3, 8, v1
-; VI-NEXT: v_lshrrev_b64 v[43:44], 24, v[1:2]
-; VI-NEXT: v_lshrrev_b32_e32 v1, 8, v37
-; VI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:424 ; 4-byte Folded Spill
-; VI-NEXT: v_lshrrev_b32_e32 v1, 8, v36
-; VI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:428 ; 4-byte Folded Spill
-; VI-NEXT: v_lshrrev_b64 v[1:2], 24, v[36:37]
-; VI-NEXT: buffer_store_dword v5, off, s[0:3], s32 offset:440 ; 4-byte Folded Spill
-; VI-NEXT: buffer_store_dword v3, off, s[0:3], s32 offset:452 ; 4-byte Folded Spill
-; VI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:108 ; 4-byte Folded Spill
-; VI-NEXT: buffer_store_dword v2, off, s[0:3], s32 offset:112 ; 4-byte Folded Spill
-; VI-NEXT: v_lshrrev_b32_e32 v1, 8, v30
-; VI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:448 ; 4-byte Folded Spill
-; VI-NEXT: v_lshrrev_b32_e32 v1, 8, v29
-; VI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:456 ; 4-byte Folded Spill
-; VI-NEXT: v_lshrrev_b64 v[1:2], 24, v[29:30]
-; VI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:116 ; 4-byte Folded Spill
-; VI-NEXT: buffer_store_dword v2, off, s[0:3], s32 offset:120 ; 4-byte Folded Spill
-; VI-NEXT: v_lshrrev_b32_e32 v1, 8, v28
-; VI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:460 ; 4-byte Folded Spill
-; VI-NEXT: v_lshrrev_b32_e32 v1, 8, v27
-; VI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:464 ; 4-byte Folded Spill
-; VI-NEXT: v_lshrrev_b64 v[1:2], 24, v[27:28]
-; VI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:124 ; 4-byte Folded Spill
-; VI-NEXT: buffer_store_dword v2, off, s[0:3], s32 offset:128 ; 4-byte Folded Spill
-; VI-NEXT: v_lshrrev_b32_e32 v1, 8, v26
-; VI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:468 ; 4-byte Folded Spill
-; VI-NEXT: v_lshrrev_b32_e32 v1, 8, v25
-; VI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:472 ; 4-byte Folded Spill
-; VI-NEXT: v_lshrrev_b64 v[1:2], 24, v[25:26]
-; VI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:140 ; 4-byte Folded Spill
-; VI-NEXT: buffer_store_dword v2, off, s[0:3], s32 offset:144 ; 4-byte Folded Spill
-; VI-NEXT: v_lshrrev_b32_e32 v1, 8, v24
-; VI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:480 ; 4-byte Folded Spill
-; VI-NEXT: v_lshrrev_b32_e32 v1, 8, v23
-; VI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:488 ; 4-byte Folded Spill
-; VI-NEXT: v_lshrrev_b32_e32 v1, 8, v22
-; VI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:500 ; 4-byte Folded Spill
-; VI-NEXT: v_lshrrev_b32_e32 v1, 8, v21
-; VI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:504 ; 4-byte Folded Spill
-; VI-NEXT: v_bfe_u32 v1, v60, 8, 8
-; VI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:268 ; 4-byte Folded Spill
-; VI-NEXT: v_bfe_u32 v1, v33, 8, 8
-; VI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:280 ; 4-byte Folded Spill
-; VI-NEXT: v_bfe_u32 v1, v63, 8, 8
-; VI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:288 ; 4-byte Folded Spill
-; VI-NEXT: v_bfe_u32 v1, v59, 8, 8
-; VI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:292 ; 4-byte Folded Spill
-; VI-NEXT: v_bfe_u32 v1, v34, 8, 8
-; VI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:296 ; 4-byte Folded Spill
-; VI-NEXT: v_bfe_u32 v1, v47, 8, 8
-; VI-NEXT: v_lshrrev_b64 v[44:45], 24, v[19:20]
-; VI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:300 ; 4-byte Folded Spill
-; VI-NEXT: v_bfe_u32 v1, v56, 8, 8
-; VI-NEXT: v_lshrrev_b64 v[45:46], 24, v[17:18]
-; VI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:476 ; 4-byte Folded Spill
-; VI-NEXT: v_bfe_u32 v1, v57, 8, 8
-; VI-NEXT: v_mov_b32_e32 v46, v35
-; VI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:492 ; 4-byte Folded Spill
-; VI-NEXT: v_bfe_u32 v1, v52, 8, 8
-; VI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:432 ; 4-byte Folded Spill
-; VI-NEXT: v_bfe_u32 v1, v46, 8, 8
-; VI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:484 ; 4-byte Folded Spill
-; VI-NEXT: buffer_store_dword v39, off, s[0:3], s32 offset:156 ; 4-byte Folded Spill
-; VI-NEXT: v_bfe_u32 v1, v39, 8, 8
-; VI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:496 ; 4-byte Folded Spill
-; VI-NEXT: buffer_store_dword v58, off, s[0:3], s32 offset:160 ; 4-byte Folded Spill
-; VI-NEXT: v_mov_b32_e32 v49, v53
-; VI-NEXT: buffer_load_dword v53, off, s[0:3], s32 offset:512 ; 4-byte Folded Reload
-; VI-NEXT: v_mov_b32_e32 v52, v51
-; VI-NEXT: v_bfe_u32 v31, v51, 8, 8
-; VI-NEXT: buffer_load_dword v51, off, s[0:3], s32 offset:508 ; 4-byte Folded Reload
-; VI-NEXT: v_lshrrev_b64 v[54:55], 24, v[23:24]
-; VI-NEXT: v_lshrrev_b64 v[41:42], 24, v[21:22]
-; VI-NEXT: v_lshrrev_b32_e32 v42, 8, v20
-; VI-NEXT: v_lshrrev_b32_e32 v38, 8, v19
-; VI-NEXT: v_lshrrev_b32_e32 v62, 8, v18
-; VI-NEXT: v_lshrrev_b32_e32 v55, 8, v17
-; VI-NEXT: v_bfe_u32 v35, v58, 8, 8
-; VI-NEXT: buffer_store_dword v61, off, s[0:3], s32 offset:164 ; 4-byte Folded Spill
-; VI-NEXT: v_bfe_u32 v39, v61, 8, 8
-; VI-NEXT: v_bfe_u32 v58, v48, 8, 8
-; VI-NEXT: buffer_store_dword v48, off, s[0:3], s32 offset:176 ; 4-byte Folded Spill
+; VI-NEXT: v_lshlrev_b32_e32 v1, 16, v59
+; VI-NEXT: v_add_u16_e32 v15, 3, v15
+; VI-NEXT: v_or_b32_e32 v34, v16, v1
+; VI-NEXT: v_lshlrev_b32_e32 v1, 16, v3
+; VI-NEXT: v_or_b32_e32 v33, v15, v1
+; VI-NEXT: v_mov_b32_e32 v31, v32
+; VI-NEXT: v_mov_b32_e32 v32, v45
+; VI-NEXT: v_lshrrev_b32_e32 v45, 8, v34
+; VI-NEXT: buffer_store_dword v3, off, s[0:3], s32 offset:164 ; 4-byte Folded Spill
+; VI-NEXT: v_mov_b32_e32 v1, v2
+; VI-NEXT: v_mov_b32_e32 v2, v17
+; VI-NEXT: v_mov_b32_e32 v3, v4
+; VI-NEXT: v_mov_b32_e32 v4, v19
+; VI-NEXT: v_mov_b32_e32 v5, v6
+; VI-NEXT: v_mov_b32_e32 v6, v21
+; VI-NEXT: v_mov_b32_e32 v7, v8
+; VI-NEXT: v_mov_b32_e32 v8, v23
+; VI-NEXT: v_mov_b32_e32 v9, v10
+; VI-NEXT: v_mov_b32_e32 v10, v25
+; VI-NEXT: v_mov_b32_e32 v11, v12
+; VI-NEXT: v_mov_b32_e32 v12, v27
+; VI-NEXT: v_mov_b32_e32 v13, v14
+; VI-NEXT: v_mov_b32_e32 v14, v29
+; VI-NEXT: v_mov_b32_e32 v17, v18
+; VI-NEXT: buffer_load_dword v18, off, s[0:3], s32 offset:88 ; 4-byte Folded Reload
+; VI-NEXT: v_mov_b32_e32 v19, v20
+; VI-NEXT: buffer_load_dword v20, off, s[0:3], s32 offset:96 ; 4-byte Folded Reload
+; VI-NEXT: v_mov_b32_e32 v21, v22
+; VI-NEXT: buffer_load_dword v22, off, s[0:3], s32 offset:104 ; 4-byte Folded Reload
+; VI-NEXT: v_mov_b32_e32 v23, v24
+; VI-NEXT: buffer_load_dword v24, off, s[0:3], s32 offset:112 ; 4-byte Folded Reload
+; VI-NEXT: v_mov_b32_e32 v25, v26
+; VI-NEXT: buffer_load_dword v26, off, s[0:3], s32 offset:120 ; 4-byte Folded Reload
+; VI-NEXT: v_mov_b32_e32 v27, v28
+; VI-NEXT: buffer_load_dword v28, off, s[0:3], s32 offset:192 ; 4-byte Folded Reload
+; VI-NEXT: v_mov_b32_e32 v29, v30
+; VI-NEXT: buffer_load_dword v30, off, s[0:3], s32 offset:196 ; 4-byte Folded Reload
+; VI-NEXT: buffer_store_dword v45, off, s[0:3], s32 offset:192 ; 4-byte Folded Spill
+; VI-NEXT: v_lshrrev_b32_e32 v45, 8, v33
+; VI-NEXT: v_lshrrev_b64 v[33:34], 24, v[33:34]
+; VI-NEXT: buffer_store_dword v45, off, s[0:3], s32 offset:196 ; 4-byte Folded Spill
+; VI-NEXT: buffer_store_dword v33, off, s[0:3], s32 offset:104 ; 4-byte Folded Spill
+; VI-NEXT: buffer_store_dword v34, off, s[0:3], s32 offset:108 ; 4-byte Folded Spill
+; VI-NEXT: v_lshrrev_b32_e32 v33, 8, v36
+; VI-NEXT: buffer_store_dword v33, off, s[0:3], s32 offset:200 ; 4-byte Folded Spill
+; VI-NEXT: v_lshrrev_b32_e32 v33, 8, v35
+; VI-NEXT: buffer_store_dword v33, off, s[0:3], s32 offset:204 ; 4-byte Folded Spill
+; VI-NEXT: v_lshrrev_b64 v[33:34], 24, v[35:36]
+; VI-NEXT: buffer_store_dword v33, off, s[0:3], s32 offset:112 ; 4-byte Folded Spill
+; VI-NEXT: buffer_store_dword v34, off, s[0:3], s32 offset:116 ; 4-byte Folded Spill
+; VI-NEXT: v_lshrrev_b32_e32 v33, 8, v51
+; VI-NEXT: buffer_store_dword v33, off, s[0:3], s32 offset:212 ; 4-byte Folded Spill
+; VI-NEXT: v_lshrrev_b32_e32 v33, 8, v50
+; VI-NEXT: buffer_store_dword v33, off, s[0:3], s32 offset:220 ; 4-byte Folded Spill
+; VI-NEXT: v_lshrrev_b64 v[33:34], 24, v[50:51]
+; VI-NEXT: buffer_store_dword v33, off, s[0:3], s32 offset:120 ; 4-byte Folded Spill
+; VI-NEXT: buffer_store_dword v34, off, s[0:3], s32 offset:124 ; 4-byte Folded Spill
+; VI-NEXT: v_lshrrev_b32_e32 v33, 8, v54
+; VI-NEXT: buffer_store_dword v33, off, s[0:3], s32 offset:240 ; 4-byte Folded Spill
+; VI-NEXT: v_lshrrev_b32_e32 v33, 8, v53
+; VI-NEXT: buffer_store_dword v33, off, s[0:3], s32 offset:248 ; 4-byte Folded Spill
+; VI-NEXT: v_lshrrev_b64 v[33:34], 24, v[53:54]
+; VI-NEXT: v_lshrrev_b32_e32 v34, 8, v42
+; VI-NEXT: buffer_store_dword v34, off, s[0:3], s32 offset:264 ; 4-byte Folded Spill
+; VI-NEXT: v_lshrrev_b32_e32 v34, 8, v41
+; VI-NEXT: buffer_store_dword v34, off, s[0:3], s32 offset:276 ; 4-byte Folded Spill
+; VI-NEXT: v_lshrrev_b32_e32 v34, 8, v44
+; VI-NEXT: buffer_store_dword v34, off, s[0:3], s32 offset:292 ; 4-byte Folded Spill
+; VI-NEXT: v_lshrrev_b32_e32 v34, 8, v43
+; VI-NEXT: buffer_store_dword v34, off, s[0:3], s32 offset:300 ; 4-byte Folded Spill
+; VI-NEXT: v_lshrrev_b32_e32 v34, 8, v47
+; VI-NEXT: buffer_store_dword v34, off, s[0:3], s32 offset:320 ; 4-byte Folded Spill
+; VI-NEXT: v_lshrrev_b32_e32 v34, 8, v46
+; VI-NEXT: buffer_store_dword v34, off, s[0:3], s32 offset:328 ; 4-byte Folded Spill
+; VI-NEXT: v_lshrrev_b32_e32 v34, 8, v49
+; VI-NEXT: buffer_store_dword v34, off, s[0:3], s32 offset:348 ; 4-byte Folded Spill
+; VI-NEXT: v_lshrrev_b32_e32 v34, 8, v48
+; VI-NEXT: buffer_store_dword v34, off, s[0:3], s32 offset:356 ; 4-byte Folded Spill
+; VI-NEXT: v_lshrrev_b32_e32 v34, 8, v38
+; VI-NEXT: v_lshrrev_b64 v[44:45], 24, v[43:44]
+; VI-NEXT: buffer_store_dword v34, off, s[0:3], s32 offset:208 ; 4-byte Folded Spill
+; VI-NEXT: v_lshrrev_b32_e32 v34, 8, v37
+; VI-NEXT: v_lshrrev_b64 v[45:46], 24, v[46:47]
+; VI-NEXT: buffer_store_dword v34, off, s[0:3], s32 offset:216 ; 4-byte Folded Spill
+; VI-NEXT: v_lshrrev_b64 v[34:35], 24, v[37:38]
+; VI-NEXT: v_lshrrev_b64 v[46:47], 24, v[48:49]
+; VI-NEXT: buffer_load_dword v47, off, s[0:3], s32 offset:388 ; 4-byte Folded Reload
+; VI-NEXT: buffer_store_dword v34, off, s[0:3], s32 offset:88 ; 4-byte Folded Spill
+; VI-NEXT: buffer_store_dword v35, off, s[0:3], s32 offset:92 ; 4-byte Folded Spill
+; VI-NEXT: v_lshrrev_b32_e32 v34, 8, v40
+; VI-NEXT: buffer_store_dword v34, off, s[0:3], s32 offset:232 ; 4-byte Folded Spill
+; VI-NEXT: v_lshrrev_b32_e32 v34, 8, v39
+; VI-NEXT: buffer_store_dword v34, off, s[0:3], s32 offset:236 ; 4-byte Folded Spill
+; VI-NEXT: v_lshrrev_b64 v[34:35], 24, v[39:40]
+; VI-NEXT: buffer_store_dword v34, off, s[0:3], s32 offset:96 ; 4-byte Folded Spill
+; VI-NEXT: buffer_store_dword v35, off, s[0:3], s32 offset:100 ; 4-byte Folded Spill
+; VI-NEXT: v_lshrrev_b32_e32 v34, 8, v56
+; VI-NEXT: buffer_store_dword v34, off, s[0:3], s32 offset:256 ; 4-byte Folded Spill
+; VI-NEXT: v_lshrrev_b32_e32 v34, 8, v55
+; VI-NEXT: buffer_store_dword v34, off, s[0:3], s32 offset:260 ; 4-byte Folded Spill
+; VI-NEXT: v_lshrrev_b32_e32 v34, 8, v58
+; VI-NEXT: buffer_store_dword v34, off, s[0:3], s32 offset:280 ; 4-byte Folded Spill
+; VI-NEXT: v_lshrrev_b32_e32 v34, 8, v57
+; VI-NEXT: buffer_store_dword v34, off, s[0:3], s32 offset:284 ; 4-byte Folded Spill
+; VI-NEXT: v_lshrrev_b32_e32 v34, 8, v63
+; VI-NEXT: buffer_store_dword v34, off, s[0:3], s32 offset:304 ; 4-byte Folded Spill
+; VI-NEXT: v_lshrrev_b32_e32 v34, 8, v62
+; VI-NEXT: buffer_store_dword v34, off, s[0:3], s32 offset:312 ; 4-byte Folded Spill
+; VI-NEXT: v_lshrrev_b32_e32 v34, 8, v61
+; VI-NEXT: buffer_store_dword v34, off, s[0:3], s32 offset:324 ; 4-byte Folded Spill
+; VI-NEXT: v_lshrrev_b32_e32 v34, 8, v60
+; VI-NEXT: v_lshrrev_b64 v[35:36], 24, v[55:56]
+; VI-NEXT: v_lshrrev_b64 v[49:50], 24, v[62:63]
+; VI-NEXT: buffer_store_dword v34, off, s[0:3], s32 offset:336 ; 4-byte Folded Spill
+; VI-NEXT: v_lshrrev_b64 v[61:62], 24, v[60:61]
+; VI-NEXT: buffer_load_dword v36, off, s[0:3], s32 offset:392 ; 4-byte Folded Reload
+; VI-NEXT: buffer_load_dword v50, off, s[0:3], s32 offset:380 ; 4-byte Folded Reload
+; VI-NEXT: buffer_load_dword v62, off, s[0:3], s32 offset:412 ; 4-byte Folded Reload
+; VI-NEXT: buffer_load_dword v53, off, s[0:3], s32 offset:376 ; 4-byte Folded Reload
+; VI-NEXT: buffer_load_dword v54, off, s[0:3], s32 offset:224 ; 4-byte Folded Reload
+; VI-NEXT: buffer_load_dword v55, off, s[0:3], s32 offset:228 ; 4-byte Folded Reload
+; VI-NEXT: v_lshrrev_b64 v[38:39], 24, v[57:58]
+; VI-NEXT: v_mov_b32_e32 v57, v52
+; VI-NEXT: v_lshrrev_b64 v[41:42], 24, v[41:42]
+; VI-NEXT: s_waitcnt vmcnt(0)
+; VI-NEXT: v_lshrrev_b32_e32 v34, 8, v55
+; VI-NEXT: buffer_store_dword v34, off, s[0:3], s32 offset:352 ; 4-byte Folded Spill
+; VI-NEXT: v_lshrrev_b32_e32 v34, 8, v54
+; VI-NEXT: buffer_store_dword v34, off, s[0:3], s32 offset:360 ; 4-byte Folded Spill
+; VI-NEXT: v_lshrrev_b64 v[39:40], 24, v[54:55]
+; VI-NEXT: buffer_load_dword v40, off, s[0:3], s32 offset:372 ; 4-byte Folded Reload
+; VI-NEXT: buffer_load_dword v54, off, s[0:3], s32 offset:152 ; 4-byte Folded Reload
+; VI-NEXT: buffer_load_dword v55, off, s[0:3], s32 offset:156 ; 4-byte Folded Reload
+; VI-NEXT: buffer_store_dword v59, off, s[0:3], s32 offset:152 ; 4-byte Folded Spill
+; VI-NEXT: v_bfe_u32 v34, v59, 8, 8
+; VI-NEXT: buffer_load_dword v59, off, s[0:3], s32 offset:416 ; 4-byte Folded Reload
+; VI-NEXT: buffer_store_dword v34, off, s[0:3], s32 offset:224 ; 4-byte Folded Spill
+; VI-NEXT: s_waitcnt vmcnt(4)
+; VI-NEXT: v_lshrrev_b32_e32 v58, 8, v54
; VI-NEXT: s_waitcnt vmcnt(3)
-; VI-NEXT: v_bfe_u32 v61, v53, 8, 8
+; VI-NEXT: v_lshrrev_b64 v[42:43], 24, v[54:55]
+; VI-NEXT: buffer_load_dword v43, off, s[0:3], s32 offset:384 ; 4-byte Folded Reload
+; VI-NEXT: v_lshrrev_b32_e32 v37, 8, v55
+; VI-NEXT: s_waitcnt vmcnt(2)
+; VI-NEXT: v_bfe_u32 v34, v59, 8, 8
+; VI-NEXT: buffer_store_dword v34, off, s[0:3], s32 offset:252 ; 4-byte Folded Spill
+; VI-NEXT: v_bfe_u32 v34, v52, 8, 8
+; VI-NEXT: buffer_store_dword v34, off, s[0:3], s32 offset:272 ; 4-byte Folded Spill
+; VI-NEXT: v_bfe_u32 v34, v62, 8, 8
+; VI-NEXT: buffer_store_dword v34, off, s[0:3], s32 offset:296 ; 4-byte Folded Spill
+; VI-NEXT: buffer_load_dword v34, off, s[0:3], s32 offset:408 ; 4-byte Folded Reload
+; VI-NEXT: s_waitcnt vmcnt(0)
+; VI-NEXT: v_mov_b32_e32 v52, v34
+; VI-NEXT: v_bfe_u32 v34, v34, 8, 8
+; VI-NEXT: buffer_store_dword v34, off, s[0:3], s32 offset:316 ; 4-byte Folded Spill
+; VI-NEXT: buffer_load_dword v34, off, s[0:3], s32 offset:404 ; 4-byte Folded Reload
+; VI-NEXT: s_waitcnt vmcnt(0)
+; VI-NEXT: v_mov_b32_e32 v54, v34
+; VI-NEXT: v_bfe_u32 v34, v34, 8, 8
+; VI-NEXT: buffer_store_dword v34, off, s[0:3], s32 offset:340 ; 4-byte Folded Spill
+; VI-NEXT: buffer_load_dword v34, off, s[0:3], s32 offset:400 ; 4-byte Folded Reload
+; VI-NEXT: s_waitcnt vmcnt(0)
+; VI-NEXT: v_mov_b32_e32 v51, v34
+; VI-NEXT: v_bfe_u32 v34, v34, 8, 8
+; VI-NEXT: buffer_store_dword v34, off, s[0:3], s32 offset:364 ; 4-byte Folded Spill
+; VI-NEXT: buffer_load_dword v34, off, s[0:3], s32 offset:396 ; 4-byte Folded Reload
+; VI-NEXT: s_waitcnt vmcnt(0)
+; VI-NEXT: v_mov_b32_e32 v56, v34
+; VI-NEXT: v_bfe_u32 v60, v34, 8, 8
+; VI-NEXT: buffer_load_dword v34, off, s[0:3], s32 offset:80 ; 4-byte Folded Reload
+; VI-NEXT: s_waitcnt vmcnt(0)
+; VI-NEXT: v_bfe_u32 v34, v34, 8, 8
+; VI-NEXT: buffer_store_dword v34, off, s[0:3], s32 offset:244 ; 4-byte Folded Spill
+; VI-NEXT: v_bfe_u32 v34, v36, 8, 8
+; VI-NEXT: buffer_store_dword v34, off, s[0:3], s32 offset:268 ; 4-byte Folded Spill
+; VI-NEXT: v_bfe_u32 v34, v50, 8, 8
+; VI-NEXT: buffer_store_dword v34, off, s[0:3], s32 offset:288 ; 4-byte Folded Spill
+; VI-NEXT: v_bfe_u32 v34, v53, 8, 8
+; VI-NEXT: buffer_store_dword v34, off, s[0:3], s32 offset:308 ; 4-byte Folded Spill
+; VI-NEXT: v_bfe_u32 v34, v40, 8, 8
+; VI-NEXT: buffer_store_dword v34, off, s[0:3], s32 offset:332 ; 4-byte Folded Spill
+; VI-NEXT: buffer_load_dword v34, off, s[0:3], s32 offset:76 ; 4-byte Folded Reload
+; VI-NEXT: s_waitcnt vmcnt(0)
+; VI-NEXT: v_bfe_u32 v34, v34, 8, 8
+; VI-NEXT: buffer_store_dword v34, off, s[0:3], s32 offset:344 ; 4-byte Folded Spill
+; VI-NEXT: buffer_load_dword v34, off, s[0:3], s32 offset:84 ; 4-byte Folded Reload
+; VI-NEXT: s_waitcnt vmcnt(0)
+; VI-NEXT: v_bfe_u32 v48, v34, 8, 8
+; VI-NEXT: buffer_load_dword v34, off, s[0:3], s32 offset:368 ; 4-byte Folded Reload
+; VI-NEXT: s_waitcnt vmcnt(0)
+; VI-NEXT: v_mov_b32_e32 v63, v34
+; VI-NEXT: v_bfe_u32 v55, v34, 8, 8
; VI-NEXT: .LBB98_4: ; %end
; VI-NEXT: s_or_b64 exec, exec, s[4:5]
-; VI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:452 ; 4-byte Folded Reload
-; VI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:404 ; 4-byte Folded Reload
-; VI-NEXT: buffer_load_dword v3, off, s[0:3], s32 offset:220 ; 4-byte Folded Reload
-; VI-NEXT: s_waitcnt vmcnt(2)
-; VI-NEXT: v_lshlrev_b16_e32 v1, 8, v1
-; VI-NEXT: s_waitcnt vmcnt(1)
-; VI-NEXT: v_or_b32_sdwa v1, v2, v1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
-; VI-NEXT: v_lshlrev_b16_e32 v2, 8, v43
+; VI-NEXT: buffer_load_dword v34, off, s[0:3], s32 offset:356 ; 4-byte Folded Reload
; VI-NEXT: s_waitcnt vmcnt(0)
-; VI-NEXT: v_or_b32_sdwa v2, v3, v2 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
-; VI-NEXT: v_or_b32_sdwa v1, v1, v2 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_0 src1_sel:DWORD
-; VI-NEXT: buffer_store_dword v1, v0, s[0:3], 0 offen
-; VI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:444 ; 4-byte Folded Reload
-; VI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:384 ; 4-byte Folded Reload
-; VI-NEXT: s_waitcnt vmcnt(1)
-; VI-NEXT: v_lshlrev_b16_e32 v1, 8, v1
+; VI-NEXT: v_lshlrev_b16_e32 v34, 8, v34
+; VI-NEXT: v_or_b32_sdwa v1, v1, v34 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; VI-NEXT: buffer_load_dword v34, off, s[0:3], s32 offset:348 ; 4-byte Folded Reload
; VI-NEXT: s_waitcnt vmcnt(0)
-; VI-NEXT: v_or_b32_sdwa v1, v2, v1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
-; VI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:492 ; 4-byte Folded Reload
+; VI-NEXT: v_lshlrev_b16_e32 v34, 8, v34
+; VI-NEXT: v_or_b32_sdwa v2, v2, v34 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; VI-NEXT: v_lshlrev_b16_e32 v34, 8, v46
+; VI-NEXT: buffer_load_dword v46, off, s[0:3], s32 offset:188 ; 4-byte Folded Reload
; VI-NEXT: s_waitcnt vmcnt(0)
-; VI-NEXT: v_lshlrev_b16_e32 v2, 8, v2
-; VI-NEXT: v_or_b32_sdwa v2, v57, v2 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
-; VI-NEXT: v_or_b32_sdwa v1, v1, v2 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_0 src1_sel:DWORD
+; VI-NEXT: v_or_b32_sdwa v34, v46, v34 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; VI-NEXT: v_or_b32_sdwa v1, v1, v34 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_0 src1_sel:DWORD
+; VI-NEXT: buffer_store_dword v1, v0, s[0:3], 0 offen
+; VI-NEXT: v_lshlrev_b16_e32 v1, 8, v60
+; VI-NEXT: v_or_b32_sdwa v1, v56, v1 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; VI-NEXT: v_or_b32_sdwa v1, v2, v1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_0 src1_sel:DWORD
; VI-NEXT: v_add_u32_e32 v2, vcc, 4, v0
; VI-NEXT: buffer_store_dword v1, v2, s[0:3], 0 offen
-; VI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:440 ; 4-byte Folded Reload
-; VI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:408 ; 4-byte Folded Reload
-; VI-NEXT: buffer_load_dword v3, off, s[0:3], s32 offset:216 ; 4-byte Folded Reload
-; VI-NEXT: s_waitcnt vmcnt(2)
-; VI-NEXT: v_lshlrev_b16_e32 v1, 8, v1
-; VI-NEXT: s_waitcnt vmcnt(1)
-; VI-NEXT: v_or_b32_sdwa v1, v2, v1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
-; VI-NEXT: v_lshlrev_b16_e32 v2, 8, v40
+; VI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:184 ; 4-byte Folded Reload
+; VI-NEXT: v_lshlrev_b16_e32 v1, 8, v45
; VI-NEXT: s_waitcnt vmcnt(0)
-; VI-NEXT: v_or_b32_sdwa v2, v3, v2 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
-; VI-NEXT: v_or_b32_sdwa v1, v1, v2 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_0 src1_sel:DWORD
+; VI-NEXT: v_or_b32_sdwa v1, v2, v1 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; VI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:328 ; 4-byte Folded Reload
+; VI-NEXT: s_waitcnt vmcnt(0)
+; VI-NEXT: v_lshlrev_b16_e32 v2, 8, v2
+; VI-NEXT: v_or_b32_sdwa v2, v3, v2 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; VI-NEXT: v_or_b32_sdwa v1, v2, v1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_0 src1_sel:DWORD
; VI-NEXT: v_add_u32_e32 v2, vcc, 8, v0
; VI-NEXT: buffer_store_dword v1, v2, s[0:3], 0 offen
-; VI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:436 ; 4-byte Folded Reload
-; VI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:388 ; 4-byte Folded Reload
+; VI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:320 ; 4-byte Folded Reload
+; VI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:364 ; 4-byte Folded Reload
; VI-NEXT: s_waitcnt vmcnt(1)
; VI-NEXT: v_lshlrev_b16_e32 v1, 8, v1
; VI-NEXT: s_waitcnt vmcnt(0)
-; VI-NEXT: v_or_b32_sdwa v1, v2, v1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
-; VI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:476 ; 4-byte Folded Reload
-; VI-NEXT: s_waitcnt vmcnt(0)
; VI-NEXT: v_lshlrev_b16_e32 v2, 8, v2
-; VI-NEXT: v_or_b32_sdwa v2, v56, v2 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; VI-NEXT: v_or_b32_sdwa v1, v4, v1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; VI-NEXT: v_or_b32_sdwa v2, v51, v2 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
; VI-NEXT: v_or_b32_sdwa v1, v1, v2 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_0 src1_sel:DWORD
; VI-NEXT: v_add_u32_e32 v2, vcc, 12, v0
; VI-NEXT: buffer_store_dword v1, v2, s[0:3], 0 offen
-; VI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:284 ; 4-byte Folded Reload
-; VI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:412 ; 4-byte Folded Reload
-; VI-NEXT: s_waitcnt vmcnt(1)
-; VI-NEXT: v_lshlrev_b16_e32 v1, 8, v1
+; VI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:180 ; 4-byte Folded Reload
+; VI-NEXT: v_lshlrev_b16_e32 v1, 8, v44
; VI-NEXT: s_waitcnt vmcnt(0)
-; VI-NEXT: v_or_b32_sdwa v1, v2, v1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
-; VI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:148 ; 4-byte Folded Reload
-; VI-NEXT: buffer_load_dword v3, off, s[0:3], s32 offset:152 ; 4-byte Folded Reload
-; VI-NEXT: buffer_load_dword v3, off, s[0:3], s32 offset:212 ; 4-byte Folded Reload
-; VI-NEXT: s_waitcnt vmcnt(2)
-; VI-NEXT: v_lshlrev_b16_e32 v2, 8, v2
+; VI-NEXT: v_or_b32_sdwa v1, v2, v1 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; VI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:300 ; 4-byte Folded Reload
; VI-NEXT: s_waitcnt vmcnt(0)
-; VI-NEXT: v_or_b32_sdwa v2, v3, v2 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
-; VI-NEXT: v_or_b32_sdwa v1, v1, v2 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_0 src1_sel:DWORD
+; VI-NEXT: v_lshlrev_b16_e32 v2, 8, v2
+; VI-NEXT: v_or_b32_sdwa v2, v5, v2 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; VI-NEXT: v_or_b32_sdwa v1, v2, v1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_0 src1_sel:DWORD
; VI-NEXT: v_add_u32_e32 v2, vcc, 16, v0
; VI-NEXT: buffer_store_dword v1, v2, s[0:3], 0 offen
-; VI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:276 ; 4-byte Folded Reload
-; VI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:392 ; 4-byte Folded Reload
+; VI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:292 ; 4-byte Folded Reload
+; VI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:340 ; 4-byte Folded Reload
; VI-NEXT: s_waitcnt vmcnt(1)
; VI-NEXT: v_lshlrev_b16_e32 v1, 8, v1
; VI-NEXT: s_waitcnt vmcnt(0)
-; VI-NEXT: v_or_b32_sdwa v1, v2, v1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
-; VI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:300 ; 4-byte Folded Reload
-; VI-NEXT: s_waitcnt vmcnt(0)
; VI-NEXT: v_lshlrev_b16_e32 v2, 8, v2
-; VI-NEXT: v_or_b32_sdwa v2, v47, v2 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; VI-NEXT: v_or_b32_sdwa v1, v6, v1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; VI-NEXT: v_or_b32_sdwa v2, v54, v2 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
; VI-NEXT: v_or_b32_sdwa v1, v1, v2 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_0 src1_sel:DWORD
; VI-NEXT: v_add_u32_e32 v2, vcc, 20, v0
; VI-NEXT: buffer_store_dword v1, v2, s[0:3], 0 offen
-; VI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:272 ; 4-byte Folded Reload
-; VI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:264 ; 4-byte Folded Reload
-; VI-NEXT: s_waitcnt vmcnt(1)
-; VI-NEXT: v_lshlrev_b16_e32 v1, 8, v1
+; VI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:176 ; 4-byte Folded Reload
+; VI-NEXT: v_lshlrev_b16_e32 v1, 8, v41
; VI-NEXT: s_waitcnt vmcnt(0)
-; VI-NEXT: v_or_b32_sdwa v1, v2, v1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
-; VI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:132 ; 4-byte Folded Reload
-; VI-NEXT: buffer_load_dword v3, off, s[0:3], s32 offset:136 ; 4-byte Folded Reload
-; VI-NEXT: buffer_load_dword v3, off, s[0:3], s32 offset:208 ; 4-byte Folded Reload
-; VI-NEXT: s_waitcnt vmcnt(2)
-; VI-NEXT: v_lshlrev_b16_e32 v2, 8, v2
+; VI-NEXT: v_or_b32_sdwa v1, v2, v1 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; VI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:276 ; 4-byte Folded Reload
; VI-NEXT: s_waitcnt vmcnt(0)
-; VI-NEXT: v_or_b32_sdwa v2, v3, v2 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
-; VI-NEXT: v_or_b32_sdwa v1, v1, v2 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_0 src1_sel:DWORD
+; VI-NEXT: v_lshlrev_b16_e32 v2, 8, v2
+; VI-NEXT: v_or_b32_sdwa v2, v7, v2 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; VI-NEXT: v_or_b32_sdwa v1, v2, v1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_0 src1_sel:DWORD
; VI-NEXT: v_add_u32_e32 v2, vcc, 24, v0
; VI-NEXT: buffer_store_dword v1, v2, s[0:3], 0 offen
-; VI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:260 ; 4-byte Folded Reload
-; VI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:396 ; 4-byte Folded Reload
+; VI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:264 ; 4-byte Folded Reload
+; VI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:316 ; 4-byte Folded Reload
; VI-NEXT: s_waitcnt vmcnt(1)
; VI-NEXT: v_lshlrev_b16_e32 v1, 8, v1
; VI-NEXT: s_waitcnt vmcnt(0)
-; VI-NEXT: v_or_b32_sdwa v1, v2, v1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
-; VI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:296 ; 4-byte Folded Reload
-; VI-NEXT: s_waitcnt vmcnt(0)
; VI-NEXT: v_lshlrev_b16_e32 v2, 8, v2
-; VI-NEXT: v_or_b32_sdwa v2, v34, v2 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; VI-NEXT: v_or_b32_sdwa v1, v8, v1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; VI-NEXT: v_or_b32_sdwa v2, v52, v2 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
; VI-NEXT: v_or_b32_sdwa v1, v1, v2 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_0 src1_sel:DWORD
; VI-NEXT: v_add_u32_e32 v2, vcc, 28, v0
; VI-NEXT: buffer_store_dword v1, v2, s[0:3], 0 offen
-; VI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:256 ; 4-byte Folded Reload
-; VI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:368 ; 4-byte Folded Reload
-; VI-NEXT: s_waitcnt vmcnt(1)
-; VI-NEXT: v_lshlrev_b16_e32 v1, 8, v1
+; VI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:172 ; 4-byte Folded Reload
+; VI-NEXT: v_lshlrev_b16_e32 v1, 8, v33
; VI-NEXT: s_waitcnt vmcnt(0)
-; VI-NEXT: v_or_b32_sdwa v1, v2, v1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
-; VI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:100 ; 4-byte Folded Reload
-; VI-NEXT: buffer_load_dword v3, off, s[0:3], s32 offset:104 ; 4-byte Folded Reload
-; VI-NEXT: buffer_load_dword v3, off, s[0:3], s32 offset:204 ; 4-byte Folded Reload
-; VI-NEXT: s_waitcnt vmcnt(2)
-; VI-NEXT: v_lshlrev_b16_e32 v2, 8, v2
+; VI-NEXT: v_or_b32_sdwa v1, v2, v1 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; VI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:248 ; 4-byte Folded Reload
; VI-NEXT: s_waitcnt vmcnt(0)
-; VI-NEXT: v_or_b32_sdwa v2, v3, v2 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
-; VI-NEXT: v_or_b32_sdwa v1, v1, v2 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_0 src1_sel:DWORD
+; VI-NEXT: v_lshlrev_b16_e32 v2, 8, v2
+; VI-NEXT: v_or_b32_sdwa v2, v9, v2 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; VI-NEXT: v_or_b32_sdwa v1, v2, v1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_0 src1_sel:DWORD
; VI-NEXT: v_add_u32_e32 v2, vcc, 32, v0
; VI-NEXT: buffer_store_dword v1, v2, s[0:3], 0 offen
-; VI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:252 ; 4-byte Folded Reload
-; VI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:344 ; 4-byte Folded Reload
+; VI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:240 ; 4-byte Folded Reload
+; VI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:296 ; 4-byte Folded Reload
; VI-NEXT: s_waitcnt vmcnt(1)
; VI-NEXT: v_lshlrev_b16_e32 v1, 8, v1
; VI-NEXT: s_waitcnt vmcnt(0)
-; VI-NEXT: v_or_b32_sdwa v1, v2, v1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
-; VI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:292 ; 4-byte Folded Reload
-; VI-NEXT: s_waitcnt vmcnt(0)
; VI-NEXT: v_lshlrev_b16_e32 v2, 8, v2
-; VI-NEXT: v_or_b32_sdwa v2, v59, v2 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; VI-NEXT: v_or_b32_sdwa v1, v10, v1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; VI-NEXT: v_or_b32_sdwa v2, v62, v2 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
; VI-NEXT: v_or_b32_sdwa v1, v1, v2 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_0 src1_sel:DWORD
; VI-NEXT: v_add_u32_e32 v2, vcc, 36, v0
; VI-NEXT: buffer_store_dword v1, v2, s[0:3], 0 offen
-; VI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:248 ; 4-byte Folded Reload
-; VI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:380 ; 4-byte Folded Reload
-; VI-NEXT: s_waitcnt vmcnt(1)
+; VI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:120 ; 4-byte Folded Reload
+; VI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:124 ; 4-byte Folded Reload
+; VI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:168 ; 4-byte Folded Reload
+; VI-NEXT: s_waitcnt vmcnt(2)
; VI-NEXT: v_lshlrev_b16_e32 v1, 8, v1
; VI-NEXT: s_waitcnt vmcnt(0)
-; VI-NEXT: v_or_b32_sdwa v1, v2, v1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
-; VI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:92 ; 4-byte Folded Reload
-; VI-NEXT: buffer_load_dword v3, off, s[0:3], s32 offset:96 ; 4-byte Folded Reload
-; VI-NEXT: buffer_load_dword v3, off, s[0:3], s32 offset:200 ; 4-byte Folded Reload
-; VI-NEXT: s_waitcnt vmcnt(2)
-; VI-NEXT: v_lshlrev_b16_e32 v2, 8, v2
+; VI-NEXT: v_or_b32_sdwa v1, v2, v1 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; VI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:220 ; 4-byte Folded Reload
; VI-NEXT: s_waitcnt vmcnt(0)
-; VI-NEXT: v_or_b32_sdwa v2, v3, v2 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
-; VI-NEXT: v_or_b32_sdwa v1, v1, v2 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_0 src1_sel:DWORD
+; VI-NEXT: v_lshlrev_b16_e32 v2, 8, v2
+; VI-NEXT: v_or_b32_sdwa v2, v11, v2 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; VI-NEXT: v_or_b32_sdwa v1, v2, v1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_0 src1_sel:DWORD
; VI-NEXT: v_add_u32_e32 v2, vcc, 40, v0
; VI-NEXT: buffer_store_dword v1, v2, s[0:3], 0 offen
-; VI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:244 ; 4-byte Folded Reload
-; VI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:376 ; 4-byte Folded Reload
+; VI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:212 ; 4-byte Folded Reload
+; VI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:272 ; 4-byte Folded Reload
; VI-NEXT: s_waitcnt vmcnt(1)
; VI-NEXT: v_lshlrev_b16_e32 v1, 8, v1
; VI-NEXT: s_waitcnt vmcnt(0)
-; VI-NEXT: v_or_b32_sdwa v1, v2, v1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
-; VI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:288 ; 4-byte Folded Reload
-; VI-NEXT: s_waitcnt vmcnt(0)
; VI-NEXT: v_lshlrev_b16_e32 v2, 8, v2
-; VI-NEXT: v_or_b32_sdwa v2, v63, v2 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; VI-NEXT: v_or_b32_sdwa v1, v12, v1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; VI-NEXT: v_or_b32_sdwa v2, v57, v2 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
; VI-NEXT: v_or_b32_sdwa v1, v1, v2 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_0 src1_sel:DWORD
; VI-NEXT: v_add_u32_e32 v2, vcc, 44, v0
; VI-NEXT: buffer_store_dword v1, v2, s[0:3], 0 offen
-; VI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:240 ; 4-byte Folded Reload
-; VI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:416 ; 4-byte Folded Reload
-; VI-NEXT: s_waitcnt vmcnt(1)
+; VI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:112 ; 4-byte Folded Reload
+; VI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:116 ; 4-byte Folded Reload
+; VI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:160 ; 4-byte Folded Reload
+; VI-NEXT: s_waitcnt vmcnt(2)
; VI-NEXT: v_lshlrev_b16_e32 v1, 8, v1
; VI-NEXT: s_waitcnt vmcnt(0)
-; VI-NEXT: v_or_b32_sdwa v1, v2, v1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
-; VI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:84 ; 4-byte Folded Reload
-; VI-NEXT: buffer_load_dword v3, off, s[0:3], s32 offset:88 ; 4-byte Folded Reload
-; VI-NEXT: buffer_load_dword v3, off, s[0:3], s32 offset:196 ; 4-byte Folded Reload
-; VI-NEXT: s_waitcnt vmcnt(2)
-; VI-NEXT: v_lshlrev_b16_e32 v2, 8, v2
+; VI-NEXT: v_or_b32_sdwa v1, v2, v1 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; VI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:204 ; 4-byte Folded Reload
; VI-NEXT: s_waitcnt vmcnt(0)
-; VI-NEXT: v_or_b32_sdwa v2, v3, v2 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
-; VI-NEXT: v_or_b32_sdwa v1, v1, v2 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_0 src1_sel:DWORD
+; VI-NEXT: v_lshlrev_b16_e32 v2, 8, v2
+; VI-NEXT: v_or_b32_sdwa v2, v13, v2 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; VI-NEXT: v_or_b32_sdwa v1, v2, v1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_0 src1_sel:DWORD
; VI-NEXT: v_add_u32_e32 v2, vcc, 48, v0
; VI-NEXT: buffer_store_dword v1, v2, s[0:3], 0 offen
-; VI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:236 ; 4-byte Folded Reload
-; VI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:400 ; 4-byte Folded Reload
+; VI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:200 ; 4-byte Folded Reload
+; VI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:252 ; 4-byte Folded Reload
; VI-NEXT: s_waitcnt vmcnt(1)
; VI-NEXT: v_lshlrev_b16_e32 v1, 8, v1
; VI-NEXT: s_waitcnt vmcnt(0)
-; VI-NEXT: v_or_b32_sdwa v1, v2, v1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
-; VI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:280 ; 4-byte Folded Reload
-; VI-NEXT: s_waitcnt vmcnt(0)
; VI-NEXT: v_lshlrev_b16_e32 v2, 8, v2
-; VI-NEXT: v_or_b32_sdwa v2, v33, v2 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; VI-NEXT: v_or_b32_sdwa v1, v14, v1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; VI-NEXT: v_or_b32_sdwa v2, v59, v2 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
; VI-NEXT: v_or_b32_sdwa v1, v1, v2 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_0 src1_sel:DWORD
; VI-NEXT: v_add_u32_e32 v2, vcc, 52, v0
; VI-NEXT: buffer_store_dword v1, v2, s[0:3], 0 offen
-; VI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:232 ; 4-byte Folded Reload
-; VI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:76 ; 4-byte Folded Reload
-; VI-NEXT: buffer_load_dword v3, off, s[0:3], s32 offset:80 ; 4-byte Folded Reload
-; VI-NEXT: buffer_load_dword v3, off, s[0:3], s32 offset:192 ; 4-byte Folded Reload
+; VI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:104 ; 4-byte Folded Reload
+; VI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:108 ; 4-byte Folded Reload
+; VI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:164 ; 4-byte Folded Reload
; VI-NEXT: s_waitcnt vmcnt(2)
-; VI-NEXT: v_lshlrev_b16_e32 v2, 8, v2
; VI-NEXT: v_lshlrev_b16_e32 v1, 8, v1
-; VI-NEXT: v_or_b32_sdwa v1, v32, v1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
; VI-NEXT: s_waitcnt vmcnt(0)
-; VI-NEXT: v_or_b32_sdwa v2, v3, v2 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
-; VI-NEXT: v_or_b32_sdwa v1, v1, v2 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_0 src1_sel:DWORD
+; VI-NEXT: v_or_b32_sdwa v1, v2, v1 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; VI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:196 ; 4-byte Folded Reload
+; VI-NEXT: s_waitcnt vmcnt(0)
+; VI-NEXT: v_lshlrev_b16_e32 v2, 8, v2
+; VI-NEXT: v_or_b32_sdwa v2, v15, v2 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; VI-NEXT: v_or_b32_sdwa v1, v2, v1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_0 src1_sel:DWORD
; VI-NEXT: v_add_u32_e32 v2, vcc, 56, v0
; VI-NEXT: buffer_store_dword v1, v2, s[0:3], 0 offen
-; VI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:228 ; 4-byte Folded Reload
-; VI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:420 ; 4-byte Folded Reload
-; VI-NEXT: s_waitcnt vmcnt(1)
+; VI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:192 ; 4-byte Folded Reload
+; VI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:224 ; 4-byte Folded Reload
+; VI-NEXT: buffer_load_dword v3, off, s[0:3], s32 offset:152 ; 4-byte Folded Reload
+; VI-NEXT: s_waitcnt vmcnt(2)
; VI-NEXT: v_lshlrev_b16_e32 v1, 8, v1
-; VI-NEXT: s_waitcnt vmcnt(0)
-; VI-NEXT: v_or_b32_sdwa v1, v2, v1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
-; VI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:268 ; 4-byte Folded Reload
-; VI-NEXT: s_waitcnt vmcnt(0)
+; VI-NEXT: s_waitcnt vmcnt(1)
; VI-NEXT: v_lshlrev_b16_e32 v2, 8, v2
-; VI-NEXT: v_or_b32_sdwa v2, v60, v2 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; VI-NEXT: v_or_b32_sdwa v1, v16, v1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; VI-NEXT: s_waitcnt vmcnt(0)
+; VI-NEXT: v_or_b32_sdwa v2, v3, v2 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
; VI-NEXT: v_or_b32_sdwa v1, v1, v2 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_0 src1_sel:DWORD
; VI-NEXT: v_add_u32_e32 v2, vcc, 60, v0
; VI-NEXT: buffer_store_dword v1, v2, s[0:3], 0 offen
-; VI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:372 ; 4-byte Folded Reload
-; VI-NEXT: buffer_load_dword v3, off, s[0:3], s32 offset:188 ; 4-byte Folded Reload
-; VI-NEXT: v_lshlrev_b16_e32 v1, 8, v55
-; VI-NEXT: s_waitcnt vmcnt(1)
-; VI-NEXT: v_or_b32_sdwa v1, v2, v1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
-; VI-NEXT: v_lshlrev_b16_e32 v2, 8, v45
+; VI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:148 ; 4-byte Folded Reload
+; VI-NEXT: v_lshlrev_b16_e32 v1, 8, v42
; VI-NEXT: s_waitcnt vmcnt(0)
-; VI-NEXT: v_or_b32_sdwa v2, v3, v2 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
-; VI-NEXT: v_or_b32_sdwa v1, v1, v2 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_0 src1_sel:DWORD
+; VI-NEXT: v_or_b32_sdwa v1, v2, v1 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; VI-NEXT: v_lshlrev_b16_e32 v2, 8, v58
+; VI-NEXT: v_or_b32_sdwa v2, v17, v2 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; VI-NEXT: v_or_b32_sdwa v1, v2, v1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_0 src1_sel:DWORD
; VI-NEXT: v_add_u32_e32 v2, vcc, 64, v0
; VI-NEXT: buffer_store_dword v1, v2, s[0:3], 0 offen
-; VI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:336 ; 4-byte Folded Reload
-; VI-NEXT: v_lshlrev_b16_e32 v1, 8, v62
-; VI-NEXT: s_waitcnt vmcnt(0)
-; VI-NEXT: v_or_b32_sdwa v1, v2, v1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
-; VI-NEXT: v_lshlrev_b16_e32 v2, 8, v31
-; VI-NEXT: v_or_b32_sdwa v2, v52, v2 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; VI-NEXT: v_lshlrev_b16_e32 v1, 8, v37
+; VI-NEXT: v_lshlrev_b16_e32 v2, 8, v55
+; VI-NEXT: v_or_b32_sdwa v1, v18, v1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; VI-NEXT: v_or_b32_sdwa v2, v63, v2 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
; VI-NEXT: v_or_b32_sdwa v1, v1, v2 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_0 src1_sel:DWORD
; VI-NEXT: v_add_u32_e32 v2, vcc, 0x44, v0
; VI-NEXT: buffer_store_dword v1, v2, s[0:3], 0 offen
-; VI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:364 ; 4-byte Folded Reload
-; VI-NEXT: v_lshlrev_b16_e32 v1, 8, v38
+; VI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:144 ; 4-byte Folded Reload
+; VI-NEXT: v_lshlrev_b16_e32 v1, 8, v39
; VI-NEXT: s_waitcnt vmcnt(0)
-; VI-NEXT: v_or_b32_sdwa v1, v2, v1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
-; VI-NEXT: v_lshlrev_b16_e32 v2, 8, v44
-; VI-NEXT: v_or_b32_sdwa v2, v51, v2 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
-; VI-NEXT: v_or_b32_sdwa v1, v1, v2 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_0 src1_sel:DWORD
+; VI-NEXT: v_or_b32_sdwa v1, v2, v1 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; VI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:360 ; 4-byte Folded Reload
+; VI-NEXT: s_waitcnt vmcnt(0)
+; VI-NEXT: v_lshlrev_b16_e32 v2, 8, v2
+; VI-NEXT: v_or_b32_sdwa v2, v19, v2 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; VI-NEXT: v_or_b32_sdwa v1, v2, v1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_0 src1_sel:DWORD
; VI-NEXT: v_add_u32_e32 v2, vcc, 0x48, v0
; VI-NEXT: buffer_store_dword v1, v2, s[0:3], 0 offen
-; VI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:332 ; 4-byte Folded Reload
-; VI-NEXT: v_lshlrev_b16_e32 v1, 8, v42
+; VI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:352 ; 4-byte Folded Reload
+; VI-NEXT: buffer_load_dword v3, off, s[0:3], s32 offset:84 ; 4-byte Folded Reload
+; VI-NEXT: v_lshlrev_b16_e32 v2, 8, v48
+; VI-NEXT: s_waitcnt vmcnt(1)
+; VI-NEXT: v_lshlrev_b16_e32 v1, 8, v1
+; VI-NEXT: v_or_b32_sdwa v1, v20, v1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
; VI-NEXT: s_waitcnt vmcnt(0)
-; VI-NEXT: v_or_b32_sdwa v1, v2, v1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
-; VI-NEXT: v_lshlrev_b16_e32 v2, 8, v61
-; VI-NEXT: v_or_b32_sdwa v2, v53, v2 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; VI-NEXT: v_or_b32_sdwa v2, v3, v2 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
; VI-NEXT: v_or_b32_sdwa v1, v1, v2 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_0 src1_sel:DWORD
; VI-NEXT: v_add_u32_e32 v2, vcc, 0x4c, v0
; VI-NEXT: buffer_store_dword v1, v2, s[0:3], 0 offen
-; VI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:504 ; 4-byte Folded Reload
-; VI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:360 ; 4-byte Folded Reload
+; VI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:336 ; 4-byte Folded Reload
+; VI-NEXT: buffer_load_dword v3, off, s[0:3], s32 offset:136 ; 4-byte Folded Reload
+; VI-NEXT: v_lshlrev_b16_e32 v2, 8, v61
; VI-NEXT: s_waitcnt vmcnt(1)
; VI-NEXT: v_lshlrev_b16_e32 v1, 8, v1
+; VI-NEXT: v_or_b32_sdwa v1, v21, v1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
; VI-NEXT: s_waitcnt vmcnt(0)
-; VI-NEXT: v_or_b32_sdwa v1, v2, v1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
-; VI-NEXT: v_lshlrev_b16_e32 v2, 8, v41
-; VI-NEXT: v_or_b32_sdwa v2, v49, v2 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; VI-NEXT: v_or_b32_sdwa v2, v3, v2 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
; VI-NEXT: v_or_b32_sdwa v1, v1, v2 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_0 src1_sel:DWORD
; VI-NEXT: v_add_u32_e32 v2, vcc, 0x50, v0
; VI-NEXT: buffer_store_dword v1, v2, s[0:3], 0 offen
-; VI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:500 ; 4-byte Folded Reload
-; VI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:324 ; 4-byte Folded Reload
-; VI-NEXT: buffer_load_dword v3, off, s[0:3], s32 offset:176 ; 4-byte Folded Reload
+; VI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:324 ; 4-byte Folded Reload
+; VI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:344 ; 4-byte Folded Reload
+; VI-NEXT: buffer_load_dword v3, off, s[0:3], s32 offset:76 ; 4-byte Folded Reload
; VI-NEXT: s_waitcnt vmcnt(2)
; VI-NEXT: v_lshlrev_b16_e32 v1, 8, v1
; VI-NEXT: s_waitcnt vmcnt(1)
-; VI-NEXT: v_or_b32_sdwa v1, v2, v1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
-; VI-NEXT: v_lshlrev_b16_e32 v2, 8, v58
+; VI-NEXT: v_lshlrev_b16_e32 v2, 8, v2
+; VI-NEXT: v_or_b32_sdwa v1, v22, v1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
; VI-NEXT: s_waitcnt vmcnt(0)
; VI-NEXT: v_or_b32_sdwa v2, v3, v2 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
; VI-NEXT: v_or_b32_sdwa v1, v1, v2 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_0 src1_sel:DWORD
; VI-NEXT: v_add_u32_e32 v2, vcc, 0x54, v0
; VI-NEXT: buffer_store_dword v1, v2, s[0:3], 0 offen
-; VI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:488 ; 4-byte Folded Reload
-; VI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:348 ; 4-byte Folded Reload
-; VI-NEXT: buffer_load_dword v3, off, s[0:3], s32 offset:184 ; 4-byte Folded Reload
-; VI-NEXT: s_waitcnt vmcnt(2)
-; VI-NEXT: v_lshlrev_b16_e32 v1, 8, v1
+; VI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:312 ; 4-byte Folded Reload
+; VI-NEXT: buffer_load_dword v3, off, s[0:3], s32 offset:132 ; 4-byte Folded Reload
+; VI-NEXT: v_lshlrev_b16_e32 v2, 8, v49
; VI-NEXT: s_waitcnt vmcnt(1)
-; VI-NEXT: v_or_b32_sdwa v1, v2, v1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
-; VI-NEXT: v_lshlrev_b16_e32 v2, 8, v54
+; VI-NEXT: v_lshlrev_b16_e32 v1, 8, v1
+; VI-NEXT: v_or_b32_sdwa v1, v23, v1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
; VI-NEXT: s_waitcnt vmcnt(0)
; VI-NEXT: v_or_b32_sdwa v2, v3, v2 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
; VI-NEXT: v_or_b32_sdwa v1, v1, v2 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_0 src1_sel:DWORD
; VI-NEXT: v_add_u32_e32 v2, vcc, 0x58, v0
; VI-NEXT: buffer_store_dword v1, v2, s[0:3], 0 offen
-; VI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:480 ; 4-byte Folded Reload
-; VI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:312 ; 4-byte Folded Reload
-; VI-NEXT: buffer_load_dword v3, off, s[0:3], s32 offset:164 ; 4-byte Folded Reload
-; VI-NEXT: s_waitcnt vmcnt(2)
-; VI-NEXT: v_lshlrev_b16_e32 v1, 8, v1
+; VI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:304 ; 4-byte Folded Reload
+; VI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:332 ; 4-byte Folded Reload
; VI-NEXT: s_waitcnt vmcnt(1)
-; VI-NEXT: v_or_b32_sdwa v1, v2, v1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
-; VI-NEXT: v_lshlrev_b16_e32 v2, 8, v39
+; VI-NEXT: v_lshlrev_b16_e32 v1, 8, v1
; VI-NEXT: s_waitcnt vmcnt(0)
-; VI-NEXT: v_or_b32_sdwa v2, v3, v2 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; VI-NEXT: v_lshlrev_b16_e32 v2, 8, v2
+; VI-NEXT: v_or_b32_sdwa v1, v24, v1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; VI-NEXT: v_or_b32_sdwa v2, v40, v2 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
; VI-NEXT: v_or_b32_sdwa v1, v1, v2 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_0 src1_sel:DWORD
; VI-NEXT: v_add_u32_e32 v2, vcc, 0x5c, v0
; VI-NEXT: buffer_store_dword v1, v2, s[0:3], 0 offen
-; VI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:472 ; 4-byte Folded Reload
-; VI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:340 ; 4-byte Folded Reload
+; VI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:284 ; 4-byte Folded Reload
+; VI-NEXT: buffer_load_dword v3, off, s[0:3], s32 offset:128 ; 4-byte Folded Reload
+; VI-NEXT: v_lshlrev_b16_e32 v2, 8, v38
; VI-NEXT: s_waitcnt vmcnt(1)
; VI-NEXT: v_lshlrev_b16_e32 v1, 8, v1
-; VI-NEXT: s_waitcnt vmcnt(0)
-; VI-NEXT: v_or_b32_sdwa v1, v2, v1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
-; VI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:140 ; 4-byte Folded Reload
-; VI-NEXT: buffer_load_dword v3, off, s[0:3], s32 offset:144 ; 4-byte Folded Reload
-; VI-NEXT: buffer_load_dword v3, off, s[0:3], s32 offset:180 ; 4-byte Folded Reload
-; VI-NEXT: s_waitcnt vmcnt(2)
-; VI-NEXT: v_lshlrev_b16_e32 v2, 8, v2
+; VI-NEXT: v_or_b32_sdwa v1, v25, v1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
; VI-NEXT: s_waitcnt vmcnt(0)
; VI-NEXT: v_or_b32_sdwa v2, v3, v2 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
; VI-NEXT: v_or_b32_sdwa v1, v1, v2 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_0 src1_sel:DWORD
; VI-NEXT: v_add_u32_e32 v2, vcc, 0x60, v0
; VI-NEXT: buffer_store_dword v1, v2, s[0:3], 0 offen
-; VI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:468 ; 4-byte Folded Reload
-; VI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:304 ; 4-byte Folded Reload
-; VI-NEXT: buffer_load_dword v3, off, s[0:3], s32 offset:160 ; 4-byte Folded Reload
-; VI-NEXT: s_waitcnt vmcnt(2)
-; VI-NEXT: v_lshlrev_b16_e32 v1, 8, v1
+; VI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:280 ; 4-byte Folded Reload
+; VI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:308 ; 4-byte Folded Reload
; VI-NEXT: s_waitcnt vmcnt(1)
-; VI-NEXT: v_or_b32_sdwa v1, v2, v1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
-; VI-NEXT: v_lshlrev_b16_e32 v2, 8, v35
+; VI-NEXT: v_lshlrev_b16_e32 v1, 8, v1
; VI-NEXT: s_waitcnt vmcnt(0)
-; VI-NEXT: v_or_b32_sdwa v2, v3, v2 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; VI-NEXT: v_lshlrev_b16_e32 v2, 8, v2
+; VI-NEXT: v_or_b32_sdwa v1, v26, v1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; VI-NEXT: v_or_b32_sdwa v2, v53, v2 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
; VI-NEXT: v_or_b32_sdwa v1, v1, v2 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_0 src1_sel:DWORD
; VI-NEXT: v_add_u32_e32 v2, vcc, 0x64, v0
; VI-NEXT: buffer_store_dword v1, v2, s[0:3], 0 offen
-; VI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:464 ; 4-byte Folded Reload
-; VI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:352 ; 4-byte Folded Reload
-; VI-NEXT: s_waitcnt vmcnt(1)
-; VI-NEXT: v_lshlrev_b16_e32 v1, 8, v1
-; VI-NEXT: s_waitcnt vmcnt(0)
-; VI-NEXT: v_or_b32_sdwa v1, v2, v1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
-; VI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:124 ; 4-byte Folded Reload
-; VI-NEXT: buffer_load_dword v3, off, s[0:3], s32 offset:128 ; 4-byte Folded Reload
-; VI-NEXT: buffer_load_dword v3, off, s[0:3], s32 offset:172 ; 4-byte Folded Reload
-; VI-NEXT: s_waitcnt vmcnt(2)
-; VI-NEXT: v_lshlrev_b16_e32 v2, 8, v2
+; VI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:260 ; 4-byte Folded Reload
+; VI-NEXT: v_lshlrev_b16_e32 v2, 8, v35
+; VI-NEXT: v_or_b32_sdwa v2, v47, v2 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
; VI-NEXT: s_waitcnt vmcnt(0)
-; VI-NEXT: v_or_b32_sdwa v2, v3, v2 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; VI-NEXT: v_lshlrev_b16_e32 v1, 8, v1
+; VI-NEXT: v_or_b32_sdwa v1, v27, v1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
; VI-NEXT: v_or_b32_sdwa v1, v1, v2 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_0 src1_sel:DWORD
; VI-NEXT: v_add_u32_e32 v2, vcc, 0x68, v0
; VI-NEXT: buffer_store_dword v1, v2, s[0:3], 0 offen
-; VI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:460 ; 4-byte Folded Reload
-; VI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:308 ; 4-byte Folded Reload
-; VI-NEXT: buffer_load_dword v3, off, s[0:3], s32 offset:156 ; 4-byte Folded Reload
-; VI-NEXT: s_waitcnt vmcnt(2)
-; VI-NEXT: v_lshlrev_b16_e32 v1, 8, v1
+; VI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:256 ; 4-byte Folded Reload
+; VI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:288 ; 4-byte Folded Reload
; VI-NEXT: s_waitcnt vmcnt(1)
-; VI-NEXT: v_or_b32_sdwa v1, v2, v1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
-; VI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:496 ; 4-byte Folded Reload
+; VI-NEXT: v_lshlrev_b16_e32 v1, 8, v1
; VI-NEXT: s_waitcnt vmcnt(0)
; VI-NEXT: v_lshlrev_b16_e32 v2, 8, v2
-; VI-NEXT: v_or_b32_sdwa v2, v3, v2 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; VI-NEXT: v_or_b32_sdwa v1, v28, v1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; VI-NEXT: v_or_b32_sdwa v2, v50, v2 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
; VI-NEXT: v_or_b32_sdwa v1, v1, v2 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_0 src1_sel:DWORD
; VI-NEXT: v_add_u32_e32 v2, vcc, 0x6c, v0
; VI-NEXT: buffer_store_dword v1, v2, s[0:3], 0 offen
-; VI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:456 ; 4-byte Folded Reload
-; VI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:356 ; 4-byte Folded Reload
-; VI-NEXT: s_waitcnt vmcnt(1)
-; VI-NEXT: v_lshlrev_b16_e32 v1, 8, v1
-; VI-NEXT: s_waitcnt vmcnt(0)
-; VI-NEXT: v_or_b32_sdwa v1, v2, v1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
-; VI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:116 ; 4-byte Folded Reload
-; VI-NEXT: buffer_load_dword v3, off, s[0:3], s32 offset:120 ; 4-byte Folded Reload
-; VI-NEXT: buffer_load_dword v3, off, s[0:3], s32 offset:168 ; 4-byte Folded Reload
+; VI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:236 ; 4-byte Folded Reload
+; VI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:96 ; 4-byte Folded Reload
+; VI-NEXT: buffer_load_dword v3, off, s[0:3], s32 offset:100 ; 4-byte Folded Reload
; VI-NEXT: s_waitcnt vmcnt(2)
+; VI-NEXT: v_lshlrev_b16_e32 v1, 8, v1
+; VI-NEXT: s_waitcnt vmcnt(1)
; VI-NEXT: v_lshlrev_b16_e32 v2, 8, v2
-; VI-NEXT: s_waitcnt vmcnt(0)
-; VI-NEXT: v_or_b32_sdwa v2, v3, v2 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; VI-NEXT: v_or_b32_sdwa v1, v29, v1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; VI-NEXT: v_or_b32_sdwa v2, v43, v2 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
; VI-NEXT: v_or_b32_sdwa v1, v1, v2 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_0 src1_sel:DWORD
; VI-NEXT: v_add_u32_e32 v2, vcc, 0x70, v0
; VI-NEXT: buffer_store_dword v1, v2, s[0:3], 0 offen
-; VI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:448 ; 4-byte Folded Reload
-; VI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:316 ; 4-byte Folded Reload
+; VI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:232 ; 4-byte Folded Reload
+; VI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:268 ; 4-byte Folded Reload
; VI-NEXT: s_waitcnt vmcnt(1)
; VI-NEXT: v_lshlrev_b16_e32 v1, 8, v1
; VI-NEXT: s_waitcnt vmcnt(0)
-; VI-NEXT: v_or_b32_sdwa v1, v2, v1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
-; VI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:484 ; 4-byte Folded Reload
-; VI-NEXT: s_waitcnt vmcnt(0)
; VI-NEXT: v_lshlrev_b16_e32 v2, 8, v2
-; VI-NEXT: v_or_b32_sdwa v2, v46, v2 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; VI-NEXT: v_or_b32_sdwa v1, v30, v1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; VI-NEXT: v_or_b32_sdwa v2, v36, v2 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
; VI-NEXT: v_or_b32_sdwa v1, v1, v2 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_0 src1_sel:DWORD
; VI-NEXT: v_add_u32_e32 v2, vcc, 0x74, v0
; VI-NEXT: buffer_store_dword v1, v2, s[0:3], 0 offen
-; VI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:428 ; 4-byte Folded Reload
-; VI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:328 ; 4-byte Folded Reload
-; VI-NEXT: s_waitcnt vmcnt(1)
+; VI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:216 ; 4-byte Folded Reload
+; VI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:88 ; 4-byte Folded Reload
+; VI-NEXT: buffer_load_dword v3, off, s[0:3], s32 offset:92 ; 4-byte Folded Reload
+; VI-NEXT: buffer_load_dword v3, off, s[0:3], s32 offset:140 ; 4-byte Folded Reload
+; VI-NEXT: s_waitcnt vmcnt(3)
; VI-NEXT: v_lshlrev_b16_e32 v1, 8, v1
-; VI-NEXT: s_waitcnt vmcnt(0)
-; VI-NEXT: v_or_b32_sdwa v1, v2, v1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
-; VI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:108 ; 4-byte Folded Reload
-; VI-NEXT: buffer_load_dword v3, off, s[0:3], s32 offset:112 ; 4-byte Folded Reload
-; VI-NEXT: s_waitcnt vmcnt(1)
+; VI-NEXT: s_waitcnt vmcnt(2)
; VI-NEXT: v_lshlrev_b16_e32 v2, 8, v2
-; VI-NEXT: v_or_b32_sdwa v2, v50, v2 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; VI-NEXT: v_or_b32_sdwa v1, v31, v1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; VI-NEXT: s_waitcnt vmcnt(0)
+; VI-NEXT: v_or_b32_sdwa v2, v3, v2 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
; VI-NEXT: v_or_b32_sdwa v1, v1, v2 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_0 src1_sel:DWORD
; VI-NEXT: v_add_u32_e32 v2, vcc, 0x78, v0
; VI-NEXT: buffer_store_dword v1, v2, s[0:3], 0 offen
-; VI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:424 ; 4-byte Folded Reload
-; VI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:320 ; 4-byte Folded Reload
-; VI-NEXT: buffer_load_dword v3, off, s[0:3], s32 offset:224 ; 4-byte Folded Reload
+; VI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:208 ; 4-byte Folded Reload
+; VI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:244 ; 4-byte Folded Reload
+; VI-NEXT: buffer_load_dword v3, off, s[0:3], s32 offset:80 ; 4-byte Folded Reload
; VI-NEXT: v_add_u32_e32 v0, vcc, 0x7c, v0
; VI-NEXT: s_waitcnt vmcnt(2)
; VI-NEXT: v_lshlrev_b16_e32 v1, 8, v1
; VI-NEXT: s_waitcnt vmcnt(1)
-; VI-NEXT: v_or_b32_sdwa v1, v2, v1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
-; VI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:432 ; 4-byte Folded Reload
-; VI-NEXT: s_waitcnt vmcnt(0)
; VI-NEXT: v_lshlrev_b16_e32 v2, 8, v2
+; VI-NEXT: v_or_b32_sdwa v1, v32, v1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; VI-NEXT: s_waitcnt vmcnt(0)
; VI-NEXT: v_or_b32_sdwa v2, v3, v2 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
; VI-NEXT: v_or_b32_sdwa v1, v1, v2 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_0 src1_sel:DWORD
; VI-NEXT: buffer_store_dword v1, v0, s[0:3], 0 offen
@@ -219846,8 +219804,8 @@ define inreg <128 x i8> @bitcast_v64i16_to_v128i8_scalar(<64 x i16> inreg %a, i3
; SI-NEXT: v_readfirstlane_b32 s82, v1
; SI-NEXT: v_lshlrev_b32_e32 v2, 16, v4
; SI-NEXT: v_lshlrev_b32_e32 v4, 16, v8
-; SI-NEXT: v_lshlrev_b32_e32 v5, 16, v12
-; SI-NEXT: v_lshlrev_b32_e32 v6, 16, v16
+; SI-NEXT: v_lshlrev_b32_e32 v6, 16, v12
+; SI-NEXT: v_lshlrev_b32_e32 v8, 16, v16
; SI-NEXT: v_lshlrev_b32_e32 v9, 16, v20
; SI-NEXT: v_lshlrev_b32_e32 v13, 16, v24
; SI-NEXT: s_waitcnt vmcnt(14)
@@ -219878,9 +219836,7 @@ define inreg <128 x i8> @bitcast_v64i16_to_v128i8_scalar(<64 x i16> inreg %a, i3
; SI-NEXT: s_waitcnt vmcnt(3)
; SI-NEXT: v_lshlrev_b32_e32 v21, 16, v43
; SI-NEXT: v_writelane_b32 v62, s6, 0
-; SI-NEXT: buffer_store_dword v2, off, s[0:3], s32 offset:220 ; 4-byte Folded Spill
-; SI-NEXT: buffer_store_dword v4, off, s[0:3], s32 offset:224 ; 4-byte Folded Spill
-; SI-NEXT: buffer_store_dword v5, off, s[0:3], s32 offset:228 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v2, off, s[0:3], s32 offset:228 ; 4-byte Folded Spill
; SI-NEXT: s_cbranch_scc0 .LBB99_4
; SI-NEXT: ; %bb.1: ; %cmp.false
; SI-NEXT: s_and_b32 s4, s16, 0xffff
@@ -219892,7 +219848,7 @@ define inreg <128 x i8> @bitcast_v64i16_to_v128i8_scalar(<64 x i16> inreg %a, i3
; SI-NEXT: v_mov_b32_e32 v1, s56
; SI-NEXT: s_and_b32 s4, s20, 0xffff
; SI-NEXT: s_lshl_b32 s5, s21, 16
-; SI-NEXT: v_alignbit_b32 v8, s57, v1, 24
+; SI-NEXT: v_alignbit_b32 v5, s57, v1, 24
; SI-NEXT: v_alignbit_b32 v50, s57, v1, 16
; SI-NEXT: v_alignbit_b32 v1, s57, v1, 8
; SI-NEXT: s_or_b32 s46, s4, s5
@@ -219904,39 +219860,43 @@ define inreg <128 x i8> @bitcast_v64i16_to_v128i8_scalar(<64 x i16> inreg %a, i3
; SI-NEXT: v_mov_b32_e32 v1, s46
; SI-NEXT: s_and_b32 s4, s24, 0xffff
; SI-NEXT: s_lshl_b32 s5, s25, 16
-; SI-NEXT: buffer_store_dword v8, off, s[0:3], s32 offset:216 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v5, off, s[0:3], s32 offset:216 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_alignbit_b32 v8, s47, v1, 24
+; SI-NEXT: v_alignbit_b32 v5, s47, v1, 24
; SI-NEXT: s_or_b32 s44, s4, s5
; SI-NEXT: s_and_b32 s4, s26, 0xffff
; SI-NEXT: s_lshl_b32 s5, s27, 16
-; SI-NEXT: buffer_store_dword v8, off, s[0:3], s32 offset:192 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v5, off, s[0:3], s32 offset:192 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_alignbit_b32 v8, s47, v1, 16
+; SI-NEXT: v_alignbit_b32 v5, s47, v1, 16
; SI-NEXT: v_alignbit_b32 v51, s47, v1, 8
; SI-NEXT: s_or_b32 s45, s4, s5
; SI-NEXT: v_mov_b32_e32 v1, s44
+; SI-NEXT: buffer_store_dword v5, off, s[0:3], s32 offset:212 ; 4-byte Folded Spill
+; SI-NEXT: s_waitcnt expcnt(0)
+; SI-NEXT: v_alignbit_b32 v5, s45, v1, 24
; SI-NEXT: s_and_b32 s4, s28, 0xffff
; SI-NEXT: s_lshl_b32 s5, s29, 16
-; SI-NEXT: buffer_store_dword v8, off, s[0:3], s32 offset:212 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v5, off, s[0:3], s32 offset:188 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_alignbit_b32 v8, s45, v1, 24
+; SI-NEXT: v_alignbit_b32 v5, s45, v1, 16
+; SI-NEXT: v_alignbit_b32 v1, s45, v1, 8
; SI-NEXT: s_or_b32 s42, s4, s5
; SI-NEXT: s_and_b32 s4, s82, 0xffff
; SI-NEXT: s_lshl_b32 s5, s81, 16
-; SI-NEXT: buffer_store_dword v8, off, s[0:3], s32 offset:188 ; 4-byte Folded Spill
-; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_alignbit_b32 v8, s45, v1, 16
-; SI-NEXT: v_alignbit_b32 v49, s45, v1, 8
+; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:224 ; 4-byte Folded Spill
; SI-NEXT: s_or_b32 s43, s4, s5
+; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_mov_b32_e32 v1, s42
-; SI-NEXT: buffer_store_dword v8, off, s[0:3], s32 offset:204 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v5, off, s[0:3], s32 offset:204 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_alignbit_b32 v8, s43, v1, 24
-; SI-NEXT: buffer_store_dword v8, off, s[0:3], s32 offset:184 ; 4-byte Folded Spill
+; SI-NEXT: v_alignbit_b32 v5, s43, v1, 24
+; SI-NEXT: buffer_store_dword v5, off, s[0:3], s32 offset:184 ; 4-byte Folded Spill
+; SI-NEXT: s_waitcnt expcnt(0)
+; SI-NEXT: v_alignbit_b32 v5, s43, v1, 16
+; SI-NEXT: v_alignbit_b32 v1, s43, v1, 8
+; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:220 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_alignbit_b32 v8, s43, v1, 16
-; SI-NEXT: v_alignbit_b32 v48, s43, v1, 8
; SI-NEXT: v_and_b32_e32 v1, 0xffff, v3
; SI-NEXT: s_and_b32 s4, s85, 0xffff
; SI-NEXT: s_lshl_b32 s5, s84, 16
@@ -219986,7 +219946,7 @@ define inreg <128 x i8> @bitcast_v64i16_to_v128i8_scalar(<64 x i16> inreg %a, i3
; SI-NEXT: s_or_b32 s10, s4, s5
; SI-NEXT: s_and_b32 s4, s75, 0xffff
; SI-NEXT: s_lshl_b32 s5, s74, 16
-; SI-NEXT: v_or_b32_e32 v12, v1, v5
+; SI-NEXT: v_or_b32_e32 v12, v1, v6
; SI-NEXT: s_or_b32 s9, s4, s5
; SI-NEXT: s_and_b32 s4, s77, 0xffff
; SI-NEXT: s_lshl_b32 s5, s76, 16
@@ -220010,7 +219970,7 @@ define inreg <128 x i8> @bitcast_v64i16_to_v128i8_scalar(<64 x i16> inreg %a, i3
; SI-NEXT: v_and_b32_e32 v1, 0xffff, v15
; SI-NEXT: v_writelane_b32 v62, s4, 1
; SI-NEXT: s_lshr_b32 s4, s10, 8
-; SI-NEXT: v_or_b32_e32 v10, v1, v6
+; SI-NEXT: v_or_b32_e32 v10, v1, v8
; SI-NEXT: v_writelane_b32 v62, s4, 3
; SI-NEXT: s_lshr_b32 s4, s9, 8
; SI-NEXT: v_alignbit_b32 v1, s14, v10, 24
@@ -220031,32 +219991,34 @@ define inreg <128 x i8> @bitcast_v64i16_to_v128i8_scalar(<64 x i16> inreg %a, i3
; SI-NEXT: v_and_b32_e32 v1, 0xffff, v19
; SI-NEXT: v_writelane_b32 v62, s4, 15
; SI-NEXT: s_and_b32 s4, s72, 0xffff
-; SI-NEXT: buffer_store_dword v8, off, s[0:3], s32 offset:196 ; 4-byte Folded Spill
-; SI-NEXT: s_waitcnt expcnt(0)
+; SI-NEXT: v_mov_b32_e32 v28, v8
; SI-NEXT: v_or_b32_e32 v8, v1, v9
; SI-NEXT: v_and_b32_e32 v1, 0xffff, v23
; SI-NEXT: v_writelane_b32 v62, s4, 2
; SI-NEXT: s_and_b32 s4, s74, 0xffff
+; SI-NEXT: buffer_store_dword v5, off, s[0:3], s32 offset:196 ; 4-byte Folded Spill
+; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_or_b32_e32 v5, v1, v13
; SI-NEXT: v_and_b32_e32 v1, 0xffff, v27
; SI-NEXT: v_writelane_b32 v62, s4, 5
; SI-NEXT: s_and_b32 s4, s76, 0xffff
-; SI-NEXT: v_mov_b32_e32 v28, v13
+; SI-NEXT: v_mov_b32_e32 v30, v13
; SI-NEXT: v_or_b32_e32 v13, v1, v17
; SI-NEXT: s_waitcnt vmcnt(14)
; SI-NEXT: v_and_b32_e32 v1, 0xffff, v31
; SI-NEXT: v_writelane_b32 v62, s4, 8
; SI-NEXT: s_and_b32 s4, s78, 0xffff
-; SI-NEXT: v_mov_b32_e32 v26, v9
+; SI-NEXT: v_mov_b32_e32 v29, v9
; SI-NEXT: v_or_b32_e32 v9, v1, v18
; SI-NEXT: v_and_b32_e32 v1, 0xffff, v32
; SI-NEXT: v_writelane_b32 v62, s4, 11
; SI-NEXT: s_and_b32 s4, s88, 0xffff
-; SI-NEXT: v_mov_b32_e32 v25, v6
+; SI-NEXT: v_mov_b32_e32 v26, v6
; SI-NEXT: v_or_b32_e32 v6, v1, v20
; SI-NEXT: v_and_b32_e32 v1, 0xffff, v33
; SI-NEXT: v_writelane_b32 v62, s4, 14
; SI-NEXT: s_bfe_u32 s4, s74, 0x80008
+; SI-NEXT: v_mov_b32_e32 v25, v4
; SI-NEXT: v_or_b32_e32 v4, v1, v21
; SI-NEXT: v_and_b32_e32 v1, 0xffff, v34
; SI-NEXT: v_writelane_b32 v62, s4, 4
@@ -220068,12 +220030,12 @@ define inreg <128 x i8> @bitcast_v64i16_to_v128i8_scalar(<64 x i16> inreg %a, i3
; SI-NEXT: v_or_b32_e32 v1, v1, v24
; SI-NEXT: v_writelane_b32 v62, s4, 10
; SI-NEXT: s_bfe_u32 s4, s88, 0x80008
-; SI-NEXT: v_mov_b32_e32 v29, v17
-; SI-NEXT: v_mov_b32_e32 v30, v18
-; SI-NEXT: v_mov_b32_e32 v36, v20
-; SI-NEXT: v_mov_b32_e32 v37, v21
-; SI-NEXT: v_mov_b32_e32 v38, v22
-; SI-NEXT: v_mov_b32_e32 v39, v24
+; SI-NEXT: v_mov_b32_e32 v36, v17
+; SI-NEXT: v_mov_b32_e32 v37, v18
+; SI-NEXT: v_mov_b32_e32 v38, v20
+; SI-NEXT: v_mov_b32_e32 v39, v21
+; SI-NEXT: v_mov_b32_e32 v48, v22
+; SI-NEXT: v_mov_b32_e32 v49, v24
; SI-NEXT: s_lshr_b32 s68, s57, 8
; SI-NEXT: s_lshr_b32 s65, s47, 8
; SI-NEXT: s_lshr_b32 s54, s45, 8
@@ -220134,9 +220096,7 @@ define inreg <128 x i8> @bitcast_v64i16_to_v128i8_scalar(<64 x i16> inreg %a, i3
; SI-NEXT: v_alignbit_b32 v54, s6, v1, 8
; SI-NEXT: s_cbranch_execnz .LBB99_3
; SI-NEXT: .LBB99_2: ; %cmp.true
-; SI-NEXT: buffer_load_dword v12, off, s[0:3], s32 offset:228 ; 4-byte Folded Reload
-; SI-NEXT: buffer_load_dword v14, off, s[0:3], s32 offset:224 ; 4-byte Folded Reload
-; SI-NEXT: buffer_load_dword v16, off, s[0:3], s32 offset:220 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v16, off, s[0:3], s32 offset:228 ; 4-byte Folded Reload
; SI-NEXT: s_add_i32 s18, s18, 3
; SI-NEXT: s_and_b32 s4, s18, 0xffff
; SI-NEXT: s_lshl_b32 s5, s88, 16
@@ -220248,50 +220208,54 @@ define inreg <128 x i8> @bitcast_v64i16_to_v128i8_scalar(<64 x i16> inreg %a, i3
; SI-NEXT: buffer_store_dword v22, off, s[0:3], s32 offset:216 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_alignbit_b32 v22, s47, v21, 24
-; SI-NEXT: s_lshr_b32 s4, s11, 8
-; SI-NEXT: v_add_i32_e32 v3, vcc, 3, v3
; SI-NEXT: buffer_store_dword v22, off, s[0:3], s32 offset:192 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_alignbit_b32 v22, s47, v21, 16
; SI-NEXT: v_alignbit_b32 v51, s47, v21, 8
; SI-NEXT: v_mov_b32_e32 v21, s44
+; SI-NEXT: buffer_store_dword v22, off, s[0:3], s32 offset:212 ; 4-byte Folded Spill
+; SI-NEXT: s_waitcnt expcnt(0)
+; SI-NEXT: v_alignbit_b32 v22, s45, v21, 24
+; SI-NEXT: s_lshr_b32 s4, s11, 8
+; SI-NEXT: v_add_i32_e32 v3, vcc, 3, v3
+; SI-NEXT: buffer_store_dword v22, off, s[0:3], s32 offset:188 ; 4-byte Folded Spill
+; SI-NEXT: s_waitcnt expcnt(0)
+; SI-NEXT: v_alignbit_b32 v22, s45, v21, 16
+; SI-NEXT: v_alignbit_b32 v21, s45, v21, 8
; SI-NEXT: v_writelane_b32 v62, s4, 1
; SI-NEXT: s_lshr_b32 s4, s10, 16
-; SI-NEXT: s_waitcnt vmcnt(10)
+; SI-NEXT: s_waitcnt vmcnt(8)
; SI-NEXT: v_add_i32_e32 v5, vcc, 3, v32
; SI-NEXT: v_and_b32_e32 v3, 0xffff, v3
-; SI-NEXT: buffer_store_dword v22, off, s[0:3], s32 offset:212 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v21, off, s[0:3], s32 offset:224 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_alignbit_b32 v22, s45, v21, 24
+; SI-NEXT: v_mov_b32_e32 v21, s42
; SI-NEXT: v_writelane_b32 v62, s4, 2
; SI-NEXT: s_lshr_b32 s4, s10, 8
; SI-NEXT: v_and_b32_e32 v5, 0xffff, v5
; SI-NEXT: v_add_i32_e32 v7, vcc, 3, v7
-; SI-NEXT: s_waitcnt vmcnt(4)
+; SI-NEXT: s_waitcnt vmcnt(6)
; SI-NEXT: v_or_b32_e32 v3, v16, v3
-; SI-NEXT: buffer_store_dword v22, off, s[0:3], s32 offset:188 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v22, off, s[0:3], s32 offset:204 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_alignbit_b32 v22, s45, v21, 16
-; SI-NEXT: v_alignbit_b32 v49, s45, v21, 8
-; SI-NEXT: v_mov_b32_e32 v21, s42
+; SI-NEXT: v_alignbit_b32 v22, s43, v21, 24
; SI-NEXT: v_writelane_b32 v62, s4, 3
; SI-NEXT: s_lshr_b32 s4, s9, 24
-; SI-NEXT: v_or_b32_e32 v5, v36, v5
+; SI-NEXT: v_or_b32_e32 v5, v38, v5
; SI-NEXT: v_and_b32_e32 v7, 0xffff, v7
; SI-NEXT: v_add_i32_e32 v16, vcc, 0x30000, v3
; SI-NEXT: v_mov_b32_e32 v3, s41
-; SI-NEXT: buffer_store_dword v22, off, s[0:3], s32 offset:204 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v22, off, s[0:3], s32 offset:184 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_alignbit_b32 v22, s43, v21, 24
+; SI-NEXT: v_alignbit_b32 v22, s43, v21, 16
+; SI-NEXT: v_alignbit_b32 v21, s43, v21, 8
; SI-NEXT: v_writelane_b32 v62, s4, 4
; SI-NEXT: s_lshr_b32 s4, s9, 16
; SI-NEXT: v_add_i32_e32 v6, vcc, 0x30000, v5
; SI-NEXT: v_add_i32_e32 v5, vcc, 3, v31
-; SI-NEXT: v_or_b32_e32 v7, v14, v7
-; SI-NEXT: buffer_store_dword v22, off, s[0:3], s32 offset:184 ; 4-byte Folded Spill
+; SI-NEXT: v_or_b32_e32 v7, v25, v7
+; SI-NEXT: buffer_store_dword v21, off, s[0:3], s32 offset:220 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_alignbit_b32 v22, s43, v21, 16
-; SI-NEXT: v_alignbit_b32 v48, s43, v21, 8
; SI-NEXT: v_alignbit_b32 v21, v3, v16, 24
; SI-NEXT: v_writelane_b32 v62, s4, 5
; SI-NEXT: s_lshr_b32 s4, s9, 8
@@ -220305,7 +220269,7 @@ define inreg <128 x i8> @bitcast_v64i16_to_v128i8_scalar(<64 x i16> inreg %a, i3
; SI-NEXT: v_alignbit_b32 v3, v3, v16, 8
; SI-NEXT: v_writelane_b32 v62, s4, 6
; SI-NEXT: s_lshr_b32 s4, s8, 24
-; SI-NEXT: v_or_b32_e32 v5, v30, v5
+; SI-NEXT: v_or_b32_e32 v5, v37, v5
; SI-NEXT: v_and_b32_e32 v11, 0xffff, v11
; SI-NEXT: buffer_store_dword v3, off, s[0:3], s32 offset:200 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
@@ -220314,7 +220278,7 @@ define inreg <128 x i8> @bitcast_v64i16_to_v128i8_scalar(<64 x i16> inreg %a, i3
; SI-NEXT: s_lshr_b32 s4, s8, 16
; SI-NEXT: v_add_i32_e32 v9, vcc, 0x30000, v5
; SI-NEXT: v_add_i32_e32 v5, vcc, 3, v27
-; SI-NEXT: v_or_b32_e32 v11, v12, v11
+; SI-NEXT: v_or_b32_e32 v11, v26, v11
; SI-NEXT: buffer_store_dword v3, off, s[0:3], s32 offset:152 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_alignbit_b32 v3, v7, v14, 16
@@ -220329,7 +220293,7 @@ define inreg <128 x i8> @bitcast_v64i16_to_v128i8_scalar(<64 x i16> inreg %a, i3
; SI-NEXT: v_alignbit_b32 v3, v7, v14, 8
; SI-NEXT: v_writelane_b32 v62, s4, 9
; SI-NEXT: s_lshr_b32 s4, s7, 24
-; SI-NEXT: v_or_b32_e32 v5, v29, v5
+; SI-NEXT: v_or_b32_e32 v5, v36, v5
; SI-NEXT: v_and_b32_e32 v10, 0xffff, v10
; SI-NEXT: buffer_store_dword v3, off, s[0:3], s32 offset:180 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
@@ -220342,7 +220306,7 @@ define inreg <128 x i8> @bitcast_v64i16_to_v128i8_scalar(<64 x i16> inreg %a, i3
; SI-NEXT: v_add_i32_e32 v13, vcc, 0x30000, v5
; SI-NEXT: v_add_i32_e32 v5, vcc, 3, v23
; SI-NEXT: v_add_i32_e32 v8, vcc, 3, v19
-; SI-NEXT: v_or_b32_e32 v10, v25, v10
+; SI-NEXT: v_or_b32_e32 v10, v28, v10
; SI-NEXT: buffer_store_dword v3, off, s[0:3], s32 offset:144 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_alignbit_b32 v3, v11, v12, 16
@@ -220360,11 +220324,11 @@ define inreg <128 x i8> @bitcast_v64i16_to_v128i8_scalar(<64 x i16> inreg %a, i3
; SI-NEXT: v_alignbit_b32 v3, v11, v12, 8
; SI-NEXT: v_writelane_b32 v62, s4, 12
; SI-NEXT: s_lshr_b32 s4, s6, 24
-; SI-NEXT: v_or_b32_e32 v1, v39, v1
-; SI-NEXT: v_or_b32_e32 v2, v38, v2
-; SI-NEXT: v_or_b32_e32 v4, v37, v4
-; SI-NEXT: v_or_b32_e32 v5, v28, v5
-; SI-NEXT: v_or_b32_e32 v8, v26, v8
+; SI-NEXT: v_or_b32_e32 v1, v49, v1
+; SI-NEXT: v_or_b32_e32 v2, v48, v2
+; SI-NEXT: v_or_b32_e32 v4, v39, v4
+; SI-NEXT: v_or_b32_e32 v5, v30, v5
+; SI-NEXT: v_or_b32_e32 v8, v29, v8
; SI-NEXT: buffer_store_dword v3, off, s[0:3], s32 offset:172 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_alignbit_b32 v3, v15, v10, 24
@@ -220532,65 +220496,64 @@ define inreg <128 x i8> @bitcast_v64i16_to_v128i8_scalar(<64 x i16> inreg %a, i3
; SI-NEXT: v_add_i32_e32 v3, vcc, 12, v0
; SI-NEXT: v_mov_b32_e32 v7, s4
; SI-NEXT: buffer_store_dword v7, v3, s[0:3], 0 offen
+; SI-NEXT: buffer_load_dword v3, off, s[0:3], s32 offset:224 ; 4-byte Folded Reload
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: buffer_load_dword v7, off, s[0:3], s32 offset:204 ; 4-byte Folded Reload
; SI-NEXT: buffer_load_dword v11, off, s[0:3], s32 offset:188 ; 4-byte Folded Reload
; SI-NEXT: s_and_b32 s4, s44, 0xff
-; SI-NEXT: v_lshlrev_b32_e32 v3, 8, v49
+; SI-NEXT: s_waitcnt vmcnt(2)
+; SI-NEXT: v_lshlrev_b32_e32 v3, 8, v3
; SI-NEXT: v_or_b32_e32 v3, s4, v3
; SI-NEXT: s_and_b32 s4, s45, 0xff
-; SI-NEXT: s_or_b32 s4, s4, s5
-; SI-NEXT: s_and_b32 s5, s66, 0xff
-; SI-NEXT: s_lshl_b32 s5, s5, 16
-; SI-NEXT: v_and_b32_e32 v3, 0xffff, v3
-; SI-NEXT: s_and_b32 s4, s4, 0xffff
-; SI-NEXT: s_or_b32 s5, s16, s5
-; SI-NEXT: s_or_b32 s4, s4, s5
-; SI-NEXT: s_lshl_b32 s5, s51, 8
-; SI-NEXT: s_lshl_b32 s16, s67, 24
-; SI-NEXT: v_readlane_b32 s67, v63, 19
-; SI-NEXT: v_readlane_b32 s66, v63, 18
-; SI-NEXT: v_readlane_b32 s51, v63, 11
; SI-NEXT: s_waitcnt vmcnt(1)
; SI-NEXT: v_and_b32_e32 v7, 0xff, v7
+; SI-NEXT: s_or_b32 s4, s4, s5
+; SI-NEXT: s_and_b32 s5, s66, 0xff
; SI-NEXT: v_lshlrev_b32_e32 v7, 16, v7
; SI-NEXT: s_waitcnt vmcnt(0)
; SI-NEXT: v_lshlrev_b32_e32 v11, 24, v11
+; SI-NEXT: s_lshl_b32 s5, s5, 16
+; SI-NEXT: v_and_b32_e32 v3, 0xffff, v3
; SI-NEXT: v_or_b32_e32 v7, v11, v7
+; SI-NEXT: s_and_b32 s4, s4, 0xffff
+; SI-NEXT: s_or_b32 s5, s16, s5
; SI-NEXT: v_or_b32_e32 v3, v3, v7
; SI-NEXT: v_add_i32_e32 v7, vcc, 16, v0
+; SI-NEXT: s_or_b32 s4, s4, s5
; SI-NEXT: buffer_store_dword v3, v7, s[0:3], 0 offen
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_add_i32_e32 v3, vcc, 20, v0
; SI-NEXT: v_mov_b32_e32 v7, s4
; SI-NEXT: buffer_store_dword v7, v3, s[0:3], 0 offen
+; SI-NEXT: buffer_load_dword v3, off, s[0:3], s32 offset:220 ; 4-byte Folded Reload
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: buffer_load_dword v7, off, s[0:3], s32 offset:196 ; 4-byte Folded Reload
; SI-NEXT: buffer_load_dword v11, off, s[0:3], s32 offset:184 ; 4-byte Folded Reload
; SI-NEXT: s_and_b32 s4, s42, 0xff
-; SI-NEXT: v_lshlrev_b32_e32 v3, 8, v48
+; SI-NEXT: s_lshl_b32 s5, s51, 8
+; SI-NEXT: s_lshl_b32 s16, s67, 24
+; SI-NEXT: v_readlane_b32 s67, v63, 19
+; SI-NEXT: v_readlane_b32 s66, v63, 18
+; SI-NEXT: v_readlane_b32 s51, v63, 11
+; SI-NEXT: s_waitcnt vmcnt(2)
+; SI-NEXT: v_lshlrev_b32_e32 v3, 8, v3
; SI-NEXT: v_or_b32_e32 v3, s4, v3
; SI-NEXT: s_and_b32 s4, s43, 0xff
-; SI-NEXT: s_or_b32 s4, s4, s5
-; SI-NEXT: s_and_b32 s5, s55, 0xff
-; SI-NEXT: s_lshl_b32 s5, s5, 16
-; SI-NEXT: v_and_b32_e32 v3, 0xffff, v3
-; SI-NEXT: s_and_b32 s4, s4, 0xffff
-; SI-NEXT: s_or_b32 s5, s16, s5
-; SI-NEXT: s_or_b32 s4, s4, s5
-; SI-NEXT: s_lshl_b32 s5, s48, 8
-; SI-NEXT: s_lshl_b32 s16, s64, 24
-; SI-NEXT: v_readlane_b32 s64, v63, 16
-; SI-NEXT: v_readlane_b32 s55, v63, 15
-; SI-NEXT: v_readlane_b32 s48, v63, 8
; SI-NEXT: s_waitcnt vmcnt(1)
; SI-NEXT: v_and_b32_e32 v7, 0xff, v7
+; SI-NEXT: s_or_b32 s4, s4, s5
+; SI-NEXT: s_and_b32 s5, s55, 0xff
; SI-NEXT: v_lshlrev_b32_e32 v7, 16, v7
; SI-NEXT: s_waitcnt vmcnt(0)
; SI-NEXT: v_lshlrev_b32_e32 v11, 24, v11
+; SI-NEXT: s_lshl_b32 s5, s5, 16
+; SI-NEXT: v_and_b32_e32 v3, 0xffff, v3
; SI-NEXT: v_or_b32_e32 v7, v11, v7
+; SI-NEXT: s_and_b32 s4, s4, 0xffff
+; SI-NEXT: s_or_b32 s5, s16, s5
; SI-NEXT: v_or_b32_e32 v3, v3, v7
; SI-NEXT: v_add_i32_e32 v7, vcc, 24, v0
+; SI-NEXT: s_or_b32 s4, s4, s5
; SI-NEXT: buffer_store_dword v3, v7, s[0:3], 0 offen
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_add_i32_e32 v3, vcc, 28, v0
@@ -220600,16 +220563,21 @@ define inreg <128 x i8> @bitcast_v64i16_to_v128i8_scalar(<64 x i16> inreg %a, i3
; SI-NEXT: buffer_load_dword v7, off, s[0:3], s32 offset:200 ; 4-byte Folded Reload
; SI-NEXT: v_and_b32_e32 v3, 0xff, v16
; SI-NEXT: s_and_b32 s4, s41, 0xff
+; SI-NEXT: s_lshl_b32 s5, s48, 8
; SI-NEXT: s_or_b32 s4, s4, s5
; SI-NEXT: s_and_b32 s5, s52, 0xff
; SI-NEXT: s_lshl_b32 s5, s5, 16
+; SI-NEXT: s_lshl_b32 s16, s64, 24
; SI-NEXT: s_and_b32 s4, s4, 0xffff
; SI-NEXT: s_or_b32 s5, s16, s5
; SI-NEXT: s_or_b32 s4, s4, s5
; SI-NEXT: s_lshl_b32 s5, s37, 8
; SI-NEXT: s_lshl_b32 s16, s53, 24
+; SI-NEXT: v_readlane_b32 s64, v63, 16
+; SI-NEXT: v_readlane_b32 s55, v63, 15
; SI-NEXT: v_readlane_b32 s53, v63, 13
; SI-NEXT: v_readlane_b32 s52, v63, 12
+; SI-NEXT: v_readlane_b32 s48, v63, 8
; SI-NEXT: v_readlane_b32 s37, v63, 5
; SI-NEXT: buffer_load_dword v11, off, s[0:3], s32 offset:168 ; 4-byte Folded Reload
; SI-NEXT: s_waitcnt vmcnt(1)
@@ -220969,55 +220937,62 @@ define inreg <128 x i8> @bitcast_v64i16_to_v128i8_scalar(<64 x i16> inreg %a, i3
; SI-NEXT: ; implicit-def: $sgpr6
; SI-NEXT: ; kill: killed $sgpr6
; SI-NEXT: ; implicit-def: $vcc_lo
-; SI-NEXT: v_mov_b32_e32 v39, v24
+; SI-NEXT: v_mov_b32_e32 v49, v24
; SI-NEXT: ; implicit-def: $vgpr1
; SI-NEXT: ; kill: killed $vgpr1
; SI-NEXT: ; implicit-def: $sgpr6
; SI-NEXT: ; kill: killed $sgpr6
-; SI-NEXT: v_mov_b32_e32 v38, v22
+; SI-NEXT: v_mov_b32_e32 v48, v22
; SI-NEXT: ; implicit-def: $vgpr1
; SI-NEXT: ; kill: killed $vgpr1
; SI-NEXT: ; implicit-def: $sgpr6
; SI-NEXT: ; kill: killed $sgpr6
-; SI-NEXT: v_mov_b32_e32 v37, v21
+; SI-NEXT: v_mov_b32_e32 v39, v21
; SI-NEXT: ; implicit-def: $vgpr1
; SI-NEXT: ; kill: killed $vgpr1
; SI-NEXT: ; implicit-def: $sgpr6
; SI-NEXT: ; kill: killed $sgpr6
-; SI-NEXT: v_mov_b32_e32 v36, v20
+; SI-NEXT: v_mov_b32_e32 v38, v20
; SI-NEXT: ; implicit-def: $vgpr1
; SI-NEXT: ; kill: killed $vgpr1
; SI-NEXT: ; implicit-def: $sgpr6
; SI-NEXT: ; kill: killed $sgpr6
-; SI-NEXT: v_mov_b32_e32 v30, v18
+; SI-NEXT: v_mov_b32_e32 v37, v18
; SI-NEXT: ; implicit-def: $vgpr1
; SI-NEXT: ; kill: killed $vgpr1
; SI-NEXT: ; implicit-def: $sgpr6
; SI-NEXT: ; kill: killed $sgpr6
-; SI-NEXT: v_mov_b32_e32 v29, v17
+; SI-NEXT: v_mov_b32_e32 v36, v17
; SI-NEXT: ; implicit-def: $vgpr1
; SI-NEXT: ; kill: killed $vgpr1
; SI-NEXT: ; implicit-def: $sgpr6
; SI-NEXT: ; kill: killed $sgpr6
-; SI-NEXT: v_mov_b32_e32 v28, v13
+; SI-NEXT: v_mov_b32_e32 v30, v13
; SI-NEXT: ; implicit-def: $vgpr1
; SI-NEXT: ; kill: killed $vgpr1
; SI-NEXT: ; implicit-def: $sgpr6
; SI-NEXT: ; kill: killed $sgpr6
-; SI-NEXT: v_mov_b32_e32 v26, v9
+; SI-NEXT: v_mov_b32_e32 v29, v9
; SI-NEXT: ; implicit-def: $vgpr1
; SI-NEXT: ; kill: killed $vgpr1
; SI-NEXT: ; implicit-def: $sgpr6
; SI-NEXT: ; kill: killed $sgpr6
-; SI-NEXT: v_mov_b32_e32 v25, v6
+; SI-NEXT: v_mov_b32_e32 v28, v8
; SI-NEXT: ; implicit-def: $vgpr1
; SI-NEXT: ; kill: killed $vgpr1
; SI-NEXT: ; implicit-def: $sgpr6
; SI-NEXT: ; kill: killed $sgpr6
+; SI-NEXT: v_mov_b32_e32 v26, v6
; SI-NEXT: ; implicit-def: $vgpr1
; SI-NEXT: ; kill: killed $vgpr1
; SI-NEXT: ; implicit-def: $sgpr6
+; SI-NEXT: v_mov_b32_e32 v25, v4
+; SI-NEXT: ; implicit-def: $vgpr1
+; SI-NEXT: ; kill: killed $vgpr1
+; SI-NEXT: ; implicit-def: $vgpr1
+; SI-NEXT: ; kill: killed $vgpr1
; SI-NEXT: ; kill: killed $sgpr6
+; SI-NEXT: ; implicit-def: $sgpr6
; SI-NEXT: ; kill: killed $vcc_lo
; SI-NEXT: ; implicit-def: $vcc_lo
; SI-NEXT: ; implicit-def: $sgpr56
@@ -221033,13 +221008,11 @@ define inreg <128 x i8> @bitcast_v64i16_to_v128i8_scalar(<64 x i16> inreg %a, i3
; SI-NEXT: ; implicit-def: $sgpr69
; SI-NEXT: ; implicit-def: $sgpr80
; SI-NEXT: ; implicit-def: $sgpr44
-; SI-NEXT: ; implicit-def: $vgpr49
; SI-NEXT: ; implicit-def: $sgpr45
; SI-NEXT: ; implicit-def: $sgpr54
; SI-NEXT: ; implicit-def: $sgpr66
; SI-NEXT: ; implicit-def: $sgpr70
; SI-NEXT: ; implicit-def: $sgpr42
-; SI-NEXT: ; implicit-def: $vgpr48
; SI-NEXT: ; implicit-def: $sgpr43
; SI-NEXT: ; implicit-def: $sgpr51
; SI-NEXT: ; implicit-def: $sgpr55
@@ -221088,12 +221061,12 @@ define inreg <128 x i8> @bitcast_v64i16_to_v128i8_scalar(<64 x i16> inreg %a, i3
; SI-NEXT: ; implicit-def: $sgpr9
; SI-NEXT: ; implicit-def: $sgpr8
; SI-NEXT: ; implicit-def: $sgpr7
-; SI-NEXT: ; kill: killed $vcc_lo
-; SI-NEXT: ; implicit-def: $vcc_lo
+; SI-NEXT: ; kill: killed $sgpr6
+; SI-NEXT: ; implicit-def: $sgpr6
; SI-NEXT: ; implicit-def: $vgpr1
; SI-NEXT: ; kill: killed $vgpr1
-; SI-NEXT: ; implicit-def: $sgpr6
-; SI-NEXT: ; kill: killed $sgpr6
+; SI-NEXT: ; kill: killed $vcc_lo
+; SI-NEXT: ; implicit-def: $vcc_lo
; SI-NEXT: ; kill: killed $vcc_lo
; SI-NEXT: ; implicit-def: $vgpr13
; SI-NEXT: ; implicit-def: $vgpr24
@@ -221120,7 +221093,6 @@ define inreg <128 x i8> @bitcast_v64i16_to_v128i8_scalar(<64 x i16> inreg %a, i3
; SI-NEXT: ; implicit-def: $vgpr52
; SI-NEXT: ; implicit-def: $vgpr1
; SI-NEXT: ; kill: killed $vgpr1
-; SI-NEXT: ; implicit-def: $sgpr6
; SI-NEXT: ; implicit-def: $vgpr1
; SI-NEXT: ; kill: killed $vgpr1
; SI-NEXT: ; implicit-def: $vgpr1
@@ -222717,7 +222689,7 @@ define inreg <128 x i8> @bitcast_v64i16_to_v128i8_scalar(<64 x i16> inreg %a, i3
; GFX9-NEXT: v_lshrrev_b32_e32 v15, 24, v36
; GFX9-NEXT: buffer_store_dword v19, off, s[0:3], s32 offset:252 ; 4-byte Folded Spill
; GFX9-NEXT: v_lshrrev_b32_e32 v19, 8, v13
-; GFX9-NEXT: buffer_store_dword v15, off, s[0:3], s32 offset:376 ; 4-byte Folded Spill
+; GFX9-NEXT: buffer_store_dword v15, off, s[0:3], s32 offset:380 ; 4-byte Folded Spill
; GFX9-NEXT: v_lshrrev_b32_e32 v15, 16, v36
; GFX9-NEXT: buffer_store_dword v19, off, s[0:3], s32 offset:224 ; 4-byte Folded Spill
; GFX9-NEXT: v_lshrrev_b32_e32 v19, 24, v22
@@ -222725,7 +222697,7 @@ define inreg <128 x i8> @bitcast_v64i16_to_v128i8_scalar(<64 x i16> inreg %a, i3
; GFX9-NEXT: v_lshrrev_b32_e32 v15, 16, v35
; GFX9-NEXT: buffer_store_dword v19, off, s[0:3], s32 offset:260 ; 4-byte Folded Spill
; GFX9-NEXT: v_lshrrev_b32_e32 v19, 16, v22
-; GFX9-NEXT: buffer_store_dword v15, off, s[0:3], s32 offset:380 ; 4-byte Folded Spill
+; GFX9-NEXT: buffer_store_dword v15, off, s[0:3], s32 offset:376 ; 4-byte Folded Spill
; GFX9-NEXT: v_lshrrev_b32_e32 v15, 16, v38
; GFX9-NEXT: buffer_store_dword v19, off, s[0:3], s32 offset:256 ; 4-byte Folded Spill
; GFX9-NEXT: v_lshrrev_b32_e32 v19, 8, v22
@@ -223024,7 +222996,7 @@ define inreg <128 x i8> @bitcast_v64i16_to_v128i8_scalar(<64 x i16> inreg %a, i3
; GFX9-NEXT: v_readlane_b32 s4, v62, 22
; GFX9-NEXT: v_mov_b32_e32 v60, s4
; GFX9-NEXT: v_readlane_b32 s4, v62, 23
-; GFX9-NEXT: buffer_store_dword v17, off, s[0:3], s32 offset:380 ; 4-byte Folded Spill
+; GFX9-NEXT: buffer_store_dword v17, off, s[0:3], s32 offset:376 ; 4-byte Folded Spill
; GFX9-NEXT: v_mov_b32_e32 v17, s4
; GFX9-NEXT: v_readlane_b32 s4, v62, 24
; GFX9-NEXT: buffer_store_dword v17, off, s[0:3], s32 offset:372 ; 4-byte Folded Spill
@@ -223032,7 +223004,7 @@ define inreg <128 x i8> @bitcast_v64i16_to_v128i8_scalar(<64 x i16> inreg %a, i3
; GFX9-NEXT: v_readlane_b32 s4, v62, 25
; GFX9-NEXT: v_mov_b32_e32 v23, s4
; GFX9-NEXT: v_readlane_b32 s4, v62, 26
-; GFX9-NEXT: buffer_store_dword v17, off, s[0:3], s32 offset:376 ; 4-byte Folded Spill
+; GFX9-NEXT: buffer_store_dword v17, off, s[0:3], s32 offset:380 ; 4-byte Folded Spill
; GFX9-NEXT: v_mov_b32_e32 v17, s4
; GFX9-NEXT: v_readlane_b32 s4, v62, 27
; GFX9-NEXT: v_mov_b32_e32 v59, s4
@@ -223302,14 +223274,14 @@ define inreg <128 x i8> @bitcast_v64i16_to_v128i8_scalar(<64 x i16> inreg %a, i3
; GFX9-NEXT: buffer_load_dword v15, off, s[0:3], s32 offset:284 ; 4-byte Folded Reload
; GFX9-NEXT: s_nop 0
; GFX9-NEXT: buffer_load_dword v16, off, s[0:3], s32 offset:288 ; 4-byte Folded Reload
-; GFX9-NEXT: buffer_load_dword v16, off, s[0:3], s32 offset:380 ; 4-byte Folded Reload
+; GFX9-NEXT: buffer_load_dword v16, off, s[0:3], s32 offset:376 ; 4-byte Folded Reload
; GFX9-NEXT: s_waitcnt vmcnt(2)
; GFX9-NEXT: v_lshlrev_b32_e32 v15, 8, v15
; GFX9-NEXT: s_waitcnt vmcnt(0)
; GFX9-NEXT: v_or_b32_sdwa v15, v16, v15 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
; GFX9-NEXT: v_or_b32_sdwa v15, v35, v15 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_0 src1_sel:DWORD
; GFX9-NEXT: buffer_store_dword v15, v0, s[0:3], 0 offen offset:16
-; GFX9-NEXT: buffer_load_dword v15, off, s[0:3], s32 offset:376 ; 4-byte Folded Reload
+; GFX9-NEXT: buffer_load_dword v15, off, s[0:3], s32 offset:380 ; 4-byte Folded Reload
; GFX9-NEXT: s_nop 0
; GFX9-NEXT: buffer_load_dword v16, off, s[0:3], s32 offset:372 ; 4-byte Folded Reload
; GFX9-NEXT: s_waitcnt vmcnt(1)
@@ -229368,8 +229340,8 @@ define inreg <64 x half> @bitcast_v64bf16_to_v64f16_scalar(<64 x bfloat> inreg %
; SI-NEXT: buffer_load_dword v42, off, s[0:3], s32 offset:72
; SI-NEXT: buffer_load_dword v43, off, s[0:3], s32 offset:76
; SI-NEXT: v_mul_f32_e32 v45, 1.0, v1
-; SI-NEXT: v_mul_f32_e32 v1, 1.0, v30
-; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:232 ; 4-byte Folded Spill
+; SI-NEXT: v_mul_f32_e32 v1, 1.0, v29
+; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:236 ; 4-byte Folded Spill
; SI-NEXT: v_mul_f32_e32 v2, 1.0, v2
; SI-NEXT: v_mul_f32_e32 v3, 1.0, v3
; SI-NEXT: v_mul_f32_e32 v4, 1.0, v4
@@ -229393,403 +229365,431 @@ define inreg <64 x half> @bitcast_v64bf16_to_v64f16_scalar(<64 x bfloat> inreg %
; SI-NEXT: s_waitcnt expcnt(1)
; SI-NEXT: v_mul_f32_e32 v63, 1.0, v18
; SI-NEXT: v_mul_f32_e32 v19, 1.0, v19
-; SI-NEXT: v_mul_f32_e32 v20, 1.0, v20
-; SI-NEXT: v_mul_f32_e32 v44, 1.0, v21
+; SI-NEXT: v_mul_f32_e32 v44, 1.0, v20
+; SI-NEXT: v_mul_f32_e32 v21, 1.0, v21
; SI-NEXT: v_mul_f32_e32 v46, 1.0, v22
; SI-NEXT: v_mul_f32_e32 v47, 1.0, v23
; SI-NEXT: v_mul_f32_e32 v56, 1.0, v24
; SI-NEXT: v_mul_f32_e32 v57, 1.0, v25
-; SI-NEXT: v_mul_f32_e32 v58, 1.0, v26
-; SI-NEXT: v_mul_f32_e32 v14, 1.0, v27
-; SI-NEXT: v_mul_f32_e32 v29, 1.0, v29
+; SI-NEXT: v_mul_f32_e32 v26, 1.0, v26
+; SI-NEXT: v_mul_f32_e32 v58, 1.0, v27
+; SI-NEXT: v_mul_f32_e32 v14, 1.0, v28
; SI-NEXT: v_mul_f32_e64 v15, 1.0, s17
; SI-NEXT: v_mul_f32_e64 v16, 1.0, s18
; SI-NEXT: v_mul_f32_e64 v17, 1.0, s19
; SI-NEXT: v_mul_f32_e64 v18, 1.0, s20
-; SI-NEXT: v_mul_f32_e64 v21, 1.0, s21
+; SI-NEXT: v_mul_f32_e64 v20, 1.0, s21
; SI-NEXT: v_mul_f32_e64 v22, 1.0, s22
; SI-NEXT: v_mul_f32_e64 v23, 1.0, s23
; SI-NEXT: v_mul_f32_e64 v24, 1.0, s24
; SI-NEXT: v_mul_f32_e64 v25, 1.0, s25
-; SI-NEXT: v_mul_f32_e64 v26, 1.0, s26
-; SI-NEXT: v_mul_f32_e64 v27, 1.0, s27
-; SI-NEXT: v_mul_f32_e64 v30, 1.0, s29
+; SI-NEXT: v_mul_f32_e64 v27, 1.0, s26
+; SI-NEXT: v_mul_f32_e64 v28, 1.0, s27
+; SI-NEXT: v_mul_f32_e64 v29, 1.0, s28
; SI-NEXT: s_waitcnt vmcnt(14)
; SI-NEXT: v_cmp_ne_u32_e32 vcc, 0, v31
-; SI-NEXT: v_mul_f32_e32 v31, 1.0, v28
+; SI-NEXT: v_mul_f32_e32 v31, 1.0, v30
; SI-NEXT: v_mul_f32_e32 v32, 1.0, v32
; SI-NEXT: v_mul_f32_e32 v33, 1.0, v33
; SI-NEXT: s_and_b64 s[4:5], vcc, exec
+; SI-NEXT: s_waitcnt expcnt(0)
+; SI-NEXT: v_mul_f32_e32 v1, 1.0, v36
+; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:228 ; 4-byte Folded Spill
; SI-NEXT: v_mul_f32_e32 v34, 1.0, v34
+; SI-NEXT: s_waitcnt vmcnt(14) expcnt(0)
+; SI-NEXT: v_mul_f32_e32 v1, 1.0, v39
+; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:240 ; 4-byte Folded Spill
; SI-NEXT: v_mul_f32_e32 v35, 1.0, v35
-; SI-NEXT: v_mul_f32_e32 v36, 1.0, v36
-; SI-NEXT: v_mul_f32_e32 v37, 1.0, v37
-; SI-NEXT: v_mul_f32_e32 v38, 1.0, v38
+; SI-NEXT: s_waitcnt vmcnt(12) expcnt(0)
+; SI-NEXT: v_mul_f32_e32 v1, 1.0, v50
+; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:244 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt vmcnt(11) expcnt(0)
-; SI-NEXT: v_mul_f32_e32 v1, 1.0, v49
-; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:236 ; 4-byte Folded Spill
-; SI-NEXT: s_waitcnt vmcnt(10) expcnt(0)
-; SI-NEXT: v_mul_f32_e32 v1, 1.0, v51
+; SI-NEXT: v_mul_f32_e32 v1, 1.0, v52
; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:248 ; 4-byte Folded Spill
-; SI-NEXT: v_mul_f32_e32 v39, 1.0, v39
-; SI-NEXT: v_mul_f32_e32 v48, 1.0, v48
-; SI-NEXT: s_waitcnt vmcnt(7) expcnt(0)
-; SI-NEXT: v_mul_f32_e32 v1, 1.0, v55
-; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:240 ; 4-byte Folded Spill
-; SI-NEXT: s_waitcnt vmcnt(6) expcnt(0)
-; SI-NEXT: v_mul_f32_e32 v1, 1.0, v41
-; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:244 ; 4-byte Folded Spill
-; SI-NEXT: s_waitcnt vmcnt(5) expcnt(0)
-; SI-NEXT: v_mul_f32_e32 v1, 1.0, v43
-; SI-NEXT: v_mul_f32_e32 v49, 1.0, v50
-; SI-NEXT: v_mul_f32_e32 v51, 1.0, v52
-; SI-NEXT: v_mul_f32_e32 v53, 1.0, v53
-; SI-NEXT: v_mul_f32_e32 v50, 1.0, v54
-; SI-NEXT: v_mul_f32_e32 v52, 1.0, v40
-; SI-NEXT: v_mul_f32_e32 v54, 1.0, v42
+; SI-NEXT: s_waitcnt vmcnt(10) expcnt(0)
+; SI-NEXT: v_mul_f32_e32 v1, 1.0, v54
; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:252 ; 4-byte Folded Spill
+; SI-NEXT: v_mul_f32_e32 v37, 1.0, v37
+; SI-NEXT: s_waitcnt vmcnt(8) expcnt(0)
+; SI-NEXT: v_mul_f32_e32 v1, 1.0, v41
+; SI-NEXT: v_mul_f32_e32 v38, 1.0, v38
+; SI-NEXT: v_mul_f32_e32 v48, 1.0, v48
+; SI-NEXT: v_mul_f32_e32 v39, 1.0, v49
+; SI-NEXT: v_mul_f32_e32 v51, 1.0, v51
+; SI-NEXT: v_mul_f32_e32 v52, 1.0, v53
+; SI-NEXT: v_mul_f32_e32 v54, 1.0, v55
+; SI-NEXT: v_mul_f32_e32 v53, 1.0, v40
+; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:256 ; 4-byte Folded Spill
+; SI-NEXT: s_waitcnt vmcnt(8)
+; SI-NEXT: v_mul_f32_e32 v41, 1.0, v42
+; SI-NEXT: s_waitcnt vmcnt(7)
+; SI-NEXT: v_mul_f32_e32 v50, 1.0, v43
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_mul_f32_e64 v1, 1.0, s16
-; SI-NEXT: v_mul_f32_e64 v28, 1.0, s28
-; SI-NEXT: buffer_store_dword v45, off, s[0:3], s32 offset:256 ; 4-byte Folded Spill
-; SI-NEXT: buffer_store_dword v2, off, s[0:3], s32 offset:260 ; 4-byte Folded Spill
-; SI-NEXT: buffer_store_dword v3, off, s[0:3], s32 offset:264 ; 4-byte Folded Spill
-; SI-NEXT: buffer_store_dword v4, off, s[0:3], s32 offset:268 ; 4-byte Folded Spill
-; SI-NEXT: buffer_store_dword v5, off, s[0:3], s32 offset:272 ; 4-byte Folded Spill
-; SI-NEXT: buffer_store_dword v6, off, s[0:3], s32 offset:276 ; 4-byte Folded Spill
-; SI-NEXT: buffer_store_dword v7, off, s[0:3], s32 offset:280 ; 4-byte Folded Spill
-; SI-NEXT: buffer_store_dword v8, off, s[0:3], s32 offset:284 ; 4-byte Folded Spill
-; SI-NEXT: buffer_store_dword v9, off, s[0:3], s32 offset:288 ; 4-byte Folded Spill
-; SI-NEXT: buffer_store_dword v10, off, s[0:3], s32 offset:292 ; 4-byte Folded Spill
-; SI-NEXT: buffer_store_dword v11, off, s[0:3], s32 offset:296 ; 4-byte Folded Spill
-; SI-NEXT: buffer_store_dword v12, off, s[0:3], s32 offset:300 ; 4-byte Folded Spill
-; SI-NEXT: buffer_store_dword v13, off, s[0:3], s32 offset:304 ; 4-byte Folded Spill
-; SI-NEXT: buffer_store_dword v59, off, s[0:3], s32 offset:308 ; 4-byte Folded Spill
-; SI-NEXT: buffer_store_dword v60, off, s[0:3], s32 offset:312 ; 4-byte Folded Spill
-; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:316 ; 4-byte Folded Spill
-; SI-NEXT: buffer_store_dword v61, off, s[0:3], s32 offset:320 ; 4-byte Folded Spill
-; SI-NEXT: buffer_store_dword v15, off, s[0:3], s32 offset:324 ; 4-byte Folded Spill
-; SI-NEXT: buffer_store_dword v62, off, s[0:3], s32 offset:328 ; 4-byte Folded Spill
-; SI-NEXT: buffer_store_dword v16, off, s[0:3], s32 offset:332 ; 4-byte Folded Spill
-; SI-NEXT: buffer_store_dword v63, off, s[0:3], s32 offset:336 ; 4-byte Folded Spill
+; SI-NEXT: v_mul_f32_e64 v30, 1.0, s29
+; SI-NEXT: buffer_store_dword v45, off, s[0:3], s32 offset:260 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v2, off, s[0:3], s32 offset:264 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v3, off, s[0:3], s32 offset:268 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v4, off, s[0:3], s32 offset:272 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v5, off, s[0:3], s32 offset:276 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v6, off, s[0:3], s32 offset:280 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v7, off, s[0:3], s32 offset:284 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v8, off, s[0:3], s32 offset:288 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v9, off, s[0:3], s32 offset:292 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v10, off, s[0:3], s32 offset:296 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v11, off, s[0:3], s32 offset:300 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v12, off, s[0:3], s32 offset:304 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v13, off, s[0:3], s32 offset:308 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v59, off, s[0:3], s32 offset:312 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v60, off, s[0:3], s32 offset:316 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:320 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v61, off, s[0:3], s32 offset:324 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v15, off, s[0:3], s32 offset:328 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v62, off, s[0:3], s32 offset:332 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v16, off, s[0:3], s32 offset:336 ; 4-byte Folded Spill
; SI-NEXT: buffer_store_dword v17, off, s[0:3], s32 offset:340 ; 4-byte Folded Spill
-; SI-NEXT: buffer_store_dword v19, off, s[0:3], s32 offset:344 ; 4-byte Folded Spill
-; SI-NEXT: buffer_store_dword v18, off, s[0:3], s32 offset:348 ; 4-byte Folded Spill
-; SI-NEXT: buffer_store_dword v20, off, s[0:3], s32 offset:352 ; 4-byte Folded Spill
-; SI-NEXT: buffer_store_dword v21, off, s[0:3], s32 offset:356 ; 4-byte Folded Spill
-; SI-NEXT: buffer_store_dword v22, off, s[0:3], s32 offset:360 ; 4-byte Folded Spill
-; SI-NEXT: buffer_store_dword v23, off, s[0:3], s32 offset:364 ; 4-byte Folded Spill
-; SI-NEXT: buffer_store_dword v24, off, s[0:3], s32 offset:368 ; 4-byte Folded Spill
-; SI-NEXT: buffer_store_dword v25, off, s[0:3], s32 offset:372 ; 4-byte Folded Spill
-; SI-NEXT: buffer_store_dword v26, off, s[0:3], s32 offset:376 ; 4-byte Folded Spill
-; SI-NEXT: buffer_store_dword v27, off, s[0:3], s32 offset:380 ; 4-byte Folded Spill
-; SI-NEXT: buffer_store_dword v28, off, s[0:3], s32 offset:384 ; 4-byte Folded Spill
-; SI-NEXT: buffer_store_dword v30, off, s[0:3], s32 offset:388 ; 4-byte Folded Spill
-; SI-NEXT: buffer_store_dword v44, off, s[0:3], s32 offset:392 ; 4-byte Folded Spill
-; SI-NEXT: buffer_store_dword v32, off, s[0:3], s32 offset:396 ; 4-byte Folded Spill
-; SI-NEXT: buffer_store_dword v33, off, s[0:3], s32 offset:400 ; 4-byte Folded Spill
-; SI-NEXT: buffer_store_dword v34, off, s[0:3], s32 offset:404 ; 4-byte Folded Spill
-; SI-NEXT: buffer_store_dword v35, off, s[0:3], s32 offset:408 ; 4-byte Folded Spill
-; SI-NEXT: buffer_store_dword v36, off, s[0:3], s32 offset:412 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v18, off, s[0:3], s32 offset:344 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v20, off, s[0:3], s32 offset:348 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v22, off, s[0:3], s32 offset:352 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v23, off, s[0:3], s32 offset:356 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v24, off, s[0:3], s32 offset:360 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v25, off, s[0:3], s32 offset:364 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v27, off, s[0:3], s32 offset:368 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v28, off, s[0:3], s32 offset:372 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v29, off, s[0:3], s32 offset:376 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v30, off, s[0:3], s32 offset:380 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v63, off, s[0:3], s32 offset:384 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v19, off, s[0:3], s32 offset:388 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v47, off, s[0:3], s32 offset:392 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v31, off, s[0:3], s32 offset:396 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v32, off, s[0:3], s32 offset:400 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v33, off, s[0:3], s32 offset:404 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v34, off, s[0:3], s32 offset:408 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v35, off, s[0:3], s32 offset:412 ; 4-byte Folded Spill
; SI-NEXT: buffer_store_dword v37, off, s[0:3], s32 offset:416 ; 4-byte Folded Spill
; SI-NEXT: buffer_store_dword v38, off, s[0:3], s32 offset:420 ; 4-byte Folded Spill
-; SI-NEXT: buffer_store_dword v39, off, s[0:3], s32 offset:424 ; 4-byte Folded Spill
-; SI-NEXT: buffer_store_dword v49, off, s[0:3], s32 offset:428 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v48, off, s[0:3], s32 offset:424 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v39, off, s[0:3], s32 offset:428 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v51, off, s[0:3], s32 offset:432 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v52, off, s[0:3], s32 offset:436 ; 4-byte Folded Spill
; SI-NEXT: s_cbranch_scc0 .LBB101_4
; SI-NEXT: ; %bb.1: ; %cmp.false
; SI-NEXT: v_lshrrev_b32_e32 v1, 16, v1
-; SI-NEXT: v_cvt_f32_f16_e32 v1, v1
-; SI-NEXT: v_mov_b32_e32 v42, v51
-; SI-NEXT: v_mov_b32_e32 v55, v50
-; SI-NEXT: v_mov_b32_e32 v40, v52
-; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:212 ; 4-byte Folded Spill
-; SI-NEXT: s_waitcnt expcnt(0)
+; SI-NEXT: v_cvt_f32_f16_e32 v55, v1
; SI-NEXT: v_lshrrev_b32_e32 v1, 16, v15
; SI-NEXT: v_cvt_f32_f16_e32 v1, v1
-; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:228 ; 4-byte Folded Spill
-; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_lshrrev_b32_e32 v1, 16, v16
-; SI-NEXT: v_cvt_f32_f16_e32 v1, v1
+; SI-NEXT: buffer_load_dword v49, off, s[0:3], s32 offset:240 ; 4-byte Folded Reload
; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:224 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
+; SI-NEXT: v_lshrrev_b32_e32 v1, 16, v16
+; SI-NEXT: v_cvt_f32_f16_e32 v40, v1
; SI-NEXT: v_lshrrev_b32_e32 v1, 16, v17
-; SI-NEXT: v_cvt_f32_f16_e32 v41, v1
-; SI-NEXT: v_lshrrev_b32_e32 v1, 16, v18
; SI-NEXT: v_cvt_f32_f16_e32 v15, v1
-; SI-NEXT: v_lshrrev_b32_e32 v1, 16, v21
+; SI-NEXT: v_lshrrev_b32_e32 v1, 16, v18
; SI-NEXT: v_cvt_f32_f16_e32 v16, v1
+; SI-NEXT: v_lshrrev_b32_e32 v1, 16, v20
+; SI-NEXT: v_cvt_f32_f16_e32 v42, v1
; SI-NEXT: v_lshrrev_b32_e32 v1, 16, v22
; SI-NEXT: v_cvt_f32_f16_e32 v1, v1
+; SI-NEXT: v_mov_b32_e32 v20, v44
+; SI-NEXT: v_mov_b32_e32 v22, v21
; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:216 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_lshrrev_b32_e32 v1, 16, v23
; SI-NEXT: v_cvt_f32_f16_e32 v18, v1
; SI-NEXT: v_lshrrev_b32_e32 v1, 16, v24
; SI-NEXT: v_cvt_f32_f16_e32 v1, v1
-; SI-NEXT: v_mov_b32_e32 v24, v47
; SI-NEXT: v_mov_b32_e32 v23, v46
-; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:208 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:212 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_lshrrev_b32_e32 v1, 16, v25
; SI-NEXT: v_cvt_f32_f16_e32 v1, v1
; SI-NEXT: v_mov_b32_e32 v25, v56
; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:220 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_lshrrev_b32_e32 v1, 16, v26
+; SI-NEXT: v_lshrrev_b32_e32 v1, 16, v27
; SI-NEXT: v_cvt_f32_f16_e32 v1, v1
-; SI-NEXT: v_mov_b32_e32 v26, v57
-; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:204 ; 4-byte Folded Spill
+; SI-NEXT: v_mov_b32_e32 v27, v57
+; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:208 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_lshrrev_b32_e32 v1, 16, v27
-; SI-NEXT: v_cvt_f32_f16_e32 v22, v1
; SI-NEXT: v_lshrrev_b32_e32 v1, 16, v28
-; SI-NEXT: v_cvt_f32_f16_e32 v43, v1
+; SI-NEXT: v_cvt_f32_f16_e32 v24, v1
+; SI-NEXT: v_lshrrev_b32_e32 v1, 16, v29
+; SI-NEXT: v_cvt_f32_f16_e32 v1, v1
+; SI-NEXT: v_mov_b32_e32 v28, v26
+; SI-NEXT: v_mov_b32_e32 v29, v58
+; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:204 ; 4-byte Folded Spill
+; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_lshrrev_b32_e32 v1, 16, v30
-; SI-NEXT: v_cvt_f32_f16_e32 v30, v1
+; SI-NEXT: v_cvt_f32_f16_e32 v17, v1
; SI-NEXT: v_lshrrev_b32_e32 v1, 16, v45
-; SI-NEXT: v_cvt_f32_f16_e32 v28, v1
+; SI-NEXT: v_cvt_f32_f16_e32 v1, v1
+; SI-NEXT: v_mov_b32_e32 v30, v14
+; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:152 ; 4-byte Folded Spill
+; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_lshrrev_b32_e32 v1, 16, v2
; SI-NEXT: v_cvt_f32_f16_e32 v45, v1
; SI-NEXT: v_lshrrev_b32_e32 v1, 16, v3
-; SI-NEXT: v_cvt_f32_f16_e32 v2, v1
-; SI-NEXT: v_lshrrev_b32_e32 v1, 16, v4
-; SI-NEXT: v_cvt_f32_f16_e32 v17, v1
-; SI-NEXT: v_lshrrev_b32_e32 v1, 16, v5
; SI-NEXT: v_cvt_f32_f16_e32 v1, v1
-; SI-NEXT: v_lshrrev_b32_e32 v5, 16, v32
-; SI-NEXT: v_cvt_f32_f16_e32 v5, v5
-; SI-NEXT: v_lshrrev_b32_e32 v4, 16, v29
-; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:160 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:156 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
+; SI-NEXT: v_lshrrev_b32_e32 v1, 16, v4
+; SI-NEXT: v_cvt_f32_f16_e32 v43, v1
+; SI-NEXT: v_lshrrev_b32_e32 v1, 16, v5
+; SI-NEXT: v_cvt_f32_f16_e32 v2, v1
; SI-NEXT: v_lshrrev_b32_e32 v1, 16, v6
-; SI-NEXT: v_cvt_f32_f16_e32 v21, v1
+; SI-NEXT: v_cvt_f32_f16_e32 v36, v1
; SI-NEXT: v_lshrrev_b32_e32 v1, 16, v7
; SI-NEXT: v_cvt_f32_f16_e32 v1, v1
-; SI-NEXT: v_lshrrev_b32_e32 v6, 16, v34
-; SI-NEXT: v_lshrrev_b32_e32 v7, 16, v36
-; SI-NEXT: v_cvt_f32_f16_e32 v6, v6
-; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:152 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:160 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_lshrrev_b32_e32 v1, 16, v8
-; SI-NEXT: v_cvt_f32_f16_e32 v8, v1
+; SI-NEXT: v_cvt_f32_f16_e32 v7, v1
; SI-NEXT: v_lshrrev_b32_e32 v1, 16, v9
; SI-NEXT: v_cvt_f32_f16_e32 v1, v1
-; SI-NEXT: buffer_store_dword v5, off, s[0:3], s32 offset:176 ; 4-byte Folded Spill
-; SI-NEXT: v_mov_b32_e32 v36, v8
-; SI-NEXT: v_lshrrev_b32_e32 v8, 16, v38
-; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:156 ; 4-byte Folded Spill
+; SI-NEXT: v_lshrrev_b32_e32 v9, 16, v48
+; SI-NEXT: v_cvt_f32_f16_e32 v9, v9
+; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:164 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_lshrrev_b32_e32 v1, 16, v10
-; SI-NEXT: v_cvt_f32_f16_e32 v9, v1
+; SI-NEXT: v_cvt_f32_f16_e32 v6, v1
; SI-NEXT: v_lshrrev_b32_e32 v1, 16, v11
-; SI-NEXT: v_cvt_f32_f16_e32 v3, v1
+; SI-NEXT: v_cvt_f32_f16_e32 v4, v1
; SI-NEXT: v_lshrrev_b32_e32 v1, 16, v12
-; SI-NEXT: v_cvt_f32_f16_e32 v10, v1
+; SI-NEXT: v_cvt_f32_f16_e32 v8, v1
; SI-NEXT: v_lshrrev_b32_e32 v1, 16, v13
; SI-NEXT: v_cvt_f32_f16_e32 v1, v1
-; SI-NEXT: v_cvt_f32_f16_e32 v8, v8
-; SI-NEXT: v_lshrrev_b32_e32 v11, 16, v51
-; SI-NEXT: v_lshrrev_b32_e32 v5, 16, v33
-; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:164 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v9, off, s[0:3], s32 offset:192 ; 4-byte Folded Spill
+; SI-NEXT: s_waitcnt expcnt(0)
+; SI-NEXT: v_lshrrev_b32_e32 v9, 16, v39
+; SI-NEXT: buffer_load_dword v39, off, s[0:3], s32 offset:244 ; 4-byte Folded Reload
+; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:168 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_lshrrev_b32_e32 v1, 16, v59
; SI-NEXT: v_cvt_f32_f16_e32 v59, v1
; SI-NEXT: v_lshrrev_b32_e32 v1, 16, v60
-; SI-NEXT: v_cvt_f32_f16_e32 v1, v1
-; SI-NEXT: buffer_store_dword v6, off, s[0:3], s32 offset:184 ; 4-byte Folded Spill
-; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_lshrrev_b32_e32 v6, 16, v35
-; SI-NEXT: v_mov_b32_e32 v35, v43
-; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:168 ; 4-byte Folded Spill
-; SI-NEXT: s_waitcnt expcnt(0)
+; SI-NEXT: v_cvt_f32_f16_e32 v3, v1
; SI-NEXT: v_lshrrev_b32_e32 v1, 16, v61
; SI-NEXT: v_cvt_f32_f16_e32 v60, v1
; SI-NEXT: v_lshrrev_b32_e32 v1, 16, v62
-; SI-NEXT: v_cvt_f32_f16_e32 v1, v1
-; SI-NEXT: v_mov_b32_e32 v38, v10
-; SI-NEXT: buffer_store_dword v8, off, s[0:3], s32 offset:192 ; 4-byte Folded Spill
-; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_lshrrev_b32_e32 v8, 16, v39
-; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:172 ; 4-byte Folded Spill
-; SI-NEXT: s_waitcnt expcnt(0)
+; SI-NEXT: v_cvt_f32_f16_e32 v5, v1
; SI-NEXT: v_lshrrev_b32_e32 v1, 16, v63
; SI-NEXT: v_cvt_f32_f16_e32 v61, v1
; SI-NEXT: v_lshrrev_b32_e32 v1, 16, v19
; SI-NEXT: v_cvt_f32_f16_e32 v1, v1
-; SI-NEXT: v_mov_b32_e32 v19, v28
-; SI-NEXT: v_mov_b32_e32 v28, v14
-; SI-NEXT: v_mov_b32_e32 v39, v22
-; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:180 ; 4-byte Folded Spill
+; SI-NEXT: v_mov_b32_e32 v19, v2
+; SI-NEXT: v_cvt_f32_f16_e32 v9, v9
+; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:172 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_lshrrev_b32_e32 v1, 16, v20
-; SI-NEXT: v_cvt_f32_f16_e32 v62, v1
; SI-NEXT: v_lshrrev_b32_e32 v1, 16, v44
-; SI-NEXT: v_cvt_f32_f16_e32 v44, v1
+; SI-NEXT: v_mov_b32_e32 v44, v3
+; SI-NEXT: v_lshrrev_b32_e32 v3, 16, v58
+; SI-NEXT: v_cvt_f32_f16_e32 v3, v3
+; SI-NEXT: v_cvt_f32_f16_e32 v62, v1
+; SI-NEXT: v_lshrrev_b32_e32 v1, 16, v21
+; SI-NEXT: v_cvt_f32_f16_e32 v2, v1
+; SI-NEXT: buffer_store_dword v3, off, s[0:3], s32 offset:180 ; 4-byte Folded Spill
+; SI-NEXT: s_waitcnt expcnt(0)
+; SI-NEXT: v_lshrrev_b32_e32 v3, 16, v14
+; SI-NEXT: buffer_load_dword v14, off, s[0:3], s32 offset:236 ; 4-byte Folded Reload
; SI-NEXT: v_lshrrev_b32_e32 v1, 16, v46
+; SI-NEXT: v_mov_b32_e32 v46, v5
+; SI-NEXT: v_lshrrev_b32_e32 v5, 16, v32
+; SI-NEXT: v_cvt_f32_f16_e32 v5, v5
; SI-NEXT: v_cvt_f32_f16_e32 v63, v1
; SI-NEXT: v_lshrrev_b32_e32 v1, 16, v47
-; SI-NEXT: v_mov_b32_e32 v47, v3
-; SI-NEXT: v_lshrrev_b32_e32 v3, 16, v14
-; SI-NEXT: buffer_load_dword v14, off, s[0:3], s32 offset:232 ; 4-byte Folded Reload
-; SI-NEXT: buffer_load_dword v22, off, s[0:3], s32 offset:236 ; 4-byte Folded Reload
-; SI-NEXT: v_lshrrev_b32_e32 v10, 16, v49
-; SI-NEXT: v_mov_b32_e32 v49, v15
-; SI-NEXT: v_mov_b32_e32 v15, v41
-; SI-NEXT: buffer_load_dword v41, off, s[0:3], s32 offset:248 ; 4-byte Folded Reload
-; SI-NEXT: v_cvt_f32_f16_e32 v33, v11
-; SI-NEXT: v_lshrrev_b32_e32 v11, 16, v53
-; SI-NEXT: v_lshrrev_b32_e32 v12, 16, v50
-; SI-NEXT: buffer_load_dword v50, off, s[0:3], s32 offset:240 ; 4-byte Folded Reload
-; SI-NEXT: v_lshrrev_b32_e32 v13, 16, v52
-; SI-NEXT: v_mov_b32_e32 v51, v53
-; SI-NEXT: v_mov_b32_e32 v53, v54
-; SI-NEXT: v_lshrrev_b32_e32 v43, 16, v54
-; SI-NEXT: buffer_load_dword v54, off, s[0:3], s32 offset:252 ; 4-byte Folded Reload
-; SI-NEXT: buffer_load_dword v52, off, s[0:3], s32 offset:244 ; 4-byte Folded Reload
-; SI-NEXT: v_cvt_f32_f16_e32 v7, v7
-; SI-NEXT: v_cvt_f32_f16_e32 v10, v10
-; SI-NEXT: v_cvt_f32_f16_e32 v12, v12
-; SI-NEXT: v_cvt_f32_f16_e32 v43, v43
-; SI-NEXT: v_mov_b32_e32 v20, v2
+; SI-NEXT: v_cvt_f32_f16_e32 v1, v1
+; SI-NEXT: buffer_store_dword v5, off, s[0:3], s32 offset:184 ; 4-byte Folded Spill
+; SI-NEXT: s_waitcnt expcnt(0)
+; SI-NEXT: v_lshrrev_b32_e32 v5, 16, v33
+; SI-NEXT: v_mov_b32_e32 v33, v6
+; SI-NEXT: v_lshrrev_b32_e32 v6, 16, v34
+; SI-NEXT: v_cvt_f32_f16_e32 v10, v6
+; SI-NEXT: v_mov_b32_e32 v47, v2
; SI-NEXT: v_lshrrev_b32_e32 v2, 16, v57
-; SI-NEXT: buffer_store_dword v7, off, s[0:3], s32 offset:188 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:176 ; 4-byte Folded Spill
+; SI-NEXT: v_mov_b32_e32 v48, v10
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_lshrrev_b32_e32 v7, 16, v37
-; SI-NEXT: v_mov_b32_e32 v37, v9
-; SI-NEXT: v_lshrrev_b32_e32 v9, 16, v48
-; SI-NEXT: v_cvt_f32_f16_e32 v46, v1
; SI-NEXT: v_lshrrev_b32_e32 v1, 16, v56
; SI-NEXT: v_cvt_f32_f16_e32 v56, v2
-; SI-NEXT: v_mov_b32_e32 v27, v58
-; SI-NEXT: v_lshrrev_b32_e32 v2, 16, v58
-; SI-NEXT: v_cvt_f32_f16_e32 v57, v3
-; SI-NEXT: v_lshrrev_b32_e32 v3, 16, v31
-; SI-NEXT: v_cvt_f32_f16_e32 v58, v4
-; SI-NEXT: v_cvt_f32_f16_e32 v32, v9
-; SI-NEXT: buffer_store_dword v10, off, s[0:3], s32 offset:196 ; 4-byte Folded Spill
-; SI-NEXT: buffer_store_dword v12, off, s[0:3], s32 offset:200 ; 4-byte Folded Spill
-; SI-NEXT: v_cvt_f32_f16_e32 v34, v13
-; SI-NEXT: buffer_store_dword v43, off, s[0:3], s32 offset:148 ; 4-byte Folded Spill
+; SI-NEXT: v_lshrrev_b32_e32 v2, 16, v26
+; SI-NEXT: v_mov_b32_e32 v26, v7
+; SI-NEXT: buffer_load_dword v7, off, s[0:3], s32 offset:228 ; 4-byte Folded Reload
+; SI-NEXT: v_mov_b32_e32 v21, v4
+; SI-NEXT: v_lshrrev_b32_e32 v6, 16, v35
+; SI-NEXT: v_mov_b32_e32 v35, v55
+; SI-NEXT: v_mov_b32_e32 v55, v53
+; SI-NEXT: v_mov_b32_e32 v34, v43
+; SI-NEXT: v_lshrrev_b32_e32 v43, 16, v41
+; SI-NEXT: v_cvt_f32_f16_e32 v43, v43
; SI-NEXT: v_cvt_f32_f16_e32 v1, v1
; SI-NEXT: v_cvt_f32_f16_e32 v2, v2
; SI-NEXT: v_cvt_f32_f16_e32 v3, v3
+; SI-NEXT: buffer_store_dword v43, off, s[0:3], s32 offset:148 ; 4-byte Folded Spill
+; SI-NEXT: s_waitcnt expcnt(0)
+; SI-NEXT: v_lshrrev_b32_e32 v43, 16, v50
; SI-NEXT: v_cvt_f32_f16_e32 v5, v5
; SI-NEXT: v_cvt_f32_f16_e32 v6, v6
+; SI-NEXT: v_mov_b32_e32 v32, v42
+; SI-NEXT: v_mov_b32_e32 v42, v41
+; SI-NEXT: v_cvt_f32_f16_e32 v43, v43
+; SI-NEXT: s_waitcnt vmcnt(8)
+; SI-NEXT: v_lshrrev_b32_e32 v10, 16, v39
+; SI-NEXT: v_cvt_f32_f16_e32 v10, v10
+; SI-NEXT: buffer_store_dword v10, off, s[0:3], s32 offset:196 ; 4-byte Folded Spill
+; SI-NEXT: s_waitcnt expcnt(0)
+; SI-NEXT: v_lshrrev_b32_e32 v10, 16, v51
+; SI-NEXT: buffer_load_dword v51, off, s[0:3], s32 offset:248 ; 4-byte Folded Reload
+; SI-NEXT: v_cvt_f32_f16_e32 v10, v10
+; SI-NEXT: s_waitcnt vmcnt(6)
+; SI-NEXT: v_lshrrev_b32_e32 v4, 16, v14
+; SI-NEXT: v_cvt_f32_f16_e32 v57, v4
+; SI-NEXT: v_lshrrev_b32_e32 v4, 16, v31
+; SI-NEXT: v_mov_b32_e32 v31, v36
+; SI-NEXT: v_mov_b32_e32 v36, v54
+; SI-NEXT: v_cvt_f32_f16_e32 v4, v4
+; SI-NEXT: s_waitcnt vmcnt(3)
+; SI-NEXT: v_lshrrev_b32_e32 v7, 16, v7
+; SI-NEXT: v_cvt_f32_f16_e32 v7, v7
+; SI-NEXT: buffer_store_dword v7, off, s[0:3], s32 offset:188 ; 4-byte Folded Spill
+; SI-NEXT: s_waitcnt expcnt(0)
+; SI-NEXT: v_lshrrev_b32_e32 v7, 16, v37
+; SI-NEXT: v_mov_b32_e32 v37, v24
+; SI-NEXT: v_mov_b32_e32 v24, v8
+; SI-NEXT: v_lshrrev_b32_e32 v8, 16, v38
+; SI-NEXT: v_cvt_f32_f16_e32 v58, v8
+; SI-NEXT: v_lshrrev_b32_e32 v8, 16, v49
; SI-NEXT: v_cvt_f32_f16_e32 v7, v7
+; SI-NEXT: v_mov_b32_e32 v38, v40
; SI-NEXT: v_cvt_f32_f16_e32 v8, v8
+; SI-NEXT: v_mov_b32_e32 v40, v50
+; SI-NEXT: s_waitcnt vmcnt(1)
+; SI-NEXT: v_lshrrev_b32_e32 v11, 16, v51
; SI-NEXT: v_cvt_f32_f16_e32 v11, v11
-; SI-NEXT: s_waitcnt vmcnt(9)
-; SI-NEXT: v_lshrrev_b32_e32 v4, 16, v14
-; SI-NEXT: s_waitcnt vmcnt(8)
-; SI-NEXT: v_lshrrev_b32_e32 v9, 16, v22
-; SI-NEXT: v_cvt_f32_f16_e32 v4, v4
-; SI-NEXT: v_cvt_f32_f16_e32 v9, v9
-; SI-NEXT: s_waitcnt vmcnt(7) expcnt(2)
-; SI-NEXT: v_lshrrev_b32_e32 v10, 16, v41
-; SI-NEXT: v_cvt_f32_f16_e32 v10, v10
-; SI-NEXT: s_waitcnt vmcnt(6) expcnt(1)
-; SI-NEXT: v_lshrrev_b32_e32 v12, 16, v50
+; SI-NEXT: buffer_store_dword v11, off, s[0:3], s32 offset:200 ; 4-byte Folded Spill
+; SI-NEXT: s_waitcnt expcnt(0)
+; SI-NEXT: v_lshrrev_b32_e32 v11, 16, v52
+; SI-NEXT: buffer_load_dword v52, off, s[0:3], s32 offset:252 ; 4-byte Folded Reload
+; SI-NEXT: v_cvt_f32_f16_e32 v11, v11
+; SI-NEXT: s_waitcnt vmcnt(0)
+; SI-NEXT: v_lshrrev_b32_e32 v12, 16, v52
+; SI-NEXT: v_cvt_f32_f16_e32 v13, v12
+; SI-NEXT: v_lshrrev_b32_e32 v12, 16, v54
; SI-NEXT: v_cvt_f32_f16_e32 v12, v12
-; SI-NEXT: s_waitcnt vmcnt(5) expcnt(0)
-; SI-NEXT: v_lshrrev_b32_e32 v43, 16, v54
-; SI-NEXT: s_waitcnt vmcnt(4)
-; SI-NEXT: v_lshrrev_b32_e32 v13, 16, v52
+; SI-NEXT: v_mov_b32_e32 v54, v13
+; SI-NEXT: v_lshrrev_b32_e32 v13, 16, v53
+; SI-NEXT: buffer_load_dword v53, off, s[0:3], s32 offset:256 ; 4-byte Folded Reload
+; SI-NEXT: v_cvt_f32_f16_e32 v13, v13
+; SI-NEXT: buffer_store_dword v13, off, s[0:3], s32 offset:232 ; 4-byte Folded Spill
+; SI-NEXT: s_waitcnt vmcnt(1) expcnt(0)
+; SI-NEXT: v_lshrrev_b32_e32 v13, 16, v53
; SI-NEXT: v_cvt_f32_f16_e32 v13, v13
-; SI-NEXT: v_cvt_f32_f16_e32 v43, v43
; SI-NEXT: s_cbranch_execnz .LBB101_3
; SI-NEXT: .LBB101_2: ; %cmp.true
-; SI-NEXT: s_waitcnt vmcnt(0)
-; SI-NEXT: v_and_b32_e32 v1, 0xffff0000, v54
+; SI-NEXT: v_and_b32_e32 v1, 0xffff0000, v40
+; SI-NEXT: v_add_f32_e32 v1, 0x40c00000, v1
+; SI-NEXT: v_lshrrev_b32_e32 v50, 16, v1
+; SI-NEXT: v_and_b32_e32 v1, 0xffff0000, v42
; SI-NEXT: v_add_f32_e32 v1, 0x40c00000, v1
; SI-NEXT: v_lshrrev_b32_e32 v1, 16, v1
-; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:232 ; 4-byte Folded Spill
-; SI-NEXT: s_waitcnt expcnt(0)
+; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:148 ; 4-byte Folded Spill
+; SI-NEXT: s_waitcnt vmcnt(1) expcnt(0)
; SI-NEXT: v_and_b32_e32 v1, 0xffff0000, v53
; SI-NEXT: v_add_f32_e32 v1, 0x40c00000, v1
; SI-NEXT: v_lshrrev_b32_e32 v1, 16, v1
-; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:148 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:236 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_and_b32_e32 v1, 0xffff0000, v52
-; SI-NEXT: v_add_f32_e32 v1, 0x40c00000, v1
-; SI-NEXT: v_lshrrev_b32_e32 v52, 16, v1
-; SI-NEXT: v_and_b32_e32 v1, 0xffff0000, v40
+; SI-NEXT: v_and_b32_e32 v1, 0xffff0000, v55
; SI-NEXT: v_add_f32_e32 v1, 0x40c00000, v1
; SI-NEXT: v_lshrrev_b32_e32 v1, 16, v1
-; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:236 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:232 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_and_b32_e32 v1, 0xffff0000, v50
+; SI-NEXT: v_and_b32_e32 v1, 0xffff0000, v36
; SI-NEXT: v_add_f32_e32 v1, 0x40c00000, v1
-; SI-NEXT: v_lshrrev_b32_e32 v50, 16, v1
-; SI-NEXT: v_and_b32_e32 v1, 0xffff0000, v55
+; SI-NEXT: v_lshrrev_b32_e32 v36, 16, v1
+; SI-NEXT: v_and_b32_e32 v1, 0xffff0000, v52
+; SI-NEXT: v_add_f32_e32 v1, 0x40c00000, v1
+; SI-NEXT: v_lshrrev_b32_e32 v52, 16, v1
+; SI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:436 ; 4-byte Folded Reload
+; SI-NEXT: s_waitcnt vmcnt(0)
+; SI-NEXT: v_and_b32_e32 v1, 0xffff0000, v1
; SI-NEXT: v_add_f32_e32 v1, 0x40c00000, v1
; SI-NEXT: v_lshrrev_b32_e32 v1, 16, v1
-; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:200 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:240 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_and_b32_e32 v1, 0xffff0000, v51
; SI-NEXT: v_add_f32_e32 v1, 0x40c00000, v1
; SI-NEXT: v_lshrrev_b32_e32 v51, 16, v1
-; SI-NEXT: v_and_b32_e32 v1, 0xffff0000, v42
+; SI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:432 ; 4-byte Folded Reload
+; SI-NEXT: s_waitcnt vmcnt(0)
+; SI-NEXT: v_and_b32_e32 v1, 0xffff0000, v1
; SI-NEXT: v_add_f32_e32 v1, 0x40c00000, v1
; SI-NEXT: v_lshrrev_b32_e32 v1, 16, v1
-; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:240 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:200 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_and_b32_e32 v1, 0xffff0000, v41
+; SI-NEXT: v_and_b32_e32 v1, 0xffff0000, v39
; SI-NEXT: v_add_f32_e32 v1, 0x40c00000, v1
; SI-NEXT: v_lshrrev_b32_e32 v1, 16, v1
-; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:244 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:196 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:428 ; 4-byte Folded Reload
; SI-NEXT: s_waitcnt vmcnt(0)
; SI-NEXT: v_and_b32_e32 v1, 0xffff0000, v1
; SI-NEXT: v_add_f32_e32 v1, 0x40c00000, v1
; SI-NEXT: v_lshrrev_b32_e32 v1, 16, v1
-; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:196 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:244 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_and_b32_e32 v1, 0xffff0000, v22
+; SI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:424 ; 4-byte Folded Reload
+; SI-NEXT: s_waitcnt vmcnt(0)
+; SI-NEXT: v_and_b32_e32 v1, 0xffff0000, v1
; SI-NEXT: v_add_f32_e32 v1, 0x40c00000, v1
; SI-NEXT: v_lshrrev_b32_e32 v1, 16, v1
-; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:248 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:192 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_and_b32_e32 v1, 0xffff0000, v48
+; SI-NEXT: v_and_b32_e32 v1, 0xffff0000, v49
; SI-NEXT: v_add_f32_e32 v1, 0x40c00000, v1
-; SI-NEXT: v_lshrrev_b32_e32 v48, 16, v1
-; SI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:424 ; 4-byte Folded Reload
+; SI-NEXT: v_lshrrev_b32_e32 v49, 16, v1
+; SI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:420 ; 4-byte Folded Reload
; SI-NEXT: s_waitcnt vmcnt(0)
; SI-NEXT: v_and_b32_e32 v1, 0xffff0000, v1
; SI-NEXT: v_add_f32_e32 v1, 0x40c00000, v1
; SI-NEXT: v_lshrrev_b32_e32 v1, 16, v1
-; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:252 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:248 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:420 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:416 ; 4-byte Folded Reload
; SI-NEXT: s_waitcnt vmcnt(0)
; SI-NEXT: v_and_b32_e32 v1, 0xffff0000, v1
; SI-NEXT: v_add_f32_e32 v1, 0x40c00000, v1
-; SI-NEXT: v_lshrrev_b32_e32 v43, 16, v1
-; SI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:416 ; 4-byte Folded Reload
+; SI-NEXT: v_lshrrev_b32_e32 v1, 16, v1
+; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:252 ; 4-byte Folded Spill
+; SI-NEXT: s_waitcnt expcnt(0)
+; SI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:228 ; 4-byte Folded Reload
; SI-NEXT: s_waitcnt vmcnt(0)
; SI-NEXT: v_and_b32_e32 v1, 0xffff0000, v1
; SI-NEXT: v_add_f32_e32 v1, 0x40c00000, v1
-; SI-NEXT: v_lshrrev_b32_e32 v34, 16, v1
+; SI-NEXT: v_lshrrev_b32_e32 v43, 16, v1
; SI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:412 ; 4-byte Folded Reload
; SI-NEXT: s_waitcnt vmcnt(0)
; SI-NEXT: v_and_b32_e32 v1, 0xffff0000, v1
; SI-NEXT: v_add_f32_e32 v1, 0x40c00000, v1
-; SI-NEXT: v_lshrrev_b32_e32 v33, 16, v1
+; SI-NEXT: v_lshrrev_b32_e32 v13, 16, v1
; SI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:408 ; 4-byte Folded Reload
; SI-NEXT: s_waitcnt vmcnt(0)
; SI-NEXT: v_and_b32_e32 v1, 0xffff0000, v1
; SI-NEXT: v_add_f32_e32 v1, 0x40c00000, v1
-; SI-NEXT: v_lshrrev_b32_e32 v13, 16, v1
+; SI-NEXT: v_lshrrev_b32_e32 v12, 16, v1
; SI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:404 ; 4-byte Folded Reload
+; SI-NEXT: v_cvt_f32_f16_e32 v48, v12
+; SI-NEXT: v_cvt_f32_f16_e32 v12, v36
; SI-NEXT: s_waitcnt vmcnt(0)
; SI-NEXT: v_and_b32_e32 v1, 0xffff0000, v1
; SI-NEXT: v_add_f32_e32 v1, 0x40c00000, v1
-; SI-NEXT: v_lshrrev_b32_e32 v12, 16, v1
+; SI-NEXT: v_lshrrev_b32_e32 v11, 16, v1
; SI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:400 ; 4-byte Folded Reload
; SI-NEXT: s_waitcnt vmcnt(0)
; SI-NEXT: v_and_b32_e32 v1, 0xffff0000, v1
; SI-NEXT: v_add_f32_e32 v1, 0x40c00000, v1
-; SI-NEXT: v_lshrrev_b32_e32 v11, 16, v1
+; SI-NEXT: v_lshrrev_b32_e32 v39, 16, v1
; SI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:396 ; 4-byte Folded Reload
; SI-NEXT: s_waitcnt vmcnt(0)
; SI-NEXT: v_and_b32_e32 v1, 0xffff0000, v1
@@ -229798,10 +229798,10 @@ define inreg <64 x half> @bitcast_v64bf16_to_v64f16_scalar(<64 x bfloat> inreg %
; SI-NEXT: v_and_b32_e32 v1, 0xffff0000, v14
; SI-NEXT: v_add_f32_e32 v1, 0x40c00000, v1
; SI-NEXT: v_lshrrev_b32_e32 v9, 16, v1
-; SI-NEXT: v_and_b32_e32 v1, 0xffff0000, v29
+; SI-NEXT: v_and_b32_e32 v1, 0xffff0000, v30
; SI-NEXT: v_add_f32_e32 v1, 0x40c00000, v1
; SI-NEXT: v_lshrrev_b32_e32 v8, 16, v1
-; SI-NEXT: v_and_b32_e32 v1, 0xffff0000, v31
+; SI-NEXT: v_and_b32_e32 v1, 0xffff0000, v29
; SI-NEXT: v_add_f32_e32 v1, 0x40c00000, v1
; SI-NEXT: v_lshrrev_b32_e32 v7, 16, v1
; SI-NEXT: v_and_b32_e32 v1, 0xffff0000, v28
@@ -229810,335 +229810,329 @@ define inreg <64 x half> @bitcast_v64bf16_to_v64f16_scalar(<64 x bfloat> inreg %
; SI-NEXT: v_and_b32_e32 v1, 0xffff0000, v27
; SI-NEXT: v_add_f32_e32 v1, 0x40c00000, v1
; SI-NEXT: v_lshrrev_b32_e32 v5, 16, v1
-; SI-NEXT: v_and_b32_e32 v1, 0xffff0000, v26
-; SI-NEXT: v_add_f32_e32 v1, 0x40c00000, v1
-; SI-NEXT: v_lshrrev_b32_e32 v4, 16, v1
; SI-NEXT: v_and_b32_e32 v1, 0xffff0000, v25
; SI-NEXT: v_add_f32_e32 v1, 0x40c00000, v1
-; SI-NEXT: v_lshrrev_b32_e32 v3, 16, v1
+; SI-NEXT: v_lshrrev_b32_e32 v4, 16, v1
; SI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:392 ; 4-byte Folded Reload
-; SI-NEXT: v_and_b32_e32 v14, 0xffff0000, v24
+; SI-NEXT: s_waitcnt vmcnt(0)
+; SI-NEXT: v_and_b32_e32 v14, 0xffff0000, v1
+; SI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:388 ; 4-byte Folded Reload
; SI-NEXT: v_add_f32_e32 v14, 0x40c00000, v14
-; SI-NEXT: v_lshrrev_b32_e32 v2, 16, v14
+; SI-NEXT: v_lshrrev_b32_e32 v30, 16, v14
; SI-NEXT: v_and_b32_e32 v14, 0xffff0000, v23
; SI-NEXT: v_add_f32_e32 v14, 0x40c00000, v14
; SI-NEXT: v_lshrrev_b32_e32 v63, 16, v14
-; SI-NEXT: v_cvt_f32_f16_e32 v63, v63
-; SI-NEXT: s_waitcnt vmcnt(0)
-; SI-NEXT: v_and_b32_e32 v14, 0xffff0000, v1
+; SI-NEXT: v_and_b32_e32 v14, 0xffff0000, v22
; SI-NEXT: v_add_f32_e32 v14, 0x40c00000, v14
-; SI-NEXT: v_lshrrev_b32_e32 v1, 16, v14
-; SI-NEXT: buffer_load_dword v14, off, s[0:3], s32 offset:352 ; 4-byte Folded Reload
-; SI-NEXT: s_waitcnt vmcnt(0)
-; SI-NEXT: v_and_b32_e32 v14, 0xffff0000, v14
+; SI-NEXT: v_lshrrev_b32_e32 v3, 16, v14
+; SI-NEXT: v_and_b32_e32 v14, 0xffff0000, v20
; SI-NEXT: v_add_f32_e32 v14, 0x40c00000, v14
; SI-NEXT: v_lshrrev_b32_e32 v62, 16, v14
-; SI-NEXT: buffer_load_dword v14, off, s[0:3], s32 offset:344 ; 4-byte Folded Reload
; SI-NEXT: v_cvt_f32_f16_e32 v62, v62
+; SI-NEXT: v_cvt_f32_f16_e32 v63, v63
; SI-NEXT: s_waitcnt vmcnt(0)
-; SI-NEXT: v_and_b32_e32 v14, 0xffff0000, v14
+; SI-NEXT: v_and_b32_e32 v14, 0xffff0000, v1
+; SI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:384 ; 4-byte Folded Reload
; SI-NEXT: v_add_f32_e32 v14, 0x40c00000, v14
; SI-NEXT: v_lshrrev_b32_e32 v28, 16, v14
-; SI-NEXT: buffer_load_dword v14, off, s[0:3], s32 offset:336 ; 4-byte Folded Reload
; SI-NEXT: s_waitcnt vmcnt(0)
-; SI-NEXT: v_and_b32_e32 v14, 0xffff0000, v14
+; SI-NEXT: v_and_b32_e32 v14, 0xffff0000, v1
; SI-NEXT: v_add_f32_e32 v14, 0x40c00000, v14
; SI-NEXT: v_lshrrev_b32_e32 v61, 16, v14
-; SI-NEXT: buffer_load_dword v14, off, s[0:3], s32 offset:328 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v14, off, s[0:3], s32 offset:332 ; 4-byte Folded Reload
; SI-NEXT: v_cvt_f32_f16_e32 v61, v61
; SI-NEXT: s_waitcnt vmcnt(0)
; SI-NEXT: v_and_b32_e32 v14, 0xffff0000, v14
; SI-NEXT: v_add_f32_e32 v14, 0x40c00000, v14
-; SI-NEXT: v_lshrrev_b32_e32 v27, 16, v14
-; SI-NEXT: buffer_load_dword v14, off, s[0:3], s32 offset:320 ; 4-byte Folded Reload
+; SI-NEXT: v_lshrrev_b32_e32 v2, 16, v14
+; SI-NEXT: buffer_load_dword v14, off, s[0:3], s32 offset:324 ; 4-byte Folded Reload
; SI-NEXT: s_waitcnt vmcnt(0)
; SI-NEXT: v_and_b32_e32 v14, 0xffff0000, v14
; SI-NEXT: v_add_f32_e32 v14, 0x40c00000, v14
; SI-NEXT: v_lshrrev_b32_e32 v60, 16, v14
-; SI-NEXT: buffer_load_dword v14, off, s[0:3], s32 offset:312 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v14, off, s[0:3], s32 offset:316 ; 4-byte Folded Reload
; SI-NEXT: v_cvt_f32_f16_e32 v60, v60
; SI-NEXT: s_waitcnt vmcnt(0)
; SI-NEXT: v_and_b32_e32 v14, 0xffff0000, v14
; SI-NEXT: v_add_f32_e32 v14, 0x40c00000, v14
-; SI-NEXT: v_lshrrev_b32_e32 v26, 16, v14
-; SI-NEXT: buffer_load_dword v14, off, s[0:3], s32 offset:308 ; 4-byte Folded Reload
+; SI-NEXT: v_lshrrev_b32_e32 v1, 16, v14
+; SI-NEXT: buffer_load_dword v14, off, s[0:3], s32 offset:312 ; 4-byte Folded Reload
; SI-NEXT: s_waitcnt vmcnt(0)
; SI-NEXT: v_and_b32_e32 v14, 0xffff0000, v14
; SI-NEXT: v_add_f32_e32 v14, 0x40c00000, v14
; SI-NEXT: v_lshrrev_b32_e32 v59, 16, v14
-; SI-NEXT: buffer_load_dword v14, off, s[0:3], s32 offset:304 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v14, off, s[0:3], s32 offset:308 ; 4-byte Folded Reload
; SI-NEXT: v_cvt_f32_f16_e32 v59, v59
; SI-NEXT: s_waitcnt vmcnt(0)
; SI-NEXT: v_and_b32_e32 v14, 0xffff0000, v14
; SI-NEXT: v_add_f32_e32 v14, 0x40c00000, v14
; SI-NEXT: v_lshrrev_b32_e32 v25, 16, v14
-; SI-NEXT: buffer_load_dword v14, off, s[0:3], s32 offset:300 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v14, off, s[0:3], s32 offset:304 ; 4-byte Folded Reload
; SI-NEXT: s_waitcnt vmcnt(0)
; SI-NEXT: v_and_b32_e32 v14, 0xffff0000, v14
; SI-NEXT: v_add_f32_e32 v14, 0x40c00000, v14
; SI-NEXT: v_lshrrev_b32_e32 v58, 16, v14
-; SI-NEXT: buffer_load_dword v14, off, s[0:3], s32 offset:296 ; 4-byte Folded Reload
-; SI-NEXT: v_cvt_f32_f16_e32 v38, v58
-; SI-NEXT: v_cvt_f32_f16_e32 v58, v8
-; SI-NEXT: v_cvt_f32_f16_e32 v8, v43
-; SI-NEXT: buffer_store_dword v8, off, s[0:3], s32 offset:192 ; 4-byte Folded Spill
-; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: buffer_load_dword v8, off, s[0:3], s32 offset:252 ; 4-byte Folded Reload
-; SI-NEXT: s_waitcnt vmcnt(2)
+; SI-NEXT: buffer_load_dword v14, off, s[0:3], s32 offset:300 ; 4-byte Folded Reload
+; SI-NEXT: s_waitcnt vmcnt(0)
; SI-NEXT: v_and_b32_e32 v14, 0xffff0000, v14
; SI-NEXT: v_add_f32_e32 v14, 0x40c00000, v14
; SI-NEXT: v_lshrrev_b32_e32 v24, 16, v14
-; SI-NEXT: buffer_load_dword v14, off, s[0:3], s32 offset:292 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v14, off, s[0:3], s32 offset:296 ; 4-byte Folded Reload
; SI-NEXT: s_waitcnt vmcnt(0)
; SI-NEXT: v_and_b32_e32 v14, 0xffff0000, v14
; SI-NEXT: v_add_f32_e32 v14, 0x40c00000, v14
; SI-NEXT: v_lshrrev_b32_e32 v57, 16, v14
-; SI-NEXT: buffer_load_dword v14, off, s[0:3], s32 offset:288 ; 4-byte Folded Reload
-; SI-NEXT: v_cvt_f32_f16_e32 v37, v57
-; SI-NEXT: v_cvt_f32_f16_e32 v57, v6
-; SI-NEXT: v_cvt_f32_f16_e32 v6, v12
-; SI-NEXT: buffer_load_dword v12, off, s[0:3], s32 offset:200 ; 4-byte Folded Reload
-; SI-NEXT: buffer_store_dword v6, off, s[0:3], s32 offset:184 ; 4-byte Folded Spill
-; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_cvt_f32_f16_e32 v6, v13
-; SI-NEXT: buffer_load_dword v13, off, s[0:3], s32 offset:236 ; 4-byte Folded Reload
-; SI-NEXT: s_waitcnt vmcnt(3)
+; SI-NEXT: buffer_load_dword v14, off, s[0:3], s32 offset:292 ; 4-byte Folded Reload
+; SI-NEXT: v_cvt_f32_f16_e32 v33, v57
+; SI-NEXT: v_cvt_f32_f16_e32 v57, v9
+; SI-NEXT: buffer_load_dword v9, off, s[0:3], s32 offset:192 ; 4-byte Folded Reload
+; SI-NEXT: s_waitcnt vmcnt(1)
; SI-NEXT: v_and_b32_e32 v14, 0xffff0000, v14
; SI-NEXT: v_add_f32_e32 v14, 0x40c00000, v14
; SI-NEXT: v_lshrrev_b32_e32 v23, 16, v14
-; SI-NEXT: buffer_load_dword v14, off, s[0:3], s32 offset:284 ; 4-byte Folded Reload
-; SI-NEXT: s_waitcnt vmcnt(3)
-; SI-NEXT: v_cvt_f32_f16_e32 v12, v12
-; SI-NEXT: buffer_store_dword v12, off, s[0:3], s32 offset:200 ; 4-byte Folded Spill
-; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_cvt_f32_f16_e32 v12, v50
-; SI-NEXT: s_waitcnt vmcnt(1)
+; SI-NEXT: buffer_load_dword v14, off, s[0:3], s32 offset:288 ; 4-byte Folded Reload
+; SI-NEXT: s_waitcnt vmcnt(0)
; SI-NEXT: v_and_b32_e32 v14, 0xffff0000, v14
; SI-NEXT: v_add_f32_e32 v14, 0x40c00000, v14
; SI-NEXT: v_lshrrev_b32_e32 v56, 16, v14
-; SI-NEXT: buffer_load_dword v14, off, s[0:3], s32 offset:280 ; 4-byte Folded Reload
-; SI-NEXT: v_cvt_f32_f16_e32 v36, v56
-; SI-NEXT: v_cvt_f32_f16_e32 v56, v4
-; SI-NEXT: v_cvt_f32_f16_e32 v4, v9
-; SI-NEXT: buffer_load_dword v9, off, s[0:3], s32 offset:248 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v14, off, s[0:3], s32 offset:284 ; 4-byte Folded Reload
+; SI-NEXT: v_cvt_f32_f16_e32 v26, v56
+; SI-NEXT: v_cvt_f32_f16_e32 v56, v5
+; SI-NEXT: v_cvt_f32_f16_e32 v5, v39
+; SI-NEXT: buffer_store_dword v5, off, s[0:3], s32 offset:184 ; 4-byte Folded Spill
+; SI-NEXT: s_waitcnt expcnt(0)
+; SI-NEXT: v_cvt_f32_f16_e32 v5, v11
+; SI-NEXT: v_cvt_f32_f16_e32 v11, v51
; SI-NEXT: s_waitcnt vmcnt(1)
; SI-NEXT: v_and_b32_e32 v14, 0xffff0000, v14
; SI-NEXT: v_add_f32_e32 v14, 0x40c00000, v14
; SI-NEXT: v_lshrrev_b32_e32 v22, 16, v14
-; SI-NEXT: buffer_load_dword v14, off, s[0:3], s32 offset:276 ; 4-byte Folded Reload
-; SI-NEXT: v_cvt_f32_f16_e32 v22, v22
-; SI-NEXT: s_waitcnt vmcnt(1)
-; SI-NEXT: v_cvt_f32_f16_e32 v9, v9
-; SI-NEXT: buffer_store_dword v22, off, s[0:3], s32 offset:152 ; 4-byte Folded Spill
-; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_cvt_f32_f16_e32 v22, v23
-; SI-NEXT: buffer_store_dword v22, off, s[0:3], s32 offset:156 ; 4-byte Folded Spill
-; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_cvt_f32_f16_e32 v22, v25
-; SI-NEXT: buffer_store_dword v22, off, s[0:3], s32 offset:164 ; 4-byte Folded Spill
-; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_cvt_f32_f16_e32 v22, v26
-; SI-NEXT: buffer_store_dword v22, off, s[0:3], s32 offset:168 ; 4-byte Folded Spill
-; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_cvt_f32_f16_e32 v22, v27
-; SI-NEXT: buffer_store_dword v22, off, s[0:3], s32 offset:172 ; 4-byte Folded Spill
-; SI-NEXT: s_waitcnt vmcnt(5)
+; SI-NEXT: buffer_load_dword v14, off, s[0:3], s32 offset:280 ; 4-byte Folded Reload
+; SI-NEXT: s_waitcnt vmcnt(0)
; SI-NEXT: v_and_b32_e32 v14, 0xffff0000, v14
; SI-NEXT: v_add_f32_e32 v14, 0x40c00000, v14
; SI-NEXT: v_lshrrev_b32_e32 v47, 16, v14
-; SI-NEXT: buffer_load_dword v14, off, s[0:3], s32 offset:272 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v14, off, s[0:3], s32 offset:276 ; 4-byte Folded Reload
; SI-NEXT: s_waitcnt vmcnt(0)
; SI-NEXT: v_and_b32_e32 v14, 0xffff0000, v14
; SI-NEXT: v_add_f32_e32 v14, 0x40c00000, v14
; SI-NEXT: v_lshrrev_b32_e32 v21, 16, v14
-; SI-NEXT: buffer_load_dword v14, off, s[0:3], s32 offset:268 ; 4-byte Folded Reload
-; SI-NEXT: v_cvt_f32_f16_e32 v21, v21
-; SI-NEXT: buffer_store_dword v21, off, s[0:3], s32 offset:160 ; 4-byte Folded Spill
-; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_cvt_f32_f16_e32 v21, v47
-; SI-NEXT: v_cvt_f32_f16_e32 v47, v24
-; SI-NEXT: s_waitcnt vmcnt(1)
+; SI-NEXT: buffer_load_dword v14, off, s[0:3], s32 offset:272 ; 4-byte Folded Reload
+; SI-NEXT: s_waitcnt vmcnt(0)
; SI-NEXT: v_and_b32_e32 v14, 0xffff0000, v14
; SI-NEXT: v_add_f32_e32 v14, 0x40c00000, v14
; SI-NEXT: v_lshrrev_b32_e32 v46, 16, v14
-; SI-NEXT: buffer_load_dword v14, off, s[0:3], s32 offset:264 ; 4-byte Folded Reload
-; SI-NEXT: s_waitcnt vmcnt(0)
+; SI-NEXT: buffer_load_dword v14, off, s[0:3], s32 offset:268 ; 4-byte Folded Reload
+; SI-NEXT: v_cvt_f32_f16_e32 v34, v46
+; SI-NEXT: v_cvt_f32_f16_e32 v46, v2
+; SI-NEXT: v_cvt_f32_f16_e32 v2, v6
+; SI-NEXT: v_cvt_f32_f16_e32 v6, v13
+; SI-NEXT: buffer_load_dword v13, off, s[0:3], s32 offset:232 ; 4-byte Folded Reload
+; SI-NEXT: s_waitcnt vmcnt(1)
; SI-NEXT: v_and_b32_e32 v14, 0xffff0000, v14
; SI-NEXT: v_add_f32_e32 v14, 0x40c00000, v14
; SI-NEXT: v_lshrrev_b32_e32 v20, 16, v14
-; SI-NEXT: buffer_load_dword v14, off, s[0:3], s32 offset:260 ; 4-byte Folded Reload
-; SI-NEXT: v_cvt_f32_f16_e32 v20, v20
-; SI-NEXT: s_waitcnt vmcnt(0)
+; SI-NEXT: buffer_load_dword v14, off, s[0:3], s32 offset:264 ; 4-byte Folded Reload
+; SI-NEXT: s_waitcnt vmcnt(1)
+; SI-NEXT: v_cvt_f32_f16_e32 v13, v13
+; SI-NEXT: buffer_store_dword v13, off, s[0:3], s32 offset:232 ; 4-byte Folded Spill
+; SI-NEXT: s_waitcnt expcnt(0)
+; SI-NEXT: buffer_load_dword v13, off, s[0:3], s32 offset:236 ; 4-byte Folded Reload
+; SI-NEXT: s_waitcnt vmcnt(2)
; SI-NEXT: v_and_b32_e32 v14, 0xffff0000, v14
; SI-NEXT: v_add_f32_e32 v14, 0x40c00000, v14
; SI-NEXT: v_lshrrev_b32_e32 v45, 16, v14
-; SI-NEXT: buffer_load_dword v14, off, s[0:3], s32 offset:256 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v14, off, s[0:3], s32 offset:260 ; 4-byte Folded Reload
; SI-NEXT: v_cvt_f32_f16_e32 v45, v45
+; SI-NEXT: s_waitcnt vmcnt(1)
+; SI-NEXT: v_cvt_f32_f16_e32 v13, v13
; SI-NEXT: s_waitcnt vmcnt(0)
; SI-NEXT: v_and_b32_e32 v14, 0xffff0000, v14
; SI-NEXT: v_add_f32_e32 v14, 0x40c00000, v14
; SI-NEXT: v_lshrrev_b32_e32 v19, 16, v14
-; SI-NEXT: buffer_load_dword v14, off, s[0:3], s32 offset:388 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v14, off, s[0:3], s32 offset:380 ; 4-byte Folded Reload
; SI-NEXT: v_cvt_f32_f16_e32 v19, v19
-; SI-NEXT: s_waitcnt vmcnt(0)
+; SI-NEXT: buffer_store_dword v19, off, s[0:3], s32 offset:152 ; 4-byte Folded Spill
+; SI-NEXT: s_waitcnt expcnt(0)
+; SI-NEXT: v_cvt_f32_f16_e32 v19, v20
+; SI-NEXT: v_cvt_f32_f16_e32 v20, v22
+; SI-NEXT: buffer_store_dword v19, off, s[0:3], s32 offset:156 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v20, off, s[0:3], s32 offset:160 ; 4-byte Folded Spill
+; SI-NEXT: s_waitcnt expcnt(0)
+; SI-NEXT: v_cvt_f32_f16_e32 v20, v23
+; SI-NEXT: v_cvt_f32_f16_e32 v19, v21
+; SI-NEXT: v_cvt_f32_f16_e32 v21, v24
+; SI-NEXT: v_cvt_f32_f16_e32 v24, v58
+; SI-NEXT: buffer_store_dword v20, off, s[0:3], s32 offset:164 ; 4-byte Folded Spill
+; SI-NEXT: s_waitcnt expcnt(0)
+; SI-NEXT: v_cvt_f32_f16_e32 v20, v25
+; SI-NEXT: buffer_store_dword v20, off, s[0:3], s32 offset:168 ; 4-byte Folded Spill
+; SI-NEXT: s_waitcnt vmcnt(5)
; SI-NEXT: v_and_b32_e32 v14, 0xffff0000, v14
; SI-NEXT: v_add_f32_e32 v14, 0x40c00000, v14
; SI-NEXT: v_lshrrev_b32_e32 v17, 16, v14
-; SI-NEXT: buffer_load_dword v14, off, s[0:3], s32 offset:384 ; 4-byte Folded Reload
-; SI-NEXT: v_cvt_f32_f16_e32 v30, v17
-; SI-NEXT: v_cvt_f32_f16_e32 v17, v46
-; SI-NEXT: v_cvt_f32_f16_e32 v46, v2
-; SI-NEXT: v_cvt_f32_f16_e32 v2, v5
-; SI-NEXT: v_cvt_f32_f16_e32 v5, v10
-; SI-NEXT: buffer_load_dword v10, off, s[0:3], s32 offset:196 ; 4-byte Folded Reload
-; SI-NEXT: buffer_store_dword v5, off, s[0:3], s32 offset:176 ; 4-byte Folded Spill
-; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_cvt_f32_f16_e32 v5, v11
-; SI-NEXT: buffer_load_dword v11, off, s[0:3], s32 offset:240 ; 4-byte Folded Reload
-; SI-NEXT: s_waitcnt vmcnt(3)
+; SI-NEXT: buffer_load_dword v14, off, s[0:3], s32 offset:376 ; 4-byte Folded Reload
+; SI-NEXT: v_cvt_f32_f16_e32 v17, v17
+; SI-NEXT: s_waitcnt vmcnt(0)
; SI-NEXT: v_and_b32_e32 v14, 0xffff0000, v14
; SI-NEXT: v_add_f32_e32 v14, 0x40c00000, v14
; SI-NEXT: v_lshrrev_b32_e32 v55, 16, v14
-; SI-NEXT: buffer_load_dword v14, off, s[0:3], s32 offset:380 ; 4-byte Folded Reload
-; SI-NEXT: v_cvt_f32_f16_e32 v35, v55
+; SI-NEXT: buffer_load_dword v14, off, s[0:3], s32 offset:372 ; 4-byte Folded Reload
; SI-NEXT: s_waitcnt vmcnt(0)
; SI-NEXT: v_and_b32_e32 v14, 0xffff0000, v14
; SI-NEXT: v_add_f32_e32 v14, 0x40c00000, v14
; SI-NEXT: v_lshrrev_b32_e32 v54, 16, v14
-; SI-NEXT: buffer_load_dword v14, off, s[0:3], s32 offset:376 ; 4-byte Folded Reload
-; SI-NEXT: v_cvt_f32_f16_e32 v39, v54
+; SI-NEXT: buffer_load_dword v14, off, s[0:3], s32 offset:368 ; 4-byte Folded Reload
+; SI-NEXT: v_cvt_f32_f16_e32 v37, v54
+; SI-NEXT: v_cvt_f32_f16_e32 v54, v52
; SI-NEXT: s_waitcnt vmcnt(0)
; SI-NEXT: v_and_b32_e32 v14, 0xffff0000, v14
; SI-NEXT: v_add_f32_e32 v14, 0x40c00000, v14
; SI-NEXT: v_lshrrev_b32_e32 v40, 16, v14
-; SI-NEXT: buffer_load_dword v14, off, s[0:3], s32 offset:372 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v14, off, s[0:3], s32 offset:364 ; 4-byte Folded Reload
; SI-NEXT: s_waitcnt vmcnt(0)
; SI-NEXT: v_and_b32_e32 v14, 0xffff0000, v14
; SI-NEXT: v_add_f32_e32 v14, 0x40c00000, v14
; SI-NEXT: v_lshrrev_b32_e32 v41, 16, v14
-; SI-NEXT: buffer_load_dword v14, off, s[0:3], s32 offset:368 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v14, off, s[0:3], s32 offset:360 ; 4-byte Folded Reload
; SI-NEXT: s_waitcnt vmcnt(0)
; SI-NEXT: v_and_b32_e32 v14, 0xffff0000, v14
; SI-NEXT: v_add_f32_e32 v14, 0x40c00000, v14
; SI-NEXT: v_lshrrev_b32_e32 v44, 16, v14
-; SI-NEXT: buffer_load_dword v14, off, s[0:3], s32 offset:364 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v14, off, s[0:3], s32 offset:356 ; 4-byte Folded Reload
; SI-NEXT: s_waitcnt vmcnt(0)
; SI-NEXT: v_and_b32_e32 v14, 0xffff0000, v14
; SI-NEXT: v_add_f32_e32 v14, 0x40c00000, v14
; SI-NEXT: v_lshrrev_b32_e32 v18, 16, v14
-; SI-NEXT: buffer_load_dword v14, off, s[0:3], s32 offset:360 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v14, off, s[0:3], s32 offset:352 ; 4-byte Folded Reload
; SI-NEXT: v_cvt_f32_f16_e32 v18, v18
; SI-NEXT: s_waitcnt vmcnt(0)
; SI-NEXT: v_and_b32_e32 v14, 0xffff0000, v14
; SI-NEXT: v_add_f32_e32 v14, 0x40c00000, v14
; SI-NEXT: v_lshrrev_b32_e32 v31, 16, v14
-; SI-NEXT: buffer_load_dword v14, off, s[0:3], s32 offset:356 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v14, off, s[0:3], s32 offset:348 ; 4-byte Folded Reload
; SI-NEXT: s_waitcnt vmcnt(0)
; SI-NEXT: v_and_b32_e32 v14, 0xffff0000, v14
; SI-NEXT: v_add_f32_e32 v14, 0x40c00000, v14
; SI-NEXT: v_lshrrev_b32_e32 v42, 16, v14
-; SI-NEXT: buffer_load_dword v14, off, s[0:3], s32 offset:348 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v14, off, s[0:3], s32 offset:344 ; 4-byte Folded Reload
; SI-NEXT: s_waitcnt vmcnt(0)
; SI-NEXT: v_and_b32_e32 v14, 0xffff0000, v14
; SI-NEXT: v_add_f32_e32 v14, 0x40c00000, v14
; SI-NEXT: v_lshrrev_b32_e32 v16, 16, v14
; SI-NEXT: buffer_load_dword v14, off, s[0:3], s32 offset:340 ; 4-byte Folded Reload
-; SI-NEXT: v_cvt_f32_f16_e32 v49, v16
-; SI-NEXT: v_cvt_f32_f16_e32 v16, v42
+; SI-NEXT: v_cvt_f32_f16_e32 v16, v16
; SI-NEXT: s_waitcnt vmcnt(0)
; SI-NEXT: v_and_b32_e32 v14, 0xffff0000, v14
; SI-NEXT: v_add_f32_e32 v14, 0x40c00000, v14
; SI-NEXT: v_lshrrev_b32_e32 v32, 16, v14
-; SI-NEXT: buffer_load_dword v14, off, s[0:3], s32 offset:332 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v14, off, s[0:3], s32 offset:336 ; 4-byte Folded Reload
; SI-NEXT: s_waitcnt vmcnt(0)
; SI-NEXT: v_and_b32_e32 v14, 0xffff0000, v14
; SI-NEXT: v_add_f32_e32 v14, 0x40c00000, v14
; SI-NEXT: v_lshrrev_b32_e32 v15, 16, v14
-; SI-NEXT: buffer_load_dword v14, off, s[0:3], s32 offset:324 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v14, off, s[0:3], s32 offset:328 ; 4-byte Folded Reload
+; SI-NEXT: v_cvt_f32_f16_e32 v38, v15
+; SI-NEXT: v_cvt_f32_f16_e32 v15, v32
+; SI-NEXT: v_cvt_f32_f16_e32 v32, v42
; SI-NEXT: s_waitcnt vmcnt(0)
; SI-NEXT: v_and_b32_e32 v14, 0xffff0000, v14
; SI-NEXT: v_add_f32_e32 v14, 0x40c00000, v14
; SI-NEXT: v_lshrrev_b32_e32 v53, 16, v14
-; SI-NEXT: buffer_load_dword v14, off, s[0:3], s32 offset:316 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v14, off, s[0:3], s32 offset:320 ; 4-byte Folded Reload
; SI-NEXT: s_waitcnt vmcnt(0)
; SI-NEXT: v_and_b32_e32 v14, 0xffff0000, v14
; SI-NEXT: v_add_f32_e32 v14, 0x40c00000, v14
; SI-NEXT: v_lshrrev_b32_e32 v14, 16, v14
-; SI-NEXT: v_cvt_f32_f16_e32 v14, v14
-; SI-NEXT: buffer_store_dword v14, off, s[0:3], s32 offset:212 ; 4-byte Folded Spill
-; SI-NEXT: s_waitcnt expcnt(0)
+; SI-NEXT: v_cvt_f32_f16_e32 v35, v14
; SI-NEXT: v_cvt_f32_f16_e32 v14, v53
-; SI-NEXT: buffer_store_dword v14, off, s[0:3], s32 offset:228 ; 4-byte Folded Spill
-; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_cvt_f32_f16_e32 v14, v15
-; SI-NEXT: v_cvt_f32_f16_e32 v15, v32
-; SI-NEXT: v_cvt_f32_f16_e32 v32, v48
; SI-NEXT: buffer_store_dword v14, off, s[0:3], s32 offset:224 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_cvt_f32_f16_e32 v14, v31
+; SI-NEXT: v_cvt_f32_f16_e32 v31, v47
+; SI-NEXT: v_cvt_f32_f16_e32 v47, v3
+; SI-NEXT: v_cvt_f32_f16_e32 v3, v7
; SI-NEXT: buffer_store_dword v14, off, s[0:3], s32 offset:216 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_cvt_f32_f16_e32 v14, v44
; SI-NEXT: v_cvt_f32_f16_e32 v44, v1
-; SI-NEXT: v_cvt_f32_f16_e32 v1, v3
-; SI-NEXT: v_cvt_f32_f16_e32 v3, v7
-; SI-NEXT: buffer_store_dword v14, off, s[0:3], s32 offset:208 ; 4-byte Folded Spill
+; SI-NEXT: v_cvt_f32_f16_e32 v1, v28
+; SI-NEXT: buffer_store_dword v3, off, s[0:3], s32 offset:180 ; 4-byte Folded Spill
+; SI-NEXT: s_waitcnt expcnt(0)
+; SI-NEXT: v_cvt_f32_f16_e32 v3, v8
+; SI-NEXT: buffer_load_dword v8, off, s[0:3], s32 offset:248 ; 4-byte Folded Reload
+; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:172 ; 4-byte Folded Spill
+; SI-NEXT: s_waitcnt expcnt(0)
+; SI-NEXT: v_cvt_f32_f16_e32 v1, v30
+; SI-NEXT: buffer_store_dword v14, off, s[0:3], s32 offset:212 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_cvt_f32_f16_e32 v14, v41
-; SI-NEXT: v_cvt_f32_f16_e32 v7, v33
+; SI-NEXT: v_cvt_f32_f16_e32 v7, v43
+; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:176 ; 4-byte Folded Spill
+; SI-NEXT: s_waitcnt expcnt(0)
+; SI-NEXT: v_cvt_f32_f16_e32 v1, v4
+; SI-NEXT: v_cvt_f32_f16_e32 v4, v10
+; SI-NEXT: buffer_load_dword v10, off, s[0:3], s32 offset:196 ; 4-byte Folded Reload
; SI-NEXT: buffer_store_dword v14, off, s[0:3], s32 offset:220 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_cvt_f32_f16_e32 v14, v40
; SI-NEXT: buffer_store_dword v7, off, s[0:3], s32 offset:188 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_cvt_f32_f16_e32 v7, v34
-; SI-NEXT: v_cvt_f32_f16_e32 v34, v13
-; SI-NEXT: buffer_store_dword v14, off, s[0:3], s32 offset:204 ; 4-byte Folded Spill
+; SI-NEXT: buffer_load_dword v7, off, s[0:3], s32 offset:252 ; 4-byte Folded Reload
+; SI-NEXT: buffer_store_dword v14, off, s[0:3], s32 offset:208 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_cvt_f32_f16_e32 v14, v28
-; SI-NEXT: v_cvt_f32_f16_e32 v13, v52
-; SI-NEXT: buffer_store_dword v14, off, s[0:3], s32 offset:180 ; 4-byte Folded Spill
+; SI-NEXT: v_cvt_f32_f16_e32 v14, v55
+; SI-NEXT: buffer_store_dword v14, off, s[0:3], s32 offset:204 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: buffer_load_dword v14, off, s[0:3], s32 offset:148 ; 4-byte Folded Reload
+; SI-NEXT: v_cvt_f32_f16_e32 v9, v9
+; SI-NEXT: buffer_store_dword v9, off, s[0:3], s32 offset:192 ; 4-byte Folded Spill
+; SI-NEXT: s_waitcnt expcnt(0)
+; SI-NEXT: buffer_load_dword v9, off, s[0:3], s32 offset:244 ; 4-byte Folded Reload
+; SI-NEXT: s_waitcnt vmcnt(12)
+; SI-NEXT: v_cvt_f32_f16_e32 v58, v8
+; SI-NEXT: v_cvt_f32_f16_e32 v8, v49
+; SI-NEXT: s_waitcnt vmcnt(8)
; SI-NEXT: v_cvt_f32_f16_e32 v10, v10
; SI-NEXT: buffer_store_dword v10, off, s[0:3], s32 offset:196 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: buffer_load_dword v10, off, s[0:3], s32 offset:244 ; 4-byte Folded Reload
-; SI-NEXT: v_cvt_f32_f16_e32 v8, v8
-; SI-NEXT: s_waitcnt vmcnt(2)
+; SI-NEXT: buffer_load_dword v10, off, s[0:3], s32 offset:200 ; 4-byte Folded Reload
+; SI-NEXT: buffer_store_dword v11, off, s[0:3], s32 offset:200 ; 4-byte Folded Spill
+; SI-NEXT: s_waitcnt expcnt(0)
+; SI-NEXT: buffer_load_dword v11, off, s[0:3], s32 offset:240 ; 4-byte Folded Reload
+; SI-NEXT: s_waitcnt vmcnt(9)
+; SI-NEXT: v_cvt_f32_f16_e32 v7, v7
+; SI-NEXT: s_waitcnt vmcnt(6)
; SI-NEXT: v_cvt_f32_f16_e32 v43, v14
-; SI-NEXT: buffer_load_dword v14, off, s[0:3], s32 offset:232 ; 4-byte Folded Reload
-; SI-NEXT: v_cvt_f32_f16_e32 v33, v11
-; SI-NEXT: v_cvt_f32_f16_e32 v11, v51
; SI-NEXT: buffer_store_dword v43, off, s[0:3], s32 offset:148 ; 4-byte Folded Spill
-; SI-NEXT: s_waitcnt vmcnt(2)
+; SI-NEXT: s_waitcnt expcnt(0)
+; SI-NEXT: v_cvt_f32_f16_e32 v43, v50
+; SI-NEXT: s_waitcnt vmcnt(5)
+; SI-NEXT: v_cvt_f32_f16_e32 v9, v9
+; SI-NEXT: s_waitcnt vmcnt(3)
; SI-NEXT: v_cvt_f32_f16_e32 v10, v10
-; SI-NEXT: s_waitcnt vmcnt(1) expcnt(0)
-; SI-NEXT: v_cvt_f32_f16_e32 v43, v14
+; SI-NEXT: s_waitcnt vmcnt(1)
+; SI-NEXT: v_cvt_f32_f16_e32 v11, v11
; SI-NEXT: .LBB101_3: ; %end
-; SI-NEXT: buffer_load_dword v14, off, s[0:3], s32 offset:228 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v14, off, s[0:3], s32 offset:224 ; 4-byte Folded Reload
; SI-NEXT: v_cvt_f16_f32_e32 v1, v1
; SI-NEXT: v_lshlrev_b32_e32 v1, 16, v1
; SI-NEXT: s_waitcnt vmcnt(0)
; SI-NEXT: v_cvt_f16_f32_e32 v53, v14
-; SI-NEXT: buffer_load_dword v14, off, s[0:3], s32 offset:212 ; 4-byte Folded Reload
+; SI-NEXT: v_cvt_f16_f32_e32 v14, v35
; SI-NEXT: v_lshlrev_b32_e32 v53, 16, v53
-; SI-NEXT: s_waitcnt vmcnt(0)
-; SI-NEXT: v_cvt_f16_f32_e32 v14, v14
; SI-NEXT: v_or_b32_e32 v14, v14, v53
; SI-NEXT: buffer_store_dword v14, v0, s[0:3], 0 offen
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_cvt_f16_f32_e32 v14, v15
-; SI-NEXT: buffer_load_dword v15, off, s[0:3], s32 offset:224 ; 4-byte Folded Reload
+; SI-NEXT: v_cvt_f16_f32_e32 v15, v38
; SI-NEXT: v_lshlrev_b32_e32 v14, 16, v14
-; SI-NEXT: s_waitcnt vmcnt(0)
-; SI-NEXT: v_cvt_f16_f32_e32 v15, v15
; SI-NEXT: v_or_b32_e32 v14, v15, v14
; SI-NEXT: v_add_i32_e32 v15, vcc, 4, v0
; SI-NEXT: buffer_store_dword v14, v15, s[0:3], 0 offen
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_cvt_f16_f32_e32 v14, v16
-; SI-NEXT: v_cvt_f16_f32_e32 v15, v49
+; SI-NEXT: v_cvt_f16_f32_e32 v14, v32
+; SI-NEXT: v_cvt_f16_f32_e32 v15, v16
; SI-NEXT: v_add_i32_e32 v16, vcc, 12, v0
; SI-NEXT: v_lshlrev_b32_e32 v14, 16, v14
; SI-NEXT: v_or_b32_e32 v14, v15, v14
@@ -230154,7 +230148,7 @@ define inreg <64 x half> @bitcast_v64bf16_to_v64f16_scalar(<64 x bfloat> inreg %
; SI-NEXT: buffer_store_dword v14, v16, s[0:3], 0 offen
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: buffer_load_dword v14, off, s[0:3], s32 offset:220 ; 4-byte Folded Reload
-; SI-NEXT: buffer_load_dword v15, off, s[0:3], s32 offset:208 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v15, off, s[0:3], s32 offset:212 ; 4-byte Folded Reload
; SI-NEXT: v_add_i32_e32 v16, vcc, 16, v0
; SI-NEXT: s_waitcnt vmcnt(1)
; SI-NEXT: v_cvt_f16_f32_e32 v14, v14
@@ -230163,57 +230157,61 @@ define inreg <64 x half> @bitcast_v64bf16_to_v64f16_scalar(<64 x bfloat> inreg %
; SI-NEXT: v_lshlrev_b32_e32 v14, 16, v14
; SI-NEXT: v_or_b32_e32 v14, v15, v14
; SI-NEXT: buffer_store_dword v14, v16, s[0:3], 0 offen
-; SI-NEXT: buffer_load_dword v15, off, s[0:3], s32 offset:204 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v15, off, s[0:3], s32 offset:208 ; 4-byte Folded Reload
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_cvt_f16_f32_e32 v14, v39
+; SI-NEXT: v_cvt_f16_f32_e32 v14, v37
; SI-NEXT: v_add_i32_e32 v16, vcc, 20, v0
; SI-NEXT: v_lshlrev_b32_e32 v14, 16, v14
; SI-NEXT: s_waitcnt vmcnt(0)
; SI-NEXT: v_cvt_f16_f32_e32 v15, v15
; SI-NEXT: v_or_b32_e32 v14, v15, v14
; SI-NEXT: buffer_store_dword v14, v16, s[0:3], 0 offen
+; SI-NEXT: buffer_load_dword v15, off, s[0:3], s32 offset:204 ; 4-byte Folded Reload
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_cvt_f16_f32_e32 v14, v30
-; SI-NEXT: v_cvt_f16_f32_e32 v15, v35
+; SI-NEXT: v_cvt_f16_f32_e32 v14, v17
; SI-NEXT: v_add_i32_e32 v16, vcc, 24, v0
; SI-NEXT: v_lshlrev_b32_e32 v14, 16, v14
+; SI-NEXT: s_waitcnt vmcnt(0)
+; SI-NEXT: v_cvt_f16_f32_e32 v15, v15
; SI-NEXT: v_or_b32_e32 v14, v15, v14
; SI-NEXT: buffer_store_dword v14, v16, s[0:3], 0 offen
+; SI-NEXT: buffer_load_dword v15, off, s[0:3], s32 offset:152 ; 4-byte Folded Reload
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_cvt_f16_f32_e32 v14, v45
-; SI-NEXT: v_cvt_f16_f32_e32 v15, v19
; SI-NEXT: v_add_i32_e32 v16, vcc, 28, v0
; SI-NEXT: v_lshlrev_b32_e32 v14, 16, v14
+; SI-NEXT: s_waitcnt vmcnt(0)
+; SI-NEXT: v_cvt_f16_f32_e32 v15, v15
; SI-NEXT: v_or_b32_e32 v14, v15, v14
; SI-NEXT: buffer_store_dword v14, v16, s[0:3], 0 offen
+; SI-NEXT: buffer_load_dword v15, off, s[0:3], s32 offset:156 ; 4-byte Folded Reload
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_cvt_f16_f32_e32 v14, v17
-; SI-NEXT: v_cvt_f16_f32_e32 v15, v20
+; SI-NEXT: v_cvt_f16_f32_e32 v14, v34
; SI-NEXT: v_add_i32_e32 v16, vcc, 32, v0
; SI-NEXT: v_lshlrev_b32_e32 v14, 16, v14
+; SI-NEXT: s_waitcnt vmcnt(0)
+; SI-NEXT: v_cvt_f16_f32_e32 v15, v15
; SI-NEXT: v_or_b32_e32 v14, v15, v14
; SI-NEXT: buffer_store_dword v14, v16, s[0:3], 0 offen
-; SI-NEXT: buffer_load_dword v15, off, s[0:3], s32 offset:160 ; 4-byte Folded Reload
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_cvt_f16_f32_e32 v14, v21
+; SI-NEXT: v_cvt_f16_f32_e32 v14, v31
+; SI-NEXT: v_cvt_f16_f32_e32 v15, v19
; SI-NEXT: v_add_i32_e32 v16, vcc, 36, v0
; SI-NEXT: v_lshlrev_b32_e32 v14, 16, v14
-; SI-NEXT: s_waitcnt vmcnt(0)
-; SI-NEXT: v_cvt_f16_f32_e32 v15, v15
; SI-NEXT: v_or_b32_e32 v14, v15, v14
; SI-NEXT: buffer_store_dword v14, v16, s[0:3], 0 offen
-; SI-NEXT: buffer_load_dword v15, off, s[0:3], s32 offset:152 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v15, off, s[0:3], s32 offset:160 ; 4-byte Folded Reload
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_cvt_f16_f32_e32 v14, v36
+; SI-NEXT: v_cvt_f16_f32_e32 v14, v26
; SI-NEXT: v_add_i32_e32 v16, vcc, 40, v0
; SI-NEXT: v_lshlrev_b32_e32 v14, 16, v14
; SI-NEXT: s_waitcnt vmcnt(0)
; SI-NEXT: v_cvt_f16_f32_e32 v15, v15
; SI-NEXT: v_or_b32_e32 v14, v15, v14
; SI-NEXT: buffer_store_dword v14, v16, s[0:3], 0 offen
-; SI-NEXT: buffer_load_dword v15, off, s[0:3], s32 offset:156 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v15, off, s[0:3], s32 offset:164 ; 4-byte Folded Reload
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_cvt_f16_f32_e32 v14, v37
+; SI-NEXT: v_cvt_f16_f32_e32 v14, v33
; SI-NEXT: v_add_i32_e32 v16, vcc, 44, v0
; SI-NEXT: v_lshlrev_b32_e32 v14, 16, v14
; SI-NEXT: s_waitcnt vmcnt(0)
@@ -230221,13 +230219,13 @@ define inreg <64 x half> @bitcast_v64bf16_to_v64f16_scalar(<64 x bfloat> inreg %
; SI-NEXT: v_or_b32_e32 v14, v15, v14
; SI-NEXT: buffer_store_dword v14, v16, s[0:3], 0 offen
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_cvt_f16_f32_e32 v14, v38
-; SI-NEXT: v_cvt_f16_f32_e32 v15, v47
+; SI-NEXT: v_cvt_f16_f32_e32 v14, v24
+; SI-NEXT: v_cvt_f16_f32_e32 v15, v21
; SI-NEXT: v_add_i32_e32 v16, vcc, 48, v0
; SI-NEXT: v_lshlrev_b32_e32 v14, 16, v14
; SI-NEXT: v_or_b32_e32 v14, v15, v14
; SI-NEXT: buffer_store_dword v14, v16, s[0:3], 0 offen
-; SI-NEXT: buffer_load_dword v15, off, s[0:3], s32 offset:164 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v15, off, s[0:3], s32 offset:168 ; 4-byte Folded Reload
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_cvt_f16_f32_e32 v14, v59
; SI-NEXT: v_add_i32_e32 v16, vcc, 52, v0
@@ -230236,25 +230234,21 @@ define inreg <64 x half> @bitcast_v64bf16_to_v64f16_scalar(<64 x bfloat> inreg %
; SI-NEXT: v_cvt_f16_f32_e32 v15, v15
; SI-NEXT: v_or_b32_e32 v14, v15, v14
; SI-NEXT: buffer_store_dword v14, v16, s[0:3], 0 offen
-; SI-NEXT: buffer_load_dword v15, off, s[0:3], s32 offset:168 ; 4-byte Folded Reload
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_cvt_f16_f32_e32 v14, v60
+; SI-NEXT: v_cvt_f16_f32_e32 v15, v44
; SI-NEXT: v_add_i32_e32 v16, vcc, 56, v0
; SI-NEXT: v_lshlrev_b32_e32 v14, 16, v14
-; SI-NEXT: s_waitcnt vmcnt(0)
-; SI-NEXT: v_cvt_f16_f32_e32 v15, v15
; SI-NEXT: v_or_b32_e32 v14, v15, v14
; SI-NEXT: buffer_store_dword v14, v16, s[0:3], 0 offen
-; SI-NEXT: buffer_load_dword v15, off, s[0:3], s32 offset:172 ; 4-byte Folded Reload
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_cvt_f16_f32_e32 v14, v61
+; SI-NEXT: v_cvt_f16_f32_e32 v15, v46
; SI-NEXT: v_add_i32_e32 v16, vcc, 60, v0
; SI-NEXT: v_lshlrev_b32_e32 v14, 16, v14
-; SI-NEXT: s_waitcnt vmcnt(0)
-; SI-NEXT: v_cvt_f16_f32_e32 v15, v15
; SI-NEXT: v_or_b32_e32 v14, v15, v14
; SI-NEXT: buffer_store_dword v14, v16, s[0:3], 0 offen
-; SI-NEXT: buffer_load_dword v15, off, s[0:3], s32 offset:180 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v15, off, s[0:3], s32 offset:172 ; 4-byte Folded Reload
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_cvt_f16_f32_e32 v14, v62
; SI-NEXT: v_add_i32_e32 v16, vcc, 64, v0
@@ -230265,14 +230259,16 @@ define inreg <64 x half> @bitcast_v64bf16_to_v64f16_scalar(<64 x bfloat> inreg %
; SI-NEXT: buffer_store_dword v14, v16, s[0:3], 0 offen
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_cvt_f16_f32_e32 v14, v63
-; SI-NEXT: v_cvt_f16_f32_e32 v15, v44
+; SI-NEXT: v_cvt_f16_f32_e32 v15, v47
; SI-NEXT: v_add_i32_e32 v16, vcc, 0x44, v0
; SI-NEXT: v_lshlrev_b32_e32 v14, 16, v14
; SI-NEXT: v_or_b32_e32 v14, v15, v14
; SI-NEXT: buffer_store_dword v14, v16, s[0:3], 0 offen
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_cvt_f16_f32_e32 v14, v46
+; SI-NEXT: buffer_load_dword v14, off, s[0:3], s32 offset:176 ; 4-byte Folded Reload
; SI-NEXT: v_add_i32_e32 v15, vcc, 0x48, v0
+; SI-NEXT: s_waitcnt vmcnt(0)
+; SI-NEXT: v_cvt_f16_f32_e32 v14, v14
; SI-NEXT: v_or_b32_e32 v1, v14, v1
; SI-NEXT: buffer_store_dword v1, v15, s[0:3], 0 offen
; SI-NEXT: s_waitcnt expcnt(0)
@@ -230282,21 +230278,23 @@ define inreg <64 x half> @bitcast_v64bf16_to_v64f16_scalar(<64 x bfloat> inreg %
; SI-NEXT: v_lshlrev_b32_e32 v1, 16, v1
; SI-NEXT: v_or_b32_e32 v1, v2, v1
; SI-NEXT: buffer_store_dword v1, v14, s[0:3], 0 offen
+; SI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:180 ; 4-byte Folded Reload
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_cvt_f16_f32_e32 v1, v3
-; SI-NEXT: v_cvt_f16_f32_e32 v2, v57
; SI-NEXT: v_add_i32_e32 v3, vcc, 0x50, v0
; SI-NEXT: v_lshlrev_b32_e32 v1, 16, v1
+; SI-NEXT: s_waitcnt vmcnt(0)
+; SI-NEXT: v_cvt_f16_f32_e32 v2, v2
; SI-NEXT: v_or_b32_e32 v1, v2, v1
; SI-NEXT: buffer_store_dword v1, v3, s[0:3], 0 offen
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_cvt_f16_f32_e32 v1, v4
-; SI-NEXT: v_cvt_f16_f32_e32 v2, v58
+; SI-NEXT: v_cvt_f16_f32_e32 v2, v57
; SI-NEXT: v_add_i32_e32 v3, vcc, 0x54, v0
; SI-NEXT: v_lshlrev_b32_e32 v1, 16, v1
; SI-NEXT: v_or_b32_e32 v1, v2, v1
; SI-NEXT: buffer_store_dword v1, v3, s[0:3], 0 offen
-; SI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:176 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:184 ; 4-byte Folded Reload
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_cvt_f16_f32_e32 v1, v5
; SI-NEXT: v_add_i32_e32 v3, vcc, 0x58, v0
@@ -230305,13 +230303,11 @@ define inreg <64 x half> @bitcast_v64bf16_to_v64f16_scalar(<64 x bfloat> inreg %
; SI-NEXT: v_cvt_f16_f32_e32 v2, v2
; SI-NEXT: v_or_b32_e32 v1, v2, v1
; SI-NEXT: buffer_store_dword v1, v3, s[0:3], 0 offen
-; SI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:184 ; 4-byte Folded Reload
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_cvt_f16_f32_e32 v1, v6
+; SI-NEXT: v_cvt_f16_f32_e32 v2, v48
; SI-NEXT: v_add_i32_e32 v3, vcc, 0x5c, v0
; SI-NEXT: v_lshlrev_b32_e32 v1, 16, v1
-; SI-NEXT: s_waitcnt vmcnt(0)
-; SI-NEXT: v_cvt_f16_f32_e32 v2, v2
; SI-NEXT: v_or_b32_e32 v1, v2, v1
; SI-NEXT: buffer_store_dword v1, v3, s[0:3], 0 offen
; SI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:188 ; 4-byte Folded Reload
@@ -230323,20 +230319,20 @@ define inreg <64 x half> @bitcast_v64bf16_to_v64f16_scalar(<64 x bfloat> inreg %
; SI-NEXT: v_cvt_f16_f32_e32 v2, v2
; SI-NEXT: v_or_b32_e32 v1, v2, v1
; SI-NEXT: buffer_store_dword v1, v3, s[0:3], 0 offen
-; SI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:192 ; 4-byte Folded Reload
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_cvt_f16_f32_e32 v1, v8
+; SI-NEXT: v_cvt_f16_f32_e32 v2, v58
; SI-NEXT: v_add_i32_e32 v3, vcc, 0x64, v0
; SI-NEXT: v_lshlrev_b32_e32 v1, 16, v1
-; SI-NEXT: s_waitcnt vmcnt(0)
-; SI-NEXT: v_cvt_f16_f32_e32 v2, v2
; SI-NEXT: v_or_b32_e32 v1, v2, v1
; SI-NEXT: buffer_store_dword v1, v3, s[0:3], 0 offen
+; SI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:192 ; 4-byte Folded Reload
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_cvt_f16_f32_e32 v1, v9
-; SI-NEXT: v_cvt_f16_f32_e32 v2, v32
; SI-NEXT: v_add_i32_e32 v3, vcc, 0x68, v0
; SI-NEXT: v_lshlrev_b32_e32 v1, 16, v1
+; SI-NEXT: s_waitcnt vmcnt(0)
+; SI-NEXT: v_cvt_f16_f32_e32 v2, v2
; SI-NEXT: v_or_b32_e32 v1, v2, v1
; SI-NEXT: buffer_store_dword v1, v3, s[0:3], 0 offen
; SI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:196 ; 4-byte Folded Reload
@@ -230348,33 +230344,35 @@ define inreg <64 x half> @bitcast_v64bf16_to_v64f16_scalar(<64 x bfloat> inreg %
; SI-NEXT: v_cvt_f16_f32_e32 v2, v2
; SI-NEXT: v_or_b32_e32 v1, v2, v1
; SI-NEXT: buffer_store_dword v1, v3, s[0:3], 0 offen
+; SI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:200 ; 4-byte Folded Reload
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_cvt_f16_f32_e32 v1, v11
-; SI-NEXT: v_cvt_f16_f32_e32 v2, v33
; SI-NEXT: v_add_i32_e32 v3, vcc, 0x70, v0
; SI-NEXT: v_lshlrev_b32_e32 v1, 16, v1
+; SI-NEXT: s_waitcnt vmcnt(0)
+; SI-NEXT: v_cvt_f16_f32_e32 v2, v2
; SI-NEXT: v_or_b32_e32 v1, v2, v1
; SI-NEXT: buffer_store_dword v1, v3, s[0:3], 0 offen
-; SI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:200 ; 4-byte Folded Reload
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_cvt_f16_f32_e32 v1, v12
+; SI-NEXT: v_cvt_f16_f32_e32 v2, v54
; SI-NEXT: v_add_i32_e32 v3, vcc, 0x74, v0
; SI-NEXT: v_lshlrev_b32_e32 v1, 16, v1
-; SI-NEXT: s_waitcnt vmcnt(0)
-; SI-NEXT: v_cvt_f16_f32_e32 v2, v2
; SI-NEXT: v_or_b32_e32 v1, v2, v1
; SI-NEXT: buffer_store_dword v1, v3, s[0:3], 0 offen
+; SI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:232 ; 4-byte Folded Reload
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_cvt_f16_f32_e32 v1, v13
-; SI-NEXT: v_cvt_f16_f32_e32 v2, v34
; SI-NEXT: v_add_i32_e32 v3, vcc, 0x78, v0
+; SI-NEXT: v_add_i32_e32 v0, vcc, 0x7c, v0
; SI-NEXT: v_lshlrev_b32_e32 v1, 16, v1
+; SI-NEXT: s_waitcnt vmcnt(0)
+; SI-NEXT: v_cvt_f16_f32_e32 v2, v2
; SI-NEXT: v_or_b32_e32 v1, v2, v1
; SI-NEXT: buffer_store_dword v1, v3, s[0:3], 0 offen
; SI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:148 ; 4-byte Folded Reload
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_cvt_f16_f32_e32 v1, v43
-; SI-NEXT: v_add_i32_e32 v0, vcc, 0x7c, v0
; SI-NEXT: v_lshlrev_b32_e32 v1, 16, v1
; SI-NEXT: s_waitcnt vmcnt(0)
; SI-NEXT: v_cvt_f16_f32_e32 v2, v2
@@ -230399,20 +230397,17 @@ define inreg <64 x half> @bitcast_v64bf16_to_v64f16_scalar(<64 x bfloat> inreg %
; SI-NEXT: s_waitcnt vmcnt(0) expcnt(0)
; SI-NEXT: s_setpc_b64 s[30:31]
; SI-NEXT: .LBB101_4:
-; SI-NEXT: buffer_store_dword v53, off, s[0:3], s32 offset:432 ; 4-byte Folded Spill
+; SI-NEXT: v_mov_b32_e32 v55, v53
+; SI-NEXT: v_mov_b32_e32 v30, v14
+; SI-NEXT: buffer_load_dword v14, off, s[0:3], s32 offset:236 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v49, off, s[0:3], s32 offset:240 ; 4-byte Folded Reload
+; SI-NEXT: s_waitcnt expcnt(2)
+; SI-NEXT: buffer_load_dword v39, off, s[0:3], s32 offset:244 ; 4-byte Folded Reload
+; SI-NEXT: s_waitcnt expcnt(1)
+; SI-NEXT: buffer_load_dword v51, off, s[0:3], s32 offset:248 ; 4-byte Folded Reload
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_mov_b32_e32 v53, v54
-; SI-NEXT: v_mov_b32_e32 v40, v52
-; SI-NEXT: v_mov_b32_e32 v55, v50
-; SI-NEXT: v_mov_b32_e32 v42, v51
-; SI-NEXT: v_mov_b32_e32 v28, v14
-; SI-NEXT: buffer_load_dword v14, off, s[0:3], s32 offset:232 ; 4-byte Folded Reload
-; SI-NEXT: buffer_load_dword v22, off, s[0:3], s32 offset:236 ; 4-byte Folded Reload
-; SI-NEXT: buffer_load_dword v41, off, s[0:3], s32 offset:248 ; 4-byte Folded Reload
-; SI-NEXT: buffer_load_dword v51, off, s[0:3], s32 offset:432 ; 4-byte Folded Reload
-; SI-NEXT: buffer_load_dword v50, off, s[0:3], s32 offset:240 ; 4-byte Folded Reload
-; SI-NEXT: buffer_load_dword v52, off, s[0:3], s32 offset:244 ; 4-byte Folded Reload
-; SI-NEXT: buffer_load_dword v54, off, s[0:3], s32 offset:252 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v52, off, s[0:3], s32 offset:252 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v53, off, s[0:3], s32 offset:256 ; 4-byte Folded Reload
; SI-NEXT: ; implicit-def: $vgpr1
; SI-NEXT: ; kill: killed $vgpr1
; SI-NEXT: ; implicit-def: $vgpr1
@@ -230436,70 +230431,75 @@ define inreg <64 x half> @bitcast_v64bf16_to_v64f16_scalar(<64 x bfloat> inreg %
; SI-NEXT: ; implicit-def: $vgpr1
; SI-NEXT: ; kill: killed $vgpr1
; SI-NEXT: ; implicit-def: $vgpr1
-; SI-NEXT: ; kill: killed $vgpr1
-; SI-NEXT: ; implicit-def: $vgpr1
-; SI-NEXT: v_mov_b32_e32 v27, v58
-; SI-NEXT: v_mov_b32_e32 v26, v57
+; SI-NEXT: v_mov_b32_e32 v40, v50
+; SI-NEXT: v_mov_b32_e32 v42, v41
+; SI-NEXT: v_mov_b32_e32 v36, v54
+; SI-NEXT: v_mov_b32_e32 v29, v58
+; SI-NEXT: v_mov_b32_e32 v28, v26
+; SI-NEXT: v_mov_b32_e32 v27, v57
; SI-NEXT: v_mov_b32_e32 v25, v56
-; SI-NEXT: v_mov_b32_e32 v24, v47
; SI-NEXT: v_mov_b32_e32 v23, v46
+; SI-NEXT: v_mov_b32_e32 v22, v21
+; SI-NEXT: v_mov_b32_e32 v20, v44
; SI-NEXT: ; kill: killed $vgpr1
; SI-NEXT: ; implicit-def: $vgpr1
+; SI-NEXT: ; implicit-def: $vgpr3
; SI-NEXT: ; implicit-def: $vgpr5
-; SI-NEXT: ; implicit-def: $vgpr6
; SI-NEXT: ; implicit-def: $vgpr7
-; SI-NEXT: ; implicit-def: $vgpr8
+; SI-NEXT: ; implicit-def: $vgpr9
; SI-NEXT: ; implicit-def: $vgpr10
-; SI-NEXT: ; implicit-def: $vgpr12
+; SI-NEXT: ; implicit-def: $vgpr11
+; SI-NEXT: ; implicit-def: $vgpr13
; SI-NEXT: ; implicit-def: $vgpr43
+; SI-NEXT: ; implicit-def: $vgpr35
+; SI-NEXT: ; implicit-def: $vgpr38
; SI-NEXT: ; implicit-def: $vgpr15
-; SI-NEXT: ; implicit-def: $vgpr49
; SI-NEXT: ; implicit-def: $vgpr16
+; SI-NEXT: ; implicit-def: $vgpr32
; SI-NEXT: ; implicit-def: $vgpr18
-; SI-NEXT: ; implicit-def: $vgpr39
-; SI-NEXT: ; implicit-def: $vgpr35
-; SI-NEXT: ; implicit-def: $vgpr30
-; SI-NEXT: ; implicit-def: $vgpr19
-; SI-NEXT: ; implicit-def: $vgpr45
-; SI-NEXT: ; implicit-def: $vgpr20
+; SI-NEXT: ; implicit-def: $vgpr37
; SI-NEXT: ; implicit-def: $vgpr17
+; SI-NEXT: ; implicit-def: $vgpr45
+; SI-NEXT: ; implicit-def: $vgpr34
+; SI-NEXT: ; implicit-def: $vgpr19
+; SI-NEXT: ; implicit-def: $vgpr31
+; SI-NEXT: ; implicit-def: $vgpr26
+; SI-NEXT: ; implicit-def: $vgpr33
; SI-NEXT: ; implicit-def: $vgpr21
-; SI-NEXT: ; implicit-def: $vgpr36
-; SI-NEXT: ; implicit-def: $vgpr37
-; SI-NEXT: ; implicit-def: $vgpr47
-; SI-NEXT: ; implicit-def: $vgpr38
+; SI-NEXT: ; implicit-def: $vgpr24
; SI-NEXT: ; implicit-def: $vgpr59
+; SI-NEXT: ; implicit-def: $vgpr44
; SI-NEXT: ; implicit-def: $vgpr60
+; SI-NEXT: ; implicit-def: $vgpr46
; SI-NEXT: ; implicit-def: $vgpr61
-; SI-NEXT: ; kill: killed $vgpr1
; SI-NEXT: ; implicit-def: $vgpr62
-; SI-NEXT: ; implicit-def: $vgpr44
+; SI-NEXT: ; implicit-def: $vgpr47
; SI-NEXT: ; implicit-def: $vgpr63
-; SI-NEXT: ; implicit-def: $vgpr46
+; SI-NEXT: ; kill: killed $vgpr1
; SI-NEXT: ; implicit-def: $vgpr1
; SI-NEXT: ; implicit-def: $vgpr56
; SI-NEXT: ; implicit-def: $vgpr2
-; SI-NEXT: ; implicit-def: $vgpr57
+; SI-NEXT: ; kill: killed $vgpr3
; SI-NEXT: ; implicit-def: $vgpr3
-; SI-NEXT: ; implicit-def: $vgpr58
+; SI-NEXT: ; implicit-def: $vgpr57
; SI-NEXT: ; implicit-def: $vgpr4
; SI-NEXT: ; kill: killed $vgpr5
; SI-NEXT: ; implicit-def: $vgpr5
-; SI-NEXT: ; kill: killed $vgpr6
+; SI-NEXT: ; implicit-def: $vgpr48
; SI-NEXT: ; implicit-def: $vgpr6
; SI-NEXT: ; kill: killed $vgpr7
; SI-NEXT: ; implicit-def: $vgpr7
-; SI-NEXT: ; kill: killed $vgpr8
+; SI-NEXT: ; implicit-def: $vgpr58
; SI-NEXT: ; implicit-def: $vgpr8
-; SI-NEXT: ; implicit-def: $vgpr32
+; SI-NEXT: ; kill: killed $vgpr9
; SI-NEXT: ; implicit-def: $vgpr9
; SI-NEXT: ; kill: killed $vgpr10
; SI-NEXT: ; implicit-def: $vgpr10
-; SI-NEXT: ; implicit-def: $vgpr33
+; SI-NEXT: ; kill: killed $vgpr11
; SI-NEXT: ; implicit-def: $vgpr11
-; SI-NEXT: ; kill: killed $vgpr12
+; SI-NEXT: ; implicit-def: $vgpr54
; SI-NEXT: ; implicit-def: $vgpr12
-; SI-NEXT: ; implicit-def: $vgpr34
+; SI-NEXT: ; kill: killed $vgpr13
; SI-NEXT: ; implicit-def: $vgpr13
; SI-NEXT: ; kill: killed $vgpr43
; SI-NEXT: ; implicit-def: $vgpr43
@@ -233773,10 +233773,11 @@ define <64 x bfloat> @bitcast_v64f16_to_v64bf16(<64 x half> %a, i32 %b) {
; SI-NEXT: v_cvt_f32_f16_e32 v63, v1
; SI-NEXT: v_cvt_f32_f16_e32 v1, v62
; SI-NEXT: v_add_f32_e32 v2, 0x38000000, v2
-; SI-NEXT: buffer_store_dword v2, off, s[0:3], s32 offset:440 ; 4-byte Folded Spill
-; SI-NEXT: v_add_f32_e32 v62, 0x38000000, v63
+; SI-NEXT: v_add_f32_e32 v62, 0x38000000, v31
+; SI-NEXT: v_add_f32_e32 v31, 0x38000000, v63
; SI-NEXT: v_add_f32_e32 v63, 0x38000000, v1
; SI-NEXT: v_cvt_f32_f16_e32 v1, v61
+; SI-NEXT: buffer_store_dword v2, off, s[0:3], s32 offset:440 ; 4-byte Folded Spill
; SI-NEXT: v_cvt_f32_f16_e32 v60, v60
; SI-NEXT: v_cvt_f32_f16_e32 v61, v59
; SI-NEXT: v_cvt_f32_f16_e32 v58, v58
@@ -233897,7 +233898,6 @@ define <64 x bfloat> @bitcast_v64f16_to_v64bf16(<64 x half> %a, i32 %b) {
; SI-NEXT: v_add_f32_e32 v23, 0x38000000, v23
; SI-NEXT: v_cvt_f16_f32_e32 v1, v1
; SI-NEXT: v_cvt_f16_f32_e32 v2, v2
-; SI-NEXT: v_add_f32_e32 v31, 0x38000000, v31
; SI-NEXT: v_lshlrev_b32_e32 v1, 16, v1
; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:292 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
@@ -234138,9 +234138,9 @@ define <64 x bfloat> @bitcast_v64f16_to_v64bf16(<64 x half> %a, i32 %b) {
; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:436 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_cvt_f16_f32_e32 v1, v63
-; SI-NEXT: v_cvt_f16_f32_e32 v2, v62
+; SI-NEXT: v_cvt_f16_f32_e32 v2, v31
; SI-NEXT: v_add_f32_e32 v58, 0x38000000, v58
-; SI-NEXT: v_cvt_f16_f32_e32 v3, v31
+; SI-NEXT: v_cvt_f16_f32_e32 v3, v62
; SI-NEXT: v_cvt_f16_f32_e32 v58, v58
; SI-NEXT: v_lshlrev_b32_e32 v1, 16, v1
; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:440 ; 4-byte Folded Spill
@@ -234806,16 +234806,15 @@ define inreg <64 x bfloat> @bitcast_v64f16_to_v64bf16_scalar(<64 x half> inreg %
; SI-NEXT: buffer_load_dword v50, off, s[0:3], s32 offset:72
; SI-NEXT: buffer_load_dword v51, off, s[0:3], s32 offset:76
; SI-NEXT: v_cvt_f16_f32_e32 v40, v1
-; SI-NEXT: v_cvt_f16_f32_e32 v1, v8
+; SI-NEXT: v_cvt_f16_f32_e32 v57, v6
; SI-NEXT: v_mov_b32_e32 v46, v26
; SI-NEXT: v_cvt_f16_f32_e32 v43, v2
-; SI-NEXT: v_cvt_f16_f32_e32 v3, v3
+; SI-NEXT: v_cvt_f16_f32_e32 v56, v3
; SI-NEXT: v_cvt_f16_f32_e32 v4, v4
; SI-NEXT: v_cvt_f16_f32_e32 v5, v5
-; SI-NEXT: v_cvt_f16_f32_e32 v6, v6
; SI-NEXT: v_cvt_f16_f32_e32 v7, v7
-; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:284 ; 4-byte Folded Spill
-; SI-NEXT: v_cvt_f16_f32_e32 v9, v9
+; SI-NEXT: v_cvt_f16_f32_e32 v44, v8
+; SI-NEXT: v_cvt_f16_f32_e32 v59, v9
; SI-NEXT: v_cvt_f16_f32_e32 v10, v10
; SI-NEXT: v_cvt_f16_f32_e32 v11, v11
; SI-NEXT: v_cvt_f16_f32_e32 v12, v12
@@ -234828,588 +234827,599 @@ define inreg <64 x bfloat> @bitcast_v64f16_to_v64bf16_scalar(<64 x half> inreg %
; SI-NEXT: v_cvt_f16_f32_e32 v20, v20
; SI-NEXT: v_cvt_f16_f32_e32 v21, v21
; SI-NEXT: v_cvt_f16_f32_e32 v22, v22
+; SI-NEXT: v_cvt_f16_f32_e32 v26, v23
; SI-NEXT: v_cvt_f16_f32_e32 v45, v24
-; SI-NEXT: v_cvt_f16_f32_e32 v26, v25
; SI-NEXT: v_cvt_f16_f32_e32 v46, v46
-; SI-NEXT: v_cvt_f16_f32_e32 v47, v27
-; SI-NEXT: v_cvt_f16_f32_e32 v28, v28
-; SI-NEXT: v_cvt_f16_f32_e32 v56, v29
-; SI-NEXT: v_cvt_f16_f32_e32 v57, v30
+; SI-NEXT: v_cvt_f16_f32_e32 v58, v27
+; SI-NEXT: v_cvt_f16_f32_e32 v60, v28
+; SI-NEXT: v_cvt_f16_f32_e32 v28, v29
+; SI-NEXT: v_cvt_f16_f32_e32 v30, v30
; SI-NEXT: v_cvt_f16_f32_e32 v8, s16
+; SI-NEXT: v_cvt_f16_f32_e32 v23, s17
; SI-NEXT: v_cvt_f16_f32_e32 v24, s18
-; SI-NEXT: v_cvt_f16_f32_e32 v25, s19
; SI-NEXT: v_cvt_f16_f32_e32 v29, s20
-; SI-NEXT: v_cvt_f16_f32_e32 v30, s21
-; SI-NEXT: v_cvt_f16_f32_e32 v27, s24
+; SI-NEXT: v_cvt_f16_f32_e32 v9, s21
+; SI-NEXT: v_cvt_f16_f32_e32 v27, s23
; SI-NEXT: s_waitcnt vmcnt(14)
; SI-NEXT: v_cmp_ne_u32_e32 vcc, 0, v31
; SI-NEXT: v_cvt_f16_f32_e32 v31, v16
-; SI-NEXT: v_cvt_f16_f32_e32 v16, v23
-; SI-NEXT: v_cvt_f16_f32_e32 v32, v32
+; SI-NEXT: v_cvt_f16_f32_e32 v16, v25
+; SI-NEXT: v_cvt_f16_f32_e32 v2, v32
+; SI-NEXT: v_cvt_f16_f32_e32 v1, v35
; SI-NEXT: v_cvt_f16_f32_e32 v33, v33
; SI-NEXT: v_cvt_f16_f32_e32 v34, v34
-; SI-NEXT: v_cvt_f16_f32_e32 v58, v35
; SI-NEXT: v_cvt_f16_f32_e32 v36, v36
-; SI-NEXT: v_cvt_f16_f32_e32 v59, v37
-; SI-NEXT: v_cvt_f16_f32_e32 v60, v38
+; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:280 ; 4-byte Folded Spill
+; SI-NEXT: s_waitcnt vmcnt(14) expcnt(0)
+; SI-NEXT: v_cvt_f16_f32_e32 v1, v38
+; SI-NEXT: v_cvt_f16_f32_e32 v37, v37
; SI-NEXT: s_waitcnt vmcnt(13)
-; SI-NEXT: v_cvt_f16_f32_e32 v39, v39
-; SI-NEXT: s_waitcnt vmcnt(12)
-; SI-NEXT: v_cvt_f16_f32_e32 v61, v48
-; SI-NEXT: s_waitcnt vmcnt(11) expcnt(0)
-; SI-NEXT: v_cvt_f16_f32_e32 v1, v49
-; SI-NEXT: s_waitcnt vmcnt(10)
+; SI-NEXT: v_cvt_f16_f32_e32 v47, v39
+; SI-NEXT: s_waitcnt vmcnt(11)
+; SI-NEXT: v_cvt_f16_f32_e32 v3, v49
+; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:284 ; 4-byte Folded Spill
+; SI-NEXT: s_waitcnt expcnt(0)
+; SI-NEXT: v_cvt_f16_f32_e32 v1, v48
+; SI-NEXT: s_waitcnt vmcnt(11)
; SI-NEXT: v_cvt_f16_f32_e32 v52, v52
-; SI-NEXT: s_waitcnt vmcnt(9)
+; SI-NEXT: s_waitcnt vmcnt(10)
; SI-NEXT: v_cvt_f16_f32_e32 v53, v53
-; SI-NEXT: s_waitcnt vmcnt(8)
+; SI-NEXT: s_waitcnt vmcnt(9)
; SI-NEXT: v_cvt_f16_f32_e32 v54, v54
-; SI-NEXT: s_waitcnt vmcnt(7)
+; SI-NEXT: s_waitcnt vmcnt(8)
; SI-NEXT: v_cvt_f16_f32_e32 v55, v55
-; SI-NEXT: s_waitcnt vmcnt(6)
+; SI-NEXT: s_waitcnt vmcnt(3)
+; SI-NEXT: v_cvt_f16_f32_e32 v6, v50
; SI-NEXT: v_cvt_f16_f32_e32 v41, v41
-; SI-NEXT: s_waitcnt vmcnt(5)
; SI-NEXT: v_cvt_f16_f32_e32 v42, v42
-; SI-NEXT: s_waitcnt vmcnt(4)
; SI-NEXT: v_cvt_f16_f32_e32 v62, v62
-; SI-NEXT: s_waitcnt vmcnt(3)
; SI-NEXT: v_cvt_f16_f32_e32 v63, v63
-; SI-NEXT: s_waitcnt vmcnt(2)
-; SI-NEXT: v_cvt_f16_f32_e32 v2, v50
-; SI-NEXT: s_waitcnt vmcnt(1)
-; SI-NEXT: v_cvt_f16_f32_e32 v44, v51
-; SI-NEXT: v_cvt_f16_f32_e32 v23, s17
-; SI-NEXT: v_cvt_f16_f32_e32 v38, s22
-; SI-NEXT: v_cvt_f16_f32_e32 v37, s23
-; SI-NEXT: v_cvt_f16_f32_e32 v48, s25
-; SI-NEXT: v_cvt_f16_f32_e32 v49, s26
-; SI-NEXT: v_cvt_f16_f32_e32 v35, s27
-; SI-NEXT: v_cvt_f16_f32_e32 v50, s28
-; SI-NEXT: v_cvt_f16_f32_e32 v51, s29
+; SI-NEXT: buffer_store_dword v6, off, s[0:3], s32 offset:288 ; 4-byte Folded Spill
+; SI-NEXT: s_waitcnt vmcnt(3)
+; SI-NEXT: v_cvt_f16_f32_e32 v50, v51
+; SI-NEXT: v_cvt_f16_f32_e32 v25, s19
+; SI-NEXT: s_waitcnt expcnt(0)
+; SI-NEXT: v_cvt_f16_f32_e32 v6, s22
+; SI-NEXT: v_cvt_f16_f32_e32 v35, s24
+; SI-NEXT: v_cvt_f16_f32_e32 v38, s25
+; SI-NEXT: v_cvt_f16_f32_e32 v48, s26
+; SI-NEXT: v_cvt_f16_f32_e32 v49, s27
+; SI-NEXT: v_cvt_f16_f32_e32 v32, s28
+; SI-NEXT: v_cvt_f16_f32_e32 v39, s29
; SI-NEXT: s_and_b64 s[4:5], vcc, exec
-; SI-NEXT: buffer_store_dword v40, off, s[0:3], s32 offset:288 ; 4-byte Folded Spill
-; SI-NEXT: buffer_store_dword v43, off, s[0:3], s32 offset:292 ; 4-byte Folded Spill
-; SI-NEXT: buffer_store_dword v3, off, s[0:3], s32 offset:296 ; 4-byte Folded Spill
-; SI-NEXT: buffer_store_dword v4, off, s[0:3], s32 offset:300 ; 4-byte Folded Spill
-; SI-NEXT: buffer_store_dword v5, off, s[0:3], s32 offset:304 ; 4-byte Folded Spill
-; SI-NEXT: buffer_store_dword v7, off, s[0:3], s32 offset:308 ; 4-byte Folded Spill
-; SI-NEXT: buffer_store_dword v10, off, s[0:3], s32 offset:312 ; 4-byte Folded Spill
-; SI-NEXT: buffer_store_dword v11, off, s[0:3], s32 offset:316 ; 4-byte Folded Spill
-; SI-NEXT: buffer_store_dword v12, off, s[0:3], s32 offset:320 ; 4-byte Folded Spill
-; SI-NEXT: buffer_store_dword v14, off, s[0:3], s32 offset:324 ; 4-byte Folded Spill
-; SI-NEXT: buffer_store_dword v15, off, s[0:3], s32 offset:328 ; 4-byte Folded Spill
-; SI-NEXT: buffer_store_dword v17, off, s[0:3], s32 offset:332 ; 4-byte Folded Spill
-; SI-NEXT: buffer_store_dword v8, off, s[0:3], s32 offset:336 ; 4-byte Folded Spill
-; SI-NEXT: buffer_store_dword v23, off, s[0:3], s32 offset:340 ; 4-byte Folded Spill
-; SI-NEXT: buffer_store_dword v24, off, s[0:3], s32 offset:344 ; 4-byte Folded Spill
-; SI-NEXT: buffer_store_dword v18, off, s[0:3], s32 offset:348 ; 4-byte Folded Spill
-; SI-NEXT: buffer_store_dword v25, off, s[0:3], s32 offset:352 ; 4-byte Folded Spill
-; SI-NEXT: buffer_store_dword v20, off, s[0:3], s32 offset:356 ; 4-byte Folded Spill
-; SI-NEXT: buffer_store_dword v29, off, s[0:3], s32 offset:360 ; 4-byte Folded Spill
-; SI-NEXT: buffer_store_dword v30, off, s[0:3], s32 offset:364 ; 4-byte Folded Spill
-; SI-NEXT: buffer_store_dword v21, off, s[0:3], s32 offset:368 ; 4-byte Folded Spill
-; SI-NEXT: buffer_store_dword v38, off, s[0:3], s32 offset:372 ; 4-byte Folded Spill
-; SI-NEXT: buffer_store_dword v37, off, s[0:3], s32 offset:376 ; 4-byte Folded Spill
-; SI-NEXT: buffer_store_dword v27, off, s[0:3], s32 offset:380 ; 4-byte Folded Spill
-; SI-NEXT: buffer_store_dword v48, off, s[0:3], s32 offset:384 ; 4-byte Folded Spill
-; SI-NEXT: buffer_store_dword v49, off, s[0:3], s32 offset:388 ; 4-byte Folded Spill
-; SI-NEXT: buffer_store_dword v35, off, s[0:3], s32 offset:392 ; 4-byte Folded Spill
-; SI-NEXT: buffer_store_dword v50, off, s[0:3], s32 offset:396 ; 4-byte Folded Spill
-; SI-NEXT: buffer_store_dword v51, off, s[0:3], s32 offset:400 ; 4-byte Folded Spill
-; SI-NEXT: buffer_store_dword v57, off, s[0:3], s32 offset:404 ; 4-byte Folded Spill
-; SI-NEXT: buffer_store_dword v32, off, s[0:3], s32 offset:408 ; 4-byte Folded Spill
-; SI-NEXT: buffer_store_dword v33, off, s[0:3], s32 offset:412 ; 4-byte Folded Spill
-; SI-NEXT: buffer_store_dword v34, off, s[0:3], s32 offset:416 ; 4-byte Folded Spill
-; SI-NEXT: buffer_store_dword v58, off, s[0:3], s32 offset:420 ; 4-byte Folded Spill
-; SI-NEXT: buffer_store_dword v36, off, s[0:3], s32 offset:424 ; 4-byte Folded Spill
-; SI-NEXT: buffer_store_dword v44, off, s[0:3], s32 offset:280 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v40, off, s[0:3], s32 offset:292 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v43, off, s[0:3], s32 offset:296 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v56, off, s[0:3], s32 offset:300 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v4, off, s[0:3], s32 offset:304 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v5, off, s[0:3], s32 offset:308 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v7, off, s[0:3], s32 offset:312 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v10, off, s[0:3], s32 offset:316 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v11, off, s[0:3], s32 offset:320 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v12, off, s[0:3], s32 offset:324 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v14, off, s[0:3], s32 offset:328 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v8, off, s[0:3], s32 offset:332 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v23, off, s[0:3], s32 offset:336 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v24, off, s[0:3], s32 offset:340 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v25, off, s[0:3], s32 offset:344 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v15, off, s[0:3], s32 offset:348 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v31, off, s[0:3], s32 offset:352 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v29, off, s[0:3], s32 offset:356 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v9, off, s[0:3], s32 offset:360 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v6, off, s[0:3], s32 offset:364 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v27, off, s[0:3], s32 offset:368 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v35, off, s[0:3], s32 offset:372 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v38, off, s[0:3], s32 offset:376 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v48, off, s[0:3], s32 offset:380 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v49, off, s[0:3], s32 offset:384 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v32, off, s[0:3], s32 offset:388 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v39, off, s[0:3], s32 offset:392 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v45, off, s[0:3], s32 offset:396 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v60, off, s[0:3], s32 offset:400 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v2, off, s[0:3], s32 offset:404 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v33, off, s[0:3], s32 offset:408 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v36, off, s[0:3], s32 offset:412 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v47, off, s[0:3], s32 offset:416 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:420 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v3, off, s[0:3], s32 offset:424 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v50, off, s[0:3], s32 offset:428 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v58, off, s[0:3], s32 offset:432 ; 4-byte Folded Spill
; SI-NEXT: s_cbranch_scc0 .LBB103_2
; SI-NEXT: ; %bb.1: ; %cmp.false
+; SI-NEXT: v_lshlrev_b32_e32 v4, 16, v4
; SI-NEXT: v_lshlrev_b32_e32 v8, 16, v8
+; SI-NEXT: buffer_store_dword v4, off, s[0:3], s32 offset:212 ; 4-byte Folded Spill
+; SI-NEXT: s_waitcnt expcnt(0)
+; SI-NEXT: v_lshlrev_b32_e32 v4, 16, v5
+; SI-NEXT: v_lshlrev_b32_e32 v5, 16, v57
; SI-NEXT: buffer_store_dword v8, off, s[0:3], s32 offset:148 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_lshlrev_b32_e32 v8, 16, v23
+; SI-NEXT: buffer_store_dword v5, off, s[0:3], s32 offset:216 ; 4-byte Folded Spill
+; SI-NEXT: s_waitcnt expcnt(0)
+; SI-NEXT: v_lshlrev_b32_e32 v5, 16, v7
; SI-NEXT: buffer_store_dword v8, off, s[0:3], s32 offset:156 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_lshlrev_b32_e32 v8, 16, v24
+; SI-NEXT: buffer_store_dword v5, off, s[0:3], s32 offset:220 ; 4-byte Folded Spill
+; SI-NEXT: s_waitcnt expcnt(0)
+; SI-NEXT: v_lshlrev_b32_e32 v5, 16, v44
; SI-NEXT: buffer_store_dword v8, off, s[0:3], s32 offset:152 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_lshlrev_b32_e32 v8, 16, v25
+; SI-NEXT: buffer_store_dword v5, off, s[0:3], s32 offset:228 ; 4-byte Folded Spill
+; SI-NEXT: s_waitcnt expcnt(0)
+; SI-NEXT: v_lshlrev_b32_e32 v5, 16, v59
; SI-NEXT: buffer_store_dword v8, off, s[0:3], s32 offset:160 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_lshlrev_b32_e32 v8, 16, v29
+; SI-NEXT: buffer_store_dword v5, off, s[0:3], s32 offset:224 ; 4-byte Folded Spill
+; SI-NEXT: s_waitcnt expcnt(0)
+; SI-NEXT: v_lshlrev_b32_e32 v5, 16, v10
; SI-NEXT: buffer_store_dword v8, off, s[0:3], s32 offset:164 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_lshlrev_b32_e32 v8, 16, v30
+; SI-NEXT: v_lshlrev_b32_e32 v8, 16, v9
+; SI-NEXT: buffer_store_dword v5, off, s[0:3], s32 offset:232 ; 4-byte Folded Spill
+; SI-NEXT: s_waitcnt expcnt(0)
+; SI-NEXT: v_lshlrev_b32_e32 v5, 16, v11
; SI-NEXT: buffer_store_dword v8, off, s[0:3], s32 offset:172 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_lshlrev_b32_e32 v8, 16, v38
+; SI-NEXT: v_lshlrev_b32_e32 v8, 16, v6
+; SI-NEXT: buffer_store_dword v5, off, s[0:3], s32 offset:236 ; 4-byte Folded Spill
+; SI-NEXT: s_waitcnt expcnt(0)
+; SI-NEXT: v_lshlrev_b32_e32 v5, 16, v12
; SI-NEXT: buffer_store_dword v8, off, s[0:3], s32 offset:168 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_lshlrev_b32_e32 v8, 16, v37
+; SI-NEXT: v_lshlrev_b32_e32 v8, 16, v27
+; SI-NEXT: buffer_store_dword v5, off, s[0:3], s32 offset:244 ; 4-byte Folded Spill
+; SI-NEXT: s_waitcnt expcnt(0)
+; SI-NEXT: v_lshlrev_b32_e32 v5, 16, v13
; SI-NEXT: buffer_store_dword v8, off, s[0:3], s32 offset:176 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_lshlrev_b32_e32 v8, 16, v27
+; SI-NEXT: v_lshlrev_b32_e32 v8, 16, v35
+; SI-NEXT: buffer_store_dword v5, off, s[0:3], s32 offset:240 ; 4-byte Folded Spill
+; SI-NEXT: s_waitcnt expcnt(0)
+; SI-NEXT: v_lshlrev_b32_e32 v5, 16, v14
; SI-NEXT: buffer_store_dword v8, off, s[0:3], s32 offset:180 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_lshlrev_b32_e32 v8, 16, v48
+; SI-NEXT: v_lshlrev_b32_e32 v8, 16, v38
+; SI-NEXT: buffer_store_dword v5, off, s[0:3], s32 offset:248 ; 4-byte Folded Spill
+; SI-NEXT: s_waitcnt expcnt(0)
+; SI-NEXT: v_lshlrev_b32_e32 v5, 16, v15
; SI-NEXT: buffer_store_dword v8, off, s[0:3], s32 offset:188 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_lshlrev_b32_e32 v8, 16, v49
+; SI-NEXT: v_lshlrev_b32_e32 v8, 16, v48
+; SI-NEXT: buffer_store_dword v5, off, s[0:3], s32 offset:252 ; 4-byte Folded Spill
+; SI-NEXT: s_waitcnt expcnt(0)
+; SI-NEXT: v_lshlrev_b32_e32 v5, 16, v31
; SI-NEXT: buffer_store_dword v8, off, s[0:3], s32 offset:184 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_lshlrev_b32_e32 v8, 16, v35
+; SI-NEXT: v_lshlrev_b32_e32 v8, 16, v49
+; SI-NEXT: buffer_store_dword v5, off, s[0:3], s32 offset:260 ; 4-byte Folded Spill
+; SI-NEXT: s_waitcnt expcnt(0)
+; SI-NEXT: v_lshlrev_b32_e32 v5, 16, v17
; SI-NEXT: buffer_store_dword v8, off, s[0:3], s32 offset:192 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_lshlrev_b32_e32 v8, 16, v50
+; SI-NEXT: v_lshlrev_b32_e32 v8, 16, v32
+; SI-NEXT: buffer_store_dword v5, off, s[0:3], s32 offset:256 ; 4-byte Folded Spill
+; SI-NEXT: s_waitcnt expcnt(0)
+; SI-NEXT: v_lshlrev_b32_e32 v5, 16, v18
; SI-NEXT: buffer_store_dword v8, off, s[0:3], s32 offset:196 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_lshlrev_b32_e32 v8, 16, v51
+; SI-NEXT: v_lshlrev_b32_e32 v8, 16, v39
+; SI-NEXT: buffer_store_dword v5, off, s[0:3], s32 offset:264 ; 4-byte Folded Spill
+; SI-NEXT: v_mov_b32_e32 v24, v19
+; SI-NEXT: s_waitcnt expcnt(0)
+; SI-NEXT: v_lshlrev_b32_e32 v5, 16, v19
+; SI-NEXT: v_lshlrev_b32_e32 v61, 16, v60
+; SI-NEXT: v_mov_b32_e32 v39, v30
+; SI-NEXT: v_lshlrev_b32_e32 v60, 16, v30
+; SI-NEXT: v_lshlrev_b32_e32 v30, 16, v33
+; SI-NEXT: buffer_load_dword v33, off, s[0:3], s32 offset:280 ; 4-byte Folded Reload
+; SI-NEXT: v_lshlrev_b32_e32 v19, 16, v36
+; SI-NEXT: buffer_load_dword v36, off, s[0:3], s32 offset:284 ; 4-byte Folded Reload
+; SI-NEXT: v_lshlrev_b32_e32 v12, 16, v3
+; SI-NEXT: buffer_load_dword v3, off, s[0:3], s32 offset:288 ; 4-byte Folded Reload
; SI-NEXT: buffer_store_dword v8, off, s[0:3], s32 offset:204 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_lshlrev_b32_e32 v8, 16, v40
+; SI-NEXT: buffer_store_dword v5, off, s[0:3], s32 offset:268 ; 4-byte Folded Spill
+; SI-NEXT: s_waitcnt expcnt(0)
+; SI-NEXT: v_lshlrev_b32_e32 v5, 16, v20
; SI-NEXT: buffer_store_dword v8, off, s[0:3], s32 offset:200 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_lshlrev_b32_e32 v8, 16, v43
-; SI-NEXT: v_mov_b32_e32 v43, v6
-; SI-NEXT: buffer_load_dword v6, off, s[0:3], s32 offset:284 ; 4-byte Folded Reload
-; SI-NEXT: v_lshlrev_b32_e32 v4, 16, v4
-; SI-NEXT: buffer_store_dword v4, off, s[0:3], s32 offset:212 ; 4-byte Folded Spill
-; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_lshlrev_b32_e32 v4, 16, v5
-; SI-NEXT: v_lshlrev_b32_e32 v5, 16, v43
-; SI-NEXT: buffer_store_dword v5, off, s[0:3], s32 offset:216 ; 4-byte Folded Spill
-; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_lshlrev_b32_e32 v5, 16, v7
-; SI-NEXT: buffer_store_dword v5, off, s[0:3], s32 offset:220 ; 4-byte Folded Spill
-; SI-NEXT: v_lshlrev_b32_e32 v7, 16, v20
-; SI-NEXT: buffer_store_dword v7, off, s[0:3], s32 offset:268 ; 4-byte Folded Spill
-; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_lshlrev_b32_e32 v7, 16, v21
-; SI-NEXT: buffer_store_dword v7, off, s[0:3], s32 offset:264 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v5, off, s[0:3], s32 offset:272 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_lshlrev_b32_e32 v7, 16, v28
+; SI-NEXT: v_lshlrev_b32_e32 v5, 16, v21
; SI-NEXT: s_mov_b64 s[4:5], 0
; SI-NEXT: buffer_store_dword v8, off, s[0:3], s32 offset:208 ; 4-byte Folded Spill
-; SI-NEXT: v_lshlrev_b32_e32 v3, 16, v3
-; SI-NEXT: v_mov_b32_e32 v50, v19
+; SI-NEXT: s_waitcnt expcnt(0)
+; SI-NEXT: v_lshlrev_b32_e32 v8, 16, v56
+; SI-NEXT: v_mov_b32_e32 v23, v17
+; SI-NEXT: v_mov_b32_e32 v25, v20
+; SI-NEXT: v_mov_b32_e32 v29, v21
+; SI-NEXT: buffer_store_dword v5, off, s[0:3], s32 offset:276 ; 4-byte Folded Spill
; SI-NEXT: v_mov_b32_e32 v51, v22
-; SI-NEXT: v_mov_b32_e32 v38, v16
-; SI-NEXT: v_lshlrev_b32_e32 v16, 16, v16
-; SI-NEXT: v_mov_b32_e32 v37, v45
-; SI-NEXT: v_mov_b32_e32 v27, v26
-; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_lshlrev_b32_e32 v8, 16, v26
-; SI-NEXT: v_mov_b32_e32 v49, v47
-; SI-NEXT: v_mov_b32_e32 v35, v28
-; SI-NEXT: v_lshlrev_b32_e32 v20, 16, v58
-; SI-NEXT: v_lshlrev_b32_e32 v23, 16, v59
-; SI-NEXT: v_lshlrev_b32_e32 v40, 16, v60
-; SI-NEXT: v_lshlrev_b32_e32 v24, 16, v39
-; SI-NEXT: v_lshlrev_b32_e32 v48, 16, v1
-; SI-NEXT: v_lshlrev_b32_e32 v25, 16, v52
+; SI-NEXT: v_lshlrev_b32_e32 v9, 16, v22
+; SI-NEXT: v_mov_b32_e32 v48, v26
+; SI-NEXT: v_lshlrev_b32_e32 v6, 16, v26
+; SI-NEXT: v_lshlrev_b32_e32 v56, 16, v45
+; SI-NEXT: v_mov_b32_e32 v49, v16
+; SI-NEXT: v_lshlrev_b32_e32 v45, 16, v16
+; SI-NEXT: v_mov_b32_e32 v35, v46
+; SI-NEXT: s_waitcnt expcnt(0)
+; SI-NEXT: v_mov_b32_e32 v5, v57
+; SI-NEXT: v_lshlrev_b32_e32 v57, 16, v46
+; SI-NEXT: v_mov_b32_e32 v43, v59
+; SI-NEXT: v_lshlrev_b32_e32 v59, 16, v58
+; SI-NEXT: v_mov_b32_e32 v32, v28
+; SI-NEXT: v_lshlrev_b32_e32 v58, 16, v28
+; SI-NEXT: v_lshlrev_b32_e32 v38, 16, v2
+; SI-NEXT: v_lshlrev_b32_e32 v27, 16, v34
+; SI-NEXT: v_lshlrev_b32_e32 v10, 16, v37
+; SI-NEXT: v_lshlrev_b32_e32 v46, 16, v47
+; SI-NEXT: v_lshlrev_b32_e32 v40, 16, v1
+; SI-NEXT: v_lshlrev_b32_e32 v28, 16, v52
; SI-NEXT: v_lshlrev_b32_e32 v21, 16, v53
-; SI-NEXT: v_lshlrev_b32_e32 v29, 16, v54
-; SI-NEXT: v_lshlrev_b32_e32 v28, 16, v55
-; SI-NEXT: v_lshlrev_b32_e32 v30, 16, v41
-; SI-NEXT: v_lshlrev_b32_e32 v26, 16, v62
+; SI-NEXT: v_lshlrev_b32_e32 v17, 16, v54
+; SI-NEXT: v_lshlrev_b32_e32 v7, 16, v55
+; SI-NEXT: v_lshlrev_b32_e32 v16, 16, v41
+; SI-NEXT: v_lshlrev_b32_e32 v11, 16, v42
+; SI-NEXT: v_lshlrev_b32_e32 v14, 16, v62
+; SI-NEXT: v_lshlrev_b32_e32 v20, 16, v63
+; SI-NEXT: v_lshlrev_b32_e32 v15, 16, v50
+; SI-NEXT: s_waitcnt vmcnt(8)
+; SI-NEXT: v_lshlrev_b32_e32 v2, 16, v33
+; SI-NEXT: s_waitcnt vmcnt(7)
+; SI-NEXT: v_lshlrev_b32_e32 v31, 16, v36
; SI-NEXT: s_waitcnt vmcnt(6)
-; SI-NEXT: v_lshlrev_b32_e32 v5, 16, v6
-; SI-NEXT: buffer_store_dword v5, off, s[0:3], s32 offset:228 ; 4-byte Folded Spill
-; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_lshlrev_b32_e32 v5, 16, v9
-; SI-NEXT: buffer_store_dword v5, off, s[0:3], s32 offset:224 ; 4-byte Folded Spill
-; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_lshlrev_b32_e32 v5, 16, v10
-; SI-NEXT: buffer_store_dword v5, off, s[0:3], s32 offset:232 ; 4-byte Folded Spill
-; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_lshlrev_b32_e32 v5, 16, v11
-; SI-NEXT: buffer_store_dword v5, off, s[0:3], s32 offset:236 ; 4-byte Folded Spill
-; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_lshlrev_b32_e32 v5, 16, v12
-; SI-NEXT: buffer_store_dword v5, off, s[0:3], s32 offset:244 ; 4-byte Folded Spill
-; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_lshlrev_b32_e32 v5, 16, v13
-; SI-NEXT: buffer_store_dword v5, off, s[0:3], s32 offset:240 ; 4-byte Folded Spill
-; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_lshlrev_b32_e32 v5, 16, v14
-; SI-NEXT: buffer_store_dword v5, off, s[0:3], s32 offset:248 ; 4-byte Folded Spill
-; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_lshlrev_b32_e32 v5, 16, v31
-; SI-NEXT: buffer_store_dword v5, off, s[0:3], s32 offset:256 ; 4-byte Folded Spill
-; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_lshlrev_b32_e32 v5, 16, v17
-; SI-NEXT: buffer_store_dword v5, off, s[0:3], s32 offset:252 ; 4-byte Folded Spill
-; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_lshlrev_b32_e32 v5, 16, v18
-; SI-NEXT: buffer_store_dword v5, off, s[0:3], s32 offset:260 ; 4-byte Folded Spill
-; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_lshlrev_b32_e32 v5, 16, v19
-; SI-NEXT: buffer_store_dword v5, off, s[0:3], s32 offset:276 ; 4-byte Folded Spill
-; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_lshlrev_b32_e32 v5, 16, v46
-; SI-NEXT: buffer_store_dword v5, off, s[0:3], s32 offset:272 ; 4-byte Folded Spill
-; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_lshlrev_b32_e32 v5, 16, v47
-; SI-NEXT: v_lshlrev_b32_e32 v12, 16, v57
-; SI-NEXT: v_mov_b32_e32 v57, v5
-; SI-NEXT: v_lshlrev_b32_e32 v5, 16, v32
-; SI-NEXT: v_mov_b32_e32 v32, v7
-; SI-NEXT: v_lshlrev_b32_e32 v18, 16, v33
-; SI-NEXT: v_lshlrev_b32_e32 v7, 16, v34
-; SI-NEXT: v_lshlrev_b32_e32 v11, 16, v15
-; SI-NEXT: v_lshlrev_b32_e32 v17, 16, v22
-; SI-NEXT: v_lshlrev_b32_e32 v19, 16, v45
-; SI-NEXT: v_lshlrev_b32_e32 v10, 16, v56
-; SI-NEXT: v_mov_b32_e32 v33, v12
-; SI-NEXT: v_mov_b32_e32 v34, v5
-; SI-NEXT: v_mov_b32_e32 v58, v7
-; SI-NEXT: v_lshlrev_b32_e32 v22, 16, v36
-; SI-NEXT: v_lshlrev_b32_e32 v45, 16, v61
-; SI-NEXT: v_lshlrev_b32_e32 v47, 16, v42
-; SI-NEXT: v_lshlrev_b32_e32 v15, 16, v63
-; SI-NEXT: v_lshlrev_b32_e32 v7, 16, v2
-; SI-NEXT: v_lshlrev_b32_e32 v14, 16, v44
-; SI-NEXT: v_mov_b32_e32 v44, v18
-; SI-NEXT: v_mov_b32_e32 v5, v43
-; SI-NEXT: v_mov_b32_e32 v18, v6
+; SI-NEXT: v_lshlrev_b32_e32 v22, 16, v3
; SI-NEXT: s_branch .LBB103_3
; SI-NEXT: .LBB103_2:
-; SI-NEXT: buffer_load_dword v18, off, s[0:3], s32 offset:284 ; 4-byte Folded Reload
-; SI-NEXT: ; implicit-def: $vgpr5
; SI-NEXT: ; implicit-def: $vgpr3
-; SI-NEXT: ; kill: killed $vgpr5
-; SI-NEXT: ; implicit-def: $vgpr5
; SI-NEXT: ; kill: killed $vgpr3
-; SI-NEXT: ; implicit-def: $vgpr3
-; SI-NEXT: ; kill: killed $vgpr5
; SI-NEXT: ; implicit-def: $vgpr5
-; SI-NEXT: ; kill: killed $vgpr3
-; SI-NEXT: ; implicit-def: $vgpr3
; SI-NEXT: ; kill: killed $vgpr5
-; SI-NEXT: ; implicit-def: $vgpr5
-; SI-NEXT: ; kill: killed $vgpr3
+; SI-NEXT: v_mov_b32_e32 v39, v30
; SI-NEXT: ; implicit-def: $vgpr3
-; SI-NEXT: ; kill: killed $vgpr5
-; SI-NEXT: ; implicit-def: $vgpr5
; SI-NEXT: ; kill: killed $vgpr3
-; SI-NEXT: ; implicit-def: $vgpr3
-; SI-NEXT: ; kill: killed $vgpr5
; SI-NEXT: ; implicit-def: $vgpr5
-; SI-NEXT: ; kill: killed $vgpr3
-; SI-NEXT: ; implicit-def: $vgpr3
; SI-NEXT: ; kill: killed $vgpr5
-; SI-NEXT: ; implicit-def: $vgpr5
-; SI-NEXT: ; kill: killed $vgpr3
+; SI-NEXT: v_mov_b32_e32 v32, v28
; SI-NEXT: ; implicit-def: $vgpr3
-; SI-NEXT: ; kill: killed $vgpr5
-; SI-NEXT: ; implicit-def: $vgpr5
; SI-NEXT: ; kill: killed $vgpr3
-; SI-NEXT: ; implicit-def: $vgpr3
-; SI-NEXT: ; kill: killed $vgpr5
; SI-NEXT: ; implicit-def: $vgpr5
-; SI-NEXT: ; kill: killed $vgpr3
-; SI-NEXT: ; implicit-def: $vgpr3
; SI-NEXT: ; kill: killed $vgpr5
-; SI-NEXT: ; implicit-def: $vgpr5
-; SI-NEXT: ; kill: killed $vgpr3
+; SI-NEXT: v_mov_b32_e32 v35, v46
; SI-NEXT: ; implicit-def: $vgpr3
-; SI-NEXT: ; kill: killed $vgpr5
-; SI-NEXT: ; implicit-def: $vgpr5
; SI-NEXT: ; kill: killed $vgpr3
-; SI-NEXT: ; implicit-def: $vgpr3
-; SI-NEXT: ; kill: killed $vgpr5
; SI-NEXT: ; implicit-def: $vgpr5
-; SI-NEXT: ; kill: killed $vgpr3
-; SI-NEXT: ; implicit-def: $vgpr3
; SI-NEXT: ; kill: killed $vgpr5
-; SI-NEXT: ; implicit-def: $vgpr5
-; SI-NEXT: ; kill: killed $vgpr3
+; SI-NEXT: v_mov_b32_e32 v49, v16
; SI-NEXT: ; implicit-def: $vgpr3
-; SI-NEXT: ; kill: killed $vgpr5
+; SI-NEXT: ; kill: killed $vgpr3
; SI-NEXT: ; implicit-def: $vgpr5
+; SI-NEXT: ; kill: killed $vgpr5
+; SI-NEXT: v_mov_b32_e32 v48, v26
+; SI-NEXT: ; implicit-def: $vgpr3
; SI-NEXT: ; kill: killed $vgpr3
+; SI-NEXT: ; implicit-def: $vgpr5
+; SI-NEXT: ; kill: killed $vgpr5
+; SI-NEXT: v_mov_b32_e32 v51, v22
; SI-NEXT: ; implicit-def: $vgpr3
+; SI-NEXT: ; kill: killed $vgpr3
+; SI-NEXT: ; implicit-def: $vgpr5
; SI-NEXT: ; kill: killed $vgpr5
+; SI-NEXT: v_mov_b32_e32 v29, v21
+; SI-NEXT: ; implicit-def: $vgpr3
+; SI-NEXT: ; kill: killed $vgpr3
; SI-NEXT: ; implicit-def: $vgpr5
+; SI-NEXT: ; kill: killed $vgpr5
+; SI-NEXT: v_mov_b32_e32 v25, v20
+; SI-NEXT: ; implicit-def: $vgpr3
; SI-NEXT: ; kill: killed $vgpr3
+; SI-NEXT: ; implicit-def: $vgpr5
+; SI-NEXT: ; kill: killed $vgpr5
+; SI-NEXT: v_mov_b32_e32 v24, v19
; SI-NEXT: ; implicit-def: $vgpr3
+; SI-NEXT: ; kill: killed $vgpr3
+; SI-NEXT: ; implicit-def: $vgpr5
; SI-NEXT: ; kill: killed $vgpr5
+; SI-NEXT: v_mov_b32_e32 v23, v17
+; SI-NEXT: ; implicit-def: $vgpr3
+; SI-NEXT: ; kill: killed $vgpr3
; SI-NEXT: ; implicit-def: $vgpr5
-; SI-NEXT: v_mov_b32_e32 v35, v28
-; SI-NEXT: v_mov_b32_e32 v49, v47
-; SI-NEXT: v_mov_b32_e32 v27, v26
-; SI-NEXT: v_mov_b32_e32 v37, v45
-; SI-NEXT: v_mov_b32_e32 v38, v16
-; SI-NEXT: v_mov_b32_e32 v51, v22
-; SI-NEXT: v_mov_b32_e32 v50, v19
+; SI-NEXT: ; kill: killed $vgpr5
; SI-NEXT: s_mov_b64 s[4:5], -1
-; SI-NEXT: ; kill: killed $vgpr3
; SI-NEXT: ; implicit-def: $vgpr3
-; SI-NEXT: ; implicit-def: $vgpr4
-; SI-NEXT: ; kill: killed $vgpr5
-; SI-NEXT: v_mov_b32_e32 v5, v6
; SI-NEXT: ; kill: killed $vgpr3
+; SI-NEXT: ; implicit-def: $vgpr5
+; SI-NEXT: ; kill: killed $vgpr5
+; SI-NEXT: ; implicit-def: $vgpr4
+; SI-NEXT: v_mov_b32_e32 v43, v59
; SI-NEXT: ; implicit-def: $vgpr3
+; SI-NEXT: ; kill: killed $vgpr3
+; SI-NEXT: ; implicit-def: $vgpr5
+; SI-NEXT: ; kill: killed $vgpr5
+; SI-NEXT: ; implicit-def: $vgpr8
; SI-NEXT: ; kill: killed $vgpr4
; SI-NEXT: ; implicit-def: $vgpr4
-; SI-NEXT: ; implicit-def: $vgpr11
-; SI-NEXT: ; implicit-def: $vgpr17
-; SI-NEXT: ; implicit-def: $vgpr16
+; SI-NEXT: ; implicit-def: $vgpr9
+; SI-NEXT: ; implicit-def: $vgpr6
+; SI-NEXT: ; implicit-def: $vgpr56
+; SI-NEXT: ; implicit-def: $vgpr45
+; SI-NEXT: ; implicit-def: $vgpr59
+; SI-NEXT: ; implicit-def: $vgpr61
+; SI-NEXT: ; implicit-def: $vgpr58
+; SI-NEXT: ; implicit-def: $vgpr60
+; SI-NEXT: ; implicit-def: $vgpr38
+; SI-NEXT: ; implicit-def: $vgpr30
+; SI-NEXT: ; implicit-def: $vgpr27
+; SI-NEXT: ; implicit-def: $vgpr2
; SI-NEXT: ; implicit-def: $vgpr19
-; SI-NEXT: ; implicit-def: $vgpr8
-; SI-NEXT: ; implicit-def: $vgpr57
-; SI-NEXT: ; implicit-def: $vgpr32
; SI-NEXT: ; implicit-def: $vgpr10
-; SI-NEXT: ; implicit-def: $vgpr33
-; SI-NEXT: ; implicit-def: $vgpr34
-; SI-NEXT: ; implicit-def: $vgpr44
-; SI-NEXT: ; implicit-def: $vgpr58
-; SI-NEXT: ; implicit-def: $vgpr20
-; SI-NEXT: ; implicit-def: $vgpr22
-; SI-NEXT: ; implicit-def: $vgpr23
+; SI-NEXT: ; implicit-def: $vgpr31
+; SI-NEXT: ; implicit-def: $vgpr46
; SI-NEXT: ; implicit-def: $vgpr40
-; SI-NEXT: ; implicit-def: $vgpr24
-; SI-NEXT: ; implicit-def: $vgpr45
-; SI-NEXT: ; implicit-def: $vgpr48
-; SI-NEXT: ; implicit-def: $vgpr25
-; SI-NEXT: ; implicit-def: $vgpr21
-; SI-NEXT: ; implicit-def: $vgpr29
+; SI-NEXT: ; implicit-def: $vgpr12
; SI-NEXT: ; implicit-def: $vgpr28
-; SI-NEXT: ; implicit-def: $vgpr30
-; SI-NEXT: ; implicit-def: $vgpr47
-; SI-NEXT: ; implicit-def: $vgpr26
-; SI-NEXT: ; implicit-def: $vgpr15
+; SI-NEXT: ; implicit-def: $vgpr21
+; SI-NEXT: ; implicit-def: $vgpr17
; SI-NEXT: ; implicit-def: $vgpr7
+; SI-NEXT: ; implicit-def: $vgpr16
+; SI-NEXT: ; implicit-def: $vgpr11
; SI-NEXT: ; implicit-def: $vgpr14
+; SI-NEXT: ; implicit-def: $vgpr20
+; SI-NEXT: ; implicit-def: $vgpr22
+; SI-NEXT: ; implicit-def: $vgpr15
+; SI-NEXT: ; implicit-def: $vgpr3
+; SI-NEXT: ; kill: killed $vgpr3
+; SI-NEXT: ; implicit-def: $vgpr5
+; SI-NEXT: ; kill: killed $vgpr5
+; SI-NEXT: ; implicit-def: $vgpr3
+; SI-NEXT: ; kill: killed $vgpr3
+; SI-NEXT: ; implicit-def: $vgpr5
+; SI-NEXT: ; kill: killed $vgpr5
+; SI-NEXT: ; implicit-def: $vgpr3
+; SI-NEXT: ; kill: killed $vgpr3
+; SI-NEXT: ; implicit-def: $vgpr5
+; SI-NEXT: ; kill: killed $vgpr5
+; SI-NEXT: s_waitcnt expcnt(6)
+; SI-NEXT: buffer_load_dword v33, off, s[0:3], s32 offset:280 ; 4-byte Folded Reload
+; SI-NEXT: s_waitcnt expcnt(5)
+; SI-NEXT: buffer_load_dword v36, off, s[0:3], s32 offset:284 ; 4-byte Folded Reload
+; SI-NEXT: s_waitcnt expcnt(2)
+; SI-NEXT: buffer_load_dword v3, off, s[0:3], s32 offset:288 ; 4-byte Folded Reload
+; SI-NEXT: v_mov_b32_e32 v5, v57
+; SI-NEXT: ; implicit-def: $vgpr57
; SI-NEXT: .LBB103_3: ; %Flow
-; SI-NEXT: s_waitcnt expcnt(1)
-; SI-NEXT: v_mov_b32_e32 v36, v2
-; SI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:280 ; 4-byte Folded Reload
-; SI-NEXT: v_mov_b32_e32 v43, v9
-; SI-NEXT: v_mov_b32_e32 v12, v31
; SI-NEXT: s_andn2_b64 vcc, exec, s[4:5]
-; SI-NEXT: v_mov_b32_e32 v31, v11
-; SI-NEXT: v_mov_b32_e32 v9, v17
+; SI-NEXT: v_mov_b32_e32 v47, v49
+; SI-NEXT: s_waitcnt expcnt(1)
+; SI-NEXT: v_mov_b32_e32 v50, v2
; SI-NEXT: s_cbranch_vccnz .LBB103_5
; SI-NEXT: ; %bb.4: ; %cmp.true
-; SI-NEXT: s_waitcnt vmcnt(0)
-; SI-NEXT: v_cvt_f32_f16_e32 v2, v2
-; SI-NEXT: v_cvt_f32_f16_e32 v8, v36
+; SI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:428 ; 4-byte Folded Reload
+; SI-NEXT: s_waitcnt vmcnt(1)
+; SI-NEXT: v_cvt_f32_f16_e32 v8, v3
; SI-NEXT: v_cvt_f32_f16_e32 v10, v63
; SI-NEXT: v_cvt_f32_f16_e32 v14, v62
-; SI-NEXT: v_add_f32_e32 v2, 0x38000000, v2
-; SI-NEXT: buffer_store_dword v2, off, s[0:3], s32 offset:280 ; 4-byte Folded Spill
-; SI-NEXT: s_waitcnt expcnt(0)
+; SI-NEXT: v_mov_b32_e32 v26, v48
+; SI-NEXT: v_cvt_f32_f16_e32 v15, v54
+; SI-NEXT: v_cvt_f32_f16_e32 v3, v34
+; SI-NEXT: v_mov_b32_e32 v1, v39
+; SI-NEXT: v_cvt_f32_f16_e32 v4, v33
+; SI-NEXT: v_cvt_f32_f16_e32 v9, v51
+; SI-NEXT: v_cvt_f32_f16_e32 v50, v13
+; SI-NEXT: v_cvt_f32_f16_e32 v45, v5
+; SI-NEXT: v_cvt_f32_f16_e32 v31, v18
+; SI-NEXT: v_cvt_f32_f16_e32 v28, v24
+; SI-NEXT: v_add_f32_e32 v50, 0x38000000, v50
+; SI-NEXT: v_add_f32_e32 v45, 0x38000000, v45
+; SI-NEXT: v_add_f32_e32 v31, 0x38000000, v31
+; SI-NEXT: v_cvt_f32_f16_e32 v21, v25
+; SI-NEXT: v_add_f32_e32 v28, 0x38000000, v28
+; SI-NEXT: v_cvt_f32_f16_e32 v11, v29
+; SI-NEXT: v_add_f32_e32 v9, 0x38000000, v9
+; SI-NEXT: v_add_f32_e32 v21, 0x38000000, v21
+; SI-NEXT: v_cvt_f16_f32_e32 v9, v9
+; SI-NEXT: v_add_f32_e32 v11, 0x38000000, v11
+; SI-NEXT: v_lshlrev_b32_e32 v9, 16, v9
+; SI-NEXT: s_waitcnt vmcnt(0)
+; SI-NEXT: v_cvt_f32_f16_e32 v2, v2
+; SI-NEXT: v_add_f32_e32 v48, 0x38000000, v2
; SI-NEXT: v_add_f32_e32 v2, 0x38000000, v8
-; SI-NEXT: buffer_store_dword v2, off, s[0:3], s32 offset:284 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v2, off, s[0:3], s32 offset:280 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_add_f32_e32 v2, 0x38000000, v10
-; SI-NEXT: buffer_store_dword v2, off, s[0:3], s32 offset:428 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v2, off, s[0:3], s32 offset:284 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_add_f32_e32 v2, 0x38000000, v14
; SI-NEXT: v_cvt_f32_f16_e32 v14, v55
-; SI-NEXT: buffer_store_dword v2, off, s[0:3], s32 offset:432 ; 4-byte Folded Spill
-; SI-NEXT: v_cvt_f32_f16_e32 v15, v54
+; SI-NEXT: buffer_store_dword v2, off, s[0:3], s32 offset:288 ; 4-byte Folded Spill
+; SI-NEXT: v_cvt_f32_f16_e32 v10, v41
; SI-NEXT: v_cvt_f32_f16_e32 v8, v42
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_add_f32_e32 v2, 0x38000000, v14
-; SI-NEXT: v_cvt_f32_f16_e32 v14, v1
-; SI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:424 ; 4-byte Folded Reload
-; SI-NEXT: buffer_store_dword v2, off, s[0:3], s32 offset:436 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v2, off, s[0:3], s32 offset:428 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_add_f32_e32 v2, 0x38000000, v15
-; SI-NEXT: v_cvt_f32_f16_e32 v15, v61
-; SI-NEXT: v_add_f32_e32 v54, 0x38000000, v8
-; SI-NEXT: v_cvt_f32_f16_e32 v8, v53
-; SI-NEXT: v_cvt_f32_f16_e32 v10, v41
-; SI-NEXT: v_add_f32_e32 v29, 0x38000000, v15
-; SI-NEXT: v_mov_b32_e32 v6, v37
-; SI-NEXT: v_add_f32_e32 v30, 0x38000000, v8
-; SI-NEXT: v_cvt_f32_f16_e32 v8, v39
+; SI-NEXT: buffer_store_dword v2, off, s[0:3], s32 offset:436 ; 4-byte Folded Spill
+; SI-NEXT: s_waitcnt expcnt(0)
+; SI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:424 ; 4-byte Folded Reload
; SI-NEXT: v_add_f32_e32 v55, 0x38000000, v10
; SI-NEXT: v_cvt_f32_f16_e32 v10, v52
-; SI-NEXT: buffer_store_dword v2, off, s[0:3], s32 offset:440 ; 4-byte Folded Spill
-; SI-NEXT: v_add_f32_e32 v25, 0x38000000, v8
-; SI-NEXT: v_cvt_f32_f16_e32 v9, v51
-; SI-NEXT: v_add_f32_e32 v37, 0x38000000, v10
-; SI-NEXT: v_cvt_f32_f16_e32 v10, v60
-; SI-NEXT: v_add_f32_e32 v48, 0x38000000, v14
-; SI-NEXT: v_cvt_f32_f16_e32 v14, v59
-; SI-NEXT: v_cvt_f32_f16_e32 v28, v50
-; SI-NEXT: v_add_f32_e32 v24, 0x38000000, v10
-; SI-NEXT: v_cvt_f32_f16_e32 v50, v13
-; SI-NEXT: v_add_f32_e32 v23, 0x38000000, v14
-; SI-NEXT: v_cvt_f32_f16_e32 v33, v12
-; SI-NEXT: v_cvt_f32_f16_e32 v45, v5
+; SI-NEXT: v_add_f32_e32 v54, 0x38000000, v8
+; SI-NEXT: v_cvt_f32_f16_e32 v8, v53
; SI-NEXT: v_cvt_f32_f16_e32 v42, v43
-; SI-NEXT: v_cvt_f32_f16_e32 v43, v18
-; SI-NEXT: v_add_f32_e32 v50, 0x38000000, v50
-; SI-NEXT: v_add_f32_e32 v45, 0x38000000, v45
+; SI-NEXT: v_cvt_f32_f16_e32 v43, v44
+; SI-NEXT: v_add_f32_e32 v53, 0x38000000, v8
; SI-NEXT: v_add_f32_e32 v42, 0x38000000, v42
; SI-NEXT: v_add_f32_e32 v43, 0x38000000, v43
-; SI-NEXT: v_add_f32_e32 v33, 0x38000000, v33
-; SI-NEXT: v_add_f32_e32 v28, 0x38000000, v28
-; SI-NEXT: v_cvt_f32_f16_e32 v6, v6
-; SI-NEXT: v_cvt_f32_f16_e32 v20, v49
-; SI-NEXT: v_add_f32_e32 v9, 0x38000000, v9
-; SI-NEXT: v_cvt_f16_f32_e32 v9, v9
-; SI-NEXT: v_add_f32_e32 v6, 0x38000000, v6
-; SI-NEXT: v_add_f32_e32 v20, 0x38000000, v20
-; SI-NEXT: v_lshlrev_b32_e32 v9, 16, v9
-; SI-NEXT: s_waitcnt vmcnt(2)
-; SI-NEXT: v_cvt_f32_f16_e32 v15, v1
-; SI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:420 ; 4-byte Folded Reload
-; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_add_f32_e32 v2, 0x38000000, v15
; SI-NEXT: s_waitcnt vmcnt(0)
-; SI-NEXT: v_cvt_f32_f16_e32 v4, v1
-; SI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:416 ; 4-byte Folded Reload
+; SI-NEXT: v_cvt_f32_f16_e32 v14, v2
+; SI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:420 ; 4-byte Folded Reload
+; SI-NEXT: s_waitcnt vmcnt(0)
+; SI-NEXT: v_cvt_f32_f16_e32 v15, v2
+; SI-NEXT: v_add_f32_e32 v2, 0x38000000, v10
+; SI-NEXT: buffer_store_dword v2, off, s[0:3], s32 offset:420 ; 4-byte Folded Spill
+; SI-NEXT: s_waitcnt expcnt(0)
+; SI-NEXT: v_add_f32_e32 v2, 0x38000000, v14
+; SI-NEXT: buffer_store_dword v2, off, s[0:3], s32 offset:424 ; 4-byte Folded Spill
+; SI-NEXT: s_waitcnt expcnt(0)
+; SI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:416 ; 4-byte Folded Reload
+; SI-NEXT: v_add_f32_e32 v52, 0x38000000, v15
+; SI-NEXT: v_cvt_f32_f16_e32 v10, v36
+; SI-NEXT: v_cvt_f32_f16_e32 v14, v37
+; SI-NEXT: v_add_f32_e32 v37, 0x38000000, v10
+; SI-NEXT: v_add_f32_e32 v30, 0x38000000, v14
; SI-NEXT: v_add_f32_e32 v14, 0x38000000, v4
-; SI-NEXT: v_cvt_f32_f16_e32 v4, v56
+; SI-NEXT: v_cvt_f32_f16_e32 v4, v32
+; SI-NEXT: v_cvt_f32_f16_e32 v32, v23
; SI-NEXT: v_add_f32_e32 v16, 0x38000000, v4
-; SI-NEXT: v_cvt_f32_f16_e32 v4, v27
+; SI-NEXT: v_cvt_f32_f16_e32 v4, v47
+; SI-NEXT: v_add_f32_e32 v32, 0x38000000, v32
; SI-NEXT: v_add_f32_e32 v4, 0x38000000, v4
; SI-NEXT: v_cvt_f16_f32_e32 v4, v4
; SI-NEXT: s_waitcnt vmcnt(0)
-; SI-NEXT: v_cvt_f32_f16_e32 v3, v1
-; SI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:412 ; 4-byte Folded Reload
-; SI-NEXT: v_add_f32_e32 v15, 0x38000000, v3
+; SI-NEXT: v_cvt_f32_f16_e32 v8, v2
+; SI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:412 ; 4-byte Folded Reload
+; SI-NEXT: v_add_f32_e32 v49, 0x38000000, v8
; SI-NEXT: s_waitcnt vmcnt(0)
-; SI-NEXT: v_cvt_f32_f16_e32 v8, v1
-; SI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:408 ; 4-byte Folded Reload
-; SI-NEXT: v_add_f32_e32 v52, 0x38000000, v8
-; SI-NEXT: v_cvt_f32_f16_e32 v8, v35
+; SI-NEXT: v_cvt_f32_f16_e32 v15, v2
+; SI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:408 ; 4-byte Folded Reload
+; SI-NEXT: v_add_f32_e32 v39, 0x38000000, v15
+; SI-NEXT: v_add_f32_e32 v15, 0x38000000, v3
+; SI-NEXT: v_cvt_f32_f16_e32 v3, v1
+; SI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:400 ; 4-byte Folded Reload
+; SI-NEXT: s_waitcnt vmcnt(1)
+; SI-NEXT: v_cvt_f32_f16_e32 v8, v2
+; SI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:404 ; 4-byte Folded Reload
; SI-NEXT: v_add_f32_e32 v17, 0x38000000, v8
-; SI-NEXT: v_cvt_f32_f16_e32 v8, v38
-; SI-NEXT: v_add_f32_e32 v8, 0x38000000, v8
-; SI-NEXT: s_waitcnt vmcnt(0)
-; SI-NEXT: v_cvt_f32_f16_e32 v10, v1
-; SI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:404 ; 4-byte Folded Reload
-; SI-NEXT: v_add_f32_e32 v39, 0x38000000, v10
+; SI-NEXT: s_waitcnt vmcnt(1)
+; SI-NEXT: v_cvt_f32_f16_e32 v8, v1
+; SI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:432 ; 4-byte Folded Reload
; SI-NEXT: s_waitcnt vmcnt(0)
-; SI-NEXT: v_cvt_f32_f16_e32 v3, v1
-; SI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:368 ; 4-byte Folded Reload
-; SI-NEXT: v_add_f32_e32 v10, 0x38000000, v3
-; SI-NEXT: v_cvt_f32_f16_e32 v3, v46
-; SI-NEXT: v_add_f32_e32 v3, 0x38000000, v3
-; SI-NEXT: v_cvt_f16_f32_e32 v3, v3
+; SI-NEXT: v_cvt_f32_f16_e32 v20, v1
+; SI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:396 ; 4-byte Folded Reload
+; SI-NEXT: v_add_f32_e32 v20, 0x38000000, v20
; SI-NEXT: s_waitcnt vmcnt(0)
-; SI-NEXT: v_cvt_f32_f16_e32 v11, v1
-; SI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:356 ; 4-byte Folded Reload
-; SI-NEXT: v_add_f32_e32 v11, 0x38000000, v11
+; SI-NEXT: v_cvt_f32_f16_e32 v6, v1
+; SI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:352 ; 4-byte Folded Reload
+; SI-NEXT: v_add_f32_e32 v6, 0x38000000, v6
; SI-NEXT: s_waitcnt vmcnt(0)
-; SI-NEXT: v_cvt_f32_f16_e32 v21, v1
+; SI-NEXT: v_cvt_f32_f16_e32 v33, v1
; SI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:348 ; 4-byte Folded Reload
-; SI-NEXT: v_add_f32_e32 v21, 0x38000000, v21
-; SI-NEXT: s_waitcnt vmcnt(0)
-; SI-NEXT: v_cvt_f32_f16_e32 v31, v1
-; SI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:332 ; 4-byte Folded Reload
-; SI-NEXT: v_add_f32_e32 v31, 0x38000000, v31
-; SI-NEXT: s_waitcnt vmcnt(0)
-; SI-NEXT: v_cvt_f32_f16_e32 v32, v1
-; SI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:328 ; 4-byte Folded Reload
-; SI-NEXT: v_add_f32_e32 v32, 0x38000000, v32
+; SI-NEXT: v_add_f32_e32 v33, 0x38000000, v33
; SI-NEXT: s_waitcnt vmcnt(0)
; SI-NEXT: v_cvt_f32_f16_e32 v34, v1
-; SI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:324 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:328 ; 4-byte Folded Reload
; SI-NEXT: v_add_f32_e32 v34, 0x38000000, v34
; SI-NEXT: s_waitcnt vmcnt(0)
; SI-NEXT: v_cvt_f32_f16_e32 v36, v1
-; SI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:320 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:324 ; 4-byte Folded Reload
; SI-NEXT: v_add_f32_e32 v36, 0x38000000, v36
; SI-NEXT: s_waitcnt vmcnt(0)
; SI-NEXT: v_cvt_f32_f16_e32 v51, v1
-; SI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:316 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:320 ; 4-byte Folded Reload
; SI-NEXT: v_add_f32_e32 v51, 0x38000000, v51
; SI-NEXT: s_waitcnt vmcnt(0)
; SI-NEXT: v_cvt_f32_f16_e32 v40, v1
-; SI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:312 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:316 ; 4-byte Folded Reload
; SI-NEXT: v_add_f32_e32 v40, 0x38000000, v40
; SI-NEXT: s_waitcnt vmcnt(0)
; SI-NEXT: v_cvt_f32_f16_e32 v41, v1
-; SI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:308 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:312 ; 4-byte Folded Reload
; SI-NEXT: v_add_f32_e32 v41, 0x38000000, v41
; SI-NEXT: s_waitcnt vmcnt(0)
; SI-NEXT: v_cvt_f32_f16_e32 v44, v1
-; SI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:304 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:308 ; 4-byte Folded Reload
; SI-NEXT: v_add_f32_e32 v44, 0x38000000, v44
; SI-NEXT: s_waitcnt vmcnt(0)
; SI-NEXT: v_cvt_f32_f16_e32 v46, v1
-; SI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:300 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:304 ; 4-byte Folded Reload
; SI-NEXT: v_add_f32_e32 v46, 0x38000000, v46
; SI-NEXT: s_waitcnt vmcnt(0)
; SI-NEXT: v_cvt_f32_f16_e32 v47, v1
-; SI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:296 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:300 ; 4-byte Folded Reload
; SI-NEXT: v_add_f32_e32 v47, 0x38000000, v47
; SI-NEXT: s_waitcnt vmcnt(0)
; SI-NEXT: v_cvt_f32_f16_e32 v56, v1
-; SI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:292 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:296 ; 4-byte Folded Reload
; SI-NEXT: v_add_f32_e32 v56, 0x38000000, v56
; SI-NEXT: s_waitcnt vmcnt(0)
; SI-NEXT: v_cvt_f32_f16_e32 v57, v1
-; SI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:288 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:292 ; 4-byte Folded Reload
+; SI-NEXT: v_cvt_f32_f16_e32 v10, v2
+; SI-NEXT: v_add_f32_e32 v2, 0x38000000, v8
+; SI-NEXT: v_cvt_f32_f16_e32 v8, v26
; SI-NEXT: v_add_f32_e32 v57, 0x38000000, v57
+; SI-NEXT: v_add_f32_e32 v27, 0x38000000, v10
+; SI-NEXT: v_add_f32_e32 v10, 0x38000000, v3
+; SI-NEXT: v_cvt_f32_f16_e32 v3, v35
+; SI-NEXT: v_add_f32_e32 v8, 0x38000000, v8
+; SI-NEXT: v_add_f32_e32 v3, 0x38000000, v3
+; SI-NEXT: v_cvt_f16_f32_e32 v3, v3
; SI-NEXT: s_waitcnt vmcnt(0)
; SI-NEXT: v_cvt_f32_f16_e32 v58, v1
-; SI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:400 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:392 ; 4-byte Folded Reload
; SI-NEXT: v_add_f32_e32 v58, 0x38000000, v58
; SI-NEXT: s_waitcnt vmcnt(0)
; SI-NEXT: v_cvt_f32_f16_e32 v26, v1
-; SI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:396 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:388 ; 4-byte Folded Reload
; SI-NEXT: v_add_f32_e32 v26, 0x38000000, v26
; SI-NEXT: s_waitcnt vmcnt(0)
; SI-NEXT: v_cvt_f32_f16_e32 v22, v1
-; SI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:392 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:384 ; 4-byte Folded Reload
; SI-NEXT: v_add_f32_e32 v22, 0x38000000, v22
; SI-NEXT: s_waitcnt vmcnt(0)
; SI-NEXT: v_cvt_f32_f16_e32 v19, v1
-; SI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:388 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:380 ; 4-byte Folded Reload
; SI-NEXT: v_add_f32_e32 v19, 0x38000000, v19
; SI-NEXT: s_waitcnt vmcnt(0)
; SI-NEXT: v_cvt_f32_f16_e32 v35, v1
-; SI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:384 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:376 ; 4-byte Folded Reload
; SI-NEXT: v_add_f32_e32 v35, 0x38000000, v35
; SI-NEXT: s_waitcnt vmcnt(0)
; SI-NEXT: v_cvt_f32_f16_e32 v13, v1
-; SI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:380 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:372 ; 4-byte Folded Reload
; SI-NEXT: v_add_f32_e32 v13, 0x38000000, v13
; SI-NEXT: s_waitcnt vmcnt(0)
; SI-NEXT: v_cvt_f32_f16_e32 v12, v1
-; SI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:376 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:368 ; 4-byte Folded Reload
; SI-NEXT: v_add_f32_e32 v12, 0x38000000, v12
; SI-NEXT: s_waitcnt vmcnt(0)
; SI-NEXT: v_cvt_f32_f16_e32 v7, v1
-; SI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:372 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:364 ; 4-byte Folded Reload
; SI-NEXT: v_add_f32_e32 v7, 0x38000000, v7
; SI-NEXT: v_cvt_f16_f32_e32 v7, v7
; SI-NEXT: s_waitcnt vmcnt(0)
; SI-NEXT: v_cvt_f32_f16_e32 v5, v1
-; SI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:364 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:360 ; 4-byte Folded Reload
; SI-NEXT: v_add_f32_e32 v5, 0x38000000, v5
; SI-NEXT: v_cvt_f16_f32_e32 v5, v5
; SI-NEXT: s_waitcnt vmcnt(0)
; SI-NEXT: v_cvt_f32_f16_e32 v59, v1
-; SI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:360 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:356 ; 4-byte Folded Reload
; SI-NEXT: v_add_f32_e32 v59, 0x38000000, v59
; SI-NEXT: v_cvt_f16_f32_e32 v59, v59
; SI-NEXT: s_waitcnt vmcnt(0)
; SI-NEXT: v_cvt_f32_f16_e32 v60, v1
-; SI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:352 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:344 ; 4-byte Folded Reload
; SI-NEXT: v_add_f32_e32 v60, 0x38000000, v60
; SI-NEXT: s_waitcnt vmcnt(0)
; SI-NEXT: v_cvt_f32_f16_e32 v61, v1
-; SI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:344 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:340 ; 4-byte Folded Reload
; SI-NEXT: v_add_f32_e32 v61, 0x38000000, v61
; SI-NEXT: v_cvt_f16_f32_e32 v61, v61
; SI-NEXT: s_waitcnt vmcnt(0)
; SI-NEXT: v_cvt_f32_f16_e32 v62, v1
-; SI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:340 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:336 ; 4-byte Folded Reload
; SI-NEXT: v_add_f32_e32 v62, 0x38000000, v62
; SI-NEXT: v_cvt_f16_f32_e32 v62, v62
; SI-NEXT: s_waitcnt vmcnt(0)
; SI-NEXT: v_cvt_f32_f16_e32 v63, v1
-; SI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:336 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:332 ; 4-byte Folded Reload
; SI-NEXT: v_add_f32_e32 v63, 0x38000000, v63
; SI-NEXT: v_cvt_f16_f32_e32 v63, v63
; SI-NEXT: s_waitcnt vmcnt(0)
@@ -235455,14 +235465,22 @@ define inreg <64 x bfloat> @bitcast_v64f16_to_v64bf16_scalar(<64 x half> inreg %
; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:184 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_lshlrev_b32_e32 v1, 16, v12
+; SI-NEXT: v_cvt_f16_f32_e32 v12, v57
+; SI-NEXT: v_lshlrev_b32_e32 v57, 16, v3
+; SI-NEXT: v_cvt_f16_f32_e32 v3, v2
; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:192 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_cvt_f16_f32_e32 v1, v22
; SI-NEXT: v_cvt_f16_f32_e32 v5, v26
+; SI-NEXT: v_lshlrev_b32_e32 v61, 16, v3
+; SI-NEXT: v_cvt_f16_f32_e32 v3, v17
; SI-NEXT: v_cvt_f16_f32_e32 v7, v58
-; SI-NEXT: v_cvt_f16_f32_e32 v12, v57
; SI-NEXT: v_lshlrev_b32_e32 v1, 16, v1
; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:196 ; 4-byte Folded Spill
+; SI-NEXT: v_lshlrev_b32_e32 v2, 16, v3
+; SI-NEXT: v_cvt_f16_f32_e32 v3, v30
+; SI-NEXT: v_mov_b32_e32 v30, v2
+; SI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:424 ; 4-byte Folded Reload
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_lshlrev_b32_e32 v1, 16, v5
; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:204 ; 4-byte Folded Spill
@@ -235486,7 +235504,7 @@ define inreg <64 x bfloat> @bitcast_v64f16_to_v64bf16_scalar(<64 x half> inreg %
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_cvt_f16_f32_e32 v1, v44
; SI-NEXT: v_cvt_f16_f32_e32 v5, v43
-; SI-NEXT: v_lshlrev_b32_e32 v27, 16, v7
+; SI-NEXT: v_lshlrev_b32_e32 v18, 16, v7
; SI-NEXT: v_cvt_f16_f32_e32 v7, v42
; SI-NEXT: v_cvt_f16_f32_e32 v12, v41
; SI-NEXT: v_lshlrev_b32_e32 v1, 16, v1
@@ -235521,99 +235539,92 @@ define inreg <64 x bfloat> @bitcast_v64f16_to_v64bf16_scalar(<64 x half> inreg %
; SI-NEXT: v_cvt_f16_f32_e32 v5, v33
; SI-NEXT: v_cvt_f16_f32_e32 v7, v32
; SI-NEXT: v_cvt_f16_f32_e32 v12, v31
-; SI-NEXT: v_lshlrev_b32_e32 v31, 16, v1
+; SI-NEXT: v_lshlrev_b32_e32 v1, 16, v1
+; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:252 ; 4-byte Folded Spill
+; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_lshlrev_b32_e32 v1, 16, v5
-; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:256 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:260 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_lshlrev_b32_e32 v1, 16, v7
-; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:252 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:256 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_lshlrev_b32_e32 v1, 16, v12
-; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:260 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:264 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_cvt_f16_f32_e32 v1, v28
; SI-NEXT: v_cvt_f16_f32_e32 v5, v21
; SI-NEXT: v_cvt_f16_f32_e32 v7, v11
+; SI-NEXT: v_lshlrev_b32_e32 v45, 16, v4
; SI-NEXT: v_lshlrev_b32_e32 v1, 16, v1
-; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:276 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:268 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_lshlrev_b32_e32 v1, 16, v5
-; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:268 ; 4-byte Folded Spill
+; SI-NEXT: v_cvt_f16_f32_e32 v5, v6
+; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:272 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_lshlrev_b32_e32 v1, 16, v7
-; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:264 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:276 ; 4-byte Folded Spill
+; SI-NEXT: v_lshlrev_b32_e32 v56, 16, v5
+; SI-NEXT: v_cvt_f16_f32_e32 v5, v10
+; SI-NEXT: v_lshlrev_b32_e32 v10, 16, v3
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_cvt_f16_f32_e32 v1, v8
-; SI-NEXT: v_cvt_f16_f32_e32 v5, v6
-; SI-NEXT: v_lshlrev_b32_e32 v8, 16, v4
; SI-NEXT: v_cvt_f16_f32_e32 v4, v16
+; SI-NEXT: v_lshlrev_b32_e32 v60, 16, v5
+; SI-NEXT: v_cvt_f16_f32_e32 v5, v14
; SI-NEXT: v_lshlrev_b32_e32 v6, 16, v1
-; SI-NEXT: v_lshlrev_b32_e32 v1, 16, v3
-; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:272 ; 4-byte Folded Spill
-; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_cvt_f16_f32_e32 v1, v20
-; SI-NEXT: v_cvt_f16_f32_e32 v3, v17
-; SI-NEXT: v_lshlrev_b32_e32 v19, 16, v5
-; SI-NEXT: v_cvt_f16_f32_e32 v5, v10
-; SI-NEXT: v_lshlrev_b32_e32 v57, 16, v1
-; SI-NEXT: v_cvt_f16_f32_e32 v1, v39
-; SI-NEXT: v_lshlrev_b32_e32 v32, 16, v3
-; SI-NEXT: v_cvt_f16_f32_e32 v3, v52
-; SI-NEXT: v_lshlrev_b32_e32 v10, 16, v4
-; SI-NEXT: v_lshlrev_b32_e32 v34, 16, v1
-; SI-NEXT: v_cvt_f16_f32_e32 v1, v2
-; SI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:436 ; 4-byte Folded Reload
-; SI-NEXT: v_lshlrev_b32_e32 v44, 16, v3
-; SI-NEXT: v_cvt_f16_f32_e32 v3, v23
-; SI-NEXT: v_lshlrev_b32_e32 v22, 16, v1
-; SI-NEXT: v_cvt_f16_f32_e32 v1, v29
-; SI-NEXT: v_cvt_f16_f32_e32 v4, v15
-; SI-NEXT: v_lshlrev_b32_e32 v23, 16, v3
-; SI-NEXT: v_cvt_f16_f32_e32 v3, v48
-; SI-NEXT: v_lshlrev_b32_e32 v45, 16, v1
-; SI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:440 ; 4-byte Folded Reload
; SI-NEXT: v_lshlrev_b32_e32 v58, 16, v4
-; SI-NEXT: v_lshlrev_b32_e32 v48, 16, v3
-; SI-NEXT: v_cvt_f16_f32_e32 v4, v24
-; SI-NEXT: v_lshlrev_b32_e32 v33, 16, v5
-; SI-NEXT: v_cvt_f16_f32_e32 v5, v14
-; SI-NEXT: v_mov_b32_e32 v16, v6
-; SI-NEXT: v_lshlrev_b32_e32 v40, 16, v4
+; SI-NEXT: v_cvt_f16_f32_e32 v4, v15
+; SI-NEXT: v_lshlrev_b32_e32 v50, 16, v5
+; SI-NEXT: v_lshlrev_b32_e32 v59, 16, v1
+; SI-NEXT: v_cvt_f16_f32_e32 v1, v27
+; SI-NEXT: v_lshlrev_b32_e32 v27, 16, v4
; SI-NEXT: v_cvt_f16_f32_e32 v4, v37
-; SI-NEXT: v_lshlrev_b32_e32 v20, 16, v5
-; SI-NEXT: v_cvt_f16_f32_e32 v5, v25
-; SI-NEXT: v_lshlrev_b32_e32 v25, 16, v4
-; SI-NEXT: v_cvt_f16_f32_e32 v4, v55
-; SI-NEXT: v_lshlrev_b32_e32 v24, 16, v5
-; SI-NEXT: v_cvt_f16_f32_e32 v5, v30
-; SI-NEXT: v_lshlrev_b32_e32 v30, 16, v4
-; SI-NEXT: buffer_load_dword v4, off, s[0:3], s32 offset:280 ; 4-byte Folded Reload
+; SI-NEXT: s_waitcnt vmcnt(14)
+; SI-NEXT: v_cvt_f16_f32_e32 v3, v2
+; SI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:420 ; 4-byte Folded Reload
+; SI-NEXT: v_lshlrev_b32_e32 v38, 16, v1
+; SI-NEXT: v_lshlrev_b32_e32 v31, 16, v4
+; SI-NEXT: v_cvt_f16_f32_e32 v1, v39
+; SI-NEXT: v_lshlrev_b32_e32 v12, 16, v3
+; SI-NEXT: v_cvt_f16_f32_e32 v5, v49
+; SI-NEXT: v_mov_b32_e32 v8, v13
+; SI-NEXT: v_lshlrev_b32_e32 v19, 16, v1
+; SI-NEXT: v_cvt_f16_f32_e32 v1, v52
+; SI-NEXT: v_lshlrev_b32_e32 v46, 16, v5
+; SI-NEXT: v_cvt_f16_f32_e32 v5, v53
+; SI-NEXT: v_lshlrev_b32_e32 v40, 16, v1
+; SI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:436 ; 4-byte Folded Reload
; SI-NEXT: v_lshlrev_b32_e32 v21, 16, v5
; SI-NEXT: v_cvt_f16_f32_e32 v5, v54
-; SI-NEXT: v_lshlrev_b32_e32 v47, 16, v5
-; SI-NEXT: s_waitcnt vmcnt(2)
-; SI-NEXT: v_cvt_f16_f32_e32 v3, v2
+; SI-NEXT: v_lshlrev_b32_e32 v11, 16, v5
+; SI-NEXT: s_waitcnt vmcnt(1)
+; SI-NEXT: v_cvt_f16_f32_e32 v4, v2
; SI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:428 ; 4-byte Folded Reload
-; SI-NEXT: v_lshlrev_b32_e32 v28, 16, v3
-; SI-NEXT: s_waitcnt vmcnt(2)
+; SI-NEXT: v_lshlrev_b32_e32 v28, 16, v4
+; SI-NEXT: v_cvt_f16_f32_e32 v4, v55
+; SI-NEXT: v_lshlrev_b32_e32 v16, 16, v4
+; SI-NEXT: v_cvt_f16_f32_e32 v4, v48
+; SI-NEXT: v_lshlrev_b32_e32 v15, 16, v4
+; SI-NEXT: v_mov_b32_e32 v4, v18
+; SI-NEXT: s_waitcnt vmcnt(0)
+; SI-NEXT: v_cvt_f16_f32_e32 v3, v2
+; SI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:284 ; 4-byte Folded Reload
; SI-NEXT: v_cvt_f16_f32_e32 v1, v1
-; SI-NEXT: v_lshlrev_b32_e32 v29, 16, v1
-; SI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:432 ; 4-byte Folded Reload
-; SI-NEXT: s_waitcnt vmcnt(2)
-; SI-NEXT: v_cvt_f16_f32_e32 v4, v4
-; SI-NEXT: v_lshlrev_b32_e32 v14, 16, v4
-; SI-NEXT: v_mov_b32_e32 v4, v27
+; SI-NEXT: v_lshlrev_b32_e32 v7, 16, v3
+; SI-NEXT: v_lshlrev_b32_e32 v17, 16, v1
+; SI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:288 ; 4-byte Folded Reload
; SI-NEXT: s_waitcnt vmcnt(1)
; SI-NEXT: v_cvt_f16_f32_e32 v3, v2
-; SI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:284 ; 4-byte Folded Reload
-; SI-NEXT: v_lshlrev_b32_e32 v15, 16, v3
-; SI-NEXT: v_mov_b32_e32 v3, v13
+; SI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:280 ; 4-byte Folded Reload
+; SI-NEXT: v_lshlrev_b32_e32 v20, 16, v3
; SI-NEXT: s_waitcnt vmcnt(1)
; SI-NEXT: v_cvt_f16_f32_e32 v1, v1
-; SI-NEXT: v_lshlrev_b32_e32 v26, 16, v1
+; SI-NEXT: v_lshlrev_b32_e32 v14, 16, v1
; SI-NEXT: s_waitcnt vmcnt(0)
; SI-NEXT: v_cvt_f16_f32_e32 v2, v2
-; SI-NEXT: v_lshlrev_b32_e32 v7, 16, v2
+; SI-NEXT: v_lshlrev_b32_e32 v22, 16, v2
; SI-NEXT: .LBB103_5: ; %end
; SI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:156 ; 4-byte Folded Reload
; SI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:148 ; 4-byte Folded Reload
@@ -235703,7 +235714,7 @@ define inreg <64 x bfloat> @bitcast_v64f16_to_v64bf16_scalar(<64 x half> inreg %
; SI-NEXT: buffer_store_dword v1, v2, s[0:3], 0 offen
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:212 ; 4-byte Folded Reload
-; SI-NEXT: v_mul_f32_e32 v2, 1.0, v3
+; SI-NEXT: v_mul_f32_e32 v2, 1.0, v8
; SI-NEXT: s_waitcnt vmcnt(0)
; SI-NEXT: v_mul_f32_e32 v1, 1.0, v1
; SI-NEXT: v_lshrrev_b32_e32 v1, 16, v1
@@ -235764,17 +235775,19 @@ define inreg <64 x bfloat> @bitcast_v64f16_to_v64bf16_scalar(<64 x half> inreg %
; SI-NEXT: v_add_i32_e32 v2, vcc, 52, v0
; SI-NEXT: buffer_store_dword v1, v2, s[0:3], 0 offen
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:256 ; 4-byte Folded Reload
-; SI-NEXT: v_mul_f32_e32 v2, 1.0, v31
-; SI-NEXT: s_waitcnt vmcnt(0)
+; SI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:260 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:252 ; 4-byte Folded Reload
+; SI-NEXT: s_waitcnt vmcnt(1)
; SI-NEXT: v_mul_f32_e32 v1, 1.0, v1
; SI-NEXT: v_lshrrev_b32_e32 v1, 16, v1
+; SI-NEXT: s_waitcnt vmcnt(0)
+; SI-NEXT: v_mul_f32_e32 v2, 1.0, v2
; SI-NEXT: v_alignbit_b32 v1, v1, v2, 16
; SI-NEXT: v_add_i32_e32 v2, vcc, 56, v0
; SI-NEXT: buffer_store_dword v1, v2, s[0:3], 0 offen
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:260 ; 4-byte Folded Reload
-; SI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:252 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:264 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:256 ; 4-byte Folded Reload
; SI-NEXT: s_waitcnt vmcnt(1)
; SI-NEXT: v_mul_f32_e32 v1, 1.0, v1
; SI-NEXT: v_lshrrev_b32_e32 v1, 16, v1
@@ -235784,8 +235797,8 @@ define inreg <64 x bfloat> @bitcast_v64f16_to_v64bf16_scalar(<64 x half> inreg %
; SI-NEXT: v_add_i32_e32 v2, vcc, 60, v0
; SI-NEXT: buffer_store_dword v1, v2, s[0:3], 0 offen
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:268 ; 4-byte Folded Reload
-; SI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:276 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:272 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:268 ; 4-byte Folded Reload
; SI-NEXT: s_waitcnt vmcnt(1)
; SI-NEXT: v_mul_f32_e32 v1, 1.0, v1
; SI-NEXT: v_lshrrev_b32_e32 v1, 16, v1
@@ -235794,7 +235807,7 @@ define inreg <64 x bfloat> @bitcast_v64f16_to_v64bf16_scalar(<64 x half> inreg %
; SI-NEXT: v_alignbit_b32 v1, v1, v2, 16
; SI-NEXT: v_add_i32_e32 v2, vcc, 64, v0
; SI-NEXT: buffer_store_dword v1, v2, s[0:3], 0 offen
-; SI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:264 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:276 ; 4-byte Folded Reload
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_mul_f32_e32 v1, 1.0, v9
; SI-NEXT: v_lshrrev_b32_e32 v1, 16, v1
@@ -235804,102 +235817,100 @@ define inreg <64 x bfloat> @bitcast_v64f16_to_v64bf16_scalar(<64 x half> inreg %
; SI-NEXT: v_add_i32_e32 v2, vcc, 0x44, v0
; SI-NEXT: buffer_store_dword v1, v2, s[0:3], 0 offen
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_mul_f32_e32 v1, 1.0, v19
+; SI-NEXT: v_mul_f32_e32 v1, 1.0, v56
; SI-NEXT: v_lshrrev_b32_e32 v1, 16, v1
-; SI-NEXT: v_mul_f32_e32 v2, 1.0, v16
+; SI-NEXT: v_mul_f32_e32 v2, 1.0, v6
; SI-NEXT: v_alignbit_b32 v1, v1, v2, 16
; SI-NEXT: v_add_i32_e32 v2, vcc, 0x48, v0
; SI-NEXT: buffer_store_dword v1, v2, s[0:3], 0 offen
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:272 ; 4-byte Folded Reload
-; SI-NEXT: v_mul_f32_e32 v2, 1.0, v8
-; SI-NEXT: s_waitcnt vmcnt(0)
-; SI-NEXT: v_mul_f32_e32 v1, 1.0, v1
+; SI-NEXT: v_mul_f32_e32 v1, 1.0, v57
; SI-NEXT: v_lshrrev_b32_e32 v1, 16, v1
+; SI-NEXT: v_mul_f32_e32 v2, 1.0, v45
; SI-NEXT: v_alignbit_b32 v1, v1, v2, 16
; SI-NEXT: v_add_i32_e32 v2, vcc, 0x4c, v0
; SI-NEXT: buffer_store_dword v1, v2, s[0:3], 0 offen
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_mul_f32_e32 v1, 1.0, v32
+; SI-NEXT: v_mul_f32_e32 v1, 1.0, v61
; SI-NEXT: v_lshrrev_b32_e32 v1, 16, v1
-; SI-NEXT: v_mul_f32_e32 v2, 1.0, v57
+; SI-NEXT: v_mul_f32_e32 v2, 1.0, v59
; SI-NEXT: v_alignbit_b32 v1, v1, v2, 16
; SI-NEXT: v_add_i32_e32 v2, vcc, 0x50, v0
; SI-NEXT: buffer_store_dword v1, v2, s[0:3], 0 offen
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_mul_f32_e32 v1, 1.0, v33
+; SI-NEXT: v_mul_f32_e32 v1, 1.0, v60
; SI-NEXT: v_lshrrev_b32_e32 v1, 16, v1
-; SI-NEXT: v_mul_f32_e32 v2, 1.0, v10
+; SI-NEXT: v_mul_f32_e32 v2, 1.0, v58
; SI-NEXT: v_alignbit_b32 v1, v1, v2, 16
; SI-NEXT: v_add_i32_e32 v2, vcc, 0x54, v0
; SI-NEXT: buffer_store_dword v1, v2, s[0:3], 0 offen
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_mul_f32_e32 v1, 1.0, v44
+; SI-NEXT: v_mul_f32_e32 v1, 1.0, v30
; SI-NEXT: v_lshrrev_b32_e32 v1, 16, v1
-; SI-NEXT: v_mul_f32_e32 v2, 1.0, v34
+; SI-NEXT: v_mul_f32_e32 v2, 1.0, v38
; SI-NEXT: v_alignbit_b32 v1, v1, v2, 16
; SI-NEXT: v_add_i32_e32 v2, vcc, 0x58, v0
; SI-NEXT: buffer_store_dword v1, v2, s[0:3], 0 offen
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_mul_f32_e32 v1, 1.0, v20
+; SI-NEXT: v_mul_f32_e32 v1, 1.0, v50
; SI-NEXT: v_lshrrev_b32_e32 v1, 16, v1
-; SI-NEXT: v_mul_f32_e32 v2, 1.0, v58
+; SI-NEXT: v_mul_f32_e32 v2, 1.0, v27
; SI-NEXT: v_alignbit_b32 v1, v1, v2, 16
; SI-NEXT: v_add_i32_e32 v2, vcc, 0x5c, v0
; SI-NEXT: buffer_store_dword v1, v2, s[0:3], 0 offen
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_mul_f32_e32 v1, 1.0, v23
+; SI-NEXT: v_mul_f32_e32 v1, 1.0, v10
; SI-NEXT: v_lshrrev_b32_e32 v1, 16, v1
-; SI-NEXT: v_mul_f32_e32 v2, 1.0, v22
+; SI-NEXT: v_mul_f32_e32 v2, 1.0, v19
; SI-NEXT: v_alignbit_b32 v1, v1, v2, 16
; SI-NEXT: v_add_i32_e32 v2, vcc, 0x60, v0
; SI-NEXT: buffer_store_dword v1, v2, s[0:3], 0 offen
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_mul_f32_e32 v1, 1.0, v24
+; SI-NEXT: v_mul_f32_e32 v1, 1.0, v46
; SI-NEXT: v_lshrrev_b32_e32 v1, 16, v1
-; SI-NEXT: v_mul_f32_e32 v2, 1.0, v40
+; SI-NEXT: v_mul_f32_e32 v2, 1.0, v31
; SI-NEXT: v_alignbit_b32 v1, v1, v2, 16
; SI-NEXT: v_add_i32_e32 v2, vcc, 0x64, v0
; SI-NEXT: buffer_store_dword v1, v2, s[0:3], 0 offen
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_mul_f32_e32 v1, 1.0, v48
+; SI-NEXT: v_mul_f32_e32 v1, 1.0, v12
; SI-NEXT: v_lshrrev_b32_e32 v1, 16, v1
-; SI-NEXT: v_mul_f32_e32 v2, 1.0, v45
+; SI-NEXT: v_mul_f32_e32 v2, 1.0, v40
; SI-NEXT: v_alignbit_b32 v1, v1, v2, 16
; SI-NEXT: v_add_i32_e32 v2, vcc, 0x68, v0
; SI-NEXT: buffer_store_dword v1, v2, s[0:3], 0 offen
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_mul_f32_e32 v1, 1.0, v21
; SI-NEXT: v_lshrrev_b32_e32 v1, 16, v1
-; SI-NEXT: v_mul_f32_e32 v2, 1.0, v25
+; SI-NEXT: v_mul_f32_e32 v2, 1.0, v28
; SI-NEXT: v_alignbit_b32 v1, v1, v2, 16
; SI-NEXT: v_add_i32_e32 v2, vcc, 0x6c, v0
; SI-NEXT: buffer_store_dword v1, v2, s[0:3], 0 offen
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_mul_f32_e32 v1, 1.0, v28
+; SI-NEXT: v_mul_f32_e32 v1, 1.0, v7
; SI-NEXT: v_lshrrev_b32_e32 v1, 16, v1
-; SI-NEXT: v_mul_f32_e32 v2, 1.0, v29
+; SI-NEXT: v_mul_f32_e32 v2, 1.0, v17
; SI-NEXT: v_alignbit_b32 v1, v1, v2, 16
; SI-NEXT: v_add_i32_e32 v2, vcc, 0x70, v0
; SI-NEXT: buffer_store_dword v1, v2, s[0:3], 0 offen
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_mul_f32_e32 v1, 1.0, v47
+; SI-NEXT: v_mul_f32_e32 v1, 1.0, v11
; SI-NEXT: v_lshrrev_b32_e32 v1, 16, v1
-; SI-NEXT: v_mul_f32_e32 v2, 1.0, v30
+; SI-NEXT: v_mul_f32_e32 v2, 1.0, v16
; SI-NEXT: v_alignbit_b32 v1, v1, v2, 16
; SI-NEXT: v_add_i32_e32 v2, vcc, 0x74, v0
; SI-NEXT: buffer_store_dword v1, v2, s[0:3], 0 offen
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_mul_f32_e32 v1, 1.0, v15
+; SI-NEXT: v_mul_f32_e32 v1, 1.0, v20
; SI-NEXT: v_lshrrev_b32_e32 v1, 16, v1
-; SI-NEXT: v_mul_f32_e32 v2, 1.0, v26
+; SI-NEXT: v_mul_f32_e32 v2, 1.0, v14
; SI-NEXT: v_alignbit_b32 v1, v1, v2, 16
; SI-NEXT: v_add_i32_e32 v2, vcc, 0x78, v0
; SI-NEXT: buffer_store_dword v1, v2, s[0:3], 0 offen
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_mul_f32_e32 v1, 1.0, v14
+; SI-NEXT: v_mul_f32_e32 v1, 1.0, v15
; SI-NEXT: v_lshrrev_b32_e32 v1, 16, v1
-; SI-NEXT: v_mul_f32_e32 v2, 1.0, v7
+; SI-NEXT: v_mul_f32_e32 v2, 1.0, v22
; SI-NEXT: v_alignbit_b32 v1, v1, v2, 16
; SI-NEXT: v_add_i32_e32 v0, vcc, 0x7c, v0
; SI-NEXT: buffer_store_dword v1, v0, s[0:3], 0 offen
@@ -236251,17 +236262,17 @@ define <64 x i16> @bitcast_v64bf16_to_v64i16(<64 x bfloat> %a, i32 %b) {
; SI-NEXT: buffer_store_dword v61, off, s[0:3], s32 offset:148 ; 4-byte Folded Spill
; SI-NEXT: buffer_store_dword v62, off, s[0:3], s32 offset:144 ; 4-byte Folded Spill
; SI-NEXT: buffer_store_dword v63, off, s[0:3], s32 offset:140 ; 4-byte Folded Spill
-; SI-NEXT: buffer_load_dword v34, off, s[0:3], s32 offset:136
-; SI-NEXT: buffer_load_dword v35, off, s[0:3], s32
-; SI-NEXT: buffer_load_dword v36, off, s[0:3], s32 offset:4
-; SI-NEXT: buffer_load_dword v37, off, s[0:3], s32 offset:8
-; SI-NEXT: buffer_load_dword v38, off, s[0:3], s32 offset:12
-; SI-NEXT: buffer_load_dword v39, off, s[0:3], s32 offset:16
-; SI-NEXT: buffer_load_dword v48, off, s[0:3], s32 offset:20
-; SI-NEXT: buffer_load_dword v49, off, s[0:3], s32 offset:24
-; SI-NEXT: buffer_load_dword v50, off, s[0:3], s32 offset:28
-; SI-NEXT: buffer_load_dword v51, off, s[0:3], s32 offset:32
-; SI-NEXT: buffer_load_dword v52, off, s[0:3], s32 offset:36
+; SI-NEXT: buffer_load_dword v35, off, s[0:3], s32 offset:136
+; SI-NEXT: buffer_load_dword v36, off, s[0:3], s32
+; SI-NEXT: buffer_load_dword v37, off, s[0:3], s32 offset:4
+; SI-NEXT: buffer_load_dword v38, off, s[0:3], s32 offset:8
+; SI-NEXT: buffer_load_dword v39, off, s[0:3], s32 offset:12
+; SI-NEXT: buffer_load_dword v48, off, s[0:3], s32 offset:16
+; SI-NEXT: buffer_load_dword v49, off, s[0:3], s32 offset:20
+; SI-NEXT: buffer_load_dword v50, off, s[0:3], s32 offset:24
+; SI-NEXT: buffer_load_dword v51, off, s[0:3], s32 offset:28
+; SI-NEXT: buffer_load_dword v52, off, s[0:3], s32 offset:32
+; SI-NEXT: buffer_load_dword v40, off, s[0:3], s32 offset:36
; SI-NEXT: buffer_load_dword v41, off, s[0:3], s32 offset:40
; SI-NEXT: buffer_load_dword v42, off, s[0:3], s32 offset:44
; SI-NEXT: buffer_load_dword v44, off, s[0:3], s32 offset:48
@@ -236280,169 +236291,170 @@ define <64 x i16> @bitcast_v64bf16_to_v64i16(<64 x bfloat> %a, i32 %b) {
; SI-NEXT: buffer_load_dword v62, off, s[0:3], s32 offset:80
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: buffer_load_dword v63, off, s[0:3], s32 offset:84
-; SI-NEXT: buffer_load_dword v45, off, s[0:3], s32 offset:92
-; SI-NEXT: buffer_load_dword v61, off, s[0:3], s32 offset:88
-; SI-NEXT: v_mul_f32_e32 v33, 1.0, v1
+; SI-NEXT: buffer_load_dword v43, off, s[0:3], s32 offset:92
+; SI-NEXT: buffer_load_dword v45, off, s[0:3], s32 offset:88
+; SI-NEXT: v_mul_f32_e32 v34, 1.0, v1
; SI-NEXT: v_mul_f32_e32 v1, 1.0, v3
-; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:328 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:332 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_mul_f32_e32 v1, 1.0, v4
-; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:324 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:328 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_mul_f32_e32 v1, 1.0, v7
-; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:360 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:364 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_mul_f32_e32 v1, 1.0, v8
-; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:352 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:356 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_mul_f32_e32 v1, 1.0, v11
-; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:408 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:412 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_mul_f32_e32 v1, 1.0, v12
-; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:404 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:408 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_mul_f32_e32 v1, 1.0, v15
-; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:416 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:420 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_mul_f32_e32 v1, 1.0, v16
-; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:412 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:416 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_mul_f32_e32 v1, 1.0, v19
-; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:424 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:428 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_mul_f32_e32 v1, 1.0, v20
-; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:420 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:424 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_mul_f32_e32 v1, 1.0, v23
-; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:432 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:436 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_mul_f32_e32 v1, 1.0, v24
-; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:428 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:432 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_mul_f32_e32 v1, 1.0, v27
-; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:440 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:444 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_mul_f32_e32 v1, 1.0, v28
-; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:436 ; 4-byte Folded Spill
-; SI-NEXT: v_mul_f32_e32 v54, 1.0, v13
-; SI-NEXT: v_mul_f32_e32 v19, 1.0, v21
-; SI-NEXT: v_mul_f32_e32 v12, 1.0, v22
-; SI-NEXT: v_mul_f32_e32 v13, 1.0, v25
-; SI-NEXT: v_mul_f32_e32 v15, 1.0, v17
-; SI-NEXT: v_mul_f32_e32 v17, 1.0, v29
-; SI-NEXT: v_mul_f32_e32 v43, 1.0, v2
-; SI-NEXT: v_mul_f32_e32 v40, 1.0, v5
-; SI-NEXT: v_mul_f32_e32 v32, 1.0, v6
-; SI-NEXT: v_mul_f32_e32 v31, 1.0, v9
-; SI-NEXT: v_mul_f32_e32 v55, 1.0, v10
-; SI-NEXT: v_mul_f32_e32 v53, 1.0, v14
-; SI-NEXT: v_mul_f32_e32 v14, 1.0, v18
-; SI-NEXT: v_mul_f32_e32 v18, 1.0, v26
-; SI-NEXT: v_mul_f32_e32 v11, 1.0, v30
+; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:440 ; 4-byte Folded Spill
+; SI-NEXT: v_mul_f32_e32 v31, 1.0, v10
+; SI-NEXT: v_mul_f32_e32 v12, 1.0, v18
+; SI-NEXT: v_mul_f32_e32 v10, 1.0, v22
+; SI-NEXT: v_mul_f32_e32 v18, 1.0, v30
+; SI-NEXT: v_mul_f32_e32 v11, 1.0, v25
+; SI-NEXT: v_mul_f32_e32 v19, 1.0, v29
+; SI-NEXT: v_mul_f32_e32 v20, 1.0, v26
+; SI-NEXT: v_mul_f32_e32 v33, 1.0, v2
+; SI-NEXT: v_mul_f32_e32 v55, 1.0, v5
+; SI-NEXT: v_mul_f32_e32 v54, 1.0, v6
+; SI-NEXT: v_mul_f32_e32 v32, 1.0, v9
+; SI-NEXT: v_mul_f32_e32 v53, 1.0, v13
+; SI-NEXT: v_mul_f32_e32 v14, 1.0, v14
+; SI-NEXT: v_mul_f32_e32 v13, 1.0, v17
+; SI-NEXT: v_mul_f32_e32 v21, 1.0, v21
; SI-NEXT: s_waitcnt vmcnt(14)
-; SI-NEXT: v_cmp_ne_u32_e32 vcc, 0, v34
-; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_mul_f32_e32 v1, 1.0, v35
-; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:448 ; 4-byte Folded Spill
+; SI-NEXT: v_cmp_ne_u32_e32 vcc, 0, v35
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_mul_f32_e32 v1, 1.0, v36
-; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:444 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:452 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_mul_f32_e32 v1, 1.0, v39
-; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:456 ; 4-byte Folded Spill
+; SI-NEXT: v_mul_f32_e32 v1, 1.0, v37
+; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:448 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_mul_f32_e32 v1, 1.0, v48
-; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:452 ; 4-byte Folded Spill
-; SI-NEXT: v_mul_f32_e32 v10, 1.0, v37
-; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_mul_f32_e32 v1, 1.0, v52
; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:460 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
+; SI-NEXT: v_mul_f32_e32 v1, 1.0, v49
+; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:456 ; 4-byte Folded Spill
+; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:96
; SI-NEXT: buffer_load_dword v3, off, s[0:3], s32 offset:100
-; SI-NEXT: buffer_load_dword v16, off, s[0:3], s32 offset:104
-; SI-NEXT: buffer_load_dword v20, off, s[0:3], s32 offset:108
-; SI-NEXT: buffer_load_dword v21, off, s[0:3], s32 offset:112
-; SI-NEXT: buffer_load_dword v22, off, s[0:3], s32 offset:116
-; SI-NEXT: buffer_load_dword v25, off, s[0:3], s32 offset:120
-; SI-NEXT: v_mul_f32_e32 v28, 1.0, v44
+; SI-NEXT: buffer_load_dword v22, off, s[0:3], s32 offset:104
+; SI-NEXT: buffer_load_dword v23, off, s[0:3], s32 offset:108
+; SI-NEXT: buffer_load_dword v24, off, s[0:3], s32 offset:112
+; SI-NEXT: buffer_load_dword v28, off, s[0:3], s32 offset:116
+; SI-NEXT: buffer_load_dword v30, off, s[0:3], s32 offset:120
+; SI-NEXT: v_mul_f32_e32 v17, 1.0, v38
+; SI-NEXT: v_mul_f32_e32 v9, 1.0, v39
+; SI-NEXT: v_mul_f32_e32 v8, 1.0, v50
+; SI-NEXT: v_mul_f32_e32 v25, 1.0, v44
; SI-NEXT: v_mul_f32_e32 v44, 1.0, v46
-; SI-NEXT: v_mul_f32_e32 v9, 1.0, v38
-; SI-NEXT: v_mul_f32_e32 v8, 1.0, v49
-; SI-NEXT: v_mul_f32_e32 v7, 1.0, v50
+; SI-NEXT: v_mul_f32_e32 v7, 1.0, v51
+; SI-NEXT: v_mul_f32_e32 v15, 1.0, v52
+; SI-NEXT: v_mul_f32_e32 v40, 1.0, v40
; SI-NEXT: v_mul_f32_e32 v46, 1.0, v58
-; SI-NEXT: v_mul_f32_e32 v29, 1.0, v59
-; SI-NEXT: v_mul_f32_e32 v30, 1.0, v51
+; SI-NEXT: v_mul_f32_e32 v61, 1.0, v59
; SI-NEXT: v_mul_f32_e32 v6, 1.0, v41
; SI-NEXT: v_mul_f32_e32 v5, 1.0, v42
-; SI-NEXT: v_mul_f32_e32 v27, 1.0, v45
-; SI-NEXT: v_mul_f32_e32 v23, 1.0, v61
-; SI-NEXT: ; implicit-def: $vgpr61
-; SI-NEXT: ; kill: killed $vgpr61
-; SI-NEXT: ; implicit-def: $vgpr61
-; SI-NEXT: ; kill: killed $vgpr61
-; SI-NEXT: ; implicit-def: $vgpr61
-; SI-NEXT: ; kill: killed $vgpr61
-; SI-NEXT: ; implicit-def: $vgpr61
-; SI-NEXT: ; kill: killed $vgpr61
-; SI-NEXT: ; implicit-def: $vgpr61
-; SI-NEXT: ; kill: killed $vgpr61
-; SI-NEXT: ; implicit-def: $vgpr61
-; SI-NEXT: ; kill: killed $vgpr61
-; SI-NEXT: ; implicit-def: $vgpr61
-; SI-NEXT: ; kill: killed $vgpr61
-; SI-NEXT: ; implicit-def: $vgpr61
-; SI-NEXT: ; kill: killed $vgpr61
-; SI-NEXT: ; implicit-def: $vgpr61
-; SI-NEXT: ; kill: killed $vgpr61
-; SI-NEXT: ; implicit-def: $vgpr61
-; SI-NEXT: ; kill: killed $vgpr61
-; SI-NEXT: ; implicit-def: $vgpr61
-; SI-NEXT: ; kill: killed $vgpr61
-; SI-NEXT: ; implicit-def: $vgpr61
-; SI-NEXT: ; kill: killed $vgpr61
-; SI-NEXT: ; implicit-def: $vgpr61
-; SI-NEXT: ; kill: killed $vgpr61
-; SI-NEXT: ; implicit-def: $vgpr61
-; SI-NEXT: ; kill: killed $vgpr61
-; SI-NEXT: ; implicit-def: $vgpr61
-; SI-NEXT: ; kill: killed $vgpr61
-; SI-NEXT: ; implicit-def: $vgpr61
-; SI-NEXT: ; kill: killed $vgpr61
-; SI-NEXT: ; implicit-def: $vgpr61
-; SI-NEXT: ; kill: killed $vgpr61
-; SI-NEXT: ; implicit-def: $vgpr61
-; SI-NEXT: ; kill: killed $vgpr61
-; SI-NEXT: ; implicit-def: $vgpr61
-; SI-NEXT: ; kill: killed $vgpr61
-; SI-NEXT: ; implicit-def: $vgpr61
-; SI-NEXT: ; kill: killed $vgpr61
-; SI-NEXT: ; implicit-def: $vgpr61
-; SI-NEXT: ; kill: killed $vgpr61
-; SI-NEXT: ; implicit-def: $vgpr61
-; SI-NEXT: ; kill: killed $vgpr61
-; SI-NEXT: ; implicit-def: $vgpr61
-; SI-NEXT: ; kill: killed $vgpr61
-; SI-NEXT: ; implicit-def: $vgpr61
-; SI-NEXT: ; kill: killed $vgpr61
-; SI-NEXT: ; implicit-def: $vgpr61
+; SI-NEXT: v_mul_f32_e32 v4, 1.0, v47
+; SI-NEXT: v_mul_f32_e32 v2, 1.0, v56
+; SI-NEXT: v_mul_f32_e32 v29, 1.0, v45
+; SI-NEXT: ; implicit-def: $vgpr45
+; SI-NEXT: ; kill: killed $vgpr45
+; SI-NEXT: ; implicit-def: $vgpr45
+; SI-NEXT: ; kill: killed $vgpr45
+; SI-NEXT: ; implicit-def: $vgpr45
+; SI-NEXT: ; kill: killed $vgpr45
+; SI-NEXT: ; implicit-def: $vgpr45
+; SI-NEXT: ; kill: killed $vgpr45
+; SI-NEXT: ; implicit-def: $vgpr45
+; SI-NEXT: ; kill: killed $vgpr45
+; SI-NEXT: ; implicit-def: $vgpr45
+; SI-NEXT: ; kill: killed $vgpr45
+; SI-NEXT: ; implicit-def: $vgpr45
+; SI-NEXT: ; kill: killed $vgpr45
+; SI-NEXT: ; implicit-def: $vgpr45
+; SI-NEXT: ; kill: killed $vgpr45
+; SI-NEXT: ; implicit-def: $vgpr45
+; SI-NEXT: ; kill: killed $vgpr45
+; SI-NEXT: ; implicit-def: $vgpr45
+; SI-NEXT: ; kill: killed $vgpr45
+; SI-NEXT: ; implicit-def: $vgpr45
+; SI-NEXT: ; kill: killed $vgpr45
+; SI-NEXT: ; implicit-def: $vgpr45
+; SI-NEXT: ; kill: killed $vgpr45
+; SI-NEXT: ; implicit-def: $vgpr45
+; SI-NEXT: ; kill: killed $vgpr45
+; SI-NEXT: ; implicit-def: $vgpr45
+; SI-NEXT: ; kill: killed $vgpr45
+; SI-NEXT: ; implicit-def: $vgpr45
+; SI-NEXT: ; kill: killed $vgpr45
+; SI-NEXT: ; implicit-def: $vgpr45
+; SI-NEXT: ; kill: killed $vgpr45
+; SI-NEXT: ; implicit-def: $vgpr45
+; SI-NEXT: ; kill: killed $vgpr45
+; SI-NEXT: ; implicit-def: $vgpr45
+; SI-NEXT: ; kill: killed $vgpr45
+; SI-NEXT: ; implicit-def: $vgpr45
+; SI-NEXT: ; kill: killed $vgpr45
+; SI-NEXT: ; implicit-def: $vgpr45
+; SI-NEXT: ; kill: killed $vgpr45
+; SI-NEXT: ; implicit-def: $vgpr45
+; SI-NEXT: ; kill: killed $vgpr45
+; SI-NEXT: ; implicit-def: $vgpr45
+; SI-NEXT: ; kill: killed $vgpr45
+; SI-NEXT: ; implicit-def: $vgpr45
+; SI-NEXT: ; kill: killed $vgpr45
+; SI-NEXT: ; implicit-def: $vgpr45
+; SI-NEXT: ; kill: killed $vgpr45
+; SI-NEXT: ; implicit-def: $vgpr45
+; SI-NEXT: ; kill: killed $vgpr45
+; SI-NEXT: ; implicit-def: $vgpr45
+; SI-NEXT: ; kill: killed $vgpr45
+; SI-NEXT: ; implicit-def: $vgpr45
+; SI-NEXT: ; kill: killed $vgpr45
+; SI-NEXT: ; implicit-def: $vgpr45
+; SI-NEXT: ; kill: killed $vgpr45
+; SI-NEXT: ; implicit-def: $vgpr45
+; SI-NEXT: ; kill: killed $vgpr45
; SI-NEXT: ; implicit-def: $vgpr45
-; SI-NEXT: ; kill: killed $vgpr61
-; SI-NEXT: ; implicit-def: $vgpr61
; SI-NEXT: ; kill: killed $vgpr45
; SI-NEXT: ; implicit-def: $vgpr45
-; SI-NEXT: ; kill: killed $vgpr61
-; SI-NEXT: ; implicit-def: $vgpr61
-; SI-NEXT: v_mul_f32_e32 v4, 1.0, v47
-; SI-NEXT: v_mul_f32_e32 v2, 1.0, v56
; SI-NEXT: v_mul_f32_e32 v47, 1.0, v57
; SI-NEXT: v_mul_f32_e32 v60, 1.0, v60
; SI-NEXT: v_mul_f32_e32 v57, 1.0, v62
; SI-NEXT: v_mul_f32_e32 v56, 1.0, v63
+; SI-NEXT: v_mul_f32_e32 v16, 1.0, v43
; SI-NEXT: ; kill: killed $vgpr45
; SI-NEXT: ; implicit-def: $vgpr45
-; SI-NEXT: ; kill: killed $vgpr61
-; SI-NEXT: ; implicit-def: $vgpr61
; SI-NEXT: ; implicit-def: $vgpr62
; SI-NEXT: ; implicit-def: $vgpr63
; SI-NEXT: ; implicit-def: $vgpr41
@@ -236457,106 +236469,96 @@ define <64 x i16> @bitcast_v64bf16_to_v64i16(<64 x bfloat> %a, i32 %b) {
; SI-NEXT: ; implicit-def: $vgpr38
; SI-NEXT: ; implicit-def: $vgpr35
; SI-NEXT: ; implicit-def: $vgpr36
+; SI-NEXT: ; implicit-def: $vgpr43
; SI-NEXT: ; kill: killed $vgpr45
; SI-NEXT: ; implicit-def: $vgpr45
-; SI-NEXT: ; kill: killed $vgpr61
-; SI-NEXT: ; implicit-def: $vgpr61
; SI-NEXT: s_waitcnt vmcnt(6)
; SI-NEXT: v_mul_f32_e32 v59, 1.0, v1
; SI-NEXT: s_waitcnt vmcnt(5)
; SI-NEXT: v_mul_f32_e32 v58, 1.0, v3
; SI-NEXT: s_waitcnt vmcnt(4)
-; SI-NEXT: v_mul_f32_e32 v16, 1.0, v16
+; SI-NEXT: v_mul_f32_e32 v27, 1.0, v22
; SI-NEXT: s_waitcnt vmcnt(3)
-; SI-NEXT: v_mul_f32_e32 v24, 1.0, v20
+; SI-NEXT: v_mul_f32_e32 v26, 1.0, v23
; SI-NEXT: s_waitcnt vmcnt(2)
-; SI-NEXT: v_mul_f32_e32 v3, 1.0, v21
+; SI-NEXT: v_mul_f32_e32 v3, 1.0, v24
; SI-NEXT: s_waitcnt vmcnt(1)
-; SI-NEXT: v_mul_f32_e32 v1, 1.0, v22
-; SI-NEXT: buffer_load_dword v21, off, s[0:3], s32 offset:124
-; SI-NEXT: buffer_load_dword v22, off, s[0:3], s32 offset:128
-; SI-NEXT: buffer_load_dword v34, off, s[0:3], s32 offset:132
+; SI-NEXT: v_mul_f32_e32 v1, 1.0, v28
+; SI-NEXT: buffer_load_dword v22, off, s[0:3], s32 offset:124
+; SI-NEXT: buffer_load_dword v24, off, s[0:3], s32 offset:128
+; SI-NEXT: buffer_load_dword v28, off, s[0:3], s32 offset:132
; SI-NEXT: s_waitcnt vmcnt(3)
-; SI-NEXT: v_mul_f32_e32 v20, 1.0, v25
+; SI-NEXT: v_mul_f32_e32 v23, 1.0, v30
; SI-NEXT: s_waitcnt vmcnt(2)
-; SI-NEXT: v_mul_f32_e32 v26, 1.0, v21
+; SI-NEXT: v_mul_f32_e32 v30, 1.0, v22
; SI-NEXT: s_waitcnt vmcnt(1)
-; SI-NEXT: v_mul_f32_e32 v25, 1.0, v22
-; SI-NEXT: ; implicit-def: $vgpr22
-; SI-NEXT: ; kill: killed $vgpr22
-; SI-NEXT: ; implicit-def: $vgpr22
-; SI-NEXT: ; kill: killed $vgpr22
-; SI-NEXT: ; implicit-def: $vgpr22
-; SI-NEXT: ; kill: killed $vgpr22
-; SI-NEXT: ; implicit-def: $vgpr22
-; SI-NEXT: ; kill: killed $vgpr22
-; SI-NEXT: ; implicit-def: $vgpr22
-; SI-NEXT: ; kill: killed $vgpr22
-; SI-NEXT: ; implicit-def: $vgpr22
-; SI-NEXT: ; kill: killed $vgpr22
-; SI-NEXT: ; implicit-def: $vgpr22
-; SI-NEXT: ; kill: killed $vgpr22
-; SI-NEXT: ; implicit-def: $vgpr22
-; SI-NEXT: ; kill: killed $vgpr22
-; SI-NEXT: ; implicit-def: $vgpr22
-; SI-NEXT: ; kill: killed $vgpr22
-; SI-NEXT: ; implicit-def: $vgpr22
-; SI-NEXT: ; kill: killed $vgpr22
-; SI-NEXT: ; implicit-def: $vgpr22
-; SI-NEXT: ; kill: killed $vgpr22
-; SI-NEXT: ; implicit-def: $vgpr22
-; SI-NEXT: ; kill: killed $vgpr22
-; SI-NEXT: ; implicit-def: $vgpr22
-; SI-NEXT: ; kill: killed $vgpr22
-; SI-NEXT: ; implicit-def: $vgpr22
+; SI-NEXT: v_mul_f32_e32 v22, 1.0, v24
; SI-NEXT: s_waitcnt vmcnt(0)
-; SI-NEXT: v_mul_f32_e32 v21, 1.0, v34
-; SI-NEXT: ; kill: killed $vgpr22
-; SI-NEXT: ; implicit-def: $vgpr22
-; SI-NEXT: ; kill: killed $vgpr22
-; SI-NEXT: ; implicit-def: $vgpr22
-; SI-NEXT: ; implicit-def: $vgpr34
+; SI-NEXT: v_mul_f32_e32 v24, 1.0, v28
+; SI-NEXT: ; implicit-def: $vgpr28
+; SI-NEXT: ; kill: killed $vgpr28
+; SI-NEXT: ; implicit-def: $vgpr28
+; SI-NEXT: ; kill: killed $vgpr28
+; SI-NEXT: ; implicit-def: $vgpr28
+; SI-NEXT: ; kill: killed $vgpr28
+; SI-NEXT: ; implicit-def: $vgpr28
+; SI-NEXT: ; kill: killed $vgpr28
+; SI-NEXT: ; implicit-def: $vgpr28
+; SI-NEXT: ; kill: killed $vgpr28
+; SI-NEXT: ; implicit-def: $vgpr28
+; SI-NEXT: ; kill: killed $vgpr28
+; SI-NEXT: ; implicit-def: $vgpr28
+; SI-NEXT: ; kill: killed $vgpr28
+; SI-NEXT: ; implicit-def: $vgpr28
+; SI-NEXT: ; kill: killed $vgpr28
+; SI-NEXT: ; implicit-def: $vgpr28
+; SI-NEXT: ; kill: killed $vgpr28
+; SI-NEXT: ; implicit-def: $vgpr28
+; SI-NEXT: ; kill: killed $vgpr28
+; SI-NEXT: ; implicit-def: $vgpr28
+; SI-NEXT: ; kill: killed $vgpr28
+; SI-NEXT: ; implicit-def: $vgpr28
+; SI-NEXT: ; kill: killed $vgpr28
+; SI-NEXT: ; implicit-def: $vgpr28
+; SI-NEXT: ; kill: killed $vgpr28
+; SI-NEXT: ; implicit-def: $vgpr28
+; SI-NEXT: ; kill: killed $vgpr28
+; SI-NEXT: ; implicit-def: $vgpr28
+; SI-NEXT: ; kill: killed $vgpr28
+; SI-NEXT: ; implicit-def: $vgpr28
; SI-NEXT: s_and_saveexec_b64 s[4:5], vcc
; SI-NEXT: s_xor_b64 s[4:5], exec, s[4:5]
; SI-NEXT: s_cbranch_execz .LBB104_2
; SI-NEXT: ; %bb.1: ; %cmp.false
-; SI-NEXT: v_lshrrev_b32_e32 v22, 16, v33
-; SI-NEXT: buffer_store_dword v22, off, s[0:3], s32 offset:244 ; 4-byte Folded Spill
+; SI-NEXT: v_lshrrev_b32_e32 v28, 16, v34
+; SI-NEXT: buffer_store_dword v28, off, s[0:3], s32 offset:240 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_lshrrev_b32_e32 v22, 16, v43
-; SI-NEXT: buffer_store_dword v22, off, s[0:3], s32 offset:396 ; 4-byte Folded Spill
+; SI-NEXT: v_lshrrev_b32_e32 v28, 16, v33
+; SI-NEXT: buffer_store_dword v28, off, s[0:3], s32 offset:400 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: buffer_load_dword v22, off, s[0:3], s32 offset:328 ; 4-byte Folded Reload
-; SI-NEXT: v_lshrrev_b32_e32 v7, 16, v7
+; SI-NEXT: buffer_load_dword v28, off, s[0:3], s32 offset:332 ; 4-byte Folded Reload
; SI-NEXT: v_lshrrev_b32_e32 v14, 16, v14
-; SI-NEXT: buffer_store_dword v7, off, s[0:3], s32 offset:364 ; 4-byte Folded Spill
-; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_lshrrev_b32_e32 v7, 16, v30
-; SI-NEXT: buffer_store_dword v14, off, s[0:3], s32 offset:384 ; 4-byte Folded Spill
+; SI-NEXT: v_lshrrev_b32_e32 v9, 16, v9
+; SI-NEXT: buffer_store_dword v14, off, s[0:3], s32 offset:392 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: buffer_load_dword v14, off, s[0:3], s32 offset:424 ; 4-byte Folded Reload
-; SI-NEXT: buffer_store_dword v7, off, s[0:3], s32 offset:312 ; 4-byte Folded Spill
+; SI-NEXT: buffer_load_dword v14, off, s[0:3], s32 offset:420 ; 4-byte Folded Reload
+; SI-NEXT: buffer_store_dword v9, off, s[0:3], s32 offset:372 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: buffer_load_dword v7, off, s[0:3], s32 offset:460 ; 4-byte Folded Reload
-; SI-NEXT: v_lshrrev_b32_e32 v9, 16, v9
-; SI-NEXT: buffer_store_dword v9, off, s[0:3], s32 offset:368 ; 4-byte Folded Spill
+; SI-NEXT: buffer_load_dword v9, off, s[0:3], s32 offset:460 ; 4-byte Folded Reload
; SI-NEXT: v_lshrrev_b32_e32 v12, 16, v12
-; SI-NEXT: buffer_store_dword v12, off, s[0:3], s32 offset:380 ; 4-byte Folded Spill
-; SI-NEXT: v_lshrrev_b32_e32 v11, 16, v11
-; SI-NEXT: buffer_store_dword v11, off, s[0:3], s32 offset:372 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v12, off, s[0:3], s32 offset:388 ; 4-byte Folded Spill
+; SI-NEXT: v_lshrrev_b32_e32 v10, 16, v10
+; SI-NEXT: buffer_store_dword v10, off, s[0:3], s32 offset:384 ; 4-byte Folded Spill
; SI-NEXT: v_lshrrev_b32_e32 v1, 16, v1
; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:276 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_lshrrev_b32_e32 v1, 16, v20
+; SI-NEXT: v_lshrrev_b32_e32 v1, 16, v23
; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:260 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_lshrrev_b32_e32 v1, 16, v26
-; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:332 ; 4-byte Folded Spill
-; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_lshrrev_b32_e32 v1, 16, v25
-; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:272 ; 4-byte Folded Spill
+; SI-NEXT: v_lshrrev_b32_e32 v1, 16, v30
+; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:336 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_lshrrev_b32_e32 v1, 16, v21
+; SI-NEXT: v_lshrrev_b32_e32 v1, 16, v24
; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:268 ; 4-byte Folded Spill
; SI-NEXT: ; implicit-def: $vgpr1
; SI-NEXT: ; kill: killed $vgpr1
@@ -236565,14 +236567,10 @@ define <64 x i16> @bitcast_v64bf16_to_v64i16(<64 x bfloat> %a, i32 %b) {
; SI-NEXT: ; implicit-def: $vgpr1
; SI-NEXT: ; kill: killed $vgpr1
; SI-NEXT: ; implicit-def: $vgpr1
-; SI-NEXT: ; kill: killed $vgpr1
-; SI-NEXT: ; implicit-def: $vgpr1
-; SI-NEXT: ; kill: killed $vgpr1
-; SI-NEXT: ; implicit-def: $vgpr1
; SI-NEXT: v_lshrrev_b32_e32 v2, 16, v2
; SI-NEXT: ; kill: killed $vgpr1
; SI-NEXT: ; implicit-def: $vgpr1
-; SI-NEXT: buffer_store_dword v2, off, s[0:3], s32 offset:348 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v2, off, s[0:3], s32 offset:352 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_lshrrev_b32_e32 v2, 16, v47
; SI-NEXT: ; kill: killed $vgpr1
@@ -236584,7 +236582,7 @@ define <64 x i16> @bitcast_v64bf16_to_v64i16(<64 x bfloat> %a, i32 %b) {
; SI-NEXT: ; implicit-def: $vgpr1
; SI-NEXT: buffer_store_dword v2, off, s[0:3], s32 offset:300 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_lshrrev_b32_e32 v2, 16, v29
+; SI-NEXT: v_lshrrev_b32_e32 v2, 16, v61
; SI-NEXT: ; kill: killed $vgpr1
; SI-NEXT: ; implicit-def: $vgpr1
; SI-NEXT: buffer_store_dword v2, off, s[0:3], s32 offset:248 ; 4-byte Folded Spill
@@ -236592,7 +236590,7 @@ define <64 x i16> @bitcast_v64bf16_to_v64i16(<64 x bfloat> %a, i32 %b) {
; SI-NEXT: v_lshrrev_b32_e32 v2, 16, v60
; SI-NEXT: ; kill: killed $vgpr1
; SI-NEXT: ; implicit-def: $vgpr1
-; SI-NEXT: buffer_store_dword v2, off, s[0:3], s32 offset:344 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v2, off, s[0:3], s32 offset:348 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_lshrrev_b32_e32 v2, 16, v57
; SI-NEXT: ; kill: killed $vgpr1
@@ -236604,265 +236602,277 @@ define <64 x i16> @bitcast_v64bf16_to_v64i16(<64 x bfloat> %a, i32 %b) {
; SI-NEXT: ; implicit-def: $vgpr1
; SI-NEXT: buffer_store_dword v2, off, s[0:3], s32 offset:292 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_lshrrev_b32_e32 v2, 16, v23
+; SI-NEXT: v_lshrrev_b32_e32 v2, 16, v29
; SI-NEXT: ; kill: killed $vgpr1
; SI-NEXT: ; implicit-def: $vgpr1
-; SI-NEXT: buffer_store_dword v2, off, s[0:3], s32 offset:252 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v2, off, s[0:3], s32 offset:264 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_lshrrev_b32_e32 v2, 16, v27
+; SI-NEXT: v_lshrrev_b32_e32 v2, 16, v16
; SI-NEXT: ; kill: killed $vgpr1
; SI-NEXT: ; implicit-def: $vgpr1
-; SI-NEXT: buffer_store_dword v2, off, s[0:3], s32 offset:340 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v2, off, s[0:3], s32 offset:344 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_lshrrev_b32_e32 v2, 16, v59
; SI-NEXT: ; kill: killed $vgpr1
; SI-NEXT: ; implicit-def: $vgpr1
-; SI-NEXT: v_lshrrev_b32_e32 v5, 16, v5
; SI-NEXT: buffer_store_dword v2, off, s[0:3], s32 offset:280 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_lshrrev_b32_e32 v2, 16, v58
; SI-NEXT: ; kill: killed $vgpr1
; SI-NEXT: ; implicit-def: $vgpr1
-; SI-NEXT: buffer_store_dword v5, off, s[0:3], s32 offset:356 ; 4-byte Folded Spill
-; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_lshrrev_b32_e32 v5, 16, v28
+; SI-NEXT: v_lshrrev_b32_e32 v7, 16, v7
+; SI-NEXT: v_lshrrev_b32_e32 v5, 16, v5
; SI-NEXT: buffer_store_dword v2, off, s[0:3], s32 offset:284 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_lshrrev_b32_e32 v2, 16, v16
+; SI-NEXT: v_lshrrev_b32_e32 v2, 16, v27
; SI-NEXT: ; kill: killed $vgpr1
; SI-NEXT: ; implicit-def: $vgpr1
-; SI-NEXT: v_lshrrev_b32_e32 v15, 16, v15
-; SI-NEXT: v_lshrrev_b32_e32 v10, 16, v10
+; SI-NEXT: buffer_store_dword v7, off, s[0:3], s32 offset:368 ; 4-byte Folded Spill
+; SI-NEXT: s_waitcnt expcnt(0)
+; SI-NEXT: v_lshrrev_b32_e32 v7, 16, v15
+; SI-NEXT: buffer_store_dword v5, off, s[0:3], s32 offset:360 ; 4-byte Folded Spill
+; SI-NEXT: s_waitcnt expcnt(0)
+; SI-NEXT: v_lshrrev_b32_e32 v5, 16, v25
+; SI-NEXT: buffer_store_dword v2, off, s[0:3], s32 offset:256 ; 4-byte Folded Spill
+; SI-NEXT: s_waitcnt expcnt(0)
+; SI-NEXT: v_lshrrev_b32_e32 v2, 16, v26
+; SI-NEXT: ; kill: killed $vgpr1
+; SI-NEXT: ; implicit-def: $vgpr1
+; SI-NEXT: v_lshrrev_b32_e32 v13, 16, v13
; SI-NEXT: v_lshrrev_b32_e32 v8, 16, v8
+; SI-NEXT: buffer_store_dword v7, off, s[0:3], s32 offset:312 ; 4-byte Folded Spill
+; SI-NEXT: s_waitcnt expcnt(0)
+; SI-NEXT: v_lshrrev_b32_e32 v7, 16, v40
; SI-NEXT: v_lshrrev_b32_e32 v6, 16, v6
; SI-NEXT: buffer_store_dword v5, off, s[0:3], s32 offset:304 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_lshrrev_b32_e32 v5, 16, v44
; SI-NEXT: v_lshrrev_b32_e32 v4, 16, v4
-; SI-NEXT: buffer_store_dword v2, off, s[0:3], s32 offset:264 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v2, off, s[0:3], s32 offset:340 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_lshrrev_b32_e32 v2, 16, v24
+; SI-NEXT: v_lshrrev_b32_e32 v2, 16, v3
; SI-NEXT: ; kill: killed $vgpr1
; SI-NEXT: ; implicit-def: $vgpr1
-; SI-NEXT: buffer_store_dword v15, off, s[0:3], s32 offset:216 ; 4-byte Folded Spill
-; SI-NEXT: buffer_store_dword v10, off, s[0:3], s32 offset:232 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v13, off, s[0:3], s32 offset:216 ; 4-byte Folded Spill
; SI-NEXT: buffer_store_dword v8, off, s[0:3], s32 offset:236 ; 4-byte Folded Spill
-; SI-NEXT: buffer_store_dword v6, off, s[0:3], s32 offset:240 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v7, off, s[0:3], s32 offset:316 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v6, off, s[0:3], s32 offset:252 ; 4-byte Folded Spill
; SI-NEXT: buffer_store_dword v5, off, s[0:3], s32 offset:308 ; 4-byte Folded Spill
-; SI-NEXT: buffer_store_dword v4, off, s[0:3], s32 offset:256 ; 4-byte Folded Spill
-; SI-NEXT: buffer_store_dword v2, off, s[0:3], s32 offset:336 ; 4-byte Folded Spill
-; SI-NEXT: v_lshrrev_b32_e32 v61, 16, v3
+; SI-NEXT: buffer_store_dword v4, off, s[0:3], s32 offset:244 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v2, off, s[0:3], s32 offset:272 ; 4-byte Folded Spill
+; SI-NEXT: v_lshrrev_b32_e32 v45, 16, v22
+; SI-NEXT: ; implicit-def: $vgpr34
; SI-NEXT: ; implicit-def: $vgpr33
-; SI-NEXT: ; implicit-def: $vgpr43
-; SI-NEXT: ; implicit-def: $vgpr15
-; SI-NEXT: ; implicit-def: $vgpr10
-; SI-NEXT: ; implicit-def: $vgpr8
-; SI-NEXT: ; implicit-def: $vgpr30
+; SI-NEXT: ; implicit-def: $vgpr13
; SI-NEXT: ; kill: killed $vgpr1
+; SI-NEXT: ; implicit-def: $vgpr8
+; SI-NEXT: ; implicit-def: $vgpr7
+; SI-NEXT: ; implicit-def: $vgpr15
+; SI-NEXT: ; implicit-def: $vgpr40
; SI-NEXT: ; implicit-def: $vgpr6
; SI-NEXT: ; implicit-def: $vgpr5
-; SI-NEXT: ; implicit-def: $vgpr28
+; SI-NEXT: ; implicit-def: $vgpr25
; SI-NEXT: ; implicit-def: $vgpr44
; SI-NEXT: ; implicit-def: $vgpr4
; SI-NEXT: ; implicit-def: $vgpr2
; SI-NEXT: ; implicit-def: $vgpr47
; SI-NEXT: ; implicit-def: $vgpr46
-; SI-NEXT: ; implicit-def: $vgpr29
+; SI-NEXT: ; implicit-def: $vgpr61
; SI-NEXT: ; implicit-def: $vgpr60
; SI-NEXT: ; implicit-def: $vgpr57
; SI-NEXT: ; implicit-def: $vgpr56
-; SI-NEXT: ; implicit-def: $vgpr23
-; SI-NEXT: ; implicit-def: $vgpr27
+; SI-NEXT: ; implicit-def: $vgpr29
+; SI-NEXT: ; implicit-def: $vgpr16
; SI-NEXT: ; implicit-def: $vgpr59
; SI-NEXT: ; implicit-def: $vgpr58
-; SI-NEXT: ; implicit-def: $vgpr16
-; SI-NEXT: ; implicit-def: $vgpr24
+; SI-NEXT: ; implicit-def: $vgpr27
+; SI-NEXT: ; implicit-def: $vgpr26
; SI-NEXT: ; implicit-def: $vgpr3
; SI-NEXT: ; implicit-def: $vgpr1
-; SI-NEXT: ; implicit-def: $vgpr20
-; SI-NEXT: ; implicit-def: $vgpr26
-; SI-NEXT: ; implicit-def: $vgpr25
-; SI-NEXT: ; implicit-def: $vgpr21
-; SI-NEXT: buffer_load_dword v9, off, s[0:3], s32 offset:456 ; 4-byte Folded Reload
-; SI-NEXT: buffer_load_dword v12, off, s[0:3], s32 offset:432 ; 4-byte Folded Reload
-; SI-NEXT: buffer_load_dword v11, off, s[0:3], s32 offset:448 ; 4-byte Folded Reload
+; SI-NEXT: ; implicit-def: $vgpr23
+; SI-NEXT: ; implicit-def: $vgpr30
+; SI-NEXT: ; implicit-def: $vgpr22
+; SI-NEXT: ; implicit-def: $vgpr24
+; SI-NEXT: buffer_load_dword v12, off, s[0:3], s32 offset:428 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v10, off, s[0:3], s32 offset:436 ; 4-byte Folded Reload
; SI-NEXT: s_waitcnt vmcnt(14)
-; SI-NEXT: v_lshrrev_b32_e32 v62, 16, v22
-; SI-NEXT: buffer_load_dword v22, off, s[0:3], s32 offset:324 ; 4-byte Folded Reload
-; SI-NEXT: v_lshrrev_b32_e32 v39, 16, v14
-; SI-NEXT: buffer_load_dword v14, off, s[0:3], s32 offset:420 ; 4-byte Folded Reload
-; SI-NEXT: v_lshrrev_b32_e32 v7, 16, v7
-; SI-NEXT: buffer_store_dword v7, off, s[0:3], s32 offset:316 ; 4-byte Folded Spill
-; SI-NEXT: ; implicit-def: $vgpr7
-; SI-NEXT: s_waitcnt vmcnt(5)
+; SI-NEXT: v_lshrrev_b32_e32 v62, 16, v28
+; SI-NEXT: buffer_load_dword v28, off, s[0:3], s32 offset:328 ; 4-byte Folded Reload
+; SI-NEXT: v_lshrrev_b32_e32 v49, 16, v14
+; SI-NEXT: buffer_load_dword v14, off, s[0:3], s32 offset:416 ; 4-byte Folded Reload
; SI-NEXT: v_lshrrev_b32_e32 v9, 16, v9
-; SI-NEXT: s_waitcnt vmcnt(4)
-; SI-NEXT: v_lshrrev_b32_e32 v37, 16, v12
-; SI-NEXT: buffer_load_dword v12, off, s[0:3], s32 offset:428 ; 4-byte Folded Reload
; SI-NEXT: buffer_store_dword v9, off, s[0:3], s32 offset:320 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: buffer_load_dword v9, off, s[0:3], s32 offset:452 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v9, off, s[0:3], s32 offset:456 ; 4-byte Folded Reload
; SI-NEXT: s_waitcnt vmcnt(5)
-; SI-NEXT: v_lshrrev_b32_e32 v63, 16, v22
-; SI-NEXT: v_lshrrev_b32_e32 v22, 16, v40
-; SI-NEXT: buffer_store_dword v22, off, s[0:3], s32 offset:204 ; 4-byte Folded Spill
+; SI-NEXT: v_lshrrev_b32_e32 v39, 16, v12
+; SI-NEXT: buffer_load_dword v12, off, s[0:3], s32 offset:424 ; 4-byte Folded Reload
+; SI-NEXT: s_waitcnt vmcnt(5)
+; SI-NEXT: v_lshrrev_b32_e32 v37, 16, v10
+; SI-NEXT: buffer_load_dword v10, off, s[0:3], s32 offset:432 ; 4-byte Folded Reload
+; SI-NEXT: s_waitcnt vmcnt(5)
+; SI-NEXT: v_lshrrev_b32_e32 v63, 16, v28
+; SI-NEXT: v_lshrrev_b32_e32 v28, 16, v55
+; SI-NEXT: buffer_store_dword v28, off, s[0:3], s32 offset:204 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_lshrrev_b32_e32 v22, 16, v32
-; SI-NEXT: buffer_store_dword v22, off, s[0:3], s32 offset:400 ; 4-byte Folded Spill
+; SI-NEXT: v_lshrrev_b32_e32 v28, 16, v54
+; SI-NEXT: buffer_store_dword v28, off, s[0:3], s32 offset:404 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: buffer_load_dword v22, off, s[0:3], s32 offset:360 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v28, off, s[0:3], s32 offset:364 ; 4-byte Folded Reload
; SI-NEXT: s_waitcnt vmcnt(7)
-; SI-NEXT: v_lshrrev_b32_e32 v48, 16, v14
-; SI-NEXT: v_lshrrev_b32_e32 v14, 16, v19
-; SI-NEXT: buffer_store_dword v14, off, s[0:3], s32 offset:220 ; 4-byte Folded Spill
-; SI-NEXT: ; implicit-def: $vgpr40
-; SI-NEXT: ; implicit-def: $vgpr32
+; SI-NEXT: v_lshrrev_b32_e32 v50, 16, v14
+; SI-NEXT: s_waitcnt vmcnt(5)
+; SI-NEXT: v_lshrrev_b32_e32 v9, 16, v9
+; SI-NEXT: buffer_store_dword v9, off, s[0:3], s32 offset:324 ; 4-byte Folded Spill
+; SI-NEXT: ; implicit-def: $vgpr55
+; SI-NEXT: ; implicit-def: $vgpr54
; SI-NEXT: ; implicit-def: $vgpr14
-; SI-NEXT: ; implicit-def: $vgpr19
-; SI-NEXT: s_waitcnt vmcnt(6)
-; SI-NEXT: v_lshrrev_b32_e32 v38, 16, v12
-; SI-NEXT: v_lshrrev_b32_e32 v12, 16, v13
-; SI-NEXT: buffer_store_dword v12, off, s[0:3], s32 offset:224 ; 4-byte Folded Spill
+; SI-NEXT: ; implicit-def: $vgpr9
+; SI-NEXT: s_waitcnt vmcnt(5)
+; SI-NEXT: v_lshrrev_b32_e32 v48, 16, v12
+; SI-NEXT: v_lshrrev_b32_e32 v12, 16, v21
+; SI-NEXT: buffer_store_dword v12, off, s[0:3], s32 offset:220 ; 4-byte Folded Spill
+; SI-NEXT: ; implicit-def: $vgpr12
+; SI-NEXT: ; implicit-def: $vgpr21
+; SI-NEXT: s_waitcnt vmcnt(5)
+; SI-NEXT: v_lshrrev_b32_e32 v38, 16, v10
+; SI-NEXT: v_lshrrev_b32_e32 v10, 16, v11
+; SI-NEXT: ; implicit-def: $vgpr11
+; SI-NEXT: buffer_store_dword v10, off, s[0:3], s32 offset:224 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_lshrrev_b32_e32 v12, 16, v18
-; SI-NEXT: buffer_store_dword v12, off, s[0:3], s32 offset:376 ; 4-byte Folded Spill
+; SI-NEXT: v_lshrrev_b32_e32 v10, 16, v20
+; SI-NEXT: ; implicit-def: $vgpr20
+; SI-NEXT: buffer_store_dword v10, off, s[0:3], s32 offset:380 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: buffer_load_dword v12, off, s[0:3], s32 offset:440 ; 4-byte Folded Reload
-; SI-NEXT: s_waitcnt vmcnt(7)
-; SI-NEXT: v_lshrrev_b32_e32 v45, 16, v9
-; SI-NEXT: ; implicit-def: $vgpr13
-; SI-NEXT: ; implicit-def: $vgpr18
-; SI-NEXT: ; implicit-def: $vgpr9
-; SI-NEXT: s_waitcnt vmcnt(4)
-; SI-NEXT: v_lshrrev_b32_e32 v41, 16, v22
-; SI-NEXT: buffer_load_dword v22, off, s[0:3], s32 offset:352 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v10, off, s[0:3], s32 offset:444 ; 4-byte Folded Reload
+; SI-NEXT: s_waitcnt vmcnt(5)
+; SI-NEXT: v_lshrrev_b32_e32 v41, 16, v28
+; SI-NEXT: buffer_load_dword v28, off, s[0:3], s32 offset:356 ; 4-byte Folded Reload
; SI-NEXT: s_waitcnt vmcnt(1)
-; SI-NEXT: v_lshrrev_b32_e32 v35, 16, v12
-; SI-NEXT: buffer_load_dword v12, off, s[0:3], s32 offset:436 ; 4-byte Folded Reload
+; SI-NEXT: v_lshrrev_b32_e32 v35, 16, v10
+; SI-NEXT: buffer_load_dword v10, off, s[0:3], s32 offset:440 ; 4-byte Folded Reload
; SI-NEXT: s_waitcnt vmcnt(1)
-; SI-NEXT: v_lshrrev_b32_e32 v42, 16, v22
-; SI-NEXT: v_lshrrev_b32_e32 v22, 16, v31
-; SI-NEXT: buffer_store_dword v22, off, s[0:3], s32 offset:208 ; 4-byte Folded Spill
+; SI-NEXT: v_lshrrev_b32_e32 v42, 16, v28
+; SI-NEXT: v_lshrrev_b32_e32 v28, 16, v32
+; SI-NEXT: buffer_store_dword v28, off, s[0:3], s32 offset:208 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_lshrrev_b32_e32 v22, 16, v55
-; SI-NEXT: buffer_store_dword v22, off, s[0:3], s32 offset:392 ; 4-byte Folded Spill
+; SI-NEXT: v_lshrrev_b32_e32 v28, 16, v31
+; SI-NEXT: buffer_store_dword v28, off, s[0:3], s32 offset:396 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: buffer_load_dword v22, off, s[0:3], s32 offset:408 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v28, off, s[0:3], s32 offset:412 ; 4-byte Folded Reload
+; SI-NEXT: ; implicit-def: $vgpr32
; SI-NEXT: ; implicit-def: $vgpr31
-; SI-NEXT: ; implicit-def: $vgpr55
; SI-NEXT: s_waitcnt vmcnt(3)
-; SI-NEXT: v_lshrrev_b32_e32 v36, 16, v12
-; SI-NEXT: v_lshrrev_b32_e32 v12, 16, v17
-; SI-NEXT: buffer_store_dword v12, off, s[0:3], s32 offset:228 ; 4-byte Folded Spill
-; SI-NEXT: ; implicit-def: $vgpr12
-; SI-NEXT: ; implicit-def: $vgpr17
-; SI-NEXT: s_waitcnt vmcnt(1)
-; SI-NEXT: v_lshrrev_b32_e32 v51, 16, v22
-; SI-NEXT: buffer_load_dword v22, off, s[0:3], s32 offset:404 ; 4-byte Folded Reload
-; SI-NEXT: s_waitcnt vmcnt(0)
-; SI-NEXT: v_lshrrev_b32_e32 v52, 16, v22
-; SI-NEXT: v_lshrrev_b32_e32 v22, 16, v54
-; SI-NEXT: buffer_store_dword v22, off, s[0:3], s32 offset:212 ; 4-byte Folded Spill
+; SI-NEXT: v_lshrrev_b32_e32 v36, 16, v10
+; SI-NEXT: v_lshrrev_b32_e32 v10, 16, v19
+; SI-NEXT: ; implicit-def: $vgpr19
+; SI-NEXT: buffer_store_dword v10, off, s[0:3], s32 offset:228 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_lshrrev_b32_e32 v22, 16, v53
-; SI-NEXT: buffer_store_dword v22, off, s[0:3], s32 offset:388 ; 4-byte Folded Spill
+; SI-NEXT: v_lshrrev_b32_e32 v10, 16, v18
+; SI-NEXT: ; implicit-def: $vgpr18
+; SI-NEXT: buffer_store_dword v10, off, s[0:3], s32 offset:376 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: buffer_load_dword v22, off, s[0:3], s32 offset:416 ; 4-byte Folded Reload
-; SI-NEXT: ; implicit-def: $vgpr54
-; SI-NEXT: ; implicit-def: $vgpr53
-; SI-NEXT: s_waitcnt vmcnt(0)
-; SI-NEXT: v_lshrrev_b32_e32 v49, 16, v22
-; SI-NEXT: buffer_load_dword v22, off, s[0:3], s32 offset:412 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v10, off, s[0:3], s32 offset:452 ; 4-byte Folded Reload
+; SI-NEXT: s_waitcnt vmcnt(3)
+; SI-NEXT: v_lshrrev_b32_e32 v51, 16, v28
+; SI-NEXT: buffer_load_dword v28, off, s[0:3], s32 offset:408 ; 4-byte Folded Reload
; SI-NEXT: s_waitcnt vmcnt(0)
-; SI-NEXT: v_lshrrev_b32_e32 v50, 16, v22
-; SI-NEXT: v_lshrrev_b32_e32 v22, 16, v11
-; SI-NEXT: buffer_load_dword v11, off, s[0:3], s32 offset:444 ; 4-byte Folded Reload
+; SI-NEXT: v_lshrrev_b32_e32 v52, 16, v28
+; SI-NEXT: v_lshrrev_b32_e32 v28, 16, v53
+; SI-NEXT: buffer_store_dword v28, off, s[0:3], s32 offset:212 ; 4-byte Folded Spill
+; SI-NEXT: ; implicit-def: $vgpr53
+; SI-NEXT: s_waitcnt expcnt(0)
+; SI-NEXT: v_lshrrev_b32_e32 v28, 16, v10
+; SI-NEXT: buffer_load_dword v10, off, s[0:3], s32 offset:448 ; 4-byte Folded Reload
; SI-NEXT: s_waitcnt vmcnt(0)
-; SI-NEXT: v_lshrrev_b32_e32 v34, 16, v11
-; SI-NEXT: ; implicit-def: $vgpr11
+; SI-NEXT: v_lshrrev_b32_e32 v43, 16, v10
+; SI-NEXT: v_lshrrev_b32_e32 v10, 16, v17
+; SI-NEXT: ; implicit-def: $vgpr17
+; SI-NEXT: buffer_store_dword v10, off, s[0:3], s32 offset:232 ; 4-byte Folded Spill
+; SI-NEXT: ; implicit-def: $vgpr10
; SI-NEXT: .LBB104_2: ; %Flow
; SI-NEXT: s_andn2_saveexec_b64 s[4:5], s[4:5]
; SI-NEXT: s_cbranch_execz .LBB104_4
; SI-NEXT: ; %bb.3: ; %cmp.true
-; SI-NEXT: v_and_b32_e32 v22, 0xffff0000, v33
-; SI-NEXT: v_and_b32_e32 v33, 0xffff0000, v43
+; SI-NEXT: v_and_b32_e32 v33, 0xffff0000, v33
+; SI-NEXT: v_and_b32_e32 v28, 0xffff0000, v34
; SI-NEXT: v_add_f32_e32 v33, 0x40c00000, v33
-; SI-NEXT: v_add_f32_e32 v22, 0x40c00000, v22
+; SI-NEXT: v_add_f32_e32 v28, 0x40c00000, v28
; SI-NEXT: v_lshrrev_b32_e32 v34, 16, v33
-; SI-NEXT: v_alignbit_b32 v22, v34, v22, 16
-; SI-NEXT: v_and_b32_e32 v32, 0xffff0000, v32
-; SI-NEXT: buffer_store_dword v22, off, s[0:3], s32 offset:244 ; 4-byte Folded Spill
+; SI-NEXT: v_alignbit_b32 v28, v34, v28, 16
+; SI-NEXT: v_and_b32_e32 v34, 0xffff0000, v54
+; SI-NEXT: buffer_store_dword v28, off, s[0:3], s32 offset:240 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_and_b32_e32 v22, 0xffff0000, v40
-; SI-NEXT: v_add_f32_e32 v32, 0x40c00000, v32
-; SI-NEXT: v_add_f32_e32 v22, 0x40c00000, v22
-; SI-NEXT: v_lshrrev_b32_e32 v34, 16, v32
-; SI-NEXT: v_alignbit_b32 v22, v34, v22, 16
-; SI-NEXT: buffer_store_dword v22, off, s[0:3], s32 offset:204 ; 4-byte Folded Spill
+; SI-NEXT: v_and_b32_e32 v28, 0xffff0000, v55
+; SI-NEXT: v_add_f32_e32 v34, 0x40c00000, v34
+; SI-NEXT: v_add_f32_e32 v28, 0x40c00000, v28
+; SI-NEXT: v_lshrrev_b32_e32 v35, 16, v34
+; SI-NEXT: v_alignbit_b32 v28, v35, v28, 16
+; SI-NEXT: v_and_b32_e32 v31, 0xffff0000, v31
+; SI-NEXT: buffer_store_dword v28, off, s[0:3], s32 offset:204 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_and_b32_e32 v22, 0xffff0000, v31
-; SI-NEXT: v_and_b32_e32 v31, 0xffff0000, v55
+; SI-NEXT: v_and_b32_e32 v28, 0xffff0000, v32
; SI-NEXT: v_add_f32_e32 v31, 0x40c00000, v31
-; SI-NEXT: v_add_f32_e32 v22, 0x40c00000, v22
-; SI-NEXT: v_lshrrev_b32_e32 v34, 16, v31
-; SI-NEXT: v_alignbit_b32 v22, v34, v22, 16
-; SI-NEXT: v_and_b32_e32 v34, 0xffff0000, v53
-; SI-NEXT: buffer_store_dword v22, off, s[0:3], s32 offset:208 ; 4-byte Folded Spill
-; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_and_b32_e32 v22, 0xffff0000, v54
-; SI-NEXT: v_add_f32_e32 v51, 0x40c00000, v34
-; SI-NEXT: v_add_f32_e32 v22, 0x40c00000, v22
-; SI-NEXT: v_lshrrev_b32_e32 v34, 16, v51
+; SI-NEXT: v_add_f32_e32 v28, 0x40c00000, v28
+; SI-NEXT: v_lshrrev_b32_e32 v32, 16, v31
+; SI-NEXT: v_alignbit_b32 v28, v32, v28, 16
; SI-NEXT: v_and_b32_e32 v14, 0xffff0000, v14
-; SI-NEXT: v_alignbit_b32 v22, v34, v22, 16
-; SI-NEXT: v_and_b32_e32 v15, 0xffff0000, v15
-; SI-NEXT: v_add_f32_e32 v14, 0x40c00000, v14
-; SI-NEXT: buffer_store_dword v22, off, s[0:3], s32 offset:212 ; 4-byte Folded Spill
-; SI-NEXT: v_add_f32_e32 v15, 0x40c00000, v15
+; SI-NEXT: buffer_store_dword v28, off, s[0:3], s32 offset:208 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_lshrrev_b32_e32 v22, 16, v14
-; SI-NEXT: v_alignbit_b32 v15, v22, v15, 16
+; SI-NEXT: v_and_b32_e32 v28, 0xffff0000, v53
+; SI-NEXT: v_add_f32_e32 v14, 0x40c00000, v14
+; SI-NEXT: v_add_f32_e32 v28, 0x40c00000, v28
+; SI-NEXT: v_lshrrev_b32_e32 v32, 16, v14
; SI-NEXT: v_and_b32_e32 v12, 0xffff0000, v12
-; SI-NEXT: buffer_store_dword v15, off, s[0:3], s32 offset:216 ; 4-byte Folded Spill
-; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_and_b32_e32 v15, 0xffff0000, v19
+; SI-NEXT: v_alignbit_b32 v28, v32, v28, 16
+; SI-NEXT: v_and_b32_e32 v13, 0xffff0000, v13
; SI-NEXT: v_add_f32_e32 v12, 0x40c00000, v12
-; SI-NEXT: v_add_f32_e32 v15, 0x40c00000, v15
-; SI-NEXT: v_lshrrev_b32_e32 v19, 16, v12
-; SI-NEXT: v_alignbit_b32 v15, v19, v15, 16
-; SI-NEXT: buffer_store_dword v15, off, s[0:3], s32 offset:220 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v28, off, s[0:3], s32 offset:212 ; 4-byte Folded Spill
+; SI-NEXT: v_add_f32_e32 v13, 0x40c00000, v13
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_and_b32_e32 v15, 0xffff0000, v18
-; SI-NEXT: v_and_b32_e32 v13, 0xffff0000, v13
-; SI-NEXT: v_add_f32_e32 v15, 0x40c00000, v15
+; SI-NEXT: v_lshrrev_b32_e32 v28, 16, v12
+; SI-NEXT: v_alignbit_b32 v13, v28, v13, 16
+; SI-NEXT: v_and_b32_e32 v10, 0xffff0000, v10
+; SI-NEXT: buffer_store_dword v13, off, s[0:3], s32 offset:216 ; 4-byte Folded Spill
+; SI-NEXT: s_waitcnt expcnt(0)
+; SI-NEXT: v_and_b32_e32 v13, 0xffff0000, v21
+; SI-NEXT: v_add_f32_e32 v10, 0x40c00000, v10
; SI-NEXT: v_add_f32_e32 v13, 0x40c00000, v13
-; SI-NEXT: v_lshrrev_b32_e32 v18, 16, v15
-; SI-NEXT: v_alignbit_b32 v13, v18, v13, 16
+; SI-NEXT: v_lshrrev_b32_e32 v21, 16, v10
+; SI-NEXT: v_alignbit_b32 v13, v21, v13, 16
+; SI-NEXT: buffer_store_dword v13, off, s[0:3], s32 offset:220 ; 4-byte Folded Spill
+; SI-NEXT: s_waitcnt expcnt(0)
+; SI-NEXT: v_and_b32_e32 v13, 0xffff0000, v20
; SI-NEXT: v_and_b32_e32 v11, 0xffff0000, v11
-; SI-NEXT: buffer_store_dword v13, off, s[0:3], s32 offset:224 ; 4-byte Folded Spill
+; SI-NEXT: v_add_f32_e32 v13, 0x40c00000, v13
+; SI-NEXT: v_add_f32_e32 v11, 0x40c00000, v11
+; SI-NEXT: v_lshrrev_b32_e32 v20, 16, v13
+; SI-NEXT: v_alignbit_b32 v11, v20, v11, 16
+; SI-NEXT: v_and_b32_e32 v18, 0xffff0000, v18
+; SI-NEXT: buffer_store_dword v11, off, s[0:3], s32 offset:224 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_and_b32_e32 v13, 0xffff0000, v17
+; SI-NEXT: v_and_b32_e32 v11, 0xffff0000, v19
+; SI-NEXT: v_add_f32_e32 v18, 0x40c00000, v18
; SI-NEXT: v_add_f32_e32 v11, 0x40c00000, v11
-; SI-NEXT: v_add_f32_e32 v13, 0x40c00000, v13
-; SI-NEXT: v_lshrrev_b32_e32 v17, 16, v11
+; SI-NEXT: v_lshrrev_b32_e32 v19, 16, v18
+; SI-NEXT: v_alignbit_b32 v11, v19, v11, 16
; SI-NEXT: v_and_b32_e32 v9, 0xffff0000, v9
-; SI-NEXT: v_alignbit_b32 v13, v17, v13, 16
-; SI-NEXT: v_and_b32_e32 v10, 0xffff0000, v10
-; SI-NEXT: v_add_f32_e32 v9, 0x40c00000, v9
-; SI-NEXT: buffer_store_dword v13, off, s[0:3], s32 offset:228 ; 4-byte Folded Spill
-; SI-NEXT: v_add_f32_e32 v10, 0x40c00000, v10
+; SI-NEXT: buffer_store_dword v11, off, s[0:3], s32 offset:228 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_lshrrev_b32_e32 v13, 16, v9
+; SI-NEXT: v_and_b32_e32 v11, 0xffff0000, v17
+; SI-NEXT: v_add_f32_e32 v9, 0x40c00000, v9
+; SI-NEXT: v_add_f32_e32 v11, 0x40c00000, v11
+; SI-NEXT: v_lshrrev_b32_e32 v17, 16, v9
; SI-NEXT: v_and_b32_e32 v7, 0xffff0000, v7
-; SI-NEXT: v_alignbit_b32 v10, v13, v10, 16
+; SI-NEXT: v_alignbit_b32 v11, v17, v11, 16
; SI-NEXT: v_and_b32_e32 v8, 0xffff0000, v8
; SI-NEXT: v_add_f32_e32 v7, 0x40c00000, v7
-; SI-NEXT: buffer_store_dword v10, off, s[0:3], s32 offset:232 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v11, off, s[0:3], s32 offset:232 ; 4-byte Folded Spill
; SI-NEXT: v_add_f32_e32 v8, 0x40c00000, v8
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_lshrrev_b32_e32 v10, 16, v7
+; SI-NEXT: v_lshrrev_b32_e32 v11, 16, v7
; SI-NEXT: v_and_b32_e32 v5, 0xffff0000, v5
-; SI-NEXT: v_alignbit_b32 v8, v10, v8, 16
+; SI-NEXT: v_alignbit_b32 v8, v11, v8, 16
; SI-NEXT: v_and_b32_e32 v6, 0xffff0000, v6
; SI-NEXT: v_add_f32_e32 v5, 0x40c00000, v5
; SI-NEXT: buffer_store_dword v8, off, s[0:3], s32 offset:236 ; 4-byte Folded Spill
@@ -236873,254 +236883,256 @@ define <64 x i16> @bitcast_v64bf16_to_v64i16(<64 x bfloat> %a, i32 %b) {
; SI-NEXT: v_alignbit_b32 v6, v8, v6, 16
; SI-NEXT: v_and_b32_e32 v4, 0xffff0000, v4
; SI-NEXT: v_add_f32_e32 v2, 0x40c00000, v2
-; SI-NEXT: buffer_store_dword v6, off, s[0:3], s32 offset:240 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v6, off, s[0:3], s32 offset:252 ; 4-byte Folded Spill
; SI-NEXT: v_add_f32_e32 v4, 0x40c00000, v4
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_lshrrev_b32_e32 v6, 16, v2
; SI-NEXT: v_alignbit_b32 v4, v6, v4, 16
; SI-NEXT: v_and_b32_e32 v6, 0xffff0000, v60
-; SI-NEXT: buffer_store_dword v4, off, s[0:3], s32 offset:256 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v4, off, s[0:3], s32 offset:244 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_and_b32_e32 v4, 0xffff0000, v29
+; SI-NEXT: v_and_b32_e32 v4, 0xffff0000, v61
; SI-NEXT: v_add_f32_e32 v6, 0x40c00000, v6
; SI-NEXT: v_add_f32_e32 v4, 0x40c00000, v4
; SI-NEXT: v_lshrrev_b32_e32 v8, 16, v6
; SI-NEXT: v_alignbit_b32 v4, v8, v4, 16
-; SI-NEXT: v_and_b32_e32 v8, 0xffff0000, v27
+; SI-NEXT: v_and_b32_e32 v8, 0xffff0000, v16
; SI-NEXT: buffer_store_dword v4, off, s[0:3], s32 offset:248 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_and_b32_e32 v4, 0xffff0000, v23
+; SI-NEXT: v_and_b32_e32 v4, 0xffff0000, v29
; SI-NEXT: v_add_f32_e32 v8, 0x40c00000, v8
; SI-NEXT: v_add_f32_e32 v4, 0x40c00000, v4
-; SI-NEXT: v_lshrrev_b32_e32 v10, 16, v8
+; SI-NEXT: v_lshrrev_b32_e32 v11, 16, v8
; SI-NEXT: v_and_b32_e32 v1, 0xffff0000, v1
-; SI-NEXT: v_alignbit_b32 v4, v10, v4, 16
-; SI-NEXT: v_and_b32_e32 v10, 0xffff0000, v24
+; SI-NEXT: v_alignbit_b32 v4, v11, v4, 16
+; SI-NEXT: v_and_b32_e32 v11, 0xffff0000, v26
; SI-NEXT: v_and_b32_e32 v3, 0xffff0000, v3
; SI-NEXT: v_add_f32_e32 v1, 0x40c00000, v1
-; SI-NEXT: v_add_f32_e32 v10, 0x40c00000, v10
+; SI-NEXT: v_add_f32_e32 v11, 0x40c00000, v11
; SI-NEXT: v_add_f32_e32 v3, 0x40c00000, v3
; SI-NEXT: v_lshrrev_b32_e32 v1, 16, v1
-; SI-NEXT: v_lshrrev_b32_e32 v13, 16, v10
+; SI-NEXT: v_lshrrev_b32_e32 v16, 16, v11
; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:276 ; 4-byte Folded Spill
-; SI-NEXT: v_alignbit_b32 v61, v1, v3, 16
+; SI-NEXT: v_alignbit_b32 v19, v1, v3, 16
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_and_b32_e32 v1, 0xffff0000, v10
-; SI-NEXT: v_and_b32_e32 v10, 0xffff0000, v58
+; SI-NEXT: v_and_b32_e32 v1, 0xffff0000, v11
+; SI-NEXT: v_and_b32_e32 v11, 0xffff0000, v58
; SI-NEXT: v_and_b32_e32 v3, 0xffff0000, v59
-; SI-NEXT: v_add_f32_e32 v10, 0x40c00000, v10
+; SI-NEXT: v_add_f32_e32 v11, 0x40c00000, v11
; SI-NEXT: v_add_f32_e32 v3, 0x40c00000, v3
-; SI-NEXT: v_lshrrev_b32_e32 v10, 16, v10
-; SI-NEXT: buffer_store_dword v4, off, s[0:3], s32 offset:252 ; 4-byte Folded Spill
+; SI-NEXT: v_lshrrev_b32_e32 v11, 16, v11
+; SI-NEXT: buffer_store_dword v4, off, s[0:3], s32 offset:264 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_and_b32_e32 v4, 0xffff0000, v16
-; SI-NEXT: buffer_store_dword v10, off, s[0:3], s32 offset:284 ; 4-byte Folded Spill
-; SI-NEXT: v_alignbit_b32 v19, v10, v3, 16
+; SI-NEXT: v_and_b32_e32 v4, 0xffff0000, v27
+; SI-NEXT: buffer_store_dword v11, off, s[0:3], s32 offset:284 ; 4-byte Folded Spill
+; SI-NEXT: v_alignbit_b32 v20, v11, v3, 16
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_and_b32_e32 v10, 0xffff0000, v56
+; SI-NEXT: v_and_b32_e32 v11, 0xffff0000, v56
; SI-NEXT: v_add_f32_e32 v4, 0x40c00000, v4
; SI-NEXT: v_and_b32_e32 v3, 0xffff0000, v8
; SI-NEXT: v_and_b32_e32 v8, 0xffff0000, v57
-; SI-NEXT: v_add_f32_e32 v10, 0x40c00000, v10
-; SI-NEXT: v_alignbit_b32 v4, v13, v4, 16
+; SI-NEXT: v_add_f32_e32 v11, 0x40c00000, v11
+; SI-NEXT: v_alignbit_b32 v4, v16, v4, 16
+; SI-NEXT: v_and_b32_e32 v16, 0xffff0000, v30
; SI-NEXT: v_add_f32_e32 v8, 0x40c00000, v8
-; SI-NEXT: v_lshrrev_b32_e32 v10, 16, v10
-; SI-NEXT: buffer_store_dword v4, off, s[0:3], s32 offset:264 ; 4-byte Folded Spill
+; SI-NEXT: v_lshrrev_b32_e32 v11, 16, v11
+; SI-NEXT: buffer_store_dword v4, off, s[0:3], s32 offset:256 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_and_b32_e32 v4, 0xffff0000, v20
-; SI-NEXT: v_and_b32_e32 v13, 0xffff0000, v26
-; SI-NEXT: buffer_store_dword v10, off, s[0:3], s32 offset:292 ; 4-byte Folded Spill
-; SI-NEXT: v_alignbit_b32 v20, v10, v8, 16
+; SI-NEXT: v_and_b32_e32 v4, 0xffff0000, v23
+; SI-NEXT: v_add_f32_e32 v16, 0x40c00000, v16
+; SI-NEXT: buffer_store_dword v11, off, s[0:3], s32 offset:292 ; 4-byte Folded Spill
+; SI-NEXT: v_alignbit_b32 v21, v11, v8, 16
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_and_b32_e32 v10, 0xffff0000, v46
-; SI-NEXT: v_add_f32_e32 v13, 0x40c00000, v13
-; SI-NEXT: v_and_b32_e32 v8, 0xffff0000, v47
-; SI-NEXT: v_add_f32_e32 v10, 0x40c00000, v10
+; SI-NEXT: v_and_b32_e32 v11, 0xffff0000, v46
; SI-NEXT: v_add_f32_e32 v4, 0x40c00000, v4
-; SI-NEXT: v_lshrrev_b32_e32 v16, 16, v13
+; SI-NEXT: v_lshrrev_b32_e32 v17, 16, v16
+; SI-NEXT: v_and_b32_e32 v8, 0xffff0000, v47
+; SI-NEXT: v_add_f32_e32 v11, 0x40c00000, v11
+; SI-NEXT: v_alignbit_b32 v4, v17, v4, 16
; SI-NEXT: v_add_f32_e32 v8, 0x40c00000, v8
-; SI-NEXT: v_lshrrev_b32_e32 v10, 16, v10
-; SI-NEXT: v_alignbit_b32 v4, v16, v4, 16
-; SI-NEXT: v_and_b32_e32 v16, 0xffff0000, v21
-; SI-NEXT: buffer_store_dword v10, off, s[0:3], s32 offset:300 ; 4-byte Folded Spill
-; SI-NEXT: v_alignbit_b32 v21, v10, v8, 16
+; SI-NEXT: v_lshrrev_b32_e32 v11, 16, v11
+; SI-NEXT: buffer_store_dword v4, off, s[0:3], s32 offset:260 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_and_b32_e32 v10, 0xffff0000, v44
-; SI-NEXT: v_and_b32_e32 v8, 0xffff0000, v28
-; SI-NEXT: v_add_f32_e32 v10, 0x40c00000, v10
+; SI-NEXT: v_and_b32_e32 v4, 0xffff0000, v22
+; SI-NEXT: buffer_store_dword v11, off, s[0:3], s32 offset:300 ; 4-byte Folded Spill
+; SI-NEXT: v_alignbit_b32 v22, v11, v8, 16
+; SI-NEXT: s_waitcnt expcnt(0)
+; SI-NEXT: v_and_b32_e32 v11, 0xffff0000, v44
+; SI-NEXT: v_and_b32_e32 v8, 0xffff0000, v25
+; SI-NEXT: v_add_f32_e32 v11, 0x40c00000, v11
; SI-NEXT: v_add_f32_e32 v8, 0x40c00000, v8
-; SI-NEXT: v_lshrrev_b32_e32 v10, 16, v10
-; SI-NEXT: buffer_store_dword v10, off, s[0:3], s32 offset:308 ; 4-byte Folded Spill
-; SI-NEXT: v_alignbit_b32 v23, v10, v8, 16
+; SI-NEXT: v_lshrrev_b32_e32 v11, 16, v11
+; SI-NEXT: buffer_store_dword v11, off, s[0:3], s32 offset:308 ; 4-byte Folded Spill
+; SI-NEXT: v_alignbit_b32 v23, v11, v8, 16
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: buffer_load_dword v10, off, s[0:3], s32 offset:460 ; 4-byte Folded Reload
-; SI-NEXT: v_and_b32_e32 v8, 0xffff0000, v30
+; SI-NEXT: v_and_b32_e32 v11, 0xffff0000, v40
+; SI-NEXT: v_and_b32_e32 v8, 0xffff0000, v15
+; SI-NEXT: v_add_f32_e32 v11, 0x40c00000, v11
; SI-NEXT: v_add_f32_e32 v8, 0x40c00000, v8
-; SI-NEXT: buffer_store_dword v4, off, s[0:3], s32 offset:260 ; 4-byte Folded Spill
+; SI-NEXT: v_lshrrev_b32_e32 v11, 16, v11
+; SI-NEXT: v_and_b32_e32 v17, 0xffff0000, v24
+; SI-NEXT: buffer_store_dword v11, off, s[0:3], s32 offset:316 ; 4-byte Folded Spill
+; SI-NEXT: v_alignbit_b32 v24, v11, v8, 16
+; SI-NEXT: buffer_load_dword v8, off, s[0:3], s32 offset:460 ; 4-byte Folded Reload
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_and_b32_e32 v4, 0xffff0000, v25
-; SI-NEXT: v_add_f32_e32 v16, 0x40c00000, v16
+; SI-NEXT: buffer_load_dword v11, off, s[0:3], s32 offset:456 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v15, off, s[0:3], s32 offset:440 ; 4-byte Folded Reload
+; SI-NEXT: v_add_f32_e32 v17, 0x40c00000, v17
; SI-NEXT: v_add_f32_e32 v4, 0x40c00000, v4
-; SI-NEXT: v_lshrrev_b32_e32 v16, 16, v16
-; SI-NEXT: v_alignbit_b32 v18, v16, v4, 16
-; SI-NEXT: v_and_b32_e32 v4, 0xffff0000, v13
-; SI-NEXT: buffer_store_dword v16, off, s[0:3], s32 offset:268 ; 4-byte Folded Spill
+; SI-NEXT: v_lshrrev_b32_e32 v17, 16, v17
+; SI-NEXT: v_alignbit_b32 v45, v17, v4, 16
+; SI-NEXT: v_and_b32_e32 v4, 0xffff0000, v16
+; SI-NEXT: buffer_store_dword v17, off, s[0:3], s32 offset:268 ; 4-byte Folded Spill
+; SI-NEXT: v_and_b32_e32 v10, 0xffff0000, v10
; SI-NEXT: v_and_b32_e32 v2, 0xffff0000, v2
; SI-NEXT: v_and_b32_e32 v5, 0xffff0000, v5
; SI-NEXT: v_and_b32_e32 v7, 0xffff0000, v7
+; SI-NEXT: v_and_b32_e32 v12, 0xffff0000, v12
; SI-NEXT: v_and_b32_e32 v6, 0xffff0000, v6
+; SI-NEXT: v_alignbit_b32 v7, v24, v7, 16
; SI-NEXT: v_alignbit_b32 v5, v23, v5, 16
-; SI-NEXT: v_alignbit_b32 v2, v21, v2, 16
-; SI-NEXT: v_alignbit_b32 v1, v61, v1, 16
-; SI-NEXT: buffer_load_dword v13, off, s[0:3], s32 offset:428 ; 4-byte Folded Reload
+; SI-NEXT: v_alignbit_b32 v2, v22, v2, 16
+; SI-NEXT: v_alignbit_b32 v1, v19, v1, 16
+; SI-NEXT: buffer_load_dword v16, off, s[0:3], s32 offset:356 ; 4-byte Folded Reload
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: buffer_load_dword v16, off, s[0:3], s32 offset:352 ; 4-byte Folded Reload
-; SI-NEXT: buffer_load_dword v17, off, s[0:3], s32 offset:324 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v17, off, s[0:3], s32 offset:328 ; 4-byte Folded Reload
; SI-NEXT: s_waitcnt vmcnt(5)
-; SI-NEXT: v_and_b32_e32 v10, 0xffff0000, v10
-; SI-NEXT: v_add_f32_e32 v10, 0x40c00000, v10
-; SI-NEXT: v_lshrrev_b32_e32 v10, 16, v10
-; SI-NEXT: buffer_store_dword v10, off, s[0:3], s32 offset:316 ; 4-byte Folded Spill
-; SI-NEXT: v_alignbit_b32 v24, v10, v8, 16
+; SI-NEXT: v_and_b32_e32 v8, 0xffff0000, v8
+; SI-NEXT: s_waitcnt vmcnt(4)
+; SI-NEXT: v_and_b32_e32 v11, 0xffff0000, v11
+; SI-NEXT: v_add_f32_e32 v11, 0x40c00000, v11
+; SI-NEXT: v_add_f32_e32 v8, 0x40c00000, v8
+; SI-NEXT: v_lshrrev_b32_e32 v11, 16, v11
+; SI-NEXT: buffer_store_dword v11, off, s[0:3], s32 offset:324 ; 4-byte Folded Spill
+; SI-NEXT: v_alignbit_b32 v25, v11, v8, 16
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: buffer_load_dword v10, off, s[0:3], s32 offset:452 ; 4-byte Folded Reload
-; SI-NEXT: buffer_load_dword v8, off, s[0:3], s32 offset:456 ; 4-byte Folded Reload
-; SI-NEXT: v_alignbit_b32 v7, v24, v7, 16
+; SI-NEXT: buffer_load_dword v11, off, s[0:3], s32 offset:448 ; 4-byte Folded Reload
; SI-NEXT: s_waitcnt vmcnt(5)
-; SI-NEXT: v_and_b32_e32 v13, 0xffff0000, v13
-; SI-NEXT: v_add_f32_e32 v13, 0x40c00000, v13
-; SI-NEXT: v_lshrrev_b32_e32 v38, 16, v13
-; SI-NEXT: buffer_load_dword v13, off, s[0:3], s32 offset:420 ; 4-byte Folded Reload
+; SI-NEXT: v_and_b32_e32 v15, 0xffff0000, v15
+; SI-NEXT: v_add_f32_e32 v15, 0x40c00000, v15
+; SI-NEXT: v_lshrrev_b32_e32 v36, 16, v15
+; SI-NEXT: buffer_load_dword v15, off, s[0:3], s32 offset:432 ; 4-byte Folded Reload
+; SI-NEXT: v_and_b32_e32 v8, 0xffff0000, v9
+; SI-NEXT: buffer_load_dword v9, off, s[0:3], s32 offset:452 ; 4-byte Folded Reload
+; SI-NEXT: v_alignbit_b32 v8, v25, v8, 16
; SI-NEXT: s_waitcnt vmcnt(5)
; SI-NEXT: v_and_b32_e32 v16, 0xffff0000, v16
; SI-NEXT: v_add_f32_e32 v16, 0x40c00000, v16
; SI-NEXT: v_lshrrev_b32_e32 v42, 16, v16
-; SI-NEXT: buffer_load_dword v16, off, s[0:3], s32 offset:328 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v16, off, s[0:3], s32 offset:332 ; 4-byte Folded Reload
; SI-NEXT: s_waitcnt vmcnt(5)
; SI-NEXT: v_and_b32_e32 v17, 0xffff0000, v17
; SI-NEXT: v_add_f32_e32 v17, 0x40c00000, v17
; SI-NEXT: v_lshrrev_b32_e32 v63, 16, v17
; SI-NEXT: s_waitcnt vmcnt(3)
-; SI-NEXT: v_and_b32_e32 v10, 0xffff0000, v10
-; SI-NEXT: v_add_f32_e32 v10, 0x40c00000, v10
-; SI-NEXT: v_lshrrev_b32_e32 v45, 16, v10
-; SI-NEXT: buffer_load_dword v10, off, s[0:3], s32 offset:444 ; 4-byte Folded Reload
-; SI-NEXT: s_waitcnt vmcnt(3)
-; SI-NEXT: v_and_b32_e32 v8, 0xffff0000, v8
-; SI-NEXT: v_add_f32_e32 v8, 0x40c00000, v8
-; SI-NEXT: v_alignbit_b32 v25, v45, v8, 16
-; SI-NEXT: v_and_b32_e32 v8, 0xffff0000, v9
-; SI-NEXT: buffer_load_dword v9, off, s[0:3], s32 offset:448 ; 4-byte Folded Reload
-; SI-NEXT: v_alignbit_b32 v8, v25, v8, 16
-; SI-NEXT: s_waitcnt vmcnt(3)
-; SI-NEXT: v_and_b32_e32 v13, 0xffff0000, v13
-; SI-NEXT: v_add_f32_e32 v13, 0x40c00000, v13
-; SI-NEXT: v_lshrrev_b32_e32 v48, 16, v13
-; SI-NEXT: buffer_load_dword v13, off, s[0:3], s32 offset:416 ; 4-byte Folded Reload
+; SI-NEXT: v_and_b32_e32 v11, 0xffff0000, v11
+; SI-NEXT: v_add_f32_e32 v11, 0x40c00000, v11
+; SI-NEXT: v_lshrrev_b32_e32 v43, 16, v11
+; SI-NEXT: buffer_load_dword v11, off, s[0:3], s32 offset:444 ; 4-byte Folded Reload
; SI-NEXT: s_waitcnt vmcnt(3)
+; SI-NEXT: v_and_b32_e32 v15, 0xffff0000, v15
+; SI-NEXT: v_add_f32_e32 v15, 0x40c00000, v15
+; SI-NEXT: v_lshrrev_b32_e32 v38, 16, v15
+; SI-NEXT: buffer_load_dword v15, off, s[0:3], s32 offset:424 ; 4-byte Folded Reload
+; SI-NEXT: s_waitcnt vmcnt(2)
; SI-NEXT: v_and_b32_e32 v16, 0xffff0000, v16
; SI-NEXT: v_add_f32_e32 v16, 0x40c00000, v16
; SI-NEXT: v_alignbit_b32 v62, v63, v16, 16
; SI-NEXT: v_and_b32_e32 v16, 0xffff0000, v33
; SI-NEXT: v_alignbit_b32 v16, v62, v16, 16
-; SI-NEXT: buffer_store_dword v16, off, s[0:3], s32 offset:396 ; 4-byte Folded Spill
-; SI-NEXT: s_waitcnt vmcnt(3)
-; SI-NEXT: v_and_b32_e32 v10, 0xffff0000, v10
-; SI-NEXT: v_add_f32_e32 v10, 0x40c00000, v10
-; SI-NEXT: v_lshrrev_b32_e32 v34, 16, v10
-; SI-NEXT: buffer_load_dword v10, off, s[0:3], s32 offset:440 ; 4-byte Folded Reload
-; SI-NEXT: s_waitcnt vmcnt(3)
-; SI-NEXT: v_and_b32_e32 v9, 0xffff0000, v9
-; SI-NEXT: v_add_f32_e32 v9, 0x40c00000, v9
-; SI-NEXT: v_alignbit_b32 v22, v34, v9, 16
-; SI-NEXT: v_and_b32_e32 v9, 0xffff0000, v11
-; SI-NEXT: buffer_load_dword v11, off, s[0:3], s32 offset:436 ; 4-byte Folded Reload
-; SI-NEXT: v_alignbit_b32 v9, v22, v9, 16
-; SI-NEXT: s_waitcnt vmcnt(0)
-; SI-NEXT: v_and_b32_e32 v11, 0xffff0000, v11
-; SI-NEXT: v_add_f32_e32 v11, 0x40c00000, v11
-; SI-NEXT: v_lshrrev_b32_e32 v36, 16, v11
-; SI-NEXT: buffer_load_dword v11, off, s[0:3], s32 offset:432 ; 4-byte Folded Reload
-; SI-NEXT: s_waitcnt vmcnt(0)
+; SI-NEXT: buffer_store_dword v16, off, s[0:3], s32 offset:400 ; 4-byte Folded Spill
+; SI-NEXT: s_waitcnt vmcnt(2)
; SI-NEXT: v_and_b32_e32 v11, 0xffff0000, v11
; SI-NEXT: v_add_f32_e32 v11, 0x40c00000, v11
-; SI-NEXT: v_alignbit_b32 v37, v38, v11, 16
-; SI-NEXT: v_and_b32_e32 v11, 0xffff0000, v12
-; SI-NEXT: buffer_load_dword v12, off, s[0:3], s32 offset:424 ; 4-byte Folded Reload
-; SI-NEXT: v_alignbit_b32 v11, v37, v11, 16
-; SI-NEXT: buffer_store_dword v11, off, s[0:3], s32 offset:380 ; 4-byte Folded Spill
+; SI-NEXT: v_alignbit_b32 v35, v36, v11, 16
+; SI-NEXT: v_and_b32_e32 v11, 0xffff0000, v13
+; SI-NEXT: buffer_load_dword v13, off, s[0:3], s32 offset:436 ; 4-byte Folded Reload
+; SI-NEXT: s_waitcnt vmcnt(2)
+; SI-NEXT: v_and_b32_e32 v15, 0xffff0000, v15
+; SI-NEXT: v_add_f32_e32 v15, 0x40c00000, v15
+; SI-NEXT: v_lshrrev_b32_e32 v48, 16, v15
+; SI-NEXT: buffer_load_dword v15, off, s[0:3], s32 offset:416 ; 4-byte Folded Reload
; SI-NEXT: s_waitcnt vmcnt(1)
-; SI-NEXT: v_and_b32_e32 v12, 0xffff0000, v12
-; SI-NEXT: v_add_f32_e32 v12, 0x40c00000, v12
-; SI-NEXT: v_alignbit_b32 v39, v48, v12, 16
-; SI-NEXT: v_and_b32_e32 v12, 0xffff0000, v14
-; SI-NEXT: buffer_load_dword v14, off, s[0:3], s32 offset:412 ; 4-byte Folded Reload
-; SI-NEXT: v_and_b32_e32 v10, 0xffff0000, v10
-; SI-NEXT: v_add_f32_e32 v10, 0x40c00000, v10
-; SI-NEXT: v_alignbit_b32 v35, v36, v10, 16
-; SI-NEXT: v_and_b32_e32 v10, 0xffff0000, v15
-; SI-NEXT: buffer_load_dword v15, off, s[0:3], s32 offset:404 ; 4-byte Folded Reload
-; SI-NEXT: v_alignbit_b32 v12, v39, v12, 16
-; SI-NEXT: v_alignbit_b32 v10, v35, v10, 16
-; SI-NEXT: buffer_store_dword v12, off, s[0:3], s32 offset:384 ; 4-byte Folded Spill
-; SI-NEXT: buffer_store_dword v10, off, s[0:3], s32 offset:376 ; 4-byte Folded Spill
+; SI-NEXT: v_and_b32_e32 v13, 0xffff0000, v13
+; SI-NEXT: v_add_f32_e32 v13, 0x40c00000, v13
+; SI-NEXT: v_alignbit_b32 v37, v38, v13, 16
+; SI-NEXT: buffer_load_dword v13, off, s[0:3], s32 offset:428 ; 4-byte Folded Reload
+; SI-NEXT: s_waitcnt vmcnt(1)
+; SI-NEXT: v_and_b32_e32 v15, 0xffff0000, v15
+; SI-NEXT: v_add_f32_e32 v15, 0x40c00000, v15
+; SI-NEXT: v_lshrrev_b32_e32 v50, 16, v15
+; SI-NEXT: buffer_load_dword v15, off, s[0:3], s32 offset:408 ; 4-byte Folded Reload
+; SI-NEXT: v_alignbit_b32 v10, v37, v10, 16
+; SI-NEXT: buffer_store_dword v10, off, s[0:3], s32 offset:384 ; 4-byte Folded Spill
+; SI-NEXT: s_waitcnt expcnt(0)
+; SI-NEXT: v_alignbit_b32 v10, v35, v11, 16
+; SI-NEXT: buffer_store_dword v10, off, s[0:3], s32 offset:380 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt vmcnt(3)
-; SI-NEXT: v_and_b32_e32 v14, 0xffff0000, v14
-; SI-NEXT: v_add_f32_e32 v14, 0x40c00000, v14
-; SI-NEXT: v_lshrrev_b32_e32 v50, 16, v14
-; SI-NEXT: buffer_load_dword v14, off, s[0:3], s32 offset:408 ; 4-byte Folded Reload
; SI-NEXT: v_and_b32_e32 v13, 0xffff0000, v13
+; SI-NEXT: v_add_f32_e32 v13, 0x40c00000, v13
+; SI-NEXT: v_alignbit_b32 v39, v48, v13, 16
+; SI-NEXT: buffer_load_dword v13, off, s[0:3], s32 offset:420 ; 4-byte Folded Reload
; SI-NEXT: s_waitcnt vmcnt(3)
; SI-NEXT: v_and_b32_e32 v15, 0xffff0000, v15
; SI-NEXT: v_add_f32_e32 v15, 0x40c00000, v15
; SI-NEXT: v_lshrrev_b32_e32 v52, 16, v15
-; SI-NEXT: buffer_load_dword v15, off, s[0:3], s32 offset:360 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v15, off, s[0:3], s32 offset:364 ; 4-byte Folded Reload
+; SI-NEXT: v_alignbit_b32 v12, v39, v12, 16
+; SI-NEXT: buffer_store_dword v12, off, s[0:3], s32 offset:388 ; 4-byte Folded Spill
+; SI-NEXT: s_waitcnt vmcnt(2)
+; SI-NEXT: v_and_b32_e32 v13, 0xffff0000, v13
; SI-NEXT: v_add_f32_e32 v13, 0x40c00000, v13
; SI-NEXT: v_alignbit_b32 v49, v50, v13, 16
-; SI-NEXT: v_and_b32_e32 v13, 0xffff0000, v51
+; SI-NEXT: v_and_b32_e32 v13, 0xffff0000, v14
+; SI-NEXT: buffer_load_dword v14, off, s[0:3], s32 offset:412 ; 4-byte Folded Reload
+; SI-NEXT: v_and_b32_e32 v9, 0xffff0000, v9
+; SI-NEXT: s_waitcnt vmcnt(2)
+; SI-NEXT: v_and_b32_e32 v15, 0xffff0000, v15
+; SI-NEXT: v_add_f32_e32 v9, 0x40c00000, v9
+; SI-NEXT: v_add_f32_e32 v15, 0x40c00000, v15
+; SI-NEXT: v_alignbit_b32 v28, v43, v9, 16
+; SI-NEXT: v_and_b32_e32 v9, 0xffff0000, v18
+; SI-NEXT: v_alignbit_b32 v41, v42, v15, 16
+; SI-NEXT: v_and_b32_e32 v15, 0xffff0000, v34
+; SI-NEXT: v_alignbit_b32 v15, v41, v15, 16
; SI-NEXT: v_alignbit_b32 v13, v49, v13, 16
-; SI-NEXT: buffer_store_dword v13, off, s[0:3], s32 offset:388 ; 4-byte Folded Spill
+; SI-NEXT: v_alignbit_b32 v9, v28, v9, 16
+; SI-NEXT: buffer_store_dword v15, off, s[0:3], s32 offset:404 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v13, off, s[0:3], s32 offset:392 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt vmcnt(2)
; SI-NEXT: v_and_b32_e32 v14, 0xffff0000, v14
; SI-NEXT: v_add_f32_e32 v14, 0x40c00000, v14
; SI-NEXT: v_alignbit_b32 v51, v52, v14, 16
; SI-NEXT: v_and_b32_e32 v14, 0xffff0000, v31
; SI-NEXT: v_alignbit_b32 v14, v51, v14, 16
-; SI-NEXT: s_waitcnt vmcnt(1)
-; SI-NEXT: v_and_b32_e32 v15, 0xffff0000, v15
-; SI-NEXT: v_add_f32_e32 v15, 0x40c00000, v15
-; SI-NEXT: v_alignbit_b32 v41, v42, v15, 16
-; SI-NEXT: v_and_b32_e32 v15, 0xffff0000, v32
-; SI-NEXT: v_alignbit_b32 v15, v41, v15, 16
-; SI-NEXT: buffer_store_dword v15, off, s[0:3], s32 offset:400 ; 4-byte Folded Spill
-; SI-NEXT: buffer_store_dword v14, off, s[0:3], s32 offset:392 ; 4-byte Folded Spill
-; SI-NEXT: buffer_store_dword v9, off, s[0:3], s32 offset:372 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v14, off, s[0:3], s32 offset:396 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v9, off, s[0:3], s32 offset:376 ; 4-byte Folded Spill
; SI-NEXT: buffer_store_dword v25, off, s[0:3], s32 offset:320 ; 4-byte Folded Spill
-; SI-NEXT: buffer_store_dword v8, off, s[0:3], s32 offset:368 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v8, off, s[0:3], s32 offset:372 ; 4-byte Folded Spill
; SI-NEXT: buffer_store_dword v24, off, s[0:3], s32 offset:312 ; 4-byte Folded Spill
-; SI-NEXT: buffer_store_dword v7, off, s[0:3], s32 offset:364 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v7, off, s[0:3], s32 offset:368 ; 4-byte Folded Spill
; SI-NEXT: buffer_store_dword v23, off, s[0:3], s32 offset:304 ; 4-byte Folded Spill
-; SI-NEXT: buffer_store_dword v5, off, s[0:3], s32 offset:356 ; 4-byte Folded Spill
-; SI-NEXT: buffer_store_dword v21, off, s[0:3], s32 offset:296 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v5, off, s[0:3], s32 offset:360 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v22, off, s[0:3], s32 offset:296 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v2, off, s[0:3], s32 offset:352 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v21, off, s[0:3], s32 offset:288 ; 4-byte Folded Spill
+; SI-NEXT: s_waitcnt expcnt(1)
+; SI-NEXT: v_alignbit_b32 v2, v21, v6, 16
; SI-NEXT: buffer_store_dword v2, off, s[0:3], s32 offset:348 ; 4-byte Folded Spill
-; SI-NEXT: buffer_store_dword v20, off, s[0:3], s32 offset:288 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v20, off, s[0:3], s32 offset:280 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(1)
-; SI-NEXT: v_alignbit_b32 v2, v20, v6, 16
+; SI-NEXT: v_alignbit_b32 v2, v20, v3, 16
; SI-NEXT: buffer_store_dword v2, off, s[0:3], s32 offset:344 ; 4-byte Folded Spill
-; SI-NEXT: buffer_store_dword v19, off, s[0:3], s32 offset:280 ; 4-byte Folded Spill
-; SI-NEXT: s_waitcnt expcnt(1)
-; SI-NEXT: v_alignbit_b32 v2, v19, v3, 16
-; SI-NEXT: buffer_store_dword v2, off, s[0:3], s32 offset:340 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v19, off, s[0:3], s32 offset:272 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:340 ; 4-byte Folded Spill
+; SI-NEXT: s_waitcnt expcnt(0)
+; SI-NEXT: v_alignbit_b32 v1, v45, v4, 16
; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:336 ; 4-byte Folded Spill
-; SI-NEXT: buffer_store_dword v18, off, s[0:3], s32 offset:272 ; 4-byte Folded Spill
-; SI-NEXT: s_waitcnt expcnt(1)
-; SI-NEXT: v_alignbit_b32 v1, v18, v4, 16
-; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:332 ; 4-byte Folded Spill
; SI-NEXT: .LBB104_4: ; %end
; SI-NEXT: s_or_b64 exec, exec, s[4:5]
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:244 ; 4-byte Folded Reload
-; SI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:396 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:240 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:400 ; 4-byte Folded Reload
; SI-NEXT: s_waitcnt vmcnt(1)
; SI-NEXT: v_and_b32_e32 v1, 0xffff, v1
; SI-NEXT: s_waitcnt vmcnt(0)
@@ -237135,7 +237147,7 @@ define <64 x i16> @bitcast_v64bf16_to_v64i16(<64 x bfloat> %a, i32 %b) {
; SI-NEXT: buffer_store_dword v1, v2, s[0:3], 0 offen
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:204 ; 4-byte Folded Reload
-; SI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:400 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:404 ; 4-byte Folded Reload
; SI-NEXT: s_waitcnt vmcnt(1)
; SI-NEXT: v_and_b32_e32 v1, 0xffff, v1
; SI-NEXT: s_waitcnt vmcnt(0)
@@ -237151,7 +237163,7 @@ define <64 x i16> @bitcast_v64bf16_to_v64i16(<64 x bfloat> %a, i32 %b) {
; SI-NEXT: buffer_store_dword v1, v2, s[0:3], 0 offen
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:208 ; 4-byte Folded Reload
-; SI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:392 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:396 ; 4-byte Folded Reload
; SI-NEXT: s_waitcnt vmcnt(1)
; SI-NEXT: v_and_b32_e32 v1, 0xffff, v1
; SI-NEXT: s_waitcnt vmcnt(0)
@@ -237167,7 +237179,7 @@ define <64 x i16> @bitcast_v64bf16_to_v64i16(<64 x bfloat> %a, i32 %b) {
; SI-NEXT: buffer_store_dword v1, v2, s[0:3], 0 offen
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:212 ; 4-byte Folded Reload
-; SI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:388 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:392 ; 4-byte Folded Reload
; SI-NEXT: s_waitcnt vmcnt(1)
; SI-NEXT: v_and_b32_e32 v1, 0xffff, v1
; SI-NEXT: s_waitcnt vmcnt(0)
@@ -237183,7 +237195,7 @@ define <64 x i16> @bitcast_v64bf16_to_v64i16(<64 x bfloat> %a, i32 %b) {
; SI-NEXT: buffer_store_dword v1, v2, s[0:3], 0 offen
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:216 ; 4-byte Folded Reload
-; SI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:384 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:388 ; 4-byte Folded Reload
; SI-NEXT: s_waitcnt vmcnt(1)
; SI-NEXT: v_and_b32_e32 v1, 0xffff, v1
; SI-NEXT: s_waitcnt vmcnt(0)
@@ -237199,7 +237211,7 @@ define <64 x i16> @bitcast_v64bf16_to_v64i16(<64 x bfloat> %a, i32 %b) {
; SI-NEXT: buffer_store_dword v1, v2, s[0:3], 0 offen
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:220 ; 4-byte Folded Reload
-; SI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:380 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:384 ; 4-byte Folded Reload
; SI-NEXT: s_waitcnt vmcnt(1)
; SI-NEXT: v_and_b32_e32 v1, 0xffff, v1
; SI-NEXT: s_waitcnt vmcnt(0)
@@ -237215,7 +237227,7 @@ define <64 x i16> @bitcast_v64bf16_to_v64i16(<64 x bfloat> %a, i32 %b) {
; SI-NEXT: buffer_store_dword v1, v2, s[0:3], 0 offen
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:224 ; 4-byte Folded Reload
-; SI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:376 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:380 ; 4-byte Folded Reload
; SI-NEXT: s_waitcnt vmcnt(1)
; SI-NEXT: v_and_b32_e32 v1, 0xffff, v1
; SI-NEXT: s_waitcnt vmcnt(0)
@@ -237231,7 +237243,7 @@ define <64 x i16> @bitcast_v64bf16_to_v64i16(<64 x bfloat> %a, i32 %b) {
; SI-NEXT: buffer_store_dword v1, v2, s[0:3], 0 offen
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:228 ; 4-byte Folded Reload
-; SI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:372 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:376 ; 4-byte Folded Reload
; SI-NEXT: s_waitcnt vmcnt(1)
; SI-NEXT: v_and_b32_e32 v1, 0xffff, v1
; SI-NEXT: s_waitcnt vmcnt(0)
@@ -237240,14 +237252,14 @@ define <64 x i16> @bitcast_v64bf16_to_v64i16(<64 x bfloat> %a, i32 %b) {
; SI-NEXT: v_add_i32_e32 v2, vcc, 56, v0
; SI-NEXT: buffer_store_dword v1, v2, s[0:3], 0 offen
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_and_b32_e32 v1, 0xffff, v22
-; SI-NEXT: v_lshlrev_b32_e32 v2, 16, v34
+; SI-NEXT: v_and_b32_e32 v1, 0xffff, v28
+; SI-NEXT: v_lshlrev_b32_e32 v2, 16, v43
; SI-NEXT: v_or_b32_e32 v1, v1, v2
; SI-NEXT: v_add_i32_e32 v2, vcc, 60, v0
; SI-NEXT: buffer_store_dword v1, v2, s[0:3], 0 offen
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:232 ; 4-byte Folded Reload
-; SI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:368 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:372 ; 4-byte Folded Reload
; SI-NEXT: s_waitcnt vmcnt(1)
; SI-NEXT: v_and_b32_e32 v1, 0xffff, v1
; SI-NEXT: s_waitcnt vmcnt(0)
@@ -237257,15 +237269,17 @@ define <64 x i16> @bitcast_v64bf16_to_v64i16(<64 x bfloat> %a, i32 %b) {
; SI-NEXT: buffer_store_dword v1, v2, s[0:3], 0 offen
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:320 ; 4-byte Folded Reload
-; SI-NEXT: v_lshlrev_b32_e32 v2, 16, v45
-; SI-NEXT: s_waitcnt vmcnt(0)
+; SI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:324 ; 4-byte Folded Reload
+; SI-NEXT: s_waitcnt vmcnt(1)
; SI-NEXT: v_and_b32_e32 v1, 0xffff, v1
+; SI-NEXT: s_waitcnt vmcnt(0)
+; SI-NEXT: v_lshlrev_b32_e32 v2, 16, v2
; SI-NEXT: v_or_b32_e32 v1, v1, v2
; SI-NEXT: v_add_i32_e32 v2, vcc, 0x44, v0
; SI-NEXT: buffer_store_dword v1, v2, s[0:3], 0 offen
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:236 ; 4-byte Folded Reload
-; SI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:364 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:368 ; 4-byte Folded Reload
; SI-NEXT: s_waitcnt vmcnt(1)
; SI-NEXT: v_and_b32_e32 v1, 0xffff, v1
; SI-NEXT: s_waitcnt vmcnt(0)
@@ -237284,8 +237298,8 @@ define <64 x i16> @bitcast_v64bf16_to_v64i16(<64 x bfloat> %a, i32 %b) {
; SI-NEXT: v_add_i32_e32 v2, vcc, 0x4c, v0
; SI-NEXT: buffer_store_dword v1, v2, s[0:3], 0 offen
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:240 ; 4-byte Folded Reload
-; SI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:356 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:252 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:360 ; 4-byte Folded Reload
; SI-NEXT: s_waitcnt vmcnt(1)
; SI-NEXT: v_and_b32_e32 v1, 0xffff, v1
; SI-NEXT: s_waitcnt vmcnt(0)
@@ -237304,8 +237318,8 @@ define <64 x i16> @bitcast_v64bf16_to_v64i16(<64 x bfloat> %a, i32 %b) {
; SI-NEXT: v_add_i32_e32 v2, vcc, 0x54, v0
; SI-NEXT: buffer_store_dword v1, v2, s[0:3], 0 offen
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:256 ; 4-byte Folded Reload
-; SI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:348 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:244 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:352 ; 4-byte Folded Reload
; SI-NEXT: s_waitcnt vmcnt(1)
; SI-NEXT: v_and_b32_e32 v1, 0xffff, v1
; SI-NEXT: s_waitcnt vmcnt(0)
@@ -237325,7 +237339,7 @@ define <64 x i16> @bitcast_v64bf16_to_v64i16(<64 x bfloat> %a, i32 %b) {
; SI-NEXT: buffer_store_dword v1, v2, s[0:3], 0 offen
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:248 ; 4-byte Folded Reload
-; SI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:344 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:348 ; 4-byte Folded Reload
; SI-NEXT: s_waitcnt vmcnt(1)
; SI-NEXT: v_and_b32_e32 v1, 0xffff, v1
; SI-NEXT: s_waitcnt vmcnt(0)
@@ -237344,8 +237358,8 @@ define <64 x i16> @bitcast_v64bf16_to_v64i16(<64 x bfloat> %a, i32 %b) {
; SI-NEXT: v_add_i32_e32 v2, vcc, 0x64, v0
; SI-NEXT: buffer_store_dword v1, v2, s[0:3], 0 offen
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:252 ; 4-byte Folded Reload
-; SI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:340 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:264 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:344 ; 4-byte Folded Reload
; SI-NEXT: s_waitcnt vmcnt(1)
; SI-NEXT: v_and_b32_e32 v1, 0xffff, v1
; SI-NEXT: s_waitcnt vmcnt(0)
@@ -237364,8 +237378,8 @@ define <64 x i16> @bitcast_v64bf16_to_v64i16(<64 x bfloat> %a, i32 %b) {
; SI-NEXT: v_add_i32_e32 v2, vcc, 0x6c, v0
; SI-NEXT: buffer_store_dword v1, v2, s[0:3], 0 offen
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:264 ; 4-byte Folded Reload
-; SI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:336 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:256 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:340 ; 4-byte Folded Reload
; SI-NEXT: s_waitcnt vmcnt(1)
; SI-NEXT: v_and_b32_e32 v1, 0xffff, v1
; SI-NEXT: s_waitcnt vmcnt(0)
@@ -237373,9 +237387,11 @@ define <64 x i16> @bitcast_v64bf16_to_v64i16(<64 x bfloat> %a, i32 %b) {
; SI-NEXT: v_or_b32_e32 v1, v1, v2
; SI-NEXT: v_add_i32_e32 v2, vcc, 0x70, v0
; SI-NEXT: buffer_store_dword v1, v2, s[0:3], 0 offen
-; SI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:276 ; 4-byte Folded Reload
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_and_b32_e32 v1, 0xffff, v61
+; SI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:272 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:276 ; 4-byte Folded Reload
+; SI-NEXT: s_waitcnt vmcnt(1)
+; SI-NEXT: v_and_b32_e32 v1, 0xffff, v1
; SI-NEXT: s_waitcnt vmcnt(0)
; SI-NEXT: v_lshlrev_b32_e32 v2, 16, v2
; SI-NEXT: v_or_b32_e32 v1, v1, v2
@@ -237383,7 +237399,7 @@ define <64 x i16> @bitcast_v64bf16_to_v64i16(<64 x bfloat> %a, i32 %b) {
; SI-NEXT: buffer_store_dword v1, v2, s[0:3], 0 offen
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:260 ; 4-byte Folded Reload
-; SI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:332 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:336 ; 4-byte Folded Reload
; SI-NEXT: s_waitcnt vmcnt(1)
; SI-NEXT: v_and_b32_e32 v1, 0xffff, v1
; SI-NEXT: s_waitcnt vmcnt(0)
@@ -237391,12 +237407,10 @@ define <64 x i16> @bitcast_v64bf16_to_v64i16(<64 x bfloat> %a, i32 %b) {
; SI-NEXT: v_or_b32_e32 v1, v1, v2
; SI-NEXT: v_add_i32_e32 v2, vcc, 0x78, v0
; SI-NEXT: buffer_store_dword v1, v2, s[0:3], 0 offen
-; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:272 ; 4-byte Folded Reload
; SI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:268 ; 4-byte Folded Reload
+; SI-NEXT: s_waitcnt expcnt(0)
+; SI-NEXT: v_and_b32_e32 v1, 0xffff, v45
; SI-NEXT: v_add_i32_e32 v0, vcc, 0x7c, v0
-; SI-NEXT: s_waitcnt vmcnt(1)
-; SI-NEXT: v_and_b32_e32 v1, 0xffff, v1
; SI-NEXT: s_waitcnt vmcnt(0)
; SI-NEXT: v_lshlrev_b32_e32 v2, 16, v2
; SI-NEXT: v_or_b32_e32 v1, v1, v2
@@ -239757,282 +239771,892 @@ define inreg <64 x i16> @bitcast_v64bf16_to_v64i16_scalar(<64 x bfloat> inreg %a
; SI-NEXT: buffer_load_dword v43, off, s[0:3], s32 offset:76
; SI-NEXT: v_mul_f32_e32 v44, 1.0, v1
; SI-NEXT: v_mul_f32_e32 v1, 1.0, v2
-; SI-NEXT: v_mul_f32_e32 v2, 1.0, v3
-; SI-NEXT: buffer_store_dword v2, off, s[0:3], s32 offset:148 ; 4-byte Folded Spill
+; SI-NEXT: v_mul_f32_e32 v2, 1.0, v4
+; SI-NEXT: buffer_store_dword v2, off, s[0:3], s32 offset:264 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_mul_f32_e32 v2, 1.0, v6
-; SI-NEXT: buffer_store_dword v2, off, s[0:3], s32 offset:152 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v2, off, s[0:3], s32 offset:148 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_mul_f32_e32 v2, 1.0, v7
-; SI-NEXT: buffer_store_dword v2, off, s[0:3], s32 offset:268 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v2, off, s[0:3], s32 offset:200 ; 4-byte Folded Spill
+; SI-NEXT: s_waitcnt expcnt(0)
+; SI-NEXT: v_mul_f32_e32 v2, 1.0, v10
+; SI-NEXT: buffer_store_dword v2, off, s[0:3], s32 offset:152 ; 4-byte Folded Spill
+; SI-NEXT: s_waitcnt expcnt(0)
+; SI-NEXT: v_mul_f32_e32 v2, 1.0, v11
+; SI-NEXT: buffer_store_dword v2, off, s[0:3], s32 offset:260 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_mul_f32_e32 v2, 1.0, v12
-; SI-NEXT: buffer_store_dword v2, off, s[0:3], s32 offset:264 ; 4-byte Folded Spill
-; SI-NEXT: v_mul_f32_e32 v61, 1.0, v4
+; SI-NEXT: buffer_store_dword v2, off, s[0:3], s32 offset:256 ; 4-byte Folded Spill
+; SI-NEXT: v_mul_f32_e32 v62, 1.0, v3
; SI-NEXT: v_mul_f32_e32 v45, 1.0, v5
; SI-NEXT: v_mul_f32_e32 v5, 1.0, v8
-; SI-NEXT: v_mul_f32_e32 v8, 1.0, v9
-; SI-NEXT: v_mul_f32_e32 v6, 1.0, v10
-; SI-NEXT: v_mul_f32_e32 v62, 1.0, v11
-; SI-NEXT: v_mul_f32_e32 v46, 1.0, v13
-; SI-NEXT: v_mul_f32_e32 v13, 1.0, v14
-; SI-NEXT: v_mul_f32_e32 v60, 1.0, v16
-; SI-NEXT: v_mul_f32_e32 v57, 1.0, v17
-; SI-NEXT: v_mul_f32_e32 v16, 1.0, v18
-; SI-NEXT: v_mul_f32_e32 v56, 1.0, v19
-; SI-NEXT: v_mul_f32_e32 v47, 1.0, v20
-; SI-NEXT: v_mul_f32_e32 v17, 1.0, v21
-; SI-NEXT: v_mul_f32_e32 v19, 1.0, v22
-; SI-NEXT: v_mul_f32_e32 v18, 1.0, v23
-; SI-NEXT: v_mul_f32_e32 v20, 1.0, v24
-; SI-NEXT: v_mul_f32_e32 v21, 1.0, v25
+; SI-NEXT: v_mul_f32_e32 v46, 1.0, v9
+; SI-NEXT: v_mul_f32_e32 v57, 1.0, v13
+; SI-NEXT: v_mul_f32_e32 v10, 1.0, v14
+; SI-NEXT: v_mul_f32_e32 v6, 1.0, v15
+; SI-NEXT: v_mul_f32_e32 v61, 1.0, v16
+; SI-NEXT: v_mul_f32_e32 v16, 1.0, v17
+; SI-NEXT: v_mul_f32_e32 v14, 1.0, v18
+; SI-NEXT: v_mul_f32_e32 v60, 1.0, v19
+; SI-NEXT: v_mul_f32_e32 v56, 1.0, v20
+; SI-NEXT: v_mul_f32_e32 v19, 1.0, v21
+; SI-NEXT: v_mul_f32_e32 v18, 1.0, v22
+; SI-NEXT: v_mul_f32_e32 v47, 1.0, v23
+; SI-NEXT: v_mul_f32_e32 v63, 1.0, v24
+; SI-NEXT: v_mul_f32_e32 v17, 1.0, v25
; SI-NEXT: v_mul_f32_e32 v22, 1.0, v26
-; SI-NEXT: v_mul_f32_e32 v63, 1.0, v27
-; SI-NEXT: v_mul_f32_e32 v58, 1.0, v28
+; SI-NEXT: v_mul_f32_e32 v12, 1.0, v27
+; SI-NEXT: v_mul_f32_e32 v13, 1.0, v28
; SI-NEXT: v_mul_f32_e32 v26, 1.0, v29
; SI-NEXT: v_mul_f32_e32 v23, 1.0, v30
-; SI-NEXT: v_mul_f32_e64 v7, 1.0, s16
+; SI-NEXT: v_mul_f32_e64 v8, 1.0, s16
; SI-NEXT: v_mul_f32_e64 v3, 1.0, s18
; SI-NEXT: v_mul_f32_e64 v9, 1.0, s22
-; SI-NEXT: v_mul_f32_e64 v11, 1.0, s23
-; SI-NEXT: v_mul_f32_e64 v29, 1.0, s25
-; SI-NEXT: v_mul_f32_e64 v14, 1.0, s27
-; SI-NEXT: v_mul_f32_e64 v25, 1.0, s28
-; SI-NEXT: v_mul_f32_e64 v24, 1.0, s29
+; SI-NEXT: v_mul_f32_e64 v7, 1.0, s23
+; SI-NEXT: v_mul_f32_e64 v29, 1.0, s24
+; SI-NEXT: v_mul_f32_e64 v25, 1.0, s25
+; SI-NEXT: v_mul_f32_e64 v20, 1.0, s26
+; SI-NEXT: v_mul_f32_e64 v15, 1.0, s27
+; SI-NEXT: v_mul_f32_e64 v24, 1.0, s28
+; SI-NEXT: v_mul_f32_e64 v11, 1.0, s29
; SI-NEXT: s_waitcnt vmcnt(14)
; SI-NEXT: v_cmp_ne_u32_e32 vcc, 0, v31
-; SI-NEXT: v_mul_f32_e32 v31, 1.0, v15
-; SI-NEXT: v_mul_f32_e32 v10, 1.0, v32
-; SI-NEXT: v_mul_f32_e32 v12, 1.0, v33
+; SI-NEXT: v_mul_f32_e32 v58, 1.0, v32
+; SI-NEXT: v_mul_f32_e32 v59, 1.0, v33
; SI-NEXT: s_and_b64 s[4:5], vcc, exec
+; SI-NEXT: v_mul_f32_e32 v21, 1.0, v34
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_mul_f32_e32 v2, 1.0, v36
-; SI-NEXT: buffer_store_dword v2, off, s[0:3], s32 offset:252 ; 4-byte Folded Spill
-; SI-NEXT: v_mul_f32_e32 v28, 1.0, v34
+; SI-NEXT: buffer_store_dword v2, off, s[0:3], s32 offset:244 ; 4-byte Folded Spill
; SI-NEXT: v_mul_f32_e32 v27, 1.0, v35
-; SI-NEXT: v_mul_f32_e32 v59, 1.0, v37
-; SI-NEXT: v_mul_f32_e32 v35, 1.0, v38
-; SI-NEXT: v_mul_f32_e32 v30, 1.0, v39
-; SI-NEXT: v_mul_f32_e32 v39, 1.0, v48
-; SI-NEXT: v_mul_f32_e32 v4, 1.0, v49
+; SI-NEXT: v_mul_f32_e32 v4, 1.0, v37
+; SI-NEXT: v_mul_f32_e32 v30, 1.0, v38
+; SI-NEXT: v_mul_f32_e32 v28, 1.0, v39
+; SI-NEXT: v_mul_f32_e32 v37, 1.0, v48
+; SI-NEXT: v_mul_f32_e32 v33, 1.0, v49
+; SI-NEXT: v_mul_f32_e32 v36, 1.0, v50
+; SI-NEXT: v_mul_f32_e32 v35, 1.0, v51
; SI-NEXT: s_waitcnt vmcnt(14)
-; SI-NEXT: v_mul_f32_e32 v48, 1.0, v50
-; SI-NEXT: s_waitcnt vmcnt(10) expcnt(0)
-; SI-NEXT: v_mul_f32_e32 v2, 1.0, v54
-; SI-NEXT: buffer_store_dword v2, off, s[0:3], s32 offset:256 ; 4-byte Folded Spill
-; SI-NEXT: v_mul_f32_e32 v36, 1.0, v51
-; SI-NEXT: v_mul_f32_e32 v37, 1.0, v52
-; SI-NEXT: v_mul_f32_e32 v38, 1.0, v53
-; SI-NEXT: s_waitcnt vmcnt(6) expcnt(0)
-; SI-NEXT: v_mul_f32_e32 v2, 1.0, v43
+; SI-NEXT: v_mul_f32_e32 v34, 1.0, v52
+; SI-NEXT: s_waitcnt vmcnt(13)
+; SI-NEXT: v_mul_f32_e32 v32, 1.0, v53
+; SI-NEXT: s_waitcnt vmcnt(12)
+; SI-NEXT: v_mul_f32_e32 v48, 1.0, v54
+; SI-NEXT: s_waitcnt vmcnt(11)
; SI-NEXT: v_mul_f32_e32 v49, 1.0, v55
-; SI-NEXT: v_mul_f32_e32 v33, 1.0, v40
-; SI-NEXT: v_mul_f32_e32 v34, 1.0, v41
-; SI-NEXT: v_mul_f32_e32 v32, 1.0, v42
-; SI-NEXT: buffer_store_dword v2, off, s[0:3], s32 offset:260 ; 4-byte Folded Spill
-; SI-NEXT: v_mul_f32_e64 v53, 1.0, s17
+; SI-NEXT: s_waitcnt vmcnt(8) expcnt(0)
+; SI-NEXT: v_mul_f32_e32 v2, 1.0, v42
+; SI-NEXT: buffer_store_dword v2, off, s[0:3], s32 offset:248 ; 4-byte Folded Spill
+; SI-NEXT: s_waitcnt vmcnt(8) expcnt(0)
+; SI-NEXT: v_mul_f32_e32 v2, 1.0, v43
+; SI-NEXT: v_mul_f32_e32 v54, 1.0, v40
+; SI-NEXT: v_mul_f32_e32 v53, 1.0, v41
+; SI-NEXT: buffer_store_dword v2, off, s[0:3], s32 offset:252 ; 4-byte Folded Spill
+; SI-NEXT: v_mul_f32_e64 v39, 1.0, s17
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_mul_f32_e64 v2, 1.0, s19
-; SI-NEXT: v_mul_f32_e64 v52, 1.0, s20
-; SI-NEXT: v_mul_f32_e64 v51, 1.0, s21
-; SI-NEXT: v_mul_f32_e64 v50, 1.0, s24
-; SI-NEXT: v_mul_f32_e64 v15, 1.0, s26
-; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:272 ; 4-byte Folded Spill
-; SI-NEXT: buffer_store_dword v44, off, s[0:3], s32 offset:276 ; 4-byte Folded Spill
-; SI-NEXT: buffer_store_dword v45, off, s[0:3], s32 offset:280 ; 4-byte Folded Spill
-; SI-NEXT: buffer_store_dword v6, off, s[0:3], s32 offset:284 ; 4-byte Folded Spill
-; SI-NEXT: buffer_store_dword v8, off, s[0:3], s32 offset:288 ; 4-byte Folded Spill
-; SI-NEXT: buffer_store_dword v13, off, s[0:3], s32 offset:292 ; 4-byte Folded Spill
-; SI-NEXT: buffer_store_dword v46, off, s[0:3], s32 offset:296 ; 4-byte Folded Spill
-; SI-NEXT: buffer_store_dword v16, off, s[0:3], s32 offset:300 ; 4-byte Folded Spill
-; SI-NEXT: buffer_store_dword v57, off, s[0:3], s32 offset:304 ; 4-byte Folded Spill
-; SI-NEXT: buffer_store_dword v17, off, s[0:3], s32 offset:308 ; 4-byte Folded Spill
-; SI-NEXT: buffer_store_dword v21, off, s[0:3], s32 offset:312 ; 4-byte Folded Spill
-; SI-NEXT: buffer_store_dword v23, off, s[0:3], s32 offset:316 ; 4-byte Folded Spill
-; SI-NEXT: buffer_store_dword v2, off, s[0:3], s32 offset:320 ; 4-byte Folded Spill
-; SI-NEXT: buffer_store_dword v3, off, s[0:3], s32 offset:324 ; 4-byte Folded Spill
-; SI-NEXT: buffer_store_dword v26, off, s[0:3], s32 offset:328 ; 4-byte Folded Spill
-; SI-NEXT: buffer_store_dword v27, off, s[0:3], s32 offset:332 ; 4-byte Folded Spill
-; SI-NEXT: buffer_store_dword v11, off, s[0:3], s32 offset:336 ; 4-byte Folded Spill
-; SI-NEXT: buffer_store_dword v9, off, s[0:3], s32 offset:340 ; 4-byte Folded Spill
-; SI-NEXT: buffer_store_dword v28, off, s[0:3], s32 offset:344 ; 4-byte Folded Spill
-; SI-NEXT: buffer_store_dword v30, off, s[0:3], s32 offset:348 ; 4-byte Folded Spill
-; SI-NEXT: buffer_store_dword v14, off, s[0:3], s32 offset:352 ; 4-byte Folded Spill
-; SI-NEXT: buffer_store_dword v15, off, s[0:3], s32 offset:356 ; 4-byte Folded Spill
-; SI-NEXT: buffer_store_dword v36, off, s[0:3], s32 offset:360 ; 4-byte Folded Spill
-; SI-NEXT: buffer_store_dword v48, off, s[0:3], s32 offset:364 ; 4-byte Folded Spill
-; SI-NEXT: buffer_store_dword v49, off, s[0:3], s32 offset:368 ; 4-byte Folded Spill
+; SI-NEXT: v_mul_f32_e64 v38, 1.0, s20
+; SI-NEXT: v_mul_f32_e64 v31, 1.0, s21
+; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:268 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v44, off, s[0:3], s32 offset:272 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v45, off, s[0:3], s32 offset:276 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v46, off, s[0:3], s32 offset:280 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v10, off, s[0:3], s32 offset:284 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v57, off, s[0:3], s32 offset:288 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v14, off, s[0:3], s32 offset:292 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v16, off, s[0:3], s32 offset:296 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v18, off, s[0:3], s32 offset:300 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v19, off, s[0:3], s32 offset:304 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v2, off, s[0:3], s32 offset:308 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v3, off, s[0:3], s32 offset:312 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v17, off, s[0:3], s32 offset:316 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v23, off, s[0:3], s32 offset:320 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v7, off, s[0:3], s32 offset:324 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v9, off, s[0:3], s32 offset:328 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v26, off, s[0:3], s32 offset:332 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v28, off, s[0:3], s32 offset:336 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v15, off, s[0:3], s32 offset:340 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v20, off, s[0:3], s32 offset:344 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v35, off, s[0:3], s32 offset:348 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v36, off, s[0:3], s32 offset:352 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v49, off, s[0:3], s32 offset:356 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v48, off, s[0:3], s32 offset:360 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v53, off, s[0:3], s32 offset:364 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v54, off, s[0:3], s32 offset:368 ; 4-byte Folded Spill
; SI-NEXT: buffer_store_dword v32, off, s[0:3], s32 offset:372 ; 4-byte Folded Spill
-; SI-NEXT: buffer_store_dword v19, off, s[0:3], s32 offset:376 ; 4-byte Folded Spill
-; SI-NEXT: buffer_store_dword v35, off, s[0:3], s32 offset:380 ; 4-byte Folded Spill
-; SI-NEXT: buffer_store_dword v33, off, s[0:3], s32 offset:384 ; 4-byte Folded Spill
-; SI-NEXT: buffer_store_dword v37, off, s[0:3], s32 offset:388 ; 4-byte Folded Spill
-; SI-NEXT: buffer_store_dword v39, off, s[0:3], s32 offset:392 ; 4-byte Folded Spill
-; SI-NEXT: buffer_store_dword v58, off, s[0:3], s32 offset:396 ; 4-byte Folded Spill
-; SI-NEXT: buffer_store_dword v22, off, s[0:3], s32 offset:400 ; 4-byte Folded Spill
-; SI-NEXT: buffer_store_dword v18, off, s[0:3], s32 offset:404 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v22, off, s[0:3], s32 offset:376 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v21, off, s[0:3], s32 offset:380 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v30, off, s[0:3], s32 offset:384 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v33, off, s[0:3], s32 offset:388 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v27, off, s[0:3], s32 offset:392 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v63, off, s[0:3], s32 offset:396 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v62, off, s[0:3], s32 offset:400 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v11, off, s[0:3], s32 offset:404 ; 4-byte Folded Spill
; SI-NEXT: buffer_store_dword v24, off, s[0:3], s32 offset:408 ; 4-byte Folded Spill
; SI-NEXT: buffer_store_dword v25, off, s[0:3], s32 offset:412 ; 4-byte Folded Spill
; SI-NEXT: buffer_store_dword v29, off, s[0:3], s32 offset:416 ; 4-byte Folded Spill
-; SI-NEXT: buffer_store_dword v50, off, s[0:3], s32 offset:420 ; 4-byte Folded Spill
-; SI-NEXT: buffer_store_dword v51, off, s[0:3], s32 offset:424 ; 4-byte Folded Spill
-; SI-NEXT: buffer_store_dword v52, off, s[0:3], s32 offset:428 ; 4-byte Folded Spill
-; SI-NEXT: buffer_store_dword v53, off, s[0:3], s32 offset:432 ; 4-byte Folded Spill
-; SI-NEXT: buffer_store_dword v7, off, s[0:3], s32 offset:436 ; 4-byte Folded Spill
-; SI-NEXT: s_cbranch_scc0 .LBB105_2
+; SI-NEXT: buffer_store_dword v31, off, s[0:3], s32 offset:420 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v38, off, s[0:3], s32 offset:424 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v39, off, s[0:3], s32 offset:428 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v8, off, s[0:3], s32 offset:432 ; 4-byte Folded Spill
+; SI-NEXT: s_cbranch_scc0 .LBB105_4
; SI-NEXT: ; %bb.1: ; %cmp.false
-; SI-NEXT: v_lshrrev_b32_e32 v2, 16, v2
-; SI-NEXT: buffer_store_dword v2, off, s[0:3], s32 offset:244 ; 4-byte Folded Spill
-; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_lshrrev_b32_e32 v2, 16, v52
-; SI-NEXT: buffer_store_dword v2, off, s[0:3], s32 offset:200 ; 4-byte Folded Spill
-; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_lshrrev_b32_e32 v2, 16, v51
-; SI-NEXT: buffer_store_dword v2, off, s[0:3], s32 offset:240 ; 4-byte Folded Spill
+; SI-NEXT: v_lshrrev_b32_e32 v51, 16, v2
+; SI-NEXT: v_lshrrev_b32_e32 v2, 16, v38
+; SI-NEXT: buffer_store_dword v2, off, s[0:3], s32 offset:192 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_lshrrev_b32_e32 v2, 16, v50
-; SI-NEXT: buffer_store_dword v2, off, s[0:3], s32 offset:196 ; 4-byte Folded Spill
+; SI-NEXT: v_lshrrev_b32_e32 v2, 16, v31
+; SI-NEXT: buffer_store_dword v2, off, s[0:3], s32 offset:236 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_lshrrev_b32_e32 v2, 16, v29
-; SI-NEXT: buffer_store_dword v2, off, s[0:3], s32 offset:232 ; 4-byte Folded Spill
-; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_lshrrev_b32_e32 v2, 16, v15
-; SI-NEXT: v_lshrrev_b32_e32 v15, 16, v44
-; SI-NEXT: v_lshrrev_b32_e32 v44, 16, v1
-; SI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:148 ; 4-byte Folded Reload
-; SI-NEXT: v_lshrrev_b32_e32 v7, 16, v7
-; SI-NEXT: buffer_store_dword v7, off, s[0:3], s32 offset:204 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v2, off, s[0:3], s32 offset:188 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_lshrrev_b32_e32 v7, 16, v53
-; SI-NEXT: buffer_store_dword v7, off, s[0:3], s32 offset:248 ; 4-byte Folded Spill
+; SI-NEXT: v_lshrrev_b32_e32 v2, 16, v25
+; SI-NEXT: buffer_store_dword v2, off, s[0:3], s32 offset:228 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_lshrrev_b32_e32 v7, 16, v14
-; SI-NEXT: buffer_store_dword v7, off, s[0:3], s32 offset:216 ; 4-byte Folded Spill
+; SI-NEXT: v_lshrrev_b32_e32 v2, 16, v20
+; SI-NEXT: v_lshrrev_b32_e32 v20, 16, v15
+; SI-NEXT: v_lshrrev_b32_e32 v15, 16, v1
+; SI-NEXT: v_lshrrev_b32_e32 v1, 16, v62
+; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:160 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
+; SI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:264 ; 4-byte Folded Reload
+; SI-NEXT: v_lshrrev_b32_e32 v50, 16, v3
+; SI-NEXT: v_lshrrev_b32_e32 v3, 16, v7
; SI-NEXT: v_lshrrev_b32_e32 v7, 16, v24
-; SI-NEXT: buffer_store_dword v7, off, s[0:3], s32 offset:236 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v7, off, s[0:3], s32 offset:156 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_mov_b32_e32 v7, v5
-; SI-NEXT: v_mov_b32_e32 v42, v62
-; SI-NEXT: v_mov_b32_e32 v43, v63
-; SI-NEXT: v_mov_b32_e32 v55, v12
-; SI-NEXT: v_lshrrev_b32_e32 v14, 16, v25
-; SI-NEXT: v_mov_b32_e32 v25, v60
-; SI-NEXT: v_mov_b32_e32 v54, v47
-; SI-NEXT: v_mov_b32_e32 v40, v20
-; SI-NEXT: v_mov_b32_e32 v51, v61
-; SI-NEXT: s_mov_b64 s[4:5], 0
-; SI-NEXT: v_lshrrev_b32_e32 v3, 16, v3
-; SI-NEXT: v_lshrrev_b32_e32 v9, 16, v9
-; SI-NEXT: v_lshrrev_b32_e32 v11, 16, v11
-; SI-NEXT: buffer_store_dword v14, off, s[0:3], s32 offset:156 ; 4-byte Folded Spill
-; SI-NEXT: v_lshrrev_b32_e32 v46, 16, v46
-; SI-NEXT: v_mov_b32_e32 v29, v31
+; SI-NEXT: v_lshrrev_b32_e32 v7, 16, v11
+; SI-NEXT: buffer_store_dword v7, off, s[0:3], s32 offset:232 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_lshrrev_b32_e32 v14, 16, v31
+; SI-NEXT: v_lshrrev_b32_e32 v7, 16, v45
+; SI-NEXT: v_lshrrev_b32_e32 v8, 16, v8
+; SI-NEXT: buffer_store_dword v8, off, s[0:3], s32 offset:196 ; 4-byte Folded Spill
+; SI-NEXT: s_waitcnt expcnt(0)
+; SI-NEXT: v_lshrrev_b32_e32 v8, 16, v39
+; SI-NEXT: buffer_store_dword v8, off, s[0:3], s32 offset:240 ; 4-byte Folded Spill
+; SI-NEXT: v_mov_b32_e32 v11, v5
+; SI-NEXT: s_waitcnt expcnt(0)
+; SI-NEXT: v_mov_b32_e32 v8, v47
+; SI-NEXT: v_mov_b32_e32 v39, v13
+; SI-NEXT: v_mov_b32_e32 v31, v6
+; SI-NEXT: v_mov_b32_e32 v29, v61
; SI-NEXT: v_mov_b32_e32 v24, v56
+; SI-NEXT: v_mov_b32_e32 v38, v4
+; SI-NEXT: v_lshrrev_b32_e32 v9, 16, v9
+; SI-NEXT: v_lshrrev_b32_e32 v44, 16, v44
+; SI-NEXT: v_lshrrev_b32_e32 v57, 16, v57
+; SI-NEXT: v_mov_b32_e32 v25, v60
; SI-NEXT: v_lshrrev_b32_e32 v18, 16, v18
-; SI-NEXT: v_mov_b32_e32 v52, v10
-; SI-NEXT: v_mov_b32_e32 v53, v59
-; SI-NEXT: v_lshrrev_b32_e32 v30, 16, v30
+; SI-NEXT: v_mov_b32_e32 v55, v12
+; SI-NEXT: v_mov_b32_e32 v40, v58
+; SI-NEXT: v_mov_b32_e32 v52, v59
+; SI-NEXT: v_mov_b32_e32 v43, v37
+; SI-NEXT: v_lshrrev_b32_e32 v37, 16, v37
+; SI-NEXT: v_lshrrev_b32_e32 v32, 16, v32
+; SI-NEXT: v_lshrrev_b32_e32 v48, 16, v48
; SI-NEXT: v_lshrrev_b32_e32 v49, 16, v49
-; SI-NEXT: s_waitcnt vmcnt(5)
+; SI-NEXT: s_waitcnt vmcnt(4)
+; SI-NEXT: v_mov_b32_e32 v62, v1
; SI-NEXT: v_lshrrev_b32_e32 v1, 16, v1
-; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:160 ; 4-byte Folded Spill
-; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_lshrrev_b32_e32 v1, 16, v61
-; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:228 ; 4-byte Folded Spill
-; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_lshrrev_b32_e32 v1, 16, v45
-; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:212 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:224 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:152 ; 4-byte Folded Reload
-; SI-NEXT: v_lshrrev_b32_e32 v61, 16, v13
-; SI-NEXT: v_lshrrev_b32_e32 v13, 16, v57
-; SI-NEXT: v_lshrrev_b32_e32 v57, 16, v26
-; SI-NEXT: v_lshrrev_b32_e32 v26, 16, v4
+; SI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:148 ; 4-byte Folded Reload
; SI-NEXT: s_waitcnt vmcnt(0)
; SI-NEXT: v_lshrrev_b32_e32 v45, 16, v1
-; SI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:268 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:200 ; 4-byte Folded Reload
; SI-NEXT: s_waitcnt vmcnt(0)
-; SI-NEXT: v_mov_b32_e32 v50, v1
; SI-NEXT: v_lshrrev_b32_e32 v1, 16, v1
; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:164 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_lshrrev_b32_e32 v1, 16, v5
-; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:224 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:220 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_lshrrev_b32_e32 v1, 16, v8
-; SI-NEXT: v_lshrrev_b32_e32 v5, 16, v6
-; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:208 ; 4-byte Folded Spill
+; SI-NEXT: v_lshrrev_b32_e32 v1, 16, v46
+; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:212 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_lshrrev_b32_e32 v1, 16, v62
-; SI-NEXT: v_mov_b32_e32 v62, v5
-; SI-NEXT: v_lshrrev_b32_e32 v5, 16, v63
-; SI-NEXT: v_lshrrev_b32_e32 v63, 16, v12
-; SI-NEXT: buffer_load_dword v12, off, s[0:3], s32 offset:252 ; 4-byte Folded Reload
-; SI-NEXT: buffer_load_dword v8, off, s[0:3], s32 offset:256 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:152 ; 4-byte Folded Reload
+; SI-NEXT: v_lshrrev_b32_e32 v5, 16, v12
+; SI-NEXT: v_lshrrev_b32_e32 v12, 16, v26
+; SI-NEXT: v_lshrrev_b32_e32 v26, 16, v36
+; SI-NEXT: s_waitcnt vmcnt(0)
+; SI-NEXT: v_lshrrev_b32_e32 v46, 16, v1
+; SI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:260 ; 4-byte Folded Reload
+; SI-NEXT: s_waitcnt vmcnt(0)
+; SI-NEXT: v_mov_b32_e32 v42, v1
+; SI-NEXT: v_lshrrev_b32_e32 v1, 16, v1
; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:168 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:264 ; 4-byte Folded Reload
-; SI-NEXT: buffer_store_dword v5, off, s[0:3], s32 offset:176 ; 4-byte Folded Spill
+; SI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:256 ; 4-byte Folded Reload
+; SI-NEXT: s_waitcnt vmcnt(0)
+; SI-NEXT: v_mov_b32_e32 v41, v1
+; SI-NEXT: v_lshrrev_b32_e32 v1, 16, v1
+; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:216 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_lshrrev_b32_e32 v5, 16, v10
-; SI-NEXT: buffer_store_dword v5, off, s[0:3], s32 offset:180 ; 4-byte Folded Spill
-; SI-NEXT: v_lshrrev_b32_e32 v10, 16, v27
-; SI-NEXT: v_lshrrev_b32_e32 v27, 16, v59
-; SI-NEXT: v_lshrrev_b32_e32 v59, 16, v36
-; SI-NEXT: v_lshrrev_b32_e32 v36, 16, v32
-; SI-NEXT: s_waitcnt vmcnt(5) expcnt(0)
-; SI-NEXT: v_lshrrev_b32_e32 v5, 16, v12
-; SI-NEXT: buffer_store_dword v5, off, s[0:3], s32 offset:184 ; 4-byte Folded Spill
+; SI-NEXT: v_lshrrev_b32_e32 v1, 16, v10
+; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:208 ; 4-byte Folded Spill
+; SI-NEXT: s_waitcnt expcnt(0)
+; SI-NEXT: v_lshrrev_b32_e32 v1, 16, v6
+; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:172 ; 4-byte Folded Spill
+; SI-NEXT: s_waitcnt expcnt(0)
+; SI-NEXT: v_lshrrev_b32_e32 v1, 16, v14
+; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:204 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_lshrrev_b32_e32 v5, 16, v39
-; SI-NEXT: s_waitcnt vmcnt(3)
-; SI-NEXT: v_mov_b32_e32 v41, v1
-; SI-NEXT: v_lshrrev_b32_e32 v6, 16, v1
; SI-NEXT: v_lshrrev_b32_e32 v1, 16, v60
-; SI-NEXT: v_lshrrev_b32_e32 v60, 16, v16
-; SI-NEXT: v_lshrrev_b32_e32 v16, 16, v47
-; SI-NEXT: v_lshrrev_b32_e32 v47, 16, v19
-; SI-NEXT: v_lshrrev_b32_e32 v19, 16, v20
-; SI-NEXT: buffer_store_dword v5, off, s[0:3], s32 offset:188 ; 4-byte Folded Spill
-; SI-NEXT: v_lshrrev_b32_e32 v20, 16, v33
+; SI-NEXT: v_lshrrev_b32_e32 v10, 16, v47
+; SI-NEXT: v_lshrrev_b32_e32 v47, 16, v63
+; SI-NEXT: v_lshrrev_b32_e32 v63, 16, v17
+; SI-NEXT: v_lshrrev_b32_e32 v17, 16, v13
+; SI-NEXT: buffer_load_dword v13, off, s[0:3], s32 offset:244 ; 4-byte Folded Reload
+; SI-NEXT: v_lshrrev_b32_e32 v6, 16, v61
+; SI-NEXT: v_lshrrev_b32_e32 v61, 16, v16
+; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:176 ; 4-byte Folded Spill
+; SI-NEXT: v_lshrrev_b32_e32 v16, 16, v56
+; SI-NEXT: s_waitcnt expcnt(0)
+; SI-NEXT: v_lshrrev_b32_e32 v1, 16, v22
+; SI-NEXT: v_lshrrev_b32_e32 v22, 16, v23
+; SI-NEXT: v_lshrrev_b32_e32 v56, 16, v21
+; SI-NEXT: v_lshrrev_b32_e32 v21, 16, v4
+; SI-NEXT: v_lshrrev_b32_e32 v23, 16, v30
+; SI-NEXT: v_lshrrev_b32_e32 v30, 16, v54
+; SI-NEXT: v_mov_b32_e32 v54, v5
+; SI-NEXT: v_lshrrev_b32_e32 v5, 16, v53
+; SI-NEXT: v_mov_b32_e32 v53, v10
+; SI-NEXT: buffer_load_dword v10, off, s[0:3], s32 offset:248 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v4, off, s[0:3], s32 offset:252 ; 4-byte Folded Reload
+; SI-NEXT: v_lshrrev_b32_e32 v14, 16, v58
+; SI-NEXT: buffer_store_dword v14, off, s[0:3], s32 offset:180 ; 4-byte Folded Spill
+; SI-NEXT: v_lshrrev_b32_e32 v60, 16, v19
+; SI-NEXT: v_lshrrev_b32_e32 v19, 16, v59
+; SI-NEXT: v_lshrrev_b32_e32 v59, 16, v27
+; SI-NEXT: v_lshrrev_b32_e32 v27, 16, v28
+; SI-NEXT: v_lshrrev_b32_e32 v58, 16, v35
+; SI-NEXT: v_lshrrev_b32_e32 v28, 16, v34
+; SI-NEXT: s_waitcnt vmcnt(4) expcnt(0)
+; SI-NEXT: v_lshrrev_b32_e32 v14, 16, v13
+; SI-NEXT: buffer_store_dword v14, off, s[0:3], s32 offset:184 ; 4-byte Folded Spill
+; SI-NEXT: s_waitcnt expcnt(0)
+; SI-NEXT: v_lshrrev_b32_e32 v14, 16, v33
; SI-NEXT: v_mov_b32_e32 v33, v34
+; SI-NEXT: v_mov_b32_e32 v34, v13
+; SI-NEXT: s_waitcnt vmcnt(3)
+; SI-NEXT: v_lshrrev_b32_e32 v36, 16, v10
+; SI-NEXT: s_waitcnt vmcnt(2)
+; SI-NEXT: v_lshrrev_b32_e32 v35, 16, v4
+; SI-NEXT: v_mov_b32_e32 v13, v4
+; SI-NEXT: s_cbranch_execnz .LBB105_3
+; SI-NEXT: .LBB105_2: ; %cmp.true
+; SI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:432 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:428 ; 4-byte Folded Reload
+; SI-NEXT: v_and_b32_e32 v23, 0xffff0000, v52
+; SI-NEXT: buffer_load_dword v7, off, s[0:3], s32 offset:396 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v12, off, s[0:3], s32 offset:336 ; 4-byte Folded Reload
+; SI-NEXT: s_waitcnt vmcnt(3)
+; SI-NEXT: v_and_b32_e32 v1, 0xffff0000, v1
+; SI-NEXT: s_waitcnt vmcnt(2)
+; SI-NEXT: v_and_b32_e32 v2, 0xffff0000, v2
+; SI-NEXT: v_add_f32_e32 v4, 0x40c00000, v2
+; SI-NEXT: v_add_f32_e32 v1, 0x40c00000, v1
+; SI-NEXT: v_lshrrev_b32_e32 v2, 16, v4
+; SI-NEXT: v_alignbit_b32 v1, v2, v1, 16
+; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:196 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_lshrrev_b32_e32 v5, 16, v34
-; SI-NEXT: buffer_load_dword v34, off, s[0:3], s32 offset:260 ; 4-byte Folded Reload
-; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:220 ; 4-byte Folded Spill
+; SI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:424 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:420 ; 4-byte Folded Reload
+; SI-NEXT: s_waitcnt vmcnt(4)
+; SI-NEXT: v_and_b32_e32 v19, 0xffff0000, v7
+; SI-NEXT: buffer_load_dword v7, off, s[0:3], s32 offset:388 ; 4-byte Folded Reload
+; SI-NEXT: v_add_f32_e32 v19, 0x40c00000, v19
+; SI-NEXT: v_lshrrev_b32_e32 v20, 16, v19
+; SI-NEXT: v_and_b32_e32 v4, 0xffff0000, v4
+; SI-NEXT: v_and_b32_e32 v19, 0xffff0000, v19
+; SI-NEXT: s_waitcnt vmcnt(2)
+; SI-NEXT: v_and_b32_e32 v1, 0xffff0000, v1
+; SI-NEXT: s_waitcnt vmcnt(1)
+; SI-NEXT: v_and_b32_e32 v2, 0xffff0000, v2
+; SI-NEXT: v_add_f32_e32 v5, 0x40c00000, v2
+; SI-NEXT: v_add_f32_e32 v1, 0x40c00000, v1
+; SI-NEXT: v_lshrrev_b32_e32 v2, 16, v5
+; SI-NEXT: v_alignbit_b32 v1, v2, v1, 16
+; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:192 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_lshrrev_b32_e32 v1, 16, v56
-; SI-NEXT: v_mov_b32_e32 v39, v4
-; SI-NEXT: v_lshrrev_b32_e32 v4, 16, v37
+; SI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:416 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:412 ; 4-byte Folded Reload
+; SI-NEXT: s_waitcnt vmcnt(3)
+; SI-NEXT: v_and_b32_e32 v27, 0xffff0000, v7
+; SI-NEXT: v_and_b32_e32 v5, 0xffff0000, v5
+; SI-NEXT: buffer_load_dword v7, off, s[0:3], s32 offset:392 ; 4-byte Folded Reload
+; SI-NEXT: s_waitcnt vmcnt(2)
+; SI-NEXT: v_and_b32_e32 v1, 0xffff0000, v1
+; SI-NEXT: s_waitcnt vmcnt(1)
+; SI-NEXT: v_and_b32_e32 v2, 0xffff0000, v2
+; SI-NEXT: v_add_f32_e32 v6, 0x40c00000, v2
+; SI-NEXT: v_add_f32_e32 v1, 0x40c00000, v1
+; SI-NEXT: v_lshrrev_b32_e32 v2, 16, v6
+; SI-NEXT: v_alignbit_b32 v1, v2, v1, 16
+; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:188 ; 4-byte Folded Spill
+; SI-NEXT: s_waitcnt expcnt(0)
+; SI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:408 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:404 ; 4-byte Folded Reload
+; SI-NEXT: v_and_b32_e32 v6, 0xffff0000, v6
+; SI-NEXT: s_waitcnt vmcnt(1)
+; SI-NEXT: v_and_b32_e32 v1, 0xffff0000, v1
+; SI-NEXT: s_waitcnt vmcnt(0)
+; SI-NEXT: v_and_b32_e32 v2, 0xffff0000, v2
+; SI-NEXT: v_add_f32_e32 v2, 0x40c00000, v2
+; SI-NEXT: v_add_f32_e32 v1, 0x40c00000, v1
+; SI-NEXT: v_lshrrev_b32_e32 v3, 16, v2
+; SI-NEXT: v_alignbit_b32 v1, v3, v1, 16
+; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:156 ; 4-byte Folded Spill
+; SI-NEXT: s_waitcnt expcnt(0)
+; SI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:400 ; 4-byte Folded Reload
+; SI-NEXT: v_and_b32_e32 v3, 0xffff0000, v62
+; SI-NEXT: v_add_f32_e32 v3, 0x40c00000, v3
+; SI-NEXT: v_lshrrev_b32_e32 v9, 16, v3
+; SI-NEXT: v_and_b32_e32 v37, 0xffff0000, v3
+; SI-NEXT: buffer_load_dword v3, off, s[0:3], s32 offset:272 ; 4-byte Folded Reload
+; SI-NEXT: s_waitcnt vmcnt(1)
+; SI-NEXT: v_and_b32_e32 v1, 0xffff0000, v1
+; SI-NEXT: v_add_f32_e32 v1, 0x40c00000, v1
+; SI-NEXT: v_alignbit_b32 v1, v9, v1, 16
+; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:160 ; 4-byte Folded Spill
+; SI-NEXT: s_waitcnt expcnt(0)
+; SI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:200 ; 4-byte Folded Reload
+; SI-NEXT: v_and_b32_e32 v9, 0xffff0000, v11
+; SI-NEXT: v_add_f32_e32 v9, 0x40c00000, v9
+; SI-NEXT: v_lshrrev_b32_e32 v11, 16, v9
+; SI-NEXT: s_waitcnt vmcnt(0)
+; SI-NEXT: v_and_b32_e32 v1, 0xffff0000, v1
+; SI-NEXT: v_add_f32_e32 v1, 0x40c00000, v1
+; SI-NEXT: v_alignbit_b32 v1, v11, v1, 16
+; SI-NEXT: v_and_b32_e32 v11, 0xffff0000, v41
+; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:164 ; 4-byte Folded Spill
+; SI-NEXT: s_waitcnt expcnt(0)
+; SI-NEXT: v_and_b32_e32 v1, 0xffff0000, v42
+; SI-NEXT: v_add_f32_e32 v11, 0x40c00000, v11
+; SI-NEXT: v_add_f32_e32 v1, 0x40c00000, v1
+; SI-NEXT: v_lshrrev_b32_e32 v14, 16, v11
+; SI-NEXT: v_alignbit_b32 v1, v14, v1, 16
+; SI-NEXT: v_and_b32_e32 v14, 0xffff0000, v29
+; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:168 ; 4-byte Folded Spill
+; SI-NEXT: s_waitcnt expcnt(0)
+; SI-NEXT: v_and_b32_e32 v1, 0xffff0000, v31
+; SI-NEXT: v_add_f32_e32 v14, 0x40c00000, v14
+; SI-NEXT: v_add_f32_e32 v1, 0x40c00000, v1
+; SI-NEXT: v_lshrrev_b32_e32 v16, 16, v14
+; SI-NEXT: v_alignbit_b32 v1, v16, v1, 16
+; SI-NEXT: v_and_b32_e32 v16, 0xffff0000, v24
; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:172 ; 4-byte Folded Spill
-; SI-NEXT: v_lshrrev_b32_e32 v56, 16, v17
-; SI-NEXT: v_lshrrev_b32_e32 v17, 16, v21
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_lshrrev_b32_e32 v1, 16, v22
-; SI-NEXT: v_lshrrev_b32_e32 v22, 16, v58
-; SI-NEXT: v_lshrrev_b32_e32 v58, 16, v23
-; SI-NEXT: v_lshrrev_b32_e32 v23, 16, v28
-; SI-NEXT: v_lshrrev_b32_e32 v21, 16, v35
-; SI-NEXT: v_lshrrev_b32_e32 v28, 16, v48
-; SI-NEXT: buffer_store_dword v4, off, s[0:3], s32 offset:192 ; 4-byte Folded Spill
-; SI-NEXT: v_mov_b32_e32 v37, v38
-; SI-NEXT: v_lshrrev_b32_e32 v38, 16, v38
-; SI-NEXT: v_lshrrev_b32_e32 v48, 16, v8
+; SI-NEXT: v_and_b32_e32 v1, 0xffff0000, v25
+; SI-NEXT: v_add_f32_e32 v16, 0x40c00000, v16
+; SI-NEXT: v_add_f32_e32 v1, 0x40c00000, v1
+; SI-NEXT: v_lshrrev_b32_e32 v18, 16, v16
+; SI-NEXT: v_alignbit_b32 v1, v18, v1, 16
+; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:176 ; 4-byte Folded Spill
+; SI-NEXT: s_waitcnt expcnt(0)
+; SI-NEXT: v_and_b32_e32 v1, 0xffff0000, v8
+; SI-NEXT: buffer_load_dword v8, off, s[0:3], s32 offset:372 ; 4-byte Folded Reload
+; SI-NEXT: v_add_f32_e32 v1, 0x40c00000, v1
+; SI-NEXT: v_alignbit_b32 v53, v20, v1, 16
+; SI-NEXT: v_and_b32_e32 v20, 0xffff0000, v39
+; SI-NEXT: v_and_b32_e32 v1, 0xffff0000, v55
+; SI-NEXT: v_add_f32_e32 v20, 0x40c00000, v20
+; SI-NEXT: v_add_f32_e32 v1, 0x40c00000, v1
+; SI-NEXT: v_lshrrev_b32_e32 v22, 16, v20
+; SI-NEXT: v_alignbit_b32 v54, v22, v1, 16
+; SI-NEXT: v_and_b32_e32 v1, 0xffff0000, v40
+; SI-NEXT: v_add_f32_e32 v24, 0x40c00000, v23
+; SI-NEXT: v_add_f32_e32 v1, 0x40c00000, v1
+; SI-NEXT: v_lshrrev_b32_e32 v23, 16, v24
+; SI-NEXT: v_alignbit_b32 v1, v23, v1, 16
+; SI-NEXT: v_and_b32_e32 v25, 0xffff0000, v38
+; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:180 ; 4-byte Folded Spill
+; SI-NEXT: s_waitcnt expcnt(0)
+; SI-NEXT: v_and_b32_e32 v1, 0xffff0000, v34
+; SI-NEXT: v_add_f32_e32 v25, 0x40c00000, v25
+; SI-NEXT: v_add_f32_e32 v1, 0x40c00000, v1
+; SI-NEXT: v_lshrrev_b32_e32 v26, 16, v25
+; SI-NEXT: v_alignbit_b32 v1, v26, v1, 16
+; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:184 ; 4-byte Folded Spill
+; SI-NEXT: s_waitcnt expcnt(0)
+; SI-NEXT: v_and_b32_e32 v1, 0xffff0000, v43
+; SI-NEXT: v_add_f32_e32 v29, 0x40c00000, v27
+; SI-NEXT: v_add_f32_e32 v1, 0x40c00000, v1
+; SI-NEXT: v_lshrrev_b32_e32 v27, 16, v29
+; SI-NEXT: v_alignbit_b32 v17, v27, v1, 16
+; SI-NEXT: v_and_b32_e32 v1, 0xffff0000, v33
+; SI-NEXT: v_add_f32_e32 v1, 0x40c00000, v1
+; SI-NEXT: v_and_b32_e32 v33, 0xffff0000, v13
+; SI-NEXT: v_add_f32_e32 v33, 0x40c00000, v33
+; SI-NEXT: v_lshrrev_b32_e32 v35, 16, v33
+; SI-NEXT: v_and_b32_e32 v21, 0xffff0000, v25
+; SI-NEXT: v_and_b32_e32 v25, 0xffff0000, v7
+; SI-NEXT: v_add_f32_e32 v25, 0x40c00000, v25
+; SI-NEXT: v_lshrrev_b32_e32 v59, 16, v25
+; SI-NEXT: buffer_load_dword v7, off, s[0:3], s32 offset:320 ; 4-byte Folded Reload
+; SI-NEXT: v_and_b32_e32 v13, 0xffff0000, v24
+; SI-NEXT: v_and_b32_e32 v34, 0xffff0000, v9
+; SI-NEXT: buffer_load_dword v9, off, s[0:3], s32 offset:276 ; 4-byte Folded Reload
+; SI-NEXT: v_and_b32_e32 v38, 0xffff0000, v2
+; SI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:344 ; 4-byte Folded Reload
+; SI-NEXT: v_and_b32_e32 v16, 0xffff0000, v16
+; SI-NEXT: v_and_b32_e32 v14, 0xffff0000, v14
+; SI-NEXT: v_and_b32_e32 v3, 0xffff0000, v3
+; SI-NEXT: v_add_f32_e32 v3, 0x40c00000, v3
+; SI-NEXT: s_waitcnt vmcnt(5)
+; SI-NEXT: v_and_b32_e32 v28, 0xffff0000, v8
+; SI-NEXT: v_add_f32_e32 v31, 0x40c00000, v28
+; SI-NEXT: v_lshrrev_b32_e32 v28, 16, v31
+; SI-NEXT: v_alignbit_b32 v28, v28, v1, 16
+; SI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:368 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v8, off, s[0:3], s32 offset:364 ; 4-byte Folded Reload
+; SI-NEXT: s_waitcnt vmcnt(4)
+; SI-NEXT: v_and_b32_e32 v24, 0xffff0000, v7
+; SI-NEXT: v_add_f32_e32 v24, 0x40c00000, v24
+; SI-NEXT: v_lshrrev_b32_e32 v22, 16, v24
+; SI-NEXT: buffer_load_dword v7, off, s[0:3], s32 offset:304 ; 4-byte Folded Reload
+; SI-NEXT: s_waitcnt vmcnt(4)
+; SI-NEXT: v_and_b32_e32 v9, 0xffff0000, v9
+; SI-NEXT: v_add_f32_e32 v9, 0x40c00000, v9
; SI-NEXT: s_waitcnt vmcnt(3)
-; SI-NEXT: v_lshrrev_b32_e32 v35, 16, v34
-; SI-NEXT: s_branch .LBB105_3
-; SI-NEXT: .LBB105_2:
-; SI-NEXT: s_waitcnt expcnt(4)
-; SI-NEXT: buffer_load_dword v50, off, s[0:3], s32 offset:268 ; 4-byte Folded Reload
-; SI-NEXT: buffer_load_dword v41, off, s[0:3], s32 offset:264 ; 4-byte Folded Reload
-; SI-NEXT: v_mov_b32_e32 v55, v12
+; SI-NEXT: v_and_b32_e32 v2, 0xffff0000, v2
+; SI-NEXT: v_add_f32_e32 v2, 0x40c00000, v2
+; SI-NEXT: s_waitcnt vmcnt(2)
+; SI-NEXT: v_and_b32_e32 v1, 0xffff0000, v1
+; SI-NEXT: s_waitcnt vmcnt(1)
+; SI-NEXT: v_and_b32_e32 v30, 0xffff0000, v8
+; SI-NEXT: v_add_f32_e32 v32, 0x40c00000, v30
+; SI-NEXT: v_add_f32_e32 v1, 0x40c00000, v1
+; SI-NEXT: v_lshrrev_b32_e32 v30, 16, v32
+; SI-NEXT: v_alignbit_b32 v30, v30, v1, 16
+; SI-NEXT: v_and_b32_e32 v1, 0xffff0000, v10
+; SI-NEXT: v_add_f32_e32 v1, 0x40c00000, v1
+; SI-NEXT: v_alignbit_b32 v36, v35, v1, 16
+; SI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:360 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v8, off, s[0:3], s32 offset:356 ; 4-byte Folded Reload
+; SI-NEXT: v_and_b32_e32 v52, 0xffff0000, v32
+; SI-NEXT: buffer_load_dword v10, off, s[0:3], s32 offset:348 ; 4-byte Folded Reload
+; SI-NEXT: s_waitcnt vmcnt(2)
+; SI-NEXT: v_and_b32_e32 v1, 0xffff0000, v1
+; SI-NEXT: s_waitcnt vmcnt(1)
+; SI-NEXT: v_and_b32_e32 v32, 0xffff0000, v8
+; SI-NEXT: v_add_f32_e32 v32, 0x40c00000, v32
+; SI-NEXT: v_add_f32_e32 v1, 0x40c00000, v1
+; SI-NEXT: v_lshrrev_b32_e32 v49, 16, v32
+; SI-NEXT: v_alignbit_b32 v48, v49, v1, 16
+; SI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:352 ; 4-byte Folded Reload
+; SI-NEXT: v_and_b32_e32 v8, 0xffff0000, v31
+; SI-NEXT: s_waitcnt vmcnt(1)
+; SI-NEXT: v_and_b32_e32 v31, 0xffff0000, v10
+; SI-NEXT: v_add_f32_e32 v31, 0x40c00000, v31
+; SI-NEXT: v_lshrrev_b32_e32 v58, 16, v31
+; SI-NEXT: v_and_b32_e32 v10, 0xffff0000, v29
+; SI-NEXT: v_and_b32_e32 v29, 0xffff0000, v12
+; SI-NEXT: v_add_f32_e32 v29, 0x40c00000, v29
+; SI-NEXT: v_lshrrev_b32_e32 v27, 16, v29
+; SI-NEXT: v_and_b32_e32 v32, 0xffff0000, v20
+; SI-NEXT: s_waitcnt vmcnt(0)
+; SI-NEXT: v_and_b32_e32 v1, 0xffff0000, v1
+; SI-NEXT: v_add_f32_e32 v1, 0x40c00000, v1
+; SI-NEXT: v_alignbit_b32 v26, v58, v1, 16
+; SI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:384 ; 4-byte Folded Reload
+; SI-NEXT: s_waitcnt vmcnt(0)
+; SI-NEXT: v_and_b32_e32 v1, 0xffff0000, v1
+; SI-NEXT: v_add_f32_e32 v1, 0x40c00000, v1
+; SI-NEXT: v_alignbit_b32 v23, v27, v1, 16
+; SI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:380 ; 4-byte Folded Reload
+; SI-NEXT: v_alignbit_b32 v21, v23, v21, 16
+; SI-NEXT: s_waitcnt vmcnt(0)
+; SI-NEXT: v_and_b32_e32 v1, 0xffff0000, v1
+; SI-NEXT: v_add_f32_e32 v1, 0x40c00000, v1
+; SI-NEXT: v_alignbit_b32 v56, v59, v1, 16
+; SI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:332 ; 4-byte Folded Reload
+; SI-NEXT: s_waitcnt vmcnt(0)
+; SI-NEXT: v_and_b32_e32 v1, 0xffff0000, v1
+; SI-NEXT: v_add_f32_e32 v1, 0x40c00000, v1
+; SI-NEXT: v_alignbit_b32 v12, v22, v1, 16
+; SI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:316 ; 4-byte Folded Reload
+; SI-NEXT: s_waitcnt vmcnt(0)
+; SI-NEXT: v_and_b32_e32 v1, 0xffff0000, v1
+; SI-NEXT: v_add_f32_e32 v20, 0x40c00000, v1
+; SI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:376 ; 4-byte Folded Reload
+; SI-NEXT: s_waitcnt vmcnt(0)
+; SI-NEXT: v_and_b32_e32 v1, 0xffff0000, v1
+; SI-NEXT: v_add_f32_e32 v1, 0x40c00000, v1
+; SI-NEXT: v_lshrrev_b32_e32 v1, 16, v1
+; SI-NEXT: v_alignbit_b32 v63, v1, v20, 16
+; SI-NEXT: v_and_b32_e32 v20, 0xffff0000, v7
+; SI-NEXT: buffer_load_dword v7, off, s[0:3], s32 offset:300 ; 4-byte Folded Reload
+; SI-NEXT: v_add_f32_e32 v20, 0x40c00000, v20
+; SI-NEXT: v_alignbit_b32 v47, v63, v19, 16
+; SI-NEXT: v_alignbit_b32 v19, v56, v13, 16
+; SI-NEXT: s_waitcnt vmcnt(0)
+; SI-NEXT: v_and_b32_e32 v24, 0xffff0000, v7
+; SI-NEXT: buffer_load_dword v7, off, s[0:3], s32 offset:296 ; 4-byte Folded Reload
+; SI-NEXT: v_add_f32_e32 v24, 0x40c00000, v24
+; SI-NEXT: v_lshrrev_b32_e32 v18, 16, v24
+; SI-NEXT: v_alignbit_b32 v60, v18, v20, 16
+; SI-NEXT: v_alignbit_b32 v16, v60, v16, 16
+; SI-NEXT: s_waitcnt vmcnt(0)
+; SI-NEXT: v_and_b32_e32 v25, 0xffff0000, v7
+; SI-NEXT: buffer_load_dword v7, off, s[0:3], s32 offset:292 ; 4-byte Folded Reload
+; SI-NEXT: v_add_f32_e32 v25, 0x40c00000, v25
+; SI-NEXT: s_waitcnt vmcnt(0)
+; SI-NEXT: v_and_b32_e32 v29, 0xffff0000, v7
+; SI-NEXT: v_add_f32_e32 v29, 0x40c00000, v29
+; SI-NEXT: v_lshrrev_b32_e32 v7, 16, v29
+; SI-NEXT: buffer_store_dword v7, off, s[0:3], s32 offset:204 ; 4-byte Folded Spill
+; SI-NEXT: v_alignbit_b32 v61, v7, v25, 16
+; SI-NEXT: s_waitcnt expcnt(0)
+; SI-NEXT: buffer_load_dword v7, off, s[0:3], s32 offset:288 ; 4-byte Folded Reload
+; SI-NEXT: s_waitcnt vmcnt(0)
+; SI-NEXT: v_and_b32_e32 v31, 0xffff0000, v7
+; SI-NEXT: buffer_load_dword v7, off, s[0:3], s32 offset:284 ; 4-byte Folded Reload
+; SI-NEXT: v_add_f32_e32 v31, 0x40c00000, v31
+; SI-NEXT: s_waitcnt vmcnt(0)
+; SI-NEXT: v_and_b32_e32 v33, 0xffff0000, v7
+; SI-NEXT: v_add_f32_e32 v33, 0x40c00000, v33
+; SI-NEXT: v_lshrrev_b32_e32 v7, 16, v33
+; SI-NEXT: buffer_store_dword v7, off, s[0:3], s32 offset:208 ; 4-byte Folded Spill
+; SI-NEXT: v_alignbit_b32 v57, v7, v31, 16
+; SI-NEXT: s_waitcnt expcnt(0)
+; SI-NEXT: buffer_load_dword v7, off, s[0:3], s32 offset:280 ; 4-byte Folded Reload
+; SI-NEXT: v_and_b32_e32 v33, 0xffff0000, v11
+; SI-NEXT: s_waitcnt vmcnt(0)
+; SI-NEXT: v_and_b32_e32 v11, 0xffff0000, v7
+; SI-NEXT: buffer_load_dword v7, off, s[0:3], s32 offset:152 ; 4-byte Folded Reload
+; SI-NEXT: v_add_f32_e32 v11, 0x40c00000, v11
+; SI-NEXT: s_waitcnt vmcnt(0)
+; SI-NEXT: v_and_b32_e32 v31, 0xffff0000, v7
+; SI-NEXT: v_add_f32_e32 v31, 0x40c00000, v31
+; SI-NEXT: v_lshrrev_b32_e32 v46, 16, v31
+; SI-NEXT: v_alignbit_b32 v24, v46, v11, 16
+; SI-NEXT: buffer_load_dword v11, off, s[0:3], s32 offset:148 ; 4-byte Folded Reload
+; SI-NEXT: buffer_store_dword v24, off, s[0:3], s32 offset:212 ; 4-byte Folded Spill
+; SI-NEXT: buffer_load_dword v7, off, s[0:3], s32 offset:328 ; 4-byte Folded Reload
+; SI-NEXT: s_waitcnt vmcnt(2)
+; SI-NEXT: v_and_b32_e32 v11, 0xffff0000, v11
+; SI-NEXT: v_add_f32_e32 v11, 0x40c00000, v11
+; SI-NEXT: v_lshrrev_b32_e32 v45, 16, v11
+; SI-NEXT: v_alignbit_b32 v25, v45, v9, 16
+; SI-NEXT: buffer_load_dword v9, off, s[0:3], s32 offset:268 ; 4-byte Folded Reload
+; SI-NEXT: s_waitcnt vmcnt(0)
+; SI-NEXT: v_and_b32_e32 v9, 0xffff0000, v9
+; SI-NEXT: v_add_f32_e32 v9, 0x40c00000, v9
+; SI-NEXT: v_lshrrev_b32_e32 v15, 16, v9
+; SI-NEXT: v_and_b32_e32 v9, 0xffff0000, v7
+; SI-NEXT: buffer_load_dword v7, off, s[0:3], s32 offset:324 ; 4-byte Folded Reload
+; SI-NEXT: v_alignbit_b32 v44, v15, v3, 16
+; SI-NEXT: buffer_load_dword v3, off, s[0:3], s32 offset:340 ; 4-byte Folded Reload
+; SI-NEXT: v_add_f32_e32 v9, 0x40c00000, v9
+; SI-NEXT: s_waitcnt vmcnt(1)
+; SI-NEXT: v_and_b32_e32 v11, 0xffff0000, v7
+; SI-NEXT: buffer_load_dword v7, off, s[0:3], s32 offset:312 ; 4-byte Folded Reload
+; SI-NEXT: v_add_f32_e32 v11, 0x40c00000, v11
+; SI-NEXT: s_waitcnt vmcnt(0)
+; SI-NEXT: v_and_b32_e32 v39, 0xffff0000, v7
+; SI-NEXT: buffer_load_dword v7, off, s[0:3], s32 offset:308 ; 4-byte Folded Reload
+; SI-NEXT: v_and_b32_e32 v3, 0xffff0000, v3
+; SI-NEXT: v_add_f32_e32 v3, 0x40c00000, v3
+; SI-NEXT: v_add_f32_e32 v39, 0x40c00000, v39
+; SI-NEXT: v_lshrrev_b32_e32 v20, 16, v3
+; SI-NEXT: v_lshrrev_b32_e32 v3, 16, v11
+; SI-NEXT: v_alignbit_b32 v9, v3, v9, 16
+; SI-NEXT: v_alignbit_b32 v2, v20, v2, 16
+; SI-NEXT: s_waitcnt vmcnt(0)
+; SI-NEXT: v_and_b32_e32 v50, 0xffff0000, v7
+; SI-NEXT: v_add_f32_e32 v50, 0x40c00000, v50
+; SI-NEXT: v_lshrrev_b32_e32 v51, 16, v50
+; SI-NEXT: v_alignbit_b32 v50, v51, v39, 16
+; SI-NEXT: v_alignbit_b32 v4, v50, v4, 16
+; SI-NEXT: buffer_store_dword v4, off, s[0:3], s32 offset:240 ; 4-byte Folded Spill
+; SI-NEXT: s_waitcnt expcnt(0)
+; SI-NEXT: v_alignbit_b32 v4, v9, v5, 16
+; SI-NEXT: buffer_store_dword v4, off, s[0:3], s32 offset:236 ; 4-byte Folded Spill
+; SI-NEXT: s_waitcnt expcnt(0)
+; SI-NEXT: v_alignbit_b32 v4, v2, v6, 16
+; SI-NEXT: buffer_store_dword v4, off, s[0:3], s32 offset:228 ; 4-byte Folded Spill
+; SI-NEXT: s_waitcnt expcnt(0)
+; SI-NEXT: v_alignbit_b32 v4, v44, v38, 16
+; SI-NEXT: buffer_store_dword v4, off, s[0:3], s32 offset:232 ; 4-byte Folded Spill
+; SI-NEXT: s_waitcnt expcnt(0)
+; SI-NEXT: v_alignbit_b32 v4, v25, v37, 16
+; SI-NEXT: buffer_store_dword v4, off, s[0:3], s32 offset:224 ; 4-byte Folded Spill
+; SI-NEXT: s_waitcnt expcnt(0)
+; SI-NEXT: v_alignbit_b32 v4, v24, v34, 16
+; SI-NEXT: v_mov_b32_e32 v7, v25
+; SI-NEXT: v_mov_b32_e32 v37, v17
+; SI-NEXT: buffer_store_dword v4, off, s[0:3], s32 offset:220 ; 4-byte Folded Spill
+; SI-NEXT: s_waitcnt expcnt(0)
+; SI-NEXT: v_alignbit_b32 v4, v57, v33, 16
+; SI-NEXT: v_alignbit_b32 v6, v61, v14, 16
+; SI-NEXT: v_alignbit_b32 v17, v12, v32, 16
+; SI-NEXT: v_alignbit_b32 v14, v26, v10, 16
+; SI-NEXT: v_alignbit_b32 v32, v48, v8, 16
+; SI-NEXT: v_alignbit_b32 v5, v36, v52, 16
+; SI-NEXT: buffer_store_dword v4, off, s[0:3], s32 offset:216 ; 4-byte Folded Spill
+; SI-NEXT: .LBB105_3: ; %end
+; SI-NEXT: s_waitcnt expcnt(0)
+; SI-NEXT: buffer_load_dword v4, off, s[0:3], s32 offset:196 ; 4-byte Folded Reload
+; SI-NEXT: v_and_b32_e32 v2, 0xffff, v2
+; SI-NEXT: v_lshlrev_b32_e32 v1, 16, v1
+; SI-NEXT: s_waitcnt vmcnt(0)
+; SI-NEXT: v_and_b32_e32 v8, 0xffff, v4
+; SI-NEXT: buffer_load_dword v4, off, s[0:3], s32 offset:240 ; 4-byte Folded Reload
+; SI-NEXT: s_waitcnt vmcnt(0)
+; SI-NEXT: v_lshlrev_b32_e32 v10, 16, v4
+; SI-NEXT: v_or_b32_e32 v8, v8, v10
+; SI-NEXT: buffer_store_dword v8, v0, s[0:3], 0 offen
+; SI-NEXT: s_waitcnt expcnt(0)
+; SI-NEXT: v_and_b32_e32 v8, 0xffff, v50
+; SI-NEXT: v_lshlrev_b32_e32 v10, 16, v51
+; SI-NEXT: v_or_b32_e32 v8, v8, v10
+; SI-NEXT: v_add_i32_e32 v10, vcc, 4, v0
+; SI-NEXT: buffer_store_dword v8, v10, s[0:3], 0 offen
+; SI-NEXT: buffer_load_dword v4, off, s[0:3], s32 offset:192 ; 4-byte Folded Reload
+; SI-NEXT: s_waitcnt vmcnt(0) expcnt(0)
+; SI-NEXT: v_and_b32_e32 v8, 0xffff, v4
+; SI-NEXT: buffer_load_dword v4, off, s[0:3], s32 offset:236 ; 4-byte Folded Reload
+; SI-NEXT: s_waitcnt vmcnt(0)
+; SI-NEXT: v_lshlrev_b32_e32 v4, 16, v4
+; SI-NEXT: v_or_b32_e32 v4, v8, v4
+; SI-NEXT: v_add_i32_e32 v8, vcc, 8, v0
+; SI-NEXT: buffer_store_dword v4, v8, s[0:3], 0 offen
+; SI-NEXT: s_waitcnt expcnt(0)
+; SI-NEXT: v_and_b32_e32 v4, 0xffff, v9
+; SI-NEXT: v_lshlrev_b32_e32 v8, 16, v3
+; SI-NEXT: v_or_b32_e32 v4, v4, v8
+; SI-NEXT: v_add_i32_e32 v8, vcc, 12, v0
+; SI-NEXT: buffer_store_dword v4, v8, s[0:3], 0 offen
+; SI-NEXT: s_waitcnt expcnt(0)
+; SI-NEXT: buffer_load_dword v4, off, s[0:3], s32 offset:188 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v3, off, s[0:3], s32 offset:228 ; 4-byte Folded Reload
+; SI-NEXT: s_waitcnt vmcnt(1)
+; SI-NEXT: v_and_b32_e32 v4, 0xffff, v4
+; SI-NEXT: s_waitcnt vmcnt(0)
+; SI-NEXT: v_lshlrev_b32_e32 v8, 16, v3
+; SI-NEXT: v_lshlrev_b32_e32 v3, 16, v20
+; SI-NEXT: v_or_b32_e32 v4, v4, v8
+; SI-NEXT: v_add_i32_e32 v8, vcc, 16, v0
+; SI-NEXT: v_or_b32_e32 v2, v2, v3
+; SI-NEXT: v_add_i32_e32 v3, vcc, 20, v0
+; SI-NEXT: buffer_store_dword v4, v8, s[0:3], 0 offen
+; SI-NEXT: buffer_store_dword v2, v3, s[0:3], 0 offen
+; SI-NEXT: s_waitcnt expcnt(0)
+; SI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:156 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v3, off, s[0:3], s32 offset:232 ; 4-byte Folded Reload
+; SI-NEXT: s_waitcnt vmcnt(1)
+; SI-NEXT: v_and_b32_e32 v2, 0xffff, v2
+; SI-NEXT: s_waitcnt vmcnt(0)
+; SI-NEXT: v_lshlrev_b32_e32 v3, 16, v3
+; SI-NEXT: v_or_b32_e32 v2, v2, v3
+; SI-NEXT: v_add_i32_e32 v3, vcc, 24, v0
+; SI-NEXT: buffer_store_dword v2, v3, s[0:3], 0 offen
+; SI-NEXT: s_waitcnt expcnt(0)
+; SI-NEXT: v_and_b32_e32 v2, 0xffff, v44
+; SI-NEXT: v_lshlrev_b32_e32 v3, 16, v15
+; SI-NEXT: v_or_b32_e32 v2, v2, v3
+; SI-NEXT: v_add_i32_e32 v3, vcc, 28, v0
+; SI-NEXT: buffer_store_dword v2, v3, s[0:3], 0 offen
+; SI-NEXT: s_waitcnt expcnt(0)
+; SI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:160 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v3, off, s[0:3], s32 offset:224 ; 4-byte Folded Reload
+; SI-NEXT: s_waitcnt vmcnt(1)
+; SI-NEXT: v_and_b32_e32 v2, 0xffff, v2
+; SI-NEXT: s_waitcnt vmcnt(0)
+; SI-NEXT: v_lshlrev_b32_e32 v3, 16, v3
+; SI-NEXT: v_or_b32_e32 v2, v2, v3
+; SI-NEXT: v_add_i32_e32 v3, vcc, 32, v0
+; SI-NEXT: buffer_store_dword v2, v3, s[0:3], 0 offen
+; SI-NEXT: s_waitcnt expcnt(0)
+; SI-NEXT: v_and_b32_e32 v2, 0xffff, v7
+; SI-NEXT: v_lshlrev_b32_e32 v3, 16, v45
+; SI-NEXT: v_or_b32_e32 v2, v2, v3
+; SI-NEXT: v_add_i32_e32 v3, vcc, 36, v0
+; SI-NEXT: buffer_store_dword v2, v3, s[0:3], 0 offen
+; SI-NEXT: s_waitcnt expcnt(0)
+; SI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:164 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v3, off, s[0:3], s32 offset:220 ; 4-byte Folded Reload
+; SI-NEXT: s_waitcnt vmcnt(1)
+; SI-NEXT: v_and_b32_e32 v2, 0xffff, v2
+; SI-NEXT: s_waitcnt vmcnt(0)
+; SI-NEXT: v_lshlrev_b32_e32 v3, 16, v3
+; SI-NEXT: v_or_b32_e32 v2, v2, v3
+; SI-NEXT: v_add_i32_e32 v3, vcc, 40, v0
+; SI-NEXT: buffer_store_dword v2, v3, s[0:3], 0 offen
+; SI-NEXT: s_waitcnt expcnt(0)
+; SI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:212 ; 4-byte Folded Reload
+; SI-NEXT: v_lshlrev_b32_e32 v3, 16, v46
+; SI-NEXT: s_waitcnt vmcnt(0)
+; SI-NEXT: v_and_b32_e32 v2, 0xffff, v2
+; SI-NEXT: v_or_b32_e32 v2, v2, v3
+; SI-NEXT: v_add_i32_e32 v3, vcc, 44, v0
+; SI-NEXT: buffer_store_dword v2, v3, s[0:3], 0 offen
+; SI-NEXT: s_waitcnt expcnt(0)
+; SI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:168 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v3, off, s[0:3], s32 offset:216 ; 4-byte Folded Reload
+; SI-NEXT: s_waitcnt vmcnt(1)
+; SI-NEXT: v_and_b32_e32 v2, 0xffff, v2
+; SI-NEXT: s_waitcnt vmcnt(0)
+; SI-NEXT: v_lshlrev_b32_e32 v3, 16, v3
+; SI-NEXT: v_or_b32_e32 v2, v2, v3
+; SI-NEXT: v_add_i32_e32 v3, vcc, 48, v0
+; SI-NEXT: buffer_store_dword v2, v3, s[0:3], 0 offen
+; SI-NEXT: buffer_load_dword v3, off, s[0:3], s32 offset:208 ; 4-byte Folded Reload
+; SI-NEXT: s_waitcnt expcnt(0)
+; SI-NEXT: v_and_b32_e32 v2, 0xffff, v57
+; SI-NEXT: s_waitcnt vmcnt(0)
+; SI-NEXT: v_lshlrev_b32_e32 v3, 16, v3
+; SI-NEXT: v_or_b32_e32 v2, v2, v3
+; SI-NEXT: v_add_i32_e32 v3, vcc, 52, v0
+; SI-NEXT: buffer_store_dword v2, v3, s[0:3], 0 offen
+; SI-NEXT: s_waitcnt expcnt(0)
+; SI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:172 ; 4-byte Folded Reload
+; SI-NEXT: v_lshlrev_b32_e32 v3, 16, v6
+; SI-NEXT: s_waitcnt vmcnt(0)
+; SI-NEXT: v_and_b32_e32 v2, 0xffff, v2
+; SI-NEXT: v_or_b32_e32 v2, v2, v3
+; SI-NEXT: v_add_i32_e32 v3, vcc, 56, v0
+; SI-NEXT: buffer_store_dword v2, v3, s[0:3], 0 offen
+; SI-NEXT: buffer_load_dword v3, off, s[0:3], s32 offset:204 ; 4-byte Folded Reload
+; SI-NEXT: s_waitcnt expcnt(0)
+; SI-NEXT: v_and_b32_e32 v2, 0xffff, v61
+; SI-NEXT: s_waitcnt vmcnt(0)
+; SI-NEXT: v_lshlrev_b32_e32 v3, 16, v3
+; SI-NEXT: v_or_b32_e32 v2, v2, v3
+; SI-NEXT: v_add_i32_e32 v3, vcc, 60, v0
+; SI-NEXT: buffer_store_dword v2, v3, s[0:3], 0 offen
+; SI-NEXT: s_waitcnt expcnt(0)
+; SI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:176 ; 4-byte Folded Reload
+; SI-NEXT: v_lshlrev_b32_e32 v3, 16, v16
+; SI-NEXT: s_waitcnt vmcnt(0)
+; SI-NEXT: v_and_b32_e32 v2, 0xffff, v2
+; SI-NEXT: v_or_b32_e32 v2, v2, v3
+; SI-NEXT: v_add_i32_e32 v3, vcc, 64, v0
+; SI-NEXT: buffer_store_dword v2, v3, s[0:3], 0 offen
+; SI-NEXT: s_waitcnt expcnt(0)
+; SI-NEXT: v_and_b32_e32 v2, 0xffff, v60
+; SI-NEXT: v_lshlrev_b32_e32 v3, 16, v18
+; SI-NEXT: v_or_b32_e32 v2, v2, v3
+; SI-NEXT: v_add_i32_e32 v3, vcc, 0x44, v0
+; SI-NEXT: buffer_store_dword v2, v3, s[0:3], 0 offen
+; SI-NEXT: s_waitcnt expcnt(0)
+; SI-NEXT: v_and_b32_e32 v2, 0xffff, v53
+; SI-NEXT: v_lshlrev_b32_e32 v3, 16, v47
+; SI-NEXT: v_or_b32_e32 v2, v2, v3
+; SI-NEXT: v_add_i32_e32 v3, vcc, 0x48, v0
+; SI-NEXT: buffer_store_dword v2, v3, s[0:3], 0 offen
+; SI-NEXT: s_waitcnt expcnt(0)
+; SI-NEXT: v_and_b32_e32 v2, 0xffff, v63
+; SI-NEXT: v_or_b32_e32 v1, v2, v1
+; SI-NEXT: v_add_i32_e32 v2, vcc, 0x4c, v0
+; SI-NEXT: buffer_store_dword v1, v2, s[0:3], 0 offen
+; SI-NEXT: s_waitcnt expcnt(0)
+; SI-NEXT: v_and_b32_e32 v1, 0xffff, v54
+; SI-NEXT: v_lshlrev_b32_e32 v2, 16, v17
+; SI-NEXT: v_or_b32_e32 v1, v1, v2
+; SI-NEXT: v_add_i32_e32 v2, vcc, 0x50, v0
+; SI-NEXT: buffer_store_dword v1, v2, s[0:3], 0 offen
+; SI-NEXT: s_waitcnt expcnt(0)
+; SI-NEXT: v_and_b32_e32 v1, 0xffff, v12
+; SI-NEXT: v_lshlrev_b32_e32 v2, 16, v22
+; SI-NEXT: v_or_b32_e32 v1, v1, v2
+; SI-NEXT: v_add_i32_e32 v2, vcc, 0x54, v0
+; SI-NEXT: buffer_store_dword v1, v2, s[0:3], 0 offen
+; SI-NEXT: s_waitcnt expcnt(0)
+; SI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:180 ; 4-byte Folded Reload
+; SI-NEXT: v_lshlrev_b32_e32 v2, 16, v19
+; SI-NEXT: s_waitcnt vmcnt(0)
+; SI-NEXT: v_and_b32_e32 v1, 0xffff, v1
+; SI-NEXT: v_or_b32_e32 v1, v1, v2
+; SI-NEXT: v_add_i32_e32 v2, vcc, 0x58, v0
+; SI-NEXT: buffer_store_dword v1, v2, s[0:3], 0 offen
+; SI-NEXT: s_waitcnt expcnt(0)
+; SI-NEXT: v_and_b32_e32 v1, 0xffff, v56
+; SI-NEXT: v_lshlrev_b32_e32 v2, 16, v59
+; SI-NEXT: v_or_b32_e32 v1, v1, v2
+; SI-NEXT: v_add_i32_e32 v2, vcc, 0x5c, v0
+; SI-NEXT: buffer_store_dword v1, v2, s[0:3], 0 offen
+; SI-NEXT: s_waitcnt expcnt(0)
+; SI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:184 ; 4-byte Folded Reload
+; SI-NEXT: v_lshlrev_b32_e32 v2, 16, v21
+; SI-NEXT: s_waitcnt vmcnt(0)
+; SI-NEXT: v_and_b32_e32 v1, 0xffff, v1
+; SI-NEXT: v_or_b32_e32 v1, v1, v2
+; SI-NEXT: v_add_i32_e32 v2, vcc, 0x60, v0
+; SI-NEXT: buffer_store_dword v1, v2, s[0:3], 0 offen
+; SI-NEXT: s_waitcnt expcnt(0)
+; SI-NEXT: v_and_b32_e32 v1, 0xffff, v23
+; SI-NEXT: v_lshlrev_b32_e32 v2, 16, v27
+; SI-NEXT: v_or_b32_e32 v1, v1, v2
+; SI-NEXT: v_add_i32_e32 v2, vcc, 0x64, v0
+; SI-NEXT: buffer_store_dword v1, v2, s[0:3], 0 offen
+; SI-NEXT: s_waitcnt expcnt(0)
+; SI-NEXT: v_and_b32_e32 v1, 0xffff, v37
+; SI-NEXT: v_lshlrev_b32_e32 v2, 16, v14
+; SI-NEXT: v_or_b32_e32 v1, v1, v2
+; SI-NEXT: v_add_i32_e32 v2, vcc, 0x68, v0
+; SI-NEXT: buffer_store_dword v1, v2, s[0:3], 0 offen
+; SI-NEXT: s_waitcnt expcnt(0)
+; SI-NEXT: v_and_b32_e32 v1, 0xffff, v26
+; SI-NEXT: v_lshlrev_b32_e32 v2, 16, v58
+; SI-NEXT: v_or_b32_e32 v1, v1, v2
+; SI-NEXT: v_add_i32_e32 v2, vcc, 0x6c, v0
+; SI-NEXT: buffer_store_dword v1, v2, s[0:3], 0 offen
+; SI-NEXT: s_waitcnt expcnt(0)
+; SI-NEXT: v_and_b32_e32 v1, 0xffff, v28
+; SI-NEXT: v_lshlrev_b32_e32 v2, 16, v32
+; SI-NEXT: v_or_b32_e32 v1, v1, v2
+; SI-NEXT: v_add_i32_e32 v2, vcc, 0x70, v0
+; SI-NEXT: buffer_store_dword v1, v2, s[0:3], 0 offen
+; SI-NEXT: s_waitcnt expcnt(0)
+; SI-NEXT: v_and_b32_e32 v1, 0xffff, v48
+; SI-NEXT: v_lshlrev_b32_e32 v2, 16, v49
+; SI-NEXT: v_or_b32_e32 v1, v1, v2
+; SI-NEXT: v_add_i32_e32 v2, vcc, 0x74, v0
+; SI-NEXT: buffer_store_dword v1, v2, s[0:3], 0 offen
+; SI-NEXT: s_waitcnt expcnt(0)
+; SI-NEXT: v_and_b32_e32 v1, 0xffff, v30
+; SI-NEXT: v_lshlrev_b32_e32 v2, 16, v5
+; SI-NEXT: v_or_b32_e32 v1, v1, v2
+; SI-NEXT: v_add_i32_e32 v2, vcc, 0x78, v0
+; SI-NEXT: buffer_store_dword v1, v2, s[0:3], 0 offen
+; SI-NEXT: s_waitcnt expcnt(0)
+; SI-NEXT: v_and_b32_e32 v1, 0xffff, v36
+; SI-NEXT: v_lshlrev_b32_e32 v2, 16, v35
+; SI-NEXT: v_or_b32_e32 v1, v1, v2
+; SI-NEXT: v_add_i32_e32 v0, vcc, 0x7c, v0
+; SI-NEXT: buffer_store_dword v1, v0, s[0:3], 0 offen
+; SI-NEXT: buffer_load_dword v63, off, s[0:3], s32 offset:84 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v62, off, s[0:3], s32 offset:88 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v61, off, s[0:3], s32 offset:92 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v60, off, s[0:3], s32 offset:96 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v59, off, s[0:3], s32 offset:100 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v58, off, s[0:3], s32 offset:104 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v57, off, s[0:3], s32 offset:108 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v56, off, s[0:3], s32 offset:112 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v47, off, s[0:3], s32 offset:116 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v46, off, s[0:3], s32 offset:120 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v45, off, s[0:3], s32 offset:124 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v44, off, s[0:3], s32 offset:128 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v43, off, s[0:3], s32 offset:132 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v42, off, s[0:3], s32 offset:136 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v41, off, s[0:3], s32 offset:140 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v40, off, s[0:3], s32 offset:144 ; 4-byte Folded Reload
+; SI-NEXT: s_waitcnt vmcnt(0) expcnt(0)
+; SI-NEXT: s_setpc_b64 s[30:31]
+; SI-NEXT: .LBB105_4:
+; SI-NEXT: buffer_load_dword v62, off, s[0:3], s32 offset:264 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v42, off, s[0:3], s32 offset:260 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v41, off, s[0:3], s32 offset:256 ; 4-byte Folded Reload
+; SI-NEXT: s_waitcnt expcnt(1)
+; SI-NEXT: v_mov_b32_e32 v39, v13
; SI-NEXT: v_mov_b32_e32 v33, v34
-; SI-NEXT: buffer_load_dword v8, off, s[0:3], s32 offset:256 ; 4-byte Folded Reload
-; SI-NEXT: buffer_load_dword v12, off, s[0:3], s32 offset:252 ; 4-byte Folded Reload
-; SI-NEXT: buffer_load_dword v34, off, s[0:3], s32 offset:260 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v10, off, s[0:3], s32 offset:248 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v34, off, s[0:3], s32 offset:244 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v13, off, s[0:3], s32 offset:252 ; 4-byte Folded Reload
; SI-NEXT: ; implicit-def: $vgpr1
; SI-NEXT: ; kill: killed $vgpr1
; SI-NEXT: ; implicit-def: $vgpr1
@@ -240064,695 +240688,73 @@ define inreg <64 x i16> @bitcast_v64bf16_to_v64i16_scalar(<64 x bfloat> inreg %a
; SI-NEXT: ; implicit-def: $vgpr1
; SI-NEXT: ; kill: killed $vgpr1
; SI-NEXT: ; implicit-def: $vgpr1
-; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_mov_b32_e32 v7, v5
; SI-NEXT: ; kill: killed $vgpr1
; SI-NEXT: ; implicit-def: $vgpr1
-; SI-NEXT: ; implicit-def: $vgpr5
; SI-NEXT: ; kill: killed $vgpr1
; SI-NEXT: ; implicit-def: $vgpr1
-; SI-NEXT: ; kill: killed $vgpr5
-; SI-NEXT: ; implicit-def: $vgpr5
-; SI-NEXT: v_mov_b32_e32 v51, v61
-; SI-NEXT: v_mov_b32_e32 v42, v62
-; SI-NEXT: v_mov_b32_e32 v29, v31
+; SI-NEXT: v_mov_b32_e32 v11, v5
+; SI-NEXT: v_mov_b32_e32 v31, v6
+; SI-NEXT: v_mov_b32_e32 v29, v61
; SI-NEXT: v_mov_b32_e32 v25, v60
; SI-NEXT: v_mov_b32_e32 v24, v56
-; SI-NEXT: v_mov_b32_e32 v54, v47
-; SI-NEXT: v_mov_b32_e32 v40, v20
-; SI-NEXT: v_mov_b32_e32 v43, v63
-; SI-NEXT: v_mov_b32_e32 v52, v10
-; SI-NEXT: v_mov_b32_e32 v53, v59
-; SI-NEXT: v_mov_b32_e32 v39, v4
-; SI-NEXT: v_mov_b32_e32 v37, v38
-; SI-NEXT: s_mov_b64 s[4:5], -1
+; SI-NEXT: s_waitcnt expcnt(0)
+; SI-NEXT: v_mov_b32_e32 v8, v47
+; SI-NEXT: v_mov_b32_e32 v55, v12
+; SI-NEXT: v_mov_b32_e32 v40, v58
+; SI-NEXT: v_mov_b32_e32 v52, v59
+; SI-NEXT: v_mov_b32_e32 v38, v4
+; SI-NEXT: v_mov_b32_e32 v43, v37
; SI-NEXT: ; kill: killed $vgpr1
; SI-NEXT: ; implicit-def: $vgpr1
-; SI-NEXT: ; kill: killed $vgpr5
; SI-NEXT: ; implicit-def: $vgpr5
-; SI-NEXT: ; implicit-def: $vgpr4
-; SI-NEXT: ; implicit-def: $vgpr3
+; SI-NEXT: ; implicit-def: $vgpr50
+; SI-NEXT: ; implicit-def: $vgpr51
; SI-NEXT: ; implicit-def: $vgpr9
-; SI-NEXT: ; implicit-def: $vgpr11
+; SI-NEXT: ; implicit-def: $vgpr3
; SI-NEXT: ; implicit-def: $vgpr2
-; SI-NEXT: ; implicit-def: $vgpr15
+; SI-NEXT: ; implicit-def: $vgpr20
; SI-NEXT: ; implicit-def: $vgpr44
+; SI-NEXT: ; implicit-def: $vgpr15
+; SI-NEXT: ; implicit-def: $vgpr7
; SI-NEXT: ; implicit-def: $vgpr45
-; SI-NEXT: ; implicit-def: $vgpr62
-; SI-NEXT: ; implicit-def: $vgpr6
; SI-NEXT: ; implicit-def: $vgpr46
+; SI-NEXT: ; implicit-def: $vgpr57
+; SI-NEXT: ; implicit-def: $vgpr6
; SI-NEXT: ; implicit-def: $vgpr61
-; SI-NEXT: ; implicit-def: $vgpr14
-; SI-NEXT: ; implicit-def: $vgpr13
-; SI-NEXT: ; implicit-def: $vgpr60
; SI-NEXT: ; kill: killed $vgpr1
; SI-NEXT: ; implicit-def: $vgpr16
-; SI-NEXT: ; implicit-def: $vgpr56
-; SI-NEXT: ; implicit-def: $vgpr47
+; SI-NEXT: ; implicit-def: $vgpr60
; SI-NEXT: ; implicit-def: $vgpr18
-; SI-NEXT: ; implicit-def: $vgpr19
-; SI-NEXT: ; implicit-def: $vgpr17
+; SI-NEXT: ; implicit-def: $vgpr53
+; SI-NEXT: ; implicit-def: $vgpr47
+; SI-NEXT: ; implicit-def: $vgpr63
; SI-NEXT: ; implicit-def: $vgpr1
+; SI-NEXT: ; implicit-def: $vgpr54
+; SI-NEXT: ; implicit-def: $vgpr17
+; SI-NEXT: ; implicit-def: $vgpr12
; SI-NEXT: ; implicit-def: $vgpr22
-; SI-NEXT: ; implicit-def: $vgpr57
-; SI-NEXT: ; implicit-def: $vgpr58
-; SI-NEXT: ; implicit-def: $vgpr63
-; SI-NEXT: ; implicit-def: $vgpr23
-; SI-NEXT: ; implicit-def: $vgpr10
; SI-NEXT: ; kill: killed $vgpr5
-; SI-NEXT: ; implicit-def: $vgpr27
-; SI-NEXT: ; implicit-def: $vgpr21
-; SI-NEXT: ; implicit-def: $vgpr30
-; SI-NEXT: ; kill: killed $vgpr4
-; SI-NEXT: ; implicit-def: $vgpr26
-; SI-NEXT: ; implicit-def: $vgpr28
+; SI-NEXT: ; implicit-def: $vgpr19
+; SI-NEXT: ; implicit-def: $vgpr56
; SI-NEXT: ; implicit-def: $vgpr59
; SI-NEXT: ; implicit-def: $vgpr4
; SI-NEXT: ; kill: killed $vgpr4
-; SI-NEXT: ; implicit-def: $vgpr38
+; SI-NEXT: ; implicit-def: $vgpr21
+; SI-NEXT: ; implicit-def: $vgpr23
+; SI-NEXT: ; implicit-def: $vgpr27
+; SI-NEXT: ; implicit-def: $vgpr37
+; SI-NEXT: ; implicit-def: $vgpr14
+; SI-NEXT: ; implicit-def: $vgpr26
+; SI-NEXT: ; implicit-def: $vgpr58
+; SI-NEXT: ; implicit-def: $vgpr28
+; SI-NEXT: ; implicit-def: $vgpr32
; SI-NEXT: ; implicit-def: $vgpr48
; SI-NEXT: ; implicit-def: $vgpr49
-; SI-NEXT: ; implicit-def: $vgpr20
+; SI-NEXT: ; implicit-def: $vgpr30
; SI-NEXT: ; implicit-def: $vgpr5
; SI-NEXT: ; implicit-def: $vgpr36
; SI-NEXT: ; implicit-def: $vgpr35
-; SI-NEXT: .LBB105_3: ; %Flow
-; SI-NEXT: buffer_load_dword v31, off, s[0:3], s32 offset:148 ; 4-byte Folded Reload
-; SI-NEXT: s_andn2_b64 vcc, exec, s[4:5]
-; SI-NEXT: s_cbranch_vccnz .LBB105_5
-; SI-NEXT: ; %bb.4: ; %cmp.true
-; SI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:436 ; 4-byte Folded Reload
-; SI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:432 ; 4-byte Folded Reload
-; SI-NEXT: v_and_b32_e32 v19, 0xffff0000, v40
-; SI-NEXT: v_add_f32_e32 v19, 0x40c00000, v19
-; SI-NEXT: v_lshrrev_b32_e32 v20, 16, v19
-; SI-NEXT: v_and_b32_e32 v23, 0xffff0000, v55
-; SI-NEXT: v_and_b32_e32 v27, 0xffff0000, v39
-; SI-NEXT: v_and_b32_e32 v28, 0xffff0000, v37
-; SI-NEXT: v_and_b32_e32 v30, 0xffff0000, v33
-; SI-NEXT: v_add_f32_e32 v32, 0x40c00000, v30
-; SI-NEXT: v_lshrrev_b32_e32 v30, 16, v32
-; SI-NEXT: s_waitcnt vmcnt(3)
-; SI-NEXT: v_and_b32_e32 v33, 0xffff0000, v34
-; SI-NEXT: v_add_f32_e32 v33, 0x40c00000, v33
-; SI-NEXT: v_lshrrev_b32_e32 v35, 16, v33
-; SI-NEXT: v_and_b32_e32 v19, 0xffff0000, v19
-; SI-NEXT: buffer_load_dword v10, off, s[0:3], s32 offset:360 ; 4-byte Folded Reload
-; SI-NEXT: buffer_load_dword v13, off, s[0:3], s32 offset:332 ; 4-byte Folded Reload
-; SI-NEXT: buffer_load_dword v15, off, s[0:3], s32 offset:316 ; 4-byte Folded Reload
-; SI-NEXT: s_waitcnt vmcnt(4)
-; SI-NEXT: v_and_b32_e32 v1, 0xffff0000, v1
-; SI-NEXT: s_waitcnt vmcnt(3)
-; SI-NEXT: v_and_b32_e32 v2, 0xffff0000, v2
-; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_add_f32_e32 v4, 0x40c00000, v2
-; SI-NEXT: v_add_f32_e32 v1, 0x40c00000, v1
-; SI-NEXT: v_lshrrev_b32_e32 v2, 16, v4
-; SI-NEXT: v_alignbit_b32 v1, v2, v1, 16
-; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:204 ; 4-byte Folded Spill
-; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:428 ; 4-byte Folded Reload
-; SI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:424 ; 4-byte Folded Reload
-; SI-NEXT: v_and_b32_e32 v4, 0xffff0000, v4
-; SI-NEXT: s_waitcnt vmcnt(1)
-; SI-NEXT: v_and_b32_e32 v1, 0xffff0000, v1
-; SI-NEXT: s_waitcnt vmcnt(0)
-; SI-NEXT: v_and_b32_e32 v2, 0xffff0000, v2
-; SI-NEXT: v_add_f32_e32 v5, 0x40c00000, v2
-; SI-NEXT: v_add_f32_e32 v1, 0x40c00000, v1
-; SI-NEXT: v_lshrrev_b32_e32 v2, 16, v5
-; SI-NEXT: v_alignbit_b32 v1, v2, v1, 16
-; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:200 ; 4-byte Folded Spill
-; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:420 ; 4-byte Folded Reload
-; SI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:416 ; 4-byte Folded Reload
-; SI-NEXT: v_and_b32_e32 v5, 0xffff0000, v5
-; SI-NEXT: s_waitcnt vmcnt(1)
-; SI-NEXT: v_and_b32_e32 v1, 0xffff0000, v1
-; SI-NEXT: s_waitcnt vmcnt(0)
-; SI-NEXT: v_and_b32_e32 v2, 0xffff0000, v2
-; SI-NEXT: v_add_f32_e32 v6, 0x40c00000, v2
-; SI-NEXT: v_add_f32_e32 v1, 0x40c00000, v1
-; SI-NEXT: v_lshrrev_b32_e32 v2, 16, v6
-; SI-NEXT: v_alignbit_b32 v1, v2, v1, 16
-; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:196 ; 4-byte Folded Spill
-; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:412 ; 4-byte Folded Reload
-; SI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:408 ; 4-byte Folded Reload
-; SI-NEXT: v_and_b32_e32 v6, 0xffff0000, v6
-; SI-NEXT: s_waitcnt vmcnt(1)
-; SI-NEXT: v_and_b32_e32 v1, 0xffff0000, v1
-; SI-NEXT: s_waitcnt vmcnt(0)
-; SI-NEXT: v_and_b32_e32 v2, 0xffff0000, v2
-; SI-NEXT: v_add_f32_e32 v2, 0x40c00000, v2
-; SI-NEXT: v_add_f32_e32 v1, 0x40c00000, v1
-; SI-NEXT: v_lshrrev_b32_e32 v3, 16, v2
-; SI-NEXT: v_alignbit_b32 v1, v3, v1, 16
-; SI-NEXT: v_and_b32_e32 v3, 0xffff0000, v51
-; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:156 ; 4-byte Folded Spill
-; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_and_b32_e32 v1, 0xffff0000, v31
-; SI-NEXT: v_add_f32_e32 v3, 0x40c00000, v3
-; SI-NEXT: v_add_f32_e32 v1, 0x40c00000, v1
-; SI-NEXT: v_lshrrev_b32_e32 v9, 16, v3
-; SI-NEXT: v_alignbit_b32 v1, v9, v1, 16
-; SI-NEXT: v_and_b32_e32 v9, 0xffff0000, v7
-; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:160 ; 4-byte Folded Spill
-; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_and_b32_e32 v1, 0xffff0000, v50
-; SI-NEXT: v_add_f32_e32 v9, 0x40c00000, v9
-; SI-NEXT: v_add_f32_e32 v1, 0x40c00000, v1
-; SI-NEXT: v_lshrrev_b32_e32 v11, 16, v9
-; SI-NEXT: v_alignbit_b32 v1, v11, v1, 16
-; SI-NEXT: v_and_b32_e32 v11, 0xffff0000, v41
-; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:164 ; 4-byte Folded Spill
-; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_and_b32_e32 v1, 0xffff0000, v42
-; SI-NEXT: v_add_f32_e32 v11, 0x40c00000, v11
-; SI-NEXT: v_add_f32_e32 v1, 0x40c00000, v1
-; SI-NEXT: v_lshrrev_b32_e32 v14, 16, v11
-; SI-NEXT: v_alignbit_b32 v1, v14, v1, 16
-; SI-NEXT: v_and_b32_e32 v14, 0xffff0000, v25
-; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:168 ; 4-byte Folded Spill
-; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_and_b32_e32 v1, 0xffff0000, v29
-; SI-NEXT: v_add_f32_e32 v14, 0x40c00000, v14
-; SI-NEXT: v_add_f32_e32 v1, 0x40c00000, v1
-; SI-NEXT: v_lshrrev_b32_e32 v16, 16, v14
-; SI-NEXT: v_alignbit_b32 v51, v16, v1, 16
-; SI-NEXT: v_and_b32_e32 v16, 0xffff0000, v54
-; SI-NEXT: v_and_b32_e32 v1, 0xffff0000, v24
-; SI-NEXT: v_add_f32_e32 v16, 0x40c00000, v16
-; SI-NEXT: v_add_f32_e32 v1, 0x40c00000, v1
-; SI-NEXT: v_lshrrev_b32_e32 v18, 16, v16
-; SI-NEXT: v_alignbit_b32 v1, v18, v1, 16
-; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:172 ; 4-byte Folded Spill
-; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:404 ; 4-byte Folded Reload
-; SI-NEXT: buffer_load_dword v7, off, s[0:3], s32 offset:396 ; 4-byte Folded Reload
-; SI-NEXT: v_add_f32_e32 v24, 0x40c00000, v23
-; SI-NEXT: v_lshrrev_b32_e32 v23, 16, v24
-; SI-NEXT: v_and_b32_e32 v25, 0xffff0000, v53
-; SI-NEXT: v_add_f32_e32 v25, 0x40c00000, v25
-; SI-NEXT: v_lshrrev_b32_e32 v26, 16, v25
-; SI-NEXT: v_add_f32_e32 v29, 0x40c00000, v27
-; SI-NEXT: v_lshrrev_b32_e32 v27, 16, v29
-; SI-NEXT: v_add_f32_e32 v31, 0x40c00000, v28
-; SI-NEXT: v_lshrrev_b32_e32 v28, 16, v31
-; SI-NEXT: v_and_b32_e32 v38, 0xffff0000, v2
-; SI-NEXT: v_and_b32_e32 v34, 0xffff0000, v9
-; SI-NEXT: v_and_b32_e32 v37, 0xffff0000, v3
-; SI-NEXT: v_and_b32_e32 v16, 0xffff0000, v16
-; SI-NEXT: v_and_b32_e32 v14, 0xffff0000, v14
-; SI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:356 ; 4-byte Folded Reload
-; SI-NEXT: buffer_load_dword v33, off, s[0:3], s32 offset:292 ; 4-byte Folded Reload
-; SI-NEXT: buffer_load_dword v9, off, s[0:3], s32 offset:280 ; 4-byte Folded Reload
-; SI-NEXT: buffer_load_dword v3, off, s[0:3], s32 offset:276 ; 4-byte Folded Reload
-; SI-NEXT: s_waitcnt vmcnt(5)
-; SI-NEXT: v_and_b32_e32 v1, 0xffff0000, v1
-; SI-NEXT: v_add_f32_e32 v1, 0x40c00000, v1
-; SI-NEXT: v_alignbit_b32 v18, v20, v1, 16
-; SI-NEXT: s_waitcnt vmcnt(4)
-; SI-NEXT: v_and_b32_e32 v20, 0xffff0000, v7
-; SI-NEXT: v_and_b32_e32 v1, 0xffff0000, v43
-; SI-NEXT: v_add_f32_e32 v20, 0x40c00000, v20
-; SI-NEXT: v_add_f32_e32 v1, 0x40c00000, v1
-; SI-NEXT: v_lshrrev_b32_e32 v22, 16, v20
-; SI-NEXT: v_alignbit_b32 v1, v22, v1, 16
-; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:176 ; 4-byte Folded Spill
-; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_and_b32_e32 v1, 0xffff0000, v52
-; SI-NEXT: v_add_f32_e32 v1, 0x40c00000, v1
-; SI-NEXT: v_alignbit_b32 v1, v23, v1, 16
-; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:180 ; 4-byte Folded Spill
-; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_and_b32_e32 v1, 0xffff0000, v12
-; SI-NEXT: v_add_f32_e32 v1, 0x40c00000, v1
-; SI-NEXT: v_alignbit_b32 v1, v26, v1, 16
-; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:184 ; 4-byte Folded Spill
-; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:392 ; 4-byte Folded Reload
-; SI-NEXT: v_and_b32_e32 v7, 0xffff0000, v32
-; SI-NEXT: buffer_load_dword v12, off, s[0:3], s32 offset:348 ; 4-byte Folded Reload
-; SI-NEXT: v_and_b32_e32 v26, 0xffff0000, v29
-; SI-NEXT: s_waitcnt vmcnt(7)
-; SI-NEXT: v_and_b32_e32 v33, 0xffff0000, v33
-; SI-NEXT: v_add_f32_e32 v33, 0x40c00000, v33
-; SI-NEXT: v_lshrrev_b32_e32 v61, 16, v33
-; SI-NEXT: v_and_b32_e32 v33, 0xffff0000, v11
-; SI-NEXT: buffer_load_dword v11, off, s[0:3], s32 offset:288 ; 4-byte Folded Reload
-; SI-NEXT: s_waitcnt vmcnt(7)
-; SI-NEXT: v_and_b32_e32 v9, 0xffff0000, v9
-; SI-NEXT: v_and_b32_e32 v2, 0xffff0000, v2
-; SI-NEXT: v_add_f32_e32 v2, 0x40c00000, v2
-; SI-NEXT: v_add_f32_e32 v9, 0x40c00000, v9
-; SI-NEXT: s_waitcnt vmcnt(6)
-; SI-NEXT: v_and_b32_e32 v3, 0xffff0000, v3
-; SI-NEXT: v_add_f32_e32 v3, 0x40c00000, v3
-; SI-NEXT: s_waitcnt vmcnt(2)
-; SI-NEXT: v_and_b32_e32 v1, 0xffff0000, v1
-; SI-NEXT: v_add_f32_e32 v1, 0x40c00000, v1
-; SI-NEXT: v_alignbit_b32 v1, v27, v1, 16
-; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:188 ; 4-byte Folded Spill
-; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:388 ; 4-byte Folded Reload
-; SI-NEXT: s_waitcnt vmcnt(3)
-; SI-NEXT: v_and_b32_e32 v29, 0xffff0000, v12
-; SI-NEXT: v_add_f32_e32 v29, 0x40c00000, v29
-; SI-NEXT: v_and_b32_e32 v12, 0xffff0000, v25
-; SI-NEXT: v_and_b32_e32 v25, 0xffff0000, v13
-; SI-NEXT: v_add_f32_e32 v25, 0x40c00000, v25
-; SI-NEXT: v_and_b32_e32 v27, 0xffff0000, v24
-; SI-NEXT: v_and_b32_e32 v24, 0xffff0000, v15
-; SI-NEXT: v_add_f32_e32 v24, 0x40c00000, v24
-; SI-NEXT: v_lshrrev_b32_e32 v58, 16, v24
-; SI-NEXT: buffer_load_dword v15, off, s[0:3], s32 offset:308 ; 4-byte Folded Reload
-; SI-NEXT: s_waitcnt vmcnt(1)
-; SI-NEXT: v_and_b32_e32 v1, 0xffff0000, v1
-; SI-NEXT: v_add_f32_e32 v1, 0x40c00000, v1
-; SI-NEXT: v_alignbit_b32 v1, v28, v1, 16
-; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:192 ; 4-byte Folded Spill
-; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:384 ; 4-byte Folded Reload
-; SI-NEXT: s_waitcnt vmcnt(0)
-; SI-NEXT: v_and_b32_e32 v1, 0xffff0000, v1
-; SI-NEXT: v_add_f32_e32 v1, 0x40c00000, v1
-; SI-NEXT: v_alignbit_b32 v52, v30, v1, 16
-; SI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:372 ; 4-byte Folded Reload
-; SI-NEXT: v_lshrrev_b32_e32 v30, 16, v29
-; SI-NEXT: s_waitcnt vmcnt(0)
-; SI-NEXT: v_and_b32_e32 v1, 0xffff0000, v1
-; SI-NEXT: v_add_f32_e32 v1, 0x40c00000, v1
-; SI-NEXT: v_alignbit_b32 v36, v35, v1, 16
-; SI-NEXT: v_and_b32_e32 v1, 0xffff0000, v8
-; SI-NEXT: buffer_load_dword v8, off, s[0:3], s32 offset:368 ; 4-byte Folded Reload
-; SI-NEXT: v_add_f32_e32 v1, 0x40c00000, v1
-; SI-NEXT: s_waitcnt vmcnt(0)
-; SI-NEXT: v_and_b32_e32 v32, 0xffff0000, v8
-; SI-NEXT: v_add_f32_e32 v32, 0x40c00000, v32
-; SI-NEXT: v_lshrrev_b32_e32 v49, 16, v32
-; SI-NEXT: v_alignbit_b32 v48, v49, v1, 16
-; SI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:364 ; 4-byte Folded Reload
-; SI-NEXT: v_and_b32_e32 v8, 0xffff0000, v31
-; SI-NEXT: v_and_b32_e32 v31, 0xffff0000, v10
-; SI-NEXT: v_add_f32_e32 v31, 0x40c00000, v31
-; SI-NEXT: v_lshrrev_b32_e32 v59, 16, v31
-; SI-NEXT: v_lshrrev_b32_e32 v10, 16, v25
-; SI-NEXT: v_and_b32_e32 v32, 0xffff0000, v20
-; SI-NEXT: buffer_load_dword v31, off, s[0:3], s32 offset:296 ; 4-byte Folded Reload
-; SI-NEXT: s_waitcnt vmcnt(1)
-; SI-NEXT: v_and_b32_e32 v1, 0xffff0000, v1
-; SI-NEXT: v_add_f32_e32 v1, 0x40c00000, v1
-; SI-NEXT: v_alignbit_b32 v28, v59, v1, 16
-; SI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:380 ; 4-byte Folded Reload
-; SI-NEXT: v_alignbit_b32 v26, v28, v26, 16
-; SI-NEXT: s_waitcnt vmcnt(1)
-; SI-NEXT: v_and_b32_e32 v31, 0xffff0000, v31
-; SI-NEXT: v_add_f32_e32 v31, 0x40c00000, v31
-; SI-NEXT: v_alignbit_b32 v46, v61, v31, 16
-; SI-NEXT: buffer_load_dword v31, off, s[0:3], s32 offset:284 ; 4-byte Folded Reload
-; SI-NEXT: v_and_b32_e32 v11, 0xffff0000, v11
-; SI-NEXT: v_add_f32_e32 v11, 0x40c00000, v11
-; SI-NEXT: s_waitcnt vmcnt(1)
-; SI-NEXT: v_and_b32_e32 v1, 0xffff0000, v1
-; SI-NEXT: v_add_f32_e32 v1, 0x40c00000, v1
-; SI-NEXT: v_alignbit_b32 v21, v30, v1, 16
-; SI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:344 ; 4-byte Folded Reload
-; SI-NEXT: s_waitcnt vmcnt(1)
-; SI-NEXT: v_and_b32_e32 v31, 0xffff0000, v31
-; SI-NEXT: v_add_f32_e32 v31, 0x40c00000, v31
-; SI-NEXT: v_lshrrev_b32_e32 v62, 16, v31
-; SI-NEXT: s_waitcnt vmcnt(0)
-; SI-NEXT: v_and_b32_e32 v1, 0xffff0000, v1
-; SI-NEXT: v_add_f32_e32 v1, 0x40c00000, v1
-; SI-NEXT: v_alignbit_b32 v23, v10, v1, 16
-; SI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:328 ; 4-byte Folded Reload
-; SI-NEXT: v_alignbit_b32 v63, v23, v27, 16
-; SI-NEXT: v_alignbit_b32 v27, v21, v12, 16
-; SI-NEXT: s_waitcnt vmcnt(0)
-; SI-NEXT: v_and_b32_e32 v1, 0xffff0000, v1
-; SI-NEXT: v_add_f32_e32 v1, 0x40c00000, v1
-; SI-NEXT: v_alignbit_b32 v57, v58, v1, 16
-; SI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:312 ; 4-byte Folded Reload
-; SI-NEXT: s_waitcnt vmcnt(0)
-; SI-NEXT: v_and_b32_e32 v1, 0xffff0000, v1
-; SI-NEXT: v_add_f32_e32 v20, 0x40c00000, v1
-; SI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:400 ; 4-byte Folded Reload
-; SI-NEXT: s_waitcnt vmcnt(0)
-; SI-NEXT: v_and_b32_e32 v1, 0xffff0000, v1
-; SI-NEXT: v_add_f32_e32 v1, 0x40c00000, v1
-; SI-NEXT: v_lshrrev_b32_e32 v1, 16, v1
-; SI-NEXT: v_alignbit_b32 v17, v1, v20, 16
-; SI-NEXT: v_and_b32_e32 v20, 0xffff0000, v15
-; SI-NEXT: buffer_load_dword v15, off, s[0:3], s32 offset:376 ; 4-byte Folded Reload
-; SI-NEXT: v_add_f32_e32 v20, 0x40c00000, v20
-; SI-NEXT: v_alignbit_b32 v19, v17, v19, 16
-; SI-NEXT: s_waitcnt vmcnt(0)
-; SI-NEXT: v_and_b32_e32 v24, 0xffff0000, v15
-; SI-NEXT: buffer_load_dword v15, off, s[0:3], s32 offset:304 ; 4-byte Folded Reload
-; SI-NEXT: v_add_f32_e32 v24, 0x40c00000, v24
-; SI-NEXT: v_lshrrev_b32_e32 v47, 16, v24
-; SI-NEXT: v_alignbit_b32 v56, v47, v20, 16
-; SI-NEXT: v_alignbit_b32 v20, v62, v11, 16
-; SI-NEXT: buffer_load_dword v11, off, s[0:3], s32 offset:152 ; 4-byte Folded Reload
-; SI-NEXT: v_alignbit_b32 v16, v56, v16, 16
-; SI-NEXT: s_waitcnt vmcnt(1)
-; SI-NEXT: v_and_b32_e32 v25, 0xffff0000, v15
-; SI-NEXT: buffer_load_dword v15, off, s[0:3], s32 offset:300 ; 4-byte Folded Reload
-; SI-NEXT: v_add_f32_e32 v25, 0x40c00000, v25
-; SI-NEXT: s_waitcnt vmcnt(1)
-; SI-NEXT: v_and_b32_e32 v11, 0xffff0000, v11
-; SI-NEXT: v_add_f32_e32 v11, 0x40c00000, v11
-; SI-NEXT: v_lshrrev_b32_e32 v45, 16, v11
-; SI-NEXT: buffer_load_dword v11, off, s[0:3], s32 offset:336 ; 4-byte Folded Reload
-; SI-NEXT: v_alignbit_b32 v22, v45, v9, 16
-; SI-NEXT: buffer_load_dword v9, off, s[0:3], s32 offset:272 ; 4-byte Folded Reload
-; SI-NEXT: s_waitcnt vmcnt(2)
-; SI-NEXT: v_and_b32_e32 v29, 0xffff0000, v15
-; SI-NEXT: v_add_f32_e32 v29, 0x40c00000, v29
-; SI-NEXT: v_lshrrev_b32_e32 v60, 16, v29
-; SI-NEXT: v_alignbit_b32 v13, v60, v25, 16
-; SI-NEXT: buffer_load_dword v15, off, s[0:3], s32 offset:324 ; 4-byte Folded Reload
-; SI-NEXT: s_waitcnt vmcnt(2)
-; SI-NEXT: v_and_b32_e32 v11, 0xffff0000, v11
-; SI-NEXT: v_add_f32_e32 v11, 0x40c00000, v11
-; SI-NEXT: s_waitcnt vmcnt(1)
-; SI-NEXT: v_and_b32_e32 v9, 0xffff0000, v9
-; SI-NEXT: v_add_f32_e32 v9, 0x40c00000, v9
-; SI-NEXT: v_lshrrev_b32_e32 v44, 16, v9
-; SI-NEXT: v_alignbit_b32 v24, v44, v3, 16
-; SI-NEXT: buffer_load_dword v3, off, s[0:3], s32 offset:352 ; 4-byte Folded Reload
-; SI-NEXT: buffer_load_dword v9, off, s[0:3], s32 offset:340 ; 4-byte Folded Reload
-; SI-NEXT: v_lshrrev_b32_e32 v11, 16, v11
-; SI-NEXT: s_waitcnt vmcnt(2)
-; SI-NEXT: v_and_b32_e32 v39, 0xffff0000, v15
-; SI-NEXT: buffer_load_dword v15, off, s[0:3], s32 offset:320 ; 4-byte Folded Reload
-; SI-NEXT: v_add_f32_e32 v39, 0x40c00000, v39
-; SI-NEXT: s_waitcnt vmcnt(2)
-; SI-NEXT: v_and_b32_e32 v3, 0xffff0000, v3
-; SI-NEXT: v_add_f32_e32 v3, 0x40c00000, v3
-; SI-NEXT: s_waitcnt vmcnt(1)
-; SI-NEXT: v_and_b32_e32 v9, 0xffff0000, v9
-; SI-NEXT: v_add_f32_e32 v9, 0x40c00000, v9
-; SI-NEXT: v_lshrrev_b32_e32 v3, 16, v3
-; SI-NEXT: v_alignbit_b32 v9, v11, v9, 16
-; SI-NEXT: buffer_store_dword v3, off, s[0:3], s32 offset:216 ; 4-byte Folded Spill
-; SI-NEXT: v_alignbit_b32 v2, v3, v2, 16
-; SI-NEXT: s_waitcnt vmcnt(1)
-; SI-NEXT: v_and_b32_e32 v50, 0xffff0000, v15
-; SI-NEXT: v_mov_b32_e32 v15, v24
-; SI-NEXT: v_add_f32_e32 v50, 0x40c00000, v50
-; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_lshrrev_b32_e32 v3, 16, v50
-; SI-NEXT: buffer_store_dword v3, off, s[0:3], s32 offset:244 ; 4-byte Folded Spill
-; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_alignbit_b32 v3, v3, v39, 16
-; SI-NEXT: v_alignbit_b32 v4, v3, v4, 16
-; SI-NEXT: buffer_store_dword v4, off, s[0:3], s32 offset:248 ; 4-byte Folded Spill
-; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_alignbit_b32 v4, v9, v5, 16
-; SI-NEXT: v_alignbit_b32 v5, v36, v7, 16
-; SI-NEXT: buffer_store_dword v4, off, s[0:3], s32 offset:240 ; 4-byte Folded Spill
-; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_alignbit_b32 v4, v2, v6, 16
-; SI-NEXT: v_alignbit_b32 v6, v46, v33, 16
-; SI-NEXT: buffer_store_dword v4, off, s[0:3], s32 offset:232 ; 4-byte Folded Spill
-; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_alignbit_b32 v4, v24, v38, 16
-; SI-NEXT: v_alignbit_b32 v38, v48, v8, 16
-; SI-NEXT: buffer_store_dword v4, off, s[0:3], s32 offset:236 ; 4-byte Folded Spill
-; SI-NEXT: buffer_store_dword v22, off, s[0:3], s32 offset:212 ; 4-byte Folded Spill
-; SI-NEXT: s_waitcnt expcnt(1)
-; SI-NEXT: v_alignbit_b32 v4, v22, v37, 16
-; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_alignbit_b32 v22, v57, v32, 16
-; SI-NEXT: buffer_store_dword v4, off, s[0:3], s32 offset:228 ; 4-byte Folded Spill
-; SI-NEXT: buffer_store_dword v20, off, s[0:3], s32 offset:208 ; 4-byte Folded Spill
-; SI-NEXT: s_waitcnt expcnt(1)
-; SI-NEXT: v_alignbit_b32 v4, v20, v34, 16
-; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_mov_b32_e32 v20, v52
-; SI-NEXT: buffer_store_dword v4, off, s[0:3], s32 offset:224 ; 4-byte Folded Spill
-; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_alignbit_b32 v4, v13, v14, 16
-; SI-NEXT: v_mov_b32_e32 v14, v51
-; SI-NEXT: buffer_store_dword v4, off, s[0:3], s32 offset:220 ; 4-byte Folded Spill
-; SI-NEXT: .LBB105_5: ; %end
-; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: buffer_load_dword v4, off, s[0:3], s32 offset:204 ; 4-byte Folded Reload
-; SI-NEXT: v_and_b32_e32 v2, 0xffff, v2
-; SI-NEXT: v_lshlrev_b32_e32 v1, 16, v1
-; SI-NEXT: s_waitcnt vmcnt(0)
-; SI-NEXT: v_and_b32_e32 v7, 0xffff, v4
-; SI-NEXT: buffer_load_dword v4, off, s[0:3], s32 offset:248 ; 4-byte Folded Reload
-; SI-NEXT: s_waitcnt vmcnt(0)
-; SI-NEXT: v_lshlrev_b32_e32 v8, 16, v4
-; SI-NEXT: v_or_b32_e32 v7, v7, v8
-; SI-NEXT: buffer_store_dword v7, v0, s[0:3], 0 offen
-; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_and_b32_e32 v7, 0xffff, v3
-; SI-NEXT: buffer_load_dword v3, off, s[0:3], s32 offset:244 ; 4-byte Folded Reload
-; SI-NEXT: s_waitcnt vmcnt(0)
-; SI-NEXT: v_lshlrev_b32_e32 v8, 16, v3
-; SI-NEXT: v_or_b32_e32 v7, v7, v8
-; SI-NEXT: v_add_i32_e32 v8, vcc, 4, v0
-; SI-NEXT: buffer_store_dword v7, v8, s[0:3], 0 offen
-; SI-NEXT: buffer_load_dword v4, off, s[0:3], s32 offset:200 ; 4-byte Folded Reload
-; SI-NEXT: buffer_load_dword v3, off, s[0:3], s32 offset:240 ; 4-byte Folded Reload
-; SI-NEXT: s_waitcnt vmcnt(1) expcnt(0)
-; SI-NEXT: v_and_b32_e32 v7, 0xffff, v4
-; SI-NEXT: s_waitcnt vmcnt(0)
-; SI-NEXT: v_lshlrev_b32_e32 v4, 16, v3
-; SI-NEXT: v_or_b32_e32 v4, v7, v4
-; SI-NEXT: v_add_i32_e32 v7, vcc, 8, v0
-; SI-NEXT: buffer_store_dword v4, v7, s[0:3], 0 offen
-; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_and_b32_e32 v4, 0xffff, v9
-; SI-NEXT: v_lshlrev_b32_e32 v7, 16, v11
-; SI-NEXT: v_or_b32_e32 v4, v4, v7
-; SI-NEXT: v_add_i32_e32 v7, vcc, 12, v0
-; SI-NEXT: buffer_store_dword v4, v7, s[0:3], 0 offen
-; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: buffer_load_dword v4, off, s[0:3], s32 offset:196 ; 4-byte Folded Reload
-; SI-NEXT: buffer_load_dword v3, off, s[0:3], s32 offset:232 ; 4-byte Folded Reload
-; SI-NEXT: s_waitcnt vmcnt(1)
-; SI-NEXT: v_and_b32_e32 v4, 0xffff, v4
-; SI-NEXT: s_waitcnt vmcnt(0)
-; SI-NEXT: v_lshlrev_b32_e32 v7, 16, v3
-; SI-NEXT: v_or_b32_e32 v4, v4, v7
-; SI-NEXT: v_add_i32_e32 v7, vcc, 16, v0
-; SI-NEXT: buffer_store_dword v4, v7, s[0:3], 0 offen
-; SI-NEXT: buffer_load_dword v3, off, s[0:3], s32 offset:216 ; 4-byte Folded Reload
-; SI-NEXT: s_waitcnt vmcnt(0)
-; SI-NEXT: v_lshlrev_b32_e32 v3, 16, v3
-; SI-NEXT: v_or_b32_e32 v2, v2, v3
-; SI-NEXT: v_add_i32_e32 v3, vcc, 20, v0
-; SI-NEXT: buffer_store_dword v2, v3, s[0:3], 0 offen
-; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:156 ; 4-byte Folded Reload
-; SI-NEXT: buffer_load_dword v3, off, s[0:3], s32 offset:236 ; 4-byte Folded Reload
-; SI-NEXT: s_waitcnt vmcnt(1)
-; SI-NEXT: v_and_b32_e32 v2, 0xffff, v2
-; SI-NEXT: s_waitcnt vmcnt(0)
-; SI-NEXT: v_lshlrev_b32_e32 v3, 16, v3
-; SI-NEXT: v_or_b32_e32 v2, v2, v3
-; SI-NEXT: v_add_i32_e32 v3, vcc, 24, v0
-; SI-NEXT: buffer_store_dword v2, v3, s[0:3], 0 offen
-; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_and_b32_e32 v2, 0xffff, v15
-; SI-NEXT: v_lshlrev_b32_e32 v3, 16, v44
-; SI-NEXT: v_or_b32_e32 v2, v2, v3
-; SI-NEXT: v_add_i32_e32 v3, vcc, 28, v0
-; SI-NEXT: buffer_store_dword v2, v3, s[0:3], 0 offen
-; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:160 ; 4-byte Folded Reload
-; SI-NEXT: buffer_load_dword v3, off, s[0:3], s32 offset:228 ; 4-byte Folded Reload
-; SI-NEXT: s_waitcnt vmcnt(1)
-; SI-NEXT: v_and_b32_e32 v2, 0xffff, v2
-; SI-NEXT: s_waitcnt vmcnt(0)
-; SI-NEXT: v_lshlrev_b32_e32 v3, 16, v3
-; SI-NEXT: v_or_b32_e32 v2, v2, v3
-; SI-NEXT: v_add_i32_e32 v3, vcc, 32, v0
-; SI-NEXT: buffer_store_dword v2, v3, s[0:3], 0 offen
-; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:212 ; 4-byte Folded Reload
-; SI-NEXT: v_lshlrev_b32_e32 v3, 16, v45
-; SI-NEXT: s_waitcnt vmcnt(0)
-; SI-NEXT: v_and_b32_e32 v2, 0xffff, v2
-; SI-NEXT: v_or_b32_e32 v2, v2, v3
-; SI-NEXT: v_add_i32_e32 v3, vcc, 36, v0
-; SI-NEXT: buffer_store_dword v2, v3, s[0:3], 0 offen
-; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:164 ; 4-byte Folded Reload
-; SI-NEXT: buffer_load_dword v3, off, s[0:3], s32 offset:224 ; 4-byte Folded Reload
-; SI-NEXT: s_waitcnt vmcnt(1)
-; SI-NEXT: v_and_b32_e32 v2, 0xffff, v2
-; SI-NEXT: s_waitcnt vmcnt(0)
-; SI-NEXT: v_lshlrev_b32_e32 v3, 16, v3
-; SI-NEXT: v_or_b32_e32 v2, v2, v3
-; SI-NEXT: v_add_i32_e32 v3, vcc, 40, v0
-; SI-NEXT: buffer_store_dword v2, v3, s[0:3], 0 offen
-; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:208 ; 4-byte Folded Reload
-; SI-NEXT: v_lshlrev_b32_e32 v3, 16, v62
-; SI-NEXT: s_waitcnt vmcnt(0)
-; SI-NEXT: v_and_b32_e32 v2, 0xffff, v2
-; SI-NEXT: v_or_b32_e32 v2, v2, v3
-; SI-NEXT: v_add_i32_e32 v3, vcc, 44, v0
-; SI-NEXT: buffer_store_dword v2, v3, s[0:3], 0 offen
-; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:168 ; 4-byte Folded Reload
-; SI-NEXT: v_lshlrev_b32_e32 v3, 16, v6
-; SI-NEXT: s_waitcnt vmcnt(0)
-; SI-NEXT: v_and_b32_e32 v2, 0xffff, v2
-; SI-NEXT: v_or_b32_e32 v2, v2, v3
-; SI-NEXT: v_add_i32_e32 v3, vcc, 48, v0
-; SI-NEXT: buffer_store_dword v2, v3, s[0:3], 0 offen
-; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_and_b32_e32 v2, 0xffff, v46
-; SI-NEXT: v_lshlrev_b32_e32 v3, 16, v61
-; SI-NEXT: v_or_b32_e32 v2, v2, v3
-; SI-NEXT: v_add_i32_e32 v3, vcc, 52, v0
-; SI-NEXT: buffer_store_dword v2, v3, s[0:3], 0 offen
-; SI-NEXT: buffer_load_dword v3, off, s[0:3], s32 offset:220 ; 4-byte Folded Reload
-; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_and_b32_e32 v2, 0xffff, v14
-; SI-NEXT: s_waitcnt vmcnt(0)
-; SI-NEXT: v_lshlrev_b32_e32 v3, 16, v3
-; SI-NEXT: v_or_b32_e32 v2, v2, v3
-; SI-NEXT: v_add_i32_e32 v3, vcc, 56, v0
-; SI-NEXT: buffer_store_dword v2, v3, s[0:3], 0 offen
-; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_and_b32_e32 v2, 0xffff, v13
-; SI-NEXT: v_lshlrev_b32_e32 v3, 16, v60
-; SI-NEXT: v_or_b32_e32 v2, v2, v3
-; SI-NEXT: v_add_i32_e32 v3, vcc, 60, v0
-; SI-NEXT: buffer_store_dword v2, v3, s[0:3], 0 offen
-; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:172 ; 4-byte Folded Reload
-; SI-NEXT: v_lshlrev_b32_e32 v3, 16, v16
-; SI-NEXT: s_waitcnt vmcnt(0)
-; SI-NEXT: v_and_b32_e32 v2, 0xffff, v2
-; SI-NEXT: v_or_b32_e32 v2, v2, v3
-; SI-NEXT: v_add_i32_e32 v3, vcc, 64, v0
-; SI-NEXT: buffer_store_dword v2, v3, s[0:3], 0 offen
-; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_and_b32_e32 v2, 0xffff, v56
-; SI-NEXT: v_lshlrev_b32_e32 v3, 16, v47
-; SI-NEXT: v_or_b32_e32 v2, v2, v3
-; SI-NEXT: v_add_i32_e32 v3, vcc, 0x44, v0
-; SI-NEXT: buffer_store_dword v2, v3, s[0:3], 0 offen
-; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_and_b32_e32 v2, 0xffff, v18
-; SI-NEXT: v_lshlrev_b32_e32 v3, 16, v19
-; SI-NEXT: v_or_b32_e32 v2, v2, v3
-; SI-NEXT: v_add_i32_e32 v3, vcc, 0x48, v0
-; SI-NEXT: buffer_store_dword v2, v3, s[0:3], 0 offen
-; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_and_b32_e32 v2, 0xffff, v17
-; SI-NEXT: v_or_b32_e32 v1, v2, v1
-; SI-NEXT: v_add_i32_e32 v2, vcc, 0x4c, v0
-; SI-NEXT: buffer_store_dword v1, v2, s[0:3], 0 offen
-; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:176 ; 4-byte Folded Reload
-; SI-NEXT: v_lshlrev_b32_e32 v2, 16, v22
-; SI-NEXT: s_waitcnt vmcnt(0)
-; SI-NEXT: v_and_b32_e32 v1, 0xffff, v1
-; SI-NEXT: v_or_b32_e32 v1, v1, v2
-; SI-NEXT: v_add_i32_e32 v2, vcc, 0x50, v0
-; SI-NEXT: buffer_store_dword v1, v2, s[0:3], 0 offen
-; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_and_b32_e32 v1, 0xffff, v57
-; SI-NEXT: v_lshlrev_b32_e32 v2, 16, v58
-; SI-NEXT: v_or_b32_e32 v1, v1, v2
-; SI-NEXT: v_add_i32_e32 v2, vcc, 0x54, v0
-; SI-NEXT: buffer_store_dword v1, v2, s[0:3], 0 offen
-; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:180 ; 4-byte Folded Reload
-; SI-NEXT: v_lshlrev_b32_e32 v2, 16, v63
-; SI-NEXT: s_waitcnt vmcnt(0)
-; SI-NEXT: v_and_b32_e32 v1, 0xffff, v1
-; SI-NEXT: v_or_b32_e32 v1, v1, v2
-; SI-NEXT: v_add_i32_e32 v2, vcc, 0x58, v0
-; SI-NEXT: buffer_store_dword v1, v2, s[0:3], 0 offen
-; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_and_b32_e32 v1, 0xffff, v23
-; SI-NEXT: v_lshlrev_b32_e32 v2, 16, v10
-; SI-NEXT: v_or_b32_e32 v1, v1, v2
-; SI-NEXT: v_add_i32_e32 v2, vcc, 0x5c, v0
-; SI-NEXT: buffer_store_dword v1, v2, s[0:3], 0 offen
-; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:184 ; 4-byte Folded Reload
-; SI-NEXT: v_lshlrev_b32_e32 v2, 16, v27
-; SI-NEXT: s_waitcnt vmcnt(0)
-; SI-NEXT: v_and_b32_e32 v1, 0xffff, v1
-; SI-NEXT: v_or_b32_e32 v1, v1, v2
-; SI-NEXT: v_add_i32_e32 v2, vcc, 0x60, v0
-; SI-NEXT: buffer_store_dword v1, v2, s[0:3], 0 offen
-; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_and_b32_e32 v1, 0xffff, v21
-; SI-NEXT: v_lshlrev_b32_e32 v2, 16, v30
-; SI-NEXT: v_or_b32_e32 v1, v1, v2
-; SI-NEXT: v_add_i32_e32 v2, vcc, 0x64, v0
-; SI-NEXT: buffer_store_dword v1, v2, s[0:3], 0 offen
-; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:188 ; 4-byte Folded Reload
-; SI-NEXT: v_lshlrev_b32_e32 v2, 16, v26
-; SI-NEXT: s_waitcnt vmcnt(0)
-; SI-NEXT: v_and_b32_e32 v1, 0xffff, v1
-; SI-NEXT: v_or_b32_e32 v1, v1, v2
-; SI-NEXT: v_add_i32_e32 v2, vcc, 0x68, v0
-; SI-NEXT: buffer_store_dword v1, v2, s[0:3], 0 offen
-; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_and_b32_e32 v1, 0xffff, v28
-; SI-NEXT: v_lshlrev_b32_e32 v2, 16, v59
-; SI-NEXT: v_or_b32_e32 v1, v1, v2
-; SI-NEXT: v_add_i32_e32 v2, vcc, 0x6c, v0
-; SI-NEXT: buffer_store_dword v1, v2, s[0:3], 0 offen
-; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:192 ; 4-byte Folded Reload
-; SI-NEXT: v_lshlrev_b32_e32 v2, 16, v38
-; SI-NEXT: s_waitcnt vmcnt(0)
-; SI-NEXT: v_and_b32_e32 v1, 0xffff, v1
-; SI-NEXT: v_or_b32_e32 v1, v1, v2
-; SI-NEXT: v_add_i32_e32 v2, vcc, 0x70, v0
-; SI-NEXT: buffer_store_dword v1, v2, s[0:3], 0 offen
-; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_and_b32_e32 v1, 0xffff, v48
-; SI-NEXT: v_lshlrev_b32_e32 v2, 16, v49
-; SI-NEXT: v_or_b32_e32 v1, v1, v2
-; SI-NEXT: v_add_i32_e32 v2, vcc, 0x74, v0
-; SI-NEXT: buffer_store_dword v1, v2, s[0:3], 0 offen
-; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_and_b32_e32 v1, 0xffff, v20
-; SI-NEXT: v_lshlrev_b32_e32 v2, 16, v5
-; SI-NEXT: v_or_b32_e32 v1, v1, v2
-; SI-NEXT: v_add_i32_e32 v2, vcc, 0x78, v0
-; SI-NEXT: buffer_store_dword v1, v2, s[0:3], 0 offen
-; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_and_b32_e32 v1, 0xffff, v36
-; SI-NEXT: v_lshlrev_b32_e32 v2, 16, v35
-; SI-NEXT: v_or_b32_e32 v1, v1, v2
-; SI-NEXT: v_add_i32_e32 v0, vcc, 0x7c, v0
-; SI-NEXT: buffer_store_dword v1, v0, s[0:3], 0 offen
-; SI-NEXT: buffer_load_dword v63, off, s[0:3], s32 offset:84 ; 4-byte Folded Reload
-; SI-NEXT: buffer_load_dword v62, off, s[0:3], s32 offset:88 ; 4-byte Folded Reload
-; SI-NEXT: buffer_load_dword v61, off, s[0:3], s32 offset:92 ; 4-byte Folded Reload
-; SI-NEXT: buffer_load_dword v60, off, s[0:3], s32 offset:96 ; 4-byte Folded Reload
-; SI-NEXT: buffer_load_dword v59, off, s[0:3], s32 offset:100 ; 4-byte Folded Reload
-; SI-NEXT: buffer_load_dword v58, off, s[0:3], s32 offset:104 ; 4-byte Folded Reload
-; SI-NEXT: buffer_load_dword v57, off, s[0:3], s32 offset:108 ; 4-byte Folded Reload
-; SI-NEXT: buffer_load_dword v56, off, s[0:3], s32 offset:112 ; 4-byte Folded Reload
-; SI-NEXT: buffer_load_dword v47, off, s[0:3], s32 offset:116 ; 4-byte Folded Reload
-; SI-NEXT: buffer_load_dword v46, off, s[0:3], s32 offset:120 ; 4-byte Folded Reload
-; SI-NEXT: buffer_load_dword v45, off, s[0:3], s32 offset:124 ; 4-byte Folded Reload
-; SI-NEXT: buffer_load_dword v44, off, s[0:3], s32 offset:128 ; 4-byte Folded Reload
-; SI-NEXT: buffer_load_dword v43, off, s[0:3], s32 offset:132 ; 4-byte Folded Reload
-; SI-NEXT: buffer_load_dword v42, off, s[0:3], s32 offset:136 ; 4-byte Folded Reload
-; SI-NEXT: buffer_load_dword v41, off, s[0:3], s32 offset:140 ; 4-byte Folded Reload
-; SI-NEXT: buffer_load_dword v40, off, s[0:3], s32 offset:144 ; 4-byte Folded Reload
-; SI-NEXT: s_waitcnt vmcnt(0) expcnt(0)
-; SI-NEXT: s_setpc_b64 s[30:31]
+; SI-NEXT: s_branch .LBB105_2
;
; VI-LABEL: bitcast_v64bf16_to_v64i16_scalar:
; VI: ; %bb.0:
@@ -243834,50 +243836,50 @@ define inreg <64 x bfloat> @bitcast_v64i16_to_v64bf16_scalar(<64 x i16> inreg %a
; SI-NEXT: s_mov_b32 s60, s16
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_writelane_b32 v41, s17, 0
-; SI-NEXT: s_mov_b32 s61, s19
; SI-NEXT: v_writelane_b32 v41, s60, 1
-; SI-NEXT: s_mov_b32 s63, s18
-; SI-NEXT: v_writelane_b32 v41, s61, 2
+; SI-NEXT: s_mov_b32 s61, s18
+; SI-NEXT: v_writelane_b32 v41, s19, 2
; SI-NEXT: s_mov_b32 s72, s21
-; SI-NEXT: v_writelane_b32 v41, s63, 3
+; SI-NEXT: v_writelane_b32 v41, s61, 3
; SI-NEXT: v_writelane_b32 v41, s72, 4
; SI-NEXT: s_mov_b32 s74, s23
; SI-NEXT: v_writelane_b32 v41, s20, 5
; SI-NEXT: v_writelane_b32 v41, s74, 6
-; SI-NEXT: s_mov_b32 s75, s25
+; SI-NEXT: s_mov_b32 s76, s25
; SI-NEXT: v_writelane_b32 v41, s22, 7
-; SI-NEXT: v_writelane_b32 v41, s75, 8
-; SI-NEXT: s_mov_b32 s76, s27
+; SI-NEXT: v_writelane_b32 v41, s76, 8
+; SI-NEXT: s_mov_b32 s78, s27
; SI-NEXT: v_writelane_b32 v41, s24, 9
-; SI-NEXT: v_writelane_b32 v41, s76, 10
-; SI-NEXT: s_mov_b32 s93, s29
+; SI-NEXT: v_writelane_b32 v41, s78, 10
+; SI-NEXT: s_mov_b32 s79, s29
; SI-NEXT: v_writelane_b32 v41, s26, 11
-; SI-NEXT: v_writelane_b32 v41, s93, 12
-; SI-NEXT: v_readfirstlane_b32 s16, v2
+; SI-NEXT: v_writelane_b32 v41, s79, 12
+; SI-NEXT: v_readfirstlane_b32 s6, v2
; SI-NEXT: v_writelane_b32 v41, s28, 13
; SI-NEXT: v_readfirstlane_b32 s73, v4
-; SI-NEXT: v_writelane_b32 v41, s16, 14
-; SI-NEXT: v_readfirstlane_b32 s89, v3
+; SI-NEXT: v_writelane_b32 v41, s6, 14
+; SI-NEXT: v_readfirstlane_b32 s95, v3
; SI-NEXT: v_writelane_b32 v41, s73, 15
-; SI-NEXT: v_readfirstlane_b32 s90, v6
-; SI-NEXT: v_writelane_b32 v41, s89, 16
-; SI-NEXT: v_readfirstlane_b32 s91, v5
-; SI-NEXT: v_writelane_b32 v41, s90, 17
-; SI-NEXT: v_readfirstlane_b32 s34, v8
-; SI-NEXT: v_writelane_b32 v41, s91, 18
-; SI-NEXT: v_readfirstlane_b32 s35, v7
-; SI-NEXT: v_writelane_b32 v41, s34, 19
-; SI-NEXT: v_readfirstlane_b32 s36, v10
-; SI-NEXT: v_writelane_b32 v41, s35, 20
+; SI-NEXT: v_readfirstlane_b32 s36, v6
+; SI-NEXT: v_writelane_b32 v41, s95, 16
+; SI-NEXT: v_readfirstlane_b32 s30, v5
+; SI-NEXT: v_writelane_b32 v41, s36, 17
+; SI-NEXT: v_readfirstlane_b32 s37, v8
+; SI-NEXT: v_writelane_b32 v41, s30, 18
+; SI-NEXT: v_readfirstlane_b32 s38, v7
+; SI-NEXT: v_writelane_b32 v41, s37, 19
+; SI-NEXT: v_readfirstlane_b32 s39, v10
+; SI-NEXT: v_writelane_b32 v41, s38, 20
; SI-NEXT: v_writelane_b32 v40, s96, 32
-; SI-NEXT: v_readfirstlane_b32 s37, v9
-; SI-NEXT: v_writelane_b32 v41, s36, 21
+; SI-NEXT: v_readfirstlane_b32 s48, v9
+; SI-NEXT: v_writelane_b32 v41, s39, 21
+; SI-NEXT: v_writelane_b32 v40, s97, 33
; SI-NEXT: s_waitcnt vmcnt(7)
-; SI-NEXT: v_readfirstlane_b32 s62, v31
+; SI-NEXT: v_readfirstlane_b32 s92, v31
; SI-NEXT: s_waitcnt vmcnt(6)
-; SI-NEXT: v_readfirstlane_b32 s80, v32
+; SI-NEXT: v_readfirstlane_b32 s93, v32
; SI-NEXT: s_waitcnt vmcnt(5)
-; SI-NEXT: v_readfirstlane_b32 s69, v33
+; SI-NEXT: v_readfirstlane_b32 s90, v33
; SI-NEXT: buffer_load_dword v31, off, s[0:3], s32 offset:44
; SI-NEXT: buffer_load_dword v32, off, s[0:3], s32 offset:40
; SI-NEXT: buffer_load_dword v33, off, s[0:3], s32 offset:36
@@ -243889,20 +243891,19 @@ define inreg <64 x bfloat> @bitcast_v64i16_to_v64bf16_scalar(<64 x i16> inreg %a
; SI-NEXT: s_waitcnt vmcnt(12)
; SI-NEXT: v_readfirstlane_b32 s84, v34
; SI-NEXT: s_waitcnt vmcnt(11)
-; SI-NEXT: v_readfirstlane_b32 s68, v35
+; SI-NEXT: v_readfirstlane_b32 s94, v35
; SI-NEXT: s_waitcnt vmcnt(10)
; SI-NEXT: v_readfirstlane_b32 s83, v36
; SI-NEXT: s_waitcnt vmcnt(8)
; SI-NEXT: v_readfirstlane_b32 s87, v38
; SI-NEXT: buffer_load_dword v38, off, s[0:3], s32 offset:80
-; SI-NEXT: v_readfirstlane_b32 s6, v37
+; SI-NEXT: v_readfirstlane_b32 s91, v37
; SI-NEXT: buffer_load_dword v34, off, s[0:3], s32 offset:12
; SI-NEXT: buffer_load_dword v35, off, s[0:3], s32 offset:8
; SI-NEXT: buffer_load_dword v36, off, s[0:3], s32 offset:4
; SI-NEXT: buffer_load_dword v37, off, s[0:3], s32
-; SI-NEXT: v_writelane_b32 v40, s97, 33
-; SI-NEXT: v_readfirstlane_b32 s38, v12
-; SI-NEXT: v_writelane_b32 v41, s37, 22
+; SI-NEXT: v_readfirstlane_b32 s49, v12
+; SI-NEXT: v_writelane_b32 v41, s48, 22
; SI-NEXT: v_writelane_b32 v40, s98, 34
; SI-NEXT: v_readfirstlane_b32 s14, v30
; SI-NEXT: v_readfirstlane_b32 s15, v29
@@ -243912,21 +243913,21 @@ define inreg <64 x bfloat> @bitcast_v64i16_to_v64bf16_scalar(<64 x i16> inreg %a
; SI-NEXT: v_readfirstlane_b32 s11, v25
; SI-NEXT: v_readfirstlane_b32 s8, v24
; SI-NEXT: v_readfirstlane_b32 s9, v23
-; SI-NEXT: v_readfirstlane_b32 s88, v22
-; SI-NEXT: v_readfirstlane_b32 s29, v21
-; SI-NEXT: v_readfirstlane_b32 s79, v20
-; SI-NEXT: v_readfirstlane_b32 s27, v19
-; SI-NEXT: v_readfirstlane_b32 s78, v18
-; SI-NEXT: v_readfirstlane_b32 s25, v17
-; SI-NEXT: v_readfirstlane_b32 s77, v16
-; SI-NEXT: v_readfirstlane_b32 s23, v15
-; SI-NEXT: v_readfirstlane_b32 s39, v14
-; SI-NEXT: v_readfirstlane_b32 s21, v13
-; SI-NEXT: v_readfirstlane_b32 s19, v11
+; SI-NEXT: v_readfirstlane_b32 s89, v22
+; SI-NEXT: v_readfirstlane_b32 s7, v21
+; SI-NEXT: v_readfirstlane_b32 s88, v20
+; SI-NEXT: v_readfirstlane_b32 s29, v19
+; SI-NEXT: v_readfirstlane_b32 s77, v18
+; SI-NEXT: v_readfirstlane_b32 s27, v17
+; SI-NEXT: v_readfirstlane_b32 s75, v16
+; SI-NEXT: v_readfirstlane_b32 s25, v15
+; SI-NEXT: v_readfirstlane_b32 s50, v14
+; SI-NEXT: v_readfirstlane_b32 s23, v13
+; SI-NEXT: v_readfirstlane_b32 s21, v11
; SI-NEXT: v_readfirstlane_b32 s18, v1
-; SI-NEXT: v_writelane_b32 v41, s38, 23
+; SI-NEXT: v_writelane_b32 v41, s49, 23
; SI-NEXT: v_writelane_b32 v40, s99, 35
-; SI-NEXT: v_writelane_b32 v41, s39, 24
+; SI-NEXT: v_writelane_b32 v41, s50, 24
; SI-NEXT: s_waitcnt vmcnt(12)
; SI-NEXT: v_readfirstlane_b32 s58, v31
; SI-NEXT: s_waitcnt vmcnt(11)
@@ -243955,165 +243956,166 @@ define inreg <64 x bfloat> @bitcast_v64i16_to_v64bf16_scalar(<64 x i16> inreg %a
; SI-NEXT: s_and_b64 s[4:5], vcc, exec
; SI-NEXT: s_cbranch_scc0 .LBB107_2
; SI-NEXT: ; %bb.1: ; %cmp.false
+; SI-NEXT: s_lshl_b32 s5, s17, 16
+; SI-NEXT: v_writelane_b32 v41, s5, 25
+; SI-NEXT: s_lshl_b32 s5, s61, 16
+; SI-NEXT: v_writelane_b32 v41, s5, 26
+; SI-NEXT: s_lshl_b32 s5, s20, 16
+; SI-NEXT: v_writelane_b32 v41, s5, 27
+; SI-NEXT: s_lshl_b32 s5, s22, 16
+; SI-NEXT: v_writelane_b32 v41, s5, 28
+; SI-NEXT: s_lshl_b32 s5, s24, 16
+; SI-NEXT: v_writelane_b32 v41, s5, 29
+; SI-NEXT: s_lshl_b32 s5, s26, 16
+; SI-NEXT: v_writelane_b32 v41, s5, 30
+; SI-NEXT: s_lshl_b32 s5, s28, 16
+; SI-NEXT: v_writelane_b32 v41, s5, 31
+; SI-NEXT: s_lshl_b32 s5, s18, 16
+; SI-NEXT: v_writelane_b32 v41, s5, 32
+; SI-NEXT: s_lshl_b32 s5, s95, 16
+; SI-NEXT: v_writelane_b32 v41, s5, 33
+; SI-NEXT: s_lshl_b32 s5, s38, 16
; SI-NEXT: s_lshl_b32 s4, s60, 16
-; SI-NEXT: v_writelane_b32 v41, s4, 25
-; SI-NEXT: s_lshl_b32 s4, s63, 16
-; SI-NEXT: v_writelane_b32 v41, s4, 26
-; SI-NEXT: s_lshl_b32 s4, s20, 16
-; SI-NEXT: v_writelane_b32 v41, s4, 27
-; SI-NEXT: s_lshl_b32 s4, s22, 16
-; SI-NEXT: v_writelane_b32 v41, s4, 28
-; SI-NEXT: s_lshl_b32 s4, s24, 16
-; SI-NEXT: v_writelane_b32 v41, s4, 29
-; SI-NEXT: s_lshl_b32 s4, s26, 16
-; SI-NEXT: v_writelane_b32 v41, s4, 30
-; SI-NEXT: s_lshl_b32 s4, s28, 16
-; SI-NEXT: v_writelane_b32 v41, s4, 31
-; SI-NEXT: s_lshl_b32 s4, s18, 16
-; SI-NEXT: v_writelane_b32 v41, s4, 32
-; SI-NEXT: s_lshl_b32 s4, s89, 16
-; SI-NEXT: v_writelane_b32 v41, s4, 33
-; SI-NEXT: s_lshl_b32 s4, s91, 16
-; SI-NEXT: v_writelane_b32 v41, s4, 34
-; SI-NEXT: s_lshl_b32 s4, s35, 16
-; SI-NEXT: v_writelane_b32 v41, s4, 35
-; SI-NEXT: s_lshl_b32 s4, s37, 16
-; SI-NEXT: s_lshl_b32 s7, s17, 16
-; SI-NEXT: s_lshl_b32 s96, s61, 16
-; SI-NEXT: s_lshl_b32 s99, s72, 16
-; SI-NEXT: s_lshl_b32 s97, s74, 16
-; SI-NEXT: s_lshl_b32 s92, s75, 16
-; SI-NEXT: s_lshl_b32 s94, s76, 16
-; SI-NEXT: s_lshl_b32 s95, s93, 16
-; SI-NEXT: s_lshl_b32 s93, s16, 16
-; SI-NEXT: s_lshl_b32 s30, s73, 16
-; SI-NEXT: s_lshl_b32 s31, s90, 16
-; SI-NEXT: s_lshl_b32 s34, s34, 16
-; SI-NEXT: v_writelane_b32 v41, s4, 36
-; SI-NEXT: s_lshl_b32 s35, s36, 16
-; SI-NEXT: s_lshl_b32 s86, s19, 16
-; SI-NEXT: s_lshl_b32 s36, s38, 16
-; SI-NEXT: s_lshl_b32 s22, s21, 16
+; SI-NEXT: v_writelane_b32 v41, s5, 34
+; SI-NEXT: s_lshl_b32 s5, s48, 16
+; SI-NEXT: s_lshl_b32 s16, s19, 16
+; SI-NEXT: s_lshl_b32 s63, s72, 16
+; SI-NEXT: s_lshl_b32 s98, s74, 16
+; SI-NEXT: s_lshl_b32 s62, s76, 16
+; SI-NEXT: s_lshl_b32 s96, s78, 16
+; SI-NEXT: s_lshl_b32 s31, s79, 16
+; SI-NEXT: s_lshl_b32 s34, s6, 16
+; SI-NEXT: s_lshl_b32 s35, s73, 16
+; SI-NEXT: s_lshl_b32 s30, s30, 16
+; SI-NEXT: s_lshl_b32 s36, s36, 16
+; SI-NEXT: s_lshl_b32 s99, s37, 16
+; SI-NEXT: v_writelane_b32 v41, s5, 35
; SI-NEXT: s_lshl_b32 s37, s39, 16
-; SI-NEXT: s_lshl_b32 s24, s23, 16
-; SI-NEXT: s_lshl_b32 s38, s77, 16
-; SI-NEXT: s_lshl_b32 s28, s25, 16
-; SI-NEXT: s_lshl_b32 s39, s78, 16
-; SI-NEXT: s_lshl_b32 s61, s27, 16
-; SI-NEXT: s_lshl_b32 s48, s79, 16
-; SI-NEXT: s_lshl_b32 s89, s29, 16
-; SI-NEXT: s_lshl_b32 s49, s88, 16
-; SI-NEXT: s_lshl_b32 s60, s9, 16
-; SI-NEXT: s_lshl_b32 s50, s8, 16
-; SI-NEXT: s_lshl_b32 s90, s11, 16
-; SI-NEXT: s_lshl_b32 s91, s10, 16
-; SI-NEXT: s_lshl_b32 s70, s13, 16
-; SI-NEXT: s_lshl_b32 s51, s12, 16
-; SI-NEXT: s_lshl_b32 s71, s15, 16
-; SI-NEXT: s_lshl_b32 s52, s14, 16
-; SI-NEXT: s_lshl_b32 s20, s41, 16
-; SI-NEXT: s_lshl_b32 s53, s40, 16
-; SI-NEXT: s_lshl_b32 s81, s43, 16
-; SI-NEXT: s_lshl_b32 s54, s42, 16
-; SI-NEXT: s_lshl_b32 s63, s45, 16
-; SI-NEXT: s_lshl_b32 s55, s44, 16
-; SI-NEXT: s_lshl_b32 s72, s47, 16
-; SI-NEXT: s_lshl_b32 s64, s46, 16
-; SI-NEXT: s_lshl_b32 s82, s57, 16
-; SI-NEXT: s_lshl_b32 s65, s56, 16
-; SI-NEXT: s_lshl_b32 s74, s59, 16
-; SI-NEXT: s_lshl_b32 s66, s58, 16
-; SI-NEXT: s_lshl_b32 s75, s87, 16
-; SI-NEXT: s_mov_b32 s73, s6
-; SI-NEXT: s_lshl_b32 s67, s6, 16
-; SI-NEXT: s_lshl_b32 s76, s83, 16
-; SI-NEXT: s_mov_b32 s16, s68
-; SI-NEXT: s_lshl_b32 s68, s68, 16
-; SI-NEXT: s_lshl_b32 s85, s84, 16
-; SI-NEXT: s_mov_b32 s98, s69
-; SI-NEXT: s_lshl_b32 s69, s69, 16
-; SI-NEXT: s_lshl_b32 s17, s80, 16
-; SI-NEXT: s_mov_b32 s6, s62
-; SI-NEXT: s_lshl_b32 s26, s62, 16
+; SI-NEXT: s_lshl_b32 s19, s21, 16
+; SI-NEXT: s_lshl_b32 s38, s49, 16
+; SI-NEXT: s_lshl_b32 s20, s23, 16
+; SI-NEXT: s_lshl_b32 s39, s50, 16
+; SI-NEXT: s_lshl_b32 s22, s25, 16
+; SI-NEXT: s_lshl_b32 s48, s75, 16
+; SI-NEXT: s_lshl_b32 s60, s27, 16
+; SI-NEXT: s_lshl_b32 s49, s77, 16
+; SI-NEXT: s_lshl_b32 s24, s29, 16
+; SI-NEXT: s_lshl_b32 s50, s88, 16
+; SI-NEXT: s_lshl_b32 s61, s7, 16
+; SI-NEXT: s_lshl_b32 s51, s89, 16
+; SI-NEXT: s_lshl_b32 s28, s9, 16
+; SI-NEXT: s_lshl_b32 s52, s8, 16
+; SI-NEXT: s_lshl_b32 s72, s11, 16
+; SI-NEXT: s_lshl_b32 s53, s10, 16
+; SI-NEXT: s_lshl_b32 s74, s13, 16
+; SI-NEXT: s_lshl_b32 s54, s12, 16
+; SI-NEXT: s_lshl_b32 s95, s15, 16
+; SI-NEXT: s_lshl_b32 s55, s14, 16
+; SI-NEXT: s_lshl_b32 s81, s41, 16
+; SI-NEXT: s_lshl_b32 s64, s40, 16
+; SI-NEXT: s_lshl_b32 s82, s43, 16
+; SI-NEXT: s_lshl_b32 s65, s42, 16
+; SI-NEXT: s_lshl_b32 s85, s45, 16
+; SI-NEXT: s_lshl_b32 s66, s44, 16
+; SI-NEXT: s_lshl_b32 s86, s47, 16
+; SI-NEXT: s_lshl_b32 s67, s46, 16
+; SI-NEXT: s_lshl_b32 s76, s57, 16
+; SI-NEXT: s_lshl_b32 s68, s56, 16
+; SI-NEXT: s_lshl_b32 s97, s59, 16
+; SI-NEXT: s_lshl_b32 s69, s58, 16
+; SI-NEXT: s_lshl_b32 s78, s87, 16
+; SI-NEXT: s_mov_b32 s6, s91
+; SI-NEXT: s_lshl_b32 s70, s91, 16
+; SI-NEXT: s_lshl_b32 s79, s83, 16
+; SI-NEXT: s_mov_b32 s73, s94
+; SI-NEXT: s_lshl_b32 s71, s94, 16
+; SI-NEXT: s_lshl_b32 s26, s84, 16
+; SI-NEXT: s_mov_b32 s91, s90
+; SI-NEXT: s_lshl_b32 s90, s90, 16
+; SI-NEXT: s_mov_b32 s94, s93
+; SI-NEXT: s_lshl_b32 s17, s93, 16
+; SI-NEXT: s_mov_b32 s93, s92
+; SI-NEXT: s_lshl_b32 s80, s92, 16
+; SI-NEXT: s_mov_b32 s92, s4
; SI-NEXT: s_mov_b64 s[4:5], 0
; SI-NEXT: s_branch .LBB107_3
; SI-NEXT: .LBB107_2:
; SI-NEXT: ; implicit-def: $sgpr17
; SI-NEXT: ; kill: killed $sgpr17
-; SI-NEXT: s_mov_b32 s16, s68
+; SI-NEXT: s_mov_b32 s73, s94
; SI-NEXT: ; implicit-def: $sgpr17
; SI-NEXT: ; kill: killed $sgpr17
-; SI-NEXT: s_mov_b32 s73, s6
+; SI-NEXT: s_mov_b32 s6, s91
; SI-NEXT: ; implicit-def: $sgpr17
; SI-NEXT: ; kill: killed $sgpr17
-; SI-NEXT: s_mov_b32 s6, s62
+; SI-NEXT: s_mov_b32 s94, s93
; SI-NEXT: ; implicit-def: $sgpr17
; SI-NEXT: ; kill: killed $sgpr17
-; SI-NEXT: s_mov_b32 s98, s69
+; SI-NEXT: s_mov_b32 s93, s92
+; SI-NEXT: ; implicit-def: $sgpr17
+; SI-NEXT: ; kill: killed $sgpr17
+; SI-NEXT: s_mov_b32 s91, s90
; SI-NEXT: ; implicit-def: $sgpr17
; SI-NEXT: ; kill: killed $sgpr17
; SI-NEXT: s_mov_b64 s[4:5], -1
; SI-NEXT: ; implicit-def: $sgpr17
; SI-NEXT: ; kill: killed $sgpr17
-; SI-NEXT: ; implicit-def: $sgpr7
-; SI-NEXT: ; implicit-def: $sgpr96
-; SI-NEXT: ; implicit-def: $sgpr99
-; SI-NEXT: ; implicit-def: $sgpr97
+; SI-NEXT: ; implicit-def: $sgpr16
; SI-NEXT: ; implicit-def: $sgpr92
-; SI-NEXT: ; implicit-def: $sgpr94
-; SI-NEXT: ; implicit-def: $sgpr95
-; SI-NEXT: ; implicit-def: $sgpr93
-; SI-NEXT: ; implicit-def: $sgpr30
+; SI-NEXT: ; kill: killed $sgpr16
+; SI-NEXT: ; implicit-def: $sgpr16
+; SI-NEXT: ; implicit-def: $sgpr17
+; SI-NEXT: ; kill: killed $sgpr17
+; SI-NEXT: ; implicit-def: $sgpr63
+; SI-NEXT: ; implicit-def: $sgpr98
+; SI-NEXT: ; implicit-def: $sgpr62
+; SI-NEXT: ; implicit-def: $sgpr96
; SI-NEXT: ; implicit-def: $sgpr31
; SI-NEXT: ; implicit-def: $sgpr34
; SI-NEXT: ; implicit-def: $sgpr35
-; SI-NEXT: ; implicit-def: $sgpr86
+; SI-NEXT: ; implicit-def: $sgpr30
; SI-NEXT: ; implicit-def: $sgpr36
-; SI-NEXT: ; implicit-def: $sgpr22
+; SI-NEXT: ; implicit-def: $sgpr99
; SI-NEXT: ; implicit-def: $sgpr37
-; SI-NEXT: ; implicit-def: $sgpr24
+; SI-NEXT: ; implicit-def: $sgpr19
; SI-NEXT: ; implicit-def: $sgpr38
-; SI-NEXT: ; implicit-def: $sgpr28
+; SI-NEXT: ; implicit-def: $sgpr20
; SI-NEXT: ; implicit-def: $sgpr39
-; SI-NEXT: ; implicit-def: $sgpr61
+; SI-NEXT: ; implicit-def: $sgpr22
; SI-NEXT: ; implicit-def: $sgpr48
-; SI-NEXT: ; implicit-def: $sgpr89
-; SI-NEXT: ; implicit-def: $sgpr49
; SI-NEXT: ; implicit-def: $sgpr60
+; SI-NEXT: ; implicit-def: $sgpr49
+; SI-NEXT: ; implicit-def: $sgpr24
; SI-NEXT: ; implicit-def: $sgpr50
-; SI-NEXT: ; implicit-def: $sgpr90
-; SI-NEXT: ; implicit-def: $sgpr91
-; SI-NEXT: ; implicit-def: $sgpr70
+; SI-NEXT: ; implicit-def: $sgpr61
; SI-NEXT: ; implicit-def: $sgpr51
-; SI-NEXT: ; implicit-def: $sgpr71
+; SI-NEXT: ; implicit-def: $sgpr28
; SI-NEXT: ; implicit-def: $sgpr52
-; SI-NEXT: ; implicit-def: $sgpr20
+; SI-NEXT: ; implicit-def: $sgpr72
; SI-NEXT: ; implicit-def: $sgpr53
-; SI-NEXT: ; implicit-def: $sgpr81
+; SI-NEXT: ; implicit-def: $sgpr74
; SI-NEXT: ; implicit-def: $sgpr54
-; SI-NEXT: ; implicit-def: $sgpr63
+; SI-NEXT: ; implicit-def: $sgpr95
; SI-NEXT: ; implicit-def: $sgpr55
-; SI-NEXT: ; implicit-def: $sgpr72
+; SI-NEXT: ; implicit-def: $sgpr81
; SI-NEXT: ; implicit-def: $sgpr64
; SI-NEXT: ; implicit-def: $sgpr82
; SI-NEXT: ; implicit-def: $sgpr65
-; SI-NEXT: ; implicit-def: $sgpr74
+; SI-NEXT: ; implicit-def: $sgpr85
; SI-NEXT: ; implicit-def: $sgpr66
-; SI-NEXT: ; implicit-def: $sgpr75
+; SI-NEXT: ; implicit-def: $sgpr86
; SI-NEXT: ; implicit-def: $sgpr67
; SI-NEXT: ; implicit-def: $sgpr76
; SI-NEXT: ; implicit-def: $sgpr68
-; SI-NEXT: ; implicit-def: $sgpr85
+; SI-NEXT: ; implicit-def: $sgpr97
; SI-NEXT: ; implicit-def: $sgpr69
+; SI-NEXT: ; implicit-def: $sgpr78
+; SI-NEXT: ; implicit-def: $sgpr70
+; SI-NEXT: ; implicit-def: $sgpr79
+; SI-NEXT: ; implicit-def: $sgpr71
; SI-NEXT: ; implicit-def: $sgpr26
-; SI-NEXT: ; implicit-def: $sgpr17
-; SI-NEXT: ; kill: killed $sgpr17
-; SI-NEXT: ; implicit-def: $sgpr17
-; SI-NEXT: ; kill: killed $sgpr17
-; SI-NEXT: ; implicit-def: $sgpr17
-; SI-NEXT: ; kill: killed $sgpr17
-; SI-NEXT: ; implicit-def: $sgpr17
-; SI-NEXT: ; kill: killed $sgpr17
+; SI-NEXT: ; implicit-def: $sgpr90
+; SI-NEXT: ; implicit-def: $sgpr80
; SI-NEXT: ; implicit-def: $sgpr17
; SI-NEXT: ; kill: killed $sgpr17
; SI-NEXT: ; implicit-def: $sgpr17
@@ -244121,86 +244123,97 @@ define inreg <64 x bfloat> @bitcast_v64i16_to_v64bf16_scalar(<64 x i16> inreg %a
; SI-NEXT: ; implicit-def: $sgpr17
; SI-NEXT: .LBB107_3: ; %Flow
; SI-NEXT: s_andn2_b64 vcc, exec, s[4:5]
-; SI-NEXT: s_mov_b32 s5, s17
-; SI-NEXT: s_mov_b32 s17, s86
-; SI-NEXT: s_mov_b32 s86, s7
+; SI-NEXT: s_mov_b32 s4, s17
+; SI-NEXT: s_mov_b32 s17, s30
+; SI-NEXT: s_mov_b32 s30, s99
+; SI-NEXT: s_mov_b32 s99, s16
; SI-NEXT: s_cbranch_vccnz .LBB107_5
; SI-NEXT: ; %bb.4: ; %cmp.true
-; SI-NEXT: s_lshl_b32 s5, s6, 16
+; SI-NEXT: s_add_i32 s4, s94, 3
+; SI-NEXT: s_and_b32 s4, s4, 0xffff
+; SI-NEXT: s_lshl_b32 s5, s93, 16
+; SI-NEXT: s_add_i32 s84, s84, 3
+; SI-NEXT: s_or_b32 s4, s5, s4
+; SI-NEXT: s_and_b32 s5, s84, 0xffff
+; SI-NEXT: s_lshl_b32 s60, s91, 16
+; SI-NEXT: s_add_i32 s83, s83, 3
+; SI-NEXT: s_or_b32 s5, s60, s5
+; SI-NEXT: s_and_b32 s60, s83, 0xffff
+; SI-NEXT: s_lshl_b32 s61, s73, 16
+; SI-NEXT: s_or_b32 s79, s61, s60
+; SI-NEXT: s_lshl_b32 s61, s6, 16
+; SI-NEXT: s_add_i32 s9, s9, 3
; SI-NEXT: v_readlane_b32 s6, v41, 24
+; SI-NEXT: s_add_i32 s11, s11, 3
+; SI-NEXT: s_and_b32 s9, s9, 0xffff
+; SI-NEXT: s_lshl_b32 s8, s8, 16
+; SI-NEXT: s_add_i32 s7, s7, 3
; SI-NEXT: s_lshl_b32 s20, s6, 16
; SI-NEXT: v_readlane_b32 s6, v41, 23
-; SI-NEXT: s_lshl_b32 s17, s6, 16
-; SI-NEXT: v_readlane_b32 s6, v41, 22
-; SI-NEXT: s_lshl_b32 s61, s16, 16
-; SI-NEXT: s_add_i32 s16, s6, 3
-; SI-NEXT: v_readlane_b32 s6, v41, 21
-; SI-NEXT: s_and_b32 s16, s16, 0xffff
-; SI-NEXT: s_lshl_b32 s7, s6, 16
-; SI-NEXT: v_readlane_b32 s6, v41, 20
-; SI-NEXT: s_or_b32 s7, s7, s16
-; SI-NEXT: s_add_i32 s6, s6, 3
-; SI-NEXT: v_readlane_b32 s16, v41, 19
-; SI-NEXT: s_add_i32 s19, s19, 3
-; SI-NEXT: s_and_b32 s6, s6, 0xffff
-; SI-NEXT: s_lshl_b32 s16, s16, 16
-; SI-NEXT: s_and_b32 s19, s19, 0xffff
-; SI-NEXT: s_or_b32 s6, s16, s6
-; SI-NEXT: v_readlane_b32 s16, v41, 18
-; SI-NEXT: s_lshl_b32 s60, s98, 16
-; SI-NEXT: s_or_b32 s17, s17, s19
-; SI-NEXT: s_add_i32 s98, s16, 3
-; SI-NEXT: v_readlane_b32 s19, v41, 17
-; SI-NEXT: s_add_i32 s21, s21, 3
-; SI-NEXT: s_and_b32 s16, s98, 0xffff
-; SI-NEXT: s_lshl_b32 s19, s19, 16
-; SI-NEXT: s_add_i32 s11, s11, 3
-; SI-NEXT: s_add_i32 s9, s9, 3
-; SI-NEXT: s_and_b32 s21, s21, 0xffff
-; SI-NEXT: s_or_b32 s16, s19, s16
-; SI-NEXT: v_readlane_b32 s19, v41, 16
; SI-NEXT: s_add_i32 s13, s13, 3
; SI-NEXT: s_and_b32 s11, s11, 0xffff
; SI-NEXT: s_lshl_b32 s10, s10, 16
-; SI-NEXT: s_and_b32 s9, s9, 0xffff
-; SI-NEXT: s_lshl_b32 s8, s8, 16
+; SI-NEXT: s_or_b32 s8, s8, s9
+; SI-NEXT: s_and_b32 s7, s7, 0xffff
+; SI-NEXT: s_lshl_b32 s9, s89, 16
; SI-NEXT: s_add_i32 s29, s29, 3
-; SI-NEXT: s_or_b32 s20, s20, s21
-; SI-NEXT: s_add_i32 s96, s19, 3
-; SI-NEXT: v_readlane_b32 s21, v41, 15
+; SI-NEXT: s_lshl_b32 s19, s6, 16
+; SI-NEXT: v_readlane_b32 s6, v41, 22
; SI-NEXT: s_add_i32 s15, s15, 3
; SI-NEXT: s_and_b32 s13, s13, 0xffff
; SI-NEXT: s_lshl_b32 s12, s12, 16
; SI-NEXT: s_or_b32 s10, s10, s11
-; SI-NEXT: s_or_b32 s8, s8, s9
+; SI-NEXT: s_or_b32 s7, s9, s7
; SI-NEXT: s_and_b32 s9, s29, 0xffff
; SI-NEXT: s_lshl_b32 s11, s88, 16
; SI-NEXT: s_add_i32 s27, s27, 3
-; SI-NEXT: s_and_b32 s19, s96, 0xffff
-; SI-NEXT: s_lshl_b32 s21, s21, 16
+; SI-NEXT: s_add_i32 s16, s6, 3
+; SI-NEXT: v_readlane_b32 s6, v41, 21
; SI-NEXT: s_and_b32 s15, s15, 0xffff
; SI-NEXT: s_lshl_b32 s14, s14, 16
; SI-NEXT: s_or_b32 s12, s12, s13
; SI-NEXT: s_or_b32 s9, s11, s9
; SI-NEXT: s_and_b32 s11, s27, 0xffff
-; SI-NEXT: s_lshl_b32 s13, s79, 16
+; SI-NEXT: s_lshl_b32 s13, s77, 16
; SI-NEXT: s_add_i32 s25, s25, 3
-; SI-NEXT: s_or_b32 s19, s21, s19
-; SI-NEXT: s_add_i32 s18, s18, 3
-; SI-NEXT: v_readlane_b32 s21, v41, 14
+; SI-NEXT: s_and_b32 s16, s16, 0xffff
+; SI-NEXT: s_lshl_b32 s17, s6, 16
+; SI-NEXT: v_readlane_b32 s6, v41, 20
; SI-NEXT: s_or_b32 s14, s14, s15
; SI-NEXT: s_or_b32 s11, s13, s11
; SI-NEXT: s_and_b32 s13, s25, 0xffff
-; SI-NEXT: s_lshl_b32 s15, s78, 16
+; SI-NEXT: s_lshl_b32 s15, s75, 16
; SI-NEXT: s_add_i32 s23, s23, 3
-; SI-NEXT: s_and_b32 s18, s18, 0xffff
-; SI-NEXT: s_lshl_b32 s21, s21, 16
+; SI-NEXT: s_or_b32 s16, s17, s16
+; SI-NEXT: s_add_i32 s6, s6, 3
+; SI-NEXT: v_readlane_b32 s17, v41, 19
; SI-NEXT: s_or_b32 s13, s15, s13
; SI-NEXT: s_and_b32 s15, s23, 0xffff
-; SI-NEXT: s_lshl_b32 s22, s77, 16
+; SI-NEXT: s_add_i32 s21, s21, 3
+; SI-NEXT: s_and_b32 s6, s6, 0xffff
+; SI-NEXT: s_lshl_b32 s17, s17, 16
+; SI-NEXT: s_or_b32 s15, s20, s15
+; SI-NEXT: s_and_b32 s20, s21, 0xffff
+; SI-NEXT: s_or_b32 s6, s17, s6
+; SI-NEXT: v_readlane_b32 s17, v41, 18
+; SI-NEXT: s_or_b32 s19, s19, s20
+; SI-NEXT: s_add_i32 s98, s17, 3
+; SI-NEXT: v_readlane_b32 s20, v41, 17
+; SI-NEXT: s_and_b32 s17, s98, 0xffff
+; SI-NEXT: s_lshl_b32 s20, s20, 16
+; SI-NEXT: s_or_b32 s17, s20, s17
+; SI-NEXT: v_readlane_b32 s20, v41, 16
+; SI-NEXT: s_add_i32 s96, s20, 3
+; SI-NEXT: v_readlane_b32 s21, v41, 15
+; SI-NEXT: s_and_b32 s20, s96, 0xffff
+; SI-NEXT: s_lshl_b32 s21, s21, 16
+; SI-NEXT: s_or_b32 s20, s21, s20
+; SI-NEXT: s_add_i32 s18, s18, 3
+; SI-NEXT: v_readlane_b32 s21, v41, 14
+; SI-NEXT: s_and_b32 s18, s18, 0xffff
+; SI-NEXT: s_lshl_b32 s21, s21, 16
; SI-NEXT: s_or_b32 s18, s21, s18
; SI-NEXT: v_readlane_b32 s21, v41, 13
-; SI-NEXT: s_or_b32 s15, s22, s15
; SI-NEXT: s_add_i32 s21, s21, 3
; SI-NEXT: v_readlane_b32 s22, v41, 12
; SI-NEXT: s_and_b32 s21, s21, 0xffff
@@ -244244,40 +244257,27 @@ define inreg <64 x bfloat> @bitcast_v64i16_to_v64bf16_scalar(<64 x i16> inreg %a
; SI-NEXT: s_or_b32 s27, s28, s27
; SI-NEXT: s_add_i32 s27, s27, 0x30000
; SI-NEXT: s_add_i32 s26, s26, 0x30000
-; SI-NEXT: s_and_b32 s86, s27, 0xffff0000
-; SI-NEXT: s_lshl_b32 s27, s27, 16
+; SI-NEXT: s_and_b32 s28, s27, 0xffff0000
; SI-NEXT: s_add_i32 s25, s25, 0x30000
-; SI-NEXT: v_writelane_b32 v41, s27, 25
-; SI-NEXT: s_and_b32 s96, s26, 0xffff0000
+; SI-NEXT: v_writelane_b32 v41, s28, 25
+; SI-NEXT: s_and_b32 s99, s26, 0xffff0000
; SI-NEXT: s_lshl_b32 s26, s26, 16
; SI-NEXT: s_add_i32 s24, s24, 0x30000
; SI-NEXT: v_writelane_b32 v41, s26, 26
-; SI-NEXT: s_and_b32 s99, s25, 0xffff0000
+; SI-NEXT: s_and_b32 s63, s25, 0xffff0000
; SI-NEXT: s_lshl_b32 s25, s25, 16
; SI-NEXT: s_add_i32 s23, s23, 0x30000
; SI-NEXT: v_writelane_b32 v41, s25, 27
-; SI-NEXT: s_and_b32 s97, s24, 0xffff0000
+; SI-NEXT: s_and_b32 s98, s24, 0xffff0000
; SI-NEXT: s_lshl_b32 s24, s24, 16
-; SI-NEXT: s_add_i32 s80, s80, 3
; SI-NEXT: s_add_i32 s22, s22, 0x30000
; SI-NEXT: v_writelane_b32 v41, s24, 28
-; SI-NEXT: s_and_b32 s92, s23, 0xffff0000
+; SI-NEXT: s_and_b32 s62, s23, 0xffff0000
; SI-NEXT: s_lshl_b32 s23, s23, 16
-; SI-NEXT: s_and_b32 s4, s80, 0xffff
-; SI-NEXT: s_add_i32 s84, s84, 3
; SI-NEXT: s_add_i32 s21, s21, 0x30000
; SI-NEXT: v_writelane_b32 v41, s23, 29
-; SI-NEXT: s_and_b32 s94, s22, 0xffff0000
+; SI-NEXT: s_and_b32 s96, s22, 0xffff0000
; SI-NEXT: s_lshl_b32 s22, s22, 16
-; SI-NEXT: s_or_b32 s4, s5, s4
-; SI-NEXT: s_and_b32 s5, s84, 0xffff
-; SI-NEXT: s_add_i32 s83, s83, 3
-; SI-NEXT: s_add_i32 s18, s18, 0x30000
-; SI-NEXT: v_writelane_b32 v41, s22, 30
-; SI-NEXT: s_and_b32 s95, s21, 0xffff0000
-; SI-NEXT: s_lshl_b32 s21, s21, 16
-; SI-NEXT: s_or_b32 s5, s60, s5
-; SI-NEXT: s_and_b32 s60, s83, 0xffff
; SI-NEXT: s_add_i32 s87, s87, 3
; SI-NEXT: s_add_i32 s59, s59, 3
; SI-NEXT: s_add_i32 s57, s57, 3
@@ -244285,13 +244285,11 @@ define inreg <64 x bfloat> @bitcast_v64i16_to_v64bf16_scalar(<64 x i16> inreg %a
; SI-NEXT: s_add_i32 s45, s45, 3
; SI-NEXT: s_add_i32 s43, s43, 3
; SI-NEXT: s_add_i32 s41, s41, 3
-; SI-NEXT: s_add_i32 s19, s19, 0x30000
-; SI-NEXT: v_writelane_b32 v41, s21, 31
-; SI-NEXT: s_and_b32 s93, s18, 0xffff0000
-; SI-NEXT: s_lshl_b32 s18, s18, 16
-; SI-NEXT: s_or_b32 s76, s61, s60
+; SI-NEXT: s_add_i32 s18, s18, 0x30000
+; SI-NEXT: v_writelane_b32 v41, s22, 30
+; SI-NEXT: s_and_b32 s31, s21, 0xffff0000
+; SI-NEXT: s_lshl_b32 s21, s21, 16
; SI-NEXT: s_and_b32 s60, s87, 0xffff
-; SI-NEXT: s_lshl_b32 s61, s73, 16
; SI-NEXT: s_and_b32 s59, s59, 0xffff
; SI-NEXT: s_lshl_b32 s58, s58, 16
; SI-NEXT: s_and_b32 s57, s57, 0xffff
@@ -244304,10 +244302,11 @@ define inreg <64 x bfloat> @bitcast_v64i16_to_v64bf16_scalar(<64 x i16> inreg %a
; SI-NEXT: s_lshl_b32 s42, s42, 16
; SI-NEXT: s_and_b32 s41, s41, 0xffff
; SI-NEXT: s_lshl_b32 s40, s40, 16
-; SI-NEXT: s_add_i32 s16, s16, 0x30000
-; SI-NEXT: v_writelane_b32 v41, s18, 32
-; SI-NEXT: s_lshl_b32 s18, s19, 16
-; SI-NEXT: s_or_b32 s75, s61, s60
+; SI-NEXT: s_add_i32 s20, s20, 0x30000
+; SI-NEXT: v_writelane_b32 v41, s21, 31
+; SI-NEXT: s_and_b32 s34, s18, 0xffff0000
+; SI-NEXT: s_lshl_b32 s18, s18, 16
+; SI-NEXT: s_or_b32 s78, s61, s60
; SI-NEXT: s_or_b32 s58, s58, s59
; SI-NEXT: s_or_b32 s56, s56, s57
; SI-NEXT: s_or_b32 s46, s46, s47
@@ -244315,13 +244314,12 @@ define inreg <64 x bfloat> @bitcast_v64i16_to_v64bf16_scalar(<64 x i16> inreg %a
; SI-NEXT: s_or_b32 s42, s42, s43
; SI-NEXT: s_or_b32 s40, s40, s41
; SI-NEXT: s_add_i32 s6, s6, 0x30000
-; SI-NEXT: v_writelane_b32 v41, s18, 33
-; SI-NEXT: s_and_b32 s31, s16, 0xffff0000
-; SI-NEXT: s_lshl_b32 s16, s16, 16
+; SI-NEXT: v_writelane_b32 v41, s18, 32
+; SI-NEXT: s_lshl_b32 s18, s20, 16
; SI-NEXT: s_add_i32 s4, s4, 0x30000
; SI-NEXT: s_add_i32 s5, s5, 0x30000
-; SI-NEXT: s_add_i32 s76, s76, 0x30000
-; SI-NEXT: s_add_i32 s75, s75, 0x30000
+; SI-NEXT: s_add_i32 s79, s79, 0x30000
+; SI-NEXT: s_add_i32 s78, s78, 0x30000
; SI-NEXT: s_add_i32 s58, s58, 0x30000
; SI-NEXT: s_add_i32 s56, s56, 0x30000
; SI-NEXT: s_add_i32 s46, s46, 0x30000
@@ -244332,293 +244330,296 @@ define inreg <64 x bfloat> @bitcast_v64i16_to_v64bf16_scalar(<64 x i16> inreg %a
; SI-NEXT: s_add_i32 s12, s12, 0x30000
; SI-NEXT: s_add_i32 s10, s10, 0x30000
; SI-NEXT: s_add_i32 s8, s8, 0x30000
+; SI-NEXT: s_add_i32 s7, s7, 0x30000
; SI-NEXT: s_add_i32 s9, s9, 0x30000
; SI-NEXT: s_add_i32 s11, s11, 0x30000
; SI-NEXT: s_add_i32 s13, s13, 0x30000
; SI-NEXT: s_add_i32 s15, s15, 0x30000
-; SI-NEXT: s_add_i32 s20, s20, 0x30000
+; SI-NEXT: s_add_i32 s19, s19, 0x30000
+; SI-NEXT: s_add_i32 s16, s16, 0x30000
; SI-NEXT: s_add_i32 s17, s17, 0x30000
-; SI-NEXT: s_add_i32 s7, s7, 0x30000
-; SI-NEXT: v_writelane_b32 v41, s16, 34
-; SI-NEXT: s_and_b32 s34, s6, 0xffff0000
+; SI-NEXT: v_writelane_b32 v41, s18, 33
+; SI-NEXT: s_and_b32 s30, s6, 0xffff0000
; SI-NEXT: s_lshl_b32 s6, s6, 16
-; SI-NEXT: s_and_b32 s30, s19, 0xffff0000
-; SI-NEXT: v_writelane_b32 v41, s6, 35
-; SI-NEXT: s_and_b32 s35, s7, 0xffff0000
-; SI-NEXT: s_lshl_b32 s6, s7, 16
+; SI-NEXT: s_lshl_b32 s92, s27, 16
+; SI-NEXT: s_and_b32 s35, s20, 0xffff0000
; SI-NEXT: s_and_b32 s36, s17, 0xffff0000
; SI-NEXT: s_lshl_b32 s17, s17, 16
-; SI-NEXT: s_and_b32 s37, s20, 0xffff0000
-; SI-NEXT: s_lshl_b32 s22, s20, 16
-; SI-NEXT: s_and_b32 s38, s15, 0xffff0000
-; SI-NEXT: s_lshl_b32 s24, s15, 16
-; SI-NEXT: s_and_b32 s39, s13, 0xffff0000
-; SI-NEXT: s_lshl_b32 s28, s13, 16
-; SI-NEXT: s_and_b32 s48, s11, 0xffff0000
-; SI-NEXT: s_lshl_b32 s61, s11, 16
-; SI-NEXT: s_and_b32 s49, s9, 0xffff0000
-; SI-NEXT: s_lshl_b32 s89, s9, 16
-; SI-NEXT: s_and_b32 s50, s8, 0xffff0000
-; SI-NEXT: s_lshl_b32 s60, s8, 16
-; SI-NEXT: s_and_b32 s91, s10, 0xffff0000
-; SI-NEXT: s_lshl_b32 s90, s10, 16
-; SI-NEXT: s_and_b32 s51, s12, 0xffff0000
-; SI-NEXT: s_lshl_b32 s70, s12, 16
-; SI-NEXT: s_and_b32 s52, s14, 0xffff0000
-; SI-NEXT: s_lshl_b32 s71, s14, 16
-; SI-NEXT: s_and_b32 s53, s40, 0xffff0000
-; SI-NEXT: s_lshl_b32 s20, s40, 16
-; SI-NEXT: s_and_b32 s54, s42, 0xffff0000
-; SI-NEXT: s_lshl_b32 s81, s42, 16
-; SI-NEXT: s_and_b32 s55, s44, 0xffff0000
-; SI-NEXT: s_lshl_b32 s63, s44, 16
-; SI-NEXT: s_and_b32 s64, s46, 0xffff0000
-; SI-NEXT: s_lshl_b32 s72, s46, 16
-; SI-NEXT: s_and_b32 s65, s56, 0xffff0000
-; SI-NEXT: s_lshl_b32 s82, s56, 16
-; SI-NEXT: s_and_b32 s66, s58, 0xffff0000
-; SI-NEXT: s_lshl_b32 s74, s58, 16
-; SI-NEXT: s_and_b32 s67, s75, 0xffff0000
-; SI-NEXT: s_lshl_b32 s75, s75, 16
-; SI-NEXT: s_and_b32 s68, s76, 0xffff0000
-; SI-NEXT: s_lshl_b32 s76, s76, 16
-; SI-NEXT: s_and_b32 s69, s5, 0xffff0000
-; SI-NEXT: s_lshl_b32 s85, s5, 16
-; SI-NEXT: s_and_b32 s26, s4, 0xffff0000
-; SI-NEXT: s_lshl_b32 s5, s4, 16
-; SI-NEXT: v_writelane_b32 v41, s6, 36
+; SI-NEXT: v_writelane_b32 v41, s6, 34
+; SI-NEXT: s_and_b32 s37, s16, 0xffff0000
+; SI-NEXT: s_lshl_b32 s6, s16, 16
+; SI-NEXT: s_and_b32 s38, s19, 0xffff0000
+; SI-NEXT: s_lshl_b32 s19, s19, 16
+; SI-NEXT: s_and_b32 s39, s15, 0xffff0000
+; SI-NEXT: s_lshl_b32 s20, s15, 16
+; SI-NEXT: s_and_b32 s48, s13, 0xffff0000
+; SI-NEXT: s_lshl_b32 s22, s13, 16
+; SI-NEXT: s_and_b32 s49, s11, 0xffff0000
+; SI-NEXT: s_lshl_b32 s60, s11, 16
+; SI-NEXT: s_and_b32 s50, s9, 0xffff0000
+; SI-NEXT: s_lshl_b32 s24, s9, 16
+; SI-NEXT: s_and_b32 s51, s7, 0xffff0000
+; SI-NEXT: s_lshl_b32 s61, s7, 16
+; SI-NEXT: s_and_b32 s52, s8, 0xffff0000
+; SI-NEXT: s_lshl_b32 s28, s8, 16
+; SI-NEXT: s_and_b32 s53, s10, 0xffff0000
+; SI-NEXT: s_lshl_b32 s72, s10, 16
+; SI-NEXT: s_and_b32 s54, s12, 0xffff0000
+; SI-NEXT: s_lshl_b32 s74, s12, 16
+; SI-NEXT: s_and_b32 s55, s14, 0xffff0000
+; SI-NEXT: s_lshl_b32 s95, s14, 16
+; SI-NEXT: s_and_b32 s64, s40, 0xffff0000
+; SI-NEXT: s_lshl_b32 s81, s40, 16
+; SI-NEXT: s_and_b32 s65, s42, 0xffff0000
+; SI-NEXT: s_lshl_b32 s82, s42, 16
+; SI-NEXT: s_and_b32 s66, s44, 0xffff0000
+; SI-NEXT: s_lshl_b32 s85, s44, 16
+; SI-NEXT: s_and_b32 s67, s46, 0xffff0000
+; SI-NEXT: s_lshl_b32 s86, s46, 16
+; SI-NEXT: s_and_b32 s68, s56, 0xffff0000
+; SI-NEXT: s_lshl_b32 s76, s56, 16
+; SI-NEXT: s_and_b32 s69, s58, 0xffff0000
+; SI-NEXT: s_lshl_b32 s97, s58, 16
+; SI-NEXT: s_and_b32 s70, s78, 0xffff0000
+; SI-NEXT: s_lshl_b32 s78, s78, 16
+; SI-NEXT: s_and_b32 s71, s79, 0xffff0000
+; SI-NEXT: s_lshl_b32 s79, s79, 16
+; SI-NEXT: s_and_b32 s90, s5, 0xffff0000
+; SI-NEXT: s_lshl_b32 s26, s5, 16
+; SI-NEXT: s_and_b32 s80, s4, 0xffff0000
+; SI-NEXT: s_lshl_b32 s4, s4, 16
+; SI-NEXT: v_writelane_b32 v41, s6, 35
; SI-NEXT: .LBB107_5: ; %end
-; SI-NEXT: v_mul_f32_e64 v1, 1.0, s86
-; SI-NEXT: v_readlane_b32 s4, v41, 25
+; SI-NEXT: v_readlane_b32 s5, v41, 25
+; SI-NEXT: v_mul_f32_e64 v1, 1.0, s5
; SI-NEXT: v_lshrrev_b32_e32 v1, 16, v1
-; SI-NEXT: v_mul_f32_e64 v2, 1.0, s4
+; SI-NEXT: v_mul_f32_e64 v2, 1.0, s92
; SI-NEXT: v_alignbit_b32 v1, v1, v2, 16
; SI-NEXT: buffer_store_dword v1, v0, s[0:3], 0 offen
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_mul_f32_e64 v1, 1.0, s96
-; SI-NEXT: v_readlane_b32 s4, v41, 26
+; SI-NEXT: v_mul_f32_e64 v1, 1.0, s99
+; SI-NEXT: v_readlane_b32 s5, v41, 26
; SI-NEXT: v_lshrrev_b32_e32 v1, 16, v1
-; SI-NEXT: v_mul_f32_e64 v2, 1.0, s4
+; SI-NEXT: v_mul_f32_e64 v2, 1.0, s5
; SI-NEXT: v_alignbit_b32 v1, v1, v2, 16
; SI-NEXT: v_add_i32_e32 v2, vcc, 4, v0
; SI-NEXT: buffer_store_dword v1, v2, s[0:3], 0 offen
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_mul_f32_e64 v1, 1.0, s99
-; SI-NEXT: v_readlane_b32 s4, v41, 27
+; SI-NEXT: v_mul_f32_e64 v1, 1.0, s63
+; SI-NEXT: v_readlane_b32 s5, v41, 27
; SI-NEXT: v_lshrrev_b32_e32 v1, 16, v1
-; SI-NEXT: v_mul_f32_e64 v2, 1.0, s4
+; SI-NEXT: v_mul_f32_e64 v2, 1.0, s5
; SI-NEXT: v_alignbit_b32 v1, v1, v2, 16
; SI-NEXT: v_add_i32_e32 v2, vcc, 8, v0
; SI-NEXT: buffer_store_dword v1, v2, s[0:3], 0 offen
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_mul_f32_e64 v1, 1.0, s97
-; SI-NEXT: v_readlane_b32 s4, v41, 28
+; SI-NEXT: v_mul_f32_e64 v1, 1.0, s98
+; SI-NEXT: v_readlane_b32 s5, v41, 28
; SI-NEXT: v_lshrrev_b32_e32 v1, 16, v1
-; SI-NEXT: v_mul_f32_e64 v2, 1.0, s4
+; SI-NEXT: v_mul_f32_e64 v2, 1.0, s5
; SI-NEXT: v_alignbit_b32 v1, v1, v2, 16
; SI-NEXT: v_add_i32_e32 v2, vcc, 12, v0
; SI-NEXT: buffer_store_dword v1, v2, s[0:3], 0 offen
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_mul_f32_e64 v1, 1.0, s92
-; SI-NEXT: v_readlane_b32 s4, v41, 29
+; SI-NEXT: v_mul_f32_e64 v1, 1.0, s62
+; SI-NEXT: v_readlane_b32 s5, v41, 29
; SI-NEXT: v_lshrrev_b32_e32 v1, 16, v1
-; SI-NEXT: v_mul_f32_e64 v2, 1.0, s4
+; SI-NEXT: v_mul_f32_e64 v2, 1.0, s5
; SI-NEXT: v_alignbit_b32 v1, v1, v2, 16
; SI-NEXT: v_add_i32_e32 v2, vcc, 16, v0
; SI-NEXT: buffer_store_dword v1, v2, s[0:3], 0 offen
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_mul_f32_e64 v1, 1.0, s94
-; SI-NEXT: v_readlane_b32 s4, v41, 30
+; SI-NEXT: v_mul_f32_e64 v1, 1.0, s96
+; SI-NEXT: v_readlane_b32 s5, v41, 30
; SI-NEXT: v_lshrrev_b32_e32 v1, 16, v1
-; SI-NEXT: v_mul_f32_e64 v2, 1.0, s4
+; SI-NEXT: v_mul_f32_e64 v2, 1.0, s5
; SI-NEXT: v_alignbit_b32 v1, v1, v2, 16
; SI-NEXT: v_add_i32_e32 v2, vcc, 20, v0
; SI-NEXT: buffer_store_dword v1, v2, s[0:3], 0 offen
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_mul_f32_e64 v1, 1.0, s95
-; SI-NEXT: v_readlane_b32 s4, v41, 31
+; SI-NEXT: v_mul_f32_e64 v1, 1.0, s31
+; SI-NEXT: v_readlane_b32 s5, v41, 31
; SI-NEXT: v_lshrrev_b32_e32 v1, 16, v1
-; SI-NEXT: v_mul_f32_e64 v2, 1.0, s4
+; SI-NEXT: v_mul_f32_e64 v2, 1.0, s5
; SI-NEXT: v_alignbit_b32 v1, v1, v2, 16
; SI-NEXT: v_add_i32_e32 v2, vcc, 24, v0
; SI-NEXT: buffer_store_dword v1, v2, s[0:3], 0 offen
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_mul_f32_e64 v1, 1.0, s93
-; SI-NEXT: v_readlane_b32 s4, v41, 32
+; SI-NEXT: v_mul_f32_e64 v1, 1.0, s34
+; SI-NEXT: v_readlane_b32 s5, v41, 32
; SI-NEXT: v_lshrrev_b32_e32 v1, 16, v1
-; SI-NEXT: v_mul_f32_e64 v2, 1.0, s4
+; SI-NEXT: v_mul_f32_e64 v2, 1.0, s5
; SI-NEXT: v_alignbit_b32 v1, v1, v2, 16
; SI-NEXT: v_add_i32_e32 v2, vcc, 28, v0
; SI-NEXT: buffer_store_dword v1, v2, s[0:3], 0 offen
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_mul_f32_e64 v1, 1.0, s30
-; SI-NEXT: v_readlane_b32 s4, v41, 33
+; SI-NEXT: v_mul_f32_e64 v1, 1.0, s35
+; SI-NEXT: v_readlane_b32 s5, v41, 33
; SI-NEXT: v_lshrrev_b32_e32 v1, 16, v1
-; SI-NEXT: v_mul_f32_e64 v2, 1.0, s4
+; SI-NEXT: v_mul_f32_e64 v2, 1.0, s5
; SI-NEXT: v_alignbit_b32 v1, v1, v2, 16
; SI-NEXT: v_add_i32_e32 v2, vcc, 32, v0
; SI-NEXT: buffer_store_dword v1, v2, s[0:3], 0 offen
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_mul_f32_e64 v1, 1.0, s31
-; SI-NEXT: v_readlane_b32 s4, v41, 34
+; SI-NEXT: v_mul_f32_e64 v1, 1.0, s36
; SI-NEXT: v_lshrrev_b32_e32 v1, 16, v1
-; SI-NEXT: v_mul_f32_e64 v2, 1.0, s4
+; SI-NEXT: v_mul_f32_e64 v2, 1.0, s17
; SI-NEXT: v_alignbit_b32 v1, v1, v2, 16
; SI-NEXT: v_add_i32_e32 v2, vcc, 36, v0
; SI-NEXT: buffer_store_dword v1, v2, s[0:3], 0 offen
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_mul_f32_e64 v1, 1.0, s34
-; SI-NEXT: v_readlane_b32 s4, v41, 35
+; SI-NEXT: v_mul_f32_e64 v1, 1.0, s30
+; SI-NEXT: v_readlane_b32 s5, v41, 34
; SI-NEXT: v_lshrrev_b32_e32 v1, 16, v1
-; SI-NEXT: v_mul_f32_e64 v2, 1.0, s4
+; SI-NEXT: v_mul_f32_e64 v2, 1.0, s5
; SI-NEXT: v_alignbit_b32 v1, v1, v2, 16
; SI-NEXT: v_add_i32_e32 v2, vcc, 40, v0
; SI-NEXT: buffer_store_dword v1, v2, s[0:3], 0 offen
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_mul_f32_e64 v1, 1.0, s35
-; SI-NEXT: v_readlane_b32 s4, v41, 36
+; SI-NEXT: v_mul_f32_e64 v1, 1.0, s37
+; SI-NEXT: v_readlane_b32 s5, v41, 35
; SI-NEXT: v_lshrrev_b32_e32 v1, 16, v1
-; SI-NEXT: v_mul_f32_e64 v2, 1.0, s4
+; SI-NEXT: v_mul_f32_e64 v2, 1.0, s5
; SI-NEXT: v_alignbit_b32 v1, v1, v2, 16
; SI-NEXT: v_add_i32_e32 v2, vcc, 44, v0
; SI-NEXT: buffer_store_dword v1, v2, s[0:3], 0 offen
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_mul_f32_e64 v1, 1.0, s36
+; SI-NEXT: v_mul_f32_e64 v1, 1.0, s38
; SI-NEXT: v_lshrrev_b32_e32 v1, 16, v1
-; SI-NEXT: v_mul_f32_e64 v2, 1.0, s17
+; SI-NEXT: v_mul_f32_e64 v2, 1.0, s19
; SI-NEXT: v_alignbit_b32 v1, v1, v2, 16
; SI-NEXT: v_add_i32_e32 v2, vcc, 48, v0
; SI-NEXT: buffer_store_dword v1, v2, s[0:3], 0 offen
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_mul_f32_e64 v1, 1.0, s37
+; SI-NEXT: v_mul_f32_e64 v1, 1.0, s39
; SI-NEXT: v_lshrrev_b32_e32 v1, 16, v1
-; SI-NEXT: v_mul_f32_e64 v2, 1.0, s22
+; SI-NEXT: v_mul_f32_e64 v2, 1.0, s20
; SI-NEXT: v_alignbit_b32 v1, v1, v2, 16
; SI-NEXT: v_add_i32_e32 v2, vcc, 52, v0
; SI-NEXT: buffer_store_dword v1, v2, s[0:3], 0 offen
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_mul_f32_e64 v1, 1.0, s38
+; SI-NEXT: v_mul_f32_e64 v1, 1.0, s48
; SI-NEXT: v_lshrrev_b32_e32 v1, 16, v1
-; SI-NEXT: v_mul_f32_e64 v2, 1.0, s24
+; SI-NEXT: v_mul_f32_e64 v2, 1.0, s22
; SI-NEXT: v_alignbit_b32 v1, v1, v2, 16
; SI-NEXT: v_add_i32_e32 v2, vcc, 56, v0
; SI-NEXT: buffer_store_dword v1, v2, s[0:3], 0 offen
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_mul_f32_e64 v1, 1.0, s39
+; SI-NEXT: v_mul_f32_e64 v1, 1.0, s49
; SI-NEXT: v_lshrrev_b32_e32 v1, 16, v1
-; SI-NEXT: v_mul_f32_e64 v2, 1.0, s28
+; SI-NEXT: v_mul_f32_e64 v2, 1.0, s60
; SI-NEXT: v_alignbit_b32 v1, v1, v2, 16
; SI-NEXT: v_add_i32_e32 v2, vcc, 60, v0
; SI-NEXT: buffer_store_dword v1, v2, s[0:3], 0 offen
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_mul_f32_e64 v1, 1.0, s48
+; SI-NEXT: v_mul_f32_e64 v1, 1.0, s50
; SI-NEXT: v_lshrrev_b32_e32 v1, 16, v1
-; SI-NEXT: v_mul_f32_e64 v2, 1.0, s61
+; SI-NEXT: v_mul_f32_e64 v2, 1.0, s24
; SI-NEXT: v_alignbit_b32 v1, v1, v2, 16
; SI-NEXT: v_add_i32_e32 v2, vcc, 64, v0
; SI-NEXT: buffer_store_dword v1, v2, s[0:3], 0 offen
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_mul_f32_e64 v1, 1.0, s49
+; SI-NEXT: v_mul_f32_e64 v1, 1.0, s51
; SI-NEXT: v_lshrrev_b32_e32 v1, 16, v1
-; SI-NEXT: v_mul_f32_e64 v2, 1.0, s89
+; SI-NEXT: v_mul_f32_e64 v2, 1.0, s61
; SI-NEXT: v_alignbit_b32 v1, v1, v2, 16
; SI-NEXT: v_add_i32_e32 v2, vcc, 0x44, v0
; SI-NEXT: buffer_store_dword v1, v2, s[0:3], 0 offen
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_mul_f32_e64 v1, 1.0, s50
+; SI-NEXT: v_mul_f32_e64 v1, 1.0, s52
; SI-NEXT: v_lshrrev_b32_e32 v1, 16, v1
-; SI-NEXT: v_mul_f32_e64 v2, 1.0, s60
+; SI-NEXT: v_mul_f32_e64 v2, 1.0, s28
; SI-NEXT: v_alignbit_b32 v1, v1, v2, 16
; SI-NEXT: v_add_i32_e32 v2, vcc, 0x48, v0
; SI-NEXT: buffer_store_dword v1, v2, s[0:3], 0 offen
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_mul_f32_e64 v1, 1.0, s91
+; SI-NEXT: v_mul_f32_e64 v1, 1.0, s53
; SI-NEXT: v_lshrrev_b32_e32 v1, 16, v1
-; SI-NEXT: v_mul_f32_e64 v2, 1.0, s90
+; SI-NEXT: v_mul_f32_e64 v2, 1.0, s72
; SI-NEXT: v_alignbit_b32 v1, v1, v2, 16
; SI-NEXT: v_add_i32_e32 v2, vcc, 0x4c, v0
; SI-NEXT: buffer_store_dword v1, v2, s[0:3], 0 offen
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_mul_f32_e64 v1, 1.0, s51
+; SI-NEXT: v_mul_f32_e64 v1, 1.0, s54
; SI-NEXT: v_lshrrev_b32_e32 v1, 16, v1
-; SI-NEXT: v_mul_f32_e64 v2, 1.0, s70
+; SI-NEXT: v_mul_f32_e64 v2, 1.0, s74
; SI-NEXT: v_alignbit_b32 v1, v1, v2, 16
; SI-NEXT: v_add_i32_e32 v2, vcc, 0x50, v0
; SI-NEXT: buffer_store_dword v1, v2, s[0:3], 0 offen
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_mul_f32_e64 v1, 1.0, s52
+; SI-NEXT: v_mul_f32_e64 v1, 1.0, s55
; SI-NEXT: v_lshrrev_b32_e32 v1, 16, v1
-; SI-NEXT: v_mul_f32_e64 v2, 1.0, s71
+; SI-NEXT: v_mul_f32_e64 v2, 1.0, s95
; SI-NEXT: v_alignbit_b32 v1, v1, v2, 16
; SI-NEXT: v_add_i32_e32 v2, vcc, 0x54, v0
; SI-NEXT: buffer_store_dword v1, v2, s[0:3], 0 offen
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_mul_f32_e64 v1, 1.0, s53
+; SI-NEXT: v_mul_f32_e64 v1, 1.0, s64
; SI-NEXT: v_lshrrev_b32_e32 v1, 16, v1
-; SI-NEXT: v_mul_f32_e64 v2, 1.0, s20
+; SI-NEXT: v_mul_f32_e64 v2, 1.0, s81
; SI-NEXT: v_alignbit_b32 v1, v1, v2, 16
; SI-NEXT: v_add_i32_e32 v2, vcc, 0x58, v0
; SI-NEXT: buffer_store_dword v1, v2, s[0:3], 0 offen
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_mul_f32_e64 v1, 1.0, s54
+; SI-NEXT: v_mul_f32_e64 v1, 1.0, s65
; SI-NEXT: v_lshrrev_b32_e32 v1, 16, v1
-; SI-NEXT: v_mul_f32_e64 v2, 1.0, s81
+; SI-NEXT: v_mul_f32_e64 v2, 1.0, s82
; SI-NEXT: v_alignbit_b32 v1, v1, v2, 16
; SI-NEXT: v_add_i32_e32 v2, vcc, 0x5c, v0
; SI-NEXT: buffer_store_dword v1, v2, s[0:3], 0 offen
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_mul_f32_e64 v1, 1.0, s55
+; SI-NEXT: v_mul_f32_e64 v1, 1.0, s66
; SI-NEXT: v_lshrrev_b32_e32 v1, 16, v1
-; SI-NEXT: v_mul_f32_e64 v2, 1.0, s63
+; SI-NEXT: v_mul_f32_e64 v2, 1.0, s85
; SI-NEXT: v_alignbit_b32 v1, v1, v2, 16
; SI-NEXT: v_add_i32_e32 v2, vcc, 0x60, v0
; SI-NEXT: buffer_store_dword v1, v2, s[0:3], 0 offen
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_mul_f32_e64 v1, 1.0, s64
+; SI-NEXT: v_mul_f32_e64 v1, 1.0, s67
; SI-NEXT: v_lshrrev_b32_e32 v1, 16, v1
-; SI-NEXT: v_mul_f32_e64 v2, 1.0, s72
+; SI-NEXT: v_mul_f32_e64 v2, 1.0, s86
; SI-NEXT: v_alignbit_b32 v1, v1, v2, 16
; SI-NEXT: v_add_i32_e32 v2, vcc, 0x64, v0
; SI-NEXT: buffer_store_dword v1, v2, s[0:3], 0 offen
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_mul_f32_e64 v1, 1.0, s65
+; SI-NEXT: v_mul_f32_e64 v1, 1.0, s68
; SI-NEXT: v_lshrrev_b32_e32 v1, 16, v1
-; SI-NEXT: v_mul_f32_e64 v2, 1.0, s82
+; SI-NEXT: v_mul_f32_e64 v2, 1.0, s76
; SI-NEXT: v_alignbit_b32 v1, v1, v2, 16
; SI-NEXT: v_add_i32_e32 v2, vcc, 0x68, v0
; SI-NEXT: buffer_store_dword v1, v2, s[0:3], 0 offen
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_mul_f32_e64 v1, 1.0, s66
+; SI-NEXT: v_mul_f32_e64 v1, 1.0, s69
; SI-NEXT: v_lshrrev_b32_e32 v1, 16, v1
-; SI-NEXT: v_mul_f32_e64 v2, 1.0, s74
+; SI-NEXT: v_mul_f32_e64 v2, 1.0, s97
; SI-NEXT: v_alignbit_b32 v1, v1, v2, 16
; SI-NEXT: v_add_i32_e32 v2, vcc, 0x6c, v0
; SI-NEXT: buffer_store_dword v1, v2, s[0:3], 0 offen
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_mul_f32_e64 v1, 1.0, s67
+; SI-NEXT: v_mul_f32_e64 v1, 1.0, s70
; SI-NEXT: v_lshrrev_b32_e32 v1, 16, v1
-; SI-NEXT: v_mul_f32_e64 v2, 1.0, s75
+; SI-NEXT: v_mul_f32_e64 v2, 1.0, s78
; SI-NEXT: v_alignbit_b32 v1, v1, v2, 16
; SI-NEXT: v_add_i32_e32 v2, vcc, 0x70, v0
; SI-NEXT: buffer_store_dword v1, v2, s[0:3], 0 offen
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_mul_f32_e64 v1, 1.0, s68
+; SI-NEXT: v_mul_f32_e64 v1, 1.0, s71
; SI-NEXT: v_lshrrev_b32_e32 v1, 16, v1
-; SI-NEXT: v_mul_f32_e64 v2, 1.0, s76
+; SI-NEXT: v_mul_f32_e64 v2, 1.0, s79
; SI-NEXT: v_alignbit_b32 v1, v1, v2, 16
; SI-NEXT: v_add_i32_e32 v2, vcc, 0x74, v0
; SI-NEXT: buffer_store_dword v1, v2, s[0:3], 0 offen
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_mul_f32_e64 v1, 1.0, s69
+; SI-NEXT: v_mul_f32_e64 v1, 1.0, s90
; SI-NEXT: v_lshrrev_b32_e32 v1, 16, v1
-; SI-NEXT: v_mul_f32_e64 v2, 1.0, s85
+; SI-NEXT: v_mul_f32_e64 v2, 1.0, s26
; SI-NEXT: v_alignbit_b32 v1, v1, v2, 16
; SI-NEXT: v_add_i32_e32 v2, vcc, 0x78, v0
; SI-NEXT: buffer_store_dword v1, v2, s[0:3], 0 offen
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_mul_f32_e64 v1, 1.0, s26
+; SI-NEXT: v_mul_f32_e64 v1, 1.0, s80
; SI-NEXT: v_lshrrev_b32_e32 v1, 16, v1
-; SI-NEXT: v_mul_f32_e64 v2, 1.0, s5
+; SI-NEXT: v_mul_f32_e64 v2, 1.0, s4
; SI-NEXT: v_alignbit_b32 v1, v1, v2, 16
; SI-NEXT: v_add_i32_e32 v0, vcc, 0x7c, v0
; SI-NEXT: buffer_store_dword v1, v0, s[0:3], 0 offen
@@ -245123,94 +245124,94 @@ define <64 x i16> @bitcast_v64f16_to_v64i16(<64 x half> %a, i32 %b) {
; SI-NEXT: buffer_load_dword v62, off, s[0:3], s32 offset:68
; SI-NEXT: buffer_load_dword v32, off, s[0:3], s32 offset:72
; SI-NEXT: buffer_load_dword v33, off, s[0:3], s32 offset:76
-; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: buffer_load_dword v63, off, s[0:3], s32 offset:80
-; SI-NEXT: buffer_load_dword v36, off, s[0:3], s32 offset:84
-; SI-NEXT: buffer_load_dword v37, off, s[0:3], s32 offset:92
-; SI-NEXT: buffer_load_dword v39, off, s[0:3], s32 offset:88
+; SI-NEXT: buffer_load_dword v36, off, s[0:3], s32 offset:80
+; SI-NEXT: buffer_load_dword v37, off, s[0:3], s32 offset:84
+; SI-NEXT: buffer_load_dword v39, off, s[0:3], s32 offset:92
+; SI-NEXT: buffer_load_dword v48, off, s[0:3], s32 offset:88
; SI-NEXT: v_cvt_f16_f32_e32 v1, v1
-; SI-NEXT: v_cvt_f16_f32_e32 v34, v2
-; SI-NEXT: v_cvt_f16_f32_e32 v2, v9
-; SI-NEXT: v_cvt_f16_f32_e32 v43, v3
-; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:220 ; 4-byte Folded Spill
+; SI-NEXT: s_waitcnt expcnt(0)
+; SI-NEXT: v_cvt_f16_f32_e32 v63, v6
+; SI-NEXT: v_cvt_f16_f32_e32 v6, v22
+; SI-NEXT: v_cvt_f16_f32_e32 v44, v4
+; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:216 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_cvt_f16_f32_e32 v1, v5
-; SI-NEXT: buffer_store_dword v2, off, s[0:3], s32 offset:212 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v6, off, s[0:3], s32 offset:228 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_cvt_f16_f32_e32 v2, v13
+; SI-NEXT: v_cvt_f16_f32_e32 v6, v30
; SI-NEXT: v_cvt_f16_f32_e32 v55, v7
-; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:216 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:212 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_cvt_f16_f32_e32 v1, v6
-; SI-NEXT: v_cvt_f16_f32_e32 v6, v22
-; SI-NEXT: buffer_store_dword v2, off, s[0:3], s32 offset:208 ; 4-byte Folded Spill
+; SI-NEXT: v_cvt_f16_f32_e32 v1, v9
+; SI-NEXT: buffer_store_dword v6, off, s[0:3], s32 offset:224 ; 4-byte Folded Spill
; SI-NEXT: v_cvt_f16_f32_e32 v40, v8
; SI-NEXT: v_cvt_f16_f32_e32 v5, v10
-; SI-NEXT: buffer_store_dword v6, off, s[0:3], s32 offset:232 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:208 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_cvt_f16_f32_e32 v6, v30
-; SI-NEXT: v_cvt_f16_f32_e32 v3, v21
+; SI-NEXT: v_cvt_f16_f32_e32 v1, v13
; SI-NEXT: v_cvt_f16_f32_e32 v53, v12
+; SI-NEXT: v_cvt_f16_f32_e32 v4, v21
+; SI-NEXT: v_cvt_f16_f32_e32 v43, v3
+; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:204 ; 4-byte Folded Spill
+; SI-NEXT: s_waitcnt expcnt(0)
+; SI-NEXT: v_cvt_f16_f32_e32 v1, v18
; SI-NEXT: v_cvt_f16_f32_e32 v9, v14
-; SI-NEXT: buffer_store_dword v6, off, s[0:3], s32 offset:228 ; 4-byte Folded Spill
-; SI-NEXT: v_cvt_f16_f32_e32 v12, v18
+; SI-NEXT: v_cvt_f16_f32_e32 v3, v16
+; SI-NEXT: v_cvt_f16_f32_e32 v34, v2
+; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:232 ; 4-byte Folded Spill
+; SI-NEXT: s_waitcnt expcnt(0)
+; SI-NEXT: v_cvt_f16_f32_e32 v1, v20
+; SI-NEXT: v_cvt_f16_f32_e32 v20, v24
+; SI-NEXT: v_cvt_f16_f32_e32 v52, v11
+; SI-NEXT: v_cvt_f16_f32_e32 v2, v15
+; SI-NEXT: v_cvt_f16_f32_e32 v13, v17
; SI-NEXT: v_cvt_f16_f32_e32 v18, v19
; SI-NEXT: v_cvt_f16_f32_e32 v19, v23
; SI-NEXT: v_cvt_f16_f32_e32 v23, v25
-; SI-NEXT: v_cvt_f16_f32_e32 v44, v4
-; SI-NEXT: v_cvt_f16_f32_e32 v52, v11
-; SI-NEXT: v_cvt_f16_f32_e32 v2, v15
-; SI-NEXT: v_cvt_f16_f32_e32 v48, v16
-; SI-NEXT: v_cvt_f16_f32_e32 v4, v17
-; SI-NEXT: v_cvt_f16_f32_e32 v13, v20
-; SI-NEXT: v_cvt_f16_f32_e32 v20, v24
; SI-NEXT: v_cvt_f16_f32_e32 v26, v26
; SI-NEXT: v_cvt_f16_f32_e32 v28, v28
; SI-NEXT: s_waitcnt vmcnt(14)
; SI-NEXT: v_cmp_ne_u32_e32 vcc, 0, v31
+; SI-NEXT: v_cvt_f16_f32_e32 v24, v35
; SI-NEXT: v_cvt_f16_f32_e32 v31, v27
-; SI-NEXT: v_cvt_f16_f32_e32 v25, v50
; SI-NEXT: v_cvt_f16_f32_e32 v27, v29
-; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_cvt_f16_f32_e32 v6, v42
; SI-NEXT: v_cvt_f16_f32_e32 v21, v47
-; SI-NEXT: v_cvt_f16_f32_e32 v22, v38
-; SI-NEXT: v_cvt_f16_f32_e32 v38, v49
-; SI-NEXT: buffer_store_dword v6, off, s[0:3], s32 offset:224 ; 4-byte Folded Spill
-; SI-NEXT: v_cvt_f16_f32_e32 v47, v54
-; SI-NEXT: v_cvt_f16_f32_e32 v24, v35
-; SI-NEXT: v_cvt_f16_f32_e32 v54, v46
-; SI-NEXT: v_cvt_f16_f32_e32 v42, v56
+; SI-NEXT: v_cvt_f16_f32_e32 v25, v50
; SI-NEXT: v_cvt_f16_f32_e32 v29, v51
+; SI-NEXT: buffer_store_dword v6, off, s[0:3], s32 offset:220 ; 4-byte Folded Spill
+; SI-NEXT: v_cvt_f16_f32_e32 v47, v54
; SI-NEXT: v_cvt_f16_f32_e32 v16, v41
-; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_cvt_f16_f32_e32 v6, v59
+; SI-NEXT: v_cvt_f16_f32_e32 v54, v46
+; SI-NEXT: v_cvt_f16_f32_e32 v22, v38
+; SI-NEXT: v_cvt_f16_f32_e32 v38, v49
+; SI-NEXT: v_cvt_f16_f32_e32 v12, v58
; SI-NEXT: v_cvt_f16_f32_e32 v17, v45
+; SI-NEXT: v_cvt_f16_f32_e32 v42, v56
; SI-NEXT: v_cvt_f16_f32_e32 v11, v57
-; SI-NEXT: v_cvt_f16_f32_e32 v58, v58
-; SI-NEXT: buffer_store_dword v6, off, s[0:3], s32 offset:204 ; 4-byte Folded Spill
-; SI-NEXT: s_waitcnt vmcnt(14) expcnt(0)
-; SI-NEXT: v_cvt_f16_f32_e32 v6, v62
+; SI-NEXT: v_cvt_f16_f32_e32 v49, v59
; SI-NEXT: s_waitcnt vmcnt(13)
-; SI-NEXT: v_cvt_f16_f32_e32 v49, v32
-; SI-NEXT: v_cvt_f16_f32_e32 v51, v60
-; SI-NEXT: s_waitcnt vmcnt(9)
-; SI-NEXT: v_cvt_f16_f32_e32 v50, v37
-; SI-NEXT: buffer_store_dword v6, off, s[0:3], s32 offset:236 ; 4-byte Folded Spill
+; SI-NEXT: v_cvt_f16_f32_e32 v58, v32
+; SI-NEXT: s_waitcnt vmcnt(12)
+; SI-NEXT: v_cvt_f16_f32_e32 v35, v33
+; SI-NEXT: s_waitcnt vmcnt(11)
+; SI-NEXT: v_cvt_f16_f32_e32 v33, v36
+; SI-NEXT: s_waitcnt vmcnt(10)
+; SI-NEXT: v_cvt_f16_f32_e32 v36, v37
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: buffer_load_dword v6, off, s[0:3], s32 offset:96
; SI-NEXT: buffer_load_dword v7, off, s[0:3], s32 offset:100
; SI-NEXT: buffer_load_dword v8, off, s[0:3], s32 offset:104
; SI-NEXT: buffer_load_dword v10, off, s[0:3], s32 offset:108
; SI-NEXT: buffer_load_dword v32, off, s[0:3], s32 offset:112
-; SI-NEXT: buffer_load_dword v46, off, s[0:3], s32 offset:116
-; SI-NEXT: buffer_load_dword v56, off, s[0:3], s32 offset:120
-; SI-NEXT: v_cvt_f16_f32_e32 v61, v61
-; SI-NEXT: v_cvt_f16_f32_e32 v35, v33
-; SI-NEXT: v_cvt_f16_f32_e32 v33, v63
-; SI-NEXT: v_cvt_f16_f32_e32 v36, v36
+; SI-NEXT: buffer_load_dword v37, off, s[0:3], s32 offset:116
+; SI-NEXT: buffer_load_dword v46, off, s[0:3], s32 offset:120
+; SI-NEXT: v_cvt_f16_f32_e32 v51, v60
+; SI-NEXT: v_cvt_f16_f32_e32 v56, v61
+; SI-NEXT: v_cvt_f16_f32_e32 v61, v62
; SI-NEXT: s_waitcnt vmcnt(14)
-; SI-NEXT: v_cvt_f16_f32_e32 v45, v39
+; SI-NEXT: v_cvt_f16_f32_e32 v45, v48
+; SI-NEXT: v_cvt_f16_f32_e32 v50, v39
; SI-NEXT: s_waitcnt vmcnt(6)
; SI-NEXT: v_cvt_f16_f32_e32 v14, v6
; SI-NEXT: s_waitcnt vmcnt(5)
@@ -245219,15 +245220,15 @@ define <64 x i16> @bitcast_v64f16_to_v64i16(<64 x half> %a, i32 %b) {
; SI-NEXT: v_cvt_f16_f32_e32 v41, v8
; SI-NEXT: s_waitcnt vmcnt(3)
; SI-NEXT: v_cvt_f16_f32_e32 v30, v10
+; SI-NEXT: s_waitcnt vmcnt(2)
+; SI-NEXT: v_cvt_f16_f32_e32 v7, v32
+; SI-NEXT: s_waitcnt vmcnt(1)
+; SI-NEXT: v_cvt_f16_f32_e32 v8, v37
; SI-NEXT: buffer_load_dword v6, off, s[0:3], s32 offset:124
; SI-NEXT: buffer_load_dword v10, off, s[0:3], s32 offset:128
; SI-NEXT: buffer_load_dword v37, off, s[0:3], s32 offset:132
-; SI-NEXT: s_waitcnt vmcnt(5)
-; SI-NEXT: v_cvt_f16_f32_e32 v7, v32
-; SI-NEXT: s_waitcnt vmcnt(4)
-; SI-NEXT: v_cvt_f16_f32_e32 v8, v46
; SI-NEXT: s_waitcnt vmcnt(3)
-; SI-NEXT: v_cvt_f16_f32_e32 v46, v56
+; SI-NEXT: v_cvt_f16_f32_e32 v46, v46
; SI-NEXT: s_waitcnt vmcnt(2)
; SI-NEXT: v_cvt_f16_f32_e32 v32, v6
; SI-NEXT: s_waitcnt vmcnt(1)
@@ -245240,17 +245241,18 @@ define <64 x i16> @bitcast_v64f16_to_v64i16(<64 x half> %a, i32 %b) {
; SI-NEXT: v_mov_b32_e32 v59, v29
; SI-NEXT: v_mov_b32_e32 v29, v27
; SI-NEXT: v_mov_b32_e32 v57, v23
-; SI-NEXT: v_mov_b32_e32 v60, v3
-; SI-NEXT: v_mov_b32_e32 v62, v4
-; SI-NEXT: v_mov_b32_e32 v63, v49
-; SI-NEXT: v_mov_b32_e32 v49, v12
+; SI-NEXT: v_mov_b32_e32 v60, v4
+; SI-NEXT: v_mov_b32_e32 v62, v13
; SI-NEXT: s_xor_b64 exec, exec, s[4:5]
; SI-NEXT: s_cbranch_execz .LBB108_2
; SI-NEXT: ; %bb.1: ; %cmp.true
-; SI-NEXT: buffer_load_dword v3, off, s[0:3], s32 offset:236 ; 4-byte Folded Reload
+; SI-NEXT: v_cvt_f32_f16_e32 v4, v61
; SI-NEXT: v_cvt_f32_f16_e32 v10, v10
; SI-NEXT: v_cvt_f32_f16_e32 v6, v6
; SI-NEXT: v_cvt_f32_f16_e32 v8, v8
+; SI-NEXT: v_add_f32_e32 v4, 0x38000000, v4
+; SI-NEXT: v_cvt_f16_f32_e32 v61, v4
+; SI-NEXT: buffer_load_dword v4, off, s[0:3], s32 offset:216 ; 4-byte Folded Reload
; SI-NEXT: v_cvt_f32_f16_e32 v7, v7
; SI-NEXT: v_cvt_f32_f16_e32 v15, v15
; SI-NEXT: v_cvt_f32_f16_e32 v14, v14
@@ -245280,17 +245282,20 @@ define <64 x i16> @bitcast_v64f16_to_v64i16(<64 x half> %a, i32 %b) {
; SI-NEXT: v_or_b32_e32 v14, v14, v37
; SI-NEXT: v_lshlrev_b32_e32 v37, 16, v36
; SI-NEXT: v_or_b32_e32 v33, v33, v37
-; SI-NEXT: v_cvt_f32_f16_e32 v12, v58
+; SI-NEXT: v_cvt_f32_f16_e32 v37, v56
+; SI-NEXT: v_cvt_f32_f16_e32 v12, v12
; SI-NEXT: v_cvt_f32_f16_e32 v11, v11
; SI-NEXT: v_cvt_f32_f16_e32 v17, v17
; SI-NEXT: v_cvt_f32_f16_e32 v16, v16
; SI-NEXT: v_cvt_f32_f16_e32 v22, v22
; SI-NEXT: v_cvt_f32_f16_e32 v21, v21
; SI-NEXT: v_cvt_f32_f16_e32 v25, v25
+; SI-NEXT: v_add_f32_e32 v37, 0x38000000, v37
; SI-NEXT: v_add_f32_e32 v12, 0x38000000, v12
; SI-NEXT: v_cvt_f32_f16_e32 v24, v24
; SI-NEXT: v_cvt_f32_f16_e32 v28, v28
-; SI-NEXT: v_cvt_f16_f32_e32 v58, v12
+; SI-NEXT: v_cvt_f16_f32_e32 v37, v37
+; SI-NEXT: v_cvt_f16_f32_e32 v12, v12
; SI-NEXT: v_add_f32_e32 v11, 0x38000000, v11
; SI-NEXT: v_add_f32_e32 v17, 0x38000000, v17
; SI-NEXT: v_cvt_f32_f16_e32 v31, v31
@@ -245300,111 +245305,106 @@ define <64 x i16> @bitcast_v64f16_to_v64i16(<64 x half> %a, i32 %b) {
; SI-NEXT: v_add_f32_e32 v16, 0x38000000, v16
; SI-NEXT: v_add_f32_e32 v22, 0x38000000, v22
; SI-NEXT: v_cvt_f32_f16_e32 v19, v19
-; SI-NEXT: v_cvt_f32_f16_e32 v13, v13
+; SI-NEXT: v_cvt_f32_f16_e32 v1, v1
; SI-NEXT: v_cvt_f16_f32_e32 v16, v16
; SI-NEXT: v_cvt_f16_f32_e32 v22, v22
; SI-NEXT: v_add_f32_e32 v21, 0x38000000, v21
; SI-NEXT: v_add_f32_e32 v25, 0x38000000, v25
; SI-NEXT: v_cvt_f32_f16_e32 v18, v18
+; SI-NEXT: v_cvt_f32_f16_e32 v3, v3
+; SI-NEXT: v_lshlrev_b32_e32 v39, 16, v61
; SI-NEXT: v_cvt_f16_f32_e32 v21, v21
; SI-NEXT: v_cvt_f16_f32_e32 v25, v25
; SI-NEXT: v_add_f32_e32 v24, 0x38000000, v24
; SI-NEXT: v_add_f32_e32 v28, 0x38000000, v28
+; SI-NEXT: v_cvt_f32_f16_e32 v2, v2
+; SI-NEXT: v_or_b32_e32 v56, v37, v39
+; SI-NEXT: v_lshlrev_b32_e32 v37, 16, v12
; SI-NEXT: v_cvt_f16_f32_e32 v24, v24
; SI-NEXT: v_cvt_f16_f32_e32 v28, v28
; SI-NEXT: v_add_f32_e32 v31, 0x38000000, v31
; SI-NEXT: v_add_f32_e32 v20, 0x38000000, v20
+; SI-NEXT: v_or_b32_e32 v11, v11, v37
+; SI-NEXT: v_lshlrev_b32_e32 v37, 16, v17
; SI-NEXT: v_cvt_f16_f32_e32 v31, v31
; SI-NEXT: v_cvt_f16_f32_e32 v20, v20
; SI-NEXT: v_add_f32_e32 v19, 0x38000000, v19
-; SI-NEXT: v_add_f32_e32 v13, 0x38000000, v13
-; SI-NEXT: v_cvt_f16_f32_e32 v19, v19
-; SI-NEXT: v_cvt_f16_f32_e32 v13, v13
-; SI-NEXT: v_add_f32_e32 v18, 0x38000000, v18
-; SI-NEXT: v_cvt_f16_f32_e32 v18, v18
-; SI-NEXT: v_cvt_f32_f16_e32 v2, v2
-; SI-NEXT: v_cvt_f32_f16_e32 v39, v52
-; SI-NEXT: v_cvt_f32_f16_e32 v34, v34
-; SI-NEXT: v_cvt_f32_f16_e32 v1, v1
-; SI-NEXT: v_add_f32_e32 v2, 0x38000000, v2
-; SI-NEXT: s_waitcnt vmcnt(0)
-; SI-NEXT: v_cvt_f32_f16_e32 v4, v3
-; SI-NEXT: v_cvt_f32_f16_e32 v3, v61
-; SI-NEXT: v_cvt_f16_f32_e32 v2, v2
-; SI-NEXT: v_add_f32_e32 v34, 0x38000000, v34
-; SI-NEXT: v_add_f32_e32 v4, 0x38000000, v4
-; SI-NEXT: v_cvt_f16_f32_e32 v4, v4
-; SI-NEXT: v_add_f32_e32 v3, 0x38000000, v3
-; SI-NEXT: v_cvt_f16_f32_e32 v3, v3
-; SI-NEXT: v_cvt_f16_f32_e32 v34, v34
-; SI-NEXT: v_lshlrev_b32_e32 v37, 16, v4
; SI-NEXT: v_add_f32_e32 v1, 0x38000000, v1
-; SI-NEXT: v_or_b32_e32 v61, v3, v37
-; SI-NEXT: buffer_load_dword v3, off, s[0:3], s32 offset:220 ; 4-byte Folded Reload
-; SI-NEXT: v_lshlrev_b32_e32 v37, 16, v58
-; SI-NEXT: v_or_b32_e32 v11, v11, v37
-; SI-NEXT: v_lshlrev_b32_e32 v37, 16, v17
; SI-NEXT: v_or_b32_e32 v16, v16, v37
; SI-NEXT: v_lshlrev_b32_e32 v37, 16, v22
+; SI-NEXT: v_cvt_f16_f32_e32 v19, v19
+; SI-NEXT: v_cvt_f16_f32_e32 v1, v1
+; SI-NEXT: v_add_f32_e32 v18, 0x38000000, v18
+; SI-NEXT: v_add_f32_e32 v3, 0x38000000, v3
; SI-NEXT: v_or_b32_e32 v21, v21, v37
; SI-NEXT: v_lshlrev_b32_e32 v37, 16, v25
+; SI-NEXT: v_cvt_f16_f32_e32 v18, v18
+; SI-NEXT: v_cvt_f16_f32_e32 v3, v3
+; SI-NEXT: v_add_f32_e32 v2, 0x38000000, v2
; SI-NEXT: v_or_b32_e32 v24, v24, v37
; SI-NEXT: v_lshlrev_b32_e32 v37, 16, v28
+; SI-NEXT: v_cvt_f16_f32_e32 v2, v2
; SI-NEXT: v_or_b32_e32 v31, v31, v37
; SI-NEXT: v_lshlrev_b32_e32 v37, 16, v20
; SI-NEXT: v_or_b32_e32 v19, v19, v37
-; SI-NEXT: v_lshlrev_b32_e32 v37, 16, v13
+; SI-NEXT: v_lshlrev_b32_e32 v37, 16, v1
; SI-NEXT: v_or_b32_e32 v18, v18, v37
-; SI-NEXT: v_cvt_f32_f16_e32 v37, v48
-; SI-NEXT: v_cvt_f16_f32_e32 v1, v1
-; SI-NEXT: v_lshlrev_b32_e32 v34, 16, v34
-; SI-NEXT: v_cvt_f32_f16_e32 v5, v5
-; SI-NEXT: v_add_f32_e32 v37, 0x38000000, v37
-; SI-NEXT: v_cvt_f16_f32_e32 v48, v37
-; SI-NEXT: v_lshlrev_b32_e32 v1, 16, v1
-; SI-NEXT: v_cvt_f32_f16_e32 v9, v9
-; SI-NEXT: v_add_f32_e32 v5, 0x38000000, v5
-; SI-NEXT: v_lshlrev_b32_e32 v37, 16, v48
+; SI-NEXT: v_lshlrev_b32_e32 v37, 16, v3
; SI-NEXT: v_or_b32_e32 v2, v2, v37
; SI-NEXT: v_cvt_f32_f16_e32 v37, v53
-; SI-NEXT: v_cvt_f16_f32_e32 v5, v5
-; SI-NEXT: v_add_f32_e32 v9, 0x38000000, v9
-; SI-NEXT: v_cvt_f16_f32_e32 v9, v9
+; SI-NEXT: v_cvt_f32_f16_e32 v39, v52
+; SI-NEXT: v_cvt_f32_f16_e32 v34, v34
+; SI-NEXT: v_cvt_f32_f16_e32 v5, v5
; SI-NEXT: v_add_f32_e32 v37, 0x38000000, v37
; SI-NEXT: v_cvt_f16_f32_e32 v53, v37
; SI-NEXT: v_add_f32_e32 v37, 0x38000000, v39
; SI-NEXT: v_cvt_f16_f32_e32 v37, v37
-; SI-NEXT: v_lshlrev_b32_e32 v5, 16, v5
+; SI-NEXT: v_add_f32_e32 v34, 0x38000000, v34
; SI-NEXT: v_lshlrev_b32_e32 v39, 16, v53
-; SI-NEXT: v_lshlrev_b32_e32 v9, 16, v9
+; SI-NEXT: v_cvt_f16_f32_e32 v34, v34
; SI-NEXT: v_or_b32_e32 v52, v37, v39
; SI-NEXT: v_cvt_f32_f16_e32 v37, v40
; SI-NEXT: v_cvt_f32_f16_e32 v39, v55
-; SI-NEXT: v_cvt_f32_f16_e32 v56, v62
-; SI-NEXT: v_cvt_f32_f16_e32 v26, v26
+; SI-NEXT: v_lshlrev_b32_e32 v34, 16, v34
+; SI-NEXT: v_add_f32_e32 v5, 0x38000000, v5
; SI-NEXT: v_add_f32_e32 v37, 0x38000000, v37
; SI-NEXT: v_cvt_f16_f32_e32 v40, v37
; SI-NEXT: v_add_f32_e32 v37, 0x38000000, v39
; SI-NEXT: v_cvt_f16_f32_e32 v37, v37
-; SI-NEXT: v_add_f32_e32 v56, 0x38000000, v56
+; SI-NEXT: v_cvt_f16_f32_e32 v5, v5
; SI-NEXT: v_lshlrev_b32_e32 v39, 16, v40
-; SI-NEXT: v_cvt_f16_f32_e32 v56, v56
+; SI-NEXT: v_cvt_f32_f16_e32 v9, v9
; SI-NEXT: v_or_b32_e32 v55, v37, v39
; SI-NEXT: v_cvt_f32_f16_e32 v37, v44
; SI-NEXT: v_cvt_f32_f16_e32 v39, v43
-; SI-NEXT: v_cvt_f32_f16_e32 v57, v57
-; SI-NEXT: v_add_f32_e32 v26, 0x38000000, v26
+; SI-NEXT: v_lshlrev_b32_e32 v5, 16, v5
+; SI-NEXT: v_add_f32_e32 v9, 0x38000000, v9
; SI-NEXT: v_add_f32_e32 v37, 0x38000000, v37
; SI-NEXT: v_cvt_f16_f32_e32 v44, v37
; SI-NEXT: v_add_f32_e32 v37, 0x38000000, v39
; SI-NEXT: v_cvt_f16_f32_e32 v37, v37
-; SI-NEXT: v_cvt_f16_f32_e32 v26, v26
+; SI-NEXT: v_cvt_f16_f32_e32 v9, v9
; SI-NEXT: v_lshlrev_b32_e32 v39, 16, v44
-; SI-NEXT: v_cvt_f32_f16_e32 v51, v51
+; SI-NEXT: v_cvt_f32_f16_e32 v26, v26
; SI-NEXT: v_or_b32_e32 v43, v37, v39
-; SI-NEXT: v_lshlrev_b32_e32 v26, 16, v26
+; SI-NEXT: s_waitcnt vmcnt(0)
+; SI-NEXT: v_cvt_f32_f16_e32 v37, v4
+; SI-NEXT: buffer_load_dword v4, off, s[0:3], s32 offset:212 ; 4-byte Folded Reload
+; SI-NEXT: v_cvt_f32_f16_e32 v39, v63
+; SI-NEXT: v_lshlrev_b32_e32 v9, 16, v9
+; SI-NEXT: v_add_f32_e32 v37, 0x38000000, v37
+; SI-NEXT: v_cvt_f16_f32_e32 v37, v37
+; SI-NEXT: v_add_f32_e32 v39, 0x38000000, v39
+; SI-NEXT: v_cvt_f16_f32_e32 v39, v39
+; SI-NEXT: v_cvt_f32_f16_e32 v57, v57
+; SI-NEXT: v_add_f32_e32 v26, 0x38000000, v26
+; SI-NEXT: v_cvt_f16_f32_e32 v26, v26
; SI-NEXT: v_cvt_f32_f16_e32 v47, v47
; SI-NEXT: v_cvt_f32_f16_e32 v42, v42
+; SI-NEXT: v_cvt_f32_f16_e32 v51, v51
+; SI-NEXT: v_lshlrev_b32_e32 v26, 16, v26
+; SI-NEXT: v_add_f32_e32 v47, 0x38000000, v47
+; SI-NEXT: v_cvt_f16_f32_e32 v47, v47
; SI-NEXT: v_cvt_f32_f16_e32 v35, v35
; SI-NEXT: v_cvt_f32_f16_e32 v50, v50
; SI-NEXT: v_cvt_f32_f16_e32 v30, v30
@@ -245412,27 +245412,22 @@ define <64 x i16> @bitcast_v64f16_to_v64i16(<64 x half> %a, i32 %b) {
; SI-NEXT: v_cvt_f32_f16_e32 v29, v29
; SI-NEXT: v_cvt_f32_f16_e32 v38, v38
; SI-NEXT: v_cvt_f32_f16_e32 v54, v54
-; SI-NEXT: v_add_f32_e32 v51, 0x38000000, v51
+; SI-NEXT: v_cvt_f32_f16_e32 v49, v49
; SI-NEXT: v_cvt_f32_f16_e32 v45, v45
; SI-NEXT: v_cvt_f32_f16_e32 v41, v41
; SI-NEXT: v_cvt_f32_f16_e32 v46, v46
-; SI-NEXT: v_cvt_f16_f32_e32 v51, v51
-; SI-NEXT: v_add_f32_e32 v47, 0x38000000, v47
; SI-NEXT: v_add_f32_e32 v42, 0x38000000, v42
+; SI-NEXT: v_add_f32_e32 v51, 0x38000000, v51
; SI-NEXT: v_add_f32_e32 v35, 0x38000000, v35
-; SI-NEXT: s_waitcnt vmcnt(0)
-; SI-NEXT: v_cvt_f32_f16_e32 v37, v3
-; SI-NEXT: buffer_load_dword v3, off, s[0:3], s32 offset:216 ; 4-byte Folded Reload
; SI-NEXT: v_add_f32_e32 v50, 0x38000000, v50
; SI-NEXT: v_add_f32_e32 v30, 0x38000000, v30
-; SI-NEXT: v_add_f32_e32 v37, 0x38000000, v37
-; SI-NEXT: v_cvt_f16_f32_e32 v37, v37
; SI-NEXT: v_add_f32_e32 v32, 0x38000000, v32
; SI-NEXT: v_add_f32_e32 v29, 0x38000000, v29
-; SI-NEXT: v_cvt_f16_f32_e32 v47, v47
; SI-NEXT: v_add_f32_e32 v38, 0x38000000, v38
; SI-NEXT: v_cvt_f16_f32_e32 v42, v42
; SI-NEXT: v_add_f32_e32 v54, 0x38000000, v54
+; SI-NEXT: v_cvt_f16_f32_e32 v51, v51
+; SI-NEXT: v_add_f32_e32 v49, 0x38000000, v49
; SI-NEXT: v_cvt_f16_f32_e32 v35, v35
; SI-NEXT: v_cvt_f16_f32_e32 v50, v50
; SI-NEXT: v_add_f32_e32 v45, 0x38000000, v45
@@ -245443,125 +245438,131 @@ define <64 x i16> @bitcast_v64f16_to_v64i16(<64 x half> %a, i32 %b) {
; SI-NEXT: v_cvt_f16_f32_e32 v29, v29
; SI-NEXT: v_cvt_f16_f32_e32 v38, v38
; SI-NEXT: v_cvt_f16_f32_e32 v54, v54
-; SI-NEXT: v_lshlrev_b32_e32 v51, 16, v51
+; SI-NEXT: v_cvt_f16_f32_e32 v49, v49
; SI-NEXT: v_cvt_f16_f32_e32 v45, v45
; SI-NEXT: v_cvt_f16_f32_e32 v41, v41
; SI-NEXT: v_cvt_f16_f32_e32 v46, v46
-; SI-NEXT: v_lshlrev_b32_e32 v47, 16, v47
; SI-NEXT: v_lshlrev_b32_e32 v42, 16, v42
+; SI-NEXT: v_lshlrev_b32_e32 v51, 16, v51
; SI-NEXT: v_lshlrev_b32_e32 v35, 16, v35
; SI-NEXT: v_lshlrev_b32_e32 v50, 16, v50
; SI-NEXT: v_lshlrev_b32_e32 v30, 16, v30
; SI-NEXT: v_lshlrev_b32_e32 v32, 16, v32
-; SI-NEXT: v_or_b32_e32 v38, v38, v47
; SI-NEXT: v_or_b32_e32 v54, v54, v42
+; SI-NEXT: v_or_b32_e32 v49, v49, v51
; SI-NEXT: v_or_b32_e32 v45, v45, v50
; SI-NEXT: v_or_b32_e32 v41, v41, v30
; SI-NEXT: v_or_b32_e32 v46, v46, v32
-; SI-NEXT: v_alignbit_b32 v47, v16, v47, 16
; SI-NEXT: v_alignbit_b32 v42, v11, v42, 16
+; SI-NEXT: v_alignbit_b32 v51, v56, v51, 16
; SI-NEXT: v_alignbit_b32 v50, v14, v50, 16
; SI-NEXT: v_alignbit_b32 v30, v7, v30, 16
; SI-NEXT: v_alignbit_b32 v32, v6, v32, 16
-; SI-NEXT: buffer_store_dword v4, off, s[0:3], s32 offset:236 ; 4-byte Folded Spill
-; SI-NEXT: s_waitcnt vmcnt(1)
-; SI-NEXT: v_cvt_f32_f16_e32 v39, v3
-; SI-NEXT: v_or_b32_e32 v3, v37, v34
-; SI-NEXT: buffer_store_dword v3, off, s[0:3], s32 offset:220 ; 4-byte Folded Spill
+; SI-NEXT: s_waitcnt vmcnt(0)
+; SI-NEXT: v_cvt_f32_f16_e32 v48, v4
+; SI-NEXT: v_or_b32_e32 v4, v37, v34
+; SI-NEXT: v_lshlrev_b32_e32 v37, 16, v39
+; SI-NEXT: buffer_store_dword v4, off, s[0:3], s32 offset:216 ; 4-byte Folded Spill
+; SI-NEXT: v_add_f32_e32 v48, 0x38000000, v48
+; SI-NEXT: v_cvt_f16_f32_e32 v48, v48
; SI-NEXT: v_alignbit_b32 v34, v43, v34, 16
-; SI-NEXT: v_add_f32_e32 v39, 0x38000000, v39
-; SI-NEXT: v_cvt_f16_f32_e32 v39, v39
+; SI-NEXT: v_alignbit_b32 v63, v55, v37, 16
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_or_b32_e32 v3, v39, v1
-; SI-NEXT: buffer_store_dword v3, off, s[0:3], s32 offset:216 ; 4-byte Folded Spill
+; SI-NEXT: v_or_b32_e32 v4, v48, v37
+; SI-NEXT: buffer_store_dword v4, off, s[0:3], s32 offset:212 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: buffer_load_dword v3, off, s[0:3], s32 offset:212 ; 4-byte Folded Reload
-; SI-NEXT: v_alignbit_b32 v1, v55, v1, 16
-; SI-NEXT: s_waitcnt vmcnt(0)
-; SI-NEXT: v_cvt_f32_f16_e32 v37, v3
-; SI-NEXT: buffer_load_dword v3, off, s[0:3], s32 offset:208 ; 4-byte Folded Reload
-; SI-NEXT: v_add_f32_e32 v37, 0x38000000, v37
-; SI-NEXT: v_cvt_f16_f32_e32 v37, v37
+; SI-NEXT: buffer_load_dword v4, off, s[0:3], s32 offset:208 ; 4-byte Folded Reload
; SI-NEXT: s_waitcnt vmcnt(0)
-; SI-NEXT: v_cvt_f32_f16_e32 v39, v3
-; SI-NEXT: v_or_b32_e32 v3, v37, v5
-; SI-NEXT: buffer_store_dword v3, off, s[0:3], s32 offset:212 ; 4-byte Folded Spill
-; SI-NEXT: v_cvt_f32_f16_e32 v37, v49
+; SI-NEXT: v_cvt_f32_f16_e32 v39, v4
+; SI-NEXT: buffer_load_dword v4, off, s[0:3], s32 offset:204 ; 4-byte Folded Reload
; SI-NEXT: v_add_f32_e32 v39, 0x38000000, v39
; SI-NEXT: v_cvt_f16_f32_e32 v39, v39
-; SI-NEXT: v_alignbit_b32 v5, v52, v5, 16
-; SI-NEXT: v_add_f32_e32 v37, 0x38000000, v37
-; SI-NEXT: v_cvt_f16_f32_e32 v37, v37
+; SI-NEXT: s_waitcnt vmcnt(0)
+; SI-NEXT: v_cvt_f32_f16_e32 v48, v4
+; SI-NEXT: v_or_b32_e32 v4, v39, v5
+; SI-NEXT: buffer_store_dword v4, off, s[0:3], s32 offset:208 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_or_b32_e32 v3, v39, v9
-; SI-NEXT: buffer_store_dword v3, off, s[0:3], s32 offset:208 ; 4-byte Folded Spill
+; SI-NEXT: buffer_load_dword v4, off, s[0:3], s32 offset:232 ; 4-byte Folded Reload
+; SI-NEXT: v_add_f32_e32 v39, 0x38000000, v48
+; SI-NEXT: v_cvt_f16_f32_e32 v39, v39
+; SI-NEXT: v_cvt_f32_f16_e32 v48, v62
+; SI-NEXT: v_alignbit_b32 v5, v52, v5, 16
+; SI-NEXT: v_add_f32_e32 v48, 0x38000000, v48
+; SI-NEXT: v_cvt_f16_f32_e32 v48, v48
+; SI-NEXT: s_waitcnt vmcnt(0)
+; SI-NEXT: v_cvt_f32_f16_e32 v13, v4
+; SI-NEXT: v_or_b32_e32 v4, v39, v9
+; SI-NEXT: buffer_store_dword v4, off, s[0:3], s32 offset:204 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: buffer_load_dword v3, off, s[0:3], s32 offset:232 ; 4-byte Folded Reload
-; SI-NEXT: v_lshlrev_b32_e32 v37, 16, v37
-; SI-NEXT: v_or_b32_e32 v62, v56, v37
-; SI-NEXT: v_cvt_f32_f16_e32 v56, v60
+; SI-NEXT: buffer_load_dword v4, off, s[0:3], s32 offset:228 ; 4-byte Folded Reload
+; SI-NEXT: v_add_f32_e32 v13, 0x38000000, v13
+; SI-NEXT: v_cvt_f16_f32_e32 v13, v13
; SI-NEXT: v_alignbit_b32 v9, v2, v9, 16
-; SI-NEXT: v_add_f32_e32 v56, 0x38000000, v56
-; SI-NEXT: v_cvt_f16_f32_e32 v56, v56
+; SI-NEXT: v_lshlrev_b32_e32 v13, 16, v13
+; SI-NEXT: v_or_b32_e32 v62, v48, v13
+; SI-NEXT: v_cvt_f32_f16_e32 v48, v60
+; SI-NEXT: v_add_f32_e32 v48, 0x38000000, v48
+; SI-NEXT: v_cvt_f16_f32_e32 v48, v48
; SI-NEXT: s_waitcnt vmcnt(0)
-; SI-NEXT: v_cvt_f32_f16_e32 v39, v3
-; SI-NEXT: buffer_load_dword v3, off, s[0:3], s32 offset:228 ; 4-byte Folded Reload
+; SI-NEXT: v_cvt_f32_f16_e32 v39, v4
+; SI-NEXT: buffer_load_dword v4, off, s[0:3], s32 offset:224 ; 4-byte Folded Reload
; SI-NEXT: v_add_f32_e32 v39, 0x38000000, v39
; SI-NEXT: v_cvt_f16_f32_e32 v39, v39
; SI-NEXT: v_lshlrev_b32_e32 v39, 16, v39
-; SI-NEXT: v_or_b32_e32 v60, v56, v39
-; SI-NEXT: v_add_f32_e32 v56, 0x38000000, v57
-; SI-NEXT: v_cvt_f16_f32_e32 v56, v56
-; SI-NEXT: v_or_b32_e32 v57, v56, v26
-; SI-NEXT: v_cvt_f32_f16_e32 v56, v59
+; SI-NEXT: v_or_b32_e32 v60, v48, v39
+; SI-NEXT: v_add_f32_e32 v48, 0x38000000, v57
+; SI-NEXT: v_cvt_f16_f32_e32 v48, v48
+; SI-NEXT: v_or_b32_e32 v57, v48, v26
+; SI-NEXT: v_cvt_f32_f16_e32 v48, v59
; SI-NEXT: v_alignbit_b32 v26, v31, v26, 16
-; SI-NEXT: v_add_f32_e32 v56, 0x38000000, v56
-; SI-NEXT: v_cvt_f16_f32_e32 v56, v56
+; SI-NEXT: v_add_f32_e32 v48, 0x38000000, v48
+; SI-NEXT: v_cvt_f16_f32_e32 v48, v48
; SI-NEXT: s_waitcnt vmcnt(0)
-; SI-NEXT: v_cvt_f32_f16_e32 v23, v3
-; SI-NEXT: buffer_load_dword v3, off, s[0:3], s32 offset:224 ; 4-byte Folded Reload
+; SI-NEXT: v_cvt_f32_f16_e32 v23, v4
+; SI-NEXT: buffer_load_dword v4, off, s[0:3], s32 offset:220 ; 4-byte Folded Reload
; SI-NEXT: v_add_f32_e32 v23, 0x38000000, v23
; SI-NEXT: v_cvt_f16_f32_e32 v23, v23
; SI-NEXT: v_lshlrev_b32_e32 v23, 16, v23
; SI-NEXT: v_or_b32_e32 v29, v29, v23
; SI-NEXT: s_waitcnt vmcnt(0)
-; SI-NEXT: v_cvt_f32_f16_e32 v27, v3
-; SI-NEXT: buffer_load_dword v3, off, s[0:3], s32 offset:204 ; 4-byte Folded Reload
+; SI-NEXT: v_cvt_f32_f16_e32 v27, v4
+; SI-NEXT: v_alignbit_b32 v4, v18, v13, 16
+; SI-NEXT: buffer_store_dword v4, off, s[0:3], s32 offset:232 ; 4-byte Folded Spill
+; SI-NEXT: s_waitcnt expcnt(0)
+; SI-NEXT: v_alignbit_b32 v4, v19, v39, 16
; SI-NEXT: v_add_f32_e32 v27, 0x38000000, v27
; SI-NEXT: v_cvt_f16_f32_e32 v27, v27
-; SI-NEXT: v_lshlrev_b32_e32 v27, 16, v27
-; SI-NEXT: v_or_b32_e32 v59, v56, v27
-; SI-NEXT: v_cvt_f32_f16_e32 v56, v63
-; SI-NEXT: v_add_f32_e32 v56, 0x38000000, v56
-; SI-NEXT: v_cvt_f16_f32_e32 v56, v56
-; SI-NEXT: v_or_b32_e32 v63, v56, v35
-; SI-NEXT: v_alignbit_b32 v35, v33, v35, 16
-; SI-NEXT: s_waitcnt vmcnt(0)
-; SI-NEXT: v_cvt_f32_f16_e32 v49, v3
-; SI-NEXT: v_add_f32_e32 v49, 0x38000000, v49
-; SI-NEXT: v_cvt_f16_f32_e32 v49, v49
-; SI-NEXT: v_or_b32_e32 v3, v49, v51
-; SI-NEXT: buffer_store_dword v3, off, s[0:3], s32 offset:204 ; 4-byte Folded Spill
-; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_alignbit_b32 v3, v19, v39, 16
-; SI-NEXT: buffer_store_dword v3, off, s[0:3], s32 offset:232 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v4, off, s[0:3], s32 offset:228 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_alignbit_b32 v3, v24, v23, 16
-; SI-NEXT: v_alignbit_b32 v49, v18, v37, 16
-; SI-NEXT: buffer_store_dword v3, off, s[0:3], s32 offset:228 ; 4-byte Folded Spill
+; SI-NEXT: v_alignbit_b32 v4, v24, v23, 16
+; SI-NEXT: buffer_store_dword v4, off, s[0:3], s32 offset:224 ; 4-byte Folded Spill
+; SI-NEXT: v_lshlrev_b32_e32 v27, 16, v27
+; SI-NEXT: v_or_b32_e32 v59, v48, v27
+; SI-NEXT: v_lshlrev_b32_e32 v48, 16, v47
+; SI-NEXT: v_cvt_f32_f16_e32 v47, v58
+; SI-NEXT: v_or_b32_e32 v38, v38, v48
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_alignbit_b32 v3, v21, v27, 16
-; SI-NEXT: v_alignbit_b32 v51, v61, v51, 16
-; SI-NEXT: buffer_store_dword v3, off, s[0:3], s32 offset:224 ; 4-byte Folded Spill
+; SI-NEXT: v_alignbit_b32 v4, v21, v27, 16
+; SI-NEXT: buffer_store_dword v4, off, s[0:3], s32 offset:220 ; 4-byte Folded Spill
+; SI-NEXT: v_add_f32_e32 v47, 0x38000000, v47
+; SI-NEXT: v_cvt_f16_f32_e32 v47, v47
+; SI-NEXT: v_or_b32_e32 v58, v47, v35
+; SI-NEXT: v_alignbit_b32 v47, v16, v48, 16
+; SI-NEXT: v_alignbit_b32 v35, v33, v35, 16
; SI-NEXT: .LBB108_2: ; %end
; SI-NEXT: s_or_b64 exec, exec, s[4:5]
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: buffer_load_dword v3, off, s[0:3], s32 offset:220 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v4, off, s[0:3], s32 offset:216 ; 4-byte Folded Reload
; SI-NEXT: v_lshlrev_b32_e32 v34, 16, v34
-; SI-NEXT: v_lshlrev_b32_e32 v1, 16, v1
; SI-NEXT: v_lshlrev_b32_e32 v5, 16, v5
+; SI-NEXT: v_lshlrev_b32_e32 v9, 16, v9
+; SI-NEXT: v_and_b32_e32 v2, 0xffff, v2
+; SI-NEXT: v_lshlrev_b32_e32 v3, 16, v3
+; SI-NEXT: v_or_b32_e32 v2, v2, v3
+; SI-NEXT: v_add_i32_e32 v3, vcc, 28, v0
+; SI-NEXT: v_lshlrev_b32_e32 v1, 16, v1
; SI-NEXT: s_waitcnt vmcnt(0)
-; SI-NEXT: v_and_b32_e32 v37, 0xffff, v3
+; SI-NEXT: v_and_b32_e32 v37, 0xffff, v4
; SI-NEXT: v_or_b32_e32 v34, v37, v34
; SI-NEXT: buffer_store_dword v34, v0, s[0:3], 0 offen
; SI-NEXT: s_waitcnt expcnt(0)
@@ -245570,58 +245571,52 @@ define <64 x i16> @bitcast_v64f16_to_v64i16(<64 x half> %a, i32 %b) {
; SI-NEXT: v_or_b32_e32 v34, v34, v37
; SI-NEXT: v_add_i32_e32 v37, vcc, 4, v0
; SI-NEXT: buffer_store_dword v34, v37, s[0:3], 0 offen
-; SI-NEXT: buffer_load_dword v3, off, s[0:3], s32 offset:216 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v4, off, s[0:3], s32 offset:212 ; 4-byte Folded Reload
+; SI-NEXT: v_lshlrev_b32_e32 v37, 16, v63
; SI-NEXT: s_waitcnt vmcnt(0) expcnt(0)
-; SI-NEXT: v_and_b32_e32 v34, 0xffff, v3
-; SI-NEXT: v_or_b32_e32 v1, v34, v1
-; SI-NEXT: v_add_i32_e32 v34, vcc, 8, v0
-; SI-NEXT: buffer_store_dword v1, v34, s[0:3], 0 offen
-; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_and_b32_e32 v1, 0xffff, v55
-; SI-NEXT: v_lshlrev_b32_e32 v34, 16, v40
-; SI-NEXT: v_or_b32_e32 v1, v1, v34
-; SI-NEXT: v_add_i32_e32 v34, vcc, 12, v0
-; SI-NEXT: buffer_store_dword v1, v34, s[0:3], 0 offen
-; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:212 ; 4-byte Folded Reload
-; SI-NEXT: s_waitcnt vmcnt(0)
-; SI-NEXT: v_and_b32_e32 v1, 0xffff, v1
-; SI-NEXT: v_or_b32_e32 v1, v1, v5
-; SI-NEXT: v_add_i32_e32 v5, vcc, 16, v0
-; SI-NEXT: buffer_store_dword v1, v5, s[0:3], 0 offen
+; SI-NEXT: v_and_b32_e32 v34, 0xffff, v4
+; SI-NEXT: v_or_b32_e32 v34, v34, v37
+; SI-NEXT: v_add_i32_e32 v37, vcc, 8, v0
+; SI-NEXT: buffer_store_dword v34, v37, s[0:3], 0 offen
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_and_b32_e32 v1, 0xffff, v52
-; SI-NEXT: v_lshlrev_b32_e32 v5, 16, v53
-; SI-NEXT: v_or_b32_e32 v1, v1, v5
-; SI-NEXT: v_add_i32_e32 v5, vcc, 20, v0
-; SI-NEXT: buffer_store_dword v1, v5, s[0:3], 0 offen
+; SI-NEXT: v_and_b32_e32 v34, 0xffff, v55
+; SI-NEXT: v_lshlrev_b32_e32 v37, 16, v40
+; SI-NEXT: v_or_b32_e32 v34, v34, v37
+; SI-NEXT: v_add_i32_e32 v37, vcc, 12, v0
+; SI-NEXT: buffer_store_dword v34, v37, s[0:3], 0 offen
+; SI-NEXT: buffer_load_dword v4, off, s[0:3], s32 offset:208 ; 4-byte Folded Reload
+; SI-NEXT: s_waitcnt vmcnt(0) expcnt(0)
+; SI-NEXT: v_and_b32_e32 v34, 0xffff, v4
+; SI-NEXT: v_or_b32_e32 v5, v34, v5
+; SI-NEXT: v_add_i32_e32 v34, vcc, 16, v0
+; SI-NEXT: buffer_store_dword v5, v34, s[0:3], 0 offen
+; SI-NEXT: s_waitcnt expcnt(0)
+; SI-NEXT: v_and_b32_e32 v5, 0xffff, v52
+; SI-NEXT: v_lshlrev_b32_e32 v34, 16, v53
+; SI-NEXT: v_or_b32_e32 v5, v5, v34
+; SI-NEXT: v_add_i32_e32 v34, vcc, 20, v0
+; SI-NEXT: buffer_store_dword v5, v34, s[0:3], 0 offen
+; SI-NEXT: buffer_load_dword v4, off, s[0:3], s32 offset:204 ; 4-byte Folded Reload
+; SI-NEXT: s_waitcnt vmcnt(0) expcnt(0)
+; SI-NEXT: v_and_b32_e32 v5, 0xffff, v4
+; SI-NEXT: v_or_b32_e32 v5, v5, v9
+; SI-NEXT: v_add_i32_e32 v9, vcc, 24, v0
+; SI-NEXT: buffer_store_dword v5, v9, s[0:3], 0 offen
+; SI-NEXT: buffer_store_dword v2, v3, s[0:3], 0 offen
+; SI-NEXT: buffer_load_dword v3, off, s[0:3], s32 offset:232 ; 4-byte Folded Reload
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:208 ; 4-byte Folded Reload
-; SI-NEXT: v_lshlrev_b32_e32 v5, 16, v9
+; SI-NEXT: v_and_b32_e32 v2, 0xffff, v62
; SI-NEXT: s_waitcnt vmcnt(0)
-; SI-NEXT: v_and_b32_e32 v1, 0xffff, v1
-; SI-NEXT: v_or_b32_e32 v1, v1, v5
-; SI-NEXT: v_add_i32_e32 v5, vcc, 24, v0
-; SI-NEXT: buffer_store_dword v1, v5, s[0:3], 0 offen
-; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_and_b32_e32 v1, 0xffff, v2
-; SI-NEXT: v_lshlrev_b32_e32 v2, 16, v48
-; SI-NEXT: v_or_b32_e32 v1, v1, v2
-; SI-NEXT: v_add_i32_e32 v2, vcc, 28, v0
-; SI-NEXT: buffer_store_dword v1, v2, s[0:3], 0 offen
-; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_and_b32_e32 v1, 0xffff, v62
-; SI-NEXT: v_lshlrev_b32_e32 v2, 16, v49
-; SI-NEXT: v_or_b32_e32 v1, v1, v2
-; SI-NEXT: v_add_i32_e32 v2, vcc, 32, v0
-; SI-NEXT: buffer_store_dword v1, v2, s[0:3], 0 offen
+; SI-NEXT: v_lshlrev_b32_e32 v3, 16, v3
+; SI-NEXT: v_or_b32_e32 v2, v2, v3
+; SI-NEXT: v_add_i32_e32 v3, vcc, 32, v0
+; SI-NEXT: buffer_store_dword v2, v3, s[0:3], 0 offen
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_and_b32_e32 v1, 0xffff, v18
-; SI-NEXT: v_lshlrev_b32_e32 v2, 16, v13
-; SI-NEXT: v_or_b32_e32 v1, v1, v2
+; SI-NEXT: v_and_b32_e32 v2, 0xffff, v18
+; SI-NEXT: v_or_b32_e32 v1, v2, v1
; SI-NEXT: v_add_i32_e32 v2, vcc, 36, v0
; SI-NEXT: buffer_store_dword v1, v2, s[0:3], 0 offen
-; SI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:232 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:228 ; 4-byte Folded Reload
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_and_b32_e32 v1, 0xffff, v60
; SI-NEXT: s_waitcnt vmcnt(0)
@@ -245647,7 +245642,7 @@ define <64 x i16> @bitcast_v64f16_to_v64i16(<64 x half> %a, i32 %b) {
; SI-NEXT: v_or_b32_e32 v1, v1, v2
; SI-NEXT: v_add_i32_e32 v2, vcc, 52, v0
; SI-NEXT: buffer_store_dword v1, v2, s[0:3], 0 offen
-; SI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:228 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:224 ; 4-byte Folded Reload
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_and_b32_e32 v1, 0xffff, v29
; SI-NEXT: s_waitcnt vmcnt(0)
@@ -245661,7 +245656,7 @@ define <64 x i16> @bitcast_v64f16_to_v64i16(<64 x half> %a, i32 %b) {
; SI-NEXT: v_or_b32_e32 v1, v1, v2
; SI-NEXT: v_add_i32_e32 v2, vcc, 60, v0
; SI-NEXT: buffer_store_dword v1, v2, s[0:3], 0 offen
-; SI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:224 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:220 ; 4-byte Folded Reload
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_and_b32_e32 v1, 0xffff, v59
; SI-NEXT: s_waitcnt vmcnt(0)
@@ -245695,28 +245690,24 @@ define <64 x i16> @bitcast_v64f16_to_v64i16(<64 x half> %a, i32 %b) {
; SI-NEXT: buffer_store_dword v1, v2, s[0:3], 0 offen
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_and_b32_e32 v1, 0xffff, v11
-; SI-NEXT: v_lshlrev_b32_e32 v2, 16, v58
+; SI-NEXT: v_lshlrev_b32_e32 v2, 16, v12
; SI-NEXT: v_or_b32_e32 v1, v1, v2
; SI-NEXT: v_add_i32_e32 v2, vcc, 0x54, v0
; SI-NEXT: buffer_store_dword v1, v2, s[0:3], 0 offen
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:204 ; 4-byte Folded Reload
+; SI-NEXT: v_and_b32_e32 v1, 0xffff, v49
; SI-NEXT: v_lshlrev_b32_e32 v2, 16, v51
-; SI-NEXT: s_waitcnt vmcnt(0)
-; SI-NEXT: v_and_b32_e32 v1, 0xffff, v1
; SI-NEXT: v_or_b32_e32 v1, v1, v2
; SI-NEXT: v_add_i32_e32 v2, vcc, 0x58, v0
; SI-NEXT: buffer_store_dword v1, v2, s[0:3], 0 offen
-; SI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:236 ; 4-byte Folded Reload
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_and_b32_e32 v1, 0xffff, v61
-; SI-NEXT: s_waitcnt vmcnt(0)
-; SI-NEXT: v_lshlrev_b32_e32 v2, 16, v2
+; SI-NEXT: v_and_b32_e32 v1, 0xffff, v56
+; SI-NEXT: v_lshlrev_b32_e32 v2, 16, v61
; SI-NEXT: v_or_b32_e32 v1, v1, v2
; SI-NEXT: v_add_i32_e32 v2, vcc, 0x5c, v0
; SI-NEXT: buffer_store_dword v1, v2, s[0:3], 0 offen
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_and_b32_e32 v1, 0xffff, v63
+; SI-NEXT: v_and_b32_e32 v1, 0xffff, v58
; SI-NEXT: v_lshlrev_b32_e32 v2, 16, v35
; SI-NEXT: v_or_b32_e32 v1, v1, v2
; SI-NEXT: v_add_i32_e32 v2, vcc, 0x60, v0
@@ -246071,29 +246062,27 @@ define inreg <64 x i16> @bitcast_v64f16_to_v64i16_scalar(<64 x half> inreg %a, i
; SI-NEXT: v_cvt_f16_f32_e32 v49, v5
; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:172 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_cvt_f16_f32_e32 v1, v7
+; SI-NEXT: v_cvt_f16_f32_e32 v1, v11
; SI-NEXT: v_cvt_f16_f32_e32 v50, v6
-; SI-NEXT: v_cvt_f16_f32_e32 v7, v8
-; SI-NEXT: v_cvt_f16_f32_e32 v38, v9
+; SI-NEXT: v_cvt_f16_f32_e32 v7, v7
+; SI-NEXT: v_cvt_f16_f32_e32 v43, v8
; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:168 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_cvt_f16_f32_e32 v1, v11
+; SI-NEXT: v_cvt_f16_f32_e32 v1, v19
+; SI-NEXT: v_cvt_f16_f32_e32 v38, v9
; SI-NEXT: v_cvt_f16_f32_e32 v39, v10
; SI-NEXT: v_cvt_f16_f32_e32 v10, v12
-; SI-NEXT: v_cvt_f16_f32_e32 v35, v13
; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:164 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_cvt_f16_f32_e32 v1, v19
+; SI-NEXT: v_cvt_f16_f32_e32 v1, v27
+; SI-NEXT: v_cvt_f16_f32_e32 v35, v13
; SI-NEXT: v_cvt_f16_f32_e32 v36, v14
; SI-NEXT: v_cvt_f16_f32_e32 v14, v15
-; SI-NEXT: v_cvt_f16_f32_e32 v13, v16
; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:160 ; 4-byte Folded Spill
-; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_cvt_f16_f32_e32 v1, v27
+; SI-NEXT: v_cvt_f16_f32_e32 v13, v16
; SI-NEXT: v_cvt_f16_f32_e32 v32, v17
; SI-NEXT: v_cvt_f16_f32_e32 v33, v18
; SI-NEXT: v_cvt_f16_f32_e32 v20, v20
-; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:156 ; 4-byte Folded Spill
; SI-NEXT: v_cvt_f16_f32_e32 v22, v22
; SI-NEXT: v_cvt_f16_f32_e32 v23, v23
; SI-NEXT: v_cvt_f16_f32_e32 v17, v24
@@ -246112,19 +246101,22 @@ define inreg <64 x i16> @bitcast_v64f16_to_v64i16_scalar(<64 x half> inreg %a, i
; SI-NEXT: v_cvt_f16_f32_e32 v31, v21
; SI-NEXT: v_cvt_f16_f32_e32 v27, v46
; SI-NEXT: v_cvt_f16_f32_e32 v11, v47
-; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:152 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:156 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_cvt_f16_f32_e32 v1, v57
; SI-NEXT: v_cvt_f16_f32_e32 v12, v56
; SI-NEXT: v_cvt_f16_f32_e32 v26, v58
; SI-NEXT: v_cvt_f16_f32_e32 v8, v59
-; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:148 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:152 ; 4-byte Folded Spill
+; SI-NEXT: s_waitcnt expcnt(0)
+; SI-NEXT: v_cvt_f16_f32_e32 v1, v61
; SI-NEXT: v_cvt_f16_f32_e32 v9, v60
-; SI-NEXT: v_cvt_f16_f32_e32 v43, v61
; SI-NEXT: v_cvt_f16_f32_e32 v24, v62
-; SI-NEXT: v_cvt_f16_f32_e32 v5, v63
; SI-NEXT: s_waitcnt vmcnt(14)
+; SI-NEXT: v_cvt_f16_f32_e32 v5, v63
+; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:148 ; 4-byte Folded Spill
; SI-NEXT: v_cvt_f16_f32_e32 v6, v34
+; SI-NEXT: s_waitcnt vmcnt(14)
; SI-NEXT: v_cvt_f16_f32_e32 v34, v37
; SI-NEXT: s_waitcnt vmcnt(13)
; SI-NEXT: v_cvt_f16_f32_e32 v21, v48
@@ -246304,9 +246296,9 @@ define inreg <64 x i16> @bitcast_v64f16_to_v64i16_scalar(<64 x half> inreg %a, i
; SI-NEXT: v_add_f32_e32 v51, 0x38000000, v51
; SI-NEXT: v_cvt_f16_f32_e32 v48, v48
; SI-NEXT: v_cvt_f16_f32_e32 v51, v51
-; SI-NEXT: v_cvt_f32_f16_e32 v40, v45
; SI-NEXT: s_waitcnt vmcnt(1)
; SI-NEXT: v_cvt_f32_f16_e32 v41, v41
+; SI-NEXT: v_cvt_f32_f16_e32 v10, v10
; SI-NEXT: v_or_b32_e32 v60, v48, v4
; SI-NEXT: v_cvt_f32_f16_e32 v48, v62
; SI-NEXT: v_lshlrev_b32_e32 v51, 16, v51
@@ -246314,153 +246306,153 @@ define inreg <64 x i16> @bitcast_v64f16_to_v64i16_scalar(<64 x half> inreg %a, i
; SI-NEXT: v_cvt_f32_f16_e32 v54, v56
; SI-NEXT: v_add_f32_e32 v48, 0x38000000, v48
; SI-NEXT: v_cvt_f16_f32_e32 v48, v48
-; SI-NEXT: v_add_f32_e32 v40, 0x38000000, v40
+; SI-NEXT: v_add_f32_e32 v41, 0x38000000, v41
; SI-NEXT: v_add_f32_e32 v54, 0x38000000, v54
; SI-NEXT: v_cvt_f16_f32_e32 v54, v54
; SI-NEXT: v_lshlrev_b32_e32 v48, 16, v48
-; SI-NEXT: v_add_f32_e32 v41, 0x38000000, v41
-; SI-NEXT: v_cvt_f16_f32_e32 v40, v40
+; SI-NEXT: v_cvt_f16_f32_e32 v41, v41
+; SI-NEXT: v_add_f32_e32 v10, 0x38000000, v10
; SI-NEXT: v_or_b32_e32 v56, v54, v48
; SI-NEXT: v_lshlrev_b32_e32 v54, 16, v55
; SI-NEXT: v_cvt_f32_f16_e32 v55, v44
-; SI-NEXT: v_cvt_f16_f32_e32 v41, v41
-; SI-NEXT: v_or_b32_e32 v45, v40, v54
-; SI-NEXT: v_cvt_f32_f16_e32 v40, v7
-; SI-NEXT: v_add_f32_e32 v55, 0x38000000, v55
-; SI-NEXT: v_cvt_f16_f32_e32 v55, v55
-; SI-NEXT: v_cvt_f32_f16_e32 v10, v10
-; SI-NEXT: v_add_f32_e32 v40, 0x38000000, v40
-; SI-NEXT: v_cvt_f16_f32_e32 v40, v40
-; SI-NEXT: v_lshlrev_b32_e32 v55, 16, v55
-; SI-NEXT: v_or_b32_e32 v7, v41, v55
-; SI-NEXT: buffer_store_dword v7, off, s[0:3], s32 offset:172 ; 4-byte Folded Spill
-; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: buffer_load_dword v7, off, s[0:3], s32 offset:168 ; 4-byte Folded Reload
-; SI-NEXT: buffer_load_dword v41, off, s[0:3], s32 offset:164 ; 4-byte Folded Reload
-; SI-NEXT: v_add_f32_e32 v10, 0x38000000, v10
; SI-NEXT: v_cvt_f16_f32_e32 v10, v10
-; SI-NEXT: v_lshlrev_b32_e32 v40, 16, v40
; SI-NEXT: v_cvt_f32_f16_e32 v20, v20
; SI-NEXT: v_cvt_f32_f16_e32 v28, v28
+; SI-NEXT: v_add_f32_e32 v55, 0x38000000, v55
+; SI-NEXT: v_cvt_f16_f32_e32 v55, v55
; SI-NEXT: v_lshlrev_b32_e32 v10, 16, v10
-; SI-NEXT: v_cvt_f32_f16_e32 v27, v27
; SI-NEXT: v_add_f32_e32 v20, 0x38000000, v20
; SI-NEXT: v_cvt_f16_f32_e32 v20, v20
+; SI-NEXT: v_lshlrev_b32_e32 v55, 16, v55
+; SI-NEXT: v_or_b32_e32 v41, v41, v55
+; SI-NEXT: buffer_store_dword v41, off, s[0:3], s32 offset:172 ; 4-byte Folded Spill
+; SI-NEXT: s_waitcnt expcnt(0)
+; SI-NEXT: buffer_load_dword v41, off, s[0:3], s32 offset:168 ; 4-byte Folded Reload
+; SI-NEXT: v_lshlrev_b32_e32 v20, 16, v20
; SI-NEXT: v_add_f32_e32 v28, 0x38000000, v28
; SI-NEXT: v_cvt_f16_f32_e32 v28, v28
-; SI-NEXT: v_add_f32_e32 v27, 0x38000000, v27
-; SI-NEXT: v_lshlrev_b32_e32 v20, 16, v20
+; SI-NEXT: v_cvt_f32_f16_e32 v27, v27
; SI-NEXT: v_cvt_f32_f16_e32 v26, v26
+; SI-NEXT: v_cvt_f32_f16_e32 v24, v24
; SI-NEXT: v_lshlrev_b32_e32 v28, 16, v28
+; SI-NEXT: v_add_f32_e32 v27, 0x38000000, v27
; SI-NEXT: v_cvt_f16_f32_e32 v27, v27
-; SI-NEXT: v_cvt_f32_f16_e32 v37, v37
; SI-NEXT: v_add_f32_e32 v26, 0x38000000, v26
; SI-NEXT: v_cvt_f16_f32_e32 v26, v26
+; SI-NEXT: v_add_f32_e32 v24, 0x38000000, v24
; SI-NEXT: v_lshlrev_b32_e32 v27, 16, v27
+; SI-NEXT: v_cvt_f16_f32_e32 v24, v24
+; SI-NEXT: v_lshlrev_b32_e32 v26, 16, v26
+; SI-NEXT: v_cvt_f32_f16_e32 v37, v37
+; SI-NEXT: v_alignbit_b32 v63, v46, v51, 16
+; SI-NEXT: v_lshlrev_b32_e32 v24, 16, v24
+; SI-NEXT: v_cvt_f32_f16_e32 v40, v45
+; SI-NEXT: v_add_f32_e32 v37, 0x38000000, v37
+; SI-NEXT: v_cvt_f16_f32_e32 v37, v37
; SI-NEXT: v_cvt_f32_f16_e32 v13, v13
+; SI-NEXT: v_add_f32_e32 v40, 0x38000000, v40
+; SI-NEXT: v_cvt_f16_f32_e32 v40, v40
+; SI-NEXT: v_lshlrev_b32_e32 v37, 16, v37
; SI-NEXT: v_cvt_f32_f16_e32 v17, v17
-; SI-NEXT: v_lshlrev_b32_e32 v26, 16, v26
-; SI-NEXT: v_cvt_f32_f16_e32 v24, v24
; SI-NEXT: v_cvt_f32_f16_e32 v21, v21
+; SI-NEXT: v_or_b32_e32 v45, v40, v54
+; SI-NEXT: v_cvt_f32_f16_e32 v40, v43
+; SI-NEXT: v_cvt_f32_f16_e32 v7, v7
; SI-NEXT: v_cvt_f32_f16_e32 v14, v14
; SI-NEXT: v_cvt_f32_f16_e32 v23, v23
; SI-NEXT: v_cvt_f32_f16_e32 v34, v34
-; SI-NEXT: v_add_f32_e32 v37, 0x38000000, v37
+; SI-NEXT: v_add_f32_e32 v40, 0x38000000, v40
; SI-NEXT: v_add_f32_e32 v13, 0x38000000, v13
; SI-NEXT: v_add_f32_e32 v17, 0x38000000, v17
-; SI-NEXT: v_add_f32_e32 v24, 0x38000000, v24
; SI-NEXT: v_add_f32_e32 v21, 0x38000000, v21
-; SI-NEXT: v_cvt_f16_f32_e32 v37, v37
+; SI-NEXT: v_cvt_f16_f32_e32 v40, v40
+; SI-NEXT: v_add_f32_e32 v7, 0x38000000, v7
; SI-NEXT: v_cvt_f16_f32_e32 v13, v13
; SI-NEXT: v_add_f32_e32 v14, 0x38000000, v14
; SI-NEXT: v_cvt_f16_f32_e32 v17, v17
; SI-NEXT: v_add_f32_e32 v23, 0x38000000, v23
-; SI-NEXT: v_cvt_f16_f32_e32 v24, v24
; SI-NEXT: v_cvt_f16_f32_e32 v21, v21
; SI-NEXT: v_add_f32_e32 v34, 0x38000000, v34
+; SI-NEXT: v_cvt_f16_f32_e32 v7, v7
; SI-NEXT: v_cvt_f16_f32_e32 v14, v14
; SI-NEXT: v_cvt_f16_f32_e32 v23, v23
; SI-NEXT: v_cvt_f16_f32_e32 v34, v34
-; SI-NEXT: v_lshlrev_b32_e32 v37, 16, v37
+; SI-NEXT: v_lshlrev_b32_e32 v40, 16, v40
; SI-NEXT: v_lshlrev_b32_e32 v13, 16, v13
; SI-NEXT: v_lshlrev_b32_e32 v17, 16, v17
-; SI-NEXT: v_lshlrev_b32_e32 v24, 16, v24
; SI-NEXT: v_lshlrev_b32_e32 v21, 16, v21
+; SI-NEXT: v_or_b32_e32 v7, v7, v40
; SI-NEXT: v_or_b32_e32 v14, v14, v13
; SI-NEXT: v_or_b32_e32 v23, v23, v17
; SI-NEXT: v_or_b32_e32 v34, v34, v21
; SI-NEXT: v_alignbit_b32 v4, v57, v4, 16
-; SI-NEXT: v_alignbit_b32 v63, v46, v51, 16
; SI-NEXT: v_alignbit_b32 v62, v29, v48, 16
; SI-NEXT: v_alignbit_b32 v61, v52, v54, 16
; SI-NEXT: v_alignbit_b32 v44, v49, v55, 16
+; SI-NEXT: v_alignbit_b32 v43, v38, v40, 16
; SI-NEXT: v_alignbit_b32 v13, v32, v13, 16
; SI-NEXT: v_alignbit_b32 v17, v18, v17, 16
; SI-NEXT: v_alignbit_b32 v21, v2, v21, 16
+; SI-NEXT: buffer_load_dword v42, off, s[0:3], s32 offset:152 ; 4-byte Folded Reload
; SI-NEXT: s_waitcnt vmcnt(1)
-; SI-NEXT: v_cvt_f32_f16_e32 v7, v7
-; SI-NEXT: s_waitcnt vmcnt(0)
; SI-NEXT: v_cvt_f32_f16_e32 v41, v41
-; SI-NEXT: v_add_f32_e32 v7, 0x38000000, v7
-; SI-NEXT: v_cvt_f16_f32_e32 v7, v7
; SI-NEXT: v_add_f32_e32 v41, 0x38000000, v41
; SI-NEXT: v_cvt_f16_f32_e32 v41, v41
-; SI-NEXT: v_or_b32_e32 v7, v7, v40
-; SI-NEXT: buffer_store_dword v7, off, s[0:3], s32 offset:168 ; 4-byte Folded Spill
-; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_or_b32_e32 v7, v41, v10
-; SI-NEXT: buffer_store_dword v7, off, s[0:3], s32 offset:164 ; 4-byte Folded Spill
+; SI-NEXT: v_or_b32_e32 v41, v41, v10
+; SI-NEXT: buffer_store_dword v41, off, s[0:3], s32 offset:168 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: buffer_load_dword v7, off, s[0:3], s32 offset:160 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v41, off, s[0:3], s32 offset:164 ; 4-byte Folded Reload
; SI-NEXT: v_alignbit_b32 v10, v35, v10, 16
+; SI-NEXT: s_waitcnt vmcnt(2)
+; SI-NEXT: v_cvt_f32_f16_e32 v42, v42
; SI-NEXT: s_waitcnt vmcnt(0)
-; SI-NEXT: v_cvt_f32_f16_e32 v41, v7
+; SI-NEXT: v_cvt_f32_f16_e32 v41, v41
; SI-NEXT: v_add_f32_e32 v41, 0x38000000, v41
; SI-NEXT: v_cvt_f16_f32_e32 v41, v41
-; SI-NEXT: v_or_b32_e32 v7, v41, v20
-; SI-NEXT: buffer_store_dword v7, off, s[0:3], s32 offset:160 ; 4-byte Folded Spill
+; SI-NEXT: v_or_b32_e32 v41, v41, v20
+; SI-NEXT: buffer_store_dword v41, off, s[0:3], s32 offset:164 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: buffer_load_dword v7, off, s[0:3], s32 offset:156 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v41, off, s[0:3], s32 offset:160 ; 4-byte Folded Reload
; SI-NEXT: v_alignbit_b32 v20, v31, v20, 16
; SI-NEXT: s_waitcnt vmcnt(0)
-; SI-NEXT: v_cvt_f32_f16_e32 v41, v7
+; SI-NEXT: v_cvt_f32_f16_e32 v41, v41
; SI-NEXT: v_add_f32_e32 v41, 0x38000000, v41
; SI-NEXT: v_cvt_f16_f32_e32 v41, v41
-; SI-NEXT: v_or_b32_e32 v7, v41, v28
-; SI-NEXT: buffer_store_dword v7, off, s[0:3], s32 offset:156 ; 4-byte Folded Spill
+; SI-NEXT: v_or_b32_e32 v41, v41, v28
+; SI-NEXT: buffer_store_dword v41, off, s[0:3], s32 offset:160 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: buffer_load_dword v7, off, s[0:3], s32 offset:152 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v41, off, s[0:3], s32 offset:156 ; 4-byte Folded Reload
; SI-NEXT: v_alignbit_b32 v28, v15, v28, 16
; SI-NEXT: s_waitcnt vmcnt(0)
-; SI-NEXT: v_cvt_f32_f16_e32 v41, v7
-; SI-NEXT: buffer_load_dword v7, off, s[0:3], s32 offset:148 ; 4-byte Folded Reload
+; SI-NEXT: v_cvt_f32_f16_e32 v41, v41
; SI-NEXT: v_add_f32_e32 v41, 0x38000000, v41
; SI-NEXT: v_cvt_f16_f32_e32 v41, v41
-; SI-NEXT: s_waitcnt vmcnt(0)
-; SI-NEXT: v_cvt_f32_f16_e32 v42, v7
-; SI-NEXT: v_or_b32_e32 v7, v41, v27
-; SI-NEXT: buffer_store_dword v7, off, s[0:3], s32 offset:152 ; 4-byte Folded Spill
-; SI-NEXT: v_alignbit_b32 v27, v11, v27, 16
+; SI-NEXT: v_or_b32_e32 v41, v41, v27
+; SI-NEXT: buffer_store_dword v41, off, s[0:3], s32 offset:156 ; 4-byte Folded Spill
+; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_add_f32_e32 v41, 0x38000000, v42
+; SI-NEXT: buffer_load_dword v42, off, s[0:3], s32 offset:148 ; 4-byte Folded Reload
; SI-NEXT: v_cvt_f16_f32_e32 v41, v41
-; SI-NEXT: v_cvt_f32_f16_e32 v42, v43
-; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_or_b32_e32 v7, v41, v26
-; SI-NEXT: v_cvt_f32_f16_e32 v41, v1
-; SI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:176 ; 4-byte Folded Reload
+; SI-NEXT: v_alignbit_b32 v27, v11, v27, 16
+; SI-NEXT: v_or_b32_e32 v41, v41, v26
+; SI-NEXT: buffer_store_dword v41, off, s[0:3], s32 offset:152 ; 4-byte Folded Spill
+; SI-NEXT: v_alignbit_b32 v26, v8, v26, 16
+; SI-NEXT: s_waitcnt vmcnt(1)
+; SI-NEXT: v_cvt_f32_f16_e32 v42, v42
; SI-NEXT: v_add_f32_e32 v42, 0x38000000, v42
; SI-NEXT: v_cvt_f16_f32_e32 v42, v42
-; SI-NEXT: v_add_f32_e32 v41, 0x38000000, v41
-; SI-NEXT: v_cvt_f16_f32_e32 v41, v41
-; SI-NEXT: buffer_store_dword v7, off, s[0:3], s32 offset:148 ; 4-byte Folded Spill
-; SI-NEXT: v_or_b32_e32 v43, v42, v24
-; SI-NEXT: v_alignbit_b32 v26, v8, v26, 16
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_or_b32_e32 v7, v41, v37
-; SI-NEXT: v_mov_b32_e32 v51, v7
-; SI-NEXT: v_alignbit_b32 v7, v38, v40, 16
+; SI-NEXT: v_or_b32_e32 v41, v42, v24
+; SI-NEXT: buffer_store_dword v41, off, s[0:3], s32 offset:148 ; 4-byte Folded Spill
+; SI-NEXT: s_waitcnt expcnt(0)
+; SI-NEXT: v_cvt_f32_f16_e32 v41, v1
; SI-NEXT: v_alignbit_b32 v24, v5, v24, 16
-; SI-NEXT: s_waitcnt vmcnt(1)
+; SI-NEXT: v_add_f32_e32 v41, 0x38000000, v41
+; SI-NEXT: v_cvt_f16_f32_e32 v41, v41
+; SI-NEXT: v_or_b32_e32 v1, v41, v37
+; SI-NEXT: v_mov_b32_e32 v51, v1
+; SI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:176 ; 4-byte Folded Reload
+; SI-NEXT: s_waitcnt vmcnt(0)
; SI-NEXT: v_alignbit_b32 v37, v1, v37, 16
; SI-NEXT: .LBB109_3: ; %end
; SI-NEXT: v_and_b32_e32 v48, 0xffff, v60
@@ -246512,7 +246504,6 @@ define inreg <64 x i16> @bitcast_v64f16_to_v64i16_scalar(<64 x half> inreg %a, i
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: buffer_load_dword v4, off, s[0:3], s32 offset:172 ; 4-byte Folded Reload
; SI-NEXT: v_lshlrev_b32_e32 v29, 16, v44
-; SI-NEXT: v_lshlrev_b32_e32 v7, 16, v7
; SI-NEXT: v_and_b32_e32 v2, 0xffff, v2
; SI-NEXT: v_lshlrev_b32_e32 v3, 16, v3
; SI-NEXT: v_or_b32_e32 v2, v2, v3
@@ -246530,9 +246521,8 @@ define inreg <64 x i16> @bitcast_v64f16_to_v64i16_scalar(<64 x half> inreg %a, i
; SI-NEXT: v_add_i32_e32 v29, vcc, 36, v0
; SI-NEXT: buffer_store_dword v4, v29, s[0:3], 0 offen
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: buffer_load_dword v4, off, s[0:3], s32 offset:168 ; 4-byte Folded Reload
-; SI-NEXT: s_waitcnt vmcnt(0)
-; SI-NEXT: v_and_b32_e32 v4, 0xffff, v4
+; SI-NEXT: v_and_b32_e32 v4, 0xffff, v7
+; SI-NEXT: v_lshlrev_b32_e32 v7, 16, v43
; SI-NEXT: v_or_b32_e32 v4, v4, v7
; SI-NEXT: v_add_i32_e32 v7, vcc, 40, v0
; SI-NEXT: buffer_store_dword v4, v7, s[0:3], 0 offen
@@ -246543,7 +246533,7 @@ define inreg <64 x i16> @bitcast_v64f16_to_v64i16_scalar(<64 x half> inreg %a, i
; SI-NEXT: v_add_i32_e32 v7, vcc, 44, v0
; SI-NEXT: buffer_store_dword v4, v7, s[0:3], 0 offen
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: buffer_load_dword v4, off, s[0:3], s32 offset:164 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v4, off, s[0:3], s32 offset:168 ; 4-byte Folded Reload
; SI-NEXT: v_lshlrev_b32_e32 v7, 16, v10
; SI-NEXT: s_waitcnt vmcnt(0)
; SI-NEXT: v_and_b32_e32 v4, 0xffff, v4
@@ -246569,7 +246559,7 @@ define inreg <64 x i16> @bitcast_v64f16_to_v64i16_scalar(<64 x half> inreg %a, i
; SI-NEXT: v_add_i32_e32 v7, vcc, 60, v0
; SI-NEXT: buffer_store_dword v4, v7, s[0:3], 0 offen
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: buffer_load_dword v4, off, s[0:3], s32 offset:160 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v4, off, s[0:3], s32 offset:164 ; 4-byte Folded Reload
; SI-NEXT: v_lshlrev_b32_e32 v7, 16, v20
; SI-NEXT: s_waitcnt vmcnt(0)
; SI-NEXT: v_and_b32_e32 v4, 0xffff, v4
@@ -246595,7 +246585,7 @@ define inreg <64 x i16> @bitcast_v64f16_to_v64i16_scalar(<64 x half> inreg %a, i
; SI-NEXT: v_add_i32_e32 v7, vcc, 0x4c, v0
; SI-NEXT: buffer_store_dword v4, v7, s[0:3], 0 offen
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: buffer_load_dword v4, off, s[0:3], s32 offset:156 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v4, off, s[0:3], s32 offset:160 ; 4-byte Folded Reload
; SI-NEXT: v_lshlrev_b32_e32 v7, 16, v28
; SI-NEXT: s_waitcnt vmcnt(0)
; SI-NEXT: v_and_b32_e32 v4, 0xffff, v4
@@ -246609,7 +246599,7 @@ define inreg <64 x i16> @bitcast_v64f16_to_v64i16_scalar(<64 x half> inreg %a, i
; SI-NEXT: v_add_i32_e32 v7, vcc, 0x54, v0
; SI-NEXT: buffer_store_dword v4, v7, s[0:3], 0 offen
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: buffer_load_dword v4, off, s[0:3], s32 offset:152 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v4, off, s[0:3], s32 offset:156 ; 4-byte Folded Reload
; SI-NEXT: v_lshlrev_b32_e32 v7, 16, v27
; SI-NEXT: s_waitcnt vmcnt(0)
; SI-NEXT: v_and_b32_e32 v4, 0xffff, v4
@@ -246623,7 +246613,7 @@ define inreg <64 x i16> @bitcast_v64f16_to_v64i16_scalar(<64 x half> inreg %a, i
; SI-NEXT: v_add_i32_e32 v7, vcc, 0x5c, v0
; SI-NEXT: buffer_store_dword v4, v7, s[0:3], 0 offen
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: buffer_load_dword v4, off, s[0:3], s32 offset:148 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v4, off, s[0:3], s32 offset:152 ; 4-byte Folded Reload
; SI-NEXT: v_lshlrev_b32_e32 v7, 16, v26
; SI-NEXT: s_waitcnt vmcnt(0)
; SI-NEXT: v_and_b32_e32 v4, 0xffff, v4
@@ -246637,8 +246627,10 @@ define inreg <64 x i16> @bitcast_v64f16_to_v64i16_scalar(<64 x half> inreg %a, i
; SI-NEXT: v_add_i32_e32 v7, vcc, 0x64, v0
; SI-NEXT: buffer_store_dword v4, v7, s[0:3], 0 offen
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_and_b32_e32 v4, 0xffff, v43
+; SI-NEXT: buffer_load_dword v4, off, s[0:3], s32 offset:148 ; 4-byte Folded Reload
; SI-NEXT: v_lshlrev_b32_e32 v7, 16, v24
+; SI-NEXT: s_waitcnt vmcnt(0)
+; SI-NEXT: v_and_b32_e32 v4, 0xffff, v4
; SI-NEXT: v_or_b32_e32 v4, v4, v7
; SI-NEXT: v_add_i32_e32 v7, vcc, 0x68, v0
; SI-NEXT: buffer_store_dword v4, v7, s[0:3], 0 offen
@@ -248341,10 +248333,9 @@ define inreg <64 x half> @bitcast_v64i16_to_v64f16_scalar(<64 x i16> inreg %a, i
; SI-NEXT: buffer_store_dword v62, off, s[0:3], s32 offset:88 ; 4-byte Folded Spill
; SI-NEXT: buffer_store_dword v63, off, s[0:3], s32 offset:84 ; 4-byte Folded Spill
; SI-NEXT: buffer_load_dword v41, off, s[0:3], s32 offset:80
-; SI-NEXT: s_waitcnt expcnt(1)
-; SI-NEXT: buffer_load_dword v62, off, s[0:3], s32 offset:12
-; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: buffer_load_dword v63, off, s[0:3], s32 offset:8
+; SI-NEXT: s_waitcnt expcnt(4)
+; SI-NEXT: buffer_load_dword v59, off, s[0:3], s32 offset:12
+; SI-NEXT: buffer_load_dword v46, off, s[0:3], s32 offset:8
; SI-NEXT: buffer_load_dword v31, off, s[0:3], s32 offset:4
; SI-NEXT: buffer_load_dword v32, off, s[0:3], s32
; SI-NEXT: buffer_load_dword v33, off, s[0:3], s32 offset:76
@@ -248364,13 +248355,13 @@ define inreg <64 x half> @bitcast_v64i16_to_v64f16_scalar(<64 x i16> inreg %a, i
; SI-NEXT: buffer_load_dword v55, off, s[0:3], s32 offset:20
; SI-NEXT: buffer_load_dword v40, off, s[0:3], s32 offset:16
; SI-NEXT: v_mov_b32_e32 v42, v4
-; SI-NEXT: buffer_store_dword v42, off, s[0:3], s32 offset:272 ; 4-byte Folded Spill
-; SI-NEXT: buffer_store_dword v5, off, s[0:3], s32 offset:276 ; 4-byte Folded Spill
-; SI-NEXT: buffer_store_dword v6, off, s[0:3], s32 offset:280 ; 4-byte Folded Spill
-; SI-NEXT: buffer_store_dword v7, off, s[0:3], s32 offset:284 ; 4-byte Folded Spill
-; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:288 ; 4-byte Folded Spill
-; SI-NEXT: buffer_store_dword v2, off, s[0:3], s32 offset:292 ; 4-byte Folded Spill
-; SI-NEXT: buffer_store_dword v3, off, s[0:3], s32 offset:296 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:272 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v2, off, s[0:3], s32 offset:276 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v3, off, s[0:3], s32 offset:280 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v6, off, s[0:3], s32 offset:284 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v8, off, s[0:3], s32 offset:288 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v9, off, s[0:3], s32 offset:292 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v10, off, s[0:3], s32 offset:296 ; 4-byte Folded Spill
; SI-NEXT: buffer_store_dword v11, off, s[0:3], s32 offset:300 ; 4-byte Folded Spill
; SI-NEXT: buffer_store_dword v12, off, s[0:3], s32 offset:304 ; 4-byte Folded Spill
; SI-NEXT: buffer_store_dword v13, off, s[0:3], s32 offset:308 ; 4-byte Folded Spill
@@ -248381,15 +248372,14 @@ define inreg <64 x half> @bitcast_v64i16_to_v64f16_scalar(<64 x i16> inreg %a, i
; SI-NEXT: buffer_store_dword v18, off, s[0:3], s32 offset:328 ; 4-byte Folded Spill
; SI-NEXT: buffer_store_dword v19, off, s[0:3], s32 offset:332 ; 4-byte Folded Spill
; SI-NEXT: buffer_store_dword v20, off, s[0:3], s32 offset:336 ; 4-byte Folded Spill
-; SI-NEXT: buffer_store_dword v8, off, s[0:3], s32 offset:340 ; 4-byte Folded Spill
-; SI-NEXT: buffer_store_dword v21, off, s[0:3], s32 offset:344 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v21, off, s[0:3], s32 offset:340 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v42, off, s[0:3], s32 offset:344 ; 4-byte Folded Spill
; SI-NEXT: buffer_store_dword v22, off, s[0:3], s32 offset:348 ; 4-byte Folded Spill
; SI-NEXT: buffer_store_dword v23, off, s[0:3], s32 offset:352 ; 4-byte Folded Spill
; SI-NEXT: buffer_store_dword v24, off, s[0:3], s32 offset:356 ; 4-byte Folded Spill
; SI-NEXT: buffer_store_dword v25, off, s[0:3], s32 offset:360 ; 4-byte Folded Spill
; SI-NEXT: buffer_store_dword v26, off, s[0:3], s32 offset:364 ; 4-byte Folded Spill
; SI-NEXT: buffer_store_dword v27, off, s[0:3], s32 offset:368 ; 4-byte Folded Spill
-; SI-NEXT: buffer_store_dword v28, off, s[0:3], s32 offset:372 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt vmcnt(14)
; SI-NEXT: v_cmp_ne_u32_e32 vcc, 0, v41
; SI-NEXT: s_and_b64 s[4:5], vcc, exec
@@ -248398,35 +248388,35 @@ define inreg <64 x half> @bitcast_v64i16_to_v64f16_scalar(<64 x i16> inreg %a, i
; SI-NEXT: v_cvt_f32_f16_e32 v4, v1
; SI-NEXT: v_cvt_f32_f16_e32 v41, s16
; SI-NEXT: v_cvt_f32_f16_e32 v44, s26
-; SI-NEXT: v_cvt_f32_f16_e32 v43, s29
+; SI-NEXT: v_cvt_f32_f16_e32 v45, s27
; SI-NEXT: buffer_store_dword v4, off, s[0:3], s32 offset:256 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_cvt_f32_f16_e32 v4, v2
; SI-NEXT: buffer_store_dword v41, off, s[0:3], s32 offset:168 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_cvt_f32_f16_e32 v41, s17
-; SI-NEXT: s_mov_b64 s[4:5], 0
+; SI-NEXT: v_cvt_f32_f16_e32 v43, s29
; SI-NEXT: buffer_store_dword v4, off, s[0:3], s32 offset:268 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_cvt_f32_f16_e32 v4, v3
; SI-NEXT: buffer_store_dword v41, off, s[0:3], s32 offset:184 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_cvt_f32_f16_e32 v41, s18
-; SI-NEXT: v_cvt_f32_f16_e32 v59, s27
+; SI-NEXT: s_mov_b64 s[4:5], 0
; SI-NEXT: buffer_store_dword v4, off, s[0:3], s32 offset:248 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_cvt_f32_f16_e32 v4, v42
; SI-NEXT: buffer_store_dword v41, off, s[0:3], s32 offset:164 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_cvt_f32_f16_e32 v41, s19
-; SI-NEXT: v_mov_b32_e32 v2, v9
+; SI-NEXT: v_mov_b32_e32 v2, v5
; SI-NEXT: buffer_store_dword v4, off, s[0:3], s32 offset:264 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_cvt_f32_f16_e32 v4, v5
; SI-NEXT: buffer_store_dword v41, off, s[0:3], s32 offset:180 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_cvt_f32_f16_e32 v41, s20
-; SI-NEXT: v_mov_b32_e32 v3, v10
+; SI-NEXT: v_mov_b32_e32 v3, v7
; SI-NEXT: buffer_store_dword v4, off, s[0:3], s32 offset:240 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_cvt_f32_f16_e32 v4, v6
@@ -248461,56 +248451,56 @@ define inreg <64 x half> @bitcast_v64i16_to_v64f16_scalar(<64 x i16> inreg %a, i
; SI-NEXT: buffer_store_dword v41, off, s[0:3], s32 offset:148 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_cvt_f32_f16_e32 v41, s25
-; SI-NEXT: v_mov_b32_e32 v60, v29
+; SI-NEXT: v_mov_b32_e32 v62, v28
; SI-NEXT: buffer_store_dword v4, off, s[0:3], s32 offset:244 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_cvt_f32_f16_e32 v4, v11
; SI-NEXT: buffer_store_dword v41, off, s[0:3], s32 offset:156 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_cvt_f32_f16_e32 v41, s28
-; SI-NEXT: v_mov_b32_e32 v61, v30
+; SI-NEXT: v_mov_b32_e32 v63, v29
; SI-NEXT: buffer_store_dword v4, off, s[0:3], s32 offset:216 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_cvt_f32_f16_e32 v4, v12
+; SI-NEXT: v_mov_b32_e32 v60, v30
+; SI-NEXT: v_mov_b32_e32 v61, v32
; SI-NEXT: v_cvt_f32_f16_e32 v11, v32
-; SI-NEXT: v_cvt_f32_f16_e32 v9, v63
-; SI-NEXT: v_cvt_f32_f16_e32 v12, v62
; SI-NEXT: buffer_store_dword v4, off, s[0:3], s32 offset:236 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_cvt_f32_f16_e32 v4, v13
; SI-NEXT: v_cvt_f32_f16_e32 v13, v29
-; SI-NEXT: v_cvt_f32_f16_e32 v7, v40
-; SI-NEXT: v_cvt_f32_f16_e32 v10, v55
+; SI-NEXT: v_mov_b32_e32 v32, v31
+; SI-NEXT: v_cvt_f32_f16_e32 v9, v46
; SI-NEXT: buffer_store_dword v4, off, s[0:3], s32 offset:208 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_cvt_f32_f16_e32 v4, v14
; SI-NEXT: v_cvt_f32_f16_e32 v14, v31
-; SI-NEXT: v_cvt_f32_f16_e32 v5, v54
-; SI-NEXT: v_cvt_f32_f16_e32 v8, v53
+; SI-NEXT: v_mov_b32_e32 v31, v46
+; SI-NEXT: v_cvt_f32_f16_e32 v12, v59
; SI-NEXT: buffer_store_dword v4, off, s[0:3], s32 offset:228 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_cvt_f32_f16_e32 v4, v15
; SI-NEXT: v_cvt_f32_f16_e32 v15, v27
-; SI-NEXT: v_cvt_f32_f16_e32 v58, v52
-; SI-NEXT: v_cvt_f32_f16_e32 v6, v51
+; SI-NEXT: v_cvt_f32_f16_e32 v7, v40
+; SI-NEXT: v_cvt_f32_f16_e32 v10, v55
; SI-NEXT: buffer_store_dword v4, off, s[0:3], s32 offset:200 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_cvt_f32_f16_e32 v4, v16
; SI-NEXT: v_cvt_f32_f16_e32 v16, v30
-; SI-NEXT: v_cvt_f32_f16_e32 v56, v50
-; SI-NEXT: v_cvt_f32_f16_e32 v46, v48
+; SI-NEXT: v_cvt_f32_f16_e32 v5, v54
+; SI-NEXT: v_cvt_f32_f16_e32 v8, v53
; SI-NEXT: buffer_store_dword v4, off, s[0:3], s32 offset:220 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_cvt_f32_f16_e32 v4, v17
; SI-NEXT: v_cvt_f32_f16_e32 v17, v25
; SI-NEXT: v_mov_b32_e32 v25, v44
-; SI-NEXT: v_cvt_f32_f16_e32 v57, v39
+; SI-NEXT: v_cvt_f32_f16_e32 v58, v52
; SI-NEXT: buffer_store_dword v4, off, s[0:3], s32 offset:196 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_cvt_f32_f16_e32 v4, v18
; SI-NEXT: v_cvt_f32_f16_e32 v18, v28
-; SI-NEXT: v_cvt_f32_f16_e32 v44, v38
-; SI-NEXT: v_cvt_f32_f16_e32 v47, v37
+; SI-NEXT: v_cvt_f32_f16_e32 v6, v51
+; SI-NEXT: v_cvt_f32_f16_e32 v56, v50
; SI-NEXT: buffer_store_dword v4, off, s[0:3], s32 offset:212 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_cvt_f32_f16_e32 v4, v19
@@ -248521,11 +248511,16 @@ define inreg <64 x half> @bitcast_v64i16_to_v64f16_scalar(<64 x i16> inreg %a, i
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_cvt_f32_f16_e32 v4, v20
; SI-NEXT: v_cvt_f32_f16_e32 v20, v26
-; SI-NEXT: v_cvt_f32_f16_e32 v42, v36
-; SI-NEXT: v_cvt_f32_f16_e32 v45, v35
+; SI-NEXT: v_mov_b32_e32 v26, v45
+; SI-NEXT: v_cvt_f32_f16_e32 v46, v48
; SI-NEXT: buffer_store_dword v4, off, s[0:3], s32 offset:204 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_cvt_f32_f16_e32 v4, v49
+; SI-NEXT: v_cvt_f32_f16_e32 v57, v39
+; SI-NEXT: v_cvt_f32_f16_e32 v44, v38
+; SI-NEXT: v_cvt_f32_f16_e32 v47, v37
+; SI-NEXT: v_cvt_f32_f16_e32 v42, v36
+; SI-NEXT: v_cvt_f32_f16_e32 v45, v35
; SI-NEXT: buffer_store_dword v41, off, s[0:3], s32 offset:188 ; 4-byte Folded Spill
; SI-NEXT: v_cvt_f32_f16_e32 v43, v33
; SI-NEXT: s_branch .LBB111_3
@@ -248534,33 +248529,53 @@ define inreg <64 x half> @bitcast_v64i16_to_v64f16_scalar(<64 x i16> inreg %a, i
; SI-NEXT: ; kill: killed $vgpr4
; SI-NEXT: ; implicit-def: $vgpr41
; SI-NEXT: ; kill: killed $vgpr41
-; SI-NEXT: v_mov_b32_e32 v61, v30
+; SI-NEXT: v_mov_b32_e32 v61, v32
; SI-NEXT: ; implicit-def: $vgpr4
; SI-NEXT: ; kill: killed $vgpr4
; SI-NEXT: ; implicit-def: $vgpr41
; SI-NEXT: ; kill: killed $vgpr41
-; SI-NEXT: v_mov_b32_e32 v60, v29
+; SI-NEXT: v_mov_b32_e32 v32, v31
; SI-NEXT: ; implicit-def: $vgpr4
; SI-NEXT: ; kill: killed $vgpr4
; SI-NEXT: ; implicit-def: $vgpr41
; SI-NEXT: ; kill: killed $vgpr41
-; SI-NEXT: v_mov_b32_e32 v3, v10
+; SI-NEXT: v_mov_b32_e32 v31, v46
; SI-NEXT: ; implicit-def: $vgpr4
; SI-NEXT: ; kill: killed $vgpr4
; SI-NEXT: ; implicit-def: $vgpr41
; SI-NEXT: ; kill: killed $vgpr41
-; SI-NEXT: v_mov_b32_e32 v2, v9
+; SI-NEXT: v_mov_b32_e32 v60, v30
; SI-NEXT: ; implicit-def: $vgpr4
; SI-NEXT: ; kill: killed $vgpr4
; SI-NEXT: ; implicit-def: $vgpr41
; SI-NEXT: ; kill: killed $vgpr41
+; SI-NEXT: v_mov_b32_e32 v63, v29
+; SI-NEXT: ; implicit-def: $vgpr4
+; SI-NEXT: ; kill: killed $vgpr4
+; SI-NEXT: ; implicit-def: $vgpr41
+; SI-NEXT: ; kill: killed $vgpr41
+; SI-NEXT: v_mov_b32_e32 v62, v28
+; SI-NEXT: ; implicit-def: $vgpr4
+; SI-NEXT: ; kill: killed $vgpr4
+; SI-NEXT: ; implicit-def: $vgpr41
+; SI-NEXT: ; kill: killed $vgpr41
+; SI-NEXT: v_mov_b32_e32 v3, v7
+; SI-NEXT: ; implicit-def: $vgpr4
+; SI-NEXT: ; kill: killed $vgpr4
+; SI-NEXT: ; implicit-def: $vgpr41
+; SI-NEXT: ; kill: killed $vgpr41
+; SI-NEXT: v_mov_b32_e32 v2, v5
+; SI-NEXT: ; implicit-def: $vgpr4
+; SI-NEXT: ; kill: killed $vgpr4
+; SI-NEXT: ; implicit-def: $vgpr41
; SI-NEXT: s_mov_b64 s[4:5], -1
; SI-NEXT: ; implicit-def: $vgpr4
; SI-NEXT: ; kill: killed $vgpr4
+; SI-NEXT: ; kill: killed $vgpr41
; SI-NEXT: ; implicit-def: $vgpr41
; SI-NEXT: ; kill: killed $vgpr41
; SI-NEXT: ; implicit-def: $vgpr25
-; SI-NEXT: ; implicit-def: $vgpr59
+; SI-NEXT: ; implicit-def: $vgpr26
; SI-NEXT: ; implicit-def: $vgpr23
; SI-NEXT: ; implicit-def: $vgpr24
; SI-NEXT: ; implicit-def: $vgpr1
@@ -248597,22 +248612,6 @@ define inreg <64 x half> @bitcast_v64i16_to_v64f16_scalar(<64 x i16> inreg %a, i
; SI-NEXT: ; kill: killed $vgpr41
; SI-NEXT: ; implicit-def: $vgpr4
; SI-NEXT: ; kill: killed $vgpr4
-; SI-NEXT: ; implicit-def: $vgpr41
-; SI-NEXT: ; kill: killed $vgpr41
-; SI-NEXT: ; implicit-def: $vgpr4
-; SI-NEXT: ; kill: killed $vgpr4
-; SI-NEXT: ; implicit-def: $vgpr41
-; SI-NEXT: ; kill: killed $vgpr41
-; SI-NEXT: ; implicit-def: $vgpr4
-; SI-NEXT: ; kill: killed $vgpr4
-; SI-NEXT: ; implicit-def: $vgpr41
-; SI-NEXT: ; kill: killed $vgpr41
-; SI-NEXT: ; implicit-def: $vgpr4
-; SI-NEXT: ; kill: killed $vgpr4
-; SI-NEXT: ; implicit-def: $vgpr41
-; SI-NEXT: ; kill: killed $vgpr41
-; SI-NEXT: ; implicit-def: $vgpr4
-; SI-NEXT: ; kill: killed $vgpr4
; SI-NEXT: ; implicit-def: $vgpr4
; SI-NEXT: ; kill: killed $vgpr4
; SI-NEXT: ; implicit-def: $vgpr4
@@ -248649,10 +248648,10 @@ define inreg <64 x half> @bitcast_v64i16_to_v64f16_scalar(<64 x i16> inreg %a, i
; SI-NEXT: v_mov_b32_e32 v19, v1
; SI-NEXT: s_cbranch_vccnz .LBB111_5
; SI-NEXT: ; %bb.4: ; %cmp.true
-; SI-NEXT: buffer_load_dword v4, off, s[0:3], s32 offset:336 ; 4-byte Folded Reload
-; SI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:372 ; 4-byte Folded Reload
-; SI-NEXT: v_add_i32_e32 v10, vcc, 3, v3
-; SI-NEXT: v_add_i32_e32 v9, vcc, 3, v2
+; SI-NEXT: buffer_load_dword v4, off, s[0:3], s32 offset:340 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:368 ; 4-byte Folded Reload
+; SI-NEXT: v_add_i32_e32 v7, vcc, 3, v3
+; SI-NEXT: v_add_i32_e32 v5, vcc, 3, v2
; SI-NEXT: s_add_i32 s16, s16, 3
; SI-NEXT: v_cvt_f32_f16_e32 v41, s16
; SI-NEXT: s_add_i32 s17, s17, 3
@@ -248707,16 +248706,17 @@ define inreg <64 x half> @bitcast_v64i16_to_v64f16_scalar(<64 x i16> inreg %a, i
; SI-NEXT: v_add_i32_e32 v54, vcc, 3, v54
; SI-NEXT: v_add_i32_e32 v55, vcc, 3, v55
; SI-NEXT: v_add_i32_e32 v40, vcc, 3, v40
-; SI-NEXT: v_add_i32_e32 v58, vcc, 3, v62
-; SI-NEXT: v_add_i32_e32 v59, vcc, 3, v63
-; SI-NEXT: v_add_i32_e32 v31, vcc, 3, v31
-; SI-NEXT: v_add_i32_e32 v32, vcc, 3, v32
-; SI-NEXT: v_add_i32_e32 v30, vcc, 3, v61
-; SI-NEXT: v_add_i32_e32 v29, vcc, 3, v60
-; SI-NEXT: s_add_i32 s29, s29, 3
+; SI-NEXT: v_add_i32_e32 v58, vcc, 3, v59
+; SI-NEXT: v_add_i32_e32 v59, vcc, 3, v31
+; SI-NEXT: v_add_i32_e32 v31, vcc, 3, v32
+; SI-NEXT: v_add_i32_e32 v32, vcc, 3, v61
+; SI-NEXT: v_add_i32_e32 v30, vcc, 3, v60
+; SI-NEXT: v_add_i32_e32 v29, vcc, 3, v63
+; SI-NEXT: v_add_i32_e32 v28, vcc, 3, v62
; SI-NEXT: s_waitcnt vmcnt(10)
-; SI-NEXT: v_add_i32_e32 v20, vcc, 3, v4
-; SI-NEXT: buffer_load_dword v4, off, s[0:3], s32 offset:332 ; 4-byte Folded Reload
+; SI-NEXT: v_add_i32_e32 v21, vcc, 3, v4
+; SI-NEXT: buffer_load_dword v4, off, s[0:3], s32 offset:336 ; 4-byte Folded Reload
+; SI-NEXT: s_add_i32 s29, s29, 3
; SI-NEXT: s_add_i32 s28, s28, 3
; SI-NEXT: s_add_i32 s27, s27, 3
; SI-NEXT: s_add_i32 s26, s26, 3
@@ -248732,75 +248732,72 @@ define inreg <64 x half> @bitcast_v64i16_to_v64f16_scalar(<64 x i16> inreg %a, i
; SI-NEXT: v_cvt_f32_f16_e32 v45, v35
; SI-NEXT: v_cvt_f32_f16_e32 v43, v33
; SI-NEXT: s_waitcnt vmcnt(1)
+; SI-NEXT: v_add_i32_e32 v20, vcc, 3, v4
+; SI-NEXT: buffer_load_dword v4, off, s[0:3], s32 offset:332 ; 4-byte Folded Reload
+; SI-NEXT: s_waitcnt vmcnt(0)
; SI-NEXT: v_add_i32_e32 v19, vcc, 3, v4
; SI-NEXT: buffer_load_dword v4, off, s[0:3], s32 offset:328 ; 4-byte Folded Reload
-; SI-NEXT: v_add_i32_e32 v28, vcc, 3, v1
-; SI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:368 ; 4-byte Folded Reload
-; SI-NEXT: s_waitcnt vmcnt(1)
+; SI-NEXT: s_waitcnt vmcnt(0)
; SI-NEXT: v_add_i32_e32 v18, vcc, 3, v4
; SI-NEXT: buffer_load_dword v4, off, s[0:3], s32 offset:324 ; 4-byte Folded Reload
-; SI-NEXT: s_waitcnt vmcnt(1)
-; SI-NEXT: v_add_i32_e32 v27, vcc, 3, v1
-; SI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:364 ; 4-byte Folded Reload
-; SI-NEXT: s_waitcnt vmcnt(1)
+; SI-NEXT: s_waitcnt vmcnt(0)
; SI-NEXT: v_add_i32_e32 v17, vcc, 3, v4
; SI-NEXT: buffer_load_dword v4, off, s[0:3], s32 offset:320 ; 4-byte Folded Reload
-; SI-NEXT: s_waitcnt vmcnt(1)
-; SI-NEXT: v_add_i32_e32 v26, vcc, 3, v1
-; SI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:360 ; 4-byte Folded Reload
-; SI-NEXT: s_waitcnt vmcnt(1)
+; SI-NEXT: s_waitcnt vmcnt(0)
; SI-NEXT: v_add_i32_e32 v16, vcc, 3, v4
; SI-NEXT: buffer_load_dword v4, off, s[0:3], s32 offset:316 ; 4-byte Folded Reload
-; SI-NEXT: s_waitcnt vmcnt(1)
-; SI-NEXT: v_add_i32_e32 v25, vcc, 3, v1
-; SI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:356 ; 4-byte Folded Reload
-; SI-NEXT: s_waitcnt vmcnt(1)
+; SI-NEXT: s_waitcnt vmcnt(0)
; SI-NEXT: v_add_i32_e32 v15, vcc, 3, v4
; SI-NEXT: buffer_load_dword v4, off, s[0:3], s32 offset:312 ; 4-byte Folded Reload
-; SI-NEXT: s_waitcnt vmcnt(1)
-; SI-NEXT: v_add_i32_e32 v24, vcc, 3, v1
-; SI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:352 ; 4-byte Folded Reload
+; SI-NEXT: v_add_i32_e32 v27, vcc, 3, v1
+; SI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:364 ; 4-byte Folded Reload
; SI-NEXT: s_waitcnt vmcnt(1)
; SI-NEXT: v_add_i32_e32 v14, vcc, 3, v4
; SI-NEXT: buffer_load_dword v4, off, s[0:3], s32 offset:308 ; 4-byte Folded Reload
; SI-NEXT: s_waitcnt vmcnt(1)
-; SI-NEXT: v_add_i32_e32 v23, vcc, 3, v1
-; SI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:348 ; 4-byte Folded Reload
+; SI-NEXT: v_add_i32_e32 v26, vcc, 3, v1
+; SI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:360 ; 4-byte Folded Reload
; SI-NEXT: s_waitcnt vmcnt(1)
; SI-NEXT: v_add_i32_e32 v13, vcc, 3, v4
; SI-NEXT: buffer_load_dword v4, off, s[0:3], s32 offset:304 ; 4-byte Folded Reload
; SI-NEXT: s_waitcnt vmcnt(1)
-; SI-NEXT: v_add_i32_e32 v22, vcc, 3, v1
-; SI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:344 ; 4-byte Folded Reload
-; SI-NEXT: v_cvt_f32_f16_e32 v22, v22
+; SI-NEXT: v_add_i32_e32 v25, vcc, 3, v1
+; SI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:356 ; 4-byte Folded Reload
; SI-NEXT: s_waitcnt vmcnt(1)
; SI-NEXT: v_add_i32_e32 v12, vcc, 3, v4
; SI-NEXT: buffer_load_dword v4, off, s[0:3], s32 offset:300 ; 4-byte Folded Reload
; SI-NEXT: s_waitcnt vmcnt(1)
-; SI-NEXT: v_add_i32_e32 v21, vcc, 3, v1
-; SI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:340 ; 4-byte Folded Reload
+; SI-NEXT: v_add_i32_e32 v24, vcc, 3, v1
+; SI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:352 ; 4-byte Folded Reload
; SI-NEXT: s_waitcnt vmcnt(1)
; SI-NEXT: v_add_i32_e32 v11, vcc, 3, v4
-; SI-NEXT: buffer_load_dword v4, off, s[0:3], s32 offset:284 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v4, off, s[0:3], s32 offset:296 ; 4-byte Folded Reload
; SI-NEXT: s_waitcnt vmcnt(1)
-; SI-NEXT: v_add_i32_e32 v8, vcc, 3, v1
-; SI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:296 ; 4-byte Folded Reload
+; SI-NEXT: v_add_i32_e32 v23, vcc, 3, v1
+; SI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:348 ; 4-byte Folded Reload
; SI-NEXT: s_waitcnt vmcnt(1)
-; SI-NEXT: v_add_i32_e32 v7, vcc, 3, v4
-; SI-NEXT: buffer_load_dword v4, off, s[0:3], s32 offset:280 ; 4-byte Folded Reload
+; SI-NEXT: v_add_i32_e32 v10, vcc, 3, v4
+; SI-NEXT: buffer_load_dword v4, off, s[0:3], s32 offset:292 ; 4-byte Folded Reload
; SI-NEXT: s_waitcnt vmcnt(1)
+; SI-NEXT: v_add_i32_e32 v22, vcc, 3, v1
+; SI-NEXT: v_cvt_f32_f16_e32 v22, v22
+; SI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:344 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v3, off, s[0:3], s32 offset:284 ; 4-byte Folded Reload
+; SI-NEXT: s_waitcnt vmcnt(2)
+; SI-NEXT: v_add_i32_e32 v9, vcc, 3, v4
+; SI-NEXT: buffer_load_dword v4, off, s[0:3], s32 offset:288 ; 4-byte Folded Reload
+; SI-NEXT: s_waitcnt vmcnt(0)
+; SI-NEXT: v_add_i32_e32 v8, vcc, 3, v4
+; SI-NEXT: v_add_i32_e32 v4, vcc, 3, v1
+; SI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:280 ; 4-byte Folded Reload
+; SI-NEXT: v_add_i32_e32 v6, vcc, 3, v3
+; SI-NEXT: s_waitcnt vmcnt(0)
; SI-NEXT: v_add_i32_e32 v3, vcc, 3, v1
-; SI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:292 ; 4-byte Folded Reload
-; SI-NEXT: s_waitcnt vmcnt(1)
-; SI-NEXT: v_add_i32_e32 v6, vcc, 3, v4
-; SI-NEXT: buffer_load_dword v4, off, s[0:3], s32 offset:276 ; 4-byte Folded Reload
-; SI-NEXT: s_waitcnt vmcnt(1)
+; SI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:276 ; 4-byte Folded Reload
+; SI-NEXT: s_waitcnt vmcnt(0)
; SI-NEXT: v_add_i32_e32 v2, vcc, 3, v1
-; SI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:288 ; 4-byte Folded Reload
-; SI-NEXT: s_waitcnt vmcnt(1)
-; SI-NEXT: v_add_i32_e32 v5, vcc, 3, v4
-; SI-NEXT: buffer_load_dword v4, off, s[0:3], s32 offset:272 ; 4-byte Folded Reload
-; SI-NEXT: s_waitcnt vmcnt(1)
+; SI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:272 ; 4-byte Folded Reload
+; SI-NEXT: s_waitcnt vmcnt(0)
; SI-NEXT: v_add_i32_e32 v1, vcc, 3, v1
; SI-NEXT: v_cvt_f32_f16_e32 v1, v1
; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:256 ; 4-byte Folded Spill
@@ -248810,8 +248807,6 @@ define inreg <64 x half> @bitcast_v64i16_to_v64f16_scalar(<64 x i16> inreg %a, i
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_cvt_f32_f16_e32 v1, v3
; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:248 ; 4-byte Folded Spill
-; SI-NEXT: s_waitcnt vmcnt(3)
-; SI-NEXT: v_add_i32_e32 v4, vcc, 3, v4
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_cvt_f32_f16_e32 v1, v4
; SI-NEXT: v_cvt_f32_f16_e32 v4, v49
@@ -248827,7 +248822,6 @@ define inreg <64 x half> @bitcast_v64i16_to_v64f16_scalar(<64 x i16> inreg %a, i
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_cvt_f32_f16_e32 v1, v7
; SI-NEXT: v_cvt_f32_f16_e32 v7, v59
-; SI-NEXT: v_cvt_f32_f16_e32 v59, s27
; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:232 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_cvt_f32_f16_e32 v1, v8
@@ -248885,6 +248879,7 @@ define inreg <64 x half> @bitcast_v64i16_to_v64f16_scalar(<64 x i16> inreg %a, i
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_cvt_f32_f16_e32 v1, v20
; SI-NEXT: v_cvt_f32_f16_e32 v20, v26
+; SI-NEXT: v_cvt_f32_f16_e32 v26, s27
; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:204 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_cvt_f32_f16_e32 v1, v34
@@ -248945,7 +248940,7 @@ define inreg <64 x half> @bitcast_v64i16_to_v64f16_scalar(<64 x i16> inreg %a, i
; SI-NEXT: v_or_b32_e32 v1, v2, v1
; SI-NEXT: buffer_store_dword v1, v3, s[0:3], 0 offen
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_cvt_f16_f32_e32 v1, v59
+; SI-NEXT: v_cvt_f16_f32_e32 v1, v26
; SI-NEXT: v_cvt_f16_f32_e32 v2, v25
; SI-NEXT: v_add_i32_e32 v3, vcc, 20, v0
; SI-NEXT: v_lshlrev_b32_e32 v1, 16, v1
diff --git a/llvm/test/CodeGen/AMDGPU/amdgcn.bitcast.128bit.ll b/llvm/test/CodeGen/AMDGPU/amdgcn.bitcast.128bit.ll
index 1056c66720b25..f219402aa5ebf 100644
--- a/llvm/test/CodeGen/AMDGPU/amdgcn.bitcast.128bit.ll
+++ b/llvm/test/CodeGen/AMDGPU/amdgcn.bitcast.128bit.ll
@@ -3743,7 +3743,7 @@ define inreg <16 x i8> @bitcast_v4i32_to_v16i8_scalar(<4 x i32> inreg %a, i32 in
; GFX11-TRUE16: ; %bb.0:
; GFX11-TRUE16-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX11-TRUE16-NEXT: s_cmp_lg_u32 s16, 0
-; GFX11-TRUE16-NEXT: s_mov_b32 s18, 0
+; GFX11-TRUE16-NEXT: s_mov_b32 s16, 0
; GFX11-TRUE16-NEXT: s_cbranch_scc0 .LBB25_4
; GFX11-TRUE16-NEXT: ; %bb.1: ; %cmp.false
; GFX11-TRUE16-NEXT: s_lshr_b32 s8, s3, 24
@@ -3754,11 +3754,11 @@ define inreg <16 x i8> @bitcast_v4i32_to_v16i8_scalar(<4 x i32> inreg %a, i32 in
; GFX11-TRUE16-NEXT: s_lshr_b32 s13, s1, 24
; GFX11-TRUE16-NEXT: s_lshr_b32 s14, s1, 16
; GFX11-TRUE16-NEXT: s_lshr_b32 s15, s1, 8
-; GFX11-TRUE16-NEXT: s_lshr_b32 s16, s0, 16
-; GFX11-TRUE16-NEXT: s_lshr_b32 s17, s0, 8
+; GFX11-TRUE16-NEXT: s_lshr_b32 s17, s0, 16
+; GFX11-TRUE16-NEXT: s_lshr_b32 s18, s0, 8
; GFX11-TRUE16-NEXT: s_lshr_b64 s[4:5], s[2:3], 24
; GFX11-TRUE16-NEXT: s_lshr_b64 s[6:7], s[0:1], 24
-; GFX11-TRUE16-NEXT: s_and_not1_b32 vcc_lo, exec_lo, s18
+; GFX11-TRUE16-NEXT: s_and_not1_b32 vcc_lo, exec_lo, s16
; GFX11-TRUE16-NEXT: s_cbranch_vccnz .LBB25_3
; GFX11-TRUE16-NEXT: .LBB25_2: ; %cmp.true
; GFX11-TRUE16-NEXT: s_add_i32 s1, s1, 3
@@ -3775,12 +3775,12 @@ define inreg <16 x i8> @bitcast_v4i32_to_v16i8_scalar(<4 x i32> inreg %a, i32 in
; GFX11-TRUE16-NEXT: s_lshr_b32 s13, s1, 24
; GFX11-TRUE16-NEXT: s_lshr_b32 s14, s1, 16
; GFX11-TRUE16-NEXT: s_lshr_b32 s15, s1, 8
-; GFX11-TRUE16-NEXT: s_lshr_b32 s16, s0, 16
-; GFX11-TRUE16-NEXT: s_lshr_b32 s17, s0, 8
+; GFX11-TRUE16-NEXT: s_lshr_b32 s17, s0, 16
+; GFX11-TRUE16-NEXT: s_lshr_b32 s18, s0, 8
; GFX11-TRUE16-NEXT: .LBB25_3: ; %end
; GFX11-TRUE16-NEXT: v_mov_b16_e32 v0.l, s0
-; GFX11-TRUE16-NEXT: v_mov_b16_e32 v1.l, s17
-; GFX11-TRUE16-NEXT: v_mov_b16_e32 v2.l, s16
+; GFX11-TRUE16-NEXT: v_mov_b16_e32 v1.l, s18
+; GFX11-TRUE16-NEXT: v_mov_b16_e32 v2.l, s17
; GFX11-TRUE16-NEXT: v_mov_b16_e32 v3.l, s6
; GFX11-TRUE16-NEXT: v_mov_b16_e32 v4.l, s1
; GFX11-TRUE16-NEXT: v_mov_b16_e32 v5.l, s15
@@ -3796,8 +3796,8 @@ define inreg <16 x i8> @bitcast_v4i32_to_v16i8_scalar(<4 x i32> inreg %a, i32 in
; GFX11-TRUE16-NEXT: v_mov_b16_e32 v15.l, s8
; GFX11-TRUE16-NEXT: s_setpc_b64 s[30:31]
; GFX11-TRUE16-NEXT: .LBB25_4:
+; GFX11-TRUE16-NEXT: ; implicit-def: $sgpr18_lo16
; GFX11-TRUE16-NEXT: ; implicit-def: $sgpr17_lo16
-; GFX11-TRUE16-NEXT: ; implicit-def: $sgpr16_lo16
; GFX11-TRUE16-NEXT: ; implicit-def: $sgpr6_lo16
; GFX11-TRUE16-NEXT: ; implicit-def: $sgpr15_lo16
; GFX11-TRUE16-NEXT: ; implicit-def: $sgpr14_lo16
@@ -3814,7 +3814,7 @@ define inreg <16 x i8> @bitcast_v4i32_to_v16i8_scalar(<4 x i32> inreg %a, i32 in
; GFX11-FAKE16: ; %bb.0:
; GFX11-FAKE16-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX11-FAKE16-NEXT: s_cmp_lg_u32 s16, 0
-; GFX11-FAKE16-NEXT: s_mov_b32 s18, 0
+; GFX11-FAKE16-NEXT: s_mov_b32 s17, 0
; GFX11-FAKE16-NEXT: s_cbranch_scc0 .LBB25_4
; GFX11-FAKE16-NEXT: ; %bb.1: ; %cmp.false
; GFX11-FAKE16-NEXT: s_lshr_b32 s8, s3, 24
@@ -3826,10 +3826,10 @@ define inreg <16 x i8> @bitcast_v4i32_to_v16i8_scalar(<4 x i32> inreg %a, i32 in
; GFX11-FAKE16-NEXT: s_lshr_b32 s14, s1, 16
; GFX11-FAKE16-NEXT: s_lshr_b32 s15, s1, 8
; GFX11-FAKE16-NEXT: s_lshr_b32 s16, s0, 16
-; GFX11-FAKE16-NEXT: s_lshr_b32 s17, s0, 8
+; GFX11-FAKE16-NEXT: s_lshr_b32 s18, s0, 8
; GFX11-FAKE16-NEXT: s_lshr_b64 s[4:5], s[2:3], 24
; GFX11-FAKE16-NEXT: s_lshr_b64 s[6:7], s[0:1], 24
-; GFX11-FAKE16-NEXT: s_and_not1_b32 vcc_lo, exec_lo, s18
+; GFX11-FAKE16-NEXT: s_and_not1_b32 vcc_lo, exec_lo, s17
; GFX11-FAKE16-NEXT: s_cbranch_vccnz .LBB25_3
; GFX11-FAKE16-NEXT: .LBB25_2: ; %cmp.true
; GFX11-FAKE16-NEXT: s_add_i32 s1, s1, 3
@@ -3847,10 +3847,10 @@ define inreg <16 x i8> @bitcast_v4i32_to_v16i8_scalar(<4 x i32> inreg %a, i32 in
; GFX11-FAKE16-NEXT: s_lshr_b32 s14, s1, 16
; GFX11-FAKE16-NEXT: s_lshr_b32 s15, s1, 8
; GFX11-FAKE16-NEXT: s_lshr_b32 s16, s0, 16
-; GFX11-FAKE16-NEXT: s_lshr_b32 s17, s0, 8
+; GFX11-FAKE16-NEXT: s_lshr_b32 s18, s0, 8
; GFX11-FAKE16-NEXT: .LBB25_3: ; %end
; GFX11-FAKE16-NEXT: s_delay_alu instid0(SALU_CYCLE_1)
-; GFX11-FAKE16-NEXT: v_dual_mov_b32 v0, s0 :: v_dual_mov_b32 v1, s17
+; GFX11-FAKE16-NEXT: v_dual_mov_b32 v0, s0 :: v_dual_mov_b32 v1, s18
; GFX11-FAKE16-NEXT: v_dual_mov_b32 v2, s16 :: v_dual_mov_b32 v3, s6
; GFX11-FAKE16-NEXT: v_dual_mov_b32 v4, s1 :: v_dual_mov_b32 v5, s15
; GFX11-FAKE16-NEXT: v_dual_mov_b32 v6, s14 :: v_dual_mov_b32 v7, s13
@@ -3860,7 +3860,7 @@ define inreg <16 x i8> @bitcast_v4i32_to_v16i8_scalar(<4 x i32> inreg %a, i32 in
; GFX11-FAKE16-NEXT: v_dual_mov_b32 v14, s9 :: v_dual_mov_b32 v15, s8
; GFX11-FAKE16-NEXT: s_setpc_b64 s[30:31]
; GFX11-FAKE16-NEXT: .LBB25_4:
-; GFX11-FAKE16-NEXT: ; implicit-def: $sgpr17
+; GFX11-FAKE16-NEXT: ; implicit-def: $sgpr18
; GFX11-FAKE16-NEXT: ; implicit-def: $sgpr16
; GFX11-FAKE16-NEXT: ; implicit-def: $sgpr6
; GFX11-FAKE16-NEXT: ; implicit-def: $sgpr15
@@ -12472,7 +12472,7 @@ define inreg <16 x i8> @bitcast_v2i64_to_v16i8_scalar(<2 x i64> inreg %a, i32 in
; GFX11-TRUE16: ; %bb.0:
; GFX11-TRUE16-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX11-TRUE16-NEXT: s_cmp_lg_u32 s16, 0
-; GFX11-TRUE16-NEXT: s_mov_b32 s18, 0
+; GFX11-TRUE16-NEXT: s_mov_b32 s16, 0
; GFX11-TRUE16-NEXT: s_cbranch_scc0 .LBB69_4
; GFX11-TRUE16-NEXT: ; %bb.1: ; %cmp.false
; GFX11-TRUE16-NEXT: s_lshr_b32 s8, s3, 24
@@ -12483,11 +12483,11 @@ define inreg <16 x i8> @bitcast_v2i64_to_v16i8_scalar(<2 x i64> inreg %a, i32 in
; GFX11-TRUE16-NEXT: s_lshr_b32 s13, s1, 24
; GFX11-TRUE16-NEXT: s_lshr_b32 s14, s1, 16
; GFX11-TRUE16-NEXT: s_lshr_b32 s15, s1, 8
-; GFX11-TRUE16-NEXT: s_lshr_b32 s16, s0, 16
-; GFX11-TRUE16-NEXT: s_lshr_b32 s17, s0, 8
+; GFX11-TRUE16-NEXT: s_lshr_b32 s17, s0, 16
+; GFX11-TRUE16-NEXT: s_lshr_b32 s18, s0, 8
; GFX11-TRUE16-NEXT: s_lshr_b64 s[4:5], s[2:3], 24
; GFX11-TRUE16-NEXT: s_lshr_b64 s[6:7], s[0:1], 24
-; GFX11-TRUE16-NEXT: s_and_not1_b32 vcc_lo, exec_lo, s18
+; GFX11-TRUE16-NEXT: s_and_not1_b32 vcc_lo, exec_lo, s16
; GFX11-TRUE16-NEXT: s_cbranch_vccnz .LBB69_3
; GFX11-TRUE16-NEXT: .LBB69_2: ; %cmp.true
; GFX11-TRUE16-NEXT: s_add_u32 s0, s0, 3
@@ -12504,12 +12504,12 @@ define inreg <16 x i8> @bitcast_v2i64_to_v16i8_scalar(<2 x i64> inreg %a, i32 in
; GFX11-TRUE16-NEXT: s_lshr_b32 s13, s1, 24
; GFX11-TRUE16-NEXT: s_lshr_b32 s14, s1, 16
; GFX11-TRUE16-NEXT: s_lshr_b32 s15, s1, 8
-; GFX11-TRUE16-NEXT: s_lshr_b32 s16, s0, 16
-; GFX11-TRUE16-NEXT: s_lshr_b32 s17, s0, 8
+; GFX11-TRUE16-NEXT: s_lshr_b32 s17, s0, 16
+; GFX11-TRUE16-NEXT: s_lshr_b32 s18, s0, 8
; GFX11-TRUE16-NEXT: .LBB69_3: ; %end
; GFX11-TRUE16-NEXT: v_mov_b16_e32 v0.l, s0
-; GFX11-TRUE16-NEXT: v_mov_b16_e32 v1.l, s17
-; GFX11-TRUE16-NEXT: v_mov_b16_e32 v2.l, s16
+; GFX11-TRUE16-NEXT: v_mov_b16_e32 v1.l, s18
+; GFX11-TRUE16-NEXT: v_mov_b16_e32 v2.l, s17
; GFX11-TRUE16-NEXT: v_mov_b16_e32 v3.l, s6
; GFX11-TRUE16-NEXT: v_mov_b16_e32 v4.l, s1
; GFX11-TRUE16-NEXT: v_mov_b16_e32 v5.l, s15
@@ -12525,8 +12525,8 @@ define inreg <16 x i8> @bitcast_v2i64_to_v16i8_scalar(<2 x i64> inreg %a, i32 in
; GFX11-TRUE16-NEXT: v_mov_b16_e32 v15.l, s8
; GFX11-TRUE16-NEXT: s_setpc_b64 s[30:31]
; GFX11-TRUE16-NEXT: .LBB69_4:
+; GFX11-TRUE16-NEXT: ; implicit-def: $sgpr18_lo16
; GFX11-TRUE16-NEXT: ; implicit-def: $sgpr17_lo16
-; GFX11-TRUE16-NEXT: ; implicit-def: $sgpr16_lo16
; GFX11-TRUE16-NEXT: ; implicit-def: $sgpr6_lo16
; GFX11-TRUE16-NEXT: ; implicit-def: $sgpr15_lo16
; GFX11-TRUE16-NEXT: ; implicit-def: $sgpr14_lo16
@@ -12543,7 +12543,7 @@ define inreg <16 x i8> @bitcast_v2i64_to_v16i8_scalar(<2 x i64> inreg %a, i32 in
; GFX11-FAKE16: ; %bb.0:
; GFX11-FAKE16-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX11-FAKE16-NEXT: s_cmp_lg_u32 s16, 0
-; GFX11-FAKE16-NEXT: s_mov_b32 s18, 0
+; GFX11-FAKE16-NEXT: s_mov_b32 s17, 0
; GFX11-FAKE16-NEXT: s_cbranch_scc0 .LBB69_4
; GFX11-FAKE16-NEXT: ; %bb.1: ; %cmp.false
; GFX11-FAKE16-NEXT: s_lshr_b32 s8, s3, 24
@@ -12555,10 +12555,10 @@ define inreg <16 x i8> @bitcast_v2i64_to_v16i8_scalar(<2 x i64> inreg %a, i32 in
; GFX11-FAKE16-NEXT: s_lshr_b32 s14, s1, 16
; GFX11-FAKE16-NEXT: s_lshr_b32 s15, s1, 8
; GFX11-FAKE16-NEXT: s_lshr_b32 s16, s0, 16
-; GFX11-FAKE16-NEXT: s_lshr_b32 s17, s0, 8
+; GFX11-FAKE16-NEXT: s_lshr_b32 s18, s0, 8
; GFX11-FAKE16-NEXT: s_lshr_b64 s[4:5], s[2:3], 24
; GFX11-FAKE16-NEXT: s_lshr_b64 s[6:7], s[0:1], 24
-; GFX11-FAKE16-NEXT: s_and_not1_b32 vcc_lo, exec_lo, s18
+; GFX11-FAKE16-NEXT: s_and_not1_b32 vcc_lo, exec_lo, s17
; GFX11-FAKE16-NEXT: s_cbranch_vccnz .LBB69_3
; GFX11-FAKE16-NEXT: .LBB69_2: ; %cmp.true
; GFX11-FAKE16-NEXT: s_add_u32 s0, s0, 3
@@ -12576,10 +12576,10 @@ define inreg <16 x i8> @bitcast_v2i64_to_v16i8_scalar(<2 x i64> inreg %a, i32 in
; GFX11-FAKE16-NEXT: s_lshr_b32 s14, s1, 16
; GFX11-FAKE16-NEXT: s_lshr_b32 s15, s1, 8
; GFX11-FAKE16-NEXT: s_lshr_b32 s16, s0, 16
-; GFX11-FAKE16-NEXT: s_lshr_b32 s17, s0, 8
+; GFX11-FAKE16-NEXT: s_lshr_b32 s18, s0, 8
; GFX11-FAKE16-NEXT: .LBB69_3: ; %end
; GFX11-FAKE16-NEXT: s_delay_alu instid0(SALU_CYCLE_1)
-; GFX11-FAKE16-NEXT: v_dual_mov_b32 v0, s0 :: v_dual_mov_b32 v1, s17
+; GFX11-FAKE16-NEXT: v_dual_mov_b32 v0, s0 :: v_dual_mov_b32 v1, s18
; GFX11-FAKE16-NEXT: v_dual_mov_b32 v2, s16 :: v_dual_mov_b32 v3, s6
; GFX11-FAKE16-NEXT: v_dual_mov_b32 v4, s1 :: v_dual_mov_b32 v5, s15
; GFX11-FAKE16-NEXT: v_dual_mov_b32 v6, s14 :: v_dual_mov_b32 v7, s13
@@ -12589,7 +12589,7 @@ define inreg <16 x i8> @bitcast_v2i64_to_v16i8_scalar(<2 x i64> inreg %a, i32 in
; GFX11-FAKE16-NEXT: v_dual_mov_b32 v14, s9 :: v_dual_mov_b32 v15, s8
; GFX11-FAKE16-NEXT: s_setpc_b64 s[30:31]
; GFX11-FAKE16-NEXT: .LBB69_4:
-; GFX11-FAKE16-NEXT: ; implicit-def: $sgpr17
+; GFX11-FAKE16-NEXT: ; implicit-def: $sgpr18
; GFX11-FAKE16-NEXT: ; implicit-def: $sgpr16
; GFX11-FAKE16-NEXT: ; implicit-def: $sgpr6
; GFX11-FAKE16-NEXT: ; implicit-def: $sgpr15
@@ -19198,78 +19198,74 @@ define <16 x i8> @bitcast_v8i16_to_v16i8(<8 x i16> %a, i32 %b) {
; VI-LABEL: bitcast_v8i16_to_v16i8:
; VI: ; %bb.0:
; VI-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
-; VI-NEXT: v_mov_b32_e32 v19, v3
-; VI-NEXT: v_mov_b32_e32 v18, v2
+; VI-NEXT: v_mov_b32_e32 v17, v3
+; VI-NEXT: v_mov_b32_e32 v16, v2
+; VI-NEXT: v_mov_b32_e32 v19, v1
+; VI-NEXT: v_mov_b32_e32 v18, v0
; VI-NEXT: v_cmp_ne_u32_e32 vcc, 0, v4
-; VI-NEXT: v_lshrrev_b32_e32 v14, 16, v19
-; VI-NEXT: v_lshrrev_b32_e32 v10, 16, v18
-; VI-NEXT: v_lshrrev_b32_e32 v6, 16, v1
-; VI-NEXT: v_lshrrev_b32_e32 v2, 16, v0
-; VI-NEXT: ; implicit-def: $vgpr16
-; VI-NEXT: ; implicit-def: $vgpr20
+; VI-NEXT: v_lshrrev_b32_e32 v14, 16, v17
+; VI-NEXT: v_lshrrev_b32_e32 v10, 16, v16
+; VI-NEXT: v_lshrrev_b32_e32 v6, 16, v19
+; VI-NEXT: v_lshrrev_b32_e32 v2, 16, v18
+; VI-NEXT: ; implicit-def: $vgpr1
; VI-NEXT: ; implicit-def: $vgpr3
-; VI-NEXT: ; implicit-def: $vgpr17
; VI-NEXT: ; implicit-def: $vgpr5
; VI-NEXT: ; implicit-def: $vgpr7
-; VI-NEXT: ; implicit-def: $vgpr8
; VI-NEXT: ; implicit-def: $vgpr9
-; VI-NEXT: ; implicit-def: $vgpr21
+; VI-NEXT: ; implicit-def: $vgpr11
; VI-NEXT: ; implicit-def: $vgpr13
; VI-NEXT: ; implicit-def: $vgpr15
-; VI-NEXT: ; implicit-def: $vgpr11
; VI-NEXT: s_and_saveexec_b64 s[4:5], vcc
; VI-NEXT: s_xor_b64 s[4:5], exec, s[4:5]
; VI-NEXT: s_cbranch_execz .LBB96_2
; VI-NEXT: ; %bb.1: ; %cmp.false
-; VI-NEXT: v_lshrrev_b32_e32 v15, 24, v19
-; VI-NEXT: v_lshrrev_b32_e32 v13, 8, v19
-; VI-NEXT: v_lshrrev_b32_e32 v9, 8, v18
-; VI-NEXT: v_lshrrev_b32_e32 v7, 24, v1
-; VI-NEXT: v_lshrrev_b32_e32 v5, 8, v1
-; VI-NEXT: v_lshrrev_b32_e32 v20, 8, v0
-; VI-NEXT: v_lshrrev_b64 v[11:12], 24, v[18:19]
-; VI-NEXT: v_lshrrev_b64 v[3:4], 24, v[0:1]
-; VI-NEXT: v_mov_b32_e32 v16, v0
-; VI-NEXT: v_mov_b32_e32 v17, v1
-; VI-NEXT: v_mov_b32_e32 v8, v18
-; VI-NEXT: v_mov_b32_e32 v21, v19
-; VI-NEXT: ; implicit-def: $vgpr1
-; VI-NEXT: ; implicit-def: $vgpr19
+; VI-NEXT: v_lshrrev_b32_e32 v15, 24, v17
+; VI-NEXT: v_lshrrev_b32_e32 v13, 8, v17
+; VI-NEXT: v_lshrrev_b32_e32 v9, 8, v16
+; VI-NEXT: v_lshrrev_b32_e32 v7, 24, v19
+; VI-NEXT: v_lshrrev_b32_e32 v5, 8, v19
+; VI-NEXT: v_lshrrev_b64 v[11:12], 24, v[16:17]
+; VI-NEXT: v_lshrrev_b64 v[3:4], 24, v[18:19]
+; VI-NEXT: v_lshrrev_b32_e32 v1, 8, v18
; VI-NEXT: .LBB96_2: ; %Flow
; VI-NEXT: s_andn2_saveexec_b64 s[4:5], s[4:5]
; VI-NEXT: s_cbranch_execz .LBB96_4
; VI-NEXT: ; %bb.3: ; %cmp.true
; VI-NEXT: v_mov_b32_e32 v3, 3
-; VI-NEXT: v_add_u16_sdwa v6, v1, v3 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_1 src1_sel:DWORD
-; VI-NEXT: v_add_u16_sdwa v2, v0, v3 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_1 src1_sel:DWORD
-; VI-NEXT: v_add_u16_sdwa v14, v19, v3 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_1 src1_sel:DWORD
-; VI-NEXT: v_add_u16_sdwa v10, v18, v3 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_1 src1_sel:DWORD
-; VI-NEXT: v_add_u16_e32 v17, 3, v1
-; VI-NEXT: v_lshlrev_b32_e32 v1, 16, v6
-; VI-NEXT: v_add_u16_e32 v16, 3, v0
+; VI-NEXT: v_add_u16_sdwa v6, v19, v3 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_1 src1_sel:DWORD
+; VI-NEXT: v_add_u16_e32 v20, 3, v19
+; VI-NEXT: v_lshlrev_b32_e32 v0, 16, v6
+; VI-NEXT: v_add_u16_sdwa v2, v18, v3 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_1 src1_sel:DWORD
+; VI-NEXT: v_add_u16_sdwa v14, v17, v3 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_1 src1_sel:DWORD
+; VI-NEXT: v_add_u16_sdwa v10, v16, v3 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_1 src1_sel:DWORD
+; VI-NEXT: v_or_b32_e32 v1, v20, v0
+; VI-NEXT: v_add_u16_e32 v19, 3, v18
; VI-NEXT: v_lshlrev_b32_e32 v0, 16, v2
-; VI-NEXT: v_add_u16_e32 v21, 3, v19
+; VI-NEXT: v_add_u16_e32 v21, 3, v17
; VI-NEXT: v_lshlrev_b32_e32 v4, 16, v14
-; VI-NEXT: v_add_u16_e32 v8, 3, v18
+; VI-NEXT: v_add_u16_e32 v17, 3, v16
; VI-NEXT: v_lshlrev_b32_e32 v3, 16, v10
-; VI-NEXT: v_or_b32_e32 v1, v17, v1
-; VI-NEXT: v_or_b32_e32 v0, v16, v0
-; VI-NEXT: v_or_b32_e32 v19, v21, v4
-; VI-NEXT: v_or_b32_e32 v18, v8, v3
-; VI-NEXT: v_lshrrev_b64 v[11:12], 24, v[18:19]
+; VI-NEXT: v_or_b32_e32 v0, v19, v0
+; VI-NEXT: v_or_b32_e32 v8, v21, v4
+; VI-NEXT: v_or_b32_e32 v7, v17, v3
+; VI-NEXT: v_lshrrev_b64 v[11:12], 24, v[7:8]
; VI-NEXT: v_lshrrev_b64 v[3:4], 24, v[0:1]
-; VI-NEXT: v_lshrrev_b32_e32 v13, 8, v19
-; VI-NEXT: v_lshrrev_b32_e32 v9, 8, v18
+; VI-NEXT: v_lshrrev_b32_e32 v13, 8, v8
+; VI-NEXT: v_lshrrev_b32_e32 v9, 8, v7
; VI-NEXT: v_lshrrev_b32_e32 v5, 8, v1
-; VI-NEXT: v_lshrrev_b32_e32 v20, 8, v0
+; VI-NEXT: v_lshrrev_b32_e32 v1, 8, v0
; VI-NEXT: v_bfe_u32 v15, v14, 8, 8
; VI-NEXT: v_bfe_u32 v7, v6, 8, 8
+; VI-NEXT: v_mov_b32_e32 v18, v19
+; VI-NEXT: v_mov_b32_e32 v19, v20
+; VI-NEXT: v_mov_b32_e32 v16, v17
+; VI-NEXT: v_mov_b32_e32 v17, v21
; VI-NEXT: .LBB96_4: ; %end
; VI-NEXT: s_or_b64 exec, exec, s[4:5]
-; VI-NEXT: v_mov_b32_e32 v0, v16
-; VI-NEXT: v_mov_b32_e32 v1, v20
-; VI-NEXT: v_mov_b32_e32 v4, v17
-; VI-NEXT: v_mov_b32_e32 v12, v21
+; VI-NEXT: v_mov_b32_e32 v0, v18
+; VI-NEXT: v_mov_b32_e32 v4, v19
+; VI-NEXT: v_mov_b32_e32 v8, v16
+; VI-NEXT: v_mov_b32_e32 v12, v17
; VI-NEXT: s_setpc_b64 s[30:31]
;
; GFX9-LABEL: bitcast_v8i16_to_v16i8:
@@ -24323,13 +24319,13 @@ define <16 x i8> @bitcast_v8bf16_to_v16i8(<8 x bfloat> %a, i32 %b) {
; GFX11-TRUE16-NEXT: v_lshrrev_b32_e32 v7, 24, v3
; GFX11-TRUE16-NEXT: v_lshrrev_b32_e32 v5, 8, v3
; GFX11-TRUE16-NEXT: v_lshrrev_b32_e32 v1, 8, v2
-; GFX11-TRUE16-NEXT: v_lshrrev_b64 v[17:18], 24, v[2:3]
-; GFX11-TRUE16-NEXT: v_mov_b16_e32 v0.h, v2.l
-; GFX11-TRUE16-NEXT: v_mov_b16_e32 v4.h, v3.l
+; GFX11-TRUE16-NEXT: v_mov_b16_e32 v14.h, v11.h
; GFX11-TRUE16-NEXT: v_mov_b16_e32 v6.h, v3.h
-; GFX11-TRUE16-NEXT: v_mov_b16_e32 v8.h, v10.l
+; GFX11-TRUE16-NEXT: v_lshrrev_b64 v[17:18], 24, v[2:3]
; GFX11-TRUE16-NEXT: v_mov_b16_e32 v12.h, v11.l
-; GFX11-TRUE16-NEXT: v_mov_b16_e32 v14.h, v11.h
+; GFX11-TRUE16-NEXT: v_mov_b16_e32 v8.h, v10.l
+; GFX11-TRUE16-NEXT: v_mov_b16_e32 v4.h, v3.l
+; GFX11-TRUE16-NEXT: v_mov_b16_e32 v0.h, v2.l
; GFX11-TRUE16-NEXT: .LBB108_2: ; %Flow
; GFX11-TRUE16-NEXT: s_and_not1_saveexec_b32 s0, s0
; GFX11-TRUE16-NEXT: s_cbranch_execz .LBB108_4
@@ -24413,6 +24409,7 @@ define <16 x i8> @bitcast_v8bf16_to_v16i8(<8 x bfloat> %a, i32 %b) {
; GFX11-TRUE16-NEXT: v_lshrrev_b32_e32 v1, 8, v2
; GFX11-TRUE16-NEXT: .LBB108_4: ; %end
; GFX11-TRUE16-NEXT: s_or_b32 exec_lo, exec_lo, s0
+; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1)
; GFX11-TRUE16-NEXT: v_mov_b16_e32 v0.l, v0.h
; GFX11-TRUE16-NEXT: v_mov_b16_e32 v2.l, v2.h
; GFX11-TRUE16-NEXT: v_mov_b16_e32 v3.l, v17.l
@@ -25860,9 +25857,9 @@ define inreg <8 x bfloat> @bitcast_v16i8_to_v8bf16_scalar(<16 x i8> inreg %a, i3
; SI: ; %bb.0:
; SI-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; SI-NEXT: v_cmp_ne_u32_e32 vcc, 0, v2
-; SI-NEXT: v_readfirstlane_b32 s9, v1
+; SI-NEXT: v_readfirstlane_b32 s8, v1
; SI-NEXT: s_and_b64 s[4:5], vcc, exec
-; SI-NEXT: v_readfirstlane_b32 s11, v0
+; SI-NEXT: v_readfirstlane_b32 s9, v0
; SI-NEXT: s_cbranch_scc0 .LBB111_4
; SI-NEXT: ; %bb.1: ; %cmp.false
; SI-NEXT: s_and_b32 s4, s16, 0xff
@@ -25876,11 +25873,11 @@ define inreg <8 x bfloat> @bitcast_v16i8_to_v8bf16_scalar(<16 x i8> inreg %a, i3
; SI-NEXT: s_and_b32 s4, s20, 0xff
; SI-NEXT: s_lshl_b32 s5, s21, 8
; SI-NEXT: s_or_b32 s4, s4, s5
-; SI-NEXT: s_lshl_b32 s8, s4, 16
+; SI-NEXT: s_lshl_b32 s10, s4, 16
; SI-NEXT: s_and_b32 s4, s22, 0xff
; SI-NEXT: s_lshl_b32 s4, s4, 16
; SI-NEXT: s_lshl_b32 s5, s23, 24
-; SI-NEXT: s_or_b32 s10, s5, s4
+; SI-NEXT: s_or_b32 s11, s5, s4
; SI-NEXT: s_and_b32 s4, s24, 0xff
; SI-NEXT: s_lshl_b32 s4, s4, 16
; SI-NEXT: s_lshl_b32 s5, s25, 24
@@ -25893,20 +25890,20 @@ define inreg <8 x bfloat> @bitcast_v16i8_to_v8bf16_scalar(<16 x i8> inreg %a, i3
; SI-NEXT: s_lshl_b32 s5, s29, 8
; SI-NEXT: s_or_b32 s4, s4, s5
; SI-NEXT: s_lshl_b32 s14, s4, 16
-; SI-NEXT: s_and_b32 s4, s11, 0xff
+; SI-NEXT: s_and_b32 s4, s9, 0xff
; SI-NEXT: s_lshl_b32 s4, s4, 16
-; SI-NEXT: s_lshl_b32 s5, s9, 24
+; SI-NEXT: s_lshl_b32 s5, s8, 24
; SI-NEXT: s_or_b32 s15, s5, s4
; SI-NEXT: s_cbranch_execnz .LBB111_3
; SI-NEXT: .LBB111_2: ; %cmp.true
; SI-NEXT: s_add_i32 s28, s28, 3
; SI-NEXT: s_and_b32 s4, s28, 0xff
; SI-NEXT: s_lshl_b32 s5, s29, 8
-; SI-NEXT: s_add_i32 s11, s11, 3
+; SI-NEXT: s_add_i32 s9, s9, 3
; SI-NEXT: s_or_b32 s4, s5, s4
-; SI-NEXT: s_and_b32 s6, s11, 0xff
+; SI-NEXT: s_and_b32 s6, s9, 0xff
; SI-NEXT: s_addk_i32 s4, 0x300
-; SI-NEXT: s_lshl_b32 s5, s9, 24
+; SI-NEXT: s_lshl_b32 s5, s8, 24
; SI-NEXT: s_lshl_b32 s6, s6, 16
; SI-NEXT: s_and_b32 s4, s4, 0xffff
; SI-NEXT: s_or_b32 s5, s5, s6
@@ -25953,8 +25950,8 @@ define inreg <8 x bfloat> @bitcast_v16i8_to_v8bf16_scalar(<16 x i8> inreg %a, i3
; SI-NEXT: s_add_i32 s6, s6, 0x3000000
; SI-NEXT: s_and_b32 s7, s6, 0xffff0000
; SI-NEXT: s_lshl_b32 s6, s6, 16
-; SI-NEXT: s_and_b32 s10, s8, 0xffff0000
-; SI-NEXT: s_lshl_b32 s8, s8, 16
+; SI-NEXT: s_and_b32 s11, s8, 0xffff0000
+; SI-NEXT: s_lshl_b32 s10, s8, 16
; SI-NEXT: s_and_b32 s13, s5, 0xffff0000
; SI-NEXT: s_lshl_b32 s12, s5, 16
; SI-NEXT: s_and_b32 s15, s4, 0xffff0000
@@ -25962,8 +25959,8 @@ define inreg <8 x bfloat> @bitcast_v16i8_to_v8bf16_scalar(<16 x i8> inreg %a, i3
; SI-NEXT: .LBB111_3: ; %end
; SI-NEXT: v_mov_b32_e32 v0, s6
; SI-NEXT: v_mov_b32_e32 v1, s7
-; SI-NEXT: v_mov_b32_e32 v2, s8
-; SI-NEXT: v_mov_b32_e32 v3, s10
+; SI-NEXT: v_mov_b32_e32 v2, s10
+; SI-NEXT: v_mov_b32_e32 v3, s11
; SI-NEXT: v_mov_b32_e32 v4, s12
; SI-NEXT: v_mov_b32_e32 v5, s13
; SI-NEXT: v_mov_b32_e32 v6, s14
@@ -25972,8 +25969,8 @@ define inreg <8 x bfloat> @bitcast_v16i8_to_v8bf16_scalar(<16 x i8> inreg %a, i3
; SI-NEXT: .LBB111_4:
; SI-NEXT: ; implicit-def: $sgpr6
; SI-NEXT: ; implicit-def: $sgpr7
-; SI-NEXT: ; implicit-def: $sgpr8
; SI-NEXT: ; implicit-def: $sgpr10
+; SI-NEXT: ; implicit-def: $sgpr11
; SI-NEXT: ; implicit-def: $sgpr12
; SI-NEXT: ; implicit-def: $sgpr13
; SI-NEXT: ; implicit-def: $sgpr14
diff --git a/llvm/test/CodeGen/AMDGPU/amdgcn.bitcast.256bit.ll b/llvm/test/CodeGen/AMDGPU/amdgcn.bitcast.256bit.ll
index 8dc00701dcfd6..12c98939a2368 100644
--- a/llvm/test/CodeGen/AMDGPU/amdgcn.bitcast.256bit.ll
+++ b/llvm/test/CodeGen/AMDGPU/amdgcn.bitcast.256bit.ll
@@ -5622,7 +5622,7 @@ define inreg <32 x i8> @bitcast_v8i32_to_v32i8_scalar(<8 x i32> inreg %a, i32 in
; GFX11-TRUE16: ; %bb.0:
; GFX11-TRUE16-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX11-TRUE16-NEXT: s_cmp_lg_u32 s20, 0
-; GFX11-TRUE16-NEXT: s_mov_b32 s46, 0
+; GFX11-TRUE16-NEXT: s_mov_b32 s43, 0
; GFX11-TRUE16-NEXT: s_cbranch_scc0 .LBB25_4
; GFX11-TRUE16-NEXT: ; %bb.1: ; %cmp.false
; GFX11-TRUE16-NEXT: s_lshr_b32 s12, s19, 24
@@ -5642,14 +5642,14 @@ define inreg <32 x i8> @bitcast_v8i32_to_v32i8_scalar(<8 x i32> inreg %a, i32 in
; GFX11-TRUE16-NEXT: s_lshr_b32 s40, s2, 8
; GFX11-TRUE16-NEXT: s_lshr_b32 s41, s1, 24
; GFX11-TRUE16-NEXT: s_lshr_b32 s42, s1, 16
-; GFX11-TRUE16-NEXT: s_lshr_b32 s43, s1, 8
-; GFX11-TRUE16-NEXT: s_lshr_b32 s44, s0, 16
-; GFX11-TRUE16-NEXT: s_lshr_b32 s45, s0, 8
+; GFX11-TRUE16-NEXT: s_lshr_b32 s44, s1, 8
+; GFX11-TRUE16-NEXT: s_lshr_b32 s45, s0, 16
+; GFX11-TRUE16-NEXT: s_lshr_b32 s46, s0, 8
; GFX11-TRUE16-NEXT: s_lshr_b64 s[4:5], s[18:19], 24
; GFX11-TRUE16-NEXT: s_lshr_b64 s[6:7], s[16:17], 24
; GFX11-TRUE16-NEXT: s_lshr_b64 s[8:9], s[2:3], 24
; GFX11-TRUE16-NEXT: s_lshr_b64 s[10:11], s[0:1], 24
-; GFX11-TRUE16-NEXT: s_and_not1_b32 vcc_lo, exec_lo, s46
+; GFX11-TRUE16-NEXT: s_and_not1_b32 vcc_lo, exec_lo, s43
; GFX11-TRUE16-NEXT: s_cbranch_vccnz .LBB25_3
; GFX11-TRUE16-NEXT: .LBB25_2: ; %cmp.true
; GFX11-TRUE16-NEXT: s_add_i32 s1, s1, 3
@@ -5681,16 +5681,16 @@ define inreg <32 x i8> @bitcast_v8i32_to_v32i8_scalar(<8 x i32> inreg %a, i32 in
; GFX11-TRUE16-NEXT: s_lshr_b32 s40, s2, 8
; GFX11-TRUE16-NEXT: s_lshr_b32 s41, s1, 24
; GFX11-TRUE16-NEXT: s_lshr_b32 s42, s1, 16
-; GFX11-TRUE16-NEXT: s_lshr_b32 s43, s1, 8
-; GFX11-TRUE16-NEXT: s_lshr_b32 s44, s0, 16
-; GFX11-TRUE16-NEXT: s_lshr_b32 s45, s0, 8
+; GFX11-TRUE16-NEXT: s_lshr_b32 s44, s1, 8
+; GFX11-TRUE16-NEXT: s_lshr_b32 s45, s0, 16
+; GFX11-TRUE16-NEXT: s_lshr_b32 s46, s0, 8
; GFX11-TRUE16-NEXT: .LBB25_3: ; %end
; GFX11-TRUE16-NEXT: v_mov_b16_e32 v0.l, s0
-; GFX11-TRUE16-NEXT: v_mov_b16_e32 v1.l, s45
-; GFX11-TRUE16-NEXT: v_mov_b16_e32 v2.l, s44
+; GFX11-TRUE16-NEXT: v_mov_b16_e32 v1.l, s46
+; GFX11-TRUE16-NEXT: v_mov_b16_e32 v2.l, s45
; GFX11-TRUE16-NEXT: v_mov_b16_e32 v3.l, s10
; GFX11-TRUE16-NEXT: v_mov_b16_e32 v4.l, s1
-; GFX11-TRUE16-NEXT: v_mov_b16_e32 v5.l, s43
+; GFX11-TRUE16-NEXT: v_mov_b16_e32 v5.l, s44
; GFX11-TRUE16-NEXT: v_mov_b16_e32 v6.l, s42
; GFX11-TRUE16-NEXT: v_mov_b16_e32 v7.l, s41
; GFX11-TRUE16-NEXT: v_mov_b16_e32 v8.l, s2
@@ -5719,10 +5719,10 @@ define inreg <32 x i8> @bitcast_v8i32_to_v32i8_scalar(<8 x i32> inreg %a, i32 in
; GFX11-TRUE16-NEXT: v_mov_b16_e32 v31.l, s12
; GFX11-TRUE16-NEXT: s_setpc_b64 s[30:31]
; GFX11-TRUE16-NEXT: .LBB25_4:
+; GFX11-TRUE16-NEXT: ; implicit-def: $sgpr46_lo16
; GFX11-TRUE16-NEXT: ; implicit-def: $sgpr45_lo16
-; GFX11-TRUE16-NEXT: ; implicit-def: $sgpr44_lo16
; GFX11-TRUE16-NEXT: ; implicit-def: $sgpr10_lo16
-; GFX11-TRUE16-NEXT: ; implicit-def: $sgpr43_lo16
+; GFX11-TRUE16-NEXT: ; implicit-def: $sgpr44_lo16
; GFX11-TRUE16-NEXT: ; implicit-def: $sgpr42_lo16
; GFX11-TRUE16-NEXT: ; implicit-def: $sgpr41_lo16
; GFX11-TRUE16-NEXT: ; implicit-def: $sgpr40_lo16
@@ -5749,7 +5749,7 @@ define inreg <32 x i8> @bitcast_v8i32_to_v32i8_scalar(<8 x i32> inreg %a, i32 in
; GFX11-FAKE16: ; %bb.0:
; GFX11-FAKE16-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX11-FAKE16-NEXT: s_cmp_lg_u32 s20, 0
-; GFX11-FAKE16-NEXT: s_mov_b32 s46, 0
+; GFX11-FAKE16-NEXT: s_mov_b32 s44, 0
; GFX11-FAKE16-NEXT: s_cbranch_scc0 .LBB25_4
; GFX11-FAKE16-NEXT: ; %bb.1: ; %cmp.false
; GFX11-FAKE16-NEXT: s_lshr_b32 s12, s19, 24
@@ -5770,13 +5770,13 @@ define inreg <32 x i8> @bitcast_v8i32_to_v32i8_scalar(<8 x i32> inreg %a, i32 in
; GFX11-FAKE16-NEXT: s_lshr_b32 s41, s1, 24
; GFX11-FAKE16-NEXT: s_lshr_b32 s42, s1, 16
; GFX11-FAKE16-NEXT: s_lshr_b32 s43, s1, 8
-; GFX11-FAKE16-NEXT: s_lshr_b32 s44, s0, 16
-; GFX11-FAKE16-NEXT: s_lshr_b32 s45, s0, 8
+; GFX11-FAKE16-NEXT: s_lshr_b32 s45, s0, 16
+; GFX11-FAKE16-NEXT: s_lshr_b32 s46, s0, 8
; GFX11-FAKE16-NEXT: s_lshr_b64 s[4:5], s[18:19], 24
; GFX11-FAKE16-NEXT: s_lshr_b64 s[6:7], s[16:17], 24
; GFX11-FAKE16-NEXT: s_lshr_b64 s[8:9], s[2:3], 24
; GFX11-FAKE16-NEXT: s_lshr_b64 s[10:11], s[0:1], 24
-; GFX11-FAKE16-NEXT: s_and_not1_b32 vcc_lo, exec_lo, s46
+; GFX11-FAKE16-NEXT: s_and_not1_b32 vcc_lo, exec_lo, s44
; GFX11-FAKE16-NEXT: s_cbranch_vccnz .LBB25_3
; GFX11-FAKE16-NEXT: .LBB25_2: ; %cmp.true
; GFX11-FAKE16-NEXT: s_add_i32 s1, s1, 3
@@ -5809,12 +5809,12 @@ define inreg <32 x i8> @bitcast_v8i32_to_v32i8_scalar(<8 x i32> inreg %a, i32 in
; GFX11-FAKE16-NEXT: s_lshr_b32 s41, s1, 24
; GFX11-FAKE16-NEXT: s_lshr_b32 s42, s1, 16
; GFX11-FAKE16-NEXT: s_lshr_b32 s43, s1, 8
-; GFX11-FAKE16-NEXT: s_lshr_b32 s44, s0, 16
-; GFX11-FAKE16-NEXT: s_lshr_b32 s45, s0, 8
+; GFX11-FAKE16-NEXT: s_lshr_b32 s45, s0, 16
+; GFX11-FAKE16-NEXT: s_lshr_b32 s46, s0, 8
; GFX11-FAKE16-NEXT: .LBB25_3: ; %end
; GFX11-FAKE16-NEXT: s_delay_alu instid0(SALU_CYCLE_1)
-; GFX11-FAKE16-NEXT: v_dual_mov_b32 v0, s0 :: v_dual_mov_b32 v1, s45
-; GFX11-FAKE16-NEXT: v_dual_mov_b32 v2, s44 :: v_dual_mov_b32 v3, s10
+; GFX11-FAKE16-NEXT: v_dual_mov_b32 v0, s0 :: v_dual_mov_b32 v1, s46
+; GFX11-FAKE16-NEXT: v_dual_mov_b32 v2, s45 :: v_dual_mov_b32 v3, s10
; GFX11-FAKE16-NEXT: v_dual_mov_b32 v4, s1 :: v_dual_mov_b32 v5, s43
; GFX11-FAKE16-NEXT: v_dual_mov_b32 v6, s42 :: v_dual_mov_b32 v7, s41
; GFX11-FAKE16-NEXT: v_dual_mov_b32 v8, s2 :: v_dual_mov_b32 v9, s40
@@ -5831,8 +5831,8 @@ define inreg <32 x i8> @bitcast_v8i32_to_v32i8_scalar(<8 x i32> inreg %a, i32 in
; GFX11-FAKE16-NEXT: v_dual_mov_b32 v30, s13 :: v_dual_mov_b32 v31, s12
; GFX11-FAKE16-NEXT: s_setpc_b64 s[30:31]
; GFX11-FAKE16-NEXT: .LBB25_4:
+; GFX11-FAKE16-NEXT: ; implicit-def: $sgpr46
; GFX11-FAKE16-NEXT: ; implicit-def: $sgpr45
-; GFX11-FAKE16-NEXT: ; implicit-def: $sgpr44
; GFX11-FAKE16-NEXT: ; implicit-def: $sgpr10
; GFX11-FAKE16-NEXT: ; implicit-def: $sgpr43
; GFX11-FAKE16-NEXT: ; implicit-def: $sgpr42
@@ -19875,7 +19875,7 @@ define inreg <32 x i8> @bitcast_v4i64_to_v32i8_scalar(<4 x i64> inreg %a, i32 in
; GFX11-TRUE16: ; %bb.0:
; GFX11-TRUE16-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX11-TRUE16-NEXT: s_cmp_lg_u32 s20, 0
-; GFX11-TRUE16-NEXT: s_mov_b32 s46, 0
+; GFX11-TRUE16-NEXT: s_mov_b32 s43, 0
; GFX11-TRUE16-NEXT: s_cbranch_scc0 .LBB69_4
; GFX11-TRUE16-NEXT: ; %bb.1: ; %cmp.false
; GFX11-TRUE16-NEXT: s_lshr_b32 s12, s19, 24
@@ -19895,14 +19895,14 @@ define inreg <32 x i8> @bitcast_v4i64_to_v32i8_scalar(<4 x i64> inreg %a, i32 in
; GFX11-TRUE16-NEXT: s_lshr_b32 s40, s2, 8
; GFX11-TRUE16-NEXT: s_lshr_b32 s41, s1, 24
; GFX11-TRUE16-NEXT: s_lshr_b32 s42, s1, 16
-; GFX11-TRUE16-NEXT: s_lshr_b32 s43, s1, 8
-; GFX11-TRUE16-NEXT: s_lshr_b32 s44, s0, 16
-; GFX11-TRUE16-NEXT: s_lshr_b32 s45, s0, 8
+; GFX11-TRUE16-NEXT: s_lshr_b32 s44, s1, 8
+; GFX11-TRUE16-NEXT: s_lshr_b32 s45, s0, 16
+; GFX11-TRUE16-NEXT: s_lshr_b32 s46, s0, 8
; GFX11-TRUE16-NEXT: s_lshr_b64 s[4:5], s[18:19], 24
; GFX11-TRUE16-NEXT: s_lshr_b64 s[6:7], s[16:17], 24
; GFX11-TRUE16-NEXT: s_lshr_b64 s[8:9], s[2:3], 24
; GFX11-TRUE16-NEXT: s_lshr_b64 s[10:11], s[0:1], 24
-; GFX11-TRUE16-NEXT: s_and_not1_b32 vcc_lo, exec_lo, s46
+; GFX11-TRUE16-NEXT: s_and_not1_b32 vcc_lo, exec_lo, s43
; GFX11-TRUE16-NEXT: s_cbranch_vccnz .LBB69_3
; GFX11-TRUE16-NEXT: .LBB69_2: ; %cmp.true
; GFX11-TRUE16-NEXT: s_add_u32 s0, s0, 3
@@ -19934,16 +19934,16 @@ define inreg <32 x i8> @bitcast_v4i64_to_v32i8_scalar(<4 x i64> inreg %a, i32 in
; GFX11-TRUE16-NEXT: s_lshr_b32 s40, s2, 8
; GFX11-TRUE16-NEXT: s_lshr_b32 s41, s1, 24
; GFX11-TRUE16-NEXT: s_lshr_b32 s42, s1, 16
-; GFX11-TRUE16-NEXT: s_lshr_b32 s43, s1, 8
-; GFX11-TRUE16-NEXT: s_lshr_b32 s44, s0, 16
-; GFX11-TRUE16-NEXT: s_lshr_b32 s45, s0, 8
+; GFX11-TRUE16-NEXT: s_lshr_b32 s44, s1, 8
+; GFX11-TRUE16-NEXT: s_lshr_b32 s45, s0, 16
+; GFX11-TRUE16-NEXT: s_lshr_b32 s46, s0, 8
; GFX11-TRUE16-NEXT: .LBB69_3: ; %end
; GFX11-TRUE16-NEXT: v_mov_b16_e32 v0.l, s0
-; GFX11-TRUE16-NEXT: v_mov_b16_e32 v1.l, s45
-; GFX11-TRUE16-NEXT: v_mov_b16_e32 v2.l, s44
+; GFX11-TRUE16-NEXT: v_mov_b16_e32 v1.l, s46
+; GFX11-TRUE16-NEXT: v_mov_b16_e32 v2.l, s45
; GFX11-TRUE16-NEXT: v_mov_b16_e32 v3.l, s10
; GFX11-TRUE16-NEXT: v_mov_b16_e32 v4.l, s1
-; GFX11-TRUE16-NEXT: v_mov_b16_e32 v5.l, s43
+; GFX11-TRUE16-NEXT: v_mov_b16_e32 v5.l, s44
; GFX11-TRUE16-NEXT: v_mov_b16_e32 v6.l, s42
; GFX11-TRUE16-NEXT: v_mov_b16_e32 v7.l, s41
; GFX11-TRUE16-NEXT: v_mov_b16_e32 v8.l, s2
@@ -19972,10 +19972,10 @@ define inreg <32 x i8> @bitcast_v4i64_to_v32i8_scalar(<4 x i64> inreg %a, i32 in
; GFX11-TRUE16-NEXT: v_mov_b16_e32 v31.l, s12
; GFX11-TRUE16-NEXT: s_setpc_b64 s[30:31]
; GFX11-TRUE16-NEXT: .LBB69_4:
+; GFX11-TRUE16-NEXT: ; implicit-def: $sgpr46_lo16
; GFX11-TRUE16-NEXT: ; implicit-def: $sgpr45_lo16
-; GFX11-TRUE16-NEXT: ; implicit-def: $sgpr44_lo16
; GFX11-TRUE16-NEXT: ; implicit-def: $sgpr10_lo16
-; GFX11-TRUE16-NEXT: ; implicit-def: $sgpr43_lo16
+; GFX11-TRUE16-NEXT: ; implicit-def: $sgpr44_lo16
; GFX11-TRUE16-NEXT: ; implicit-def: $sgpr42_lo16
; GFX11-TRUE16-NEXT: ; implicit-def: $sgpr41_lo16
; GFX11-TRUE16-NEXT: ; implicit-def: $sgpr40_lo16
@@ -20002,7 +20002,7 @@ define inreg <32 x i8> @bitcast_v4i64_to_v32i8_scalar(<4 x i64> inreg %a, i32 in
; GFX11-FAKE16: ; %bb.0:
; GFX11-FAKE16-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX11-FAKE16-NEXT: s_cmp_lg_u32 s20, 0
-; GFX11-FAKE16-NEXT: s_mov_b32 s46, 0
+; GFX11-FAKE16-NEXT: s_mov_b32 s44, 0
; GFX11-FAKE16-NEXT: s_cbranch_scc0 .LBB69_4
; GFX11-FAKE16-NEXT: ; %bb.1: ; %cmp.false
; GFX11-FAKE16-NEXT: s_lshr_b32 s12, s19, 24
@@ -20023,13 +20023,13 @@ define inreg <32 x i8> @bitcast_v4i64_to_v32i8_scalar(<4 x i64> inreg %a, i32 in
; GFX11-FAKE16-NEXT: s_lshr_b32 s41, s1, 24
; GFX11-FAKE16-NEXT: s_lshr_b32 s42, s1, 16
; GFX11-FAKE16-NEXT: s_lshr_b32 s43, s1, 8
-; GFX11-FAKE16-NEXT: s_lshr_b32 s44, s0, 16
-; GFX11-FAKE16-NEXT: s_lshr_b32 s45, s0, 8
+; GFX11-FAKE16-NEXT: s_lshr_b32 s45, s0, 16
+; GFX11-FAKE16-NEXT: s_lshr_b32 s46, s0, 8
; GFX11-FAKE16-NEXT: s_lshr_b64 s[4:5], s[18:19], 24
; GFX11-FAKE16-NEXT: s_lshr_b64 s[6:7], s[16:17], 24
; GFX11-FAKE16-NEXT: s_lshr_b64 s[8:9], s[2:3], 24
; GFX11-FAKE16-NEXT: s_lshr_b64 s[10:11], s[0:1], 24
-; GFX11-FAKE16-NEXT: s_and_not1_b32 vcc_lo, exec_lo, s46
+; GFX11-FAKE16-NEXT: s_and_not1_b32 vcc_lo, exec_lo, s44
; GFX11-FAKE16-NEXT: s_cbranch_vccnz .LBB69_3
; GFX11-FAKE16-NEXT: .LBB69_2: ; %cmp.true
; GFX11-FAKE16-NEXT: s_add_u32 s0, s0, 3
@@ -20062,12 +20062,12 @@ define inreg <32 x i8> @bitcast_v4i64_to_v32i8_scalar(<4 x i64> inreg %a, i32 in
; GFX11-FAKE16-NEXT: s_lshr_b32 s41, s1, 24
; GFX11-FAKE16-NEXT: s_lshr_b32 s42, s1, 16
; GFX11-FAKE16-NEXT: s_lshr_b32 s43, s1, 8
-; GFX11-FAKE16-NEXT: s_lshr_b32 s44, s0, 16
-; GFX11-FAKE16-NEXT: s_lshr_b32 s45, s0, 8
+; GFX11-FAKE16-NEXT: s_lshr_b32 s45, s0, 16
+; GFX11-FAKE16-NEXT: s_lshr_b32 s46, s0, 8
; GFX11-FAKE16-NEXT: .LBB69_3: ; %end
; GFX11-FAKE16-NEXT: s_delay_alu instid0(SALU_CYCLE_1)
-; GFX11-FAKE16-NEXT: v_dual_mov_b32 v0, s0 :: v_dual_mov_b32 v1, s45
-; GFX11-FAKE16-NEXT: v_dual_mov_b32 v2, s44 :: v_dual_mov_b32 v3, s10
+; GFX11-FAKE16-NEXT: v_dual_mov_b32 v0, s0 :: v_dual_mov_b32 v1, s46
+; GFX11-FAKE16-NEXT: v_dual_mov_b32 v2, s45 :: v_dual_mov_b32 v3, s10
; GFX11-FAKE16-NEXT: v_dual_mov_b32 v4, s1 :: v_dual_mov_b32 v5, s43
; GFX11-FAKE16-NEXT: v_dual_mov_b32 v6, s42 :: v_dual_mov_b32 v7, s41
; GFX11-FAKE16-NEXT: v_dual_mov_b32 v8, s2 :: v_dual_mov_b32 v9, s40
@@ -20084,8 +20084,8 @@ define inreg <32 x i8> @bitcast_v4i64_to_v32i8_scalar(<4 x i64> inreg %a, i32 in
; GFX11-FAKE16-NEXT: v_dual_mov_b32 v30, s13 :: v_dual_mov_b32 v31, s12
; GFX11-FAKE16-NEXT: s_setpc_b64 s[30:31]
; GFX11-FAKE16-NEXT: .LBB69_4:
+; GFX11-FAKE16-NEXT: ; implicit-def: $sgpr46
; GFX11-FAKE16-NEXT: ; implicit-def: $sgpr45
-; GFX11-FAKE16-NEXT: ; implicit-def: $sgpr44
; GFX11-FAKE16-NEXT: ; implicit-def: $sgpr10
; GFX11-FAKE16-NEXT: ; implicit-def: $sgpr43
; GFX11-FAKE16-NEXT: ; implicit-def: $sgpr42
@@ -31362,36 +31362,32 @@ define <32 x i8> @bitcast_v16i16_to_v32i8(<16 x i16> %a, i32 %b) {
; VI-LABEL: bitcast_v16i16_to_v32i8:
; VI: ; %bb.0:
; VI-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; VI-NEXT: v_mov_b32_e32 v33, v5
+; VI-NEXT: v_mov_b32_e32 v32, v4
+; VI-NEXT: v_mov_b32_e32 v35, v3
+; VI-NEXT: v_mov_b32_e32 v34, v2
; VI-NEXT: v_cmp_ne_u32_e32 vcc, 0, v8
; VI-NEXT: v_lshrrev_b32_e32 v30, 16, v7
; VI-NEXT: v_lshrrev_b32_e32 v26, 16, v6
-; VI-NEXT: v_lshrrev_b32_e32 v22, 16, v5
-; VI-NEXT: v_lshrrev_b32_e32 v18, 16, v4
-; VI-NEXT: v_lshrrev_b32_e32 v14, 16, v3
-; VI-NEXT: v_lshrrev_b32_e32 v10, 16, v2
+; VI-NEXT: v_lshrrev_b32_e32 v22, 16, v33
+; VI-NEXT: v_lshrrev_b32_e32 v18, 16, v32
+; VI-NEXT: v_lshrrev_b32_e32 v14, 16, v35
+; VI-NEXT: v_lshrrev_b32_e32 v10, 16, v34
; VI-NEXT: v_lshrrev_b32_e32 v36, 16, v1
-; VI-NEXT: v_lshrrev_b32_e32 v32, 16, v0
-; VI-NEXT: ; implicit-def: $vgpr50
+; VI-NEXT: v_lshrrev_b32_e32 v2, 16, v0
; VI-NEXT: ; implicit-def: $vgpr38
-; VI-NEXT: ; implicit-def: $vgpr48
+; VI-NEXT: ; implicit-def: $vgpr5
; VI-NEXT: ; implicit-def: $vgpr37
-; VI-NEXT: ; implicit-def: $vgpr39
-; VI-NEXT: ; implicit-def: $vgpr8
; VI-NEXT: ; implicit-def: $vgpr9
-; VI-NEXT: ; implicit-def: $vgpr35
; VI-NEXT: ; implicit-def: $vgpr13
; VI-NEXT: ; implicit-def: $vgpr15
-; VI-NEXT: ; implicit-def: $vgpr16
; VI-NEXT: ; implicit-def: $vgpr17
-; VI-NEXT: ; implicit-def: $vgpr49
; VI-NEXT: ; implicit-def: $vgpr21
; VI-NEXT: ; implicit-def: $vgpr23
-; VI-NEXT: ; implicit-def: $vgpr24
; VI-NEXT: ; implicit-def: $vgpr25
-; VI-NEXT: ; implicit-def: $vgpr51
; VI-NEXT: ; implicit-def: $vgpr29
; VI-NEXT: ; implicit-def: $vgpr31
-; VI-NEXT: ; implicit-def: $vgpr33
+; VI-NEXT: ; implicit-def: $vgpr3
; VI-NEXT: ; implicit-def: $vgpr11
; VI-NEXT: ; implicit-def: $vgpr19
; VI-NEXT: ; implicit-def: $vgpr27
@@ -31402,97 +31398,92 @@ define <32 x i8> @bitcast_v16i16_to_v32i8(<16 x i16> %a, i32 %b) {
; VI-NEXT: v_lshrrev_b32_e32 v31, 24, v7
; VI-NEXT: v_lshrrev_b32_e32 v29, 8, v7
; VI-NEXT: v_lshrrev_b32_e32 v25, 8, v6
-; VI-NEXT: v_lshrrev_b32_e32 v23, 24, v5
-; VI-NEXT: v_lshrrev_b32_e32 v21, 8, v5
-; VI-NEXT: v_lshrrev_b32_e32 v17, 8, v4
-; VI-NEXT: v_lshrrev_b32_e32 v15, 24, v3
-; VI-NEXT: v_lshrrev_b32_e32 v13, 8, v3
-; VI-NEXT: v_lshrrev_b32_e32 v9, 8, v2
-; VI-NEXT: v_lshrrev_b32_e32 v39, 24, v1
-; VI-NEXT: v_lshrrev_b32_e32 v37, 8, v1
-; VI-NEXT: v_lshrrev_b32_e32 v38, 8, v0
+; VI-NEXT: v_lshrrev_b32_e32 v23, 24, v33
+; VI-NEXT: v_lshrrev_b32_e32 v21, 8, v33
+; VI-NEXT: v_lshrrev_b32_e32 v17, 8, v32
+; VI-NEXT: v_lshrrev_b32_e32 v15, 24, v35
+; VI-NEXT: v_lshrrev_b32_e32 v13, 8, v35
+; VI-NEXT: v_lshrrev_b32_e32 v9, 8, v34
+; VI-NEXT: v_lshrrev_b32_e32 v37, 24, v1
+; VI-NEXT: v_lshrrev_b32_e32 v5, 8, v1
; VI-NEXT: v_lshrrev_b64 v[27:28], 24, v[6:7]
-; VI-NEXT: v_lshrrev_b64 v[19:20], 24, v[4:5]
-; VI-NEXT: v_lshrrev_b64 v[11:12], 24, v[2:3]
-; VI-NEXT: v_lshrrev_b64 v[33:34], 24, v[0:1]
-; VI-NEXT: v_mov_b32_e32 v50, v0
-; VI-NEXT: v_mov_b32_e32 v48, v1
-; VI-NEXT: v_mov_b32_e32 v8, v2
-; VI-NEXT: v_mov_b32_e32 v35, v3
-; VI-NEXT: v_mov_b32_e32 v16, v4
-; VI-NEXT: v_mov_b32_e32 v49, v5
-; VI-NEXT: v_mov_b32_e32 v24, v6
-; VI-NEXT: v_mov_b32_e32 v51, v7
-; VI-NEXT: ; implicit-def: $vgpr1
-; VI-NEXT: ; implicit-def: $vgpr3
-; VI-NEXT: ; implicit-def: $vgpr5
-; VI-NEXT: ; implicit-def: $vgpr7
+; VI-NEXT: v_lshrrev_b64 v[19:20], 24, v[32:33]
+; VI-NEXT: v_lshrrev_b64 v[11:12], 24, v[34:35]
+; VI-NEXT: v_lshrrev_b64 v[3:4], 24, v[0:1]
+; VI-NEXT: v_lshrrev_b32_e32 v38, 8, v0
; VI-NEXT: .LBB96_2: ; %Flow
; VI-NEXT: s_andn2_saveexec_b64 s[4:5], s[4:5]
; VI-NEXT: s_cbranch_execz .LBB96_4
; VI-NEXT: ; %bb.3: ; %cmp.true
-; VI-NEXT: v_mov_b32_e32 v9, 3
-; VI-NEXT: v_add_u16_sdwa v36, v1, v9 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_1 src1_sel:DWORD
-; VI-NEXT: v_add_u16_sdwa v32, v0, v9 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_1 src1_sel:DWORD
-; VI-NEXT: v_add_u16_sdwa v14, v3, v9 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_1 src1_sel:DWORD
-; VI-NEXT: v_add_u16_sdwa v10, v2, v9 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_1 src1_sel:DWORD
-; VI-NEXT: v_add_u16_sdwa v22, v5, v9 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_1 src1_sel:DWORD
-; VI-NEXT: v_add_u16_sdwa v18, v4, v9 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_1 src1_sel:DWORD
-; VI-NEXT: v_add_u16_sdwa v30, v7, v9 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_1 src1_sel:DWORD
-; VI-NEXT: v_add_u16_sdwa v26, v6, v9 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_1 src1_sel:DWORD
-; VI-NEXT: v_add_u16_e32 v48, 3, v1
+; VI-NEXT: v_mov_b32_e32 v13, 3
+; VI-NEXT: v_add_u16_sdwa v2, v0, v13 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_1 src1_sel:DWORD
+; VI-NEXT: v_add_u16_e32 v15, 3, v0
+; VI-NEXT: v_lshlrev_b32_e32 v0, 16, v2
+; VI-NEXT: v_add_u16_sdwa v14, v35, v13 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_1 src1_sel:DWORD
+; VI-NEXT: v_or_b32_e32 v3, v15, v0
+; VI-NEXT: v_add_u16_e32 v16, 3, v35
+; VI-NEXT: v_lshlrev_b32_e32 v0, 16, v14
+; VI-NEXT: v_add_u16_sdwa v10, v34, v13 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_1 src1_sel:DWORD
+; VI-NEXT: v_or_b32_e32 v12, v16, v0
+; VI-NEXT: v_add_u16_e32 v17, 3, v34
+; VI-NEXT: v_lshlrev_b32_e32 v0, 16, v10
+; VI-NEXT: v_add_u16_sdwa v22, v33, v13 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_1 src1_sel:DWORD
+; VI-NEXT: v_or_b32_e32 v11, v17, v0
+; VI-NEXT: v_add_u16_e32 v19, 3, v33
+; VI-NEXT: v_lshlrev_b32_e32 v0, 16, v22
+; VI-NEXT: v_add_u16_sdwa v18, v32, v13 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_1 src1_sel:DWORD
+; VI-NEXT: v_add_u16_sdwa v36, v1, v13 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_1 src1_sel:DWORD
+; VI-NEXT: v_or_b32_e32 v9, v19, v0
+; VI-NEXT: v_add_u16_e32 v20, 3, v32
+; VI-NEXT: v_lshlrev_b32_e32 v0, 16, v18
+; VI-NEXT: v_add_u16_sdwa v30, v7, v13 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_1 src1_sel:DWORD
+; VI-NEXT: v_add_u16_e32 v5, 3, v1
; VI-NEXT: v_lshlrev_b32_e32 v1, 16, v36
-; VI-NEXT: v_add_u16_e32 v50, 3, v0
-; VI-NEXT: v_lshlrev_b32_e32 v0, 16, v32
-; VI-NEXT: v_add_u16_e32 v35, 3, v3
-; VI-NEXT: v_lshlrev_b32_e32 v3, 16, v14
-; VI-NEXT: v_add_u16_e32 v8, 3, v2
-; VI-NEXT: v_lshlrev_b32_e32 v2, 16, v10
-; VI-NEXT: v_add_u16_e32 v49, 3, v5
-; VI-NEXT: v_lshlrev_b32_e32 v5, 16, v22
-; VI-NEXT: v_add_u16_e32 v16, 3, v4
-; VI-NEXT: v_lshlrev_b32_e32 v4, 16, v18
-; VI-NEXT: v_add_u16_e32 v51, 3, v7
-; VI-NEXT: v_lshlrev_b32_e32 v7, 16, v30
-; VI-NEXT: v_add_u16_e32 v24, 3, v6
-; VI-NEXT: v_lshlrev_b32_e32 v6, 16, v26
-; VI-NEXT: v_or_b32_e32 v1, v48, v1
-; VI-NEXT: v_or_b32_e32 v0, v50, v0
-; VI-NEXT: v_or_b32_e32 v3, v35, v3
-; VI-NEXT: v_or_b32_e32 v2, v8, v2
-; VI-NEXT: v_or_b32_e32 v5, v49, v5
-; VI-NEXT: v_or_b32_e32 v4, v16, v4
-; VI-NEXT: v_or_b32_e32 v7, v51, v7
-; VI-NEXT: v_or_b32_e32 v6, v24, v6
-; VI-NEXT: v_lshrrev_b64 v[27:28], 24, v[6:7]
-; VI-NEXT: v_lshrrev_b64 v[19:20], 24, v[4:5]
-; VI-NEXT: v_lshrrev_b64 v[11:12], 24, v[2:3]
-; VI-NEXT: v_lshrrev_b64 v[33:34], 24, v[0:1]
-; VI-NEXT: v_lshrrev_b32_e32 v29, 8, v7
-; VI-NEXT: v_lshrrev_b32_e32 v25, 8, v6
-; VI-NEXT: v_lshrrev_b32_e32 v21, 8, v5
-; VI-NEXT: v_lshrrev_b32_e32 v17, 8, v4
-; VI-NEXT: v_lshrrev_b32_e32 v13, 8, v3
-; VI-NEXT: v_lshrrev_b32_e32 v9, 8, v2
-; VI-NEXT: v_lshrrev_b32_e32 v37, 8, v1
-; VI-NEXT: v_lshrrev_b32_e32 v38, 8, v0
+; VI-NEXT: v_or_b32_e32 v8, v20, v0
+; VI-NEXT: v_add_u16_e32 v23, 3, v7
+; VI-NEXT: v_lshlrev_b32_e32 v0, 16, v30
+; VI-NEXT: v_add_u16_sdwa v26, v6, v13 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_1 src1_sel:DWORD
+; VI-NEXT: v_or_b32_e32 v4, v5, v1
+; VI-NEXT: v_or_b32_e32 v1, v23, v0
+; VI-NEXT: v_add_u16_e32 v7, 3, v6
+; VI-NEXT: v_lshlrev_b32_e32 v0, 16, v26
+; VI-NEXT: v_or_b32_e32 v0, v7, v0
+; VI-NEXT: v_lshrrev_b64 v[27:28], 24, v[0:1]
+; VI-NEXT: v_lshrrev_b32_e32 v29, 8, v1
+; VI-NEXT: v_lshrrev_b32_e32 v21, 8, v9
+; VI-NEXT: v_mov_b32_e32 v1, v5
+; VI-NEXT: v_mov_b32_e32 v32, v20
+; VI-NEXT: v_mov_b32_e32 v33, v19
+; VI-NEXT: v_lshrrev_b64 v[19:20], 24, v[8:9]
+; VI-NEXT: v_lshrrev_b32_e32 v13, 8, v12
+; VI-NEXT: v_lshrrev_b32_e32 v9, 8, v11
+; VI-NEXT: v_lshrrev_b64 v[11:12], 24, v[11:12]
+; VI-NEXT: v_lshrrev_b32_e32 v5, 8, v4
+; VI-NEXT: v_lshrrev_b32_e32 v38, 8, v3
+; VI-NEXT: v_lshrrev_b64 v[3:4], 24, v[3:4]
+; VI-NEXT: v_lshrrev_b32_e32 v25, 8, v0
+; VI-NEXT: v_mov_b32_e32 v0, v15
+; VI-NEXT: v_mov_b32_e32 v34, v17
+; VI-NEXT: v_mov_b32_e32 v35, v16
+; VI-NEXT: v_mov_b32_e32 v6, v7
+; VI-NEXT: v_mov_b32_e32 v7, v23
+; VI-NEXT: v_lshrrev_b32_e32 v17, 8, v8
; VI-NEXT: v_bfe_u32 v31, v30, 8, 8
; VI-NEXT: v_bfe_u32 v23, v22, 8, 8
; VI-NEXT: v_bfe_u32 v15, v14, 8, 8
-; VI-NEXT: v_bfe_u32 v39, v36, 8, 8
+; VI-NEXT: v_bfe_u32 v37, v36, 8, 8
; VI-NEXT: .LBB96_4: ; %end
; VI-NEXT: s_or_b64 exec, exec, s[4:5]
-; VI-NEXT: v_mov_b32_e32 v0, v50
+; VI-NEXT: v_mov_b32_e32 v4, v1
+; VI-NEXT: v_mov_b32_e32 v8, v34
+; VI-NEXT: v_mov_b32_e32 v12, v35
+; VI-NEXT: v_mov_b32_e32 v16, v32
+; VI-NEXT: v_mov_b32_e32 v20, v33
+; VI-NEXT: v_mov_b32_e32 v24, v6
+; VI-NEXT: v_mov_b32_e32 v28, v7
; VI-NEXT: v_mov_b32_e32 v1, v38
-; VI-NEXT: v_mov_b32_e32 v2, v32
-; VI-NEXT: v_mov_b32_e32 v3, v33
-; VI-NEXT: v_mov_b32_e32 v4, v48
-; VI-NEXT: v_mov_b32_e32 v5, v37
; VI-NEXT: v_mov_b32_e32 v6, v36
-; VI-NEXT: v_mov_b32_e32 v7, v39
-; VI-NEXT: v_mov_b32_e32 v12, v35
-; VI-NEXT: v_mov_b32_e32 v20, v49
-; VI-NEXT: v_mov_b32_e32 v28, v51
+; VI-NEXT: v_mov_b32_e32 v7, v37
; VI-NEXT: s_setpc_b64 s[30:31]
;
; GFX9-LABEL: bitcast_v16i16_to_v32i8:
@@ -40362,19 +40353,19 @@ define <32 x i8> @bitcast_v16bf16_to_v32i8(<16 x bfloat> %a, i32 %b) {
; GFX11-TRUE16-NEXT: v_lshrrev_b32_e32 v7, 24, v3
; GFX11-TRUE16-NEXT: v_lshrrev_b32_e32 v5, 8, v3
; GFX11-TRUE16-NEXT: v_lshrrev_b32_e32 v1, 8, v2
-; GFX11-TRUE16-NEXT: v_lshrrev_b64 v[35:36], 24, v[2:3]
-; GFX11-TRUE16-NEXT: v_mov_b16_e32 v0.h, v2.l
-; GFX11-TRUE16-NEXT: v_mov_b16_e32 v4.h, v3.l
-; GFX11-TRUE16-NEXT: v_mov_b16_e32 v6.h, v3.h
-; GFX11-TRUE16-NEXT: v_mov_b16_e32 v8.h, v10.l
-; GFX11-TRUE16-NEXT: v_mov_b16_e32 v12.h, v11.l
-; GFX11-TRUE16-NEXT: v_mov_b16_e32 v14.h, v11.h
-; GFX11-TRUE16-NEXT: v_mov_b16_e32 v16.h, v18.l
-; GFX11-TRUE16-NEXT: v_mov_b16_e32 v20.h, v19.l
+; GFX11-TRUE16-NEXT: v_mov_b16_e32 v30.h, v27.h
; GFX11-TRUE16-NEXT: v_mov_b16_e32 v22.h, v19.h
-; GFX11-TRUE16-NEXT: v_mov_b16_e32 v24.h, v26.l
+; GFX11-TRUE16-NEXT: v_mov_b16_e32 v14.h, v11.h
+; GFX11-TRUE16-NEXT: v_mov_b16_e32 v6.h, v3.h
+; GFX11-TRUE16-NEXT: v_lshrrev_b64 v[35:36], 24, v[2:3]
; GFX11-TRUE16-NEXT: v_mov_b16_e32 v28.h, v27.l
-; GFX11-TRUE16-NEXT: v_mov_b16_e32 v30.h, v27.h
+; GFX11-TRUE16-NEXT: v_mov_b16_e32 v24.h, v26.l
+; GFX11-TRUE16-NEXT: v_mov_b16_e32 v20.h, v19.l
+; GFX11-TRUE16-NEXT: v_mov_b16_e32 v16.h, v18.l
+; GFX11-TRUE16-NEXT: v_mov_b16_e32 v12.h, v11.l
+; GFX11-TRUE16-NEXT: v_mov_b16_e32 v8.h, v10.l
+; GFX11-TRUE16-NEXT: v_mov_b16_e32 v4.h, v3.l
+; GFX11-TRUE16-NEXT: v_mov_b16_e32 v0.h, v2.l
; GFX11-TRUE16-NEXT: .LBB108_2: ; %Flow
; GFX11-TRUE16-NEXT: s_and_not1_saveexec_b32 s0, s0
; GFX11-TRUE16-NEXT: s_cbranch_execz .LBB108_4
@@ -40536,6 +40527,7 @@ define <32 x i8> @bitcast_v16bf16_to_v32i8(<16 x bfloat> %a, i32 %b) {
; GFX11-TRUE16-NEXT: v_lshrrev_b32_e32 v1, 8, v2
; GFX11-TRUE16-NEXT: .LBB108_4: ; %end
; GFX11-TRUE16-NEXT: s_or_b32 exec_lo, exec_lo, s0
+; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1)
; GFX11-TRUE16-NEXT: v_mov_b16_e32 v0.l, v0.h
; GFX11-TRUE16-NEXT: v_mov_b16_e32 v2.l, v2.h
; GFX11-TRUE16-NEXT: v_mov_b16_e32 v3.l, v35.l
@@ -43074,10 +43066,10 @@ define inreg <16 x bfloat> @bitcast_v32i8_to_v16bf16_scalar(<32 x i8> inreg %a,
; SI-NEXT: v_cmp_ne_u32_e32 vcc, 0, v18
; SI-NEXT: v_readfirstlane_b32 s42, v15
; SI-NEXT: v_readfirstlane_b32 s43, v14
-; SI-NEXT: v_readfirstlane_b32 s40, v7
-; SI-NEXT: v_readfirstlane_b32 s41, v6
-; SI-NEXT: v_readfirstlane_b32 s10, v1
-; SI-NEXT: v_readfirstlane_b32 s9, v0
+; SI-NEXT: v_readfirstlane_b32 s13, v7
+; SI-NEXT: v_readfirstlane_b32 s15, v6
+; SI-NEXT: v_readfirstlane_b32 s7, v1
+; SI-NEXT: v_readfirstlane_b32 s6, v0
; SI-NEXT: s_and_b64 s[4:5], vcc, exec
; SI-NEXT: v_lshlrev_b32_e32 v0, 24, v5
; SI-NEXT: v_lshlrev_b32_e32 v1, 24, v9
@@ -43088,15 +43080,15 @@ define inreg <16 x bfloat> @bitcast_v32i8_to_v16bf16_scalar(<32 x i8> inreg %a,
; SI-NEXT: s_and_b32 s4, s16, 0xff
; SI-NEXT: s_lshl_b32 s4, s4, 16
; SI-NEXT: s_lshl_b32 s5, s17, 24
-; SI-NEXT: s_or_b32 s6, s5, s4
+; SI-NEXT: s_or_b32 s8, s5, s4
; SI-NEXT: s_and_b32 s4, s18, 0xff
; SI-NEXT: s_lshl_b32 s4, s4, 16
; SI-NEXT: s_lshl_b32 s5, s19, 24
-; SI-NEXT: s_or_b32 s7, s5, s4
+; SI-NEXT: s_or_b32 s9, s5, s4
; SI-NEXT: s_and_b32 s4, s20, 0xff
; SI-NEXT: s_lshl_b32 s5, s21, 8
; SI-NEXT: s_or_b32 s4, s4, s5
-; SI-NEXT: s_lshl_b32 s8, s4, 16
+; SI-NEXT: s_lshl_b32 s10, s4, 16
; SI-NEXT: s_and_b32 s4, s22, 0xff
; SI-NEXT: s_lshl_b32 s4, s4, 16
; SI-NEXT: s_lshl_b32 s5, s23, 24
@@ -43114,24 +43106,24 @@ define inreg <16 x bfloat> @bitcast_v32i8_to_v16bf16_scalar(<32 x i8> inreg %a,
; SI-NEXT: s_lshl_b32 s4, s4, 16
; SI-NEXT: s_lshl_b32 s5, s27, 24
; SI-NEXT: v_lshlrev_b32_e32 v7, 16, v7
-; SI-NEXT: s_or_b32 s13, s5, s4
+; SI-NEXT: s_or_b32 s14, s5, s4
; SI-NEXT: s_and_b32 s4, s28, 0xff
; SI-NEXT: s_lshl_b32 s5, s29, 8
; SI-NEXT: v_or_b32_e32 v9, v0, v7
; SI-NEXT: v_and_b32_e32 v7, 0xff, v8
; SI-NEXT: s_or_b32 s4, s4, s5
; SI-NEXT: v_lshlrev_b32_e32 v7, 16, v7
-; SI-NEXT: s_lshl_b32 s14, s4, 16
-; SI-NEXT: s_and_b32 s4, s9, 0xff
+; SI-NEXT: s_lshl_b32 s40, s4, 16
+; SI-NEXT: s_and_b32 s4, s6, 0xff
; SI-NEXT: v_or_b32_e32 v19, v1, v7
; SI-NEXT: v_and_b32_e32 v7, 0xff, v10
; SI-NEXT: s_lshl_b32 s4, s4, 16
-; SI-NEXT: s_lshl_b32 s5, s10, 24
+; SI-NEXT: s_lshl_b32 s5, s7, 24
; SI-NEXT: v_lshlrev_b32_e32 v7, 16, v7
; SI-NEXT: v_lshlrev_b32_e32 v13, 24, v11
-; SI-NEXT: s_or_b32 s15, s5, s4
-; SI-NEXT: s_and_b32 s4, s41, 0xff
-; SI-NEXT: s_lshl_b32 s5, s40, 8
+; SI-NEXT: s_or_b32 s41, s5, s4
+; SI-NEXT: s_and_b32 s4, s15, 0xff
+; SI-NEXT: s_lshl_b32 s5, s13, 8
; SI-NEXT: v_or_b32_e32 v18, v13, v7
; SI-NEXT: v_and_b32_e32 v7, 0xff, v12
; SI-NEXT: s_or_b32 s4, s4, s5
@@ -43167,11 +43159,11 @@ define inreg <16 x bfloat> @bitcast_v32i8_to_v16bf16_scalar(<32 x i8> inreg %a,
; SI-NEXT: s_and_b32 s4, s4, 0xffff
; SI-NEXT: v_and_b32_e32 v7, 0xffff, v7
; SI-NEXT: v_or_b32_e32 v5, v5, v9
-; SI-NEXT: s_add_i32 s41, s41, 3
+; SI-NEXT: s_add_i32 s15, s15, 3
; SI-NEXT: v_or_b32_e32 v6, s4, v6
; SI-NEXT: v_or_b32_e32 v5, v5, v7
-; SI-NEXT: s_and_b32 s4, s41, 0xff
-; SI-NEXT: s_lshl_b32 s5, s40, 8
+; SI-NEXT: s_and_b32 s4, s15, 0xff
+; SI-NEXT: s_lshl_b32 s5, s13, 8
; SI-NEXT: v_add_i32_e32 v7, vcc, 3, v8
; SI-NEXT: s_or_b32 s4, s5, s4
; SI-NEXT: v_and_b32_e32 v7, 0xff, v7
@@ -43183,11 +43175,11 @@ define inreg <16 x bfloat> @bitcast_v32i8_to_v16bf16_scalar(<32 x i8> inreg %a,
; SI-NEXT: v_or_b32_e32 v1, s4, v1
; SI-NEXT: s_and_b32 s4, s28, 0xff
; SI-NEXT: s_lshl_b32 s5, s29, 8
-; SI-NEXT: s_add_i32 s9, s9, 3
+; SI-NEXT: s_add_i32 s6, s6, 3
; SI-NEXT: s_or_b32 s4, s5, s4
-; SI-NEXT: s_and_b32 s6, s9, 0xff
+; SI-NEXT: s_and_b32 s6, s6, 0xff
; SI-NEXT: s_addk_i32 s4, 0x300
-; SI-NEXT: s_lshl_b32 s5, s10, 24
+; SI-NEXT: s_lshl_b32 s5, s7, 24
; SI-NEXT: s_lshl_b32 s6, s6, 16
; SI-NEXT: s_and_b32 s4, s4, 0xffff
; SI-NEXT: s_or_b32 s5, s5, s6
@@ -43210,36 +43202,35 @@ define inreg <16 x bfloat> @bitcast_v32i8_to_v16bf16_scalar(<32 x i8> inreg %a,
; SI-NEXT: s_add_i32 s22, s22, 3
; SI-NEXT: s_or_b32 s6, s7, s6
; SI-NEXT: s_and_b32 s8, s22, 0xff
+; SI-NEXT: v_add_i32_e32 v2, vcc, 3, v2
; SI-NEXT: s_addk_i32 s6, 0x300
; SI-NEXT: s_lshl_b32 s7, s23, 24
; SI-NEXT: s_lshl_b32 s8, s8, 16
-; SI-NEXT: v_add_i32_e32 v2, vcc, 3, v2
-; SI-NEXT: s_and_b32 s6, s6, 0xffff
-; SI-NEXT: s_or_b32 s7, s7, s8
; SI-NEXT: v_and_b32_e32 v2, 0xff, v2
; SI-NEXT: v_lshlrev_b32_e32 v3, 8, v3
-; SI-NEXT: s_or_b32 s6, s7, s6
+; SI-NEXT: s_and_b32 s6, s6, 0xffff
+; SI-NEXT: s_or_b32 s7, s7, s8
; SI-NEXT: s_add_i32 s16, s16, 3
; SI-NEXT: v_or_b32_e32 v2, v3, v2
; SI-NEXT: v_add_i32_e32 v3, vcc, 3, v4
-; SI-NEXT: s_add_i32 s8, s6, 0x3000000
-; SI-NEXT: s_and_b32 s6, s16, 0xff
-; SI-NEXT: s_lshl_b32 s7, s17, 8
+; SI-NEXT: s_or_b32 s6, s7, s6
+; SI-NEXT: s_and_b32 s7, s16, 0xff
+; SI-NEXT: s_lshl_b32 s8, s17, 8
; SI-NEXT: s_add_i32 s18, s18, 3
; SI-NEXT: v_and_b32_e32 v3, 0xff, v3
-; SI-NEXT: s_or_b32 s6, s7, s6
+; SI-NEXT: s_or_b32 s7, s8, s7
; SI-NEXT: s_and_b32 s9, s18, 0xff
; SI-NEXT: v_add_i32_e32 v2, vcc, 0x300, v2
; SI-NEXT: v_lshlrev_b32_e32 v3, 16, v3
-; SI-NEXT: s_addk_i32 s6, 0x300
-; SI-NEXT: s_lshl_b32 s7, s19, 24
+; SI-NEXT: s_addk_i32 s7, 0x300
+; SI-NEXT: s_lshl_b32 s8, s19, 24
; SI-NEXT: s_lshl_b32 s9, s9, 16
; SI-NEXT: v_and_b32_e32 v2, 0xffff, v2
; SI-NEXT: v_or_b32_e32 v0, v0, v3
-; SI-NEXT: s_and_b32 s6, s6, 0xffff
-; SI-NEXT: s_or_b32 s7, s7, s9
+; SI-NEXT: s_and_b32 s7, s7, 0xffff
+; SI-NEXT: s_or_b32 s8, s8, s9
; SI-NEXT: v_or_b32_e32 v0, v0, v2
-; SI-NEXT: s_or_b32 s6, s7, s6
+; SI-NEXT: s_or_b32 s7, s8, s7
; SI-NEXT: v_add_i32_e32 v6, vcc, 0x3000000, v6
; SI-NEXT: v_add_i32_e32 v5, vcc, 0x3000000, v5
; SI-NEXT: v_add_i32_e32 v1, vcc, 0x3000000, v1
@@ -43247,14 +43238,15 @@ define inreg <16 x bfloat> @bitcast_v32i8_to_v16bf16_scalar(<32 x i8> inreg %a,
; SI-NEXT: s_add_i32 s4, s4, 0x3000000
; SI-NEXT: s_add_i32 s5, s5, 0x3000000
; SI-NEXT: s_add_i32 s6, s6, 0x3000000
-; SI-NEXT: s_and_b32 s7, s6, 0xffff0000
-; SI-NEXT: s_lshl_b32 s6, s6, 16
-; SI-NEXT: s_and_b32 s11, s8, 0xffff0000
-; SI-NEXT: s_lshl_b32 s8, s8, 16
-; SI-NEXT: s_and_b32 s13, s5, 0xffff0000
+; SI-NEXT: s_add_i32 s7, s7, 0x3000000
+; SI-NEXT: s_and_b32 s9, s7, 0xffff0000
+; SI-NEXT: s_lshl_b32 s8, s7, 16
+; SI-NEXT: s_and_b32 s11, s6, 0xffff0000
+; SI-NEXT: s_lshl_b32 s10, s6, 16
+; SI-NEXT: s_and_b32 s14, s5, 0xffff0000
; SI-NEXT: s_lshl_b32 s12, s5, 16
-; SI-NEXT: s_and_b32 s15, s4, 0xffff0000
-; SI-NEXT: s_lshl_b32 s14, s4, 16
+; SI-NEXT: s_and_b32 s41, s4, 0xffff0000
+; SI-NEXT: s_lshl_b32 s40, s4, 16
; SI-NEXT: v_and_b32_e32 v9, 0xffff0000, v0
; SI-NEXT: v_lshlrev_b32_e32 v17, 16, v0
; SI-NEXT: v_and_b32_e32 v19, 0xffff0000, v1
@@ -43265,14 +43257,14 @@ define inreg <16 x bfloat> @bitcast_v32i8_to_v16bf16_scalar(<32 x i8> inreg %a,
; SI-NEXT: v_lshlrev_b32_e32 v14, 16, v6
; SI-NEXT: s_branch .LBB111_5
; SI-NEXT: .LBB111_3:
-; SI-NEXT: ; implicit-def: $sgpr6
-; SI-NEXT: ; implicit-def: $sgpr7
; SI-NEXT: ; implicit-def: $sgpr8
+; SI-NEXT: ; implicit-def: $sgpr9
+; SI-NEXT: ; implicit-def: $sgpr10
; SI-NEXT: ; implicit-def: $sgpr11
; SI-NEXT: ; implicit-def: $sgpr12
-; SI-NEXT: ; implicit-def: $sgpr13
; SI-NEXT: ; implicit-def: $sgpr14
-; SI-NEXT: ; implicit-def: $sgpr15
+; SI-NEXT: ; implicit-def: $sgpr40
+; SI-NEXT: ; implicit-def: $sgpr41
; SI-NEXT: ; implicit-def: $vgpr17
; SI-NEXT: ; implicit-def: $vgpr9
; SI-NEXT: ; implicit-def: $sgpr44
@@ -43286,14 +43278,14 @@ define inreg <16 x bfloat> @bitcast_v32i8_to_v16bf16_scalar(<32 x i8> inreg %a,
; SI-NEXT: v_mov_b32_e32 v10, s44
; SI-NEXT: v_mov_b32_e32 v14, s45
; SI-NEXT: .LBB111_5: ; %end
-; SI-NEXT: v_mov_b32_e32 v0, s6
-; SI-NEXT: v_mov_b32_e32 v1, s7
-; SI-NEXT: v_mov_b32_e32 v2, s8
+; SI-NEXT: v_mov_b32_e32 v0, s8
+; SI-NEXT: v_mov_b32_e32 v1, s9
+; SI-NEXT: v_mov_b32_e32 v2, s10
; SI-NEXT: v_mov_b32_e32 v3, s11
; SI-NEXT: v_mov_b32_e32 v4, s12
-; SI-NEXT: v_mov_b32_e32 v5, s13
-; SI-NEXT: v_mov_b32_e32 v6, s14
-; SI-NEXT: v_mov_b32_e32 v7, s15
+; SI-NEXT: v_mov_b32_e32 v5, s14
+; SI-NEXT: v_mov_b32_e32 v6, s40
+; SI-NEXT: v_mov_b32_e32 v7, s41
; SI-NEXT: v_mov_b32_e32 v8, v17
; SI-NEXT: v_mov_b32_e32 v11, v19
; SI-NEXT: v_mov_b32_e32 v12, v18
diff --git a/llvm/test/CodeGen/AMDGPU/amdgcn.bitcast.320bit.ll b/llvm/test/CodeGen/AMDGPU/amdgcn.bitcast.320bit.ll
index e5a1260fa4538..5b82535bf956a 100644
--- a/llvm/test/CodeGen/AMDGPU/amdgcn.bitcast.320bit.ll
+++ b/llvm/test/CodeGen/AMDGPU/amdgcn.bitcast.320bit.ll
@@ -4076,7 +4076,7 @@ define inreg <40 x i8> @bitcast_v10i32_to_v40i8_scalar(<10 x i32> inreg %a, i32
; GFX11-TRUE16: ; %bb.0:
; GFX11-TRUE16-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX11-TRUE16-NEXT: s_cmp_lg_u32 s22, 0
-; GFX11-TRUE16-NEXT: s_mov_b32 s63, 0
+; GFX11-TRUE16-NEXT: s_mov_b32 s61, 0
; GFX11-TRUE16-NEXT: s_cbranch_scc0 .LBB13_4
; GFX11-TRUE16-NEXT: ; %bb.1: ; %cmp.false
; GFX11-TRUE16-NEXT: s_lshr_b32 s14, s21, 24
@@ -4102,14 +4102,14 @@ define inreg <40 x i8> @bitcast_v10i32_to_v40i8_scalar(<10 x i32> inreg %a, i32
; GFX11-TRUE16-NEXT: s_lshr_b32 s58, s1, 24
; GFX11-TRUE16-NEXT: s_lshr_b32 s59, s1, 16
; GFX11-TRUE16-NEXT: s_lshr_b32 s60, s1, 8
-; GFX11-TRUE16-NEXT: s_lshr_b32 s61, s0, 16
-; GFX11-TRUE16-NEXT: s_lshr_b32 s62, s0, 8
+; GFX11-TRUE16-NEXT: s_lshr_b32 s62, s0, 16
+; GFX11-TRUE16-NEXT: s_lshr_b32 s63, s0, 8
; GFX11-TRUE16-NEXT: s_lshr_b64 s[4:5], s[20:21], 24
; GFX11-TRUE16-NEXT: s_lshr_b64 s[6:7], s[18:19], 24
; GFX11-TRUE16-NEXT: s_lshr_b64 s[8:9], s[16:17], 24
; GFX11-TRUE16-NEXT: s_lshr_b64 s[10:11], s[2:3], 24
; GFX11-TRUE16-NEXT: s_lshr_b64 s[12:13], s[0:1], 24
-; GFX11-TRUE16-NEXT: s_and_not1_b32 vcc_lo, exec_lo, s63
+; GFX11-TRUE16-NEXT: s_and_not1_b32 vcc_lo, exec_lo, s61
; GFX11-TRUE16-NEXT: s_cbranch_vccnz .LBB13_3
; GFX11-TRUE16-NEXT: .LBB13_2: ; %cmp.true
; GFX11-TRUE16-NEXT: s_add_i32 s1, s1, 3
@@ -4150,12 +4150,12 @@ define inreg <40 x i8> @bitcast_v10i32_to_v40i8_scalar(<10 x i32> inreg %a, i32
; GFX11-TRUE16-NEXT: s_lshr_b32 s58, s1, 24
; GFX11-TRUE16-NEXT: s_lshr_b32 s59, s1, 16
; GFX11-TRUE16-NEXT: s_lshr_b32 s60, s1, 8
-; GFX11-TRUE16-NEXT: s_lshr_b32 s61, s0, 16
-; GFX11-TRUE16-NEXT: s_lshr_b32 s62, s0, 8
+; GFX11-TRUE16-NEXT: s_lshr_b32 s62, s0, 16
+; GFX11-TRUE16-NEXT: s_lshr_b32 s63, s0, 8
; GFX11-TRUE16-NEXT: .LBB13_3: ; %end
; GFX11-TRUE16-NEXT: s_delay_alu instid0(SALU_CYCLE_1)
-; GFX11-TRUE16-NEXT: s_mov_b32 s5, s62
-; GFX11-TRUE16-NEXT: s_mov_b32 s7, s61
+; GFX11-TRUE16-NEXT: s_mov_b32 s5, s63
+; GFX11-TRUE16-NEXT: s_mov_b32 s7, s62
; GFX11-TRUE16-NEXT: s_mov_b32 s9, s12
; GFX11-TRUE16-NEXT: s_and_b32 s0, s0, 0xff
; GFX11-TRUE16-NEXT: s_lshl_b32 s5, s5, 8
@@ -4291,8 +4291,8 @@ define inreg <40 x i8> @bitcast_v10i32_to_v40i8_scalar(<10 x i32> inreg %a, i32
; GFX11-TRUE16-NEXT: scratch_store_b64 v0, v[9:10], off offset:32
; GFX11-TRUE16-NEXT: s_setpc_b64 s[30:31]
; GFX11-TRUE16-NEXT: .LBB13_4:
+; GFX11-TRUE16-NEXT: ; implicit-def: $sgpr63_lo16
; GFX11-TRUE16-NEXT: ; implicit-def: $sgpr62_lo16
-; GFX11-TRUE16-NEXT: ; implicit-def: $sgpr61_lo16
; GFX11-TRUE16-NEXT: ; implicit-def: $sgpr12_lo16
; GFX11-TRUE16-NEXT: ; implicit-def: $sgpr60_lo16
; GFX11-TRUE16-NEXT: ; implicit-def: $sgpr59_lo16
@@ -4327,7 +4327,7 @@ define inreg <40 x i8> @bitcast_v10i32_to_v40i8_scalar(<10 x i32> inreg %a, i32
; GFX11-FAKE16: ; %bb.0:
; GFX11-FAKE16-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX11-FAKE16-NEXT: s_cmp_lg_u32 s22, 0
-; GFX11-FAKE16-NEXT: s_mov_b32 s63, 0
+; GFX11-FAKE16-NEXT: s_mov_b32 s61, 0
; GFX11-FAKE16-NEXT: s_cbranch_scc0 .LBB13_4
; GFX11-FAKE16-NEXT: ; %bb.1: ; %cmp.false
; GFX11-FAKE16-NEXT: s_lshr_b32 s14, s21, 24
@@ -4353,14 +4353,14 @@ define inreg <40 x i8> @bitcast_v10i32_to_v40i8_scalar(<10 x i32> inreg %a, i32
; GFX11-FAKE16-NEXT: s_lshr_b32 s58, s1, 24
; GFX11-FAKE16-NEXT: s_lshr_b32 s59, s1, 16
; GFX11-FAKE16-NEXT: s_lshr_b32 s60, s1, 8
-; GFX11-FAKE16-NEXT: s_lshr_b32 s61, s0, 16
-; GFX11-FAKE16-NEXT: s_lshr_b32 s62, s0, 8
+; GFX11-FAKE16-NEXT: s_lshr_b32 s62, s0, 16
+; GFX11-FAKE16-NEXT: s_lshr_b32 s63, s0, 8
; GFX11-FAKE16-NEXT: s_lshr_b64 s[4:5], s[20:21], 24
; GFX11-FAKE16-NEXT: s_lshr_b64 s[6:7], s[18:19], 24
; GFX11-FAKE16-NEXT: s_lshr_b64 s[8:9], s[16:17], 24
; GFX11-FAKE16-NEXT: s_lshr_b64 s[10:11], s[2:3], 24
; GFX11-FAKE16-NEXT: s_lshr_b64 s[12:13], s[0:1], 24
-; GFX11-FAKE16-NEXT: s_and_not1_b32 vcc_lo, exec_lo, s63
+; GFX11-FAKE16-NEXT: s_and_not1_b32 vcc_lo, exec_lo, s61
; GFX11-FAKE16-NEXT: s_cbranch_vccnz .LBB13_3
; GFX11-FAKE16-NEXT: .LBB13_2: ; %cmp.true
; GFX11-FAKE16-NEXT: s_add_i32 s1, s1, 3
@@ -4401,12 +4401,12 @@ define inreg <40 x i8> @bitcast_v10i32_to_v40i8_scalar(<10 x i32> inreg %a, i32
; GFX11-FAKE16-NEXT: s_lshr_b32 s58, s1, 24
; GFX11-FAKE16-NEXT: s_lshr_b32 s59, s1, 16
; GFX11-FAKE16-NEXT: s_lshr_b32 s60, s1, 8
-; GFX11-FAKE16-NEXT: s_lshr_b32 s61, s0, 16
-; GFX11-FAKE16-NEXT: s_lshr_b32 s62, s0, 8
+; GFX11-FAKE16-NEXT: s_lshr_b32 s62, s0, 16
+; GFX11-FAKE16-NEXT: s_lshr_b32 s63, s0, 8
; GFX11-FAKE16-NEXT: .LBB13_3: ; %end
; GFX11-FAKE16-NEXT: s_and_b32 s0, s0, 0xff
-; GFX11-FAKE16-NEXT: s_lshl_b32 s5, s62, 8
-; GFX11-FAKE16-NEXT: s_and_b32 s7, s61, 0xff
+; GFX11-FAKE16-NEXT: s_lshl_b32 s5, s63, 8
+; GFX11-FAKE16-NEXT: s_and_b32 s7, s62, 0xff
; GFX11-FAKE16-NEXT: s_lshl_b32 s9, s12, 8
; GFX11-FAKE16-NEXT: s_or_b32 s0, s0, s5
; GFX11-FAKE16-NEXT: s_or_b32 s5, s7, s9
@@ -4506,8 +4506,8 @@ define inreg <40 x i8> @bitcast_v10i32_to_v40i8_scalar(<10 x i32> inreg %a, i32
; GFX11-FAKE16-NEXT: scratch_store_b64 v0, v[9:10], off offset:32
; GFX11-FAKE16-NEXT: s_setpc_b64 s[30:31]
; GFX11-FAKE16-NEXT: .LBB13_4:
+; GFX11-FAKE16-NEXT: ; implicit-def: $sgpr63
; GFX11-FAKE16-NEXT: ; implicit-def: $sgpr62
-; GFX11-FAKE16-NEXT: ; implicit-def: $sgpr61
; GFX11-FAKE16-NEXT: ; implicit-def: $sgpr12
; GFX11-FAKE16-NEXT: ; implicit-def: $sgpr60
; GFX11-FAKE16-NEXT: ; implicit-def: $sgpr59
@@ -16547,10 +16547,9 @@ define <40 x i8> @bitcast_v20i16_to_v40i8(<20 x i16> %a, i32 %b) {
; SI-NEXT: buffer_store_dword v61, off, s[0:3], s32 offset:8 ; 4-byte Folded Spill
; SI-NEXT: buffer_store_dword v62, off, s[0:3], s32 offset:4 ; 4-byte Folded Spill
; SI-NEXT: buffer_store_dword v63, off, s[0:3], s32 ; 4-byte Folded Spill
+; SI-NEXT: v_cmp_ne_u32_e32 vcc, 0, v21
; SI-NEXT: s_waitcnt expcnt(6)
; SI-NEXT: v_lshlrev_b32_e32 v57, 16, v2
-; SI-NEXT: ; implicit-def: $vgpr2
-; SI-NEXT: v_cmp_ne_u32_e32 vcc, 0, v21
; SI-NEXT: v_lshlrev_b32_e32 v56, 16, v4
; SI-NEXT: s_waitcnt expcnt(4)
; SI-NEXT: v_lshlrev_b32_e32 v59, 16, v6
@@ -16563,47 +16562,48 @@ define <40 x i8> @bitcast_v20i16_to_v40i8(<20 x i16> %a, i32 %b) {
; SI-NEXT: v_lshlrev_b32_e32 v62, 16, v16
; SI-NEXT: v_lshlrev_b32_e32 v25, 16, v18
; SI-NEXT: v_lshlrev_b32_e32 v24, 16, v20
-; SI-NEXT: ; kill: killed $vgpr2
; SI-NEXT: ; implicit-def: $vgpr2
+; SI-NEXT: ; implicit-def: $vgpr51
; SI-NEXT: ; implicit-def: $vgpr30
; SI-NEXT: ; implicit-def: $vgpr52
; SI-NEXT: ; implicit-def: $vgpr48
; SI-NEXT: ; implicit-def: $vgpr39
; SI-NEXT: ; implicit-def: $vgpr31
+; SI-NEXT: ; implicit-def: $vgpr45
+; SI-NEXT: ; implicit-def: $vgpr46
; SI-NEXT: ; implicit-def: $vgpr47
-; SI-NEXT: ; kill: killed $vgpr2
; SI-NEXT: ; implicit-def: $vgpr23
; SI-NEXT: ; implicit-def: $vgpr49
; SI-NEXT: ; implicit-def: $vgpr37
; SI-NEXT: ; implicit-def: $vgpr36
; SI-NEXT: ; implicit-def: $vgpr22
+; SI-NEXT: ; implicit-def: $vgpr42
+; SI-NEXT: ; implicit-def: $vgpr43
; SI-NEXT: ; implicit-def: $vgpr44
-; SI-NEXT: ; implicit-def: $vgpr45
-; SI-NEXT: ; implicit-def: $vgpr46
; SI-NEXT: ; implicit-def: $vgpr21
; SI-NEXT: ; implicit-def: $vgpr38
; SI-NEXT: ; implicit-def: $vgpr34
; SI-NEXT: ; implicit-def: $vgpr33
; SI-NEXT: ; implicit-def: $vgpr18
+; SI-NEXT: ; implicit-def: $vgpr55
+; SI-NEXT: ; implicit-def: $vgpr40
; SI-NEXT: ; implicit-def: $vgpr41
-; SI-NEXT: ; implicit-def: $vgpr42
-; SI-NEXT: ; implicit-def: $vgpr43
; SI-NEXT: ; implicit-def: $vgpr14
; SI-NEXT: ; implicit-def: $vgpr35
; SI-NEXT: ; implicit-def: $vgpr29
; SI-NEXT: ; implicit-def: $vgpr28
; SI-NEXT: ; implicit-def: $vgpr10
+; SI-NEXT: ; kill: killed $vgpr2
+; SI-NEXT: ; implicit-def: $vgpr53
; SI-NEXT: ; implicit-def: $vgpr54
-; SI-NEXT: ; implicit-def: $vgpr55
-; SI-NEXT: ; implicit-def: $vgpr40
; SI-NEXT: ; implicit-def: $vgpr6
; SI-NEXT: ; implicit-def: $vgpr32
; SI-NEXT: ; implicit-def: $vgpr27
; SI-NEXT: ; implicit-def: $vgpr26
; SI-NEXT: ; implicit-def: $vgpr2
; SI-NEXT: ; implicit-def: $vgpr50
+; SI-NEXT: ; kill: killed $vgpr51
; SI-NEXT: ; implicit-def: $vgpr51
-; SI-NEXT: ; implicit-def: $vgpr53
; SI-NEXT: s_and_saveexec_b64 s[4:5], vcc
; SI-NEXT: s_xor_b64 s[4:5], exec, s[4:5]
; SI-NEXT: s_cbranch_execz .LBB48_2
@@ -16628,10 +16628,10 @@ define <40 x i8> @bitcast_v20i16_to_v40i8(<20 x i16> %a, i32 %b) {
; SI-NEXT: v_or_b32_e32 v6, v1, v25
; SI-NEXT: v_and_b32_e32 v1, 0xffff, v19
; SI-NEXT: v_or_b32_e32 v2, v1, v24
-; SI-NEXT: v_and_b32_e32 v1, 0xffff, v4
-; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:64 ; 4-byte Folded Spill
+; SI-NEXT: v_lshrrev_b32_e32 v1, 8, v10
+; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:68 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_bfe_u32 v1, v4, 8, 8
+; SI-NEXT: v_and_b32_e32 v1, 0xffff, v20
; SI-NEXT: v_alignbit_b32 v39, v31, v30, 24
; SI-NEXT: v_alignbit_b32 v48, v31, v30, 16
; SI-NEXT: v_alignbit_b32 v52, v31, v30, 8
@@ -16647,20 +16647,20 @@ define <40 x i8> @bitcast_v20i16_to_v40i8(<20 x i16> %a, i32 %b) {
; SI-NEXT: v_alignbit_b32 v26, v2, v6, 24
; SI-NEXT: v_alignbit_b32 v27, v2, v6, 16
; SI-NEXT: v_alignbit_b32 v32, v2, v6, 8
-; SI-NEXT: v_lshrrev_b32_e32 v47, 8, v31
-; SI-NEXT: v_lshrrev_b32_e32 v44, 8, v22
-; SI-NEXT: v_lshrrev_b32_e32 v41, 8, v18
-; SI-NEXT: v_lshrrev_b32_e32 v54, 8, v10
+; SI-NEXT: v_lshrrev_b32_e32 v45, 8, v31
+; SI-NEXT: v_lshrrev_b32_e32 v42, 8, v22
+; SI-NEXT: v_lshrrev_b32_e32 v55, 8, v18
; SI-NEXT: v_lshrrev_b32_e32 v50, 8, v2
-; SI-NEXT: v_and_b32_e32 v45, 0xffff, v8
-; SI-NEXT: v_and_b32_e32 v42, 0xffff, v12
-; SI-NEXT: v_and_b32_e32 v55, 0xffff, v16
-; SI-NEXT: v_and_b32_e32 v51, 0xffff, v20
-; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:68 ; 4-byte Folded Spill
-; SI-NEXT: v_bfe_u32 v46, v8, 8, 8
-; SI-NEXT: v_bfe_u32 v43, v12, 8, 8
-; SI-NEXT: v_bfe_u32 v40, v16, 8, 8
-; SI-NEXT: v_bfe_u32 v53, v20, 8, 8
+; SI-NEXT: v_and_b32_e32 v46, 0xffff, v4
+; SI-NEXT: v_and_b32_e32 v43, 0xffff, v8
+; SI-NEXT: v_and_b32_e32 v40, 0xffff, v12
+; SI-NEXT: v_and_b32_e32 v53, 0xffff, v16
+; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:64 ; 4-byte Folded Spill
+; SI-NEXT: v_bfe_u32 v47, v4, 8, 8
+; SI-NEXT: v_bfe_u32 v44, v8, 8, 8
+; SI-NEXT: v_bfe_u32 v41, v12, 8, 8
+; SI-NEXT: v_bfe_u32 v54, v16, 8, 8
+; SI-NEXT: v_bfe_u32 v51, v20, 8, 8
; SI-NEXT: ; implicit-def: $vgpr1
; SI-NEXT: ; implicit-def: $vgpr3
; SI-NEXT: ; implicit-def: $vgpr5
@@ -16682,10 +16682,7 @@ define <40 x i8> @bitcast_v20i16_to_v40i8(<20 x i16> %a, i32 %b) {
; SI-NEXT: ; implicit-def: $vgpr25
; SI-NEXT: ; implicit-def: $vgpr24
; SI-NEXT: .LBB48_2: ; %Flow
-; SI-NEXT: s_or_saveexec_b64 s[4:5], s[4:5]
-; SI-NEXT: buffer_load_dword v8, off, s[0:3], s32 offset:64 ; 4-byte Folded Reload
-; SI-NEXT: buffer_load_dword v12, off, s[0:3], s32 offset:68 ; 4-byte Folded Reload
-; SI-NEXT: s_xor_b64 exec, exec, s[4:5]
+; SI-NEXT: s_andn2_saveexec_b64 s[4:5], s[4:5]
; SI-NEXT: s_cbranch_execz .LBB48_4
; SI-NEXT: ; %bb.3: ; %cmp.true
; SI-NEXT: v_add_i32_e32 v4, vcc, 3, v13
@@ -16730,6 +16727,7 @@ define <40 x i8> @bitcast_v20i16_to_v40i8(<20 x i16> %a, i32 %b) {
; SI-NEXT: v_add_i32_e32 v2, vcc, s6, v2
; SI-NEXT: v_add_i32_e32 v22, vcc, s6, v4
; SI-NEXT: v_add_i32_e32 v31, vcc, s6, v1
+; SI-NEXT: v_lshrrev_b32_e32 v1, 8, v10
; SI-NEXT: v_alignbit_b32 v39, v31, v30, 24
; SI-NEXT: v_alignbit_b32 v48, v31, v30, 16
; SI-NEXT: v_alignbit_b32 v52, v31, v30, 8
@@ -16745,22 +16743,23 @@ define <40 x i8> @bitcast_v20i16_to_v40i8(<20 x i16> %a, i32 %b) {
; SI-NEXT: v_alignbit_b32 v26, v2, v6, 24
; SI-NEXT: v_alignbit_b32 v27, v2, v6, 16
; SI-NEXT: v_alignbit_b32 v32, v2, v6, 8
-; SI-NEXT: s_waitcnt vmcnt(0)
-; SI-NEXT: v_lshrrev_b32_e32 v12, 24, v31
-; SI-NEXT: v_lshrrev_b32_e32 v8, 16, v31
-; SI-NEXT: v_lshrrev_b32_e32 v47, 8, v31
-; SI-NEXT: v_lshrrev_b32_e32 v46, 24, v22
-; SI-NEXT: v_lshrrev_b32_e32 v45, 16, v22
-; SI-NEXT: v_lshrrev_b32_e32 v44, 8, v22
-; SI-NEXT: v_lshrrev_b32_e32 v43, 24, v18
-; SI-NEXT: v_lshrrev_b32_e32 v42, 16, v18
-; SI-NEXT: v_lshrrev_b32_e32 v41, 8, v18
-; SI-NEXT: v_lshrrev_b32_e32 v40, 24, v10
-; SI-NEXT: v_lshrrev_b32_e32 v55, 16, v10
-; SI-NEXT: v_lshrrev_b32_e32 v54, 8, v10
-; SI-NEXT: v_lshrrev_b32_e32 v53, 24, v2
-; SI-NEXT: v_lshrrev_b32_e32 v51, 16, v2
+; SI-NEXT: v_lshrrev_b32_e32 v47, 24, v31
+; SI-NEXT: v_lshrrev_b32_e32 v46, 16, v31
+; SI-NEXT: v_lshrrev_b32_e32 v45, 8, v31
+; SI-NEXT: v_lshrrev_b32_e32 v44, 24, v22
+; SI-NEXT: v_lshrrev_b32_e32 v43, 16, v22
+; SI-NEXT: v_lshrrev_b32_e32 v42, 8, v22
+; SI-NEXT: v_lshrrev_b32_e32 v41, 24, v18
+; SI-NEXT: v_lshrrev_b32_e32 v40, 16, v18
+; SI-NEXT: v_lshrrev_b32_e32 v55, 8, v18
+; SI-NEXT: v_lshrrev_b32_e32 v54, 24, v10
+; SI-NEXT: v_lshrrev_b32_e32 v53, 16, v10
+; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:68 ; 4-byte Folded Spill
+; SI-NEXT: v_lshrrev_b32_e32 v51, 24, v2
+; SI-NEXT: s_waitcnt expcnt(0)
+; SI-NEXT: v_lshrrev_b32_e32 v1, 16, v2
; SI-NEXT: v_lshrrev_b32_e32 v50, 8, v2
+; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:64 ; 4-byte Folded Spill
; SI-NEXT: .LBB48_4: ; %end
; SI-NEXT: s_or_b64 exec, exec, s[4:5]
; SI-NEXT: s_waitcnt expcnt(0)
@@ -16776,13 +16775,11 @@ define <40 x i8> @bitcast_v20i16_to_v40i8(<20 x i16> %a, i32 %b) {
; SI-NEXT: buffer_store_dword v1, v0, s[0:3], 0 offen
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_and_b32_e32 v1, 0xff, v31
-; SI-NEXT: v_lshlrev_b32_e32 v3, 8, v47
+; SI-NEXT: v_lshlrev_b32_e32 v3, 8, v45
; SI-NEXT: v_or_b32_e32 v1, v1, v3
-; SI-NEXT: s_waitcnt vmcnt(2)
-; SI-NEXT: v_and_b32_e32 v3, 0xff, v8
+; SI-NEXT: v_and_b32_e32 v3, 0xff, v46
; SI-NEXT: v_lshlrev_b32_e32 v3, 16, v3
-; SI-NEXT: s_waitcnt vmcnt(1)
-; SI-NEXT: v_lshlrev_b32_e32 v4, 24, v12
+; SI-NEXT: v_lshlrev_b32_e32 v4, 24, v47
; SI-NEXT: v_and_b32_e32 v1, 0xffff, v1
; SI-NEXT: v_or_b32_e32 v3, v4, v3
; SI-NEXT: v_or_b32_e32 v1, v1, v3
@@ -16802,11 +16799,11 @@ define <40 x i8> @bitcast_v20i16_to_v40i8(<20 x i16> %a, i32 %b) {
; SI-NEXT: buffer_store_dword v1, v3, s[0:3], 0 offen
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_and_b32_e32 v1, 0xff, v22
-; SI-NEXT: v_lshlrev_b32_e32 v3, 8, v44
+; SI-NEXT: v_lshlrev_b32_e32 v3, 8, v42
; SI-NEXT: v_or_b32_e32 v1, v1, v3
-; SI-NEXT: v_and_b32_e32 v3, 0xff, v45
+; SI-NEXT: v_and_b32_e32 v3, 0xff, v43
; SI-NEXT: v_lshlrev_b32_e32 v3, 16, v3
-; SI-NEXT: v_lshlrev_b32_e32 v4, 24, v46
+; SI-NEXT: v_lshlrev_b32_e32 v4, 24, v44
; SI-NEXT: v_and_b32_e32 v1, 0xffff, v1
; SI-NEXT: v_or_b32_e32 v3, v4, v3
; SI-NEXT: v_or_b32_e32 v1, v1, v3
@@ -16826,11 +16823,11 @@ define <40 x i8> @bitcast_v20i16_to_v40i8(<20 x i16> %a, i32 %b) {
; SI-NEXT: buffer_store_dword v1, v3, s[0:3], 0 offen
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_and_b32_e32 v1, 0xff, v18
-; SI-NEXT: v_lshlrev_b32_e32 v3, 8, v41
+; SI-NEXT: v_lshlrev_b32_e32 v3, 8, v55
; SI-NEXT: v_or_b32_e32 v1, v1, v3
-; SI-NEXT: v_and_b32_e32 v3, 0xff, v42
+; SI-NEXT: v_and_b32_e32 v3, 0xff, v40
; SI-NEXT: v_lshlrev_b32_e32 v3, 16, v3
-; SI-NEXT: v_lshlrev_b32_e32 v4, 24, v43
+; SI-NEXT: v_lshlrev_b32_e32 v4, 24, v41
; SI-NEXT: v_and_b32_e32 v1, 0xffff, v1
; SI-NEXT: v_or_b32_e32 v3, v4, v3
; SI-NEXT: v_or_b32_e32 v1, v1, v3
@@ -16848,13 +16845,15 @@ define <40 x i8> @bitcast_v20i16_to_v40i8(<20 x i16> %a, i32 %b) {
; SI-NEXT: v_or_b32_e32 v1, v1, v3
; SI-NEXT: v_add_i32_e32 v3, vcc, 24, v0
; SI-NEXT: buffer_store_dword v1, v3, s[0:3], 0 offen
+; SI-NEXT: buffer_load_dword v3, off, s[0:3], s32 offset:68 ; 4-byte Folded Reload
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_and_b32_e32 v1, 0xff, v10
-; SI-NEXT: v_lshlrev_b32_e32 v3, 8, v54
+; SI-NEXT: v_lshlrev_b32_e32 v4, 24, v54
+; SI-NEXT: s_waitcnt vmcnt(0)
+; SI-NEXT: v_lshlrev_b32_e32 v3, 8, v3
; SI-NEXT: v_or_b32_e32 v1, v1, v3
-; SI-NEXT: v_and_b32_e32 v3, 0xff, v55
+; SI-NEXT: v_and_b32_e32 v3, 0xff, v53
; SI-NEXT: v_lshlrev_b32_e32 v3, 16, v3
-; SI-NEXT: v_lshlrev_b32_e32 v4, 24, v40
; SI-NEXT: v_and_b32_e32 v1, 0xffff, v1
; SI-NEXT: v_or_b32_e32 v3, v4, v3
; SI-NEXT: v_or_b32_e32 v1, v1, v3
@@ -16876,13 +16875,15 @@ define <40 x i8> @bitcast_v20i16_to_v40i8(<20 x i16> %a, i32 %b) {
; SI-NEXT: v_and_b32_e32 v1, 0xff, v2
; SI-NEXT: v_lshlrev_b32_e32 v2, 8, v50
; SI-NEXT: v_or_b32_e32 v1, v1, v2
-; SI-NEXT: v_and_b32_e32 v2, 0xff, v51
-; SI-NEXT: v_lshlrev_b32_e32 v2, 16, v2
-; SI-NEXT: v_lshlrev_b32_e32 v3, 24, v53
+; SI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:64 ; 4-byte Folded Reload
+; SI-NEXT: v_lshlrev_b32_e32 v3, 24, v51
; SI-NEXT: v_and_b32_e32 v1, 0xffff, v1
+; SI-NEXT: v_add_i32_e32 v0, vcc, 36, v0
+; SI-NEXT: s_waitcnt vmcnt(0)
+; SI-NEXT: v_and_b32_e32 v2, 0xff, v2
+; SI-NEXT: v_lshlrev_b32_e32 v2, 16, v2
; SI-NEXT: v_or_b32_e32 v2, v3, v2
; SI-NEXT: v_or_b32_e32 v1, v1, v2
-; SI-NEXT: v_add_i32_e32 v0, vcc, 36, v0
; SI-NEXT: buffer_store_dword v1, v0, s[0:3], 0 offen
; SI-NEXT: buffer_load_dword v63, off, s[0:3], s32 ; 4-byte Folded Reload
; SI-NEXT: buffer_load_dword v62, off, s[0:3], s32 offset:4 ; 4-byte Folded Reload
@@ -16917,40 +16918,27 @@ define <40 x i8> @bitcast_v20i16_to_v40i8(<20 x i16> %a, i32 %b) {
; VI-NEXT: v_lshrrev_b32_e32 v25, 16, v3
; VI-NEXT: v_lshrrev_b32_e32 v23, 16, v2
; VI-NEXT: v_lshrrev_b32_e32 v26, 16, v1
-; VI-NEXT: buffer_store_dword v40, off, s[0:3], s32 offset:12 ; 4-byte Folded Spill
-; VI-NEXT: buffer_store_dword v41, off, s[0:3], s32 offset:8 ; 4-byte Folded Spill
-; VI-NEXT: buffer_store_dword v42, off, s[0:3], s32 offset:4 ; 4-byte Folded Spill
-; VI-NEXT: buffer_store_dword v43, off, s[0:3], s32 ; 4-byte Folded Spill
-; VI-NEXT: ; implicit-def: $vgpr34
-; VI-NEXT: ; implicit-def: $vgpr40
+; VI-NEXT: buffer_store_dword v40, off, s[0:3], s32 ; 4-byte Folded Spill
+; VI-NEXT: ; implicit-def: $vgpr16
; VI-NEXT: ; implicit-def: $vgpr15
-; VI-NEXT: ; implicit-def: $vgpr32
-; VI-NEXT: ; implicit-def: $vgpr54
-; VI-NEXT: ; implicit-def: $vgpr41
; VI-NEXT: ; implicit-def: $vgpr38
-; VI-NEXT: ; implicit-def: $vgpr51
-; VI-NEXT: ; implicit-def: $vgpr14
-; VI-NEXT: ; implicit-def: $vgpr37
; VI-NEXT: ; implicit-def: $vgpr48
-; VI-NEXT: ; implicit-def: $vgpr52
-; VI-NEXT: ; implicit-def: $vgpr50
+; VI-NEXT: ; implicit-def: $vgpr37
+; VI-NEXT: ; implicit-def: $vgpr14
; VI-NEXT: ; implicit-def: $vgpr36
+; VI-NEXT: ; implicit-def: $vgpr39
+; VI-NEXT: ; implicit-def: $vgpr34
; VI-NEXT: ; implicit-def: $vgpr13
-; VI-NEXT: ; implicit-def: $vgpr49
+; VI-NEXT: ; implicit-def: $vgpr33
; VI-NEXT: ; implicit-def: $vgpr35
-; VI-NEXT: ; implicit-def: $vgpr39
-; VI-NEXT: ; implicit-def: $vgpr55
; VI-NEXT: ; implicit-def: $vgpr31
; VI-NEXT: ; implicit-def: $vgpr12
-; VI-NEXT: ; implicit-def: $vgpr53
; VI-NEXT: ; implicit-def: $vgpr30
-; VI-NEXT: ; implicit-def: $vgpr33
-; VI-NEXT: ; implicit-def: $vgpr43
+; VI-NEXT: ; implicit-def: $vgpr32
; VI-NEXT: ; implicit-def: $vgpr28
-; VI-NEXT: ; implicit-def: $vgpr11
-; VI-NEXT: ; implicit-def: $vgpr42
; VI-NEXT: ; implicit-def: $vgpr27
; VI-NEXT: ; implicit-def: $vgpr29
+; VI-NEXT: ; implicit-def: $vgpr11
; VI-NEXT: s_and_saveexec_b64 s[4:5], vcc
; VI-NEXT: s_xor_b64 s[4:5], exec, s[4:5]
; VI-NEXT: s_cbranch_execz .LBB48_2
@@ -16959,37 +16947,22 @@ define <40 x i8> @bitcast_v20i16_to_v40i8(<20 x i16> %a, i32 %b) {
; VI-NEXT: v_lshrrev_b64 v[12:13], 24, v[7:8]
; VI-NEXT: v_lshrrev_b64 v[13:14], 24, v[5:6]
; VI-NEXT: v_lshrrev_b64 v[14:15], 24, v[3:4]
+; VI-NEXT: v_lshrrev_b64 v[15:16], 24, v[1:2]
; VI-NEXT: v_lshrrev_b32_e32 v29, 24, v10
; VI-NEXT: v_lshrrev_b32_e32 v27, 8, v10
; VI-NEXT: v_lshrrev_b32_e32 v28, 8, v9
-; VI-NEXT: v_lshrrev_b32_e32 v33, 24, v8
+; VI-NEXT: v_lshrrev_b32_e32 v32, 24, v8
; VI-NEXT: v_lshrrev_b32_e32 v30, 8, v8
; VI-NEXT: v_lshrrev_b32_e32 v31, 8, v7
-; VI-NEXT: v_lshrrev_b32_e32 v39, 24, v6
-; VI-NEXT: v_lshrrev_b32_e32 v35, 8, v6
-; VI-NEXT: v_lshrrev_b32_e32 v36, 8, v5
-; VI-NEXT: v_lshrrev_b32_e32 v52, 24, v4
-; VI-NEXT: v_lshrrev_b32_e32 v48, 8, v4
-; VI-NEXT: v_lshrrev_b32_e32 v51, 8, v3
-; VI-NEXT: v_lshrrev_b32_e32 v41, 24, v2
-; VI-NEXT: v_lshrrev_b32_e32 v54, 8, v2
-; VI-NEXT: v_lshrrev_b32_e32 v40, 8, v1
-; VI-NEXT: v_lshrrev_b64 v[15:16], 24, v[1:2]
-; VI-NEXT: v_mov_b32_e32 v34, v1
-; VI-NEXT: v_mov_b32_e32 v32, v2
-; VI-NEXT: v_mov_b32_e32 v38, v3
-; VI-NEXT: v_mov_b32_e32 v37, v4
-; VI-NEXT: v_mov_b32_e32 v50, v5
-; VI-NEXT: v_mov_b32_e32 v49, v6
-; VI-NEXT: v_mov_b32_e32 v55, v7
-; VI-NEXT: v_mov_b32_e32 v53, v8
-; VI-NEXT: v_mov_b32_e32 v43, v9
-; VI-NEXT: v_mov_b32_e32 v42, v10
-; VI-NEXT: ; implicit-def: $vgpr1
-; VI-NEXT: ; implicit-def: $vgpr3
-; VI-NEXT: ; implicit-def: $vgpr5
-; VI-NEXT: ; implicit-def: $vgpr7
-; VI-NEXT: ; implicit-def: $vgpr9
+; VI-NEXT: v_lshrrev_b32_e32 v35, 24, v6
+; VI-NEXT: v_lshrrev_b32_e32 v33, 8, v6
+; VI-NEXT: v_lshrrev_b32_e32 v34, 8, v5
+; VI-NEXT: v_lshrrev_b32_e32 v39, 24, v4
+; VI-NEXT: v_lshrrev_b32_e32 v36, 8, v4
+; VI-NEXT: v_lshrrev_b32_e32 v37, 8, v3
+; VI-NEXT: v_lshrrev_b32_e32 v48, 24, v2
+; VI-NEXT: v_lshrrev_b32_e32 v38, 8, v2
+; VI-NEXT: v_lshrrev_b32_e32 v16, 8, v1
; VI-NEXT: .LBB48_2: ; %Flow
; VI-NEXT: s_andn2_saveexec_b64 s[4:5], s[4:5]
; VI-NEXT: s_cbranch_execz .LBB48_4
@@ -16999,137 +16972,144 @@ define <40 x i8> @bitcast_v20i16_to_v40i8(<20 x i16> %a, i32 %b) {
; VI-NEXT: v_add_u16_sdwa v20, v9, v11 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_1 src1_sel:DWORD
; VI-NEXT: v_add_u16_sdwa v18, v8, v11 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_1 src1_sel:DWORD
; VI-NEXT: v_add_u16_sdwa v22, v7, v11 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_1 src1_sel:DWORD
-; VI-NEXT: v_add_u16_e32 v42, 3, v10
+; VI-NEXT: v_add_u16_e32 v55, 3, v10
; VI-NEXT: v_lshlrev_b32_e32 v10, 16, v17
-; VI-NEXT: v_add_u16_e32 v43, 3, v9
+; VI-NEXT: v_add_u16_e32 v40, 3, v9
; VI-NEXT: v_lshlrev_b32_e32 v9, 16, v20
; VI-NEXT: v_add_u16_sdwa v19, v6, v11 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_1 src1_sel:DWORD
; VI-NEXT: v_add_u16_sdwa v24, v5, v11 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_1 src1_sel:DWORD
; VI-NEXT: v_add_u16_e32 v53, 3, v8
; VI-NEXT: v_lshlrev_b32_e32 v8, 16, v18
-; VI-NEXT: v_add_u16_e32 v55, 3, v7
+; VI-NEXT: v_add_u16_e32 v54, 3, v7
; VI-NEXT: v_lshlrev_b32_e32 v7, 16, v22
-; VI-NEXT: v_or_b32_e32 v10, v42, v10
-; VI-NEXT: v_or_b32_e32 v9, v43, v9
+; VI-NEXT: v_or_b32_e32 v10, v55, v10
+; VI-NEXT: v_or_b32_e32 v9, v40, v9
; VI-NEXT: v_add_u16_sdwa v23, v2, v11 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_1 src1_sel:DWORD
; VI-NEXT: v_add_u16_sdwa v26, v1, v11 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_1 src1_sel:DWORD
; VI-NEXT: v_add_u16_sdwa v21, v4, v11 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_1 src1_sel:DWORD
; VI-NEXT: v_add_u16_sdwa v25, v3, v11 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_1 src1_sel:DWORD
-; VI-NEXT: v_add_u16_e32 v49, 3, v6
+; VI-NEXT: v_add_u16_e32 v51, 3, v6
; VI-NEXT: v_lshlrev_b32_e32 v6, 16, v19
-; VI-NEXT: v_add_u16_e32 v50, 3, v5
+; VI-NEXT: v_add_u16_e32 v52, 3, v5
; VI-NEXT: v_lshlrev_b32_e32 v5, 16, v24
; VI-NEXT: v_or_b32_e32 v8, v53, v8
-; VI-NEXT: v_or_b32_e32 v7, v55, v7
+; VI-NEXT: v_or_b32_e32 v7, v54, v7
; VI-NEXT: v_lshrrev_b64 v[11:12], 24, v[9:10]
-; VI-NEXT: v_add_u16_e32 v37, 3, v4
+; VI-NEXT: v_add_u16_e32 v49, 3, v4
; VI-NEXT: v_lshlrev_b32_e32 v4, 16, v21
-; VI-NEXT: v_add_u16_e32 v38, 3, v3
+; VI-NEXT: v_add_u16_e32 v50, 3, v3
; VI-NEXT: v_lshlrev_b32_e32 v3, 16, v25
-; VI-NEXT: v_or_b32_e32 v6, v49, v6
-; VI-NEXT: v_or_b32_e32 v5, v50, v5
+; VI-NEXT: v_or_b32_e32 v6, v51, v6
+; VI-NEXT: v_or_b32_e32 v5, v52, v5
; VI-NEXT: v_lshrrev_b64 v[12:13], 24, v[7:8]
-; VI-NEXT: v_add_u16_e32 v32, 3, v2
+; VI-NEXT: v_add_u16_e32 v39, 3, v2
; VI-NEXT: v_lshlrev_b32_e32 v2, 16, v23
-; VI-NEXT: v_add_u16_e32 v34, 3, v1
+; VI-NEXT: v_add_u16_e32 v48, 3, v1
; VI-NEXT: v_lshlrev_b32_e32 v1, 16, v26
-; VI-NEXT: v_or_b32_e32 v4, v37, v4
-; VI-NEXT: v_or_b32_e32 v3, v38, v3
+; VI-NEXT: v_or_b32_e32 v4, v49, v4
+; VI-NEXT: v_or_b32_e32 v3, v50, v3
; VI-NEXT: v_lshrrev_b64 v[13:14], 24, v[5:6]
-; VI-NEXT: v_or_b32_e32 v2, v32, v2
-; VI-NEXT: v_or_b32_e32 v1, v34, v1
+; VI-NEXT: v_or_b32_e32 v2, v39, v2
+; VI-NEXT: v_or_b32_e32 v1, v48, v1
; VI-NEXT: v_lshrrev_b64 v[14:15], 24, v[3:4]
; VI-NEXT: v_lshrrev_b64 v[15:16], 24, v[1:2]
; VI-NEXT: v_lshrrev_b32_e32 v27, 8, v10
; VI-NEXT: v_lshrrev_b32_e32 v28, 8, v9
; VI-NEXT: v_lshrrev_b32_e32 v30, 8, v8
; VI-NEXT: v_lshrrev_b32_e32 v31, 8, v7
-; VI-NEXT: v_lshrrev_b32_e32 v35, 8, v6
-; VI-NEXT: v_lshrrev_b32_e32 v36, 8, v5
-; VI-NEXT: v_lshrrev_b32_e32 v48, 8, v4
-; VI-NEXT: v_lshrrev_b32_e32 v51, 8, v3
-; VI-NEXT: v_lshrrev_b32_e32 v54, 8, v2
-; VI-NEXT: v_lshrrev_b32_e32 v40, 8, v1
+; VI-NEXT: v_lshrrev_b32_e32 v33, 8, v6
+; VI-NEXT: v_lshrrev_b32_e32 v34, 8, v5
+; VI-NEXT: v_lshrrev_b32_e32 v36, 8, v4
+; VI-NEXT: v_lshrrev_b32_e32 v37, 8, v3
+; VI-NEXT: v_lshrrev_b32_e32 v38, 8, v2
+; VI-NEXT: v_lshrrev_b32_e32 v16, 8, v1
; VI-NEXT: v_bfe_u32 v29, v17, 8, 8
-; VI-NEXT: v_bfe_u32 v33, v18, 8, 8
-; VI-NEXT: v_bfe_u32 v39, v19, 8, 8
-; VI-NEXT: v_bfe_u32 v52, v21, 8, 8
-; VI-NEXT: v_bfe_u32 v41, v23, 8, 8
+; VI-NEXT: v_bfe_u32 v32, v18, 8, 8
+; VI-NEXT: v_bfe_u32 v35, v19, 8, 8
+; VI-NEXT: v_mov_b32_e32 v1, v48
+; VI-NEXT: v_mov_b32_e32 v2, v39
+; VI-NEXT: v_mov_b32_e32 v3, v50
+; VI-NEXT: v_mov_b32_e32 v4, v49
+; VI-NEXT: v_mov_b32_e32 v5, v52
+; VI-NEXT: v_mov_b32_e32 v6, v51
+; VI-NEXT: v_mov_b32_e32 v7, v54
+; VI-NEXT: v_mov_b32_e32 v8, v53
+; VI-NEXT: v_mov_b32_e32 v9, v40
+; VI-NEXT: v_mov_b32_e32 v10, v55
+; VI-NEXT: v_bfe_u32 v39, v21, 8, 8
+; VI-NEXT: v_bfe_u32 v48, v23, 8, 8
; VI-NEXT: .LBB48_4: ; %end
; VI-NEXT: s_or_b64 exec, exec, s[4:5]
-; VI-NEXT: v_lshlrev_b16_e32 v1, 8, v40
-; VI-NEXT: v_lshlrev_b16_e32 v2, 8, v15
-; VI-NEXT: v_or_b32_sdwa v1, v34, v1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
-; VI-NEXT: v_or_b32_sdwa v2, v26, v2 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
-; VI-NEXT: v_or_b32_sdwa v1, v1, v2 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_0 src1_sel:DWORD
+; VI-NEXT: v_lshlrev_b16_e32 v15, 8, v15
+; VI-NEXT: v_lshlrev_b16_e32 v16, 8, v16
+; VI-NEXT: v_or_b32_sdwa v15, v26, v15 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; VI-NEXT: v_or_b32_sdwa v1, v1, v16 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; VI-NEXT: v_or_b32_sdwa v1, v1, v15 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_0 src1_sel:DWORD
; VI-NEXT: buffer_store_dword v1, v0, s[0:3], 0 offen
-; VI-NEXT: v_lshlrev_b16_e32 v1, 8, v54
-; VI-NEXT: v_lshlrev_b16_e32 v2, 8, v41
-; VI-NEXT: v_or_b32_sdwa v1, v32, v1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; VI-NEXT: v_lshlrev_b16_e32 v1, 8, v38
+; VI-NEXT: v_or_b32_sdwa v1, v2, v1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; VI-NEXT: v_lshlrev_b16_e32 v2, 8, v48
; VI-NEXT: v_or_b32_sdwa v2, v23, v2 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
; VI-NEXT: v_or_b32_sdwa v1, v1, v2 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_0 src1_sel:DWORD
; VI-NEXT: v_add_u32_e32 v2, vcc, 4, v0
; VI-NEXT: buffer_store_dword v1, v2, s[0:3], 0 offen
-; VI-NEXT: v_lshlrev_b16_e32 v1, 8, v51
+; VI-NEXT: v_lshlrev_b16_e32 v1, 8, v37
; VI-NEXT: v_lshlrev_b16_e32 v2, 8, v14
-; VI-NEXT: v_or_b32_sdwa v1, v38, v1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; VI-NEXT: v_or_b32_sdwa v1, v3, v1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
; VI-NEXT: v_or_b32_sdwa v2, v25, v2 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
; VI-NEXT: v_or_b32_sdwa v1, v1, v2 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_0 src1_sel:DWORD
; VI-NEXT: v_add_u32_e32 v2, vcc, 8, v0
; VI-NEXT: buffer_store_dword v1, v2, s[0:3], 0 offen
-; VI-NEXT: v_lshlrev_b16_e32 v1, 8, v48
-; VI-NEXT: v_lshlrev_b16_e32 v2, 8, v52
-; VI-NEXT: v_or_b32_sdwa v1, v37, v1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; VI-NEXT: v_lshlrev_b16_e32 v1, 8, v36
+; VI-NEXT: v_lshlrev_b16_e32 v2, 8, v39
+; VI-NEXT: v_or_b32_sdwa v1, v4, v1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
; VI-NEXT: v_or_b32_sdwa v2, v21, v2 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
; VI-NEXT: v_or_b32_sdwa v1, v1, v2 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_0 src1_sel:DWORD
; VI-NEXT: v_add_u32_e32 v2, vcc, 12, v0
; VI-NEXT: buffer_store_dword v1, v2, s[0:3], 0 offen
-; VI-NEXT: v_lshlrev_b16_e32 v1, 8, v36
+; VI-NEXT: v_lshlrev_b16_e32 v1, 8, v34
; VI-NEXT: v_lshlrev_b16_e32 v2, 8, v13
-; VI-NEXT: v_or_b32_sdwa v1, v50, v1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; VI-NEXT: v_or_b32_sdwa v1, v5, v1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
; VI-NEXT: v_or_b32_sdwa v2, v24, v2 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
; VI-NEXT: v_or_b32_sdwa v1, v1, v2 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_0 src1_sel:DWORD
; VI-NEXT: v_add_u32_e32 v2, vcc, 16, v0
; VI-NEXT: buffer_store_dword v1, v2, s[0:3], 0 offen
-; VI-NEXT: v_lshlrev_b16_e32 v1, 8, v35
-; VI-NEXT: v_lshlrev_b16_e32 v2, 8, v39
-; VI-NEXT: v_or_b32_sdwa v1, v49, v1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; VI-NEXT: v_lshlrev_b16_e32 v1, 8, v33
+; VI-NEXT: v_lshlrev_b16_e32 v2, 8, v35
+; VI-NEXT: v_or_b32_sdwa v1, v6, v1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
; VI-NEXT: v_or_b32_sdwa v2, v19, v2 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
; VI-NEXT: v_or_b32_sdwa v1, v1, v2 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_0 src1_sel:DWORD
; VI-NEXT: v_add_u32_e32 v2, vcc, 20, v0
; VI-NEXT: buffer_store_dword v1, v2, s[0:3], 0 offen
; VI-NEXT: v_lshlrev_b16_e32 v1, 8, v31
; VI-NEXT: v_lshlrev_b16_e32 v2, 8, v12
-; VI-NEXT: v_or_b32_sdwa v1, v55, v1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; VI-NEXT: v_or_b32_sdwa v1, v7, v1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
; VI-NEXT: v_or_b32_sdwa v2, v22, v2 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
; VI-NEXT: v_or_b32_sdwa v1, v1, v2 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_0 src1_sel:DWORD
; VI-NEXT: v_add_u32_e32 v2, vcc, 24, v0
; VI-NEXT: buffer_store_dword v1, v2, s[0:3], 0 offen
; VI-NEXT: v_lshlrev_b16_e32 v1, 8, v30
-; VI-NEXT: v_lshlrev_b16_e32 v2, 8, v33
-; VI-NEXT: v_or_b32_sdwa v1, v53, v1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; VI-NEXT: v_lshlrev_b16_e32 v2, 8, v32
+; VI-NEXT: v_or_b32_sdwa v1, v8, v1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
; VI-NEXT: v_or_b32_sdwa v2, v18, v2 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
; VI-NEXT: v_or_b32_sdwa v1, v1, v2 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_0 src1_sel:DWORD
; VI-NEXT: v_add_u32_e32 v2, vcc, 28, v0
; VI-NEXT: buffer_store_dword v1, v2, s[0:3], 0 offen
; VI-NEXT: v_lshlrev_b16_e32 v1, 8, v28
; VI-NEXT: v_lshlrev_b16_e32 v2, 8, v11
-; VI-NEXT: v_or_b32_sdwa v1, v43, v1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; VI-NEXT: v_or_b32_sdwa v1, v9, v1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
; VI-NEXT: v_or_b32_sdwa v2, v20, v2 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
; VI-NEXT: v_or_b32_sdwa v1, v1, v2 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_0 src1_sel:DWORD
; VI-NEXT: v_add_u32_e32 v2, vcc, 32, v0
; VI-NEXT: buffer_store_dword v1, v2, s[0:3], 0 offen
; VI-NEXT: v_lshlrev_b16_e32 v1, 8, v27
; VI-NEXT: v_lshlrev_b16_e32 v2, 8, v29
-; VI-NEXT: v_or_b32_sdwa v1, v42, v1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; VI-NEXT: v_or_b32_sdwa v1, v10, v1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
; VI-NEXT: v_or_b32_sdwa v2, v17, v2 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
; VI-NEXT: v_or_b32_sdwa v1, v1, v2 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_0 src1_sel:DWORD
; VI-NEXT: v_add_u32_e32 v0, vcc, 36, v0
; VI-NEXT: buffer_store_dword v1, v0, s[0:3], 0 offen
-; VI-NEXT: buffer_load_dword v43, off, s[0:3], s32 ; 4-byte Folded Reload
-; VI-NEXT: buffer_load_dword v42, off, s[0:3], s32 offset:4 ; 4-byte Folded Reload
-; VI-NEXT: buffer_load_dword v41, off, s[0:3], s32 offset:8 ; 4-byte Folded Reload
-; VI-NEXT: buffer_load_dword v40, off, s[0:3], s32 offset:12 ; 4-byte Folded Reload
+; VI-NEXT: buffer_load_dword v40, off, s[0:3], s32 ; 4-byte Folded Reload
; VI-NEXT: s_waitcnt vmcnt(0)
; VI-NEXT: s_setpc_b64 s[30:31]
;
@@ -20353,8 +20333,8 @@ define inreg <20 x i16> @bitcast_v40i8_to_v20i16_scalar(<40 x i8> inreg %a, i32
; SI-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; SI-NEXT: v_cmp_ne_u32_e32 vcc, 0, v26
; SI-NEXT: v_readfirstlane_b32 s14, v19
-; SI-NEXT: v_readfirstlane_b32 s40, v18
-; SI-NEXT: v_readfirstlane_b32 s12, v11
+; SI-NEXT: v_readfirstlane_b32 s15, v18
+; SI-NEXT: v_readfirstlane_b32 s11, v11
; SI-NEXT: v_readfirstlane_b32 s13, v10
; SI-NEXT: v_readfirstlane_b32 s8, v3
; SI-NEXT: v_readfirstlane_b32 s9, v2
@@ -20380,22 +20360,22 @@ define inreg <20 x i16> @bitcast_v40i8_to_v20i16_scalar(<40 x i8> inreg %a, i32
; SI-NEXT: s_lshl_b32 s10, s23, 24
; SI-NEXT: s_and_b32 s4, s4, 0xffff
; SI-NEXT: s_or_b32 s5, s10, s5
-; SI-NEXT: s_or_b32 s11, s4, s5
+; SI-NEXT: s_or_b32 s12, s4, s5
; SI-NEXT: s_and_b32 s4, s18, 0xff
; SI-NEXT: s_lshl_b32 s4, s4, 16
; SI-NEXT: s_lshl_b32 s10, s19, 24
; SI-NEXT: s_or_b32 s4, s10, s4
; SI-NEXT: s_and_b32 s10, s28, 0xff
-; SI-NEXT: s_lshl_b32 s15, s29, 8
-; SI-NEXT: s_or_b32 s10, s10, s15
-; SI-NEXT: s_and_b32 s15, s6, 0xff
-; SI-NEXT: s_lshl_b32 s15, s15, 16
+; SI-NEXT: s_lshl_b32 s40, s29, 8
+; SI-NEXT: s_or_b32 s10, s10, s40
+; SI-NEXT: s_and_b32 s40, s6, 0xff
+; SI-NEXT: s_lshl_b32 s40, s40, 16
; SI-NEXT: s_lshl_b32 s41, s7, 24
-; SI-NEXT: s_or_b32 s43, s41, s15
-; SI-NEXT: s_and_b32 s15, s26, 0xff
-; SI-NEXT: s_lshl_b32 s15, s15, 16
+; SI-NEXT: s_or_b32 s43, s41, s40
+; SI-NEXT: s_and_b32 s40, s26, 0xff
+; SI-NEXT: s_lshl_b32 s40, s40, 16
; SI-NEXT: s_lshl_b32 s41, s27, 24
-; SI-NEXT: s_or_b32 s15, s41, s15
+; SI-NEXT: s_or_b32 s40, s41, s40
; SI-NEXT: s_and_b32 s41, s16, 0xff
; SI-NEXT: s_lshl_b32 s42, s17, 8
; SI-NEXT: s_or_b32 s41, s41, s42
@@ -20412,12 +20392,12 @@ define inreg <20 x i16> @bitcast_v40i8_to_v20i16_scalar(<40 x i8> inreg %a, i32
; SI-NEXT: v_and_b32_e32 v9, 0xffff, v9
; SI-NEXT: v_or_b32_e32 v11, v0, v10
; SI-NEXT: s_and_b32 s4, s4, 0xffff
-; SI-NEXT: v_mov_b32_e32 v5, s15
+; SI-NEXT: v_mov_b32_e32 v5, s40
; SI-NEXT: v_or_b32_e32 v10, v9, v11
; SI-NEXT: v_and_b32_e32 v9, 0xff, v4
; SI-NEXT: v_and_b32_e32 v13, 0xff, v14
; SI-NEXT: v_and_b32_e32 v17, 0xff, v16
-; SI-NEXT: s_or_b32 s15, s4, s15
+; SI-NEXT: s_or_b32 s40, s4, s40
; SI-NEXT: s_and_b32 s4, s9, 0xff
; SI-NEXT: s_lshl_b32 s42, s8, 8
; SI-NEXT: v_lshlrev_b32_e32 v9, 16, v9
@@ -20434,7 +20414,7 @@ define inreg <20 x i16> @bitcast_v40i8_to_v20i16_scalar(<40 x i8> inreg %a, i32
; SI-NEXT: v_and_b32_e32 v18, 0xff, v24
; SI-NEXT: v_or_b32_e32 v23, s4, v15
; SI-NEXT: s_and_b32 s4, s13, 0xff
-; SI-NEXT: s_lshl_b32 s42, s12, 8
+; SI-NEXT: s_lshl_b32 s42, s11, 8
; SI-NEXT: v_lshlrev_b32_e32 v13, 16, v13
; SI-NEXT: v_or_b32_e32 v17, v17, v30
; SI-NEXT: v_lshlrev_b32_e32 v18, 16, v18
@@ -20446,7 +20426,7 @@ define inreg <20 x i16> @bitcast_v40i8_to_v20i16_scalar(<40 x i8> inreg %a, i32
; SI-NEXT: v_or_b32_e32 v18, v17, v32
; SI-NEXT: v_and_b32_e32 v17, 0xff, v20
; SI-NEXT: v_or_b32_e32 v26, s4, v21
-; SI-NEXT: s_and_b32 s4, s40, 0xff
+; SI-NEXT: s_and_b32 s4, s15, 0xff
; SI-NEXT: s_lshl_b32 s42, s14, 8
; SI-NEXT: s_and_b32 s10, s10, 0xffff
; SI-NEXT: v_lshlrev_b32_e32 v17, 16, v17
@@ -20454,7 +20434,7 @@ define inreg <20 x i16> @bitcast_v40i8_to_v20i16_scalar(<40 x i8> inreg %a, i32
; SI-NEXT: s_or_b32 s10, s10, s43
; SI-NEXT: v_or_b32_e32 v33, v31, v17
; SI-NEXT: s_and_b32 s4, s4, 0xffff
-; SI-NEXT: v_alignbit_b32 v1, s11, v1, 16
+; SI-NEXT: v_alignbit_b32 v1, s12, v1, 16
; SI-NEXT: v_alignbit_b32 v5, s10, v5, 16
; SI-NEXT: v_alignbit_b32 v9, v10, v15, 16
; SI-NEXT: v_alignbit_b32 v13, v25, v21, 16
@@ -20467,8 +20447,8 @@ define inreg <20 x i16> @bitcast_v40i8_to_v20i16_scalar(<40 x i8> inreg %a, i32
; SI-NEXT: v_lshrrev_b32_e32 v19, 16, v32
; SI-NEXT: s_cbranch_execnz .LBB51_3
; SI-NEXT: .LBB51_2: ; %cmp.true
-; SI-NEXT: s_add_i32 s40, s40, 3
-; SI-NEXT: s_and_b32 s4, s40, 0xff
+; SI-NEXT: s_add_i32 s15, s15, 3
+; SI-NEXT: s_and_b32 s4, s15, 0xff
; SI-NEXT: s_lshl_b32 s5, s14, 8
; SI-NEXT: v_add_i32_e32 v1, vcc, 3, v20
; SI-NEXT: s_or_b32 s4, s5, s4
@@ -20492,7 +20472,7 @@ define inreg <20 x i16> @bitcast_v40i8_to_v20i16_scalar(<40 x i8> inreg %a, i32
; SI-NEXT: s_add_i32 s13, s13, 3
; SI-NEXT: v_add_i32_e32 v18, vcc, 0x3000000, v1
; SI-NEXT: s_and_b32 s4, s13, 0xff
-; SI-NEXT: s_lshl_b32 s5, s12, 8
+; SI-NEXT: s_lshl_b32 s5, s11, 8
; SI-NEXT: v_add_i32_e32 v1, vcc, 3, v12
; SI-NEXT: s_or_b32 s4, s5, s4
; SI-NEXT: v_and_b32_e32 v1, 0xff, v1
@@ -20537,7 +20517,7 @@ define inreg <20 x i16> @bitcast_v40i8_to_v20i16_scalar(<40 x i8> inreg %a, i32
; SI-NEXT: s_or_b32 s5, s5, s8
; SI-NEXT: s_or_b32 s4, s5, s4
; SI-NEXT: s_add_i32 s28, s28, 3
-; SI-NEXT: s_add_i32 s15, s4, 0x3000000
+; SI-NEXT: s_add_i32 s40, s4, 0x3000000
; SI-NEXT: s_and_b32 s4, s28, 0xff
; SI-NEXT: s_lshl_b32 s5, s29, 8
; SI-NEXT: s_add_i32 s6, s6, 3
@@ -20587,24 +20567,24 @@ define inreg <20 x i16> @bitcast_v40i8_to_v20i16_scalar(<40 x i8> inreg %a, i32
; SI-NEXT: v_or_b32_e32 v0, v0, v1
; SI-NEXT: s_or_b32 s4, s5, s4
; SI-NEXT: v_add_i32_e32 v10, vcc, 0x3000000, v0
-; SI-NEXT: s_add_i32 s11, s4, 0x3000000
+; SI-NEXT: s_add_i32 s12, s4, 0x3000000
; SI-NEXT: v_mov_b32_e32 v0, s41
-; SI-NEXT: v_alignbit_b32 v1, s11, v0, 16
-; SI-NEXT: v_mov_b32_e32 v0, s15
+; SI-NEXT: v_alignbit_b32 v1, s12, v0, 16
+; SI-NEXT: v_mov_b32_e32 v0, s40
; SI-NEXT: v_alignbit_b32 v5, s10, v0, 16
; SI-NEXT: v_alignbit_b32 v9, v10, v23, 16
; SI-NEXT: v_alignbit_b32 v13, v25, v26, 16
; SI-NEXT: v_alignbit_b32 v17, v18, v21, 16
-; SI-NEXT: s_lshr_b32 s42, s11, 16
+; SI-NEXT: s_lshr_b32 s42, s12, 16
; SI-NEXT: s_lshr_b32 s43, s10, 16
; SI-NEXT: v_lshrrev_b32_e32 v11, 16, v10
; SI-NEXT: v_lshrrev_b32_e32 v15, 16, v25
; SI-NEXT: v_lshrrev_b32_e32 v19, 16, v18
; SI-NEXT: .LBB51_3: ; %end
; SI-NEXT: v_mov_b32_e32 v0, s41
-; SI-NEXT: v_mov_b32_e32 v2, s11
+; SI-NEXT: v_mov_b32_e32 v2, s12
; SI-NEXT: v_mov_b32_e32 v3, s42
-; SI-NEXT: v_mov_b32_e32 v4, s15
+; SI-NEXT: v_mov_b32_e32 v4, s40
; SI-NEXT: v_mov_b32_e32 v6, s10
; SI-NEXT: v_mov_b32_e32 v7, s43
; SI-NEXT: v_mov_b32_e32 v8, v23
@@ -20615,9 +20595,9 @@ define inreg <20 x i16> @bitcast_v40i8_to_v20i16_scalar(<40 x i8> inreg %a, i32
; SI-NEXT: .LBB51_4:
; SI-NEXT: ; implicit-def: $sgpr41
; SI-NEXT: ; implicit-def: $vgpr1
-; SI-NEXT: ; implicit-def: $sgpr11
+; SI-NEXT: ; implicit-def: $sgpr12
; SI-NEXT: ; implicit-def: $sgpr42
-; SI-NEXT: ; implicit-def: $sgpr15
+; SI-NEXT: ; implicit-def: $sgpr40
; SI-NEXT: ; implicit-def: $vgpr5
; SI-NEXT: ; implicit-def: $sgpr10
; SI-NEXT: ; implicit-def: $sgpr43
@@ -38803,7 +38783,7 @@ define inreg <40 x i8> @bitcast_v5i64_to_v40i8_scalar(<5 x i64> inreg %a, i32 in
; GFX11-TRUE16: ; %bb.0:
; GFX11-TRUE16-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX11-TRUE16-NEXT: s_cmp_lg_u32 s22, 0
-; GFX11-TRUE16-NEXT: s_mov_b32 s63, 0
+; GFX11-TRUE16-NEXT: s_mov_b32 s61, 0
; GFX11-TRUE16-NEXT: s_cbranch_scc0 .LBB79_4
; GFX11-TRUE16-NEXT: ; %bb.1: ; %cmp.false
; GFX11-TRUE16-NEXT: s_lshr_b32 s14, s21, 24
@@ -38829,14 +38809,14 @@ define inreg <40 x i8> @bitcast_v5i64_to_v40i8_scalar(<5 x i64> inreg %a, i32 in
; GFX11-TRUE16-NEXT: s_lshr_b32 s58, s1, 24
; GFX11-TRUE16-NEXT: s_lshr_b32 s59, s1, 16
; GFX11-TRUE16-NEXT: s_lshr_b32 s60, s1, 8
-; GFX11-TRUE16-NEXT: s_lshr_b32 s61, s0, 16
-; GFX11-TRUE16-NEXT: s_lshr_b32 s62, s0, 8
+; GFX11-TRUE16-NEXT: s_lshr_b32 s62, s0, 16
+; GFX11-TRUE16-NEXT: s_lshr_b32 s63, s0, 8
; GFX11-TRUE16-NEXT: s_lshr_b64 s[4:5], s[20:21], 24
; GFX11-TRUE16-NEXT: s_lshr_b64 s[6:7], s[18:19], 24
; GFX11-TRUE16-NEXT: s_lshr_b64 s[8:9], s[16:17], 24
; GFX11-TRUE16-NEXT: s_lshr_b64 s[10:11], s[2:3], 24
; GFX11-TRUE16-NEXT: s_lshr_b64 s[12:13], s[0:1], 24
-; GFX11-TRUE16-NEXT: s_and_not1_b32 vcc_lo, exec_lo, s63
+; GFX11-TRUE16-NEXT: s_and_not1_b32 vcc_lo, exec_lo, s61
; GFX11-TRUE16-NEXT: s_cbranch_vccnz .LBB79_3
; GFX11-TRUE16-NEXT: .LBB79_2: ; %cmp.true
; GFX11-TRUE16-NEXT: s_add_u32 s0, s0, 3
@@ -38877,12 +38857,12 @@ define inreg <40 x i8> @bitcast_v5i64_to_v40i8_scalar(<5 x i64> inreg %a, i32 in
; GFX11-TRUE16-NEXT: s_lshr_b32 s58, s1, 24
; GFX11-TRUE16-NEXT: s_lshr_b32 s59, s1, 16
; GFX11-TRUE16-NEXT: s_lshr_b32 s60, s1, 8
-; GFX11-TRUE16-NEXT: s_lshr_b32 s61, s0, 16
-; GFX11-TRUE16-NEXT: s_lshr_b32 s62, s0, 8
+; GFX11-TRUE16-NEXT: s_lshr_b32 s62, s0, 16
+; GFX11-TRUE16-NEXT: s_lshr_b32 s63, s0, 8
; GFX11-TRUE16-NEXT: .LBB79_3: ; %end
; GFX11-TRUE16-NEXT: s_delay_alu instid0(SALU_CYCLE_1)
-; GFX11-TRUE16-NEXT: s_mov_b32 s5, s62
-; GFX11-TRUE16-NEXT: s_mov_b32 s7, s61
+; GFX11-TRUE16-NEXT: s_mov_b32 s5, s63
+; GFX11-TRUE16-NEXT: s_mov_b32 s7, s62
; GFX11-TRUE16-NEXT: s_mov_b32 s9, s12
; GFX11-TRUE16-NEXT: s_and_b32 s0, s0, 0xff
; GFX11-TRUE16-NEXT: s_lshl_b32 s5, s5, 8
@@ -39018,8 +38998,8 @@ define inreg <40 x i8> @bitcast_v5i64_to_v40i8_scalar(<5 x i64> inreg %a, i32 in
; GFX11-TRUE16-NEXT: scratch_store_b64 v0, v[9:10], off offset:32
; GFX11-TRUE16-NEXT: s_setpc_b64 s[30:31]
; GFX11-TRUE16-NEXT: .LBB79_4:
+; GFX11-TRUE16-NEXT: ; implicit-def: $sgpr63_lo16
; GFX11-TRUE16-NEXT: ; implicit-def: $sgpr62_lo16
-; GFX11-TRUE16-NEXT: ; implicit-def: $sgpr61_lo16
; GFX11-TRUE16-NEXT: ; implicit-def: $sgpr12_lo16
; GFX11-TRUE16-NEXT: ; implicit-def: $sgpr60_lo16
; GFX11-TRUE16-NEXT: ; implicit-def: $sgpr59_lo16
@@ -39054,7 +39034,7 @@ define inreg <40 x i8> @bitcast_v5i64_to_v40i8_scalar(<5 x i64> inreg %a, i32 in
; GFX11-FAKE16: ; %bb.0:
; GFX11-FAKE16-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX11-FAKE16-NEXT: s_cmp_lg_u32 s22, 0
-; GFX11-FAKE16-NEXT: s_mov_b32 s63, 0
+; GFX11-FAKE16-NEXT: s_mov_b32 s61, 0
; GFX11-FAKE16-NEXT: s_cbranch_scc0 .LBB79_4
; GFX11-FAKE16-NEXT: ; %bb.1: ; %cmp.false
; GFX11-FAKE16-NEXT: s_lshr_b32 s14, s21, 24
@@ -39080,14 +39060,14 @@ define inreg <40 x i8> @bitcast_v5i64_to_v40i8_scalar(<5 x i64> inreg %a, i32 in
; GFX11-FAKE16-NEXT: s_lshr_b32 s58, s1, 24
; GFX11-FAKE16-NEXT: s_lshr_b32 s59, s1, 16
; GFX11-FAKE16-NEXT: s_lshr_b32 s60, s1, 8
-; GFX11-FAKE16-NEXT: s_lshr_b32 s61, s0, 16
-; GFX11-FAKE16-NEXT: s_lshr_b32 s62, s0, 8
+; GFX11-FAKE16-NEXT: s_lshr_b32 s62, s0, 16
+; GFX11-FAKE16-NEXT: s_lshr_b32 s63, s0, 8
; GFX11-FAKE16-NEXT: s_lshr_b64 s[4:5], s[20:21], 24
; GFX11-FAKE16-NEXT: s_lshr_b64 s[6:7], s[18:19], 24
; GFX11-FAKE16-NEXT: s_lshr_b64 s[8:9], s[16:17], 24
; GFX11-FAKE16-NEXT: s_lshr_b64 s[10:11], s[2:3], 24
; GFX11-FAKE16-NEXT: s_lshr_b64 s[12:13], s[0:1], 24
-; GFX11-FAKE16-NEXT: s_and_not1_b32 vcc_lo, exec_lo, s63
+; GFX11-FAKE16-NEXT: s_and_not1_b32 vcc_lo, exec_lo, s61
; GFX11-FAKE16-NEXT: s_cbranch_vccnz .LBB79_3
; GFX11-FAKE16-NEXT: .LBB79_2: ; %cmp.true
; GFX11-FAKE16-NEXT: s_add_u32 s0, s0, 3
@@ -39128,12 +39108,12 @@ define inreg <40 x i8> @bitcast_v5i64_to_v40i8_scalar(<5 x i64> inreg %a, i32 in
; GFX11-FAKE16-NEXT: s_lshr_b32 s58, s1, 24
; GFX11-FAKE16-NEXT: s_lshr_b32 s59, s1, 16
; GFX11-FAKE16-NEXT: s_lshr_b32 s60, s1, 8
-; GFX11-FAKE16-NEXT: s_lshr_b32 s61, s0, 16
-; GFX11-FAKE16-NEXT: s_lshr_b32 s62, s0, 8
+; GFX11-FAKE16-NEXT: s_lshr_b32 s62, s0, 16
+; GFX11-FAKE16-NEXT: s_lshr_b32 s63, s0, 8
; GFX11-FAKE16-NEXT: .LBB79_3: ; %end
; GFX11-FAKE16-NEXT: s_and_b32 s0, s0, 0xff
-; GFX11-FAKE16-NEXT: s_lshl_b32 s5, s62, 8
-; GFX11-FAKE16-NEXT: s_and_b32 s7, s61, 0xff
+; GFX11-FAKE16-NEXT: s_lshl_b32 s5, s63, 8
+; GFX11-FAKE16-NEXT: s_and_b32 s7, s62, 0xff
; GFX11-FAKE16-NEXT: s_lshl_b32 s9, s12, 8
; GFX11-FAKE16-NEXT: s_or_b32 s0, s0, s5
; GFX11-FAKE16-NEXT: s_or_b32 s5, s7, s9
@@ -39233,8 +39213,8 @@ define inreg <40 x i8> @bitcast_v5i64_to_v40i8_scalar(<5 x i64> inreg %a, i32 in
; GFX11-FAKE16-NEXT: scratch_store_b64 v0, v[9:10], off offset:32
; GFX11-FAKE16-NEXT: s_setpc_b64 s[30:31]
; GFX11-FAKE16-NEXT: .LBB79_4:
+; GFX11-FAKE16-NEXT: ; implicit-def: $sgpr63
; GFX11-FAKE16-NEXT: ; implicit-def: $sgpr62
-; GFX11-FAKE16-NEXT: ; implicit-def: $sgpr61
; GFX11-FAKE16-NEXT: ; implicit-def: $sgpr12
; GFX11-FAKE16-NEXT: ; implicit-def: $sgpr60
; GFX11-FAKE16-NEXT: ; implicit-def: $sgpr59
diff --git a/llvm/test/CodeGen/AMDGPU/amdgcn.bitcast.32bit.ll b/llvm/test/CodeGen/AMDGPU/amdgcn.bitcast.32bit.ll
index 7b756bce857bc..fbdee6eea327e 100644
--- a/llvm/test/CodeGen/AMDGPU/amdgcn.bitcast.32bit.ll
+++ b/llvm/test/CodeGen/AMDGPU/amdgcn.bitcast.32bit.ll
@@ -6112,10 +6112,8 @@ define <4 x i8> @bitcast_v2i16_to_v4i8(<2 x i16> %a, i32 %b) {
; VI-LABEL: bitcast_v2i16_to_v4i8:
; VI: ; %bb.0:
; VI-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
-; VI-NEXT: v_mov_b32_e32 v4, v0
; VI-NEXT: v_cmp_ne_u32_e32 vcc, 0, v1
-; VI-NEXT: v_lshrrev_b32_e32 v2, 16, v4
-; VI-NEXT: ; implicit-def: $vgpr0
+; VI-NEXT: v_lshrrev_b32_e32 v2, 16, v0
; VI-NEXT: ; implicit-def: $vgpr1
; VI-NEXT: ; implicit-def: $vgpr3
; VI-NEXT: s_and_saveexec_b64 s[4:5], vcc
@@ -6128,20 +6126,19 @@ define <4 x i8> @bitcast_v2i16_to_v4i8(<2 x i16> %a, i32 %b) {
; VI-NEXT: s_or_b64 exec, exec, s[4:5]
; VI-NEXT: s_setpc_b64 s[30:31]
; VI-NEXT: .LBB56_3: ; %cmp.false
-; VI-NEXT: v_lshrrev_b32_e32 v3, 24, v4
-; VI-NEXT: v_lshrrev_b32_e32 v1, 8, v4
-; VI-NEXT: v_mov_b32_e32 v0, v4
-; VI-NEXT: ; implicit-def: $vgpr4
+; VI-NEXT: v_lshrrev_b32_e32 v3, 24, v0
+; VI-NEXT: v_lshrrev_b32_e32 v1, 8, v0
; VI-NEXT: s_andn2_saveexec_b64 s[4:5], s[4:5]
; VI-NEXT: s_cbranch_execz .LBB56_2
; VI-NEXT: .LBB56_4: ; %cmp.true
; VI-NEXT: v_mov_b32_e32 v1, 3
-; VI-NEXT: v_add_u16_sdwa v2, v4, v1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_1 src1_sel:DWORD
-; VI-NEXT: v_add_u16_e32 v0, 3, v4
-; VI-NEXT: v_lshlrev_b32_e32 v1, 16, v2
-; VI-NEXT: v_or_b32_e32 v1, v0, v1
-; VI-NEXT: v_lshrrev_b32_e32 v1, 8, v1
+; VI-NEXT: v_add_u16_sdwa v2, v0, v1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_1 src1_sel:DWORD
+; VI-NEXT: v_add_u16_e32 v4, 3, v0
+; VI-NEXT: v_lshlrev_b32_e32 v0, 16, v2
+; VI-NEXT: v_or_b32_e32 v0, v4, v0
+; VI-NEXT: v_lshrrev_b32_e32 v1, 8, v0
; VI-NEXT: v_bfe_u32 v3, v2, 8, 8
+; VI-NEXT: v_mov_b32_e32 v0, v4
; VI-NEXT: s_or_b64 exec, exec, s[4:5]
; VI-NEXT: s_setpc_b64 s[30:31]
;
@@ -6292,31 +6289,30 @@ define inreg <4 x i8> @bitcast_v2i16_to_v4i8_scalar(<2 x i16> inreg %a, i32 inre
; VI-NEXT: s_cmp_lg_u32 s17, 0
; VI-NEXT: s_cbranch_scc0 .LBB57_4
; VI-NEXT: ; %bb.1: ; %cmp.false
-; VI-NEXT: s_lshr_b32 s8, s16, 24
+; VI-NEXT: s_lshr_b32 s7, s16, 24
; VI-NEXT: s_lshr_b32 s6, s16, 16
-; VI-NEXT: s_lshr_b32 s9, s16, 8
-; VI-NEXT: s_mov_b32 s7, s16
+; VI-NEXT: s_lshr_b32 s8, s16, 8
; VI-NEXT: s_cbranch_execnz .LBB57_3
; VI-NEXT: .LBB57_2: ; %cmp.true
-; VI-NEXT: s_lshr_b32 s5, s16, 16
-; VI-NEXT: s_add_i32 s7, s16, 3
-; VI-NEXT: s_add_i32 s6, s5, 3
-; VI-NEXT: s_and_b32 s4, s7, 0xffff
-; VI-NEXT: s_lshl_b32 s5, s6, 16
-; VI-NEXT: s_or_b32 s4, s4, s5
-; VI-NEXT: s_lshr_b32 s9, s4, 8
-; VI-NEXT: s_bfe_u32 s8, s6, 0x80008
+; VI-NEXT: s_lshr_b32 s6, s16, 16
+; VI-NEXT: s_add_i32 s4, s16, 3
+; VI-NEXT: s_add_i32 s6, s6, 3
+; VI-NEXT: s_and_b32 s5, s4, 0xffff
+; VI-NEXT: s_lshl_b32 s7, s6, 16
+; VI-NEXT: s_or_b32 s5, s5, s7
+; VI-NEXT: s_mov_b32 s16, s4
+; VI-NEXT: s_lshr_b32 s8, s5, 8
+; VI-NEXT: s_bfe_u32 s7, s6, 0x80008
; VI-NEXT: .LBB57_3: ; %end
-; VI-NEXT: v_mov_b32_e32 v0, s7
-; VI-NEXT: v_mov_b32_e32 v1, s9
+; VI-NEXT: v_mov_b32_e32 v0, s16
+; VI-NEXT: v_mov_b32_e32 v1, s8
; VI-NEXT: v_mov_b32_e32 v2, s6
-; VI-NEXT: v_mov_b32_e32 v3, s8
+; VI-NEXT: v_mov_b32_e32 v3, s7
; VI-NEXT: s_setpc_b64 s[30:31]
; VI-NEXT: .LBB57_4:
-; VI-NEXT: ; implicit-def: $sgpr7
-; VI-NEXT: ; implicit-def: $sgpr9
-; VI-NEXT: ; implicit-def: $sgpr6
; VI-NEXT: ; implicit-def: $sgpr8
+; VI-NEXT: ; implicit-def: $sgpr6
+; VI-NEXT: ; implicit-def: $sgpr7
; VI-NEXT: s_branch .LBB57_2
;
; GFX9-LABEL: bitcast_v2i16_to_v4i8_scalar:
@@ -9174,8 +9170,8 @@ define <4 x i8> @bitcast_v2bf16_to_v4i8(<2 x bfloat> %a, i32 %b) {
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_2)
; GFX11-TRUE16-NEXT: v_lshrrev_b32_e32 v3, 24, v4
; GFX11-TRUE16-NEXT: v_lshrrev_b32_e32 v1, 8, v4
-; GFX11-TRUE16-NEXT: v_mov_b16_e32 v0.h, v4.l
; GFX11-TRUE16-NEXT: v_mov_b16_e32 v2.h, v4.h
+; GFX11-TRUE16-NEXT: v_mov_b16_e32 v0.h, v4.l
; GFX11-TRUE16-NEXT: ; implicit-def: $vgpr4
; GFX11-TRUE16-NEXT: ; %bb.2: ; %Flow
; GFX11-TRUE16-NEXT: s_and_not1_saveexec_b32 s0, s0
@@ -9204,6 +9200,7 @@ define <4 x i8> @bitcast_v2bf16_to_v4i8(<2 x bfloat> %a, i32 %b) {
; GFX11-TRUE16-NEXT: v_lshrrev_b32_e32 v1, 8, v1
; GFX11-TRUE16-NEXT: .LBB76_4: ; %end
; GFX11-TRUE16-NEXT: s_or_b32 exec_lo, exec_lo, s0
+; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1)
; GFX11-TRUE16-NEXT: v_mov_b16_e32 v0.l, v0.h
; GFX11-TRUE16-NEXT: v_mov_b16_e32 v2.l, v2.h
; GFX11-TRUE16-NEXT: s_setpc_b64 s[30:31]
diff --git a/llvm/test/CodeGen/AMDGPU/amdgcn.bitcast.512bit.ll b/llvm/test/CodeGen/AMDGPU/amdgcn.bitcast.512bit.ll
index 1024c2a7f066a..68498462f489e 100644
--- a/llvm/test/CodeGen/AMDGPU/amdgcn.bitcast.512bit.ll
+++ b/llvm/test/CodeGen/AMDGPU/amdgcn.bitcast.512bit.ll
@@ -64649,11 +64649,10 @@ define <64 x i8> @bitcast_v32i16_to_v64i8(<32 x i16> %a, i32 %b) {
; SI-NEXT: buffer_store_dword v61, off, s[0:3], s32 offset:20 ; 4-byte Folded Spill
; SI-NEXT: buffer_store_dword v62, off, s[0:3], s32 offset:16 ; 4-byte Folded Spill
; SI-NEXT: buffer_store_dword v63, off, s[0:3], s32 offset:12 ; 4-byte Folded Spill
-; SI-NEXT: buffer_store_dword v27, off, s[0:3], s32 offset:204 ; 4-byte Folded Spill
; SI-NEXT: buffer_load_dword v31, off, s[0:3], s32 offset:8
; SI-NEXT: buffer_load_dword v44, off, s[0:3], s32 offset:4
; SI-NEXT: buffer_load_dword v35, off, s[0:3], s32
-; SI-NEXT: s_waitcnt expcnt(1)
+; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_lshlrev_b32_e32 v63, 16, v2
; SI-NEXT: ; implicit-def: $vgpr2
; SI-NEXT: ; kill: killed $vgpr2
@@ -64728,9 +64727,10 @@ define <64 x i8> @bitcast_v32i16_to_v64i8(<32 x i16> %a, i32 %b) {
; SI-NEXT: ; implicit-def: $vgpr2
; SI-NEXT: ; kill: killed $vgpr2
; SI-NEXT: ; implicit-def: $vgpr2
-; SI-NEXT: ; implicit-def: $vgpr62
-; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_lshlrev_b32_e32 v27, 16, v4
+; SI-NEXT: ; kill: killed $vgpr2
+; SI-NEXT: ; implicit-def: $vgpr2
+; SI-NEXT: ; implicit-def: $vgpr58
+; SI-NEXT: v_lshlrev_b32_e32 v62, 16, v4
; SI-NEXT: v_lshlrev_b32_e32 v37, 16, v6
; SI-NEXT: v_lshlrev_b32_e32 v36, 16, v8
; SI-NEXT: v_lshlrev_b32_e32 v48, 16, v10
@@ -64746,8 +64746,8 @@ define <64 x i8> @bitcast_v32i16_to_v64i8(<32 x i16> %a, i32 %b) {
; SI-NEXT: v_lshlrev_b32_e32 v43, 16, v30
; SI-NEXT: ; kill: killed $vgpr2
; SI-NEXT: ; implicit-def: $vgpr2
-; SI-NEXT: ; kill: killed $vgpr62
-; SI-NEXT: ; implicit-def: $vgpr62
+; SI-NEXT: ; kill: killed $vgpr58
+; SI-NEXT: ; implicit-def: $vgpr58
; SI-NEXT: ; implicit-def: $vgpr56
; SI-NEXT: ; implicit-def: $vgpr47
; SI-NEXT: ; implicit-def: $vgpr50
@@ -64758,11 +64758,10 @@ define <64 x i8> @bitcast_v32i16_to_v64i8(<32 x i16> %a, i32 %b) {
; SI-NEXT: ; implicit-def: $vgpr30
; SI-NEXT: ; implicit-def: $vgpr26
; SI-NEXT: ; implicit-def: $vgpr22
-; SI-NEXT: ; implicit-def: $vgpr61
; SI-NEXT: ; implicit-def: $vgpr60
; SI-NEXT: ; implicit-def: $vgpr18
; SI-NEXT: ; implicit-def: $vgpr14
-; SI-NEXT: ; implicit-def: $vgpr58
+; SI-NEXT: ; implicit-def: $vgpr61
; SI-NEXT: ; implicit-def: $vgpr57
; SI-NEXT: ; implicit-def: $vgpr10
; SI-NEXT: ; kill: killed $vgpr2
@@ -64771,8 +64770,8 @@ define <64 x i8> @bitcast_v32i16_to_v64i8(<32 x i16> %a, i32 %b) {
; SI-NEXT: ; implicit-def: $vgpr46
; SI-NEXT: ; implicit-def: $vgpr45
; SI-NEXT: ; implicit-def: $vgpr2
-; SI-NEXT: ; kill: killed $vgpr62
-; SI-NEXT: ; implicit-def: $vgpr62
+; SI-NEXT: ; kill: killed $vgpr58
+; SI-NEXT: ; implicit-def: $vgpr58
; SI-NEXT: s_waitcnt vmcnt(2)
; SI-NEXT: v_cmp_ne_u32_e32 vcc, 0, v31
; SI-NEXT: s_waitcnt vmcnt(1)
@@ -64785,207 +64784,204 @@ define <64 x i8> @bitcast_v32i16_to_v64i8(<32 x i16> %a, i32 %b) {
; SI-NEXT: v_and_b32_e32 v1, 0xffff, v1
; SI-NEXT: v_or_b32_e32 v56, v1, v63
; SI-NEXT: v_and_b32_e32 v1, 0xffff, v3
-; SI-NEXT: v_or_b32_e32 v47, v1, v27
+; SI-NEXT: v_or_b32_e32 v47, v1, v62
; SI-NEXT: v_alignbit_b32 v1, v47, v56, 24
-; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:236 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:232 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_alignbit_b32 v1, v47, v56, 16
-; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:104 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:124 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_alignbit_b32 v1, v47, v56, 8
-; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:108 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:128 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_and_b32_e32 v1, 0xffff, v5
; SI-NEXT: v_or_b32_e32 v50, v1, v37
; SI-NEXT: v_and_b32_e32 v1, 0xffff, v7
; SI-NEXT: v_or_b32_e32 v38, v1, v36
; SI-NEXT: v_alignbit_b32 v1, v38, v50, 24
-; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:88 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:108 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_alignbit_b32 v1, v38, v50, 16
-; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:96 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:116 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_alignbit_b32 v1, v38, v50, 8
-; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:100 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:120 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_and_b32_e32 v1, 0xffff, v9
; SI-NEXT: v_or_b32_e32 v34, v1, v48
; SI-NEXT: v_and_b32_e32 v1, 0xffff, v11
; SI-NEXT: v_or_b32_e32 v33, v1, v39
; SI-NEXT: v_alignbit_b32 v1, v33, v34, 24
-; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:76 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:96 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_alignbit_b32 v1, v33, v34, 16
-; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:80 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:100 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_alignbit_b32 v1, v33, v34, 8
-; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:92 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:112 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_and_b32_e32 v1, 0xffff, v13
; SI-NEXT: v_or_b32_e32 v32, v1, v51
; SI-NEXT: v_and_b32_e32 v1, 0xffff, v15
; SI-NEXT: v_or_b32_e32 v31, v1, v49
; SI-NEXT: v_alignbit_b32 v1, v31, v32, 24
-; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:224 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:84 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_alignbit_b32 v1, v31, v32, 16
-; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:228 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:88 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_alignbit_b32 v1, v31, v32, 8
-; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:84 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:104 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_and_b32_e32 v1, 0xffff, v17
; SI-NEXT: v_or_b32_e32 v30, v1, v53
; SI-NEXT: v_and_b32_e32 v1, 0xffff, v19
; SI-NEXT: v_or_b32_e32 v26, v1, v52
; SI-NEXT: v_alignbit_b32 v1, v26, v30, 24
-; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:212 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:228 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_alignbit_b32 v1, v26, v30, 16
-; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:216 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:76 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_alignbit_b32 v1, v26, v30, 8
-; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:232 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:92 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_and_b32_e32 v1, 0xffff, v21
; SI-NEXT: v_or_b32_e32 v22, v1, v55
; SI-NEXT: v_and_b32_e32 v1, 0xffff, v23
; SI-NEXT: v_or_b32_e32 v18, v1, v54
+; SI-NEXT: v_alignbit_b32 v1, v18, v22, 16
+; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:236 ; 4-byte Folded Spill
+; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_alignbit_b32 v1, v18, v22, 8
-; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:220 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:80 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_and_b32_e32 v1, 0xffff, v25
; SI-NEXT: v_or_b32_e32 v14, v1, v41
-; SI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:204 ; 4-byte Folded Reload
-; SI-NEXT: v_alignbit_b32 v60, v18, v22, 24
-; SI-NEXT: v_alignbit_b32 v61, v18, v22, 16
-; SI-NEXT: v_bfe_u32 v62, v44, 8, 8
-; SI-NEXT: ; implicit-def: $vgpr3
-; SI-NEXT: ; implicit-def: $vgpr5
-; SI-NEXT: ; implicit-def: $vgpr7
-; SI-NEXT: ; implicit-def: $vgpr9
-; SI-NEXT: ; implicit-def: $vgpr11
-; SI-NEXT: ; implicit-def: $vgpr13
-; SI-NEXT: ; implicit-def: $vgpr15
-; SI-NEXT: ; implicit-def: $vgpr17
-; SI-NEXT: ; implicit-def: $vgpr19
-; SI-NEXT: ; implicit-def: $vgpr21
-; SI-NEXT: ; implicit-def: $vgpr23
-; SI-NEXT: ; implicit-def: $vgpr25
-; SI-NEXT: ; implicit-def: $vgpr63
-; SI-NEXT: ; implicit-def: $vgpr27
-; SI-NEXT: ; implicit-def: $vgpr37
-; SI-NEXT: ; implicit-def: $vgpr36
-; SI-NEXT: ; implicit-def: $vgpr48
-; SI-NEXT: ; implicit-def: $vgpr39
-; SI-NEXT: ; implicit-def: $vgpr51
-; SI-NEXT: ; implicit-def: $vgpr49
-; SI-NEXT: ; implicit-def: $vgpr53
-; SI-NEXT: ; implicit-def: $vgpr52
-; SI-NEXT: ; implicit-def: $vgpr55
-; SI-NEXT: ; implicit-def: $vgpr54
-; SI-NEXT: ; implicit-def: $vgpr41
-; SI-NEXT: s_waitcnt vmcnt(0)
-; SI-NEXT: v_and_b32_e32 v1, 0xffff, v1
+; SI-NEXT: v_and_b32_e32 v1, 0xffff, v27
; SI-NEXT: v_or_b32_e32 v10, v1, v40
; SI-NEXT: v_alignbit_b32 v1, v10, v14, 8
-; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:208 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:224 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_and_b32_e32 v1, 0xffff, v29
; SI-NEXT: v_or_b32_e32 v6, v1, v43
+; SI-NEXT: s_waitcnt vmcnt(14)
; SI-NEXT: v_and_b32_e32 v1, 0xffff, v35
; SI-NEXT: v_or_b32_e32 v2, v1, v42
; SI-NEXT: v_lshrrev_b32_e32 v1, 8, v47
-; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:188 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:208 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_lshrrev_b32_e32 v1, 8, v38
-; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:168 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:188 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_lshrrev_b32_e32 v1, 8, v33
-; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:152 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:172 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_lshrrev_b32_e32 v1, 8, v31
-; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:136 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:156 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_and_b32_e32 v1, 0xffff, v4
-; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:196 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:216 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_and_b32_e32 v1, 0xffff, v8
-; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:184 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:204 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_and_b32_e32 v1, 0xffff, v12
-; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:172 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:192 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_and_b32_e32 v1, 0xffff, v16
-; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:156 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:176 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_and_b32_e32 v1, 0xffff, v20
-; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:140 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:160 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_and_b32_e32 v1, 0xffff, v24
-; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:124 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:144 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_and_b32_e32 v1, 0xffff, v28
-; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:116 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:136 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_and_b32_e32 v1, 0xffff, v44
-; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:112 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:132 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_bfe_u32 v1, v4, 8, 8
-; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:200 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:220 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_bfe_u32 v1, v8, 8, 8
-; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:192 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:212 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_bfe_u32 v1, v12, 8, 8
-; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:180 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:200 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_bfe_u32 v1, v16, 8, 8
-; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:164 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:184 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_bfe_u32 v1, v20, 8, 8
-; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:148 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:168 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_bfe_u32 v1, v24, 8, 8
-; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:132 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:152 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_bfe_u32 v1, v28, 8, 8
-; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:120 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:140 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_lshrrev_b32_e32 v1, 8, v26
-; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:176 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:196 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_lshrrev_b32_e32 v1, 8, v18
-; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:160 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:180 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_lshrrev_b32_e32 v1, 8, v10
-; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:144 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:164 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_lshrrev_b32_e32 v1, 8, v2
+; SI-NEXT: v_alignbit_b32 v60, v18, v22, 24
; SI-NEXT: v_alignbit_b32 v57, v10, v14, 24
-; SI-NEXT: v_alignbit_b32 v58, v10, v14, 16
+; SI-NEXT: v_alignbit_b32 v61, v10, v14, 16
; SI-NEXT: v_alignbit_b32 v45, v2, v6, 24
; SI-NEXT: v_alignbit_b32 v46, v2, v6, 16
; SI-NEXT: v_alignbit_b32 v59, v2, v6, 8
-; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:128 ; 4-byte Folded Spill
+; SI-NEXT: v_bfe_u32 v58, v44, 8, 8
+; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:148 ; 4-byte Folded Spill
; SI-NEXT: ; implicit-def: $vgpr1
-; SI-NEXT: ; implicit-def: $vgpr4
-; SI-NEXT: ; kill: killed $vgpr4
+; SI-NEXT: ; implicit-def: $vgpr3
+; SI-NEXT: ; implicit-def: $vgpr5
+; SI-NEXT: ; implicit-def: $vgpr7
+; SI-NEXT: ; implicit-def: $vgpr9
+; SI-NEXT: ; implicit-def: $vgpr11
+; SI-NEXT: ; implicit-def: $vgpr13
+; SI-NEXT: ; implicit-def: $vgpr15
+; SI-NEXT: ; implicit-def: $vgpr17
+; SI-NEXT: ; implicit-def: $vgpr19
+; SI-NEXT: ; implicit-def: $vgpr21
+; SI-NEXT: ; implicit-def: $vgpr23
+; SI-NEXT: ; implicit-def: $vgpr25
+; SI-NEXT: ; implicit-def: $vgpr27
; SI-NEXT: ; implicit-def: $vgpr29
; SI-NEXT: ; implicit-def: $vgpr35
+; SI-NEXT: ; implicit-def: $vgpr63
+; SI-NEXT: ; implicit-def: $vgpr62
+; SI-NEXT: ; implicit-def: $vgpr37
+; SI-NEXT: ; implicit-def: $vgpr36
+; SI-NEXT: ; implicit-def: $vgpr48
+; SI-NEXT: ; implicit-def: $vgpr39
+; SI-NEXT: ; implicit-def: $vgpr51
+; SI-NEXT: ; implicit-def: $vgpr49
+; SI-NEXT: ; implicit-def: $vgpr53
+; SI-NEXT: ; implicit-def: $vgpr52
+; SI-NEXT: ; implicit-def: $vgpr55
+; SI-NEXT: ; implicit-def: $vgpr54
+; SI-NEXT: ; implicit-def: $vgpr41
; SI-NEXT: ; implicit-def: $vgpr40
; SI-NEXT: ; implicit-def: $vgpr43
; SI-NEXT: ; implicit-def: $vgpr42
; SI-NEXT: .LBB96_2: ; %Flow
; SI-NEXT: s_or_saveexec_b64 s[4:5], s[4:5]
-; SI-NEXT: buffer_load_dword v8, off, s[0:3], s32 offset:208 ; 4-byte Folded Reload
-; SI-NEXT: buffer_load_dword v12, off, s[0:3], s32 offset:212 ; 4-byte Folded Reload
-; SI-NEXT: buffer_load_dword v16, off, s[0:3], s32 offset:216 ; 4-byte Folded Reload
-; SI-NEXT: buffer_load_dword v20, off, s[0:3], s32 offset:220 ; 4-byte Folded Reload
-; SI-NEXT: buffer_load_dword v24, off, s[0:3], s32 offset:224 ; 4-byte Folded Reload
-; SI-NEXT: buffer_load_dword v28, off, s[0:3], s32 offset:228 ; 4-byte Folded Reload
-; SI-NEXT: buffer_load_dword v44, off, s[0:3], s32 offset:232 ; 4-byte Folded Reload
-; SI-NEXT: buffer_load_dword v4, off, s[0:3], s32 offset:236 ; 4-byte Folded Reload
+; SI-NEXT: v_mov_b32_e32 v20, v61
+; SI-NEXT: buffer_load_dword v61, off, s[0:3], s32 offset:236 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v28, off, s[0:3], s32 offset:224 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v44, off, s[0:3], s32 offset:228 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v4, off, s[0:3], s32 offset:232 ; 4-byte Folded Reload
; SI-NEXT: s_xor_b64 exec, exec, s[4:5]
; SI-NEXT: s_cbranch_execz .LBB96_4
; SI-NEXT: ; %bb.3: ; %cmp.true
@@ -64995,35 +64991,7 @@ define <64 x i8> @bitcast_v32i16_to_v64i8(<32 x i16> %a, i32 %b) {
; SI-NEXT: s_mov_b32 s6, 0x30000
; SI-NEXT: v_or_b32_e32 v4, v41, v4
; SI-NEXT: v_add_i32_e32 v14, vcc, s6, v4
-; SI-NEXT: buffer_load_dword v4, off, s[0:3], s32 offset:204 ; 4-byte Folded Reload
-; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_add_i32_e32 v1, vcc, 3, v1
-; SI-NEXT: v_and_b32_e32 v1, 0xffff, v1
-; SI-NEXT: v_or_b32_e32 v1, v63, v1
-; SI-NEXT: v_add_i32_e32 v56, vcc, s6, v1
-; SI-NEXT: v_add_i32_e32 v1, vcc, 3, v3
-; SI-NEXT: v_and_b32_e32 v1, 0xffff, v1
-; SI-NEXT: v_or_b32_e32 v1, v27, v1
-; SI-NEXT: v_add_i32_e32 v47, vcc, s6, v1
-; SI-NEXT: v_alignbit_b32 v1, v47, v56, 16
-; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:104 ; 4-byte Folded Spill
-; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_alignbit_b32 v1, v47, v56, 8
-; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:108 ; 4-byte Folded Spill
-; SI-NEXT: v_add_i32_e32 v2, vcc, 3, v29
-; SI-NEXT: v_and_b32_e32 v2, 0xffff, v2
-; SI-NEXT: v_or_b32_e32 v2, v43, v2
-; SI-NEXT: v_add_i32_e32 v6, vcc, 0x30000, v2
-; SI-NEXT: v_add_i32_e32 v2, vcc, 3, v35
-; SI-NEXT: v_and_b32_e32 v2, 0xffff, v2
-; SI-NEXT: v_or_b32_e32 v2, v42, v2
-; SI-NEXT: v_add_i32_e32 v2, vcc, s6, v2
-; SI-NEXT: v_alignbit_b32 v45, v2, v6, 24
-; SI-NEXT: v_alignbit_b32 v46, v2, v6, 16
-; SI-NEXT: v_alignbit_b32 v59, v2, v6, 8
-; SI-NEXT: v_lshrrev_b32_e32 v62, 24, v2
-; SI-NEXT: s_waitcnt vmcnt(2)
-; SI-NEXT: v_add_i32_e32 v4, vcc, 3, v4
+; SI-NEXT: v_add_i32_e32 v4, vcc, 3, v27
; SI-NEXT: v_and_b32_e32 v4, 0xffff, v4
; SI-NEXT: v_or_b32_e32 v4, v40, v4
; SI-NEXT: v_add_i32_e32 v10, vcc, s6, v4
@@ -65057,121 +65025,157 @@ define <64 x i8> @bitcast_v32i16_to_v64i8(<32 x i16> %a, i32 %b) {
; SI-NEXT: v_add_i32_e32 v34, vcc, s6, v4
; SI-NEXT: v_add_i32_e32 v4, vcc, 3, v11
; SI-NEXT: v_and_b32_e32 v4, 0xffff, v4
+; SI-NEXT: s_waitcnt expcnt(0)
+; SI-NEXT: v_add_i32_e32 v1, vcc, 3, v1
; SI-NEXT: v_or_b32_e32 v4, v39, v4
+; SI-NEXT: v_and_b32_e32 v1, 0xffff, v1
; SI-NEXT: v_add_i32_e32 v33, vcc, s6, v4
; SI-NEXT: v_add_i32_e32 v4, vcc, 3, v5
+; SI-NEXT: v_or_b32_e32 v1, v63, v1
; SI-NEXT: v_and_b32_e32 v4, 0xffff, v4
+; SI-NEXT: v_add_i32_e32 v56, vcc, s6, v1
+; SI-NEXT: v_add_i32_e32 v1, vcc, 3, v3
; SI-NEXT: v_or_b32_e32 v4, v37, v4
+; SI-NEXT: v_and_b32_e32 v1, 0xffff, v1
; SI-NEXT: v_add_i32_e32 v50, vcc, s6, v4
; SI-NEXT: v_add_i32_e32 v4, vcc, 3, v7
+; SI-NEXT: v_or_b32_e32 v1, v62, v1
; SI-NEXT: v_and_b32_e32 v4, 0xffff, v4
+; SI-NEXT: v_add_i32_e32 v47, vcc, s6, v1
; SI-NEXT: v_or_b32_e32 v4, v36, v4
+; SI-NEXT: v_alignbit_b32 v1, v47, v56, 16
; SI-NEXT: v_add_i32_e32 v38, vcc, s6, v4
+; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:124 ; 4-byte Folded Spill
+; SI-NEXT: s_waitcnt expcnt(0)
+; SI-NEXT: v_alignbit_b32 v1, v47, v56, 8
+; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:128 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_alignbit_b32 v1, v38, v50, 24
-; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:88 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:108 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_alignbit_b32 v1, v38, v50, 16
-; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:96 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:116 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_alignbit_b32 v1, v38, v50, 8
-; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:100 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:120 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_alignbit_b32 v1, v33, v34, 24
-; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:76 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:96 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_alignbit_b32 v1, v33, v34, 16
-; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:80 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:100 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_alignbit_b32 v1, v33, v34, 8
-; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:92 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:112 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_alignbit_b32 v1, v31, v32, 8
+; SI-NEXT: v_alignbit_b32 v1, v31, v32, 24
; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:84 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
+; SI-NEXT: v_alignbit_b32 v1, v31, v32, 16
+; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:88 ; 4-byte Folded Spill
+; SI-NEXT: s_waitcnt expcnt(0)
+; SI-NEXT: v_alignbit_b32 v1, v31, v32, 8
+; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:104 ; 4-byte Folded Spill
+; SI-NEXT: s_waitcnt expcnt(0)
+; SI-NEXT: v_alignbit_b32 v1, v26, v30, 16
+; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:76 ; 4-byte Folded Spill
+; SI-NEXT: s_waitcnt expcnt(0)
+; SI-NEXT: v_alignbit_b32 v1, v26, v30, 8
+; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:92 ; 4-byte Folded Spill
+; SI-NEXT: s_waitcnt expcnt(0)
+; SI-NEXT: v_alignbit_b32 v1, v18, v22, 8
+; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:80 ; 4-byte Folded Spill
+; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_lshrrev_b32_e32 v1, 24, v47
-; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:200 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:220 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_lshrrev_b32_e32 v1, 16, v47
-; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:196 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:216 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_lshrrev_b32_e32 v1, 8, v47
-; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:188 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:208 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_lshrrev_b32_e32 v1, 24, v38
-; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:192 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:212 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_lshrrev_b32_e32 v1, 16, v38
-; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:184 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:204 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_lshrrev_b32_e32 v1, 8, v38
-; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:168 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:188 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_lshrrev_b32_e32 v1, 24, v33
-; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:180 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:200 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_lshrrev_b32_e32 v1, 16, v33
-; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:172 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:192 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_lshrrev_b32_e32 v1, 8, v33
-; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:152 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:172 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_lshrrev_b32_e32 v1, 24, v31
-; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:164 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:184 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_lshrrev_b32_e32 v1, 16, v31
-; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:156 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:176 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_lshrrev_b32_e32 v1, 8, v31
-; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:136 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:156 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_lshrrev_b32_e32 v1, 24, v26
-; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:148 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:168 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_lshrrev_b32_e32 v1, 16, v26
-; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:140 ; 4-byte Folded Spill
+; SI-NEXT: v_add_i32_e32 v2, vcc, 3, v29
+; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:160 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_lshrrev_b32_e32 v1, 8, v26
-; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:176 ; 4-byte Folded Spill
+; SI-NEXT: v_and_b32_e32 v2, 0xffff, v2
+; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:196 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_lshrrev_b32_e32 v1, 24, v18
-; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:132 ; 4-byte Folded Spill
+; SI-NEXT: v_or_b32_e32 v2, v43, v2
+; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:152 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_lshrrev_b32_e32 v1, 16, v18
-; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:124 ; 4-byte Folded Spill
+; SI-NEXT: v_add_i32_e32 v6, vcc, 0x30000, v2
+; SI-NEXT: v_add_i32_e32 v2, vcc, 3, v35
+; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:144 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_lshrrev_b32_e32 v1, 8, v18
-; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:160 ; 4-byte Folded Spill
+; SI-NEXT: v_and_b32_e32 v2, 0xffff, v2
+; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:180 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_lshrrev_b32_e32 v1, 24, v10
-; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:120 ; 4-byte Folded Spill
+; SI-NEXT: v_or_b32_e32 v2, v42, v2
+; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:140 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_lshrrev_b32_e32 v1, 16, v10
-; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:116 ; 4-byte Folded Spill
+; SI-NEXT: v_add_i32_e32 v2, vcc, s6, v2
+; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:136 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_lshrrev_b32_e32 v1, 8, v10
-; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:144 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:164 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_lshrrev_b32_e32 v1, 16, v2
; SI-NEXT: v_alignbit_b32 v4, v47, v56, 24
-; SI-NEXT: v_alignbit_b32 v24, v31, v32, 24
-; SI-NEXT: v_alignbit_b32 v28, v31, v32, 16
-; SI-NEXT: v_alignbit_b32 v12, v26, v30, 24
-; SI-NEXT: v_alignbit_b32 v16, v26, v30, 16
-; SI-NEXT: v_alignbit_b32 v44, v26, v30, 8
+; SI-NEXT: v_alignbit_b32 v44, v26, v30, 24
; SI-NEXT: v_alignbit_b32 v60, v18, v22, 24
; SI-NEXT: v_alignbit_b32 v61, v18, v22, 16
-; SI-NEXT: v_alignbit_b32 v20, v18, v22, 8
; SI-NEXT: v_alignbit_b32 v57, v10, v14, 24
-; SI-NEXT: v_alignbit_b32 v58, v10, v14, 16
-; SI-NEXT: v_alignbit_b32 v8, v10, v14, 8
-; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:112 ; 4-byte Folded Spill
+; SI-NEXT: v_alignbit_b32 v20, v10, v14, 16
+; SI-NEXT: v_alignbit_b32 v28, v10, v14, 8
+; SI-NEXT: v_alignbit_b32 v45, v2, v6, 24
+; SI-NEXT: v_alignbit_b32 v46, v2, v6, 16
+; SI-NEXT: v_alignbit_b32 v59, v2, v6, 8
+; SI-NEXT: v_lshrrev_b32_e32 v58, 24, v2
+; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:132 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_lshrrev_b32_e32 v1, 8, v2
-; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:128 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:148 ; 4-byte Folded Spill
; SI-NEXT: .LBB96_4: ; %end
; SI-NEXT: s_or_b64 exec, exec, s[4:5]
-; SI-NEXT: buffer_load_dword v3, off, s[0:3], s32 offset:108 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v3, off, s[0:3], s32 offset:128 ; 4-byte Folded Reload
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_and_b32_e32 v1, 0xff, v56
; SI-NEXT: s_waitcnt vmcnt(1)
@@ -65179,7 +65183,7 @@ define <64 x i8> @bitcast_v32i16_to_v64i8(<32 x i16> %a, i32 %b) {
; SI-NEXT: s_waitcnt vmcnt(0)
; SI-NEXT: v_lshlrev_b32_e32 v3, 8, v3
; SI-NEXT: v_or_b32_e32 v1, v1, v3
-; SI-NEXT: buffer_load_dword v3, off, s[0:3], s32 offset:104 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v3, off, s[0:3], s32 offset:124 ; 4-byte Folded Reload
; SI-NEXT: v_and_b32_e32 v1, 0xffff, v1
; SI-NEXT: s_waitcnt vmcnt(0)
; SI-NEXT: v_and_b32_e32 v3, 0xff, v3
@@ -65187,14 +65191,14 @@ define <64 x i8> @bitcast_v32i16_to_v64i8(<32 x i16> %a, i32 %b) {
; SI-NEXT: v_or_b32_e32 v3, v4, v3
; SI-NEXT: v_or_b32_e32 v1, v1, v3
; SI-NEXT: buffer_store_dword v1, v0, s[0:3], 0 offen
-; SI-NEXT: buffer_load_dword v3, off, s[0:3], s32 offset:188 ; 4-byte Folded Reload
-; SI-NEXT: buffer_load_dword v4, off, s[0:3], s32 offset:200 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v3, off, s[0:3], s32 offset:208 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v4, off, s[0:3], s32 offset:220 ; 4-byte Folded Reload
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_and_b32_e32 v1, 0xff, v47
; SI-NEXT: s_waitcnt vmcnt(1)
; SI-NEXT: v_lshlrev_b32_e32 v3, 8, v3
; SI-NEXT: v_or_b32_e32 v1, v1, v3
-; SI-NEXT: buffer_load_dword v3, off, s[0:3], s32 offset:196 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v3, off, s[0:3], s32 offset:216 ; 4-byte Folded Reload
; SI-NEXT: s_waitcnt vmcnt(1)
; SI-NEXT: v_lshlrev_b32_e32 v4, 24, v4
; SI-NEXT: v_and_b32_e32 v1, 0xffff, v1
@@ -65205,14 +65209,14 @@ define <64 x i8> @bitcast_v32i16_to_v64i8(<32 x i16> %a, i32 %b) {
; SI-NEXT: v_or_b32_e32 v1, v1, v3
; SI-NEXT: v_add_i32_e32 v3, vcc, 4, v0
; SI-NEXT: buffer_store_dword v1, v3, s[0:3], 0 offen
-; SI-NEXT: buffer_load_dword v3, off, s[0:3], s32 offset:100 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v3, off, s[0:3], s32 offset:120 ; 4-byte Folded Reload
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_and_b32_e32 v1, 0xff, v50
-; SI-NEXT: buffer_load_dword v4, off, s[0:3], s32 offset:88 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v4, off, s[0:3], s32 offset:108 ; 4-byte Folded Reload
; SI-NEXT: s_waitcnt vmcnt(1)
; SI-NEXT: v_lshlrev_b32_e32 v3, 8, v3
; SI-NEXT: v_or_b32_e32 v1, v1, v3
-; SI-NEXT: buffer_load_dword v3, off, s[0:3], s32 offset:96 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v3, off, s[0:3], s32 offset:116 ; 4-byte Folded Reload
; SI-NEXT: v_and_b32_e32 v1, 0xffff, v1
; SI-NEXT: s_waitcnt vmcnt(1)
; SI-NEXT: v_lshlrev_b32_e32 v4, 24, v4
@@ -65223,14 +65227,14 @@ define <64 x i8> @bitcast_v32i16_to_v64i8(<32 x i16> %a, i32 %b) {
; SI-NEXT: v_or_b32_e32 v1, v1, v3
; SI-NEXT: v_add_i32_e32 v3, vcc, 8, v0
; SI-NEXT: buffer_store_dword v1, v3, s[0:3], 0 offen
-; SI-NEXT: buffer_load_dword v3, off, s[0:3], s32 offset:168 ; 4-byte Folded Reload
-; SI-NEXT: buffer_load_dword v4, off, s[0:3], s32 offset:192 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v3, off, s[0:3], s32 offset:188 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v4, off, s[0:3], s32 offset:212 ; 4-byte Folded Reload
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_and_b32_e32 v1, 0xff, v38
; SI-NEXT: s_waitcnt vmcnt(1)
; SI-NEXT: v_lshlrev_b32_e32 v3, 8, v3
; SI-NEXT: v_or_b32_e32 v1, v1, v3
-; SI-NEXT: buffer_load_dword v3, off, s[0:3], s32 offset:184 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v3, off, s[0:3], s32 offset:204 ; 4-byte Folded Reload
; SI-NEXT: s_waitcnt vmcnt(1)
; SI-NEXT: v_lshlrev_b32_e32 v4, 24, v4
; SI-NEXT: v_and_b32_e32 v1, 0xffff, v1
@@ -65241,14 +65245,14 @@ define <64 x i8> @bitcast_v32i16_to_v64i8(<32 x i16> %a, i32 %b) {
; SI-NEXT: v_or_b32_e32 v1, v1, v3
; SI-NEXT: v_add_i32_e32 v3, vcc, 12, v0
; SI-NEXT: buffer_store_dword v1, v3, s[0:3], 0 offen
-; SI-NEXT: buffer_load_dword v3, off, s[0:3], s32 offset:92 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v3, off, s[0:3], s32 offset:112 ; 4-byte Folded Reload
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_and_b32_e32 v1, 0xff, v34
-; SI-NEXT: buffer_load_dword v4, off, s[0:3], s32 offset:76 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v4, off, s[0:3], s32 offset:96 ; 4-byte Folded Reload
; SI-NEXT: s_waitcnt vmcnt(1)
; SI-NEXT: v_lshlrev_b32_e32 v3, 8, v3
; SI-NEXT: v_or_b32_e32 v1, v1, v3
-; SI-NEXT: buffer_load_dword v3, off, s[0:3], s32 offset:80 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v3, off, s[0:3], s32 offset:100 ; 4-byte Folded Reload
; SI-NEXT: v_and_b32_e32 v1, 0xffff, v1
; SI-NEXT: s_waitcnt vmcnt(1)
; SI-NEXT: v_lshlrev_b32_e32 v4, 24, v4
@@ -65259,14 +65263,14 @@ define <64 x i8> @bitcast_v32i16_to_v64i8(<32 x i16> %a, i32 %b) {
; SI-NEXT: v_or_b32_e32 v1, v1, v3
; SI-NEXT: v_add_i32_e32 v3, vcc, 16, v0
; SI-NEXT: buffer_store_dword v1, v3, s[0:3], 0 offen
-; SI-NEXT: buffer_load_dword v3, off, s[0:3], s32 offset:152 ; 4-byte Folded Reload
-; SI-NEXT: buffer_load_dword v4, off, s[0:3], s32 offset:180 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v3, off, s[0:3], s32 offset:172 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v4, off, s[0:3], s32 offset:200 ; 4-byte Folded Reload
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_and_b32_e32 v1, 0xff, v33
; SI-NEXT: s_waitcnt vmcnt(1)
; SI-NEXT: v_lshlrev_b32_e32 v3, 8, v3
; SI-NEXT: v_or_b32_e32 v1, v1, v3
-; SI-NEXT: buffer_load_dword v3, off, s[0:3], s32 offset:172 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v3, off, s[0:3], s32 offset:192 ; 4-byte Folded Reload
; SI-NEXT: s_waitcnt vmcnt(1)
; SI-NEXT: v_lshlrev_b32_e32 v4, 24, v4
; SI-NEXT: v_and_b32_e32 v1, 0xffff, v1
@@ -65277,28 +65281,32 @@ define <64 x i8> @bitcast_v32i16_to_v64i8(<32 x i16> %a, i32 %b) {
; SI-NEXT: v_or_b32_e32 v1, v1, v3
; SI-NEXT: v_add_i32_e32 v3, vcc, 20, v0
; SI-NEXT: buffer_store_dword v1, v3, s[0:3], 0 offen
-; SI-NEXT: buffer_load_dword v3, off, s[0:3], s32 offset:84 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v3, off, s[0:3], s32 offset:104 ; 4-byte Folded Reload
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_and_b32_e32 v1, 0xff, v32
-; SI-NEXT: v_lshlrev_b32_e32 v4, 24, v24
-; SI-NEXT: s_waitcnt vmcnt(0)
+; SI-NEXT: buffer_load_dword v4, off, s[0:3], s32 offset:84 ; 4-byte Folded Reload
+; SI-NEXT: s_waitcnt vmcnt(1)
; SI-NEXT: v_lshlrev_b32_e32 v3, 8, v3
; SI-NEXT: v_or_b32_e32 v1, v1, v3
-; SI-NEXT: v_and_b32_e32 v3, 0xff, v28
-; SI-NEXT: v_lshlrev_b32_e32 v3, 16, v3
+; SI-NEXT: buffer_load_dword v3, off, s[0:3], s32 offset:88 ; 4-byte Folded Reload
; SI-NEXT: v_and_b32_e32 v1, 0xffff, v1
+; SI-NEXT: s_waitcnt vmcnt(1)
+; SI-NEXT: v_lshlrev_b32_e32 v4, 24, v4
+; SI-NEXT: s_waitcnt vmcnt(0)
+; SI-NEXT: v_and_b32_e32 v3, 0xff, v3
+; SI-NEXT: v_lshlrev_b32_e32 v3, 16, v3
; SI-NEXT: v_or_b32_e32 v3, v4, v3
; SI-NEXT: v_or_b32_e32 v1, v1, v3
; SI-NEXT: v_add_i32_e32 v3, vcc, 24, v0
; SI-NEXT: buffer_store_dword v1, v3, s[0:3], 0 offen
-; SI-NEXT: buffer_load_dword v3, off, s[0:3], s32 offset:136 ; 4-byte Folded Reload
-; SI-NEXT: buffer_load_dword v4, off, s[0:3], s32 offset:164 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v3, off, s[0:3], s32 offset:156 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v4, off, s[0:3], s32 offset:184 ; 4-byte Folded Reload
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_and_b32_e32 v1, 0xff, v31
; SI-NEXT: s_waitcnt vmcnt(1)
; SI-NEXT: v_lshlrev_b32_e32 v3, 8, v3
; SI-NEXT: v_or_b32_e32 v1, v1, v3
-; SI-NEXT: buffer_load_dword v3, off, s[0:3], s32 offset:156 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v3, off, s[0:3], s32 offset:176 ; 4-byte Folded Reload
; SI-NEXT: s_waitcnt vmcnt(1)
; SI-NEXT: v_lshlrev_b32_e32 v4, 24, v4
; SI-NEXT: v_and_b32_e32 v1, 0xffff, v1
@@ -65309,26 +65317,30 @@ define <64 x i8> @bitcast_v32i16_to_v64i8(<32 x i16> %a, i32 %b) {
; SI-NEXT: v_or_b32_e32 v1, v1, v3
; SI-NEXT: v_add_i32_e32 v3, vcc, 28, v0
; SI-NEXT: buffer_store_dword v1, v3, s[0:3], 0 offen
+; SI-NEXT: buffer_load_dword v3, off, s[0:3], s32 offset:92 ; 4-byte Folded Reload
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_and_b32_e32 v1, 0xff, v30
-; SI-NEXT: v_lshlrev_b32_e32 v3, 8, v44
+; SI-NEXT: v_lshlrev_b32_e32 v4, 24, v44
+; SI-NEXT: s_waitcnt vmcnt(0)
+; SI-NEXT: v_lshlrev_b32_e32 v3, 8, v3
; SI-NEXT: v_or_b32_e32 v1, v1, v3
-; SI-NEXT: v_and_b32_e32 v3, 0xff, v16
-; SI-NEXT: v_lshlrev_b32_e32 v3, 16, v3
-; SI-NEXT: v_lshlrev_b32_e32 v4, 24, v12
+; SI-NEXT: buffer_load_dword v3, off, s[0:3], s32 offset:76 ; 4-byte Folded Reload
; SI-NEXT: v_and_b32_e32 v1, 0xffff, v1
+; SI-NEXT: s_waitcnt vmcnt(0)
+; SI-NEXT: v_and_b32_e32 v3, 0xff, v3
+; SI-NEXT: v_lshlrev_b32_e32 v3, 16, v3
; SI-NEXT: v_or_b32_e32 v3, v4, v3
; SI-NEXT: v_or_b32_e32 v1, v1, v3
; SI-NEXT: v_add_i32_e32 v3, vcc, 32, v0
; SI-NEXT: buffer_store_dword v1, v3, s[0:3], 0 offen
-; SI-NEXT: buffer_load_dword v3, off, s[0:3], s32 offset:176 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v3, off, s[0:3], s32 offset:196 ; 4-byte Folded Reload
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_and_b32_e32 v1, 0xff, v26
-; SI-NEXT: buffer_load_dword v4, off, s[0:3], s32 offset:148 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v4, off, s[0:3], s32 offset:168 ; 4-byte Folded Reload
; SI-NEXT: s_waitcnt vmcnt(1)
; SI-NEXT: v_lshlrev_b32_e32 v3, 8, v3
; SI-NEXT: v_or_b32_e32 v1, v1, v3
-; SI-NEXT: buffer_load_dword v3, off, s[0:3], s32 offset:140 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v3, off, s[0:3], s32 offset:160 ; 4-byte Folded Reload
; SI-NEXT: v_and_b32_e32 v1, 0xffff, v1
; SI-NEXT: s_waitcnt vmcnt(1)
; SI-NEXT: v_lshlrev_b32_e32 v4, 24, v4
@@ -65339,26 +65351,28 @@ define <64 x i8> @bitcast_v32i16_to_v64i8(<32 x i16> %a, i32 %b) {
; SI-NEXT: v_or_b32_e32 v1, v1, v3
; SI-NEXT: v_add_i32_e32 v3, vcc, 36, v0
; SI-NEXT: buffer_store_dword v1, v3, s[0:3], 0 offen
+; SI-NEXT: buffer_load_dword v3, off, s[0:3], s32 offset:80 ; 4-byte Folded Reload
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_and_b32_e32 v1, 0xff, v22
-; SI-NEXT: v_lshlrev_b32_e32 v3, 8, v20
+; SI-NEXT: v_lshlrev_b32_e32 v4, 24, v60
+; SI-NEXT: s_waitcnt vmcnt(0)
+; SI-NEXT: v_lshlrev_b32_e32 v3, 8, v3
; SI-NEXT: v_or_b32_e32 v1, v1, v3
; SI-NEXT: v_and_b32_e32 v3, 0xff, v61
; SI-NEXT: v_lshlrev_b32_e32 v3, 16, v3
-; SI-NEXT: v_lshlrev_b32_e32 v4, 24, v60
; SI-NEXT: v_and_b32_e32 v1, 0xffff, v1
; SI-NEXT: v_or_b32_e32 v3, v4, v3
; SI-NEXT: v_or_b32_e32 v1, v1, v3
; SI-NEXT: v_add_i32_e32 v3, vcc, 40, v0
; SI-NEXT: buffer_store_dword v1, v3, s[0:3], 0 offen
-; SI-NEXT: buffer_load_dword v3, off, s[0:3], s32 offset:160 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v3, off, s[0:3], s32 offset:180 ; 4-byte Folded Reload
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_and_b32_e32 v1, 0xff, v18
-; SI-NEXT: buffer_load_dword v4, off, s[0:3], s32 offset:132 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v4, off, s[0:3], s32 offset:152 ; 4-byte Folded Reload
; SI-NEXT: s_waitcnt vmcnt(1)
; SI-NEXT: v_lshlrev_b32_e32 v3, 8, v3
; SI-NEXT: v_or_b32_e32 v1, v1, v3
-; SI-NEXT: buffer_load_dword v3, off, s[0:3], s32 offset:124 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v3, off, s[0:3], s32 offset:144 ; 4-byte Folded Reload
; SI-NEXT: v_and_b32_e32 v1, 0xffff, v1
; SI-NEXT: s_waitcnt vmcnt(1)
; SI-NEXT: v_lshlrev_b32_e32 v4, 24, v4
@@ -65371,9 +65385,9 @@ define <64 x i8> @bitcast_v32i16_to_v64i8(<32 x i16> %a, i32 %b) {
; SI-NEXT: buffer_store_dword v1, v3, s[0:3], 0 offen
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_and_b32_e32 v1, 0xff, v14
-; SI-NEXT: v_lshlrev_b32_e32 v3, 8, v8
+; SI-NEXT: v_lshlrev_b32_e32 v3, 8, v28
; SI-NEXT: v_or_b32_e32 v1, v1, v3
-; SI-NEXT: v_and_b32_e32 v3, 0xff, v58
+; SI-NEXT: v_and_b32_e32 v3, 0xff, v20
; SI-NEXT: v_lshlrev_b32_e32 v3, 16, v3
; SI-NEXT: v_lshlrev_b32_e32 v4, 24, v57
; SI-NEXT: v_and_b32_e32 v1, 0xffff, v1
@@ -65381,14 +65395,14 @@ define <64 x i8> @bitcast_v32i16_to_v64i8(<32 x i16> %a, i32 %b) {
; SI-NEXT: v_or_b32_e32 v1, v1, v3
; SI-NEXT: v_add_i32_e32 v3, vcc, 48, v0
; SI-NEXT: buffer_store_dword v1, v3, s[0:3], 0 offen
-; SI-NEXT: buffer_load_dword v3, off, s[0:3], s32 offset:144 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v3, off, s[0:3], s32 offset:164 ; 4-byte Folded Reload
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_and_b32_e32 v1, 0xff, v10
-; SI-NEXT: buffer_load_dword v4, off, s[0:3], s32 offset:120 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v4, off, s[0:3], s32 offset:140 ; 4-byte Folded Reload
; SI-NEXT: s_waitcnt vmcnt(1)
; SI-NEXT: v_lshlrev_b32_e32 v3, 8, v3
; SI-NEXT: v_or_b32_e32 v1, v1, v3
-; SI-NEXT: buffer_load_dword v3, off, s[0:3], s32 offset:116 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v3, off, s[0:3], s32 offset:136 ; 4-byte Folded Reload
; SI-NEXT: v_and_b32_e32 v1, 0xffff, v1
; SI-NEXT: s_waitcnt vmcnt(1)
; SI-NEXT: v_lshlrev_b32_e32 v4, 24, v4
@@ -65413,13 +65427,13 @@ define <64 x i8> @bitcast_v32i16_to_v64i8(<32 x i16> %a, i32 %b) {
; SI-NEXT: buffer_store_dword v1, v3, s[0:3], 0 offen
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_and_b32_e32 v1, 0xff, v2
-; SI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:128 ; 4-byte Folded Reload
-; SI-NEXT: v_lshlrev_b32_e32 v3, 24, v62
+; SI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:148 ; 4-byte Folded Reload
+; SI-NEXT: v_lshlrev_b32_e32 v3, 24, v58
; SI-NEXT: v_add_i32_e32 v0, vcc, 60, v0
; SI-NEXT: s_waitcnt vmcnt(0)
; SI-NEXT: v_lshlrev_b32_e32 v2, 8, v2
; SI-NEXT: v_or_b32_e32 v1, v1, v2
-; SI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:112 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:132 ; 4-byte Folded Reload
; SI-NEXT: v_and_b32_e32 v1, 0xffff, v1
; SI-NEXT: s_waitcnt vmcnt(0)
; SI-NEXT: v_and_b32_e32 v2, 0xff, v2
@@ -65448,28 +65462,27 @@ define <64 x i8> @bitcast_v32i16_to_v64i8(<32 x i16> %a, i32 %b) {
;
; VI-LABEL: bitcast_v32i16_to_v64i8:
; VI: ; %bb.0:
-; VI-NEXT: ; implicit-def: $vgpr19
; VI-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; VI-NEXT: v_cmp_ne_u32_e32 vcc, 0, v17
; VI-NEXT: ; implicit-def: $vgpr17
-; VI-NEXT: ; kill: killed $vgpr19
-; VI-NEXT: ; implicit-def: $vgpr19
-; VI-NEXT: ; kill: killed $vgpr17
-; VI-NEXT: ; implicit-def: $vgpr17
-; VI-NEXT: ; kill: killed $vgpr19
-; VI-NEXT: ; implicit-def: $vgpr19
-; VI-NEXT: ; kill: killed $vgpr17
-; VI-NEXT: ; implicit-def: $vgpr17
-; VI-NEXT: ; kill: killed $vgpr19
-; VI-NEXT: ; implicit-def: $vgpr19
-; VI-NEXT: ; kill: killed $vgpr17
-; VI-NEXT: ; implicit-def: $vgpr17
-; VI-NEXT: ; kill: killed $vgpr19
-; VI-NEXT: ; implicit-def: $vgpr19
+; VI-NEXT: v_lshrrev_b32_e32 v29, 16, v16
+; VI-NEXT: v_lshrrev_b32_e32 v32, 16, v15
+; VI-NEXT: v_lshrrev_b32_e32 v30, 16, v14
+; VI-NEXT: v_lshrrev_b32_e32 v34, 16, v13
+; VI-NEXT: v_lshrrev_b32_e32 v31, 16, v12
+; VI-NEXT: v_lshrrev_b32_e32 v36, 16, v11
+; VI-NEXT: v_lshrrev_b32_e32 v33, 16, v10
+; VI-NEXT: v_lshrrev_b32_e32 v38, 16, v9
+; VI-NEXT: v_lshrrev_b32_e32 v35, 16, v8
+; VI-NEXT: v_lshrrev_b32_e32 v39, 16, v7
+; VI-NEXT: v_lshrrev_b32_e32 v37, 16, v6
+; VI-NEXT: v_lshrrev_b32_e32 v49, 16, v5
+; VI-NEXT: v_lshrrev_b32_e32 v48, 16, v4
+; VI-NEXT: v_lshrrev_b32_e32 v51, 16, v3
+; VI-NEXT: v_lshrrev_b32_e32 v50, 16, v2
+; VI-NEXT: v_lshrrev_b32_e32 v52, 16, v1
; VI-NEXT: ; kill: killed $vgpr17
; VI-NEXT: ; implicit-def: $vgpr17
-; VI-NEXT: ; kill: killed $vgpr19
-; VI-NEXT: ; implicit-def: $vgpr19
; VI-NEXT: buffer_store_dword v40, off, s[0:3], s32 offset:60 ; 4-byte Folded Spill
; VI-NEXT: buffer_store_dword v41, off, s[0:3], s32 offset:56 ; 4-byte Folded Spill
; VI-NEXT: buffer_store_dword v42, off, s[0:3], s32 offset:52 ; 4-byte Folded Spill
@@ -65486,415 +65499,312 @@ define <64 x i8> @bitcast_v32i16_to_v64i8(<32 x i16> %a, i32 %b) {
; VI-NEXT: buffer_store_dword v61, off, s[0:3], s32 offset:8 ; 4-byte Folded Spill
; VI-NEXT: buffer_store_dword v62, off, s[0:3], s32 offset:4 ; 4-byte Folded Spill
; VI-NEXT: buffer_store_dword v63, off, s[0:3], s32 ; 4-byte Folded Spill
-; VI-NEXT: ; implicit-def: $vgpr43
-; VI-NEXT: ; kill: killed $vgpr17
-; VI-NEXT: ; implicit-def: $vgpr17
-; VI-NEXT: ; kill: killed $vgpr19
-; VI-NEXT: ; implicit-def: $vgpr19
-; VI-NEXT: ; implicit-def: $vgpr42
-; VI-NEXT: v_lshrrev_b32_e32 v26, 16, v16
-; VI-NEXT: v_lshrrev_b32_e32 v29, 16, v15
-; VI-NEXT: v_lshrrev_b32_e32 v27, 16, v14
-; VI-NEXT: v_lshrrev_b32_e32 v31, 16, v13
-; VI-NEXT: v_lshrrev_b32_e32 v28, 16, v12
-; VI-NEXT: v_lshrrev_b32_e32 v33, 16, v11
-; VI-NEXT: v_lshrrev_b32_e32 v30, 16, v10
-; VI-NEXT: v_lshrrev_b32_e32 v35, 16, v9
-; VI-NEXT: v_lshrrev_b32_e32 v32, 16, v8
-; VI-NEXT: v_lshrrev_b32_e32 v37, 16, v7
-; VI-NEXT: v_lshrrev_b32_e32 v34, 16, v6
-; VI-NEXT: v_lshrrev_b32_e32 v39, 16, v5
-; VI-NEXT: v_lshrrev_b32_e32 v36, 16, v4
-; VI-NEXT: v_lshrrev_b32_e32 v48, 16, v3
-; VI-NEXT: v_lshrrev_b32_e32 v38, 16, v2
-; VI-NEXT: v_lshrrev_b32_e32 v49, 16, v1
-; VI-NEXT: ; kill: killed $vgpr17
-; VI-NEXT: ; implicit-def: $vgpr17
-; VI-NEXT: ; kill: killed $vgpr19
-; VI-NEXT: ; implicit-def: $vgpr19
-; VI-NEXT: buffer_store_dword v42, off, s[0:3], s32 offset:72 ; 4-byte Folded Spill
-; VI-NEXT: buffer_store_dword v43, off, s[0:3], s32 offset:76 ; 4-byte Folded Spill
-; VI-NEXT: ; implicit-def: $vgpr42
-; VI-NEXT: buffer_store_dword v42, off, s[0:3], s32 offset:64 ; 4-byte Folded Spill
-; VI-NEXT: buffer_store_dword v43, off, s[0:3], s32 offset:68 ; 4-byte Folded Spill
-; VI-NEXT: ; implicit-def: $vgpr41
-; VI-NEXT: ; implicit-def: $vgpr18
+; VI-NEXT: ; implicit-def: $vgpr26
+; VI-NEXT: ; implicit-def: $vgpr23
+; VI-NEXT: ; implicit-def: $vgpr53
; VI-NEXT: ; implicit-def: $vgpr54
-; VI-NEXT: ; implicit-def: $vgpr45
+; VI-NEXT: ; implicit-def: $vgpr62
+; VI-NEXT: ; implicit-def: $vgpr22
+; VI-NEXT: ; implicit-def: $vgpr61
+; VI-NEXT: ; implicit-def: $vgpr63
; VI-NEXT: ; implicit-def: $vgpr59
-; VI-NEXT: ; implicit-def: $vgpr57
-; VI-NEXT: ; implicit-def: $vgpr51
; VI-NEXT: ; implicit-def: $vgpr47
-; VI-NEXT: ; implicit-def: $vgpr55
-; VI-NEXT: ; implicit-def: $vgpr46
-; VI-NEXT: ; implicit-def: $vgpr61
; VI-NEXT: ; implicit-def: $vgpr60
-; VI-NEXT: ; implicit-def: $vgpr52
-; VI-NEXT: ; implicit-def: $vgpr63
-; VI-NEXT: ; implicit-def: $vgpr40
-; VI-NEXT: ; implicit-def: $vgpr53
-; VI-NEXT: ; kill: killed $vgpr17
-; VI-NEXT: ; implicit-def: $vgpr17
-; VI-NEXT: ; implicit-def: $vgpr44
+; VI-NEXT: ; implicit-def: $vgpr46
+; VI-NEXT: ; implicit-def: $vgpr45
; VI-NEXT: ; implicit-def: $vgpr58
+; VI-NEXT: ; implicit-def: $vgpr44
+; VI-NEXT: ; implicit-def: $vgpr43
+; VI-NEXT: ; implicit-def: $vgpr57
+; VI-NEXT: ; implicit-def: $vgpr42
+; VI-NEXT: ; implicit-def: $vgpr41
; VI-NEXT: ; implicit-def: $vgpr56
-; VI-NEXT: ; implicit-def: $vgpr50
-; VI-NEXT: ; implicit-def: $vgpr62
-; VI-NEXT: ; kill: killed $vgpr19
+; VI-NEXT: ; kill: killed $vgpr17
+; VI-NEXT: ; implicit-def: $vgpr28
+; VI-NEXT: ; implicit-def: $vgpr40
+; VI-NEXT: ; implicit-def: $vgpr55
+; VI-NEXT: ; implicit-def: $vgpr27
; VI-NEXT: ; implicit-def: $vgpr25
-; VI-NEXT: ; implicit-def: $vgpr24
-; VI-NEXT: ; implicit-def: $vgpr23
-; VI-NEXT: ; implicit-def: $vgpr22
; VI-NEXT: ; implicit-def: $vgpr21
; VI-NEXT: ; implicit-def: $vgpr20
; VI-NEXT: ; implicit-def: $vgpr19
-; VI-NEXT: ; implicit-def: $vgpr42
+; VI-NEXT: ; implicit-def: $vgpr18
+; VI-NEXT: ; implicit-def: $vgpr17
; VI-NEXT: s_and_saveexec_b64 s[4:5], vcc
; VI-NEXT: s_xor_b64 s[4:5], exec, s[4:5]
; VI-NEXT: s_cbranch_execz .LBB96_2
; VI-NEXT: ; %bb.1: ; %cmp.false
-; VI-NEXT: v_lshrrev_b32_e32 v17, 8, v16
-; VI-NEXT: buffer_store_dword v17, off, s[0:3], s32 offset:80 ; 4-byte Folded Spill
-; VI-NEXT: v_lshrrev_b32_e32 v17, 8, v15
-; VI-NEXT: buffer_store_dword v17, off, s[0:3], s32 offset:84 ; 4-byte Folded Spill
-; VI-NEXT: v_lshrrev_b32_e32 v17, 24, v14
-; VI-NEXT: buffer_store_dword v17, off, s[0:3], s32 offset:104 ; 4-byte Folded Spill
; VI-NEXT: v_lshrrev_b32_e32 v17, 8, v14
-; VI-NEXT: buffer_store_dword v17, off, s[0:3], s32 offset:88 ; 4-byte Folded Spill
+; VI-NEXT: buffer_store_dword v17, off, s[0:3], s32 offset:64 ; 4-byte Folded Spill
; VI-NEXT: v_lshrrev_b32_e32 v17, 8, v13
-; VI-NEXT: buffer_store_dword v17, off, s[0:3], s32 offset:92 ; 4-byte Folded Spill
-; VI-NEXT: v_lshrrev_b32_e32 v17, 24, v12
-; VI-NEXT: buffer_store_dword v17, off, s[0:3], s32 offset:112 ; 4-byte Folded Spill
-; VI-NEXT: v_lshrrev_b32_e32 v17, 8, v12
-; VI-NEXT: buffer_store_dword v17, off, s[0:3], s32 offset:96 ; 4-byte Folded Spill
-; VI-NEXT: v_lshrrev_b32_e32 v17, 8, v11
-; VI-NEXT: buffer_store_dword v17, off, s[0:3], s32 offset:100 ; 4-byte Folded Spill
-; VI-NEXT: v_lshrrev_b32_e32 v17, 24, v10
-; VI-NEXT: buffer_store_dword v17, off, s[0:3], s32 offset:124 ; 4-byte Folded Spill
-; VI-NEXT: v_lshrrev_b32_e32 v17, 8, v10
-; VI-NEXT: buffer_store_dword v17, off, s[0:3], s32 offset:108 ; 4-byte Folded Spill
-; VI-NEXT: v_lshrrev_b32_e32 v17, 24, v8
-; VI-NEXT: buffer_store_dword v17, off, s[0:3], s32 offset:132 ; 4-byte Folded Spill
-; VI-NEXT: v_lshrrev_b32_e32 v17, 8, v8
-; VI-NEXT: buffer_store_dword v17, off, s[0:3], s32 offset:116 ; 4-byte Folded Spill
-; VI-NEXT: v_lshrrev_b32_e32 v17, 8, v7
-; VI-NEXT: buffer_store_dword v17, off, s[0:3], s32 offset:120 ; 4-byte Folded Spill
-; VI-NEXT: v_lshrrev_b32_e32 v17, 8, v6
-; VI-NEXT: v_lshrrev_b64 v[19:20], 24, v[15:16]
-; VI-NEXT: buffer_store_dword v17, off, s[0:3], s32 offset:128 ; 4-byte Folded Spill
-; VI-NEXT: v_lshrrev_b32_e32 v17, 8, v5
-; VI-NEXT: buffer_store_dword v17, off, s[0:3], s32 offset:136 ; 4-byte Folded Spill
-; VI-NEXT: buffer_store_dword v19, off, s[0:3], s32 offset:64 ; 4-byte Folded Spill
-; VI-NEXT: buffer_store_dword v20, off, s[0:3], s32 offset:68 ; 4-byte Folded Spill
-; VI-NEXT: v_lshrrev_b64 v[19:20], 24, v[13:14]
-; VI-NEXT: buffer_store_dword v19, off, s[0:3], s32 offset:72 ; 4-byte Folded Spill
-; VI-NEXT: buffer_store_dword v20, off, s[0:3], s32 offset:76 ; 4-byte Folded Spill
+; VI-NEXT: buffer_store_dword v17, off, s[0:3], s32 offset:68 ; 4-byte Folded Spill
+; VI-NEXT: v_lshrrev_b64 v[17:18], 24, v[15:16]
+; VI-NEXT: v_lshrrev_b64 v[18:19], 24, v[13:14]
; VI-NEXT: v_lshrrev_b64 v[19:20], 24, v[11:12]
; VI-NEXT: v_lshrrev_b64 v[20:21], 24, v[9:10]
; VI-NEXT: v_lshrrev_b64 v[21:22], 24, v[7:8]
-; VI-NEXT: v_lshrrev_b64 v[22:23], 24, v[5:6]
-; VI-NEXT: v_lshrrev_b64 v[23:24], 24, v[3:4]
-; VI-NEXT: v_lshrrev_b32_e32 v50, 24, v16
-; VI-NEXT: v_lshrrev_b64 v[24:25], 24, v[1:2]
-; VI-NEXT: v_lshrrev_b32_e32 v42, 8, v9
-; VI-NEXT: v_lshrrev_b32_e32 v43, 24, v6
-; VI-NEXT: v_lshrrev_b32_e32 v46, 24, v4
-; VI-NEXT: v_lshrrev_b32_e32 v55, 8, v4
-; VI-NEXT: v_lshrrev_b32_e32 v51, 8, v3
-; VI-NEXT: v_lshrrev_b32_e32 v59, 24, v2
-; VI-NEXT: v_lshrrev_b32_e32 v45, 8, v2
-; VI-NEXT: v_lshrrev_b32_e32 v18, 8, v1
-; VI-NEXT: v_mov_b32_e32 v25, v50
-; VI-NEXT: v_mov_b32_e32 v41, v1
-; VI-NEXT: v_mov_b32_e32 v54, v2
-; VI-NEXT: v_mov_b32_e32 v57, v3
-; VI-NEXT: v_mov_b32_e32 v47, v4
-; VI-NEXT: v_mov_b32_e32 v61, v5
-; VI-NEXT: v_mov_b32_e32 v60, v6
-; VI-NEXT: v_mov_b32_e32 v52, v7
-; VI-NEXT: v_mov_b32_e32 v63, v8
-; VI-NEXT: v_mov_b32_e32 v40, v9
-; VI-NEXT: v_mov_b32_e32 v53, v10
-; VI-NEXT: v_mov_b32_e32 v17, v11
-; VI-NEXT: v_mov_b32_e32 v44, v12
-; VI-NEXT: v_mov_b32_e32 v58, v13
-; VI-NEXT: v_mov_b32_e32 v56, v14
-; VI-NEXT: v_mov_b32_e32 v50, v15
-; VI-NEXT: v_mov_b32_e32 v62, v16
-; VI-NEXT: ; implicit-def: $vgpr1
-; VI-NEXT: ; implicit-def: $vgpr3
-; VI-NEXT: ; implicit-def: $vgpr5
-; VI-NEXT: ; implicit-def: $vgpr7
-; VI-NEXT: ; implicit-def: $vgpr9
-; VI-NEXT: ; implicit-def: $vgpr11
-; VI-NEXT: ; implicit-def: $vgpr13
-; VI-NEXT: ; implicit-def: $vgpr15
+; VI-NEXT: v_lshrrev_b64 v[25:26], 24, v[5:6]
+; VI-NEXT: v_lshrrev_b64 v[22:23], 24, v[3:4]
+; VI-NEXT: v_lshrrev_b32_e32 v27, 24, v16
+; VI-NEXT: v_lshrrev_b32_e32 v55, 8, v16
+; VI-NEXT: v_lshrrev_b32_e32 v40, 8, v15
+; VI-NEXT: v_lshrrev_b32_e32 v28, 24, v14
+; VI-NEXT: v_lshrrev_b32_e32 v56, 24, v12
+; VI-NEXT: v_lshrrev_b32_e32 v41, 8, v12
+; VI-NEXT: v_lshrrev_b32_e32 v42, 8, v11
+; VI-NEXT: v_lshrrev_b32_e32 v57, 24, v10
+; VI-NEXT: v_lshrrev_b32_e32 v43, 8, v10
+; VI-NEXT: v_lshrrev_b32_e32 v44, 8, v9
+; VI-NEXT: v_lshrrev_b32_e32 v58, 24, v8
+; VI-NEXT: v_lshrrev_b32_e32 v45, 8, v8
+; VI-NEXT: v_lshrrev_b32_e32 v46, 8, v7
+; VI-NEXT: v_lshrrev_b32_e32 v60, 24, v6
+; VI-NEXT: v_lshrrev_b32_e32 v47, 8, v6
+; VI-NEXT: v_lshrrev_b32_e32 v59, 8, v5
+; VI-NEXT: v_lshrrev_b32_e32 v63, 24, v4
+; VI-NEXT: v_lshrrev_b32_e32 v61, 8, v4
+; VI-NEXT: v_lshrrev_b32_e32 v62, 8, v3
+; VI-NEXT: v_lshrrev_b32_e32 v54, 24, v2
+; VI-NEXT: v_lshrrev_b32_e32 v53, 8, v2
+; VI-NEXT: v_lshrrev_b64 v[23:24], 24, v[1:2]
+; VI-NEXT: v_lshrrev_b32_e32 v26, 8, v1
; VI-NEXT: .LBB96_2: ; %Flow
; VI-NEXT: s_andn2_saveexec_b64 s[4:5], s[4:5]
; VI-NEXT: s_cbranch_execz .LBB96_4
; VI-NEXT: ; %bb.3: ; %cmp.true
-; VI-NEXT: v_mov_b32_e32 v18, 3
-; VI-NEXT: v_add_u16_sdwa v26, v16, v18 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_1 src1_sel:DWORD
-; VI-NEXT: v_add_u16_sdwa v29, v15, v18 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_1 src1_sel:DWORD
-; VI-NEXT: v_add_u16_e32 v62, 3, v16
-; VI-NEXT: v_lshlrev_b32_e32 v16, 16, v26
-; VI-NEXT: v_add_u16_e32 v50, 3, v15
-; VI-NEXT: v_lshlrev_b32_e32 v15, 16, v29
-; VI-NEXT: v_or_b32_e32 v16, v62, v16
-; VI-NEXT: v_or_b32_e32 v15, v50, v15
-; VI-NEXT: v_add_u16_sdwa v38, v2, v18 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_1 src1_sel:DWORD
-; VI-NEXT: v_add_u16_sdwa v49, v1, v18 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_1 src1_sel:DWORD
-; VI-NEXT: v_add_u16_sdwa v36, v4, v18 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_1 src1_sel:DWORD
-; VI-NEXT: v_add_u16_sdwa v48, v3, v18 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_1 src1_sel:DWORD
-; VI-NEXT: v_add_u16_sdwa v34, v6, v18 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_1 src1_sel:DWORD
-; VI-NEXT: v_add_u16_sdwa v39, v5, v18 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_1 src1_sel:DWORD
-; VI-NEXT: v_add_u16_sdwa v32, v8, v18 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_1 src1_sel:DWORD
-; VI-NEXT: v_add_u16_sdwa v37, v7, v18 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_1 src1_sel:DWORD
-; VI-NEXT: v_add_u16_sdwa v30, v10, v18 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_1 src1_sel:DWORD
-; VI-NEXT: v_add_u16_sdwa v35, v9, v18 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_1 src1_sel:DWORD
-; VI-NEXT: v_add_u16_sdwa v28, v12, v18 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_1 src1_sel:DWORD
-; VI-NEXT: v_add_u16_sdwa v33, v11, v18 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_1 src1_sel:DWORD
-; VI-NEXT: v_add_u16_sdwa v27, v14, v18 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_1 src1_sel:DWORD
-; VI-NEXT: v_add_u16_sdwa v31, v13, v18 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_1 src1_sel:DWORD
-; VI-NEXT: v_lshrrev_b64 v[18:19], 24, v[15:16]
-; VI-NEXT: v_add_u16_e32 v56, 3, v14
-; VI-NEXT: v_lshlrev_b32_e32 v14, 16, v27
-; VI-NEXT: v_add_u16_e32 v58, 3, v13
-; VI-NEXT: v_lshlrev_b32_e32 v13, 16, v31
-; VI-NEXT: v_or_b32_e32 v14, v56, v14
-; VI-NEXT: v_or_b32_e32 v13, v58, v13
-; VI-NEXT: buffer_store_dword v18, off, s[0:3], s32 offset:64 ; 4-byte Folded Spill
-; VI-NEXT: buffer_store_dword v19, off, s[0:3], s32 offset:68 ; 4-byte Folded Spill
+; VI-NEXT: v_mov_b32_e32 v17, 3
+; VI-NEXT: v_add_u16_sdwa v50, v2, v17 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_1 src1_sel:DWORD
+; VI-NEXT: v_add_u16_e32 v53, 3, v2
+; VI-NEXT: v_lshlrev_b32_e32 v2, 16, v50
+; VI-NEXT: v_add_u16_sdwa v52, v1, v17 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_1 src1_sel:DWORD
+; VI-NEXT: v_or_b32_e32 v28, v53, v2
+; VI-NEXT: v_add_u16_e32 v2, 3, v1
+; VI-NEXT: v_lshlrev_b32_e32 v1, 16, v52
+; VI-NEXT: v_add_u16_sdwa v48, v4, v17 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_1 src1_sel:DWORD
+; VI-NEXT: v_or_b32_e32 v27, v2, v1
+; VI-NEXT: v_add_u16_e32 v54, 3, v4
+; VI-NEXT: v_lshlrev_b32_e32 v1, 16, v48
+; VI-NEXT: v_add_u16_sdwa v51, v3, v17 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_1 src1_sel:DWORD
+; VI-NEXT: v_or_b32_e32 v24, v54, v1
+; VI-NEXT: v_add_u16_e32 v4, 3, v3
+; VI-NEXT: v_lshlrev_b32_e32 v1, 16, v51
+; VI-NEXT: v_add_u16_sdwa v37, v6, v17 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_1 src1_sel:DWORD
+; VI-NEXT: v_or_b32_e32 v23, v4, v1
+; VI-NEXT: v_add_u16_e32 v56, 3, v6
+; VI-NEXT: v_lshlrev_b32_e32 v1, 16, v37
+; VI-NEXT: v_add_u16_sdwa v49, v5, v17 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_1 src1_sel:DWORD
+; VI-NEXT: v_or_b32_e32 v26, v56, v1
+; VI-NEXT: v_add_u16_e32 v6, 3, v5
+; VI-NEXT: v_lshlrev_b32_e32 v1, 16, v49
+; VI-NEXT: v_add_u16_sdwa v35, v8, v17 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_1 src1_sel:DWORD
+; VI-NEXT: v_or_b32_e32 v25, v6, v1
+; VI-NEXT: v_add_u16_e32 v57, 3, v8
+; VI-NEXT: v_lshlrev_b32_e32 v1, 16, v35
+; VI-NEXT: v_add_u16_sdwa v39, v7, v17 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_1 src1_sel:DWORD
+; VI-NEXT: v_or_b32_e32 v47, v57, v1
+; VI-NEXT: v_add_u16_e32 v8, 3, v7
+; VI-NEXT: v_lshlrev_b32_e32 v1, 16, v39
+; VI-NEXT: v_add_u16_sdwa v33, v10, v17 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_1 src1_sel:DWORD
+; VI-NEXT: v_or_b32_e32 v46, v8, v1
+; VI-NEXT: v_add_u16_e32 v58, 3, v10
+; VI-NEXT: v_lshlrev_b32_e32 v1, 16, v33
+; VI-NEXT: v_add_u16_sdwa v38, v9, v17 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_1 src1_sel:DWORD
+; VI-NEXT: v_or_b32_e32 v10, v58, v1
+; VI-NEXT: v_add_u16_e32 v59, 3, v9
+; VI-NEXT: v_lshlrev_b32_e32 v1, 16, v38
+; VI-NEXT: v_add_u16_sdwa v31, v12, v17 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_1 src1_sel:DWORD
+; VI-NEXT: v_or_b32_e32 v9, v59, v1
+; VI-NEXT: v_add_u16_e32 v60, 3, v12
+; VI-NEXT: v_lshlrev_b32_e32 v1, 16, v31
+; VI-NEXT: v_add_u16_sdwa v36, v11, v17 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_1 src1_sel:DWORD
+; VI-NEXT: v_or_b32_e32 v12, v60, v1
+; VI-NEXT: v_add_u16_e32 v61, 3, v11
+; VI-NEXT: v_lshlrev_b32_e32 v1, 16, v36
+; VI-NEXT: v_add_u16_sdwa v30, v14, v17 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_1 src1_sel:DWORD
+; VI-NEXT: v_or_b32_e32 v11, v61, v1
+; VI-NEXT: v_add_u16_e32 v62, 3, v14
+; VI-NEXT: v_lshlrev_b32_e32 v1, 16, v30
+; VI-NEXT: v_add_u16_sdwa v34, v13, v17 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_1 src1_sel:DWORD
+; VI-NEXT: v_or_b32_e32 v14, v62, v1
+; VI-NEXT: v_add_u16_e32 v63, 3, v13
+; VI-NEXT: v_lshlrev_b32_e32 v1, 16, v34
+; VI-NEXT: v_add_u16_sdwa v29, v16, v17 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_1 src1_sel:DWORD
+; VI-NEXT: v_or_b32_e32 v13, v63, v1
+; VI-NEXT: v_add_u16_e32 v55, 3, v16
+; VI-NEXT: v_lshlrev_b32_e32 v1, 16, v29
+; VI-NEXT: v_add_u16_sdwa v32, v15, v17 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_1 src1_sel:DWORD
+; VI-NEXT: v_or_b32_e32 v16, v55, v1
+; VI-NEXT: v_add_u16_e32 v40, 3, v15
+; VI-NEXT: v_lshlrev_b32_e32 v1, 16, v32
+; VI-NEXT: v_or_b32_e32 v15, v40, v1
+; VI-NEXT: v_lshrrev_b64 v[17:18], 24, v[15:16]
; VI-NEXT: v_lshrrev_b64 v[18:19], 24, v[13:14]
-; VI-NEXT: v_add_u16_e32 v44, 3, v12
-; VI-NEXT: v_lshlrev_b32_e32 v12, 16, v28
-; VI-NEXT: v_add_u16_e32 v17, 3, v11
-; VI-NEXT: v_lshlrev_b32_e32 v11, 16, v33
-; VI-NEXT: v_add_u16_e32 v53, 3, v10
-; VI-NEXT: v_lshlrev_b32_e32 v10, 16, v30
-; VI-NEXT: v_add_u16_e32 v40, 3, v9
-; VI-NEXT: v_lshlrev_b32_e32 v9, 16, v35
-; VI-NEXT: v_or_b32_e32 v12, v44, v12
-; VI-NEXT: v_or_b32_e32 v11, v17, v11
-; VI-NEXT: buffer_store_dword v18, off, s[0:3], s32 offset:72 ; 4-byte Folded Spill
-; VI-NEXT: buffer_store_dword v19, off, s[0:3], s32 offset:76 ; 4-byte Folded Spill
-; VI-NEXT: v_add_u16_e32 v63, 3, v8
-; VI-NEXT: v_lshlrev_b32_e32 v8, 16, v32
-; VI-NEXT: v_add_u16_e32 v52, 3, v7
-; VI-NEXT: v_lshlrev_b32_e32 v7, 16, v37
-; VI-NEXT: v_or_b32_e32 v10, v53, v10
-; VI-NEXT: v_or_b32_e32 v9, v40, v9
; VI-NEXT: v_lshrrev_b64 v[19:20], 24, v[11:12]
-; VI-NEXT: v_add_u16_e32 v60, 3, v6
-; VI-NEXT: v_lshlrev_b32_e32 v6, 16, v34
-; VI-NEXT: v_add_u16_e32 v61, 3, v5
-; VI-NEXT: v_lshlrev_b32_e32 v5, 16, v39
-; VI-NEXT: v_or_b32_e32 v8, v63, v8
-; VI-NEXT: v_or_b32_e32 v7, v52, v7
; VI-NEXT: v_lshrrev_b64 v[20:21], 24, v[9:10]
-; VI-NEXT: v_add_u16_e32 v47, 3, v4
-; VI-NEXT: v_lshlrev_b32_e32 v4, 16, v36
-; VI-NEXT: v_add_u16_e32 v57, 3, v3
-; VI-NEXT: v_lshlrev_b32_e32 v3, 16, v48
-; VI-NEXT: v_or_b32_e32 v6, v60, v6
-; VI-NEXT: v_or_b32_e32 v5, v61, v5
-; VI-NEXT: v_lshrrev_b64 v[21:22], 24, v[7:8]
-; VI-NEXT: v_add_u16_e32 v54, 3, v2
-; VI-NEXT: v_lshlrev_b32_e32 v2, 16, v38
-; VI-NEXT: v_add_u16_e32 v41, 3, v1
-; VI-NEXT: v_lshlrev_b32_e32 v1, 16, v49
-; VI-NEXT: v_or_b32_e32 v4, v47, v4
-; VI-NEXT: v_or_b32_e32 v3, v57, v3
-; VI-NEXT: v_lshrrev_b64 v[22:23], 24, v[5:6]
-; VI-NEXT: v_or_b32_e32 v2, v54, v2
-; VI-NEXT: v_or_b32_e32 v1, v41, v1
-; VI-NEXT: v_lshrrev_b64 v[23:24], 24, v[3:4]
-; VI-NEXT: v_lshrrev_b32_e32 v18, 8, v1
-; VI-NEXT: v_lshrrev_b64 v[24:25], 24, v[1:2]
-; VI-NEXT: v_bfe_u32 v1, v27, 8, 8
-; VI-NEXT: v_lshrrev_b32_e32 v13, 8, v13
-; VI-NEXT: v_lshrrev_b32_e32 v42, 8, v9
-; VI-NEXT: v_lshrrev_b32_e32 v9, 8, v8
-; VI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:104 ; 4-byte Folded Spill
-; VI-NEXT: v_bfe_u32 v1, v28, 8, 8
-; VI-NEXT: buffer_store_dword v13, off, s[0:3], s32 offset:92 ; 4-byte Folded Spill
-; VI-NEXT: v_lshrrev_b32_e32 v13, 8, v12
-; VI-NEXT: buffer_store_dword v9, off, s[0:3], s32 offset:116 ; 4-byte Folded Spill
-; VI-NEXT: v_lshrrev_b32_e32 v9, 8, v7
-; VI-NEXT: v_lshrrev_b32_e32 v7, 8, v6
-; VI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:112 ; 4-byte Folded Spill
-; VI-NEXT: v_bfe_u32 v1, v30, 8, 8
-; VI-NEXT: v_lshrrev_b32_e32 v16, 8, v16
-; VI-NEXT: v_lshrrev_b32_e32 v15, 8, v15
-; VI-NEXT: v_lshrrev_b32_e32 v14, 8, v14
-; VI-NEXT: buffer_store_dword v13, off, s[0:3], s32 offset:96 ; 4-byte Folded Spill
-; VI-NEXT: v_lshrrev_b32_e32 v13, 8, v11
-; VI-NEXT: v_lshrrev_b32_e32 v11, 8, v10
-; VI-NEXT: buffer_store_dword v7, off, s[0:3], s32 offset:128 ; 4-byte Folded Spill
-; VI-NEXT: v_lshrrev_b32_e32 v7, 8, v5
-; VI-NEXT: v_lshrrev_b32_e32 v55, 8, v4
-; VI-NEXT: v_lshrrev_b32_e32 v51, 8, v3
-; VI-NEXT: v_lshrrev_b32_e32 v45, 8, v2
-; VI-NEXT: v_bfe_u32 v25, v26, 8, 8
-; VI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:124 ; 4-byte Folded Spill
-; VI-NEXT: v_bfe_u32 v1, v32, 8, 8
-; VI-NEXT: v_bfe_u32 v43, v34, 8, 8
-; VI-NEXT: v_bfe_u32 v46, v36, 8, 8
-; VI-NEXT: v_bfe_u32 v59, v38, 8, 8
-; VI-NEXT: buffer_store_dword v16, off, s[0:3], s32 offset:80 ; 4-byte Folded Spill
-; VI-NEXT: buffer_store_dword v15, off, s[0:3], s32 offset:84 ; 4-byte Folded Spill
-; VI-NEXT: buffer_store_dword v14, off, s[0:3], s32 offset:88 ; 4-byte Folded Spill
-; VI-NEXT: buffer_store_dword v13, off, s[0:3], s32 offset:100 ; 4-byte Folded Spill
-; VI-NEXT: buffer_store_dword v11, off, s[0:3], s32 offset:108 ; 4-byte Folded Spill
-; VI-NEXT: buffer_store_dword v9, off, s[0:3], s32 offset:120 ; 4-byte Folded Spill
-; VI-NEXT: buffer_store_dword v7, off, s[0:3], s32 offset:136 ; 4-byte Folded Spill
-; VI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:132 ; 4-byte Folded Spill
+; VI-NEXT: v_lshrrev_b64 v[21:22], 24, v[46:47]
+; VI-NEXT: v_lshrrev_b32_e32 v22, 8, v16
+; VI-NEXT: v_lshrrev_b32_e32 v16, 8, v15
+; VI-NEXT: v_lshrrev_b32_e32 v1, 8, v14
+; VI-NEXT: v_mov_b32_e32 v14, v62
+; VI-NEXT: v_mov_b32_e32 v15, v40
+; VI-NEXT: v_mov_b32_e32 v40, v16
+; VI-NEXT: v_mov_b32_e32 v16, v55
+; VI-NEXT: v_mov_b32_e32 v55, v22
+; VI-NEXT: v_lshrrev_b32_e32 v62, 8, v23
+; VI-NEXT: v_lshrrev_b64 v[22:23], 24, v[23:24]
+; VI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:64 ; 4-byte Folded Spill
+; VI-NEXT: v_lshrrev_b32_e32 v1, 8, v13
+; VI-NEXT: v_lshrrev_b32_e32 v42, 8, v11
+; VI-NEXT: v_lshrrev_b32_e32 v44, 8, v9
+; VI-NEXT: v_lshrrev_b32_e32 v45, 8, v47
+; VI-NEXT: v_lshrrev_b32_e32 v47, 8, v26
+; VI-NEXT: v_mov_b32_e32 v9, v59
+; VI-NEXT: v_mov_b32_e32 v11, v61
+; VI-NEXT: v_lshrrev_b32_e32 v59, 8, v25
+; VI-NEXT: v_lshrrev_b64 v[25:26], 24, v[25:26]
+; VI-NEXT: v_lshrrev_b32_e32 v61, 8, v24
+; VI-NEXT: v_lshrrev_b64 v[23:24], 24, v[27:28]
+; VI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:68 ; 4-byte Folded Spill
+; VI-NEXT: v_lshrrev_b32_e32 v41, 8, v12
+; VI-NEXT: v_lshrrev_b32_e32 v43, 8, v10
+; VI-NEXT: v_lshrrev_b32_e32 v46, 8, v46
+; VI-NEXT: v_mov_b32_e32 v1, v2
+; VI-NEXT: v_mov_b32_e32 v2, v53
+; VI-NEXT: v_mov_b32_e32 v3, v4
+; VI-NEXT: v_mov_b32_e32 v4, v54
+; VI-NEXT: v_mov_b32_e32 v5, v6
+; VI-NEXT: v_mov_b32_e32 v6, v56
+; VI-NEXT: v_mov_b32_e32 v7, v8
+; VI-NEXT: v_mov_b32_e32 v8, v57
+; VI-NEXT: v_mov_b32_e32 v10, v58
+; VI-NEXT: v_mov_b32_e32 v12, v60
+; VI-NEXT: v_mov_b32_e32 v13, v63
+; VI-NEXT: v_lshrrev_b32_e32 v53, 8, v28
+; VI-NEXT: v_lshrrev_b32_e32 v26, 8, v27
+; VI-NEXT: v_bfe_u32 v27, v29, 8, 8
+; VI-NEXT: v_bfe_u32 v28, v30, 8, 8
+; VI-NEXT: v_bfe_u32 v56, v31, 8, 8
+; VI-NEXT: v_bfe_u32 v57, v33, 8, 8
+; VI-NEXT: v_bfe_u32 v58, v35, 8, 8
+; VI-NEXT: v_bfe_u32 v60, v37, 8, 8
+; VI-NEXT: v_bfe_u32 v63, v48, 8, 8
+; VI-NEXT: v_bfe_u32 v54, v50, 8, 8
; VI-NEXT: .LBB96_4: ; %end
; VI-NEXT: s_or_b64 exec, exec, s[4:5]
-; VI-NEXT: v_lshlrev_b16_e32 v1, 8, v18
-; VI-NEXT: v_lshlrev_b16_e32 v2, 8, v24
-; VI-NEXT: v_or_b32_sdwa v1, v41, v1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
-; VI-NEXT: v_or_b32_sdwa v2, v49, v2 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
-; VI-NEXT: v_or_b32_sdwa v1, v1, v2 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_0 src1_sel:DWORD
+; VI-NEXT: v_lshlrev_b16_e32 v24, 8, v26
+; VI-NEXT: v_lshlrev_b16_e32 v23, 8, v23
+; VI-NEXT: v_or_b32_sdwa v1, v1, v24 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; VI-NEXT: v_or_b32_sdwa v23, v52, v23 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; VI-NEXT: v_or_b32_sdwa v1, v1, v23 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_0 src1_sel:DWORD
+; VI-NEXT: v_lshlrev_b16_e32 v24, 8, v53
; VI-NEXT: buffer_store_dword v1, v0, s[0:3], 0 offen
-; VI-NEXT: v_lshlrev_b16_e32 v1, 8, v45
-; VI-NEXT: v_lshlrev_b16_e32 v2, 8, v59
-; VI-NEXT: v_or_b32_sdwa v1, v54, v1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
-; VI-NEXT: v_or_b32_sdwa v2, v38, v2 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
-; VI-NEXT: v_or_b32_sdwa v1, v1, v2 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_0 src1_sel:DWORD
+; VI-NEXT: v_lshlrev_b16_e32 v1, 8, v54
+; VI-NEXT: v_or_b32_sdwa v2, v2, v24 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; VI-NEXT: v_or_b32_sdwa v1, v50, v1 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; VI-NEXT: v_or_b32_sdwa v1, v2, v1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_0 src1_sel:DWORD
; VI-NEXT: v_add_u32_e32 v2, vcc, 4, v0
; VI-NEXT: buffer_store_dword v1, v2, s[0:3], 0 offen
-; VI-NEXT: v_lshlrev_b16_e32 v1, 8, v51
-; VI-NEXT: v_lshlrev_b16_e32 v2, 8, v23
-; VI-NEXT: v_or_b32_sdwa v1, v57, v1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
-; VI-NEXT: v_or_b32_sdwa v2, v48, v2 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
-; VI-NEXT: v_or_b32_sdwa v1, v1, v2 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_0 src1_sel:DWORD
+; VI-NEXT: v_lshlrev_b16_e32 v1, 8, v22
+; VI-NEXT: v_lshlrev_b16_e32 v2, 8, v62
+; VI-NEXT: v_or_b32_sdwa v1, v51, v1 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; VI-NEXT: v_or_b32_sdwa v2, v3, v2 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; VI-NEXT: v_or_b32_sdwa v1, v2, v1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_0 src1_sel:DWORD
; VI-NEXT: v_add_u32_e32 v2, vcc, 8, v0
; VI-NEXT: buffer_store_dword v1, v2, s[0:3], 0 offen
-; VI-NEXT: v_lshlrev_b16_e32 v1, 8, v55
-; VI-NEXT: v_lshlrev_b16_e32 v2, 8, v46
-; VI-NEXT: v_or_b32_sdwa v1, v47, v1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
-; VI-NEXT: v_or_b32_sdwa v2, v36, v2 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; VI-NEXT: v_lshlrev_b16_e32 v1, 8, v61
+; VI-NEXT: v_lshlrev_b16_e32 v2, 8, v63
+; VI-NEXT: v_or_b32_sdwa v1, v4, v1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; VI-NEXT: v_or_b32_sdwa v2, v48, v2 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
; VI-NEXT: v_or_b32_sdwa v1, v1, v2 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_0 src1_sel:DWORD
; VI-NEXT: v_add_u32_e32 v2, vcc, 12, v0
; VI-NEXT: buffer_store_dword v1, v2, s[0:3], 0 offen
-; VI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:136 ; 4-byte Folded Reload
-; VI-NEXT: v_lshlrev_b16_e32 v2, 8, v22
-; VI-NEXT: v_or_b32_sdwa v2, v39, v2 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
-; VI-NEXT: s_waitcnt vmcnt(0)
-; VI-NEXT: v_lshlrev_b16_e32 v1, 8, v1
-; VI-NEXT: v_or_b32_sdwa v1, v61, v1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; VI-NEXT: v_lshlrev_b16_e32 v1, 8, v59
+; VI-NEXT: v_lshlrev_b16_e32 v2, 8, v25
+; VI-NEXT: v_or_b32_sdwa v1, v5, v1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; VI-NEXT: v_or_b32_sdwa v2, v49, v2 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
; VI-NEXT: v_or_b32_sdwa v1, v1, v2 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_0 src1_sel:DWORD
; VI-NEXT: v_add_u32_e32 v2, vcc, 16, v0
; VI-NEXT: buffer_store_dword v1, v2, s[0:3], 0 offen
-; VI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:128 ; 4-byte Folded Reload
-; VI-NEXT: v_lshlrev_b16_e32 v2, 8, v43
-; VI-NEXT: v_or_b32_sdwa v2, v34, v2 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
-; VI-NEXT: s_waitcnt vmcnt(0)
-; VI-NEXT: v_lshlrev_b16_e32 v1, 8, v1
-; VI-NEXT: v_or_b32_sdwa v1, v60, v1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; VI-NEXT: v_lshlrev_b16_e32 v1, 8, v47
+; VI-NEXT: v_lshlrev_b16_e32 v2, 8, v60
+; VI-NEXT: v_or_b32_sdwa v1, v6, v1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; VI-NEXT: v_or_b32_sdwa v2, v37, v2 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
; VI-NEXT: v_or_b32_sdwa v1, v1, v2 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_0 src1_sel:DWORD
; VI-NEXT: v_add_u32_e32 v2, vcc, 20, v0
; VI-NEXT: buffer_store_dword v1, v2, s[0:3], 0 offen
-; VI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:120 ; 4-byte Folded Reload
+; VI-NEXT: v_lshlrev_b16_e32 v1, 8, v46
; VI-NEXT: v_lshlrev_b16_e32 v2, 8, v21
-; VI-NEXT: v_or_b32_sdwa v2, v37, v2 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
-; VI-NEXT: s_waitcnt vmcnt(0)
-; VI-NEXT: v_lshlrev_b16_e32 v1, 8, v1
-; VI-NEXT: v_or_b32_sdwa v1, v52, v1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; VI-NEXT: v_or_b32_sdwa v1, v7, v1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; VI-NEXT: v_or_b32_sdwa v2, v39, v2 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
; VI-NEXT: v_or_b32_sdwa v1, v1, v2 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_0 src1_sel:DWORD
; VI-NEXT: v_add_u32_e32 v2, vcc, 24, v0
; VI-NEXT: buffer_store_dword v1, v2, s[0:3], 0 offen
-; VI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:116 ; 4-byte Folded Reload
-; VI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:132 ; 4-byte Folded Reload
-; VI-NEXT: s_waitcnt vmcnt(1)
-; VI-NEXT: v_lshlrev_b16_e32 v1, 8, v1
-; VI-NEXT: s_waitcnt vmcnt(0)
-; VI-NEXT: v_lshlrev_b16_e32 v2, 8, v2
-; VI-NEXT: v_or_b32_sdwa v1, v63, v1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
-; VI-NEXT: v_or_b32_sdwa v2, v32, v2 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; VI-NEXT: v_lshlrev_b16_e32 v1, 8, v45
+; VI-NEXT: v_lshlrev_b16_e32 v2, 8, v58
+; VI-NEXT: v_or_b32_sdwa v1, v8, v1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; VI-NEXT: v_or_b32_sdwa v2, v35, v2 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
; VI-NEXT: v_or_b32_sdwa v1, v1, v2 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_0 src1_sel:DWORD
; VI-NEXT: v_add_u32_e32 v2, vcc, 28, v0
; VI-NEXT: buffer_store_dword v1, v2, s[0:3], 0 offen
-; VI-NEXT: v_lshlrev_b16_e32 v1, 8, v42
+; VI-NEXT: v_lshlrev_b16_e32 v1, 8, v44
; VI-NEXT: v_lshlrev_b16_e32 v2, 8, v20
-; VI-NEXT: v_or_b32_sdwa v1, v40, v1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
-; VI-NEXT: v_or_b32_sdwa v2, v35, v2 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; VI-NEXT: v_or_b32_sdwa v1, v9, v1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; VI-NEXT: v_or_b32_sdwa v2, v38, v2 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
; VI-NEXT: v_or_b32_sdwa v1, v1, v2 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_0 src1_sel:DWORD
; VI-NEXT: v_add_u32_e32 v2, vcc, 32, v0
; VI-NEXT: buffer_store_dword v1, v2, s[0:3], 0 offen
-; VI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:108 ; 4-byte Folded Reload
-; VI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:124 ; 4-byte Folded Reload
-; VI-NEXT: s_waitcnt vmcnt(1)
-; VI-NEXT: v_lshlrev_b16_e32 v1, 8, v1
-; VI-NEXT: s_waitcnt vmcnt(0)
-; VI-NEXT: v_lshlrev_b16_e32 v2, 8, v2
-; VI-NEXT: v_or_b32_sdwa v1, v53, v1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
-; VI-NEXT: v_or_b32_sdwa v2, v30, v2 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; VI-NEXT: v_lshlrev_b16_e32 v1, 8, v43
+; VI-NEXT: v_lshlrev_b16_e32 v2, 8, v57
+; VI-NEXT: v_or_b32_sdwa v1, v10, v1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; VI-NEXT: v_or_b32_sdwa v2, v33, v2 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
; VI-NEXT: v_or_b32_sdwa v1, v1, v2 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_0 src1_sel:DWORD
; VI-NEXT: v_add_u32_e32 v2, vcc, 36, v0
; VI-NEXT: buffer_store_dword v1, v2, s[0:3], 0 offen
-; VI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:100 ; 4-byte Folded Reload
+; VI-NEXT: v_lshlrev_b16_e32 v1, 8, v42
; VI-NEXT: v_lshlrev_b16_e32 v2, 8, v19
-; VI-NEXT: v_or_b32_sdwa v2, v33, v2 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
-; VI-NEXT: s_waitcnt vmcnt(0)
-; VI-NEXT: v_lshlrev_b16_e32 v1, 8, v1
-; VI-NEXT: v_or_b32_sdwa v1, v17, v1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; VI-NEXT: v_or_b32_sdwa v1, v11, v1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; VI-NEXT: v_or_b32_sdwa v2, v36, v2 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
; VI-NEXT: v_or_b32_sdwa v1, v1, v2 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_0 src1_sel:DWORD
; VI-NEXT: v_add_u32_e32 v2, vcc, 40, v0
; VI-NEXT: buffer_store_dword v1, v2, s[0:3], 0 offen
-; VI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:96 ; 4-byte Folded Reload
-; VI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:112 ; 4-byte Folded Reload
-; VI-NEXT: s_waitcnt vmcnt(1)
-; VI-NEXT: v_lshlrev_b16_e32 v1, 8, v1
-; VI-NEXT: s_waitcnt vmcnt(0)
-; VI-NEXT: v_lshlrev_b16_e32 v2, 8, v2
-; VI-NEXT: v_or_b32_sdwa v1, v44, v1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
-; VI-NEXT: v_or_b32_sdwa v2, v28, v2 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; VI-NEXT: v_lshlrev_b16_e32 v1, 8, v41
+; VI-NEXT: v_lshlrev_b16_e32 v2, 8, v56
+; VI-NEXT: v_or_b32_sdwa v1, v12, v1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; VI-NEXT: v_or_b32_sdwa v2, v31, v2 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
; VI-NEXT: v_or_b32_sdwa v1, v1, v2 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_0 src1_sel:DWORD
; VI-NEXT: v_add_u32_e32 v2, vcc, 44, v0
; VI-NEXT: buffer_store_dword v1, v2, s[0:3], 0 offen
-; VI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:92 ; 4-byte Folded Reload
-; VI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:72 ; 4-byte Folded Reload
-; VI-NEXT: buffer_load_dword v3, off, s[0:3], s32 offset:76 ; 4-byte Folded Reload
-; VI-NEXT: s_waitcnt vmcnt(1)
-; VI-NEXT: v_lshlrev_b16_e32 v2, 8, v2
+; VI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:68 ; 4-byte Folded Reload
+; VI-NEXT: v_lshlrev_b16_e32 v2, 8, v18
+; VI-NEXT: v_or_b32_sdwa v2, v34, v2 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; VI-NEXT: s_waitcnt vmcnt(0)
; VI-NEXT: v_lshlrev_b16_e32 v1, 8, v1
-; VI-NEXT: v_or_b32_sdwa v1, v58, v1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
-; VI-NEXT: v_or_b32_sdwa v2, v31, v2 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; VI-NEXT: v_or_b32_sdwa v1, v13, v1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
; VI-NEXT: v_or_b32_sdwa v1, v1, v2 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_0 src1_sel:DWORD
; VI-NEXT: v_add_u32_e32 v2, vcc, 48, v0
; VI-NEXT: buffer_store_dword v1, v2, s[0:3], 0 offen
-; VI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:88 ; 4-byte Folded Reload
-; VI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:104 ; 4-byte Folded Reload
-; VI-NEXT: s_waitcnt vmcnt(1)
-; VI-NEXT: v_lshlrev_b16_e32 v1, 8, v1
+; VI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:64 ; 4-byte Folded Reload
+; VI-NEXT: v_lshlrev_b16_e32 v2, 8, v28
+; VI-NEXT: v_or_b32_sdwa v2, v30, v2 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
; VI-NEXT: s_waitcnt vmcnt(0)
-; VI-NEXT: v_lshlrev_b16_e32 v2, 8, v2
-; VI-NEXT: v_or_b32_sdwa v1, v56, v1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
-; VI-NEXT: v_or_b32_sdwa v2, v27, v2 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; VI-NEXT: v_lshlrev_b16_e32 v1, 8, v1
+; VI-NEXT: v_or_b32_sdwa v1, v14, v1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
; VI-NEXT: v_or_b32_sdwa v1, v1, v2 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_0 src1_sel:DWORD
; VI-NEXT: v_add_u32_e32 v2, vcc, 52, v0
; VI-NEXT: buffer_store_dword v1, v2, s[0:3], 0 offen
-; VI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:84 ; 4-byte Folded Reload
-; VI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:64 ; 4-byte Folded Reload
-; VI-NEXT: buffer_load_dword v3, off, s[0:3], s32 offset:68 ; 4-byte Folded Reload
-; VI-NEXT: s_waitcnt vmcnt(1)
-; VI-NEXT: v_lshlrev_b16_e32 v2, 8, v2
-; VI-NEXT: v_lshlrev_b16_e32 v1, 8, v1
-; VI-NEXT: v_or_b32_sdwa v1, v50, v1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
-; VI-NEXT: v_or_b32_sdwa v2, v29, v2 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; VI-NEXT: v_lshlrev_b16_e32 v1, 8, v40
+; VI-NEXT: v_lshlrev_b16_e32 v2, 8, v17
+; VI-NEXT: v_or_b32_sdwa v1, v15, v1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; VI-NEXT: v_or_b32_sdwa v2, v32, v2 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
; VI-NEXT: v_or_b32_sdwa v1, v1, v2 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_0 src1_sel:DWORD
; VI-NEXT: v_add_u32_e32 v2, vcc, 56, v0
; VI-NEXT: buffer_store_dword v1, v2, s[0:3], 0 offen
-; VI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:80 ; 4-byte Folded Reload
-; VI-NEXT: v_lshlrev_b16_e32 v2, 8, v25
-; VI-NEXT: v_or_b32_sdwa v2, v26, v2 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
-; VI-NEXT: v_add_u32_e32 v0, vcc, 60, v0
-; VI-NEXT: s_waitcnt vmcnt(0)
-; VI-NEXT: v_lshlrev_b16_e32 v1, 8, v1
-; VI-NEXT: v_or_b32_sdwa v1, v62, v1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; VI-NEXT: v_lshlrev_b16_e32 v1, 8, v55
+; VI-NEXT: v_lshlrev_b16_e32 v2, 8, v27
+; VI-NEXT: v_or_b32_sdwa v1, v16, v1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; VI-NEXT: v_or_b32_sdwa v2, v29, v2 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
; VI-NEXT: v_or_b32_sdwa v1, v1, v2 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_0 src1_sel:DWORD
+; VI-NEXT: v_add_u32_e32 v0, vcc, 60, v0
; VI-NEXT: buffer_store_dword v1, v0, s[0:3], 0 offen
; VI-NEXT: buffer_load_dword v63, off, s[0:3], s32 ; 4-byte Folded Reload
; VI-NEXT: buffer_load_dword v62, off, s[0:3], s32 offset:4 ; 4-byte Folded Reload
@@ -71641,7 +71551,7 @@ define inreg <32 x i16> @bitcast_v64i8_to_v32i16_scalar(<64 x i8> inreg %a, i32
; SI-NEXT: buffer_load_dword v62, off, s[0:3], s32 offset:72
; SI-NEXT: buffer_load_dword v57, off, s[0:3], s32 offset:52
; SI-NEXT: buffer_load_dword v46, off, s[0:3], s32 offset:68
-; SI-NEXT: v_readfirstlane_b32 s15, v27
+; SI-NEXT: v_readfirstlane_b32 s14, v27
; SI-NEXT: v_readfirstlane_b32 s40, v26
; SI-NEXT: v_readfirstlane_b32 s12, v19
; SI-NEXT: v_readfirstlane_b32 s13, v18
@@ -71690,17 +71600,17 @@ define inreg <32 x i16> @bitcast_v64i8_to_v32i16_scalar(<64 x i8> inreg %a, i32
; SI-NEXT: s_or_b32 s4, s4, s5
; SI-NEXT: s_and_b32 s5, s22, 0xff
; SI-NEXT: s_lshl_b32 s5, s5, 16
-; SI-NEXT: s_lshl_b32 s14, s23, 24
+; SI-NEXT: s_lshl_b32 s15, s23, 24
; SI-NEXT: s_and_b32 s4, s4, 0xffff
-; SI-NEXT: s_or_b32 s5, s14, s5
+; SI-NEXT: s_or_b32 s5, s15, s5
; SI-NEXT: s_or_b32 s41, s4, s5
; SI-NEXT: s_and_b32 s4, s18, 0xff
; SI-NEXT: s_lshl_b32 s4, s4, 16
-; SI-NEXT: s_lshl_b32 s14, s19, 24
-; SI-NEXT: s_or_b32 s4, s14, s4
-; SI-NEXT: s_and_b32 s14, s28, 0xff
+; SI-NEXT: s_lshl_b32 s15, s19, 24
+; SI-NEXT: s_or_b32 s4, s15, s4
+; SI-NEXT: s_and_b32 s15, s28, 0xff
; SI-NEXT: s_lshl_b32 s46, s29, 8
-; SI-NEXT: s_or_b32 s14, s14, s46
+; SI-NEXT: s_or_b32 s15, s15, s46
; SI-NEXT: s_and_b32 s46, s6, 0xff
; SI-NEXT: s_lshl_b32 s46, s46, 16
; SI-NEXT: s_lshl_b32 s47, s7, 24
@@ -71789,7 +71699,7 @@ define inreg <32 x i16> @bitcast_v64i8_to_v32i16_scalar(<64 x i8> inreg %a, i32
; SI-NEXT: v_or_b32_e32 v63, v59, v34
; SI-NEXT: v_or_b32_e32 v39, s4, v25
; SI-NEXT: s_and_b32 s4, s40, 0xff
-; SI-NEXT: s_lshl_b32 s56, s15, 8
+; SI-NEXT: s_lshl_b32 s56, s14, 8
; SI-NEXT: v_lshlrev_b32_e32 v21, 16, v21
; SI-NEXT: v_or_b32_e32 v48, v32, v63
; SI-NEXT: v_and_b32_e32 v32, 0xff, v57
@@ -71812,12 +71722,12 @@ define inreg <32 x i16> @bitcast_v64i8_to_v32i16_scalar(<64 x i8> inreg %a, i32
; SI-NEXT: v_or_b32_e32 v33, s4, v33
; SI-NEXT: s_and_b32 s4, s45, 0xff
; SI-NEXT: s_lshl_b32 s56, s44, 8
-; SI-NEXT: s_and_b32 s14, s14, 0xffff
+; SI-NEXT: s_and_b32 s15, s15, 0xffff
; SI-NEXT: s_or_b32 s4, s4, s56
-; SI-NEXT: s_or_b32 s14, s14, s57
+; SI-NEXT: s_or_b32 s15, s15, s57
; SI-NEXT: s_and_b32 s4, s4, 0xffff
; SI-NEXT: v_alignbit_b32 v1, s41, v1, 16
-; SI-NEXT: v_alignbit_b32 v5, s14, v5, 16
+; SI-NEXT: v_alignbit_b32 v5, s15, v5, 16
; SI-NEXT: v_alignbit_b32 v9, v10, v15, 16
; SI-NEXT: v_alignbit_b32 v13, v36, v23, 16
; SI-NEXT: v_alignbit_b32 v21, v38, v29, 16
@@ -71883,7 +71793,7 @@ define inreg <32 x i16> @bitcast_v64i8_to_v32i16_scalar(<64 x i8> inreg %a, i32
; SI-NEXT: s_add_i32 s40, s40, 3
; SI-NEXT: v_add_i32_e32 v26, vcc, 0x3000000, v1
; SI-NEXT: s_and_b32 s4, s40, 0xff
-; SI-NEXT: s_lshl_b32 s5, s15, 8
+; SI-NEXT: s_lshl_b32 s5, s14, 8
; SI-NEXT: v_add_i32_e32 v1, vcc, 3, v28
; SI-NEXT: s_or_b32 s4, s5, s4
; SI-NEXT: v_and_b32_e32 v1, 0xff, v1
@@ -71987,7 +71897,7 @@ define inreg <32 x i16> @bitcast_v64i8_to_v32i16_scalar(<64 x i8> inreg %a, i32
; SI-NEXT: s_or_b32 s5, s5, s6
; SI-NEXT: s_or_b32 s4, s5, s4
; SI-NEXT: s_add_i32 s16, s16, 3
-; SI-NEXT: s_add_i32 s14, s4, 0x3000000
+; SI-NEXT: s_add_i32 s15, s4, 0x3000000
; SI-NEXT: s_and_b32 s4, s16, 0xff
; SI-NEXT: s_lshl_b32 s5, s17, 8
; SI-NEXT: s_add_i32 s18, s18, 3
@@ -72028,7 +71938,7 @@ define inreg <32 x i16> @bitcast_v64i8_to_v32i16_scalar(<64 x i8> inreg %a, i32
; SI-NEXT: v_mov_b32_e32 v0, s47
; SI-NEXT: v_alignbit_b32 v1, s41, v0, 16
; SI-NEXT: v_mov_b32_e32 v0, s46
-; SI-NEXT: v_alignbit_b32 v5, s14, v0, 16
+; SI-NEXT: v_alignbit_b32 v5, s15, v0, 16
; SI-NEXT: v_alignbit_b32 v9, v10, v35, 16
; SI-NEXT: v_alignbit_b32 v13, v36, v37, 16
; SI-NEXT: v_alignbit_b32 v17, v18, v39, 16
@@ -72036,7 +71946,7 @@ define inreg <32 x i16> @bitcast_v64i8_to_v32i16_scalar(<64 x i8> inreg %a, i32
; SI-NEXT: v_alignbit_b32 v25, v26, v33, 16
; SI-NEXT: v_alignbit_b32 v29, v48, v34, 16
; SI-NEXT: s_lshr_b32 s56, s41, 16
-; SI-NEXT: s_lshr_b32 s57, s14, 16
+; SI-NEXT: s_lshr_b32 s57, s15, 16
; SI-NEXT: v_lshrrev_b32_e32 v11, 16, v10
; SI-NEXT: v_lshrrev_b32_e32 v15, 16, v36
; SI-NEXT: v_lshrrev_b32_e32 v19, 16, v18
@@ -72065,7 +71975,7 @@ define inreg <32 x i16> @bitcast_v64i8_to_v32i16_scalar(<64 x i8> inreg %a, i32
; SI-NEXT: v_mov_b32_e32 v2, s41
; SI-NEXT: v_mov_b32_e32 v3, s56
; SI-NEXT: v_mov_b32_e32 v4, s46
-; SI-NEXT: v_mov_b32_e32 v6, s14
+; SI-NEXT: v_mov_b32_e32 v6, s15
; SI-NEXT: v_mov_b32_e32 v7, s57
; SI-NEXT: v_mov_b32_e32 v8, v35
; SI-NEXT: v_mov_b32_e32 v12, v37
@@ -72085,7 +71995,7 @@ define inreg <32 x i16> @bitcast_v64i8_to_v32i16_scalar(<64 x i8> inreg %a, i32
; SI-NEXT: ; implicit-def: $sgpr56
; SI-NEXT: ; implicit-def: $sgpr46
; SI-NEXT: ; implicit-def: $vgpr5
-; SI-NEXT: ; implicit-def: $sgpr14
+; SI-NEXT: ; implicit-def: $sgpr15
; SI-NEXT: ; implicit-def: $sgpr57
; SI-NEXT: ; implicit-def: $vgpr35
; SI-NEXT: ; implicit-def: $vgpr9
@@ -72192,11 +72102,11 @@ define inreg <32 x i16> @bitcast_v64i8_to_v32i16_scalar(<64 x i8> inreg %a, i32
; VI-NEXT: s_waitcnt vmcnt(11)
; VI-NEXT: v_lshlrev_b32_e32 v14, 8, v14
; VI-NEXT: s_waitcnt vmcnt(9)
-; VI-NEXT: v_lshlrev_b32_e32 v38, 8, v39
+; VI-NEXT: v_lshlrev_b32_e32 v22, 8, v39
; VI-NEXT: s_waitcnt vmcnt(7)
; VI-NEXT: v_lshlrev_b32_e32 v51, 8, v48
; VI-NEXT: s_waitcnt vmcnt(5)
-; VI-NEXT: v_lshlrev_b32_e32 v22, 8, v49
+; VI-NEXT: v_lshlrev_b32_e32 v38, 8, v49
; VI-NEXT: s_cbranch_scc0 .LBB99_4
; VI-NEXT: ; %bb.1: ; %cmp.false
; VI-NEXT: v_or_b32_sdwa v0, v2, v1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
@@ -72255,10 +72165,10 @@ define inreg <32 x i16> @bitcast_v64i8_to_v32i16_scalar(<64 x i8> inreg %a, i32
; VI-NEXT: v_or_b32_sdwa v1, v43, v33 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
; VI-NEXT: v_or_b32_sdwa v13, v0, v1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_0 src1_sel:DWORD
; VI-NEXT: v_or_b32_sdwa v0, v44, v14 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
-; VI-NEXT: v_or_b32_sdwa v1, v45, v38 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; VI-NEXT: v_or_b32_sdwa v1, v45, v22 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
; VI-NEXT: v_or_b32_sdwa v14, v0, v1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_0 src1_sel:DWORD
; VI-NEXT: v_or_b32_sdwa v0, v47, v51 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
-; VI-NEXT: v_or_b32_sdwa v1, v57, v22 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; VI-NEXT: v_or_b32_sdwa v1, v57, v38 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
; VI-NEXT: v_or_b32_sdwa v15, v0, v1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_0 src1_sel:DWORD
; VI-NEXT: v_or_b32_sdwa v0, v32, v50 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
; VI-NEXT: v_or_b32_e32 v3, s4, v0
@@ -72318,9 +72228,9 @@ define inreg <32 x i16> @bitcast_v64i8_to_v32i16_scalar(<64 x i8> inreg %a, i32
; VI-NEXT: v_add_u32_e32 v3, vcc, 3, v24
; VI-NEXT: v_or_b32_sdwa v9, v58, v3 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:BYTE_0
; VI-NEXT: v_add_u32_e32 v3, vcc, 3, v23
-; VI-NEXT: v_add_u32_e32 v0, vcc, 3, v57
+; VI-NEXT: v_add_u32_e32 v2, vcc, 3, v45
; VI-NEXT: v_or_b32_sdwa v3, v56, v3 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:BYTE_0
-; VI-NEXT: v_or_b32_sdwa v0, v22, v0 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:BYTE_0
+; VI-NEXT: v_or_b32_sdwa v2, v22, v2 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:BYTE_0
; VI-NEXT: v_add_u32_e32 v22, vcc, 0x300, v3
; VI-NEXT: v_add_u32_e32 v3, vcc, 3, v55
; VI-NEXT: v_or_b32_sdwa v8, v46, v3 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:BYTE_0
@@ -72377,8 +72287,8 @@ define inreg <32 x i16> @bitcast_v64i8_to_v32i16_scalar(<64 x i8> inreg %a, i32
; VI-NEXT: s_addk_i32 s6, 0x300
; VI-NEXT: s_addk_i32 s8, 0x300
; VI-NEXT: s_addk_i32 s10, 0x300
+; VI-NEXT: v_add_u32_e32 v0, vcc, 3, v57
; VI-NEXT: v_or_b32_sdwa v1, v48, v1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:BYTE_0
-; VI-NEXT: v_add_u32_e32 v2, vcc, 3, v45
; VI-NEXT: s_addk_i32 s4, 0x300
; VI-NEXT: s_lshl_b32 s5, s5, 16
; VI-NEXT: s_lshl_b32 s7, s7, 16
@@ -72386,8 +72296,8 @@ define inreg <32 x i16> @bitcast_v64i8_to_v32i16_scalar(<64 x i8> inreg %a, i32
; VI-NEXT: s_and_b32 s10, s10, 0xffff
; VI-NEXT: s_and_b32 s8, s8, 0xffff
; VI-NEXT: s_and_b32 s6, s6, 0xffff
+; VI-NEXT: v_or_b32_sdwa v0, v38, v0 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:BYTE_0
; VI-NEXT: v_add_u32_e32 v1, vcc, 0x300, v1
-; VI-NEXT: v_or_b32_sdwa v2, v38, v2 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:BYTE_0
; VI-NEXT: s_or_b32 s9, s9, s10
; VI-NEXT: s_or_b32 s7, s7, s8
; VI-NEXT: s_or_b32 s5, s5, s6
@@ -80753,7 +80663,6 @@ define inreg <64 x i8> @bitcast_v32f16_to_v64i8_scalar(<32 x half> inreg %a, i32
; VI-NEXT: buffer_store_dword v52, off, s[0:3], s32 offset:72 ; 4-byte Folded Spill
; VI-NEXT: buffer_store_dword v53, off, s[0:3], s32 offset:76 ; 4-byte Folded Spill
; VI-NEXT: v_mov_b32_e32 v52, s44
-; VI-NEXT: v_mov_b32_e32 v19, s67
; VI-NEXT: v_mov_b32_e32 v12, s66
; VI-NEXT: v_mov_b32_e32 v20, s65
; VI-NEXT: v_mov_b32_e32 v13, s64
@@ -80812,6 +80721,7 @@ define inreg <64 x i8> @bitcast_v32f16_to_v64i8_scalar(<32 x half> inreg %a, i32
; VI-NEXT: v_mov_b32_e32 v4, s14
; VI-NEXT: v_mov_b32_e32 v3, s40
; VI-NEXT: v_mov_b32_e32 v9, s75
+; VI-NEXT: v_mov_b32_e32 v19, s67
; VI-NEXT: buffer_store_dword v52, off, s[0:3], s32 offset:64 ; 4-byte Folded Spill
; VI-NEXT: buffer_store_dword v53, off, s[0:3], s32 offset:68 ; 4-byte Folded Spill
; VI-NEXT: v_mov_b32_e32 v52, s62
@@ -85175,11 +85085,11 @@ define inreg <32 x half> @bitcast_v64i8_to_v32f16_scalar(<64 x i8> inreg %a, i32
; VI-NEXT: s_waitcnt vmcnt(11)
; VI-NEXT: v_lshlrev_b32_e32 v14, 8, v14
; VI-NEXT: s_waitcnt vmcnt(9)
-; VI-NEXT: v_lshlrev_b32_e32 v38, 8, v39
+; VI-NEXT: v_lshlrev_b32_e32 v22, 8, v39
; VI-NEXT: s_waitcnt vmcnt(7)
; VI-NEXT: v_lshlrev_b32_e32 v51, 8, v48
; VI-NEXT: s_waitcnt vmcnt(5)
-; VI-NEXT: v_lshlrev_b32_e32 v22, 8, v49
+; VI-NEXT: v_lshlrev_b32_e32 v38, 8, v49
; VI-NEXT: s_cbranch_scc0 .LBB107_4
; VI-NEXT: ; %bb.1: ; %cmp.false
; VI-NEXT: v_or_b32_sdwa v0, v2, v1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
@@ -85238,10 +85148,10 @@ define inreg <32 x half> @bitcast_v64i8_to_v32f16_scalar(<64 x i8> inreg %a, i32
; VI-NEXT: v_or_b32_sdwa v1, v43, v33 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
; VI-NEXT: v_or_b32_sdwa v13, v0, v1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_0 src1_sel:DWORD
; VI-NEXT: v_or_b32_sdwa v0, v44, v14 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
-; VI-NEXT: v_or_b32_sdwa v1, v45, v38 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; VI-NEXT: v_or_b32_sdwa v1, v45, v22 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
; VI-NEXT: v_or_b32_sdwa v14, v0, v1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_0 src1_sel:DWORD
; VI-NEXT: v_or_b32_sdwa v0, v47, v51 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
-; VI-NEXT: v_or_b32_sdwa v1, v57, v22 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; VI-NEXT: v_or_b32_sdwa v1, v57, v38 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
; VI-NEXT: v_or_b32_sdwa v15, v0, v1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_0 src1_sel:DWORD
; VI-NEXT: v_or_b32_sdwa v0, v32, v50 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
; VI-NEXT: v_or_b32_e32 v3, s4, v0
@@ -85301,9 +85211,9 @@ define inreg <32 x half> @bitcast_v64i8_to_v32f16_scalar(<64 x i8> inreg %a, i32
; VI-NEXT: v_add_u32_e32 v3, vcc, 3, v24
; VI-NEXT: v_or_b32_sdwa v9, v58, v3 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:BYTE_0
; VI-NEXT: v_add_u32_e32 v3, vcc, 3, v23
-; VI-NEXT: v_add_u32_e32 v0, vcc, 3, v57
+; VI-NEXT: v_add_u32_e32 v2, vcc, 3, v45
; VI-NEXT: v_or_b32_sdwa v3, v56, v3 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:BYTE_0
-; VI-NEXT: v_or_b32_sdwa v0, v22, v0 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:BYTE_0
+; VI-NEXT: v_or_b32_sdwa v2, v22, v2 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:BYTE_0
; VI-NEXT: v_add_u32_e32 v22, vcc, 0x300, v3
; VI-NEXT: v_add_u32_e32 v3, vcc, 3, v55
; VI-NEXT: v_or_b32_sdwa v8, v46, v3 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:BYTE_0
@@ -85360,8 +85270,8 @@ define inreg <32 x half> @bitcast_v64i8_to_v32f16_scalar(<64 x i8> inreg %a, i32
; VI-NEXT: s_addk_i32 s6, 0x300
; VI-NEXT: s_addk_i32 s8, 0x300
; VI-NEXT: s_addk_i32 s10, 0x300
+; VI-NEXT: v_add_u32_e32 v0, vcc, 3, v57
; VI-NEXT: v_or_b32_sdwa v1, v48, v1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:BYTE_0
-; VI-NEXT: v_add_u32_e32 v2, vcc, 3, v45
; VI-NEXT: s_addk_i32 s4, 0x300
; VI-NEXT: s_lshl_b32 s5, s5, 16
; VI-NEXT: s_lshl_b32 s7, s7, 16
@@ -85369,8 +85279,8 @@ define inreg <32 x half> @bitcast_v64i8_to_v32f16_scalar(<64 x i8> inreg %a, i32
; VI-NEXT: s_and_b32 s10, s10, 0xffff
; VI-NEXT: s_and_b32 s8, s8, 0xffff
; VI-NEXT: s_and_b32 s6, s6, 0xffff
+; VI-NEXT: v_or_b32_sdwa v0, v38, v0 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:BYTE_0
; VI-NEXT: v_add_u32_e32 v1, vcc, 0x300, v1
-; VI-NEXT: v_or_b32_sdwa v2, v38, v2 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:BYTE_0
; VI-NEXT: s_or_b32 s9, s9, s10
; VI-NEXT: s_or_b32 s7, s7, s8
; VI-NEXT: s_or_b32 s5, s5, s6
@@ -88009,6 +87919,8 @@ define <64 x i8> @bitcast_v32bf16_to_v64i8(<32 x bfloat> %a, i32 %b) {
; GFX9-NEXT: ; implicit-def: $vgpr17
; GFX9-NEXT: ; kill: killed $vgpr17
; GFX9-NEXT: ; implicit-def: $vgpr17
+; GFX9-NEXT: ; kill: killed $vgpr17
+; GFX9-NEXT: ; implicit-def: $vgpr17
; GFX9-NEXT: buffer_store_dword v40, off, s[0:3], s32 offset:60 ; 4-byte Folded Spill
; GFX9-NEXT: buffer_store_dword v41, off, s[0:3], s32 offset:56 ; 4-byte Folded Spill
; GFX9-NEXT: buffer_store_dword v42, off, s[0:3], s32 offset:52 ; 4-byte Folded Spill
@@ -88025,47 +87937,46 @@ define <64 x i8> @bitcast_v32bf16_to_v64i8(<32 x bfloat> %a, i32 %b) {
; GFX9-NEXT: buffer_store_dword v61, off, s[0:3], s32 offset:8 ; 4-byte Folded Spill
; GFX9-NEXT: buffer_store_dword v62, off, s[0:3], s32 offset:4 ; 4-byte Folded Spill
; GFX9-NEXT: buffer_store_dword v63, off, s[0:3], s32 ; 4-byte Folded Spill
-; GFX9-NEXT: ; implicit-def: $vgpr27
-; GFX9-NEXT: ; implicit-def: $vgpr28
-; GFX9-NEXT: ; implicit-def: $vgpr23
-; GFX9-NEXT: ; implicit-def: $vgpr51
-; GFX9-NEXT: ; implicit-def: $vgpr58
-; GFX9-NEXT: ; implicit-def: $vgpr50
-; GFX9-NEXT: ; implicit-def: $vgpr29
-; GFX9-NEXT: ; implicit-def: $vgpr39
-; GFX9-NEXT: ; implicit-def: $vgpr22
+; GFX9-NEXT: ; implicit-def: $vgpr23
+; GFX9-NEXT: ; implicit-def: $vgpr24
+; GFX9-NEXT: ; implicit-def: $vgpr26
+; GFX9-NEXT: ; implicit-def: $vgpr53
+; GFX9-NEXT: ; implicit-def: $vgpr59
+; GFX9-NEXT: ; implicit-def: $vgpr52
; GFX9-NEXT: ; implicit-def: $vgpr30
-; GFX9-NEXT: ; implicit-def: $vgpr45
-; GFX9-NEXT: ; implicit-def: $vgpr63
+; GFX9-NEXT: ; implicit-def: $vgpr51
+; GFX9-NEXT: ; implicit-def: $vgpr25
; GFX9-NEXT: ; implicit-def: $vgpr31
-; GFX9-NEXT: ; implicit-def: $vgpr62
+; GFX9-NEXT: ; implicit-def: $vgpr46
+; GFX9-NEXT: ; implicit-def: $vgpr50
; GFX9-NEXT: ; implicit-def: $vgpr32
-; GFX9-NEXT: ; implicit-def: $vgpr42
-; GFX9-NEXT: ; implicit-def: $vgpr61
+; GFX9-NEXT: ; implicit-def: $vgpr63
; GFX9-NEXT: ; implicit-def: $vgpr33
-; GFX9-NEXT: ; implicit-def: $vgpr60
+; GFX9-NEXT: ; implicit-def: $vgpr43
+; GFX9-NEXT: ; implicit-def: $vgpr62
; GFX9-NEXT: ; implicit-def: $vgpr34
-; GFX9-NEXT: ; implicit-def: $vgpr53
-; GFX9-NEXT: ; implicit-def: $vgpr59
+; GFX9-NEXT: ; implicit-def: $vgpr61
; GFX9-NEXT: ; implicit-def: $vgpr35
-; GFX9-NEXT: ; implicit-def: $vgpr57
+; GFX9-NEXT: ; implicit-def: $vgpr54
+; GFX9-NEXT: ; implicit-def: $vgpr60
; GFX9-NEXT: ; implicit-def: $vgpr36
-; GFX9-NEXT: ; implicit-def: $vgpr52
-; GFX9-NEXT: ; implicit-def: $vgpr56
+; GFX9-NEXT: ; implicit-def: $vgpr58
; GFX9-NEXT: ; implicit-def: $vgpr37
-; GFX9-NEXT: ; implicit-def: $vgpr47
+; GFX9-NEXT: ; implicit-def: $vgpr57
; GFX9-NEXT: ; implicit-def: $vgpr38
-; GFX9-NEXT: ; implicit-def: $vgpr46
+; GFX9-NEXT: ; implicit-def: $vgpr56
+; GFX9-NEXT: ; implicit-def: $vgpr39
+; GFX9-NEXT: ; implicit-def: $vgpr47
; GFX9-NEXT: ; implicit-def: $vgpr48
-; GFX9-NEXT: ; implicit-def: $vgpr44
+; GFX9-NEXT: ; implicit-def: $vgpr45
; GFX9-NEXT: ; implicit-def: $vgpr49
-; GFX9-NEXT: ; implicit-def: $vgpr43
+; GFX9-NEXT: ; implicit-def: $vgpr44
+; GFX9-NEXT: ; implicit-def: $vgpr42
; GFX9-NEXT: ; implicit-def: $vgpr41
; GFX9-NEXT: ; implicit-def: $vgpr40
-; GFX9-NEXT: ; implicit-def: $vgpr55
; GFX9-NEXT: ; kill: killed $vgpr17
-; GFX9-NEXT: ; implicit-def: $vgpr54
-; GFX9-NEXT: ; implicit-def: $vgpr25
+; GFX9-NEXT: ; implicit-def: $vgpr55
+; GFX9-NEXT: ; implicit-def: $vgpr28
; GFX9-NEXT: ; implicit-def: $vgpr21
; GFX9-NEXT: ; implicit-def: $vgpr20
; GFX9-NEXT: ; implicit-def: $vgpr19
@@ -88081,51 +87992,52 @@ define <64 x i8> @bitcast_v32bf16_to_v64i8(<32 x bfloat> %a, i32 %b) {
; GFX9-NEXT: buffer_store_dword v17, off, s[0:3], s32 offset:68 ; 4-byte Folded Spill
; GFX9-NEXT: v_lshrrev_b32_e32 v17, 16, v12
; GFX9-NEXT: buffer_store_dword v17, off, s[0:3], s32 offset:72 ; 4-byte Folded Spill
+; GFX9-NEXT: v_lshrrev_b32_e32 v17, 16, v10
+; GFX9-NEXT: buffer_store_dword v17, off, s[0:3], s32 offset:76 ; 4-byte Folded Spill
; GFX9-NEXT: v_lshrrev_b64 v[17:18], 24, v[15:16]
; GFX9-NEXT: v_lshrrev_b64 v[18:19], 24, v[13:14]
; GFX9-NEXT: v_lshrrev_b64 v[19:20], 24, v[11:12]
; GFX9-NEXT: v_lshrrev_b64 v[20:21], 24, v[9:10]
-; GFX9-NEXT: v_lshrrev_b64 v[21:22], 24, v[7:8]
-; GFX9-NEXT: v_lshrrev_b64 v[22:23], 24, v[3:4]
-; GFX9-NEXT: v_lshrrev_b32_e32 v54, 24, v16
-; GFX9-NEXT: v_lshrrev_b32_e32 v55, 8, v16
-; GFX9-NEXT: v_lshrrev_b32_e32 v40, 16, v15
-; GFX9-NEXT: v_lshrrev_b32_e32 v41, 8, v15
-; GFX9-NEXT: v_lshrrev_b32_e32 v43, 24, v14
+; GFX9-NEXT: v_lshrrev_b64 v[25:26], 24, v[3:4]
+; GFX9-NEXT: v_lshrrev_b32_e32 v55, 24, v16
+; GFX9-NEXT: v_lshrrev_b32_e32 v40, 8, v16
+; GFX9-NEXT: v_lshrrev_b32_e32 v41, 16, v15
+; GFX9-NEXT: v_lshrrev_b32_e32 v42, 8, v15
+; GFX9-NEXT: v_lshrrev_b32_e32 v44, 24, v14
; GFX9-NEXT: v_lshrrev_b32_e32 v49, 8, v14
-; GFX9-NEXT: v_lshrrev_b32_e32 v44, 16, v13
+; GFX9-NEXT: v_lshrrev_b32_e32 v45, 16, v13
; GFX9-NEXT: v_lshrrev_b32_e32 v48, 8, v13
-; GFX9-NEXT: v_lshrrev_b32_e32 v46, 24, v12
-; GFX9-NEXT: v_lshrrev_b32_e32 v38, 8, v12
-; GFX9-NEXT: v_lshrrev_b32_e32 v47, 16, v11
-; GFX9-NEXT: v_lshrrev_b32_e32 v37, 8, v11
-; GFX9-NEXT: v_lshrrev_b32_e32 v56, 24, v10
-; GFX9-NEXT: v_lshrrev_b32_e32 v52, 16, v10
-; GFX9-NEXT: v_lshrrev_b32_e32 v36, 8, v10
-; GFX9-NEXT: v_lshrrev_b32_e32 v57, 16, v9
-; GFX9-NEXT: v_lshrrev_b32_e32 v35, 8, v9
-; GFX9-NEXT: v_lshrrev_b32_e32 v59, 24, v8
-; GFX9-NEXT: v_lshrrev_b32_e32 v53, 16, v8
-; GFX9-NEXT: v_lshrrev_b32_e32 v34, 8, v8
-; GFX9-NEXT: v_lshrrev_b32_e32 v60, 16, v7
-; GFX9-NEXT: v_lshrrev_b32_e32 v33, 8, v7
-; GFX9-NEXT: v_lshrrev_b32_e32 v61, 24, v6
-; GFX9-NEXT: v_lshrrev_b32_e32 v42, 16, v6
-; GFX9-NEXT: v_lshrrev_b32_e32 v32, 8, v6
-; GFX9-NEXT: v_lshrrev_b32_e32 v62, 16, v5
-; GFX9-NEXT: v_lshrrev_b32_e32 v31, 8, v5
-; GFX9-NEXT: v_lshrrev_b32_e32 v63, 24, v4
-; GFX9-NEXT: v_lshrrev_b32_e32 v45, 16, v4
-; GFX9-NEXT: v_lshrrev_b32_e32 v30, 8, v4
-; GFX9-NEXT: v_lshrrev_b32_e32 v39, 16, v3
-; GFX9-NEXT: v_lshrrev_b32_e32 v29, 8, v3
-; GFX9-NEXT: v_lshrrev_b32_e32 v50, 24, v2
-; GFX9-NEXT: v_lshrrev_b32_e32 v58, 16, v2
-; GFX9-NEXT: v_lshrrev_b32_e32 v51, 8, v2
-; GFX9-NEXT: v_lshrrev_b32_e32 v28, 16, v1
-; GFX9-NEXT: v_lshrrev_b32_e32 v27, 8, v1
-; GFX9-NEXT: v_lshrrev_b64 v[25:26], 24, v[5:6]
-; GFX9-NEXT: v_lshrrev_b64 v[23:24], 24, v[1:2]
+; GFX9-NEXT: v_lshrrev_b32_e32 v47, 24, v12
+; GFX9-NEXT: v_lshrrev_b32_e32 v39, 8, v12
+; GFX9-NEXT: v_lshrrev_b32_e32 v56, 16, v11
+; GFX9-NEXT: v_lshrrev_b32_e32 v38, 8, v11
+; GFX9-NEXT: v_lshrrev_b32_e32 v57, 24, v10
+; GFX9-NEXT: v_lshrrev_b32_e32 v37, 8, v10
+; GFX9-NEXT: v_lshrrev_b32_e32 v58, 16, v9
+; GFX9-NEXT: v_lshrrev_b32_e32 v36, 8, v9
+; GFX9-NEXT: v_lshrrev_b32_e32 v60, 24, v8
+; GFX9-NEXT: v_lshrrev_b32_e32 v54, 16, v8
+; GFX9-NEXT: v_lshrrev_b32_e32 v35, 8, v8
+; GFX9-NEXT: v_lshrrev_b32_e32 v61, 16, v7
+; GFX9-NEXT: v_lshrrev_b32_e32 v34, 8, v7
+; GFX9-NEXT: v_lshrrev_b32_e32 v62, 24, v6
+; GFX9-NEXT: v_lshrrev_b32_e32 v43, 16, v6
+; GFX9-NEXT: v_lshrrev_b32_e32 v33, 8, v6
+; GFX9-NEXT: v_lshrrev_b32_e32 v63, 16, v5
+; GFX9-NEXT: v_lshrrev_b32_e32 v32, 8, v5
+; GFX9-NEXT: v_lshrrev_b32_e32 v50, 24, v4
+; GFX9-NEXT: v_lshrrev_b32_e32 v46, 16, v4
+; GFX9-NEXT: v_lshrrev_b32_e32 v31, 8, v4
+; GFX9-NEXT: v_lshrrev_b32_e32 v51, 16, v3
+; GFX9-NEXT: v_lshrrev_b32_e32 v30, 8, v3
+; GFX9-NEXT: v_lshrrev_b32_e32 v52, 24, v2
+; GFX9-NEXT: v_lshrrev_b32_e32 v59, 16, v2
+; GFX9-NEXT: v_lshrrev_b32_e32 v53, 8, v2
+; GFX9-NEXT: v_lshrrev_b32_e32 v24, 16, v1
+; GFX9-NEXT: v_lshrrev_b32_e32 v23, 8, v1
+; GFX9-NEXT: v_lshrrev_b64 v[21:22], 24, v[7:8]
+; GFX9-NEXT: v_lshrrev_b64 v[28:29], 24, v[5:6]
+; GFX9-NEXT: v_lshrrev_b64 v[26:27], 24, v[1:2]
; GFX9-NEXT: .LBB108_2: ; %Flow
; GFX9-NEXT: s_andn2_saveexec_b64 s[4:5], s[4:5]
; GFX9-NEXT: s_cbranch_execz .LBB108_4
@@ -88160,7 +88072,7 @@ define <64 x i8> @bitcast_v32bf16_to_v64i8(<32 x bfloat> %a, i32 %b) {
; GFX9-NEXT: v_cmp_u_f32_e32 vcc, v1, v1
; GFX9-NEXT: s_mov_b32 s7, 0x7060302
; GFX9-NEXT: v_cndmask_b32_e32 v1, v19, v20, vcc
-; GFX9-NEXT: v_perm_b32 v27, v1, v18, s7
+; GFX9-NEXT: v_perm_b32 v23, v1, v18, s7
; GFX9-NEXT: v_lshlrev_b32_e32 v1, 16, v4
; GFX9-NEXT: v_add_f32_e32 v1, 0x40c00000, v1
; GFX9-NEXT: v_bfe_u32 v19, v1, 16, 1
@@ -88189,7 +88101,7 @@ define <64 x i8> @bitcast_v32bf16_to_v64i8(<32 x bfloat> %a, i32 %b) {
; GFX9-NEXT: v_or_b32_e32 v21, 0x400000, v3
; GFX9-NEXT: v_cmp_u_f32_e32 vcc, v3, v3
; GFX9-NEXT: v_cndmask_b32_e32 v3, v20, v21, vcc
-; GFX9-NEXT: v_perm_b32 v29, v3, v19, s7
+; GFX9-NEXT: v_perm_b32 v30, v3, v19, s7
; GFX9-NEXT: v_lshlrev_b32_e32 v3, 16, v6
; GFX9-NEXT: v_add_f32_e32 v3, 0x40c00000, v3
; GFX9-NEXT: v_bfe_u32 v20, v3, 16, 1
@@ -88218,7 +88130,7 @@ define <64 x i8> @bitcast_v32bf16_to_v64i8(<32 x bfloat> %a, i32 %b) {
; GFX9-NEXT: v_or_b32_e32 v22, 0x400000, v5
; GFX9-NEXT: v_cmp_u_f32_e32 vcc, v5, v5
; GFX9-NEXT: v_cndmask_b32_e32 v5, v21, v22, vcc
-; GFX9-NEXT: v_perm_b32 v31, v5, v20, s7
+; GFX9-NEXT: v_perm_b32 v32, v5, v20, s7
; GFX9-NEXT: v_lshlrev_b32_e32 v5, 16, v8
; GFX9-NEXT: v_add_f32_e32 v5, 0x40c00000, v5
; GFX9-NEXT: v_bfe_u32 v21, v5, 16, 1
@@ -88238,278 +88150,281 @@ define <64 x i8> @bitcast_v32bf16_to_v64i8(<32 x bfloat> %a, i32 %b) {
; GFX9-NEXT: v_bfe_u32 v22, v21, 16, 1
; GFX9-NEXT: v_and_b32_e32 v7, 0xffff0000, v7
; GFX9-NEXT: v_add3_u32 v22, v22, v21, s6
-; GFX9-NEXT: v_or_b32_e32 v23, 0x400000, v21
+; GFX9-NEXT: v_or_b32_e32 v25, 0x400000, v21
; GFX9-NEXT: v_cmp_u_f32_e32 vcc, v21, v21
; GFX9-NEXT: v_add_f32_e32 v7, 0x40c00000, v7
-; GFX9-NEXT: v_cndmask_b32_e32 v21, v22, v23, vcc
+; GFX9-NEXT: v_cndmask_b32_e32 v21, v22, v25, vcc
; GFX9-NEXT: v_bfe_u32 v22, v7, 16, 1
; GFX9-NEXT: v_add3_u32 v22, v22, v7, s6
-; GFX9-NEXT: v_or_b32_e32 v23, 0x400000, v7
+; GFX9-NEXT: v_or_b32_e32 v25, 0x400000, v7
; GFX9-NEXT: v_cmp_u_f32_e32 vcc, v7, v7
-; GFX9-NEXT: v_cndmask_b32_e32 v7, v22, v23, vcc
-; GFX9-NEXT: v_perm_b32 v33, v7, v21, s7
+; GFX9-NEXT: v_cndmask_b32_e32 v7, v22, v25, vcc
+; GFX9-NEXT: v_perm_b32 v34, v7, v21, s7
; GFX9-NEXT: v_lshlrev_b32_e32 v7, 16, v10
; GFX9-NEXT: v_add_f32_e32 v7, 0x40c00000, v7
; GFX9-NEXT: v_bfe_u32 v22, v7, 16, 1
; GFX9-NEXT: v_and_b32_e32 v10, 0xffff0000, v10
; GFX9-NEXT: v_add3_u32 v22, v22, v7, s6
-; GFX9-NEXT: v_or_b32_e32 v23, 0x400000, v7
+; GFX9-NEXT: v_or_b32_e32 v25, 0x400000, v7
; GFX9-NEXT: v_cmp_u_f32_e32 vcc, v7, v7
; GFX9-NEXT: v_add_f32_e32 v10, 0x40c00000, v10
-; GFX9-NEXT: v_cndmask_b32_e32 v7, v22, v23, vcc
+; GFX9-NEXT: v_cndmask_b32_e32 v7, v22, v25, vcc
; GFX9-NEXT: v_bfe_u32 v22, v10, 16, 1
; GFX9-NEXT: v_add3_u32 v22, v22, v10, s6
-; GFX9-NEXT: v_or_b32_e32 v23, 0x400000, v10
+; GFX9-NEXT: v_or_b32_e32 v25, 0x400000, v10
; GFX9-NEXT: v_cmp_u_f32_e32 vcc, v10, v10
-; GFX9-NEXT: v_cndmask_b32_e32 v10, v22, v23, vcc
+; GFX9-NEXT: v_cndmask_b32_e32 v10, v22, v25, vcc
; GFX9-NEXT: v_lshlrev_b32_e32 v22, 16, v9
; GFX9-NEXT: v_add_f32_e32 v22, 0x40c00000, v22
-; GFX9-NEXT: v_bfe_u32 v23, v22, 16, 1
+; GFX9-NEXT: v_bfe_u32 v25, v22, 16, 1
; GFX9-NEXT: v_and_b32_e32 v9, 0xffff0000, v9
-; GFX9-NEXT: v_add3_u32 v23, v23, v22, s6
-; GFX9-NEXT: v_or_b32_e32 v24, 0x400000, v22
+; GFX9-NEXT: v_add3_u32 v25, v25, v22, s6
+; GFX9-NEXT: v_or_b32_e32 v26, 0x400000, v22
; GFX9-NEXT: v_cmp_u_f32_e32 vcc, v22, v22
; GFX9-NEXT: v_add_f32_e32 v9, 0x40c00000, v9
-; GFX9-NEXT: v_cndmask_b32_e32 v22, v23, v24, vcc
-; GFX9-NEXT: v_bfe_u32 v23, v9, 16, 1
-; GFX9-NEXT: v_add3_u32 v23, v23, v9, s6
-; GFX9-NEXT: v_or_b32_e32 v24, 0x400000, v9
+; GFX9-NEXT: v_cndmask_b32_e32 v22, v25, v26, vcc
+; GFX9-NEXT: v_bfe_u32 v25, v9, 16, 1
+; GFX9-NEXT: v_add3_u32 v25, v25, v9, s6
+; GFX9-NEXT: v_or_b32_e32 v26, 0x400000, v9
; GFX9-NEXT: v_cmp_u_f32_e32 vcc, v9, v9
-; GFX9-NEXT: v_cndmask_b32_e32 v9, v23, v24, vcc
-; GFX9-NEXT: v_perm_b32 v35, v9, v22, s7
+; GFX9-NEXT: v_cndmask_b32_e32 v9, v25, v26, vcc
+; GFX9-NEXT: v_perm_b32 v36, v9, v22, s7
; GFX9-NEXT: v_lshlrev_b32_e32 v9, 16, v12
; GFX9-NEXT: v_add_f32_e32 v9, 0x40c00000, v9
-; GFX9-NEXT: v_bfe_u32 v23, v9, 16, 1
+; GFX9-NEXT: v_bfe_u32 v25, v9, 16, 1
; GFX9-NEXT: v_and_b32_e32 v12, 0xffff0000, v12
-; GFX9-NEXT: v_add3_u32 v23, v23, v9, s6
-; GFX9-NEXT: v_or_b32_e32 v24, 0x400000, v9
+; GFX9-NEXT: v_add3_u32 v25, v25, v9, s6
+; GFX9-NEXT: v_or_b32_e32 v26, 0x400000, v9
; GFX9-NEXT: v_cmp_u_f32_e32 vcc, v9, v9
; GFX9-NEXT: v_add_f32_e32 v12, 0x40c00000, v12
-; GFX9-NEXT: v_cndmask_b32_e32 v9, v23, v24, vcc
-; GFX9-NEXT: v_bfe_u32 v23, v12, 16, 1
-; GFX9-NEXT: v_add3_u32 v23, v23, v12, s6
-; GFX9-NEXT: v_or_b32_e32 v24, 0x400000, v12
+; GFX9-NEXT: v_cndmask_b32_e32 v9, v25, v26, vcc
+; GFX9-NEXT: v_bfe_u32 v25, v12, 16, 1
+; GFX9-NEXT: v_add3_u32 v25, v25, v12, s6
+; GFX9-NEXT: v_or_b32_e32 v26, 0x400000, v12
; GFX9-NEXT: v_cmp_u_f32_e32 vcc, v12, v12
-; GFX9-NEXT: v_cndmask_b32_e32 v12, v23, v24, vcc
-; GFX9-NEXT: v_lshlrev_b32_e32 v23, 16, v11
-; GFX9-NEXT: v_add_f32_e32 v23, 0x40c00000, v23
-; GFX9-NEXT: v_bfe_u32 v24, v23, 16, 1
+; GFX9-NEXT: v_cndmask_b32_e32 v12, v25, v26, vcc
+; GFX9-NEXT: v_lshlrev_b32_e32 v25, 16, v11
+; GFX9-NEXT: v_add_f32_e32 v25, 0x40c00000, v25
+; GFX9-NEXT: v_bfe_u32 v26, v25, 16, 1
; GFX9-NEXT: v_and_b32_e32 v11, 0xffff0000, v11
-; GFX9-NEXT: v_add3_u32 v24, v24, v23, s6
-; GFX9-NEXT: v_or_b32_e32 v25, 0x400000, v23
-; GFX9-NEXT: v_cmp_u_f32_e32 vcc, v23, v23
+; GFX9-NEXT: v_add3_u32 v26, v26, v25, s6
+; GFX9-NEXT: v_or_b32_e32 v27, 0x400000, v25
+; GFX9-NEXT: v_cmp_u_f32_e32 vcc, v25, v25
; GFX9-NEXT: v_add_f32_e32 v11, 0x40c00000, v11
-; GFX9-NEXT: v_cndmask_b32_e32 v23, v24, v25, vcc
-; GFX9-NEXT: v_bfe_u32 v24, v11, 16, 1
-; GFX9-NEXT: v_add3_u32 v24, v24, v11, s6
-; GFX9-NEXT: v_or_b32_e32 v25, 0x400000, v11
+; GFX9-NEXT: v_cndmask_b32_e32 v25, v26, v27, vcc
+; GFX9-NEXT: v_bfe_u32 v26, v11, 16, 1
+; GFX9-NEXT: v_add3_u32 v26, v26, v11, s6
+; GFX9-NEXT: v_or_b32_e32 v27, 0x400000, v11
; GFX9-NEXT: v_cmp_u_f32_e32 vcc, v11, v11
-; GFX9-NEXT: v_cndmask_b32_e32 v11, v24, v25, vcc
-; GFX9-NEXT: v_perm_b32 v37, v11, v23, s7
+; GFX9-NEXT: v_cndmask_b32_e32 v11, v26, v27, vcc
+; GFX9-NEXT: v_perm_b32 v38, v11, v25, s7
; GFX9-NEXT: v_lshlrev_b32_e32 v11, 16, v14
; GFX9-NEXT: v_add_f32_e32 v11, 0x40c00000, v11
-; GFX9-NEXT: v_bfe_u32 v24, v11, 16, 1
+; GFX9-NEXT: v_bfe_u32 v26, v11, 16, 1
; GFX9-NEXT: v_and_b32_e32 v14, 0xffff0000, v14
-; GFX9-NEXT: v_add3_u32 v24, v24, v11, s6
-; GFX9-NEXT: v_or_b32_e32 v25, 0x400000, v11
+; GFX9-NEXT: v_add3_u32 v26, v26, v11, s6
+; GFX9-NEXT: v_or_b32_e32 v27, 0x400000, v11
; GFX9-NEXT: v_cmp_u_f32_e32 vcc, v11, v11
; GFX9-NEXT: v_add_f32_e32 v14, 0x40c00000, v14
-; GFX9-NEXT: v_cndmask_b32_e32 v11, v24, v25, vcc
-; GFX9-NEXT: v_bfe_u32 v24, v14, 16, 1
-; GFX9-NEXT: v_add3_u32 v24, v24, v14, s6
-; GFX9-NEXT: v_or_b32_e32 v25, 0x400000, v14
+; GFX9-NEXT: v_cndmask_b32_e32 v11, v26, v27, vcc
+; GFX9-NEXT: v_bfe_u32 v26, v14, 16, 1
+; GFX9-NEXT: v_add3_u32 v26, v26, v14, s6
+; GFX9-NEXT: v_or_b32_e32 v27, 0x400000, v14
; GFX9-NEXT: v_cmp_u_f32_e32 vcc, v14, v14
-; GFX9-NEXT: v_cndmask_b32_e32 v14, v24, v25, vcc
-; GFX9-NEXT: v_lshlrev_b32_e32 v24, 16, v13
-; GFX9-NEXT: v_add_f32_e32 v24, 0x40c00000, v24
-; GFX9-NEXT: v_bfe_u32 v25, v24, 16, 1
+; GFX9-NEXT: v_cndmask_b32_e32 v14, v26, v27, vcc
+; GFX9-NEXT: v_lshlrev_b32_e32 v26, 16, v13
+; GFX9-NEXT: v_add_f32_e32 v26, 0x40c00000, v26
+; GFX9-NEXT: v_bfe_u32 v27, v26, 16, 1
; GFX9-NEXT: v_and_b32_e32 v13, 0xffff0000, v13
-; GFX9-NEXT: v_add3_u32 v25, v25, v24, s6
-; GFX9-NEXT: v_or_b32_e32 v26, 0x400000, v24
-; GFX9-NEXT: v_cmp_u_f32_e32 vcc, v24, v24
+; GFX9-NEXT: v_add3_u32 v27, v27, v26, s6
+; GFX9-NEXT: v_or_b32_e32 v28, 0x400000, v26
+; GFX9-NEXT: v_cmp_u_f32_e32 vcc, v26, v26
; GFX9-NEXT: v_add_f32_e32 v13, 0x40c00000, v13
-; GFX9-NEXT: v_cndmask_b32_e32 v24, v25, v26, vcc
-; GFX9-NEXT: v_bfe_u32 v25, v13, 16, 1
-; GFX9-NEXT: v_add3_u32 v25, v25, v13, s6
-; GFX9-NEXT: v_or_b32_e32 v26, 0x400000, v13
+; GFX9-NEXT: v_cndmask_b32_e32 v26, v27, v28, vcc
+; GFX9-NEXT: v_bfe_u32 v27, v13, 16, 1
+; GFX9-NEXT: v_add3_u32 v27, v27, v13, s6
+; GFX9-NEXT: v_or_b32_e32 v28, 0x400000, v13
; GFX9-NEXT: v_cmp_u_f32_e32 vcc, v13, v13
-; GFX9-NEXT: v_cndmask_b32_e32 v13, v25, v26, vcc
-; GFX9-NEXT: v_perm_b32 v48, v13, v24, s7
+; GFX9-NEXT: v_cndmask_b32_e32 v13, v27, v28, vcc
+; GFX9-NEXT: v_perm_b32 v48, v13, v26, s7
; GFX9-NEXT: v_lshlrev_b32_e32 v13, 16, v16
; GFX9-NEXT: v_add_f32_e32 v13, 0x40c00000, v13
-; GFX9-NEXT: v_bfe_u32 v25, v13, 16, 1
+; GFX9-NEXT: v_bfe_u32 v27, v13, 16, 1
; GFX9-NEXT: v_and_b32_e32 v16, 0xffff0000, v16
-; GFX9-NEXT: v_add3_u32 v25, v25, v13, s6
-; GFX9-NEXT: v_or_b32_e32 v26, 0x400000, v13
+; GFX9-NEXT: v_add3_u32 v27, v27, v13, s6
+; GFX9-NEXT: v_or_b32_e32 v28, 0x400000, v13
; GFX9-NEXT: v_cmp_u_f32_e32 vcc, v13, v13
; GFX9-NEXT: v_add_f32_e32 v16, 0x40c00000, v16
-; GFX9-NEXT: v_cndmask_b32_e32 v13, v25, v26, vcc
-; GFX9-NEXT: v_bfe_u32 v25, v16, 16, 1
-; GFX9-NEXT: v_add3_u32 v25, v25, v16, s6
-; GFX9-NEXT: v_or_b32_e32 v26, 0x400000, v16
+; GFX9-NEXT: v_cndmask_b32_e32 v13, v27, v28, vcc
+; GFX9-NEXT: v_bfe_u32 v27, v16, 16, 1
+; GFX9-NEXT: v_add3_u32 v27, v27, v16, s6
+; GFX9-NEXT: v_or_b32_e32 v28, 0x400000, v16
; GFX9-NEXT: v_cmp_u_f32_e32 vcc, v16, v16
-; GFX9-NEXT: v_cndmask_b32_e32 v16, v25, v26, vcc
-; GFX9-NEXT: v_lshlrev_b32_e32 v25, 16, v15
-; GFX9-NEXT: v_add_f32_e32 v25, 0x40c00000, v25
-; GFX9-NEXT: v_bfe_u32 v26, v25, 16, 1
+; GFX9-NEXT: v_cndmask_b32_e32 v16, v27, v28, vcc
+; GFX9-NEXT: v_lshlrev_b32_e32 v27, 16, v15
+; GFX9-NEXT: v_add_f32_e32 v27, 0x40c00000, v27
+; GFX9-NEXT: v_bfe_u32 v28, v27, 16, 1
; GFX9-NEXT: v_and_b32_e32 v15, 0xffff0000, v15
-; GFX9-NEXT: v_add3_u32 v26, v26, v25, s6
-; GFX9-NEXT: v_or_b32_e32 v39, 0x400000, v25
-; GFX9-NEXT: v_cmp_u_f32_e32 vcc, v25, v25
+; GFX9-NEXT: v_add3_u32 v28, v28, v27, s6
+; GFX9-NEXT: v_or_b32_e32 v29, 0x400000, v27
+; GFX9-NEXT: v_cmp_u_f32_e32 vcc, v27, v27
; GFX9-NEXT: v_add_f32_e32 v15, 0x40c00000, v15
-; GFX9-NEXT: v_cndmask_b32_e32 v25, v26, v39, vcc
-; GFX9-NEXT: v_bfe_u32 v26, v15, 16, 1
-; GFX9-NEXT: v_add3_u32 v26, v26, v15, s6
-; GFX9-NEXT: v_or_b32_e32 v39, 0x400000, v15
+; GFX9-NEXT: v_cndmask_b32_e32 v27, v28, v29, vcc
+; GFX9-NEXT: v_bfe_u32 v28, v15, 16, 1
+; GFX9-NEXT: v_add3_u32 v28, v28, v15, s6
+; GFX9-NEXT: v_or_b32_e32 v29, 0x400000, v15
; GFX9-NEXT: v_cmp_u_f32_e32 vcc, v15, v15
-; GFX9-NEXT: v_cndmask_b32_e32 v15, v26, v39, vcc
+; GFX9-NEXT: v_cndmask_b32_e32 v15, v28, v29, vcc
; GFX9-NEXT: v_perm_b32 v51, v16, v13, s7
-; GFX9-NEXT: v_perm_b32 v50, v15, v25, s7
-; GFX9-NEXT: v_perm_b32 v28, v2, v17, s7
-; GFX9-NEXT: v_perm_b32 v30, v4, v1, s7
+; GFX9-NEXT: v_perm_b32 v50, v15, v27, s7
+; GFX9-NEXT: v_perm_b32 v24, v2, v17, s7
+; GFX9-NEXT: v_perm_b32 v31, v4, v1, s7
; GFX9-NEXT: v_perm_b32 v49, v14, v11, s7
-; GFX9-NEXT: v_lshrrev_b32_e32 v45, 16, v4
+; GFX9-NEXT: v_lshrrev_b32_e32 v46, 16, v4
; GFX9-NEXT: v_lshrrev_b32_e32 v4, 16, v1
-; GFX9-NEXT: v_lshrrev_b32_e32 v58, 16, v2
+; GFX9-NEXT: v_lshrrev_b32_e32 v59, 16, v2
; GFX9-NEXT: v_lshrrev_b32_e32 v2, 16, v17
; GFX9-NEXT: v_lshrrev_b32_e32 v1, 16, v18
; GFX9-NEXT: v_lshrrev_b64 v[17:18], 24, v[50:51]
-; GFX9-NEXT: v_perm_b32 v32, v6, v3, s7
-; GFX9-NEXT: v_perm_b32 v38, v12, v9, s7
-; GFX9-NEXT: v_lshrrev_b32_e32 v42, 16, v6
+; GFX9-NEXT: v_perm_b32 v33, v6, v3, s7
+; GFX9-NEXT: v_perm_b32 v39, v12, v9, s7
+; GFX9-NEXT: v_lshrrev_b32_e32 v43, 16, v6
; GFX9-NEXT: v_lshrrev_b32_e32 v6, 16, v3
; GFX9-NEXT: v_lshrrev_b32_e32 v3, 16, v19
; GFX9-NEXT: v_lshrrev_b64 v[18:19], 24, v[48:49]
-; GFX9-NEXT: v_perm_b32 v34, v8, v5, s7
-; GFX9-NEXT: v_perm_b32 v36, v10, v7, s7
-; GFX9-NEXT: v_lshrrev_b32_e32 v53, 16, v8
-; GFX9-NEXT: v_lshrrev_b32_e32 v8, 16, v5
-; GFX9-NEXT: v_lshrrev_b32_e32 v5, 16, v20
-; GFX9-NEXT: v_lshrrev_b64 v[19:20], 24, v[37:38]
-; GFX9-NEXT: v_lshrrev_b32_e32 v52, 16, v10
-; GFX9-NEXT: v_lshrrev_b32_e32 v10, 16, v7
-; GFX9-NEXT: v_lshrrev_b32_e32 v7, 16, v21
-; GFX9-NEXT: v_lshrrev_b64 v[20:21], 24, v[35:36]
+; GFX9-NEXT: v_perm_b32 v35, v8, v5, s7
+; GFX9-NEXT: v_perm_b32 v37, v10, v7, s7
; GFX9-NEXT: v_lshrrev_b32_e32 v15, 16, v16
; GFX9-NEXT: v_lshrrev_b32_e32 v16, 16, v13
; GFX9-NEXT: v_lshrrev_b32_e32 v13, 16, v14
; GFX9-NEXT: v_lshrrev_b32_e32 v14, 16, v11
; GFX9-NEXT: v_lshrrev_b32_e32 v11, 16, v12
-; GFX9-NEXT: v_lshrrev_b32_e32 v12, 16, v9
-; GFX9-NEXT: v_lshrrev_b32_e32 v9, 16, v22
-; GFX9-NEXT: v_lshrrev_b64 v[21:22], 24, v[33:34]
+; GFX9-NEXT: v_lshrrev_b32_e32 v54, 16, v8
+; GFX9-NEXT: v_lshrrev_b32_e32 v8, 16, v5
+; GFX9-NEXT: v_lshrrev_b32_e32 v5, 16, v20
+; GFX9-NEXT: v_lshrrev_b64 v[19:20], 24, v[38:39]
+; GFX9-NEXT: buffer_store_dword v13, off, s[0:3], s32 offset:68 ; 4-byte Folded Spill
+; GFX9-NEXT: v_lshrrev_b32_e32 v13, 16, v26
; GFX9-NEXT: buffer_store_dword v11, off, s[0:3], s32 offset:72 ; 4-byte Folded Spill
-; GFX9-NEXT: v_lshrrev_b32_e32 v11, 16, v23
-; GFX9-NEXT: v_lshrrev_b64 v[22:23], 24, v[29:30]
+; GFX9-NEXT: v_lshrrev_b32_e32 v12, 16, v9
+; GFX9-NEXT: v_lshrrev_b32_e32 v11, 16, v25
+; GFX9-NEXT: v_lshrrev_b32_e32 v9, 16, v10
+; GFX9-NEXT: v_lshrrev_b32_e32 v10, 16, v7
+; GFX9-NEXT: v_lshrrev_b32_e32 v7, 16, v21
+; GFX9-NEXT: v_lshrrev_b64 v[20:21], 24, v[36:37]
+; GFX9-NEXT: v_lshrrev_b64 v[25:26], 24, v[30:31]
; GFX9-NEXT: buffer_store_dword v15, off, s[0:3], s32 offset:64 ; 4-byte Folded Spill
-; GFX9-NEXT: v_lshrrev_b32_e32 v15, 16, v25
-; GFX9-NEXT: buffer_store_dword v13, off, s[0:3], s32 offset:68 ; 4-byte Folded Spill
-; GFX9-NEXT: v_lshrrev_b32_e32 v13, 16, v24
-; GFX9-NEXT: v_lshrrev_b64 v[25:26], 24, v[31:32]
-; GFX9-NEXT: v_lshrrev_b64 v[23:24], 24, v[27:28]
-; GFX9-NEXT: v_lshrrev_b32_e32 v54, 24, v51
-; GFX9-NEXT: v_lshrrev_b32_e32 v55, 8, v51
-; GFX9-NEXT: v_lshrrev_b32_e32 v40, 16, v50
-; GFX9-NEXT: v_lshrrev_b32_e32 v41, 8, v50
-; GFX9-NEXT: v_lshrrev_b32_e32 v43, 24, v49
+; GFX9-NEXT: v_lshrrev_b32_e32 v15, 16, v27
+; GFX9-NEXT: buffer_store_dword v9, off, s[0:3], s32 offset:76 ; 4-byte Folded Spill
+; GFX9-NEXT: v_lshrrev_b32_e32 v9, 16, v22
+; GFX9-NEXT: v_lshrrev_b64 v[21:22], 24, v[34:35]
+; GFX9-NEXT: v_lshrrev_b64 v[28:29], 24, v[32:33]
+; GFX9-NEXT: v_lshrrev_b64 v[26:27], 24, v[23:24]
+; GFX9-NEXT: v_lshrrev_b32_e32 v55, 24, v51
+; GFX9-NEXT: v_lshrrev_b32_e32 v40, 8, v51
+; GFX9-NEXT: v_lshrrev_b32_e32 v41, 16, v50
+; GFX9-NEXT: v_lshrrev_b32_e32 v42, 8, v50
+; GFX9-NEXT: v_lshrrev_b32_e32 v44, 24, v49
; GFX9-NEXT: v_lshrrev_b32_e32 v49, 8, v49
-; GFX9-NEXT: v_lshrrev_b32_e32 v44, 16, v48
+; GFX9-NEXT: v_lshrrev_b32_e32 v45, 16, v48
; GFX9-NEXT: v_lshrrev_b32_e32 v48, 8, v48
-; GFX9-NEXT: v_lshrrev_b32_e32 v46, 24, v38
+; GFX9-NEXT: v_lshrrev_b32_e32 v47, 24, v39
+; GFX9-NEXT: v_lshrrev_b32_e32 v39, 8, v39
+; GFX9-NEXT: v_lshrrev_b32_e32 v56, 16, v38
; GFX9-NEXT: v_lshrrev_b32_e32 v38, 8, v38
-; GFX9-NEXT: v_lshrrev_b32_e32 v47, 16, v37
+; GFX9-NEXT: v_lshrrev_b32_e32 v57, 24, v37
; GFX9-NEXT: v_lshrrev_b32_e32 v37, 8, v37
-; GFX9-NEXT: v_lshrrev_b32_e32 v56, 24, v36
+; GFX9-NEXT: v_lshrrev_b32_e32 v58, 16, v36
; GFX9-NEXT: v_lshrrev_b32_e32 v36, 8, v36
-; GFX9-NEXT: v_lshrrev_b32_e32 v57, 16, v35
+; GFX9-NEXT: v_lshrrev_b32_e32 v60, 24, v35
; GFX9-NEXT: v_lshrrev_b32_e32 v35, 8, v35
-; GFX9-NEXT: v_lshrrev_b32_e32 v59, 24, v34
+; GFX9-NEXT: v_lshrrev_b32_e32 v61, 16, v34
; GFX9-NEXT: v_lshrrev_b32_e32 v34, 8, v34
-; GFX9-NEXT: v_lshrrev_b32_e32 v60, 16, v33
+; GFX9-NEXT: v_lshrrev_b32_e32 v62, 24, v33
; GFX9-NEXT: v_lshrrev_b32_e32 v33, 8, v33
-; GFX9-NEXT: v_lshrrev_b32_e32 v61, 24, v32
+; GFX9-NEXT: v_lshrrev_b32_e32 v63, 16, v32
; GFX9-NEXT: v_lshrrev_b32_e32 v32, 8, v32
-; GFX9-NEXT: v_lshrrev_b32_e32 v62, 16, v31
+; GFX9-NEXT: v_lshrrev_b32_e32 v50, 24, v31
; GFX9-NEXT: v_lshrrev_b32_e32 v31, 8, v31
-; GFX9-NEXT: v_lshrrev_b32_e32 v63, 24, v30
+; GFX9-NEXT: v_lshrrev_b32_e32 v51, 16, v30
; GFX9-NEXT: v_lshrrev_b32_e32 v30, 8, v30
-; GFX9-NEXT: v_lshrrev_b32_e32 v39, 16, v29
-; GFX9-NEXT: v_lshrrev_b32_e32 v29, 8, v29
-; GFX9-NEXT: v_lshrrev_b32_e32 v50, 24, v28
-; GFX9-NEXT: v_lshrrev_b32_e32 v51, 8, v28
-; GFX9-NEXT: v_lshrrev_b32_e32 v28, 16, v27
-; GFX9-NEXT: v_lshrrev_b32_e32 v27, 8, v27
+; GFX9-NEXT: v_lshrrev_b32_e32 v52, 24, v24
+; GFX9-NEXT: v_lshrrev_b32_e32 v53, 8, v24
+; GFX9-NEXT: v_lshrrev_b32_e32 v24, 16, v23
+; GFX9-NEXT: v_lshrrev_b32_e32 v23, 8, v23
; GFX9-NEXT: .LBB108_4: ; %end
; GFX9-NEXT: s_or_b64 exec, exec, s[4:5]
-; GFX9-NEXT: v_lshlrev_b16_e32 v24, 8, v27
-; GFX9-NEXT: v_lshlrev_b16_e32 v23, 8, v23
-; GFX9-NEXT: v_or_b32_sdwa v1, v1, v24 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
-; GFX9-NEXT: v_or_b32_sdwa v23, v28, v23 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
-; GFX9-NEXT: v_or_b32_sdwa v1, v1, v23 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_0 src1_sel:DWORD
-; GFX9-NEXT: v_lshlrev_b16_e32 v24, 8, v51
+; GFX9-NEXT: v_lshlrev_b16_e32 v22, 8, v23
+; GFX9-NEXT: v_or_b32_sdwa v1, v1, v22 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; GFX9-NEXT: v_lshlrev_b16_e32 v22, 8, v53
+; GFX9-NEXT: v_or_b32_sdwa v2, v2, v22 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; GFX9-NEXT: v_lshlrev_b16_e32 v22, 8, v26
+; GFX9-NEXT: v_or_b32_sdwa v22, v24, v22 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; GFX9-NEXT: v_or_b32_sdwa v1, v1, v22 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_0 src1_sel:DWORD
; GFX9-NEXT: buffer_store_dword v1, v0, s[0:3], 0 offen
-; GFX9-NEXT: v_lshlrev_b16_e32 v1, 8, v50
-; GFX9-NEXT: v_or_b32_sdwa v2, v2, v24 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
-; GFX9-NEXT: v_or_b32_sdwa v1, v58, v1 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; GFX9-NEXT: v_lshlrev_b16_e32 v1, 8, v52
+; GFX9-NEXT: v_or_b32_sdwa v1, v59, v1 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
; GFX9-NEXT: v_or_b32_sdwa v1, v2, v1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_0 src1_sel:DWORD
; GFX9-NEXT: buffer_store_dword v1, v0, s[0:3], 0 offen offset:4
-; GFX9-NEXT: v_lshlrev_b16_e32 v1, 8, v22
-; GFX9-NEXT: v_lshlrev_b16_e32 v2, 8, v29
-; GFX9-NEXT: v_or_b32_sdwa v1, v39, v1 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; GFX9-NEXT: v_lshlrev_b16_e32 v1, 8, v25
+; GFX9-NEXT: v_lshlrev_b16_e32 v2, 8, v30
+; GFX9-NEXT: v_or_b32_sdwa v1, v51, v1 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
; GFX9-NEXT: v_or_b32_sdwa v2, v3, v2 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
; GFX9-NEXT: v_or_b32_sdwa v1, v2, v1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_0 src1_sel:DWORD
; GFX9-NEXT: buffer_store_dword v1, v0, s[0:3], 0 offen offset:8
-; GFX9-NEXT: v_lshlrev_b16_e32 v1, 8, v30
-; GFX9-NEXT: v_lshlrev_b16_e32 v2, 8, v63
+; GFX9-NEXT: v_lshlrev_b16_e32 v1, 8, v31
+; GFX9-NEXT: v_lshlrev_b16_e32 v2, 8, v50
; GFX9-NEXT: v_or_b32_sdwa v1, v4, v1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
-; GFX9-NEXT: v_or_b32_sdwa v2, v45, v2 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; GFX9-NEXT: v_or_b32_sdwa v2, v46, v2 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
; GFX9-NEXT: v_or_b32_sdwa v1, v1, v2 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_0 src1_sel:DWORD
; GFX9-NEXT: buffer_store_dword v1, v0, s[0:3], 0 offen offset:12
-; GFX9-NEXT: v_lshlrev_b16_e32 v1, 8, v31
-; GFX9-NEXT: v_lshlrev_b16_e32 v2, 8, v25
+; GFX9-NEXT: v_lshlrev_b16_e32 v1, 8, v32
+; GFX9-NEXT: v_lshlrev_b16_e32 v2, 8, v28
; GFX9-NEXT: v_or_b32_sdwa v1, v5, v1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
-; GFX9-NEXT: v_or_b32_sdwa v2, v62, v2 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; GFX9-NEXT: v_or_b32_sdwa v2, v63, v2 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
; GFX9-NEXT: v_or_b32_sdwa v1, v1, v2 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_0 src1_sel:DWORD
; GFX9-NEXT: buffer_store_dword v1, v0, s[0:3], 0 offen offset:16
-; GFX9-NEXT: v_lshlrev_b16_e32 v1, 8, v32
-; GFX9-NEXT: v_lshlrev_b16_e32 v2, 8, v61
+; GFX9-NEXT: v_lshlrev_b16_e32 v1, 8, v33
+; GFX9-NEXT: v_lshlrev_b16_e32 v2, 8, v62
; GFX9-NEXT: v_or_b32_sdwa v1, v6, v1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
-; GFX9-NEXT: v_or_b32_sdwa v2, v42, v2 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; GFX9-NEXT: v_or_b32_sdwa v2, v43, v2 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
; GFX9-NEXT: v_or_b32_sdwa v1, v1, v2 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_0 src1_sel:DWORD
; GFX9-NEXT: buffer_store_dword v1, v0, s[0:3], 0 offen offset:20
-; GFX9-NEXT: v_lshlrev_b16_e32 v1, 8, v33
+; GFX9-NEXT: v_lshlrev_b16_e32 v1, 8, v34
; GFX9-NEXT: v_lshlrev_b16_e32 v2, 8, v21
; GFX9-NEXT: v_or_b32_sdwa v1, v7, v1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
-; GFX9-NEXT: v_or_b32_sdwa v2, v60, v2 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; GFX9-NEXT: v_or_b32_sdwa v2, v61, v2 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
; GFX9-NEXT: v_or_b32_sdwa v1, v1, v2 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_0 src1_sel:DWORD
; GFX9-NEXT: buffer_store_dword v1, v0, s[0:3], 0 offen offset:24
-; GFX9-NEXT: v_lshlrev_b16_e32 v1, 8, v34
-; GFX9-NEXT: v_lshlrev_b16_e32 v2, 8, v59
+; GFX9-NEXT: v_lshlrev_b16_e32 v1, 8, v35
+; GFX9-NEXT: v_lshlrev_b16_e32 v2, 8, v60
; GFX9-NEXT: v_or_b32_sdwa v1, v8, v1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
-; GFX9-NEXT: v_or_b32_sdwa v2, v53, v2 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; GFX9-NEXT: v_or_b32_sdwa v2, v54, v2 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
; GFX9-NEXT: v_or_b32_sdwa v1, v1, v2 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_0 src1_sel:DWORD
; GFX9-NEXT: buffer_store_dword v1, v0, s[0:3], 0 offen offset:28
-; GFX9-NEXT: v_lshlrev_b16_e32 v1, 8, v35
+; GFX9-NEXT: v_lshlrev_b16_e32 v1, 8, v36
; GFX9-NEXT: v_lshlrev_b16_e32 v2, 8, v20
; GFX9-NEXT: v_or_b32_sdwa v1, v9, v1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
-; GFX9-NEXT: v_or_b32_sdwa v2, v57, v2 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; GFX9-NEXT: v_or_b32_sdwa v2, v58, v2 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
; GFX9-NEXT: v_or_b32_sdwa v1, v1, v2 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_0 src1_sel:DWORD
; GFX9-NEXT: buffer_store_dword v1, v0, s[0:3], 0 offen offset:32
-; GFX9-NEXT: v_lshlrev_b16_e32 v1, 8, v36
-; GFX9-NEXT: v_lshlrev_b16_e32 v2, 8, v56
+; GFX9-NEXT: buffer_load_dword v3, off, s[0:3], s32 offset:76 ; 4-byte Folded Reload
+; GFX9-NEXT: v_lshlrev_b16_e32 v1, 8, v37
+; GFX9-NEXT: v_lshlrev_b16_e32 v2, 8, v57
; GFX9-NEXT: v_or_b32_sdwa v1, v10, v1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
-; GFX9-NEXT: v_or_b32_sdwa v2, v52, v2 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; GFX9-NEXT: s_waitcnt vmcnt(0)
+; GFX9-NEXT: v_or_b32_sdwa v2, v3, v2 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
; GFX9-NEXT: v_or_b32_sdwa v1, v1, v2 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_0 src1_sel:DWORD
; GFX9-NEXT: buffer_store_dword v1, v0, s[0:3], 0 offen offset:36
-; GFX9-NEXT: v_lshlrev_b16_e32 v1, 8, v37
+; GFX9-NEXT: v_lshlrev_b16_e32 v1, 8, v38
; GFX9-NEXT: v_lshlrev_b16_e32 v2, 8, v19
; GFX9-NEXT: v_or_b32_sdwa v1, v11, v1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
-; GFX9-NEXT: v_or_b32_sdwa v2, v47, v2 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; GFX9-NEXT: v_or_b32_sdwa v2, v56, v2 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
; GFX9-NEXT: v_or_b32_sdwa v1, v1, v2 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_0 src1_sel:DWORD
; GFX9-NEXT: buffer_store_dword v1, v0, s[0:3], 0 offen offset:40
; GFX9-NEXT: buffer_load_dword v3, off, s[0:3], s32 offset:72 ; 4-byte Folded Reload
-; GFX9-NEXT: v_lshlrev_b16_e32 v1, 8, v38
-; GFX9-NEXT: v_lshlrev_b16_e32 v2, 8, v46
+; GFX9-NEXT: v_lshlrev_b16_e32 v1, 8, v39
+; GFX9-NEXT: v_lshlrev_b16_e32 v2, 8, v47
; GFX9-NEXT: v_or_b32_sdwa v1, v12, v1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
; GFX9-NEXT: s_waitcnt vmcnt(0)
; GFX9-NEXT: v_or_b32_sdwa v2, v3, v2 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
@@ -88518,26 +88433,26 @@ define <64 x i8> @bitcast_v32bf16_to_v64i8(<32 x bfloat> %a, i32 %b) {
; GFX9-NEXT: v_lshlrev_b16_e32 v1, 8, v48
; GFX9-NEXT: v_lshlrev_b16_e32 v2, 8, v18
; GFX9-NEXT: v_or_b32_sdwa v1, v13, v1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
-; GFX9-NEXT: v_or_b32_sdwa v2, v44, v2 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; GFX9-NEXT: v_or_b32_sdwa v2, v45, v2 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
; GFX9-NEXT: v_or_b32_sdwa v1, v1, v2 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_0 src1_sel:DWORD
; GFX9-NEXT: buffer_store_dword v1, v0, s[0:3], 0 offen offset:48
; GFX9-NEXT: buffer_load_dword v3, off, s[0:3], s32 offset:68 ; 4-byte Folded Reload
; GFX9-NEXT: v_lshlrev_b16_e32 v1, 8, v49
-; GFX9-NEXT: v_lshlrev_b16_e32 v2, 8, v43
+; GFX9-NEXT: v_lshlrev_b16_e32 v2, 8, v44
; GFX9-NEXT: v_or_b32_sdwa v1, v14, v1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
; GFX9-NEXT: s_waitcnt vmcnt(0)
; GFX9-NEXT: v_or_b32_sdwa v2, v3, v2 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
; GFX9-NEXT: v_or_b32_sdwa v1, v1, v2 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_0 src1_sel:DWORD
; GFX9-NEXT: buffer_store_dword v1, v0, s[0:3], 0 offen offset:52
-; GFX9-NEXT: v_lshlrev_b16_e32 v1, 8, v41
+; GFX9-NEXT: v_lshlrev_b16_e32 v1, 8, v42
; GFX9-NEXT: v_lshlrev_b16_e32 v2, 8, v17
; GFX9-NEXT: v_or_b32_sdwa v1, v15, v1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
-; GFX9-NEXT: v_or_b32_sdwa v2, v40, v2 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; GFX9-NEXT: v_or_b32_sdwa v2, v41, v2 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
; GFX9-NEXT: v_or_b32_sdwa v1, v1, v2 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_0 src1_sel:DWORD
; GFX9-NEXT: buffer_store_dword v1, v0, s[0:3], 0 offen offset:56
; GFX9-NEXT: buffer_load_dword v3, off, s[0:3], s32 offset:64 ; 4-byte Folded Reload
-; GFX9-NEXT: v_lshlrev_b16_e32 v1, 8, v55
-; GFX9-NEXT: v_lshlrev_b16_e32 v2, 8, v54
+; GFX9-NEXT: v_lshlrev_b16_e32 v1, 8, v40
+; GFX9-NEXT: v_lshlrev_b16_e32 v2, 8, v55
; GFX9-NEXT: v_or_b32_sdwa v1, v16, v1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
; GFX9-NEXT: s_waitcnt vmcnt(0)
; GFX9-NEXT: v_or_b32_sdwa v2, v3, v2 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
@@ -88566,62 +88481,62 @@ define <64 x i8> @bitcast_v32bf16_to_v64i8(<32 x bfloat> %a, i32 %b) {
; GFX11-TRUE16: ; %bb.0:
; GFX11-TRUE16-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX11-TRUE16-NEXT: v_cmp_ne_u32_e32 vcc_lo, 0, v17
-; GFX11-TRUE16-NEXT: ; implicit-def: $vgpr28_hi16
+; GFX11-TRUE16-NEXT: ; implicit-def: $vgpr31_hi16
; GFX11-TRUE16-NEXT: ; implicit-def: $vgpr113_lo16
; GFX11-TRUE16-NEXT: ; implicit-def: $vgpr24_lo16
-; GFX11-TRUE16-NEXT: ; implicit-def: $vgpr26_hi16
+; GFX11-TRUE16-NEXT: ; implicit-def: $vgpr28_hi16
; GFX11-TRUE16-NEXT: ; implicit-def: $vgpr112_lo16
-; GFX11-TRUE16-NEXT: ; implicit-def: $vgpr27_hi16
+; GFX11-TRUE16-NEXT: ; implicit-def: $vgpr26_hi16
; GFX11-TRUE16-NEXT: ; implicit-def: $vgpr103_lo16
-; GFX11-TRUE16-NEXT: ; implicit-def: $vgpr31_hi16
+; GFX11-TRUE16-NEXT: ; implicit-def: $vgpr34_hi16
; GFX11-TRUE16-NEXT: ; implicit-def: $vgpr102_lo16
; GFX11-TRUE16-NEXT: ; implicit-def: $vgpr23_lo16
; GFX11-TRUE16-NEXT: ; implicit-def: $vgpr30_hi16
; GFX11-TRUE16-NEXT: ; implicit-def: $vgpr101_lo16
-; GFX11-TRUE16-NEXT: ; implicit-def: $vgpr29_hi16
+; GFX11-TRUE16-NEXT: ; implicit-def: $vgpr27_hi16
; GFX11-TRUE16-NEXT: ; implicit-def: $vgpr100_lo16
-; GFX11-TRUE16-NEXT: ; implicit-def: $vgpr35_hi16
+; GFX11-TRUE16-NEXT: ; implicit-def: $vgpr38_hi16
; GFX11-TRUE16-NEXT: ; implicit-def: $vgpr99_lo16
; GFX11-TRUE16-NEXT: ; implicit-def: $vgpr22_lo16
-; GFX11-TRUE16-NEXT: ; implicit-def: $vgpr32_hi16
-; GFX11-TRUE16-NEXT: ; implicit-def: $vgpr98_lo16
; GFX11-TRUE16-NEXT: ; implicit-def: $vgpr33_hi16
+; GFX11-TRUE16-NEXT: ; implicit-def: $vgpr98_lo16
+; GFX11-TRUE16-NEXT: ; implicit-def: $vgpr29_hi16
; GFX11-TRUE16-NEXT: ; implicit-def: $vgpr97_lo16
-; GFX11-TRUE16-NEXT: ; implicit-def: $vgpr49_hi16
+; GFX11-TRUE16-NEXT: ; implicit-def: $vgpr50_hi16
; GFX11-TRUE16-NEXT: ; implicit-def: $vgpr96_lo16
; GFX11-TRUE16-NEXT: ; implicit-def: $vgpr21_lo16
-; GFX11-TRUE16-NEXT: ; implicit-def: $vgpr34_hi16
-; GFX11-TRUE16-NEXT: ; implicit-def: $vgpr87_lo16
; GFX11-TRUE16-NEXT: ; implicit-def: $vgpr36_hi16
+; GFX11-TRUE16-NEXT: ; implicit-def: $vgpr87_lo16
+; GFX11-TRUE16-NEXT: ; implicit-def: $vgpr32_hi16
; GFX11-TRUE16-NEXT: ; implicit-def: $vgpr86_lo16
; GFX11-TRUE16-NEXT: ; implicit-def: $vgpr65_hi16
-; GFX11-TRUE16-NEXT: ; implicit-def: $vgpr83_lo16
+; GFX11-TRUE16-NEXT: ; implicit-def: $vgpr85_lo16
; GFX11-TRUE16-NEXT: ; implicit-def: $vgpr20_lo16
-; GFX11-TRUE16-NEXT: ; implicit-def: $vgpr38_hi16
-; GFX11-TRUE16-NEXT: ; implicit-def: $vgpr80_lo16
-; GFX11-TRUE16-NEXT: ; implicit-def: $vgpr48_hi16
-; GFX11-TRUE16-NEXT: ; implicit-def: $vgpr71_lo16
-; GFX11-TRUE16-NEXT: ; implicit-def: $vgpr70_hi16
-; GFX11-TRUE16-NEXT: ; implicit-def: $vgpr69_lo16
+; GFX11-TRUE16-NEXT: ; implicit-def: $vgpr39_hi16
+; GFX11-TRUE16-NEXT: ; implicit-def: $vgpr84_lo16
+; GFX11-TRUE16-NEXT: ; implicit-def: $vgpr35_hi16
+; GFX11-TRUE16-NEXT: ; implicit-def: $vgpr83_lo16
+; GFX11-TRUE16-NEXT: ; implicit-def: $vgpr68_hi16
+; GFX11-TRUE16-NEXT: ; implicit-def: $vgpr82_lo16
; GFX11-TRUE16-NEXT: ; implicit-def: $vgpr19_lo16
-; GFX11-TRUE16-NEXT: ; implicit-def: $vgpr52_hi16
+; GFX11-TRUE16-NEXT: ; implicit-def: $vgpr49_hi16
+; GFX11-TRUE16-NEXT: ; implicit-def: $vgpr80_lo16
+; GFX11-TRUE16-NEXT: ; implicit-def: $vgpr37_hi16
+; GFX11-TRUE16-NEXT: ; implicit-def: $vgpr70_lo16
+; GFX11-TRUE16-NEXT: ; implicit-def: $vgpr81_hi16
+; GFX11-TRUE16-NEXT: ; implicit-def: $vgpr67_lo16
+; GFX11-TRUE16-NEXT: ; implicit-def: $vgpr18_lo16
+; GFX11-TRUE16-NEXT: ; implicit-def: $vgpr64_hi16
; GFX11-TRUE16-NEXT: ; implicit-def: $vgpr66_lo16
-; GFX11-TRUE16-NEXT: ; implicit-def: $vgpr53_hi16
-; GFX11-TRUE16-NEXT: ; implicit-def: $vgpr64_lo16
-; GFX11-TRUE16-NEXT: ; implicit-def: $vgpr85_hi16
+; GFX11-TRUE16-NEXT: ; implicit-def: $vgpr48_hi16
; GFX11-TRUE16-NEXT: ; implicit-def: $vgpr55_lo16
-; GFX11-TRUE16-NEXT: ; implicit-def: $vgpr18_lo16
-; GFX11-TRUE16-NEXT: ; implicit-def: $vgpr68_hi16
-; GFX11-TRUE16-NEXT: ; implicit-def: $vgpr54_lo16
-; GFX11-TRUE16-NEXT: ; implicit-def: $vgpr67_hi16
-; GFX11-TRUE16-NEXT: ; implicit-def: $vgpr51_lo16
-; GFX11-TRUE16-NEXT: ; implicit-def: $vgpr84_hi16
-; GFX11-TRUE16-NEXT: ; implicit-def: $vgpr50_lo16
+; GFX11-TRUE16-NEXT: ; implicit-def: $vgpr71_hi16
+; GFX11-TRUE16-NEXT: ; implicit-def: $vgpr53_lo16
; GFX11-TRUE16-NEXT: ; implicit-def: $vgpr17_lo16
-; GFX11-TRUE16-NEXT: ; implicit-def: $vgpr81_hi16
-; GFX11-TRUE16-NEXT: ; implicit-def: $vgpr39_lo16
-; GFX11-TRUE16-NEXT: ; implicit-def: $vgpr82_hi16
-; GFX11-TRUE16-NEXT: ; implicit-def: $vgpr37_lo16
+; GFX11-TRUE16-NEXT: ; implicit-def: $vgpr69_hi16
+; GFX11-TRUE16-NEXT: ; implicit-def: $vgpr52_lo16
+; GFX11-TRUE16-NEXT: ; implicit-def: $vgpr54_hi16
+; GFX11-TRUE16-NEXT: ; implicit-def: $vgpr51_lo16
; GFX11-TRUE16-NEXT: s_and_saveexec_b32 s0, vcc_lo
; GFX11-TRUE16-NEXT: s_delay_alu instid0(SALU_CYCLE_1)
; GFX11-TRUE16-NEXT: s_xor_b32 s0, exec_lo, s0
@@ -88634,18 +88549,18 @@ define <64 x i8> @bitcast_v32bf16_to_v64i8(<32 x bfloat> %a, i32 %b) {
; GFX11-TRUE16-NEXT: v_lshrrev_b64 v[21:22], 24, v[7:8]
; GFX11-TRUE16-NEXT: v_lshrrev_b64 v[22:23], 24, v[5:6]
; GFX11-TRUE16-NEXT: v_lshrrev_b64 v[23:24], 24, v[3:4]
-; GFX11-TRUE16-NEXT: v_lshrrev_b32_e32 v37, 24, v16
-; GFX11-TRUE16-NEXT: v_lshrrev_b32_e32 v39, 8, v16
-; GFX11-TRUE16-NEXT: v_lshrrev_b32_e32 v50, 8, v15
-; GFX11-TRUE16-NEXT: v_lshrrev_b32_e32 v51, 24, v14
-; GFX11-TRUE16-NEXT: v_lshrrev_b32_e32 v54, 8, v14
-; GFX11-TRUE16-NEXT: v_lshrrev_b32_e32 v55, 8, v13
-; GFX11-TRUE16-NEXT: v_lshrrev_b32_e32 v64, 24, v12
-; GFX11-TRUE16-NEXT: v_lshrrev_b32_e32 v66, 8, v12
-; GFX11-TRUE16-NEXT: v_lshrrev_b32_e32 v69, 8, v11
-; GFX11-TRUE16-NEXT: v_lshrrev_b32_e32 v71, 24, v10
-; GFX11-TRUE16-NEXT: v_lshrrev_b32_e32 v80, 8, v10
-; GFX11-TRUE16-NEXT: v_lshrrev_b32_e32 v83, 8, v9
+; GFX11-TRUE16-NEXT: v_lshrrev_b32_e32 v51, 24, v16
+; GFX11-TRUE16-NEXT: v_lshrrev_b32_e32 v52, 8, v16
+; GFX11-TRUE16-NEXT: v_lshrrev_b32_e32 v53, 8, v15
+; GFX11-TRUE16-NEXT: v_lshrrev_b32_e32 v55, 24, v14
+; GFX11-TRUE16-NEXT: v_lshrrev_b32_e32 v66, 8, v14
+; GFX11-TRUE16-NEXT: v_lshrrev_b32_e32 v67, 8, v13
+; GFX11-TRUE16-NEXT: v_lshrrev_b32_e32 v70, 24, v12
+; GFX11-TRUE16-NEXT: v_lshrrev_b32_e32 v80, 8, v12
+; GFX11-TRUE16-NEXT: v_lshrrev_b32_e32 v82, 8, v11
+; GFX11-TRUE16-NEXT: v_lshrrev_b32_e32 v83, 24, v10
+; GFX11-TRUE16-NEXT: v_lshrrev_b32_e32 v84, 8, v10
+; GFX11-TRUE16-NEXT: v_lshrrev_b32_e32 v85, 8, v9
; GFX11-TRUE16-NEXT: v_lshrrev_b32_e32 v86, 24, v8
; GFX11-TRUE16-NEXT: v_lshrrev_b32_e32 v87, 8, v8
; GFX11-TRUE16-NEXT: v_lshrrev_b32_e32 v96, 8, v7
@@ -88658,246 +88573,247 @@ define <64 x i8> @bitcast_v32bf16_to_v64i8(<32 x bfloat> %a, i32 %b) {
; GFX11-TRUE16-NEXT: v_lshrrev_b32_e32 v103, 24, v2
; GFX11-TRUE16-NEXT: v_lshrrev_b32_e32 v112, 8, v2
; GFX11-TRUE16-NEXT: v_lshrrev_b32_e32 v113, 8, v1
+; GFX11-TRUE16-NEXT: v_mov_b16_e32 v54.h, v16.h
+; GFX11-TRUE16-NEXT: v_mov_b16_e32 v48.h, v14.h
+; GFX11-TRUE16-NEXT: v_mov_b16_e32 v37.h, v12.h
+; GFX11-TRUE16-NEXT: v_mov_b16_e32 v35.h, v10.h
+; GFX11-TRUE16-NEXT: v_mov_b16_e32 v32.h, v8.h
+; GFX11-TRUE16-NEXT: v_mov_b16_e32 v29.h, v6.h
+; GFX11-TRUE16-NEXT: v_mov_b16_e32 v27.h, v4.h
+; GFX11-TRUE16-NEXT: v_mov_b16_e32 v26.h, v2.h
; GFX11-TRUE16-NEXT: v_lshrrev_b64 v[24:25], 24, v[1:2]
-; GFX11-TRUE16-NEXT: v_mov_b16_e32 v28.h, v1.l
-; GFX11-TRUE16-NEXT: v_mov_b16_e32 v26.h, v2.l
-; GFX11-TRUE16-NEXT: v_mov_b16_e32 v27.h, v2.h
-; GFX11-TRUE16-NEXT: v_mov_b16_e32 v31.h, v3.l
-; GFX11-TRUE16-NEXT: v_mov_b16_e32 v30.h, v4.l
-; GFX11-TRUE16-NEXT: v_mov_b16_e32 v29.h, v4.h
-; GFX11-TRUE16-NEXT: v_mov_b16_e32 v35.h, v5.l
-; GFX11-TRUE16-NEXT: v_mov_b16_e32 v32.h, v6.l
-; GFX11-TRUE16-NEXT: v_mov_b16_e32 v33.h, v6.h
-; GFX11-TRUE16-NEXT: v_mov_b16_e32 v49.h, v7.l
-; GFX11-TRUE16-NEXT: v_mov_b16_e32 v34.h, v8.l
-; GFX11-TRUE16-NEXT: v_mov_b16_e32 v36.h, v8.h
+; GFX11-TRUE16-NEXT: v_mov_b16_e32 v69.h, v16.l
+; GFX11-TRUE16-NEXT: v_mov_b16_e32 v71.h, v15.l
+; GFX11-TRUE16-NEXT: v_mov_b16_e32 v64.h, v14.l
+; GFX11-TRUE16-NEXT: v_mov_b16_e32 v81.h, v13.l
+; GFX11-TRUE16-NEXT: v_mov_b16_e32 v49.h, v12.l
+; GFX11-TRUE16-NEXT: v_mov_b16_e32 v68.h, v11.l
+; GFX11-TRUE16-NEXT: v_mov_b16_e32 v39.h, v10.l
; GFX11-TRUE16-NEXT: v_mov_b16_e32 v65.h, v9.l
-; GFX11-TRUE16-NEXT: v_mov_b16_e32 v38.h, v10.l
-; GFX11-TRUE16-NEXT: v_mov_b16_e32 v48.h, v10.h
-; GFX11-TRUE16-NEXT: v_mov_b16_e32 v70.h, v11.l
-; GFX11-TRUE16-NEXT: v_mov_b16_e32 v52.h, v12.l
-; GFX11-TRUE16-NEXT: v_mov_b16_e32 v53.h, v12.h
-; GFX11-TRUE16-NEXT: v_mov_b16_e32 v85.h, v13.l
-; GFX11-TRUE16-NEXT: v_mov_b16_e32 v68.h, v14.l
-; GFX11-TRUE16-NEXT: v_mov_b16_e32 v67.h, v14.h
-; GFX11-TRUE16-NEXT: v_mov_b16_e32 v84.h, v15.l
-; GFX11-TRUE16-NEXT: v_mov_b16_e32 v81.h, v16.l
-; GFX11-TRUE16-NEXT: v_mov_b16_e32 v82.h, v16.h
+; GFX11-TRUE16-NEXT: v_mov_b16_e32 v36.h, v8.l
+; GFX11-TRUE16-NEXT: v_mov_b16_e32 v50.h, v7.l
+; GFX11-TRUE16-NEXT: v_mov_b16_e32 v33.h, v6.l
+; GFX11-TRUE16-NEXT: v_mov_b16_e32 v38.h, v5.l
+; GFX11-TRUE16-NEXT: v_mov_b16_e32 v30.h, v4.l
+; GFX11-TRUE16-NEXT: v_mov_b16_e32 v34.h, v3.l
+; GFX11-TRUE16-NEXT: v_mov_b16_e32 v28.h, v2.l
+; GFX11-TRUE16-NEXT: v_mov_b16_e32 v31.h, v1.l
; GFX11-TRUE16-NEXT: .LBB108_2: ; %Flow
; GFX11-TRUE16-NEXT: s_and_not1_saveexec_b32 s0, s0
; GFX11-TRUE16-NEXT: s_cbranch_execz .LBB108_4
; GFX11-TRUE16-NEXT: ; %bb.3: ; %cmp.true
; GFX11-TRUE16-NEXT: v_and_b32_e32 v18, 0xffff0000, v1
+; GFX11-TRUE16-NEXT: v_lshlrev_b32_e32 v1, 16, v1
; GFX11-TRUE16-NEXT: v_and_b32_e32 v19, 0xffff0000, v4
; GFX11-TRUE16-NEXT: v_lshlrev_b32_e32 v4, 16, v4
-; GFX11-TRUE16-NEXT: v_lshlrev_b32_e32 v17, 16, v2
-; GFX11-TRUE16-NEXT: v_and_b32_e32 v2, 0xffff0000, v2
-; GFX11-TRUE16-NEXT: v_dual_add_f32 v18, 0x40c00000, v18 :: v_dual_lshlrev_b32 v1, 16, v1
-; GFX11-TRUE16-NEXT: v_and_b32_e32 v39, 0xffff0000, v11
-; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_2) | instskip(SKIP_1) | instid1(VALU_DEP_2)
-; GFX11-TRUE16-NEXT: v_dual_add_f32 v2, 0x40c00000, v2 :: v_dual_add_f32 v1, 0x40c00000, v1
-; GFX11-TRUE16-NEXT: v_dual_add_f32 v4, 0x40c00000, v4 :: v_dual_add_f32 v17, 0x40c00000, v17
-; GFX11-TRUE16-NEXT: v_bfe_u32 v21, v2, 16, 1
-; GFX11-TRUE16-NEXT: v_or_b32_e32 v23, 0x400000, v2
-; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_3) | instskip(SKIP_3) | instid1(VALU_DEP_4)
+; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_3) | instskip(NEXT) | instid1(VALU_DEP_2)
+; GFX11-TRUE16-NEXT: v_dual_add_f32 v18, 0x40c00000, v18 :: v_dual_add_f32 v1, 0x40c00000, v1
+; GFX11-TRUE16-NEXT: v_dual_add_f32 v4, 0x40c00000, v4 :: v_dual_lshlrev_b32 v17, 16, v2
+; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
+; GFX11-TRUE16-NEXT: v_add_f32_e32 v17, 0x40c00000, v17
; GFX11-TRUE16-NEXT: v_bfe_u32 v20, v17, 16, 1
; GFX11-TRUE16-NEXT: v_or_b32_e32 v22, 0x400000, v17
; GFX11-TRUE16-NEXT: v_cmp_u_f32_e32 vcc_lo, v17, v17
-; GFX11-TRUE16-NEXT: v_add3_u32 v21, v21, v2, 0x7fff
+; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_3) | instskip(SKIP_1) | instid1(VALU_DEP_2)
; GFX11-TRUE16-NEXT: v_add3_u32 v20, v20, v17, 0x7fff
; GFX11-TRUE16-NEXT: v_bfe_u32 v17, v18, 16, 1
-; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_2)
-; GFX11-TRUE16-NEXT: v_cndmask_b32_e32 v26, v20, v22, vcc_lo
+; GFX11-TRUE16-NEXT: v_cndmask_b32_e32 v28, v20, v22, vcc_lo
; GFX11-TRUE16-NEXT: v_bfe_u32 v20, v1, 16, 1
-; GFX11-TRUE16-NEXT: v_cmp_u_f32_e32 vcc_lo, v2, v2
; GFX11-TRUE16-NEXT: v_or_b32_e32 v22, 0x400000, v18
+; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_4) | instskip(NEXT) | instid1(VALU_DEP_3)
; GFX11-TRUE16-NEXT: v_add3_u32 v17, v17, v18, 0x7fff
-; GFX11-TRUE16-NEXT: v_mov_b16_e32 v2.l, v26.h
; GFX11-TRUE16-NEXT: v_add3_u32 v20, v20, v1, 0x7fff
-; GFX11-TRUE16-NEXT: v_cndmask_b32_e32 v27, v21, v23, vcc_lo
+; GFX11-TRUE16-NEXT: v_and_b32_e32 v2, 0xffff0000, v2
+; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
+; GFX11-TRUE16-NEXT: v_add_f32_e32 v2, 0x40c00000, v2
+; GFX11-TRUE16-NEXT: v_bfe_u32 v21, v2, 16, 1
+; GFX11-TRUE16-NEXT: v_or_b32_e32 v23, 0x400000, v2
+; GFX11-TRUE16-NEXT: v_cmp_u_f32_e32 vcc_lo, v2, v2
+; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_3) | instskip(SKIP_1) | instid1(VALU_DEP_2)
+; GFX11-TRUE16-NEXT: v_add3_u32 v21, v21, v2, 0x7fff
+; GFX11-TRUE16-NEXT: v_mov_b16_e32 v2.l, v28.h
+; GFX11-TRUE16-NEXT: v_cndmask_b32_e32 v26, v21, v23, vcc_lo
; GFX11-TRUE16-NEXT: v_or_b32_e32 v21, 0x400000, v1
; GFX11-TRUE16-NEXT: v_cmp_u_f32_e32 vcc_lo, v1, v1
-; GFX11-TRUE16-NEXT: v_add_f32_e32 v19, 0x40c00000, v19
-; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_4) | instskip(NEXT) | instid1(VALU_DEP_4)
-; GFX11-TRUE16-NEXT: v_bfi_b32 v2, 0xffff, v2, v27
-; GFX11-TRUE16-NEXT: v_cndmask_b32_e32 v28, v20, v21, vcc_lo
+; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_2) | instskip(SKIP_3) | instid1(VALU_DEP_4)
+; GFX11-TRUE16-NEXT: v_cndmask_b32_e32 v31, v20, v21, vcc_lo
; GFX11-TRUE16-NEXT: v_cmp_u_f32_e32 vcc_lo, v18, v18
; GFX11-TRUE16-NEXT: v_bfe_u32 v18, v4, 16, 1
-; GFX11-TRUE16-NEXT: v_or_b32_e32 v23, 0x400000, v19
-; GFX11-TRUE16-NEXT: v_lshrrev_b32_e32 v103, 24, v2
-; GFX11-TRUE16-NEXT: v_mov_b16_e32 v20.l, v28.h
+; GFX11-TRUE16-NEXT: v_and_b32_e32 v21, 0xffff0000, v3
+; GFX11-TRUE16-NEXT: v_mov_b16_e32 v20.l, v31.h
; GFX11-TRUE16-NEXT: v_cndmask_b32_e32 v1, v17, v22, vcc_lo
-; GFX11-TRUE16-NEXT: v_bfe_u32 v17, v19, 16, 1
+; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_4)
; GFX11-TRUE16-NEXT: v_add3_u32 v18, v18, v4, 0x7fff
; GFX11-TRUE16-NEXT: v_or_b32_e32 v22, 0x400000, v4
; GFX11-TRUE16-NEXT: v_cmp_u_f32_e32 vcc_lo, v4, v4
-; GFX11-TRUE16-NEXT: v_and_b32_e32 v21, 0xffff0000, v3
-; GFX11-TRUE16-NEXT: v_add3_u32 v17, v17, v19, 0x7fff
+; GFX11-TRUE16-NEXT: v_add_f32_e32 v19, 0x40c00000, v19
+; GFX11-TRUE16-NEXT: v_add_f32_e32 v21, 0x40c00000, v21
; GFX11-TRUE16-NEXT: v_bfi_b32 v1, 0xffff, v20, v1
-; GFX11-TRUE16-NEXT: v_lshrrev_b32_e32 v112, 8, v2
+; GFX11-TRUE16-NEXT: v_bfi_b32 v2, 0xffff, v2, v26
; GFX11-TRUE16-NEXT: v_cndmask_b32_e32 v30, v18, v22, vcc_lo
+; GFX11-TRUE16-NEXT: v_bfe_u32 v17, v19, 16, 1
+; GFX11-TRUE16-NEXT: v_or_b32_e32 v23, 0x400000, v19
; GFX11-TRUE16-NEXT: v_cmp_u_f32_e32 vcc_lo, v19, v19
; GFX11-TRUE16-NEXT: v_and_b32_e32 v22, 0xffff0000, v6
; GFX11-TRUE16-NEXT: v_lshlrev_b32_e32 v3, 16, v3
-; GFX11-TRUE16-NEXT: v_dual_add_f32 v21, 0x40c00000, v21 :: v_dual_lshlrev_b32 v6, 16, v6
-; GFX11-TRUE16-NEXT: v_cndmask_b32_e32 v29, v17, v23, vcc_lo
-; GFX11-TRUE16-NEXT: v_mov_b16_e32 v17.l, v30.h
-; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_4) | instskip(NEXT) | instid1(VALU_DEP_4)
-; GFX11-TRUE16-NEXT: v_add_f32_e32 v3, 0x40c00000, v3
+; GFX11-TRUE16-NEXT: v_add3_u32 v17, v17, v19, 0x7fff
+; GFX11-TRUE16-NEXT: v_lshlrev_b32_e32 v6, 16, v6
; GFX11-TRUE16-NEXT: v_bfe_u32 v4, v21, 16, 1
-; GFX11-TRUE16-NEXT: v_add_f32_e32 v6, 0x40c00000, v6
+; GFX11-TRUE16-NEXT: v_lshrrev_b32_e32 v103, 24, v2
+; GFX11-TRUE16-NEXT: v_add_f32_e32 v3, 0x40c00000, v3
+; GFX11-TRUE16-NEXT: v_cndmask_b32_e32 v27, v17, v23, vcc_lo
; GFX11-TRUE16-NEXT: v_or_b32_e32 v23, 0x400000, v21
-; GFX11-TRUE16-NEXT: v_and_b32_e32 v20, 0xffff0000, v5
-; GFX11-TRUE16-NEXT: v_bfe_u32 v18, v3, 16, 1
-; GFX11-TRUE16-NEXT: v_or_b32_e32 v19, 0x400000, v3
-; GFX11-TRUE16-NEXT: v_cmp_u_f32_e32 vcc_lo, v3, v3
; GFX11-TRUE16-NEXT: v_add3_u32 v4, v4, v21, 0x7fff
-; GFX11-TRUE16-NEXT: v_add_f32_e32 v20, 0x40c00000, v20
-; GFX11-TRUE16-NEXT: v_add3_u32 v18, v18, v3, 0x7fff
+; GFX11-TRUE16-NEXT: v_mov_b16_e32 v17.l, v30.h
+; GFX11-TRUE16-NEXT: v_bfe_u32 v18, v3, 16, 1
+; GFX11-TRUE16-NEXT: v_or_b32_e32 v19, 0x400000, v3
+; GFX11-TRUE16-NEXT: v_cmp_u_f32_e32 vcc_lo, v3, v3
+; GFX11-TRUE16-NEXT: v_lshrrev_b32_e32 v112, 8, v2
; GFX11-TRUE16-NEXT: v_lshrrev_b32_e32 v113, 8, v1
-; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_2) | instskip(SKIP_1) | instid1(VALU_DEP_2)
-; GFX11-TRUE16-NEXT: v_dual_cndmask_b32 v31, v18, v19 :: v_dual_add_f32 v18, 0x40c00000, v22
+; GFX11-TRUE16-NEXT: v_add3_u32 v18, v18, v3, 0x7fff
+; GFX11-TRUE16-NEXT: v_add_f32_e32 v6, 0x40c00000, v6
+; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_2) | instskip(SKIP_2) | instid1(VALU_DEP_1)
+; GFX11-TRUE16-NEXT: v_cndmask_b32_e32 v34, v18, v19, vcc_lo
; GFX11-TRUE16-NEXT: v_cmp_u_f32_e32 vcc_lo, v21, v21
-; GFX11-TRUE16-NEXT: v_mov_b16_e32 v19.l, v31.h
-; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_3)
+; GFX11-TRUE16-NEXT: v_dual_add_f32 v18, 0x40c00000, v22 :: v_dual_cndmask_b32 v3, v4, v23
; GFX11-TRUE16-NEXT: v_bfe_u32 v21, v18, 16, 1
-; GFX11-TRUE16-NEXT: v_cndmask_b32_e32 v3, v4, v23, vcc_lo
-; GFX11-TRUE16-NEXT: v_bfi_b32 v4, 0xffff, v17, v29
+; GFX11-TRUE16-NEXT: v_bfi_b32 v4, 0xffff, v17, v27
; GFX11-TRUE16-NEXT: v_bfe_u32 v17, v6, 16, 1
-; GFX11-TRUE16-NEXT: v_cmp_u_f32_e32 vcc_lo, v6, v6
; GFX11-TRUE16-NEXT: v_or_b32_e32 v22, 0x400000, v18
-; GFX11-TRUE16-NEXT: v_bfi_b32 v3, 0xffff, v19, v3
-; GFX11-TRUE16-NEXT: v_add3_u32 v19, v21, v18, 0x7fff
+; GFX11-TRUE16-NEXT: v_and_b32_e32 v20, 0xffff0000, v5
+; GFX11-TRUE16-NEXT: v_lshlrev_b32_e32 v5, 16, v5
+; GFX11-TRUE16-NEXT: v_mov_b16_e32 v19.l, v34.h
; GFX11-TRUE16-NEXT: v_add3_u32 v17, v17, v6, 0x7fff
-; GFX11-TRUE16-NEXT: v_or_b32_e32 v21, 0x400000, v6
-; GFX11-TRUE16-NEXT: v_bfe_u32 v6, v20, 16, 1
+; GFX11-TRUE16-NEXT: v_cmp_u_f32_e32 vcc_lo, v6, v6
; GFX11-TRUE16-NEXT: v_lshrrev_b32_e32 v100, 24, v4
; GFX11-TRUE16-NEXT: v_lshrrev_b32_e32 v101, 8, v4
+; GFX11-TRUE16-NEXT: v_bfi_b32 v3, 0xffff, v19, v3
+; GFX11-TRUE16-NEXT: v_add3_u32 v19, v21, v18, 0x7fff
+; GFX11-TRUE16-NEXT: v_or_b32_e32 v21, 0x400000, v6
+; GFX11-TRUE16-NEXT: v_add_f32_e32 v5, 0x40c00000, v5
+; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_4) | instskip(NEXT) | instid1(VALU_DEP_3)
; GFX11-TRUE16-NEXT: v_lshrrev_b32_e32 v102, 8, v3
-; GFX11-TRUE16-NEXT: v_cndmask_b32_e32 v32, v17, v21, vcc_lo
+; GFX11-TRUE16-NEXT: v_cndmask_b32_e32 v33, v17, v21, vcc_lo
; GFX11-TRUE16-NEXT: v_cmp_u_f32_e32 vcc_lo, v18, v18
-; GFX11-TRUE16-NEXT: v_add3_u32 v6, v6, v20, 0x7fff
-; GFX11-TRUE16-NEXT: v_cndmask_b32_e32 v33, v19, v22, vcc_lo
-; GFX11-TRUE16-NEXT: v_and_b32_e32 v19, 0xffff0000, v8
-; GFX11-TRUE16-NEXT: v_lshlrev_b32_e32 v8, 16, v8
-; GFX11-TRUE16-NEXT: v_lshlrev_b32_e32 v5, 16, v5
-; GFX11-TRUE16-NEXT: v_or_b32_e32 v22, 0x400000, v20
-; GFX11-TRUE16-NEXT: v_mov_b16_e32 v18.l, v32.h
-; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_3) | instskip(NEXT) | instid1(VALU_DEP_1)
-; GFX11-TRUE16-NEXT: v_dual_add_f32 v8, 0x40c00000, v8 :: v_dual_add_f32 v5, 0x40c00000, v5
+; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_4) | instskip(SKIP_2) | instid1(VALU_DEP_3)
; GFX11-TRUE16-NEXT: v_bfe_u32 v17, v5, 16, 1
; GFX11-TRUE16-NEXT: v_or_b32_e32 v21, 0x400000, v5
-; GFX11-TRUE16-NEXT: v_cmp_u_f32_e32 vcc_lo, v5, v5
-; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_3) | instskip(NEXT) | instid1(VALU_DEP_1)
+; GFX11-TRUE16-NEXT: v_dual_cndmask_b32 v29, v19, v22 :: v_dual_add_f32 v20, 0x40c00000, v20
; GFX11-TRUE16-NEXT: v_add3_u32 v17, v17, v5, 0x7fff
-; GFX11-TRUE16-NEXT: v_cndmask_b32_e32 v35, v17, v21, vcc_lo
-; GFX11-TRUE16-NEXT: v_bfe_u32 v17, v8, 16, 1
+; GFX11-TRUE16-NEXT: v_cmp_u_f32_e32 vcc_lo, v5, v5
+; GFX11-TRUE16-NEXT: v_and_b32_e32 v19, 0xffff0000, v8
+; GFX11-TRUE16-NEXT: v_lshlrev_b32_e32 v8, 16, v8
+; GFX11-TRUE16-NEXT: v_bfe_u32 v6, v20, 16, 1
+; GFX11-TRUE16-NEXT: v_or_b32_e32 v22, 0x400000, v20
+; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_4) | instskip(NEXT) | instid1(VALU_DEP_4)
+; GFX11-TRUE16-NEXT: v_dual_cndmask_b32 v38, v17, v21 :: v_dual_add_f32 v19, 0x40c00000, v19
+; GFX11-TRUE16-NEXT: v_add_f32_e32 v8, 0x40c00000, v8
+; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_4) | instskip(SKIP_1) | instid1(VALU_DEP_4)
+; GFX11-TRUE16-NEXT: v_add3_u32 v6, v6, v20, 0x7fff
; GFX11-TRUE16-NEXT: v_cmp_u_f32_e32 vcc_lo, v20, v20
-; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_3) | instskip(SKIP_1) | instid1(VALU_DEP_4)
-; GFX11-TRUE16-NEXT: v_mov_b16_e32 v21.l, v35.h
+; GFX11-TRUE16-NEXT: v_mov_b16_e32 v21.l, v38.h
+; GFX11-TRUE16-NEXT: v_bfe_u32 v5, v19, 16, 1
+; GFX11-TRUE16-NEXT: v_bfe_u32 v17, v8, 16, 1
+; GFX11-TRUE16-NEXT: v_mov_b16_e32 v18.l, v33.h
; GFX11-TRUE16-NEXT: v_cndmask_b32_e32 v20, v6, v22, vcc_lo
+; GFX11-TRUE16-NEXT: v_cmp_u_f32_e32 vcc_lo, v8, v8
+; GFX11-TRUE16-NEXT: v_add3_u32 v5, v5, v19, 0x7fff
; GFX11-TRUE16-NEXT: v_add3_u32 v6, v17, v8, 0x7fff
; GFX11-TRUE16-NEXT: v_or_b32_e32 v17, 0x400000, v8
-; GFX11-TRUE16-NEXT: v_cmp_u_f32_e32 vcc_lo, v8, v8
-; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_2) | instskip(SKIP_1) | instid1(VALU_DEP_2)
-; GFX11-TRUE16-NEXT: v_dual_cndmask_b32 v34, v6, v17 :: v_dual_add_f32 v19, 0x40c00000, v19
-; GFX11-TRUE16-NEXT: v_bfi_b32 v6, 0xffff, v18, v33
-; GFX11-TRUE16-NEXT: v_mov_b16_e32 v8.l, v34.h
-; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_3)
-; GFX11-TRUE16-NEXT: v_bfe_u32 v5, v19, 16, 1
; GFX11-TRUE16-NEXT: v_or_b32_e32 v22, 0x400000, v19
+; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_2) | instskip(SKIP_2) | instid1(VALU_DEP_3)
+; GFX11-TRUE16-NEXT: v_dual_cndmask_b32 v36, v6, v17 :: v_dual_and_b32 v23, 0xffff0000, v7
; GFX11-TRUE16-NEXT: v_cmp_u_f32_e32 vcc_lo, v19, v19
-; GFX11-TRUE16-NEXT: v_lshrrev_b32_e32 v97, 24, v6
-; GFX11-TRUE16-NEXT: v_lshrrev_b32_e32 v98, 8, v6
-; GFX11-TRUE16-NEXT: v_add3_u32 v5, v5, v19, 0x7fff
-; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(SKIP_3) | instid1(VALU_DEP_4)
-; GFX11-TRUE16-NEXT: v_dual_cndmask_b32 v36, v5, v22 :: v_dual_and_b32 v23, 0xffff0000, v7
-; GFX11-TRUE16-NEXT: v_bfi_b32 v5, 0xffff, v21, v20
-; GFX11-TRUE16-NEXT: v_lshlrev_b32_e32 v20, 16, v10
; GFX11-TRUE16-NEXT: v_lshlrev_b32_e32 v7, 16, v7
-; GFX11-TRUE16-NEXT: v_bfi_b32 v8, 0xffff, v8, v36
-; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_4) | instskip(NEXT) | instid1(VALU_DEP_3)
-; GFX11-TRUE16-NEXT: v_lshrrev_b32_e32 v99, 8, v5
-; GFX11-TRUE16-NEXT: v_dual_add_f32 v20, 0x40c00000, v20 :: v_dual_add_f32 v7, 0x40c00000, v7
-; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_3) | instskip(SKIP_1) | instid1(VALU_DEP_3)
-; GFX11-TRUE16-NEXT: v_lshrrev_b32_e32 v86, 24, v8
-; GFX11-TRUE16-NEXT: v_lshrrev_b32_e32 v87, 8, v8
+; GFX11-TRUE16-NEXT: v_add_f32_e32 v17, 0x40c00000, v23
+; GFX11-TRUE16-NEXT: v_bfi_b32 v6, 0xffff, v18, v29
+; GFX11-TRUE16-NEXT: v_dual_cndmask_b32 v32, v5, v22 :: v_dual_and_b32 v23, 0xffff0000, v9
+; GFX11-TRUE16-NEXT: v_bfi_b32 v5, 0xffff, v21, v20
+; GFX11-TRUE16-NEXT: v_dual_add_f32 v7, 0x40c00000, v7 :: v_dual_lshlrev_b32 v20, 16, v10
+; GFX11-TRUE16-NEXT: v_bfe_u32 v18, v17, 16, 1
+; GFX11-TRUE16-NEXT: v_or_b32_e32 v22, 0x400000, v17
+; GFX11-TRUE16-NEXT: v_mov_b16_e32 v8.l, v36.h
+; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_4)
+; GFX11-TRUE16-NEXT: v_add_f32_e32 v20, 0x40c00000, v20
; GFX11-TRUE16-NEXT: v_bfe_u32 v19, v7, 16, 1
; GFX11-TRUE16-NEXT: v_or_b32_e32 v21, 0x400000, v7
; GFX11-TRUE16-NEXT: v_cmp_u_f32_e32 vcc_lo, v7, v7
-; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_3) | instskip(SKIP_2) | instid1(VALU_DEP_3)
+; GFX11-TRUE16-NEXT: v_add3_u32 v18, v18, v17, 0x7fff
+; GFX11-TRUE16-NEXT: v_bfi_b32 v8, 0xffff, v8, v32
; GFX11-TRUE16-NEXT: v_add3_u32 v19, v19, v7, 0x7fff
-; GFX11-TRUE16-NEXT: v_and_b32_e32 v10, 0xffff0000, v10
; GFX11-TRUE16-NEXT: v_bfe_u32 v7, v20, 16, 1
-; GFX11-TRUE16-NEXT: v_cndmask_b32_e32 v49, v19, v21, vcc_lo
-; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_3) | instskip(NEXT) | instid1(VALU_DEP_3)
-; GFX11-TRUE16-NEXT: v_dual_add_f32 v17, 0x40c00000, v23 :: v_dual_add_f32 v10, 0x40c00000, v10
+; GFX11-TRUE16-NEXT: v_lshrrev_b32_e32 v97, 24, v6
+; GFX11-TRUE16-NEXT: v_lshrrev_b32_e32 v98, 8, v6
+; GFX11-TRUE16-NEXT: v_lshrrev_b32_e32 v86, 24, v8
+; GFX11-TRUE16-NEXT: v_cndmask_b32_e32 v50, v19, v21, vcc_lo
+; GFX11-TRUE16-NEXT: v_cmp_u_f32_e32 vcc_lo, v17, v17
; GFX11-TRUE16-NEXT: v_add3_u32 v7, v7, v20, 0x7fff
; GFX11-TRUE16-NEXT: v_or_b32_e32 v21, 0x400000, v20
-; GFX11-TRUE16-NEXT: v_and_b32_e32 v23, 0xffff0000, v9
-; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_4) | instskip(SKIP_3) | instid1(VALU_DEP_4)
-; GFX11-TRUE16-NEXT: v_bfe_u32 v18, v17, 16, 1
-; GFX11-TRUE16-NEXT: v_or_b32_e32 v22, 0x400000, v17
-; GFX11-TRUE16-NEXT: v_cmp_u_f32_e32 vcc_lo, v17, v17
-; GFX11-TRUE16-NEXT: v_bfe_u32 v19, v10, 16, 1
-; GFX11-TRUE16-NEXT: v_add3_u32 v18, v18, v17, 0x7fff
-; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_2) | instskip(NEXT) | instid1(VALU_DEP_2)
-; GFX11-TRUE16-NEXT: v_add3_u32 v19, v19, v10, 0x7fff
+; GFX11-TRUE16-NEXT: v_lshrrev_b32_e32 v87, 8, v8
+; GFX11-TRUE16-NEXT: v_lshrrev_b32_e32 v99, 8, v5
; GFX11-TRUE16-NEXT: v_cndmask_b32_e32 v17, v18, v22, vcc_lo
; GFX11-TRUE16-NEXT: v_cmp_u_f32_e32 vcc_lo, v20, v20
+; GFX11-TRUE16-NEXT: v_and_b32_e32 v10, 0xffff0000, v10
+; GFX11-TRUE16-NEXT: v_mov_b16_e32 v18.l, v50.h
+; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_2) | instskip(SKIP_1) | instid1(VALU_DEP_2)
+; GFX11-TRUE16-NEXT: v_dual_cndmask_b32 v39, v7, v21 :: v_dual_add_f32 v10, 0x40c00000, v10
+; GFX11-TRUE16-NEXT: v_lshlrev_b32_e32 v7, 16, v9
+; GFX11-TRUE16-NEXT: v_bfe_u32 v19, v10, 16, 1
; GFX11-TRUE16-NEXT: v_or_b32_e32 v22, 0x400000, v10
-; GFX11-TRUE16-NEXT: v_mov_b16_e32 v18.l, v49.h
-; GFX11-TRUE16-NEXT: v_cndmask_b32_e32 v38, v7, v21, vcc_lo
; GFX11-TRUE16-NEXT: v_cmp_u_f32_e32 vcc_lo, v10, v10
-; GFX11-TRUE16-NEXT: v_lshlrev_b32_e32 v10, 16, v12
-; GFX11-TRUE16-NEXT: v_and_b32_e32 v12, 0xffff0000, v12
-; GFX11-TRUE16-NEXT: v_dual_cndmask_b32 v48, v19, v22 :: v_dual_lshlrev_b32 v7, 16, v9
-; GFX11-TRUE16-NEXT: v_mov_b16_e32 v19.l, v38.h
-; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_2) | instskip(SKIP_1) | instid1(VALU_DEP_3)
-; GFX11-TRUE16-NEXT: v_dual_add_f32 v12, 0x40c00000, v12 :: v_dual_add_f32 v7, 0x40c00000, v7
+; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_4) | instskip(NEXT) | instid1(VALU_DEP_4)
+; GFX11-TRUE16-NEXT: v_add_f32_e32 v7, 0x40c00000, v7
+; GFX11-TRUE16-NEXT: v_add3_u32 v19, v19, v10, 0x7fff
+; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
+; GFX11-TRUE16-NEXT: v_dual_cndmask_b32 v35, v19, v22 :: v_dual_lshlrev_b32 v10, 16, v12
; GFX11-TRUE16-NEXT: v_add_f32_e32 v21, 0x40c00000, v10
-; GFX11-TRUE16-NEXT: v_bfi_b32 v10, 0xffff, v19, v48
-; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_3) | instskip(NEXT) | instid1(VALU_DEP_4)
-; GFX11-TRUE16-NEXT: v_bfe_u32 v24, v12, 16, 1
-; GFX11-TRUE16-NEXT: v_bfe_u32 v19, v7, 16, 1
-; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_4)
+; GFX11-TRUE16-NEXT: v_mov_b16_e32 v19.l, v39.h
+; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_2) | instskip(SKIP_2) | instid1(VALU_DEP_4)
; GFX11-TRUE16-NEXT: v_bfe_u32 v22, v21, 16, 1
; GFX11-TRUE16-NEXT: v_or_b32_e32 v37, 0x400000, v21
; GFX11-TRUE16-NEXT: v_cmp_u_f32_e32 vcc_lo, v21, v21
-; GFX11-TRUE16-NEXT: v_or_b32_e32 v25, 0x400000, v7
-; GFX11-TRUE16-NEXT: v_add3_u32 v19, v19, v7, 0x7fff
+; GFX11-TRUE16-NEXT: v_bfi_b32 v10, 0xffff, v19, v35
+; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_4) | instskip(SKIP_1) | instid1(VALU_DEP_3)
; GFX11-TRUE16-NEXT: v_add3_u32 v22, v22, v21, 0x7fff
-; GFX11-TRUE16-NEXT: v_add_f32_e32 v9, 0x40c00000, v23
-; GFX11-TRUE16-NEXT: v_add3_u32 v24, v24, v12, 0x7fff
-; GFX11-TRUE16-NEXT: v_or_b32_e32 v50, 0x400000, v12
-; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_4)
-; GFX11-TRUE16-NEXT: v_dual_add_f32 v21, 0x40c00000, v39 :: v_dual_cndmask_b32 v52, v22, v37
+; GFX11-TRUE16-NEXT: v_and_b32_e32 v12, 0xffff0000, v12
+; GFX11-TRUE16-NEXT: v_lshrrev_b32_e32 v83, 24, v10
+; GFX11-TRUE16-NEXT: v_lshrrev_b32_e32 v84, 8, v10
+; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_4) | instskip(NEXT) | instid1(VALU_DEP_4)
+; GFX11-TRUE16-NEXT: v_cndmask_b32_e32 v49, v22, v37, vcc_lo
+; GFX11-TRUE16-NEXT: v_dual_add_f32 v9, 0x40c00000, v23 :: v_dual_add_f32 v12, 0x40c00000, v12
+; GFX11-TRUE16-NEXT: v_and_b32_e32 v22, 0xffff0000, v14
+; GFX11-TRUE16-NEXT: v_lshlrev_b32_e32 v14, 16, v14
+; GFX11-TRUE16-NEXT: v_bfe_u32 v19, v7, 16, 1
+; GFX11-TRUE16-NEXT: v_or_b32_e32 v25, 0x400000, v7
+; GFX11-TRUE16-NEXT: v_bfe_u32 v24, v12, 16, 1
; GFX11-TRUE16-NEXT: v_cmp_u_f32_e32 vcc_lo, v7, v7
; GFX11-TRUE16-NEXT: v_bfe_u32 v20, v9, 16, 1
-; GFX11-TRUE16-NEXT: v_or_b32_e32 v23, 0x400000, v9
-; GFX11-TRUE16-NEXT: v_and_b32_e32 v22, 0xffff0000, v14
-; GFX11-TRUE16-NEXT: v_mov_b16_e32 v7.l, v52.h
+; GFX11-TRUE16-NEXT: v_add3_u32 v19, v19, v7, 0x7fff
+; GFX11-TRUE16-NEXT: v_and_b32_e32 v48, 0xffff0000, v11
+; GFX11-TRUE16-NEXT: v_add3_u32 v24, v24, v12, 0x7fff
+; GFX11-TRUE16-NEXT: v_or_b32_e32 v51, 0x400000, v12
+; GFX11-TRUE16-NEXT: v_add3_u32 v20, v20, v9, 0x7fff
; GFX11-TRUE16-NEXT: v_cndmask_b32_e32 v65, v19, v25, vcc_lo
+; GFX11-TRUE16-NEXT: v_dual_add_f32 v21, 0x40c00000, v48 :: v_dual_add_f32 v14, 0x40c00000, v14
; GFX11-TRUE16-NEXT: v_cmp_u_f32_e32 vcc_lo, v12, v12
-; GFX11-TRUE16-NEXT: v_add3_u32 v20, v20, v9, 0x7fff
-; GFX11-TRUE16-NEXT: v_lshlrev_b32_e32 v14, 16, v14
+; GFX11-TRUE16-NEXT: v_or_b32_e32 v23, 0x400000, v9
+; GFX11-TRUE16-NEXT: v_mov_b16_e32 v7.l, v49.h
+; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_4) | instskip(SKIP_4) | instid1(VALU_DEP_3)
; GFX11-TRUE16-NEXT: v_bfe_u32 v19, v21, 16, 1
; GFX11-TRUE16-NEXT: v_lshlrev_b32_e32 v11, 16, v11
-; GFX11-TRUE16-NEXT: v_cndmask_b32_e32 v53, v24, v50, vcc_lo
+; GFX11-TRUE16-NEXT: v_cndmask_b32_e32 v37, v24, v51, vcc_lo
; GFX11-TRUE16-NEXT: v_cmp_u_f32_e32 vcc_lo, v9, v9
-; GFX11-TRUE16-NEXT: v_add_f32_e32 v14, 0x40c00000, v14
-; GFX11-TRUE16-NEXT: v_lshrrev_b32_e32 v71, 24, v10
-; GFX11-TRUE16-NEXT: v_lshrrev_b32_e32 v80, 8, v10
-; GFX11-TRUE16-NEXT: v_bfi_b32 v12, 0xffff, v7, v53
+; GFX11-TRUE16-NEXT: v_bfe_u32 v25, v14, 16, 1
+; GFX11-TRUE16-NEXT: v_bfi_b32 v12, 0xffff, v7, v37
; GFX11-TRUE16-NEXT: v_cndmask_b32_e32 v9, v20, v23, vcc_lo
; GFX11-TRUE16-NEXT: v_add_f32_e32 v7, 0x40c00000, v11
; GFX11-TRUE16-NEXT: v_add3_u32 v11, v19, v21, 0x7fff
; GFX11-TRUE16-NEXT: v_or_b32_e32 v19, 0x400000, v21
; GFX11-TRUE16-NEXT: v_cmp_u_f32_e32 vcc_lo, v21, v21
-; GFX11-TRUE16-NEXT: v_bfe_u32 v25, v14, 16, 1
-; GFX11-TRUE16-NEXT: v_bfe_u32 v23, v7, 16, 1
; GFX11-TRUE16-NEXT: v_mov_b16_e32 v20.l, v65.h
-; GFX11-TRUE16-NEXT: v_lshrrev_b32_e32 v64, 24, v12
+; GFX11-TRUE16-NEXT: v_bfe_u32 v23, v7, 16, 1
+; GFX11-TRUE16-NEXT: v_lshrrev_b32_e32 v70, 24, v12
+; GFX11-TRUE16-NEXT: v_lshrrev_b32_e32 v80, 8, v12
; GFX11-TRUE16-NEXT: v_cndmask_b32_e32 v11, v11, v19, vcc_lo
; GFX11-TRUE16-NEXT: v_add_f32_e32 v19, 0x40c00000, v22
; GFX11-TRUE16-NEXT: v_add3_u32 v21, v23, v7, 0x7fff
@@ -88906,28 +88822,27 @@ define <64 x i8> @bitcast_v32bf16_to_v64i8(<32 x bfloat> %a, i32 %b) {
; GFX11-TRUE16-NEXT: v_cmp_u_f32_e32 vcc_lo, v7, v7
; GFX11-TRUE16-NEXT: v_bfe_u32 v24, v19, 16, 1
; GFX11-TRUE16-NEXT: v_lshlrev_b32_e32 v13, 16, v13
-; GFX11-TRUE16-NEXT: v_bfi_b32 v9, 0xffff, v20, v9
-; GFX11-TRUE16-NEXT: v_dual_add_f32 v7, 0x40c00000, v23 :: v_dual_cndmask_b32 v70, v21, v22
-; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_4)
+; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_4) | instskip(NEXT) | instid1(VALU_DEP_3)
+; GFX11-TRUE16-NEXT: v_dual_add_f32 v7, 0x40c00000, v23 :: v_dual_cndmask_b32 v68, v21, v22
; GFX11-TRUE16-NEXT: v_add3_u32 v21, v24, v19, 0x7fff
; GFX11-TRUE16-NEXT: v_or_b32_e32 v22, 0x400000, v19
; GFX11-TRUE16-NEXT: v_cmp_u_f32_e32 vcc_lo, v19, v19
; GFX11-TRUE16-NEXT: v_add3_u32 v23, v25, v14, 0x7fff
; GFX11-TRUE16-NEXT: v_or_b32_e32 v24, 0x400000, v14
; GFX11-TRUE16-NEXT: v_bfe_u32 v25, v7, 16, 1
-; GFX11-TRUE16-NEXT: v_add_f32_e32 v13, 0x40c00000, v13
-; GFX11-TRUE16-NEXT: v_cndmask_b32_e32 v67, v21, v22, vcc_lo
+; GFX11-TRUE16-NEXT: v_dual_add_f32 v13, 0x40c00000, v13 :: v_dual_cndmask_b32 v48, v21, v22
; GFX11-TRUE16-NEXT: v_cmp_u_f32_e32 vcc_lo, v14, v14
; GFX11-TRUE16-NEXT: v_or_b32_e32 v19, 0x400000, v7
+; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_4)
; GFX11-TRUE16-NEXT: v_add3_u32 v14, v25, v7, 0x7fff
; GFX11-TRUE16-NEXT: v_and_b32_e32 v21, 0xffff0000, v16
; GFX11-TRUE16-NEXT: v_lshlrev_b32_e32 v16, 16, v16
-; GFX11-TRUE16-NEXT: v_cndmask_b32_e32 v68, v23, v24, vcc_lo
+; GFX11-TRUE16-NEXT: v_cndmask_b32_e32 v64, v23, v24, vcc_lo
; GFX11-TRUE16-NEXT: v_cmp_u_f32_e32 vcc_lo, v7, v7
; GFX11-TRUE16-NEXT: v_bfe_u32 v23, v13, 16, 1
-; GFX11-TRUE16-NEXT: v_lshrrev_b32_e32 v66, 8, v12
+; GFX11-TRUE16-NEXT: v_bfi_b32 v9, 0xffff, v20, v9
; GFX11-TRUE16-NEXT: v_add_f32_e32 v16, 0x40c00000, v16
-; GFX11-TRUE16-NEXT: v_mov_b16_e32 v22.l, v68.h
+; GFX11-TRUE16-NEXT: v_mov_b16_e32 v22.l, v64.h
; GFX11-TRUE16-NEXT: v_dual_cndmask_b32 v7, v14, v19 :: v_dual_add_f32 v14, 0x40c00000, v21
; GFX11-TRUE16-NEXT: v_lshlrev_b32_e32 v21, 16, v15
; GFX11-TRUE16-NEXT: v_add3_u32 v19, v23, v13, 0x7fff
@@ -88937,42 +88852,42 @@ define <64 x i8> @bitcast_v32bf16_to_v64i8(<32 x bfloat> %a, i32 %b) {
; GFX11-TRUE16-NEXT: v_add_f32_e32 v21, 0x40c00000, v21
; GFX11-TRUE16-NEXT: v_and_b32_e32 v15, 0xffff0000, v15
; GFX11-TRUE16-NEXT: v_or_b32_e32 v25, 0x400000, v16
-; GFX11-TRUE16-NEXT: v_cndmask_b32_e32 v85, v19, v23, vcc_lo
+; GFX11-TRUE16-NEXT: v_cndmask_b32_e32 v81, v19, v23, vcc_lo
; GFX11-TRUE16-NEXT: v_add3_u32 v13, v13, v16, 0x7fff
; GFX11-TRUE16-NEXT: v_bfe_u32 v23, v21, 16, 1
; GFX11-TRUE16-NEXT: v_cmp_u_f32_e32 vcc_lo, v16, v16
; GFX11-TRUE16-NEXT: v_bfe_u32 v24, v14, 16, 1
; GFX11-TRUE16-NEXT: v_add_f32_e32 v15, 0x40c00000, v15
-; GFX11-TRUE16-NEXT: v_or_b32_e32 v39, 0x400000, v21
+; GFX11-TRUE16-NEXT: v_or_b32_e32 v52, 0x400000, v21
; GFX11-TRUE16-NEXT: v_add3_u32 v23, v23, v21, 0x7fff
-; GFX11-TRUE16-NEXT: v_cndmask_b32_e32 v81, v13, v25, vcc_lo
+; GFX11-TRUE16-NEXT: v_cndmask_b32_e32 v69, v13, v25, vcc_lo
; GFX11-TRUE16-NEXT: v_cmp_u_f32_e32 vcc_lo, v21, v21
; GFX11-TRUE16-NEXT: v_add3_u32 v19, v24, v14, 0x7fff
; GFX11-TRUE16-NEXT: v_or_b32_e32 v24, 0x400000, v14
-; GFX11-TRUE16-NEXT: v_bfe_u32 v37, v15, 16, 1
+; GFX11-TRUE16-NEXT: v_bfe_u32 v51, v15, 16, 1
; GFX11-TRUE16-NEXT: v_or_b32_e32 v16, 0x400000, v15
-; GFX11-TRUE16-NEXT: v_cndmask_b32_e32 v84, v23, v39, vcc_lo
+; GFX11-TRUE16-NEXT: v_cndmask_b32_e32 v71, v23, v52, vcc_lo
; GFX11-TRUE16-NEXT: v_cmp_u_f32_e32 vcc_lo, v14, v14
-; GFX11-TRUE16-NEXT: v_mov_b16_e32 v21.l, v85.h
-; GFX11-TRUE16-NEXT: v_add3_u32 v13, v37, v15, 0x7fff
-; GFX11-TRUE16-NEXT: v_mov_b16_e32 v23.l, v70.h
-; GFX11-TRUE16-NEXT: v_bfi_b32 v14, 0xffff, v22, v67
-; GFX11-TRUE16-NEXT: v_cndmask_b32_e32 v82, v19, v24, vcc_lo
+; GFX11-TRUE16-NEXT: v_mov_b16_e32 v21.l, v81.h
+; GFX11-TRUE16-NEXT: v_add3_u32 v13, v51, v15, 0x7fff
+; GFX11-TRUE16-NEXT: v_mov_b16_e32 v23.l, v68.h
+; GFX11-TRUE16-NEXT: v_bfi_b32 v14, 0xffff, v22, v48
+; GFX11-TRUE16-NEXT: v_cndmask_b32_e32 v54, v19, v24, vcc_lo
; GFX11-TRUE16-NEXT: v_cmp_u_f32_e32 vcc_lo, v15, v15
-; GFX11-TRUE16-NEXT: v_mov_b16_e32 v19.l, v81.h
-; GFX11-TRUE16-NEXT: v_mov_b16_e32 v15.l, v84.h
+; GFX11-TRUE16-NEXT: v_mov_b16_e32 v19.l, v69.h
+; GFX11-TRUE16-NEXT: v_mov_b16_e32 v15.l, v71.h
; GFX11-TRUE16-NEXT: v_bfi_b32 v11, 0xffff, v23, v11
-; GFX11-TRUE16-NEXT: v_lshrrev_b32_e32 v51, 24, v14
+; GFX11-TRUE16-NEXT: v_lshrrev_b32_e32 v55, 24, v14
; GFX11-TRUE16-NEXT: v_cndmask_b32_e32 v13, v13, v16, vcc_lo
-; GFX11-TRUE16-NEXT: v_bfi_b32 v16, 0xffff, v19, v82
-; GFX11-TRUE16-NEXT: v_lshrrev_b32_e32 v54, 8, v14
-; GFX11-TRUE16-NEXT: v_lshrrev_b32_e32 v69, 8, v11
-; GFX11-TRUE16-NEXT: v_lshrrev_b32_e32 v83, 8, v9
+; GFX11-TRUE16-NEXT: v_bfi_b32 v16, 0xffff, v19, v54
+; GFX11-TRUE16-NEXT: v_lshrrev_b32_e32 v66, 8, v14
+; GFX11-TRUE16-NEXT: v_lshrrev_b32_e32 v82, 8, v11
+; GFX11-TRUE16-NEXT: v_lshrrev_b32_e32 v85, 8, v9
; GFX11-TRUE16-NEXT: v_bfi_b32 v15, 0xffff, v15, v13
; GFX11-TRUE16-NEXT: v_bfi_b32 v13, 0xffff, v21, v7
; GFX11-TRUE16-NEXT: v_bfi_b32 v7, 0xffff, v18, v17
-; GFX11-TRUE16-NEXT: v_lshrrev_b32_e32 v37, 24, v16
-; GFX11-TRUE16-NEXT: v_lshrrev_b32_e32 v39, 8, v16
+; GFX11-TRUE16-NEXT: v_lshrrev_b32_e32 v51, 24, v16
+; GFX11-TRUE16-NEXT: v_lshrrev_b32_e32 v52, 8, v16
; GFX11-TRUE16-NEXT: v_lshrrev_b64 v[17:18], 24, v[15:16]
; GFX11-TRUE16-NEXT: v_lshrrev_b64 v[18:19], 24, v[13:14]
; GFX11-TRUE16-NEXT: v_lshrrev_b64 v[19:20], 24, v[11:12]
@@ -88981,21 +88896,22 @@ define <64 x i8> @bitcast_v32bf16_to_v64i8(<32 x bfloat> %a, i32 %b) {
; GFX11-TRUE16-NEXT: v_lshrrev_b64 v[22:23], 24, v[5:6]
; GFX11-TRUE16-NEXT: v_lshrrev_b64 v[23:24], 24, v[3:4]
; GFX11-TRUE16-NEXT: v_lshrrev_b64 v[24:25], 24, v[1:2]
-; GFX11-TRUE16-NEXT: v_lshrrev_b32_e32 v50, 8, v15
-; GFX11-TRUE16-NEXT: v_lshrrev_b32_e32 v55, 8, v13
+; GFX11-TRUE16-NEXT: v_lshrrev_b32_e32 v53, 8, v15
+; GFX11-TRUE16-NEXT: v_lshrrev_b32_e32 v67, 8, v13
; GFX11-TRUE16-NEXT: v_lshrrev_b32_e32 v96, 8, v7
; GFX11-TRUE16-NEXT: .LBB108_4: ; %end
; GFX11-TRUE16-NEXT: s_or_b32 exec_lo, exec_lo, s0
-; GFX11-TRUE16-NEXT: v_and_b16 v1.l, 0xff, v28.h
+; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1)
+; GFX11-TRUE16-NEXT: v_and_b16 v1.l, 0xff, v31.h
; GFX11-TRUE16-NEXT: v_lshlrev_b16 v2.l, 8, v113.l
; GFX11-TRUE16-NEXT: v_and_b16 v1.h, 0xff, v1.h
; GFX11-TRUE16-NEXT: v_lshlrev_b16 v2.h, 8, v24.l
; GFX11-TRUE16-NEXT: v_lshlrev_b16 v3.l, 8, v112.l
-; GFX11-TRUE16-NEXT: v_and_b16 v4.l, 0xff, v27.h
+; GFX11-TRUE16-NEXT: v_and_b16 v4.l, 0xff, v26.h
; GFX11-TRUE16-NEXT: v_or_b16 v1.l, v1.l, v2.l
-; GFX11-TRUE16-NEXT: v_and_b16 v4.h, 0xff, v31.h
+; GFX11-TRUE16-NEXT: v_and_b16 v4.h, 0xff, v34.h
; GFX11-TRUE16-NEXT: v_or_b16 v2.l, v1.h, v2.h
-; GFX11-TRUE16-NEXT: v_and_b16 v2.h, 0xff, v26.h
+; GFX11-TRUE16-NEXT: v_and_b16 v2.h, 0xff, v28.h
; GFX11-TRUE16-NEXT: v_lshlrev_b16 v5.l, 8, v102.l
; GFX11-TRUE16-NEXT: v_and_b16 v3.h, 0xff, v3.h
; GFX11-TRUE16-NEXT: v_and_b32_e32 v1, 0xffff, v1
@@ -89011,13 +88927,13 @@ define <64 x i8> @bitcast_v32bf16_to_v64i8(<32 x bfloat> %a, i32 %b) {
; GFX11-TRUE16-NEXT: v_and_b32_e32 v10, 0xffff, v3
; GFX11-TRUE16-NEXT: v_or_b32_e32 v1, v1, v8
; GFX11-TRUE16-NEXT: v_or_b16 v5.l, v3.h, v6.l
-; GFX11-TRUE16-NEXT: v_and_b16 v3.l, 0xff, v35.h
+; GFX11-TRUE16-NEXT: v_and_b16 v3.l, 0xff, v38.h
; GFX11-TRUE16-NEXT: v_lshlrev_b16 v3.h, 8, v99.l
; GFX11-TRUE16-NEXT: v_and_b16 v6.h, 0xff, v30.h
; GFX11-TRUE16-NEXT: v_lshlrev_b32_e32 v12, 16, v2
; GFX11-TRUE16-NEXT: v_mov_b16_e32 v8.l, v5.l
; GFX11-TRUE16-NEXT: v_and_b32_e32 v14, 0xffff, v4
-; GFX11-TRUE16-NEXT: v_and_b16 v2.l, 0xff, v29.h
+; GFX11-TRUE16-NEXT: v_and_b16 v2.l, 0xff, v27.h
; GFX11-TRUE16-NEXT: v_lshlrev_b16 v2.h, 8, v100.l
; GFX11-TRUE16-NEXT: v_or_b16 v3.l, v3.l, v3.h
; GFX11-TRUE16-NEXT: v_lshlrev_b32_e32 v8, 16, v8
@@ -89027,19 +88943,19 @@ define <64 x i8> @bitcast_v32bf16_to_v64i8(<32 x bfloat> %a, i32 %b) {
; GFX11-TRUE16-NEXT: v_or_b16 v2.l, v2.l, v2.h
; GFX11-TRUE16-NEXT: v_and_b32_e32 v23, 0xffff, v3
; GFX11-TRUE16-NEXT: v_or_b32_e32 v3, v14, v8
-; GFX11-TRUE16-NEXT: v_and_b16 v7.l, 0xff, v49.h
+; GFX11-TRUE16-NEXT: v_and_b16 v7.l, 0xff, v50.h
; GFX11-TRUE16-NEXT: v_lshlrev_b16 v8.l, 8, v96.l
; GFX11-TRUE16-NEXT: v_and_b32_e32 v16, 0xffff, v6
; GFX11-TRUE16-NEXT: v_or_b16 v4.l, v4.l, v4.h
; GFX11-TRUE16-NEXT: v_lshlrev_b32_e32 v22, 16, v2
; GFX11-TRUE16-NEXT: v_or_b32_e32 v2, v10, v12
; GFX11-TRUE16-NEXT: v_lshlrev_b16 v8.h, 8, v21.l
-; GFX11-TRUE16-NEXT: v_and_b16 v9.l, 0xff, v34.h
+; GFX11-TRUE16-NEXT: v_and_b16 v9.l, 0xff, v36.h
; GFX11-TRUE16-NEXT: v_lshlrev_b16 v10.l, 8, v87.l
-; GFX11-TRUE16-NEXT: v_and_b16 v10.h, 0xff, v36.h
+; GFX11-TRUE16-NEXT: v_and_b16 v10.h, 0xff, v32.h
; GFX11-TRUE16-NEXT: v_lshlrev_b16 v11.l, 8, v86.l
; GFX11-TRUE16-NEXT: v_and_b16 v12.l, 0xff, v65.h
-; GFX11-TRUE16-NEXT: v_lshlrev_b16 v12.h, 8, v83.l
+; GFX11-TRUE16-NEXT: v_lshlrev_b16 v12.h, 8, v85.l
; GFX11-TRUE16-NEXT: v_or_b16 v7.l, v7.l, v8.l
; GFX11-TRUE16-NEXT: v_lshlrev_b32_e32 v24, 16, v4
; GFX11-TRUE16-NEXT: v_or_b32_e32 v4, v16, v22
@@ -89050,15 +88966,15 @@ define <64 x i8> @bitcast_v32bf16_to_v64i8(<32 x bfloat> %a, i32 %b) {
; GFX11-TRUE16-NEXT: v_and_b32_e32 v16, 0xffff, v7
; GFX11-TRUE16-NEXT: v_and_b16 v7.l, 0xff, v9.h
; GFX11-TRUE16-NEXT: v_lshlrev_b16 v7.h, 8, v20.l
-; GFX11-TRUE16-NEXT: v_and_b16 v5.l, 0xff, v32.h
+; GFX11-TRUE16-NEXT: v_and_b16 v5.l, 0xff, v33.h
; GFX11-TRUE16-NEXT: v_lshlrev_b16 v5.h, 8, v98.l
; GFX11-TRUE16-NEXT: v_mov_b16_e32 v12.l, v9.l
; GFX11-TRUE16-NEXT: v_mov_b16_e32 v14.l, v11.l
; GFX11-TRUE16-NEXT: v_lshlrev_b32_e32 v21, 16, v8
-; GFX11-TRUE16-NEXT: v_and_b16 v8.l, 0xff, v38.h
-; GFX11-TRUE16-NEXT: v_lshlrev_b16 v8.h, 8, v80.l
-; GFX11-TRUE16-NEXT: v_and_b16 v9.l, 0xff, v48.h
-; GFX11-TRUE16-NEXT: v_lshlrev_b16 v9.h, 8, v71.l
+; GFX11-TRUE16-NEXT: v_and_b16 v8.l, 0xff, v39.h
+; GFX11-TRUE16-NEXT: v_lshlrev_b16 v8.h, 8, v84.l
+; GFX11-TRUE16-NEXT: v_and_b16 v9.l, 0xff, v35.h
+; GFX11-TRUE16-NEXT: v_lshlrev_b16 v9.h, 8, v83.l
; GFX11-TRUE16-NEXT: v_or_b16 v7.l, v7.l, v7.h
; GFX11-TRUE16-NEXT: v_or_b16 v5.l, v5.l, v5.h
; GFX11-TRUE16-NEXT: v_and_b32_e32 v12, 0xffff, v12
@@ -89071,24 +88987,24 @@ define <64 x i8> @bitcast_v32bf16_to_v64i8(<32 x bfloat> %a, i32 %b) {
; GFX11-TRUE16-NEXT: v_lshlrev_b32_e32 v19, 16, v7
; GFX11-TRUE16-NEXT: v_and_b32_e32 v25, 0xffff, v5
; GFX11-TRUE16-NEXT: v_or_b32_e32 v5, v23, v24
-; GFX11-TRUE16-NEXT: v_and_b16 v10.l, 0xff, v70.h
-; GFX11-TRUE16-NEXT: v_lshlrev_b16 v10.h, 8, v69.l
+; GFX11-TRUE16-NEXT: v_and_b16 v10.l, 0xff, v68.h
+; GFX11-TRUE16-NEXT: v_lshlrev_b16 v10.h, 8, v82.l
; GFX11-TRUE16-NEXT: v_and_b32_e32 v20, 0xffff, v8
; GFX11-TRUE16-NEXT: v_lshlrev_b32_e32 v23, 16, v9
; GFX11-TRUE16-NEXT: v_or_b32_e32 v7, v16, v21
; GFX11-TRUE16-NEXT: v_or_b32_e32 v8, v12, v22
; GFX11-TRUE16-NEXT: v_or_b32_e32 v9, v14, v19
-; GFX11-TRUE16-NEXT: v_and_b16 v12.l, 0xff, v52.h
-; GFX11-TRUE16-NEXT: v_lshlrev_b16 v12.h, 8, v66.l
-; GFX11-TRUE16-NEXT: v_and_b16 v13.l, 0xff, v53.h
-; GFX11-TRUE16-NEXT: v_lshlrev_b16 v14.l, 8, v64.l
-; GFX11-TRUE16-NEXT: v_and_b16 v14.h, 0xff, v85.h
-; GFX11-TRUE16-NEXT: v_lshlrev_b16 v15.l, 8, v55.l
+; GFX11-TRUE16-NEXT: v_and_b16 v12.l, 0xff, v49.h
+; GFX11-TRUE16-NEXT: v_lshlrev_b16 v12.h, 8, v80.l
+; GFX11-TRUE16-NEXT: v_and_b16 v13.l, 0xff, v37.h
+; GFX11-TRUE16-NEXT: v_lshlrev_b16 v14.l, 8, v70.l
+; GFX11-TRUE16-NEXT: v_and_b16 v14.h, 0xff, v81.h
+; GFX11-TRUE16-NEXT: v_lshlrev_b16 v15.l, 8, v67.l
; GFX11-TRUE16-NEXT: v_and_b16 v13.h, 0xff, v13.h
; GFX11-TRUE16-NEXT: v_lshlrev_b16 v16.l, 8, v18.l
-; GFX11-TRUE16-NEXT: v_and_b16 v16.h, 0xff, v68.h
-; GFX11-TRUE16-NEXT: v_lshlrev_b16 v17.h, 8, v54.l
-; GFX11-TRUE16-NEXT: v_and_b16 v6.l, 0xff, v33.h
+; GFX11-TRUE16-NEXT: v_and_b16 v16.h, 0xff, v64.h
+; GFX11-TRUE16-NEXT: v_lshlrev_b16 v17.h, 8, v66.l
+; GFX11-TRUE16-NEXT: v_and_b16 v6.l, 0xff, v29.h
; GFX11-TRUE16-NEXT: v_lshlrev_b16 v6.h, 8, v97.l
; GFX11-TRUE16-NEXT: v_or_b16 v10.l, v10.l, v10.h
; GFX11-TRUE16-NEXT: v_or_b16 v12.l, v12.l, v12.h
@@ -89105,16 +89021,16 @@ define <64 x i8> @bitcast_v32bf16_to_v64i8(<32 x bfloat> %a, i32 %b) {
; GFX11-TRUE16-NEXT: v_lshlrev_b32_e32 v20, 16, v13
; GFX11-TRUE16-NEXT: v_and_b32_e32 v21, 0xffff, v14
; GFX11-TRUE16-NEXT: v_and_b32_e32 v22, 0xffff, v16
-; GFX11-TRUE16-NEXT: v_and_b16 v12.l, 0xff, v67.h
-; GFX11-TRUE16-NEXT: v_lshlrev_b16 v12.h, 8, v51.l
-; GFX11-TRUE16-NEXT: v_and_b16 v13.l, 0xff, v84.h
-; GFX11-TRUE16-NEXT: v_lshlrev_b16 v13.h, 8, v50.l
+; GFX11-TRUE16-NEXT: v_and_b16 v12.l, 0xff, v48.h
+; GFX11-TRUE16-NEXT: v_lshlrev_b16 v12.h, 8, v55.l
+; GFX11-TRUE16-NEXT: v_and_b16 v13.l, 0xff, v71.h
+; GFX11-TRUE16-NEXT: v_lshlrev_b16 v13.h, 8, v53.l
; GFX11-TRUE16-NEXT: v_and_b16 v14.l, 0xff, v15.h
; GFX11-TRUE16-NEXT: v_lshlrev_b16 v14.h, 8, v17.l
-; GFX11-TRUE16-NEXT: v_and_b16 v15.l, 0xff, v81.h
-; GFX11-TRUE16-NEXT: v_lshlrev_b16 v15.h, 8, v39.l
-; GFX11-TRUE16-NEXT: v_and_b16 v16.l, 0xff, v82.h
-; GFX11-TRUE16-NEXT: v_lshlrev_b16 v16.h, 8, v37.l
+; GFX11-TRUE16-NEXT: v_and_b16 v15.l, 0xff, v69.h
+; GFX11-TRUE16-NEXT: v_lshlrev_b16 v15.h, 8, v52.l
+; GFX11-TRUE16-NEXT: v_and_b16 v16.l, 0xff, v54.h
+; GFX11-TRUE16-NEXT: v_lshlrev_b16 v16.h, 8, v51.l
; GFX11-TRUE16-NEXT: v_lshlrev_b32_e32 v6, 16, v6
; GFX11-TRUE16-NEXT: v_lshlrev_b32_e32 v11, 16, v11
; GFX11-TRUE16-NEXT: v_or_b16 v12.l, v12.l, v12.h
@@ -89146,107 +89062,107 @@ define <64 x i8> @bitcast_v32bf16_to_v64i8(<32 x bfloat> %a, i32 %b) {
; GFX11-FAKE16: ; %bb.0:
; GFX11-FAKE16-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX11-FAKE16-NEXT: v_cmp_ne_u32_e32 vcc_lo, 0, v17
+; GFX11-FAKE16-NEXT: ; implicit-def: $vgpr19
+; GFX11-FAKE16-NEXT: ; implicit-def: $vgpr99
+; GFX11-FAKE16-NEXT: ; implicit-def: $vgpr30
+; GFX11-FAKE16-NEXT: ; implicit-def: $vgpr21
+; GFX11-FAKE16-NEXT: ; implicit-def: $vgpr83
+; GFX11-FAKE16-NEXT: ; implicit-def: $vgpr98
; GFX11-FAKE16-NEXT: ; implicit-def: $vgpr25
; GFX11-FAKE16-NEXT: ; implicit-def: $vgpr97
-; GFX11-FAKE16-NEXT: ; implicit-def: $vgpr24
-; GFX11-FAKE16-NEXT: ; implicit-def: $vgpr27
-; GFX11-FAKE16-NEXT: ; implicit-def: $vgpr81
+; GFX11-FAKE16-NEXT: ; implicit-def: $vgpr29
+; GFX11-FAKE16-NEXT: ; implicit-def: $vgpr26
+; GFX11-FAKE16-NEXT: ; implicit-def: $vgpr70
; GFX11-FAKE16-NEXT: ; implicit-def: $vgpr96
-; GFX11-FAKE16-NEXT: ; implicit-def: $vgpr28
; GFX11-FAKE16-NEXT: ; implicit-def: $vgpr87
-; GFX11-FAKE16-NEXT: ; implicit-def: $vgpr23
-; GFX11-FAKE16-NEXT: ; implicit-def: $vgpr29
-; GFX11-FAKE16-NEXT: ; implicit-def: $vgpr68
; GFX11-FAKE16-NEXT: ; implicit-def: $vgpr86
+; GFX11-FAKE16-NEXT: ; implicit-def: $vgpr28
; GFX11-FAKE16-NEXT: ; implicit-def: $vgpr85
+; GFX11-FAKE16-NEXT: ; implicit-def: $vgpr67
; GFX11-FAKE16-NEXT: ; implicit-def: $vgpr84
-; GFX11-FAKE16-NEXT: ; implicit-def: $vgpr22
-; GFX11-FAKE16-NEXT: ; implicit-def: $vgpr83
-; GFX11-FAKE16-NEXT: ; implicit-def: $vgpr65
; GFX11-FAKE16-NEXT: ; implicit-def: $vgpr82
+; GFX11-FAKE16-NEXT: ; implicit-def: $vgpr81
+; GFX11-FAKE16-NEXT: ; implicit-def: $vgpr27
; GFX11-FAKE16-NEXT: ; implicit-def: $vgpr80
+; GFX11-FAKE16-NEXT: ; implicit-def: $vgpr54
; GFX11-FAKE16-NEXT: ; implicit-def: $vgpr71
-; GFX11-FAKE16-NEXT: ; implicit-def: $vgpr21
-; GFX11-FAKE16-NEXT: ; implicit-def: $vgpr70
-; GFX11-FAKE16-NEXT: ; implicit-def: $vgpr52
; GFX11-FAKE16-NEXT: ; implicit-def: $vgpr69
-; GFX11-FAKE16-NEXT: ; implicit-def: $vgpr67
+; GFX11-FAKE16-NEXT: ; implicit-def: $vgpr68
+; GFX11-FAKE16-NEXT: ; implicit-def: $vgpr23
; GFX11-FAKE16-NEXT: ; implicit-def: $vgpr66
-; GFX11-FAKE16-NEXT: ; implicit-def: $vgpr20
+; GFX11-FAKE16-NEXT: ; implicit-def: $vgpr48
+; GFX11-FAKE16-NEXT: ; implicit-def: $vgpr65
; GFX11-FAKE16-NEXT: ; implicit-def: $vgpr64
-; GFX11-FAKE16-NEXT: ; implicit-def: $vgpr38
; GFX11-FAKE16-NEXT: ; implicit-def: $vgpr55
-; GFX11-FAKE16-NEXT: ; implicit-def: $vgpr54
+; GFX11-FAKE16-NEXT: ; implicit-def: $vgpr22
; GFX11-FAKE16-NEXT: ; implicit-def: $vgpr53
-; GFX11-FAKE16-NEXT: ; implicit-def: $vgpr19
+; GFX11-FAKE16-NEXT: ; implicit-def: $vgpr35
+; GFX11-FAKE16-NEXT: ; implicit-def: $vgpr52
; GFX11-FAKE16-NEXT: ; implicit-def: $vgpr51
-; GFX11-FAKE16-NEXT: ; implicit-def: $vgpr33
; GFX11-FAKE16-NEXT: ; implicit-def: $vgpr50
-; GFX11-FAKE16-NEXT: ; implicit-def: $vgpr49
-; GFX11-FAKE16-NEXT: ; implicit-def: $vgpr48
; GFX11-FAKE16-NEXT: ; implicit-def: $vgpr18
+; GFX11-FAKE16-NEXT: ; implicit-def: $vgpr49
+; GFX11-FAKE16-NEXT: ; implicit-def: $vgpr33
; GFX11-FAKE16-NEXT: ; implicit-def: $vgpr39
-; GFX11-FAKE16-NEXT: ; implicit-def: $vgpr31
+; GFX11-FAKE16-NEXT: ; implicit-def: $vgpr38
; GFX11-FAKE16-NEXT: ; implicit-def: $vgpr37
-; GFX11-FAKE16-NEXT: ; implicit-def: $vgpr36
-; GFX11-FAKE16-NEXT: ; implicit-def: $vgpr35
; GFX11-FAKE16-NEXT: ; implicit-def: $vgpr17
-; GFX11-FAKE16-NEXT: ; implicit-def: $vgpr34
-; GFX11-FAKE16-NEXT: ; implicit-def: $vgpr30
+; GFX11-FAKE16-NEXT: ; implicit-def: $vgpr36
; GFX11-FAKE16-NEXT: ; implicit-def: $vgpr32
+; GFX11-FAKE16-NEXT: ; implicit-def: $vgpr34
; GFX11-FAKE16-NEXT: s_and_saveexec_b32 s0, vcc_lo
; GFX11-FAKE16-NEXT: s_delay_alu instid0(SALU_CYCLE_1)
; GFX11-FAKE16-NEXT: s_xor_b32 s0, exec_lo, s0
; GFX11-FAKE16-NEXT: s_cbranch_execz .LBB108_2
; GFX11-FAKE16-NEXT: ; %bb.1: ; %cmp.false
+; GFX11-FAKE16-NEXT: v_lshrrev_b64 v[27:28], 24, v[7:8]
; GFX11-FAKE16-NEXT: v_lshrrev_b64 v[17:18], 24, v[15:16]
+; GFX11-FAKE16-NEXT: v_lshrrev_b64 v[28:29], 24, v[5:6]
; GFX11-FAKE16-NEXT: v_lshrrev_b64 v[18:19], 24, v[13:14]
-; GFX11-FAKE16-NEXT: v_lshrrev_b64 v[19:20], 24, v[11:12]
-; GFX11-FAKE16-NEXT: v_lshrrev_b64 v[20:21], 24, v[9:10]
-; GFX11-FAKE16-NEXT: v_lshrrev_b64 v[21:22], 24, v[7:8]
-; GFX11-FAKE16-NEXT: v_lshrrev_b64 v[22:23], 24, v[5:6]
-; GFX11-FAKE16-NEXT: v_lshrrev_b64 v[23:24], 24, v[3:4]
-; GFX11-FAKE16-NEXT: v_lshrrev_b64 v[24:25], 24, v[1:2]
-; GFX11-FAKE16-NEXT: v_lshrrev_b32_e32 v32, 24, v16
-; GFX11-FAKE16-NEXT: v_lshrrev_b32_e32 v30, 16, v16
-; GFX11-FAKE16-NEXT: v_lshrrev_b32_e32 v34, 8, v16
-; GFX11-FAKE16-NEXT: v_lshrrev_b32_e32 v35, 16, v15
-; GFX11-FAKE16-NEXT: v_lshrrev_b32_e32 v36, 8, v15
-; GFX11-FAKE16-NEXT: v_lshrrev_b32_e32 v37, 24, v14
-; GFX11-FAKE16-NEXT: v_lshrrev_b32_e32 v31, 16, v14
-; GFX11-FAKE16-NEXT: v_lshrrev_b32_e32 v39, 8, v14
-; GFX11-FAKE16-NEXT: v_lshrrev_b32_e32 v48, 16, v13
-; GFX11-FAKE16-NEXT: v_lshrrev_b32_e32 v49, 8, v13
-; GFX11-FAKE16-NEXT: v_lshrrev_b32_e32 v50, 24, v12
-; GFX11-FAKE16-NEXT: v_lshrrev_b32_e32 v33, 16, v12
-; GFX11-FAKE16-NEXT: v_lshrrev_b32_e32 v51, 8, v12
-; GFX11-FAKE16-NEXT: v_lshrrev_b32_e32 v53, 16, v11
-; GFX11-FAKE16-NEXT: v_lshrrev_b32_e32 v54, 8, v11
-; GFX11-FAKE16-NEXT: v_lshrrev_b32_e32 v55, 24, v10
-; GFX11-FAKE16-NEXT: v_lshrrev_b32_e32 v38, 16, v10
-; GFX11-FAKE16-NEXT: v_lshrrev_b32_e32 v64, 8, v10
-; GFX11-FAKE16-NEXT: v_lshrrev_b32_e32 v66, 16, v9
-; GFX11-FAKE16-NEXT: v_lshrrev_b32_e32 v67, 8, v9
-; GFX11-FAKE16-NEXT: v_lshrrev_b32_e32 v69, 24, v8
-; GFX11-FAKE16-NEXT: v_lshrrev_b32_e32 v52, 16, v8
-; GFX11-FAKE16-NEXT: v_lshrrev_b32_e32 v70, 8, v8
-; GFX11-FAKE16-NEXT: v_lshrrev_b32_e32 v71, 16, v7
-; GFX11-FAKE16-NEXT: v_lshrrev_b32_e32 v80, 8, v7
-; GFX11-FAKE16-NEXT: v_lshrrev_b32_e32 v82, 24, v6
-; GFX11-FAKE16-NEXT: v_lshrrev_b32_e32 v65, 16, v6
-; GFX11-FAKE16-NEXT: v_lshrrev_b32_e32 v83, 8, v6
-; GFX11-FAKE16-NEXT: v_lshrrev_b32_e32 v84, 16, v5
-; GFX11-FAKE16-NEXT: v_lshrrev_b32_e32 v85, 8, v5
-; GFX11-FAKE16-NEXT: v_lshrrev_b32_e32 v86, 24, v4
-; GFX11-FAKE16-NEXT: v_lshrrev_b32_e32 v68, 16, v4
-; GFX11-FAKE16-NEXT: v_lshrrev_b32_e32 v29, 8, v4
-; GFX11-FAKE16-NEXT: v_lshrrev_b32_e32 v87, 16, v3
-; GFX11-FAKE16-NEXT: v_lshrrev_b32_e32 v28, 8, v3
-; GFX11-FAKE16-NEXT: v_lshrrev_b32_e32 v96, 24, v2
-; GFX11-FAKE16-NEXT: v_lshrrev_b32_e32 v81, 16, v2
-; GFX11-FAKE16-NEXT: v_lshrrev_b32_e32 v27, 8, v2
-; GFX11-FAKE16-NEXT: v_lshrrev_b32_e32 v97, 16, v1
-; GFX11-FAKE16-NEXT: v_lshrrev_b32_e32 v25, 8, v1
+; GFX11-FAKE16-NEXT: v_lshrrev_b64 v[22:23], 24, v[11:12]
+; GFX11-FAKE16-NEXT: v_lshrrev_b64 v[29:30], 24, v[3:4]
+; GFX11-FAKE16-NEXT: v_lshrrev_b32_e32 v34, 24, v16
+; GFX11-FAKE16-NEXT: v_lshrrev_b32_e32 v32, 16, v16
+; GFX11-FAKE16-NEXT: v_lshrrev_b32_e32 v36, 8, v16
+; GFX11-FAKE16-NEXT: v_lshrrev_b32_e32 v37, 16, v15
+; GFX11-FAKE16-NEXT: v_lshrrev_b32_e32 v38, 8, v15
+; GFX11-FAKE16-NEXT: v_lshrrev_b32_e32 v39, 24, v14
+; GFX11-FAKE16-NEXT: v_lshrrev_b32_e32 v33, 16, v14
+; GFX11-FAKE16-NEXT: v_lshrrev_b32_e32 v49, 8, v14
+; GFX11-FAKE16-NEXT: v_lshrrev_b32_e32 v50, 16, v13
+; GFX11-FAKE16-NEXT: v_lshrrev_b32_e32 v51, 8, v13
+; GFX11-FAKE16-NEXT: v_lshrrev_b32_e32 v52, 24, v12
+; GFX11-FAKE16-NEXT: v_lshrrev_b32_e32 v35, 16, v12
+; GFX11-FAKE16-NEXT: v_lshrrev_b32_e32 v53, 8, v12
+; GFX11-FAKE16-NEXT: v_lshrrev_b32_e32 v55, 16, v11
+; GFX11-FAKE16-NEXT: v_lshrrev_b32_e32 v64, 8, v11
+; GFX11-FAKE16-NEXT: v_lshrrev_b32_e32 v65, 24, v10
+; GFX11-FAKE16-NEXT: v_lshrrev_b32_e32 v48, 16, v10
+; GFX11-FAKE16-NEXT: v_lshrrev_b32_e32 v66, 8, v10
+; GFX11-FAKE16-NEXT: v_lshrrev_b32_e32 v68, 16, v9
+; GFX11-FAKE16-NEXT: v_lshrrev_b32_e32 v69, 8, v9
+; GFX11-FAKE16-NEXT: v_lshrrev_b32_e32 v71, 24, v8
+; GFX11-FAKE16-NEXT: v_lshrrev_b32_e32 v54, 16, v8
+; GFX11-FAKE16-NEXT: v_lshrrev_b32_e32 v80, 8, v8
+; GFX11-FAKE16-NEXT: v_lshrrev_b32_e32 v81, 16, v7
+; GFX11-FAKE16-NEXT: v_lshrrev_b32_e32 v82, 8, v7
+; GFX11-FAKE16-NEXT: v_lshrrev_b32_e32 v84, 24, v6
+; GFX11-FAKE16-NEXT: v_lshrrev_b32_e32 v67, 16, v6
+; GFX11-FAKE16-NEXT: v_lshrrev_b32_e32 v85, 8, v6
+; GFX11-FAKE16-NEXT: v_lshrrev_b32_e32 v86, 16, v5
+; GFX11-FAKE16-NEXT: v_lshrrev_b32_e32 v87, 8, v5
+; GFX11-FAKE16-NEXT: v_lshrrev_b32_e32 v96, 24, v4
+; GFX11-FAKE16-NEXT: v_lshrrev_b32_e32 v70, 16, v4
+; GFX11-FAKE16-NEXT: v_lshrrev_b32_e32 v26, 8, v4
+; GFX11-FAKE16-NEXT: v_lshrrev_b32_e32 v97, 16, v3
+; GFX11-FAKE16-NEXT: v_lshrrev_b32_e32 v25, 8, v3
+; GFX11-FAKE16-NEXT: v_lshrrev_b32_e32 v98, 24, v2
+; GFX11-FAKE16-NEXT: v_lshrrev_b32_e32 v83, 16, v2
+; GFX11-FAKE16-NEXT: v_lshrrev_b32_e32 v21, 8, v2
+; GFX11-FAKE16-NEXT: v_lshrrev_b32_e32 v99, 16, v1
+; GFX11-FAKE16-NEXT: v_lshrrev_b64 v[23:24], 24, v[9:10]
+; GFX11-FAKE16-NEXT: v_lshrrev_b64 v[30:31], 24, v[1:2]
+; GFX11-FAKE16-NEXT: v_lshrrev_b32_e32 v19, 8, v1
; GFX11-FAKE16-NEXT: .LBB108_2: ; %Flow
; GFX11-FAKE16-NEXT: s_and_not1_saveexec_b32 s0, s0
; GFX11-FAKE16-NEXT: s_cbranch_execz .LBB108_4
@@ -89279,222 +89195,227 @@ define <64 x i8> @bitcast_v32bf16_to_v64i8(<32 x bfloat> %a, i32 %b) {
; GFX11-FAKE16-NEXT: v_cmp_u_f32_e32 vcc_lo, v18, v18
; GFX11-FAKE16-NEXT: v_or_b32_e32 v23, 0x400000, v19
; GFX11-FAKE16-NEXT: v_and_b32_e32 v4, 0xffff0000, v4
-; GFX11-FAKE16-NEXT: s_delay_alu instid0(VALU_DEP_4) | instskip(SKIP_4) | instid1(VALU_DEP_4)
-; GFX11-FAKE16-NEXT: v_perm_b32 v27, v2, v1, 0x7060302
-; GFX11-FAKE16-NEXT: v_dual_add_f32 v18, 0x40c00000, v21 :: v_dual_lshlrev_b32 v21, 16, v3
+; GFX11-FAKE16-NEXT: s_delay_alu instid0(VALU_DEP_4)
+; GFX11-FAKE16-NEXT: v_lshrrev_b32_e32 v83, 16, v2
+; GFX11-FAKE16-NEXT: v_add_f32_e32 v18, 0x40c00000, v21
; GFX11-FAKE16-NEXT: v_cndmask_b32_e32 v17, v22, v17, vcc_lo
; GFX11-FAKE16-NEXT: v_cmp_u_f32_e32 vcc_lo, v19, v19
-; GFX11-FAKE16-NEXT: v_add_f32_e32 v4, 0x40c00000, v4
-; GFX11-FAKE16-NEXT: v_or_b32_e32 v22, 0x400000, v18
-; GFX11-FAKE16-NEXT: v_add_f32_e32 v21, 0x40c00000, v21
-; GFX11-FAKE16-NEXT: v_and_b32_e32 v3, 0xffff0000, v3
-; GFX11-FAKE16-NEXT: v_cndmask_b32_e32 v19, v20, v23, vcc_lo
-; GFX11-FAKE16-NEXT: v_bfe_u32 v20, v18, 16, 1
-; GFX11-FAKE16-NEXT: v_cmp_u_f32_e32 vcc_lo, v18, v18
-; GFX11-FAKE16-NEXT: v_or_b32_e32 v23, 0x400000, v4
-; GFX11-FAKE16-NEXT: v_bfe_u32 v24, v21, 16, 1
-; GFX11-FAKE16-NEXT: v_perm_b32 v26, v19, v17, 0x7060302
-; GFX11-FAKE16-NEXT: v_bfe_u32 v19, v4, 16, 1
-; GFX11-FAKE16-NEXT: v_add3_u32 v20, v20, v18, 0x7fff
-; GFX11-FAKE16-NEXT: v_lshrrev_b32_e32 v81, 16, v2
+; GFX11-FAKE16-NEXT: v_perm_b32 v21, v2, v1, 0x7060302
; GFX11-FAKE16-NEXT: v_lshrrev_b32_e32 v2, 16, v1
+; GFX11-FAKE16-NEXT: s_delay_alu instid0(VALU_DEP_4) | instskip(SKIP_4) | instid1(VALU_DEP_4)
; GFX11-FAKE16-NEXT: v_lshrrev_b32_e32 v1, 16, v17
+; GFX11-FAKE16-NEXT: v_cndmask_b32_e32 v19, v20, v23, vcc_lo
+; GFX11-FAKE16-NEXT: v_dual_add_f32 v4, 0x40c00000, v4 :: v_dual_lshlrev_b32 v23, 16, v3
+; GFX11-FAKE16-NEXT: v_bfe_u32 v22, v18, 16, 1
+; GFX11-FAKE16-NEXT: v_or_b32_e32 v24, 0x400000, v18
+; GFX11-FAKE16-NEXT: v_perm_b32 v20, v19, v17, 0x7060302
+; GFX11-FAKE16-NEXT: s_delay_alu instid0(VALU_DEP_4)
+; GFX11-FAKE16-NEXT: v_add_f32_e32 v23, 0x40c00000, v23
+; GFX11-FAKE16-NEXT: v_bfe_u32 v19, v4, 16, 1
+; GFX11-FAKE16-NEXT: v_add3_u32 v22, v22, v18, 0x7fff
+; GFX11-FAKE16-NEXT: v_cmp_u_f32_e32 vcc_lo, v18, v18
+; GFX11-FAKE16-NEXT: v_or_b32_e32 v25, 0x400000, v4
+; GFX11-FAKE16-NEXT: v_bfe_u32 v26, v23, 16, 1
; GFX11-FAKE16-NEXT: v_add3_u32 v19, v19, v4, 0x7fff
-; GFX11-FAKE16-NEXT: v_cndmask_b32_e32 v18, v20, v22, vcc_lo
+; GFX11-FAKE16-NEXT: v_dual_cndmask_b32 v18, v22, v24 :: v_dual_and_b32 v3, 0xffff0000, v3
; GFX11-FAKE16-NEXT: v_cmp_u_f32_e32 vcc_lo, v4, v4
-; GFX11-FAKE16-NEXT: v_or_b32_e32 v22, 0x400000, v21
-; GFX11-FAKE16-NEXT: v_lshrrev_b32_e32 v96, 24, v27
-; GFX11-FAKE16-NEXT: v_lshrrev_b32_e32 v97, 16, v26
-; GFX11-FAKE16-NEXT: v_dual_cndmask_b32 v4, v19, v23 :: v_dual_lshlrev_b32 v23, 16, v6
+; GFX11-FAKE16-NEXT: v_or_b32_e32 v24, 0x400000, v23
+; GFX11-FAKE16-NEXT: v_lshrrev_b32_e32 v98, 24, v21
+; GFX11-FAKE16-NEXT: v_lshrrev_b32_e32 v99, 16, v20
+; GFX11-FAKE16-NEXT: v_cndmask_b32_e32 v4, v19, v25, vcc_lo
+; GFX11-FAKE16-NEXT: v_add3_u32 v19, v26, v23, 0x7fff
+; GFX11-FAKE16-NEXT: v_cmp_u_f32_e32 vcc_lo, v23, v23
+; GFX11-FAKE16-NEXT: s_delay_alu instid0(VALU_DEP_3) | instskip(NEXT) | instid1(VALU_DEP_3)
+; GFX11-FAKE16-NEXT: v_lshrrev_b32_e32 v70, 16, v4
+; GFX11-FAKE16-NEXT: v_dual_cndmask_b32 v19, v19, v24 :: v_dual_lshlrev_b32 v24, 16, v5
+; GFX11-FAKE16-NEXT: v_lshlrev_b32_e32 v25, 16, v6
; GFX11-FAKE16-NEXT: v_and_b32_e32 v6, 0xffff0000, v6
-; GFX11-FAKE16-NEXT: v_add_f32_e32 v3, 0x40c00000, v3
-; GFX11-FAKE16-NEXT: v_add3_u32 v19, v24, v21, 0x7fff
-; GFX11-FAKE16-NEXT: v_cmp_u_f32_e32 vcc_lo, v21, v21
-; GFX11-FAKE16-NEXT: v_perm_b32 v29, v4, v18, 0x7060302
-; GFX11-FAKE16-NEXT: v_add_f32_e32 v6, 0x40c00000, v6
-; GFX11-FAKE16-NEXT: v_bfe_u32 v20, v3, 16, 1
-; GFX11-FAKE16-NEXT: v_or_b32_e32 v24, 0x400000, v3
-; GFX11-FAKE16-NEXT: v_dual_cndmask_b32 v19, v19, v22 :: v_dual_lshlrev_b32 v22, 16, v5
-; GFX11-FAKE16-NEXT: v_add_f32_e32 v21, 0x40c00000, v23
-; GFX11-FAKE16-NEXT: s_delay_alu instid0(VALU_DEP_4)
-; GFX11-FAKE16-NEXT: v_add3_u32 v20, v20, v3, 0x7fff
+; GFX11-FAKE16-NEXT: s_delay_alu instid0(VALU_DEP_3) | instskip(NEXT) | instid1(VALU_DEP_2)
+; GFX11-FAKE16-NEXT: v_dual_add_f32 v3, 0x40c00000, v3 :: v_dual_add_f32 v24, 0x40c00000, v24
+; GFX11-FAKE16-NEXT: v_dual_add_f32 v23, 0x40c00000, v25 :: v_dual_add_f32 v6, 0x40c00000, v6
+; GFX11-FAKE16-NEXT: s_delay_alu instid0(VALU_DEP_2) | instskip(SKIP_2) | instid1(VALU_DEP_4)
+; GFX11-FAKE16-NEXT: v_bfe_u32 v22, v3, 16, 1
+; GFX11-FAKE16-NEXT: v_or_b32_e32 v26, 0x400000, v3
; GFX11-FAKE16-NEXT: v_cmp_u_f32_e32 vcc_lo, v3, v3
-; GFX11-FAKE16-NEXT: v_lshrrev_b32_e32 v68, 16, v4
-; GFX11-FAKE16-NEXT: v_add_f32_e32 v22, 0x40c00000, v22
-; GFX11-FAKE16-NEXT: v_or_b32_e32 v23, 0x400000, v21
+; GFX11-FAKE16-NEXT: v_or_b32_e32 v27, 0x400000, v23
+; GFX11-FAKE16-NEXT: v_or_b32_e32 v28, 0x400000, v6
+; GFX11-FAKE16-NEXT: v_add3_u32 v22, v22, v3, 0x7fff
+; GFX11-FAKE16-NEXT: v_bfe_u32 v29, v24, 16, 1
+; GFX11-FAKE16-NEXT: s_delay_alu instid0(VALU_DEP_2)
+; GFX11-FAKE16-NEXT: v_cndmask_b32_e32 v3, v22, v26, vcc_lo
+; GFX11-FAKE16-NEXT: v_bfe_u32 v22, v23, 16, 1
+; GFX11-FAKE16-NEXT: v_cmp_u_f32_e32 vcc_lo, v23, v23
+; GFX11-FAKE16-NEXT: v_perm_b32 v26, v4, v18, 0x7060302
; GFX11-FAKE16-NEXT: v_lshrrev_b32_e32 v4, 16, v18
-; GFX11-FAKE16-NEXT: v_cndmask_b32_e32 v3, v20, v24, vcc_lo
-; GFX11-FAKE16-NEXT: v_bfe_u32 v20, v21, 16, 1
-; GFX11-FAKE16-NEXT: v_cmp_u_f32_e32 vcc_lo, v21, v21
-; GFX11-FAKE16-NEXT: v_or_b32_e32 v24, 0x400000, v6
-; GFX11-FAKE16-NEXT: v_bfe_u32 v25, v22, 16, 1
-; GFX11-FAKE16-NEXT: v_perm_b32 v28, v3, v19, 0x7060302
+; GFX11-FAKE16-NEXT: v_perm_b32 v25, v3, v19, 0x7060302
; GFX11-FAKE16-NEXT: v_bfe_u32 v3, v6, 16, 1
-; GFX11-FAKE16-NEXT: v_add3_u32 v20, v20, v21, 0x7fff
-; GFX11-FAKE16-NEXT: v_and_b32_e32 v5, 0xffff0000, v5
-; GFX11-FAKE16-NEXT: s_delay_alu instid0(VALU_DEP_4) | instskip(NEXT) | instid1(VALU_DEP_4)
-; GFX11-FAKE16-NEXT: v_lshrrev_b32_e32 v87, 16, v28
+; GFX11-FAKE16-NEXT: v_add3_u32 v22, v22, v23, 0x7fff
+; GFX11-FAKE16-NEXT: s_delay_alu instid0(VALU_DEP_2) | instskip(NEXT) | instid1(VALU_DEP_2)
; GFX11-FAKE16-NEXT: v_add3_u32 v3, v3, v6, 0x7fff
-; GFX11-FAKE16-NEXT: s_delay_alu instid0(VALU_DEP_3)
-; GFX11-FAKE16-NEXT: v_dual_cndmask_b32 v20, v20, v23 :: v_dual_add_f32 v5, 0x40c00000, v5
+; GFX11-FAKE16-NEXT: v_cndmask_b32_e32 v22, v22, v27, vcc_lo
; GFX11-FAKE16-NEXT: v_cmp_u_f32_e32 vcc_lo, v6, v6
-; GFX11-FAKE16-NEXT: v_add3_u32 v6, v25, v22, 0x7fff
-; GFX11-FAKE16-NEXT: v_or_b32_e32 v23, 0x400000, v22
-; GFX11-FAKE16-NEXT: v_cndmask_b32_e32 v3, v3, v24, vcc_lo
-; GFX11-FAKE16-NEXT: v_lshlrev_b32_e32 v24, 16, v8
-; GFX11-FAKE16-NEXT: v_bfe_u32 v21, v5, 16, 1
-; GFX11-FAKE16-NEXT: v_cmp_u_f32_e32 vcc_lo, v22, v22
-; GFX11-FAKE16-NEXT: v_or_b32_e32 v25, 0x400000, v5
-; GFX11-FAKE16-NEXT: s_delay_alu instid0(VALU_DEP_3)
-; GFX11-FAKE16-NEXT: v_add3_u32 v21, v21, v5, 0x7fff
-; GFX11-FAKE16-NEXT: v_cndmask_b32_e32 v22, v6, v23, vcc_lo
-; GFX11-FAKE16-NEXT: v_add_f32_e32 v6, 0x40c00000, v24
+; GFX11-FAKE16-NEXT: v_add3_u32 v6, v29, v24, 0x7fff
+; GFX11-FAKE16-NEXT: v_or_b32_e32 v27, 0x400000, v24
+; GFX11-FAKE16-NEXT: v_cndmask_b32_e32 v3, v3, v28, vcc_lo
+; GFX11-FAKE16-NEXT: v_cmp_u_f32_e32 vcc_lo, v24, v24
+; GFX11-FAKE16-NEXT: v_and_b32_e32 v5, 0xffff0000, v5
+; GFX11-FAKE16-NEXT: v_lshlrev_b32_e32 v28, 16, v8
+; GFX11-FAKE16-NEXT: s_delay_alu instid0(VALU_DEP_4) | instskip(NEXT) | instid1(VALU_DEP_3)
+; GFX11-FAKE16-NEXT: v_perm_b32 v97, v3, v22, 0x7060302
+; GFX11-FAKE16-NEXT: v_dual_cndmask_b32 v24, v6, v27 :: v_dual_add_f32 v5, 0x40c00000, v5
+; GFX11-FAKE16-NEXT: s_delay_alu instid0(VALU_DEP_3) | instskip(SKIP_2) | instid1(VALU_DEP_4)
+; GFX11-FAKE16-NEXT: v_dual_add_f32 v6, 0x40c00000, v28 :: v_dual_lshlrev_b32 v27, 16, v7
+; GFX11-FAKE16-NEXT: v_lshrrev_b32_e32 v67, 16, v3
+; GFX11-FAKE16-NEXT: v_lshrrev_b32_e32 v3, 16, v19
+; GFX11-FAKE16-NEXT: v_bfe_u32 v23, v5, 16, 1
+; GFX11-FAKE16-NEXT: v_or_b32_e32 v29, 0x400000, v5
; GFX11-FAKE16-NEXT: v_cmp_u_f32_e32 vcc_lo, v5, v5
-; GFX11-FAKE16-NEXT: v_and_b32_e32 v8, 0xffff0000, v8
-; GFX11-FAKE16-NEXT: v_lshlrev_b32_e32 v23, 16, v7
-; GFX11-FAKE16-NEXT: v_perm_b32 v86, v3, v20, 0x7060302
-; GFX11-FAKE16-NEXT: v_or_b32_e32 v24, 0x400000, v6
-; GFX11-FAKE16-NEXT: s_delay_alu instid0(VALU_DEP_4) | instskip(SKIP_3) | instid1(VALU_DEP_4)
-; GFX11-FAKE16-NEXT: v_dual_cndmask_b32 v5, v21, v25 :: v_dual_add_f32 v8, 0x40c00000, v8
-; GFX11-FAKE16-NEXT: v_bfe_u32 v21, v6, 16, 1
+; GFX11-FAKE16-NEXT: v_or_b32_e32 v28, 0x400000, v6
+; GFX11-FAKE16-NEXT: s_delay_alu instid0(VALU_DEP_4) | instskip(NEXT) | instid1(VALU_DEP_1)
+; GFX11-FAKE16-NEXT: v_add3_u32 v23, v23, v5, 0x7fff
+; GFX11-FAKE16-NEXT: v_dual_cndmask_b32 v5, v23, v29 :: v_dual_and_b32 v8, 0xffff0000, v8
+; GFX11-FAKE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(SKIP_2) | instid1(VALU_DEP_4)
+; GFX11-FAKE16-NEXT: v_add_f32_e32 v8, 0x40c00000, v8
+; GFX11-FAKE16-NEXT: v_bfe_u32 v23, v6, 16, 1
; GFX11-FAKE16-NEXT: v_cmp_u_f32_e32 vcc_lo, v6, v6
-; GFX11-FAKE16-NEXT: v_lshrrev_b32_e32 v65, 16, v3
-; GFX11-FAKE16-NEXT: v_perm_b32 v85, v5, v22, 0x7060302
+; GFX11-FAKE16-NEXT: v_perm_b32 v96, v5, v24, 0x7060302
+; GFX11-FAKE16-NEXT: s_delay_alu instid0(VALU_DEP_4) | instskip(NEXT) | instid1(VALU_DEP_4)
; GFX11-FAKE16-NEXT: v_bfe_u32 v5, v8, 16, 1
-; GFX11-FAKE16-NEXT: v_add3_u32 v21, v21, v6, 0x7fff
-; GFX11-FAKE16-NEXT: v_or_b32_e32 v25, 0x400000, v8
-; GFX11-FAKE16-NEXT: v_lshrrev_b32_e32 v3, 16, v19
-; GFX11-FAKE16-NEXT: v_lshrrev_b32_e32 v84, 16, v85
+; GFX11-FAKE16-NEXT: v_add3_u32 v23, v23, v6, 0x7fff
+; GFX11-FAKE16-NEXT: v_or_b32_e32 v29, 0x400000, v8
+; GFX11-FAKE16-NEXT: s_delay_alu instid0(VALU_DEP_4) | instskip(NEXT) | instid1(VALU_DEP_4)
+; GFX11-FAKE16-NEXT: v_lshrrev_b32_e32 v86, 16, v96
; GFX11-FAKE16-NEXT: v_add3_u32 v5, v5, v8, 0x7fff
-; GFX11-FAKE16-NEXT: v_cndmask_b32_e32 v6, v21, v24, vcc_lo
+; GFX11-FAKE16-NEXT: s_delay_alu instid0(VALU_DEP_4) | instskip(SKIP_2) | instid1(VALU_DEP_4)
+; GFX11-FAKE16-NEXT: v_cndmask_b32_e32 v6, v23, v28, vcc_lo
; GFX11-FAKE16-NEXT: v_cmp_u_f32_e32 vcc_lo, v8, v8
-; GFX11-FAKE16-NEXT: s_delay_alu instid0(VALU_DEP_3) | instskip(SKIP_3) | instid1(VALU_DEP_4)
-; GFX11-FAKE16-NEXT: v_cndmask_b32_e32 v5, v5, v25, vcc_lo
-; GFX11-FAKE16-NEXT: v_lshlrev_b32_e32 v25, 16, v10
-; GFX11-FAKE16-NEXT: v_add_f32_e32 v23, 0x40c00000, v23
+; GFX11-FAKE16-NEXT: v_lshrrev_b32_e32 v87, 8, v96
+; GFX11-FAKE16-NEXT: v_cndmask_b32_e32 v5, v5, v29, vcc_lo
+; GFX11-FAKE16-NEXT: v_lshlrev_b32_e32 v29, 16, v10
+; GFX11-FAKE16-NEXT: v_add_f32_e32 v27, 0x40c00000, v27
; GFX11-FAKE16-NEXT: v_and_b32_e32 v10, 0xffff0000, v10
-; GFX11-FAKE16-NEXT: v_perm_b32 v83, v5, v6, 0x7060302
-; GFX11-FAKE16-NEXT: v_lshrrev_b32_e32 v52, 16, v5
-; GFX11-FAKE16-NEXT: s_delay_alu instid0(VALU_DEP_4)
-; GFX11-FAKE16-NEXT: v_bfe_u32 v30, v23, 16, 1
-; GFX11-FAKE16-NEXT: v_or_b32_e32 v24, 0x400000, v23
-; GFX11-FAKE16-NEXT: v_cmp_u_f32_e32 vcc_lo, v23, v23
+; GFX11-FAKE16-NEXT: s_delay_alu instid0(VALU_DEP_4) | instskip(SKIP_1) | instid1(VALU_DEP_4)
+; GFX11-FAKE16-NEXT: v_perm_b32 v85, v5, v6, 0x7060302
+; GFX11-FAKE16-NEXT: v_lshrrev_b32_e32 v54, 16, v5
+; GFX11-FAKE16-NEXT: v_bfe_u32 v30, v27, 16, 1
+; GFX11-FAKE16-NEXT: v_or_b32_e32 v28, 0x400000, v27
+; GFX11-FAKE16-NEXT: v_cmp_u_f32_e32 vcc_lo, v27, v27
; GFX11-FAKE16-NEXT: v_add_f32_e32 v10, 0x40c00000, v10
-; GFX11-FAKE16-NEXT: v_lshrrev_b32_e32 v5, 16, v22
-; GFX11-FAKE16-NEXT: v_add3_u32 v8, v30, v23, 0x7fff
-; GFX11-FAKE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
-; GFX11-FAKE16-NEXT: v_dual_cndmask_b32 v23, v8, v24 :: v_dual_lshlrev_b32 v24, 16, v9
-; GFX11-FAKE16-NEXT: v_add_f32_e32 v24, 0x40c00000, v24
-; GFX11-FAKE16-NEXT: v_add_f32_e32 v8, 0x40c00000, v25
+; GFX11-FAKE16-NEXT: v_lshrrev_b32_e32 v5, 16, v24
+; GFX11-FAKE16-NEXT: v_add3_u32 v8, v30, v27, 0x7fff
+; GFX11-FAKE16-NEXT: v_lshrrev_b32_e32 v71, 24, v85
+; GFX11-FAKE16-NEXT: s_delay_alu instid0(VALU_DEP_2) | instskip(NEXT) | instid1(VALU_DEP_1)
+; GFX11-FAKE16-NEXT: v_dual_cndmask_b32 v27, v8, v28 :: v_dual_lshlrev_b32 v28, 16, v9
+; GFX11-FAKE16-NEXT: v_add_f32_e32 v28, 0x40c00000, v28
+; GFX11-FAKE16-NEXT: v_add_f32_e32 v8, 0x40c00000, v29
; GFX11-FAKE16-NEXT: s_delay_alu instid0(VALU_DEP_2) | instskip(NEXT) | instid1(VALU_DEP_2)
-; GFX11-FAKE16-NEXT: v_bfe_u32 v31, v24, 16, 1
-; GFX11-FAKE16-NEXT: v_or_b32_e32 v25, 0x400000, v8
+; GFX11-FAKE16-NEXT: v_bfe_u32 v31, v28, 16, 1
+; GFX11-FAKE16-NEXT: v_or_b32_e32 v29, 0x400000, v8
; GFX11-FAKE16-NEXT: v_and_b32_e32 v7, 0xffff0000, v7
; GFX11-FAKE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
; GFX11-FAKE16-NEXT: v_add_f32_e32 v7, 0x40c00000, v7
-; GFX11-FAKE16-NEXT: v_bfe_u32 v21, v7, 16, 1
+; GFX11-FAKE16-NEXT: v_bfe_u32 v23, v7, 16, 1
; GFX11-FAKE16-NEXT: v_or_b32_e32 v30, 0x400000, v7
; GFX11-FAKE16-NEXT: v_cmp_u_f32_e32 vcc_lo, v7, v7
; GFX11-FAKE16-NEXT: s_delay_alu instid0(VALU_DEP_3) | instskip(NEXT) | instid1(VALU_DEP_1)
-; GFX11-FAKE16-NEXT: v_add3_u32 v21, v21, v7, 0x7fff
-; GFX11-FAKE16-NEXT: v_cndmask_b32_e32 v7, v21, v30, vcc_lo
-; GFX11-FAKE16-NEXT: v_bfe_u32 v21, v8, 16, 1
+; GFX11-FAKE16-NEXT: v_add3_u32 v23, v23, v7, 0x7fff
+; GFX11-FAKE16-NEXT: v_cndmask_b32_e32 v7, v23, v30, vcc_lo
+; GFX11-FAKE16-NEXT: v_bfe_u32 v23, v8, 16, 1
; GFX11-FAKE16-NEXT: v_cmp_u_f32_e32 vcc_lo, v8, v8
; GFX11-FAKE16-NEXT: v_or_b32_e32 v30, 0x400000, v10
; GFX11-FAKE16-NEXT: s_delay_alu instid0(VALU_DEP_4) | instskip(SKIP_2) | instid1(VALU_DEP_3)
-; GFX11-FAKE16-NEXT: v_perm_b32 v82, v7, v23, 0x7060302
+; GFX11-FAKE16-NEXT: v_perm_b32 v84, v7, v27, 0x7060302
; GFX11-FAKE16-NEXT: v_bfe_u32 v7, v10, 16, 1
-; GFX11-FAKE16-NEXT: v_add3_u32 v21, v21, v8, 0x7fff
-; GFX11-FAKE16-NEXT: v_lshrrev_b32_e32 v71, 16, v82
+; GFX11-FAKE16-NEXT: v_add3_u32 v23, v23, v8, 0x7fff
+; GFX11-FAKE16-NEXT: v_lshrrev_b32_e32 v82, 8, v84
; GFX11-FAKE16-NEXT: s_delay_alu instid0(VALU_DEP_3) | instskip(NEXT) | instid1(VALU_DEP_3)
; GFX11-FAKE16-NEXT: v_add3_u32 v7, v7, v10, 0x7fff
-; GFX11-FAKE16-NEXT: v_cndmask_b32_e32 v8, v21, v25, vcc_lo
+; GFX11-FAKE16-NEXT: v_cndmask_b32_e32 v8, v23, v29, vcc_lo
; GFX11-FAKE16-NEXT: v_cmp_u_f32_e32 vcc_lo, v10, v10
-; GFX11-FAKE16-NEXT: v_add3_u32 v10, v31, v24, 0x7fff
-; GFX11-FAKE16-NEXT: v_or_b32_e32 v25, 0x400000, v24
-; GFX11-FAKE16-NEXT: v_lshrrev_b32_e32 v80, 8, v82
+; GFX11-FAKE16-NEXT: v_add3_u32 v10, v31, v28, 0x7fff
+; GFX11-FAKE16-NEXT: v_or_b32_e32 v29, 0x400000, v28
; GFX11-FAKE16-NEXT: v_dual_cndmask_b32 v7, v7, v30 :: v_dual_lshlrev_b32 v30, 16, v12
; GFX11-FAKE16-NEXT: v_and_b32_e32 v12, 0xffff0000, v12
-; GFX11-FAKE16-NEXT: v_cmp_u_f32_e32 vcc_lo, v24, v24
+; GFX11-FAKE16-NEXT: v_cmp_u_f32_e32 vcc_lo, v28, v28
; GFX11-FAKE16-NEXT: s_delay_alu instid0(VALU_DEP_3) | instskip(SKIP_1) | instid1(VALU_DEP_4)
-; GFX11-FAKE16-NEXT: v_perm_b32 v70, v7, v8, 0x7060302
-; GFX11-FAKE16-NEXT: v_lshrrev_b32_e32 v38, 16, v7
+; GFX11-FAKE16-NEXT: v_perm_b32 v81, v7, v8, 0x7060302
+; GFX11-FAKE16-NEXT: v_lshrrev_b32_e32 v48, 16, v7
; GFX11-FAKE16-NEXT: v_add_f32_e32 v12, 0x40c00000, v12
-; GFX11-FAKE16-NEXT: v_cndmask_b32_e32 v24, v10, v25, vcc_lo
+; GFX11-FAKE16-NEXT: v_cndmask_b32_e32 v28, v10, v29, vcc_lo
; GFX11-FAKE16-NEXT: v_add_f32_e32 v10, 0x40c00000, v30
-; GFX11-FAKE16-NEXT: v_lshrrev_b32_e32 v7, 16, v23
-; GFX11-FAKE16-NEXT: v_lshrrev_b32_e32 v64, 8, v70
+; GFX11-FAKE16-NEXT: v_lshrrev_b32_e32 v7, 16, v27
+; GFX11-FAKE16-NEXT: v_lshrrev_b32_e32 v66, 8, v81
; GFX11-FAKE16-NEXT: v_bfe_u32 v30, v12, 16, 1
; GFX11-FAKE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(SKIP_1) | instid1(VALU_DEP_1)
; GFX11-FAKE16-NEXT: v_add3_u32 v30, v30, v12, 0x7fff
; GFX11-FAKE16-NEXT: v_and_b32_e32 v9, 0xffff0000, v9
; GFX11-FAKE16-NEXT: v_add_f32_e32 v9, 0x40c00000, v9
; GFX11-FAKE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(SKIP_2) | instid1(VALU_DEP_3)
-; GFX11-FAKE16-NEXT: v_bfe_u32 v21, v9, 16, 1
+; GFX11-FAKE16-NEXT: v_bfe_u32 v23, v9, 16, 1
; GFX11-FAKE16-NEXT: v_or_b32_e32 v31, 0x400000, v9
; GFX11-FAKE16-NEXT: v_cmp_u_f32_e32 vcc_lo, v9, v9
-; GFX11-FAKE16-NEXT: v_add3_u32 v21, v21, v9, 0x7fff
+; GFX11-FAKE16-NEXT: v_add3_u32 v23, v23, v9, 0x7fff
; GFX11-FAKE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(SKIP_3) | instid1(VALU_DEP_4)
-; GFX11-FAKE16-NEXT: v_cndmask_b32_e32 v9, v21, v31, vcc_lo
-; GFX11-FAKE16-NEXT: v_lshlrev_b32_e32 v21, 16, v11
-; GFX11-FAKE16-NEXT: v_bfe_u32 v25, v10, 16, 1
+; GFX11-FAKE16-NEXT: v_cndmask_b32_e32 v9, v23, v31, vcc_lo
+; GFX11-FAKE16-NEXT: v_lshlrev_b32_e32 v23, 16, v11
+; GFX11-FAKE16-NEXT: v_bfe_u32 v29, v10, 16, 1
; GFX11-FAKE16-NEXT: v_cmp_u_f32_e32 vcc_lo, v10, v10
-; GFX11-FAKE16-NEXT: v_perm_b32 v69, v9, v24, 0x7060302
+; GFX11-FAKE16-NEXT: v_perm_b32 v80, v9, v28, 0x7060302
; GFX11-FAKE16-NEXT: s_delay_alu instid0(VALU_DEP_4) | instskip(NEXT) | instid1(VALU_DEP_4)
-; GFX11-FAKE16-NEXT: v_add_f32_e32 v9, 0x40c00000, v21
-; GFX11-FAKE16-NEXT: v_add3_u32 v21, v25, v10, 0x7fff
-; GFX11-FAKE16-NEXT: v_or_b32_e32 v25, 0x400000, v10
+; GFX11-FAKE16-NEXT: v_add_f32_e32 v9, 0x40c00000, v23
+; GFX11-FAKE16-NEXT: v_add3_u32 v23, v29, v10, 0x7fff
+; GFX11-FAKE16-NEXT: v_or_b32_e32 v29, 0x400000, v10
; GFX11-FAKE16-NEXT: v_or_b32_e32 v31, 0x400000, v12
; GFX11-FAKE16-NEXT: v_and_b32_e32 v11, 0xffff0000, v11
; GFX11-FAKE16-NEXT: v_bfe_u32 v32, v9, 16, 1
-; GFX11-FAKE16-NEXT: v_lshrrev_b32_e32 v66, 16, v69
-; GFX11-FAKE16-NEXT: v_cndmask_b32_e32 v10, v21, v25, vcc_lo
+; GFX11-FAKE16-NEXT: v_lshrrev_b32_e32 v68, 16, v80
+; GFX11-FAKE16-NEXT: v_cndmask_b32_e32 v10, v23, v29, vcc_lo
; GFX11-FAKE16-NEXT: v_cmp_u_f32_e32 vcc_lo, v12, v12
-; GFX11-FAKE16-NEXT: v_or_b32_e32 v25, 0x400000, v9
-; GFX11-FAKE16-NEXT: v_add3_u32 v21, v32, v9, 0x7fff
-; GFX11-FAKE16-NEXT: v_lshrrev_b32_e32 v67, 8, v69
+; GFX11-FAKE16-NEXT: v_or_b32_e32 v29, 0x400000, v9
+; GFX11-FAKE16-NEXT: v_add3_u32 v23, v32, v9, 0x7fff
+; GFX11-FAKE16-NEXT: v_lshrrev_b32_e32 v69, 8, v80
; GFX11-FAKE16-NEXT: v_cndmask_b32_e32 v12, v30, v31, vcc_lo
; GFX11-FAKE16-NEXT: v_cmp_u_f32_e32 vcc_lo, v9, v9
; GFX11-FAKE16-NEXT: v_lshlrev_b32_e32 v30, 16, v14
; GFX11-FAKE16-NEXT: v_and_b32_e32 v14, 0xffff0000, v14
; GFX11-FAKE16-NEXT: v_add_f32_e32 v11, 0x40c00000, v11
-; GFX11-FAKE16-NEXT: v_perm_b32 v55, v12, v10, 0x7060302
-; GFX11-FAKE16-NEXT: v_cndmask_b32_e32 v9, v21, v25, vcc_lo
-; GFX11-FAKE16-NEXT: v_add_f32_e32 v21, 0x40c00000, v30
+; GFX11-FAKE16-NEXT: v_perm_b32 v65, v12, v10, 0x7060302
+; GFX11-FAKE16-NEXT: v_cndmask_b32_e32 v9, v23, v29, vcc_lo
+; GFX11-FAKE16-NEXT: v_add_f32_e32 v23, 0x40c00000, v30
; GFX11-FAKE16-NEXT: v_add_f32_e32 v14, 0x40c00000, v14
; GFX11-FAKE16-NEXT: v_bfe_u32 v31, v11, 16, 1
; GFX11-FAKE16-NEXT: v_or_b32_e32 v30, 0x400000, v11
; GFX11-FAKE16-NEXT: v_cmp_u_f32_e32 vcc_lo, v11, v11
-; GFX11-FAKE16-NEXT: v_bfe_u32 v32, v21, 16, 1
-; GFX11-FAKE16-NEXT: v_lshrrev_b32_e32 v51, 8, v55
-; GFX11-FAKE16-NEXT: v_add3_u32 v25, v31, v11, 0x7fff
+; GFX11-FAKE16-NEXT: v_bfe_u32 v32, v23, 16, 1
+; GFX11-FAKE16-NEXT: v_lshrrev_b32_e32 v53, 8, v65
+; GFX11-FAKE16-NEXT: v_add3_u32 v29, v31, v11, 0x7fff
; GFX11-FAKE16-NEXT: v_lshlrev_b32_e32 v31, 16, v13
; GFX11-FAKE16-NEXT: v_and_b32_e32 v13, 0xffff0000, v13
; GFX11-FAKE16-NEXT: s_delay_alu instid0(VALU_DEP_3) | instskip(NEXT) | instid1(VALU_DEP_3)
-; GFX11-FAKE16-NEXT: v_cndmask_b32_e32 v11, v25, v30, vcc_lo
-; GFX11-FAKE16-NEXT: v_add_f32_e32 v25, 0x40c00000, v31
-; GFX11-FAKE16-NEXT: v_add3_u32 v30, v32, v21, 0x7fff
-; GFX11-FAKE16-NEXT: v_or_b32_e32 v31, 0x400000, v21
+; GFX11-FAKE16-NEXT: v_cndmask_b32_e32 v11, v29, v30, vcc_lo
+; GFX11-FAKE16-NEXT: v_add_f32_e32 v29, 0x40c00000, v31
+; GFX11-FAKE16-NEXT: v_add3_u32 v30, v32, v23, 0x7fff
+; GFX11-FAKE16-NEXT: v_or_b32_e32 v31, 0x400000, v23
; GFX11-FAKE16-NEXT: v_bfe_u32 v32, v14, 16, 1
-; GFX11-FAKE16-NEXT: v_cmp_u_f32_e32 vcc_lo, v21, v21
-; GFX11-FAKE16-NEXT: v_bfe_u32 v33, v25, 16, 1
+; GFX11-FAKE16-NEXT: v_cmp_u_f32_e32 vcc_lo, v23, v23
+; GFX11-FAKE16-NEXT: v_bfe_u32 v33, v29, 16, 1
; GFX11-FAKE16-NEXT: v_add_f32_e32 v13, 0x40c00000, v13
-; GFX11-FAKE16-NEXT: v_or_b32_e32 v34, 0x400000, v25
-; GFX11-FAKE16-NEXT: v_cndmask_b32_e32 v21, v30, v31, vcc_lo
+; GFX11-FAKE16-NEXT: v_or_b32_e32 v34, 0x400000, v29
+; GFX11-FAKE16-NEXT: v_cndmask_b32_e32 v23, v30, v31, vcc_lo
; GFX11-FAKE16-NEXT: v_add3_u32 v30, v32, v14, 0x7fff
; GFX11-FAKE16-NEXT: v_or_b32_e32 v31, 0x400000, v14
; GFX11-FAKE16-NEXT: v_cmp_u_f32_e32 vcc_lo, v14, v14
-; GFX11-FAKE16-NEXT: v_add3_u32 v32, v33, v25, 0x7fff
+; GFX11-FAKE16-NEXT: v_add3_u32 v32, v33, v29, 0x7fff
; GFX11-FAKE16-NEXT: v_lshlrev_b32_e32 v33, 16, v16
; GFX11-FAKE16-NEXT: v_and_b32_e32 v16, 0xffff0000, v16
; GFX11-FAKE16-NEXT: v_cndmask_b32_e32 v14, v30, v31, vcc_lo
-; GFX11-FAKE16-NEXT: v_cmp_u_f32_e32 vcc_lo, v25, v25
+; GFX11-FAKE16-NEXT: v_cmp_u_f32_e32 vcc_lo, v29, v29
; GFX11-FAKE16-NEXT: s_delay_alu instid0(VALU_DEP_4) | instskip(NEXT) | instid1(VALU_DEP_4)
; GFX11-FAKE16-NEXT: v_add_f32_e32 v30, 0x40c00000, v33
; GFX11-FAKE16-NEXT: v_add_f32_e32 v16, 0x40c00000, v16
-; GFX11-FAKE16-NEXT: v_perm_b32 v54, v11, v9, 0x7060302
-; GFX11-FAKE16-NEXT: v_perm_b32 v50, v14, v21, 0x7060302
-; GFX11-FAKE16-NEXT: v_dual_cndmask_b32 v25, v32, v34 :: v_dual_lshlrev_b32 v34, 16, v15
+; GFX11-FAKE16-NEXT: v_perm_b32 v64, v11, v9, 0x7060302
+; GFX11-FAKE16-NEXT: v_perm_b32 v52, v14, v23, 0x7060302
+; GFX11-FAKE16-NEXT: v_dual_cndmask_b32 v29, v32, v34 :: v_dual_lshlrev_b32 v34, 16, v15
; GFX11-FAKE16-NEXT: v_bfe_u32 v35, v13, 16, 1
; GFX11-FAKE16-NEXT: v_or_b32_e32 v32, 0x400000, v13
; GFX11-FAKE16-NEXT: v_bfe_u32 v33, v30, 16, 1
@@ -89502,8 +89423,8 @@ define <64 x i8> @bitcast_v32bf16_to_v64i8(<32 x bfloat> %a, i32 %b) {
; GFX11-FAKE16-NEXT: v_and_b32_e32 v15, 0xffff0000, v15
; GFX11-FAKE16-NEXT: v_add3_u32 v31, v35, v13, 0x7fff
; GFX11-FAKE16-NEXT: v_lshrrev_b32_e32 v11, 16, v9
-; GFX11-FAKE16-NEXT: v_lshrrev_b32_e32 v9, 16, v24
-; GFX11-FAKE16-NEXT: v_lshrrev_b32_e32 v39, 8, v50
+; GFX11-FAKE16-NEXT: v_lshrrev_b32_e32 v9, 16, v28
+; GFX11-FAKE16-NEXT: v_lshrrev_b64 v[27:28], 24, v[84:85]
; GFX11-FAKE16-NEXT: v_add_f32_e32 v15, 0x40c00000, v15
; GFX11-FAKE16-NEXT: v_cndmask_b32_e32 v13, v31, v32, vcc_lo
; GFX11-FAKE16-NEXT: v_add_f32_e32 v31, 0x40c00000, v34
@@ -89512,191 +89433,190 @@ define <64 x i8> @bitcast_v32bf16_to_v64i8(<32 x bfloat> %a, i32 %b) {
; GFX11-FAKE16-NEXT: v_bfe_u32 v34, v16, 16, 1
; GFX11-FAKE16-NEXT: v_cmp_u_f32_e32 vcc_lo, v30, v30
; GFX11-FAKE16-NEXT: v_bfe_u32 v35, v31, 16, 1
-; GFX11-FAKE16-NEXT: v_bfe_u32 v30, v15, 16, 1
; GFX11-FAKE16-NEXT: v_or_b32_e32 v36, 0x400000, v31
; GFX11-FAKE16-NEXT: v_or_b32_e32 v37, 0x400000, v15
-; GFX11-FAKE16-NEXT: v_cndmask_b32_e32 v32, v32, v33, vcc_lo
+; GFX11-FAKE16-NEXT: v_perm_b32 v51, v13, v29, 0x7060302
+; GFX11-FAKE16-NEXT: v_cndmask_b32_e32 v30, v32, v33, vcc_lo
; GFX11-FAKE16-NEXT: v_add3_u32 v33, v34, v16, 0x7fff
; GFX11-FAKE16-NEXT: v_or_b32_e32 v34, 0x400000, v16
; GFX11-FAKE16-NEXT: v_cmp_u_f32_e32 vcc_lo, v16, v16
+; GFX11-FAKE16-NEXT: v_bfe_u32 v32, v15, 16, 1
; GFX11-FAKE16-NEXT: v_add3_u32 v35, v35, v31, 0x7fff
-; GFX11-FAKE16-NEXT: v_add3_u32 v30, v30, v15, 0x7fff
-; GFX11-FAKE16-NEXT: v_perm_b32 v49, v13, v25, 0x7060302
-; GFX11-FAKE16-NEXT: v_lshrrev_b32_e32 v13, 16, v25
+; GFX11-FAKE16-NEXT: v_lshrrev_b32_e32 v13, 16, v29
+; GFX11-FAKE16-NEXT: v_lshrrev_b64 v[28:29], 24, v[96:97]
; GFX11-FAKE16-NEXT: v_cndmask_b32_e32 v16, v33, v34, vcc_lo
; GFX11-FAKE16-NEXT: v_cmp_u_f32_e32 vcc_lo, v31, v31
-; GFX11-FAKE16-NEXT: v_lshrrev_b32_e32 v33, 16, v12
+; GFX11-FAKE16-NEXT: v_add3_u32 v32, v32, v15, 0x7fff
+; GFX11-FAKE16-NEXT: v_lshrrev_b32_e32 v33, 16, v14
+; GFX11-FAKE16-NEXT: v_lshrrev_b32_e32 v14, 16, v23
+; GFX11-FAKE16-NEXT: v_perm_b32 v39, v16, v30, 0x7060302
+; GFX11-FAKE16-NEXT: v_cndmask_b32_e32 v31, v35, v36, vcc_lo
+; GFX11-FAKE16-NEXT: v_cmp_u_f32_e32 vcc_lo, v15, v15
+; GFX11-FAKE16-NEXT: v_lshrrev_b32_e32 v35, 16, v12
; GFX11-FAKE16-NEXT: v_lshrrev_b32_e32 v12, 16, v10
; GFX11-FAKE16-NEXT: v_lshrrev_b32_e32 v10, 16, v8
; GFX11-FAKE16-NEXT: v_lshrrev_b32_e32 v8, 16, v6
-; GFX11-FAKE16-NEXT: v_cndmask_b32_e32 v31, v35, v36, vcc_lo
-; GFX11-FAKE16-NEXT: v_cmp_u_f32_e32 vcc_lo, v15, v15
-; GFX11-FAKE16-NEXT: v_lshrrev_b32_e32 v6, 16, v20
-; GFX11-FAKE16-NEXT: v_lshrrev_b32_e32 v48, 16, v49
-; GFX11-FAKE16-NEXT: v_lshrrev_b32_e32 v53, 16, v54
-; GFX11-FAKE16-NEXT: v_cndmask_b32_e32 v15, v30, v37, vcc_lo
-; GFX11-FAKE16-NEXT: v_perm_b32 v37, v16, v32, 0x7060302
-; GFX11-FAKE16-NEXT: v_lshrrev_b32_e32 v30, 16, v16
-; GFX11-FAKE16-NEXT: v_lshrrev_b32_e32 v16, 16, v32
-; GFX11-FAKE16-NEXT: s_delay_alu instid0(VALU_DEP_4)
-; GFX11-FAKE16-NEXT: v_perm_b32 v36, v15, v31, 0x7060302
+; GFX11-FAKE16-NEXT: v_cndmask_b32_e32 v15, v32, v37, vcc_lo
+; GFX11-FAKE16-NEXT: v_lshrrev_b32_e32 v32, 16, v16
+; GFX11-FAKE16-NEXT: v_lshrrev_b32_e32 v16, 16, v30
+; GFX11-FAKE16-NEXT: v_lshrrev_b32_e32 v6, 16, v22
+; GFX11-FAKE16-NEXT: v_lshrrev_b64 v[22:23], 24, v[64:65]
+; GFX11-FAKE16-NEXT: v_perm_b32 v38, v15, v31, 0x7060302
+; GFX11-FAKE16-NEXT: v_lshrrev_b64 v[29:30], 24, v[25:26]
; GFX11-FAKE16-NEXT: v_lshrrev_b32_e32 v15, 16, v31
-; GFX11-FAKE16-NEXT: v_lshrrev_b32_e32 v31, 16, v14
-; GFX11-FAKE16-NEXT: v_lshrrev_b32_e32 v14, 16, v21
-; GFX11-FAKE16-NEXT: v_lshrrev_b32_e32 v32, 24, v37
-; GFX11-FAKE16-NEXT: v_lshrrev_b64 v[17:18], 24, v[36:37]
-; GFX11-FAKE16-NEXT: v_lshrrev_b64 v[18:19], 24, v[49:50]
-; GFX11-FAKE16-NEXT: v_lshrrev_b64 v[19:20], 24, v[54:55]
-; GFX11-FAKE16-NEXT: v_lshrrev_b64 v[20:21], 24, v[69:70]
-; GFX11-FAKE16-NEXT: v_lshrrev_b64 v[21:22], 24, v[82:83]
-; GFX11-FAKE16-NEXT: v_lshrrev_b64 v[22:23], 24, v[85:86]
-; GFX11-FAKE16-NEXT: v_lshrrev_b64 v[23:24], 24, v[28:29]
-; GFX11-FAKE16-NEXT: v_lshrrev_b64 v[24:25], 24, v[26:27]
-; GFX11-FAKE16-NEXT: v_lshrrev_b32_e32 v34, 8, v37
-; GFX11-FAKE16-NEXT: v_lshrrev_b32_e32 v35, 16, v36
-; GFX11-FAKE16-NEXT: v_lshrrev_b32_e32 v36, 8, v36
-; GFX11-FAKE16-NEXT: v_lshrrev_b32_e32 v37, 24, v50
-; GFX11-FAKE16-NEXT: v_lshrrev_b32_e32 v49, 8, v49
-; GFX11-FAKE16-NEXT: v_lshrrev_b32_e32 v50, 24, v55
-; GFX11-FAKE16-NEXT: v_lshrrev_b32_e32 v54, 8, v54
-; GFX11-FAKE16-NEXT: v_lshrrev_b32_e32 v55, 24, v70
-; GFX11-FAKE16-NEXT: v_lshrrev_b32_e32 v69, 24, v83
-; GFX11-FAKE16-NEXT: v_lshrrev_b32_e32 v70, 8, v83
-; GFX11-FAKE16-NEXT: v_lshrrev_b32_e32 v82, 24, v86
-; GFX11-FAKE16-NEXT: v_lshrrev_b32_e32 v83, 8, v86
-; GFX11-FAKE16-NEXT: v_lshrrev_b32_e32 v85, 8, v85
-; GFX11-FAKE16-NEXT: v_lshrrev_b32_e32 v86, 24, v29
-; GFX11-FAKE16-NEXT: v_lshrrev_b32_e32 v29, 8, v29
-; GFX11-FAKE16-NEXT: v_lshrrev_b32_e32 v28, 8, v28
-; GFX11-FAKE16-NEXT: v_lshrrev_b32_e32 v27, 8, v27
-; GFX11-FAKE16-NEXT: v_lshrrev_b32_e32 v25, 8, v26
+; GFX11-FAKE16-NEXT: v_lshrrev_b64 v[23:24], 24, v[80:81]
+; GFX11-FAKE16-NEXT: v_lshrrev_b64 v[30:31], 24, v[20:21]
+; GFX11-FAKE16-NEXT: v_lshrrev_b64 v[17:18], 24, v[38:39]
+; GFX11-FAKE16-NEXT: v_lshrrev_b64 v[18:19], 24, v[51:52]
+; GFX11-FAKE16-NEXT: v_lshrrev_b32_e32 v34, 24, v39
+; GFX11-FAKE16-NEXT: v_lshrrev_b32_e32 v36, 8, v39
+; GFX11-FAKE16-NEXT: v_lshrrev_b32_e32 v37, 16, v38
+; GFX11-FAKE16-NEXT: v_lshrrev_b32_e32 v38, 8, v38
+; GFX11-FAKE16-NEXT: v_lshrrev_b32_e32 v39, 24, v52
+; GFX11-FAKE16-NEXT: v_lshrrev_b32_e32 v49, 8, v52
+; GFX11-FAKE16-NEXT: v_lshrrev_b32_e32 v50, 16, v51
+; GFX11-FAKE16-NEXT: v_lshrrev_b32_e32 v51, 8, v51
+; GFX11-FAKE16-NEXT: v_lshrrev_b32_e32 v52, 24, v65
+; GFX11-FAKE16-NEXT: v_lshrrev_b32_e32 v55, 16, v64
+; GFX11-FAKE16-NEXT: v_lshrrev_b32_e32 v64, 8, v64
+; GFX11-FAKE16-NEXT: v_lshrrev_b32_e32 v65, 24, v81
+; GFX11-FAKE16-NEXT: v_lshrrev_b32_e32 v80, 8, v85
+; GFX11-FAKE16-NEXT: v_lshrrev_b32_e32 v81, 16, v84
+; GFX11-FAKE16-NEXT: v_lshrrev_b32_e32 v84, 24, v97
+; GFX11-FAKE16-NEXT: v_lshrrev_b32_e32 v85, 8, v97
+; GFX11-FAKE16-NEXT: v_lshrrev_b32_e32 v96, 24, v26
+; GFX11-FAKE16-NEXT: v_lshrrev_b32_e32 v26, 8, v26
+; GFX11-FAKE16-NEXT: v_lshrrev_b32_e32 v97, 16, v25
+; GFX11-FAKE16-NEXT: v_lshrrev_b32_e32 v25, 8, v25
+; GFX11-FAKE16-NEXT: v_lshrrev_b32_e32 v21, 8, v21
+; GFX11-FAKE16-NEXT: v_lshrrev_b32_e32 v19, 8, v20
; GFX11-FAKE16-NEXT: .LBB108_4: ; %end
; GFX11-FAKE16-NEXT: s_or_b32 exec_lo, exec_lo, s0
; GFX11-FAKE16-NEXT: v_and_b32_e32 v1, 0xff, v1
; GFX11-FAKE16-NEXT: s_delay_alu instid0(VALU_DEP_2)
-; GFX11-FAKE16-NEXT: v_lshlrev_b16 v25, 8, v25
-; GFX11-FAKE16-NEXT: v_and_b32_e32 v26, 0xff, v97
-; GFX11-FAKE16-NEXT: v_lshlrev_b16 v24, 8, v24
+; GFX11-FAKE16-NEXT: v_lshlrev_b16 v19, 8, v19
+; GFX11-FAKE16-NEXT: v_and_b32_e32 v20, 0xff, v99
+; GFX11-FAKE16-NEXT: v_lshlrev_b16 v24, 8, v30
; GFX11-FAKE16-NEXT: v_and_b32_e32 v2, 0xff, v2
; GFX11-FAKE16-NEXT: v_and_b32_e32 v3, 0xff, v3
-; GFX11-FAKE16-NEXT: v_or_b32_e32 v1, v1, v25
-; GFX11-FAKE16-NEXT: v_lshlrev_b16 v25, 8, v27
-; GFX11-FAKE16-NEXT: v_or_b32_e32 v24, v26, v24
-; GFX11-FAKE16-NEXT: v_and_b32_e32 v26, 0xff, v81
-; GFX11-FAKE16-NEXT: v_lshlrev_b16 v27, 8, v96
+; GFX11-FAKE16-NEXT: v_or_b32_e32 v1, v1, v19
+; GFX11-FAKE16-NEXT: v_lshlrev_b16 v25, 8, v25
+; GFX11-FAKE16-NEXT: v_or_b32_e32 v19, v20, v24
+; GFX11-FAKE16-NEXT: v_lshlrev_b16 v20, 8, v21
+; GFX11-FAKE16-NEXT: v_and_b32_e32 v21, 0xff, v83
; GFX11-FAKE16-NEXT: v_and_b32_e32 v1, 0xffff, v1
-; GFX11-FAKE16-NEXT: v_lshlrev_b16 v28, 8, v28
-; GFX11-FAKE16-NEXT: v_lshlrev_b32_e32 v24, 16, v24
+; GFX11-FAKE16-NEXT: v_lshlrev_b16 v24, 8, v98
+; GFX11-FAKE16-NEXT: v_lshlrev_b32_e32 v19, 16, v19
+; GFX11-FAKE16-NEXT: v_and_b32_e32 v30, 0xff, v97
; GFX11-FAKE16-NEXT: v_and_b32_e32 v4, 0xff, v4
-; GFX11-FAKE16-NEXT: v_or_b32_e32 v2, v2, v25
-; GFX11-FAKE16-NEXT: v_or_b32_e32 v25, v26, v27
-; GFX11-FAKE16-NEXT: v_and_b32_e32 v81, 0xff, v87
-; GFX11-FAKE16-NEXT: v_or_b32_e32 v1, v1, v24
-; GFX11-FAKE16-NEXT: v_lshlrev_b16 v24, 8, v29
-; GFX11-FAKE16-NEXT: v_lshlrev_b16 v23, 8, v23
-; GFX11-FAKE16-NEXT: v_or_b32_e32 v3, v3, v28
-; GFX11-FAKE16-NEXT: v_lshlrev_b16 v26, 8, v86
+; GFX11-FAKE16-NEXT: v_lshlrev_b16 v26, 8, v26
+; GFX11-FAKE16-NEXT: v_or_b32_e32 v2, v2, v20
+; GFX11-FAKE16-NEXT: v_or_b32_e32 v1, v1, v19
+; GFX11-FAKE16-NEXT: v_lshlrev_b16 v19, 8, v29
+; GFX11-FAKE16-NEXT: v_or_b32_e32 v20, v21, v24
+; GFX11-FAKE16-NEXT: v_or_b32_e32 v3, v3, v25
+; GFX11-FAKE16-NEXT: v_or_b32_e32 v4, v4, v26
+; GFX11-FAKE16-NEXT: v_and_b32_e32 v21, 0xff, v70
+; GFX11-FAKE16-NEXT: v_or_b32_e32 v19, v30, v19
+; GFX11-FAKE16-NEXT: v_lshlrev_b16 v24, 8, v96
; GFX11-FAKE16-NEXT: v_and_b32_e32 v5, 0xff, v5
-; GFX11-FAKE16-NEXT: v_or_b32_e32 v4, v4, v24
-; GFX11-FAKE16-NEXT: v_lshlrev_b32_e32 v24, 16, v25
-; GFX11-FAKE16-NEXT: v_and_b32_e32 v25, 0xff, v68
-; GFX11-FAKE16-NEXT: v_lshlrev_b16 v27, 8, v85
-; GFX11-FAKE16-NEXT: v_and_b32_e32 v28, 0xff, v84
-; GFX11-FAKE16-NEXT: v_lshlrev_b16 v22, 8, v22
+; GFX11-FAKE16-NEXT: v_lshlrev_b16 v25, 8, v87
+; GFX11-FAKE16-NEXT: v_and_b32_e32 v26, 0xff, v86
+; GFX11-FAKE16-NEXT: v_lshlrev_b16 v28, 8, v28
; GFX11-FAKE16-NEXT: v_and_b32_e32 v6, 0xff, v6
-; GFX11-FAKE16-NEXT: v_lshlrev_b16 v29, 8, v83
-; GFX11-FAKE16-NEXT: v_and_b32_e32 v65, 0xff, v65
-; GFX11-FAKE16-NEXT: v_lshlrev_b16 v68, 8, v82
-; GFX11-FAKE16-NEXT: v_or_b32_e32 v23, v81, v23
-; GFX11-FAKE16-NEXT: v_or_b32_e32 v25, v25, v26
-; GFX11-FAKE16-NEXT: v_or_b32_e32 v5, v5, v27
-; GFX11-FAKE16-NEXT: v_or_b32_e32 v22, v28, v22
+; GFX11-FAKE16-NEXT: v_lshlrev_b16 v29, 8, v85
+; GFX11-FAKE16-NEXT: v_and_b32_e32 v30, 0xff, v67
+; GFX11-FAKE16-NEXT: v_lshlrev_b16 v31, 8, v84
+; GFX11-FAKE16-NEXT: v_or_b32_e32 v21, v21, v24
+; GFX11-FAKE16-NEXT: v_or_b32_e32 v5, v5, v25
+; GFX11-FAKE16-NEXT: v_or_b32_e32 v24, v26, v28
; GFX11-FAKE16-NEXT: v_or_b32_e32 v6, v6, v29
-; GFX11-FAKE16-NEXT: v_or_b32_e32 v26, v65, v68
+; GFX11-FAKE16-NEXT: v_or_b32_e32 v25, v30, v31
; GFX11-FAKE16-NEXT: v_and_b32_e32 v2, 0xffff, v2
+; GFX11-FAKE16-NEXT: v_lshlrev_b32_e32 v20, 16, v20
; GFX11-FAKE16-NEXT: v_and_b32_e32 v3, 0xffff, v3
-; GFX11-FAKE16-NEXT: v_lshlrev_b32_e32 v23, 16, v23
+; GFX11-FAKE16-NEXT: v_lshlrev_b32_e32 v19, 16, v19
; GFX11-FAKE16-NEXT: v_and_b32_e32 v4, 0xffff, v4
-; GFX11-FAKE16-NEXT: v_lshlrev_b32_e32 v25, 16, v25
+; GFX11-FAKE16-NEXT: v_lshlrev_b32_e32 v21, 16, v21
; GFX11-FAKE16-NEXT: v_and_b32_e32 v5, 0xffff, v5
-; GFX11-FAKE16-NEXT: v_lshlrev_b32_e32 v22, 16, v22
+; GFX11-FAKE16-NEXT: v_lshlrev_b32_e32 v24, 16, v24
; GFX11-FAKE16-NEXT: v_and_b32_e32 v6, 0xffff, v6
-; GFX11-FAKE16-NEXT: v_lshlrev_b32_e32 v26, 16, v26
-; GFX11-FAKE16-NEXT: v_or_b32_e32 v2, v2, v24
-; GFX11-FAKE16-NEXT: v_or_b32_e32 v3, v3, v23
-; GFX11-FAKE16-NEXT: v_or_b32_e32 v4, v4, v25
-; GFX11-FAKE16-NEXT: v_or_b32_e32 v5, v5, v22
-; GFX11-FAKE16-NEXT: v_or_b32_e32 v6, v6, v26
+; GFX11-FAKE16-NEXT: v_lshlrev_b32_e32 v25, 16, v25
+; GFX11-FAKE16-NEXT: v_or_b32_e32 v2, v2, v20
+; GFX11-FAKE16-NEXT: v_or_b32_e32 v3, v3, v19
+; GFX11-FAKE16-NEXT: v_or_b32_e32 v4, v4, v21
+; GFX11-FAKE16-NEXT: v_or_b32_e32 v5, v5, v24
+; GFX11-FAKE16-NEXT: v_or_b32_e32 v6, v6, v25
; GFX11-FAKE16-NEXT: v_and_b32_e32 v7, 0xff, v7
-; GFX11-FAKE16-NEXT: v_lshlrev_b16 v22, 8, v80
-; GFX11-FAKE16-NEXT: v_and_b32_e32 v23, 0xff, v71
-; GFX11-FAKE16-NEXT: v_lshlrev_b16 v21, 8, v21
+; GFX11-FAKE16-NEXT: v_lshlrev_b16 v19, 8, v82
+; GFX11-FAKE16-NEXT: v_and_b32_e32 v20, 0xff, v81
+; GFX11-FAKE16-NEXT: v_lshlrev_b16 v21, 8, v27
; GFX11-FAKE16-NEXT: v_and_b32_e32 v8, 0xff, v8
-; GFX11-FAKE16-NEXT: v_lshlrev_b16 v24, 8, v70
-; GFX11-FAKE16-NEXT: v_and_b32_e32 v25, 0xff, v52
-; GFX11-FAKE16-NEXT: v_lshlrev_b16 v26, 8, v69
+; GFX11-FAKE16-NEXT: v_lshlrev_b16 v24, 8, v80
+; GFX11-FAKE16-NEXT: v_and_b32_e32 v25, 0xff, v54
+; GFX11-FAKE16-NEXT: v_lshlrev_b16 v26, 8, v71
; GFX11-FAKE16-NEXT: v_and_b32_e32 v9, 0xff, v9
-; GFX11-FAKE16-NEXT: v_lshlrev_b16 v27, 8, v67
-; GFX11-FAKE16-NEXT: v_or_b32_e32 v7, v7, v22
-; GFX11-FAKE16-NEXT: v_or_b32_e32 v21, v23, v21
+; GFX11-FAKE16-NEXT: v_lshlrev_b16 v27, 8, v69
+; GFX11-FAKE16-NEXT: v_or_b32_e32 v7, v7, v19
+; GFX11-FAKE16-NEXT: v_or_b32_e32 v19, v20, v21
; GFX11-FAKE16-NEXT: v_or_b32_e32 v8, v8, v24
-; GFX11-FAKE16-NEXT: v_or_b32_e32 v22, v25, v26
+; GFX11-FAKE16-NEXT: v_or_b32_e32 v20, v25, v26
; GFX11-FAKE16-NEXT: v_or_b32_e32 v9, v9, v27
-; GFX11-FAKE16-NEXT: v_and_b32_e32 v23, 0xff, v66
-; GFX11-FAKE16-NEXT: v_lshlrev_b16 v20, 8, v20
+; GFX11-FAKE16-NEXT: v_and_b32_e32 v21, 0xff, v68
+; GFX11-FAKE16-NEXT: v_lshlrev_b16 v23, 8, v23
; GFX11-FAKE16-NEXT: v_and_b32_e32 v10, 0xff, v10
-; GFX11-FAKE16-NEXT: v_lshlrev_b16 v24, 8, v64
-; GFX11-FAKE16-NEXT: v_and_b32_e32 v25, 0xff, v38
-; GFX11-FAKE16-NEXT: v_lshlrev_b16 v26, 8, v55
+; GFX11-FAKE16-NEXT: v_lshlrev_b16 v24, 8, v66
+; GFX11-FAKE16-NEXT: v_and_b32_e32 v25, 0xff, v48
+; GFX11-FAKE16-NEXT: v_lshlrev_b16 v26, 8, v65
; GFX11-FAKE16-NEXT: v_and_b32_e32 v11, 0xff, v11
-; GFX11-FAKE16-NEXT: v_lshlrev_b16 v27, 8, v54
-; GFX11-FAKE16-NEXT: v_and_b32_e32 v28, 0xff, v53
-; GFX11-FAKE16-NEXT: v_lshlrev_b16 v19, 8, v19
-; GFX11-FAKE16-NEXT: v_or_b32_e32 v20, v23, v20
+; GFX11-FAKE16-NEXT: v_lshlrev_b16 v27, 8, v64
+; GFX11-FAKE16-NEXT: v_and_b32_e32 v28, 0xff, v55
+; GFX11-FAKE16-NEXT: v_lshlrev_b16 v22, 8, v22
+; GFX11-FAKE16-NEXT: v_or_b32_e32 v21, v21, v23
; GFX11-FAKE16-NEXT: v_or_b32_e32 v10, v10, v24
; GFX11-FAKE16-NEXT: v_or_b32_e32 v23, v25, v26
; GFX11-FAKE16-NEXT: v_or_b32_e32 v11, v11, v27
-; GFX11-FAKE16-NEXT: v_or_b32_e32 v19, v28, v19
+; GFX11-FAKE16-NEXT: v_or_b32_e32 v22, v28, v22
; GFX11-FAKE16-NEXT: v_and_b32_e32 v7, 0xffff, v7
-; GFX11-FAKE16-NEXT: v_lshlrev_b32_e32 v21, 16, v21
+; GFX11-FAKE16-NEXT: v_lshlrev_b32_e32 v19, 16, v19
; GFX11-FAKE16-NEXT: v_and_b32_e32 v8, 0xffff, v8
-; GFX11-FAKE16-NEXT: v_lshlrev_b32_e32 v22, 16, v22
-; GFX11-FAKE16-NEXT: v_and_b32_e32 v9, 0xffff, v9
; GFX11-FAKE16-NEXT: v_lshlrev_b32_e32 v20, 16, v20
+; GFX11-FAKE16-NEXT: v_and_b32_e32 v9, 0xffff, v9
+; GFX11-FAKE16-NEXT: v_lshlrev_b32_e32 v21, 16, v21
; GFX11-FAKE16-NEXT: v_and_b32_e32 v10, 0xffff, v10
; GFX11-FAKE16-NEXT: v_lshlrev_b32_e32 v23, 16, v23
; GFX11-FAKE16-NEXT: v_and_b32_e32 v11, 0xffff, v11
-; GFX11-FAKE16-NEXT: v_lshlrev_b32_e32 v19, 16, v19
-; GFX11-FAKE16-NEXT: v_or_b32_e32 v7, v7, v21
-; GFX11-FAKE16-NEXT: v_or_b32_e32 v8, v8, v22
-; GFX11-FAKE16-NEXT: v_or_b32_e32 v9, v9, v20
+; GFX11-FAKE16-NEXT: v_lshlrev_b32_e32 v22, 16, v22
+; GFX11-FAKE16-NEXT: v_or_b32_e32 v7, v7, v19
+; GFX11-FAKE16-NEXT: v_or_b32_e32 v8, v8, v20
+; GFX11-FAKE16-NEXT: v_or_b32_e32 v9, v9, v21
; GFX11-FAKE16-NEXT: v_or_b32_e32 v10, v10, v23
-; GFX11-FAKE16-NEXT: v_or_b32_e32 v11, v11, v19
+; GFX11-FAKE16-NEXT: v_or_b32_e32 v11, v11, v22
; GFX11-FAKE16-NEXT: v_and_b32_e32 v12, 0xff, v12
-; GFX11-FAKE16-NEXT: v_lshlrev_b16 v19, 8, v51
-; GFX11-FAKE16-NEXT: v_and_b32_e32 v20, 0xff, v33
-; GFX11-FAKE16-NEXT: v_lshlrev_b16 v21, 8, v50
+; GFX11-FAKE16-NEXT: v_lshlrev_b16 v19, 8, v53
+; GFX11-FAKE16-NEXT: v_and_b32_e32 v20, 0xff, v35
+; GFX11-FAKE16-NEXT: v_lshlrev_b16 v21, 8, v52
; GFX11-FAKE16-NEXT: v_and_b32_e32 v13, 0xff, v13
-; GFX11-FAKE16-NEXT: v_lshlrev_b16 v22, 8, v49
-; GFX11-FAKE16-NEXT: v_and_b32_e32 v23, 0xff, v48
+; GFX11-FAKE16-NEXT: v_lshlrev_b16 v22, 8, v51
+; GFX11-FAKE16-NEXT: v_and_b32_e32 v23, 0xff, v50
; GFX11-FAKE16-NEXT: v_lshlrev_b16 v18, 8, v18
; GFX11-FAKE16-NEXT: v_and_b32_e32 v14, 0xff, v14
-; GFX11-FAKE16-NEXT: v_lshlrev_b16 v24, 8, v39
+; GFX11-FAKE16-NEXT: v_lshlrev_b16 v24, 8, v49
; GFX11-FAKE16-NEXT: v_or_b32_e32 v12, v12, v19
; GFX11-FAKE16-NEXT: v_or_b32_e32 v19, v20, v21
; GFX11-FAKE16-NEXT: v_or_b32_e32 v13, v13, v22
; GFX11-FAKE16-NEXT: v_or_b32_e32 v18, v23, v18
; GFX11-FAKE16-NEXT: v_or_b32_e32 v14, v14, v24
-; GFX11-FAKE16-NEXT: v_and_b32_e32 v20, 0xff, v31
-; GFX11-FAKE16-NEXT: v_lshlrev_b16 v21, 8, v37
+; GFX11-FAKE16-NEXT: v_and_b32_e32 v20, 0xff, v33
+; GFX11-FAKE16-NEXT: v_lshlrev_b16 v21, 8, v39
; GFX11-FAKE16-NEXT: v_and_b32_e32 v15, 0xff, v15
-; GFX11-FAKE16-NEXT: v_lshlrev_b16 v22, 8, v36
-; GFX11-FAKE16-NEXT: v_and_b32_e32 v23, 0xff, v35
+; GFX11-FAKE16-NEXT: v_lshlrev_b16 v22, 8, v38
+; GFX11-FAKE16-NEXT: v_and_b32_e32 v23, 0xff, v37
; GFX11-FAKE16-NEXT: v_lshlrev_b16 v17, 8, v17
; GFX11-FAKE16-NEXT: v_and_b32_e32 v16, 0xff, v16
-; GFX11-FAKE16-NEXT: v_lshlrev_b16 v24, 8, v34
-; GFX11-FAKE16-NEXT: v_and_b32_e32 v25, 0xff, v30
-; GFX11-FAKE16-NEXT: v_lshlrev_b16 v26, 8, v32
+; GFX11-FAKE16-NEXT: v_lshlrev_b16 v24, 8, v36
+; GFX11-FAKE16-NEXT: v_and_b32_e32 v25, 0xff, v32
+; GFX11-FAKE16-NEXT: v_lshlrev_b16 v26, 8, v34
; GFX11-FAKE16-NEXT: v_or_b32_e32 v20, v20, v21
; GFX11-FAKE16-NEXT: v_or_b32_e32 v15, v15, v22
; GFX11-FAKE16-NEXT: v_or_b32_e32 v17, v23, v17
@@ -89766,21 +89686,21 @@ define inreg <64 x i8> @bitcast_v32bf16_to_v64i8_scalar(<32 x bfloat> inreg %a,
; SI-NEXT: v_mul_f32_e32 v42, 1.0, v2
; SI-NEXT: v_mul_f32_e32 v20, 1.0, v1
; SI-NEXT: v_mul_f32_e32 v25, 1.0, v4
-; SI-NEXT: v_mul_f32_e32 v28, 1.0, v3
+; SI-NEXT: v_mul_f32_e32 v29, 1.0, v3
; SI-NEXT: v_mul_f32_e32 v43, 1.0, v6
; SI-NEXT: v_mul_f32_e32 v23, 1.0, v5
; SI-NEXT: v_mul_f32_e32 v31, 1.0, v8
-; SI-NEXT: v_mul_f32_e32 v34, 1.0, v7
+; SI-NEXT: v_mul_f32_e32 v32, 1.0, v7
; SI-NEXT: v_mul_f32_e32 v44, 1.0, v10
-; SI-NEXT: v_mul_f32_e32 v29, 1.0, v9
+; SI-NEXT: v_mul_f32_e32 v28, 1.0, v9
; SI-NEXT: v_mul_f32_e32 v12, 1.0, v12
; SI-NEXT: v_mul_f32_e32 v35, 1.0, v11
; SI-NEXT: v_mul_f32_e32 v56, 1.0, v14
-; SI-NEXT: v_mul_f32_e32 v33, 1.0, v13
-; SI-NEXT: v_mul_f32_e32 v36, 1.0, v16
+; SI-NEXT: v_mul_f32_e32 v38, 1.0, v13
+; SI-NEXT: v_mul_f32_e32 v14, 1.0, v16
; SI-NEXT: v_mul_f32_e32 v39, 1.0, v15
-; SI-NEXT: v_mul_f32_e32 v48, 1.0, v18
-; SI-NEXT: v_mul_f32_e32 v32, 1.0, v17
+; SI-NEXT: v_mul_f32_e32 v33, 1.0, v18
+; SI-NEXT: v_mul_f32_e32 v45, 1.0, v17
; SI-NEXT: v_mul_f32_e64 v3, 1.0, s16
; SI-NEXT: v_mul_f32_e64 v22, 1.0, s19
; SI-NEXT: v_mul_f32_e64 v2, 1.0, s18
@@ -89794,25 +89714,19 @@ define inreg <64 x i8> @bitcast_v32bf16_to_v64i8_scalar(<32 x bfloat> inreg %a,
; SI-NEXT: v_mul_f32_e64 v7, 1.0, s26
; SI-NEXT: v_mul_f32_e64 v10, 1.0, s29
; SI-NEXT: v_mul_f32_e64 v11, 1.0, s28
-; SI-NEXT: buffer_store_dword v19, off, s[0:3], s32 offset:124 ; 4-byte Folded Spill
-; SI-NEXT: buffer_store_dword v2, off, s[0:3], s32 offset:128 ; 4-byte Folded Spill
-; SI-NEXT: buffer_store_dword v3, off, s[0:3], s32 offset:132 ; 4-byte Folded Spill
-; SI-NEXT: buffer_store_dword v4, off, s[0:3], s32 offset:136 ; 4-byte Folded Spill
-; SI-NEXT: buffer_store_dword v5, off, s[0:3], s32 offset:140 ; 4-byte Folded Spill
-; SI-NEXT: buffer_store_dword v20, off, s[0:3], s32 offset:144 ; 4-byte Folded Spill
-; SI-NEXT: buffer_store_dword v6, off, s[0:3], s32 offset:148 ; 4-byte Folded Spill
-; SI-NEXT: buffer_store_dword v7, off, s[0:3], s32 offset:152 ; 4-byte Folded Spill
-; SI-NEXT: buffer_store_dword v8, off, s[0:3], s32 offset:156 ; 4-byte Folded Spill
-; SI-NEXT: buffer_store_dword v9, off, s[0:3], s32 offset:160 ; 4-byte Folded Spill
-; SI-NEXT: buffer_store_dword v10, off, s[0:3], s32 offset:164 ; 4-byte Folded Spill
-; SI-NEXT: buffer_store_dword v25, off, s[0:3], s32 offset:168 ; 4-byte Folded Spill
-; SI-NEXT: buffer_store_dword v11, off, s[0:3], s32 offset:172 ; 4-byte Folded Spill
-; SI-NEXT: buffer_store_dword v28, off, s[0:3], s32 offset:176 ; 4-byte Folded Spill
-; SI-NEXT: buffer_store_dword v31, off, s[0:3], s32 offset:180 ; 4-byte Folded Spill
-; SI-NEXT: buffer_store_dword v22, off, s[0:3], s32 offset:184 ; 4-byte Folded Spill
-; SI-NEXT: buffer_store_dword v34, off, s[0:3], s32 offset:188 ; 4-byte Folded Spill
-; SI-NEXT: buffer_store_dword v12, off, s[0:3], s32 offset:192 ; 4-byte Folded Spill
-; SI-NEXT: buffer_store_dword v35, off, s[0:3], s32 offset:196 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v19, off, s[0:3], s32 offset:144 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v2, off, s[0:3], s32 offset:148 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v3, off, s[0:3], s32 offset:152 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v4, off, s[0:3], s32 offset:156 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v5, off, s[0:3], s32 offset:160 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v6, off, s[0:3], s32 offset:164 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v7, off, s[0:3], s32 offset:168 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v8, off, s[0:3], s32 offset:172 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v9, off, s[0:3], s32 offset:176 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v20, off, s[0:3], s32 offset:180 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v10, off, s[0:3], s32 offset:184 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v11, off, s[0:3], s32 offset:188 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v25, off, s[0:3], s32 offset:192 ; 4-byte Folded Spill
; SI-NEXT: s_cbranch_scc0 .LBB109_4
; SI-NEXT: ; %bb.1: ; %cmp.false
; SI-NEXT: v_lshrrev_b32_e32 v1, 16, v19
@@ -89868,213 +89782,219 @@ define inreg <64 x i8> @bitcast_v32bf16_to_v64i8_scalar(<32 x bfloat> inreg %a,
; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:84 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_lshrrev_b32_e32 v1, 16, v25
-; SI-NEXT: v_alignbit_b32 v6, v1, v28, 16
+; SI-NEXT: v_alignbit_b32 v6, v1, v29, 16
; SI-NEXT: v_lshrrev_b32_e32 v1, 16, v31
-; SI-NEXT: v_alignbit_b32 v3, v1, v34, 16
+; SI-NEXT: v_alignbit_b32 v3, v1, v32, 16
+; SI-NEXT: v_mov_b32_e32 v48, v12
; SI-NEXT: v_lshrrev_b32_e32 v1, 16, v12
-; SI-NEXT: v_lshrrev_b32_e32 v7, 16, v56
+; SI-NEXT: v_lshrrev_b32_e32 v12, 16, v44
; SI-NEXT: v_alignbit_b32 v2, v1, v35, 16
-; SI-NEXT: v_alignbit_b32 v8, v7, v33, 16
+; SI-NEXT: v_mov_b32_e32 v51, v14
+; SI-NEXT: v_lshrrev_b32_e32 v1, 16, v14
+; SI-NEXT: v_alignbit_b32 v14, v12, v28, 16
+; SI-NEXT: v_alignbit_b32 v4, v14, v3, 24
+; SI-NEXT: v_lshrrev_b32_e32 v7, 16, v56
+; SI-NEXT: buffer_store_dword v4, off, s[0:3], s32 offset:132 ; 4-byte Folded Spill
+; SI-NEXT: s_waitcnt expcnt(0)
+; SI-NEXT: v_alignbit_b32 v4, v14, v3, 16
+; SI-NEXT: v_alignbit_b32 v8, v7, v38, 16
+; SI-NEXT: buffer_store_dword v4, off, s[0:3], s32 offset:140 ; 4-byte Folded Spill
+; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_alignbit_b32 v4, v8, v2, 24
-; SI-NEXT: v_lshrrev_b32_e32 v1, 16, v36
; SI-NEXT: buffer_store_dword v4, off, s[0:3], s32 offset:120 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_lshrrev_b32_e32 v4, 16, v48
+; SI-NEXT: v_alignbit_b32 v4, v8, v2, 16
+; SI-NEXT: buffer_store_dword v4, off, s[0:3], s32 offset:128 ; 4-byte Folded Spill
+; SI-NEXT: s_waitcnt expcnt(0)
+; SI-NEXT: v_alignbit_b32 v4, v8, v2, 8
+; SI-NEXT: buffer_store_dword v4, off, s[0:3], s32 offset:136 ; 4-byte Folded Spill
+; SI-NEXT: s_waitcnt expcnt(0)
+; SI-NEXT: v_lshrrev_b32_e32 v4, 16, v33
; SI-NEXT: v_alignbit_b32 v1, v1, v39, 16
; SI-NEXT: v_lshrrev_b32_e32 v18, 16, v43
-; SI-NEXT: v_lshrrev_b32_e32 v12, 16, v44
-; SI-NEXT: v_alignbit_b32 v5, v4, v32, 16
-; SI-NEXT: v_mov_b32_e32 v31, v23
+; SI-NEXT: v_alignbit_b32 v5, v4, v45, 16
+; SI-NEXT: v_mov_b32_e32 v37, v32
+; SI-NEXT: v_mov_b32_e32 v32, v23
; SI-NEXT: v_alignbit_b32 v20, v18, v23, 16
-; SI-NEXT: v_alignbit_b32 v14, v12, v29, 16
; SI-NEXT: v_alignbit_b32 v23, v5, v1, 24
-; SI-NEXT: v_mov_b32_e32 v38, v36
-; SI-NEXT: v_alignbit_b32 v36, v20, v6, 24
-; SI-NEXT: v_alignbit_b32 v25, v14, v3, 24
-; SI-NEXT: v_alignbit_b32 v50, v8, v2, 16
-; SI-NEXT: v_mov_b32_e32 v53, v32
; SI-NEXT: buffer_store_dword v23, off, s[0:3], s32 offset:112 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_alignbit_b32 v23, v5, v1, 16
-; SI-NEXT: v_alignbit_b32 v32, v5, v1, 8
+; SI-NEXT: v_mov_b32_e32 v34, v29
+; SI-NEXT: buffer_store_dword v23, off, s[0:3], s32 offset:116 ; 4-byte Folded Spill
+; SI-NEXT: s_waitcnt expcnt(0)
+; SI-NEXT: v_alignbit_b32 v23, v5, v1, 8
+; SI-NEXT: v_mov_b32_e32 v29, v26
+; SI-NEXT: v_lshrrev_b32_e32 v61, 24, v26
+; SI-NEXT: v_mov_b32_e32 v26, v42
+; SI-NEXT: v_lshrrev_b32_e32 v63, 24, v42
+; SI-NEXT: v_mov_b32_e32 v42, v33
+; SI-NEXT: v_mov_b32_e32 v36, v31
+; SI-NEXT: v_mov_b32_e32 v49, v35
+; SI-NEXT: v_mov_b32_e32 v52, v39
+; SI-NEXT: v_alignbit_b32 v53, v20, v6, 24
; SI-NEXT: v_alignbit_b32 v55, v20, v6, 16
; SI-NEXT: v_alignbit_b32 v40, v20, v6, 8
-; SI-NEXT: v_mov_b32_e32 v35, v29
-; SI-NEXT: v_alignbit_b32 v52, v14, v3, 16
+; SI-NEXT: v_mov_b32_e32 v35, v28
; SI-NEXT: v_alignbit_b32 v54, v14, v3, 8
-; SI-NEXT: v_mov_b32_e32 v37, v33
-; SI-NEXT: v_alignbit_b32 v51, v8, v2, 8
-; SI-NEXT: buffer_store_dword v23, off, s[0:3], s32 offset:116 ; 4-byte Folded Spill
+; SI-NEXT: v_mov_b32_e32 v39, v38
+; SI-NEXT: v_mov_b32_e32 v50, v45
+; SI-NEXT: buffer_store_dword v23, off, s[0:3], s32 offset:124 ; 4-byte Folded Spill
+; SI-NEXT: v_mov_b32_e32 v25, v22
; SI-NEXT: v_lshrrev_b32_e32 v22, 24, v22
; SI-NEXT: v_lshrrev_b32_e32 v62, 8, v30
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_mov_b32_e32 v23, v41
; SI-NEXT: v_lshrrev_b32_e32 v41, 24, v41
; SI-NEXT: v_lshrrev_b32_e32 v46, 8, v19
-; SI-NEXT: v_mov_b32_e32 v28, v26
-; SI-NEXT: v_lshrrev_b32_e32 v61, 24, v26
; SI-NEXT: v_lshrrev_b32_e32 v57, 8, v16
-; SI-NEXT: v_mov_b32_e32 v26, v42
-; SI-NEXT: v_lshrrev_b32_e32 v63, 24, v42
; SI-NEXT: v_lshrrev_b32_e32 v58, 8, v11
-; SI-NEXT: v_mov_b32_e32 v29, v43
+; SI-NEXT: v_mov_b32_e32 v31, v43
; SI-NEXT: v_lshrrev_b32_e32 v59, 24, v43
; SI-NEXT: v_lshrrev_b32_e32 v47, 8, v20
-; SI-NEXT: v_mov_b32_e32 v34, v44
+; SI-NEXT: v_mov_b32_e32 v28, v44
; SI-NEXT: v_lshrrev_b32_e32 v45, 24, v44
; SI-NEXT: v_lshrrev_b32_e32 v60, 8, v14
-; SI-NEXT: v_mov_b32_e32 v33, v56
+; SI-NEXT: v_mov_b32_e32 v38, v56
; SI-NEXT: v_lshrrev_b32_e32 v43, 24, v56
; SI-NEXT: v_lshrrev_b32_e32 v56, 8, v8
-; SI-NEXT: v_mov_b32_e32 v49, v48
-; SI-NEXT: v_lshrrev_b32_e32 v42, 24, v48
-; SI-NEXT: v_mov_b32_e32 v48, v32
-; SI-NEXT: v_mov_b32_e32 v32, v50
-; SI-NEXT: v_mov_b32_e32 v50, v25
-; SI-NEXT: v_mov_b32_e32 v25, v36
-; SI-NEXT: v_mov_b32_e32 v36, v38
+; SI-NEXT: v_lshrrev_b32_e32 v42, 24, v42
; SI-NEXT: v_lshrrev_b32_e32 v44, 8, v5
; SI-NEXT: s_cbranch_execnz .LBB109_3
; SI-NEXT: .LBB109_2: ; %cmp.true
-; SI-NEXT: v_and_b32_e32 v2, 0xffff0000, v36
-; SI-NEXT: v_and_b32_e32 v1, 0xffff0000, v39
+; SI-NEXT: v_and_b32_e32 v9, 0xffff0000, v28
+; SI-NEXT: v_add_f32_e32 v44, 0x40c00000, v9
+; SI-NEXT: buffer_load_dword v9, off, s[0:3], s32 offset:192 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v15, off, s[0:3], s32 offset:172 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v22, off, s[0:3], s32 offset:152 ; 4-byte Folded Reload
+; SI-NEXT: v_and_b32_e32 v2, 0xffff0000, v51
+; SI-NEXT: v_and_b32_e32 v1, 0xffff0000, v52
; SI-NEXT: v_add_f32_e32 v2, 0x40c00000, v2
; SI-NEXT: v_add_f32_e32 v1, 0x40c00000, v1
; SI-NEXT: v_lshrrev_b32_e32 v2, 16, v2
-; SI-NEXT: v_and_b32_e32 v3, 0xffff0000, v49
+; SI-NEXT: v_and_b32_e32 v3, 0xffff0000, v33
; SI-NEXT: v_alignbit_b32 v1, v2, v1, 16
-; SI-NEXT: v_and_b32_e32 v2, 0xffff0000, v53
+; SI-NEXT: v_and_b32_e32 v2, 0xffff0000, v50
; SI-NEXT: v_add_f32_e32 v42, 0x40c00000, v3
; SI-NEXT: v_add_f32_e32 v2, 0x40c00000, v2
; SI-NEXT: v_lshrrev_b32_e32 v4, 16, v42
+; SI-NEXT: v_and_b32_e32 v3, 0xffff0000, v48
; SI-NEXT: v_alignbit_b32 v5, v4, v2, 16
-; SI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:196 ; 4-byte Folded Reload
-; SI-NEXT: buffer_load_dword v3, off, s[0:3], s32 offset:192 ; 4-byte Folded Reload
-; SI-NEXT: v_and_b32_e32 v6, 0xffff0000, v33
-; SI-NEXT: v_add_f32_e32 v43, 0x40c00000, v6
-; SI-NEXT: v_lshrrev_b32_e32 v7, 16, v43
-; SI-NEXT: v_and_b32_e32 v9, 0xffff0000, v34
-; SI-NEXT: v_add_f32_e32 v44, 0x40c00000, v9
-; SI-NEXT: v_lshrrev_b32_e32 v12, 16, v44
-; SI-NEXT: v_and_b32_e32 v10, 0xffff0000, v29
-; SI-NEXT: v_add_f32_e32 v45, 0x40c00000, v10
-; SI-NEXT: v_lshrrev_b32_e32 v18, 16, v45
-; SI-NEXT: v_alignbit_b32 v48, v5, v1, 8
-; SI-NEXT: v_lshrrev_b32_e32 v43, 24, v43
-; SI-NEXT: v_lshrrev_b32_e32 v42, 24, v42
-; SI-NEXT: buffer_load_dword v6, off, s[0:3], s32 offset:180 ; 4-byte Folded Reload
-; SI-NEXT: buffer_load_dword v9, off, s[0:3], s32 offset:168 ; 4-byte Folded Reload
-; SI-NEXT: buffer_load_dword v10, off, s[0:3], s32 offset:164 ; 4-byte Folded Reload
-; SI-NEXT: buffer_load_dword v13, off, s[0:3], s32 offset:160 ; 4-byte Folded Reload
-; SI-NEXT: buffer_load_dword v17, off, s[0:3], s32 offset:148 ; 4-byte Folded Reload
-; SI-NEXT: buffer_load_dword v19, off, s[0:3], s32 offset:140 ; 4-byte Folded Reload
-; SI-NEXT: buffer_load_dword v15, off, s[0:3], s32 offset:156 ; 4-byte Folded Reload
-; SI-NEXT: buffer_load_dword v22, off, s[0:3], s32 offset:132 ; 4-byte Folded Reload
-; SI-NEXT: s_waitcnt vmcnt(9)
-; SI-NEXT: v_and_b32_e32 v2, 0xffff0000, v2
-; SI-NEXT: s_waitcnt vmcnt(8)
-; SI-NEXT: v_and_b32_e32 v3, 0xffff0000, v3
+; SI-NEXT: v_and_b32_e32 v2, 0xffff0000, v49
; SI-NEXT: v_add_f32_e32 v3, 0x40c00000, v3
; SI-NEXT: v_add_f32_e32 v2, 0x40c00000, v2
; SI-NEXT: v_lshrrev_b32_e32 v3, 16, v3
+; SI-NEXT: v_and_b32_e32 v6, 0xffff0000, v38
; SI-NEXT: v_alignbit_b32 v2, v3, v2, 16
-; SI-NEXT: v_and_b32_e32 v3, 0xffff0000, v37
+; SI-NEXT: v_and_b32_e32 v3, 0xffff0000, v39
+; SI-NEXT: v_add_f32_e32 v43, 0x40c00000, v6
; SI-NEXT: v_add_f32_e32 v3, 0x40c00000, v3
+; SI-NEXT: v_lshrrev_b32_e32 v7, 16, v43
+; SI-NEXT: v_and_b32_e32 v6, 0xffff0000, v36
; SI-NEXT: v_alignbit_b32 v8, v7, v3, 16
-; SI-NEXT: buffer_load_dword v3, off, s[0:3], s32 offset:188 ; 4-byte Folded Reload
-; SI-NEXT: v_alignbit_b32 v32, v8, v2, 16
-; SI-NEXT: v_alignbit_b32 v51, v8, v2, 8
-; SI-NEXT: s_waitcnt vmcnt(8)
-; SI-NEXT: v_and_b32_e32 v6, 0xffff0000, v6
+; SI-NEXT: v_and_b32_e32 v3, 0xffff0000, v37
; SI-NEXT: v_add_f32_e32 v6, 0x40c00000, v6
+; SI-NEXT: v_add_f32_e32 v3, 0x40c00000, v3
; SI-NEXT: v_lshrrev_b32_e32 v6, 16, v6
-; SI-NEXT: s_waitcnt vmcnt(7)
+; SI-NEXT: v_alignbit_b32 v3, v6, v3, 16
+; SI-NEXT: v_and_b32_e32 v6, 0xffff0000, v35
+; SI-NEXT: v_and_b32_e32 v10, 0xffff0000, v31
+; SI-NEXT: v_add_f32_e32 v6, 0x40c00000, v6
+; SI-NEXT: v_lshrrev_b32_e32 v12, 16, v44
+; SI-NEXT: v_add_f32_e32 v45, 0x40c00000, v10
+; SI-NEXT: buffer_load_dword v10, off, s[0:3], s32 offset:184 ; 4-byte Folded Reload
+; SI-NEXT: v_alignbit_b32 v14, v12, v6, 16
+; SI-NEXT: v_and_b32_e32 v6, 0xffff0000, v34
+; SI-NEXT: v_add_f32_e32 v6, 0x40c00000, v6
+; SI-NEXT: v_lshrrev_b32_e32 v18, 16, v45
+; SI-NEXT: buffer_load_dword v13, off, s[0:3], s32 offset:176 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v17, off, s[0:3], s32 offset:164 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v19, off, s[0:3], s32 offset:160 ; 4-byte Folded Reload
+; SI-NEXT: v_alignbit_b32 v54, v14, v3, 8
+; SI-NEXT: v_lshrrev_b32_e32 v43, 24, v43
+; SI-NEXT: v_lshrrev_b32_e32 v42, 24, v42
+; SI-NEXT: v_lshrrev_b32_e32 v60, 8, v14
+; SI-NEXT: s_waitcnt vmcnt(6)
; SI-NEXT: v_and_b32_e32 v9, 0xffff0000, v9
; SI-NEXT: v_add_f32_e32 v9, 0x40c00000, v9
; SI-NEXT: v_lshrrev_b32_e32 v9, 16, v9
-; SI-NEXT: s_waitcnt vmcnt(6)
+; SI-NEXT: v_alignbit_b32 v6, v9, v6, 16
+; SI-NEXT: v_and_b32_e32 v9, 0xffff0000, v32
+; SI-NEXT: v_add_f32_e32 v9, 0x40c00000, v9
+; SI-NEXT: v_alignbit_b32 v20, v18, v9, 16
+; SI-NEXT: buffer_load_dword v9, off, s[0:3], s32 offset:188 ; 4-byte Folded Reload
+; SI-NEXT: v_alignbit_b32 v53, v20, v6, 24
+; SI-NEXT: v_alignbit_b32 v55, v20, v6, 16
+; SI-NEXT: v_alignbit_b32 v40, v20, v6, 8
+; SI-NEXT: s_waitcnt vmcnt(4)
; SI-NEXT: v_and_b32_e32 v10, 0xffff0000, v10
; SI-NEXT: v_add_f32_e32 v10, 0x40c00000, v10
; SI-NEXT: v_lshrrev_b32_e32 v10, 16, v10
-; SI-NEXT: s_waitcnt vmcnt(2)
-; SI-NEXT: v_and_b32_e32 v15, 0xffff0000, v15
-; SI-NEXT: v_and_b32_e32 v19, 0xffff0000, v19
+; SI-NEXT: s_waitcnt vmcnt(3)
; SI-NEXT: v_and_b32_e32 v13, 0xffff0000, v13
-; SI-NEXT: v_add_f32_e32 v15, 0x40c00000, v15
+; SI-NEXT: s_waitcnt vmcnt(2)
; SI-NEXT: v_and_b32_e32 v17, 0xffff0000, v17
+; SI-NEXT: s_waitcnt vmcnt(1)
+; SI-NEXT: v_and_b32_e32 v19, 0xffff0000, v19
; SI-NEXT: v_add_f32_e32 v19, 0x40c00000, v19
-; SI-NEXT: v_add_f32_e32 v13, 0x40c00000, v13
-; SI-NEXT: v_lshrrev_b32_e32 v15, 16, v15
; SI-NEXT: v_add_f32_e32 v17, 0x40c00000, v17
; SI-NEXT: v_lshrrev_b32_e32 v19, 16, v19
-; SI-NEXT: v_alignbit_b32 v15, v15, v13, 16
-; SI-NEXT: buffer_load_dword v13, off, s[0:3], s32 offset:152 ; 4-byte Folded Reload
; SI-NEXT: v_alignbit_b32 v21, v19, v17, 16
-; SI-NEXT: buffer_load_dword v17, off, s[0:3], s32 offset:136 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v17, off, s[0:3], s32 offset:156 ; 4-byte Folded Reload
+; SI-NEXT: v_and_b32_e32 v15, 0xffff0000, v15
+; SI-NEXT: v_add_f32_e32 v13, 0x40c00000, v13
+; SI-NEXT: v_add_f32_e32 v15, 0x40c00000, v15
+; SI-NEXT: v_lshrrev_b32_e32 v15, 16, v15
+; SI-NEXT: v_alignbit_b32 v15, v15, v13, 16
+; SI-NEXT: buffer_load_dword v13, off, s[0:3], s32 offset:168 ; 4-byte Folded Reload
; SI-NEXT: s_waitcnt vmcnt(2)
-; SI-NEXT: v_and_b32_e32 v3, 0xffff0000, v3
-; SI-NEXT: v_add_f32_e32 v3, 0x40c00000, v3
-; SI-NEXT: v_alignbit_b32 v3, v6, v3, 16
-; SI-NEXT: v_and_b32_e32 v6, 0xffff0000, v35
-; SI-NEXT: v_add_f32_e32 v6, 0x40c00000, v6
-; SI-NEXT: v_alignbit_b32 v14, v12, v6, 16
-; SI-NEXT: buffer_load_dword v6, off, s[0:3], s32 offset:176 ; 4-byte Folded Reload
-; SI-NEXT: v_alignbit_b32 v50, v14, v3, 24
-; SI-NEXT: v_alignbit_b32 v52, v14, v3, 16
-; SI-NEXT: v_alignbit_b32 v54, v14, v3, 8
-; SI-NEXT: v_lshrrev_b32_e32 v60, 8, v14
+; SI-NEXT: v_and_b32_e32 v9, 0xffff0000, v9
+; SI-NEXT: v_add_f32_e32 v9, 0x40c00000, v9
+; SI-NEXT: v_alignbit_b32 v10, v10, v9, 16
+; SI-NEXT: buffer_load_dword v9, off, s[0:3], s32 offset:180 ; 4-byte Folded Reload
; SI-NEXT: s_waitcnt vmcnt(2)
-; SI-NEXT: v_and_b32_e32 v13, 0xffff0000, v13
-; SI-NEXT: v_add_f32_e32 v16, 0x40c00000, v13
-; SI-NEXT: s_waitcnt vmcnt(1)
; SI-NEXT: v_and_b32_e32 v17, 0xffff0000, v17
; SI-NEXT: v_add_f32_e32 v19, 0x40c00000, v17
; SI-NEXT: v_and_b32_e32 v17, 0xffff0000, v23
-; SI-NEXT: buffer_load_dword v23, off, s[0:3], s32 offset:124 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v23, off, s[0:3], s32 offset:144 ; 4-byte Folded Reload
; SI-NEXT: v_and_b32_e32 v22, 0xffff0000, v22
-; SI-NEXT: v_and_b32_e32 v13, 0xffff0000, v28
; SI-NEXT: v_add_f32_e32 v22, 0x40c00000, v22
; SI-NEXT: v_add_f32_e32 v41, 0x40c00000, v17
-; SI-NEXT: v_add_f32_e32 v56, 0x40c00000, v13
; SI-NEXT: v_lshrrev_b32_e32 v17, 16, v41
-; SI-NEXT: v_lshrrev_b32_e32 v13, 16, v56
; SI-NEXT: v_lshrrev_b32_e32 v41, 24, v41
-; SI-NEXT: v_lshrrev_b32_e32 v61, 24, v56
-; SI-NEXT: v_lshrrev_b32_e32 v56, 8, v8
; SI-NEXT: v_alignbit_b32 v19, v17, v19, 16
-; SI-NEXT: v_alignbit_b32 v16, v13, v16, 16
; SI-NEXT: v_lshrrev_b32_e32 v46, 8, v19
-; SI-NEXT: v_lshrrev_b32_e32 v57, 8, v16
-; SI-NEXT: s_waitcnt vmcnt(1)
-; SI-NEXT: v_and_b32_e32 v6, 0xffff0000, v6
-; SI-NEXT: v_add_f32_e32 v6, 0x40c00000, v6
-; SI-NEXT: v_alignbit_b32 v6, v9, v6, 16
-; SI-NEXT: v_and_b32_e32 v9, 0xffff0000, v31
-; SI-NEXT: v_add_f32_e32 v9, 0x40c00000, v9
-; SI-NEXT: v_alignbit_b32 v20, v18, v9, 16
-; SI-NEXT: buffer_load_dword v9, off, s[0:3], s32 offset:172 ; 4-byte Folded Reload
-; SI-NEXT: v_alignbit_b32 v25, v20, v6, 24
-; SI-NEXT: v_alignbit_b32 v55, v20, v6, 16
-; SI-NEXT: v_alignbit_b32 v40, v20, v6, 8
; SI-NEXT: s_waitcnt vmcnt(1)
+; SI-NEXT: v_and_b32_e32 v9, 0xffff0000, v9
+; SI-NEXT: v_add_f32_e32 v11, 0x40c00000, v9
+; SI-NEXT: v_and_b32_e32 v9, 0xffff0000, v26
+; SI-NEXT: v_add_f32_e32 v47, 0x40c00000, v9
+; SI-NEXT: v_lshrrev_b32_e32 v9, 16, v47
+; SI-NEXT: v_alignbit_b32 v11, v9, v11, 16
+; SI-NEXT: v_lshrrev_b32_e32 v58, 8, v11
+; SI-NEXT: v_lshrrev_b32_e32 v63, 24, v47
+; SI-NEXT: v_lshrrev_b32_e32 v47, 8, v20
+; SI-NEXT: s_waitcnt vmcnt(0)
; SI-NEXT: v_and_b32_e32 v23, 0xffff0000, v23
; SI-NEXT: v_add_f32_e32 v23, 0x40c00000, v23
; SI-NEXT: v_lshrrev_b32_e32 v23, 16, v23
; SI-NEXT: v_alignbit_b32 v27, v23, v22, 16
-; SI-NEXT: buffer_load_dword v23, off, s[0:3], s32 offset:184 ; 4-byte Folded Reload
-; SI-NEXT: buffer_load_dword v22, off, s[0:3], s32 offset:128 ; 4-byte Folded Reload
-; SI-NEXT: s_waitcnt vmcnt(2)
-; SI-NEXT: v_and_b32_e32 v9, 0xffff0000, v9
-; SI-NEXT: v_add_f32_e32 v9, 0x40c00000, v9
-; SI-NEXT: v_alignbit_b32 v10, v10, v9, 16
-; SI-NEXT: buffer_load_dword v9, off, s[0:3], s32 offset:144 ; 4-byte Folded Reload
-; SI-NEXT: s_waitcnt vmcnt(2)
-; SI-NEXT: v_and_b32_e32 v23, 0xffff0000, v23
-; SI-NEXT: s_waitcnt vmcnt(1)
-; SI-NEXT: v_and_b32_e32 v22, 0xffff0000, v22
+; SI-NEXT: buffer_load_dword v22, off, s[0:3], s32 offset:148 ; 4-byte Folded Reload
+; SI-NEXT: v_and_b32_e32 v23, 0xffff0000, v25
+; SI-NEXT: v_and_b32_e32 v13, 0xffff0000, v13
; SI-NEXT: v_add_f32_e32 v59, 0x40c00000, v23
-; SI-NEXT: v_add_f32_e32 v22, 0x40c00000, v22
+; SI-NEXT: v_add_f32_e32 v16, 0x40c00000, v13
+; SI-NEXT: v_and_b32_e32 v13, 0xffff0000, v29
; SI-NEXT: v_lshrrev_b32_e32 v24, 16, v59
+; SI-NEXT: v_add_f32_e32 v56, 0x40c00000, v13
+; SI-NEXT: v_lshrrev_b32_e32 v13, 16, v56
+; SI-NEXT: v_lshrrev_b32_e32 v61, 24, v56
+; SI-NEXT: v_lshrrev_b32_e32 v56, 8, v8
+; SI-NEXT: v_alignbit_b32 v16, v13, v16, 16
+; SI-NEXT: v_lshrrev_b32_e32 v57, 8, v16
+; SI-NEXT: s_waitcnt vmcnt(0)
+; SI-NEXT: v_and_b32_e32 v22, 0xffff0000, v22
+; SI-NEXT: v_add_f32_e32 v22, 0x40c00000, v22
; SI-NEXT: v_alignbit_b32 v30, v24, v22, 16
; SI-NEXT: v_alignbit_b32 v22, v30, v27, 24
; SI-NEXT: v_lshrrev_b32_e32 v62, 8, v30
@@ -90103,18 +90023,8 @@ define inreg <64 x i8> @bitcast_v32bf16_to_v64i8_scalar(<32 x bfloat> inreg %a,
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_alignbit_b32 v22, v16, v15, 8
; SI-NEXT: buffer_store_dword v22, off, s[0:3], s32 offset:88 ; 4-byte Folded Spill
-; SI-NEXT: s_waitcnt vmcnt(9)
-; SI-NEXT: v_and_b32_e32 v9, 0xffff0000, v9
-; SI-NEXT: v_add_f32_e32 v11, 0x40c00000, v9
-; SI-NEXT: v_and_b32_e32 v9, 0xffff0000, v26
-; SI-NEXT: v_add_f32_e32 v47, 0x40c00000, v9
-; SI-NEXT: v_lshrrev_b32_e32 v9, 16, v47
-; SI-NEXT: v_alignbit_b32 v11, v9, v11, 16
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_alignbit_b32 v22, v11, v10, 24
-; SI-NEXT: v_lshrrev_b32_e32 v58, 8, v11
-; SI-NEXT: v_lshrrev_b32_e32 v63, 24, v47
-; SI-NEXT: v_lshrrev_b32_e32 v47, 8, v20
; SI-NEXT: buffer_store_dword v22, off, s[0:3], s32 offset:64 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_alignbit_b32 v22, v11, v10, 16
@@ -90123,15 +90033,30 @@ define inreg <64 x i8> @bitcast_v32bf16_to_v64i8_scalar(<32 x bfloat> inreg %a,
; SI-NEXT: v_alignbit_b32 v22, v11, v10, 8
; SI-NEXT: buffer_store_dword v22, off, s[0:3], s32 offset:84 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
+; SI-NEXT: v_alignbit_b32 v22, v14, v3, 24
+; SI-NEXT: buffer_store_dword v22, off, s[0:3], s32 offset:132 ; 4-byte Folded Spill
+; SI-NEXT: s_waitcnt expcnt(0)
+; SI-NEXT: v_alignbit_b32 v22, v14, v3, 16
+; SI-NEXT: buffer_store_dword v22, off, s[0:3], s32 offset:140 ; 4-byte Folded Spill
+; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_alignbit_b32 v22, v8, v2, 24
; SI-NEXT: buffer_store_dword v22, off, s[0:3], s32 offset:120 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
+; SI-NEXT: v_alignbit_b32 v22, v8, v2, 16
+; SI-NEXT: buffer_store_dword v22, off, s[0:3], s32 offset:128 ; 4-byte Folded Spill
+; SI-NEXT: s_waitcnt expcnt(0)
+; SI-NEXT: v_alignbit_b32 v22, v8, v2, 8
+; SI-NEXT: buffer_store_dword v22, off, s[0:3], s32 offset:136 ; 4-byte Folded Spill
+; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_alignbit_b32 v22, v5, v1, 24
; SI-NEXT: buffer_store_dword v22, off, s[0:3], s32 offset:112 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_alignbit_b32 v22, v5, v1, 16
; SI-NEXT: buffer_store_dword v22, off, s[0:3], s32 offset:116 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
+; SI-NEXT: v_alignbit_b32 v22, v5, v1, 8
+; SI-NEXT: buffer_store_dword v22, off, s[0:3], s32 offset:124 ; 4-byte Folded Spill
+; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_lshrrev_b32_e32 v22, 24, v59
; SI-NEXT: v_lshrrev_b32_e32 v59, 24, v45
; SI-NEXT: v_lshrrev_b32_e32 v45, 24, v44
@@ -90264,7 +90189,7 @@ define inreg <64 x i8> @bitcast_v32bf16_to_v64i8_scalar(<32 x bfloat> inreg %a,
; SI-NEXT: v_or_b32_e32 v6, v6, v9
; SI-NEXT: v_and_b32_e32 v9, 0xff, v55
; SI-NEXT: v_lshlrev_b32_e32 v9, 16, v9
-; SI-NEXT: v_lshlrev_b32_e32 v10, 24, v25
+; SI-NEXT: v_lshlrev_b32_e32 v10, 24, v53
; SI-NEXT: v_and_b32_e32 v6, 0xffff, v6
; SI-NEXT: v_or_b32_e32 v9, v10, v9
; SI-NEXT: v_or_b32_e32 v6, v6, v9
@@ -90285,10 +90210,14 @@ define inreg <64 x i8> @bitcast_v32bf16_to_v64i8_scalar(<32 x bfloat> inreg %a,
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_lshlrev_b32_e32 v6, 8, v54
; SI-NEXT: v_or_b32_e32 v3, v3, v6
-; SI-NEXT: v_and_b32_e32 v6, 0xff, v52
-; SI-NEXT: v_lshlrev_b32_e32 v6, 16, v6
-; SI-NEXT: v_lshlrev_b32_e32 v9, 24, v50
+; SI-NEXT: buffer_load_dword v6, off, s[0:3], s32 offset:140 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v9, off, s[0:3], s32 offset:132 ; 4-byte Folded Reload
; SI-NEXT: v_and_b32_e32 v3, 0xffff, v3
+; SI-NEXT: s_waitcnt vmcnt(1)
+; SI-NEXT: v_and_b32_e32 v6, 0xff, v6
+; SI-NEXT: v_lshlrev_b32_e32 v6, 16, v6
+; SI-NEXT: s_waitcnt vmcnt(0)
+; SI-NEXT: v_lshlrev_b32_e32 v9, 24, v9
; SI-NEXT: v_or_b32_e32 v6, v9, v6
; SI-NEXT: v_or_b32_e32 v3, v3, v6
; SI-NEXT: v_add_i32_e32 v6, vcc, 40, v0
@@ -90305,15 +90234,19 @@ define inreg <64 x i8> @bitcast_v32bf16_to_v64i8_scalar(<32 x bfloat> inreg %a,
; SI-NEXT: v_or_b32_e32 v3, v3, v6
; SI-NEXT: v_add_i32_e32 v6, vcc, 44, v0
; SI-NEXT: buffer_store_dword v3, v6, s[0:3], 0 offen
-; SI-NEXT: buffer_load_dword v6, off, s[0:3], s32 offset:120 ; 4-byte Folded Reload
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_lshlrev_b32_e32 v3, 8, v51
+; SI-NEXT: buffer_load_dword v3, off, s[0:3], s32 offset:136 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v6, off, s[0:3], s32 offset:120 ; 4-byte Folded Reload
+; SI-NEXT: s_waitcnt vmcnt(1)
+; SI-NEXT: v_lshlrev_b32_e32 v3, 8, v3
; SI-NEXT: v_or_b32_e32 v2, v2, v3
-; SI-NEXT: v_and_b32_e32 v3, 0xff, v32
-; SI-NEXT: v_lshlrev_b32_e32 v3, 16, v3
+; SI-NEXT: buffer_load_dword v3, off, s[0:3], s32 offset:128 ; 4-byte Folded Reload
; SI-NEXT: v_and_b32_e32 v2, 0xffff, v2
-; SI-NEXT: s_waitcnt vmcnt(0)
+; SI-NEXT: s_waitcnt vmcnt(1)
; SI-NEXT: v_lshlrev_b32_e32 v6, 24, v6
+; SI-NEXT: s_waitcnt vmcnt(0)
+; SI-NEXT: v_and_b32_e32 v3, 0xff, v3
+; SI-NEXT: v_lshlrev_b32_e32 v3, 16, v3
; SI-NEXT: v_or_b32_e32 v3, v6, v3
; SI-NEXT: v_or_b32_e32 v2, v2, v3
; SI-NEXT: v_add_i32_e32 v3, vcc, 48, v0
@@ -90331,16 +90264,18 @@ define inreg <64 x i8> @bitcast_v32bf16_to_v64i8_scalar(<32 x bfloat> inreg %a,
; SI-NEXT: v_add_i32_e32 v3, vcc, 52, v0
; SI-NEXT: buffer_store_dword v2, v3, s[0:3], 0 offen
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_lshlrev_b32_e32 v2, 8, v48
+; SI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:124 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v3, off, s[0:3], s32 offset:112 ; 4-byte Folded Reload
+; SI-NEXT: s_waitcnt vmcnt(1)
+; SI-NEXT: v_lshlrev_b32_e32 v2, 8, v2
; SI-NEXT: v_or_b32_e32 v1, v1, v2
; SI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:116 ; 4-byte Folded Reload
-; SI-NEXT: buffer_load_dword v3, off, s[0:3], s32 offset:112 ; 4-byte Folded Reload
; SI-NEXT: v_and_b32_e32 v1, 0xffff, v1
; SI-NEXT: s_waitcnt vmcnt(1)
+; SI-NEXT: v_lshlrev_b32_e32 v3, 24, v3
+; SI-NEXT: s_waitcnt vmcnt(0)
; SI-NEXT: v_and_b32_e32 v2, 0xff, v2
; SI-NEXT: v_lshlrev_b32_e32 v2, 16, v2
-; SI-NEXT: s_waitcnt vmcnt(0)
-; SI-NEXT: v_lshlrev_b32_e32 v3, 24, v3
; SI-NEXT: v_or_b32_e32 v2, v3, v2
; SI-NEXT: v_or_b32_e32 v1, v1, v2
; SI-NEXT: v_add_i32_e32 v2, vcc, 56, v0
@@ -90379,39 +90314,56 @@ define inreg <64 x i8> @bitcast_v32bf16_to_v64i8_scalar(<32 x bfloat> inreg %a,
; SI-NEXT: ; implicit-def: $vgpr1
; SI-NEXT: ; kill: killed $vgpr1
; SI-NEXT: ; implicit-def: $vgpr4
-; SI-NEXT: v_mov_b32_e32 v53, v32
+; SI-NEXT: ; kill: killed $vgpr4
+; SI-NEXT: v_mov_b32_e32 v52, v39
; SI-NEXT: ; implicit-def: $vgpr1
; SI-NEXT: ; kill: killed $vgpr1
-; SI-NEXT: v_mov_b32_e32 v49, v48
+; SI-NEXT: ; implicit-def: $vgpr4
+; SI-NEXT: v_mov_b32_e32 v51, v14
; SI-NEXT: ; implicit-def: $vgpr1
; SI-NEXT: ; kill: killed $vgpr1
-; SI-NEXT: v_mov_b32_e32 v37, v33
+; SI-NEXT: v_mov_b32_e32 v50, v45
; SI-NEXT: ; implicit-def: $vgpr1
; SI-NEXT: ; kill: killed $vgpr1
-; SI-NEXT: v_mov_b32_e32 v33, v56
+; SI-NEXT: v_mov_b32_e32 v49, v35
; SI-NEXT: ; implicit-def: $vgpr1
; SI-NEXT: ; kill: killed $vgpr1
-; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_mov_b32_e32 v35, v29
+; SI-NEXT: v_mov_b32_e32 v48, v12
; SI-NEXT: ; implicit-def: $vgpr1
; SI-NEXT: ; kill: killed $vgpr1
-; SI-NEXT: v_mov_b32_e32 v34, v44
+; SI-NEXT: v_mov_b32_e32 v39, v38
; SI-NEXT: ; implicit-def: $vgpr1
; SI-NEXT: ; kill: killed $vgpr1
-; SI-NEXT: v_mov_b32_e32 v31, v23
+; SI-NEXT: v_mov_b32_e32 v38, v56
; SI-NEXT: ; implicit-def: $vgpr1
; SI-NEXT: ; kill: killed $vgpr1
-; SI-NEXT: v_mov_b32_e32 v29, v43
+; SI-NEXT: v_mov_b32_e32 v37, v32
; SI-NEXT: ; implicit-def: $vgpr1
; SI-NEXT: ; kill: killed $vgpr1
-; SI-NEXT: v_mov_b32_e32 v28, v26
+; SI-NEXT: v_mov_b32_e32 v36, v31
; SI-NEXT: ; implicit-def: $vgpr1
; SI-NEXT: ; kill: killed $vgpr1
-; SI-NEXT: v_mov_b32_e32 v26, v42
+; SI-NEXT: v_mov_b32_e32 v35, v28
; SI-NEXT: ; implicit-def: $vgpr1
; SI-NEXT: ; kill: killed $vgpr1
-; SI-NEXT: v_mov_b32_e32 v23, v41
+; SI-NEXT: v_mov_b32_e32 v28, v44
+; SI-NEXT: ; implicit-def: $vgpr1
+; SI-NEXT: ; kill: killed $vgpr1
+; SI-NEXT: v_mov_b32_e32 v34, v29
+; SI-NEXT: ; implicit-def: $vgpr1
+; SI-NEXT: ; kill: killed $vgpr1
+; SI-NEXT: v_mov_b32_e32 v32, v23
+; SI-NEXT: ; implicit-def: $vgpr1
+; SI-NEXT: ; kill: killed $vgpr1
+; SI-NEXT: v_mov_b32_e32 v31, v43
+; SI-NEXT: ; implicit-def: $vgpr1
+; SI-NEXT: ; kill: killed $vgpr1
+; SI-NEXT: v_mov_b32_e32 v29, v26
; SI-NEXT: ; implicit-def: $vgpr1
+; SI-NEXT: v_mov_b32_e32 v26, v42
+; SI-NEXT: v_mov_b32_e32 v23, v41
+; SI-NEXT: s_waitcnt expcnt(0)
+; SI-NEXT: v_mov_b32_e32 v25, v22
; SI-NEXT: ; kill: killed $vgpr1
; SI-NEXT: ; implicit-def: $vgpr1
; SI-NEXT: ; kill: killed $vgpr4
@@ -90439,33 +90391,28 @@ define inreg <64 x i8> @bitcast_v32bf16_to_v64i8_scalar(<32 x bfloat> inreg %a,
; SI-NEXT: ; implicit-def: $vgpr6
; SI-NEXT: ; implicit-def: $vgpr40
; SI-NEXT: ; implicit-def: $vgpr55
-; SI-NEXT: ; implicit-def: $vgpr25
+; SI-NEXT: ; implicit-def: $vgpr53
; SI-NEXT: ; implicit-def: $vgpr20
; SI-NEXT: ; implicit-def: $vgpr47
; SI-NEXT: ; implicit-def: $vgpr18
; SI-NEXT: ; implicit-def: $vgpr59
; SI-NEXT: ; implicit-def: $vgpr3
; SI-NEXT: ; implicit-def: $vgpr54
-; SI-NEXT: ; implicit-def: $vgpr52
-; SI-NEXT: ; implicit-def: $vgpr50
; SI-NEXT: ; implicit-def: $vgpr14
; SI-NEXT: ; implicit-def: $vgpr60
; SI-NEXT: ; implicit-def: $vgpr12
; SI-NEXT: ; implicit-def: $vgpr45
; SI-NEXT: ; implicit-def: $vgpr2
-; SI-NEXT: ; implicit-def: $vgpr51
-; SI-NEXT: ; implicit-def: $vgpr32
; SI-NEXT: ; kill: killed $vgpr1
; SI-NEXT: ; implicit-def: $vgpr8
; SI-NEXT: ; implicit-def: $vgpr56
; SI-NEXT: ; implicit-def: $vgpr7
; SI-NEXT: ; implicit-def: $vgpr43
-; SI-NEXT: ; implicit-def: $vgpr48
+; SI-NEXT: ; implicit-def: $vgpr1
; SI-NEXT: ; kill: killed $vgpr4
; SI-NEXT: ; implicit-def: $vgpr5
; SI-NEXT: ; implicit-def: $vgpr44
; SI-NEXT: ; implicit-def: $vgpr4
-; SI-NEXT: ; implicit-def: $vgpr1
; SI-NEXT: ; implicit-def: $vgpr42
; SI-NEXT: s_branch .LBB109_2
;
@@ -91241,27 +91188,27 @@ define inreg <64 x i8> @bitcast_v32bf16_to_v64i8_scalar(<32 x bfloat> inreg %a,
; GFX9-NEXT: s_lshr_b32 s60, s22, 8
; GFX9-NEXT: s_lshr_b32 s48, s21, 24
; GFX9-NEXT: s_lshr_b32 s78, s21, 16
-; GFX9-NEXT: s_lshr_b32 s59, s21, 8
+; GFX9-NEXT: s_lshr_b32 s45, s21, 8
; GFX9-NEXT: s_lshr_b32 s49, s20, 16
-; GFX9-NEXT: s_lshr_b32 s58, s20, 8
+; GFX9-NEXT: s_lshr_b32 s44, s20, 8
; GFX9-NEXT: s_lshr_b32 s50, s19, 24
; GFX9-NEXT: s_lshr_b32 s77, s19, 16
-; GFX9-NEXT: s_lshr_b32 s57, s19, 8
+; GFX9-NEXT: s_lshr_b32 s13, s19, 8
; GFX9-NEXT: s_lshr_b32 s51, s18, 16
-; GFX9-NEXT: s_lshr_b32 s56, s18, 8
+; GFX9-NEXT: s_lshr_b32 s12, s18, 8
; GFX9-NEXT: s_lshr_b32 s52, s17, 24
; GFX9-NEXT: s_lshr_b32 s76, s17, 16
; GFX9-NEXT: s_lshr_b32 s53, s17, 8
; GFX9-NEXT: s_lshr_b32 s54, s16, 16
; GFX9-NEXT: s_lshr_b32 s55, s16, 8
-; GFX9-NEXT: s_lshr_b64 s[6:7], s[4:5], 24
-; GFX9-NEXT: s_lshr_b64 s[8:9], s[28:29], 24
-; GFX9-NEXT: s_lshr_b64 s[10:11], s[26:27], 24
-; GFX9-NEXT: s_lshr_b64 s[12:13], s[24:25], 24
-; GFX9-NEXT: s_lshr_b64 s[14:15], s[22:23], 24
-; GFX9-NEXT: s_lshr_b64 s[40:41], s[20:21], 24
-; GFX9-NEXT: s_lshr_b64 s[42:43], s[18:19], 24
-; GFX9-NEXT: s_lshr_b64 s[44:45], s[16:17], 24
+; GFX9-NEXT: s_lshr_b64 s[8:9], s[4:5], 24
+; GFX9-NEXT: s_lshr_b64 s[10:11], s[28:29], 24
+; GFX9-NEXT: s_lshr_b64 s[14:15], s[26:27], 24
+; GFX9-NEXT: s_lshr_b64 s[40:41], s[24:25], 24
+; GFX9-NEXT: s_lshr_b64 s[42:43], s[22:23], 24
+; GFX9-NEXT: s_lshr_b64 s[46:47], s[20:21], 24
+; GFX9-NEXT: s_lshr_b64 s[56:57], s[18:19], 24
+; GFX9-NEXT: s_lshr_b64 s[58:59], s[16:17], 24
; GFX9-NEXT: s_cbranch_execnz .LBB109_3
; GFX9-NEXT: .LBB109_2: ; %cmp.true
; GFX9-NEXT: s_and_b32 s6, s17, 0xffff0000
@@ -91290,357 +91237,357 @@ define inreg <64 x i8> @bitcast_v32bf16_to_v64i8_scalar(<32 x bfloat> inreg %a,
; GFX9-NEXT: s_and_b32 s6, s16, 0xffff0000
; GFX9-NEXT: v_add_f32_e32 v2, s6, v1
; GFX9-NEXT: v_readfirstlane_b32 s6, v2
-; GFX9-NEXT: s_bfe_u32 s7, s6, 0x10010
-; GFX9-NEXT: s_add_i32 s7, s7, s6
-; GFX9-NEXT: s_add_i32 s8, s7, 0x7fff
-; GFX9-NEXT: s_or_b32 s9, s6, 0x400000
+; GFX9-NEXT: s_bfe_u32 s8, s6, 0x10010
+; GFX9-NEXT: s_add_i32 s8, s8, s6
+; GFX9-NEXT: s_add_i32 s10, s8, 0x7fff
+; GFX9-NEXT: s_bitset1_b32 s6, 22
; GFX9-NEXT: v_cmp_u_f32_e32 vcc, v2, v2
-; GFX9-NEXT: s_and_b64 s[6:7], vcc, exec
-; GFX9-NEXT: s_cselect_b32 s6, s9, s8
-; GFX9-NEXT: s_lshr_b32 s8, s6, 16
-; GFX9-NEXT: s_lshl_b32 s6, s16, 16
-; GFX9-NEXT: v_add_f32_e32 v2, s6, v1
-; GFX9-NEXT: v_readfirstlane_b32 s6, v2
-; GFX9-NEXT: s_bfe_u32 s7, s6, 0x10010
-; GFX9-NEXT: s_add_i32 s7, s7, s6
-; GFX9-NEXT: s_add_i32 s9, s7, 0x7fff
-; GFX9-NEXT: s_or_b32 s10, s6, 0x400000
+; GFX9-NEXT: s_and_b64 s[8:9], vcc, exec
+; GFX9-NEXT: s_cselect_b32 s6, s6, s10
+; GFX9-NEXT: s_lshl_b32 s8, s16, 16
+; GFX9-NEXT: v_add_f32_e32 v2, s8, v1
+; GFX9-NEXT: v_readfirstlane_b32 s8, v2
+; GFX9-NEXT: s_bfe_u32 s9, s8, 0x10010
+; GFX9-NEXT: s_add_i32 s9, s9, s8
+; GFX9-NEXT: s_lshr_b32 s6, s6, 16
+; GFX9-NEXT: s_add_i32 s10, s9, 0x7fff
+; GFX9-NEXT: s_or_b32 s11, s8, 0x400000
; GFX9-NEXT: v_cmp_u_f32_e32 vcc, v2, v2
-; GFX9-NEXT: s_and_b64 s[6:7], vcc, exec
-; GFX9-NEXT: s_cselect_b32 s6, s10, s9
-; GFX9-NEXT: s_lshr_b32 s16, s6, 16
-; GFX9-NEXT: s_and_b32 s6, s19, 0xffff0000
-; GFX9-NEXT: v_add_f32_e32 v2, s6, v1
-; GFX9-NEXT: v_readfirstlane_b32 s6, v2
-; GFX9-NEXT: s_bfe_u32 s7, s6, 0x10010
-; GFX9-NEXT: s_add_i32 s7, s7, s6
-; GFX9-NEXT: s_pack_ll_b32_b16 s46, s16, s8
-; GFX9-NEXT: s_add_i32 s8, s7, 0x7fff
-; GFX9-NEXT: s_or_b32 s9, s6, 0x400000
+; GFX9-NEXT: s_and_b64 s[8:9], vcc, exec
+; GFX9-NEXT: s_cselect_b32 s8, s11, s10
+; GFX9-NEXT: s_lshr_b32 s16, s8, 16
+; GFX9-NEXT: s_and_b32 s8, s19, 0xffff0000
+; GFX9-NEXT: v_add_f32_e32 v2, s8, v1
+; GFX9-NEXT: v_readfirstlane_b32 s8, v2
+; GFX9-NEXT: s_bfe_u32 s9, s8, 0x10010
+; GFX9-NEXT: s_add_i32 s9, s9, s8
+; GFX9-NEXT: s_add_i32 s10, s9, 0x7fff
+; GFX9-NEXT: s_or_b32 s11, s8, 0x400000
; GFX9-NEXT: v_cmp_u_f32_e32 vcc, v2, v2
-; GFX9-NEXT: s_and_b64 s[6:7], vcc, exec
-; GFX9-NEXT: s_cselect_b32 s6, s9, s8
-; GFX9-NEXT: s_lshr_b32 s77, s6, 16
-; GFX9-NEXT: s_lshl_b32 s6, s19, 16
-; GFX9-NEXT: v_add_f32_e32 v2, s6, v1
-; GFX9-NEXT: v_readfirstlane_b32 s6, v2
-; GFX9-NEXT: s_bfe_u32 s7, s6, 0x10010
-; GFX9-NEXT: s_add_i32 s7, s7, s6
-; GFX9-NEXT: s_add_i32 s8, s7, 0x7fff
-; GFX9-NEXT: s_or_b32 s9, s6, 0x400000
+; GFX9-NEXT: s_and_b64 s[8:9], vcc, exec
+; GFX9-NEXT: s_cselect_b32 s8, s11, s10
+; GFX9-NEXT: s_lshr_b32 s77, s8, 16
+; GFX9-NEXT: s_lshl_b32 s8, s19, 16
+; GFX9-NEXT: v_add_f32_e32 v2, s8, v1
+; GFX9-NEXT: v_readfirstlane_b32 s8, v2
+; GFX9-NEXT: s_bfe_u32 s9, s8, 0x10010
+; GFX9-NEXT: s_add_i32 s9, s9, s8
+; GFX9-NEXT: s_add_i32 s10, s9, 0x7fff
+; GFX9-NEXT: s_or_b32 s11, s8, 0x400000
; GFX9-NEXT: v_cmp_u_f32_e32 vcc, v2, v2
-; GFX9-NEXT: s_and_b64 s[6:7], vcc, exec
-; GFX9-NEXT: s_cselect_b32 s6, s9, s8
-; GFX9-NEXT: s_lshr_b32 s19, s6, 16
-; GFX9-NEXT: s_and_b32 s6, s18, 0xffff0000
-; GFX9-NEXT: v_add_f32_e32 v2, s6, v1
-; GFX9-NEXT: v_readfirstlane_b32 s6, v2
-; GFX9-NEXT: s_bfe_u32 s7, s6, 0x10010
-; GFX9-NEXT: s_add_i32 s7, s7, s6
-; GFX9-NEXT: s_add_i32 s8, s7, 0x7fff
-; GFX9-NEXT: s_or_b32 s9, s6, 0x400000
+; GFX9-NEXT: s_and_b64 s[8:9], vcc, exec
+; GFX9-NEXT: s_cselect_b32 s8, s11, s10
+; GFX9-NEXT: s_lshr_b32 s19, s8, 16
+; GFX9-NEXT: s_and_b32 s8, s18, 0xffff0000
+; GFX9-NEXT: v_add_f32_e32 v2, s8, v1
+; GFX9-NEXT: v_readfirstlane_b32 s8, v2
+; GFX9-NEXT: s_bfe_u32 s9, s8, 0x10010
+; GFX9-NEXT: s_add_i32 s9, s9, s8
+; GFX9-NEXT: s_add_i32 s10, s9, 0x7fff
+; GFX9-NEXT: s_or_b32 s11, s8, 0x400000
; GFX9-NEXT: v_cmp_u_f32_e32 vcc, v2, v2
-; GFX9-NEXT: s_and_b64 s[6:7], vcc, exec
-; GFX9-NEXT: s_cselect_b32 s6, s9, s8
-; GFX9-NEXT: s_lshr_b32 s8, s6, 16
-; GFX9-NEXT: s_lshl_b32 s6, s18, 16
-; GFX9-NEXT: v_add_f32_e32 v2, s6, v1
-; GFX9-NEXT: v_readfirstlane_b32 s6, v2
-; GFX9-NEXT: s_bfe_u32 s7, s6, 0x10010
-; GFX9-NEXT: s_add_i32 s7, s7, s6
-; GFX9-NEXT: s_add_i32 s9, s7, 0x7fff
-; GFX9-NEXT: s_or_b32 s10, s6, 0x400000
+; GFX9-NEXT: s_and_b64 s[8:9], vcc, exec
+; GFX9-NEXT: s_cselect_b32 s8, s11, s10
+; GFX9-NEXT: s_lshr_b32 s10, s8, 16
+; GFX9-NEXT: s_lshl_b32 s8, s18, 16
+; GFX9-NEXT: v_add_f32_e32 v2, s8, v1
+; GFX9-NEXT: v_readfirstlane_b32 s8, v2
+; GFX9-NEXT: s_bfe_u32 s9, s8, 0x10010
+; GFX9-NEXT: s_add_i32 s9, s9, s8
+; GFX9-NEXT: s_add_i32 s11, s9, 0x7fff
+; GFX9-NEXT: s_or_b32 s12, s8, 0x400000
; GFX9-NEXT: v_cmp_u_f32_e32 vcc, v2, v2
-; GFX9-NEXT: s_and_b64 s[6:7], vcc, exec
-; GFX9-NEXT: s_cselect_b32 s6, s10, s9
-; GFX9-NEXT: s_lshr_b32 s18, s6, 16
-; GFX9-NEXT: s_and_b32 s6, s21, 0xffff0000
-; GFX9-NEXT: v_add_f32_e32 v2, s6, v1
-; GFX9-NEXT: v_readfirstlane_b32 s6, v2
-; GFX9-NEXT: s_bfe_u32 s7, s6, 0x10010
-; GFX9-NEXT: s_add_i32 s7, s7, s6
-; GFX9-NEXT: s_pack_ll_b32_b16 s56, s18, s8
-; GFX9-NEXT: s_add_i32 s8, s7, 0x7fff
-; GFX9-NEXT: s_or_b32 s9, s6, 0x400000
+; GFX9-NEXT: s_and_b64 s[8:9], vcc, exec
+; GFX9-NEXT: s_cselect_b32 s8, s12, s11
+; GFX9-NEXT: s_lshr_b32 s18, s8, 16
+; GFX9-NEXT: s_and_b32 s8, s21, 0xffff0000
+; GFX9-NEXT: v_add_f32_e32 v2, s8, v1
+; GFX9-NEXT: v_readfirstlane_b32 s8, v2
+; GFX9-NEXT: s_bfe_u32 s9, s8, 0x10010
+; GFX9-NEXT: s_add_i32 s9, s9, s8
+; GFX9-NEXT: s_pack_ll_b32_b16 s12, s18, s10
+; GFX9-NEXT: s_add_i32 s10, s9, 0x7fff
+; GFX9-NEXT: s_or_b32 s11, s8, 0x400000
; GFX9-NEXT: v_cmp_u_f32_e32 vcc, v2, v2
-; GFX9-NEXT: s_and_b64 s[6:7], vcc, exec
-; GFX9-NEXT: s_cselect_b32 s6, s9, s8
-; GFX9-NEXT: s_lshr_b32 s78, s6, 16
-; GFX9-NEXT: s_lshl_b32 s6, s21, 16
-; GFX9-NEXT: v_add_f32_e32 v2, s6, v1
-; GFX9-NEXT: v_readfirstlane_b32 s6, v2
-; GFX9-NEXT: s_bfe_u32 s7, s6, 0x10010
-; GFX9-NEXT: s_add_i32 s7, s7, s6
-; GFX9-NEXT: s_add_i32 s8, s7, 0x7fff
-; GFX9-NEXT: s_or_b32 s9, s6, 0x400000
+; GFX9-NEXT: s_and_b64 s[8:9], vcc, exec
+; GFX9-NEXT: s_cselect_b32 s8, s11, s10
+; GFX9-NEXT: s_lshr_b32 s78, s8, 16
+; GFX9-NEXT: s_lshl_b32 s8, s21, 16
+; GFX9-NEXT: v_add_f32_e32 v2, s8, v1
+; GFX9-NEXT: v_readfirstlane_b32 s8, v2
+; GFX9-NEXT: s_bfe_u32 s9, s8, 0x10010
+; GFX9-NEXT: s_add_i32 s9, s9, s8
+; GFX9-NEXT: s_add_i32 s10, s9, 0x7fff
+; GFX9-NEXT: s_or_b32 s11, s8, 0x400000
; GFX9-NEXT: v_cmp_u_f32_e32 vcc, v2, v2
-; GFX9-NEXT: s_and_b64 s[6:7], vcc, exec
-; GFX9-NEXT: s_cselect_b32 s6, s9, s8
-; GFX9-NEXT: s_lshr_b32 s21, s6, 16
-; GFX9-NEXT: s_and_b32 s6, s20, 0xffff0000
-; GFX9-NEXT: v_add_f32_e32 v2, s6, v1
-; GFX9-NEXT: v_readfirstlane_b32 s6, v2
-; GFX9-NEXT: s_bfe_u32 s7, s6, 0x10010
-; GFX9-NEXT: s_add_i32 s7, s7, s6
-; GFX9-NEXT: s_add_i32 s8, s7, 0x7fff
-; GFX9-NEXT: s_or_b32 s9, s6, 0x400000
+; GFX9-NEXT: s_and_b64 s[8:9], vcc, exec
+; GFX9-NEXT: s_cselect_b32 s8, s11, s10
+; GFX9-NEXT: s_lshr_b32 s21, s8, 16
+; GFX9-NEXT: s_and_b32 s8, s20, 0xffff0000
+; GFX9-NEXT: v_add_f32_e32 v2, s8, v1
+; GFX9-NEXT: v_readfirstlane_b32 s8, v2
+; GFX9-NEXT: s_bfe_u32 s9, s8, 0x10010
+; GFX9-NEXT: s_add_i32 s9, s9, s8
+; GFX9-NEXT: s_add_i32 s10, s9, 0x7fff
+; GFX9-NEXT: s_or_b32 s11, s8, 0x400000
; GFX9-NEXT: v_cmp_u_f32_e32 vcc, v2, v2
-; GFX9-NEXT: s_and_b64 s[6:7], vcc, exec
-; GFX9-NEXT: s_cselect_b32 s6, s9, s8
-; GFX9-NEXT: s_lshr_b32 s8, s6, 16
-; GFX9-NEXT: s_lshl_b32 s6, s20, 16
-; GFX9-NEXT: v_add_f32_e32 v2, s6, v1
-; GFX9-NEXT: v_readfirstlane_b32 s6, v2
-; GFX9-NEXT: s_bfe_u32 s7, s6, 0x10010
-; GFX9-NEXT: s_add_i32 s7, s7, s6
-; GFX9-NEXT: s_add_i32 s9, s7, 0x7fff
-; GFX9-NEXT: s_or_b32 s10, s6, 0x400000
+; GFX9-NEXT: s_and_b64 s[8:9], vcc, exec
+; GFX9-NEXT: s_cselect_b32 s8, s11, s10
+; GFX9-NEXT: s_lshr_b32 s10, s8, 16
+; GFX9-NEXT: s_lshl_b32 s8, s20, 16
+; GFX9-NEXT: v_add_f32_e32 v2, s8, v1
+; GFX9-NEXT: v_readfirstlane_b32 s8, v2
+; GFX9-NEXT: s_bfe_u32 s9, s8, 0x10010
+; GFX9-NEXT: s_add_i32 s9, s9, s8
+; GFX9-NEXT: s_add_i32 s11, s9, 0x7fff
+; GFX9-NEXT: s_or_b32 s14, s8, 0x400000
; GFX9-NEXT: v_cmp_u_f32_e32 vcc, v2, v2
-; GFX9-NEXT: s_and_b64 s[6:7], vcc, exec
-; GFX9-NEXT: s_cselect_b32 s6, s10, s9
-; GFX9-NEXT: s_lshr_b32 s20, s6, 16
-; GFX9-NEXT: s_and_b32 s6, s23, 0xffff0000
-; GFX9-NEXT: v_add_f32_e32 v2, s6, v1
-; GFX9-NEXT: v_readfirstlane_b32 s6, v2
-; GFX9-NEXT: s_bfe_u32 s7, s6, 0x10010
-; GFX9-NEXT: s_add_i32 s7, s7, s6
-; GFX9-NEXT: s_pack_ll_b32_b16 s58, s20, s8
-; GFX9-NEXT: s_add_i32 s8, s7, 0x7fff
-; GFX9-NEXT: s_or_b32 s9, s6, 0x400000
+; GFX9-NEXT: s_and_b64 s[8:9], vcc, exec
+; GFX9-NEXT: s_cselect_b32 s8, s14, s11
+; GFX9-NEXT: s_lshr_b32 s20, s8, 16
+; GFX9-NEXT: s_and_b32 s8, s23, 0xffff0000
+; GFX9-NEXT: v_add_f32_e32 v2, s8, v1
+; GFX9-NEXT: v_readfirstlane_b32 s8, v2
+; GFX9-NEXT: s_bfe_u32 s9, s8, 0x10010
+; GFX9-NEXT: s_add_i32 s9, s9, s8
+; GFX9-NEXT: s_pack_ll_b32_b16 s44, s20, s10
+; GFX9-NEXT: s_add_i32 s10, s9, 0x7fff
+; GFX9-NEXT: s_or_b32 s11, s8, 0x400000
; GFX9-NEXT: v_cmp_u_f32_e32 vcc, v2, v2
-; GFX9-NEXT: s_and_b64 s[6:7], vcc, exec
-; GFX9-NEXT: s_cselect_b32 s6, s9, s8
-; GFX9-NEXT: s_lshr_b32 s79, s6, 16
-; GFX9-NEXT: s_lshl_b32 s6, s23, 16
-; GFX9-NEXT: v_add_f32_e32 v2, s6, v1
-; GFX9-NEXT: v_readfirstlane_b32 s6, v2
-; GFX9-NEXT: s_bfe_u32 s7, s6, 0x10010
-; GFX9-NEXT: s_add_i32 s7, s7, s6
-; GFX9-NEXT: s_add_i32 s8, s7, 0x7fff
-; GFX9-NEXT: s_or_b32 s9, s6, 0x400000
+; GFX9-NEXT: s_and_b64 s[8:9], vcc, exec
+; GFX9-NEXT: s_cselect_b32 s8, s11, s10
+; GFX9-NEXT: s_lshr_b32 s79, s8, 16
+; GFX9-NEXT: s_lshl_b32 s8, s23, 16
+; GFX9-NEXT: v_add_f32_e32 v2, s8, v1
+; GFX9-NEXT: v_readfirstlane_b32 s8, v2
+; GFX9-NEXT: s_bfe_u32 s9, s8, 0x10010
+; GFX9-NEXT: s_add_i32 s9, s9, s8
+; GFX9-NEXT: s_add_i32 s10, s9, 0x7fff
+; GFX9-NEXT: s_or_b32 s11, s8, 0x400000
; GFX9-NEXT: v_cmp_u_f32_e32 vcc, v2, v2
-; GFX9-NEXT: s_and_b64 s[6:7], vcc, exec
-; GFX9-NEXT: s_cselect_b32 s6, s9, s8
-; GFX9-NEXT: s_lshr_b32 s23, s6, 16
-; GFX9-NEXT: s_and_b32 s6, s22, 0xffff0000
-; GFX9-NEXT: v_add_f32_e32 v2, s6, v1
-; GFX9-NEXT: v_readfirstlane_b32 s6, v2
-; GFX9-NEXT: s_bfe_u32 s7, s6, 0x10010
-; GFX9-NEXT: s_add_i32 s7, s7, s6
-; GFX9-NEXT: s_add_i32 s8, s7, 0x7fff
-; GFX9-NEXT: s_or_b32 s9, s6, 0x400000
+; GFX9-NEXT: s_and_b64 s[8:9], vcc, exec
+; GFX9-NEXT: s_cselect_b32 s8, s11, s10
+; GFX9-NEXT: s_lshr_b32 s23, s8, 16
+; GFX9-NEXT: s_and_b32 s8, s22, 0xffff0000
+; GFX9-NEXT: v_add_f32_e32 v2, s8, v1
+; GFX9-NEXT: v_readfirstlane_b32 s8, v2
+; GFX9-NEXT: s_bfe_u32 s9, s8, 0x10010
+; GFX9-NEXT: s_add_i32 s9, s9, s8
+; GFX9-NEXT: s_add_i32 s10, s9, 0x7fff
+; GFX9-NEXT: s_or_b32 s11, s8, 0x400000
; GFX9-NEXT: v_cmp_u_f32_e32 vcc, v2, v2
-; GFX9-NEXT: s_and_b64 s[6:7], vcc, exec
-; GFX9-NEXT: s_cselect_b32 s6, s9, s8
-; GFX9-NEXT: s_lshr_b32 s8, s6, 16
-; GFX9-NEXT: s_lshl_b32 s6, s22, 16
-; GFX9-NEXT: v_add_f32_e32 v2, s6, v1
-; GFX9-NEXT: v_readfirstlane_b32 s6, v2
-; GFX9-NEXT: s_bfe_u32 s7, s6, 0x10010
-; GFX9-NEXT: s_add_i32 s7, s7, s6
-; GFX9-NEXT: s_add_i32 s9, s7, 0x7fff
-; GFX9-NEXT: s_or_b32 s10, s6, 0x400000
+; GFX9-NEXT: s_and_b64 s[8:9], vcc, exec
+; GFX9-NEXT: s_cselect_b32 s8, s11, s10
+; GFX9-NEXT: s_lshr_b32 s10, s8, 16
+; GFX9-NEXT: s_lshl_b32 s8, s22, 16
+; GFX9-NEXT: v_add_f32_e32 v2, s8, v1
+; GFX9-NEXT: v_readfirstlane_b32 s8, v2
+; GFX9-NEXT: s_bfe_u32 s9, s8, 0x10010
+; GFX9-NEXT: s_add_i32 s9, s9, s8
+; GFX9-NEXT: s_add_i32 s11, s9, 0x7fff
+; GFX9-NEXT: s_or_b32 s14, s8, 0x400000
; GFX9-NEXT: v_cmp_u_f32_e32 vcc, v2, v2
-; GFX9-NEXT: s_and_b64 s[6:7], vcc, exec
-; GFX9-NEXT: s_cselect_b32 s6, s10, s9
-; GFX9-NEXT: s_lshr_b32 s22, s6, 16
-; GFX9-NEXT: s_and_b32 s6, s25, 0xffff0000
-; GFX9-NEXT: v_add_f32_e32 v2, s6, v1
-; GFX9-NEXT: v_readfirstlane_b32 s6, v2
-; GFX9-NEXT: s_bfe_u32 s7, s6, 0x10010
-; GFX9-NEXT: s_add_i32 s7, s7, s6
-; GFX9-NEXT: s_pack_ll_b32_b16 s60, s22, s8
-; GFX9-NEXT: s_add_i32 s8, s7, 0x7fff
-; GFX9-NEXT: s_or_b32 s9, s6, 0x400000
+; GFX9-NEXT: s_and_b64 s[8:9], vcc, exec
+; GFX9-NEXT: s_cselect_b32 s8, s14, s11
+; GFX9-NEXT: s_lshr_b32 s22, s8, 16
+; GFX9-NEXT: s_and_b32 s8, s25, 0xffff0000
+; GFX9-NEXT: v_add_f32_e32 v2, s8, v1
+; GFX9-NEXT: v_readfirstlane_b32 s8, v2
+; GFX9-NEXT: s_bfe_u32 s9, s8, 0x10010
+; GFX9-NEXT: s_add_i32 s9, s9, s8
+; GFX9-NEXT: s_pack_ll_b32_b16 s60, s22, s10
+; GFX9-NEXT: s_add_i32 s10, s9, 0x7fff
+; GFX9-NEXT: s_or_b32 s11, s8, 0x400000
; GFX9-NEXT: v_cmp_u_f32_e32 vcc, v2, v2
-; GFX9-NEXT: s_and_b64 s[6:7], vcc, exec
-; GFX9-NEXT: s_cselect_b32 s6, s9, s8
-; GFX9-NEXT: s_lshr_b32 s88, s6, 16
-; GFX9-NEXT: s_lshl_b32 s6, s25, 16
-; GFX9-NEXT: v_add_f32_e32 v2, s6, v1
-; GFX9-NEXT: v_readfirstlane_b32 s6, v2
-; GFX9-NEXT: s_bfe_u32 s7, s6, 0x10010
-; GFX9-NEXT: s_add_i32 s7, s7, s6
-; GFX9-NEXT: s_add_i32 s8, s7, 0x7fff
-; GFX9-NEXT: s_or_b32 s9, s6, 0x400000
+; GFX9-NEXT: s_and_b64 s[8:9], vcc, exec
+; GFX9-NEXT: s_cselect_b32 s8, s11, s10
+; GFX9-NEXT: s_lshr_b32 s88, s8, 16
+; GFX9-NEXT: s_lshl_b32 s8, s25, 16
+; GFX9-NEXT: v_add_f32_e32 v2, s8, v1
+; GFX9-NEXT: v_readfirstlane_b32 s8, v2
+; GFX9-NEXT: s_bfe_u32 s9, s8, 0x10010
+; GFX9-NEXT: s_add_i32 s9, s9, s8
+; GFX9-NEXT: s_add_i32 s10, s9, 0x7fff
+; GFX9-NEXT: s_or_b32 s11, s8, 0x400000
; GFX9-NEXT: v_cmp_u_f32_e32 vcc, v2, v2
-; GFX9-NEXT: s_and_b64 s[6:7], vcc, exec
-; GFX9-NEXT: s_cselect_b32 s6, s9, s8
-; GFX9-NEXT: s_lshr_b32 s25, s6, 16
-; GFX9-NEXT: s_and_b32 s6, s24, 0xffff0000
-; GFX9-NEXT: v_add_f32_e32 v2, s6, v1
-; GFX9-NEXT: v_readfirstlane_b32 s6, v2
-; GFX9-NEXT: s_bfe_u32 s7, s6, 0x10010
-; GFX9-NEXT: s_add_i32 s7, s7, s6
-; GFX9-NEXT: s_add_i32 s8, s7, 0x7fff
-; GFX9-NEXT: s_or_b32 s9, s6, 0x400000
+; GFX9-NEXT: s_and_b64 s[8:9], vcc, exec
+; GFX9-NEXT: s_cselect_b32 s8, s11, s10
+; GFX9-NEXT: s_lshr_b32 s25, s8, 16
+; GFX9-NEXT: s_and_b32 s8, s24, 0xffff0000
+; GFX9-NEXT: v_add_f32_e32 v2, s8, v1
+; GFX9-NEXT: v_readfirstlane_b32 s8, v2
+; GFX9-NEXT: s_bfe_u32 s9, s8, 0x10010
+; GFX9-NEXT: s_add_i32 s9, s9, s8
+; GFX9-NEXT: s_add_i32 s10, s9, 0x7fff
+; GFX9-NEXT: s_or_b32 s11, s8, 0x400000
; GFX9-NEXT: v_cmp_u_f32_e32 vcc, v2, v2
-; GFX9-NEXT: s_and_b64 s[6:7], vcc, exec
-; GFX9-NEXT: s_cselect_b32 s6, s9, s8
-; GFX9-NEXT: s_lshr_b32 s8, s6, 16
-; GFX9-NEXT: s_lshl_b32 s6, s24, 16
-; GFX9-NEXT: v_add_f32_e32 v2, s6, v1
-; GFX9-NEXT: v_readfirstlane_b32 s6, v2
-; GFX9-NEXT: s_bfe_u32 s7, s6, 0x10010
-; GFX9-NEXT: s_add_i32 s7, s7, s6
-; GFX9-NEXT: s_add_i32 s9, s7, 0x7fff
-; GFX9-NEXT: s_or_b32 s10, s6, 0x400000
+; GFX9-NEXT: s_and_b64 s[8:9], vcc, exec
+; GFX9-NEXT: s_cselect_b32 s8, s11, s10
+; GFX9-NEXT: s_lshr_b32 s10, s8, 16
+; GFX9-NEXT: s_lshl_b32 s8, s24, 16
+; GFX9-NEXT: v_add_f32_e32 v2, s8, v1
+; GFX9-NEXT: v_readfirstlane_b32 s8, v2
+; GFX9-NEXT: s_bfe_u32 s9, s8, 0x10010
+; GFX9-NEXT: s_add_i32 s9, s9, s8
+; GFX9-NEXT: s_add_i32 s11, s9, 0x7fff
+; GFX9-NEXT: s_or_b32 s14, s8, 0x400000
; GFX9-NEXT: v_cmp_u_f32_e32 vcc, v2, v2
-; GFX9-NEXT: s_and_b64 s[6:7], vcc, exec
-; GFX9-NEXT: s_cselect_b32 s6, s10, s9
-; GFX9-NEXT: s_lshr_b32 s24, s6, 16
-; GFX9-NEXT: s_and_b32 s6, s27, 0xffff0000
-; GFX9-NEXT: v_add_f32_e32 v2, s6, v1
-; GFX9-NEXT: v_readfirstlane_b32 s6, v2
-; GFX9-NEXT: s_bfe_u32 s7, s6, 0x10010
-; GFX9-NEXT: s_add_i32 s7, s7, s6
-; GFX9-NEXT: s_pack_ll_b32_b16 s62, s24, s8
-; GFX9-NEXT: s_add_i32 s8, s7, 0x7fff
-; GFX9-NEXT: s_or_b32 s9, s6, 0x400000
+; GFX9-NEXT: s_and_b64 s[8:9], vcc, exec
+; GFX9-NEXT: s_cselect_b32 s8, s14, s11
+; GFX9-NEXT: s_lshr_b32 s24, s8, 16
+; GFX9-NEXT: s_and_b32 s8, s27, 0xffff0000
+; GFX9-NEXT: v_add_f32_e32 v2, s8, v1
+; GFX9-NEXT: v_readfirstlane_b32 s8, v2
+; GFX9-NEXT: s_bfe_u32 s9, s8, 0x10010
+; GFX9-NEXT: s_add_i32 s9, s9, s8
+; GFX9-NEXT: s_pack_ll_b32_b16 s62, s24, s10
+; GFX9-NEXT: s_add_i32 s10, s9, 0x7fff
+; GFX9-NEXT: s_or_b32 s11, s8, 0x400000
; GFX9-NEXT: v_cmp_u_f32_e32 vcc, v2, v2
-; GFX9-NEXT: s_and_b64 s[6:7], vcc, exec
-; GFX9-NEXT: s_cselect_b32 s6, s9, s8
-; GFX9-NEXT: s_lshr_b32 s89, s6, 16
-; GFX9-NEXT: s_lshl_b32 s6, s27, 16
-; GFX9-NEXT: v_add_f32_e32 v2, s6, v1
-; GFX9-NEXT: v_readfirstlane_b32 s6, v2
-; GFX9-NEXT: s_bfe_u32 s7, s6, 0x10010
-; GFX9-NEXT: s_add_i32 s7, s7, s6
-; GFX9-NEXT: s_add_i32 s8, s7, 0x7fff
-; GFX9-NEXT: s_or_b32 s9, s6, 0x400000
+; GFX9-NEXT: s_and_b64 s[8:9], vcc, exec
+; GFX9-NEXT: s_cselect_b32 s8, s11, s10
+; GFX9-NEXT: s_lshr_b32 s89, s8, 16
+; GFX9-NEXT: s_lshl_b32 s8, s27, 16
+; GFX9-NEXT: v_add_f32_e32 v2, s8, v1
+; GFX9-NEXT: v_readfirstlane_b32 s8, v2
+; GFX9-NEXT: s_bfe_u32 s9, s8, 0x10010
+; GFX9-NEXT: s_add_i32 s9, s9, s8
+; GFX9-NEXT: s_add_i32 s10, s9, 0x7fff
+; GFX9-NEXT: s_or_b32 s11, s8, 0x400000
; GFX9-NEXT: v_cmp_u_f32_e32 vcc, v2, v2
-; GFX9-NEXT: s_and_b64 s[6:7], vcc, exec
-; GFX9-NEXT: s_cselect_b32 s6, s9, s8
-; GFX9-NEXT: s_lshr_b32 s27, s6, 16
-; GFX9-NEXT: s_and_b32 s6, s26, 0xffff0000
-; GFX9-NEXT: v_add_f32_e32 v2, s6, v1
-; GFX9-NEXT: v_readfirstlane_b32 s6, v2
-; GFX9-NEXT: s_bfe_u32 s7, s6, 0x10010
-; GFX9-NEXT: s_add_i32 s7, s7, s6
-; GFX9-NEXT: s_add_i32 s8, s7, 0x7fff
-; GFX9-NEXT: s_or_b32 s9, s6, 0x400000
+; GFX9-NEXT: s_and_b64 s[8:9], vcc, exec
+; GFX9-NEXT: s_cselect_b32 s8, s11, s10
+; GFX9-NEXT: s_lshr_b32 s27, s8, 16
+; GFX9-NEXT: s_and_b32 s8, s26, 0xffff0000
+; GFX9-NEXT: v_add_f32_e32 v2, s8, v1
+; GFX9-NEXT: v_readfirstlane_b32 s8, v2
+; GFX9-NEXT: s_bfe_u32 s9, s8, 0x10010
+; GFX9-NEXT: s_add_i32 s9, s9, s8
+; GFX9-NEXT: s_add_i32 s10, s9, 0x7fff
+; GFX9-NEXT: s_or_b32 s11, s8, 0x400000
; GFX9-NEXT: v_cmp_u_f32_e32 vcc, v2, v2
-; GFX9-NEXT: s_and_b64 s[6:7], vcc, exec
-; GFX9-NEXT: s_cselect_b32 s6, s9, s8
-; GFX9-NEXT: s_lshr_b32 s8, s6, 16
-; GFX9-NEXT: s_lshl_b32 s6, s26, 16
-; GFX9-NEXT: v_add_f32_e32 v2, s6, v1
-; GFX9-NEXT: v_readfirstlane_b32 s6, v2
-; GFX9-NEXT: s_bfe_u32 s7, s6, 0x10010
-; GFX9-NEXT: s_add_i32 s7, s7, s6
-; GFX9-NEXT: s_add_i32 s9, s7, 0x7fff
-; GFX9-NEXT: s_or_b32 s10, s6, 0x400000
+; GFX9-NEXT: s_and_b64 s[8:9], vcc, exec
+; GFX9-NEXT: s_cselect_b32 s8, s11, s10
+; GFX9-NEXT: s_lshr_b32 s10, s8, 16
+; GFX9-NEXT: s_lshl_b32 s8, s26, 16
+; GFX9-NEXT: v_add_f32_e32 v2, s8, v1
+; GFX9-NEXT: v_readfirstlane_b32 s8, v2
+; GFX9-NEXT: s_bfe_u32 s9, s8, 0x10010
+; GFX9-NEXT: s_add_i32 s9, s9, s8
+; GFX9-NEXT: s_add_i32 s11, s9, 0x7fff
+; GFX9-NEXT: s_or_b32 s14, s8, 0x400000
; GFX9-NEXT: v_cmp_u_f32_e32 vcc, v2, v2
-; GFX9-NEXT: s_and_b64 s[6:7], vcc, exec
-; GFX9-NEXT: s_cselect_b32 s6, s10, s9
-; GFX9-NEXT: s_lshr_b32 s26, s6, 16
-; GFX9-NEXT: s_and_b32 s6, s29, 0xffff0000
-; GFX9-NEXT: v_add_f32_e32 v2, s6, v1
-; GFX9-NEXT: v_readfirstlane_b32 s6, v2
-; GFX9-NEXT: s_bfe_u32 s7, s6, 0x10010
-; GFX9-NEXT: s_add_i32 s7, s7, s6
-; GFX9-NEXT: s_pack_ll_b32_b16 s72, s26, s8
-; GFX9-NEXT: s_add_i32 s8, s7, 0x7fff
-; GFX9-NEXT: s_or_b32 s9, s6, 0x400000
+; GFX9-NEXT: s_and_b64 s[8:9], vcc, exec
+; GFX9-NEXT: s_cselect_b32 s8, s14, s11
+; GFX9-NEXT: s_lshr_b32 s26, s8, 16
+; GFX9-NEXT: s_and_b32 s8, s29, 0xffff0000
+; GFX9-NEXT: v_add_f32_e32 v2, s8, v1
+; GFX9-NEXT: v_readfirstlane_b32 s8, v2
+; GFX9-NEXT: s_bfe_u32 s9, s8, 0x10010
+; GFX9-NEXT: s_add_i32 s9, s9, s8
+; GFX9-NEXT: s_pack_ll_b32_b16 s72, s26, s10
+; GFX9-NEXT: s_add_i32 s10, s9, 0x7fff
+; GFX9-NEXT: s_or_b32 s11, s8, 0x400000
; GFX9-NEXT: v_cmp_u_f32_e32 vcc, v2, v2
-; GFX9-NEXT: s_and_b64 s[6:7], vcc, exec
-; GFX9-NEXT: s_cselect_b32 s6, s9, s8
-; GFX9-NEXT: s_lshr_b32 s90, s6, 16
-; GFX9-NEXT: s_lshl_b32 s6, s29, 16
-; GFX9-NEXT: v_add_f32_e32 v2, s6, v1
-; GFX9-NEXT: v_readfirstlane_b32 s6, v2
-; GFX9-NEXT: s_bfe_u32 s7, s6, 0x10010
-; GFX9-NEXT: s_add_i32 s7, s7, s6
-; GFX9-NEXT: s_add_i32 s8, s7, 0x7fff
-; GFX9-NEXT: s_or_b32 s9, s6, 0x400000
+; GFX9-NEXT: s_and_b64 s[8:9], vcc, exec
+; GFX9-NEXT: s_cselect_b32 s8, s11, s10
+; GFX9-NEXT: s_lshr_b32 s90, s8, 16
+; GFX9-NEXT: s_lshl_b32 s8, s29, 16
+; GFX9-NEXT: v_add_f32_e32 v2, s8, v1
+; GFX9-NEXT: v_readfirstlane_b32 s8, v2
+; GFX9-NEXT: s_bfe_u32 s9, s8, 0x10010
+; GFX9-NEXT: s_add_i32 s9, s9, s8
+; GFX9-NEXT: s_add_i32 s10, s9, 0x7fff
+; GFX9-NEXT: s_or_b32 s11, s8, 0x400000
; GFX9-NEXT: v_cmp_u_f32_e32 vcc, v2, v2
-; GFX9-NEXT: s_and_b64 s[6:7], vcc, exec
-; GFX9-NEXT: s_cselect_b32 s6, s9, s8
-; GFX9-NEXT: s_lshr_b32 s29, s6, 16
-; GFX9-NEXT: s_and_b32 s6, s28, 0xffff0000
-; GFX9-NEXT: v_add_f32_e32 v2, s6, v1
-; GFX9-NEXT: v_readfirstlane_b32 s6, v2
-; GFX9-NEXT: s_bfe_u32 s7, s6, 0x10010
-; GFX9-NEXT: s_add_i32 s7, s7, s6
-; GFX9-NEXT: s_add_i32 s8, s7, 0x7fff
-; GFX9-NEXT: s_or_b32 s9, s6, 0x400000
+; GFX9-NEXT: s_and_b64 s[8:9], vcc, exec
+; GFX9-NEXT: s_cselect_b32 s8, s11, s10
+; GFX9-NEXT: s_lshr_b32 s29, s8, 16
+; GFX9-NEXT: s_and_b32 s8, s28, 0xffff0000
+; GFX9-NEXT: v_add_f32_e32 v2, s8, v1
+; GFX9-NEXT: v_readfirstlane_b32 s8, v2
+; GFX9-NEXT: s_bfe_u32 s9, s8, 0x10010
+; GFX9-NEXT: s_add_i32 s9, s9, s8
+; GFX9-NEXT: s_add_i32 s10, s9, 0x7fff
+; GFX9-NEXT: s_or_b32 s11, s8, 0x400000
; GFX9-NEXT: v_cmp_u_f32_e32 vcc, v2, v2
-; GFX9-NEXT: s_and_b64 s[6:7], vcc, exec
-; GFX9-NEXT: s_cselect_b32 s6, s9, s8
-; GFX9-NEXT: s_lshr_b32 s8, s6, 16
-; GFX9-NEXT: s_lshl_b32 s6, s28, 16
-; GFX9-NEXT: v_add_f32_e32 v2, s6, v1
-; GFX9-NEXT: v_readfirstlane_b32 s6, v2
-; GFX9-NEXT: s_bfe_u32 s7, s6, 0x10010
-; GFX9-NEXT: s_add_i32 s7, s7, s6
-; GFX9-NEXT: s_add_i32 s9, s7, 0x7fff
-; GFX9-NEXT: s_or_b32 s10, s6, 0x400000
+; GFX9-NEXT: s_and_b64 s[8:9], vcc, exec
+; GFX9-NEXT: s_cselect_b32 s8, s11, s10
+; GFX9-NEXT: s_lshr_b32 s10, s8, 16
+; GFX9-NEXT: s_lshl_b32 s8, s28, 16
+; GFX9-NEXT: v_add_f32_e32 v2, s8, v1
+; GFX9-NEXT: v_readfirstlane_b32 s8, v2
+; GFX9-NEXT: s_bfe_u32 s9, s8, 0x10010
+; GFX9-NEXT: s_add_i32 s9, s9, s8
+; GFX9-NEXT: s_add_i32 s11, s9, 0x7fff
+; GFX9-NEXT: s_or_b32 s14, s8, 0x400000
; GFX9-NEXT: v_cmp_u_f32_e32 vcc, v2, v2
-; GFX9-NEXT: s_and_b64 s[6:7], vcc, exec
-; GFX9-NEXT: s_cselect_b32 s6, s10, s9
-; GFX9-NEXT: s_lshr_b32 s28, s6, 16
-; GFX9-NEXT: s_and_b32 s6, s5, 0xffff0000
-; GFX9-NEXT: v_add_f32_e32 v2, s6, v1
-; GFX9-NEXT: v_readfirstlane_b32 s6, v2
-; GFX9-NEXT: s_bfe_u32 s7, s6, 0x10010
-; GFX9-NEXT: s_add_i32 s7, s7, s6
-; GFX9-NEXT: s_pack_ll_b32_b16 s74, s28, s8
-; GFX9-NEXT: s_add_i32 s8, s7, 0x7fff
-; GFX9-NEXT: s_or_b32 s9, s6, 0x400000
+; GFX9-NEXT: s_and_b64 s[8:9], vcc, exec
+; GFX9-NEXT: s_cselect_b32 s8, s14, s11
+; GFX9-NEXT: s_lshr_b32 s28, s8, 16
+; GFX9-NEXT: s_and_b32 s8, s5, 0xffff0000
+; GFX9-NEXT: v_add_f32_e32 v2, s8, v1
+; GFX9-NEXT: v_readfirstlane_b32 s8, v2
+; GFX9-NEXT: s_bfe_u32 s9, s8, 0x10010
+; GFX9-NEXT: s_add_i32 s9, s9, s8
+; GFX9-NEXT: s_pack_ll_b32_b16 s74, s28, s10
+; GFX9-NEXT: s_add_i32 s10, s9, 0x7fff
+; GFX9-NEXT: s_or_b32 s11, s8, 0x400000
; GFX9-NEXT: v_cmp_u_f32_e32 vcc, v2, v2
-; GFX9-NEXT: s_and_b64 s[6:7], vcc, exec
-; GFX9-NEXT: s_cselect_b32 s6, s9, s8
+; GFX9-NEXT: s_and_b64 s[8:9], vcc, exec
+; GFX9-NEXT: s_cselect_b32 s8, s11, s10
; GFX9-NEXT: s_lshl_b32 s5, s5, 16
; GFX9-NEXT: v_add_f32_e32 v2, s5, v1
; GFX9-NEXT: v_readfirstlane_b32 s5, v2
-; GFX9-NEXT: s_lshr_b32 s91, s6, 16
-; GFX9-NEXT: s_bfe_u32 s6, s5, 0x10010
-; GFX9-NEXT: s_add_i32 s6, s6, s5
-; GFX9-NEXT: s_add_i32 s8, s6, 0x7fff
+; GFX9-NEXT: s_lshr_b32 s91, s8, 16
+; GFX9-NEXT: s_bfe_u32 s8, s5, 0x10010
+; GFX9-NEXT: s_add_i32 s8, s8, s5
+; GFX9-NEXT: s_add_i32 s10, s8, 0x7fff
; GFX9-NEXT: s_bitset1_b32 s5, 22
; GFX9-NEXT: v_cmp_u_f32_e32 vcc, v2, v2
-; GFX9-NEXT: s_and_b64 s[6:7], vcc, exec
-; GFX9-NEXT: s_cselect_b32 s5, s5, s8
-; GFX9-NEXT: s_and_b32 s6, s4, 0xffff0000
-; GFX9-NEXT: v_add_f32_e32 v2, s6, v1
-; GFX9-NEXT: v_readfirstlane_b32 s6, v2
-; GFX9-NEXT: s_bfe_u32 s7, s6, 0x10010
-; GFX9-NEXT: s_add_i32 s7, s7, s6
+; GFX9-NEXT: s_and_b64 s[8:9], vcc, exec
+; GFX9-NEXT: s_cselect_b32 s5, s5, s10
+; GFX9-NEXT: s_and_b32 s8, s4, 0xffff0000
+; GFX9-NEXT: v_add_f32_e32 v2, s8, v1
+; GFX9-NEXT: v_readfirstlane_b32 s8, v2
+; GFX9-NEXT: s_bfe_u32 s9, s8, 0x10010
+; GFX9-NEXT: s_add_i32 s9, s9, s8
; GFX9-NEXT: s_lshr_b32 s5, s5, 16
-; GFX9-NEXT: s_add_i32 s8, s7, 0x7fff
-; GFX9-NEXT: s_or_b32 s9, s6, 0x400000
+; GFX9-NEXT: s_add_i32 s10, s9, 0x7fff
+; GFX9-NEXT: s_or_b32 s11, s8, 0x400000
; GFX9-NEXT: v_cmp_u_f32_e32 vcc, v2, v2
-; GFX9-NEXT: s_and_b64 s[6:7], vcc, exec
-; GFX9-NEXT: s_cselect_b32 s6, s9, s8
+; GFX9-NEXT: s_and_b64 s[8:9], vcc, exec
+; GFX9-NEXT: s_cselect_b32 s8, s11, s10
; GFX9-NEXT: s_lshl_b32 s4, s4, 16
; GFX9-NEXT: v_add_f32_e32 v1, s4, v1
; GFX9-NEXT: v_readfirstlane_b32 s4, v1
-; GFX9-NEXT: s_lshr_b32 s8, s6, 16
-; GFX9-NEXT: s_bfe_u32 s6, s4, 0x10010
-; GFX9-NEXT: s_add_i32 s6, s6, s4
-; GFX9-NEXT: s_add_i32 s9, s6, 0x7fff
+; GFX9-NEXT: s_lshr_b32 s10, s8, 16
+; GFX9-NEXT: s_bfe_u32 s8, s4, 0x10010
+; GFX9-NEXT: s_add_i32 s8, s8, s4
+; GFX9-NEXT: s_add_i32 s11, s8, 0x7fff
; GFX9-NEXT: s_bitset1_b32 s4, 22
; GFX9-NEXT: v_cmp_u_f32_e32 vcc, v1, v1
-; GFX9-NEXT: s_and_b64 s[6:7], vcc, exec
-; GFX9-NEXT: s_cselect_b32 s4, s4, s9
+; GFX9-NEXT: s_and_b64 s[8:9], vcc, exec
+; GFX9-NEXT: s_cselect_b32 s4, s4, s11
; GFX9-NEXT: s_lshr_b32 s4, s4, 16
-; GFX9-NEXT: s_pack_ll_b32_b16 s47, s17, s76
-; GFX9-NEXT: s_pack_ll_b32_b16 s57, s19, s77
-; GFX9-NEXT: s_pack_ll_b32_b16 s59, s21, s78
+; GFX9-NEXT: s_pack_ll_b32_b16 s7, s17, s76
+; GFX9-NEXT: s_pack_ll_b32_b16 s6, s16, s6
+; GFX9-NEXT: s_pack_ll_b32_b16 s13, s19, s77
+; GFX9-NEXT: s_pack_ll_b32_b16 s45, s21, s78
; GFX9-NEXT: s_pack_ll_b32_b16 s61, s23, s79
; GFX9-NEXT: s_pack_ll_b32_b16 s63, s25, s88
; GFX9-NEXT: s_pack_ll_b32_b16 s73, s27, s89
; GFX9-NEXT: s_pack_ll_b32_b16 s75, s29, s90
; GFX9-NEXT: s_pack_ll_b32_b16 s31, s5, s91
-; GFX9-NEXT: s_pack_ll_b32_b16 s30, s4, s8
-; GFX9-NEXT: s_lshr_b64 s[6:7], s[30:31], 24
-; GFX9-NEXT: s_lshr_b64 s[8:9], s[74:75], 24
-; GFX9-NEXT: s_lshr_b64 s[10:11], s[72:73], 24
-; GFX9-NEXT: s_lshr_b64 s[12:13], s[62:63], 24
-; GFX9-NEXT: s_lshr_b64 s[14:15], s[60:61], 24
-; GFX9-NEXT: s_lshr_b64 s[40:41], s[58:59], 24
-; GFX9-NEXT: s_lshr_b64 s[42:43], s[56:57], 24
-; GFX9-NEXT: s_lshr_b64 s[44:45], s[46:47], 24
+; GFX9-NEXT: s_pack_ll_b32_b16 s30, s4, s10
+; GFX9-NEXT: s_lshr_b64 s[8:9], s[30:31], 24
+; GFX9-NEXT: s_lshr_b64 s[10:11], s[74:75], 24
+; GFX9-NEXT: s_lshr_b64 s[14:15], s[72:73], 24
+; GFX9-NEXT: s_lshr_b64 s[40:41], s[62:63], 24
+; GFX9-NEXT: s_lshr_b64 s[42:43], s[60:61], 24
+; GFX9-NEXT: s_lshr_b64 s[46:47], s[44:45], 24
+; GFX9-NEXT: s_lshr_b64 s[56:57], s[12:13], 24
+; GFX9-NEXT: s_lshr_b64 s[58:59], s[6:7], 24
; GFX9-NEXT: s_lshr_b32 s92, s31, 24
; GFX9-NEXT: s_lshr_b32 s93, s31, 8
; GFX9-NEXT: s_lshr_b32 s94, s30, 16
@@ -91661,178 +91608,178 @@ define inreg <64 x i8> @bitcast_v32bf16_to_v64i8_scalar(<32 x bfloat> inreg %a,
; GFX9-NEXT: s_lshr_b32 s61, s61, 8
; GFX9-NEXT: s_lshr_b32 s39, s60, 16
; GFX9-NEXT: s_lshr_b32 s60, s60, 8
-; GFX9-NEXT: s_lshr_b32 s48, s59, 24
-; GFX9-NEXT: s_lshr_b32 s59, s59, 8
-; GFX9-NEXT: s_lshr_b32 s49, s58, 16
-; GFX9-NEXT: s_lshr_b32 s58, s58, 8
-; GFX9-NEXT: s_lshr_b32 s50, s57, 24
-; GFX9-NEXT: s_lshr_b32 s57, s57, 8
-; GFX9-NEXT: s_lshr_b32 s51, s56, 16
-; GFX9-NEXT: s_lshr_b32 s56, s56, 8
-; GFX9-NEXT: s_lshr_b32 s52, s47, 24
-; GFX9-NEXT: s_lshr_b32 s53, s47, 8
-; GFX9-NEXT: s_lshr_b32 s54, s46, 16
-; GFX9-NEXT: s_lshr_b32 s55, s46, 8
+; GFX9-NEXT: s_lshr_b32 s48, s45, 24
+; GFX9-NEXT: s_lshr_b32 s45, s45, 8
+; GFX9-NEXT: s_lshr_b32 s49, s44, 16
+; GFX9-NEXT: s_lshr_b32 s44, s44, 8
+; GFX9-NEXT: s_lshr_b32 s50, s13, 24
+; GFX9-NEXT: s_lshr_b32 s13, s13, 8
+; GFX9-NEXT: s_lshr_b32 s51, s12, 16
+; GFX9-NEXT: s_lshr_b32 s12, s12, 8
+; GFX9-NEXT: s_lshr_b32 s52, s7, 24
+; GFX9-NEXT: s_lshr_b32 s53, s7, 8
+; GFX9-NEXT: s_lshr_b32 s54, s6, 16
+; GFX9-NEXT: s_lshr_b32 s55, s6, 8
; GFX9-NEXT: .LBB109_3: ; %end
-; GFX9-NEXT: s_and_b32 s7, s16, 0xff
-; GFX9-NEXT: s_lshl_b32 s9, s55, 8
-; GFX9-NEXT: s_or_b32 s7, s7, s9
-; GFX9-NEXT: s_and_b32 s9, s54, 0xff
-; GFX9-NEXT: s_lshl_b32 s11, s44, 8
-; GFX9-NEXT: s_or_b32 s9, s9, s11
-; GFX9-NEXT: s_and_b32 s7, s7, 0xffff
-; GFX9-NEXT: s_lshl_b32 s9, s9, 16
-; GFX9-NEXT: s_or_b32 s7, s7, s9
-; GFX9-NEXT: v_mov_b32_e32 v1, s7
-; GFX9-NEXT: s_and_b32 s7, s17, 0xff
-; GFX9-NEXT: s_lshl_b32 s9, s53, 8
+; GFX9-NEXT: s_and_b32 s6, s16, 0xff
+; GFX9-NEXT: s_lshl_b32 s7, s55, 8
+; GFX9-NEXT: s_or_b32 s6, s6, s7
+; GFX9-NEXT: s_and_b32 s7, s54, 0xff
+; GFX9-NEXT: s_lshl_b32 s9, s58, 8
; GFX9-NEXT: s_or_b32 s7, s7, s9
-; GFX9-NEXT: s_and_b32 s9, s76, 0xff
-; GFX9-NEXT: s_lshl_b32 s11, s52, 8
-; GFX9-NEXT: s_or_b32 s9, s9, s11
-; GFX9-NEXT: s_and_b32 s7, s7, 0xffff
-; GFX9-NEXT: s_lshl_b32 s9, s9, 16
+; GFX9-NEXT: s_and_b32 s6, s6, 0xffff
+; GFX9-NEXT: s_lshl_b32 s7, s7, 16
+; GFX9-NEXT: s_or_b32 s6, s6, s7
+; GFX9-NEXT: v_mov_b32_e32 v1, s6
+; GFX9-NEXT: s_and_b32 s6, s17, 0xff
+; GFX9-NEXT: s_lshl_b32 s7, s53, 8
+; GFX9-NEXT: s_or_b32 s6, s6, s7
+; GFX9-NEXT: s_and_b32 s7, s76, 0xff
+; GFX9-NEXT: s_lshl_b32 s9, s52, 8
; GFX9-NEXT: s_or_b32 s7, s7, s9
+; GFX9-NEXT: s_and_b32 s6, s6, 0xffff
+; GFX9-NEXT: s_lshl_b32 s7, s7, 16
+; GFX9-NEXT: s_or_b32 s6, s6, s7
; GFX9-NEXT: buffer_store_dword v1, v0, s[0:3], 0 offen
-; GFX9-NEXT: v_mov_b32_e32 v1, s7
-; GFX9-NEXT: s_and_b32 s7, s18, 0xff
+; GFX9-NEXT: v_mov_b32_e32 v1, s6
+; GFX9-NEXT: s_and_b32 s6, s18, 0xff
+; GFX9-NEXT: s_lshl_b32 s7, s12, 8
+; GFX9-NEXT: s_or_b32 s6, s6, s7
+; GFX9-NEXT: s_and_b32 s7, s51, 0xff
; GFX9-NEXT: s_lshl_b32 s9, s56, 8
; GFX9-NEXT: s_or_b32 s7, s7, s9
-; GFX9-NEXT: s_and_b32 s9, s51, 0xff
-; GFX9-NEXT: s_lshl_b32 s11, s42, 8
-; GFX9-NEXT: s_or_b32 s9, s9, s11
-; GFX9-NEXT: s_and_b32 s7, s7, 0xffff
-; GFX9-NEXT: s_lshl_b32 s9, s9, 16
-; GFX9-NEXT: s_or_b32 s7, s7, s9
+; GFX9-NEXT: s_and_b32 s6, s6, 0xffff
+; GFX9-NEXT: s_lshl_b32 s7, s7, 16
+; GFX9-NEXT: s_or_b32 s6, s6, s7
; GFX9-NEXT: buffer_store_dword v1, v0, s[0:3], 0 offen offset:4
-; GFX9-NEXT: v_mov_b32_e32 v1, s7
-; GFX9-NEXT: s_and_b32 s7, s19, 0xff
-; GFX9-NEXT: s_lshl_b32 s9, s57, 8
-; GFX9-NEXT: s_or_b32 s7, s7, s9
-; GFX9-NEXT: s_and_b32 s9, s77, 0xff
-; GFX9-NEXT: s_lshl_b32 s11, s50, 8
-; GFX9-NEXT: s_or_b32 s9, s9, s11
-; GFX9-NEXT: s_and_b32 s7, s7, 0xffff
-; GFX9-NEXT: s_lshl_b32 s9, s9, 16
+; GFX9-NEXT: v_mov_b32_e32 v1, s6
+; GFX9-NEXT: s_and_b32 s6, s19, 0xff
+; GFX9-NEXT: s_lshl_b32 s7, s13, 8
+; GFX9-NEXT: s_or_b32 s6, s6, s7
+; GFX9-NEXT: s_and_b32 s7, s77, 0xff
+; GFX9-NEXT: s_lshl_b32 s9, s50, 8
; GFX9-NEXT: s_or_b32 s7, s7, s9
+; GFX9-NEXT: s_and_b32 s6, s6, 0xffff
+; GFX9-NEXT: s_lshl_b32 s7, s7, 16
+; GFX9-NEXT: s_or_b32 s6, s6, s7
; GFX9-NEXT: buffer_store_dword v1, v0, s[0:3], 0 offen offset:8
-; GFX9-NEXT: v_mov_b32_e32 v1, s7
-; GFX9-NEXT: s_and_b32 s7, s20, 0xff
-; GFX9-NEXT: s_lshl_b32 s9, s58, 8
-; GFX9-NEXT: s_or_b32 s7, s7, s9
-; GFX9-NEXT: s_and_b32 s9, s49, 0xff
-; GFX9-NEXT: s_lshl_b32 s11, s40, 8
-; GFX9-NEXT: s_or_b32 s9, s9, s11
-; GFX9-NEXT: s_and_b32 s7, s7, 0xffff
-; GFX9-NEXT: s_lshl_b32 s9, s9, 16
+; GFX9-NEXT: v_mov_b32_e32 v1, s6
+; GFX9-NEXT: s_and_b32 s6, s20, 0xff
+; GFX9-NEXT: s_lshl_b32 s7, s44, 8
+; GFX9-NEXT: s_or_b32 s6, s6, s7
+; GFX9-NEXT: s_and_b32 s7, s49, 0xff
+; GFX9-NEXT: s_lshl_b32 s9, s46, 8
; GFX9-NEXT: s_or_b32 s7, s7, s9
+; GFX9-NEXT: s_and_b32 s6, s6, 0xffff
+; GFX9-NEXT: s_lshl_b32 s7, s7, 16
+; GFX9-NEXT: s_or_b32 s6, s6, s7
; GFX9-NEXT: buffer_store_dword v1, v0, s[0:3], 0 offen offset:12
-; GFX9-NEXT: v_mov_b32_e32 v1, s7
-; GFX9-NEXT: s_and_b32 s7, s21, 0xff
-; GFX9-NEXT: s_lshl_b32 s9, s59, 8
-; GFX9-NEXT: s_or_b32 s7, s7, s9
-; GFX9-NEXT: s_and_b32 s9, s78, 0xff
-; GFX9-NEXT: s_lshl_b32 s11, s48, 8
-; GFX9-NEXT: s_or_b32 s9, s9, s11
-; GFX9-NEXT: s_and_b32 s7, s7, 0xffff
-; GFX9-NEXT: s_lshl_b32 s9, s9, 16
+; GFX9-NEXT: v_mov_b32_e32 v1, s6
+; GFX9-NEXT: s_and_b32 s6, s21, 0xff
+; GFX9-NEXT: s_lshl_b32 s7, s45, 8
+; GFX9-NEXT: s_or_b32 s6, s6, s7
+; GFX9-NEXT: s_and_b32 s7, s78, 0xff
+; GFX9-NEXT: s_lshl_b32 s9, s48, 8
; GFX9-NEXT: s_or_b32 s7, s7, s9
+; GFX9-NEXT: s_and_b32 s6, s6, 0xffff
+; GFX9-NEXT: s_lshl_b32 s7, s7, 16
+; GFX9-NEXT: s_or_b32 s6, s6, s7
; GFX9-NEXT: buffer_store_dword v1, v0, s[0:3], 0 offen offset:16
-; GFX9-NEXT: v_mov_b32_e32 v1, s7
-; GFX9-NEXT: s_and_b32 s7, s22, 0xff
-; GFX9-NEXT: s_lshl_b32 s9, s60, 8
-; GFX9-NEXT: s_or_b32 s7, s7, s9
-; GFX9-NEXT: s_and_b32 s9, s39, 0xff
-; GFX9-NEXT: s_lshl_b32 s11, s14, 8
-; GFX9-NEXT: s_or_b32 s9, s9, s11
-; GFX9-NEXT: s_and_b32 s7, s7, 0xffff
-; GFX9-NEXT: s_lshl_b32 s9, s9, 16
+; GFX9-NEXT: v_mov_b32_e32 v1, s6
+; GFX9-NEXT: s_and_b32 s6, s22, 0xff
+; GFX9-NEXT: s_lshl_b32 s7, s60, 8
+; GFX9-NEXT: s_or_b32 s6, s6, s7
+; GFX9-NEXT: s_and_b32 s7, s39, 0xff
+; GFX9-NEXT: s_lshl_b32 s9, s42, 8
; GFX9-NEXT: s_or_b32 s7, s7, s9
+; GFX9-NEXT: s_and_b32 s6, s6, 0xffff
+; GFX9-NEXT: s_lshl_b32 s7, s7, 16
+; GFX9-NEXT: s_or_b32 s6, s6, s7
; GFX9-NEXT: buffer_store_dword v1, v0, s[0:3], 0 offen offset:20
-; GFX9-NEXT: v_mov_b32_e32 v1, s7
-; GFX9-NEXT: s_and_b32 s7, s23, 0xff
-; GFX9-NEXT: s_lshl_b32 s9, s61, 8
-; GFX9-NEXT: s_or_b32 s7, s7, s9
-; GFX9-NEXT: s_and_b32 s9, s79, 0xff
-; GFX9-NEXT: s_lshl_b32 s11, s38, 8
-; GFX9-NEXT: s_or_b32 s9, s9, s11
-; GFX9-NEXT: s_and_b32 s7, s7, 0xffff
-; GFX9-NEXT: s_lshl_b32 s9, s9, 16
+; GFX9-NEXT: v_mov_b32_e32 v1, s6
+; GFX9-NEXT: s_and_b32 s6, s23, 0xff
+; GFX9-NEXT: s_lshl_b32 s7, s61, 8
+; GFX9-NEXT: s_or_b32 s6, s6, s7
+; GFX9-NEXT: s_and_b32 s7, s79, 0xff
+; GFX9-NEXT: s_lshl_b32 s9, s38, 8
; GFX9-NEXT: s_or_b32 s7, s7, s9
+; GFX9-NEXT: s_and_b32 s6, s6, 0xffff
+; GFX9-NEXT: s_lshl_b32 s7, s7, 16
+; GFX9-NEXT: s_or_b32 s6, s6, s7
; GFX9-NEXT: buffer_store_dword v1, v0, s[0:3], 0 offen offset:24
-; GFX9-NEXT: v_mov_b32_e32 v1, s7
-; GFX9-NEXT: s_and_b32 s7, s24, 0xff
-; GFX9-NEXT: s_lshl_b32 s9, s62, 8
-; GFX9-NEXT: s_or_b32 s7, s7, s9
-; GFX9-NEXT: s_and_b32 s9, s37, 0xff
-; GFX9-NEXT: s_lshl_b32 s11, s12, 8
-; GFX9-NEXT: s_or_b32 s9, s9, s11
-; GFX9-NEXT: s_and_b32 s7, s7, 0xffff
-; GFX9-NEXT: s_lshl_b32 s9, s9, 16
+; GFX9-NEXT: v_mov_b32_e32 v1, s6
+; GFX9-NEXT: s_and_b32 s6, s24, 0xff
+; GFX9-NEXT: s_lshl_b32 s7, s62, 8
+; GFX9-NEXT: s_or_b32 s6, s6, s7
+; GFX9-NEXT: s_and_b32 s7, s37, 0xff
+; GFX9-NEXT: s_lshl_b32 s9, s40, 8
; GFX9-NEXT: s_or_b32 s7, s7, s9
+; GFX9-NEXT: s_and_b32 s6, s6, 0xffff
+; GFX9-NEXT: s_lshl_b32 s7, s7, 16
+; GFX9-NEXT: s_or_b32 s6, s6, s7
; GFX9-NEXT: buffer_store_dword v1, v0, s[0:3], 0 offen offset:28
-; GFX9-NEXT: v_mov_b32_e32 v1, s7
-; GFX9-NEXT: s_and_b32 s7, s25, 0xff
-; GFX9-NEXT: s_lshl_b32 s9, s63, 8
-; GFX9-NEXT: s_or_b32 s7, s7, s9
-; GFX9-NEXT: s_and_b32 s9, s88, 0xff
-; GFX9-NEXT: s_lshl_b32 s11, s36, 8
-; GFX9-NEXT: s_or_b32 s9, s9, s11
-; GFX9-NEXT: s_and_b32 s7, s7, 0xffff
-; GFX9-NEXT: s_lshl_b32 s9, s9, 16
+; GFX9-NEXT: v_mov_b32_e32 v1, s6
+; GFX9-NEXT: s_and_b32 s6, s25, 0xff
+; GFX9-NEXT: s_lshl_b32 s7, s63, 8
+; GFX9-NEXT: s_or_b32 s6, s6, s7
+; GFX9-NEXT: s_and_b32 s7, s88, 0xff
+; GFX9-NEXT: s_lshl_b32 s9, s36, 8
; GFX9-NEXT: s_or_b32 s7, s7, s9
+; GFX9-NEXT: s_and_b32 s6, s6, 0xffff
+; GFX9-NEXT: s_lshl_b32 s7, s7, 16
+; GFX9-NEXT: s_or_b32 s6, s6, s7
; GFX9-NEXT: buffer_store_dword v1, v0, s[0:3], 0 offen offset:32
-; GFX9-NEXT: v_mov_b32_e32 v1, s7
-; GFX9-NEXT: s_and_b32 s7, s26, 0xff
-; GFX9-NEXT: s_lshl_b32 s9, s72, 8
-; GFX9-NEXT: s_or_b32 s7, s7, s9
-; GFX9-NEXT: s_and_b32 s9, s35, 0xff
-; GFX9-NEXT: s_lshl_b32 s10, s10, 8
-; GFX9-NEXT: s_or_b32 s9, s9, s10
-; GFX9-NEXT: s_and_b32 s7, s7, 0xffff
-; GFX9-NEXT: s_lshl_b32 s9, s9, 16
+; GFX9-NEXT: v_mov_b32_e32 v1, s6
+; GFX9-NEXT: s_and_b32 s6, s26, 0xff
+; GFX9-NEXT: s_lshl_b32 s7, s72, 8
+; GFX9-NEXT: s_or_b32 s6, s6, s7
+; GFX9-NEXT: s_and_b32 s7, s35, 0xff
+; GFX9-NEXT: s_lshl_b32 s9, s14, 8
; GFX9-NEXT: s_or_b32 s7, s7, s9
+; GFX9-NEXT: s_and_b32 s6, s6, 0xffff
+; GFX9-NEXT: s_lshl_b32 s7, s7, 16
+; GFX9-NEXT: s_or_b32 s6, s6, s7
; GFX9-NEXT: buffer_store_dword v1, v0, s[0:3], 0 offen offset:36
-; GFX9-NEXT: v_mov_b32_e32 v1, s7
-; GFX9-NEXT: s_and_b32 s7, s27, 0xff
-; GFX9-NEXT: s_lshl_b32 s9, s73, 8
-; GFX9-NEXT: s_or_b32 s7, s7, s9
-; GFX9-NEXT: s_and_b32 s9, s89, 0xff
-; GFX9-NEXT: s_lshl_b32 s10, s34, 8
-; GFX9-NEXT: s_or_b32 s9, s9, s10
-; GFX9-NEXT: s_and_b32 s7, s7, 0xffff
-; GFX9-NEXT: s_lshl_b32 s9, s9, 16
+; GFX9-NEXT: v_mov_b32_e32 v1, s6
+; GFX9-NEXT: s_and_b32 s6, s27, 0xff
+; GFX9-NEXT: s_lshl_b32 s7, s73, 8
+; GFX9-NEXT: s_or_b32 s6, s6, s7
+; GFX9-NEXT: s_and_b32 s7, s89, 0xff
+; GFX9-NEXT: s_lshl_b32 s9, s34, 8
; GFX9-NEXT: s_or_b32 s7, s7, s9
+; GFX9-NEXT: s_and_b32 s6, s6, 0xffff
+; GFX9-NEXT: s_lshl_b32 s7, s7, 16
+; GFX9-NEXT: s_or_b32 s6, s6, s7
; GFX9-NEXT: buffer_store_dword v1, v0, s[0:3], 0 offen offset:40
-; GFX9-NEXT: v_mov_b32_e32 v1, s7
-; GFX9-NEXT: s_and_b32 s7, s28, 0xff
-; GFX9-NEXT: s_lshl_b32 s9, s74, 8
+; GFX9-NEXT: v_mov_b32_e32 v1, s6
+; GFX9-NEXT: s_and_b32 s6, s28, 0xff
+; GFX9-NEXT: s_lshl_b32 s7, s74, 8
+; GFX9-NEXT: s_or_b32 s6, s6, s7
+; GFX9-NEXT: s_and_b32 s7, s31, 0xff
+; GFX9-NEXT: s_lshl_b32 s9, s10, 8
; GFX9-NEXT: s_or_b32 s7, s7, s9
-; GFX9-NEXT: s_and_b32 s9, s31, 0xff
-; GFX9-NEXT: s_lshl_b32 s8, s8, 8
-; GFX9-NEXT: s_or_b32 s8, s9, s8
-; GFX9-NEXT: s_and_b32 s7, s7, 0xffff
-; GFX9-NEXT: s_lshl_b32 s8, s8, 16
-; GFX9-NEXT: s_or_b32 s7, s7, s8
+; GFX9-NEXT: s_and_b32 s6, s6, 0xffff
+; GFX9-NEXT: s_lshl_b32 s7, s7, 16
+; GFX9-NEXT: s_or_b32 s6, s6, s7
; GFX9-NEXT: buffer_store_dword v1, v0, s[0:3], 0 offen offset:44
-; GFX9-NEXT: v_mov_b32_e32 v1, s7
-; GFX9-NEXT: s_and_b32 s7, s29, 0xff
-; GFX9-NEXT: s_lshl_b32 s8, s75, 8
-; GFX9-NEXT: s_or_b32 s7, s7, s8
-; GFX9-NEXT: s_and_b32 s8, s90, 0xff
+; GFX9-NEXT: v_mov_b32_e32 v1, s6
+; GFX9-NEXT: s_and_b32 s6, s29, 0xff
+; GFX9-NEXT: s_lshl_b32 s7, s75, 8
+; GFX9-NEXT: s_or_b32 s6, s6, s7
+; GFX9-NEXT: s_and_b32 s7, s90, 0xff
; GFX9-NEXT: s_lshl_b32 s9, s30, 8
-; GFX9-NEXT: s_or_b32 s8, s8, s9
-; GFX9-NEXT: s_and_b32 s7, s7, 0xffff
-; GFX9-NEXT: s_lshl_b32 s8, s8, 16
-; GFX9-NEXT: s_or_b32 s7, s7, s8
+; GFX9-NEXT: s_or_b32 s7, s7, s9
+; GFX9-NEXT: s_and_b32 s6, s6, 0xffff
+; GFX9-NEXT: s_lshl_b32 s7, s7, 16
+; GFX9-NEXT: s_or_b32 s6, s6, s7
; GFX9-NEXT: buffer_store_dword v1, v0, s[0:3], 0 offen offset:48
-; GFX9-NEXT: v_mov_b32_e32 v1, s7
+; GFX9-NEXT: v_mov_b32_e32 v1, s6
; GFX9-NEXT: s_and_b32 s4, s4, 0xff
-; GFX9-NEXT: s_lshl_b32 s7, s95, 8
-; GFX9-NEXT: s_or_b32 s4, s4, s7
-; GFX9-NEXT: s_and_b32 s7, s94, 0xff
-; GFX9-NEXT: s_lshl_b32 s6, s6, 8
-; GFX9-NEXT: s_or_b32 s6, s7, s6
+; GFX9-NEXT: s_lshl_b32 s6, s95, 8
+; GFX9-NEXT: s_or_b32 s4, s4, s6
+; GFX9-NEXT: s_and_b32 s6, s94, 0xff
+; GFX9-NEXT: s_lshl_b32 s7, s8, 8
+; GFX9-NEXT: s_or_b32 s6, s6, s7
; GFX9-NEXT: s_and_b32 s4, s4, 0xffff
; GFX9-NEXT: s_lshl_b32 s6, s6, 16
; GFX9-NEXT: s_or_b32 s4, s4, s6
@@ -91874,49 +91821,49 @@ define inreg <64 x i8> @bitcast_v32bf16_to_v64i8_scalar(<32 x bfloat> inreg %a,
; GFX9-NEXT: .LBB109_4:
; GFX9-NEXT: ; implicit-def: $sgpr55
; GFX9-NEXT: ; implicit-def: $sgpr54
-; GFX9-NEXT: ; implicit-def: $sgpr44
+; GFX9-NEXT: ; implicit-def: $sgpr58
; GFX9-NEXT: ; implicit-def: $sgpr53
; GFX9-NEXT: ; implicit-def: $sgpr76
; GFX9-NEXT: ; implicit-def: $sgpr52
-; GFX9-NEXT: ; implicit-def: $sgpr56
+; GFX9-NEXT: ; implicit-def: $sgpr12
; GFX9-NEXT: ; implicit-def: $sgpr51
-; GFX9-NEXT: ; implicit-def: $sgpr42
-; GFX9-NEXT: ; implicit-def: $sgpr57
+; GFX9-NEXT: ; implicit-def: $sgpr56
+; GFX9-NEXT: ; implicit-def: $sgpr13
; GFX9-NEXT: ; implicit-def: $sgpr77
; GFX9-NEXT: ; implicit-def: $sgpr50
-; GFX9-NEXT: ; implicit-def: $sgpr58
+; GFX9-NEXT: ; implicit-def: $sgpr44
; GFX9-NEXT: ; implicit-def: $sgpr49
-; GFX9-NEXT: ; implicit-def: $sgpr40
-; GFX9-NEXT: ; implicit-def: $sgpr59
+; GFX9-NEXT: ; implicit-def: $sgpr46
+; GFX9-NEXT: ; implicit-def: $sgpr45
; GFX9-NEXT: ; implicit-def: $sgpr78
; GFX9-NEXT: ; implicit-def: $sgpr48
; GFX9-NEXT: ; implicit-def: $sgpr60
; GFX9-NEXT: ; implicit-def: $sgpr39
-; GFX9-NEXT: ; implicit-def: $sgpr14
+; GFX9-NEXT: ; implicit-def: $sgpr42
; GFX9-NEXT: ; implicit-def: $sgpr61
; GFX9-NEXT: ; implicit-def: $sgpr79
; GFX9-NEXT: ; implicit-def: $sgpr38
; GFX9-NEXT: ; implicit-def: $sgpr62
; GFX9-NEXT: ; implicit-def: $sgpr37
-; GFX9-NEXT: ; implicit-def: $sgpr12
+; GFX9-NEXT: ; implicit-def: $sgpr40
; GFX9-NEXT: ; implicit-def: $sgpr63
; GFX9-NEXT: ; implicit-def: $sgpr88
; GFX9-NEXT: ; implicit-def: $sgpr36
; GFX9-NEXT: ; implicit-def: $sgpr72
; GFX9-NEXT: ; implicit-def: $sgpr35
-; GFX9-NEXT: ; implicit-def: $sgpr10
+; GFX9-NEXT: ; implicit-def: $sgpr14
; GFX9-NEXT: ; implicit-def: $sgpr73
; GFX9-NEXT: ; implicit-def: $sgpr89
; GFX9-NEXT: ; implicit-def: $sgpr34
; GFX9-NEXT: ; implicit-def: $sgpr74
; GFX9-NEXT: ; implicit-def: $sgpr31
-; GFX9-NEXT: ; implicit-def: $sgpr8
+; GFX9-NEXT: ; implicit-def: $sgpr10
; GFX9-NEXT: ; implicit-def: $sgpr75
; GFX9-NEXT: ; implicit-def: $sgpr90
; GFX9-NEXT: ; implicit-def: $sgpr30
; GFX9-NEXT: ; implicit-def: $sgpr95
; GFX9-NEXT: ; implicit-def: $sgpr94
-; GFX9-NEXT: ; implicit-def: $sgpr6
+; GFX9-NEXT: ; implicit-def: $sgpr8
; GFX9-NEXT: ; implicit-def: $sgpr93
; GFX9-NEXT: ; implicit-def: $sgpr91
; GFX9-NEXT: ; implicit-def: $sgpr92
@@ -91930,7 +91877,7 @@ define inreg <64 x i8> @bitcast_v32bf16_to_v64i8_scalar(<32 x bfloat> inreg %a,
; GFX11-TRUE16-NEXT: s_mov_b32 exec_lo, s4
; GFX11-TRUE16-NEXT: v_writelane_b32 v17, s30, 0
; GFX11-TRUE16-NEXT: s_cmp_lg_u32 s28, 0
-; GFX11-TRUE16-NEXT: s_mov_b32 vcc_lo, 0
+; GFX11-TRUE16-NEXT: s_mov_b32 s42, 0
; GFX11-TRUE16-NEXT: v_writelane_b32 v17, s31, 1
; GFX11-TRUE16-NEXT: v_writelane_b32 v17, s34, 2
; GFX11-TRUE16-NEXT: v_writelane_b32 v17, s35, 3
@@ -91940,6 +91887,7 @@ define inreg <64 x i8> @bitcast_v32bf16_to_v64i8_scalar(<32 x bfloat> inreg %a,
; GFX11-TRUE16-NEXT: v_writelane_b32 v17, s39, 7
; GFX11-TRUE16-NEXT: v_writelane_b32 v17, s48, 8
; GFX11-TRUE16-NEXT: v_writelane_b32 v17, s49, 9
+; GFX11-TRUE16-NEXT: v_writelane_b32 v17, s50, 10
; GFX11-TRUE16-NEXT: s_cbranch_scc0 .LBB109_4
; GFX11-TRUE16-NEXT: ; %bb.1: ; %cmp.false
; GFX11-TRUE16-NEXT: s_lshr_b32 s62, s27, 24
@@ -91981,7 +91929,7 @@ define inreg <64 x i8> @bitcast_v32bf16_to_v64i8_scalar(<32 x bfloat> inreg %a,
; GFX11-TRUE16-NEXT: s_lshr_b32 s46, s1, 16
; GFX11-TRUE16-NEXT: s_lshr_b32 s48, s1, 8
; GFX11-TRUE16-NEXT: s_lshr_b32 s49, s0, 16
-; GFX11-TRUE16-NEXT: s_lshr_b32 s42, s0, 8
+; GFX11-TRUE16-NEXT: s_lshr_b32 s50, s0, 8
; GFX11-TRUE16-NEXT: s_lshr_b64 s[8:9], s[26:27], 24
; GFX11-TRUE16-NEXT: s_lshr_b64 s[4:5], s[24:25], 24
; GFX11-TRUE16-NEXT: s_lshr_b64 s[6:7], s[22:23], 24
@@ -91990,7 +91938,7 @@ define inreg <64 x i8> @bitcast_v32bf16_to_v64i8_scalar(<32 x bfloat> inreg %a,
; GFX11-TRUE16-NEXT: s_lshr_b64 s[14:15], s[16:17], 24
; GFX11-TRUE16-NEXT: s_lshr_b64 s[28:29], s[2:3], 24
; GFX11-TRUE16-NEXT: s_lshr_b64 s[40:41], s[0:1], 24
-; GFX11-TRUE16-NEXT: s_and_not1_b32 vcc_lo, exec_lo, vcc_lo
+; GFX11-TRUE16-NEXT: s_and_not1_b32 vcc_lo, exec_lo, s42
; GFX11-TRUE16-NEXT: s_cbranch_vccnz .LBB109_3
; GFX11-TRUE16-NEXT: .LBB109_2: ; %cmp.true
; GFX11-TRUE16-NEXT: s_and_b32 s4, s1, 0xffff0000
@@ -92433,10 +92381,10 @@ define inreg <64 x i8> @bitcast_v32bf16_to_v64i8_scalar(<32 x bfloat> inreg %a,
; GFX11-TRUE16-NEXT: s_lshr_b64 s[8:9], vcc, 24
; GFX11-TRUE16-NEXT: s_lshr_b32 s43, vcc_lo, 16
; GFX11-TRUE16-NEXT: s_lshr_b32 s44, vcc_lo, 8
-; GFX11-TRUE16-NEXT: s_lshr_b32 s42, s42, 8
+; GFX11-TRUE16-NEXT: s_lshr_b32 s50, s42, 8
; GFX11-TRUE16-NEXT: .LBB109_3: ; %end
; GFX11-TRUE16-NEXT: s_delay_alu instid0(SALU_CYCLE_1)
-; GFX11-TRUE16-NEXT: s_mov_b32 s5, s42
+; GFX11-TRUE16-NEXT: s_mov_b32 s5, s50
; GFX11-TRUE16-NEXT: s_mov_b32 s7, s49
; GFX11-TRUE16-NEXT: s_mov_b32 s9, s40
; GFX11-TRUE16-NEXT: s_and_b32 s0, s0, 0xff
@@ -92652,6 +92600,7 @@ define inreg <64 x i8> @bitcast_v32bf16_to_v64i8_scalar(<32 x bfloat> inreg %a,
; GFX11-TRUE16-NEXT: scratch_store_b128 v0, v[5:8], off offset:16
; GFX11-TRUE16-NEXT: scratch_store_b128 v0, v[9:12], off offset:32
; GFX11-TRUE16-NEXT: scratch_store_b128 v0, v[13:16], off offset:48
+; GFX11-TRUE16-NEXT: v_readlane_b32 s50, v17, 10
; GFX11-TRUE16-NEXT: v_readlane_b32 s49, v17, 9
; GFX11-TRUE16-NEXT: v_readlane_b32 s48, v17, 8
; GFX11-TRUE16-NEXT: v_readlane_b32 s39, v17, 7
@@ -92668,7 +92617,7 @@ define inreg <64 x i8> @bitcast_v32bf16_to_v64i8_scalar(<32 x bfloat> inreg %a,
; GFX11-TRUE16-NEXT: s_waitcnt vmcnt(0)
; GFX11-TRUE16-NEXT: s_setpc_b64 s[30:31]
; GFX11-TRUE16-NEXT: .LBB109_4:
-; GFX11-TRUE16-NEXT: ; implicit-def: $sgpr42_lo16
+; GFX11-TRUE16-NEXT: ; implicit-def: $sgpr50_lo16
; GFX11-TRUE16-NEXT: ; implicit-def: $sgpr49_lo16
; GFX11-TRUE16-NEXT: ; implicit-def: $sgpr40_lo16
; GFX11-TRUE16-NEXT: ; implicit-def: $sgpr48_lo16
@@ -92726,7 +92675,7 @@ define inreg <64 x i8> @bitcast_v32bf16_to_v64i8_scalar(<32 x bfloat> inreg %a,
; GFX11-FAKE16-NEXT: s_mov_b32 exec_lo, s4
; GFX11-FAKE16-NEXT: v_writelane_b32 v17, s30, 0
; GFX11-FAKE16-NEXT: s_cmp_lg_u32 s28, 0
-; GFX11-FAKE16-NEXT: s_mov_b32 vcc_lo, 0
+; GFX11-FAKE16-NEXT: s_mov_b32 s42, 0
; GFX11-FAKE16-NEXT: v_writelane_b32 v17, s31, 1
; GFX11-FAKE16-NEXT: v_writelane_b32 v17, s34, 2
; GFX11-FAKE16-NEXT: v_writelane_b32 v17, s35, 3
@@ -92735,8 +92684,9 @@ define inreg <64 x i8> @bitcast_v32bf16_to_v64i8_scalar(<32 x bfloat> inreg %a,
; GFX11-FAKE16-NEXT: v_writelane_b32 v17, s38, 6
; GFX11-FAKE16-NEXT: v_writelane_b32 v17, s39, 7
; GFX11-FAKE16-NEXT: v_writelane_b32 v17, s48, 8
-; GFX11-FAKE16-NEXT: v_writelane_b32 v17, s50, 9
-; GFX11-FAKE16-NEXT: v_writelane_b32 v17, s51, 10
+; GFX11-FAKE16-NEXT: v_writelane_b32 v17, s49, 9
+; GFX11-FAKE16-NEXT: v_writelane_b32 v17, s50, 10
+; GFX11-FAKE16-NEXT: v_writelane_b32 v17, s51, 11
; GFX11-FAKE16-NEXT: s_cbranch_scc0 .LBB109_4
; GFX11-FAKE16-NEXT: ; %bb.1: ; %cmp.false
; GFX11-FAKE16-NEXT: s_lshr_b32 s62, s27, 24
@@ -92778,7 +92728,7 @@ define inreg <64 x i8> @bitcast_v32bf16_to_v64i8_scalar(<32 x bfloat> inreg %a,
; GFX11-FAKE16-NEXT: s_lshr_b32 s46, s1, 16
; GFX11-FAKE16-NEXT: s_lshr_b32 s39, s1, 8
; GFX11-FAKE16-NEXT: s_lshr_b32 s48, s0, 16
-; GFX11-FAKE16-NEXT: s_lshr_b32 s42, s0, 8
+; GFX11-FAKE16-NEXT: s_lshr_b32 s49, s0, 8
; GFX11-FAKE16-NEXT: s_lshr_b64 s[10:11], s[26:27], 24
; GFX11-FAKE16-NEXT: s_lshr_b64 s[4:5], s[24:25], 24
; GFX11-FAKE16-NEXT: s_lshr_b64 s[6:7], s[22:23], 24
@@ -92787,7 +92737,7 @@ define inreg <64 x i8> @bitcast_v32bf16_to_v64i8_scalar(<32 x bfloat> inreg %a,
; GFX11-FAKE16-NEXT: s_lshr_b64 s[14:15], s[16:17], 24
; GFX11-FAKE16-NEXT: s_lshr_b64 s[28:29], s[2:3], 24
; GFX11-FAKE16-NEXT: s_lshr_b64 s[40:41], s[0:1], 24
-; GFX11-FAKE16-NEXT: s_and_not1_b32 vcc_lo, exec_lo, vcc_lo
+; GFX11-FAKE16-NEXT: s_and_not1_b32 vcc_lo, exec_lo, s42
; GFX11-FAKE16-NEXT: s_cbranch_vccnz .LBB109_3
; GFX11-FAKE16-NEXT: .LBB109_2: ; %cmp.true
; GFX11-FAKE16-NEXT: s_and_b32 s4, s1, 0xffff0000
@@ -93230,10 +93180,10 @@ define inreg <64 x i8> @bitcast_v32bf16_to_v64i8_scalar(<32 x bfloat> inreg %a,
; GFX11-FAKE16-NEXT: s_lshr_b64 s[10:11], s[50:51], 24
; GFX11-FAKE16-NEXT: s_lshr_b32 s43, s50, 16
; GFX11-FAKE16-NEXT: s_lshr_b32 s44, s50, 8
-; GFX11-FAKE16-NEXT: s_lshr_b32 s42, s42, 8
+; GFX11-FAKE16-NEXT: s_lshr_b32 s49, s42, 8
; GFX11-FAKE16-NEXT: .LBB109_3: ; %end
; GFX11-FAKE16-NEXT: s_and_b32 s0, s0, 0xff
-; GFX11-FAKE16-NEXT: s_lshl_b32 s5, s42, 8
+; GFX11-FAKE16-NEXT: s_lshl_b32 s5, s49, 8
; GFX11-FAKE16-NEXT: s_and_b32 s7, s48, 0xff
; GFX11-FAKE16-NEXT: s_lshl_b32 s9, s40, 8
; GFX11-FAKE16-NEXT: s_or_b32 s0, s0, s5
@@ -93389,8 +93339,9 @@ define inreg <64 x i8> @bitcast_v32bf16_to_v64i8_scalar(<32 x bfloat> inreg %a,
; GFX11-FAKE16-NEXT: scratch_store_b128 v0, v[5:8], off offset:16
; GFX11-FAKE16-NEXT: scratch_store_b128 v0, v[9:12], off offset:32
; GFX11-FAKE16-NEXT: scratch_store_b128 v0, v[13:16], off offset:48
-; GFX11-FAKE16-NEXT: v_readlane_b32 s51, v17, 10
-; GFX11-FAKE16-NEXT: v_readlane_b32 s50, v17, 9
+; GFX11-FAKE16-NEXT: v_readlane_b32 s51, v17, 11
+; GFX11-FAKE16-NEXT: v_readlane_b32 s50, v17, 10
+; GFX11-FAKE16-NEXT: v_readlane_b32 s49, v17, 9
; GFX11-FAKE16-NEXT: v_readlane_b32 s48, v17, 8
; GFX11-FAKE16-NEXT: v_readlane_b32 s39, v17, 7
; GFX11-FAKE16-NEXT: v_readlane_b32 s38, v17, 6
@@ -93406,7 +93357,7 @@ define inreg <64 x i8> @bitcast_v32bf16_to_v64i8_scalar(<32 x bfloat> inreg %a,
; GFX11-FAKE16-NEXT: s_waitcnt vmcnt(0)
; GFX11-FAKE16-NEXT: s_setpc_b64 s[30:31]
; GFX11-FAKE16-NEXT: .LBB109_4:
-; GFX11-FAKE16-NEXT: ; implicit-def: $sgpr42
+; GFX11-FAKE16-NEXT: ; implicit-def: $sgpr49
; GFX11-FAKE16-NEXT: ; implicit-def: $sgpr48
; GFX11-FAKE16-NEXT: ; implicit-def: $sgpr40
; GFX11-FAKE16-NEXT: ; implicit-def: $sgpr39
@@ -95984,12 +95935,12 @@ define inreg <32 x bfloat> @bitcast_v64i8_to_v32bf16_scalar(<64 x i8> inreg %a,
; SI-NEXT: buffer_load_dword v45, off, s[0:3], s32 offset:72
; SI-NEXT: buffer_load_dword v43, off, s[0:3], s32 offset:68
; SI-NEXT: v_readfirstlane_b32 s46, v30
-; SI-NEXT: v_readfirstlane_b32 s44, v23
-; SI-NEXT: v_readfirstlane_b32 s45, v22
-; SI-NEXT: v_readfirstlane_b32 s41, v15
-; SI-NEXT: v_readfirstlane_b32 s43, v14
-; SI-NEXT: v_readfirstlane_b32 s10, v7
-; SI-NEXT: v_readfirstlane_b32 s12, v6
+; SI-NEXT: v_readfirstlane_b32 s41, v23
+; SI-NEXT: v_readfirstlane_b32 s43, v22
+; SI-NEXT: v_readfirstlane_b32 s10, v15
+; SI-NEXT: v_readfirstlane_b32 s12, v14
+; SI-NEXT: v_readfirstlane_b32 s8, v7
+; SI-NEXT: v_readfirstlane_b32 s9, v6
; SI-NEXT: v_readfirstlane_b32 s7, v1
; SI-NEXT: v_readfirstlane_b32 s6, v0
; SI-NEXT: v_lshlrev_b32_e32 v0, 24, v5
@@ -96022,47 +95973,47 @@ define inreg <32 x bfloat> @bitcast_v64i8_to_v32bf16_scalar(<64 x i8> inreg %a,
; SI-NEXT: s_and_b32 s4, s16, 0xff
; SI-NEXT: s_lshl_b32 s4, s4, 16
; SI-NEXT: s_lshl_b32 s5, s17, 24
-; SI-NEXT: s_or_b32 s8, s5, s4
+; SI-NEXT: s_or_b32 s11, s5, s4
; SI-NEXT: s_and_b32 s4, s18, 0xff
; SI-NEXT: s_lshl_b32 s4, s4, 16
; SI-NEXT: s_lshl_b32 s5, s19, 24
-; SI-NEXT: s_or_b32 s9, s5, s4
+; SI-NEXT: s_or_b32 s13, s5, s4
; SI-NEXT: s_and_b32 s4, s20, 0xff
; SI-NEXT: s_lshl_b32 s5, s21, 8
; SI-NEXT: s_or_b32 s4, s4, s5
-; SI-NEXT: s_lshl_b32 s11, s4, 16
+; SI-NEXT: s_lshl_b32 s14, s4, 16
; SI-NEXT: s_and_b32 s4, s22, 0xff
; SI-NEXT: s_lshl_b32 s4, s4, 16
; SI-NEXT: s_lshl_b32 s5, s23, 24
-; SI-NEXT: s_or_b32 s13, s5, s4
+; SI-NEXT: s_or_b32 s15, s5, s4
; SI-NEXT: s_and_b32 s4, s24, 0xff
; SI-NEXT: s_lshl_b32 s4, s4, 16
; SI-NEXT: s_lshl_b32 s5, s25, 24
-; SI-NEXT: s_or_b32 s14, s5, s4
+; SI-NEXT: s_or_b32 s40, s5, s4
; SI-NEXT: s_and_b32 s4, s26, 0xff
; SI-NEXT: s_lshl_b32 s4, s4, 16
; SI-NEXT: s_lshl_b32 s5, s27, 24
-; SI-NEXT: s_or_b32 s15, s5, s4
+; SI-NEXT: s_or_b32 s42, s5, s4
; SI-NEXT: s_and_b32 s4, s28, 0xff
; SI-NEXT: s_lshl_b32 s5, s29, 8
; SI-NEXT: s_or_b32 s4, s4, s5
-; SI-NEXT: s_lshl_b32 s40, s4, 16
+; SI-NEXT: s_lshl_b32 s44, s4, 16
; SI-NEXT: s_and_b32 s4, s6, 0xff
; SI-NEXT: s_lshl_b32 s4, s4, 16
; SI-NEXT: s_lshl_b32 s5, s7, 24
-; SI-NEXT: s_or_b32 s42, s5, s4
-; SI-NEXT: s_and_b32 s4, s12, 0xff
-; SI-NEXT: s_lshl_b32 s5, s10, 8
+; SI-NEXT: s_or_b32 s45, s5, s4
+; SI-NEXT: s_and_b32 s4, s9, 0xff
+; SI-NEXT: s_lshl_b32 s5, s8, 8
; SI-NEXT: s_or_b32 s4, s4, s5
; SI-NEXT: s_lshl_b32 s60, s4, 16
-; SI-NEXT: s_and_b32 s4, s43, 0xff
-; SI-NEXT: s_lshl_b32 s5, s41, 8
+; SI-NEXT: s_and_b32 s4, s12, 0xff
+; SI-NEXT: s_lshl_b32 s5, s10, 8
; SI-NEXT: s_or_b32 s4, s4, s5
; SI-NEXT: v_and_b32_e32 v9, 0xff, v2
; SI-NEXT: s_lshl_b32 s61, s4, 16
; SI-NEXT: v_and_b32_e32 v17, 0xff, v18
-; SI-NEXT: s_and_b32 s4, s45, 0xff
-; SI-NEXT: s_lshl_b32 s5, s44, 8
+; SI-NEXT: s_and_b32 s4, s43, 0xff
+; SI-NEXT: s_lshl_b32 s5, s41, 8
; SI-NEXT: v_and_b32_e32 v25, 0xff, v52
; SI-NEXT: v_lshlrev_b32_e32 v9, 16, v9
; SI-NEXT: v_lshlrev_b32_e32 v13, 24, v3
@@ -96210,10 +96161,10 @@ define inreg <32 x bfloat> @bitcast_v64i8_to_v32bf16_scalar(<64 x i8> inreg %a,
; SI-NEXT: v_and_b32_e32 v9, 0xffff, v9
; SI-NEXT: v_or_b32_e32 v13, v22, v13
; SI-NEXT: v_or_b32_e32 v9, v13, v9
-; SI-NEXT: s_add_i32 s45, s45, 3
+; SI-NEXT: s_add_i32 s43, s43, 3
; SI-NEXT: v_add_i32_e32 v22, vcc, 0x3000000, v9
-; SI-NEXT: s_and_b32 s4, s45, 0xff
-; SI-NEXT: s_lshl_b32 s5, s44, 8
+; SI-NEXT: s_and_b32 s4, s43, 0xff
+; SI-NEXT: s_lshl_b32 s5, s41, 8
; SI-NEXT: v_add_i32_e32 v9, vcc, 3, v24
; SI-NEXT: s_or_b32 s4, s5, s4
; SI-NEXT: v_and_b32_e32 v9, 0xff, v9
@@ -96233,10 +96184,10 @@ define inreg <32 x bfloat> @bitcast_v64i8_to_v32bf16_scalar(<64 x i8> inreg %a,
; SI-NEXT: v_lshlrev_b32_e32 v13, 16, v13
; SI-NEXT: v_and_b32_e32 v9, 0xffff, v9
; SI-NEXT: v_or_b32_e32 v7, v7, v13
-; SI-NEXT: s_add_i32 s43, s43, 3
+; SI-NEXT: s_add_i32 s12, s12, 3
; SI-NEXT: v_or_b32_e32 v7, v7, v9
-; SI-NEXT: s_and_b32 s4, s43, 0xff
-; SI-NEXT: s_lshl_b32 s5, s41, 8
+; SI-NEXT: s_and_b32 s4, s12, 0xff
+; SI-NEXT: s_lshl_b32 s5, s10, 8
; SI-NEXT: v_add_i32_e32 v9, vcc, 3, v16
; SI-NEXT: s_or_b32 s4, s5, s4
; SI-NEXT: v_and_b32_e32 v9, 0xff, v9
@@ -96244,10 +96195,10 @@ define inreg <32 x bfloat> @bitcast_v64i8_to_v32bf16_scalar(<64 x i8> inreg %a,
; SI-NEXT: v_lshlrev_b32_e32 v9, 16, v9
; SI-NEXT: s_and_b32 s4, s4, 0xffff
; SI-NEXT: v_or_b32_e32 v6, v6, v9
-; SI-NEXT: s_add_i32 s12, s12, 3
+; SI-NEXT: s_add_i32 s9, s9, 3
; SI-NEXT: v_or_b32_e32 v6, s4, v6
-; SI-NEXT: s_and_b32 s4, s12, 0xff
-; SI-NEXT: s_lshl_b32 s5, s10, 8
+; SI-NEXT: s_and_b32 s4, s9, 0xff
+; SI-NEXT: s_lshl_b32 s5, s8, 8
; SI-NEXT: v_add_i32_e32 v8, vcc, 3, v8
; SI-NEXT: s_or_b32 s4, s5, s4
; SI-NEXT: v_and_b32_e32 v8, 0xff, v8
@@ -96335,14 +96286,14 @@ define inreg <32 x bfloat> @bitcast_v64i8_to_v32bf16_scalar(<64 x i8> inreg %a,
; SI-NEXT: s_add_i32 s5, s5, 0x3000000
; SI-NEXT: s_add_i32 s6, s6, 0x3000000
; SI-NEXT: s_add_i32 s7, s7, 0x3000000
-; SI-NEXT: s_and_b32 s9, s7, 0xffff0000
-; SI-NEXT: s_lshl_b32 s8, s7, 16
-; SI-NEXT: s_and_b32 s13, s6, 0xffff0000
-; SI-NEXT: s_lshl_b32 s11, s6, 16
-; SI-NEXT: s_and_b32 s15, s5, 0xffff0000
-; SI-NEXT: s_lshl_b32 s14, s5, 16
-; SI-NEXT: s_and_b32 s42, s4, 0xffff0000
-; SI-NEXT: s_lshl_b32 s40, s4, 16
+; SI-NEXT: s_and_b32 s13, s7, 0xffff0000
+; SI-NEXT: s_lshl_b32 s11, s7, 16
+; SI-NEXT: s_and_b32 s15, s6, 0xffff0000
+; SI-NEXT: s_lshl_b32 s14, s6, 16
+; SI-NEXT: s_and_b32 s42, s5, 0xffff0000
+; SI-NEXT: s_lshl_b32 s40, s5, 16
+; SI-NEXT: s_and_b32 s45, s4, 0xffff0000
+; SI-NEXT: s_lshl_b32 s44, s4, 16
; SI-NEXT: v_and_b32_e32 v9, 0xffff0000, v0
; SI-NEXT: v_lshlrev_b32_e32 v37, 16, v0
; SI-NEXT: v_and_b32_e32 v38, 0xffff0000, v1
@@ -96369,14 +96320,14 @@ define inreg <32 x bfloat> @bitcast_v64i8_to_v32bf16_scalar(<64 x i8> inreg %a,
; SI-NEXT: v_lshlrev_b32_e32 v30, 16, v43
; SI-NEXT: s_branch .LBB111_5
; SI-NEXT: .LBB111_3:
-; SI-NEXT: ; implicit-def: $sgpr8
-; SI-NEXT: ; implicit-def: $sgpr9
; SI-NEXT: ; implicit-def: $sgpr11
; SI-NEXT: ; implicit-def: $sgpr13
; SI-NEXT: ; implicit-def: $sgpr14
; SI-NEXT: ; implicit-def: $sgpr15
; SI-NEXT: ; implicit-def: $sgpr40
; SI-NEXT: ; implicit-def: $sgpr42
+; SI-NEXT: ; implicit-def: $sgpr44
+; SI-NEXT: ; implicit-def: $sgpr45
; SI-NEXT: ; implicit-def: $vgpr37
; SI-NEXT: ; implicit-def: $vgpr9
; SI-NEXT: ; implicit-def: $sgpr60
@@ -96416,14 +96367,14 @@ define inreg <32 x bfloat> @bitcast_v64i8_to_v32bf16_scalar(<64 x i8> inreg %a,
; SI-NEXT: buffer_load_dword v42, off, s[0:3], s32 offset:92 ; 4-byte Folded Reload
; SI-NEXT: buffer_load_dword v41, off, s[0:3], s32 offset:96 ; 4-byte Folded Reload
; SI-NEXT: buffer_load_dword v40, off, s[0:3], s32 offset:100 ; 4-byte Folded Reload
-; SI-NEXT: v_mov_b32_e32 v0, s8
-; SI-NEXT: v_mov_b32_e32 v1, s9
-; SI-NEXT: v_mov_b32_e32 v2, s11
-; SI-NEXT: v_mov_b32_e32 v3, s13
-; SI-NEXT: v_mov_b32_e32 v4, s14
-; SI-NEXT: v_mov_b32_e32 v5, s15
-; SI-NEXT: v_mov_b32_e32 v6, s40
-; SI-NEXT: v_mov_b32_e32 v7, s42
+; SI-NEXT: v_mov_b32_e32 v0, s11
+; SI-NEXT: v_mov_b32_e32 v1, s13
+; SI-NEXT: v_mov_b32_e32 v2, s14
+; SI-NEXT: v_mov_b32_e32 v3, s15
+; SI-NEXT: v_mov_b32_e32 v4, s40
+; SI-NEXT: v_mov_b32_e32 v5, s42
+; SI-NEXT: v_mov_b32_e32 v6, s44
+; SI-NEXT: v_mov_b32_e32 v7, s45
; SI-NEXT: v_mov_b32_e32 v8, v37
; SI-NEXT: v_mov_b32_e32 v11, v38
; SI-NEXT: v_mov_b32_e32 v12, v48
@@ -96515,11 +96466,11 @@ define inreg <32 x bfloat> @bitcast_v64i8_to_v32bf16_scalar(<64 x i8> inreg %a,
; VI-NEXT: s_waitcnt vmcnt(11)
; VI-NEXT: v_lshlrev_b32_e32 v14, 8, v14
; VI-NEXT: s_waitcnt vmcnt(9)
-; VI-NEXT: v_lshlrev_b32_e32 v38, 8, v39
+; VI-NEXT: v_lshlrev_b32_e32 v22, 8, v39
; VI-NEXT: s_waitcnt vmcnt(7)
; VI-NEXT: v_lshlrev_b32_e32 v51, 8, v48
; VI-NEXT: s_waitcnt vmcnt(5)
-; VI-NEXT: v_lshlrev_b32_e32 v22, 8, v49
+; VI-NEXT: v_lshlrev_b32_e32 v38, 8, v49
; VI-NEXT: s_cbranch_scc0 .LBB111_4
; VI-NEXT: ; %bb.1: ; %cmp.false
; VI-NEXT: v_or_b32_sdwa v0, v2, v1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
@@ -96578,10 +96529,10 @@ define inreg <32 x bfloat> @bitcast_v64i8_to_v32bf16_scalar(<64 x i8> inreg %a,
; VI-NEXT: v_or_b32_sdwa v1, v43, v33 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
; VI-NEXT: v_or_b32_sdwa v13, v0, v1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_0 src1_sel:DWORD
; VI-NEXT: v_or_b32_sdwa v0, v44, v14 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
-; VI-NEXT: v_or_b32_sdwa v1, v45, v38 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; VI-NEXT: v_or_b32_sdwa v1, v45, v22 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
; VI-NEXT: v_or_b32_sdwa v14, v0, v1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_0 src1_sel:DWORD
; VI-NEXT: v_or_b32_sdwa v0, v47, v51 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
-; VI-NEXT: v_or_b32_sdwa v1, v57, v22 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; VI-NEXT: v_or_b32_sdwa v1, v57, v38 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
; VI-NEXT: v_or_b32_sdwa v15, v0, v1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_0 src1_sel:DWORD
; VI-NEXT: v_or_b32_sdwa v0, v32, v50 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
; VI-NEXT: v_or_b32_e32 v3, s4, v0
@@ -96641,9 +96592,9 @@ define inreg <32 x bfloat> @bitcast_v64i8_to_v32bf16_scalar(<64 x i8> inreg %a,
; VI-NEXT: v_add_u32_e32 v3, vcc, 3, v24
; VI-NEXT: v_or_b32_sdwa v9, v58, v3 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:BYTE_0
; VI-NEXT: v_add_u32_e32 v3, vcc, 3, v23
-; VI-NEXT: v_add_u32_e32 v0, vcc, 3, v57
+; VI-NEXT: v_add_u32_e32 v2, vcc, 3, v45
; VI-NEXT: v_or_b32_sdwa v3, v56, v3 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:BYTE_0
-; VI-NEXT: v_or_b32_sdwa v0, v22, v0 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:BYTE_0
+; VI-NEXT: v_or_b32_sdwa v2, v22, v2 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:BYTE_0
; VI-NEXT: v_add_u32_e32 v22, vcc, 0x300, v3
; VI-NEXT: v_add_u32_e32 v3, vcc, 3, v55
; VI-NEXT: v_or_b32_sdwa v8, v46, v3 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:BYTE_0
@@ -96700,8 +96651,8 @@ define inreg <32 x bfloat> @bitcast_v64i8_to_v32bf16_scalar(<64 x i8> inreg %a,
; VI-NEXT: s_addk_i32 s6, 0x300
; VI-NEXT: s_addk_i32 s8, 0x300
; VI-NEXT: s_addk_i32 s10, 0x300
+; VI-NEXT: v_add_u32_e32 v0, vcc, 3, v57
; VI-NEXT: v_or_b32_sdwa v1, v48, v1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:BYTE_0
-; VI-NEXT: v_add_u32_e32 v2, vcc, 3, v45
; VI-NEXT: s_addk_i32 s4, 0x300
; VI-NEXT: s_lshl_b32 s5, s5, 16
; VI-NEXT: s_lshl_b32 s7, s7, 16
@@ -96709,8 +96660,8 @@ define inreg <32 x bfloat> @bitcast_v64i8_to_v32bf16_scalar(<64 x i8> inreg %a,
; VI-NEXT: s_and_b32 s10, s10, 0xffff
; VI-NEXT: s_and_b32 s8, s8, 0xffff
; VI-NEXT: s_and_b32 s6, s6, 0xffff
+; VI-NEXT: v_or_b32_sdwa v0, v38, v0 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:BYTE_0
; VI-NEXT: v_add_u32_e32 v1, vcc, 0x300, v1
-; VI-NEXT: v_or_b32_sdwa v2, v38, v2 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:BYTE_0
; VI-NEXT: s_or_b32 s9, s9, s10
; VI-NEXT: s_or_b32 s7, s7, s8
; VI-NEXT: s_or_b32 s5, s5, s6
diff --git a/llvm/test/CodeGen/AMDGPU/amdgcn.bitcast.576bit.ll b/llvm/test/CodeGen/AMDGPU/amdgcn.bitcast.576bit.ll
index 3a26a5c263d78..70a4fdce8d9a0 100644
--- a/llvm/test/CodeGen/AMDGPU/amdgcn.bitcast.576bit.ll
+++ b/llvm/test/CodeGen/AMDGPU/amdgcn.bitcast.576bit.ll
@@ -3264,7 +3264,7 @@ define inreg <36 x i16> @bitcast_v18i32_to_v36i16_scalar(<18 x i32> inreg %a, i3
; GFX11-TRUE16: ; %bb.0:
; GFX11-TRUE16-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX11-TRUE16-NEXT: v_cmp_ne_u32_e32 vcc_lo, 0, v0
-; GFX11-TRUE16-NEXT: s_mov_b32 s46, 0
+; GFX11-TRUE16-NEXT: s_mov_b32 s43, 0
; GFX11-TRUE16-NEXT: s_and_b32 s4, vcc_lo, exec_lo
; GFX11-TRUE16-NEXT: s_cbranch_scc0 .LBB13_4
; GFX11-TRUE16-NEXT: ; %bb.1: ; %cmp.false
@@ -3283,10 +3283,10 @@ define inreg <36 x i16> @bitcast_v18i32_to_v36i16_scalar(<18 x i32> inreg %a, i3
; GFX11-TRUE16-NEXT: s_lshr_b32 s40, s17, 16
; GFX11-TRUE16-NEXT: s_lshr_b32 s41, s16, 16
; GFX11-TRUE16-NEXT: s_lshr_b32 s42, s3, 16
-; GFX11-TRUE16-NEXT: s_lshr_b32 s43, s2, 16
-; GFX11-TRUE16-NEXT: s_lshr_b32 s44, s1, 16
-; GFX11-TRUE16-NEXT: s_lshr_b32 s45, s0, 16
-; GFX11-TRUE16-NEXT: s_and_not1_b32 vcc_lo, exec_lo, s46
+; GFX11-TRUE16-NEXT: s_lshr_b32 s44, s2, 16
+; GFX11-TRUE16-NEXT: s_lshr_b32 s45, s1, 16
+; GFX11-TRUE16-NEXT: s_lshr_b32 s46, s0, 16
+; GFX11-TRUE16-NEXT: s_and_not1_b32 vcc_lo, exec_lo, s43
; GFX11-TRUE16-NEXT: s_cbranch_vccnz .LBB13_3
; GFX11-TRUE16-NEXT: .LBB13_2: ; %cmp.true
; GFX11-TRUE16-NEXT: s_add_i32 s29, s29, 3
@@ -3322,9 +3322,9 @@ define inreg <36 x i16> @bitcast_v18i32_to_v36i16_scalar(<18 x i32> inreg %a, i3
; GFX11-TRUE16-NEXT: s_lshr_b32 s40, s17, 16
; GFX11-TRUE16-NEXT: s_lshr_b32 s41, s16, 16
; GFX11-TRUE16-NEXT: s_lshr_b32 s42, s3, 16
-; GFX11-TRUE16-NEXT: s_lshr_b32 s43, s2, 16
-; GFX11-TRUE16-NEXT: s_lshr_b32 s44, s1, 16
-; GFX11-TRUE16-NEXT: s_lshr_b32 s45, s0, 16
+; GFX11-TRUE16-NEXT: s_lshr_b32 s44, s2, 16
+; GFX11-TRUE16-NEXT: s_lshr_b32 s45, s1, 16
+; GFX11-TRUE16-NEXT: s_lshr_b32 s46, s0, 16
; GFX11-TRUE16-NEXT: .LBB13_3: ; %end
; GFX11-TRUE16-NEXT: s_pack_ll_b32_b16 s15, s18, s15
; GFX11-TRUE16-NEXT: s_pack_ll_b32_b16 s14, s19, s14
@@ -3338,16 +3338,18 @@ define inreg <36 x i16> @bitcast_v18i32_to_v36i16_scalar(<18 x i32> inreg %a, i3
; GFX11-TRUE16-NEXT: s_pack_ll_b32_b16 s10, s19, s10
; GFX11-TRUE16-NEXT: s_mov_b32 s18, s24
; GFX11-TRUE16-NEXT: s_mov_b32 s19, s25
+; GFX11-TRUE16-NEXT: s_mov_b32 s43, s46
; GFX11-TRUE16-NEXT: s_pack_ll_b32_b16 s9, s18, s9
; GFX11-TRUE16-NEXT: s_pack_ll_b32_b16 s8, s19, s8
; GFX11-TRUE16-NEXT: s_mov_b32 s18, s26
; GFX11-TRUE16-NEXT: s_mov_b32 s19, s27
+; GFX11-TRUE16-NEXT: s_pack_ll_b32_b16 s0, s0, s43
+; GFX11-TRUE16-NEXT: s_mov_b32 s43, s44
; GFX11-TRUE16-NEXT: s_pack_ll_b32_b16 s7, s18, s7
; GFX11-TRUE16-NEXT: s_pack_ll_b32_b16 s6, s19, s6
; GFX11-TRUE16-NEXT: s_mov_b32 s18, s28
; GFX11-TRUE16-NEXT: s_mov_b32 s19, s29
-; GFX11-TRUE16-NEXT: s_pack_ll_b32_b16 s0, s0, s45
-; GFX11-TRUE16-NEXT: s_pack_ll_b32_b16 s1, s1, s44
+; GFX11-TRUE16-NEXT: s_pack_ll_b32_b16 s1, s1, s45
; GFX11-TRUE16-NEXT: s_pack_ll_b32_b16 s2, s2, s43
; GFX11-TRUE16-NEXT: s_pack_ll_b32_b16 s3, s3, s42
; GFX11-TRUE16-NEXT: s_pack_ll_b32_b16 s16, s16, s41
@@ -3365,9 +3367,9 @@ define inreg <36 x i16> @bitcast_v18i32_to_v36i16_scalar(<18 x i32> inreg %a, i3
; GFX11-TRUE16-NEXT: v_dual_mov_b32 v16, s5 :: v_dual_mov_b32 v17, s4
; GFX11-TRUE16-NEXT: s_setpc_b64 s[30:31]
; GFX11-TRUE16-NEXT: .LBB13_4:
+; GFX11-TRUE16-NEXT: ; implicit-def: $sgpr46_lo16
; GFX11-TRUE16-NEXT: ; implicit-def: $sgpr45_lo16
; GFX11-TRUE16-NEXT: ; implicit-def: $sgpr44_lo16
-; GFX11-TRUE16-NEXT: ; implicit-def: $sgpr43_lo16
; GFX11-TRUE16-NEXT: ; implicit-def: $sgpr42_lo16
; GFX11-TRUE16-NEXT: ; implicit-def: $sgpr41_lo16
; GFX11-TRUE16-NEXT: ; implicit-def: $sgpr40_lo16
@@ -6339,7 +6341,7 @@ define inreg <36 x half> @bitcast_v18i32_to_v36f16_scalar(<18 x i32> inreg %a, i
; GFX11-TRUE16: ; %bb.0:
; GFX11-TRUE16-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX11-TRUE16-NEXT: v_cmp_ne_u32_e32 vcc_lo, 0, v0
-; GFX11-TRUE16-NEXT: s_mov_b32 s46, 0
+; GFX11-TRUE16-NEXT: s_mov_b32 s43, 0
; GFX11-TRUE16-NEXT: s_and_b32 s4, vcc_lo, exec_lo
; GFX11-TRUE16-NEXT: s_cbranch_scc0 .LBB17_4
; GFX11-TRUE16-NEXT: ; %bb.1: ; %cmp.false
@@ -6358,10 +6360,10 @@ define inreg <36 x half> @bitcast_v18i32_to_v36f16_scalar(<18 x i32> inreg %a, i
; GFX11-TRUE16-NEXT: s_lshr_b32 s40, s17, 16
; GFX11-TRUE16-NEXT: s_lshr_b32 s41, s16, 16
; GFX11-TRUE16-NEXT: s_lshr_b32 s42, s3, 16
-; GFX11-TRUE16-NEXT: s_lshr_b32 s43, s2, 16
-; GFX11-TRUE16-NEXT: s_lshr_b32 s44, s1, 16
-; GFX11-TRUE16-NEXT: s_lshr_b32 s45, s0, 16
-; GFX11-TRUE16-NEXT: s_and_not1_b32 vcc_lo, exec_lo, s46
+; GFX11-TRUE16-NEXT: s_lshr_b32 s44, s2, 16
+; GFX11-TRUE16-NEXT: s_lshr_b32 s45, s1, 16
+; GFX11-TRUE16-NEXT: s_lshr_b32 s46, s0, 16
+; GFX11-TRUE16-NEXT: s_and_not1_b32 vcc_lo, exec_lo, s43
; GFX11-TRUE16-NEXT: s_cbranch_vccnz .LBB17_3
; GFX11-TRUE16-NEXT: .LBB17_2: ; %cmp.true
; GFX11-TRUE16-NEXT: s_add_i32 s29, s29, 3
@@ -6397,9 +6399,9 @@ define inreg <36 x half> @bitcast_v18i32_to_v36f16_scalar(<18 x i32> inreg %a, i
; GFX11-TRUE16-NEXT: s_lshr_b32 s40, s17, 16
; GFX11-TRUE16-NEXT: s_lshr_b32 s41, s16, 16
; GFX11-TRUE16-NEXT: s_lshr_b32 s42, s3, 16
-; GFX11-TRUE16-NEXT: s_lshr_b32 s43, s2, 16
-; GFX11-TRUE16-NEXT: s_lshr_b32 s44, s1, 16
-; GFX11-TRUE16-NEXT: s_lshr_b32 s45, s0, 16
+; GFX11-TRUE16-NEXT: s_lshr_b32 s44, s2, 16
+; GFX11-TRUE16-NEXT: s_lshr_b32 s45, s1, 16
+; GFX11-TRUE16-NEXT: s_lshr_b32 s46, s0, 16
; GFX11-TRUE16-NEXT: .LBB17_3: ; %end
; GFX11-TRUE16-NEXT: s_pack_ll_b32_b16 s15, s18, s15
; GFX11-TRUE16-NEXT: s_pack_ll_b32_b16 s14, s19, s14
@@ -6413,16 +6415,18 @@ define inreg <36 x half> @bitcast_v18i32_to_v36f16_scalar(<18 x i32> inreg %a, i
; GFX11-TRUE16-NEXT: s_pack_ll_b32_b16 s10, s19, s10
; GFX11-TRUE16-NEXT: s_mov_b32 s18, s24
; GFX11-TRUE16-NEXT: s_mov_b32 s19, s25
+; GFX11-TRUE16-NEXT: s_mov_b32 s43, s46
; GFX11-TRUE16-NEXT: s_pack_ll_b32_b16 s9, s18, s9
; GFX11-TRUE16-NEXT: s_pack_ll_b32_b16 s8, s19, s8
; GFX11-TRUE16-NEXT: s_mov_b32 s18, s26
; GFX11-TRUE16-NEXT: s_mov_b32 s19, s27
+; GFX11-TRUE16-NEXT: s_pack_ll_b32_b16 s0, s0, s43
+; GFX11-TRUE16-NEXT: s_mov_b32 s43, s44
; GFX11-TRUE16-NEXT: s_pack_ll_b32_b16 s7, s18, s7
; GFX11-TRUE16-NEXT: s_pack_ll_b32_b16 s6, s19, s6
; GFX11-TRUE16-NEXT: s_mov_b32 s18, s28
; GFX11-TRUE16-NEXT: s_mov_b32 s19, s29
-; GFX11-TRUE16-NEXT: s_pack_ll_b32_b16 s0, s0, s45
-; GFX11-TRUE16-NEXT: s_pack_ll_b32_b16 s1, s1, s44
+; GFX11-TRUE16-NEXT: s_pack_ll_b32_b16 s1, s1, s45
; GFX11-TRUE16-NEXT: s_pack_ll_b32_b16 s2, s2, s43
; GFX11-TRUE16-NEXT: s_pack_ll_b32_b16 s3, s3, s42
; GFX11-TRUE16-NEXT: s_pack_ll_b32_b16 s16, s16, s41
@@ -6440,9 +6444,9 @@ define inreg <36 x half> @bitcast_v18i32_to_v36f16_scalar(<18 x i32> inreg %a, i
; GFX11-TRUE16-NEXT: v_dual_mov_b32 v16, s5 :: v_dual_mov_b32 v17, s4
; GFX11-TRUE16-NEXT: s_setpc_b64 s[30:31]
; GFX11-TRUE16-NEXT: .LBB17_4:
+; GFX11-TRUE16-NEXT: ; implicit-def: $sgpr46_lo16
; GFX11-TRUE16-NEXT: ; implicit-def: $sgpr45_lo16
; GFX11-TRUE16-NEXT: ; implicit-def: $sgpr44_lo16
-; GFX11-TRUE16-NEXT: ; implicit-def: $sgpr43_lo16
; GFX11-TRUE16-NEXT: ; implicit-def: $sgpr42_lo16
; GFX11-TRUE16-NEXT: ; implicit-def: $sgpr41_lo16
; GFX11-TRUE16-NEXT: ; implicit-def: $sgpr40_lo16
@@ -17770,7 +17774,7 @@ define inreg <36 x i16> @bitcast_v9i64_to_v36i16_scalar(<9 x i64> inreg %a, i32
; GFX11-TRUE16: ; %bb.0:
; GFX11-TRUE16-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX11-TRUE16-NEXT: v_cmp_ne_u32_e32 vcc_lo, 0, v0
-; GFX11-TRUE16-NEXT: s_mov_b32 s46, 0
+; GFX11-TRUE16-NEXT: s_mov_b32 s43, 0
; GFX11-TRUE16-NEXT: s_and_b32 s4, vcc_lo, exec_lo
; GFX11-TRUE16-NEXT: s_cbranch_scc0 .LBB41_4
; GFX11-TRUE16-NEXT: ; %bb.1: ; %cmp.false
@@ -17789,10 +17793,10 @@ define inreg <36 x i16> @bitcast_v9i64_to_v36i16_scalar(<9 x i64> inreg %a, i32
; GFX11-TRUE16-NEXT: s_lshr_b32 s40, s17, 16
; GFX11-TRUE16-NEXT: s_lshr_b32 s41, s16, 16
; GFX11-TRUE16-NEXT: s_lshr_b32 s42, s3, 16
-; GFX11-TRUE16-NEXT: s_lshr_b32 s43, s2, 16
-; GFX11-TRUE16-NEXT: s_lshr_b32 s44, s1, 16
-; GFX11-TRUE16-NEXT: s_lshr_b32 s45, s0, 16
-; GFX11-TRUE16-NEXT: s_and_not1_b32 vcc_lo, exec_lo, s46
+; GFX11-TRUE16-NEXT: s_lshr_b32 s44, s2, 16
+; GFX11-TRUE16-NEXT: s_lshr_b32 s45, s1, 16
+; GFX11-TRUE16-NEXT: s_lshr_b32 s46, s0, 16
+; GFX11-TRUE16-NEXT: s_and_not1_b32 vcc_lo, exec_lo, s43
; GFX11-TRUE16-NEXT: s_cbranch_vccnz .LBB41_3
; GFX11-TRUE16-NEXT: .LBB41_2: ; %cmp.true
; GFX11-TRUE16-NEXT: s_add_u32 s28, s28, 3
@@ -17828,9 +17832,9 @@ define inreg <36 x i16> @bitcast_v9i64_to_v36i16_scalar(<9 x i64> inreg %a, i32
; GFX11-TRUE16-NEXT: s_lshr_b32 s40, s17, 16
; GFX11-TRUE16-NEXT: s_lshr_b32 s41, s16, 16
; GFX11-TRUE16-NEXT: s_lshr_b32 s42, s3, 16
-; GFX11-TRUE16-NEXT: s_lshr_b32 s43, s2, 16
-; GFX11-TRUE16-NEXT: s_lshr_b32 s44, s1, 16
-; GFX11-TRUE16-NEXT: s_lshr_b32 s45, s0, 16
+; GFX11-TRUE16-NEXT: s_lshr_b32 s44, s2, 16
+; GFX11-TRUE16-NEXT: s_lshr_b32 s45, s1, 16
+; GFX11-TRUE16-NEXT: s_lshr_b32 s46, s0, 16
; GFX11-TRUE16-NEXT: .LBB41_3: ; %end
; GFX11-TRUE16-NEXT: s_pack_ll_b32_b16 s15, s18, s15
; GFX11-TRUE16-NEXT: s_pack_ll_b32_b16 s14, s19, s14
@@ -17844,16 +17848,18 @@ define inreg <36 x i16> @bitcast_v9i64_to_v36i16_scalar(<9 x i64> inreg %a, i32
; GFX11-TRUE16-NEXT: s_pack_ll_b32_b16 s10, s19, s10
; GFX11-TRUE16-NEXT: s_mov_b32 s18, s24
; GFX11-TRUE16-NEXT: s_mov_b32 s19, s25
+; GFX11-TRUE16-NEXT: s_mov_b32 s43, s46
; GFX11-TRUE16-NEXT: s_pack_ll_b32_b16 s9, s18, s9
; GFX11-TRUE16-NEXT: s_pack_ll_b32_b16 s8, s19, s8
; GFX11-TRUE16-NEXT: s_mov_b32 s18, s26
; GFX11-TRUE16-NEXT: s_mov_b32 s19, s27
+; GFX11-TRUE16-NEXT: s_pack_ll_b32_b16 s0, s0, s43
+; GFX11-TRUE16-NEXT: s_mov_b32 s43, s44
; GFX11-TRUE16-NEXT: s_pack_ll_b32_b16 s7, s18, s7
; GFX11-TRUE16-NEXT: s_pack_ll_b32_b16 s6, s19, s6
; GFX11-TRUE16-NEXT: s_mov_b32 s18, s28
; GFX11-TRUE16-NEXT: s_mov_b32 s19, s29
-; GFX11-TRUE16-NEXT: s_pack_ll_b32_b16 s0, s0, s45
-; GFX11-TRUE16-NEXT: s_pack_ll_b32_b16 s1, s1, s44
+; GFX11-TRUE16-NEXT: s_pack_ll_b32_b16 s1, s1, s45
; GFX11-TRUE16-NEXT: s_pack_ll_b32_b16 s2, s2, s43
; GFX11-TRUE16-NEXT: s_pack_ll_b32_b16 s3, s3, s42
; GFX11-TRUE16-NEXT: s_pack_ll_b32_b16 s16, s16, s41
@@ -17871,9 +17877,9 @@ define inreg <36 x i16> @bitcast_v9i64_to_v36i16_scalar(<9 x i64> inreg %a, i32
; GFX11-TRUE16-NEXT: v_dual_mov_b32 v16, s5 :: v_dual_mov_b32 v17, s4
; GFX11-TRUE16-NEXT: s_setpc_b64 s[30:31]
; GFX11-TRUE16-NEXT: .LBB41_4:
+; GFX11-TRUE16-NEXT: ; implicit-def: $sgpr46_lo16
; GFX11-TRUE16-NEXT: ; implicit-def: $sgpr45_lo16
; GFX11-TRUE16-NEXT: ; implicit-def: $sgpr44_lo16
-; GFX11-TRUE16-NEXT: ; implicit-def: $sgpr43_lo16
; GFX11-TRUE16-NEXT: ; implicit-def: $sgpr42_lo16
; GFX11-TRUE16-NEXT: ; implicit-def: $sgpr41_lo16
; GFX11-TRUE16-NEXT: ; implicit-def: $sgpr40_lo16
@@ -20855,7 +20861,7 @@ define inreg <36 x half> @bitcast_v9i64_to_v36f16_scalar(<9 x i64> inreg %a, i32
; GFX11-TRUE16: ; %bb.0:
; GFX11-TRUE16-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX11-TRUE16-NEXT: v_cmp_ne_u32_e32 vcc_lo, 0, v0
-; GFX11-TRUE16-NEXT: s_mov_b32 s46, 0
+; GFX11-TRUE16-NEXT: s_mov_b32 s43, 0
; GFX11-TRUE16-NEXT: s_and_b32 s4, vcc_lo, exec_lo
; GFX11-TRUE16-NEXT: s_cbranch_scc0 .LBB45_4
; GFX11-TRUE16-NEXT: ; %bb.1: ; %cmp.false
@@ -20874,10 +20880,10 @@ define inreg <36 x half> @bitcast_v9i64_to_v36f16_scalar(<9 x i64> inreg %a, i32
; GFX11-TRUE16-NEXT: s_lshr_b32 s40, s17, 16
; GFX11-TRUE16-NEXT: s_lshr_b32 s41, s16, 16
; GFX11-TRUE16-NEXT: s_lshr_b32 s42, s3, 16
-; GFX11-TRUE16-NEXT: s_lshr_b32 s43, s2, 16
-; GFX11-TRUE16-NEXT: s_lshr_b32 s44, s1, 16
-; GFX11-TRUE16-NEXT: s_lshr_b32 s45, s0, 16
-; GFX11-TRUE16-NEXT: s_and_not1_b32 vcc_lo, exec_lo, s46
+; GFX11-TRUE16-NEXT: s_lshr_b32 s44, s2, 16
+; GFX11-TRUE16-NEXT: s_lshr_b32 s45, s1, 16
+; GFX11-TRUE16-NEXT: s_lshr_b32 s46, s0, 16
+; GFX11-TRUE16-NEXT: s_and_not1_b32 vcc_lo, exec_lo, s43
; GFX11-TRUE16-NEXT: s_cbranch_vccnz .LBB45_3
; GFX11-TRUE16-NEXT: .LBB45_2: ; %cmp.true
; GFX11-TRUE16-NEXT: s_add_u32 s28, s28, 3
@@ -20913,9 +20919,9 @@ define inreg <36 x half> @bitcast_v9i64_to_v36f16_scalar(<9 x i64> inreg %a, i32
; GFX11-TRUE16-NEXT: s_lshr_b32 s40, s17, 16
; GFX11-TRUE16-NEXT: s_lshr_b32 s41, s16, 16
; GFX11-TRUE16-NEXT: s_lshr_b32 s42, s3, 16
-; GFX11-TRUE16-NEXT: s_lshr_b32 s43, s2, 16
-; GFX11-TRUE16-NEXT: s_lshr_b32 s44, s1, 16
-; GFX11-TRUE16-NEXT: s_lshr_b32 s45, s0, 16
+; GFX11-TRUE16-NEXT: s_lshr_b32 s44, s2, 16
+; GFX11-TRUE16-NEXT: s_lshr_b32 s45, s1, 16
+; GFX11-TRUE16-NEXT: s_lshr_b32 s46, s0, 16
; GFX11-TRUE16-NEXT: .LBB45_3: ; %end
; GFX11-TRUE16-NEXT: s_pack_ll_b32_b16 s15, s18, s15
; GFX11-TRUE16-NEXT: s_pack_ll_b32_b16 s14, s19, s14
@@ -20929,16 +20935,18 @@ define inreg <36 x half> @bitcast_v9i64_to_v36f16_scalar(<9 x i64> inreg %a, i32
; GFX11-TRUE16-NEXT: s_pack_ll_b32_b16 s10, s19, s10
; GFX11-TRUE16-NEXT: s_mov_b32 s18, s24
; GFX11-TRUE16-NEXT: s_mov_b32 s19, s25
+; GFX11-TRUE16-NEXT: s_mov_b32 s43, s46
; GFX11-TRUE16-NEXT: s_pack_ll_b32_b16 s9, s18, s9
; GFX11-TRUE16-NEXT: s_pack_ll_b32_b16 s8, s19, s8
; GFX11-TRUE16-NEXT: s_mov_b32 s18, s26
; GFX11-TRUE16-NEXT: s_mov_b32 s19, s27
+; GFX11-TRUE16-NEXT: s_pack_ll_b32_b16 s0, s0, s43
+; GFX11-TRUE16-NEXT: s_mov_b32 s43, s44
; GFX11-TRUE16-NEXT: s_pack_ll_b32_b16 s7, s18, s7
; GFX11-TRUE16-NEXT: s_pack_ll_b32_b16 s6, s19, s6
; GFX11-TRUE16-NEXT: s_mov_b32 s18, s28
; GFX11-TRUE16-NEXT: s_mov_b32 s19, s29
-; GFX11-TRUE16-NEXT: s_pack_ll_b32_b16 s0, s0, s45
-; GFX11-TRUE16-NEXT: s_pack_ll_b32_b16 s1, s1, s44
+; GFX11-TRUE16-NEXT: s_pack_ll_b32_b16 s1, s1, s45
; GFX11-TRUE16-NEXT: s_pack_ll_b32_b16 s2, s2, s43
; GFX11-TRUE16-NEXT: s_pack_ll_b32_b16 s3, s3, s42
; GFX11-TRUE16-NEXT: s_pack_ll_b32_b16 s16, s16, s41
@@ -20956,9 +20964,9 @@ define inreg <36 x half> @bitcast_v9i64_to_v36f16_scalar(<9 x i64> inreg %a, i32
; GFX11-TRUE16-NEXT: v_dual_mov_b32 v16, s5 :: v_dual_mov_b32 v17, s4
; GFX11-TRUE16-NEXT: s_setpc_b64 s[30:31]
; GFX11-TRUE16-NEXT: .LBB45_4:
+; GFX11-TRUE16-NEXT: ; implicit-def: $sgpr46_lo16
; GFX11-TRUE16-NEXT: ; implicit-def: $sgpr45_lo16
; GFX11-TRUE16-NEXT: ; implicit-def: $sgpr44_lo16
-; GFX11-TRUE16-NEXT: ; implicit-def: $sgpr43_lo16
; GFX11-TRUE16-NEXT: ; implicit-def: $sgpr42_lo16
; GFX11-TRUE16-NEXT: ; implicit-def: $sgpr41_lo16
; GFX11-TRUE16-NEXT: ; implicit-def: $sgpr40_lo16
diff --git a/llvm/test/CodeGen/AMDGPU/amdgcn.bitcast.640bit.ll b/llvm/test/CodeGen/AMDGPU/amdgcn.bitcast.640bit.ll
index cc55ba1d84df6..9d2b0df4280e7 100644
--- a/llvm/test/CodeGen/AMDGPU/amdgcn.bitcast.640bit.ll
+++ b/llvm/test/CodeGen/AMDGPU/amdgcn.bitcast.640bit.ll
@@ -3400,7 +3400,7 @@ define inreg <40 x i16> @bitcast_v20i32_to_v40i16_scalar(<20 x i32> inreg %a, i3
; GFX11-TRUE16-NEXT: v_cmp_ne_u32_e32 vcc_lo, 0, v2
; GFX11-TRUE16-NEXT: v_readfirstlane_b32 s5, v0
; GFX11-TRUE16-NEXT: v_readfirstlane_b32 s4, v1
-; GFX11-TRUE16-NEXT: s_mov_b32 s58, 0
+; GFX11-TRUE16-NEXT: s_mov_b32 s47, 0
; GFX11-TRUE16-NEXT: s_and_b32 s6, vcc_lo, exec_lo
; GFX11-TRUE16-NEXT: s_cbranch_scc0 .LBB13_4
; GFX11-TRUE16-NEXT: ; %bb.1: ; %cmp.false
@@ -3421,10 +3421,10 @@ define inreg <40 x i16> @bitcast_v20i32_to_v40i16_scalar(<20 x i32> inreg %a, i3
; GFX11-TRUE16-NEXT: s_lshr_b32 s44, s17, 16
; GFX11-TRUE16-NEXT: s_lshr_b32 s45, s16, 16
; GFX11-TRUE16-NEXT: s_lshr_b32 s46, s3, 16
-; GFX11-TRUE16-NEXT: s_lshr_b32 s47, s2, 16
-; GFX11-TRUE16-NEXT: s_lshr_b32 s56, s1, 16
-; GFX11-TRUE16-NEXT: s_lshr_b32 s57, s0, 16
-; GFX11-TRUE16-NEXT: s_and_not1_b32 vcc_lo, exec_lo, s58
+; GFX11-TRUE16-NEXT: s_lshr_b32 s56, s2, 16
+; GFX11-TRUE16-NEXT: s_lshr_b32 s57, s1, 16
+; GFX11-TRUE16-NEXT: s_lshr_b32 s58, s0, 16
+; GFX11-TRUE16-NEXT: s_and_not1_b32 vcc_lo, exec_lo, s47
; GFX11-TRUE16-NEXT: s_cbranch_vccnz .LBB13_3
; GFX11-TRUE16-NEXT: .LBB13_2: ; %cmp.true
; GFX11-TRUE16-NEXT: s_add_i32 s4, s4, 3
@@ -3464,24 +3464,26 @@ define inreg <40 x i16> @bitcast_v20i32_to_v40i16_scalar(<20 x i32> inreg %a, i3
; GFX11-TRUE16-NEXT: s_lshr_b32 s44, s17, 16
; GFX11-TRUE16-NEXT: s_lshr_b32 s45, s16, 16
; GFX11-TRUE16-NEXT: s_lshr_b32 s46, s3, 16
-; GFX11-TRUE16-NEXT: s_lshr_b32 s47, s2, 16
-; GFX11-TRUE16-NEXT: s_lshr_b32 s56, s1, 16
-; GFX11-TRUE16-NEXT: s_lshr_b32 s57, s0, 16
+; GFX11-TRUE16-NEXT: s_lshr_b32 s56, s2, 16
+; GFX11-TRUE16-NEXT: s_lshr_b32 s57, s1, 16
+; GFX11-TRUE16-NEXT: s_lshr_b32 s58, s0, 16
; GFX11-TRUE16-NEXT: .LBB13_3: ; %end
; GFX11-TRUE16-NEXT: s_pack_ll_b32_b16 s15, s22, s15
; GFX11-TRUE16-NEXT: s_pack_ll_b32_b16 s14, s23, s14
; GFX11-TRUE16-NEXT: s_mov_b32 s22, s24
; GFX11-TRUE16-NEXT: s_mov_b32 s23, s25
+; GFX11-TRUE16-NEXT: s_mov_b32 s47, s58
; GFX11-TRUE16-NEXT: s_pack_ll_b32_b16 s13, s22, s13
; GFX11-TRUE16-NEXT: s_pack_ll_b32_b16 s12, s23, s12
; GFX11-TRUE16-NEXT: s_mov_b32 s22, s26
; GFX11-TRUE16-NEXT: s_mov_b32 s23, s27
+; GFX11-TRUE16-NEXT: s_pack_ll_b32_b16 s0, s0, s47
+; GFX11-TRUE16-NEXT: s_mov_b32 s47, s56
; GFX11-TRUE16-NEXT: s_pack_ll_b32_b16 s11, s22, s11
; GFX11-TRUE16-NEXT: s_pack_ll_b32_b16 s10, s23, s10
; GFX11-TRUE16-NEXT: s_mov_b32 s22, s28
; GFX11-TRUE16-NEXT: s_mov_b32 s23, s29
-; GFX11-TRUE16-NEXT: s_pack_ll_b32_b16 s0, s0, s57
-; GFX11-TRUE16-NEXT: s_pack_ll_b32_b16 s1, s1, s56
+; GFX11-TRUE16-NEXT: s_pack_ll_b32_b16 s1, s1, s57
; GFX11-TRUE16-NEXT: s_pack_ll_b32_b16 s2, s2, s47
; GFX11-TRUE16-NEXT: s_pack_ll_b32_b16 s3, s3, s46
; GFX11-TRUE16-NEXT: s_pack_ll_b32_b16 s16, s16, s45
@@ -3506,9 +3508,9 @@ define inreg <40 x i16> @bitcast_v20i32_to_v40i16_scalar(<20 x i32> inreg %a, i3
; GFX11-TRUE16-NEXT: v_dual_mov_b32 v18, s5 :: v_dual_mov_b32 v19, s4
; GFX11-TRUE16-NEXT: s_setpc_b64 s[30:31]
; GFX11-TRUE16-NEXT: .LBB13_4:
+; GFX11-TRUE16-NEXT: ; implicit-def: $sgpr58_lo16
; GFX11-TRUE16-NEXT: ; implicit-def: $sgpr57_lo16
; GFX11-TRUE16-NEXT: ; implicit-def: $sgpr56_lo16
-; GFX11-TRUE16-NEXT: ; implicit-def: $sgpr47_lo16
; GFX11-TRUE16-NEXT: ; implicit-def: $sgpr46_lo16
; GFX11-TRUE16-NEXT: ; implicit-def: $sgpr45_lo16
; GFX11-TRUE16-NEXT: ; implicit-def: $sgpr44_lo16
@@ -6774,7 +6776,7 @@ define inreg <40 x half> @bitcast_v20i32_to_v40f16_scalar(<20 x i32> inreg %a, i
; GFX11-TRUE16-NEXT: v_cmp_ne_u32_e32 vcc_lo, 0, v2
; GFX11-TRUE16-NEXT: v_readfirstlane_b32 s5, v0
; GFX11-TRUE16-NEXT: v_readfirstlane_b32 s4, v1
-; GFX11-TRUE16-NEXT: s_mov_b32 s58, 0
+; GFX11-TRUE16-NEXT: s_mov_b32 s47, 0
; GFX11-TRUE16-NEXT: s_and_b32 s6, vcc_lo, exec_lo
; GFX11-TRUE16-NEXT: s_cbranch_scc0 .LBB17_4
; GFX11-TRUE16-NEXT: ; %bb.1: ; %cmp.false
@@ -6795,10 +6797,10 @@ define inreg <40 x half> @bitcast_v20i32_to_v40f16_scalar(<20 x i32> inreg %a, i
; GFX11-TRUE16-NEXT: s_lshr_b32 s44, s17, 16
; GFX11-TRUE16-NEXT: s_lshr_b32 s45, s16, 16
; GFX11-TRUE16-NEXT: s_lshr_b32 s46, s3, 16
-; GFX11-TRUE16-NEXT: s_lshr_b32 s47, s2, 16
-; GFX11-TRUE16-NEXT: s_lshr_b32 s56, s1, 16
-; GFX11-TRUE16-NEXT: s_lshr_b32 s57, s0, 16
-; GFX11-TRUE16-NEXT: s_and_not1_b32 vcc_lo, exec_lo, s58
+; GFX11-TRUE16-NEXT: s_lshr_b32 s56, s2, 16
+; GFX11-TRUE16-NEXT: s_lshr_b32 s57, s1, 16
+; GFX11-TRUE16-NEXT: s_lshr_b32 s58, s0, 16
+; GFX11-TRUE16-NEXT: s_and_not1_b32 vcc_lo, exec_lo, s47
; GFX11-TRUE16-NEXT: s_cbranch_vccnz .LBB17_3
; GFX11-TRUE16-NEXT: .LBB17_2: ; %cmp.true
; GFX11-TRUE16-NEXT: s_add_i32 s4, s4, 3
@@ -6838,24 +6840,26 @@ define inreg <40 x half> @bitcast_v20i32_to_v40f16_scalar(<20 x i32> inreg %a, i
; GFX11-TRUE16-NEXT: s_lshr_b32 s44, s17, 16
; GFX11-TRUE16-NEXT: s_lshr_b32 s45, s16, 16
; GFX11-TRUE16-NEXT: s_lshr_b32 s46, s3, 16
-; GFX11-TRUE16-NEXT: s_lshr_b32 s47, s2, 16
-; GFX11-TRUE16-NEXT: s_lshr_b32 s56, s1, 16
-; GFX11-TRUE16-NEXT: s_lshr_b32 s57, s0, 16
+; GFX11-TRUE16-NEXT: s_lshr_b32 s56, s2, 16
+; GFX11-TRUE16-NEXT: s_lshr_b32 s57, s1, 16
+; GFX11-TRUE16-NEXT: s_lshr_b32 s58, s0, 16
; GFX11-TRUE16-NEXT: .LBB17_3: ; %end
; GFX11-TRUE16-NEXT: s_pack_ll_b32_b16 s15, s22, s15
; GFX11-TRUE16-NEXT: s_pack_ll_b32_b16 s14, s23, s14
; GFX11-TRUE16-NEXT: s_mov_b32 s22, s24
; GFX11-TRUE16-NEXT: s_mov_b32 s23, s25
+; GFX11-TRUE16-NEXT: s_mov_b32 s47, s58
; GFX11-TRUE16-NEXT: s_pack_ll_b32_b16 s13, s22, s13
; GFX11-TRUE16-NEXT: s_pack_ll_b32_b16 s12, s23, s12
; GFX11-TRUE16-NEXT: s_mov_b32 s22, s26
; GFX11-TRUE16-NEXT: s_mov_b32 s23, s27
+; GFX11-TRUE16-NEXT: s_pack_ll_b32_b16 s0, s0, s47
+; GFX11-TRUE16-NEXT: s_mov_b32 s47, s56
; GFX11-TRUE16-NEXT: s_pack_ll_b32_b16 s11, s22, s11
; GFX11-TRUE16-NEXT: s_pack_ll_b32_b16 s10, s23, s10
; GFX11-TRUE16-NEXT: s_mov_b32 s22, s28
; GFX11-TRUE16-NEXT: s_mov_b32 s23, s29
-; GFX11-TRUE16-NEXT: s_pack_ll_b32_b16 s0, s0, s57
-; GFX11-TRUE16-NEXT: s_pack_ll_b32_b16 s1, s1, s56
+; GFX11-TRUE16-NEXT: s_pack_ll_b32_b16 s1, s1, s57
; GFX11-TRUE16-NEXT: s_pack_ll_b32_b16 s2, s2, s47
; GFX11-TRUE16-NEXT: s_pack_ll_b32_b16 s3, s3, s46
; GFX11-TRUE16-NEXT: s_pack_ll_b32_b16 s16, s16, s45
@@ -6880,9 +6884,9 @@ define inreg <40 x half> @bitcast_v20i32_to_v40f16_scalar(<20 x i32> inreg %a, i
; GFX11-TRUE16-NEXT: v_dual_mov_b32 v18, s5 :: v_dual_mov_b32 v19, s4
; GFX11-TRUE16-NEXT: s_setpc_b64 s[30:31]
; GFX11-TRUE16-NEXT: .LBB17_4:
+; GFX11-TRUE16-NEXT: ; implicit-def: $sgpr58_lo16
; GFX11-TRUE16-NEXT: ; implicit-def: $sgpr57_lo16
; GFX11-TRUE16-NEXT: ; implicit-def: $sgpr56_lo16
-; GFX11-TRUE16-NEXT: ; implicit-def: $sgpr47_lo16
; GFX11-TRUE16-NEXT: ; implicit-def: $sgpr46_lo16
; GFX11-TRUE16-NEXT: ; implicit-def: $sgpr45_lo16
; GFX11-TRUE16-NEXT: ; implicit-def: $sgpr44_lo16
@@ -8025,7 +8029,7 @@ define inreg <20 x i32> @bitcast_v40f16_to_v20i32_scalar(<40 x half> inreg %a, i
; SI-NEXT: v_cvt_f16_f32_e32 v62, v2
; SI-NEXT: v_cvt_f16_f32_e32 v44, v5
; SI-NEXT: v_cvt_f16_f32_e32 v43, v4
-; SI-NEXT: v_cvt_f16_f32_e32 v42, v7
+; SI-NEXT: v_cvt_f16_f32_e32 v56, v7
; SI-NEXT: v_cvt_f16_f32_e32 v41, v6
; SI-NEXT: v_cvt_f16_f32_e32 v40, v9
; SI-NEXT: v_cvt_f16_f32_e32 v55, v8
@@ -8045,26 +8049,26 @@ define inreg <20 x i32> @bitcast_v40f16_to_v20i32_scalar(<40 x half> inreg %a, i
; SI-NEXT: v_cvt_f16_f32_e32 v22, v22
; SI-NEXT: v_cvt_f16_f32_e32 v20, v25
; SI-NEXT: v_cvt_f16_f32_e32 v21, v24
-; SI-NEXT: v_cvt_f16_f32_e32 v58, s17
-; SI-NEXT: v_cvt_f16_f32_e32 v60, s16
-; SI-NEXT: v_cvt_f16_f32_e32 v59, s19
-; SI-NEXT: v_cvt_f16_f32_e32 v57, s18
+; SI-NEXT: v_cvt_f16_f32_e32 v61, s17
+; SI-NEXT: v_cvt_f16_f32_e32 v57, s16
+; SI-NEXT: v_cvt_f16_f32_e32 v60, s19
+; SI-NEXT: v_cvt_f16_f32_e32 v58, s18
; SI-NEXT: v_cvt_f16_f32_e32 v39, s21
-; SI-NEXT: v_cvt_f16_f32_e32 v56, s20
+; SI-NEXT: v_cvt_f16_f32_e32 v42, s20
; SI-NEXT: v_cvt_f16_f32_e32 v38, s23
; SI-NEXT: v_cvt_f16_f32_e32 v37, s22
; SI-NEXT: v_cvt_f16_f32_e32 v36, s25
; SI-NEXT: v_cvt_f16_f32_e32 v35, s24
; SI-NEXT: v_cvt_f16_f32_e32 v34, s27
-; SI-NEXT: v_cvt_f16_f32_e32 v61, s26
+; SI-NEXT: v_cvt_f16_f32_e32 v33, s26
; SI-NEXT: v_cvt_f16_f32_e32 v25, s29
; SI-NEXT: v_cvt_f16_f32_e32 v24, s28
; SI-NEXT: v_cmp_ne_u32_e32 vcc, 0, v26
; SI-NEXT: s_and_b64 s[4:5], vcc, exec
; SI-NEXT: s_cbranch_scc0 .LBB19_4
; SI-NEXT: ; %bb.1: ; %cmp.false
-; SI-NEXT: v_lshlrev_b32_e32 v0, 16, v58
-; SI-NEXT: v_lshlrev_b32_e32 v1, 16, v59
+; SI-NEXT: v_lshlrev_b32_e32 v0, 16, v61
+; SI-NEXT: v_lshlrev_b32_e32 v1, 16, v60
; SI-NEXT: v_lshlrev_b32_e32 v2, 16, v39
; SI-NEXT: v_lshlrev_b32_e32 v3, 16, v38
; SI-NEXT: v_lshlrev_b32_e32 v4, 16, v36
@@ -8073,7 +8077,7 @@ define inreg <20 x i32> @bitcast_v40f16_to_v20i32_scalar(<40 x half> inreg %a, i
; SI-NEXT: v_lshlrev_b32_e32 v7, 16, v32
; SI-NEXT: v_lshlrev_b32_e32 v8, 16, v63
; SI-NEXT: v_lshlrev_b32_e32 v9, 16, v44
-; SI-NEXT: v_lshlrev_b32_e32 v10, 16, v42
+; SI-NEXT: v_lshlrev_b32_e32 v10, 16, v56
; SI-NEXT: v_lshlrev_b32_e32 v11, 16, v40
; SI-NEXT: v_lshlrev_b32_e32 v12, 16, v54
; SI-NEXT: v_lshlrev_b32_e32 v13, 16, v52
@@ -8083,12 +8087,12 @@ define inreg <20 x i32> @bitcast_v40f16_to_v20i32_scalar(<40 x half> inreg %a, i
; SI-NEXT: v_lshlrev_b32_e32 v17, 16, v28
; SI-NEXT: v_lshlrev_b32_e32 v18, 16, v23
; SI-NEXT: v_lshlrev_b32_e32 v19, 16, v20
-; SI-NEXT: v_or_b32_e32 v0, v60, v0
-; SI-NEXT: v_or_b32_e32 v1, v57, v1
-; SI-NEXT: v_or_b32_e32 v2, v56, v2
+; SI-NEXT: v_or_b32_e32 v0, v57, v0
+; SI-NEXT: v_or_b32_e32 v1, v58, v1
+; SI-NEXT: v_or_b32_e32 v2, v42, v2
; SI-NEXT: v_or_b32_e32 v3, v37, v3
; SI-NEXT: v_or_b32_e32 v4, v35, v4
-; SI-NEXT: v_or_b32_e32 v5, v61, v5
+; SI-NEXT: v_or_b32_e32 v5, v33, v5
; SI-NEXT: v_or_b32_e32 v6, v24, v6
; SI-NEXT: v_or_b32_e32 v7, v47, v7
; SI-NEXT: v_or_b32_e32 v8, v62, v8
@@ -8105,11 +8109,12 @@ define inreg <20 x i32> @bitcast_v40f16_to_v20i32_scalar(<40 x half> inreg %a, i
; SI-NEXT: v_or_b32_e32 v19, v21, v19
; SI-NEXT: s_cbranch_execnz .LBB19_3
; SI-NEXT: .LBB19_2: ; %cmp.true
-; SI-NEXT: v_cvt_f32_f16_e32 v0, v58
-; SI-NEXT: s_waitcnt vmcnt(1)
-; SI-NEXT: v_cvt_f32_f16_e32 v2, v59
-; SI-NEXT: v_cvt_f32_f16_e32 v1, v60
-; SI-NEXT: v_cvt_f32_f16_e32 v3, v57
+; SI-NEXT: v_cvt_f32_f16_e32 v0, v61
+; SI-NEXT: s_waitcnt vmcnt(7)
+; SI-NEXT: v_cvt_f32_f16_e32 v2, v60
+; SI-NEXT: v_cvt_f32_f16_e32 v1, v57
+; SI-NEXT: s_waitcnt vmcnt(3)
+; SI-NEXT: v_cvt_f32_f16_e32 v3, v58
; SI-NEXT: v_add_f32_e32 v0, 0x38000000, v0
; SI-NEXT: v_add_f32_e32 v2, 0x38000000, v2
; SI-NEXT: v_add_f32_e32 v1, 0x38000000, v1
@@ -8123,7 +8128,8 @@ define inreg <20 x i32> @bitcast_v40f16_to_v20i32_scalar(<40 x half> inreg %a, i
; SI-NEXT: v_or_b32_e32 v0, v1, v0
; SI-NEXT: v_or_b32_e32 v1, v3, v2
; SI-NEXT: v_cvt_f32_f16_e32 v2, v39
-; SI-NEXT: v_cvt_f32_f16_e32 v3, v56
+; SI-NEXT: s_waitcnt vmcnt(1)
+; SI-NEXT: v_cvt_f32_f16_e32 v3, v42
; SI-NEXT: v_cvt_f32_f16_e32 v4, v37
; SI-NEXT: v_cvt_f32_f16_e32 v5, v35
; SI-NEXT: v_add_f32_e32 v2, 0x38000000, v2
@@ -8137,7 +8143,7 @@ define inreg <20 x i32> @bitcast_v40f16_to_v20i32_scalar(<40 x half> inreg %a, i
; SI-NEXT: v_cvt_f32_f16_e32 v3, v38
; SI-NEXT: v_add_f32_e32 v5, 0x38000000, v5
; SI-NEXT: v_cvt_f16_f32_e32 v5, v5
-; SI-NEXT: v_cvt_f32_f16_e32 v6, v61
+; SI-NEXT: v_cvt_f32_f16_e32 v6, v33
; SI-NEXT: v_add_f32_e32 v3, 0x38000000, v3
; SI-NEXT: v_cvt_f16_f32_e32 v3, v3
; SI-NEXT: v_cvt_f32_f16_e32 v7, v24
@@ -8164,7 +8170,7 @@ define inreg <20 x i32> @bitcast_v40f16_to_v20i32_scalar(<40 x half> inreg %a, i
; SI-NEXT: v_cvt_f16_f32_e32 v5, v5
; SI-NEXT: v_add_f32_e32 v10, 0x38000000, v10
; SI-NEXT: v_cvt_f16_f32_e32 v10, v10
-; SI-NEXT: v_cvt_f32_f16_e32 v11, v42
+; SI-NEXT: v_cvt_f32_f16_e32 v11, v56
; SI-NEXT: v_lshlrev_b32_e32 v5, 16, v5
; SI-NEXT: v_or_b32_e32 v5, v6, v5
; SI-NEXT: v_cvt_f32_f16_e32 v6, v25
@@ -8288,20 +8294,20 @@ define inreg <20 x i32> @bitcast_v40f16_to_v20i32_scalar(<40 x half> inreg %a, i
; SI-NEXT: s_setpc_b64 s[30:31]
; SI-NEXT: .LBB19_4:
; SI-NEXT: buffer_store_dword v41, off, s[0:3], s32 offset:64 ; 4-byte Folded Spill
-; SI-NEXT: buffer_store_dword v59, off, s[0:3], s32 offset:88 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v42, off, s[0:3], s32 offset:76 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_mov_b32_e32 v59, v48
+; SI-NEXT: v_mov_b32_e32 v42, v48
; SI-NEXT: v_mov_b32_e32 v48, v21
-; SI-NEXT: buffer_store_dword v42, off, s[0:3], s32 offset:68 ; 4-byte Folded Spill
-; SI-NEXT: buffer_store_dword v60, off, s[0:3], s32 offset:92 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v56, off, s[0:3], s32 offset:68 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v58, off, s[0:3], s32 offset:84 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_mov_b32_e32 v60, v49
+; SI-NEXT: v_mov_b32_e32 v58, v49
; SI-NEXT: v_mov_b32_e32 v49, v20
; SI-NEXT: buffer_store_dword v43, off, s[0:3], s32 offset:72 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_mov_b32_e32 v43, v50
; SI-NEXT: v_mov_b32_e32 v50, v22
-; SI-NEXT: buffer_store_dword v44, off, s[0:3], s32 offset:84 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v44, off, s[0:3], s32 offset:88 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_mov_b32_e32 v44, v51
; SI-NEXT: v_mov_b32_e32 v51, v23
@@ -8309,22 +8315,21 @@ define inreg <20 x i32> @bitcast_v40f16_to_v20i32_scalar(<40 x half> inreg %a, i
; SI-NEXT: v_mov_b32_e32 v52, v27
; SI-NEXT: v_mov_b32_e32 v46, v53
; SI-NEXT: v_mov_b32_e32 v53, v28
-; SI-NEXT: buffer_store_dword v36, off, s[0:3], s32 offset:76 ; 4-byte Folded Spill
-; SI-NEXT: buffer_store_dword v39, off, s[0:3], s32 offset:80 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v36, off, s[0:3], s32 offset:80 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v60, off, s[0:3], s32 offset:92 ; 4-byte Folded Spill
+; SI-NEXT: s_waitcnt expcnt(0)
+; SI-NEXT: v_mov_b32_e32 v60, v33
; SI-NEXT: v_mov_b32_e32 v41, v32
; SI-NEXT: v_mov_b32_e32 v33, v47
; SI-NEXT: v_mov_b32_e32 v47, v54
; SI-NEXT: v_mov_b32_e32 v54, v29
-; SI-NEXT: v_mov_b32_e32 v42, v56
; SI-NEXT: v_mov_b32_e32 v56, v55
; SI-NEXT: v_mov_b32_e32 v55, v30
-; SI-NEXT: s_waitcnt expcnt(1)
-; SI-NEXT: v_mov_b32_e32 v36, v57
+; SI-NEXT: v_mov_b32_e32 v59, v57
; SI-NEXT: v_mov_b32_e32 v57, v40
; SI-NEXT: v_mov_b32_e32 v40, v31
-; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_mov_b32_e32 v39, v58
-; SI-NEXT: v_mov_b32_e32 v58, v37
+; SI-NEXT: v_mov_b32_e32 v36, v39
+; SI-NEXT: v_mov_b32_e32 v39, v37
; SI-NEXT: v_mov_b32_e32 v37, v34
; SI-NEXT: v_mov_b32_e32 v34, v24
; SI-NEXT: v_mov_b32_e32 v32, v38
@@ -8336,34 +8341,34 @@ define inreg <20 x i32> @bitcast_v40f16_to_v20i32_scalar(<40 x half> inreg %a, i
; SI-NEXT: v_mov_b32_e32 v38, v32
; SI-NEXT: v_mov_b32_e32 v24, v34
; SI-NEXT: v_mov_b32_e32 v34, v37
-; SI-NEXT: v_mov_b32_e32 v37, v58
-; SI-NEXT: v_mov_b32_e32 v58, v39
-; SI-NEXT: v_mov_b32_e32 v31, v40
-; SI-NEXT: v_mov_b32_e32 v40, v57
-; SI-NEXT: v_mov_b32_e32 v57, v36
+; SI-NEXT: v_mov_b32_e32 v37, v39
+; SI-NEXT: v_mov_b32_e32 v39, v36
; SI-NEXT: v_mov_b32_e32 v30, v55
; SI-NEXT: v_mov_b32_e32 v55, v56
-; SI-NEXT: v_mov_b32_e32 v56, v42
+; SI-NEXT: v_mov_b32_e32 v29, v54
+; SI-NEXT: v_mov_b32_e32 v54, v47
+; SI-NEXT: v_mov_b32_e32 v47, v33
; SI-NEXT: v_mov_b32_e32 v32, v41
-; SI-NEXT: buffer_load_dword v39, off, s[0:3], s32 offset:80 ; 4-byte Folded Reload
-; SI-NEXT: buffer_load_dword v36, off, s[0:3], s32 offset:76 ; 4-byte Folded Reload
+; SI-NEXT: v_mov_b32_e32 v33, v60
+; SI-NEXT: buffer_load_dword v60, off, s[0:3], s32 offset:92 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v36, off, s[0:3], s32 offset:80 ; 4-byte Folded Reload
; SI-NEXT: v_mov_b32_e32 v23, v51
; SI-NEXT: v_mov_b32_e32 v51, v44
-; SI-NEXT: buffer_load_dword v44, off, s[0:3], s32 offset:84 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v44, off, s[0:3], s32 offset:88 ; 4-byte Folded Reload
; SI-NEXT: v_mov_b32_e32 v22, v50
; SI-NEXT: v_mov_b32_e32 v50, v43
; SI-NEXT: buffer_load_dword v43, off, s[0:3], s32 offset:72 ; 4-byte Folded Reload
; SI-NEXT: v_mov_b32_e32 v20, v49
-; SI-NEXT: v_mov_b32_e32 v49, v60
-; SI-NEXT: buffer_load_dword v60, off, s[0:3], s32 offset:92 ; 4-byte Folded Reload
-; SI-NEXT: buffer_load_dword v42, off, s[0:3], s32 offset:68 ; 4-byte Folded Reload
+; SI-NEXT: v_mov_b32_e32 v49, v58
+; SI-NEXT: buffer_load_dword v58, off, s[0:3], s32 offset:84 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v56, off, s[0:3], s32 offset:68 ; 4-byte Folded Reload
; SI-NEXT: v_mov_b32_e32 v21, v48
-; SI-NEXT: v_mov_b32_e32 v48, v59
-; SI-NEXT: buffer_load_dword v59, off, s[0:3], s32 offset:88 ; 4-byte Folded Reload
+; SI-NEXT: v_mov_b32_e32 v48, v42
+; SI-NEXT: buffer_load_dword v42, off, s[0:3], s32 offset:76 ; 4-byte Folded Reload
; SI-NEXT: buffer_load_dword v41, off, s[0:3], s32 offset:64 ; 4-byte Folded Reload
-; SI-NEXT: v_mov_b32_e32 v29, v54
-; SI-NEXT: v_mov_b32_e32 v54, v47
-; SI-NEXT: v_mov_b32_e32 v47, v33
+; SI-NEXT: v_mov_b32_e32 v31, v40
+; SI-NEXT: v_mov_b32_e32 v40, v57
+; SI-NEXT: v_mov_b32_e32 v57, v59
; SI-NEXT: v_mov_b32_e32 v28, v53
; SI-NEXT: v_mov_b32_e32 v53, v46
; SI-NEXT: v_mov_b32_e32 v27, v52
@@ -16112,7 +16117,7 @@ define inreg <20 x float> @bitcast_v40f16_to_v20f32_scalar(<40 x half> inreg %a,
; SI-NEXT: v_cvt_f16_f32_e32 v62, v2
; SI-NEXT: v_cvt_f16_f32_e32 v44, v5
; SI-NEXT: v_cvt_f16_f32_e32 v43, v4
-; SI-NEXT: v_cvt_f16_f32_e32 v42, v7
+; SI-NEXT: v_cvt_f16_f32_e32 v56, v7
; SI-NEXT: v_cvt_f16_f32_e32 v41, v6
; SI-NEXT: v_cvt_f16_f32_e32 v40, v9
; SI-NEXT: v_cvt_f16_f32_e32 v55, v8
@@ -16132,26 +16137,26 @@ define inreg <20 x float> @bitcast_v40f16_to_v20f32_scalar(<40 x half> inreg %a,
; SI-NEXT: v_cvt_f16_f32_e32 v22, v22
; SI-NEXT: v_cvt_f16_f32_e32 v20, v25
; SI-NEXT: v_cvt_f16_f32_e32 v21, v24
-; SI-NEXT: v_cvt_f16_f32_e32 v58, s17
-; SI-NEXT: v_cvt_f16_f32_e32 v60, s16
-; SI-NEXT: v_cvt_f16_f32_e32 v59, s19
-; SI-NEXT: v_cvt_f16_f32_e32 v57, s18
+; SI-NEXT: v_cvt_f16_f32_e32 v61, s17
+; SI-NEXT: v_cvt_f16_f32_e32 v57, s16
+; SI-NEXT: v_cvt_f16_f32_e32 v60, s19
+; SI-NEXT: v_cvt_f16_f32_e32 v58, s18
; SI-NEXT: v_cvt_f16_f32_e32 v39, s21
-; SI-NEXT: v_cvt_f16_f32_e32 v56, s20
+; SI-NEXT: v_cvt_f16_f32_e32 v42, s20
; SI-NEXT: v_cvt_f16_f32_e32 v38, s23
; SI-NEXT: v_cvt_f16_f32_e32 v37, s22
; SI-NEXT: v_cvt_f16_f32_e32 v36, s25
; SI-NEXT: v_cvt_f16_f32_e32 v35, s24
; SI-NEXT: v_cvt_f16_f32_e32 v34, s27
-; SI-NEXT: v_cvt_f16_f32_e32 v61, s26
+; SI-NEXT: v_cvt_f16_f32_e32 v33, s26
; SI-NEXT: v_cvt_f16_f32_e32 v25, s29
; SI-NEXT: v_cvt_f16_f32_e32 v24, s28
; SI-NEXT: v_cmp_ne_u32_e32 vcc, 0, v26
; SI-NEXT: s_and_b64 s[4:5], vcc, exec
; SI-NEXT: s_cbranch_scc0 .LBB35_4
; SI-NEXT: ; %bb.1: ; %cmp.false
-; SI-NEXT: v_lshlrev_b32_e32 v0, 16, v58
-; SI-NEXT: v_lshlrev_b32_e32 v1, 16, v59
+; SI-NEXT: v_lshlrev_b32_e32 v0, 16, v61
+; SI-NEXT: v_lshlrev_b32_e32 v1, 16, v60
; SI-NEXT: v_lshlrev_b32_e32 v2, 16, v39
; SI-NEXT: v_lshlrev_b32_e32 v3, 16, v38
; SI-NEXT: v_lshlrev_b32_e32 v4, 16, v36
@@ -16160,7 +16165,7 @@ define inreg <20 x float> @bitcast_v40f16_to_v20f32_scalar(<40 x half> inreg %a,
; SI-NEXT: v_lshlrev_b32_e32 v7, 16, v32
; SI-NEXT: v_lshlrev_b32_e32 v8, 16, v63
; SI-NEXT: v_lshlrev_b32_e32 v9, 16, v44
-; SI-NEXT: v_lshlrev_b32_e32 v10, 16, v42
+; SI-NEXT: v_lshlrev_b32_e32 v10, 16, v56
; SI-NEXT: v_lshlrev_b32_e32 v11, 16, v40
; SI-NEXT: v_lshlrev_b32_e32 v12, 16, v54
; SI-NEXT: v_lshlrev_b32_e32 v13, 16, v52
@@ -16170,12 +16175,12 @@ define inreg <20 x float> @bitcast_v40f16_to_v20f32_scalar(<40 x half> inreg %a,
; SI-NEXT: v_lshlrev_b32_e32 v17, 16, v28
; SI-NEXT: v_lshlrev_b32_e32 v18, 16, v23
; SI-NEXT: v_lshlrev_b32_e32 v19, 16, v20
-; SI-NEXT: v_or_b32_e32 v0, v60, v0
-; SI-NEXT: v_or_b32_e32 v1, v57, v1
-; SI-NEXT: v_or_b32_e32 v2, v56, v2
+; SI-NEXT: v_or_b32_e32 v0, v57, v0
+; SI-NEXT: v_or_b32_e32 v1, v58, v1
+; SI-NEXT: v_or_b32_e32 v2, v42, v2
; SI-NEXT: v_or_b32_e32 v3, v37, v3
; SI-NEXT: v_or_b32_e32 v4, v35, v4
-; SI-NEXT: v_or_b32_e32 v5, v61, v5
+; SI-NEXT: v_or_b32_e32 v5, v33, v5
; SI-NEXT: v_or_b32_e32 v6, v24, v6
; SI-NEXT: v_or_b32_e32 v7, v47, v7
; SI-NEXT: v_or_b32_e32 v8, v62, v8
@@ -16192,11 +16197,12 @@ define inreg <20 x float> @bitcast_v40f16_to_v20f32_scalar(<40 x half> inreg %a,
; SI-NEXT: v_or_b32_e32 v19, v21, v19
; SI-NEXT: s_cbranch_execnz .LBB35_3
; SI-NEXT: .LBB35_2: ; %cmp.true
-; SI-NEXT: v_cvt_f32_f16_e32 v0, v58
-; SI-NEXT: s_waitcnt vmcnt(1)
-; SI-NEXT: v_cvt_f32_f16_e32 v2, v59
-; SI-NEXT: v_cvt_f32_f16_e32 v1, v60
-; SI-NEXT: v_cvt_f32_f16_e32 v3, v57
+; SI-NEXT: v_cvt_f32_f16_e32 v0, v61
+; SI-NEXT: s_waitcnt vmcnt(7)
+; SI-NEXT: v_cvt_f32_f16_e32 v2, v60
+; SI-NEXT: v_cvt_f32_f16_e32 v1, v57
+; SI-NEXT: s_waitcnt vmcnt(3)
+; SI-NEXT: v_cvt_f32_f16_e32 v3, v58
; SI-NEXT: v_add_f32_e32 v0, 0x38000000, v0
; SI-NEXT: v_add_f32_e32 v2, 0x38000000, v2
; SI-NEXT: v_add_f32_e32 v1, 0x38000000, v1
@@ -16210,7 +16216,8 @@ define inreg <20 x float> @bitcast_v40f16_to_v20f32_scalar(<40 x half> inreg %a,
; SI-NEXT: v_or_b32_e32 v0, v1, v0
; SI-NEXT: v_or_b32_e32 v1, v3, v2
; SI-NEXT: v_cvt_f32_f16_e32 v2, v39
-; SI-NEXT: v_cvt_f32_f16_e32 v3, v56
+; SI-NEXT: s_waitcnt vmcnt(1)
+; SI-NEXT: v_cvt_f32_f16_e32 v3, v42
; SI-NEXT: v_cvt_f32_f16_e32 v4, v37
; SI-NEXT: v_cvt_f32_f16_e32 v5, v35
; SI-NEXT: v_add_f32_e32 v2, 0x38000000, v2
@@ -16224,7 +16231,7 @@ define inreg <20 x float> @bitcast_v40f16_to_v20f32_scalar(<40 x half> inreg %a,
; SI-NEXT: v_cvt_f32_f16_e32 v3, v38
; SI-NEXT: v_add_f32_e32 v5, 0x38000000, v5
; SI-NEXT: v_cvt_f16_f32_e32 v5, v5
-; SI-NEXT: v_cvt_f32_f16_e32 v6, v61
+; SI-NEXT: v_cvt_f32_f16_e32 v6, v33
; SI-NEXT: v_add_f32_e32 v3, 0x38000000, v3
; SI-NEXT: v_cvt_f16_f32_e32 v3, v3
; SI-NEXT: v_cvt_f32_f16_e32 v7, v24
@@ -16251,7 +16258,7 @@ define inreg <20 x float> @bitcast_v40f16_to_v20f32_scalar(<40 x half> inreg %a,
; SI-NEXT: v_cvt_f16_f32_e32 v5, v5
; SI-NEXT: v_add_f32_e32 v10, 0x38000000, v10
; SI-NEXT: v_cvt_f16_f32_e32 v10, v10
-; SI-NEXT: v_cvt_f32_f16_e32 v11, v42
+; SI-NEXT: v_cvt_f32_f16_e32 v11, v56
; SI-NEXT: v_lshlrev_b32_e32 v5, 16, v5
; SI-NEXT: v_or_b32_e32 v5, v6, v5
; SI-NEXT: v_cvt_f32_f16_e32 v6, v25
@@ -16375,20 +16382,20 @@ define inreg <20 x float> @bitcast_v40f16_to_v20f32_scalar(<40 x half> inreg %a,
; SI-NEXT: s_setpc_b64 s[30:31]
; SI-NEXT: .LBB35_4:
; SI-NEXT: buffer_store_dword v41, off, s[0:3], s32 offset:64 ; 4-byte Folded Spill
-; SI-NEXT: buffer_store_dword v59, off, s[0:3], s32 offset:88 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v42, off, s[0:3], s32 offset:76 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_mov_b32_e32 v59, v48
+; SI-NEXT: v_mov_b32_e32 v42, v48
; SI-NEXT: v_mov_b32_e32 v48, v21
-; SI-NEXT: buffer_store_dword v42, off, s[0:3], s32 offset:68 ; 4-byte Folded Spill
-; SI-NEXT: buffer_store_dword v60, off, s[0:3], s32 offset:92 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v56, off, s[0:3], s32 offset:68 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v58, off, s[0:3], s32 offset:84 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_mov_b32_e32 v60, v49
+; SI-NEXT: v_mov_b32_e32 v58, v49
; SI-NEXT: v_mov_b32_e32 v49, v20
; SI-NEXT: buffer_store_dword v43, off, s[0:3], s32 offset:72 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_mov_b32_e32 v43, v50
; SI-NEXT: v_mov_b32_e32 v50, v22
-; SI-NEXT: buffer_store_dword v44, off, s[0:3], s32 offset:84 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v44, off, s[0:3], s32 offset:88 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_mov_b32_e32 v44, v51
; SI-NEXT: v_mov_b32_e32 v51, v23
@@ -16396,22 +16403,21 @@ define inreg <20 x float> @bitcast_v40f16_to_v20f32_scalar(<40 x half> inreg %a,
; SI-NEXT: v_mov_b32_e32 v52, v27
; SI-NEXT: v_mov_b32_e32 v46, v53
; SI-NEXT: v_mov_b32_e32 v53, v28
-; SI-NEXT: buffer_store_dword v36, off, s[0:3], s32 offset:76 ; 4-byte Folded Spill
-; SI-NEXT: buffer_store_dword v39, off, s[0:3], s32 offset:80 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v36, off, s[0:3], s32 offset:80 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v60, off, s[0:3], s32 offset:92 ; 4-byte Folded Spill
+; SI-NEXT: s_waitcnt expcnt(0)
+; SI-NEXT: v_mov_b32_e32 v60, v33
; SI-NEXT: v_mov_b32_e32 v41, v32
; SI-NEXT: v_mov_b32_e32 v33, v47
; SI-NEXT: v_mov_b32_e32 v47, v54
; SI-NEXT: v_mov_b32_e32 v54, v29
-; SI-NEXT: v_mov_b32_e32 v42, v56
; SI-NEXT: v_mov_b32_e32 v56, v55
; SI-NEXT: v_mov_b32_e32 v55, v30
-; SI-NEXT: s_waitcnt expcnt(1)
-; SI-NEXT: v_mov_b32_e32 v36, v57
+; SI-NEXT: v_mov_b32_e32 v59, v57
; SI-NEXT: v_mov_b32_e32 v57, v40
; SI-NEXT: v_mov_b32_e32 v40, v31
-; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_mov_b32_e32 v39, v58
-; SI-NEXT: v_mov_b32_e32 v58, v37
+; SI-NEXT: v_mov_b32_e32 v36, v39
+; SI-NEXT: v_mov_b32_e32 v39, v37
; SI-NEXT: v_mov_b32_e32 v37, v34
; SI-NEXT: v_mov_b32_e32 v34, v24
; SI-NEXT: v_mov_b32_e32 v32, v38
@@ -16423,34 +16429,34 @@ define inreg <20 x float> @bitcast_v40f16_to_v20f32_scalar(<40 x half> inreg %a,
; SI-NEXT: v_mov_b32_e32 v38, v32
; SI-NEXT: v_mov_b32_e32 v24, v34
; SI-NEXT: v_mov_b32_e32 v34, v37
-; SI-NEXT: v_mov_b32_e32 v37, v58
-; SI-NEXT: v_mov_b32_e32 v58, v39
-; SI-NEXT: v_mov_b32_e32 v31, v40
-; SI-NEXT: v_mov_b32_e32 v40, v57
-; SI-NEXT: v_mov_b32_e32 v57, v36
+; SI-NEXT: v_mov_b32_e32 v37, v39
+; SI-NEXT: v_mov_b32_e32 v39, v36
; SI-NEXT: v_mov_b32_e32 v30, v55
; SI-NEXT: v_mov_b32_e32 v55, v56
-; SI-NEXT: v_mov_b32_e32 v56, v42
+; SI-NEXT: v_mov_b32_e32 v29, v54
+; SI-NEXT: v_mov_b32_e32 v54, v47
+; SI-NEXT: v_mov_b32_e32 v47, v33
; SI-NEXT: v_mov_b32_e32 v32, v41
-; SI-NEXT: buffer_load_dword v39, off, s[0:3], s32 offset:80 ; 4-byte Folded Reload
-; SI-NEXT: buffer_load_dword v36, off, s[0:3], s32 offset:76 ; 4-byte Folded Reload
+; SI-NEXT: v_mov_b32_e32 v33, v60
+; SI-NEXT: buffer_load_dword v60, off, s[0:3], s32 offset:92 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v36, off, s[0:3], s32 offset:80 ; 4-byte Folded Reload
; SI-NEXT: v_mov_b32_e32 v23, v51
; SI-NEXT: v_mov_b32_e32 v51, v44
-; SI-NEXT: buffer_load_dword v44, off, s[0:3], s32 offset:84 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v44, off, s[0:3], s32 offset:88 ; 4-byte Folded Reload
; SI-NEXT: v_mov_b32_e32 v22, v50
; SI-NEXT: v_mov_b32_e32 v50, v43
; SI-NEXT: buffer_load_dword v43, off, s[0:3], s32 offset:72 ; 4-byte Folded Reload
; SI-NEXT: v_mov_b32_e32 v20, v49
-; SI-NEXT: v_mov_b32_e32 v49, v60
-; SI-NEXT: buffer_load_dword v60, off, s[0:3], s32 offset:92 ; 4-byte Folded Reload
-; SI-NEXT: buffer_load_dword v42, off, s[0:3], s32 offset:68 ; 4-byte Folded Reload
+; SI-NEXT: v_mov_b32_e32 v49, v58
+; SI-NEXT: buffer_load_dword v58, off, s[0:3], s32 offset:84 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v56, off, s[0:3], s32 offset:68 ; 4-byte Folded Reload
; SI-NEXT: v_mov_b32_e32 v21, v48
-; SI-NEXT: v_mov_b32_e32 v48, v59
-; SI-NEXT: buffer_load_dword v59, off, s[0:3], s32 offset:88 ; 4-byte Folded Reload
+; SI-NEXT: v_mov_b32_e32 v48, v42
+; SI-NEXT: buffer_load_dword v42, off, s[0:3], s32 offset:76 ; 4-byte Folded Reload
; SI-NEXT: buffer_load_dword v41, off, s[0:3], s32 offset:64 ; 4-byte Folded Reload
-; SI-NEXT: v_mov_b32_e32 v29, v54
-; SI-NEXT: v_mov_b32_e32 v54, v47
-; SI-NEXT: v_mov_b32_e32 v47, v33
+; SI-NEXT: v_mov_b32_e32 v31, v40
+; SI-NEXT: v_mov_b32_e32 v40, v57
+; SI-NEXT: v_mov_b32_e32 v57, v59
; SI-NEXT: v_mov_b32_e32 v28, v53
; SI-NEXT: v_mov_b32_e32 v53, v46
; SI-NEXT: v_mov_b32_e32 v27, v52
@@ -18910,7 +18916,7 @@ define inreg <40 x i16> @bitcast_v10i64_to_v40i16_scalar(<10 x i64> inreg %a, i3
; GFX11-TRUE16-NEXT: v_cmp_ne_u32_e32 vcc_lo, 0, v2
; GFX11-TRUE16-NEXT: v_readfirstlane_b32 s5, v0
; GFX11-TRUE16-NEXT: v_readfirstlane_b32 s4, v1
-; GFX11-TRUE16-NEXT: s_mov_b32 s58, 0
+; GFX11-TRUE16-NEXT: s_mov_b32 s47, 0
; GFX11-TRUE16-NEXT: s_and_b32 s6, vcc_lo, exec_lo
; GFX11-TRUE16-NEXT: s_cbranch_scc0 .LBB41_4
; GFX11-TRUE16-NEXT: ; %bb.1: ; %cmp.false
@@ -18931,10 +18937,10 @@ define inreg <40 x i16> @bitcast_v10i64_to_v40i16_scalar(<10 x i64> inreg %a, i3
; GFX11-TRUE16-NEXT: s_lshr_b32 s44, s17, 16
; GFX11-TRUE16-NEXT: s_lshr_b32 s45, s16, 16
; GFX11-TRUE16-NEXT: s_lshr_b32 s46, s3, 16
-; GFX11-TRUE16-NEXT: s_lshr_b32 s47, s2, 16
-; GFX11-TRUE16-NEXT: s_lshr_b32 s56, s1, 16
-; GFX11-TRUE16-NEXT: s_lshr_b32 s57, s0, 16
-; GFX11-TRUE16-NEXT: s_and_not1_b32 vcc_lo, exec_lo, s58
+; GFX11-TRUE16-NEXT: s_lshr_b32 s56, s2, 16
+; GFX11-TRUE16-NEXT: s_lshr_b32 s57, s1, 16
+; GFX11-TRUE16-NEXT: s_lshr_b32 s58, s0, 16
+; GFX11-TRUE16-NEXT: s_and_not1_b32 vcc_lo, exec_lo, s47
; GFX11-TRUE16-NEXT: s_cbranch_vccnz .LBB41_3
; GFX11-TRUE16-NEXT: .LBB41_2: ; %cmp.true
; GFX11-TRUE16-NEXT: s_add_u32 s5, s5, 3
@@ -18974,24 +18980,26 @@ define inreg <40 x i16> @bitcast_v10i64_to_v40i16_scalar(<10 x i64> inreg %a, i3
; GFX11-TRUE16-NEXT: s_lshr_b32 s44, s17, 16
; GFX11-TRUE16-NEXT: s_lshr_b32 s45, s16, 16
; GFX11-TRUE16-NEXT: s_lshr_b32 s46, s3, 16
-; GFX11-TRUE16-NEXT: s_lshr_b32 s47, s2, 16
-; GFX11-TRUE16-NEXT: s_lshr_b32 s56, s1, 16
-; GFX11-TRUE16-NEXT: s_lshr_b32 s57, s0, 16
+; GFX11-TRUE16-NEXT: s_lshr_b32 s56, s2, 16
+; GFX11-TRUE16-NEXT: s_lshr_b32 s57, s1, 16
+; GFX11-TRUE16-NEXT: s_lshr_b32 s58, s0, 16
; GFX11-TRUE16-NEXT: .LBB41_3: ; %end
; GFX11-TRUE16-NEXT: s_pack_ll_b32_b16 s15, s22, s15
; GFX11-TRUE16-NEXT: s_pack_ll_b32_b16 s14, s23, s14
; GFX11-TRUE16-NEXT: s_mov_b32 s22, s24
; GFX11-TRUE16-NEXT: s_mov_b32 s23, s25
+; GFX11-TRUE16-NEXT: s_mov_b32 s47, s58
; GFX11-TRUE16-NEXT: s_pack_ll_b32_b16 s13, s22, s13
; GFX11-TRUE16-NEXT: s_pack_ll_b32_b16 s12, s23, s12
; GFX11-TRUE16-NEXT: s_mov_b32 s22, s26
; GFX11-TRUE16-NEXT: s_mov_b32 s23, s27
+; GFX11-TRUE16-NEXT: s_pack_ll_b32_b16 s0, s0, s47
+; GFX11-TRUE16-NEXT: s_mov_b32 s47, s56
; GFX11-TRUE16-NEXT: s_pack_ll_b32_b16 s11, s22, s11
; GFX11-TRUE16-NEXT: s_pack_ll_b32_b16 s10, s23, s10
; GFX11-TRUE16-NEXT: s_mov_b32 s22, s28
; GFX11-TRUE16-NEXT: s_mov_b32 s23, s29
-; GFX11-TRUE16-NEXT: s_pack_ll_b32_b16 s0, s0, s57
-; GFX11-TRUE16-NEXT: s_pack_ll_b32_b16 s1, s1, s56
+; GFX11-TRUE16-NEXT: s_pack_ll_b32_b16 s1, s1, s57
; GFX11-TRUE16-NEXT: s_pack_ll_b32_b16 s2, s2, s47
; GFX11-TRUE16-NEXT: s_pack_ll_b32_b16 s3, s3, s46
; GFX11-TRUE16-NEXT: s_pack_ll_b32_b16 s16, s16, s45
@@ -19016,9 +19024,9 @@ define inreg <40 x i16> @bitcast_v10i64_to_v40i16_scalar(<10 x i64> inreg %a, i3
; GFX11-TRUE16-NEXT: v_dual_mov_b32 v18, s5 :: v_dual_mov_b32 v19, s4
; GFX11-TRUE16-NEXT: s_setpc_b64 s[30:31]
; GFX11-TRUE16-NEXT: .LBB41_4:
+; GFX11-TRUE16-NEXT: ; implicit-def: $sgpr58_lo16
; GFX11-TRUE16-NEXT: ; implicit-def: $sgpr57_lo16
; GFX11-TRUE16-NEXT: ; implicit-def: $sgpr56_lo16
-; GFX11-TRUE16-NEXT: ; implicit-def: $sgpr47_lo16
; GFX11-TRUE16-NEXT: ; implicit-def: $sgpr46_lo16
; GFX11-TRUE16-NEXT: ; implicit-def: $sgpr45_lo16
; GFX11-TRUE16-NEXT: ; implicit-def: $sgpr44_lo16
@@ -22294,7 +22302,7 @@ define inreg <40 x half> @bitcast_v10i64_to_v40f16_scalar(<10 x i64> inreg %a, i
; GFX11-TRUE16-NEXT: v_cmp_ne_u32_e32 vcc_lo, 0, v2
; GFX11-TRUE16-NEXT: v_readfirstlane_b32 s5, v0
; GFX11-TRUE16-NEXT: v_readfirstlane_b32 s4, v1
-; GFX11-TRUE16-NEXT: s_mov_b32 s58, 0
+; GFX11-TRUE16-NEXT: s_mov_b32 s47, 0
; GFX11-TRUE16-NEXT: s_and_b32 s6, vcc_lo, exec_lo
; GFX11-TRUE16-NEXT: s_cbranch_scc0 .LBB45_4
; GFX11-TRUE16-NEXT: ; %bb.1: ; %cmp.false
@@ -22315,10 +22323,10 @@ define inreg <40 x half> @bitcast_v10i64_to_v40f16_scalar(<10 x i64> inreg %a, i
; GFX11-TRUE16-NEXT: s_lshr_b32 s44, s17, 16
; GFX11-TRUE16-NEXT: s_lshr_b32 s45, s16, 16
; GFX11-TRUE16-NEXT: s_lshr_b32 s46, s3, 16
-; GFX11-TRUE16-NEXT: s_lshr_b32 s47, s2, 16
-; GFX11-TRUE16-NEXT: s_lshr_b32 s56, s1, 16
-; GFX11-TRUE16-NEXT: s_lshr_b32 s57, s0, 16
-; GFX11-TRUE16-NEXT: s_and_not1_b32 vcc_lo, exec_lo, s58
+; GFX11-TRUE16-NEXT: s_lshr_b32 s56, s2, 16
+; GFX11-TRUE16-NEXT: s_lshr_b32 s57, s1, 16
+; GFX11-TRUE16-NEXT: s_lshr_b32 s58, s0, 16
+; GFX11-TRUE16-NEXT: s_and_not1_b32 vcc_lo, exec_lo, s47
; GFX11-TRUE16-NEXT: s_cbranch_vccnz .LBB45_3
; GFX11-TRUE16-NEXT: .LBB45_2: ; %cmp.true
; GFX11-TRUE16-NEXT: s_add_u32 s5, s5, 3
@@ -22358,24 +22366,26 @@ define inreg <40 x half> @bitcast_v10i64_to_v40f16_scalar(<10 x i64> inreg %a, i
; GFX11-TRUE16-NEXT: s_lshr_b32 s44, s17, 16
; GFX11-TRUE16-NEXT: s_lshr_b32 s45, s16, 16
; GFX11-TRUE16-NEXT: s_lshr_b32 s46, s3, 16
-; GFX11-TRUE16-NEXT: s_lshr_b32 s47, s2, 16
-; GFX11-TRUE16-NEXT: s_lshr_b32 s56, s1, 16
-; GFX11-TRUE16-NEXT: s_lshr_b32 s57, s0, 16
+; GFX11-TRUE16-NEXT: s_lshr_b32 s56, s2, 16
+; GFX11-TRUE16-NEXT: s_lshr_b32 s57, s1, 16
+; GFX11-TRUE16-NEXT: s_lshr_b32 s58, s0, 16
; GFX11-TRUE16-NEXT: .LBB45_3: ; %end
; GFX11-TRUE16-NEXT: s_pack_ll_b32_b16 s15, s22, s15
; GFX11-TRUE16-NEXT: s_pack_ll_b32_b16 s14, s23, s14
; GFX11-TRUE16-NEXT: s_mov_b32 s22, s24
; GFX11-TRUE16-NEXT: s_mov_b32 s23, s25
+; GFX11-TRUE16-NEXT: s_mov_b32 s47, s58
; GFX11-TRUE16-NEXT: s_pack_ll_b32_b16 s13, s22, s13
; GFX11-TRUE16-NEXT: s_pack_ll_b32_b16 s12, s23, s12
; GFX11-TRUE16-NEXT: s_mov_b32 s22, s26
; GFX11-TRUE16-NEXT: s_mov_b32 s23, s27
+; GFX11-TRUE16-NEXT: s_pack_ll_b32_b16 s0, s0, s47
+; GFX11-TRUE16-NEXT: s_mov_b32 s47, s56
; GFX11-TRUE16-NEXT: s_pack_ll_b32_b16 s11, s22, s11
; GFX11-TRUE16-NEXT: s_pack_ll_b32_b16 s10, s23, s10
; GFX11-TRUE16-NEXT: s_mov_b32 s22, s28
; GFX11-TRUE16-NEXT: s_mov_b32 s23, s29
-; GFX11-TRUE16-NEXT: s_pack_ll_b32_b16 s0, s0, s57
-; GFX11-TRUE16-NEXT: s_pack_ll_b32_b16 s1, s1, s56
+; GFX11-TRUE16-NEXT: s_pack_ll_b32_b16 s1, s1, s57
; GFX11-TRUE16-NEXT: s_pack_ll_b32_b16 s2, s2, s47
; GFX11-TRUE16-NEXT: s_pack_ll_b32_b16 s3, s3, s46
; GFX11-TRUE16-NEXT: s_pack_ll_b32_b16 s16, s16, s45
@@ -22400,9 +22410,9 @@ define inreg <40 x half> @bitcast_v10i64_to_v40f16_scalar(<10 x i64> inreg %a, i
; GFX11-TRUE16-NEXT: v_dual_mov_b32 v18, s5 :: v_dual_mov_b32 v19, s4
; GFX11-TRUE16-NEXT: s_setpc_b64 s[30:31]
; GFX11-TRUE16-NEXT: .LBB45_4:
+; GFX11-TRUE16-NEXT: ; implicit-def: $sgpr58_lo16
; GFX11-TRUE16-NEXT: ; implicit-def: $sgpr57_lo16
; GFX11-TRUE16-NEXT: ; implicit-def: $sgpr56_lo16
-; GFX11-TRUE16-NEXT: ; implicit-def: $sgpr47_lo16
; GFX11-TRUE16-NEXT: ; implicit-def: $sgpr46_lo16
; GFX11-TRUE16-NEXT: ; implicit-def: $sgpr45_lo16
; GFX11-TRUE16-NEXT: ; implicit-def: $sgpr44_lo16
@@ -23545,7 +23555,7 @@ define inreg <10 x i64> @bitcast_v40f16_to_v10i64_scalar(<40 x half> inreg %a, i
; SI-NEXT: v_cvt_f16_f32_e32 v62, v2
; SI-NEXT: v_cvt_f16_f32_e32 v44, v5
; SI-NEXT: v_cvt_f16_f32_e32 v43, v4
-; SI-NEXT: v_cvt_f16_f32_e32 v42, v7
+; SI-NEXT: v_cvt_f16_f32_e32 v56, v7
; SI-NEXT: v_cvt_f16_f32_e32 v41, v6
; SI-NEXT: v_cvt_f16_f32_e32 v40, v9
; SI-NEXT: v_cvt_f16_f32_e32 v55, v8
@@ -23565,26 +23575,26 @@ define inreg <10 x i64> @bitcast_v40f16_to_v10i64_scalar(<40 x half> inreg %a, i
; SI-NEXT: v_cvt_f16_f32_e32 v22, v22
; SI-NEXT: v_cvt_f16_f32_e32 v20, v25
; SI-NEXT: v_cvt_f16_f32_e32 v21, v24
-; SI-NEXT: v_cvt_f16_f32_e32 v58, s17
-; SI-NEXT: v_cvt_f16_f32_e32 v60, s16
-; SI-NEXT: v_cvt_f16_f32_e32 v59, s19
-; SI-NEXT: v_cvt_f16_f32_e32 v57, s18
+; SI-NEXT: v_cvt_f16_f32_e32 v61, s17
+; SI-NEXT: v_cvt_f16_f32_e32 v57, s16
+; SI-NEXT: v_cvt_f16_f32_e32 v60, s19
+; SI-NEXT: v_cvt_f16_f32_e32 v58, s18
; SI-NEXT: v_cvt_f16_f32_e32 v39, s21
-; SI-NEXT: v_cvt_f16_f32_e32 v56, s20
+; SI-NEXT: v_cvt_f16_f32_e32 v42, s20
; SI-NEXT: v_cvt_f16_f32_e32 v38, s23
; SI-NEXT: v_cvt_f16_f32_e32 v37, s22
; SI-NEXT: v_cvt_f16_f32_e32 v36, s25
; SI-NEXT: v_cvt_f16_f32_e32 v35, s24
; SI-NEXT: v_cvt_f16_f32_e32 v34, s27
-; SI-NEXT: v_cvt_f16_f32_e32 v61, s26
+; SI-NEXT: v_cvt_f16_f32_e32 v33, s26
; SI-NEXT: v_cvt_f16_f32_e32 v25, s29
; SI-NEXT: v_cvt_f16_f32_e32 v24, s28
; SI-NEXT: v_cmp_ne_u32_e32 vcc, 0, v26
; SI-NEXT: s_and_b64 s[4:5], vcc, exec
; SI-NEXT: s_cbranch_scc0 .LBB47_4
; SI-NEXT: ; %bb.1: ; %cmp.false
-; SI-NEXT: v_lshlrev_b32_e32 v0, 16, v58
-; SI-NEXT: v_lshlrev_b32_e32 v1, 16, v59
+; SI-NEXT: v_lshlrev_b32_e32 v0, 16, v61
+; SI-NEXT: v_lshlrev_b32_e32 v1, 16, v60
; SI-NEXT: v_lshlrev_b32_e32 v2, 16, v39
; SI-NEXT: v_lshlrev_b32_e32 v3, 16, v38
; SI-NEXT: v_lshlrev_b32_e32 v4, 16, v36
@@ -23593,7 +23603,7 @@ define inreg <10 x i64> @bitcast_v40f16_to_v10i64_scalar(<40 x half> inreg %a, i
; SI-NEXT: v_lshlrev_b32_e32 v7, 16, v32
; SI-NEXT: v_lshlrev_b32_e32 v8, 16, v63
; SI-NEXT: v_lshlrev_b32_e32 v9, 16, v44
-; SI-NEXT: v_lshlrev_b32_e32 v10, 16, v42
+; SI-NEXT: v_lshlrev_b32_e32 v10, 16, v56
; SI-NEXT: v_lshlrev_b32_e32 v11, 16, v40
; SI-NEXT: v_lshlrev_b32_e32 v12, 16, v54
; SI-NEXT: v_lshlrev_b32_e32 v13, 16, v52
@@ -23603,12 +23613,12 @@ define inreg <10 x i64> @bitcast_v40f16_to_v10i64_scalar(<40 x half> inreg %a, i
; SI-NEXT: v_lshlrev_b32_e32 v17, 16, v28
; SI-NEXT: v_lshlrev_b32_e32 v18, 16, v23
; SI-NEXT: v_lshlrev_b32_e32 v19, 16, v20
-; SI-NEXT: v_or_b32_e32 v0, v60, v0
-; SI-NEXT: v_or_b32_e32 v1, v57, v1
-; SI-NEXT: v_or_b32_e32 v2, v56, v2
+; SI-NEXT: v_or_b32_e32 v0, v57, v0
+; SI-NEXT: v_or_b32_e32 v1, v58, v1
+; SI-NEXT: v_or_b32_e32 v2, v42, v2
; SI-NEXT: v_or_b32_e32 v3, v37, v3
; SI-NEXT: v_or_b32_e32 v4, v35, v4
-; SI-NEXT: v_or_b32_e32 v5, v61, v5
+; SI-NEXT: v_or_b32_e32 v5, v33, v5
; SI-NEXT: v_or_b32_e32 v6, v24, v6
; SI-NEXT: v_or_b32_e32 v7, v47, v7
; SI-NEXT: v_or_b32_e32 v8, v62, v8
@@ -23625,11 +23635,12 @@ define inreg <10 x i64> @bitcast_v40f16_to_v10i64_scalar(<40 x half> inreg %a, i
; SI-NEXT: v_or_b32_e32 v19, v21, v19
; SI-NEXT: s_cbranch_execnz .LBB47_3
; SI-NEXT: .LBB47_2: ; %cmp.true
-; SI-NEXT: v_cvt_f32_f16_e32 v0, v58
-; SI-NEXT: s_waitcnt vmcnt(1)
-; SI-NEXT: v_cvt_f32_f16_e32 v2, v59
-; SI-NEXT: v_cvt_f32_f16_e32 v1, v60
-; SI-NEXT: v_cvt_f32_f16_e32 v3, v57
+; SI-NEXT: v_cvt_f32_f16_e32 v0, v61
+; SI-NEXT: s_waitcnt vmcnt(7)
+; SI-NEXT: v_cvt_f32_f16_e32 v2, v60
+; SI-NEXT: v_cvt_f32_f16_e32 v1, v57
+; SI-NEXT: s_waitcnt vmcnt(3)
+; SI-NEXT: v_cvt_f32_f16_e32 v3, v58
; SI-NEXT: v_add_f32_e32 v0, 0x38000000, v0
; SI-NEXT: v_add_f32_e32 v2, 0x38000000, v2
; SI-NEXT: v_add_f32_e32 v1, 0x38000000, v1
@@ -23643,7 +23654,8 @@ define inreg <10 x i64> @bitcast_v40f16_to_v10i64_scalar(<40 x half> inreg %a, i
; SI-NEXT: v_or_b32_e32 v0, v1, v0
; SI-NEXT: v_or_b32_e32 v1, v3, v2
; SI-NEXT: v_cvt_f32_f16_e32 v2, v39
-; SI-NEXT: v_cvt_f32_f16_e32 v3, v56
+; SI-NEXT: s_waitcnt vmcnt(1)
+; SI-NEXT: v_cvt_f32_f16_e32 v3, v42
; SI-NEXT: v_cvt_f32_f16_e32 v4, v37
; SI-NEXT: v_cvt_f32_f16_e32 v5, v35
; SI-NEXT: v_add_f32_e32 v2, 0x38000000, v2
@@ -23657,7 +23669,7 @@ define inreg <10 x i64> @bitcast_v40f16_to_v10i64_scalar(<40 x half> inreg %a, i
; SI-NEXT: v_cvt_f32_f16_e32 v3, v38
; SI-NEXT: v_add_f32_e32 v5, 0x38000000, v5
; SI-NEXT: v_cvt_f16_f32_e32 v5, v5
-; SI-NEXT: v_cvt_f32_f16_e32 v6, v61
+; SI-NEXT: v_cvt_f32_f16_e32 v6, v33
; SI-NEXT: v_add_f32_e32 v3, 0x38000000, v3
; SI-NEXT: v_cvt_f16_f32_e32 v3, v3
; SI-NEXT: v_cvt_f32_f16_e32 v7, v24
@@ -23684,7 +23696,7 @@ define inreg <10 x i64> @bitcast_v40f16_to_v10i64_scalar(<40 x half> inreg %a, i
; SI-NEXT: v_cvt_f16_f32_e32 v5, v5
; SI-NEXT: v_add_f32_e32 v10, 0x38000000, v10
; SI-NEXT: v_cvt_f16_f32_e32 v10, v10
-; SI-NEXT: v_cvt_f32_f16_e32 v11, v42
+; SI-NEXT: v_cvt_f32_f16_e32 v11, v56
; SI-NEXT: v_lshlrev_b32_e32 v5, 16, v5
; SI-NEXT: v_or_b32_e32 v5, v6, v5
; SI-NEXT: v_cvt_f32_f16_e32 v6, v25
@@ -23808,20 +23820,20 @@ define inreg <10 x i64> @bitcast_v40f16_to_v10i64_scalar(<40 x half> inreg %a, i
; SI-NEXT: s_setpc_b64 s[30:31]
; SI-NEXT: .LBB47_4:
; SI-NEXT: buffer_store_dword v41, off, s[0:3], s32 offset:64 ; 4-byte Folded Spill
-; SI-NEXT: buffer_store_dword v59, off, s[0:3], s32 offset:88 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v42, off, s[0:3], s32 offset:76 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_mov_b32_e32 v59, v48
+; SI-NEXT: v_mov_b32_e32 v42, v48
; SI-NEXT: v_mov_b32_e32 v48, v21
-; SI-NEXT: buffer_store_dword v42, off, s[0:3], s32 offset:68 ; 4-byte Folded Spill
-; SI-NEXT: buffer_store_dword v60, off, s[0:3], s32 offset:92 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v56, off, s[0:3], s32 offset:68 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v58, off, s[0:3], s32 offset:84 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_mov_b32_e32 v60, v49
+; SI-NEXT: v_mov_b32_e32 v58, v49
; SI-NEXT: v_mov_b32_e32 v49, v20
; SI-NEXT: buffer_store_dword v43, off, s[0:3], s32 offset:72 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_mov_b32_e32 v43, v50
; SI-NEXT: v_mov_b32_e32 v50, v22
-; SI-NEXT: buffer_store_dword v44, off, s[0:3], s32 offset:84 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v44, off, s[0:3], s32 offset:88 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_mov_b32_e32 v44, v51
; SI-NEXT: v_mov_b32_e32 v51, v23
@@ -23829,22 +23841,21 @@ define inreg <10 x i64> @bitcast_v40f16_to_v10i64_scalar(<40 x half> inreg %a, i
; SI-NEXT: v_mov_b32_e32 v52, v27
; SI-NEXT: v_mov_b32_e32 v46, v53
; SI-NEXT: v_mov_b32_e32 v53, v28
-; SI-NEXT: buffer_store_dword v36, off, s[0:3], s32 offset:76 ; 4-byte Folded Spill
-; SI-NEXT: buffer_store_dword v39, off, s[0:3], s32 offset:80 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v36, off, s[0:3], s32 offset:80 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v60, off, s[0:3], s32 offset:92 ; 4-byte Folded Spill
+; SI-NEXT: s_waitcnt expcnt(0)
+; SI-NEXT: v_mov_b32_e32 v60, v33
; SI-NEXT: v_mov_b32_e32 v41, v32
; SI-NEXT: v_mov_b32_e32 v33, v47
; SI-NEXT: v_mov_b32_e32 v47, v54
; SI-NEXT: v_mov_b32_e32 v54, v29
-; SI-NEXT: v_mov_b32_e32 v42, v56
; SI-NEXT: v_mov_b32_e32 v56, v55
; SI-NEXT: v_mov_b32_e32 v55, v30
-; SI-NEXT: s_waitcnt expcnt(1)
-; SI-NEXT: v_mov_b32_e32 v36, v57
+; SI-NEXT: v_mov_b32_e32 v59, v57
; SI-NEXT: v_mov_b32_e32 v57, v40
; SI-NEXT: v_mov_b32_e32 v40, v31
-; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_mov_b32_e32 v39, v58
-; SI-NEXT: v_mov_b32_e32 v58, v37
+; SI-NEXT: v_mov_b32_e32 v36, v39
+; SI-NEXT: v_mov_b32_e32 v39, v37
; SI-NEXT: v_mov_b32_e32 v37, v34
; SI-NEXT: v_mov_b32_e32 v34, v24
; SI-NEXT: v_mov_b32_e32 v32, v38
@@ -23856,34 +23867,34 @@ define inreg <10 x i64> @bitcast_v40f16_to_v10i64_scalar(<40 x half> inreg %a, i
; SI-NEXT: v_mov_b32_e32 v38, v32
; SI-NEXT: v_mov_b32_e32 v24, v34
; SI-NEXT: v_mov_b32_e32 v34, v37
-; SI-NEXT: v_mov_b32_e32 v37, v58
-; SI-NEXT: v_mov_b32_e32 v58, v39
-; SI-NEXT: v_mov_b32_e32 v31, v40
-; SI-NEXT: v_mov_b32_e32 v40, v57
-; SI-NEXT: v_mov_b32_e32 v57, v36
+; SI-NEXT: v_mov_b32_e32 v37, v39
+; SI-NEXT: v_mov_b32_e32 v39, v36
; SI-NEXT: v_mov_b32_e32 v30, v55
; SI-NEXT: v_mov_b32_e32 v55, v56
-; SI-NEXT: v_mov_b32_e32 v56, v42
+; SI-NEXT: v_mov_b32_e32 v29, v54
+; SI-NEXT: v_mov_b32_e32 v54, v47
+; SI-NEXT: v_mov_b32_e32 v47, v33
; SI-NEXT: v_mov_b32_e32 v32, v41
-; SI-NEXT: buffer_load_dword v39, off, s[0:3], s32 offset:80 ; 4-byte Folded Reload
-; SI-NEXT: buffer_load_dword v36, off, s[0:3], s32 offset:76 ; 4-byte Folded Reload
+; SI-NEXT: v_mov_b32_e32 v33, v60
+; SI-NEXT: buffer_load_dword v60, off, s[0:3], s32 offset:92 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v36, off, s[0:3], s32 offset:80 ; 4-byte Folded Reload
; SI-NEXT: v_mov_b32_e32 v23, v51
; SI-NEXT: v_mov_b32_e32 v51, v44
-; SI-NEXT: buffer_load_dword v44, off, s[0:3], s32 offset:84 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v44, off, s[0:3], s32 offset:88 ; 4-byte Folded Reload
; SI-NEXT: v_mov_b32_e32 v22, v50
; SI-NEXT: v_mov_b32_e32 v50, v43
; SI-NEXT: buffer_load_dword v43, off, s[0:3], s32 offset:72 ; 4-byte Folded Reload
; SI-NEXT: v_mov_b32_e32 v20, v49
-; SI-NEXT: v_mov_b32_e32 v49, v60
-; SI-NEXT: buffer_load_dword v60, off, s[0:3], s32 offset:92 ; 4-byte Folded Reload
-; SI-NEXT: buffer_load_dword v42, off, s[0:3], s32 offset:68 ; 4-byte Folded Reload
+; SI-NEXT: v_mov_b32_e32 v49, v58
+; SI-NEXT: buffer_load_dword v58, off, s[0:3], s32 offset:84 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v56, off, s[0:3], s32 offset:68 ; 4-byte Folded Reload
; SI-NEXT: v_mov_b32_e32 v21, v48
-; SI-NEXT: v_mov_b32_e32 v48, v59
-; SI-NEXT: buffer_load_dword v59, off, s[0:3], s32 offset:88 ; 4-byte Folded Reload
+; SI-NEXT: v_mov_b32_e32 v48, v42
+; SI-NEXT: buffer_load_dword v42, off, s[0:3], s32 offset:76 ; 4-byte Folded Reload
; SI-NEXT: buffer_load_dword v41, off, s[0:3], s32 offset:64 ; 4-byte Folded Reload
-; SI-NEXT: v_mov_b32_e32 v29, v54
-; SI-NEXT: v_mov_b32_e32 v54, v47
-; SI-NEXT: v_mov_b32_e32 v47, v33
+; SI-NEXT: v_mov_b32_e32 v31, v40
+; SI-NEXT: v_mov_b32_e32 v40, v57
+; SI-NEXT: v_mov_b32_e32 v57, v59
; SI-NEXT: v_mov_b32_e32 v28, v53
; SI-NEXT: v_mov_b32_e32 v53, v46
; SI-NEXT: v_mov_b32_e32 v27, v52
@@ -30093,7 +30104,7 @@ define inreg <10 x double> @bitcast_v40f16_to_v10f64_scalar(<40 x half> inreg %a
; SI-NEXT: v_cvt_f16_f32_e32 v62, v2
; SI-NEXT: v_cvt_f16_f32_e32 v44, v5
; SI-NEXT: v_cvt_f16_f32_e32 v43, v4
-; SI-NEXT: v_cvt_f16_f32_e32 v42, v7
+; SI-NEXT: v_cvt_f16_f32_e32 v56, v7
; SI-NEXT: v_cvt_f16_f32_e32 v41, v6
; SI-NEXT: v_cvt_f16_f32_e32 v40, v9
; SI-NEXT: v_cvt_f16_f32_e32 v55, v8
@@ -30113,26 +30124,26 @@ define inreg <10 x double> @bitcast_v40f16_to_v10f64_scalar(<40 x half> inreg %a
; SI-NEXT: v_cvt_f16_f32_e32 v22, v22
; SI-NEXT: v_cvt_f16_f32_e32 v20, v25
; SI-NEXT: v_cvt_f16_f32_e32 v21, v24
-; SI-NEXT: v_cvt_f16_f32_e32 v58, s17
-; SI-NEXT: v_cvt_f16_f32_e32 v60, s16
-; SI-NEXT: v_cvt_f16_f32_e32 v59, s19
-; SI-NEXT: v_cvt_f16_f32_e32 v57, s18
+; SI-NEXT: v_cvt_f16_f32_e32 v61, s17
+; SI-NEXT: v_cvt_f16_f32_e32 v57, s16
+; SI-NEXT: v_cvt_f16_f32_e32 v60, s19
+; SI-NEXT: v_cvt_f16_f32_e32 v58, s18
; SI-NEXT: v_cvt_f16_f32_e32 v39, s21
-; SI-NEXT: v_cvt_f16_f32_e32 v56, s20
+; SI-NEXT: v_cvt_f16_f32_e32 v42, s20
; SI-NEXT: v_cvt_f16_f32_e32 v38, s23
; SI-NEXT: v_cvt_f16_f32_e32 v37, s22
; SI-NEXT: v_cvt_f16_f32_e32 v36, s25
; SI-NEXT: v_cvt_f16_f32_e32 v35, s24
; SI-NEXT: v_cvt_f16_f32_e32 v34, s27
-; SI-NEXT: v_cvt_f16_f32_e32 v61, s26
+; SI-NEXT: v_cvt_f16_f32_e32 v33, s26
; SI-NEXT: v_cvt_f16_f32_e32 v25, s29
; SI-NEXT: v_cvt_f16_f32_e32 v24, s28
; SI-NEXT: v_cmp_ne_u32_e32 vcc, 0, v26
; SI-NEXT: s_and_b64 s[4:5], vcc, exec
; SI-NEXT: s_cbranch_scc0 .LBB55_4
; SI-NEXT: ; %bb.1: ; %cmp.false
-; SI-NEXT: v_lshlrev_b32_e32 v0, 16, v58
-; SI-NEXT: v_lshlrev_b32_e32 v1, 16, v59
+; SI-NEXT: v_lshlrev_b32_e32 v0, 16, v61
+; SI-NEXT: v_lshlrev_b32_e32 v1, 16, v60
; SI-NEXT: v_lshlrev_b32_e32 v2, 16, v39
; SI-NEXT: v_lshlrev_b32_e32 v3, 16, v38
; SI-NEXT: v_lshlrev_b32_e32 v4, 16, v36
@@ -30141,7 +30152,7 @@ define inreg <10 x double> @bitcast_v40f16_to_v10f64_scalar(<40 x half> inreg %a
; SI-NEXT: v_lshlrev_b32_e32 v7, 16, v32
; SI-NEXT: v_lshlrev_b32_e32 v8, 16, v63
; SI-NEXT: v_lshlrev_b32_e32 v9, 16, v44
-; SI-NEXT: v_lshlrev_b32_e32 v10, 16, v42
+; SI-NEXT: v_lshlrev_b32_e32 v10, 16, v56
; SI-NEXT: v_lshlrev_b32_e32 v11, 16, v40
; SI-NEXT: v_lshlrev_b32_e32 v12, 16, v54
; SI-NEXT: v_lshlrev_b32_e32 v13, 16, v52
@@ -30151,12 +30162,12 @@ define inreg <10 x double> @bitcast_v40f16_to_v10f64_scalar(<40 x half> inreg %a
; SI-NEXT: v_lshlrev_b32_e32 v17, 16, v28
; SI-NEXT: v_lshlrev_b32_e32 v18, 16, v23
; SI-NEXT: v_lshlrev_b32_e32 v19, 16, v20
-; SI-NEXT: v_or_b32_e32 v0, v60, v0
-; SI-NEXT: v_or_b32_e32 v1, v57, v1
-; SI-NEXT: v_or_b32_e32 v2, v56, v2
+; SI-NEXT: v_or_b32_e32 v0, v57, v0
+; SI-NEXT: v_or_b32_e32 v1, v58, v1
+; SI-NEXT: v_or_b32_e32 v2, v42, v2
; SI-NEXT: v_or_b32_e32 v3, v37, v3
; SI-NEXT: v_or_b32_e32 v4, v35, v4
-; SI-NEXT: v_or_b32_e32 v5, v61, v5
+; SI-NEXT: v_or_b32_e32 v5, v33, v5
; SI-NEXT: v_or_b32_e32 v6, v24, v6
; SI-NEXT: v_or_b32_e32 v7, v47, v7
; SI-NEXT: v_or_b32_e32 v8, v62, v8
@@ -30173,11 +30184,12 @@ define inreg <10 x double> @bitcast_v40f16_to_v10f64_scalar(<40 x half> inreg %a
; SI-NEXT: v_or_b32_e32 v19, v21, v19
; SI-NEXT: s_cbranch_execnz .LBB55_3
; SI-NEXT: .LBB55_2: ; %cmp.true
-; SI-NEXT: v_cvt_f32_f16_e32 v0, v58
-; SI-NEXT: s_waitcnt vmcnt(1)
-; SI-NEXT: v_cvt_f32_f16_e32 v2, v59
-; SI-NEXT: v_cvt_f32_f16_e32 v1, v60
-; SI-NEXT: v_cvt_f32_f16_e32 v3, v57
+; SI-NEXT: v_cvt_f32_f16_e32 v0, v61
+; SI-NEXT: s_waitcnt vmcnt(7)
+; SI-NEXT: v_cvt_f32_f16_e32 v2, v60
+; SI-NEXT: v_cvt_f32_f16_e32 v1, v57
+; SI-NEXT: s_waitcnt vmcnt(3)
+; SI-NEXT: v_cvt_f32_f16_e32 v3, v58
; SI-NEXT: v_add_f32_e32 v0, 0x38000000, v0
; SI-NEXT: v_add_f32_e32 v2, 0x38000000, v2
; SI-NEXT: v_add_f32_e32 v1, 0x38000000, v1
@@ -30191,7 +30203,8 @@ define inreg <10 x double> @bitcast_v40f16_to_v10f64_scalar(<40 x half> inreg %a
; SI-NEXT: v_or_b32_e32 v0, v1, v0
; SI-NEXT: v_or_b32_e32 v1, v3, v2
; SI-NEXT: v_cvt_f32_f16_e32 v2, v39
-; SI-NEXT: v_cvt_f32_f16_e32 v3, v56
+; SI-NEXT: s_waitcnt vmcnt(1)
+; SI-NEXT: v_cvt_f32_f16_e32 v3, v42
; SI-NEXT: v_cvt_f32_f16_e32 v4, v37
; SI-NEXT: v_cvt_f32_f16_e32 v5, v35
; SI-NEXT: v_add_f32_e32 v2, 0x38000000, v2
@@ -30205,7 +30218,7 @@ define inreg <10 x double> @bitcast_v40f16_to_v10f64_scalar(<40 x half> inreg %a
; SI-NEXT: v_cvt_f32_f16_e32 v3, v38
; SI-NEXT: v_add_f32_e32 v5, 0x38000000, v5
; SI-NEXT: v_cvt_f16_f32_e32 v5, v5
-; SI-NEXT: v_cvt_f32_f16_e32 v6, v61
+; SI-NEXT: v_cvt_f32_f16_e32 v6, v33
; SI-NEXT: v_add_f32_e32 v3, 0x38000000, v3
; SI-NEXT: v_cvt_f16_f32_e32 v3, v3
; SI-NEXT: v_cvt_f32_f16_e32 v7, v24
@@ -30232,7 +30245,7 @@ define inreg <10 x double> @bitcast_v40f16_to_v10f64_scalar(<40 x half> inreg %a
; SI-NEXT: v_cvt_f16_f32_e32 v5, v5
; SI-NEXT: v_add_f32_e32 v10, 0x38000000, v10
; SI-NEXT: v_cvt_f16_f32_e32 v10, v10
-; SI-NEXT: v_cvt_f32_f16_e32 v11, v42
+; SI-NEXT: v_cvt_f32_f16_e32 v11, v56
; SI-NEXT: v_lshlrev_b32_e32 v5, 16, v5
; SI-NEXT: v_or_b32_e32 v5, v6, v5
; SI-NEXT: v_cvt_f32_f16_e32 v6, v25
@@ -30356,20 +30369,20 @@ define inreg <10 x double> @bitcast_v40f16_to_v10f64_scalar(<40 x half> inreg %a
; SI-NEXT: s_setpc_b64 s[30:31]
; SI-NEXT: .LBB55_4:
; SI-NEXT: buffer_store_dword v41, off, s[0:3], s32 offset:64 ; 4-byte Folded Spill
-; SI-NEXT: buffer_store_dword v59, off, s[0:3], s32 offset:88 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v42, off, s[0:3], s32 offset:76 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_mov_b32_e32 v59, v48
+; SI-NEXT: v_mov_b32_e32 v42, v48
; SI-NEXT: v_mov_b32_e32 v48, v21
-; SI-NEXT: buffer_store_dword v42, off, s[0:3], s32 offset:68 ; 4-byte Folded Spill
-; SI-NEXT: buffer_store_dword v60, off, s[0:3], s32 offset:92 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v56, off, s[0:3], s32 offset:68 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v58, off, s[0:3], s32 offset:84 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_mov_b32_e32 v60, v49
+; SI-NEXT: v_mov_b32_e32 v58, v49
; SI-NEXT: v_mov_b32_e32 v49, v20
; SI-NEXT: buffer_store_dword v43, off, s[0:3], s32 offset:72 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_mov_b32_e32 v43, v50
; SI-NEXT: v_mov_b32_e32 v50, v22
-; SI-NEXT: buffer_store_dword v44, off, s[0:3], s32 offset:84 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v44, off, s[0:3], s32 offset:88 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_mov_b32_e32 v44, v51
; SI-NEXT: v_mov_b32_e32 v51, v23
@@ -30377,22 +30390,21 @@ define inreg <10 x double> @bitcast_v40f16_to_v10f64_scalar(<40 x half> inreg %a
; SI-NEXT: v_mov_b32_e32 v52, v27
; SI-NEXT: v_mov_b32_e32 v46, v53
; SI-NEXT: v_mov_b32_e32 v53, v28
-; SI-NEXT: buffer_store_dword v36, off, s[0:3], s32 offset:76 ; 4-byte Folded Spill
-; SI-NEXT: buffer_store_dword v39, off, s[0:3], s32 offset:80 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v36, off, s[0:3], s32 offset:80 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v60, off, s[0:3], s32 offset:92 ; 4-byte Folded Spill
+; SI-NEXT: s_waitcnt expcnt(0)
+; SI-NEXT: v_mov_b32_e32 v60, v33
; SI-NEXT: v_mov_b32_e32 v41, v32
; SI-NEXT: v_mov_b32_e32 v33, v47
; SI-NEXT: v_mov_b32_e32 v47, v54
; SI-NEXT: v_mov_b32_e32 v54, v29
-; SI-NEXT: v_mov_b32_e32 v42, v56
; SI-NEXT: v_mov_b32_e32 v56, v55
; SI-NEXT: v_mov_b32_e32 v55, v30
-; SI-NEXT: s_waitcnt expcnt(1)
-; SI-NEXT: v_mov_b32_e32 v36, v57
+; SI-NEXT: v_mov_b32_e32 v59, v57
; SI-NEXT: v_mov_b32_e32 v57, v40
; SI-NEXT: v_mov_b32_e32 v40, v31
-; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_mov_b32_e32 v39, v58
-; SI-NEXT: v_mov_b32_e32 v58, v37
+; SI-NEXT: v_mov_b32_e32 v36, v39
+; SI-NEXT: v_mov_b32_e32 v39, v37
; SI-NEXT: v_mov_b32_e32 v37, v34
; SI-NEXT: v_mov_b32_e32 v34, v24
; SI-NEXT: v_mov_b32_e32 v32, v38
@@ -30404,34 +30416,34 @@ define inreg <10 x double> @bitcast_v40f16_to_v10f64_scalar(<40 x half> inreg %a
; SI-NEXT: v_mov_b32_e32 v38, v32
; SI-NEXT: v_mov_b32_e32 v24, v34
; SI-NEXT: v_mov_b32_e32 v34, v37
-; SI-NEXT: v_mov_b32_e32 v37, v58
-; SI-NEXT: v_mov_b32_e32 v58, v39
-; SI-NEXT: v_mov_b32_e32 v31, v40
-; SI-NEXT: v_mov_b32_e32 v40, v57
-; SI-NEXT: v_mov_b32_e32 v57, v36
+; SI-NEXT: v_mov_b32_e32 v37, v39
+; SI-NEXT: v_mov_b32_e32 v39, v36
; SI-NEXT: v_mov_b32_e32 v30, v55
; SI-NEXT: v_mov_b32_e32 v55, v56
-; SI-NEXT: v_mov_b32_e32 v56, v42
+; SI-NEXT: v_mov_b32_e32 v29, v54
+; SI-NEXT: v_mov_b32_e32 v54, v47
+; SI-NEXT: v_mov_b32_e32 v47, v33
; SI-NEXT: v_mov_b32_e32 v32, v41
-; SI-NEXT: buffer_load_dword v39, off, s[0:3], s32 offset:80 ; 4-byte Folded Reload
-; SI-NEXT: buffer_load_dword v36, off, s[0:3], s32 offset:76 ; 4-byte Folded Reload
+; SI-NEXT: v_mov_b32_e32 v33, v60
+; SI-NEXT: buffer_load_dword v60, off, s[0:3], s32 offset:92 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v36, off, s[0:3], s32 offset:80 ; 4-byte Folded Reload
; SI-NEXT: v_mov_b32_e32 v23, v51
; SI-NEXT: v_mov_b32_e32 v51, v44
-; SI-NEXT: buffer_load_dword v44, off, s[0:3], s32 offset:84 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v44, off, s[0:3], s32 offset:88 ; 4-byte Folded Reload
; SI-NEXT: v_mov_b32_e32 v22, v50
; SI-NEXT: v_mov_b32_e32 v50, v43
; SI-NEXT: buffer_load_dword v43, off, s[0:3], s32 offset:72 ; 4-byte Folded Reload
; SI-NEXT: v_mov_b32_e32 v20, v49
-; SI-NEXT: v_mov_b32_e32 v49, v60
-; SI-NEXT: buffer_load_dword v60, off, s[0:3], s32 offset:92 ; 4-byte Folded Reload
-; SI-NEXT: buffer_load_dword v42, off, s[0:3], s32 offset:68 ; 4-byte Folded Reload
+; SI-NEXT: v_mov_b32_e32 v49, v58
+; SI-NEXT: buffer_load_dword v58, off, s[0:3], s32 offset:84 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v56, off, s[0:3], s32 offset:68 ; 4-byte Folded Reload
; SI-NEXT: v_mov_b32_e32 v21, v48
-; SI-NEXT: v_mov_b32_e32 v48, v59
-; SI-NEXT: buffer_load_dword v59, off, s[0:3], s32 offset:88 ; 4-byte Folded Reload
+; SI-NEXT: v_mov_b32_e32 v48, v42
+; SI-NEXT: buffer_load_dword v42, off, s[0:3], s32 offset:76 ; 4-byte Folded Reload
; SI-NEXT: buffer_load_dword v41, off, s[0:3], s32 offset:64 ; 4-byte Folded Reload
-; SI-NEXT: v_mov_b32_e32 v29, v54
-; SI-NEXT: v_mov_b32_e32 v54, v47
-; SI-NEXT: v_mov_b32_e32 v47, v33
+; SI-NEXT: v_mov_b32_e32 v31, v40
+; SI-NEXT: v_mov_b32_e32 v40, v57
+; SI-NEXT: v_mov_b32_e32 v57, v59
; SI-NEXT: v_mov_b32_e32 v28, v53
; SI-NEXT: v_mov_b32_e32 v53, v46
; SI-NEXT: v_mov_b32_e32 v27, v52
diff --git a/llvm/test/CodeGen/AMDGPU/amdgcn.bitcast.64bit.ll b/llvm/test/CodeGen/AMDGPU/amdgcn.bitcast.64bit.ll
index b73870977c429..50b13db0e15d4 100644
--- a/llvm/test/CodeGen/AMDGPU/amdgcn.bitcast.64bit.ll
+++ b/llvm/test/CodeGen/AMDGPU/amdgcn.bitcast.64bit.ll
@@ -2889,7 +2889,7 @@ define inreg <8 x i8> @bitcast_i64_to_v8i8_scalar(i64 inreg %a, i32 inreg %b) {
; GFX11-TRUE16: ; %bb.0:
; GFX11-TRUE16-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX11-TRUE16-NEXT: s_cmp_lg_u32 s2, 0
-; GFX11-TRUE16-NEXT: s_mov_b32 s8, 0
+; GFX11-TRUE16-NEXT: s_mov_b32 s7, 0
; GFX11-TRUE16-NEXT: s_cbranch_scc0 .LBB25_4
; GFX11-TRUE16-NEXT: ; %bb.1: ; %cmp.false
; GFX11-TRUE16-NEXT: s_lshr_b64 s[2:3], s[0:1], 24
@@ -2897,8 +2897,8 @@ define inreg <8 x i8> @bitcast_i64_to_v8i8_scalar(i64 inreg %a, i32 inreg %b) {
; GFX11-TRUE16-NEXT: s_lshr_b32 s4, s1, 16
; GFX11-TRUE16-NEXT: s_lshr_b32 s5, s1, 8
; GFX11-TRUE16-NEXT: s_lshr_b32 s6, s0, 16
-; GFX11-TRUE16-NEXT: s_lshr_b32 s7, s0, 8
-; GFX11-TRUE16-NEXT: s_and_not1_b32 vcc_lo, exec_lo, s8
+; GFX11-TRUE16-NEXT: s_lshr_b32 s8, s0, 8
+; GFX11-TRUE16-NEXT: s_and_not1_b32 vcc_lo, exec_lo, s7
; GFX11-TRUE16-NEXT: s_cbranch_vccnz .LBB25_3
; GFX11-TRUE16-NEXT: .LBB25_2: ; %cmp.true
; GFX11-TRUE16-NEXT: s_add_u32 s0, s0, 3
@@ -2908,10 +2908,10 @@ define inreg <8 x i8> @bitcast_i64_to_v8i8_scalar(i64 inreg %a, i32 inreg %b) {
; GFX11-TRUE16-NEXT: s_lshr_b32 s3, s1, 24
; GFX11-TRUE16-NEXT: s_lshr_b32 s4, s1, 16
; GFX11-TRUE16-NEXT: s_lshr_b32 s5, s1, 8
-; GFX11-TRUE16-NEXT: s_lshr_b32 s7, s0, 8
+; GFX11-TRUE16-NEXT: s_lshr_b32 s8, s0, 8
; GFX11-TRUE16-NEXT: .LBB25_3: ; %end
; GFX11-TRUE16-NEXT: v_mov_b16_e32 v0.l, s0
-; GFX11-TRUE16-NEXT: v_mov_b16_e32 v1.l, s7
+; GFX11-TRUE16-NEXT: v_mov_b16_e32 v1.l, s8
; GFX11-TRUE16-NEXT: v_mov_b16_e32 v2.l, s6
; GFX11-TRUE16-NEXT: v_mov_b16_e32 v3.l, s2
; GFX11-TRUE16-NEXT: v_mov_b16_e32 v4.l, s1
@@ -2920,7 +2920,7 @@ define inreg <8 x i8> @bitcast_i64_to_v8i8_scalar(i64 inreg %a, i32 inreg %b) {
; GFX11-TRUE16-NEXT: v_mov_b16_e32 v7.l, s3
; GFX11-TRUE16-NEXT: s_setpc_b64 s[30:31]
; GFX11-TRUE16-NEXT: .LBB25_4:
-; GFX11-TRUE16-NEXT: ; implicit-def: $sgpr7_lo16
+; GFX11-TRUE16-NEXT: ; implicit-def: $sgpr8_lo16
; GFX11-TRUE16-NEXT: ; implicit-def: $sgpr6_lo16
; GFX11-TRUE16-NEXT: ; implicit-def: $sgpr2_lo16
; GFX11-TRUE16-NEXT: ; implicit-def: $sgpr5_lo16
@@ -2932,7 +2932,7 @@ define inreg <8 x i8> @bitcast_i64_to_v8i8_scalar(i64 inreg %a, i32 inreg %b) {
; GFX11-FAKE16: ; %bb.0:
; GFX11-FAKE16-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX11-FAKE16-NEXT: s_cmp_lg_u32 s2, 0
-; GFX11-FAKE16-NEXT: s_mov_b32 s8, 0
+; GFX11-FAKE16-NEXT: s_mov_b32 s7, 0
; GFX11-FAKE16-NEXT: s_cbranch_scc0 .LBB25_4
; GFX11-FAKE16-NEXT: ; %bb.1: ; %cmp.false
; GFX11-FAKE16-NEXT: s_lshr_b64 s[2:3], s[0:1], 24
@@ -2940,8 +2940,8 @@ define inreg <8 x i8> @bitcast_i64_to_v8i8_scalar(i64 inreg %a, i32 inreg %b) {
; GFX11-FAKE16-NEXT: s_lshr_b32 s4, s1, 16
; GFX11-FAKE16-NEXT: s_lshr_b32 s5, s1, 8
; GFX11-FAKE16-NEXT: s_lshr_b32 s6, s0, 16
-; GFX11-FAKE16-NEXT: s_lshr_b32 s7, s0, 8
-; GFX11-FAKE16-NEXT: s_and_not1_b32 vcc_lo, exec_lo, s8
+; GFX11-FAKE16-NEXT: s_lshr_b32 s8, s0, 8
+; GFX11-FAKE16-NEXT: s_and_not1_b32 vcc_lo, exec_lo, s7
; GFX11-FAKE16-NEXT: s_cbranch_vccnz .LBB25_3
; GFX11-FAKE16-NEXT: .LBB25_2: ; %cmp.true
; GFX11-FAKE16-NEXT: s_add_u32 s0, s0, 3
@@ -2951,16 +2951,16 @@ define inreg <8 x i8> @bitcast_i64_to_v8i8_scalar(i64 inreg %a, i32 inreg %b) {
; GFX11-FAKE16-NEXT: s_lshr_b32 s3, s1, 24
; GFX11-FAKE16-NEXT: s_lshr_b32 s4, s1, 16
; GFX11-FAKE16-NEXT: s_lshr_b32 s5, s1, 8
-; GFX11-FAKE16-NEXT: s_lshr_b32 s7, s0, 8
+; GFX11-FAKE16-NEXT: s_lshr_b32 s8, s0, 8
; GFX11-FAKE16-NEXT: .LBB25_3: ; %end
; GFX11-FAKE16-NEXT: s_delay_alu instid0(SALU_CYCLE_1)
-; GFX11-FAKE16-NEXT: v_dual_mov_b32 v0, s0 :: v_dual_mov_b32 v1, s7
+; GFX11-FAKE16-NEXT: v_dual_mov_b32 v0, s0 :: v_dual_mov_b32 v1, s8
; GFX11-FAKE16-NEXT: v_dual_mov_b32 v2, s6 :: v_dual_mov_b32 v3, s2
; GFX11-FAKE16-NEXT: v_dual_mov_b32 v4, s1 :: v_dual_mov_b32 v5, s5
; GFX11-FAKE16-NEXT: v_dual_mov_b32 v6, s4 :: v_dual_mov_b32 v7, s3
; GFX11-FAKE16-NEXT: s_setpc_b64 s[30:31]
; GFX11-FAKE16-NEXT: .LBB25_4:
-; GFX11-FAKE16-NEXT: ; implicit-def: $sgpr7
+; GFX11-FAKE16-NEXT: ; implicit-def: $sgpr8
; GFX11-FAKE16-NEXT: ; implicit-def: $sgpr6
; GFX11-FAKE16-NEXT: ; implicit-def: $sgpr2
; GFX11-FAKE16-NEXT: ; implicit-def: $sgpr5
@@ -9107,7 +9107,7 @@ define inreg <8 x i8> @bitcast_v2i32_to_v8i8_scalar(<2 x i32> inreg %a, i32 inre
; GFX11-TRUE16: ; %bb.0:
; GFX11-TRUE16-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX11-TRUE16-NEXT: s_cmp_lg_u32 s2, 0
-; GFX11-TRUE16-NEXT: s_mov_b32 s8, 0
+; GFX11-TRUE16-NEXT: s_mov_b32 s7, 0
; GFX11-TRUE16-NEXT: s_cbranch_scc0 .LBB69_4
; GFX11-TRUE16-NEXT: ; %bb.1: ; %cmp.false
; GFX11-TRUE16-NEXT: s_lshr_b64 s[2:3], s[0:1], 24
@@ -9115,8 +9115,8 @@ define inreg <8 x i8> @bitcast_v2i32_to_v8i8_scalar(<2 x i32> inreg %a, i32 inre
; GFX11-TRUE16-NEXT: s_lshr_b32 s4, s1, 16
; GFX11-TRUE16-NEXT: s_lshr_b32 s5, s1, 8
; GFX11-TRUE16-NEXT: s_lshr_b32 s6, s0, 16
-; GFX11-TRUE16-NEXT: s_lshr_b32 s7, s0, 8
-; GFX11-TRUE16-NEXT: s_and_not1_b32 vcc_lo, exec_lo, s8
+; GFX11-TRUE16-NEXT: s_lshr_b32 s8, s0, 8
+; GFX11-TRUE16-NEXT: s_and_not1_b32 vcc_lo, exec_lo, s7
; GFX11-TRUE16-NEXT: s_cbranch_vccnz .LBB69_3
; GFX11-TRUE16-NEXT: .LBB69_2: ; %cmp.true
; GFX11-TRUE16-NEXT: s_add_i32 s1, s1, 3
@@ -9126,10 +9126,10 @@ define inreg <8 x i8> @bitcast_v2i32_to_v8i8_scalar(<2 x i32> inreg %a, i32 inre
; GFX11-TRUE16-NEXT: s_lshr_b32 s3, s1, 24
; GFX11-TRUE16-NEXT: s_lshr_b32 s5, s1, 8
; GFX11-TRUE16-NEXT: s_lshr_b32 s6, s0, 16
-; GFX11-TRUE16-NEXT: s_lshr_b32 s7, s0, 8
+; GFX11-TRUE16-NEXT: s_lshr_b32 s8, s0, 8
; GFX11-TRUE16-NEXT: .LBB69_3: ; %end
; GFX11-TRUE16-NEXT: v_mov_b16_e32 v0.l, s0
-; GFX11-TRUE16-NEXT: v_mov_b16_e32 v1.l, s7
+; GFX11-TRUE16-NEXT: v_mov_b16_e32 v1.l, s8
; GFX11-TRUE16-NEXT: v_mov_b16_e32 v2.l, s6
; GFX11-TRUE16-NEXT: v_mov_b16_e32 v3.l, s2
; GFX11-TRUE16-NEXT: v_mov_b16_e32 v4.l, s1
@@ -9138,7 +9138,7 @@ define inreg <8 x i8> @bitcast_v2i32_to_v8i8_scalar(<2 x i32> inreg %a, i32 inre
; GFX11-TRUE16-NEXT: v_mov_b16_e32 v7.l, s3
; GFX11-TRUE16-NEXT: s_setpc_b64 s[30:31]
; GFX11-TRUE16-NEXT: .LBB69_4:
-; GFX11-TRUE16-NEXT: ; implicit-def: $sgpr7_lo16
+; GFX11-TRUE16-NEXT: ; implicit-def: $sgpr8_lo16
; GFX11-TRUE16-NEXT: ; implicit-def: $sgpr6_lo16
; GFX11-TRUE16-NEXT: ; implicit-def: $sgpr2_lo16
; GFX11-TRUE16-NEXT: ; implicit-def: $sgpr5_lo16
@@ -9150,7 +9150,7 @@ define inreg <8 x i8> @bitcast_v2i32_to_v8i8_scalar(<2 x i32> inreg %a, i32 inre
; GFX11-FAKE16: ; %bb.0:
; GFX11-FAKE16-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX11-FAKE16-NEXT: s_cmp_lg_u32 s2, 0
-; GFX11-FAKE16-NEXT: s_mov_b32 s8, 0
+; GFX11-FAKE16-NEXT: s_mov_b32 s7, 0
; GFX11-FAKE16-NEXT: s_cbranch_scc0 .LBB69_4
; GFX11-FAKE16-NEXT: ; %bb.1: ; %cmp.false
; GFX11-FAKE16-NEXT: s_lshr_b64 s[2:3], s[0:1], 24
@@ -9158,8 +9158,8 @@ define inreg <8 x i8> @bitcast_v2i32_to_v8i8_scalar(<2 x i32> inreg %a, i32 inre
; GFX11-FAKE16-NEXT: s_lshr_b32 s4, s1, 16
; GFX11-FAKE16-NEXT: s_lshr_b32 s5, s1, 8
; GFX11-FAKE16-NEXT: s_lshr_b32 s6, s0, 16
-; GFX11-FAKE16-NEXT: s_lshr_b32 s7, s0, 8
-; GFX11-FAKE16-NEXT: s_and_not1_b32 vcc_lo, exec_lo, s8
+; GFX11-FAKE16-NEXT: s_lshr_b32 s8, s0, 8
+; GFX11-FAKE16-NEXT: s_and_not1_b32 vcc_lo, exec_lo, s7
; GFX11-FAKE16-NEXT: s_cbranch_vccnz .LBB69_3
; GFX11-FAKE16-NEXT: .LBB69_2: ; %cmp.true
; GFX11-FAKE16-NEXT: s_add_i32 s1, s1, 3
@@ -9169,16 +9169,16 @@ define inreg <8 x i8> @bitcast_v2i32_to_v8i8_scalar(<2 x i32> inreg %a, i32 inre
; GFX11-FAKE16-NEXT: s_lshr_b32 s3, s1, 24
; GFX11-FAKE16-NEXT: s_lshr_b32 s5, s1, 8
; GFX11-FAKE16-NEXT: s_lshr_b32 s6, s0, 16
-; GFX11-FAKE16-NEXT: s_lshr_b32 s7, s0, 8
+; GFX11-FAKE16-NEXT: s_lshr_b32 s8, s0, 8
; GFX11-FAKE16-NEXT: .LBB69_3: ; %end
; GFX11-FAKE16-NEXT: s_delay_alu instid0(SALU_CYCLE_1)
-; GFX11-FAKE16-NEXT: v_dual_mov_b32 v0, s0 :: v_dual_mov_b32 v1, s7
+; GFX11-FAKE16-NEXT: v_dual_mov_b32 v0, s0 :: v_dual_mov_b32 v1, s8
; GFX11-FAKE16-NEXT: v_dual_mov_b32 v2, s6 :: v_dual_mov_b32 v3, s2
; GFX11-FAKE16-NEXT: v_dual_mov_b32 v4, s1 :: v_dual_mov_b32 v5, s5
; GFX11-FAKE16-NEXT: v_dual_mov_b32 v6, s4 :: v_dual_mov_b32 v7, s3
; GFX11-FAKE16-NEXT: s_setpc_b64 s[30:31]
; GFX11-FAKE16-NEXT: .LBB69_4:
-; GFX11-FAKE16-NEXT: ; implicit-def: $sgpr7
+; GFX11-FAKE16-NEXT: ; implicit-def: $sgpr8
; GFX11-FAKE16-NEXT: ; implicit-def: $sgpr6
; GFX11-FAKE16-NEXT: ; implicit-def: $sgpr2
; GFX11-FAKE16-NEXT: ; implicit-def: $sgpr5
@@ -13747,47 +13747,45 @@ define <8 x i8> @bitcast_v4i16_to_v8i8(<4 x i16> %a, i32 %b) {
; VI-LABEL: bitcast_v4i16_to_v8i8:
; VI: ; %bb.0:
; VI-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; VI-NEXT: v_mov_b32_e32 v9, v1
+; VI-NEXT: v_mov_b32_e32 v8, v0
; VI-NEXT: v_cmp_ne_u32_e32 vcc, 0, v2
-; VI-NEXT: v_lshrrev_b32_e32 v6, 16, v1
-; VI-NEXT: v_lshrrev_b32_e32 v2, 16, v0
-; VI-NEXT: ; implicit-def: $vgpr9
-; VI-NEXT: ; implicit-def: $vgpr4
+; VI-NEXT: v_lshrrev_b32_e32 v6, 16, v9
+; VI-NEXT: v_lshrrev_b32_e32 v2, 16, v8
+; VI-NEXT: ; implicit-def: $vgpr1
; VI-NEXT: ; implicit-def: $vgpr3
-; VI-NEXT: ; implicit-def: $vgpr8
; VI-NEXT: ; implicit-def: $vgpr5
; VI-NEXT: ; implicit-def: $vgpr7
; VI-NEXT: s_and_saveexec_b64 s[4:5], vcc
; VI-NEXT: s_xor_b64 s[4:5], exec, s[4:5]
; VI-NEXT: ; %bb.1: ; %cmp.false
-; VI-NEXT: v_lshrrev_b64 v[3:4], 24, v[0:1]
-; VI-NEXT: v_lshrrev_b32_e32 v7, 24, v1
-; VI-NEXT: v_lshrrev_b32_e32 v5, 8, v1
-; VI-NEXT: v_lshrrev_b32_e32 v4, 8, v0
-; VI-NEXT: v_mov_b32_e32 v9, v0
-; VI-NEXT: v_mov_b32_e32 v8, v1
-; VI-NEXT: ; implicit-def: $vgpr0_vgpr1
+; VI-NEXT: v_lshrrev_b64 v[3:4], 24, v[8:9]
+; VI-NEXT: v_lshrrev_b32_e32 v7, 24, v9
+; VI-NEXT: v_lshrrev_b32_e32 v5, 8, v9
+; VI-NEXT: v_lshrrev_b32_e32 v1, 8, v8
; VI-NEXT: ; %bb.2: ; %Flow
; VI-NEXT: s_andn2_saveexec_b64 s[4:5], s[4:5]
; VI-NEXT: s_cbranch_execz .LBB96_4
; VI-NEXT: ; %bb.3: ; %cmp.true
-; VI-NEXT: v_mov_b32_e32 v2, 3
-; VI-NEXT: v_add_u16_sdwa v6, v1, v2 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_1 src1_sel:DWORD
-; VI-NEXT: v_add_u16_sdwa v2, v0, v2 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_1 src1_sel:DWORD
-; VI-NEXT: v_add_u16_e32 v8, 3, v1
+; VI-NEXT: v_mov_b32_e32 v0, 3
+; VI-NEXT: v_add_u16_sdwa v6, v9, v0 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_1 src1_sel:DWORD
+; VI-NEXT: v_add_u16_sdwa v2, v8, v0 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_1 src1_sel:DWORD
+; VI-NEXT: v_add_u16_e32 v10, 3, v9
; VI-NEXT: v_lshlrev_b32_e32 v1, 16, v6
-; VI-NEXT: v_add_u16_e32 v9, 3, v0
+; VI-NEXT: v_add_u16_e32 v9, 3, v8
; VI-NEXT: v_lshlrev_b32_e32 v0, 16, v2
-; VI-NEXT: v_or_b32_e32 v1, v8, v1
+; VI-NEXT: v_or_b32_e32 v1, v10, v1
; VI-NEXT: v_or_b32_e32 v0, v9, v0
; VI-NEXT: v_lshrrev_b64 v[3:4], 24, v[0:1]
; VI-NEXT: v_lshrrev_b32_e32 v5, 8, v1
-; VI-NEXT: v_lshrrev_b32_e32 v4, 8, v0
+; VI-NEXT: v_lshrrev_b32_e32 v1, 8, v0
; VI-NEXT: v_bfe_u32 v7, v6, 8, 8
+; VI-NEXT: v_mov_b32_e32 v8, v9
+; VI-NEXT: v_mov_b32_e32 v9, v10
; VI-NEXT: .LBB96_4: ; %end
; VI-NEXT: s_or_b64 exec, exec, s[4:5]
-; VI-NEXT: v_mov_b32_e32 v0, v9
-; VI-NEXT: v_mov_b32_e32 v1, v4
-; VI-NEXT: v_mov_b32_e32 v4, v8
+; VI-NEXT: v_mov_b32_e32 v0, v8
+; VI-NEXT: v_mov_b32_e32 v4, v9
; VI-NEXT: s_setpc_b64 s[30:31]
;
; GFX9-LABEL: bitcast_v4i16_to_v8i8:
@@ -13984,48 +13982,47 @@ define inreg <8 x i8> @bitcast_v4i16_to_v8i8_scalar(<4 x i16> inreg %a, i32 inre
; VI-NEXT: s_cbranch_scc0 .LBB97_4
; VI-NEXT: ; %bb.1: ; %cmp.false
; VI-NEXT: s_lshr_b64 s[4:5], s[16:17], 24
-; VI-NEXT: s_lshr_b32 s10, s17, 24
+; VI-NEXT: s_lshr_b32 s9, s17, 24
; VI-NEXT: s_lshr_b32 s8, s17, 16
; VI-NEXT: s_lshr_b32 s5, s17, 8
-; VI-NEXT: s_lshr_b32 s11, s16, 16
-; VI-NEXT: s_lshr_b32 s12, s16, 8
-; VI-NEXT: s_mov_b32 s9, s17
+; VI-NEXT: s_lshr_b32 s10, s16, 16
+; VI-NEXT: s_lshr_b32 s11, s16, 8
; VI-NEXT: s_cbranch_execnz .LBB97_3
; VI-NEXT: .LBB97_2: ; %cmp.true
-; VI-NEXT: s_lshr_b32 s5, s17, 16
-; VI-NEXT: s_add_i32 s9, s17, 3
-; VI-NEXT: s_add_i32 s8, s5, 3
-; VI-NEXT: s_and_b32 s4, s9, 0xffff
-; VI-NEXT: s_lshl_b32 s5, s8, 16
-; VI-NEXT: s_or_b32 s7, s4, s5
-; VI-NEXT: s_and_b32 s4, s16, 0xffff0000
+; VI-NEXT: s_lshr_b32 s6, s17, 16
+; VI-NEXT: s_add_i32 s4, s17, 3
+; VI-NEXT: s_add_i32 s8, s6, 3
+; VI-NEXT: s_and_b32 s5, s4, 0xffff
+; VI-NEXT: s_lshl_b32 s6, s8, 16
+; VI-NEXT: s_or_b32 s7, s5, s6
+; VI-NEXT: s_and_b32 s5, s16, 0xffff0000
; VI-NEXT: s_add_i32 s16, s16, 3
-; VI-NEXT: s_and_b32 s5, s16, 0xffff
-; VI-NEXT: s_or_b32 s4, s4, s5
-; VI-NEXT: s_add_i32 s6, s4, 0x30000
+; VI-NEXT: s_and_b32 s6, s16, 0xffff
+; VI-NEXT: s_or_b32 s5, s5, s6
+; VI-NEXT: s_add_i32 s6, s5, 0x30000
+; VI-NEXT: s_mov_b32 s17, s4
; VI-NEXT: s_lshr_b64 s[4:5], s[6:7], 24
; VI-NEXT: s_lshr_b32 s5, s7, 8
-; VI-NEXT: s_lshr_b32 s11, s6, 16
-; VI-NEXT: s_lshr_b32 s12, s6, 8
-; VI-NEXT: s_bfe_u32 s10, s8, 0x80008
+; VI-NEXT: s_lshr_b32 s10, s6, 16
+; VI-NEXT: s_lshr_b32 s11, s6, 8
+; VI-NEXT: s_bfe_u32 s9, s8, 0x80008
; VI-NEXT: .LBB97_3: ; %end
; VI-NEXT: v_mov_b32_e32 v0, s16
-; VI-NEXT: v_mov_b32_e32 v1, s12
-; VI-NEXT: v_mov_b32_e32 v2, s11
+; VI-NEXT: v_mov_b32_e32 v1, s11
+; VI-NEXT: v_mov_b32_e32 v2, s10
; VI-NEXT: v_mov_b32_e32 v3, s4
-; VI-NEXT: v_mov_b32_e32 v4, s9
+; VI-NEXT: v_mov_b32_e32 v4, s17
; VI-NEXT: v_mov_b32_e32 v5, s5
; VI-NEXT: v_mov_b32_e32 v6, s8
-; VI-NEXT: v_mov_b32_e32 v7, s10
+; VI-NEXT: v_mov_b32_e32 v7, s9
; VI-NEXT: s_setpc_b64 s[30:31]
; VI-NEXT: .LBB97_4:
-; VI-NEXT: ; implicit-def: $sgpr12
; VI-NEXT: ; implicit-def: $sgpr11
+; VI-NEXT: ; implicit-def: $sgpr10
; VI-NEXT: ; implicit-def: $sgpr4
-; VI-NEXT: ; implicit-def: $sgpr9
; VI-NEXT: ; implicit-def: $sgpr5
; VI-NEXT: ; implicit-def: $sgpr8
-; VI-NEXT: ; implicit-def: $sgpr10
+; VI-NEXT: ; implicit-def: $sgpr9
; VI-NEXT: s_branch .LBB97_2
;
; GFX9-LABEL: bitcast_v4i16_to_v8i8_scalar:
@@ -16967,77 +16964,78 @@ define <8 x i8> @bitcast_v4bf16_to_v8i8(<4 x bfloat> %a, i32 %b) {
; GFX11-TRUE16-LABEL: bitcast_v4bf16_to_v8i8:
; GFX11-TRUE16: ; %bb.0:
; GFX11-TRUE16-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
-; GFX11-TRUE16-NEXT: v_dual_mov_b32 v9, v1 :: v_dual_mov_b32 v8, v0
-; GFX11-TRUE16-NEXT: s_mov_b32 s0, exec_lo
-; GFX11-TRUE16-NEXT: ; implicit-def: $vgpr0_hi16
-; GFX11-TRUE16-NEXT: ; implicit-def: $vgpr1_lo16
-; GFX11-TRUE16-NEXT: ; implicit-def: $vgpr10_lo16
+; GFX11-TRUE16-NEXT: v_cmp_ne_u32_e32 vcc_lo, 0, v2
+; GFX11-TRUE16-NEXT: ; implicit-def: $vgpr2_lo16
+; GFX11-TRUE16-NEXT: ; implicit-def: $vgpr10_hi16
+; GFX11-TRUE16-NEXT: ; implicit-def: $vgpr8_lo16
; GFX11-TRUE16-NEXT: ; implicit-def: $vgpr4_hi16
; GFX11-TRUE16-NEXT: ; implicit-def: $vgpr5_lo16
; GFX11-TRUE16-NEXT: ; implicit-def: $vgpr6_hi16
; GFX11-TRUE16-NEXT: ; implicit-def: $vgpr7_lo16
-; GFX11-TRUE16-NEXT: v_cmpx_ne_u32_e32 0, v2
+; GFX11-TRUE16-NEXT: s_and_saveexec_b32 s0, vcc_lo
+; GFX11-TRUE16-NEXT: s_delay_alu instid0(SALU_CYCLE_1)
; GFX11-TRUE16-NEXT: s_xor_b32 s0, exec_lo, s0
; GFX11-TRUE16-NEXT: s_cbranch_execz .LBB108_2
; GFX11-TRUE16-NEXT: ; %bb.1: ; %cmp.false
-; GFX11-TRUE16-NEXT: v_lshrrev_b64 v[10:11], 24, v[8:9]
-; GFX11-TRUE16-NEXT: v_lshrrev_b32_e32 v7, 24, v9
-; GFX11-TRUE16-NEXT: v_lshrrev_b32_e32 v5, 8, v9
-; GFX11-TRUE16-NEXT: v_lshrrev_b32_e32 v1, 8, v8
-; GFX11-TRUE16-NEXT: v_mov_b16_e32 v0.h, v8.l
-; GFX11-TRUE16-NEXT: v_mov_b16_e32 v4.h, v9.l
-; GFX11-TRUE16-NEXT: v_mov_b16_e32 v6.h, v9.h
+; GFX11-TRUE16-NEXT: v_lshrrev_b64 v[8:9], 24, v[0:1]
+; GFX11-TRUE16-NEXT: v_lshrrev_b32_e32 v7, 24, v1
+; GFX11-TRUE16-NEXT: v_lshrrev_b32_e32 v5, 8, v1
+; GFX11-TRUE16-NEXT: v_lshrrev_b32_e32 v2, 8, v0
+; GFX11-TRUE16-NEXT: v_mov_b16_e32 v6.h, v1.h
+; GFX11-TRUE16-NEXT: v_mov_b16_e32 v10.h, v0.h
+; GFX11-TRUE16-NEXT: v_mov_b16_e32 v4.h, v1.l
; GFX11-TRUE16-NEXT: .LBB108_2: ; %Flow
; GFX11-TRUE16-NEXT: s_and_not1_saveexec_b32 s0, s0
; GFX11-TRUE16-NEXT: s_cbranch_execz .LBB108_4
; GFX11-TRUE16-NEXT: ; %bb.3: ; %cmp.true
-; GFX11-TRUE16-NEXT: v_and_b32_e32 v0, 0xffff0000, v8
-; GFX11-TRUE16-NEXT: v_mov_b16_e32 v1.l, v8.l
-; GFX11-TRUE16-NEXT: v_and_b32_e32 v2, 0xffff0000, v9
-; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_2) | instskip(SKIP_1) | instid1(VALU_DEP_2)
-; GFX11-TRUE16-NEXT: v_dual_add_f32 v5, 0x40c00000, v0 :: v_dual_lshlrev_b32 v0, 16, v1
-; GFX11-TRUE16-NEXT: v_lshlrev_b32_e32 v3, 16, v9
-; GFX11-TRUE16-NEXT: v_or_b32_e32 v6, 0x400000, v5
-; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_3) | instskip(NEXT) | instid1(VALU_DEP_3)
-; GFX11-TRUE16-NEXT: v_dual_add_f32 v0, 0x40c00000, v0 :: v_dual_add_f32 v1, 0x40c00000, v2
-; GFX11-TRUE16-NEXT: v_add_f32_e32 v2, 0x40c00000, v3
-; GFX11-TRUE16-NEXT: v_bfe_u32 v3, v5, 16, 1
-; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_3) | instskip(NEXT) | instid1(VALU_DEP_4)
+; GFX11-TRUE16-NEXT: v_and_b32_e32 v2, 0xffff0000, v0
+; GFX11-TRUE16-NEXT: v_and_b32_e32 v3, 0xffff0000, v1
+; GFX11-TRUE16-NEXT: v_lshlrev_b32_e32 v1, 16, v1
+; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
+; GFX11-TRUE16-NEXT: v_dual_add_f32 v2, 0x40c00000, v2 :: v_dual_add_f32 v1, 0x40c00000, v1
+; GFX11-TRUE16-NEXT: v_bfe_u32 v4, v2, 16, 1
+; GFX11-TRUE16-NEXT: v_or_b32_e32 v5, 0x400000, v2
+; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_3) | instskip(SKIP_3) | instid1(VALU_DEP_4)
+; GFX11-TRUE16-NEXT: v_bfe_u32 v7, v1, 16, 1
+; GFX11-TRUE16-NEXT: v_or_b32_e32 v10, 0x400000, v1
+; GFX11-TRUE16-NEXT: v_cmp_u_f32_e32 vcc_lo, v1, v1
+; GFX11-TRUE16-NEXT: v_add3_u32 v11, v4, v2, 0x7fff
+; GFX11-TRUE16-NEXT: v_add3_u32 v4, v7, v1, 0x7fff
+; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(SKIP_1) | instid1(VALU_DEP_1)
+; GFX11-TRUE16-NEXT: v_cndmask_b32_e32 v4, v4, v10, vcc_lo
+; GFX11-TRUE16-NEXT: v_dual_add_f32 v3, 0x40c00000, v3 :: v_dual_lshlrev_b32 v0, 16, v0
+; GFX11-TRUE16-NEXT: v_add_f32_e32 v0, 0x40c00000, v0
+; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_2) | instskip(SKIP_1) | instid1(VALU_DEP_3)
+; GFX11-TRUE16-NEXT: v_bfe_u32 v6, v3, 16, 1
+; GFX11-TRUE16-NEXT: v_or_b32_e32 v8, 0x400000, v3
; GFX11-TRUE16-NEXT: v_bfe_u32 v9, v0, 16, 1
-; GFX11-TRUE16-NEXT: v_bfe_u32 v4, v1, 16, 1
-; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_4)
-; GFX11-TRUE16-NEXT: v_bfe_u32 v7, v2, 16, 1
-; GFX11-TRUE16-NEXT: v_or_b32_e32 v10, 0x400000, v2
-; GFX11-TRUE16-NEXT: v_cmp_u_f32_e32 vcc_lo, v2, v2
-; GFX11-TRUE16-NEXT: v_or_b32_e32 v11, 0x400000, v0
-; GFX11-TRUE16-NEXT: v_add3_u32 v9, v9, v0, 0x7fff
-; GFX11-TRUE16-NEXT: v_add3_u32 v7, v7, v2, 0x7fff
-; GFX11-TRUE16-NEXT: v_add3_u32 v12, v4, v1, 0x7fff
-; GFX11-TRUE16-NEXT: v_add3_u32 v3, v3, v5, 0x7fff
-; GFX11-TRUE16-NEXT: v_or_b32_e32 v8, 0x400000, v1
-; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_4)
-; GFX11-TRUE16-NEXT: v_cndmask_b32_e32 v4, v7, v10, vcc_lo
+; GFX11-TRUE16-NEXT: v_or_b32_e32 v7, 0x400000, v0
; GFX11-TRUE16-NEXT: v_cmp_u_f32_e32 vcc_lo, v0, v0
-; GFX11-TRUE16-NEXT: v_cndmask_b32_e32 v0, v9, v11, vcc_lo
-; GFX11-TRUE16-NEXT: v_cmp_u_f32_e32 vcc_lo, v5, v5
-; GFX11-TRUE16-NEXT: v_cndmask_b32_e32 v2, v3, v6, vcc_lo
-; GFX11-TRUE16-NEXT: v_cmp_u_f32_e32 vcc_lo, v1, v1
-; GFX11-TRUE16-NEXT: v_mov_b16_e32 v1.l, v4.h
+; GFX11-TRUE16-NEXT: v_add3_u32 v6, v6, v3, 0x7fff
+; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_4) | instskip(NEXT) | instid1(VALU_DEP_1)
+; GFX11-TRUE16-NEXT: v_add3_u32 v9, v9, v0, 0x7fff
+; GFX11-TRUE16-NEXT: v_cndmask_b32_e32 v0, v9, v7, vcc_lo
+; GFX11-TRUE16-NEXT: v_cmp_u_f32_e32 vcc_lo, v2, v2
+; GFX11-TRUE16-NEXT: v_mov_b16_e32 v2.l, v4.h
+; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_3) | instskip(SKIP_4) | instid1(VALU_DEP_2)
+; GFX11-TRUE16-NEXT: v_mov_b16_e32 v0.l, v0.h
+; GFX11-TRUE16-NEXT: v_cndmask_b32_e32 v1, v11, v5, vcc_lo
+; GFX11-TRUE16-NEXT: v_cmp_u_f32_e32 vcc_lo, v3, v3
; GFX11-TRUE16-NEXT: v_mov_b16_e32 v3.l, v0.h
-; GFX11-TRUE16-NEXT: v_cndmask_b32_e32 v6, v12, v8, vcc_lo
-; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_2) | instskip(NEXT) | instid1(VALU_DEP_2)
-; GFX11-TRUE16-NEXT: v_bfi_b32 v8, 0xffff, v3, v2
-; GFX11-TRUE16-NEXT: v_bfi_b32 v9, 0xffff, v1, v6
+; GFX11-TRUE16-NEXT: v_cndmask_b32_e32 v6, v6, v8, vcc_lo
+; GFX11-TRUE16-NEXT: v_bfi_b32 v10, 0xffff, v3, v1
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_2) | instskip(NEXT) | instid1(VALU_DEP_2)
-; GFX11-TRUE16-NEXT: v_lshrrev_b32_e32 v1, 8, v8
-; GFX11-TRUE16-NEXT: v_lshrrev_b64 v[10:11], 24, v[8:9]
-; GFX11-TRUE16-NEXT: v_lshrrev_b32_e32 v7, 24, v9
-; GFX11-TRUE16-NEXT: v_lshrrev_b32_e32 v5, 8, v9
+; GFX11-TRUE16-NEXT: v_bfi_b32 v11, 0xffff, v2, v6
+; GFX11-TRUE16-NEXT: v_lshrrev_b32_e32 v2, 8, v10
+; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_2)
+; GFX11-TRUE16-NEXT: v_lshrrev_b64 v[8:9], 24, v[10:11]
+; GFX11-TRUE16-NEXT: v_lshrrev_b32_e32 v7, 24, v11
+; GFX11-TRUE16-NEXT: v_lshrrev_b32_e32 v5, 8, v11
; GFX11-TRUE16-NEXT: .LBB108_4: ; %end
; GFX11-TRUE16-NEXT: s_or_b32 exec_lo, exec_lo, s0
-; GFX11-TRUE16-NEXT: v_mov_b16_e32 v0.l, v0.h
-; GFX11-TRUE16-NEXT: v_mov_b16_e32 v2.l, v8.h
-; GFX11-TRUE16-NEXT: v_mov_b16_e32 v3.l, v10.l
+; GFX11-TRUE16-NEXT: v_mov_b16_e32 v1.l, v2.l
+; GFX11-TRUE16-NEXT: v_mov_b16_e32 v2.l, v10.h
+; GFX11-TRUE16-NEXT: v_mov_b16_e32 v3.l, v8.l
; GFX11-TRUE16-NEXT: v_mov_b16_e32 v4.l, v4.h
; GFX11-TRUE16-NEXT: v_mov_b16_e32 v6.l, v6.h
; GFX11-TRUE16-NEXT: s_setpc_b64 s[30:31]
diff --git a/llvm/test/CodeGen/AMDGPU/amdgcn.bitcast.704bit.ll b/llvm/test/CodeGen/AMDGPU/amdgcn.bitcast.704bit.ll
index 2a96722ccce0b..51ed72a3a16fe 100644
--- a/llvm/test/CodeGen/AMDGPU/amdgcn.bitcast.704bit.ll
+++ b/llvm/test/CodeGen/AMDGPU/amdgcn.bitcast.704bit.ll
@@ -3642,7 +3642,7 @@ define inreg <44 x i16> @bitcast_v22i32_to_v44i16_scalar(<22 x i32> inreg %a, i3
; GFX11-TRUE16-NEXT: v_readfirstlane_b32 s5, v1
; GFX11-TRUE16-NEXT: v_readfirstlane_b32 s6, v2
; GFX11-TRUE16-NEXT: v_readfirstlane_b32 s4, v3
-; GFX11-TRUE16-NEXT: s_mov_b32 s62, 0
+; GFX11-TRUE16-NEXT: s_mov_b32 s58, 0
; GFX11-TRUE16-NEXT: s_and_b32 s8, vcc_lo, exec_lo
; GFX11-TRUE16-NEXT: s_cbranch_scc0 .LBB13_4
; GFX11-TRUE16-NEXT: ; %bb.1: ; %cmp.false
@@ -3664,11 +3664,11 @@ define inreg <44 x i16> @bitcast_v22i32_to_v44i16_scalar(<22 x i32> inreg %a, i3
; GFX11-TRUE16-NEXT: s_lshr_b32 s47, s18, 16
; GFX11-TRUE16-NEXT: s_lshr_b32 s56, s17, 16
; GFX11-TRUE16-NEXT: s_lshr_b32 s57, s16, 16
-; GFX11-TRUE16-NEXT: s_lshr_b32 s58, s3, 16
-; GFX11-TRUE16-NEXT: s_lshr_b32 s59, s2, 16
-; GFX11-TRUE16-NEXT: s_lshr_b32 s60, s1, 16
-; GFX11-TRUE16-NEXT: s_lshr_b32 s61, s0, 16
-; GFX11-TRUE16-NEXT: s_and_not1_b32 vcc_lo, exec_lo, s62
+; GFX11-TRUE16-NEXT: s_lshr_b32 s59, s3, 16
+; GFX11-TRUE16-NEXT: s_lshr_b32 s60, s2, 16
+; GFX11-TRUE16-NEXT: s_lshr_b32 s61, s1, 16
+; GFX11-TRUE16-NEXT: s_lshr_b32 s62, s0, 16
+; GFX11-TRUE16-NEXT: s_and_not1_b32 vcc_lo, exec_lo, s58
; GFX11-TRUE16-NEXT: s_cbranch_vccnz .LBB13_3
; GFX11-TRUE16-NEXT: .LBB13_2: ; %cmp.true
; GFX11-TRUE16-NEXT: s_add_i32 s4, s4, 3
@@ -3711,19 +3711,22 @@ define inreg <44 x i16> @bitcast_v22i32_to_v44i16_scalar(<22 x i32> inreg %a, i3
; GFX11-TRUE16-NEXT: s_lshr_b32 s47, s18, 16
; GFX11-TRUE16-NEXT: s_lshr_b32 s56, s17, 16
; GFX11-TRUE16-NEXT: s_lshr_b32 s57, s16, 16
-; GFX11-TRUE16-NEXT: s_lshr_b32 s58, s3, 16
-; GFX11-TRUE16-NEXT: s_lshr_b32 s59, s2, 16
-; GFX11-TRUE16-NEXT: s_lshr_b32 s60, s1, 16
-; GFX11-TRUE16-NEXT: s_lshr_b32 s61, s0, 16
+; GFX11-TRUE16-NEXT: s_lshr_b32 s59, s3, 16
+; GFX11-TRUE16-NEXT: s_lshr_b32 s60, s2, 16
+; GFX11-TRUE16-NEXT: s_lshr_b32 s61, s1, 16
+; GFX11-TRUE16-NEXT: s_lshr_b32 s62, s0, 16
; GFX11-TRUE16-NEXT: .LBB13_3: ; %end
+; GFX11-TRUE16-NEXT: s_delay_alu instid0(SALU_CYCLE_1)
+; GFX11-TRUE16-NEXT: s_mov_b32 s58, s62
; GFX11-TRUE16-NEXT: s_pack_ll_b32_b16 s15, s26, s15
+; GFX11-TRUE16-NEXT: s_pack_ll_b32_b16 s0, s0, s58
+; GFX11-TRUE16-NEXT: s_mov_b32 s58, s60
; GFX11-TRUE16-NEXT: s_pack_ll_b32_b16 s14, s27, s14
; GFX11-TRUE16-NEXT: s_mov_b32 s26, s28
; GFX11-TRUE16-NEXT: s_mov_b32 s27, s29
-; GFX11-TRUE16-NEXT: s_pack_ll_b32_b16 s0, s0, s61
-; GFX11-TRUE16-NEXT: s_pack_ll_b32_b16 s1, s1, s60
-; GFX11-TRUE16-NEXT: s_pack_ll_b32_b16 s2, s2, s59
-; GFX11-TRUE16-NEXT: s_pack_ll_b32_b16 s3, s3, s58
+; GFX11-TRUE16-NEXT: s_pack_ll_b32_b16 s1, s1, s61
+; GFX11-TRUE16-NEXT: s_pack_ll_b32_b16 s2, s2, s58
+; GFX11-TRUE16-NEXT: s_pack_ll_b32_b16 s3, s3, s59
; GFX11-TRUE16-NEXT: s_pack_ll_b32_b16 s16, s16, s57
; GFX11-TRUE16-NEXT: s_pack_ll_b32_b16 s17, s17, s56
; GFX11-TRUE16-NEXT: s_pack_ll_b32_b16 s18, s18, s47
@@ -3753,10 +3756,10 @@ define inreg <44 x i16> @bitcast_v22i32_to_v44i16_scalar(<22 x i32> inreg %a, i3
; GFX11-TRUE16-NEXT: v_dual_mov_b32 v20, s6 :: v_dual_mov_b32 v21, s4
; GFX11-TRUE16-NEXT: s_setpc_b64 s[30:31]
; GFX11-TRUE16-NEXT: .LBB13_4:
+; GFX11-TRUE16-NEXT: ; implicit-def: $sgpr62_lo16
; GFX11-TRUE16-NEXT: ; implicit-def: $sgpr61_lo16
; GFX11-TRUE16-NEXT: ; implicit-def: $sgpr60_lo16
; GFX11-TRUE16-NEXT: ; implicit-def: $sgpr59_lo16
-; GFX11-TRUE16-NEXT: ; implicit-def: $sgpr58_lo16
; GFX11-TRUE16-NEXT: ; implicit-def: $sgpr57_lo16
; GFX11-TRUE16-NEXT: ; implicit-def: $sgpr56_lo16
; GFX11-TRUE16-NEXT: ; implicit-def: $sgpr47_lo16
@@ -7383,7 +7386,7 @@ define inreg <44 x half> @bitcast_v22i32_to_v44f16_scalar(<22 x i32> inreg %a, i
; GFX11-TRUE16-NEXT: v_readfirstlane_b32 s5, v1
; GFX11-TRUE16-NEXT: v_readfirstlane_b32 s6, v2
; GFX11-TRUE16-NEXT: v_readfirstlane_b32 s4, v3
-; GFX11-TRUE16-NEXT: s_mov_b32 s62, 0
+; GFX11-TRUE16-NEXT: s_mov_b32 s58, 0
; GFX11-TRUE16-NEXT: s_and_b32 s8, vcc_lo, exec_lo
; GFX11-TRUE16-NEXT: s_cbranch_scc0 .LBB17_4
; GFX11-TRUE16-NEXT: ; %bb.1: ; %cmp.false
@@ -7405,11 +7408,11 @@ define inreg <44 x half> @bitcast_v22i32_to_v44f16_scalar(<22 x i32> inreg %a, i
; GFX11-TRUE16-NEXT: s_lshr_b32 s47, s18, 16
; GFX11-TRUE16-NEXT: s_lshr_b32 s56, s17, 16
; GFX11-TRUE16-NEXT: s_lshr_b32 s57, s16, 16
-; GFX11-TRUE16-NEXT: s_lshr_b32 s58, s3, 16
-; GFX11-TRUE16-NEXT: s_lshr_b32 s59, s2, 16
-; GFX11-TRUE16-NEXT: s_lshr_b32 s60, s1, 16
-; GFX11-TRUE16-NEXT: s_lshr_b32 s61, s0, 16
-; GFX11-TRUE16-NEXT: s_and_not1_b32 vcc_lo, exec_lo, s62
+; GFX11-TRUE16-NEXT: s_lshr_b32 s59, s3, 16
+; GFX11-TRUE16-NEXT: s_lshr_b32 s60, s2, 16
+; GFX11-TRUE16-NEXT: s_lshr_b32 s61, s1, 16
+; GFX11-TRUE16-NEXT: s_lshr_b32 s62, s0, 16
+; GFX11-TRUE16-NEXT: s_and_not1_b32 vcc_lo, exec_lo, s58
; GFX11-TRUE16-NEXT: s_cbranch_vccnz .LBB17_3
; GFX11-TRUE16-NEXT: .LBB17_2: ; %cmp.true
; GFX11-TRUE16-NEXT: s_add_i32 s4, s4, 3
@@ -7452,19 +7455,22 @@ define inreg <44 x half> @bitcast_v22i32_to_v44f16_scalar(<22 x i32> inreg %a, i
; GFX11-TRUE16-NEXT: s_lshr_b32 s47, s18, 16
; GFX11-TRUE16-NEXT: s_lshr_b32 s56, s17, 16
; GFX11-TRUE16-NEXT: s_lshr_b32 s57, s16, 16
-; GFX11-TRUE16-NEXT: s_lshr_b32 s58, s3, 16
-; GFX11-TRUE16-NEXT: s_lshr_b32 s59, s2, 16
-; GFX11-TRUE16-NEXT: s_lshr_b32 s60, s1, 16
-; GFX11-TRUE16-NEXT: s_lshr_b32 s61, s0, 16
+; GFX11-TRUE16-NEXT: s_lshr_b32 s59, s3, 16
+; GFX11-TRUE16-NEXT: s_lshr_b32 s60, s2, 16
+; GFX11-TRUE16-NEXT: s_lshr_b32 s61, s1, 16
+; GFX11-TRUE16-NEXT: s_lshr_b32 s62, s0, 16
; GFX11-TRUE16-NEXT: .LBB17_3: ; %end
+; GFX11-TRUE16-NEXT: s_delay_alu instid0(SALU_CYCLE_1)
+; GFX11-TRUE16-NEXT: s_mov_b32 s58, s62
; GFX11-TRUE16-NEXT: s_pack_ll_b32_b16 s15, s26, s15
+; GFX11-TRUE16-NEXT: s_pack_ll_b32_b16 s0, s0, s58
+; GFX11-TRUE16-NEXT: s_mov_b32 s58, s60
; GFX11-TRUE16-NEXT: s_pack_ll_b32_b16 s14, s27, s14
; GFX11-TRUE16-NEXT: s_mov_b32 s26, s28
; GFX11-TRUE16-NEXT: s_mov_b32 s27, s29
-; GFX11-TRUE16-NEXT: s_pack_ll_b32_b16 s0, s0, s61
-; GFX11-TRUE16-NEXT: s_pack_ll_b32_b16 s1, s1, s60
-; GFX11-TRUE16-NEXT: s_pack_ll_b32_b16 s2, s2, s59
-; GFX11-TRUE16-NEXT: s_pack_ll_b32_b16 s3, s3, s58
+; GFX11-TRUE16-NEXT: s_pack_ll_b32_b16 s1, s1, s61
+; GFX11-TRUE16-NEXT: s_pack_ll_b32_b16 s2, s2, s58
+; GFX11-TRUE16-NEXT: s_pack_ll_b32_b16 s3, s3, s59
; GFX11-TRUE16-NEXT: s_pack_ll_b32_b16 s16, s16, s57
; GFX11-TRUE16-NEXT: s_pack_ll_b32_b16 s17, s17, s56
; GFX11-TRUE16-NEXT: s_pack_ll_b32_b16 s18, s18, s47
@@ -7494,10 +7500,10 @@ define inreg <44 x half> @bitcast_v22i32_to_v44f16_scalar(<22 x i32> inreg %a, i
; GFX11-TRUE16-NEXT: v_dual_mov_b32 v20, s6 :: v_dual_mov_b32 v21, s4
; GFX11-TRUE16-NEXT: s_setpc_b64 s[30:31]
; GFX11-TRUE16-NEXT: .LBB17_4:
+; GFX11-TRUE16-NEXT: ; implicit-def: $sgpr62_lo16
; GFX11-TRUE16-NEXT: ; implicit-def: $sgpr61_lo16
; GFX11-TRUE16-NEXT: ; implicit-def: $sgpr60_lo16
; GFX11-TRUE16-NEXT: ; implicit-def: $sgpr59_lo16
-; GFX11-TRUE16-NEXT: ; implicit-def: $sgpr58_lo16
; GFX11-TRUE16-NEXT: ; implicit-def: $sgpr57_lo16
; GFX11-TRUE16-NEXT: ; implicit-def: $sgpr56_lo16
; GFX11-TRUE16-NEXT: ; implicit-def: $sgpr47_lo16
@@ -8775,13 +8781,13 @@ define inreg <22 x i32> @bitcast_v44f16_to_v22i32_scalar(<44 x half> inreg %a, i
; SI-NEXT: buffer_store_dword v61, off, s[0:3], s32 offset:8 ; 4-byte Folded Spill
; SI-NEXT: buffer_store_dword v62, off, s[0:3], s32 offset:4 ; 4-byte Folded Spill
; SI-NEXT: buffer_store_dword v63, off, s[0:3], s32 ; 4-byte Folded Spill
+; SI-NEXT: v_cvt_f16_f32_e32 v39, v1
; SI-NEXT: s_waitcnt expcnt(3)
-; SI-NEXT: v_cvt_f16_f32_e32 v60, v1
-; SI-NEXT: v_cvt_f16_f32_e32 v48, v0
-; SI-NEXT: v_cvt_f16_f32_e32 v39, v3
+; SI-NEXT: v_cvt_f16_f32_e32 v60, v0
+; SI-NEXT: v_cvt_f16_f32_e32 v48, v3
; SI-NEXT: v_cvt_f16_f32_e32 v38, v2
; SI-NEXT: v_cvt_f16_f32_e32 v37, v5
-; SI-NEXT: v_cvt_f16_f32_e32 v49, v4
+; SI-NEXT: v_cvt_f16_f32_e32 v36, v4
; SI-NEXT: v_cvt_f16_f32_e32 v35, v7
; SI-NEXT: v_cvt_f16_f32_e32 v50, v6
; SI-NEXT: v_cvt_f16_f32_e32 v32, v9
@@ -8809,37 +8815,36 @@ define inreg <22 x i32> @bitcast_v44f16_to_v22i32_scalar(<44 x half> inreg %a, i
; SI-NEXT: v_cvt_f16_f32_e32 v51, s17
; SI-NEXT: v_cvt_f16_f32_e32 v33, s16
; SI-NEXT: v_cvt_f16_f32_e32 v1, s19
-; SI-NEXT: v_cvt_f16_f32_e32 v34, s18
-; SI-NEXT: v_cvt_f16_f32_e32 v2, s21
-; SI-NEXT: v_cvt_f16_f32_e32 v3, s20
+; SI-NEXT: v_cvt_f16_f32_e32 v2, s18
+; SI-NEXT: v_cvt_f16_f32_e32 v3, s21
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_cvt_f16_f32_e32 v63, s23
-; SI-NEXT: v_cvt_f16_f32_e32 v62, s22
+; SI-NEXT: v_cvt_f16_f32_e32 v63, s20
+; SI-NEXT: v_cvt_f16_f32_e32 v62, s23
+; SI-NEXT: v_cvt_f16_f32_e32 v34, s22
; SI-NEXT: v_cvt_f16_f32_e32 v61, s25
-; SI-NEXT: v_cvt_f16_f32_e32 v36, s24
+; SI-NEXT: v_cvt_f16_f32_e32 v49, s24
; SI-NEXT: v_cvt_f16_f32_e32 v29, s27
; SI-NEXT: v_cvt_f16_f32_e32 v28, s26
; SI-NEXT: v_cvt_f16_f32_e32 v27, s29
; SI-NEXT: v_cvt_f16_f32_e32 v26, s28
; SI-NEXT: v_cmp_ne_u32_e32 vcc, 0, v30
; SI-NEXT: s_and_b64 s[4:5], vcc, exec
-; SI-NEXT: buffer_store_dword v3, off, s[0:3], s32 offset:64 ; 4-byte Folded Spill
-; SI-NEXT: buffer_store_dword v2, off, s[0:3], s32 offset:68 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v2, off, s[0:3], s32 offset:64 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v3, off, s[0:3], s32 offset:68 ; 4-byte Folded Spill
; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:72 ; 4-byte Folded Spill
; SI-NEXT: s_cbranch_scc0 .LBB19_4
; SI-NEXT: ; %bb.1: ; %cmp.false
-; SI-NEXT: s_waitcnt expcnt(1)
-; SI-NEXT: v_lshlrev_b32_e32 v2, 16, v2
-; SI-NEXT: v_lshlrev_b32_e32 v0, 16, v51
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_lshlrev_b32_e32 v1, 16, v1
-; SI-NEXT: v_or_b32_e32 v2, v3, v2
-; SI-NEXT: v_lshlrev_b32_e32 v3, 16, v63
+; SI-NEXT: v_lshlrev_b32_e32 v0, 16, v51
+; SI-NEXT: v_or_b32_e32 v1, v2, v1
+; SI-NEXT: v_lshlrev_b32_e32 v2, 16, v3
+; SI-NEXT: v_lshlrev_b32_e32 v3, 16, v62
; SI-NEXT: v_lshlrev_b32_e32 v4, 16, v61
; SI-NEXT: v_lshlrev_b32_e32 v5, 16, v29
; SI-NEXT: v_lshlrev_b32_e32 v6, 16, v27
-; SI-NEXT: v_lshlrev_b32_e32 v7, 16, v60
-; SI-NEXT: v_lshlrev_b32_e32 v8, 16, v39
+; SI-NEXT: v_lshlrev_b32_e32 v7, 16, v39
+; SI-NEXT: v_lshlrev_b32_e32 v8, 16, v48
; SI-NEXT: v_lshlrev_b32_e32 v9, 16, v37
; SI-NEXT: v_lshlrev_b32_e32 v10, 16, v35
; SI-NEXT: v_lshlrev_b32_e32 v11, 16, v32
@@ -8854,14 +8859,14 @@ define inreg <22 x i32> @bitcast_v44f16_to_v22i32_scalar(<44 x half> inreg %a, i
; SI-NEXT: v_lshlrev_b32_e32 v20, 16, v25
; SI-NEXT: v_lshlrev_b32_e32 v21, 16, v22
; SI-NEXT: v_or_b32_e32 v0, v33, v0
-; SI-NEXT: v_or_b32_e32 v1, v34, v1
-; SI-NEXT: v_or_b32_e32 v3, v62, v3
-; SI-NEXT: v_or_b32_e32 v4, v36, v4
+; SI-NEXT: v_or_b32_e32 v2, v63, v2
+; SI-NEXT: v_or_b32_e32 v3, v34, v3
+; SI-NEXT: v_or_b32_e32 v4, v49, v4
; SI-NEXT: v_or_b32_e32 v5, v28, v5
; SI-NEXT: v_or_b32_e32 v6, v26, v6
-; SI-NEXT: v_or_b32_e32 v7, v48, v7
+; SI-NEXT: v_or_b32_e32 v7, v60, v7
; SI-NEXT: v_or_b32_e32 v8, v38, v8
-; SI-NEXT: v_or_b32_e32 v9, v49, v9
+; SI-NEXT: v_or_b32_e32 v9, v36, v9
; SI-NEXT: v_or_b32_e32 v10, v50, v10
; SI-NEXT: v_or_b32_e32 v11, v59, v11
; SI-NEXT: v_or_b32_e32 v12, v57, v12
@@ -8877,65 +8882,64 @@ define inreg <22 x i32> @bitcast_v44f16_to_v22i32_scalar(<44 x half> inreg %a, i
; SI-NEXT: s_cbranch_execnz .LBB19_3
; SI-NEXT: .LBB19_2: ; %cmp.true
; SI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:72 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v3, off, s[0:3], s32 offset:64 ; 4-byte Folded Reload
; SI-NEXT: v_cvt_f32_f16_e32 v0, v51
; SI-NEXT: v_cvt_f32_f16_e32 v1, v33
-; SI-NEXT: v_cvt_f32_f16_e32 v3, v34
-; SI-NEXT: v_cvt_f32_f16_e32 v4, v62
+; SI-NEXT: s_waitcnt vmcnt(10)
+; SI-NEXT: v_cvt_f32_f16_e32 v4, v34
+; SI-NEXT: v_cvt_f32_f16_e32 v5, v49
; SI-NEXT: v_add_f32_e32 v0, 0x38000000, v0
; SI-NEXT: v_add_f32_e32 v1, 0x38000000, v1
; SI-NEXT: v_cvt_f16_f32_e32 v0, v0
-; SI-NEXT: v_add_f32_e32 v3, 0x38000000, v3
; SI-NEXT: v_cvt_f16_f32_e32 v1, v1
-; SI-NEXT: v_cvt_f16_f32_e32 v3, v3
-; SI-NEXT: v_lshlrev_b32_e32 v0, 16, v0
; SI-NEXT: v_add_f32_e32 v4, 0x38000000, v4
-; SI-NEXT: v_or_b32_e32 v0, v1, v0
; SI-NEXT: v_cvt_f16_f32_e32 v4, v4
-; SI-NEXT: v_cvt_f32_f16_e32 v5, v36
-; SI-NEXT: v_cvt_f32_f16_e32 v6, v28
-; SI-NEXT: v_cvt_f32_f16_e32 v7, v26
-; SI-NEXT: s_waitcnt vmcnt(1)
-; SI-NEXT: v_cvt_f32_f16_e32 v8, v48
+; SI-NEXT: v_lshlrev_b32_e32 v0, 16, v0
+; SI-NEXT: v_or_b32_e32 v0, v1, v0
; SI-NEXT: v_add_f32_e32 v5, 0x38000000, v5
; SI-NEXT: v_cvt_f16_f32_e32 v5, v5
+; SI-NEXT: v_cvt_f32_f16_e32 v6, v28
+; SI-NEXT: v_cvt_f32_f16_e32 v7, v26
+; SI-NEXT: v_cvt_f32_f16_e32 v8, v60
+; SI-NEXT: s_waitcnt vmcnt(3)
+; SI-NEXT: v_cvt_f32_f16_e32 v9, v38
; SI-NEXT: v_add_f32_e32 v6, 0x38000000, v6
; SI-NEXT: v_cvt_f16_f32_e32 v6, v6
; SI-NEXT: v_add_f32_e32 v7, 0x38000000, v7
; SI-NEXT: v_cvt_f16_f32_e32 v7, v7
; SI-NEXT: v_add_f32_e32 v8, 0x38000000, v8
; SI-NEXT: v_cvt_f16_f32_e32 v8, v8
-; SI-NEXT: v_cvt_f32_f16_e32 v9, v38
; SI-NEXT: v_cvt_f32_f16_e32 v10, v37
-; SI-NEXT: v_cvt_f32_f16_e32 v11, v49
-; SI-NEXT: v_cvt_f32_f16_e32 v12, v50
; SI-NEXT: v_add_f32_e32 v9, 0x38000000, v9
-; SI-NEXT: v_add_f32_e32 v10, 0x38000000, v10
; SI-NEXT: v_cvt_f16_f32_e32 v9, v9
+; SI-NEXT: v_cvt_f32_f16_e32 v11, v36
+; SI-NEXT: v_add_f32_e32 v10, 0x38000000, v10
; SI-NEXT: v_cvt_f16_f32_e32 v10, v10
+; SI-NEXT: v_cvt_f32_f16_e32 v12, v50
; SI-NEXT: v_add_f32_e32 v11, 0x38000000, v11
; SI-NEXT: v_cvt_f16_f32_e32 v11, v11
+; SI-NEXT: v_cvt_f32_f16_e32 v13, v58
; SI-NEXT: v_add_f32_e32 v12, 0x38000000, v12
; SI-NEXT: v_cvt_f16_f32_e32 v12, v12
-; SI-NEXT: v_cvt_f32_f16_e32 v13, v58
; SI-NEXT: v_cvt_f32_f16_e32 v14, v57
-; SI-NEXT: v_cvt_f32_f16_e32 v15, v47
-; SI-NEXT: v_cvt_f32_f16_e32 v16, v44
; SI-NEXT: v_add_f32_e32 v13, 0x38000000, v13
; SI-NEXT: v_cvt_f16_f32_e32 v13, v13
+; SI-NEXT: v_cvt_f32_f16_e32 v15, v47
; SI-NEXT: v_add_f32_e32 v14, 0x38000000, v14
; SI-NEXT: v_cvt_f16_f32_e32 v14, v14
+; SI-NEXT: v_cvt_f32_f16_e32 v16, v44
; SI-NEXT: v_add_f32_e32 v15, 0x38000000, v15
; SI-NEXT: v_cvt_f16_f32_e32 v15, v15
+; SI-NEXT: v_cvt_f32_f16_e32 v17, v43
; SI-NEXT: v_add_f32_e32 v16, 0x38000000, v16
; SI-NEXT: v_cvt_f16_f32_e32 v16, v16
-; SI-NEXT: v_cvt_f32_f16_e32 v17, v43
; SI-NEXT: v_cvt_f32_f16_e32 v18, v41
-; SI-NEXT: v_cvt_f32_f16_e32 v19, v54
-; SI-NEXT: v_cvt_f32_f16_e32 v20, v53
; SI-NEXT: v_add_f32_e32 v17, 0x38000000, v17
; SI-NEXT: v_cvt_f16_f32_e32 v17, v17
+; SI-NEXT: v_cvt_f32_f16_e32 v19, v54
; SI-NEXT: v_add_f32_e32 v18, 0x38000000, v18
; SI-NEXT: v_cvt_f16_f32_e32 v18, v18
+; SI-NEXT: v_cvt_f32_f16_e32 v20, v53
; SI-NEXT: v_add_f32_e32 v19, 0x38000000, v19
; SI-NEXT: v_cvt_f16_f32_e32 v19, v19
; SI-NEXT: v_cvt_f32_f16_e32 v21, v31
@@ -8949,25 +8953,27 @@ define inreg <22 x i32> @bitcast_v44f16_to_v22i32_scalar(<44 x half> inreg %a, i
; SI-NEXT: v_cvt_f16_f32_e32 v22, v22
; SI-NEXT: v_add_f32_e32 v23, 0x38000000, v23
; SI-NEXT: v_cvt_f16_f32_e32 v23, v23
-; SI-NEXT: s_waitcnt vmcnt(0)
+; SI-NEXT: s_waitcnt vmcnt(1)
; SI-NEXT: v_cvt_f32_f16_e32 v2, v2
+; SI-NEXT: s_waitcnt vmcnt(0)
+; SI-NEXT: v_cvt_f32_f16_e32 v3, v3
; SI-NEXT: v_add_f32_e32 v2, 0x38000000, v2
; SI-NEXT: v_cvt_f16_f32_e32 v2, v2
+; SI-NEXT: v_add_f32_e32 v3, 0x38000000, v3
+; SI-NEXT: v_cvt_f16_f32_e32 v3, v3
; SI-NEXT: v_lshlrev_b32_e32 v2, 16, v2
; SI-NEXT: v_or_b32_e32 v1, v3, v2
; SI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:68 ; 4-byte Folded Reload
-; SI-NEXT: buffer_load_dword v3, off, s[0:3], s32 offset:64 ; 4-byte Folded Reload
-; SI-NEXT: s_waitcnt vmcnt(1)
-; SI-NEXT: v_cvt_f32_f16_e32 v2, v2
+; SI-NEXT: v_cvt_f32_f16_e32 v3, v63
+; SI-NEXT: v_add_f32_e32 v3, 0x38000000, v3
+; SI-NEXT: v_cvt_f16_f32_e32 v3, v3
; SI-NEXT: s_waitcnt vmcnt(0)
-; SI-NEXT: v_cvt_f32_f16_e32 v3, v3
+; SI-NEXT: v_cvt_f32_f16_e32 v2, v2
; SI-NEXT: v_add_f32_e32 v2, 0x38000000, v2
; SI-NEXT: v_cvt_f16_f32_e32 v2, v2
-; SI-NEXT: v_add_f32_e32 v3, 0x38000000, v3
-; SI-NEXT: v_cvt_f16_f32_e32 v3, v3
; SI-NEXT: v_lshlrev_b32_e32 v2, 16, v2
; SI-NEXT: v_or_b32_e32 v2, v3, v2
-; SI-NEXT: v_cvt_f32_f16_e32 v3, v63
+; SI-NEXT: v_cvt_f32_f16_e32 v3, v62
; SI-NEXT: v_add_f32_e32 v3, 0x38000000, v3
; SI-NEXT: v_cvt_f16_f32_e32 v3, v3
; SI-NEXT: v_lshlrev_b32_e32 v3, 16, v3
@@ -8987,12 +8993,12 @@ define inreg <22 x i32> @bitcast_v44f16_to_v22i32_scalar(<44 x half> inreg %a, i
; SI-NEXT: v_cvt_f16_f32_e32 v6, v6
; SI-NEXT: v_lshlrev_b32_e32 v6, 16, v6
; SI-NEXT: v_or_b32_e32 v6, v7, v6
-; SI-NEXT: v_cvt_f32_f16_e32 v7, v60
+; SI-NEXT: v_cvt_f32_f16_e32 v7, v39
; SI-NEXT: v_add_f32_e32 v7, 0x38000000, v7
; SI-NEXT: v_cvt_f16_f32_e32 v7, v7
; SI-NEXT: v_lshlrev_b32_e32 v7, 16, v7
; SI-NEXT: v_or_b32_e32 v7, v8, v7
-; SI-NEXT: v_cvt_f32_f16_e32 v8, v39
+; SI-NEXT: v_cvt_f32_f16_e32 v8, v48
; SI-NEXT: v_add_f32_e32 v8, 0x38000000, v8
; SI-NEXT: v_cvt_f16_f32_e32 v8, v8
; SI-NEXT: v_lshlrev_b32_e32 v8, 16, v8
@@ -9080,86 +9086,89 @@ define inreg <22 x i32> @bitcast_v44f16_to_v22i32_scalar(<44 x half> inreg %a, i
; SI-NEXT: s_setpc_b64 s[30:31]
; SI-NEXT: .LBB19_4:
; SI-NEXT: buffer_store_dword v48, off, s[0:3], s32 offset:76 ; 4-byte Folded Spill
-; SI-NEXT: buffer_store_dword v39, off, s[0:3], s32 offset:80 ; 4-byte Folded Spill
-; SI-NEXT: buffer_store_dword v38, off, s[0:3], s32 offset:84 ; 4-byte Folded Spill
-; SI-NEXT: buffer_store_dword v37, off, s[0:3], s32 offset:88 ; 4-byte Folded Spill
-; SI-NEXT: buffer_store_dword v49, off, s[0:3], s32 offset:92 ; 4-byte Folded Spill
-; SI-NEXT: buffer_store_dword v35, off, s[0:3], s32 offset:96 ; 4-byte Folded Spill
-; SI-NEXT: buffer_store_dword v50, off, s[0:3], s32 offset:100 ; 4-byte Folded Spill
-; SI-NEXT: buffer_store_dword v32, off, s[0:3], s32 offset:104 ; 4-byte Folded Spill
-; SI-NEXT: buffer_store_dword v59, off, s[0:3], s32 offset:108 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v38, off, s[0:3], s32 offset:80 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v37, off, s[0:3], s32 offset:84 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v36, off, s[0:3], s32 offset:88 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v35, off, s[0:3], s32 offset:92 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v50, off, s[0:3], s32 offset:96 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v32, off, s[0:3], s32 offset:100 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v59, off, s[0:3], s32 offset:104 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_mov_b32_e32 v59, v46
; SI-NEXT: v_mov_b32_e32 v46, v41
; SI-NEXT: v_mov_b32_e32 v41, v52
; SI-NEXT: v_mov_b32_e32 v52, v23
-; SI-NEXT: v_mov_b32_e32 v48, v60
+; SI-NEXT: v_mov_b32_e32 v48, v39
+; SI-NEXT: v_mov_b32_e32 v39, v60
; SI-NEXT: v_mov_b32_e32 v60, v47
; SI-NEXT: v_mov_b32_e32 v47, v42
; SI-NEXT: v_mov_b32_e32 v42, v53
; SI-NEXT: v_mov_b32_e32 v53, v22
-; SI-NEXT: v_mov_b32_e32 v35, v61
+; SI-NEXT: buffer_store_dword v34, off, s[0:3], s32 offset:108 ; 4-byte Folded Spill
+; SI-NEXT: s_waitcnt expcnt(0)
+; SI-NEXT: v_mov_b32_e32 v34, v61
; SI-NEXT: v_mov_b32_e32 v61, v56
; SI-NEXT: v_mov_b32_e32 v56, v43
; SI-NEXT: v_mov_b32_e32 v43, v54
; SI-NEXT: v_mov_b32_e32 v54, v24
-; SI-NEXT: v_mov_b32_e32 v50, v34
-; SI-NEXT: v_mov_b32_e32 v34, v62
+; SI-NEXT: v_mov_b32_e32 v50, v62
; SI-NEXT: v_mov_b32_e32 v62, v57
; SI-NEXT: v_mov_b32_e32 v57, v44
; SI-NEXT: v_mov_b32_e32 v44, v55
; SI-NEXT: v_mov_b32_e32 v55, v25
-; SI-NEXT: v_mov_b32_e32 v32, v33
+; SI-NEXT: v_mov_b32_e32 v32, v51
+; SI-NEXT: v_mov_b32_e32 v51, v33
; SI-NEXT: v_mov_b32_e32 v33, v63
; SI-NEXT: v_mov_b32_e32 v63, v58
; SI-NEXT: v_mov_b32_e32 v58, v45
; SI-NEXT: v_mov_b32_e32 v45, v40
; SI-NEXT: v_mov_b32_e32 v40, v31
-; SI-NEXT: v_mov_b32_e32 v39, v26
-; SI-NEXT: v_mov_b32_e32 v38, v27
-; SI-NEXT: v_mov_b32_e32 v37, v28
-; SI-NEXT: v_mov_b32_e32 v49, v36
-; SI-NEXT: v_mov_b32_e32 v36, v29
+; SI-NEXT: v_mov_b32_e32 v38, v26
+; SI-NEXT: v_mov_b32_e32 v37, v27
+; SI-NEXT: v_mov_b32_e32 v36, v28
+; SI-NEXT: v_mov_b32_e32 v35, v49
+; SI-NEXT: v_mov_b32_e32 v49, v29
; SI-NEXT: ; implicit-def: $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7_vgpr8_vgpr9_vgpr10_vgpr11_vgpr12_vgpr13_vgpr14_vgpr15_vgpr16_vgpr17_vgpr18_vgpr19_vgpr20_vgpr21_vgpr22_vgpr23_vgpr24_vgpr25_vgpr26_vgpr27_vgpr28_vgpr29_vgpr30_vgpr31
-; SI-NEXT: v_mov_b32_e32 v29, v36
-; SI-NEXT: v_mov_b32_e32 v36, v49
-; SI-NEXT: v_mov_b32_e32 v28, v37
-; SI-NEXT: v_mov_b32_e32 v27, v38
-; SI-NEXT: v_mov_b32_e32 v26, v39
+; SI-NEXT: v_mov_b32_e32 v29, v49
+; SI-NEXT: v_mov_b32_e32 v49, v35
+; SI-NEXT: v_mov_b32_e32 v28, v36
+; SI-NEXT: v_mov_b32_e32 v27, v37
+; SI-NEXT: v_mov_b32_e32 v26, v38
; SI-NEXT: v_mov_b32_e32 v31, v40
; SI-NEXT: v_mov_b32_e32 v40, v45
; SI-NEXT: v_mov_b32_e32 v45, v58
; SI-NEXT: v_mov_b32_e32 v58, v63
; SI-NEXT: v_mov_b32_e32 v63, v33
-; SI-NEXT: v_mov_b32_e32 v33, v32
+; SI-NEXT: v_mov_b32_e32 v33, v51
+; SI-NEXT: v_mov_b32_e32 v51, v32
; SI-NEXT: v_mov_b32_e32 v25, v55
; SI-NEXT: v_mov_b32_e32 v55, v44
; SI-NEXT: v_mov_b32_e32 v44, v57
; SI-NEXT: v_mov_b32_e32 v57, v62
-; SI-NEXT: v_mov_b32_e32 v62, v34
-; SI-NEXT: v_mov_b32_e32 v34, v50
+; SI-NEXT: v_mov_b32_e32 v62, v50
; SI-NEXT: v_mov_b32_e32 v24, v54
; SI-NEXT: v_mov_b32_e32 v54, v43
; SI-NEXT: v_mov_b32_e32 v43, v56
; SI-NEXT: v_mov_b32_e32 v56, v61
-; SI-NEXT: v_mov_b32_e32 v61, v35
+; SI-NEXT: v_mov_b32_e32 v61, v34
+; SI-NEXT: buffer_load_dword v34, off, s[0:3], s32 offset:108 ; 4-byte Folded Reload
; SI-NEXT: v_mov_b32_e32 v22, v53
; SI-NEXT: v_mov_b32_e32 v53, v42
; SI-NEXT: v_mov_b32_e32 v42, v47
; SI-NEXT: v_mov_b32_e32 v47, v60
-; SI-NEXT: v_mov_b32_e32 v60, v48
+; SI-NEXT: v_mov_b32_e32 v60, v39
+; SI-NEXT: v_mov_b32_e32 v39, v48
; SI-NEXT: v_mov_b32_e32 v23, v52
; SI-NEXT: v_mov_b32_e32 v52, v41
; SI-NEXT: v_mov_b32_e32 v41, v46
; SI-NEXT: v_mov_b32_e32 v46, v59
-; SI-NEXT: buffer_load_dword v59, off, s[0:3], s32 offset:108 ; 4-byte Folded Reload
-; SI-NEXT: buffer_load_dword v32, off, s[0:3], s32 offset:104 ; 4-byte Folded Reload
-; SI-NEXT: buffer_load_dword v50, off, s[0:3], s32 offset:100 ; 4-byte Folded Reload
-; SI-NEXT: buffer_load_dword v35, off, s[0:3], s32 offset:96 ; 4-byte Folded Reload
-; SI-NEXT: buffer_load_dword v49, off, s[0:3], s32 offset:92 ; 4-byte Folded Reload
-; SI-NEXT: buffer_load_dword v37, off, s[0:3], s32 offset:88 ; 4-byte Folded Reload
-; SI-NEXT: buffer_load_dword v38, off, s[0:3], s32 offset:84 ; 4-byte Folded Reload
-; SI-NEXT: buffer_load_dword v39, off, s[0:3], s32 offset:80 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v59, off, s[0:3], s32 offset:104 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v32, off, s[0:3], s32 offset:100 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v50, off, s[0:3], s32 offset:96 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v35, off, s[0:3], s32 offset:92 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v36, off, s[0:3], s32 offset:88 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v37, off, s[0:3], s32 offset:84 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v38, off, s[0:3], s32 offset:80 ; 4-byte Folded Reload
; SI-NEXT: buffer_load_dword v48, off, s[0:3], s32 offset:76 ; 4-byte Folded Reload
; SI-NEXT: s_branch .LBB19_2
;
@@ -17665,13 +17674,13 @@ define inreg <22 x float> @bitcast_v44f16_to_v22f32_scalar(<44 x half> inreg %a,
; SI-NEXT: buffer_store_dword v61, off, s[0:3], s32 offset:8 ; 4-byte Folded Spill
; SI-NEXT: buffer_store_dword v62, off, s[0:3], s32 offset:4 ; 4-byte Folded Spill
; SI-NEXT: buffer_store_dword v63, off, s[0:3], s32 ; 4-byte Folded Spill
+; SI-NEXT: v_cvt_f16_f32_e32 v39, v1
; SI-NEXT: s_waitcnt expcnt(3)
-; SI-NEXT: v_cvt_f16_f32_e32 v60, v1
-; SI-NEXT: v_cvt_f16_f32_e32 v48, v0
-; SI-NEXT: v_cvt_f16_f32_e32 v39, v3
+; SI-NEXT: v_cvt_f16_f32_e32 v60, v0
+; SI-NEXT: v_cvt_f16_f32_e32 v48, v3
; SI-NEXT: v_cvt_f16_f32_e32 v38, v2
; SI-NEXT: v_cvt_f16_f32_e32 v37, v5
-; SI-NEXT: v_cvt_f16_f32_e32 v49, v4
+; SI-NEXT: v_cvt_f16_f32_e32 v36, v4
; SI-NEXT: v_cvt_f16_f32_e32 v35, v7
; SI-NEXT: v_cvt_f16_f32_e32 v50, v6
; SI-NEXT: v_cvt_f16_f32_e32 v32, v9
@@ -17699,37 +17708,36 @@ define inreg <22 x float> @bitcast_v44f16_to_v22f32_scalar(<44 x half> inreg %a,
; SI-NEXT: v_cvt_f16_f32_e32 v51, s17
; SI-NEXT: v_cvt_f16_f32_e32 v33, s16
; SI-NEXT: v_cvt_f16_f32_e32 v1, s19
-; SI-NEXT: v_cvt_f16_f32_e32 v34, s18
-; SI-NEXT: v_cvt_f16_f32_e32 v2, s21
-; SI-NEXT: v_cvt_f16_f32_e32 v3, s20
+; SI-NEXT: v_cvt_f16_f32_e32 v2, s18
+; SI-NEXT: v_cvt_f16_f32_e32 v3, s21
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_cvt_f16_f32_e32 v63, s23
-; SI-NEXT: v_cvt_f16_f32_e32 v62, s22
+; SI-NEXT: v_cvt_f16_f32_e32 v63, s20
+; SI-NEXT: v_cvt_f16_f32_e32 v62, s23
+; SI-NEXT: v_cvt_f16_f32_e32 v34, s22
; SI-NEXT: v_cvt_f16_f32_e32 v61, s25
-; SI-NEXT: v_cvt_f16_f32_e32 v36, s24
+; SI-NEXT: v_cvt_f16_f32_e32 v49, s24
; SI-NEXT: v_cvt_f16_f32_e32 v29, s27
; SI-NEXT: v_cvt_f16_f32_e32 v28, s26
; SI-NEXT: v_cvt_f16_f32_e32 v27, s29
; SI-NEXT: v_cvt_f16_f32_e32 v26, s28
; SI-NEXT: v_cmp_ne_u32_e32 vcc, 0, v30
; SI-NEXT: s_and_b64 s[4:5], vcc, exec
-; SI-NEXT: buffer_store_dword v3, off, s[0:3], s32 offset:64 ; 4-byte Folded Spill
-; SI-NEXT: buffer_store_dword v2, off, s[0:3], s32 offset:68 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v2, off, s[0:3], s32 offset:64 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v3, off, s[0:3], s32 offset:68 ; 4-byte Folded Spill
; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:72 ; 4-byte Folded Spill
; SI-NEXT: s_cbranch_scc0 .LBB35_4
; SI-NEXT: ; %bb.1: ; %cmp.false
-; SI-NEXT: s_waitcnt expcnt(1)
-; SI-NEXT: v_lshlrev_b32_e32 v2, 16, v2
-; SI-NEXT: v_lshlrev_b32_e32 v0, 16, v51
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_lshlrev_b32_e32 v1, 16, v1
-; SI-NEXT: v_or_b32_e32 v2, v3, v2
-; SI-NEXT: v_lshlrev_b32_e32 v3, 16, v63
+; SI-NEXT: v_lshlrev_b32_e32 v0, 16, v51
+; SI-NEXT: v_or_b32_e32 v1, v2, v1
+; SI-NEXT: v_lshlrev_b32_e32 v2, 16, v3
+; SI-NEXT: v_lshlrev_b32_e32 v3, 16, v62
; SI-NEXT: v_lshlrev_b32_e32 v4, 16, v61
; SI-NEXT: v_lshlrev_b32_e32 v5, 16, v29
; SI-NEXT: v_lshlrev_b32_e32 v6, 16, v27
-; SI-NEXT: v_lshlrev_b32_e32 v7, 16, v60
-; SI-NEXT: v_lshlrev_b32_e32 v8, 16, v39
+; SI-NEXT: v_lshlrev_b32_e32 v7, 16, v39
+; SI-NEXT: v_lshlrev_b32_e32 v8, 16, v48
; SI-NEXT: v_lshlrev_b32_e32 v9, 16, v37
; SI-NEXT: v_lshlrev_b32_e32 v10, 16, v35
; SI-NEXT: v_lshlrev_b32_e32 v11, 16, v32
@@ -17744,14 +17752,14 @@ define inreg <22 x float> @bitcast_v44f16_to_v22f32_scalar(<44 x half> inreg %a,
; SI-NEXT: v_lshlrev_b32_e32 v20, 16, v25
; SI-NEXT: v_lshlrev_b32_e32 v21, 16, v22
; SI-NEXT: v_or_b32_e32 v0, v33, v0
-; SI-NEXT: v_or_b32_e32 v1, v34, v1
-; SI-NEXT: v_or_b32_e32 v3, v62, v3
-; SI-NEXT: v_or_b32_e32 v4, v36, v4
+; SI-NEXT: v_or_b32_e32 v2, v63, v2
+; SI-NEXT: v_or_b32_e32 v3, v34, v3
+; SI-NEXT: v_or_b32_e32 v4, v49, v4
; SI-NEXT: v_or_b32_e32 v5, v28, v5
; SI-NEXT: v_or_b32_e32 v6, v26, v6
-; SI-NEXT: v_or_b32_e32 v7, v48, v7
+; SI-NEXT: v_or_b32_e32 v7, v60, v7
; SI-NEXT: v_or_b32_e32 v8, v38, v8
-; SI-NEXT: v_or_b32_e32 v9, v49, v9
+; SI-NEXT: v_or_b32_e32 v9, v36, v9
; SI-NEXT: v_or_b32_e32 v10, v50, v10
; SI-NEXT: v_or_b32_e32 v11, v59, v11
; SI-NEXT: v_or_b32_e32 v12, v57, v12
@@ -17767,65 +17775,64 @@ define inreg <22 x float> @bitcast_v44f16_to_v22f32_scalar(<44 x half> inreg %a,
; SI-NEXT: s_cbranch_execnz .LBB35_3
; SI-NEXT: .LBB35_2: ; %cmp.true
; SI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:72 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v3, off, s[0:3], s32 offset:64 ; 4-byte Folded Reload
; SI-NEXT: v_cvt_f32_f16_e32 v0, v51
; SI-NEXT: v_cvt_f32_f16_e32 v1, v33
-; SI-NEXT: v_cvt_f32_f16_e32 v3, v34
-; SI-NEXT: v_cvt_f32_f16_e32 v4, v62
+; SI-NEXT: s_waitcnt vmcnt(10)
+; SI-NEXT: v_cvt_f32_f16_e32 v4, v34
+; SI-NEXT: v_cvt_f32_f16_e32 v5, v49
; SI-NEXT: v_add_f32_e32 v0, 0x38000000, v0
; SI-NEXT: v_add_f32_e32 v1, 0x38000000, v1
; SI-NEXT: v_cvt_f16_f32_e32 v0, v0
-; SI-NEXT: v_add_f32_e32 v3, 0x38000000, v3
; SI-NEXT: v_cvt_f16_f32_e32 v1, v1
-; SI-NEXT: v_cvt_f16_f32_e32 v3, v3
-; SI-NEXT: v_lshlrev_b32_e32 v0, 16, v0
; SI-NEXT: v_add_f32_e32 v4, 0x38000000, v4
-; SI-NEXT: v_or_b32_e32 v0, v1, v0
; SI-NEXT: v_cvt_f16_f32_e32 v4, v4
-; SI-NEXT: v_cvt_f32_f16_e32 v5, v36
-; SI-NEXT: v_cvt_f32_f16_e32 v6, v28
-; SI-NEXT: v_cvt_f32_f16_e32 v7, v26
-; SI-NEXT: s_waitcnt vmcnt(1)
-; SI-NEXT: v_cvt_f32_f16_e32 v8, v48
+; SI-NEXT: v_lshlrev_b32_e32 v0, 16, v0
+; SI-NEXT: v_or_b32_e32 v0, v1, v0
; SI-NEXT: v_add_f32_e32 v5, 0x38000000, v5
; SI-NEXT: v_cvt_f16_f32_e32 v5, v5
+; SI-NEXT: v_cvt_f32_f16_e32 v6, v28
+; SI-NEXT: v_cvt_f32_f16_e32 v7, v26
+; SI-NEXT: v_cvt_f32_f16_e32 v8, v60
+; SI-NEXT: s_waitcnt vmcnt(3)
+; SI-NEXT: v_cvt_f32_f16_e32 v9, v38
; SI-NEXT: v_add_f32_e32 v6, 0x38000000, v6
; SI-NEXT: v_cvt_f16_f32_e32 v6, v6
; SI-NEXT: v_add_f32_e32 v7, 0x38000000, v7
; SI-NEXT: v_cvt_f16_f32_e32 v7, v7
; SI-NEXT: v_add_f32_e32 v8, 0x38000000, v8
; SI-NEXT: v_cvt_f16_f32_e32 v8, v8
-; SI-NEXT: v_cvt_f32_f16_e32 v9, v38
; SI-NEXT: v_cvt_f32_f16_e32 v10, v37
-; SI-NEXT: v_cvt_f32_f16_e32 v11, v49
-; SI-NEXT: v_cvt_f32_f16_e32 v12, v50
; SI-NEXT: v_add_f32_e32 v9, 0x38000000, v9
-; SI-NEXT: v_add_f32_e32 v10, 0x38000000, v10
; SI-NEXT: v_cvt_f16_f32_e32 v9, v9
+; SI-NEXT: v_cvt_f32_f16_e32 v11, v36
+; SI-NEXT: v_add_f32_e32 v10, 0x38000000, v10
; SI-NEXT: v_cvt_f16_f32_e32 v10, v10
+; SI-NEXT: v_cvt_f32_f16_e32 v12, v50
; SI-NEXT: v_add_f32_e32 v11, 0x38000000, v11
; SI-NEXT: v_cvt_f16_f32_e32 v11, v11
+; SI-NEXT: v_cvt_f32_f16_e32 v13, v58
; SI-NEXT: v_add_f32_e32 v12, 0x38000000, v12
; SI-NEXT: v_cvt_f16_f32_e32 v12, v12
-; SI-NEXT: v_cvt_f32_f16_e32 v13, v58
; SI-NEXT: v_cvt_f32_f16_e32 v14, v57
-; SI-NEXT: v_cvt_f32_f16_e32 v15, v47
-; SI-NEXT: v_cvt_f32_f16_e32 v16, v44
; SI-NEXT: v_add_f32_e32 v13, 0x38000000, v13
; SI-NEXT: v_cvt_f16_f32_e32 v13, v13
+; SI-NEXT: v_cvt_f32_f16_e32 v15, v47
; SI-NEXT: v_add_f32_e32 v14, 0x38000000, v14
; SI-NEXT: v_cvt_f16_f32_e32 v14, v14
+; SI-NEXT: v_cvt_f32_f16_e32 v16, v44
; SI-NEXT: v_add_f32_e32 v15, 0x38000000, v15
; SI-NEXT: v_cvt_f16_f32_e32 v15, v15
+; SI-NEXT: v_cvt_f32_f16_e32 v17, v43
; SI-NEXT: v_add_f32_e32 v16, 0x38000000, v16
; SI-NEXT: v_cvt_f16_f32_e32 v16, v16
-; SI-NEXT: v_cvt_f32_f16_e32 v17, v43
; SI-NEXT: v_cvt_f32_f16_e32 v18, v41
-; SI-NEXT: v_cvt_f32_f16_e32 v19, v54
-; SI-NEXT: v_cvt_f32_f16_e32 v20, v53
; SI-NEXT: v_add_f32_e32 v17, 0x38000000, v17
; SI-NEXT: v_cvt_f16_f32_e32 v17, v17
+; SI-NEXT: v_cvt_f32_f16_e32 v19, v54
; SI-NEXT: v_add_f32_e32 v18, 0x38000000, v18
; SI-NEXT: v_cvt_f16_f32_e32 v18, v18
+; SI-NEXT: v_cvt_f32_f16_e32 v20, v53
; SI-NEXT: v_add_f32_e32 v19, 0x38000000, v19
; SI-NEXT: v_cvt_f16_f32_e32 v19, v19
; SI-NEXT: v_cvt_f32_f16_e32 v21, v31
@@ -17839,25 +17846,27 @@ define inreg <22 x float> @bitcast_v44f16_to_v22f32_scalar(<44 x half> inreg %a,
; SI-NEXT: v_cvt_f16_f32_e32 v22, v22
; SI-NEXT: v_add_f32_e32 v23, 0x38000000, v23
; SI-NEXT: v_cvt_f16_f32_e32 v23, v23
-; SI-NEXT: s_waitcnt vmcnt(0)
+; SI-NEXT: s_waitcnt vmcnt(1)
; SI-NEXT: v_cvt_f32_f16_e32 v2, v2
+; SI-NEXT: s_waitcnt vmcnt(0)
+; SI-NEXT: v_cvt_f32_f16_e32 v3, v3
; SI-NEXT: v_add_f32_e32 v2, 0x38000000, v2
; SI-NEXT: v_cvt_f16_f32_e32 v2, v2
+; SI-NEXT: v_add_f32_e32 v3, 0x38000000, v3
+; SI-NEXT: v_cvt_f16_f32_e32 v3, v3
; SI-NEXT: v_lshlrev_b32_e32 v2, 16, v2
; SI-NEXT: v_or_b32_e32 v1, v3, v2
; SI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:68 ; 4-byte Folded Reload
-; SI-NEXT: buffer_load_dword v3, off, s[0:3], s32 offset:64 ; 4-byte Folded Reload
-; SI-NEXT: s_waitcnt vmcnt(1)
-; SI-NEXT: v_cvt_f32_f16_e32 v2, v2
+; SI-NEXT: v_cvt_f32_f16_e32 v3, v63
+; SI-NEXT: v_add_f32_e32 v3, 0x38000000, v3
+; SI-NEXT: v_cvt_f16_f32_e32 v3, v3
; SI-NEXT: s_waitcnt vmcnt(0)
-; SI-NEXT: v_cvt_f32_f16_e32 v3, v3
+; SI-NEXT: v_cvt_f32_f16_e32 v2, v2
; SI-NEXT: v_add_f32_e32 v2, 0x38000000, v2
; SI-NEXT: v_cvt_f16_f32_e32 v2, v2
-; SI-NEXT: v_add_f32_e32 v3, 0x38000000, v3
-; SI-NEXT: v_cvt_f16_f32_e32 v3, v3
; SI-NEXT: v_lshlrev_b32_e32 v2, 16, v2
; SI-NEXT: v_or_b32_e32 v2, v3, v2
-; SI-NEXT: v_cvt_f32_f16_e32 v3, v63
+; SI-NEXT: v_cvt_f32_f16_e32 v3, v62
; SI-NEXT: v_add_f32_e32 v3, 0x38000000, v3
; SI-NEXT: v_cvt_f16_f32_e32 v3, v3
; SI-NEXT: v_lshlrev_b32_e32 v3, 16, v3
@@ -17877,12 +17886,12 @@ define inreg <22 x float> @bitcast_v44f16_to_v22f32_scalar(<44 x half> inreg %a,
; SI-NEXT: v_cvt_f16_f32_e32 v6, v6
; SI-NEXT: v_lshlrev_b32_e32 v6, 16, v6
; SI-NEXT: v_or_b32_e32 v6, v7, v6
-; SI-NEXT: v_cvt_f32_f16_e32 v7, v60
+; SI-NEXT: v_cvt_f32_f16_e32 v7, v39
; SI-NEXT: v_add_f32_e32 v7, 0x38000000, v7
; SI-NEXT: v_cvt_f16_f32_e32 v7, v7
; SI-NEXT: v_lshlrev_b32_e32 v7, 16, v7
; SI-NEXT: v_or_b32_e32 v7, v8, v7
-; SI-NEXT: v_cvt_f32_f16_e32 v8, v39
+; SI-NEXT: v_cvt_f32_f16_e32 v8, v48
; SI-NEXT: v_add_f32_e32 v8, 0x38000000, v8
; SI-NEXT: v_cvt_f16_f32_e32 v8, v8
; SI-NEXT: v_lshlrev_b32_e32 v8, 16, v8
@@ -17970,86 +17979,89 @@ define inreg <22 x float> @bitcast_v44f16_to_v22f32_scalar(<44 x half> inreg %a,
; SI-NEXT: s_setpc_b64 s[30:31]
; SI-NEXT: .LBB35_4:
; SI-NEXT: buffer_store_dword v48, off, s[0:3], s32 offset:76 ; 4-byte Folded Spill
-; SI-NEXT: buffer_store_dword v39, off, s[0:3], s32 offset:80 ; 4-byte Folded Spill
-; SI-NEXT: buffer_store_dword v38, off, s[0:3], s32 offset:84 ; 4-byte Folded Spill
-; SI-NEXT: buffer_store_dword v37, off, s[0:3], s32 offset:88 ; 4-byte Folded Spill
-; SI-NEXT: buffer_store_dword v49, off, s[0:3], s32 offset:92 ; 4-byte Folded Spill
-; SI-NEXT: buffer_store_dword v35, off, s[0:3], s32 offset:96 ; 4-byte Folded Spill
-; SI-NEXT: buffer_store_dword v50, off, s[0:3], s32 offset:100 ; 4-byte Folded Spill
-; SI-NEXT: buffer_store_dword v32, off, s[0:3], s32 offset:104 ; 4-byte Folded Spill
-; SI-NEXT: buffer_store_dword v59, off, s[0:3], s32 offset:108 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v38, off, s[0:3], s32 offset:80 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v37, off, s[0:3], s32 offset:84 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v36, off, s[0:3], s32 offset:88 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v35, off, s[0:3], s32 offset:92 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v50, off, s[0:3], s32 offset:96 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v32, off, s[0:3], s32 offset:100 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v59, off, s[0:3], s32 offset:104 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_mov_b32_e32 v59, v46
; SI-NEXT: v_mov_b32_e32 v46, v41
; SI-NEXT: v_mov_b32_e32 v41, v52
; SI-NEXT: v_mov_b32_e32 v52, v23
-; SI-NEXT: v_mov_b32_e32 v48, v60
+; SI-NEXT: v_mov_b32_e32 v48, v39
+; SI-NEXT: v_mov_b32_e32 v39, v60
; SI-NEXT: v_mov_b32_e32 v60, v47
; SI-NEXT: v_mov_b32_e32 v47, v42
; SI-NEXT: v_mov_b32_e32 v42, v53
; SI-NEXT: v_mov_b32_e32 v53, v22
-; SI-NEXT: v_mov_b32_e32 v35, v61
+; SI-NEXT: buffer_store_dword v34, off, s[0:3], s32 offset:108 ; 4-byte Folded Spill
+; SI-NEXT: s_waitcnt expcnt(0)
+; SI-NEXT: v_mov_b32_e32 v34, v61
; SI-NEXT: v_mov_b32_e32 v61, v56
; SI-NEXT: v_mov_b32_e32 v56, v43
; SI-NEXT: v_mov_b32_e32 v43, v54
; SI-NEXT: v_mov_b32_e32 v54, v24
-; SI-NEXT: v_mov_b32_e32 v50, v34
-; SI-NEXT: v_mov_b32_e32 v34, v62
+; SI-NEXT: v_mov_b32_e32 v50, v62
; SI-NEXT: v_mov_b32_e32 v62, v57
; SI-NEXT: v_mov_b32_e32 v57, v44
; SI-NEXT: v_mov_b32_e32 v44, v55
; SI-NEXT: v_mov_b32_e32 v55, v25
-; SI-NEXT: v_mov_b32_e32 v32, v33
+; SI-NEXT: v_mov_b32_e32 v32, v51
+; SI-NEXT: v_mov_b32_e32 v51, v33
; SI-NEXT: v_mov_b32_e32 v33, v63
; SI-NEXT: v_mov_b32_e32 v63, v58
; SI-NEXT: v_mov_b32_e32 v58, v45
; SI-NEXT: v_mov_b32_e32 v45, v40
; SI-NEXT: v_mov_b32_e32 v40, v31
-; SI-NEXT: v_mov_b32_e32 v39, v26
-; SI-NEXT: v_mov_b32_e32 v38, v27
-; SI-NEXT: v_mov_b32_e32 v37, v28
-; SI-NEXT: v_mov_b32_e32 v49, v36
-; SI-NEXT: v_mov_b32_e32 v36, v29
+; SI-NEXT: v_mov_b32_e32 v38, v26
+; SI-NEXT: v_mov_b32_e32 v37, v27
+; SI-NEXT: v_mov_b32_e32 v36, v28
+; SI-NEXT: v_mov_b32_e32 v35, v49
+; SI-NEXT: v_mov_b32_e32 v49, v29
; SI-NEXT: ; implicit-def: $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7_vgpr8_vgpr9_vgpr10_vgpr11_vgpr12_vgpr13_vgpr14_vgpr15_vgpr16_vgpr17_vgpr18_vgpr19_vgpr20_vgpr21_vgpr22_vgpr23_vgpr24_vgpr25_vgpr26_vgpr27_vgpr28_vgpr29_vgpr30_vgpr31
-; SI-NEXT: v_mov_b32_e32 v29, v36
-; SI-NEXT: v_mov_b32_e32 v36, v49
-; SI-NEXT: v_mov_b32_e32 v28, v37
-; SI-NEXT: v_mov_b32_e32 v27, v38
-; SI-NEXT: v_mov_b32_e32 v26, v39
+; SI-NEXT: v_mov_b32_e32 v29, v49
+; SI-NEXT: v_mov_b32_e32 v49, v35
+; SI-NEXT: v_mov_b32_e32 v28, v36
+; SI-NEXT: v_mov_b32_e32 v27, v37
+; SI-NEXT: v_mov_b32_e32 v26, v38
; SI-NEXT: v_mov_b32_e32 v31, v40
; SI-NEXT: v_mov_b32_e32 v40, v45
; SI-NEXT: v_mov_b32_e32 v45, v58
; SI-NEXT: v_mov_b32_e32 v58, v63
; SI-NEXT: v_mov_b32_e32 v63, v33
-; SI-NEXT: v_mov_b32_e32 v33, v32
+; SI-NEXT: v_mov_b32_e32 v33, v51
+; SI-NEXT: v_mov_b32_e32 v51, v32
; SI-NEXT: v_mov_b32_e32 v25, v55
; SI-NEXT: v_mov_b32_e32 v55, v44
; SI-NEXT: v_mov_b32_e32 v44, v57
; SI-NEXT: v_mov_b32_e32 v57, v62
-; SI-NEXT: v_mov_b32_e32 v62, v34
-; SI-NEXT: v_mov_b32_e32 v34, v50
+; SI-NEXT: v_mov_b32_e32 v62, v50
; SI-NEXT: v_mov_b32_e32 v24, v54
; SI-NEXT: v_mov_b32_e32 v54, v43
; SI-NEXT: v_mov_b32_e32 v43, v56
; SI-NEXT: v_mov_b32_e32 v56, v61
-; SI-NEXT: v_mov_b32_e32 v61, v35
+; SI-NEXT: v_mov_b32_e32 v61, v34
+; SI-NEXT: buffer_load_dword v34, off, s[0:3], s32 offset:108 ; 4-byte Folded Reload
; SI-NEXT: v_mov_b32_e32 v22, v53
; SI-NEXT: v_mov_b32_e32 v53, v42
; SI-NEXT: v_mov_b32_e32 v42, v47
; SI-NEXT: v_mov_b32_e32 v47, v60
-; SI-NEXT: v_mov_b32_e32 v60, v48
+; SI-NEXT: v_mov_b32_e32 v60, v39
+; SI-NEXT: v_mov_b32_e32 v39, v48
; SI-NEXT: v_mov_b32_e32 v23, v52
; SI-NEXT: v_mov_b32_e32 v52, v41
; SI-NEXT: v_mov_b32_e32 v41, v46
; SI-NEXT: v_mov_b32_e32 v46, v59
-; SI-NEXT: buffer_load_dword v59, off, s[0:3], s32 offset:108 ; 4-byte Folded Reload
-; SI-NEXT: buffer_load_dword v32, off, s[0:3], s32 offset:104 ; 4-byte Folded Reload
-; SI-NEXT: buffer_load_dword v50, off, s[0:3], s32 offset:100 ; 4-byte Folded Reload
-; SI-NEXT: buffer_load_dword v35, off, s[0:3], s32 offset:96 ; 4-byte Folded Reload
-; SI-NEXT: buffer_load_dword v49, off, s[0:3], s32 offset:92 ; 4-byte Folded Reload
-; SI-NEXT: buffer_load_dword v37, off, s[0:3], s32 offset:88 ; 4-byte Folded Reload
-; SI-NEXT: buffer_load_dword v38, off, s[0:3], s32 offset:84 ; 4-byte Folded Reload
-; SI-NEXT: buffer_load_dword v39, off, s[0:3], s32 offset:80 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v59, off, s[0:3], s32 offset:104 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v32, off, s[0:3], s32 offset:100 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v50, off, s[0:3], s32 offset:96 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v35, off, s[0:3], s32 offset:92 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v36, off, s[0:3], s32 offset:88 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v37, off, s[0:3], s32 offset:84 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v38, off, s[0:3], s32 offset:80 ; 4-byte Folded Reload
; SI-NEXT: buffer_load_dword v48, off, s[0:3], s32 offset:76 ; 4-byte Folded Reload
; SI-NEXT: s_branch .LBB35_2
;
@@ -20712,7 +20724,7 @@ define inreg <44 x i16> @bitcast_v11i64_to_v44i16_scalar(<11 x i64> inreg %a, i3
; GFX11-TRUE16-NEXT: v_readfirstlane_b32 s5, v1
; GFX11-TRUE16-NEXT: v_readfirstlane_b32 s6, v2
; GFX11-TRUE16-NEXT: v_readfirstlane_b32 s4, v3
-; GFX11-TRUE16-NEXT: s_mov_b32 s62, 0
+; GFX11-TRUE16-NEXT: s_mov_b32 s58, 0
; GFX11-TRUE16-NEXT: s_and_b32 s8, vcc_lo, exec_lo
; GFX11-TRUE16-NEXT: s_cbranch_scc0 .LBB41_4
; GFX11-TRUE16-NEXT: ; %bb.1: ; %cmp.false
@@ -20734,11 +20746,11 @@ define inreg <44 x i16> @bitcast_v11i64_to_v44i16_scalar(<11 x i64> inreg %a, i3
; GFX11-TRUE16-NEXT: s_lshr_b32 s47, s18, 16
; GFX11-TRUE16-NEXT: s_lshr_b32 s56, s17, 16
; GFX11-TRUE16-NEXT: s_lshr_b32 s57, s16, 16
-; GFX11-TRUE16-NEXT: s_lshr_b32 s58, s3, 16
-; GFX11-TRUE16-NEXT: s_lshr_b32 s59, s2, 16
-; GFX11-TRUE16-NEXT: s_lshr_b32 s60, s1, 16
-; GFX11-TRUE16-NEXT: s_lshr_b32 s61, s0, 16
-; GFX11-TRUE16-NEXT: s_and_not1_b32 vcc_lo, exec_lo, s62
+; GFX11-TRUE16-NEXT: s_lshr_b32 s59, s3, 16
+; GFX11-TRUE16-NEXT: s_lshr_b32 s60, s2, 16
+; GFX11-TRUE16-NEXT: s_lshr_b32 s61, s1, 16
+; GFX11-TRUE16-NEXT: s_lshr_b32 s62, s0, 16
+; GFX11-TRUE16-NEXT: s_and_not1_b32 vcc_lo, exec_lo, s58
; GFX11-TRUE16-NEXT: s_cbranch_vccnz .LBB41_3
; GFX11-TRUE16-NEXT: .LBB41_2: ; %cmp.true
; GFX11-TRUE16-NEXT: s_add_u32 s6, s6, 3
@@ -20781,19 +20793,22 @@ define inreg <44 x i16> @bitcast_v11i64_to_v44i16_scalar(<11 x i64> inreg %a, i3
; GFX11-TRUE16-NEXT: s_lshr_b32 s47, s18, 16
; GFX11-TRUE16-NEXT: s_lshr_b32 s56, s17, 16
; GFX11-TRUE16-NEXT: s_lshr_b32 s57, s16, 16
-; GFX11-TRUE16-NEXT: s_lshr_b32 s58, s3, 16
-; GFX11-TRUE16-NEXT: s_lshr_b32 s59, s2, 16
-; GFX11-TRUE16-NEXT: s_lshr_b32 s60, s1, 16
-; GFX11-TRUE16-NEXT: s_lshr_b32 s61, s0, 16
+; GFX11-TRUE16-NEXT: s_lshr_b32 s59, s3, 16
+; GFX11-TRUE16-NEXT: s_lshr_b32 s60, s2, 16
+; GFX11-TRUE16-NEXT: s_lshr_b32 s61, s1, 16
+; GFX11-TRUE16-NEXT: s_lshr_b32 s62, s0, 16
; GFX11-TRUE16-NEXT: .LBB41_3: ; %end
+; GFX11-TRUE16-NEXT: s_delay_alu instid0(SALU_CYCLE_1)
+; GFX11-TRUE16-NEXT: s_mov_b32 s58, s62
; GFX11-TRUE16-NEXT: s_pack_ll_b32_b16 s15, s26, s15
+; GFX11-TRUE16-NEXT: s_pack_ll_b32_b16 s0, s0, s58
+; GFX11-TRUE16-NEXT: s_mov_b32 s58, s60
; GFX11-TRUE16-NEXT: s_pack_ll_b32_b16 s14, s27, s14
; GFX11-TRUE16-NEXT: s_mov_b32 s26, s28
; GFX11-TRUE16-NEXT: s_mov_b32 s27, s29
-; GFX11-TRUE16-NEXT: s_pack_ll_b32_b16 s0, s0, s61
-; GFX11-TRUE16-NEXT: s_pack_ll_b32_b16 s1, s1, s60
-; GFX11-TRUE16-NEXT: s_pack_ll_b32_b16 s2, s2, s59
-; GFX11-TRUE16-NEXT: s_pack_ll_b32_b16 s3, s3, s58
+; GFX11-TRUE16-NEXT: s_pack_ll_b32_b16 s1, s1, s61
+; GFX11-TRUE16-NEXT: s_pack_ll_b32_b16 s2, s2, s58
+; GFX11-TRUE16-NEXT: s_pack_ll_b32_b16 s3, s3, s59
; GFX11-TRUE16-NEXT: s_pack_ll_b32_b16 s16, s16, s57
; GFX11-TRUE16-NEXT: s_pack_ll_b32_b16 s17, s17, s56
; GFX11-TRUE16-NEXT: s_pack_ll_b32_b16 s18, s18, s47
@@ -20823,10 +20838,10 @@ define inreg <44 x i16> @bitcast_v11i64_to_v44i16_scalar(<11 x i64> inreg %a, i3
; GFX11-TRUE16-NEXT: v_dual_mov_b32 v20, s6 :: v_dual_mov_b32 v21, s4
; GFX11-TRUE16-NEXT: s_setpc_b64 s[30:31]
; GFX11-TRUE16-NEXT: .LBB41_4:
+; GFX11-TRUE16-NEXT: ; implicit-def: $sgpr62_lo16
; GFX11-TRUE16-NEXT: ; implicit-def: $sgpr61_lo16
; GFX11-TRUE16-NEXT: ; implicit-def: $sgpr60_lo16
; GFX11-TRUE16-NEXT: ; implicit-def: $sgpr59_lo16
-; GFX11-TRUE16-NEXT: ; implicit-def: $sgpr58_lo16
; GFX11-TRUE16-NEXT: ; implicit-def: $sgpr57_lo16
; GFX11-TRUE16-NEXT: ; implicit-def: $sgpr56_lo16
; GFX11-TRUE16-NEXT: ; implicit-def: $sgpr47_lo16
@@ -24465,7 +24480,7 @@ define inreg <44 x half> @bitcast_v11i64_to_v44f16_scalar(<11 x i64> inreg %a, i
; GFX11-TRUE16-NEXT: v_readfirstlane_b32 s5, v1
; GFX11-TRUE16-NEXT: v_readfirstlane_b32 s6, v2
; GFX11-TRUE16-NEXT: v_readfirstlane_b32 s4, v3
-; GFX11-TRUE16-NEXT: s_mov_b32 s62, 0
+; GFX11-TRUE16-NEXT: s_mov_b32 s58, 0
; GFX11-TRUE16-NEXT: s_and_b32 s8, vcc_lo, exec_lo
; GFX11-TRUE16-NEXT: s_cbranch_scc0 .LBB45_4
; GFX11-TRUE16-NEXT: ; %bb.1: ; %cmp.false
@@ -24487,11 +24502,11 @@ define inreg <44 x half> @bitcast_v11i64_to_v44f16_scalar(<11 x i64> inreg %a, i
; GFX11-TRUE16-NEXT: s_lshr_b32 s47, s18, 16
; GFX11-TRUE16-NEXT: s_lshr_b32 s56, s17, 16
; GFX11-TRUE16-NEXT: s_lshr_b32 s57, s16, 16
-; GFX11-TRUE16-NEXT: s_lshr_b32 s58, s3, 16
-; GFX11-TRUE16-NEXT: s_lshr_b32 s59, s2, 16
-; GFX11-TRUE16-NEXT: s_lshr_b32 s60, s1, 16
-; GFX11-TRUE16-NEXT: s_lshr_b32 s61, s0, 16
-; GFX11-TRUE16-NEXT: s_and_not1_b32 vcc_lo, exec_lo, s62
+; GFX11-TRUE16-NEXT: s_lshr_b32 s59, s3, 16
+; GFX11-TRUE16-NEXT: s_lshr_b32 s60, s2, 16
+; GFX11-TRUE16-NEXT: s_lshr_b32 s61, s1, 16
+; GFX11-TRUE16-NEXT: s_lshr_b32 s62, s0, 16
+; GFX11-TRUE16-NEXT: s_and_not1_b32 vcc_lo, exec_lo, s58
; GFX11-TRUE16-NEXT: s_cbranch_vccnz .LBB45_3
; GFX11-TRUE16-NEXT: .LBB45_2: ; %cmp.true
; GFX11-TRUE16-NEXT: s_add_u32 s6, s6, 3
@@ -24534,19 +24549,22 @@ define inreg <44 x half> @bitcast_v11i64_to_v44f16_scalar(<11 x i64> inreg %a, i
; GFX11-TRUE16-NEXT: s_lshr_b32 s47, s18, 16
; GFX11-TRUE16-NEXT: s_lshr_b32 s56, s17, 16
; GFX11-TRUE16-NEXT: s_lshr_b32 s57, s16, 16
-; GFX11-TRUE16-NEXT: s_lshr_b32 s58, s3, 16
-; GFX11-TRUE16-NEXT: s_lshr_b32 s59, s2, 16
-; GFX11-TRUE16-NEXT: s_lshr_b32 s60, s1, 16
-; GFX11-TRUE16-NEXT: s_lshr_b32 s61, s0, 16
+; GFX11-TRUE16-NEXT: s_lshr_b32 s59, s3, 16
+; GFX11-TRUE16-NEXT: s_lshr_b32 s60, s2, 16
+; GFX11-TRUE16-NEXT: s_lshr_b32 s61, s1, 16
+; GFX11-TRUE16-NEXT: s_lshr_b32 s62, s0, 16
; GFX11-TRUE16-NEXT: .LBB45_3: ; %end
+; GFX11-TRUE16-NEXT: s_delay_alu instid0(SALU_CYCLE_1)
+; GFX11-TRUE16-NEXT: s_mov_b32 s58, s62
; GFX11-TRUE16-NEXT: s_pack_ll_b32_b16 s15, s26, s15
+; GFX11-TRUE16-NEXT: s_pack_ll_b32_b16 s0, s0, s58
+; GFX11-TRUE16-NEXT: s_mov_b32 s58, s60
; GFX11-TRUE16-NEXT: s_pack_ll_b32_b16 s14, s27, s14
; GFX11-TRUE16-NEXT: s_mov_b32 s26, s28
; GFX11-TRUE16-NEXT: s_mov_b32 s27, s29
-; GFX11-TRUE16-NEXT: s_pack_ll_b32_b16 s0, s0, s61
-; GFX11-TRUE16-NEXT: s_pack_ll_b32_b16 s1, s1, s60
-; GFX11-TRUE16-NEXT: s_pack_ll_b32_b16 s2, s2, s59
-; GFX11-TRUE16-NEXT: s_pack_ll_b32_b16 s3, s3, s58
+; GFX11-TRUE16-NEXT: s_pack_ll_b32_b16 s1, s1, s61
+; GFX11-TRUE16-NEXT: s_pack_ll_b32_b16 s2, s2, s58
+; GFX11-TRUE16-NEXT: s_pack_ll_b32_b16 s3, s3, s59
; GFX11-TRUE16-NEXT: s_pack_ll_b32_b16 s16, s16, s57
; GFX11-TRUE16-NEXT: s_pack_ll_b32_b16 s17, s17, s56
; GFX11-TRUE16-NEXT: s_pack_ll_b32_b16 s18, s18, s47
@@ -24576,10 +24594,10 @@ define inreg <44 x half> @bitcast_v11i64_to_v44f16_scalar(<11 x i64> inreg %a, i
; GFX11-TRUE16-NEXT: v_dual_mov_b32 v20, s6 :: v_dual_mov_b32 v21, s4
; GFX11-TRUE16-NEXT: s_setpc_b64 s[30:31]
; GFX11-TRUE16-NEXT: .LBB45_4:
+; GFX11-TRUE16-NEXT: ; implicit-def: $sgpr62_lo16
; GFX11-TRUE16-NEXT: ; implicit-def: $sgpr61_lo16
; GFX11-TRUE16-NEXT: ; implicit-def: $sgpr60_lo16
; GFX11-TRUE16-NEXT: ; implicit-def: $sgpr59_lo16
-; GFX11-TRUE16-NEXT: ; implicit-def: $sgpr58_lo16
; GFX11-TRUE16-NEXT: ; implicit-def: $sgpr57_lo16
; GFX11-TRUE16-NEXT: ; implicit-def: $sgpr56_lo16
; GFX11-TRUE16-NEXT: ; implicit-def: $sgpr47_lo16
@@ -25857,13 +25875,13 @@ define inreg <11 x i64> @bitcast_v44f16_to_v11i64_scalar(<44 x half> inreg %a, i
; SI-NEXT: buffer_store_dword v61, off, s[0:3], s32 offset:8 ; 4-byte Folded Spill
; SI-NEXT: buffer_store_dword v62, off, s[0:3], s32 offset:4 ; 4-byte Folded Spill
; SI-NEXT: buffer_store_dword v63, off, s[0:3], s32 ; 4-byte Folded Spill
+; SI-NEXT: v_cvt_f16_f32_e32 v39, v1
; SI-NEXT: s_waitcnt expcnt(3)
-; SI-NEXT: v_cvt_f16_f32_e32 v60, v1
-; SI-NEXT: v_cvt_f16_f32_e32 v48, v0
-; SI-NEXT: v_cvt_f16_f32_e32 v39, v3
+; SI-NEXT: v_cvt_f16_f32_e32 v60, v0
+; SI-NEXT: v_cvt_f16_f32_e32 v48, v3
; SI-NEXT: v_cvt_f16_f32_e32 v38, v2
; SI-NEXT: v_cvt_f16_f32_e32 v37, v5
-; SI-NEXT: v_cvt_f16_f32_e32 v49, v4
+; SI-NEXT: v_cvt_f16_f32_e32 v36, v4
; SI-NEXT: v_cvt_f16_f32_e32 v35, v7
; SI-NEXT: v_cvt_f16_f32_e32 v50, v6
; SI-NEXT: v_cvt_f16_f32_e32 v32, v9
@@ -25891,37 +25909,36 @@ define inreg <11 x i64> @bitcast_v44f16_to_v11i64_scalar(<44 x half> inreg %a, i
; SI-NEXT: v_cvt_f16_f32_e32 v51, s17
; SI-NEXT: v_cvt_f16_f32_e32 v33, s16
; SI-NEXT: v_cvt_f16_f32_e32 v1, s19
-; SI-NEXT: v_cvt_f16_f32_e32 v34, s18
-; SI-NEXT: v_cvt_f16_f32_e32 v2, s21
-; SI-NEXT: v_cvt_f16_f32_e32 v3, s20
+; SI-NEXT: v_cvt_f16_f32_e32 v2, s18
+; SI-NEXT: v_cvt_f16_f32_e32 v3, s21
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_cvt_f16_f32_e32 v63, s23
-; SI-NEXT: v_cvt_f16_f32_e32 v62, s22
+; SI-NEXT: v_cvt_f16_f32_e32 v63, s20
+; SI-NEXT: v_cvt_f16_f32_e32 v62, s23
+; SI-NEXT: v_cvt_f16_f32_e32 v34, s22
; SI-NEXT: v_cvt_f16_f32_e32 v61, s25
-; SI-NEXT: v_cvt_f16_f32_e32 v36, s24
+; SI-NEXT: v_cvt_f16_f32_e32 v49, s24
; SI-NEXT: v_cvt_f16_f32_e32 v29, s27
; SI-NEXT: v_cvt_f16_f32_e32 v28, s26
; SI-NEXT: v_cvt_f16_f32_e32 v27, s29
; SI-NEXT: v_cvt_f16_f32_e32 v26, s28
; SI-NEXT: v_cmp_ne_u32_e32 vcc, 0, v30
; SI-NEXT: s_and_b64 s[4:5], vcc, exec
-; SI-NEXT: buffer_store_dword v3, off, s[0:3], s32 offset:64 ; 4-byte Folded Spill
-; SI-NEXT: buffer_store_dword v2, off, s[0:3], s32 offset:68 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v2, off, s[0:3], s32 offset:64 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v3, off, s[0:3], s32 offset:68 ; 4-byte Folded Spill
; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:72 ; 4-byte Folded Spill
; SI-NEXT: s_cbranch_scc0 .LBB47_4
; SI-NEXT: ; %bb.1: ; %cmp.false
-; SI-NEXT: s_waitcnt expcnt(1)
-; SI-NEXT: v_lshlrev_b32_e32 v2, 16, v2
-; SI-NEXT: v_lshlrev_b32_e32 v0, 16, v51
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_lshlrev_b32_e32 v1, 16, v1
-; SI-NEXT: v_or_b32_e32 v2, v3, v2
-; SI-NEXT: v_lshlrev_b32_e32 v3, 16, v63
+; SI-NEXT: v_lshlrev_b32_e32 v0, 16, v51
+; SI-NEXT: v_or_b32_e32 v1, v2, v1
+; SI-NEXT: v_lshlrev_b32_e32 v2, 16, v3
+; SI-NEXT: v_lshlrev_b32_e32 v3, 16, v62
; SI-NEXT: v_lshlrev_b32_e32 v4, 16, v61
; SI-NEXT: v_lshlrev_b32_e32 v5, 16, v29
; SI-NEXT: v_lshlrev_b32_e32 v6, 16, v27
-; SI-NEXT: v_lshlrev_b32_e32 v7, 16, v60
-; SI-NEXT: v_lshlrev_b32_e32 v8, 16, v39
+; SI-NEXT: v_lshlrev_b32_e32 v7, 16, v39
+; SI-NEXT: v_lshlrev_b32_e32 v8, 16, v48
; SI-NEXT: v_lshlrev_b32_e32 v9, 16, v37
; SI-NEXT: v_lshlrev_b32_e32 v10, 16, v35
; SI-NEXT: v_lshlrev_b32_e32 v11, 16, v32
@@ -25936,14 +25953,14 @@ define inreg <11 x i64> @bitcast_v44f16_to_v11i64_scalar(<44 x half> inreg %a, i
; SI-NEXT: v_lshlrev_b32_e32 v20, 16, v25
; SI-NEXT: v_lshlrev_b32_e32 v21, 16, v22
; SI-NEXT: v_or_b32_e32 v0, v33, v0
-; SI-NEXT: v_or_b32_e32 v1, v34, v1
-; SI-NEXT: v_or_b32_e32 v3, v62, v3
-; SI-NEXT: v_or_b32_e32 v4, v36, v4
+; SI-NEXT: v_or_b32_e32 v2, v63, v2
+; SI-NEXT: v_or_b32_e32 v3, v34, v3
+; SI-NEXT: v_or_b32_e32 v4, v49, v4
; SI-NEXT: v_or_b32_e32 v5, v28, v5
; SI-NEXT: v_or_b32_e32 v6, v26, v6
-; SI-NEXT: v_or_b32_e32 v7, v48, v7
+; SI-NEXT: v_or_b32_e32 v7, v60, v7
; SI-NEXT: v_or_b32_e32 v8, v38, v8
-; SI-NEXT: v_or_b32_e32 v9, v49, v9
+; SI-NEXT: v_or_b32_e32 v9, v36, v9
; SI-NEXT: v_or_b32_e32 v10, v50, v10
; SI-NEXT: v_or_b32_e32 v11, v59, v11
; SI-NEXT: v_or_b32_e32 v12, v57, v12
@@ -25959,65 +25976,64 @@ define inreg <11 x i64> @bitcast_v44f16_to_v11i64_scalar(<44 x half> inreg %a, i
; SI-NEXT: s_cbranch_execnz .LBB47_3
; SI-NEXT: .LBB47_2: ; %cmp.true
; SI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:72 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v3, off, s[0:3], s32 offset:64 ; 4-byte Folded Reload
; SI-NEXT: v_cvt_f32_f16_e32 v0, v51
; SI-NEXT: v_cvt_f32_f16_e32 v1, v33
-; SI-NEXT: v_cvt_f32_f16_e32 v3, v34
-; SI-NEXT: v_cvt_f32_f16_e32 v4, v62
+; SI-NEXT: s_waitcnt vmcnt(10)
+; SI-NEXT: v_cvt_f32_f16_e32 v4, v34
+; SI-NEXT: v_cvt_f32_f16_e32 v5, v49
; SI-NEXT: v_add_f32_e32 v0, 0x38000000, v0
; SI-NEXT: v_add_f32_e32 v1, 0x38000000, v1
; SI-NEXT: v_cvt_f16_f32_e32 v0, v0
-; SI-NEXT: v_add_f32_e32 v3, 0x38000000, v3
; SI-NEXT: v_cvt_f16_f32_e32 v1, v1
-; SI-NEXT: v_cvt_f16_f32_e32 v3, v3
-; SI-NEXT: v_lshlrev_b32_e32 v0, 16, v0
; SI-NEXT: v_add_f32_e32 v4, 0x38000000, v4
-; SI-NEXT: v_or_b32_e32 v0, v1, v0
; SI-NEXT: v_cvt_f16_f32_e32 v4, v4
-; SI-NEXT: v_cvt_f32_f16_e32 v5, v36
-; SI-NEXT: v_cvt_f32_f16_e32 v6, v28
-; SI-NEXT: v_cvt_f32_f16_e32 v7, v26
-; SI-NEXT: s_waitcnt vmcnt(1)
-; SI-NEXT: v_cvt_f32_f16_e32 v8, v48
+; SI-NEXT: v_lshlrev_b32_e32 v0, 16, v0
+; SI-NEXT: v_or_b32_e32 v0, v1, v0
; SI-NEXT: v_add_f32_e32 v5, 0x38000000, v5
; SI-NEXT: v_cvt_f16_f32_e32 v5, v5
+; SI-NEXT: v_cvt_f32_f16_e32 v6, v28
+; SI-NEXT: v_cvt_f32_f16_e32 v7, v26
+; SI-NEXT: v_cvt_f32_f16_e32 v8, v60
+; SI-NEXT: s_waitcnt vmcnt(3)
+; SI-NEXT: v_cvt_f32_f16_e32 v9, v38
; SI-NEXT: v_add_f32_e32 v6, 0x38000000, v6
; SI-NEXT: v_cvt_f16_f32_e32 v6, v6
; SI-NEXT: v_add_f32_e32 v7, 0x38000000, v7
; SI-NEXT: v_cvt_f16_f32_e32 v7, v7
; SI-NEXT: v_add_f32_e32 v8, 0x38000000, v8
; SI-NEXT: v_cvt_f16_f32_e32 v8, v8
-; SI-NEXT: v_cvt_f32_f16_e32 v9, v38
; SI-NEXT: v_cvt_f32_f16_e32 v10, v37
-; SI-NEXT: v_cvt_f32_f16_e32 v11, v49
-; SI-NEXT: v_cvt_f32_f16_e32 v12, v50
; SI-NEXT: v_add_f32_e32 v9, 0x38000000, v9
-; SI-NEXT: v_add_f32_e32 v10, 0x38000000, v10
; SI-NEXT: v_cvt_f16_f32_e32 v9, v9
+; SI-NEXT: v_cvt_f32_f16_e32 v11, v36
+; SI-NEXT: v_add_f32_e32 v10, 0x38000000, v10
; SI-NEXT: v_cvt_f16_f32_e32 v10, v10
+; SI-NEXT: v_cvt_f32_f16_e32 v12, v50
; SI-NEXT: v_add_f32_e32 v11, 0x38000000, v11
; SI-NEXT: v_cvt_f16_f32_e32 v11, v11
+; SI-NEXT: v_cvt_f32_f16_e32 v13, v58
; SI-NEXT: v_add_f32_e32 v12, 0x38000000, v12
; SI-NEXT: v_cvt_f16_f32_e32 v12, v12
-; SI-NEXT: v_cvt_f32_f16_e32 v13, v58
; SI-NEXT: v_cvt_f32_f16_e32 v14, v57
-; SI-NEXT: v_cvt_f32_f16_e32 v15, v47
-; SI-NEXT: v_cvt_f32_f16_e32 v16, v44
; SI-NEXT: v_add_f32_e32 v13, 0x38000000, v13
; SI-NEXT: v_cvt_f16_f32_e32 v13, v13
+; SI-NEXT: v_cvt_f32_f16_e32 v15, v47
; SI-NEXT: v_add_f32_e32 v14, 0x38000000, v14
; SI-NEXT: v_cvt_f16_f32_e32 v14, v14
+; SI-NEXT: v_cvt_f32_f16_e32 v16, v44
; SI-NEXT: v_add_f32_e32 v15, 0x38000000, v15
; SI-NEXT: v_cvt_f16_f32_e32 v15, v15
+; SI-NEXT: v_cvt_f32_f16_e32 v17, v43
; SI-NEXT: v_add_f32_e32 v16, 0x38000000, v16
; SI-NEXT: v_cvt_f16_f32_e32 v16, v16
-; SI-NEXT: v_cvt_f32_f16_e32 v17, v43
; SI-NEXT: v_cvt_f32_f16_e32 v18, v41
-; SI-NEXT: v_cvt_f32_f16_e32 v19, v54
-; SI-NEXT: v_cvt_f32_f16_e32 v20, v53
; SI-NEXT: v_add_f32_e32 v17, 0x38000000, v17
; SI-NEXT: v_cvt_f16_f32_e32 v17, v17
+; SI-NEXT: v_cvt_f32_f16_e32 v19, v54
; SI-NEXT: v_add_f32_e32 v18, 0x38000000, v18
; SI-NEXT: v_cvt_f16_f32_e32 v18, v18
+; SI-NEXT: v_cvt_f32_f16_e32 v20, v53
; SI-NEXT: v_add_f32_e32 v19, 0x38000000, v19
; SI-NEXT: v_cvt_f16_f32_e32 v19, v19
; SI-NEXT: v_cvt_f32_f16_e32 v21, v31
@@ -26031,25 +26047,27 @@ define inreg <11 x i64> @bitcast_v44f16_to_v11i64_scalar(<44 x half> inreg %a, i
; SI-NEXT: v_cvt_f16_f32_e32 v22, v22
; SI-NEXT: v_add_f32_e32 v23, 0x38000000, v23
; SI-NEXT: v_cvt_f16_f32_e32 v23, v23
-; SI-NEXT: s_waitcnt vmcnt(0)
+; SI-NEXT: s_waitcnt vmcnt(1)
; SI-NEXT: v_cvt_f32_f16_e32 v2, v2
+; SI-NEXT: s_waitcnt vmcnt(0)
+; SI-NEXT: v_cvt_f32_f16_e32 v3, v3
; SI-NEXT: v_add_f32_e32 v2, 0x38000000, v2
; SI-NEXT: v_cvt_f16_f32_e32 v2, v2
+; SI-NEXT: v_add_f32_e32 v3, 0x38000000, v3
+; SI-NEXT: v_cvt_f16_f32_e32 v3, v3
; SI-NEXT: v_lshlrev_b32_e32 v2, 16, v2
; SI-NEXT: v_or_b32_e32 v1, v3, v2
; SI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:68 ; 4-byte Folded Reload
-; SI-NEXT: buffer_load_dword v3, off, s[0:3], s32 offset:64 ; 4-byte Folded Reload
-; SI-NEXT: s_waitcnt vmcnt(1)
-; SI-NEXT: v_cvt_f32_f16_e32 v2, v2
+; SI-NEXT: v_cvt_f32_f16_e32 v3, v63
+; SI-NEXT: v_add_f32_e32 v3, 0x38000000, v3
+; SI-NEXT: v_cvt_f16_f32_e32 v3, v3
; SI-NEXT: s_waitcnt vmcnt(0)
-; SI-NEXT: v_cvt_f32_f16_e32 v3, v3
+; SI-NEXT: v_cvt_f32_f16_e32 v2, v2
; SI-NEXT: v_add_f32_e32 v2, 0x38000000, v2
; SI-NEXT: v_cvt_f16_f32_e32 v2, v2
-; SI-NEXT: v_add_f32_e32 v3, 0x38000000, v3
-; SI-NEXT: v_cvt_f16_f32_e32 v3, v3
; SI-NEXT: v_lshlrev_b32_e32 v2, 16, v2
; SI-NEXT: v_or_b32_e32 v2, v3, v2
-; SI-NEXT: v_cvt_f32_f16_e32 v3, v63
+; SI-NEXT: v_cvt_f32_f16_e32 v3, v62
; SI-NEXT: v_add_f32_e32 v3, 0x38000000, v3
; SI-NEXT: v_cvt_f16_f32_e32 v3, v3
; SI-NEXT: v_lshlrev_b32_e32 v3, 16, v3
@@ -26069,12 +26087,12 @@ define inreg <11 x i64> @bitcast_v44f16_to_v11i64_scalar(<44 x half> inreg %a, i
; SI-NEXT: v_cvt_f16_f32_e32 v6, v6
; SI-NEXT: v_lshlrev_b32_e32 v6, 16, v6
; SI-NEXT: v_or_b32_e32 v6, v7, v6
-; SI-NEXT: v_cvt_f32_f16_e32 v7, v60
+; SI-NEXT: v_cvt_f32_f16_e32 v7, v39
; SI-NEXT: v_add_f32_e32 v7, 0x38000000, v7
; SI-NEXT: v_cvt_f16_f32_e32 v7, v7
; SI-NEXT: v_lshlrev_b32_e32 v7, 16, v7
; SI-NEXT: v_or_b32_e32 v7, v8, v7
-; SI-NEXT: v_cvt_f32_f16_e32 v8, v39
+; SI-NEXT: v_cvt_f32_f16_e32 v8, v48
; SI-NEXT: v_add_f32_e32 v8, 0x38000000, v8
; SI-NEXT: v_cvt_f16_f32_e32 v8, v8
; SI-NEXT: v_lshlrev_b32_e32 v8, 16, v8
@@ -26162,86 +26180,89 @@ define inreg <11 x i64> @bitcast_v44f16_to_v11i64_scalar(<44 x half> inreg %a, i
; SI-NEXT: s_setpc_b64 s[30:31]
; SI-NEXT: .LBB47_4:
; SI-NEXT: buffer_store_dword v48, off, s[0:3], s32 offset:76 ; 4-byte Folded Spill
-; SI-NEXT: buffer_store_dword v39, off, s[0:3], s32 offset:80 ; 4-byte Folded Spill
-; SI-NEXT: buffer_store_dword v38, off, s[0:3], s32 offset:84 ; 4-byte Folded Spill
-; SI-NEXT: buffer_store_dword v37, off, s[0:3], s32 offset:88 ; 4-byte Folded Spill
-; SI-NEXT: buffer_store_dword v49, off, s[0:3], s32 offset:92 ; 4-byte Folded Spill
-; SI-NEXT: buffer_store_dword v35, off, s[0:3], s32 offset:96 ; 4-byte Folded Spill
-; SI-NEXT: buffer_store_dword v50, off, s[0:3], s32 offset:100 ; 4-byte Folded Spill
-; SI-NEXT: buffer_store_dword v32, off, s[0:3], s32 offset:104 ; 4-byte Folded Spill
-; SI-NEXT: buffer_store_dword v59, off, s[0:3], s32 offset:108 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v38, off, s[0:3], s32 offset:80 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v37, off, s[0:3], s32 offset:84 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v36, off, s[0:3], s32 offset:88 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v35, off, s[0:3], s32 offset:92 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v50, off, s[0:3], s32 offset:96 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v32, off, s[0:3], s32 offset:100 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v59, off, s[0:3], s32 offset:104 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_mov_b32_e32 v59, v46
; SI-NEXT: v_mov_b32_e32 v46, v41
; SI-NEXT: v_mov_b32_e32 v41, v52
; SI-NEXT: v_mov_b32_e32 v52, v23
-; SI-NEXT: v_mov_b32_e32 v48, v60
+; SI-NEXT: v_mov_b32_e32 v48, v39
+; SI-NEXT: v_mov_b32_e32 v39, v60
; SI-NEXT: v_mov_b32_e32 v60, v47
; SI-NEXT: v_mov_b32_e32 v47, v42
; SI-NEXT: v_mov_b32_e32 v42, v53
; SI-NEXT: v_mov_b32_e32 v53, v22
-; SI-NEXT: v_mov_b32_e32 v35, v61
+; SI-NEXT: buffer_store_dword v34, off, s[0:3], s32 offset:108 ; 4-byte Folded Spill
+; SI-NEXT: s_waitcnt expcnt(0)
+; SI-NEXT: v_mov_b32_e32 v34, v61
; SI-NEXT: v_mov_b32_e32 v61, v56
; SI-NEXT: v_mov_b32_e32 v56, v43
; SI-NEXT: v_mov_b32_e32 v43, v54
; SI-NEXT: v_mov_b32_e32 v54, v24
-; SI-NEXT: v_mov_b32_e32 v50, v34
-; SI-NEXT: v_mov_b32_e32 v34, v62
+; SI-NEXT: v_mov_b32_e32 v50, v62
; SI-NEXT: v_mov_b32_e32 v62, v57
; SI-NEXT: v_mov_b32_e32 v57, v44
; SI-NEXT: v_mov_b32_e32 v44, v55
; SI-NEXT: v_mov_b32_e32 v55, v25
-; SI-NEXT: v_mov_b32_e32 v32, v33
+; SI-NEXT: v_mov_b32_e32 v32, v51
+; SI-NEXT: v_mov_b32_e32 v51, v33
; SI-NEXT: v_mov_b32_e32 v33, v63
; SI-NEXT: v_mov_b32_e32 v63, v58
; SI-NEXT: v_mov_b32_e32 v58, v45
; SI-NEXT: v_mov_b32_e32 v45, v40
; SI-NEXT: v_mov_b32_e32 v40, v31
-; SI-NEXT: v_mov_b32_e32 v39, v26
-; SI-NEXT: v_mov_b32_e32 v38, v27
-; SI-NEXT: v_mov_b32_e32 v37, v28
-; SI-NEXT: v_mov_b32_e32 v49, v36
-; SI-NEXT: v_mov_b32_e32 v36, v29
+; SI-NEXT: v_mov_b32_e32 v38, v26
+; SI-NEXT: v_mov_b32_e32 v37, v27
+; SI-NEXT: v_mov_b32_e32 v36, v28
+; SI-NEXT: v_mov_b32_e32 v35, v49
+; SI-NEXT: v_mov_b32_e32 v49, v29
; SI-NEXT: ; implicit-def: $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7_vgpr8_vgpr9_vgpr10_vgpr11_vgpr12_vgpr13_vgpr14_vgpr15_vgpr16_vgpr17_vgpr18_vgpr19_vgpr20_vgpr21_vgpr22_vgpr23_vgpr24_vgpr25_vgpr26_vgpr27_vgpr28_vgpr29_vgpr30_vgpr31
-; SI-NEXT: v_mov_b32_e32 v29, v36
-; SI-NEXT: v_mov_b32_e32 v36, v49
-; SI-NEXT: v_mov_b32_e32 v28, v37
-; SI-NEXT: v_mov_b32_e32 v27, v38
-; SI-NEXT: v_mov_b32_e32 v26, v39
+; SI-NEXT: v_mov_b32_e32 v29, v49
+; SI-NEXT: v_mov_b32_e32 v49, v35
+; SI-NEXT: v_mov_b32_e32 v28, v36
+; SI-NEXT: v_mov_b32_e32 v27, v37
+; SI-NEXT: v_mov_b32_e32 v26, v38
; SI-NEXT: v_mov_b32_e32 v31, v40
; SI-NEXT: v_mov_b32_e32 v40, v45
; SI-NEXT: v_mov_b32_e32 v45, v58
; SI-NEXT: v_mov_b32_e32 v58, v63
; SI-NEXT: v_mov_b32_e32 v63, v33
-; SI-NEXT: v_mov_b32_e32 v33, v32
+; SI-NEXT: v_mov_b32_e32 v33, v51
+; SI-NEXT: v_mov_b32_e32 v51, v32
; SI-NEXT: v_mov_b32_e32 v25, v55
; SI-NEXT: v_mov_b32_e32 v55, v44
; SI-NEXT: v_mov_b32_e32 v44, v57
; SI-NEXT: v_mov_b32_e32 v57, v62
-; SI-NEXT: v_mov_b32_e32 v62, v34
-; SI-NEXT: v_mov_b32_e32 v34, v50
+; SI-NEXT: v_mov_b32_e32 v62, v50
; SI-NEXT: v_mov_b32_e32 v24, v54
; SI-NEXT: v_mov_b32_e32 v54, v43
; SI-NEXT: v_mov_b32_e32 v43, v56
; SI-NEXT: v_mov_b32_e32 v56, v61
-; SI-NEXT: v_mov_b32_e32 v61, v35
+; SI-NEXT: v_mov_b32_e32 v61, v34
+; SI-NEXT: buffer_load_dword v34, off, s[0:3], s32 offset:108 ; 4-byte Folded Reload
; SI-NEXT: v_mov_b32_e32 v22, v53
; SI-NEXT: v_mov_b32_e32 v53, v42
; SI-NEXT: v_mov_b32_e32 v42, v47
; SI-NEXT: v_mov_b32_e32 v47, v60
-; SI-NEXT: v_mov_b32_e32 v60, v48
+; SI-NEXT: v_mov_b32_e32 v60, v39
+; SI-NEXT: v_mov_b32_e32 v39, v48
; SI-NEXT: v_mov_b32_e32 v23, v52
; SI-NEXT: v_mov_b32_e32 v52, v41
; SI-NEXT: v_mov_b32_e32 v41, v46
; SI-NEXT: v_mov_b32_e32 v46, v59
-; SI-NEXT: buffer_load_dword v59, off, s[0:3], s32 offset:108 ; 4-byte Folded Reload
-; SI-NEXT: buffer_load_dword v32, off, s[0:3], s32 offset:104 ; 4-byte Folded Reload
-; SI-NEXT: buffer_load_dword v50, off, s[0:3], s32 offset:100 ; 4-byte Folded Reload
-; SI-NEXT: buffer_load_dword v35, off, s[0:3], s32 offset:96 ; 4-byte Folded Reload
-; SI-NEXT: buffer_load_dword v49, off, s[0:3], s32 offset:92 ; 4-byte Folded Reload
-; SI-NEXT: buffer_load_dword v37, off, s[0:3], s32 offset:88 ; 4-byte Folded Reload
-; SI-NEXT: buffer_load_dword v38, off, s[0:3], s32 offset:84 ; 4-byte Folded Reload
-; SI-NEXT: buffer_load_dword v39, off, s[0:3], s32 offset:80 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v59, off, s[0:3], s32 offset:104 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v32, off, s[0:3], s32 offset:100 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v50, off, s[0:3], s32 offset:96 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v35, off, s[0:3], s32 offset:92 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v36, off, s[0:3], s32 offset:88 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v37, off, s[0:3], s32 offset:84 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v38, off, s[0:3], s32 offset:80 ; 4-byte Folded Reload
; SI-NEXT: buffer_load_dword v48, off, s[0:3], s32 offset:76 ; 4-byte Folded Reload
; SI-NEXT: s_branch .LBB47_2
;
@@ -33110,13 +33131,13 @@ define inreg <11 x double> @bitcast_v44f16_to_v11f64_scalar(<44 x half> inreg %a
; SI-NEXT: buffer_store_dword v61, off, s[0:3], s32 offset:8 ; 4-byte Folded Spill
; SI-NEXT: buffer_store_dword v62, off, s[0:3], s32 offset:4 ; 4-byte Folded Spill
; SI-NEXT: buffer_store_dword v63, off, s[0:3], s32 ; 4-byte Folded Spill
+; SI-NEXT: v_cvt_f16_f32_e32 v39, v1
; SI-NEXT: s_waitcnt expcnt(3)
-; SI-NEXT: v_cvt_f16_f32_e32 v60, v1
-; SI-NEXT: v_cvt_f16_f32_e32 v48, v0
-; SI-NEXT: v_cvt_f16_f32_e32 v39, v3
+; SI-NEXT: v_cvt_f16_f32_e32 v60, v0
+; SI-NEXT: v_cvt_f16_f32_e32 v48, v3
; SI-NEXT: v_cvt_f16_f32_e32 v38, v2
; SI-NEXT: v_cvt_f16_f32_e32 v37, v5
-; SI-NEXT: v_cvt_f16_f32_e32 v49, v4
+; SI-NEXT: v_cvt_f16_f32_e32 v36, v4
; SI-NEXT: v_cvt_f16_f32_e32 v35, v7
; SI-NEXT: v_cvt_f16_f32_e32 v50, v6
; SI-NEXT: v_cvt_f16_f32_e32 v32, v9
@@ -33144,37 +33165,36 @@ define inreg <11 x double> @bitcast_v44f16_to_v11f64_scalar(<44 x half> inreg %a
; SI-NEXT: v_cvt_f16_f32_e32 v51, s17
; SI-NEXT: v_cvt_f16_f32_e32 v33, s16
; SI-NEXT: v_cvt_f16_f32_e32 v1, s19
-; SI-NEXT: v_cvt_f16_f32_e32 v34, s18
-; SI-NEXT: v_cvt_f16_f32_e32 v2, s21
-; SI-NEXT: v_cvt_f16_f32_e32 v3, s20
+; SI-NEXT: v_cvt_f16_f32_e32 v2, s18
+; SI-NEXT: v_cvt_f16_f32_e32 v3, s21
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_cvt_f16_f32_e32 v63, s23
-; SI-NEXT: v_cvt_f16_f32_e32 v62, s22
+; SI-NEXT: v_cvt_f16_f32_e32 v63, s20
+; SI-NEXT: v_cvt_f16_f32_e32 v62, s23
+; SI-NEXT: v_cvt_f16_f32_e32 v34, s22
; SI-NEXT: v_cvt_f16_f32_e32 v61, s25
-; SI-NEXT: v_cvt_f16_f32_e32 v36, s24
+; SI-NEXT: v_cvt_f16_f32_e32 v49, s24
; SI-NEXT: v_cvt_f16_f32_e32 v29, s27
; SI-NEXT: v_cvt_f16_f32_e32 v28, s26
; SI-NEXT: v_cvt_f16_f32_e32 v27, s29
; SI-NEXT: v_cvt_f16_f32_e32 v26, s28
; SI-NEXT: v_cmp_ne_u32_e32 vcc, 0, v30
; SI-NEXT: s_and_b64 s[4:5], vcc, exec
-; SI-NEXT: buffer_store_dword v3, off, s[0:3], s32 offset:64 ; 4-byte Folded Spill
-; SI-NEXT: buffer_store_dword v2, off, s[0:3], s32 offset:68 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v2, off, s[0:3], s32 offset:64 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v3, off, s[0:3], s32 offset:68 ; 4-byte Folded Spill
; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:72 ; 4-byte Folded Spill
; SI-NEXT: s_cbranch_scc0 .LBB55_4
; SI-NEXT: ; %bb.1: ; %cmp.false
-; SI-NEXT: s_waitcnt expcnt(1)
-; SI-NEXT: v_lshlrev_b32_e32 v2, 16, v2
-; SI-NEXT: v_lshlrev_b32_e32 v0, 16, v51
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_lshlrev_b32_e32 v1, 16, v1
-; SI-NEXT: v_or_b32_e32 v2, v3, v2
-; SI-NEXT: v_lshlrev_b32_e32 v3, 16, v63
+; SI-NEXT: v_lshlrev_b32_e32 v0, 16, v51
+; SI-NEXT: v_or_b32_e32 v1, v2, v1
+; SI-NEXT: v_lshlrev_b32_e32 v2, 16, v3
+; SI-NEXT: v_lshlrev_b32_e32 v3, 16, v62
; SI-NEXT: v_lshlrev_b32_e32 v4, 16, v61
; SI-NEXT: v_lshlrev_b32_e32 v5, 16, v29
; SI-NEXT: v_lshlrev_b32_e32 v6, 16, v27
-; SI-NEXT: v_lshlrev_b32_e32 v7, 16, v60
-; SI-NEXT: v_lshlrev_b32_e32 v8, 16, v39
+; SI-NEXT: v_lshlrev_b32_e32 v7, 16, v39
+; SI-NEXT: v_lshlrev_b32_e32 v8, 16, v48
; SI-NEXT: v_lshlrev_b32_e32 v9, 16, v37
; SI-NEXT: v_lshlrev_b32_e32 v10, 16, v35
; SI-NEXT: v_lshlrev_b32_e32 v11, 16, v32
@@ -33189,14 +33209,14 @@ define inreg <11 x double> @bitcast_v44f16_to_v11f64_scalar(<44 x half> inreg %a
; SI-NEXT: v_lshlrev_b32_e32 v20, 16, v25
; SI-NEXT: v_lshlrev_b32_e32 v21, 16, v22
; SI-NEXT: v_or_b32_e32 v0, v33, v0
-; SI-NEXT: v_or_b32_e32 v1, v34, v1
-; SI-NEXT: v_or_b32_e32 v3, v62, v3
-; SI-NEXT: v_or_b32_e32 v4, v36, v4
+; SI-NEXT: v_or_b32_e32 v2, v63, v2
+; SI-NEXT: v_or_b32_e32 v3, v34, v3
+; SI-NEXT: v_or_b32_e32 v4, v49, v4
; SI-NEXT: v_or_b32_e32 v5, v28, v5
; SI-NEXT: v_or_b32_e32 v6, v26, v6
-; SI-NEXT: v_or_b32_e32 v7, v48, v7
+; SI-NEXT: v_or_b32_e32 v7, v60, v7
; SI-NEXT: v_or_b32_e32 v8, v38, v8
-; SI-NEXT: v_or_b32_e32 v9, v49, v9
+; SI-NEXT: v_or_b32_e32 v9, v36, v9
; SI-NEXT: v_or_b32_e32 v10, v50, v10
; SI-NEXT: v_or_b32_e32 v11, v59, v11
; SI-NEXT: v_or_b32_e32 v12, v57, v12
@@ -33212,65 +33232,64 @@ define inreg <11 x double> @bitcast_v44f16_to_v11f64_scalar(<44 x half> inreg %a
; SI-NEXT: s_cbranch_execnz .LBB55_3
; SI-NEXT: .LBB55_2: ; %cmp.true
; SI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:72 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v3, off, s[0:3], s32 offset:64 ; 4-byte Folded Reload
; SI-NEXT: v_cvt_f32_f16_e32 v0, v51
; SI-NEXT: v_cvt_f32_f16_e32 v1, v33
-; SI-NEXT: v_cvt_f32_f16_e32 v3, v34
-; SI-NEXT: v_cvt_f32_f16_e32 v4, v62
+; SI-NEXT: s_waitcnt vmcnt(10)
+; SI-NEXT: v_cvt_f32_f16_e32 v4, v34
+; SI-NEXT: v_cvt_f32_f16_e32 v5, v49
; SI-NEXT: v_add_f32_e32 v0, 0x38000000, v0
; SI-NEXT: v_add_f32_e32 v1, 0x38000000, v1
; SI-NEXT: v_cvt_f16_f32_e32 v0, v0
-; SI-NEXT: v_add_f32_e32 v3, 0x38000000, v3
; SI-NEXT: v_cvt_f16_f32_e32 v1, v1
-; SI-NEXT: v_cvt_f16_f32_e32 v3, v3
-; SI-NEXT: v_lshlrev_b32_e32 v0, 16, v0
; SI-NEXT: v_add_f32_e32 v4, 0x38000000, v4
-; SI-NEXT: v_or_b32_e32 v0, v1, v0
; SI-NEXT: v_cvt_f16_f32_e32 v4, v4
-; SI-NEXT: v_cvt_f32_f16_e32 v5, v36
-; SI-NEXT: v_cvt_f32_f16_e32 v6, v28
-; SI-NEXT: v_cvt_f32_f16_e32 v7, v26
-; SI-NEXT: s_waitcnt vmcnt(1)
-; SI-NEXT: v_cvt_f32_f16_e32 v8, v48
+; SI-NEXT: v_lshlrev_b32_e32 v0, 16, v0
+; SI-NEXT: v_or_b32_e32 v0, v1, v0
; SI-NEXT: v_add_f32_e32 v5, 0x38000000, v5
; SI-NEXT: v_cvt_f16_f32_e32 v5, v5
+; SI-NEXT: v_cvt_f32_f16_e32 v6, v28
+; SI-NEXT: v_cvt_f32_f16_e32 v7, v26
+; SI-NEXT: v_cvt_f32_f16_e32 v8, v60
+; SI-NEXT: s_waitcnt vmcnt(3)
+; SI-NEXT: v_cvt_f32_f16_e32 v9, v38
; SI-NEXT: v_add_f32_e32 v6, 0x38000000, v6
; SI-NEXT: v_cvt_f16_f32_e32 v6, v6
; SI-NEXT: v_add_f32_e32 v7, 0x38000000, v7
; SI-NEXT: v_cvt_f16_f32_e32 v7, v7
; SI-NEXT: v_add_f32_e32 v8, 0x38000000, v8
; SI-NEXT: v_cvt_f16_f32_e32 v8, v8
-; SI-NEXT: v_cvt_f32_f16_e32 v9, v38
; SI-NEXT: v_cvt_f32_f16_e32 v10, v37
-; SI-NEXT: v_cvt_f32_f16_e32 v11, v49
-; SI-NEXT: v_cvt_f32_f16_e32 v12, v50
; SI-NEXT: v_add_f32_e32 v9, 0x38000000, v9
-; SI-NEXT: v_add_f32_e32 v10, 0x38000000, v10
; SI-NEXT: v_cvt_f16_f32_e32 v9, v9
+; SI-NEXT: v_cvt_f32_f16_e32 v11, v36
+; SI-NEXT: v_add_f32_e32 v10, 0x38000000, v10
; SI-NEXT: v_cvt_f16_f32_e32 v10, v10
+; SI-NEXT: v_cvt_f32_f16_e32 v12, v50
; SI-NEXT: v_add_f32_e32 v11, 0x38000000, v11
; SI-NEXT: v_cvt_f16_f32_e32 v11, v11
+; SI-NEXT: v_cvt_f32_f16_e32 v13, v58
; SI-NEXT: v_add_f32_e32 v12, 0x38000000, v12
; SI-NEXT: v_cvt_f16_f32_e32 v12, v12
-; SI-NEXT: v_cvt_f32_f16_e32 v13, v58
; SI-NEXT: v_cvt_f32_f16_e32 v14, v57
-; SI-NEXT: v_cvt_f32_f16_e32 v15, v47
-; SI-NEXT: v_cvt_f32_f16_e32 v16, v44
; SI-NEXT: v_add_f32_e32 v13, 0x38000000, v13
; SI-NEXT: v_cvt_f16_f32_e32 v13, v13
+; SI-NEXT: v_cvt_f32_f16_e32 v15, v47
; SI-NEXT: v_add_f32_e32 v14, 0x38000000, v14
; SI-NEXT: v_cvt_f16_f32_e32 v14, v14
+; SI-NEXT: v_cvt_f32_f16_e32 v16, v44
; SI-NEXT: v_add_f32_e32 v15, 0x38000000, v15
; SI-NEXT: v_cvt_f16_f32_e32 v15, v15
+; SI-NEXT: v_cvt_f32_f16_e32 v17, v43
; SI-NEXT: v_add_f32_e32 v16, 0x38000000, v16
; SI-NEXT: v_cvt_f16_f32_e32 v16, v16
-; SI-NEXT: v_cvt_f32_f16_e32 v17, v43
; SI-NEXT: v_cvt_f32_f16_e32 v18, v41
-; SI-NEXT: v_cvt_f32_f16_e32 v19, v54
-; SI-NEXT: v_cvt_f32_f16_e32 v20, v53
; SI-NEXT: v_add_f32_e32 v17, 0x38000000, v17
; SI-NEXT: v_cvt_f16_f32_e32 v17, v17
+; SI-NEXT: v_cvt_f32_f16_e32 v19, v54
; SI-NEXT: v_add_f32_e32 v18, 0x38000000, v18
; SI-NEXT: v_cvt_f16_f32_e32 v18, v18
+; SI-NEXT: v_cvt_f32_f16_e32 v20, v53
; SI-NEXT: v_add_f32_e32 v19, 0x38000000, v19
; SI-NEXT: v_cvt_f16_f32_e32 v19, v19
; SI-NEXT: v_cvt_f32_f16_e32 v21, v31
@@ -33284,25 +33303,27 @@ define inreg <11 x double> @bitcast_v44f16_to_v11f64_scalar(<44 x half> inreg %a
; SI-NEXT: v_cvt_f16_f32_e32 v22, v22
; SI-NEXT: v_add_f32_e32 v23, 0x38000000, v23
; SI-NEXT: v_cvt_f16_f32_e32 v23, v23
-; SI-NEXT: s_waitcnt vmcnt(0)
+; SI-NEXT: s_waitcnt vmcnt(1)
; SI-NEXT: v_cvt_f32_f16_e32 v2, v2
+; SI-NEXT: s_waitcnt vmcnt(0)
+; SI-NEXT: v_cvt_f32_f16_e32 v3, v3
; SI-NEXT: v_add_f32_e32 v2, 0x38000000, v2
; SI-NEXT: v_cvt_f16_f32_e32 v2, v2
+; SI-NEXT: v_add_f32_e32 v3, 0x38000000, v3
+; SI-NEXT: v_cvt_f16_f32_e32 v3, v3
; SI-NEXT: v_lshlrev_b32_e32 v2, 16, v2
; SI-NEXT: v_or_b32_e32 v1, v3, v2
; SI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:68 ; 4-byte Folded Reload
-; SI-NEXT: buffer_load_dword v3, off, s[0:3], s32 offset:64 ; 4-byte Folded Reload
-; SI-NEXT: s_waitcnt vmcnt(1)
-; SI-NEXT: v_cvt_f32_f16_e32 v2, v2
+; SI-NEXT: v_cvt_f32_f16_e32 v3, v63
+; SI-NEXT: v_add_f32_e32 v3, 0x38000000, v3
+; SI-NEXT: v_cvt_f16_f32_e32 v3, v3
; SI-NEXT: s_waitcnt vmcnt(0)
-; SI-NEXT: v_cvt_f32_f16_e32 v3, v3
+; SI-NEXT: v_cvt_f32_f16_e32 v2, v2
; SI-NEXT: v_add_f32_e32 v2, 0x38000000, v2
; SI-NEXT: v_cvt_f16_f32_e32 v2, v2
-; SI-NEXT: v_add_f32_e32 v3, 0x38000000, v3
-; SI-NEXT: v_cvt_f16_f32_e32 v3, v3
; SI-NEXT: v_lshlrev_b32_e32 v2, 16, v2
; SI-NEXT: v_or_b32_e32 v2, v3, v2
-; SI-NEXT: v_cvt_f32_f16_e32 v3, v63
+; SI-NEXT: v_cvt_f32_f16_e32 v3, v62
; SI-NEXT: v_add_f32_e32 v3, 0x38000000, v3
; SI-NEXT: v_cvt_f16_f32_e32 v3, v3
; SI-NEXT: v_lshlrev_b32_e32 v3, 16, v3
@@ -33322,12 +33343,12 @@ define inreg <11 x double> @bitcast_v44f16_to_v11f64_scalar(<44 x half> inreg %a
; SI-NEXT: v_cvt_f16_f32_e32 v6, v6
; SI-NEXT: v_lshlrev_b32_e32 v6, 16, v6
; SI-NEXT: v_or_b32_e32 v6, v7, v6
-; SI-NEXT: v_cvt_f32_f16_e32 v7, v60
+; SI-NEXT: v_cvt_f32_f16_e32 v7, v39
; SI-NEXT: v_add_f32_e32 v7, 0x38000000, v7
; SI-NEXT: v_cvt_f16_f32_e32 v7, v7
; SI-NEXT: v_lshlrev_b32_e32 v7, 16, v7
; SI-NEXT: v_or_b32_e32 v7, v8, v7
-; SI-NEXT: v_cvt_f32_f16_e32 v8, v39
+; SI-NEXT: v_cvt_f32_f16_e32 v8, v48
; SI-NEXT: v_add_f32_e32 v8, 0x38000000, v8
; SI-NEXT: v_cvt_f16_f32_e32 v8, v8
; SI-NEXT: v_lshlrev_b32_e32 v8, 16, v8
@@ -33415,86 +33436,89 @@ define inreg <11 x double> @bitcast_v44f16_to_v11f64_scalar(<44 x half> inreg %a
; SI-NEXT: s_setpc_b64 s[30:31]
; SI-NEXT: .LBB55_4:
; SI-NEXT: buffer_store_dword v48, off, s[0:3], s32 offset:76 ; 4-byte Folded Spill
-; SI-NEXT: buffer_store_dword v39, off, s[0:3], s32 offset:80 ; 4-byte Folded Spill
-; SI-NEXT: buffer_store_dword v38, off, s[0:3], s32 offset:84 ; 4-byte Folded Spill
-; SI-NEXT: buffer_store_dword v37, off, s[0:3], s32 offset:88 ; 4-byte Folded Spill
-; SI-NEXT: buffer_store_dword v49, off, s[0:3], s32 offset:92 ; 4-byte Folded Spill
-; SI-NEXT: buffer_store_dword v35, off, s[0:3], s32 offset:96 ; 4-byte Folded Spill
-; SI-NEXT: buffer_store_dword v50, off, s[0:3], s32 offset:100 ; 4-byte Folded Spill
-; SI-NEXT: buffer_store_dword v32, off, s[0:3], s32 offset:104 ; 4-byte Folded Spill
-; SI-NEXT: buffer_store_dword v59, off, s[0:3], s32 offset:108 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v38, off, s[0:3], s32 offset:80 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v37, off, s[0:3], s32 offset:84 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v36, off, s[0:3], s32 offset:88 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v35, off, s[0:3], s32 offset:92 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v50, off, s[0:3], s32 offset:96 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v32, off, s[0:3], s32 offset:100 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v59, off, s[0:3], s32 offset:104 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_mov_b32_e32 v59, v46
; SI-NEXT: v_mov_b32_e32 v46, v41
; SI-NEXT: v_mov_b32_e32 v41, v52
; SI-NEXT: v_mov_b32_e32 v52, v23
-; SI-NEXT: v_mov_b32_e32 v48, v60
+; SI-NEXT: v_mov_b32_e32 v48, v39
+; SI-NEXT: v_mov_b32_e32 v39, v60
; SI-NEXT: v_mov_b32_e32 v60, v47
; SI-NEXT: v_mov_b32_e32 v47, v42
; SI-NEXT: v_mov_b32_e32 v42, v53
; SI-NEXT: v_mov_b32_e32 v53, v22
-; SI-NEXT: v_mov_b32_e32 v35, v61
+; SI-NEXT: buffer_store_dword v34, off, s[0:3], s32 offset:108 ; 4-byte Folded Spill
+; SI-NEXT: s_waitcnt expcnt(0)
+; SI-NEXT: v_mov_b32_e32 v34, v61
; SI-NEXT: v_mov_b32_e32 v61, v56
; SI-NEXT: v_mov_b32_e32 v56, v43
; SI-NEXT: v_mov_b32_e32 v43, v54
; SI-NEXT: v_mov_b32_e32 v54, v24
-; SI-NEXT: v_mov_b32_e32 v50, v34
-; SI-NEXT: v_mov_b32_e32 v34, v62
+; SI-NEXT: v_mov_b32_e32 v50, v62
; SI-NEXT: v_mov_b32_e32 v62, v57
; SI-NEXT: v_mov_b32_e32 v57, v44
; SI-NEXT: v_mov_b32_e32 v44, v55
; SI-NEXT: v_mov_b32_e32 v55, v25
-; SI-NEXT: v_mov_b32_e32 v32, v33
+; SI-NEXT: v_mov_b32_e32 v32, v51
+; SI-NEXT: v_mov_b32_e32 v51, v33
; SI-NEXT: v_mov_b32_e32 v33, v63
; SI-NEXT: v_mov_b32_e32 v63, v58
; SI-NEXT: v_mov_b32_e32 v58, v45
; SI-NEXT: v_mov_b32_e32 v45, v40
; SI-NEXT: v_mov_b32_e32 v40, v31
-; SI-NEXT: v_mov_b32_e32 v39, v26
-; SI-NEXT: v_mov_b32_e32 v38, v27
-; SI-NEXT: v_mov_b32_e32 v37, v28
-; SI-NEXT: v_mov_b32_e32 v49, v36
-; SI-NEXT: v_mov_b32_e32 v36, v29
+; SI-NEXT: v_mov_b32_e32 v38, v26
+; SI-NEXT: v_mov_b32_e32 v37, v27
+; SI-NEXT: v_mov_b32_e32 v36, v28
+; SI-NEXT: v_mov_b32_e32 v35, v49
+; SI-NEXT: v_mov_b32_e32 v49, v29
; SI-NEXT: ; implicit-def: $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7_vgpr8_vgpr9_vgpr10_vgpr11_vgpr12_vgpr13_vgpr14_vgpr15_vgpr16_vgpr17_vgpr18_vgpr19_vgpr20_vgpr21_vgpr22_vgpr23_vgpr24_vgpr25_vgpr26_vgpr27_vgpr28_vgpr29_vgpr30_vgpr31
-; SI-NEXT: v_mov_b32_e32 v29, v36
-; SI-NEXT: v_mov_b32_e32 v36, v49
-; SI-NEXT: v_mov_b32_e32 v28, v37
-; SI-NEXT: v_mov_b32_e32 v27, v38
-; SI-NEXT: v_mov_b32_e32 v26, v39
+; SI-NEXT: v_mov_b32_e32 v29, v49
+; SI-NEXT: v_mov_b32_e32 v49, v35
+; SI-NEXT: v_mov_b32_e32 v28, v36
+; SI-NEXT: v_mov_b32_e32 v27, v37
+; SI-NEXT: v_mov_b32_e32 v26, v38
; SI-NEXT: v_mov_b32_e32 v31, v40
; SI-NEXT: v_mov_b32_e32 v40, v45
; SI-NEXT: v_mov_b32_e32 v45, v58
; SI-NEXT: v_mov_b32_e32 v58, v63
; SI-NEXT: v_mov_b32_e32 v63, v33
-; SI-NEXT: v_mov_b32_e32 v33, v32
+; SI-NEXT: v_mov_b32_e32 v33, v51
+; SI-NEXT: v_mov_b32_e32 v51, v32
; SI-NEXT: v_mov_b32_e32 v25, v55
; SI-NEXT: v_mov_b32_e32 v55, v44
; SI-NEXT: v_mov_b32_e32 v44, v57
; SI-NEXT: v_mov_b32_e32 v57, v62
-; SI-NEXT: v_mov_b32_e32 v62, v34
-; SI-NEXT: v_mov_b32_e32 v34, v50
+; SI-NEXT: v_mov_b32_e32 v62, v50
; SI-NEXT: v_mov_b32_e32 v24, v54
; SI-NEXT: v_mov_b32_e32 v54, v43
; SI-NEXT: v_mov_b32_e32 v43, v56
; SI-NEXT: v_mov_b32_e32 v56, v61
-; SI-NEXT: v_mov_b32_e32 v61, v35
+; SI-NEXT: v_mov_b32_e32 v61, v34
+; SI-NEXT: buffer_load_dword v34, off, s[0:3], s32 offset:108 ; 4-byte Folded Reload
; SI-NEXT: v_mov_b32_e32 v22, v53
; SI-NEXT: v_mov_b32_e32 v53, v42
; SI-NEXT: v_mov_b32_e32 v42, v47
; SI-NEXT: v_mov_b32_e32 v47, v60
-; SI-NEXT: v_mov_b32_e32 v60, v48
+; SI-NEXT: v_mov_b32_e32 v60, v39
+; SI-NEXT: v_mov_b32_e32 v39, v48
; SI-NEXT: v_mov_b32_e32 v23, v52
; SI-NEXT: v_mov_b32_e32 v52, v41
; SI-NEXT: v_mov_b32_e32 v41, v46
; SI-NEXT: v_mov_b32_e32 v46, v59
-; SI-NEXT: buffer_load_dword v59, off, s[0:3], s32 offset:108 ; 4-byte Folded Reload
-; SI-NEXT: buffer_load_dword v32, off, s[0:3], s32 offset:104 ; 4-byte Folded Reload
-; SI-NEXT: buffer_load_dword v50, off, s[0:3], s32 offset:100 ; 4-byte Folded Reload
-; SI-NEXT: buffer_load_dword v35, off, s[0:3], s32 offset:96 ; 4-byte Folded Reload
-; SI-NEXT: buffer_load_dword v49, off, s[0:3], s32 offset:92 ; 4-byte Folded Reload
-; SI-NEXT: buffer_load_dword v37, off, s[0:3], s32 offset:88 ; 4-byte Folded Reload
-; SI-NEXT: buffer_load_dword v38, off, s[0:3], s32 offset:84 ; 4-byte Folded Reload
-; SI-NEXT: buffer_load_dword v39, off, s[0:3], s32 offset:80 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v59, off, s[0:3], s32 offset:104 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v32, off, s[0:3], s32 offset:100 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v50, off, s[0:3], s32 offset:96 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v35, off, s[0:3], s32 offset:92 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v36, off, s[0:3], s32 offset:88 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v37, off, s[0:3], s32 offset:84 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v38, off, s[0:3], s32 offset:80 ; 4-byte Folded Reload
; SI-NEXT: buffer_load_dword v48, off, s[0:3], s32 offset:76 ; 4-byte Folded Reload
; SI-NEXT: s_branch .LBB55_2
;
diff --git a/llvm/test/CodeGen/AMDGPU/amdgcn.bitcast.768bit.ll b/llvm/test/CodeGen/AMDGPU/amdgcn.bitcast.768bit.ll
index a2bd1d30cc634..d84d3230f9538 100644
--- a/llvm/test/CodeGen/AMDGPU/amdgcn.bitcast.768bit.ll
+++ b/llvm/test/CodeGen/AMDGPU/amdgcn.bitcast.768bit.ll
@@ -3868,7 +3868,7 @@ define inreg <48 x i16> @bitcast_v24i32_to_v48i16_scalar(<24 x i32> inreg %a, i3
; GFX11-TRUE16-NEXT: v_readfirstlane_b32 s5, v3
; GFX11-TRUE16-NEXT: v_readfirstlane_b32 s6, v4
; GFX11-TRUE16-NEXT: v_readfirstlane_b32 s4, v5
-; GFX11-TRUE16-NEXT: s_mov_b32 s74, 0
+; GFX11-TRUE16-NEXT: s_mov_b32 s62, 0
; GFX11-TRUE16-NEXT: s_and_b32 s10, vcc_lo, exec_lo
; GFX11-TRUE16-NEXT: s_cbranch_scc0 .LBB13_4
; GFX11-TRUE16-NEXT: ; %bb.1: ; %cmp.false
@@ -3892,11 +3892,11 @@ define inreg <48 x i16> @bitcast_v24i32_to_v48i16_scalar(<24 x i32> inreg %a, i3
; GFX11-TRUE16-NEXT: s_lshr_b32 s59, s18, 16
; GFX11-TRUE16-NEXT: s_lshr_b32 s60, s17, 16
; GFX11-TRUE16-NEXT: s_lshr_b32 s61, s16, 16
-; GFX11-TRUE16-NEXT: s_lshr_b32 s62, s3, 16
-; GFX11-TRUE16-NEXT: s_lshr_b32 s63, s2, 16
-; GFX11-TRUE16-NEXT: s_lshr_b32 s72, s1, 16
-; GFX11-TRUE16-NEXT: s_lshr_b32 s73, s0, 16
-; GFX11-TRUE16-NEXT: s_and_not1_b32 vcc_lo, exec_lo, s74
+; GFX11-TRUE16-NEXT: s_lshr_b32 s63, s3, 16
+; GFX11-TRUE16-NEXT: s_lshr_b32 s72, s2, 16
+; GFX11-TRUE16-NEXT: s_lshr_b32 s73, s1, 16
+; GFX11-TRUE16-NEXT: s_lshr_b32 s74, s0, 16
+; GFX11-TRUE16-NEXT: s_and_not1_b32 vcc_lo, exec_lo, s62
; GFX11-TRUE16-NEXT: s_cbranch_vccnz .LBB13_3
; GFX11-TRUE16-NEXT: .LBB13_2: ; %cmp.true
; GFX11-TRUE16-NEXT: s_add_i32 s4, s4, 3
@@ -3943,16 +3943,18 @@ define inreg <48 x i16> @bitcast_v24i32_to_v48i16_scalar(<24 x i32> inreg %a, i3
; GFX11-TRUE16-NEXT: s_lshr_b32 s59, s18, 16
; GFX11-TRUE16-NEXT: s_lshr_b32 s60, s17, 16
; GFX11-TRUE16-NEXT: s_lshr_b32 s61, s16, 16
-; GFX11-TRUE16-NEXT: s_lshr_b32 s62, s3, 16
-; GFX11-TRUE16-NEXT: s_lshr_b32 s63, s2, 16
-; GFX11-TRUE16-NEXT: s_lshr_b32 s72, s1, 16
-; GFX11-TRUE16-NEXT: s_lshr_b32 s73, s0, 16
+; GFX11-TRUE16-NEXT: s_lshr_b32 s63, s3, 16
+; GFX11-TRUE16-NEXT: s_lshr_b32 s72, s2, 16
+; GFX11-TRUE16-NEXT: s_lshr_b32 s73, s1, 16
+; GFX11-TRUE16-NEXT: s_lshr_b32 s74, s0, 16
; GFX11-TRUE16-NEXT: .LBB13_3: ; %end
; GFX11-TRUE16-NEXT: s_delay_alu instid0(SALU_CYCLE_1)
-; GFX11-TRUE16-NEXT: s_pack_ll_b32_b16 s0, s0, s73
-; GFX11-TRUE16-NEXT: s_pack_ll_b32_b16 s1, s1, s72
-; GFX11-TRUE16-NEXT: s_pack_ll_b32_b16 s2, s2, s63
-; GFX11-TRUE16-NEXT: s_pack_ll_b32_b16 s3, s3, s62
+; GFX11-TRUE16-NEXT: s_mov_b32 s62, s74
+; GFX11-TRUE16-NEXT: s_pack_ll_b32_b16 s1, s1, s73
+; GFX11-TRUE16-NEXT: s_pack_ll_b32_b16 s0, s0, s62
+; GFX11-TRUE16-NEXT: s_mov_b32 s62, s72
+; GFX11-TRUE16-NEXT: s_pack_ll_b32_b16 s3, s3, s63
+; GFX11-TRUE16-NEXT: s_pack_ll_b32_b16 s2, s2, s62
; GFX11-TRUE16-NEXT: s_pack_ll_b32_b16 s16, s16, s61
; GFX11-TRUE16-NEXT: s_pack_ll_b32_b16 s17, s17, s60
; GFX11-TRUE16-NEXT: s_pack_ll_b32_b16 s18, s18, s59
@@ -3987,10 +3989,10 @@ define inreg <48 x i16> @bitcast_v24i32_to_v48i16_scalar(<24 x i32> inreg %a, i3
; GFX11-TRUE16-NEXT: v_dual_mov_b32 v22, s6 :: v_dual_mov_b32 v23, s4
; GFX11-TRUE16-NEXT: s_setpc_b64 s[30:31]
; GFX11-TRUE16-NEXT: .LBB13_4:
+; GFX11-TRUE16-NEXT: ; implicit-def: $sgpr74_lo16
; GFX11-TRUE16-NEXT: ; implicit-def: $sgpr73_lo16
; GFX11-TRUE16-NEXT: ; implicit-def: $sgpr72_lo16
; GFX11-TRUE16-NEXT: ; implicit-def: $sgpr63_lo16
-; GFX11-TRUE16-NEXT: ; implicit-def: $sgpr62_lo16
; GFX11-TRUE16-NEXT: ; implicit-def: $sgpr61_lo16
; GFX11-TRUE16-NEXT: ; implicit-def: $sgpr60_lo16
; GFX11-TRUE16-NEXT: ; implicit-def: $sgpr59_lo16
@@ -6144,9 +6146,9 @@ define inreg <24 x i32> @bitcast_v48i16_to_v24i32_scalar(<48 x i16> inreg %a, i3
; GFX11-FAKE16-NEXT: v_and_b32_e32 v48, 0xffff, v3
; GFX11-FAKE16-NEXT: v_and_b32_e32 v39, 0xffff, v4
; GFX11-FAKE16-NEXT: v_and_b32_e32 v38, 0xffff, v5
-; GFX11-FAKE16-NEXT: s_lshr_b32 s41, s29, 16
-; GFX11-FAKE16-NEXT: s_lshr_b32 s42, s28, 16
-; GFX11-FAKE16-NEXT: s_lshr_b32 s15, s27, 16
+; GFX11-FAKE16-NEXT: s_lshr_b32 s40, s29, 16
+; GFX11-FAKE16-NEXT: s_lshr_b32 s41, s28, 16
+; GFX11-FAKE16-NEXT: s_lshr_b32 s42, s27, 16
; GFX11-FAKE16-NEXT: s_lshr_b32 s43, s26, 16
; GFX11-FAKE16-NEXT: s_lshr_b32 s14, s25, 16
; GFX11-FAKE16-NEXT: s_lshr_b32 s13, s24, 16
@@ -6162,7 +6164,7 @@ define inreg <24 x i32> @bitcast_v48i16_to_v24i32_scalar(<48 x i16> inreg %a, i3
; GFX11-FAKE16-NEXT: s_lshr_b32 s45, s2, 16
; GFX11-FAKE16-NEXT: s_lshr_b32 s46, s1, 16
; GFX11-FAKE16-NEXT: s_lshr_b32 s4, s0, 16
-; GFX11-FAKE16-NEXT: s_mov_b32 s40, 0
+; GFX11-FAKE16-NEXT: s_mov_b32 s15, 0
; GFX11-FAKE16-NEXT: s_pack_ll_b32_b16 s4, s0, s4
; GFX11-FAKE16-NEXT: s_pack_ll_b32_b16 s1, s1, s46
; GFX11-FAKE16-NEXT: s_pack_ll_b32_b16 s2, s2, s45
@@ -6178,9 +6180,9 @@ define inreg <24 x i32> @bitcast_v48i16_to_v24i32_scalar(<48 x i16> inreg %a, i3
; GFX11-FAKE16-NEXT: s_pack_ll_b32_b16 s13, s24, s13
; GFX11-FAKE16-NEXT: s_pack_ll_b32_b16 s14, s25, s14
; GFX11-FAKE16-NEXT: s_pack_ll_b32_b16 s0, s26, s43
-; GFX11-FAKE16-NEXT: s_pack_ll_b32_b16 s15, s27, s15
-; GFX11-FAKE16-NEXT: s_pack_ll_b32_b16 s16, s28, s42
-; GFX11-FAKE16-NEXT: s_pack_ll_b32_b16 s17, s29, s41
+; GFX11-FAKE16-NEXT: s_pack_ll_b32_b16 s16, s27, s42
+; GFX11-FAKE16-NEXT: s_pack_ll_b32_b16 s17, s28, s41
+; GFX11-FAKE16-NEXT: s_pack_ll_b32_b16 s18, s29, s40
; GFX11-FAKE16-NEXT: s_and_b32 s47, vcc_lo, exec_lo
; GFX11-FAKE16-NEXT: s_cbranch_scc0 .LBB15_4
; GFX11-FAKE16-NEXT: ; %bb.1: ; %cmp.false
@@ -6197,9 +6199,9 @@ define inreg <24 x i32> @bitcast_v48i16_to_v24i32_scalar(<48 x i16> inreg %a, i3
; GFX11-FAKE16-NEXT: v_dual_mov_b32 v8, s9 :: v_dual_mov_b32 v9, s10
; GFX11-FAKE16-NEXT: v_dual_mov_b32 v10, s11 :: v_dual_mov_b32 v11, s12
; GFX11-FAKE16-NEXT: v_dual_mov_b32 v12, s13 :: v_dual_mov_b32 v13, s14
-; GFX11-FAKE16-NEXT: v_dual_mov_b32 v14, s0 :: v_dual_mov_b32 v15, s15
-; GFX11-FAKE16-NEXT: v_dual_mov_b32 v16, s16 :: v_dual_mov_b32 v17, s17
-; GFX11-FAKE16-NEXT: s_and_not1_b32 vcc_lo, exec_lo, s40
+; GFX11-FAKE16-NEXT: v_dual_mov_b32 v14, s0 :: v_dual_mov_b32 v15, s16
+; GFX11-FAKE16-NEXT: v_dual_mov_b32 v16, s17 :: v_dual_mov_b32 v17, s18
+; GFX11-FAKE16-NEXT: s_and_not1_b32 vcc_lo, exec_lo, s15
; GFX11-FAKE16-NEXT: s_cbranch_vccnz .LBB15_3
; GFX11-FAKE16-NEXT: .LBB15_2: ; %cmp.true
; GFX11-FAKE16-NEXT: v_lshl_or_b32 v18, v37, 16, v51
@@ -6223,9 +6225,9 @@ define inreg <24 x i32> @bitcast_v48i16_to_v24i32_scalar(<48 x i16> inreg %a, i3
; GFX11-FAKE16-NEXT: v_pk_add_u16 v12, s13, 3 op_sel_hi:[1,0]
; GFX11-FAKE16-NEXT: v_pk_add_u16 v13, s14, 3 op_sel_hi:[1,0]
; GFX11-FAKE16-NEXT: v_pk_add_u16 v14, s0, 3 op_sel_hi:[1,0]
-; GFX11-FAKE16-NEXT: v_pk_add_u16 v15, s15, 3 op_sel_hi:[1,0]
-; GFX11-FAKE16-NEXT: v_pk_add_u16 v16, s16, 3 op_sel_hi:[1,0]
-; GFX11-FAKE16-NEXT: v_pk_add_u16 v17, s17, 3 op_sel_hi:[1,0]
+; GFX11-FAKE16-NEXT: v_pk_add_u16 v15, s16, 3 op_sel_hi:[1,0]
+; GFX11-FAKE16-NEXT: v_pk_add_u16 v16, s17, 3 op_sel_hi:[1,0]
+; GFX11-FAKE16-NEXT: v_pk_add_u16 v17, s18, 3 op_sel_hi:[1,0]
; GFX11-FAKE16-NEXT: v_pk_add_u16 v18, v18, 3 op_sel_hi:[1,0]
; GFX11-FAKE16-NEXT: v_pk_add_u16 v19, v19, 3 op_sel_hi:[1,0]
; GFX11-FAKE16-NEXT: v_pk_add_u16 v20, v20, 3 op_sel_hi:[1,0]
@@ -8055,7 +8057,7 @@ define inreg <48 x half> @bitcast_v24i32_to_v48f16_scalar(<24 x i32> inreg %a, i
; GFX11-TRUE16-NEXT: v_readfirstlane_b32 s5, v3
; GFX11-TRUE16-NEXT: v_readfirstlane_b32 s6, v4
; GFX11-TRUE16-NEXT: v_readfirstlane_b32 s4, v5
-; GFX11-TRUE16-NEXT: s_mov_b32 s74, 0
+; GFX11-TRUE16-NEXT: s_mov_b32 s62, 0
; GFX11-TRUE16-NEXT: s_and_b32 s10, vcc_lo, exec_lo
; GFX11-TRUE16-NEXT: s_cbranch_scc0 .LBB17_4
; GFX11-TRUE16-NEXT: ; %bb.1: ; %cmp.false
@@ -8079,11 +8081,11 @@ define inreg <48 x half> @bitcast_v24i32_to_v48f16_scalar(<24 x i32> inreg %a, i
; GFX11-TRUE16-NEXT: s_lshr_b32 s59, s18, 16
; GFX11-TRUE16-NEXT: s_lshr_b32 s60, s17, 16
; GFX11-TRUE16-NEXT: s_lshr_b32 s61, s16, 16
-; GFX11-TRUE16-NEXT: s_lshr_b32 s62, s3, 16
-; GFX11-TRUE16-NEXT: s_lshr_b32 s63, s2, 16
-; GFX11-TRUE16-NEXT: s_lshr_b32 s72, s1, 16
-; GFX11-TRUE16-NEXT: s_lshr_b32 s73, s0, 16
-; GFX11-TRUE16-NEXT: s_and_not1_b32 vcc_lo, exec_lo, s74
+; GFX11-TRUE16-NEXT: s_lshr_b32 s63, s3, 16
+; GFX11-TRUE16-NEXT: s_lshr_b32 s72, s2, 16
+; GFX11-TRUE16-NEXT: s_lshr_b32 s73, s1, 16
+; GFX11-TRUE16-NEXT: s_lshr_b32 s74, s0, 16
+; GFX11-TRUE16-NEXT: s_and_not1_b32 vcc_lo, exec_lo, s62
; GFX11-TRUE16-NEXT: s_cbranch_vccnz .LBB17_3
; GFX11-TRUE16-NEXT: .LBB17_2: ; %cmp.true
; GFX11-TRUE16-NEXT: s_add_i32 s4, s4, 3
@@ -8130,16 +8132,18 @@ define inreg <48 x half> @bitcast_v24i32_to_v48f16_scalar(<24 x i32> inreg %a, i
; GFX11-TRUE16-NEXT: s_lshr_b32 s59, s18, 16
; GFX11-TRUE16-NEXT: s_lshr_b32 s60, s17, 16
; GFX11-TRUE16-NEXT: s_lshr_b32 s61, s16, 16
-; GFX11-TRUE16-NEXT: s_lshr_b32 s62, s3, 16
-; GFX11-TRUE16-NEXT: s_lshr_b32 s63, s2, 16
-; GFX11-TRUE16-NEXT: s_lshr_b32 s72, s1, 16
-; GFX11-TRUE16-NEXT: s_lshr_b32 s73, s0, 16
+; GFX11-TRUE16-NEXT: s_lshr_b32 s63, s3, 16
+; GFX11-TRUE16-NEXT: s_lshr_b32 s72, s2, 16
+; GFX11-TRUE16-NEXT: s_lshr_b32 s73, s1, 16
+; GFX11-TRUE16-NEXT: s_lshr_b32 s74, s0, 16
; GFX11-TRUE16-NEXT: .LBB17_3: ; %end
; GFX11-TRUE16-NEXT: s_delay_alu instid0(SALU_CYCLE_1)
-; GFX11-TRUE16-NEXT: s_pack_ll_b32_b16 s0, s0, s73
-; GFX11-TRUE16-NEXT: s_pack_ll_b32_b16 s1, s1, s72
-; GFX11-TRUE16-NEXT: s_pack_ll_b32_b16 s2, s2, s63
-; GFX11-TRUE16-NEXT: s_pack_ll_b32_b16 s3, s3, s62
+; GFX11-TRUE16-NEXT: s_mov_b32 s62, s74
+; GFX11-TRUE16-NEXT: s_pack_ll_b32_b16 s1, s1, s73
+; GFX11-TRUE16-NEXT: s_pack_ll_b32_b16 s0, s0, s62
+; GFX11-TRUE16-NEXT: s_mov_b32 s62, s72
+; GFX11-TRUE16-NEXT: s_pack_ll_b32_b16 s3, s3, s63
+; GFX11-TRUE16-NEXT: s_pack_ll_b32_b16 s2, s2, s62
; GFX11-TRUE16-NEXT: s_pack_ll_b32_b16 s16, s16, s61
; GFX11-TRUE16-NEXT: s_pack_ll_b32_b16 s17, s17, s60
; GFX11-TRUE16-NEXT: s_pack_ll_b32_b16 s18, s18, s59
@@ -8174,10 +8178,10 @@ define inreg <48 x half> @bitcast_v24i32_to_v48f16_scalar(<24 x i32> inreg %a, i
; GFX11-TRUE16-NEXT: v_dual_mov_b32 v22, s6 :: v_dual_mov_b32 v23, s4
; GFX11-TRUE16-NEXT: s_setpc_b64 s[30:31]
; GFX11-TRUE16-NEXT: .LBB17_4:
+; GFX11-TRUE16-NEXT: ; implicit-def: $sgpr74_lo16
; GFX11-TRUE16-NEXT: ; implicit-def: $sgpr73_lo16
; GFX11-TRUE16-NEXT: ; implicit-def: $sgpr72_lo16
; GFX11-TRUE16-NEXT: ; implicit-def: $sgpr63_lo16
-; GFX11-TRUE16-NEXT: ; implicit-def: $sgpr62_lo16
; GFX11-TRUE16-NEXT: ; implicit-def: $sgpr61_lo16
; GFX11-TRUE16-NEXT: ; implicit-def: $sgpr60_lo16
; GFX11-TRUE16-NEXT: ; implicit-def: $sgpr59_lo16
@@ -10603,9 +10607,9 @@ define inreg <24 x i32> @bitcast_v48f16_to_v24i32_scalar(<48 x half> inreg %a, i
; GFX11-FAKE16-NEXT: v_and_b32_e32 v48, 0xffff, v3
; GFX11-FAKE16-NEXT: v_and_b32_e32 v39, 0xffff, v4
; GFX11-FAKE16-NEXT: v_and_b32_e32 v38, 0xffff, v5
-; GFX11-FAKE16-NEXT: s_lshr_b32 s41, s29, 16
-; GFX11-FAKE16-NEXT: s_lshr_b32 s42, s28, 16
-; GFX11-FAKE16-NEXT: s_lshr_b32 s15, s27, 16
+; GFX11-FAKE16-NEXT: s_lshr_b32 s40, s29, 16
+; GFX11-FAKE16-NEXT: s_lshr_b32 s41, s28, 16
+; GFX11-FAKE16-NEXT: s_lshr_b32 s42, s27, 16
; GFX11-FAKE16-NEXT: s_lshr_b32 s43, s26, 16
; GFX11-FAKE16-NEXT: s_lshr_b32 s14, s25, 16
; GFX11-FAKE16-NEXT: s_lshr_b32 s13, s24, 16
@@ -10621,7 +10625,7 @@ define inreg <24 x i32> @bitcast_v48f16_to_v24i32_scalar(<48 x half> inreg %a, i
; GFX11-FAKE16-NEXT: s_lshr_b32 s45, s2, 16
; GFX11-FAKE16-NEXT: s_lshr_b32 s46, s1, 16
; GFX11-FAKE16-NEXT: s_lshr_b32 s4, s0, 16
-; GFX11-FAKE16-NEXT: s_mov_b32 s40, 0
+; GFX11-FAKE16-NEXT: s_mov_b32 s15, 0
; GFX11-FAKE16-NEXT: s_pack_ll_b32_b16 s4, s0, s4
; GFX11-FAKE16-NEXT: s_pack_ll_b32_b16 s1, s1, s46
; GFX11-FAKE16-NEXT: s_pack_ll_b32_b16 s2, s2, s45
@@ -10637,9 +10641,9 @@ define inreg <24 x i32> @bitcast_v48f16_to_v24i32_scalar(<48 x half> inreg %a, i
; GFX11-FAKE16-NEXT: s_pack_ll_b32_b16 s13, s24, s13
; GFX11-FAKE16-NEXT: s_pack_ll_b32_b16 s14, s25, s14
; GFX11-FAKE16-NEXT: s_pack_ll_b32_b16 s0, s26, s43
-; GFX11-FAKE16-NEXT: s_pack_ll_b32_b16 s15, s27, s15
-; GFX11-FAKE16-NEXT: s_pack_ll_b32_b16 s16, s28, s42
-; GFX11-FAKE16-NEXT: s_pack_ll_b32_b16 s17, s29, s41
+; GFX11-FAKE16-NEXT: s_pack_ll_b32_b16 s16, s27, s42
+; GFX11-FAKE16-NEXT: s_pack_ll_b32_b16 s17, s28, s41
+; GFX11-FAKE16-NEXT: s_pack_ll_b32_b16 s18, s29, s40
; GFX11-FAKE16-NEXT: s_and_b32 s47, vcc_lo, exec_lo
; GFX11-FAKE16-NEXT: s_cbranch_scc0 .LBB19_4
; GFX11-FAKE16-NEXT: ; %bb.1: ; %cmp.false
@@ -10656,9 +10660,9 @@ define inreg <24 x i32> @bitcast_v48f16_to_v24i32_scalar(<48 x half> inreg %a, i
; GFX11-FAKE16-NEXT: v_dual_mov_b32 v8, s9 :: v_dual_mov_b32 v9, s10
; GFX11-FAKE16-NEXT: v_dual_mov_b32 v10, s11 :: v_dual_mov_b32 v11, s12
; GFX11-FAKE16-NEXT: v_dual_mov_b32 v12, s13 :: v_dual_mov_b32 v13, s14
-; GFX11-FAKE16-NEXT: v_dual_mov_b32 v14, s0 :: v_dual_mov_b32 v15, s15
-; GFX11-FAKE16-NEXT: v_dual_mov_b32 v16, s16 :: v_dual_mov_b32 v17, s17
-; GFX11-FAKE16-NEXT: s_and_not1_b32 vcc_lo, exec_lo, s40
+; GFX11-FAKE16-NEXT: v_dual_mov_b32 v14, s0 :: v_dual_mov_b32 v15, s16
+; GFX11-FAKE16-NEXT: v_dual_mov_b32 v16, s17 :: v_dual_mov_b32 v17, s18
+; GFX11-FAKE16-NEXT: s_and_not1_b32 vcc_lo, exec_lo, s15
; GFX11-FAKE16-NEXT: s_cbranch_vccnz .LBB19_3
; GFX11-FAKE16-NEXT: .LBB19_2: ; %cmp.true
; GFX11-FAKE16-NEXT: v_lshl_or_b32 v18, v37, 16, v51
@@ -10682,9 +10686,9 @@ define inreg <24 x i32> @bitcast_v48f16_to_v24i32_scalar(<48 x half> inreg %a, i
; GFX11-FAKE16-NEXT: v_pk_add_f16 v12, 0x200, s13 op_sel_hi:[0,1]
; GFX11-FAKE16-NEXT: v_pk_add_f16 v13, 0x200, s14 op_sel_hi:[0,1]
; GFX11-FAKE16-NEXT: v_pk_add_f16 v14, 0x200, s0 op_sel_hi:[0,1]
-; GFX11-FAKE16-NEXT: v_pk_add_f16 v15, 0x200, s15 op_sel_hi:[0,1]
-; GFX11-FAKE16-NEXT: v_pk_add_f16 v16, 0x200, s16 op_sel_hi:[0,1]
-; GFX11-FAKE16-NEXT: v_pk_add_f16 v17, 0x200, s17 op_sel_hi:[0,1]
+; GFX11-FAKE16-NEXT: v_pk_add_f16 v15, 0x200, s16 op_sel_hi:[0,1]
+; GFX11-FAKE16-NEXT: v_pk_add_f16 v16, 0x200, s17 op_sel_hi:[0,1]
+; GFX11-FAKE16-NEXT: v_pk_add_f16 v17, 0x200, s18 op_sel_hi:[0,1]
; GFX11-FAKE16-NEXT: v_pk_add_f16 v18, 0x200, v18 op_sel_hi:[0,1]
; GFX11-FAKE16-NEXT: v_pk_add_f16 v19, 0x200, v19 op_sel_hi:[0,1]
; GFX11-FAKE16-NEXT: v_pk_add_f16 v20, 0x200, v20 op_sel_hi:[0,1]
@@ -15932,9 +15936,9 @@ define inreg <24 x float> @bitcast_v48i16_to_v24f32_scalar(<48 x i16> inreg %a,
; GFX11-FAKE16-NEXT: v_and_b32_e32 v48, 0xffff, v3
; GFX11-FAKE16-NEXT: v_and_b32_e32 v39, 0xffff, v4
; GFX11-FAKE16-NEXT: v_and_b32_e32 v38, 0xffff, v5
-; GFX11-FAKE16-NEXT: s_lshr_b32 s41, s29, 16
-; GFX11-FAKE16-NEXT: s_lshr_b32 s42, s28, 16
-; GFX11-FAKE16-NEXT: s_lshr_b32 s15, s27, 16
+; GFX11-FAKE16-NEXT: s_lshr_b32 s40, s29, 16
+; GFX11-FAKE16-NEXT: s_lshr_b32 s41, s28, 16
+; GFX11-FAKE16-NEXT: s_lshr_b32 s42, s27, 16
; GFX11-FAKE16-NEXT: s_lshr_b32 s43, s26, 16
; GFX11-FAKE16-NEXT: s_lshr_b32 s14, s25, 16
; GFX11-FAKE16-NEXT: s_lshr_b32 s13, s24, 16
@@ -15950,7 +15954,7 @@ define inreg <24 x float> @bitcast_v48i16_to_v24f32_scalar(<48 x i16> inreg %a,
; GFX11-FAKE16-NEXT: s_lshr_b32 s45, s2, 16
; GFX11-FAKE16-NEXT: s_lshr_b32 s46, s1, 16
; GFX11-FAKE16-NEXT: s_lshr_b32 s4, s0, 16
-; GFX11-FAKE16-NEXT: s_mov_b32 s40, 0
+; GFX11-FAKE16-NEXT: s_mov_b32 s15, 0
; GFX11-FAKE16-NEXT: s_pack_ll_b32_b16 s4, s0, s4
; GFX11-FAKE16-NEXT: s_pack_ll_b32_b16 s1, s1, s46
; GFX11-FAKE16-NEXT: s_pack_ll_b32_b16 s2, s2, s45
@@ -15966,9 +15970,9 @@ define inreg <24 x float> @bitcast_v48i16_to_v24f32_scalar(<48 x i16> inreg %a,
; GFX11-FAKE16-NEXT: s_pack_ll_b32_b16 s13, s24, s13
; GFX11-FAKE16-NEXT: s_pack_ll_b32_b16 s14, s25, s14
; GFX11-FAKE16-NEXT: s_pack_ll_b32_b16 s0, s26, s43
-; GFX11-FAKE16-NEXT: s_pack_ll_b32_b16 s15, s27, s15
-; GFX11-FAKE16-NEXT: s_pack_ll_b32_b16 s16, s28, s42
-; GFX11-FAKE16-NEXT: s_pack_ll_b32_b16 s17, s29, s41
+; GFX11-FAKE16-NEXT: s_pack_ll_b32_b16 s16, s27, s42
+; GFX11-FAKE16-NEXT: s_pack_ll_b32_b16 s17, s28, s41
+; GFX11-FAKE16-NEXT: s_pack_ll_b32_b16 s18, s29, s40
; GFX11-FAKE16-NEXT: s_and_b32 s47, vcc_lo, exec_lo
; GFX11-FAKE16-NEXT: s_cbranch_scc0 .LBB31_4
; GFX11-FAKE16-NEXT: ; %bb.1: ; %cmp.false
@@ -15985,9 +15989,9 @@ define inreg <24 x float> @bitcast_v48i16_to_v24f32_scalar(<48 x i16> inreg %a,
; GFX11-FAKE16-NEXT: v_dual_mov_b32 v8, s9 :: v_dual_mov_b32 v9, s10
; GFX11-FAKE16-NEXT: v_dual_mov_b32 v10, s11 :: v_dual_mov_b32 v11, s12
; GFX11-FAKE16-NEXT: v_dual_mov_b32 v12, s13 :: v_dual_mov_b32 v13, s14
-; GFX11-FAKE16-NEXT: v_dual_mov_b32 v14, s0 :: v_dual_mov_b32 v15, s15
-; GFX11-FAKE16-NEXT: v_dual_mov_b32 v16, s16 :: v_dual_mov_b32 v17, s17
-; GFX11-FAKE16-NEXT: s_and_not1_b32 vcc_lo, exec_lo, s40
+; GFX11-FAKE16-NEXT: v_dual_mov_b32 v14, s0 :: v_dual_mov_b32 v15, s16
+; GFX11-FAKE16-NEXT: v_dual_mov_b32 v16, s17 :: v_dual_mov_b32 v17, s18
+; GFX11-FAKE16-NEXT: s_and_not1_b32 vcc_lo, exec_lo, s15
; GFX11-FAKE16-NEXT: s_cbranch_vccnz .LBB31_3
; GFX11-FAKE16-NEXT: .LBB31_2: ; %cmp.true
; GFX11-FAKE16-NEXT: v_lshl_or_b32 v18, v37, 16, v51
@@ -16011,9 +16015,9 @@ define inreg <24 x float> @bitcast_v48i16_to_v24f32_scalar(<48 x i16> inreg %a,
; GFX11-FAKE16-NEXT: v_pk_add_u16 v12, s13, 3 op_sel_hi:[1,0]
; GFX11-FAKE16-NEXT: v_pk_add_u16 v13, s14, 3 op_sel_hi:[1,0]
; GFX11-FAKE16-NEXT: v_pk_add_u16 v14, s0, 3 op_sel_hi:[1,0]
-; GFX11-FAKE16-NEXT: v_pk_add_u16 v15, s15, 3 op_sel_hi:[1,0]
-; GFX11-FAKE16-NEXT: v_pk_add_u16 v16, s16, 3 op_sel_hi:[1,0]
-; GFX11-FAKE16-NEXT: v_pk_add_u16 v17, s17, 3 op_sel_hi:[1,0]
+; GFX11-FAKE16-NEXT: v_pk_add_u16 v15, s16, 3 op_sel_hi:[1,0]
+; GFX11-FAKE16-NEXT: v_pk_add_u16 v16, s17, 3 op_sel_hi:[1,0]
+; GFX11-FAKE16-NEXT: v_pk_add_u16 v17, s18, 3 op_sel_hi:[1,0]
; GFX11-FAKE16-NEXT: v_pk_add_u16 v18, v18, 3 op_sel_hi:[1,0]
; GFX11-FAKE16-NEXT: v_pk_add_u16 v19, v19, 3 op_sel_hi:[1,0]
; GFX11-FAKE16-NEXT: v_pk_add_u16 v20, v20, 3 op_sel_hi:[1,0]
@@ -20387,9 +20391,9 @@ define inreg <24 x float> @bitcast_v48f16_to_v24f32_scalar(<48 x half> inreg %a,
; GFX11-FAKE16-NEXT: v_and_b32_e32 v48, 0xffff, v3
; GFX11-FAKE16-NEXT: v_and_b32_e32 v39, 0xffff, v4
; GFX11-FAKE16-NEXT: v_and_b32_e32 v38, 0xffff, v5
-; GFX11-FAKE16-NEXT: s_lshr_b32 s41, s29, 16
-; GFX11-FAKE16-NEXT: s_lshr_b32 s42, s28, 16
-; GFX11-FAKE16-NEXT: s_lshr_b32 s15, s27, 16
+; GFX11-FAKE16-NEXT: s_lshr_b32 s40, s29, 16
+; GFX11-FAKE16-NEXT: s_lshr_b32 s41, s28, 16
+; GFX11-FAKE16-NEXT: s_lshr_b32 s42, s27, 16
; GFX11-FAKE16-NEXT: s_lshr_b32 s43, s26, 16
; GFX11-FAKE16-NEXT: s_lshr_b32 s14, s25, 16
; GFX11-FAKE16-NEXT: s_lshr_b32 s13, s24, 16
@@ -20405,7 +20409,7 @@ define inreg <24 x float> @bitcast_v48f16_to_v24f32_scalar(<48 x half> inreg %a,
; GFX11-FAKE16-NEXT: s_lshr_b32 s45, s2, 16
; GFX11-FAKE16-NEXT: s_lshr_b32 s46, s1, 16
; GFX11-FAKE16-NEXT: s_lshr_b32 s4, s0, 16
-; GFX11-FAKE16-NEXT: s_mov_b32 s40, 0
+; GFX11-FAKE16-NEXT: s_mov_b32 s15, 0
; GFX11-FAKE16-NEXT: s_pack_ll_b32_b16 s4, s0, s4
; GFX11-FAKE16-NEXT: s_pack_ll_b32_b16 s1, s1, s46
; GFX11-FAKE16-NEXT: s_pack_ll_b32_b16 s2, s2, s45
@@ -20421,9 +20425,9 @@ define inreg <24 x float> @bitcast_v48f16_to_v24f32_scalar(<48 x half> inreg %a,
; GFX11-FAKE16-NEXT: s_pack_ll_b32_b16 s13, s24, s13
; GFX11-FAKE16-NEXT: s_pack_ll_b32_b16 s14, s25, s14
; GFX11-FAKE16-NEXT: s_pack_ll_b32_b16 s0, s26, s43
-; GFX11-FAKE16-NEXT: s_pack_ll_b32_b16 s15, s27, s15
-; GFX11-FAKE16-NEXT: s_pack_ll_b32_b16 s16, s28, s42
-; GFX11-FAKE16-NEXT: s_pack_ll_b32_b16 s17, s29, s41
+; GFX11-FAKE16-NEXT: s_pack_ll_b32_b16 s16, s27, s42
+; GFX11-FAKE16-NEXT: s_pack_ll_b32_b16 s17, s28, s41
+; GFX11-FAKE16-NEXT: s_pack_ll_b32_b16 s18, s29, s40
; GFX11-FAKE16-NEXT: s_and_b32 s47, vcc_lo, exec_lo
; GFX11-FAKE16-NEXT: s_cbranch_scc0 .LBB35_4
; GFX11-FAKE16-NEXT: ; %bb.1: ; %cmp.false
@@ -20440,9 +20444,9 @@ define inreg <24 x float> @bitcast_v48f16_to_v24f32_scalar(<48 x half> inreg %a,
; GFX11-FAKE16-NEXT: v_dual_mov_b32 v8, s9 :: v_dual_mov_b32 v9, s10
; GFX11-FAKE16-NEXT: v_dual_mov_b32 v10, s11 :: v_dual_mov_b32 v11, s12
; GFX11-FAKE16-NEXT: v_dual_mov_b32 v12, s13 :: v_dual_mov_b32 v13, s14
-; GFX11-FAKE16-NEXT: v_dual_mov_b32 v14, s0 :: v_dual_mov_b32 v15, s15
-; GFX11-FAKE16-NEXT: v_dual_mov_b32 v16, s16 :: v_dual_mov_b32 v17, s17
-; GFX11-FAKE16-NEXT: s_and_not1_b32 vcc_lo, exec_lo, s40
+; GFX11-FAKE16-NEXT: v_dual_mov_b32 v14, s0 :: v_dual_mov_b32 v15, s16
+; GFX11-FAKE16-NEXT: v_dual_mov_b32 v16, s17 :: v_dual_mov_b32 v17, s18
+; GFX11-FAKE16-NEXT: s_and_not1_b32 vcc_lo, exec_lo, s15
; GFX11-FAKE16-NEXT: s_cbranch_vccnz .LBB35_3
; GFX11-FAKE16-NEXT: .LBB35_2: ; %cmp.true
; GFX11-FAKE16-NEXT: v_lshl_or_b32 v18, v37, 16, v51
@@ -20466,9 +20470,9 @@ define inreg <24 x float> @bitcast_v48f16_to_v24f32_scalar(<48 x half> inreg %a,
; GFX11-FAKE16-NEXT: v_pk_add_f16 v12, 0x200, s13 op_sel_hi:[0,1]
; GFX11-FAKE16-NEXT: v_pk_add_f16 v13, 0x200, s14 op_sel_hi:[0,1]
; GFX11-FAKE16-NEXT: v_pk_add_f16 v14, 0x200, s0 op_sel_hi:[0,1]
-; GFX11-FAKE16-NEXT: v_pk_add_f16 v15, 0x200, s15 op_sel_hi:[0,1]
-; GFX11-FAKE16-NEXT: v_pk_add_f16 v16, 0x200, s16 op_sel_hi:[0,1]
-; GFX11-FAKE16-NEXT: v_pk_add_f16 v17, 0x200, s17 op_sel_hi:[0,1]
+; GFX11-FAKE16-NEXT: v_pk_add_f16 v15, 0x200, s16 op_sel_hi:[0,1]
+; GFX11-FAKE16-NEXT: v_pk_add_f16 v16, 0x200, s17 op_sel_hi:[0,1]
+; GFX11-FAKE16-NEXT: v_pk_add_f16 v17, 0x200, s18 op_sel_hi:[0,1]
; GFX11-FAKE16-NEXT: v_pk_add_f16 v18, 0x200, v18 op_sel_hi:[0,1]
; GFX11-FAKE16-NEXT: v_pk_add_f16 v19, 0x200, v19 op_sel_hi:[0,1]
; GFX11-FAKE16-NEXT: v_pk_add_f16 v20, 0x200, v20 op_sel_hi:[0,1]
@@ -22703,7 +22707,7 @@ define inreg <48 x i16> @bitcast_v12i64_to_v48i16_scalar(<12 x i64> inreg %a, i3
; GFX11-TRUE16-NEXT: v_readfirstlane_b32 s5, v3
; GFX11-TRUE16-NEXT: v_readfirstlane_b32 s6, v4
; GFX11-TRUE16-NEXT: v_readfirstlane_b32 s4, v5
-; GFX11-TRUE16-NEXT: s_mov_b32 s74, 0
+; GFX11-TRUE16-NEXT: s_mov_b32 s62, 0
; GFX11-TRUE16-NEXT: s_and_b32 s10, vcc_lo, exec_lo
; GFX11-TRUE16-NEXT: s_cbranch_scc0 .LBB41_4
; GFX11-TRUE16-NEXT: ; %bb.1: ; %cmp.false
@@ -22727,11 +22731,11 @@ define inreg <48 x i16> @bitcast_v12i64_to_v48i16_scalar(<12 x i64> inreg %a, i3
; GFX11-TRUE16-NEXT: s_lshr_b32 s59, s18, 16
; GFX11-TRUE16-NEXT: s_lshr_b32 s60, s17, 16
; GFX11-TRUE16-NEXT: s_lshr_b32 s61, s16, 16
-; GFX11-TRUE16-NEXT: s_lshr_b32 s62, s3, 16
-; GFX11-TRUE16-NEXT: s_lshr_b32 s63, s2, 16
-; GFX11-TRUE16-NEXT: s_lshr_b32 s72, s1, 16
-; GFX11-TRUE16-NEXT: s_lshr_b32 s73, s0, 16
-; GFX11-TRUE16-NEXT: s_and_not1_b32 vcc_lo, exec_lo, s74
+; GFX11-TRUE16-NEXT: s_lshr_b32 s63, s3, 16
+; GFX11-TRUE16-NEXT: s_lshr_b32 s72, s2, 16
+; GFX11-TRUE16-NEXT: s_lshr_b32 s73, s1, 16
+; GFX11-TRUE16-NEXT: s_lshr_b32 s74, s0, 16
+; GFX11-TRUE16-NEXT: s_and_not1_b32 vcc_lo, exec_lo, s62
; GFX11-TRUE16-NEXT: s_cbranch_vccnz .LBB41_3
; GFX11-TRUE16-NEXT: .LBB41_2: ; %cmp.true
; GFX11-TRUE16-NEXT: s_add_u32 s6, s6, 3
@@ -22778,16 +22782,18 @@ define inreg <48 x i16> @bitcast_v12i64_to_v48i16_scalar(<12 x i64> inreg %a, i3
; GFX11-TRUE16-NEXT: s_lshr_b32 s59, s18, 16
; GFX11-TRUE16-NEXT: s_lshr_b32 s60, s17, 16
; GFX11-TRUE16-NEXT: s_lshr_b32 s61, s16, 16
-; GFX11-TRUE16-NEXT: s_lshr_b32 s62, s3, 16
-; GFX11-TRUE16-NEXT: s_lshr_b32 s63, s2, 16
-; GFX11-TRUE16-NEXT: s_lshr_b32 s72, s1, 16
-; GFX11-TRUE16-NEXT: s_lshr_b32 s73, s0, 16
+; GFX11-TRUE16-NEXT: s_lshr_b32 s63, s3, 16
+; GFX11-TRUE16-NEXT: s_lshr_b32 s72, s2, 16
+; GFX11-TRUE16-NEXT: s_lshr_b32 s73, s1, 16
+; GFX11-TRUE16-NEXT: s_lshr_b32 s74, s0, 16
; GFX11-TRUE16-NEXT: .LBB41_3: ; %end
; GFX11-TRUE16-NEXT: s_delay_alu instid0(SALU_CYCLE_1)
-; GFX11-TRUE16-NEXT: s_pack_ll_b32_b16 s0, s0, s73
-; GFX11-TRUE16-NEXT: s_pack_ll_b32_b16 s1, s1, s72
-; GFX11-TRUE16-NEXT: s_pack_ll_b32_b16 s2, s2, s63
-; GFX11-TRUE16-NEXT: s_pack_ll_b32_b16 s3, s3, s62
+; GFX11-TRUE16-NEXT: s_mov_b32 s62, s74
+; GFX11-TRUE16-NEXT: s_pack_ll_b32_b16 s1, s1, s73
+; GFX11-TRUE16-NEXT: s_pack_ll_b32_b16 s0, s0, s62
+; GFX11-TRUE16-NEXT: s_mov_b32 s62, s72
+; GFX11-TRUE16-NEXT: s_pack_ll_b32_b16 s3, s3, s63
+; GFX11-TRUE16-NEXT: s_pack_ll_b32_b16 s2, s2, s62
; GFX11-TRUE16-NEXT: s_pack_ll_b32_b16 s16, s16, s61
; GFX11-TRUE16-NEXT: s_pack_ll_b32_b16 s17, s17, s60
; GFX11-TRUE16-NEXT: s_pack_ll_b32_b16 s18, s18, s59
@@ -22822,10 +22828,10 @@ define inreg <48 x i16> @bitcast_v12i64_to_v48i16_scalar(<12 x i64> inreg %a, i3
; GFX11-TRUE16-NEXT: v_dual_mov_b32 v22, s6 :: v_dual_mov_b32 v23, s4
; GFX11-TRUE16-NEXT: s_setpc_b64 s[30:31]
; GFX11-TRUE16-NEXT: .LBB41_4:
+; GFX11-TRUE16-NEXT: ; implicit-def: $sgpr74_lo16
; GFX11-TRUE16-NEXT: ; implicit-def: $sgpr73_lo16
; GFX11-TRUE16-NEXT: ; implicit-def: $sgpr72_lo16
; GFX11-TRUE16-NEXT: ; implicit-def: $sgpr63_lo16
-; GFX11-TRUE16-NEXT: ; implicit-def: $sgpr62_lo16
; GFX11-TRUE16-NEXT: ; implicit-def: $sgpr61_lo16
; GFX11-TRUE16-NEXT: ; implicit-def: $sgpr60_lo16
; GFX11-TRUE16-NEXT: ; implicit-def: $sgpr59_lo16
@@ -24979,9 +24985,9 @@ define inreg <12 x i64> @bitcast_v48i16_to_v12i64_scalar(<48 x i16> inreg %a, i3
; GFX11-FAKE16-NEXT: v_and_b32_e32 v48, 0xffff, v3
; GFX11-FAKE16-NEXT: v_and_b32_e32 v39, 0xffff, v4
; GFX11-FAKE16-NEXT: v_and_b32_e32 v38, 0xffff, v5
-; GFX11-FAKE16-NEXT: s_lshr_b32 s41, s29, 16
-; GFX11-FAKE16-NEXT: s_lshr_b32 s42, s28, 16
-; GFX11-FAKE16-NEXT: s_lshr_b32 s15, s27, 16
+; GFX11-FAKE16-NEXT: s_lshr_b32 s40, s29, 16
+; GFX11-FAKE16-NEXT: s_lshr_b32 s41, s28, 16
+; GFX11-FAKE16-NEXT: s_lshr_b32 s42, s27, 16
; GFX11-FAKE16-NEXT: s_lshr_b32 s43, s26, 16
; GFX11-FAKE16-NEXT: s_lshr_b32 s14, s25, 16
; GFX11-FAKE16-NEXT: s_lshr_b32 s13, s24, 16
@@ -24997,7 +25003,7 @@ define inreg <12 x i64> @bitcast_v48i16_to_v12i64_scalar(<48 x i16> inreg %a, i3
; GFX11-FAKE16-NEXT: s_lshr_b32 s45, s2, 16
; GFX11-FAKE16-NEXT: s_lshr_b32 s46, s1, 16
; GFX11-FAKE16-NEXT: s_lshr_b32 s4, s0, 16
-; GFX11-FAKE16-NEXT: s_mov_b32 s40, 0
+; GFX11-FAKE16-NEXT: s_mov_b32 s15, 0
; GFX11-FAKE16-NEXT: s_pack_ll_b32_b16 s4, s0, s4
; GFX11-FAKE16-NEXT: s_pack_ll_b32_b16 s1, s1, s46
; GFX11-FAKE16-NEXT: s_pack_ll_b32_b16 s2, s2, s45
@@ -25013,9 +25019,9 @@ define inreg <12 x i64> @bitcast_v48i16_to_v12i64_scalar(<48 x i16> inreg %a, i3
; GFX11-FAKE16-NEXT: s_pack_ll_b32_b16 s13, s24, s13
; GFX11-FAKE16-NEXT: s_pack_ll_b32_b16 s14, s25, s14
; GFX11-FAKE16-NEXT: s_pack_ll_b32_b16 s0, s26, s43
-; GFX11-FAKE16-NEXT: s_pack_ll_b32_b16 s15, s27, s15
-; GFX11-FAKE16-NEXT: s_pack_ll_b32_b16 s16, s28, s42
-; GFX11-FAKE16-NEXT: s_pack_ll_b32_b16 s17, s29, s41
+; GFX11-FAKE16-NEXT: s_pack_ll_b32_b16 s16, s27, s42
+; GFX11-FAKE16-NEXT: s_pack_ll_b32_b16 s17, s28, s41
+; GFX11-FAKE16-NEXT: s_pack_ll_b32_b16 s18, s29, s40
; GFX11-FAKE16-NEXT: s_and_b32 s47, vcc_lo, exec_lo
; GFX11-FAKE16-NEXT: s_cbranch_scc0 .LBB43_4
; GFX11-FAKE16-NEXT: ; %bb.1: ; %cmp.false
@@ -25032,9 +25038,9 @@ define inreg <12 x i64> @bitcast_v48i16_to_v12i64_scalar(<48 x i16> inreg %a, i3
; GFX11-FAKE16-NEXT: v_dual_mov_b32 v8, s9 :: v_dual_mov_b32 v9, s10
; GFX11-FAKE16-NEXT: v_dual_mov_b32 v10, s11 :: v_dual_mov_b32 v11, s12
; GFX11-FAKE16-NEXT: v_dual_mov_b32 v12, s13 :: v_dual_mov_b32 v13, s14
-; GFX11-FAKE16-NEXT: v_dual_mov_b32 v14, s0 :: v_dual_mov_b32 v15, s15
-; GFX11-FAKE16-NEXT: v_dual_mov_b32 v16, s16 :: v_dual_mov_b32 v17, s17
-; GFX11-FAKE16-NEXT: s_and_not1_b32 vcc_lo, exec_lo, s40
+; GFX11-FAKE16-NEXT: v_dual_mov_b32 v14, s0 :: v_dual_mov_b32 v15, s16
+; GFX11-FAKE16-NEXT: v_dual_mov_b32 v16, s17 :: v_dual_mov_b32 v17, s18
+; GFX11-FAKE16-NEXT: s_and_not1_b32 vcc_lo, exec_lo, s15
; GFX11-FAKE16-NEXT: s_cbranch_vccnz .LBB43_3
; GFX11-FAKE16-NEXT: .LBB43_2: ; %cmp.true
; GFX11-FAKE16-NEXT: v_lshl_or_b32 v18, v37, 16, v51
@@ -25058,9 +25064,9 @@ define inreg <12 x i64> @bitcast_v48i16_to_v12i64_scalar(<48 x i16> inreg %a, i3
; GFX11-FAKE16-NEXT: v_pk_add_u16 v12, s13, 3 op_sel_hi:[1,0]
; GFX11-FAKE16-NEXT: v_pk_add_u16 v13, s14, 3 op_sel_hi:[1,0]
; GFX11-FAKE16-NEXT: v_pk_add_u16 v14, s0, 3 op_sel_hi:[1,0]
-; GFX11-FAKE16-NEXT: v_pk_add_u16 v15, s15, 3 op_sel_hi:[1,0]
-; GFX11-FAKE16-NEXT: v_pk_add_u16 v16, s16, 3 op_sel_hi:[1,0]
-; GFX11-FAKE16-NEXT: v_pk_add_u16 v17, s17, 3 op_sel_hi:[1,0]
+; GFX11-FAKE16-NEXT: v_pk_add_u16 v15, s16, 3 op_sel_hi:[1,0]
+; GFX11-FAKE16-NEXT: v_pk_add_u16 v16, s17, 3 op_sel_hi:[1,0]
+; GFX11-FAKE16-NEXT: v_pk_add_u16 v17, s18, 3 op_sel_hi:[1,0]
; GFX11-FAKE16-NEXT: v_pk_add_u16 v18, v18, 3 op_sel_hi:[1,0]
; GFX11-FAKE16-NEXT: v_pk_add_u16 v19, v19, 3 op_sel_hi:[1,0]
; GFX11-FAKE16-NEXT: v_pk_add_u16 v20, v20, 3 op_sel_hi:[1,0]
@@ -26902,7 +26908,7 @@ define inreg <48 x half> @bitcast_v12i64_to_v48f16_scalar(<12 x i64> inreg %a, i
; GFX11-TRUE16-NEXT: v_readfirstlane_b32 s5, v3
; GFX11-TRUE16-NEXT: v_readfirstlane_b32 s6, v4
; GFX11-TRUE16-NEXT: v_readfirstlane_b32 s4, v5
-; GFX11-TRUE16-NEXT: s_mov_b32 s74, 0
+; GFX11-TRUE16-NEXT: s_mov_b32 s62, 0
; GFX11-TRUE16-NEXT: s_and_b32 s10, vcc_lo, exec_lo
; GFX11-TRUE16-NEXT: s_cbranch_scc0 .LBB45_4
; GFX11-TRUE16-NEXT: ; %bb.1: ; %cmp.false
@@ -26926,11 +26932,11 @@ define inreg <48 x half> @bitcast_v12i64_to_v48f16_scalar(<12 x i64> inreg %a, i
; GFX11-TRUE16-NEXT: s_lshr_b32 s59, s18, 16
; GFX11-TRUE16-NEXT: s_lshr_b32 s60, s17, 16
; GFX11-TRUE16-NEXT: s_lshr_b32 s61, s16, 16
-; GFX11-TRUE16-NEXT: s_lshr_b32 s62, s3, 16
-; GFX11-TRUE16-NEXT: s_lshr_b32 s63, s2, 16
-; GFX11-TRUE16-NEXT: s_lshr_b32 s72, s1, 16
-; GFX11-TRUE16-NEXT: s_lshr_b32 s73, s0, 16
-; GFX11-TRUE16-NEXT: s_and_not1_b32 vcc_lo, exec_lo, s74
+; GFX11-TRUE16-NEXT: s_lshr_b32 s63, s3, 16
+; GFX11-TRUE16-NEXT: s_lshr_b32 s72, s2, 16
+; GFX11-TRUE16-NEXT: s_lshr_b32 s73, s1, 16
+; GFX11-TRUE16-NEXT: s_lshr_b32 s74, s0, 16
+; GFX11-TRUE16-NEXT: s_and_not1_b32 vcc_lo, exec_lo, s62
; GFX11-TRUE16-NEXT: s_cbranch_vccnz .LBB45_3
; GFX11-TRUE16-NEXT: .LBB45_2: ; %cmp.true
; GFX11-TRUE16-NEXT: s_add_u32 s6, s6, 3
@@ -26977,16 +26983,18 @@ define inreg <48 x half> @bitcast_v12i64_to_v48f16_scalar(<12 x i64> inreg %a, i
; GFX11-TRUE16-NEXT: s_lshr_b32 s59, s18, 16
; GFX11-TRUE16-NEXT: s_lshr_b32 s60, s17, 16
; GFX11-TRUE16-NEXT: s_lshr_b32 s61, s16, 16
-; GFX11-TRUE16-NEXT: s_lshr_b32 s62, s3, 16
-; GFX11-TRUE16-NEXT: s_lshr_b32 s63, s2, 16
-; GFX11-TRUE16-NEXT: s_lshr_b32 s72, s1, 16
-; GFX11-TRUE16-NEXT: s_lshr_b32 s73, s0, 16
+; GFX11-TRUE16-NEXT: s_lshr_b32 s63, s3, 16
+; GFX11-TRUE16-NEXT: s_lshr_b32 s72, s2, 16
+; GFX11-TRUE16-NEXT: s_lshr_b32 s73, s1, 16
+; GFX11-TRUE16-NEXT: s_lshr_b32 s74, s0, 16
; GFX11-TRUE16-NEXT: .LBB45_3: ; %end
; GFX11-TRUE16-NEXT: s_delay_alu instid0(SALU_CYCLE_1)
-; GFX11-TRUE16-NEXT: s_pack_ll_b32_b16 s0, s0, s73
-; GFX11-TRUE16-NEXT: s_pack_ll_b32_b16 s1, s1, s72
-; GFX11-TRUE16-NEXT: s_pack_ll_b32_b16 s2, s2, s63
-; GFX11-TRUE16-NEXT: s_pack_ll_b32_b16 s3, s3, s62
+; GFX11-TRUE16-NEXT: s_mov_b32 s62, s74
+; GFX11-TRUE16-NEXT: s_pack_ll_b32_b16 s1, s1, s73
+; GFX11-TRUE16-NEXT: s_pack_ll_b32_b16 s0, s0, s62
+; GFX11-TRUE16-NEXT: s_mov_b32 s62, s72
+; GFX11-TRUE16-NEXT: s_pack_ll_b32_b16 s3, s3, s63
+; GFX11-TRUE16-NEXT: s_pack_ll_b32_b16 s2, s2, s62
; GFX11-TRUE16-NEXT: s_pack_ll_b32_b16 s16, s16, s61
; GFX11-TRUE16-NEXT: s_pack_ll_b32_b16 s17, s17, s60
; GFX11-TRUE16-NEXT: s_pack_ll_b32_b16 s18, s18, s59
@@ -27021,10 +27029,10 @@ define inreg <48 x half> @bitcast_v12i64_to_v48f16_scalar(<12 x i64> inreg %a, i
; GFX11-TRUE16-NEXT: v_dual_mov_b32 v22, s6 :: v_dual_mov_b32 v23, s4
; GFX11-TRUE16-NEXT: s_setpc_b64 s[30:31]
; GFX11-TRUE16-NEXT: .LBB45_4:
+; GFX11-TRUE16-NEXT: ; implicit-def: $sgpr74_lo16
; GFX11-TRUE16-NEXT: ; implicit-def: $sgpr73_lo16
; GFX11-TRUE16-NEXT: ; implicit-def: $sgpr72_lo16
; GFX11-TRUE16-NEXT: ; implicit-def: $sgpr63_lo16
-; GFX11-TRUE16-NEXT: ; implicit-def: $sgpr62_lo16
; GFX11-TRUE16-NEXT: ; implicit-def: $sgpr61_lo16
; GFX11-TRUE16-NEXT: ; implicit-def: $sgpr60_lo16
; GFX11-TRUE16-NEXT: ; implicit-def: $sgpr59_lo16
@@ -29450,9 +29458,9 @@ define inreg <12 x i64> @bitcast_v48f16_to_v12i64_scalar(<48 x half> inreg %a, i
; GFX11-FAKE16-NEXT: v_and_b32_e32 v48, 0xffff, v3
; GFX11-FAKE16-NEXT: v_and_b32_e32 v39, 0xffff, v4
; GFX11-FAKE16-NEXT: v_and_b32_e32 v38, 0xffff, v5
-; GFX11-FAKE16-NEXT: s_lshr_b32 s41, s29, 16
-; GFX11-FAKE16-NEXT: s_lshr_b32 s42, s28, 16
-; GFX11-FAKE16-NEXT: s_lshr_b32 s15, s27, 16
+; GFX11-FAKE16-NEXT: s_lshr_b32 s40, s29, 16
+; GFX11-FAKE16-NEXT: s_lshr_b32 s41, s28, 16
+; GFX11-FAKE16-NEXT: s_lshr_b32 s42, s27, 16
; GFX11-FAKE16-NEXT: s_lshr_b32 s43, s26, 16
; GFX11-FAKE16-NEXT: s_lshr_b32 s14, s25, 16
; GFX11-FAKE16-NEXT: s_lshr_b32 s13, s24, 16
@@ -29468,7 +29476,7 @@ define inreg <12 x i64> @bitcast_v48f16_to_v12i64_scalar(<48 x half> inreg %a, i
; GFX11-FAKE16-NEXT: s_lshr_b32 s45, s2, 16
; GFX11-FAKE16-NEXT: s_lshr_b32 s46, s1, 16
; GFX11-FAKE16-NEXT: s_lshr_b32 s4, s0, 16
-; GFX11-FAKE16-NEXT: s_mov_b32 s40, 0
+; GFX11-FAKE16-NEXT: s_mov_b32 s15, 0
; GFX11-FAKE16-NEXT: s_pack_ll_b32_b16 s4, s0, s4
; GFX11-FAKE16-NEXT: s_pack_ll_b32_b16 s1, s1, s46
; GFX11-FAKE16-NEXT: s_pack_ll_b32_b16 s2, s2, s45
@@ -29484,9 +29492,9 @@ define inreg <12 x i64> @bitcast_v48f16_to_v12i64_scalar(<48 x half> inreg %a, i
; GFX11-FAKE16-NEXT: s_pack_ll_b32_b16 s13, s24, s13
; GFX11-FAKE16-NEXT: s_pack_ll_b32_b16 s14, s25, s14
; GFX11-FAKE16-NEXT: s_pack_ll_b32_b16 s0, s26, s43
-; GFX11-FAKE16-NEXT: s_pack_ll_b32_b16 s15, s27, s15
-; GFX11-FAKE16-NEXT: s_pack_ll_b32_b16 s16, s28, s42
-; GFX11-FAKE16-NEXT: s_pack_ll_b32_b16 s17, s29, s41
+; GFX11-FAKE16-NEXT: s_pack_ll_b32_b16 s16, s27, s42
+; GFX11-FAKE16-NEXT: s_pack_ll_b32_b16 s17, s28, s41
+; GFX11-FAKE16-NEXT: s_pack_ll_b32_b16 s18, s29, s40
; GFX11-FAKE16-NEXT: s_and_b32 s47, vcc_lo, exec_lo
; GFX11-FAKE16-NEXT: s_cbranch_scc0 .LBB47_4
; GFX11-FAKE16-NEXT: ; %bb.1: ; %cmp.false
@@ -29503,9 +29511,9 @@ define inreg <12 x i64> @bitcast_v48f16_to_v12i64_scalar(<48 x half> inreg %a, i
; GFX11-FAKE16-NEXT: v_dual_mov_b32 v8, s9 :: v_dual_mov_b32 v9, s10
; GFX11-FAKE16-NEXT: v_dual_mov_b32 v10, s11 :: v_dual_mov_b32 v11, s12
; GFX11-FAKE16-NEXT: v_dual_mov_b32 v12, s13 :: v_dual_mov_b32 v13, s14
-; GFX11-FAKE16-NEXT: v_dual_mov_b32 v14, s0 :: v_dual_mov_b32 v15, s15
-; GFX11-FAKE16-NEXT: v_dual_mov_b32 v16, s16 :: v_dual_mov_b32 v17, s17
-; GFX11-FAKE16-NEXT: s_and_not1_b32 vcc_lo, exec_lo, s40
+; GFX11-FAKE16-NEXT: v_dual_mov_b32 v14, s0 :: v_dual_mov_b32 v15, s16
+; GFX11-FAKE16-NEXT: v_dual_mov_b32 v16, s17 :: v_dual_mov_b32 v17, s18
+; GFX11-FAKE16-NEXT: s_and_not1_b32 vcc_lo, exec_lo, s15
; GFX11-FAKE16-NEXT: s_cbranch_vccnz .LBB47_3
; GFX11-FAKE16-NEXT: .LBB47_2: ; %cmp.true
; GFX11-FAKE16-NEXT: v_lshl_or_b32 v18, v37, 16, v51
@@ -29529,9 +29537,9 @@ define inreg <12 x i64> @bitcast_v48f16_to_v12i64_scalar(<48 x half> inreg %a, i
; GFX11-FAKE16-NEXT: v_pk_add_f16 v12, 0x200, s13 op_sel_hi:[0,1]
; GFX11-FAKE16-NEXT: v_pk_add_f16 v13, 0x200, s14 op_sel_hi:[0,1]
; GFX11-FAKE16-NEXT: v_pk_add_f16 v14, 0x200, s0 op_sel_hi:[0,1]
-; GFX11-FAKE16-NEXT: v_pk_add_f16 v15, 0x200, s15 op_sel_hi:[0,1]
-; GFX11-FAKE16-NEXT: v_pk_add_f16 v16, 0x200, s16 op_sel_hi:[0,1]
-; GFX11-FAKE16-NEXT: v_pk_add_f16 v17, 0x200, s17 op_sel_hi:[0,1]
+; GFX11-FAKE16-NEXT: v_pk_add_f16 v15, 0x200, s16 op_sel_hi:[0,1]
+; GFX11-FAKE16-NEXT: v_pk_add_f16 v16, 0x200, s17 op_sel_hi:[0,1]
+; GFX11-FAKE16-NEXT: v_pk_add_f16 v17, 0x200, s18 op_sel_hi:[0,1]
; GFX11-FAKE16-NEXT: v_pk_add_f16 v18, 0x200, v18 op_sel_hi:[0,1]
; GFX11-FAKE16-NEXT: v_pk_add_f16 v19, 0x200, v19 op_sel_hi:[0,1]
; GFX11-FAKE16-NEXT: v_pk_add_f16 v20, 0x200, v20 op_sel_hi:[0,1]
@@ -33147,9 +33155,9 @@ define inreg <12 x double> @bitcast_v48i16_to_v12f64_scalar(<48 x i16> inreg %a,
; GFX11-FAKE16-NEXT: v_and_b32_e32 v48, 0xffff, v3
; GFX11-FAKE16-NEXT: v_and_b32_e32 v39, 0xffff, v4
; GFX11-FAKE16-NEXT: v_and_b32_e32 v38, 0xffff, v5
-; GFX11-FAKE16-NEXT: s_lshr_b32 s41, s29, 16
-; GFX11-FAKE16-NEXT: s_lshr_b32 s42, s28, 16
-; GFX11-FAKE16-NEXT: s_lshr_b32 s15, s27, 16
+; GFX11-FAKE16-NEXT: s_lshr_b32 s40, s29, 16
+; GFX11-FAKE16-NEXT: s_lshr_b32 s41, s28, 16
+; GFX11-FAKE16-NEXT: s_lshr_b32 s42, s27, 16
; GFX11-FAKE16-NEXT: s_lshr_b32 s43, s26, 16
; GFX11-FAKE16-NEXT: s_lshr_b32 s14, s25, 16
; GFX11-FAKE16-NEXT: s_lshr_b32 s13, s24, 16
@@ -33165,7 +33173,7 @@ define inreg <12 x double> @bitcast_v48i16_to_v12f64_scalar(<48 x i16> inreg %a,
; GFX11-FAKE16-NEXT: s_lshr_b32 s45, s2, 16
; GFX11-FAKE16-NEXT: s_lshr_b32 s46, s1, 16
; GFX11-FAKE16-NEXT: s_lshr_b32 s4, s0, 16
-; GFX11-FAKE16-NEXT: s_mov_b32 s40, 0
+; GFX11-FAKE16-NEXT: s_mov_b32 s15, 0
; GFX11-FAKE16-NEXT: s_pack_ll_b32_b16 s4, s0, s4
; GFX11-FAKE16-NEXT: s_pack_ll_b32_b16 s1, s1, s46
; GFX11-FAKE16-NEXT: s_pack_ll_b32_b16 s2, s2, s45
@@ -33181,9 +33189,9 @@ define inreg <12 x double> @bitcast_v48i16_to_v12f64_scalar(<48 x i16> inreg %a,
; GFX11-FAKE16-NEXT: s_pack_ll_b32_b16 s13, s24, s13
; GFX11-FAKE16-NEXT: s_pack_ll_b32_b16 s14, s25, s14
; GFX11-FAKE16-NEXT: s_pack_ll_b32_b16 s0, s26, s43
-; GFX11-FAKE16-NEXT: s_pack_ll_b32_b16 s15, s27, s15
-; GFX11-FAKE16-NEXT: s_pack_ll_b32_b16 s16, s28, s42
-; GFX11-FAKE16-NEXT: s_pack_ll_b32_b16 s17, s29, s41
+; GFX11-FAKE16-NEXT: s_pack_ll_b32_b16 s16, s27, s42
+; GFX11-FAKE16-NEXT: s_pack_ll_b32_b16 s17, s28, s41
+; GFX11-FAKE16-NEXT: s_pack_ll_b32_b16 s18, s29, s40
; GFX11-FAKE16-NEXT: s_and_b32 s47, vcc_lo, exec_lo
; GFX11-FAKE16-NEXT: s_cbranch_scc0 .LBB51_4
; GFX11-FAKE16-NEXT: ; %bb.1: ; %cmp.false
@@ -33200,9 +33208,9 @@ define inreg <12 x double> @bitcast_v48i16_to_v12f64_scalar(<48 x i16> inreg %a,
; GFX11-FAKE16-NEXT: v_dual_mov_b32 v8, s9 :: v_dual_mov_b32 v9, s10
; GFX11-FAKE16-NEXT: v_dual_mov_b32 v10, s11 :: v_dual_mov_b32 v11, s12
; GFX11-FAKE16-NEXT: v_dual_mov_b32 v12, s13 :: v_dual_mov_b32 v13, s14
-; GFX11-FAKE16-NEXT: v_dual_mov_b32 v14, s0 :: v_dual_mov_b32 v15, s15
-; GFX11-FAKE16-NEXT: v_dual_mov_b32 v16, s16 :: v_dual_mov_b32 v17, s17
-; GFX11-FAKE16-NEXT: s_and_not1_b32 vcc_lo, exec_lo, s40
+; GFX11-FAKE16-NEXT: v_dual_mov_b32 v14, s0 :: v_dual_mov_b32 v15, s16
+; GFX11-FAKE16-NEXT: v_dual_mov_b32 v16, s17 :: v_dual_mov_b32 v17, s18
+; GFX11-FAKE16-NEXT: s_and_not1_b32 vcc_lo, exec_lo, s15
; GFX11-FAKE16-NEXT: s_cbranch_vccnz .LBB51_3
; GFX11-FAKE16-NEXT: .LBB51_2: ; %cmp.true
; GFX11-FAKE16-NEXT: v_lshl_or_b32 v18, v37, 16, v51
@@ -33226,9 +33234,9 @@ define inreg <12 x double> @bitcast_v48i16_to_v12f64_scalar(<48 x i16> inreg %a,
; GFX11-FAKE16-NEXT: v_pk_add_u16 v12, s13, 3 op_sel_hi:[1,0]
; GFX11-FAKE16-NEXT: v_pk_add_u16 v13, s14, 3 op_sel_hi:[1,0]
; GFX11-FAKE16-NEXT: v_pk_add_u16 v14, s0, 3 op_sel_hi:[1,0]
-; GFX11-FAKE16-NEXT: v_pk_add_u16 v15, s15, 3 op_sel_hi:[1,0]
-; GFX11-FAKE16-NEXT: v_pk_add_u16 v16, s16, 3 op_sel_hi:[1,0]
-; GFX11-FAKE16-NEXT: v_pk_add_u16 v17, s17, 3 op_sel_hi:[1,0]
+; GFX11-FAKE16-NEXT: v_pk_add_u16 v15, s16, 3 op_sel_hi:[1,0]
+; GFX11-FAKE16-NEXT: v_pk_add_u16 v16, s17, 3 op_sel_hi:[1,0]
+; GFX11-FAKE16-NEXT: v_pk_add_u16 v17, s18, 3 op_sel_hi:[1,0]
; GFX11-FAKE16-NEXT: v_pk_add_u16 v18, v18, 3 op_sel_hi:[1,0]
; GFX11-FAKE16-NEXT: v_pk_add_u16 v19, v19, 3 op_sel_hi:[1,0]
; GFX11-FAKE16-NEXT: v_pk_add_u16 v20, v20, 3 op_sel_hi:[1,0]
@@ -37520,9 +37528,9 @@ define inreg <12 x double> @bitcast_v48f16_to_v12f64_scalar(<48 x half> inreg %a
; GFX11-FAKE16-NEXT: v_and_b32_e32 v48, 0xffff, v3
; GFX11-FAKE16-NEXT: v_and_b32_e32 v39, 0xffff, v4
; GFX11-FAKE16-NEXT: v_and_b32_e32 v38, 0xffff, v5
-; GFX11-FAKE16-NEXT: s_lshr_b32 s41, s29, 16
-; GFX11-FAKE16-NEXT: s_lshr_b32 s42, s28, 16
-; GFX11-FAKE16-NEXT: s_lshr_b32 s15, s27, 16
+; GFX11-FAKE16-NEXT: s_lshr_b32 s40, s29, 16
+; GFX11-FAKE16-NEXT: s_lshr_b32 s41, s28, 16
+; GFX11-FAKE16-NEXT: s_lshr_b32 s42, s27, 16
; GFX11-FAKE16-NEXT: s_lshr_b32 s43, s26, 16
; GFX11-FAKE16-NEXT: s_lshr_b32 s14, s25, 16
; GFX11-FAKE16-NEXT: s_lshr_b32 s13, s24, 16
@@ -37538,7 +37546,7 @@ define inreg <12 x double> @bitcast_v48f16_to_v12f64_scalar(<48 x half> inreg %a
; GFX11-FAKE16-NEXT: s_lshr_b32 s45, s2, 16
; GFX11-FAKE16-NEXT: s_lshr_b32 s46, s1, 16
; GFX11-FAKE16-NEXT: s_lshr_b32 s4, s0, 16
-; GFX11-FAKE16-NEXT: s_mov_b32 s40, 0
+; GFX11-FAKE16-NEXT: s_mov_b32 s15, 0
; GFX11-FAKE16-NEXT: s_pack_ll_b32_b16 s4, s0, s4
; GFX11-FAKE16-NEXT: s_pack_ll_b32_b16 s1, s1, s46
; GFX11-FAKE16-NEXT: s_pack_ll_b32_b16 s2, s2, s45
@@ -37554,9 +37562,9 @@ define inreg <12 x double> @bitcast_v48f16_to_v12f64_scalar(<48 x half> inreg %a
; GFX11-FAKE16-NEXT: s_pack_ll_b32_b16 s13, s24, s13
; GFX11-FAKE16-NEXT: s_pack_ll_b32_b16 s14, s25, s14
; GFX11-FAKE16-NEXT: s_pack_ll_b32_b16 s0, s26, s43
-; GFX11-FAKE16-NEXT: s_pack_ll_b32_b16 s15, s27, s15
-; GFX11-FAKE16-NEXT: s_pack_ll_b32_b16 s16, s28, s42
-; GFX11-FAKE16-NEXT: s_pack_ll_b32_b16 s17, s29, s41
+; GFX11-FAKE16-NEXT: s_pack_ll_b32_b16 s16, s27, s42
+; GFX11-FAKE16-NEXT: s_pack_ll_b32_b16 s17, s28, s41
+; GFX11-FAKE16-NEXT: s_pack_ll_b32_b16 s18, s29, s40
; GFX11-FAKE16-NEXT: s_and_b32 s47, vcc_lo, exec_lo
; GFX11-FAKE16-NEXT: s_cbranch_scc0 .LBB55_4
; GFX11-FAKE16-NEXT: ; %bb.1: ; %cmp.false
@@ -37573,9 +37581,9 @@ define inreg <12 x double> @bitcast_v48f16_to_v12f64_scalar(<48 x half> inreg %a
; GFX11-FAKE16-NEXT: v_dual_mov_b32 v8, s9 :: v_dual_mov_b32 v9, s10
; GFX11-FAKE16-NEXT: v_dual_mov_b32 v10, s11 :: v_dual_mov_b32 v11, s12
; GFX11-FAKE16-NEXT: v_dual_mov_b32 v12, s13 :: v_dual_mov_b32 v13, s14
-; GFX11-FAKE16-NEXT: v_dual_mov_b32 v14, s0 :: v_dual_mov_b32 v15, s15
-; GFX11-FAKE16-NEXT: v_dual_mov_b32 v16, s16 :: v_dual_mov_b32 v17, s17
-; GFX11-FAKE16-NEXT: s_and_not1_b32 vcc_lo, exec_lo, s40
+; GFX11-FAKE16-NEXT: v_dual_mov_b32 v14, s0 :: v_dual_mov_b32 v15, s16
+; GFX11-FAKE16-NEXT: v_dual_mov_b32 v16, s17 :: v_dual_mov_b32 v17, s18
+; GFX11-FAKE16-NEXT: s_and_not1_b32 vcc_lo, exec_lo, s15
; GFX11-FAKE16-NEXT: s_cbranch_vccnz .LBB55_3
; GFX11-FAKE16-NEXT: .LBB55_2: ; %cmp.true
; GFX11-FAKE16-NEXT: v_lshl_or_b32 v18, v37, 16, v51
@@ -37599,9 +37607,9 @@ define inreg <12 x double> @bitcast_v48f16_to_v12f64_scalar(<48 x half> inreg %a
; GFX11-FAKE16-NEXT: v_pk_add_f16 v12, 0x200, s13 op_sel_hi:[0,1]
; GFX11-FAKE16-NEXT: v_pk_add_f16 v13, 0x200, s14 op_sel_hi:[0,1]
; GFX11-FAKE16-NEXT: v_pk_add_f16 v14, 0x200, s0 op_sel_hi:[0,1]
-; GFX11-FAKE16-NEXT: v_pk_add_f16 v15, 0x200, s15 op_sel_hi:[0,1]
-; GFX11-FAKE16-NEXT: v_pk_add_f16 v16, 0x200, s16 op_sel_hi:[0,1]
-; GFX11-FAKE16-NEXT: v_pk_add_f16 v17, 0x200, s17 op_sel_hi:[0,1]
+; GFX11-FAKE16-NEXT: v_pk_add_f16 v15, 0x200, s16 op_sel_hi:[0,1]
+; GFX11-FAKE16-NEXT: v_pk_add_f16 v16, 0x200, s17 op_sel_hi:[0,1]
+; GFX11-FAKE16-NEXT: v_pk_add_f16 v17, 0x200, s18 op_sel_hi:[0,1]
; GFX11-FAKE16-NEXT: v_pk_add_f16 v18, 0x200, v18 op_sel_hi:[0,1]
; GFX11-FAKE16-NEXT: v_pk_add_f16 v19, 0x200, v19 op_sel_hi:[0,1]
; GFX11-FAKE16-NEXT: v_pk_add_f16 v20, 0x200, v20 op_sel_hi:[0,1]
@@ -39746,16 +39754,17 @@ define inreg <48 x half> @bitcast_v48i16_to_v48f16_scalar(<48 x i16> inreg %a, i
; GFX11-TRUE16-NEXT: v_and_b32_e32 v5, 0xffff, v5
; GFX11-TRUE16-NEXT: v_mov_b16_e32 v7.l, v4.h
; GFX11-TRUE16-NEXT: v_and_b32_e32 v4, 0xffff, v4
+; GFX11-TRUE16-NEXT: v_mov_b16_e32 v9.l, v0.h
+; GFX11-TRUE16-NEXT: v_and_b32_e32 v0, 0xffff, v0
; GFX11-TRUE16-NEXT: v_mov_b16_e32 v8.l, v3.h
-; GFX11-TRUE16-NEXT: v_and_b32_e32 v3, 0xffff, v3
; GFX11-TRUE16-NEXT: v_lshl_or_b32 v5, v6, 16, v5
+; GFX11-TRUE16-NEXT: v_and_b32_e32 v3, 0xffff, v3
; GFX11-TRUE16-NEXT: v_mov_b16_e32 v6.l, v2.h
; GFX11-TRUE16-NEXT: v_and_b32_e32 v2, 0xffff, v2
; GFX11-TRUE16-NEXT: v_lshl_or_b32 v4, v7, 16, v4
; GFX11-TRUE16-NEXT: v_mov_b16_e32 v7.l, v1.h
; GFX11-TRUE16-NEXT: v_and_b32_e32 v1, 0xffff, v1
-; GFX11-TRUE16-NEXT: v_mov_b16_e32 v9.l, v0.h
-; GFX11-TRUE16-NEXT: v_and_b32_e32 v0, 0xffff, v0
+; GFX11-TRUE16-NEXT: v_lshl_or_b32 v0, v9, 16, v0
; GFX11-TRUE16-NEXT: s_pack_ll_b32_b16 s15, s23, s15
; GFX11-TRUE16-NEXT: s_pack_ll_b32_b16 s14, s22, s14
; GFX11-TRUE16-NEXT: v_pk_add_u16 v11, s15, 3 op_sel_hi:[1,0]
@@ -39765,7 +39774,7 @@ define inreg <48 x half> @bitcast_v48i16_to_v48f16_scalar(<48 x i16> inreg %a, i
; GFX11-TRUE16-NEXT: v_lshl_or_b32 v3, v8, 16, v3
; GFX11-TRUE16-NEXT: v_lshl_or_b32 v2, v6, 16, v2
; GFX11-TRUE16-NEXT: v_lshl_or_b32 v1, v7, 16, v1
-; GFX11-TRUE16-NEXT: v_lshl_or_b32 v0, v9, 16, v0
+; GFX11-TRUE16-NEXT: v_pk_add_u16 v0, v0, 3 op_sel_hi:[1,0]
; GFX11-TRUE16-NEXT: s_pack_ll_b32_b16 s13, s15, s13
; GFX11-TRUE16-NEXT: s_pack_ll_b32_b16 s11, s14, s11
; GFX11-TRUE16-NEXT: v_pk_add_u16 v13, s13, 3 op_sel_hi:[1,0]
@@ -39779,7 +39788,7 @@ define inreg <48 x half> @bitcast_v48i16_to_v48f16_scalar(<48 x i16> inreg %a, i
; GFX11-TRUE16-NEXT: v_pk_add_u16 v3, v3, 3 op_sel_hi:[1,0]
; GFX11-TRUE16-NEXT: v_pk_add_u16 v2, v2, 3 op_sel_hi:[1,0]
; GFX11-TRUE16-NEXT: v_pk_add_u16 v1, v1, 3 op_sel_hi:[1,0]
-; GFX11-TRUE16-NEXT: v_pk_add_u16 v0, v0, 3 op_sel_hi:[1,0]
+; GFX11-TRUE16-NEXT: v_lshrrev_b32_e32 v50, 16, v0
; GFX11-TRUE16-NEXT: v_pk_add_u16 v7, s10, 3 op_sel_hi:[1,0]
; GFX11-TRUE16-NEXT: s_mov_b32 s10, s17
; GFX11-TRUE16-NEXT: v_pk_add_u16 v8, s7, 3 op_sel_hi:[1,0]
@@ -39807,18 +39816,18 @@ define inreg <48 x half> @bitcast_v48i16_to_v48f16_scalar(<48 x i16> inreg %a, i
; GFX11-TRUE16-NEXT: v_pk_add_u16 v26, s1, 3 op_sel_hi:[1,0]
; GFX11-TRUE16-NEXT: v_pk_add_u16 v25, s2, 3 op_sel_hi:[1,0]
; GFX11-TRUE16-NEXT: v_pk_add_u16 v24, s3, 3 op_sel_hi:[1,0]
-; GFX11-TRUE16-NEXT: v_pk_add_u16 v23, s5, 3 op_sel_hi:[1,0]
-; GFX11-TRUE16-NEXT: v_lshrrev_b32_e32 v50, 16, v0
-; GFX11-TRUE16-NEXT: v_lshrrev_b32_e32 v51, 16, v1
-; GFX11-TRUE16-NEXT: v_lshrrev_b32_e32 v52, 16, v2
-; GFX11-TRUE16-NEXT: v_lshrrev_b32_e32 v53, 16, v3
-; GFX11-TRUE16-NEXT: v_lshrrev_b32_e32 v54, 16, v4
-; GFX11-TRUE16-NEXT: v_lshrrev_b32_e32 v55, 16, v5
+; GFX11-TRUE16-NEXT: v_pk_add_u16 v22, s5, 3 op_sel_hi:[1,0]
+; GFX11-TRUE16-NEXT: v_mov_b16_e32 v0.h, v50.l
+; GFX11-TRUE16-NEXT: v_lshrrev_b32_e32 v50, 16, v1
+; GFX11-TRUE16-NEXT: v_lshrrev_b32_e32 v51, 16, v2
+; GFX11-TRUE16-NEXT: v_lshrrev_b32_e32 v52, 16, v3
+; GFX11-TRUE16-NEXT: v_lshrrev_b32_e32 v53, 16, v4
+; GFX11-TRUE16-NEXT: v_lshrrev_b32_e32 v54, 16, v5
; GFX11-TRUE16-NEXT: v_lshrrev_b32_e32 v38, 16, v27
; GFX11-TRUE16-NEXT: v_lshrrev_b32_e32 v39, 16, v26
; GFX11-TRUE16-NEXT: v_lshrrev_b32_e32 v48, 16, v25
; GFX11-TRUE16-NEXT: v_lshrrev_b32_e32 v49, 16, v24
-; GFX11-TRUE16-NEXT: v_lshrrev_b32_e32 v28, 16, v23
+; GFX11-TRUE16-NEXT: v_lshrrev_b32_e32 v28, 16, v22
; GFX11-TRUE16-NEXT: v_lshrrev_b32_e32 v29, 16, v18
; GFX11-TRUE16-NEXT: v_lshrrev_b32_e32 v35, 16, v8
; GFX11-TRUE16-NEXT: v_lshrrev_b32_e32 v36, 16, v7
@@ -39828,16 +39837,15 @@ define inreg <48 x half> @bitcast_v48i16_to_v48f16_scalar(<48 x i16> inreg %a, i
; GFX11-TRUE16-NEXT: v_lshrrev_b32_e32 v32, 16, v11
; GFX11-TRUE16-NEXT: v_lshrrev_b32_e32 v33, 16, v10
; GFX11-TRUE16-NEXT: v_lshrrev_b32_e32 v34, 16, v9
-; GFX11-TRUE16-NEXT: v_lshrrev_b32_e32 v19, 16, v17
-; GFX11-TRUE16-NEXT: v_lshrrev_b32_e32 v20, 16, v16
-; GFX11-TRUE16-NEXT: v_lshrrev_b32_e32 v21, 16, v15
-; GFX11-TRUE16-NEXT: v_lshrrev_b32_e32 v22, 16, v14
-; GFX11-TRUE16-NEXT: v_mov_b16_e32 v0.h, v50.l
-; GFX11-TRUE16-NEXT: v_mov_b16_e32 v1.h, v51.l
-; GFX11-TRUE16-NEXT: v_mov_b16_e32 v2.h, v52.l
-; GFX11-TRUE16-NEXT: v_mov_b16_e32 v3.h, v53.l
-; GFX11-TRUE16-NEXT: v_mov_b16_e32 v4.h, v54.l
-; GFX11-TRUE16-NEXT: v_mov_b16_e32 v5.h, v55.l
+; GFX11-TRUE16-NEXT: v_lshrrev_b32_e32 v23, 16, v17
+; GFX11-TRUE16-NEXT: v_lshrrev_b32_e32 v21, 16, v16
+; GFX11-TRUE16-NEXT: v_lshrrev_b32_e32 v20, 16, v15
+; GFX11-TRUE16-NEXT: v_lshrrev_b32_e32 v19, 16, v14
+; GFX11-TRUE16-NEXT: v_mov_b16_e32 v1.h, v50.l
+; GFX11-TRUE16-NEXT: v_mov_b16_e32 v2.h, v51.l
+; GFX11-TRUE16-NEXT: v_mov_b16_e32 v3.h, v52.l
+; GFX11-TRUE16-NEXT: v_mov_b16_e32 v4.h, v53.l
+; GFX11-TRUE16-NEXT: v_mov_b16_e32 v5.h, v54.l
; GFX11-TRUE16-NEXT: s_branch .LBB57_5
; GFX11-TRUE16-NEXT: .LBB57_3:
; GFX11-TRUE16-NEXT: s_branch .LBB57_2
@@ -39855,15 +39863,15 @@ define inreg <48 x half> @bitcast_v48i16_to_v48f16_scalar(<48 x i16> inreg %a, i
; GFX11-TRUE16-NEXT: v_mov_b16_e32 v7.l, s19
; GFX11-TRUE16-NEXT: v_mov_b16_e32 v8.l, s18
; GFX11-TRUE16-NEXT: v_mov_b16_e32 v18.l, s17
-; GFX11-TRUE16-NEXT: v_mov_b16_e32 v23.l, s16
+; GFX11-TRUE16-NEXT: v_mov_b16_e32 v22.l, s16
; GFX11-TRUE16-NEXT: v_mov_b16_e32 v24.l, s3
; GFX11-TRUE16-NEXT: v_mov_b16_e32 v25.l, s2
; GFX11-TRUE16-NEXT: v_mov_b16_e32 v26.l, s1
; GFX11-TRUE16-NEXT: v_mov_b16_e32 v27.l, s0
-; GFX11-TRUE16-NEXT: v_mov_b16_e32 v22.l, s45
-; GFX11-TRUE16-NEXT: v_mov_b16_e32 v21.l, s44
-; GFX11-TRUE16-NEXT: v_mov_b16_e32 v20.l, s43
-; GFX11-TRUE16-NEXT: v_mov_b16_e32 v19.l, s42
+; GFX11-TRUE16-NEXT: v_mov_b16_e32 v19.l, s45
+; GFX11-TRUE16-NEXT: v_mov_b16_e32 v20.l, s44
+; GFX11-TRUE16-NEXT: v_mov_b16_e32 v21.l, s43
+; GFX11-TRUE16-NEXT: v_mov_b16_e32 v23.l, s42
; GFX11-TRUE16-NEXT: v_mov_b16_e32 v34.l, s41
; GFX11-TRUE16-NEXT: v_mov_b16_e32 v33.l, s40
; GFX11-TRUE16-NEXT: v_mov_b16_e32 v32.l, s15
@@ -39880,31 +39888,30 @@ define inreg <48 x half> @bitcast_v48i16_to_v48f16_scalar(<48 x i16> inreg %a, i
; GFX11-TRUE16-NEXT: v_mov_b16_e32 v38.l, s9
; GFX11-TRUE16-NEXT: .LBB57_5: ; %end
; GFX11-TRUE16-NEXT: v_and_b32_e32 v18, 0xffff, v18
-; GFX11-TRUE16-NEXT: v_and_b32_e32 v23, 0xffff, v23
-; GFX11-TRUE16-NEXT: v_and_b32_e32 v12, 0xffff, v12
; GFX11-TRUE16-NEXT: v_and_b32_e32 v27, 0xffff, v27
; GFX11-TRUE16-NEXT: v_and_b32_e32 v51, 0xffff, v24
+; GFX11-TRUE16-NEXT: v_and_b32_e32 v13, 0xffff, v13
+; GFX11-TRUE16-NEXT: v_and_b32_e32 v12, 0xffff, v12
; GFX11-TRUE16-NEXT: v_lshl_or_b32 v29, v29, 16, v18
+; GFX11-TRUE16-NEXT: v_and_b32_e32 v22, 0xffff, v22
; GFX11-TRUE16-NEXT: v_and_b32_e32 v18, 0xffff, v10
-; GFX11-TRUE16-NEXT: v_lshl_or_b32 v28, v28, 16, v23
-; GFX11-TRUE16-NEXT: v_and_b32_e32 v13, 0xffff, v13
-; GFX11-TRUE16-NEXT: v_and_b32_e32 v23, 0xffff, v9
-; GFX11-TRUE16-NEXT: v_lshl_or_b32 v10, v31, 16, v12
-; GFX11-TRUE16-NEXT: v_lshl_or_b32 v12, v33, 16, v18
; GFX11-TRUE16-NEXT: v_and_b32_e32 v16, 0xffff, v16
-; GFX11-TRUE16-NEXT: v_and_b32_e32 v18, 0xffff, v15
; GFX11-TRUE16-NEXT: v_and_b32_e32 v26, 0xffff, v26
; GFX11-TRUE16-NEXT: v_lshl_or_b32 v24, v38, 16, v27
+; GFX11-TRUE16-NEXT: v_lshl_or_b32 v28, v28, 16, v22
+; GFX11-TRUE16-NEXT: v_and_b32_e32 v22, 0xffff, v9
; GFX11-TRUE16-NEXT: v_lshl_or_b32 v27, v49, 16, v51
; GFX11-TRUE16-NEXT: v_and_b32_e32 v50, 0xffff, v25
; GFX11-TRUE16-NEXT: v_lshl_or_b32 v9, v30, 16, v13
-; GFX11-TRUE16-NEXT: v_lshl_or_b32 v13, v34, 16, v23
+; GFX11-TRUE16-NEXT: v_lshl_or_b32 v10, v31, 16, v12
+; GFX11-TRUE16-NEXT: v_lshl_or_b32 v12, v33, 16, v18
+; GFX11-TRUE16-NEXT: v_lshl_or_b32 v13, v34, 16, v22
; GFX11-TRUE16-NEXT: v_and_b32_e32 v17, 0xffff, v17
-; GFX11-TRUE16-NEXT: v_and_b32_e32 v23, 0xffff, v14
+; GFX11-TRUE16-NEXT: v_and_b32_e32 v18, 0xffff, v15
+; GFX11-TRUE16-NEXT: v_and_b32_e32 v22, 0xffff, v14
; GFX11-TRUE16-NEXT: v_mov_b16_e32 v30.l, v0.h
; GFX11-TRUE16-NEXT: v_and_b32_e32 v0, 0xffff, v0
-; GFX11-TRUE16-NEXT: v_lshl_or_b32 v15, v20, 16, v16
-; GFX11-TRUE16-NEXT: v_lshl_or_b32 v16, v21, 16, v18
+; GFX11-TRUE16-NEXT: v_lshl_or_b32 v15, v21, 16, v16
; GFX11-TRUE16-NEXT: v_mov_b16_e32 v21.l, v3.h
; GFX11-TRUE16-NEXT: v_and_b32_e32 v3, 0xffff, v3
; GFX11-TRUE16-NEXT: v_lshl_or_b32 v25, v39, 16, v26
@@ -39912,8 +39919,9 @@ define inreg <48 x half> @bitcast_v48i16_to_v48f16_scalar(<48 x i16> inreg %a, i
; GFX11-TRUE16-NEXT: v_and_b32_e32 v7, 0xffff, v7
; GFX11-TRUE16-NEXT: v_and_b32_e32 v38, 0xffff, v6
; GFX11-TRUE16-NEXT: v_and_b32_e32 v11, 0xffff, v11
-; GFX11-TRUE16-NEXT: v_lshl_or_b32 v14, v19, 16, v17
-; GFX11-TRUE16-NEXT: v_lshl_or_b32 v17, v22, 16, v23
+; GFX11-TRUE16-NEXT: v_lshl_or_b32 v14, v23, 16, v17
+; GFX11-TRUE16-NEXT: v_lshl_or_b32 v16, v20, 16, v18
+; GFX11-TRUE16-NEXT: v_lshl_or_b32 v17, v19, 16, v22
; GFX11-TRUE16-NEXT: v_lshl_or_b32 v18, v30, 16, v0
; GFX11-TRUE16-NEXT: v_mov_b16_e32 v0.l, v1.h
; GFX11-TRUE16-NEXT: v_and_b32_e32 v1, 0xffff, v1
@@ -39927,7 +39935,6 @@ define inreg <48 x half> @bitcast_v48i16_to_v48f16_scalar(<48 x i16> inreg %a, i
; GFX11-TRUE16-NEXT: v_mov_b32_e32 v3, v27
; GFX11-TRUE16-NEXT: v_lshl_or_b32 v26, v48, 16, v50
; GFX11-TRUE16-NEXT: v_lshl_or_b32 v6, v35, 16, v8
-; GFX11-TRUE16-NEXT: v_lshl_or_b32 v7, v36, 16, v7
; GFX11-TRUE16-NEXT: v_lshl_or_b32 v8, v37, 16, v38
; GFX11-TRUE16-NEXT: v_lshl_or_b32 v11, v32, 16, v11
; GFX11-TRUE16-NEXT: v_lshl_or_b32 v19, v0, 16, v1
@@ -39936,7 +39943,9 @@ define inreg <48 x half> @bitcast_v48i16_to_v48f16_scalar(<48 x i16> inreg %a, i
; GFX11-TRUE16-NEXT: v_lshl_or_b32 v23, v23, 16, v5
; GFX11-TRUE16-NEXT: v_dual_mov_b32 v0, v24 :: v_dual_mov_b32 v1, v25
; GFX11-TRUE16-NEXT: v_mov_b32_e32 v2, v26
-; GFX11-TRUE16-NEXT: v_dual_mov_b32 v4, v28 :: v_dual_mov_b32 v5, v29
+; GFX11-TRUE16-NEXT: v_mov_b32_e32 v4, v28
+; GFX11-TRUE16-NEXT: v_lshl_or_b32 v7, v36, 16, v7
+; GFX11-TRUE16-NEXT: v_mov_b32_e32 v5, v29
; GFX11-TRUE16-NEXT: s_setpc_b64 s[30:31]
;
; GFX11-FAKE16-LABEL: bitcast_v48i16_to_v48f16_scalar:
@@ -41941,16 +41950,17 @@ define inreg <48 x i16> @bitcast_v48f16_to_v48i16_scalar(<48 x half> inreg %a, i
; GFX11-TRUE16-NEXT: v_and_b32_e32 v5, 0xffff, v5
; GFX11-TRUE16-NEXT: v_mov_b16_e32 v7.l, v4.h
; GFX11-TRUE16-NEXT: v_and_b32_e32 v4, 0xffff, v4
+; GFX11-TRUE16-NEXT: v_mov_b16_e32 v9.l, v0.h
+; GFX11-TRUE16-NEXT: v_and_b32_e32 v0, 0xffff, v0
; GFX11-TRUE16-NEXT: v_mov_b16_e32 v8.l, v3.h
-; GFX11-TRUE16-NEXT: v_and_b32_e32 v3, 0xffff, v3
; GFX11-TRUE16-NEXT: v_lshl_or_b32 v5, v6, 16, v5
+; GFX11-TRUE16-NEXT: v_and_b32_e32 v3, 0xffff, v3
; GFX11-TRUE16-NEXT: v_mov_b16_e32 v6.l, v2.h
; GFX11-TRUE16-NEXT: v_and_b32_e32 v2, 0xffff, v2
; GFX11-TRUE16-NEXT: v_lshl_or_b32 v4, v7, 16, v4
; GFX11-TRUE16-NEXT: v_mov_b16_e32 v7.l, v1.h
; GFX11-TRUE16-NEXT: v_and_b32_e32 v1, 0xffff, v1
-; GFX11-TRUE16-NEXT: v_mov_b16_e32 v9.l, v0.h
-; GFX11-TRUE16-NEXT: v_and_b32_e32 v0, 0xffff, v0
+; GFX11-TRUE16-NEXT: v_lshl_or_b32 v0, v9, 16, v0
; GFX11-TRUE16-NEXT: s_pack_ll_b32_b16 s15, s23, s15
; GFX11-TRUE16-NEXT: s_pack_ll_b32_b16 s14, s22, s14
; GFX11-TRUE16-NEXT: v_pk_add_f16 v11, 0x200, s15 op_sel_hi:[0,1]
@@ -41960,7 +41970,7 @@ define inreg <48 x i16> @bitcast_v48f16_to_v48i16_scalar(<48 x half> inreg %a, i
; GFX11-TRUE16-NEXT: v_lshl_or_b32 v3, v8, 16, v3
; GFX11-TRUE16-NEXT: v_lshl_or_b32 v2, v6, 16, v2
; GFX11-TRUE16-NEXT: v_lshl_or_b32 v1, v7, 16, v1
-; GFX11-TRUE16-NEXT: v_lshl_or_b32 v0, v9, 16, v0
+; GFX11-TRUE16-NEXT: v_pk_add_f16 v0, 0x200, v0 op_sel_hi:[0,1]
; GFX11-TRUE16-NEXT: s_pack_ll_b32_b16 s13, s15, s13
; GFX11-TRUE16-NEXT: s_pack_ll_b32_b16 s11, s14, s11
; GFX11-TRUE16-NEXT: v_pk_add_f16 v13, 0x200, s13 op_sel_hi:[0,1]
@@ -41974,7 +41984,7 @@ define inreg <48 x i16> @bitcast_v48f16_to_v48i16_scalar(<48 x half> inreg %a, i
; GFX11-TRUE16-NEXT: v_pk_add_f16 v3, 0x200, v3 op_sel_hi:[0,1]
; GFX11-TRUE16-NEXT: v_pk_add_f16 v2, 0x200, v2 op_sel_hi:[0,1]
; GFX11-TRUE16-NEXT: v_pk_add_f16 v1, 0x200, v1 op_sel_hi:[0,1]
-; GFX11-TRUE16-NEXT: v_pk_add_f16 v0, 0x200, v0 op_sel_hi:[0,1]
+; GFX11-TRUE16-NEXT: v_lshrrev_b32_e32 v50, 16, v0
; GFX11-TRUE16-NEXT: v_pk_add_f16 v7, 0x200, s10 op_sel_hi:[0,1]
; GFX11-TRUE16-NEXT: s_mov_b32 s10, s17
; GFX11-TRUE16-NEXT: v_pk_add_f16 v8, 0x200, s7 op_sel_hi:[0,1]
@@ -42002,18 +42012,18 @@ define inreg <48 x i16> @bitcast_v48f16_to_v48i16_scalar(<48 x half> inreg %a, i
; GFX11-TRUE16-NEXT: v_pk_add_f16 v26, 0x200, s1 op_sel_hi:[0,1]
; GFX11-TRUE16-NEXT: v_pk_add_f16 v25, 0x200, s2 op_sel_hi:[0,1]
; GFX11-TRUE16-NEXT: v_pk_add_f16 v24, 0x200, s3 op_sel_hi:[0,1]
-; GFX11-TRUE16-NEXT: v_pk_add_f16 v23, 0x200, s5 op_sel_hi:[0,1]
-; GFX11-TRUE16-NEXT: v_lshrrev_b32_e32 v50, 16, v0
-; GFX11-TRUE16-NEXT: v_lshrrev_b32_e32 v51, 16, v1
-; GFX11-TRUE16-NEXT: v_lshrrev_b32_e32 v52, 16, v2
-; GFX11-TRUE16-NEXT: v_lshrrev_b32_e32 v53, 16, v3
-; GFX11-TRUE16-NEXT: v_lshrrev_b32_e32 v54, 16, v4
-; GFX11-TRUE16-NEXT: v_lshrrev_b32_e32 v55, 16, v5
+; GFX11-TRUE16-NEXT: v_pk_add_f16 v22, 0x200, s5 op_sel_hi:[0,1]
+; GFX11-TRUE16-NEXT: v_mov_b16_e32 v0.h, v50.l
+; GFX11-TRUE16-NEXT: v_lshrrev_b32_e32 v50, 16, v1
+; GFX11-TRUE16-NEXT: v_lshrrev_b32_e32 v51, 16, v2
+; GFX11-TRUE16-NEXT: v_lshrrev_b32_e32 v52, 16, v3
+; GFX11-TRUE16-NEXT: v_lshrrev_b32_e32 v53, 16, v4
+; GFX11-TRUE16-NEXT: v_lshrrev_b32_e32 v54, 16, v5
; GFX11-TRUE16-NEXT: v_lshrrev_b32_e32 v38, 16, v27
; GFX11-TRUE16-NEXT: v_lshrrev_b32_e32 v39, 16, v26
; GFX11-TRUE16-NEXT: v_lshrrev_b32_e32 v48, 16, v25
; GFX11-TRUE16-NEXT: v_lshrrev_b32_e32 v49, 16, v24
-; GFX11-TRUE16-NEXT: v_lshrrev_b32_e32 v28, 16, v23
+; GFX11-TRUE16-NEXT: v_lshrrev_b32_e32 v28, 16, v22
; GFX11-TRUE16-NEXT: v_lshrrev_b32_e32 v29, 16, v18
; GFX11-TRUE16-NEXT: v_lshrrev_b32_e32 v35, 16, v8
; GFX11-TRUE16-NEXT: v_lshrrev_b32_e32 v36, 16, v7
@@ -42023,16 +42033,15 @@ define inreg <48 x i16> @bitcast_v48f16_to_v48i16_scalar(<48 x half> inreg %a, i
; GFX11-TRUE16-NEXT: v_lshrrev_b32_e32 v32, 16, v11
; GFX11-TRUE16-NEXT: v_lshrrev_b32_e32 v33, 16, v10
; GFX11-TRUE16-NEXT: v_lshrrev_b32_e32 v34, 16, v9
-; GFX11-TRUE16-NEXT: v_lshrrev_b32_e32 v19, 16, v17
-; GFX11-TRUE16-NEXT: v_lshrrev_b32_e32 v20, 16, v16
-; GFX11-TRUE16-NEXT: v_lshrrev_b32_e32 v21, 16, v15
-; GFX11-TRUE16-NEXT: v_lshrrev_b32_e32 v22, 16, v14
-; GFX11-TRUE16-NEXT: v_mov_b16_e32 v0.h, v50.l
-; GFX11-TRUE16-NEXT: v_mov_b16_e32 v1.h, v51.l
-; GFX11-TRUE16-NEXT: v_mov_b16_e32 v2.h, v52.l
-; GFX11-TRUE16-NEXT: v_mov_b16_e32 v3.h, v53.l
-; GFX11-TRUE16-NEXT: v_mov_b16_e32 v4.h, v54.l
-; GFX11-TRUE16-NEXT: v_mov_b16_e32 v5.h, v55.l
+; GFX11-TRUE16-NEXT: v_lshrrev_b32_e32 v23, 16, v17
+; GFX11-TRUE16-NEXT: v_lshrrev_b32_e32 v21, 16, v16
+; GFX11-TRUE16-NEXT: v_lshrrev_b32_e32 v20, 16, v15
+; GFX11-TRUE16-NEXT: v_lshrrev_b32_e32 v19, 16, v14
+; GFX11-TRUE16-NEXT: v_mov_b16_e32 v1.h, v50.l
+; GFX11-TRUE16-NEXT: v_mov_b16_e32 v2.h, v51.l
+; GFX11-TRUE16-NEXT: v_mov_b16_e32 v3.h, v52.l
+; GFX11-TRUE16-NEXT: v_mov_b16_e32 v4.h, v53.l
+; GFX11-TRUE16-NEXT: v_mov_b16_e32 v5.h, v54.l
; GFX11-TRUE16-NEXT: s_branch .LBB59_5
; GFX11-TRUE16-NEXT: .LBB59_3:
; GFX11-TRUE16-NEXT: s_branch .LBB59_2
@@ -42050,15 +42059,15 @@ define inreg <48 x i16> @bitcast_v48f16_to_v48i16_scalar(<48 x half> inreg %a, i
; GFX11-TRUE16-NEXT: v_mov_b16_e32 v7.l, s19
; GFX11-TRUE16-NEXT: v_mov_b16_e32 v8.l, s18
; GFX11-TRUE16-NEXT: v_mov_b16_e32 v18.l, s17
-; GFX11-TRUE16-NEXT: v_mov_b16_e32 v23.l, s16
+; GFX11-TRUE16-NEXT: v_mov_b16_e32 v22.l, s16
; GFX11-TRUE16-NEXT: v_mov_b16_e32 v24.l, s3
; GFX11-TRUE16-NEXT: v_mov_b16_e32 v25.l, s2
; GFX11-TRUE16-NEXT: v_mov_b16_e32 v26.l, s1
; GFX11-TRUE16-NEXT: v_mov_b16_e32 v27.l, s0
-; GFX11-TRUE16-NEXT: v_mov_b16_e32 v22.l, s45
-; GFX11-TRUE16-NEXT: v_mov_b16_e32 v21.l, s44
-; GFX11-TRUE16-NEXT: v_mov_b16_e32 v20.l, s43
-; GFX11-TRUE16-NEXT: v_mov_b16_e32 v19.l, s42
+; GFX11-TRUE16-NEXT: v_mov_b16_e32 v19.l, s45
+; GFX11-TRUE16-NEXT: v_mov_b16_e32 v20.l, s44
+; GFX11-TRUE16-NEXT: v_mov_b16_e32 v21.l, s43
+; GFX11-TRUE16-NEXT: v_mov_b16_e32 v23.l, s42
; GFX11-TRUE16-NEXT: v_mov_b16_e32 v34.l, s41
; GFX11-TRUE16-NEXT: v_mov_b16_e32 v33.l, s40
; GFX11-TRUE16-NEXT: v_mov_b16_e32 v32.l, s15
@@ -42075,31 +42084,30 @@ define inreg <48 x i16> @bitcast_v48f16_to_v48i16_scalar(<48 x half> inreg %a, i
; GFX11-TRUE16-NEXT: v_mov_b16_e32 v38.l, s9
; GFX11-TRUE16-NEXT: .LBB59_5: ; %end
; GFX11-TRUE16-NEXT: v_and_b32_e32 v18, 0xffff, v18
-; GFX11-TRUE16-NEXT: v_and_b32_e32 v23, 0xffff, v23
-; GFX11-TRUE16-NEXT: v_and_b32_e32 v12, 0xffff, v12
; GFX11-TRUE16-NEXT: v_and_b32_e32 v27, 0xffff, v27
; GFX11-TRUE16-NEXT: v_and_b32_e32 v51, 0xffff, v24
+; GFX11-TRUE16-NEXT: v_and_b32_e32 v13, 0xffff, v13
+; GFX11-TRUE16-NEXT: v_and_b32_e32 v12, 0xffff, v12
; GFX11-TRUE16-NEXT: v_lshl_or_b32 v29, v29, 16, v18
+; GFX11-TRUE16-NEXT: v_and_b32_e32 v22, 0xffff, v22
; GFX11-TRUE16-NEXT: v_and_b32_e32 v18, 0xffff, v10
-; GFX11-TRUE16-NEXT: v_lshl_or_b32 v28, v28, 16, v23
-; GFX11-TRUE16-NEXT: v_and_b32_e32 v13, 0xffff, v13
-; GFX11-TRUE16-NEXT: v_and_b32_e32 v23, 0xffff, v9
-; GFX11-TRUE16-NEXT: v_lshl_or_b32 v10, v31, 16, v12
-; GFX11-TRUE16-NEXT: v_lshl_or_b32 v12, v33, 16, v18
; GFX11-TRUE16-NEXT: v_and_b32_e32 v16, 0xffff, v16
-; GFX11-TRUE16-NEXT: v_and_b32_e32 v18, 0xffff, v15
; GFX11-TRUE16-NEXT: v_and_b32_e32 v26, 0xffff, v26
; GFX11-TRUE16-NEXT: v_lshl_or_b32 v24, v38, 16, v27
+; GFX11-TRUE16-NEXT: v_lshl_or_b32 v28, v28, 16, v22
+; GFX11-TRUE16-NEXT: v_and_b32_e32 v22, 0xffff, v9
; GFX11-TRUE16-NEXT: v_lshl_or_b32 v27, v49, 16, v51
; GFX11-TRUE16-NEXT: v_and_b32_e32 v50, 0xffff, v25
; GFX11-TRUE16-NEXT: v_lshl_or_b32 v9, v30, 16, v13
-; GFX11-TRUE16-NEXT: v_lshl_or_b32 v13, v34, 16, v23
+; GFX11-TRUE16-NEXT: v_lshl_or_b32 v10, v31, 16, v12
+; GFX11-TRUE16-NEXT: v_lshl_or_b32 v12, v33, 16, v18
+; GFX11-TRUE16-NEXT: v_lshl_or_b32 v13, v34, 16, v22
; GFX11-TRUE16-NEXT: v_and_b32_e32 v17, 0xffff, v17
-; GFX11-TRUE16-NEXT: v_and_b32_e32 v23, 0xffff, v14
+; GFX11-TRUE16-NEXT: v_and_b32_e32 v18, 0xffff, v15
+; GFX11-TRUE16-NEXT: v_and_b32_e32 v22, 0xffff, v14
; GFX11-TRUE16-NEXT: v_mov_b16_e32 v30.l, v0.h
; GFX11-TRUE16-NEXT: v_and_b32_e32 v0, 0xffff, v0
-; GFX11-TRUE16-NEXT: v_lshl_or_b32 v15, v20, 16, v16
-; GFX11-TRUE16-NEXT: v_lshl_or_b32 v16, v21, 16, v18
+; GFX11-TRUE16-NEXT: v_lshl_or_b32 v15, v21, 16, v16
; GFX11-TRUE16-NEXT: v_mov_b16_e32 v21.l, v3.h
; GFX11-TRUE16-NEXT: v_and_b32_e32 v3, 0xffff, v3
; GFX11-TRUE16-NEXT: v_lshl_or_b32 v25, v39, 16, v26
@@ -42107,8 +42115,9 @@ define inreg <48 x i16> @bitcast_v48f16_to_v48i16_scalar(<48 x half> inreg %a, i
; GFX11-TRUE16-NEXT: v_and_b32_e32 v7, 0xffff, v7
; GFX11-TRUE16-NEXT: v_and_b32_e32 v38, 0xffff, v6
; GFX11-TRUE16-NEXT: v_and_b32_e32 v11, 0xffff, v11
-; GFX11-TRUE16-NEXT: v_lshl_or_b32 v14, v19, 16, v17
-; GFX11-TRUE16-NEXT: v_lshl_or_b32 v17, v22, 16, v23
+; GFX11-TRUE16-NEXT: v_lshl_or_b32 v14, v23, 16, v17
+; GFX11-TRUE16-NEXT: v_lshl_or_b32 v16, v20, 16, v18
+; GFX11-TRUE16-NEXT: v_lshl_or_b32 v17, v19, 16, v22
; GFX11-TRUE16-NEXT: v_lshl_or_b32 v18, v30, 16, v0
; GFX11-TRUE16-NEXT: v_mov_b16_e32 v0.l, v1.h
; GFX11-TRUE16-NEXT: v_and_b32_e32 v1, 0xffff, v1
@@ -42122,7 +42131,6 @@ define inreg <48 x i16> @bitcast_v48f16_to_v48i16_scalar(<48 x half> inreg %a, i
; GFX11-TRUE16-NEXT: v_mov_b32_e32 v3, v27
; GFX11-TRUE16-NEXT: v_lshl_or_b32 v26, v48, 16, v50
; GFX11-TRUE16-NEXT: v_lshl_or_b32 v6, v35, 16, v8
-; GFX11-TRUE16-NEXT: v_lshl_or_b32 v7, v36, 16, v7
; GFX11-TRUE16-NEXT: v_lshl_or_b32 v8, v37, 16, v38
; GFX11-TRUE16-NEXT: v_lshl_or_b32 v11, v32, 16, v11
; GFX11-TRUE16-NEXT: v_lshl_or_b32 v19, v0, 16, v1
@@ -42131,7 +42139,9 @@ define inreg <48 x i16> @bitcast_v48f16_to_v48i16_scalar(<48 x half> inreg %a, i
; GFX11-TRUE16-NEXT: v_lshl_or_b32 v23, v23, 16, v5
; GFX11-TRUE16-NEXT: v_dual_mov_b32 v0, v24 :: v_dual_mov_b32 v1, v25
; GFX11-TRUE16-NEXT: v_mov_b32_e32 v2, v26
-; GFX11-TRUE16-NEXT: v_dual_mov_b32 v4, v28 :: v_dual_mov_b32 v5, v29
+; GFX11-TRUE16-NEXT: v_mov_b32_e32 v4, v28
+; GFX11-TRUE16-NEXT: v_lshl_or_b32 v7, v36, 16, v7
+; GFX11-TRUE16-NEXT: v_mov_b32_e32 v5, v29
; GFX11-TRUE16-NEXT: s_setpc_b64 s[30:31]
;
; GFX11-FAKE16-LABEL: bitcast_v48f16_to_v48i16_scalar:
diff --git a/llvm/test/CodeGen/AMDGPU/amdgcn.bitcast.832bit.ll b/llvm/test/CodeGen/AMDGPU/amdgcn.bitcast.832bit.ll
index fbee320c82c7f..f4b487be4316b 100644
--- a/llvm/test/CodeGen/AMDGPU/amdgcn.bitcast.832bit.ll
+++ b/llvm/test/CodeGen/AMDGPU/amdgcn.bitcast.832bit.ll
@@ -4145,7 +4145,7 @@ define inreg <52 x i16> @bitcast_v26i32_to_v52i16_scalar(<26 x i32> inreg %a, i3
; GFX11-TRUE16-NEXT: v_readfirstlane_b32 s5, v5
; GFX11-TRUE16-NEXT: v_readfirstlane_b32 s6, v6
; GFX11-TRUE16-NEXT: v_readfirstlane_b32 s4, v7
-; GFX11-TRUE16-NEXT: s_mov_b32 s78, 0
+; GFX11-TRUE16-NEXT: s_mov_b32 s74, 0
; GFX11-TRUE16-NEXT: s_and_b32 s12, vcc_lo, exec_lo
; GFX11-TRUE16-NEXT: s_cbranch_scc0 .LBB13_4
; GFX11-TRUE16-NEXT: ; %bb.1: ; %cmp.false
@@ -4171,11 +4171,11 @@ define inreg <52 x i16> @bitcast_v26i32_to_v52i16_scalar(<26 x i32> inreg %a, i3
; GFX11-TRUE16-NEXT: s_lshr_b32 s63, s18, 16
; GFX11-TRUE16-NEXT: s_lshr_b32 s72, s17, 16
; GFX11-TRUE16-NEXT: s_lshr_b32 s73, s16, 16
-; GFX11-TRUE16-NEXT: s_lshr_b32 s74, s3, 16
-; GFX11-TRUE16-NEXT: s_lshr_b32 s75, s2, 16
-; GFX11-TRUE16-NEXT: s_lshr_b32 s76, s1, 16
-; GFX11-TRUE16-NEXT: s_lshr_b32 s77, s0, 16
-; GFX11-TRUE16-NEXT: s_and_not1_b32 vcc_lo, exec_lo, s78
+; GFX11-TRUE16-NEXT: s_lshr_b32 s75, s3, 16
+; GFX11-TRUE16-NEXT: s_lshr_b32 s76, s2, 16
+; GFX11-TRUE16-NEXT: s_lshr_b32 s77, s1, 16
+; GFX11-TRUE16-NEXT: s_lshr_b32 s78, s0, 16
+; GFX11-TRUE16-NEXT: s_and_not1_b32 vcc_lo, exec_lo, s74
; GFX11-TRUE16-NEXT: s_cbranch_vccnz .LBB13_3
; GFX11-TRUE16-NEXT: .LBB13_2: ; %cmp.true
; GFX11-TRUE16-NEXT: s_add_i32 s4, s4, 3
@@ -4226,16 +4226,18 @@ define inreg <52 x i16> @bitcast_v26i32_to_v52i16_scalar(<26 x i32> inreg %a, i3
; GFX11-TRUE16-NEXT: s_lshr_b32 s63, s18, 16
; GFX11-TRUE16-NEXT: s_lshr_b32 s72, s17, 16
; GFX11-TRUE16-NEXT: s_lshr_b32 s73, s16, 16
-; GFX11-TRUE16-NEXT: s_lshr_b32 s74, s3, 16
-; GFX11-TRUE16-NEXT: s_lshr_b32 s75, s2, 16
-; GFX11-TRUE16-NEXT: s_lshr_b32 s76, s1, 16
-; GFX11-TRUE16-NEXT: s_lshr_b32 s77, s0, 16
+; GFX11-TRUE16-NEXT: s_lshr_b32 s75, s3, 16
+; GFX11-TRUE16-NEXT: s_lshr_b32 s76, s2, 16
+; GFX11-TRUE16-NEXT: s_lshr_b32 s77, s1, 16
+; GFX11-TRUE16-NEXT: s_lshr_b32 s78, s0, 16
; GFX11-TRUE16-NEXT: .LBB13_3: ; %end
; GFX11-TRUE16-NEXT: s_delay_alu instid0(SALU_CYCLE_1)
-; GFX11-TRUE16-NEXT: s_pack_ll_b32_b16 s0, s0, s77
-; GFX11-TRUE16-NEXT: s_pack_ll_b32_b16 s1, s1, s76
-; GFX11-TRUE16-NEXT: s_pack_ll_b32_b16 s2, s2, s75
-; GFX11-TRUE16-NEXT: s_pack_ll_b32_b16 s3, s3, s74
+; GFX11-TRUE16-NEXT: s_mov_b32 s74, s78
+; GFX11-TRUE16-NEXT: s_pack_ll_b32_b16 s1, s1, s77
+; GFX11-TRUE16-NEXT: s_pack_ll_b32_b16 s0, s0, s74
+; GFX11-TRUE16-NEXT: s_mov_b32 s74, s76
+; GFX11-TRUE16-NEXT: s_pack_ll_b32_b16 s3, s3, s75
+; GFX11-TRUE16-NEXT: s_pack_ll_b32_b16 s2, s2, s74
; GFX11-TRUE16-NEXT: s_pack_ll_b32_b16 s16, s16, s73
; GFX11-TRUE16-NEXT: s_pack_ll_b32_b16 s17, s17, s72
; GFX11-TRUE16-NEXT: s_pack_ll_b32_b16 s18, s18, s63
@@ -4273,10 +4275,10 @@ define inreg <52 x i16> @bitcast_v26i32_to_v52i16_scalar(<26 x i32> inreg %a, i3
; GFX11-TRUE16-NEXT: v_dual_mov_b32 v24, s6 :: v_dual_mov_b32 v25, s4
; GFX11-TRUE16-NEXT: s_setpc_b64 s[30:31]
; GFX11-TRUE16-NEXT: .LBB13_4:
+; GFX11-TRUE16-NEXT: ; implicit-def: $sgpr78_lo16
; GFX11-TRUE16-NEXT: ; implicit-def: $sgpr77_lo16
; GFX11-TRUE16-NEXT: ; implicit-def: $sgpr76_lo16
; GFX11-TRUE16-NEXT: ; implicit-def: $sgpr75_lo16
-; GFX11-TRUE16-NEXT: ; implicit-def: $sgpr74_lo16
; GFX11-TRUE16-NEXT: ; implicit-def: $sgpr73_lo16
; GFX11-TRUE16-NEXT: ; implicit-def: $sgpr72_lo16
; GFX11-TRUE16-NEXT: ; implicit-def: $sgpr63_lo16
@@ -6638,8 +6640,8 @@ define inreg <26 x i32> @bitcast_v52i16_to_v26i32_scalar(<52 x i16> inreg %a, i3
; GFX11-FAKE16-NEXT: v_and_b32_e32 v50, 0xffff, v5
; GFX11-FAKE16-NEXT: v_and_b32_e32 v49, 0xffff, v6
; GFX11-FAKE16-NEXT: v_and_b32_e32 v48, 0xffff, v7
-; GFX11-FAKE16-NEXT: s_lshr_b32 s41, s29, 16
-; GFX11-FAKE16-NEXT: s_lshr_b32 s15, s28, 16
+; GFX11-FAKE16-NEXT: s_lshr_b32 s40, s29, 16
+; GFX11-FAKE16-NEXT: s_lshr_b32 s41, s28, 16
; GFX11-FAKE16-NEXT: s_lshr_b32 s42, s27, 16
; GFX11-FAKE16-NEXT: s_lshr_b32 s14, s26, 16
; GFX11-FAKE16-NEXT: s_lshr_b32 s13, s25, 16
@@ -6656,7 +6658,7 @@ define inreg <26 x i32> @bitcast_v52i16_to_v26i32_scalar(<52 x i16> inreg %a, i3
; GFX11-FAKE16-NEXT: s_lshr_b32 s44, s2, 16
; GFX11-FAKE16-NEXT: s_lshr_b32 s45, s1, 16
; GFX11-FAKE16-NEXT: s_lshr_b32 s46, s0, 16
-; GFX11-FAKE16-NEXT: s_mov_b32 s40, 0
+; GFX11-FAKE16-NEXT: s_mov_b32 s15, 0
; GFX11-FAKE16-NEXT: s_pack_ll_b32_b16 s0, s0, s46
; GFX11-FAKE16-NEXT: s_pack_ll_b32_b16 s1, s1, s45
; GFX11-FAKE16-NEXT: s_pack_ll_b32_b16 s2, s2, s44
@@ -6672,9 +6674,9 @@ define inreg <26 x i32> @bitcast_v52i16_to_v26i32_scalar(<52 x i16> inreg %a, i3
; GFX11-FAKE16-NEXT: s_pack_ll_b32_b16 s12, s24, s12
; GFX11-FAKE16-NEXT: s_pack_ll_b32_b16 s13, s25, s13
; GFX11-FAKE16-NEXT: s_pack_ll_b32_b16 s14, s26, s14
-; GFX11-FAKE16-NEXT: s_pack_ll_b32_b16 s17, s27, s42
-; GFX11-FAKE16-NEXT: s_pack_ll_b32_b16 s15, s28, s15
-; GFX11-FAKE16-NEXT: s_pack_ll_b32_b16 s16, s29, s41
+; GFX11-FAKE16-NEXT: s_pack_ll_b32_b16 s18, s27, s42
+; GFX11-FAKE16-NEXT: s_pack_ll_b32_b16 s16, s28, s41
+; GFX11-FAKE16-NEXT: s_pack_ll_b32_b16 s17, s29, s40
; GFX11-FAKE16-NEXT: s_and_b32 s47, vcc_lo, exec_lo
; GFX11-FAKE16-NEXT: s_cbranch_scc0 .LBB15_4
; GFX11-FAKE16-NEXT: ; %bb.1: ; %cmp.false
@@ -6693,9 +6695,9 @@ define inreg <26 x i32> @bitcast_v52i16_to_v26i32_scalar(<52 x i16> inreg %a, i3
; GFX11-FAKE16-NEXT: v_dual_mov_b32 v8, s8 :: v_dual_mov_b32 v9, s9
; GFX11-FAKE16-NEXT: v_dual_mov_b32 v10, s10 :: v_dual_mov_b32 v11, s11
; GFX11-FAKE16-NEXT: v_dual_mov_b32 v12, s12 :: v_dual_mov_b32 v13, s13
-; GFX11-FAKE16-NEXT: v_dual_mov_b32 v14, s14 :: v_dual_mov_b32 v15, s17
-; GFX11-FAKE16-NEXT: v_dual_mov_b32 v16, s15 :: v_dual_mov_b32 v17, s16
-; GFX11-FAKE16-NEXT: s_and_not1_b32 vcc_lo, exec_lo, s40
+; GFX11-FAKE16-NEXT: v_dual_mov_b32 v14, s14 :: v_dual_mov_b32 v15, s18
+; GFX11-FAKE16-NEXT: v_dual_mov_b32 v16, s16 :: v_dual_mov_b32 v17, s17
+; GFX11-FAKE16-NEXT: s_and_not1_b32 vcc_lo, exec_lo, s15
; GFX11-FAKE16-NEXT: s_cbranch_vccnz .LBB15_3
; GFX11-FAKE16-NEXT: .LBB15_2: ; %cmp.true
; GFX11-FAKE16-NEXT: v_lshl_or_b32 v18, v39, 16, v55
@@ -6721,9 +6723,9 @@ define inreg <26 x i32> @bitcast_v52i16_to_v26i32_scalar(<52 x i16> inreg %a, i3
; GFX11-FAKE16-NEXT: v_pk_add_u16 v12, s12, 3 op_sel_hi:[1,0]
; GFX11-FAKE16-NEXT: v_pk_add_u16 v13, s13, 3 op_sel_hi:[1,0]
; GFX11-FAKE16-NEXT: v_pk_add_u16 v14, s14, 3 op_sel_hi:[1,0]
-; GFX11-FAKE16-NEXT: v_pk_add_u16 v15, s17, 3 op_sel_hi:[1,0]
-; GFX11-FAKE16-NEXT: v_pk_add_u16 v16, s15, 3 op_sel_hi:[1,0]
-; GFX11-FAKE16-NEXT: v_pk_add_u16 v17, s16, 3 op_sel_hi:[1,0]
+; GFX11-FAKE16-NEXT: v_pk_add_u16 v15, s18, 3 op_sel_hi:[1,0]
+; GFX11-FAKE16-NEXT: v_pk_add_u16 v16, s16, 3 op_sel_hi:[1,0]
+; GFX11-FAKE16-NEXT: v_pk_add_u16 v17, s17, 3 op_sel_hi:[1,0]
; GFX11-FAKE16-NEXT: v_pk_add_u16 v18, v18, 3 op_sel_hi:[1,0]
; GFX11-FAKE16-NEXT: v_pk_add_u16 v19, v19, 3 op_sel_hi:[1,0]
; GFX11-FAKE16-NEXT: v_pk_add_u16 v20, v20, 3 op_sel_hi:[1,0]
@@ -8756,7 +8758,7 @@ define inreg <52 x half> @bitcast_v26i32_to_v52f16_scalar(<26 x i32> inreg %a, i
; GFX11-TRUE16-NEXT: v_readfirstlane_b32 s5, v5
; GFX11-TRUE16-NEXT: v_readfirstlane_b32 s6, v6
; GFX11-TRUE16-NEXT: v_readfirstlane_b32 s4, v7
-; GFX11-TRUE16-NEXT: s_mov_b32 s78, 0
+; GFX11-TRUE16-NEXT: s_mov_b32 s74, 0
; GFX11-TRUE16-NEXT: s_and_b32 s12, vcc_lo, exec_lo
; GFX11-TRUE16-NEXT: s_cbranch_scc0 .LBB17_4
; GFX11-TRUE16-NEXT: ; %bb.1: ; %cmp.false
@@ -8782,11 +8784,11 @@ define inreg <52 x half> @bitcast_v26i32_to_v52f16_scalar(<26 x i32> inreg %a, i
; GFX11-TRUE16-NEXT: s_lshr_b32 s63, s18, 16
; GFX11-TRUE16-NEXT: s_lshr_b32 s72, s17, 16
; GFX11-TRUE16-NEXT: s_lshr_b32 s73, s16, 16
-; GFX11-TRUE16-NEXT: s_lshr_b32 s74, s3, 16
-; GFX11-TRUE16-NEXT: s_lshr_b32 s75, s2, 16
-; GFX11-TRUE16-NEXT: s_lshr_b32 s76, s1, 16
-; GFX11-TRUE16-NEXT: s_lshr_b32 s77, s0, 16
-; GFX11-TRUE16-NEXT: s_and_not1_b32 vcc_lo, exec_lo, s78
+; GFX11-TRUE16-NEXT: s_lshr_b32 s75, s3, 16
+; GFX11-TRUE16-NEXT: s_lshr_b32 s76, s2, 16
+; GFX11-TRUE16-NEXT: s_lshr_b32 s77, s1, 16
+; GFX11-TRUE16-NEXT: s_lshr_b32 s78, s0, 16
+; GFX11-TRUE16-NEXT: s_and_not1_b32 vcc_lo, exec_lo, s74
; GFX11-TRUE16-NEXT: s_cbranch_vccnz .LBB17_3
; GFX11-TRUE16-NEXT: .LBB17_2: ; %cmp.true
; GFX11-TRUE16-NEXT: s_add_i32 s4, s4, 3
@@ -8837,16 +8839,18 @@ define inreg <52 x half> @bitcast_v26i32_to_v52f16_scalar(<26 x i32> inreg %a, i
; GFX11-TRUE16-NEXT: s_lshr_b32 s63, s18, 16
; GFX11-TRUE16-NEXT: s_lshr_b32 s72, s17, 16
; GFX11-TRUE16-NEXT: s_lshr_b32 s73, s16, 16
-; GFX11-TRUE16-NEXT: s_lshr_b32 s74, s3, 16
-; GFX11-TRUE16-NEXT: s_lshr_b32 s75, s2, 16
-; GFX11-TRUE16-NEXT: s_lshr_b32 s76, s1, 16
-; GFX11-TRUE16-NEXT: s_lshr_b32 s77, s0, 16
+; GFX11-TRUE16-NEXT: s_lshr_b32 s75, s3, 16
+; GFX11-TRUE16-NEXT: s_lshr_b32 s76, s2, 16
+; GFX11-TRUE16-NEXT: s_lshr_b32 s77, s1, 16
+; GFX11-TRUE16-NEXT: s_lshr_b32 s78, s0, 16
; GFX11-TRUE16-NEXT: .LBB17_3: ; %end
; GFX11-TRUE16-NEXT: s_delay_alu instid0(SALU_CYCLE_1)
-; GFX11-TRUE16-NEXT: s_pack_ll_b32_b16 s0, s0, s77
-; GFX11-TRUE16-NEXT: s_pack_ll_b32_b16 s1, s1, s76
-; GFX11-TRUE16-NEXT: s_pack_ll_b32_b16 s2, s2, s75
-; GFX11-TRUE16-NEXT: s_pack_ll_b32_b16 s3, s3, s74
+; GFX11-TRUE16-NEXT: s_mov_b32 s74, s78
+; GFX11-TRUE16-NEXT: s_pack_ll_b32_b16 s1, s1, s77
+; GFX11-TRUE16-NEXT: s_pack_ll_b32_b16 s0, s0, s74
+; GFX11-TRUE16-NEXT: s_mov_b32 s74, s76
+; GFX11-TRUE16-NEXT: s_pack_ll_b32_b16 s3, s3, s75
+; GFX11-TRUE16-NEXT: s_pack_ll_b32_b16 s2, s2, s74
; GFX11-TRUE16-NEXT: s_pack_ll_b32_b16 s16, s16, s73
; GFX11-TRUE16-NEXT: s_pack_ll_b32_b16 s17, s17, s72
; GFX11-TRUE16-NEXT: s_pack_ll_b32_b16 s18, s18, s63
@@ -8884,10 +8888,10 @@ define inreg <52 x half> @bitcast_v26i32_to_v52f16_scalar(<26 x i32> inreg %a, i
; GFX11-TRUE16-NEXT: v_dual_mov_b32 v24, s6 :: v_dual_mov_b32 v25, s4
; GFX11-TRUE16-NEXT: s_setpc_b64 s[30:31]
; GFX11-TRUE16-NEXT: .LBB17_4:
+; GFX11-TRUE16-NEXT: ; implicit-def: $sgpr78_lo16
; GFX11-TRUE16-NEXT: ; implicit-def: $sgpr77_lo16
; GFX11-TRUE16-NEXT: ; implicit-def: $sgpr76_lo16
; GFX11-TRUE16-NEXT: ; implicit-def: $sgpr75_lo16
-; GFX11-TRUE16-NEXT: ; implicit-def: $sgpr74_lo16
; GFX11-TRUE16-NEXT: ; implicit-def: $sgpr73_lo16
; GFX11-TRUE16-NEXT: ; implicit-def: $sgpr72_lo16
; GFX11-TRUE16-NEXT: ; implicit-def: $sgpr63_lo16
@@ -10464,15 +10468,15 @@ define inreg <26 x i32> @bitcast_v52f16_to_v26i32_scalar(<52 x half> inreg %a, i
; SI-NEXT: buffer_store_dword v0, off, s[0:3], s32 offset:124 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_cvt_f16_f32_e32 v0, v28
-; SI-NEXT: v_cvt_f16_f32_e32 v42, v5
-; SI-NEXT: v_cvt_f16_f32_e32 v55, v4
+; SI-NEXT: v_cvt_f16_f32_e32 v55, v5
+; SI-NEXT: v_cvt_f16_f32_e32 v42, v4
; SI-NEXT: v_cvt_f16_f32_e32 v56, v7
; SI-NEXT: buffer_store_dword v0, off, s[0:3], s32 offset:120 ; 4-byte Folded Spill
; SI-NEXT: v_cvt_f16_f32_e32 v54, v6
-; SI-NEXT: v_cvt_f16_f32_e32 v43, v9
-; SI-NEXT: v_cvt_f16_f32_e32 v47, v8
-; SI-NEXT: v_cvt_f16_f32_e32 v57, v11
-; SI-NEXT: v_cvt_f16_f32_e32 v60, v10
+; SI-NEXT: v_cvt_f16_f32_e32 v47, v9
+; SI-NEXT: v_cvt_f16_f32_e32 v57, v8
+; SI-NEXT: v_cvt_f16_f32_e32 v60, v11
+; SI-NEXT: v_cvt_f16_f32_e32 v53, v10
; SI-NEXT: v_cvt_f16_f32_e32 v13, v13
; SI-NEXT: v_cvt_f16_f32_e32 v52, v12
; SI-NEXT: v_cvt_f16_f32_e32 v15, v15
@@ -10490,13 +10494,13 @@ define inreg <26 x i32> @bitcast_v52f16_to_v26i32_scalar(<52 x half> inreg %a, i
; SI-NEXT: v_cvt_f16_f32_e32 v27, v26
; SI-NEXT: v_cvt_f16_f32_e32 v26, v29
; SI-NEXT: v_cvt_f16_f32_e32 v30, v30
-; SI-NEXT: v_cvt_f16_f32_e32 v53, s17
-; SI-NEXT: v_cvt_f16_f32_e32 v11, s16
+; SI-NEXT: v_cvt_f16_f32_e32 v43, s17
+; SI-NEXT: v_cvt_f16_f32_e32 v2, s16
; SI-NEXT: v_cvt_f16_f32_e32 v1, s19
-; SI-NEXT: v_cvt_f16_f32_e32 v2, s18
+; SI-NEXT: v_cvt_f16_f32_e32 v3, s18
; SI-NEXT: v_cvt_f16_f32_e32 v12, s21
; SI-NEXT: v_cvt_f16_f32_e32 v14, s20
-; SI-NEXT: v_cvt_f16_f32_e32 v3, s23
+; SI-NEXT: v_cvt_f16_f32_e32 v11, s23
; SI-NEXT: v_cvt_f16_f32_e32 v10, s22
; SI-NEXT: v_cvt_f16_f32_e32 v4, s25
; SI-NEXT: v_cvt_f16_f32_e32 v9, s24
@@ -10538,9 +10542,9 @@ define inreg <26 x i32> @bitcast_v52f16_to_v26i32_scalar(<52 x half> inreg %a, i
; SI-NEXT: buffer_store_dword v4, off, s[0:3], s32 offset:160 ; 4-byte Folded Spill
; SI-NEXT: buffer_store_dword v10, off, s[0:3], s32 offset:164 ; 4-byte Folded Spill
; SI-NEXT: buffer_store_dword v2, off, s[0:3], s32 offset:168 ; 4-byte Folded Spill
-; SI-NEXT: buffer_store_dword v3, off, s[0:3], s32 offset:172 ; 4-byte Folded Spill
-; SI-NEXT: buffer_store_dword v11, off, s[0:3], s32 offset:176 ; 4-byte Folded Spill
-; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:180 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:172 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v3, off, s[0:3], s32 offset:176 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v11, off, s[0:3], s32 offset:180 ; 4-byte Folded Spill
; SI-NEXT: buffer_store_dword v12, off, s[0:3], s32 offset:184 ; 4-byte Folded Spill
; SI-NEXT: buffer_store_dword v14, off, s[0:3], s32 offset:188 ; 4-byte Folded Spill
; SI-NEXT: s_cbranch_scc0 .LBB19_4
@@ -10551,17 +10555,18 @@ define inreg <26 x i32> @bitcast_v52f16_to_v26i32_scalar(<52 x half> inreg %a, i
; SI-NEXT: buffer_load_dword v25, off, s[0:3], s32 offset:104 ; 4-byte Folded Reload
; SI-NEXT: buffer_load_dword v22, off, s[0:3], s32 offset:120 ; 4-byte Folded Reload
; SI-NEXT: buffer_load_dword v38, off, s[0:3], s32 offset:96 ; 4-byte Folded Reload
-; SI-NEXT: s_waitcnt expcnt(2)
+; SI-NEXT: v_lshlrev_b32_e32 v0, 16, v43
+; SI-NEXT: s_waitcnt expcnt(4)
; SI-NEXT: v_lshlrev_b32_e32 v1, 16, v1
-; SI-NEXT: v_lshlrev_b32_e32 v0, 16, v53
-; SI-NEXT: v_or_b32_e32 v1, v2, v1
+; SI-NEXT: v_or_b32_e32 v0, v2, v0
+; SI-NEXT: v_or_b32_e32 v1, v3, v1
; SI-NEXT: v_lshlrev_b32_e32 v2, 16, v12
-; SI-NEXT: v_lshlrev_b32_e32 v3, 16, v3
+; SI-NEXT: s_waitcnt expcnt(3)
+; SI-NEXT: v_lshlrev_b32_e32 v3, 16, v11
; SI-NEXT: v_lshlrev_b32_e32 v4, 16, v4
; SI-NEXT: v_lshlrev_b32_e32 v5, 16, v5
; SI-NEXT: v_lshlrev_b32_e32 v6, 16, v6
; SI-NEXT: v_lshlrev_b32_e32 v21, 16, v26
-; SI-NEXT: v_or_b32_e32 v0, v11, v0
; SI-NEXT: v_or_b32_e32 v2, v14, v2
; SI-NEXT: v_or_b32_e32 v3, v10, v3
; SI-NEXT: v_or_b32_e32 v4, v9, v4
@@ -10569,11 +10574,12 @@ define inreg <26 x i32> @bitcast_v52f16_to_v26i32_scalar(<52 x half> inreg %a, i
; SI-NEXT: v_or_b32_e32 v6, v7, v6
; SI-NEXT: v_lshlrev_b32_e32 v7, 16, v46
; SI-NEXT: v_lshlrev_b32_e32 v8, 16, v41
-; SI-NEXT: v_lshlrev_b32_e32 v9, 16, v42
+; SI-NEXT: v_lshlrev_b32_e32 v9, 16, v55
; SI-NEXT: v_lshlrev_b32_e32 v10, 16, v56
-; SI-NEXT: v_lshlrev_b32_e32 v11, 16, v43
+; SI-NEXT: s_waitcnt expcnt(2)
+; SI-NEXT: v_lshlrev_b32_e32 v11, 16, v47
; SI-NEXT: s_waitcnt expcnt(1)
-; SI-NEXT: v_lshlrev_b32_e32 v12, 16, v57
+; SI-NEXT: v_lshlrev_b32_e32 v12, 16, v60
; SI-NEXT: v_lshlrev_b32_e32 v13, 16, v13
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_lshlrev_b32_e32 v14, 16, v15
@@ -10585,10 +10591,10 @@ define inreg <26 x i32> @bitcast_v52f16_to_v26i32_scalar(<52 x half> inreg %a, i
; SI-NEXT: v_mov_b32_e32 v51, v46
; SI-NEXT: v_or_b32_e32 v7, v45, v7
; SI-NEXT: v_or_b32_e32 v8, v40, v8
-; SI-NEXT: v_or_b32_e32 v9, v55, v9
+; SI-NEXT: v_or_b32_e32 v9, v42, v9
; SI-NEXT: v_or_b32_e32 v10, v54, v10
-; SI-NEXT: v_or_b32_e32 v11, v47, v11
-; SI-NEXT: v_or_b32_e32 v12, v60, v12
+; SI-NEXT: v_or_b32_e32 v11, v57, v11
+; SI-NEXT: v_or_b32_e32 v12, v53, v12
; SI-NEXT: v_or_b32_e32 v13, v52, v13
; SI-NEXT: v_or_b32_e32 v14, v63, v14
; SI-NEXT: v_or_b32_e32 v15, v61, v15
@@ -10616,14 +10622,16 @@ define inreg <26 x i32> @bitcast_v52f16_to_v26i32_scalar(<52 x half> inreg %a, i
; SI-NEXT: v_or_b32_e32 v25, v38, v25
; SI-NEXT: s_cbranch_execnz .LBB19_3
; SI-NEXT: .LBB19_2: ; %cmp.true
-; SI-NEXT: s_waitcnt expcnt(2)
-; SI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:176 ; 4-byte Folded Reload
-; SI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:180 ; 4-byte Folded Reload
-; SI-NEXT: buffer_load_dword v3, off, s[0:3], s32 offset:168 ; 4-byte Folded Reload
+; SI-NEXT: s_waitcnt expcnt(4)
+; SI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:168 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:172 ; 4-byte Folded Reload
+; SI-NEXT: s_waitcnt expcnt(3)
+; SI-NEXT: buffer_load_dword v3, off, s[0:3], s32 offset:176 ; 4-byte Folded Reload
; SI-NEXT: buffer_load_dword v4, off, s[0:3], s32 offset:164 ; 4-byte Folded Reload
-; SI-NEXT: v_cvt_f32_f16_e32 v0, v53
+; SI-NEXT: v_cvt_f32_f16_e32 v0, v43
; SI-NEXT: v_cvt_f32_f16_e32 v9, v40
-; SI-NEXT: v_cvt_f32_f16_e32 v10, v55
+; SI-NEXT: v_cvt_f32_f16_e32 v10, v42
+; SI-NEXT: s_waitcnt expcnt(2)
; SI-NEXT: v_cvt_f32_f16_e32 v11, v54
; SI-NEXT: v_add_f32_e32 v0, 0x38000000, v0
; SI-NEXT: v_cvt_f16_f32_e32 v0, v0
@@ -10635,8 +10643,8 @@ define inreg <26 x i32> @bitcast_v52f16_to_v26i32_scalar(<52 x half> inreg %a, i
; SI-NEXT: v_add_f32_e32 v11, 0x38000000, v11
; SI-NEXT: v_cvt_f16_f32_e32 v11, v11
; SI-NEXT: s_waitcnt expcnt(1)
-; SI-NEXT: v_cvt_f32_f16_e32 v12, v47
-; SI-NEXT: v_cvt_f32_f16_e32 v13, v60
+; SI-NEXT: v_cvt_f32_f16_e32 v12, v57
+; SI-NEXT: v_cvt_f32_f16_e32 v13, v53
; SI-NEXT: v_cvt_f32_f16_e32 v15, v52
; SI-NEXT: v_cvt_f32_f16_e32 v16, v63
; SI-NEXT: v_add_f32_e32 v12, 0x38000000, v12
@@ -10736,7 +10744,7 @@ define inreg <26 x i32> @bitcast_v52f16_to_v26i32_scalar(<52 x half> inreg %a, i
; SI-NEXT: v_cvt_f16_f32_e32 v3, v3
; SI-NEXT: v_lshlrev_b32_e32 v2, 16, v2
; SI-NEXT: v_or_b32_e32 v2, v3, v2
-; SI-NEXT: buffer_load_dword v3, off, s[0:3], s32 offset:172 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v3, off, s[0:3], s32 offset:180 ; 4-byte Folded Reload
; SI-NEXT: s_waitcnt vmcnt(0)
; SI-NEXT: v_cvt_f32_f16_e32 v3, v3
; SI-NEXT: v_add_f32_e32 v3, 0x38000000, v3
@@ -10774,7 +10782,7 @@ define inreg <26 x i32> @bitcast_v52f16_to_v26i32_scalar(<52 x half> inreg %a, i
; SI-NEXT: v_cvt_f16_f32_e32 v8, v8
; SI-NEXT: v_lshlrev_b32_e32 v8, 16, v8
; SI-NEXT: v_or_b32_e32 v8, v9, v8
-; SI-NEXT: v_cvt_f32_f16_e32 v9, v42
+; SI-NEXT: v_cvt_f32_f16_e32 v9, v55
; SI-NEXT: v_add_f32_e32 v9, 0x38000000, v9
; SI-NEXT: v_cvt_f16_f32_e32 v9, v9
; SI-NEXT: v_lshlrev_b32_e32 v9, 16, v9
@@ -10784,12 +10792,12 @@ define inreg <26 x i32> @bitcast_v52f16_to_v26i32_scalar(<52 x half> inreg %a, i
; SI-NEXT: v_cvt_f16_f32_e32 v10, v10
; SI-NEXT: v_lshlrev_b32_e32 v10, 16, v10
; SI-NEXT: v_or_b32_e32 v10, v11, v10
-; SI-NEXT: v_cvt_f32_f16_e32 v11, v43
+; SI-NEXT: v_cvt_f32_f16_e32 v11, v47
; SI-NEXT: v_add_f32_e32 v11, 0x38000000, v11
; SI-NEXT: v_cvt_f16_f32_e32 v11, v11
; SI-NEXT: v_lshlrev_b32_e32 v11, 16, v11
; SI-NEXT: v_or_b32_e32 v11, v12, v11
-; SI-NEXT: v_cvt_f32_f16_e32 v12, v57
+; SI-NEXT: v_cvt_f32_f16_e32 v12, v60
; SI-NEXT: v_add_f32_e32 v12, 0x38000000, v12
; SI-NEXT: v_cvt_f16_f32_e32 v12, v12
; SI-NEXT: v_lshlrev_b32_e32 v12, 16, v12
@@ -11532,8 +11540,8 @@ define inreg <26 x i32> @bitcast_v52f16_to_v26i32_scalar(<52 x half> inreg %a, i
; GFX11-FAKE16-NEXT: v_and_b32_e32 v50, 0xffff, v5
; GFX11-FAKE16-NEXT: v_and_b32_e32 v49, 0xffff, v6
; GFX11-FAKE16-NEXT: v_and_b32_e32 v48, 0xffff, v7
-; GFX11-FAKE16-NEXT: s_lshr_b32 s41, s29, 16
-; GFX11-FAKE16-NEXT: s_lshr_b32 s15, s28, 16
+; GFX11-FAKE16-NEXT: s_lshr_b32 s40, s29, 16
+; GFX11-FAKE16-NEXT: s_lshr_b32 s41, s28, 16
; GFX11-FAKE16-NEXT: s_lshr_b32 s42, s27, 16
; GFX11-FAKE16-NEXT: s_lshr_b32 s14, s26, 16
; GFX11-FAKE16-NEXT: s_lshr_b32 s13, s25, 16
@@ -11550,7 +11558,7 @@ define inreg <26 x i32> @bitcast_v52f16_to_v26i32_scalar(<52 x half> inreg %a, i
; GFX11-FAKE16-NEXT: s_lshr_b32 s44, s2, 16
; GFX11-FAKE16-NEXT: s_lshr_b32 s45, s1, 16
; GFX11-FAKE16-NEXT: s_lshr_b32 s46, s0, 16
-; GFX11-FAKE16-NEXT: s_mov_b32 s40, 0
+; GFX11-FAKE16-NEXT: s_mov_b32 s15, 0
; GFX11-FAKE16-NEXT: s_pack_ll_b32_b16 s0, s0, s46
; GFX11-FAKE16-NEXT: s_pack_ll_b32_b16 s1, s1, s45
; GFX11-FAKE16-NEXT: s_pack_ll_b32_b16 s2, s2, s44
@@ -11566,9 +11574,9 @@ define inreg <26 x i32> @bitcast_v52f16_to_v26i32_scalar(<52 x half> inreg %a, i
; GFX11-FAKE16-NEXT: s_pack_ll_b32_b16 s12, s24, s12
; GFX11-FAKE16-NEXT: s_pack_ll_b32_b16 s13, s25, s13
; GFX11-FAKE16-NEXT: s_pack_ll_b32_b16 s14, s26, s14
-; GFX11-FAKE16-NEXT: s_pack_ll_b32_b16 s17, s27, s42
-; GFX11-FAKE16-NEXT: s_pack_ll_b32_b16 s15, s28, s15
-; GFX11-FAKE16-NEXT: s_pack_ll_b32_b16 s16, s29, s41
+; GFX11-FAKE16-NEXT: s_pack_ll_b32_b16 s18, s27, s42
+; GFX11-FAKE16-NEXT: s_pack_ll_b32_b16 s16, s28, s41
+; GFX11-FAKE16-NEXT: s_pack_ll_b32_b16 s17, s29, s40
; GFX11-FAKE16-NEXT: s_and_b32 s47, vcc_lo, exec_lo
; GFX11-FAKE16-NEXT: s_cbranch_scc0 .LBB19_4
; GFX11-FAKE16-NEXT: ; %bb.1: ; %cmp.false
@@ -11587,9 +11595,9 @@ define inreg <26 x i32> @bitcast_v52f16_to_v26i32_scalar(<52 x half> inreg %a, i
; GFX11-FAKE16-NEXT: v_dual_mov_b32 v8, s8 :: v_dual_mov_b32 v9, s9
; GFX11-FAKE16-NEXT: v_dual_mov_b32 v10, s10 :: v_dual_mov_b32 v11, s11
; GFX11-FAKE16-NEXT: v_dual_mov_b32 v12, s12 :: v_dual_mov_b32 v13, s13
-; GFX11-FAKE16-NEXT: v_dual_mov_b32 v14, s14 :: v_dual_mov_b32 v15, s17
-; GFX11-FAKE16-NEXT: v_dual_mov_b32 v16, s15 :: v_dual_mov_b32 v17, s16
-; GFX11-FAKE16-NEXT: s_and_not1_b32 vcc_lo, exec_lo, s40
+; GFX11-FAKE16-NEXT: v_dual_mov_b32 v14, s14 :: v_dual_mov_b32 v15, s18
+; GFX11-FAKE16-NEXT: v_dual_mov_b32 v16, s16 :: v_dual_mov_b32 v17, s17
+; GFX11-FAKE16-NEXT: s_and_not1_b32 vcc_lo, exec_lo, s15
; GFX11-FAKE16-NEXT: s_cbranch_vccnz .LBB19_3
; GFX11-FAKE16-NEXT: .LBB19_2: ; %cmp.true
; GFX11-FAKE16-NEXT: v_lshl_or_b32 v18, v39, 16, v55
@@ -11615,9 +11623,9 @@ define inreg <26 x i32> @bitcast_v52f16_to_v26i32_scalar(<52 x half> inreg %a, i
; GFX11-FAKE16-NEXT: v_pk_add_f16 v12, 0x200, s12 op_sel_hi:[0,1]
; GFX11-FAKE16-NEXT: v_pk_add_f16 v13, 0x200, s13 op_sel_hi:[0,1]
; GFX11-FAKE16-NEXT: v_pk_add_f16 v14, 0x200, s14 op_sel_hi:[0,1]
-; GFX11-FAKE16-NEXT: v_pk_add_f16 v15, 0x200, s17 op_sel_hi:[0,1]
-; GFX11-FAKE16-NEXT: v_pk_add_f16 v16, 0x200, s15 op_sel_hi:[0,1]
-; GFX11-FAKE16-NEXT: v_pk_add_f16 v17, 0x200, s16 op_sel_hi:[0,1]
+; GFX11-FAKE16-NEXT: v_pk_add_f16 v15, 0x200, s18 op_sel_hi:[0,1]
+; GFX11-FAKE16-NEXT: v_pk_add_f16 v16, 0x200, s16 op_sel_hi:[0,1]
+; GFX11-FAKE16-NEXT: v_pk_add_f16 v17, 0x200, s17 op_sel_hi:[0,1]
; GFX11-FAKE16-NEXT: v_pk_add_f16 v18, 0x200, v18 op_sel_hi:[0,1]
; GFX11-FAKE16-NEXT: v_pk_add_f16 v19, 0x200, v19 op_sel_hi:[0,1]
; GFX11-FAKE16-NEXT: v_pk_add_f16 v20, 0x200, v20 op_sel_hi:[0,1]
@@ -17336,8 +17344,8 @@ define inreg <26 x float> @bitcast_v52i16_to_v26f32_scalar(<52 x i16> inreg %a,
; GFX11-FAKE16-NEXT: v_and_b32_e32 v50, 0xffff, v5
; GFX11-FAKE16-NEXT: v_and_b32_e32 v49, 0xffff, v6
; GFX11-FAKE16-NEXT: v_and_b32_e32 v48, 0xffff, v7
-; GFX11-FAKE16-NEXT: s_lshr_b32 s41, s29, 16
-; GFX11-FAKE16-NEXT: s_lshr_b32 s15, s28, 16
+; GFX11-FAKE16-NEXT: s_lshr_b32 s40, s29, 16
+; GFX11-FAKE16-NEXT: s_lshr_b32 s41, s28, 16
; GFX11-FAKE16-NEXT: s_lshr_b32 s42, s27, 16
; GFX11-FAKE16-NEXT: s_lshr_b32 s14, s26, 16
; GFX11-FAKE16-NEXT: s_lshr_b32 s13, s25, 16
@@ -17354,7 +17362,7 @@ define inreg <26 x float> @bitcast_v52i16_to_v26f32_scalar(<52 x i16> inreg %a,
; GFX11-FAKE16-NEXT: s_lshr_b32 s44, s2, 16
; GFX11-FAKE16-NEXT: s_lshr_b32 s45, s1, 16
; GFX11-FAKE16-NEXT: s_lshr_b32 s46, s0, 16
-; GFX11-FAKE16-NEXT: s_mov_b32 s40, 0
+; GFX11-FAKE16-NEXT: s_mov_b32 s15, 0
; GFX11-FAKE16-NEXT: s_pack_ll_b32_b16 s0, s0, s46
; GFX11-FAKE16-NEXT: s_pack_ll_b32_b16 s1, s1, s45
; GFX11-FAKE16-NEXT: s_pack_ll_b32_b16 s2, s2, s44
@@ -17370,9 +17378,9 @@ define inreg <26 x float> @bitcast_v52i16_to_v26f32_scalar(<52 x i16> inreg %a,
; GFX11-FAKE16-NEXT: s_pack_ll_b32_b16 s12, s24, s12
; GFX11-FAKE16-NEXT: s_pack_ll_b32_b16 s13, s25, s13
; GFX11-FAKE16-NEXT: s_pack_ll_b32_b16 s14, s26, s14
-; GFX11-FAKE16-NEXT: s_pack_ll_b32_b16 s17, s27, s42
-; GFX11-FAKE16-NEXT: s_pack_ll_b32_b16 s15, s28, s15
-; GFX11-FAKE16-NEXT: s_pack_ll_b32_b16 s16, s29, s41
+; GFX11-FAKE16-NEXT: s_pack_ll_b32_b16 s18, s27, s42
+; GFX11-FAKE16-NEXT: s_pack_ll_b32_b16 s16, s28, s41
+; GFX11-FAKE16-NEXT: s_pack_ll_b32_b16 s17, s29, s40
; GFX11-FAKE16-NEXT: s_and_b32 s47, vcc_lo, exec_lo
; GFX11-FAKE16-NEXT: s_cbranch_scc0 .LBB31_4
; GFX11-FAKE16-NEXT: ; %bb.1: ; %cmp.false
@@ -17391,9 +17399,9 @@ define inreg <26 x float> @bitcast_v52i16_to_v26f32_scalar(<52 x i16> inreg %a,
; GFX11-FAKE16-NEXT: v_dual_mov_b32 v8, s8 :: v_dual_mov_b32 v9, s9
; GFX11-FAKE16-NEXT: v_dual_mov_b32 v10, s10 :: v_dual_mov_b32 v11, s11
; GFX11-FAKE16-NEXT: v_dual_mov_b32 v12, s12 :: v_dual_mov_b32 v13, s13
-; GFX11-FAKE16-NEXT: v_dual_mov_b32 v14, s14 :: v_dual_mov_b32 v15, s17
-; GFX11-FAKE16-NEXT: v_dual_mov_b32 v16, s15 :: v_dual_mov_b32 v17, s16
-; GFX11-FAKE16-NEXT: s_and_not1_b32 vcc_lo, exec_lo, s40
+; GFX11-FAKE16-NEXT: v_dual_mov_b32 v14, s14 :: v_dual_mov_b32 v15, s18
+; GFX11-FAKE16-NEXT: v_dual_mov_b32 v16, s16 :: v_dual_mov_b32 v17, s17
+; GFX11-FAKE16-NEXT: s_and_not1_b32 vcc_lo, exec_lo, s15
; GFX11-FAKE16-NEXT: s_cbranch_vccnz .LBB31_3
; GFX11-FAKE16-NEXT: .LBB31_2: ; %cmp.true
; GFX11-FAKE16-NEXT: v_lshl_or_b32 v18, v39, 16, v55
@@ -17419,9 +17427,9 @@ define inreg <26 x float> @bitcast_v52i16_to_v26f32_scalar(<52 x i16> inreg %a,
; GFX11-FAKE16-NEXT: v_pk_add_u16 v12, s12, 3 op_sel_hi:[1,0]
; GFX11-FAKE16-NEXT: v_pk_add_u16 v13, s13, 3 op_sel_hi:[1,0]
; GFX11-FAKE16-NEXT: v_pk_add_u16 v14, s14, 3 op_sel_hi:[1,0]
-; GFX11-FAKE16-NEXT: v_pk_add_u16 v15, s17, 3 op_sel_hi:[1,0]
-; GFX11-FAKE16-NEXT: v_pk_add_u16 v16, s15, 3 op_sel_hi:[1,0]
-; GFX11-FAKE16-NEXT: v_pk_add_u16 v17, s16, 3 op_sel_hi:[1,0]
+; GFX11-FAKE16-NEXT: v_pk_add_u16 v15, s18, 3 op_sel_hi:[1,0]
+; GFX11-FAKE16-NEXT: v_pk_add_u16 v16, s16, 3 op_sel_hi:[1,0]
+; GFX11-FAKE16-NEXT: v_pk_add_u16 v17, s17, 3 op_sel_hi:[1,0]
; GFX11-FAKE16-NEXT: v_pk_add_u16 v18, v18, 3 op_sel_hi:[1,0]
; GFX11-FAKE16-NEXT: v_pk_add_u16 v19, v19, 3 op_sel_hi:[1,0]
; GFX11-FAKE16-NEXT: v_pk_add_u16 v20, v20, 3 op_sel_hi:[1,0]
@@ -18588,9 +18596,9 @@ define inreg <52 x half> @bitcast_v26f32_to_v52f16_scalar(<26 x float> inreg %a,
; SI-NEXT: s_lshr_b32 s4, s10, 16
; SI-NEXT: v_cvt_f32_f16_e32 v7, s4
; SI-NEXT: s_lshr_b32 s4, s11, 16
-; SI-NEXT: s_waitcnt expcnt(3)
-; SI-NEXT: v_cvt_f32_f16_e32 v60, s4
+; SI-NEXT: v_cvt_f32_f16_e32 v9, s4
; SI-NEXT: s_lshr_b32 s4, s12, 16
+; SI-NEXT: s_waitcnt expcnt(4)
; SI-NEXT: v_cvt_f32_f16_e32 v59, s4
; SI-NEXT: s_lshr_b32 s4, s13, 16
; SI-NEXT: v_cvt_f32_f16_e32 v14, s4
@@ -18634,8 +18642,8 @@ define inreg <52 x half> @bitcast_v26f32_to_v52f16_scalar(<26 x float> inreg %a,
; SI-NEXT: v_cvt_f32_f16_e32 v32, s6
; SI-NEXT: v_cvt_f32_f16_e32 v34, s7
; SI-NEXT: v_cvt_f32_f16_e32 v36, s8
-; SI-NEXT: v_cvt_f32_f16_e32 v38, s10
-; SI-NEXT: v_cvt_f32_f16_e32 v48, s11
+; SI-NEXT: v_cvt_f32_f16_e32 v46, s10
+; SI-NEXT: v_cvt_f32_f16_e32 v38, s11
; SI-NEXT: v_cvt_f32_f16_e32 v15, s12
; SI-NEXT: v_cvt_f32_f16_e32 v17, s13
; SI-NEXT: v_cvt_f32_f16_e32 v19, s14
@@ -18650,7 +18658,7 @@ define inreg <52 x half> @bitcast_v26f32_to_v52f16_scalar(<26 x float> inreg %a,
; SI-NEXT: v_cvt_f32_f16_e32 v12, s24
; SI-NEXT: v_cvt_f32_f16_e32 v13, s23
; SI-NEXT: v_cvt_f32_f16_e32 v11, s22
-; SI-NEXT: v_cvt_f32_f16_e32 v9, s21
+; SI-NEXT: v_cvt_f32_f16_e32 v50, s21
; SI-NEXT: v_cvt_f32_f16_e32 v52, s20
; SI-NEXT: v_cvt_f32_f16_e32 v53, s19
; SI-NEXT: v_cvt_f32_f16_e32 v55, s18
@@ -18659,7 +18667,16 @@ define inreg <52 x half> @bitcast_v26f32_to_v52f16_scalar(<26 x float> inreg %a,
; SI-NEXT: s_cbranch_execnz .LBB33_3
; SI-NEXT: .LBB33_2: ; %cmp.true
; SI-NEXT: v_add_f32_e64 v1, s16, 1.0
+; SI-NEXT: v_lshrrev_b32_e32 v2, 16, v1
+; SI-NEXT: v_add_f32_e64 v6, s6, 1.0
+; SI-NEXT: buffer_store_dword v2, off, s[0:3], s32 offset:64 ; 4-byte Folded Spill
+; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_add_f32_e64 v2, s17, 1.0
+; SI-NEXT: v_lshrrev_b32_e32 v4, 16, v6
+; SI-NEXT: v_lshrrev_b32_e32 v44, 16, v2
+; SI-NEXT: v_cvt_f32_f16_e32 v41, v2
+; SI-NEXT: v_cvt_f32_f16_e32 v2, v4
+; SI-NEXT: buffer_load_dword v4, off, s[0:3], s32 offset:64 ; 4-byte Folded Reload
; SI-NEXT: v_add_f32_e64 v3, s18, 1.0
; SI-NEXT: v_add_f32_e64 v5, s19, 1.0
; SI-NEXT: v_add_f32_e64 v7, s20, 1.0
@@ -18682,47 +18699,36 @@ define inreg <52 x half> @bitcast_v26f32_to_v52f16_scalar(<26 x float> inreg %a,
; SI-NEXT: v_add_f32_e64 v12, s10, 1.0
; SI-NEXT: v_add_f32_e64 v10, s8, 1.0
; SI-NEXT: v_add_f32_e64 v8, s7, 1.0
-; SI-NEXT: v_add_f32_e64 v6, s6, 1.0
; SI-NEXT: v_add_f32_e64 v29, s9, 1.0
-; SI-NEXT: v_lshrrev_b32_e32 v44, 16, v1
-; SI-NEXT: v_lshrrev_b32_e32 v42, 16, v2
-; SI-NEXT: v_lshrrev_b32_e32 v40, 16, v3
-; SI-NEXT: v_lshrrev_b32_e32 v54, 16, v5
-; SI-NEXT: v_lshrrev_b32_e32 v51, 16, v7
-; SI-NEXT: v_lshrrev_b32_e32 v49, 16, v9
-; SI-NEXT: v_lshrrev_b32_e32 v39, 16, v11
-; SI-NEXT: v_lshrrev_b32_e32 v37, 16, v14
-; SI-NEXT: v_lshrrev_b32_e32 v35, 16, v16
-; SI-NEXT: v_lshrrev_b32_e32 v33, 16, v18
-; SI-NEXT: v_lshrrev_b32_e32 v31, 16, v20
-; SI-NEXT: v_lshrrev_b32_e32 v28, 16, v22
-; SI-NEXT: v_lshrrev_b32_e32 v26, 16, v24
-; SI-NEXT: v_lshrrev_b32_e32 v50, 16, v27
-; SI-NEXT: v_lshrrev_b32_e32 v46, 16, v25
+; SI-NEXT: v_lshrrev_b32_e32 v42, 16, v3
+; SI-NEXT: v_lshrrev_b32_e32 v40, 16, v5
+; SI-NEXT: v_lshrrev_b32_e32 v54, 16, v7
+; SI-NEXT: v_lshrrev_b32_e32 v51, 16, v9
+; SI-NEXT: v_lshrrev_b32_e32 v49, 16, v11
+; SI-NEXT: v_lshrrev_b32_e32 v39, 16, v14
+; SI-NEXT: v_lshrrev_b32_e32 v37, 16, v16
+; SI-NEXT: v_lshrrev_b32_e32 v35, 16, v18
+; SI-NEXT: v_lshrrev_b32_e32 v33, 16, v20
+; SI-NEXT: v_lshrrev_b32_e32 v31, 16, v22
+; SI-NEXT: v_lshrrev_b32_e32 v28, 16, v24
+; SI-NEXT: v_lshrrev_b32_e32 v26, 16, v27
+; SI-NEXT: v_lshrrev_b32_e32 v48, 16, v25
; SI-NEXT: v_lshrrev_b32_e32 v47, 16, v23
; SI-NEXT: v_lshrrev_b32_e32 v56, 16, v21
-; SI-NEXT: s_waitcnt expcnt(6)
; SI-NEXT: v_lshrrev_b32_e32 v57, 16, v19
-; SI-NEXT: s_waitcnt expcnt(5)
; SI-NEXT: v_lshrrev_b32_e32 v58, 16, v17
-; SI-NEXT: s_waitcnt expcnt(4)
; SI-NEXT: v_lshrrev_b32_e32 v59, 16, v15
-; SI-NEXT: s_waitcnt expcnt(3)
; SI-NEXT: v_lshrrev_b32_e32 v60, 16, v13
-; SI-NEXT: s_waitcnt expcnt(2)
; SI-NEXT: v_lshrrev_b32_e32 v61, 16, v12
-; SI-NEXT: s_waitcnt expcnt(1)
; SI-NEXT: v_lshrrev_b32_e32 v62, 16, v10
-; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_lshrrev_b32_e32 v63, 16, v8
-; SI-NEXT: v_lshrrev_b32_e32 v4, 16, v6
; SI-NEXT: v_lshrrev_b32_e32 v30, 16, v29
; SI-NEXT: v_cvt_f32_f16_e32 v45, v29
; SI-NEXT: v_cvt_f32_f16_e32 v32, v6
; SI-NEXT: v_cvt_f32_f16_e32 v34, v8
; SI-NEXT: v_cvt_f32_f16_e32 v36, v10
-; SI-NEXT: v_cvt_f32_f16_e32 v38, v12
-; SI-NEXT: v_cvt_f32_f16_e32 v48, v13
+; SI-NEXT: v_cvt_f32_f16_e32 v46, v12
+; SI-NEXT: v_cvt_f32_f16_e32 v38, v13
; SI-NEXT: v_cvt_f32_f16_e32 v15, v15
; SI-NEXT: v_cvt_f32_f16_e32 v17, v17
; SI-NEXT: v_cvt_f32_f16_e32 v19, v19
@@ -18737,38 +18743,37 @@ define inreg <52 x half> @bitcast_v26f32_to_v52f16_scalar(<26 x float> inreg %a,
; SI-NEXT: v_cvt_f32_f16_e32 v12, v16
; SI-NEXT: v_cvt_f32_f16_e32 v13, v14
; SI-NEXT: v_cvt_f32_f16_e32 v11, v11
-; SI-NEXT: v_cvt_f32_f16_e32 v9, v9
+; SI-NEXT: v_cvt_f32_f16_e32 v50, v9
; SI-NEXT: v_cvt_f32_f16_e32 v52, v7
; SI-NEXT: v_cvt_f32_f16_e32 v53, v5
; SI-NEXT: v_cvt_f32_f16_e32 v55, v3
-; SI-NEXT: v_cvt_f32_f16_e32 v41, v2
; SI-NEXT: v_cvt_f32_f16_e32 v43, v1
; SI-NEXT: v_cvt_f32_f16_e32 v1, v30
-; SI-NEXT: v_cvt_f32_f16_e32 v2, v4
; SI-NEXT: v_cvt_f32_f16_e32 v3, v63
; SI-NEXT: v_cvt_f32_f16_e32 v5, v62
; SI-NEXT: v_cvt_f32_f16_e32 v7, v61
-; SI-NEXT: v_cvt_f32_f16_e32 v60, v60
+; SI-NEXT: v_cvt_f32_f16_e32 v9, v60
; SI-NEXT: v_cvt_f32_f16_e32 v59, v59
; SI-NEXT: v_cvt_f32_f16_e32 v14, v58
; SI-NEXT: v_cvt_f32_f16_e32 v16, v57
; SI-NEXT: v_cvt_f32_f16_e32 v18, v56
; SI-NEXT: v_cvt_f32_f16_e32 v20, v47
-; SI-NEXT: v_cvt_f32_f16_e32 v22, v46
-; SI-NEXT: v_cvt_f32_f16_e32 v24, v50
-; SI-NEXT: v_cvt_f32_f16_e32 v26, v26
-; SI-NEXT: v_cvt_f32_f16_e32 v28, v28
-; SI-NEXT: v_cvt_f32_f16_e32 v31, v31
-; SI-NEXT: v_cvt_f32_f16_e32 v33, v33
-; SI-NEXT: v_cvt_f32_f16_e32 v35, v35
-; SI-NEXT: v_cvt_f32_f16_e32 v37, v37
-; SI-NEXT: v_cvt_f32_f16_e32 v39, v39
-; SI-NEXT: v_cvt_f32_f16_e32 v49, v49
-; SI-NEXT: v_cvt_f32_f16_e32 v51, v51
-; SI-NEXT: v_cvt_f32_f16_e32 v54, v54
-; SI-NEXT: v_cvt_f32_f16_e32 v40, v40
-; SI-NEXT: v_cvt_f32_f16_e32 v42, v42
-; SI-NEXT: v_cvt_f32_f16_e32 v44, v44
+; SI-NEXT: v_cvt_f32_f16_e32 v22, v48
+; SI-NEXT: v_cvt_f32_f16_e32 v24, v26
+; SI-NEXT: v_cvt_f32_f16_e32 v26, v28
+; SI-NEXT: v_cvt_f32_f16_e32 v28, v31
+; SI-NEXT: v_cvt_f32_f16_e32 v31, v33
+; SI-NEXT: v_cvt_f32_f16_e32 v33, v35
+; SI-NEXT: v_cvt_f32_f16_e32 v35, v37
+; SI-NEXT: v_cvt_f32_f16_e32 v37, v39
+; SI-NEXT: v_cvt_f32_f16_e32 v39, v49
+; SI-NEXT: v_cvt_f32_f16_e32 v49, v51
+; SI-NEXT: v_cvt_f32_f16_e32 v51, v54
+; SI-NEXT: v_cvt_f32_f16_e32 v54, v40
+; SI-NEXT: v_cvt_f32_f16_e32 v40, v42
+; SI-NEXT: v_cvt_f32_f16_e32 v42, v44
+; SI-NEXT: s_waitcnt vmcnt(0)
+; SI-NEXT: v_cvt_f32_f16_e32 v44, v4
; SI-NEXT: .LBB33_3: ; %end
; SI-NEXT: v_cvt_f16_f32_e32 v4, v44
; SI-NEXT: v_cvt_f16_f32_e32 v30, v43
@@ -18805,7 +18810,7 @@ define inreg <52 x half> @bitcast_v26f32_to_v52f16_scalar(<26 x float> inreg %a,
; SI-NEXT: buffer_store_dword v4, v51, s[0:3], 0 offen
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_cvt_f16_f32_e32 v4, v49
-; SI-NEXT: v_cvt_f16_f32_e32 v30, v9
+; SI-NEXT: v_cvt_f16_f32_e32 v30, v50
; SI-NEXT: v_add_i32_e32 v49, vcc, 20, v0
; SI-NEXT: v_lshlrev_b32_e32 v4, 16, v4
; SI-NEXT: v_or_b32_e32 v4, v30, v4
@@ -18909,15 +18914,15 @@ define inreg <52 x half> @bitcast_v26f32_to_v52f16_scalar(<26 x float> inreg %a,
; SI-NEXT: v_or_b32_e32 v4, v6, v4
; SI-NEXT: buffer_store_dword v4, v11, s[0:3], 0 offen
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_cvt_f16_f32_e32 v4, v60
-; SI-NEXT: v_cvt_f16_f32_e32 v6, v48
+; SI-NEXT: v_cvt_f16_f32_e32 v4, v9
+; SI-NEXT: v_cvt_f16_f32_e32 v6, v38
; SI-NEXT: v_add_i32_e32 v9, vcc, 0x50, v0
; SI-NEXT: v_lshlrev_b32_e32 v4, 16, v4
; SI-NEXT: v_or_b32_e32 v4, v6, v4
; SI-NEXT: buffer_store_dword v4, v9, s[0:3], 0 offen
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_cvt_f16_f32_e32 v4, v7
-; SI-NEXT: v_cvt_f16_f32_e32 v6, v38
+; SI-NEXT: v_cvt_f16_f32_e32 v6, v46
; SI-NEXT: v_add_i32_e32 v7, vcc, 0x54, v0
; SI-NEXT: v_lshlrev_b32_e32 v4, 16, v4
; SI-NEXT: v_or_b32_e32 v4, v6, v4
@@ -18979,7 +18984,7 @@ define inreg <52 x half> @bitcast_v26f32_to_v52f16_scalar(<26 x float> inreg %a,
; SI-NEXT: ; implicit-def: $vgpr54
; SI-NEXT: ; implicit-def: $vgpr52
; SI-NEXT: ; implicit-def: $vgpr51
-; SI-NEXT: ; implicit-def: $vgpr9
+; SI-NEXT: ; implicit-def: $vgpr50
; SI-NEXT: ; implicit-def: $vgpr49
; SI-NEXT: ; implicit-def: $vgpr11
; SI-NEXT: ; implicit-def: $vgpr39
@@ -19009,9 +19014,9 @@ define inreg <52 x half> @bitcast_v26f32_to_v52f16_scalar(<26 x float> inreg %a,
; SI-NEXT: ; implicit-def: $vgpr14
; SI-NEXT: ; implicit-def: $vgpr15
; SI-NEXT: ; implicit-def: $vgpr59
-; SI-NEXT: ; implicit-def: $vgpr48
-; SI-NEXT: ; implicit-def: $vgpr60
; SI-NEXT: ; implicit-def: $vgpr38
+; SI-NEXT: ; implicit-def: $vgpr9
+; SI-NEXT: ; implicit-def: $vgpr46
; SI-NEXT: ; implicit-def: $vgpr7
; SI-NEXT: ; implicit-def: $vgpr36
; SI-NEXT: ; implicit-def: $vgpr5
@@ -21163,15 +21168,15 @@ define inreg <26 x float> @bitcast_v52f16_to_v26f32_scalar(<52 x half> inreg %a,
; SI-NEXT: buffer_store_dword v0, off, s[0:3], s32 offset:124 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_cvt_f16_f32_e32 v0, v28
-; SI-NEXT: v_cvt_f16_f32_e32 v42, v5
-; SI-NEXT: v_cvt_f16_f32_e32 v55, v4
+; SI-NEXT: v_cvt_f16_f32_e32 v55, v5
+; SI-NEXT: v_cvt_f16_f32_e32 v42, v4
; SI-NEXT: v_cvt_f16_f32_e32 v56, v7
; SI-NEXT: buffer_store_dword v0, off, s[0:3], s32 offset:120 ; 4-byte Folded Spill
; SI-NEXT: v_cvt_f16_f32_e32 v54, v6
-; SI-NEXT: v_cvt_f16_f32_e32 v43, v9
-; SI-NEXT: v_cvt_f16_f32_e32 v47, v8
-; SI-NEXT: v_cvt_f16_f32_e32 v57, v11
-; SI-NEXT: v_cvt_f16_f32_e32 v60, v10
+; SI-NEXT: v_cvt_f16_f32_e32 v47, v9
+; SI-NEXT: v_cvt_f16_f32_e32 v57, v8
+; SI-NEXT: v_cvt_f16_f32_e32 v60, v11
+; SI-NEXT: v_cvt_f16_f32_e32 v53, v10
; SI-NEXT: v_cvt_f16_f32_e32 v13, v13
; SI-NEXT: v_cvt_f16_f32_e32 v52, v12
; SI-NEXT: v_cvt_f16_f32_e32 v15, v15
@@ -21189,13 +21194,13 @@ define inreg <26 x float> @bitcast_v52f16_to_v26f32_scalar(<52 x half> inreg %a,
; SI-NEXT: v_cvt_f16_f32_e32 v27, v26
; SI-NEXT: v_cvt_f16_f32_e32 v26, v29
; SI-NEXT: v_cvt_f16_f32_e32 v30, v30
-; SI-NEXT: v_cvt_f16_f32_e32 v53, s17
-; SI-NEXT: v_cvt_f16_f32_e32 v11, s16
+; SI-NEXT: v_cvt_f16_f32_e32 v43, s17
+; SI-NEXT: v_cvt_f16_f32_e32 v2, s16
; SI-NEXT: v_cvt_f16_f32_e32 v1, s19
-; SI-NEXT: v_cvt_f16_f32_e32 v2, s18
+; SI-NEXT: v_cvt_f16_f32_e32 v3, s18
; SI-NEXT: v_cvt_f16_f32_e32 v12, s21
; SI-NEXT: v_cvt_f16_f32_e32 v14, s20
-; SI-NEXT: v_cvt_f16_f32_e32 v3, s23
+; SI-NEXT: v_cvt_f16_f32_e32 v11, s23
; SI-NEXT: v_cvt_f16_f32_e32 v10, s22
; SI-NEXT: v_cvt_f16_f32_e32 v4, s25
; SI-NEXT: v_cvt_f16_f32_e32 v9, s24
@@ -21237,9 +21242,9 @@ define inreg <26 x float> @bitcast_v52f16_to_v26f32_scalar(<52 x half> inreg %a,
; SI-NEXT: buffer_store_dword v4, off, s[0:3], s32 offset:160 ; 4-byte Folded Spill
; SI-NEXT: buffer_store_dword v10, off, s[0:3], s32 offset:164 ; 4-byte Folded Spill
; SI-NEXT: buffer_store_dword v2, off, s[0:3], s32 offset:168 ; 4-byte Folded Spill
-; SI-NEXT: buffer_store_dword v3, off, s[0:3], s32 offset:172 ; 4-byte Folded Spill
-; SI-NEXT: buffer_store_dword v11, off, s[0:3], s32 offset:176 ; 4-byte Folded Spill
-; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:180 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:172 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v3, off, s[0:3], s32 offset:176 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v11, off, s[0:3], s32 offset:180 ; 4-byte Folded Spill
; SI-NEXT: buffer_store_dword v12, off, s[0:3], s32 offset:184 ; 4-byte Folded Spill
; SI-NEXT: buffer_store_dword v14, off, s[0:3], s32 offset:188 ; 4-byte Folded Spill
; SI-NEXT: s_cbranch_scc0 .LBB35_4
@@ -21250,17 +21255,18 @@ define inreg <26 x float> @bitcast_v52f16_to_v26f32_scalar(<52 x half> inreg %a,
; SI-NEXT: buffer_load_dword v25, off, s[0:3], s32 offset:104 ; 4-byte Folded Reload
; SI-NEXT: buffer_load_dword v22, off, s[0:3], s32 offset:120 ; 4-byte Folded Reload
; SI-NEXT: buffer_load_dword v38, off, s[0:3], s32 offset:96 ; 4-byte Folded Reload
-; SI-NEXT: s_waitcnt expcnt(2)
+; SI-NEXT: v_lshlrev_b32_e32 v0, 16, v43
+; SI-NEXT: s_waitcnt expcnt(4)
; SI-NEXT: v_lshlrev_b32_e32 v1, 16, v1
-; SI-NEXT: v_lshlrev_b32_e32 v0, 16, v53
-; SI-NEXT: v_or_b32_e32 v1, v2, v1
+; SI-NEXT: v_or_b32_e32 v0, v2, v0
+; SI-NEXT: v_or_b32_e32 v1, v3, v1
; SI-NEXT: v_lshlrev_b32_e32 v2, 16, v12
-; SI-NEXT: v_lshlrev_b32_e32 v3, 16, v3
+; SI-NEXT: s_waitcnt expcnt(3)
+; SI-NEXT: v_lshlrev_b32_e32 v3, 16, v11
; SI-NEXT: v_lshlrev_b32_e32 v4, 16, v4
; SI-NEXT: v_lshlrev_b32_e32 v5, 16, v5
; SI-NEXT: v_lshlrev_b32_e32 v6, 16, v6
; SI-NEXT: v_lshlrev_b32_e32 v21, 16, v26
-; SI-NEXT: v_or_b32_e32 v0, v11, v0
; SI-NEXT: v_or_b32_e32 v2, v14, v2
; SI-NEXT: v_or_b32_e32 v3, v10, v3
; SI-NEXT: v_or_b32_e32 v4, v9, v4
@@ -21268,11 +21274,12 @@ define inreg <26 x float> @bitcast_v52f16_to_v26f32_scalar(<52 x half> inreg %a,
; SI-NEXT: v_or_b32_e32 v6, v7, v6
; SI-NEXT: v_lshlrev_b32_e32 v7, 16, v46
; SI-NEXT: v_lshlrev_b32_e32 v8, 16, v41
-; SI-NEXT: v_lshlrev_b32_e32 v9, 16, v42
+; SI-NEXT: v_lshlrev_b32_e32 v9, 16, v55
; SI-NEXT: v_lshlrev_b32_e32 v10, 16, v56
-; SI-NEXT: v_lshlrev_b32_e32 v11, 16, v43
+; SI-NEXT: s_waitcnt expcnt(2)
+; SI-NEXT: v_lshlrev_b32_e32 v11, 16, v47
; SI-NEXT: s_waitcnt expcnt(1)
-; SI-NEXT: v_lshlrev_b32_e32 v12, 16, v57
+; SI-NEXT: v_lshlrev_b32_e32 v12, 16, v60
; SI-NEXT: v_lshlrev_b32_e32 v13, 16, v13
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_lshlrev_b32_e32 v14, 16, v15
@@ -21284,10 +21291,10 @@ define inreg <26 x float> @bitcast_v52f16_to_v26f32_scalar(<52 x half> inreg %a,
; SI-NEXT: v_mov_b32_e32 v51, v46
; SI-NEXT: v_or_b32_e32 v7, v45, v7
; SI-NEXT: v_or_b32_e32 v8, v40, v8
-; SI-NEXT: v_or_b32_e32 v9, v55, v9
+; SI-NEXT: v_or_b32_e32 v9, v42, v9
; SI-NEXT: v_or_b32_e32 v10, v54, v10
-; SI-NEXT: v_or_b32_e32 v11, v47, v11
-; SI-NEXT: v_or_b32_e32 v12, v60, v12
+; SI-NEXT: v_or_b32_e32 v11, v57, v11
+; SI-NEXT: v_or_b32_e32 v12, v53, v12
; SI-NEXT: v_or_b32_e32 v13, v52, v13
; SI-NEXT: v_or_b32_e32 v14, v63, v14
; SI-NEXT: v_or_b32_e32 v15, v61, v15
@@ -21315,14 +21322,16 @@ define inreg <26 x float> @bitcast_v52f16_to_v26f32_scalar(<52 x half> inreg %a,
; SI-NEXT: v_or_b32_e32 v25, v38, v25
; SI-NEXT: s_cbranch_execnz .LBB35_3
; SI-NEXT: .LBB35_2: ; %cmp.true
-; SI-NEXT: s_waitcnt expcnt(2)
-; SI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:176 ; 4-byte Folded Reload
-; SI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:180 ; 4-byte Folded Reload
-; SI-NEXT: buffer_load_dword v3, off, s[0:3], s32 offset:168 ; 4-byte Folded Reload
+; SI-NEXT: s_waitcnt expcnt(4)
+; SI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:168 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:172 ; 4-byte Folded Reload
+; SI-NEXT: s_waitcnt expcnt(3)
+; SI-NEXT: buffer_load_dword v3, off, s[0:3], s32 offset:176 ; 4-byte Folded Reload
; SI-NEXT: buffer_load_dword v4, off, s[0:3], s32 offset:164 ; 4-byte Folded Reload
-; SI-NEXT: v_cvt_f32_f16_e32 v0, v53
+; SI-NEXT: v_cvt_f32_f16_e32 v0, v43
; SI-NEXT: v_cvt_f32_f16_e32 v9, v40
-; SI-NEXT: v_cvt_f32_f16_e32 v10, v55
+; SI-NEXT: v_cvt_f32_f16_e32 v10, v42
+; SI-NEXT: s_waitcnt expcnt(2)
; SI-NEXT: v_cvt_f32_f16_e32 v11, v54
; SI-NEXT: v_add_f32_e32 v0, 0x38000000, v0
; SI-NEXT: v_cvt_f16_f32_e32 v0, v0
@@ -21334,8 +21343,8 @@ define inreg <26 x float> @bitcast_v52f16_to_v26f32_scalar(<52 x half> inreg %a,
; SI-NEXT: v_add_f32_e32 v11, 0x38000000, v11
; SI-NEXT: v_cvt_f16_f32_e32 v11, v11
; SI-NEXT: s_waitcnt expcnt(1)
-; SI-NEXT: v_cvt_f32_f16_e32 v12, v47
-; SI-NEXT: v_cvt_f32_f16_e32 v13, v60
+; SI-NEXT: v_cvt_f32_f16_e32 v12, v57
+; SI-NEXT: v_cvt_f32_f16_e32 v13, v53
; SI-NEXT: v_cvt_f32_f16_e32 v15, v52
; SI-NEXT: v_cvt_f32_f16_e32 v16, v63
; SI-NEXT: v_add_f32_e32 v12, 0x38000000, v12
@@ -21435,7 +21444,7 @@ define inreg <26 x float> @bitcast_v52f16_to_v26f32_scalar(<52 x half> inreg %a,
; SI-NEXT: v_cvt_f16_f32_e32 v3, v3
; SI-NEXT: v_lshlrev_b32_e32 v2, 16, v2
; SI-NEXT: v_or_b32_e32 v2, v3, v2
-; SI-NEXT: buffer_load_dword v3, off, s[0:3], s32 offset:172 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v3, off, s[0:3], s32 offset:180 ; 4-byte Folded Reload
; SI-NEXT: s_waitcnt vmcnt(0)
; SI-NEXT: v_cvt_f32_f16_e32 v3, v3
; SI-NEXT: v_add_f32_e32 v3, 0x38000000, v3
@@ -21473,7 +21482,7 @@ define inreg <26 x float> @bitcast_v52f16_to_v26f32_scalar(<52 x half> inreg %a,
; SI-NEXT: v_cvt_f16_f32_e32 v8, v8
; SI-NEXT: v_lshlrev_b32_e32 v8, 16, v8
; SI-NEXT: v_or_b32_e32 v8, v9, v8
-; SI-NEXT: v_cvt_f32_f16_e32 v9, v42
+; SI-NEXT: v_cvt_f32_f16_e32 v9, v55
; SI-NEXT: v_add_f32_e32 v9, 0x38000000, v9
; SI-NEXT: v_cvt_f16_f32_e32 v9, v9
; SI-NEXT: v_lshlrev_b32_e32 v9, 16, v9
@@ -21483,12 +21492,12 @@ define inreg <26 x float> @bitcast_v52f16_to_v26f32_scalar(<52 x half> inreg %a,
; SI-NEXT: v_cvt_f16_f32_e32 v10, v10
; SI-NEXT: v_lshlrev_b32_e32 v10, 16, v10
; SI-NEXT: v_or_b32_e32 v10, v11, v10
-; SI-NEXT: v_cvt_f32_f16_e32 v11, v43
+; SI-NEXT: v_cvt_f32_f16_e32 v11, v47
; SI-NEXT: v_add_f32_e32 v11, 0x38000000, v11
; SI-NEXT: v_cvt_f16_f32_e32 v11, v11
; SI-NEXT: v_lshlrev_b32_e32 v11, 16, v11
; SI-NEXT: v_or_b32_e32 v11, v12, v11
-; SI-NEXT: v_cvt_f32_f16_e32 v12, v57
+; SI-NEXT: v_cvt_f32_f16_e32 v12, v60
; SI-NEXT: v_add_f32_e32 v12, 0x38000000, v12
; SI-NEXT: v_cvt_f16_f32_e32 v12, v12
; SI-NEXT: v_lshlrev_b32_e32 v12, 16, v12
@@ -22231,8 +22240,8 @@ define inreg <26 x float> @bitcast_v52f16_to_v26f32_scalar(<52 x half> inreg %a,
; GFX11-FAKE16-NEXT: v_and_b32_e32 v50, 0xffff, v5
; GFX11-FAKE16-NEXT: v_and_b32_e32 v49, 0xffff, v6
; GFX11-FAKE16-NEXT: v_and_b32_e32 v48, 0xffff, v7
-; GFX11-FAKE16-NEXT: s_lshr_b32 s41, s29, 16
-; GFX11-FAKE16-NEXT: s_lshr_b32 s15, s28, 16
+; GFX11-FAKE16-NEXT: s_lshr_b32 s40, s29, 16
+; GFX11-FAKE16-NEXT: s_lshr_b32 s41, s28, 16
; GFX11-FAKE16-NEXT: s_lshr_b32 s42, s27, 16
; GFX11-FAKE16-NEXT: s_lshr_b32 s14, s26, 16
; GFX11-FAKE16-NEXT: s_lshr_b32 s13, s25, 16
@@ -22249,7 +22258,7 @@ define inreg <26 x float> @bitcast_v52f16_to_v26f32_scalar(<52 x half> inreg %a,
; GFX11-FAKE16-NEXT: s_lshr_b32 s44, s2, 16
; GFX11-FAKE16-NEXT: s_lshr_b32 s45, s1, 16
; GFX11-FAKE16-NEXT: s_lshr_b32 s46, s0, 16
-; GFX11-FAKE16-NEXT: s_mov_b32 s40, 0
+; GFX11-FAKE16-NEXT: s_mov_b32 s15, 0
; GFX11-FAKE16-NEXT: s_pack_ll_b32_b16 s0, s0, s46
; GFX11-FAKE16-NEXT: s_pack_ll_b32_b16 s1, s1, s45
; GFX11-FAKE16-NEXT: s_pack_ll_b32_b16 s2, s2, s44
@@ -22265,9 +22274,9 @@ define inreg <26 x float> @bitcast_v52f16_to_v26f32_scalar(<52 x half> inreg %a,
; GFX11-FAKE16-NEXT: s_pack_ll_b32_b16 s12, s24, s12
; GFX11-FAKE16-NEXT: s_pack_ll_b32_b16 s13, s25, s13
; GFX11-FAKE16-NEXT: s_pack_ll_b32_b16 s14, s26, s14
-; GFX11-FAKE16-NEXT: s_pack_ll_b32_b16 s17, s27, s42
-; GFX11-FAKE16-NEXT: s_pack_ll_b32_b16 s15, s28, s15
-; GFX11-FAKE16-NEXT: s_pack_ll_b32_b16 s16, s29, s41
+; GFX11-FAKE16-NEXT: s_pack_ll_b32_b16 s18, s27, s42
+; GFX11-FAKE16-NEXT: s_pack_ll_b32_b16 s16, s28, s41
+; GFX11-FAKE16-NEXT: s_pack_ll_b32_b16 s17, s29, s40
; GFX11-FAKE16-NEXT: s_and_b32 s47, vcc_lo, exec_lo
; GFX11-FAKE16-NEXT: s_cbranch_scc0 .LBB35_4
; GFX11-FAKE16-NEXT: ; %bb.1: ; %cmp.false
@@ -22286,9 +22295,9 @@ define inreg <26 x float> @bitcast_v52f16_to_v26f32_scalar(<52 x half> inreg %a,
; GFX11-FAKE16-NEXT: v_dual_mov_b32 v8, s8 :: v_dual_mov_b32 v9, s9
; GFX11-FAKE16-NEXT: v_dual_mov_b32 v10, s10 :: v_dual_mov_b32 v11, s11
; GFX11-FAKE16-NEXT: v_dual_mov_b32 v12, s12 :: v_dual_mov_b32 v13, s13
-; GFX11-FAKE16-NEXT: v_dual_mov_b32 v14, s14 :: v_dual_mov_b32 v15, s17
-; GFX11-FAKE16-NEXT: v_dual_mov_b32 v16, s15 :: v_dual_mov_b32 v17, s16
-; GFX11-FAKE16-NEXT: s_and_not1_b32 vcc_lo, exec_lo, s40
+; GFX11-FAKE16-NEXT: v_dual_mov_b32 v14, s14 :: v_dual_mov_b32 v15, s18
+; GFX11-FAKE16-NEXT: v_dual_mov_b32 v16, s16 :: v_dual_mov_b32 v17, s17
+; GFX11-FAKE16-NEXT: s_and_not1_b32 vcc_lo, exec_lo, s15
; GFX11-FAKE16-NEXT: s_cbranch_vccnz .LBB35_3
; GFX11-FAKE16-NEXT: .LBB35_2: ; %cmp.true
; GFX11-FAKE16-NEXT: v_lshl_or_b32 v18, v39, 16, v55
@@ -22314,9 +22323,9 @@ define inreg <26 x float> @bitcast_v52f16_to_v26f32_scalar(<52 x half> inreg %a,
; GFX11-FAKE16-NEXT: v_pk_add_f16 v12, 0x200, s12 op_sel_hi:[0,1]
; GFX11-FAKE16-NEXT: v_pk_add_f16 v13, 0x200, s13 op_sel_hi:[0,1]
; GFX11-FAKE16-NEXT: v_pk_add_f16 v14, 0x200, s14 op_sel_hi:[0,1]
-; GFX11-FAKE16-NEXT: v_pk_add_f16 v15, 0x200, s17 op_sel_hi:[0,1]
-; GFX11-FAKE16-NEXT: v_pk_add_f16 v16, 0x200, s15 op_sel_hi:[0,1]
-; GFX11-FAKE16-NEXT: v_pk_add_f16 v17, 0x200, s16 op_sel_hi:[0,1]
+; GFX11-FAKE16-NEXT: v_pk_add_f16 v15, 0x200, s18 op_sel_hi:[0,1]
+; GFX11-FAKE16-NEXT: v_pk_add_f16 v16, 0x200, s16 op_sel_hi:[0,1]
+; GFX11-FAKE16-NEXT: v_pk_add_f16 v17, 0x200, s17 op_sel_hi:[0,1]
; GFX11-FAKE16-NEXT: v_pk_add_f16 v18, 0x200, v18 op_sel_hi:[0,1]
; GFX11-FAKE16-NEXT: v_pk_add_f16 v19, 0x200, v19 op_sel_hi:[0,1]
; GFX11-FAKE16-NEXT: v_pk_add_f16 v20, 0x200, v20 op_sel_hi:[0,1]
@@ -24742,7 +24751,7 @@ define inreg <52 x i16> @bitcast_v13i64_to_v52i16_scalar(<13 x i64> inreg %a, i3
; GFX11-TRUE16-NEXT: v_readfirstlane_b32 s5, v5
; GFX11-TRUE16-NEXT: v_readfirstlane_b32 s6, v6
; GFX11-TRUE16-NEXT: v_readfirstlane_b32 s4, v7
-; GFX11-TRUE16-NEXT: s_mov_b32 s78, 0
+; GFX11-TRUE16-NEXT: s_mov_b32 s74, 0
; GFX11-TRUE16-NEXT: s_and_b32 s12, vcc_lo, exec_lo
; GFX11-TRUE16-NEXT: s_cbranch_scc0 .LBB41_4
; GFX11-TRUE16-NEXT: ; %bb.1: ; %cmp.false
@@ -24768,11 +24777,11 @@ define inreg <52 x i16> @bitcast_v13i64_to_v52i16_scalar(<13 x i64> inreg %a, i3
; GFX11-TRUE16-NEXT: s_lshr_b32 s63, s18, 16
; GFX11-TRUE16-NEXT: s_lshr_b32 s72, s17, 16
; GFX11-TRUE16-NEXT: s_lshr_b32 s73, s16, 16
-; GFX11-TRUE16-NEXT: s_lshr_b32 s74, s3, 16
-; GFX11-TRUE16-NEXT: s_lshr_b32 s75, s2, 16
-; GFX11-TRUE16-NEXT: s_lshr_b32 s76, s1, 16
-; GFX11-TRUE16-NEXT: s_lshr_b32 s77, s0, 16
-; GFX11-TRUE16-NEXT: s_and_not1_b32 vcc_lo, exec_lo, s78
+; GFX11-TRUE16-NEXT: s_lshr_b32 s75, s3, 16
+; GFX11-TRUE16-NEXT: s_lshr_b32 s76, s2, 16
+; GFX11-TRUE16-NEXT: s_lshr_b32 s77, s1, 16
+; GFX11-TRUE16-NEXT: s_lshr_b32 s78, s0, 16
+; GFX11-TRUE16-NEXT: s_and_not1_b32 vcc_lo, exec_lo, s74
; GFX11-TRUE16-NEXT: s_cbranch_vccnz .LBB41_3
; GFX11-TRUE16-NEXT: .LBB41_2: ; %cmp.true
; GFX11-TRUE16-NEXT: s_add_u32 s6, s6, 3
@@ -24823,16 +24832,18 @@ define inreg <52 x i16> @bitcast_v13i64_to_v52i16_scalar(<13 x i64> inreg %a, i3
; GFX11-TRUE16-NEXT: s_lshr_b32 s63, s18, 16
; GFX11-TRUE16-NEXT: s_lshr_b32 s72, s17, 16
; GFX11-TRUE16-NEXT: s_lshr_b32 s73, s16, 16
-; GFX11-TRUE16-NEXT: s_lshr_b32 s74, s3, 16
-; GFX11-TRUE16-NEXT: s_lshr_b32 s75, s2, 16
-; GFX11-TRUE16-NEXT: s_lshr_b32 s76, s1, 16
-; GFX11-TRUE16-NEXT: s_lshr_b32 s77, s0, 16
+; GFX11-TRUE16-NEXT: s_lshr_b32 s75, s3, 16
+; GFX11-TRUE16-NEXT: s_lshr_b32 s76, s2, 16
+; GFX11-TRUE16-NEXT: s_lshr_b32 s77, s1, 16
+; GFX11-TRUE16-NEXT: s_lshr_b32 s78, s0, 16
; GFX11-TRUE16-NEXT: .LBB41_3: ; %end
; GFX11-TRUE16-NEXT: s_delay_alu instid0(SALU_CYCLE_1)
-; GFX11-TRUE16-NEXT: s_pack_ll_b32_b16 s0, s0, s77
-; GFX11-TRUE16-NEXT: s_pack_ll_b32_b16 s1, s1, s76
-; GFX11-TRUE16-NEXT: s_pack_ll_b32_b16 s2, s2, s75
-; GFX11-TRUE16-NEXT: s_pack_ll_b32_b16 s3, s3, s74
+; GFX11-TRUE16-NEXT: s_mov_b32 s74, s78
+; GFX11-TRUE16-NEXT: s_pack_ll_b32_b16 s1, s1, s77
+; GFX11-TRUE16-NEXT: s_pack_ll_b32_b16 s0, s0, s74
+; GFX11-TRUE16-NEXT: s_mov_b32 s74, s76
+; GFX11-TRUE16-NEXT: s_pack_ll_b32_b16 s3, s3, s75
+; GFX11-TRUE16-NEXT: s_pack_ll_b32_b16 s2, s2, s74
; GFX11-TRUE16-NEXT: s_pack_ll_b32_b16 s16, s16, s73
; GFX11-TRUE16-NEXT: s_pack_ll_b32_b16 s17, s17, s72
; GFX11-TRUE16-NEXT: s_pack_ll_b32_b16 s18, s18, s63
@@ -24870,10 +24881,10 @@ define inreg <52 x i16> @bitcast_v13i64_to_v52i16_scalar(<13 x i64> inreg %a, i3
; GFX11-TRUE16-NEXT: v_dual_mov_b32 v24, s6 :: v_dual_mov_b32 v25, s4
; GFX11-TRUE16-NEXT: s_setpc_b64 s[30:31]
; GFX11-TRUE16-NEXT: .LBB41_4:
+; GFX11-TRUE16-NEXT: ; implicit-def: $sgpr78_lo16
; GFX11-TRUE16-NEXT: ; implicit-def: $sgpr77_lo16
; GFX11-TRUE16-NEXT: ; implicit-def: $sgpr76_lo16
; GFX11-TRUE16-NEXT: ; implicit-def: $sgpr75_lo16
-; GFX11-TRUE16-NEXT: ; implicit-def: $sgpr74_lo16
; GFX11-TRUE16-NEXT: ; implicit-def: $sgpr73_lo16
; GFX11-TRUE16-NEXT: ; implicit-def: $sgpr72_lo16
; GFX11-TRUE16-NEXT: ; implicit-def: $sgpr63_lo16
@@ -27235,8 +27246,8 @@ define inreg <13 x i64> @bitcast_v52i16_to_v13i64_scalar(<52 x i16> inreg %a, i3
; GFX11-FAKE16-NEXT: v_and_b32_e32 v50, 0xffff, v5
; GFX11-FAKE16-NEXT: v_and_b32_e32 v49, 0xffff, v6
; GFX11-FAKE16-NEXT: v_and_b32_e32 v48, 0xffff, v7
-; GFX11-FAKE16-NEXT: s_lshr_b32 s41, s29, 16
-; GFX11-FAKE16-NEXT: s_lshr_b32 s15, s28, 16
+; GFX11-FAKE16-NEXT: s_lshr_b32 s40, s29, 16
+; GFX11-FAKE16-NEXT: s_lshr_b32 s41, s28, 16
; GFX11-FAKE16-NEXT: s_lshr_b32 s42, s27, 16
; GFX11-FAKE16-NEXT: s_lshr_b32 s14, s26, 16
; GFX11-FAKE16-NEXT: s_lshr_b32 s13, s25, 16
@@ -27253,7 +27264,7 @@ define inreg <13 x i64> @bitcast_v52i16_to_v13i64_scalar(<52 x i16> inreg %a, i3
; GFX11-FAKE16-NEXT: s_lshr_b32 s44, s2, 16
; GFX11-FAKE16-NEXT: s_lshr_b32 s45, s1, 16
; GFX11-FAKE16-NEXT: s_lshr_b32 s46, s0, 16
-; GFX11-FAKE16-NEXT: s_mov_b32 s40, 0
+; GFX11-FAKE16-NEXT: s_mov_b32 s15, 0
; GFX11-FAKE16-NEXT: s_pack_ll_b32_b16 s0, s0, s46
; GFX11-FAKE16-NEXT: s_pack_ll_b32_b16 s1, s1, s45
; GFX11-FAKE16-NEXT: s_pack_ll_b32_b16 s2, s2, s44
@@ -27269,9 +27280,9 @@ define inreg <13 x i64> @bitcast_v52i16_to_v13i64_scalar(<52 x i16> inreg %a, i3
; GFX11-FAKE16-NEXT: s_pack_ll_b32_b16 s12, s24, s12
; GFX11-FAKE16-NEXT: s_pack_ll_b32_b16 s13, s25, s13
; GFX11-FAKE16-NEXT: s_pack_ll_b32_b16 s14, s26, s14
-; GFX11-FAKE16-NEXT: s_pack_ll_b32_b16 s17, s27, s42
-; GFX11-FAKE16-NEXT: s_pack_ll_b32_b16 s15, s28, s15
-; GFX11-FAKE16-NEXT: s_pack_ll_b32_b16 s16, s29, s41
+; GFX11-FAKE16-NEXT: s_pack_ll_b32_b16 s18, s27, s42
+; GFX11-FAKE16-NEXT: s_pack_ll_b32_b16 s16, s28, s41
+; GFX11-FAKE16-NEXT: s_pack_ll_b32_b16 s17, s29, s40
; GFX11-FAKE16-NEXT: s_and_b32 s47, vcc_lo, exec_lo
; GFX11-FAKE16-NEXT: s_cbranch_scc0 .LBB43_4
; GFX11-FAKE16-NEXT: ; %bb.1: ; %cmp.false
@@ -27290,9 +27301,9 @@ define inreg <13 x i64> @bitcast_v52i16_to_v13i64_scalar(<52 x i16> inreg %a, i3
; GFX11-FAKE16-NEXT: v_dual_mov_b32 v8, s8 :: v_dual_mov_b32 v9, s9
; GFX11-FAKE16-NEXT: v_dual_mov_b32 v10, s10 :: v_dual_mov_b32 v11, s11
; GFX11-FAKE16-NEXT: v_dual_mov_b32 v12, s12 :: v_dual_mov_b32 v13, s13
-; GFX11-FAKE16-NEXT: v_dual_mov_b32 v14, s14 :: v_dual_mov_b32 v15, s17
-; GFX11-FAKE16-NEXT: v_dual_mov_b32 v16, s15 :: v_dual_mov_b32 v17, s16
-; GFX11-FAKE16-NEXT: s_and_not1_b32 vcc_lo, exec_lo, s40
+; GFX11-FAKE16-NEXT: v_dual_mov_b32 v14, s14 :: v_dual_mov_b32 v15, s18
+; GFX11-FAKE16-NEXT: v_dual_mov_b32 v16, s16 :: v_dual_mov_b32 v17, s17
+; GFX11-FAKE16-NEXT: s_and_not1_b32 vcc_lo, exec_lo, s15
; GFX11-FAKE16-NEXT: s_cbranch_vccnz .LBB43_3
; GFX11-FAKE16-NEXT: .LBB43_2: ; %cmp.true
; GFX11-FAKE16-NEXT: v_lshl_or_b32 v18, v39, 16, v55
@@ -27318,9 +27329,9 @@ define inreg <13 x i64> @bitcast_v52i16_to_v13i64_scalar(<52 x i16> inreg %a, i3
; GFX11-FAKE16-NEXT: v_pk_add_u16 v12, s12, 3 op_sel_hi:[1,0]
; GFX11-FAKE16-NEXT: v_pk_add_u16 v13, s13, 3 op_sel_hi:[1,0]
; GFX11-FAKE16-NEXT: v_pk_add_u16 v14, s14, 3 op_sel_hi:[1,0]
-; GFX11-FAKE16-NEXT: v_pk_add_u16 v15, s17, 3 op_sel_hi:[1,0]
-; GFX11-FAKE16-NEXT: v_pk_add_u16 v16, s15, 3 op_sel_hi:[1,0]
-; GFX11-FAKE16-NEXT: v_pk_add_u16 v17, s16, 3 op_sel_hi:[1,0]
+; GFX11-FAKE16-NEXT: v_pk_add_u16 v15, s18, 3 op_sel_hi:[1,0]
+; GFX11-FAKE16-NEXT: v_pk_add_u16 v16, s16, 3 op_sel_hi:[1,0]
+; GFX11-FAKE16-NEXT: v_pk_add_u16 v17, s17, 3 op_sel_hi:[1,0]
; GFX11-FAKE16-NEXT: v_pk_add_u16 v18, v18, 3 op_sel_hi:[1,0]
; GFX11-FAKE16-NEXT: v_pk_add_u16 v19, v19, 3 op_sel_hi:[1,0]
; GFX11-FAKE16-NEXT: v_pk_add_u16 v20, v20, 3 op_sel_hi:[1,0]
@@ -29368,7 +29379,7 @@ define inreg <52 x half> @bitcast_v13i64_to_v52f16_scalar(<13 x i64> inreg %a, i
; GFX11-TRUE16-NEXT: v_readfirstlane_b32 s5, v5
; GFX11-TRUE16-NEXT: v_readfirstlane_b32 s6, v6
; GFX11-TRUE16-NEXT: v_readfirstlane_b32 s4, v7
-; GFX11-TRUE16-NEXT: s_mov_b32 s78, 0
+; GFX11-TRUE16-NEXT: s_mov_b32 s74, 0
; GFX11-TRUE16-NEXT: s_and_b32 s12, vcc_lo, exec_lo
; GFX11-TRUE16-NEXT: s_cbranch_scc0 .LBB45_4
; GFX11-TRUE16-NEXT: ; %bb.1: ; %cmp.false
@@ -29394,11 +29405,11 @@ define inreg <52 x half> @bitcast_v13i64_to_v52f16_scalar(<13 x i64> inreg %a, i
; GFX11-TRUE16-NEXT: s_lshr_b32 s63, s18, 16
; GFX11-TRUE16-NEXT: s_lshr_b32 s72, s17, 16
; GFX11-TRUE16-NEXT: s_lshr_b32 s73, s16, 16
-; GFX11-TRUE16-NEXT: s_lshr_b32 s74, s3, 16
-; GFX11-TRUE16-NEXT: s_lshr_b32 s75, s2, 16
-; GFX11-TRUE16-NEXT: s_lshr_b32 s76, s1, 16
-; GFX11-TRUE16-NEXT: s_lshr_b32 s77, s0, 16
-; GFX11-TRUE16-NEXT: s_and_not1_b32 vcc_lo, exec_lo, s78
+; GFX11-TRUE16-NEXT: s_lshr_b32 s75, s3, 16
+; GFX11-TRUE16-NEXT: s_lshr_b32 s76, s2, 16
+; GFX11-TRUE16-NEXT: s_lshr_b32 s77, s1, 16
+; GFX11-TRUE16-NEXT: s_lshr_b32 s78, s0, 16
+; GFX11-TRUE16-NEXT: s_and_not1_b32 vcc_lo, exec_lo, s74
; GFX11-TRUE16-NEXT: s_cbranch_vccnz .LBB45_3
; GFX11-TRUE16-NEXT: .LBB45_2: ; %cmp.true
; GFX11-TRUE16-NEXT: s_add_u32 s6, s6, 3
@@ -29449,16 +29460,18 @@ define inreg <52 x half> @bitcast_v13i64_to_v52f16_scalar(<13 x i64> inreg %a, i
; GFX11-TRUE16-NEXT: s_lshr_b32 s63, s18, 16
; GFX11-TRUE16-NEXT: s_lshr_b32 s72, s17, 16
; GFX11-TRUE16-NEXT: s_lshr_b32 s73, s16, 16
-; GFX11-TRUE16-NEXT: s_lshr_b32 s74, s3, 16
-; GFX11-TRUE16-NEXT: s_lshr_b32 s75, s2, 16
-; GFX11-TRUE16-NEXT: s_lshr_b32 s76, s1, 16
-; GFX11-TRUE16-NEXT: s_lshr_b32 s77, s0, 16
+; GFX11-TRUE16-NEXT: s_lshr_b32 s75, s3, 16
+; GFX11-TRUE16-NEXT: s_lshr_b32 s76, s2, 16
+; GFX11-TRUE16-NEXT: s_lshr_b32 s77, s1, 16
+; GFX11-TRUE16-NEXT: s_lshr_b32 s78, s0, 16
; GFX11-TRUE16-NEXT: .LBB45_3: ; %end
; GFX11-TRUE16-NEXT: s_delay_alu instid0(SALU_CYCLE_1)
-; GFX11-TRUE16-NEXT: s_pack_ll_b32_b16 s0, s0, s77
-; GFX11-TRUE16-NEXT: s_pack_ll_b32_b16 s1, s1, s76
-; GFX11-TRUE16-NEXT: s_pack_ll_b32_b16 s2, s2, s75
-; GFX11-TRUE16-NEXT: s_pack_ll_b32_b16 s3, s3, s74
+; GFX11-TRUE16-NEXT: s_mov_b32 s74, s78
+; GFX11-TRUE16-NEXT: s_pack_ll_b32_b16 s1, s1, s77
+; GFX11-TRUE16-NEXT: s_pack_ll_b32_b16 s0, s0, s74
+; GFX11-TRUE16-NEXT: s_mov_b32 s74, s76
+; GFX11-TRUE16-NEXT: s_pack_ll_b32_b16 s3, s3, s75
+; GFX11-TRUE16-NEXT: s_pack_ll_b32_b16 s2, s2, s74
; GFX11-TRUE16-NEXT: s_pack_ll_b32_b16 s16, s16, s73
; GFX11-TRUE16-NEXT: s_pack_ll_b32_b16 s17, s17, s72
; GFX11-TRUE16-NEXT: s_pack_ll_b32_b16 s18, s18, s63
@@ -29496,10 +29509,10 @@ define inreg <52 x half> @bitcast_v13i64_to_v52f16_scalar(<13 x i64> inreg %a, i
; GFX11-TRUE16-NEXT: v_dual_mov_b32 v24, s6 :: v_dual_mov_b32 v25, s4
; GFX11-TRUE16-NEXT: s_setpc_b64 s[30:31]
; GFX11-TRUE16-NEXT: .LBB45_4:
+; GFX11-TRUE16-NEXT: ; implicit-def: $sgpr78_lo16
; GFX11-TRUE16-NEXT: ; implicit-def: $sgpr77_lo16
; GFX11-TRUE16-NEXT: ; implicit-def: $sgpr76_lo16
; GFX11-TRUE16-NEXT: ; implicit-def: $sgpr75_lo16
-; GFX11-TRUE16-NEXT: ; implicit-def: $sgpr74_lo16
; GFX11-TRUE16-NEXT: ; implicit-def: $sgpr73_lo16
; GFX11-TRUE16-NEXT: ; implicit-def: $sgpr72_lo16
; GFX11-TRUE16-NEXT: ; implicit-def: $sgpr63_lo16
@@ -31076,15 +31089,15 @@ define inreg <13 x i64> @bitcast_v52f16_to_v13i64_scalar(<52 x half> inreg %a, i
; SI-NEXT: buffer_store_dword v0, off, s[0:3], s32 offset:124 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_cvt_f16_f32_e32 v0, v28
-; SI-NEXT: v_cvt_f16_f32_e32 v42, v5
-; SI-NEXT: v_cvt_f16_f32_e32 v55, v4
+; SI-NEXT: v_cvt_f16_f32_e32 v55, v5
+; SI-NEXT: v_cvt_f16_f32_e32 v42, v4
; SI-NEXT: v_cvt_f16_f32_e32 v56, v7
; SI-NEXT: buffer_store_dword v0, off, s[0:3], s32 offset:120 ; 4-byte Folded Spill
; SI-NEXT: v_cvt_f16_f32_e32 v54, v6
-; SI-NEXT: v_cvt_f16_f32_e32 v43, v9
-; SI-NEXT: v_cvt_f16_f32_e32 v47, v8
-; SI-NEXT: v_cvt_f16_f32_e32 v57, v11
-; SI-NEXT: v_cvt_f16_f32_e32 v60, v10
+; SI-NEXT: v_cvt_f16_f32_e32 v47, v9
+; SI-NEXT: v_cvt_f16_f32_e32 v57, v8
+; SI-NEXT: v_cvt_f16_f32_e32 v60, v11
+; SI-NEXT: v_cvt_f16_f32_e32 v53, v10
; SI-NEXT: v_cvt_f16_f32_e32 v13, v13
; SI-NEXT: v_cvt_f16_f32_e32 v52, v12
; SI-NEXT: v_cvt_f16_f32_e32 v15, v15
@@ -31102,13 +31115,13 @@ define inreg <13 x i64> @bitcast_v52f16_to_v13i64_scalar(<52 x half> inreg %a, i
; SI-NEXT: v_cvt_f16_f32_e32 v27, v26
; SI-NEXT: v_cvt_f16_f32_e32 v26, v29
; SI-NEXT: v_cvt_f16_f32_e32 v30, v30
-; SI-NEXT: v_cvt_f16_f32_e32 v53, s17
-; SI-NEXT: v_cvt_f16_f32_e32 v11, s16
+; SI-NEXT: v_cvt_f16_f32_e32 v43, s17
+; SI-NEXT: v_cvt_f16_f32_e32 v2, s16
; SI-NEXT: v_cvt_f16_f32_e32 v1, s19
-; SI-NEXT: v_cvt_f16_f32_e32 v2, s18
+; SI-NEXT: v_cvt_f16_f32_e32 v3, s18
; SI-NEXT: v_cvt_f16_f32_e32 v12, s21
; SI-NEXT: v_cvt_f16_f32_e32 v14, s20
-; SI-NEXT: v_cvt_f16_f32_e32 v3, s23
+; SI-NEXT: v_cvt_f16_f32_e32 v11, s23
; SI-NEXT: v_cvt_f16_f32_e32 v10, s22
; SI-NEXT: v_cvt_f16_f32_e32 v4, s25
; SI-NEXT: v_cvt_f16_f32_e32 v9, s24
@@ -31150,9 +31163,9 @@ define inreg <13 x i64> @bitcast_v52f16_to_v13i64_scalar(<52 x half> inreg %a, i
; SI-NEXT: buffer_store_dword v4, off, s[0:3], s32 offset:160 ; 4-byte Folded Spill
; SI-NEXT: buffer_store_dword v10, off, s[0:3], s32 offset:164 ; 4-byte Folded Spill
; SI-NEXT: buffer_store_dword v2, off, s[0:3], s32 offset:168 ; 4-byte Folded Spill
-; SI-NEXT: buffer_store_dword v3, off, s[0:3], s32 offset:172 ; 4-byte Folded Spill
-; SI-NEXT: buffer_store_dword v11, off, s[0:3], s32 offset:176 ; 4-byte Folded Spill
-; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:180 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:172 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v3, off, s[0:3], s32 offset:176 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v11, off, s[0:3], s32 offset:180 ; 4-byte Folded Spill
; SI-NEXT: buffer_store_dword v12, off, s[0:3], s32 offset:184 ; 4-byte Folded Spill
; SI-NEXT: buffer_store_dword v14, off, s[0:3], s32 offset:188 ; 4-byte Folded Spill
; SI-NEXT: s_cbranch_scc0 .LBB47_4
@@ -31163,17 +31176,18 @@ define inreg <13 x i64> @bitcast_v52f16_to_v13i64_scalar(<52 x half> inreg %a, i
; SI-NEXT: buffer_load_dword v25, off, s[0:3], s32 offset:104 ; 4-byte Folded Reload
; SI-NEXT: buffer_load_dword v22, off, s[0:3], s32 offset:120 ; 4-byte Folded Reload
; SI-NEXT: buffer_load_dword v38, off, s[0:3], s32 offset:96 ; 4-byte Folded Reload
-; SI-NEXT: s_waitcnt expcnt(2)
+; SI-NEXT: v_lshlrev_b32_e32 v0, 16, v43
+; SI-NEXT: s_waitcnt expcnt(4)
; SI-NEXT: v_lshlrev_b32_e32 v1, 16, v1
-; SI-NEXT: v_lshlrev_b32_e32 v0, 16, v53
-; SI-NEXT: v_or_b32_e32 v1, v2, v1
+; SI-NEXT: v_or_b32_e32 v0, v2, v0
+; SI-NEXT: v_or_b32_e32 v1, v3, v1
; SI-NEXT: v_lshlrev_b32_e32 v2, 16, v12
-; SI-NEXT: v_lshlrev_b32_e32 v3, 16, v3
+; SI-NEXT: s_waitcnt expcnt(3)
+; SI-NEXT: v_lshlrev_b32_e32 v3, 16, v11
; SI-NEXT: v_lshlrev_b32_e32 v4, 16, v4
; SI-NEXT: v_lshlrev_b32_e32 v5, 16, v5
; SI-NEXT: v_lshlrev_b32_e32 v6, 16, v6
; SI-NEXT: v_lshlrev_b32_e32 v21, 16, v26
-; SI-NEXT: v_or_b32_e32 v0, v11, v0
; SI-NEXT: v_or_b32_e32 v2, v14, v2
; SI-NEXT: v_or_b32_e32 v3, v10, v3
; SI-NEXT: v_or_b32_e32 v4, v9, v4
@@ -31181,11 +31195,12 @@ define inreg <13 x i64> @bitcast_v52f16_to_v13i64_scalar(<52 x half> inreg %a, i
; SI-NEXT: v_or_b32_e32 v6, v7, v6
; SI-NEXT: v_lshlrev_b32_e32 v7, 16, v46
; SI-NEXT: v_lshlrev_b32_e32 v8, 16, v41
-; SI-NEXT: v_lshlrev_b32_e32 v9, 16, v42
+; SI-NEXT: v_lshlrev_b32_e32 v9, 16, v55
; SI-NEXT: v_lshlrev_b32_e32 v10, 16, v56
-; SI-NEXT: v_lshlrev_b32_e32 v11, 16, v43
+; SI-NEXT: s_waitcnt expcnt(2)
+; SI-NEXT: v_lshlrev_b32_e32 v11, 16, v47
; SI-NEXT: s_waitcnt expcnt(1)
-; SI-NEXT: v_lshlrev_b32_e32 v12, 16, v57
+; SI-NEXT: v_lshlrev_b32_e32 v12, 16, v60
; SI-NEXT: v_lshlrev_b32_e32 v13, 16, v13
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_lshlrev_b32_e32 v14, 16, v15
@@ -31197,10 +31212,10 @@ define inreg <13 x i64> @bitcast_v52f16_to_v13i64_scalar(<52 x half> inreg %a, i
; SI-NEXT: v_mov_b32_e32 v51, v46
; SI-NEXT: v_or_b32_e32 v7, v45, v7
; SI-NEXT: v_or_b32_e32 v8, v40, v8
-; SI-NEXT: v_or_b32_e32 v9, v55, v9
+; SI-NEXT: v_or_b32_e32 v9, v42, v9
; SI-NEXT: v_or_b32_e32 v10, v54, v10
-; SI-NEXT: v_or_b32_e32 v11, v47, v11
-; SI-NEXT: v_or_b32_e32 v12, v60, v12
+; SI-NEXT: v_or_b32_e32 v11, v57, v11
+; SI-NEXT: v_or_b32_e32 v12, v53, v12
; SI-NEXT: v_or_b32_e32 v13, v52, v13
; SI-NEXT: v_or_b32_e32 v14, v63, v14
; SI-NEXT: v_or_b32_e32 v15, v61, v15
@@ -31228,14 +31243,16 @@ define inreg <13 x i64> @bitcast_v52f16_to_v13i64_scalar(<52 x half> inreg %a, i
; SI-NEXT: v_or_b32_e32 v25, v38, v25
; SI-NEXT: s_cbranch_execnz .LBB47_3
; SI-NEXT: .LBB47_2: ; %cmp.true
-; SI-NEXT: s_waitcnt expcnt(2)
-; SI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:176 ; 4-byte Folded Reload
-; SI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:180 ; 4-byte Folded Reload
-; SI-NEXT: buffer_load_dword v3, off, s[0:3], s32 offset:168 ; 4-byte Folded Reload
+; SI-NEXT: s_waitcnt expcnt(4)
+; SI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:168 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:172 ; 4-byte Folded Reload
+; SI-NEXT: s_waitcnt expcnt(3)
+; SI-NEXT: buffer_load_dword v3, off, s[0:3], s32 offset:176 ; 4-byte Folded Reload
; SI-NEXT: buffer_load_dword v4, off, s[0:3], s32 offset:164 ; 4-byte Folded Reload
-; SI-NEXT: v_cvt_f32_f16_e32 v0, v53
+; SI-NEXT: v_cvt_f32_f16_e32 v0, v43
; SI-NEXT: v_cvt_f32_f16_e32 v9, v40
-; SI-NEXT: v_cvt_f32_f16_e32 v10, v55
+; SI-NEXT: v_cvt_f32_f16_e32 v10, v42
+; SI-NEXT: s_waitcnt expcnt(2)
; SI-NEXT: v_cvt_f32_f16_e32 v11, v54
; SI-NEXT: v_add_f32_e32 v0, 0x38000000, v0
; SI-NEXT: v_cvt_f16_f32_e32 v0, v0
@@ -31247,8 +31264,8 @@ define inreg <13 x i64> @bitcast_v52f16_to_v13i64_scalar(<52 x half> inreg %a, i
; SI-NEXT: v_add_f32_e32 v11, 0x38000000, v11
; SI-NEXT: v_cvt_f16_f32_e32 v11, v11
; SI-NEXT: s_waitcnt expcnt(1)
-; SI-NEXT: v_cvt_f32_f16_e32 v12, v47
-; SI-NEXT: v_cvt_f32_f16_e32 v13, v60
+; SI-NEXT: v_cvt_f32_f16_e32 v12, v57
+; SI-NEXT: v_cvt_f32_f16_e32 v13, v53
; SI-NEXT: v_cvt_f32_f16_e32 v15, v52
; SI-NEXT: v_cvt_f32_f16_e32 v16, v63
; SI-NEXT: v_add_f32_e32 v12, 0x38000000, v12
@@ -31348,7 +31365,7 @@ define inreg <13 x i64> @bitcast_v52f16_to_v13i64_scalar(<52 x half> inreg %a, i
; SI-NEXT: v_cvt_f16_f32_e32 v3, v3
; SI-NEXT: v_lshlrev_b32_e32 v2, 16, v2
; SI-NEXT: v_or_b32_e32 v2, v3, v2
-; SI-NEXT: buffer_load_dword v3, off, s[0:3], s32 offset:172 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v3, off, s[0:3], s32 offset:180 ; 4-byte Folded Reload
; SI-NEXT: s_waitcnt vmcnt(0)
; SI-NEXT: v_cvt_f32_f16_e32 v3, v3
; SI-NEXT: v_add_f32_e32 v3, 0x38000000, v3
@@ -31386,7 +31403,7 @@ define inreg <13 x i64> @bitcast_v52f16_to_v13i64_scalar(<52 x half> inreg %a, i
; SI-NEXT: v_cvt_f16_f32_e32 v8, v8
; SI-NEXT: v_lshlrev_b32_e32 v8, 16, v8
; SI-NEXT: v_or_b32_e32 v8, v9, v8
-; SI-NEXT: v_cvt_f32_f16_e32 v9, v42
+; SI-NEXT: v_cvt_f32_f16_e32 v9, v55
; SI-NEXT: v_add_f32_e32 v9, 0x38000000, v9
; SI-NEXT: v_cvt_f16_f32_e32 v9, v9
; SI-NEXT: v_lshlrev_b32_e32 v9, 16, v9
@@ -31396,12 +31413,12 @@ define inreg <13 x i64> @bitcast_v52f16_to_v13i64_scalar(<52 x half> inreg %a, i
; SI-NEXT: v_cvt_f16_f32_e32 v10, v10
; SI-NEXT: v_lshlrev_b32_e32 v10, 16, v10
; SI-NEXT: v_or_b32_e32 v10, v11, v10
-; SI-NEXT: v_cvt_f32_f16_e32 v11, v43
+; SI-NEXT: v_cvt_f32_f16_e32 v11, v47
; SI-NEXT: v_add_f32_e32 v11, 0x38000000, v11
; SI-NEXT: v_cvt_f16_f32_e32 v11, v11
; SI-NEXT: v_lshlrev_b32_e32 v11, 16, v11
; SI-NEXT: v_or_b32_e32 v11, v12, v11
-; SI-NEXT: v_cvt_f32_f16_e32 v12, v57
+; SI-NEXT: v_cvt_f32_f16_e32 v12, v60
; SI-NEXT: v_add_f32_e32 v12, 0x38000000, v12
; SI-NEXT: v_cvt_f16_f32_e32 v12, v12
; SI-NEXT: v_lshlrev_b32_e32 v12, 16, v12
@@ -32144,8 +32161,8 @@ define inreg <13 x i64> @bitcast_v52f16_to_v13i64_scalar(<52 x half> inreg %a, i
; GFX11-FAKE16-NEXT: v_and_b32_e32 v50, 0xffff, v5
; GFX11-FAKE16-NEXT: v_and_b32_e32 v49, 0xffff, v6
; GFX11-FAKE16-NEXT: v_and_b32_e32 v48, 0xffff, v7
-; GFX11-FAKE16-NEXT: s_lshr_b32 s41, s29, 16
-; GFX11-FAKE16-NEXT: s_lshr_b32 s15, s28, 16
+; GFX11-FAKE16-NEXT: s_lshr_b32 s40, s29, 16
+; GFX11-FAKE16-NEXT: s_lshr_b32 s41, s28, 16
; GFX11-FAKE16-NEXT: s_lshr_b32 s42, s27, 16
; GFX11-FAKE16-NEXT: s_lshr_b32 s14, s26, 16
; GFX11-FAKE16-NEXT: s_lshr_b32 s13, s25, 16
@@ -32162,7 +32179,7 @@ define inreg <13 x i64> @bitcast_v52f16_to_v13i64_scalar(<52 x half> inreg %a, i
; GFX11-FAKE16-NEXT: s_lshr_b32 s44, s2, 16
; GFX11-FAKE16-NEXT: s_lshr_b32 s45, s1, 16
; GFX11-FAKE16-NEXT: s_lshr_b32 s46, s0, 16
-; GFX11-FAKE16-NEXT: s_mov_b32 s40, 0
+; GFX11-FAKE16-NEXT: s_mov_b32 s15, 0
; GFX11-FAKE16-NEXT: s_pack_ll_b32_b16 s0, s0, s46
; GFX11-FAKE16-NEXT: s_pack_ll_b32_b16 s1, s1, s45
; GFX11-FAKE16-NEXT: s_pack_ll_b32_b16 s2, s2, s44
@@ -32178,9 +32195,9 @@ define inreg <13 x i64> @bitcast_v52f16_to_v13i64_scalar(<52 x half> inreg %a, i
; GFX11-FAKE16-NEXT: s_pack_ll_b32_b16 s12, s24, s12
; GFX11-FAKE16-NEXT: s_pack_ll_b32_b16 s13, s25, s13
; GFX11-FAKE16-NEXT: s_pack_ll_b32_b16 s14, s26, s14
-; GFX11-FAKE16-NEXT: s_pack_ll_b32_b16 s17, s27, s42
-; GFX11-FAKE16-NEXT: s_pack_ll_b32_b16 s15, s28, s15
-; GFX11-FAKE16-NEXT: s_pack_ll_b32_b16 s16, s29, s41
+; GFX11-FAKE16-NEXT: s_pack_ll_b32_b16 s18, s27, s42
+; GFX11-FAKE16-NEXT: s_pack_ll_b32_b16 s16, s28, s41
+; GFX11-FAKE16-NEXT: s_pack_ll_b32_b16 s17, s29, s40
; GFX11-FAKE16-NEXT: s_and_b32 s47, vcc_lo, exec_lo
; GFX11-FAKE16-NEXT: s_cbranch_scc0 .LBB47_4
; GFX11-FAKE16-NEXT: ; %bb.1: ; %cmp.false
@@ -32199,9 +32216,9 @@ define inreg <13 x i64> @bitcast_v52f16_to_v13i64_scalar(<52 x half> inreg %a, i
; GFX11-FAKE16-NEXT: v_dual_mov_b32 v8, s8 :: v_dual_mov_b32 v9, s9
; GFX11-FAKE16-NEXT: v_dual_mov_b32 v10, s10 :: v_dual_mov_b32 v11, s11
; GFX11-FAKE16-NEXT: v_dual_mov_b32 v12, s12 :: v_dual_mov_b32 v13, s13
-; GFX11-FAKE16-NEXT: v_dual_mov_b32 v14, s14 :: v_dual_mov_b32 v15, s17
-; GFX11-FAKE16-NEXT: v_dual_mov_b32 v16, s15 :: v_dual_mov_b32 v17, s16
-; GFX11-FAKE16-NEXT: s_and_not1_b32 vcc_lo, exec_lo, s40
+; GFX11-FAKE16-NEXT: v_dual_mov_b32 v14, s14 :: v_dual_mov_b32 v15, s18
+; GFX11-FAKE16-NEXT: v_dual_mov_b32 v16, s16 :: v_dual_mov_b32 v17, s17
+; GFX11-FAKE16-NEXT: s_and_not1_b32 vcc_lo, exec_lo, s15
; GFX11-FAKE16-NEXT: s_cbranch_vccnz .LBB47_3
; GFX11-FAKE16-NEXT: .LBB47_2: ; %cmp.true
; GFX11-FAKE16-NEXT: v_lshl_or_b32 v18, v39, 16, v55
@@ -32227,9 +32244,9 @@ define inreg <13 x i64> @bitcast_v52f16_to_v13i64_scalar(<52 x half> inreg %a, i
; GFX11-FAKE16-NEXT: v_pk_add_f16 v12, 0x200, s12 op_sel_hi:[0,1]
; GFX11-FAKE16-NEXT: v_pk_add_f16 v13, 0x200, s13 op_sel_hi:[0,1]
; GFX11-FAKE16-NEXT: v_pk_add_f16 v14, 0x200, s14 op_sel_hi:[0,1]
-; GFX11-FAKE16-NEXT: v_pk_add_f16 v15, 0x200, s17 op_sel_hi:[0,1]
-; GFX11-FAKE16-NEXT: v_pk_add_f16 v16, 0x200, s15 op_sel_hi:[0,1]
-; GFX11-FAKE16-NEXT: v_pk_add_f16 v17, 0x200, s16 op_sel_hi:[0,1]
+; GFX11-FAKE16-NEXT: v_pk_add_f16 v15, 0x200, s18 op_sel_hi:[0,1]
+; GFX11-FAKE16-NEXT: v_pk_add_f16 v16, 0x200, s16 op_sel_hi:[0,1]
+; GFX11-FAKE16-NEXT: v_pk_add_f16 v17, 0x200, s17 op_sel_hi:[0,1]
; GFX11-FAKE16-NEXT: v_pk_add_f16 v18, 0x200, v18 op_sel_hi:[0,1]
; GFX11-FAKE16-NEXT: v_pk_add_f16 v19, 0x200, v19 op_sel_hi:[0,1]
; GFX11-FAKE16-NEXT: v_pk_add_f16 v20, 0x200, v20 op_sel_hi:[0,1]
@@ -36228,8 +36245,8 @@ define inreg <13 x double> @bitcast_v52i16_to_v13f64_scalar(<52 x i16> inreg %a,
; GFX11-FAKE16-NEXT: v_and_b32_e32 v50, 0xffff, v5
; GFX11-FAKE16-NEXT: v_and_b32_e32 v49, 0xffff, v6
; GFX11-FAKE16-NEXT: v_and_b32_e32 v48, 0xffff, v7
-; GFX11-FAKE16-NEXT: s_lshr_b32 s41, s29, 16
-; GFX11-FAKE16-NEXT: s_lshr_b32 s15, s28, 16
+; GFX11-FAKE16-NEXT: s_lshr_b32 s40, s29, 16
+; GFX11-FAKE16-NEXT: s_lshr_b32 s41, s28, 16
; GFX11-FAKE16-NEXT: s_lshr_b32 s42, s27, 16
; GFX11-FAKE16-NEXT: s_lshr_b32 s14, s26, 16
; GFX11-FAKE16-NEXT: s_lshr_b32 s13, s25, 16
@@ -36246,7 +36263,7 @@ define inreg <13 x double> @bitcast_v52i16_to_v13f64_scalar(<52 x i16> inreg %a,
; GFX11-FAKE16-NEXT: s_lshr_b32 s44, s2, 16
; GFX11-FAKE16-NEXT: s_lshr_b32 s45, s1, 16
; GFX11-FAKE16-NEXT: s_lshr_b32 s46, s0, 16
-; GFX11-FAKE16-NEXT: s_mov_b32 s40, 0
+; GFX11-FAKE16-NEXT: s_mov_b32 s15, 0
; GFX11-FAKE16-NEXT: s_pack_ll_b32_b16 s0, s0, s46
; GFX11-FAKE16-NEXT: s_pack_ll_b32_b16 s1, s1, s45
; GFX11-FAKE16-NEXT: s_pack_ll_b32_b16 s2, s2, s44
@@ -36262,9 +36279,9 @@ define inreg <13 x double> @bitcast_v52i16_to_v13f64_scalar(<52 x i16> inreg %a,
; GFX11-FAKE16-NEXT: s_pack_ll_b32_b16 s12, s24, s12
; GFX11-FAKE16-NEXT: s_pack_ll_b32_b16 s13, s25, s13
; GFX11-FAKE16-NEXT: s_pack_ll_b32_b16 s14, s26, s14
-; GFX11-FAKE16-NEXT: s_pack_ll_b32_b16 s17, s27, s42
-; GFX11-FAKE16-NEXT: s_pack_ll_b32_b16 s15, s28, s15
-; GFX11-FAKE16-NEXT: s_pack_ll_b32_b16 s16, s29, s41
+; GFX11-FAKE16-NEXT: s_pack_ll_b32_b16 s18, s27, s42
+; GFX11-FAKE16-NEXT: s_pack_ll_b32_b16 s16, s28, s41
+; GFX11-FAKE16-NEXT: s_pack_ll_b32_b16 s17, s29, s40
; GFX11-FAKE16-NEXT: s_and_b32 s47, vcc_lo, exec_lo
; GFX11-FAKE16-NEXT: s_cbranch_scc0 .LBB51_4
; GFX11-FAKE16-NEXT: ; %bb.1: ; %cmp.false
@@ -36283,9 +36300,9 @@ define inreg <13 x double> @bitcast_v52i16_to_v13f64_scalar(<52 x i16> inreg %a,
; GFX11-FAKE16-NEXT: v_dual_mov_b32 v8, s8 :: v_dual_mov_b32 v9, s9
; GFX11-FAKE16-NEXT: v_dual_mov_b32 v10, s10 :: v_dual_mov_b32 v11, s11
; GFX11-FAKE16-NEXT: v_dual_mov_b32 v12, s12 :: v_dual_mov_b32 v13, s13
-; GFX11-FAKE16-NEXT: v_dual_mov_b32 v14, s14 :: v_dual_mov_b32 v15, s17
-; GFX11-FAKE16-NEXT: v_dual_mov_b32 v16, s15 :: v_dual_mov_b32 v17, s16
-; GFX11-FAKE16-NEXT: s_and_not1_b32 vcc_lo, exec_lo, s40
+; GFX11-FAKE16-NEXT: v_dual_mov_b32 v14, s14 :: v_dual_mov_b32 v15, s18
+; GFX11-FAKE16-NEXT: v_dual_mov_b32 v16, s16 :: v_dual_mov_b32 v17, s17
+; GFX11-FAKE16-NEXT: s_and_not1_b32 vcc_lo, exec_lo, s15
; GFX11-FAKE16-NEXT: s_cbranch_vccnz .LBB51_3
; GFX11-FAKE16-NEXT: .LBB51_2: ; %cmp.true
; GFX11-FAKE16-NEXT: v_lshl_or_b32 v18, v39, 16, v55
@@ -36311,9 +36328,9 @@ define inreg <13 x double> @bitcast_v52i16_to_v13f64_scalar(<52 x i16> inreg %a,
; GFX11-FAKE16-NEXT: v_pk_add_u16 v12, s12, 3 op_sel_hi:[1,0]
; GFX11-FAKE16-NEXT: v_pk_add_u16 v13, s13, 3 op_sel_hi:[1,0]
; GFX11-FAKE16-NEXT: v_pk_add_u16 v14, s14, 3 op_sel_hi:[1,0]
-; GFX11-FAKE16-NEXT: v_pk_add_u16 v15, s17, 3 op_sel_hi:[1,0]
-; GFX11-FAKE16-NEXT: v_pk_add_u16 v16, s15, 3 op_sel_hi:[1,0]
-; GFX11-FAKE16-NEXT: v_pk_add_u16 v17, s16, 3 op_sel_hi:[1,0]
+; GFX11-FAKE16-NEXT: v_pk_add_u16 v15, s18, 3 op_sel_hi:[1,0]
+; GFX11-FAKE16-NEXT: v_pk_add_u16 v16, s16, 3 op_sel_hi:[1,0]
+; GFX11-FAKE16-NEXT: v_pk_add_u16 v17, s17, 3 op_sel_hi:[1,0]
; GFX11-FAKE16-NEXT: v_pk_add_u16 v18, v18, 3 op_sel_hi:[1,0]
; GFX11-FAKE16-NEXT: v_pk_add_u16 v19, v19, 3 op_sel_hi:[1,0]
; GFX11-FAKE16-NEXT: v_pk_add_u16 v20, v20, 3 op_sel_hi:[1,0]
@@ -37469,12 +37486,12 @@ define inreg <52 x half> @bitcast_v13f64_to_v52f16_scalar(<13 x double> inreg %a
; SI-NEXT: v_cvt_f32_f16_e32 v42, s40
; SI-NEXT: s_lshr_b32 s40, s16, 16
; SI-NEXT: v_cvt_f32_f16_e32 v44, s40
-; SI-NEXT: v_cvt_f32_f16_e32 v46, s5
+; SI-NEXT: v_cvt_f32_f16_e32 v4, s5
; SI-NEXT: v_cvt_f32_f16_e32 v3, s4
; SI-NEXT: v_cvt_f32_f16_e32 v8, s7
-; SI-NEXT: v_cvt_f32_f16_e32 v4, s6
+; SI-NEXT: v_cvt_f32_f16_e32 v10, s6
; SI-NEXT: v_cvt_f32_f16_e32 v12, s9
-; SI-NEXT: v_cvt_f32_f16_e32 v45, s8
+; SI-NEXT: v_cvt_f32_f16_e32 v13, s8
; SI-NEXT: v_cvt_f32_f16_e32 v15, s11
; SI-NEXT: v_cvt_f32_f16_e32 v17, s10
; SI-NEXT: v_cvt_f32_f16_e32 v19, s13
@@ -37500,22 +37517,24 @@ define inreg <52 x half> @bitcast_v13f64_to_v52f16_scalar(<13 x double> inreg %a
; SI-NEXT: v_add_f64 v[1:2], s[16:17], 1.0
; SI-NEXT: v_add_f64 v[54:55], s[18:19], 1.0
; SI-NEXT: v_lshrrev_b32_e32 v3, 16, v1
+; SI-NEXT: v_add_f64 v[49:50], s[20:21], 1.0
; SI-NEXT: buffer_store_dword v3, off, s[0:3], s32 offset:64 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_lshrrev_b32_e32 v3, 16, v2
; SI-NEXT: buffer_store_dword v3, off, s[0:3], s32 offset:68 ; 4-byte Folded Spill
-; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_add_f64 v[3:4], s[4:5], 1.0
-; SI-NEXT: v_cvt_f32_f16_e32 v41, v2
-; SI-NEXT: v_lshrrev_b32_e32 v10, 16, v3
-; SI-NEXT: v_cvt_f32_f16_e32 v2, v10
-; SI-NEXT: buffer_load_dword v10, off, s[0:3], s32 offset:68 ; 4-byte Folded Reload
-; SI-NEXT: v_lshrrev_b32_e32 v42, 16, v55
; SI-NEXT: v_lshrrev_b32_e32 v44, 16, v54
+; SI-NEXT: v_lshrrev_b32_e32 v42, 16, v55
+; SI-NEXT: v_lshrrev_b32_e32 v40, 16, v49
+; SI-NEXT: v_lshrrev_b32_e32 v51, 16, v50
+; SI-NEXT: v_cvt_f32_f16_e32 v52, v49
; SI-NEXT: v_cvt_f32_f16_e32 v53, v55
; SI-NEXT: v_cvt_f32_f16_e32 v55, v54
+; SI-NEXT: v_cvt_f32_f16_e32 v49, v51
+; SI-NEXT: v_cvt_f32_f16_e32 v51, v40
; SI-NEXT: v_cvt_f32_f16_e32 v54, v42
-; SI-NEXT: v_add_f64 v[49:50], s[20:21], 1.0
+; SI-NEXT: v_cvt_f32_f16_e32 v40, v44
+; SI-NEXT: buffer_load_dword v42, off, s[0:3], s32 offset:68 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v44, off, s[0:3], s32 offset:64 ; 4-byte Folded Reload
; SI-NEXT: v_add_f64 v[37:38], s[22:23], 1.0
; SI-NEXT: v_add_f64 v[33:34], s[24:25], 1.0
; SI-NEXT: v_add_f64 v[31:32], s[26:27], 1.0
@@ -37525,8 +37544,8 @@ define inreg <52 x half> @bitcast_v13f64_to_v52f16_scalar(<13 x double> inreg %a
; SI-NEXT: v_add_f64 v[14:15], s[10:11], 1.0
; SI-NEXT: v_add_f64 v[11:12], s[8:9], 1.0
; SI-NEXT: v_add_f64 v[7:8], s[6:7], 1.0
-; SI-NEXT: v_lshrrev_b32_e32 v40, 16, v49
-; SI-NEXT: v_lshrrev_b32_e32 v51, 16, v50
+; SI-NEXT: s_waitcnt expcnt(0)
+; SI-NEXT: v_add_f64 v[3:4], s[4:5], 1.0
; SI-NEXT: v_lshrrev_b32_e32 v39, 16, v37
; SI-NEXT: v_lshrrev_b32_e32 v47, 16, v38
; SI-NEXT: v_lshrrev_b32_e32 v35, 16, v33
@@ -37545,13 +37564,14 @@ define inreg <52 x half> @bitcast_v13f64_to_v52f16_scalar(<13 x double> inreg %a
; SI-NEXT: v_lshrrev_b32_e32 v63, 16, v12
; SI-NEXT: v_lshrrev_b32_e32 v5, 16, v7
; SI-NEXT: v_lshrrev_b32_e32 v6, 16, v8
-; SI-NEXT: v_lshrrev_b32_e32 v13, 16, v4
-; SI-NEXT: v_cvt_f32_f16_e32 v46, v4
+; SI-NEXT: v_lshrrev_b32_e32 v45, 16, v3
+; SI-NEXT: v_lshrrev_b32_e32 v46, 16, v4
+; SI-NEXT: v_cvt_f32_f16_e32 v4, v4
; SI-NEXT: v_cvt_f32_f16_e32 v3, v3
; SI-NEXT: v_cvt_f32_f16_e32 v8, v8
-; SI-NEXT: v_cvt_f32_f16_e32 v4, v7
+; SI-NEXT: v_cvt_f32_f16_e32 v10, v7
; SI-NEXT: v_cvt_f32_f16_e32 v12, v12
-; SI-NEXT: v_cvt_f32_f16_e32 v45, v11
+; SI-NEXT: v_cvt_f32_f16_e32 v13, v11
; SI-NEXT: v_cvt_f32_f16_e32 v15, v15
; SI-NEXT: v_cvt_f32_f16_e32 v17, v14
; SI-NEXT: v_cvt_f32_f16_e32 v19, v19
@@ -37567,9 +37587,10 @@ define inreg <52 x half> @bitcast_v13f64_to_v52f16_scalar(<13 x double> inreg %a
; SI-NEXT: v_cvt_f32_f16_e32 v38, v38
; SI-NEXT: v_cvt_f32_f16_e32 v48, v37
; SI-NEXT: v_cvt_f32_f16_e32 v50, v50
-; SI-NEXT: v_cvt_f32_f16_e32 v52, v49
+; SI-NEXT: v_cvt_f32_f16_e32 v41, v2
; SI-NEXT: v_cvt_f32_f16_e32 v43, v1
-; SI-NEXT: v_cvt_f32_f16_e32 v1, v13
+; SI-NEXT: v_cvt_f32_f16_e32 v1, v46
+; SI-NEXT: v_cvt_f32_f16_e32 v2, v45
; SI-NEXT: v_cvt_f32_f16_e32 v6, v6
; SI-NEXT: v_cvt_f32_f16_e32 v5, v5
; SI-NEXT: v_cvt_f32_f16_e32 v7, v63
@@ -37588,191 +37609,187 @@ define inreg <52 x half> @bitcast_v13f64_to_v52f16_scalar(<13 x double> inreg %a
; SI-NEXT: v_cvt_f32_f16_e32 v35, v35
; SI-NEXT: v_cvt_f32_f16_e32 v37, v47
; SI-NEXT: v_cvt_f32_f16_e32 v39, v39
-; SI-NEXT: v_cvt_f32_f16_e32 v49, v51
-; SI-NEXT: s_waitcnt vmcnt(0)
-; SI-NEXT: v_cvt_f32_f16_e32 v42, v10
-; SI-NEXT: buffer_load_dword v10, off, s[0:3], s32 offset:64 ; 4-byte Folded Reload
-; SI-NEXT: v_cvt_f32_f16_e32 v51, v40
-; SI-NEXT: v_cvt_f32_f16_e32 v40, v44
+; SI-NEXT: s_waitcnt vmcnt(1)
+; SI-NEXT: v_cvt_f32_f16_e32 v42, v42
; SI-NEXT: s_waitcnt vmcnt(0)
-; SI-NEXT: v_cvt_f32_f16_e32 v44, v10
+; SI-NEXT: v_cvt_f32_f16_e32 v44, v44
; SI-NEXT: .LBB53_3: ; %end
-; SI-NEXT: v_cvt_f16_f32_e32 v10, v44
-; SI-NEXT: v_cvt_f16_f32_e32 v13, v43
+; SI-NEXT: v_cvt_f16_f32_e32 v44, v44
+; SI-NEXT: v_cvt_f16_f32_e32 v43, v43
; SI-NEXT: v_cvt_f16_f32_e32 v42, v42
; SI-NEXT: v_cvt_f16_f32_e32 v41, v41
-; SI-NEXT: v_lshlrev_b32_e32 v10, 16, v10
-; SI-NEXT: v_or_b32_e32 v10, v13, v10
-; SI-NEXT: v_lshlrev_b32_e32 v13, 16, v42
-; SI-NEXT: buffer_store_dword v10, v0, s[0:3], 0 offen
+; SI-NEXT: v_cvt_f16_f32_e32 v40, v40
+; SI-NEXT: v_cvt_f16_f32_e32 v55, v55
+; SI-NEXT: v_cvt_f16_f32_e32 v54, v54
+; SI-NEXT: v_lshlrev_b32_e32 v44, 16, v44
+; SI-NEXT: v_cvt_f16_f32_e32 v53, v53
+; SI-NEXT: v_cvt_f16_f32_e32 v51, v51
+; SI-NEXT: v_or_b32_e32 v43, v43, v44
+; SI-NEXT: v_lshlrev_b32_e32 v42, 16, v42
+; SI-NEXT: v_cvt_f16_f32_e32 v52, v52
+; SI-NEXT: v_cvt_f16_f32_e32 v49, v49
+; SI-NEXT: buffer_store_dword v43, v0, s[0:3], 0 offen
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_add_i32_e32 v10, vcc, 4, v0
-; SI-NEXT: v_or_b32_e32 v13, v41, v13
-; SI-NEXT: buffer_store_dword v13, v10, s[0:3], 0 offen
-; SI-NEXT: v_cvt_f16_f32_e32 v10, v40
+; SI-NEXT: v_add_i32_e32 v43, vcc, 4, v0
+; SI-NEXT: v_or_b32_e32 v41, v41, v42
+; SI-NEXT: v_lshlrev_b32_e32 v40, 16, v40
+; SI-NEXT: v_cvt_f16_f32_e32 v50, v50
+; SI-NEXT: v_cvt_f16_f32_e32 v39, v39
+; SI-NEXT: buffer_store_dword v41, v43, s[0:3], 0 offen
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_cvt_f16_f32_e32 v13, v55
-; SI-NEXT: v_add_i32_e32 v55, vcc, 8, v0
-; SI-NEXT: v_lshlrev_b32_e32 v10, 16, v10
-; SI-NEXT: v_or_b32_e32 v10, v13, v10
-; SI-NEXT: buffer_store_dword v10, v55, s[0:3], 0 offen
+; SI-NEXT: v_add_i32_e32 v41, vcc, 8, v0
+; SI-NEXT: v_or_b32_e32 v55, v55, v40
+; SI-NEXT: v_lshlrev_b32_e32 v54, 16, v54
+; SI-NEXT: v_cvt_f16_f32_e32 v48, v48
+; SI-NEXT: v_cvt_f16_f32_e32 v37, v37
+; SI-NEXT: buffer_store_dword v55, v41, s[0:3], 0 offen
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_cvt_f16_f32_e32 v10, v54
-; SI-NEXT: v_cvt_f16_f32_e32 v13, v53
-; SI-NEXT: v_add_i32_e32 v53, vcc, 12, v0
-; SI-NEXT: v_lshlrev_b32_e32 v10, 16, v10
-; SI-NEXT: v_or_b32_e32 v10, v13, v10
-; SI-NEXT: buffer_store_dword v10, v53, s[0:3], 0 offen
+; SI-NEXT: v_add_i32_e32 v55, vcc, 12, v0
+; SI-NEXT: v_or_b32_e32 v53, v53, v54
+; SI-NEXT: v_lshlrev_b32_e32 v51, 16, v51
+; SI-NEXT: v_cvt_f16_f32_e32 v38, v38
+; SI-NEXT: v_cvt_f16_f32_e32 v35, v35
+; SI-NEXT: buffer_store_dword v53, v55, s[0:3], 0 offen
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_cvt_f16_f32_e32 v10, v51
-; SI-NEXT: v_cvt_f16_f32_e32 v13, v52
-; SI-NEXT: v_add_i32_e32 v51, vcc, 16, v0
-; SI-NEXT: v_lshlrev_b32_e32 v10, 16, v10
-; SI-NEXT: v_or_b32_e32 v10, v13, v10
-; SI-NEXT: buffer_store_dword v10, v51, s[0:3], 0 offen
+; SI-NEXT: v_add_i32_e32 v53, vcc, 16, v0
+; SI-NEXT: v_or_b32_e32 v51, v52, v51
+; SI-NEXT: v_lshlrev_b32_e32 v49, 16, v49
+; SI-NEXT: v_cvt_f16_f32_e32 v36, v36
+; SI-NEXT: v_cvt_f16_f32_e32 v33, v33
+; SI-NEXT: buffer_store_dword v51, v53, s[0:3], 0 offen
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_cvt_f16_f32_e32 v10, v49
-; SI-NEXT: v_cvt_f16_f32_e32 v13, v50
-; SI-NEXT: v_add_i32_e32 v49, vcc, 20, v0
-; SI-NEXT: v_lshlrev_b32_e32 v10, 16, v10
-; SI-NEXT: v_or_b32_e32 v10, v13, v10
-; SI-NEXT: buffer_store_dword v10, v49, s[0:3], 0 offen
+; SI-NEXT: v_add_i32_e32 v51, vcc, 20, v0
+; SI-NEXT: v_or_b32_e32 v49, v50, v49
+; SI-NEXT: v_lshlrev_b32_e32 v39, 16, v39
+; SI-NEXT: v_cvt_f16_f32_e32 v34, v34
+; SI-NEXT: v_cvt_f16_f32_e32 v31, v31
+; SI-NEXT: buffer_store_dword v49, v51, s[0:3], 0 offen
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_cvt_f16_f32_e32 v10, v39
-; SI-NEXT: v_cvt_f16_f32_e32 v13, v48
-; SI-NEXT: v_add_i32_e32 v39, vcc, 24, v0
-; SI-NEXT: v_lshlrev_b32_e32 v10, 16, v10
-; SI-NEXT: v_or_b32_e32 v10, v13, v10
-; SI-NEXT: buffer_store_dword v10, v39, s[0:3], 0 offen
+; SI-NEXT: v_add_i32_e32 v49, vcc, 24, v0
+; SI-NEXT: v_or_b32_e32 v39, v48, v39
+; SI-NEXT: v_lshlrev_b32_e32 v37, 16, v37
+; SI-NEXT: v_cvt_f16_f32_e32 v32, v32
+; SI-NEXT: v_cvt_f16_f32_e32 v28, v28
+; SI-NEXT: buffer_store_dword v39, v49, s[0:3], 0 offen
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_cvt_f16_f32_e32 v10, v37
-; SI-NEXT: v_cvt_f16_f32_e32 v13, v38
-; SI-NEXT: v_add_i32_e32 v37, vcc, 28, v0
-; SI-NEXT: v_lshlrev_b32_e32 v10, 16, v10
-; SI-NEXT: v_or_b32_e32 v10, v13, v10
-; SI-NEXT: buffer_store_dword v10, v37, s[0:3], 0 offen
+; SI-NEXT: v_add_i32_e32 v39, vcc, 28, v0
+; SI-NEXT: v_or_b32_e32 v37, v38, v37
+; SI-NEXT: v_lshlrev_b32_e32 v35, 16, v35
+; SI-NEXT: v_cvt_f16_f32_e32 v30, v30
+; SI-NEXT: buffer_store_dword v37, v39, s[0:3], 0 offen
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_cvt_f16_f32_e32 v10, v35
-; SI-NEXT: v_cvt_f16_f32_e32 v13, v36
-; SI-NEXT: v_add_i32_e32 v35, vcc, 32, v0
-; SI-NEXT: v_lshlrev_b32_e32 v10, 16, v10
-; SI-NEXT: v_or_b32_e32 v10, v13, v10
-; SI-NEXT: buffer_store_dword v10, v35, s[0:3], 0 offen
+; SI-NEXT: v_add_i32_e32 v37, vcc, 32, v0
+; SI-NEXT: v_or_b32_e32 v35, v36, v35
+; SI-NEXT: v_lshlrev_b32_e32 v33, 16, v33
+; SI-NEXT: buffer_store_dword v35, v37, s[0:3], 0 offen
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_cvt_f16_f32_e32 v10, v33
-; SI-NEXT: v_cvt_f16_f32_e32 v13, v34
-; SI-NEXT: v_add_i32_e32 v33, vcc, 36, v0
-; SI-NEXT: v_lshlrev_b32_e32 v10, 16, v10
-; SI-NEXT: v_or_b32_e32 v10, v13, v10
-; SI-NEXT: buffer_store_dword v10, v33, s[0:3], 0 offen
+; SI-NEXT: v_add_i32_e32 v35, vcc, 36, v0
+; SI-NEXT: v_or_b32_e32 v33, v34, v33
+; SI-NEXT: v_lshlrev_b32_e32 v31, 16, v31
+; SI-NEXT: buffer_store_dword v33, v35, s[0:3], 0 offen
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_cvt_f16_f32_e32 v10, v31
-; SI-NEXT: v_cvt_f16_f32_e32 v13, v32
-; SI-NEXT: v_add_i32_e32 v31, vcc, 40, v0
-; SI-NEXT: v_lshlrev_b32_e32 v10, 16, v10
-; SI-NEXT: v_or_b32_e32 v10, v13, v10
-; SI-NEXT: buffer_store_dword v10, v31, s[0:3], 0 offen
+; SI-NEXT: v_add_i32_e32 v33, vcc, 40, v0
+; SI-NEXT: v_or_b32_e32 v31, v32, v31
+; SI-NEXT: v_lshlrev_b32_e32 v28, 16, v28
+; SI-NEXT: buffer_store_dword v31, v33, s[0:3], 0 offen
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_cvt_f16_f32_e32 v10, v28
-; SI-NEXT: v_cvt_f16_f32_e32 v13, v30
-; SI-NEXT: v_add_i32_e32 v28, vcc, 44, v0
-; SI-NEXT: v_lshlrev_b32_e32 v10, 16, v10
-; SI-NEXT: v_or_b32_e32 v10, v13, v10
-; SI-NEXT: buffer_store_dword v10, v28, s[0:3], 0 offen
+; SI-NEXT: v_add_i32_e32 v31, vcc, 44, v0
+; SI-NEXT: v_or_b32_e32 v28, v30, v28
+; SI-NEXT: v_cvt_f16_f32_e32 v26, v26
+; SI-NEXT: buffer_store_dword v28, v31, s[0:3], 0 offen
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_cvt_f16_f32_e32 v10, v26
-; SI-NEXT: v_cvt_f16_f32_e32 v13, v29
-; SI-NEXT: v_add_i32_e32 v26, vcc, 48, v0
-; SI-NEXT: v_lshlrev_b32_e32 v10, 16, v10
-; SI-NEXT: v_or_b32_e32 v10, v13, v10
-; SI-NEXT: buffer_store_dword v10, v26, s[0:3], 0 offen
+; SI-NEXT: v_cvt_f16_f32_e32 v28, v29
+; SI-NEXT: v_add_i32_e32 v29, vcc, 48, v0
+; SI-NEXT: v_lshlrev_b32_e32 v26, 16, v26
+; SI-NEXT: v_or_b32_e32 v26, v28, v26
+; SI-NEXT: v_cvt_f16_f32_e32 v24, v24
+; SI-NEXT: buffer_store_dword v26, v29, s[0:3], 0 offen
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_cvt_f16_f32_e32 v10, v24
-; SI-NEXT: v_cvt_f16_f32_e32 v13, v27
-; SI-NEXT: v_add_i32_e32 v24, vcc, 52, v0
-; SI-NEXT: v_lshlrev_b32_e32 v10, 16, v10
-; SI-NEXT: v_or_b32_e32 v10, v13, v10
-; SI-NEXT: buffer_store_dword v10, v24, s[0:3], 0 offen
+; SI-NEXT: v_cvt_f16_f32_e32 v26, v27
+; SI-NEXT: v_add_i32_e32 v27, vcc, 52, v0
+; SI-NEXT: v_lshlrev_b32_e32 v24, 16, v24
+; SI-NEXT: v_or_b32_e32 v24, v26, v24
+; SI-NEXT: v_cvt_f16_f32_e32 v22, v22
+; SI-NEXT: buffer_store_dword v24, v27, s[0:3], 0 offen
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_cvt_f16_f32_e32 v10, v22
-; SI-NEXT: v_cvt_f16_f32_e32 v13, v25
-; SI-NEXT: v_add_i32_e32 v22, vcc, 56, v0
-; SI-NEXT: v_lshlrev_b32_e32 v10, 16, v10
-; SI-NEXT: v_or_b32_e32 v10, v13, v10
-; SI-NEXT: buffer_store_dword v10, v22, s[0:3], 0 offen
+; SI-NEXT: v_cvt_f16_f32_e32 v24, v25
+; SI-NEXT: v_add_i32_e32 v25, vcc, 56, v0
+; SI-NEXT: v_lshlrev_b32_e32 v22, 16, v22
+; SI-NEXT: v_or_b32_e32 v22, v24, v22
+; SI-NEXT: v_cvt_f16_f32_e32 v20, v20
+; SI-NEXT: buffer_store_dword v22, v25, s[0:3], 0 offen
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_cvt_f16_f32_e32 v10, v20
-; SI-NEXT: v_cvt_f16_f32_e32 v13, v23
-; SI-NEXT: v_add_i32_e32 v20, vcc, 60, v0
-; SI-NEXT: v_lshlrev_b32_e32 v10, 16, v10
-; SI-NEXT: v_or_b32_e32 v10, v13, v10
-; SI-NEXT: buffer_store_dword v10, v20, s[0:3], 0 offen
+; SI-NEXT: v_cvt_f16_f32_e32 v22, v23
+; SI-NEXT: v_add_i32_e32 v23, vcc, 60, v0
+; SI-NEXT: v_lshlrev_b32_e32 v20, 16, v20
+; SI-NEXT: v_or_b32_e32 v20, v22, v20
+; SI-NEXT: v_cvt_f16_f32_e32 v18, v18
+; SI-NEXT: buffer_store_dword v20, v23, s[0:3], 0 offen
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_cvt_f16_f32_e32 v10, v18
-; SI-NEXT: v_cvt_f16_f32_e32 v13, v21
-; SI-NEXT: v_add_i32_e32 v18, vcc, 64, v0
-; SI-NEXT: v_lshlrev_b32_e32 v10, 16, v10
-; SI-NEXT: v_or_b32_e32 v10, v13, v10
-; SI-NEXT: buffer_store_dword v10, v18, s[0:3], 0 offen
+; SI-NEXT: v_cvt_f16_f32_e32 v20, v21
+; SI-NEXT: v_add_i32_e32 v21, vcc, 64, v0
+; SI-NEXT: v_lshlrev_b32_e32 v18, 16, v18
+; SI-NEXT: v_or_b32_e32 v18, v20, v18
+; SI-NEXT: v_cvt_f16_f32_e32 v16, v16
+; SI-NEXT: buffer_store_dword v18, v21, s[0:3], 0 offen
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_cvt_f16_f32_e32 v10, v16
-; SI-NEXT: v_cvt_f16_f32_e32 v13, v19
-; SI-NEXT: v_add_i32_e32 v16, vcc, 0x44, v0
-; SI-NEXT: v_lshlrev_b32_e32 v10, 16, v10
-; SI-NEXT: v_or_b32_e32 v10, v13, v10
-; SI-NEXT: buffer_store_dword v10, v16, s[0:3], 0 offen
+; SI-NEXT: v_cvt_f16_f32_e32 v18, v19
+; SI-NEXT: v_add_i32_e32 v19, vcc, 0x44, v0
+; SI-NEXT: v_lshlrev_b32_e32 v16, 16, v16
+; SI-NEXT: v_or_b32_e32 v16, v18, v16
+; SI-NEXT: v_cvt_f16_f32_e32 v14, v14
+; SI-NEXT: buffer_store_dword v16, v19, s[0:3], 0 offen
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_cvt_f16_f32_e32 v10, v14
-; SI-NEXT: v_cvt_f16_f32_e32 v13, v17
-; SI-NEXT: v_add_i32_e32 v14, vcc, 0x48, v0
-; SI-NEXT: v_lshlrev_b32_e32 v10, 16, v10
-; SI-NEXT: v_or_b32_e32 v10, v13, v10
-; SI-NEXT: buffer_store_dword v10, v14, s[0:3], 0 offen
+; SI-NEXT: v_cvt_f16_f32_e32 v16, v17
+; SI-NEXT: v_add_i32_e32 v17, vcc, 0x48, v0
+; SI-NEXT: v_lshlrev_b32_e32 v14, 16, v14
+; SI-NEXT: v_or_b32_e32 v14, v16, v14
+; SI-NEXT: v_cvt_f16_f32_e32 v11, v11
+; SI-NEXT: buffer_store_dword v14, v17, s[0:3], 0 offen
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_cvt_f16_f32_e32 v10, v11
-; SI-NEXT: v_cvt_f16_f32_e32 v11, v15
-; SI-NEXT: v_add_i32_e32 v13, vcc, 0x4c, v0
-; SI-NEXT: v_lshlrev_b32_e32 v10, 16, v10
-; SI-NEXT: v_or_b32_e32 v10, v11, v10
+; SI-NEXT: v_cvt_f16_f32_e32 v14, v15
+; SI-NEXT: v_add_i32_e32 v15, vcc, 0x4c, v0
+; SI-NEXT: v_lshlrev_b32_e32 v11, 16, v11
+; SI-NEXT: v_or_b32_e32 v11, v14, v11
; SI-NEXT: v_cvt_f16_f32_e32 v9, v9
-; SI-NEXT: buffer_store_dword v10, v13, s[0:3], 0 offen
+; SI-NEXT: buffer_store_dword v11, v15, s[0:3], 0 offen
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_cvt_f16_f32_e32 v10, v45
-; SI-NEXT: v_add_i32_e32 v11, vcc, 0x50, v0
+; SI-NEXT: v_cvt_f16_f32_e32 v11, v13
+; SI-NEXT: v_add_i32_e32 v13, vcc, 0x50, v0
; SI-NEXT: v_lshlrev_b32_e32 v9, 16, v9
-; SI-NEXT: v_or_b32_e32 v9, v10, v9
+; SI-NEXT: v_or_b32_e32 v9, v11, v9
; SI-NEXT: v_cvt_f16_f32_e32 v7, v7
-; SI-NEXT: buffer_store_dword v9, v11, s[0:3], 0 offen
+; SI-NEXT: buffer_store_dword v9, v13, s[0:3], 0 offen
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_cvt_f16_f32_e32 v9, v12
-; SI-NEXT: v_cvt_f16_f32_e32 v5, v5
-; SI-NEXT: v_cvt_f16_f32_e32 v4, v4
+; SI-NEXT: v_add_i32_e32 v11, vcc, 0x54, v0
; SI-NEXT: v_lshlrev_b32_e32 v7, 16, v7
-; SI-NEXT: v_add_i32_e32 v10, vcc, 0x54, v0
; SI-NEXT: v_or_b32_e32 v7, v9, v7
-; SI-NEXT: v_lshlrev_b32_e32 v5, 16, v5
-; SI-NEXT: buffer_store_dword v7, v10, s[0:3], 0 offen
+; SI-NEXT: v_cvt_f16_f32_e32 v5, v5
+; SI-NEXT: buffer_store_dword v7, v11, s[0:3], 0 offen
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_add_i32_e32 v7, vcc, 0x58, v0
-; SI-NEXT: v_or_b32_e32 v4, v4, v5
-; SI-NEXT: buffer_store_dword v4, v7, s[0:3], 0 offen
+; SI-NEXT: v_cvt_f16_f32_e32 v7, v10
+; SI-NEXT: v_add_i32_e32 v9, vcc, 0x58, v0
+; SI-NEXT: v_lshlrev_b32_e32 v5, 16, v5
+; SI-NEXT: v_or_b32_e32 v5, v7, v5
+; SI-NEXT: buffer_store_dword v5, v9, s[0:3], 0 offen
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_cvt_f16_f32_e32 v4, v6
-; SI-NEXT: v_cvt_f16_f32_e32 v5, v8
+; SI-NEXT: v_cvt_f16_f32_e32 v5, v6
+; SI-NEXT: v_cvt_f16_f32_e32 v6, v8
; SI-NEXT: v_cvt_f16_f32_e32 v2, v2
; SI-NEXT: v_cvt_f16_f32_e32 v3, v3
-; SI-NEXT: v_lshlrev_b32_e32 v4, 16, v4
-; SI-NEXT: v_add_i32_e32 v6, vcc, 0x5c, v0
-; SI-NEXT: v_or_b32_e32 v4, v5, v4
+; SI-NEXT: v_lshlrev_b32_e32 v5, 16, v5
+; SI-NEXT: v_add_i32_e32 v7, vcc, 0x5c, v0
+; SI-NEXT: v_or_b32_e32 v5, v6, v5
; SI-NEXT: v_lshlrev_b32_e32 v2, 16, v2
-; SI-NEXT: buffer_store_dword v4, v6, s[0:3], 0 offen
+; SI-NEXT: buffer_store_dword v5, v7, s[0:3], 0 offen
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_add_i32_e32 v4, vcc, 0x60, v0
+; SI-NEXT: v_add_i32_e32 v5, vcc, 0x60, v0
; SI-NEXT: v_or_b32_e32 v2, v3, v2
; SI-NEXT: v_cvt_f16_f32_e32 v1, v1
-; SI-NEXT: buffer_store_dword v2, v4, s[0:3], 0 offen
+; SI-NEXT: buffer_store_dword v2, v5, s[0:3], 0 offen
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_cvt_f16_f32_e32 v2, v46
+; SI-NEXT: v_cvt_f16_f32_e32 v2, v4
; SI-NEXT: v_add_i32_e32 v0, vcc, 0x64, v0
; SI-NEXT: v_lshlrev_b32_e32 v1, 16, v1
; SI-NEXT: v_or_b32_e32 v1, v2, v1
@@ -37836,17 +37853,17 @@ define inreg <52 x half> @bitcast_v13f64_to_v52f16_scalar(<13 x double> inreg %a
; SI-NEXT: ; implicit-def: $vgpr14
; SI-NEXT: ; implicit-def: $vgpr15
; SI-NEXT: ; implicit-def: $vgpr11
-; SI-NEXT: ; implicit-def: $vgpr45
+; SI-NEXT: ; implicit-def: $vgpr13
; SI-NEXT: ; implicit-def: $vgpr9
; SI-NEXT: ; implicit-def: $vgpr12
; SI-NEXT: ; implicit-def: $vgpr7
-; SI-NEXT: ; implicit-def: $vgpr4
+; SI-NEXT: ; implicit-def: $vgpr10
; SI-NEXT: ; implicit-def: $vgpr5
; SI-NEXT: ; implicit-def: $vgpr8
; SI-NEXT: ; implicit-def: $vgpr6
; SI-NEXT: ; implicit-def: $vgpr3
; SI-NEXT: ; implicit-def: $vgpr2
-; SI-NEXT: ; implicit-def: $vgpr46
+; SI-NEXT: ; implicit-def: $vgpr4
; SI-NEXT: ; implicit-def: $vgpr1
; SI-NEXT: s_branch .LBB53_2
;
@@ -39964,15 +39981,15 @@ define inreg <13 x double> @bitcast_v52f16_to_v13f64_scalar(<52 x half> inreg %a
; SI-NEXT: buffer_store_dword v0, off, s[0:3], s32 offset:124 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_cvt_f16_f32_e32 v0, v28
-; SI-NEXT: v_cvt_f16_f32_e32 v42, v5
-; SI-NEXT: v_cvt_f16_f32_e32 v55, v4
+; SI-NEXT: v_cvt_f16_f32_e32 v55, v5
+; SI-NEXT: v_cvt_f16_f32_e32 v42, v4
; SI-NEXT: v_cvt_f16_f32_e32 v56, v7
; SI-NEXT: buffer_store_dword v0, off, s[0:3], s32 offset:120 ; 4-byte Folded Spill
; SI-NEXT: v_cvt_f16_f32_e32 v54, v6
-; SI-NEXT: v_cvt_f16_f32_e32 v43, v9
-; SI-NEXT: v_cvt_f16_f32_e32 v47, v8
-; SI-NEXT: v_cvt_f16_f32_e32 v57, v11
-; SI-NEXT: v_cvt_f16_f32_e32 v60, v10
+; SI-NEXT: v_cvt_f16_f32_e32 v47, v9
+; SI-NEXT: v_cvt_f16_f32_e32 v57, v8
+; SI-NEXT: v_cvt_f16_f32_e32 v60, v11
+; SI-NEXT: v_cvt_f16_f32_e32 v53, v10
; SI-NEXT: v_cvt_f16_f32_e32 v13, v13
; SI-NEXT: v_cvt_f16_f32_e32 v52, v12
; SI-NEXT: v_cvt_f16_f32_e32 v15, v15
@@ -39990,13 +40007,13 @@ define inreg <13 x double> @bitcast_v52f16_to_v13f64_scalar(<52 x half> inreg %a
; SI-NEXT: v_cvt_f16_f32_e32 v27, v26
; SI-NEXT: v_cvt_f16_f32_e32 v26, v29
; SI-NEXT: v_cvt_f16_f32_e32 v30, v30
-; SI-NEXT: v_cvt_f16_f32_e32 v53, s17
-; SI-NEXT: v_cvt_f16_f32_e32 v11, s16
+; SI-NEXT: v_cvt_f16_f32_e32 v43, s17
+; SI-NEXT: v_cvt_f16_f32_e32 v2, s16
; SI-NEXT: v_cvt_f16_f32_e32 v1, s19
-; SI-NEXT: v_cvt_f16_f32_e32 v2, s18
+; SI-NEXT: v_cvt_f16_f32_e32 v3, s18
; SI-NEXT: v_cvt_f16_f32_e32 v12, s21
; SI-NEXT: v_cvt_f16_f32_e32 v14, s20
-; SI-NEXT: v_cvt_f16_f32_e32 v3, s23
+; SI-NEXT: v_cvt_f16_f32_e32 v11, s23
; SI-NEXT: v_cvt_f16_f32_e32 v10, s22
; SI-NEXT: v_cvt_f16_f32_e32 v4, s25
; SI-NEXT: v_cvt_f16_f32_e32 v9, s24
@@ -40038,9 +40055,9 @@ define inreg <13 x double> @bitcast_v52f16_to_v13f64_scalar(<52 x half> inreg %a
; SI-NEXT: buffer_store_dword v4, off, s[0:3], s32 offset:160 ; 4-byte Folded Spill
; SI-NEXT: buffer_store_dword v10, off, s[0:3], s32 offset:164 ; 4-byte Folded Spill
; SI-NEXT: buffer_store_dword v2, off, s[0:3], s32 offset:168 ; 4-byte Folded Spill
-; SI-NEXT: buffer_store_dword v3, off, s[0:3], s32 offset:172 ; 4-byte Folded Spill
-; SI-NEXT: buffer_store_dword v11, off, s[0:3], s32 offset:176 ; 4-byte Folded Spill
-; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:180 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:172 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v3, off, s[0:3], s32 offset:176 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v11, off, s[0:3], s32 offset:180 ; 4-byte Folded Spill
; SI-NEXT: buffer_store_dword v12, off, s[0:3], s32 offset:184 ; 4-byte Folded Spill
; SI-NEXT: buffer_store_dword v14, off, s[0:3], s32 offset:188 ; 4-byte Folded Spill
; SI-NEXT: s_cbranch_scc0 .LBB55_4
@@ -40051,17 +40068,18 @@ define inreg <13 x double> @bitcast_v52f16_to_v13f64_scalar(<52 x half> inreg %a
; SI-NEXT: buffer_load_dword v25, off, s[0:3], s32 offset:104 ; 4-byte Folded Reload
; SI-NEXT: buffer_load_dword v22, off, s[0:3], s32 offset:120 ; 4-byte Folded Reload
; SI-NEXT: buffer_load_dword v38, off, s[0:3], s32 offset:96 ; 4-byte Folded Reload
-; SI-NEXT: s_waitcnt expcnt(2)
+; SI-NEXT: v_lshlrev_b32_e32 v0, 16, v43
+; SI-NEXT: s_waitcnt expcnt(4)
; SI-NEXT: v_lshlrev_b32_e32 v1, 16, v1
-; SI-NEXT: v_lshlrev_b32_e32 v0, 16, v53
-; SI-NEXT: v_or_b32_e32 v1, v2, v1
+; SI-NEXT: v_or_b32_e32 v0, v2, v0
+; SI-NEXT: v_or_b32_e32 v1, v3, v1
; SI-NEXT: v_lshlrev_b32_e32 v2, 16, v12
-; SI-NEXT: v_lshlrev_b32_e32 v3, 16, v3
+; SI-NEXT: s_waitcnt expcnt(3)
+; SI-NEXT: v_lshlrev_b32_e32 v3, 16, v11
; SI-NEXT: v_lshlrev_b32_e32 v4, 16, v4
; SI-NEXT: v_lshlrev_b32_e32 v5, 16, v5
; SI-NEXT: v_lshlrev_b32_e32 v6, 16, v6
; SI-NEXT: v_lshlrev_b32_e32 v21, 16, v26
-; SI-NEXT: v_or_b32_e32 v0, v11, v0
; SI-NEXT: v_or_b32_e32 v2, v14, v2
; SI-NEXT: v_or_b32_e32 v3, v10, v3
; SI-NEXT: v_or_b32_e32 v4, v9, v4
@@ -40069,11 +40087,12 @@ define inreg <13 x double> @bitcast_v52f16_to_v13f64_scalar(<52 x half> inreg %a
; SI-NEXT: v_or_b32_e32 v6, v7, v6
; SI-NEXT: v_lshlrev_b32_e32 v7, 16, v46
; SI-NEXT: v_lshlrev_b32_e32 v8, 16, v41
-; SI-NEXT: v_lshlrev_b32_e32 v9, 16, v42
+; SI-NEXT: v_lshlrev_b32_e32 v9, 16, v55
; SI-NEXT: v_lshlrev_b32_e32 v10, 16, v56
-; SI-NEXT: v_lshlrev_b32_e32 v11, 16, v43
+; SI-NEXT: s_waitcnt expcnt(2)
+; SI-NEXT: v_lshlrev_b32_e32 v11, 16, v47
; SI-NEXT: s_waitcnt expcnt(1)
-; SI-NEXT: v_lshlrev_b32_e32 v12, 16, v57
+; SI-NEXT: v_lshlrev_b32_e32 v12, 16, v60
; SI-NEXT: v_lshlrev_b32_e32 v13, 16, v13
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_lshlrev_b32_e32 v14, 16, v15
@@ -40085,10 +40104,10 @@ define inreg <13 x double> @bitcast_v52f16_to_v13f64_scalar(<52 x half> inreg %a
; SI-NEXT: v_mov_b32_e32 v51, v46
; SI-NEXT: v_or_b32_e32 v7, v45, v7
; SI-NEXT: v_or_b32_e32 v8, v40, v8
-; SI-NEXT: v_or_b32_e32 v9, v55, v9
+; SI-NEXT: v_or_b32_e32 v9, v42, v9
; SI-NEXT: v_or_b32_e32 v10, v54, v10
-; SI-NEXT: v_or_b32_e32 v11, v47, v11
-; SI-NEXT: v_or_b32_e32 v12, v60, v12
+; SI-NEXT: v_or_b32_e32 v11, v57, v11
+; SI-NEXT: v_or_b32_e32 v12, v53, v12
; SI-NEXT: v_or_b32_e32 v13, v52, v13
; SI-NEXT: v_or_b32_e32 v14, v63, v14
; SI-NEXT: v_or_b32_e32 v15, v61, v15
@@ -40116,14 +40135,16 @@ define inreg <13 x double> @bitcast_v52f16_to_v13f64_scalar(<52 x half> inreg %a
; SI-NEXT: v_or_b32_e32 v25, v38, v25
; SI-NEXT: s_cbranch_execnz .LBB55_3
; SI-NEXT: .LBB55_2: ; %cmp.true
-; SI-NEXT: s_waitcnt expcnt(2)
-; SI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:176 ; 4-byte Folded Reload
-; SI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:180 ; 4-byte Folded Reload
-; SI-NEXT: buffer_load_dword v3, off, s[0:3], s32 offset:168 ; 4-byte Folded Reload
+; SI-NEXT: s_waitcnt expcnt(4)
+; SI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:168 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:172 ; 4-byte Folded Reload
+; SI-NEXT: s_waitcnt expcnt(3)
+; SI-NEXT: buffer_load_dword v3, off, s[0:3], s32 offset:176 ; 4-byte Folded Reload
; SI-NEXT: buffer_load_dword v4, off, s[0:3], s32 offset:164 ; 4-byte Folded Reload
-; SI-NEXT: v_cvt_f32_f16_e32 v0, v53
+; SI-NEXT: v_cvt_f32_f16_e32 v0, v43
; SI-NEXT: v_cvt_f32_f16_e32 v9, v40
-; SI-NEXT: v_cvt_f32_f16_e32 v10, v55
+; SI-NEXT: v_cvt_f32_f16_e32 v10, v42
+; SI-NEXT: s_waitcnt expcnt(2)
; SI-NEXT: v_cvt_f32_f16_e32 v11, v54
; SI-NEXT: v_add_f32_e32 v0, 0x38000000, v0
; SI-NEXT: v_cvt_f16_f32_e32 v0, v0
@@ -40135,8 +40156,8 @@ define inreg <13 x double> @bitcast_v52f16_to_v13f64_scalar(<52 x half> inreg %a
; SI-NEXT: v_add_f32_e32 v11, 0x38000000, v11
; SI-NEXT: v_cvt_f16_f32_e32 v11, v11
; SI-NEXT: s_waitcnt expcnt(1)
-; SI-NEXT: v_cvt_f32_f16_e32 v12, v47
-; SI-NEXT: v_cvt_f32_f16_e32 v13, v60
+; SI-NEXT: v_cvt_f32_f16_e32 v12, v57
+; SI-NEXT: v_cvt_f32_f16_e32 v13, v53
; SI-NEXT: v_cvt_f32_f16_e32 v15, v52
; SI-NEXT: v_cvt_f32_f16_e32 v16, v63
; SI-NEXT: v_add_f32_e32 v12, 0x38000000, v12
@@ -40236,7 +40257,7 @@ define inreg <13 x double> @bitcast_v52f16_to_v13f64_scalar(<52 x half> inreg %a
; SI-NEXT: v_cvt_f16_f32_e32 v3, v3
; SI-NEXT: v_lshlrev_b32_e32 v2, 16, v2
; SI-NEXT: v_or_b32_e32 v2, v3, v2
-; SI-NEXT: buffer_load_dword v3, off, s[0:3], s32 offset:172 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v3, off, s[0:3], s32 offset:180 ; 4-byte Folded Reload
; SI-NEXT: s_waitcnt vmcnt(0)
; SI-NEXT: v_cvt_f32_f16_e32 v3, v3
; SI-NEXT: v_add_f32_e32 v3, 0x38000000, v3
@@ -40274,7 +40295,7 @@ define inreg <13 x double> @bitcast_v52f16_to_v13f64_scalar(<52 x half> inreg %a
; SI-NEXT: v_cvt_f16_f32_e32 v8, v8
; SI-NEXT: v_lshlrev_b32_e32 v8, 16, v8
; SI-NEXT: v_or_b32_e32 v8, v9, v8
-; SI-NEXT: v_cvt_f32_f16_e32 v9, v42
+; SI-NEXT: v_cvt_f32_f16_e32 v9, v55
; SI-NEXT: v_add_f32_e32 v9, 0x38000000, v9
; SI-NEXT: v_cvt_f16_f32_e32 v9, v9
; SI-NEXT: v_lshlrev_b32_e32 v9, 16, v9
@@ -40284,12 +40305,12 @@ define inreg <13 x double> @bitcast_v52f16_to_v13f64_scalar(<52 x half> inreg %a
; SI-NEXT: v_cvt_f16_f32_e32 v10, v10
; SI-NEXT: v_lshlrev_b32_e32 v10, 16, v10
; SI-NEXT: v_or_b32_e32 v10, v11, v10
-; SI-NEXT: v_cvt_f32_f16_e32 v11, v43
+; SI-NEXT: v_cvt_f32_f16_e32 v11, v47
; SI-NEXT: v_add_f32_e32 v11, 0x38000000, v11
; SI-NEXT: v_cvt_f16_f32_e32 v11, v11
; SI-NEXT: v_lshlrev_b32_e32 v11, 16, v11
; SI-NEXT: v_or_b32_e32 v11, v12, v11
-; SI-NEXT: v_cvt_f32_f16_e32 v12, v57
+; SI-NEXT: v_cvt_f32_f16_e32 v12, v60
; SI-NEXT: v_add_f32_e32 v12, 0x38000000, v12
; SI-NEXT: v_cvt_f16_f32_e32 v12, v12
; SI-NEXT: v_lshlrev_b32_e32 v12, 16, v12
@@ -41032,8 +41053,8 @@ define inreg <13 x double> @bitcast_v52f16_to_v13f64_scalar(<52 x half> inreg %a
; GFX11-FAKE16-NEXT: v_and_b32_e32 v50, 0xffff, v5
; GFX11-FAKE16-NEXT: v_and_b32_e32 v49, 0xffff, v6
; GFX11-FAKE16-NEXT: v_and_b32_e32 v48, 0xffff, v7
-; GFX11-FAKE16-NEXT: s_lshr_b32 s41, s29, 16
-; GFX11-FAKE16-NEXT: s_lshr_b32 s15, s28, 16
+; GFX11-FAKE16-NEXT: s_lshr_b32 s40, s29, 16
+; GFX11-FAKE16-NEXT: s_lshr_b32 s41, s28, 16
; GFX11-FAKE16-NEXT: s_lshr_b32 s42, s27, 16
; GFX11-FAKE16-NEXT: s_lshr_b32 s14, s26, 16
; GFX11-FAKE16-NEXT: s_lshr_b32 s13, s25, 16
@@ -41050,7 +41071,7 @@ define inreg <13 x double> @bitcast_v52f16_to_v13f64_scalar(<52 x half> inreg %a
; GFX11-FAKE16-NEXT: s_lshr_b32 s44, s2, 16
; GFX11-FAKE16-NEXT: s_lshr_b32 s45, s1, 16
; GFX11-FAKE16-NEXT: s_lshr_b32 s46, s0, 16
-; GFX11-FAKE16-NEXT: s_mov_b32 s40, 0
+; GFX11-FAKE16-NEXT: s_mov_b32 s15, 0
; GFX11-FAKE16-NEXT: s_pack_ll_b32_b16 s0, s0, s46
; GFX11-FAKE16-NEXT: s_pack_ll_b32_b16 s1, s1, s45
; GFX11-FAKE16-NEXT: s_pack_ll_b32_b16 s2, s2, s44
@@ -41066,9 +41087,9 @@ define inreg <13 x double> @bitcast_v52f16_to_v13f64_scalar(<52 x half> inreg %a
; GFX11-FAKE16-NEXT: s_pack_ll_b32_b16 s12, s24, s12
; GFX11-FAKE16-NEXT: s_pack_ll_b32_b16 s13, s25, s13
; GFX11-FAKE16-NEXT: s_pack_ll_b32_b16 s14, s26, s14
-; GFX11-FAKE16-NEXT: s_pack_ll_b32_b16 s17, s27, s42
-; GFX11-FAKE16-NEXT: s_pack_ll_b32_b16 s15, s28, s15
-; GFX11-FAKE16-NEXT: s_pack_ll_b32_b16 s16, s29, s41
+; GFX11-FAKE16-NEXT: s_pack_ll_b32_b16 s18, s27, s42
+; GFX11-FAKE16-NEXT: s_pack_ll_b32_b16 s16, s28, s41
+; GFX11-FAKE16-NEXT: s_pack_ll_b32_b16 s17, s29, s40
; GFX11-FAKE16-NEXT: s_and_b32 s47, vcc_lo, exec_lo
; GFX11-FAKE16-NEXT: s_cbranch_scc0 .LBB55_4
; GFX11-FAKE16-NEXT: ; %bb.1: ; %cmp.false
@@ -41087,9 +41108,9 @@ define inreg <13 x double> @bitcast_v52f16_to_v13f64_scalar(<52 x half> inreg %a
; GFX11-FAKE16-NEXT: v_dual_mov_b32 v8, s8 :: v_dual_mov_b32 v9, s9
; GFX11-FAKE16-NEXT: v_dual_mov_b32 v10, s10 :: v_dual_mov_b32 v11, s11
; GFX11-FAKE16-NEXT: v_dual_mov_b32 v12, s12 :: v_dual_mov_b32 v13, s13
-; GFX11-FAKE16-NEXT: v_dual_mov_b32 v14, s14 :: v_dual_mov_b32 v15, s17
-; GFX11-FAKE16-NEXT: v_dual_mov_b32 v16, s15 :: v_dual_mov_b32 v17, s16
-; GFX11-FAKE16-NEXT: s_and_not1_b32 vcc_lo, exec_lo, s40
+; GFX11-FAKE16-NEXT: v_dual_mov_b32 v14, s14 :: v_dual_mov_b32 v15, s18
+; GFX11-FAKE16-NEXT: v_dual_mov_b32 v16, s16 :: v_dual_mov_b32 v17, s17
+; GFX11-FAKE16-NEXT: s_and_not1_b32 vcc_lo, exec_lo, s15
; GFX11-FAKE16-NEXT: s_cbranch_vccnz .LBB55_3
; GFX11-FAKE16-NEXT: .LBB55_2: ; %cmp.true
; GFX11-FAKE16-NEXT: v_lshl_or_b32 v18, v39, 16, v55
@@ -41115,9 +41136,9 @@ define inreg <13 x double> @bitcast_v52f16_to_v13f64_scalar(<52 x half> inreg %a
; GFX11-FAKE16-NEXT: v_pk_add_f16 v12, 0x200, s12 op_sel_hi:[0,1]
; GFX11-FAKE16-NEXT: v_pk_add_f16 v13, 0x200, s13 op_sel_hi:[0,1]
; GFX11-FAKE16-NEXT: v_pk_add_f16 v14, 0x200, s14 op_sel_hi:[0,1]
-; GFX11-FAKE16-NEXT: v_pk_add_f16 v15, 0x200, s17 op_sel_hi:[0,1]
-; GFX11-FAKE16-NEXT: v_pk_add_f16 v16, 0x200, s15 op_sel_hi:[0,1]
-; GFX11-FAKE16-NEXT: v_pk_add_f16 v17, 0x200, s16 op_sel_hi:[0,1]
+; GFX11-FAKE16-NEXT: v_pk_add_f16 v15, 0x200, s18 op_sel_hi:[0,1]
+; GFX11-FAKE16-NEXT: v_pk_add_f16 v16, 0x200, s16 op_sel_hi:[0,1]
+; GFX11-FAKE16-NEXT: v_pk_add_f16 v17, 0x200, s17 op_sel_hi:[0,1]
; GFX11-FAKE16-NEXT: v_pk_add_f16 v18, 0x200, v18 op_sel_hi:[0,1]
; GFX11-FAKE16-NEXT: v_pk_add_f16 v19, 0x200, v19 op_sel_hi:[0,1]
; GFX11-FAKE16-NEXT: v_pk_add_f16 v20, 0x200, v20 op_sel_hi:[0,1]
@@ -42485,12 +42506,12 @@ define inreg <52 x half> @bitcast_v52i16_to_v52f16_scalar(<52 x i16> inreg %a, i
; SI-NEXT: buffer_load_dword v43, off, s[0:3], s32 offset:28
; SI-NEXT: s_waitcnt expcnt(5)
; SI-NEXT: buffer_load_dword v58, off, s[0:3], s32 offset:24
-; SI-NEXT: buffer_load_dword v44, off, s[0:3], s32 offset:20
-; SI-NEXT: buffer_load_dword v45, off, s[0:3], s32 offset:16
-; SI-NEXT: buffer_load_dword v46, off, s[0:3], s32 offset:12
-; SI-NEXT: buffer_load_dword v47, off, s[0:3], s32 offset:8
-; SI-NEXT: buffer_load_dword v56, off, s[0:3], s32 offset:4
-; SI-NEXT: buffer_load_dword v57, off, s[0:3], s32
+; SI-NEXT: buffer_load_dword v36, off, s[0:3], s32 offset:20
+; SI-NEXT: buffer_load_dword v44, off, s[0:3], s32 offset:16
+; SI-NEXT: buffer_load_dword v38, off, s[0:3], s32 offset:12
+; SI-NEXT: buffer_load_dword v32, off, s[0:3], s32 offset:8
+; SI-NEXT: buffer_load_dword v48, off, s[0:3], s32 offset:4
+; SI-NEXT: buffer_load_dword v37, off, s[0:3], s32
; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:208 ; 4-byte Folded Spill
; SI-NEXT: buffer_store_dword v2, off, s[0:3], s32 offset:212 ; 4-byte Folded Spill
; SI-NEXT: buffer_store_dword v3, off, s[0:3], s32 offset:216 ; 4-byte Folded Spill
@@ -42522,70 +42543,75 @@ define inreg <52 x half> @bitcast_v52i16_to_v52f16_scalar(<52 x i16> inreg %a, i
; SI-NEXT: buffer_store_dword v29, off, s[0:3], s32 offset:320 ; 4-byte Folded Spill
; SI-NEXT: buffer_store_dword v30, off, s[0:3], s32 offset:324 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt vmcnt(14)
-; SI-NEXT: buffer_store_dword v58, off, s[0:3], s32 offset:328 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v37, off, s[0:3], s32 offset:328 ; 4-byte Folded Spill
; SI-NEXT: v_cmp_ne_u32_e32 vcc, 0, v31
; SI-NEXT: s_and_b64 s[4:5], vcc, exec
; SI-NEXT: s_cbranch_scc0 .LBB57_2
; SI-NEXT: ; %bb.1: ; %cmp.false
; SI-NEXT: v_cvt_f32_f16_e32 v31, s22
+; SI-NEXT: v_mov_b32_e32 v46, v44
+; SI-NEXT: v_cvt_f32_f16_e32 v62, v44
+; SI-NEXT: v_mov_b32_e32 v44, v58
+; SI-NEXT: buffer_store_dword v31, off, s[0:3], s32 offset:148 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
+; SI-NEXT: v_cvt_f32_f16_e32 v31, s24
; SI-NEXT: v_cvt_f32_f16_e32 v58, v58
; SI-NEXT: v_cvt_f32_f16_e32 v33, s16
; SI-NEXT: v_cvt_f32_f16_e32 v49, s17
-; SI-NEXT: buffer_store_dword v31, off, s[0:3], s32 offset:148 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v31, off, s[0:3], s32 offset:140 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_cvt_f32_f16_e32 v31, s24
+; SI-NEXT: v_cvt_f32_f16_e32 v31, s26
; SI-NEXT: v_cvt_f32_f16_e32 v34, s18
; SI-NEXT: v_cvt_f32_f16_e32 v50, s19
; SI-NEXT: v_cvt_f32_f16_e32 v60, s20
-; SI-NEXT: buffer_store_dword v31, off, s[0:3], s32 offset:140 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v31, off, s[0:3], s32 offset:132 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_cvt_f32_f16_e32 v31, s26
+; SI-NEXT: v_cvt_f32_f16_e32 v31, s28
; SI-NEXT: v_cvt_f32_f16_e32 v35, s21
; SI-NEXT: v_cvt_f32_f16_e32 v63, s23
; SI-NEXT: v_cvt_f32_f16_e32 v61, s25
-; SI-NEXT: buffer_store_dword v31, off, s[0:3], s32 offset:132 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v31, off, s[0:3], s32 offset:124 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_cvt_f32_f16_e32 v31, s28
+; SI-NEXT: v_cvt_f32_f16_e32 v31, s29
; SI-NEXT: v_cvt_f32_f16_e32 v59, s27
; SI-NEXT: s_mov_b64 s[4:5], 0
; SI-NEXT: v_cvt_f32_f16_e32 v42, v19
-; SI-NEXT: buffer_store_dword v31, off, s[0:3], s32 offset:124 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v31, off, s[0:3], s32 offset:144 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_cvt_f32_f16_e32 v31, s29
+; SI-NEXT: v_cvt_f32_f16_e32 v31, v1
; SI-NEXT: v_cvt_f32_f16_e32 v41, v21
; SI-NEXT: v_cvt_f32_f16_e32 v55, v23
; SI-NEXT: v_cvt_f32_f16_e32 v53, v25
-; SI-NEXT: buffer_store_dword v31, off, s[0:3], s32 offset:144 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v31, off, s[0:3], s32 offset:116 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_cvt_f32_f16_e32 v31, v1
+; SI-NEXT: v_cvt_f32_f16_e32 v31, v2
; SI-NEXT: v_cvt_f32_f16_e32 v40, v26
; SI-NEXT: v_cvt_f32_f16_e32 v51, v27
; SI-NEXT: v_cvt_f32_f16_e32 v54, v28
-; SI-NEXT: buffer_store_dword v31, off, s[0:3], s32 offset:116 ; 4-byte Folded Spill
-; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_cvt_f32_f16_e32 v31, v2
-; SI-NEXT: v_cvt_f32_f16_e32 v39, v29
-; SI-NEXT: v_cvt_f32_f16_e32 v52, v30
-; SI-NEXT: v_cvt_f32_f16_e32 v37, v57
; SI-NEXT: buffer_store_dword v31, off, s[0:3], s32 offset:136 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_cvt_f32_f16_e32 v31, v3
-; SI-NEXT: v_cvt_f32_f16_e32 v48, v56
-; SI-NEXT: v_cvt_f32_f16_e32 v32, v47
-; SI-NEXT: v_cvt_f32_f16_e32 v38, v46
+; SI-NEXT: v_cvt_f32_f16_e32 v39, v29
+; SI-NEXT: v_cvt_f32_f16_e32 v52, v30
+; SI-NEXT: v_cvt_f32_f16_e32 v37, v37
; SI-NEXT: buffer_store_dword v31, off, s[0:3], s32 offset:108 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_cvt_f32_f16_e32 v31, v4
-; SI-NEXT: v_cvt_f32_f16_e32 v62, v45
-; SI-NEXT: v_cvt_f32_f16_e32 v36, v44
-; SI-NEXT: buffer_store_dword v58, off, s[0:3], s32 offset:152 ; 4-byte Folded Spill
+; SI-NEXT: v_mov_b32_e32 v57, v48
+; SI-NEXT: v_cvt_f32_f16_e32 v48, v48
+; SI-NEXT: v_mov_b32_e32 v56, v32
; SI-NEXT: buffer_store_dword v31, off, s[0:3], s32 offset:128 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_cvt_f32_f16_e32 v31, v5
+; SI-NEXT: v_cvt_f32_f16_e32 v32, v32
+; SI-NEXT: v_mov_b32_e32 v47, v38
+; SI-NEXT: v_cvt_f32_f16_e32 v38, v38
; SI-NEXT: buffer_store_dword v31, off, s[0:3], s32 offset:104 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_cvt_f32_f16_e32 v31, v6
+; SI-NEXT: v_mov_b32_e32 v45, v36
+; SI-NEXT: v_cvt_f32_f16_e32 v36, v36
+; SI-NEXT: buffer_store_dword v58, off, s[0:3], s32 offset:152 ; 4-byte Folded Spill
; SI-NEXT: buffer_store_dword v31, off, s[0:3], s32 offset:120 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_cvt_f32_f16_e32 v31, v7
@@ -42638,6 +42664,24 @@ define inreg <52 x half> @bitcast_v52i16_to_v52f16_scalar(<52 x i16> inreg %a, i
; SI-NEXT: .LBB57_2:
; SI-NEXT: ; implicit-def: $vgpr31
; SI-NEXT: ; kill: killed $vgpr31
+; SI-NEXT: v_mov_b32_e32 v57, v48
+; SI-NEXT: ; implicit-def: $vgpr31
+; SI-NEXT: ; kill: killed $vgpr31
+; SI-NEXT: v_mov_b32_e32 v56, v32
+; SI-NEXT: ; implicit-def: $vgpr31
+; SI-NEXT: ; kill: killed $vgpr31
+; SI-NEXT: v_mov_b32_e32 v47, v38
+; SI-NEXT: ; implicit-def: $vgpr31
+; SI-NEXT: ; kill: killed $vgpr31
+; SI-NEXT: v_mov_b32_e32 v46, v44
+; SI-NEXT: ; implicit-def: $vgpr31
+; SI-NEXT: ; kill: killed $vgpr31
+; SI-NEXT: v_mov_b32_e32 v45, v36
+; SI-NEXT: ; implicit-def: $vgpr31
+; SI-NEXT: ; kill: killed $vgpr31
+; SI-NEXT: v_mov_b32_e32 v44, v58
+; SI-NEXT: ; implicit-def: $vgpr31
+; SI-NEXT: ; kill: killed $vgpr31
; SI-NEXT: s_mov_b64 s[4:5], -1
; SI-NEXT: ; implicit-def: $vgpr31
; SI-NEXT: ; kill: killed $vgpr31
@@ -42704,24 +42748,12 @@ define inreg <52 x half> @bitcast_v52i16_to_v52f16_scalar(<52 x i16> inreg %a, i
; SI-NEXT: ; implicit-def: $vgpr31
; SI-NEXT: ; kill: killed $vgpr31
; SI-NEXT: ; implicit-def: $vgpr31
-; SI-NEXT: ; kill: killed $vgpr31
-; SI-NEXT: ; implicit-def: $vgpr31
-; SI-NEXT: ; kill: killed $vgpr31
-; SI-NEXT: ; implicit-def: $vgpr31
-; SI-NEXT: ; kill: killed $vgpr31
-; SI-NEXT: ; implicit-def: $vgpr31
-; SI-NEXT: ; kill: killed $vgpr31
-; SI-NEXT: ; implicit-def: $vgpr31
-; SI-NEXT: ; kill: killed $vgpr31
-; SI-NEXT: ; implicit-def: $vgpr31
-; SI-NEXT: ; kill: killed $vgpr31
-; SI-NEXT: ; implicit-def: $vgpr31
; SI-NEXT: .LBB57_3: ; %Flow
; SI-NEXT: s_andn2_b64 vcc, exec, s[4:5]
-; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_mov_b32_e32 v58, v62
; SI-NEXT: v_mov_b32_e32 v62, v32
; SI-NEXT: v_mov_b32_e32 v32, v37
+; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_mov_b32_e32 v37, v39
; SI-NEXT: v_mov_b32_e32 v39, v51
; SI-NEXT: v_mov_b32_e32 v51, v53
@@ -42732,6 +42764,12 @@ define inreg <52 x half> @bitcast_v52i16_to_v52f16_scalar(<52 x i16> inreg %a, i
; SI-NEXT: ; %bb.4: ; %cmp.true
; SI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:328 ; 4-byte Folded Reload
; SI-NEXT: v_add_i32_e32 v42, vcc, 3, v43
+; SI-NEXT: v_add_i32_e32 v43, vcc, 3, v44
+; SI-NEXT: v_add_i32_e32 v44, vcc, 3, v45
+; SI-NEXT: v_add_i32_e32 v45, vcc, 3, v46
+; SI-NEXT: v_add_i32_e32 v46, vcc, 3, v47
+; SI-NEXT: v_add_i32_e32 v47, vcc, 3, v56
+; SI-NEXT: v_add_i32_e32 v56, vcc, 3, v57
; SI-NEXT: s_add_i32 s22, s22, 3
; SI-NEXT: v_cvt_f32_f16_e32 v31, s22
; SI-NEXT: s_add_i32 s24, s24, 3
@@ -42741,29 +42779,23 @@ define inreg <52 x half> @bitcast_v52i16_to_v52f16_scalar(<52 x i16> inreg %a, i
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_cvt_f32_f16_e32 v31, s24
; SI-NEXT: s_add_i32 s29, s29, 3
-; SI-NEXT: v_add_i32_e32 v44, vcc, 3, v44
+; SI-NEXT: s_add_i32 s27, s27, 3
+; SI-NEXT: s_add_i32 s25, s25, 3
; SI-NEXT: buffer_store_dword v31, off, s[0:3], s32 offset:140 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_cvt_f32_f16_e32 v31, s26
-; SI-NEXT: v_add_i32_e32 v45, vcc, 3, v45
-; SI-NEXT: v_add_i32_e32 v46, vcc, 3, v46
-; SI-NEXT: buffer_store_dword v31, off, s[0:3], s32 offset:132 ; 4-byte Folded Spill
-; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_cvt_f32_f16_e32 v31, s28
-; SI-NEXT: v_add_i32_e32 v47, vcc, 3, v47
-; SI-NEXT: v_add_i32_e32 v56, vcc, 3, v56
-; SI-NEXT: buffer_store_dword v31, off, s[0:3], s32 offset:124 ; 4-byte Folded Spill
-; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_cvt_f32_f16_e32 v31, s29
-; SI-NEXT: v_add_i32_e32 v57, vcc, 3, v57
-; SI-NEXT: s_add_i32 s27, s27, 3
-; SI-NEXT: s_add_i32 s25, s25, 3
; SI-NEXT: s_add_i32 s23, s23, 3
; SI-NEXT: s_add_i32 s21, s21, 3
; SI-NEXT: s_add_i32 s20, s20, 3
+; SI-NEXT: buffer_store_dword v31, off, s[0:3], s32 offset:132 ; 4-byte Folded Spill
+; SI-NEXT: s_waitcnt expcnt(0)
+; SI-NEXT: v_cvt_f32_f16_e32 v31, s28
; SI-NEXT: s_add_i32 s19, s19, 3
; SI-NEXT: s_add_i32 s18, s18, 3
; SI-NEXT: s_add_i32 s17, s17, 3
+; SI-NEXT: buffer_store_dword v31, off, s[0:3], s32 offset:124 ; 4-byte Folded Spill
+; SI-NEXT: s_waitcnt expcnt(0)
+; SI-NEXT: v_cvt_f32_f16_e32 v31, s29
; SI-NEXT: s_add_i32 s16, s16, 3
; SI-NEXT: v_cvt_f32_f16_e32 v33, s16
; SI-NEXT: v_cvt_f32_f16_e32 v49, s17
@@ -42775,7 +42807,6 @@ define inreg <52 x half> @bitcast_v52i16_to_v52f16_scalar(<52 x i16> inreg %a, i
; SI-NEXT: v_cvt_f32_f16_e32 v61, s25
; SI-NEXT: v_cvt_f32_f16_e32 v59, s27
; SI-NEXT: buffer_store_dword v31, off, s[0:3], s32 offset:144 ; 4-byte Folded Spill
-; SI-NEXT: v_cvt_f32_f16_e32 v32, v57
; SI-NEXT: v_cvt_f32_f16_e32 v48, v56
; SI-NEXT: v_cvt_f32_f16_e32 v62, v47
; SI-NEXT: v_cvt_f32_f16_e32 v38, v46
@@ -42784,8 +42815,9 @@ define inreg <52 x half> @bitcast_v52i16_to_v52f16_scalar(<52 x i16> inreg %a, i
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_cvt_f32_f16_e32 v31, v42
; SI-NEXT: s_waitcnt vmcnt(5)
-; SI-NEXT: v_add_i32_e32 v43, vcc, 3, v1
+; SI-NEXT: v_add_i32_e32 v57, vcc, 3, v1
; SI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:324 ; 4-byte Folded Reload
+; SI-NEXT: v_cvt_f32_f16_e32 v32, v57
; SI-NEXT: s_waitcnt vmcnt(0)
; SI-NEXT: v_add_i32_e32 v30, vcc, 3, v1
; SI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:320 ; 4-byte Folded Reload
@@ -43651,13 +43683,16 @@ define inreg <52 x half> @bitcast_v52i16_to_v52f16_scalar(<52 x i16> inreg %a, i
; GFX11-TRUE16-NEXT: v_lshl_or_b32 v5, v10, 16, v5
; GFX11-TRUE16-NEXT: v_mov_b16_e32 v9.l, v2.h
; GFX11-TRUE16-NEXT: v_and_b32_e32 v2, 0xffff, v2
-; GFX11-TRUE16-NEXT: v_lshl_or_b32 v4, v8, 16, v4
-; GFX11-TRUE16-NEXT: v_mov_b16_e32 v8.l, v3.h
-; GFX11-TRUE16-NEXT: v_and_b32_e32 v3, 0xffff, v3
; GFX11-TRUE16-NEXT: v_mov_b16_e32 v10.l, v1.h
; GFX11-TRUE16-NEXT: v_and_b32_e32 v1, 0xffff, v1
; GFX11-TRUE16-NEXT: v_mov_b16_e32 v11.l, v0.h
; GFX11-TRUE16-NEXT: v_and_b32_e32 v0, 0xffff, v0
+; GFX11-TRUE16-NEXT: v_lshl_or_b32 v4, v8, 16, v4
+; GFX11-TRUE16-NEXT: v_mov_b16_e32 v8.l, v3.h
+; GFX11-TRUE16-NEXT: v_and_b32_e32 v3, 0xffff, v3
+; GFX11-TRUE16-NEXT: v_lshl_or_b32 v2, v9, 16, v2
+; GFX11-TRUE16-NEXT: v_lshl_or_b32 v1, v10, 16, v1
+; GFX11-TRUE16-NEXT: v_lshl_or_b32 v0, v11, 16, v0
; GFX11-TRUE16-NEXT: s_pack_ll_b32_b16 s15, s23, s15
; GFX11-TRUE16-NEXT: s_pack_ll_b32_b16 s14, s22, s14
; GFX11-TRUE16-NEXT: v_lshl_or_b32 v3, v8, 16, v3
@@ -43665,9 +43700,9 @@ define inreg <52 x half> @bitcast_v52i16_to_v52f16_scalar(<52 x i16> inreg %a, i
; GFX11-TRUE16-NEXT: s_mov_b32 s15, s21
; GFX11-TRUE16-NEXT: v_pk_add_u16 v8, s14, 3 op_sel_hi:[1,0]
; GFX11-TRUE16-NEXT: s_mov_b32 s14, s20
-; GFX11-TRUE16-NEXT: v_lshl_or_b32 v2, v9, 16, v2
-; GFX11-TRUE16-NEXT: v_lshl_or_b32 v1, v10, 16, v1
-; GFX11-TRUE16-NEXT: v_lshl_or_b32 v0, v11, 16, v0
+; GFX11-TRUE16-NEXT: v_pk_add_u16 v2, v2, 3 op_sel_hi:[1,0]
+; GFX11-TRUE16-NEXT: v_pk_add_u16 v1, v1, 3 op_sel_hi:[1,0]
+; GFX11-TRUE16-NEXT: v_pk_add_u16 v0, v0, 3 op_sel_hi:[1,0]
; GFX11-TRUE16-NEXT: s_pack_ll_b32_b16 s12, s15, s12
; GFX11-TRUE16-NEXT: s_pack_ll_b32_b16 s11, s14, s11
; GFX11-TRUE16-NEXT: v_pk_add_u16 v9, s12, 3 op_sel_hi:[1,0]
@@ -43681,9 +43716,9 @@ define inreg <52 x half> @bitcast_v52i16_to_v52f16_scalar(<52 x i16> inreg %a, i
; GFX11-TRUE16-NEXT: v_pk_add_u16 v5, v5, 3 op_sel_hi:[1,0]
; GFX11-TRUE16-NEXT: v_pk_add_u16 v4, v4, 3 op_sel_hi:[1,0]
; GFX11-TRUE16-NEXT: v_pk_add_u16 v3, v3, 3 op_sel_hi:[1,0]
-; GFX11-TRUE16-NEXT: v_pk_add_u16 v2, v2, 3 op_sel_hi:[1,0]
-; GFX11-TRUE16-NEXT: v_pk_add_u16 v1, v1, 3 op_sel_hi:[1,0]
-; GFX11-TRUE16-NEXT: v_pk_add_u16 v0, v0, 3 op_sel_hi:[1,0]
+; GFX11-TRUE16-NEXT: v_lshrrev_b32_e32 v52, 16, v0
+; GFX11-TRUE16-NEXT: v_lshrrev_b32_e32 v53, 16, v1
+; GFX11-TRUE16-NEXT: v_lshrrev_b32_e32 v54, 16, v2
; GFX11-TRUE16-NEXT: v_pk_add_u16 v18, s8, 3 op_sel_hi:[1,0]
; GFX11-TRUE16-NEXT: s_mov_b32 s8, s17
; GFX11-TRUE16-NEXT: v_pk_add_u16 v19, s7, 3 op_sel_hi:[1,0]
@@ -43712,14 +43747,14 @@ define inreg <52 x half> @bitcast_v52i16_to_v52f16_scalar(<52 x i16> inreg %a, i
; GFX11-TRUE16-NEXT: v_pk_add_u16 v28, s2, 3 op_sel_hi:[1,0]
; GFX11-TRUE16-NEXT: v_pk_add_u16 v27, s3, 3 op_sel_hi:[1,0]
; GFX11-TRUE16-NEXT: v_pk_add_u16 v26, s5, 3 op_sel_hi:[1,0]
-; GFX11-TRUE16-NEXT: v_lshrrev_b32_e32 v52, 16, v0
-; GFX11-TRUE16-NEXT: v_lshrrev_b32_e32 v53, 16, v1
-; GFX11-TRUE16-NEXT: v_lshrrev_b32_e32 v54, 16, v2
-; GFX11-TRUE16-NEXT: v_lshrrev_b32_e32 v55, 16, v3
-; GFX11-TRUE16-NEXT: v_lshrrev_b32_e32 v64, 16, v4
-; GFX11-TRUE16-NEXT: v_lshrrev_b32_e32 v65, 16, v5
-; GFX11-TRUE16-NEXT: v_lshrrev_b32_e32 v66, 16, v6
-; GFX11-TRUE16-NEXT: v_lshrrev_b32_e32 v67, 16, v7
+; GFX11-TRUE16-NEXT: v_mov_b16_e32 v0.h, v52.l
+; GFX11-TRUE16-NEXT: v_mov_b16_e32 v1.h, v53.l
+; GFX11-TRUE16-NEXT: v_mov_b16_e32 v2.h, v54.l
+; GFX11-TRUE16-NEXT: v_lshrrev_b32_e32 v52, 16, v3
+; GFX11-TRUE16-NEXT: v_lshrrev_b32_e32 v53, 16, v4
+; GFX11-TRUE16-NEXT: v_lshrrev_b32_e32 v54, 16, v5
+; GFX11-TRUE16-NEXT: v_lshrrev_b32_e32 v55, 16, v6
+; GFX11-TRUE16-NEXT: v_lshrrev_b32_e32 v64, 16, v7
; GFX11-TRUE16-NEXT: v_lshrrev_b32_e32 v50, 16, v32
; GFX11-TRUE16-NEXT: v_lshrrev_b32_e32 v33, 16, v29
; GFX11-TRUE16-NEXT: v_lshrrev_b32_e32 v39, 16, v28
@@ -43736,16 +43771,13 @@ define inreg <52 x half> @bitcast_v52i16_to_v52f16_scalar(<52 x i16> inreg %a, i
; GFX11-TRUE16-NEXT: v_lshrrev_b32_e32 v25, 16, v13
; GFX11-TRUE16-NEXT: v_lshrrev_b32_e32 v34, 16, v12
; GFX11-TRUE16-NEXT: v_lshrrev_b32_e32 v35, 16, v11
-; GFX11-TRUE16-NEXT: v_lshrrev_b32_e32 v21, 16, v17
-; GFX11-TRUE16-NEXT: v_lshrrev_b32_e32 v22, 16, v16
-; GFX11-TRUE16-NEXT: v_mov_b16_e32 v0.h, v52.l
-; GFX11-TRUE16-NEXT: v_mov_b16_e32 v1.h, v53.l
-; GFX11-TRUE16-NEXT: v_mov_b16_e32 v2.h, v54.l
-; GFX11-TRUE16-NEXT: v_mov_b16_e32 v3.h, v55.l
-; GFX11-TRUE16-NEXT: v_mov_b16_e32 v4.h, v64.l
-; GFX11-TRUE16-NEXT: v_mov_b16_e32 v5.h, v65.l
-; GFX11-TRUE16-NEXT: v_mov_b16_e32 v6.h, v66.l
-; GFX11-TRUE16-NEXT: v_mov_b16_e32 v7.h, v67.l
+; GFX11-TRUE16-NEXT: v_lshrrev_b32_e32 v22, 16, v17
+; GFX11-TRUE16-NEXT: v_lshrrev_b32_e32 v21, 16, v16
+; GFX11-TRUE16-NEXT: v_mov_b16_e32 v3.h, v52.l
+; GFX11-TRUE16-NEXT: v_mov_b16_e32 v4.h, v53.l
+; GFX11-TRUE16-NEXT: v_mov_b16_e32 v5.h, v54.l
+; GFX11-TRUE16-NEXT: v_mov_b16_e32 v6.h, v55.l
+; GFX11-TRUE16-NEXT: v_mov_b16_e32 v7.h, v64.l
; GFX11-TRUE16-NEXT: s_branch .LBB57_5
; GFX11-TRUE16-NEXT: .LBB57_3:
; GFX11-TRUE16-NEXT: s_branch .LBB57_2
@@ -43768,8 +43800,8 @@ define inreg <52 x half> @bitcast_v52i16_to_v52f16_scalar(<52 x i16> inreg %a, i
; GFX11-TRUE16-NEXT: v_mov_b16_e32 v28.l, s2
; GFX11-TRUE16-NEXT: v_mov_b16_e32 v29.l, s1
; GFX11-TRUE16-NEXT: v_mov_b16_e32 v32.l, s0
-; GFX11-TRUE16-NEXT: v_mov_b16_e32 v22.l, s45
-; GFX11-TRUE16-NEXT: v_mov_b16_e32 v21.l, s44
+; GFX11-TRUE16-NEXT: v_mov_b16_e32 v21.l, s45
+; GFX11-TRUE16-NEXT: v_mov_b16_e32 v22.l, s44
; GFX11-TRUE16-NEXT: v_mov_b16_e32 v35.l, s43
; GFX11-TRUE16-NEXT: v_mov_b16_e32 v34.l, s42
; GFX11-TRUE16-NEXT: v_mov_b16_e32 v25.l, s41
@@ -43826,8 +43858,8 @@ define inreg <52 x half> @bitcast_v52i16_to_v52f16_scalar(<52 x i16> inreg %a, i
; GFX11-TRUE16-NEXT: v_and_b32_e32 v52, 0xffff, v26
; GFX11-TRUE16-NEXT: v_lshl_or_b32 v27, v48, 16, v27
; GFX11-TRUE16-NEXT: v_and_b32_e32 v28, 0xffff, v28
-; GFX11-TRUE16-NEXT: v_lshl_or_b32 v16, v21, 16, v17
-; GFX11-TRUE16-NEXT: v_lshl_or_b32 v17, v22, 16, v18
+; GFX11-TRUE16-NEXT: v_lshl_or_b32 v16, v22, 16, v17
+; GFX11-TRUE16-NEXT: v_lshl_or_b32 v17, v21, 16, v18
; GFX11-TRUE16-NEXT: v_lshl_or_b32 v18, v19, 16, v0
; GFX11-TRUE16-NEXT: v_lshl_or_b32 v19, v20, 16, v1
; GFX11-TRUE16-NEXT: v_lshl_or_b32 v20, v23, 16, v2
@@ -45215,7 +45247,7 @@ define inreg <52 x i16> @bitcast_v52f16_to_v52i16_scalar(<52 x half> inreg %a, i
; SI-NEXT: v_cvt_f16_f32_e32 v5, v56
; SI-NEXT: s_waitcnt vmcnt(1)
; SI-NEXT: v_cvt_f16_f32_e32 v1, v57
-; SI-NEXT: v_cvt_f16_f32_e32 v44, s17
+; SI-NEXT: v_cvt_f16_f32_e32 v45, s17
; SI-NEXT: v_cvt_f16_f32_e32 v43, s21
; SI-NEXT: v_cvt_f16_f32_e32 v42, s25
; SI-NEXT: s_waitcnt vmcnt(0)
@@ -45225,25 +45257,26 @@ define inreg <52 x i16> @bitcast_v52f16_to_v52i16_scalar(<52 x half> inreg %a, i
; SI-NEXT: ; %bb.1: ; %cmp.false
; SI-NEXT: s_cbranch_execnz .LBB59_3
; SI-NEXT: .LBB59_2: ; %cmp.true
-; SI-NEXT: v_cvt_f32_f16_e32 v44, v44
+; SI-NEXT: v_cvt_f32_f16_e32 v44, v45
; SI-NEXT: v_cvt_f32_f16_e32 v43, v43
; SI-NEXT: v_cvt_f32_f16_e32 v42, v42
; SI-NEXT: v_cvt_f32_f16_e32 v41, v41
; SI-NEXT: v_add_f32_e32 v44, 0x38000000, v44
-; SI-NEXT: v_cvt_f16_f32_e32 v44, v44
; SI-NEXT: v_add_f32_e32 v43, 0x38000000, v43
-; SI-NEXT: v_add_f32_e32 v42, 0x38000000, v42
+; SI-NEXT: v_cvt_f16_f32_e32 v44, v44
; SI-NEXT: v_cvt_f16_f32_e32 v45, v43
-; SI-NEXT: v_lshlrev_b32_e32 v43, 16, v44
+; SI-NEXT: v_add_f32_e32 v42, 0x38000000, v42
; SI-NEXT: v_cvt_f16_f32_e32 v42, v42
-; SI-NEXT: v_cvt_f32_f16_e32 v44, v30
+; SI-NEXT: v_lshlrev_b32_e32 v43, 16, v44
+; SI-NEXT: v_lshlrev_b32_e32 v44, 16, v45
+; SI-NEXT: v_cvt_f32_f16_e32 v45, v30
+; SI-NEXT: v_lshlrev_b32_e32 v30, 16, v42
; SI-NEXT: v_cvt_f32_f16_e32 v40, v40
; SI-NEXT: v_add_f32_e32 v41, 0x38000000, v41
-; SI-NEXT: v_lshlrev_b32_e32 v30, 16, v42
-; SI-NEXT: v_add_f32_e32 v42, 0x38000000, v44
+; SI-NEXT: v_add_f32_e32 v42, 0x38000000, v45
; SI-NEXT: v_cvt_f16_f32_e32 v42, v42
; SI-NEXT: v_add_f32_e32 v40, 0x38000000, v40
-; SI-NEXT: v_cvt_f16_f32_e32 v44, v41
+; SI-NEXT: v_cvt_f16_f32_e32 v45, v41
; SI-NEXT: v_cvt_f16_f32_e32 v40, v40
; SI-NEXT: v_lshlrev_b32_e32 v41, 16, v42
; SI-NEXT: v_cvt_f32_f16_e32 v42, v55
@@ -45415,13 +45448,12 @@ define inreg <52 x i16> @bitcast_v52f16_to_v52i16_scalar(<52 x half> inreg %a, i
; SI-NEXT: v_lshlrev_b32_e32 v29, 16, v22
; SI-NEXT: v_or_b32_e32 v25, v25, v29
; SI-NEXT: v_lshlrev_b32_e32 v29, 16, v23
-; SI-NEXT: v_lshlrev_b32_e32 v45, 16, v45
-; SI-NEXT: v_lshlrev_b32_e32 v46, 16, v44
+; SI-NEXT: v_lshlrev_b32_e32 v46, 16, v45
; SI-NEXT: v_lshlrev_b32_e32 v47, 16, v42
; SI-NEXT: v_lshlrev_b32_e32 v56, 16, v40
; SI-NEXT: v_or_b32_e32 v24, v24, v29
; SI-NEXT: v_or_b32_e32 v27, v27, v43
-; SI-NEXT: v_or_b32_e32 v26, v26, v45
+; SI-NEXT: v_or_b32_e32 v26, v26, v44
; SI-NEXT: v_or_b32_e32 v21, v21, v30
; SI-NEXT: v_or_b32_e32 v20, v20, v41
; SI-NEXT: v_or_b32_e32 v49, v49, v46
@@ -45433,8 +45465,8 @@ define inreg <52 x i16> @bitcast_v52f16_to_v52i16_scalar(<52 x half> inreg %a, i
; SI-NEXT: v_or_b32_e32 v11, v11, v56
; SI-NEXT: v_or_b32_e32 v6, v6, v28
; SI-NEXT: v_or_b32_e32 v4, v4, v57
-; SI-NEXT: v_alignbit_b32 v44, v24, v43, 16
-; SI-NEXT: v_alignbit_b32 v43, v25, v45, 16
+; SI-NEXT: v_alignbit_b32 v45, v24, v43, 16
+; SI-NEXT: v_alignbit_b32 v43, v25, v44, 16
; SI-NEXT: v_alignbit_b32 v42, v19, v30, 16
; SI-NEXT: v_alignbit_b32 v30, v50, v41, 16
; SI-NEXT: v_alignbit_b32 v41, v48, v46, 16
@@ -45448,7 +45480,7 @@ define inreg <52 x i16> @bitcast_v52f16_to_v52i16_scalar(<52 x half> inreg %a, i
; SI-NEXT: v_alignbit_b32 v28, v5, v57, 16
; SI-NEXT: .LBB59_3: ; %end
; SI-NEXT: v_and_b32_e32 v27, 0xffff, v27
-; SI-NEXT: v_lshlrev_b32_e32 v44, 16, v44
+; SI-NEXT: v_lshlrev_b32_e32 v44, 16, v45
; SI-NEXT: v_and_b32_e32 v24, 0xffff, v24
; SI-NEXT: v_lshlrev_b32_e32 v23, 16, v23
; SI-NEXT: v_or_b32_e32 v27, v27, v44
@@ -46085,13 +46117,16 @@ define inreg <52 x i16> @bitcast_v52f16_to_v52i16_scalar(<52 x half> inreg %a, i
; GFX11-TRUE16-NEXT: v_lshl_or_b32 v5, v10, 16, v5
; GFX11-TRUE16-NEXT: v_mov_b16_e32 v9.l, v2.h
; GFX11-TRUE16-NEXT: v_and_b32_e32 v2, 0xffff, v2
-; GFX11-TRUE16-NEXT: v_lshl_or_b32 v4, v8, 16, v4
-; GFX11-TRUE16-NEXT: v_mov_b16_e32 v8.l, v3.h
-; GFX11-TRUE16-NEXT: v_and_b32_e32 v3, 0xffff, v3
; GFX11-TRUE16-NEXT: v_mov_b16_e32 v10.l, v1.h
; GFX11-TRUE16-NEXT: v_and_b32_e32 v1, 0xffff, v1
; GFX11-TRUE16-NEXT: v_mov_b16_e32 v11.l, v0.h
; GFX11-TRUE16-NEXT: v_and_b32_e32 v0, 0xffff, v0
+; GFX11-TRUE16-NEXT: v_lshl_or_b32 v4, v8, 16, v4
+; GFX11-TRUE16-NEXT: v_mov_b16_e32 v8.l, v3.h
+; GFX11-TRUE16-NEXT: v_and_b32_e32 v3, 0xffff, v3
+; GFX11-TRUE16-NEXT: v_lshl_or_b32 v2, v9, 16, v2
+; GFX11-TRUE16-NEXT: v_lshl_or_b32 v1, v10, 16, v1
+; GFX11-TRUE16-NEXT: v_lshl_or_b32 v0, v11, 16, v0
; GFX11-TRUE16-NEXT: s_pack_ll_b32_b16 s15, s23, s15
; GFX11-TRUE16-NEXT: s_pack_ll_b32_b16 s14, s22, s14
; GFX11-TRUE16-NEXT: v_lshl_or_b32 v3, v8, 16, v3
@@ -46099,9 +46134,9 @@ define inreg <52 x i16> @bitcast_v52f16_to_v52i16_scalar(<52 x half> inreg %a, i
; GFX11-TRUE16-NEXT: s_mov_b32 s15, s21
; GFX11-TRUE16-NEXT: v_pk_add_f16 v8, 0x200, s14 op_sel_hi:[0,1]
; GFX11-TRUE16-NEXT: s_mov_b32 s14, s20
-; GFX11-TRUE16-NEXT: v_lshl_or_b32 v2, v9, 16, v2
-; GFX11-TRUE16-NEXT: v_lshl_or_b32 v1, v10, 16, v1
-; GFX11-TRUE16-NEXT: v_lshl_or_b32 v0, v11, 16, v0
+; GFX11-TRUE16-NEXT: v_pk_add_f16 v2, 0x200, v2 op_sel_hi:[0,1]
+; GFX11-TRUE16-NEXT: v_pk_add_f16 v1, 0x200, v1 op_sel_hi:[0,1]
+; GFX11-TRUE16-NEXT: v_pk_add_f16 v0, 0x200, v0 op_sel_hi:[0,1]
; GFX11-TRUE16-NEXT: s_pack_ll_b32_b16 s12, s15, s12
; GFX11-TRUE16-NEXT: s_pack_ll_b32_b16 s11, s14, s11
; GFX11-TRUE16-NEXT: v_pk_add_f16 v9, 0x200, s12 op_sel_hi:[0,1]
@@ -46115,9 +46150,9 @@ define inreg <52 x i16> @bitcast_v52f16_to_v52i16_scalar(<52 x half> inreg %a, i
; GFX11-TRUE16-NEXT: v_pk_add_f16 v5, 0x200, v5 op_sel_hi:[0,1]
; GFX11-TRUE16-NEXT: v_pk_add_f16 v4, 0x200, v4 op_sel_hi:[0,1]
; GFX11-TRUE16-NEXT: v_pk_add_f16 v3, 0x200, v3 op_sel_hi:[0,1]
-; GFX11-TRUE16-NEXT: v_pk_add_f16 v2, 0x200, v2 op_sel_hi:[0,1]
-; GFX11-TRUE16-NEXT: v_pk_add_f16 v1, 0x200, v1 op_sel_hi:[0,1]
-; GFX11-TRUE16-NEXT: v_pk_add_f16 v0, 0x200, v0 op_sel_hi:[0,1]
+; GFX11-TRUE16-NEXT: v_lshrrev_b32_e32 v52, 16, v0
+; GFX11-TRUE16-NEXT: v_lshrrev_b32_e32 v53, 16, v1
+; GFX11-TRUE16-NEXT: v_lshrrev_b32_e32 v54, 16, v2
; GFX11-TRUE16-NEXT: v_pk_add_f16 v18, 0x200, s8 op_sel_hi:[0,1]
; GFX11-TRUE16-NEXT: s_mov_b32 s8, s17
; GFX11-TRUE16-NEXT: v_pk_add_f16 v19, 0x200, s7 op_sel_hi:[0,1]
@@ -46146,14 +46181,14 @@ define inreg <52 x i16> @bitcast_v52f16_to_v52i16_scalar(<52 x half> inreg %a, i
; GFX11-TRUE16-NEXT: v_pk_add_f16 v28, 0x200, s2 op_sel_hi:[0,1]
; GFX11-TRUE16-NEXT: v_pk_add_f16 v27, 0x200, s3 op_sel_hi:[0,1]
; GFX11-TRUE16-NEXT: v_pk_add_f16 v26, 0x200, s5 op_sel_hi:[0,1]
-; GFX11-TRUE16-NEXT: v_lshrrev_b32_e32 v52, 16, v0
-; GFX11-TRUE16-NEXT: v_lshrrev_b32_e32 v53, 16, v1
-; GFX11-TRUE16-NEXT: v_lshrrev_b32_e32 v54, 16, v2
-; GFX11-TRUE16-NEXT: v_lshrrev_b32_e32 v55, 16, v3
-; GFX11-TRUE16-NEXT: v_lshrrev_b32_e32 v64, 16, v4
-; GFX11-TRUE16-NEXT: v_lshrrev_b32_e32 v65, 16, v5
-; GFX11-TRUE16-NEXT: v_lshrrev_b32_e32 v66, 16, v6
-; GFX11-TRUE16-NEXT: v_lshrrev_b32_e32 v67, 16, v7
+; GFX11-TRUE16-NEXT: v_mov_b16_e32 v0.h, v52.l
+; GFX11-TRUE16-NEXT: v_mov_b16_e32 v1.h, v53.l
+; GFX11-TRUE16-NEXT: v_mov_b16_e32 v2.h, v54.l
+; GFX11-TRUE16-NEXT: v_lshrrev_b32_e32 v52, 16, v3
+; GFX11-TRUE16-NEXT: v_lshrrev_b32_e32 v53, 16, v4
+; GFX11-TRUE16-NEXT: v_lshrrev_b32_e32 v54, 16, v5
+; GFX11-TRUE16-NEXT: v_lshrrev_b32_e32 v55, 16, v6
+; GFX11-TRUE16-NEXT: v_lshrrev_b32_e32 v64, 16, v7
; GFX11-TRUE16-NEXT: v_lshrrev_b32_e32 v50, 16, v32
; GFX11-TRUE16-NEXT: v_lshrrev_b32_e32 v33, 16, v29
; GFX11-TRUE16-NEXT: v_lshrrev_b32_e32 v39, 16, v28
@@ -46170,16 +46205,13 @@ define inreg <52 x i16> @bitcast_v52f16_to_v52i16_scalar(<52 x half> inreg %a, i
; GFX11-TRUE16-NEXT: v_lshrrev_b32_e32 v25, 16, v13
; GFX11-TRUE16-NEXT: v_lshrrev_b32_e32 v34, 16, v12
; GFX11-TRUE16-NEXT: v_lshrrev_b32_e32 v35, 16, v11
-; GFX11-TRUE16-NEXT: v_lshrrev_b32_e32 v21, 16, v17
-; GFX11-TRUE16-NEXT: v_lshrrev_b32_e32 v22, 16, v16
-; GFX11-TRUE16-NEXT: v_mov_b16_e32 v0.h, v52.l
-; GFX11-TRUE16-NEXT: v_mov_b16_e32 v1.h, v53.l
-; GFX11-TRUE16-NEXT: v_mov_b16_e32 v2.h, v54.l
-; GFX11-TRUE16-NEXT: v_mov_b16_e32 v3.h, v55.l
-; GFX11-TRUE16-NEXT: v_mov_b16_e32 v4.h, v64.l
-; GFX11-TRUE16-NEXT: v_mov_b16_e32 v5.h, v65.l
-; GFX11-TRUE16-NEXT: v_mov_b16_e32 v6.h, v66.l
-; GFX11-TRUE16-NEXT: v_mov_b16_e32 v7.h, v67.l
+; GFX11-TRUE16-NEXT: v_lshrrev_b32_e32 v22, 16, v17
+; GFX11-TRUE16-NEXT: v_lshrrev_b32_e32 v21, 16, v16
+; GFX11-TRUE16-NEXT: v_mov_b16_e32 v3.h, v52.l
+; GFX11-TRUE16-NEXT: v_mov_b16_e32 v4.h, v53.l
+; GFX11-TRUE16-NEXT: v_mov_b16_e32 v5.h, v54.l
+; GFX11-TRUE16-NEXT: v_mov_b16_e32 v6.h, v55.l
+; GFX11-TRUE16-NEXT: v_mov_b16_e32 v7.h, v64.l
; GFX11-TRUE16-NEXT: s_branch .LBB59_5
; GFX11-TRUE16-NEXT: .LBB59_3:
; GFX11-TRUE16-NEXT: s_branch .LBB59_2
@@ -46202,8 +46234,8 @@ define inreg <52 x i16> @bitcast_v52f16_to_v52i16_scalar(<52 x half> inreg %a, i
; GFX11-TRUE16-NEXT: v_mov_b16_e32 v28.l, s2
; GFX11-TRUE16-NEXT: v_mov_b16_e32 v29.l, s1
; GFX11-TRUE16-NEXT: v_mov_b16_e32 v32.l, s0
-; GFX11-TRUE16-NEXT: v_mov_b16_e32 v22.l, s45
-; GFX11-TRUE16-NEXT: v_mov_b16_e32 v21.l, s44
+; GFX11-TRUE16-NEXT: v_mov_b16_e32 v21.l, s45
+; GFX11-TRUE16-NEXT: v_mov_b16_e32 v22.l, s44
; GFX11-TRUE16-NEXT: v_mov_b16_e32 v35.l, s43
; GFX11-TRUE16-NEXT: v_mov_b16_e32 v34.l, s42
; GFX11-TRUE16-NEXT: v_mov_b16_e32 v25.l, s41
@@ -46260,8 +46292,8 @@ define inreg <52 x i16> @bitcast_v52f16_to_v52i16_scalar(<52 x half> inreg %a, i
; GFX11-TRUE16-NEXT: v_and_b32_e32 v52, 0xffff, v26
; GFX11-TRUE16-NEXT: v_lshl_or_b32 v27, v48, 16, v27
; GFX11-TRUE16-NEXT: v_and_b32_e32 v28, 0xffff, v28
-; GFX11-TRUE16-NEXT: v_lshl_or_b32 v16, v21, 16, v17
-; GFX11-TRUE16-NEXT: v_lshl_or_b32 v17, v22, 16, v18
+; GFX11-TRUE16-NEXT: v_lshl_or_b32 v16, v22, 16, v17
+; GFX11-TRUE16-NEXT: v_lshl_or_b32 v17, v21, 16, v18
; GFX11-TRUE16-NEXT: v_lshl_or_b32 v18, v19, 16, v0
; GFX11-TRUE16-NEXT: v_lshl_or_b32 v19, v20, 16, v1
; GFX11-TRUE16-NEXT: v_lshl_or_b32 v20, v23, 16, v2
diff --git a/llvm/test/CodeGen/AMDGPU/amdgcn.bitcast.896bit.ll b/llvm/test/CodeGen/AMDGPU/amdgcn.bitcast.896bit.ll
index 462e50ac8412c..6c2d2b33a04b1 100644
--- a/llvm/test/CodeGen/AMDGPU/amdgcn.bitcast.896bit.ll
+++ b/llvm/test/CodeGen/AMDGPU/amdgcn.bitcast.896bit.ll
@@ -4447,7 +4447,7 @@ define inreg <56 x i16> @bitcast_v28i32_to_v56i16_scalar(<28 x i32> inreg %a, i3
; GFX11-TRUE16-NEXT: v_readfirstlane_b32 s5, v7
; GFX11-TRUE16-NEXT: v_readfirstlane_b32 s6, v8
; GFX11-TRUE16-NEXT: v_readfirstlane_b32 s4, v9
-; GFX11-TRUE16-NEXT: s_mov_b32 s90, 0
+; GFX11-TRUE16-NEXT: s_mov_b32 s78, 0
; GFX11-TRUE16-NEXT: s_and_b32 s14, vcc_lo, exec_lo
; GFX11-TRUE16-NEXT: s_cbranch_scc0 .LBB13_4
; GFX11-TRUE16-NEXT: ; %bb.1: ; %cmp.false
@@ -4475,11 +4475,11 @@ define inreg <56 x i16> @bitcast_v28i32_to_v56i16_scalar(<28 x i32> inreg %a, i3
; GFX11-TRUE16-NEXT: s_lshr_b32 s75, s18, 16
; GFX11-TRUE16-NEXT: s_lshr_b32 s76, s17, 16
; GFX11-TRUE16-NEXT: s_lshr_b32 s77, s16, 16
-; GFX11-TRUE16-NEXT: s_lshr_b32 s78, s3, 16
-; GFX11-TRUE16-NEXT: s_lshr_b32 s79, s2, 16
-; GFX11-TRUE16-NEXT: s_lshr_b32 s88, s1, 16
-; GFX11-TRUE16-NEXT: s_lshr_b32 s89, s0, 16
-; GFX11-TRUE16-NEXT: s_and_not1_b32 vcc_lo, exec_lo, s90
+; GFX11-TRUE16-NEXT: s_lshr_b32 s79, s3, 16
+; GFX11-TRUE16-NEXT: s_lshr_b32 s88, s2, 16
+; GFX11-TRUE16-NEXT: s_lshr_b32 s89, s1, 16
+; GFX11-TRUE16-NEXT: s_lshr_b32 s90, s0, 16
+; GFX11-TRUE16-NEXT: s_and_not1_b32 vcc_lo, exec_lo, s78
; GFX11-TRUE16-NEXT: s_cbranch_vccnz .LBB13_3
; GFX11-TRUE16-NEXT: .LBB13_2: ; %cmp.true
; GFX11-TRUE16-NEXT: s_add_i32 s4, s4, 3
@@ -4534,16 +4534,18 @@ define inreg <56 x i16> @bitcast_v28i32_to_v56i16_scalar(<28 x i32> inreg %a, i3
; GFX11-TRUE16-NEXT: s_lshr_b32 s75, s18, 16
; GFX11-TRUE16-NEXT: s_lshr_b32 s76, s17, 16
; GFX11-TRUE16-NEXT: s_lshr_b32 s77, s16, 16
-; GFX11-TRUE16-NEXT: s_lshr_b32 s78, s3, 16
-; GFX11-TRUE16-NEXT: s_lshr_b32 s79, s2, 16
-; GFX11-TRUE16-NEXT: s_lshr_b32 s88, s1, 16
-; GFX11-TRUE16-NEXT: s_lshr_b32 s89, s0, 16
+; GFX11-TRUE16-NEXT: s_lshr_b32 s79, s3, 16
+; GFX11-TRUE16-NEXT: s_lshr_b32 s88, s2, 16
+; GFX11-TRUE16-NEXT: s_lshr_b32 s89, s1, 16
+; GFX11-TRUE16-NEXT: s_lshr_b32 s90, s0, 16
; GFX11-TRUE16-NEXT: .LBB13_3: ; %end
; GFX11-TRUE16-NEXT: s_delay_alu instid0(SALU_CYCLE_1)
-; GFX11-TRUE16-NEXT: s_pack_ll_b32_b16 s0, s0, s89
-; GFX11-TRUE16-NEXT: s_pack_ll_b32_b16 s1, s1, s88
-; GFX11-TRUE16-NEXT: s_pack_ll_b32_b16 s2, s2, s79
-; GFX11-TRUE16-NEXT: s_pack_ll_b32_b16 s3, s3, s78
+; GFX11-TRUE16-NEXT: s_mov_b32 s78, s90
+; GFX11-TRUE16-NEXT: s_pack_ll_b32_b16 s1, s1, s89
+; GFX11-TRUE16-NEXT: s_pack_ll_b32_b16 s0, s0, s78
+; GFX11-TRUE16-NEXT: s_mov_b32 s78, s88
+; GFX11-TRUE16-NEXT: s_pack_ll_b32_b16 s3, s3, s79
+; GFX11-TRUE16-NEXT: s_pack_ll_b32_b16 s2, s2, s78
; GFX11-TRUE16-NEXT: s_pack_ll_b32_b16 s16, s16, s77
; GFX11-TRUE16-NEXT: s_pack_ll_b32_b16 s17, s17, s76
; GFX11-TRUE16-NEXT: s_pack_ll_b32_b16 s18, s18, s75
@@ -4584,10 +4586,10 @@ define inreg <56 x i16> @bitcast_v28i32_to_v56i16_scalar(<28 x i32> inreg %a, i3
; GFX11-TRUE16-NEXT: v_dual_mov_b32 v26, s6 :: v_dual_mov_b32 v27, s4
; GFX11-TRUE16-NEXT: s_setpc_b64 s[30:31]
; GFX11-TRUE16-NEXT: .LBB13_4:
+; GFX11-TRUE16-NEXT: ; implicit-def: $sgpr90_lo16
; GFX11-TRUE16-NEXT: ; implicit-def: $sgpr89_lo16
; GFX11-TRUE16-NEXT: ; implicit-def: $sgpr88_lo16
; GFX11-TRUE16-NEXT: ; implicit-def: $sgpr79_lo16
-; GFX11-TRUE16-NEXT: ; implicit-def: $sgpr78_lo16
; GFX11-TRUE16-NEXT: ; implicit-def: $sgpr77_lo16
; GFX11-TRUE16-NEXT: ; implicit-def: $sgpr76_lo16
; GFX11-TRUE16-NEXT: ; implicit-def: $sgpr75_lo16
@@ -5632,8 +5634,8 @@ define <28 x i32> @bitcast_v56i16_to_v28i32(<56 x i16> %a, i32 %b) {
; GFX9-NEXT: v_lshrrev_b32_e32 v61, 16, v46
; GFX9-NEXT: v_lshrrev_b32_e32 v33, 16, v47
; GFX9-NEXT: v_lshrrev_b32_e32 v34, 16, v56
-; GFX9-NEXT: v_lshrrev_b32_e32 v62, 16, v57
-; GFX9-NEXT: v_lshrrev_b32_e32 v35, 16, v58
+; GFX9-NEXT: v_lshrrev_b32_e32 v35, 16, v57
+; GFX9-NEXT: v_lshrrev_b32_e32 v62, 16, v58
; GFX9-NEXT: v_lshrrev_b32_e32 v36, 16, v59
; GFX9-NEXT: v_cmp_ne_u32_e32 vcc, 0, v28
; GFX9-NEXT: buffer_store_dword v0, off, s[0:3], s32 offset:156 ; 4-byte Folded Spill
@@ -5660,8 +5662,8 @@ define <28 x i32> @bitcast_v56i16_to_v28i32(<56 x i16> %a, i32 %b) {
; GFX9-NEXT: v_lshrrev_b32_e32 v21, 16, v38
; GFX9-NEXT: v_lshrrev_b32_e32 v22, 16, v37
; GFX9-NEXT: v_perm_b32 v0, v36, v59, s6
-; GFX9-NEXT: v_perm_b32 v1, v35, v58, s6
-; GFX9-NEXT: v_perm_b32 v2, v62, v57, s6
+; GFX9-NEXT: v_perm_b32 v1, v62, v58, s6
+; GFX9-NEXT: v_perm_b32 v2, v35, v57, s6
; GFX9-NEXT: v_perm_b32 v3, v34, v56, s6
; GFX9-NEXT: v_perm_b32 v4, v33, v47, s6
; GFX9-NEXT: v_perm_b32 v5, v61, v46, s6
@@ -5711,8 +5713,8 @@ define <28 x i32> @bitcast_v56i16_to_v28i32(<56 x i16> %a, i32 %b) {
; GFX9-NEXT: ; implicit-def: $vgpr61
; GFX9-NEXT: ; implicit-def: $vgpr33
; GFX9-NEXT: ; implicit-def: $vgpr34
-; GFX9-NEXT: ; implicit-def: $vgpr62
; GFX9-NEXT: ; implicit-def: $vgpr35
+; GFX9-NEXT: ; implicit-def: $vgpr62
; GFX9-NEXT: ; implicit-def: $vgpr36
; GFX9-NEXT: s_waitcnt vmcnt(0)
; GFX9-NEXT: v_perm_b32 v23, v24, v23, s6
@@ -5802,8 +5804,8 @@ define <28 x i32> @bitcast_v56i16_to_v28i32(<56 x i16> %a, i32 %b) {
; GFX9-NEXT: buffer_load_dword v24, off, s[0:3], s32 offset:100 ; 4-byte Folded Reload
; GFX9-NEXT: s_mov_b32 s6, 0x5040100
; GFX9-NEXT: v_perm_b32 v0, v36, v59, s6
-; GFX9-NEXT: v_perm_b32 v1, v35, v58, s6
-; GFX9-NEXT: v_perm_b32 v2, v62, v57, s6
+; GFX9-NEXT: v_perm_b32 v1, v62, v58, s6
+; GFX9-NEXT: v_perm_b32 v2, v35, v57, s6
; GFX9-NEXT: v_perm_b32 v3, v34, v56, s6
; GFX9-NEXT: v_perm_b32 v4, v33, v47, s6
; GFX9-NEXT: v_perm_b32 v5, v61, v46, s6
@@ -9480,7 +9482,7 @@ define inreg <56 x half> @bitcast_v28i32_to_v56f16_scalar(<28 x i32> inreg %a, i
; GFX11-TRUE16-NEXT: v_readfirstlane_b32 s5, v7
; GFX11-TRUE16-NEXT: v_readfirstlane_b32 s6, v8
; GFX11-TRUE16-NEXT: v_readfirstlane_b32 s4, v9
-; GFX11-TRUE16-NEXT: s_mov_b32 s90, 0
+; GFX11-TRUE16-NEXT: s_mov_b32 s78, 0
; GFX11-TRUE16-NEXT: s_and_b32 s14, vcc_lo, exec_lo
; GFX11-TRUE16-NEXT: s_cbranch_scc0 .LBB17_4
; GFX11-TRUE16-NEXT: ; %bb.1: ; %cmp.false
@@ -9508,11 +9510,11 @@ define inreg <56 x half> @bitcast_v28i32_to_v56f16_scalar(<28 x i32> inreg %a, i
; GFX11-TRUE16-NEXT: s_lshr_b32 s75, s18, 16
; GFX11-TRUE16-NEXT: s_lshr_b32 s76, s17, 16
; GFX11-TRUE16-NEXT: s_lshr_b32 s77, s16, 16
-; GFX11-TRUE16-NEXT: s_lshr_b32 s78, s3, 16
-; GFX11-TRUE16-NEXT: s_lshr_b32 s79, s2, 16
-; GFX11-TRUE16-NEXT: s_lshr_b32 s88, s1, 16
-; GFX11-TRUE16-NEXT: s_lshr_b32 s89, s0, 16
-; GFX11-TRUE16-NEXT: s_and_not1_b32 vcc_lo, exec_lo, s90
+; GFX11-TRUE16-NEXT: s_lshr_b32 s79, s3, 16
+; GFX11-TRUE16-NEXT: s_lshr_b32 s88, s2, 16
+; GFX11-TRUE16-NEXT: s_lshr_b32 s89, s1, 16
+; GFX11-TRUE16-NEXT: s_lshr_b32 s90, s0, 16
+; GFX11-TRUE16-NEXT: s_and_not1_b32 vcc_lo, exec_lo, s78
; GFX11-TRUE16-NEXT: s_cbranch_vccnz .LBB17_3
; GFX11-TRUE16-NEXT: .LBB17_2: ; %cmp.true
; GFX11-TRUE16-NEXT: s_add_i32 s4, s4, 3
@@ -9567,16 +9569,18 @@ define inreg <56 x half> @bitcast_v28i32_to_v56f16_scalar(<28 x i32> inreg %a, i
; GFX11-TRUE16-NEXT: s_lshr_b32 s75, s18, 16
; GFX11-TRUE16-NEXT: s_lshr_b32 s76, s17, 16
; GFX11-TRUE16-NEXT: s_lshr_b32 s77, s16, 16
-; GFX11-TRUE16-NEXT: s_lshr_b32 s78, s3, 16
-; GFX11-TRUE16-NEXT: s_lshr_b32 s79, s2, 16
-; GFX11-TRUE16-NEXT: s_lshr_b32 s88, s1, 16
-; GFX11-TRUE16-NEXT: s_lshr_b32 s89, s0, 16
+; GFX11-TRUE16-NEXT: s_lshr_b32 s79, s3, 16
+; GFX11-TRUE16-NEXT: s_lshr_b32 s88, s2, 16
+; GFX11-TRUE16-NEXT: s_lshr_b32 s89, s1, 16
+; GFX11-TRUE16-NEXT: s_lshr_b32 s90, s0, 16
; GFX11-TRUE16-NEXT: .LBB17_3: ; %end
; GFX11-TRUE16-NEXT: s_delay_alu instid0(SALU_CYCLE_1)
-; GFX11-TRUE16-NEXT: s_pack_ll_b32_b16 s0, s0, s89
-; GFX11-TRUE16-NEXT: s_pack_ll_b32_b16 s1, s1, s88
-; GFX11-TRUE16-NEXT: s_pack_ll_b32_b16 s2, s2, s79
-; GFX11-TRUE16-NEXT: s_pack_ll_b32_b16 s3, s3, s78
+; GFX11-TRUE16-NEXT: s_mov_b32 s78, s90
+; GFX11-TRUE16-NEXT: s_pack_ll_b32_b16 s1, s1, s89
+; GFX11-TRUE16-NEXT: s_pack_ll_b32_b16 s0, s0, s78
+; GFX11-TRUE16-NEXT: s_mov_b32 s78, s88
+; GFX11-TRUE16-NEXT: s_pack_ll_b32_b16 s3, s3, s79
+; GFX11-TRUE16-NEXT: s_pack_ll_b32_b16 s2, s2, s78
; GFX11-TRUE16-NEXT: s_pack_ll_b32_b16 s16, s16, s77
; GFX11-TRUE16-NEXT: s_pack_ll_b32_b16 s17, s17, s76
; GFX11-TRUE16-NEXT: s_pack_ll_b32_b16 s18, s18, s75
@@ -9617,10 +9621,10 @@ define inreg <56 x half> @bitcast_v28i32_to_v56f16_scalar(<28 x i32> inreg %a, i
; GFX11-TRUE16-NEXT: v_dual_mov_b32 v26, s6 :: v_dual_mov_b32 v27, s4
; GFX11-TRUE16-NEXT: s_setpc_b64 s[30:31]
; GFX11-TRUE16-NEXT: .LBB17_4:
+; GFX11-TRUE16-NEXT: ; implicit-def: $sgpr90_lo16
; GFX11-TRUE16-NEXT: ; implicit-def: $sgpr89_lo16
; GFX11-TRUE16-NEXT: ; implicit-def: $sgpr88_lo16
; GFX11-TRUE16-NEXT: ; implicit-def: $sgpr79_lo16
-; GFX11-TRUE16-NEXT: ; implicit-def: $sgpr78_lo16
; GFX11-TRUE16-NEXT: ; implicit-def: $sgpr77_lo16
; GFX11-TRUE16-NEXT: ; implicit-def: $sgpr76_lo16
; GFX11-TRUE16-NEXT: ; implicit-def: $sgpr75_lo16
@@ -11338,15 +11342,15 @@ define inreg <28 x i32> @bitcast_v56f16_to_v28i32_scalar(<56 x half> inreg %a, i
; SI-NEXT: buffer_store_dword v0, off, s[0:3], s32 offset:180 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_cvt_f16_f32_e32 v0, v24
-; SI-NEXT: v_cvt_f16_f32_e32 v63, v11
-; SI-NEXT: v_cvt_f16_f32_e32 v38, v10
+; SI-NEXT: v_cvt_f16_f32_e32 v39, v11
+; SI-NEXT: v_cvt_f16_f32_e32 v33, v10
; SI-NEXT: v_cvt_f16_f32_e32 v13, v13
; SI-NEXT: buffer_store_dword v0, off, s[0:3], s32 offset:176 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_cvt_f16_f32_e32 v0, v27
-; SI-NEXT: v_cvt_f16_f32_e32 v36, v12
+; SI-NEXT: v_cvt_f16_f32_e32 v38, v12
; SI-NEXT: v_cvt_f16_f32_e32 v34, v15
-; SI-NEXT: v_cvt_f16_f32_e32 v35, v14
+; SI-NEXT: v_cvt_f16_f32_e32 v36, v14
; SI-NEXT: buffer_store_dword v0, off, s[0:3], s32 offset:172 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_cvt_f16_f32_e32 v0, v26
@@ -11439,9 +11443,9 @@ define inreg <28 x i32> @bitcast_v56f16_to_v28i32_scalar(<56 x half> inreg %a, i
; SI-NEXT: buffer_load_dword v23, off, s[0:3], s32 offset:156 ; 4-byte Folded Reload
; SI-NEXT: buffer_load_dword v24, off, s[0:3], s32 offset:148 ; 4-byte Folded Reload
; SI-NEXT: buffer_load_dword v25, off, s[0:3], s32 offset:136 ; 4-byte Folded Reload
-; SI-NEXT: v_mov_b32_e32 v49, v2
+; SI-NEXT: v_mov_b32_e32 v63, v2
; SI-NEXT: v_lshlrev_b32_e32 v2, 16, v2
-; SI-NEXT: v_mov_b32_e32 v48, v3
+; SI-NEXT: v_mov_b32_e32 v49, v3
; SI-NEXT: v_lshlrev_b32_e32 v3, 16, v3
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_lshlrev_b32_e32 v4, 16, v4
@@ -11450,9 +11454,9 @@ define inreg <28 x i32> @bitcast_v56f16_to_v28i32_scalar(<56 x half> inreg %a, i
; SI-NEXT: v_mov_b32_e32 v61, v44
; SI-NEXT: v_lshlrev_b32_e32 v0, 16, v43
; SI-NEXT: v_lshlrev_b32_e32 v1, 16, v54
-; SI-NEXT: v_mov_b32_e32 v39, v11
+; SI-NEXT: v_mov_b32_e32 v48, v11
; SI-NEXT: v_or_b32_e32 v2, v11, v2
-; SI-NEXT: v_mov_b32_e32 v33, v10
+; SI-NEXT: v_mov_b32_e32 v35, v10
; SI-NEXT: v_or_b32_e32 v3, v10, v3
; SI-NEXT: v_or_b32_e32 v4, v9, v4
; SI-NEXT: v_or_b32_e32 v5, v8, v5
@@ -11462,7 +11466,7 @@ define inreg <28 x i32> @bitcast_v56f16_to_v28i32_scalar(<56 x half> inreg %a, i
; SI-NEXT: v_lshlrev_b32_e32 v9, 16, v60
; SI-NEXT: v_lshlrev_b32_e32 v10, 16, v62
; SI-NEXT: v_lshlrev_b32_e32 v11, 16, v57
-; SI-NEXT: v_lshlrev_b32_e32 v12, 16, v63
+; SI-NEXT: v_lshlrev_b32_e32 v12, 16, v39
; SI-NEXT: v_lshlrev_b32_e32 v13, 16, v13
; SI-NEXT: v_lshlrev_b32_e32 v14, 16, v34
; SI-NEXT: v_lshlrev_b32_e32 v15, 16, v15
@@ -11477,9 +11481,9 @@ define inreg <28 x i32> @bitcast_v56f16_to_v28i32_scalar(<56 x half> inreg %a, i
; SI-NEXT: v_mov_b32_e32 v40, v56
; SI-NEXT: v_or_b32_e32 v10, v56, v10
; SI-NEXT: v_or_b32_e32 v11, v45, v11
-; SI-NEXT: v_or_b32_e32 v12, v38, v12
-; SI-NEXT: v_or_b32_e32 v13, v36, v13
-; SI-NEXT: v_or_b32_e32 v14, v35, v14
+; SI-NEXT: v_or_b32_e32 v12, v33, v12
+; SI-NEXT: v_or_b32_e32 v13, v38, v13
+; SI-NEXT: v_or_b32_e32 v14, v36, v14
; SI-NEXT: v_or_b32_e32 v15, v32, v15
; SI-NEXT: v_or_b32_e32 v17, v37, v17
; SI-NEXT: buffer_load_dword v26, off, s[0:3], s32 offset:128 ; 4-byte Folded Reload
@@ -11541,10 +11545,10 @@ define inreg <28 x i32> @bitcast_v56f16_to_v28i32_scalar(<56 x half> inreg %a, i
; SI-NEXT: v_lshlrev_b32_e32 v2, 16, v2
; SI-NEXT: v_or_b32_e32 v0, v1, v0
; SI-NEXT: v_or_b32_e32 v1, v3, v2
-; SI-NEXT: v_cvt_f32_f16_e32 v2, v49
-; SI-NEXT: v_cvt_f32_f16_e32 v3, v39
+; SI-NEXT: v_cvt_f32_f16_e32 v2, v63
+; SI-NEXT: v_cvt_f32_f16_e32 v3, v48
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_cvt_f32_f16_e32 v4, v33
+; SI-NEXT: v_cvt_f32_f16_e32 v4, v35
; SI-NEXT: v_cvt_f32_f16_e32 v8, v47
; SI-NEXT: v_add_f32_e32 v2, 0x38000000, v2
; SI-NEXT: v_cvt_f16_f32_e32 v2, v2
@@ -11554,7 +11558,7 @@ define inreg <28 x i32> @bitcast_v56f16_to_v28i32_scalar(<56 x half> inreg %a, i
; SI-NEXT: v_lshlrev_b32_e32 v2, 16, v2
; SI-NEXT: v_cvt_f16_f32_e32 v4, v4
; SI-NEXT: v_or_b32_e32 v2, v3, v2
-; SI-NEXT: v_cvt_f32_f16_e32 v3, v48
+; SI-NEXT: v_cvt_f32_f16_e32 v3, v49
; SI-NEXT: v_add_f32_e32 v8, 0x38000000, v8
; SI-NEXT: buffer_load_dword v5, off, s[0:3], s32 offset:232 ; 4-byte Folded Reload
; SI-NEXT: v_cvt_f16_f32_e32 v8, v8
@@ -11572,9 +11576,9 @@ define inreg <28 x i32> @bitcast_v56f16_to_v28i32_scalar(<56 x half> inreg %a, i
; SI-NEXT: v_cvt_f16_f32_e32 v11, v11
; SI-NEXT: v_add_f32_e32 v12, 0x38000000, v12
; SI-NEXT: v_cvt_f16_f32_e32 v12, v12
-; SI-NEXT: v_cvt_f32_f16_e32 v13, v38
-; SI-NEXT: v_cvt_f32_f16_e32 v14, v36
-; SI-NEXT: v_cvt_f32_f16_e32 v15, v35
+; SI-NEXT: v_cvt_f32_f16_e32 v13, v33
+; SI-NEXT: v_cvt_f32_f16_e32 v14, v38
+; SI-NEXT: v_cvt_f32_f16_e32 v15, v36
; SI-NEXT: v_cvt_f32_f16_e32 v17, v32
; SI-NEXT: v_add_f32_e32 v13, 0x38000000, v13
; SI-NEXT: v_cvt_f16_f32_e32 v13, v13
@@ -11687,7 +11691,7 @@ define inreg <28 x i32> @bitcast_v56f16_to_v28i32_scalar(<56 x half> inreg %a, i
; SI-NEXT: v_cvt_f16_f32_e32 v11, v11
; SI-NEXT: v_lshlrev_b32_e32 v11, 16, v11
; SI-NEXT: v_or_b32_e32 v11, v12, v11
-; SI-NEXT: v_cvt_f32_f16_e32 v12, v63
+; SI-NEXT: v_cvt_f32_f16_e32 v12, v39
; SI-NEXT: v_add_f32_e32 v12, 0x38000000, v12
; SI-NEXT: v_cvt_f16_f32_e32 v12, v12
; SI-NEXT: v_lshlrev_b32_e32 v12, 16, v12
@@ -11827,10 +11831,10 @@ define inreg <28 x i32> @bitcast_v56f16_to_v28i32_scalar(<56 x half> inreg %a, i
; SI-NEXT: s_waitcnt vmcnt(0) expcnt(0)
; SI-NEXT: s_setpc_b64 s[30:31]
; SI-NEXT: .LBB19_4:
-; SI-NEXT: v_mov_b32_e32 v39, v11
-; SI-NEXT: v_mov_b32_e32 v33, v10
-; SI-NEXT: v_mov_b32_e32 v49, v2
-; SI-NEXT: v_mov_b32_e32 v48, v3
+; SI-NEXT: v_mov_b32_e32 v48, v11
+; SI-NEXT: v_mov_b32_e32 v35, v10
+; SI-NEXT: v_mov_b32_e32 v63, v2
+; SI-NEXT: v_mov_b32_e32 v49, v3
; SI-NEXT: v_mov_b32_e32 v52, v37
; SI-NEXT: v_mov_b32_e32 v37, v29
; SI-NEXT: ; implicit-def: $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7_vgpr8_vgpr9_vgpr10_vgpr11_vgpr12_vgpr13_vgpr14_vgpr15_vgpr16_vgpr17_vgpr18_vgpr19_vgpr20_vgpr21_vgpr22_vgpr23_vgpr24_vgpr25_vgpr26_vgpr27_vgpr28_vgpr29_vgpr30_vgpr31
@@ -17248,8 +17252,8 @@ define <28 x float> @bitcast_v56i16_to_v28f32(<56 x i16> %a, i32 %b) {
; GFX9-NEXT: v_lshrrev_b32_e32 v61, 16, v46
; GFX9-NEXT: v_lshrrev_b32_e32 v33, 16, v47
; GFX9-NEXT: v_lshrrev_b32_e32 v34, 16, v56
-; GFX9-NEXT: v_lshrrev_b32_e32 v62, 16, v57
-; GFX9-NEXT: v_lshrrev_b32_e32 v35, 16, v58
+; GFX9-NEXT: v_lshrrev_b32_e32 v35, 16, v57
+; GFX9-NEXT: v_lshrrev_b32_e32 v62, 16, v58
; GFX9-NEXT: v_lshrrev_b32_e32 v36, 16, v59
; GFX9-NEXT: v_cmp_ne_u32_e32 vcc, 0, v28
; GFX9-NEXT: buffer_store_dword v0, off, s[0:3], s32 offset:156 ; 4-byte Folded Spill
@@ -17276,8 +17280,8 @@ define <28 x float> @bitcast_v56i16_to_v28f32(<56 x i16> %a, i32 %b) {
; GFX9-NEXT: v_lshrrev_b32_e32 v21, 16, v38
; GFX9-NEXT: v_lshrrev_b32_e32 v22, 16, v37
; GFX9-NEXT: v_perm_b32 v0, v36, v59, s6
-; GFX9-NEXT: v_perm_b32 v1, v35, v58, s6
-; GFX9-NEXT: v_perm_b32 v2, v62, v57, s6
+; GFX9-NEXT: v_perm_b32 v1, v62, v58, s6
+; GFX9-NEXT: v_perm_b32 v2, v35, v57, s6
; GFX9-NEXT: v_perm_b32 v3, v34, v56, s6
; GFX9-NEXT: v_perm_b32 v4, v33, v47, s6
; GFX9-NEXT: v_perm_b32 v5, v61, v46, s6
@@ -17327,8 +17331,8 @@ define <28 x float> @bitcast_v56i16_to_v28f32(<56 x i16> %a, i32 %b) {
; GFX9-NEXT: ; implicit-def: $vgpr61
; GFX9-NEXT: ; implicit-def: $vgpr33
; GFX9-NEXT: ; implicit-def: $vgpr34
-; GFX9-NEXT: ; implicit-def: $vgpr62
; GFX9-NEXT: ; implicit-def: $vgpr35
+; GFX9-NEXT: ; implicit-def: $vgpr62
; GFX9-NEXT: ; implicit-def: $vgpr36
; GFX9-NEXT: s_waitcnt vmcnt(0)
; GFX9-NEXT: v_perm_b32 v23, v24, v23, s6
@@ -17418,8 +17422,8 @@ define <28 x float> @bitcast_v56i16_to_v28f32(<56 x i16> %a, i32 %b) {
; GFX9-NEXT: buffer_load_dword v24, off, s[0:3], s32 offset:100 ; 4-byte Folded Reload
; GFX9-NEXT: s_mov_b32 s6, 0x5040100
; GFX9-NEXT: v_perm_b32 v0, v36, v59, s6
-; GFX9-NEXT: v_perm_b32 v1, v35, v58, s6
-; GFX9-NEXT: v_perm_b32 v2, v62, v57, s6
+; GFX9-NEXT: v_perm_b32 v1, v62, v58, s6
+; GFX9-NEXT: v_perm_b32 v2, v35, v57, s6
; GFX9-NEXT: v_perm_b32 v3, v34, v56, s6
; GFX9-NEXT: v_perm_b32 v4, v33, v47, s6
; GFX9-NEXT: v_perm_b32 v5, v61, v46, s6
@@ -20129,13 +20133,13 @@ define inreg <56 x half> @bitcast_v28f32_to_v56f16_scalar(<28 x float> inreg %a,
; SI-NEXT: s_lshr_b32 s4, s9, 16
; SI-NEXT: v_cvt_f32_f16_e32 v40, s4
; SI-NEXT: s_lshr_b32 s4, s6, 16
-; SI-NEXT: v_cvt_f32_f16_e32 v41, s4
+; SI-NEXT: v_cvt_f32_f16_e32 v2, s4
; SI-NEXT: s_lshr_b32 s4, s7, 16
-; SI-NEXT: v_cvt_f32_f16_e32 v43, s4
+; SI-NEXT: v_cvt_f32_f16_e32 v3, s4
; SI-NEXT: s_lshr_b32 s4, s8, 16
-; SI-NEXT: v_cvt_f32_f16_e32 v45, s4
+; SI-NEXT: v_cvt_f32_f16_e32 v5, s4
; SI-NEXT: s_lshr_b32 s4, s10, 16
-; SI-NEXT: v_cvt_f32_f16_e32 v47, s4
+; SI-NEXT: v_cvt_f32_f16_e32 v1, s4
; SI-NEXT: s_lshr_b32 s4, s11, 16
; SI-NEXT: v_cvt_f32_f16_e32 v54, s4
; SI-NEXT: s_lshr_b32 s4, s12, 16
@@ -20210,83 +20214,98 @@ define inreg <56 x half> @bitcast_v28f32_to_v56f16_scalar(<28 x float> inreg %a,
; SI-NEXT: v_cvt_f32_f16_e32 v11, s22
; SI-NEXT: v_cvt_f32_f16_e32 v9, s21
; SI-NEXT: v_cvt_f32_f16_e32 v7, s20
-; SI-NEXT: v_cvt_f32_f16_e32 v5, s19
-; SI-NEXT: v_cvt_f32_f16_e32 v3, s18
-; SI-NEXT: v_cvt_f32_f16_e32 v2, s17
-; SI-NEXT: v_cvt_f32_f16_e32 v1, s16
+; SI-NEXT: v_cvt_f32_f16_e32 v41, s19
+; SI-NEXT: v_cvt_f32_f16_e32 v43, s18
+; SI-NEXT: v_cvt_f32_f16_e32 v45, s17
+; SI-NEXT: v_cvt_f32_f16_e32 v47, s16
; SI-NEXT: s_cbranch_execnz .LBB33_3
; SI-NEXT: .LBB33_2: ; %cmp.true
; SI-NEXT: v_add_f32_e64 v1, s16, 1.0
; SI-NEXT: v_lshrrev_b32_e32 v2, 16, v1
+; SI-NEXT: buffer_store_dword v2, off, s[0:3], s32 offset:64 ; 4-byte Folded Spill
+; SI-NEXT: s_waitcnt expcnt(0)
+; SI-NEXT: v_add_f32_e64 v2, s17, 1.0
+; SI-NEXT: v_lshrrev_b32_e32 v3, 16, v2
+; SI-NEXT: buffer_store_dword v3, off, s[0:3], s32 offset:68 ; 4-byte Folded Spill
+; SI-NEXT: s_waitcnt expcnt(0)
+; SI-NEXT: v_add_f32_e64 v3, s18, 1.0
+; SI-NEXT: v_lshrrev_b32_e32 v4, 16, v3
; SI-NEXT: v_add_f32_e64 v14, s11, 1.0
; SI-NEXT: v_add_f32_e64 v36, s6, 1.0
-; SI-NEXT: buffer_store_dword v2, off, s[0:3], s32 offset:64 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v4, off, s[0:3], s32 offset:72 ; 4-byte Folded Spill
; SI-NEXT: v_lshrrev_b32_e32 v6, 16, v14
; SI-NEXT: v_lshrrev_b32_e32 v54, 16, v36
-; SI-NEXT: v_cvt_f32_f16_e32 v41, v54
+; SI-NEXT: v_cvt_f32_f16_e32 v45, v2
+; SI-NEXT: v_cvt_f32_f16_e32 v2, v54
; SI-NEXT: v_cvt_f32_f16_e32 v54, v6
-; SI-NEXT: buffer_load_dword v6, off, s[0:3], s32 offset:64 ; 4-byte Folded Reload
-; SI-NEXT: v_add_f32_e64 v10, s8, 1.0
+; SI-NEXT: buffer_load_dword v6, off, s[0:3], s32 offset:72 ; 4-byte Folded Reload
+; SI-NEXT: v_add_f32_e64 v9, s21, 1.0
+; SI-NEXT: v_add_f32_e64 v15, s24, 1.0
+; SI-NEXT: v_add_f32_e64 v22, s27, 1.0
+; SI-NEXT: v_add_f32_e64 v29, s43, 1.0
+; SI-NEXT: v_lshrrev_b32_e32 v44, 16, v9
+; SI-NEXT: v_lshrrev_b32_e32 v53, 16, v15
+; SI-NEXT: v_lshrrev_b32_e32 v39, 16, v22
; SI-NEXT: v_add_f32_e64 v26, s29, 1.0
+; SI-NEXT: v_lshrrev_b32_e32 v32, 16, v29
+; SI-NEXT: v_add_f32_e64 v10, s8, 1.0
+; SI-NEXT: v_lshrrev_b32_e32 v35, 16, v26
; SI-NEXT: v_lshrrev_b32_e32 v31, 16, v10
-; SI-NEXT: v_lshrrev_b32_e32 v30, 16, v26
-; SI-NEXT: s_waitcnt expcnt(6)
; SI-NEXT: v_cvt_f32_f16_e32 v58, v10
; SI-NEXT: v_cvt_f32_f16_e32 v10, v26
-; SI-NEXT: v_cvt_f32_f16_e32 v26, v31
+; SI-NEXT: v_cvt_f32_f16_e32 v26, v32
+; SI-NEXT: v_cvt_f32_f16_e32 v32, v39
+; SI-NEXT: v_cvt_f32_f16_e32 v39, v53
+; SI-NEXT: v_cvt_f32_f16_e32 v53, v44
+; SI-NEXT: v_add_f32_e64 v48, s9, 1.0
; SI-NEXT: v_add_f32_e64 v12, s10, 1.0
-; SI-NEXT: v_add_f32_e64 v33, s7, 1.0
-; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_add_f32_e64 v2, s17, 1.0
-; SI-NEXT: v_add_f32_e64 v3, s18, 1.0
-; SI-NEXT: v_add_f32_e64 v5, s19, 1.0
+; SI-NEXT: v_lshrrev_b32_e32 v50, 16, v48
+; SI-NEXT: v_lshrrev_b32_e32 v8, 16, v12
+; SI-NEXT: v_cvt_f32_f16_e32 v59, v12
+; SI-NEXT: v_cvt_f32_f16_e32 v12, v22
+; SI-NEXT: v_cvt_f32_f16_e32 v22, v50
; SI-NEXT: v_add_f32_e64 v7, s20, 1.0
-; SI-NEXT: v_add_f32_e64 v9, s21, 1.0
-; SI-NEXT: v_add_f32_e64 v11, s22, 1.0
; SI-NEXT: v_add_f32_e64 v13, s23, 1.0
-; SI-NEXT: v_add_f32_e64 v15, s24, 1.0
-; SI-NEXT: v_add_f32_e64 v18, s25, 1.0
; SI-NEXT: v_add_f32_e64 v20, s26, 1.0
-; SI-NEXT: v_add_f32_e64 v22, s27, 1.0
-; SI-NEXT: v_add_f32_e64 v24, s28, 1.0
-; SI-NEXT: v_add_f32_e64 v29, s43, 1.0
-; SI-NEXT: v_add_f32_e64 v27, s42, 1.0
; SI-NEXT: v_add_f32_e64 v25, s41, 1.0
; SI-NEXT: v_add_f32_e64 v23, s40, 1.0
+; SI-NEXT: v_lshrrev_b32_e32 v46, 16, v7
+; SI-NEXT: v_lshrrev_b32_e32 v55, 16, v13
+; SI-NEXT: v_lshrrev_b32_e32 v49, 16, v20
+; SI-NEXT: v_lshrrev_b32_e32 v28, 16, v25
+; SI-NEXT: v_lshrrev_b32_e32 v40, 16, v23
+; SI-NEXT: v_cvt_f32_f16_e32 v52, v36
+; SI-NEXT: v_cvt_f32_f16_e32 v36, v20
+; SI-NEXT: v_cvt_f32_f16_e32 v20, v40
+; SI-NEXT: v_mov_b32_e32 v40, v22
+; SI-NEXT: v_cvt_f32_f16_e32 v22, v28
+; SI-NEXT: v_cvt_f32_f16_e32 v28, v35
+; SI-NEXT: v_cvt_f32_f16_e32 v35, v49
+; SI-NEXT: v_cvt_f32_f16_e32 v49, v55
+; SI-NEXT: v_cvt_f32_f16_e32 v55, v46
+; SI-NEXT: v_add_f32_e64 v5, s19, 1.0
+; SI-NEXT: v_add_f32_e64 v11, s22, 1.0
+; SI-NEXT: v_add_f32_e64 v18, s25, 1.0
+; SI-NEXT: v_add_f32_e64 v24, s28, 1.0
+; SI-NEXT: v_add_f32_e64 v27, s42, 1.0
; SI-NEXT: v_add_f32_e64 v21, s15, 1.0
; SI-NEXT: v_add_f32_e64 v19, s14, 1.0
; SI-NEXT: v_add_f32_e64 v17, s13, 1.0
; SI-NEXT: v_add_f32_e64 v16, s12, 1.0
-; SI-NEXT: v_lshrrev_b32_e32 v8, 16, v12
-; SI-NEXT: v_lshrrev_b32_e32 v34, 16, v33
-; SI-NEXT: v_add_f32_e64 v48, s9, 1.0
-; SI-NEXT: v_lshrrev_b32_e32 v56, 16, v2
-; SI-NEXT: v_lshrrev_b32_e32 v46, 16, v3
-; SI-NEXT: v_lshrrev_b32_e32 v44, 16, v5
-; SI-NEXT: v_lshrrev_b32_e32 v42, 16, v7
-; SI-NEXT: v_lshrrev_b32_e32 v55, 16, v9
-; SI-NEXT: v_lshrrev_b32_e32 v53, 16, v11
-; SI-NEXT: v_lshrrev_b32_e32 v51, 16, v13
-; SI-NEXT: v_lshrrev_b32_e32 v49, 16, v15
-; SI-NEXT: v_lshrrev_b32_e32 v39, 16, v18
-; SI-NEXT: v_lshrrev_b32_e32 v37, 16, v20
-; SI-NEXT: v_lshrrev_b32_e32 v35, 16, v22
-; SI-NEXT: v_lshrrev_b32_e32 v32, 16, v24
-; SI-NEXT: v_lshrrev_b32_e32 v43, 16, v29
-; SI-NEXT: v_lshrrev_b32_e32 v45, 16, v27
-; SI-NEXT: v_lshrrev_b32_e32 v28, 16, v25
-; SI-NEXT: v_lshrrev_b32_e32 v47, 16, v23
+; SI-NEXT: v_add_f32_e64 v33, s7, 1.0
+; SI-NEXT: v_lshrrev_b32_e32 v56, 16, v5
+; SI-NEXT: v_lshrrev_b32_e32 v42, 16, v11
+; SI-NEXT: v_lshrrev_b32_e32 v51, 16, v18
+; SI-NEXT: v_lshrrev_b32_e32 v37, 16, v24
+; SI-NEXT: v_lshrrev_b32_e32 v30, 16, v27
; SI-NEXT: v_lshrrev_b32_e32 v61, 16, v21
; SI-NEXT: v_lshrrev_b32_e32 v62, 16, v19
; SI-NEXT: v_lshrrev_b32_e32 v63, 16, v17
+; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_lshrrev_b32_e32 v4, 16, v16
-; SI-NEXT: v_lshrrev_b32_e32 v50, 16, v48
-; SI-NEXT: v_cvt_f32_f16_e32 v34, v34
-; SI-NEXT: v_cvt_f32_f16_e32 v8, v8
+; SI-NEXT: v_lshrrev_b32_e32 v34, 16, v33
; SI-NEXT: v_cvt_f32_f16_e32 v38, v48
-; SI-NEXT: v_cvt_f32_f16_e32 v52, v36
; SI-NEXT: v_cvt_f32_f16_e32 v57, v33
-; SI-NEXT: v_cvt_f32_f16_e32 v59, v12
; SI-NEXT: v_cvt_f32_f16_e32 v60, v14
; SI-NEXT: v_cvt_f32_f16_e32 v16, v16
; SI-NEXT: v_cvt_f32_f16_e32 v17, v17
@@ -20297,50 +20316,40 @@ define inreg <56 x half> @bitcast_v28f32_to_v56f16_scalar(<28 x float> inreg %a,
; SI-NEXT: v_cvt_f32_f16_e32 v27, v27
; SI-NEXT: v_cvt_f32_f16_e32 v29, v29
; SI-NEXT: v_cvt_f32_f16_e32 v33, v24
-; SI-NEXT: v_cvt_f32_f16_e32 v12, v22
-; SI-NEXT: v_cvt_f32_f16_e32 v36, v20
; SI-NEXT: v_cvt_f32_f16_e32 v14, v18
; SI-NEXT: v_cvt_f32_f16_e32 v48, v15
; SI-NEXT: v_cvt_f32_f16_e32 v13, v13
+; SI-NEXT: s_waitcnt vmcnt(0)
+; SI-NEXT: v_cvt_f32_f16_e32 v44, v6
+; SI-NEXT: buffer_load_dword v6, off, s[0:3], s32 offset:68 ; 4-byte Folded Reload
; SI-NEXT: v_cvt_f32_f16_e32 v11, v11
; SI-NEXT: v_cvt_f32_f16_e32 v9, v9
; SI-NEXT: v_cvt_f32_f16_e32 v7, v7
-; SI-NEXT: v_cvt_f32_f16_e32 v5, v5
-; SI-NEXT: v_cvt_f32_f16_e32 v3, v3
-; SI-NEXT: v_cvt_f32_f16_e32 v2, v2
-; SI-NEXT: v_cvt_f32_f16_e32 v1, v1
-; SI-NEXT: v_cvt_f32_f16_e32 v40, v50
+; SI-NEXT: v_cvt_f32_f16_e32 v41, v5
+; SI-NEXT: v_cvt_f32_f16_e32 v43, v3
+; SI-NEXT: v_cvt_f32_f16_e32 v47, v1
+; SI-NEXT: v_cvt_f32_f16_e32 v3, v34
+; SI-NEXT: v_cvt_f32_f16_e32 v5, v31
+; SI-NEXT: v_cvt_f32_f16_e32 v1, v8
; SI-NEXT: v_cvt_f32_f16_e32 v50, v4
; SI-NEXT: v_cvt_f32_f16_e32 v4, v63
; SI-NEXT: v_cvt_f32_f16_e32 v15, v62
; SI-NEXT: v_cvt_f32_f16_e32 v18, v61
-; SI-NEXT: v_cvt_f32_f16_e32 v20, v47
-; SI-NEXT: v_cvt_f32_f16_e32 v22, v28
-; SI-NEXT: v_cvt_f32_f16_e32 v24, v45
-; SI-NEXT: v_mov_b32_e32 v45, v26
-; SI-NEXT: v_cvt_f32_f16_e32 v26, v43
-; SI-NEXT: v_cvt_f32_f16_e32 v28, v30
-; SI-NEXT: v_cvt_f32_f16_e32 v30, v32
-; SI-NEXT: v_cvt_f32_f16_e32 v32, v35
-; SI-NEXT: v_cvt_f32_f16_e32 v35, v37
-; SI-NEXT: v_cvt_f32_f16_e32 v37, v39
-; SI-NEXT: v_cvt_f32_f16_e32 v39, v49
-; SI-NEXT: v_cvt_f32_f16_e32 v49, v51
-; SI-NEXT: v_cvt_f32_f16_e32 v51, v53
-; SI-NEXT: v_cvt_f32_f16_e32 v53, v55
-; SI-NEXT: v_cvt_f32_f16_e32 v55, v42
-; SI-NEXT: v_cvt_f32_f16_e32 v42, v44
-; SI-NEXT: v_cvt_f32_f16_e32 v44, v46
-; SI-NEXT: v_cvt_f32_f16_e32 v46, v56
+; SI-NEXT: v_cvt_f32_f16_e32 v24, v30
+; SI-NEXT: v_cvt_f32_f16_e32 v30, v37
+; SI-NEXT: v_cvt_f32_f16_e32 v37, v51
+; SI-NEXT: v_cvt_f32_f16_e32 v51, v42
+; SI-NEXT: v_cvt_f32_f16_e32 v42, v56
+; SI-NEXT: s_waitcnt vmcnt(0)
+; SI-NEXT: v_cvt_f32_f16_e32 v46, v6
+; SI-NEXT: buffer_load_dword v6, off, s[0:3], s32 offset:64 ; 4-byte Folded Reload
; SI-NEXT: s_waitcnt vmcnt(0)
; SI-NEXT: v_cvt_f32_f16_e32 v56, v6
-; SI-NEXT: v_mov_b32_e32 v47, v8
-; SI-NEXT: v_mov_b32_e32 v43, v34
; SI-NEXT: .LBB33_3: ; %end
; SI-NEXT: v_cvt_f16_f32_e32 v6, v56
-; SI-NEXT: v_cvt_f16_f32_e32 v8, v1
+; SI-NEXT: v_cvt_f16_f32_e32 v8, v47
; SI-NEXT: v_cvt_f16_f32_e32 v31, v46
-; SI-NEXT: v_cvt_f16_f32_e32 v34, v2
+; SI-NEXT: v_cvt_f16_f32_e32 v34, v45
; SI-NEXT: v_lshlrev_b32_e32 v6, 16, v6
; SI-NEXT: v_or_b32_e32 v6, v8, v6
; SI-NEXT: v_lshlrev_b32_e32 v8, 16, v31
@@ -20351,14 +20360,14 @@ define inreg <56 x half> @bitcast_v28f32_to_v56f16_scalar(<28 x float> inreg %a,
; SI-NEXT: buffer_store_dword v8, v6, s[0:3], 0 offen
; SI-NEXT: v_cvt_f16_f32_e32 v6, v44
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_cvt_f16_f32_e32 v8, v3
+; SI-NEXT: v_cvt_f16_f32_e32 v8, v43
; SI-NEXT: v_add_i32_e32 v31, vcc, 8, v0
; SI-NEXT: v_lshlrev_b32_e32 v6, 16, v6
; SI-NEXT: v_or_b32_e32 v6, v8, v6
; SI-NEXT: buffer_store_dword v6, v31, s[0:3], 0 offen
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_cvt_f16_f32_e32 v6, v42
-; SI-NEXT: v_cvt_f16_f32_e32 v8, v5
+; SI-NEXT: v_cvt_f16_f32_e32 v8, v41
; SI-NEXT: v_add_i32_e32 v31, vcc, 12, v0
; SI-NEXT: v_lshlrev_b32_e32 v6, 16, v6
; SI-NEXT: v_or_b32_e32 v6, v8, v6
@@ -20497,26 +20506,26 @@ define inreg <56 x half> @bitcast_v28f32_to_v56f16_scalar(<28 x float> inreg %a,
; SI-NEXT: v_or_b32_e32 v4, v6, v4
; SI-NEXT: buffer_store_dword v4, v8, s[0:3], 0 offen
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_cvt_f16_f32_e32 v4, v47
+; SI-NEXT: v_cvt_f16_f32_e32 v4, v1
; SI-NEXT: v_cvt_f16_f32_e32 v6, v59
; SI-NEXT: v_add_i32_e32 v7, vcc, 0x5c, v0
; SI-NEXT: v_lshlrev_b32_e32 v4, 16, v4
; SI-NEXT: v_or_b32_e32 v4, v6, v4
; SI-NEXT: buffer_store_dword v4, v7, s[0:3], 0 offen
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_cvt_f16_f32_e32 v4, v45
+; SI-NEXT: v_cvt_f16_f32_e32 v4, v5
; SI-NEXT: v_cvt_f16_f32_e32 v5, v58
; SI-NEXT: v_add_i32_e32 v6, vcc, 0x60, v0
; SI-NEXT: v_lshlrev_b32_e32 v4, 16, v4
; SI-NEXT: v_or_b32_e32 v4, v5, v4
-; SI-NEXT: v_cvt_f16_f32_e32 v3, v43
+; SI-NEXT: v_cvt_f16_f32_e32 v3, v3
; SI-NEXT: buffer_store_dword v4, v6, s[0:3], 0 offen
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_cvt_f16_f32_e32 v4, v57
; SI-NEXT: v_add_i32_e32 v5, vcc, 0x64, v0
; SI-NEXT: v_lshlrev_b32_e32 v3, 16, v3
; SI-NEXT: v_or_b32_e32 v3, v4, v3
-; SI-NEXT: v_cvt_f16_f32_e32 v2, v41
+; SI-NEXT: v_cvt_f16_f32_e32 v2, v2
; SI-NEXT: buffer_store_dword v3, v5, s[0:3], 0 offen
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_cvt_f16_f32_e32 v3, v52
@@ -20550,13 +20559,13 @@ define inreg <56 x half> @bitcast_v28f32_to_v56f16_scalar(<28 x float> inreg %a,
; SI-NEXT: s_waitcnt vmcnt(0) expcnt(0)
; SI-NEXT: s_setpc_b64 s[30:31]
; SI-NEXT: .LBB33_4:
-; SI-NEXT: ; implicit-def: $vgpr1
+; SI-NEXT: ; implicit-def: $vgpr47
; SI-NEXT: ; implicit-def: $vgpr56
-; SI-NEXT: ; implicit-def: $vgpr2
+; SI-NEXT: ; implicit-def: $vgpr45
; SI-NEXT: ; implicit-def: $vgpr46
-; SI-NEXT: ; implicit-def: $vgpr3
+; SI-NEXT: ; implicit-def: $vgpr43
; SI-NEXT: ; implicit-def: $vgpr44
-; SI-NEXT: ; implicit-def: $vgpr5
+; SI-NEXT: ; implicit-def: $vgpr41
; SI-NEXT: ; implicit-def: $vgpr42
; SI-NEXT: ; implicit-def: $vgpr7
; SI-NEXT: ; implicit-def: $vgpr55
@@ -20597,13 +20606,13 @@ define inreg <56 x half> @bitcast_v28f32_to_v56f16_scalar(<28 x float> inreg %a,
; SI-NEXT: ; implicit-def: $vgpr60
; SI-NEXT: ; implicit-def: $vgpr54
; SI-NEXT: ; implicit-def: $vgpr59
-; SI-NEXT: ; implicit-def: $vgpr47
+; SI-NEXT: ; implicit-def: $vgpr1
; SI-NEXT: ; implicit-def: $vgpr58
-; SI-NEXT: ; implicit-def: $vgpr45
+; SI-NEXT: ; implicit-def: $vgpr5
; SI-NEXT: ; implicit-def: $vgpr57
-; SI-NEXT: ; implicit-def: $vgpr43
+; SI-NEXT: ; implicit-def: $vgpr3
; SI-NEXT: ; implicit-def: $vgpr52
-; SI-NEXT: ; implicit-def: $vgpr41
+; SI-NEXT: ; implicit-def: $vgpr2
; SI-NEXT: ; implicit-def: $vgpr38
; SI-NEXT: ; implicit-def: $vgpr40
; SI-NEXT: s_branch .LBB33_2
@@ -22944,15 +22953,15 @@ define inreg <28 x float> @bitcast_v56f16_to_v28f32_scalar(<56 x half> inreg %a,
; SI-NEXT: buffer_store_dword v0, off, s[0:3], s32 offset:180 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_cvt_f16_f32_e32 v0, v24
-; SI-NEXT: v_cvt_f16_f32_e32 v63, v11
-; SI-NEXT: v_cvt_f16_f32_e32 v38, v10
+; SI-NEXT: v_cvt_f16_f32_e32 v39, v11
+; SI-NEXT: v_cvt_f16_f32_e32 v33, v10
; SI-NEXT: v_cvt_f16_f32_e32 v13, v13
; SI-NEXT: buffer_store_dword v0, off, s[0:3], s32 offset:176 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_cvt_f16_f32_e32 v0, v27
-; SI-NEXT: v_cvt_f16_f32_e32 v36, v12
+; SI-NEXT: v_cvt_f16_f32_e32 v38, v12
; SI-NEXT: v_cvt_f16_f32_e32 v34, v15
-; SI-NEXT: v_cvt_f16_f32_e32 v35, v14
+; SI-NEXT: v_cvt_f16_f32_e32 v36, v14
; SI-NEXT: buffer_store_dword v0, off, s[0:3], s32 offset:172 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_cvt_f16_f32_e32 v0, v26
@@ -23045,9 +23054,9 @@ define inreg <28 x float> @bitcast_v56f16_to_v28f32_scalar(<56 x half> inreg %a,
; SI-NEXT: buffer_load_dword v23, off, s[0:3], s32 offset:156 ; 4-byte Folded Reload
; SI-NEXT: buffer_load_dword v24, off, s[0:3], s32 offset:148 ; 4-byte Folded Reload
; SI-NEXT: buffer_load_dword v25, off, s[0:3], s32 offset:136 ; 4-byte Folded Reload
-; SI-NEXT: v_mov_b32_e32 v49, v2
+; SI-NEXT: v_mov_b32_e32 v63, v2
; SI-NEXT: v_lshlrev_b32_e32 v2, 16, v2
-; SI-NEXT: v_mov_b32_e32 v48, v3
+; SI-NEXT: v_mov_b32_e32 v49, v3
; SI-NEXT: v_lshlrev_b32_e32 v3, 16, v3
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_lshlrev_b32_e32 v4, 16, v4
@@ -23056,9 +23065,9 @@ define inreg <28 x float> @bitcast_v56f16_to_v28f32_scalar(<56 x half> inreg %a,
; SI-NEXT: v_mov_b32_e32 v61, v44
; SI-NEXT: v_lshlrev_b32_e32 v0, 16, v43
; SI-NEXT: v_lshlrev_b32_e32 v1, 16, v54
-; SI-NEXT: v_mov_b32_e32 v39, v11
+; SI-NEXT: v_mov_b32_e32 v48, v11
; SI-NEXT: v_or_b32_e32 v2, v11, v2
-; SI-NEXT: v_mov_b32_e32 v33, v10
+; SI-NEXT: v_mov_b32_e32 v35, v10
; SI-NEXT: v_or_b32_e32 v3, v10, v3
; SI-NEXT: v_or_b32_e32 v4, v9, v4
; SI-NEXT: v_or_b32_e32 v5, v8, v5
@@ -23068,7 +23077,7 @@ define inreg <28 x float> @bitcast_v56f16_to_v28f32_scalar(<56 x half> inreg %a,
; SI-NEXT: v_lshlrev_b32_e32 v9, 16, v60
; SI-NEXT: v_lshlrev_b32_e32 v10, 16, v62
; SI-NEXT: v_lshlrev_b32_e32 v11, 16, v57
-; SI-NEXT: v_lshlrev_b32_e32 v12, 16, v63
+; SI-NEXT: v_lshlrev_b32_e32 v12, 16, v39
; SI-NEXT: v_lshlrev_b32_e32 v13, 16, v13
; SI-NEXT: v_lshlrev_b32_e32 v14, 16, v34
; SI-NEXT: v_lshlrev_b32_e32 v15, 16, v15
@@ -23083,9 +23092,9 @@ define inreg <28 x float> @bitcast_v56f16_to_v28f32_scalar(<56 x half> inreg %a,
; SI-NEXT: v_mov_b32_e32 v40, v56
; SI-NEXT: v_or_b32_e32 v10, v56, v10
; SI-NEXT: v_or_b32_e32 v11, v45, v11
-; SI-NEXT: v_or_b32_e32 v12, v38, v12
-; SI-NEXT: v_or_b32_e32 v13, v36, v13
-; SI-NEXT: v_or_b32_e32 v14, v35, v14
+; SI-NEXT: v_or_b32_e32 v12, v33, v12
+; SI-NEXT: v_or_b32_e32 v13, v38, v13
+; SI-NEXT: v_or_b32_e32 v14, v36, v14
; SI-NEXT: v_or_b32_e32 v15, v32, v15
; SI-NEXT: v_or_b32_e32 v17, v37, v17
; SI-NEXT: buffer_load_dword v26, off, s[0:3], s32 offset:128 ; 4-byte Folded Reload
@@ -23147,10 +23156,10 @@ define inreg <28 x float> @bitcast_v56f16_to_v28f32_scalar(<56 x half> inreg %a,
; SI-NEXT: v_lshlrev_b32_e32 v2, 16, v2
; SI-NEXT: v_or_b32_e32 v0, v1, v0
; SI-NEXT: v_or_b32_e32 v1, v3, v2
-; SI-NEXT: v_cvt_f32_f16_e32 v2, v49
-; SI-NEXT: v_cvt_f32_f16_e32 v3, v39
+; SI-NEXT: v_cvt_f32_f16_e32 v2, v63
+; SI-NEXT: v_cvt_f32_f16_e32 v3, v48
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_cvt_f32_f16_e32 v4, v33
+; SI-NEXT: v_cvt_f32_f16_e32 v4, v35
; SI-NEXT: v_cvt_f32_f16_e32 v8, v47
; SI-NEXT: v_add_f32_e32 v2, 0x38000000, v2
; SI-NEXT: v_cvt_f16_f32_e32 v2, v2
@@ -23160,7 +23169,7 @@ define inreg <28 x float> @bitcast_v56f16_to_v28f32_scalar(<56 x half> inreg %a,
; SI-NEXT: v_lshlrev_b32_e32 v2, 16, v2
; SI-NEXT: v_cvt_f16_f32_e32 v4, v4
; SI-NEXT: v_or_b32_e32 v2, v3, v2
-; SI-NEXT: v_cvt_f32_f16_e32 v3, v48
+; SI-NEXT: v_cvt_f32_f16_e32 v3, v49
; SI-NEXT: v_add_f32_e32 v8, 0x38000000, v8
; SI-NEXT: buffer_load_dword v5, off, s[0:3], s32 offset:232 ; 4-byte Folded Reload
; SI-NEXT: v_cvt_f16_f32_e32 v8, v8
@@ -23178,9 +23187,9 @@ define inreg <28 x float> @bitcast_v56f16_to_v28f32_scalar(<56 x half> inreg %a,
; SI-NEXT: v_cvt_f16_f32_e32 v11, v11
; SI-NEXT: v_add_f32_e32 v12, 0x38000000, v12
; SI-NEXT: v_cvt_f16_f32_e32 v12, v12
-; SI-NEXT: v_cvt_f32_f16_e32 v13, v38
-; SI-NEXT: v_cvt_f32_f16_e32 v14, v36
-; SI-NEXT: v_cvt_f32_f16_e32 v15, v35
+; SI-NEXT: v_cvt_f32_f16_e32 v13, v33
+; SI-NEXT: v_cvt_f32_f16_e32 v14, v38
+; SI-NEXT: v_cvt_f32_f16_e32 v15, v36
; SI-NEXT: v_cvt_f32_f16_e32 v17, v32
; SI-NEXT: v_add_f32_e32 v13, 0x38000000, v13
; SI-NEXT: v_cvt_f16_f32_e32 v13, v13
@@ -23293,7 +23302,7 @@ define inreg <28 x float> @bitcast_v56f16_to_v28f32_scalar(<56 x half> inreg %a,
; SI-NEXT: v_cvt_f16_f32_e32 v11, v11
; SI-NEXT: v_lshlrev_b32_e32 v11, 16, v11
; SI-NEXT: v_or_b32_e32 v11, v12, v11
-; SI-NEXT: v_cvt_f32_f16_e32 v12, v63
+; SI-NEXT: v_cvt_f32_f16_e32 v12, v39
; SI-NEXT: v_add_f32_e32 v12, 0x38000000, v12
; SI-NEXT: v_cvt_f16_f32_e32 v12, v12
; SI-NEXT: v_lshlrev_b32_e32 v12, 16, v12
@@ -23433,10 +23442,10 @@ define inreg <28 x float> @bitcast_v56f16_to_v28f32_scalar(<56 x half> inreg %a,
; SI-NEXT: s_waitcnt vmcnt(0) expcnt(0)
; SI-NEXT: s_setpc_b64 s[30:31]
; SI-NEXT: .LBB35_4:
-; SI-NEXT: v_mov_b32_e32 v39, v11
-; SI-NEXT: v_mov_b32_e32 v33, v10
-; SI-NEXT: v_mov_b32_e32 v49, v2
-; SI-NEXT: v_mov_b32_e32 v48, v3
+; SI-NEXT: v_mov_b32_e32 v48, v11
+; SI-NEXT: v_mov_b32_e32 v35, v10
+; SI-NEXT: v_mov_b32_e32 v63, v2
+; SI-NEXT: v_mov_b32_e32 v49, v3
; SI-NEXT: v_mov_b32_e32 v52, v37
; SI-NEXT: v_mov_b32_e32 v37, v29
; SI-NEXT: ; implicit-def: $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7_vgpr8_vgpr9_vgpr10_vgpr11_vgpr12_vgpr13_vgpr14_vgpr15_vgpr16_vgpr17_vgpr18_vgpr19_vgpr20_vgpr21_vgpr22_vgpr23_vgpr24_vgpr25_vgpr26_vgpr27_vgpr28_vgpr29_vgpr30_vgpr31
@@ -26820,7 +26829,7 @@ define inreg <56 x i16> @bitcast_v14i64_to_v56i16_scalar(<14 x i64> inreg %a, i3
; GFX11-TRUE16-NEXT: v_readfirstlane_b32 s5, v7
; GFX11-TRUE16-NEXT: v_readfirstlane_b32 s6, v8
; GFX11-TRUE16-NEXT: v_readfirstlane_b32 s4, v9
-; GFX11-TRUE16-NEXT: s_mov_b32 s90, 0
+; GFX11-TRUE16-NEXT: s_mov_b32 s78, 0
; GFX11-TRUE16-NEXT: s_and_b32 s14, vcc_lo, exec_lo
; GFX11-TRUE16-NEXT: s_cbranch_scc0 .LBB41_4
; GFX11-TRUE16-NEXT: ; %bb.1: ; %cmp.false
@@ -26848,11 +26857,11 @@ define inreg <56 x i16> @bitcast_v14i64_to_v56i16_scalar(<14 x i64> inreg %a, i3
; GFX11-TRUE16-NEXT: s_lshr_b32 s75, s18, 16
; GFX11-TRUE16-NEXT: s_lshr_b32 s76, s17, 16
; GFX11-TRUE16-NEXT: s_lshr_b32 s77, s16, 16
-; GFX11-TRUE16-NEXT: s_lshr_b32 s78, s3, 16
-; GFX11-TRUE16-NEXT: s_lshr_b32 s79, s2, 16
-; GFX11-TRUE16-NEXT: s_lshr_b32 s88, s1, 16
-; GFX11-TRUE16-NEXT: s_lshr_b32 s89, s0, 16
-; GFX11-TRUE16-NEXT: s_and_not1_b32 vcc_lo, exec_lo, s90
+; GFX11-TRUE16-NEXT: s_lshr_b32 s79, s3, 16
+; GFX11-TRUE16-NEXT: s_lshr_b32 s88, s2, 16
+; GFX11-TRUE16-NEXT: s_lshr_b32 s89, s1, 16
+; GFX11-TRUE16-NEXT: s_lshr_b32 s90, s0, 16
+; GFX11-TRUE16-NEXT: s_and_not1_b32 vcc_lo, exec_lo, s78
; GFX11-TRUE16-NEXT: s_cbranch_vccnz .LBB41_3
; GFX11-TRUE16-NEXT: .LBB41_2: ; %cmp.true
; GFX11-TRUE16-NEXT: s_add_u32 s6, s6, 3
@@ -26907,16 +26916,18 @@ define inreg <56 x i16> @bitcast_v14i64_to_v56i16_scalar(<14 x i64> inreg %a, i3
; GFX11-TRUE16-NEXT: s_lshr_b32 s75, s18, 16
; GFX11-TRUE16-NEXT: s_lshr_b32 s76, s17, 16
; GFX11-TRUE16-NEXT: s_lshr_b32 s77, s16, 16
-; GFX11-TRUE16-NEXT: s_lshr_b32 s78, s3, 16
-; GFX11-TRUE16-NEXT: s_lshr_b32 s79, s2, 16
-; GFX11-TRUE16-NEXT: s_lshr_b32 s88, s1, 16
-; GFX11-TRUE16-NEXT: s_lshr_b32 s89, s0, 16
+; GFX11-TRUE16-NEXT: s_lshr_b32 s79, s3, 16
+; GFX11-TRUE16-NEXT: s_lshr_b32 s88, s2, 16
+; GFX11-TRUE16-NEXT: s_lshr_b32 s89, s1, 16
+; GFX11-TRUE16-NEXT: s_lshr_b32 s90, s0, 16
; GFX11-TRUE16-NEXT: .LBB41_3: ; %end
; GFX11-TRUE16-NEXT: s_delay_alu instid0(SALU_CYCLE_1)
-; GFX11-TRUE16-NEXT: s_pack_ll_b32_b16 s0, s0, s89
-; GFX11-TRUE16-NEXT: s_pack_ll_b32_b16 s1, s1, s88
-; GFX11-TRUE16-NEXT: s_pack_ll_b32_b16 s2, s2, s79
-; GFX11-TRUE16-NEXT: s_pack_ll_b32_b16 s3, s3, s78
+; GFX11-TRUE16-NEXT: s_mov_b32 s78, s90
+; GFX11-TRUE16-NEXT: s_pack_ll_b32_b16 s1, s1, s89
+; GFX11-TRUE16-NEXT: s_pack_ll_b32_b16 s0, s0, s78
+; GFX11-TRUE16-NEXT: s_mov_b32 s78, s88
+; GFX11-TRUE16-NEXT: s_pack_ll_b32_b16 s3, s3, s79
+; GFX11-TRUE16-NEXT: s_pack_ll_b32_b16 s2, s2, s78
; GFX11-TRUE16-NEXT: s_pack_ll_b32_b16 s16, s16, s77
; GFX11-TRUE16-NEXT: s_pack_ll_b32_b16 s17, s17, s76
; GFX11-TRUE16-NEXT: s_pack_ll_b32_b16 s18, s18, s75
@@ -26957,10 +26968,10 @@ define inreg <56 x i16> @bitcast_v14i64_to_v56i16_scalar(<14 x i64> inreg %a, i3
; GFX11-TRUE16-NEXT: v_dual_mov_b32 v26, s6 :: v_dual_mov_b32 v27, s4
; GFX11-TRUE16-NEXT: s_setpc_b64 s[30:31]
; GFX11-TRUE16-NEXT: .LBB41_4:
+; GFX11-TRUE16-NEXT: ; implicit-def: $sgpr90_lo16
; GFX11-TRUE16-NEXT: ; implicit-def: $sgpr89_lo16
; GFX11-TRUE16-NEXT: ; implicit-def: $sgpr88_lo16
; GFX11-TRUE16-NEXT: ; implicit-def: $sgpr79_lo16
-; GFX11-TRUE16-NEXT: ; implicit-def: $sgpr78_lo16
; GFX11-TRUE16-NEXT: ; implicit-def: $sgpr77_lo16
; GFX11-TRUE16-NEXT: ; implicit-def: $sgpr76_lo16
; GFX11-TRUE16-NEXT: ; implicit-def: $sgpr75_lo16
@@ -28005,8 +28016,8 @@ define <14 x i64> @bitcast_v56i16_to_v14i64(<56 x i16> %a, i32 %b) {
; GFX9-NEXT: v_lshrrev_b32_e32 v61, 16, v46
; GFX9-NEXT: v_lshrrev_b32_e32 v33, 16, v47
; GFX9-NEXT: v_lshrrev_b32_e32 v34, 16, v56
-; GFX9-NEXT: v_lshrrev_b32_e32 v62, 16, v57
-; GFX9-NEXT: v_lshrrev_b32_e32 v35, 16, v58
+; GFX9-NEXT: v_lshrrev_b32_e32 v35, 16, v57
+; GFX9-NEXT: v_lshrrev_b32_e32 v62, 16, v58
; GFX9-NEXT: v_lshrrev_b32_e32 v36, 16, v59
; GFX9-NEXT: v_cmp_ne_u32_e32 vcc, 0, v28
; GFX9-NEXT: buffer_store_dword v0, off, s[0:3], s32 offset:156 ; 4-byte Folded Spill
@@ -28033,8 +28044,8 @@ define <14 x i64> @bitcast_v56i16_to_v14i64(<56 x i16> %a, i32 %b) {
; GFX9-NEXT: v_lshrrev_b32_e32 v21, 16, v38
; GFX9-NEXT: v_lshrrev_b32_e32 v22, 16, v37
; GFX9-NEXT: v_perm_b32 v0, v36, v59, s6
-; GFX9-NEXT: v_perm_b32 v1, v35, v58, s6
-; GFX9-NEXT: v_perm_b32 v2, v62, v57, s6
+; GFX9-NEXT: v_perm_b32 v1, v62, v58, s6
+; GFX9-NEXT: v_perm_b32 v2, v35, v57, s6
; GFX9-NEXT: v_perm_b32 v3, v34, v56, s6
; GFX9-NEXT: v_perm_b32 v4, v33, v47, s6
; GFX9-NEXT: v_perm_b32 v5, v61, v46, s6
@@ -28084,8 +28095,8 @@ define <14 x i64> @bitcast_v56i16_to_v14i64(<56 x i16> %a, i32 %b) {
; GFX9-NEXT: ; implicit-def: $vgpr61
; GFX9-NEXT: ; implicit-def: $vgpr33
; GFX9-NEXT: ; implicit-def: $vgpr34
-; GFX9-NEXT: ; implicit-def: $vgpr62
; GFX9-NEXT: ; implicit-def: $vgpr35
+; GFX9-NEXT: ; implicit-def: $vgpr62
; GFX9-NEXT: ; implicit-def: $vgpr36
; GFX9-NEXT: s_waitcnt vmcnt(0)
; GFX9-NEXT: v_perm_b32 v23, v24, v23, s6
@@ -28175,8 +28186,8 @@ define <14 x i64> @bitcast_v56i16_to_v14i64(<56 x i16> %a, i32 %b) {
; GFX9-NEXT: buffer_load_dword v24, off, s[0:3], s32 offset:100 ; 4-byte Folded Reload
; GFX9-NEXT: s_mov_b32 s6, 0x5040100
; GFX9-NEXT: v_perm_b32 v0, v36, v59, s6
-; GFX9-NEXT: v_perm_b32 v1, v35, v58, s6
-; GFX9-NEXT: v_perm_b32 v2, v62, v57, s6
+; GFX9-NEXT: v_perm_b32 v1, v62, v58, s6
+; GFX9-NEXT: v_perm_b32 v2, v35, v57, s6
; GFX9-NEXT: v_perm_b32 v3, v34, v56, s6
; GFX9-NEXT: v_perm_b32 v4, v33, v47, s6
; GFX9-NEXT: v_perm_b32 v5, v61, v46, s6
@@ -31867,7 +31878,7 @@ define inreg <56 x half> @bitcast_v14i64_to_v56f16_scalar(<14 x i64> inreg %a, i
; GFX11-TRUE16-NEXT: v_readfirstlane_b32 s5, v7
; GFX11-TRUE16-NEXT: v_readfirstlane_b32 s6, v8
; GFX11-TRUE16-NEXT: v_readfirstlane_b32 s4, v9
-; GFX11-TRUE16-NEXT: s_mov_b32 s90, 0
+; GFX11-TRUE16-NEXT: s_mov_b32 s78, 0
; GFX11-TRUE16-NEXT: s_and_b32 s14, vcc_lo, exec_lo
; GFX11-TRUE16-NEXT: s_cbranch_scc0 .LBB45_4
; GFX11-TRUE16-NEXT: ; %bb.1: ; %cmp.false
@@ -31895,11 +31906,11 @@ define inreg <56 x half> @bitcast_v14i64_to_v56f16_scalar(<14 x i64> inreg %a, i
; GFX11-TRUE16-NEXT: s_lshr_b32 s75, s18, 16
; GFX11-TRUE16-NEXT: s_lshr_b32 s76, s17, 16
; GFX11-TRUE16-NEXT: s_lshr_b32 s77, s16, 16
-; GFX11-TRUE16-NEXT: s_lshr_b32 s78, s3, 16
-; GFX11-TRUE16-NEXT: s_lshr_b32 s79, s2, 16
-; GFX11-TRUE16-NEXT: s_lshr_b32 s88, s1, 16
-; GFX11-TRUE16-NEXT: s_lshr_b32 s89, s0, 16
-; GFX11-TRUE16-NEXT: s_and_not1_b32 vcc_lo, exec_lo, s90
+; GFX11-TRUE16-NEXT: s_lshr_b32 s79, s3, 16
+; GFX11-TRUE16-NEXT: s_lshr_b32 s88, s2, 16
+; GFX11-TRUE16-NEXT: s_lshr_b32 s89, s1, 16
+; GFX11-TRUE16-NEXT: s_lshr_b32 s90, s0, 16
+; GFX11-TRUE16-NEXT: s_and_not1_b32 vcc_lo, exec_lo, s78
; GFX11-TRUE16-NEXT: s_cbranch_vccnz .LBB45_3
; GFX11-TRUE16-NEXT: .LBB45_2: ; %cmp.true
; GFX11-TRUE16-NEXT: s_add_u32 s6, s6, 3
@@ -31954,16 +31965,18 @@ define inreg <56 x half> @bitcast_v14i64_to_v56f16_scalar(<14 x i64> inreg %a, i
; GFX11-TRUE16-NEXT: s_lshr_b32 s75, s18, 16
; GFX11-TRUE16-NEXT: s_lshr_b32 s76, s17, 16
; GFX11-TRUE16-NEXT: s_lshr_b32 s77, s16, 16
-; GFX11-TRUE16-NEXT: s_lshr_b32 s78, s3, 16
-; GFX11-TRUE16-NEXT: s_lshr_b32 s79, s2, 16
-; GFX11-TRUE16-NEXT: s_lshr_b32 s88, s1, 16
-; GFX11-TRUE16-NEXT: s_lshr_b32 s89, s0, 16
+; GFX11-TRUE16-NEXT: s_lshr_b32 s79, s3, 16
+; GFX11-TRUE16-NEXT: s_lshr_b32 s88, s2, 16
+; GFX11-TRUE16-NEXT: s_lshr_b32 s89, s1, 16
+; GFX11-TRUE16-NEXT: s_lshr_b32 s90, s0, 16
; GFX11-TRUE16-NEXT: .LBB45_3: ; %end
; GFX11-TRUE16-NEXT: s_delay_alu instid0(SALU_CYCLE_1)
-; GFX11-TRUE16-NEXT: s_pack_ll_b32_b16 s0, s0, s89
-; GFX11-TRUE16-NEXT: s_pack_ll_b32_b16 s1, s1, s88
-; GFX11-TRUE16-NEXT: s_pack_ll_b32_b16 s2, s2, s79
-; GFX11-TRUE16-NEXT: s_pack_ll_b32_b16 s3, s3, s78
+; GFX11-TRUE16-NEXT: s_mov_b32 s78, s90
+; GFX11-TRUE16-NEXT: s_pack_ll_b32_b16 s1, s1, s89
+; GFX11-TRUE16-NEXT: s_pack_ll_b32_b16 s0, s0, s78
+; GFX11-TRUE16-NEXT: s_mov_b32 s78, s88
+; GFX11-TRUE16-NEXT: s_pack_ll_b32_b16 s3, s3, s79
+; GFX11-TRUE16-NEXT: s_pack_ll_b32_b16 s2, s2, s78
; GFX11-TRUE16-NEXT: s_pack_ll_b32_b16 s16, s16, s77
; GFX11-TRUE16-NEXT: s_pack_ll_b32_b16 s17, s17, s76
; GFX11-TRUE16-NEXT: s_pack_ll_b32_b16 s18, s18, s75
@@ -32004,10 +32017,10 @@ define inreg <56 x half> @bitcast_v14i64_to_v56f16_scalar(<14 x i64> inreg %a, i
; GFX11-TRUE16-NEXT: v_dual_mov_b32 v26, s6 :: v_dual_mov_b32 v27, s4
; GFX11-TRUE16-NEXT: s_setpc_b64 s[30:31]
; GFX11-TRUE16-NEXT: .LBB45_4:
+; GFX11-TRUE16-NEXT: ; implicit-def: $sgpr90_lo16
; GFX11-TRUE16-NEXT: ; implicit-def: $sgpr89_lo16
; GFX11-TRUE16-NEXT: ; implicit-def: $sgpr88_lo16
; GFX11-TRUE16-NEXT: ; implicit-def: $sgpr79_lo16
-; GFX11-TRUE16-NEXT: ; implicit-def: $sgpr78_lo16
; GFX11-TRUE16-NEXT: ; implicit-def: $sgpr77_lo16
; GFX11-TRUE16-NEXT: ; implicit-def: $sgpr76_lo16
; GFX11-TRUE16-NEXT: ; implicit-def: $sgpr75_lo16
@@ -33725,15 +33738,15 @@ define inreg <14 x i64> @bitcast_v56f16_to_v14i64_scalar(<56 x half> inreg %a, i
; SI-NEXT: buffer_store_dword v0, off, s[0:3], s32 offset:180 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_cvt_f16_f32_e32 v0, v24
-; SI-NEXT: v_cvt_f16_f32_e32 v63, v11
-; SI-NEXT: v_cvt_f16_f32_e32 v38, v10
+; SI-NEXT: v_cvt_f16_f32_e32 v39, v11
+; SI-NEXT: v_cvt_f16_f32_e32 v33, v10
; SI-NEXT: v_cvt_f16_f32_e32 v13, v13
; SI-NEXT: buffer_store_dword v0, off, s[0:3], s32 offset:176 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_cvt_f16_f32_e32 v0, v27
-; SI-NEXT: v_cvt_f16_f32_e32 v36, v12
+; SI-NEXT: v_cvt_f16_f32_e32 v38, v12
; SI-NEXT: v_cvt_f16_f32_e32 v34, v15
-; SI-NEXT: v_cvt_f16_f32_e32 v35, v14
+; SI-NEXT: v_cvt_f16_f32_e32 v36, v14
; SI-NEXT: buffer_store_dword v0, off, s[0:3], s32 offset:172 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_cvt_f16_f32_e32 v0, v26
@@ -33826,9 +33839,9 @@ define inreg <14 x i64> @bitcast_v56f16_to_v14i64_scalar(<56 x half> inreg %a, i
; SI-NEXT: buffer_load_dword v23, off, s[0:3], s32 offset:156 ; 4-byte Folded Reload
; SI-NEXT: buffer_load_dword v24, off, s[0:3], s32 offset:148 ; 4-byte Folded Reload
; SI-NEXT: buffer_load_dword v25, off, s[0:3], s32 offset:136 ; 4-byte Folded Reload
-; SI-NEXT: v_mov_b32_e32 v49, v2
+; SI-NEXT: v_mov_b32_e32 v63, v2
; SI-NEXT: v_lshlrev_b32_e32 v2, 16, v2
-; SI-NEXT: v_mov_b32_e32 v48, v3
+; SI-NEXT: v_mov_b32_e32 v49, v3
; SI-NEXT: v_lshlrev_b32_e32 v3, 16, v3
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_lshlrev_b32_e32 v4, 16, v4
@@ -33837,9 +33850,9 @@ define inreg <14 x i64> @bitcast_v56f16_to_v14i64_scalar(<56 x half> inreg %a, i
; SI-NEXT: v_mov_b32_e32 v61, v44
; SI-NEXT: v_lshlrev_b32_e32 v0, 16, v43
; SI-NEXT: v_lshlrev_b32_e32 v1, 16, v54
-; SI-NEXT: v_mov_b32_e32 v39, v11
+; SI-NEXT: v_mov_b32_e32 v48, v11
; SI-NEXT: v_or_b32_e32 v2, v11, v2
-; SI-NEXT: v_mov_b32_e32 v33, v10
+; SI-NEXT: v_mov_b32_e32 v35, v10
; SI-NEXT: v_or_b32_e32 v3, v10, v3
; SI-NEXT: v_or_b32_e32 v4, v9, v4
; SI-NEXT: v_or_b32_e32 v5, v8, v5
@@ -33849,7 +33862,7 @@ define inreg <14 x i64> @bitcast_v56f16_to_v14i64_scalar(<56 x half> inreg %a, i
; SI-NEXT: v_lshlrev_b32_e32 v9, 16, v60
; SI-NEXT: v_lshlrev_b32_e32 v10, 16, v62
; SI-NEXT: v_lshlrev_b32_e32 v11, 16, v57
-; SI-NEXT: v_lshlrev_b32_e32 v12, 16, v63
+; SI-NEXT: v_lshlrev_b32_e32 v12, 16, v39
; SI-NEXT: v_lshlrev_b32_e32 v13, 16, v13
; SI-NEXT: v_lshlrev_b32_e32 v14, 16, v34
; SI-NEXT: v_lshlrev_b32_e32 v15, 16, v15
@@ -33864,9 +33877,9 @@ define inreg <14 x i64> @bitcast_v56f16_to_v14i64_scalar(<56 x half> inreg %a, i
; SI-NEXT: v_mov_b32_e32 v40, v56
; SI-NEXT: v_or_b32_e32 v10, v56, v10
; SI-NEXT: v_or_b32_e32 v11, v45, v11
-; SI-NEXT: v_or_b32_e32 v12, v38, v12
-; SI-NEXT: v_or_b32_e32 v13, v36, v13
-; SI-NEXT: v_or_b32_e32 v14, v35, v14
+; SI-NEXT: v_or_b32_e32 v12, v33, v12
+; SI-NEXT: v_or_b32_e32 v13, v38, v13
+; SI-NEXT: v_or_b32_e32 v14, v36, v14
; SI-NEXT: v_or_b32_e32 v15, v32, v15
; SI-NEXT: v_or_b32_e32 v17, v37, v17
; SI-NEXT: buffer_load_dword v26, off, s[0:3], s32 offset:128 ; 4-byte Folded Reload
@@ -33928,10 +33941,10 @@ define inreg <14 x i64> @bitcast_v56f16_to_v14i64_scalar(<56 x half> inreg %a, i
; SI-NEXT: v_lshlrev_b32_e32 v2, 16, v2
; SI-NEXT: v_or_b32_e32 v0, v1, v0
; SI-NEXT: v_or_b32_e32 v1, v3, v2
-; SI-NEXT: v_cvt_f32_f16_e32 v2, v49
-; SI-NEXT: v_cvt_f32_f16_e32 v3, v39
+; SI-NEXT: v_cvt_f32_f16_e32 v2, v63
+; SI-NEXT: v_cvt_f32_f16_e32 v3, v48
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_cvt_f32_f16_e32 v4, v33
+; SI-NEXT: v_cvt_f32_f16_e32 v4, v35
; SI-NEXT: v_cvt_f32_f16_e32 v8, v47
; SI-NEXT: v_add_f32_e32 v2, 0x38000000, v2
; SI-NEXT: v_cvt_f16_f32_e32 v2, v2
@@ -33941,7 +33954,7 @@ define inreg <14 x i64> @bitcast_v56f16_to_v14i64_scalar(<56 x half> inreg %a, i
; SI-NEXT: v_lshlrev_b32_e32 v2, 16, v2
; SI-NEXT: v_cvt_f16_f32_e32 v4, v4
; SI-NEXT: v_or_b32_e32 v2, v3, v2
-; SI-NEXT: v_cvt_f32_f16_e32 v3, v48
+; SI-NEXT: v_cvt_f32_f16_e32 v3, v49
; SI-NEXT: v_add_f32_e32 v8, 0x38000000, v8
; SI-NEXT: buffer_load_dword v5, off, s[0:3], s32 offset:232 ; 4-byte Folded Reload
; SI-NEXT: v_cvt_f16_f32_e32 v8, v8
@@ -33959,9 +33972,9 @@ define inreg <14 x i64> @bitcast_v56f16_to_v14i64_scalar(<56 x half> inreg %a, i
; SI-NEXT: v_cvt_f16_f32_e32 v11, v11
; SI-NEXT: v_add_f32_e32 v12, 0x38000000, v12
; SI-NEXT: v_cvt_f16_f32_e32 v12, v12
-; SI-NEXT: v_cvt_f32_f16_e32 v13, v38
-; SI-NEXT: v_cvt_f32_f16_e32 v14, v36
-; SI-NEXT: v_cvt_f32_f16_e32 v15, v35
+; SI-NEXT: v_cvt_f32_f16_e32 v13, v33
+; SI-NEXT: v_cvt_f32_f16_e32 v14, v38
+; SI-NEXT: v_cvt_f32_f16_e32 v15, v36
; SI-NEXT: v_cvt_f32_f16_e32 v17, v32
; SI-NEXT: v_add_f32_e32 v13, 0x38000000, v13
; SI-NEXT: v_cvt_f16_f32_e32 v13, v13
@@ -34074,7 +34087,7 @@ define inreg <14 x i64> @bitcast_v56f16_to_v14i64_scalar(<56 x half> inreg %a, i
; SI-NEXT: v_cvt_f16_f32_e32 v11, v11
; SI-NEXT: v_lshlrev_b32_e32 v11, 16, v11
; SI-NEXT: v_or_b32_e32 v11, v12, v11
-; SI-NEXT: v_cvt_f32_f16_e32 v12, v63
+; SI-NEXT: v_cvt_f32_f16_e32 v12, v39
; SI-NEXT: v_add_f32_e32 v12, 0x38000000, v12
; SI-NEXT: v_cvt_f16_f32_e32 v12, v12
; SI-NEXT: v_lshlrev_b32_e32 v12, 16, v12
@@ -34214,10 +34227,10 @@ define inreg <14 x i64> @bitcast_v56f16_to_v14i64_scalar(<56 x half> inreg %a, i
; SI-NEXT: s_waitcnt vmcnt(0) expcnt(0)
; SI-NEXT: s_setpc_b64 s[30:31]
; SI-NEXT: .LBB47_4:
-; SI-NEXT: v_mov_b32_e32 v39, v11
-; SI-NEXT: v_mov_b32_e32 v33, v10
-; SI-NEXT: v_mov_b32_e32 v49, v2
-; SI-NEXT: v_mov_b32_e32 v48, v3
+; SI-NEXT: v_mov_b32_e32 v48, v11
+; SI-NEXT: v_mov_b32_e32 v35, v10
+; SI-NEXT: v_mov_b32_e32 v63, v2
+; SI-NEXT: v_mov_b32_e32 v49, v3
; SI-NEXT: v_mov_b32_e32 v52, v37
; SI-NEXT: v_mov_b32_e32 v37, v29
; SI-NEXT: ; implicit-def: $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7_vgpr8_vgpr9_vgpr10_vgpr11_vgpr12_vgpr13_vgpr14_vgpr15_vgpr16_vgpr17_vgpr18_vgpr19_vgpr20_vgpr21_vgpr22_vgpr23_vgpr24_vgpr25_vgpr26_vgpr27_vgpr28_vgpr29_vgpr30_vgpr31
@@ -37812,8 +37825,8 @@ define <14 x double> @bitcast_v56i16_to_v14f64(<56 x i16> %a, i32 %b) {
; GFX9-NEXT: v_lshrrev_b32_e32 v61, 16, v46
; GFX9-NEXT: v_lshrrev_b32_e32 v33, 16, v47
; GFX9-NEXT: v_lshrrev_b32_e32 v34, 16, v56
-; GFX9-NEXT: v_lshrrev_b32_e32 v62, 16, v57
-; GFX9-NEXT: v_lshrrev_b32_e32 v35, 16, v58
+; GFX9-NEXT: v_lshrrev_b32_e32 v35, 16, v57
+; GFX9-NEXT: v_lshrrev_b32_e32 v62, 16, v58
; GFX9-NEXT: v_lshrrev_b32_e32 v36, 16, v59
; GFX9-NEXT: v_cmp_ne_u32_e32 vcc, 0, v28
; GFX9-NEXT: buffer_store_dword v0, off, s[0:3], s32 offset:156 ; 4-byte Folded Spill
@@ -37840,8 +37853,8 @@ define <14 x double> @bitcast_v56i16_to_v14f64(<56 x i16> %a, i32 %b) {
; GFX9-NEXT: v_lshrrev_b32_e32 v21, 16, v38
; GFX9-NEXT: v_lshrrev_b32_e32 v22, 16, v37
; GFX9-NEXT: v_perm_b32 v0, v36, v59, s6
-; GFX9-NEXT: v_perm_b32 v1, v35, v58, s6
-; GFX9-NEXT: v_perm_b32 v2, v62, v57, s6
+; GFX9-NEXT: v_perm_b32 v1, v62, v58, s6
+; GFX9-NEXT: v_perm_b32 v2, v35, v57, s6
; GFX9-NEXT: v_perm_b32 v3, v34, v56, s6
; GFX9-NEXT: v_perm_b32 v4, v33, v47, s6
; GFX9-NEXT: v_perm_b32 v5, v61, v46, s6
@@ -37891,8 +37904,8 @@ define <14 x double> @bitcast_v56i16_to_v14f64(<56 x i16> %a, i32 %b) {
; GFX9-NEXT: ; implicit-def: $vgpr61
; GFX9-NEXT: ; implicit-def: $vgpr33
; GFX9-NEXT: ; implicit-def: $vgpr34
-; GFX9-NEXT: ; implicit-def: $vgpr62
; GFX9-NEXT: ; implicit-def: $vgpr35
+; GFX9-NEXT: ; implicit-def: $vgpr62
; GFX9-NEXT: ; implicit-def: $vgpr36
; GFX9-NEXT: s_waitcnt vmcnt(0)
; GFX9-NEXT: v_perm_b32 v23, v24, v23, s6
@@ -37982,8 +37995,8 @@ define <14 x double> @bitcast_v56i16_to_v14f64(<56 x i16> %a, i32 %b) {
; GFX9-NEXT: buffer_load_dword v24, off, s[0:3], s32 offset:100 ; 4-byte Folded Reload
; GFX9-NEXT: s_mov_b32 s6, 0x5040100
; GFX9-NEXT: v_perm_b32 v0, v36, v59, s6
-; GFX9-NEXT: v_perm_b32 v1, v35, v58, s6
-; GFX9-NEXT: v_perm_b32 v2, v62, v57, s6
+; GFX9-NEXT: v_perm_b32 v1, v62, v58, s6
+; GFX9-NEXT: v_perm_b32 v2, v35, v57, s6
; GFX9-NEXT: v_perm_b32 v3, v34, v56, s6
; GFX9-NEXT: v_perm_b32 v4, v33, v47, s6
; GFX9-NEXT: v_perm_b32 v5, v61, v46, s6
@@ -39530,18 +39543,18 @@ define <56 x half> @bitcast_v14f64_to_v56f16(<14 x double> %a, i32 %b) {
; SI-NEXT: ; implicit-def: $vgpr56
; SI-NEXT: ; implicit-def: $vgpr46
; SI-NEXT: ; implicit-def: $vgpr44
+; SI-NEXT: ; implicit-def: $vgpr47
; SI-NEXT: ; implicit-def: $vgpr57
; SI-NEXT: ; implicit-def: $vgpr59
; SI-NEXT: ; implicit-def: $vgpr60
-; SI-NEXT: ; implicit-def: $vgpr42
; SI-NEXT: ; implicit-def: $vgpr62
; SI-NEXT: ; kill: killed $vgpr29
; SI-NEXT: ; implicit-def: $vgpr29
-; SI-NEXT: ; implicit-def: $vgpr47
-; SI-NEXT: ; implicit-def: $vgpr31
; SI-NEXT: ; implicit-def: $vgpr45
-; SI-NEXT: ; implicit-def: $vgpr33
+; SI-NEXT: ; implicit-def: $vgpr31
; SI-NEXT: ; implicit-def: $vgpr43
+; SI-NEXT: ; implicit-def: $vgpr33
+; SI-NEXT: ; implicit-def: $vgpr42
; SI-NEXT: ; implicit-def: $vgpr34
; SI-NEXT: s_and_saveexec_b64 s[4:5], vcc
; SI-NEXT: s_xor_b64 s[4:5], exec, s[4:5]
@@ -39559,13 +39572,13 @@ define <56 x half> @bitcast_v14f64_to_v56f16(<14 x double> %a, i32 %b) {
; SI-NEXT: s_waitcnt expcnt(1)
; SI-NEXT: v_cvt_f32_f16_e32 v62, v29
; SI-NEXT: v_lshrrev_b32_e32 v29, 16, v23
-; SI-NEXT: v_cvt_f32_f16_e32 v42, v29
-; SI-NEXT: v_lshrrev_b32_e32 v29, 16, v22
; SI-NEXT: v_cvt_f32_f16_e32 v60, v29
-; SI-NEXT: v_lshrrev_b32_e32 v29, 16, v21
+; SI-NEXT: v_lshrrev_b32_e32 v29, 16, v22
; SI-NEXT: v_cvt_f32_f16_e32 v59, v29
-; SI-NEXT: v_lshrrev_b32_e32 v29, 16, v20
+; SI-NEXT: v_lshrrev_b32_e32 v29, 16, v21
; SI-NEXT: v_cvt_f32_f16_e32 v57, v29
+; SI-NEXT: v_lshrrev_b32_e32 v29, 16, v20
+; SI-NEXT: v_cvt_f32_f16_e32 v47, v29
; SI-NEXT: v_lshrrev_b32_e32 v29, 16, v19
; SI-NEXT: v_cvt_f32_f16_e32 v29, v29
; SI-NEXT: v_cvt_f32_f16_e32 v25, v25
@@ -39590,8 +39603,8 @@ define <56 x half> @bitcast_v14f64_to_v56f16(<14 x double> %a, i32 %b) {
; SI-NEXT: v_lshrrev_b32_e32 v29, 16, v16
; SI-NEXT: v_cvt_f32_f16_e32 v29, v29
; SI-NEXT: v_cvt_f32_f16_e32 v16, v16
-; SI-NEXT: v_cvt_f32_f16_e32 v43, v28
-; SI-NEXT: v_cvt_f32_f16_e32 v45, v27
+; SI-NEXT: v_cvt_f32_f16_e32 v42, v28
+; SI-NEXT: v_cvt_f32_f16_e32 v43, v27
; SI-NEXT: buffer_store_dword v29, off, s[0:3], s32 offset:96 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_lshrrev_b32_e32 v29, 16, v15
@@ -39632,7 +39645,7 @@ define <56 x half> @bitcast_v14f64_to_v56f16(<14 x double> %a, i32 %b) {
; SI-NEXT: v_lshrrev_b32_e32 v29, 16, v1
; SI-NEXT: v_cvt_f32_f16_e32 v55, v29
; SI-NEXT: v_mov_b32_e32 v29, v36
-; SI-NEXT: v_cvt_f32_f16_e32 v47, v26
+; SI-NEXT: v_cvt_f32_f16_e32 v45, v26
; SI-NEXT: buffer_store_dword v25, off, s[0:3], s32 offset:64 ; 4-byte Folded Spill
; SI-NEXT: buffer_store_dword v24, off, s[0:3], s32 offset:68 ; 4-byte Folded Spill
; SI-NEXT: buffer_store_dword v23, off, s[0:3], s32 offset:76 ; 4-byte Folded Spill
@@ -39676,41 +39689,35 @@ define <56 x half> @bitcast_v14f64_to_v56f16(<14 x double> %a, i32 %b) {
; SI-NEXT: s_andn2_saveexec_b64 s[4:5], s[4:5]
; SI-NEXT: s_cbranch_execz .LBB52_4
; SI-NEXT: ; %bb.3: ; %cmp.true
-; SI-NEXT: v_add_f64 v[19:20], v[19:20], 1.0
-; SI-NEXT: v_add_f64 v[3:4], v[3:4], 1.0
-; SI-NEXT: v_lshrrev_b32_e32 v47, 16, v19
-; SI-NEXT: v_lshrrev_b32_e32 v51, 16, v3
-; SI-NEXT: v_cvt_f32_f16_e32 v53, v3
-; SI-NEXT: v_cvt_f32_f16_e32 v3, v47
-; SI-NEXT: v_add_f64 v[17:18], v[17:18], 1.0
; SI-NEXT: v_add_f64 v[54:55], v[1:2], 1.0
-; SI-NEXT: v_lshrrev_b32_e32 v45, 16, v18
-; SI-NEXT: buffer_store_dword v3, off, s[0:3], s32 offset:72 ; 4-byte Folded Spill
-; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_cvt_f32_f16_e32 v3, v45
-; SI-NEXT: v_add_f64 v[23:24], v[23:24], 1.0
+; SI-NEXT: s_waitcnt expcnt(6)
+; SI-NEXT: v_add_f64 v[15:16], v[15:16], 1.0
; SI-NEXT: v_lshrrev_b32_e32 v1, 16, v54
; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:148 ; 4-byte Folded Spill
-; SI-NEXT: v_lshrrev_b32_e32 v43, 16, v17
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_lshrrev_b32_e32 v1, 16, v23
+; SI-NEXT: v_lshrrev_b32_e32 v1, 16, v16
; SI-NEXT: v_cvt_f32_f16_e32 v1, v1
-; SI-NEXT: buffer_store_dword v3, off, s[0:3], s32 offset:80 ; 4-byte Folded Spill
+; SI-NEXT: v_add_f64 v[19:20], v[19:20], 1.0
+; SI-NEXT: v_add_f64 v[3:4], v[3:4], 1.0
+; SI-NEXT: v_lshrrev_b32_e32 v45, 16, v19
+; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:96 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_cvt_f32_f16_e32 v3, v43
-; SI-NEXT: v_add_f64 v[15:16], v[15:16], 1.0
+; SI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:148 ; 4-byte Folded Reload
+; SI-NEXT: v_lshrrev_b32_e32 v51, 16, v3
+; SI-NEXT: v_cvt_f32_f16_e32 v53, v3
+; SI-NEXT: v_cvt_f32_f16_e32 v3, v45
+; SI-NEXT: v_add_f64 v[17:18], v[17:18], 1.0
; SI-NEXT: v_add_f64 v[5:6], v[5:6], 1.0
-; SI-NEXT: v_lshrrev_b32_e32 v42, 16, v16
-; SI-NEXT: buffer_store_dword v3, off, s[0:3], s32 offset:88 ; 4-byte Folded Spill
+; SI-NEXT: v_lshrrev_b32_e32 v43, 16, v18
+; SI-NEXT: buffer_store_dword v3, off, s[0:3], s32 offset:72 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_cvt_f32_f16_e32 v3, v42
-; SI-NEXT: v_mov_b32_e32 v42, v1
-; SI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:148 ; 4-byte Folded Reload
+; SI-NEXT: v_cvt_f32_f16_e32 v3, v43
; SI-NEXT: v_add_f64 v[7:8], v[7:8], 1.0
; SI-NEXT: v_add_f64 v[9:10], v[9:10], 1.0
; SI-NEXT: v_add_f64 v[11:12], v[11:12], 1.0
; SI-NEXT: v_add_f64 v[13:14], v[13:14], 1.0
; SI-NEXT: v_add_f64 v[21:22], v[21:22], 1.0
+; SI-NEXT: v_add_f64 v[23:24], v[23:24], 1.0
; SI-NEXT: v_add_f64 v[25:26], v[25:26], 1.0
; SI-NEXT: v_add_f64 v[27:28], v[27:28], 1.0
; SI-NEXT: v_lshrrev_b32_e32 v2, 16, v55
@@ -39726,9 +39733,11 @@ define <56 x half> @bitcast_v14f64_to_v56f16(<14 x double> %a, i32 %b) {
; SI-NEXT: v_lshrrev_b32_e32 v56, 16, v13
; SI-NEXT: v_lshrrev_b32_e32 v46, 16, v14
; SI-NEXT: v_lshrrev_b32_e32 v44, 16, v15
-; SI-NEXT: v_lshrrev_b32_e32 v57, 16, v20
-; SI-NEXT: v_lshrrev_b32_e32 v59, 16, v21
-; SI-NEXT: v_lshrrev_b32_e32 v60, 16, v22
+; SI-NEXT: v_lshrrev_b32_e32 v42, 16, v17
+; SI-NEXT: v_lshrrev_b32_e32 v47, 16, v20
+; SI-NEXT: v_lshrrev_b32_e32 v57, 16, v21
+; SI-NEXT: v_lshrrev_b32_e32 v59, 16, v22
+; SI-NEXT: v_lshrrev_b32_e32 v60, 16, v23
; SI-NEXT: v_lshrrev_b32_e32 v62, 16, v24
; SI-NEXT: v_lshrrev_b32_e32 v29, 16, v25
; SI-NEXT: v_lshrrev_b32_e32 v31, 16, v26
@@ -39769,6 +39778,10 @@ define <56 x half> @bitcast_v14f64_to_v56f16(<14 x double> %a, i32 %b) {
; SI-NEXT: v_cvt_f32_f16_e32 v60, v60
; SI-NEXT: v_cvt_f32_f16_e32 v59, v59
; SI-NEXT: v_cvt_f32_f16_e32 v57, v57
+; SI-NEXT: v_cvt_f32_f16_e32 v47, v47
+; SI-NEXT: buffer_store_dword v3, off, s[0:3], s32 offset:80 ; 4-byte Folded Spill
+; SI-NEXT: s_waitcnt expcnt(0)
+; SI-NEXT: v_cvt_f32_f16_e32 v3, v42
; SI-NEXT: v_cvt_f32_f16_e32 v44, v44
; SI-NEXT: v_cvt_f32_f16_e32 v46, v46
; SI-NEXT: v_cvt_f32_f16_e32 v56, v56
@@ -39783,11 +39796,11 @@ define <56 x half> @bitcast_v14f64_to_v56f16(<14 x double> %a, i32 %b) {
; SI-NEXT: v_cvt_f32_f16_e32 v49, v49
; SI-NEXT: v_cvt_f32_f16_e32 v51, v51
; SI-NEXT: v_cvt_f32_f16_e32 v54, v2
-; SI-NEXT: v_mov_b32_e32 v47, v26
-; SI-NEXT: v_mov_b32_e32 v45, v27
-; SI-NEXT: s_waitcnt vmcnt(0)
+; SI-NEXT: s_waitcnt vmcnt(2)
; SI-NEXT: v_cvt_f32_f16_e32 v55, v1
-; SI-NEXT: v_mov_b32_e32 v43, v28
+; SI-NEXT: v_mov_b32_e32 v45, v26
+; SI-NEXT: v_mov_b32_e32 v43, v27
+; SI-NEXT: v_mov_b32_e32 v42, v28
; SI-NEXT: buffer_store_dword v25, off, s[0:3], s32 offset:64 ; 4-byte Folded Spill
; SI-NEXT: buffer_store_dword v24, off, s[0:3], s32 offset:68 ; 4-byte Folded Spill
; SI-NEXT: buffer_store_dword v23, off, s[0:3], s32 offset:76 ; 4-byte Folded Spill
@@ -39805,7 +39818,7 @@ define <56 x half> @bitcast_v14f64_to_v56f16(<14 x double> %a, i32 %b) {
; SI-NEXT: buffer_store_dword v11, off, s[0:3], s32 offset:136 ; 4-byte Folded Spill
; SI-NEXT: buffer_store_dword v10, off, s[0:3], s32 offset:140 ; 4-byte Folded Spill
; SI-NEXT: buffer_store_dword v9, off, s[0:3], s32 offset:144 ; 4-byte Folded Spill
-; SI-NEXT: buffer_store_dword v3, off, s[0:3], s32 offset:96 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v3, off, s[0:3], s32 offset:88 ; 4-byte Folded Spill
; SI-NEXT: .LBB52_4: ; %end
; SI-NEXT: s_or_b64 exec, exec, s[4:5]
; SI-NEXT: v_cvt_f16_f32_e32 v1, v55
@@ -39972,7 +39985,7 @@ define <56 x half> @bitcast_v14f64_to_v56f16(<14 x double> %a, i32 %b) {
; SI-NEXT: buffer_store_dword v1, v3, s[0:3], 0 offen
; SI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:100 ; 4-byte Folded Reload
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_cvt_f16_f32_e32 v1, v57
+; SI-NEXT: v_cvt_f16_f32_e32 v1, v47
; SI-NEXT: v_add_i32_e32 v3, vcc, 0x4c, v0
; SI-NEXT: v_lshlrev_b32_e32 v1, 16, v1
; SI-NEXT: s_waitcnt vmcnt(0)
@@ -39981,7 +39994,7 @@ define <56 x half> @bitcast_v14f64_to_v56f16(<14 x double> %a, i32 %b) {
; SI-NEXT: buffer_store_dword v1, v3, s[0:3], 0 offen
; SI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:92 ; 4-byte Folded Reload
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_cvt_f16_f32_e32 v1, v59
+; SI-NEXT: v_cvt_f16_f32_e32 v1, v57
; SI-NEXT: v_add_i32_e32 v3, vcc, 0x50, v0
; SI-NEXT: v_lshlrev_b32_e32 v1, 16, v1
; SI-NEXT: s_waitcnt vmcnt(0)
@@ -39990,7 +40003,7 @@ define <56 x half> @bitcast_v14f64_to_v56f16(<14 x double> %a, i32 %b) {
; SI-NEXT: buffer_store_dword v1, v3, s[0:3], 0 offen
; SI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:84 ; 4-byte Folded Reload
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_cvt_f16_f32_e32 v1, v60
+; SI-NEXT: v_cvt_f16_f32_e32 v1, v59
; SI-NEXT: v_add_i32_e32 v3, vcc, 0x54, v0
; SI-NEXT: v_lshlrev_b32_e32 v1, 16, v1
; SI-NEXT: s_waitcnt vmcnt(0)
@@ -39999,7 +40012,7 @@ define <56 x half> @bitcast_v14f64_to_v56f16(<14 x double> %a, i32 %b) {
; SI-NEXT: buffer_store_dword v1, v3, s[0:3], 0 offen
; SI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:76 ; 4-byte Folded Reload
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_cvt_f16_f32_e32 v1, v42
+; SI-NEXT: v_cvt_f16_f32_e32 v1, v60
; SI-NEXT: v_add_i32_e32 v3, vcc, 0x58, v0
; SI-NEXT: v_lshlrev_b32_e32 v1, 16, v1
; SI-NEXT: s_waitcnt vmcnt(0)
@@ -40026,21 +40039,21 @@ define <56 x half> @bitcast_v14f64_to_v56f16(<14 x double> %a, i32 %b) {
; SI-NEXT: buffer_store_dword v1, v3, s[0:3], 0 offen
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_cvt_f16_f32_e32 v1, v31
-; SI-NEXT: v_cvt_f16_f32_e32 v2, v47
+; SI-NEXT: v_cvt_f16_f32_e32 v2, v45
; SI-NEXT: v_add_i32_e32 v3, vcc, 0x64, v0
; SI-NEXT: v_lshlrev_b32_e32 v1, 16, v1
; SI-NEXT: v_or_b32_e32 v1, v2, v1
; SI-NEXT: buffer_store_dword v1, v3, s[0:3], 0 offen
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_cvt_f16_f32_e32 v1, v33
-; SI-NEXT: v_cvt_f16_f32_e32 v2, v45
+; SI-NEXT: v_cvt_f16_f32_e32 v2, v43
; SI-NEXT: v_add_i32_e32 v3, vcc, 0x68, v0
; SI-NEXT: v_lshlrev_b32_e32 v1, 16, v1
; SI-NEXT: v_or_b32_e32 v1, v2, v1
; SI-NEXT: buffer_store_dword v1, v3, s[0:3], 0 offen
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_cvt_f16_f32_e32 v1, v34
-; SI-NEXT: v_cvt_f16_f32_e32 v2, v43
+; SI-NEXT: v_cvt_f16_f32_e32 v2, v42
; SI-NEXT: v_add_i32_e32 v0, vcc, 0x6c, v0
; SI-NEXT: v_lshlrev_b32_e32 v1, 16, v1
; SI-NEXT: v_or_b32_e32 v1, v2, v1
@@ -40638,23 +40651,24 @@ define inreg <56 x half> @bitcast_v14f64_to_v56f16_scalar(<14 x double> inreg %a
; SI-NEXT: s_cbranch_scc0 .LBB53_4
; SI-NEXT: ; %bb.1: ; %cmp.false
; SI-NEXT: s_lshr_b32 s42, s5, 16
-; SI-NEXT: v_cvt_f32_f16_e32 v29, s42
+; SI-NEXT: s_waitcnt expcnt(3)
+; SI-NEXT: v_cvt_f32_f16_e32 v60, s42
; SI-NEXT: s_lshr_b32 s42, s4, 16
-; SI-NEXT: v_cvt_f32_f16_e32 v25, s42
-; SI-NEXT: s_lshr_b32 s42, s7, 16
; SI-NEXT: v_cvt_f32_f16_e32 v21, s42
+; SI-NEXT: s_lshr_b32 s42, s7, 16
+; SI-NEXT: v_cvt_f32_f16_e32 v2, s42
; SI-NEXT: s_lshr_b32 s42, s6, 16
; SI-NEXT: v_cvt_f32_f16_e32 v1, s42
; SI-NEXT: s_lshr_b32 s42, s9, 16
-; SI-NEXT: v_cvt_f32_f16_e32 v17, s42
+; SI-NEXT: v_cvt_f32_f16_e32 v10, s42
; SI-NEXT: s_lshr_b32 s42, s8, 16
-; SI-NEXT: v_cvt_f32_f16_e32 v2, s42
+; SI-NEXT: v_cvt_f32_f16_e32 v9, s42
; SI-NEXT: s_lshr_b32 s42, s11, 16
-; SI-NEXT: v_cvt_f32_f16_e32 v10, s42
+; SI-NEXT: v_cvt_f32_f16_e32 v58, s42
; SI-NEXT: s_lshr_b32 s42, s10, 16
; SI-NEXT: v_cvt_f32_f16_e32 v13, s42
; SI-NEXT: s_lshr_b32 s42, s13, 16
-; SI-NEXT: v_cvt_f32_f16_e32 v48, s42
+; SI-NEXT: v_cvt_f32_f16_e32 v59, s42
; SI-NEXT: s_lshr_b32 s42, s12, 16
; SI-NEXT: v_cvt_f32_f16_e32 v18, s42
; SI-NEXT: s_lshr_b32 s42, s15, 16
@@ -40693,16 +40707,12 @@ define inreg <56 x half> @bitcast_v14f64_to_v56f16_scalar(<14 x double> inreg %a
; SI-NEXT: v_cvt_f32_f16_e32 v46, s42
; SI-NEXT: s_lshr_b32 s42, s16, 16
; SI-NEXT: v_cvt_f32_f16_e32 v56, s42
-; SI-NEXT: s_waitcnt expcnt(6)
; SI-NEXT: v_cvt_f32_f16_e32 v57, s5
-; SI-NEXT: v_cvt_f32_f16_e32 v33, s4
-; SI-NEXT: v_cvt_f32_f16_e32 v34, s7
-; SI-NEXT: s_waitcnt expcnt(5)
-; SI-NEXT: v_cvt_f32_f16_e32 v58, s6
-; SI-NEXT: s_waitcnt expcnt(4)
-; SI-NEXT: v_cvt_f32_f16_e32 v59, s9
-; SI-NEXT: s_waitcnt expcnt(3)
-; SI-NEXT: v_cvt_f32_f16_e32 v60, s8
+; SI-NEXT: v_cvt_f32_f16_e32 v25, s4
+; SI-NEXT: v_cvt_f32_f16_e32 v29, s7
+; SI-NEXT: v_cvt_f32_f16_e32 v33, s6
+; SI-NEXT: v_cvt_f32_f16_e32 v34, s9
+; SI-NEXT: v_cvt_f32_f16_e32 v48, s8
; SI-NEXT: v_cvt_f32_f16_e32 v16, s11
; SI-NEXT: v_cvt_f32_f16_e32 v7, s10
; SI-NEXT: v_cvt_f32_f16_e32 v19, s13
@@ -40739,14 +40749,23 @@ define inreg <56 x half> @bitcast_v14f64_to_v56f16_scalar(<14 x double> inreg %a
; SI-NEXT: buffer_store_dword v3, off, s[0:3], s32 offset:72 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_lshrrev_b32_e32 v3, 16, v43
-; SI-NEXT: v_add_f64 v[22:23], s[14:15], 1.0
; SI-NEXT: buffer_store_dword v3, off, s[0:3], s32 offset:76 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_add_f64 v[3:4], s[4:5], 1.0
+; SI-NEXT: v_add_f64 v[22:23], s[14:15], 1.0
+; SI-NEXT: v_lshrrev_b32_e32 v59, 16, v4
; SI-NEXT: v_lshrrev_b32_e32 v5, 16, v22
-; SI-NEXT: v_lshrrev_b32_e32 v29, 16, v4
; SI-NEXT: v_cvt_f32_f16_e32 v57, v4
; SI-NEXT: v_cvt_f32_f16_e32 v4, v22
+; SI-NEXT: v_cvt_f32_f16_e32 v22, v59
+; SI-NEXT: v_add_f64 v[18:19], s[12:13], 1.0
+; SI-NEXT: v_add_f64 v[7:8], s[6:7], 1.0
+; SI-NEXT: v_lshrrev_b32_e32 v60, 16, v18
+; SI-NEXT: v_lshrrev_b32_e32 v17, 16, v8
+; SI-NEXT: v_cvt_f32_f16_e32 v29, v8
+; SI-NEXT: v_cvt_f32_f16_e32 v8, v18
+; SI-NEXT: v_cvt_f32_f16_e32 v18, v60
+; SI-NEXT: v_mov_b32_e32 v60, v22
; SI-NEXT: v_cvt_f32_f16_e32 v22, v5
; SI-NEXT: buffer_load_dword v5, off, s[0:3], s32 offset:76 ; 4-byte Folded Reload
; SI-NEXT: v_cvt_f32_f16_e32 v41, v43
@@ -40754,13 +40773,12 @@ define inreg <56 x half> @bitcast_v14f64_to_v56f16_scalar(<14 x double> inreg %a
; SI-NEXT: v_add_f64 v[49:50], s[22:23], 1.0
; SI-NEXT: v_add_f64 v[37:38], s[24:25], 1.0
; SI-NEXT: v_add_f64 v[15:16], s[10:11], 1.0
-; SI-NEXT: v_add_f64 v[7:8], s[6:7], 1.0
; SI-NEXT: v_lshrrev_b32_e32 v44, 16, v49
; SI-NEXT: v_lshrrev_b32_e32 v51, 16, v37
; SI-NEXT: v_lshrrev_b32_e32 v39, 16, v38
; SI-NEXT: v_lshrrev_b32_e32 v13, 16, v15
-; SI-NEXT: v_lshrrev_b32_e32 v17, 16, v7
-; SI-NEXT: v_cvt_f32_f16_e32 v58, v7
+; SI-NEXT: v_lshrrev_b32_e32 v14, 16, v7
+; SI-NEXT: v_cvt_f32_f16_e32 v33, v7
; SI-NEXT: v_cvt_f32_f16_e32 v7, v15
; SI-NEXT: v_cvt_f32_f16_e32 v15, v37
; SI-NEXT: v_cvt_f32_f16_e32 v37, v39
@@ -40774,7 +40792,6 @@ define inreg <56 x half> @bitcast_v14f64_to_v56f16_scalar(<14 x double> inreg %a
; SI-NEXT: v_cvt_f32_f16_e32 v53, v46
; SI-NEXT: v_add_f64 v[30:31], s[28:29], 1.0
; SI-NEXT: v_add_f64 v[26:27], s[40:41], 1.0
-; SI-NEXT: v_add_f64 v[18:19], s[12:13], 1.0
; SI-NEXT: v_add_f64 v[11:12], s[8:9], 1.0
; SI-NEXT: v_lshrrev_b32_e32 v55, 16, v50
; SI-NEXT: v_lshrrev_b32_e32 v61, 16, v35
@@ -40784,20 +40801,16 @@ define inreg <56 x half> @bitcast_v14f64_to_v56f16_scalar(<14 x double> inreg %a
; SI-NEXT: v_lshrrev_b32_e32 v63, 16, v26
; SI-NEXT: v_lshrrev_b32_e32 v24, 16, v27
; SI-NEXT: v_lshrrev_b32_e32 v20, 16, v23
-; SI-NEXT: v_lshrrev_b32_e32 v6, 16, v18
-; SI-NEXT: v_lshrrev_b32_e32 v9, 16, v19
-; SI-NEXT: v_lshrrev_b32_e32 v10, 16, v16
-; SI-NEXT: v_lshrrev_b32_e32 v14, 16, v11
-; SI-NEXT: v_lshrrev_b32_e32 v48, 16, v12
-; SI-NEXT: v_lshrrev_b32_e32 v21, 16, v8
-; SI-NEXT: v_lshrrev_b32_e32 v25, 16, v3
-; SI-NEXT: v_cvt_f32_f16_e32 v33, v3
-; SI-NEXT: v_cvt_f32_f16_e32 v34, v8
-; SI-NEXT: v_cvt_f32_f16_e32 v59, v12
-; SI-NEXT: v_cvt_f32_f16_e32 v60, v11
+; SI-NEXT: v_lshrrev_b32_e32 v6, 16, v19
+; SI-NEXT: v_lshrrev_b32_e32 v58, 16, v16
+; SI-NEXT: v_lshrrev_b32_e32 v9, 16, v11
+; SI-NEXT: v_lshrrev_b32_e32 v10, 16, v12
+; SI-NEXT: v_lshrrev_b32_e32 v21, 16, v3
+; SI-NEXT: v_cvt_f32_f16_e32 v25, v3
+; SI-NEXT: v_cvt_f32_f16_e32 v34, v12
+; SI-NEXT: v_cvt_f32_f16_e32 v48, v11
; SI-NEXT: v_cvt_f32_f16_e32 v16, v16
; SI-NEXT: v_cvt_f32_f16_e32 v19, v19
-; SI-NEXT: v_cvt_f32_f16_e32 v8, v18
; SI-NEXT: v_cvt_f32_f16_e32 v23, v23
; SI-NEXT: v_cvt_f32_f16_e32 v27, v27
; SI-NEXT: v_cvt_f32_f16_e32 v11, v26
@@ -40811,16 +40824,14 @@ define inreg <56 x half> @bitcast_v14f64_to_v56f16_scalar(<14 x double> inreg %a
; SI-NEXT: v_cvt_f32_f16_e32 v54, v54
; SI-NEXT: v_cvt_f32_f16_e32 v45, v2
; SI-NEXT: v_cvt_f32_f16_e32 v47, v1
-; SI-NEXT: v_cvt_f32_f16_e32 v29, v29
-; SI-NEXT: v_cvt_f32_f16_e32 v25, v25
; SI-NEXT: v_cvt_f32_f16_e32 v21, v21
-; SI-NEXT: v_cvt_f32_f16_e32 v1, v17
-; SI-NEXT: v_cvt_f32_f16_e32 v17, v48
-; SI-NEXT: v_cvt_f32_f16_e32 v2, v14
+; SI-NEXT: v_cvt_f32_f16_e32 v2, v17
+; SI-NEXT: v_cvt_f32_f16_e32 v1, v14
; SI-NEXT: v_cvt_f32_f16_e32 v10, v10
+; SI-NEXT: v_cvt_f32_f16_e32 v9, v9
+; SI-NEXT: v_cvt_f32_f16_e32 v58, v58
; SI-NEXT: v_cvt_f32_f16_e32 v13, v13
-; SI-NEXT: v_cvt_f32_f16_e32 v48, v9
-; SI-NEXT: v_cvt_f32_f16_e32 v18, v6
+; SI-NEXT: v_cvt_f32_f16_e32 v59, v6
; SI-NEXT: v_cvt_f32_f16_e32 v20, v20
; SI-NEXT: v_cvt_f32_f16_e32 v24, v24
; SI-NEXT: v_cvt_f32_f16_e32 v26, v63
@@ -40828,11 +40839,11 @@ define inreg <56 x half> @bitcast_v14f64_to_v56f16_scalar(<14 x double> inreg %a
; SI-NEXT: v_cvt_f32_f16_e32 v30, v62
; SI-NEXT: v_cvt_f32_f16_e32 v32, v32
; SI-NEXT: v_cvt_f32_f16_e32 v35, v61
+; SI-NEXT: v_cvt_f32_f16_e32 v49, v55
+; SI-NEXT: v_cvt_f32_f16_e32 v55, v56
; SI-NEXT: s_waitcnt vmcnt(0)
; SI-NEXT: v_cvt_f32_f16_e32 v42, v5
; SI-NEXT: buffer_load_dword v5, off, s[0:3], s32 offset:72 ; 4-byte Folded Reload
-; SI-NEXT: v_cvt_f32_f16_e32 v49, v55
-; SI-NEXT: v_cvt_f32_f16_e32 v55, v56
; SI-NEXT: s_waitcnt vmcnt(0)
; SI-NEXT: v_cvt_f32_f16_e32 v44, v5
; SI-NEXT: buffer_load_dword v5, off, s[0:3], s32 offset:68 ; 4-byte Folded Reload
@@ -40844,193 +40855,193 @@ define inreg <56 x half> @bitcast_v14f64_to_v56f16_scalar(<14 x double> inreg %a
; SI-NEXT: .LBB53_3: ; %end
; SI-NEXT: v_cvt_f16_f32_e32 v5, v56
; SI-NEXT: v_cvt_f16_f32_e32 v6, v47
-; SI-NEXT: v_cvt_f16_f32_e32 v9, v46
-; SI-NEXT: v_cvt_f16_f32_e32 v14, v45
+; SI-NEXT: v_cvt_f16_f32_e32 v14, v46
+; SI-NEXT: v_cvt_f16_f32_e32 v17, v45
; SI-NEXT: v_lshlrev_b32_e32 v5, 16, v5
; SI-NEXT: v_or_b32_e32 v5, v6, v5
-; SI-NEXT: v_lshlrev_b32_e32 v6, 16, v9
+; SI-NEXT: v_lshlrev_b32_e32 v6, 16, v14
; SI-NEXT: buffer_store_dword v5, v0, s[0:3], 0 offen
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_add_i32_e32 v5, vcc, 4, v0
-; SI-NEXT: v_or_b32_e32 v6, v14, v6
+; SI-NEXT: v_or_b32_e32 v6, v17, v6
; SI-NEXT: buffer_store_dword v6, v5, s[0:3], 0 offen
; SI-NEXT: v_cvt_f16_f32_e32 v5, v44
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_cvt_f16_f32_e32 v6, v43
-; SI-NEXT: v_add_i32_e32 v9, vcc, 8, v0
+; SI-NEXT: v_add_i32_e32 v14, vcc, 8, v0
; SI-NEXT: v_lshlrev_b32_e32 v5, 16, v5
; SI-NEXT: v_or_b32_e32 v5, v6, v5
-; SI-NEXT: buffer_store_dword v5, v9, s[0:3], 0 offen
+; SI-NEXT: buffer_store_dword v5, v14, s[0:3], 0 offen
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_cvt_f16_f32_e32 v5, v42
; SI-NEXT: v_cvt_f16_f32_e32 v6, v41
-; SI-NEXT: v_add_i32_e32 v9, vcc, 12, v0
+; SI-NEXT: v_add_i32_e32 v14, vcc, 12, v0
; SI-NEXT: v_lshlrev_b32_e32 v5, 16, v5
; SI-NEXT: v_or_b32_e32 v5, v6, v5
-; SI-NEXT: buffer_store_dword v5, v9, s[0:3], 0 offen
+; SI-NEXT: buffer_store_dword v5, v14, s[0:3], 0 offen
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_cvt_f16_f32_e32 v5, v55
; SI-NEXT: v_cvt_f16_f32_e32 v6, v40
-; SI-NEXT: v_add_i32_e32 v9, vcc, 16, v0
+; SI-NEXT: v_add_i32_e32 v14, vcc, 16, v0
; SI-NEXT: v_lshlrev_b32_e32 v5, 16, v5
; SI-NEXT: v_or_b32_e32 v5, v6, v5
-; SI-NEXT: buffer_store_dword v5, v9, s[0:3], 0 offen
+; SI-NEXT: buffer_store_dword v5, v14, s[0:3], 0 offen
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_cvt_f16_f32_e32 v5, v53
; SI-NEXT: v_cvt_f16_f32_e32 v6, v54
-; SI-NEXT: v_add_i32_e32 v9, vcc, 20, v0
+; SI-NEXT: v_add_i32_e32 v14, vcc, 20, v0
; SI-NEXT: v_lshlrev_b32_e32 v5, 16, v5
; SI-NEXT: v_or_b32_e32 v5, v6, v5
-; SI-NEXT: buffer_store_dword v5, v9, s[0:3], 0 offen
+; SI-NEXT: buffer_store_dword v5, v14, s[0:3], 0 offen
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_cvt_f16_f32_e32 v5, v51
; SI-NEXT: v_cvt_f16_f32_e32 v6, v52
-; SI-NEXT: v_add_i32_e32 v9, vcc, 24, v0
+; SI-NEXT: v_add_i32_e32 v14, vcc, 24, v0
; SI-NEXT: v_lshlrev_b32_e32 v5, 16, v5
; SI-NEXT: v_or_b32_e32 v5, v6, v5
-; SI-NEXT: buffer_store_dword v5, v9, s[0:3], 0 offen
+; SI-NEXT: buffer_store_dword v5, v14, s[0:3], 0 offen
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_cvt_f16_f32_e32 v5, v49
; SI-NEXT: v_cvt_f16_f32_e32 v6, v50
-; SI-NEXT: v_add_i32_e32 v9, vcc, 28, v0
+; SI-NEXT: v_add_i32_e32 v14, vcc, 28, v0
; SI-NEXT: v_lshlrev_b32_e32 v5, 16, v5
; SI-NEXT: v_or_b32_e32 v5, v6, v5
-; SI-NEXT: buffer_store_dword v5, v9, s[0:3], 0 offen
+; SI-NEXT: buffer_store_dword v5, v14, s[0:3], 0 offen
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_cvt_f16_f32_e32 v5, v39
; SI-NEXT: v_cvt_f16_f32_e32 v6, v15
-; SI-NEXT: v_add_i32_e32 v9, vcc, 32, v0
+; SI-NEXT: v_add_i32_e32 v14, vcc, 32, v0
; SI-NEXT: v_lshlrev_b32_e32 v5, 16, v5
; SI-NEXT: v_or_b32_e32 v5, v6, v5
-; SI-NEXT: buffer_store_dword v5, v9, s[0:3], 0 offen
+; SI-NEXT: buffer_store_dword v5, v14, s[0:3], 0 offen
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_cvt_f16_f32_e32 v5, v37
; SI-NEXT: v_cvt_f16_f32_e32 v6, v38
-; SI-NEXT: v_add_i32_e32 v9, vcc, 36, v0
+; SI-NEXT: v_add_i32_e32 v14, vcc, 36, v0
; SI-NEXT: v_lshlrev_b32_e32 v5, 16, v5
; SI-NEXT: v_or_b32_e32 v5, v6, v5
-; SI-NEXT: buffer_store_dword v5, v9, s[0:3], 0 offen
+; SI-NEXT: buffer_store_dword v5, v14, s[0:3], 0 offen
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_cvt_f16_f32_e32 v5, v35
; SI-NEXT: v_cvt_f16_f32_e32 v6, v36
-; SI-NEXT: v_add_i32_e32 v9, vcc, 40, v0
+; SI-NEXT: v_add_i32_e32 v14, vcc, 40, v0
; SI-NEXT: v_lshlrev_b32_e32 v5, 16, v5
; SI-NEXT: v_or_b32_e32 v5, v6, v5
-; SI-NEXT: buffer_store_dword v5, v9, s[0:3], 0 offen
+; SI-NEXT: buffer_store_dword v5, v14, s[0:3], 0 offen
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_cvt_f16_f32_e32 v5, v32
; SI-NEXT: v_cvt_f16_f32_e32 v6, v3
-; SI-NEXT: v_add_i32_e32 v9, vcc, 44, v0
+; SI-NEXT: v_add_i32_e32 v14, vcc, 44, v0
; SI-NEXT: v_lshlrev_b32_e32 v5, 16, v5
; SI-NEXT: v_or_b32_e32 v5, v6, v5
-; SI-NEXT: buffer_store_dword v5, v9, s[0:3], 0 offen
+; SI-NEXT: buffer_store_dword v5, v14, s[0:3], 0 offen
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_cvt_f16_f32_e32 v5, v30
; SI-NEXT: v_cvt_f16_f32_e32 v6, v12
-; SI-NEXT: v_add_i32_e32 v9, vcc, 48, v0
+; SI-NEXT: v_add_i32_e32 v14, vcc, 48, v0
; SI-NEXT: v_lshlrev_b32_e32 v5, 16, v5
; SI-NEXT: v_or_b32_e32 v5, v6, v5
-; SI-NEXT: buffer_store_dword v5, v9, s[0:3], 0 offen
+; SI-NEXT: buffer_store_dword v5, v14, s[0:3], 0 offen
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_cvt_f16_f32_e32 v5, v28
; SI-NEXT: v_cvt_f16_f32_e32 v6, v31
-; SI-NEXT: v_add_i32_e32 v9, vcc, 52, v0
+; SI-NEXT: v_add_i32_e32 v14, vcc, 52, v0
; SI-NEXT: v_lshlrev_b32_e32 v5, 16, v5
; SI-NEXT: v_or_b32_e32 v5, v6, v5
-; SI-NEXT: buffer_store_dword v5, v9, s[0:3], 0 offen
+; SI-NEXT: buffer_store_dword v5, v14, s[0:3], 0 offen
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_cvt_f16_f32_e32 v5, v26
; SI-NEXT: v_cvt_f16_f32_e32 v6, v11
-; SI-NEXT: v_add_i32_e32 v9, vcc, 56, v0
+; SI-NEXT: v_add_i32_e32 v14, vcc, 56, v0
; SI-NEXT: v_lshlrev_b32_e32 v5, 16, v5
; SI-NEXT: v_or_b32_e32 v5, v6, v5
-; SI-NEXT: buffer_store_dword v5, v9, s[0:3], 0 offen
+; SI-NEXT: buffer_store_dword v5, v14, s[0:3], 0 offen
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_cvt_f16_f32_e32 v5, v24
; SI-NEXT: v_cvt_f16_f32_e32 v6, v27
-; SI-NEXT: v_add_i32_e32 v9, vcc, 60, v0
+; SI-NEXT: v_add_i32_e32 v14, vcc, 60, v0
; SI-NEXT: v_lshlrev_b32_e32 v5, 16, v5
; SI-NEXT: v_or_b32_e32 v5, v6, v5
-; SI-NEXT: buffer_store_dword v5, v9, s[0:3], 0 offen
+; SI-NEXT: buffer_store_dword v5, v14, s[0:3], 0 offen
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_cvt_f16_f32_e32 v5, v22
-; SI-NEXT: v_cvt_f16_f32_e32 v4, v4
-; SI-NEXT: v_add_i32_e32 v6, vcc, 64, v0
+; SI-NEXT: v_cvt_f16_f32_e32 v6, v4
+; SI-NEXT: v_add_i32_e32 v14, vcc, 64, v0
; SI-NEXT: v_lshlrev_b32_e32 v5, 16, v5
-; SI-NEXT: v_or_b32_e32 v4, v4, v5
-; SI-NEXT: buffer_store_dword v4, v6, s[0:3], 0 offen
+; SI-NEXT: v_or_b32_e32 v5, v6, v5
+; SI-NEXT: buffer_store_dword v5, v14, s[0:3], 0 offen
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_cvt_f16_f32_e32 v4, v20
-; SI-NEXT: v_cvt_f16_f32_e32 v5, v23
-; SI-NEXT: v_add_i32_e32 v6, vcc, 0x44, v0
-; SI-NEXT: v_lshlrev_b32_e32 v4, 16, v4
-; SI-NEXT: v_or_b32_e32 v4, v5, v4
-; SI-NEXT: buffer_store_dword v4, v6, s[0:3], 0 offen
+; SI-NEXT: v_cvt_f16_f32_e32 v5, v20
+; SI-NEXT: v_cvt_f16_f32_e32 v6, v23
+; SI-NEXT: v_add_i32_e32 v14, vcc, 0x44, v0
+; SI-NEXT: v_lshlrev_b32_e32 v5, 16, v5
+; SI-NEXT: v_or_b32_e32 v5, v6, v5
+; SI-NEXT: buffer_store_dword v5, v14, s[0:3], 0 offen
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_cvt_f16_f32_e32 v4, v18
-; SI-NEXT: v_cvt_f16_f32_e32 v5, v8
-; SI-NEXT: v_add_i32_e32 v6, vcc, 0x48, v0
-; SI-NEXT: v_lshlrev_b32_e32 v4, 16, v4
-; SI-NEXT: v_or_b32_e32 v4, v5, v4
-; SI-NEXT: buffer_store_dword v4, v6, s[0:3], 0 offen
+; SI-NEXT: v_cvt_f16_f32_e32 v5, v18
+; SI-NEXT: v_cvt_f16_f32_e32 v6, v8
+; SI-NEXT: v_add_i32_e32 v8, vcc, 0x48, v0
+; SI-NEXT: v_lshlrev_b32_e32 v5, 16, v5
+; SI-NEXT: v_or_b32_e32 v5, v6, v5
+; SI-NEXT: buffer_store_dword v5, v8, s[0:3], 0 offen
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_cvt_f16_f32_e32 v4, v48
-; SI-NEXT: v_cvt_f16_f32_e32 v5, v19
-; SI-NEXT: v_add_i32_e32 v6, vcc, 0x4c, v0
-; SI-NEXT: v_lshlrev_b32_e32 v4, 16, v4
-; SI-NEXT: v_or_b32_e32 v4, v5, v4
-; SI-NEXT: buffer_store_dword v4, v6, s[0:3], 0 offen
+; SI-NEXT: v_cvt_f16_f32_e32 v5, v59
+; SI-NEXT: v_cvt_f16_f32_e32 v6, v19
+; SI-NEXT: v_add_i32_e32 v8, vcc, 0x4c, v0
+; SI-NEXT: v_lshlrev_b32_e32 v5, 16, v5
+; SI-NEXT: v_or_b32_e32 v5, v6, v5
+; SI-NEXT: buffer_store_dword v5, v8, s[0:3], 0 offen
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_cvt_f16_f32_e32 v4, v13
-; SI-NEXT: v_cvt_f16_f32_e32 v5, v7
-; SI-NEXT: v_add_i32_e32 v6, vcc, 0x50, v0
-; SI-NEXT: v_lshlrev_b32_e32 v4, 16, v4
-; SI-NEXT: v_or_b32_e32 v4, v5, v4
-; SI-NEXT: buffer_store_dword v4, v6, s[0:3], 0 offen
+; SI-NEXT: v_cvt_f16_f32_e32 v5, v13
+; SI-NEXT: v_cvt_f16_f32_e32 v6, v7
+; SI-NEXT: v_add_i32_e32 v7, vcc, 0x50, v0
+; SI-NEXT: v_lshlrev_b32_e32 v5, 16, v5
+; SI-NEXT: v_or_b32_e32 v5, v6, v5
+; SI-NEXT: buffer_store_dword v5, v7, s[0:3], 0 offen
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_cvt_f16_f32_e32 v4, v10
-; SI-NEXT: v_cvt_f16_f32_e32 v5, v16
-; SI-NEXT: v_cvt_f16_f32_e32 v2, v2
-; SI-NEXT: v_cvt_f16_f32_e32 v3, v60
-; SI-NEXT: v_lshlrev_b32_e32 v4, 16, v4
-; SI-NEXT: v_add_i32_e32 v6, vcc, 0x54, v0
-; SI-NEXT: v_or_b32_e32 v4, v5, v4
-; SI-NEXT: v_lshlrev_b32_e32 v2, 16, v2
-; SI-NEXT: buffer_store_dword v4, v6, s[0:3], 0 offen
+; SI-NEXT: v_cvt_f16_f32_e32 v5, v58
+; SI-NEXT: v_cvt_f16_f32_e32 v6, v16
+; SI-NEXT: v_add_i32_e32 v7, vcc, 0x54, v0
+; SI-NEXT: v_lshlrev_b32_e32 v5, 16, v5
+; SI-NEXT: v_or_b32_e32 v5, v6, v5
+; SI-NEXT: buffer_store_dword v5, v7, s[0:3], 0 offen
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_add_i32_e32 v4, vcc, 0x58, v0
-; SI-NEXT: v_or_b32_e32 v2, v3, v2
-; SI-NEXT: buffer_store_dword v2, v4, s[0:3], 0 offen
+; SI-NEXT: v_cvt_f16_f32_e32 v5, v9
+; SI-NEXT: v_cvt_f16_f32_e32 v3, v48
+; SI-NEXT: v_add_i32_e32 v6, vcc, 0x58, v0
+; SI-NEXT: v_lshlrev_b32_e32 v5, 16, v5
+; SI-NEXT: v_or_b32_e32 v3, v3, v5
+; SI-NEXT: buffer_store_dword v3, v6, s[0:3], 0 offen
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_cvt_f16_f32_e32 v2, v17
-; SI-NEXT: v_cvt_f16_f32_e32 v3, v59
-; SI-NEXT: v_add_i32_e32 v4, vcc, 0x5c, v0
-; SI-NEXT: v_lshlrev_b32_e32 v2, 16, v2
-; SI-NEXT: v_or_b32_e32 v2, v3, v2
+; SI-NEXT: v_cvt_f16_f32_e32 v3, v10
+; SI-NEXT: v_cvt_f16_f32_e32 v5, v34
+; SI-NEXT: v_add_i32_e32 v6, vcc, 0x5c, v0
+; SI-NEXT: v_lshlrev_b32_e32 v3, 16, v3
+; SI-NEXT: v_or_b32_e32 v3, v5, v3
; SI-NEXT: v_cvt_f16_f32_e32 v1, v1
-; SI-NEXT: buffer_store_dword v2, v4, s[0:3], 0 offen
+; SI-NEXT: buffer_store_dword v3, v6, s[0:3], 0 offen
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_cvt_f16_f32_e32 v2, v58
-; SI-NEXT: v_add_i32_e32 v3, vcc, 0x60, v0
+; SI-NEXT: v_cvt_f16_f32_e32 v3, v33
+; SI-NEXT: v_add_i32_e32 v4, vcc, 0x60, v0
; SI-NEXT: v_lshlrev_b32_e32 v1, 16, v1
-; SI-NEXT: v_or_b32_e32 v1, v2, v1
-; SI-NEXT: buffer_store_dword v1, v3, s[0:3], 0 offen
+; SI-NEXT: v_or_b32_e32 v1, v3, v1
+; SI-NEXT: buffer_store_dword v1, v4, s[0:3], 0 offen
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_cvt_f16_f32_e32 v1, v21
-; SI-NEXT: v_cvt_f16_f32_e32 v2, v34
+; SI-NEXT: v_cvt_f16_f32_e32 v1, v2
+; SI-NEXT: v_cvt_f16_f32_e32 v2, v29
; SI-NEXT: v_add_i32_e32 v3, vcc, 0x64, v0
; SI-NEXT: v_lshlrev_b32_e32 v1, 16, v1
; SI-NEXT: v_or_b32_e32 v1, v2, v1
; SI-NEXT: buffer_store_dword v1, v3, s[0:3], 0 offen
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_cvt_f16_f32_e32 v1, v25
-; SI-NEXT: v_cvt_f16_f32_e32 v2, v33
+; SI-NEXT: v_cvt_f16_f32_e32 v1, v21
+; SI-NEXT: v_cvt_f16_f32_e32 v2, v25
; SI-NEXT: v_add_i32_e32 v3, vcc, 0x68, v0
; SI-NEXT: v_lshlrev_b32_e32 v1, 16, v1
; SI-NEXT: v_or_b32_e32 v1, v2, v1
; SI-NEXT: buffer_store_dword v1, v3, s[0:3], 0 offen
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_cvt_f16_f32_e32 v1, v29
+; SI-NEXT: v_cvt_f16_f32_e32 v1, v60
; SI-NEXT: v_cvt_f16_f32_e32 v2, v57
; SI-NEXT: v_add_i32_e32 v0, vcc, 0x6c, v0
; SI-NEXT: v_lshlrev_b32_e32 v1, 16, v1
@@ -41094,23 +41105,23 @@ define inreg <56 x half> @bitcast_v14f64_to_v56f16_scalar(<14 x double> inreg %a
; SI-NEXT: ; implicit-def: $vgpr8
; SI-NEXT: ; implicit-def: $vgpr18
; SI-NEXT: ; implicit-def: $vgpr19
-; SI-NEXT: ; implicit-def: $vgpr48
+; SI-NEXT: ; implicit-def: $vgpr59
; SI-NEXT: ; implicit-def: $vgpr7
; SI-NEXT: ; implicit-def: $vgpr13
; SI-NEXT: ; implicit-def: $vgpr16
-; SI-NEXT: ; implicit-def: $vgpr10
-; SI-NEXT: ; implicit-def: $vgpr60
-; SI-NEXT: ; implicit-def: $vgpr2
-; SI-NEXT: ; implicit-def: $vgpr59
-; SI-NEXT: ; implicit-def: $vgpr17
; SI-NEXT: ; implicit-def: $vgpr58
-; SI-NEXT: ; implicit-def: $vgpr1
+; SI-NEXT: ; implicit-def: $vgpr48
+; SI-NEXT: ; implicit-def: $vgpr9
; SI-NEXT: ; implicit-def: $vgpr34
-; SI-NEXT: ; implicit-def: $vgpr21
+; SI-NEXT: ; implicit-def: $vgpr10
; SI-NEXT: ; implicit-def: $vgpr33
+; SI-NEXT: ; implicit-def: $vgpr1
+; SI-NEXT: ; implicit-def: $vgpr29
+; SI-NEXT: ; implicit-def: $vgpr2
; SI-NEXT: ; implicit-def: $vgpr25
+; SI-NEXT: ; implicit-def: $vgpr21
; SI-NEXT: ; implicit-def: $vgpr57
-; SI-NEXT: ; implicit-def: $vgpr29
+; SI-NEXT: ; implicit-def: $vgpr60
; SI-NEXT: s_branch .LBB53_2
;
; VI-LABEL: bitcast_v14f64_to_v56f16_scalar:
@@ -43419,15 +43430,15 @@ define inreg <14 x double> @bitcast_v56f16_to_v14f64_scalar(<56 x half> inreg %a
; SI-NEXT: buffer_store_dword v0, off, s[0:3], s32 offset:180 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_cvt_f16_f32_e32 v0, v24
-; SI-NEXT: v_cvt_f16_f32_e32 v63, v11
-; SI-NEXT: v_cvt_f16_f32_e32 v38, v10
+; SI-NEXT: v_cvt_f16_f32_e32 v39, v11
+; SI-NEXT: v_cvt_f16_f32_e32 v33, v10
; SI-NEXT: v_cvt_f16_f32_e32 v13, v13
; SI-NEXT: buffer_store_dword v0, off, s[0:3], s32 offset:176 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_cvt_f16_f32_e32 v0, v27
-; SI-NEXT: v_cvt_f16_f32_e32 v36, v12
+; SI-NEXT: v_cvt_f16_f32_e32 v38, v12
; SI-NEXT: v_cvt_f16_f32_e32 v34, v15
-; SI-NEXT: v_cvt_f16_f32_e32 v35, v14
+; SI-NEXT: v_cvt_f16_f32_e32 v36, v14
; SI-NEXT: buffer_store_dword v0, off, s[0:3], s32 offset:172 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_cvt_f16_f32_e32 v0, v26
@@ -43520,9 +43531,9 @@ define inreg <14 x double> @bitcast_v56f16_to_v14f64_scalar(<56 x half> inreg %a
; SI-NEXT: buffer_load_dword v23, off, s[0:3], s32 offset:156 ; 4-byte Folded Reload
; SI-NEXT: buffer_load_dword v24, off, s[0:3], s32 offset:148 ; 4-byte Folded Reload
; SI-NEXT: buffer_load_dword v25, off, s[0:3], s32 offset:136 ; 4-byte Folded Reload
-; SI-NEXT: v_mov_b32_e32 v49, v2
+; SI-NEXT: v_mov_b32_e32 v63, v2
; SI-NEXT: v_lshlrev_b32_e32 v2, 16, v2
-; SI-NEXT: v_mov_b32_e32 v48, v3
+; SI-NEXT: v_mov_b32_e32 v49, v3
; SI-NEXT: v_lshlrev_b32_e32 v3, 16, v3
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_lshlrev_b32_e32 v4, 16, v4
@@ -43531,9 +43542,9 @@ define inreg <14 x double> @bitcast_v56f16_to_v14f64_scalar(<56 x half> inreg %a
; SI-NEXT: v_mov_b32_e32 v61, v44
; SI-NEXT: v_lshlrev_b32_e32 v0, 16, v43
; SI-NEXT: v_lshlrev_b32_e32 v1, 16, v54
-; SI-NEXT: v_mov_b32_e32 v39, v11
+; SI-NEXT: v_mov_b32_e32 v48, v11
; SI-NEXT: v_or_b32_e32 v2, v11, v2
-; SI-NEXT: v_mov_b32_e32 v33, v10
+; SI-NEXT: v_mov_b32_e32 v35, v10
; SI-NEXT: v_or_b32_e32 v3, v10, v3
; SI-NEXT: v_or_b32_e32 v4, v9, v4
; SI-NEXT: v_or_b32_e32 v5, v8, v5
@@ -43543,7 +43554,7 @@ define inreg <14 x double> @bitcast_v56f16_to_v14f64_scalar(<56 x half> inreg %a
; SI-NEXT: v_lshlrev_b32_e32 v9, 16, v60
; SI-NEXT: v_lshlrev_b32_e32 v10, 16, v62
; SI-NEXT: v_lshlrev_b32_e32 v11, 16, v57
-; SI-NEXT: v_lshlrev_b32_e32 v12, 16, v63
+; SI-NEXT: v_lshlrev_b32_e32 v12, 16, v39
; SI-NEXT: v_lshlrev_b32_e32 v13, 16, v13
; SI-NEXT: v_lshlrev_b32_e32 v14, 16, v34
; SI-NEXT: v_lshlrev_b32_e32 v15, 16, v15
@@ -43558,9 +43569,9 @@ define inreg <14 x double> @bitcast_v56f16_to_v14f64_scalar(<56 x half> inreg %a
; SI-NEXT: v_mov_b32_e32 v40, v56
; SI-NEXT: v_or_b32_e32 v10, v56, v10
; SI-NEXT: v_or_b32_e32 v11, v45, v11
-; SI-NEXT: v_or_b32_e32 v12, v38, v12
-; SI-NEXT: v_or_b32_e32 v13, v36, v13
-; SI-NEXT: v_or_b32_e32 v14, v35, v14
+; SI-NEXT: v_or_b32_e32 v12, v33, v12
+; SI-NEXT: v_or_b32_e32 v13, v38, v13
+; SI-NEXT: v_or_b32_e32 v14, v36, v14
; SI-NEXT: v_or_b32_e32 v15, v32, v15
; SI-NEXT: v_or_b32_e32 v17, v37, v17
; SI-NEXT: buffer_load_dword v26, off, s[0:3], s32 offset:128 ; 4-byte Folded Reload
@@ -43622,10 +43633,10 @@ define inreg <14 x double> @bitcast_v56f16_to_v14f64_scalar(<56 x half> inreg %a
; SI-NEXT: v_lshlrev_b32_e32 v2, 16, v2
; SI-NEXT: v_or_b32_e32 v0, v1, v0
; SI-NEXT: v_or_b32_e32 v1, v3, v2
-; SI-NEXT: v_cvt_f32_f16_e32 v2, v49
-; SI-NEXT: v_cvt_f32_f16_e32 v3, v39
+; SI-NEXT: v_cvt_f32_f16_e32 v2, v63
+; SI-NEXT: v_cvt_f32_f16_e32 v3, v48
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_cvt_f32_f16_e32 v4, v33
+; SI-NEXT: v_cvt_f32_f16_e32 v4, v35
; SI-NEXT: v_cvt_f32_f16_e32 v8, v47
; SI-NEXT: v_add_f32_e32 v2, 0x38000000, v2
; SI-NEXT: v_cvt_f16_f32_e32 v2, v2
@@ -43635,7 +43646,7 @@ define inreg <14 x double> @bitcast_v56f16_to_v14f64_scalar(<56 x half> inreg %a
; SI-NEXT: v_lshlrev_b32_e32 v2, 16, v2
; SI-NEXT: v_cvt_f16_f32_e32 v4, v4
; SI-NEXT: v_or_b32_e32 v2, v3, v2
-; SI-NEXT: v_cvt_f32_f16_e32 v3, v48
+; SI-NEXT: v_cvt_f32_f16_e32 v3, v49
; SI-NEXT: v_add_f32_e32 v8, 0x38000000, v8
; SI-NEXT: buffer_load_dword v5, off, s[0:3], s32 offset:232 ; 4-byte Folded Reload
; SI-NEXT: v_cvt_f16_f32_e32 v8, v8
@@ -43653,9 +43664,9 @@ define inreg <14 x double> @bitcast_v56f16_to_v14f64_scalar(<56 x half> inreg %a
; SI-NEXT: v_cvt_f16_f32_e32 v11, v11
; SI-NEXT: v_add_f32_e32 v12, 0x38000000, v12
; SI-NEXT: v_cvt_f16_f32_e32 v12, v12
-; SI-NEXT: v_cvt_f32_f16_e32 v13, v38
-; SI-NEXT: v_cvt_f32_f16_e32 v14, v36
-; SI-NEXT: v_cvt_f32_f16_e32 v15, v35
+; SI-NEXT: v_cvt_f32_f16_e32 v13, v33
+; SI-NEXT: v_cvt_f32_f16_e32 v14, v38
+; SI-NEXT: v_cvt_f32_f16_e32 v15, v36
; SI-NEXT: v_cvt_f32_f16_e32 v17, v32
; SI-NEXT: v_add_f32_e32 v13, 0x38000000, v13
; SI-NEXT: v_cvt_f16_f32_e32 v13, v13
@@ -43768,7 +43779,7 @@ define inreg <14 x double> @bitcast_v56f16_to_v14f64_scalar(<56 x half> inreg %a
; SI-NEXT: v_cvt_f16_f32_e32 v11, v11
; SI-NEXT: v_lshlrev_b32_e32 v11, 16, v11
; SI-NEXT: v_or_b32_e32 v11, v12, v11
-; SI-NEXT: v_cvt_f32_f16_e32 v12, v63
+; SI-NEXT: v_cvt_f32_f16_e32 v12, v39
; SI-NEXT: v_add_f32_e32 v12, 0x38000000, v12
; SI-NEXT: v_cvt_f16_f32_e32 v12, v12
; SI-NEXT: v_lshlrev_b32_e32 v12, 16, v12
@@ -43908,10 +43919,10 @@ define inreg <14 x double> @bitcast_v56f16_to_v14f64_scalar(<56 x half> inreg %a
; SI-NEXT: s_waitcnt vmcnt(0) expcnt(0)
; SI-NEXT: s_setpc_b64 s[30:31]
; SI-NEXT: .LBB55_4:
-; SI-NEXT: v_mov_b32_e32 v39, v11
-; SI-NEXT: v_mov_b32_e32 v33, v10
-; SI-NEXT: v_mov_b32_e32 v49, v2
-; SI-NEXT: v_mov_b32_e32 v48, v3
+; SI-NEXT: v_mov_b32_e32 v48, v11
+; SI-NEXT: v_mov_b32_e32 v35, v10
+; SI-NEXT: v_mov_b32_e32 v63, v2
+; SI-NEXT: v_mov_b32_e32 v49, v3
; SI-NEXT: v_mov_b32_e32 v52, v37
; SI-NEXT: v_mov_b32_e32 v37, v29
; SI-NEXT: ; implicit-def: $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7_vgpr8_vgpr9_vgpr10_vgpr11_vgpr12_vgpr13_vgpr14_vgpr15_vgpr16_vgpr17_vgpr18_vgpr19_vgpr20_vgpr21_vgpr22_vgpr23_vgpr24_vgpr25_vgpr26_vgpr27_vgpr28_vgpr29_vgpr30_vgpr31
@@ -46178,9 +46189,8 @@ define inreg <56 x half> @bitcast_v56i16_to_v56f16_scalar(<56 x i16> inreg %a, i
; SI-NEXT: buffer_load_dword v33, off, s[0:3], s32 offset:48
; SI-NEXT: buffer_load_dword v34, off, s[0:3], s32 offset:12
; SI-NEXT: buffer_load_dword v35, off, s[0:3], s32 offset:8
-; SI-NEXT: buffer_load_dword v56, off, s[0:3], s32 offset:4
-; SI-NEXT: s_waitcnt expcnt(6)
-; SI-NEXT: buffer_load_dword v57, off, s[0:3], s32
+; SI-NEXT: buffer_load_dword v36, off, s[0:3], s32 offset:4
+; SI-NEXT: buffer_load_dword v37, off, s[0:3], s32
; SI-NEXT: s_waitcnt expcnt(5)
; SI-NEXT: buffer_load_dword v58, off, s[0:3], s32 offset:44
; SI-NEXT: s_waitcnt expcnt(4)
@@ -46226,7 +46236,7 @@ define inreg <56 x half> @bitcast_v56i16_to_v56f16_scalar(<56 x i16> inreg %a, i
; SI-NEXT: buffer_store_dword v29, off, s[0:3], s32 offset:368 ; 4-byte Folded Spill
; SI-NEXT: buffer_store_dword v30, off, s[0:3], s32 offset:372 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt vmcnt(14)
-; SI-NEXT: buffer_store_dword v35, off, s[0:3], s32 offset:376 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v37, off, s[0:3], s32 offset:376 ; 4-byte Folded Spill
; SI-NEXT: v_cmp_ne_u32_e32 vcc, 0, v33
; SI-NEXT: s_and_b64 s[4:5], vcc, exec
; SI-NEXT: s_cbranch_scc0 .LBB57_2
@@ -46252,33 +46262,35 @@ define inreg <56 x half> @bitcast_v56i16_to_v56f16_scalar(<56 x i16> inreg %a, i
; SI-NEXT: v_cvt_f32_f16_e32 v33, s21
; SI-NEXT: v_cvt_f32_f16_e32 v54, v29
; SI-NEXT: v_cvt_f32_f16_e32 v41, v30
-; SI-NEXT: v_cvt_f32_f16_e32 v52, v57
+; SI-NEXT: v_cvt_f32_f16_e32 v52, v37
; SI-NEXT: buffer_store_dword v33, off, s[0:3], s32 offset:180 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_cvt_f32_f16_e32 v33, s22
-; SI-NEXT: v_cvt_f32_f16_e32 v55, v56
-; SI-NEXT: v_cvt_f32_f16_e32 v50, v35
-; SI-NEXT: v_mov_b32_e32 v47, v34
+; SI-NEXT: v_mov_b32_e32 v57, v36
+; SI-NEXT: v_cvt_f32_f16_e32 v55, v36
+; SI-NEXT: v_mov_b32_e32 v56, v35
; SI-NEXT: buffer_store_dword v33, off, s[0:3], s32 offset:148 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_cvt_f32_f16_e32 v33, s23
+; SI-NEXT: v_cvt_f32_f16_e32 v50, v35
+; SI-NEXT: v_mov_b32_e32 v47, v34
; SI-NEXT: v_cvt_f32_f16_e32 v53, v34
-; SI-NEXT: v_cvt_f32_f16_e32 v48, v32
-; SI-NEXT: v_cvt_f32_f16_e32 v51, v31
; SI-NEXT: buffer_store_dword v33, off, s[0:3], s32 offset:168 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_cvt_f32_f16_e32 v33, s24
+; SI-NEXT: v_cvt_f32_f16_e32 v48, v32
+; SI-NEXT: v_cvt_f32_f16_e32 v51, v31
; SI-NEXT: v_cvt_f32_f16_e32 v36, v63
-; SI-NEXT: v_cvt_f32_f16_e32 v49, v62
-; SI-NEXT: v_cvt_f32_f16_e32 v34, v61
; SI-NEXT: buffer_store_dword v33, off, s[0:3], s32 offset:140 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_cvt_f32_f16_e32 v33, s25
+; SI-NEXT: v_cvt_f32_f16_e32 v49, v62
+; SI-NEXT: v_cvt_f32_f16_e32 v34, v61
; SI-NEXT: v_cvt_f32_f16_e32 v37, v60
-; SI-NEXT: v_cvt_f32_f16_e32 v35, v58
; SI-NEXT: buffer_store_dword v33, off, s[0:3], s32 offset:160 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_cvt_f32_f16_e32 v33, s26
+; SI-NEXT: v_cvt_f32_f16_e32 v35, v58
; SI-NEXT: buffer_store_dword v33, off, s[0:3], s32 offset:132 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_cvt_f32_f16_e32 v33, s27
@@ -46362,6 +46374,12 @@ define inreg <56 x half> @bitcast_v56i16_to_v56f16_scalar(<56 x i16> inreg %a, i
; SI-NEXT: .LBB57_2:
; SI-NEXT: ; implicit-def: $vgpr33
; SI-NEXT: ; kill: killed $vgpr33
+; SI-NEXT: v_mov_b32_e32 v57, v36
+; SI-NEXT: ; implicit-def: $vgpr33
+; SI-NEXT: ; kill: killed $vgpr33
+; SI-NEXT: v_mov_b32_e32 v56, v35
+; SI-NEXT: ; implicit-def: $vgpr33
+; SI-NEXT: ; kill: killed $vgpr33
; SI-NEXT: v_mov_b32_e32 v47, v34
; SI-NEXT: ; implicit-def: $vgpr33
; SI-NEXT: ; kill: killed $vgpr33
@@ -46449,10 +46467,6 @@ define inreg <56 x half> @bitcast_v56i16_to_v56f16_scalar(<56 x i16> inreg %a, i
; SI-NEXT: ; kill: killed $vgpr33
; SI-NEXT: ; implicit-def: $vgpr33
; SI-NEXT: ; kill: killed $vgpr33
-; SI-NEXT: ; implicit-def: $vgpr33
-; SI-NEXT: ; kill: killed $vgpr33
-; SI-NEXT: ; implicit-def: $vgpr33
-; SI-NEXT: ; kill: killed $vgpr33
; SI-NEXT: .LBB57_3: ; %Flow
; SI-NEXT: s_andn2_b64 vcc, exec, s[4:5]
; SI-NEXT: s_waitcnt expcnt(0)
@@ -46470,6 +46484,8 @@ define inreg <56 x half> @bitcast_v56i16_to_v56f16_scalar(<56 x i16> inreg %a, i
; SI-NEXT: ; %bb.4: ; %cmp.true
; SI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:376 ; 4-byte Folded Reload
; SI-NEXT: v_add_i32_e32 v46, vcc, 3, v47
+; SI-NEXT: v_add_i32_e32 v47, vcc, 3, v56
+; SI-NEXT: v_add_i32_e32 v56, vcc, 3, v57
; SI-NEXT: s_add_i32 s16, s16, 3
; SI-NEXT: v_cvt_f32_f16_e32 v33, s16
; SI-NEXT: s_add_i32 s18, s18, 3
@@ -46512,40 +46528,38 @@ define inreg <56 x half> @bitcast_v56i16_to_v56f16_scalar(<56 x i16> inreg %a, i
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_cvt_f32_f16_e32 v33, s25
; SI-NEXT: v_add_i32_e32 v32, vcc, 3, v32
-; SI-NEXT: v_add_i32_e32 v56, vcc, 3, v56
+; SI-NEXT: s_add_i32 s19, s19, 3
; SI-NEXT: buffer_store_dword v33, off, s[0:3], s32 offset:160 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_cvt_f32_f16_e32 v33, s26
-; SI-NEXT: v_add_i32_e32 v57, vcc, 3, v57
-; SI-NEXT: s_add_i32 s19, s19, 3
-; SI-NEXT: buffer_store_dword v33, off, s[0:3], s32 offset:132 ; 4-byte Folded Spill
-; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_cvt_f32_f16_e32 v33, s27
; SI-NEXT: s_add_i32 s17, s17, 3
; SI-NEXT: v_cvt_f32_f16_e32 v38, s17
; SI-NEXT: v_cvt_f32_f16_e32 v39, s19
-; SI-NEXT: buffer_store_dword v33, off, s[0:3], s32 offset:152 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v33, off, s[0:3], s32 offset:132 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_cvt_f32_f16_e32 v33, s28
-; SI-NEXT: v_cvt_f32_f16_e32 v50, v57
+; SI-NEXT: v_cvt_f32_f16_e32 v33, s27
; SI-NEXT: v_cvt_f32_f16_e32 v55, v56
+; SI-NEXT: v_cvt_f32_f16_e32 v48, v47
; SI-NEXT: v_cvt_f32_f16_e32 v53, v46
-; SI-NEXT: buffer_store_dword v33, off, s[0:3], s32 offset:124 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v33, off, s[0:3], s32 offset:152 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_cvt_f32_f16_e32 v33, s29
+; SI-NEXT: v_cvt_f32_f16_e32 v33, s28
; SI-NEXT: v_cvt_f32_f16_e32 v36, v32
; SI-NEXT: v_cvt_f32_f16_e32 v51, v31
; SI-NEXT: v_cvt_f32_f16_e32 v34, v63
-; SI-NEXT: buffer_store_dword v33, off, s[0:3], s32 offset:144 ; 4-byte Folded Spill
-; SI-NEXT: v_cvt_f32_f16_e32 v49, v62
+; SI-NEXT: buffer_store_dword v33, off, s[0:3], s32 offset:124 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_cvt_f32_f16_e32 v33, v61
+; SI-NEXT: v_cvt_f32_f16_e32 v33, s29
+; SI-NEXT: v_cvt_f32_f16_e32 v49, v62
; SI-NEXT: v_cvt_f32_f16_e32 v37, v60
; SI-NEXT: v_cvt_f32_f16_e32 v35, v58
+; SI-NEXT: buffer_store_dword v33, off, s[0:3], s32 offset:144 ; 4-byte Folded Spill
+; SI-NEXT: s_waitcnt expcnt(0)
+; SI-NEXT: v_cvt_f32_f16_e32 v33, v61
; SI-NEXT: s_waitcnt vmcnt(12)
-; SI-NEXT: v_add_i32_e32 v47, vcc, 3, v1
+; SI-NEXT: v_add_i32_e32 v57, vcc, 3, v1
; SI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:372 ; 4-byte Folded Reload
-; SI-NEXT: v_cvt_f32_f16_e32 v48, v47
+; SI-NEXT: v_cvt_f32_f16_e32 v50, v57
; SI-NEXT: s_waitcnt vmcnt(0)
; SI-NEXT: v_add_i32_e32 v30, vcc, 3, v1
; SI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:368 ; 4-byte Folded Reload
@@ -47480,26 +47494,31 @@ define inreg <56 x half> @bitcast_v56i16_to_v56f16_scalar(<56 x i16> inreg %a, i
; GFX11-TRUE16-NEXT: v_lshl_or_b32 v6, v10, 16, v6
; GFX11-TRUE16-NEXT: v_mov_b16_e32 v10.l, v4.h
; GFX11-TRUE16-NEXT: v_and_b32_e32 v4, 0xffff, v4
+; GFX11-TRUE16-NEXT: v_lshl_or_b32 v5, v13, 16, v5
; GFX11-TRUE16-NEXT: v_mov_b16_e32 v11.l, v2.h
; GFX11-TRUE16-NEXT: v_and_b32_e32 v2, 0xffff, v2
; GFX11-TRUE16-NEXT: v_mov_b16_e32 v12.l, v1.h
-; GFX11-TRUE16-NEXT: v_and_b32_e32 v1, 0xffff, v1
-; GFX11-TRUE16-NEXT: v_lshl_or_b32 v5, v13, 16, v5
; GFX11-TRUE16-NEXT: v_lshl_or_b32 v4, v10, 16, v4
; GFX11-TRUE16-NEXT: v_mov_b16_e32 v10.l, v3.h
; GFX11-TRUE16-NEXT: v_and_b32_e32 v3, 0xffff, v3
+; GFX11-TRUE16-NEXT: v_and_b32_e32 v1, 0xffff, v1
; GFX11-TRUE16-NEXT: v_mov_b16_e32 v13.l, v0.h
; GFX11-TRUE16-NEXT: v_and_b32_e32 v0, 0xffff, v0
-; GFX11-TRUE16-NEXT: s_pack_ll_b32_b16 s14, s23, s14
-; GFX11-TRUE16-NEXT: s_pack_ll_b32_b16 s13, s22, s13
; GFX11-TRUE16-NEXT: v_lshl_or_b32 v2, v11, 16, v2
+; GFX11-TRUE16-NEXT: v_lshl_or_b32 v3, v10, 16, v3
; GFX11-TRUE16-NEXT: v_lshl_or_b32 v1, v12, 16, v1
+; GFX11-TRUE16-NEXT: s_pack_ll_b32_b16 s14, s23, s14
+; GFX11-TRUE16-NEXT: v_lshl_or_b32 v0, v13, 16, v0
+; GFX11-TRUE16-NEXT: s_pack_ll_b32_b16 s13, s22, s13
; GFX11-TRUE16-NEXT: v_pk_add_u16 v11, s14, 3 op_sel_hi:[1,0]
; GFX11-TRUE16-NEXT: s_mov_b32 s14, s21
; GFX11-TRUE16-NEXT: v_pk_add_u16 v12, s13, 3 op_sel_hi:[1,0]
; GFX11-TRUE16-NEXT: s_mov_b32 s13, s20
-; GFX11-TRUE16-NEXT: v_lshl_or_b32 v3, v10, 16, v3
-; GFX11-TRUE16-NEXT: v_lshl_or_b32 v0, v13, 16, v0
+; GFX11-TRUE16-NEXT: v_pk_add_u16 v4, v4, 3 op_sel_hi:[1,0]
+; GFX11-TRUE16-NEXT: v_pk_add_u16 v3, v3, 3 op_sel_hi:[1,0]
+; GFX11-TRUE16-NEXT: v_pk_add_u16 v2, v2, 3 op_sel_hi:[1,0]
+; GFX11-TRUE16-NEXT: v_pk_add_u16 v1, v1, 3 op_sel_hi:[1,0]
+; GFX11-TRUE16-NEXT: v_pk_add_u16 v0, v0, 3 op_sel_hi:[1,0]
; GFX11-TRUE16-NEXT: s_pack_ll_b32_b16 s10, s14, s10
; GFX11-TRUE16-NEXT: s_pack_ll_b32_b16 s9, s13, s9
; GFX11-TRUE16-NEXT: v_pk_add_u16 v18, s10, 3 op_sel_hi:[1,0]
@@ -47513,11 +47532,11 @@ define inreg <56 x half> @bitcast_v56i16_to_v56f16_scalar(<56 x i16> inreg %a, i
; GFX11-TRUE16-NEXT: v_pk_add_u16 v7, v7, 3 op_sel_hi:[1,0]
; GFX11-TRUE16-NEXT: v_pk_add_u16 v6, v6, 3 op_sel_hi:[1,0]
; GFX11-TRUE16-NEXT: v_pk_add_u16 v5, v5, 3 op_sel_hi:[1,0]
-; GFX11-TRUE16-NEXT: v_pk_add_u16 v4, v4, 3 op_sel_hi:[1,0]
-; GFX11-TRUE16-NEXT: v_pk_add_u16 v3, v3, 3 op_sel_hi:[1,0]
-; GFX11-TRUE16-NEXT: v_pk_add_u16 v2, v2, 3 op_sel_hi:[1,0]
-; GFX11-TRUE16-NEXT: v_pk_add_u16 v1, v1, 3 op_sel_hi:[1,0]
-; GFX11-TRUE16-NEXT: v_pk_add_u16 v0, v0, 3 op_sel_hi:[1,0]
+; GFX11-TRUE16-NEXT: v_lshrrev_b32_e32 v54, 16, v0
+; GFX11-TRUE16-NEXT: v_lshrrev_b32_e32 v55, 16, v1
+; GFX11-TRUE16-NEXT: v_lshrrev_b32_e32 v64, 16, v2
+; GFX11-TRUE16-NEXT: v_lshrrev_b32_e32 v65, 16, v3
+; GFX11-TRUE16-NEXT: v_lshrrev_b32_e32 v66, 16, v4
; GFX11-TRUE16-NEXT: v_pk_add_u16 v20, s8, 3 op_sel_hi:[1,0]
; GFX11-TRUE16-NEXT: s_mov_b32 s8, s17
; GFX11-TRUE16-NEXT: v_pk_add_u16 v21, s7, 3 op_sel_hi:[1,0]
@@ -47546,16 +47565,16 @@ define inreg <56 x half> @bitcast_v56i16_to_v56f16_scalar(<56 x i16> inreg %a, i
; GFX11-TRUE16-NEXT: v_pk_add_u16 v30, s2, 3 op_sel_hi:[1,0]
; GFX11-TRUE16-NEXT: v_pk_add_u16 v33, s3, 3 op_sel_hi:[1,0]
; GFX11-TRUE16-NEXT: v_pk_add_u16 v28, s5, 3 op_sel_hi:[1,0]
-; GFX11-TRUE16-NEXT: v_lshrrev_b32_e32 v54, 16, v0
-; GFX11-TRUE16-NEXT: v_lshrrev_b32_e32 v55, 16, v1
-; GFX11-TRUE16-NEXT: v_lshrrev_b32_e32 v64, 16, v2
-; GFX11-TRUE16-NEXT: v_lshrrev_b32_e32 v65, 16, v3
-; GFX11-TRUE16-NEXT: v_lshrrev_b32_e32 v66, 16, v4
-; GFX11-TRUE16-NEXT: v_lshrrev_b32_e32 v67, 16, v5
-; GFX11-TRUE16-NEXT: v_lshrrev_b32_e32 v68, 16, v6
-; GFX11-TRUE16-NEXT: v_lshrrev_b32_e32 v69, 16, v7
-; GFX11-TRUE16-NEXT: v_lshrrev_b32_e32 v70, 16, v8
-; GFX11-TRUE16-NEXT: v_lshrrev_b32_e32 v71, 16, v9
+; GFX11-TRUE16-NEXT: v_mov_b16_e32 v0.h, v54.l
+; GFX11-TRUE16-NEXT: v_mov_b16_e32 v1.h, v55.l
+; GFX11-TRUE16-NEXT: v_mov_b16_e32 v2.h, v64.l
+; GFX11-TRUE16-NEXT: v_mov_b16_e32 v3.h, v65.l
+; GFX11-TRUE16-NEXT: v_mov_b16_e32 v4.h, v66.l
+; GFX11-TRUE16-NEXT: v_lshrrev_b32_e32 v54, 16, v5
+; GFX11-TRUE16-NEXT: v_lshrrev_b32_e32 v55, 16, v6
+; GFX11-TRUE16-NEXT: v_lshrrev_b32_e32 v64, 16, v7
+; GFX11-TRUE16-NEXT: v_lshrrev_b32_e32 v65, 16, v8
+; GFX11-TRUE16-NEXT: v_lshrrev_b32_e32 v66, 16, v9
; GFX11-TRUE16-NEXT: v_lshrrev_b32_e32 v51, 16, v32
; GFX11-TRUE16-NEXT: v_lshrrev_b32_e32 v52, 16, v31
; GFX11-TRUE16-NEXT: v_lshrrev_b32_e32 v53, 16, v30
@@ -47574,16 +47593,11 @@ define inreg <56 x half> @bitcast_v56i16_to_v56f16_scalar(<56 x i16> inreg %a, i
; GFX11-TRUE16-NEXT: v_lshrrev_b32_e32 v25, 16, v15
; GFX11-TRUE16-NEXT: v_lshrrev_b32_e32 v26, 16, v14
; GFX11-TRUE16-NEXT: v_lshrrev_b32_e32 v27, 16, v13
-; GFX11-TRUE16-NEXT: v_mov_b16_e32 v0.h, v54.l
-; GFX11-TRUE16-NEXT: v_mov_b16_e32 v1.h, v55.l
-; GFX11-TRUE16-NEXT: v_mov_b16_e32 v2.h, v64.l
-; GFX11-TRUE16-NEXT: v_mov_b16_e32 v3.h, v65.l
-; GFX11-TRUE16-NEXT: v_mov_b16_e32 v4.h, v66.l
-; GFX11-TRUE16-NEXT: v_mov_b16_e32 v5.h, v67.l
-; GFX11-TRUE16-NEXT: v_mov_b16_e32 v6.h, v68.l
-; GFX11-TRUE16-NEXT: v_mov_b16_e32 v7.h, v69.l
-; GFX11-TRUE16-NEXT: v_mov_b16_e32 v8.h, v70.l
-; GFX11-TRUE16-NEXT: v_mov_b16_e32 v9.h, v71.l
+; GFX11-TRUE16-NEXT: v_mov_b16_e32 v5.h, v54.l
+; GFX11-TRUE16-NEXT: v_mov_b16_e32 v6.h, v55.l
+; GFX11-TRUE16-NEXT: v_mov_b16_e32 v7.h, v64.l
+; GFX11-TRUE16-NEXT: v_mov_b16_e32 v8.h, v65.l
+; GFX11-TRUE16-NEXT: v_mov_b16_e32 v9.h, v66.l
; GFX11-TRUE16-NEXT: s_branch .LBB57_5
; GFX11-TRUE16-NEXT: .LBB57_3:
; GFX11-TRUE16-NEXT: s_branch .LBB57_2
@@ -50146,26 +50160,31 @@ define inreg <56 x i16> @bitcast_v56f16_to_v56i16_scalar(<56 x half> inreg %a, i
; GFX11-TRUE16-NEXT: v_lshl_or_b32 v6, v10, 16, v6
; GFX11-TRUE16-NEXT: v_mov_b16_e32 v10.l, v4.h
; GFX11-TRUE16-NEXT: v_and_b32_e32 v4, 0xffff, v4
+; GFX11-TRUE16-NEXT: v_lshl_or_b32 v5, v13, 16, v5
; GFX11-TRUE16-NEXT: v_mov_b16_e32 v11.l, v2.h
; GFX11-TRUE16-NEXT: v_and_b32_e32 v2, 0xffff, v2
; GFX11-TRUE16-NEXT: v_mov_b16_e32 v12.l, v1.h
-; GFX11-TRUE16-NEXT: v_and_b32_e32 v1, 0xffff, v1
-; GFX11-TRUE16-NEXT: v_lshl_or_b32 v5, v13, 16, v5
; GFX11-TRUE16-NEXT: v_lshl_or_b32 v4, v10, 16, v4
; GFX11-TRUE16-NEXT: v_mov_b16_e32 v10.l, v3.h
; GFX11-TRUE16-NEXT: v_and_b32_e32 v3, 0xffff, v3
+; GFX11-TRUE16-NEXT: v_and_b32_e32 v1, 0xffff, v1
; GFX11-TRUE16-NEXT: v_mov_b16_e32 v13.l, v0.h
; GFX11-TRUE16-NEXT: v_and_b32_e32 v0, 0xffff, v0
-; GFX11-TRUE16-NEXT: s_pack_ll_b32_b16 s14, s23, s14
-; GFX11-TRUE16-NEXT: s_pack_ll_b32_b16 s13, s22, s13
; GFX11-TRUE16-NEXT: v_lshl_or_b32 v2, v11, 16, v2
+; GFX11-TRUE16-NEXT: v_lshl_or_b32 v3, v10, 16, v3
; GFX11-TRUE16-NEXT: v_lshl_or_b32 v1, v12, 16, v1
+; GFX11-TRUE16-NEXT: s_pack_ll_b32_b16 s14, s23, s14
+; GFX11-TRUE16-NEXT: v_lshl_or_b32 v0, v13, 16, v0
+; GFX11-TRUE16-NEXT: s_pack_ll_b32_b16 s13, s22, s13
; GFX11-TRUE16-NEXT: v_pk_add_f16 v11, 0x200, s14 op_sel_hi:[0,1]
; GFX11-TRUE16-NEXT: s_mov_b32 s14, s21
; GFX11-TRUE16-NEXT: v_pk_add_f16 v12, 0x200, s13 op_sel_hi:[0,1]
; GFX11-TRUE16-NEXT: s_mov_b32 s13, s20
-; GFX11-TRUE16-NEXT: v_lshl_or_b32 v3, v10, 16, v3
-; GFX11-TRUE16-NEXT: v_lshl_or_b32 v0, v13, 16, v0
+; GFX11-TRUE16-NEXT: v_pk_add_f16 v4, 0x200, v4 op_sel_hi:[0,1]
+; GFX11-TRUE16-NEXT: v_pk_add_f16 v3, 0x200, v3 op_sel_hi:[0,1]
+; GFX11-TRUE16-NEXT: v_pk_add_f16 v2, 0x200, v2 op_sel_hi:[0,1]
+; GFX11-TRUE16-NEXT: v_pk_add_f16 v1, 0x200, v1 op_sel_hi:[0,1]
+; GFX11-TRUE16-NEXT: v_pk_add_f16 v0, 0x200, v0 op_sel_hi:[0,1]
; GFX11-TRUE16-NEXT: s_pack_ll_b32_b16 s10, s14, s10
; GFX11-TRUE16-NEXT: s_pack_ll_b32_b16 s9, s13, s9
; GFX11-TRUE16-NEXT: v_pk_add_f16 v18, 0x200, s10 op_sel_hi:[0,1]
@@ -50179,11 +50198,11 @@ define inreg <56 x i16> @bitcast_v56f16_to_v56i16_scalar(<56 x half> inreg %a, i
; GFX11-TRUE16-NEXT: v_pk_add_f16 v7, 0x200, v7 op_sel_hi:[0,1]
; GFX11-TRUE16-NEXT: v_pk_add_f16 v6, 0x200, v6 op_sel_hi:[0,1]
; GFX11-TRUE16-NEXT: v_pk_add_f16 v5, 0x200, v5 op_sel_hi:[0,1]
-; GFX11-TRUE16-NEXT: v_pk_add_f16 v4, 0x200, v4 op_sel_hi:[0,1]
-; GFX11-TRUE16-NEXT: v_pk_add_f16 v3, 0x200, v3 op_sel_hi:[0,1]
-; GFX11-TRUE16-NEXT: v_pk_add_f16 v2, 0x200, v2 op_sel_hi:[0,1]
-; GFX11-TRUE16-NEXT: v_pk_add_f16 v1, 0x200, v1 op_sel_hi:[0,1]
-; GFX11-TRUE16-NEXT: v_pk_add_f16 v0, 0x200, v0 op_sel_hi:[0,1]
+; GFX11-TRUE16-NEXT: v_lshrrev_b32_e32 v54, 16, v0
+; GFX11-TRUE16-NEXT: v_lshrrev_b32_e32 v55, 16, v1
+; GFX11-TRUE16-NEXT: v_lshrrev_b32_e32 v64, 16, v2
+; GFX11-TRUE16-NEXT: v_lshrrev_b32_e32 v65, 16, v3
+; GFX11-TRUE16-NEXT: v_lshrrev_b32_e32 v66, 16, v4
; GFX11-TRUE16-NEXT: v_pk_add_f16 v20, 0x200, s8 op_sel_hi:[0,1]
; GFX11-TRUE16-NEXT: s_mov_b32 s8, s17
; GFX11-TRUE16-NEXT: v_pk_add_f16 v21, 0x200, s7 op_sel_hi:[0,1]
@@ -50212,16 +50231,16 @@ define inreg <56 x i16> @bitcast_v56f16_to_v56i16_scalar(<56 x half> inreg %a, i
; GFX11-TRUE16-NEXT: v_pk_add_f16 v30, 0x200, s2 op_sel_hi:[0,1]
; GFX11-TRUE16-NEXT: v_pk_add_f16 v33, 0x200, s3 op_sel_hi:[0,1]
; GFX11-TRUE16-NEXT: v_pk_add_f16 v28, 0x200, s5 op_sel_hi:[0,1]
-; GFX11-TRUE16-NEXT: v_lshrrev_b32_e32 v54, 16, v0
-; GFX11-TRUE16-NEXT: v_lshrrev_b32_e32 v55, 16, v1
-; GFX11-TRUE16-NEXT: v_lshrrev_b32_e32 v64, 16, v2
-; GFX11-TRUE16-NEXT: v_lshrrev_b32_e32 v65, 16, v3
-; GFX11-TRUE16-NEXT: v_lshrrev_b32_e32 v66, 16, v4
-; GFX11-TRUE16-NEXT: v_lshrrev_b32_e32 v67, 16, v5
-; GFX11-TRUE16-NEXT: v_lshrrev_b32_e32 v68, 16, v6
-; GFX11-TRUE16-NEXT: v_lshrrev_b32_e32 v69, 16, v7
-; GFX11-TRUE16-NEXT: v_lshrrev_b32_e32 v70, 16, v8
-; GFX11-TRUE16-NEXT: v_lshrrev_b32_e32 v71, 16, v9
+; GFX11-TRUE16-NEXT: v_mov_b16_e32 v0.h, v54.l
+; GFX11-TRUE16-NEXT: v_mov_b16_e32 v1.h, v55.l
+; GFX11-TRUE16-NEXT: v_mov_b16_e32 v2.h, v64.l
+; GFX11-TRUE16-NEXT: v_mov_b16_e32 v3.h, v65.l
+; GFX11-TRUE16-NEXT: v_mov_b16_e32 v4.h, v66.l
+; GFX11-TRUE16-NEXT: v_lshrrev_b32_e32 v54, 16, v5
+; GFX11-TRUE16-NEXT: v_lshrrev_b32_e32 v55, 16, v6
+; GFX11-TRUE16-NEXT: v_lshrrev_b32_e32 v64, 16, v7
+; GFX11-TRUE16-NEXT: v_lshrrev_b32_e32 v65, 16, v8
+; GFX11-TRUE16-NEXT: v_lshrrev_b32_e32 v66, 16, v9
; GFX11-TRUE16-NEXT: v_lshrrev_b32_e32 v51, 16, v32
; GFX11-TRUE16-NEXT: v_lshrrev_b32_e32 v52, 16, v31
; GFX11-TRUE16-NEXT: v_lshrrev_b32_e32 v53, 16, v30
@@ -50240,16 +50259,11 @@ define inreg <56 x i16> @bitcast_v56f16_to_v56i16_scalar(<56 x half> inreg %a, i
; GFX11-TRUE16-NEXT: v_lshrrev_b32_e32 v25, 16, v15
; GFX11-TRUE16-NEXT: v_lshrrev_b32_e32 v26, 16, v14
; GFX11-TRUE16-NEXT: v_lshrrev_b32_e32 v27, 16, v13
-; GFX11-TRUE16-NEXT: v_mov_b16_e32 v0.h, v54.l
-; GFX11-TRUE16-NEXT: v_mov_b16_e32 v1.h, v55.l
-; GFX11-TRUE16-NEXT: v_mov_b16_e32 v2.h, v64.l
-; GFX11-TRUE16-NEXT: v_mov_b16_e32 v3.h, v65.l
-; GFX11-TRUE16-NEXT: v_mov_b16_e32 v4.h, v66.l
-; GFX11-TRUE16-NEXT: v_mov_b16_e32 v5.h, v67.l
-; GFX11-TRUE16-NEXT: v_mov_b16_e32 v6.h, v68.l
-; GFX11-TRUE16-NEXT: v_mov_b16_e32 v7.h, v69.l
-; GFX11-TRUE16-NEXT: v_mov_b16_e32 v8.h, v70.l
-; GFX11-TRUE16-NEXT: v_mov_b16_e32 v9.h, v71.l
+; GFX11-TRUE16-NEXT: v_mov_b16_e32 v5.h, v54.l
+; GFX11-TRUE16-NEXT: v_mov_b16_e32 v6.h, v55.l
+; GFX11-TRUE16-NEXT: v_mov_b16_e32 v7.h, v64.l
+; GFX11-TRUE16-NEXT: v_mov_b16_e32 v8.h, v65.l
+; GFX11-TRUE16-NEXT: v_mov_b16_e32 v9.h, v66.l
; GFX11-TRUE16-NEXT: s_branch .LBB59_5
; GFX11-TRUE16-NEXT: .LBB59_3:
; GFX11-TRUE16-NEXT: s_branch .LBB59_2
diff --git a/llvm/test/CodeGen/AMDGPU/amdgcn.bitcast.960bit.ll b/llvm/test/CodeGen/AMDGPU/amdgcn.bitcast.960bit.ll
index 94ed6276bd051..a5bfff7b0e5d3 100644
--- a/llvm/test/CodeGen/AMDGPU/amdgcn.bitcast.960bit.ll
+++ b/llvm/test/CodeGen/AMDGPU/amdgcn.bitcast.960bit.ll
@@ -4738,7 +4738,7 @@ define inreg <60 x i16> @bitcast_v30i32_to_v60i16_scalar(<30 x i32> inreg %a, i3
; GFX11-TRUE16-NEXT: v_readfirstlane_b32 s5, v9
; GFX11-TRUE16-NEXT: v_readfirstlane_b32 s6, v10
; GFX11-TRUE16-NEXT: v_readfirstlane_b32 s4, v11
-; GFX11-TRUE16-NEXT: s_mov_b32 s94, 0
+; GFX11-TRUE16-NEXT: s_mov_b32 s89, 0
; GFX11-TRUE16-NEXT: s_and_b32 s40, vcc_lo, exec_lo
; GFX11-TRUE16-NEXT: s_cbranch_scc0 .LBB13_4
; GFX11-TRUE16-NEXT: ; %bb.1: ; %cmp.false
@@ -4767,12 +4767,12 @@ define inreg <60 x i16> @bitcast_v30i32_to_v60i16_scalar(<30 x i32> inreg %a, i3
; GFX11-TRUE16-NEXT: s_lshr_b32 s78, s19, 16
; GFX11-TRUE16-NEXT: s_lshr_b32 s79, s18, 16
; GFX11-TRUE16-NEXT: s_lshr_b32 s88, s17, 16
-; GFX11-TRUE16-NEXT: s_lshr_b32 s89, s16, 16
-; GFX11-TRUE16-NEXT: s_lshr_b32 s90, s3, 16
-; GFX11-TRUE16-NEXT: s_lshr_b32 s91, s2, 16
-; GFX11-TRUE16-NEXT: s_lshr_b32 s92, s1, 16
-; GFX11-TRUE16-NEXT: s_lshr_b32 s93, s0, 16
-; GFX11-TRUE16-NEXT: s_and_not1_b32 vcc_lo, exec_lo, s94
+; GFX11-TRUE16-NEXT: s_lshr_b32 s90, s16, 16
+; GFX11-TRUE16-NEXT: s_lshr_b32 s91, s3, 16
+; GFX11-TRUE16-NEXT: s_lshr_b32 s92, s2, 16
+; GFX11-TRUE16-NEXT: s_lshr_b32 s93, s1, 16
+; GFX11-TRUE16-NEXT: s_lshr_b32 s94, s0, 16
+; GFX11-TRUE16-NEXT: s_and_not1_b32 vcc_lo, exec_lo, s89
; GFX11-TRUE16-NEXT: s_cbranch_vccnz .LBB13_3
; GFX11-TRUE16-NEXT: .LBB13_2: ; %cmp.true
; GFX11-TRUE16-NEXT: s_add_i32 s4, s4, 3
@@ -4830,19 +4830,22 @@ define inreg <60 x i16> @bitcast_v30i32_to_v60i16_scalar(<30 x i32> inreg %a, i3
; GFX11-TRUE16-NEXT: s_lshr_b32 s78, s19, 16
; GFX11-TRUE16-NEXT: s_lshr_b32 s79, s18, 16
; GFX11-TRUE16-NEXT: s_lshr_b32 s88, s17, 16
-; GFX11-TRUE16-NEXT: s_lshr_b32 s89, s16, 16
-; GFX11-TRUE16-NEXT: s_lshr_b32 s90, s3, 16
-; GFX11-TRUE16-NEXT: s_lshr_b32 s91, s2, 16
-; GFX11-TRUE16-NEXT: s_lshr_b32 s92, s1, 16
-; GFX11-TRUE16-NEXT: s_lshr_b32 s93, s0, 16
+; GFX11-TRUE16-NEXT: s_lshr_b32 s90, s16, 16
+; GFX11-TRUE16-NEXT: s_lshr_b32 s91, s3, 16
+; GFX11-TRUE16-NEXT: s_lshr_b32 s92, s2, 16
+; GFX11-TRUE16-NEXT: s_lshr_b32 s93, s1, 16
+; GFX11-TRUE16-NEXT: s_lshr_b32 s94, s0, 16
; GFX11-TRUE16-NEXT: .LBB13_3: ; %end
; GFX11-TRUE16-NEXT: s_delay_alu instid0(SALU_CYCLE_1)
-; GFX11-TRUE16-NEXT: s_pack_ll_b32_b16 s0, s0, s93
-; GFX11-TRUE16-NEXT: s_pack_ll_b32_b16 s1, s1, s92
-; GFX11-TRUE16-NEXT: s_pack_ll_b32_b16 s2, s2, s91
-; GFX11-TRUE16-NEXT: s_pack_ll_b32_b16 s3, s3, s90
-; GFX11-TRUE16-NEXT: s_pack_ll_b32_b16 s16, s16, s89
+; GFX11-TRUE16-NEXT: s_mov_b32 s89, s94
+; GFX11-TRUE16-NEXT: s_pack_ll_b32_b16 s1, s1, s93
+; GFX11-TRUE16-NEXT: s_pack_ll_b32_b16 s0, s0, s89
+; GFX11-TRUE16-NEXT: s_mov_b32 s89, s92
+; GFX11-TRUE16-NEXT: s_pack_ll_b32_b16 s3, s3, s91
+; GFX11-TRUE16-NEXT: s_pack_ll_b32_b16 s2, s2, s89
+; GFX11-TRUE16-NEXT: s_mov_b32 s89, s90
; GFX11-TRUE16-NEXT: s_pack_ll_b32_b16 s17, s17, s88
+; GFX11-TRUE16-NEXT: s_pack_ll_b32_b16 s16, s16, s89
; GFX11-TRUE16-NEXT: s_pack_ll_b32_b16 s18, s18, s79
; GFX11-TRUE16-NEXT: s_pack_ll_b32_b16 s19, s19, s78
; GFX11-TRUE16-NEXT: s_pack_ll_b32_b16 s20, s20, s77
@@ -4884,11 +4887,11 @@ define inreg <60 x i16> @bitcast_v30i32_to_v60i16_scalar(<30 x i32> inreg %a, i3
; GFX11-TRUE16-NEXT: v_dual_mov_b32 v28, s6 :: v_dual_mov_b32 v29, s4
; GFX11-TRUE16-NEXT: s_setpc_b64 s[30:31]
; GFX11-TRUE16-NEXT: .LBB13_4:
+; GFX11-TRUE16-NEXT: ; implicit-def: $sgpr94_lo16
; GFX11-TRUE16-NEXT: ; implicit-def: $sgpr93_lo16
; GFX11-TRUE16-NEXT: ; implicit-def: $sgpr92_lo16
; GFX11-TRUE16-NEXT: ; implicit-def: $sgpr91_lo16
; GFX11-TRUE16-NEXT: ; implicit-def: $sgpr90_lo16
-; GFX11-TRUE16-NEXT: ; implicit-def: $sgpr89_lo16
; GFX11-TRUE16-NEXT: ; implicit-def: $sgpr88_lo16
; GFX11-TRUE16-NEXT: ; implicit-def: $sgpr79_lo16
; GFX11-TRUE16-NEXT: ; implicit-def: $sgpr78_lo16
@@ -6006,8 +6009,8 @@ define <30 x i32> @bitcast_v60i16_to_v30i32(<60 x i16> %a, i32 %b) {
; GFX9-NEXT: v_lshrrev_b32_e32 v0, 16, v44
; GFX9-NEXT: v_lshrrev_b32_e32 v32, 16, v45
; GFX9-NEXT: v_lshrrev_b32_e32 v33, 16, v46
-; GFX9-NEXT: v_lshrrev_b32_e32 v63, 16, v47
-; GFX9-NEXT: v_lshrrev_b32_e32 v34, 16, v56
+; GFX9-NEXT: v_lshrrev_b32_e32 v34, 16, v47
+; GFX9-NEXT: v_lshrrev_b32_e32 v63, 16, v56
; GFX9-NEXT: v_lshrrev_b32_e32 v35, 16, v57
; GFX9-NEXT: v_lshrrev_b32_e32 v36, 16, v58
; GFX9-NEXT: v_lshrrev_b32_e32 v62, 16, v59
@@ -6042,8 +6045,8 @@ define <30 x i32> @bitcast_v60i16_to_v30i32(<60 x i16> %a, i32 %b) {
; GFX9-NEXT: v_perm_b32 v2, v62, v59, s6
; GFX9-NEXT: v_perm_b32 v3, v36, v58, s6
; GFX9-NEXT: v_perm_b32 v4, v35, v57, s6
-; GFX9-NEXT: v_perm_b32 v5, v34, v56, s6
-; GFX9-NEXT: v_perm_b32 v6, v63, v47, s6
+; GFX9-NEXT: v_perm_b32 v5, v63, v56, s6
+; GFX9-NEXT: v_perm_b32 v6, v34, v47, s6
; GFX9-NEXT: v_perm_b32 v7, v33, v46, s6
; GFX9-NEXT: v_perm_b32 v8, v32, v45, s6
; GFX9-NEXT: v_perm_b32 v9, v9, v44, s6
@@ -6085,8 +6088,8 @@ define <30 x i32> @bitcast_v60i16_to_v30i32(<60 x i16> %a, i32 %b) {
; GFX9-NEXT: ; implicit-def: $vgpr39
; GFX9-NEXT: ; implicit-def: $vgpr32
; GFX9-NEXT: ; implicit-def: $vgpr33
-; GFX9-NEXT: ; implicit-def: $vgpr63
; GFX9-NEXT: ; implicit-def: $vgpr34
+; GFX9-NEXT: ; implicit-def: $vgpr63
; GFX9-NEXT: ; implicit-def: $vgpr35
; GFX9-NEXT: ; implicit-def: $vgpr36
; GFX9-NEXT: ; implicit-def: $vgpr62
@@ -6200,8 +6203,8 @@ define <30 x i32> @bitcast_v60i16_to_v30i32(<60 x i16> %a, i32 %b) {
; GFX9-NEXT: v_perm_b32 v2, v62, v59, s6
; GFX9-NEXT: v_perm_b32 v3, v36, v58, s6
; GFX9-NEXT: v_perm_b32 v4, v35, v57, s6
-; GFX9-NEXT: v_perm_b32 v5, v34, v56, s6
-; GFX9-NEXT: v_perm_b32 v6, v63, v47, s6
+; GFX9-NEXT: v_perm_b32 v5, v63, v56, s6
+; GFX9-NEXT: v_perm_b32 v6, v34, v47, s6
; GFX9-NEXT: v_perm_b32 v7, v33, v46, s6
; GFX9-NEXT: v_perm_b32 v8, v32, v45, s6
; GFX9-NEXT: v_pk_add_u16 v0, v0, 3 op_sel_hi:[1,0]
@@ -7748,39 +7751,43 @@ define <60 x half> @bitcast_v30i32_to_v60f16(<30 x i32> %a, i32 %b) {
; SI: ; %bb.0:
; SI-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; SI-NEXT: buffer_load_dword v31, off, s[0:3], s32
-; SI-NEXT: ; implicit-def: $vgpr32
-; SI-NEXT: ; kill: killed $vgpr32
-; SI-NEXT: ; implicit-def: $vgpr32
-; SI-NEXT: ; kill: killed $vgpr32
-; SI-NEXT: ; implicit-def: $vgpr32
-; SI-NEXT: ; kill: killed $vgpr32
-; SI-NEXT: ; implicit-def: $vgpr32
-; SI-NEXT: ; kill: killed $vgpr32
-; SI-NEXT: ; implicit-def: $vgpr32
-; SI-NEXT: ; kill: killed $vgpr32
-; SI-NEXT: ; implicit-def: $vgpr32
-; SI-NEXT: ; kill: killed $vgpr32
-; SI-NEXT: ; implicit-def: $vgpr32
-; SI-NEXT: ; kill: killed $vgpr32
-; SI-NEXT: ; implicit-def: $vgpr32
-; SI-NEXT: ; kill: killed $vgpr32
-; SI-NEXT: ; implicit-def: $vgpr32
-; SI-NEXT: ; kill: killed $vgpr32
-; SI-NEXT: ; implicit-def: $vgpr32
-; SI-NEXT: ; kill: killed $vgpr32
-; SI-NEXT: ; implicit-def: $vgpr32
-; SI-NEXT: ; kill: killed $vgpr32
-; SI-NEXT: ; implicit-def: $vgpr32
-; SI-NEXT: ; kill: killed $vgpr32
-; SI-NEXT: ; implicit-def: $vgpr32
-; SI-NEXT: ; kill: killed $vgpr32
-; SI-NEXT: ; implicit-def: $vgpr32
-; SI-NEXT: ; kill: killed $vgpr32
-; SI-NEXT: ; implicit-def: $vgpr32
-; SI-NEXT: ; kill: killed $vgpr32
-; SI-NEXT: ; implicit-def: $vgpr32
-; SI-NEXT: ; kill: killed $vgpr32
-; SI-NEXT: ; implicit-def: $vgpr32
+; SI-NEXT: ; implicit-def: $vgpr35
+; SI-NEXT: ; kill: killed $vgpr35
+; SI-NEXT: ; implicit-def: $vgpr35
+; SI-NEXT: ; kill: killed $vgpr35
+; SI-NEXT: ; implicit-def: $vgpr35
+; SI-NEXT: ; kill: killed $vgpr35
+; SI-NEXT: ; implicit-def: $vgpr35
+; SI-NEXT: ; kill: killed $vgpr35
+; SI-NEXT: ; implicit-def: $vgpr35
+; SI-NEXT: ; kill: killed $vgpr35
+; SI-NEXT: ; implicit-def: $vgpr35
+; SI-NEXT: ; kill: killed $vgpr35
+; SI-NEXT: ; implicit-def: $vgpr35
+; SI-NEXT: ; kill: killed $vgpr35
+; SI-NEXT: ; implicit-def: $vgpr35
+; SI-NEXT: ; kill: killed $vgpr35
+; SI-NEXT: ; implicit-def: $vgpr35
+; SI-NEXT: ; kill: killed $vgpr35
+; SI-NEXT: ; implicit-def: $vgpr35
+; SI-NEXT: ; kill: killed $vgpr35
+; SI-NEXT: ; implicit-def: $vgpr35
+; SI-NEXT: ; kill: killed $vgpr35
+; SI-NEXT: ; implicit-def: $vgpr35
+; SI-NEXT: ; kill: killed $vgpr35
+; SI-NEXT: ; implicit-def: $vgpr35
+; SI-NEXT: ; kill: killed $vgpr35
+; SI-NEXT: ; implicit-def: $vgpr35
+; SI-NEXT: ; kill: killed $vgpr35
+; SI-NEXT: ; implicit-def: $vgpr35
+; SI-NEXT: ; kill: killed $vgpr35
+; SI-NEXT: ; implicit-def: $vgpr35
+; SI-NEXT: ; kill: killed $vgpr35
+; SI-NEXT: ; implicit-def: $vgpr35
+; SI-NEXT: ; kill: killed $vgpr35
+; SI-NEXT: ; implicit-def: $vgpr35
+; SI-NEXT: ; kill: killed $vgpr35
+; SI-NEXT: ; implicit-def: $vgpr35
; SI-NEXT: buffer_store_dword v40, off, s[0:3], s32 offset:64 ; 4-byte Folded Spill
; SI-NEXT: buffer_store_dword v41, off, s[0:3], s32 offset:60 ; 4-byte Folded Spill
; SI-NEXT: buffer_store_dword v42, off, s[0:3], s32 offset:56 ; 4-byte Folded Spill
@@ -7803,10 +7810,12 @@ define <60 x half> @bitcast_v30i32_to_v60f16(<30 x i32> %a, i32 %b) {
; SI-NEXT: ; implicit-def: $vgpr47
; SI-NEXT: ; implicit-def: $vgpr44
; SI-NEXT: ; implicit-def: $vgpr45
+; SI-NEXT: ; implicit-def: $vgpr42
; SI-NEXT: ; implicit-def: $vgpr43
-; SI-NEXT: ; implicit-def: $vgpr40
-; SI-NEXT: ; implicit-def: $vgpr54
; SI-NEXT: ; implicit-def: $vgpr52
+; SI-NEXT: ; implicit-def: $vgpr40
+; SI-NEXT: ; implicit-def: $vgpr32
+; SI-NEXT: ; implicit-def: $vgpr34
; SI-NEXT: ; implicit-def: $vgpr50
; SI-NEXT: ; implicit-def: $vgpr48
; SI-NEXT: ; implicit-def: $vgpr38
@@ -7816,20 +7825,18 @@ define <60 x half> @bitcast_v30i32_to_v60f16(<30 x i32> %a, i32 %b) {
; SI-NEXT: ; implicit-def: $vgpr60
; SI-NEXT: ; implicit-def: $vgpr58
; SI-NEXT: ; implicit-def: $vgpr61
-; SI-NEXT: ; kill: killed $vgpr32
-; SI-NEXT: ; implicit-def: $vgpr39
-; SI-NEXT: ; implicit-def: $vgpr63
+; SI-NEXT: ; kill: killed $vgpr35
; SI-NEXT: ; implicit-def: $vgpr49
-; SI-NEXT: ; implicit-def: $vgpr59
+; SI-NEXT: ; implicit-def: $vgpr63
; SI-NEXT: ; implicit-def: $vgpr51
-; SI-NEXT: ; implicit-def: $vgpr37
+; SI-NEXT: ; implicit-def: $vgpr59
; SI-NEXT: ; implicit-def: $vgpr53
-; SI-NEXT: ; implicit-def: $vgpr35
+; SI-NEXT: ; implicit-def: $vgpr39
; SI-NEXT: ; implicit-def: $vgpr55
-; SI-NEXT: ; implicit-def: $vgpr34
+; SI-NEXT: ; implicit-def: $vgpr37
+; SI-NEXT: ; implicit-def: $vgpr54
+; SI-NEXT: ; implicit-def: $vgpr35
; SI-NEXT: ; implicit-def: $vgpr41
-; SI-NEXT: ; implicit-def: $vgpr32
-; SI-NEXT: ; implicit-def: $vgpr42
; SI-NEXT: s_waitcnt vmcnt(14)
; SI-NEXT: v_cmp_ne_u32_e32 vcc, 0, v31
; SI-NEXT: ; implicit-def: $vgpr31
@@ -7849,18 +7856,14 @@ define <60 x half> @bitcast_v30i32_to_v60f16(<30 x i32> %a, i32 %b) {
; SI-NEXT: ; implicit-def: $vgpr31
; SI-NEXT: ; kill: killed $vgpr31
; SI-NEXT: ; implicit-def: $vgpr31
-; SI-NEXT: ; kill: killed $vgpr31
-; SI-NEXT: ; implicit-def: $vgpr31
-; SI-NEXT: ; kill: killed $vgpr31
-; SI-NEXT: ; implicit-def: $vgpr31
; SI-NEXT: s_and_saveexec_b64 s[4:5], vcc
; SI-NEXT: s_xor_b64 s[4:5], exec, s[4:5]
; SI-NEXT: s_cbranch_execz .LBB16_2
; SI-NEXT: ; %bb.1: ; %cmp.false
; SI-NEXT: v_lshrrev_b32_e32 v31, 16, v30
-; SI-NEXT: v_cvt_f32_f16_e32 v42, v31
-; SI-NEXT: v_lshrrev_b32_e32 v31, 16, v29
; SI-NEXT: v_cvt_f32_f16_e32 v41, v31
+; SI-NEXT: v_lshrrev_b32_e32 v31, 16, v29
+; SI-NEXT: v_cvt_f32_f16_e32 v54, v31
; SI-NEXT: v_lshrrev_b32_e32 v31, 16, v28
; SI-NEXT: v_cvt_f32_f16_e32 v55, v31
; SI-NEXT: v_lshrrev_b32_e32 v31, 16, v27
@@ -7870,52 +7873,57 @@ define <60 x half> @bitcast_v30i32_to_v60f16(<30 x i32> %a, i32 %b) {
; SI-NEXT: v_lshrrev_b32_e32 v31, 16, v25
; SI-NEXT: v_cvt_f32_f16_e32 v49, v31
; SI-NEXT: v_lshrrev_b32_e32 v31, 16, v24
-; SI-NEXT: v_cvt_f32_f16_e32 v39, v31
-; SI-NEXT: v_lshrrev_b32_e32 v31, 16, v23
; SI-NEXT: s_waitcnt expcnt(2)
; SI-NEXT: v_cvt_f32_f16_e32 v61, v31
-; SI-NEXT: v_lshrrev_b32_e32 v31, 16, v22
+; SI-NEXT: v_lshrrev_b32_e32 v31, 16, v23
; SI-NEXT: v_cvt_f32_f16_e32 v31, v31
; SI-NEXT: v_lshrrev_b32_e32 v32, 16, v12
; SI-NEXT: v_cvt_f32_f16_e32 v33, v32
; SI-NEXT: v_lshrrev_b32_e32 v32, 16, v11
; SI-NEXT: buffer_store_dword v31, off, s[0:3], s32 offset:68 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_lshrrev_b32_e32 v31, 16, v21
+; SI-NEXT: v_lshrrev_b32_e32 v31, 16, v22
; SI-NEXT: v_cvt_f32_f16_e32 v31, v31
; SI-NEXT: v_cvt_f32_f16_e32 v36, v32
; SI-NEXT: v_lshrrev_b32_e32 v32, 16, v10
; SI-NEXT: v_cvt_f32_f16_e32 v38, v32
; SI-NEXT: buffer_store_dword v31, off, s[0:3], s32 offset:76 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_lshrrev_b32_e32 v31, 16, v20
+; SI-NEXT: v_lshrrev_b32_e32 v31, 16, v21
; SI-NEXT: v_cvt_f32_f16_e32 v31, v31
; SI-NEXT: v_lshrrev_b32_e32 v32, 16, v9
; SI-NEXT: v_cvt_f32_f16_e32 v48, v32
; SI-NEXT: v_lshrrev_b32_e32 v32, 16, v8
; SI-NEXT: buffer_store_dword v31, off, s[0:3], s32 offset:84 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_lshrrev_b32_e32 v31, 16, v19
+; SI-NEXT: v_lshrrev_b32_e32 v31, 16, v20
; SI-NEXT: v_cvt_f32_f16_e32 v31, v31
; SI-NEXT: v_cvt_f32_f16_e32 v50, v32
; SI-NEXT: v_lshrrev_b32_e32 v32, 16, v7
-; SI-NEXT: v_cvt_f32_f16_e32 v52, v32
-; SI-NEXT: buffer_store_dword v31, off, s[0:3], s32 offset:96 ; 4-byte Folded Spill
+; SI-NEXT: v_lshrrev_b32_e32 v35, 16, v5
+; SI-NEXT: buffer_store_dword v31, off, s[0:3], s32 offset:92 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_lshrrev_b32_e32 v31, 16, v18
+; SI-NEXT: v_lshrrev_b32_e32 v31, 16, v19
; SI-NEXT: v_cvt_f32_f16_e32 v31, v31
+; SI-NEXT: v_cvt_f32_f16_e32 v34, v32
; SI-NEXT: v_lshrrev_b32_e32 v32, 16, v6
-; SI-NEXT: v_cvt_f32_f16_e32 v54, v32
-; SI-NEXT: v_lshrrev_b32_e32 v32, 16, v5
+; SI-NEXT: v_cvt_f32_f16_e32 v40, v35
; SI-NEXT: buffer_store_dword v31, off, s[0:3], s32 offset:104 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_lshrrev_b32_e32 v31, 16, v17
+; SI-NEXT: v_lshrrev_b32_e32 v31, 16, v18
; SI-NEXT: v_cvt_f32_f16_e32 v31, v31
-; SI-NEXT: v_cvt_f32_f16_e32 v40, v32
-; SI-NEXT: v_lshrrev_b32_e32 v32, 16, v4
+; SI-NEXT: v_lshrrev_b32_e32 v35, 16, v4
+; SI-NEXT: v_cvt_f32_f16_e32 v25, v25
; SI-NEXT: v_cvt_f32_f16_e32 v24, v24
; SI-NEXT: buffer_store_dword v31, off, s[0:3], s32 offset:112 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
+; SI-NEXT: v_lshrrev_b32_e32 v31, 16, v17
+; SI-NEXT: v_cvt_f32_f16_e32 v31, v31
+; SI-NEXT: v_cvt_f32_f16_e32 v23, v23
+; SI-NEXT: v_cvt_f32_f16_e32 v22, v22
+; SI-NEXT: v_cvt_f32_f16_e32 v21, v21
+; SI-NEXT: buffer_store_dword v31, off, s[0:3], s32 offset:120 ; 4-byte Folded Spill
+; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_lshrrev_b32_e32 v31, 16, v16
; SI-NEXT: v_cvt_f32_f16_e32 v58, v31
; SI-NEXT: v_lshrrev_b32_e32 v31, 16, v15
@@ -7923,9 +7931,6 @@ define <60 x half> @bitcast_v30i32_to_v60f16(<30 x i32> %a, i32 %b) {
; SI-NEXT: v_lshrrev_b32_e32 v31, 16, v14
; SI-NEXT: v_cvt_f32_f16_e32 v62, v31
; SI-NEXT: v_lshrrev_b32_e32 v31, 16, v13
-; SI-NEXT: v_cvt_f32_f16_e32 v23, v23
-; SI-NEXT: v_cvt_f32_f16_e32 v22, v22
-; SI-NEXT: v_cvt_f32_f16_e32 v21, v21
; SI-NEXT: v_cvt_f32_f16_e32 v20, v20
; SI-NEXT: v_cvt_f32_f16_e32 v19, v19
; SI-NEXT: v_cvt_f32_f16_e32 v18, v18
@@ -7941,43 +7946,42 @@ define <60 x half> @bitcast_v30i32_to_v60f16(<30 x i32> %a, i32 %b) {
; SI-NEXT: v_cvt_f32_f16_e32 v8, v8
; SI-NEXT: v_cvt_f32_f16_e32 v7, v7
; SI-NEXT: v_cvt_f32_f16_e32 v6, v6
-; SI-NEXT: v_cvt_f32_f16_e32 v5, v5
-; SI-NEXT: v_cvt_f32_f16_e32 v4, v4
-; SI-NEXT: v_cvt_f32_f16_e32 v43, v32
-; SI-NEXT: v_lshrrev_b32_e32 v32, 16, v3
-; SI-NEXT: v_cvt_f32_f16_e32 v45, v32
-; SI-NEXT: v_lshrrev_b32_e32 v32, 16, v2
-; SI-NEXT: v_cvt_f32_f16_e32 v47, v32
-; SI-NEXT: v_lshrrev_b32_e32 v32, 16, v1
+; SI-NEXT: v_cvt_f32_f16_e32 v43, v35
+; SI-NEXT: v_lshrrev_b32_e32 v35, 16, v3
+; SI-NEXT: v_cvt_f32_f16_e32 v45, v35
+; SI-NEXT: v_lshrrev_b32_e32 v35, 16, v2
+; SI-NEXT: v_cvt_f32_f16_e32 v47, v35
+; SI-NEXT: v_lshrrev_b32_e32 v35, 16, v1
; SI-NEXT: v_cvt_f32_f16_e32 v31, v31
-; SI-NEXT: v_cvt_f32_f16_e32 v57, v32
-; SI-NEXT: v_cvt_f32_f16_e32 v32, v30
-; SI-NEXT: v_cvt_f32_f16_e32 v34, v29
-; SI-NEXT: v_cvt_f32_f16_e32 v35, v28
-; SI-NEXT: v_cvt_f32_f16_e32 v37, v27
-; SI-NEXT: v_cvt_f32_f16_e32 v59, v26
-; SI-NEXT: v_cvt_f32_f16_e32 v63, v25
-; SI-NEXT: buffer_store_dword v24, off, s[0:3], s32 offset:72 ; 4-byte Folded Spill
-; SI-NEXT: buffer_store_dword v23, off, s[0:3], s32 offset:80 ; 4-byte Folded Spill
-; SI-NEXT: buffer_store_dword v22, off, s[0:3], s32 offset:88 ; 4-byte Folded Spill
-; SI-NEXT: buffer_store_dword v21, off, s[0:3], s32 offset:92 ; 4-byte Folded Spill
-; SI-NEXT: buffer_store_dword v20, off, s[0:3], s32 offset:100 ; 4-byte Folded Spill
-; SI-NEXT: buffer_store_dword v19, off, s[0:3], s32 offset:108 ; 4-byte Folded Spill
-; SI-NEXT: buffer_store_dword v18, off, s[0:3], s32 offset:116 ; 4-byte Folded Spill
-; SI-NEXT: buffer_store_dword v17, off, s[0:3], s32 offset:120 ; 4-byte Folded Spill
-; SI-NEXT: buffer_store_dword v16, off, s[0:3], s32 offset:124 ; 4-byte Folded Spill
-; SI-NEXT: buffer_store_dword v15, off, s[0:3], s32 offset:128 ; 4-byte Folded Spill
-; SI-NEXT: buffer_store_dword v14, off, s[0:3], s32 offset:132 ; 4-byte Folded Spill
-; SI-NEXT: buffer_store_dword v13, off, s[0:3], s32 offset:136 ; 4-byte Folded Spill
-; SI-NEXT: buffer_store_dword v12, off, s[0:3], s32 offset:140 ; 4-byte Folded Spill
-; SI-NEXT: buffer_store_dword v11, off, s[0:3], s32 offset:144 ; 4-byte Folded Spill
-; SI-NEXT: buffer_store_dword v10, off, s[0:3], s32 offset:148 ; 4-byte Folded Spill
-; SI-NEXT: buffer_store_dword v9, off, s[0:3], s32 offset:152 ; 4-byte Folded Spill
-; SI-NEXT: buffer_store_dword v8, off, s[0:3], s32 offset:156 ; 4-byte Folded Spill
-; SI-NEXT: buffer_store_dword v7, off, s[0:3], s32 offset:160 ; 4-byte Folded Spill
-; SI-NEXT: buffer_store_dword v6, off, s[0:3], s32 offset:164 ; 4-byte Folded Spill
-; SI-NEXT: buffer_store_dword v5, off, s[0:3], s32 offset:168 ; 4-byte Folded Spill
-; SI-NEXT: buffer_store_dword v4, off, s[0:3], s32 offset:172 ; 4-byte Folded Spill
+; SI-NEXT: v_cvt_f32_f16_e32 v32, v32
+; SI-NEXT: v_cvt_f32_f16_e32 v57, v35
+; SI-NEXT: v_cvt_f32_f16_e32 v35, v30
+; SI-NEXT: v_cvt_f32_f16_e32 v37, v29
+; SI-NEXT: v_cvt_f32_f16_e32 v39, v28
+; SI-NEXT: v_cvt_f32_f16_e32 v59, v27
+; SI-NEXT: v_cvt_f32_f16_e32 v63, v26
+; SI-NEXT: buffer_store_dword v25, off, s[0:3], s32 offset:72 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v24, off, s[0:3], s32 offset:80 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v23, off, s[0:3], s32 offset:88 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v22, off, s[0:3], s32 offset:96 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v21, off, s[0:3], s32 offset:100 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v20, off, s[0:3], s32 offset:108 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v19, off, s[0:3], s32 offset:116 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v18, off, s[0:3], s32 offset:124 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v17, off, s[0:3], s32 offset:128 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v16, off, s[0:3], s32 offset:132 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v15, off, s[0:3], s32 offset:136 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v14, off, s[0:3], s32 offset:140 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v13, off, s[0:3], s32 offset:144 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v12, off, s[0:3], s32 offset:148 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v11, off, s[0:3], s32 offset:152 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v10, off, s[0:3], s32 offset:156 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v9, off, s[0:3], s32 offset:160 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v8, off, s[0:3], s32 offset:164 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v7, off, s[0:3], s32 offset:168 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v6, off, s[0:3], s32 offset:172 ; 4-byte Folded Spill
+; SI-NEXT: v_cvt_f32_f16_e32 v52, v5
+; SI-NEXT: v_cvt_f32_f16_e32 v42, v4
; SI-NEXT: v_cvt_f32_f16_e32 v44, v3
; SI-NEXT: v_cvt_f32_f16_e32 v46, v2
; SI-NEXT: v_cvt_f32_f16_e32 v56, v1
@@ -8015,37 +8019,61 @@ define <60 x half> @bitcast_v30i32_to_v60f16(<30 x i32> %a, i32 %b) {
; SI-NEXT: s_andn2_saveexec_b64 s[4:5], s[4:5]
; SI-NEXT: s_cbranch_execz .LBB16_4
; SI-NEXT: ; %bb.3: ; %cmp.true
-; SI-NEXT: v_add_i32_e32 v22, vcc, 3, v22
+; SI-NEXT: v_add_i32_e32 v1, vcc, 3, v1
+; SI-NEXT: v_add_i32_e32 v23, vcc, 3, v23
+; SI-NEXT: v_lshrrev_b32_e32 v31, 16, v1
; SI-NEXT: v_add_i32_e32 v2, vcc, 3, v2
-; SI-NEXT: v_lshrrev_b32_e32 v35, 16, v22
-; SI-NEXT: v_lshrrev_b32_e32 v47, 16, v2
+; SI-NEXT: v_lshrrev_b32_e32 v37, 16, v23
+; SI-NEXT: buffer_store_dword v31, off, s[0:3], s32 offset:176 ; 4-byte Folded Spill
+; SI-NEXT: s_waitcnt expcnt(0)
+; SI-NEXT: v_lshrrev_b32_e32 v31, 16, v2
; SI-NEXT: v_cvt_f32_f16_e32 v46, v2
+; SI-NEXT: v_cvt_f32_f16_e32 v2, v37
+; SI-NEXT: v_add_i32_e32 v22, vcc, 3, v22
+; SI-NEXT: v_lshrrev_b32_e32 v35, 16, v22
+; SI-NEXT: buffer_store_dword v2, off, s[0:3], s32 offset:68 ; 4-byte Folded Spill
+; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_cvt_f32_f16_e32 v2, v35
; SI-NEXT: v_add_i32_e32 v21, vcc, 3, v21
; SI-NEXT: v_lshrrev_b32_e32 v34, 16, v21
-; SI-NEXT: buffer_store_dword v2, off, s[0:3], s32 offset:68 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v2, off, s[0:3], s32 offset:76 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_cvt_f32_f16_e32 v2, v34
; SI-NEXT: v_add_i32_e32 v20, vcc, 3, v20
; SI-NEXT: v_lshrrev_b32_e32 v32, 16, v20
-; SI-NEXT: buffer_store_dword v2, off, s[0:3], s32 offset:76 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v2, off, s[0:3], s32 offset:84 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_cvt_f32_f16_e32 v2, v32
; SI-NEXT: v_add_i32_e32 v19, vcc, 3, v19
-; SI-NEXT: v_add_i32_e32 v23, vcc, 3, v23
-; SI-NEXT: v_add_i32_e32 v1, vcc, 3, v1
+; SI-NEXT: v_add_i32_e32 v24, vcc, 3, v24
; SI-NEXT: v_lshrrev_b32_e32 v63, 16, v19
-; SI-NEXT: v_lshrrev_b32_e32 v37, 16, v23
-; SI-NEXT: v_lshrrev_b32_e32 v57, 16, v1
+; SI-NEXT: v_lshrrev_b32_e32 v39, 16, v24
; SI-NEXT: v_cvt_f32_f16_e32 v56, v1
-; SI-NEXT: v_cvt_f32_f16_e32 v1, v37
-; SI-NEXT: buffer_store_dword v2, off, s[0:3], s32 offset:84 ; 4-byte Folded Spill
+; SI-NEXT: v_cvt_f32_f16_e32 v1, v39
+; SI-NEXT: buffer_store_dword v2, off, s[0:3], s32 offset:92 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_cvt_f32_f16_e32 v2, v63
-; SI-NEXT: v_add_i32_e32 v3, vcc, 3, v3
+; SI-NEXT: v_add_i32_e32 v17, vcc, 3, v17
+; SI-NEXT: v_add_i32_e32 v18, vcc, 3, v18
+; SI-NEXT: v_lshrrev_b32_e32 v59, 16, v17
+; SI-NEXT: v_lshrrev_b32_e32 v61, 16, v18
+; SI-NEXT: buffer_store_dword v2, off, s[0:3], s32 offset:104 ; 4-byte Folded Spill
+; SI-NEXT: s_waitcnt expcnt(0)
+; SI-NEXT: v_cvt_f32_f16_e32 v2, v61
+; SI-NEXT: v_mov_b32_e32 v61, v1
+; SI-NEXT: v_cvt_f32_f16_e32 v1, v59
+; SI-NEXT: buffer_store_dword v31, off, s[0:3], s32 offset:180 ; 4-byte Folded Spill
; SI-NEXT: v_add_i32_e32 v4, vcc, 3, v4
-; SI-NEXT: v_add_i32_e32 v5, vcc, 3, v5
+; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:120 ; 4-byte Folded Spill
+; SI-NEXT: s_waitcnt expcnt(0)
+; SI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:180 ; 4-byte Folded Reload
; SI-NEXT: v_add_i32_e32 v6, vcc, 3, v6
+; SI-NEXT: v_lshrrev_b32_e32 v47, 16, v4
+; SI-NEXT: v_lshrrev_b32_e32 v43, 16, v6
+; SI-NEXT: v_cvt_f32_f16_e32 v32, v43
+; SI-NEXT: v_cvt_f32_f16_e32 v43, v47
+; SI-NEXT: v_add_i32_e32 v3, vcc, 3, v3
+; SI-NEXT: v_add_i32_e32 v5, vcc, 3, v5
; SI-NEXT: v_add_i32_e32 v7, vcc, 3, v7
; SI-NEXT: v_add_i32_e32 v8, vcc, 3, v8
; SI-NEXT: v_add_i32_e32 v9, vcc, 3, v9
@@ -8056,20 +8084,15 @@ define <60 x half> @bitcast_v30i32_to_v60f16(<30 x i32> %a, i32 %b) {
; SI-NEXT: v_add_i32_e32 v14, vcc, 3, v14
; SI-NEXT: v_add_i32_e32 v15, vcc, 3, v15
; SI-NEXT: v_add_i32_e32 v16, vcc, 3, v16
-; SI-NEXT: v_add_i32_e32 v17, vcc, 3, v17
-; SI-NEXT: v_add_i32_e32 v18, vcc, 3, v18
-; SI-NEXT: v_add_i32_e32 v24, vcc, 3, v24
; SI-NEXT: v_add_i32_e32 v25, vcc, 3, v25
; SI-NEXT: v_add_i32_e32 v26, vcc, 3, v26
; SI-NEXT: v_add_i32_e32 v27, vcc, 3, v27
; SI-NEXT: v_add_i32_e32 v28, vcc, 3, v28
; SI-NEXT: v_add_i32_e32 v29, vcc, 3, v29
; SI-NEXT: v_add_i32_e32 v30, vcc, 3, v30
-; SI-NEXT: v_lshrrev_b32_e32 v45, 16, v3
-; SI-NEXT: v_lshrrev_b32_e32 v43, 16, v4
-; SI-NEXT: v_lshrrev_b32_e32 v40, 16, v5
-; SI-NEXT: v_lshrrev_b32_e32 v54, 16, v6
-; SI-NEXT: v_lshrrev_b32_e32 v52, 16, v7
+; SI-NEXT: v_lshrrev_b32_e32 v57, 16, v3
+; SI-NEXT: v_lshrrev_b32_e32 v45, 16, v5
+; SI-NEXT: v_lshrrev_b32_e32 v40, 16, v7
; SI-NEXT: v_lshrrev_b32_e32 v50, 16, v8
; SI-NEXT: v_lshrrev_b32_e32 v48, 16, v9
; SI-NEXT: v_lshrrev_b32_e32 v38, 16, v10
@@ -8079,15 +8102,12 @@ define <60 x half> @bitcast_v30i32_to_v60f16(<30 x i32> %a, i32 %b) {
; SI-NEXT: v_lshrrev_b32_e32 v62, 16, v14
; SI-NEXT: v_lshrrev_b32_e32 v60, 16, v15
; SI-NEXT: v_lshrrev_b32_e32 v58, 16, v16
-; SI-NEXT: v_lshrrev_b32_e32 v59, 16, v17
-; SI-NEXT: v_lshrrev_b32_e32 v61, 16, v18
-; SI-NEXT: v_lshrrev_b32_e32 v39, 16, v24
; SI-NEXT: v_lshrrev_b32_e32 v49, 16, v25
; SI-NEXT: v_lshrrev_b32_e32 v51, 16, v26
; SI-NEXT: v_lshrrev_b32_e32 v53, 16, v27
; SI-NEXT: v_lshrrev_b32_e32 v55, 16, v28
-; SI-NEXT: v_lshrrev_b32_e32 v41, 16, v29
-; SI-NEXT: v_lshrrev_b32_e32 v42, 16, v30
+; SI-NEXT: v_lshrrev_b32_e32 v54, 16, v29
+; SI-NEXT: v_lshrrev_b32_e32 v41, 16, v30
; SI-NEXT: v_cvt_f32_f16_e32 v30, v30
; SI-NEXT: v_cvt_f32_f16_e32 v29, v29
; SI-NEXT: v_cvt_f32_f16_e32 v28, v28
@@ -8107,27 +8127,24 @@ define <60 x half> @bitcast_v30i32_to_v60f16(<30 x i32> %a, i32 %b) {
; SI-NEXT: v_cvt_f32_f16_e32 v14, v14
; SI-NEXT: v_cvt_f32_f16_e32 v13, v13
; SI-NEXT: v_cvt_f32_f16_e32 v12, v12
+; SI-NEXT: s_waitcnt vmcnt(0)
+; SI-NEXT: v_cvt_f32_f16_e32 v47, v1
+; SI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:176 ; 4-byte Folded Reload
; SI-NEXT: v_cvt_f32_f16_e32 v11, v11
; SI-NEXT: v_cvt_f32_f16_e32 v10, v10
; SI-NEXT: v_cvt_f32_f16_e32 v9, v9
; SI-NEXT: v_cvt_f32_f16_e32 v8, v8
; SI-NEXT: v_cvt_f32_f16_e32 v7, v7
; SI-NEXT: v_cvt_f32_f16_e32 v6, v6
-; SI-NEXT: v_cvt_f32_f16_e32 v5, v5
-; SI-NEXT: v_cvt_f32_f16_e32 v4, v4
+; SI-NEXT: v_cvt_f32_f16_e32 v52, v5
+; SI-NEXT: v_cvt_f32_f16_e32 v42, v4
; SI-NEXT: v_cvt_f32_f16_e32 v44, v3
-; SI-NEXT: v_cvt_f32_f16_e32 v42, v42
; SI-NEXT: v_cvt_f32_f16_e32 v41, v41
+; SI-NEXT: v_cvt_f32_f16_e32 v54, v54
; SI-NEXT: v_cvt_f32_f16_e32 v55, v55
; SI-NEXT: v_cvt_f32_f16_e32 v53, v53
; SI-NEXT: v_cvt_f32_f16_e32 v51, v51
; SI-NEXT: v_cvt_f32_f16_e32 v49, v49
-; SI-NEXT: v_cvt_f32_f16_e32 v39, v39
-; SI-NEXT: buffer_store_dword v2, off, s[0:3], s32 offset:96 ; 4-byte Folded Spill
-; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_cvt_f32_f16_e32 v2, v61
-; SI-NEXT: v_mov_b32_e32 v61, v1
-; SI-NEXT: v_cvt_f32_f16_e32 v1, v59
; SI-NEXT: v_cvt_f32_f16_e32 v58, v58
; SI-NEXT: v_cvt_f32_f16_e32 v60, v60
; SI-NEXT: v_cvt_f32_f16_e32 v62, v62
@@ -8137,46 +8154,41 @@ define <60 x half> @bitcast_v30i32_to_v60f16(<30 x i32> %a, i32 %b) {
; SI-NEXT: v_cvt_f32_f16_e32 v38, v38
; SI-NEXT: v_cvt_f32_f16_e32 v48, v48
; SI-NEXT: v_cvt_f32_f16_e32 v50, v50
-; SI-NEXT: v_cvt_f32_f16_e32 v52, v52
-; SI-NEXT: v_cvt_f32_f16_e32 v54, v54
-; SI-NEXT: v_cvt_f32_f16_e32 v40, v40
-; SI-NEXT: v_cvt_f32_f16_e32 v43, v43
-; SI-NEXT: v_cvt_f32_f16_e32 v45, v45
-; SI-NEXT: v_cvt_f32_f16_e32 v47, v47
-; SI-NEXT: v_cvt_f32_f16_e32 v57, v57
-; SI-NEXT: v_mov_b32_e32 v37, v27
-; SI-NEXT: v_mov_b32_e32 v35, v28
-; SI-NEXT: v_mov_b32_e32 v34, v29
-; SI-NEXT: v_mov_b32_e32 v32, v30
-; SI-NEXT: v_mov_b32_e32 v63, v25
-; SI-NEXT: v_mov_b32_e32 v59, v26
-; SI-NEXT: buffer_store_dword v24, off, s[0:3], s32 offset:72 ; 4-byte Folded Spill
-; SI-NEXT: buffer_store_dword v23, off, s[0:3], s32 offset:80 ; 4-byte Folded Spill
-; SI-NEXT: buffer_store_dword v22, off, s[0:3], s32 offset:88 ; 4-byte Folded Spill
-; SI-NEXT: buffer_store_dword v21, off, s[0:3], s32 offset:92 ; 4-byte Folded Spill
-; SI-NEXT: buffer_store_dword v20, off, s[0:3], s32 offset:100 ; 4-byte Folded Spill
-; SI-NEXT: buffer_store_dword v19, off, s[0:3], s32 offset:108 ; 4-byte Folded Spill
-; SI-NEXT: buffer_store_dword v18, off, s[0:3], s32 offset:116 ; 4-byte Folded Spill
-; SI-NEXT: buffer_store_dword v17, off, s[0:3], s32 offset:120 ; 4-byte Folded Spill
-; SI-NEXT: buffer_store_dword v16, off, s[0:3], s32 offset:124 ; 4-byte Folded Spill
-; SI-NEXT: buffer_store_dword v15, off, s[0:3], s32 offset:128 ; 4-byte Folded Spill
-; SI-NEXT: buffer_store_dword v14, off, s[0:3], s32 offset:132 ; 4-byte Folded Spill
-; SI-NEXT: buffer_store_dword v13, off, s[0:3], s32 offset:136 ; 4-byte Folded Spill
-; SI-NEXT: buffer_store_dword v12, off, s[0:3], s32 offset:140 ; 4-byte Folded Spill
-; SI-NEXT: buffer_store_dword v11, off, s[0:3], s32 offset:144 ; 4-byte Folded Spill
-; SI-NEXT: buffer_store_dword v10, off, s[0:3], s32 offset:148 ; 4-byte Folded Spill
-; SI-NEXT: buffer_store_dword v9, off, s[0:3], s32 offset:152 ; 4-byte Folded Spill
-; SI-NEXT: buffer_store_dword v8, off, s[0:3], s32 offset:156 ; 4-byte Folded Spill
-; SI-NEXT: buffer_store_dword v7, off, s[0:3], s32 offset:160 ; 4-byte Folded Spill
-; SI-NEXT: buffer_store_dword v6, off, s[0:3], s32 offset:164 ; 4-byte Folded Spill
-; SI-NEXT: buffer_store_dword v5, off, s[0:3], s32 offset:168 ; 4-byte Folded Spill
-; SI-NEXT: buffer_store_dword v4, off, s[0:3], s32 offset:172 ; 4-byte Folded Spill
-; SI-NEXT: buffer_store_dword v2, off, s[0:3], s32 offset:104 ; 4-byte Folded Spill
-; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:112 ; 4-byte Folded Spill
+; SI-NEXT: v_cvt_f32_f16_e32 v34, v40
+; SI-NEXT: v_cvt_f32_f16_e32 v40, v45
+; SI-NEXT: v_cvt_f32_f16_e32 v45, v57
+; SI-NEXT: v_mov_b32_e32 v39, v28
+; SI-NEXT: v_mov_b32_e32 v37, v29
+; SI-NEXT: v_mov_b32_e32 v35, v30
+; SI-NEXT: v_mov_b32_e32 v63, v26
+; SI-NEXT: v_mov_b32_e32 v59, v27
+; SI-NEXT: buffer_store_dword v25, off, s[0:3], s32 offset:72 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v24, off, s[0:3], s32 offset:80 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v23, off, s[0:3], s32 offset:88 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v22, off, s[0:3], s32 offset:96 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v21, off, s[0:3], s32 offset:100 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v20, off, s[0:3], s32 offset:108 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v19, off, s[0:3], s32 offset:116 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v18, off, s[0:3], s32 offset:124 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v17, off, s[0:3], s32 offset:128 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v16, off, s[0:3], s32 offset:132 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v15, off, s[0:3], s32 offset:136 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v14, off, s[0:3], s32 offset:140 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v13, off, s[0:3], s32 offset:144 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v12, off, s[0:3], s32 offset:148 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v11, off, s[0:3], s32 offset:152 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v10, off, s[0:3], s32 offset:156 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v9, off, s[0:3], s32 offset:160 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v8, off, s[0:3], s32 offset:164 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v7, off, s[0:3], s32 offset:168 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v6, off, s[0:3], s32 offset:172 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v2, off, s[0:3], s32 offset:112 ; 4-byte Folded Spill
+; SI-NEXT: s_waitcnt vmcnt(14)
+; SI-NEXT: v_cvt_f32_f16_e32 v57, v1
; SI-NEXT: .LBB16_4: ; %end
; SI-NEXT: s_or_b64 exec, exec, s[4:5]
-; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_cvt_f16_f32_e32 v1, v57
+; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_cvt_f16_f32_e32 v2, v56
; SI-NEXT: v_cvt_f16_f32_e32 v3, v46
; SI-NEXT: v_lshlrev_b32_e32 v1, 16, v1
@@ -8195,43 +8207,39 @@ define <60 x half> @bitcast_v30i32_to_v60f16(<30 x i32> %a, i32 %b) {
; SI-NEXT: v_lshlrev_b32_e32 v1, 16, v1
; SI-NEXT: v_or_b32_e32 v1, v2, v1
; SI-NEXT: buffer_store_dword v1, v3, s[0:3], 0 offen
-; SI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:172 ; 4-byte Folded Reload
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_cvt_f16_f32_e32 v1, v43
+; SI-NEXT: v_cvt_f16_f32_e32 v2, v42
; SI-NEXT: v_add_i32_e32 v3, vcc, 12, v0
; SI-NEXT: v_lshlrev_b32_e32 v1, 16, v1
-; SI-NEXT: s_waitcnt vmcnt(0)
-; SI-NEXT: v_cvt_f16_f32_e32 v2, v2
; SI-NEXT: v_or_b32_e32 v1, v2, v1
; SI-NEXT: buffer_store_dword v1, v3, s[0:3], 0 offen
-; SI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:168 ; 4-byte Folded Reload
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_cvt_f16_f32_e32 v1, v40
+; SI-NEXT: v_cvt_f16_f32_e32 v2, v52
; SI-NEXT: v_add_i32_e32 v3, vcc, 16, v0
; SI-NEXT: v_lshlrev_b32_e32 v1, 16, v1
-; SI-NEXT: s_waitcnt vmcnt(0)
-; SI-NEXT: v_cvt_f16_f32_e32 v2, v2
; SI-NEXT: v_or_b32_e32 v1, v2, v1
; SI-NEXT: buffer_store_dword v1, v3, s[0:3], 0 offen
-; SI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:164 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:172 ; 4-byte Folded Reload
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_cvt_f16_f32_e32 v1, v54
+; SI-NEXT: v_cvt_f16_f32_e32 v1, v32
; SI-NEXT: v_add_i32_e32 v3, vcc, 20, v0
; SI-NEXT: v_lshlrev_b32_e32 v1, 16, v1
; SI-NEXT: s_waitcnt vmcnt(0)
; SI-NEXT: v_cvt_f16_f32_e32 v2, v2
; SI-NEXT: v_or_b32_e32 v1, v2, v1
; SI-NEXT: buffer_store_dword v1, v3, s[0:3], 0 offen
-; SI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:160 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:168 ; 4-byte Folded Reload
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_cvt_f16_f32_e32 v1, v52
+; SI-NEXT: v_cvt_f16_f32_e32 v1, v34
; SI-NEXT: v_add_i32_e32 v3, vcc, 24, v0
; SI-NEXT: v_lshlrev_b32_e32 v1, 16, v1
; SI-NEXT: s_waitcnt vmcnt(0)
; SI-NEXT: v_cvt_f16_f32_e32 v2, v2
; SI-NEXT: v_or_b32_e32 v1, v2, v1
; SI-NEXT: buffer_store_dword v1, v3, s[0:3], 0 offen
-; SI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:156 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:164 ; 4-byte Folded Reload
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_cvt_f16_f32_e32 v1, v50
; SI-NEXT: v_add_i32_e32 v3, vcc, 28, v0
@@ -8240,7 +8248,7 @@ define <60 x half> @bitcast_v30i32_to_v60f16(<30 x i32> %a, i32 %b) {
; SI-NEXT: v_cvt_f16_f32_e32 v2, v2
; SI-NEXT: v_or_b32_e32 v1, v2, v1
; SI-NEXT: buffer_store_dword v1, v3, s[0:3], 0 offen
-; SI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:152 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:160 ; 4-byte Folded Reload
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_cvt_f16_f32_e32 v1, v48
; SI-NEXT: v_add_i32_e32 v3, vcc, 32, v0
@@ -8249,7 +8257,7 @@ define <60 x half> @bitcast_v30i32_to_v60f16(<30 x i32> %a, i32 %b) {
; SI-NEXT: v_cvt_f16_f32_e32 v2, v2
; SI-NEXT: v_or_b32_e32 v1, v2, v1
; SI-NEXT: buffer_store_dword v1, v3, s[0:3], 0 offen
-; SI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:148 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:156 ; 4-byte Folded Reload
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_cvt_f16_f32_e32 v1, v38
; SI-NEXT: v_add_i32_e32 v3, vcc, 36, v0
@@ -8258,7 +8266,7 @@ define <60 x half> @bitcast_v30i32_to_v60f16(<30 x i32> %a, i32 %b) {
; SI-NEXT: v_cvt_f16_f32_e32 v2, v2
; SI-NEXT: v_or_b32_e32 v1, v2, v1
; SI-NEXT: buffer_store_dword v1, v3, s[0:3], 0 offen
-; SI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:144 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:152 ; 4-byte Folded Reload
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_cvt_f16_f32_e32 v1, v36
; SI-NEXT: v_add_i32_e32 v3, vcc, 40, v0
@@ -8267,7 +8275,7 @@ define <60 x half> @bitcast_v30i32_to_v60f16(<30 x i32> %a, i32 %b) {
; SI-NEXT: v_cvt_f16_f32_e32 v2, v2
; SI-NEXT: v_or_b32_e32 v1, v2, v1
; SI-NEXT: buffer_store_dword v1, v3, s[0:3], 0 offen
-; SI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:140 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:148 ; 4-byte Folded Reload
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_cvt_f16_f32_e32 v1, v33
; SI-NEXT: v_add_i32_e32 v3, vcc, 44, v0
@@ -8276,7 +8284,7 @@ define <60 x half> @bitcast_v30i32_to_v60f16(<30 x i32> %a, i32 %b) {
; SI-NEXT: v_cvt_f16_f32_e32 v2, v2
; SI-NEXT: v_or_b32_e32 v1, v2, v1
; SI-NEXT: buffer_store_dword v1, v3, s[0:3], 0 offen
-; SI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:136 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:144 ; 4-byte Folded Reload
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_cvt_f16_f32_e32 v1, v31
; SI-NEXT: v_add_i32_e32 v3, vcc, 48, v0
@@ -8285,7 +8293,7 @@ define <60 x half> @bitcast_v30i32_to_v60f16(<30 x i32> %a, i32 %b) {
; SI-NEXT: v_cvt_f16_f32_e32 v2, v2
; SI-NEXT: v_or_b32_e32 v1, v2, v1
; SI-NEXT: buffer_store_dword v1, v3, s[0:3], 0 offen
-; SI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:132 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:140 ; 4-byte Folded Reload
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_cvt_f16_f32_e32 v1, v62
; SI-NEXT: v_add_i32_e32 v3, vcc, 52, v0
@@ -8294,7 +8302,7 @@ define <60 x half> @bitcast_v30i32_to_v60f16(<30 x i32> %a, i32 %b) {
; SI-NEXT: v_cvt_f16_f32_e32 v2, v2
; SI-NEXT: v_or_b32_e32 v1, v2, v1
; SI-NEXT: buffer_store_dword v1, v3, s[0:3], 0 offen
-; SI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:128 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:136 ; 4-byte Folded Reload
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_cvt_f16_f32_e32 v1, v60
; SI-NEXT: v_add_i32_e32 v3, vcc, 56, v0
@@ -8303,7 +8311,7 @@ define <60 x half> @bitcast_v30i32_to_v60f16(<30 x i32> %a, i32 %b) {
; SI-NEXT: v_cvt_f16_f32_e32 v2, v2
; SI-NEXT: v_or_b32_e32 v1, v2, v1
; SI-NEXT: buffer_store_dword v1, v3, s[0:3], 0 offen
-; SI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:124 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:132 ; 4-byte Folded Reload
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_cvt_f16_f32_e32 v1, v58
; SI-NEXT: v_add_i32_e32 v3, vcc, 60, v0
@@ -8313,8 +8321,8 @@ define <60 x half> @bitcast_v30i32_to_v60f16(<30 x i32> %a, i32 %b) {
; SI-NEXT: v_or_b32_e32 v1, v2, v1
; SI-NEXT: buffer_store_dword v1, v3, s[0:3], 0 offen
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:112 ; 4-byte Folded Reload
-; SI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:120 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:120 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:128 ; 4-byte Folded Reload
; SI-NEXT: v_add_i32_e32 v3, vcc, 64, v0
; SI-NEXT: s_waitcnt vmcnt(1)
; SI-NEXT: v_cvt_f16_f32_e32 v1, v1
@@ -8324,8 +8332,8 @@ define <60 x half> @bitcast_v30i32_to_v60f16(<30 x i32> %a, i32 %b) {
; SI-NEXT: v_or_b32_e32 v1, v2, v1
; SI-NEXT: buffer_store_dword v1, v3, s[0:3], 0 offen
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:104 ; 4-byte Folded Reload
-; SI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:116 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:112 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:124 ; 4-byte Folded Reload
; SI-NEXT: v_add_i32_e32 v3, vcc, 0x44, v0
; SI-NEXT: s_waitcnt vmcnt(1)
; SI-NEXT: v_cvt_f16_f32_e32 v1, v1
@@ -8335,8 +8343,8 @@ define <60 x half> @bitcast_v30i32_to_v60f16(<30 x i32> %a, i32 %b) {
; SI-NEXT: v_or_b32_e32 v1, v2, v1
; SI-NEXT: buffer_store_dword v1, v3, s[0:3], 0 offen
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:96 ; 4-byte Folded Reload
-; SI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:108 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:104 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:116 ; 4-byte Folded Reload
; SI-NEXT: v_add_i32_e32 v3, vcc, 0x48, v0
; SI-NEXT: s_waitcnt vmcnt(1)
; SI-NEXT: v_cvt_f16_f32_e32 v1, v1
@@ -8346,8 +8354,8 @@ define <60 x half> @bitcast_v30i32_to_v60f16(<30 x i32> %a, i32 %b) {
; SI-NEXT: v_or_b32_e32 v1, v2, v1
; SI-NEXT: buffer_store_dword v1, v3, s[0:3], 0 offen
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:84 ; 4-byte Folded Reload
-; SI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:100 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:92 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:108 ; 4-byte Folded Reload
; SI-NEXT: v_add_i32_e32 v3, vcc, 0x4c, v0
; SI-NEXT: s_waitcnt vmcnt(1)
; SI-NEXT: v_cvt_f16_f32_e32 v1, v1
@@ -8357,8 +8365,8 @@ define <60 x half> @bitcast_v30i32_to_v60f16(<30 x i32> %a, i32 %b) {
; SI-NEXT: v_or_b32_e32 v1, v2, v1
; SI-NEXT: buffer_store_dword v1, v3, s[0:3], 0 offen
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:76 ; 4-byte Folded Reload
-; SI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:92 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:84 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:100 ; 4-byte Folded Reload
; SI-NEXT: v_add_i32_e32 v3, vcc, 0x50, v0
; SI-NEXT: s_waitcnt vmcnt(1)
; SI-NEXT: v_cvt_f16_f32_e32 v1, v1
@@ -8368,8 +8376,8 @@ define <60 x half> @bitcast_v30i32_to_v60f16(<30 x i32> %a, i32 %b) {
; SI-NEXT: v_or_b32_e32 v1, v2, v1
; SI-NEXT: buffer_store_dword v1, v3, s[0:3], 0 offen
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:68 ; 4-byte Folded Reload
-; SI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:88 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:76 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:96 ; 4-byte Folded Reload
; SI-NEXT: v_add_i32_e32 v3, vcc, 0x54, v0
; SI-NEXT: s_waitcnt vmcnt(1)
; SI-NEXT: v_cvt_f16_f32_e32 v1, v1
@@ -8378,62 +8386,66 @@ define <60 x half> @bitcast_v30i32_to_v60f16(<30 x i32> %a, i32 %b) {
; SI-NEXT: v_lshlrev_b32_e32 v1, 16, v1
; SI-NEXT: v_or_b32_e32 v1, v2, v1
; SI-NEXT: buffer_store_dword v1, v3, s[0:3], 0 offen
-; SI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:80 ; 4-byte Folded Reload
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_cvt_f16_f32_e32 v1, v61
+; SI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:68 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:88 ; 4-byte Folded Reload
; SI-NEXT: v_add_i32_e32 v3, vcc, 0x58, v0
-; SI-NEXT: v_lshlrev_b32_e32 v1, 16, v1
+; SI-NEXT: s_waitcnt vmcnt(1)
+; SI-NEXT: v_cvt_f16_f32_e32 v1, v1
; SI-NEXT: s_waitcnt vmcnt(0)
; SI-NEXT: v_cvt_f16_f32_e32 v2, v2
+; SI-NEXT: v_lshlrev_b32_e32 v1, 16, v1
; SI-NEXT: v_or_b32_e32 v1, v2, v1
; SI-NEXT: buffer_store_dword v1, v3, s[0:3], 0 offen
-; SI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:72 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:80 ; 4-byte Folded Reload
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_cvt_f16_f32_e32 v1, v39
+; SI-NEXT: v_cvt_f16_f32_e32 v1, v61
; SI-NEXT: v_add_i32_e32 v3, vcc, 0x5c, v0
; SI-NEXT: v_lshlrev_b32_e32 v1, 16, v1
; SI-NEXT: s_waitcnt vmcnt(0)
; SI-NEXT: v_cvt_f16_f32_e32 v2, v2
; SI-NEXT: v_or_b32_e32 v1, v2, v1
; SI-NEXT: buffer_store_dword v1, v3, s[0:3], 0 offen
+; SI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:72 ; 4-byte Folded Reload
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_cvt_f16_f32_e32 v1, v49
-; SI-NEXT: v_cvt_f16_f32_e32 v2, v63
; SI-NEXT: v_add_i32_e32 v3, vcc, 0x60, v0
; SI-NEXT: v_lshlrev_b32_e32 v1, 16, v1
+; SI-NEXT: s_waitcnt vmcnt(0)
+; SI-NEXT: v_cvt_f16_f32_e32 v2, v2
; SI-NEXT: v_or_b32_e32 v1, v2, v1
; SI-NEXT: buffer_store_dword v1, v3, s[0:3], 0 offen
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_cvt_f16_f32_e32 v1, v51
-; SI-NEXT: v_cvt_f16_f32_e32 v2, v59
+; SI-NEXT: v_cvt_f16_f32_e32 v2, v63
; SI-NEXT: v_add_i32_e32 v3, vcc, 0x64, v0
; SI-NEXT: v_lshlrev_b32_e32 v1, 16, v1
; SI-NEXT: v_or_b32_e32 v1, v2, v1
; SI-NEXT: buffer_store_dword v1, v3, s[0:3], 0 offen
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_cvt_f16_f32_e32 v1, v53
-; SI-NEXT: v_cvt_f16_f32_e32 v2, v37
+; SI-NEXT: v_cvt_f16_f32_e32 v2, v59
; SI-NEXT: v_add_i32_e32 v3, vcc, 0x68, v0
; SI-NEXT: v_lshlrev_b32_e32 v1, 16, v1
; SI-NEXT: v_or_b32_e32 v1, v2, v1
; SI-NEXT: buffer_store_dword v1, v3, s[0:3], 0 offen
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_cvt_f16_f32_e32 v1, v55
-; SI-NEXT: v_cvt_f16_f32_e32 v2, v35
+; SI-NEXT: v_cvt_f16_f32_e32 v2, v39
; SI-NEXT: v_add_i32_e32 v3, vcc, 0x6c, v0
; SI-NEXT: v_lshlrev_b32_e32 v1, 16, v1
; SI-NEXT: v_or_b32_e32 v1, v2, v1
; SI-NEXT: buffer_store_dword v1, v3, s[0:3], 0 offen
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_cvt_f16_f32_e32 v1, v41
-; SI-NEXT: v_cvt_f16_f32_e32 v2, v34
+; SI-NEXT: v_cvt_f16_f32_e32 v1, v54
+; SI-NEXT: v_cvt_f16_f32_e32 v2, v37
; SI-NEXT: v_add_i32_e32 v3, vcc, 0x70, v0
; SI-NEXT: v_lshlrev_b32_e32 v1, 16, v1
; SI-NEXT: v_or_b32_e32 v1, v2, v1
; SI-NEXT: buffer_store_dword v1, v3, s[0:3], 0 offen
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_cvt_f16_f32_e32 v1, v42
-; SI-NEXT: v_cvt_f16_f32_e32 v2, v32
+; SI-NEXT: v_cvt_f16_f32_e32 v1, v41
+; SI-NEXT: v_cvt_f16_f32_e32 v2, v35
; SI-NEXT: v_add_i32_e32 v0, vcc, 0x74, v0
; SI-NEXT: v_lshlrev_b32_e32 v1, 16, v1
; SI-NEXT: v_or_b32_e32 v1, v2, v1
@@ -10177,7 +10189,7 @@ define inreg <60 x half> @bitcast_v30i32_to_v60f16_scalar(<30 x i32> inreg %a, i
; GFX11-TRUE16-NEXT: v_readfirstlane_b32 s5, v9
; GFX11-TRUE16-NEXT: v_readfirstlane_b32 s6, v10
; GFX11-TRUE16-NEXT: v_readfirstlane_b32 s4, v11
-; GFX11-TRUE16-NEXT: s_mov_b32 s94, 0
+; GFX11-TRUE16-NEXT: s_mov_b32 s89, 0
; GFX11-TRUE16-NEXT: s_and_b32 s40, vcc_lo, exec_lo
; GFX11-TRUE16-NEXT: s_cbranch_scc0 .LBB17_4
; GFX11-TRUE16-NEXT: ; %bb.1: ; %cmp.false
@@ -10206,12 +10218,12 @@ define inreg <60 x half> @bitcast_v30i32_to_v60f16_scalar(<30 x i32> inreg %a, i
; GFX11-TRUE16-NEXT: s_lshr_b32 s78, s19, 16
; GFX11-TRUE16-NEXT: s_lshr_b32 s79, s18, 16
; GFX11-TRUE16-NEXT: s_lshr_b32 s88, s17, 16
-; GFX11-TRUE16-NEXT: s_lshr_b32 s89, s16, 16
-; GFX11-TRUE16-NEXT: s_lshr_b32 s90, s3, 16
-; GFX11-TRUE16-NEXT: s_lshr_b32 s91, s2, 16
-; GFX11-TRUE16-NEXT: s_lshr_b32 s92, s1, 16
-; GFX11-TRUE16-NEXT: s_lshr_b32 s93, s0, 16
-; GFX11-TRUE16-NEXT: s_and_not1_b32 vcc_lo, exec_lo, s94
+; GFX11-TRUE16-NEXT: s_lshr_b32 s90, s16, 16
+; GFX11-TRUE16-NEXT: s_lshr_b32 s91, s3, 16
+; GFX11-TRUE16-NEXT: s_lshr_b32 s92, s2, 16
+; GFX11-TRUE16-NEXT: s_lshr_b32 s93, s1, 16
+; GFX11-TRUE16-NEXT: s_lshr_b32 s94, s0, 16
+; GFX11-TRUE16-NEXT: s_and_not1_b32 vcc_lo, exec_lo, s89
; GFX11-TRUE16-NEXT: s_cbranch_vccnz .LBB17_3
; GFX11-TRUE16-NEXT: .LBB17_2: ; %cmp.true
; GFX11-TRUE16-NEXT: s_add_i32 s4, s4, 3
@@ -10269,19 +10281,22 @@ define inreg <60 x half> @bitcast_v30i32_to_v60f16_scalar(<30 x i32> inreg %a, i
; GFX11-TRUE16-NEXT: s_lshr_b32 s78, s19, 16
; GFX11-TRUE16-NEXT: s_lshr_b32 s79, s18, 16
; GFX11-TRUE16-NEXT: s_lshr_b32 s88, s17, 16
-; GFX11-TRUE16-NEXT: s_lshr_b32 s89, s16, 16
-; GFX11-TRUE16-NEXT: s_lshr_b32 s90, s3, 16
-; GFX11-TRUE16-NEXT: s_lshr_b32 s91, s2, 16
-; GFX11-TRUE16-NEXT: s_lshr_b32 s92, s1, 16
-; GFX11-TRUE16-NEXT: s_lshr_b32 s93, s0, 16
+; GFX11-TRUE16-NEXT: s_lshr_b32 s90, s16, 16
+; GFX11-TRUE16-NEXT: s_lshr_b32 s91, s3, 16
+; GFX11-TRUE16-NEXT: s_lshr_b32 s92, s2, 16
+; GFX11-TRUE16-NEXT: s_lshr_b32 s93, s1, 16
+; GFX11-TRUE16-NEXT: s_lshr_b32 s94, s0, 16
; GFX11-TRUE16-NEXT: .LBB17_3: ; %end
; GFX11-TRUE16-NEXT: s_delay_alu instid0(SALU_CYCLE_1)
-; GFX11-TRUE16-NEXT: s_pack_ll_b32_b16 s0, s0, s93
-; GFX11-TRUE16-NEXT: s_pack_ll_b32_b16 s1, s1, s92
-; GFX11-TRUE16-NEXT: s_pack_ll_b32_b16 s2, s2, s91
-; GFX11-TRUE16-NEXT: s_pack_ll_b32_b16 s3, s3, s90
-; GFX11-TRUE16-NEXT: s_pack_ll_b32_b16 s16, s16, s89
+; GFX11-TRUE16-NEXT: s_mov_b32 s89, s94
+; GFX11-TRUE16-NEXT: s_pack_ll_b32_b16 s1, s1, s93
+; GFX11-TRUE16-NEXT: s_pack_ll_b32_b16 s0, s0, s89
+; GFX11-TRUE16-NEXT: s_mov_b32 s89, s92
+; GFX11-TRUE16-NEXT: s_pack_ll_b32_b16 s3, s3, s91
+; GFX11-TRUE16-NEXT: s_pack_ll_b32_b16 s2, s2, s89
+; GFX11-TRUE16-NEXT: s_mov_b32 s89, s90
; GFX11-TRUE16-NEXT: s_pack_ll_b32_b16 s17, s17, s88
+; GFX11-TRUE16-NEXT: s_pack_ll_b32_b16 s16, s16, s89
; GFX11-TRUE16-NEXT: s_pack_ll_b32_b16 s18, s18, s79
; GFX11-TRUE16-NEXT: s_pack_ll_b32_b16 s19, s19, s78
; GFX11-TRUE16-NEXT: s_pack_ll_b32_b16 s20, s20, s77
@@ -10323,11 +10338,11 @@ define inreg <60 x half> @bitcast_v30i32_to_v60f16_scalar(<30 x i32> inreg %a, i
; GFX11-TRUE16-NEXT: v_dual_mov_b32 v28, s6 :: v_dual_mov_b32 v29, s4
; GFX11-TRUE16-NEXT: s_setpc_b64 s[30:31]
; GFX11-TRUE16-NEXT: .LBB17_4:
+; GFX11-TRUE16-NEXT: ; implicit-def: $sgpr94_lo16
; GFX11-TRUE16-NEXT: ; implicit-def: $sgpr93_lo16
; GFX11-TRUE16-NEXT: ; implicit-def: $sgpr92_lo16
; GFX11-TRUE16-NEXT: ; implicit-def: $sgpr91_lo16
; GFX11-TRUE16-NEXT: ; implicit-def: $sgpr90_lo16
-; GFX11-TRUE16-NEXT: ; implicit-def: $sgpr89_lo16
; GFX11-TRUE16-NEXT: ; implicit-def: $sgpr88_lo16
; GFX11-TRUE16-NEXT: ; implicit-def: $sgpr79_lo16
; GFX11-TRUE16-NEXT: ; implicit-def: $sgpr78_lo16
@@ -12219,9 +12234,9 @@ define inreg <30 x i32> @bitcast_v60f16_to_v30i32_scalar(<60 x half> inreg %a, i
; SI-NEXT: v_cvt_f16_f32_e32 v1, s19
; SI-NEXT: v_cvt_f16_f32_e32 v12, s18
; SI-NEXT: v_cvt_f16_f32_e32 v2, s21
-; SI-NEXT: v_cvt_f16_f32_e32 v11, s20
+; SI-NEXT: v_cvt_f16_f32_e32 v10, s20
; SI-NEXT: v_cvt_f16_f32_e32 v3, s23
-; SI-NEXT: v_cvt_f16_f32_e32 v10, s22
+; SI-NEXT: v_cvt_f16_f32_e32 v11, s22
; SI-NEXT: v_cvt_f16_f32_e32 v4, s25
; SI-NEXT: v_cvt_f16_f32_e32 v9, s24
; SI-NEXT: v_cvt_f16_f32_e32 v5, s27
@@ -12300,9 +12315,9 @@ define inreg <30 x i32> @bitcast_v60f16_to_v30i32_scalar(<60 x half> inreg %a, i
; SI-NEXT: ; %bb.1: ; %cmp.false
; SI-NEXT: buffer_load_dword v24, off, s[0:3], s32 offset:172 ; 4-byte Folded Reload
; SI-NEXT: buffer_load_dword v25, off, s[0:3], s32 offset:168 ; 4-byte Folded Reload
-; SI-NEXT: s_waitcnt expcnt(4)
-; SI-NEXT: v_lshlrev_b32_e32 v3, 16, v3
-; SI-NEXT: v_or_b32_e32 v3, v10, v3
+; SI-NEXT: s_waitcnt expcnt(5)
+; SI-NEXT: v_lshlrev_b32_e32 v2, 16, v2
+; SI-NEXT: v_or_b32_e32 v2, v10, v2
; SI-NEXT: s_waitcnt expcnt(3)
; SI-NEXT: v_lshlrev_b32_e32 v10, 16, v34
; SI-NEXT: v_mov_b32_e32 v33, v32
@@ -12330,13 +12345,13 @@ define inreg <30 x i32> @bitcast_v60f16_to_v30i32_scalar(<60 x half> inreg %a, i
; SI-NEXT: buffer_load_dword v54, off, s[0:3], s32 offset:128 ; 4-byte Folded Reload
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_lshlrev_b32_e32 v1, 16, v1
-; SI-NEXT: v_lshlrev_b32_e32 v2, 16, v2
+; SI-NEXT: v_lshlrev_b32_e32 v3, 16, v3
; SI-NEXT: v_lshlrev_b32_e32 v4, 16, v4
; SI-NEXT: v_lshlrev_b32_e32 v18, 16, v18
; SI-NEXT: v_lshlrev_b32_e32 v19, 16, v19
; SI-NEXT: v_lshlrev_b32_e32 v0, 16, v59
; SI-NEXT: v_or_b32_e32 v1, v12, v1
-; SI-NEXT: v_or_b32_e32 v2, v11, v2
+; SI-NEXT: v_or_b32_e32 v3, v11, v3
; SI-NEXT: v_or_b32_e32 v4, v9, v4
; SI-NEXT: v_lshlrev_b32_e32 v11, 16, v36
; SI-NEXT: v_lshlrev_b32_e32 v12, 16, v63
@@ -12501,7 +12516,7 @@ define inreg <30 x i32> @bitcast_v60f16_to_v30i32_scalar(<60 x half> inreg %a, i
; SI-NEXT: buffer_load_dword v28, off, s[0:3], s32 offset:152 ; 4-byte Folded Reload
; SI-NEXT: buffer_load_dword v29, off, s[0:3], s32 offset:144 ; 4-byte Folded Reload
; SI-NEXT: buffer_load_dword v31, off, s[0:3], s32 offset:128 ; 4-byte Folded Reload
-; SI-NEXT: buffer_load_dword v4, off, s[0:3], s32 offset:276 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v4, off, s[0:3], s32 offset:280 ; 4-byte Folded Reload
; SI-NEXT: buffer_load_dword v5, off, s[0:3], s32 offset:260 ; 4-byte Folded Reload
; SI-NEXT: buffer_load_dword v6, off, s[0:3], s32 offset:252 ; 4-byte Folded Reload
; SI-NEXT: buffer_load_dword v7, off, s[0:3], s32 offset:244 ; 4-byte Folded Reload
@@ -12523,7 +12538,7 @@ define inreg <30 x i32> @bitcast_v60f16_to_v30i32_scalar(<60 x half> inreg %a, i
; SI-NEXT: v_lshlrev_b32_e32 v2, 16, v2
; SI-NEXT: v_or_b32_e32 v1, v3, v2
; SI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:268 ; 4-byte Folded Reload
-; SI-NEXT: buffer_load_dword v3, off, s[0:3], s32 offset:280 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v3, off, s[0:3], s32 offset:276 ; 4-byte Folded Reload
; SI-NEXT: s_waitcnt vmcnt(14)
; SI-NEXT: v_cvt_f32_f16_e32 v27, v27
; SI-NEXT: s_waitcnt vmcnt(11)
@@ -18557,8 +18572,8 @@ define <30 x float> @bitcast_v60i16_to_v30f32(<60 x i16> %a, i32 %b) {
; GFX9-NEXT: v_lshrrev_b32_e32 v0, 16, v44
; GFX9-NEXT: v_lshrrev_b32_e32 v32, 16, v45
; GFX9-NEXT: v_lshrrev_b32_e32 v33, 16, v46
-; GFX9-NEXT: v_lshrrev_b32_e32 v63, 16, v47
-; GFX9-NEXT: v_lshrrev_b32_e32 v34, 16, v56
+; GFX9-NEXT: v_lshrrev_b32_e32 v34, 16, v47
+; GFX9-NEXT: v_lshrrev_b32_e32 v63, 16, v56
; GFX9-NEXT: v_lshrrev_b32_e32 v35, 16, v57
; GFX9-NEXT: v_lshrrev_b32_e32 v36, 16, v58
; GFX9-NEXT: v_lshrrev_b32_e32 v62, 16, v59
@@ -18593,8 +18608,8 @@ define <30 x float> @bitcast_v60i16_to_v30f32(<60 x i16> %a, i32 %b) {
; GFX9-NEXT: v_perm_b32 v2, v62, v59, s6
; GFX9-NEXT: v_perm_b32 v3, v36, v58, s6
; GFX9-NEXT: v_perm_b32 v4, v35, v57, s6
-; GFX9-NEXT: v_perm_b32 v5, v34, v56, s6
-; GFX9-NEXT: v_perm_b32 v6, v63, v47, s6
+; GFX9-NEXT: v_perm_b32 v5, v63, v56, s6
+; GFX9-NEXT: v_perm_b32 v6, v34, v47, s6
; GFX9-NEXT: v_perm_b32 v7, v33, v46, s6
; GFX9-NEXT: v_perm_b32 v8, v32, v45, s6
; GFX9-NEXT: v_perm_b32 v9, v9, v44, s6
@@ -18636,8 +18651,8 @@ define <30 x float> @bitcast_v60i16_to_v30f32(<60 x i16> %a, i32 %b) {
; GFX9-NEXT: ; implicit-def: $vgpr39
; GFX9-NEXT: ; implicit-def: $vgpr32
; GFX9-NEXT: ; implicit-def: $vgpr33
-; GFX9-NEXT: ; implicit-def: $vgpr63
; GFX9-NEXT: ; implicit-def: $vgpr34
+; GFX9-NEXT: ; implicit-def: $vgpr63
; GFX9-NEXT: ; implicit-def: $vgpr35
; GFX9-NEXT: ; implicit-def: $vgpr36
; GFX9-NEXT: ; implicit-def: $vgpr62
@@ -18751,8 +18766,8 @@ define <30 x float> @bitcast_v60i16_to_v30f32(<60 x i16> %a, i32 %b) {
; GFX9-NEXT: v_perm_b32 v2, v62, v59, s6
; GFX9-NEXT: v_perm_b32 v3, v36, v58, s6
; GFX9-NEXT: v_perm_b32 v4, v35, v57, s6
-; GFX9-NEXT: v_perm_b32 v5, v34, v56, s6
-; GFX9-NEXT: v_perm_b32 v6, v63, v47, s6
+; GFX9-NEXT: v_perm_b32 v5, v63, v56, s6
+; GFX9-NEXT: v_perm_b32 v6, v34, v47, s6
; GFX9-NEXT: v_perm_b32 v7, v33, v46, s6
; GFX9-NEXT: v_perm_b32 v8, v32, v45, s6
; GFX9-NEXT: v_pk_add_u16 v0, v0, 3 op_sel_hi:[1,0]
@@ -20299,39 +20314,43 @@ define <60 x half> @bitcast_v30f32_to_v60f16(<30 x float> %a, i32 %b) {
; SI: ; %bb.0:
; SI-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; SI-NEXT: buffer_load_dword v31, off, s[0:3], s32
-; SI-NEXT: ; implicit-def: $vgpr32
-; SI-NEXT: ; kill: killed $vgpr32
-; SI-NEXT: ; implicit-def: $vgpr32
-; SI-NEXT: ; kill: killed $vgpr32
-; SI-NEXT: ; implicit-def: $vgpr32
-; SI-NEXT: ; kill: killed $vgpr32
-; SI-NEXT: ; implicit-def: $vgpr32
-; SI-NEXT: ; kill: killed $vgpr32
-; SI-NEXT: ; implicit-def: $vgpr32
-; SI-NEXT: ; kill: killed $vgpr32
-; SI-NEXT: ; implicit-def: $vgpr32
-; SI-NEXT: ; kill: killed $vgpr32
-; SI-NEXT: ; implicit-def: $vgpr32
-; SI-NEXT: ; kill: killed $vgpr32
-; SI-NEXT: ; implicit-def: $vgpr32
-; SI-NEXT: ; kill: killed $vgpr32
-; SI-NEXT: ; implicit-def: $vgpr32
-; SI-NEXT: ; kill: killed $vgpr32
-; SI-NEXT: ; implicit-def: $vgpr32
-; SI-NEXT: ; kill: killed $vgpr32
-; SI-NEXT: ; implicit-def: $vgpr32
-; SI-NEXT: ; kill: killed $vgpr32
-; SI-NEXT: ; implicit-def: $vgpr32
-; SI-NEXT: ; kill: killed $vgpr32
-; SI-NEXT: ; implicit-def: $vgpr32
-; SI-NEXT: ; kill: killed $vgpr32
-; SI-NEXT: ; implicit-def: $vgpr32
-; SI-NEXT: ; kill: killed $vgpr32
-; SI-NEXT: ; implicit-def: $vgpr32
-; SI-NEXT: ; kill: killed $vgpr32
-; SI-NEXT: ; implicit-def: $vgpr32
-; SI-NEXT: ; kill: killed $vgpr32
-; SI-NEXT: ; implicit-def: $vgpr32
+; SI-NEXT: ; implicit-def: $vgpr35
+; SI-NEXT: ; kill: killed $vgpr35
+; SI-NEXT: ; implicit-def: $vgpr35
+; SI-NEXT: ; kill: killed $vgpr35
+; SI-NEXT: ; implicit-def: $vgpr35
+; SI-NEXT: ; kill: killed $vgpr35
+; SI-NEXT: ; implicit-def: $vgpr35
+; SI-NEXT: ; kill: killed $vgpr35
+; SI-NEXT: ; implicit-def: $vgpr35
+; SI-NEXT: ; kill: killed $vgpr35
+; SI-NEXT: ; implicit-def: $vgpr35
+; SI-NEXT: ; kill: killed $vgpr35
+; SI-NEXT: ; implicit-def: $vgpr35
+; SI-NEXT: ; kill: killed $vgpr35
+; SI-NEXT: ; implicit-def: $vgpr35
+; SI-NEXT: ; kill: killed $vgpr35
+; SI-NEXT: ; implicit-def: $vgpr35
+; SI-NEXT: ; kill: killed $vgpr35
+; SI-NEXT: ; implicit-def: $vgpr35
+; SI-NEXT: ; kill: killed $vgpr35
+; SI-NEXT: ; implicit-def: $vgpr35
+; SI-NEXT: ; kill: killed $vgpr35
+; SI-NEXT: ; implicit-def: $vgpr35
+; SI-NEXT: ; kill: killed $vgpr35
+; SI-NEXT: ; implicit-def: $vgpr35
+; SI-NEXT: ; kill: killed $vgpr35
+; SI-NEXT: ; implicit-def: $vgpr35
+; SI-NEXT: ; kill: killed $vgpr35
+; SI-NEXT: ; implicit-def: $vgpr35
+; SI-NEXT: ; kill: killed $vgpr35
+; SI-NEXT: ; implicit-def: $vgpr35
+; SI-NEXT: ; kill: killed $vgpr35
+; SI-NEXT: ; implicit-def: $vgpr35
+; SI-NEXT: ; kill: killed $vgpr35
+; SI-NEXT: ; implicit-def: $vgpr35
+; SI-NEXT: ; kill: killed $vgpr35
+; SI-NEXT: ; implicit-def: $vgpr35
; SI-NEXT: buffer_store_dword v40, off, s[0:3], s32 offset:64 ; 4-byte Folded Spill
; SI-NEXT: buffer_store_dword v41, off, s[0:3], s32 offset:60 ; 4-byte Folded Spill
; SI-NEXT: buffer_store_dword v42, off, s[0:3], s32 offset:56 ; 4-byte Folded Spill
@@ -20354,10 +20373,12 @@ define <60 x half> @bitcast_v30f32_to_v60f16(<30 x float> %a, i32 %b) {
; SI-NEXT: ; implicit-def: $vgpr47
; SI-NEXT: ; implicit-def: $vgpr44
; SI-NEXT: ; implicit-def: $vgpr45
+; SI-NEXT: ; implicit-def: $vgpr42
; SI-NEXT: ; implicit-def: $vgpr43
-; SI-NEXT: ; implicit-def: $vgpr40
-; SI-NEXT: ; implicit-def: $vgpr54
; SI-NEXT: ; implicit-def: $vgpr52
+; SI-NEXT: ; implicit-def: $vgpr40
+; SI-NEXT: ; implicit-def: $vgpr32
+; SI-NEXT: ; implicit-def: $vgpr34
; SI-NEXT: ; implicit-def: $vgpr50
; SI-NEXT: ; implicit-def: $vgpr48
; SI-NEXT: ; implicit-def: $vgpr38
@@ -20367,20 +20388,18 @@ define <60 x half> @bitcast_v30f32_to_v60f16(<30 x float> %a, i32 %b) {
; SI-NEXT: ; implicit-def: $vgpr60
; SI-NEXT: ; implicit-def: $vgpr58
; SI-NEXT: ; implicit-def: $vgpr61
-; SI-NEXT: ; kill: killed $vgpr32
-; SI-NEXT: ; implicit-def: $vgpr39
-; SI-NEXT: ; implicit-def: $vgpr63
+; SI-NEXT: ; kill: killed $vgpr35
; SI-NEXT: ; implicit-def: $vgpr49
-; SI-NEXT: ; implicit-def: $vgpr59
+; SI-NEXT: ; implicit-def: $vgpr63
; SI-NEXT: ; implicit-def: $vgpr51
-; SI-NEXT: ; implicit-def: $vgpr37
+; SI-NEXT: ; implicit-def: $vgpr59
; SI-NEXT: ; implicit-def: $vgpr53
-; SI-NEXT: ; implicit-def: $vgpr35
+; SI-NEXT: ; implicit-def: $vgpr39
; SI-NEXT: ; implicit-def: $vgpr55
-; SI-NEXT: ; implicit-def: $vgpr34
+; SI-NEXT: ; implicit-def: $vgpr37
+; SI-NEXT: ; implicit-def: $vgpr54
+; SI-NEXT: ; implicit-def: $vgpr35
; SI-NEXT: ; implicit-def: $vgpr41
-; SI-NEXT: ; implicit-def: $vgpr32
-; SI-NEXT: ; implicit-def: $vgpr42
; SI-NEXT: s_waitcnt vmcnt(14)
; SI-NEXT: v_cmp_ne_u32_e32 vcc, 0, v31
; SI-NEXT: ; implicit-def: $vgpr31
@@ -20400,18 +20419,14 @@ define <60 x half> @bitcast_v30f32_to_v60f16(<30 x float> %a, i32 %b) {
; SI-NEXT: ; implicit-def: $vgpr31
; SI-NEXT: ; kill: killed $vgpr31
; SI-NEXT: ; implicit-def: $vgpr31
-; SI-NEXT: ; kill: killed $vgpr31
-; SI-NEXT: ; implicit-def: $vgpr31
-; SI-NEXT: ; kill: killed $vgpr31
-; SI-NEXT: ; implicit-def: $vgpr31
; SI-NEXT: s_and_saveexec_b64 s[4:5], vcc
; SI-NEXT: s_xor_b64 s[4:5], exec, s[4:5]
; SI-NEXT: s_cbranch_execz .LBB32_2
; SI-NEXT: ; %bb.1: ; %cmp.false
; SI-NEXT: v_lshrrev_b32_e32 v31, 16, v30
-; SI-NEXT: v_cvt_f32_f16_e32 v42, v31
-; SI-NEXT: v_lshrrev_b32_e32 v31, 16, v29
; SI-NEXT: v_cvt_f32_f16_e32 v41, v31
+; SI-NEXT: v_lshrrev_b32_e32 v31, 16, v29
+; SI-NEXT: v_cvt_f32_f16_e32 v54, v31
; SI-NEXT: v_lshrrev_b32_e32 v31, 16, v28
; SI-NEXT: v_cvt_f32_f16_e32 v55, v31
; SI-NEXT: v_lshrrev_b32_e32 v31, 16, v27
@@ -20421,52 +20436,57 @@ define <60 x half> @bitcast_v30f32_to_v60f16(<30 x float> %a, i32 %b) {
; SI-NEXT: v_lshrrev_b32_e32 v31, 16, v25
; SI-NEXT: v_cvt_f32_f16_e32 v49, v31
; SI-NEXT: v_lshrrev_b32_e32 v31, 16, v24
-; SI-NEXT: v_cvt_f32_f16_e32 v39, v31
-; SI-NEXT: v_lshrrev_b32_e32 v31, 16, v23
; SI-NEXT: s_waitcnt expcnt(2)
; SI-NEXT: v_cvt_f32_f16_e32 v61, v31
-; SI-NEXT: v_lshrrev_b32_e32 v31, 16, v22
+; SI-NEXT: v_lshrrev_b32_e32 v31, 16, v23
; SI-NEXT: v_cvt_f32_f16_e32 v31, v31
; SI-NEXT: v_lshrrev_b32_e32 v32, 16, v12
; SI-NEXT: v_cvt_f32_f16_e32 v33, v32
; SI-NEXT: v_lshrrev_b32_e32 v32, 16, v11
; SI-NEXT: buffer_store_dword v31, off, s[0:3], s32 offset:68 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_lshrrev_b32_e32 v31, 16, v21
+; SI-NEXT: v_lshrrev_b32_e32 v31, 16, v22
; SI-NEXT: v_cvt_f32_f16_e32 v31, v31
; SI-NEXT: v_cvt_f32_f16_e32 v36, v32
; SI-NEXT: v_lshrrev_b32_e32 v32, 16, v10
; SI-NEXT: v_cvt_f32_f16_e32 v38, v32
; SI-NEXT: buffer_store_dword v31, off, s[0:3], s32 offset:76 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_lshrrev_b32_e32 v31, 16, v20
+; SI-NEXT: v_lshrrev_b32_e32 v31, 16, v21
; SI-NEXT: v_cvt_f32_f16_e32 v31, v31
; SI-NEXT: v_lshrrev_b32_e32 v32, 16, v9
; SI-NEXT: v_cvt_f32_f16_e32 v48, v32
; SI-NEXT: v_lshrrev_b32_e32 v32, 16, v8
; SI-NEXT: buffer_store_dword v31, off, s[0:3], s32 offset:84 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_lshrrev_b32_e32 v31, 16, v19
+; SI-NEXT: v_lshrrev_b32_e32 v31, 16, v20
; SI-NEXT: v_cvt_f32_f16_e32 v31, v31
; SI-NEXT: v_cvt_f32_f16_e32 v50, v32
; SI-NEXT: v_lshrrev_b32_e32 v32, 16, v7
-; SI-NEXT: v_cvt_f32_f16_e32 v52, v32
-; SI-NEXT: buffer_store_dword v31, off, s[0:3], s32 offset:96 ; 4-byte Folded Spill
+; SI-NEXT: v_lshrrev_b32_e32 v35, 16, v5
+; SI-NEXT: buffer_store_dword v31, off, s[0:3], s32 offset:92 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_lshrrev_b32_e32 v31, 16, v18
+; SI-NEXT: v_lshrrev_b32_e32 v31, 16, v19
; SI-NEXT: v_cvt_f32_f16_e32 v31, v31
+; SI-NEXT: v_cvt_f32_f16_e32 v34, v32
; SI-NEXT: v_lshrrev_b32_e32 v32, 16, v6
-; SI-NEXT: v_cvt_f32_f16_e32 v54, v32
-; SI-NEXT: v_lshrrev_b32_e32 v32, 16, v5
+; SI-NEXT: v_cvt_f32_f16_e32 v40, v35
; SI-NEXT: buffer_store_dword v31, off, s[0:3], s32 offset:104 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_lshrrev_b32_e32 v31, 16, v17
+; SI-NEXT: v_lshrrev_b32_e32 v31, 16, v18
; SI-NEXT: v_cvt_f32_f16_e32 v31, v31
-; SI-NEXT: v_cvt_f32_f16_e32 v40, v32
-; SI-NEXT: v_lshrrev_b32_e32 v32, 16, v4
+; SI-NEXT: v_lshrrev_b32_e32 v35, 16, v4
+; SI-NEXT: v_cvt_f32_f16_e32 v25, v25
; SI-NEXT: v_cvt_f32_f16_e32 v24, v24
; SI-NEXT: buffer_store_dword v31, off, s[0:3], s32 offset:112 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
+; SI-NEXT: v_lshrrev_b32_e32 v31, 16, v17
+; SI-NEXT: v_cvt_f32_f16_e32 v31, v31
+; SI-NEXT: v_cvt_f32_f16_e32 v23, v23
+; SI-NEXT: v_cvt_f32_f16_e32 v22, v22
+; SI-NEXT: v_cvt_f32_f16_e32 v21, v21
+; SI-NEXT: buffer_store_dword v31, off, s[0:3], s32 offset:120 ; 4-byte Folded Spill
+; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_lshrrev_b32_e32 v31, 16, v16
; SI-NEXT: v_cvt_f32_f16_e32 v58, v31
; SI-NEXT: v_lshrrev_b32_e32 v31, 16, v15
@@ -20474,9 +20494,6 @@ define <60 x half> @bitcast_v30f32_to_v60f16(<30 x float> %a, i32 %b) {
; SI-NEXT: v_lshrrev_b32_e32 v31, 16, v14
; SI-NEXT: v_cvt_f32_f16_e32 v62, v31
; SI-NEXT: v_lshrrev_b32_e32 v31, 16, v13
-; SI-NEXT: v_cvt_f32_f16_e32 v23, v23
-; SI-NEXT: v_cvt_f32_f16_e32 v22, v22
-; SI-NEXT: v_cvt_f32_f16_e32 v21, v21
; SI-NEXT: v_cvt_f32_f16_e32 v20, v20
; SI-NEXT: v_cvt_f32_f16_e32 v19, v19
; SI-NEXT: v_cvt_f32_f16_e32 v18, v18
@@ -20492,43 +20509,42 @@ define <60 x half> @bitcast_v30f32_to_v60f16(<30 x float> %a, i32 %b) {
; SI-NEXT: v_cvt_f32_f16_e32 v8, v8
; SI-NEXT: v_cvt_f32_f16_e32 v7, v7
; SI-NEXT: v_cvt_f32_f16_e32 v6, v6
-; SI-NEXT: v_cvt_f32_f16_e32 v5, v5
-; SI-NEXT: v_cvt_f32_f16_e32 v4, v4
-; SI-NEXT: v_cvt_f32_f16_e32 v43, v32
-; SI-NEXT: v_lshrrev_b32_e32 v32, 16, v3
-; SI-NEXT: v_cvt_f32_f16_e32 v45, v32
-; SI-NEXT: v_lshrrev_b32_e32 v32, 16, v2
-; SI-NEXT: v_cvt_f32_f16_e32 v47, v32
-; SI-NEXT: v_lshrrev_b32_e32 v32, 16, v1
+; SI-NEXT: v_cvt_f32_f16_e32 v43, v35
+; SI-NEXT: v_lshrrev_b32_e32 v35, 16, v3
+; SI-NEXT: v_cvt_f32_f16_e32 v45, v35
+; SI-NEXT: v_lshrrev_b32_e32 v35, 16, v2
+; SI-NEXT: v_cvt_f32_f16_e32 v47, v35
+; SI-NEXT: v_lshrrev_b32_e32 v35, 16, v1
; SI-NEXT: v_cvt_f32_f16_e32 v31, v31
-; SI-NEXT: v_cvt_f32_f16_e32 v57, v32
-; SI-NEXT: v_cvt_f32_f16_e32 v32, v30
-; SI-NEXT: v_cvt_f32_f16_e32 v34, v29
-; SI-NEXT: v_cvt_f32_f16_e32 v35, v28
-; SI-NEXT: v_cvt_f32_f16_e32 v37, v27
-; SI-NEXT: v_cvt_f32_f16_e32 v59, v26
-; SI-NEXT: v_cvt_f32_f16_e32 v63, v25
-; SI-NEXT: buffer_store_dword v24, off, s[0:3], s32 offset:72 ; 4-byte Folded Spill
-; SI-NEXT: buffer_store_dword v23, off, s[0:3], s32 offset:80 ; 4-byte Folded Spill
-; SI-NEXT: buffer_store_dword v22, off, s[0:3], s32 offset:88 ; 4-byte Folded Spill
-; SI-NEXT: buffer_store_dword v21, off, s[0:3], s32 offset:92 ; 4-byte Folded Spill
-; SI-NEXT: buffer_store_dword v20, off, s[0:3], s32 offset:100 ; 4-byte Folded Spill
-; SI-NEXT: buffer_store_dword v19, off, s[0:3], s32 offset:108 ; 4-byte Folded Spill
-; SI-NEXT: buffer_store_dword v18, off, s[0:3], s32 offset:116 ; 4-byte Folded Spill
-; SI-NEXT: buffer_store_dword v17, off, s[0:3], s32 offset:120 ; 4-byte Folded Spill
-; SI-NEXT: buffer_store_dword v16, off, s[0:3], s32 offset:124 ; 4-byte Folded Spill
-; SI-NEXT: buffer_store_dword v15, off, s[0:3], s32 offset:128 ; 4-byte Folded Spill
-; SI-NEXT: buffer_store_dword v14, off, s[0:3], s32 offset:132 ; 4-byte Folded Spill
-; SI-NEXT: buffer_store_dword v13, off, s[0:3], s32 offset:136 ; 4-byte Folded Spill
-; SI-NEXT: buffer_store_dword v12, off, s[0:3], s32 offset:140 ; 4-byte Folded Spill
-; SI-NEXT: buffer_store_dword v11, off, s[0:3], s32 offset:144 ; 4-byte Folded Spill
-; SI-NEXT: buffer_store_dword v10, off, s[0:3], s32 offset:148 ; 4-byte Folded Spill
-; SI-NEXT: buffer_store_dword v9, off, s[0:3], s32 offset:152 ; 4-byte Folded Spill
-; SI-NEXT: buffer_store_dword v8, off, s[0:3], s32 offset:156 ; 4-byte Folded Spill
-; SI-NEXT: buffer_store_dword v7, off, s[0:3], s32 offset:160 ; 4-byte Folded Spill
-; SI-NEXT: buffer_store_dword v6, off, s[0:3], s32 offset:164 ; 4-byte Folded Spill
-; SI-NEXT: buffer_store_dword v5, off, s[0:3], s32 offset:168 ; 4-byte Folded Spill
-; SI-NEXT: buffer_store_dword v4, off, s[0:3], s32 offset:172 ; 4-byte Folded Spill
+; SI-NEXT: v_cvt_f32_f16_e32 v32, v32
+; SI-NEXT: v_cvt_f32_f16_e32 v57, v35
+; SI-NEXT: v_cvt_f32_f16_e32 v35, v30
+; SI-NEXT: v_cvt_f32_f16_e32 v37, v29
+; SI-NEXT: v_cvt_f32_f16_e32 v39, v28
+; SI-NEXT: v_cvt_f32_f16_e32 v59, v27
+; SI-NEXT: v_cvt_f32_f16_e32 v63, v26
+; SI-NEXT: buffer_store_dword v25, off, s[0:3], s32 offset:72 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v24, off, s[0:3], s32 offset:80 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v23, off, s[0:3], s32 offset:88 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v22, off, s[0:3], s32 offset:96 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v21, off, s[0:3], s32 offset:100 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v20, off, s[0:3], s32 offset:108 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v19, off, s[0:3], s32 offset:116 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v18, off, s[0:3], s32 offset:124 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v17, off, s[0:3], s32 offset:128 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v16, off, s[0:3], s32 offset:132 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v15, off, s[0:3], s32 offset:136 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v14, off, s[0:3], s32 offset:140 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v13, off, s[0:3], s32 offset:144 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v12, off, s[0:3], s32 offset:148 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v11, off, s[0:3], s32 offset:152 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v10, off, s[0:3], s32 offset:156 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v9, off, s[0:3], s32 offset:160 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v8, off, s[0:3], s32 offset:164 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v7, off, s[0:3], s32 offset:168 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v6, off, s[0:3], s32 offset:172 ; 4-byte Folded Spill
+; SI-NEXT: v_cvt_f32_f16_e32 v52, v5
+; SI-NEXT: v_cvt_f32_f16_e32 v42, v4
; SI-NEXT: v_cvt_f32_f16_e32 v44, v3
; SI-NEXT: v_cvt_f32_f16_e32 v46, v2
; SI-NEXT: v_cvt_f32_f16_e32 v56, v1
@@ -20566,37 +20582,61 @@ define <60 x half> @bitcast_v30f32_to_v60f16(<30 x float> %a, i32 %b) {
; SI-NEXT: s_andn2_saveexec_b64 s[4:5], s[4:5]
; SI-NEXT: s_cbranch_execz .LBB32_4
; SI-NEXT: ; %bb.3: ; %cmp.true
-; SI-NEXT: v_add_f32_e32 v22, 1.0, v22
+; SI-NEXT: v_add_f32_e32 v1, 1.0, v1
+; SI-NEXT: v_add_f32_e32 v23, 1.0, v23
+; SI-NEXT: v_lshrrev_b32_e32 v31, 16, v1
; SI-NEXT: v_add_f32_e32 v2, 1.0, v2
-; SI-NEXT: v_lshrrev_b32_e32 v35, 16, v22
-; SI-NEXT: v_lshrrev_b32_e32 v47, 16, v2
+; SI-NEXT: v_lshrrev_b32_e32 v37, 16, v23
+; SI-NEXT: buffer_store_dword v31, off, s[0:3], s32 offset:176 ; 4-byte Folded Spill
+; SI-NEXT: s_waitcnt expcnt(0)
+; SI-NEXT: v_lshrrev_b32_e32 v31, 16, v2
; SI-NEXT: v_cvt_f32_f16_e32 v46, v2
-; SI-NEXT: v_cvt_f32_f16_e32 v2, v35
+; SI-NEXT: v_cvt_f32_f16_e32 v2, v37
+; SI-NEXT: v_add_f32_e32 v22, 1.0, v22
+; SI-NEXT: v_lshrrev_b32_e32 v35, 16, v22
; SI-NEXT: v_add_f32_e32 v21, 1.0, v21
+; SI-NEXT: buffer_store_dword v2, off, s[0:3], s32 offset:68 ; 4-byte Folded Spill
+; SI-NEXT: s_waitcnt expcnt(0)
+; SI-NEXT: v_cvt_f32_f16_e32 v2, v35
; SI-NEXT: v_lshrrev_b32_e32 v34, 16, v21
; SI-NEXT: v_add_f32_e32 v20, 1.0, v20
-; SI-NEXT: buffer_store_dword v2, off, s[0:3], s32 offset:68 ; 4-byte Folded Spill
+; SI-NEXT: v_lshrrev_b32_e32 v32, 16, v20
+; SI-NEXT: buffer_store_dword v2, off, s[0:3], s32 offset:76 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_cvt_f32_f16_e32 v2, v34
-; SI-NEXT: v_lshrrev_b32_e32 v32, 16, v20
; SI-NEXT: v_add_f32_e32 v19, 1.0, v19
-; SI-NEXT: v_add_f32_e32 v23, 1.0, v23
-; SI-NEXT: buffer_store_dword v2, off, s[0:3], s32 offset:76 ; 4-byte Folded Spill
+; SI-NEXT: v_add_f32_e32 v24, 1.0, v24
+; SI-NEXT: v_lshrrev_b32_e32 v63, 16, v19
+; SI-NEXT: buffer_store_dword v2, off, s[0:3], s32 offset:84 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_cvt_f32_f16_e32 v2, v32
-; SI-NEXT: v_add_f32_e32 v1, 1.0, v1
-; SI-NEXT: v_lshrrev_b32_e32 v63, 16, v19
-; SI-NEXT: v_lshrrev_b32_e32 v37, 16, v23
-; SI-NEXT: v_lshrrev_b32_e32 v57, 16, v1
+; SI-NEXT: v_lshrrev_b32_e32 v39, 16, v24
; SI-NEXT: v_cvt_f32_f16_e32 v56, v1
-; SI-NEXT: v_cvt_f32_f16_e32 v1, v37
-; SI-NEXT: buffer_store_dword v2, off, s[0:3], s32 offset:84 ; 4-byte Folded Spill
+; SI-NEXT: v_cvt_f32_f16_e32 v1, v39
+; SI-NEXT: buffer_store_dword v2, off, s[0:3], s32 offset:92 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_cvt_f32_f16_e32 v2, v63
-; SI-NEXT: v_add_f32_e32 v3, 1.0, v3
+; SI-NEXT: v_add_f32_e32 v17, 1.0, v17
+; SI-NEXT: v_add_f32_e32 v18, 1.0, v18
+; SI-NEXT: v_lshrrev_b32_e32 v59, 16, v17
+; SI-NEXT: v_lshrrev_b32_e32 v61, 16, v18
+; SI-NEXT: buffer_store_dword v2, off, s[0:3], s32 offset:104 ; 4-byte Folded Spill
+; SI-NEXT: s_waitcnt expcnt(0)
+; SI-NEXT: v_cvt_f32_f16_e32 v2, v61
+; SI-NEXT: v_mov_b32_e32 v61, v1
+; SI-NEXT: v_cvt_f32_f16_e32 v1, v59
+; SI-NEXT: buffer_store_dword v31, off, s[0:3], s32 offset:180 ; 4-byte Folded Spill
; SI-NEXT: v_add_f32_e32 v4, 1.0, v4
-; SI-NEXT: v_add_f32_e32 v5, 1.0, v5
; SI-NEXT: v_add_f32_e32 v6, 1.0, v6
+; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:120 ; 4-byte Folded Spill
+; SI-NEXT: s_waitcnt expcnt(0)
+; SI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:180 ; 4-byte Folded Reload
+; SI-NEXT: v_lshrrev_b32_e32 v47, 16, v4
+; SI-NEXT: v_lshrrev_b32_e32 v43, 16, v6
+; SI-NEXT: v_cvt_f32_f16_e32 v32, v43
+; SI-NEXT: v_cvt_f32_f16_e32 v43, v47
+; SI-NEXT: v_add_f32_e32 v3, 1.0, v3
+; SI-NEXT: v_add_f32_e32 v5, 1.0, v5
; SI-NEXT: v_add_f32_e32 v7, 1.0, v7
; SI-NEXT: v_add_f32_e32 v8, 1.0, v8
; SI-NEXT: v_add_f32_e32 v9, 1.0, v9
@@ -20607,20 +20647,15 @@ define <60 x half> @bitcast_v30f32_to_v60f16(<30 x float> %a, i32 %b) {
; SI-NEXT: v_add_f32_e32 v14, 1.0, v14
; SI-NEXT: v_add_f32_e32 v15, 1.0, v15
; SI-NEXT: v_add_f32_e32 v16, 1.0, v16
-; SI-NEXT: v_add_f32_e32 v17, 1.0, v17
-; SI-NEXT: v_add_f32_e32 v18, 1.0, v18
-; SI-NEXT: v_add_f32_e32 v24, 1.0, v24
; SI-NEXT: v_add_f32_e32 v25, 1.0, v25
; SI-NEXT: v_add_f32_e32 v26, 1.0, v26
; SI-NEXT: v_add_f32_e32 v27, 1.0, v27
; SI-NEXT: v_add_f32_e32 v28, 1.0, v28
; SI-NEXT: v_add_f32_e32 v29, 1.0, v29
; SI-NEXT: v_add_f32_e32 v30, 1.0, v30
-; SI-NEXT: v_lshrrev_b32_e32 v45, 16, v3
-; SI-NEXT: v_lshrrev_b32_e32 v43, 16, v4
-; SI-NEXT: v_lshrrev_b32_e32 v40, 16, v5
-; SI-NEXT: v_lshrrev_b32_e32 v54, 16, v6
-; SI-NEXT: v_lshrrev_b32_e32 v52, 16, v7
+; SI-NEXT: v_lshrrev_b32_e32 v57, 16, v3
+; SI-NEXT: v_lshrrev_b32_e32 v45, 16, v5
+; SI-NEXT: v_lshrrev_b32_e32 v40, 16, v7
; SI-NEXT: v_lshrrev_b32_e32 v50, 16, v8
; SI-NEXT: v_lshrrev_b32_e32 v48, 16, v9
; SI-NEXT: v_lshrrev_b32_e32 v38, 16, v10
@@ -20630,15 +20665,12 @@ define <60 x half> @bitcast_v30f32_to_v60f16(<30 x float> %a, i32 %b) {
; SI-NEXT: v_lshrrev_b32_e32 v62, 16, v14
; SI-NEXT: v_lshrrev_b32_e32 v60, 16, v15
; SI-NEXT: v_lshrrev_b32_e32 v58, 16, v16
-; SI-NEXT: v_lshrrev_b32_e32 v59, 16, v17
-; SI-NEXT: v_lshrrev_b32_e32 v61, 16, v18
-; SI-NEXT: v_lshrrev_b32_e32 v39, 16, v24
; SI-NEXT: v_lshrrev_b32_e32 v49, 16, v25
; SI-NEXT: v_lshrrev_b32_e32 v51, 16, v26
; SI-NEXT: v_lshrrev_b32_e32 v53, 16, v27
; SI-NEXT: v_lshrrev_b32_e32 v55, 16, v28
-; SI-NEXT: v_lshrrev_b32_e32 v41, 16, v29
-; SI-NEXT: v_lshrrev_b32_e32 v42, 16, v30
+; SI-NEXT: v_lshrrev_b32_e32 v54, 16, v29
+; SI-NEXT: v_lshrrev_b32_e32 v41, 16, v30
; SI-NEXT: v_cvt_f32_f16_e32 v30, v30
; SI-NEXT: v_cvt_f32_f16_e32 v29, v29
; SI-NEXT: v_cvt_f32_f16_e32 v28, v28
@@ -20664,70 +20696,62 @@ define <60 x half> @bitcast_v30f32_to_v60f16(<30 x float> %a, i32 %b) {
; SI-NEXT: v_cvt_f32_f16_e32 v8, v8
; SI-NEXT: v_cvt_f32_f16_e32 v7, v7
; SI-NEXT: v_cvt_f32_f16_e32 v6, v6
-; SI-NEXT: v_cvt_f32_f16_e32 v5, v5
-; SI-NEXT: v_cvt_f32_f16_e32 v4, v4
+; SI-NEXT: v_cvt_f32_f16_e32 v52, v5
+; SI-NEXT: v_cvt_f32_f16_e32 v42, v4
; SI-NEXT: v_cvt_f32_f16_e32 v44, v3
-; SI-NEXT: v_cvt_f32_f16_e32 v42, v42
; SI-NEXT: v_cvt_f32_f16_e32 v41, v41
+; SI-NEXT: v_cvt_f32_f16_e32 v54, v54
; SI-NEXT: v_cvt_f32_f16_e32 v55, v55
; SI-NEXT: v_cvt_f32_f16_e32 v53, v53
; SI-NEXT: v_cvt_f32_f16_e32 v51, v51
; SI-NEXT: v_cvt_f32_f16_e32 v49, v49
-; SI-NEXT: v_cvt_f32_f16_e32 v39, v39
-; SI-NEXT: buffer_store_dword v2, off, s[0:3], s32 offset:96 ; 4-byte Folded Spill
-; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_cvt_f32_f16_e32 v2, v61
-; SI-NEXT: v_mov_b32_e32 v61, v1
-; SI-NEXT: v_cvt_f32_f16_e32 v1, v59
; SI-NEXT: v_cvt_f32_f16_e32 v58, v58
; SI-NEXT: v_cvt_f32_f16_e32 v60, v60
; SI-NEXT: v_cvt_f32_f16_e32 v62, v62
; SI-NEXT: v_cvt_f32_f16_e32 v31, v31
; SI-NEXT: v_cvt_f32_f16_e32 v33, v33
+; SI-NEXT: s_waitcnt vmcnt(0)
+; SI-NEXT: v_cvt_f32_f16_e32 v47, v1
+; SI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:176 ; 4-byte Folded Reload
; SI-NEXT: v_cvt_f32_f16_e32 v36, v36
; SI-NEXT: v_cvt_f32_f16_e32 v38, v38
; SI-NEXT: v_cvt_f32_f16_e32 v48, v48
; SI-NEXT: v_cvt_f32_f16_e32 v50, v50
-; SI-NEXT: v_cvt_f32_f16_e32 v52, v52
-; SI-NEXT: v_cvt_f32_f16_e32 v54, v54
-; SI-NEXT: v_cvt_f32_f16_e32 v40, v40
-; SI-NEXT: v_cvt_f32_f16_e32 v43, v43
-; SI-NEXT: v_cvt_f32_f16_e32 v45, v45
-; SI-NEXT: v_cvt_f32_f16_e32 v47, v47
-; SI-NEXT: v_cvt_f32_f16_e32 v57, v57
-; SI-NEXT: v_mov_b32_e32 v37, v27
-; SI-NEXT: v_mov_b32_e32 v35, v28
-; SI-NEXT: v_mov_b32_e32 v34, v29
-; SI-NEXT: v_mov_b32_e32 v32, v30
-; SI-NEXT: v_mov_b32_e32 v63, v25
-; SI-NEXT: v_mov_b32_e32 v59, v26
-; SI-NEXT: buffer_store_dword v24, off, s[0:3], s32 offset:72 ; 4-byte Folded Spill
-; SI-NEXT: buffer_store_dword v23, off, s[0:3], s32 offset:80 ; 4-byte Folded Spill
-; SI-NEXT: buffer_store_dword v22, off, s[0:3], s32 offset:88 ; 4-byte Folded Spill
-; SI-NEXT: buffer_store_dword v21, off, s[0:3], s32 offset:92 ; 4-byte Folded Spill
-; SI-NEXT: buffer_store_dword v20, off, s[0:3], s32 offset:100 ; 4-byte Folded Spill
-; SI-NEXT: buffer_store_dword v19, off, s[0:3], s32 offset:108 ; 4-byte Folded Spill
-; SI-NEXT: buffer_store_dword v18, off, s[0:3], s32 offset:116 ; 4-byte Folded Spill
-; SI-NEXT: buffer_store_dword v17, off, s[0:3], s32 offset:120 ; 4-byte Folded Spill
-; SI-NEXT: buffer_store_dword v16, off, s[0:3], s32 offset:124 ; 4-byte Folded Spill
-; SI-NEXT: buffer_store_dword v15, off, s[0:3], s32 offset:128 ; 4-byte Folded Spill
-; SI-NEXT: buffer_store_dword v14, off, s[0:3], s32 offset:132 ; 4-byte Folded Spill
-; SI-NEXT: buffer_store_dword v13, off, s[0:3], s32 offset:136 ; 4-byte Folded Spill
-; SI-NEXT: buffer_store_dword v12, off, s[0:3], s32 offset:140 ; 4-byte Folded Spill
-; SI-NEXT: buffer_store_dword v11, off, s[0:3], s32 offset:144 ; 4-byte Folded Spill
-; SI-NEXT: buffer_store_dword v10, off, s[0:3], s32 offset:148 ; 4-byte Folded Spill
-; SI-NEXT: buffer_store_dword v9, off, s[0:3], s32 offset:152 ; 4-byte Folded Spill
-; SI-NEXT: buffer_store_dword v8, off, s[0:3], s32 offset:156 ; 4-byte Folded Spill
-; SI-NEXT: buffer_store_dword v7, off, s[0:3], s32 offset:160 ; 4-byte Folded Spill
-; SI-NEXT: buffer_store_dword v6, off, s[0:3], s32 offset:164 ; 4-byte Folded Spill
-; SI-NEXT: buffer_store_dword v5, off, s[0:3], s32 offset:168 ; 4-byte Folded Spill
-; SI-NEXT: buffer_store_dword v4, off, s[0:3], s32 offset:172 ; 4-byte Folded Spill
-; SI-NEXT: buffer_store_dword v2, off, s[0:3], s32 offset:104 ; 4-byte Folded Spill
-; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:112 ; 4-byte Folded Spill
+; SI-NEXT: v_cvt_f32_f16_e32 v34, v40
+; SI-NEXT: v_cvt_f32_f16_e32 v40, v45
+; SI-NEXT: v_cvt_f32_f16_e32 v45, v57
+; SI-NEXT: v_mov_b32_e32 v39, v28
+; SI-NEXT: v_mov_b32_e32 v37, v29
+; SI-NEXT: v_mov_b32_e32 v35, v30
+; SI-NEXT: v_mov_b32_e32 v63, v26
+; SI-NEXT: v_mov_b32_e32 v59, v27
+; SI-NEXT: buffer_store_dword v25, off, s[0:3], s32 offset:72 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v24, off, s[0:3], s32 offset:80 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v23, off, s[0:3], s32 offset:88 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v22, off, s[0:3], s32 offset:96 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v21, off, s[0:3], s32 offset:100 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v20, off, s[0:3], s32 offset:108 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v19, off, s[0:3], s32 offset:116 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v18, off, s[0:3], s32 offset:124 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v17, off, s[0:3], s32 offset:128 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v16, off, s[0:3], s32 offset:132 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v15, off, s[0:3], s32 offset:136 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v14, off, s[0:3], s32 offset:140 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v13, off, s[0:3], s32 offset:144 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v12, off, s[0:3], s32 offset:148 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v11, off, s[0:3], s32 offset:152 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v10, off, s[0:3], s32 offset:156 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v9, off, s[0:3], s32 offset:160 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v8, off, s[0:3], s32 offset:164 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v7, off, s[0:3], s32 offset:168 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v6, off, s[0:3], s32 offset:172 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v2, off, s[0:3], s32 offset:112 ; 4-byte Folded Spill
+; SI-NEXT: s_waitcnt vmcnt(14)
+; SI-NEXT: v_cvt_f32_f16_e32 v57, v1
; SI-NEXT: .LBB32_4: ; %end
; SI-NEXT: s_or_b64 exec, exec, s[4:5]
-; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_cvt_f16_f32_e32 v1, v57
+; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_cvt_f16_f32_e32 v2, v56
; SI-NEXT: v_cvt_f16_f32_e32 v3, v46
; SI-NEXT: v_lshlrev_b32_e32 v1, 16, v1
@@ -20746,43 +20770,39 @@ define <60 x half> @bitcast_v30f32_to_v60f16(<30 x float> %a, i32 %b) {
; SI-NEXT: v_lshlrev_b32_e32 v1, 16, v1
; SI-NEXT: v_or_b32_e32 v1, v2, v1
; SI-NEXT: buffer_store_dword v1, v3, s[0:3], 0 offen
-; SI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:172 ; 4-byte Folded Reload
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_cvt_f16_f32_e32 v1, v43
+; SI-NEXT: v_cvt_f16_f32_e32 v2, v42
; SI-NEXT: v_add_i32_e32 v3, vcc, 12, v0
; SI-NEXT: v_lshlrev_b32_e32 v1, 16, v1
-; SI-NEXT: s_waitcnt vmcnt(0)
-; SI-NEXT: v_cvt_f16_f32_e32 v2, v2
; SI-NEXT: v_or_b32_e32 v1, v2, v1
; SI-NEXT: buffer_store_dword v1, v3, s[0:3], 0 offen
-; SI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:168 ; 4-byte Folded Reload
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_cvt_f16_f32_e32 v1, v40
+; SI-NEXT: v_cvt_f16_f32_e32 v2, v52
; SI-NEXT: v_add_i32_e32 v3, vcc, 16, v0
; SI-NEXT: v_lshlrev_b32_e32 v1, 16, v1
-; SI-NEXT: s_waitcnt vmcnt(0)
-; SI-NEXT: v_cvt_f16_f32_e32 v2, v2
; SI-NEXT: v_or_b32_e32 v1, v2, v1
; SI-NEXT: buffer_store_dword v1, v3, s[0:3], 0 offen
-; SI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:164 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:172 ; 4-byte Folded Reload
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_cvt_f16_f32_e32 v1, v54
+; SI-NEXT: v_cvt_f16_f32_e32 v1, v32
; SI-NEXT: v_add_i32_e32 v3, vcc, 20, v0
; SI-NEXT: v_lshlrev_b32_e32 v1, 16, v1
; SI-NEXT: s_waitcnt vmcnt(0)
; SI-NEXT: v_cvt_f16_f32_e32 v2, v2
; SI-NEXT: v_or_b32_e32 v1, v2, v1
; SI-NEXT: buffer_store_dword v1, v3, s[0:3], 0 offen
-; SI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:160 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:168 ; 4-byte Folded Reload
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_cvt_f16_f32_e32 v1, v52
+; SI-NEXT: v_cvt_f16_f32_e32 v1, v34
; SI-NEXT: v_add_i32_e32 v3, vcc, 24, v0
; SI-NEXT: v_lshlrev_b32_e32 v1, 16, v1
; SI-NEXT: s_waitcnt vmcnt(0)
; SI-NEXT: v_cvt_f16_f32_e32 v2, v2
; SI-NEXT: v_or_b32_e32 v1, v2, v1
; SI-NEXT: buffer_store_dword v1, v3, s[0:3], 0 offen
-; SI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:156 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:164 ; 4-byte Folded Reload
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_cvt_f16_f32_e32 v1, v50
; SI-NEXT: v_add_i32_e32 v3, vcc, 28, v0
@@ -20791,7 +20811,7 @@ define <60 x half> @bitcast_v30f32_to_v60f16(<30 x float> %a, i32 %b) {
; SI-NEXT: v_cvt_f16_f32_e32 v2, v2
; SI-NEXT: v_or_b32_e32 v1, v2, v1
; SI-NEXT: buffer_store_dword v1, v3, s[0:3], 0 offen
-; SI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:152 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:160 ; 4-byte Folded Reload
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_cvt_f16_f32_e32 v1, v48
; SI-NEXT: v_add_i32_e32 v3, vcc, 32, v0
@@ -20800,7 +20820,7 @@ define <60 x half> @bitcast_v30f32_to_v60f16(<30 x float> %a, i32 %b) {
; SI-NEXT: v_cvt_f16_f32_e32 v2, v2
; SI-NEXT: v_or_b32_e32 v1, v2, v1
; SI-NEXT: buffer_store_dword v1, v3, s[0:3], 0 offen
-; SI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:148 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:156 ; 4-byte Folded Reload
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_cvt_f16_f32_e32 v1, v38
; SI-NEXT: v_add_i32_e32 v3, vcc, 36, v0
@@ -20809,7 +20829,7 @@ define <60 x half> @bitcast_v30f32_to_v60f16(<30 x float> %a, i32 %b) {
; SI-NEXT: v_cvt_f16_f32_e32 v2, v2
; SI-NEXT: v_or_b32_e32 v1, v2, v1
; SI-NEXT: buffer_store_dword v1, v3, s[0:3], 0 offen
-; SI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:144 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:152 ; 4-byte Folded Reload
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_cvt_f16_f32_e32 v1, v36
; SI-NEXT: v_add_i32_e32 v3, vcc, 40, v0
@@ -20818,7 +20838,7 @@ define <60 x half> @bitcast_v30f32_to_v60f16(<30 x float> %a, i32 %b) {
; SI-NEXT: v_cvt_f16_f32_e32 v2, v2
; SI-NEXT: v_or_b32_e32 v1, v2, v1
; SI-NEXT: buffer_store_dword v1, v3, s[0:3], 0 offen
-; SI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:140 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:148 ; 4-byte Folded Reload
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_cvt_f16_f32_e32 v1, v33
; SI-NEXT: v_add_i32_e32 v3, vcc, 44, v0
@@ -20827,7 +20847,7 @@ define <60 x half> @bitcast_v30f32_to_v60f16(<30 x float> %a, i32 %b) {
; SI-NEXT: v_cvt_f16_f32_e32 v2, v2
; SI-NEXT: v_or_b32_e32 v1, v2, v1
; SI-NEXT: buffer_store_dword v1, v3, s[0:3], 0 offen
-; SI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:136 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:144 ; 4-byte Folded Reload
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_cvt_f16_f32_e32 v1, v31
; SI-NEXT: v_add_i32_e32 v3, vcc, 48, v0
@@ -20836,7 +20856,7 @@ define <60 x half> @bitcast_v30f32_to_v60f16(<30 x float> %a, i32 %b) {
; SI-NEXT: v_cvt_f16_f32_e32 v2, v2
; SI-NEXT: v_or_b32_e32 v1, v2, v1
; SI-NEXT: buffer_store_dword v1, v3, s[0:3], 0 offen
-; SI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:132 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:140 ; 4-byte Folded Reload
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_cvt_f16_f32_e32 v1, v62
; SI-NEXT: v_add_i32_e32 v3, vcc, 52, v0
@@ -20845,7 +20865,7 @@ define <60 x half> @bitcast_v30f32_to_v60f16(<30 x float> %a, i32 %b) {
; SI-NEXT: v_cvt_f16_f32_e32 v2, v2
; SI-NEXT: v_or_b32_e32 v1, v2, v1
; SI-NEXT: buffer_store_dword v1, v3, s[0:3], 0 offen
-; SI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:128 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:136 ; 4-byte Folded Reload
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_cvt_f16_f32_e32 v1, v60
; SI-NEXT: v_add_i32_e32 v3, vcc, 56, v0
@@ -20854,7 +20874,7 @@ define <60 x half> @bitcast_v30f32_to_v60f16(<30 x float> %a, i32 %b) {
; SI-NEXT: v_cvt_f16_f32_e32 v2, v2
; SI-NEXT: v_or_b32_e32 v1, v2, v1
; SI-NEXT: buffer_store_dword v1, v3, s[0:3], 0 offen
-; SI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:124 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:132 ; 4-byte Folded Reload
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_cvt_f16_f32_e32 v1, v58
; SI-NEXT: v_add_i32_e32 v3, vcc, 60, v0
@@ -20864,8 +20884,8 @@ define <60 x half> @bitcast_v30f32_to_v60f16(<30 x float> %a, i32 %b) {
; SI-NEXT: v_or_b32_e32 v1, v2, v1
; SI-NEXT: buffer_store_dword v1, v3, s[0:3], 0 offen
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:112 ; 4-byte Folded Reload
-; SI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:120 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:120 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:128 ; 4-byte Folded Reload
; SI-NEXT: v_add_i32_e32 v3, vcc, 64, v0
; SI-NEXT: s_waitcnt vmcnt(1)
; SI-NEXT: v_cvt_f16_f32_e32 v1, v1
@@ -20875,8 +20895,8 @@ define <60 x half> @bitcast_v30f32_to_v60f16(<30 x float> %a, i32 %b) {
; SI-NEXT: v_or_b32_e32 v1, v2, v1
; SI-NEXT: buffer_store_dword v1, v3, s[0:3], 0 offen
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:104 ; 4-byte Folded Reload
-; SI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:116 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:112 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:124 ; 4-byte Folded Reload
; SI-NEXT: v_add_i32_e32 v3, vcc, 0x44, v0
; SI-NEXT: s_waitcnt vmcnt(1)
; SI-NEXT: v_cvt_f16_f32_e32 v1, v1
@@ -20886,8 +20906,8 @@ define <60 x half> @bitcast_v30f32_to_v60f16(<30 x float> %a, i32 %b) {
; SI-NEXT: v_or_b32_e32 v1, v2, v1
; SI-NEXT: buffer_store_dword v1, v3, s[0:3], 0 offen
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:96 ; 4-byte Folded Reload
-; SI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:108 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:104 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:116 ; 4-byte Folded Reload
; SI-NEXT: v_add_i32_e32 v3, vcc, 0x48, v0
; SI-NEXT: s_waitcnt vmcnt(1)
; SI-NEXT: v_cvt_f16_f32_e32 v1, v1
@@ -20897,8 +20917,8 @@ define <60 x half> @bitcast_v30f32_to_v60f16(<30 x float> %a, i32 %b) {
; SI-NEXT: v_or_b32_e32 v1, v2, v1
; SI-NEXT: buffer_store_dword v1, v3, s[0:3], 0 offen
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:84 ; 4-byte Folded Reload
-; SI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:100 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:92 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:108 ; 4-byte Folded Reload
; SI-NEXT: v_add_i32_e32 v3, vcc, 0x4c, v0
; SI-NEXT: s_waitcnt vmcnt(1)
; SI-NEXT: v_cvt_f16_f32_e32 v1, v1
@@ -20908,8 +20928,8 @@ define <60 x half> @bitcast_v30f32_to_v60f16(<30 x float> %a, i32 %b) {
; SI-NEXT: v_or_b32_e32 v1, v2, v1
; SI-NEXT: buffer_store_dword v1, v3, s[0:3], 0 offen
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:76 ; 4-byte Folded Reload
-; SI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:92 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:84 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:100 ; 4-byte Folded Reload
; SI-NEXT: v_add_i32_e32 v3, vcc, 0x50, v0
; SI-NEXT: s_waitcnt vmcnt(1)
; SI-NEXT: v_cvt_f16_f32_e32 v1, v1
@@ -20919,8 +20939,8 @@ define <60 x half> @bitcast_v30f32_to_v60f16(<30 x float> %a, i32 %b) {
; SI-NEXT: v_or_b32_e32 v1, v2, v1
; SI-NEXT: buffer_store_dword v1, v3, s[0:3], 0 offen
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:68 ; 4-byte Folded Reload
-; SI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:88 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:76 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:96 ; 4-byte Folded Reload
; SI-NEXT: v_add_i32_e32 v3, vcc, 0x54, v0
; SI-NEXT: s_waitcnt vmcnt(1)
; SI-NEXT: v_cvt_f16_f32_e32 v1, v1
@@ -20929,62 +20949,66 @@ define <60 x half> @bitcast_v30f32_to_v60f16(<30 x float> %a, i32 %b) {
; SI-NEXT: v_lshlrev_b32_e32 v1, 16, v1
; SI-NEXT: v_or_b32_e32 v1, v2, v1
; SI-NEXT: buffer_store_dword v1, v3, s[0:3], 0 offen
-; SI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:80 ; 4-byte Folded Reload
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_cvt_f16_f32_e32 v1, v61
+; SI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:68 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:88 ; 4-byte Folded Reload
; SI-NEXT: v_add_i32_e32 v3, vcc, 0x58, v0
-; SI-NEXT: v_lshlrev_b32_e32 v1, 16, v1
+; SI-NEXT: s_waitcnt vmcnt(1)
+; SI-NEXT: v_cvt_f16_f32_e32 v1, v1
; SI-NEXT: s_waitcnt vmcnt(0)
; SI-NEXT: v_cvt_f16_f32_e32 v2, v2
+; SI-NEXT: v_lshlrev_b32_e32 v1, 16, v1
; SI-NEXT: v_or_b32_e32 v1, v2, v1
; SI-NEXT: buffer_store_dword v1, v3, s[0:3], 0 offen
-; SI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:72 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:80 ; 4-byte Folded Reload
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_cvt_f16_f32_e32 v1, v39
+; SI-NEXT: v_cvt_f16_f32_e32 v1, v61
; SI-NEXT: v_add_i32_e32 v3, vcc, 0x5c, v0
; SI-NEXT: v_lshlrev_b32_e32 v1, 16, v1
; SI-NEXT: s_waitcnt vmcnt(0)
; SI-NEXT: v_cvt_f16_f32_e32 v2, v2
; SI-NEXT: v_or_b32_e32 v1, v2, v1
; SI-NEXT: buffer_store_dword v1, v3, s[0:3], 0 offen
+; SI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:72 ; 4-byte Folded Reload
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_cvt_f16_f32_e32 v1, v49
-; SI-NEXT: v_cvt_f16_f32_e32 v2, v63
; SI-NEXT: v_add_i32_e32 v3, vcc, 0x60, v0
; SI-NEXT: v_lshlrev_b32_e32 v1, 16, v1
+; SI-NEXT: s_waitcnt vmcnt(0)
+; SI-NEXT: v_cvt_f16_f32_e32 v2, v2
; SI-NEXT: v_or_b32_e32 v1, v2, v1
; SI-NEXT: buffer_store_dword v1, v3, s[0:3], 0 offen
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_cvt_f16_f32_e32 v1, v51
-; SI-NEXT: v_cvt_f16_f32_e32 v2, v59
+; SI-NEXT: v_cvt_f16_f32_e32 v2, v63
; SI-NEXT: v_add_i32_e32 v3, vcc, 0x64, v0
; SI-NEXT: v_lshlrev_b32_e32 v1, 16, v1
; SI-NEXT: v_or_b32_e32 v1, v2, v1
; SI-NEXT: buffer_store_dword v1, v3, s[0:3], 0 offen
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_cvt_f16_f32_e32 v1, v53
-; SI-NEXT: v_cvt_f16_f32_e32 v2, v37
+; SI-NEXT: v_cvt_f16_f32_e32 v2, v59
; SI-NEXT: v_add_i32_e32 v3, vcc, 0x68, v0
; SI-NEXT: v_lshlrev_b32_e32 v1, 16, v1
; SI-NEXT: v_or_b32_e32 v1, v2, v1
; SI-NEXT: buffer_store_dword v1, v3, s[0:3], 0 offen
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_cvt_f16_f32_e32 v1, v55
-; SI-NEXT: v_cvt_f16_f32_e32 v2, v35
+; SI-NEXT: v_cvt_f16_f32_e32 v2, v39
; SI-NEXT: v_add_i32_e32 v3, vcc, 0x6c, v0
; SI-NEXT: v_lshlrev_b32_e32 v1, 16, v1
; SI-NEXT: v_or_b32_e32 v1, v2, v1
; SI-NEXT: buffer_store_dword v1, v3, s[0:3], 0 offen
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_cvt_f16_f32_e32 v1, v41
-; SI-NEXT: v_cvt_f16_f32_e32 v2, v34
+; SI-NEXT: v_cvt_f16_f32_e32 v1, v54
+; SI-NEXT: v_cvt_f16_f32_e32 v2, v37
; SI-NEXT: v_add_i32_e32 v3, vcc, 0x70, v0
; SI-NEXT: v_lshlrev_b32_e32 v1, 16, v1
; SI-NEXT: v_or_b32_e32 v1, v2, v1
; SI-NEXT: buffer_store_dword v1, v3, s[0:3], 0 offen
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_cvt_f16_f32_e32 v1, v42
-; SI-NEXT: v_cvt_f16_f32_e32 v2, v32
+; SI-NEXT: v_cvt_f16_f32_e32 v1, v41
+; SI-NEXT: v_cvt_f16_f32_e32 v2, v35
; SI-NEXT: v_add_i32_e32 v0, vcc, 0x74, v0
; SI-NEXT: v_lshlrev_b32_e32 v1, 16, v1
; SI-NEXT: v_or_b32_e32 v1, v2, v1
@@ -24752,9 +24776,9 @@ define inreg <30 x float> @bitcast_v60f16_to_v30f32_scalar(<60 x half> inreg %a,
; SI-NEXT: v_cvt_f16_f32_e32 v1, s19
; SI-NEXT: v_cvt_f16_f32_e32 v12, s18
; SI-NEXT: v_cvt_f16_f32_e32 v2, s21
-; SI-NEXT: v_cvt_f16_f32_e32 v11, s20
+; SI-NEXT: v_cvt_f16_f32_e32 v10, s20
; SI-NEXT: v_cvt_f16_f32_e32 v3, s23
-; SI-NEXT: v_cvt_f16_f32_e32 v10, s22
+; SI-NEXT: v_cvt_f16_f32_e32 v11, s22
; SI-NEXT: v_cvt_f16_f32_e32 v4, s25
; SI-NEXT: v_cvt_f16_f32_e32 v9, s24
; SI-NEXT: v_cvt_f16_f32_e32 v5, s27
@@ -24833,9 +24857,9 @@ define inreg <30 x float> @bitcast_v60f16_to_v30f32_scalar(<60 x half> inreg %a,
; SI-NEXT: ; %bb.1: ; %cmp.false
; SI-NEXT: buffer_load_dword v24, off, s[0:3], s32 offset:172 ; 4-byte Folded Reload
; SI-NEXT: buffer_load_dword v25, off, s[0:3], s32 offset:168 ; 4-byte Folded Reload
-; SI-NEXT: s_waitcnt expcnt(4)
-; SI-NEXT: v_lshlrev_b32_e32 v3, 16, v3
-; SI-NEXT: v_or_b32_e32 v3, v10, v3
+; SI-NEXT: s_waitcnt expcnt(5)
+; SI-NEXT: v_lshlrev_b32_e32 v2, 16, v2
+; SI-NEXT: v_or_b32_e32 v2, v10, v2
; SI-NEXT: s_waitcnt expcnt(3)
; SI-NEXT: v_lshlrev_b32_e32 v10, 16, v34
; SI-NEXT: v_mov_b32_e32 v33, v32
@@ -24863,13 +24887,13 @@ define inreg <30 x float> @bitcast_v60f16_to_v30f32_scalar(<60 x half> inreg %a,
; SI-NEXT: buffer_load_dword v54, off, s[0:3], s32 offset:128 ; 4-byte Folded Reload
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_lshlrev_b32_e32 v1, 16, v1
-; SI-NEXT: v_lshlrev_b32_e32 v2, 16, v2
+; SI-NEXT: v_lshlrev_b32_e32 v3, 16, v3
; SI-NEXT: v_lshlrev_b32_e32 v4, 16, v4
; SI-NEXT: v_lshlrev_b32_e32 v18, 16, v18
; SI-NEXT: v_lshlrev_b32_e32 v19, 16, v19
; SI-NEXT: v_lshlrev_b32_e32 v0, 16, v59
; SI-NEXT: v_or_b32_e32 v1, v12, v1
-; SI-NEXT: v_or_b32_e32 v2, v11, v2
+; SI-NEXT: v_or_b32_e32 v3, v11, v3
; SI-NEXT: v_or_b32_e32 v4, v9, v4
; SI-NEXT: v_lshlrev_b32_e32 v11, 16, v36
; SI-NEXT: v_lshlrev_b32_e32 v12, 16, v63
@@ -25034,7 +25058,7 @@ define inreg <30 x float> @bitcast_v60f16_to_v30f32_scalar(<60 x half> inreg %a,
; SI-NEXT: buffer_load_dword v28, off, s[0:3], s32 offset:152 ; 4-byte Folded Reload
; SI-NEXT: buffer_load_dword v29, off, s[0:3], s32 offset:144 ; 4-byte Folded Reload
; SI-NEXT: buffer_load_dword v31, off, s[0:3], s32 offset:128 ; 4-byte Folded Reload
-; SI-NEXT: buffer_load_dword v4, off, s[0:3], s32 offset:276 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v4, off, s[0:3], s32 offset:280 ; 4-byte Folded Reload
; SI-NEXT: buffer_load_dword v5, off, s[0:3], s32 offset:260 ; 4-byte Folded Reload
; SI-NEXT: buffer_load_dword v6, off, s[0:3], s32 offset:252 ; 4-byte Folded Reload
; SI-NEXT: buffer_load_dword v7, off, s[0:3], s32 offset:244 ; 4-byte Folded Reload
@@ -25056,7 +25080,7 @@ define inreg <30 x float> @bitcast_v60f16_to_v30f32_scalar(<60 x half> inreg %a,
; SI-NEXT: v_lshlrev_b32_e32 v2, 16, v2
; SI-NEXT: v_or_b32_e32 v1, v3, v2
; SI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:268 ; 4-byte Folded Reload
-; SI-NEXT: buffer_load_dword v3, off, s[0:3], s32 offset:280 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v3, off, s[0:3], s32 offset:276 ; 4-byte Folded Reload
; SI-NEXT: s_waitcnt vmcnt(14)
; SI-NEXT: v_cvt_f32_f16_e32 v27, v27
; SI-NEXT: s_waitcnt vmcnt(11)
@@ -28942,7 +28966,7 @@ define inreg <60 x i16> @bitcast_v15i64_to_v60i16_scalar(<15 x i64> inreg %a, i3
; GFX11-TRUE16-NEXT: v_readfirstlane_b32 s5, v9
; GFX11-TRUE16-NEXT: v_readfirstlane_b32 s6, v10
; GFX11-TRUE16-NEXT: v_readfirstlane_b32 s4, v11
-; GFX11-TRUE16-NEXT: s_mov_b32 s94, 0
+; GFX11-TRUE16-NEXT: s_mov_b32 s89, 0
; GFX11-TRUE16-NEXT: s_and_b32 s40, vcc_lo, exec_lo
; GFX11-TRUE16-NEXT: s_cbranch_scc0 .LBB41_4
; GFX11-TRUE16-NEXT: ; %bb.1: ; %cmp.false
@@ -28971,12 +28995,12 @@ define inreg <60 x i16> @bitcast_v15i64_to_v60i16_scalar(<15 x i64> inreg %a, i3
; GFX11-TRUE16-NEXT: s_lshr_b32 s78, s19, 16
; GFX11-TRUE16-NEXT: s_lshr_b32 s79, s18, 16
; GFX11-TRUE16-NEXT: s_lshr_b32 s88, s17, 16
-; GFX11-TRUE16-NEXT: s_lshr_b32 s89, s16, 16
-; GFX11-TRUE16-NEXT: s_lshr_b32 s90, s3, 16
-; GFX11-TRUE16-NEXT: s_lshr_b32 s91, s2, 16
-; GFX11-TRUE16-NEXT: s_lshr_b32 s92, s1, 16
-; GFX11-TRUE16-NEXT: s_lshr_b32 s93, s0, 16
-; GFX11-TRUE16-NEXT: s_and_not1_b32 vcc_lo, exec_lo, s94
+; GFX11-TRUE16-NEXT: s_lshr_b32 s90, s16, 16
+; GFX11-TRUE16-NEXT: s_lshr_b32 s91, s3, 16
+; GFX11-TRUE16-NEXT: s_lshr_b32 s92, s2, 16
+; GFX11-TRUE16-NEXT: s_lshr_b32 s93, s1, 16
+; GFX11-TRUE16-NEXT: s_lshr_b32 s94, s0, 16
+; GFX11-TRUE16-NEXT: s_and_not1_b32 vcc_lo, exec_lo, s89
; GFX11-TRUE16-NEXT: s_cbranch_vccnz .LBB41_3
; GFX11-TRUE16-NEXT: .LBB41_2: ; %cmp.true
; GFX11-TRUE16-NEXT: s_add_u32 s6, s6, 3
@@ -29034,19 +29058,22 @@ define inreg <60 x i16> @bitcast_v15i64_to_v60i16_scalar(<15 x i64> inreg %a, i3
; GFX11-TRUE16-NEXT: s_lshr_b32 s78, s19, 16
; GFX11-TRUE16-NEXT: s_lshr_b32 s79, s18, 16
; GFX11-TRUE16-NEXT: s_lshr_b32 s88, s17, 16
-; GFX11-TRUE16-NEXT: s_lshr_b32 s89, s16, 16
-; GFX11-TRUE16-NEXT: s_lshr_b32 s90, s3, 16
-; GFX11-TRUE16-NEXT: s_lshr_b32 s91, s2, 16
-; GFX11-TRUE16-NEXT: s_lshr_b32 s92, s1, 16
-; GFX11-TRUE16-NEXT: s_lshr_b32 s93, s0, 16
+; GFX11-TRUE16-NEXT: s_lshr_b32 s90, s16, 16
+; GFX11-TRUE16-NEXT: s_lshr_b32 s91, s3, 16
+; GFX11-TRUE16-NEXT: s_lshr_b32 s92, s2, 16
+; GFX11-TRUE16-NEXT: s_lshr_b32 s93, s1, 16
+; GFX11-TRUE16-NEXT: s_lshr_b32 s94, s0, 16
; GFX11-TRUE16-NEXT: .LBB41_3: ; %end
; GFX11-TRUE16-NEXT: s_delay_alu instid0(SALU_CYCLE_1)
-; GFX11-TRUE16-NEXT: s_pack_ll_b32_b16 s0, s0, s93
-; GFX11-TRUE16-NEXT: s_pack_ll_b32_b16 s1, s1, s92
-; GFX11-TRUE16-NEXT: s_pack_ll_b32_b16 s2, s2, s91
-; GFX11-TRUE16-NEXT: s_pack_ll_b32_b16 s3, s3, s90
-; GFX11-TRUE16-NEXT: s_pack_ll_b32_b16 s16, s16, s89
+; GFX11-TRUE16-NEXT: s_mov_b32 s89, s94
+; GFX11-TRUE16-NEXT: s_pack_ll_b32_b16 s1, s1, s93
+; GFX11-TRUE16-NEXT: s_pack_ll_b32_b16 s0, s0, s89
+; GFX11-TRUE16-NEXT: s_mov_b32 s89, s92
+; GFX11-TRUE16-NEXT: s_pack_ll_b32_b16 s3, s3, s91
+; GFX11-TRUE16-NEXT: s_pack_ll_b32_b16 s2, s2, s89
+; GFX11-TRUE16-NEXT: s_mov_b32 s89, s90
; GFX11-TRUE16-NEXT: s_pack_ll_b32_b16 s17, s17, s88
+; GFX11-TRUE16-NEXT: s_pack_ll_b32_b16 s16, s16, s89
; GFX11-TRUE16-NEXT: s_pack_ll_b32_b16 s18, s18, s79
; GFX11-TRUE16-NEXT: s_pack_ll_b32_b16 s19, s19, s78
; GFX11-TRUE16-NEXT: s_pack_ll_b32_b16 s20, s20, s77
@@ -29088,11 +29115,11 @@ define inreg <60 x i16> @bitcast_v15i64_to_v60i16_scalar(<15 x i64> inreg %a, i3
; GFX11-TRUE16-NEXT: v_dual_mov_b32 v28, s6 :: v_dual_mov_b32 v29, s4
; GFX11-TRUE16-NEXT: s_setpc_b64 s[30:31]
; GFX11-TRUE16-NEXT: .LBB41_4:
+; GFX11-TRUE16-NEXT: ; implicit-def: $sgpr94_lo16
; GFX11-TRUE16-NEXT: ; implicit-def: $sgpr93_lo16
; GFX11-TRUE16-NEXT: ; implicit-def: $sgpr92_lo16
; GFX11-TRUE16-NEXT: ; implicit-def: $sgpr91_lo16
; GFX11-TRUE16-NEXT: ; implicit-def: $sgpr90_lo16
-; GFX11-TRUE16-NEXT: ; implicit-def: $sgpr89_lo16
; GFX11-TRUE16-NEXT: ; implicit-def: $sgpr88_lo16
; GFX11-TRUE16-NEXT: ; implicit-def: $sgpr79_lo16
; GFX11-TRUE16-NEXT: ; implicit-def: $sgpr78_lo16
@@ -30210,8 +30237,8 @@ define <15 x i64> @bitcast_v60i16_to_v15i64(<60 x i16> %a, i32 %b) {
; GFX9-NEXT: v_lshrrev_b32_e32 v0, 16, v44
; GFX9-NEXT: v_lshrrev_b32_e32 v32, 16, v45
; GFX9-NEXT: v_lshrrev_b32_e32 v33, 16, v46
-; GFX9-NEXT: v_lshrrev_b32_e32 v63, 16, v47
-; GFX9-NEXT: v_lshrrev_b32_e32 v34, 16, v56
+; GFX9-NEXT: v_lshrrev_b32_e32 v34, 16, v47
+; GFX9-NEXT: v_lshrrev_b32_e32 v63, 16, v56
; GFX9-NEXT: v_lshrrev_b32_e32 v35, 16, v57
; GFX9-NEXT: v_lshrrev_b32_e32 v36, 16, v58
; GFX9-NEXT: v_lshrrev_b32_e32 v62, 16, v59
@@ -30246,8 +30273,8 @@ define <15 x i64> @bitcast_v60i16_to_v15i64(<60 x i16> %a, i32 %b) {
; GFX9-NEXT: v_perm_b32 v2, v62, v59, s6
; GFX9-NEXT: v_perm_b32 v3, v36, v58, s6
; GFX9-NEXT: v_perm_b32 v4, v35, v57, s6
-; GFX9-NEXT: v_perm_b32 v5, v34, v56, s6
-; GFX9-NEXT: v_perm_b32 v6, v63, v47, s6
+; GFX9-NEXT: v_perm_b32 v5, v63, v56, s6
+; GFX9-NEXT: v_perm_b32 v6, v34, v47, s6
; GFX9-NEXT: v_perm_b32 v7, v33, v46, s6
; GFX9-NEXT: v_perm_b32 v8, v32, v45, s6
; GFX9-NEXT: v_perm_b32 v9, v9, v44, s6
@@ -30289,8 +30316,8 @@ define <15 x i64> @bitcast_v60i16_to_v15i64(<60 x i16> %a, i32 %b) {
; GFX9-NEXT: ; implicit-def: $vgpr39
; GFX9-NEXT: ; implicit-def: $vgpr32
; GFX9-NEXT: ; implicit-def: $vgpr33
-; GFX9-NEXT: ; implicit-def: $vgpr63
; GFX9-NEXT: ; implicit-def: $vgpr34
+; GFX9-NEXT: ; implicit-def: $vgpr63
; GFX9-NEXT: ; implicit-def: $vgpr35
; GFX9-NEXT: ; implicit-def: $vgpr36
; GFX9-NEXT: ; implicit-def: $vgpr62
@@ -30404,8 +30431,8 @@ define <15 x i64> @bitcast_v60i16_to_v15i64(<60 x i16> %a, i32 %b) {
; GFX9-NEXT: v_perm_b32 v2, v62, v59, s6
; GFX9-NEXT: v_perm_b32 v3, v36, v58, s6
; GFX9-NEXT: v_perm_b32 v4, v35, v57, s6
-; GFX9-NEXT: v_perm_b32 v5, v34, v56, s6
-; GFX9-NEXT: v_perm_b32 v6, v63, v47, s6
+; GFX9-NEXT: v_perm_b32 v5, v63, v56, s6
+; GFX9-NEXT: v_perm_b32 v6, v34, v47, s6
; GFX9-NEXT: v_perm_b32 v7, v33, v46, s6
; GFX9-NEXT: v_perm_b32 v8, v32, v45, s6
; GFX9-NEXT: v_pk_add_u16 v0, v0, 3 op_sel_hi:[1,0]
@@ -31952,39 +31979,43 @@ define <60 x half> @bitcast_v15i64_to_v60f16(<15 x i64> %a, i32 %b) {
; SI: ; %bb.0:
; SI-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; SI-NEXT: buffer_load_dword v31, off, s[0:3], s32
-; SI-NEXT: ; implicit-def: $vgpr32
-; SI-NEXT: ; kill: killed $vgpr32
-; SI-NEXT: ; implicit-def: $vgpr32
-; SI-NEXT: ; kill: killed $vgpr32
-; SI-NEXT: ; implicit-def: $vgpr32
-; SI-NEXT: ; kill: killed $vgpr32
-; SI-NEXT: ; implicit-def: $vgpr32
-; SI-NEXT: ; kill: killed $vgpr32
-; SI-NEXT: ; implicit-def: $vgpr32
-; SI-NEXT: ; kill: killed $vgpr32
-; SI-NEXT: ; implicit-def: $vgpr32
-; SI-NEXT: ; kill: killed $vgpr32
-; SI-NEXT: ; implicit-def: $vgpr32
-; SI-NEXT: ; kill: killed $vgpr32
-; SI-NEXT: ; implicit-def: $vgpr32
-; SI-NEXT: ; kill: killed $vgpr32
-; SI-NEXT: ; implicit-def: $vgpr32
-; SI-NEXT: ; kill: killed $vgpr32
-; SI-NEXT: ; implicit-def: $vgpr32
-; SI-NEXT: ; kill: killed $vgpr32
-; SI-NEXT: ; implicit-def: $vgpr32
-; SI-NEXT: ; kill: killed $vgpr32
-; SI-NEXT: ; implicit-def: $vgpr32
-; SI-NEXT: ; kill: killed $vgpr32
-; SI-NEXT: ; implicit-def: $vgpr32
-; SI-NEXT: ; kill: killed $vgpr32
-; SI-NEXT: ; implicit-def: $vgpr32
-; SI-NEXT: ; kill: killed $vgpr32
-; SI-NEXT: ; implicit-def: $vgpr32
-; SI-NEXT: ; kill: killed $vgpr32
-; SI-NEXT: ; implicit-def: $vgpr32
-; SI-NEXT: ; kill: killed $vgpr32
-; SI-NEXT: ; implicit-def: $vgpr32
+; SI-NEXT: ; implicit-def: $vgpr35
+; SI-NEXT: ; kill: killed $vgpr35
+; SI-NEXT: ; implicit-def: $vgpr35
+; SI-NEXT: ; kill: killed $vgpr35
+; SI-NEXT: ; implicit-def: $vgpr35
+; SI-NEXT: ; kill: killed $vgpr35
+; SI-NEXT: ; implicit-def: $vgpr35
+; SI-NEXT: ; kill: killed $vgpr35
+; SI-NEXT: ; implicit-def: $vgpr35
+; SI-NEXT: ; kill: killed $vgpr35
+; SI-NEXT: ; implicit-def: $vgpr35
+; SI-NEXT: ; kill: killed $vgpr35
+; SI-NEXT: ; implicit-def: $vgpr35
+; SI-NEXT: ; kill: killed $vgpr35
+; SI-NEXT: ; implicit-def: $vgpr35
+; SI-NEXT: ; kill: killed $vgpr35
+; SI-NEXT: ; implicit-def: $vgpr35
+; SI-NEXT: ; kill: killed $vgpr35
+; SI-NEXT: ; implicit-def: $vgpr35
+; SI-NEXT: ; kill: killed $vgpr35
+; SI-NEXT: ; implicit-def: $vgpr35
+; SI-NEXT: ; kill: killed $vgpr35
+; SI-NEXT: ; implicit-def: $vgpr35
+; SI-NEXT: ; kill: killed $vgpr35
+; SI-NEXT: ; implicit-def: $vgpr35
+; SI-NEXT: ; kill: killed $vgpr35
+; SI-NEXT: ; implicit-def: $vgpr35
+; SI-NEXT: ; kill: killed $vgpr35
+; SI-NEXT: ; implicit-def: $vgpr35
+; SI-NEXT: ; kill: killed $vgpr35
+; SI-NEXT: ; implicit-def: $vgpr35
+; SI-NEXT: ; kill: killed $vgpr35
+; SI-NEXT: ; implicit-def: $vgpr35
+; SI-NEXT: ; kill: killed $vgpr35
+; SI-NEXT: ; implicit-def: $vgpr35
+; SI-NEXT: ; kill: killed $vgpr35
+; SI-NEXT: ; implicit-def: $vgpr35
; SI-NEXT: buffer_store_dword v40, off, s[0:3], s32 offset:64 ; 4-byte Folded Spill
; SI-NEXT: buffer_store_dword v41, off, s[0:3], s32 offset:60 ; 4-byte Folded Spill
; SI-NEXT: buffer_store_dword v42, off, s[0:3], s32 offset:56 ; 4-byte Folded Spill
@@ -32007,9 +32038,11 @@ define <60 x half> @bitcast_v15i64_to_v60f16(<15 x i64> %a, i32 %b) {
; SI-NEXT: ; implicit-def: $vgpr47
; SI-NEXT: ; implicit-def: $vgpr44
; SI-NEXT: ; implicit-def: $vgpr45
+; SI-NEXT: ; implicit-def: $vgpr42
; SI-NEXT: ; implicit-def: $vgpr43
; SI-NEXT: ; implicit-def: $vgpr40
-; SI-NEXT: ; implicit-def: $vgpr54
+; SI-NEXT: ; implicit-def: $vgpr34
+; SI-NEXT: ; implicit-def: $vgpr32
; SI-NEXT: ; implicit-def: $vgpr52
; SI-NEXT: ; implicit-def: $vgpr50
; SI-NEXT: ; implicit-def: $vgpr48
@@ -32020,20 +32053,18 @@ define <60 x half> @bitcast_v15i64_to_v60f16(<15 x i64> %a, i32 %b) {
; SI-NEXT: ; implicit-def: $vgpr60
; SI-NEXT: ; implicit-def: $vgpr58
; SI-NEXT: ; implicit-def: $vgpr61
-; SI-NEXT: ; kill: killed $vgpr32
-; SI-NEXT: ; implicit-def: $vgpr39
-; SI-NEXT: ; implicit-def: $vgpr63
+; SI-NEXT: ; kill: killed $vgpr35
; SI-NEXT: ; implicit-def: $vgpr49
-; SI-NEXT: ; implicit-def: $vgpr59
+; SI-NEXT: ; implicit-def: $vgpr63
; SI-NEXT: ; implicit-def: $vgpr51
-; SI-NEXT: ; implicit-def: $vgpr37
+; SI-NEXT: ; implicit-def: $vgpr59
; SI-NEXT: ; implicit-def: $vgpr53
-; SI-NEXT: ; implicit-def: $vgpr35
+; SI-NEXT: ; implicit-def: $vgpr39
; SI-NEXT: ; implicit-def: $vgpr55
-; SI-NEXT: ; implicit-def: $vgpr34
+; SI-NEXT: ; implicit-def: $vgpr37
+; SI-NEXT: ; implicit-def: $vgpr54
+; SI-NEXT: ; implicit-def: $vgpr35
; SI-NEXT: ; implicit-def: $vgpr41
-; SI-NEXT: ; implicit-def: $vgpr32
-; SI-NEXT: ; implicit-def: $vgpr42
; SI-NEXT: s_waitcnt vmcnt(14)
; SI-NEXT: v_cmp_ne_u32_e32 vcc, 0, v31
; SI-NEXT: ; implicit-def: $vgpr31
@@ -32053,18 +32084,14 @@ define <60 x half> @bitcast_v15i64_to_v60f16(<15 x i64> %a, i32 %b) {
; SI-NEXT: ; implicit-def: $vgpr31
; SI-NEXT: ; kill: killed $vgpr31
; SI-NEXT: ; implicit-def: $vgpr31
-; SI-NEXT: ; kill: killed $vgpr31
-; SI-NEXT: ; implicit-def: $vgpr31
-; SI-NEXT: ; kill: killed $vgpr31
-; SI-NEXT: ; implicit-def: $vgpr31
; SI-NEXT: s_and_saveexec_b64 s[4:5], vcc
; SI-NEXT: s_xor_b64 s[4:5], exec, s[4:5]
; SI-NEXT: s_cbranch_execz .LBB44_2
; SI-NEXT: ; %bb.1: ; %cmp.false
; SI-NEXT: v_lshrrev_b32_e32 v31, 16, v30
-; SI-NEXT: v_cvt_f32_f16_e32 v42, v31
-; SI-NEXT: v_lshrrev_b32_e32 v31, 16, v29
; SI-NEXT: v_cvt_f32_f16_e32 v41, v31
+; SI-NEXT: v_lshrrev_b32_e32 v31, 16, v29
+; SI-NEXT: v_cvt_f32_f16_e32 v54, v31
; SI-NEXT: v_lshrrev_b32_e32 v31, 16, v28
; SI-NEXT: v_cvt_f32_f16_e32 v55, v31
; SI-NEXT: v_lshrrev_b32_e32 v31, 16, v27
@@ -32074,52 +32101,57 @@ define <60 x half> @bitcast_v15i64_to_v60f16(<15 x i64> %a, i32 %b) {
; SI-NEXT: v_lshrrev_b32_e32 v31, 16, v25
; SI-NEXT: v_cvt_f32_f16_e32 v49, v31
; SI-NEXT: v_lshrrev_b32_e32 v31, 16, v24
-; SI-NEXT: v_cvt_f32_f16_e32 v39, v31
-; SI-NEXT: v_lshrrev_b32_e32 v31, 16, v23
; SI-NEXT: s_waitcnt expcnt(2)
; SI-NEXT: v_cvt_f32_f16_e32 v61, v31
-; SI-NEXT: v_lshrrev_b32_e32 v31, 16, v22
+; SI-NEXT: v_lshrrev_b32_e32 v31, 16, v23
; SI-NEXT: v_cvt_f32_f16_e32 v31, v31
; SI-NEXT: v_lshrrev_b32_e32 v32, 16, v12
; SI-NEXT: v_cvt_f32_f16_e32 v33, v32
; SI-NEXT: v_lshrrev_b32_e32 v32, 16, v11
; SI-NEXT: buffer_store_dword v31, off, s[0:3], s32 offset:68 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_lshrrev_b32_e32 v31, 16, v21
+; SI-NEXT: v_lshrrev_b32_e32 v31, 16, v22
; SI-NEXT: v_cvt_f32_f16_e32 v31, v31
; SI-NEXT: v_cvt_f32_f16_e32 v36, v32
; SI-NEXT: v_lshrrev_b32_e32 v32, 16, v10
; SI-NEXT: v_cvt_f32_f16_e32 v38, v32
; SI-NEXT: buffer_store_dword v31, off, s[0:3], s32 offset:76 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_lshrrev_b32_e32 v31, 16, v20
+; SI-NEXT: v_lshrrev_b32_e32 v31, 16, v21
; SI-NEXT: v_cvt_f32_f16_e32 v31, v31
; SI-NEXT: v_lshrrev_b32_e32 v32, 16, v9
; SI-NEXT: v_cvt_f32_f16_e32 v48, v32
; SI-NEXT: v_lshrrev_b32_e32 v32, 16, v8
; SI-NEXT: buffer_store_dword v31, off, s[0:3], s32 offset:84 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_lshrrev_b32_e32 v31, 16, v19
+; SI-NEXT: v_lshrrev_b32_e32 v31, 16, v20
; SI-NEXT: v_cvt_f32_f16_e32 v31, v31
; SI-NEXT: v_cvt_f32_f16_e32 v50, v32
; SI-NEXT: v_lshrrev_b32_e32 v32, 16, v7
; SI-NEXT: v_cvt_f32_f16_e32 v52, v32
-; SI-NEXT: buffer_store_dword v31, off, s[0:3], s32 offset:96 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v31, off, s[0:3], s32 offset:92 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_lshrrev_b32_e32 v31, 16, v18
+; SI-NEXT: v_lshrrev_b32_e32 v31, 16, v19
; SI-NEXT: v_cvt_f32_f16_e32 v31, v31
; SI-NEXT: v_lshrrev_b32_e32 v32, 16, v6
-; SI-NEXT: v_cvt_f32_f16_e32 v54, v32
-; SI-NEXT: v_lshrrev_b32_e32 v32, 16, v5
+; SI-NEXT: v_lshrrev_b32_e32 v35, 16, v4
+; SI-NEXT: v_cvt_f32_f16_e32 v25, v25
; SI-NEXT: buffer_store_dword v31, off, s[0:3], s32 offset:104 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_lshrrev_b32_e32 v31, 16, v17
+; SI-NEXT: v_lshrrev_b32_e32 v31, 16, v18
; SI-NEXT: v_cvt_f32_f16_e32 v31, v31
-; SI-NEXT: v_cvt_f32_f16_e32 v40, v32
-; SI-NEXT: v_lshrrev_b32_e32 v32, 16, v4
; SI-NEXT: v_cvt_f32_f16_e32 v24, v24
+; SI-NEXT: v_cvt_f32_f16_e32 v23, v23
+; SI-NEXT: v_cvt_f32_f16_e32 v22, v22
; SI-NEXT: buffer_store_dword v31, off, s[0:3], s32 offset:112 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
+; SI-NEXT: v_lshrrev_b32_e32 v31, 16, v17
+; SI-NEXT: v_cvt_f32_f16_e32 v31, v31
+; SI-NEXT: v_cvt_f32_f16_e32 v21, v21
+; SI-NEXT: v_cvt_f32_f16_e32 v20, v20
+; SI-NEXT: v_cvt_f32_f16_e32 v19, v19
+; SI-NEXT: buffer_store_dword v31, off, s[0:3], s32 offset:120 ; 4-byte Folded Spill
+; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_lshrrev_b32_e32 v31, 16, v16
; SI-NEXT: v_cvt_f32_f16_e32 v58, v31
; SI-NEXT: v_lshrrev_b32_e32 v31, 16, v15
@@ -32127,11 +32159,6 @@ define <60 x half> @bitcast_v15i64_to_v60f16(<15 x i64> %a, i32 %b) {
; SI-NEXT: v_lshrrev_b32_e32 v31, 16, v14
; SI-NEXT: v_cvt_f32_f16_e32 v62, v31
; SI-NEXT: v_lshrrev_b32_e32 v31, 16, v13
-; SI-NEXT: v_cvt_f32_f16_e32 v23, v23
-; SI-NEXT: v_cvt_f32_f16_e32 v22, v22
-; SI-NEXT: v_cvt_f32_f16_e32 v21, v21
-; SI-NEXT: v_cvt_f32_f16_e32 v20, v20
-; SI-NEXT: v_cvt_f32_f16_e32 v19, v19
; SI-NEXT: v_cvt_f32_f16_e32 v18, v18
; SI-NEXT: v_cvt_f32_f16_e32 v17, v17
; SI-NEXT: v_cvt_f32_f16_e32 v16, v16
@@ -32145,43 +32172,44 @@ define <60 x half> @bitcast_v15i64_to_v60f16(<15 x i64> %a, i32 %b) {
; SI-NEXT: v_cvt_f32_f16_e32 v8, v8
; SI-NEXT: v_cvt_f32_f16_e32 v7, v7
; SI-NEXT: v_cvt_f32_f16_e32 v6, v6
-; SI-NEXT: v_cvt_f32_f16_e32 v5, v5
-; SI-NEXT: v_cvt_f32_f16_e32 v4, v4
-; SI-NEXT: v_cvt_f32_f16_e32 v43, v32
-; SI-NEXT: v_lshrrev_b32_e32 v32, 16, v3
-; SI-NEXT: v_cvt_f32_f16_e32 v45, v32
-; SI-NEXT: v_lshrrev_b32_e32 v32, 16, v2
-; SI-NEXT: v_cvt_f32_f16_e32 v47, v32
-; SI-NEXT: v_lshrrev_b32_e32 v32, 16, v1
+; SI-NEXT: v_cvt_f32_f16_e32 v43, v35
+; SI-NEXT: v_lshrrev_b32_e32 v35, 16, v3
+; SI-NEXT: v_cvt_f32_f16_e32 v45, v35
+; SI-NEXT: v_lshrrev_b32_e32 v35, 16, v2
+; SI-NEXT: v_lshrrev_b32_e32 v34, 16, v5
+; SI-NEXT: v_cvt_f32_f16_e32 v47, v35
+; SI-NEXT: v_lshrrev_b32_e32 v35, 16, v1
; SI-NEXT: v_cvt_f32_f16_e32 v31, v31
-; SI-NEXT: v_cvt_f32_f16_e32 v57, v32
-; SI-NEXT: v_cvt_f32_f16_e32 v32, v30
-; SI-NEXT: v_cvt_f32_f16_e32 v34, v29
-; SI-NEXT: v_cvt_f32_f16_e32 v35, v28
-; SI-NEXT: v_cvt_f32_f16_e32 v37, v27
-; SI-NEXT: v_cvt_f32_f16_e32 v59, v26
-; SI-NEXT: v_cvt_f32_f16_e32 v63, v25
-; SI-NEXT: buffer_store_dword v24, off, s[0:3], s32 offset:72 ; 4-byte Folded Spill
-; SI-NEXT: buffer_store_dword v23, off, s[0:3], s32 offset:80 ; 4-byte Folded Spill
-; SI-NEXT: buffer_store_dword v22, off, s[0:3], s32 offset:88 ; 4-byte Folded Spill
-; SI-NEXT: buffer_store_dword v21, off, s[0:3], s32 offset:92 ; 4-byte Folded Spill
-; SI-NEXT: buffer_store_dword v20, off, s[0:3], s32 offset:100 ; 4-byte Folded Spill
-; SI-NEXT: buffer_store_dword v19, off, s[0:3], s32 offset:108 ; 4-byte Folded Spill
-; SI-NEXT: buffer_store_dword v18, off, s[0:3], s32 offset:116 ; 4-byte Folded Spill
-; SI-NEXT: buffer_store_dword v17, off, s[0:3], s32 offset:120 ; 4-byte Folded Spill
-; SI-NEXT: buffer_store_dword v16, off, s[0:3], s32 offset:124 ; 4-byte Folded Spill
-; SI-NEXT: buffer_store_dword v15, off, s[0:3], s32 offset:128 ; 4-byte Folded Spill
-; SI-NEXT: buffer_store_dword v14, off, s[0:3], s32 offset:132 ; 4-byte Folded Spill
-; SI-NEXT: buffer_store_dword v13, off, s[0:3], s32 offset:136 ; 4-byte Folded Spill
-; SI-NEXT: buffer_store_dword v12, off, s[0:3], s32 offset:140 ; 4-byte Folded Spill
-; SI-NEXT: buffer_store_dword v11, off, s[0:3], s32 offset:144 ; 4-byte Folded Spill
-; SI-NEXT: buffer_store_dword v10, off, s[0:3], s32 offset:148 ; 4-byte Folded Spill
-; SI-NEXT: buffer_store_dword v9, off, s[0:3], s32 offset:152 ; 4-byte Folded Spill
-; SI-NEXT: buffer_store_dword v8, off, s[0:3], s32 offset:156 ; 4-byte Folded Spill
-; SI-NEXT: buffer_store_dword v7, off, s[0:3], s32 offset:160 ; 4-byte Folded Spill
-; SI-NEXT: buffer_store_dword v6, off, s[0:3], s32 offset:164 ; 4-byte Folded Spill
-; SI-NEXT: buffer_store_dword v5, off, s[0:3], s32 offset:168 ; 4-byte Folded Spill
-; SI-NEXT: buffer_store_dword v4, off, s[0:3], s32 offset:172 ; 4-byte Folded Spill
+; SI-NEXT: v_cvt_f32_f16_e32 v32, v32
+; SI-NEXT: v_cvt_f32_f16_e32 v34, v34
+; SI-NEXT: v_cvt_f32_f16_e32 v57, v35
+; SI-NEXT: v_cvt_f32_f16_e32 v35, v30
+; SI-NEXT: v_cvt_f32_f16_e32 v37, v29
+; SI-NEXT: v_cvt_f32_f16_e32 v39, v28
+; SI-NEXT: v_cvt_f32_f16_e32 v59, v27
+; SI-NEXT: v_cvt_f32_f16_e32 v63, v26
+; SI-NEXT: buffer_store_dword v25, off, s[0:3], s32 offset:72 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v24, off, s[0:3], s32 offset:80 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v23, off, s[0:3], s32 offset:88 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v22, off, s[0:3], s32 offset:96 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v21, off, s[0:3], s32 offset:100 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v20, off, s[0:3], s32 offset:108 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v19, off, s[0:3], s32 offset:116 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v18, off, s[0:3], s32 offset:124 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v17, off, s[0:3], s32 offset:128 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v16, off, s[0:3], s32 offset:132 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v15, off, s[0:3], s32 offset:136 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v14, off, s[0:3], s32 offset:140 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v13, off, s[0:3], s32 offset:144 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v12, off, s[0:3], s32 offset:148 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v11, off, s[0:3], s32 offset:152 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v10, off, s[0:3], s32 offset:156 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v9, off, s[0:3], s32 offset:160 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v8, off, s[0:3], s32 offset:164 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v7, off, s[0:3], s32 offset:168 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v6, off, s[0:3], s32 offset:172 ; 4-byte Folded Spill
+; SI-NEXT: v_cvt_f32_f16_e32 v40, v5
+; SI-NEXT: v_cvt_f32_f16_e32 v42, v4
; SI-NEXT: v_cvt_f32_f16_e32 v44, v3
; SI-NEXT: v_cvt_f32_f16_e32 v46, v2
; SI-NEXT: v_cvt_f32_f16_e32 v56, v1
@@ -32222,9 +32250,9 @@ define <60 x half> @bitcast_v15i64_to_v60f16(<15 x i64> %a, i32 %b) {
; SI-NEXT: v_add_i32_e32 v1, vcc, 3, v1
; SI-NEXT: v_addc_u32_e32 v2, vcc, 0, v2, vcc
; SI-NEXT: v_add_i32_e32 v3, vcc, 3, v3
-; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_addc_u32_e32 v4, vcc, 0, v4, vcc
; SI-NEXT: v_add_i32_e32 v5, vcc, 3, v5
+; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_addc_u32_e32 v6, vcc, 0, v6, vcc
; SI-NEXT: v_add_i32_e32 v7, vcc, 3, v7
; SI-NEXT: v_addc_u32_e32 v8, vcc, 0, v8, vcc
@@ -32242,38 +32270,57 @@ define <60 x half> @bitcast_v15i64_to_v60f16(<15 x i64> %a, i32 %b) {
; SI-NEXT: v_addc_u32_e32 v20, vcc, 0, v20, vcc
; SI-NEXT: v_add_i32_e32 v21, vcc, 3, v21
; SI-NEXT: v_addc_u32_e32 v22, vcc, 0, v22, vcc
-; SI-NEXT: v_lshrrev_b32_e32 v35, 16, v22
-; SI-NEXT: v_lshrrev_b32_e32 v47, 16, v2
+; SI-NEXT: v_add_i32_e32 v23, vcc, 3, v23
+; SI-NEXT: v_lshrrev_b32_e32 v31, 16, v1
+; SI-NEXT: v_lshrrev_b32_e32 v37, 16, v23
+; SI-NEXT: buffer_store_dword v31, off, s[0:3], s32 offset:176 ; 4-byte Folded Spill
+; SI-NEXT: s_waitcnt expcnt(0)
+; SI-NEXT: v_lshrrev_b32_e32 v31, 16, v2
; SI-NEXT: v_cvt_f32_f16_e32 v46, v2
-; SI-NEXT: v_cvt_f32_f16_e32 v2, v35
+; SI-NEXT: v_cvt_f32_f16_e32 v2, v37
+; SI-NEXT: v_lshrrev_b32_e32 v35, 16, v22
; SI-NEXT: v_lshrrev_b32_e32 v34, 16, v21
; SI-NEXT: v_lshrrev_b32_e32 v32, 16, v20
-; SI-NEXT: v_add_i32_e32 v23, vcc, 3, v23
; SI-NEXT: buffer_store_dword v2, off, s[0:3], s32 offset:68 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_cvt_f32_f16_e32 v2, v34
+; SI-NEXT: v_cvt_f32_f16_e32 v2, v35
; SI-NEXT: v_addc_u32_e32 v24, vcc, 0, v24, vcc
-; SI-NEXT: v_add_i32_e32 v25, vcc, 3, v25
+; SI-NEXT: v_lshrrev_b32_e32 v63, 16, v19
; SI-NEXT: buffer_store_dword v2, off, s[0:3], s32 offset:76 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_cvt_f32_f16_e32 v2, v32
-; SI-NEXT: v_addc_u32_e32 v26, vcc, 0, v26, vcc
-; SI-NEXT: v_lshrrev_b32_e32 v63, 16, v19
-; SI-NEXT: v_lshrrev_b32_e32 v37, 16, v23
-; SI-NEXT: v_add_i32_e32 v27, vcc, 3, v27
-; SI-NEXT: v_lshrrev_b32_e32 v57, 16, v1
-; SI-NEXT: v_addc_u32_e32 v28, vcc, 0, v28, vcc
+; SI-NEXT: v_cvt_f32_f16_e32 v2, v34
+; SI-NEXT: v_lshrrev_b32_e32 v39, 16, v24
; SI-NEXT: v_cvt_f32_f16_e32 v56, v1
-; SI-NEXT: v_cvt_f32_f16_e32 v1, v37
+; SI-NEXT: v_cvt_f32_f16_e32 v1, v39
; SI-NEXT: buffer_store_dword v2, off, s[0:3], s32 offset:84 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
+; SI-NEXT: v_cvt_f32_f16_e32 v2, v32
+; SI-NEXT: v_lshrrev_b32_e32 v59, 16, v17
+; SI-NEXT: v_lshrrev_b32_e32 v61, 16, v18
+; SI-NEXT: buffer_store_dword v31, off, s[0:3], s32 offset:180 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v2, off, s[0:3], s32 offset:92 ; 4-byte Folded Spill
+; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_cvt_f32_f16_e32 v2, v63
+; SI-NEXT: v_lshrrev_b32_e32 v47, 16, v4
+; SI-NEXT: v_lshrrev_b32_e32 v43, 16, v6
+; SI-NEXT: v_cvt_f32_f16_e32 v32, v43
+; SI-NEXT: buffer_store_dword v2, off, s[0:3], s32 offset:104 ; 4-byte Folded Spill
+; SI-NEXT: s_waitcnt expcnt(0)
+; SI-NEXT: v_cvt_f32_f16_e32 v2, v61
+; SI-NEXT: v_mov_b32_e32 v61, v1
+; SI-NEXT: v_cvt_f32_f16_e32 v1, v59
+; SI-NEXT: v_cvt_f32_f16_e32 v43, v47
+; SI-NEXT: v_add_i32_e32 v25, vcc, 3, v25
+; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:120 ; 4-byte Folded Spill
+; SI-NEXT: s_waitcnt expcnt(0)
+; SI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:180 ; 4-byte Folded Reload
+; SI-NEXT: v_addc_u32_e32 v26, vcc, 0, v26, vcc
+; SI-NEXT: v_add_i32_e32 v27, vcc, 3, v27
+; SI-NEXT: v_addc_u32_e32 v28, vcc, 0, v28, vcc
; SI-NEXT: v_add_i32_e32 v29, vcc, 3, v29
; SI-NEXT: v_addc_u32_e32 v30, vcc, 0, v30, vcc
-; SI-NEXT: v_lshrrev_b32_e32 v45, 16, v3
-; SI-NEXT: v_lshrrev_b32_e32 v43, 16, v4
-; SI-NEXT: v_lshrrev_b32_e32 v40, 16, v5
-; SI-NEXT: v_lshrrev_b32_e32 v54, 16, v6
+; SI-NEXT: v_lshrrev_b32_e32 v57, 16, v3
+; SI-NEXT: v_lshrrev_b32_e32 v45, 16, v5
; SI-NEXT: v_lshrrev_b32_e32 v52, 16, v7
; SI-NEXT: v_lshrrev_b32_e32 v50, 16, v8
; SI-NEXT: v_lshrrev_b32_e32 v48, 16, v9
@@ -32284,15 +32331,12 @@ define <60 x half> @bitcast_v15i64_to_v60f16(<15 x i64> %a, i32 %b) {
; SI-NEXT: v_lshrrev_b32_e32 v62, 16, v14
; SI-NEXT: v_lshrrev_b32_e32 v60, 16, v15
; SI-NEXT: v_lshrrev_b32_e32 v58, 16, v16
-; SI-NEXT: v_lshrrev_b32_e32 v59, 16, v17
-; SI-NEXT: v_lshrrev_b32_e32 v61, 16, v18
-; SI-NEXT: v_lshrrev_b32_e32 v39, 16, v24
; SI-NEXT: v_lshrrev_b32_e32 v49, 16, v25
; SI-NEXT: v_lshrrev_b32_e32 v51, 16, v26
; SI-NEXT: v_lshrrev_b32_e32 v53, 16, v27
; SI-NEXT: v_lshrrev_b32_e32 v55, 16, v28
-; SI-NEXT: v_lshrrev_b32_e32 v41, 16, v29
-; SI-NEXT: v_lshrrev_b32_e32 v42, 16, v30
+; SI-NEXT: v_lshrrev_b32_e32 v54, 16, v29
+; SI-NEXT: v_lshrrev_b32_e32 v41, 16, v30
; SI-NEXT: v_cvt_f32_f16_e32 v30, v30
; SI-NEXT: v_cvt_f32_f16_e32 v29, v29
; SI-NEXT: v_cvt_f32_f16_e32 v28, v28
@@ -32318,21 +32362,15 @@ define <60 x half> @bitcast_v15i64_to_v60f16(<15 x i64> %a, i32 %b) {
; SI-NEXT: v_cvt_f32_f16_e32 v8, v8
; SI-NEXT: v_cvt_f32_f16_e32 v7, v7
; SI-NEXT: v_cvt_f32_f16_e32 v6, v6
-; SI-NEXT: v_cvt_f32_f16_e32 v5, v5
-; SI-NEXT: v_cvt_f32_f16_e32 v4, v4
+; SI-NEXT: v_cvt_f32_f16_e32 v40, v5
+; SI-NEXT: v_cvt_f32_f16_e32 v42, v4
; SI-NEXT: v_cvt_f32_f16_e32 v44, v3
-; SI-NEXT: v_cvt_f32_f16_e32 v42, v42
; SI-NEXT: v_cvt_f32_f16_e32 v41, v41
+; SI-NEXT: v_cvt_f32_f16_e32 v54, v54
; SI-NEXT: v_cvt_f32_f16_e32 v55, v55
; SI-NEXT: v_cvt_f32_f16_e32 v53, v53
; SI-NEXT: v_cvt_f32_f16_e32 v51, v51
; SI-NEXT: v_cvt_f32_f16_e32 v49, v49
-; SI-NEXT: v_cvt_f32_f16_e32 v39, v39
-; SI-NEXT: buffer_store_dword v2, off, s[0:3], s32 offset:96 ; 4-byte Folded Spill
-; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_cvt_f32_f16_e32 v2, v61
-; SI-NEXT: v_mov_b32_e32 v61, v1
-; SI-NEXT: v_cvt_f32_f16_e32 v1, v59
; SI-NEXT: v_cvt_f32_f16_e32 v58, v58
; SI-NEXT: v_cvt_f32_f16_e32 v60, v60
; SI-NEXT: v_cvt_f32_f16_e32 v62, v62
@@ -32343,45 +32381,43 @@ define <60 x half> @bitcast_v15i64_to_v60f16(<15 x i64> %a, i32 %b) {
; SI-NEXT: v_cvt_f32_f16_e32 v48, v48
; SI-NEXT: v_cvt_f32_f16_e32 v50, v50
; SI-NEXT: v_cvt_f32_f16_e32 v52, v52
-; SI-NEXT: v_cvt_f32_f16_e32 v54, v54
-; SI-NEXT: v_cvt_f32_f16_e32 v40, v40
-; SI-NEXT: v_cvt_f32_f16_e32 v43, v43
-; SI-NEXT: v_cvt_f32_f16_e32 v45, v45
-; SI-NEXT: v_cvt_f32_f16_e32 v47, v47
-; SI-NEXT: v_cvt_f32_f16_e32 v57, v57
-; SI-NEXT: v_mov_b32_e32 v37, v27
-; SI-NEXT: v_mov_b32_e32 v35, v28
-; SI-NEXT: v_mov_b32_e32 v34, v29
-; SI-NEXT: v_mov_b32_e32 v32, v30
-; SI-NEXT: v_mov_b32_e32 v63, v25
-; SI-NEXT: v_mov_b32_e32 v59, v26
-; SI-NEXT: buffer_store_dword v24, off, s[0:3], s32 offset:72 ; 4-byte Folded Spill
-; SI-NEXT: buffer_store_dword v23, off, s[0:3], s32 offset:80 ; 4-byte Folded Spill
-; SI-NEXT: buffer_store_dword v22, off, s[0:3], s32 offset:88 ; 4-byte Folded Spill
-; SI-NEXT: buffer_store_dword v21, off, s[0:3], s32 offset:92 ; 4-byte Folded Spill
-; SI-NEXT: buffer_store_dword v20, off, s[0:3], s32 offset:100 ; 4-byte Folded Spill
-; SI-NEXT: buffer_store_dword v19, off, s[0:3], s32 offset:108 ; 4-byte Folded Spill
-; SI-NEXT: buffer_store_dword v18, off, s[0:3], s32 offset:116 ; 4-byte Folded Spill
-; SI-NEXT: buffer_store_dword v17, off, s[0:3], s32 offset:120 ; 4-byte Folded Spill
-; SI-NEXT: buffer_store_dword v16, off, s[0:3], s32 offset:124 ; 4-byte Folded Spill
-; SI-NEXT: buffer_store_dword v15, off, s[0:3], s32 offset:128 ; 4-byte Folded Spill
-; SI-NEXT: buffer_store_dword v14, off, s[0:3], s32 offset:132 ; 4-byte Folded Spill
-; SI-NEXT: buffer_store_dword v13, off, s[0:3], s32 offset:136 ; 4-byte Folded Spill
-; SI-NEXT: buffer_store_dword v12, off, s[0:3], s32 offset:140 ; 4-byte Folded Spill
-; SI-NEXT: buffer_store_dword v11, off, s[0:3], s32 offset:144 ; 4-byte Folded Spill
-; SI-NEXT: buffer_store_dword v10, off, s[0:3], s32 offset:148 ; 4-byte Folded Spill
-; SI-NEXT: buffer_store_dword v9, off, s[0:3], s32 offset:152 ; 4-byte Folded Spill
-; SI-NEXT: buffer_store_dword v8, off, s[0:3], s32 offset:156 ; 4-byte Folded Spill
-; SI-NEXT: buffer_store_dword v7, off, s[0:3], s32 offset:160 ; 4-byte Folded Spill
-; SI-NEXT: buffer_store_dword v6, off, s[0:3], s32 offset:164 ; 4-byte Folded Spill
-; SI-NEXT: buffer_store_dword v5, off, s[0:3], s32 offset:168 ; 4-byte Folded Spill
-; SI-NEXT: buffer_store_dword v4, off, s[0:3], s32 offset:172 ; 4-byte Folded Spill
-; SI-NEXT: buffer_store_dword v2, off, s[0:3], s32 offset:104 ; 4-byte Folded Spill
-; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:112 ; 4-byte Folded Spill
+; SI-NEXT: v_cvt_f32_f16_e32 v34, v45
+; SI-NEXT: v_cvt_f32_f16_e32 v45, v57
+; SI-NEXT: v_mov_b32_e32 v39, v28
+; SI-NEXT: v_mov_b32_e32 v37, v29
+; SI-NEXT: v_mov_b32_e32 v35, v30
+; SI-NEXT: v_mov_b32_e32 v63, v26
+; SI-NEXT: v_mov_b32_e32 v59, v27
+; SI-NEXT: s_waitcnt vmcnt(0)
+; SI-NEXT: v_cvt_f32_f16_e32 v47, v1
+; SI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:176 ; 4-byte Folded Reload
+; SI-NEXT: buffer_store_dword v25, off, s[0:3], s32 offset:72 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v24, off, s[0:3], s32 offset:80 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v23, off, s[0:3], s32 offset:88 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v22, off, s[0:3], s32 offset:96 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v21, off, s[0:3], s32 offset:100 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v20, off, s[0:3], s32 offset:108 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v19, off, s[0:3], s32 offset:116 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v18, off, s[0:3], s32 offset:124 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v17, off, s[0:3], s32 offset:128 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v16, off, s[0:3], s32 offset:132 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v15, off, s[0:3], s32 offset:136 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v14, off, s[0:3], s32 offset:140 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v13, off, s[0:3], s32 offset:144 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v12, off, s[0:3], s32 offset:148 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v11, off, s[0:3], s32 offset:152 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v10, off, s[0:3], s32 offset:156 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v9, off, s[0:3], s32 offset:160 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v8, off, s[0:3], s32 offset:164 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v7, off, s[0:3], s32 offset:168 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v6, off, s[0:3], s32 offset:172 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v2, off, s[0:3], s32 offset:112 ; 4-byte Folded Spill
+; SI-NEXT: s_waitcnt vmcnt(14)
+; SI-NEXT: v_cvt_f32_f16_e32 v57, v1
; SI-NEXT: .LBB44_4: ; %end
; SI-NEXT: s_or_b64 exec, exec, s[4:5]
-; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_cvt_f16_f32_e32 v1, v57
+; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_cvt_f16_f32_e32 v2, v56
; SI-NEXT: v_cvt_f16_f32_e32 v3, v46
; SI-NEXT: v_lshlrev_b32_e32 v1, 16, v1
@@ -32400,34 +32436,30 @@ define <60 x half> @bitcast_v15i64_to_v60f16(<15 x i64> %a, i32 %b) {
; SI-NEXT: v_lshlrev_b32_e32 v1, 16, v1
; SI-NEXT: v_or_b32_e32 v1, v2, v1
; SI-NEXT: buffer_store_dword v1, v3, s[0:3], 0 offen
-; SI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:172 ; 4-byte Folded Reload
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_cvt_f16_f32_e32 v1, v43
+; SI-NEXT: v_cvt_f16_f32_e32 v2, v42
; SI-NEXT: v_add_i32_e32 v3, vcc, 12, v0
; SI-NEXT: v_lshlrev_b32_e32 v1, 16, v1
-; SI-NEXT: s_waitcnt vmcnt(0)
-; SI-NEXT: v_cvt_f16_f32_e32 v2, v2
; SI-NEXT: v_or_b32_e32 v1, v2, v1
; SI-NEXT: buffer_store_dword v1, v3, s[0:3], 0 offen
-; SI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:168 ; 4-byte Folded Reload
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_cvt_f16_f32_e32 v1, v40
+; SI-NEXT: v_cvt_f16_f32_e32 v1, v34
+; SI-NEXT: v_cvt_f16_f32_e32 v2, v40
; SI-NEXT: v_add_i32_e32 v3, vcc, 16, v0
; SI-NEXT: v_lshlrev_b32_e32 v1, 16, v1
-; SI-NEXT: s_waitcnt vmcnt(0)
-; SI-NEXT: v_cvt_f16_f32_e32 v2, v2
; SI-NEXT: v_or_b32_e32 v1, v2, v1
; SI-NEXT: buffer_store_dword v1, v3, s[0:3], 0 offen
-; SI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:164 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:172 ; 4-byte Folded Reload
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_cvt_f16_f32_e32 v1, v54
+; SI-NEXT: v_cvt_f16_f32_e32 v1, v32
; SI-NEXT: v_add_i32_e32 v3, vcc, 20, v0
; SI-NEXT: v_lshlrev_b32_e32 v1, 16, v1
; SI-NEXT: s_waitcnt vmcnt(0)
; SI-NEXT: v_cvt_f16_f32_e32 v2, v2
; SI-NEXT: v_or_b32_e32 v1, v2, v1
; SI-NEXT: buffer_store_dword v1, v3, s[0:3], 0 offen
-; SI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:160 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:168 ; 4-byte Folded Reload
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_cvt_f16_f32_e32 v1, v52
; SI-NEXT: v_add_i32_e32 v3, vcc, 24, v0
@@ -32436,7 +32468,7 @@ define <60 x half> @bitcast_v15i64_to_v60f16(<15 x i64> %a, i32 %b) {
; SI-NEXT: v_cvt_f16_f32_e32 v2, v2
; SI-NEXT: v_or_b32_e32 v1, v2, v1
; SI-NEXT: buffer_store_dword v1, v3, s[0:3], 0 offen
-; SI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:156 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:164 ; 4-byte Folded Reload
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_cvt_f16_f32_e32 v1, v50
; SI-NEXT: v_add_i32_e32 v3, vcc, 28, v0
@@ -32445,7 +32477,7 @@ define <60 x half> @bitcast_v15i64_to_v60f16(<15 x i64> %a, i32 %b) {
; SI-NEXT: v_cvt_f16_f32_e32 v2, v2
; SI-NEXT: v_or_b32_e32 v1, v2, v1
; SI-NEXT: buffer_store_dword v1, v3, s[0:3], 0 offen
-; SI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:152 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:160 ; 4-byte Folded Reload
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_cvt_f16_f32_e32 v1, v48
; SI-NEXT: v_add_i32_e32 v3, vcc, 32, v0
@@ -32454,7 +32486,7 @@ define <60 x half> @bitcast_v15i64_to_v60f16(<15 x i64> %a, i32 %b) {
; SI-NEXT: v_cvt_f16_f32_e32 v2, v2
; SI-NEXT: v_or_b32_e32 v1, v2, v1
; SI-NEXT: buffer_store_dword v1, v3, s[0:3], 0 offen
-; SI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:148 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:156 ; 4-byte Folded Reload
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_cvt_f16_f32_e32 v1, v38
; SI-NEXT: v_add_i32_e32 v3, vcc, 36, v0
@@ -32463,7 +32495,7 @@ define <60 x half> @bitcast_v15i64_to_v60f16(<15 x i64> %a, i32 %b) {
; SI-NEXT: v_cvt_f16_f32_e32 v2, v2
; SI-NEXT: v_or_b32_e32 v1, v2, v1
; SI-NEXT: buffer_store_dword v1, v3, s[0:3], 0 offen
-; SI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:144 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:152 ; 4-byte Folded Reload
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_cvt_f16_f32_e32 v1, v36
; SI-NEXT: v_add_i32_e32 v3, vcc, 40, v0
@@ -32472,7 +32504,7 @@ define <60 x half> @bitcast_v15i64_to_v60f16(<15 x i64> %a, i32 %b) {
; SI-NEXT: v_cvt_f16_f32_e32 v2, v2
; SI-NEXT: v_or_b32_e32 v1, v2, v1
; SI-NEXT: buffer_store_dword v1, v3, s[0:3], 0 offen
-; SI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:140 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:148 ; 4-byte Folded Reload
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_cvt_f16_f32_e32 v1, v33
; SI-NEXT: v_add_i32_e32 v3, vcc, 44, v0
@@ -32481,7 +32513,7 @@ define <60 x half> @bitcast_v15i64_to_v60f16(<15 x i64> %a, i32 %b) {
; SI-NEXT: v_cvt_f16_f32_e32 v2, v2
; SI-NEXT: v_or_b32_e32 v1, v2, v1
; SI-NEXT: buffer_store_dword v1, v3, s[0:3], 0 offen
-; SI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:136 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:144 ; 4-byte Folded Reload
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_cvt_f16_f32_e32 v1, v31
; SI-NEXT: v_add_i32_e32 v3, vcc, 48, v0
@@ -32490,7 +32522,7 @@ define <60 x half> @bitcast_v15i64_to_v60f16(<15 x i64> %a, i32 %b) {
; SI-NEXT: v_cvt_f16_f32_e32 v2, v2
; SI-NEXT: v_or_b32_e32 v1, v2, v1
; SI-NEXT: buffer_store_dword v1, v3, s[0:3], 0 offen
-; SI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:132 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:140 ; 4-byte Folded Reload
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_cvt_f16_f32_e32 v1, v62
; SI-NEXT: v_add_i32_e32 v3, vcc, 52, v0
@@ -32499,7 +32531,7 @@ define <60 x half> @bitcast_v15i64_to_v60f16(<15 x i64> %a, i32 %b) {
; SI-NEXT: v_cvt_f16_f32_e32 v2, v2
; SI-NEXT: v_or_b32_e32 v1, v2, v1
; SI-NEXT: buffer_store_dword v1, v3, s[0:3], 0 offen
-; SI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:128 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:136 ; 4-byte Folded Reload
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_cvt_f16_f32_e32 v1, v60
; SI-NEXT: v_add_i32_e32 v3, vcc, 56, v0
@@ -32508,7 +32540,7 @@ define <60 x half> @bitcast_v15i64_to_v60f16(<15 x i64> %a, i32 %b) {
; SI-NEXT: v_cvt_f16_f32_e32 v2, v2
; SI-NEXT: v_or_b32_e32 v1, v2, v1
; SI-NEXT: buffer_store_dword v1, v3, s[0:3], 0 offen
-; SI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:124 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:132 ; 4-byte Folded Reload
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_cvt_f16_f32_e32 v1, v58
; SI-NEXT: v_add_i32_e32 v3, vcc, 60, v0
@@ -32518,8 +32550,8 @@ define <60 x half> @bitcast_v15i64_to_v60f16(<15 x i64> %a, i32 %b) {
; SI-NEXT: v_or_b32_e32 v1, v2, v1
; SI-NEXT: buffer_store_dword v1, v3, s[0:3], 0 offen
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:112 ; 4-byte Folded Reload
-; SI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:120 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:120 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:128 ; 4-byte Folded Reload
; SI-NEXT: v_add_i32_e32 v3, vcc, 64, v0
; SI-NEXT: s_waitcnt vmcnt(1)
; SI-NEXT: v_cvt_f16_f32_e32 v1, v1
@@ -32529,8 +32561,8 @@ define <60 x half> @bitcast_v15i64_to_v60f16(<15 x i64> %a, i32 %b) {
; SI-NEXT: v_or_b32_e32 v1, v2, v1
; SI-NEXT: buffer_store_dword v1, v3, s[0:3], 0 offen
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:104 ; 4-byte Folded Reload
-; SI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:116 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:112 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:124 ; 4-byte Folded Reload
; SI-NEXT: v_add_i32_e32 v3, vcc, 0x44, v0
; SI-NEXT: s_waitcnt vmcnt(1)
; SI-NEXT: v_cvt_f16_f32_e32 v1, v1
@@ -32540,8 +32572,8 @@ define <60 x half> @bitcast_v15i64_to_v60f16(<15 x i64> %a, i32 %b) {
; SI-NEXT: v_or_b32_e32 v1, v2, v1
; SI-NEXT: buffer_store_dword v1, v3, s[0:3], 0 offen
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:96 ; 4-byte Folded Reload
-; SI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:108 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:104 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:116 ; 4-byte Folded Reload
; SI-NEXT: v_add_i32_e32 v3, vcc, 0x48, v0
; SI-NEXT: s_waitcnt vmcnt(1)
; SI-NEXT: v_cvt_f16_f32_e32 v1, v1
@@ -32551,8 +32583,8 @@ define <60 x half> @bitcast_v15i64_to_v60f16(<15 x i64> %a, i32 %b) {
; SI-NEXT: v_or_b32_e32 v1, v2, v1
; SI-NEXT: buffer_store_dword v1, v3, s[0:3], 0 offen
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:84 ; 4-byte Folded Reload
-; SI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:100 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:92 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:108 ; 4-byte Folded Reload
; SI-NEXT: v_add_i32_e32 v3, vcc, 0x4c, v0
; SI-NEXT: s_waitcnt vmcnt(1)
; SI-NEXT: v_cvt_f16_f32_e32 v1, v1
@@ -32562,8 +32594,8 @@ define <60 x half> @bitcast_v15i64_to_v60f16(<15 x i64> %a, i32 %b) {
; SI-NEXT: v_or_b32_e32 v1, v2, v1
; SI-NEXT: buffer_store_dword v1, v3, s[0:3], 0 offen
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:76 ; 4-byte Folded Reload
-; SI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:92 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:84 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:100 ; 4-byte Folded Reload
; SI-NEXT: v_add_i32_e32 v3, vcc, 0x50, v0
; SI-NEXT: s_waitcnt vmcnt(1)
; SI-NEXT: v_cvt_f16_f32_e32 v1, v1
@@ -32573,8 +32605,8 @@ define <60 x half> @bitcast_v15i64_to_v60f16(<15 x i64> %a, i32 %b) {
; SI-NEXT: v_or_b32_e32 v1, v2, v1
; SI-NEXT: buffer_store_dword v1, v3, s[0:3], 0 offen
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:68 ; 4-byte Folded Reload
-; SI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:88 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:76 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:96 ; 4-byte Folded Reload
; SI-NEXT: v_add_i32_e32 v3, vcc, 0x54, v0
; SI-NEXT: s_waitcnt vmcnt(1)
; SI-NEXT: v_cvt_f16_f32_e32 v1, v1
@@ -32583,62 +32615,66 @@ define <60 x half> @bitcast_v15i64_to_v60f16(<15 x i64> %a, i32 %b) {
; SI-NEXT: v_lshlrev_b32_e32 v1, 16, v1
; SI-NEXT: v_or_b32_e32 v1, v2, v1
; SI-NEXT: buffer_store_dword v1, v3, s[0:3], 0 offen
-; SI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:80 ; 4-byte Folded Reload
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_cvt_f16_f32_e32 v1, v61
+; SI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:68 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:88 ; 4-byte Folded Reload
; SI-NEXT: v_add_i32_e32 v3, vcc, 0x58, v0
-; SI-NEXT: v_lshlrev_b32_e32 v1, 16, v1
+; SI-NEXT: s_waitcnt vmcnt(1)
+; SI-NEXT: v_cvt_f16_f32_e32 v1, v1
; SI-NEXT: s_waitcnt vmcnt(0)
; SI-NEXT: v_cvt_f16_f32_e32 v2, v2
+; SI-NEXT: v_lshlrev_b32_e32 v1, 16, v1
; SI-NEXT: v_or_b32_e32 v1, v2, v1
; SI-NEXT: buffer_store_dword v1, v3, s[0:3], 0 offen
-; SI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:72 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:80 ; 4-byte Folded Reload
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_cvt_f16_f32_e32 v1, v39
+; SI-NEXT: v_cvt_f16_f32_e32 v1, v61
; SI-NEXT: v_add_i32_e32 v3, vcc, 0x5c, v0
; SI-NEXT: v_lshlrev_b32_e32 v1, 16, v1
; SI-NEXT: s_waitcnt vmcnt(0)
; SI-NEXT: v_cvt_f16_f32_e32 v2, v2
; SI-NEXT: v_or_b32_e32 v1, v2, v1
; SI-NEXT: buffer_store_dword v1, v3, s[0:3], 0 offen
+; SI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:72 ; 4-byte Folded Reload
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_cvt_f16_f32_e32 v1, v49
-; SI-NEXT: v_cvt_f16_f32_e32 v2, v63
; SI-NEXT: v_add_i32_e32 v3, vcc, 0x60, v0
; SI-NEXT: v_lshlrev_b32_e32 v1, 16, v1
+; SI-NEXT: s_waitcnt vmcnt(0)
+; SI-NEXT: v_cvt_f16_f32_e32 v2, v2
; SI-NEXT: v_or_b32_e32 v1, v2, v1
; SI-NEXT: buffer_store_dword v1, v3, s[0:3], 0 offen
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_cvt_f16_f32_e32 v1, v51
-; SI-NEXT: v_cvt_f16_f32_e32 v2, v59
+; SI-NEXT: v_cvt_f16_f32_e32 v2, v63
; SI-NEXT: v_add_i32_e32 v3, vcc, 0x64, v0
; SI-NEXT: v_lshlrev_b32_e32 v1, 16, v1
; SI-NEXT: v_or_b32_e32 v1, v2, v1
; SI-NEXT: buffer_store_dword v1, v3, s[0:3], 0 offen
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_cvt_f16_f32_e32 v1, v53
-; SI-NEXT: v_cvt_f16_f32_e32 v2, v37
+; SI-NEXT: v_cvt_f16_f32_e32 v2, v59
; SI-NEXT: v_add_i32_e32 v3, vcc, 0x68, v0
; SI-NEXT: v_lshlrev_b32_e32 v1, 16, v1
; SI-NEXT: v_or_b32_e32 v1, v2, v1
; SI-NEXT: buffer_store_dword v1, v3, s[0:3], 0 offen
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_cvt_f16_f32_e32 v1, v55
-; SI-NEXT: v_cvt_f16_f32_e32 v2, v35
+; SI-NEXT: v_cvt_f16_f32_e32 v2, v39
; SI-NEXT: v_add_i32_e32 v3, vcc, 0x6c, v0
; SI-NEXT: v_lshlrev_b32_e32 v1, 16, v1
; SI-NEXT: v_or_b32_e32 v1, v2, v1
; SI-NEXT: buffer_store_dword v1, v3, s[0:3], 0 offen
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_cvt_f16_f32_e32 v1, v41
-; SI-NEXT: v_cvt_f16_f32_e32 v2, v34
+; SI-NEXT: v_cvt_f16_f32_e32 v1, v54
+; SI-NEXT: v_cvt_f16_f32_e32 v2, v37
; SI-NEXT: v_add_i32_e32 v3, vcc, 0x70, v0
; SI-NEXT: v_lshlrev_b32_e32 v1, 16, v1
; SI-NEXT: v_or_b32_e32 v1, v2, v1
; SI-NEXT: buffer_store_dword v1, v3, s[0:3], 0 offen
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_cvt_f16_f32_e32 v1, v42
-; SI-NEXT: v_cvt_f16_f32_e32 v2, v32
+; SI-NEXT: v_cvt_f16_f32_e32 v1, v41
+; SI-NEXT: v_cvt_f16_f32_e32 v2, v35
; SI-NEXT: v_add_i32_e32 v0, vcc, 0x74, v0
; SI-NEXT: v_lshlrev_b32_e32 v1, 16, v1
; SI-NEXT: v_or_b32_e32 v1, v2, v1
@@ -34398,7 +34434,7 @@ define inreg <60 x half> @bitcast_v15i64_to_v60f16_scalar(<15 x i64> inreg %a, i
; GFX11-TRUE16-NEXT: v_readfirstlane_b32 s5, v9
; GFX11-TRUE16-NEXT: v_readfirstlane_b32 s6, v10
; GFX11-TRUE16-NEXT: v_readfirstlane_b32 s4, v11
-; GFX11-TRUE16-NEXT: s_mov_b32 s94, 0
+; GFX11-TRUE16-NEXT: s_mov_b32 s89, 0
; GFX11-TRUE16-NEXT: s_and_b32 s40, vcc_lo, exec_lo
; GFX11-TRUE16-NEXT: s_cbranch_scc0 .LBB45_4
; GFX11-TRUE16-NEXT: ; %bb.1: ; %cmp.false
@@ -34427,12 +34463,12 @@ define inreg <60 x half> @bitcast_v15i64_to_v60f16_scalar(<15 x i64> inreg %a, i
; GFX11-TRUE16-NEXT: s_lshr_b32 s78, s19, 16
; GFX11-TRUE16-NEXT: s_lshr_b32 s79, s18, 16
; GFX11-TRUE16-NEXT: s_lshr_b32 s88, s17, 16
-; GFX11-TRUE16-NEXT: s_lshr_b32 s89, s16, 16
-; GFX11-TRUE16-NEXT: s_lshr_b32 s90, s3, 16
-; GFX11-TRUE16-NEXT: s_lshr_b32 s91, s2, 16
-; GFX11-TRUE16-NEXT: s_lshr_b32 s92, s1, 16
-; GFX11-TRUE16-NEXT: s_lshr_b32 s93, s0, 16
-; GFX11-TRUE16-NEXT: s_and_not1_b32 vcc_lo, exec_lo, s94
+; GFX11-TRUE16-NEXT: s_lshr_b32 s90, s16, 16
+; GFX11-TRUE16-NEXT: s_lshr_b32 s91, s3, 16
+; GFX11-TRUE16-NEXT: s_lshr_b32 s92, s2, 16
+; GFX11-TRUE16-NEXT: s_lshr_b32 s93, s1, 16
+; GFX11-TRUE16-NEXT: s_lshr_b32 s94, s0, 16
+; GFX11-TRUE16-NEXT: s_and_not1_b32 vcc_lo, exec_lo, s89
; GFX11-TRUE16-NEXT: s_cbranch_vccnz .LBB45_3
; GFX11-TRUE16-NEXT: .LBB45_2: ; %cmp.true
; GFX11-TRUE16-NEXT: s_add_u32 s6, s6, 3
@@ -34490,19 +34526,22 @@ define inreg <60 x half> @bitcast_v15i64_to_v60f16_scalar(<15 x i64> inreg %a, i
; GFX11-TRUE16-NEXT: s_lshr_b32 s78, s19, 16
; GFX11-TRUE16-NEXT: s_lshr_b32 s79, s18, 16
; GFX11-TRUE16-NEXT: s_lshr_b32 s88, s17, 16
-; GFX11-TRUE16-NEXT: s_lshr_b32 s89, s16, 16
-; GFX11-TRUE16-NEXT: s_lshr_b32 s90, s3, 16
-; GFX11-TRUE16-NEXT: s_lshr_b32 s91, s2, 16
-; GFX11-TRUE16-NEXT: s_lshr_b32 s92, s1, 16
-; GFX11-TRUE16-NEXT: s_lshr_b32 s93, s0, 16
+; GFX11-TRUE16-NEXT: s_lshr_b32 s90, s16, 16
+; GFX11-TRUE16-NEXT: s_lshr_b32 s91, s3, 16
+; GFX11-TRUE16-NEXT: s_lshr_b32 s92, s2, 16
+; GFX11-TRUE16-NEXT: s_lshr_b32 s93, s1, 16
+; GFX11-TRUE16-NEXT: s_lshr_b32 s94, s0, 16
; GFX11-TRUE16-NEXT: .LBB45_3: ; %end
; GFX11-TRUE16-NEXT: s_delay_alu instid0(SALU_CYCLE_1)
-; GFX11-TRUE16-NEXT: s_pack_ll_b32_b16 s0, s0, s93
-; GFX11-TRUE16-NEXT: s_pack_ll_b32_b16 s1, s1, s92
-; GFX11-TRUE16-NEXT: s_pack_ll_b32_b16 s2, s2, s91
-; GFX11-TRUE16-NEXT: s_pack_ll_b32_b16 s3, s3, s90
-; GFX11-TRUE16-NEXT: s_pack_ll_b32_b16 s16, s16, s89
+; GFX11-TRUE16-NEXT: s_mov_b32 s89, s94
+; GFX11-TRUE16-NEXT: s_pack_ll_b32_b16 s1, s1, s93
+; GFX11-TRUE16-NEXT: s_pack_ll_b32_b16 s0, s0, s89
+; GFX11-TRUE16-NEXT: s_mov_b32 s89, s92
+; GFX11-TRUE16-NEXT: s_pack_ll_b32_b16 s3, s3, s91
+; GFX11-TRUE16-NEXT: s_pack_ll_b32_b16 s2, s2, s89
+; GFX11-TRUE16-NEXT: s_mov_b32 s89, s90
; GFX11-TRUE16-NEXT: s_pack_ll_b32_b16 s17, s17, s88
+; GFX11-TRUE16-NEXT: s_pack_ll_b32_b16 s16, s16, s89
; GFX11-TRUE16-NEXT: s_pack_ll_b32_b16 s18, s18, s79
; GFX11-TRUE16-NEXT: s_pack_ll_b32_b16 s19, s19, s78
; GFX11-TRUE16-NEXT: s_pack_ll_b32_b16 s20, s20, s77
@@ -34544,11 +34583,11 @@ define inreg <60 x half> @bitcast_v15i64_to_v60f16_scalar(<15 x i64> inreg %a, i
; GFX11-TRUE16-NEXT: v_dual_mov_b32 v28, s6 :: v_dual_mov_b32 v29, s4
; GFX11-TRUE16-NEXT: s_setpc_b64 s[30:31]
; GFX11-TRUE16-NEXT: .LBB45_4:
+; GFX11-TRUE16-NEXT: ; implicit-def: $sgpr94_lo16
; GFX11-TRUE16-NEXT: ; implicit-def: $sgpr93_lo16
; GFX11-TRUE16-NEXT: ; implicit-def: $sgpr92_lo16
; GFX11-TRUE16-NEXT: ; implicit-def: $sgpr91_lo16
; GFX11-TRUE16-NEXT: ; implicit-def: $sgpr90_lo16
-; GFX11-TRUE16-NEXT: ; implicit-def: $sgpr89_lo16
; GFX11-TRUE16-NEXT: ; implicit-def: $sgpr88_lo16
; GFX11-TRUE16-NEXT: ; implicit-def: $sgpr79_lo16
; GFX11-TRUE16-NEXT: ; implicit-def: $sgpr78_lo16
@@ -36440,9 +36479,9 @@ define inreg <15 x i64> @bitcast_v60f16_to_v15i64_scalar(<60 x half> inreg %a, i
; SI-NEXT: v_cvt_f16_f32_e32 v1, s19
; SI-NEXT: v_cvt_f16_f32_e32 v12, s18
; SI-NEXT: v_cvt_f16_f32_e32 v2, s21
-; SI-NEXT: v_cvt_f16_f32_e32 v11, s20
+; SI-NEXT: v_cvt_f16_f32_e32 v10, s20
; SI-NEXT: v_cvt_f16_f32_e32 v3, s23
-; SI-NEXT: v_cvt_f16_f32_e32 v10, s22
+; SI-NEXT: v_cvt_f16_f32_e32 v11, s22
; SI-NEXT: v_cvt_f16_f32_e32 v4, s25
; SI-NEXT: v_cvt_f16_f32_e32 v9, s24
; SI-NEXT: v_cvt_f16_f32_e32 v5, s27
@@ -36521,9 +36560,9 @@ define inreg <15 x i64> @bitcast_v60f16_to_v15i64_scalar(<60 x half> inreg %a, i
; SI-NEXT: ; %bb.1: ; %cmp.false
; SI-NEXT: buffer_load_dword v24, off, s[0:3], s32 offset:172 ; 4-byte Folded Reload
; SI-NEXT: buffer_load_dword v25, off, s[0:3], s32 offset:168 ; 4-byte Folded Reload
-; SI-NEXT: s_waitcnt expcnt(4)
-; SI-NEXT: v_lshlrev_b32_e32 v3, 16, v3
-; SI-NEXT: v_or_b32_e32 v3, v10, v3
+; SI-NEXT: s_waitcnt expcnt(5)
+; SI-NEXT: v_lshlrev_b32_e32 v2, 16, v2
+; SI-NEXT: v_or_b32_e32 v2, v10, v2
; SI-NEXT: s_waitcnt expcnt(3)
; SI-NEXT: v_lshlrev_b32_e32 v10, 16, v34
; SI-NEXT: v_mov_b32_e32 v33, v32
@@ -36551,13 +36590,13 @@ define inreg <15 x i64> @bitcast_v60f16_to_v15i64_scalar(<60 x half> inreg %a, i
; SI-NEXT: buffer_load_dword v54, off, s[0:3], s32 offset:128 ; 4-byte Folded Reload
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_lshlrev_b32_e32 v1, 16, v1
-; SI-NEXT: v_lshlrev_b32_e32 v2, 16, v2
+; SI-NEXT: v_lshlrev_b32_e32 v3, 16, v3
; SI-NEXT: v_lshlrev_b32_e32 v4, 16, v4
; SI-NEXT: v_lshlrev_b32_e32 v18, 16, v18
; SI-NEXT: v_lshlrev_b32_e32 v19, 16, v19
; SI-NEXT: v_lshlrev_b32_e32 v0, 16, v59
; SI-NEXT: v_or_b32_e32 v1, v12, v1
-; SI-NEXT: v_or_b32_e32 v2, v11, v2
+; SI-NEXT: v_or_b32_e32 v3, v11, v3
; SI-NEXT: v_or_b32_e32 v4, v9, v4
; SI-NEXT: v_lshlrev_b32_e32 v11, 16, v36
; SI-NEXT: v_lshlrev_b32_e32 v12, 16, v63
@@ -36722,7 +36761,7 @@ define inreg <15 x i64> @bitcast_v60f16_to_v15i64_scalar(<60 x half> inreg %a, i
; SI-NEXT: buffer_load_dword v28, off, s[0:3], s32 offset:152 ; 4-byte Folded Reload
; SI-NEXT: buffer_load_dword v29, off, s[0:3], s32 offset:144 ; 4-byte Folded Reload
; SI-NEXT: buffer_load_dword v31, off, s[0:3], s32 offset:128 ; 4-byte Folded Reload
-; SI-NEXT: buffer_load_dword v4, off, s[0:3], s32 offset:276 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v4, off, s[0:3], s32 offset:280 ; 4-byte Folded Reload
; SI-NEXT: buffer_load_dword v5, off, s[0:3], s32 offset:260 ; 4-byte Folded Reload
; SI-NEXT: buffer_load_dword v6, off, s[0:3], s32 offset:252 ; 4-byte Folded Reload
; SI-NEXT: buffer_load_dword v7, off, s[0:3], s32 offset:244 ; 4-byte Folded Reload
@@ -36744,7 +36783,7 @@ define inreg <15 x i64> @bitcast_v60f16_to_v15i64_scalar(<60 x half> inreg %a, i
; SI-NEXT: v_lshlrev_b32_e32 v2, 16, v2
; SI-NEXT: v_or_b32_e32 v1, v3, v2
; SI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:268 ; 4-byte Folded Reload
-; SI-NEXT: buffer_load_dword v3, off, s[0:3], s32 offset:280 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v3, off, s[0:3], s32 offset:276 ; 4-byte Folded Reload
; SI-NEXT: s_waitcnt vmcnt(14)
; SI-NEXT: v_cvt_f32_f16_e32 v27, v27
; SI-NEXT: s_waitcnt vmcnt(11)
@@ -40866,8 +40905,8 @@ define <15 x double> @bitcast_v60i16_to_v15f64(<60 x i16> %a, i32 %b) {
; GFX9-NEXT: v_lshrrev_b32_e32 v0, 16, v44
; GFX9-NEXT: v_lshrrev_b32_e32 v32, 16, v45
; GFX9-NEXT: v_lshrrev_b32_e32 v33, 16, v46
-; GFX9-NEXT: v_lshrrev_b32_e32 v63, 16, v47
-; GFX9-NEXT: v_lshrrev_b32_e32 v34, 16, v56
+; GFX9-NEXT: v_lshrrev_b32_e32 v34, 16, v47
+; GFX9-NEXT: v_lshrrev_b32_e32 v63, 16, v56
; GFX9-NEXT: v_lshrrev_b32_e32 v35, 16, v57
; GFX9-NEXT: v_lshrrev_b32_e32 v36, 16, v58
; GFX9-NEXT: v_lshrrev_b32_e32 v62, 16, v59
@@ -40902,8 +40941,8 @@ define <15 x double> @bitcast_v60i16_to_v15f64(<60 x i16> %a, i32 %b) {
; GFX9-NEXT: v_perm_b32 v2, v62, v59, s6
; GFX9-NEXT: v_perm_b32 v3, v36, v58, s6
; GFX9-NEXT: v_perm_b32 v4, v35, v57, s6
-; GFX9-NEXT: v_perm_b32 v5, v34, v56, s6
-; GFX9-NEXT: v_perm_b32 v6, v63, v47, s6
+; GFX9-NEXT: v_perm_b32 v5, v63, v56, s6
+; GFX9-NEXT: v_perm_b32 v6, v34, v47, s6
; GFX9-NEXT: v_perm_b32 v7, v33, v46, s6
; GFX9-NEXT: v_perm_b32 v8, v32, v45, s6
; GFX9-NEXT: v_perm_b32 v9, v9, v44, s6
@@ -40945,8 +40984,8 @@ define <15 x double> @bitcast_v60i16_to_v15f64(<60 x i16> %a, i32 %b) {
; GFX9-NEXT: ; implicit-def: $vgpr39
; GFX9-NEXT: ; implicit-def: $vgpr32
; GFX9-NEXT: ; implicit-def: $vgpr33
-; GFX9-NEXT: ; implicit-def: $vgpr63
; GFX9-NEXT: ; implicit-def: $vgpr34
+; GFX9-NEXT: ; implicit-def: $vgpr63
; GFX9-NEXT: ; implicit-def: $vgpr35
; GFX9-NEXT: ; implicit-def: $vgpr36
; GFX9-NEXT: ; implicit-def: $vgpr62
@@ -41060,8 +41099,8 @@ define <15 x double> @bitcast_v60i16_to_v15f64(<60 x i16> %a, i32 %b) {
; GFX9-NEXT: v_perm_b32 v2, v62, v59, s6
; GFX9-NEXT: v_perm_b32 v3, v36, v58, s6
; GFX9-NEXT: v_perm_b32 v4, v35, v57, s6
-; GFX9-NEXT: v_perm_b32 v5, v34, v56, s6
-; GFX9-NEXT: v_perm_b32 v6, v63, v47, s6
+; GFX9-NEXT: v_perm_b32 v5, v63, v56, s6
+; GFX9-NEXT: v_perm_b32 v6, v34, v47, s6
; GFX9-NEXT: v_perm_b32 v7, v33, v46, s6
; GFX9-NEXT: v_perm_b32 v8, v32, v45, s6
; GFX9-NEXT: v_pk_add_u16 v0, v0, 3 op_sel_hi:[1,0]
@@ -43915,29 +43954,27 @@ define inreg <60 x half> @bitcast_v15f64_to_v60f16_scalar(<15 x double> inreg %a
; SI-NEXT: s_cbranch_scc0 .LBB53_4
; SI-NEXT: ; %bb.1: ; %cmp.false
; SI-NEXT: s_lshr_b32 s44, s5, 16
-; SI-NEXT: s_waitcnt expcnt(4)
-; SI-NEXT: v_cvt_f32_f16_e32 v59, s44
-; SI-NEXT: s_lshr_b32 s44, s4, 16
; SI-NEXT: v_cvt_f32_f16_e32 v36, s44
+; SI-NEXT: s_lshr_b32 s44, s4, 16
+; SI-NEXT: v_cvt_f32_f16_e32 v33, s44
; SI-NEXT: s_lshr_b32 s44, s7, 16
-; SI-NEXT: v_cvt_f32_f16_e32 v13, s44
+; SI-NEXT: v_cvt_f32_f16_e32 v29, s44
; SI-NEXT: s_lshr_b32 s44, s6, 16
-; SI-NEXT: s_waitcnt expcnt(2)
-; SI-NEXT: v_cvt_f32_f16_e32 v61, s44
+; SI-NEXT: v_cvt_f32_f16_e32 v25, s44
; SI-NEXT: s_lshr_b32 s44, s9, 16
-; SI-NEXT: v_cvt_f32_f16_e32 v33, s44
+; SI-NEXT: v_cvt_f32_f16_e32 v18, s44
; SI-NEXT: s_lshr_b32 s44, s8, 16
-; SI-NEXT: v_cvt_f32_f16_e32 v48, s44
+; SI-NEXT: v_cvt_f32_f16_e32 v2, s44
; SI-NEXT: s_lshr_b32 s44, s11, 16
-; SI-NEXT: v_cvt_f32_f16_e32 v25, s44
+; SI-NEXT: v_cvt_f32_f16_e32 v5, s44
; SI-NEXT: s_lshr_b32 s44, s10, 16
; SI-NEXT: v_cvt_f32_f16_e32 v1, s44
; SI-NEXT: s_lshr_b32 s44, s13, 16
-; SI-NEXT: v_cvt_f32_f16_e32 v5, s44
+; SI-NEXT: v_cvt_f32_f16_e32 v48, s44
; SI-NEXT: s_lshr_b32 s44, s12, 16
; SI-NEXT: v_cvt_f32_f16_e32 v17, s44
; SI-NEXT: s_lshr_b32 s44, s15, 16
-; SI-NEXT: v_cvt_f32_f16_e32 v43, s44
+; SI-NEXT: v_cvt_f32_f16_e32 v45, s44
; SI-NEXT: s_lshr_b32 s44, s14, 16
; SI-NEXT: v_cvt_f32_f16_e32 v22, s44
; SI-NEXT: s_lshr_b32 s44, s41, 16
@@ -43973,21 +44010,24 @@ define inreg <60 x half> @bitcast_v15f64_to_v60f16_scalar(<15 x double> inreg %a
; SI-NEXT: s_lshr_b32 s44, s18, 16
; SI-NEXT: v_cvt_f32_f16_e32 v56, s44
; SI-NEXT: s_lshr_b32 s44, s17, 16
+; SI-NEXT: s_waitcnt expcnt(5)
; SI-NEXT: v_cvt_f32_f16_e32 v58, s44
; SI-NEXT: s_lshr_b32 s44, s16, 16
+; SI-NEXT: s_waitcnt expcnt(3)
; SI-NEXT: v_cvt_f32_f16_e32 v60, s44
-; SI-NEXT: v_cvt_f32_f16_e32 v14, s5
-; SI-NEXT: v_cvt_f32_f16_e32 v18, s4
+; SI-NEXT: v_cvt_f32_f16_e32 v13, s5
+; SI-NEXT: v_cvt_f32_f16_e32 v14, s4
; SI-NEXT: v_cvt_f32_f16_e32 v19, s7
-; SI-NEXT: v_cvt_f32_f16_e32 v29, s6
-; SI-NEXT: v_cvt_f32_f16_e32 v52, s9
-; SI-NEXT: v_cvt_f32_f16_e32 v40, s8
-; SI-NEXT: v_cvt_f32_f16_e32 v45, s11
-; SI-NEXT: v_cvt_f32_f16_e32 v47, s10
+; SI-NEXT: s_waitcnt expcnt(2)
+; SI-NEXT: v_cvt_f32_f16_e32 v61, s6
; SI-NEXT: s_waitcnt expcnt(1)
-; SI-NEXT: v_cvt_f32_f16_e32 v62, s13
+; SI-NEXT: v_cvt_f32_f16_e32 v62, s9
+; SI-NEXT: v_cvt_f32_f16_e32 v52, s8
+; SI-NEXT: v_cvt_f32_f16_e32 v40, s11
+; SI-NEXT: v_cvt_f32_f16_e32 v43, s10
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_cvt_f32_f16_e32 v63, s12
+; SI-NEXT: v_cvt_f32_f16_e32 v63, s13
+; SI-NEXT: v_cvt_f32_f16_e32 v21, s12
; SI-NEXT: v_cvt_f32_f16_e32 v23, s15
; SI-NEXT: v_cvt_f32_f16_e32 v4, s14
; SI-NEXT: v_cvt_f32_f16_e32 v27, s41
@@ -44005,9 +44045,9 @@ define inreg <60 x half> @bitcast_v15f64_to_v60f16_scalar(<15 x double> inreg %a
; SI-NEXT: v_cvt_f32_f16_e32 v42, s21
; SI-NEXT: v_cvt_f32_f16_e32 v8, s20
; SI-NEXT: v_cvt_f32_f16_e32 v20, s19
-; SI-NEXT: v_cvt_f32_f16_e32 v21, s18
+; SI-NEXT: v_cvt_f32_f16_e32 v47, s18
; SI-NEXT: v_cvt_f32_f16_e32 v57, s17
-; SI-NEXT: v_cvt_f32_f16_e32 v2, s16
+; SI-NEXT: v_cvt_f32_f16_e32 v59, s16
; SI-NEXT: s_cbranch_execnz .LBB53_3
; SI-NEXT: .LBB53_2: ; %cmp.true
; SI-NEXT: v_add_f64 v[1:2], s[16:17], 1.0
@@ -44028,104 +44068,107 @@ define inreg <60 x half> @bitcast_v15f64_to_v60f16_scalar(<15 x double> inreg %a
; SI-NEXT: buffer_store_dword v3, off, s[0:3], s32 offset:76 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_lshrrev_b32_e32 v3, 16, v42
-; SI-NEXT: v_add_f64 v[15:16], s[10:11], 1.0
; SI-NEXT: buffer_store_dword v3, off, s[0:3], s32 offset:80 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_lshrrev_b32_e32 v3, 16, v53
; SI-NEXT: v_add_f64 v[20:21], s[12:13], 1.0
+; SI-NEXT: v_add_f64 v[11:12], s[8:9], 1.0
; SI-NEXT: buffer_store_dword v3, off, s[0:3], s32 offset:84 ; 4-byte Folded Spill
-; SI-NEXT: v_lshrrev_b32_e32 v6, 16, v15
+; SI-NEXT: s_waitcnt expcnt(0)
+; SI-NEXT: v_lshrrev_b32_e32 v3, 16, v54
+; SI-NEXT: buffer_store_dword v3, off, s[0:3], s32 offset:88 ; 4-byte Folded Spill
+; SI-NEXT: v_lshrrev_b32_e32 v6, 16, v21
+; SI-NEXT: v_lshrrev_b32_e32 v48, 16, v11
; SI-NEXT: v_lshrrev_b32_e32 v17, 16, v20
-; SI-NEXT: v_lshrrev_b32_e32 v59, 16, v21
-; SI-NEXT: v_cvt_f32_f16_e32 v62, v21
-; SI-NEXT: v_cvt_f32_f16_e32 v63, v20
+; SI-NEXT: v_cvt_f32_f16_e32 v63, v21
+; SI-NEXT: v_cvt_f32_f16_e32 v21, v20
; SI-NEXT: v_cvt_f32_f16_e32 v20, v58
-; SI-NEXT: v_cvt_f32_f16_e32 v21, v57
+; SI-NEXT: v_cvt_f32_f16_e32 v47, v57
; SI-NEXT: v_lshrrev_b32_e32 v58, 16, v2
; SI-NEXT: v_cvt_f32_f16_e32 v57, v2
-; SI-NEXT: v_cvt_f32_f16_e32 v2, v1
-; SI-NEXT: v_cvt_f32_f16_e32 v1, v6
-; SI-NEXT: buffer_load_dword v6, off, s[0:3], s32 offset:84 ; 4-byte Folded Reload
-; SI-NEXT: v_add_f64 v[37:38], s[26:27], 1.0
-; SI-NEXT: v_cvt_f32_f16_e32 v47, v15
-; SI-NEXT: v_lshrrev_b32_e32 v55, 16, v38
-; SI-NEXT: v_lshrrev_b32_e32 v44, 16, v37
-; SI-NEXT: v_cvt_f32_f16_e32 v15, v37
-; SI-NEXT: v_cvt_f32_f16_e32 v37, v55
+; SI-NEXT: v_cvt_f32_f16_e32 v2, v48
+; SI-NEXT: v_cvt_f32_f16_e32 v48, v6
+; SI-NEXT: buffer_load_dword v6, off, s[0:3], s32 offset:88 ; 4-byte Folded Reload
; SI-NEXT: v_add_f64 v[7:8], s[6:7], 1.0
; SI-NEXT: v_add_f64 v[34:35], s[28:29], 1.0
-; SI-NEXT: v_lshrrev_b32_e32 v5, 16, v8
+; SI-NEXT: v_lshrrev_b32_e32 v25, 16, v7
+; SI-NEXT: v_cvt_f32_f16_e32 v61, v7
+; SI-NEXT: v_cvt_f32_f16_e32 v7, v53
+; SI-NEXT: s_waitcnt expcnt(0)
+; SI-NEXT: v_add_f64 v[3:4], s[4:5], 1.0
+; SI-NEXT: v_lshrrev_b32_e32 v55, 16, v34
+; SI-NEXT: v_lshrrev_b32_e32 v33, 16, v3
+; SI-NEXT: v_cvt_f32_f16_e32 v14, v3
+; SI-NEXT: v_cvt_f32_f16_e32 v3, v34
+; SI-NEXT: v_cvt_f32_f16_e32 v34, v55
+; SI-NEXT: v_lshrrev_b32_e32 v29, 16, v8
; SI-NEXT: v_cvt_f32_f16_e32 v19, v8
; SI-NEXT: v_cvt_f32_f16_e32 v8, v41
+; SI-NEXT: v_add_f64 v[37:38], s[26:27], 1.0
+; SI-NEXT: v_add_f64 v[15:16], s[10:11], 1.0
+; SI-NEXT: v_lshrrev_b32_e32 v44, 16, v38
+; SI-NEXT: v_lshrrev_b32_e32 v46, 16, v37
+; SI-NEXT: v_lshrrev_b32_e32 v45, 16, v15
+; SI-NEXT: v_cvt_f32_f16_e32 v43, v15
+; SI-NEXT: v_cvt_f32_f16_e32 v15, v37
+; SI-NEXT: v_cvt_f32_f16_e32 v37, v44
; SI-NEXT: v_add_f64 v[30:31], s[42:43], 1.0
-; SI-NEXT: v_add_f64 v[11:12], s[8:9], 1.0
-; SI-NEXT: v_lshrrev_b32_e32 v39, 16, v35
-; SI-NEXT: v_lshrrev_b32_e32 v32, 16, v30
-; SI-NEXT: v_lshrrev_b32_e32 v33, 16, v12
-; SI-NEXT: v_cvt_f32_f16_e32 v52, v12
+; SI-NEXT: v_lshrrev_b32_e32 v5, 16, v12
+; SI-NEXT: v_lshrrev_b32_e32 v39, 16, v30
+; SI-NEXT: v_cvt_f32_f16_e32 v62, v12
; SI-NEXT: v_cvt_f32_f16_e32 v12, v30
-; SI-NEXT: v_cvt_f32_f16_e32 v30, v32
-; SI-NEXT: v_cvt_f32_f16_e32 v32, v39
-; SI-NEXT: v_cvt_f32_f16_e32 v39, v44
+; SI-NEXT: v_cvt_f32_f16_e32 v30, v39
+; SI-NEXT: v_cvt_f32_f16_e32 v39, v46
; SI-NEXT: v_add_f64 v[49:50], s[24:25], 1.0
-; SI-NEXT: v_lshrrev_b32_e32 v25, 16, v16
-; SI-NEXT: v_lshrrev_b32_e32 v46, 16, v50
-; SI-NEXT: v_lshrrev_b32_e32 v56, 16, v49
-; SI-NEXT: v_cvt_f32_f16_e32 v45, v16
+; SI-NEXT: v_lshrrev_b32_e32 v18, 16, v16
+; SI-NEXT: v_lshrrev_b32_e32 v56, 16, v50
+; SI-NEXT: v_lshrrev_b32_e32 v60, 16, v49
+; SI-NEXT: v_cvt_f32_f16_e32 v40, v16
; SI-NEXT: v_cvt_f32_f16_e32 v16, v49
-; SI-NEXT: v_cvt_f32_f16_e32 v49, v46
-; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_add_f64 v[3:4], s[4:5], 1.0
-; SI-NEXT: v_lshrrev_b32_e32 v51, 16, v34
-; SI-NEXT: v_lshrrev_b32_e32 v36, 16, v3
-; SI-NEXT: v_cvt_f32_f16_e32 v18, v3
-; SI-NEXT: v_cvt_f32_f16_e32 v3, v34
-; SI-NEXT: v_cvt_f32_f16_e32 v34, v51
-; SI-NEXT: v_cvt_f32_f16_e32 v51, v56
+; SI-NEXT: v_cvt_f32_f16_e32 v49, v56
+; SI-NEXT: v_cvt_f32_f16_e32 v59, v1
+; SI-NEXT: v_cvt_f32_f16_e32 v1, v5
; SI-NEXT: v_add_f64 v[26:27], s[40:41], 1.0
; SI-NEXT: v_add_f64 v[22:23], s[14:15], 1.0
-; SI-NEXT: v_lshrrev_b32_e32 v48, 16, v4
-; SI-NEXT: v_lshrrev_b32_e32 v10, 16, v26
-; SI-NEXT: v_lshrrev_b32_e32 v13, 16, v22
-; SI-NEXT: v_lshrrev_b32_e32 v43, 16, v11
-; SI-NEXT: v_cvt_f32_f16_e32 v14, v4
-; SI-NEXT: v_cvt_f32_f16_e32 v40, v11
-; SI-NEXT: v_cvt_f32_f16_e32 v4, v22
-; SI-NEXT: v_cvt_f32_f16_e32 v11, v26
-; SI-NEXT: v_cvt_f32_f16_e32 v22, v48
-; SI-NEXT: v_cvt_f32_f16_e32 v26, v5
-; SI-NEXT: v_lshrrev_b32_e32 v60, 16, v54
-; SI-NEXT: v_lshrrev_b32_e32 v28, 16, v31
+; SI-NEXT: v_lshrrev_b32_e32 v51, 16, v35
+; SI-NEXT: v_lshrrev_b32_e32 v32, 16, v31
+; SI-NEXT: v_lshrrev_b32_e32 v28, 16, v26
; SI-NEXT: v_lshrrev_b32_e32 v24, 16, v27
+; SI-NEXT: v_lshrrev_b32_e32 v10, 16, v22
; SI-NEXT: v_lshrrev_b32_e32 v9, 16, v23
-; SI-NEXT: v_lshrrev_b32_e32 v61, 16, v7
-; SI-NEXT: v_cvt_f32_f16_e32 v29, v7
+; SI-NEXT: v_lshrrev_b32_e32 v36, 16, v4
+; SI-NEXT: v_cvt_f32_f16_e32 v13, v4
+; SI-NEXT: v_cvt_f32_f16_e32 v52, v11
; SI-NEXT: v_cvt_f32_f16_e32 v23, v23
+; SI-NEXT: v_cvt_f32_f16_e32 v4, v22
; SI-NEXT: v_cvt_f32_f16_e32 v27, v27
+; SI-NEXT: v_cvt_f32_f16_e32 v11, v26
; SI-NEXT: v_cvt_f32_f16_e32 v31, v31
; SI-NEXT: v_cvt_f32_f16_e32 v35, v35
; SI-NEXT: v_cvt_f32_f16_e32 v38, v38
; SI-NEXT: v_cvt_f32_f16_e32 v50, v50
; SI-NEXT: v_cvt_f32_f16_e32 v54, v54
-; SI-NEXT: v_cvt_f32_f16_e32 v7, v53
; SI-NEXT: v_cvt_f32_f16_e32 v42, v42
; SI-NEXT: v_cvt_f32_f16_e32 v36, v36
-; SI-NEXT: v_cvt_f32_f16_e32 v61, v61
; SI-NEXT: v_cvt_f32_f16_e32 v33, v33
-; SI-NEXT: v_cvt_f32_f16_e32 v48, v43
+; SI-NEXT: v_cvt_f32_f16_e32 v29, v29
; SI-NEXT: v_cvt_f32_f16_e32 v25, v25
-; SI-NEXT: v_cvt_f32_f16_e32 v5, v59
-; SI-NEXT: v_mov_b32_e32 v59, v22
+; SI-NEXT: v_cvt_f32_f16_e32 v5, v18
+; SI-NEXT: v_mov_b32_e32 v18, v1
+; SI-NEXT: v_cvt_f32_f16_e32 v1, v45
; SI-NEXT: v_cvt_f32_f16_e32 v17, v17
-; SI-NEXT: v_cvt_f32_f16_e32 v43, v9
-; SI-NEXT: v_cvt_f32_f16_e32 v22, v13
-; SI-NEXT: v_mov_b32_e32 v13, v26
+; SI-NEXT: v_cvt_f32_f16_e32 v45, v9
+; SI-NEXT: v_cvt_f32_f16_e32 v22, v10
; SI-NEXT: v_cvt_f32_f16_e32 v24, v24
-; SI-NEXT: v_cvt_f32_f16_e32 v26, v10
-; SI-NEXT: v_cvt_f32_f16_e32 v28, v28
-; SI-NEXT: v_cvt_f32_f16_e32 v53, v60
+; SI-NEXT: v_cvt_f32_f16_e32 v26, v28
+; SI-NEXT: v_cvt_f32_f16_e32 v28, v32
+; SI-NEXT: v_cvt_f32_f16_e32 v32, v51
+; SI-NEXT: v_cvt_f32_f16_e32 v51, v60
; SI-NEXT: v_cvt_f32_f16_e32 v58, v58
; SI-NEXT: s_waitcnt vmcnt(0)
+; SI-NEXT: v_cvt_f32_f16_e32 v53, v6
+; SI-NEXT: buffer_load_dword v6, off, s[0:3], s32 offset:84 ; 4-byte Folded Reload
+; SI-NEXT: s_waitcnt vmcnt(0)
; SI-NEXT: v_cvt_f32_f16_e32 v55, v6
; SI-NEXT: buffer_load_dword v6, off, s[0:3], s32 offset:80 ; 4-byte Folded Reload
; SI-NEXT: s_waitcnt vmcnt(0)
@@ -44144,9 +44187,9 @@ define inreg <60 x half> @bitcast_v15f64_to_v60f16_scalar(<15 x double> inreg %a
; SI-NEXT: v_cvt_f32_f16_e32 v60, v6
; SI-NEXT: .LBB53_3: ; %end
; SI-NEXT: v_cvt_f16_f32_e32 v6, v60
-; SI-NEXT: v_cvt_f16_f32_e32 v9, v2
+; SI-NEXT: v_cvt_f16_f32_e32 v9, v59
; SI-NEXT: v_cvt_f16_f32_e32 v10, v57
-; SI-NEXT: v_cvt_f16_f32_e32 v3, v3
+; SI-NEXT: v_cvt_f16_f32_e32 v4, v4
; SI-NEXT: v_lshlrev_b32_e32 v6, 16, v6
; SI-NEXT: v_or_b32_e32 v6, v9, v6
; SI-NEXT: v_cvt_f16_f32_e32 v9, v58
@@ -44158,7 +44201,7 @@ define inreg <60 x half> @bitcast_v15f64_to_v60f16_scalar(<15 x double> inreg %a
; SI-NEXT: buffer_store_dword v9, v6, s[0:3], 0 offen
; SI-NEXT: v_cvt_f16_f32_e32 v6, v56
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_cvt_f16_f32_e32 v9, v21
+; SI-NEXT: v_cvt_f16_f32_e32 v9, v47
; SI-NEXT: v_add_i32_e32 v10, vcc, 8, v0
; SI-NEXT: v_lshlrev_b32_e32 v6, 16, v6
; SI-NEXT: v_or_b32_e32 v6, v9, v6
@@ -44228,125 +44271,125 @@ define inreg <60 x half> @bitcast_v15f64_to_v60f16_scalar(<15 x double> inreg %a
; SI-NEXT: buffer_store_dword v6, v10, s[0:3], 0 offen
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_cvt_f16_f32_e32 v6, v34
-; SI-NEXT: v_add_i32_e32 v9, vcc, 48, v0
-; SI-NEXT: v_cvt_f16_f32_e32 v4, v4
+; SI-NEXT: v_cvt_f16_f32_e32 v9, v3
+; SI-NEXT: v_add_i32_e32 v10, vcc, 48, v0
; SI-NEXT: v_lshlrev_b32_e32 v6, 16, v6
-; SI-NEXT: v_or_b32_e32 v3, v3, v6
-; SI-NEXT: buffer_store_dword v3, v9, s[0:3], 0 offen
+; SI-NEXT: v_or_b32_e32 v6, v9, v6
+; SI-NEXT: buffer_store_dword v6, v10, s[0:3], 0 offen
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_cvt_f16_f32_e32 v3, v32
-; SI-NEXT: v_cvt_f16_f32_e32 v6, v35
-; SI-NEXT: v_add_i32_e32 v9, vcc, 52, v0
-; SI-NEXT: v_lshlrev_b32_e32 v3, 16, v3
-; SI-NEXT: v_or_b32_e32 v3, v6, v3
-; SI-NEXT: buffer_store_dword v3, v9, s[0:3], 0 offen
+; SI-NEXT: v_cvt_f16_f32_e32 v6, v32
+; SI-NEXT: v_cvt_f16_f32_e32 v9, v35
+; SI-NEXT: v_add_i32_e32 v10, vcc, 52, v0
+; SI-NEXT: v_lshlrev_b32_e32 v6, 16, v6
+; SI-NEXT: v_or_b32_e32 v6, v9, v6
+; SI-NEXT: buffer_store_dword v6, v10, s[0:3], 0 offen
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_cvt_f16_f32_e32 v3, v30
-; SI-NEXT: v_cvt_f16_f32_e32 v6, v12
-; SI-NEXT: v_add_i32_e32 v9, vcc, 56, v0
-; SI-NEXT: v_lshlrev_b32_e32 v3, 16, v3
-; SI-NEXT: v_or_b32_e32 v3, v6, v3
-; SI-NEXT: buffer_store_dword v3, v9, s[0:3], 0 offen
+; SI-NEXT: v_cvt_f16_f32_e32 v6, v30
+; SI-NEXT: v_cvt_f16_f32_e32 v9, v12
+; SI-NEXT: v_add_i32_e32 v10, vcc, 56, v0
+; SI-NEXT: v_lshlrev_b32_e32 v6, 16, v6
+; SI-NEXT: v_or_b32_e32 v6, v9, v6
+; SI-NEXT: buffer_store_dword v6, v10, s[0:3], 0 offen
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_cvt_f16_f32_e32 v3, v28
-; SI-NEXT: v_cvt_f16_f32_e32 v6, v31
-; SI-NEXT: v_add_i32_e32 v9, vcc, 60, v0
-; SI-NEXT: v_lshlrev_b32_e32 v3, 16, v3
-; SI-NEXT: v_or_b32_e32 v3, v6, v3
-; SI-NEXT: buffer_store_dword v3, v9, s[0:3], 0 offen
+; SI-NEXT: v_cvt_f16_f32_e32 v6, v28
+; SI-NEXT: v_cvt_f16_f32_e32 v9, v31
+; SI-NEXT: v_add_i32_e32 v10, vcc, 60, v0
+; SI-NEXT: v_lshlrev_b32_e32 v6, 16, v6
+; SI-NEXT: v_or_b32_e32 v6, v9, v6
+; SI-NEXT: buffer_store_dword v6, v10, s[0:3], 0 offen
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_cvt_f16_f32_e32 v3, v26
-; SI-NEXT: v_cvt_f16_f32_e32 v6, v11
-; SI-NEXT: v_add_i32_e32 v9, vcc, 64, v0
-; SI-NEXT: v_lshlrev_b32_e32 v3, 16, v3
-; SI-NEXT: v_or_b32_e32 v3, v6, v3
-; SI-NEXT: buffer_store_dword v3, v9, s[0:3], 0 offen
+; SI-NEXT: v_cvt_f16_f32_e32 v6, v26
+; SI-NEXT: v_cvt_f16_f32_e32 v9, v11
+; SI-NEXT: v_add_i32_e32 v10, vcc, 64, v0
+; SI-NEXT: v_lshlrev_b32_e32 v6, 16, v6
+; SI-NEXT: v_or_b32_e32 v6, v9, v6
+; SI-NEXT: buffer_store_dword v6, v10, s[0:3], 0 offen
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_cvt_f16_f32_e32 v3, v24
-; SI-NEXT: v_cvt_f16_f32_e32 v6, v27
-; SI-NEXT: v_add_i32_e32 v9, vcc, 0x44, v0
-; SI-NEXT: v_lshlrev_b32_e32 v3, 16, v3
-; SI-NEXT: v_or_b32_e32 v3, v6, v3
-; SI-NEXT: buffer_store_dword v3, v9, s[0:3], 0 offen
+; SI-NEXT: v_cvt_f16_f32_e32 v6, v24
+; SI-NEXT: v_cvt_f16_f32_e32 v9, v27
+; SI-NEXT: v_add_i32_e32 v10, vcc, 0x44, v0
+; SI-NEXT: v_lshlrev_b32_e32 v6, 16, v6
+; SI-NEXT: v_or_b32_e32 v6, v9, v6
+; SI-NEXT: buffer_store_dword v6, v10, s[0:3], 0 offen
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_cvt_f16_f32_e32 v3, v22
-; SI-NEXT: v_add_i32_e32 v6, vcc, 0x48, v0
-; SI-NEXT: v_cvt_f16_f32_e32 v1, v1
-; SI-NEXT: v_lshlrev_b32_e32 v3, 16, v3
-; SI-NEXT: v_or_b32_e32 v3, v4, v3
-; SI-NEXT: buffer_store_dword v3, v6, s[0:3], 0 offen
+; SI-NEXT: v_cvt_f16_f32_e32 v6, v22
+; SI-NEXT: v_add_i32_e32 v9, vcc, 0x48, v0
+; SI-NEXT: v_add_i32_e32 v8, vcc, 0x54, v0
+; SI-NEXT: v_lshlrev_b32_e32 v6, 16, v6
+; SI-NEXT: v_or_b32_e32 v4, v4, v6
+; SI-NEXT: buffer_store_dword v4, v9, s[0:3], 0 offen
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_cvt_f16_f32_e32 v3, v43
-; SI-NEXT: v_cvt_f16_f32_e32 v4, v23
-; SI-NEXT: v_add_i32_e32 v6, vcc, 0x4c, v0
-; SI-NEXT: v_lshlrev_b32_e32 v3, 16, v3
-; SI-NEXT: v_or_b32_e32 v3, v4, v3
-; SI-NEXT: buffer_store_dword v3, v6, s[0:3], 0 offen
+; SI-NEXT: v_cvt_f16_f32_e32 v4, v45
+; SI-NEXT: v_cvt_f16_f32_e32 v6, v23
+; SI-NEXT: v_add_i32_e32 v9, vcc, 0x4c, v0
+; SI-NEXT: v_lshlrev_b32_e32 v4, 16, v4
+; SI-NEXT: v_or_b32_e32 v4, v6, v4
+; SI-NEXT: buffer_store_dword v4, v9, s[0:3], 0 offen
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_cvt_f16_f32_e32 v3, v17
-; SI-NEXT: v_cvt_f16_f32_e32 v4, v63
-; SI-NEXT: v_add_i32_e32 v6, vcc, 0x50, v0
-; SI-NEXT: v_lshlrev_b32_e32 v3, 16, v3
-; SI-NEXT: v_or_b32_e32 v3, v4, v3
-; SI-NEXT: buffer_store_dword v3, v6, s[0:3], 0 offen
+; SI-NEXT: v_cvt_f16_f32_e32 v4, v17
+; SI-NEXT: v_cvt_f16_f32_e32 v6, v21
+; SI-NEXT: v_add_i32_e32 v9, vcc, 0x50, v0
+; SI-NEXT: v_lshlrev_b32_e32 v4, 16, v4
+; SI-NEXT: v_or_b32_e32 v4, v6, v4
+; SI-NEXT: buffer_store_dword v4, v9, s[0:3], 0 offen
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_cvt_f16_f32_e32 v3, v5
-; SI-NEXT: v_cvt_f16_f32_e32 v4, v62
-; SI-NEXT: v_add_i32_e32 v5, vcc, 0x54, v0
-; SI-NEXT: v_lshlrev_b32_e32 v3, 16, v3
-; SI-NEXT: v_or_b32_e32 v3, v4, v3
-; SI-NEXT: buffer_store_dword v3, v5, s[0:3], 0 offen
+; SI-NEXT: v_cvt_f16_f32_e32 v4, v48
+; SI-NEXT: v_cvt_f16_f32_e32 v6, v63
+; SI-NEXT: v_cvt_f16_f32_e32 v1, v1
+; SI-NEXT: v_add_i32_e32 v3, vcc, 0x60, v0
+; SI-NEXT: v_lshlrev_b32_e32 v4, 16, v4
+; SI-NEXT: v_or_b32_e32 v4, v6, v4
+; SI-NEXT: buffer_store_dword v4, v8, s[0:3], 0 offen
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_cvt_f16_f32_e32 v3, v47
+; SI-NEXT: v_cvt_f16_f32_e32 v4, v43
; SI-NEXT: v_lshlrev_b32_e32 v1, 16, v1
-; SI-NEXT: v_add_i32_e32 v4, vcc, 0x58, v0
-; SI-NEXT: v_or_b32_e32 v1, v3, v1
-; SI-NEXT: buffer_store_dword v1, v4, s[0:3], 0 offen
+; SI-NEXT: v_add_i32_e32 v6, vcc, 0x58, v0
+; SI-NEXT: v_or_b32_e32 v1, v4, v1
+; SI-NEXT: buffer_store_dword v1, v6, s[0:3], 0 offen
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_cvt_f16_f32_e32 v1, v25
-; SI-NEXT: v_cvt_f16_f32_e32 v3, v45
-; SI-NEXT: v_add_i32_e32 v4, vcc, 0x5c, v0
+; SI-NEXT: v_cvt_f16_f32_e32 v1, v5
+; SI-NEXT: v_cvt_f16_f32_e32 v4, v40
+; SI-NEXT: v_add_i32_e32 v5, vcc, 0x5c, v0
; SI-NEXT: v_lshlrev_b32_e32 v1, 16, v1
-; SI-NEXT: v_or_b32_e32 v1, v3, v1
-; SI-NEXT: buffer_store_dword v1, v4, s[0:3], 0 offen
+; SI-NEXT: v_or_b32_e32 v1, v4, v1
+; SI-NEXT: buffer_store_dword v1, v5, s[0:3], 0 offen
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_cvt_f16_f32_e32 v1, v48
-; SI-NEXT: v_cvt_f16_f32_e32 v2, v40
-; SI-NEXT: v_add_i32_e32 v3, vcc, 0x60, v0
+; SI-NEXT: v_cvt_f16_f32_e32 v1, v2
+; SI-NEXT: v_cvt_f16_f32_e32 v2, v52
; SI-NEXT: v_lshlrev_b32_e32 v1, 16, v1
; SI-NEXT: v_or_b32_e32 v1, v2, v1
; SI-NEXT: buffer_store_dword v1, v3, s[0:3], 0 offen
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_cvt_f16_f32_e32 v1, v33
-; SI-NEXT: v_cvt_f16_f32_e32 v2, v52
+; SI-NEXT: v_cvt_f16_f32_e32 v1, v18
+; SI-NEXT: v_cvt_f16_f32_e32 v2, v62
; SI-NEXT: v_add_i32_e32 v3, vcc, 0x64, v0
; SI-NEXT: v_lshlrev_b32_e32 v1, 16, v1
; SI-NEXT: v_or_b32_e32 v1, v2, v1
; SI-NEXT: buffer_store_dword v1, v3, s[0:3], 0 offen
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_cvt_f16_f32_e32 v1, v61
-; SI-NEXT: v_cvt_f16_f32_e32 v2, v29
+; SI-NEXT: v_cvt_f16_f32_e32 v1, v25
+; SI-NEXT: v_cvt_f16_f32_e32 v2, v61
; SI-NEXT: v_add_i32_e32 v3, vcc, 0x68, v0
; SI-NEXT: v_lshlrev_b32_e32 v1, 16, v1
; SI-NEXT: v_or_b32_e32 v1, v2, v1
; SI-NEXT: buffer_store_dword v1, v3, s[0:3], 0 offen
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_cvt_f16_f32_e32 v1, v13
+; SI-NEXT: v_cvt_f16_f32_e32 v1, v29
; SI-NEXT: v_cvt_f16_f32_e32 v2, v19
; SI-NEXT: v_add_i32_e32 v3, vcc, 0x6c, v0
; SI-NEXT: v_lshlrev_b32_e32 v1, 16, v1
; SI-NEXT: v_or_b32_e32 v1, v2, v1
; SI-NEXT: buffer_store_dword v1, v3, s[0:3], 0 offen
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_cvt_f16_f32_e32 v1, v36
-; SI-NEXT: v_cvt_f16_f32_e32 v2, v18
+; SI-NEXT: v_cvt_f16_f32_e32 v1, v33
+; SI-NEXT: v_cvt_f16_f32_e32 v2, v14
; SI-NEXT: v_add_i32_e32 v3, vcc, 0x70, v0
; SI-NEXT: v_lshlrev_b32_e32 v1, 16, v1
; SI-NEXT: v_or_b32_e32 v1, v2, v1
; SI-NEXT: buffer_store_dword v1, v3, s[0:3], 0 offen
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_cvt_f16_f32_e32 v1, v59
-; SI-NEXT: v_cvt_f16_f32_e32 v2, v14
+; SI-NEXT: v_cvt_f16_f32_e32 v1, v36
+; SI-NEXT: v_cvt_f16_f32_e32 v2, v13
; SI-NEXT: v_add_i32_e32 v0, vcc, 0x74, v0
; SI-NEXT: v_lshlrev_b32_e32 v1, 16, v1
; SI-NEXT: v_or_b32_e32 v1, v2, v1
@@ -44370,11 +44413,11 @@ define inreg <60 x half> @bitcast_v15f64_to_v60f16_scalar(<15 x double> inreg %a
; SI-NEXT: s_waitcnt vmcnt(0) expcnt(0)
; SI-NEXT: s_setpc_b64 s[30:31]
; SI-NEXT: .LBB53_4:
-; SI-NEXT: ; implicit-def: $vgpr2
+; SI-NEXT: ; implicit-def: $vgpr59
; SI-NEXT: ; implicit-def: $vgpr60
; SI-NEXT: ; implicit-def: $vgpr57
; SI-NEXT: ; implicit-def: $vgpr58
-; SI-NEXT: ; implicit-def: $vgpr21
+; SI-NEXT: ; implicit-def: $vgpr47
; SI-NEXT: ; implicit-def: $vgpr56
; SI-NEXT: ; implicit-def: $vgpr20
; SI-NEXT: ; implicit-def: $vgpr46
@@ -44409,27 +44452,27 @@ define inreg <60 x half> @bitcast_v15f64_to_v60f16_scalar(<15 x double> inreg %a
; SI-NEXT: ; implicit-def: $vgpr4
; SI-NEXT: ; implicit-def: $vgpr22
; SI-NEXT: ; implicit-def: $vgpr23
-; SI-NEXT: ; implicit-def: $vgpr43
-; SI-NEXT: ; implicit-def: $vgpr63
+; SI-NEXT: ; implicit-def: $vgpr45
+; SI-NEXT: ; implicit-def: $vgpr21
; SI-NEXT: ; implicit-def: $vgpr17
-; SI-NEXT: ; implicit-def: $vgpr62
-; SI-NEXT: ; implicit-def: $vgpr5
-; SI-NEXT: ; implicit-def: $vgpr47
+; SI-NEXT: ; implicit-def: $vgpr63
+; SI-NEXT: ; implicit-def: $vgpr48
+; SI-NEXT: ; implicit-def: $vgpr43
; SI-NEXT: ; implicit-def: $vgpr1
-; SI-NEXT: ; implicit-def: $vgpr45
-; SI-NEXT: ; implicit-def: $vgpr25
; SI-NEXT: ; implicit-def: $vgpr40
-; SI-NEXT: ; implicit-def: $vgpr48
+; SI-NEXT: ; implicit-def: $vgpr5
; SI-NEXT: ; implicit-def: $vgpr52
-; SI-NEXT: ; implicit-def: $vgpr33
-; SI-NEXT: ; implicit-def: $vgpr29
+; SI-NEXT: ; implicit-def: $vgpr2
+; SI-NEXT: ; implicit-def: $vgpr62
+; SI-NEXT: ; implicit-def: $vgpr18
; SI-NEXT: ; implicit-def: $vgpr61
+; SI-NEXT: ; implicit-def: $vgpr25
; SI-NEXT: ; implicit-def: $vgpr19
+; SI-NEXT: ; implicit-def: $vgpr29
+; SI-NEXT: ; implicit-def: $vgpr14
+; SI-NEXT: ; implicit-def: $vgpr33
; SI-NEXT: ; implicit-def: $vgpr13
-; SI-NEXT: ; implicit-def: $vgpr18
; SI-NEXT: ; implicit-def: $vgpr36
-; SI-NEXT: ; implicit-def: $vgpr14
-; SI-NEXT: ; implicit-def: $vgpr59
; SI-NEXT: s_branch .LBB53_2
;
; VI-LABEL: bitcast_v15f64_to_v60f16_scalar:
@@ -46965,9 +47008,9 @@ define inreg <15 x double> @bitcast_v60f16_to_v15f64_scalar(<60 x half> inreg %a
; SI-NEXT: v_cvt_f16_f32_e32 v1, s19
; SI-NEXT: v_cvt_f16_f32_e32 v12, s18
; SI-NEXT: v_cvt_f16_f32_e32 v2, s21
-; SI-NEXT: v_cvt_f16_f32_e32 v11, s20
+; SI-NEXT: v_cvt_f16_f32_e32 v10, s20
; SI-NEXT: v_cvt_f16_f32_e32 v3, s23
-; SI-NEXT: v_cvt_f16_f32_e32 v10, s22
+; SI-NEXT: v_cvt_f16_f32_e32 v11, s22
; SI-NEXT: v_cvt_f16_f32_e32 v4, s25
; SI-NEXT: v_cvt_f16_f32_e32 v9, s24
; SI-NEXT: v_cvt_f16_f32_e32 v5, s27
@@ -47046,9 +47089,9 @@ define inreg <15 x double> @bitcast_v60f16_to_v15f64_scalar(<60 x half> inreg %a
; SI-NEXT: ; %bb.1: ; %cmp.false
; SI-NEXT: buffer_load_dword v24, off, s[0:3], s32 offset:172 ; 4-byte Folded Reload
; SI-NEXT: buffer_load_dword v25, off, s[0:3], s32 offset:168 ; 4-byte Folded Reload
-; SI-NEXT: s_waitcnt expcnt(4)
-; SI-NEXT: v_lshlrev_b32_e32 v3, 16, v3
-; SI-NEXT: v_or_b32_e32 v3, v10, v3
+; SI-NEXT: s_waitcnt expcnt(5)
+; SI-NEXT: v_lshlrev_b32_e32 v2, 16, v2
+; SI-NEXT: v_or_b32_e32 v2, v10, v2
; SI-NEXT: s_waitcnt expcnt(3)
; SI-NEXT: v_lshlrev_b32_e32 v10, 16, v34
; SI-NEXT: v_mov_b32_e32 v33, v32
@@ -47076,13 +47119,13 @@ define inreg <15 x double> @bitcast_v60f16_to_v15f64_scalar(<60 x half> inreg %a
; SI-NEXT: buffer_load_dword v54, off, s[0:3], s32 offset:128 ; 4-byte Folded Reload
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_lshlrev_b32_e32 v1, 16, v1
-; SI-NEXT: v_lshlrev_b32_e32 v2, 16, v2
+; SI-NEXT: v_lshlrev_b32_e32 v3, 16, v3
; SI-NEXT: v_lshlrev_b32_e32 v4, 16, v4
; SI-NEXT: v_lshlrev_b32_e32 v18, 16, v18
; SI-NEXT: v_lshlrev_b32_e32 v19, 16, v19
; SI-NEXT: v_lshlrev_b32_e32 v0, 16, v59
; SI-NEXT: v_or_b32_e32 v1, v12, v1
-; SI-NEXT: v_or_b32_e32 v2, v11, v2
+; SI-NEXT: v_or_b32_e32 v3, v11, v3
; SI-NEXT: v_or_b32_e32 v4, v9, v4
; SI-NEXT: v_lshlrev_b32_e32 v11, 16, v36
; SI-NEXT: v_lshlrev_b32_e32 v12, 16, v63
@@ -47247,7 +47290,7 @@ define inreg <15 x double> @bitcast_v60f16_to_v15f64_scalar(<60 x half> inreg %a
; SI-NEXT: buffer_load_dword v28, off, s[0:3], s32 offset:152 ; 4-byte Folded Reload
; SI-NEXT: buffer_load_dword v29, off, s[0:3], s32 offset:144 ; 4-byte Folded Reload
; SI-NEXT: buffer_load_dword v31, off, s[0:3], s32 offset:128 ; 4-byte Folded Reload
-; SI-NEXT: buffer_load_dword v4, off, s[0:3], s32 offset:276 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v4, off, s[0:3], s32 offset:280 ; 4-byte Folded Reload
; SI-NEXT: buffer_load_dword v5, off, s[0:3], s32 offset:260 ; 4-byte Folded Reload
; SI-NEXT: buffer_load_dword v6, off, s[0:3], s32 offset:252 ; 4-byte Folded Reload
; SI-NEXT: buffer_load_dword v7, off, s[0:3], s32 offset:244 ; 4-byte Folded Reload
@@ -47269,7 +47312,7 @@ define inreg <15 x double> @bitcast_v60f16_to_v15f64_scalar(<60 x half> inreg %a
; SI-NEXT: v_lshlrev_b32_e32 v2, 16, v2
; SI-NEXT: v_or_b32_e32 v1, v3, v2
; SI-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:268 ; 4-byte Folded Reload
-; SI-NEXT: buffer_load_dword v3, off, s[0:3], s32 offset:280 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v3, off, s[0:3], s32 offset:276 ; 4-byte Folded Reload
; SI-NEXT: s_waitcnt vmcnt(14)
; SI-NEXT: v_cvt_f32_f16_e32 v27, v27
; SI-NEXT: s_waitcnt vmcnt(11)
@@ -50025,7 +50068,7 @@ define inreg <60 x half> @bitcast_v60i16_to_v60f16_scalar(<60 x i16> inreg %a, i
; SI-NEXT: buffer_store_dword v29, off, s[0:3], s32 offset:416 ; 4-byte Folded Spill
; SI-NEXT: buffer_store_dword v30, off, s[0:3], s32 offset:420 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt vmcnt(14)
-; SI-NEXT: buffer_store_dword v35, off, s[0:3], s32 offset:424 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v48, off, s[0:3], s32 offset:424 ; 4-byte Folded Spill
; SI-NEXT: v_cmp_ne_u32_e32 vcc, 0, v49
; SI-NEXT: s_and_b64 s[4:5], vcc, exec
; SI-NEXT: s_cbranch_scc0 .LBB57_2
@@ -50047,118 +50090,123 @@ define inreg <60 x half> @bitcast_v60i16_to_v60f16_scalar(<60 x i16> inreg %a, i
; SI-NEXT: buffer_store_dword v49, off, s[0:3], s32 offset:188 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_cvt_f32_f16_e32 v49, s18
-; SI-NEXT: v_cvt_f32_f16_e32 v57, v39
+; SI-NEXT: v_mov_b32_e32 v48, v39
; SI-NEXT: buffer_store_dword v31, off, s[0:3], s32 offset:280 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_cvt_f32_f16_e32 v31, v4
; SI-NEXT: buffer_store_dword v49, off, s[0:3], s32 offset:164 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_cvt_f32_f16_e32 v49, s19
-; SI-NEXT: v_cvt_f32_f16_e32 v44, v38
+; SI-NEXT: v_cvt_f32_f16_e32 v57, v39
; SI-NEXT: buffer_store_dword v31, off, s[0:3], s32 offset:296 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_cvt_f32_f16_e32 v31, v5
; SI-NEXT: buffer_store_dword v49, off, s[0:3], s32 offset:184 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_cvt_f32_f16_e32 v49, s20
-; SI-NEXT: v_cvt_f32_f16_e32 v47, v37
+; SI-NEXT: v_mov_b32_e32 v39, v38
; SI-NEXT: buffer_store_dword v31, off, s[0:3], s32 offset:272 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_cvt_f32_f16_e32 v31, v6
; SI-NEXT: buffer_store_dword v49, off, s[0:3], s32 offset:156 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_cvt_f32_f16_e32 v49, s21
-; SI-NEXT: v_cvt_f32_f16_e32 v42, v36
+; SI-NEXT: v_cvt_f32_f16_e32 v44, v38
; SI-NEXT: buffer_store_dword v31, off, s[0:3], s32 offset:292 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_cvt_f32_f16_e32 v31, v7
; SI-NEXT: buffer_store_dword v49, off, s[0:3], s32 offset:176 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_cvt_f32_f16_e32 v49, s22
-; SI-NEXT: v_cvt_f32_f16_e32 v45, v35
+; SI-NEXT: v_mov_b32_e32 v38, v37
; SI-NEXT: buffer_store_dword v31, off, s[0:3], s32 offset:264 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_cvt_f32_f16_e32 v31, v8
; SI-NEXT: buffer_store_dword v49, off, s[0:3], s32 offset:148 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_cvt_f32_f16_e32 v49, s23
-; SI-NEXT: v_mov_b32_e32 v35, v34
+; SI-NEXT: v_cvt_f32_f16_e32 v47, v37
; SI-NEXT: buffer_store_dword v31, off, s[0:3], s32 offset:284 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_cvt_f32_f16_e32 v31, v9
; SI-NEXT: buffer_store_dword v49, off, s[0:3], s32 offset:168 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_cvt_f32_f16_e32 v49, s24
-; SI-NEXT: v_cvt_f32_f16_e32 v40, v34
+; SI-NEXT: v_mov_b32_e32 v37, v36
; SI-NEXT: buffer_store_dword v31, off, s[0:3], s32 offset:256 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_cvt_f32_f16_e32 v31, v10
; SI-NEXT: buffer_store_dword v49, off, s[0:3], s32 offset:140 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_cvt_f32_f16_e32 v49, s25
-; SI-NEXT: v_mov_b32_e32 v34, v33
+; SI-NEXT: v_cvt_f32_f16_e32 v42, v36
; SI-NEXT: buffer_store_dword v31, off, s[0:3], s32 offset:276 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_cvt_f32_f16_e32 v31, v11
; SI-NEXT: buffer_store_dword v49, off, s[0:3], s32 offset:160 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_cvt_f32_f16_e32 v49, s26
-; SI-NEXT: v_cvt_f32_f16_e32 v43, v33
+; SI-NEXT: v_mov_b32_e32 v36, v35
; SI-NEXT: buffer_store_dword v31, off, s[0:3], s32 offset:248 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_cvt_f32_f16_e32 v31, v12
; SI-NEXT: buffer_store_dword v49, off, s[0:3], s32 offset:136 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_cvt_f32_f16_e32 v49, s27
-; SI-NEXT: v_mov_b32_e32 v33, v32
+; SI-NEXT: v_cvt_f32_f16_e32 v45, v35
; SI-NEXT: buffer_store_dword v31, off, s[0:3], s32 offset:268 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_cvt_f32_f16_e32 v31, v13
; SI-NEXT: buffer_store_dword v49, off, s[0:3], s32 offset:152 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_cvt_f32_f16_e32 v49, s28
-; SI-NEXT: v_cvt_f32_f16_e32 v54, v32
+; SI-NEXT: v_mov_b32_e32 v35, v34
; SI-NEXT: buffer_store_dword v31, off, s[0:3], s32 offset:240 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_cvt_f32_f16_e32 v31, v14
; SI-NEXT: buffer_store_dword v49, off, s[0:3], s32 offset:132 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_cvt_f32_f16_e32 v49, s29
-; SI-NEXT: v_mov_b32_e32 v32, v50
+; SI-NEXT: v_cvt_f32_f16_e32 v40, v34
; SI-NEXT: buffer_store_dword v31, off, s[0:3], s32 offset:260 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_cvt_f32_f16_e32 v31, v15
; SI-NEXT: buffer_store_dword v49, off, s[0:3], s32 offset:144 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_cvt_f32_f16_e32 v49, v28
-; SI-NEXT: v_cvt_f32_f16_e32 v41, v50
+; SI-NEXT: v_mov_b32_e32 v34, v33
; SI-NEXT: buffer_store_dword v31, off, s[0:3], s32 offset:232 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_cvt_f32_f16_e32 v31, v16
; SI-NEXT: buffer_store_dword v49, off, s[0:3], s32 offset:204 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_cvt_f32_f16_e32 v49, v30
-; SI-NEXT: v_cvt_f32_f16_e32 v52, v63
+; SI-NEXT: v_cvt_f32_f16_e32 v43, v33
; SI-NEXT: buffer_store_dword v31, off, s[0:3], s32 offset:252 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_cvt_f32_f16_e32 v31, v17
; SI-NEXT: buffer_store_dword v49, off, s[0:3], s32 offset:196 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_cvt_f32_f16_e32 v49, v59
-; SI-NEXT: v_cvt_f32_f16_e32 v55, v62
+; SI-NEXT: v_mov_b32_e32 v33, v32
; SI-NEXT: buffer_store_dword v31, off, s[0:3], s32 offset:224 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_cvt_f32_f16_e32 v31, v18
-; SI-NEXT: v_cvt_f32_f16_e32 v50, v61
-; SI-NEXT: v_cvt_f32_f16_e32 v53, v60
-; SI-NEXT: buffer_store_dword v49, off, s[0:3], s32 offset:180 ; 4-byte Folded Spill
+; SI-NEXT: v_cvt_f32_f16_e32 v54, v32
+; SI-NEXT: v_mov_b32_e32 v32, v50
+; SI-NEXT: v_cvt_f32_f16_e32 v41, v50
; SI-NEXT: buffer_store_dword v31, off, s[0:3], s32 offset:244 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_cvt_f32_f16_e32 v31, v19
-; SI-NEXT: v_cvt_f32_f16_e32 v51, v58
+; SI-NEXT: v_cvt_f32_f16_e32 v52, v63
+; SI-NEXT: v_cvt_f32_f16_e32 v55, v62
+; SI-NEXT: v_cvt_f32_f16_e32 v50, v61
; SI-NEXT: buffer_store_dword v31, off, s[0:3], s32 offset:216 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_cvt_f32_f16_e32 v31, v20
+; SI-NEXT: v_cvt_f32_f16_e32 v53, v60
+; SI-NEXT: buffer_store_dword v49, off, s[0:3], s32 offset:180 ; 4-byte Folded Spill
+; SI-NEXT: v_cvt_f32_f16_e32 v51, v58
; SI-NEXT: buffer_store_dword v31, off, s[0:3], s32 offset:236 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_cvt_f32_f16_e32 v31, v21
@@ -50187,6 +50235,31 @@ define inreg <60 x half> @bitcast_v60i16_to_v60f16_scalar(<60 x i16> inreg %a, i
; SI-NEXT: ; implicit-def: $vgpr49
; SI-NEXT: ; kill: killed $vgpr49
; SI-NEXT: s_waitcnt expcnt(0)
+; SI-NEXT: v_mov_b32_e32 v48, v39
+; SI-NEXT: ; implicit-def: $vgpr31
+; SI-NEXT: ; kill: killed $vgpr31
+; SI-NEXT: ; implicit-def: $vgpr49
+; SI-NEXT: ; kill: killed $vgpr49
+; SI-NEXT: v_mov_b32_e32 v39, v38
+; SI-NEXT: ; implicit-def: $vgpr31
+; SI-NEXT: ; kill: killed $vgpr31
+; SI-NEXT: ; implicit-def: $vgpr49
+; SI-NEXT: ; kill: killed $vgpr49
+; SI-NEXT: v_mov_b32_e32 v38, v37
+; SI-NEXT: ; implicit-def: $vgpr31
+; SI-NEXT: ; kill: killed $vgpr31
+; SI-NEXT: ; implicit-def: $vgpr49
+; SI-NEXT: ; kill: killed $vgpr49
+; SI-NEXT: v_mov_b32_e32 v37, v36
+; SI-NEXT: ; implicit-def: $vgpr31
+; SI-NEXT: ; kill: killed $vgpr31
+; SI-NEXT: ; implicit-def: $vgpr49
+; SI-NEXT: ; kill: killed $vgpr49
+; SI-NEXT: v_mov_b32_e32 v36, v35
+; SI-NEXT: ; implicit-def: $vgpr31
+; SI-NEXT: ; kill: killed $vgpr31
+; SI-NEXT: ; implicit-def: $vgpr49
+; SI-NEXT: ; kill: killed $vgpr49
; SI-NEXT: v_mov_b32_e32 v35, v34
; SI-NEXT: ; implicit-def: $vgpr31
; SI-NEXT: ; kill: killed $vgpr31
@@ -50254,26 +50327,6 @@ define inreg <60 x half> @bitcast_v60i16_to_v60f16_scalar(<60 x i16> inreg %a, i
; SI-NEXT: ; kill: killed $vgpr49
; SI-NEXT: ; implicit-def: $vgpr31
; SI-NEXT: ; kill: killed $vgpr31
-; SI-NEXT: ; implicit-def: $vgpr49
-; SI-NEXT: ; kill: killed $vgpr49
-; SI-NEXT: ; implicit-def: $vgpr31
-; SI-NEXT: ; kill: killed $vgpr31
-; SI-NEXT: ; implicit-def: $vgpr49
-; SI-NEXT: ; kill: killed $vgpr49
-; SI-NEXT: ; implicit-def: $vgpr31
-; SI-NEXT: ; kill: killed $vgpr31
-; SI-NEXT: ; implicit-def: $vgpr49
-; SI-NEXT: ; kill: killed $vgpr49
-; SI-NEXT: ; implicit-def: $vgpr31
-; SI-NEXT: ; kill: killed $vgpr31
-; SI-NEXT: ; implicit-def: $vgpr49
-; SI-NEXT: ; kill: killed $vgpr49
-; SI-NEXT: ; implicit-def: $vgpr31
-; SI-NEXT: ; kill: killed $vgpr31
-; SI-NEXT: ; implicit-def: $vgpr49
-; SI-NEXT: ; kill: killed $vgpr49
-; SI-NEXT: ; implicit-def: $vgpr31
-; SI-NEXT: ; kill: killed $vgpr31
; SI-NEXT: ; implicit-def: $vgpr31
; SI-NEXT: ; kill: killed $vgpr31
; SI-NEXT: ; implicit-def: $vgpr31
@@ -50309,6 +50362,11 @@ define inreg <60 x half> @bitcast_v60i16_to_v60f16_scalar(<60 x i16> inreg %a, i
; SI-NEXT: v_add_i32_e32 v32, vcc, 3, v33
; SI-NEXT: v_add_i32_e32 v33, vcc, 3, v34
; SI-NEXT: v_add_i32_e32 v34, vcc, 3, v35
+; SI-NEXT: v_add_i32_e32 v35, vcc, 3, v36
+; SI-NEXT: v_add_i32_e32 v36, vcc, 3, v37
+; SI-NEXT: v_add_i32_e32 v37, vcc, 3, v38
+; SI-NEXT: v_add_i32_e32 v38, vcc, 3, v39
+; SI-NEXT: v_add_i32_e32 v39, vcc, 3, v48
; SI-NEXT: s_add_i32 s16, s16, 3
; SI-NEXT: v_cvt_f32_f16_e32 v49, s16
; SI-NEXT: s_add_i32 s17, s17, 3
@@ -50351,50 +50409,45 @@ define inreg <60 x half> @bitcast_v60i16_to_v60f16_scalar(<60 x i16> inreg %a, i
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_cvt_f32_f16_e32 v49, s23
; SI-NEXT: v_add_i32_e32 v63, vcc, 3, v63
-; SI-NEXT: v_add_i32_e32 v36, vcc, 3, v36
+; SI-NEXT: v_cvt_f32_f16_e32 v57, v39
; SI-NEXT: buffer_store_dword v49, off, s[0:3], s32 offset:168 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_cvt_f32_f16_e32 v49, s24
-; SI-NEXT: v_add_i32_e32 v37, vcc, 3, v37
-; SI-NEXT: v_add_i32_e32 v38, vcc, 3, v38
+; SI-NEXT: v_cvt_f32_f16_e32 v42, v38
+; SI-NEXT: v_cvt_f32_f16_e32 v47, v37
+; SI-NEXT: v_cvt_f32_f16_e32 v40, v36
; SI-NEXT: buffer_store_dword v49, off, s[0:3], s32 offset:140 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_cvt_f32_f16_e32 v49, s25
-; SI-NEXT: v_add_i32_e32 v39, vcc, 3, v39
-; SI-NEXT: v_add_i32_e32 v48, vcc, 3, v48
+; SI-NEXT: v_cvt_f32_f16_e32 v45, v35
+; SI-NEXT: v_cvt_f32_f16_e32 v54, v34
+; SI-NEXT: v_cvt_f32_f16_e32 v43, v33
; SI-NEXT: buffer_store_dword v49, off, s[0:3], s32 offset:160 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_cvt_f32_f16_e32 v49, s26
-; SI-NEXT: v_cvt_f32_f16_e32 v44, v48
-; SI-NEXT: v_cvt_f32_f16_e32 v57, v39
-; SI-NEXT: v_cvt_f32_f16_e32 v42, v38
+; SI-NEXT: v_cvt_f32_f16_e32 v52, v32
+; SI-NEXT: v_cvt_f32_f16_e32 v41, v31
+; SI-NEXT: v_cvt_f32_f16_e32 v50, v63
; SI-NEXT: buffer_store_dword v49, off, s[0:3], s32 offset:136 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_cvt_f32_f16_e32 v49, s27
-; SI-NEXT: v_cvt_f32_f16_e32 v47, v37
-; SI-NEXT: v_cvt_f32_f16_e32 v40, v36
-; SI-NEXT: v_cvt_f32_f16_e32 v54, v34
+; SI-NEXT: v_cvt_f32_f16_e32 v55, v62
+; SI-NEXT: v_cvt_f32_f16_e32 v53, v60
+; SI-NEXT: s_waitcnt vmcnt(11)
+; SI-NEXT: v_add_i32_e32 v48, vcc, 3, v1
+; SI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:420 ; 4-byte Folded Reload
; SI-NEXT: buffer_store_dword v49, off, s[0:3], s32 offset:152 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_cvt_f32_f16_e32 v49, s28
-; SI-NEXT: v_cvt_f32_f16_e32 v43, v33
-; SI-NEXT: v_cvt_f32_f16_e32 v52, v32
-; SI-NEXT: v_cvt_f32_f16_e32 v41, v31
+; SI-NEXT: v_cvt_f32_f16_e32 v44, v48
+; SI-NEXT: v_cvt_f32_f16_e32 v51, v58
; SI-NEXT: buffer_store_dword v49, off, s[0:3], s32 offset:132 ; 4-byte Folded Spill
-; SI-NEXT: s_waitcnt vmcnt(13)
-; SI-NEXT: v_add_i32_e32 v35, vcc, 3, v1
-; SI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:420 ; 4-byte Folded Reload
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_cvt_f32_f16_e32 v49, s29
-; SI-NEXT: v_cvt_f32_f16_e32 v45, v35
-; SI-NEXT: v_cvt_f32_f16_e32 v50, v63
-; SI-NEXT: v_cvt_f32_f16_e32 v55, v62
; SI-NEXT: buffer_store_dword v49, off, s[0:3], s32 offset:144 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_cvt_f32_f16_e32 v49, v61
-; SI-NEXT: v_cvt_f32_f16_e32 v53, v60
-; SI-NEXT: v_cvt_f32_f16_e32 v51, v58
-; SI-NEXT: s_waitcnt vmcnt(1)
+; SI-NEXT: s_waitcnt vmcnt(3)
; SI-NEXT: v_add_i32_e32 v30, vcc, 3, v1
; SI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:416 ; 4-byte Folded Reload
; SI-NEXT: s_waitcnt vmcnt(0)
@@ -51390,53 +51443,62 @@ define inreg <60 x half> @bitcast_v60i16_to_v60f16_scalar(<60 x i16> inreg %a, i
; GFX11-TRUE16-NEXT: s_and_not1_b32 vcc_lo, exec_lo, s46
; GFX11-TRUE16-NEXT: s_cbranch_vccnz .LBB57_4
; GFX11-TRUE16-NEXT: .LBB57_2: ; %cmp.true
-; GFX11-TRUE16-NEXT: v_mov_b16_e32 v14.l, v9.h
-; GFX11-TRUE16-NEXT: v_and_b32_e32 v9, 0xffff, v9
; GFX11-TRUE16-NEXT: v_mov_b16_e32 v12.l, v11.h
; GFX11-TRUE16-NEXT: v_and_b32_e32 v11, 0xffff, v11
+; GFX11-TRUE16-NEXT: v_mov_b16_e32 v14.l, v9.h
+; GFX11-TRUE16-NEXT: v_and_b32_e32 v9, 0xffff, v9
; GFX11-TRUE16-NEXT: v_mov_b16_e32 v13.l, v10.h
; GFX11-TRUE16-NEXT: v_and_b32_e32 v10, 0xffff, v10
-; GFX11-TRUE16-NEXT: v_lshl_or_b32 v9, v14, 16, v9
-; GFX11-TRUE16-NEXT: v_mov_b16_e32 v14.l, v6.h
-; GFX11-TRUE16-NEXT: v_and_b32_e32 v6, 0xffff, v6
; GFX11-TRUE16-NEXT: v_lshl_or_b32 v11, v12, 16, v11
; GFX11-TRUE16-NEXT: v_mov_b16_e32 v12.l, v8.h
+; GFX11-TRUE16-NEXT: v_lshl_or_b32 v9, v14, 16, v9
; GFX11-TRUE16-NEXT: v_and_b32_e32 v8, 0xffff, v8
+; GFX11-TRUE16-NEXT: v_mov_b16_e32 v14.l, v6.h
+; GFX11-TRUE16-NEXT: v_and_b32_e32 v6, 0xffff, v6
+; GFX11-TRUE16-NEXT: v_mov_b16_e32 v15.l, v5.h
+; GFX11-TRUE16-NEXT: v_and_b32_e32 v5, 0xffff, v5
; GFX11-TRUE16-NEXT: v_lshl_or_b32 v10, v13, 16, v10
-; GFX11-TRUE16-NEXT: v_lshl_or_b32 v6, v14, 16, v6
-; GFX11-TRUE16-NEXT: v_mov_b16_e32 v14.l, v1.h
-; GFX11-TRUE16-NEXT: v_and_b32_e32 v1, 0xffff, v1
; GFX11-TRUE16-NEXT: v_mov_b16_e32 v13.l, v7.h
; GFX11-TRUE16-NEXT: v_and_b32_e32 v7, 0xffff, v7
-; GFX11-TRUE16-NEXT: v_mov_b16_e32 v15.l, v5.h
-; GFX11-TRUE16-NEXT: v_and_b32_e32 v5, 0xffff, v5
; GFX11-TRUE16-NEXT: v_lshl_or_b32 v8, v12, 16, v8
+; GFX11-TRUE16-NEXT: v_lshl_or_b32 v6, v14, 16, v6
+; GFX11-TRUE16-NEXT: v_lshl_or_b32 v5, v15, 16, v5
; GFX11-TRUE16-NEXT: v_mov_b16_e32 v12.l, v4.h
; GFX11-TRUE16-NEXT: v_and_b32_e32 v4, 0xffff, v4
-; GFX11-TRUE16-NEXT: s_pack_ll_b32_b16 s15, s24, s15
-; GFX11-TRUE16-NEXT: v_lshl_or_b32 v1, v14, 16, v1
-; GFX11-TRUE16-NEXT: v_pk_add_u16 v14, s15, 3 op_sel_hi:[1,0]
-; GFX11-TRUE16-NEXT: s_mov_b32 s15, s22
+; GFX11-TRUE16-NEXT: v_mov_b16_e32 v14.l, v1.h
+; GFX11-TRUE16-NEXT: v_and_b32_e32 v1, 0xffff, v1
+; GFX11-TRUE16-NEXT: v_mov_b16_e32 v15.l, v0.h
+; GFX11-TRUE16-NEXT: v_and_b32_e32 v0, 0xffff, v0
; GFX11-TRUE16-NEXT: v_lshl_or_b32 v7, v13, 16, v7
-; GFX11-TRUE16-NEXT: v_lshl_or_b32 v5, v15, 16, v5
; GFX11-TRUE16-NEXT: v_lshl_or_b32 v4, v12, 16, v4
; GFX11-TRUE16-NEXT: v_mov_b16_e32 v12.l, v3.h
; GFX11-TRUE16-NEXT: v_and_b32_e32 v3, 0xffff, v3
; GFX11-TRUE16-NEXT: v_mov_b16_e32 v13.l, v2.h
; GFX11-TRUE16-NEXT: v_and_b32_e32 v2, 0xffff, v2
-; GFX11-TRUE16-NEXT: v_mov_b16_e32 v15.l, v0.h
-; GFX11-TRUE16-NEXT: v_and_b32_e32 v0, 0xffff, v0
+; GFX11-TRUE16-NEXT: v_lshl_or_b32 v1, v14, 16, v1
+; GFX11-TRUE16-NEXT: v_lshl_or_b32 v0, v15, 16, v0
+; GFX11-TRUE16-NEXT: s_pack_ll_b32_b16 s15, s24, s15
+; GFX11-TRUE16-NEXT: v_lshl_or_b32 v3, v12, 16, v3
+; GFX11-TRUE16-NEXT: v_pk_add_u16 v14, s15, 3 op_sel_hi:[1,0]
+; GFX11-TRUE16-NEXT: s_mov_b32 s15, s22
+; GFX11-TRUE16-NEXT: v_lshl_or_b32 v2, v13, 16, v2
+; GFX11-TRUE16-NEXT: v_pk_add_u16 v1, v1, 3 op_sel_hi:[1,0]
+; GFX11-TRUE16-NEXT: v_pk_add_u16 v0, v0, 3 op_sel_hi:[1,0]
; GFX11-TRUE16-NEXT: s_pack_ll_b32_b16 s14, s23, s14
; GFX11-TRUE16-NEXT: s_pack_ll_b32_b16 s11, s15, s11
; GFX11-TRUE16-NEXT: v_pk_add_u16 v18, s14, 3 op_sel_hi:[1,0]
; GFX11-TRUE16-NEXT: s_mov_b32 s14, s21
; GFX11-TRUE16-NEXT: v_pk_add_u16 v19, s11, 3 op_sel_hi:[1,0]
; GFX11-TRUE16-NEXT: s_mov_b32 s11, s20
-; GFX11-TRUE16-NEXT: v_lshl_or_b32 v3, v12, 16, v3
-; GFX11-TRUE16-NEXT: v_lshl_or_b32 v2, v13, 16, v2
-; GFX11-TRUE16-NEXT: v_lshl_or_b32 v0, v15, 16, v0
+; GFX11-TRUE16-NEXT: v_pk_add_u16 v6, v6, 3 op_sel_hi:[1,0]
+; GFX11-TRUE16-NEXT: v_pk_add_u16 v5, v5, 3 op_sel_hi:[1,0]
+; GFX11-TRUE16-NEXT: v_pk_add_u16 v4, v4, 3 op_sel_hi:[1,0]
+; GFX11-TRUE16-NEXT: v_pk_add_u16 v3, v3, 3 op_sel_hi:[1,0]
+; GFX11-TRUE16-NEXT: v_pk_add_u16 v2, v2, 3 op_sel_hi:[1,0]
; GFX11-TRUE16-NEXT: s_pack_ll_b32_b16 s10, s14, s10
; GFX11-TRUE16-NEXT: s_pack_ll_b32_b16 s9, s11, s9
+; GFX11-TRUE16-NEXT: v_lshrrev_b32_e32 v64, 16, v0
+; GFX11-TRUE16-NEXT: v_lshrrev_b32_e32 v65, 16, v1
; GFX11-TRUE16-NEXT: v_pk_add_u16 v20, s10, 3 op_sel_hi:[1,0]
; GFX11-TRUE16-NEXT: s_mov_b32 s10, s19
; GFX11-TRUE16-NEXT: v_pk_add_u16 v21, s9, 3 op_sel_hi:[1,0]
@@ -51448,13 +51510,13 @@ define inreg <60 x half> @bitcast_v60i16_to_v60f16_scalar(<60 x i16> inreg %a, i
; GFX11-TRUE16-NEXT: v_pk_add_u16 v9, v9, 3 op_sel_hi:[1,0]
; GFX11-TRUE16-NEXT: v_pk_add_u16 v8, v8, 3 op_sel_hi:[1,0]
; GFX11-TRUE16-NEXT: v_pk_add_u16 v7, v7, 3 op_sel_hi:[1,0]
-; GFX11-TRUE16-NEXT: v_pk_add_u16 v6, v6, 3 op_sel_hi:[1,0]
-; GFX11-TRUE16-NEXT: v_pk_add_u16 v5, v5, 3 op_sel_hi:[1,0]
-; GFX11-TRUE16-NEXT: v_pk_add_u16 v4, v4, 3 op_sel_hi:[1,0]
-; GFX11-TRUE16-NEXT: v_pk_add_u16 v3, v3, 3 op_sel_hi:[1,0]
-; GFX11-TRUE16-NEXT: v_pk_add_u16 v2, v2, 3 op_sel_hi:[1,0]
-; GFX11-TRUE16-NEXT: v_pk_add_u16 v1, v1, 3 op_sel_hi:[1,0]
-; GFX11-TRUE16-NEXT: v_pk_add_u16 v0, v0, 3 op_sel_hi:[1,0]
+; GFX11-TRUE16-NEXT: v_mov_b16_e32 v0.h, v64.l
+; GFX11-TRUE16-NEXT: v_mov_b16_e32 v1.h, v65.l
+; GFX11-TRUE16-NEXT: v_lshrrev_b32_e32 v64, 16, v2
+; GFX11-TRUE16-NEXT: v_lshrrev_b32_e32 v65, 16, v3
+; GFX11-TRUE16-NEXT: v_lshrrev_b32_e32 v66, 16, v4
+; GFX11-TRUE16-NEXT: v_lshrrev_b32_e32 v67, 16, v5
+; GFX11-TRUE16-NEXT: v_lshrrev_b32_e32 v68, 16, v6
; GFX11-TRUE16-NEXT: v_pk_add_u16 v22, s8, 3 op_sel_hi:[1,0]
; GFX11-TRUE16-NEXT: s_mov_b32 s8, s17
; GFX11-TRUE16-NEXT: v_pk_add_u16 v23, s7, 3 op_sel_hi:[1,0]
@@ -51481,18 +51543,16 @@ define inreg <60 x half> @bitcast_v60i16_to_v60f16_scalar(<60 x i16> inreg %a, i
; GFX11-TRUE16-NEXT: v_pk_add_u16 v32, s2, 3 op_sel_hi:[1,0]
; GFX11-TRUE16-NEXT: v_pk_add_u16 v31, s3, 3 op_sel_hi:[1,0]
; GFX11-TRUE16-NEXT: v_pk_add_u16 v30, s5, 3 op_sel_hi:[1,0]
-; GFX11-TRUE16-NEXT: v_lshrrev_b32_e32 v64, 16, v0
-; GFX11-TRUE16-NEXT: v_lshrrev_b32_e32 v65, 16, v1
-; GFX11-TRUE16-NEXT: v_lshrrev_b32_e32 v66, 16, v2
-; GFX11-TRUE16-NEXT: v_lshrrev_b32_e32 v67, 16, v3
-; GFX11-TRUE16-NEXT: v_lshrrev_b32_e32 v68, 16, v4
-; GFX11-TRUE16-NEXT: v_lshrrev_b32_e32 v69, 16, v5
-; GFX11-TRUE16-NEXT: v_lshrrev_b32_e32 v70, 16, v6
-; GFX11-TRUE16-NEXT: v_lshrrev_b32_e32 v71, 16, v7
-; GFX11-TRUE16-NEXT: v_lshrrev_b32_e32 v80, 16, v8
-; GFX11-TRUE16-NEXT: v_lshrrev_b32_e32 v81, 16, v9
-; GFX11-TRUE16-NEXT: v_lshrrev_b32_e32 v82, 16, v10
-; GFX11-TRUE16-NEXT: v_lshrrev_b32_e32 v83, 16, v11
+; GFX11-TRUE16-NEXT: v_mov_b16_e32 v2.h, v64.l
+; GFX11-TRUE16-NEXT: v_mov_b16_e32 v3.h, v65.l
+; GFX11-TRUE16-NEXT: v_mov_b16_e32 v4.h, v66.l
+; GFX11-TRUE16-NEXT: v_mov_b16_e32 v5.h, v67.l
+; GFX11-TRUE16-NEXT: v_mov_b16_e32 v6.h, v68.l
+; GFX11-TRUE16-NEXT: v_lshrrev_b32_e32 v64, 16, v7
+; GFX11-TRUE16-NEXT: v_lshrrev_b32_e32 v65, 16, v8
+; GFX11-TRUE16-NEXT: v_lshrrev_b32_e32 v66, 16, v9
+; GFX11-TRUE16-NEXT: v_lshrrev_b32_e32 v67, 16, v10
+; GFX11-TRUE16-NEXT: v_lshrrev_b32_e32 v68, 16, v11
; GFX11-TRUE16-NEXT: v_lshrrev_b32_e32 v38, 16, v37
; GFX11-TRUE16-NEXT: v_lshrrev_b32_e32 v39, 16, v36
; GFX11-TRUE16-NEXT: v_lshrrev_b32_e32 v48, 16, v32
@@ -51508,21 +51568,14 @@ define inreg <60 x half> @bitcast_v60i16_to_v60f16_scalar(<60 x i16> inreg %a, i
; GFX11-TRUE16-NEXT: v_lshrrev_b32_e32 v50, 16, v14
; GFX11-TRUE16-NEXT: v_lshrrev_b32_e32 v51, 16, v13
; GFX11-TRUE16-NEXT: v_lshrrev_b32_e32 v52, 16, v12
-; GFX11-TRUE16-NEXT: v_lshrrev_b32_e32 v25, 16, v17
+; GFX11-TRUE16-NEXT: v_lshrrev_b32_e32 v27, 16, v17
; GFX11-TRUE16-NEXT: v_lshrrev_b32_e32 v26, 16, v16
-; GFX11-TRUE16-NEXT: v_lshrrev_b32_e32 v27, 16, v15
-; GFX11-TRUE16-NEXT: v_mov_b16_e32 v0.h, v64.l
-; GFX11-TRUE16-NEXT: v_mov_b16_e32 v1.h, v65.l
-; GFX11-TRUE16-NEXT: v_mov_b16_e32 v2.h, v66.l
-; GFX11-TRUE16-NEXT: v_mov_b16_e32 v3.h, v67.l
-; GFX11-TRUE16-NEXT: v_mov_b16_e32 v4.h, v68.l
-; GFX11-TRUE16-NEXT: v_mov_b16_e32 v5.h, v69.l
-; GFX11-TRUE16-NEXT: v_mov_b16_e32 v6.h, v70.l
-; GFX11-TRUE16-NEXT: v_mov_b16_e32 v7.h, v71.l
-; GFX11-TRUE16-NEXT: v_mov_b16_e32 v8.h, v80.l
-; GFX11-TRUE16-NEXT: v_mov_b16_e32 v9.h, v81.l
-; GFX11-TRUE16-NEXT: v_mov_b16_e32 v10.h, v82.l
-; GFX11-TRUE16-NEXT: v_mov_b16_e32 v11.h, v83.l
+; GFX11-TRUE16-NEXT: v_lshrrev_b32_e32 v25, 16, v15
+; GFX11-TRUE16-NEXT: v_mov_b16_e32 v7.h, v64.l
+; GFX11-TRUE16-NEXT: v_mov_b16_e32 v8.h, v65.l
+; GFX11-TRUE16-NEXT: v_mov_b16_e32 v9.h, v66.l
+; GFX11-TRUE16-NEXT: v_mov_b16_e32 v10.h, v67.l
+; GFX11-TRUE16-NEXT: v_mov_b16_e32 v11.h, v68.l
; GFX11-TRUE16-NEXT: s_branch .LBB57_5
; GFX11-TRUE16-NEXT: .LBB57_3:
; GFX11-TRUE16-NEXT: s_branch .LBB57_2
@@ -51545,9 +51598,9 @@ define inreg <60 x half> @bitcast_v60i16_to_v60f16_scalar(<60 x i16> inreg %a, i
; GFX11-TRUE16-NEXT: v_mov_b16_e32 v32.l, s2
; GFX11-TRUE16-NEXT: v_mov_b16_e32 v36.l, s1
; GFX11-TRUE16-NEXT: v_mov_b16_e32 v37.l, s0
-; GFX11-TRUE16-NEXT: v_mov_b16_e32 v27.l, s45
+; GFX11-TRUE16-NEXT: v_mov_b16_e32 v25.l, s45
; GFX11-TRUE16-NEXT: v_mov_b16_e32 v26.l, s44
-; GFX11-TRUE16-NEXT: v_mov_b16_e32 v25.l, s43
+; GFX11-TRUE16-NEXT: v_mov_b16_e32 v27.l, s43
; GFX11-TRUE16-NEXT: v_mov_b16_e32 v52.l, s42
; GFX11-TRUE16-NEXT: v_mov_b16_e32 v51.l, s41
; GFX11-TRUE16-NEXT: v_mov_b16_e32 v50.l, s15
@@ -51596,8 +51649,8 @@ define inreg <60 x half> @bitcast_v60i16_to_v60f16_scalar(<60 x i16> inreg %a, i
; GFX11-TRUE16-NEXT: v_and_b32_e32 v0, 0xffff, v0
; GFX11-TRUE16-NEXT: v_mov_b16_e32 v20.l, v1.h
; GFX11-TRUE16-NEXT: v_and_b32_e32 v1, 0xffff, v1
-; GFX11-TRUE16-NEXT: v_lshl_or_b32 v15, v25, 16, v17
-; GFX11-TRUE16-NEXT: v_lshl_or_b32 v17, v27, 16, v18
+; GFX11-TRUE16-NEXT: v_lshl_or_b32 v15, v27, 16, v17
+; GFX11-TRUE16-NEXT: v_lshl_or_b32 v17, v25, 16, v18
; GFX11-TRUE16-NEXT: v_lshl_or_b32 v18, v19, 16, v0
; GFX11-TRUE16-NEXT: v_mov_b16_e32 v0.l, v2.h
; GFX11-TRUE16-NEXT: v_lshl_or_b32 v19, v20, 16, v1
@@ -51939,48 +51992,48 @@ define <60 x i16> @bitcast_v60f16_to_v60i16(<60 x half> %a, i32 %b) {
; SI-NEXT: buffer_load_dword v47, off, s[0:3], s32 offset:56
; SI-NEXT: buffer_load_dword v56, off, s[0:3], s32 offset:60
; SI-NEXT: buffer_load_dword v57, off, s[0:3], s32 offset:64
-; SI-NEXT: buffer_load_dword v58, off, s[0:3], s32 offset:68
-; SI-NEXT: buffer_load_dword v31, off, s[0:3], s32 offset:72
-; SI-NEXT: buffer_load_dword v32, off, s[0:3], s32 offset:76
-; SI-NEXT: buffer_load_dword v59, off, s[0:3], s32 offset:80
-; SI-NEXT: buffer_load_dword v34, off, s[0:3], s32 offset:84
-; SI-NEXT: buffer_load_dword v35, off, s[0:3], s32 offset:88
+; SI-NEXT: buffer_load_dword v31, off, s[0:3], s32 offset:68
+; SI-NEXT: buffer_load_dword v32, off, s[0:3], s32 offset:72
+; SI-NEXT: buffer_load_dword v58, off, s[0:3], s32 offset:76
+; SI-NEXT: buffer_load_dword v34, off, s[0:3], s32 offset:80
+; SI-NEXT: buffer_load_dword v35, off, s[0:3], s32 offset:84
+; SI-NEXT: buffer_load_dword v37, off, s[0:3], s32 offset:88
; SI-NEXT: v_cvt_f16_f32_e32 v1, v1
; SI-NEXT: v_cvt_f16_f32_e32 v55, v3
; SI-NEXT: v_cvt_f16_f32_e32 v3, v22
-; SI-NEXT: v_cvt_f16_f32_e32 v60, v2
+; SI-NEXT: v_cvt_f16_f32_e32 v40, v4
; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:196 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_cvt_f16_f32_e32 v1, v5
-; SI-NEXT: v_cvt_f16_f32_e32 v2, v18
; SI-NEXT: buffer_store_dword v3, off, s[0:3], s32 offset:200 ; 4-byte Folded Spill
-; SI-NEXT: v_cvt_f16_f32_e32 v40, v4
+; SI-NEXT: v_cvt_f16_f32_e32 v59, v6
+; SI-NEXT: v_cvt_f16_f32_e32 v52, v7
; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:192 ; 4-byte Folded Spill
-; SI-NEXT: buffer_store_dword v2, off, s[0:3], s32 offset:204 ; 4-byte Folded Spill
-; SI-NEXT: s_waitcnt expcnt(1)
-; SI-NEXT: v_cvt_f16_f32_e32 v1, v6
-; SI-NEXT: v_cvt_f16_f32_e32 v53, v8
-; SI-NEXT: v_cvt_f16_f32_e32 v8, v10
+; SI-NEXT: s_waitcnt expcnt(0)
+; SI-NEXT: v_cvt_f16_f32_e32 v1, v9
+; SI-NEXT: v_cvt_f16_f32_e32 v5, v10
; SI-NEXT: v_cvt_f16_f32_e32 v49, v12
-; SI-NEXT: v_cvt_f16_f32_e32 v6, v13
-; SI-NEXT: v_cvt_f16_f32_e32 v37, v15
-; SI-NEXT: v_cvt_f16_f32_e32 v52, v7
-; SI-NEXT: v_cvt_f16_f32_e32 v7, v9
+; SI-NEXT: v_cvt_f16_f32_e32 v7, v13
+; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:188 ; 4-byte Folded Spill
+; SI-NEXT: v_cvt_f16_f32_e32 v6, v17
+; SI-NEXT: v_cvt_f16_f32_e32 v17, v18
+; SI-NEXT: s_waitcnt expcnt(0)
+; SI-NEXT: v_cvt_f16_f32_e32 v1, v19
+; SI-NEXT: v_cvt_f16_f32_e32 v60, v2
+; SI-NEXT: v_cvt_f16_f32_e32 v53, v8
; SI-NEXT: v_cvt_f16_f32_e32 v48, v11
; SI-NEXT: v_cvt_f16_f32_e32 v14, v14
+; SI-NEXT: v_cvt_f16_f32_e32 v2, v15
; SI-NEXT: v_cvt_f16_f32_e32 v38, v16
-; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_cvt_f16_f32_e32 v2, v19
; SI-NEXT: v_cvt_f16_f32_e32 v20, v20
; SI-NEXT: v_cvt_f16_f32_e32 v11, v23
-; SI-NEXT: v_cvt_f16_f32_e32 v5, v24
+; SI-NEXT: v_cvt_f16_f32_e32 v8, v24
; SI-NEXT: v_cvt_f16_f32_e32 v25, v25
; SI-NEXT: v_cvt_f16_f32_e32 v24, v26
; SI-NEXT: v_cvt_f16_f32_e32 v26, v27
; SI-NEXT: v_cvt_f16_f32_e32 v27, v28
; SI-NEXT: v_cvt_f16_f32_e32 v29, v29
; SI-NEXT: v_cvt_f16_f32_e32 v28, v30
-; SI-NEXT: v_cvt_f16_f32_e32 v17, v17
; SI-NEXT: v_cvt_f16_f32_e32 v21, v21
; SI-NEXT: s_waitcnt vmcnt(14)
; SI-NEXT: v_cmp_ne_u32_e32 vcc, 0, v51
@@ -51990,7 +52043,7 @@ define <60 x i16> @bitcast_v60f16_to_v60i16(<60 x half> %a, i32 %b) {
; SI-NEXT: v_cvt_f16_f32_e32 v44, v62
; SI-NEXT: v_cvt_f16_f32_e32 v18, v63
; SI-NEXT: v_cvt_f16_f32_e32 v19, v33
-; SI-NEXT: v_cvt_f16_f32_e32 v33, v36
+; SI-NEXT: v_cvt_f16_f32_e32 v62, v36
; SI-NEXT: v_cvt_f16_f32_e32 v43, v39
; SI-NEXT: v_cvt_f16_f32_e32 v15, v50
; SI-NEXT: v_cvt_f16_f32_e32 v16, v54
@@ -52004,28 +52057,26 @@ define <60 x i16> @bitcast_v60f16_to_v60i16(<60 x half> %a, i32 %b) {
; SI-NEXT: s_waitcnt vmcnt(11)
; SI-NEXT: v_cvt_f16_f32_e32 v30, v56
; SI-NEXT: s_waitcnt vmcnt(10)
-; SI-NEXT: v_cvt_f16_f32_e32 v63, v57
+; SI-NEXT: v_cvt_f16_f32_e32 v57, v57
; SI-NEXT: s_waitcnt vmcnt(9)
-; SI-NEXT: v_cvt_f16_f32_e32 v62, v58
+; SI-NEXT: v_cvt_f16_f32_e32 v63, v31
; SI-NEXT: s_waitcnt vmcnt(8)
-; SI-NEXT: v_cvt_f16_f32_e32 v36, v31
+; SI-NEXT: v_cvt_f16_f32_e32 v36, v32
; SI-NEXT: s_waitcnt vmcnt(7)
-; SI-NEXT: v_cvt_f16_f32_e32 v45, v32
+; SI-NEXT: v_cvt_f16_f32_e32 v32, v58
; SI-NEXT: s_waitcnt vmcnt(6)
-; SI-NEXT: v_cvt_f16_f32_e32 v31, v59
+; SI-NEXT: v_cvt_f16_f32_e32 v31, v34
; SI-NEXT: s_waitcnt vmcnt(5)
-; SI-NEXT: v_cvt_f16_f32_e32 v32, v34
+; SI-NEXT: v_cvt_f16_f32_e32 v34, v35
; SI-NEXT: s_waitcnt vmcnt(4)
-; SI-NEXT: v_cvt_f16_f32_e32 v3, v35
-; SI-NEXT: buffer_store_dword v3, off, s[0:3], s32 offset:188 ; 4-byte Folded Spill
-; SI-NEXT: s_waitcnt expcnt(0)
+; SI-NEXT: v_cvt_f16_f32_e32 v33, v37
; SI-NEXT: buffer_load_dword v3, off, s[0:3], s32 offset:92
; SI-NEXT: buffer_load_dword v4, off, s[0:3], s32 offset:96
; SI-NEXT: buffer_load_dword v10, off, s[0:3], s32 offset:100
-; SI-NEXT: buffer_load_dword v34, off, s[0:3], s32 offset:104
-; SI-NEXT: buffer_load_dword v35, off, s[0:3], s32 offset:108
-; SI-NEXT: buffer_load_dword v46, off, s[0:3], s32 offset:112
-; SI-NEXT: buffer_load_dword v47, off, s[0:3], s32 offset:116
+; SI-NEXT: buffer_load_dword v35, off, s[0:3], s32 offset:104
+; SI-NEXT: buffer_load_dword v37, off, s[0:3], s32 offset:108
+; SI-NEXT: buffer_load_dword v45, off, s[0:3], s32 offset:112
+; SI-NEXT: buffer_load_dword v46, off, s[0:3], s32 offset:116
; SI-NEXT: s_waitcnt vmcnt(6)
; SI-NEXT: v_cvt_f16_f32_e32 v42, v3
; SI-NEXT: s_waitcnt vmcnt(5)
@@ -52033,21 +52084,19 @@ define <60 x i16> @bitcast_v60f16_to_v60i16(<60 x half> %a, i32 %b) {
; SI-NEXT: s_waitcnt vmcnt(4)
; SI-NEXT: v_cvt_f16_f32_e32 v10, v10
; SI-NEXT: s_waitcnt vmcnt(3)
-; SI-NEXT: v_cvt_f16_f32_e32 v39, v34
+; SI-NEXT: v_cvt_f16_f32_e32 v39, v35
; SI-NEXT: s_waitcnt vmcnt(2)
-; SI-NEXT: v_cvt_f16_f32_e32 v41, v35
+; SI-NEXT: v_cvt_f16_f32_e32 v41, v37
; SI-NEXT: s_waitcnt vmcnt(1)
-; SI-NEXT: v_cvt_f16_f32_e32 v3, v46
+; SI-NEXT: v_cvt_f16_f32_e32 v3, v45
; SI-NEXT: s_waitcnt vmcnt(0)
-; SI-NEXT: v_cvt_f16_f32_e32 v4, v47
+; SI-NEXT: v_cvt_f16_f32_e32 v4, v46
; SI-NEXT: s_and_saveexec_b64 s[4:5], vcc
; SI-NEXT: s_xor_b64 s[4:5], exec, s[4:5]
; SI-NEXT: s_or_saveexec_b64 s[4:5], s[4:5]
; SI-NEXT: v_mov_b32_e32 v47, v21
-; SI-NEXT: v_mov_b32_e32 v56, v17
-; SI-NEXT: v_mov_b32_e32 v57, v6
+; SI-NEXT: v_mov_b32_e32 v56, v6
; SI-NEXT: v_mov_b32_e32 v58, v7
-; SI-NEXT: v_mov_b32_e32 v59, v33
; SI-NEXT: s_xor_b64 exec, exec, s[4:5]
; SI-NEXT: s_cbranch_execz .LBB58_2
; SI-NEXT: ; %bb.1: ; %cmp.true
@@ -52055,34 +52104,34 @@ define <60 x i16> @bitcast_v60f16_to_v60i16(<60 x half> %a, i32 %b) {
; SI-NEXT: v_cvt_f32_f16_e32 v3, v3
; SI-NEXT: v_cvt_f32_f16_e32 v10, v10
; SI-NEXT: v_cvt_f32_f16_e32 v9, v9
-; SI-NEXT: v_cvt_f32_f16_e32 v32, v32
+; SI-NEXT: v_cvt_f32_f16_e32 v34, v34
; SI-NEXT: v_cvt_f32_f16_e32 v31, v31
-; SI-NEXT: v_cvt_f32_f16_e32 v7, v62
+; SI-NEXT: v_cvt_f32_f16_e32 v7, v63
; SI-NEXT: v_add_f32_e32 v4, 0x38000000, v4
-; SI-NEXT: v_cvt_f32_f16_e32 v6, v63
+; SI-NEXT: v_cvt_f32_f16_e32 v6, v57
; SI-NEXT: v_add_f32_e32 v3, 0x38000000, v3
; SI-NEXT: v_cvt_f16_f32_e32 v4, v4
; SI-NEXT: v_add_f32_e32 v10, 0x38000000, v10
; SI-NEXT: v_cvt_f16_f32_e32 v3, v3
; SI-NEXT: v_cvt_f16_f32_e32 v10, v10
; SI-NEXT: v_add_f32_e32 v9, 0x38000000, v9
-; SI-NEXT: v_add_f32_e32 v32, 0x38000000, v32
+; SI-NEXT: v_add_f32_e32 v34, 0x38000000, v34
; SI-NEXT: v_cvt_f16_f32_e32 v9, v9
-; SI-NEXT: v_cvt_f16_f32_e32 v32, v32
+; SI-NEXT: v_cvt_f16_f32_e32 v34, v34
; SI-NEXT: v_add_f32_e32 v31, 0x38000000, v31
; SI-NEXT: v_add_f32_e32 v7, 0x38000000, v7
; SI-NEXT: v_cvt_f16_f32_e32 v31, v31
-; SI-NEXT: v_cvt_f16_f32_e32 v62, v7
+; SI-NEXT: v_cvt_f16_f32_e32 v63, v7
; SI-NEXT: v_add_f32_e32 v6, 0x38000000, v6
-; SI-NEXT: v_lshlrev_b32_e32 v34, 16, v4
+; SI-NEXT: v_lshlrev_b32_e32 v35, 16, v4
; SI-NEXT: v_cvt_f16_f32_e32 v6, v6
-; SI-NEXT: v_or_b32_e32 v3, v3, v34
-; SI-NEXT: v_lshlrev_b32_e32 v34, 16, v10
-; SI-NEXT: v_or_b32_e32 v9, v9, v34
-; SI-NEXT: v_lshlrev_b32_e32 v34, 16, v32
-; SI-NEXT: v_or_b32_e32 v31, v31, v34
-; SI-NEXT: v_lshlrev_b32_e32 v34, 16, v62
-; SI-NEXT: v_or_b32_e32 v63, v6, v34
+; SI-NEXT: v_or_b32_e32 v3, v3, v35
+; SI-NEXT: v_lshlrev_b32_e32 v35, 16, v10
+; SI-NEXT: v_or_b32_e32 v9, v9, v35
+; SI-NEXT: v_lshlrev_b32_e32 v35, 16, v34
+; SI-NEXT: v_or_b32_e32 v31, v31, v35
+; SI-NEXT: v_lshlrev_b32_e32 v35, 16, v63
+; SI-NEXT: v_or_b32_e32 v57, v6, v35
; SI-NEXT: buffer_load_dword v6, off, s[0:3], s32 offset:196 ; 4-byte Folded Reload
; SI-NEXT: v_cvt_f32_f16_e32 v13, v13
; SI-NEXT: v_cvt_f32_f16_e32 v12, v12
@@ -52098,7 +52147,7 @@ define <60 x i16> @bitcast_v60f16_to_v60i16(<60 x half> %a, i32 %b) {
; SI-NEXT: v_add_f32_e32 v12, 0x38000000, v12
; SI-NEXT: v_add_f32_e32 v16, 0x38000000, v16
; SI-NEXT: v_cvt_f32_f16_e32 v26, v26
-; SI-NEXT: v_cvt_f32_f16_e32 v5, v5
+; SI-NEXT: v_cvt_f32_f16_e32 v8, v8
; SI-NEXT: v_cvt_f16_f32_e32 v12, v12
; SI-NEXT: v_cvt_f16_f32_e32 v16, v16
; SI-NEXT: v_add_f32_e32 v15, 0x38000000, v15
@@ -52109,114 +52158,119 @@ define <60 x i16> @bitcast_v60f16_to_v60i16(<60 x half> %a, i32 %b) {
; SI-NEXT: v_cvt_f16_f32_e32 v19, v19
; SI-NEXT: v_add_f32_e32 v18, 0x38000000, v18
; SI-NEXT: v_add_f32_e32 v23, 0x38000000, v23
-; SI-NEXT: v_cvt_f32_f16_e32 v2, v2
+; SI-NEXT: v_cvt_f32_f16_e32 v1, v1
; SI-NEXT: v_cvt_f16_f32_e32 v18, v18
; SI-NEXT: v_cvt_f16_f32_e32 v23, v23
; SI-NEXT: v_add_f32_e32 v22, 0x38000000, v22
; SI-NEXT: v_add_f32_e32 v27, 0x38000000, v27
-; SI-NEXT: v_lshlrev_b32_e32 v34, 16, v13
+; SI-NEXT: v_lshlrev_b32_e32 v35, 16, v13
; SI-NEXT: v_cvt_f16_f32_e32 v22, v22
; SI-NEXT: v_cvt_f16_f32_e32 v27, v27
; SI-NEXT: v_add_f32_e32 v26, 0x38000000, v26
-; SI-NEXT: v_add_f32_e32 v5, 0x38000000, v5
-; SI-NEXT: v_or_b32_e32 v12, v12, v34
-; SI-NEXT: v_lshlrev_b32_e32 v34, 16, v16
+; SI-NEXT: v_add_f32_e32 v8, 0x38000000, v8
+; SI-NEXT: v_or_b32_e32 v12, v12, v35
+; SI-NEXT: v_lshlrev_b32_e32 v35, 16, v16
; SI-NEXT: v_cvt_f16_f32_e32 v26, v26
-; SI-NEXT: v_cvt_f16_f32_e32 v5, v5
+; SI-NEXT: v_cvt_f16_f32_e32 v8, v8
; SI-NEXT: v_add_f32_e32 v11, 0x38000000, v11
; SI-NEXT: v_add_f32_e32 v20, 0x38000000, v20
-; SI-NEXT: v_or_b32_e32 v15, v15, v34
-; SI-NEXT: v_lshlrev_b32_e32 v34, 16, v19
+; SI-NEXT: v_or_b32_e32 v15, v15, v35
+; SI-NEXT: v_lshlrev_b32_e32 v35, 16, v19
; SI-NEXT: v_cvt_f16_f32_e32 v11, v11
; SI-NEXT: v_cvt_f16_f32_e32 v20, v20
+; SI-NEXT: v_add_f32_e32 v1, 0x38000000, v1
+; SI-NEXT: v_or_b32_e32 v18, v18, v35
+; SI-NEXT: v_lshlrev_b32_e32 v35, 16, v23
+; SI-NEXT: v_cvt_f16_f32_e32 v1, v1
+; SI-NEXT: v_or_b32_e32 v22, v22, v35
+; SI-NEXT: v_lshlrev_b32_e32 v35, 16, v27
+; SI-NEXT: v_or_b32_e32 v26, v26, v35
+; SI-NEXT: v_lshlrev_b32_e32 v35, 16, v8
+; SI-NEXT: v_or_b32_e32 v11, v11, v35
+; SI-NEXT: v_lshlrev_b32_e32 v35, 16, v20
+; SI-NEXT: v_or_b32_e32 v1, v1, v35
+; SI-NEXT: v_cvt_f32_f16_e32 v35, v38
+; SI-NEXT: v_cvt_f32_f16_e32 v2, v2
+; SI-NEXT: v_cvt_f32_f16_e32 v37, v48
+; SI-NEXT: v_cvt_f32_f16_e32 v45, v59
+; SI-NEXT: v_add_f32_e32 v35, 0x38000000, v35
+; SI-NEXT: v_cvt_f16_f32_e32 v38, v35
; SI-NEXT: v_add_f32_e32 v2, 0x38000000, v2
-; SI-NEXT: v_or_b32_e32 v18, v18, v34
-; SI-NEXT: v_lshlrev_b32_e32 v34, 16, v23
; SI-NEXT: v_cvt_f16_f32_e32 v2, v2
-; SI-NEXT: v_or_b32_e32 v22, v22, v34
-; SI-NEXT: v_lshlrev_b32_e32 v34, 16, v27
-; SI-NEXT: v_or_b32_e32 v26, v26, v34
-; SI-NEXT: v_lshlrev_b32_e32 v34, 16, v5
-; SI-NEXT: v_or_b32_e32 v11, v11, v34
-; SI-NEXT: v_lshlrev_b32_e32 v34, 16, v20
-; SI-NEXT: v_or_b32_e32 v2, v2, v34
-; SI-NEXT: v_cvt_f32_f16_e32 v34, v38
-; SI-NEXT: v_cvt_f32_f16_e32 v35, v37
-; SI-NEXT: v_cvt_f32_f16_e32 v1, v1
-; SI-NEXT: v_cvt_f32_f16_e32 v8, v8
-; SI-NEXT: v_add_f32_e32 v34, 0x38000000, v34
-; SI-NEXT: v_cvt_f16_f32_e32 v38, v34
-; SI-NEXT: v_add_f32_e32 v34, 0x38000000, v35
-; SI-NEXT: v_cvt_f16_f32_e32 v34, v34
-; SI-NEXT: v_add_f32_e32 v1, 0x38000000, v1
+; SI-NEXT: v_add_f32_e32 v45, 0x38000000, v45
; SI-NEXT: v_lshlrev_b32_e32 v35, 16, v38
-; SI-NEXT: v_cvt_f16_f32_e32 v1, v1
-; SI-NEXT: v_or_b32_e32 v37, v34, v35
-; SI-NEXT: v_cvt_f32_f16_e32 v34, v49
-; SI-NEXT: v_cvt_f32_f16_e32 v35, v48
-; SI-NEXT: v_lshlrev_b32_e32 v1, 16, v1
-; SI-NEXT: v_cvt_f32_f16_e32 v46, v58
-; SI-NEXT: v_add_f32_e32 v34, 0x38000000, v34
-; SI-NEXT: v_cvt_f16_f32_e32 v49, v34
-; SI-NEXT: v_add_f32_e32 v34, 0x38000000, v35
-; SI-NEXT: v_cvt_f16_f32_e32 v34, v34
-; SI-NEXT: v_add_f32_e32 v8, 0x38000000, v8
-; SI-NEXT: v_lshlrev_b32_e32 v35, 16, v49
-; SI-NEXT: v_cvt_f16_f32_e32 v8, v8
-; SI-NEXT: v_or_b32_e32 v48, v34, v35
-; SI-NEXT: v_cvt_f32_f16_e32 v34, v53
-; SI-NEXT: v_cvt_f32_f16_e32 v35, v52
+; SI-NEXT: v_cvt_f32_f16_e32 v5, v5
+; SI-NEXT: v_or_b32_e32 v2, v2, v35
+; SI-NEXT: v_cvt_f32_f16_e32 v35, v49
+; SI-NEXT: v_cvt_f16_f32_e32 v45, v45
+; SI-NEXT: v_add_f32_e32 v5, 0x38000000, v5
+; SI-NEXT: v_cvt_f16_f32_e32 v5, v5
+; SI-NEXT: v_add_f32_e32 v35, 0x38000000, v35
+; SI-NEXT: v_cvt_f16_f32_e32 v49, v35
+; SI-NEXT: v_add_f32_e32 v35, 0x38000000, v37
+; SI-NEXT: v_cvt_f16_f32_e32 v35, v35
+; SI-NEXT: v_lshlrev_b32_e32 v45, 16, v45
+; SI-NEXT: v_lshlrev_b32_e32 v37, 16, v49
+; SI-NEXT: v_lshlrev_b32_e32 v5, 16, v5
+; SI-NEXT: v_or_b32_e32 v48, v35, v37
+; SI-NEXT: v_cvt_f32_f16_e32 v35, v53
+; SI-NEXT: v_cvt_f32_f16_e32 v37, v52
; SI-NEXT: v_cvt_f32_f16_e32 v14, v14
-; SI-NEXT: v_lshlrev_b32_e32 v8, 16, v8
-; SI-NEXT: v_add_f32_e32 v34, 0x38000000, v34
-; SI-NEXT: v_cvt_f16_f32_e32 v53, v34
-; SI-NEXT: v_add_f32_e32 v34, 0x38000000, v35
-; SI-NEXT: v_cvt_f16_f32_e32 v34, v34
+; SI-NEXT: v_cvt_f32_f16_e32 v17, v17
+; SI-NEXT: v_add_f32_e32 v35, 0x38000000, v35
+; SI-NEXT: v_cvt_f16_f32_e32 v53, v35
+; SI-NEXT: v_add_f32_e32 v35, 0x38000000, v37
+; SI-NEXT: v_cvt_f16_f32_e32 v35, v35
; SI-NEXT: v_add_f32_e32 v14, 0x38000000, v14
-; SI-NEXT: v_lshlrev_b32_e32 v35, 16, v53
+; SI-NEXT: v_lshlrev_b32_e32 v37, 16, v53
; SI-NEXT: v_cvt_f16_f32_e32 v14, v14
-; SI-NEXT: v_or_b32_e32 v52, v34, v35
-; SI-NEXT: v_cvt_f32_f16_e32 v34, v40
-; SI-NEXT: v_cvt_f32_f16_e32 v35, v55
+; SI-NEXT: v_or_b32_e32 v52, v35, v37
+; SI-NEXT: v_cvt_f32_f16_e32 v35, v40
+; SI-NEXT: v_cvt_f32_f16_e32 v37, v55
+; SI-NEXT: v_add_f32_e32 v17, 0x38000000, v17
; SI-NEXT: v_lshlrev_b32_e32 v14, 16, v14
-; SI-NEXT: v_cvt_f32_f16_e32 v42, v42
-; SI-NEXT: v_add_f32_e32 v34, 0x38000000, v34
-; SI-NEXT: v_cvt_f16_f32_e32 v40, v34
-; SI-NEXT: v_add_f32_e32 v34, 0x38000000, v35
-; SI-NEXT: v_cvt_f16_f32_e32 v34, v34
+; SI-NEXT: v_add_f32_e32 v35, 0x38000000, v35
+; SI-NEXT: v_cvt_f16_f32_e32 v40, v35
+; SI-NEXT: v_add_f32_e32 v35, 0x38000000, v37
+; SI-NEXT: v_cvt_f16_f32_e32 v35, v35
+; SI-NEXT: v_cvt_f16_f32_e32 v17, v17
+; SI-NEXT: v_lshlrev_b32_e32 v37, 16, v40
; SI-NEXT: v_cvt_f32_f16_e32 v24, v24
-; SI-NEXT: v_lshlrev_b32_e32 v35, 16, v40
-; SI-NEXT: v_cvt_f32_f16_e32 v28, v28
-; SI-NEXT: v_or_b32_e32 v55, v34, v35
-; SI-NEXT: v_cvt_f32_f16_e32 v34, v60
+; SI-NEXT: v_or_b32_e32 v55, v35, v37
+; SI-NEXT: v_cvt_f32_f16_e32 v35, v60
; SI-NEXT: s_waitcnt vmcnt(0)
-; SI-NEXT: v_cvt_f32_f16_e32 v35, v6
-; SI-NEXT: v_cvt_f32_f16_e32 v43, v43
-; SI-NEXT: v_cvt_f32_f16_e32 v51, v51
-; SI-NEXT: v_add_f32_e32 v34, 0x38000000, v34
-; SI-NEXT: v_cvt_f16_f32_e32 v34, v34
+; SI-NEXT: v_cvt_f32_f16_e32 v37, v6
+; SI-NEXT: v_lshlrev_b32_e32 v17, 16, v17
+; SI-NEXT: v_cvt_f32_f16_e32 v28, v28
; SI-NEXT: v_add_f32_e32 v35, 0x38000000, v35
; SI-NEXT: v_cvt_f16_f32_e32 v35, v35
-; SI-NEXT: v_cvt_f32_f16_e32 v30, v30
-; SI-NEXT: v_lshlrev_b32_e32 v34, 16, v34
-; SI-NEXT: v_cvt_f32_f16_e32 v41, v41
-; SI-NEXT: v_or_b32_e32 v6, v35, v34
+; SI-NEXT: v_add_f32_e32 v37, 0x38000000, v37
+; SI-NEXT: v_cvt_f16_f32_e32 v37, v37
+; SI-NEXT: v_cvt_f32_f16_e32 v43, v43
+; SI-NEXT: v_lshlrev_b32_e32 v35, 16, v35
+; SI-NEXT: v_cvt_f32_f16_e32 v51, v51
+; SI-NEXT: v_or_b32_e32 v6, v37, v35
; SI-NEXT: buffer_store_dword v6, off, s[0:3], s32 offset:196 ; 4-byte Folded Spill
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: buffer_load_dword v6, off, s[0:3], s32 offset:192 ; 4-byte Folded Reload
+; SI-NEXT: v_cvt_f32_f16_e32 v30, v30
+; SI-NEXT: v_cvt_f32_f16_e32 v32, v32
+; SI-NEXT: v_cvt_f32_f16_e32 v42, v42
+; SI-NEXT: v_cvt_f32_f16_e32 v41, v41
; SI-NEXT: v_cvt_f32_f16_e32 v25, v25
; SI-NEXT: v_cvt_f32_f16_e32 v29, v29
; SI-NEXT: v_cvt_f32_f16_e32 v54, v54
; SI-NEXT: v_cvt_f32_f16_e32 v50, v50
; SI-NEXT: v_cvt_f32_f16_e32 v36, v36
+; SI-NEXT: v_cvt_f32_f16_e32 v33, v33
; SI-NEXT: v_cvt_f32_f16_e32 v39, v39
-; SI-NEXT: v_add_f32_e32 v42, 0x38000000, v42
-; SI-NEXT: v_cvt_f16_f32_e32 v42, v42
; SI-NEXT: v_add_f32_e32 v24, 0x38000000, v24
; SI-NEXT: v_add_f32_e32 v28, 0x38000000, v28
; SI-NEXT: v_add_f32_e32 v43, 0x38000000, v43
; SI-NEXT: v_add_f32_e32 v51, 0x38000000, v51
; SI-NEXT: v_add_f32_e32 v30, 0x38000000, v30
+; SI-NEXT: v_add_f32_e32 v32, 0x38000000, v32
+; SI-NEXT: v_add_f32_e32 v42, 0x38000000, v42
; SI-NEXT: v_add_f32_e32 v41, 0x38000000, v41
; SI-NEXT: v_cvt_f16_f32_e32 v24, v24
; SI-NEXT: v_add_f32_e32 v25, 0x38000000, v25
@@ -52227,7 +52281,10 @@ define <60 x i16> @bitcast_v60f16_to_v60i16(<60 x half> %a, i32 %b) {
; SI-NEXT: v_add_f32_e32 v54, 0x38000000, v54
; SI-NEXT: v_cvt_f16_f32_e32 v30, v30
; SI-NEXT: v_add_f32_e32 v50, 0x38000000, v50
+; SI-NEXT: v_cvt_f16_f32_e32 v32, v32
; SI-NEXT: v_add_f32_e32 v36, 0x38000000, v36
+; SI-NEXT: v_cvt_f16_f32_e32 v42, v42
+; SI-NEXT: v_add_f32_e32 v33, 0x38000000, v33
; SI-NEXT: v_cvt_f16_f32_e32 v41, v41
; SI-NEXT: v_add_f32_e32 v39, 0x38000000, v39
; SI-NEXT: v_cvt_f16_f32_e32 v25, v25
@@ -52235,158 +52292,145 @@ define <60 x i16> @bitcast_v60f16_to_v60i16(<60 x half> %a, i32 %b) {
; SI-NEXT: v_cvt_f16_f32_e32 v54, v54
; SI-NEXT: v_cvt_f16_f32_e32 v50, v50
; SI-NEXT: v_cvt_f16_f32_e32 v36, v36
+; SI-NEXT: v_cvt_f16_f32_e32 v33, v33
; SI-NEXT: v_cvt_f16_f32_e32 v39, v39
-; SI-NEXT: v_lshlrev_b32_e32 v42, 16, v42
; SI-NEXT: v_lshlrev_b32_e32 v24, 16, v24
; SI-NEXT: v_lshlrev_b32_e32 v28, 16, v28
; SI-NEXT: v_lshlrev_b32_e32 v43, 16, v43
; SI-NEXT: v_lshlrev_b32_e32 v51, 16, v51
; SI-NEXT: v_lshlrev_b32_e32 v30, 16, v30
+; SI-NEXT: v_lshlrev_b32_e32 v32, 16, v32
+; SI-NEXT: v_lshlrev_b32_e32 v42, 16, v42
; SI-NEXT: v_lshlrev_b32_e32 v41, 16, v41
; SI-NEXT: v_or_b32_e32 v25, v25, v24
; SI-NEXT: v_or_b32_e32 v29, v29, v28
; SI-NEXT: v_or_b32_e32 v54, v54, v51
; SI-NEXT: v_or_b32_e32 v50, v50, v30
+; SI-NEXT: v_or_b32_e32 v36, v36, v32
+; SI-NEXT: v_or_b32_e32 v33, v33, v42
; SI-NEXT: v_or_b32_e32 v39, v39, v41
-; SI-NEXT: v_alignbit_b32 v60, v55, v34, 16
+; SI-NEXT: v_alignbit_b32 v60, v55, v35, 16
+; SI-NEXT: v_alignbit_b32 v59, v52, v45, 16
; SI-NEXT: v_alignbit_b32 v24, v26, v24, 16
; SI-NEXT: v_alignbit_b32 v28, v22, v28, 16
; SI-NEXT: v_alignbit_b32 v51, v12, v51, 16
-; SI-NEXT: v_alignbit_b32 v30, v63, v30, 16
+; SI-NEXT: v_alignbit_b32 v30, v57, v30, 16
+; SI-NEXT: v_alignbit_b32 v32, v31, v32, 16
+; SI-NEXT: v_alignbit_b32 v42, v9, v42, 16
; SI-NEXT: v_alignbit_b32 v41, v3, v41, 16
; SI-NEXT: s_waitcnt vmcnt(0)
-; SI-NEXT: v_cvt_f32_f16_e32 v35, v6
-; SI-NEXT: v_add_f32_e32 v35, 0x38000000, v35
-; SI-NEXT: v_cvt_f16_f32_e32 v35, v35
-; SI-NEXT: v_or_b32_e32 v6, v35, v1
+; SI-NEXT: v_cvt_f32_f16_e32 v37, v6
+; SI-NEXT: buffer_load_dword v6, off, s[0:3], s32 offset:188 ; 4-byte Folded Reload
+; SI-NEXT: v_add_f32_e32 v37, 0x38000000, v37
+; SI-NEXT: v_cvt_f16_f32_e32 v37, v37
+; SI-NEXT: s_waitcnt vmcnt(0)
+; SI-NEXT: v_cvt_f32_f16_e32 v46, v6
+; SI-NEXT: v_or_b32_e32 v6, v37, v45
; SI-NEXT: buffer_store_dword v6, off, s[0:3], s32 offset:192 ; 4-byte Folded Spill
+; SI-NEXT: v_add_f32_e32 v37, 0x38000000, v46
+; SI-NEXT: v_cvt_f16_f32_e32 v37, v37
+; SI-NEXT: v_cvt_f32_f16_e32 v46, v58
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: buffer_load_dword v6, off, s[0:3], s32 offset:204 ; 4-byte Folded Reload
-; SI-NEXT: v_add_f32_e32 v35, 0x38000000, v46
-; SI-NEXT: v_cvt_f32_f16_e32 v46, v57
-; SI-NEXT: v_cvt_f16_f32_e32 v35, v35
-; SI-NEXT: v_alignbit_b32 v1, v52, v1, 16
+; SI-NEXT: v_or_b32_e32 v6, v37, v5
+; SI-NEXT: buffer_store_dword v6, off, s[0:3], s32 offset:188 ; 4-byte Folded Spill
+; SI-NEXT: s_waitcnt expcnt(0)
+; SI-NEXT: buffer_load_dword v6, off, s[0:3], s32 offset:200 ; 4-byte Folded Reload
; SI-NEXT: v_add_f32_e32 v46, 0x38000000, v46
-; SI-NEXT: v_or_b32_e32 v58, v35, v8
-; SI-NEXT: v_cvt_f32_f16_e32 v35, v56
+; SI-NEXT: v_cvt_f32_f16_e32 v37, v56
; SI-NEXT: v_cvt_f16_f32_e32 v46, v46
-; SI-NEXT: v_alignbit_b32 v8, v48, v8, 16
-; SI-NEXT: v_add_f32_e32 v35, 0x38000000, v35
-; SI-NEXT: v_or_b32_e32 v57, v46, v14
-; SI-NEXT: v_cvt_f16_f32_e32 v35, v35
+; SI-NEXT: v_alignbit_b32 v5, v48, v5, 16
+; SI-NEXT: v_add_f32_e32 v37, 0x38000000, v37
+; SI-NEXT: v_or_b32_e32 v58, v46, v14
+; SI-NEXT: v_cvt_f16_f32_e32 v37, v37
; SI-NEXT: v_cvt_f32_f16_e32 v46, v47
-; SI-NEXT: v_alignbit_b32 v14, v37, v14, 16
-; SI-NEXT: s_waitcnt vmcnt(0)
-; SI-NEXT: v_cvt_f32_f16_e32 v17, v6
-; SI-NEXT: buffer_load_dword v6, off, s[0:3], s32 offset:200 ; 4-byte Folded Reload
-; SI-NEXT: v_add_f32_e32 v17, 0x38000000, v17
-; SI-NEXT: v_cvt_f16_f32_e32 v17, v17
-; SI-NEXT: v_lshlrev_b32_e32 v17, 16, v17
-; SI-NEXT: v_or_b32_e32 v56, v35, v17
-; SI-NEXT: v_add_f32_e32 v35, 0x38000000, v46
-; SI-NEXT: v_cvt_f16_f32_e32 v35, v35
-; SI-NEXT: v_cvt_f32_f16_e32 v46, v59
+; SI-NEXT: v_alignbit_b32 v14, v2, v14, 16
+; SI-NEXT: v_or_b32_e32 v56, v37, v17
+; SI-NEXT: v_add_f32_e32 v37, 0x38000000, v46
+; SI-NEXT: v_cvt_f16_f32_e32 v37, v37
+; SI-NEXT: v_cvt_f32_f16_e32 v46, v62
+; SI-NEXT: v_alignbit_b32 v17, v1, v17, 16
; SI-NEXT: v_add_f32_e32 v46, 0x38000000, v46
; SI-NEXT: v_cvt_f16_f32_e32 v46, v46
-; SI-NEXT: v_or_b32_e32 v59, v46, v43
+; SI-NEXT: v_or_b32_e32 v62, v46, v43
; SI-NEXT: v_alignbit_b32 v43, v15, v43, 16
; SI-NEXT: s_waitcnt vmcnt(0)
; SI-NEXT: v_cvt_f32_f16_e32 v21, v6
-; SI-NEXT: buffer_load_dword v6, off, s[0:3], s32 offset:188 ; 4-byte Folded Reload
; SI-NEXT: v_add_f32_e32 v21, 0x38000000, v21
; SI-NEXT: v_cvt_f16_f32_e32 v21, v21
; SI-NEXT: v_lshlrev_b32_e32 v21, 16, v21
-; SI-NEXT: v_or_b32_e32 v47, v35, v21
-; SI-NEXT: v_cvt_f32_f16_e32 v35, v44
+; SI-NEXT: v_or_b32_e32 v47, v37, v21
+; SI-NEXT: v_cvt_f32_f16_e32 v37, v44
; SI-NEXT: v_cvt_f32_f16_e32 v44, v61
-; SI-NEXT: v_add_f32_e32 v35, 0x38000000, v35
-; SI-NEXT: v_cvt_f16_f32_e32 v35, v35
-; SI-NEXT: v_add_f32_e32 v44, 0x38000000, v44
-; SI-NEXT: v_cvt_f16_f32_e32 v44, v44
-; SI-NEXT: v_lshlrev_b32_e32 v35, 16, v35
-; SI-NEXT: v_or_b32_e32 v61, v44, v35
-; SI-NEXT: v_cvt_f32_f16_e32 v44, v45
-; SI-NEXT: v_add_f32_e32 v44, 0x38000000, v44
-; SI-NEXT: v_cvt_f16_f32_e32 v44, v44
-; SI-NEXT: v_lshlrev_b32_e32 v45, 16, v44
-; SI-NEXT: v_or_b32_e32 v36, v36, v45
-; SI-NEXT: v_alignbit_b32 v44, v18, v35, 16
-; SI-NEXT: v_alignbit_b32 v45, v31, v45, 16
-; SI-NEXT: s_waitcnt vmcnt(0)
-; SI-NEXT: v_cvt_f32_f16_e32 v33, v6
-; SI-NEXT: v_add_f32_e32 v33, 0x38000000, v33
-; SI-NEXT: v_cvt_f16_f32_e32 v33, v33
-; SI-NEXT: v_or_b32_e32 v6, v33, v42
-; SI-NEXT: buffer_store_dword v6, off, s[0:3], s32 offset:188 ; 4-byte Folded Spill
-; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_alignbit_b32 v6, v2, v17, 16
-; SI-NEXT: buffer_store_dword v6, off, s[0:3], s32 offset:204 ; 4-byte Folded Spill
-; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_alignbit_b32 v6, v11, v21, 16
-; SI-NEXT: v_alignbit_b32 v42, v9, v42, 16
; SI-NEXT: buffer_store_dword v6, off, s[0:3], s32 offset:200 ; 4-byte Folded Spill
+; SI-NEXT: v_add_f32_e32 v37, 0x38000000, v37
+; SI-NEXT: v_cvt_f16_f32_e32 v37, v37
+; SI-NEXT: v_add_f32_e32 v44, 0x38000000, v44
+; SI-NEXT: v_cvt_f16_f32_e32 v44, v44
+; SI-NEXT: v_lshlrev_b32_e32 v37, 16, v37
+; SI-NEXT: v_or_b32_e32 v61, v44, v37
+; SI-NEXT: v_alignbit_b32 v44, v18, v37, 16
; SI-NEXT: .LBB58_2: ; %end
; SI-NEXT: s_or_b64 exec, exec, s[4:5]
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: buffer_load_dword v6, off, s[0:3], s32 offset:196 ; 4-byte Folded Reload
-; SI-NEXT: v_lshlrev_b32_e32 v35, 16, v60
-; SI-NEXT: v_lshlrev_b32_e32 v1, 16, v1
-; SI-NEXT: v_lshlrev_b32_e32 v8, 16, v8
+; SI-NEXT: v_lshlrev_b32_e32 v37, 16, v60
+; SI-NEXT: v_lshlrev_b32_e32 v5, 16, v5
+; SI-NEXT: v_lshlrev_b32_e32 v14, 16, v14
+; SI-NEXT: v_and_b32_e32 v2, 0xffff, v2
+; SI-NEXT: v_and_b32_e32 v1, 0xffff, v1
; SI-NEXT: s_waitcnt vmcnt(0)
-; SI-NEXT: v_and_b32_e32 v34, 0xffff, v6
-; SI-NEXT: v_or_b32_e32 v34, v34, v35
-; SI-NEXT: buffer_store_dword v34, v0, s[0:3], 0 offen
+; SI-NEXT: v_and_b32_e32 v35, 0xffff, v6
+; SI-NEXT: v_or_b32_e32 v35, v35, v37
+; SI-NEXT: buffer_store_dword v35, v0, s[0:3], 0 offen
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_and_b32_e32 v34, 0xffff, v55
-; SI-NEXT: v_lshlrev_b32_e32 v35, 16, v40
-; SI-NEXT: v_or_b32_e32 v34, v34, v35
-; SI-NEXT: v_add_i32_e32 v35, vcc, 4, v0
-; SI-NEXT: buffer_store_dword v34, v35, s[0:3], 0 offen
+; SI-NEXT: v_and_b32_e32 v35, 0xffff, v55
+; SI-NEXT: v_lshlrev_b32_e32 v37, 16, v40
+; SI-NEXT: v_or_b32_e32 v35, v35, v37
+; SI-NEXT: v_add_i32_e32 v37, vcc, 4, v0
+; SI-NEXT: buffer_store_dword v35, v37, s[0:3], 0 offen
; SI-NEXT: buffer_load_dword v6, off, s[0:3], s32 offset:192 ; 4-byte Folded Reload
+; SI-NEXT: v_lshlrev_b32_e32 v37, 16, v59
; SI-NEXT: s_waitcnt vmcnt(0) expcnt(0)
-; SI-NEXT: v_and_b32_e32 v34, 0xffff, v6
-; SI-NEXT: v_or_b32_e32 v1, v34, v1
-; SI-NEXT: v_add_i32_e32 v34, vcc, 8, v0
-; SI-NEXT: buffer_store_dword v1, v34, s[0:3], 0 offen
-; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_and_b32_e32 v1, 0xffff, v52
-; SI-NEXT: v_lshlrev_b32_e32 v34, 16, v53
-; SI-NEXT: v_or_b32_e32 v1, v1, v34
-; SI-NEXT: v_add_i32_e32 v34, vcc, 12, v0
-; SI-NEXT: buffer_store_dword v1, v34, s[0:3], 0 offen
-; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_and_b32_e32 v1, 0xffff, v58
-; SI-NEXT: v_or_b32_e32 v1, v1, v8
-; SI-NEXT: v_add_i32_e32 v8, vcc, 16, v0
-; SI-NEXT: buffer_store_dword v1, v8, s[0:3], 0 offen
-; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_and_b32_e32 v1, 0xffff, v48
-; SI-NEXT: v_lshlrev_b32_e32 v8, 16, v49
-; SI-NEXT: v_or_b32_e32 v1, v1, v8
-; SI-NEXT: v_add_i32_e32 v8, vcc, 20, v0
-; SI-NEXT: buffer_store_dword v1, v8, s[0:3], 0 offen
-; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_and_b32_e32 v1, 0xffff, v57
-; SI-NEXT: v_lshlrev_b32_e32 v8, 16, v14
-; SI-NEXT: v_or_b32_e32 v1, v1, v8
-; SI-NEXT: v_add_i32_e32 v8, vcc, 24, v0
-; SI-NEXT: buffer_store_dword v1, v8, s[0:3], 0 offen
+; SI-NEXT: v_and_b32_e32 v35, 0xffff, v6
+; SI-NEXT: v_or_b32_e32 v35, v35, v37
+; SI-NEXT: v_add_i32_e32 v37, vcc, 8, v0
+; SI-NEXT: buffer_store_dword v35, v37, s[0:3], 0 offen
+; SI-NEXT: s_waitcnt expcnt(0)
+; SI-NEXT: v_and_b32_e32 v35, 0xffff, v52
+; SI-NEXT: v_lshlrev_b32_e32 v37, 16, v53
+; SI-NEXT: v_or_b32_e32 v35, v35, v37
+; SI-NEXT: v_add_i32_e32 v37, vcc, 12, v0
+; SI-NEXT: buffer_store_dword v35, v37, s[0:3], 0 offen
+; SI-NEXT: buffer_load_dword v6, off, s[0:3], s32 offset:188 ; 4-byte Folded Reload
+; SI-NEXT: s_waitcnt vmcnt(0) expcnt(0)
+; SI-NEXT: v_and_b32_e32 v35, 0xffff, v6
+; SI-NEXT: v_or_b32_e32 v5, v35, v5
+; SI-NEXT: v_add_i32_e32 v35, vcc, 16, v0
+; SI-NEXT: buffer_store_dword v5, v35, s[0:3], 0 offen
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_and_b32_e32 v1, 0xffff, v37
-; SI-NEXT: v_lshlrev_b32_e32 v8, 16, v38
-; SI-NEXT: v_or_b32_e32 v1, v1, v8
-; SI-NEXT: v_add_i32_e32 v8, vcc, 28, v0
-; SI-NEXT: buffer_store_dword v1, v8, s[0:3], 0 offen
-; SI-NEXT: buffer_load_dword v6, off, s[0:3], s32 offset:204 ; 4-byte Folded Reload
+; SI-NEXT: v_and_b32_e32 v5, 0xffff, v48
+; SI-NEXT: v_lshlrev_b32_e32 v35, 16, v49
+; SI-NEXT: v_or_b32_e32 v5, v5, v35
+; SI-NEXT: v_add_i32_e32 v35, vcc, 20, v0
+; SI-NEXT: buffer_store_dword v5, v35, s[0:3], 0 offen
+; SI-NEXT: s_waitcnt expcnt(0)
+; SI-NEXT: v_and_b32_e32 v5, 0xffff, v58
+; SI-NEXT: v_or_b32_e32 v5, v5, v14
+; SI-NEXT: v_add_i32_e32 v14, vcc, 24, v0
+; SI-NEXT: buffer_store_dword v5, v14, s[0:3], 0 offen
+; SI-NEXT: s_waitcnt expcnt(0)
+; SI-NEXT: v_lshlrev_b32_e32 v5, 16, v38
+; SI-NEXT: v_or_b32_e32 v2, v2, v5
+; SI-NEXT: v_add_i32_e32 v5, vcc, 28, v0
+; SI-NEXT: buffer_store_dword v2, v5, s[0:3], 0 offen
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_and_b32_e32 v1, 0xffff, v56
-; SI-NEXT: s_waitcnt vmcnt(0)
-; SI-NEXT: v_lshlrev_b32_e32 v8, 16, v6
-; SI-NEXT: v_or_b32_e32 v1, v1, v8
-; SI-NEXT: v_add_i32_e32 v8, vcc, 32, v0
-; SI-NEXT: buffer_store_dword v1, v8, s[0:3], 0 offen
+; SI-NEXT: v_and_b32_e32 v2, 0xffff, v56
+; SI-NEXT: v_lshlrev_b32_e32 v5, 16, v17
+; SI-NEXT: v_or_b32_e32 v2, v2, v5
+; SI-NEXT: v_add_i32_e32 v5, vcc, 32, v0
+; SI-NEXT: buffer_store_dword v2, v5, s[0:3], 0 offen
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_and_b32_e32 v1, 0xffff, v2
; SI-NEXT: v_lshlrev_b32_e32 v2, 16, v20
; SI-NEXT: v_or_b32_e32 v1, v1, v2
; SI-NEXT: v_add_i32_e32 v2, vcc, 36, v0
@@ -52401,7 +52445,7 @@ define <60 x i16> @bitcast_v60f16_to_v60i16(<60 x half> %a, i32 %b) {
; SI-NEXT: buffer_store_dword v1, v2, s[0:3], 0 offen
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_and_b32_e32 v1, 0xffff, v11
-; SI-NEXT: v_lshlrev_b32_e32 v2, 16, v5
+; SI-NEXT: v_lshlrev_b32_e32 v2, 16, v8
; SI-NEXT: v_or_b32_e32 v1, v1, v2
; SI-NEXT: v_add_i32_e32 v2, vcc, 44, v0
; SI-NEXT: buffer_store_dword v1, v2, s[0:3], 0 offen
@@ -52442,7 +52486,7 @@ define <60 x i16> @bitcast_v60f16_to_v60i16(<60 x half> %a, i32 %b) {
; SI-NEXT: v_add_i32_e32 v2, vcc, 0x44, v0
; SI-NEXT: buffer_store_dword v1, v2, s[0:3], 0 offen
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_and_b32_e32 v1, 0xffff, v59
+; SI-NEXT: v_and_b32_e32 v1, 0xffff, v62
; SI-NEXT: v_lshlrev_b32_e32 v2, 16, v43
; SI-NEXT: v_or_b32_e32 v1, v1, v2
; SI-NEXT: v_add_i32_e32 v2, vcc, 0x48, v0
@@ -52472,28 +52516,26 @@ define <60 x i16> @bitcast_v60f16_to_v60i16(<60 x half> %a, i32 %b) {
; SI-NEXT: v_add_i32_e32 v2, vcc, 0x58, v0
; SI-NEXT: buffer_store_dword v1, v2, s[0:3], 0 offen
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_and_b32_e32 v1, 0xffff, v63
-; SI-NEXT: v_lshlrev_b32_e32 v2, 16, v62
+; SI-NEXT: v_and_b32_e32 v1, 0xffff, v57
+; SI-NEXT: v_lshlrev_b32_e32 v2, 16, v63
; SI-NEXT: v_or_b32_e32 v1, v1, v2
; SI-NEXT: v_add_i32_e32 v2, vcc, 0x5c, v0
; SI-NEXT: buffer_store_dword v1, v2, s[0:3], 0 offen
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_and_b32_e32 v1, 0xffff, v36
-; SI-NEXT: v_lshlrev_b32_e32 v2, 16, v45
+; SI-NEXT: v_lshlrev_b32_e32 v2, 16, v32
; SI-NEXT: v_or_b32_e32 v1, v1, v2
; SI-NEXT: v_add_i32_e32 v2, vcc, 0x60, v0
; SI-NEXT: buffer_store_dword v1, v2, s[0:3], 0 offen
; SI-NEXT: s_waitcnt expcnt(0)
; SI-NEXT: v_and_b32_e32 v1, 0xffff, v31
-; SI-NEXT: v_lshlrev_b32_e32 v2, 16, v32
+; SI-NEXT: v_lshlrev_b32_e32 v2, 16, v34
; SI-NEXT: v_or_b32_e32 v1, v1, v2
; SI-NEXT: v_add_i32_e32 v2, vcc, 0x64, v0
; SI-NEXT: buffer_store_dword v1, v2, s[0:3], 0 offen
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:188 ; 4-byte Folded Reload
+; SI-NEXT: v_and_b32_e32 v1, 0xffff, v33
; SI-NEXT: v_lshlrev_b32_e32 v2, 16, v42
-; SI-NEXT: s_waitcnt vmcnt(0)
-; SI-NEXT: v_and_b32_e32 v1, 0xffff, v1
; SI-NEXT: v_or_b32_e32 v1, v1, v2
; SI-NEXT: v_add_i32_e32 v2, vcc, 0x68, v0
; SI-NEXT: buffer_store_dword v1, v2, s[0:3], 0 offen
@@ -53188,9 +53230,9 @@ define inreg <60 x i16> @bitcast_v60f16_to_v60i16_scalar(<60 x half> inreg %a, i
; SI-NEXT: v_cvt_f16_f32_e32 v48, v5
; SI-NEXT: v_cvt_f16_f32_e32 v49, v6
; SI-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:132 ; 4-byte Folded Spill
-; SI-NEXT: v_cvt_f16_f32_e32 v42, v7
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_cvt_f16_f32_e32 v1, v8
+; SI-NEXT: v_cvt_f16_f32_e32 v1, v7
+; SI-NEXT: v_cvt_f16_f32_e32 v42, v8
; SI-NEXT: v_cvt_f16_f32_e32 v37, v9
; SI-NEXT: v_cvt_f16_f32_e32 v38, v10
; SI-NEXT: v_cvt_f16_f32_e32 v5, v11
@@ -53275,8 +53317,7 @@ define inreg <60 x i16> @bitcast_v60f16_to_v60i16_scalar(<60 x half> inreg %a, i
; SI-NEXT: v_mov_b32_e32 v33, v11
; SI-NEXT: v_mov_b32_e32 v11, v8
; SI-NEXT: v_mov_b32_e32 v8, v5
-; SI-NEXT: v_mov_b32_e32 v5, v42
-; SI-NEXT: v_mov_b32_e32 v42, v1
+; SI-NEXT: v_mov_b32_e32 v5, v1
; SI-NEXT: s_cbranch_vccnz .LBB59_5
; SI-NEXT: ; %bb.4: ; %cmp.true
; SI-NEXT: buffer_load_dword v14, off, s[0:3], s32 offset:132 ; 4-byte Folded Reload
@@ -54285,53 +54326,62 @@ define inreg <60 x i16> @bitcast_v60f16_to_v60i16_scalar(<60 x half> inreg %a, i
; GFX11-TRUE16-NEXT: s_and_not1_b32 vcc_lo, exec_lo, s46
; GFX11-TRUE16-NEXT: s_cbranch_vccnz .LBB59_4
; GFX11-TRUE16-NEXT: .LBB59_2: ; %cmp.true
-; GFX11-TRUE16-NEXT: v_mov_b16_e32 v14.l, v9.h
-; GFX11-TRUE16-NEXT: v_and_b32_e32 v9, 0xffff, v9
; GFX11-TRUE16-NEXT: v_mov_b16_e32 v12.l, v11.h
; GFX11-TRUE16-NEXT: v_and_b32_e32 v11, 0xffff, v11
+; GFX11-TRUE16-NEXT: v_mov_b16_e32 v14.l, v9.h
+; GFX11-TRUE16-NEXT: v_and_b32_e32 v9, 0xffff, v9
; GFX11-TRUE16-NEXT: v_mov_b16_e32 v13.l, v10.h
; GFX11-TRUE16-NEXT: v_and_b32_e32 v10, 0xffff, v10
-; GFX11-TRUE16-NEXT: v_lshl_or_b32 v9, v14, 16, v9
-; GFX11-TRUE16-NEXT: v_mov_b16_e32 v14.l, v6.h
-; GFX11-TRUE16-NEXT: v_and_b32_e32 v6, 0xffff, v6
; GFX11-TRUE16-NEXT: v_lshl_or_b32 v11, v12, 16, v11
; GFX11-TRUE16-NEXT: v_mov_b16_e32 v12.l, v8.h
+; GFX11-TRUE16-NEXT: v_lshl_or_b32 v9, v14, 16, v9
; GFX11-TRUE16-NEXT: v_and_b32_e32 v8, 0xffff, v8
+; GFX11-TRUE16-NEXT: v_mov_b16_e32 v14.l, v6.h
+; GFX11-TRUE16-NEXT: v_and_b32_e32 v6, 0xffff, v6
+; GFX11-TRUE16-NEXT: v_mov_b16_e32 v15.l, v5.h
+; GFX11-TRUE16-NEXT: v_and_b32_e32 v5, 0xffff, v5
; GFX11-TRUE16-NEXT: v_lshl_or_b32 v10, v13, 16, v10
-; GFX11-TRUE16-NEXT: v_lshl_or_b32 v6, v14, 16, v6
-; GFX11-TRUE16-NEXT: v_mov_b16_e32 v14.l, v1.h
-; GFX11-TRUE16-NEXT: v_and_b32_e32 v1, 0xffff, v1
; GFX11-TRUE16-NEXT: v_mov_b16_e32 v13.l, v7.h
; GFX11-TRUE16-NEXT: v_and_b32_e32 v7, 0xffff, v7
-; GFX11-TRUE16-NEXT: v_mov_b16_e32 v15.l, v5.h
-; GFX11-TRUE16-NEXT: v_and_b32_e32 v5, 0xffff, v5
; GFX11-TRUE16-NEXT: v_lshl_or_b32 v8, v12, 16, v8
+; GFX11-TRUE16-NEXT: v_lshl_or_b32 v6, v14, 16, v6
+; GFX11-TRUE16-NEXT: v_lshl_or_b32 v5, v15, 16, v5
; GFX11-TRUE16-NEXT: v_mov_b16_e32 v12.l, v4.h
; GFX11-TRUE16-NEXT: v_and_b32_e32 v4, 0xffff, v4
-; GFX11-TRUE16-NEXT: s_pack_ll_b32_b16 s15, s24, s15
-; GFX11-TRUE16-NEXT: v_lshl_or_b32 v1, v14, 16, v1
-; GFX11-TRUE16-NEXT: v_pk_add_f16 v14, 0x200, s15 op_sel_hi:[0,1]
-; GFX11-TRUE16-NEXT: s_mov_b32 s15, s22
+; GFX11-TRUE16-NEXT: v_mov_b16_e32 v14.l, v1.h
+; GFX11-TRUE16-NEXT: v_and_b32_e32 v1, 0xffff, v1
+; GFX11-TRUE16-NEXT: v_mov_b16_e32 v15.l, v0.h
+; GFX11-TRUE16-NEXT: v_and_b32_e32 v0, 0xffff, v0
; GFX11-TRUE16-NEXT: v_lshl_or_b32 v7, v13, 16, v7
-; GFX11-TRUE16-NEXT: v_lshl_or_b32 v5, v15, 16, v5
; GFX11-TRUE16-NEXT: v_lshl_or_b32 v4, v12, 16, v4
; GFX11-TRUE16-NEXT: v_mov_b16_e32 v12.l, v3.h
; GFX11-TRUE16-NEXT: v_and_b32_e32 v3, 0xffff, v3
; GFX11-TRUE16-NEXT: v_mov_b16_e32 v13.l, v2.h
; GFX11-TRUE16-NEXT: v_and_b32_e32 v2, 0xffff, v2
-; GFX11-TRUE16-NEXT: v_mov_b16_e32 v15.l, v0.h
-; GFX11-TRUE16-NEXT: v_and_b32_e32 v0, 0xffff, v0
+; GFX11-TRUE16-NEXT: v_lshl_or_b32 v1, v14, 16, v1
+; GFX11-TRUE16-NEXT: v_lshl_or_b32 v0, v15, 16, v0
+; GFX11-TRUE16-NEXT: s_pack_ll_b32_b16 s15, s24, s15
+; GFX11-TRUE16-NEXT: v_lshl_or_b32 v3, v12, 16, v3
+; GFX11-TRUE16-NEXT: v_pk_add_f16 v14, 0x200, s15 op_sel_hi:[0,1]
+; GFX11-TRUE16-NEXT: s_mov_b32 s15, s22
+; GFX11-TRUE16-NEXT: v_lshl_or_b32 v2, v13, 16, v2
+; GFX11-TRUE16-NEXT: v_pk_add_f16 v1, 0x200, v1 op_sel_hi:[0,1]
+; GFX11-TRUE16-NEXT: v_pk_add_f16 v0, 0x200, v0 op_sel_hi:[0,1]
; GFX11-TRUE16-NEXT: s_pack_ll_b32_b16 s14, s23, s14
; GFX11-TRUE16-NEXT: s_pack_ll_b32_b16 s11, s15, s11
; GFX11-TRUE16-NEXT: v_pk_add_f16 v18, 0x200, s14 op_sel_hi:[0,1]
; GFX11-TRUE16-NEXT: s_mov_b32 s14, s21
; GFX11-TRUE16-NEXT: v_pk_add_f16 v19, 0x200, s11 op_sel_hi:[0,1]
; GFX11-TRUE16-NEXT: s_mov_b32 s11, s20
-; GFX11-TRUE16-NEXT: v_lshl_or_b32 v3, v12, 16, v3
-; GFX11-TRUE16-NEXT: v_lshl_or_b32 v2, v13, 16, v2
-; GFX11-TRUE16-NEXT: v_lshl_or_b32 v0, v15, 16, v0
+; GFX11-TRUE16-NEXT: v_pk_add_f16 v6, 0x200, v6 op_sel_hi:[0,1]
+; GFX11-TRUE16-NEXT: v_pk_add_f16 v5, 0x200, v5 op_sel_hi:[0,1]
+; GFX11-TRUE16-NEXT: v_pk_add_f16 v4, 0x200, v4 op_sel_hi:[0,1]
+; GFX11-TRUE16-NEXT: v_pk_add_f16 v3, 0x200, v3 op_sel_hi:[0,1]
+; GFX11-TRUE16-NEXT: v_pk_add_f16 v2, 0x200, v2 op_sel_hi:[0,1]
; GFX11-TRUE16-NEXT: s_pack_ll_b32_b16 s10, s14, s10
; GFX11-TRUE16-NEXT: s_pack_ll_b32_b16 s9, s11, s9
+; GFX11-TRUE16-NEXT: v_lshrrev_b32_e32 v64, 16, v0
+; GFX11-TRUE16-NEXT: v_lshrrev_b32_e32 v65, 16, v1
; GFX11-TRUE16-NEXT: v_pk_add_f16 v20, 0x200, s10 op_sel_hi:[0,1]
; GFX11-TRUE16-NEXT: s_mov_b32 s10, s19
; GFX11-TRUE16-NEXT: v_pk_add_f16 v21, 0x200, s9 op_sel_hi:[0,1]
@@ -54343,13 +54393,13 @@ define inreg <60 x i16> @bitcast_v60f16_to_v60i16_scalar(<60 x half> inreg %a, i
; GFX11-TRUE16-NEXT: v_pk_add_f16 v9, 0x200, v9 op_sel_hi:[0,1]
; GFX11-TRUE16-NEXT: v_pk_add_f16 v8, 0x200, v8 op_sel_hi:[0,1]
; GFX11-TRUE16-NEXT: v_pk_add_f16 v7, 0x200, v7 op_sel_hi:[0,1]
-; GFX11-TRUE16-NEXT: v_pk_add_f16 v6, 0x200, v6 op_sel_hi:[0,1]
-; GFX11-TRUE16-NEXT: v_pk_add_f16 v5, 0x200, v5 op_sel_hi:[0,1]
-; GFX11-TRUE16-NEXT: v_pk_add_f16 v4, 0x200, v4 op_sel_hi:[0,1]
-; GFX11-TRUE16-NEXT: v_pk_add_f16 v3, 0x200, v3 op_sel_hi:[0,1]
-; GFX11-TRUE16-NEXT: v_pk_add_f16 v2, 0x200, v2 op_sel_hi:[0,1]
-; GFX11-TRUE16-NEXT: v_pk_add_f16 v1, 0x200, v1 op_sel_hi:[0,1]
-; GFX11-TRUE16-NEXT: v_pk_add_f16 v0, 0x200, v0 op_sel_hi:[0,1]
+; GFX11-TRUE16-NEXT: v_mov_b16_e32 v0.h, v64.l
+; GFX11-TRUE16-NEXT: v_mov_b16_e32 v1.h, v65.l
+; GFX11-TRUE16-NEXT: v_lshrrev_b32_e32 v64, 16, v2
+; GFX11-TRUE16-NEXT: v_lshrrev_b32_e32 v65, 16, v3
+; GFX11-TRUE16-NEXT: v_lshrrev_b32_e32 v66, 16, v4
+; GFX11-TRUE16-NEXT: v_lshrrev_b32_e32 v67, 16, v5
+; GFX11-TRUE16-NEXT: v_lshrrev_b32_e32 v68, 16, v6
; GFX11-TRUE16-NEXT: v_pk_add_f16 v22, 0x200, s8 op_sel_hi:[0,1]
; GFX11-TRUE16-NEXT: s_mov_b32 s8, s17
; GFX11-TRUE16-NEXT: v_pk_add_f16 v23, 0x200, s7 op_sel_hi:[0,1]
@@ -54376,18 +54426,16 @@ define inreg <60 x i16> @bitcast_v60f16_to_v60i16_scalar(<60 x half> inreg %a, i
; GFX11-TRUE16-NEXT: v_pk_add_f16 v32, 0x200, s2 op_sel_hi:[0,1]
; GFX11-TRUE16-NEXT: v_pk_add_f16 v31, 0x200, s3 op_sel_hi:[0,1]
; GFX11-TRUE16-NEXT: v_pk_add_f16 v30, 0x200, s5 op_sel_hi:[0,1]
-; GFX11-TRUE16-NEXT: v_lshrrev_b32_e32 v64, 16, v0
-; GFX11-TRUE16-NEXT: v_lshrrev_b32_e32 v65, 16, v1
-; GFX11-TRUE16-NEXT: v_lshrrev_b32_e32 v66, 16, v2
-; GFX11-TRUE16-NEXT: v_lshrrev_b32_e32 v67, 16, v3
-; GFX11-TRUE16-NEXT: v_lshrrev_b32_e32 v68, 16, v4
-; GFX11-TRUE16-NEXT: v_lshrrev_b32_e32 v69, 16, v5
-; GFX11-TRUE16-NEXT: v_lshrrev_b32_e32 v70, 16, v6
-; GFX11-TRUE16-NEXT: v_lshrrev_b32_e32 v71, 16, v7
-; GFX11-TRUE16-NEXT: v_lshrrev_b32_e32 v80, 16, v8
-; GFX11-TRUE16-NEXT: v_lshrrev_b32_e32 v81, 16, v9
-; GFX11-TRUE16-NEXT: v_lshrrev_b32_e32 v82, 16, v10
-; GFX11-TRUE16-NEXT: v_lshrrev_b32_e32 v83, 16, v11
+; GFX11-TRUE16-NEXT: v_mov_b16_e32 v2.h, v64.l
+; GFX11-TRUE16-NEXT: v_mov_b16_e32 v3.h, v65.l
+; GFX11-TRUE16-NEXT: v_mov_b16_e32 v4.h, v66.l
+; GFX11-TRUE16-NEXT: v_mov_b16_e32 v5.h, v67.l
+; GFX11-TRUE16-NEXT: v_mov_b16_e32 v6.h, v68.l
+; GFX11-TRUE16-NEXT: v_lshrrev_b32_e32 v64, 16, v7
+; GFX11-TRUE16-NEXT: v_lshrrev_b32_e32 v65, 16, v8
+; GFX11-TRUE16-NEXT: v_lshrrev_b32_e32 v66, 16, v9
+; GFX11-TRUE16-NEXT: v_lshrrev_b32_e32 v67, 16, v10
+; GFX11-TRUE16-NEXT: v_lshrrev_b32_e32 v68, 16, v11
; GFX11-TRUE16-NEXT: v_lshrrev_b32_e32 v38, 16, v37
; GFX11-TRUE16-NEXT: v_lshrrev_b32_e32 v39, 16, v36
; GFX11-TRUE16-NEXT: v_lshrrev_b32_e32 v48, 16, v32
@@ -54403,21 +54451,14 @@ define inreg <60 x i16> @bitcast_v60f16_to_v60i16_scalar(<60 x half> inreg %a, i
; GFX11-TRUE16-NEXT: v_lshrrev_b32_e32 v50, 16, v14
; GFX11-TRUE16-NEXT: v_lshrrev_b32_e32 v51, 16, v13
; GFX11-TRUE16-NEXT: v_lshrrev_b32_e32 v52, 16, v12
-; GFX11-TRUE16-NEXT: v_lshrrev_b32_e32 v25, 16, v17
+; GFX11-TRUE16-NEXT: v_lshrrev_b32_e32 v27, 16, v17
; GFX11-TRUE16-NEXT: v_lshrrev_b32_e32 v26, 16, v16
-; GFX11-TRUE16-NEXT: v_lshrrev_b32_e32 v27, 16, v15
-; GFX11-TRUE16-NEXT: v_mov_b16_e32 v0.h, v64.l
-; GFX11-TRUE16-NEXT: v_mov_b16_e32 v1.h, v65.l
-; GFX11-TRUE16-NEXT: v_mov_b16_e32 v2.h, v66.l
-; GFX11-TRUE16-NEXT: v_mov_b16_e32 v3.h, v67.l
-; GFX11-TRUE16-NEXT: v_mov_b16_e32 v4.h, v68.l
-; GFX11-TRUE16-NEXT: v_mov_b16_e32 v5.h, v69.l
-; GFX11-TRUE16-NEXT: v_mov_b16_e32 v6.h, v70.l
-; GFX11-TRUE16-NEXT: v_mov_b16_e32 v7.h, v71.l
-; GFX11-TRUE16-NEXT: v_mov_b16_e32 v8.h, v80.l
-; GFX11-TRUE16-NEXT: v_mov_b16_e32 v9.h, v81.l
-; GFX11-TRUE16-NEXT: v_mov_b16_e32 v10.h, v82.l
-; GFX11-TRUE16-NEXT: v_mov_b16_e32 v11.h, v83.l
+; GFX11-TRUE16-NEXT: v_lshrrev_b32_e32 v25, 16, v15
+; GFX11-TRUE16-NEXT: v_mov_b16_e32 v7.h, v64.l
+; GFX11-TRUE16-NEXT: v_mov_b16_e32 v8.h, v65.l
+; GFX11-TRUE16-NEXT: v_mov_b16_e32 v9.h, v66.l
+; GFX11-TRUE16-NEXT: v_mov_b16_e32 v10.h, v67.l
+; GFX11-TRUE16-NEXT: v_mov_b16_e32 v11.h, v68.l
; GFX11-TRUE16-NEXT: s_branch .LBB59_5
; GFX11-TRUE16-NEXT: .LBB59_3:
; GFX11-TRUE16-NEXT: s_branch .LBB59_2
@@ -54440,9 +54481,9 @@ define inreg <60 x i16> @bitcast_v60f16_to_v60i16_scalar(<60 x half> inreg %a, i
; GFX11-TRUE16-NEXT: v_mov_b16_e32 v32.l, s2
; GFX11-TRUE16-NEXT: v_mov_b16_e32 v36.l, s1
; GFX11-TRUE16-NEXT: v_mov_b16_e32 v37.l, s0
-; GFX11-TRUE16-NEXT: v_mov_b16_e32 v27.l, s45
+; GFX11-TRUE16-NEXT: v_mov_b16_e32 v25.l, s45
; GFX11-TRUE16-NEXT: v_mov_b16_e32 v26.l, s44
-; GFX11-TRUE16-NEXT: v_mov_b16_e32 v25.l, s43
+; GFX11-TRUE16-NEXT: v_mov_b16_e32 v27.l, s43
; GFX11-TRUE16-NEXT: v_mov_b16_e32 v52.l, s42
; GFX11-TRUE16-NEXT: v_mov_b16_e32 v51.l, s41
; GFX11-TRUE16-NEXT: v_mov_b16_e32 v50.l, s15
@@ -54491,8 +54532,8 @@ define inreg <60 x i16> @bitcast_v60f16_to_v60i16_scalar(<60 x half> inreg %a, i
; GFX11-TRUE16-NEXT: v_and_b32_e32 v0, 0xffff, v0
; GFX11-TRUE16-NEXT: v_mov_b16_e32 v20.l, v1.h
; GFX11-TRUE16-NEXT: v_and_b32_e32 v1, 0xffff, v1
-; GFX11-TRUE16-NEXT: v_lshl_or_b32 v15, v25, 16, v17
-; GFX11-TRUE16-NEXT: v_lshl_or_b32 v17, v27, 16, v18
+; GFX11-TRUE16-NEXT: v_lshl_or_b32 v15, v27, 16, v17
+; GFX11-TRUE16-NEXT: v_lshl_or_b32 v17, v25, 16, v18
; GFX11-TRUE16-NEXT: v_lshl_or_b32 v18, v19, 16, v0
; GFX11-TRUE16-NEXT: v_mov_b16_e32 v0.l, v2.h
; GFX11-TRUE16-NEXT: v_lshl_or_b32 v19, v20, 16, v1
diff --git a/llvm/test/CodeGen/AMDGPU/amdgcn.bitcast.96bit.ll b/llvm/test/CodeGen/AMDGPU/amdgcn.bitcast.96bit.ll
index f888f4f3b1407..8a2602e668a15 100644
--- a/llvm/test/CodeGen/AMDGPU/amdgcn.bitcast.96bit.ll
+++ b/llvm/test/CodeGen/AMDGPU/amdgcn.bitcast.96bit.ll
@@ -7895,11 +7895,11 @@ define <12 x i8> @bitcast_v6bf16_to_v12i8(<6 x bfloat> %a, i32 %b) {
; GFX11-TRUE16-NEXT: v_lshrrev_b32_e32 v5, 8, v13
; GFX11-TRUE16-NEXT: v_lshrrev_b32_e32 v1, 8, v12
; GFX11-TRUE16-NEXT: v_lshrrev_b64 v[16:17], 24, v[10:11]
-; GFX11-TRUE16-NEXT: v_lshrrev_b64 v[14:15], 24, v[12:13]
-; GFX11-TRUE16-NEXT: v_mov_b16_e32 v0.h, v12.l
-; GFX11-TRUE16-NEXT: v_mov_b16_e32 v4.h, v13.l
; GFX11-TRUE16-NEXT: v_mov_b16_e32 v6.h, v13.h
+; GFX11-TRUE16-NEXT: v_lshrrev_b64 v[14:15], 24, v[12:13]
; GFX11-TRUE16-NEXT: v_mov_b16_e32 v8.h, v10.l
+; GFX11-TRUE16-NEXT: v_mov_b16_e32 v4.h, v13.l
+; GFX11-TRUE16-NEXT: v_mov_b16_e32 v0.h, v12.l
; GFX11-TRUE16-NEXT: .LBB38_2: ; %Flow
; GFX11-TRUE16-NEXT: s_and_not1_saveexec_b32 s0, s0
; GFX11-TRUE16-NEXT: s_cbranch_execz .LBB38_4
@@ -7969,6 +7969,7 @@ define <12 x i8> @bitcast_v6bf16_to_v12i8(<6 x bfloat> %a, i32 %b) {
; GFX11-TRUE16-NEXT: v_lshrrev_b32_e32 v9, 8, v10
; GFX11-TRUE16-NEXT: .LBB38_4: ; %end
; GFX11-TRUE16-NEXT: s_or_b32 exec_lo, exec_lo, s0
+; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1)
; GFX11-TRUE16-NEXT: v_mov_b16_e32 v0.l, v0.h
; GFX11-TRUE16-NEXT: v_mov_b16_e32 v2.l, v12.h
; GFX11-TRUE16-NEXT: v_mov_b16_e32 v3.l, v14.l
@@ -11041,62 +11042,57 @@ define <12 x i8> @bitcast_v6i16_to_v12i8(<6 x i16> %a, i32 %b) {
; VI-LABEL: bitcast_v6i16_to_v12i8:
; VI: ; %bb.0:
; VI-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; VI-NEXT: v_mov_b32_e32 v8, v2
+; VI-NEXT: v_mov_b32_e32 v14, v1
+; VI-NEXT: v_mov_b32_e32 v13, v0
; VI-NEXT: v_cmp_ne_u32_e32 vcc, 0, v3
-; VI-NEXT: v_lshrrev_b32_e32 v10, 16, v2
-; VI-NEXT: v_lshrrev_b32_e32 v6, 16, v1
-; VI-NEXT: v_lshrrev_b32_e32 v13, 16, v0
-; VI-NEXT: ; implicit-def: $vgpr16
-; VI-NEXT: ; implicit-def: $vgpr15
+; VI-NEXT: v_lshrrev_b32_e32 v10, 16, v8
+; VI-NEXT: v_lshrrev_b32_e32 v6, 16, v14
+; VI-NEXT: v_lshrrev_b32_e32 v2, 16, v13
+; VI-NEXT: ; implicit-def: $vgpr1
; VI-NEXT: ; implicit-def: $vgpr3
-; VI-NEXT: ; implicit-def: $vgpr14
; VI-NEXT: ; implicit-def: $vgpr5
; VI-NEXT: ; implicit-def: $vgpr7
-; VI-NEXT: ; implicit-def: $vgpr8
; VI-NEXT: ; implicit-def: $vgpr9
; VI-NEXT: ; implicit-def: $vgpr11
; VI-NEXT: s_and_saveexec_b64 s[4:5], vcc
; VI-NEXT: s_xor_b64 s[4:5], exec, s[4:5]
-; VI-NEXT: s_cbranch_execz .LBB46_2
; VI-NEXT: ; %bb.1: ; %cmp.false
-; VI-NEXT: v_lshrrev_b32_e32 v9, 8, v2
-; VI-NEXT: v_lshrrev_b32_e32 v7, 24, v1
-; VI-NEXT: v_lshrrev_b32_e32 v5, 8, v1
-; VI-NEXT: v_lshrrev_b32_e32 v15, 8, v0
-; VI-NEXT: v_lshrrev_b64 v[11:12], 24, v[2:3]
-; VI-NEXT: v_lshrrev_b64 v[3:4], 24, v[0:1]
-; VI-NEXT: v_mov_b32_e32 v16, v0
-; VI-NEXT: v_mov_b32_e32 v14, v1
-; VI-NEXT: v_mov_b32_e32 v8, v2
-; VI-NEXT: ; implicit-def: $vgpr1
-; VI-NEXT: ; implicit-def: $vgpr2
-; VI-NEXT: .LBB46_2: ; %Flow
+; VI-NEXT: v_lshrrev_b32_e32 v9, 8, v8
+; VI-NEXT: v_lshrrev_b32_e32 v7, 24, v14
+; VI-NEXT: v_lshrrev_b32_e32 v5, 8, v14
+; VI-NEXT: v_lshrrev_b64 v[11:12], 24, v[8:9]
+; VI-NEXT: v_lshrrev_b64 v[3:4], 24, v[13:14]
+; VI-NEXT: v_lshrrev_b32_e32 v1, 8, v13
+; VI-NEXT: ; %bb.2: ; %Flow
; VI-NEXT: s_andn2_saveexec_b64 s[4:5], s[4:5]
; VI-NEXT: s_cbranch_execz .LBB46_4
; VI-NEXT: ; %bb.3: ; %cmp.true
; VI-NEXT: v_mov_b32_e32 v3, 3
-; VI-NEXT: v_add_u16_sdwa v6, v1, v3 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_1 src1_sel:DWORD
-; VI-NEXT: v_add_u16_sdwa v13, v0, v3 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_1 src1_sel:DWORD
-; VI-NEXT: v_add_u16_e32 v14, 3, v1
-; VI-NEXT: v_lshlrev_b32_e32 v1, 16, v6
-; VI-NEXT: v_add_u16_e32 v16, 3, v0
-; VI-NEXT: v_lshlrev_b32_e32 v0, 16, v13
-; VI-NEXT: v_or_b32_e32 v1, v14, v1
-; VI-NEXT: v_or_b32_e32 v0, v16, v0
-; VI-NEXT: v_add_u16_sdwa v10, v2, v3 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_1 src1_sel:DWORD
-; VI-NEXT: v_add_u16_e32 v8, 3, v2
-; VI-NEXT: v_lshlrev_b32_e32 v2, 16, v10
+; VI-NEXT: v_add_u16_sdwa v6, v14, v3 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_1 src1_sel:DWORD
+; VI-NEXT: v_add_u16_e32 v15, 3, v14
+; VI-NEXT: v_lshlrev_b32_e32 v0, 16, v6
+; VI-NEXT: v_add_u16_sdwa v2, v13, v3 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_1 src1_sel:DWORD
+; VI-NEXT: v_add_u16_sdwa v10, v8, v3 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_1 src1_sel:DWORD
+; VI-NEXT: v_or_b32_e32 v1, v15, v0
+; VI-NEXT: v_add_u16_e32 v14, 3, v13
+; VI-NEXT: v_lshlrev_b32_e32 v0, 16, v2
+; VI-NEXT: v_add_u16_e32 v16, 3, v8
+; VI-NEXT: v_lshlrev_b32_e32 v3, 16, v10
+; VI-NEXT: v_or_b32_e32 v0, v14, v0
+; VI-NEXT: v_or_b32_e32 v7, v16, v3
; VI-NEXT: v_lshrrev_b64 v[3:4], 24, v[0:1]
-; VI-NEXT: v_or_b32_e32 v2, v8, v2
-; VI-NEXT: v_lshrrev_b64 v[11:12], 24, v[2:3]
-; VI-NEXT: v_lshrrev_b32_e32 v9, 8, v2
+; VI-NEXT: v_lshrrev_b64 v[11:12], 24, v[7:8]
+; VI-NEXT: v_lshrrev_b32_e32 v9, 8, v7
; VI-NEXT: v_lshrrev_b32_e32 v5, 8, v1
-; VI-NEXT: v_lshrrev_b32_e32 v15, 8, v0
+; VI-NEXT: v_lshrrev_b32_e32 v1, 8, v0
; VI-NEXT: v_bfe_u32 v7, v6, 8, 8
+; VI-NEXT: v_mov_b32_e32 v13, v14
+; VI-NEXT: v_mov_b32_e32 v14, v15
+; VI-NEXT: v_mov_b32_e32 v8, v16
; VI-NEXT: .LBB46_4: ; %end
; VI-NEXT: s_or_b64 exec, exec, s[4:5]
-; VI-NEXT: v_mov_b32_e32 v0, v16
-; VI-NEXT: v_mov_b32_e32 v1, v15
-; VI-NEXT: v_mov_b32_e32 v2, v13
+; VI-NEXT: v_mov_b32_e32 v0, v13
; VI-NEXT: v_mov_b32_e32 v4, v14
; VI-NEXT: s_setpc_b64 s[30:31]
;
diff --git a/llvm/test/CodeGen/AMDGPU/atomic_optimizations_global_pointer.ll b/llvm/test/CodeGen/AMDGPU/atomic_optimizations_global_pointer.ll
index f979d01e495ba..e9071a012a2a5 100644
--- a/llvm/test/CodeGen/AMDGPU/atomic_optimizations_global_pointer.ll
+++ b/llvm/test/CodeGen/AMDGPU/atomic_optimizations_global_pointer.ll
@@ -7628,8 +7628,8 @@ define amdgpu_kernel void @uniform_add_i8(ptr addrspace(1) %result, ptr addrspac
; GFX7LESS-NEXT: s_load_dwordx4 s[0:3], s[4:5], 0x9
; GFX7LESS-NEXT: s_load_dword s10, s[4:5], 0xd
; GFX7LESS-NEXT: v_mbcnt_lo_u32_b32_e64 v0, s6, 0
-; GFX7LESS-NEXT: v_mbcnt_hi_u32_b32_e32 v4, s7, v0
-; GFX7LESS-NEXT: v_cmp_eq_u32_e32 vcc, 0, v4
+; GFX7LESS-NEXT: v_mbcnt_hi_u32_b32_e32 v2, s7, v0
+; GFX7LESS-NEXT: v_cmp_eq_u32_e32 vcc, 0, v2
; GFX7LESS-NEXT: ; implicit-def: $vgpr0
; GFX7LESS-NEXT: s_and_saveexec_b64 s[8:9], vcc
; GFX7LESS-NEXT: s_cbranch_execz .LBB13_4
@@ -7649,27 +7649,28 @@ define amdgpu_kernel void @uniform_add_i8(ptr addrspace(1) %result, ptr addrspac
; GFX7LESS-NEXT: s_not_b32 s13, s12
; GFX7LESS-NEXT: s_lshl_b32 s14, s6, s11
; GFX7LESS-NEXT: s_waitcnt lgkmcnt(0)
-; GFX7LESS-NEXT: v_mov_b32_e32 v1, s15
+; GFX7LESS-NEXT: v_mov_b32_e32 v0, s15
; GFX7LESS-NEXT: s_mov_b32 s6, -1
; GFX7LESS-NEXT: .LBB13_2: ; %atomicrmw.start
; GFX7LESS-NEXT: ; =>This Inner Loop Header: Depth=1
-; GFX7LESS-NEXT: v_add_i32_e32 v0, vcc, s14, v1
-; GFX7LESS-NEXT: v_and_b32_e32 v0, s12, v0
+; GFX7LESS-NEXT: v_mov_b32_e32 v4, v0
; GFX7LESS-NEXT: s_waitcnt expcnt(0)
-; GFX7LESS-NEXT: v_and_b32_e32 v2, s13, v1
-; GFX7LESS-NEXT: v_or_b32_e32 v0, v2, v0
-; GFX7LESS-NEXT: v_mov_b32_e32 v3, v1
-; GFX7LESS-NEXT: v_mov_b32_e32 v2, v0
-; GFX7LESS-NEXT: buffer_atomic_cmpswap v[2:3], off, s[4:7], 0 glc
+; GFX7LESS-NEXT: v_add_i32_e32 v0, vcc, s14, v4
+; GFX7LESS-NEXT: v_and_b32_e32 v0, s12, v0
+; GFX7LESS-NEXT: v_and_b32_e32 v1, s13, v4
+; GFX7LESS-NEXT: v_or_b32_e32 v3, v1, v0
+; GFX7LESS-NEXT: v_mov_b32_e32 v0, v3
+; GFX7LESS-NEXT: v_mov_b32_e32 v1, v4
+; GFX7LESS-NEXT: buffer_atomic_cmpswap v[0:1], off, s[4:7], 0 glc
; GFX7LESS-NEXT: s_waitcnt vmcnt(0)
-; GFX7LESS-NEXT: v_cmp_eq_u32_e32 vcc, v2, v1
+; GFX7LESS-NEXT: v_cmp_eq_u32_e32 vcc, v0, v4
; GFX7LESS-NEXT: s_or_b64 s[2:3], vcc, s[2:3]
-; GFX7LESS-NEXT: v_mov_b32_e32 v1, v2
; GFX7LESS-NEXT: s_andn2_b64 exec, exec, s[2:3]
; GFX7LESS-NEXT: s_cbranch_execnz .LBB13_2
; GFX7LESS-NEXT: ; %bb.3: ; %atomicrmw.end
; GFX7LESS-NEXT: s_or_b64 exec, exec, s[2:3]
-; GFX7LESS-NEXT: v_lshrrev_b32_e32 v0, s11, v2
+; GFX7LESS-NEXT: s_waitcnt expcnt(0)
+; GFX7LESS-NEXT: v_lshrrev_b32_e32 v0, s11, v0
; GFX7LESS-NEXT: .LBB13_4: ; %Flow
; GFX7LESS-NEXT: s_or_b64 exec, exec, s[8:9]
; GFX7LESS-NEXT: s_waitcnt lgkmcnt(0)
@@ -7679,7 +7680,7 @@ define amdgpu_kernel void @uniform_add_i8(ptr addrspace(1) %result, ptr addrspac
; GFX7LESS-NEXT: v_readfirstlane_b32 s4, v0
; GFX7LESS-NEXT: s_and_b32 s5, s10, 0xff
; GFX7LESS-NEXT: v_mov_b32_e32 v0, s4
-; GFX7LESS-NEXT: v_mad_u32_u24 v0, s5, v4, v0
+; GFX7LESS-NEXT: v_mad_u32_u24 v0, s5, v2, v0
; GFX7LESS-NEXT: buffer_store_byte v0, off, s[0:3], 0
; GFX7LESS-NEXT: s_endpgm
;
@@ -7689,8 +7690,8 @@ define amdgpu_kernel void @uniform_add_i8(ptr addrspace(1) %result, ptr addrspac
; GFX8-NEXT: s_load_dword s10, s[4:5], 0x34
; GFX8-NEXT: s_mov_b64 s[6:7], exec
; GFX8-NEXT: v_mbcnt_lo_u32_b32 v0, s6, 0
-; GFX8-NEXT: v_mbcnt_hi_u32_b32 v4, s7, v0
-; GFX8-NEXT: v_cmp_eq_u32_e32 vcc, 0, v4
+; GFX8-NEXT: v_mbcnt_hi_u32_b32 v2, s7, v0
+; GFX8-NEXT: v_cmp_eq_u32_e32 vcc, 0, v2
; GFX8-NEXT: ; implicit-def: $vgpr0
; GFX8-NEXT: s_and_saveexec_b64 s[8:9], vcc
; GFX8-NEXT: s_cbranch_execz .LBB13_4
@@ -7709,27 +7710,27 @@ define amdgpu_kernel void @uniform_add_i8(ptr addrspace(1) %result, ptr addrspac
; GFX8-NEXT: s_lshl_b32 s14, s2, s11
; GFX8-NEXT: s_mov_b64 s[2:3], 0
; GFX8-NEXT: s_waitcnt lgkmcnt(0)
-; GFX8-NEXT: v_mov_b32_e32 v1, s7
+; GFX8-NEXT: v_mov_b32_e32 v0, s7
; GFX8-NEXT: s_mov_b32 s7, 0xf000
; GFX8-NEXT: s_mov_b32 s6, -1
; GFX8-NEXT: .LBB13_2: ; %atomicrmw.start
; GFX8-NEXT: ; =>This Inner Loop Header: Depth=1
-; GFX8-NEXT: v_add_u32_e32 v0, vcc, s14, v1
-; GFX8-NEXT: v_and_b32_e32 v2, s13, v1
+; GFX8-NEXT: v_mov_b32_e32 v4, v0
+; GFX8-NEXT: v_add_u32_e32 v0, vcc, s14, v4
+; GFX8-NEXT: v_and_b32_e32 v1, s13, v4
; GFX8-NEXT: v_and_b32_e32 v0, s12, v0
-; GFX8-NEXT: v_or_b32_e32 v0, v2, v0
-; GFX8-NEXT: v_mov_b32_e32 v3, v1
-; GFX8-NEXT: v_mov_b32_e32 v2, v0
-; GFX8-NEXT: buffer_atomic_cmpswap v[2:3], off, s[4:7], 0 glc
+; GFX8-NEXT: v_or_b32_e32 v3, v1, v0
+; GFX8-NEXT: v_mov_b32_e32 v0, v3
+; GFX8-NEXT: v_mov_b32_e32 v1, v4
+; GFX8-NEXT: buffer_atomic_cmpswap v[0:1], off, s[4:7], 0 glc
; GFX8-NEXT: s_waitcnt vmcnt(0)
-; GFX8-NEXT: v_cmp_eq_u32_e32 vcc, v2, v1
+; GFX8-NEXT: v_cmp_eq_u32_e32 vcc, v0, v4
; GFX8-NEXT: s_or_b64 s[2:3], vcc, s[2:3]
-; GFX8-NEXT: v_mov_b32_e32 v1, v2
; GFX8-NEXT: s_andn2_b64 exec, exec, s[2:3]
; GFX8-NEXT: s_cbranch_execnz .LBB13_2
; GFX8-NEXT: ; %bb.3: ; %atomicrmw.end
; GFX8-NEXT: s_or_b64 exec, exec, s[2:3]
-; GFX8-NEXT: v_lshrrev_b32_e32 v0, s11, v2
+; GFX8-NEXT: v_lshrrev_b32_e32 v0, s11, v0
; GFX8-NEXT: .LBB13_4: ; %Flow
; GFX8-NEXT: s_or_b64 exec, exec, s[8:9]
; GFX8-NEXT: v_and_b32_e32 v0, 0xff, v0
@@ -7738,7 +7739,7 @@ define amdgpu_kernel void @uniform_add_i8(ptr addrspace(1) %result, ptr addrspac
; GFX8-NEXT: s_waitcnt lgkmcnt(0)
; GFX8-NEXT: s_mov_b32 s3, 0xf000
; GFX8-NEXT: s_mov_b32 s2, -1
-; GFX8-NEXT: v_mad_u16 v0, s10, v4, v0
+; GFX8-NEXT: v_mad_u16 v0, s10, v2, v0
; GFX8-NEXT: buffer_store_byte v0, off, s[0:3], 0
; GFX8-NEXT: s_endpgm
;
@@ -7748,8 +7749,8 @@ define amdgpu_kernel void @uniform_add_i8(ptr addrspace(1) %result, ptr addrspac
; GFX9-NEXT: s_load_dword s10, s[4:5], 0x34
; GFX9-NEXT: s_mov_b64 s[6:7], exec
; GFX9-NEXT: v_mbcnt_lo_u32_b32 v0, s6, 0
-; GFX9-NEXT: v_mbcnt_hi_u32_b32 v4, s7, v0
-; GFX9-NEXT: v_cmp_eq_u32_e32 vcc, 0, v4
+; GFX9-NEXT: v_mbcnt_hi_u32_b32 v2, s7, v0
+; GFX9-NEXT: v_cmp_eq_u32_e32 vcc, 0, v2
; GFX9-NEXT: ; implicit-def: $vgpr0
; GFX9-NEXT: s_and_saveexec_b64 s[8:9], vcc
; GFX9-NEXT: s_cbranch_execz .LBB13_4
@@ -7768,26 +7769,26 @@ define amdgpu_kernel void @uniform_add_i8(ptr addrspace(1) %result, ptr addrspac
; GFX9-NEXT: s_lshl_b32 s14, s2, s11
; GFX9-NEXT: s_mov_b64 s[2:3], 0
; GFX9-NEXT: s_waitcnt lgkmcnt(0)
-; GFX9-NEXT: v_mov_b32_e32 v1, s7
+; GFX9-NEXT: v_mov_b32_e32 v0, s7
; GFX9-NEXT: s_mov_b32 s7, 0xf000
; GFX9-NEXT: s_mov_b32 s6, -1
; GFX9-NEXT: .LBB13_2: ; %atomicrmw.start
; GFX9-NEXT: ; =>This Inner Loop Header: Depth=1
-; GFX9-NEXT: v_add_u32_e32 v0, s14, v1
+; GFX9-NEXT: v_mov_b32_e32 v4, v0
+; GFX9-NEXT: v_add_u32_e32 v0, s14, v4
; GFX9-NEXT: v_and_b32_e32 v0, s12, v0
-; GFX9-NEXT: v_and_or_b32 v0, v1, s13, v0
-; GFX9-NEXT: v_mov_b32_e32 v3, v1
-; GFX9-NEXT: v_mov_b32_e32 v2, v0
-; GFX9-NEXT: buffer_atomic_cmpswap v[2:3], off, s[4:7], 0 glc
+; GFX9-NEXT: v_and_or_b32 v3, v4, s13, v0
+; GFX9-NEXT: v_mov_b32_e32 v0, v3
+; GFX9-NEXT: v_mov_b32_e32 v1, v4
+; GFX9-NEXT: buffer_atomic_cmpswap v[0:1], off, s[4:7], 0 glc
; GFX9-NEXT: s_waitcnt vmcnt(0)
-; GFX9-NEXT: v_cmp_eq_u32_e32 vcc, v2, v1
+; GFX9-NEXT: v_cmp_eq_u32_e32 vcc, v0, v4
; GFX9-NEXT: s_or_b64 s[2:3], vcc, s[2:3]
-; GFX9-NEXT: v_mov_b32_e32 v1, v2
; GFX9-NEXT: s_andn2_b64 exec, exec, s[2:3]
; GFX9-NEXT: s_cbranch_execnz .LBB13_2
; GFX9-NEXT: ; %bb.3: ; %atomicrmw.end
; GFX9-NEXT: s_or_b64 exec, exec, s[2:3]
-; GFX9-NEXT: v_lshrrev_b32_e32 v0, s11, v2
+; GFX9-NEXT: v_lshrrev_b32_e32 v0, s11, v0
; GFX9-NEXT: .LBB13_4: ; %Flow
; GFX9-NEXT: s_or_b64 exec, exec, s[8:9]
; GFX9-NEXT: v_and_b32_e32 v0, 0xff, v0
@@ -7796,7 +7797,7 @@ define amdgpu_kernel void @uniform_add_i8(ptr addrspace(1) %result, ptr addrspac
; GFX9-NEXT: s_waitcnt lgkmcnt(0)
; GFX9-NEXT: s_mov_b32 s3, 0xf000
; GFX9-NEXT: s_mov_b32 s2, -1
-; GFX9-NEXT: v_mad_legacy_u16 v0, s10, v4, v0
+; GFX9-NEXT: v_mad_legacy_u16 v0, s10, v2, v0
; GFX9-NEXT: buffer_store_byte v0, off, s[0:3], 0
; GFX9-NEXT: s_endpgm
;
@@ -7807,9 +7808,9 @@ define amdgpu_kernel void @uniform_add_i8(ptr addrspace(1) %result, ptr addrspac
; GFX1064-NEXT: s_load_dword s10, s[4:5], 0x34
; GFX1064-NEXT: s_mov_b64 s[6:7], exec
; GFX1064-NEXT: v_mbcnt_lo_u32_b32 v0, s6, 0
-; GFX1064-NEXT: v_mbcnt_hi_u32_b32 v4, s7, v0
+; GFX1064-NEXT: v_mbcnt_hi_u32_b32 v2, s7, v0
; GFX1064-NEXT: ; implicit-def: $vgpr0
-; GFX1064-NEXT: v_cmp_eq_u32_e32 vcc, 0, v4
+; GFX1064-NEXT: v_cmp_eq_u32_e32 vcc, 0, v2
; GFX1064-NEXT: s_and_saveexec_b64 s[8:9], vcc
; GFX1064-NEXT: s_cbranch_execz .LBB13_4
; GFX1064-NEXT: ; %bb.1:
@@ -7828,32 +7829,32 @@ define amdgpu_kernel void @uniform_add_i8(ptr addrspace(1) %result, ptr addrspac
; GFX1064-NEXT: s_mov_b32 s7, 0x31016000
; GFX1064-NEXT: s_mov_b32 s6, -1
; GFX1064-NEXT: s_waitcnt lgkmcnt(0)
-; GFX1064-NEXT: v_mov_b32_e32 v1, s3
+; GFX1064-NEXT: v_mov_b32_e32 v0, s3
; GFX1064-NEXT: s_mov_b64 s[2:3], 0
; GFX1064-NEXT: .LBB13_2: ; %atomicrmw.start
; GFX1064-NEXT: ; =>This Inner Loop Header: Depth=1
-; GFX1064-NEXT: v_add_nc_u32_e32 v0, s14, v1
+; GFX1064-NEXT: v_mov_b32_e32 v4, v0
+; GFX1064-NEXT: v_add_nc_u32_e32 v0, s14, v4
; GFX1064-NEXT: v_and_b32_e32 v0, s12, v0
-; GFX1064-NEXT: v_and_or_b32 v0, v1, s13, v0
-; GFX1064-NEXT: v_mov_b32_e32 v3, v1
-; GFX1064-NEXT: v_mov_b32_e32 v2, v0
-; GFX1064-NEXT: buffer_atomic_cmpswap v[2:3], off, s[4:7], 0 glc
+; GFX1064-NEXT: v_and_or_b32 v3, v4, s13, v0
+; GFX1064-NEXT: v_mov_b32_e32 v0, v3
+; GFX1064-NEXT: v_mov_b32_e32 v1, v4
+; GFX1064-NEXT: buffer_atomic_cmpswap v[0:1], off, s[4:7], 0 glc
; GFX1064-NEXT: s_waitcnt vmcnt(0)
-; GFX1064-NEXT: v_cmp_eq_u32_e32 vcc, v2, v1
-; GFX1064-NEXT: v_mov_b32_e32 v1, v2
+; GFX1064-NEXT: v_cmp_eq_u32_e32 vcc, v0, v4
; GFX1064-NEXT: s_or_b64 s[2:3], vcc, s[2:3]
; GFX1064-NEXT: s_andn2_b64 exec, exec, s[2:3]
; GFX1064-NEXT: s_cbranch_execnz .LBB13_2
; GFX1064-NEXT: ; %bb.3: ; %atomicrmw.end
; GFX1064-NEXT: s_or_b64 exec, exec, s[2:3]
-; GFX1064-NEXT: v_lshrrev_b32_e32 v0, s11, v2
+; GFX1064-NEXT: v_lshrrev_b32_e32 v0, s11, v0
; GFX1064-NEXT: .LBB13_4: ; %Flow
; GFX1064-NEXT: s_or_b64 exec, exec, s[8:9]
; GFX1064-NEXT: v_and_b32_e32 v0, 0xff, v0
; GFX1064-NEXT: s_waitcnt lgkmcnt(0)
; GFX1064-NEXT: s_mov_b32 s3, 0x31016000
; GFX1064-NEXT: v_readfirstlane_b32 s2, v0
-; GFX1064-NEXT: v_mad_u16 v0, s10, v4, s2
+; GFX1064-NEXT: v_mad_u16 v0, s10, v2, s2
; GFX1064-NEXT: s_mov_b32 s2, -1
; GFX1064-NEXT: buffer_store_byte v0, off, s[0:3], 0
; GFX1064-NEXT: s_endpgm
@@ -7865,9 +7866,9 @@ define amdgpu_kernel void @uniform_add_i8(ptr addrspace(1) %result, ptr addrspac
; GFX1032-NEXT: s_load_dword s8, s[4:5], 0x34
; GFX1032-NEXT: s_mov_b32 s6, exec_lo
; GFX1032-NEXT: s_mov_b32 s10, 0
-; GFX1032-NEXT: v_mbcnt_lo_u32_b32 v4, s6, 0
+; GFX1032-NEXT: v_mbcnt_lo_u32_b32 v2, s6, 0
; GFX1032-NEXT: ; implicit-def: $vgpr0
-; GFX1032-NEXT: v_cmp_eq_u32_e32 vcc_lo, 0, v4
+; GFX1032-NEXT: v_cmp_eq_u32_e32 vcc_lo, 0, v2
; GFX1032-NEXT: s_and_saveexec_b32 s9, vcc_lo
; GFX1032-NEXT: s_cbranch_execz .LBB13_4
; GFX1032-NEXT: ; %bb.1:
@@ -7885,32 +7886,32 @@ define amdgpu_kernel void @uniform_add_i8(ptr addrspace(1) %result, ptr addrspac
; GFX1032-NEXT: s_lshl_b32 s12, s6, s2
; GFX1032-NEXT: s_mov_b32 s6, -1
; GFX1032-NEXT: s_waitcnt lgkmcnt(0)
-; GFX1032-NEXT: v_mov_b32_e32 v1, s7
+; GFX1032-NEXT: v_mov_b32_e32 v0, s7
; GFX1032-NEXT: s_mov_b32 s7, 0x31016000
; GFX1032-NEXT: .LBB13_2: ; %atomicrmw.start
; GFX1032-NEXT: ; =>This Inner Loop Header: Depth=1
-; GFX1032-NEXT: v_add_nc_u32_e32 v0, s12, v1
+; GFX1032-NEXT: v_mov_b32_e32 v4, v0
+; GFX1032-NEXT: v_add_nc_u32_e32 v0, s12, v4
; GFX1032-NEXT: v_and_b32_e32 v0, s3, v0
-; GFX1032-NEXT: v_and_or_b32 v0, v1, s11, v0
-; GFX1032-NEXT: v_mov_b32_e32 v3, v1
-; GFX1032-NEXT: v_mov_b32_e32 v2, v0
-; GFX1032-NEXT: buffer_atomic_cmpswap v[2:3], off, s[4:7], 0 glc
+; GFX1032-NEXT: v_and_or_b32 v3, v4, s11, v0
+; GFX1032-NEXT: v_mov_b32_e32 v0, v3
+; GFX1032-NEXT: v_mov_b32_e32 v1, v4
+; GFX1032-NEXT: buffer_atomic_cmpswap v[0:1], off, s[4:7], 0 glc
; GFX1032-NEXT: s_waitcnt vmcnt(0)
-; GFX1032-NEXT: v_cmp_eq_u32_e32 vcc_lo, v2, v1
-; GFX1032-NEXT: v_mov_b32_e32 v1, v2
+; GFX1032-NEXT: v_cmp_eq_u32_e32 vcc_lo, v0, v4
; GFX1032-NEXT: s_or_b32 s10, vcc_lo, s10
; GFX1032-NEXT: s_andn2_b32 exec_lo, exec_lo, s10
; GFX1032-NEXT: s_cbranch_execnz .LBB13_2
; GFX1032-NEXT: ; %bb.3: ; %atomicrmw.end
; GFX1032-NEXT: s_or_b32 exec_lo, exec_lo, s10
-; GFX1032-NEXT: v_lshrrev_b32_e32 v0, s2, v2
+; GFX1032-NEXT: v_lshrrev_b32_e32 v0, s2, v0
; GFX1032-NEXT: .LBB13_4: ; %Flow
; GFX1032-NEXT: s_or_b32 exec_lo, exec_lo, s9
; GFX1032-NEXT: v_and_b32_e32 v0, 0xff, v0
; GFX1032-NEXT: s_waitcnt lgkmcnt(0)
; GFX1032-NEXT: s_mov_b32 s3, 0x31016000
; GFX1032-NEXT: v_readfirstlane_b32 s2, v0
-; GFX1032-NEXT: v_mad_u16 v0, s8, v4, s2
+; GFX1032-NEXT: v_mad_u16 v0, s8, v2, s2
; GFX1032-NEXT: s_mov_b32 s2, -1
; GFX1032-NEXT: buffer_store_byte v0, off, s[0:3], 0
; GFX1032-NEXT: s_endpgm
@@ -7924,9 +7925,9 @@ define amdgpu_kernel void @uniform_add_i8(ptr addrspace(1) %result, ptr addrspac
; GFX1164-TRUE16-NEXT: s_mov_b64 s[8:9], exec
; GFX1164-TRUE16-NEXT: v_mbcnt_lo_u32_b32 v0, s6, 0
; GFX1164-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
-; GFX1164-TRUE16-NEXT: v_mbcnt_hi_u32_b32 v4, s7, v0
+; GFX1164-TRUE16-NEXT: v_mbcnt_hi_u32_b32 v2, s7, v0
; GFX1164-TRUE16-NEXT: ; implicit-def: $vgpr0_lo16
-; GFX1164-TRUE16-NEXT: v_cmpx_eq_u32_e32 0, v4
+; GFX1164-TRUE16-NEXT: v_cmpx_eq_u32_e32 0, v2
; GFX1164-TRUE16-NEXT: s_cbranch_execz .LBB13_4
; GFX1164-TRUE16-NEXT: ; %bb.1:
; GFX1164-TRUE16-NEXT: s_waitcnt lgkmcnt(0)
@@ -7945,29 +7946,29 @@ define amdgpu_kernel void @uniform_add_i8(ptr addrspace(1) %result, ptr addrspac
; GFX1164-TRUE16-NEXT: s_mov_b32 s7, 0x31016000
; GFX1164-TRUE16-NEXT: s_mov_b32 s6, -1
; GFX1164-TRUE16-NEXT: s_waitcnt lgkmcnt(0)
-; GFX1164-TRUE16-NEXT: v_mov_b32_e32 v1, s3
+; GFX1164-TRUE16-NEXT: v_mov_b32_e32 v0, s3
; GFX1164-TRUE16-NEXT: s_mov_b64 s[2:3], 0
; GFX1164-TRUE16-NEXT: .LBB13_2: ; %atomicrmw.start
; GFX1164-TRUE16-NEXT: ; =>This Inner Loop Header: Depth=1
; GFX1164-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
-; GFX1164-TRUE16-NEXT: v_add_nc_u32_e32 v0, s14, v1
+; GFX1164-TRUE16-NEXT: v_mov_b32_e32 v4, v0
+; GFX1164-TRUE16-NEXT: v_add_nc_u32_e32 v0, s14, v4
+; GFX1164-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
; GFX1164-TRUE16-NEXT: v_and_b32_e32 v0, s12, v0
-; GFX1164-TRUE16-NEXT: s_waitcnt_depctr 0xfff
-; GFX1164-TRUE16-NEXT: v_and_or_b32 v0, v1, s13, v0
-; GFX1164-TRUE16-NEXT: v_mov_b32_e32 v3, v1
-; GFX1164-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_2)
-; GFX1164-TRUE16-NEXT: v_mov_b32_e32 v2, v0
-; GFX1164-TRUE16-NEXT: buffer_atomic_cmpswap_b32 v[2:3], off, s[4:7], 0 glc
+; GFX1164-TRUE16-NEXT: v_and_or_b32 v3, v4, s13, v0
+; GFX1164-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1)
+; GFX1164-TRUE16-NEXT: v_mov_b32_e32 v0, v3
+; GFX1164-TRUE16-NEXT: v_mov_b32_e32 v1, v4
+; GFX1164-TRUE16-NEXT: buffer_atomic_cmpswap_b32 v[0:1], off, s[4:7], 0 glc
; GFX1164-TRUE16-NEXT: s_waitcnt vmcnt(0)
-; GFX1164-TRUE16-NEXT: v_cmp_eq_u32_e32 vcc, v2, v1
-; GFX1164-TRUE16-NEXT: v_mov_b32_e32 v1, v2
+; GFX1164-TRUE16-NEXT: v_cmp_eq_u32_e32 vcc, v0, v4
; GFX1164-TRUE16-NEXT: s_or_b64 s[2:3], vcc, s[2:3]
; GFX1164-TRUE16-NEXT: s_delay_alu instid0(SALU_CYCLE_1)
; GFX1164-TRUE16-NEXT: s_and_not1_b64 exec, exec, s[2:3]
; GFX1164-TRUE16-NEXT: s_cbranch_execnz .LBB13_2
; GFX1164-TRUE16-NEXT: ; %bb.3: ; %atomicrmw.end
; GFX1164-TRUE16-NEXT: s_or_b64 exec, exec, s[2:3]
-; GFX1164-TRUE16-NEXT: v_lshrrev_b32_e32 v0, s11, v2
+; GFX1164-TRUE16-NEXT: v_lshrrev_b32_e32 v0, s11, v0
; GFX1164-TRUE16-NEXT: .LBB13_4: ; %Flow
; GFX1164-TRUE16-NEXT: s_or_b64 exec, exec, s[8:9]
; GFX1164-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(SKIP_2) | instid1(VALU_DEP_1)
@@ -7976,7 +7977,7 @@ define amdgpu_kernel void @uniform_add_i8(ptr addrspace(1) %result, ptr addrspac
; GFX1164-TRUE16-NEXT: s_mov_b32 s3, 0x31016000
; GFX1164-TRUE16-NEXT: v_readfirstlane_b32 s2, v0
; GFX1164-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1)
-; GFX1164-TRUE16-NEXT: v_mad_u16 v0.l, s10, v4.l, s2
+; GFX1164-TRUE16-NEXT: v_mad_u16 v0.l, s10, v2.l, s2
; GFX1164-TRUE16-NEXT: s_mov_b32 s2, -1
; GFX1164-TRUE16-NEXT: buffer_store_b8 v0, off, s[0:3], 0
; GFX1164-TRUE16-NEXT: s_endpgm
@@ -7990,9 +7991,9 @@ define amdgpu_kernel void @uniform_add_i8(ptr addrspace(1) %result, ptr addrspac
; GFX1164-FAKE16-NEXT: s_mov_b64 s[8:9], exec
; GFX1164-FAKE16-NEXT: v_mbcnt_lo_u32_b32 v0, s6, 0
; GFX1164-FAKE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
-; GFX1164-FAKE16-NEXT: v_mbcnt_hi_u32_b32 v4, s7, v0
+; GFX1164-FAKE16-NEXT: v_mbcnt_hi_u32_b32 v2, s7, v0
; GFX1164-FAKE16-NEXT: ; implicit-def: $vgpr0
-; GFX1164-FAKE16-NEXT: v_cmpx_eq_u32_e32 0, v4
+; GFX1164-FAKE16-NEXT: v_cmpx_eq_u32_e32 0, v2
; GFX1164-FAKE16-NEXT: s_cbranch_execz .LBB13_4
; GFX1164-FAKE16-NEXT: ; %bb.1:
; GFX1164-FAKE16-NEXT: s_waitcnt lgkmcnt(0)
@@ -8010,29 +8011,29 @@ define amdgpu_kernel void @uniform_add_i8(ptr addrspace(1) %result, ptr addrspac
; GFX1164-FAKE16-NEXT: s_mov_b32 s7, 0x31016000
; GFX1164-FAKE16-NEXT: s_mov_b32 s6, -1
; GFX1164-FAKE16-NEXT: s_waitcnt lgkmcnt(0)
-; GFX1164-FAKE16-NEXT: v_mov_b32_e32 v1, s3
+; GFX1164-FAKE16-NEXT: v_mov_b32_e32 v0, s3
; GFX1164-FAKE16-NEXT: s_mov_b64 s[2:3], 0
; GFX1164-FAKE16-NEXT: .LBB13_2: ; %atomicrmw.start
; GFX1164-FAKE16-NEXT: ; =>This Inner Loop Header: Depth=1
; GFX1164-FAKE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
-; GFX1164-FAKE16-NEXT: v_add_nc_u32_e32 v0, s14, v1
+; GFX1164-FAKE16-NEXT: v_mov_b32_e32 v4, v0
+; GFX1164-FAKE16-NEXT: v_add_nc_u32_e32 v0, s14, v4
+; GFX1164-FAKE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
; GFX1164-FAKE16-NEXT: v_and_b32_e32 v0, s12, v0
-; GFX1164-FAKE16-NEXT: s_waitcnt_depctr 0xfff
-; GFX1164-FAKE16-NEXT: v_and_or_b32 v0, v1, s13, v0
-; GFX1164-FAKE16-NEXT: v_mov_b32_e32 v3, v1
-; GFX1164-FAKE16-NEXT: s_delay_alu instid0(VALU_DEP_2)
-; GFX1164-FAKE16-NEXT: v_mov_b32_e32 v2, v0
-; GFX1164-FAKE16-NEXT: buffer_atomic_cmpswap_b32 v[2:3], off, s[4:7], 0 glc
+; GFX1164-FAKE16-NEXT: v_and_or_b32 v3, v4, s13, v0
+; GFX1164-FAKE16-NEXT: s_delay_alu instid0(VALU_DEP_1)
+; GFX1164-FAKE16-NEXT: v_mov_b32_e32 v0, v3
+; GFX1164-FAKE16-NEXT: v_mov_b32_e32 v1, v4
+; GFX1164-FAKE16-NEXT: buffer_atomic_cmpswap_b32 v[0:1], off, s[4:7], 0 glc
; GFX1164-FAKE16-NEXT: s_waitcnt vmcnt(0)
-; GFX1164-FAKE16-NEXT: v_cmp_eq_u32_e32 vcc, v2, v1
-; GFX1164-FAKE16-NEXT: v_mov_b32_e32 v1, v2
+; GFX1164-FAKE16-NEXT: v_cmp_eq_u32_e32 vcc, v0, v4
; GFX1164-FAKE16-NEXT: s_or_b64 s[2:3], vcc, s[2:3]
; GFX1164-FAKE16-NEXT: s_delay_alu instid0(SALU_CYCLE_1)
; GFX1164-FAKE16-NEXT: s_and_not1_b64 exec, exec, s[2:3]
; GFX1164-FAKE16-NEXT: s_cbranch_execnz .LBB13_2
; GFX1164-FAKE16-NEXT: ; %bb.3: ; %atomicrmw.end
; GFX1164-FAKE16-NEXT: s_or_b64 exec, exec, s[2:3]
-; GFX1164-FAKE16-NEXT: v_lshrrev_b32_e32 v0, s11, v2
+; GFX1164-FAKE16-NEXT: v_lshrrev_b32_e32 v0, s11, v0
; GFX1164-FAKE16-NEXT: .LBB13_4: ; %Flow
; GFX1164-FAKE16-NEXT: s_or_b64 exec, exec, s[8:9]
; GFX1164-FAKE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(SKIP_2) | instid1(VALU_DEP_1)
@@ -8041,7 +8042,7 @@ define amdgpu_kernel void @uniform_add_i8(ptr addrspace(1) %result, ptr addrspac
; GFX1164-FAKE16-NEXT: s_mov_b32 s3, 0x31016000
; GFX1164-FAKE16-NEXT: v_readfirstlane_b32 s2, v0
; GFX1164-FAKE16-NEXT: s_delay_alu instid0(VALU_DEP_1)
-; GFX1164-FAKE16-NEXT: v_mad_u16 v0, s10, v4, s2
+; GFX1164-FAKE16-NEXT: v_mad_u16 v0, s10, v2, s2
; GFX1164-FAKE16-NEXT: s_mov_b32 s2, -1
; GFX1164-FAKE16-NEXT: buffer_store_b8 v0, off, s[0:3], 0
; GFX1164-FAKE16-NEXT: s_endpgm
@@ -8053,11 +8054,11 @@ define amdgpu_kernel void @uniform_add_i8(ptr addrspace(1) %result, ptr addrspac
; GFX1132-TRUE16-NEXT: s_load_b32 s8, s[4:5], 0x34
; GFX1132-TRUE16-NEXT: s_mov_b32 s6, exec_lo
; GFX1132-TRUE16-NEXT: s_mov_b32 s10, 0
-; GFX1132-TRUE16-NEXT: v_mbcnt_lo_u32_b32 v4, s6, 0
+; GFX1132-TRUE16-NEXT: v_mbcnt_lo_u32_b32 v2, s6, 0
; GFX1132-TRUE16-NEXT: s_mov_b32 s9, exec_lo
; GFX1132-TRUE16-NEXT: ; implicit-def: $vgpr0_lo16
; GFX1132-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1)
-; GFX1132-TRUE16-NEXT: v_cmpx_eq_u32_e32 0, v4
+; GFX1132-TRUE16-NEXT: v_cmpx_eq_u32_e32 0, v2
; GFX1132-TRUE16-NEXT: s_cbranch_execz .LBB13_4
; GFX1132-TRUE16-NEXT: ; %bb.1:
; GFX1132-TRUE16-NEXT: s_waitcnt lgkmcnt(0)
@@ -8075,27 +8076,27 @@ define amdgpu_kernel void @uniform_add_i8(ptr addrspace(1) %result, ptr addrspac
; GFX1132-TRUE16-NEXT: s_lshl_b32 s12, s6, s2
; GFX1132-TRUE16-NEXT: s_mov_b32 s6, -1
; GFX1132-TRUE16-NEXT: s_waitcnt lgkmcnt(0)
-; GFX1132-TRUE16-NEXT: v_mov_b32_e32 v1, s7
+; GFX1132-TRUE16-NEXT: v_mov_b32_e32 v0, s7
; GFX1132-TRUE16-NEXT: s_mov_b32 s7, 0x31016000
; GFX1132-TRUE16-NEXT: .LBB13_2: ; %atomicrmw.start
; GFX1132-TRUE16-NEXT: ; =>This Inner Loop Header: Depth=1
; GFX1132-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
-; GFX1132-TRUE16-NEXT: v_add_nc_u32_e32 v0, s12, v1
-; GFX1132-TRUE16-NEXT: v_and_b32_e32 v0, s3, v0
+; GFX1132-TRUE16-NEXT: v_mov_b32_e32 v4, v0
+; GFX1132-TRUE16-NEXT: v_add_nc_u32_e32 v0, s12, v4
; GFX1132-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
-; GFX1132-TRUE16-NEXT: v_and_or_b32 v0, v1, s11, v0
-; GFX1132-TRUE16-NEXT: v_dual_mov_b32 v3, v1 :: v_dual_mov_b32 v2, v0
-; GFX1132-TRUE16-NEXT: buffer_atomic_cmpswap_b32 v[2:3], off, s[4:7], 0 glc
+; GFX1132-TRUE16-NEXT: v_and_b32_e32 v0, s3, v0
+; GFX1132-TRUE16-NEXT: v_and_or_b32 v3, v4, s11, v0
+; GFX1132-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(SKIP_4) | instid1(SALU_CYCLE_1)
+; GFX1132-TRUE16-NEXT: v_dual_mov_b32 v0, v3 :: v_dual_mov_b32 v1, v4
+; GFX1132-TRUE16-NEXT: buffer_atomic_cmpswap_b32 v[0:1], off, s[4:7], 0 glc
; GFX1132-TRUE16-NEXT: s_waitcnt vmcnt(0)
-; GFX1132-TRUE16-NEXT: v_cmp_eq_u32_e32 vcc_lo, v2, v1
-; GFX1132-TRUE16-NEXT: v_mov_b32_e32 v1, v2
+; GFX1132-TRUE16-NEXT: v_cmp_eq_u32_e32 vcc_lo, v0, v4
; GFX1132-TRUE16-NEXT: s_or_b32 s10, vcc_lo, s10
-; GFX1132-TRUE16-NEXT: s_delay_alu instid0(SALU_CYCLE_1)
; GFX1132-TRUE16-NEXT: s_and_not1_b32 exec_lo, exec_lo, s10
; GFX1132-TRUE16-NEXT: s_cbranch_execnz .LBB13_2
; GFX1132-TRUE16-NEXT: ; %bb.3: ; %atomicrmw.end
; GFX1132-TRUE16-NEXT: s_or_b32 exec_lo, exec_lo, s10
-; GFX1132-TRUE16-NEXT: v_lshrrev_b32_e32 v0, s2, v2
+; GFX1132-TRUE16-NEXT: v_lshrrev_b32_e32 v0, s2, v0
; GFX1132-TRUE16-NEXT: .LBB13_4: ; %Flow
; GFX1132-TRUE16-NEXT: s_or_b32 exec_lo, exec_lo, s9
; GFX1132-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(SKIP_2) | instid1(VALU_DEP_1)
@@ -8104,7 +8105,7 @@ define amdgpu_kernel void @uniform_add_i8(ptr addrspace(1) %result, ptr addrspac
; GFX1132-TRUE16-NEXT: s_mov_b32 s3, 0x31016000
; GFX1132-TRUE16-NEXT: v_readfirstlane_b32 s2, v0
; GFX1132-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1)
-; GFX1132-TRUE16-NEXT: v_mad_u16 v0.l, s8, v4.l, s2
+; GFX1132-TRUE16-NEXT: v_mad_u16 v0.l, s8, v2.l, s2
; GFX1132-TRUE16-NEXT: s_mov_b32 s2, -1
; GFX1132-TRUE16-NEXT: buffer_store_b8 v0, off, s[0:3], 0
; GFX1132-TRUE16-NEXT: s_endpgm
@@ -8116,11 +8117,11 @@ define amdgpu_kernel void @uniform_add_i8(ptr addrspace(1) %result, ptr addrspac
; GFX1132-FAKE16-NEXT: s_load_b32 s8, s[4:5], 0x34
; GFX1132-FAKE16-NEXT: s_mov_b32 s6, exec_lo
; GFX1132-FAKE16-NEXT: s_mov_b32 s10, 0
-; GFX1132-FAKE16-NEXT: v_mbcnt_lo_u32_b32 v4, s6, 0
+; GFX1132-FAKE16-NEXT: v_mbcnt_lo_u32_b32 v2, s6, 0
; GFX1132-FAKE16-NEXT: s_mov_b32 s9, exec_lo
; GFX1132-FAKE16-NEXT: ; implicit-def: $vgpr0
; GFX1132-FAKE16-NEXT: s_delay_alu instid0(VALU_DEP_1)
-; GFX1132-FAKE16-NEXT: v_cmpx_eq_u32_e32 0, v4
+; GFX1132-FAKE16-NEXT: v_cmpx_eq_u32_e32 0, v2
; GFX1132-FAKE16-NEXT: s_cbranch_execz .LBB13_4
; GFX1132-FAKE16-NEXT: ; %bb.1:
; GFX1132-FAKE16-NEXT: s_waitcnt lgkmcnt(0)
@@ -8137,27 +8138,27 @@ define amdgpu_kernel void @uniform_add_i8(ptr addrspace(1) %result, ptr addrspac
; GFX1132-FAKE16-NEXT: s_lshl_b32 s12, s6, s2
; GFX1132-FAKE16-NEXT: s_mov_b32 s6, -1
; GFX1132-FAKE16-NEXT: s_waitcnt lgkmcnt(0)
-; GFX1132-FAKE16-NEXT: v_mov_b32_e32 v1, s7
+; GFX1132-FAKE16-NEXT: v_mov_b32_e32 v0, s7
; GFX1132-FAKE16-NEXT: s_mov_b32 s7, 0x31016000
; GFX1132-FAKE16-NEXT: .LBB13_2: ; %atomicrmw.start
; GFX1132-FAKE16-NEXT: ; =>This Inner Loop Header: Depth=1
; GFX1132-FAKE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
-; GFX1132-FAKE16-NEXT: v_add_nc_u32_e32 v0, s12, v1
-; GFX1132-FAKE16-NEXT: v_and_b32_e32 v0, s3, v0
+; GFX1132-FAKE16-NEXT: v_mov_b32_e32 v4, v0
+; GFX1132-FAKE16-NEXT: v_add_nc_u32_e32 v0, s12, v4
; GFX1132-FAKE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
-; GFX1132-FAKE16-NEXT: v_and_or_b32 v0, v1, s11, v0
-; GFX1132-FAKE16-NEXT: v_dual_mov_b32 v3, v1 :: v_dual_mov_b32 v2, v0
-; GFX1132-FAKE16-NEXT: buffer_atomic_cmpswap_b32 v[2:3], off, s[4:7], 0 glc
+; GFX1132-FAKE16-NEXT: v_and_b32_e32 v0, s3, v0
+; GFX1132-FAKE16-NEXT: v_and_or_b32 v3, v4, s11, v0
+; GFX1132-FAKE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(SKIP_4) | instid1(SALU_CYCLE_1)
+; GFX1132-FAKE16-NEXT: v_dual_mov_b32 v0, v3 :: v_dual_mov_b32 v1, v4
+; GFX1132-FAKE16-NEXT: buffer_atomic_cmpswap_b32 v[0:1], off, s[4:7], 0 glc
; GFX1132-FAKE16-NEXT: s_waitcnt vmcnt(0)
-; GFX1132-FAKE16-NEXT: v_cmp_eq_u32_e32 vcc_lo, v2, v1
-; GFX1132-FAKE16-NEXT: v_mov_b32_e32 v1, v2
+; GFX1132-FAKE16-NEXT: v_cmp_eq_u32_e32 vcc_lo, v0, v4
; GFX1132-FAKE16-NEXT: s_or_b32 s10, vcc_lo, s10
-; GFX1132-FAKE16-NEXT: s_delay_alu instid0(SALU_CYCLE_1)
; GFX1132-FAKE16-NEXT: s_and_not1_b32 exec_lo, exec_lo, s10
; GFX1132-FAKE16-NEXT: s_cbranch_execnz .LBB13_2
; GFX1132-FAKE16-NEXT: ; %bb.3: ; %atomicrmw.end
; GFX1132-FAKE16-NEXT: s_or_b32 exec_lo, exec_lo, s10
-; GFX1132-FAKE16-NEXT: v_lshrrev_b32_e32 v0, s2, v2
+; GFX1132-FAKE16-NEXT: v_lshrrev_b32_e32 v0, s2, v0
; GFX1132-FAKE16-NEXT: .LBB13_4: ; %Flow
; GFX1132-FAKE16-NEXT: s_or_b32 exec_lo, exec_lo, s9
; GFX1132-FAKE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(SKIP_2) | instid1(VALU_DEP_1)
@@ -8166,7 +8167,7 @@ define amdgpu_kernel void @uniform_add_i8(ptr addrspace(1) %result, ptr addrspac
; GFX1132-FAKE16-NEXT: s_mov_b32 s3, 0x31016000
; GFX1132-FAKE16-NEXT: v_readfirstlane_b32 s2, v0
; GFX1132-FAKE16-NEXT: s_delay_alu instid0(VALU_DEP_1)
-; GFX1132-FAKE16-NEXT: v_mad_u16 v0, s8, v4, s2
+; GFX1132-FAKE16-NEXT: v_mad_u16 v0, s8, v2, s2
; GFX1132-FAKE16-NEXT: s_mov_b32 s2, -1
; GFX1132-FAKE16-NEXT: buffer_store_b8 v0, off, s[0:3], 0
; GFX1132-FAKE16-NEXT: s_endpgm
@@ -8180,9 +8181,9 @@ define amdgpu_kernel void @uniform_add_i8(ptr addrspace(1) %result, ptr addrspac
; GFX1264-TRUE16-NEXT: s_mov_b64 s[8:9], exec
; GFX1264-TRUE16-NEXT: v_mbcnt_lo_u32_b32 v0, s6, 0
; GFX1264-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
-; GFX1264-TRUE16-NEXT: v_mbcnt_hi_u32_b32 v4, s7, v0
+; GFX1264-TRUE16-NEXT: v_mbcnt_hi_u32_b32 v2, s7, v0
; GFX1264-TRUE16-NEXT: ; implicit-def: $vgpr0_lo16
-; GFX1264-TRUE16-NEXT: v_cmpx_eq_u32_e32 0, v4
+; GFX1264-TRUE16-NEXT: v_cmpx_eq_u32_e32 0, v2
; GFX1264-TRUE16-NEXT: s_cbranch_execz .LBB13_4
; GFX1264-TRUE16-NEXT: ; %bb.1:
; GFX1264-TRUE16-NEXT: s_wait_kmcnt 0x0
@@ -8203,28 +8204,29 @@ define amdgpu_kernel void @uniform_add_i8(ptr addrspace(1) %result, ptr addrspac
; GFX1264-TRUE16-NEXT: s_mov_b32 s7, 0x31016000
; GFX1264-TRUE16-NEXT: s_mov_b32 s6, -1
; GFX1264-TRUE16-NEXT: s_wait_kmcnt 0x0
-; GFX1264-TRUE16-NEXT: v_mov_b32_e32 v1, s3
+; GFX1264-TRUE16-NEXT: v_mov_b32_e32 v0, s3
; GFX1264-TRUE16-NEXT: s_mov_b64 s[2:3], 0
; GFX1264-TRUE16-NEXT: .LBB13_2: ; %atomicrmw.start
; GFX1264-TRUE16-NEXT: ; =>This Inner Loop Header: Depth=1
; GFX1264-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
-; GFX1264-TRUE16-NEXT: v_add_nc_u32_e32 v0, s14, v1
+; GFX1264-TRUE16-NEXT: v_mov_b32_e32 v4, v0
+; GFX1264-TRUE16-NEXT: v_add_nc_u32_e32 v0, s14, v4
+; GFX1264-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
; GFX1264-TRUE16-NEXT: v_and_b32_e32 v0, s12, v0
-; GFX1264-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(SKIP_1) | instid1(VALU_DEP_2)
-; GFX1264-TRUE16-NEXT: v_and_or_b32 v0, v1, s13, v0
-; GFX1264-TRUE16-NEXT: v_mov_b32_e32 v3, v1
-; GFX1264-TRUE16-NEXT: v_mov_b32_e32 v2, v0
-; GFX1264-TRUE16-NEXT: buffer_atomic_cmpswap_b32 v[2:3], off, s[4:7], null th:TH_ATOMIC_RETURN scope:SCOPE_SYS
+; GFX1264-TRUE16-NEXT: v_and_or_b32 v3, v4, s13, v0
+; GFX1264-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1)
+; GFX1264-TRUE16-NEXT: v_mov_b32_e32 v0, v3
+; GFX1264-TRUE16-NEXT: v_mov_b32_e32 v1, v4
+; GFX1264-TRUE16-NEXT: buffer_atomic_cmpswap_b32 v[0:1], off, s[4:7], null th:TH_ATOMIC_RETURN scope:SCOPE_SYS
; GFX1264-TRUE16-NEXT: s_wait_loadcnt 0x0
-; GFX1264-TRUE16-NEXT: v_cmp_eq_u32_e32 vcc, v2, v1
-; GFX1264-TRUE16-NEXT: v_mov_b32_e32 v1, v2
+; GFX1264-TRUE16-NEXT: v_cmp_eq_u32_e32 vcc, v0, v4
; GFX1264-TRUE16-NEXT: s_or_b64 s[2:3], vcc, s[2:3]
; GFX1264-TRUE16-NEXT: s_wait_alu 0xfffe
; GFX1264-TRUE16-NEXT: s_and_not1_b64 exec, exec, s[2:3]
; GFX1264-TRUE16-NEXT: s_cbranch_execnz .LBB13_2
; GFX1264-TRUE16-NEXT: ; %bb.3: ; %atomicrmw.end
; GFX1264-TRUE16-NEXT: s_or_b64 exec, exec, s[2:3]
-; GFX1264-TRUE16-NEXT: v_lshrrev_b32_e32 v0, s11, v2
+; GFX1264-TRUE16-NEXT: v_lshrrev_b32_e32 v0, s11, v0
; GFX1264-TRUE16-NEXT: .LBB13_4: ; %Flow
; GFX1264-TRUE16-NEXT: s_or_b64 exec, exec, s[8:9]
; GFX1264-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(SKIP_2) | instid1(VALU_DEP_1)
@@ -8233,7 +8235,7 @@ define amdgpu_kernel void @uniform_add_i8(ptr addrspace(1) %result, ptr addrspac
; GFX1264-TRUE16-NEXT: s_mov_b32 s3, 0x31016000
; GFX1264-TRUE16-NEXT: v_readfirstlane_b32 s2, v0
; GFX1264-TRUE16-NEXT: s_wait_alu 0xf1ff
-; GFX1264-TRUE16-NEXT: v_mad_u16 v0.l, s10, v4.l, s2
+; GFX1264-TRUE16-NEXT: v_mad_u16 v0.l, s10, v2.l, s2
; GFX1264-TRUE16-NEXT: s_mov_b32 s2, -1
; GFX1264-TRUE16-NEXT: buffer_store_b8 v0, off, s[0:3], null
; GFX1264-TRUE16-NEXT: s_endpgm
@@ -8247,9 +8249,9 @@ define amdgpu_kernel void @uniform_add_i8(ptr addrspace(1) %result, ptr addrspac
; GFX1264-FAKE16-NEXT: s_mov_b64 s[8:9], exec
; GFX1264-FAKE16-NEXT: v_mbcnt_lo_u32_b32 v0, s6, 0
; GFX1264-FAKE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
-; GFX1264-FAKE16-NEXT: v_mbcnt_hi_u32_b32 v4, s7, v0
+; GFX1264-FAKE16-NEXT: v_mbcnt_hi_u32_b32 v2, s7, v0
; GFX1264-FAKE16-NEXT: ; implicit-def: $vgpr0
-; GFX1264-FAKE16-NEXT: v_cmpx_eq_u32_e32 0, v4
+; GFX1264-FAKE16-NEXT: v_cmpx_eq_u32_e32 0, v2
; GFX1264-FAKE16-NEXT: s_cbranch_execz .LBB13_4
; GFX1264-FAKE16-NEXT: ; %bb.1:
; GFX1264-FAKE16-NEXT: s_wait_kmcnt 0x0
@@ -8268,28 +8270,29 @@ define amdgpu_kernel void @uniform_add_i8(ptr addrspace(1) %result, ptr addrspac
; GFX1264-FAKE16-NEXT: s_mov_b32 s7, 0x31016000
; GFX1264-FAKE16-NEXT: s_mov_b32 s6, -1
; GFX1264-FAKE16-NEXT: s_wait_kmcnt 0x0
-; GFX1264-FAKE16-NEXT: v_mov_b32_e32 v1, s3
+; GFX1264-FAKE16-NEXT: v_mov_b32_e32 v0, s3
; GFX1264-FAKE16-NEXT: s_mov_b64 s[2:3], 0
; GFX1264-FAKE16-NEXT: .LBB13_2: ; %atomicrmw.start
; GFX1264-FAKE16-NEXT: ; =>This Inner Loop Header: Depth=1
; GFX1264-FAKE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
-; GFX1264-FAKE16-NEXT: v_add_nc_u32_e32 v0, s14, v1
+; GFX1264-FAKE16-NEXT: v_mov_b32_e32 v4, v0
+; GFX1264-FAKE16-NEXT: v_add_nc_u32_e32 v0, s14, v4
+; GFX1264-FAKE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
; GFX1264-FAKE16-NEXT: v_and_b32_e32 v0, s12, v0
-; GFX1264-FAKE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(SKIP_1) | instid1(VALU_DEP_2)
-; GFX1264-FAKE16-NEXT: v_and_or_b32 v0, v1, s13, v0
-; GFX1264-FAKE16-NEXT: v_mov_b32_e32 v3, v1
-; GFX1264-FAKE16-NEXT: v_mov_b32_e32 v2, v0
-; GFX1264-FAKE16-NEXT: buffer_atomic_cmpswap_b32 v[2:3], off, s[4:7], null th:TH_ATOMIC_RETURN scope:SCOPE_SYS
+; GFX1264-FAKE16-NEXT: v_and_or_b32 v3, v4, s13, v0
+; GFX1264-FAKE16-NEXT: s_delay_alu instid0(VALU_DEP_1)
+; GFX1264-FAKE16-NEXT: v_mov_b32_e32 v0, v3
+; GFX1264-FAKE16-NEXT: v_mov_b32_e32 v1, v4
+; GFX1264-FAKE16-NEXT: buffer_atomic_cmpswap_b32 v[0:1], off, s[4:7], null th:TH_ATOMIC_RETURN scope:SCOPE_SYS
; GFX1264-FAKE16-NEXT: s_wait_loadcnt 0x0
-; GFX1264-FAKE16-NEXT: v_cmp_eq_u32_e32 vcc, v2, v1
-; GFX1264-FAKE16-NEXT: v_mov_b32_e32 v1, v2
+; GFX1264-FAKE16-NEXT: v_cmp_eq_u32_e32 vcc, v0, v4
; GFX1264-FAKE16-NEXT: s_or_b64 s[2:3], vcc, s[2:3]
; GFX1264-FAKE16-NEXT: s_wait_alu 0xfffe
; GFX1264-FAKE16-NEXT: s_and_not1_b64 exec, exec, s[2:3]
; GFX1264-FAKE16-NEXT: s_cbranch_execnz .LBB13_2
; GFX1264-FAKE16-NEXT: ; %bb.3: ; %atomicrmw.end
; GFX1264-FAKE16-NEXT: s_or_b64 exec, exec, s[2:3]
-; GFX1264-FAKE16-NEXT: v_lshrrev_b32_e32 v0, s11, v2
+; GFX1264-FAKE16-NEXT: v_lshrrev_b32_e32 v0, s11, v0
; GFX1264-FAKE16-NEXT: .LBB13_4: ; %Flow
; GFX1264-FAKE16-NEXT: s_or_b64 exec, exec, s[8:9]
; GFX1264-FAKE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(SKIP_2) | instid1(VALU_DEP_1)
@@ -8298,7 +8301,7 @@ define amdgpu_kernel void @uniform_add_i8(ptr addrspace(1) %result, ptr addrspac
; GFX1264-FAKE16-NEXT: s_mov_b32 s3, 0x31016000
; GFX1264-FAKE16-NEXT: v_readfirstlane_b32 s2, v0
; GFX1264-FAKE16-NEXT: s_wait_alu 0xf1ff
-; GFX1264-FAKE16-NEXT: v_mad_u16 v0, s10, v4, s2
+; GFX1264-FAKE16-NEXT: v_mad_u16 v0, s10, v2, s2
; GFX1264-FAKE16-NEXT: s_mov_b32 s2, -1
; GFX1264-FAKE16-NEXT: buffer_store_b8 v0, off, s[0:3], null
; GFX1264-FAKE16-NEXT: s_endpgm
@@ -8310,11 +8313,11 @@ define amdgpu_kernel void @uniform_add_i8(ptr addrspace(1) %result, ptr addrspac
; GFX1232-TRUE16-NEXT: s_load_b32 s8, s[4:5], 0x34
; GFX1232-TRUE16-NEXT: s_mov_b32 s6, exec_lo
; GFX1232-TRUE16-NEXT: s_mov_b32 s10, 0
-; GFX1232-TRUE16-NEXT: v_mbcnt_lo_u32_b32 v4, s6, 0
+; GFX1232-TRUE16-NEXT: v_mbcnt_lo_u32_b32 v2, s6, 0
; GFX1232-TRUE16-NEXT: s_mov_b32 s9, exec_lo
; GFX1232-TRUE16-NEXT: ; implicit-def: $vgpr0_lo16
; GFX1232-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1)
-; GFX1232-TRUE16-NEXT: v_cmpx_eq_u32_e32 0, v4
+; GFX1232-TRUE16-NEXT: v_cmpx_eq_u32_e32 0, v2
; GFX1232-TRUE16-NEXT: s_cbranch_execz .LBB13_4
; GFX1232-TRUE16-NEXT: ; %bb.1:
; GFX1232-TRUE16-NEXT: s_wait_kmcnt 0x0
@@ -8335,27 +8338,28 @@ define amdgpu_kernel void @uniform_add_i8(ptr addrspace(1) %result, ptr addrspac
; GFX1232-TRUE16-NEXT: s_lshl_b32 s12, s6, s2
; GFX1232-TRUE16-NEXT: s_mov_b32 s6, -1
; GFX1232-TRUE16-NEXT: s_wait_kmcnt 0x0
-; GFX1232-TRUE16-NEXT: v_mov_b32_e32 v1, s7
+; GFX1232-TRUE16-NEXT: v_mov_b32_e32 v0, s7
; GFX1232-TRUE16-NEXT: s_mov_b32 s7, 0x31016000
; GFX1232-TRUE16-NEXT: .LBB13_2: ; %atomicrmw.start
; GFX1232-TRUE16-NEXT: ; =>This Inner Loop Header: Depth=1
; GFX1232-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
-; GFX1232-TRUE16-NEXT: v_add_nc_u32_e32 v0, s12, v1
-; GFX1232-TRUE16-NEXT: v_and_b32_e32 v0, s3, v0
+; GFX1232-TRUE16-NEXT: v_mov_b32_e32 v4, v0
+; GFX1232-TRUE16-NEXT: v_add_nc_u32_e32 v0, s12, v4
; GFX1232-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
-; GFX1232-TRUE16-NEXT: v_and_or_b32 v0, v1, s11, v0
-; GFX1232-TRUE16-NEXT: v_dual_mov_b32 v3, v1 :: v_dual_mov_b32 v2, v0
-; GFX1232-TRUE16-NEXT: buffer_atomic_cmpswap_b32 v[2:3], off, s[4:7], null th:TH_ATOMIC_RETURN scope:SCOPE_SYS
+; GFX1232-TRUE16-NEXT: v_and_b32_e32 v0, s3, v0
+; GFX1232-TRUE16-NEXT: v_and_or_b32 v3, v4, s11, v0
+; GFX1232-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1)
+; GFX1232-TRUE16-NEXT: v_dual_mov_b32 v0, v3 :: v_dual_mov_b32 v1, v4
+; GFX1232-TRUE16-NEXT: buffer_atomic_cmpswap_b32 v[0:1], off, s[4:7], null th:TH_ATOMIC_RETURN scope:SCOPE_SYS
; GFX1232-TRUE16-NEXT: s_wait_loadcnt 0x0
-; GFX1232-TRUE16-NEXT: v_cmp_eq_u32_e32 vcc_lo, v2, v1
-; GFX1232-TRUE16-NEXT: v_mov_b32_e32 v1, v2
+; GFX1232-TRUE16-NEXT: v_cmp_eq_u32_e32 vcc_lo, v0, v4
; GFX1232-TRUE16-NEXT: s_or_b32 s10, vcc_lo, s10
; GFX1232-TRUE16-NEXT: s_wait_alu 0xfffe
; GFX1232-TRUE16-NEXT: s_and_not1_b32 exec_lo, exec_lo, s10
; GFX1232-TRUE16-NEXT: s_cbranch_execnz .LBB13_2
; GFX1232-TRUE16-NEXT: ; %bb.3: ; %atomicrmw.end
; GFX1232-TRUE16-NEXT: s_or_b32 exec_lo, exec_lo, s10
-; GFX1232-TRUE16-NEXT: v_lshrrev_b32_e32 v0, s2, v2
+; GFX1232-TRUE16-NEXT: v_lshrrev_b32_e32 v0, s2, v0
; GFX1232-TRUE16-NEXT: .LBB13_4: ; %Flow
; GFX1232-TRUE16-NEXT: s_or_b32 exec_lo, exec_lo, s9
; GFX1232-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(SKIP_2) | instid1(VALU_DEP_1)
@@ -8364,7 +8368,7 @@ define amdgpu_kernel void @uniform_add_i8(ptr addrspace(1) %result, ptr addrspac
; GFX1232-TRUE16-NEXT: s_mov_b32 s3, 0x31016000
; GFX1232-TRUE16-NEXT: v_readfirstlane_b32 s2, v0
; GFX1232-TRUE16-NEXT: s_wait_alu 0xf1ff
-; GFX1232-TRUE16-NEXT: v_mad_u16 v0.l, s8, v4.l, s2
+; GFX1232-TRUE16-NEXT: v_mad_u16 v0.l, s8, v2.l, s2
; GFX1232-TRUE16-NEXT: s_mov_b32 s2, -1
; GFX1232-TRUE16-NEXT: buffer_store_b8 v0, off, s[0:3], null
; GFX1232-TRUE16-NEXT: s_endpgm
@@ -8376,11 +8380,11 @@ define amdgpu_kernel void @uniform_add_i8(ptr addrspace(1) %result, ptr addrspac
; GFX1232-FAKE16-NEXT: s_load_b32 s8, s[4:5], 0x34
; GFX1232-FAKE16-NEXT: s_mov_b32 s6, exec_lo
; GFX1232-FAKE16-NEXT: s_mov_b32 s10, 0
-; GFX1232-FAKE16-NEXT: v_mbcnt_lo_u32_b32 v4, s6, 0
+; GFX1232-FAKE16-NEXT: v_mbcnt_lo_u32_b32 v2, s6, 0
; GFX1232-FAKE16-NEXT: s_mov_b32 s9, exec_lo
; GFX1232-FAKE16-NEXT: ; implicit-def: $vgpr0
; GFX1232-FAKE16-NEXT: s_delay_alu instid0(VALU_DEP_1)
-; GFX1232-FAKE16-NEXT: v_cmpx_eq_u32_e32 0, v4
+; GFX1232-FAKE16-NEXT: v_cmpx_eq_u32_e32 0, v2
; GFX1232-FAKE16-NEXT: s_cbranch_execz .LBB13_4
; GFX1232-FAKE16-NEXT: ; %bb.1:
; GFX1232-FAKE16-NEXT: s_wait_kmcnt 0x0
@@ -8400,27 +8404,28 @@ define amdgpu_kernel void @uniform_add_i8(ptr addrspace(1) %result, ptr addrspac
; GFX1232-FAKE16-NEXT: s_lshl_b32 s12, s6, s2
; GFX1232-FAKE16-NEXT: s_mov_b32 s6, -1
; GFX1232-FAKE16-NEXT: s_wait_kmcnt 0x0
-; GFX1232-FAKE16-NEXT: v_mov_b32_e32 v1, s7
+; GFX1232-FAKE16-NEXT: v_mov_b32_e32 v0, s7
; GFX1232-FAKE16-NEXT: s_mov_b32 s7, 0x31016000
; GFX1232-FAKE16-NEXT: .LBB13_2: ; %atomicrmw.start
; GFX1232-FAKE16-NEXT: ; =>This Inner Loop Header: Depth=1
; GFX1232-FAKE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
-; GFX1232-FAKE16-NEXT: v_add_nc_u32_e32 v0, s12, v1
-; GFX1232-FAKE16-NEXT: v_and_b32_e32 v0, s3, v0
+; GFX1232-FAKE16-NEXT: v_mov_b32_e32 v4, v0
+; GFX1232-FAKE16-NEXT: v_add_nc_u32_e32 v0, s12, v4
; GFX1232-FAKE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
-; GFX1232-FAKE16-NEXT: v_and_or_b32 v0, v1, s11, v0
-; GFX1232-FAKE16-NEXT: v_dual_mov_b32 v3, v1 :: v_dual_mov_b32 v2, v0
-; GFX1232-FAKE16-NEXT: buffer_atomic_cmpswap_b32 v[2:3], off, s[4:7], null th:TH_ATOMIC_RETURN scope:SCOPE_SYS
+; GFX1232-FAKE16-NEXT: v_and_b32_e32 v0, s3, v0
+; GFX1232-FAKE16-NEXT: v_and_or_b32 v3, v4, s11, v0
+; GFX1232-FAKE16-NEXT: s_delay_alu instid0(VALU_DEP_1)
+; GFX1232-FAKE16-NEXT: v_dual_mov_b32 v0, v3 :: v_dual_mov_b32 v1, v4
+; GFX1232-FAKE16-NEXT: buffer_atomic_cmpswap_b32 v[0:1], off, s[4:7], null th:TH_ATOMIC_RETURN scope:SCOPE_SYS
; GFX1232-FAKE16-NEXT: s_wait_loadcnt 0x0
-; GFX1232-FAKE16-NEXT: v_cmp_eq_u32_e32 vcc_lo, v2, v1
-; GFX1232-FAKE16-NEXT: v_mov_b32_e32 v1, v2
+; GFX1232-FAKE16-NEXT: v_cmp_eq_u32_e32 vcc_lo, v0, v4
; GFX1232-FAKE16-NEXT: s_or_b32 s10, vcc_lo, s10
; GFX1232-FAKE16-NEXT: s_wait_alu 0xfffe
; GFX1232-FAKE16-NEXT: s_and_not1_b32 exec_lo, exec_lo, s10
; GFX1232-FAKE16-NEXT: s_cbranch_execnz .LBB13_2
; GFX1232-FAKE16-NEXT: ; %bb.3: ; %atomicrmw.end
; GFX1232-FAKE16-NEXT: s_or_b32 exec_lo, exec_lo, s10
-; GFX1232-FAKE16-NEXT: v_lshrrev_b32_e32 v0, s2, v2
+; GFX1232-FAKE16-NEXT: v_lshrrev_b32_e32 v0, s2, v0
; GFX1232-FAKE16-NEXT: .LBB13_4: ; %Flow
; GFX1232-FAKE16-NEXT: s_or_b32 exec_lo, exec_lo, s9
; GFX1232-FAKE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(SKIP_2) | instid1(VALU_DEP_1)
@@ -8429,7 +8434,7 @@ define amdgpu_kernel void @uniform_add_i8(ptr addrspace(1) %result, ptr addrspac
; GFX1232-FAKE16-NEXT: s_mov_b32 s3, 0x31016000
; GFX1232-FAKE16-NEXT: v_readfirstlane_b32 s2, v0
; GFX1232-FAKE16-NEXT: s_wait_alu 0xf1ff
-; GFX1232-FAKE16-NEXT: v_mad_u16 v0, s8, v4, s2
+; GFX1232-FAKE16-NEXT: v_mad_u16 v0, s8, v2, s2
; GFX1232-FAKE16-NEXT: s_mov_b32 s2, -1
; GFX1232-FAKE16-NEXT: buffer_store_b8 v0, off, s[0:3], null
; GFX1232-FAKE16-NEXT: s_endpgm
@@ -8456,27 +8461,28 @@ define amdgpu_kernel void @uniform_xchg_i8(ptr addrspace(1) %result, ptr addrspa
; GFX7LESS-NEXT: s_lshl_b32 s2, s3, s10
; GFX7LESS-NEXT: s_not_b32 s3, s11
; GFX7LESS-NEXT: s_waitcnt lgkmcnt(0)
-; GFX7LESS-NEXT: v_mov_b32_e32 v1, s6
+; GFX7LESS-NEXT: v_mov_b32_e32 v0, s6
; GFX7LESS-NEXT: s_mov_b32 s6, -1
; GFX7LESS-NEXT: .LBB14_1: ; %atomicrmw.start
; GFX7LESS-NEXT: ; =>This Inner Loop Header: Depth=1
-; GFX7LESS-NEXT: v_and_b32_e32 v0, s3, v1
-; GFX7LESS-NEXT: v_or_b32_e32 v0, s2, v0
-; GFX7LESS-NEXT: s_waitcnt expcnt(0)
-; GFX7LESS-NEXT: v_mov_b32_e32 v3, v1
; GFX7LESS-NEXT: v_mov_b32_e32 v2, v0
-; GFX7LESS-NEXT: buffer_atomic_cmpswap v[2:3], off, s[4:7], 0 glc
+; GFX7LESS-NEXT: s_waitcnt expcnt(0)
+; GFX7LESS-NEXT: v_and_b32_e32 v0, s3, v2
+; GFX7LESS-NEXT: v_or_b32_e32 v1, s2, v0
+; GFX7LESS-NEXT: v_mov_b32_e32 v0, v1
+; GFX7LESS-NEXT: v_mov_b32_e32 v1, v2
+; GFX7LESS-NEXT: buffer_atomic_cmpswap v[0:1], off, s[4:7], 0 glc
; GFX7LESS-NEXT: s_waitcnt vmcnt(0)
-; GFX7LESS-NEXT: v_cmp_eq_u32_e32 vcc, v2, v1
+; GFX7LESS-NEXT: v_cmp_eq_u32_e32 vcc, v0, v2
; GFX7LESS-NEXT: s_or_b64 s[8:9], vcc, s[8:9]
-; GFX7LESS-NEXT: v_mov_b32_e32 v1, v2
; GFX7LESS-NEXT: s_andn2_b64 exec, exec, s[8:9]
; GFX7LESS-NEXT: s_cbranch_execnz .LBB14_1
; GFX7LESS-NEXT: ; %bb.2: ; %atomicrmw.end
; GFX7LESS-NEXT: s_or_b64 exec, exec, s[8:9]
; GFX7LESS-NEXT: s_mov_b32 s3, 0xf000
; GFX7LESS-NEXT: s_mov_b32 s2, -1
-; GFX7LESS-NEXT: v_lshrrev_b32_e32 v0, s10, v2
+; GFX7LESS-NEXT: s_waitcnt expcnt(0)
+; GFX7LESS-NEXT: v_lshrrev_b32_e32 v0, s10, v0
; GFX7LESS-NEXT: buffer_store_byte v0, off, s[0:3], 0
; GFX7LESS-NEXT: s_endpgm
;
@@ -8496,27 +8502,27 @@ define amdgpu_kernel void @uniform_xchg_i8(ptr addrspace(1) %result, ptr addrspa
; GFX8-NEXT: s_lshl_b32 s10, s2, s8
; GFX8-NEXT: s_mov_b64 s[2:3], 0
; GFX8-NEXT: s_waitcnt lgkmcnt(0)
-; GFX8-NEXT: v_mov_b32_e32 v1, s7
+; GFX8-NEXT: v_mov_b32_e32 v0, s7
; GFX8-NEXT: s_mov_b32 s7, 0xf000
; GFX8-NEXT: s_mov_b32 s6, -1
; GFX8-NEXT: .LBB14_1: ; %atomicrmw.start
; GFX8-NEXT: ; =>This Inner Loop Header: Depth=1
-; GFX8-NEXT: v_and_b32_e32 v0, s9, v1
-; GFX8-NEXT: v_or_b32_e32 v0, s10, v0
-; GFX8-NEXT: v_mov_b32_e32 v3, v1
; GFX8-NEXT: v_mov_b32_e32 v2, v0
-; GFX8-NEXT: buffer_atomic_cmpswap v[2:3], off, s[4:7], 0 glc
+; GFX8-NEXT: v_and_b32_e32 v0, s9, v2
+; GFX8-NEXT: v_or_b32_e32 v1, s10, v0
+; GFX8-NEXT: v_mov_b32_e32 v0, v1
+; GFX8-NEXT: v_mov_b32_e32 v1, v2
+; GFX8-NEXT: buffer_atomic_cmpswap v[0:1], off, s[4:7], 0 glc
; GFX8-NEXT: s_waitcnt vmcnt(0)
-; GFX8-NEXT: v_cmp_eq_u32_e32 vcc, v2, v1
+; GFX8-NEXT: v_cmp_eq_u32_e32 vcc, v0, v2
; GFX8-NEXT: s_or_b64 s[2:3], vcc, s[2:3]
-; GFX8-NEXT: v_mov_b32_e32 v1, v2
; GFX8-NEXT: s_andn2_b64 exec, exec, s[2:3]
; GFX8-NEXT: s_cbranch_execnz .LBB14_1
; GFX8-NEXT: ; %bb.2: ; %atomicrmw.end
; GFX8-NEXT: s_or_b64 exec, exec, s[2:3]
; GFX8-NEXT: s_mov_b32 s3, 0xf000
; GFX8-NEXT: s_mov_b32 s2, -1
-; GFX8-NEXT: v_lshrrev_b32_e32 v0, s8, v2
+; GFX8-NEXT: v_lshrrev_b32_e32 v0, s8, v0
; GFX8-NEXT: buffer_store_byte v0, off, s[0:3], 0
; GFX8-NEXT: s_endpgm
;
@@ -8536,27 +8542,27 @@ define amdgpu_kernel void @uniform_xchg_i8(ptr addrspace(1) %result, ptr addrspa
; GFX9-NEXT: s_lshl_b32 s10, s2, s8
; GFX9-NEXT: s_mov_b64 s[2:3], 0
; GFX9-NEXT: s_waitcnt lgkmcnt(0)
-; GFX9-NEXT: v_mov_b32_e32 v1, s7
+; GFX9-NEXT: v_mov_b32_e32 v0, s7
; GFX9-NEXT: s_mov_b32 s7, 0xf000
; GFX9-NEXT: s_mov_b32 s6, -1
; GFX9-NEXT: .LBB14_1: ; %atomicrmw.start
; GFX9-NEXT: ; =>This Inner Loop Header: Depth=1
-; GFX9-NEXT: v_and_b32_e32 v0, s9, v1
-; GFX9-NEXT: v_or_b32_e32 v0, s10, v0
-; GFX9-NEXT: v_mov_b32_e32 v3, v1
; GFX9-NEXT: v_mov_b32_e32 v2, v0
-; GFX9-NEXT: buffer_atomic_cmpswap v[2:3], off, s[4:7], 0 glc
+; GFX9-NEXT: v_and_b32_e32 v0, s9, v2
+; GFX9-NEXT: v_or_b32_e32 v1, s10, v0
+; GFX9-NEXT: v_mov_b32_e32 v0, v1
+; GFX9-NEXT: v_mov_b32_e32 v1, v2
+; GFX9-NEXT: buffer_atomic_cmpswap v[0:1], off, s[4:7], 0 glc
; GFX9-NEXT: s_waitcnt vmcnt(0)
-; GFX9-NEXT: v_cmp_eq_u32_e32 vcc, v2, v1
+; GFX9-NEXT: v_cmp_eq_u32_e32 vcc, v0, v2
; GFX9-NEXT: s_or_b64 s[2:3], vcc, s[2:3]
-; GFX9-NEXT: v_mov_b32_e32 v1, v2
; GFX9-NEXT: s_andn2_b64 exec, exec, s[2:3]
; GFX9-NEXT: s_cbranch_execnz .LBB14_1
; GFX9-NEXT: ; %bb.2: ; %atomicrmw.end
; GFX9-NEXT: s_or_b64 exec, exec, s[2:3]
; GFX9-NEXT: s_mov_b32 s3, 0xf000
; GFX9-NEXT: s_mov_b32 s2, -1
-; GFX9-NEXT: v_lshrrev_b32_e32 v0, s8, v2
+; GFX9-NEXT: v_lshrrev_b32_e32 v0, s8, v0
; GFX9-NEXT: buffer_store_byte v0, off, s[0:3], 0
; GFX9-NEXT: s_endpgm
;
@@ -8578,23 +8584,23 @@ define amdgpu_kernel void @uniform_xchg_i8(ptr addrspace(1) %result, ptr addrspa
; GFX1064-NEXT: s_not_b32 s9, s2
; GFX1064-NEXT: s_mov_b32 s6, -1
; GFX1064-NEXT: s_waitcnt lgkmcnt(0)
-; GFX1064-NEXT: v_mov_b32_e32 v1, s3
+; GFX1064-NEXT: v_mov_b32_e32 v0, s3
; GFX1064-NEXT: s_mov_b64 s[2:3], 0
; GFX1064-NEXT: .LBB14_1: ; %atomicrmw.start
; GFX1064-NEXT: ; =>This Inner Loop Header: Depth=1
-; GFX1064-NEXT: v_and_or_b32 v0, v1, s9, s10
-; GFX1064-NEXT: v_mov_b32_e32 v3, v1
; GFX1064-NEXT: v_mov_b32_e32 v2, v0
-; GFX1064-NEXT: buffer_atomic_cmpswap v[2:3], off, s[4:7], 0 glc
-; GFX1064-NEXT: s_waitcnt vmcnt(0)
-; GFX1064-NEXT: v_cmp_eq_u32_e32 vcc, v2, v1
+; GFX1064-NEXT: v_and_or_b32 v1, v2, s9, s10
+; GFX1064-NEXT: v_mov_b32_e32 v0, v1
; GFX1064-NEXT: v_mov_b32_e32 v1, v2
+; GFX1064-NEXT: buffer_atomic_cmpswap v[0:1], off, s[4:7], 0 glc
+; GFX1064-NEXT: s_waitcnt vmcnt(0)
+; GFX1064-NEXT: v_cmp_eq_u32_e32 vcc, v0, v2
; GFX1064-NEXT: s_or_b64 s[2:3], vcc, s[2:3]
; GFX1064-NEXT: s_andn2_b64 exec, exec, s[2:3]
; GFX1064-NEXT: s_cbranch_execnz .LBB14_1
; GFX1064-NEXT: ; %bb.2: ; %atomicrmw.end
; GFX1064-NEXT: s_or_b64 exec, exec, s[2:3]
-; GFX1064-NEXT: v_lshrrev_b32_e32 v0, s8, v2
+; GFX1064-NEXT: v_lshrrev_b32_e32 v0, s8, v0
; GFX1064-NEXT: s_mov_b32 s3, 0x31016000
; GFX1064-NEXT: s_mov_b32 s2, -1
; GFX1064-NEXT: buffer_store_byte v0, off, s[0:3], 0
@@ -8618,23 +8624,23 @@ define amdgpu_kernel void @uniform_xchg_i8(ptr addrspace(1) %result, ptr addrspa
; GFX1032-NEXT: s_not_b32 s3, s3
; GFX1032-NEXT: s_mov_b32 s6, -1
; GFX1032-NEXT: s_waitcnt lgkmcnt(0)
-; GFX1032-NEXT: v_mov_b32_e32 v1, s7
+; GFX1032-NEXT: v_mov_b32_e32 v0, s7
; GFX1032-NEXT: s_mov_b32 s7, 0x31016000
; GFX1032-NEXT: .LBB14_1: ; %atomicrmw.start
; GFX1032-NEXT: ; =>This Inner Loop Header: Depth=1
-; GFX1032-NEXT: v_and_or_b32 v0, v1, s3, s8
-; GFX1032-NEXT: v_mov_b32_e32 v3, v1
; GFX1032-NEXT: v_mov_b32_e32 v2, v0
-; GFX1032-NEXT: buffer_atomic_cmpswap v[2:3], off, s[4:7], 0 glc
-; GFX1032-NEXT: s_waitcnt vmcnt(0)
-; GFX1032-NEXT: v_cmp_eq_u32_e32 vcc_lo, v2, v1
+; GFX1032-NEXT: v_and_or_b32 v1, v2, s3, s8
+; GFX1032-NEXT: v_mov_b32_e32 v0, v1
; GFX1032-NEXT: v_mov_b32_e32 v1, v2
+; GFX1032-NEXT: buffer_atomic_cmpswap v[0:1], off, s[4:7], 0 glc
+; GFX1032-NEXT: s_waitcnt vmcnt(0)
+; GFX1032-NEXT: v_cmp_eq_u32_e32 vcc_lo, v0, v2
; GFX1032-NEXT: s_or_b32 s9, vcc_lo, s9
; GFX1032-NEXT: s_andn2_b32 exec_lo, exec_lo, s9
; GFX1032-NEXT: s_cbranch_execnz .LBB14_1
; GFX1032-NEXT: ; %bb.2: ; %atomicrmw.end
; GFX1032-NEXT: s_or_b32 exec_lo, exec_lo, s9
-; GFX1032-NEXT: v_lshrrev_b32_e32 v0, s2, v2
+; GFX1032-NEXT: v_lshrrev_b32_e32 v0, s2, v0
; GFX1032-NEXT: s_mov_b32 s3, 0x31016000
; GFX1032-NEXT: s_mov_b32 s2, -1
; GFX1032-NEXT: buffer_store_byte v0, off, s[0:3], 0
@@ -8658,25 +8664,26 @@ define amdgpu_kernel void @uniform_xchg_i8(ptr addrspace(1) %result, ptr addrspa
; GFX1164-NEXT: s_not_b32 s9, s2
; GFX1164-NEXT: s_mov_b32 s6, -1
; GFX1164-NEXT: s_waitcnt lgkmcnt(0)
-; GFX1164-NEXT: v_mov_b32_e32 v1, s3
+; GFX1164-NEXT: v_mov_b32_e32 v0, s3
; GFX1164-NEXT: s_mov_b64 s[2:3], 0
; GFX1164-NEXT: .LBB14_1: ; %atomicrmw.start
; GFX1164-NEXT: ; =>This Inner Loop Header: Depth=1
-; GFX1164-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(SKIP_1) | instid1(VALU_DEP_2)
-; GFX1164-NEXT: v_and_or_b32 v0, v1, s9, s10
-; GFX1164-NEXT: v_mov_b32_e32 v3, v1
+; GFX1164-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
; GFX1164-NEXT: v_mov_b32_e32 v2, v0
-; GFX1164-NEXT: buffer_atomic_cmpswap_b32 v[2:3], off, s[4:7], 0 glc
-; GFX1164-NEXT: s_waitcnt vmcnt(0)
-; GFX1164-NEXT: v_cmp_eq_u32_e32 vcc, v2, v1
+; GFX1164-NEXT: v_and_or_b32 v1, v2, s9, s10
+; GFX1164-NEXT: s_delay_alu instid0(VALU_DEP_1)
+; GFX1164-NEXT: v_mov_b32_e32 v0, v1
; GFX1164-NEXT: v_mov_b32_e32 v1, v2
+; GFX1164-NEXT: buffer_atomic_cmpswap_b32 v[0:1], off, s[4:7], 0 glc
+; GFX1164-NEXT: s_waitcnt vmcnt(0)
+; GFX1164-NEXT: v_cmp_eq_u32_e32 vcc, v0, v2
; GFX1164-NEXT: s_or_b64 s[2:3], vcc, s[2:3]
; GFX1164-NEXT: s_delay_alu instid0(SALU_CYCLE_1)
; GFX1164-NEXT: s_and_not1_b64 exec, exec, s[2:3]
; GFX1164-NEXT: s_cbranch_execnz .LBB14_1
; GFX1164-NEXT: ; %bb.2: ; %atomicrmw.end
; GFX1164-NEXT: s_or_b64 exec, exec, s[2:3]
-; GFX1164-NEXT: v_lshrrev_b32_e32 v0, s8, v2
+; GFX1164-NEXT: v_lshrrev_b32_e32 v0, s8, v0
; GFX1164-NEXT: s_mov_b32 s3, 0x31016000
; GFX1164-NEXT: s_mov_b32 s2, -1
; GFX1164-NEXT: buffer_store_b8 v0, off, s[0:3], 0
@@ -8700,24 +8707,26 @@ define amdgpu_kernel void @uniform_xchg_i8(ptr addrspace(1) %result, ptr addrspa
; GFX1132-NEXT: s_not_b32 s3, s3
; GFX1132-NEXT: s_mov_b32 s6, -1
; GFX1132-NEXT: s_waitcnt lgkmcnt(0)
-; GFX1132-NEXT: v_mov_b32_e32 v1, s7
+; GFX1132-NEXT: v_mov_b32_e32 v0, s7
; GFX1132-NEXT: s_mov_b32 s7, 0x31016000
; GFX1132-NEXT: .LBB14_1: ; %atomicrmw.start
; GFX1132-NEXT: ; =>This Inner Loop Header: Depth=1
; GFX1132-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
-; GFX1132-NEXT: v_and_or_b32 v0, v1, s3, s8
-; GFX1132-NEXT: v_dual_mov_b32 v3, v1 :: v_dual_mov_b32 v2, v0
-; GFX1132-NEXT: buffer_atomic_cmpswap_b32 v[2:3], off, s[4:7], 0 glc
-; GFX1132-NEXT: s_waitcnt vmcnt(0)
-; GFX1132-NEXT: v_cmp_eq_u32_e32 vcc_lo, v2, v1
+; GFX1132-NEXT: v_mov_b32_e32 v2, v0
+; GFX1132-NEXT: v_and_or_b32 v1, v2, s3, s8
+; GFX1132-NEXT: s_delay_alu instid0(VALU_DEP_1)
+; GFX1132-NEXT: v_mov_b32_e32 v0, v1
; GFX1132-NEXT: v_mov_b32_e32 v1, v2
+; GFX1132-NEXT: buffer_atomic_cmpswap_b32 v[0:1], off, s[4:7], 0 glc
+; GFX1132-NEXT: s_waitcnt vmcnt(0)
+; GFX1132-NEXT: v_cmp_eq_u32_e32 vcc_lo, v0, v2
; GFX1132-NEXT: s_or_b32 s9, vcc_lo, s9
; GFX1132-NEXT: s_delay_alu instid0(SALU_CYCLE_1)
; GFX1132-NEXT: s_and_not1_b32 exec_lo, exec_lo, s9
; GFX1132-NEXT: s_cbranch_execnz .LBB14_1
; GFX1132-NEXT: ; %bb.2: ; %atomicrmw.end
; GFX1132-NEXT: s_or_b32 exec_lo, exec_lo, s9
-; GFX1132-NEXT: v_lshrrev_b32_e32 v0, s2, v2
+; GFX1132-NEXT: v_lshrrev_b32_e32 v0, s2, v0
; GFX1132-NEXT: s_mov_b32 s3, 0x31016000
; GFX1132-NEXT: s_mov_b32 s2, -1
; GFX1132-NEXT: buffer_store_b8 v0, off, s[0:3], 0
@@ -8741,25 +8750,26 @@ define amdgpu_kernel void @uniform_xchg_i8(ptr addrspace(1) %result, ptr addrspa
; GFX1264-NEXT: s_not_b32 s9, s2
; GFX1264-NEXT: s_mov_b32 s6, -1
; GFX1264-NEXT: s_wait_kmcnt 0x0
-; GFX1264-NEXT: v_mov_b32_e32 v1, s3
+; GFX1264-NEXT: v_mov_b32_e32 v0, s3
; GFX1264-NEXT: s_mov_b64 s[2:3], 0
; GFX1264-NEXT: .LBB14_1: ; %atomicrmw.start
; GFX1264-NEXT: ; =>This Inner Loop Header: Depth=1
-; GFX1264-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(SKIP_1) | instid1(VALU_DEP_2)
-; GFX1264-NEXT: v_and_or_b32 v0, v1, s9, s10
-; GFX1264-NEXT: v_mov_b32_e32 v3, v1
+; GFX1264-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
; GFX1264-NEXT: v_mov_b32_e32 v2, v0
-; GFX1264-NEXT: buffer_atomic_cmpswap_b32 v[2:3], off, s[4:7], null th:TH_ATOMIC_RETURN scope:SCOPE_SYS
-; GFX1264-NEXT: s_wait_loadcnt 0x0
-; GFX1264-NEXT: v_cmp_eq_u32_e32 vcc, v2, v1
+; GFX1264-NEXT: v_and_or_b32 v1, v2, s9, s10
+; GFX1264-NEXT: s_delay_alu instid0(VALU_DEP_1)
+; GFX1264-NEXT: v_mov_b32_e32 v0, v1
; GFX1264-NEXT: v_mov_b32_e32 v1, v2
+; GFX1264-NEXT: buffer_atomic_cmpswap_b32 v[0:1], off, s[4:7], null th:TH_ATOMIC_RETURN scope:SCOPE_SYS
+; GFX1264-NEXT: s_wait_loadcnt 0x0
+; GFX1264-NEXT: v_cmp_eq_u32_e32 vcc, v0, v2
; GFX1264-NEXT: s_or_b64 s[2:3], vcc, s[2:3]
; GFX1264-NEXT: s_wait_alu 0xfffe
; GFX1264-NEXT: s_and_not1_b64 exec, exec, s[2:3]
; GFX1264-NEXT: s_cbranch_execnz .LBB14_1
; GFX1264-NEXT: ; %bb.2: ; %atomicrmw.end
; GFX1264-NEXT: s_or_b64 exec, exec, s[2:3]
-; GFX1264-NEXT: v_lshrrev_b32_e32 v0, s8, v2
+; GFX1264-NEXT: v_lshrrev_b32_e32 v0, s8, v0
; GFX1264-NEXT: s_mov_b32 s3, 0x31016000
; GFX1264-NEXT: s_mov_b32 s2, -1
; GFX1264-NEXT: buffer_store_b8 v0, off, s[0:3], null
@@ -8783,24 +8793,26 @@ define amdgpu_kernel void @uniform_xchg_i8(ptr addrspace(1) %result, ptr addrspa
; GFX1232-NEXT: s_not_b32 s3, s3
; GFX1232-NEXT: s_mov_b32 s6, -1
; GFX1232-NEXT: s_wait_kmcnt 0x0
-; GFX1232-NEXT: v_mov_b32_e32 v1, s7
+; GFX1232-NEXT: v_mov_b32_e32 v0, s7
; GFX1232-NEXT: s_mov_b32 s7, 0x31016000
; GFX1232-NEXT: .LBB14_1: ; %atomicrmw.start
; GFX1232-NEXT: ; =>This Inner Loop Header: Depth=1
; GFX1232-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
-; GFX1232-NEXT: v_and_or_b32 v0, v1, s3, s8
-; GFX1232-NEXT: v_dual_mov_b32 v3, v1 :: v_dual_mov_b32 v2, v0
-; GFX1232-NEXT: buffer_atomic_cmpswap_b32 v[2:3], off, s[4:7], null th:TH_ATOMIC_RETURN scope:SCOPE_SYS
-; GFX1232-NEXT: s_wait_loadcnt 0x0
-; GFX1232-NEXT: v_cmp_eq_u32_e32 vcc_lo, v2, v1
+; GFX1232-NEXT: v_mov_b32_e32 v2, v0
+; GFX1232-NEXT: v_and_or_b32 v1, v2, s3, s8
+; GFX1232-NEXT: s_delay_alu instid0(VALU_DEP_1)
+; GFX1232-NEXT: v_mov_b32_e32 v0, v1
; GFX1232-NEXT: v_mov_b32_e32 v1, v2
+; GFX1232-NEXT: buffer_atomic_cmpswap_b32 v[0:1], off, s[4:7], null th:TH_ATOMIC_RETURN scope:SCOPE_SYS
+; GFX1232-NEXT: s_wait_loadcnt 0x0
+; GFX1232-NEXT: v_cmp_eq_u32_e32 vcc_lo, v0, v2
; GFX1232-NEXT: s_or_b32 s9, vcc_lo, s9
; GFX1232-NEXT: s_wait_alu 0xfffe
; GFX1232-NEXT: s_and_not1_b32 exec_lo, exec_lo, s9
; GFX1232-NEXT: s_cbranch_execnz .LBB14_1
; GFX1232-NEXT: ; %bb.2: ; %atomicrmw.end
; GFX1232-NEXT: s_or_b32 exec_lo, exec_lo, s9
-; GFX1232-NEXT: v_lshrrev_b32_e32 v0, s2, v2
+; GFX1232-NEXT: v_lshrrev_b32_e32 v0, s2, v0
; GFX1232-NEXT: s_mov_b32 s3, 0x31016000
; GFX1232-NEXT: s_mov_b32 s2, -1
; GFX1232-NEXT: buffer_store_b8 v0, off, s[0:3], null
@@ -8818,7 +8830,7 @@ define amdgpu_kernel void @uniform_or_i16(ptr addrspace(1) %result, ptr addrspac
; GFX7LESS-NEXT: v_mbcnt_lo_u32_b32_e64 v0, exec_lo, 0
; GFX7LESS-NEXT: v_mbcnt_hi_u32_b32_e32 v0, exec_hi, v0
; GFX7LESS-NEXT: v_cmp_eq_u32_e32 vcc, 0, v0
-; GFX7LESS-NEXT: ; implicit-def: $vgpr0
+; GFX7LESS-NEXT: ; implicit-def: $vgpr0
; GFX7LESS-NEXT: s_and_saveexec_b64 s[4:5], vcc
; GFX7LESS-NEXT: s_cbranch_execz .LBB15_2
; GFX7LESS-NEXT: ; %bb.1:
@@ -9326,9 +9338,9 @@ define amdgpu_kernel void @uniform_add_i16(ptr addrspace(1) %result, ptr addrspa
; GFX7LESS-NEXT: s_load_dwordx4 s[0:3], s[4:5], 0x9
; GFX7LESS-NEXT: s_load_dword s10, s[4:5], 0xd
; GFX7LESS-NEXT: v_mbcnt_lo_u32_b32_e64 v0, s6, 0
-; GFX7LESS-NEXT: v_mbcnt_hi_u32_b32_e32 v4, s7, v0
-; GFX7LESS-NEXT: v_cmp_eq_u32_e32 vcc, 0, v4
-; GFX7LESS-NEXT: ; implicit-def: $vgpr0
+; GFX7LESS-NEXT: v_mbcnt_hi_u32_b32_e32 v2, s7, v0
+; GFX7LESS-NEXT: v_cmp_eq_u32_e32 vcc, 0, v2
+; GFX7LESS-NEXT: ; implicit-def: $vgpr0
; GFX7LESS-NEXT: s_and_saveexec_b64 s[8:9], vcc
; GFX7LESS-NEXT: s_cbranch_execz .LBB16_4
; GFX7LESS-NEXT: ; %bb.1:
@@ -9347,27 +9359,28 @@ define amdgpu_kernel void @uniform_add_i16(ptr addrspace(1) %result, ptr addrspa
; GFX7LESS-NEXT: s_not_b32 s13, s12
; GFX7LESS-NEXT: s_lshl_b32 s14, s6, s11
; GFX7LESS-NEXT: s_waitcnt lgkmcnt(0)
-; GFX7LESS-NEXT: v_mov_b32_e32 v1, s15
+; GFX7LESS-NEXT: v_mov_b32_e32 v0, s15
; GFX7LESS-NEXT: s_mov_b32 s6, -1
; GFX7LESS-NEXT: .LBB16_2: ; %atomicrmw.start
; GFX7LESS-NEXT: ; =>This Inner Loop Header: Depth=1
-; GFX7LESS-NEXT: v_add_i32_e32 v0, vcc, s14, v1
-; GFX7LESS-NEXT: v_and_b32_e32 v0, s12, v0
+; GFX7LESS-NEXT: v_mov_b32_e32 v4, v0
; GFX7LESS-NEXT: s_waitcnt expcnt(0)
-; GFX7LESS-NEXT: v_and_b32_e32 v2, s13, v1
-; GFX7LESS-NEXT: v_or_b32_e32 v0, v2, v0
-; GFX7LESS-NEXT: v_mov_b32_e32 v3, v1
-; GFX7LESS-NEXT: v_mov_b32_e32 v2, v0
-; GFX7LESS-NEXT: buffer_atomic_cmpswap v[2:3], off, s[4:7], 0 glc
+; GFX7LESS-NEXT: v_add_i32_e32 v0, vcc, s14, v4
+; GFX7LESS-NEXT: v_and_b32_e32 v0, s12, v0
+; GFX7LESS-NEXT: v_and_b32_e32 v1, s13, v4
+; GFX7LESS-NEXT: v_or_b32_e32 v3, v1, v0
+; GFX7LESS-NEXT: v_mov_b32_e32 v0, v3
+; GFX7LESS-NEXT: v_mov_b32_e32 v1, v4
+; GFX7LESS-NEXT: buffer_atomic_cmpswap v[0:1], off, s[4:7], 0 glc
; GFX7LESS-NEXT: s_waitcnt vmcnt(0)
-; GFX7LESS-NEXT: v_cmp_eq_u32_e32 vcc, v2, v1
+; GFX7LESS-NEXT: v_cmp_eq_u32_e32 vcc, v0, v4
; GFX7LESS-NEXT: s_or_b64 s[2:3], vcc, s[2:3]
-; GFX7LESS-NEXT: v_mov_b32_e32 v1, v2
; GFX7LESS-NEXT: s_andn2_b64 exec, exec, s[2:3]
; GFX7LESS-NEXT: s_cbranch_execnz .LBB16_2
; GFX7LESS-NEXT: ; %bb.3: ; %atomicrmw.end
; GFX7LESS-NEXT: s_or_b64 exec, exec, s[2:3]
-; GFX7LESS-NEXT: v_bfe_u32 v0, v2, s11, 16
+; GFX7LESS-NEXT: s_waitcnt expcnt(0)
+; GFX7LESS-NEXT: v_bfe_u32 v0, v0, s11, 16
; GFX7LESS-NEXT: .LBB16_4: ; %Flow
; GFX7LESS-NEXT: s_or_b64 exec, exec, s[8:9]
; GFX7LESS-NEXT: s_waitcnt lgkmcnt(0)
@@ -9377,7 +9390,7 @@ define amdgpu_kernel void @uniform_add_i16(ptr addrspace(1) %result, ptr addrspa
; GFX7LESS-NEXT: v_readfirstlane_b32 s4, v0
; GFX7LESS-NEXT: s_and_b32 s5, s10, 0xffff
; GFX7LESS-NEXT: v_mov_b32_e32 v0, s4
-; GFX7LESS-NEXT: v_mad_u32_u24 v0, s5, v4, v0
+; GFX7LESS-NEXT: v_mad_u32_u24 v0, s5, v2, v0
; GFX7LESS-NEXT: buffer_store_short v0, off, s[0:3], 0
; GFX7LESS-NEXT: s_endpgm
;
@@ -9387,8 +9400,8 @@ define amdgpu_kernel void @uniform_add_i16(ptr addrspace(1) %result, ptr addrspa
; GFX8-NEXT: s_load_dword s10, s[4:5], 0x34
; GFX8-NEXT: s_mov_b64 s[6:7], exec
; GFX8-NEXT: v_mbcnt_lo_u32_b32 v0, s6, 0
-; GFX8-NEXT: v_mbcnt_hi_u32_b32 v4, s7, v0
-; GFX8-NEXT: v_cmp_eq_u32_e32 vcc, 0, v4
+; GFX8-NEXT: v_mbcnt_hi_u32_b32 v2, s7, v0
+; GFX8-NEXT: v_cmp_eq_u32_e32 vcc, 0, v2
; GFX8-NEXT: ; implicit-def: $vgpr0
; GFX8-NEXT: s_and_saveexec_b64 s[8:9], vcc
; GFX8-NEXT: s_cbranch_execz .LBB16_4
@@ -9407,27 +9420,27 @@ define amdgpu_kernel void @uniform_add_i16(ptr addrspace(1) %result, ptr addrspa
; GFX8-NEXT: s_lshl_b32 s14, s2, s11
; GFX8-NEXT: s_mov_b64 s[2:3], 0
; GFX8-NEXT: s_waitcnt lgkmcnt(0)
-; GFX8-NEXT: v_mov_b32_e32 v1, s7
+; GFX8-NEXT: v_mov_b32_e32 v0, s7
; GFX8-NEXT: s_mov_b32 s7, 0xf000
; GFX8-NEXT: s_mov_b32 s6, -1
; GFX8-NEXT: .LBB16_2: ; %atomicrmw.start
; GFX8-NEXT: ; =>This Inner Loop Header: Depth=1
-; GFX8-NEXT: v_add_u32_e32 v0, vcc, s14, v1
-; GFX8-NEXT: v_and_b32_e32 v2, s13, v1
+; GFX8-NEXT: v_mov_b32_e32 v4, v0
+; GFX8-NEXT: v_add_u32_e32 v0, vcc, s14, v4
+; GFX8-NEXT: v_and_b32_e32 v1, s13, v4
; GFX8-NEXT: v_and_b32_e32 v0, s12, v0
-; GFX8-NEXT: v_or_b32_e32 v0, v2, v0
-; GFX8-NEXT: v_mov_b32_e32 v3, v1
-; GFX8-NEXT: v_mov_b32_e32 v2, v0
-; GFX8-NEXT: buffer_atomic_cmpswap v[2:3], off, s[4:7], 0 glc
+; GFX8-NEXT: v_or_b32_e32 v3, v1, v0
+; GFX8-NEXT: v_mov_b32_e32 v0, v3
+; GFX8-NEXT: v_mov_b32_e32 v1, v4
+; GFX8-NEXT: buffer_atomic_cmpswap v[0:1], off, s[4:7], 0 glc
; GFX8-NEXT: s_waitcnt vmcnt(0)
-; GFX8-NEXT: v_cmp_eq_u32_e32 vcc, v2, v1
+; GFX8-NEXT: v_cmp_eq_u32_e32 vcc, v0, v4
; GFX8-NEXT: s_or_b64 s[2:3], vcc, s[2:3]
-; GFX8-NEXT: v_mov_b32_e32 v1, v2
; GFX8-NEXT: s_andn2_b64 exec, exec, s[2:3]
; GFX8-NEXT: s_cbranch_execnz .LBB16_2
; GFX8-NEXT: ; %bb.3: ; %atomicrmw.end
; GFX8-NEXT: s_or_b64 exec, exec, s[2:3]
-; GFX8-NEXT: v_lshrrev_b32_e32 v0, s11, v2
+; GFX8-NEXT: v_lshrrev_b32_e32 v0, s11, v0
; GFX8-NEXT: .LBB16_4: ; %Flow
; GFX8-NEXT: s_or_b64 exec, exec, s[8:9]
; GFX8-NEXT: v_and_b32_e32 v0, 0xffff, v0
@@ -9436,7 +9449,7 @@ define amdgpu_kernel void @uniform_add_i16(ptr addrspace(1) %result, ptr addrspa
; GFX8-NEXT: s_waitcnt lgkmcnt(0)
; GFX8-NEXT: s_mov_b32 s3, 0xf000
; GFX8-NEXT: s_mov_b32 s2, -1
-; GFX8-NEXT: v_mad_u16 v0, s10, v4, v0
+; GFX8-NEXT: v_mad_u16 v0, s10, v2, v0
; GFX8-NEXT: buffer_store_short v0, off, s[0:3], 0
; GFX8-NEXT: s_endpgm
;
@@ -9446,8 +9459,8 @@ define amdgpu_kernel void @uniform_add_i16(ptr addrspace(1) %result, ptr addrspa
; GFX9-NEXT: s_load_dword s10, s[4:5], 0x34
; GFX9-NEXT: s_mov_b64 s[6:7], exec
; GFX9-NEXT: v_mbcnt_lo_u32_b32 v0, s6, 0
-; GFX9-NEXT: v_mbcnt_hi_u32_b32 v4, s7, v0
-; GFX9-NEXT: v_cmp_eq_u32_e32 vcc, 0, v4
+; GFX9-NEXT: v_mbcnt_hi_u32_b32 v2, s7, v0
+; GFX9-NEXT: v_cmp_eq_u32_e32 vcc, 0, v2
; GFX9-NEXT: ; implicit-def: $vgpr0
; GFX9-NEXT: s_and_saveexec_b64 s[8:9], vcc
; GFX9-NEXT: s_cbranch_execz .LBB16_4
@@ -9466,26 +9479,26 @@ define amdgpu_kernel void @uniform_add_i16(ptr addrspace(1) %result, ptr addrspa
; GFX9-NEXT: s_lshl_b32 s14, s2, s11
; GFX9-NEXT: s_mov_b64 s[2:3], 0
; GFX9-NEXT: s_waitcnt lgkmcnt(0)
-; GFX9-NEXT: v_mov_b32_e32 v1, s7
+; GFX9-NEXT: v_mov_b32_e32 v0, s7
; GFX9-NEXT: s_mov_b32 s7, 0xf000
; GFX9-NEXT: s_mov_b32 s6, -1
; GFX9-NEXT: .LBB16_2: ; %atomicrmw.start
; GFX9-NEXT: ; =>This Inner Loop Header: Depth=1
-; GFX9-NEXT: v_add_u32_e32 v0, s14, v1
+; GFX9-NEXT: v_mov_b32_e32 v4, v0
+; GFX9-NEXT: v_add_u32_e32 v0, s14, v4
; GFX9-NEXT: v_and_b32_e32 v0, s12, v0
-; GFX9-NEXT: v_and_or_b32 v0, v1, s13, v0
-; GFX9-NEXT: v_mov_b32_e32 v3, v1
-; GFX9-NEXT: v_mov_b32_e32 v2, v0
-; GFX9-NEXT: buffer_atomic_cmpswap v[2:3], off, s[4:7], 0 glc
+; GFX9-NEXT: v_and_or_b32 v3, v4, s13, v0
+; GFX9-NEXT: v_mov_b32_e32 v0, v3
+; GFX9-NEXT: v_mov_b32_e32 v1, v4
+; GFX9-NEXT: buffer_atomic_cmpswap v[0:1], off, s[4:7], 0 glc
; GFX9-NEXT: s_waitcnt vmcnt(0)
-; GFX9-NEXT: v_cmp_eq_u32_e32 vcc, v2, v1
+; GFX9-NEXT: v_cmp_eq_u32_e32 vcc, v0, v4
; GFX9-NEXT: s_or_b64 s[2:3], vcc, s[2:3]
-; GFX9-NEXT: v_mov_b32_e32 v1, v2
; GFX9-NEXT: s_andn2_b64 exec, exec, s[2:3]
; GFX9-NEXT: s_cbranch_execnz .LBB16_2
; GFX9-NEXT: ; %bb.3: ; %atomicrmw.end
; GFX9-NEXT: s_or_b64 exec, exec, s[2:3]
-; GFX9-NEXT: v_lshrrev_b32_e32 v0, s11, v2
+; GFX9-NEXT: v_lshrrev_b32_e32 v0, s11, v0
; GFX9-NEXT: .LBB16_4: ; %Flow
; GFX9-NEXT: s_or_b64 exec, exec, s[8:9]
; GFX9-NEXT: v_and_b32_e32 v0, 0xffff, v0
@@ -9494,7 +9507,7 @@ define amdgpu_kernel void @uniform_add_i16(ptr addrspace(1) %result, ptr addrspa
; GFX9-NEXT: s_waitcnt lgkmcnt(0)
; GFX9-NEXT: s_mov_b32 s3, 0xf000
; GFX9-NEXT: s_mov_b32 s2, -1
-; GFX9-NEXT: v_mad_legacy_u16 v0, s10, v4, v0
+; GFX9-NEXT: v_mad_legacy_u16 v0, s10, v2, v0
; GFX9-NEXT: buffer_store_short v0, off, s[0:3], 0
; GFX9-NEXT: s_endpgm
;
@@ -9505,9 +9518,9 @@ define amdgpu_kernel void @uniform_add_i16(ptr addrspace(1) %result, ptr addrspa
; GFX1064-NEXT: s_load_dword s10, s[4:5], 0x34
; GFX1064-NEXT: s_mov_b64 s[6:7], exec
; GFX1064-NEXT: v_mbcnt_lo_u32_b32 v0, s6, 0
-; GFX1064-NEXT: v_mbcnt_hi_u32_b32 v4, s7, v0
+; GFX1064-NEXT: v_mbcnt_hi_u32_b32 v2, s7, v0
; GFX1064-NEXT: ; implicit-def: $vgpr0
-; GFX1064-NEXT: v_cmp_eq_u32_e32 vcc, 0, v4
+; GFX1064-NEXT: v_cmp_eq_u32_e32 vcc, 0, v2
; GFX1064-NEXT: s_and_saveexec_b64 s[8:9], vcc
; GFX1064-NEXT: s_cbranch_execz .LBB16_4
; GFX1064-NEXT: ; %bb.1:
@@ -9526,32 +9539,32 @@ define amdgpu_kernel void @uniform_add_i16(ptr addrspace(1) %result, ptr addrspa
; GFX1064-NEXT: s_mov_b32 s7, 0x31016000
; GFX1064-NEXT: s_mov_b32 s6, -1
; GFX1064-NEXT: s_waitcnt lgkmcnt(0)
-; GFX1064-NEXT: v_mov_b32_e32 v1, s3
+; GFX1064-NEXT: v_mov_b32_e32 v0, s3
; GFX1064-NEXT: s_mov_b64 s[2:3], 0
; GFX1064-NEXT: .LBB16_2: ; %atomicrmw.start
; GFX1064-NEXT: ; =>This Inner Loop Header: Depth=1
-; GFX1064-NEXT: v_add_nc_u32_e32 v0, s14, v1
+; GFX1064-NEXT: v_mov_b32_e32 v4, v0
+; GFX1064-NEXT: v_add_nc_u32_e32 v0, s14, v4
; GFX1064-NEXT: v_and_b32_e32 v0, s12, v0
-; GFX1064-NEXT: v_and_or_b32 v0, v1, s13, v0
-; GFX1064-NEXT: v_mov_b32_e32 v3, v1
-; GFX1064-NEXT: v_mov_b32_e32 v2, v0
-; GFX1064-NEXT: buffer_atomic_cmpswap v[2:3], off, s[4:7], 0 glc
+; GFX1064-NEXT: v_and_or_b32 v3, v4, s13, v0
+; GFX1064-NEXT: v_mov_b32_e32 v0, v3
+; GFX1064-NEXT: v_mov_b32_e32 v1, v4
+; GFX1064-NEXT: buffer_atomic_cmpswap v[0:1], off, s[4:7], 0 glc
; GFX1064-NEXT: s_waitcnt vmcnt(0)
-; GFX1064-NEXT: v_cmp_eq_u32_e32 vcc, v2, v1
-; GFX1064-NEXT: v_mov_b32_e32 v1, v2
+; GFX1064-NEXT: v_cmp_eq_u32_e32 vcc, v0, v4
; GFX1064-NEXT: s_or_b64 s[2:3], vcc, s[2:3]
; GFX1064-NEXT: s_andn2_b64 exec, exec, s[2:3]
; GFX1064-NEXT: s_cbranch_execnz .LBB16_2
; GFX1064-NEXT: ; %bb.3: ; %atomicrmw.end
; GFX1064-NEXT: s_or_b64 exec, exec, s[2:3]
-; GFX1064-NEXT: v_lshrrev_b32_e32 v0, s11, v2
+; GFX1064-NEXT: v_lshrrev_b32_e32 v0, s11, v0
; GFX1064-NEXT: .LBB16_4: ; %Flow
; GFX1064-NEXT: s_or_b64 exec, exec, s[8:9]
; GFX1064-NEXT: v_and_b32_e32 v0, 0xffff, v0
; GFX1064-NEXT: s_waitcnt lgkmcnt(0)
; GFX1064-NEXT: s_mov_b32 s3, 0x31016000
; GFX1064-NEXT: v_readfirstlane_b32 s2, v0
-; GFX1064-NEXT: v_mad_u16 v0, s10, v4, s2
+; GFX1064-NEXT: v_mad_u16 v0, s10, v2, s2
; GFX1064-NEXT: s_mov_b32 s2, -1
; GFX1064-NEXT: buffer_store_short v0, off, s[0:3], 0
; GFX1064-NEXT: s_endpgm
@@ -9563,9 +9576,9 @@ define amdgpu_kernel void @uniform_add_i16(ptr addrspace(1) %result, ptr addrspa
; GFX1032-NEXT: s_load_dword s8, s[4:5], 0x34
; GFX1032-NEXT: s_mov_b32 s6, exec_lo
; GFX1032-NEXT: s_mov_b32 s10, 0
-; GFX1032-NEXT: v_mbcnt_lo_u32_b32 v4, s6, 0
+; GFX1032-NEXT: v_mbcnt_lo_u32_b32 v2, s6, 0
; GFX1032-NEXT: ; implicit-def: $vgpr0
-; GFX1032-NEXT: v_cmp_eq_u32_e32 vcc_lo, 0, v4
+; GFX1032-NEXT: v_cmp_eq_u32_e32 vcc_lo, 0, v2
; GFX1032-NEXT: s_and_saveexec_b32 s9, vcc_lo
; GFX1032-NEXT: s_cbranch_execz .LBB16_4
; GFX1032-NEXT: ; %bb.1:
@@ -9583,32 +9596,32 @@ define amdgpu_kernel void @uniform_add_i16(ptr addrspace(1) %result, ptr addrspa
; GFX1032-NEXT: s_lshl_b32 s12, s6, s2
; GFX1032-NEXT: s_mov_b32 s6, -1
; GFX1032-NEXT: s_waitcnt lgkmcnt(0)
-; GFX1032-NEXT: v_mov_b32_e32 v1, s7
+; GFX1032-NEXT: v_mov_b32_e32 v0, s7
; GFX1032-NEXT: s_mov_b32 s7, 0x31016000
; GFX1032-NEXT: .LBB16_2: ; %atomicrmw.start
; GFX1032-NEXT: ; =>This Inner Loop Header: Depth=1
-; GFX1032-NEXT: v_add_nc_u32_e32 v0, s12, v1
+; GFX1032-NEXT: v_mov_b32_e32 v4, v0
+; GFX1032-NEXT: v_add_nc_u32_e32 v0, s12, v4
; GFX1032-NEXT: v_and_b32_e32 v0, s3, v0
-; GFX1032-NEXT: v_and_or_b32 v0, v1, s11, v0
-; GFX1032-NEXT: v_mov_b32_e32 v3, v1
-; GFX1032-NEXT: v_mov_b32_e32 v2, v0
-; GFX1032-NEXT: buffer_atomic_cmpswap v[2:3], off, s[4:7], 0 glc
+; GFX1032-NEXT: v_and_or_b32 v3, v4, s11, v0
+; GFX1032-NEXT: v_mov_b32_e32 v0, v3
+; GFX1032-NEXT: v_mov_b32_e32 v1, v4
+; GFX1032-NEXT: buffer_atomic_cmpswap v[0:1], off, s[4:7], 0 glc
; GFX1032-NEXT: s_waitcnt vmcnt(0)
-; GFX1032-NEXT: v_cmp_eq_u32_e32 vcc_lo, v2, v1
-; GFX1032-NEXT: v_mov_b32_e32 v1, v2
+; GFX1032-NEXT: v_cmp_eq_u32_e32 vcc_lo, v0, v4
; GFX1032-NEXT: s_or_b32 s10, vcc_lo, s10
; GFX1032-NEXT: s_andn2_b32 exec_lo, exec_lo, s10
; GFX1032-NEXT: s_cbranch_execnz .LBB16_2
; GFX1032-NEXT: ; %bb.3: ; %atomicrmw.end
; GFX1032-NEXT: s_or_b32 exec_lo, exec_lo, s10
-; GFX1032-NEXT: v_lshrrev_b32_e32 v0, s2, v2
+; GFX1032-NEXT: v_lshrrev_b32_e32 v0, s2, v0
; GFX1032-NEXT: .LBB16_4: ; %Flow
; GFX1032-NEXT: s_or_b32 exec_lo, exec_lo, s9
; GFX1032-NEXT: v_and_b32_e32 v0, 0xffff, v0
; GFX1032-NEXT: s_waitcnt lgkmcnt(0)
; GFX1032-NEXT: s_mov_b32 s3, 0x31016000
; GFX1032-NEXT: v_readfirstlane_b32 s2, v0
-; GFX1032-NEXT: v_mad_u16 v0, s8, v4, s2
+; GFX1032-NEXT: v_mad_u16 v0, s8, v2, s2
; GFX1032-NEXT: s_mov_b32 s2, -1
; GFX1032-NEXT: buffer_store_short v0, off, s[0:3], 0
; GFX1032-NEXT: s_endpgm
@@ -9622,9 +9635,9 @@ define amdgpu_kernel void @uniform_add_i16(ptr addrspace(1) %result, ptr addrspa
; GFX1164-TRUE16-NEXT: s_mov_b64 s[8:9], exec
; GFX1164-TRUE16-NEXT: v_mbcnt_lo_u32_b32 v0, s6, 0
; GFX1164-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
-; GFX1164-TRUE16-NEXT: v_mbcnt_hi_u32_b32 v4, s7, v0
+; GFX1164-TRUE16-NEXT: v_mbcnt_hi_u32_b32 v2, s7, v0
; GFX1164-TRUE16-NEXT: ; implicit-def: $vgpr0_lo16
-; GFX1164-TRUE16-NEXT: v_cmpx_eq_u32_e32 0, v4
+; GFX1164-TRUE16-NEXT: v_cmpx_eq_u32_e32 0, v2
; GFX1164-TRUE16-NEXT: s_cbranch_execz .LBB16_4
; GFX1164-TRUE16-NEXT: ; %bb.1:
; GFX1164-TRUE16-NEXT: s_waitcnt lgkmcnt(0)
@@ -9643,29 +9656,29 @@ define amdgpu_kernel void @uniform_add_i16(ptr addrspace(1) %result, ptr addrspa
; GFX1164-TRUE16-NEXT: s_mov_b32 s7, 0x31016000
; GFX1164-TRUE16-NEXT: s_mov_b32 s6, -1
; GFX1164-TRUE16-NEXT: s_waitcnt lgkmcnt(0)
-; GFX1164-TRUE16-NEXT: v_mov_b32_e32 v1, s3
+; GFX1164-TRUE16-NEXT: v_mov_b32_e32 v0, s3
; GFX1164-TRUE16-NEXT: s_mov_b64 s[2:3], 0
; GFX1164-TRUE16-NEXT: .LBB16_2: ; %atomicrmw.start
; GFX1164-TRUE16-NEXT: ; =>This Inner Loop Header: Depth=1
; GFX1164-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
-; GFX1164-TRUE16-NEXT: v_add_nc_u32_e32 v0, s14, v1
+; GFX1164-TRUE16-NEXT: v_mov_b32_e32 v4, v0
+; GFX1164-TRUE16-NEXT: v_add_nc_u32_e32 v0, s14, v4
+; GFX1164-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
; GFX1164-TRUE16-NEXT: v_and_b32_e32 v0, s12, v0
-; GFX1164-TRUE16-NEXT: s_waitcnt_depctr 0xfff
-; GFX1164-TRUE16-NEXT: v_and_or_b32 v0, v1, s13, v0
-; GFX1164-TRUE16-NEXT: v_mov_b32_e32 v3, v1
-; GFX1164-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_2)
-; GFX1164-TRUE16-NEXT: v_mov_b32_e32 v2, v0
-; GFX1164-TRUE16-NEXT: buffer_atomic_cmpswap_b32 v[2:3], off, s[4:7], 0 glc
+; GFX1164-TRUE16-NEXT: v_and_or_b32 v3, v4, s13, v0
+; GFX1164-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1)
+; GFX1164-TRUE16-NEXT: v_mov_b32_e32 v0, v3
+; GFX1164-TRUE16-NEXT: v_mov_b32_e32 v1, v4
+; GFX1164-TRUE16-NEXT: buffer_atomic_cmpswap_b32 v[0:1], off, s[4:7], 0 glc
; GFX1164-TRUE16-NEXT: s_waitcnt vmcnt(0)
-; GFX1164-TRUE16-NEXT: v_cmp_eq_u32_e32 vcc, v2, v1
-; GFX1164-TRUE16-NEXT: v_mov_b32_e32 v1, v2
+; GFX1164-TRUE16-NEXT: v_cmp_eq_u32_e32 vcc, v0, v4
; GFX1164-TRUE16-NEXT: s_or_b64 s[2:3], vcc, s[2:3]
; GFX1164-TRUE16-NEXT: s_delay_alu instid0(SALU_CYCLE_1)
; GFX1164-TRUE16-NEXT: s_and_not1_b64 exec, exec, s[2:3]
; GFX1164-TRUE16-NEXT: s_cbranch_execnz .LBB16_2
; GFX1164-TRUE16-NEXT: ; %bb.3: ; %atomicrmw.end
; GFX1164-TRUE16-NEXT: s_or_b64 exec, exec, s[2:3]
-; GFX1164-TRUE16-NEXT: v_lshrrev_b32_e32 v0, s11, v2
+; GFX1164-TRUE16-NEXT: v_lshrrev_b32_e32 v0, s11, v0
; GFX1164-TRUE16-NEXT: .LBB16_4: ; %Flow
; GFX1164-TRUE16-NEXT: s_or_b64 exec, exec, s[8:9]
; GFX1164-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(SKIP_2) | instid1(VALU_DEP_1)
@@ -9674,7 +9687,7 @@ define amdgpu_kernel void @uniform_add_i16(ptr addrspace(1) %result, ptr addrspa
; GFX1164-TRUE16-NEXT: s_mov_b32 s3, 0x31016000
; GFX1164-TRUE16-NEXT: v_readfirstlane_b32 s2, v0
; GFX1164-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1)
-; GFX1164-TRUE16-NEXT: v_mad_u16 v0.l, s10, v4.l, s2
+; GFX1164-TRUE16-NEXT: v_mad_u16 v0.l, s10, v2.l, s2
; GFX1164-TRUE16-NEXT: s_mov_b32 s2, -1
; GFX1164-TRUE16-NEXT: buffer_store_b16 v0, off, s[0:3], 0
; GFX1164-TRUE16-NEXT: s_endpgm
@@ -9688,9 +9701,9 @@ define amdgpu_kernel void @uniform_add_i16(ptr addrspace(1) %result, ptr addrspa
; GFX1164-FAKE16-NEXT: s_mov_b64 s[8:9], exec
; GFX1164-FAKE16-NEXT: v_mbcnt_lo_u32_b32 v0, s6, 0
; GFX1164-FAKE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
-; GFX1164-FAKE16-NEXT: v_mbcnt_hi_u32_b32 v4, s7, v0
+; GFX1164-FAKE16-NEXT: v_mbcnt_hi_u32_b32 v2, s7, v0
; GFX1164-FAKE16-NEXT: ; implicit-def: $vgpr0
-; GFX1164-FAKE16-NEXT: v_cmpx_eq_u32_e32 0, v4
+; GFX1164-FAKE16-NEXT: v_cmpx_eq_u32_e32 0, v2
; GFX1164-FAKE16-NEXT: s_cbranch_execz .LBB16_4
; GFX1164-FAKE16-NEXT: ; %bb.1:
; GFX1164-FAKE16-NEXT: s_waitcnt lgkmcnt(0)
@@ -9708,29 +9721,29 @@ define amdgpu_kernel void @uniform_add_i16(ptr addrspace(1) %result, ptr addrspa
; GFX1164-FAKE16-NEXT: s_mov_b32 s7, 0x31016000
; GFX1164-FAKE16-NEXT: s_mov_b32 s6, -1
; GFX1164-FAKE16-NEXT: s_waitcnt lgkmcnt(0)
-; GFX1164-FAKE16-NEXT: v_mov_b32_e32 v1, s3
+; GFX1164-FAKE16-NEXT: v_mov_b32_e32 v0, s3
; GFX1164-FAKE16-NEXT: s_mov_b64 s[2:3], 0
; GFX1164-FAKE16-NEXT: .LBB16_2: ; %atomicrmw.start
; GFX1164-FAKE16-NEXT: ; =>This Inner Loop Header: Depth=1
; GFX1164-FAKE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
-; GFX1164-FAKE16-NEXT: v_add_nc_u32_e32 v0, s14, v1
+; GFX1164-FAKE16-NEXT: v_mov_b32_e32 v4, v0
+; GFX1164-FAKE16-NEXT: v_add_nc_u32_e32 v0, s14, v4
+; GFX1164-FAKE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
; GFX1164-FAKE16-NEXT: v_and_b32_e32 v0, s12, v0
-; GFX1164-FAKE16-NEXT: s_waitcnt_depctr 0xfff
-; GFX1164-FAKE16-NEXT: v_and_or_b32 v0, v1, s13, v0
-; GFX1164-FAKE16-NEXT: v_mov_b32_e32 v3, v1
-; GFX1164-FAKE16-NEXT: s_delay_alu instid0(VALU_DEP_2)
-; GFX1164-FAKE16-NEXT: v_mov_b32_e32 v2, v0
-; GFX1164-FAKE16-NEXT: buffer_atomic_cmpswap_b32 v[2:3], off, s[4:7], 0 glc
+; GFX1164-FAKE16-NEXT: v_and_or_b32 v3, v4, s13, v0
+; GFX1164-FAKE16-NEXT: s_delay_alu instid0(VALU_DEP_1)
+; GFX1164-FAKE16-NEXT: v_mov_b32_e32 v0, v3
+; GFX1164-FAKE16-NEXT: v_mov_b32_e32 v1, v4
+; GFX1164-FAKE16-NEXT: buffer_atomic_cmpswap_b32 v[0:1], off, s[4:7], 0 glc
; GFX1164-FAKE16-NEXT: s_waitcnt vmcnt(0)
-; GFX1164-FAKE16-NEXT: v_cmp_eq_u32_e32 vcc, v2, v1
-; GFX1164-FAKE16-NEXT: v_mov_b32_e32 v1, v2
+; GFX1164-FAKE16-NEXT: v_cmp_eq_u32_e32 vcc, v0, v4
; GFX1164-FAKE16-NEXT: s_or_b64 s[2:3], vcc, s[2:3]
; GFX1164-FAKE16-NEXT: s_delay_alu instid0(SALU_CYCLE_1)
; GFX1164-FAKE16-NEXT: s_and_not1_b64 exec, exec, s[2:3]
; GFX1164-FAKE16-NEXT: s_cbranch_execnz .LBB16_2
; GFX1164-FAKE16-NEXT: ; %bb.3: ; %atomicrmw.end
; GFX1164-FAKE16-NEXT: s_or_b64 exec, exec, s[2:3]
-; GFX1164-FAKE16-NEXT: v_lshrrev_b32_e32 v0, s11, v2
+; GFX1164-FAKE16-NEXT: v_lshrrev_b32_e32 v0, s11, v0
; GFX1164-FAKE16-NEXT: .LBB16_4: ; %Flow
; GFX1164-FAKE16-NEXT: s_or_b64 exec, exec, s[8:9]
; GFX1164-FAKE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(SKIP_2) | instid1(VALU_DEP_1)
@@ -9739,7 +9752,7 @@ define amdgpu_kernel void @uniform_add_i16(ptr addrspace(1) %result, ptr addrspa
; GFX1164-FAKE16-NEXT: s_mov_b32 s3, 0x31016000
; GFX1164-FAKE16-NEXT: v_readfirstlane_b32 s2, v0
; GFX1164-FAKE16-NEXT: s_delay_alu instid0(VALU_DEP_1)
-; GFX1164-FAKE16-NEXT: v_mad_u16 v0, s10, v4, s2
+; GFX1164-FAKE16-NEXT: v_mad_u16 v0, s10, v2, s2
; GFX1164-FAKE16-NEXT: s_mov_b32 s2, -1
; GFX1164-FAKE16-NEXT: buffer_store_b16 v0, off, s[0:3], 0
; GFX1164-FAKE16-NEXT: s_endpgm
@@ -9751,11 +9764,11 @@ define amdgpu_kernel void @uniform_add_i16(ptr addrspace(1) %result, ptr addrspa
; GFX1132-TRUE16-NEXT: s_load_b32 s8, s[4:5], 0x34
; GFX1132-TRUE16-NEXT: s_mov_b32 s6, exec_lo
; GFX1132-TRUE16-NEXT: s_mov_b32 s10, 0
-; GFX1132-TRUE16-NEXT: v_mbcnt_lo_u32_b32 v4, s6, 0
+; GFX1132-TRUE16-NEXT: v_mbcnt_lo_u32_b32 v2, s6, 0
; GFX1132-TRUE16-NEXT: s_mov_b32 s9, exec_lo
; GFX1132-TRUE16-NEXT: ; implicit-def: $vgpr0_lo16
; GFX1132-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1)
-; GFX1132-TRUE16-NEXT: v_cmpx_eq_u32_e32 0, v4
+; GFX1132-TRUE16-NEXT: v_cmpx_eq_u32_e32 0, v2
; GFX1132-TRUE16-NEXT: s_cbranch_execz .LBB16_4
; GFX1132-TRUE16-NEXT: ; %bb.1:
; GFX1132-TRUE16-NEXT: s_waitcnt lgkmcnt(0)
@@ -9773,27 +9786,27 @@ define amdgpu_kernel void @uniform_add_i16(ptr addrspace(1) %result, ptr addrspa
; GFX1132-TRUE16-NEXT: s_lshl_b32 s12, s6, s2
; GFX1132-TRUE16-NEXT: s_mov_b32 s6, -1
; GFX1132-TRUE16-NEXT: s_waitcnt lgkmcnt(0)
-; GFX1132-TRUE16-NEXT: v_mov_b32_e32 v1, s7
+; GFX1132-TRUE16-NEXT: v_mov_b32_e32 v0, s7
; GFX1132-TRUE16-NEXT: s_mov_b32 s7, 0x31016000
; GFX1132-TRUE16-NEXT: .LBB16_2: ; %atomicrmw.start
; GFX1132-TRUE16-NEXT: ; =>This Inner Loop Header: Depth=1
; GFX1132-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
-; GFX1132-TRUE16-NEXT: v_add_nc_u32_e32 v0, s12, v1
-; GFX1132-TRUE16-NEXT: v_and_b32_e32 v0, s3, v0
+; GFX1132-TRUE16-NEXT: v_mov_b32_e32 v4, v0
+; GFX1132-TRUE16-NEXT: v_add_nc_u32_e32 v0, s12, v4
; GFX1132-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
-; GFX1132-TRUE16-NEXT: v_and_or_b32 v0, v1, s11, v0
-; GFX1132-TRUE16-NEXT: v_dual_mov_b32 v3, v1 :: v_dual_mov_b32 v2, v0
-; GFX1132-TRUE16-NEXT: buffer_atomic_cmpswap_b32 v[2:3], off, s[4:7], 0 glc
+; GFX1132-TRUE16-NEXT: v_and_b32_e32 v0, s3, v0
+; GFX1132-TRUE16-NEXT: v_and_or_b32 v3, v4, s11, v0
+; GFX1132-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(SKIP_4) | instid1(SALU_CYCLE_1)
+; GFX1132-TRUE16-NEXT: v_dual_mov_b32 v0, v3 :: v_dual_mov_b32 v1, v4
+; GFX1132-TRUE16-NEXT: buffer_atomic_cmpswap_b32 v[0:1], off, s[4:7], 0 glc
; GFX1132-TRUE16-NEXT: s_waitcnt vmcnt(0)
-; GFX1132-TRUE16-NEXT: v_cmp_eq_u32_e32 vcc_lo, v2, v1
-; GFX1132-TRUE16-NEXT: v_mov_b32_e32 v1, v2
+; GFX1132-TRUE16-NEXT: v_cmp_eq_u32_e32 vcc_lo, v0, v4
; GFX1132-TRUE16-NEXT: s_or_b32 s10, vcc_lo, s10
-; GFX1132-TRUE16-NEXT: s_delay_alu instid0(SALU_CYCLE_1)
; GFX1132-TRUE16-NEXT: s_and_not1_b32 exec_lo, exec_lo, s10
; GFX1132-TRUE16-NEXT: s_cbranch_execnz .LBB16_2
; GFX1132-TRUE16-NEXT: ; %bb.3: ; %atomicrmw.end
; GFX1132-TRUE16-NEXT: s_or_b32 exec_lo, exec_lo, s10
-; GFX1132-TRUE16-NEXT: v_lshrrev_b32_e32 v0, s2, v2
+; GFX1132-TRUE16-NEXT: v_lshrrev_b32_e32 v0, s2, v0
; GFX1132-TRUE16-NEXT: .LBB16_4: ; %Flow
; GFX1132-TRUE16-NEXT: s_or_b32 exec_lo, exec_lo, s9
; GFX1132-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(SKIP_2) | instid1(VALU_DEP_1)
@@ -9802,7 +9815,7 @@ define amdgpu_kernel void @uniform_add_i16(ptr addrspace(1) %result, ptr addrspa
; GFX1132-TRUE16-NEXT: s_mov_b32 s3, 0x31016000
; GFX1132-TRUE16-NEXT: v_readfirstlane_b32 s2, v0
; GFX1132-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1)
-; GFX1132-TRUE16-NEXT: v_mad_u16 v0.l, s8, v4.l, s2
+; GFX1132-TRUE16-NEXT: v_mad_u16 v0.l, s8, v2.l, s2
; GFX1132-TRUE16-NEXT: s_mov_b32 s2, -1
; GFX1132-TRUE16-NEXT: buffer_store_b16 v0, off, s[0:3], 0
; GFX1132-TRUE16-NEXT: s_endpgm
@@ -9814,11 +9827,11 @@ define amdgpu_kernel void @uniform_add_i16(ptr addrspace(1) %result, ptr addrspa
; GFX1132-FAKE16-NEXT: s_load_b32 s8, s[4:5], 0x34
; GFX1132-FAKE16-NEXT: s_mov_b32 s6, exec_lo
; GFX1132-FAKE16-NEXT: s_mov_b32 s10, 0
-; GFX1132-FAKE16-NEXT: v_mbcnt_lo_u32_b32 v4, s6, 0
+; GFX1132-FAKE16-NEXT: v_mbcnt_lo_u32_b32 v2, s6, 0
; GFX1132-FAKE16-NEXT: s_mov_b32 s9, exec_lo
; GFX1132-FAKE16-NEXT: ; implicit-def: $vgpr0
; GFX1132-FAKE16-NEXT: s_delay_alu instid0(VALU_DEP_1)
-; GFX1132-FAKE16-NEXT: v_cmpx_eq_u32_e32 0, v4
+; GFX1132-FAKE16-NEXT: v_cmpx_eq_u32_e32 0, v2
; GFX1132-FAKE16-NEXT: s_cbranch_execz .LBB16_4
; GFX1132-FAKE16-NEXT: ; %bb.1:
; GFX1132-FAKE16-NEXT: s_waitcnt lgkmcnt(0)
@@ -9835,27 +9848,27 @@ define amdgpu_kernel void @uniform_add_i16(ptr addrspace(1) %result, ptr addrspa
; GFX1132-FAKE16-NEXT: s_lshl_b32 s12, s6, s2
; GFX1132-FAKE16-NEXT: s_mov_b32 s6, -1
; GFX1132-FAKE16-NEXT: s_waitcnt lgkmcnt(0)
-; GFX1132-FAKE16-NEXT: v_mov_b32_e32 v1, s7
+; GFX1132-FAKE16-NEXT: v_mov_b32_e32 v0, s7
; GFX1132-FAKE16-NEXT: s_mov_b32 s7, 0x31016000
; GFX1132-FAKE16-NEXT: .LBB16_2: ; %atomicrmw.start
; GFX1132-FAKE16-NEXT: ; =>This Inner Loop Header: Depth=1
; GFX1132-FAKE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
-; GFX1132-FAKE16-NEXT: v_add_nc_u32_e32 v0, s12, v1
-; GFX1132-FAKE16-NEXT: v_and_b32_e32 v0, s3, v0
+; GFX1132-FAKE16-NEXT: v_mov_b32_e32 v4, v0
+; GFX1132-FAKE16-NEXT: v_add_nc_u32_e32 v0, s12, v4
; GFX1132-FAKE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
-; GFX1132-FAKE16-NEXT: v_and_or_b32 v0, v1, s11, v0
-; GFX1132-FAKE16-NEXT: v_dual_mov_b32 v3, v1 :: v_dual_mov_b32 v2, v0
-; GFX1132-FAKE16-NEXT: buffer_atomic_cmpswap_b32 v[2:3], off, s[4:7], 0 glc
+; GFX1132-FAKE16-NEXT: v_and_b32_e32 v0, s3, v0
+; GFX1132-FAKE16-NEXT: v_and_or_b32 v3, v4, s11, v0
+; GFX1132-FAKE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(SKIP_4) | instid1(SALU_CYCLE_1)
+; GFX1132-FAKE16-NEXT: v_dual_mov_b32 v0, v3 :: v_dual_mov_b32 v1, v4
+; GFX1132-FAKE16-NEXT: buffer_atomic_cmpswap_b32 v[0:1], off, s[4:7], 0 glc
; GFX1132-FAKE16-NEXT: s_waitcnt vmcnt(0)
-; GFX1132-FAKE16-NEXT: v_cmp_eq_u32_e32 vcc_lo, v2, v1
-; GFX1132-FAKE16-NEXT: v_mov_b32_e32 v1, v2
+; GFX1132-FAKE16-NEXT: v_cmp_eq_u32_e32 vcc_lo, v0, v4
; GFX1132-FAKE16-NEXT: s_or_b32 s10, vcc_lo, s10
-; GFX1132-FAKE16-NEXT: s_delay_alu instid0(SALU_CYCLE_1)
; GFX1132-FAKE16-NEXT: s_and_not1_b32 exec_lo, exec_lo, s10
; GFX1132-FAKE16-NEXT: s_cbranch_execnz .LBB16_2
; GFX1132-FAKE16-NEXT: ; %bb.3: ; %atomicrmw.end
; GFX1132-FAKE16-NEXT: s_or_b32 exec_lo, exec_lo, s10
-; GFX1132-FAKE16-NEXT: v_lshrrev_b32_e32 v0, s2, v2
+; GFX1132-FAKE16-NEXT: v_lshrrev_b32_e32 v0, s2, v0
; GFX1132-FAKE16-NEXT: .LBB16_4: ; %Flow
; GFX1132-FAKE16-NEXT: s_or_b32 exec_lo, exec_lo, s9
; GFX1132-FAKE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(SKIP_2) | instid1(VALU_DEP_1)
@@ -9864,7 +9877,7 @@ define amdgpu_kernel void @uniform_add_i16(ptr addrspace(1) %result, ptr addrspa
; GFX1132-FAKE16-NEXT: s_mov_b32 s3, 0x31016000
; GFX1132-FAKE16-NEXT: v_readfirstlane_b32 s2, v0
; GFX1132-FAKE16-NEXT: s_delay_alu instid0(VALU_DEP_1)
-; GFX1132-FAKE16-NEXT: v_mad_u16 v0, s8, v4, s2
+; GFX1132-FAKE16-NEXT: v_mad_u16 v0, s8, v2, s2
; GFX1132-FAKE16-NEXT: s_mov_b32 s2, -1
; GFX1132-FAKE16-NEXT: buffer_store_b16 v0, off, s[0:3], 0
; GFX1132-FAKE16-NEXT: s_endpgm
@@ -9878,9 +9891,9 @@ define amdgpu_kernel void @uniform_add_i16(ptr addrspace(1) %result, ptr addrspa
; GFX1264-TRUE16-NEXT: s_mov_b64 s[8:9], exec
; GFX1264-TRUE16-NEXT: v_mbcnt_lo_u32_b32 v0, s6, 0
; GFX1264-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
-; GFX1264-TRUE16-NEXT: v_mbcnt_hi_u32_b32 v4, s7, v0
+; GFX1264-TRUE16-NEXT: v_mbcnt_hi_u32_b32 v2, s7, v0
; GFX1264-TRUE16-NEXT: ; implicit-def: $vgpr0_lo16
-; GFX1264-TRUE16-NEXT: v_cmpx_eq_u32_e32 0, v4
+; GFX1264-TRUE16-NEXT: v_cmpx_eq_u32_e32 0, v2
; GFX1264-TRUE16-NEXT: s_cbranch_execz .LBB16_4
; GFX1264-TRUE16-NEXT: ; %bb.1:
; GFX1264-TRUE16-NEXT: s_wait_kmcnt 0x0
@@ -9901,28 +9914,29 @@ define amdgpu_kernel void @uniform_add_i16(ptr addrspace(1) %result, ptr addrspa
; GFX1264-TRUE16-NEXT: s_mov_b32 s7, 0x31016000
; GFX1264-TRUE16-NEXT: s_mov_b32 s6, -1
; GFX1264-TRUE16-NEXT: s_wait_kmcnt 0x0
-; GFX1264-TRUE16-NEXT: v_mov_b32_e32 v1, s3
+; GFX1264-TRUE16-NEXT: v_mov_b32_e32 v0, s3
; GFX1264-TRUE16-NEXT: s_mov_b64 s[2:3], 0
; GFX1264-TRUE16-NEXT: .LBB16_2: ; %atomicrmw.start
; GFX1264-TRUE16-NEXT: ; =>This Inner Loop Header: Depth=1
; GFX1264-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
-; GFX1264-TRUE16-NEXT: v_add_nc_u32_e32 v0, s14, v1
+; GFX1264-TRUE16-NEXT: v_mov_b32_e32 v4, v0
+; GFX1264-TRUE16-NEXT: v_add_nc_u32_e32 v0, s14, v4
+; GFX1264-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
; GFX1264-TRUE16-NEXT: v_and_b32_e32 v0, s12, v0
-; GFX1264-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(SKIP_1) | instid1(VALU_DEP_2)
-; GFX1264-TRUE16-NEXT: v_and_or_b32 v0, v1, s13, v0
-; GFX1264-TRUE16-NEXT: v_mov_b32_e32 v3, v1
-; GFX1264-TRUE16-NEXT: v_mov_b32_e32 v2, v0
-; GFX1264-TRUE16-NEXT: buffer_atomic_cmpswap_b32 v[2:3], off, s[4:7], null th:TH_ATOMIC_RETURN scope:SCOPE_SYS
+; GFX1264-TRUE16-NEXT: v_and_or_b32 v3, v4, s13, v0
+; GFX1264-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1)
+; GFX1264-TRUE16-NEXT: v_mov_b32_e32 v0, v3
+; GFX1264-TRUE16-NEXT: v_mov_b32_e32 v1, v4
+; GFX1264-TRUE16-NEXT: buffer_atomic_cmpswap_b32 v[0:1], off, s[4:7], null th:TH_ATOMIC_RETURN scope:SCOPE_SYS
; GFX1264-TRUE16-NEXT: s_wait_loadcnt 0x0
-; GFX1264-TRUE16-NEXT: v_cmp_eq_u32_e32 vcc, v2, v1
-; GFX1264-TRUE16-NEXT: v_mov_b32_e32 v1, v2
+; GFX1264-TRUE16-NEXT: v_cmp_eq_u32_e32 vcc, v0, v4
; GFX1264-TRUE16-NEXT: s_or_b64 s[2:3], vcc, s[2:3]
; GFX1264-TRUE16-NEXT: s_wait_alu 0xfffe
; GFX1264-TRUE16-NEXT: s_and_not1_b64 exec, exec, s[2:3]
; GFX1264-TRUE16-NEXT: s_cbranch_execnz .LBB16_2
; GFX1264-TRUE16-NEXT: ; %bb.3: ; %atomicrmw.end
; GFX1264-TRUE16-NEXT: s_or_b64 exec, exec, s[2:3]
-; GFX1264-TRUE16-NEXT: v_lshrrev_b32_e32 v0, s11, v2
+; GFX1264-TRUE16-NEXT: v_lshrrev_b32_e32 v0, s11, v0
; GFX1264-TRUE16-NEXT: .LBB16_4: ; %Flow
; GFX1264-TRUE16-NEXT: s_or_b64 exec, exec, s[8:9]
; GFX1264-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(SKIP_2) | instid1(VALU_DEP_1)
@@ -9931,7 +9945,7 @@ define amdgpu_kernel void @uniform_add_i16(ptr addrspace(1) %result, ptr addrspa
; GFX1264-TRUE16-NEXT: s_mov_b32 s3, 0x31016000
; GFX1264-TRUE16-NEXT: v_readfirstlane_b32 s2, v0
; GFX1264-TRUE16-NEXT: s_wait_alu 0xf1ff
-; GFX1264-TRUE16-NEXT: v_mad_u16 v0.l, s10, v4.l, s2
+; GFX1264-TRUE16-NEXT: v_mad_u16 v0.l, s10, v2.l, s2
; GFX1264-TRUE16-NEXT: s_mov_b32 s2, -1
; GFX1264-TRUE16-NEXT: buffer_store_b16 v0, off, s[0:3], null
; GFX1264-TRUE16-NEXT: s_endpgm
@@ -9945,9 +9959,9 @@ define amdgpu_kernel void @uniform_add_i16(ptr addrspace(1) %result, ptr addrspa
; GFX1264-FAKE16-NEXT: s_mov_b64 s[8:9], exec
; GFX1264-FAKE16-NEXT: v_mbcnt_lo_u32_b32 v0, s6, 0
; GFX1264-FAKE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
-; GFX1264-FAKE16-NEXT: v_mbcnt_hi_u32_b32 v4, s7, v0
+; GFX1264-FAKE16-NEXT: v_mbcnt_hi_u32_b32 v2, s7, v0
; GFX1264-FAKE16-NEXT: ; implicit-def: $vgpr0
-; GFX1264-FAKE16-NEXT: v_cmpx_eq_u32_e32 0, v4
+; GFX1264-FAKE16-NEXT: v_cmpx_eq_u32_e32 0, v2
; GFX1264-FAKE16-NEXT: s_cbranch_execz .LBB16_4
; GFX1264-FAKE16-NEXT: ; %bb.1:
; GFX1264-FAKE16-NEXT: s_wait_kmcnt 0x0
@@ -9966,28 +9980,29 @@ define amdgpu_kernel void @uniform_add_i16(ptr addrspace(1) %result, ptr addrspa
; GFX1264-FAKE16-NEXT: s_mov_b32 s7, 0x31016000
; GFX1264-FAKE16-NEXT: s_mov_b32 s6, -1
; GFX1264-FAKE16-NEXT: s_wait_kmcnt 0x0
-; GFX1264-FAKE16-NEXT: v_mov_b32_e32 v1, s3
+; GFX1264-FAKE16-NEXT: v_mov_b32_e32 v0, s3
; GFX1264-FAKE16-NEXT: s_mov_b64 s[2:3], 0
; GFX1264-FAKE16-NEXT: .LBB16_2: ; %atomicrmw.start
; GFX1264-FAKE16-NEXT: ; =>This Inner Loop Header: Depth=1
; GFX1264-FAKE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
-; GFX1264-FAKE16-NEXT: v_add_nc_u32_e32 v0, s14, v1
+; GFX1264-FAKE16-NEXT: v_mov_b32_e32 v4, v0
+; GFX1264-FAKE16-NEXT: v_add_nc_u32_e32 v0, s14, v4
+; GFX1264-FAKE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
; GFX1264-FAKE16-NEXT: v_and_b32_e32 v0, s12, v0
-; GFX1264-FAKE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(SKIP_1) | instid1(VALU_DEP_2)
-; GFX1264-FAKE16-NEXT: v_and_or_b32 v0, v1, s13, v0
-; GFX1264-FAKE16-NEXT: v_mov_b32_e32 v3, v1
-; GFX1264-FAKE16-NEXT: v_mov_b32_e32 v2, v0
-; GFX1264-FAKE16-NEXT: buffer_atomic_cmpswap_b32 v[2:3], off, s[4:7], null th:TH_ATOMIC_RETURN scope:SCOPE_SYS
+; GFX1264-FAKE16-NEXT: v_and_or_b32 v3, v4, s13, v0
+; GFX1264-FAKE16-NEXT: s_delay_alu instid0(VALU_DEP_1)
+; GFX1264-FAKE16-NEXT: v_mov_b32_e32 v0, v3
+; GFX1264-FAKE16-NEXT: v_mov_b32_e32 v1, v4
+; GFX1264-FAKE16-NEXT: buffer_atomic_cmpswap_b32 v[0:1], off, s[4:7], null th:TH_ATOMIC_RETURN scope:SCOPE_SYS
; GFX1264-FAKE16-NEXT: s_wait_loadcnt 0x0
-; GFX1264-FAKE16-NEXT: v_cmp_eq_u32_e32 vcc, v2, v1
-; GFX1264-FAKE16-NEXT: v_mov_b32_e32 v1, v2
+; GFX1264-FAKE16-NEXT: v_cmp_eq_u32_e32 vcc, v0, v4
; GFX1264-FAKE16-NEXT: s_or_b64 s[2:3], vcc, s[2:3]
; GFX1264-FAKE16-NEXT: s_wait_alu 0xfffe
; GFX1264-FAKE16-NEXT: s_and_not1_b64 exec, exec, s[2:3]
; GFX1264-FAKE16-NEXT: s_cbranch_execnz .LBB16_2
; GFX1264-FAKE16-NEXT: ; %bb.3: ; %atomicrmw.end
; GFX1264-FAKE16-NEXT: s_or_b64 exec, exec, s[2:3]
-; GFX1264-FAKE16-NEXT: v_lshrrev_b32_e32 v0, s11, v2
+; GFX1264-FAKE16-NEXT: v_lshrrev_b32_e32 v0, s11, v0
; GFX1264-FAKE16-NEXT: .LBB16_4: ; %Flow
; GFX1264-FAKE16-NEXT: s_or_b64 exec, exec, s[8:9]
; GFX1264-FAKE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(SKIP_2) | instid1(VALU_DEP_1)
@@ -9996,7 +10011,7 @@ define amdgpu_kernel void @uniform_add_i16(ptr addrspace(1) %result, ptr addrspa
; GFX1264-FAKE16-NEXT: s_mov_b32 s3, 0x31016000
; GFX1264-FAKE16-NEXT: v_readfirstlane_b32 s2, v0
; GFX1264-FAKE16-NEXT: s_wait_alu 0xf1ff
-; GFX1264-FAKE16-NEXT: v_mad_u16 v0, s10, v4, s2
+; GFX1264-FAKE16-NEXT: v_mad_u16 v0, s10, v2, s2
; GFX1264-FAKE16-NEXT: s_mov_b32 s2, -1
; GFX1264-FAKE16-NEXT: buffer_store_b16 v0, off, s[0:3], null
; GFX1264-FAKE16-NEXT: s_endpgm
@@ -10008,11 +10023,11 @@ define amdgpu_kernel void @uniform_add_i16(ptr addrspace(1) %result, ptr addrspa
; GFX1232-TRUE16-NEXT: s_load_b32 s8, s[4:5], 0x34
; GFX1232-TRUE16-NEXT: s_mov_b32 s6, exec_lo
; GFX1232-TRUE16-NEXT: s_mov_b32 s10, 0
-; GFX1232-TRUE16-NEXT: v_mbcnt_lo_u32_b32 v4, s6, 0
+; GFX1232-TRUE16-NEXT: v_mbcnt_lo_u32_b32 v2, s6, 0
; GFX1232-TRUE16-NEXT: s_mov_b32 s9, exec_lo
; GFX1232-TRUE16-NEXT: ; implicit-def: $vgpr0_lo16
; GFX1232-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1)
-; GFX1232-TRUE16-NEXT: v_cmpx_eq_u32_e32 0, v4
+; GFX1232-TRUE16-NEXT: v_cmpx_eq_u32_e32 0, v2
; GFX1232-TRUE16-NEXT: s_cbranch_execz .LBB16_4
; GFX1232-TRUE16-NEXT: ; %bb.1:
; GFX1232-TRUE16-NEXT: s_wait_kmcnt 0x0
@@ -10033,27 +10048,28 @@ define amdgpu_kernel void @uniform_add_i16(ptr addrspace(1) %result, ptr addrspa
; GFX1232-TRUE16-NEXT: s_lshl_b32 s12, s6, s2
; GFX1232-TRUE16-NEXT: s_mov_b32 s6, -1
; GFX1232-TRUE16-NEXT: s_wait_kmcnt 0x0
-; GFX1232-TRUE16-NEXT: v_mov_b32_e32 v1, s7
+; GFX1232-TRUE16-NEXT: v_mov_b32_e32 v0, s7
; GFX1232-TRUE16-NEXT: s_mov_b32 s7, 0x31016000
; GFX1232-TRUE16-NEXT: .LBB16_2: ; %atomicrmw.start
; GFX1232-TRUE16-NEXT: ; =>This Inner Loop Header: Depth=1
; GFX1232-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
-; GFX1232-TRUE16-NEXT: v_add_nc_u32_e32 v0, s12, v1
-; GFX1232-TRUE16-NEXT: v_and_b32_e32 v0, s3, v0
+; GFX1232-TRUE16-NEXT: v_mov_b32_e32 v4, v0
+; GFX1232-TRUE16-NEXT: v_add_nc_u32_e32 v0, s12, v4
; GFX1232-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
-; GFX1232-TRUE16-NEXT: v_and_or_b32 v0, v1, s11, v0
-; GFX1232-TRUE16-NEXT: v_dual_mov_b32 v3, v1 :: v_dual_mov_b32 v2, v0
-; GFX1232-TRUE16-NEXT: buffer_atomic_cmpswap_b32 v[2:3], off, s[4:7], null th:TH_ATOMIC_RETURN scope:SCOPE_SYS
+; GFX1232-TRUE16-NEXT: v_and_b32_e32 v0, s3, v0
+; GFX1232-TRUE16-NEXT: v_and_or_b32 v3, v4, s11, v0
+; GFX1232-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1)
+; GFX1232-TRUE16-NEXT: v_dual_mov_b32 v0, v3 :: v_dual_mov_b32 v1, v4
+; GFX1232-TRUE16-NEXT: buffer_atomic_cmpswap_b32 v[0:1], off, s[4:7], null th:TH_ATOMIC_RETURN scope:SCOPE_SYS
; GFX1232-TRUE16-NEXT: s_wait_loadcnt 0x0
-; GFX1232-TRUE16-NEXT: v_cmp_eq_u32_e32 vcc_lo, v2, v1
-; GFX1232-TRUE16-NEXT: v_mov_b32_e32 v1, v2
+; GFX1232-TRUE16-NEXT: v_cmp_eq_u32_e32 vcc_lo, v0, v4
; GFX1232-TRUE16-NEXT: s_or_b32 s10, vcc_lo, s10
; GFX1232-TRUE16-NEXT: s_wait_alu 0xfffe
; GFX1232-TRUE16-NEXT: s_and_not1_b32 exec_lo, exec_lo, s10
; GFX1232-TRUE16-NEXT: s_cbranch_execnz .LBB16_2
; GFX1232-TRUE16-NEXT: ; %bb.3: ; %atomicrmw.end
; GFX1232-TRUE16-NEXT: s_or_b32 exec_lo, exec_lo, s10
-; GFX1232-TRUE16-NEXT: v_lshrrev_b32_e32 v0, s2, v2
+; GFX1232-TRUE16-NEXT: v_lshrrev_b32_e32 v0, s2, v0
; GFX1232-TRUE16-NEXT: .LBB16_4: ; %Flow
; GFX1232-TRUE16-NEXT: s_or_b32 exec_lo, exec_lo, s9
; GFX1232-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(SKIP_2) | instid1(VALU_DEP_1)
@@ -10062,7 +10078,7 @@ define amdgpu_kernel void @uniform_add_i16(ptr addrspace(1) %result, ptr addrspa
; GFX1232-TRUE16-NEXT: s_mov_b32 s3, 0x31016000
; GFX1232-TRUE16-NEXT: v_readfirstlane_b32 s2, v0
; GFX1232-TRUE16-NEXT: s_wait_alu 0xf1ff
-; GFX1232-TRUE16-NEXT: v_mad_u16 v0.l, s8, v4.l, s2
+; GFX1232-TRUE16-NEXT: v_mad_u16 v0.l, s8, v2.l, s2
; GFX1232-TRUE16-NEXT: s_mov_b32 s2, -1
; GFX1232-TRUE16-NEXT: buffer_store_b16 v0, off, s[0:3], null
; GFX1232-TRUE16-NEXT: s_endpgm
@@ -10074,11 +10090,11 @@ define amdgpu_kernel void @uniform_add_i16(ptr addrspace(1) %result, ptr addrspa
; GFX1232-FAKE16-NEXT: s_load_b32 s8, s[4:5], 0x34
; GFX1232-FAKE16-NEXT: s_mov_b32 s6, exec_lo
; GFX1232-FAKE16-NEXT: s_mov_b32 s10, 0
-; GFX1232-FAKE16-NEXT: v_mbcnt_lo_u32_b32 v4, s6, 0
+; GFX1232-FAKE16-NEXT: v_mbcnt_lo_u32_b32 v2, s6, 0
; GFX1232-FAKE16-NEXT: s_mov_b32 s9, exec_lo
; GFX1232-FAKE16-NEXT: ; implicit-def: $vgpr0
; GFX1232-FAKE16-NEXT: s_delay_alu instid0(VALU_DEP_1)
-; GFX1232-FAKE16-NEXT: v_cmpx_eq_u32_e32 0, v4
+; GFX1232-FAKE16-NEXT: v_cmpx_eq_u32_e32 0, v2
; GFX1232-FAKE16-NEXT: s_cbranch_execz .LBB16_4
; GFX1232-FAKE16-NEXT: ; %bb.1:
; GFX1232-FAKE16-NEXT: s_wait_kmcnt 0x0
@@ -10098,27 +10114,28 @@ define amdgpu_kernel void @uniform_add_i16(ptr addrspace(1) %result, ptr addrspa
; GFX1232-FAKE16-NEXT: s_lshl_b32 s12, s6, s2
; GFX1232-FAKE16-NEXT: s_mov_b32 s6, -1
; GFX1232-FAKE16-NEXT: s_wait_kmcnt 0x0
-; GFX1232-FAKE16-NEXT: v_mov_b32_e32 v1, s7
+; GFX1232-FAKE16-NEXT: v_mov_b32_e32 v0, s7
; GFX1232-FAKE16-NEXT: s_mov_b32 s7, 0x31016000
; GFX1232-FAKE16-NEXT: .LBB16_2: ; %atomicrmw.start
; GFX1232-FAKE16-NEXT: ; =>This Inner Loop Header: Depth=1
; GFX1232-FAKE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
-; GFX1232-FAKE16-NEXT: v_add_nc_u32_e32 v0, s12, v1
-; GFX1232-FAKE16-NEXT: v_and_b32_e32 v0, s3, v0
+; GFX1232-FAKE16-NEXT: v_mov_b32_e32 v4, v0
+; GFX1232-FAKE16-NEXT: v_add_nc_u32_e32 v0, s12, v4
; GFX1232-FAKE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
-; GFX1232-FAKE16-NEXT: v_and_or_b32 v0, v1, s11, v0
-; GFX1232-FAKE16-NEXT: v_dual_mov_b32 v3, v1 :: v_dual_mov_b32 v2, v0
-; GFX1232-FAKE16-NEXT: buffer_atomic_cmpswap_b32 v[2:3], off, s[4:7], null th:TH_ATOMIC_RETURN scope:SCOPE_SYS
+; GFX1232-FAKE16-NEXT: v_and_b32_e32 v0, s3, v0
+; GFX1232-FAKE16-NEXT: v_and_or_b32 v3, v4, s11, v0
+; GFX1232-FAKE16-NEXT: s_delay_alu instid0(VALU_DEP_1)
+; GFX1232-FAKE16-NEXT: v_dual_mov_b32 v0, v3 :: v_dual_mov_b32 v1, v4
+; GFX1232-FAKE16-NEXT: buffer_atomic_cmpswap_b32 v[0:1], off, s[4:7], null th:TH_ATOMIC_RETURN scope:SCOPE_SYS
; GFX1232-FAKE16-NEXT: s_wait_loadcnt 0x0
-; GFX1232-FAKE16-NEXT: v_cmp_eq_u32_e32 vcc_lo, v2, v1
-; GFX1232-FAKE16-NEXT: v_mov_b32_e32 v1, v2
+; GFX1232-FAKE16-NEXT: v_cmp_eq_u32_e32 vcc_lo, v0, v4
; GFX1232-FAKE16-NEXT: s_or_b32 s10, vcc_lo, s10
; GFX1232-FAKE16-NEXT: s_wait_alu 0xfffe
; GFX1232-FAKE16-NEXT: s_and_not1_b32 exec_lo, exec_lo, s10
; GFX1232-FAKE16-NEXT: s_cbranch_execnz .LBB16_2
; GFX1232-FAKE16-NEXT: ; %bb.3: ; %atomicrmw.end
; GFX1232-FAKE16-NEXT: s_or_b32 exec_lo, exec_lo, s10
-; GFX1232-FAKE16-NEXT: v_lshrrev_b32_e32 v0, s2, v2
+; GFX1232-FAKE16-NEXT: v_lshrrev_b32_e32 v0, s2, v0
; GFX1232-FAKE16-NEXT: .LBB16_4: ; %Flow
; GFX1232-FAKE16-NEXT: s_or_b32 exec_lo, exec_lo, s9
; GFX1232-FAKE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(SKIP_2) | instid1(VALU_DEP_1)
@@ -10127,7 +10144,7 @@ define amdgpu_kernel void @uniform_add_i16(ptr addrspace(1) %result, ptr addrspa
; GFX1232-FAKE16-NEXT: s_mov_b32 s3, 0x31016000
; GFX1232-FAKE16-NEXT: v_readfirstlane_b32 s2, v0
; GFX1232-FAKE16-NEXT: s_wait_alu 0xf1ff
-; GFX1232-FAKE16-NEXT: v_mad_u16 v0, s8, v4, s2
+; GFX1232-FAKE16-NEXT: v_mad_u16 v0, s8, v2, s2
; GFX1232-FAKE16-NEXT: s_mov_b32 s2, -1
; GFX1232-FAKE16-NEXT: buffer_store_b16 v0, off, s[0:3], null
; GFX1232-FAKE16-NEXT: s_endpgm
@@ -10154,27 +10171,28 @@ define amdgpu_kernel void @uniform_xchg_i16(ptr addrspace(1) %result, ptr addrsp
; GFX7LESS-NEXT: s_lshl_b32 s2, s3, s10
; GFX7LESS-NEXT: s_not_b32 s3, s11
; GFX7LESS-NEXT: s_waitcnt lgkmcnt(0)
-; GFX7LESS-NEXT: v_mov_b32_e32 v1, s6
+; GFX7LESS-NEXT: v_mov_b32_e32 v0, s6
; GFX7LESS-NEXT: s_mov_b32 s6, -1
; GFX7LESS-NEXT: .LBB17_1: ; %atomicrmw.start
; GFX7LESS-NEXT: ; =>This Inner Loop Header: Depth=1
-; GFX7LESS-NEXT: v_and_b32_e32 v0, s3, v1
-; GFX7LESS-NEXT: v_or_b32_e32 v0, s2, v0
-; GFX7LESS-NEXT: s_waitcnt expcnt(0)
-; GFX7LESS-NEXT: v_mov_b32_e32 v3, v1
; GFX7LESS-NEXT: v_mov_b32_e32 v2, v0
-; GFX7LESS-NEXT: buffer_atomic_cmpswap v[2:3], off, s[4:7], 0 glc
+; GFX7LESS-NEXT: s_waitcnt expcnt(0)
+; GFX7LESS-NEXT: v_and_b32_e32 v0, s3, v2
+; GFX7LESS-NEXT: v_or_b32_e32 v1, s2, v0
+; GFX7LESS-NEXT: v_mov_b32_e32 v0, v1
+; GFX7LESS-NEXT: v_mov_b32_e32 v1, v2
+; GFX7LESS-NEXT: buffer_atomic_cmpswap v[0:1], off, s[4:7], 0 glc
; GFX7LESS-NEXT: s_waitcnt vmcnt(0)
-; GFX7LESS-NEXT: v_cmp_eq_u32_e32 vcc, v2, v1
+; GFX7LESS-NEXT: v_cmp_eq_u32_e32 vcc, v0, v2
; GFX7LESS-NEXT: s_or_b64 s[8:9], vcc, s[8:9]
-; GFX7LESS-NEXT: v_mov_b32_e32 v1, v2
; GFX7LESS-NEXT: s_andn2_b64 exec, exec, s[8:9]
; GFX7LESS-NEXT: s_cbranch_execnz .LBB17_1
; GFX7LESS-NEXT: ; %bb.2: ; %atomicrmw.end
; GFX7LESS-NEXT: s_or_b64 exec, exec, s[8:9]
; GFX7LESS-NEXT: s_mov_b32 s3, 0xf000
; GFX7LESS-NEXT: s_mov_b32 s2, -1
-; GFX7LESS-NEXT: v_lshrrev_b32_e32 v0, s10, v2
+; GFX7LESS-NEXT: s_waitcnt expcnt(0)
+; GFX7LESS-NEXT: v_lshrrev_b32_e32 v0, s10, v0
; GFX7LESS-NEXT: buffer_store_short v0, off, s[0:3], 0
; GFX7LESS-NEXT: s_endpgm
;
@@ -10194,27 +10212,27 @@ define amdgpu_kernel void @uniform_xchg_i16(ptr addrspace(1) %result, ptr addrsp
; GFX8-NEXT: s_lshl_b32 s10, s2, s8
; GFX8-NEXT: s_mov_b64 s[2:3], 0
; GFX8-NEXT: s_waitcnt lgkmcnt(0)
-; GFX8-NEXT: v_mov_b32_e32 v1, s7
+; GFX8-NEXT: v_mov_b32_e32 v0, s7
; GFX8-NEXT: s_mov_b32 s7, 0xf000
; GFX8-NEXT: s_mov_b32 s6, -1
; GFX8-NEXT: .LBB17_1: ; %atomicrmw.start
; GFX8-NEXT: ; =>This Inner Loop Header: Depth=1
-; GFX8-NEXT: v_and_b32_e32 v0, s9, v1
-; GFX8-NEXT: v_or_b32_e32 v0, s10, v0
-; GFX8-NEXT: v_mov_b32_e32 v3, v1
; GFX8-NEXT: v_mov_b32_e32 v2, v0
-; GFX8-NEXT: buffer_atomic_cmpswap v[2:3], off, s[4:7], 0 glc
+; GFX8-NEXT: v_and_b32_e32 v0, s9, v2
+; GFX8-NEXT: v_or_b32_e32 v1, s10, v0
+; GFX8-NEXT: v_mov_b32_e32 v0, v1
+; GFX8-NEXT: v_mov_b32_e32 v1, v2
+; GFX8-NEXT: buffer_atomic_cmpswap v[0:1], off, s[4:7], 0 glc
; GFX8-NEXT: s_waitcnt vmcnt(0)
-; GFX8-NEXT: v_cmp_eq_u32_e32 vcc, v2, v1
+; GFX8-NEXT: v_cmp_eq_u32_e32 vcc, v0, v2
; GFX8-NEXT: s_or_b64 s[2:3], vcc, s[2:3]
-; GFX8-NEXT: v_mov_b32_e32 v1, v2
; GFX8-NEXT: s_andn2_b64 exec, exec, s[2:3]
; GFX8-NEXT: s_cbranch_execnz .LBB17_1
; GFX8-NEXT: ; %bb.2: ; %atomicrmw.end
; GFX8-NEXT: s_or_b64 exec, exec, s[2:3]
; GFX8-NEXT: s_mov_b32 s3, 0xf000
; GFX8-NEXT: s_mov_b32 s2, -1
-; GFX8-NEXT: v_lshrrev_b32_e32 v0, s8, v2
+; GFX8-NEXT: v_lshrrev_b32_e32 v0, s8, v0
; GFX8-NEXT: buffer_store_short v0, off, s[0:3], 0
; GFX8-NEXT: s_endpgm
;
@@ -10234,27 +10252,27 @@ define amdgpu_kernel void @uniform_xchg_i16(ptr addrspace(1) %result, ptr addrsp
; GFX9-NEXT: s_lshl_b32 s10, s2, s8
; GFX9-NEXT: s_mov_b64 s[2:3], 0
; GFX9-NEXT: s_waitcnt lgkmcnt(0)
-; GFX9-NEXT: v_mov_b32_e32 v1, s7
+; GFX9-NEXT: v_mov_b32_e32 v0, s7
; GFX9-NEXT: s_mov_b32 s7, 0xf000
; GFX9-NEXT: s_mov_b32 s6, -1
; GFX9-NEXT: .LBB17_1: ; %atomicrmw.start
; GFX9-NEXT: ; =>This Inner Loop Header: Depth=1
-; GFX9-NEXT: v_and_b32_e32 v0, s9, v1
-; GFX9-NEXT: v_or_b32_e32 v0, s10, v0
-; GFX9-NEXT: v_mov_b32_e32 v3, v1
; GFX9-NEXT: v_mov_b32_e32 v2, v0
-; GFX9-NEXT: buffer_atomic_cmpswap v[2:3], off, s[4:7], 0 glc
+; GFX9-NEXT: v_and_b32_e32 v0, s9, v2
+; GFX9-NEXT: v_or_b32_e32 v1, s10, v0
+; GFX9-NEXT: v_mov_b32_e32 v0, v1
+; GFX9-NEXT: v_mov_b32_e32 v1, v2
+; GFX9-NEXT: buffer_atomic_cmpswap v[0:1], off, s[4:7], 0 glc
; GFX9-NEXT: s_waitcnt vmcnt(0)
-; GFX9-NEXT: v_cmp_eq_u32_e32 vcc, v2, v1
+; GFX9-NEXT: v_cmp_eq_u32_e32 vcc, v0, v2
; GFX9-NEXT: s_or_b64 s[2:3], vcc, s[2:3]
-; GFX9-NEXT: v_mov_b32_e32 v1, v2
; GFX9-NEXT: s_andn2_b64 exec, exec, s[2:3]
; GFX9-NEXT: s_cbranch_execnz .LBB17_1
; GFX9-NEXT: ; %bb.2: ; %atomicrmw.end
; GFX9-NEXT: s_or_b64 exec, exec, s[2:3]
; GFX9-NEXT: s_mov_b32 s3, 0xf000
; GFX9-NEXT: s_mov_b32 s2, -1
-; GFX9-NEXT: v_lshrrev_b32_e32 v0, s8, v2
+; GFX9-NEXT: v_lshrrev_b32_e32 v0, s8, v0
; GFX9-NEXT: buffer_store_short v0, off, s[0:3], 0
; GFX9-NEXT: s_endpgm
;
@@ -10276,23 +10294,23 @@ define amdgpu_kernel void @uniform_xchg_i16(ptr addrspace(1) %result, ptr addrsp
; GFX1064-NEXT: s_not_b32 s9, s2
; GFX1064-NEXT: s_mov_b32 s6, -1
; GFX1064-NEXT: s_waitcnt lgkmcnt(0)
-; GFX1064-NEXT: v_mov_b32_e32 v1, s3
+; GFX1064-NEXT: v_mov_b32_e32 v0, s3
; GFX1064-NEXT: s_mov_b64 s[2:3], 0
; GFX1064-NEXT: .LBB17_1: ; %atomicrmw.start
; GFX1064-NEXT: ; =>This Inner Loop Header: Depth=1
-; GFX1064-NEXT: v_and_or_b32 v0, v1, s9, s10
-; GFX1064-NEXT: v_mov_b32_e32 v3, v1
; GFX1064-NEXT: v_mov_b32_e32 v2, v0
-; GFX1064-NEXT: buffer_atomic_cmpswap v[2:3], off, s[4:7], 0 glc
-; GFX1064-NEXT: s_waitcnt vmcnt(0)
-; GFX1064-NEXT: v_cmp_eq_u32_e32 vcc, v2, v1
+; GFX1064-NEXT: v_and_or_b32 v1, v2, s9, s10
+; GFX1064-NEXT: v_mov_b32_e32 v0, v1
; GFX1064-NEXT: v_mov_b32_e32 v1, v2
+; GFX1064-NEXT: buffer_atomic_cmpswap v[0:1], off, s[4:7], 0 glc
+; GFX1064-NEXT: s_waitcnt vmcnt(0)
+; GFX1064-NEXT: v_cmp_eq_u32_e32 vcc, v0, v2
; GFX1064-NEXT: s_or_b64 s[2:3], vcc, s[2:3]
; GFX1064-NEXT: s_andn2_b64 exec, exec, s[2:3]
; GFX1064-NEXT: s_cbranch_execnz .LBB17_1
; GFX1064-NEXT: ; %bb.2: ; %atomicrmw.end
; GFX1064-NEXT: s_or_b64 exec, exec, s[2:3]
-; GFX1064-NEXT: v_lshrrev_b32_e32 v0, s8, v2
+; GFX1064-NEXT: v_lshrrev_b32_e32 v0, s8, v0
; GFX1064-NEXT: s_mov_b32 s3, 0x31016000
; GFX1064-NEXT: s_mov_b32 s2, -1
; GFX1064-NEXT: buffer_store_short v0, off, s[0:3], 0
@@ -10316,23 +10334,23 @@ define amdgpu_kernel void @uniform_xchg_i16(ptr addrspace(1) %result, ptr addrsp
; GFX1032-NEXT: s_not_b32 s3, s3
; GFX1032-NEXT: s_mov_b32 s6, -1
; GFX1032-NEXT: s_waitcnt lgkmcnt(0)
-; GFX1032-NEXT: v_mov_b32_e32 v1, s7
+; GFX1032-NEXT: v_mov_b32_e32 v0, s7
; GFX1032-NEXT: s_mov_b32 s7, 0x31016000
; GFX1032-NEXT: .LBB17_1: ; %atomicrmw.start
; GFX1032-NEXT: ; =>This Inner Loop Header: Depth=1
-; GFX1032-NEXT: v_and_or_b32 v0, v1, s3, s8
-; GFX1032-NEXT: v_mov_b32_e32 v3, v1
; GFX1032-NEXT: v_mov_b32_e32 v2, v0
-; GFX1032-NEXT: buffer_atomic_cmpswap v[2:3], off, s[4:7], 0 glc
-; GFX1032-NEXT: s_waitcnt vmcnt(0)
-; GFX1032-NEXT: v_cmp_eq_u32_e32 vcc_lo, v2, v1
+; GFX1032-NEXT: v_and_or_b32 v1, v2, s3, s8
+; GFX1032-NEXT: v_mov_b32_e32 v0, v1
; GFX1032-NEXT: v_mov_b32_e32 v1, v2
+; GFX1032-NEXT: buffer_atomic_cmpswap v[0:1], off, s[4:7], 0 glc
+; GFX1032-NEXT: s_waitcnt vmcnt(0)
+; GFX1032-NEXT: v_cmp_eq_u32_e32 vcc_lo, v0, v2
; GFX1032-NEXT: s_or_b32 s9, vcc_lo, s9
; GFX1032-NEXT: s_andn2_b32 exec_lo, exec_lo, s9
; GFX1032-NEXT: s_cbranch_execnz .LBB17_1
; GFX1032-NEXT: ; %bb.2: ; %atomicrmw.end
; GFX1032-NEXT: s_or_b32 exec_lo, exec_lo, s9
-; GFX1032-NEXT: v_lshrrev_b32_e32 v0, s2, v2
+; GFX1032-NEXT: v_lshrrev_b32_e32 v0, s2, v0
; GFX1032-NEXT: s_mov_b32 s3, 0x31016000
; GFX1032-NEXT: s_mov_b32 s2, -1
; GFX1032-NEXT: buffer_store_short v0, off, s[0:3], 0
@@ -10356,25 +10374,26 @@ define amdgpu_kernel void @uniform_xchg_i16(ptr addrspace(1) %result, ptr addrsp
; GFX1164-NEXT: s_not_b32 s9, s2
; GFX1164-NEXT: s_mov_b32 s6, -1
; GFX1164-NEXT: s_waitcnt lgkmcnt(0)
-; GFX1164-NEXT: v_mov_b32_e32 v1, s3
+; GFX1164-NEXT: v_mov_b32_e32 v0, s3
; GFX1164-NEXT: s_mov_b64 s[2:3], 0
; GFX1164-NEXT: .LBB17_1: ; %atomicrmw.start
; GFX1164-NEXT: ; =>This Inner Loop Header: Depth=1
-; GFX1164-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(SKIP_1) | instid1(VALU_DEP_2)
-; GFX1164-NEXT: v_and_or_b32 v0, v1, s9, s10
-; GFX1164-NEXT: v_mov_b32_e32 v3, v1
+; GFX1164-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
; GFX1164-NEXT: v_mov_b32_e32 v2, v0
-; GFX1164-NEXT: buffer_atomic_cmpswap_b32 v[2:3], off, s[4:7], 0 glc
-; GFX1164-NEXT: s_waitcnt vmcnt(0)
-; GFX1164-NEXT: v_cmp_eq_u32_e32 vcc, v2, v1
+; GFX1164-NEXT: v_and_or_b32 v1, v2, s9, s10
+; GFX1164-NEXT: s_delay_alu instid0(VALU_DEP_1)
+; GFX1164-NEXT: v_mov_b32_e32 v0, v1
; GFX1164-NEXT: v_mov_b32_e32 v1, v2
+; GFX1164-NEXT: buffer_atomic_cmpswap_b32 v[0:1], off, s[4:7], 0 glc
+; GFX1164-NEXT: s_waitcnt vmcnt(0)
+; GFX1164-NEXT: v_cmp_eq_u32_e32 vcc, v0, v2
; GFX1164-NEXT: s_or_b64 s[2:3], vcc, s[2:3]
; GFX1164-NEXT: s_delay_alu instid0(SALU_CYCLE_1)
; GFX1164-NEXT: s_and_not1_b64 exec, exec, s[2:3]
; GFX1164-NEXT: s_cbranch_execnz .LBB17_1
; GFX1164-NEXT: ; %bb.2: ; %atomicrmw.end
; GFX1164-NEXT: s_or_b64 exec, exec, s[2:3]
-; GFX1164-NEXT: v_lshrrev_b32_e32 v0, s8, v2
+; GFX1164-NEXT: v_lshrrev_b32_e32 v0, s8, v0
; GFX1164-NEXT: s_mov_b32 s3, 0x31016000
; GFX1164-NEXT: s_mov_b32 s2, -1
; GFX1164-NEXT: buffer_store_b16 v0, off, s[0:3], 0
@@ -10398,24 +10417,26 @@ define amdgpu_kernel void @uniform_xchg_i16(ptr addrspace(1) %result, ptr addrsp
; GFX1132-NEXT: s_not_b32 s3, s3
; GFX1132-NEXT: s_mov_b32 s6, -1
; GFX1132-NEXT: s_waitcnt lgkmcnt(0)
-; GFX1132-NEXT: v_mov_b32_e32 v1, s7
+; GFX1132-NEXT: v_mov_b32_e32 v0, s7
; GFX1132-NEXT: s_mov_b32 s7, 0x31016000
; GFX1132-NEXT: .LBB17_1: ; %atomicrmw.start
; GFX1132-NEXT: ; =>This Inner Loop Header: Depth=1
; GFX1132-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
-; GFX1132-NEXT: v_and_or_b32 v0, v1, s3, s8
-; GFX1132-NEXT: v_dual_mov_b32 v3, v1 :: v_dual_mov_b32 v2, v0
-; GFX1132-NEXT: buffer_atomic_cmpswap_b32 v[2:3], off, s[4:7], 0 glc
-; GFX1132-NEXT: s_waitcnt vmcnt(0)
-; GFX1132-NEXT: v_cmp_eq_u32_e32 vcc_lo, v2, v1
+; GFX1132-NEXT: v_mov_b32_e32 v2, v0
+; GFX1132-NEXT: v_and_or_b32 v1, v2, s3, s8
+; GFX1132-NEXT: s_delay_alu instid0(VALU_DEP_1)
+; GFX1132-NEXT: v_mov_b32_e32 v0, v1
; GFX1132-NEXT: v_mov_b32_e32 v1, v2
+; GFX1132-NEXT: buffer_atomic_cmpswap_b32 v[0:1], off, s[4:7], 0 glc
+; GFX1132-NEXT: s_waitcnt vmcnt(0)
+; GFX1132-NEXT: v_cmp_eq_u32_e32 vcc_lo, v0, v2
; GFX1132-NEXT: s_or_b32 s9, vcc_lo, s9
; GFX1132-NEXT: s_delay_alu instid0(SALU_CYCLE_1)
; GFX1132-NEXT: s_and_not1_b32 exec_lo, exec_lo, s9
; GFX1132-NEXT: s_cbranch_execnz .LBB17_1
; GFX1132-NEXT: ; %bb.2: ; %atomicrmw.end
; GFX1132-NEXT: s_or_b32 exec_lo, exec_lo, s9
-; GFX1132-NEXT: v_lshrrev_b32_e32 v0, s2, v2
+; GFX1132-NEXT: v_lshrrev_b32_e32 v0, s2, v0
; GFX1132-NEXT: s_mov_b32 s3, 0x31016000
; GFX1132-NEXT: s_mov_b32 s2, -1
; GFX1132-NEXT: buffer_store_b16 v0, off, s[0:3], 0
@@ -10439,25 +10460,26 @@ define amdgpu_kernel void @uniform_xchg_i16(ptr addrspace(1) %result, ptr addrsp
; GFX1264-NEXT: s_not_b32 s9, s2
; GFX1264-NEXT: s_mov_b32 s6, -1
; GFX1264-NEXT: s_wait_kmcnt 0x0
-; GFX1264-NEXT: v_mov_b32_e32 v1, s3
+; GFX1264-NEXT: v_mov_b32_e32 v0, s3
; GFX1264-NEXT: s_mov_b64 s[2:3], 0
; GFX1264-NEXT: .LBB17_1: ; %atomicrmw.start
; GFX1264-NEXT: ; =>This Inner Loop Header: Depth=1
-; GFX1264-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(SKIP_1) | instid1(VALU_DEP_2)
-; GFX1264-NEXT: v_and_or_b32 v0, v1, s9, s10
-; GFX1264-NEXT: v_mov_b32_e32 v3, v1
+; GFX1264-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
; GFX1264-NEXT: v_mov_b32_e32 v2, v0
-; GFX1264-NEXT: buffer_atomic_cmpswap_b32 v[2:3], off, s[4:7], null th:TH_ATOMIC_RETURN scope:SCOPE_SYS
-; GFX1264-NEXT: s_wait_loadcnt 0x0
-; GFX1264-NEXT: v_cmp_eq_u32_e32 vcc, v2, v1
+; GFX1264-NEXT: v_and_or_b32 v1, v2, s9, s10
+; GFX1264-NEXT: s_delay_alu instid0(VALU_DEP_1)
+; GFX1264-NEXT: v_mov_b32_e32 v0, v1
; GFX1264-NEXT: v_mov_b32_e32 v1, v2
+; GFX1264-NEXT: buffer_atomic_cmpswap_b32 v[0:1], off, s[4:7], null th:TH_ATOMIC_RETURN scope:SCOPE_SYS
+; GFX1264-NEXT: s_wait_loadcnt 0x0
+; GFX1264-NEXT: v_cmp_eq_u32_e32 vcc, v0, v2
; GFX1264-NEXT: s_or_b64 s[2:3], vcc, s[2:3]
; GFX1264-NEXT: s_wait_alu 0xfffe
; GFX1264-NEXT: s_and_not1_b64 exec, exec, s[2:3]
; GFX1264-NEXT: s_cbranch_execnz .LBB17_1
; GFX1264-NEXT: ; %bb.2: ; %atomicrmw.end
; GFX1264-NEXT: s_or_b64 exec, exec, s[2:3]
-; GFX1264-NEXT: v_lshrrev_b32_e32 v0, s8, v2
+; GFX1264-NEXT: v_lshrrev_b32_e32 v0, s8, v0
; GFX1264-NEXT: s_mov_b32 s3, 0x31016000
; GFX1264-NEXT: s_mov_b32 s2, -1
; GFX1264-NEXT: buffer_store_b16 v0, off, s[0:3], null
@@ -10481,24 +10503,26 @@ define amdgpu_kernel void @uniform_xchg_i16(ptr addrspace(1) %result, ptr addrsp
; GFX1232-NEXT: s_not_b32 s3, s3
; GFX1232-NEXT: s_mov_b32 s6, -1
; GFX1232-NEXT: s_wait_kmcnt 0x0
-; GFX1232-NEXT: v_mov_b32_e32 v1, s7
+; GFX1232-NEXT: v_mov_b32_e32 v0, s7
; GFX1232-NEXT: s_mov_b32 s7, 0x31016000
; GFX1232-NEXT: .LBB17_1: ; %atomicrmw.start
; GFX1232-NEXT: ; =>This Inner Loop Header: Depth=1
; GFX1232-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
-; GFX1232-NEXT: v_and_or_b32 v0, v1, s3, s8
-; GFX1232-NEXT: v_dual_mov_b32 v3, v1 :: v_dual_mov_b32 v2, v0
-; GFX1232-NEXT: buffer_atomic_cmpswap_b32 v[2:3], off, s[4:7], null th:TH_ATOMIC_RETURN scope:SCOPE_SYS
-; GFX1232-NEXT: s_wait_loadcnt 0x0
-; GFX1232-NEXT: v_cmp_eq_u32_e32 vcc_lo, v2, v1
+; GFX1232-NEXT: v_mov_b32_e32 v2, v0
+; GFX1232-NEXT: v_and_or_b32 v1, v2, s3, s8
+; GFX1232-NEXT: s_delay_alu instid0(VALU_DEP_1)
+; GFX1232-NEXT: v_mov_b32_e32 v0, v1
; GFX1232-NEXT: v_mov_b32_e32 v1, v2
+; GFX1232-NEXT: buffer_atomic_cmpswap_b32 v[0:1], off, s[4:7], null th:TH_ATOMIC_RETURN scope:SCOPE_SYS
+; GFX1232-NEXT: s_wait_loadcnt 0x0
+; GFX1232-NEXT: v_cmp_eq_u32_e32 vcc_lo, v0, v2
; GFX1232-NEXT: s_or_b32 s9, vcc_lo, s9
; GFX1232-NEXT: s_wait_alu 0xfffe
; GFX1232-NEXT: s_and_not1_b32 exec_lo, exec_lo, s9
; GFX1232-NEXT: s_cbranch_execnz .LBB17_1
; GFX1232-NEXT: ; %bb.2: ; %atomicrmw.end
; GFX1232-NEXT: s_or_b32 exec_lo, exec_lo, s9
-; GFX1232-NEXT: v_lshrrev_b32_e32 v0, s2, v2
+; GFX1232-NEXT: v_lshrrev_b32_e32 v0, s2, v0
; GFX1232-NEXT: s_mov_b32 s3, 0x31016000
; GFX1232-NEXT: s_mov_b32 s2, -1
; GFX1232-NEXT: buffer_store_b16 v0, off, s[0:3], null
@@ -10524,35 +10548,36 @@ define amdgpu_kernel void @uniform_fadd_f16(ptr addrspace(1) %result, ptr addrsp
; GFX7LESS-NEXT: s_load_dword s3, s[4:5], 0x0
; GFX7LESS-NEXT: v_cvt_f16_f32_e32 v0, v0
; GFX7LESS-NEXT: s_lshl_b32 s2, 0xffff, s10
-; GFX7LESS-NEXT: v_cvt_f32_f16_e32 v4, v0
+; GFX7LESS-NEXT: v_cvt_f32_f16_e32 v2, v0
; GFX7LESS-NEXT: s_not_b32 s2, s2
; GFX7LESS-NEXT: s_waitcnt lgkmcnt(0)
-; GFX7LESS-NEXT: v_mov_b32_e32 v1, s3
+; GFX7LESS-NEXT: v_mov_b32_e32 v0, s3
; GFX7LESS-NEXT: s_mov_b32 s6, -1
; GFX7LESS-NEXT: .LBB18_1: ; %atomicrmw.start
; GFX7LESS-NEXT: ; =>This Inner Loop Header: Depth=1
-; GFX7LESS-NEXT: v_lshrrev_b32_e32 v0, s10, v1
+; GFX7LESS-NEXT: v_mov_b32_e32 v4, v0
+; GFX7LESS-NEXT: s_waitcnt expcnt(0)
+; GFX7LESS-NEXT: v_lshrrev_b32_e32 v0, s10, v4
; GFX7LESS-NEXT: v_cvt_f32_f16_e32 v0, v0
-; GFX7LESS-NEXT: v_add_f32_e32 v0, v0, v4
+; GFX7LESS-NEXT: v_add_f32_e32 v0, v0, v2
; GFX7LESS-NEXT: v_cvt_f16_f32_e32 v0, v0
; GFX7LESS-NEXT: v_lshlrev_b32_e32 v0, s10, v0
-; GFX7LESS-NEXT: s_waitcnt expcnt(0)
-; GFX7LESS-NEXT: v_and_b32_e32 v2, s2, v1
-; GFX7LESS-NEXT: v_or_b32_e32 v0, v2, v0
-; GFX7LESS-NEXT: v_mov_b32_e32 v3, v1
-; GFX7LESS-NEXT: v_mov_b32_e32 v2, v0
-; GFX7LESS-NEXT: buffer_atomic_cmpswap v[2:3], off, s[4:7], 0 glc
+; GFX7LESS-NEXT: v_and_b32_e32 v1, s2, v4
+; GFX7LESS-NEXT: v_or_b32_e32 v3, v1, v0
+; GFX7LESS-NEXT: v_mov_b32_e32 v0, v3
+; GFX7LESS-NEXT: v_mov_b32_e32 v1, v4
+; GFX7LESS-NEXT: buffer_atomic_cmpswap v[0:1], off, s[4:7], 0 glc
; GFX7LESS-NEXT: s_waitcnt vmcnt(0)
-; GFX7LESS-NEXT: v_cmp_eq_u32_e32 vcc, v2, v1
+; GFX7LESS-NEXT: v_cmp_eq_u32_e32 vcc, v0, v4
; GFX7LESS-NEXT: s_or_b64 s[8:9], vcc, s[8:9]
-; GFX7LESS-NEXT: v_mov_b32_e32 v1, v2
; GFX7LESS-NEXT: s_andn2_b64 exec, exec, s[8:9]
; GFX7LESS-NEXT: s_cbranch_execnz .LBB18_1
; GFX7LESS-NEXT: ; %bb.2: ; %atomicrmw.end
; GFX7LESS-NEXT: s_or_b64 exec, exec, s[8:9]
; GFX7LESS-NEXT: s_mov_b32 s3, 0xf000
; GFX7LESS-NEXT: s_mov_b32 s2, -1
-; GFX7LESS-NEXT: v_lshrrev_b32_e32 v0, s10, v2
+; GFX7LESS-NEXT: s_waitcnt expcnt(0)
+; GFX7LESS-NEXT: v_lshrrev_b32_e32 v0, s10, v0
; GFX7LESS-NEXT: buffer_store_short v0, off, s[0:3], 0
; GFX7LESS-NEXT: s_endpgm
;
@@ -10572,28 +10597,28 @@ define amdgpu_kernel void @uniform_fadd_f16(ptr addrspace(1) %result, ptr addrsp
; GFX8-NEXT: s_lshl_b32 s2, 0xffff, s10
; GFX8-NEXT: s_not_b32 s2, s2
; GFX8-NEXT: s_waitcnt lgkmcnt(0)
-; GFX8-NEXT: v_mov_b32_e32 v1, s3
+; GFX8-NEXT: v_mov_b32_e32 v0, s3
; GFX8-NEXT: .LBB18_1: ; %atomicrmw.start
; GFX8-NEXT: ; =>This Inner Loop Header: Depth=1
-; GFX8-NEXT: v_lshrrev_b32_e32 v0, s10, v1
+; GFX8-NEXT: v_mov_b32_e32 v2, v0
+; GFX8-NEXT: v_lshrrev_b32_e32 v0, s10, v2
; GFX8-NEXT: v_add_f16_e32 v0, s11, v0
-; GFX8-NEXT: v_and_b32_e32 v2, s2, v1
+; GFX8-NEXT: v_and_b32_e32 v1, s2, v2
; GFX8-NEXT: v_lshlrev_b32_e32 v0, s10, v0
-; GFX8-NEXT: v_or_b32_e32 v0, v2, v0
-; GFX8-NEXT: v_mov_b32_e32 v3, v1
-; GFX8-NEXT: v_mov_b32_e32 v2, v0
-; GFX8-NEXT: buffer_atomic_cmpswap v[2:3], off, s[4:7], 0 glc
+; GFX8-NEXT: v_or_b32_e32 v1, v1, v0
+; GFX8-NEXT: v_mov_b32_e32 v0, v1
+; GFX8-NEXT: v_mov_b32_e32 v1, v2
+; GFX8-NEXT: buffer_atomic_cmpswap v[0:1], off, s[4:7], 0 glc
; GFX8-NEXT: s_waitcnt vmcnt(0)
-; GFX8-NEXT: v_cmp_eq_u32_e32 vcc, v2, v1
+; GFX8-NEXT: v_cmp_eq_u32_e32 vcc, v0, v2
; GFX8-NEXT: s_or_b64 s[8:9], vcc, s[8:9]
-; GFX8-NEXT: v_mov_b32_e32 v1, v2
; GFX8-NEXT: s_andn2_b64 exec, exec, s[8:9]
; GFX8-NEXT: s_cbranch_execnz .LBB18_1
; GFX8-NEXT: ; %bb.2: ; %atomicrmw.end
; GFX8-NEXT: s_or_b64 exec, exec, s[8:9]
; GFX8-NEXT: s_mov_b32 s3, 0xf000
; GFX8-NEXT: s_mov_b32 s2, -1
-; GFX8-NEXT: v_lshrrev_b32_e32 v0, s10, v2
+; GFX8-NEXT: v_lshrrev_b32_e32 v0, s10, v0
; GFX8-NEXT: buffer_store_short v0, off, s[0:3], 0
; GFX8-NEXT: s_endpgm
;
@@ -10613,27 +10638,27 @@ define amdgpu_kernel void @uniform_fadd_f16(ptr addrspace(1) %result, ptr addrsp
; GFX9-NEXT: s_lshl_b32 s2, 0xffff, s10
; GFX9-NEXT: s_not_b32 s2, s2
; GFX9-NEXT: s_waitcnt lgkmcnt(0)
-; GFX9-NEXT: v_mov_b32_e32 v1, s3
+; GFX9-NEXT: v_mov_b32_e32 v0, s3
; GFX9-NEXT: .LBB18_1: ; %atomicrmw.start
; GFX9-NEXT: ; =>This Inner Loop Header: Depth=1
-; GFX9-NEXT: v_lshrrev_b32_e32 v0, s10, v1
+; GFX9-NEXT: v_mov_b32_e32 v2, v0
+; GFX9-NEXT: v_lshrrev_b32_e32 v0, s10, v2
; GFX9-NEXT: v_add_f16_e32 v0, s11, v0
; GFX9-NEXT: v_lshlrev_b32_e32 v0, s10, v0
-; GFX9-NEXT: v_and_or_b32 v0, v1, s2, v0
-; GFX9-NEXT: v_mov_b32_e32 v3, v1
-; GFX9-NEXT: v_mov_b32_e32 v2, v0
-; GFX9-NEXT: buffer_atomic_cmpswap v[2:3], off, s[4:7], 0 glc
+; GFX9-NEXT: v_and_or_b32 v1, v2, s2, v0
+; GFX9-NEXT: v_mov_b32_e32 v0, v1
+; GFX9-NEXT: v_mov_b32_e32 v1, v2
+; GFX9-NEXT: buffer_atomic_cmpswap v[0:1], off, s[4:7], 0 glc
; GFX9-NEXT: s_waitcnt vmcnt(0)
-; GFX9-NEXT: v_cmp_eq_u32_e32 vcc, v2, v1
+; GFX9-NEXT: v_cmp_eq_u32_e32 vcc, v0, v2
; GFX9-NEXT: s_or_b64 s[8:9], vcc, s[8:9]
-; GFX9-NEXT: v_mov_b32_e32 v1, v2
; GFX9-NEXT: s_andn2_b64 exec, exec, s[8:9]
; GFX9-NEXT: s_cbranch_execnz .LBB18_1
; GFX9-NEXT: ; %bb.2: ; %atomicrmw.end
; GFX9-NEXT: s_or_b64 exec, exec, s[8:9]
; GFX9-NEXT: s_mov_b32 s3, 0xf000
; GFX9-NEXT: s_mov_b32 s2, -1
-; GFX9-NEXT: v_lshrrev_b32_e32 v0, s10, v2
+; GFX9-NEXT: v_lshrrev_b32_e32 v0, s10, v0
; GFX9-NEXT: buffer_store_short v0, off, s[0:3], 0
; GFX9-NEXT: s_endpgm
;
@@ -10653,26 +10678,26 @@ define amdgpu_kernel void @uniform_fadd_f16(ptr addrspace(1) %result, ptr addrsp
; GFX1064-NEXT: s_lshl_b32 s2, 0xffff, s9
; GFX1064-NEXT: s_not_b32 s10, s2
; GFX1064-NEXT: s_waitcnt lgkmcnt(0)
-; GFX1064-NEXT: v_mov_b32_e32 v1, s3
+; GFX1064-NEXT: v_mov_b32_e32 v0, s3
; GFX1064-NEXT: s_mov_b64 s[2:3], 0
; GFX1064-NEXT: .LBB18_1: ; %atomicrmw.start
; GFX1064-NEXT: ; =>This Inner Loop Header: Depth=1
-; GFX1064-NEXT: v_lshrrev_b32_e32 v0, s9, v1
+; GFX1064-NEXT: v_mov_b32_e32 v2, v0
+; GFX1064-NEXT: v_lshrrev_b32_e32 v0, s9, v2
; GFX1064-NEXT: v_add_f16_e32 v0, s8, v0
; GFX1064-NEXT: v_lshlrev_b32_sdwa v0, s9, v0 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:WORD_0
-; GFX1064-NEXT: v_and_or_b32 v0, v1, s10, v0
-; GFX1064-NEXT: v_mov_b32_e32 v3, v1
-; GFX1064-NEXT: v_mov_b32_e32 v2, v0
-; GFX1064-NEXT: buffer_atomic_cmpswap v[2:3], off, s[4:7], 0 glc
-; GFX1064-NEXT: s_waitcnt vmcnt(0)
-; GFX1064-NEXT: v_cmp_eq_u32_e32 vcc, v2, v1
+; GFX1064-NEXT: v_and_or_b32 v1, v2, s10, v0
+; GFX1064-NEXT: v_mov_b32_e32 v0, v1
; GFX1064-NEXT: v_mov_b32_e32 v1, v2
+; GFX1064-NEXT: buffer_atomic_cmpswap v[0:1], off, s[4:7], 0 glc
+; GFX1064-NEXT: s_waitcnt vmcnt(0)
+; GFX1064-NEXT: v_cmp_eq_u32_e32 vcc, v0, v2
; GFX1064-NEXT: s_or_b64 s[2:3], vcc, s[2:3]
; GFX1064-NEXT: s_andn2_b64 exec, exec, s[2:3]
; GFX1064-NEXT: s_cbranch_execnz .LBB18_1
; GFX1064-NEXT: ; %bb.2: ; %atomicrmw.end
; GFX1064-NEXT: s_or_b64 exec, exec, s[2:3]
-; GFX1064-NEXT: v_lshrrev_b32_e32 v0, s9, v2
+; GFX1064-NEXT: v_lshrrev_b32_e32 v0, s9, v0
; GFX1064-NEXT: s_mov_b32 s3, 0x31016000
; GFX1064-NEXT: s_mov_b32 s2, -1
; GFX1064-NEXT: buffer_store_short v0, off, s[0:3], 0
@@ -10694,26 +10719,26 @@ define amdgpu_kernel void @uniform_fadd_f16(ptr addrspace(1) %result, ptr addrsp
; GFX1032-NEXT: s_lshl_b32 s3, 0xffff, s2
; GFX1032-NEXT: s_not_b32 s3, s3
; GFX1032-NEXT: s_waitcnt lgkmcnt(0)
-; GFX1032-NEXT: v_mov_b32_e32 v1, s6
+; GFX1032-NEXT: v_mov_b32_e32 v0, s6
; GFX1032-NEXT: s_mov_b32 s6, -1
; GFX1032-NEXT: .LBB18_1: ; %atomicrmw.start
; GFX1032-NEXT: ; =>This Inner Loop Header: Depth=1
-; GFX1032-NEXT: v_lshrrev_b32_e32 v0, s2, v1
+; GFX1032-NEXT: v_mov_b32_e32 v2, v0
+; GFX1032-NEXT: v_lshrrev_b32_e32 v0, s2, v2
; GFX1032-NEXT: v_add_f16_e32 v0, s8, v0
; GFX1032-NEXT: v_lshlrev_b32_sdwa v0, s2, v0 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:WORD_0
-; GFX1032-NEXT: v_and_or_b32 v0, v1, s3, v0
-; GFX1032-NEXT: v_mov_b32_e32 v3, v1
-; GFX1032-NEXT: v_mov_b32_e32 v2, v0
-; GFX1032-NEXT: buffer_atomic_cmpswap v[2:3], off, s[4:7], 0 glc
-; GFX1032-NEXT: s_waitcnt vmcnt(0)
-; GFX1032-NEXT: v_cmp_eq_u32_e32 vcc_lo, v2, v1
+; GFX1032-NEXT: v_and_or_b32 v1, v2, s3, v0
+; GFX1032-NEXT: v_mov_b32_e32 v0, v1
; GFX1032-NEXT: v_mov_b32_e32 v1, v2
+; GFX1032-NEXT: buffer_atomic_cmpswap v[0:1], off, s[4:7], 0 glc
+; GFX1032-NEXT: s_waitcnt vmcnt(0)
+; GFX1032-NEXT: v_cmp_eq_u32_e32 vcc_lo, v0, v2
; GFX1032-NEXT: s_or_b32 s9, vcc_lo, s9
; GFX1032-NEXT: s_andn2_b32 exec_lo, exec_lo, s9
; GFX1032-NEXT: s_cbranch_execnz .LBB18_1
; GFX1032-NEXT: ; %bb.2: ; %atomicrmw.end
; GFX1032-NEXT: s_or_b32 exec_lo, exec_lo, s9
-; GFX1032-NEXT: v_lshrrev_b32_e32 v0, s2, v2
+; GFX1032-NEXT: v_lshrrev_b32_e32 v0, s2, v0
; GFX1032-NEXT: s_mov_b32 s3, 0x31016000
; GFX1032-NEXT: s_mov_b32 s2, -1
; GFX1032-NEXT: buffer_store_short v0, off, s[0:3], 0
@@ -10736,31 +10761,32 @@ define amdgpu_kernel void @uniform_fadd_f16(ptr addrspace(1) %result, ptr addrsp
; GFX1164-TRUE16-NEXT: s_lshl_b32 s2, 0xffff, s9
; GFX1164-TRUE16-NEXT: s_not_b32 s10, s2
; GFX1164-TRUE16-NEXT: s_waitcnt lgkmcnt(0)
-; GFX1164-TRUE16-NEXT: v_mov_b32_e32 v1, s3
+; GFX1164-TRUE16-NEXT: v_mov_b32_e32 v0, s3
; GFX1164-TRUE16-NEXT: s_mov_b64 s[2:3], 0
; GFX1164-TRUE16-NEXT: .LBB18_1: ; %atomicrmw.start
; GFX1164-TRUE16-NEXT: ; =>This Inner Loop Header: Depth=1
; GFX1164-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
-; GFX1164-TRUE16-NEXT: v_lshrrev_b32_e32 v0, s9, v1
-; GFX1164-TRUE16-NEXT: v_add_f16_e32 v0.l, s8, v0.l
+; GFX1164-TRUE16-NEXT: v_mov_b32_e32 v2, v0
+; GFX1164-TRUE16-NEXT: v_lshrrev_b32_e32 v0, s9, v2
; GFX1164-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
+; GFX1164-TRUE16-NEXT: v_add_f16_e32 v0.l, s8, v0.l
; GFX1164-TRUE16-NEXT: v_and_b32_e32 v0, 0xffff, v0
+; GFX1164-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
; GFX1164-TRUE16-NEXT: v_lshlrev_b32_e32 v0, s9, v0
-; GFX1164-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(SKIP_1) | instid1(VALU_DEP_2)
-; GFX1164-TRUE16-NEXT: v_and_or_b32 v0, v1, s10, v0
-; GFX1164-TRUE16-NEXT: v_mov_b32_e32 v3, v1
-; GFX1164-TRUE16-NEXT: v_mov_b32_e32 v2, v0
-; GFX1164-TRUE16-NEXT: buffer_atomic_cmpswap_b32 v[2:3], off, s[4:7], 0 glc
-; GFX1164-TRUE16-NEXT: s_waitcnt vmcnt(0)
-; GFX1164-TRUE16-NEXT: v_cmp_eq_u32_e32 vcc, v2, v1
+; GFX1164-TRUE16-NEXT: v_and_or_b32 v1, v2, s10, v0
+; GFX1164-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1)
+; GFX1164-TRUE16-NEXT: v_mov_b32_e32 v0, v1
; GFX1164-TRUE16-NEXT: v_mov_b32_e32 v1, v2
+; GFX1164-TRUE16-NEXT: buffer_atomic_cmpswap_b32 v[0:1], off, s[4:7], 0 glc
+; GFX1164-TRUE16-NEXT: s_waitcnt vmcnt(0)
+; GFX1164-TRUE16-NEXT: v_cmp_eq_u32_e32 vcc, v0, v2
; GFX1164-TRUE16-NEXT: s_or_b64 s[2:3], vcc, s[2:3]
; GFX1164-TRUE16-NEXT: s_delay_alu instid0(SALU_CYCLE_1)
; GFX1164-TRUE16-NEXT: s_and_not1_b64 exec, exec, s[2:3]
; GFX1164-TRUE16-NEXT: s_cbranch_execnz .LBB18_1
; GFX1164-TRUE16-NEXT: ; %bb.2: ; %atomicrmw.end
; GFX1164-TRUE16-NEXT: s_or_b64 exec, exec, s[2:3]
-; GFX1164-TRUE16-NEXT: v_lshrrev_b32_e32 v0, s9, v2
+; GFX1164-TRUE16-NEXT: v_lshrrev_b32_e32 v0, s9, v0
; GFX1164-TRUE16-NEXT: s_mov_b32 s3, 0x31016000
; GFX1164-TRUE16-NEXT: s_mov_b32 s2, -1
; GFX1164-TRUE16-NEXT: buffer_store_b16 v0, off, s[0:3], 0
@@ -10783,31 +10809,32 @@ define amdgpu_kernel void @uniform_fadd_f16(ptr addrspace(1) %result, ptr addrsp
; GFX1164-FAKE16-NEXT: s_lshl_b32 s2, 0xffff, s9
; GFX1164-FAKE16-NEXT: s_not_b32 s10, s2
; GFX1164-FAKE16-NEXT: s_waitcnt lgkmcnt(0)
-; GFX1164-FAKE16-NEXT: v_mov_b32_e32 v1, s3
+; GFX1164-FAKE16-NEXT: v_mov_b32_e32 v0, s3
; GFX1164-FAKE16-NEXT: s_mov_b64 s[2:3], 0
; GFX1164-FAKE16-NEXT: .LBB18_1: ; %atomicrmw.start
; GFX1164-FAKE16-NEXT: ; =>This Inner Loop Header: Depth=1
; GFX1164-FAKE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
-; GFX1164-FAKE16-NEXT: v_lshrrev_b32_e32 v0, s9, v1
-; GFX1164-FAKE16-NEXT: v_add_f16_e32 v0, s8, v0
+; GFX1164-FAKE16-NEXT: v_mov_b32_e32 v2, v0
+; GFX1164-FAKE16-NEXT: v_lshrrev_b32_e32 v0, s9, v2
; GFX1164-FAKE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
+; GFX1164-FAKE16-NEXT: v_add_f16_e32 v0, s8, v0
; GFX1164-FAKE16-NEXT: v_and_b32_e32 v0, 0xffff, v0
+; GFX1164-FAKE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
; GFX1164-FAKE16-NEXT: v_lshlrev_b32_e32 v0, s9, v0
-; GFX1164-FAKE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(SKIP_1) | instid1(VALU_DEP_2)
-; GFX1164-FAKE16-NEXT: v_and_or_b32 v0, v1, s10, v0
-; GFX1164-FAKE16-NEXT: v_mov_b32_e32 v3, v1
-; GFX1164-FAKE16-NEXT: v_mov_b32_e32 v2, v0
-; GFX1164-FAKE16-NEXT: buffer_atomic_cmpswap_b32 v[2:3], off, s[4:7], 0 glc
-; GFX1164-FAKE16-NEXT: s_waitcnt vmcnt(0)
-; GFX1164-FAKE16-NEXT: v_cmp_eq_u32_e32 vcc, v2, v1
+; GFX1164-FAKE16-NEXT: v_and_or_b32 v1, v2, s10, v0
+; GFX1164-FAKE16-NEXT: s_delay_alu instid0(VALU_DEP_1)
+; GFX1164-FAKE16-NEXT: v_mov_b32_e32 v0, v1
; GFX1164-FAKE16-NEXT: v_mov_b32_e32 v1, v2
+; GFX1164-FAKE16-NEXT: buffer_atomic_cmpswap_b32 v[0:1], off, s[4:7], 0 glc
+; GFX1164-FAKE16-NEXT: s_waitcnt vmcnt(0)
+; GFX1164-FAKE16-NEXT: v_cmp_eq_u32_e32 vcc, v0, v2
; GFX1164-FAKE16-NEXT: s_or_b64 s[2:3], vcc, s[2:3]
; GFX1164-FAKE16-NEXT: s_delay_alu instid0(SALU_CYCLE_1)
; GFX1164-FAKE16-NEXT: s_and_not1_b64 exec, exec, s[2:3]
; GFX1164-FAKE16-NEXT: s_cbranch_execnz .LBB18_1
; GFX1164-FAKE16-NEXT: ; %bb.2: ; %atomicrmw.end
; GFX1164-FAKE16-NEXT: s_or_b64 exec, exec, s[2:3]
-; GFX1164-FAKE16-NEXT: v_lshrrev_b32_e32 v0, s9, v2
+; GFX1164-FAKE16-NEXT: v_lshrrev_b32_e32 v0, s9, v0
; GFX1164-FAKE16-NEXT: s_mov_b32 s3, 0x31016000
; GFX1164-FAKE16-NEXT: s_mov_b32 s2, -1
; GFX1164-FAKE16-NEXT: buffer_store_b16 v0, off, s[0:3], 0
@@ -10830,30 +10857,32 @@ define amdgpu_kernel void @uniform_fadd_f16(ptr addrspace(1) %result, ptr addrsp
; GFX1132-TRUE16-NEXT: s_lshl_b32 s3, 0xffff, s2
; GFX1132-TRUE16-NEXT: s_not_b32 s3, s3
; GFX1132-TRUE16-NEXT: s_waitcnt lgkmcnt(0)
-; GFX1132-TRUE16-NEXT: v_mov_b32_e32 v1, s6
+; GFX1132-TRUE16-NEXT: v_mov_b32_e32 v0, s6
; GFX1132-TRUE16-NEXT: s_mov_b32 s6, -1
; GFX1132-TRUE16-NEXT: .LBB18_1: ; %atomicrmw.start
; GFX1132-TRUE16-NEXT: ; =>This Inner Loop Header: Depth=1
; GFX1132-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
-; GFX1132-TRUE16-NEXT: v_lshrrev_b32_e32 v0, s2, v1
-; GFX1132-TRUE16-NEXT: v_add_f16_e32 v0.l, s8, v0.l
+; GFX1132-TRUE16-NEXT: v_mov_b32_e32 v2, v0
+; GFX1132-TRUE16-NEXT: v_lshrrev_b32_e32 v0, s2, v2
; GFX1132-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
+; GFX1132-TRUE16-NEXT: v_add_f16_e32 v0.l, s8, v0.l
; GFX1132-TRUE16-NEXT: v_and_b32_e32 v0, 0xffff, v0
-; GFX1132-TRUE16-NEXT: v_lshlrev_b32_e32 v0, s2, v0
; GFX1132-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
-; GFX1132-TRUE16-NEXT: v_and_or_b32 v0, v1, s3, v0
-; GFX1132-TRUE16-NEXT: v_dual_mov_b32 v3, v1 :: v_dual_mov_b32 v2, v0
-; GFX1132-TRUE16-NEXT: buffer_atomic_cmpswap_b32 v[2:3], off, s[4:7], 0 glc
-; GFX1132-TRUE16-NEXT: s_waitcnt vmcnt(0)
-; GFX1132-TRUE16-NEXT: v_cmp_eq_u32_e32 vcc_lo, v2, v1
+; GFX1132-TRUE16-NEXT: v_lshlrev_b32_e32 v0, s2, v0
+; GFX1132-TRUE16-NEXT: v_and_or_b32 v1, v2, s3, v0
+; GFX1132-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1)
+; GFX1132-TRUE16-NEXT: v_mov_b32_e32 v0, v1
; GFX1132-TRUE16-NEXT: v_mov_b32_e32 v1, v2
+; GFX1132-TRUE16-NEXT: buffer_atomic_cmpswap_b32 v[0:1], off, s[4:7], 0 glc
+; GFX1132-TRUE16-NEXT: s_waitcnt vmcnt(0)
+; GFX1132-TRUE16-NEXT: v_cmp_eq_u32_e32 vcc_lo, v0, v2
; GFX1132-TRUE16-NEXT: s_or_b32 s9, vcc_lo, s9
; GFX1132-TRUE16-NEXT: s_delay_alu instid0(SALU_CYCLE_1)
; GFX1132-TRUE16-NEXT: s_and_not1_b32 exec_lo, exec_lo, s9
; GFX1132-TRUE16-NEXT: s_cbranch_execnz .LBB18_1
; GFX1132-TRUE16-NEXT: ; %bb.2: ; %atomicrmw.end
; GFX1132-TRUE16-NEXT: s_or_b32 exec_lo, exec_lo, s9
-; GFX1132-TRUE16-NEXT: v_lshrrev_b32_e32 v0, s2, v2
+; GFX1132-TRUE16-NEXT: v_lshrrev_b32_e32 v0, s2, v0
; GFX1132-TRUE16-NEXT: s_mov_b32 s3, 0x31016000
; GFX1132-TRUE16-NEXT: s_mov_b32 s2, -1
; GFX1132-TRUE16-NEXT: buffer_store_b16 v0, off, s[0:3], 0
@@ -10876,30 +10905,32 @@ define amdgpu_kernel void @uniform_fadd_f16(ptr addrspace(1) %result, ptr addrsp
; GFX1132-FAKE16-NEXT: s_lshl_b32 s3, 0xffff, s2
; GFX1132-FAKE16-NEXT: s_not_b32 s3, s3
; GFX1132-FAKE16-NEXT: s_waitcnt lgkmcnt(0)
-; GFX1132-FAKE16-NEXT: v_mov_b32_e32 v1, s6
+; GFX1132-FAKE16-NEXT: v_mov_b32_e32 v0, s6
; GFX1132-FAKE16-NEXT: s_mov_b32 s6, -1
; GFX1132-FAKE16-NEXT: .LBB18_1: ; %atomicrmw.start
; GFX1132-FAKE16-NEXT: ; =>This Inner Loop Header: Depth=1
; GFX1132-FAKE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
-; GFX1132-FAKE16-NEXT: v_lshrrev_b32_e32 v0, s2, v1
-; GFX1132-FAKE16-NEXT: v_add_f16_e32 v0, s8, v0
+; GFX1132-FAKE16-NEXT: v_mov_b32_e32 v2, v0
+; GFX1132-FAKE16-NEXT: v_lshrrev_b32_e32 v0, s2, v2
; GFX1132-FAKE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
+; GFX1132-FAKE16-NEXT: v_add_f16_e32 v0, s8, v0
; GFX1132-FAKE16-NEXT: v_and_b32_e32 v0, 0xffff, v0
-; GFX1132-FAKE16-NEXT: v_lshlrev_b32_e32 v0, s2, v0
; GFX1132-FAKE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
-; GFX1132-FAKE16-NEXT: v_and_or_b32 v0, v1, s3, v0
-; GFX1132-FAKE16-NEXT: v_dual_mov_b32 v3, v1 :: v_dual_mov_b32 v2, v0
-; GFX1132-FAKE16-NEXT: buffer_atomic_cmpswap_b32 v[2:3], off, s[4:7], 0 glc
-; GFX1132-FAKE16-NEXT: s_waitcnt vmcnt(0)
-; GFX1132-FAKE16-NEXT: v_cmp_eq_u32_e32 vcc_lo, v2, v1
+; GFX1132-FAKE16-NEXT: v_lshlrev_b32_e32 v0, s2, v0
+; GFX1132-FAKE16-NEXT: v_and_or_b32 v1, v2, s3, v0
+; GFX1132-FAKE16-NEXT: s_delay_alu instid0(VALU_DEP_1)
+; GFX1132-FAKE16-NEXT: v_mov_b32_e32 v0, v1
; GFX1132-FAKE16-NEXT: v_mov_b32_e32 v1, v2
+; GFX1132-FAKE16-NEXT: buffer_atomic_cmpswap_b32 v[0:1], off, s[4:7], 0 glc
+; GFX1132-FAKE16-NEXT: s_waitcnt vmcnt(0)
+; GFX1132-FAKE16-NEXT: v_cmp_eq_u32_e32 vcc_lo, v0, v2
; GFX1132-FAKE16-NEXT: s_or_b32 s9, vcc_lo, s9
; GFX1132-FAKE16-NEXT: s_delay_alu instid0(SALU_CYCLE_1)
; GFX1132-FAKE16-NEXT: s_and_not1_b32 exec_lo, exec_lo, s9
; GFX1132-FAKE16-NEXT: s_cbranch_execnz .LBB18_1
; GFX1132-FAKE16-NEXT: ; %bb.2: ; %atomicrmw.end
; GFX1132-FAKE16-NEXT: s_or_b32 exec_lo, exec_lo, s9
-; GFX1132-FAKE16-NEXT: v_lshrrev_b32_e32 v0, s2, v2
+; GFX1132-FAKE16-NEXT: v_lshrrev_b32_e32 v0, s2, v0
; GFX1132-FAKE16-NEXT: s_mov_b32 s3, 0x31016000
; GFX1132-FAKE16-NEXT: s_mov_b32 s2, -1
; GFX1132-FAKE16-NEXT: buffer_store_b16 v0, off, s[0:3], 0
@@ -10922,31 +10953,32 @@ define amdgpu_kernel void @uniform_fadd_f16(ptr addrspace(1) %result, ptr addrsp
; GFX1264-TRUE16-NEXT: s_lshl_b32 s2, 0xffff, s9
; GFX1264-TRUE16-NEXT: s_not_b32 s10, s2
; GFX1264-TRUE16-NEXT: s_wait_kmcnt 0x0
-; GFX1264-TRUE16-NEXT: v_mov_b32_e32 v1, s3
+; GFX1264-TRUE16-NEXT: v_mov_b32_e32 v0, s3
; GFX1264-TRUE16-NEXT: s_mov_b64 s[2:3], 0
; GFX1264-TRUE16-NEXT: .LBB18_1: ; %atomicrmw.start
; GFX1264-TRUE16-NEXT: ; =>This Inner Loop Header: Depth=1
; GFX1264-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
-; GFX1264-TRUE16-NEXT: v_lshrrev_b32_e32 v0, s9, v1
-; GFX1264-TRUE16-NEXT: v_add_f16_e32 v0.l, s8, v0.l
+; GFX1264-TRUE16-NEXT: v_mov_b32_e32 v2, v0
+; GFX1264-TRUE16-NEXT: v_lshrrev_b32_e32 v0, s9, v2
; GFX1264-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
+; GFX1264-TRUE16-NEXT: v_add_f16_e32 v0.l, s8, v0.l
; GFX1264-TRUE16-NEXT: v_and_b32_e32 v0, 0xffff, v0
+; GFX1264-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
; GFX1264-TRUE16-NEXT: v_lshlrev_b32_e32 v0, s9, v0
-; GFX1264-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(SKIP_1) | instid1(VALU_DEP_2)
-; GFX1264-TRUE16-NEXT: v_and_or_b32 v0, v1, s10, v0
-; GFX1264-TRUE16-NEXT: v_mov_b32_e32 v3, v1
-; GFX1264-TRUE16-NEXT: v_mov_b32_e32 v2, v0
-; GFX1264-TRUE16-NEXT: buffer_atomic_cmpswap_b32 v[2:3], off, s[4:7], null th:TH_ATOMIC_RETURN scope:SCOPE_SYS
-; GFX1264-TRUE16-NEXT: s_wait_loadcnt 0x0
-; GFX1264-TRUE16-NEXT: v_cmp_eq_u32_e32 vcc, v2, v1
+; GFX1264-TRUE16-NEXT: v_and_or_b32 v1, v2, s10, v0
+; GFX1264-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1)
+; GFX1264-TRUE16-NEXT: v_mov_b32_e32 v0, v1
; GFX1264-TRUE16-NEXT: v_mov_b32_e32 v1, v2
+; GFX1264-TRUE16-NEXT: buffer_atomic_cmpswap_b32 v[0:1], off, s[4:7], null th:TH_ATOMIC_RETURN scope:SCOPE_SYS
+; GFX1264-TRUE16-NEXT: s_wait_loadcnt 0x0
+; GFX1264-TRUE16-NEXT: v_cmp_eq_u32_e32 vcc, v0, v2
; GFX1264-TRUE16-NEXT: s_or_b64 s[2:3], vcc, s[2:3]
; GFX1264-TRUE16-NEXT: s_wait_alu 0xfffe
; GFX1264-TRUE16-NEXT: s_and_not1_b64 exec, exec, s[2:3]
; GFX1264-TRUE16-NEXT: s_cbranch_execnz .LBB18_1
; GFX1264-TRUE16-NEXT: ; %bb.2: ; %atomicrmw.end
; GFX1264-TRUE16-NEXT: s_or_b64 exec, exec, s[2:3]
-; GFX1264-TRUE16-NEXT: v_lshrrev_b32_e32 v0, s9, v2
+; GFX1264-TRUE16-NEXT: v_lshrrev_b32_e32 v0, s9, v0
; GFX1264-TRUE16-NEXT: s_mov_b32 s3, 0x31016000
; GFX1264-TRUE16-NEXT: s_mov_b32 s2, -1
; GFX1264-TRUE16-NEXT: buffer_store_b16 v0, off, s[0:3], null
@@ -10969,31 +11001,32 @@ define amdgpu_kernel void @uniform_fadd_f16(ptr addrspace(1) %result, ptr addrsp
; GFX1264-FAKE16-NEXT: s_lshl_b32 s2, 0xffff, s9
; GFX1264-FAKE16-NEXT: s_not_b32 s10, s2
; GFX1264-FAKE16-NEXT: s_wait_kmcnt 0x0
-; GFX1264-FAKE16-NEXT: v_mov_b32_e32 v1, s3
+; GFX1264-FAKE16-NEXT: v_mov_b32_e32 v0, s3
; GFX1264-FAKE16-NEXT: s_mov_b64 s[2:3], 0
; GFX1264-FAKE16-NEXT: .LBB18_1: ; %atomicrmw.start
; GFX1264-FAKE16-NEXT: ; =>This Inner Loop Header: Depth=1
; GFX1264-FAKE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
-; GFX1264-FAKE16-NEXT: v_lshrrev_b32_e32 v0, s9, v1
-; GFX1264-FAKE16-NEXT: v_add_f16_e32 v0, s8, v0
+; GFX1264-FAKE16-NEXT: v_mov_b32_e32 v2, v0
+; GFX1264-FAKE16-NEXT: v_lshrrev_b32_e32 v0, s9, v2
; GFX1264-FAKE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
+; GFX1264-FAKE16-NEXT: v_add_f16_e32 v0, s8, v0
; GFX1264-FAKE16-NEXT: v_and_b32_e32 v0, 0xffff, v0
+; GFX1264-FAKE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
; GFX1264-FAKE16-NEXT: v_lshlrev_b32_e32 v0, s9, v0
-; GFX1264-FAKE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(SKIP_1) | instid1(VALU_DEP_2)
-; GFX1264-FAKE16-NEXT: v_and_or_b32 v0, v1, s10, v0
-; GFX1264-FAKE16-NEXT: v_mov_b32_e32 v3, v1
-; GFX1264-FAKE16-NEXT: v_mov_b32_e32 v2, v0
-; GFX1264-FAKE16-NEXT: buffer_atomic_cmpswap_b32 v[2:3], off, s[4:7], null th:TH_ATOMIC_RETURN scope:SCOPE_SYS
-; GFX1264-FAKE16-NEXT: s_wait_loadcnt 0x0
-; GFX1264-FAKE16-NEXT: v_cmp_eq_u32_e32 vcc, v2, v1
+; GFX1264-FAKE16-NEXT: v_and_or_b32 v1, v2, s10, v0
+; GFX1264-FAKE16-NEXT: s_delay_alu instid0(VALU_DEP_1)
+; GFX1264-FAKE16-NEXT: v_mov_b32_e32 v0, v1
; GFX1264-FAKE16-NEXT: v_mov_b32_e32 v1, v2
+; GFX1264-FAKE16-NEXT: buffer_atomic_cmpswap_b32 v[0:1], off, s[4:7], null th:TH_ATOMIC_RETURN scope:SCOPE_SYS
+; GFX1264-FAKE16-NEXT: s_wait_loadcnt 0x0
+; GFX1264-FAKE16-NEXT: v_cmp_eq_u32_e32 vcc, v0, v2
; GFX1264-FAKE16-NEXT: s_or_b64 s[2:3], vcc, s[2:3]
; GFX1264-FAKE16-NEXT: s_wait_alu 0xfffe
; GFX1264-FAKE16-NEXT: s_and_not1_b64 exec, exec, s[2:3]
; GFX1264-FAKE16-NEXT: s_cbranch_execnz .LBB18_1
; GFX1264-FAKE16-NEXT: ; %bb.2: ; %atomicrmw.end
; GFX1264-FAKE16-NEXT: s_or_b64 exec, exec, s[2:3]
-; GFX1264-FAKE16-NEXT: v_lshrrev_b32_e32 v0, s9, v2
+; GFX1264-FAKE16-NEXT: v_lshrrev_b32_e32 v0, s9, v0
; GFX1264-FAKE16-NEXT: s_mov_b32 s3, 0x31016000
; GFX1264-FAKE16-NEXT: s_mov_b32 s2, -1
; GFX1264-FAKE16-NEXT: buffer_store_b16 v0, off, s[0:3], null
@@ -11016,30 +11049,32 @@ define amdgpu_kernel void @uniform_fadd_f16(ptr addrspace(1) %result, ptr addrsp
; GFX1232-TRUE16-NEXT: s_lshl_b32 s3, 0xffff, s2
; GFX1232-TRUE16-NEXT: s_not_b32 s3, s3
; GFX1232-TRUE16-NEXT: s_wait_kmcnt 0x0
-; GFX1232-TRUE16-NEXT: v_mov_b32_e32 v1, s6
+; GFX1232-TRUE16-NEXT: v_mov_b32_e32 v0, s6
; GFX1232-TRUE16-NEXT: s_mov_b32 s6, -1
; GFX1232-TRUE16-NEXT: .LBB18_1: ; %atomicrmw.start
; GFX1232-TRUE16-NEXT: ; =>This Inner Loop Header: Depth=1
; GFX1232-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
-; GFX1232-TRUE16-NEXT: v_lshrrev_b32_e32 v0, s2, v1
-; GFX1232-TRUE16-NEXT: v_add_f16_e32 v0.l, s8, v0.l
+; GFX1232-TRUE16-NEXT: v_mov_b32_e32 v2, v0
+; GFX1232-TRUE16-NEXT: v_lshrrev_b32_e32 v0, s2, v2
; GFX1232-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
+; GFX1232-TRUE16-NEXT: v_add_f16_e32 v0.l, s8, v0.l
; GFX1232-TRUE16-NEXT: v_and_b32_e32 v0, 0xffff, v0
-; GFX1232-TRUE16-NEXT: v_lshlrev_b32_e32 v0, s2, v0
; GFX1232-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
-; GFX1232-TRUE16-NEXT: v_and_or_b32 v0, v1, s3, v0
-; GFX1232-TRUE16-NEXT: v_dual_mov_b32 v3, v1 :: v_dual_mov_b32 v2, v0
-; GFX1232-TRUE16-NEXT: buffer_atomic_cmpswap_b32 v[2:3], off, s[4:7], null th:TH_ATOMIC_RETURN scope:SCOPE_SYS
-; GFX1232-TRUE16-NEXT: s_wait_loadcnt 0x0
-; GFX1232-TRUE16-NEXT: v_cmp_eq_u32_e32 vcc_lo, v2, v1
+; GFX1232-TRUE16-NEXT: v_lshlrev_b32_e32 v0, s2, v0
+; GFX1232-TRUE16-NEXT: v_and_or_b32 v1, v2, s3, v0
+; GFX1232-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1)
+; GFX1232-TRUE16-NEXT: v_mov_b32_e32 v0, v1
; GFX1232-TRUE16-NEXT: v_mov_b32_e32 v1, v2
+; GFX1232-TRUE16-NEXT: buffer_atomic_cmpswap_b32 v[0:1], off, s[4:7], null th:TH_ATOMIC_RETURN scope:SCOPE_SYS
+; GFX1232-TRUE16-NEXT: s_wait_loadcnt 0x0
+; GFX1232-TRUE16-NEXT: v_cmp_eq_u32_e32 vcc_lo, v0, v2
; GFX1232-TRUE16-NEXT: s_or_b32 s9, vcc_lo, s9
; GFX1232-TRUE16-NEXT: s_wait_alu 0xfffe
; GFX1232-TRUE16-NEXT: s_and_not1_b32 exec_lo, exec_lo, s9
; GFX1232-TRUE16-NEXT: s_cbranch_execnz .LBB18_1
; GFX1232-TRUE16-NEXT: ; %bb.2: ; %atomicrmw.end
; GFX1232-TRUE16-NEXT: s_or_b32 exec_lo, exec_lo, s9
-; GFX1232-TRUE16-NEXT: v_lshrrev_b32_e32 v0, s2, v2
+; GFX1232-TRUE16-NEXT: v_lshrrev_b32_e32 v0, s2, v0
; GFX1232-TRUE16-NEXT: s_mov_b32 s3, 0x31016000
; GFX1232-TRUE16-NEXT: s_mov_b32 s2, -1
; GFX1232-TRUE16-NEXT: buffer_store_b16 v0, off, s[0:3], null
@@ -11062,30 +11097,32 @@ define amdgpu_kernel void @uniform_fadd_f16(ptr addrspace(1) %result, ptr addrsp
; GFX1232-FAKE16-NEXT: s_lshl_b32 s3, 0xffff, s2
; GFX1232-FAKE16-NEXT: s_not_b32 s3, s3
; GFX1232-FAKE16-NEXT: s_wait_kmcnt 0x0
-; GFX1232-FAKE16-NEXT: v_mov_b32_e32 v1, s6
+; GFX1232-FAKE16-NEXT: v_mov_b32_e32 v0, s6
; GFX1232-FAKE16-NEXT: s_mov_b32 s6, -1
; GFX1232-FAKE16-NEXT: .LBB18_1: ; %atomicrmw.start
; GFX1232-FAKE16-NEXT: ; =>This Inner Loop Header: Depth=1
; GFX1232-FAKE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
-; GFX1232-FAKE16-NEXT: v_lshrrev_b32_e32 v0, s2, v1
-; GFX1232-FAKE16-NEXT: v_add_f16_e32 v0, s8, v0
+; GFX1232-FAKE16-NEXT: v_mov_b32_e32 v2, v0
+; GFX1232-FAKE16-NEXT: v_lshrrev_b32_e32 v0, s2, v2
; GFX1232-FAKE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
+; GFX1232-FAKE16-NEXT: v_add_f16_e32 v0, s8, v0
; GFX1232-FAKE16-NEXT: v_and_b32_e32 v0, 0xffff, v0
-; GFX1232-FAKE16-NEXT: v_lshlrev_b32_e32 v0, s2, v0
; GFX1232-FAKE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
-; GFX1232-FAKE16-NEXT: v_and_or_b32 v0, v1, s3, v0
-; GFX1232-FAKE16-NEXT: v_dual_mov_b32 v3, v1 :: v_dual_mov_b32 v2, v0
-; GFX1232-FAKE16-NEXT: buffer_atomic_cmpswap_b32 v[2:3], off, s[4:7], null th:TH_ATOMIC_RETURN scope:SCOPE_SYS
-; GFX1232-FAKE16-NEXT: s_wait_loadcnt 0x0
-; GFX1232-FAKE16-NEXT: v_cmp_eq_u32_e32 vcc_lo, v2, v1
+; GFX1232-FAKE16-NEXT: v_lshlrev_b32_e32 v0, s2, v0
+; GFX1232-FAKE16-NEXT: v_and_or_b32 v1, v2, s3, v0
+; GFX1232-FAKE16-NEXT: s_delay_alu instid0(VALU_DEP_1)
+; GFX1232-FAKE16-NEXT: v_mov_b32_e32 v0, v1
; GFX1232-FAKE16-NEXT: v_mov_b32_e32 v1, v2
+; GFX1232-FAKE16-NEXT: buffer_atomic_cmpswap_b32 v[0:1], off, s[4:7], null th:TH_ATOMIC_RETURN scope:SCOPE_SYS
+; GFX1232-FAKE16-NEXT: s_wait_loadcnt 0x0
+; GFX1232-FAKE16-NEXT: v_cmp_eq_u32_e32 vcc_lo, v0, v2
; GFX1232-FAKE16-NEXT: s_or_b32 s9, vcc_lo, s9
; GFX1232-FAKE16-NEXT: s_wait_alu 0xfffe
; GFX1232-FAKE16-NEXT: s_and_not1_b32 exec_lo, exec_lo, s9
; GFX1232-FAKE16-NEXT: s_cbranch_execnz .LBB18_1
; GFX1232-FAKE16-NEXT: ; %bb.2: ; %atomicrmw.end
; GFX1232-FAKE16-NEXT: s_or_b32 exec_lo, exec_lo, s9
-; GFX1232-FAKE16-NEXT: v_lshrrev_b32_e32 v0, s2, v2
+; GFX1232-FAKE16-NEXT: v_lshrrev_b32_e32 v0, s2, v0
; GFX1232-FAKE16-NEXT: s_mov_b32 s3, 0x31016000
; GFX1232-FAKE16-NEXT: s_mov_b32 s2, -1
; GFX1232-FAKE16-NEXT: buffer_store_b16 v0, off, s[0:3], null
@@ -11111,35 +11148,36 @@ define amdgpu_kernel void @uniform_fadd_bf16(ptr addrspace(1) %result, ptr addrs
; GFX7LESS-NEXT: s_load_dword s3, s[4:5], 0x0
; GFX7LESS-NEXT: v_mul_f32_e64 v0, 1.0, s6
; GFX7LESS-NEXT: s_lshl_b32 s2, 0xffff, s10
-; GFX7LESS-NEXT: v_and_b32_e32 v4, 0xffff0000, v0
+; GFX7LESS-NEXT: v_and_b32_e32 v2, 0xffff0000, v0
; GFX7LESS-NEXT: s_not_b32 s2, s2
; GFX7LESS-NEXT: s_waitcnt lgkmcnt(0)
-; GFX7LESS-NEXT: v_mov_b32_e32 v1, s3
+; GFX7LESS-NEXT: v_mov_b32_e32 v0, s3
; GFX7LESS-NEXT: s_mov_b32 s6, -1
; GFX7LESS-NEXT: .LBB19_1: ; %atomicrmw.start
; GFX7LESS-NEXT: ; =>This Inner Loop Header: Depth=1
-; GFX7LESS-NEXT: v_lshrrev_b32_e32 v0, s10, v1
+; GFX7LESS-NEXT: v_mov_b32_e32 v4, v0
+; GFX7LESS-NEXT: s_waitcnt expcnt(0)
+; GFX7LESS-NEXT: v_lshrrev_b32_e32 v0, s10, v4
; GFX7LESS-NEXT: v_lshlrev_b32_e32 v0, 16, v0
-; GFX7LESS-NEXT: v_add_f32_e32 v0, v0, v4
+; GFX7LESS-NEXT: v_add_f32_e32 v0, v0, v2
; GFX7LESS-NEXT: v_lshrrev_b32_e32 v0, 16, v0
; GFX7LESS-NEXT: v_lshlrev_b32_e32 v0, s10, v0
-; GFX7LESS-NEXT: s_waitcnt expcnt(0)
-; GFX7LESS-NEXT: v_and_b32_e32 v2, s2, v1
-; GFX7LESS-NEXT: v_or_b32_e32 v0, v2, v0
-; GFX7LESS-NEXT: v_mov_b32_e32 v3, v1
-; GFX7LESS-NEXT: v_mov_b32_e32 v2, v0
-; GFX7LESS-NEXT: buffer_atomic_cmpswap v[2:3], off, s[4:7], 0 glc
+; GFX7LESS-NEXT: v_and_b32_e32 v1, s2, v4
+; GFX7LESS-NEXT: v_or_b32_e32 v3, v1, v0
+; GFX7LESS-NEXT: v_mov_b32_e32 v0, v3
+; GFX7LESS-NEXT: v_mov_b32_e32 v1, v4
+; GFX7LESS-NEXT: buffer_atomic_cmpswap v[0:1], off, s[4:7], 0 glc
; GFX7LESS-NEXT: s_waitcnt vmcnt(0)
-; GFX7LESS-NEXT: v_cmp_eq_u32_e32 vcc, v2, v1
+; GFX7LESS-NEXT: v_cmp_eq_u32_e32 vcc, v0, v4
; GFX7LESS-NEXT: s_or_b64 s[8:9], vcc, s[8:9]
-; GFX7LESS-NEXT: v_mov_b32_e32 v1, v2
; GFX7LESS-NEXT: s_andn2_b64 exec, exec, s[8:9]
; GFX7LESS-NEXT: s_cbranch_execnz .LBB19_1
; GFX7LESS-NEXT: ; %bb.2: ; %atomicrmw.end
; GFX7LESS-NEXT: s_or_b64 exec, exec, s[8:9]
; GFX7LESS-NEXT: s_mov_b32 s3, 0xf000
; GFX7LESS-NEXT: s_mov_b32 s2, -1
-; GFX7LESS-NEXT: v_lshrrev_b32_e32 v0, s10, v2
+; GFX7LESS-NEXT: s_waitcnt expcnt(0)
+; GFX7LESS-NEXT: v_lshrrev_b32_e32 v0, s10, v0
; GFX7LESS-NEXT: buffer_store_short v0, off, s[0:3], 0
; GFX7LESS-NEXT: s_endpgm
;
@@ -11158,37 +11196,37 @@ define amdgpu_kernel void @uniform_fadd_bf16(ptr addrspace(1) %result, ptr addrs
; GFX8-NEXT: s_lshl_b32 s2, s6, 16
; GFX8-NEXT: s_not_b32 s3, s3
; GFX8-NEXT: s_waitcnt lgkmcnt(0)
-; GFX8-NEXT: v_mov_b32_e32 v1, s7
+; GFX8-NEXT: v_mov_b32_e32 v0, s7
; GFX8-NEXT: s_mov_b32 s7, 0xf000
; GFX8-NEXT: s_mov_b32 s6, -1
; GFX8-NEXT: .LBB19_1: ; %atomicrmw.start
; GFX8-NEXT: ; =>This Inner Loop Header: Depth=1
-; GFX8-NEXT: v_mov_b32_e32 v0, s10
-; GFX8-NEXT: v_lshrrev_b32_sdwa v3, v0, v1 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:DWORD
-; GFX8-NEXT: v_add_f32_e32 v3, s2, v3
-; GFX8-NEXT: v_bfe_u32 v4, v3, 16, 1
-; GFX8-NEXT: v_add_u32_e32 v4, vcc, v4, v3
-; GFX8-NEXT: v_add_u32_e32 v4, vcc, 0x7fff, v4
-; GFX8-NEXT: v_or_b32_e32 v5, 0x400000, v3
-; GFX8-NEXT: v_cmp_u_f32_e32 vcc, v3, v3
-; GFX8-NEXT: v_cndmask_b32_e32 v3, v4, v5, vcc
-; GFX8-NEXT: v_and_b32_e32 v2, s3, v1
-; GFX8-NEXT: v_lshlrev_b32_sdwa v0, v0, v3 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:WORD_1
-; GFX8-NEXT: v_or_b32_e32 v0, v2, v0
-; GFX8-NEXT: v_mov_b32_e32 v3, v1
+; GFX8-NEXT: v_mov_b32_e32 v1, s10
; GFX8-NEXT: v_mov_b32_e32 v2, v0
-; GFX8-NEXT: buffer_atomic_cmpswap v[2:3], off, s[4:7], 0 glc
+; GFX8-NEXT: v_lshrrev_b32_sdwa v0, v1, v2 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:DWORD
+; GFX8-NEXT: v_add_f32_e32 v0, s2, v0
+; GFX8-NEXT: v_bfe_u32 v4, v0, 16, 1
+; GFX8-NEXT: v_add_u32_e32 v4, vcc, v4, v0
+; GFX8-NEXT: v_add_u32_e32 v4, vcc, 0x7fff, v4
+; GFX8-NEXT: v_or_b32_e32 v5, 0x400000, v0
+; GFX8-NEXT: v_cmp_u_f32_e32 vcc, v0, v0
+; GFX8-NEXT: v_cndmask_b32_e32 v0, v4, v5, vcc
+; GFX8-NEXT: v_and_b32_e32 v3, s3, v2
+; GFX8-NEXT: v_lshlrev_b32_sdwa v0, v1, v0 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:WORD_1
+; GFX8-NEXT: v_or_b32_e32 v1, v3, v0
+; GFX8-NEXT: v_mov_b32_e32 v0, v1
+; GFX8-NEXT: v_mov_b32_e32 v1, v2
+; GFX8-NEXT: buffer_atomic_cmpswap v[0:1], off, s[4:7], 0 glc
; GFX8-NEXT: s_waitcnt vmcnt(0)
-; GFX8-NEXT: v_cmp_eq_u32_e32 vcc, v2, v1
+; GFX8-NEXT: v_cmp_eq_u32_e32 vcc, v0, v2
; GFX8-NEXT: s_or_b64 s[8:9], vcc, s[8:9]
-; GFX8-NEXT: v_mov_b32_e32 v1, v2
; GFX8-NEXT: s_andn2_b64 exec, exec, s[8:9]
; GFX8-NEXT: s_cbranch_execnz .LBB19_1
; GFX8-NEXT: ; %bb.2: ; %atomicrmw.end
; GFX8-NEXT: s_or_b64 exec, exec, s[8:9]
; GFX8-NEXT: s_mov_b32 s3, 0xf000
; GFX8-NEXT: s_mov_b32 s2, -1
-; GFX8-NEXT: v_lshrrev_b32_e32 v0, s10, v2
+; GFX8-NEXT: v_lshrrev_b32_e32 v0, s10, v0
; GFX8-NEXT: buffer_store_short v0, off, s[0:3], 0
; GFX8-NEXT: s_endpgm
;
@@ -11208,34 +11246,34 @@ define amdgpu_kernel void @uniform_fadd_bf16(ptr addrspace(1) %result, ptr addrs
; GFX9-NEXT: s_lshl_b32 s2, s6, 16
; GFX9-NEXT: s_not_b32 s3, s3
; GFX9-NEXT: s_waitcnt lgkmcnt(0)
-; GFX9-NEXT: v_mov_b32_e32 v1, s7
+; GFX9-NEXT: v_mov_b32_e32 v0, s7
; GFX9-NEXT: s_mov_b32 s7, 0xf000
; GFX9-NEXT: s_mov_b32 s6, -1
; GFX9-NEXT: .LBB19_1: ; %atomicrmw.start
; GFX9-NEXT: ; =>This Inner Loop Header: Depth=1
-; GFX9-NEXT: v_lshrrev_b32_sdwa v0, s10, v1 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:DWORD
+; GFX9-NEXT: v_mov_b32_e32 v2, v0
+; GFX9-NEXT: v_lshrrev_b32_sdwa v0, s10, v2 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:DWORD
; GFX9-NEXT: v_add_f32_e32 v0, s2, v0
-; GFX9-NEXT: v_bfe_u32 v2, v0, 16, 1
+; GFX9-NEXT: v_bfe_u32 v1, v0, 16, 1
; GFX9-NEXT: v_or_b32_e32 v3, 0x400000, v0
-; GFX9-NEXT: v_add3_u32 v2, v2, v0, s11
+; GFX9-NEXT: v_add3_u32 v1, v1, v0, s11
; GFX9-NEXT: v_cmp_u_f32_e32 vcc, v0, v0
-; GFX9-NEXT: v_cndmask_b32_e32 v0, v2, v3, vcc
+; GFX9-NEXT: v_cndmask_b32_e32 v0, v1, v3, vcc
; GFX9-NEXT: v_lshlrev_b32_sdwa v0, s10, v0 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:WORD_1
-; GFX9-NEXT: v_and_or_b32 v0, v1, s3, v0
-; GFX9-NEXT: v_mov_b32_e32 v3, v1
-; GFX9-NEXT: v_mov_b32_e32 v2, v0
-; GFX9-NEXT: buffer_atomic_cmpswap v[2:3], off, s[4:7], 0 glc
+; GFX9-NEXT: v_and_or_b32 v1, v2, s3, v0
+; GFX9-NEXT: v_mov_b32_e32 v0, v1
+; GFX9-NEXT: v_mov_b32_e32 v1, v2
+; GFX9-NEXT: buffer_atomic_cmpswap v[0:1], off, s[4:7], 0 glc
; GFX9-NEXT: s_waitcnt vmcnt(0)
-; GFX9-NEXT: v_cmp_eq_u32_e32 vcc, v2, v1
+; GFX9-NEXT: v_cmp_eq_u32_e32 vcc, v0, v2
; GFX9-NEXT: s_or_b64 s[8:9], vcc, s[8:9]
-; GFX9-NEXT: v_mov_b32_e32 v1, v2
; GFX9-NEXT: s_andn2_b64 exec, exec, s[8:9]
; GFX9-NEXT: s_cbranch_execnz .LBB19_1
; GFX9-NEXT: ; %bb.2: ; %atomicrmw.end
; GFX9-NEXT: s_or_b64 exec, exec, s[8:9]
; GFX9-NEXT: s_mov_b32 s3, 0xf000
; GFX9-NEXT: s_mov_b32 s2, -1
-; GFX9-NEXT: v_lshrrev_b32_e32 v0, s10, v2
+; GFX9-NEXT: v_lshrrev_b32_e32 v0, s10, v0
; GFX9-NEXT: buffer_store_short v0, off, s[0:3], 0
; GFX9-NEXT: s_endpgm
;
@@ -11256,31 +11294,31 @@ define amdgpu_kernel void @uniform_fadd_bf16(ptr addrspace(1) %result, ptr addrs
; GFX1064-NEXT: s_not_b32 s9, s2
; GFX1064-NEXT: s_mov_b64 s[2:3], 0
; GFX1064-NEXT: s_waitcnt lgkmcnt(0)
-; GFX1064-NEXT: v_mov_b32_e32 v1, s7
+; GFX1064-NEXT: v_mov_b32_e32 v0, s7
; GFX1064-NEXT: s_mov_b32 s7, 0x31016000
; GFX1064-NEXT: .LBB19_1: ; %atomicrmw.start
; GFX1064-NEXT: ; =>This Inner Loop Header: Depth=1
-; GFX1064-NEXT: v_lshrrev_b32_sdwa v0, s8, v1 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:DWORD
+; GFX1064-NEXT: v_mov_b32_e32 v2, v0
+; GFX1064-NEXT: v_lshrrev_b32_sdwa v0, s8, v2 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:DWORD
; GFX1064-NEXT: v_add_f32_e32 v0, s10, v0
-; GFX1064-NEXT: v_bfe_u32 v2, v0, 16, 1
+; GFX1064-NEXT: v_bfe_u32 v1, v0, 16, 1
; GFX1064-NEXT: v_or_b32_e32 v3, 0x400000, v0
; GFX1064-NEXT: v_cmp_u_f32_e32 vcc, v0, v0
-; GFX1064-NEXT: v_add3_u32 v2, v2, v0, 0x7fff
-; GFX1064-NEXT: v_cndmask_b32_e32 v0, v2, v3, vcc
+; GFX1064-NEXT: v_add3_u32 v1, v1, v0, 0x7fff
+; GFX1064-NEXT: v_cndmask_b32_e32 v0, v1, v3, vcc
; GFX1064-NEXT: v_lshlrev_b32_sdwa v0, s8, v0 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:WORD_1
-; GFX1064-NEXT: v_and_or_b32 v0, v1, s9, v0
-; GFX1064-NEXT: v_mov_b32_e32 v3, v1
-; GFX1064-NEXT: v_mov_b32_e32 v2, v0
-; GFX1064-NEXT: buffer_atomic_cmpswap v[2:3], off, s[4:7], 0 glc
-; GFX1064-NEXT: s_waitcnt vmcnt(0)
-; GFX1064-NEXT: v_cmp_eq_u32_e32 vcc, v2, v1
+; GFX1064-NEXT: v_and_or_b32 v1, v2, s9, v0
+; GFX1064-NEXT: v_mov_b32_e32 v0, v1
; GFX1064-NEXT: v_mov_b32_e32 v1, v2
+; GFX1064-NEXT: buffer_atomic_cmpswap v[0:1], off, s[4:7], 0 glc
+; GFX1064-NEXT: s_waitcnt vmcnt(0)
+; GFX1064-NEXT: v_cmp_eq_u32_e32 vcc, v0, v2
; GFX1064-NEXT: s_or_b64 s[2:3], vcc, s[2:3]
; GFX1064-NEXT: s_andn2_b64 exec, exec, s[2:3]
; GFX1064-NEXT: s_cbranch_execnz .LBB19_1
; GFX1064-NEXT: ; %bb.2: ; %atomicrmw.end
; GFX1064-NEXT: s_or_b64 exec, exec, s[2:3]
-; GFX1064-NEXT: v_lshrrev_b32_e32 v0, s8, v2
+; GFX1064-NEXT: v_lshrrev_b32_e32 v0, s8, v0
; GFX1064-NEXT: s_mov_b32 s3, 0x31016000
; GFX1064-NEXT: s_mov_b32 s2, -1
; GFX1064-NEXT: buffer_store_short v0, off, s[0:3], 0
@@ -11303,31 +11341,31 @@ define amdgpu_kernel void @uniform_fadd_bf16(ptr addrspace(1) %result, ptr addrs
; GFX1032-NEXT: s_not_b32 s8, s3
; GFX1032-NEXT: s_mov_b32 s3, 0
; GFX1032-NEXT: s_waitcnt lgkmcnt(0)
-; GFX1032-NEXT: v_mov_b32_e32 v1, s7
+; GFX1032-NEXT: v_mov_b32_e32 v0, s7
; GFX1032-NEXT: s_mov_b32 s7, 0x31016000
; GFX1032-NEXT: .LBB19_1: ; %atomicrmw.start
; GFX1032-NEXT: ; =>This Inner Loop Header: Depth=1
-; GFX1032-NEXT: v_lshrrev_b32_sdwa v0, s2, v1 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:DWORD
+; GFX1032-NEXT: v_mov_b32_e32 v2, v0
+; GFX1032-NEXT: v_lshrrev_b32_sdwa v0, s2, v2 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:DWORD
; GFX1032-NEXT: v_add_f32_e32 v0, s9, v0
-; GFX1032-NEXT: v_bfe_u32 v2, v0, 16, 1
+; GFX1032-NEXT: v_bfe_u32 v1, v0, 16, 1
; GFX1032-NEXT: v_or_b32_e32 v3, 0x400000, v0
; GFX1032-NEXT: v_cmp_u_f32_e32 vcc_lo, v0, v0
-; GFX1032-NEXT: v_add3_u32 v2, v2, v0, 0x7fff
-; GFX1032-NEXT: v_cndmask_b32_e32 v0, v2, v3, vcc_lo
+; GFX1032-NEXT: v_add3_u32 v1, v1, v0, 0x7fff
+; GFX1032-NEXT: v_cndmask_b32_e32 v0, v1, v3, vcc_lo
; GFX1032-NEXT: v_lshlrev_b32_sdwa v0, s2, v0 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:WORD_1
-; GFX1032-NEXT: v_and_or_b32 v0, v1, s8, v0
-; GFX1032-NEXT: v_mov_b32_e32 v3, v1
-; GFX1032-NEXT: v_mov_b32_e32 v2, v0
-; GFX1032-NEXT: buffer_atomic_cmpswap v[2:3], off, s[4:7], 0 glc
-; GFX1032-NEXT: s_waitcnt vmcnt(0)
-; GFX1032-NEXT: v_cmp_eq_u32_e32 vcc_lo, v2, v1
+; GFX1032-NEXT: v_and_or_b32 v1, v2, s8, v0
+; GFX1032-NEXT: v_mov_b32_e32 v0, v1
; GFX1032-NEXT: v_mov_b32_e32 v1, v2
+; GFX1032-NEXT: buffer_atomic_cmpswap v[0:1], off, s[4:7], 0 glc
+; GFX1032-NEXT: s_waitcnt vmcnt(0)
+; GFX1032-NEXT: v_cmp_eq_u32_e32 vcc_lo, v0, v2
; GFX1032-NEXT: s_or_b32 s3, vcc_lo, s3
; GFX1032-NEXT: s_andn2_b32 exec_lo, exec_lo, s3
; GFX1032-NEXT: s_cbranch_execnz .LBB19_1
; GFX1032-NEXT: ; %bb.2: ; %atomicrmw.end
; GFX1032-NEXT: s_or_b32 exec_lo, exec_lo, s3
-; GFX1032-NEXT: v_lshrrev_b32_e32 v0, s2, v2
+; GFX1032-NEXT: v_lshrrev_b32_e32 v0, s2, v0
; GFX1032-NEXT: s_mov_b32 s3, 0x31016000
; GFX1032-NEXT: s_mov_b32 s2, -1
; GFX1032-NEXT: buffer_store_short v0, off, s[0:3], 0
@@ -11350,41 +11388,42 @@ define amdgpu_kernel void @uniform_fadd_bf16(ptr addrspace(1) %result, ptr addrs
; GFX1164-TRUE16-NEXT: s_not_b32 s9, s2
; GFX1164-TRUE16-NEXT: s_mov_b64 s[2:3], 0
; GFX1164-TRUE16-NEXT: s_waitcnt lgkmcnt(0)
-; GFX1164-TRUE16-NEXT: v_mov_b32_e32 v1, s7
+; GFX1164-TRUE16-NEXT: v_mov_b32_e32 v0, s7
; GFX1164-TRUE16-NEXT: s_mov_b32 s7, 0x31016000
; GFX1164-TRUE16-NEXT: .p2align 6
; GFX1164-TRUE16-NEXT: .LBB19_1: ; %atomicrmw.start
; GFX1164-TRUE16-NEXT: ; =>This Inner Loop Header: Depth=1
; GFX1164-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
-; GFX1164-TRUE16-NEXT: v_lshrrev_b32_e32 v0, s8, v1
-; GFX1164-TRUE16-NEXT: v_lshlrev_b32_e32 v0, 16, v0
+; GFX1164-TRUE16-NEXT: v_mov_b32_e32 v2, v0
+; GFX1164-TRUE16-NEXT: v_lshrrev_b32_e32 v0, s8, v2
; GFX1164-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
+; GFX1164-TRUE16-NEXT: v_lshlrev_b32_e32 v0, 16, v0
; GFX1164-TRUE16-NEXT: v_add_f32_e32 v0, s10, v0
-; GFX1164-TRUE16-NEXT: v_bfe_u32 v2, v0, 16, 1
+; GFX1164-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(SKIP_2) | instid1(VALU_DEP_3)
+; GFX1164-TRUE16-NEXT: v_bfe_u32 v1, v0, 16, 1
; GFX1164-TRUE16-NEXT: v_or_b32_e32 v3, 0x400000, v0
; GFX1164-TRUE16-NEXT: v_cmp_u_f32_e32 vcc, v0, v0
-; GFX1164-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_3) | instskip(NEXT) | instid1(VALU_DEP_1)
-; GFX1164-TRUE16-NEXT: v_add3_u32 v2, v2, v0, 0x7fff
-; GFX1164-TRUE16-NEXT: v_cndmask_b32_e32 v0, v2, v3, vcc
-; GFX1164-TRUE16-NEXT: v_mov_b16_e32 v2.h, 0
-; GFX1164-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_2) | instskip(NEXT) | instid1(VALU_DEP_1)
-; GFX1164-TRUE16-NEXT: v_mov_b16_e32 v2.l, v0.h
-; GFX1164-TRUE16-NEXT: v_lshlrev_b32_e32 v0, s8, v2
+; GFX1164-TRUE16-NEXT: v_add3_u32 v1, v1, v0, 0x7fff
; GFX1164-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(SKIP_1) | instid1(VALU_DEP_2)
-; GFX1164-TRUE16-NEXT: v_and_or_b32 v0, v1, s9, v0
-; GFX1164-TRUE16-NEXT: v_mov_b32_e32 v3, v1
-; GFX1164-TRUE16-NEXT: v_mov_b32_e32 v2, v0
-; GFX1164-TRUE16-NEXT: buffer_atomic_cmpswap_b32 v[2:3], off, s[4:7], 0 glc
-; GFX1164-TRUE16-NEXT: s_waitcnt vmcnt(0)
-; GFX1164-TRUE16-NEXT: v_cmp_eq_u32_e32 vcc, v2, v1
+; GFX1164-TRUE16-NEXT: v_cndmask_b32_e32 v0, v1, v3, vcc
+; GFX1164-TRUE16-NEXT: v_mov_b16_e32 v1.h, 0
+; GFX1164-TRUE16-NEXT: v_mov_b16_e32 v1.l, v0.h
+; GFX1164-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
+; GFX1164-TRUE16-NEXT: v_lshlrev_b32_e32 v0, s8, v1
+; GFX1164-TRUE16-NEXT: v_and_or_b32 v1, v2, s9, v0
+; GFX1164-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1)
+; GFX1164-TRUE16-NEXT: v_mov_b32_e32 v0, v1
; GFX1164-TRUE16-NEXT: v_mov_b32_e32 v1, v2
+; GFX1164-TRUE16-NEXT: buffer_atomic_cmpswap_b32 v[0:1], off, s[4:7], 0 glc
+; GFX1164-TRUE16-NEXT: s_waitcnt vmcnt(0)
+; GFX1164-TRUE16-NEXT: v_cmp_eq_u32_e32 vcc, v0, v2
; GFX1164-TRUE16-NEXT: s_or_b64 s[2:3], vcc, s[2:3]
; GFX1164-TRUE16-NEXT: s_delay_alu instid0(SALU_CYCLE_1)
; GFX1164-TRUE16-NEXT: s_and_not1_b64 exec, exec, s[2:3]
; GFX1164-TRUE16-NEXT: s_cbranch_execnz .LBB19_1
; GFX1164-TRUE16-NEXT: ; %bb.2: ; %atomicrmw.end
; GFX1164-TRUE16-NEXT: s_or_b64 exec, exec, s[2:3]
-; GFX1164-TRUE16-NEXT: v_lshrrev_b32_e32 v0, s8, v2
+; GFX1164-TRUE16-NEXT: v_lshrrev_b32_e32 v0, s8, v0
; GFX1164-TRUE16-NEXT: s_mov_b32 s3, 0x31016000
; GFX1164-TRUE16-NEXT: s_mov_b32 s2, -1
; GFX1164-TRUE16-NEXT: buffer_store_b16 v0, off, s[0:3], 0
@@ -11407,40 +11446,41 @@ define amdgpu_kernel void @uniform_fadd_bf16(ptr addrspace(1) %result, ptr addrs
; GFX1164-FAKE16-NEXT: s_not_b32 s9, s2
; GFX1164-FAKE16-NEXT: s_mov_b64 s[2:3], 0
; GFX1164-FAKE16-NEXT: s_waitcnt lgkmcnt(0)
-; GFX1164-FAKE16-NEXT: v_mov_b32_e32 v1, s7
+; GFX1164-FAKE16-NEXT: v_mov_b32_e32 v0, s7
; GFX1164-FAKE16-NEXT: s_mov_b32 s7, 0x31016000
; GFX1164-FAKE16-NEXT: .p2align 6
; GFX1164-FAKE16-NEXT: .LBB19_1: ; %atomicrmw.start
; GFX1164-FAKE16-NEXT: ; =>This Inner Loop Header: Depth=1
; GFX1164-FAKE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
-; GFX1164-FAKE16-NEXT: v_lshrrev_b32_e32 v0, s8, v1
-; GFX1164-FAKE16-NEXT: v_lshlrev_b32_e32 v0, 16, v0
+; GFX1164-FAKE16-NEXT: v_mov_b32_e32 v2, v0
+; GFX1164-FAKE16-NEXT: v_lshrrev_b32_e32 v0, s8, v2
; GFX1164-FAKE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
+; GFX1164-FAKE16-NEXT: v_lshlrev_b32_e32 v0, 16, v0
; GFX1164-FAKE16-NEXT: v_add_f32_e32 v0, s10, v0
-; GFX1164-FAKE16-NEXT: v_bfe_u32 v2, v0, 16, 1
+; GFX1164-FAKE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(SKIP_2) | instid1(VALU_DEP_3)
+; GFX1164-FAKE16-NEXT: v_bfe_u32 v1, v0, 16, 1
; GFX1164-FAKE16-NEXT: v_or_b32_e32 v3, 0x400000, v0
; GFX1164-FAKE16-NEXT: v_cmp_u_f32_e32 vcc, v0, v0
-; GFX1164-FAKE16-NEXT: s_delay_alu instid0(VALU_DEP_3) | instskip(NEXT) | instid1(VALU_DEP_1)
-; GFX1164-FAKE16-NEXT: v_add3_u32 v2, v2, v0, 0x7fff
-; GFX1164-FAKE16-NEXT: v_cndmask_b32_e32 v0, v2, v3, vcc
+; GFX1164-FAKE16-NEXT: v_add3_u32 v1, v1, v0, 0x7fff
; GFX1164-FAKE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
+; GFX1164-FAKE16-NEXT: v_cndmask_b32_e32 v0, v1, v3, vcc
; GFX1164-FAKE16-NEXT: v_lshrrev_b32_e32 v0, 16, v0
+; GFX1164-FAKE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
; GFX1164-FAKE16-NEXT: v_lshlrev_b32_e32 v0, s8, v0
-; GFX1164-FAKE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(SKIP_1) | instid1(VALU_DEP_2)
-; GFX1164-FAKE16-NEXT: v_and_or_b32 v0, v1, s9, v0
-; GFX1164-FAKE16-NEXT: v_mov_b32_e32 v3, v1
-; GFX1164-FAKE16-NEXT: v_mov_b32_e32 v2, v0
-; GFX1164-FAKE16-NEXT: buffer_atomic_cmpswap_b32 v[2:3], off, s[4:7], 0 glc
-; GFX1164-FAKE16-NEXT: s_waitcnt vmcnt(0)
-; GFX1164-FAKE16-NEXT: v_cmp_eq_u32_e32 vcc, v2, v1
+; GFX1164-FAKE16-NEXT: v_and_or_b32 v1, v2, s9, v0
+; GFX1164-FAKE16-NEXT: s_delay_alu instid0(VALU_DEP_1)
+; GFX1164-FAKE16-NEXT: v_mov_b32_e32 v0, v1
; GFX1164-FAKE16-NEXT: v_mov_b32_e32 v1, v2
+; GFX1164-FAKE16-NEXT: buffer_atomic_cmpswap_b32 v[0:1], off, s[4:7], 0 glc
+; GFX1164-FAKE16-NEXT: s_waitcnt vmcnt(0)
+; GFX1164-FAKE16-NEXT: v_cmp_eq_u32_e32 vcc, v0, v2
; GFX1164-FAKE16-NEXT: s_or_b64 s[2:3], vcc, s[2:3]
; GFX1164-FAKE16-NEXT: s_delay_alu instid0(SALU_CYCLE_1)
; GFX1164-FAKE16-NEXT: s_and_not1_b64 exec, exec, s[2:3]
; GFX1164-FAKE16-NEXT: s_cbranch_execnz .LBB19_1
; GFX1164-FAKE16-NEXT: ; %bb.2: ; %atomicrmw.end
; GFX1164-FAKE16-NEXT: s_or_b64 exec, exec, s[2:3]
-; GFX1164-FAKE16-NEXT: v_lshrrev_b32_e32 v0, s8, v2
+; GFX1164-FAKE16-NEXT: v_lshrrev_b32_e32 v0, s8, v0
; GFX1164-FAKE16-NEXT: s_mov_b32 s3, 0x31016000
; GFX1164-FAKE16-NEXT: s_mov_b32 s2, -1
; GFX1164-FAKE16-NEXT: buffer_store_b16 v0, off, s[0:3], 0
@@ -11463,40 +11503,42 @@ define amdgpu_kernel void @uniform_fadd_bf16(ptr addrspace(1) %result, ptr addrs
; GFX1132-TRUE16-NEXT: s_not_b32 s8, s3
; GFX1132-TRUE16-NEXT: s_mov_b32 s3, 0
; GFX1132-TRUE16-NEXT: s_waitcnt lgkmcnt(0)
-; GFX1132-TRUE16-NEXT: v_mov_b32_e32 v1, s7
+; GFX1132-TRUE16-NEXT: v_mov_b32_e32 v0, s7
; GFX1132-TRUE16-NEXT: s_mov_b32 s7, 0x31016000
; GFX1132-TRUE16-NEXT: .p2align 6
; GFX1132-TRUE16-NEXT: .LBB19_1: ; %atomicrmw.start
; GFX1132-TRUE16-NEXT: ; =>This Inner Loop Header: Depth=1
; GFX1132-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
-; GFX1132-TRUE16-NEXT: v_lshrrev_b32_e32 v0, s2, v1
-; GFX1132-TRUE16-NEXT: v_lshlrev_b32_e32 v0, 16, v0
+; GFX1132-TRUE16-NEXT: v_mov_b32_e32 v2, v0
+; GFX1132-TRUE16-NEXT: v_lshrrev_b32_e32 v0, s2, v2
; GFX1132-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
+; GFX1132-TRUE16-NEXT: v_lshlrev_b32_e32 v0, 16, v0
; GFX1132-TRUE16-NEXT: v_add_f32_e32 v0, s9, v0
-; GFX1132-TRUE16-NEXT: v_bfe_u32 v2, v0, 16, 1
+; GFX1132-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(SKIP_2) | instid1(VALU_DEP_3)
+; GFX1132-TRUE16-NEXT: v_bfe_u32 v1, v0, 16, 1
; GFX1132-TRUE16-NEXT: v_or_b32_e32 v3, 0x400000, v0
; GFX1132-TRUE16-NEXT: v_cmp_u_f32_e32 vcc_lo, v0, v0
-; GFX1132-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_3) | instskip(NEXT) | instid1(VALU_DEP_1)
-; GFX1132-TRUE16-NEXT: v_add3_u32 v2, v2, v0, 0x7fff
-; GFX1132-TRUE16-NEXT: v_cndmask_b32_e32 v0, v2, v3, vcc_lo
-; GFX1132-TRUE16-NEXT: v_mov_b16_e32 v2.h, 0
-; GFX1132-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_2) | instskip(NEXT) | instid1(VALU_DEP_1)
-; GFX1132-TRUE16-NEXT: v_mov_b16_e32 v2.l, v0.h
-; GFX1132-TRUE16-NEXT: v_lshlrev_b32_e32 v0, s2, v2
+; GFX1132-TRUE16-NEXT: v_add3_u32 v1, v1, v0, 0x7fff
+; GFX1132-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(SKIP_1) | instid1(VALU_DEP_2)
+; GFX1132-TRUE16-NEXT: v_cndmask_b32_e32 v0, v1, v3, vcc_lo
+; GFX1132-TRUE16-NEXT: v_mov_b16_e32 v1.h, 0
+; GFX1132-TRUE16-NEXT: v_mov_b16_e32 v1.l, v0.h
; GFX1132-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
-; GFX1132-TRUE16-NEXT: v_and_or_b32 v0, v1, s8, v0
-; GFX1132-TRUE16-NEXT: v_dual_mov_b32 v3, v1 :: v_dual_mov_b32 v2, v0
-; GFX1132-TRUE16-NEXT: buffer_atomic_cmpswap_b32 v[2:3], off, s[4:7], 0 glc
-; GFX1132-TRUE16-NEXT: s_waitcnt vmcnt(0)
-; GFX1132-TRUE16-NEXT: v_cmp_eq_u32_e32 vcc_lo, v2, v1
+; GFX1132-TRUE16-NEXT: v_lshlrev_b32_e32 v0, s2, v1
+; GFX1132-TRUE16-NEXT: v_and_or_b32 v1, v2, s8, v0
+; GFX1132-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1)
+; GFX1132-TRUE16-NEXT: v_mov_b32_e32 v0, v1
; GFX1132-TRUE16-NEXT: v_mov_b32_e32 v1, v2
+; GFX1132-TRUE16-NEXT: buffer_atomic_cmpswap_b32 v[0:1], off, s[4:7], 0 glc
+; GFX1132-TRUE16-NEXT: s_waitcnt vmcnt(0)
+; GFX1132-TRUE16-NEXT: v_cmp_eq_u32_e32 vcc_lo, v0, v2
; GFX1132-TRUE16-NEXT: s_or_b32 s3, vcc_lo, s3
; GFX1132-TRUE16-NEXT: s_delay_alu instid0(SALU_CYCLE_1)
; GFX1132-TRUE16-NEXT: s_and_not1_b32 exec_lo, exec_lo, s3
; GFX1132-TRUE16-NEXT: s_cbranch_execnz .LBB19_1
; GFX1132-TRUE16-NEXT: ; %bb.2: ; %atomicrmw.end
; GFX1132-TRUE16-NEXT: s_or_b32 exec_lo, exec_lo, s3
-; GFX1132-TRUE16-NEXT: v_lshrrev_b32_e32 v0, s2, v2
+; GFX1132-TRUE16-NEXT: v_lshrrev_b32_e32 v0, s2, v0
; GFX1132-TRUE16-NEXT: s_mov_b32 s3, 0x31016000
; GFX1132-TRUE16-NEXT: s_mov_b32 s2, -1
; GFX1132-TRUE16-NEXT: buffer_store_b16 v0, off, s[0:3], 0
@@ -11519,39 +11561,41 @@ define amdgpu_kernel void @uniform_fadd_bf16(ptr addrspace(1) %result, ptr addrs
; GFX1132-FAKE16-NEXT: s_not_b32 s8, s3
; GFX1132-FAKE16-NEXT: s_mov_b32 s3, 0
; GFX1132-FAKE16-NEXT: s_waitcnt lgkmcnt(0)
-; GFX1132-FAKE16-NEXT: v_mov_b32_e32 v1, s7
+; GFX1132-FAKE16-NEXT: v_mov_b32_e32 v0, s7
; GFX1132-FAKE16-NEXT: s_mov_b32 s7, 0x31016000
; GFX1132-FAKE16-NEXT: .p2align 6
; GFX1132-FAKE16-NEXT: .LBB19_1: ; %atomicrmw.start
; GFX1132-FAKE16-NEXT: ; =>This Inner Loop Header: Depth=1
; GFX1132-FAKE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
-; GFX1132-FAKE16-NEXT: v_lshrrev_b32_e32 v0, s2, v1
-; GFX1132-FAKE16-NEXT: v_lshlrev_b32_e32 v0, 16, v0
+; GFX1132-FAKE16-NEXT: v_mov_b32_e32 v2, v0
+; GFX1132-FAKE16-NEXT: v_lshrrev_b32_e32 v0, s2, v2
; GFX1132-FAKE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
+; GFX1132-FAKE16-NEXT: v_lshlrev_b32_e32 v0, 16, v0
; GFX1132-FAKE16-NEXT: v_add_f32_e32 v0, s9, v0
-; GFX1132-FAKE16-NEXT: v_bfe_u32 v2, v0, 16, 1
+; GFX1132-FAKE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(SKIP_2) | instid1(VALU_DEP_3)
+; GFX1132-FAKE16-NEXT: v_bfe_u32 v1, v0, 16, 1
; GFX1132-FAKE16-NEXT: v_or_b32_e32 v3, 0x400000, v0
; GFX1132-FAKE16-NEXT: v_cmp_u_f32_e32 vcc_lo, v0, v0
-; GFX1132-FAKE16-NEXT: s_delay_alu instid0(VALU_DEP_3) | instskip(NEXT) | instid1(VALU_DEP_1)
-; GFX1132-FAKE16-NEXT: v_add3_u32 v2, v2, v0, 0x7fff
-; GFX1132-FAKE16-NEXT: v_cndmask_b32_e32 v0, v2, v3, vcc_lo
+; GFX1132-FAKE16-NEXT: v_add3_u32 v1, v1, v0, 0x7fff
; GFX1132-FAKE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
+; GFX1132-FAKE16-NEXT: v_cndmask_b32_e32 v0, v1, v3, vcc_lo
; GFX1132-FAKE16-NEXT: v_lshrrev_b32_e32 v0, 16, v0
-; GFX1132-FAKE16-NEXT: v_lshlrev_b32_e32 v0, s2, v0
; GFX1132-FAKE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
-; GFX1132-FAKE16-NEXT: v_and_or_b32 v0, v1, s8, v0
-; GFX1132-FAKE16-NEXT: v_dual_mov_b32 v3, v1 :: v_dual_mov_b32 v2, v0
-; GFX1132-FAKE16-NEXT: buffer_atomic_cmpswap_b32 v[2:3], off, s[4:7], 0 glc
-; GFX1132-FAKE16-NEXT: s_waitcnt vmcnt(0)
-; GFX1132-FAKE16-NEXT: v_cmp_eq_u32_e32 vcc_lo, v2, v1
+; GFX1132-FAKE16-NEXT: v_lshlrev_b32_e32 v0, s2, v0
+; GFX1132-FAKE16-NEXT: v_and_or_b32 v1, v2, s8, v0
+; GFX1132-FAKE16-NEXT: s_delay_alu instid0(VALU_DEP_1)
+; GFX1132-FAKE16-NEXT: v_mov_b32_e32 v0, v1
; GFX1132-FAKE16-NEXT: v_mov_b32_e32 v1, v2
+; GFX1132-FAKE16-NEXT: buffer_atomic_cmpswap_b32 v[0:1], off, s[4:7], 0 glc
+; GFX1132-FAKE16-NEXT: s_waitcnt vmcnt(0)
+; GFX1132-FAKE16-NEXT: v_cmp_eq_u32_e32 vcc_lo, v0, v2
; GFX1132-FAKE16-NEXT: s_or_b32 s3, vcc_lo, s3
; GFX1132-FAKE16-NEXT: s_delay_alu instid0(SALU_CYCLE_1)
; GFX1132-FAKE16-NEXT: s_and_not1_b32 exec_lo, exec_lo, s3
; GFX1132-FAKE16-NEXT: s_cbranch_execnz .LBB19_1
; GFX1132-FAKE16-NEXT: ; %bb.2: ; %atomicrmw.end
; GFX1132-FAKE16-NEXT: s_or_b32 exec_lo, exec_lo, s3
-; GFX1132-FAKE16-NEXT: v_lshrrev_b32_e32 v0, s2, v2
+; GFX1132-FAKE16-NEXT: v_lshrrev_b32_e32 v0, s2, v0
; GFX1132-FAKE16-NEXT: s_mov_b32 s3, 0x31016000
; GFX1132-FAKE16-NEXT: s_mov_b32 s2, -1
; GFX1132-FAKE16-NEXT: buffer_store_b16 v0, off, s[0:3], 0
@@ -11574,41 +11618,42 @@ define amdgpu_kernel void @uniform_fadd_bf16(ptr addrspace(1) %result, ptr addrs
; GFX1264-TRUE16-NEXT: s_not_b32 s9, s2
; GFX1264-TRUE16-NEXT: s_mov_b64 s[2:3], 0
; GFX1264-TRUE16-NEXT: s_wait_kmcnt 0x0
-; GFX1264-TRUE16-NEXT: v_mov_b32_e32 v1, s7
+; GFX1264-TRUE16-NEXT: v_mov_b32_e32 v0, s7
; GFX1264-TRUE16-NEXT: s_mov_b32 s7, 0x31016000
; GFX1264-TRUE16-NEXT: .LBB19_1: ; %atomicrmw.start
; GFX1264-TRUE16-NEXT: ; =>This Inner Loop Header: Depth=1
; GFX1264-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
-; GFX1264-TRUE16-NEXT: v_lshrrev_b32_e32 v0, s8, v1
-; GFX1264-TRUE16-NEXT: v_lshlrev_b32_e32 v0, 16, v0
+; GFX1264-TRUE16-NEXT: v_mov_b32_e32 v2, v0
+; GFX1264-TRUE16-NEXT: v_lshrrev_b32_e32 v0, s8, v2
; GFX1264-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
+; GFX1264-TRUE16-NEXT: v_lshlrev_b32_e32 v0, 16, v0
; GFX1264-TRUE16-NEXT: v_add_f32_e32 v0, s10, v0
-; GFX1264-TRUE16-NEXT: v_bfe_u32 v2, v0, 16, 1
+; GFX1264-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(SKIP_2) | instid1(VALU_DEP_3)
+; GFX1264-TRUE16-NEXT: v_bfe_u32 v1, v0, 16, 1
; GFX1264-TRUE16-NEXT: v_or_b32_e32 v3, 0x400000, v0
; GFX1264-TRUE16-NEXT: v_cmp_u_f32_e32 vcc, v0, v0
-; GFX1264-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_3) | instskip(SKIP_1) | instid1(VALU_DEP_1)
-; GFX1264-TRUE16-NEXT: v_add3_u32 v2, v2, v0, 0x7fff
+; GFX1264-TRUE16-NEXT: v_add3_u32 v1, v1, v0, 0x7fff
; GFX1264-TRUE16-NEXT: s_wait_alu 0xfffd
-; GFX1264-TRUE16-NEXT: v_cndmask_b32_e32 v0, v2, v3, vcc
-; GFX1264-TRUE16-NEXT: v_mov_b16_e32 v2.h, 0
-; GFX1264-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_2) | instskip(NEXT) | instid1(VALU_DEP_1)
-; GFX1264-TRUE16-NEXT: v_mov_b16_e32 v2.l, v0.h
-; GFX1264-TRUE16-NEXT: v_lshlrev_b32_e32 v0, s8, v2
; GFX1264-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(SKIP_1) | instid1(VALU_DEP_2)
-; GFX1264-TRUE16-NEXT: v_and_or_b32 v0, v1, s9, v0
-; GFX1264-TRUE16-NEXT: v_mov_b32_e32 v3, v1
-; GFX1264-TRUE16-NEXT: v_mov_b32_e32 v2, v0
-; GFX1264-TRUE16-NEXT: buffer_atomic_cmpswap_b32 v[2:3], off, s[4:7], null th:TH_ATOMIC_RETURN scope:SCOPE_SYS
-; GFX1264-TRUE16-NEXT: s_wait_loadcnt 0x0
-; GFX1264-TRUE16-NEXT: v_cmp_eq_u32_e32 vcc, v2, v1
+; GFX1264-TRUE16-NEXT: v_cndmask_b32_e32 v0, v1, v3, vcc
+; GFX1264-TRUE16-NEXT: v_mov_b16_e32 v1.h, 0
+; GFX1264-TRUE16-NEXT: v_mov_b16_e32 v1.l, v0.h
+; GFX1264-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
+; GFX1264-TRUE16-NEXT: v_lshlrev_b32_e32 v0, s8, v1
+; GFX1264-TRUE16-NEXT: v_and_or_b32 v1, v2, s9, v0
+; GFX1264-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1)
+; GFX1264-TRUE16-NEXT: v_mov_b32_e32 v0, v1
; GFX1264-TRUE16-NEXT: v_mov_b32_e32 v1, v2
+; GFX1264-TRUE16-NEXT: buffer_atomic_cmpswap_b32 v[0:1], off, s[4:7], null th:TH_ATOMIC_RETURN scope:SCOPE_SYS
+; GFX1264-TRUE16-NEXT: s_wait_loadcnt 0x0
+; GFX1264-TRUE16-NEXT: v_cmp_eq_u32_e32 vcc, v0, v2
; GFX1264-TRUE16-NEXT: s_or_b64 s[2:3], vcc, s[2:3]
; GFX1264-TRUE16-NEXT: s_delay_alu instid0(SALU_CYCLE_1)
; GFX1264-TRUE16-NEXT: s_and_not1_b64 exec, exec, s[2:3]
; GFX1264-TRUE16-NEXT: s_cbranch_execnz .LBB19_1
; GFX1264-TRUE16-NEXT: ; %bb.2: ; %atomicrmw.end
; GFX1264-TRUE16-NEXT: s_or_b64 exec, exec, s[2:3]
-; GFX1264-TRUE16-NEXT: v_lshrrev_b32_e32 v0, s8, v2
+; GFX1264-TRUE16-NEXT: v_lshrrev_b32_e32 v0, s8, v0
; GFX1264-TRUE16-NEXT: s_mov_b32 s3, 0x31016000
; GFX1264-TRUE16-NEXT: s_mov_b32 s2, -1
; GFX1264-TRUE16-NEXT: buffer_store_b16 v0, off, s[0:3], null
@@ -11631,40 +11676,41 @@ define amdgpu_kernel void @uniform_fadd_bf16(ptr addrspace(1) %result, ptr addrs
; GFX1264-FAKE16-NEXT: s_not_b32 s9, s2
; GFX1264-FAKE16-NEXT: s_mov_b64 s[2:3], 0
; GFX1264-FAKE16-NEXT: s_wait_kmcnt 0x0
-; GFX1264-FAKE16-NEXT: v_mov_b32_e32 v1, s7
+; GFX1264-FAKE16-NEXT: v_mov_b32_e32 v0, s7
; GFX1264-FAKE16-NEXT: s_mov_b32 s7, 0x31016000
; GFX1264-FAKE16-NEXT: .LBB19_1: ; %atomicrmw.start
; GFX1264-FAKE16-NEXT: ; =>This Inner Loop Header: Depth=1
; GFX1264-FAKE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
-; GFX1264-FAKE16-NEXT: v_lshrrev_b32_e32 v0, s8, v1
-; GFX1264-FAKE16-NEXT: v_lshlrev_b32_e32 v0, 16, v0
+; GFX1264-FAKE16-NEXT: v_mov_b32_e32 v2, v0
+; GFX1264-FAKE16-NEXT: v_lshrrev_b32_e32 v0, s8, v2
; GFX1264-FAKE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
+; GFX1264-FAKE16-NEXT: v_lshlrev_b32_e32 v0, 16, v0
; GFX1264-FAKE16-NEXT: v_add_f32_e32 v0, s10, v0
-; GFX1264-FAKE16-NEXT: v_bfe_u32 v2, v0, 16, 1
+; GFX1264-FAKE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(SKIP_2) | instid1(VALU_DEP_3)
+; GFX1264-FAKE16-NEXT: v_bfe_u32 v1, v0, 16, 1
; GFX1264-FAKE16-NEXT: v_or_b32_e32 v3, 0x400000, v0
; GFX1264-FAKE16-NEXT: v_cmp_u_f32_e32 vcc, v0, v0
-; GFX1264-FAKE16-NEXT: s_delay_alu instid0(VALU_DEP_3) | instskip(SKIP_1) | instid1(VALU_DEP_1)
-; GFX1264-FAKE16-NEXT: v_add3_u32 v2, v2, v0, 0x7fff
+; GFX1264-FAKE16-NEXT: v_add3_u32 v1, v1, v0, 0x7fff
; GFX1264-FAKE16-NEXT: s_wait_alu 0xfffd
-; GFX1264-FAKE16-NEXT: v_cndmask_b32_e32 v0, v2, v3, vcc
; GFX1264-FAKE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
+; GFX1264-FAKE16-NEXT: v_cndmask_b32_e32 v0, v1, v3, vcc
; GFX1264-FAKE16-NEXT: v_lshrrev_b32_e32 v0, 16, v0
+; GFX1264-FAKE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
; GFX1264-FAKE16-NEXT: v_lshlrev_b32_e32 v0, s8, v0
-; GFX1264-FAKE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(SKIP_1) | instid1(VALU_DEP_2)
-; GFX1264-FAKE16-NEXT: v_and_or_b32 v0, v1, s9, v0
-; GFX1264-FAKE16-NEXT: v_mov_b32_e32 v3, v1
-; GFX1264-FAKE16-NEXT: v_mov_b32_e32 v2, v0
-; GFX1264-FAKE16-NEXT: buffer_atomic_cmpswap_b32 v[2:3], off, s[4:7], null th:TH_ATOMIC_RETURN scope:SCOPE_SYS
-; GFX1264-FAKE16-NEXT: s_wait_loadcnt 0x0
-; GFX1264-FAKE16-NEXT: v_cmp_eq_u32_e32 vcc, v2, v1
+; GFX1264-FAKE16-NEXT: v_and_or_b32 v1, v2, s9, v0
+; GFX1264-FAKE16-NEXT: s_delay_alu instid0(VALU_DEP_1)
+; GFX1264-FAKE16-NEXT: v_mov_b32_e32 v0, v1
; GFX1264-FAKE16-NEXT: v_mov_b32_e32 v1, v2
+; GFX1264-FAKE16-NEXT: buffer_atomic_cmpswap_b32 v[0:1], off, s[4:7], null th:TH_ATOMIC_RETURN scope:SCOPE_SYS
+; GFX1264-FAKE16-NEXT: s_wait_loadcnt 0x0
+; GFX1264-FAKE16-NEXT: v_cmp_eq_u32_e32 vcc, v0, v2
; GFX1264-FAKE16-NEXT: s_or_b64 s[2:3], vcc, s[2:3]
; GFX1264-FAKE16-NEXT: s_delay_alu instid0(SALU_CYCLE_1)
; GFX1264-FAKE16-NEXT: s_and_not1_b64 exec, exec, s[2:3]
; GFX1264-FAKE16-NEXT: s_cbranch_execnz .LBB19_1
; GFX1264-FAKE16-NEXT: ; %bb.2: ; %atomicrmw.end
; GFX1264-FAKE16-NEXT: s_or_b64 exec, exec, s[2:3]
-; GFX1264-FAKE16-NEXT: v_lshrrev_b32_e32 v0, s8, v2
+; GFX1264-FAKE16-NEXT: v_lshrrev_b32_e32 v0, s8, v0
; GFX1264-FAKE16-NEXT: s_mov_b32 s3, 0x31016000
; GFX1264-FAKE16-NEXT: s_mov_b32 s2, -1
; GFX1264-FAKE16-NEXT: buffer_store_b16 v0, off, s[0:3], null
@@ -11687,40 +11733,42 @@ define amdgpu_kernel void @uniform_fadd_bf16(ptr addrspace(1) %result, ptr addrs
; GFX1232-TRUE16-NEXT: s_not_b32 s8, s3
; GFX1232-TRUE16-NEXT: s_mov_b32 s3, 0
; GFX1232-TRUE16-NEXT: s_wait_kmcnt 0x0
-; GFX1232-TRUE16-NEXT: v_mov_b32_e32 v1, s7
+; GFX1232-TRUE16-NEXT: v_mov_b32_e32 v0, s7
; GFX1232-TRUE16-NEXT: s_mov_b32 s7, 0x31016000
; GFX1232-TRUE16-NEXT: .LBB19_1: ; %atomicrmw.start
; GFX1232-TRUE16-NEXT: ; =>This Inner Loop Header: Depth=1
; GFX1232-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
-; GFX1232-TRUE16-NEXT: v_lshrrev_b32_e32 v0, s2, v1
-; GFX1232-TRUE16-NEXT: v_lshlrev_b32_e32 v0, 16, v0
+; GFX1232-TRUE16-NEXT: v_mov_b32_e32 v2, v0
+; GFX1232-TRUE16-NEXT: v_lshrrev_b32_e32 v0, s2, v2
; GFX1232-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
+; GFX1232-TRUE16-NEXT: v_lshlrev_b32_e32 v0, 16, v0
; GFX1232-TRUE16-NEXT: v_add_f32_e32 v0, s9, v0
-; GFX1232-TRUE16-NEXT: v_bfe_u32 v2, v0, 16, 1
+; GFX1232-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(SKIP_2) | instid1(VALU_DEP_3)
+; GFX1232-TRUE16-NEXT: v_bfe_u32 v1, v0, 16, 1
; GFX1232-TRUE16-NEXT: v_or_b32_e32 v3, 0x400000, v0
; GFX1232-TRUE16-NEXT: v_cmp_u_f32_e32 vcc_lo, v0, v0
-; GFX1232-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_3) | instskip(SKIP_1) | instid1(VALU_DEP_1)
-; GFX1232-TRUE16-NEXT: v_add3_u32 v2, v2, v0, 0x7fff
+; GFX1232-TRUE16-NEXT: v_add3_u32 v1, v1, v0, 0x7fff
; GFX1232-TRUE16-NEXT: s_wait_alu 0xfffd
-; GFX1232-TRUE16-NEXT: v_cndmask_b32_e32 v0, v2, v3, vcc_lo
-; GFX1232-TRUE16-NEXT: v_mov_b16_e32 v2.h, 0
-; GFX1232-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_2) | instskip(NEXT) | instid1(VALU_DEP_1)
-; GFX1232-TRUE16-NEXT: v_mov_b16_e32 v2.l, v0.h
-; GFX1232-TRUE16-NEXT: v_lshlrev_b32_e32 v0, s2, v2
+; GFX1232-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(SKIP_1) | instid1(VALU_DEP_2)
+; GFX1232-TRUE16-NEXT: v_cndmask_b32_e32 v0, v1, v3, vcc_lo
+; GFX1232-TRUE16-NEXT: v_mov_b16_e32 v1.h, 0
+; GFX1232-TRUE16-NEXT: v_mov_b16_e32 v1.l, v0.h
; GFX1232-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
-; GFX1232-TRUE16-NEXT: v_and_or_b32 v0, v1, s8, v0
-; GFX1232-TRUE16-NEXT: v_dual_mov_b32 v3, v1 :: v_dual_mov_b32 v2, v0
-; GFX1232-TRUE16-NEXT: buffer_atomic_cmpswap_b32 v[2:3], off, s[4:7], null th:TH_ATOMIC_RETURN scope:SCOPE_SYS
-; GFX1232-TRUE16-NEXT: s_wait_loadcnt 0x0
-; GFX1232-TRUE16-NEXT: v_cmp_eq_u32_e32 vcc_lo, v2, v1
+; GFX1232-TRUE16-NEXT: v_lshlrev_b32_e32 v0, s2, v1
+; GFX1232-TRUE16-NEXT: v_and_or_b32 v1, v2, s8, v0
+; GFX1232-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1)
+; GFX1232-TRUE16-NEXT: v_mov_b32_e32 v0, v1
; GFX1232-TRUE16-NEXT: v_mov_b32_e32 v1, v2
+; GFX1232-TRUE16-NEXT: buffer_atomic_cmpswap_b32 v[0:1], off, s[4:7], null th:TH_ATOMIC_RETURN scope:SCOPE_SYS
+; GFX1232-TRUE16-NEXT: s_wait_loadcnt 0x0
+; GFX1232-TRUE16-NEXT: v_cmp_eq_u32_e32 vcc_lo, v0, v2
; GFX1232-TRUE16-NEXT: s_or_b32 s3, vcc_lo, s3
; GFX1232-TRUE16-NEXT: s_wait_alu 0xfffe
; GFX1232-TRUE16-NEXT: s_and_not1_b32 exec_lo, exec_lo, s3
; GFX1232-TRUE16-NEXT: s_cbranch_execnz .LBB19_1
; GFX1232-TRUE16-NEXT: ; %bb.2: ; %atomicrmw.end
; GFX1232-TRUE16-NEXT: s_or_b32 exec_lo, exec_lo, s3
-; GFX1232-TRUE16-NEXT: v_lshrrev_b32_e32 v0, s2, v2
+; GFX1232-TRUE16-NEXT: v_lshrrev_b32_e32 v0, s2, v0
; GFX1232-TRUE16-NEXT: s_mov_b32 s3, 0x31016000
; GFX1232-TRUE16-NEXT: s_mov_b32 s2, -1
; GFX1232-TRUE16-NEXT: buffer_store_b16 v0, off, s[0:3], null
@@ -11743,39 +11791,41 @@ define amdgpu_kernel void @uniform_fadd_bf16(ptr addrspace(1) %result, ptr addrs
; GFX1232-FAKE16-NEXT: s_not_b32 s8, s3
; GFX1232-FAKE16-NEXT: s_mov_b32 s3, 0
; GFX1232-FAKE16-NEXT: s_wait_kmcnt 0x0
-; GFX1232-FAKE16-NEXT: v_mov_b32_e32 v1, s7
+; GFX1232-FAKE16-NEXT: v_mov_b32_e32 v0, s7
; GFX1232-FAKE16-NEXT: s_mov_b32 s7, 0x31016000
; GFX1232-FAKE16-NEXT: .LBB19_1: ; %atomicrmw.start
; GFX1232-FAKE16-NEXT: ; =>This Inner Loop Header: Depth=1
; GFX1232-FAKE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
-; GFX1232-FAKE16-NEXT: v_lshrrev_b32_e32 v0, s2, v1
-; GFX1232-FAKE16-NEXT: v_lshlrev_b32_e32 v0, 16, v0
+; GFX1232-FAKE16-NEXT: v_mov_b32_e32 v2, v0
+; GFX1232-FAKE16-NEXT: v_lshrrev_b32_e32 v0, s2, v2
; GFX1232-FAKE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
+; GFX1232-FAKE16-NEXT: v_lshlrev_b32_e32 v0, 16, v0
; GFX1232-FAKE16-NEXT: v_add_f32_e32 v0, s9, v0
-; GFX1232-FAKE16-NEXT: v_bfe_u32 v2, v0, 16, 1
+; GFX1232-FAKE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(SKIP_2) | instid1(VALU_DEP_3)
+; GFX1232-FAKE16-NEXT: v_bfe_u32 v1, v0, 16, 1
; GFX1232-FAKE16-NEXT: v_or_b32_e32 v3, 0x400000, v0
; GFX1232-FAKE16-NEXT: v_cmp_u_f32_e32 vcc_lo, v0, v0
-; GFX1232-FAKE16-NEXT: s_delay_alu instid0(VALU_DEP_3) | instskip(SKIP_1) | instid1(VALU_DEP_1)
-; GFX1232-FAKE16-NEXT: v_add3_u32 v2, v2, v0, 0x7fff
+; GFX1232-FAKE16-NEXT: v_add3_u32 v1, v1, v0, 0x7fff
; GFX1232-FAKE16-NEXT: s_wait_alu 0xfffd
-; GFX1232-FAKE16-NEXT: v_cndmask_b32_e32 v0, v2, v3, vcc_lo
; GFX1232-FAKE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
+; GFX1232-FAKE16-NEXT: v_cndmask_b32_e32 v0, v1, v3, vcc_lo
; GFX1232-FAKE16-NEXT: v_lshrrev_b32_e32 v0, 16, v0
-; GFX1232-FAKE16-NEXT: v_lshlrev_b32_e32 v0, s2, v0
; GFX1232-FAKE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
-; GFX1232-FAKE16-NEXT: v_and_or_b32 v0, v1, s8, v0
-; GFX1232-FAKE16-NEXT: v_dual_mov_b32 v3, v1 :: v_dual_mov_b32 v2, v0
-; GFX1232-FAKE16-NEXT: buffer_atomic_cmpswap_b32 v[2:3], off, s[4:7], null th:TH_ATOMIC_RETURN scope:SCOPE_SYS
-; GFX1232-FAKE16-NEXT: s_wait_loadcnt 0x0
-; GFX1232-FAKE16-NEXT: v_cmp_eq_u32_e32 vcc_lo, v2, v1
+; GFX1232-FAKE16-NEXT: v_lshlrev_b32_e32 v0, s2, v0
+; GFX1232-FAKE16-NEXT: v_and_or_b32 v1, v2, s8, v0
+; GFX1232-FAKE16-NEXT: s_delay_alu instid0(VALU_DEP_1)
+; GFX1232-FAKE16-NEXT: v_mov_b32_e32 v0, v1
; GFX1232-FAKE16-NEXT: v_mov_b32_e32 v1, v2
+; GFX1232-FAKE16-NEXT: buffer_atomic_cmpswap_b32 v[0:1], off, s[4:7], null th:TH_ATOMIC_RETURN scope:SCOPE_SYS
+; GFX1232-FAKE16-NEXT: s_wait_loadcnt 0x0
+; GFX1232-FAKE16-NEXT: v_cmp_eq_u32_e32 vcc_lo, v0, v2
; GFX1232-FAKE16-NEXT: s_or_b32 s3, vcc_lo, s3
; GFX1232-FAKE16-NEXT: s_wait_alu 0xfffe
; GFX1232-FAKE16-NEXT: s_and_not1_b32 exec_lo, exec_lo, s3
; GFX1232-FAKE16-NEXT: s_cbranch_execnz .LBB19_1
; GFX1232-FAKE16-NEXT: ; %bb.2: ; %atomicrmw.end
; GFX1232-FAKE16-NEXT: s_or_b32 exec_lo, exec_lo, s3
-; GFX1232-FAKE16-NEXT: v_lshrrev_b32_e32 v0, s2, v2
+; GFX1232-FAKE16-NEXT: v_lshrrev_b32_e32 v0, s2, v0
; GFX1232-FAKE16-NEXT: s_mov_b32 s3, 0x31016000
; GFX1232-FAKE16-NEXT: s_mov_b32 s2, -1
; GFX1232-FAKE16-NEXT: buffer_store_b16 v0, off, s[0:3], null
@@ -11857,28 +11907,28 @@ define amdgpu_kernel void @uniform_fadd_v2f16(ptr addrspace(1) %result, ptr addr
; GFX8-NEXT: s_lshr_b32 s11, s10, 16
; GFX8-NEXT: s_mov_b32 s4, s2
; GFX8-NEXT: s_waitcnt lgkmcnt(0)
-; GFX8-NEXT: v_mov_b32_e32 v1, s5
+; GFX8-NEXT: v_mov_b32_e32 v0, s5
; GFX8-NEXT: s_mov_b32 s5, s3
; GFX8-NEXT: .LBB20_1: ; %atomicrmw.start
; GFX8-NEXT: ; =>This Inner Loop Header: Depth=1
-; GFX8-NEXT: v_mov_b32_e32 v0, s11
-; GFX8-NEXT: v_add_f16_e32 v2, s10, v1
-; GFX8-NEXT: v_add_f16_sdwa v0, v1, v0 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:WORD_1 src1_sel:DWORD
-; GFX8-NEXT: v_or_b32_e32 v0, v2, v0
-; GFX8-NEXT: v_mov_b32_e32 v3, v1
+; GFX8-NEXT: v_mov_b32_e32 v1, s11
; GFX8-NEXT: v_mov_b32_e32 v2, v0
-; GFX8-NEXT: buffer_atomic_cmpswap v[2:3], off, s[4:7], 0 glc
+; GFX8-NEXT: v_add_f16_sdwa v0, v2, v1 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:WORD_1 src1_sel:DWORD
+; GFX8-NEXT: v_add_f16_e32 v1, s10, v2
+; GFX8-NEXT: v_or_b32_e32 v1, v1, v0
+; GFX8-NEXT: v_mov_b32_e32 v0, v1
+; GFX8-NEXT: v_mov_b32_e32 v1, v2
+; GFX8-NEXT: buffer_atomic_cmpswap v[0:1], off, s[4:7], 0 glc
; GFX8-NEXT: s_waitcnt vmcnt(0)
-; GFX8-NEXT: v_cmp_eq_u32_e32 vcc, v2, v1
+; GFX8-NEXT: v_cmp_eq_u32_e32 vcc, v0, v2
; GFX8-NEXT: s_or_b64 s[8:9], vcc, s[8:9]
-; GFX8-NEXT: v_mov_b32_e32 v1, v2
; GFX8-NEXT: s_andn2_b64 exec, exec, s[8:9]
; GFX8-NEXT: s_cbranch_execnz .LBB20_1
; GFX8-NEXT: ; %bb.2: ; %atomicrmw.end
; GFX8-NEXT: s_or_b64 exec, exec, s[8:9]
; GFX8-NEXT: s_mov_b32 s3, 0xf000
; GFX8-NEXT: s_mov_b32 s2, -1
-; GFX8-NEXT: buffer_store_dword v2, off, s[0:3], 0
+; GFX8-NEXT: buffer_store_dword v0, off, s[0:3], 0
; GFX8-NEXT: s_endpgm
;
; GFX9-LABEL: uniform_fadd_v2f16:
@@ -11892,25 +11942,25 @@ define amdgpu_kernel void @uniform_fadd_v2f16(ptr addrspace(1) %result, ptr addr
; GFX9-NEXT: s_load_dword s5, s[2:3], 0x0
; GFX9-NEXT: s_mov_b32 s4, s2
; GFX9-NEXT: s_waitcnt lgkmcnt(0)
-; GFX9-NEXT: v_mov_b32_e32 v1, s5
+; GFX9-NEXT: v_mov_b32_e32 v0, s5
; GFX9-NEXT: s_mov_b32 s5, s3
; GFX9-NEXT: .LBB20_1: ; %atomicrmw.start
; GFX9-NEXT: ; =>This Inner Loop Header: Depth=1
-; GFX9-NEXT: v_pk_add_f16 v0, v1, s10
-; GFX9-NEXT: v_mov_b32_e32 v3, v1
; GFX9-NEXT: v_mov_b32_e32 v2, v0
-; GFX9-NEXT: buffer_atomic_cmpswap v[2:3], off, s[4:7], 0 glc
+; GFX9-NEXT: v_pk_add_f16 v1, v2, s10
+; GFX9-NEXT: v_mov_b32_e32 v0, v1
+; GFX9-NEXT: v_mov_b32_e32 v1, v2
+; GFX9-NEXT: buffer_atomic_cmpswap v[0:1], off, s[4:7], 0 glc
; GFX9-NEXT: s_waitcnt vmcnt(0)
-; GFX9-NEXT: v_cmp_eq_u32_e32 vcc, v2, v1
+; GFX9-NEXT: v_cmp_eq_u32_e32 vcc, v0, v2
; GFX9-NEXT: s_or_b64 s[8:9], vcc, s[8:9]
-; GFX9-NEXT: v_mov_b32_e32 v1, v2
; GFX9-NEXT: s_andn2_b64 exec, exec, s[8:9]
; GFX9-NEXT: s_cbranch_execnz .LBB20_1
; GFX9-NEXT: ; %bb.2: ; %atomicrmw.end
; GFX9-NEXT: s_or_b64 exec, exec, s[8:9]
; GFX9-NEXT: s_mov_b32 s3, 0xf000
; GFX9-NEXT: s_mov_b32 s2, -1
-; GFX9-NEXT: buffer_store_dword v2, off, s[0:3], 0
+; GFX9-NEXT: buffer_store_dword v0, off, s[0:3], 0
; GFX9-NEXT: s_endpgm
;
; GFX1064-LABEL: uniform_fadd_v2f16:
@@ -11925,17 +11975,17 @@ define amdgpu_kernel void @uniform_fadd_v2f16(ptr addrspace(1) %result, ptr addr
; GFX1064-NEXT: s_load_dword s4, s[2:3], 0x0
; GFX1064-NEXT: s_mov_b32 s5, s3
; GFX1064-NEXT: s_waitcnt lgkmcnt(0)
-; GFX1064-NEXT: v_mov_b32_e32 v1, s4
+; GFX1064-NEXT: v_mov_b32_e32 v0, s4
; GFX1064-NEXT: s_mov_b32 s4, s2
; GFX1064-NEXT: .LBB20_1: ; %atomicrmw.start
; GFX1064-NEXT: ; =>This Inner Loop Header: Depth=1
-; GFX1064-NEXT: v_pk_add_f16 v0, v1, s10
-; GFX1064-NEXT: v_mov_b32_e32 v3, v1
; GFX1064-NEXT: v_mov_b32_e32 v2, v0
-; GFX1064-NEXT: buffer_atomic_cmpswap v[2:3], off, s[4:7], 0 glc
-; GFX1064-NEXT: s_waitcnt vmcnt(0)
-; GFX1064-NEXT: v_cmp_eq_u32_e32 vcc, v2, v1
+; GFX1064-NEXT: v_pk_add_f16 v1, v2, s10
+; GFX1064-NEXT: v_mov_b32_e32 v0, v1
; GFX1064-NEXT: v_mov_b32_e32 v1, v2
+; GFX1064-NEXT: buffer_atomic_cmpswap v[0:1], off, s[4:7], 0 glc
+; GFX1064-NEXT: s_waitcnt vmcnt(0)
+; GFX1064-NEXT: v_cmp_eq_u32_e32 vcc, v0, v2
; GFX1064-NEXT: s_or_b64 s[8:9], vcc, s[8:9]
; GFX1064-NEXT: s_andn2_b64 exec, exec, s[8:9]
; GFX1064-NEXT: s_cbranch_execnz .LBB20_1
@@ -11943,7 +11993,7 @@ define amdgpu_kernel void @uniform_fadd_v2f16(ptr addrspace(1) %result, ptr addr
; GFX1064-NEXT: s_or_b64 exec, exec, s[8:9]
; GFX1064-NEXT: s_mov_b32 s3, 0x31016000
; GFX1064-NEXT: s_mov_b32 s2, -1
-; GFX1064-NEXT: buffer_store_dword v2, off, s[0:3], 0
+; GFX1064-NEXT: buffer_store_dword v0, off, s[0:3], 0
; GFX1064-NEXT: s_endpgm
;
; GFX1032-LABEL: uniform_fadd_v2f16:
@@ -11958,17 +12008,17 @@ define amdgpu_kernel void @uniform_fadd_v2f16(ptr addrspace(1) %result, ptr addr
; GFX1032-NEXT: s_load_dword s4, s[2:3], 0x0
; GFX1032-NEXT: s_mov_b32 s5, s3
; GFX1032-NEXT: s_waitcnt lgkmcnt(0)
-; GFX1032-NEXT: v_mov_b32_e32 v1, s4
+; GFX1032-NEXT: v_mov_b32_e32 v0, s4
; GFX1032-NEXT: s_mov_b32 s4, s2
; GFX1032-NEXT: .LBB20_1: ; %atomicrmw.start
; GFX1032-NEXT: ; =>This Inner Loop Header: Depth=1
-; GFX1032-NEXT: v_pk_add_f16 v0, v1, s8
-; GFX1032-NEXT: v_mov_b32_e32 v3, v1
; GFX1032-NEXT: v_mov_b32_e32 v2, v0
-; GFX1032-NEXT: buffer_atomic_cmpswap v[2:3], off, s[4:7], 0 glc
-; GFX1032-NEXT: s_waitcnt vmcnt(0)
-; GFX1032-NEXT: v_cmp_eq_u32_e32 vcc_lo, v2, v1
+; GFX1032-NEXT: v_pk_add_f16 v1, v2, s8
+; GFX1032-NEXT: v_mov_b32_e32 v0, v1
; GFX1032-NEXT: v_mov_b32_e32 v1, v2
+; GFX1032-NEXT: buffer_atomic_cmpswap v[0:1], off, s[4:7], 0 glc
+; GFX1032-NEXT: s_waitcnt vmcnt(0)
+; GFX1032-NEXT: v_cmp_eq_u32_e32 vcc_lo, v0, v2
; GFX1032-NEXT: s_or_b32 s9, vcc_lo, s9
; GFX1032-NEXT: s_andn2_b32 exec_lo, exec_lo, s9
; GFX1032-NEXT: s_cbranch_execnz .LBB20_1
@@ -11976,7 +12026,7 @@ define amdgpu_kernel void @uniform_fadd_v2f16(ptr addrspace(1) %result, ptr addr
; GFX1032-NEXT: s_or_b32 exec_lo, exec_lo, s9
; GFX1032-NEXT: s_mov_b32 s3, 0x31016000
; GFX1032-NEXT: s_mov_b32 s2, -1
-; GFX1032-NEXT: buffer_store_dword v2, off, s[0:3], 0
+; GFX1032-NEXT: buffer_store_dword v0, off, s[0:3], 0
; GFX1032-NEXT: s_endpgm
;
; GFX1164-LABEL: uniform_fadd_v2f16:
@@ -11991,18 +12041,19 @@ define amdgpu_kernel void @uniform_fadd_v2f16(ptr addrspace(1) %result, ptr addr
; GFX1164-NEXT: s_load_b32 s4, s[2:3], 0x0
; GFX1164-NEXT: s_mov_b32 s5, s3
; GFX1164-NEXT: s_waitcnt lgkmcnt(0)
-; GFX1164-NEXT: v_mov_b32_e32 v1, s4
+; GFX1164-NEXT: v_mov_b32_e32 v0, s4
; GFX1164-NEXT: s_mov_b32 s4, s2
; GFX1164-NEXT: .LBB20_1: ; %atomicrmw.start
; GFX1164-NEXT: ; =>This Inner Loop Header: Depth=1
-; GFX1164-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(SKIP_1) | instid1(VALU_DEP_2)
-; GFX1164-NEXT: v_pk_add_f16 v0, v1, s10
-; GFX1164-NEXT: v_mov_b32_e32 v3, v1
+; GFX1164-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
; GFX1164-NEXT: v_mov_b32_e32 v2, v0
-; GFX1164-NEXT: buffer_atomic_cmpswap_b32 v[2:3], off, s[4:7], 0 glc
-; GFX1164-NEXT: s_waitcnt vmcnt(0)
-; GFX1164-NEXT: v_cmp_eq_u32_e32 vcc, v2, v1
+; GFX1164-NEXT: v_pk_add_f16 v1, v2, s10
+; GFX1164-NEXT: s_delay_alu instid0(VALU_DEP_1)
+; GFX1164-NEXT: v_mov_b32_e32 v0, v1
; GFX1164-NEXT: v_mov_b32_e32 v1, v2
+; GFX1164-NEXT: buffer_atomic_cmpswap_b32 v[0:1], off, s[4:7], 0 glc
+; GFX1164-NEXT: s_waitcnt vmcnt(0)
+; GFX1164-NEXT: v_cmp_eq_u32_e32 vcc, v0, v2
; GFX1164-NEXT: s_or_b64 s[8:9], vcc, s[8:9]
; GFX1164-NEXT: s_delay_alu instid0(SALU_CYCLE_1)
; GFX1164-NEXT: s_and_not1_b64 exec, exec, s[8:9]
@@ -12011,7 +12062,7 @@ define amdgpu_kernel void @uniform_fadd_v2f16(ptr addrspace(1) %result, ptr addr
; GFX1164-NEXT: s_or_b64 exec, exec, s[8:9]
; GFX1164-NEXT: s_mov_b32 s3, 0x31016000
; GFX1164-NEXT: s_mov_b32 s2, -1
-; GFX1164-NEXT: buffer_store_b32 v2, off, s[0:3], 0
+; GFX1164-NEXT: buffer_store_b32 v0, off, s[0:3], 0
; GFX1164-NEXT: s_endpgm
;
; GFX1132-LABEL: uniform_fadd_v2f16:
@@ -12026,17 +12077,19 @@ define amdgpu_kernel void @uniform_fadd_v2f16(ptr addrspace(1) %result, ptr addr
; GFX1132-NEXT: s_load_b32 s4, s[2:3], 0x0
; GFX1132-NEXT: s_mov_b32 s5, s3
; GFX1132-NEXT: s_waitcnt lgkmcnt(0)
-; GFX1132-NEXT: v_mov_b32_e32 v1, s4
+; GFX1132-NEXT: v_mov_b32_e32 v0, s4
; GFX1132-NEXT: s_mov_b32 s4, s2
; GFX1132-NEXT: .LBB20_1: ; %atomicrmw.start
; GFX1132-NEXT: ; =>This Inner Loop Header: Depth=1
; GFX1132-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
-; GFX1132-NEXT: v_pk_add_f16 v0, v1, s8
-; GFX1132-NEXT: v_dual_mov_b32 v3, v1 :: v_dual_mov_b32 v2, v0
-; GFX1132-NEXT: buffer_atomic_cmpswap_b32 v[2:3], off, s[4:7], 0 glc
-; GFX1132-NEXT: s_waitcnt vmcnt(0)
-; GFX1132-NEXT: v_cmp_eq_u32_e32 vcc_lo, v2, v1
+; GFX1132-NEXT: v_mov_b32_e32 v2, v0
+; GFX1132-NEXT: v_pk_add_f16 v1, v2, s8
+; GFX1132-NEXT: s_delay_alu instid0(VALU_DEP_1)
+; GFX1132-NEXT: v_mov_b32_e32 v0, v1
; GFX1132-NEXT: v_mov_b32_e32 v1, v2
+; GFX1132-NEXT: buffer_atomic_cmpswap_b32 v[0:1], off, s[4:7], 0 glc
+; GFX1132-NEXT: s_waitcnt vmcnt(0)
+; GFX1132-NEXT: v_cmp_eq_u32_e32 vcc_lo, v0, v2
; GFX1132-NEXT: s_or_b32 s9, vcc_lo, s9
; GFX1132-NEXT: s_delay_alu instid0(SALU_CYCLE_1)
; GFX1132-NEXT: s_and_not1_b32 exec_lo, exec_lo, s9
@@ -12045,7 +12098,7 @@ define amdgpu_kernel void @uniform_fadd_v2f16(ptr addrspace(1) %result, ptr addr
; GFX1132-NEXT: s_or_b32 exec_lo, exec_lo, s9
; GFX1132-NEXT: s_mov_b32 s3, 0x31016000
; GFX1132-NEXT: s_mov_b32 s2, -1
-; GFX1132-NEXT: buffer_store_b32 v2, off, s[0:3], 0
+; GFX1132-NEXT: buffer_store_b32 v0, off, s[0:3], 0
; GFX1132-NEXT: s_endpgm
;
; GFX1264-LABEL: uniform_fadd_v2f16:
@@ -12060,18 +12113,19 @@ define amdgpu_kernel void @uniform_fadd_v2f16(ptr addrspace(1) %result, ptr addr
; GFX1264-NEXT: s_load_b32 s4, s[2:3], 0x0
; GFX1264-NEXT: s_mov_b32 s5, s3
; GFX1264-NEXT: s_wait_kmcnt 0x0
-; GFX1264-NEXT: v_mov_b32_e32 v1, s4
+; GFX1264-NEXT: v_mov_b32_e32 v0, s4
; GFX1264-NEXT: s_mov_b32 s4, s2
; GFX1264-NEXT: .LBB20_1: ; %atomicrmw.start
; GFX1264-NEXT: ; =>This Inner Loop Header: Depth=1
-; GFX1264-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(SKIP_1) | instid1(VALU_DEP_2)
-; GFX1264-NEXT: v_pk_add_f16 v0, v1, s10
-; GFX1264-NEXT: v_mov_b32_e32 v3, v1
+; GFX1264-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
; GFX1264-NEXT: v_mov_b32_e32 v2, v0
-; GFX1264-NEXT: buffer_atomic_cmpswap_b32 v[2:3], off, s[4:7], null th:TH_ATOMIC_RETURN scope:SCOPE_SYS
-; GFX1264-NEXT: s_wait_loadcnt 0x0
-; GFX1264-NEXT: v_cmp_eq_u32_e32 vcc, v2, v1
+; GFX1264-NEXT: v_pk_add_f16 v1, v2, s10
+; GFX1264-NEXT: s_delay_alu instid0(VALU_DEP_1)
+; GFX1264-NEXT: v_mov_b32_e32 v0, v1
; GFX1264-NEXT: v_mov_b32_e32 v1, v2
+; GFX1264-NEXT: buffer_atomic_cmpswap_b32 v[0:1], off, s[4:7], null th:TH_ATOMIC_RETURN scope:SCOPE_SYS
+; GFX1264-NEXT: s_wait_loadcnt 0x0
+; GFX1264-NEXT: v_cmp_eq_u32_e32 vcc, v0, v2
; GFX1264-NEXT: s_or_b64 s[8:9], vcc, s[8:9]
; GFX1264-NEXT: s_delay_alu instid0(SALU_CYCLE_1)
; GFX1264-NEXT: s_and_not1_b64 exec, exec, s[8:9]
@@ -12080,7 +12134,7 @@ define amdgpu_kernel void @uniform_fadd_v2f16(ptr addrspace(1) %result, ptr addr
; GFX1264-NEXT: s_or_b64 exec, exec, s[8:9]
; GFX1264-NEXT: s_mov_b32 s3, 0x31016000
; GFX1264-NEXT: s_mov_b32 s2, -1
-; GFX1264-NEXT: buffer_store_b32 v2, off, s[0:3], null
+; GFX1264-NEXT: buffer_store_b32 v0, off, s[0:3], null
; GFX1264-NEXT: s_endpgm
;
; GFX1232-LABEL: uniform_fadd_v2f16:
@@ -12095,17 +12149,19 @@ define amdgpu_kernel void @uniform_fadd_v2f16(ptr addrspace(1) %result, ptr addr
; GFX1232-NEXT: s_load_b32 s4, s[2:3], 0x0
; GFX1232-NEXT: s_mov_b32 s5, s3
; GFX1232-NEXT: s_wait_kmcnt 0x0
-; GFX1232-NEXT: v_mov_b32_e32 v1, s4
+; GFX1232-NEXT: v_mov_b32_e32 v0, s4
; GFX1232-NEXT: s_mov_b32 s4, s2
; GFX1232-NEXT: .LBB20_1: ; %atomicrmw.start
; GFX1232-NEXT: ; =>This Inner Loop Header: Depth=1
; GFX1232-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
-; GFX1232-NEXT: v_pk_add_f16 v0, v1, s8
-; GFX1232-NEXT: v_dual_mov_b32 v3, v1 :: v_dual_mov_b32 v2, v0
-; GFX1232-NEXT: buffer_atomic_cmpswap_b32 v[2:3], off, s[4:7], null th:TH_ATOMIC_RETURN scope:SCOPE_SYS
-; GFX1232-NEXT: s_wait_loadcnt 0x0
-; GFX1232-NEXT: v_cmp_eq_u32_e32 vcc_lo, v2, v1
+; GFX1232-NEXT: v_mov_b32_e32 v2, v0
+; GFX1232-NEXT: v_pk_add_f16 v1, v2, s8
+; GFX1232-NEXT: s_delay_alu instid0(VALU_DEP_1)
+; GFX1232-NEXT: v_mov_b32_e32 v0, v1
; GFX1232-NEXT: v_mov_b32_e32 v1, v2
+; GFX1232-NEXT: buffer_atomic_cmpswap_b32 v[0:1], off, s[4:7], null th:TH_ATOMIC_RETURN scope:SCOPE_SYS
+; GFX1232-NEXT: s_wait_loadcnt 0x0
+; GFX1232-NEXT: v_cmp_eq_u32_e32 vcc_lo, v0, v2
; GFX1232-NEXT: s_or_b32 s9, vcc_lo, s9
; GFX1232-NEXT: s_wait_alu 0xfffe
; GFX1232-NEXT: s_and_not1_b32 exec_lo, exec_lo, s9
@@ -12114,7 +12170,7 @@ define amdgpu_kernel void @uniform_fadd_v2f16(ptr addrspace(1) %result, ptr addr
; GFX1232-NEXT: s_or_b32 exec_lo, exec_lo, s9
; GFX1232-NEXT: s_mov_b32 s3, 0x31016000
; GFX1232-NEXT: s_mov_b32 s2, -1
-; GFX1232-NEXT: buffer_store_b32 v2, off, s[0:3], null
+; GFX1232-NEXT: buffer_store_b32 v0, off, s[0:3], null
; GFX1232-NEXT: s_endpgm
%rmw = atomicrmw fadd ptr addrspace(1) %uniform.ptr, <2 x half> %val monotonic, align 4
store <2 x half> %rmw, ptr addrspace(1) %result
@@ -12192,41 +12248,41 @@ define amdgpu_kernel void @uniform_fadd_v2bf16(ptr addrspace(1) %result, ptr add
; GFX8-NEXT: s_mov_b32 s4, s10
; GFX8-NEXT: s_mov_b32 s5, s11
; GFX8-NEXT: s_waitcnt lgkmcnt(0)
-; GFX8-NEXT: v_mov_b32_e32 v1, s1
+; GFX8-NEXT: v_mov_b32_e32 v0, s1
; GFX8-NEXT: .LBB21_1: ; %atomicrmw.start
; GFX8-NEXT: ; =>This Inner Loop Header: Depth=1
-; GFX8-NEXT: v_lshlrev_b32_e32 v0, 16, v1
-; GFX8-NEXT: v_and_b32_e32 v2, 0xffff0000, v1
+; GFX8-NEXT: v_mov_b32_e32 v2, v0
+; GFX8-NEXT: v_lshlrev_b32_e32 v0, 16, v2
+; GFX8-NEXT: v_and_b32_e32 v1, 0xffff0000, v2
; GFX8-NEXT: v_add_f32_e32 v0, s12, v0
-; GFX8-NEXT: v_add_f32_e32 v2, s13, v2
+; GFX8-NEXT: v_add_f32_e32 v1, s13, v1
; GFX8-NEXT: v_bfe_u32 v3, v0, 16, 1
-; GFX8-NEXT: v_bfe_u32 v5, v2, 16, 1
+; GFX8-NEXT: v_bfe_u32 v5, v1, 16, 1
; GFX8-NEXT: v_add_u32_e32 v3, vcc, v3, v0
-; GFX8-NEXT: v_add_u32_e32 v5, vcc, v5, v2
+; GFX8-NEXT: v_add_u32_e32 v5, vcc, v5, v1
; GFX8-NEXT: v_add_u32_e32 v3, vcc, 0x7fff, v3
; GFX8-NEXT: v_add_u32_e32 v5, vcc, 0x7fff, v5
-; GFX8-NEXT: v_or_b32_e32 v6, 0x400000, v2
-; GFX8-NEXT: v_cmp_u_f32_e32 vcc, v2, v2
+; GFX8-NEXT: v_or_b32_e32 v6, 0x400000, v1
+; GFX8-NEXT: v_cmp_u_f32_e32 vcc, v1, v1
; GFX8-NEXT: v_or_b32_e32 v4, 0x400000, v0
; GFX8-NEXT: v_cmp_u_f32_e64 s[0:1], v0, v0
-; GFX8-NEXT: v_cndmask_b32_e32 v2, v5, v6, vcc
+; GFX8-NEXT: v_cndmask_b32_e32 v1, v5, v6, vcc
; GFX8-NEXT: v_cndmask_b32_e64 v0, v3, v4, s[0:1]
-; GFX8-NEXT: v_lshrrev_b32_e32 v2, 16, v2
-; GFX8-NEXT: v_alignbit_b32 v0, v2, v0, 16
-; GFX8-NEXT: v_mov_b32_e32 v3, v1
-; GFX8-NEXT: v_mov_b32_e32 v2, v0
-; GFX8-NEXT: buffer_atomic_cmpswap v[2:3], off, s[4:7], 0 glc
+; GFX8-NEXT: v_lshrrev_b32_e32 v1, 16, v1
+; GFX8-NEXT: v_alignbit_b32 v1, v1, v0, 16
+; GFX8-NEXT: v_mov_b32_e32 v0, v1
+; GFX8-NEXT: v_mov_b32_e32 v1, v2
+; GFX8-NEXT: buffer_atomic_cmpswap v[0:1], off, s[4:7], 0 glc
; GFX8-NEXT: s_waitcnt vmcnt(0)
-; GFX8-NEXT: v_cmp_eq_u32_e32 vcc, v2, v1
+; GFX8-NEXT: v_cmp_eq_u32_e32 vcc, v0, v2
; GFX8-NEXT: s_or_b64 s[2:3], vcc, s[2:3]
-; GFX8-NEXT: v_mov_b32_e32 v1, v2
; GFX8-NEXT: s_andn2_b64 exec, exec, s[2:3]
; GFX8-NEXT: s_cbranch_execnz .LBB21_1
; GFX8-NEXT: ; %bb.2: ; %atomicrmw.end
; GFX8-NEXT: s_or_b64 exec, exec, s[2:3]
; GFX8-NEXT: s_mov_b32 s11, 0xf000
; GFX8-NEXT: s_mov_b32 s10, -1
-; GFX8-NEXT: buffer_store_dword v2, off, s[8:11], 0
+; GFX8-NEXT: buffer_store_dword v0, off, s[8:11], 0
; GFX8-NEXT: s_endpgm
;
; GFX9-LABEL: uniform_fadd_v2bf16:
@@ -12243,40 +12299,40 @@ define amdgpu_kernel void @uniform_fadd_v2bf16(ptr addrspace(1) %result, ptr add
; GFX9-NEXT: s_lshl_b32 s14, s0, 16
; GFX9-NEXT: s_and_b32 s15, s0, 0xffff0000
; GFX9-NEXT: s_waitcnt lgkmcnt(0)
-; GFX9-NEXT: v_mov_b32_e32 v1, s1
+; GFX9-NEXT: v_mov_b32_e32 v0, s1
; GFX9-NEXT: s_mov_b32 s4, s10
; GFX9-NEXT: s_mov_b32 s5, s11
; GFX9-NEXT: .LBB21_1: ; %atomicrmw.start
; GFX9-NEXT: ; =>This Inner Loop Header: Depth=1
-; GFX9-NEXT: v_lshlrev_b32_e32 v0, 16, v1
-; GFX9-NEXT: v_and_b32_e32 v2, 0xffff0000, v1
+; GFX9-NEXT: v_mov_b32_e32 v2, v0
+; GFX9-NEXT: v_lshlrev_b32_e32 v0, 16, v2
+; GFX9-NEXT: v_and_b32_e32 v1, 0xffff0000, v2
; GFX9-NEXT: v_add_f32_e32 v0, s14, v0
-; GFX9-NEXT: v_add_f32_e32 v2, s15, v2
+; GFX9-NEXT: v_add_f32_e32 v1, s15, v1
; GFX9-NEXT: v_bfe_u32 v3, v0, 16, 1
-; GFX9-NEXT: v_bfe_u32 v5, v2, 16, 1
+; GFX9-NEXT: v_bfe_u32 v5, v1, 16, 1
; GFX9-NEXT: v_or_b32_e32 v4, 0x400000, v0
-; GFX9-NEXT: v_or_b32_e32 v6, 0x400000, v2
+; GFX9-NEXT: v_or_b32_e32 v6, 0x400000, v1
; GFX9-NEXT: v_add3_u32 v3, v3, v0, s12
-; GFX9-NEXT: v_add3_u32 v5, v5, v2, s12
-; GFX9-NEXT: v_cmp_u_f32_e32 vcc, v2, v2
+; GFX9-NEXT: v_add3_u32 v5, v5, v1, s12
+; GFX9-NEXT: v_cmp_u_f32_e32 vcc, v1, v1
; GFX9-NEXT: v_cmp_u_f32_e64 s[0:1], v0, v0
; GFX9-NEXT: v_cndmask_b32_e64 v0, v3, v4, s[0:1]
-; GFX9-NEXT: v_cndmask_b32_e32 v2, v5, v6, vcc
-; GFX9-NEXT: v_perm_b32 v0, v2, v0, s13
-; GFX9-NEXT: v_mov_b32_e32 v3, v1
-; GFX9-NEXT: v_mov_b32_e32 v2, v0
-; GFX9-NEXT: buffer_atomic_cmpswap v[2:3], off, s[4:7], 0 glc
+; GFX9-NEXT: v_cndmask_b32_e32 v1, v5, v6, vcc
+; GFX9-NEXT: v_perm_b32 v1, v1, v0, s13
+; GFX9-NEXT: v_mov_b32_e32 v0, v1
+; GFX9-NEXT: v_mov_b32_e32 v1, v2
+; GFX9-NEXT: buffer_atomic_cmpswap v[0:1], off, s[4:7], 0 glc
; GFX9-NEXT: s_waitcnt vmcnt(0)
-; GFX9-NEXT: v_cmp_eq_u32_e32 vcc, v2, v1
+; GFX9-NEXT: v_cmp_eq_u32_e32 vcc, v0, v2
; GFX9-NEXT: s_or_b64 s[2:3], vcc, s[2:3]
-; GFX9-NEXT: v_mov_b32_e32 v1, v2
; GFX9-NEXT: s_andn2_b64 exec, exec, s[2:3]
; GFX9-NEXT: s_cbranch_execnz .LBB21_1
; GFX9-NEXT: ; %bb.2: ; %atomicrmw.end
; GFX9-NEXT: s_or_b64 exec, exec, s[2:3]
; GFX9-NEXT: s_mov_b32 s11, 0xf000
; GFX9-NEXT: s_mov_b32 s10, -1
-; GFX9-NEXT: buffer_store_dword v2, off, s[8:11], 0
+; GFX9-NEXT: buffer_store_dword v0, off, s[8:11], 0
; GFX9-NEXT: s_endpgm
;
; GFX1064-LABEL: uniform_fadd_v2bf16:
@@ -12294,30 +12350,30 @@ define amdgpu_kernel void @uniform_fadd_v2bf16(ptr addrspace(1) %result, ptr add
; GFX1064-NEXT: s_mov_b32 s4, s10
; GFX1064-NEXT: s_mov_b32 s5, s11
; GFX1064-NEXT: s_waitcnt lgkmcnt(0)
-; GFX1064-NEXT: v_mov_b32_e32 v1, s1
+; GFX1064-NEXT: v_mov_b32_e32 v0, s1
; GFX1064-NEXT: .LBB21_1: ; %atomicrmw.start
; GFX1064-NEXT: ; =>This Inner Loop Header: Depth=1
-; GFX1064-NEXT: v_lshlrev_b32_e32 v0, 16, v1
-; GFX1064-NEXT: v_and_b32_e32 v2, 0xffff0000, v1
+; GFX1064-NEXT: v_mov_b32_e32 v2, v0
+; GFX1064-NEXT: v_lshlrev_b32_e32 v0, 16, v2
+; GFX1064-NEXT: v_and_b32_e32 v1, 0xffff0000, v2
; GFX1064-NEXT: v_add_f32_e32 v0, s12, v0
-; GFX1064-NEXT: v_add_f32_e32 v2, s13, v2
+; GFX1064-NEXT: v_add_f32_e32 v1, s13, v1
; GFX1064-NEXT: v_bfe_u32 v3, v0, 16, 1
-; GFX1064-NEXT: v_bfe_u32 v4, v2, 16, 1
+; GFX1064-NEXT: v_bfe_u32 v4, v1, 16, 1
; GFX1064-NEXT: v_or_b32_e32 v5, 0x400000, v0
-; GFX1064-NEXT: v_or_b32_e32 v6, 0x400000, v2
-; GFX1064-NEXT: v_cmp_u_f32_e32 vcc, v2, v2
+; GFX1064-NEXT: v_or_b32_e32 v6, 0x400000, v1
+; GFX1064-NEXT: v_cmp_u_f32_e32 vcc, v1, v1
; GFX1064-NEXT: v_add3_u32 v3, v3, v0, 0x7fff
-; GFX1064-NEXT: v_add3_u32 v4, v4, v2, 0x7fff
+; GFX1064-NEXT: v_add3_u32 v4, v4, v1, 0x7fff
; GFX1064-NEXT: v_cmp_u_f32_e64 s[0:1], v0, v0
-; GFX1064-NEXT: v_cndmask_b32_e32 v2, v4, v6, vcc
+; GFX1064-NEXT: v_cndmask_b32_e32 v1, v4, v6, vcc
; GFX1064-NEXT: v_cndmask_b32_e64 v0, v3, v5, s[0:1]
-; GFX1064-NEXT: v_perm_b32 v0, v2, v0, 0x7060302
-; GFX1064-NEXT: v_mov_b32_e32 v3, v1
-; GFX1064-NEXT: v_mov_b32_e32 v2, v0
-; GFX1064-NEXT: buffer_atomic_cmpswap v[2:3], off, s[4:7], 0 glc
-; GFX1064-NEXT: s_waitcnt vmcnt(0)
-; GFX1064-NEXT: v_cmp_eq_u32_e32 vcc, v2, v1
+; GFX1064-NEXT: v_perm_b32 v1, v1, v0, 0x7060302
+; GFX1064-NEXT: v_mov_b32_e32 v0, v1
; GFX1064-NEXT: v_mov_b32_e32 v1, v2
+; GFX1064-NEXT: buffer_atomic_cmpswap v[0:1], off, s[4:7], 0 glc
+; GFX1064-NEXT: s_waitcnt vmcnt(0)
+; GFX1064-NEXT: v_cmp_eq_u32_e32 vcc, v0, v2
; GFX1064-NEXT: s_or_b64 s[2:3], vcc, s[2:3]
; GFX1064-NEXT: s_andn2_b64 exec, exec, s[2:3]
; GFX1064-NEXT: s_cbranch_execnz .LBB21_1
@@ -12325,7 +12381,7 @@ define amdgpu_kernel void @uniform_fadd_v2bf16(ptr addrspace(1) %result, ptr add
; GFX1064-NEXT: s_or_b64 exec, exec, s[2:3]
; GFX1064-NEXT: s_mov_b32 s11, 0x31016000
; GFX1064-NEXT: s_mov_b32 s10, -1
-; GFX1064-NEXT: buffer_store_dword v2, off, s[8:11], 0
+; GFX1064-NEXT: buffer_store_dword v0, off, s[8:11], 0
; GFX1064-NEXT: s_endpgm
;
; GFX1032-LABEL: uniform_fadd_v2bf16:
@@ -12342,31 +12398,31 @@ define amdgpu_kernel void @uniform_fadd_v2bf16(ptr addrspace(1) %result, ptr add
; GFX1032-NEXT: s_and_b32 s3, s0, 0xffff0000
; GFX1032-NEXT: s_mov_b32 s5, s11
; GFX1032-NEXT: s_waitcnt lgkmcnt(0)
-; GFX1032-NEXT: v_mov_b32_e32 v1, s4
+; GFX1032-NEXT: v_mov_b32_e32 v0, s4
; GFX1032-NEXT: s_mov_b32 s4, s10
; GFX1032-NEXT: .LBB21_1: ; %atomicrmw.start
; GFX1032-NEXT: ; =>This Inner Loop Header: Depth=1
-; GFX1032-NEXT: v_lshlrev_b32_e32 v0, 16, v1
-; GFX1032-NEXT: v_and_b32_e32 v2, 0xffff0000, v1
+; GFX1032-NEXT: v_mov_b32_e32 v2, v0
+; GFX1032-NEXT: v_lshlrev_b32_e32 v0, 16, v2
+; GFX1032-NEXT: v_and_b32_e32 v1, 0xffff0000, v2
; GFX1032-NEXT: v_add_f32_e32 v0, s2, v0
-; GFX1032-NEXT: v_add_f32_e32 v2, s3, v2
+; GFX1032-NEXT: v_add_f32_e32 v1, s3, v1
; GFX1032-NEXT: v_bfe_u32 v3, v0, 16, 1
-; GFX1032-NEXT: v_bfe_u32 v4, v2, 16, 1
+; GFX1032-NEXT: v_bfe_u32 v4, v1, 16, 1
; GFX1032-NEXT: v_or_b32_e32 v5, 0x400000, v0
-; GFX1032-NEXT: v_or_b32_e32 v6, 0x400000, v2
-; GFX1032-NEXT: v_cmp_u_f32_e32 vcc_lo, v2, v2
+; GFX1032-NEXT: v_or_b32_e32 v6, 0x400000, v1
+; GFX1032-NEXT: v_cmp_u_f32_e32 vcc_lo, v1, v1
; GFX1032-NEXT: v_add3_u32 v3, v3, v0, 0x7fff
-; GFX1032-NEXT: v_add3_u32 v4, v4, v2, 0x7fff
+; GFX1032-NEXT: v_add3_u32 v4, v4, v1, 0x7fff
; GFX1032-NEXT: v_cmp_u_f32_e64 s0, v0, v0
-; GFX1032-NEXT: v_cndmask_b32_e32 v2, v4, v6, vcc_lo
+; GFX1032-NEXT: v_cndmask_b32_e32 v1, v4, v6, vcc_lo
; GFX1032-NEXT: v_cndmask_b32_e64 v0, v3, v5, s0
-; GFX1032-NEXT: v_perm_b32 v0, v2, v0, 0x7060302
-; GFX1032-NEXT: v_mov_b32_e32 v3, v1
-; GFX1032-NEXT: v_mov_b32_e32 v2, v0
-; GFX1032-NEXT: buffer_atomic_cmpswap v[2:3], off, s[4:7], 0 glc
-; GFX1032-NEXT: s_waitcnt vmcnt(0)
-; GFX1032-NEXT: v_cmp_eq_u32_e32 vcc_lo, v2, v1
+; GFX1032-NEXT: v_perm_b32 v1, v1, v0, 0x7060302
+; GFX1032-NEXT: v_mov_b32_e32 v0, v1
; GFX1032-NEXT: v_mov_b32_e32 v1, v2
+; GFX1032-NEXT: buffer_atomic_cmpswap v[0:1], off, s[4:7], 0 glc
+; GFX1032-NEXT: s_waitcnt vmcnt(0)
+; GFX1032-NEXT: v_cmp_eq_u32_e32 vcc_lo, v0, v2
; GFX1032-NEXT: s_or_b32 s1, vcc_lo, s1
; GFX1032-NEXT: s_andn2_b32 exec_lo, exec_lo, s1
; GFX1032-NEXT: s_cbranch_execnz .LBB21_1
@@ -12374,7 +12430,7 @@ define amdgpu_kernel void @uniform_fadd_v2bf16(ptr addrspace(1) %result, ptr add
; GFX1032-NEXT: s_or_b32 exec_lo, exec_lo, s1
; GFX1032-NEXT: s_mov_b32 s11, 0x31016000
; GFX1032-NEXT: s_mov_b32 s10, -1
-; GFX1032-NEXT: buffer_store_dword v2, off, s[8:11], 0
+; GFX1032-NEXT: buffer_store_dword v0, off, s[8:11], 0
; GFX1032-NEXT: s_endpgm
;
; GFX1164-TRUE16-LABEL: uniform_fadd_v2bf16:
@@ -12391,40 +12447,40 @@ define amdgpu_kernel void @uniform_fadd_v2bf16(ptr addrspace(1) %result, ptr add
; GFX1164-TRUE16-NEXT: s_lshl_b32 s11, s4, 16
; GFX1164-TRUE16-NEXT: s_mov_b32 s4, s2
; GFX1164-TRUE16-NEXT: s_waitcnt lgkmcnt(0)
-; GFX1164-TRUE16-NEXT: v_mov_b32_e32 v1, s5
+; GFX1164-TRUE16-NEXT: v_mov_b32_e32 v0, s5
; GFX1164-TRUE16-NEXT: s_mov_b32 s5, s3
; GFX1164-TRUE16-NEXT: s_set_inst_prefetch_distance 0x1
; GFX1164-TRUE16-NEXT: .p2align 6
; GFX1164-TRUE16-NEXT: .LBB21_1: ; %atomicrmw.start
; GFX1164-TRUE16-NEXT: ; =>This Inner Loop Header: Depth=1
-; GFX1164-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(SKIP_1) | instid1(VALU_DEP_2)
-; GFX1164-TRUE16-NEXT: v_lshlrev_b32_e32 v0, 16, v1
-; GFX1164-TRUE16-NEXT: v_and_b32_e32 v2, 0xffff0000, v1
+; GFX1164-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
+; GFX1164-TRUE16-NEXT: v_mov_b32_e32 v2, v0
+; GFX1164-TRUE16-NEXT: v_lshlrev_b32_e32 v0, 16, v2
+; GFX1164-TRUE16-NEXT: v_and_b32_e32 v1, 0xffff0000, v2
+; GFX1164-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_2) | instskip(NEXT) | instid1(VALU_DEP_2)
; GFX1164-TRUE16-NEXT: v_add_f32_e32 v0, s11, v0
+; GFX1164-TRUE16-NEXT: v_add_f32_e32 v1, s10, v1
; GFX1164-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_2) | instskip(NEXT) | instid1(VALU_DEP_2)
-; GFX1164-TRUE16-NEXT: v_add_f32_e32 v2, s10, v2
; GFX1164-TRUE16-NEXT: v_bfe_u32 v3, v0, 16, 1
-; GFX1164-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_2)
-; GFX1164-TRUE16-NEXT: v_bfe_u32 v4, v2, 16, 1
+; GFX1164-TRUE16-NEXT: v_bfe_u32 v4, v1, 16, 1
; GFX1164-TRUE16-NEXT: v_or_b32_e32 v5, 0x400000, v0
; GFX1164-TRUE16-NEXT: v_cmp_u_f32_e32 vcc, v0, v0
-; GFX1164-TRUE16-NEXT: v_or_b32_e32 v6, 0x400000, v2
+; GFX1164-TRUE16-NEXT: v_or_b32_e32 v6, 0x400000, v1
; GFX1164-TRUE16-NEXT: v_add3_u32 v3, v3, v0, 0x7fff
-; GFX1164-TRUE16-NEXT: v_add3_u32 v4, v4, v2, 0x7fff
+; GFX1164-TRUE16-NEXT: v_add3_u32 v4, v4, v1, 0x7fff
; GFX1164-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_2) | instskip(SKIP_1) | instid1(VALU_DEP_2)
; GFX1164-TRUE16-NEXT: v_cndmask_b32_e32 v0, v3, v5, vcc
-; GFX1164-TRUE16-NEXT: v_cmp_u_f32_e32 vcc, v2, v2
+; GFX1164-TRUE16-NEXT: v_cmp_u_f32_e32 vcc, v1, v1
; GFX1164-TRUE16-NEXT: v_mov_b16_e32 v0.l, v0.h
; GFX1164-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_4) | instskip(NEXT) | instid1(VALU_DEP_1)
-; GFX1164-TRUE16-NEXT: v_cndmask_b32_e32 v2, v4, v6, vcc
-; GFX1164-TRUE16-NEXT: v_bfi_b32 v0, 0xffff, v0, v2
-; GFX1164-TRUE16-NEXT: v_mov_b32_e32 v3, v1
-; GFX1164-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_2)
-; GFX1164-TRUE16-NEXT: v_mov_b32_e32 v2, v0
-; GFX1164-TRUE16-NEXT: buffer_atomic_cmpswap_b32 v[2:3], off, s[4:7], 0 glc
-; GFX1164-TRUE16-NEXT: s_waitcnt vmcnt(0)
-; GFX1164-TRUE16-NEXT: v_cmp_eq_u32_e32 vcc, v2, v1
+; GFX1164-TRUE16-NEXT: v_cndmask_b32_e32 v1, v4, v6, vcc
+; GFX1164-TRUE16-NEXT: v_bfi_b32 v1, 0xffff, v0, v1
+; GFX1164-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1)
+; GFX1164-TRUE16-NEXT: v_mov_b32_e32 v0, v1
; GFX1164-TRUE16-NEXT: v_mov_b32_e32 v1, v2
+; GFX1164-TRUE16-NEXT: buffer_atomic_cmpswap_b32 v[0:1], off, s[4:7], 0 glc
+; GFX1164-TRUE16-NEXT: s_waitcnt vmcnt(0)
+; GFX1164-TRUE16-NEXT: v_cmp_eq_u32_e32 vcc, v0, v2
; GFX1164-TRUE16-NEXT: s_or_b64 s[8:9], vcc, s[8:9]
; GFX1164-TRUE16-NEXT: s_delay_alu instid0(SALU_CYCLE_1)
; GFX1164-TRUE16-NEXT: s_and_not1_b64 exec, exec, s[8:9]
@@ -12434,7 +12490,7 @@ define amdgpu_kernel void @uniform_fadd_v2bf16(ptr addrspace(1) %result, ptr add
; GFX1164-TRUE16-NEXT: s_or_b64 exec, exec, s[8:9]
; GFX1164-TRUE16-NEXT: s_mov_b32 s3, 0x31016000
; GFX1164-TRUE16-NEXT: s_mov_b32 s2, -1
-; GFX1164-TRUE16-NEXT: buffer_store_b32 v2, off, s[0:3], 0
+; GFX1164-TRUE16-NEXT: buffer_store_b32 v0, off, s[0:3], 0
; GFX1164-TRUE16-NEXT: s_endpgm
;
; GFX1164-FAKE16-LABEL: uniform_fadd_v2bf16:
@@ -12452,37 +12508,37 @@ define amdgpu_kernel void @uniform_fadd_v2bf16(ptr addrspace(1) %result, ptr add
; GFX1164-FAKE16-NEXT: s_mov_b32 s4, s10
; GFX1164-FAKE16-NEXT: s_mov_b32 s5, s11
; GFX1164-FAKE16-NEXT: s_waitcnt lgkmcnt(0)
-; GFX1164-FAKE16-NEXT: v_mov_b32_e32 v1, s1
+; GFX1164-FAKE16-NEXT: v_mov_b32_e32 v0, s1
; GFX1164-FAKE16-NEXT: s_set_inst_prefetch_distance 0x1
; GFX1164-FAKE16-NEXT: .p2align 6
; GFX1164-FAKE16-NEXT: .LBB21_1: ; %atomicrmw.start
; GFX1164-FAKE16-NEXT: ; =>This Inner Loop Header: Depth=1
-; GFX1164-FAKE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(SKIP_1) | instid1(VALU_DEP_2)
-; GFX1164-FAKE16-NEXT: v_lshlrev_b32_e32 v0, 16, v1
-; GFX1164-FAKE16-NEXT: v_and_b32_e32 v2, 0xffff0000, v1
+; GFX1164-FAKE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
+; GFX1164-FAKE16-NEXT: v_mov_b32_e32 v2, v0
+; GFX1164-FAKE16-NEXT: v_lshlrev_b32_e32 v0, 16, v2
+; GFX1164-FAKE16-NEXT: v_and_b32_e32 v1, 0xffff0000, v2
+; GFX1164-FAKE16-NEXT: s_delay_alu instid0(VALU_DEP_2) | instskip(NEXT) | instid1(VALU_DEP_2)
; GFX1164-FAKE16-NEXT: v_add_f32_e32 v0, s12, v0
+; GFX1164-FAKE16-NEXT: v_add_f32_e32 v1, s13, v1
; GFX1164-FAKE16-NEXT: s_delay_alu instid0(VALU_DEP_2) | instskip(NEXT) | instid1(VALU_DEP_2)
-; GFX1164-FAKE16-NEXT: v_add_f32_e32 v2, s13, v2
; GFX1164-FAKE16-NEXT: v_bfe_u32 v3, v0, 16, 1
-; GFX1164-FAKE16-NEXT: s_delay_alu instid0(VALU_DEP_2)
-; GFX1164-FAKE16-NEXT: v_bfe_u32 v4, v2, 16, 1
+; GFX1164-FAKE16-NEXT: v_bfe_u32 v4, v1, 16, 1
; GFX1164-FAKE16-NEXT: v_or_b32_e32 v5, 0x400000, v0
-; GFX1164-FAKE16-NEXT: v_or_b32_e32 v6, 0x400000, v2
-; GFX1164-FAKE16-NEXT: v_cmp_u_f32_e32 vcc, v2, v2
+; GFX1164-FAKE16-NEXT: v_or_b32_e32 v6, 0x400000, v1
+; GFX1164-FAKE16-NEXT: v_cmp_u_f32_e32 vcc, v1, v1
; GFX1164-FAKE16-NEXT: v_add3_u32 v3, v3, v0, 0x7fff
-; GFX1164-FAKE16-NEXT: v_add3_u32 v4, v4, v2, 0x7fff
+; GFX1164-FAKE16-NEXT: v_add3_u32 v4, v4, v1, 0x7fff
; GFX1164-FAKE16-NEXT: v_cmp_u_f32_e64 s[0:1], v0, v0
; GFX1164-FAKE16-NEXT: s_delay_alu instid0(VALU_DEP_2) | instskip(NEXT) | instid1(VALU_DEP_2)
-; GFX1164-FAKE16-NEXT: v_cndmask_b32_e32 v2, v4, v6, vcc
+; GFX1164-FAKE16-NEXT: v_cndmask_b32_e32 v1, v4, v6, vcc
; GFX1164-FAKE16-NEXT: v_cndmask_b32_e64 v0, v3, v5, s[0:1]
-; GFX1164-FAKE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(SKIP_1) | instid1(VALU_DEP_2)
-; GFX1164-FAKE16-NEXT: v_perm_b32 v0, v2, v0, 0x7060302
-; GFX1164-FAKE16-NEXT: v_mov_b32_e32 v3, v1
-; GFX1164-FAKE16-NEXT: v_mov_b32_e32 v2, v0
-; GFX1164-FAKE16-NEXT: buffer_atomic_cmpswap_b32 v[2:3], off, s[4:7], 0 glc
-; GFX1164-FAKE16-NEXT: s_waitcnt vmcnt(0)
-; GFX1164-FAKE16-NEXT: v_cmp_eq_u32_e32 vcc, v2, v1
+; GFX1164-FAKE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
+; GFX1164-FAKE16-NEXT: v_perm_b32 v1, v1, v0, 0x7060302
+; GFX1164-FAKE16-NEXT: v_mov_b32_e32 v0, v1
; GFX1164-FAKE16-NEXT: v_mov_b32_e32 v1, v2
+; GFX1164-FAKE16-NEXT: buffer_atomic_cmpswap_b32 v[0:1], off, s[4:7], 0 glc
+; GFX1164-FAKE16-NEXT: s_waitcnt vmcnt(0)
+; GFX1164-FAKE16-NEXT: v_cmp_eq_u32_e32 vcc, v0, v2
; GFX1164-FAKE16-NEXT: s_or_b64 s[2:3], vcc, s[2:3]
; GFX1164-FAKE16-NEXT: s_delay_alu instid0(SALU_CYCLE_1)
; GFX1164-FAKE16-NEXT: s_and_not1_b64 exec, exec, s[2:3]
@@ -12492,7 +12548,7 @@ define amdgpu_kernel void @uniform_fadd_v2bf16(ptr addrspace(1) %result, ptr add
; GFX1164-FAKE16-NEXT: s_or_b64 exec, exec, s[2:3]
; GFX1164-FAKE16-NEXT: s_mov_b32 s11, 0x31016000
; GFX1164-FAKE16-NEXT: s_mov_b32 s10, -1
-; GFX1164-FAKE16-NEXT: buffer_store_b32 v2, off, s[8:11], 0
+; GFX1164-FAKE16-NEXT: buffer_store_b32 v0, off, s[8:11], 0
; GFX1164-FAKE16-NEXT: s_endpgm
;
; GFX1132-TRUE16-LABEL: uniform_fadd_v2bf16:
@@ -12509,39 +12565,39 @@ define amdgpu_kernel void @uniform_fadd_v2bf16(ptr addrspace(1) %result, ptr add
; GFX1132-TRUE16-NEXT: s_lshl_b32 s10, s4, 16
; GFX1132-TRUE16-NEXT: s_mov_b32 s4, s2
; GFX1132-TRUE16-NEXT: s_waitcnt lgkmcnt(0)
-; GFX1132-TRUE16-NEXT: v_mov_b32_e32 v1, s5
+; GFX1132-TRUE16-NEXT: v_mov_b32_e32 v0, s5
; GFX1132-TRUE16-NEXT: s_mov_b32 s5, s3
; GFX1132-TRUE16-NEXT: s_set_inst_prefetch_distance 0x1
; GFX1132-TRUE16-NEXT: .p2align 6
; GFX1132-TRUE16-NEXT: .LBB21_1: ; %atomicrmw.start
; GFX1132-TRUE16-NEXT: ; =>This Inner Loop Header: Depth=1
-; GFX1132-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(SKIP_1) | instid1(VALU_DEP_2)
-; GFX1132-TRUE16-NEXT: v_lshlrev_b32_e32 v0, 16, v1
-; GFX1132-TRUE16-NEXT: v_and_b32_e32 v2, 0xffff0000, v1
+; GFX1132-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
+; GFX1132-TRUE16-NEXT: v_mov_b32_e32 v2, v0
+; GFX1132-TRUE16-NEXT: v_and_b32_e32 v1, 0xffff0000, v2
+; GFX1132-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
+; GFX1132-TRUE16-NEXT: v_dual_add_f32 v1, s9, v1 :: v_dual_lshlrev_b32 v0, 16, v2
+; GFX1132-TRUE16-NEXT: v_bfe_u32 v4, v1, 16, 1
+; GFX1132-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_2) | instskip(SKIP_1) | instid1(VALU_DEP_3)
; GFX1132-TRUE16-NEXT: v_add_f32_e32 v0, s10, v0
-; GFX1132-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_2) | instskip(NEXT) | instid1(VALU_DEP_2)
-; GFX1132-TRUE16-NEXT: v_add_f32_e32 v2, s9, v2
+; GFX1132-TRUE16-NEXT: v_or_b32_e32 v6, 0x400000, v1
+; GFX1132-TRUE16-NEXT: v_add3_u32 v4, v4, v1, 0x7fff
+; GFX1132-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_3) | instskip(SKIP_2) | instid1(VALU_DEP_3)
; GFX1132-TRUE16-NEXT: v_bfe_u32 v3, v0, 16, 1
-; GFX1132-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_2)
-; GFX1132-TRUE16-NEXT: v_bfe_u32 v4, v2, 16, 1
; GFX1132-TRUE16-NEXT: v_or_b32_e32 v5, 0x400000, v0
; GFX1132-TRUE16-NEXT: v_cmp_u_f32_e32 vcc_lo, v0, v0
-; GFX1132-TRUE16-NEXT: v_or_b32_e32 v6, 0x400000, v2
; GFX1132-TRUE16-NEXT: v_add3_u32 v3, v3, v0, 0x7fff
-; GFX1132-TRUE16-NEXT: v_add3_u32 v4, v4, v2, 0x7fff
-; GFX1132-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_2) | instskip(SKIP_1) | instid1(VALU_DEP_2)
+; GFX1132-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(SKIP_2) | instid1(VALU_DEP_3)
; GFX1132-TRUE16-NEXT: v_cndmask_b32_e32 v0, v3, v5, vcc_lo
-; GFX1132-TRUE16-NEXT: v_cmp_u_f32_e32 vcc_lo, v2, v2
+; GFX1132-TRUE16-NEXT: v_cmp_u_f32_e32 vcc_lo, v1, v1
+; GFX1132-TRUE16-NEXT: v_cndmask_b32_e32 v1, v4, v6, vcc_lo
; GFX1132-TRUE16-NEXT: v_mov_b16_e32 v0.l, v0.h
-; GFX1132-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_4) | instskip(NEXT) | instid1(VALU_DEP_1)
-; GFX1132-TRUE16-NEXT: v_cndmask_b32_e32 v2, v4, v6, vcc_lo
-; GFX1132-TRUE16-NEXT: v_bfi_b32 v0, 0xffff, v0, v2
-; GFX1132-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1)
-; GFX1132-TRUE16-NEXT: v_dual_mov_b32 v3, v1 :: v_dual_mov_b32 v2, v0
-; GFX1132-TRUE16-NEXT: buffer_atomic_cmpswap_b32 v[2:3], off, s[4:7], 0 glc
-; GFX1132-TRUE16-NEXT: s_waitcnt vmcnt(0)
-; GFX1132-TRUE16-NEXT: v_cmp_eq_u32_e32 vcc_lo, v2, v1
+; GFX1132-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
+; GFX1132-TRUE16-NEXT: v_bfi_b32 v1, 0xffff, v0, v1
+; GFX1132-TRUE16-NEXT: v_mov_b32_e32 v0, v1
; GFX1132-TRUE16-NEXT: v_mov_b32_e32 v1, v2
+; GFX1132-TRUE16-NEXT: buffer_atomic_cmpswap_b32 v[0:1], off, s[4:7], 0 glc
+; GFX1132-TRUE16-NEXT: s_waitcnt vmcnt(0)
+; GFX1132-TRUE16-NEXT: v_cmp_eq_u32_e32 vcc_lo, v0, v2
; GFX1132-TRUE16-NEXT: s_or_b32 s8, vcc_lo, s8
; GFX1132-TRUE16-NEXT: s_delay_alu instid0(SALU_CYCLE_1)
; GFX1132-TRUE16-NEXT: s_and_not1_b32 exec_lo, exec_lo, s8
@@ -12551,7 +12607,7 @@ define amdgpu_kernel void @uniform_fadd_v2bf16(ptr addrspace(1) %result, ptr add
; GFX1132-TRUE16-NEXT: s_or_b32 exec_lo, exec_lo, s8
; GFX1132-TRUE16-NEXT: s_mov_b32 s3, 0x31016000
; GFX1132-TRUE16-NEXT: s_mov_b32 s2, -1
-; GFX1132-TRUE16-NEXT: buffer_store_b32 v2, off, s[0:3], 0
+; GFX1132-TRUE16-NEXT: buffer_store_b32 v0, off, s[0:3], 0
; GFX1132-TRUE16-NEXT: s_endpgm
;
; GFX1132-FAKE16-LABEL: uniform_fadd_v2bf16:
@@ -12568,37 +12624,38 @@ define amdgpu_kernel void @uniform_fadd_v2bf16(ptr addrspace(1) %result, ptr add
; GFX1132-FAKE16-NEXT: s_and_b32 s3, s0, 0xffff0000
; GFX1132-FAKE16-NEXT: s_mov_b32 s5, s11
; GFX1132-FAKE16-NEXT: s_waitcnt lgkmcnt(0)
-; GFX1132-FAKE16-NEXT: v_mov_b32_e32 v1, s4
+; GFX1132-FAKE16-NEXT: v_mov_b32_e32 v0, s4
; GFX1132-FAKE16-NEXT: s_mov_b32 s4, s10
; GFX1132-FAKE16-NEXT: s_set_inst_prefetch_distance 0x1
; GFX1132-FAKE16-NEXT: .p2align 6
; GFX1132-FAKE16-NEXT: .LBB21_1: ; %atomicrmw.start
; GFX1132-FAKE16-NEXT: ; =>This Inner Loop Header: Depth=1
-; GFX1132-FAKE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(SKIP_1) | instid1(VALU_DEP_2)
-; GFX1132-FAKE16-NEXT: v_lshlrev_b32_e32 v0, 16, v1
-; GFX1132-FAKE16-NEXT: v_and_b32_e32 v2, 0xffff0000, v1
+; GFX1132-FAKE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
+; GFX1132-FAKE16-NEXT: v_mov_b32_e32 v2, v0
+; GFX1132-FAKE16-NEXT: v_and_b32_e32 v1, 0xffff0000, v2
+; GFX1132-FAKE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
+; GFX1132-FAKE16-NEXT: v_dual_add_f32 v1, s3, v1 :: v_dual_lshlrev_b32 v0, 16, v2
+; GFX1132-FAKE16-NEXT: v_bfe_u32 v4, v1, 16, 1
+; GFX1132-FAKE16-NEXT: s_delay_alu instid0(VALU_DEP_2) | instskip(SKIP_2) | instid1(VALU_DEP_4)
; GFX1132-FAKE16-NEXT: v_add_f32_e32 v0, s2, v0
-; GFX1132-FAKE16-NEXT: s_delay_alu instid0(VALU_DEP_2) | instskip(NEXT) | instid1(VALU_DEP_2)
-; GFX1132-FAKE16-NEXT: v_add_f32_e32 v2, s3, v2
+; GFX1132-FAKE16-NEXT: v_or_b32_e32 v6, 0x400000, v1
+; GFX1132-FAKE16-NEXT: v_cmp_u_f32_e32 vcc_lo, v1, v1
+; GFX1132-FAKE16-NEXT: v_add3_u32 v4, v4, v1, 0x7fff
+; GFX1132-FAKE16-NEXT: s_delay_alu instid0(VALU_DEP_4) | instskip(SKIP_2) | instid1(VALU_DEP_4)
; GFX1132-FAKE16-NEXT: v_bfe_u32 v3, v0, 16, 1
-; GFX1132-FAKE16-NEXT: s_delay_alu instid0(VALU_DEP_2)
-; GFX1132-FAKE16-NEXT: v_bfe_u32 v4, v2, 16, 1
; GFX1132-FAKE16-NEXT: v_or_b32_e32 v5, 0x400000, v0
-; GFX1132-FAKE16-NEXT: v_or_b32_e32 v6, 0x400000, v2
-; GFX1132-FAKE16-NEXT: v_cmp_u_f32_e32 vcc_lo, v2, v2
-; GFX1132-FAKE16-NEXT: v_add3_u32 v3, v3, v0, 0x7fff
-; GFX1132-FAKE16-NEXT: v_add3_u32 v4, v4, v2, 0x7fff
; GFX1132-FAKE16-NEXT: v_cmp_u_f32_e64 s0, v0, v0
-; GFX1132-FAKE16-NEXT: s_delay_alu instid0(VALU_DEP_2) | instskip(NEXT) | instid1(VALU_DEP_2)
-; GFX1132-FAKE16-NEXT: v_cndmask_b32_e32 v2, v4, v6, vcc_lo
+; GFX1132-FAKE16-NEXT: v_cndmask_b32_e32 v1, v4, v6, vcc_lo
+; GFX1132-FAKE16-NEXT: s_delay_alu instid0(VALU_DEP_4) | instskip(NEXT) | instid1(VALU_DEP_1)
+; GFX1132-FAKE16-NEXT: v_add3_u32 v3, v3, v0, 0x7fff
; GFX1132-FAKE16-NEXT: v_cndmask_b32_e64 v0, v3, v5, s0
; GFX1132-FAKE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
-; GFX1132-FAKE16-NEXT: v_perm_b32 v0, v2, v0, 0x7060302
-; GFX1132-FAKE16-NEXT: v_dual_mov_b32 v3, v1 :: v_dual_mov_b32 v2, v0
-; GFX1132-FAKE16-NEXT: buffer_atomic_cmpswap_b32 v[2:3], off, s[4:7], 0 glc
-; GFX1132-FAKE16-NEXT: s_waitcnt vmcnt(0)
-; GFX1132-FAKE16-NEXT: v_cmp_eq_u32_e32 vcc_lo, v2, v1
+; GFX1132-FAKE16-NEXT: v_perm_b32 v1, v1, v0, 0x7060302
+; GFX1132-FAKE16-NEXT: v_mov_b32_e32 v0, v1
; GFX1132-FAKE16-NEXT: v_mov_b32_e32 v1, v2
+; GFX1132-FAKE16-NEXT: buffer_atomic_cmpswap_b32 v[0:1], off, s[4:7], 0 glc
+; GFX1132-FAKE16-NEXT: s_waitcnt vmcnt(0)
+; GFX1132-FAKE16-NEXT: v_cmp_eq_u32_e32 vcc_lo, v0, v2
; GFX1132-FAKE16-NEXT: s_or_b32 s1, vcc_lo, s1
; GFX1132-FAKE16-NEXT: s_delay_alu instid0(SALU_CYCLE_1)
; GFX1132-FAKE16-NEXT: s_and_not1_b32 exec_lo, exec_lo, s1
@@ -12608,7 +12665,7 @@ define amdgpu_kernel void @uniform_fadd_v2bf16(ptr addrspace(1) %result, ptr add
; GFX1132-FAKE16-NEXT: s_or_b32 exec_lo, exec_lo, s1
; GFX1132-FAKE16-NEXT: s_mov_b32 s11, 0x31016000
; GFX1132-FAKE16-NEXT: s_mov_b32 s10, -1
-; GFX1132-FAKE16-NEXT: buffer_store_b32 v2, off, s[8:11], 0
+; GFX1132-FAKE16-NEXT: buffer_store_b32 v0, off, s[8:11], 0
; GFX1132-FAKE16-NEXT: s_endpgm
;
; GFX1264-TRUE16-LABEL: uniform_fadd_v2bf16:
@@ -12625,39 +12682,39 @@ define amdgpu_kernel void @uniform_fadd_v2bf16(ptr addrspace(1) %result, ptr add
; GFX1264-TRUE16-NEXT: s_lshl_b32 s11, s4, 16
; GFX1264-TRUE16-NEXT: s_mov_b32 s4, s2
; GFX1264-TRUE16-NEXT: s_wait_kmcnt 0x0
-; GFX1264-TRUE16-NEXT: v_mov_b32_e32 v1, s5
+; GFX1264-TRUE16-NEXT: v_mov_b32_e32 v0, s5
; GFX1264-TRUE16-NEXT: s_mov_b32 s5, s3
; GFX1264-TRUE16-NEXT: .LBB21_1: ; %atomicrmw.start
; GFX1264-TRUE16-NEXT: ; =>This Inner Loop Header: Depth=1
-; GFX1264-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(SKIP_1) | instid1(VALU_DEP_2)
-; GFX1264-TRUE16-NEXT: v_lshlrev_b32_e32 v0, 16, v1
-; GFX1264-TRUE16-NEXT: v_and_b32_e32 v2, 0xffff0000, v1
+; GFX1264-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
+; GFX1264-TRUE16-NEXT: v_mov_b32_e32 v2, v0
+; GFX1264-TRUE16-NEXT: v_lshlrev_b32_e32 v0, 16, v2
+; GFX1264-TRUE16-NEXT: v_and_b32_e32 v1, 0xffff0000, v2
+; GFX1264-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_2) | instskip(NEXT) | instid1(VALU_DEP_2)
; GFX1264-TRUE16-NEXT: v_add_f32_e32 v0, s11, v0
+; GFX1264-TRUE16-NEXT: v_add_f32_e32 v1, s10, v1
; GFX1264-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_2) | instskip(NEXT) | instid1(VALU_DEP_2)
-; GFX1264-TRUE16-NEXT: v_add_f32_e32 v2, s10, v2
; GFX1264-TRUE16-NEXT: v_bfe_u32 v3, v0, 16, 1
-; GFX1264-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_2)
-; GFX1264-TRUE16-NEXT: v_bfe_u32 v4, v2, 16, 1
+; GFX1264-TRUE16-NEXT: v_bfe_u32 v4, v1, 16, 1
; GFX1264-TRUE16-NEXT: v_or_b32_e32 v5, 0x400000, v0
; GFX1264-TRUE16-NEXT: v_cmp_u_f32_e32 vcc, v0, v0
-; GFX1264-TRUE16-NEXT: v_or_b32_e32 v6, 0x400000, v2
+; GFX1264-TRUE16-NEXT: v_or_b32_e32 v6, 0x400000, v1
; GFX1264-TRUE16-NEXT: v_add3_u32 v3, v3, v0, 0x7fff
-; GFX1264-TRUE16-NEXT: v_add3_u32 v4, v4, v2, 0x7fff
+; GFX1264-TRUE16-NEXT: v_add3_u32 v4, v4, v1, 0x7fff
; GFX1264-TRUE16-NEXT: s_wait_alu 0xfffd
; GFX1264-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_2) | instskip(SKIP_1) | instid1(VALU_DEP_2)
; GFX1264-TRUE16-NEXT: v_cndmask_b32_e32 v0, v3, v5, vcc
-; GFX1264-TRUE16-NEXT: v_cmp_u_f32_e32 vcc, v2, v2
+; GFX1264-TRUE16-NEXT: v_cmp_u_f32_e32 vcc, v1, v1
; GFX1264-TRUE16-NEXT: v_mov_b16_e32 v0.l, v0.h
; GFX1264-TRUE16-NEXT: s_wait_alu 0xfffd
-; GFX1264-TRUE16-NEXT: v_cndmask_b32_e32 v2, v4, v6, vcc
-; GFX1264-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(SKIP_1) | instid1(VALU_DEP_2)
-; GFX1264-TRUE16-NEXT: v_bfi_b32 v0, 0xffff, v0, v2
-; GFX1264-TRUE16-NEXT: v_mov_b32_e32 v3, v1
-; GFX1264-TRUE16-NEXT: v_mov_b32_e32 v2, v0
-; GFX1264-TRUE16-NEXT: buffer_atomic_cmpswap_b32 v[2:3], off, s[4:7], null th:TH_ATOMIC_RETURN scope:SCOPE_SYS
-; GFX1264-TRUE16-NEXT: s_wait_loadcnt 0x0
-; GFX1264-TRUE16-NEXT: v_cmp_eq_u32_e32 vcc, v2, v1
+; GFX1264-TRUE16-NEXT: v_cndmask_b32_e32 v1, v4, v6, vcc
+; GFX1264-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
+; GFX1264-TRUE16-NEXT: v_bfi_b32 v1, 0xffff, v0, v1
+; GFX1264-TRUE16-NEXT: v_mov_b32_e32 v0, v1
; GFX1264-TRUE16-NEXT: v_mov_b32_e32 v1, v2
+; GFX1264-TRUE16-NEXT: buffer_atomic_cmpswap_b32 v[0:1], off, s[4:7], null th:TH_ATOMIC_RETURN scope:SCOPE_SYS
+; GFX1264-TRUE16-NEXT: s_wait_loadcnt 0x0
+; GFX1264-TRUE16-NEXT: v_cmp_eq_u32_e32 vcc, v0, v2
; GFX1264-TRUE16-NEXT: s_or_b64 s[8:9], vcc, s[8:9]
; GFX1264-TRUE16-NEXT: s_delay_alu instid0(SALU_CYCLE_1)
; GFX1264-TRUE16-NEXT: s_and_not1_b64 exec, exec, s[8:9]
@@ -12666,7 +12723,7 @@ define amdgpu_kernel void @uniform_fadd_v2bf16(ptr addrspace(1) %result, ptr add
; GFX1264-TRUE16-NEXT: s_or_b64 exec, exec, s[8:9]
; GFX1264-TRUE16-NEXT: s_mov_b32 s3, 0x31016000
; GFX1264-TRUE16-NEXT: s_mov_b32 s2, -1
-; GFX1264-TRUE16-NEXT: buffer_store_b32 v2, off, s[0:3], null
+; GFX1264-TRUE16-NEXT: buffer_store_b32 v0, off, s[0:3], null
; GFX1264-TRUE16-NEXT: s_endpgm
;
; GFX1264-FAKE16-LABEL: uniform_fadd_v2bf16:
@@ -12684,36 +12741,36 @@ define amdgpu_kernel void @uniform_fadd_v2bf16(ptr addrspace(1) %result, ptr add
; GFX1264-FAKE16-NEXT: s_mov_b32 s4, s10
; GFX1264-FAKE16-NEXT: s_mov_b32 s5, s11
; GFX1264-FAKE16-NEXT: s_wait_kmcnt 0x0
-; GFX1264-FAKE16-NEXT: v_mov_b32_e32 v1, s1
+; GFX1264-FAKE16-NEXT: v_mov_b32_e32 v0, s1
; GFX1264-FAKE16-NEXT: .LBB21_1: ; %atomicrmw.start
; GFX1264-FAKE16-NEXT: ; =>This Inner Loop Header: Depth=1
-; GFX1264-FAKE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(SKIP_1) | instid1(VALU_DEP_2)
-; GFX1264-FAKE16-NEXT: v_lshlrev_b32_e32 v0, 16, v1
-; GFX1264-FAKE16-NEXT: v_and_b32_e32 v2, 0xffff0000, v1
+; GFX1264-FAKE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
+; GFX1264-FAKE16-NEXT: v_mov_b32_e32 v2, v0
+; GFX1264-FAKE16-NEXT: v_lshlrev_b32_e32 v0, 16, v2
+; GFX1264-FAKE16-NEXT: v_and_b32_e32 v1, 0xffff0000, v2
+; GFX1264-FAKE16-NEXT: s_delay_alu instid0(VALU_DEP_2) | instskip(NEXT) | instid1(VALU_DEP_2)
; GFX1264-FAKE16-NEXT: v_add_f32_e32 v0, s12, v0
+; GFX1264-FAKE16-NEXT: v_add_f32_e32 v1, s13, v1
; GFX1264-FAKE16-NEXT: s_delay_alu instid0(VALU_DEP_2) | instskip(NEXT) | instid1(VALU_DEP_2)
-; GFX1264-FAKE16-NEXT: v_add_f32_e32 v2, s13, v2
; GFX1264-FAKE16-NEXT: v_bfe_u32 v3, v0, 16, 1
-; GFX1264-FAKE16-NEXT: s_delay_alu instid0(VALU_DEP_2)
-; GFX1264-FAKE16-NEXT: v_bfe_u32 v4, v2, 16, 1
+; GFX1264-FAKE16-NEXT: v_bfe_u32 v4, v1, 16, 1
; GFX1264-FAKE16-NEXT: v_or_b32_e32 v5, 0x400000, v0
-; GFX1264-FAKE16-NEXT: v_or_b32_e32 v6, 0x400000, v2
-; GFX1264-FAKE16-NEXT: v_cmp_u_f32_e32 vcc, v2, v2
+; GFX1264-FAKE16-NEXT: v_or_b32_e32 v6, 0x400000, v1
+; GFX1264-FAKE16-NEXT: v_cmp_u_f32_e32 vcc, v1, v1
; GFX1264-FAKE16-NEXT: v_add3_u32 v3, v3, v0, 0x7fff
-; GFX1264-FAKE16-NEXT: v_add3_u32 v4, v4, v2, 0x7fff
+; GFX1264-FAKE16-NEXT: v_add3_u32 v4, v4, v1, 0x7fff
; GFX1264-FAKE16-NEXT: v_cmp_u_f32_e64 s[0:1], v0, v0
; GFX1264-FAKE16-NEXT: s_wait_alu 0xfffd
-; GFX1264-FAKE16-NEXT: v_cndmask_b32_e32 v2, v4, v6, vcc
+; GFX1264-FAKE16-NEXT: v_cndmask_b32_e32 v1, v4, v6, vcc
; GFX1264-FAKE16-NEXT: s_wait_alu 0xf1ff
; GFX1264-FAKE16-NEXT: v_cndmask_b32_e64 v0, v3, v5, s[0:1]
-; GFX1264-FAKE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(SKIP_1) | instid1(VALU_DEP_2)
-; GFX1264-FAKE16-NEXT: v_perm_b32 v0, v2, v0, 0x7060302
-; GFX1264-FAKE16-NEXT: v_mov_b32_e32 v3, v1
-; GFX1264-FAKE16-NEXT: v_mov_b32_e32 v2, v0
-; GFX1264-FAKE16-NEXT: buffer_atomic_cmpswap_b32 v[2:3], off, s[4:7], null th:TH_ATOMIC_RETURN scope:SCOPE_SYS
-; GFX1264-FAKE16-NEXT: s_wait_loadcnt 0x0
-; GFX1264-FAKE16-NEXT: v_cmp_eq_u32_e32 vcc, v2, v1
+; GFX1264-FAKE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
+; GFX1264-FAKE16-NEXT: v_perm_b32 v1, v1, v0, 0x7060302
+; GFX1264-FAKE16-NEXT: v_mov_b32_e32 v0, v1
; GFX1264-FAKE16-NEXT: v_mov_b32_e32 v1, v2
+; GFX1264-FAKE16-NEXT: buffer_atomic_cmpswap_b32 v[0:1], off, s[4:7], null th:TH_ATOMIC_RETURN scope:SCOPE_SYS
+; GFX1264-FAKE16-NEXT: s_wait_loadcnt 0x0
+; GFX1264-FAKE16-NEXT: v_cmp_eq_u32_e32 vcc, v0, v2
; GFX1264-FAKE16-NEXT: s_or_b64 s[2:3], vcc, s[2:3]
; GFX1264-FAKE16-NEXT: s_delay_alu instid0(SALU_CYCLE_1)
; GFX1264-FAKE16-NEXT: s_and_not1_b64 exec, exec, s[2:3]
@@ -12722,7 +12779,7 @@ define amdgpu_kernel void @uniform_fadd_v2bf16(ptr addrspace(1) %result, ptr add
; GFX1264-FAKE16-NEXT: s_or_b64 exec, exec, s[2:3]
; GFX1264-FAKE16-NEXT: s_mov_b32 s11, 0x31016000
; GFX1264-FAKE16-NEXT: s_mov_b32 s10, -1
-; GFX1264-FAKE16-NEXT: buffer_store_b32 v2, off, s[8:11], null
+; GFX1264-FAKE16-NEXT: buffer_store_b32 v0, off, s[8:11], null
; GFX1264-FAKE16-NEXT: s_endpgm
;
; GFX1232-TRUE16-LABEL: uniform_fadd_v2bf16:
@@ -12739,38 +12796,39 @@ define amdgpu_kernel void @uniform_fadd_v2bf16(ptr addrspace(1) %result, ptr add
; GFX1232-TRUE16-NEXT: s_lshl_b32 s10, s4, 16
; GFX1232-TRUE16-NEXT: s_mov_b32 s4, s2
; GFX1232-TRUE16-NEXT: s_wait_kmcnt 0x0
-; GFX1232-TRUE16-NEXT: v_mov_b32_e32 v1, s5
+; GFX1232-TRUE16-NEXT: v_mov_b32_e32 v0, s5
; GFX1232-TRUE16-NEXT: s_mov_b32 s5, s3
; GFX1232-TRUE16-NEXT: .LBB21_1: ; %atomicrmw.start
; GFX1232-TRUE16-NEXT: ; =>This Inner Loop Header: Depth=1
-; GFX1232-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(SKIP_1) | instid1(VALU_DEP_2)
-; GFX1232-TRUE16-NEXT: v_lshlrev_b32_e32 v0, 16, v1
-; GFX1232-TRUE16-NEXT: v_and_b32_e32 v2, 0xffff0000, v1
+; GFX1232-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
+; GFX1232-TRUE16-NEXT: v_mov_b32_e32 v2, v0
+; GFX1232-TRUE16-NEXT: v_and_b32_e32 v1, 0xffff0000, v2
+; GFX1232-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
+; GFX1232-TRUE16-NEXT: v_dual_add_f32 v1, s9, v1 :: v_dual_lshlrev_b32 v0, 16, v2
+; GFX1232-TRUE16-NEXT: v_bfe_u32 v4, v1, 16, 1
+; GFX1232-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_2) | instskip(SKIP_1) | instid1(VALU_DEP_3)
; GFX1232-TRUE16-NEXT: v_add_f32_e32 v0, s10, v0
-; GFX1232-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_2) | instskip(NEXT) | instid1(VALU_DEP_2)
-; GFX1232-TRUE16-NEXT: v_add_f32_e32 v2, s9, v2
+; GFX1232-TRUE16-NEXT: v_or_b32_e32 v6, 0x400000, v1
+; GFX1232-TRUE16-NEXT: v_add3_u32 v4, v4, v1, 0x7fff
+; GFX1232-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_3) | instskip(SKIP_2) | instid1(VALU_DEP_3)
; GFX1232-TRUE16-NEXT: v_bfe_u32 v3, v0, 16, 1
-; GFX1232-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_2)
-; GFX1232-TRUE16-NEXT: v_bfe_u32 v4, v2, 16, 1
; GFX1232-TRUE16-NEXT: v_or_b32_e32 v5, 0x400000, v0
; GFX1232-TRUE16-NEXT: v_cmp_u_f32_e32 vcc_lo, v0, v0
-; GFX1232-TRUE16-NEXT: v_or_b32_e32 v6, 0x400000, v2
; GFX1232-TRUE16-NEXT: v_add3_u32 v3, v3, v0, 0x7fff
-; GFX1232-TRUE16-NEXT: v_add3_u32 v4, v4, v2, 0x7fff
; GFX1232-TRUE16-NEXT: s_wait_alu 0xfffd
-; GFX1232-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_2) | instskip(SKIP_1) | instid1(VALU_DEP_2)
+; GFX1232-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(SKIP_3) | instid1(VALU_DEP_3)
; GFX1232-TRUE16-NEXT: v_cndmask_b32_e32 v0, v3, v5, vcc_lo
-; GFX1232-TRUE16-NEXT: v_cmp_u_f32_e32 vcc_lo, v2, v2
-; GFX1232-TRUE16-NEXT: v_mov_b16_e32 v0.l, v0.h
+; GFX1232-TRUE16-NEXT: v_cmp_u_f32_e32 vcc_lo, v1, v1
; GFX1232-TRUE16-NEXT: s_wait_alu 0xfffd
-; GFX1232-TRUE16-NEXT: v_cndmask_b32_e32 v2, v4, v6, vcc_lo
+; GFX1232-TRUE16-NEXT: v_cndmask_b32_e32 v1, v4, v6, vcc_lo
+; GFX1232-TRUE16-NEXT: v_mov_b16_e32 v0.l, v0.h
; GFX1232-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
-; GFX1232-TRUE16-NEXT: v_bfi_b32 v0, 0xffff, v0, v2
-; GFX1232-TRUE16-NEXT: v_dual_mov_b32 v3, v1 :: v_dual_mov_b32 v2, v0
-; GFX1232-TRUE16-NEXT: buffer_atomic_cmpswap_b32 v[2:3], off, s[4:7], null th:TH_ATOMIC_RETURN scope:SCOPE_SYS
-; GFX1232-TRUE16-NEXT: s_wait_loadcnt 0x0
-; GFX1232-TRUE16-NEXT: v_cmp_eq_u32_e32 vcc_lo, v2, v1
+; GFX1232-TRUE16-NEXT: v_bfi_b32 v1, 0xffff, v0, v1
+; GFX1232-TRUE16-NEXT: v_mov_b32_e32 v0, v1
; GFX1232-TRUE16-NEXT: v_mov_b32_e32 v1, v2
+; GFX1232-TRUE16-NEXT: buffer_atomic_cmpswap_b32 v[0:1], off, s[4:7], null th:TH_ATOMIC_RETURN scope:SCOPE_SYS
+; GFX1232-TRUE16-NEXT: s_wait_loadcnt 0x0
+; GFX1232-TRUE16-NEXT: v_cmp_eq_u32_e32 vcc_lo, v0, v2
; GFX1232-TRUE16-NEXT: s_or_b32 s8, vcc_lo, s8
; GFX1232-TRUE16-NEXT: s_wait_alu 0xfffe
; GFX1232-TRUE16-NEXT: s_and_not1_b32 exec_lo, exec_lo, s8
@@ -12779,7 +12837,7 @@ define amdgpu_kernel void @uniform_fadd_v2bf16(ptr addrspace(1) %result, ptr add
; GFX1232-TRUE16-NEXT: s_or_b32 exec_lo, exec_lo, s8
; GFX1232-TRUE16-NEXT: s_mov_b32 s3, 0x31016000
; GFX1232-TRUE16-NEXT: s_mov_b32 s2, -1
-; GFX1232-TRUE16-NEXT: buffer_store_b32 v2, off, s[0:3], null
+; GFX1232-TRUE16-NEXT: buffer_store_b32 v0, off, s[0:3], null
; GFX1232-TRUE16-NEXT: s_endpgm
;
; GFX1232-FAKE16-LABEL: uniform_fadd_v2bf16:
@@ -12796,36 +12854,38 @@ define amdgpu_kernel void @uniform_fadd_v2bf16(ptr addrspace(1) %result, ptr add
; GFX1232-FAKE16-NEXT: s_and_b32 s3, s0, 0xffff0000
; GFX1232-FAKE16-NEXT: s_mov_b32 s5, s11
; GFX1232-FAKE16-NEXT: s_wait_kmcnt 0x0
-; GFX1232-FAKE16-NEXT: v_mov_b32_e32 v1, s4
+; GFX1232-FAKE16-NEXT: v_mov_b32_e32 v0, s4
; GFX1232-FAKE16-NEXT: s_mov_b32 s4, s10
; GFX1232-FAKE16-NEXT: .LBB21_1: ; %atomicrmw.start
; GFX1232-FAKE16-NEXT: ; =>This Inner Loop Header: Depth=1
-; GFX1232-FAKE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(SKIP_1) | instid1(VALU_DEP_2)
-; GFX1232-FAKE16-NEXT: v_lshlrev_b32_e32 v0, 16, v1
-; GFX1232-FAKE16-NEXT: v_and_b32_e32 v2, 0xffff0000, v1
+; GFX1232-FAKE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
+; GFX1232-FAKE16-NEXT: v_mov_b32_e32 v2, v0
+; GFX1232-FAKE16-NEXT: v_and_b32_e32 v1, 0xffff0000, v2
+; GFX1232-FAKE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
+; GFX1232-FAKE16-NEXT: v_dual_add_f32 v1, s3, v1 :: v_dual_lshlrev_b32 v0, 16, v2
+; GFX1232-FAKE16-NEXT: v_bfe_u32 v4, v1, 16, 1
+; GFX1232-FAKE16-NEXT: s_delay_alu instid0(VALU_DEP_2) | instskip(SKIP_2) | instid1(VALU_DEP_4)
; GFX1232-FAKE16-NEXT: v_add_f32_e32 v0, s2, v0
-; GFX1232-FAKE16-NEXT: s_delay_alu instid0(VALU_DEP_2) | instskip(NEXT) | instid1(VALU_DEP_2)
-; GFX1232-FAKE16-NEXT: v_add_f32_e32 v2, s3, v2
+; GFX1232-FAKE16-NEXT: v_or_b32_e32 v6, 0x400000, v1
+; GFX1232-FAKE16-NEXT: v_cmp_u_f32_e32 vcc_lo, v1, v1
+; GFX1232-FAKE16-NEXT: v_add3_u32 v4, v4, v1, 0x7fff
+; GFX1232-FAKE16-NEXT: s_delay_alu instid0(VALU_DEP_4)
; GFX1232-FAKE16-NEXT: v_bfe_u32 v3, v0, 16, 1
-; GFX1232-FAKE16-NEXT: s_delay_alu instid0(VALU_DEP_2)
-; GFX1232-FAKE16-NEXT: v_bfe_u32 v4, v2, 16, 1
; GFX1232-FAKE16-NEXT: v_or_b32_e32 v5, 0x400000, v0
-; GFX1232-FAKE16-NEXT: v_or_b32_e32 v6, 0x400000, v2
-; GFX1232-FAKE16-NEXT: v_cmp_u_f32_e32 vcc_lo, v2, v2
-; GFX1232-FAKE16-NEXT: v_add3_u32 v3, v3, v0, 0x7fff
-; GFX1232-FAKE16-NEXT: v_add3_u32 v4, v4, v2, 0x7fff
; GFX1232-FAKE16-NEXT: v_cmp_u_f32_e64 s0, v0, v0
; GFX1232-FAKE16-NEXT: s_wait_alu 0xfffd
-; GFX1232-FAKE16-NEXT: v_cndmask_b32_e32 v2, v4, v6, vcc_lo
+; GFX1232-FAKE16-NEXT: v_cndmask_b32_e32 v1, v4, v6, vcc_lo
+; GFX1232-FAKE16-NEXT: v_add3_u32 v3, v3, v0, 0x7fff
; GFX1232-FAKE16-NEXT: s_wait_alu 0xf1ff
-; GFX1232-FAKE16-NEXT: v_cndmask_b32_e64 v0, v3, v5, s0
; GFX1232-FAKE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
-; GFX1232-FAKE16-NEXT: v_perm_b32 v0, v2, v0, 0x7060302
-; GFX1232-FAKE16-NEXT: v_dual_mov_b32 v3, v1 :: v_dual_mov_b32 v2, v0
-; GFX1232-FAKE16-NEXT: buffer_atomic_cmpswap_b32 v[2:3], off, s[4:7], null th:TH_ATOMIC_RETURN scope:SCOPE_SYS
-; GFX1232-FAKE16-NEXT: s_wait_loadcnt 0x0
-; GFX1232-FAKE16-NEXT: v_cmp_eq_u32_e32 vcc_lo, v2, v1
+; GFX1232-FAKE16-NEXT: v_cndmask_b32_e64 v0, v3, v5, s0
+; GFX1232-FAKE16-NEXT: v_perm_b32 v1, v1, v0, 0x7060302
+; GFX1232-FAKE16-NEXT: s_delay_alu instid0(VALU_DEP_1)
+; GFX1232-FAKE16-NEXT: v_mov_b32_e32 v0, v1
; GFX1232-FAKE16-NEXT: v_mov_b32_e32 v1, v2
+; GFX1232-FAKE16-NEXT: buffer_atomic_cmpswap_b32 v[0:1], off, s[4:7], null th:TH_ATOMIC_RETURN scope:SCOPE_SYS
+; GFX1232-FAKE16-NEXT: s_wait_loadcnt 0x0
+; GFX1232-FAKE16-NEXT: v_cmp_eq_u32_e32 vcc_lo, v0, v2
; GFX1232-FAKE16-NEXT: s_or_b32 s1, vcc_lo, s1
; GFX1232-FAKE16-NEXT: s_wait_alu 0xfffe
; GFX1232-FAKE16-NEXT: s_and_not1_b32 exec_lo, exec_lo, s1
@@ -12834,7 +12894,7 @@ define amdgpu_kernel void @uniform_fadd_v2bf16(ptr addrspace(1) %result, ptr add
; GFX1232-FAKE16-NEXT: s_or_b32 exec_lo, exec_lo, s1
; GFX1232-FAKE16-NEXT: s_mov_b32 s11, 0x31016000
; GFX1232-FAKE16-NEXT: s_mov_b32 s10, -1
-; GFX1232-FAKE16-NEXT: buffer_store_b32 v2, off, s[8:11], null
+; GFX1232-FAKE16-NEXT: buffer_store_b32 v0, off, s[8:11], null
; GFX1232-FAKE16-NEXT: s_endpgm
%rmw = atomicrmw fadd ptr addrspace(1) %uniform.ptr, <2 x bfloat> %val monotonic, align 4
store <2 x bfloat> %rmw, ptr addrspace(1) %result
diff --git a/llvm/test/CodeGen/AMDGPU/branch-folding-implicit-def-subreg.ll b/llvm/test/CodeGen/AMDGPU/branch-folding-implicit-def-subreg.ll
index ae90cfb631e8d..509ba295ea7f7 100644
--- a/llvm/test/CodeGen/AMDGPU/branch-folding-implicit-def-subreg.ll
+++ b/llvm/test/CodeGen/AMDGPU/branch-folding-implicit-def-subreg.ll
@@ -32,12 +32,12 @@ define amdgpu_kernel void @f1(ptr addrspace(1) %arg, ptr addrspace(1) %arg1, i64
; GFX90A-NEXT: S_CBRANCH_VCCZ %bb.2, implicit $vcc
; GFX90A-NEXT: {{ $}}
; GFX90A-NEXT: bb.1.bb103:
- ; GFX90A-NEXT: successors: %bb.58(0x40000000), %bb.2(0x40000000)
+ ; GFX90A-NEXT: successors: %bb.59(0x40000000), %bb.2(0x40000000)
; GFX90A-NEXT: liveins: $sgpr14, $sgpr15, $sgpr16, $sgpr17, $sgpr33, $vgpr31, $sgpr4_sgpr5, $sgpr6_sgpr7, $sgpr8_sgpr9:0x000000000000000F, $sgpr10_sgpr11, $sgpr12_sgpr13, $sgpr18_sgpr19, $sgpr28_sgpr29, $sgpr30_sgpr31, $sgpr46_sgpr47:0x000000000000000F, $sgpr20_sgpr21_sgpr22_sgpr23:0x00000000000000FF, $sgpr24_sgpr25_sgpr26_sgpr27:0x00000000000000FF, $vgpr2_vgpr3:0x000000000000000F, $sgpr0_sgpr1_sgpr2_sgpr3
; GFX90A-NEXT: {{ $}}
; GFX90A-NEXT: renamable $sgpr34_sgpr35 = S_MOV_B64 0
; GFX90A-NEXT: renamable $vcc = S_AND_B64 $exec, renamable $sgpr30_sgpr31, implicit-def dead $scc
- ; GFX90A-NEXT: S_CBRANCH_VCCNZ %bb.58, implicit $vcc
+ ; GFX90A-NEXT: S_CBRANCH_VCCNZ %bb.59, implicit $vcc
; GFX90A-NEXT: {{ $}}
; GFX90A-NEXT: bb.2:
; GFX90A-NEXT: successors: %bb.3(0x80000000)
@@ -51,14 +51,14 @@ define amdgpu_kernel void @f1(ptr addrspace(1) %arg, ptr addrspace(1) %arg1, i64
; GFX90A-NEXT: renamable $sgpr36_sgpr37 = S_MOV_B64 0
; GFX90A-NEXT: {{ $}}
; GFX90A-NEXT: bb.3.Flow17:
- ; GFX90A-NEXT: successors: %bb.4(0x40000000), %bb.57(0x40000000)
+ ; GFX90A-NEXT: successors: %bb.4(0x40000000), %bb.58(0x40000000)
; GFX90A-NEXT: liveins: $sgpr14, $sgpr15, $sgpr16, $sgpr17, $vgpr31, $sgpr4_sgpr5, $sgpr6_sgpr7, $sgpr8_sgpr9:0x000000000000000F, $sgpr10_sgpr11, $sgpr12_sgpr13, $sgpr18_sgpr19, $sgpr28_sgpr29, $sgpr30_sgpr31, $sgpr34_sgpr35, $sgpr36_sgpr37, $sgpr46_sgpr47:0x000000000000000F, $sgpr20_sgpr21_sgpr22_sgpr23:0x000000000000003F, $sgpr24_sgpr25_sgpr26_sgpr27:0x00000000000000FF, $vgpr2_vgpr3:0x000000000000000F, $vgpr10_vgpr11:0x000000000000000F, $vgpr18_vgpr19:0x000000000000000F, $vgpr20_vgpr21:0x000000000000000F, $vgpr22_vgpr23:0x000000000000000F, $vgpr24_vgpr25:0x000000000000000F, $sgpr0_sgpr1_sgpr2_sgpr3
; GFX90A-NEXT: {{ $}}
; GFX90A-NEXT: renamable $vgpr30 = V_AND_B32_e32 1023, $vgpr31, implicit $exec
; GFX90A-NEXT: renamable $vcc = S_AND_B64 $exec, killed renamable $sgpr34_sgpr35, implicit-def dead $scc
; GFX90A-NEXT: renamable $vgpr15 = V_MOV_B32_e32 0, implicit $exec
; GFX90A-NEXT: renamable $vgpr17 = V_MOV_B32_e32 0, implicit $exec
- ; GFX90A-NEXT: S_CBRANCH_VCCZ %bb.57, implicit $vcc
+ ; GFX90A-NEXT: S_CBRANCH_VCCZ %bb.58, implicit $vcc
; GFX90A-NEXT: {{ $}}
; GFX90A-NEXT: bb.4.bb15:
; GFX90A-NEXT: successors: %bb.35(0x40000000), %bb.5(0x40000000)
@@ -122,12 +122,12 @@ define amdgpu_kernel void @f1(ptr addrspace(1) %arg, ptr addrspace(1) %arg1, i64
; GFX90A-NEXT: renamable $vgpr24 = V_MOV_B32_e32 0, implicit $exec
; GFX90A-NEXT: {{ $}}
; GFX90A-NEXT: bb.7.Flow19:
- ; GFX90A-NEXT: successors: %bb.62(0x40000000), %bb.8(0x40000000)
+ ; GFX90A-NEXT: successors: %bb.63(0x40000000), %bb.8(0x40000000)
; GFX90A-NEXT: liveins: $sgpr14, $sgpr15, $sgpr16, $vgpr15, $vgpr17, $vgpr30, $vgpr31, $vgpr52, $vgpr53, $sgpr4_sgpr5, $sgpr6_sgpr7, $sgpr8_sgpr9:0x000000000000000F, $sgpr10_sgpr11, $sgpr12_sgpr13, $sgpr18_sgpr19, $sgpr28_sgpr29, $sgpr30_sgpr31, $sgpr34_sgpr35, $sgpr36_sgpr37, $sgpr38_sgpr39, $sgpr40_sgpr41, $sgpr42_sgpr43, $sgpr44_sgpr45, $sgpr48_sgpr49, $sgpr50_sgpr51, $sgpr52_sgpr53, $sgpr54_sgpr55, $sgpr64_sgpr65, $sgpr66_sgpr67, $sgpr20_sgpr21_sgpr22_sgpr23:0x000000000000003C, $sgpr24_sgpr25_sgpr26_sgpr27:0x00000000000000F0, $vgpr0_vgpr1:0x000000000000000F, $vgpr2_vgpr3:0x000000000000000F, $vgpr4_vgpr5:0x000000000000000F, $vgpr6_vgpr7:0x000000000000000F, $vgpr8_vgpr9:0x000000000000000F, $vgpr10_vgpr11:0x000000000000000F, $vgpr12_vgpr13:0x000000000000000F, $vgpr14_vgpr15:0x0000000000000003, $vgpr16_vgpr17:0x0000000000000003, $vgpr18_vgpr19:0x000000000000000F, $vgpr20_vgpr21:0x000000000000000F, $vgpr22_vgpr23:0x000000000000000F, $vgpr24_vgpr25:0x000000000000000F, $vgpr40_vgpr41:0x000000000000000F, $vgpr42_vgpr43:0x000000000000000F, $vgpr44_vgpr45:0x000000000000000F, $vgpr46_vgpr47:0x000000000000000F, $vgpr56_vgpr57:0x000000000000000F, $vgpr58_vgpr59:0x000000000000000F, $vgpr60_vgpr61:0x000000000000000F, $vgpr62_vgpr63:0x000000000000000F, $sgpr0_sgpr1_sgpr2_sgpr3
; GFX90A-NEXT: {{ $}}
; GFX90A-NEXT: renamable $sgpr68_sgpr69 = S_MOV_B64 0
; GFX90A-NEXT: $sgpr24_sgpr25 = S_AND_SAVEEXEC_B64 $sgpr36_sgpr37, implicit-def $exec, implicit-def $scc, implicit $exec
- ; GFX90A-NEXT: S_CBRANCH_EXECNZ %bb.62, implicit $exec
+ ; GFX90A-NEXT: S_CBRANCH_EXECNZ %bb.63, implicit $exec
; GFX90A-NEXT: {{ $}}
; GFX90A-NEXT: bb.8.Flow32:
; GFX90A-NEXT: successors: %bb.9(0x40000000), %bb.10(0x40000000)
@@ -506,8 +506,8 @@ define amdgpu_kernel void @f1(ptr addrspace(1) %arg, ptr addrspace(1) %arg1, i64
; GFX90A-NEXT: S_BRANCH %bb.38
; GFX90A-NEXT: {{ $}}
; GFX90A-NEXT: bb.41.bb41:
- ; GFX90A-NEXT: successors: %bb.46(0x40000000), %bb.42(0x40000000)
- ; GFX90A-NEXT: liveins: $sgpr14, $sgpr15, $sgpr16, $sgpr17, $vgpr30, $vgpr31, $sgpr4_sgpr5, $sgpr6_sgpr7, $sgpr8_sgpr9:0x000000000000000F, $sgpr10_sgpr11, $sgpr12_sgpr13, $sgpr24_sgpr25, $sgpr28_sgpr29, $sgpr30_sgpr31, $sgpr34_sgpr35, $sgpr36_sgpr37, $sgpr38_sgpr39, $sgpr40_sgpr41, $sgpr46_sgpr47:0x000000000000000F, $sgpr20_sgpr21_sgpr22_sgpr23:0x000000000000003F, $sgpr24_sgpr25_sgpr26_sgpr27:0x00000000000000F0, $vgpr2_vgpr3:0x000000000000000F, $vgpr40_vgpr41:0x000000000000000F, $vgpr42_vgpr43:0x000000000000000F, $vgpr44_vgpr45:0x000000000000000F, $vgpr46_vgpr47:0x000000000000000F, $vgpr56_vgpr57:0x000000000000000F, $sgpr0_sgpr1_sgpr2_sgpr3, $sgpr52_sgpr53, $sgpr50_sgpr51, $sgpr56_sgpr57, $sgpr58_sgpr59, $sgpr60_sgpr61, $sgpr54_sgpr55, $sgpr64_sgpr65, $sgpr66_sgpr67
+ ; GFX90A-NEXT: successors: %bb.47(0x40000000), %bb.42(0x40000000)
+ ; GFX90A-NEXT: liveins: $sgpr14, $sgpr15, $sgpr16, $sgpr17, $vgpr30, $vgpr31, $sgpr4_sgpr5, $sgpr6_sgpr7, $sgpr8_sgpr9:0x000000000000000F, $sgpr10_sgpr11, $sgpr12_sgpr13, $sgpr24_sgpr25, $sgpr28_sgpr29, $sgpr30_sgpr31, $sgpr34_sgpr35, $sgpr36_sgpr37, $sgpr38_sgpr39, $sgpr40_sgpr41, $sgpr46_sgpr47:0x000000000000000F, $sgpr20_sgpr21_sgpr22_sgpr23:0x000000000000003F, $sgpr24_sgpr25_sgpr26_sgpr27:0x00000000000000F0, $vgpr2_vgpr3:0x000000000000000F, $vgpr40_vgpr41:0x000000000000000F, $vgpr42_vgpr43:0x000000000000000F, $vgpr44_vgpr45:0x000000000000000F, $vgpr46_vgpr47:0x000000000000000F, $vgpr56_vgpr57:0x000000000000000F, $sgpr0_sgpr1_sgpr2_sgpr3, $sgpr52_sgpr53, $sgpr50_sgpr51, $sgpr56_sgpr57, $sgpr58_sgpr59, $sgpr60_sgpr61, $sgpr64_sgpr65, $sgpr62_sgpr63, $sgpr66_sgpr67
; GFX90A-NEXT: {{ $}}
; GFX90A-NEXT: renamable $vgpr58 = V_ADD_CO_U32_e32 4096, $vgpr40, implicit-def $vcc, implicit $exec
; GFX90A-NEXT: renamable $sgpr18_sgpr19 = COPY $vcc
@@ -533,17 +533,17 @@ define amdgpu_kernel void @f1(ptr addrspace(1) %arg, ptr addrspace(1) %arg1, i64
; GFX90A-NEXT: renamable $vgpr13 = IMPLICIT_DEF implicit-def $vgpr12
; GFX90A-NEXT: renamable $vgpr11 = IMPLICIT_DEF implicit-def $vgpr10
; GFX90A-NEXT: $sgpr42_sgpr43 = S_AND_SAVEEXEC_B64 $vcc, implicit-def $exec, implicit-def $scc, implicit $exec
- ; GFX90A-NEXT: S_CBRANCH_EXECNZ %bb.46, implicit $exec
+ ; GFX90A-NEXT: S_CBRANCH_EXECNZ %bb.47, implicit $exec
; GFX90A-NEXT: {{ $}}
; GFX90A-NEXT: bb.42.Flow24:
; GFX90A-NEXT: successors: %bb.40(0x80000000)
- ; GFX90A-NEXT: liveins: $sgpr14, $sgpr15, $sgpr16, $vgpr15, $vgpr17, $vgpr18, $vgpr30, $vgpr31, $vgpr52, $vgpr53, $sgpr4_sgpr5, $sgpr6_sgpr7, $sgpr8_sgpr9:0x000000000000000F, $sgpr10_sgpr11, $sgpr12_sgpr13, $sgpr18_sgpr19, $sgpr24_sgpr25, $sgpr28_sgpr29, $sgpr30_sgpr31, $sgpr34_sgpr35, $sgpr36_sgpr37, $sgpr38_sgpr39, $sgpr40_sgpr41, $sgpr42_sgpr43, $sgpr44_sgpr45, $sgpr48_sgpr49, $sgpr50_sgpr51, $sgpr52_sgpr53, $sgpr54_sgpr55, $sgpr56_sgpr57, $sgpr58_sgpr59, $sgpr60_sgpr61, $sgpr20_sgpr21_sgpr22_sgpr23:0x000000000000003C, $sgpr24_sgpr25_sgpr26_sgpr27:0x00000000000000F0, $vgpr0_vgpr1:0x000000000000000F, $vgpr2_vgpr3:0x000000000000000F, $vgpr4_vgpr5:0x000000000000000F, $vgpr6_vgpr7:0x000000000000000F, $vgpr8_vgpr9:0x000000000000000F, $vgpr10_vgpr11:0x000000000000000F, $vgpr12_vgpr13:0x000000000000000F, $vgpr14_vgpr15:0x0000000000000003, $vgpr16_vgpr17:0x0000000000000003, $vgpr40_vgpr41:0x000000000000000F, $vgpr42_vgpr43:0x000000000000000F, $vgpr44_vgpr45:0x000000000000000F, $vgpr46_vgpr47:0x000000000000000F, $vgpr56_vgpr57:0x000000000000000F, $vgpr58_vgpr59:0x0000000000000003, $vgpr60_vgpr61:0x000000000000000F, $vgpr62_vgpr63:0x000000000000000F, $sgpr0_sgpr1_sgpr2_sgpr3
+ ; GFX90A-NEXT: liveins: $sgpr14, $sgpr15, $sgpr16, $vgpr15, $vgpr17, $vgpr18, $vgpr30, $vgpr31, $vgpr52, $vgpr53, $sgpr4_sgpr5, $sgpr6_sgpr7, $sgpr8_sgpr9:0x000000000000000F, $sgpr10_sgpr11, $sgpr12_sgpr13, $sgpr18_sgpr19, $sgpr24_sgpr25, $sgpr28_sgpr29, $sgpr30_sgpr31, $sgpr34_sgpr35, $sgpr36_sgpr37, $sgpr38_sgpr39, $sgpr40_sgpr41, $sgpr42_sgpr43, $sgpr44_sgpr45, $sgpr48_sgpr49, $sgpr50_sgpr51, $sgpr52_sgpr53, $sgpr56_sgpr57, $sgpr58_sgpr59, $sgpr60_sgpr61, $sgpr64_sgpr65, $sgpr20_sgpr21_sgpr22_sgpr23:0x000000000000003C, $sgpr24_sgpr25_sgpr26_sgpr27:0x00000000000000F0, $vgpr0_vgpr1:0x000000000000000F, $vgpr2_vgpr3:0x000000000000000F, $vgpr4_vgpr5:0x000000000000000F, $vgpr6_vgpr7:0x000000000000000F, $vgpr8_vgpr9:0x000000000000000F, $vgpr10_vgpr11:0x000000000000000F, $vgpr12_vgpr13:0x000000000000000F, $vgpr14_vgpr15:0x0000000000000003, $vgpr16_vgpr17:0x0000000000000003, $vgpr40_vgpr41:0x000000000000000F, $vgpr42_vgpr43:0x000000000000000F, $vgpr44_vgpr45:0x000000000000000F, $vgpr46_vgpr47:0x000000000000000F, $vgpr56_vgpr57:0x000000000000000F, $vgpr58_vgpr59:0x0000000000000003, $vgpr60_vgpr61:0x000000000000000F, $vgpr62_vgpr63:0x000000000000000F, $sgpr0_sgpr1_sgpr2_sgpr3
; GFX90A-NEXT: {{ $}}
; GFX90A-NEXT: $exec = S_OR_B64 $exec, killed renamable $sgpr42_sgpr43, implicit-def $scc
; GFX90A-NEXT: renamable $vgpr59 = COPY killed renamable $vgpr18, implicit $exec
; GFX90A-NEXT: renamable $sgpr42_sgpr43 = S_XOR_B64 $exec, -1, implicit-def dead $scc
; GFX90A-NEXT: renamable $sgpr62_sgpr63 = S_AND_B64 killed renamable $sgpr44_sgpr45, $exec, implicit-def dead $scc
- ; GFX90A-NEXT: renamable $sgpr54_sgpr55 = S_AND_B64 killed renamable $sgpr54_sgpr55, $exec, implicit-def dead $scc
+ ; GFX90A-NEXT: renamable $sgpr54_sgpr55 = S_AND_B64 killed renamable $sgpr64_sgpr65, $exec, implicit-def dead $scc
; GFX90A-NEXT: renamable $sgpr60_sgpr61 = S_AND_B64 killed renamable $sgpr60_sgpr61, $exec, implicit-def dead $scc
; GFX90A-NEXT: renamable $sgpr58_sgpr59 = S_AND_B64 killed renamable $sgpr58_sgpr59, $exec, implicit-def dead $scc
; GFX90A-NEXT: renamable $sgpr56_sgpr57 = S_AND_B64 killed renamable $sgpr56_sgpr57, $exec, implicit-def dead $scc
@@ -556,8 +556,8 @@ define amdgpu_kernel void @f1(ptr addrspace(1) %arg, ptr addrspace(1) %arg1, i64
; GFX90A-NEXT: S_BRANCH %bb.40
; GFX90A-NEXT: {{ $}}
; GFX90A-NEXT: bb.43.bb55:
- ; GFX90A-NEXT: successors: %bb.48(0x40000000), %bb.44(0x40000000)
- ; GFX90A-NEXT: liveins: $sgpr14, $sgpr15, $sgpr16, $sgpr17, $vgpr18, $vgpr30, $vgpr31, $sgpr4_sgpr5, $sgpr6_sgpr7, $sgpr8_sgpr9:0x000000000000000F, $sgpr10_sgpr11, $sgpr12_sgpr13, $sgpr18_sgpr19, $sgpr24_sgpr25, $sgpr28_sgpr29, $sgpr30_sgpr31, $sgpr34_sgpr35, $sgpr36_sgpr37, $sgpr38_sgpr39, $sgpr40_sgpr41, $sgpr42_sgpr43, $sgpr46_sgpr47:0x000000000000000F, $sgpr20_sgpr21_sgpr22_sgpr23:0x000000000000003F, $sgpr24_sgpr25_sgpr26_sgpr27:0x00000000000000F0, $vgpr2_vgpr3:0x000000000000000F, $vgpr40_vgpr41:0x000000000000000F, $vgpr42_vgpr43:0x000000000000000F, $vgpr44_vgpr45:0x000000000000000F, $vgpr46_vgpr47:0x000000000000000F, $vgpr56_vgpr57:0x000000000000000F, $vgpr58_vgpr59:0x0000000000000003, $vgpr60_vgpr61:0x000000000000000F, $sgpr0_sgpr1_sgpr2_sgpr3, $sgpr44_sgpr45, $sgpr54_sgpr55, $sgpr58_sgpr59, $sgpr56_sgpr57, $sgpr48_sgpr49
+ ; GFX90A-NEXT: successors: %bb.49(0x40000000), %bb.44(0x40000000)
+ ; GFX90A-NEXT: liveins: $sgpr14, $sgpr15, $sgpr16, $sgpr17, $vgpr18, $vgpr30, $vgpr31, $sgpr4_sgpr5, $sgpr6_sgpr7, $sgpr8_sgpr9:0x000000000000000F, $sgpr10_sgpr11, $sgpr12_sgpr13, $sgpr18_sgpr19, $sgpr24_sgpr25, $sgpr28_sgpr29, $sgpr30_sgpr31, $sgpr34_sgpr35, $sgpr36_sgpr37, $sgpr38_sgpr39, $sgpr40_sgpr41, $sgpr42_sgpr43, $sgpr46_sgpr47:0x000000000000000F, $sgpr20_sgpr21_sgpr22_sgpr23:0x000000000000003F, $sgpr24_sgpr25_sgpr26_sgpr27:0x00000000000000F0, $vgpr2_vgpr3:0x000000000000000F, $vgpr40_vgpr41:0x000000000000000F, $vgpr42_vgpr43:0x000000000000000F, $vgpr44_vgpr45:0x000000000000000F, $vgpr46_vgpr47:0x000000000000000F, $vgpr56_vgpr57:0x000000000000000F, $vgpr58_vgpr59:0x0000000000000003, $vgpr60_vgpr61:0x000000000000000F, $sgpr0_sgpr1_sgpr2_sgpr3, $sgpr44_sgpr45, $sgpr58_sgpr59, $sgpr56_sgpr57, $sgpr48_sgpr49
; GFX90A-NEXT: {{ $}}
; GFX90A-NEXT: S_BITCMP1_B32 killed renamable $sgpr17, 16, implicit-def $scc
; GFX90A-NEXT: renamable $sgpr64_sgpr65 = S_CSELECT_B64 -1, 0, implicit killed $scc
@@ -565,13 +565,19 @@ define amdgpu_kernel void @f1(ptr addrspace(1) %arg, ptr addrspace(1) %arg1, i64
; GFX90A-NEXT: renamable $vgpr62 = V_ADD_CO_U32_e32 6144, $vgpr40, implicit-def $vcc, implicit $exec
; GFX90A-NEXT: renamable $vgpr63, dead renamable $vcc = V_ADDC_U32_e64 0, $vgpr41, killed $vcc, 0, implicit $exec
; GFX90A-NEXT: renamable $vcc = S_AND_B64 $exec, renamable $sgpr50_sgpr51, implicit-def dead $scc
- ; GFX90A-NEXT: S_CBRANCH_VCCNZ %bb.48, implicit $vcc
+ ; GFX90A-NEXT: S_CBRANCH_VCCNZ %bb.49, implicit $vcc
; GFX90A-NEXT: {{ $}}
; GFX90A-NEXT: bb.44:
; GFX90A-NEXT: successors: %bb.45(0x80000000)
- ; GFX90A-NEXT: liveins: $sgpr14, $sgpr15, $sgpr16, $vgpr57, $vgpr56, $vgpr18, $vgpr30, $vgpr31, $vgpr60, $vgpr62, $sgpr4_sgpr5, $sgpr6_sgpr7, $sgpr8, $sgpr9, $sgpr10_sgpr11, $sgpr12_sgpr13, $sgpr18_sgpr19, $sgpr28_sgpr29, $sgpr30_sgpr31, $sgpr34_sgpr35, $sgpr36_sgpr37, $sgpr38_sgpr39, $sgpr40_sgpr41, $sgpr42_sgpr43, $sgpr44_sgpr45, $sgpr48_sgpr49, $vgpr61, $vgpr58, $sgpr54_sgpr55, $sgpr56_sgpr57, $sgpr58_sgpr59, $sgpr20_sgpr21_sgpr22, $sgpr22_sgpr23, $sgpr24_sgpr25_sgpr26, $sgpr26_sgpr27, $vgpr47, $vgpr46, $vgpr2, $vgpr3, $vgpr45, $vgpr44, $vgpr43, $vgpr42, $vgpr41, $vgpr40, $vgpr63
+ ; GFX90A-NEXT: liveins: $sgpr14, $sgpr15, $sgpr16, $vgpr18, $vgpr30, $vgpr31, $sgpr4_sgpr5, $sgpr6_sgpr7, $sgpr8_sgpr9:0x000000000000000F, $sgpr10_sgpr11, $sgpr12_sgpr13, $sgpr18_sgpr19, $sgpr24_sgpr25, $sgpr28_sgpr29, $sgpr30_sgpr31, $sgpr34_sgpr35, $sgpr36_sgpr37, $sgpr38_sgpr39, $sgpr40_sgpr41, $sgpr42_sgpr43, $sgpr44_sgpr45, $sgpr20_sgpr21_sgpr22_sgpr23:0x000000000000003C, $sgpr24_sgpr25_sgpr26_sgpr27:0x00000000000000F0, $vgpr2_vgpr3:0x000000000000000F, $vgpr40_vgpr41:0x000000000000000F, $vgpr42_vgpr43:0x000000000000000F, $vgpr44_vgpr45:0x000000000000000F, $vgpr46_vgpr47:0x000000000000000F, $vgpr56_vgpr57:0x000000000000000F, $vgpr58_vgpr59:0x0000000000000003, $vgpr60_vgpr61:0x000000000000000F, $vgpr62_vgpr63:0x000000000000000F, $sgpr0_sgpr1_sgpr2_sgpr3, $sgpr58_sgpr59, $sgpr56_sgpr57, $sgpr48_sgpr49
; GFX90A-NEXT: {{ $}}
+ ; GFX90A-NEXT: renamable $sgpr54_sgpr55 = S_MOV_B64 0
; GFX90A-NEXT: renamable $sgpr52_sgpr53 = COPY renamable $sgpr36_sgpr37
+ ; GFX90A-NEXT: {{ $}}
+ ; GFX90A-NEXT: bb.45.Flow26:
+ ; GFX90A-NEXT: successors: %bb.46(0x80000000)
+ ; GFX90A-NEXT: liveins: $sgpr14, $sgpr15, $sgpr16, $vgpr57, $vgpr56, $vgpr18, $vgpr30, $vgpr31, $vgpr60, $vgpr62, $sgpr4_sgpr5, $sgpr6_sgpr7, $sgpr8, $sgpr9, $sgpr10_sgpr11, $sgpr12_sgpr13, $sgpr18_sgpr19, $sgpr28_sgpr29, $sgpr30_sgpr31, $sgpr34_sgpr35, $sgpr36_sgpr37, $sgpr38_sgpr39, $sgpr40_sgpr41, $sgpr42_sgpr43, $sgpr44_sgpr45, $sgpr48_sgpr49, $sgpr52_sgpr53, $sgpr54_sgpr55, $sgpr56_sgpr57, $sgpr58_sgpr59, $sgpr20_sgpr21_sgpr22, $sgpr22_sgpr23, $sgpr24_sgpr25_sgpr26, $sgpr26_sgpr27, $vgpr47, $vgpr46, $vgpr2, $vgpr3, $vgpr45, $vgpr44, $vgpr43, $vgpr42, $vgpr41, $vgpr40, $vgpr63, $vgpr58, $vgpr61
+ ; GFX90A-NEXT: {{ $}}
; GFX90A-NEXT: renamable $vgpr8_vgpr9 = IMPLICIT_DEF
; GFX90A-NEXT: renamable $vgpr6_vgpr7 = IMPLICIT_DEF
; GFX90A-NEXT: renamable $vgpr4_vgpr5 = IMPLICIT_DEF
@@ -586,8 +592,8 @@ define amdgpu_kernel void @f1(ptr addrspace(1) %arg, ptr addrspace(1) %arg1, i64
; GFX90A-NEXT: renamable $vgpr11 = IMPLICIT_DEF implicit-def $vgpr10
; GFX90A-NEXT: renamable $sgpr50_sgpr51 = S_MOV_B64 0
; GFX90A-NEXT: {{ $}}
- ; GFX90A-NEXT: bb.45.Flow26:
- ; GFX90A-NEXT: successors: %bb.47(0x80000000)
+ ; GFX90A-NEXT: bb.46.Flow26:
+ ; GFX90A-NEXT: successors: %bb.48(0x80000000)
; GFX90A-NEXT: liveins: $sgpr14, $sgpr15, $sgpr16, $vgpr15, $vgpr17, $vgpr18, $vgpr30, $vgpr31, $vgpr52, $vgpr53, $sgpr4_sgpr5, $sgpr6_sgpr7, $sgpr8_sgpr9:0x000000000000000F, $sgpr10_sgpr11, $sgpr12_sgpr13, $sgpr18_sgpr19, $sgpr24_sgpr25, $sgpr28_sgpr29, $sgpr30_sgpr31, $sgpr34_sgpr35, $sgpr36_sgpr37, $sgpr38_sgpr39, $sgpr40_sgpr41, $sgpr42_sgpr43, $sgpr44_sgpr45, $sgpr48_sgpr49, $sgpr50_sgpr51, $sgpr52_sgpr53, $sgpr54_sgpr55, $sgpr56_sgpr57, $sgpr58_sgpr59, $sgpr20_sgpr21_sgpr22_sgpr23:0x000000000000003C, $sgpr24_sgpr25_sgpr26_sgpr27:0x00000000000000F0, $vgpr0_vgpr1:0x000000000000000F, $vgpr2_vgpr3:0x000000000000000F, $vgpr4_vgpr5:0x000000000000000F, $vgpr6_vgpr7:0x000000000000000F, $vgpr8_vgpr9:0x000000000000000F, $vgpr10_vgpr11:0x000000000000000F, $vgpr12_vgpr13:0x000000000000000F, $vgpr14_vgpr15:0x0000000000000003, $vgpr16_vgpr17:0x0000000000000003, $vgpr40_vgpr41:0x000000000000000F, $vgpr42_vgpr43:0x000000000000000F, $vgpr44_vgpr45:0x000000000000000F, $vgpr46_vgpr47:0x000000000000000F, $vgpr56_vgpr57:0x000000000000000F, $vgpr58_vgpr59:0x0000000000000003, $vgpr60_vgpr61:0x000000000000000F, $vgpr62_vgpr63:0x000000000000000F, $sgpr0_sgpr1_sgpr2_sgpr3
; GFX90A-NEXT: {{ $}}
; GFX90A-NEXT: renamable $sgpr60_sgpr61 = S_XOR_B64 $exec, -1, implicit-def dead $scc
@@ -595,16 +601,16 @@ define amdgpu_kernel void @f1(ptr addrspace(1) %arg, ptr addrspace(1) %arg1, i64
; GFX90A-NEXT: renamable $sgpr68_sgpr69 = S_AND_B64 killed renamable $sgpr48_sgpr49, $exec, implicit-def dead $scc
; GFX90A-NEXT: renamable $sgpr66_sgpr67 = S_AND_B64 killed renamable $sgpr50_sgpr51, $exec, implicit-def dead $scc
; GFX90A-NEXT: renamable $sgpr50_sgpr51 = S_AND_B64 killed renamable $sgpr56_sgpr57, $exec, implicit-def dead $scc
- ; GFX90A-NEXT: renamable $sgpr64_sgpr65 = S_AND_B64 killed renamable $sgpr58_sgpr59, $exec, implicit-def dead $scc
+ ; GFX90A-NEXT: renamable $sgpr62_sgpr63 = S_AND_B64 killed renamable $sgpr58_sgpr59, $exec, implicit-def dead $scc
; GFX90A-NEXT: renamable $sgpr48_sgpr49 = S_AND_B64 killed renamable $sgpr54_sgpr55, $exec, implicit-def dead $scc
; GFX90A-NEXT: renamable $sgpr44_sgpr45 = S_ANDN2_B64 renamable $sgpr36_sgpr37, $exec, implicit-def dead $scc
; GFX90A-NEXT: renamable $sgpr46_sgpr47 = S_AND_B64 killed renamable $sgpr52_sgpr53, $exec, implicit-def dead $scc
- ; GFX90A-NEXT: renamable $sgpr62_sgpr63 = S_OR_B64 killed renamable $sgpr44_sgpr45, killed renamable $sgpr46_sgpr47, implicit-def dead $scc
- ; GFX90A-NEXT: S_BRANCH %bb.47
+ ; GFX90A-NEXT: renamable $sgpr54_sgpr55 = S_OR_B64 killed renamable $sgpr44_sgpr45, killed renamable $sgpr46_sgpr47, implicit-def dead $scc
+ ; GFX90A-NEXT: S_BRANCH %bb.48
; GFX90A-NEXT: {{ $}}
- ; GFX90A-NEXT: bb.46.bb48:
- ; GFX90A-NEXT: successors: %bb.43(0x40000000), %bb.47(0x40000000)
- ; GFX90A-NEXT: liveins: $sgpr14, $sgpr15, $sgpr16, $sgpr17, $vgpr18, $vgpr30, $vgpr31, $sgpr4_sgpr5, $sgpr6_sgpr7, $sgpr8_sgpr9:0x000000000000000F, $sgpr10_sgpr11, $sgpr12_sgpr13, $sgpr24_sgpr25, $sgpr28_sgpr29, $sgpr30_sgpr31, $sgpr34_sgpr35, $sgpr36_sgpr37, $sgpr38_sgpr39, $sgpr40_sgpr41, $sgpr42_sgpr43, $sgpr46_sgpr47:0x000000000000000F, $sgpr20_sgpr21_sgpr22_sgpr23:0x000000000000003F, $sgpr24_sgpr25_sgpr26_sgpr27:0x00000000000000F0, $vgpr2_vgpr3:0x000000000000000F, $vgpr40_vgpr41:0x000000000000000F, $vgpr42_vgpr43:0x000000000000000F, $vgpr44_vgpr45:0x000000000000000F, $vgpr46_vgpr47:0x000000000000000F, $vgpr56_vgpr57:0x000000000000000F, $vgpr58_vgpr59:0x0000000000000003, $sgpr0_sgpr1_sgpr2_sgpr3, $sgpr64_sgpr65, $sgpr50_sgpr51, $sgpr66_sgpr67, $sgpr44_sgpr45, $sgpr54_sgpr55, $sgpr58_sgpr59, $sgpr56_sgpr57
+ ; GFX90A-NEXT: bb.47.bb48:
+ ; GFX90A-NEXT: successors: %bb.43(0x40000000), %bb.48(0x40000000)
+ ; GFX90A-NEXT: liveins: $sgpr14, $sgpr15, $sgpr16, $sgpr17, $vgpr18, $vgpr30, $vgpr31, $sgpr4_sgpr5, $sgpr6_sgpr7, $sgpr8_sgpr9:0x000000000000000F, $sgpr10_sgpr11, $sgpr12_sgpr13, $sgpr24_sgpr25, $sgpr28_sgpr29, $sgpr30_sgpr31, $sgpr34_sgpr35, $sgpr36_sgpr37, $sgpr38_sgpr39, $sgpr40_sgpr41, $sgpr42_sgpr43, $sgpr46_sgpr47:0x000000000000000F, $sgpr20_sgpr21_sgpr22_sgpr23:0x000000000000003F, $sgpr24_sgpr25_sgpr26_sgpr27:0x00000000000000F0, $vgpr2_vgpr3:0x000000000000000F, $vgpr40_vgpr41:0x000000000000000F, $vgpr42_vgpr43:0x000000000000000F, $vgpr44_vgpr45:0x000000000000000F, $vgpr46_vgpr47:0x000000000000000F, $vgpr56_vgpr57:0x000000000000000F, $vgpr58_vgpr59:0x0000000000000003, $sgpr0_sgpr1_sgpr2_sgpr3, $sgpr62_sgpr63, $sgpr50_sgpr51, $sgpr66_sgpr67, $sgpr44_sgpr45, $sgpr58_sgpr59, $sgpr56_sgpr57
; GFX90A-NEXT: {{ $}}
; GFX90A-NEXT: renamable $vgpr60 = V_ADD_CO_U32_e32 5120, $vgpr40, implicit-def $vcc, implicit $exec
; GFX90A-NEXT: renamable $sgpr18_sgpr19 = COPY $vcc
@@ -613,7 +619,7 @@ define amdgpu_kernel void @f1(ptr addrspace(1) %arg, ptr addrspace(1) %arg1, i64
; GFX90A-NEXT: renamable $vgpr0 = FLAT_LOAD_UBYTE killed renamable $vgpr0_vgpr1, 1024, 0, implicit $exec, implicit $flat_scr :: (load (s8) from %ir.i51)
; GFX90A-NEXT: renamable $sgpr48_sgpr49 = S_MOV_B64 0
; GFX90A-NEXT: renamable $sgpr60_sgpr61 = S_MOV_B64 -1
- ; GFX90A-NEXT: renamable $sgpr62_sgpr63 = COPY renamable $sgpr36_sgpr37
+ ; GFX90A-NEXT: renamable $sgpr54_sgpr55 = COPY renamable $sgpr36_sgpr37
; GFX90A-NEXT: renamable $sgpr68_sgpr69 = S_MOV_B64 0
; GFX90A-NEXT: renamable $vgpr61, dead renamable $vcc = V_ADDC_U32_e64 0, $vgpr41, killed $sgpr18_sgpr19, 0, implicit $exec
; GFX90A-NEXT: renamable $vcc = V_CMP_EQ_U16_e64 0, killed $vgpr0, implicit $exec
@@ -634,53 +640,56 @@ define amdgpu_kernel void @f1(ptr addrspace(1) %arg, ptr addrspace(1) %arg1, i64
; GFX90A-NEXT: $sgpr18_sgpr19 = S_AND_SAVEEXEC_B64 $vcc, implicit-def $exec, implicit-def $scc, implicit $exec
; GFX90A-NEXT: S_CBRANCH_EXECNZ %bb.43, implicit $exec
; GFX90A-NEXT: {{ $}}
- ; GFX90A-NEXT: bb.47.Flow25:
+ ; GFX90A-NEXT: bb.48.Flow25:
; GFX90A-NEXT: successors: %bb.42(0x80000000)
- ; GFX90A-NEXT: liveins: $sgpr14, $sgpr15, $sgpr16, $vgpr15, $vgpr17, $vgpr18, $vgpr30, $vgpr31, $vgpr52, $vgpr53, $sgpr4_sgpr5, $sgpr6_sgpr7, $sgpr8_sgpr9:0x000000000000000F, $sgpr10_sgpr11, $sgpr12_sgpr13, $sgpr18_sgpr19, $sgpr24_sgpr25, $sgpr28_sgpr29, $sgpr30_sgpr31, $sgpr34_sgpr35, $sgpr36_sgpr37, $sgpr38_sgpr39, $sgpr40_sgpr41, $sgpr42_sgpr43, $sgpr48_sgpr49, $sgpr50_sgpr51, $sgpr60_sgpr61, $sgpr62_sgpr63, $sgpr64_sgpr65, $sgpr66_sgpr67, $sgpr68_sgpr69, $sgpr70_sgpr71, $sgpr20_sgpr21_sgpr22_sgpr23:0x000000000000003C, $sgpr24_sgpr25_sgpr26_sgpr27:0x00000000000000F0, $vgpr0_vgpr1:0x000000000000000F, $vgpr2_vgpr3:0x000000000000000F, $vgpr4_vgpr5:0x000000000000000F, $vgpr6_vgpr7:0x000000000000000F, $vgpr8_vgpr9:0x000000000000000F, $vgpr10_vgpr11:0x000000000000000F, $vgpr12_vgpr13:0x000000000000000F, $vgpr14_vgpr15:0x0000000000000003, $vgpr16_vgpr17:0x0000000000000003, $vgpr40_vgpr41:0x000000000000000F, $vgpr42_vgpr43:0x000000000000000F, $vgpr44_vgpr45:0x000000000000000F, $vgpr46_vgpr47:0x000000000000000F, $vgpr56_vgpr57:0x000000000000000F, $vgpr58_vgpr59:0x0000000000000003, $vgpr60_vgpr61:0x000000000000000F, $vgpr62_vgpr63:0x000000000000000F, $sgpr0_sgpr1_sgpr2_sgpr3
+ ; GFX90A-NEXT: liveins: $sgpr14, $sgpr15, $sgpr16, $vgpr15, $vgpr17, $vgpr18, $vgpr30, $vgpr31, $vgpr52, $vgpr53, $sgpr4_sgpr5, $sgpr6_sgpr7, $sgpr8_sgpr9:0x000000000000000F, $sgpr10_sgpr11, $sgpr12_sgpr13, $sgpr18_sgpr19, $sgpr24_sgpr25, $sgpr28_sgpr29, $sgpr30_sgpr31, $sgpr34_sgpr35, $sgpr36_sgpr37, $sgpr38_sgpr39, $sgpr40_sgpr41, $sgpr42_sgpr43, $sgpr48_sgpr49, $sgpr50_sgpr51, $sgpr54_sgpr55, $sgpr60_sgpr61, $sgpr62_sgpr63, $sgpr66_sgpr67, $sgpr68_sgpr69, $sgpr70_sgpr71, $sgpr20_sgpr21_sgpr22_sgpr23:0x000000000000003C, $sgpr24_sgpr25_sgpr26_sgpr27:0x00000000000000F0, $vgpr0_vgpr1:0x000000000000000F, $vgpr2_vgpr3:0x000000000000000F, $vgpr4_vgpr5:0x000000000000000F, $vgpr6_vgpr7:0x000000000000000F, $vgpr8_vgpr9:0x000000000000000F, $vgpr10_vgpr11:0x000000000000000F, $vgpr12_vgpr13:0x000000000000000F, $vgpr14_vgpr15:0x0000000000000003, $vgpr16_vgpr17:0x0000000000000003, $vgpr40_vgpr41:0x000000000000000F, $vgpr42_vgpr43:0x000000000000000F, $vgpr44_vgpr45:0x000000000000000F, $vgpr46_vgpr47:0x000000000000000F, $vgpr56_vgpr57:0x000000000000000F, $vgpr58_vgpr59:0x0000000000000003, $vgpr60_vgpr61:0x000000000000000F, $vgpr62_vgpr63:0x000000000000000F, $sgpr0_sgpr1_sgpr2_sgpr3
; GFX90A-NEXT: {{ $}}
; GFX90A-NEXT: $exec = S_OR_B64 $exec, killed renamable $sgpr18_sgpr19, implicit-def $scc
; GFX90A-NEXT: renamable $sgpr44_sgpr45 = S_XOR_B64 $exec, -1, implicit-def dead $scc
- ; GFX90A-NEXT: renamable $sgpr54_sgpr55 = S_AND_B64 killed renamable $sgpr60_sgpr61, $exec, implicit-def dead $scc
+ ; GFX90A-NEXT: renamable $sgpr64_sgpr65 = S_AND_B64 killed renamable $sgpr60_sgpr61, $exec, implicit-def dead $scc
; GFX90A-NEXT: renamable $sgpr60_sgpr61 = S_AND_B64 killed renamable $sgpr70_sgpr71, $exec, implicit-def dead $scc
; GFX90A-NEXT: renamable $sgpr58_sgpr59 = S_AND_B64 killed renamable $sgpr68_sgpr69, $exec, implicit-def dead $scc
; GFX90A-NEXT: renamable $sgpr56_sgpr57 = S_AND_B64 killed renamable $sgpr66_sgpr67, $exec, implicit-def dead $scc
; GFX90A-NEXT: renamable $sgpr50_sgpr51 = S_AND_B64 killed renamable $sgpr50_sgpr51, $exec, implicit-def dead $scc
- ; GFX90A-NEXT: renamable $sgpr52_sgpr53 = S_AND_B64 killed renamable $sgpr64_sgpr65, $exec, implicit-def dead $scc
+ ; GFX90A-NEXT: renamable $sgpr52_sgpr53 = S_AND_B64 killed renamable $sgpr62_sgpr63, $exec, implicit-def dead $scc
; GFX90A-NEXT: renamable $sgpr18_sgpr19 = S_AND_B64 killed renamable $sgpr48_sgpr49, $exec, implicit-def dead $scc
; GFX90A-NEXT: renamable $sgpr46_sgpr47 = S_ANDN2_B64 renamable $sgpr36_sgpr37, $exec, implicit-def dead $scc
- ; GFX90A-NEXT: renamable $sgpr48_sgpr49 = S_AND_B64 killed renamable $sgpr62_sgpr63, $exec, implicit-def dead $scc
+ ; GFX90A-NEXT: renamable $sgpr48_sgpr49 = S_AND_B64 killed renamable $sgpr54_sgpr55, $exec, implicit-def dead $scc
; GFX90A-NEXT: renamable $sgpr48_sgpr49 = S_OR_B64 killed renamable $sgpr46_sgpr47, killed renamable $sgpr48_sgpr49, implicit-def dead $scc
; GFX90A-NEXT: S_BRANCH %bb.42
; GFX90A-NEXT: {{ $}}
- ; GFX90A-NEXT: bb.48.bb63:
- ; GFX90A-NEXT: successors: %bb.50(0x40000000), %bb.49(0x40000000)
- ; GFX90A-NEXT: liveins: $vcc, $sgpr14, $sgpr15, $sgpr16, $vgpr18, $vgpr30, $vgpr31, $sgpr4_sgpr5, $sgpr6_sgpr7, $sgpr8_sgpr9:0x000000000000000F, $sgpr10_sgpr11, $sgpr12_sgpr13, $sgpr18_sgpr19, $sgpr24_sgpr25, $sgpr28_sgpr29, $sgpr30_sgpr31, $sgpr34_sgpr35, $sgpr36_sgpr37, $sgpr38_sgpr39, $sgpr40_sgpr41, $sgpr42_sgpr43, $sgpr46_sgpr47:0x000000000000000F, $sgpr50_sgpr51, $sgpr64_sgpr65, $sgpr20_sgpr21_sgpr22_sgpr23:0x000000000000003F, $sgpr24_sgpr25_sgpr26_sgpr27:0x00000000000000F0, $vgpr2_vgpr3:0x000000000000000F, $vgpr40_vgpr41:0x000000000000000F, $vgpr42_vgpr43:0x000000000000000F, $vgpr44_vgpr45:0x000000000000000F, $vgpr46_vgpr47:0x000000000000000F, $vgpr56_vgpr57:0x000000000000000F, $vgpr58_vgpr59:0x0000000000000003, $vgpr60_vgpr61:0x000000000000000F, $vgpr62_vgpr63:0x000000000000000F, $sgpr0_sgpr1_sgpr2_sgpr3, $sgpr54_sgpr55, $sgpr58_sgpr59, $sgpr56_sgpr57, $sgpr48_sgpr49
+ ; GFX90A-NEXT: bb.49.bb63:
+ ; GFX90A-NEXT: successors: %bb.51(0x40000000), %bb.50(0x40000000)
+ ; GFX90A-NEXT: liveins: $vcc, $sgpr14, $sgpr15, $sgpr16, $vgpr18, $vgpr30, $vgpr31, $sgpr4_sgpr5, $sgpr6_sgpr7, $sgpr8_sgpr9:0x000000000000000F, $sgpr10_sgpr11, $sgpr12_sgpr13, $sgpr18_sgpr19, $sgpr24_sgpr25, $sgpr28_sgpr29, $sgpr30_sgpr31, $sgpr34_sgpr35, $sgpr36_sgpr37, $sgpr38_sgpr39, $sgpr40_sgpr41, $sgpr42_sgpr43, $sgpr46_sgpr47:0x000000000000000F, $sgpr50_sgpr51, $sgpr64_sgpr65, $sgpr20_sgpr21_sgpr22_sgpr23:0x000000000000003F, $sgpr24_sgpr25_sgpr26_sgpr27:0x00000000000000F0, $vgpr2_vgpr3:0x000000000000000F, $vgpr40_vgpr41:0x000000000000000F, $vgpr42_vgpr43:0x000000000000000F, $vgpr44_vgpr45:0x000000000000000F, $vgpr46_vgpr47:0x000000000000000F, $vgpr56_vgpr57:0x000000000000000F, $vgpr58_vgpr59:0x0000000000000003, $vgpr60_vgpr61:0x000000000000000F, $vgpr62_vgpr63:0x000000000000000F, $sgpr0_sgpr1_sgpr2_sgpr3, $sgpr58_sgpr59, $sgpr56_sgpr57, $sgpr48_sgpr49
; GFX90A-NEXT: {{ $}}
; GFX90A-NEXT: renamable $sgpr44_sgpr45 = S_MOV_B64 0
- ; GFX90A-NEXT: S_CBRANCH_VCCNZ %bb.50, implicit $vcc
+ ; GFX90A-NEXT: S_CBRANCH_VCCNZ %bb.51, implicit $vcc
; GFX90A-NEXT: {{ $}}
- ; GFX90A-NEXT: bb.49:
- ; GFX90A-NEXT: successors: %bb.44(0x80000000)
- ; GFX90A-NEXT: liveins: $sgpr14, $sgpr15, $sgpr16, $vgpr18, $vgpr30, $vgpr31, $sgpr4_sgpr5, $sgpr6_sgpr7, $sgpr8_sgpr9:0x000000000000000F, $sgpr10_sgpr11, $sgpr12_sgpr13, $sgpr18_sgpr19, $sgpr24_sgpr25, $sgpr28_sgpr29, $sgpr30_sgpr31, $sgpr34_sgpr35, $sgpr36_sgpr37, $sgpr38_sgpr39, $sgpr40_sgpr41, $sgpr42_sgpr43, $sgpr44_sgpr45, $sgpr20_sgpr21_sgpr22_sgpr23:0x000000000000003C, $sgpr24_sgpr25_sgpr26_sgpr27:0x00000000000000F0, $vgpr2_vgpr3:0x000000000000000F, $vgpr40_vgpr41:0x000000000000000F, $vgpr42_vgpr43:0x000000000000000F, $vgpr44_vgpr45:0x000000000000000F, $vgpr46_vgpr47:0x000000000000000F, $vgpr56_vgpr57:0x000000000000000F, $vgpr58_vgpr59:0x0000000000000003, $vgpr60_vgpr61:0x000000000000000F, $vgpr62_vgpr63:0x000000000000000F, $sgpr0_sgpr1_sgpr2_sgpr3, $sgpr54_sgpr55, $sgpr58_sgpr59, $sgpr56_sgpr57
+ ; GFX90A-NEXT: bb.50:
+ ; GFX90A-NEXT: successors: %bb.45(0x80000000)
+ ; GFX90A-NEXT: liveins: $sgpr14, $sgpr15, $sgpr16, $vgpr18, $vgpr30, $vgpr31, $sgpr4_sgpr5, $sgpr6_sgpr7, $sgpr8_sgpr9:0x000000000000000F, $sgpr10_sgpr11, $sgpr12_sgpr13, $sgpr18_sgpr19, $sgpr24_sgpr25, $sgpr28_sgpr29, $sgpr30_sgpr31, $sgpr34_sgpr35, $sgpr36_sgpr37, $sgpr38_sgpr39, $sgpr40_sgpr41, $sgpr42_sgpr43, $sgpr44_sgpr45, $sgpr20_sgpr21_sgpr22_sgpr23:0x000000000000003C, $sgpr24_sgpr25_sgpr26_sgpr27:0x00000000000000F0, $vgpr2_vgpr3:0x000000000000000F, $vgpr40_vgpr41:0x000000000000000F, $vgpr42_vgpr43:0x000000000000000F, $vgpr44_vgpr45:0x000000000000000F, $vgpr46_vgpr47:0x000000000000000F, $vgpr56_vgpr57:0x000000000000000F, $vgpr58_vgpr59:0x0000000000000003, $vgpr60_vgpr61:0x000000000000000F, $vgpr62_vgpr63:0x000000000000000F, $sgpr0_sgpr1_sgpr2_sgpr3, $sgpr58_sgpr59, $sgpr56_sgpr57
; GFX90A-NEXT: {{ $}}
; GFX90A-NEXT: renamable $sgpr48_sgpr49 = S_MOV_B64 -1
- ; GFX90A-NEXT: S_BRANCH %bb.44
+ ; GFX90A-NEXT: renamable $sgpr52_sgpr53 = COPY renamable $sgpr36_sgpr37
+ ; GFX90A-NEXT: renamable $sgpr54_sgpr55 = S_MOV_B64 0
+ ; GFX90A-NEXT: S_BRANCH %bb.45
; GFX90A-NEXT: {{ $}}
- ; GFX90A-NEXT: bb.50.bb68:
- ; GFX90A-NEXT: successors: %bb.54(0x40000000), %bb.51(0x40000000)
- ; GFX90A-NEXT: liveins: $sgpr14, $sgpr15, $sgpr16, $vgpr18, $vgpr30, $vgpr31, $sgpr4_sgpr5, $sgpr6_sgpr7, $sgpr8_sgpr9:0x000000000000000F, $sgpr10_sgpr11, $sgpr12_sgpr13, $sgpr18_sgpr19, $sgpr24_sgpr25, $sgpr28_sgpr29, $sgpr30_sgpr31, $sgpr34_sgpr35, $sgpr36_sgpr37, $sgpr38_sgpr39, $sgpr40_sgpr41, $sgpr42_sgpr43, $sgpr44_sgpr45, $sgpr46_sgpr47:0x000000000000000F, $sgpr50_sgpr51, $sgpr64_sgpr65, $sgpr20_sgpr21_sgpr22_sgpr23:0x000000000000003F, $sgpr24_sgpr25_sgpr26_sgpr27:0x00000000000000F0, $vgpr2_vgpr3:0x000000000000000F, $vgpr40_vgpr41:0x000000000000000F, $vgpr42_vgpr43:0x000000000000000F, $vgpr44_vgpr45:0x000000000000000F, $vgpr46_vgpr47:0x000000000000000F, $vgpr56_vgpr57:0x000000000000000F, $vgpr58_vgpr59:0x0000000000000003, $vgpr60_vgpr61:0x000000000000000F, $vgpr62_vgpr63:0x000000000000000F, $sgpr0_sgpr1_sgpr2_sgpr3, $sgpr48_sgpr49, $sgpr54_sgpr55, $sgpr58_sgpr59, $sgpr56_sgpr57
+ ; GFX90A-NEXT: bb.51.bb68:
+ ; GFX90A-NEXT: successors: %bb.55(0x40000000), %bb.52(0x40000000)
+ ; GFX90A-NEXT: liveins: $sgpr14, $sgpr15, $sgpr16, $vgpr18, $vgpr30, $vgpr31, $sgpr4_sgpr5, $sgpr6_sgpr7, $sgpr8_sgpr9:0x000000000000000F, $sgpr10_sgpr11, $sgpr12_sgpr13, $sgpr18_sgpr19, $sgpr24_sgpr25, $sgpr28_sgpr29, $sgpr30_sgpr31, $sgpr34_sgpr35, $sgpr36_sgpr37, $sgpr38_sgpr39, $sgpr40_sgpr41, $sgpr42_sgpr43, $sgpr44_sgpr45, $sgpr46_sgpr47:0x000000000000000F, $sgpr50_sgpr51, $sgpr64_sgpr65, $sgpr20_sgpr21_sgpr22_sgpr23:0x000000000000003F, $sgpr24_sgpr25_sgpr26_sgpr27:0x00000000000000F0, $vgpr2_vgpr3:0x000000000000000F, $vgpr40_vgpr41:0x000000000000000F, $vgpr42_vgpr43:0x000000000000000F, $vgpr44_vgpr45:0x000000000000000F, $vgpr46_vgpr47:0x000000000000000F, $vgpr56_vgpr57:0x000000000000000F, $vgpr58_vgpr59:0x0000000000000003, $vgpr60_vgpr61:0x000000000000000F, $vgpr62_vgpr63:0x000000000000000F, $sgpr0_sgpr1_sgpr2_sgpr3, $sgpr48_sgpr49, $sgpr58_sgpr59, $sgpr56_sgpr57
; GFX90A-NEXT: {{ $}}
; GFX90A-NEXT: renamable $vgpr0 = nuw nsw V_LSHLREV_B32_e32 3, $vgpr30, implicit $exec
; GFX90A-NEXT: renamable $vgpr1 = V_MOV_B32_e32 0, implicit $exec
; GFX90A-NEXT: renamable $vcc = S_AND_B64 $exec, killed renamable $sgpr50_sgpr51, implicit-def dead $scc
- ; GFX90A-NEXT: S_CBRANCH_VCCNZ %bb.54, implicit $vcc
+ ; GFX90A-NEXT: S_CBRANCH_VCCNZ %bb.55, implicit $vcc
; GFX90A-NEXT: {{ $}}
- ; GFX90A-NEXT: bb.51:
- ; GFX90A-NEXT: successors: %bb.45(0x80000000)
- ; GFX90A-NEXT: liveins: $sgpr14, $sgpr15, $sgpr16, $vgpr18, $vgpr30, $vgpr31, $sgpr4_sgpr5, $sgpr6_sgpr7, $sgpr8_sgpr9:0x000000000000000F, $sgpr10_sgpr11, $sgpr12_sgpr13, $sgpr18_sgpr19, $sgpr24_sgpr25, $sgpr28_sgpr29, $sgpr30_sgpr31, $sgpr34_sgpr35, $sgpr36_sgpr37, $sgpr38_sgpr39, $sgpr40_sgpr41, $sgpr42_sgpr43, $sgpr44_sgpr45, $sgpr48_sgpr49, $sgpr20_sgpr21_sgpr22_sgpr23:0x000000000000003C, $sgpr24_sgpr25_sgpr26_sgpr27:0x00000000000000F0, $vgpr0_vgpr1:0x000000000000000F, $vgpr2_vgpr3:0x000000000000000F, $vgpr40_vgpr41:0x000000000000000F, $vgpr42_vgpr43:0x000000000000000F, $vgpr44_vgpr45:0x000000000000000F, $vgpr46_vgpr47:0x000000000000000F, $vgpr56_vgpr57:0x000000000000000F, $vgpr58_vgpr59:0x0000000000000003, $vgpr60_vgpr61:0x000000000000000F, $vgpr62_vgpr63:0x000000000000000F, $sgpr0_sgpr1_sgpr2_sgpr3, $sgpr54_sgpr55, $sgpr58_sgpr59, $sgpr56_sgpr57
+ ; GFX90A-NEXT: bb.52:
+ ; GFX90A-NEXT: successors: %bb.46(0x80000000)
+ ; GFX90A-NEXT: liveins: $sgpr14, $sgpr15, $sgpr16, $vgpr18, $vgpr30, $vgpr31, $sgpr4_sgpr5, $sgpr6_sgpr7, $sgpr8_sgpr9:0x000000000000000F, $sgpr10_sgpr11, $sgpr12_sgpr13, $sgpr18_sgpr19, $sgpr24_sgpr25, $sgpr28_sgpr29, $sgpr30_sgpr31, $sgpr34_sgpr35, $sgpr36_sgpr37, $sgpr38_sgpr39, $sgpr40_sgpr41, $sgpr42_sgpr43, $sgpr44_sgpr45, $sgpr48_sgpr49, $sgpr20_sgpr21_sgpr22_sgpr23:0x000000000000003C, $sgpr24_sgpr25_sgpr26_sgpr27:0x00000000000000F0, $vgpr0_vgpr1:0x000000000000000F, $vgpr2_vgpr3:0x000000000000000F, $vgpr40_vgpr41:0x000000000000000F, $vgpr42_vgpr43:0x000000000000000F, $vgpr44_vgpr45:0x000000000000000F, $vgpr46_vgpr47:0x000000000000000F, $vgpr56_vgpr57:0x000000000000000F, $vgpr58_vgpr59:0x0000000000000003, $vgpr60_vgpr61:0x000000000000000F, $vgpr62_vgpr63:0x000000000000000F, $sgpr0_sgpr1_sgpr2_sgpr3, $sgpr58_sgpr59, $sgpr56_sgpr57
; GFX90A-NEXT: {{ $}}
; GFX90A-NEXT: renamable $sgpr50_sgpr51 = S_MOV_B64 -1
; GFX90A-NEXT: renamable $sgpr52_sgpr53 = COPY renamable $sgpr36_sgpr37
+ ; GFX90A-NEXT: renamable $sgpr54_sgpr55 = S_MOV_B64 0
; GFX90A-NEXT: renamable $vgpr8_vgpr9 = IMPLICIT_DEF
; GFX90A-NEXT: renamable $vgpr6_vgpr7 = IMPLICIT_DEF
; GFX90A-NEXT: renamable $vgpr4_vgpr5 = IMPLICIT_DEF
@@ -692,20 +701,20 @@ define amdgpu_kernel void @f1(ptr addrspace(1) %arg, ptr addrspace(1) %arg1, i64
; GFX90A-NEXT: renamable $vgpr53 = IMPLICIT_DEF
; GFX90A-NEXT: renamable $vgpr13 = IMPLICIT_DEF implicit-def $vgpr12
; GFX90A-NEXT: renamable $vgpr11 = IMPLICIT_DEF implicit-def $vgpr10
- ; GFX90A-NEXT: S_BRANCH %bb.45
+ ; GFX90A-NEXT: S_BRANCH %bb.46
; GFX90A-NEXT: {{ $}}
- ; GFX90A-NEXT: bb.52.bb80:
- ; GFX90A-NEXT: successors: %bb.59(0x40000000), %bb.53(0x40000000)
+ ; GFX90A-NEXT: bb.53.bb80:
+ ; GFX90A-NEXT: successors: %bb.60(0x40000000), %bb.54(0x40000000)
; GFX90A-NEXT: liveins: $sgpr14, $sgpr15, $sgpr16, $vgpr18, $vgpr30, $vgpr31, $sgpr4_sgpr5, $sgpr6_sgpr7, $sgpr8_sgpr9:0x000000000000000F, $sgpr10_sgpr11, $sgpr12_sgpr13, $sgpr18_sgpr19, $sgpr24_sgpr25, $sgpr28_sgpr29, $sgpr30_sgpr31, $sgpr34_sgpr35, $sgpr36_sgpr37, $sgpr38_sgpr39, $sgpr40_sgpr41, $sgpr42_sgpr43, $sgpr44_sgpr45, $sgpr46_sgpr47:0x000000000000000F, $sgpr48_sgpr49, $sgpr50_sgpr51, $sgpr60_sgpr61, $sgpr64_sgpr65, $sgpr20_sgpr21_sgpr22_sgpr23:0x000000000000003F, $sgpr24_sgpr25_sgpr26_sgpr27:0x00000000000000F0, $vgpr0_vgpr1:0x000000000000000F, $vgpr2_vgpr3:0x000000000000000F, $vgpr4_vgpr5:0x000000000000000F, $vgpr40_vgpr41:0x000000000000000F, $vgpr42_vgpr43:0x000000000000000F, $vgpr44_vgpr45:0x000000000000000F, $vgpr46_vgpr47:0x000000000000000F, $vgpr56_vgpr57:0x000000000000000F, $vgpr58_vgpr59:0x0000000000000003, $vgpr60_vgpr61:0x000000000000000F, $vgpr62_vgpr63:0x000000000000000F, $sgpr0_sgpr1_sgpr2_sgpr3
; GFX90A-NEXT: {{ $}}
; GFX90A-NEXT: renamable $sgpr17 = S_BFE_U32 renamable $sgpr20, 65560, implicit-def dead $scc
; GFX90A-NEXT: S_CMP_EQ_U32 killed renamable $sgpr17, 0, implicit-def $scc
; GFX90A-NEXT: renamable $vgpr6 = V_ADD_CO_U32_e32 4096, $vgpr0, implicit-def $vcc, implicit $exec
; GFX90A-NEXT: renamable $vgpr7, dead renamable $sgpr52_sgpr53 = V_ADDC_U32_e64 0, 0, killed $vcc, 0, implicit $exec
- ; GFX90A-NEXT: S_CBRANCH_SCC1 %bb.59, implicit killed $scc
+ ; GFX90A-NEXT: S_CBRANCH_SCC1 %bb.60, implicit killed $scc
; GFX90A-NEXT: {{ $}}
- ; GFX90A-NEXT: bb.53:
- ; GFX90A-NEXT: successors: %bb.61(0x80000000)
+ ; GFX90A-NEXT: bb.54:
+ ; GFX90A-NEXT: successors: %bb.62(0x80000000)
; GFX90A-NEXT: liveins: $sgpr14, $sgpr15, $sgpr16, $vgpr18, $vgpr30, $vgpr31, $sgpr4_sgpr5, $sgpr6_sgpr7, $sgpr8_sgpr9:0x000000000000000F, $sgpr10_sgpr11, $sgpr12_sgpr13, $sgpr18_sgpr19, $sgpr24_sgpr25, $sgpr28_sgpr29, $sgpr30_sgpr31, $sgpr34_sgpr35, $sgpr36_sgpr37, $sgpr38_sgpr39, $sgpr40_sgpr41, $sgpr42_sgpr43, $sgpr44_sgpr45, $sgpr48_sgpr49, $sgpr50_sgpr51, $sgpr60_sgpr61, $sgpr20_sgpr21_sgpr22_sgpr23:0x000000000000003C, $sgpr24_sgpr25_sgpr26_sgpr27:0x00000000000000F0, $vgpr0_vgpr1:0x000000000000000F, $vgpr2_vgpr3:0x000000000000000F, $vgpr4_vgpr5:0x000000000000000F, $vgpr6_vgpr7:0x000000000000000F, $vgpr40_vgpr41:0x000000000000000F, $vgpr42_vgpr43:0x000000000000000F, $vgpr44_vgpr45:0x000000000000000F, $vgpr46_vgpr47:0x000000000000000F, $vgpr56_vgpr57:0x000000000000000F, $vgpr58_vgpr59:0x0000000000000003, $vgpr60_vgpr61:0x000000000000000F, $vgpr62_vgpr63:0x000000000000000F, $sgpr0_sgpr1_sgpr2_sgpr3
; GFX90A-NEXT: {{ $}}
; GFX90A-NEXT: renamable $sgpr52_sgpr53 = S_MOV_B64 0
@@ -720,17 +729,18 @@ define amdgpu_kernel void @f1(ptr addrspace(1) %arg, ptr addrspace(1) %arg1, i64
; GFX90A-NEXT: renamable $vgpr53 = IMPLICIT_DEF
; GFX90A-NEXT: renamable $vgpr13 = IMPLICIT_DEF implicit-def $vgpr12
; GFX90A-NEXT: renamable $vgpr11 = IMPLICIT_DEF implicit-def $vgpr10
- ; GFX90A-NEXT: S_BRANCH %bb.61
+ ; GFX90A-NEXT: S_BRANCH %bb.62
; GFX90A-NEXT: {{ $}}
- ; GFX90A-NEXT: bb.54.bb73:
- ; GFX90A-NEXT: successors: %bb.52(0x40000000), %bb.55(0x40000000)
- ; GFX90A-NEXT: liveins: $sgpr14, $sgpr15, $sgpr16, $vgpr18, $vgpr30, $vgpr31, $sgpr4_sgpr5, $sgpr6_sgpr7, $sgpr8_sgpr9:0x000000000000000F, $sgpr10_sgpr11, $sgpr12_sgpr13, $sgpr18_sgpr19, $sgpr24_sgpr25, $sgpr28_sgpr29, $sgpr30_sgpr31, $sgpr34_sgpr35, $sgpr36_sgpr37, $sgpr38_sgpr39, $sgpr40_sgpr41, $sgpr42_sgpr43, $sgpr44_sgpr45, $sgpr46_sgpr47:0x000000000000000F, $sgpr48_sgpr49, $sgpr64_sgpr65, $sgpr20_sgpr21_sgpr22_sgpr23:0x000000000000003F, $sgpr24_sgpr25_sgpr26_sgpr27:0x00000000000000F0, $vgpr0_vgpr1:0x000000000000000F, $vgpr2_vgpr3:0x000000000000000F, $vgpr40_vgpr41:0x000000000000000F, $vgpr42_vgpr43:0x000000000000000F, $vgpr44_vgpr45:0x000000000000000F, $vgpr46_vgpr47:0x000000000000000F, $vgpr56_vgpr57:0x000000000000000F, $vgpr58_vgpr59:0x0000000000000003, $vgpr60_vgpr61:0x000000000000000F, $vgpr62_vgpr63:0x000000000000000F, $sgpr0_sgpr1_sgpr2_sgpr3, $sgpr54_sgpr55
+ ; GFX90A-NEXT: bb.55.bb73:
+ ; GFX90A-NEXT: successors: %bb.53(0x40000000), %bb.56(0x40000000)
+ ; GFX90A-NEXT: liveins: $sgpr14, $sgpr15, $sgpr16, $vgpr18, $vgpr30, $vgpr31, $sgpr4_sgpr5, $sgpr6_sgpr7, $sgpr8_sgpr9:0x000000000000000F, $sgpr10_sgpr11, $sgpr12_sgpr13, $sgpr18_sgpr19, $sgpr24_sgpr25, $sgpr28_sgpr29, $sgpr30_sgpr31, $sgpr34_sgpr35, $sgpr36_sgpr37, $sgpr38_sgpr39, $sgpr40_sgpr41, $sgpr42_sgpr43, $sgpr44_sgpr45, $sgpr46_sgpr47:0x000000000000000F, $sgpr48_sgpr49, $sgpr64_sgpr65, $sgpr20_sgpr21_sgpr22_sgpr23:0x000000000000003F, $sgpr24_sgpr25_sgpr26_sgpr27:0x00000000000000F0, $vgpr0_vgpr1:0x000000000000000F, $vgpr2_vgpr3:0x000000000000000F, $vgpr40_vgpr41:0x000000000000000F, $vgpr42_vgpr43:0x000000000000000F, $vgpr44_vgpr45:0x000000000000000F, $vgpr46_vgpr47:0x000000000000000F, $vgpr56_vgpr57:0x000000000000000F, $vgpr58_vgpr59:0x0000000000000003, $vgpr60_vgpr61:0x000000000000000F, $vgpr62_vgpr63:0x000000000000000F, $sgpr0_sgpr1_sgpr2_sgpr3
; GFX90A-NEXT: {{ $}}
; GFX90A-NEXT: renamable $vgpr6 = FLAT_LOAD_UBYTE renamable $vgpr0_vgpr1, 2048, 0, implicit $exec, implicit $flat_scr :: (load (s8) from %ir.i76)
; GFX90A-NEXT: renamable $vgpr4 = V_ADD_CO_U32_e32 2048, $vgpr0, implicit-def $vcc, implicit $exec
; GFX90A-NEXT: renamable $sgpr50_sgpr51 = S_MOV_B64 0
; GFX90A-NEXT: renamable $sgpr56_sgpr57 = S_MOV_B64 -1
; GFX90A-NEXT: renamable $sgpr52_sgpr53 = COPY renamable $sgpr36_sgpr37
+ ; GFX90A-NEXT: renamable $sgpr54_sgpr55 = S_MOV_B64 0
; GFX90A-NEXT: renamable $vgpr5, dead renamable $sgpr58_sgpr59 = V_ADDC_U32_e64 0, 0, killed $vcc, 0, implicit $exec
; GFX90A-NEXT: renamable $vcc = V_CMP_EQ_U16_e64 0, killed $vgpr6, implicit $exec
; GFX90A-NEXT: renamable $sgpr58_sgpr59 = S_MOV_B64 0
@@ -745,17 +755,17 @@ define amdgpu_kernel void @f1(ptr addrspace(1) %arg, ptr addrspace(1) %arg1, i64
; GFX90A-NEXT: renamable $vgpr13 = IMPLICIT_DEF implicit-def $vgpr12
; GFX90A-NEXT: renamable $vgpr11 = IMPLICIT_DEF implicit-def $vgpr10
; GFX90A-NEXT: $sgpr60_sgpr61 = S_AND_SAVEEXEC_B64 $vcc, implicit-def $exec, implicit-def $scc, implicit $exec
- ; GFX90A-NEXT: S_CBRANCH_EXECNZ %bb.52, implicit $exec
+ ; GFX90A-NEXT: S_CBRANCH_EXECNZ %bb.53, implicit $exec
; GFX90A-NEXT: {{ $}}
- ; GFX90A-NEXT: bb.55.Flow29:
- ; GFX90A-NEXT: successors: %bb.45(0x80000000)
+ ; GFX90A-NEXT: bb.56.Flow29:
+ ; GFX90A-NEXT: successors: %bb.46(0x80000000)
; GFX90A-NEXT: liveins: $sgpr14, $sgpr15, $sgpr16, $vgpr15, $vgpr17, $vgpr18, $vgpr30, $vgpr31, $vgpr52, $vgpr53, $sgpr4_sgpr5, $sgpr6_sgpr7, $sgpr8_sgpr9:0x000000000000000F, $sgpr10_sgpr11, $sgpr12_sgpr13, $sgpr18_sgpr19, $sgpr24_sgpr25, $sgpr28_sgpr29, $sgpr30_sgpr31, $sgpr34_sgpr35, $sgpr36_sgpr37, $sgpr38_sgpr39, $sgpr40_sgpr41, $sgpr42_sgpr43, $sgpr44_sgpr45, $sgpr48_sgpr49, $sgpr50_sgpr51, $sgpr52_sgpr53, $sgpr54_sgpr55, $sgpr56_sgpr57, $sgpr58_sgpr59, $sgpr60_sgpr61, $sgpr20_sgpr21_sgpr22_sgpr23:0x000000000000003C, $sgpr24_sgpr25_sgpr26_sgpr27:0x00000000000000F0, $vgpr0_vgpr1:0x000000000000000F, $vgpr2_vgpr3:0x000000000000000F, $vgpr4_vgpr5:0x000000000000000F, $vgpr6_vgpr7:0x000000000000000F, $vgpr8_vgpr9:0x000000000000000F, $vgpr10_vgpr11:0x000000000000000F, $vgpr12_vgpr13:0x000000000000000F, $vgpr14_vgpr15:0x0000000000000003, $vgpr16_vgpr17:0x0000000000000003, $vgpr40_vgpr41:0x000000000000000F, $vgpr42_vgpr43:0x000000000000000F, $vgpr44_vgpr45:0x000000000000000F, $vgpr46_vgpr47:0x000000000000000F, $vgpr56_vgpr57:0x000000000000000F, $vgpr58_vgpr59:0x0000000000000003, $vgpr60_vgpr61:0x000000000000000F, $vgpr62_vgpr63:0x000000000000000F, $sgpr0_sgpr1_sgpr2_sgpr3
; GFX90A-NEXT: {{ $}}
; GFX90A-NEXT: $exec = S_OR_B64 $exec, killed renamable $sgpr60_sgpr61, implicit-def $scc
- ; GFX90A-NEXT: S_BRANCH %bb.45
+ ; GFX90A-NEXT: S_BRANCH %bb.46
; GFX90A-NEXT: {{ $}}
- ; GFX90A-NEXT: bb.56.bb90:
- ; GFX90A-NEXT: successors: %bb.60(0x80000000)
+ ; GFX90A-NEXT: bb.57.bb90:
+ ; GFX90A-NEXT: successors: %bb.61(0x80000000)
; GFX90A-NEXT: liveins: $sgpr14, $sgpr15, $sgpr16, $vgpr18, $vgpr30, $vgpr31, $sgpr4_sgpr5, $sgpr6_sgpr7, $sgpr8_sgpr9:0x000000000000000F, $sgpr10_sgpr11, $sgpr12_sgpr13, $sgpr18_sgpr19, $sgpr24_sgpr25, $sgpr28_sgpr29, $sgpr30_sgpr31, $sgpr34_sgpr35, $sgpr36_sgpr37, $sgpr38_sgpr39, $sgpr40_sgpr41, $sgpr42_sgpr43, $sgpr44_sgpr45, $sgpr46_sgpr47:0x000000000000000F, $sgpr48_sgpr49, $sgpr50_sgpr51, $sgpr54_sgpr55, $sgpr60_sgpr61, $sgpr64_sgpr65, $sgpr20_sgpr21_sgpr22_sgpr23:0x000000000000003C, $sgpr24_sgpr25_sgpr26_sgpr27:0x00000000000000F0, $vgpr0_vgpr1:0x000000000000000F, $vgpr2_vgpr3:0x000000000000000F, $vgpr4_vgpr5:0x000000000000000F, $vgpr6_vgpr7:0x000000000000000F, $vgpr8_vgpr9:0x000000000000000F, $vgpr40_vgpr41:0x000000000000000F, $vgpr42_vgpr43:0x000000000000000F, $vgpr44_vgpr45:0x000000000000000F, $vgpr46_vgpr47:0x000000000000000F, $vgpr56_vgpr57:0x000000000000000F, $vgpr58_vgpr59:0x0000000000000003, $vgpr60_vgpr61:0x000000000000000F, $vgpr62_vgpr63:0x000000000000000F, $sgpr0_sgpr1_sgpr2_sgpr3
; GFX90A-NEXT: {{ $}}
; GFX90A-NEXT: renamable $vgpr53 = V_CNDMASK_B32_e64 0, 0, 0, 1, killed $sgpr64_sgpr65, implicit $exec
@@ -773,9 +783,9 @@ define amdgpu_kernel void @f1(ptr addrspace(1) %arg, ptr addrspace(1) %arg1, i64
; GFX90A-NEXT: renamable $sgpr52_sgpr53 = S_XOR_B64 $exec, -1, implicit-def dead $scc
; GFX90A-NEXT: renamable $sgpr62_sgpr63 = S_OR_B64 renamable $sgpr36_sgpr37, $exec, implicit-def dead $scc
; GFX90A-NEXT: renamable $vgpr10 = COPY renamable $vgpr14, implicit $exec
- ; GFX90A-NEXT: S_BRANCH %bb.60
+ ; GFX90A-NEXT: S_BRANCH %bb.61
; GFX90A-NEXT: {{ $}}
- ; GFX90A-NEXT: bb.57:
+ ; GFX90A-NEXT: bb.58:
; GFX90A-NEXT: successors: %bb.7(0x80000000)
; GFX90A-NEXT: liveins: $sgpr14, $sgpr15, $sgpr16, $vgpr15, $vgpr17, $vgpr30, $vgpr31, $sgpr4_sgpr5, $sgpr6_sgpr7, $sgpr8_sgpr9:0x000000000000000F, $sgpr10_sgpr11, $sgpr12_sgpr13, $sgpr18_sgpr19, $sgpr28_sgpr29, $sgpr30_sgpr31, $sgpr36_sgpr37, $sgpr20_sgpr21_sgpr22_sgpr23:0x000000000000003C, $sgpr24_sgpr25_sgpr26_sgpr27:0x00000000000000F0, $vgpr2_vgpr3:0x000000000000000F, $vgpr10_vgpr11:0x000000000000000F, $vgpr18_vgpr19:0x000000000000000F, $vgpr20_vgpr21:0x000000000000000F, $vgpr22_vgpr23:0x000000000000000F, $vgpr24_vgpr25:0x000000000000000F, $sgpr0_sgpr1_sgpr2_sgpr3
; GFX90A-NEXT: {{ $}}
@@ -810,7 +820,7 @@ define amdgpu_kernel void @f1(ptr addrspace(1) %arg, ptr addrspace(1) %arg1, i64
; GFX90A-NEXT: renamable $sgpr34_sgpr35 = S_MOV_B64 0
; GFX90A-NEXT: S_BRANCH %bb.7
; GFX90A-NEXT: {{ $}}
- ; GFX90A-NEXT: bb.58.bb105:
+ ; GFX90A-NEXT: bb.59.bb105:
; GFX90A-NEXT: successors: %bb.3(0x80000000)
; GFX90A-NEXT: liveins: $sgpr14, $sgpr15, $sgpr16, $sgpr17, $sgpr33, $vgpr31, $sgpr4_sgpr5, $sgpr6_sgpr7, $sgpr8_sgpr9:0x000000000000000F, $sgpr10_sgpr11, $sgpr12_sgpr13, $sgpr18_sgpr19, $sgpr28_sgpr29, $sgpr30_sgpr31, $sgpr34_sgpr35, $sgpr46_sgpr47:0x000000000000000F, $sgpr20_sgpr21_sgpr22_sgpr23:0x00000000000000FF, $sgpr24_sgpr25_sgpr26_sgpr27:0x00000000000000FF, $vgpr2_vgpr3:0x000000000000000F, $sgpr0_sgpr1_sgpr2_sgpr3
; GFX90A-NEXT: {{ $}}
@@ -827,8 +837,8 @@ define amdgpu_kernel void @f1(ptr addrspace(1) %arg, ptr addrspace(1) %arg1, i64
; GFX90A-NEXT: renamable $sgpr36_sgpr37 = S_MOV_B64 -1
; GFX90A-NEXT: S_BRANCH %bb.3
; GFX90A-NEXT: {{ $}}
- ; GFX90A-NEXT: bb.59.bb85:
- ; GFX90A-NEXT: successors: %bb.56(0x40000000), %bb.60(0x40000000)
+ ; GFX90A-NEXT: bb.60.bb85:
+ ; GFX90A-NEXT: successors: %bb.57(0x40000000), %bb.61(0x40000000)
; GFX90A-NEXT: liveins: $sgpr14, $sgpr15, $sgpr16, $vgpr18, $vgpr30, $vgpr31, $sgpr4_sgpr5, $sgpr6_sgpr7, $sgpr8_sgpr9:0x000000000000000F, $sgpr10_sgpr11, $sgpr12_sgpr13, $sgpr18_sgpr19, $sgpr24_sgpr25, $sgpr28_sgpr29, $sgpr30_sgpr31, $sgpr34_sgpr35, $sgpr36_sgpr37, $sgpr38_sgpr39, $sgpr40_sgpr41, $sgpr42_sgpr43, $sgpr44_sgpr45, $sgpr46_sgpr47:0x000000000000000F, $sgpr48_sgpr49, $sgpr50_sgpr51, $sgpr60_sgpr61, $sgpr64_sgpr65, $sgpr20_sgpr21_sgpr22_sgpr23:0x000000000000003C, $sgpr24_sgpr25_sgpr26_sgpr27:0x00000000000000F0, $vgpr0_vgpr1:0x000000000000000F, $vgpr2_vgpr3:0x000000000000000F, $vgpr4_vgpr5:0x000000000000000F, $vgpr6_vgpr7:0x000000000000000F, $vgpr40_vgpr41:0x000000000000000F, $vgpr42_vgpr43:0x000000000000000F, $vgpr44_vgpr45:0x000000000000000F, $vgpr46_vgpr47:0x000000000000000F, $vgpr56_vgpr57:0x000000000000000F, $vgpr58_vgpr59:0x0000000000000003, $vgpr60_vgpr61:0x000000000000000F, $vgpr62_vgpr63:0x000000000000000F, $sgpr0_sgpr1_sgpr2_sgpr3
; GFX90A-NEXT: {{ $}}
; GFX90A-NEXT: renamable $vgpr8 = V_OR_B32_e32 1, $vgpr6, implicit $exec
@@ -846,17 +856,17 @@ define amdgpu_kernel void @f1(ptr addrspace(1) %arg, ptr addrspace(1) %arg1, i64
; GFX90A-NEXT: renamable $vgpr13 = IMPLICIT_DEF implicit-def $vgpr12
; GFX90A-NEXT: renamable $vgpr11 = IMPLICIT_DEF implicit-def $vgpr10
; GFX90A-NEXT: $sgpr54_sgpr55 = S_AND_SAVEEXEC_B64 $vcc, implicit-def $exec, implicit-def $scc, implicit $exec
- ; GFX90A-NEXT: S_CBRANCH_EXECNZ %bb.56, implicit $exec
+ ; GFX90A-NEXT: S_CBRANCH_EXECNZ %bb.57, implicit $exec
; GFX90A-NEXT: {{ $}}
- ; GFX90A-NEXT: bb.60.Flow31:
- ; GFX90A-NEXT: successors: %bb.61(0x80000000)
+ ; GFX90A-NEXT: bb.61.Flow31:
+ ; GFX90A-NEXT: successors: %bb.62(0x80000000)
; GFX90A-NEXT: liveins: $sgpr14, $sgpr15, $sgpr16, $vgpr15, $vgpr17, $vgpr18, $vgpr30, $vgpr31, $vgpr52, $vgpr53, $sgpr4_sgpr5, $sgpr6_sgpr7, $sgpr8_sgpr9:0x000000000000000F, $sgpr10_sgpr11, $sgpr12_sgpr13, $sgpr18_sgpr19, $sgpr24_sgpr25, $sgpr28_sgpr29, $sgpr30_sgpr31, $sgpr34_sgpr35, $sgpr36_sgpr37, $sgpr38_sgpr39, $sgpr40_sgpr41, $sgpr42_sgpr43, $sgpr44_sgpr45, $sgpr48_sgpr49, $sgpr50_sgpr51, $sgpr52_sgpr53, $sgpr54_sgpr55, $sgpr60_sgpr61, $sgpr62_sgpr63, $sgpr20_sgpr21_sgpr22_sgpr23:0x000000000000003C, $sgpr24_sgpr25_sgpr26_sgpr27:0x00000000000000F0, $vgpr0_vgpr1:0x000000000000000F, $vgpr2_vgpr3:0x000000000000000F, $vgpr4_vgpr5:0x000000000000000F, $vgpr6_vgpr7:0x000000000000000F, $vgpr8_vgpr9:0x000000000000000F, $vgpr10_vgpr11:0x000000000000000F, $vgpr12_vgpr13:0x000000000000000F, $vgpr14_vgpr15:0x0000000000000003, $vgpr16_vgpr17:0x0000000000000003, $vgpr40_vgpr41:0x000000000000000F, $vgpr42_vgpr43:0x000000000000000F, $vgpr44_vgpr45:0x000000000000000F, $vgpr46_vgpr47:0x000000000000000F, $vgpr56_vgpr57:0x000000000000000F, $vgpr58_vgpr59:0x0000000000000003, $vgpr60_vgpr61:0x000000000000000F, $vgpr62_vgpr63:0x000000000000000F, $sgpr0_sgpr1_sgpr2_sgpr3
; GFX90A-NEXT: {{ $}}
; GFX90A-NEXT: $exec = S_OR_B64 $exec, killed renamable $sgpr54_sgpr55, implicit-def $scc
; GFX90A-NEXT: renamable $sgpr46_sgpr47 = S_MOV_B64 0
; GFX90A-NEXT: {{ $}}
- ; GFX90A-NEXT: bb.61.Flow30:
- ; GFX90A-NEXT: successors: %bb.55(0x80000000)
+ ; GFX90A-NEXT: bb.62.Flow30:
+ ; GFX90A-NEXT: successors: %bb.56(0x80000000)
; GFX90A-NEXT: liveins: $sgpr14, $sgpr15, $sgpr16, $vgpr15, $vgpr17, $vgpr18, $vgpr30, $vgpr31, $vgpr52, $vgpr53, $sgpr4_sgpr5, $sgpr6_sgpr7, $sgpr8_sgpr9:0x000000000000000F, $sgpr10_sgpr11, $sgpr12_sgpr13, $sgpr18_sgpr19, $sgpr24_sgpr25, $sgpr28_sgpr29, $sgpr30_sgpr31, $sgpr34_sgpr35, $sgpr36_sgpr37, $sgpr38_sgpr39, $sgpr40_sgpr41, $sgpr42_sgpr43, $sgpr44_sgpr45, $sgpr46_sgpr47, $sgpr48_sgpr49, $sgpr50_sgpr51, $sgpr52_sgpr53, $sgpr60_sgpr61, $sgpr62_sgpr63, $sgpr20_sgpr21_sgpr22_sgpr23:0x000000000000003C, $sgpr24_sgpr25_sgpr26_sgpr27:0x00000000000000F0, $vgpr0_vgpr1:0x000000000000000F, $vgpr2_vgpr3:0x000000000000000F, $vgpr4_vgpr5:0x000000000000000F, $vgpr6_vgpr7:0x000000000000000F, $vgpr8_vgpr9:0x000000000000000F, $vgpr10_vgpr11:0x000000000000000F, $vgpr12_vgpr13:0x000000000000000F, $vgpr14_vgpr15:0x0000000000000003, $vgpr16_vgpr17:0x0000000000000003, $vgpr40_vgpr41:0x000000000000000F, $vgpr42_vgpr43:0x000000000000000F, $vgpr44_vgpr45:0x000000000000000F, $vgpr46_vgpr47:0x000000000000000F, $vgpr56_vgpr57:0x000000000000000F, $vgpr58_vgpr59:0x0000000000000003, $vgpr60_vgpr61:0x000000000000000F, $vgpr62_vgpr63:0x000000000000000F, $sgpr0_sgpr1_sgpr2_sgpr3
; GFX90A-NEXT: {{ $}}
; GFX90A-NEXT: renamable $sgpr56_sgpr57 = S_XOR_B64 $exec, -1, implicit-def dead $scc
@@ -865,48 +875,48 @@ define amdgpu_kernel void @f1(ptr addrspace(1) %arg, ptr addrspace(1) %arg1, i64
; GFX90A-NEXT: renamable $sgpr46_sgpr47 = S_ANDN2_B64 renamable $sgpr36_sgpr37, $exec, implicit-def dead $scc
; GFX90A-NEXT: renamable $sgpr52_sgpr53 = S_AND_B64 killed renamable $sgpr62_sgpr63, $exec, implicit-def dead $scc
; GFX90A-NEXT: renamable $sgpr52_sgpr53 = S_OR_B64 killed renamable $sgpr46_sgpr47, killed renamable $sgpr52_sgpr53, implicit-def dead $scc
- ; GFX90A-NEXT: S_BRANCH %bb.55
+ ; GFX90A-NEXT: S_BRANCH %bb.56
; GFX90A-NEXT: {{ $}}
- ; GFX90A-NEXT: bb.62.bb140:
- ; GFX90A-NEXT: successors: %bb.68(0x40000000), %bb.63(0x40000000)
+ ; GFX90A-NEXT: bb.63.bb140:
+ ; GFX90A-NEXT: successors: %bb.69(0x40000000), %bb.64(0x40000000)
; GFX90A-NEXT: liveins: $sgpr14, $sgpr15, $sgpr16, $vgpr15, $vgpr17, $vgpr30, $vgpr31, $vgpr52, $vgpr53, $sgpr4_sgpr5, $sgpr6_sgpr7, $sgpr8_sgpr9:0x000000000000000F, $sgpr10_sgpr11, $sgpr12_sgpr13, $sgpr18_sgpr19, $sgpr24_sgpr25, $sgpr28_sgpr29, $sgpr30_sgpr31, $sgpr34_sgpr35, $sgpr38_sgpr39, $sgpr40_sgpr41, $sgpr42_sgpr43, $sgpr44_sgpr45, $sgpr48_sgpr49, $sgpr50_sgpr51, $sgpr52_sgpr53, $sgpr54_sgpr55, $sgpr64_sgpr65, $sgpr66_sgpr67, $sgpr20_sgpr21_sgpr22_sgpr23:0x000000000000003C, $sgpr24_sgpr25_sgpr26_sgpr27:0x00000000000000F0, $vgpr0_vgpr1:0x000000000000000F, $vgpr2_vgpr3:0x000000000000000F, $vgpr4_vgpr5:0x000000000000000F, $vgpr6_vgpr7:0x000000000000000F, $vgpr8_vgpr9:0x000000000000000F, $vgpr10_vgpr11:0x000000000000000F, $vgpr12_vgpr13:0x000000000000000F, $vgpr14_vgpr15:0x0000000000000003, $vgpr16_vgpr17:0x0000000000000003, $vgpr18_vgpr19:0x000000000000000F, $vgpr20_vgpr21:0x000000000000000F, $vgpr22_vgpr23:0x000000000000000F, $vgpr24_vgpr25:0x000000000000000F, $vgpr40_vgpr41:0x000000000000000F, $vgpr42_vgpr43:0x000000000000000F, $vgpr44_vgpr45:0x000000000000000F, $vgpr46_vgpr47:0x000000000000000F, $vgpr56_vgpr57:0x000000000000000F, $vgpr58_vgpr59:0x000000000000000F, $vgpr60_vgpr61:0x000000000000000F, $vgpr62_vgpr63:0x000000000000000F, $sgpr0_sgpr1_sgpr2_sgpr3
; GFX90A-NEXT: {{ $}}
; GFX90A-NEXT: renamable $sgpr36_sgpr37 = S_MOV_B64 -1
; GFX90A-NEXT: renamable $vcc = S_AND_B64 $exec, killed renamable $sgpr30_sgpr31, implicit-def dead $scc
- ; GFX90A-NEXT: S_CBRANCH_VCCNZ %bb.68, implicit $vcc
+ ; GFX90A-NEXT: S_CBRANCH_VCCNZ %bb.69, implicit $vcc
; GFX90A-NEXT: {{ $}}
- ; GFX90A-NEXT: bb.63.Flow13:
- ; GFX90A-NEXT: successors: %bb.64(0x40000000), %bb.66(0x40000000)
+ ; GFX90A-NEXT: bb.64.Flow13:
+ ; GFX90A-NEXT: successors: %bb.65(0x40000000), %bb.67(0x40000000)
; GFX90A-NEXT: liveins: $sgpr14, $sgpr15, $sgpr16, $vgpr15, $vgpr17, $vgpr30, $vgpr31, $vgpr52, $vgpr53, $sgpr4_sgpr5, $sgpr6_sgpr7, $sgpr8_sgpr9:0x000000000000000F, $sgpr10_sgpr11, $sgpr18_sgpr19, $sgpr24_sgpr25, $sgpr34_sgpr35, $sgpr36_sgpr37, $sgpr38_sgpr39, $sgpr40_sgpr41, $sgpr42_sgpr43, $sgpr44_sgpr45, $sgpr48_sgpr49, $sgpr50_sgpr51, $sgpr52_sgpr53, $sgpr54_sgpr55, $sgpr64_sgpr65, $sgpr66_sgpr67, $vgpr0_vgpr1:0x000000000000000F, $vgpr4_vgpr5:0x000000000000000F, $vgpr6_vgpr7:0x000000000000000F, $vgpr8_vgpr9:0x000000000000000F, $vgpr10_vgpr11:0x000000000000000C, $vgpr12_vgpr13:0x000000000000000C, $vgpr18_vgpr19:0x000000000000000C, $vgpr20_vgpr21:0x000000000000000C, $vgpr22_vgpr23:0x000000000000000C, $vgpr24_vgpr25:0x000000000000000C, $vgpr40_vgpr41:0x000000000000000F, $vgpr42_vgpr43:0x000000000000000F, $vgpr44_vgpr45:0x000000000000000F, $vgpr46_vgpr47:0x000000000000000F, $vgpr56_vgpr57:0x000000000000000F, $vgpr58_vgpr59:0x000000000000000F, $vgpr60_vgpr61:0x000000000000000F, $vgpr62_vgpr63:0x000000000000000F, $sgpr0_sgpr1_sgpr2_sgpr3
; GFX90A-NEXT: {{ $}}
; GFX90A-NEXT: $vcc = S_ANDN2_B64 $exec, killed renamable $sgpr36_sgpr37, implicit-def dead $scc
- ; GFX90A-NEXT: S_CBRANCH_VCCNZ %bb.66, implicit $vcc
+ ; GFX90A-NEXT: S_CBRANCH_VCCNZ %bb.67, implicit $vcc
; GFX90A-NEXT: {{ $}}
- ; GFX90A-NEXT: bb.64.bb159:
- ; GFX90A-NEXT: successors: %bb.67(0x40000000), %bb.65(0x40000000)
+ ; GFX90A-NEXT: bb.65.bb159:
+ ; GFX90A-NEXT: successors: %bb.68(0x40000000), %bb.66(0x40000000)
; GFX90A-NEXT: liveins: $sgpr14, $sgpr15, $sgpr16, $vgpr15, $vgpr17, $vgpr30, $vgpr31, $vgpr52, $vgpr53, $sgpr4_sgpr5, $sgpr6_sgpr7, $sgpr8_sgpr9:0x000000000000000F, $sgpr10_sgpr11, $sgpr18_sgpr19, $sgpr24_sgpr25, $sgpr34_sgpr35, $sgpr38_sgpr39, $sgpr40_sgpr41, $sgpr42_sgpr43, $sgpr44_sgpr45, $sgpr48_sgpr49, $sgpr50_sgpr51, $sgpr52_sgpr53, $sgpr54_sgpr55, $sgpr64_sgpr65, $sgpr66_sgpr67, $vgpr0_vgpr1:0x000000000000000F, $vgpr4_vgpr5:0x000000000000000F, $vgpr6_vgpr7:0x000000000000000F, $vgpr8_vgpr9:0x000000000000000F, $vgpr10_vgpr11:0x000000000000000C, $vgpr12_vgpr13:0x000000000000000C, $vgpr18_vgpr19:0x000000000000000C, $vgpr20_vgpr21:0x000000000000000C, $vgpr22_vgpr23:0x000000000000000C, $vgpr24_vgpr25:0x000000000000000C, $vgpr40_vgpr41:0x000000000000000F, $vgpr42_vgpr43:0x000000000000000F, $vgpr44_vgpr45:0x000000000000000F, $vgpr46_vgpr47:0x000000000000000F, $vgpr56_vgpr57:0x000000000000000F, $vgpr58_vgpr59:0x000000000000000F, $vgpr60_vgpr61:0x000000000000000F, $vgpr62_vgpr63:0x000000000000000F, $sgpr0_sgpr1_sgpr2_sgpr3
; GFX90A-NEXT: {{ $}}
; GFX90A-NEXT: renamable $vcc = V_CMP_NE_U32_e64 0, killed $vgpr30, implicit $exec
; GFX90A-NEXT: $sgpr12_sgpr13 = S_AND_SAVEEXEC_B64 $vcc, implicit-def $exec, implicit-def $scc, implicit $exec
; GFX90A-NEXT: renamable $sgpr12_sgpr13 = S_XOR_B64 $exec, killed renamable $sgpr12_sgpr13, implicit-def dead $scc
- ; GFX90A-NEXT: S_CBRANCH_EXECNZ %bb.67, implicit $exec
+ ; GFX90A-NEXT: S_CBRANCH_EXECNZ %bb.68, implicit $exec
; GFX90A-NEXT: {{ $}}
- ; GFX90A-NEXT: bb.65.Flow10:
- ; GFX90A-NEXT: successors: %bb.66(0x80000000)
+ ; GFX90A-NEXT: bb.66.Flow10:
+ ; GFX90A-NEXT: successors: %bb.67(0x80000000)
; GFX90A-NEXT: liveins: $sgpr14, $sgpr15, $sgpr16, $vgpr31, $sgpr4_sgpr5, $sgpr6_sgpr7, $sgpr8_sgpr9:0x000000000000000F, $sgpr10_sgpr11, $sgpr12_sgpr13, $sgpr18_sgpr19, $sgpr24_sgpr25, $sgpr34_sgpr35, $sgpr38_sgpr39, $sgpr40_sgpr41, $sgpr42_sgpr43, $sgpr44_sgpr45, $sgpr48_sgpr49, $sgpr50_sgpr51, $sgpr52_sgpr53, $sgpr54_sgpr55, $sgpr64_sgpr65, $sgpr66_sgpr67, $vgpr0_vgpr1:0x000000000000000F, $vgpr4_vgpr5:0x000000000000000F, $vgpr6_vgpr7:0x000000000000000F, $vgpr8_vgpr9:0x000000000000000F, $vgpr40_vgpr41:0x000000000000000F, $vgpr42_vgpr43:0x000000000000000F, $vgpr44_vgpr45:0x000000000000000F, $vgpr46_vgpr47:0x000000000000000F, $vgpr56_vgpr57:0x000000000000000F, $vgpr58_vgpr59:0x000000000000000F, $vgpr60_vgpr61:0x000000000000000F, $vgpr62_vgpr63:0x000000000000000F, $sgpr0_sgpr1_sgpr2_sgpr3
; GFX90A-NEXT: {{ $}}
; GFX90A-NEXT: $sgpr12_sgpr13 = S_ANDN2_SAVEEXEC_B64 $sgpr12_sgpr13, implicit-def $exec, implicit-def $scc, implicit $exec
; GFX90A-NEXT: $exec = S_OR_B64 $exec, killed renamable $sgpr12_sgpr13, implicit-def $scc
; GFX90A-NEXT: {{ $}}
- ; GFX90A-NEXT: bb.66.Flow14:
+ ; GFX90A-NEXT: bb.67.Flow14:
; GFX90A-NEXT: successors: %bb.8(0x80000000)
; GFX90A-NEXT: liveins: $sgpr14, $sgpr15, $sgpr16, $vgpr31, $sgpr4_sgpr5, $sgpr6_sgpr7, $sgpr8_sgpr9:0x000000000000000F, $sgpr10_sgpr11, $sgpr18_sgpr19, $sgpr24_sgpr25, $sgpr34_sgpr35, $sgpr38_sgpr39, $sgpr40_sgpr41, $sgpr42_sgpr43, $sgpr44_sgpr45, $sgpr48_sgpr49, $sgpr50_sgpr51, $sgpr52_sgpr53, $sgpr54_sgpr55, $sgpr64_sgpr65, $sgpr66_sgpr67, $vgpr0_vgpr1:0x000000000000000F, $vgpr4_vgpr5:0x000000000000000F, $vgpr6_vgpr7:0x000000000000000F, $vgpr8_vgpr9:0x000000000000000F, $vgpr40_vgpr41:0x000000000000000F, $vgpr42_vgpr43:0x000000000000000F, $vgpr44_vgpr45:0x000000000000000F, $vgpr46_vgpr47:0x000000000000000F, $vgpr56_vgpr57:0x000000000000000F, $vgpr58_vgpr59:0x000000000000000F, $vgpr60_vgpr61:0x000000000000000F, $vgpr62_vgpr63:0x000000000000000F, $sgpr0_sgpr1_sgpr2_sgpr3
; GFX90A-NEXT: {{ $}}
; GFX90A-NEXT: renamable $sgpr68_sgpr69 = COPY $exec
; GFX90A-NEXT: S_BRANCH %bb.8
; GFX90A-NEXT: {{ $}}
- ; GFX90A-NEXT: bb.67.bb161:
- ; GFX90A-NEXT: successors: %bb.65(0x80000000)
+ ; GFX90A-NEXT: bb.68.bb161:
+ ; GFX90A-NEXT: successors: %bb.66(0x80000000)
; GFX90A-NEXT: liveins: $sgpr14, $sgpr15, $sgpr16, $vgpr15, $vgpr17, $vgpr31, $vgpr52, $vgpr53, $sgpr4_sgpr5, $sgpr6_sgpr7, $sgpr8_sgpr9:0x000000000000000F, $sgpr10_sgpr11, $sgpr12_sgpr13, $sgpr18_sgpr19, $sgpr24_sgpr25, $sgpr34_sgpr35, $sgpr38_sgpr39, $sgpr40_sgpr41, $sgpr42_sgpr43, $sgpr44_sgpr45, $sgpr48_sgpr49, $sgpr50_sgpr51, $sgpr52_sgpr53, $sgpr54_sgpr55, $sgpr64_sgpr65, $sgpr66_sgpr67, $vgpr0_vgpr1:0x000000000000000F, $vgpr4_vgpr5:0x000000000000000F, $vgpr6_vgpr7:0x000000000000000F, $vgpr8_vgpr9:0x000000000000000F, $vgpr10_vgpr11:0x000000000000000C, $vgpr12_vgpr13:0x000000000000000C, $vgpr18_vgpr19:0x000000000000000C, $vgpr20_vgpr21:0x000000000000000C, $vgpr22_vgpr23:0x000000000000000C, $vgpr24_vgpr25:0x000000000000000C, $vgpr40_vgpr41:0x000000000000000F, $vgpr42_vgpr43:0x000000000000000F, $vgpr44_vgpr45:0x000000000000000F, $vgpr46_vgpr47:0x000000000000000F, $vgpr56_vgpr57:0x000000000000000F, $vgpr58_vgpr59:0x000000000000000F, $vgpr60_vgpr61:0x000000000000000F, $vgpr62_vgpr63:0x000000000000000F, $sgpr0_sgpr1_sgpr2_sgpr3
; GFX90A-NEXT: {{ $}}
; GFX90A-NEXT: renamable $vgpr2 = V_OR_B32_e32 killed $vgpr21, killed $vgpr23, implicit $exec
@@ -922,10 +932,10 @@ define amdgpu_kernel void @f1(ptr addrspace(1) %arg, ptr addrspace(1) %arg1, i64
; GFX90A-NEXT: renamable $vgpr2 = V_CNDMASK_B32_e64 0, 0, 0, killed $vgpr2, killed $vcc, implicit $exec
; GFX90A-NEXT: renamable $vgpr2 = V_OR_B32_e32 killed $vgpr2, killed $vgpr15, implicit $exec
; GFX90A-NEXT: DS_WRITE2_B32_gfx9 killed renamable $vgpr3, killed renamable $vgpr2, renamable $vgpr3, 0, 1, 0, implicit $exec :: (store (s64) into `ptr addrspace(3) null`, align 4, addrspace 3)
- ; GFX90A-NEXT: S_BRANCH %bb.65
+ ; GFX90A-NEXT: S_BRANCH %bb.66
; GFX90A-NEXT: {{ $}}
- ; GFX90A-NEXT: bb.68.bb174:
- ; GFX90A-NEXT: successors: %bb.72(0x40000000), %bb.69(0x40000000)
+ ; GFX90A-NEXT: bb.69.bb174:
+ ; GFX90A-NEXT: successors: %bb.73(0x40000000), %bb.70(0x40000000)
; GFX90A-NEXT: liveins: $sgpr14, $sgpr15, $sgpr16, $vgpr15, $vgpr17, $vgpr30, $vgpr31, $vgpr52, $vgpr53, $sgpr4_sgpr5, $sgpr6_sgpr7, $sgpr8_sgpr9:0x000000000000000F, $sgpr10_sgpr11, $sgpr12_sgpr13, $sgpr18_sgpr19, $sgpr24_sgpr25, $sgpr28_sgpr29, $sgpr34_sgpr35, $sgpr38_sgpr39, $sgpr40_sgpr41, $sgpr42_sgpr43, $sgpr44_sgpr45, $sgpr48_sgpr49, $sgpr50_sgpr51, $sgpr52_sgpr53, $sgpr54_sgpr55, $sgpr64_sgpr65, $sgpr66_sgpr67, $sgpr20_sgpr21_sgpr22_sgpr23:0x000000000000003C, $sgpr24_sgpr25_sgpr26_sgpr27:0x00000000000000F0, $vgpr0_vgpr1:0x000000000000000F, $vgpr2_vgpr3:0x000000000000000F, $vgpr4_vgpr5:0x000000000000000F, $vgpr6_vgpr7:0x000000000000000F, $vgpr8_vgpr9:0x000000000000000F, $vgpr10_vgpr11:0x000000000000000F, $vgpr12_vgpr13:0x000000000000000F, $vgpr14_vgpr15:0x0000000000000003, $vgpr16_vgpr17:0x0000000000000003, $vgpr18_vgpr19:0x000000000000000F, $vgpr20_vgpr21:0x000000000000000F, $vgpr22_vgpr23:0x000000000000000F, $vgpr24_vgpr25:0x000000000000000F, $vgpr40_vgpr41:0x000000000000000F, $vgpr42_vgpr43:0x000000000000000F, $vgpr44_vgpr45:0x000000000000000F, $vgpr46_vgpr47:0x000000000000000F, $vgpr56_vgpr57:0x000000000000000F, $vgpr58_vgpr59:0x000000000000000F, $vgpr60_vgpr61:0x000000000000000F, $vgpr62_vgpr63:0x000000000000000F, $sgpr0_sgpr1_sgpr2_sgpr3
; GFX90A-NEXT: {{ $}}
; GFX90A-NEXT: renamable $vgpr26 = V_OR_B32_e32 1, $vgpr24, implicit $exec
@@ -938,17 +948,17 @@ define amdgpu_kernel void @f1(ptr addrspace(1) %arg, ptr addrspace(1) %arg1, i64
; GFX90A-NEXT: renamable $vgpr50 = V_CNDMASK_B32_e64 0, 0, 0, $vgpr32, killed $sgpr12_sgpr13, implicit $exec
; GFX90A-NEXT: renamable $sgpr12_sgpr13 = S_MOV_B64 -1
; GFX90A-NEXT: renamable $vcc = S_AND_B64 $exec, killed renamable $sgpr28_sgpr29, implicit-def dead $scc
- ; GFX90A-NEXT: S_CBRANCH_VCCNZ %bb.72, implicit $vcc
+ ; GFX90A-NEXT: S_CBRANCH_VCCNZ %bb.73, implicit $vcc
; GFX90A-NEXT: {{ $}}
- ; GFX90A-NEXT: bb.69.Flow:
- ; GFX90A-NEXT: successors: %bb.70(0x40000000), %bb.71(0x40000000)
+ ; GFX90A-NEXT: bb.70.Flow:
+ ; GFX90A-NEXT: successors: %bb.71(0x40000000), %bb.72(0x40000000)
; GFX90A-NEXT: liveins: $sgpr14, $sgpr15, $sgpr16, $vgpr15, $vgpr17, $vgpr30, $vgpr31, $vgpr52, $vgpr53, $sgpr4_sgpr5, $sgpr6_sgpr7, $sgpr8_sgpr9:0x000000000000000F, $sgpr10_sgpr11, $sgpr12_sgpr13, $sgpr18_sgpr19, $sgpr24_sgpr25, $sgpr34_sgpr35, $sgpr38_sgpr39, $sgpr40_sgpr41, $sgpr42_sgpr43, $sgpr44_sgpr45, $sgpr48_sgpr49, $sgpr50_sgpr51, $sgpr52_sgpr53, $sgpr54_sgpr55, $sgpr64_sgpr65, $sgpr66_sgpr67, $sgpr20_sgpr21_sgpr22_sgpr23:0x000000000000003C, $sgpr24_sgpr25_sgpr26_sgpr27:0x00000000000000F0, $vgpr0_vgpr1:0x000000000000000F, $vgpr2_vgpr3:0x000000000000000F, $vgpr4_vgpr5:0x000000000000000F, $vgpr6_vgpr7:0x000000000000000F, $vgpr8_vgpr9:0x000000000000000F, $vgpr10_vgpr11:0x000000000000000C, $vgpr12_vgpr13:0x000000000000000C, $vgpr18_vgpr19:0x000000000000000C, $vgpr20_vgpr21:0x000000000000000C, $vgpr22_vgpr23:0x000000000000000C, $vgpr24_vgpr25:0x000000000000000C, $vgpr26_vgpr27:0x0000000000000003, $vgpr28_vgpr29:0x0000000000000003, $vgpr32_vgpr33:0x0000000000000003, $vgpr34_vgpr35:0x0000000000000003, $vgpr36_vgpr37:0x0000000000000003, $vgpr38_vgpr39:0x0000000000000003, $vgpr40_vgpr41:0x000000000000000F, $vgpr42_vgpr43:0x000000000000000F, $vgpr44_vgpr45:0x000000000000000F, $vgpr46_vgpr47:0x000000000000000F, $vgpr48_vgpr49:0x0000000000000003, $vgpr50_vgpr51:0x0000000000000003, $vgpr56_vgpr57:0x000000000000000F, $vgpr58_vgpr59:0x000000000000000F, $vgpr60_vgpr61:0x000000000000000F, $vgpr62_vgpr63:0x000000000000000F, $sgpr0_sgpr1_sgpr2_sgpr3
; GFX90A-NEXT: {{ $}}
; GFX90A-NEXT: $vcc = S_ANDN2_B64 $exec, killed renamable $sgpr12_sgpr13, implicit-def dead $scc
- ; GFX90A-NEXT: S_CBRANCH_VCCNZ %bb.71, implicit $vcc
+ ; GFX90A-NEXT: S_CBRANCH_VCCNZ %bb.72, implicit $vcc
; GFX90A-NEXT: {{ $}}
- ; GFX90A-NEXT: bb.70.bb186:
- ; GFX90A-NEXT: successors: %bb.71(0x80000000)
+ ; GFX90A-NEXT: bb.71.bb186:
+ ; GFX90A-NEXT: successors: %bb.72(0x80000000)
; GFX90A-NEXT: liveins: $sgpr14, $sgpr15, $sgpr16, $vgpr15, $vgpr17, $vgpr30, $vgpr31, $vgpr52, $vgpr53, $sgpr4_sgpr5, $sgpr6_sgpr7, $sgpr8_sgpr9:0x000000000000000F, $sgpr10_sgpr11, $sgpr18_sgpr19, $sgpr24_sgpr25, $sgpr34_sgpr35, $sgpr38_sgpr39, $sgpr40_sgpr41, $sgpr42_sgpr43, $sgpr44_sgpr45, $sgpr48_sgpr49, $sgpr50_sgpr51, $sgpr52_sgpr53, $sgpr54_sgpr55, $sgpr64_sgpr65, $sgpr66_sgpr67, $sgpr20_sgpr21_sgpr22_sgpr23:0x000000000000003C, $sgpr24_sgpr25_sgpr26_sgpr27:0x00000000000000F0, $vgpr0_vgpr1:0x000000000000000F, $vgpr2_vgpr3:0x000000000000000F, $vgpr4_vgpr5:0x000000000000000F, $vgpr6_vgpr7:0x000000000000000F, $vgpr8_vgpr9:0x000000000000000F, $vgpr10_vgpr11:0x000000000000000C, $vgpr12_vgpr13:0x000000000000000C, $vgpr18_vgpr19:0x000000000000000C, $vgpr20_vgpr21:0x000000000000000C, $vgpr22_vgpr23:0x000000000000000C, $vgpr24_vgpr25:0x000000000000000C, $vgpr26_vgpr27:0x0000000000000003, $vgpr28_vgpr29:0x0000000000000003, $vgpr32_vgpr33:0x0000000000000003, $vgpr34_vgpr35:0x0000000000000003, $vgpr36_vgpr37:0x0000000000000003, $vgpr38_vgpr39:0x0000000000000003, $vgpr40_vgpr41:0x000000000000000F, $vgpr42_vgpr43:0x000000000000000F, $vgpr44_vgpr45:0x000000000000000F, $vgpr46_vgpr47:0x000000000000000F, $vgpr48_vgpr49:0x0000000000000003, $vgpr50_vgpr51:0x0000000000000003, $vgpr56_vgpr57:0x000000000000000F, $vgpr58_vgpr59:0x000000000000000F, $vgpr60_vgpr61:0x000000000000000F, $vgpr62_vgpr63:0x000000000000000F, $sgpr0_sgpr1_sgpr2_sgpr3
; GFX90A-NEXT: {{ $}}
; GFX90A-NEXT: renamable $vgpr2_vgpr3 = V_LSHLREV_B64_e64 3, killed $vgpr2_vgpr3, implicit $exec
@@ -976,15 +986,15 @@ define amdgpu_kernel void @f1(ptr addrspace(1) %arg, ptr addrspace(1) %arg1, i64
; GFX90A-NEXT: BUFFER_STORE_DWORD_OFFSET killed renamable $vgpr3, $sgpr0_sgpr1_sgpr2_sgpr3, 0, 4, 0, 0, implicit $exec :: (store (s32) into `ptr addrspace(5) null` + 4, basealign 8, addrspace 5)
; GFX90A-NEXT: BUFFER_STORE_DWORD_OFFSET killed renamable $vgpr2, $sgpr0_sgpr1_sgpr2_sgpr3, 0, 0, 0, 0, implicit $exec :: (store (s32) into `ptr addrspace(5) null`, align 8, addrspace 5)
; GFX90A-NEXT: {{ $}}
- ; GFX90A-NEXT: bb.71.Flow9:
- ; GFX90A-NEXT: successors: %bb.63(0x80000000)
+ ; GFX90A-NEXT: bb.72.Flow9:
+ ; GFX90A-NEXT: successors: %bb.64(0x80000000)
; GFX90A-NEXT: liveins: $sgpr14, $sgpr15, $sgpr16, $vgpr15, $vgpr17, $vgpr30, $vgpr31, $vgpr52, $vgpr53, $sgpr4_sgpr5, $sgpr6_sgpr7, $sgpr8_sgpr9:0x000000000000000F, $sgpr10_sgpr11, $sgpr18_sgpr19, $sgpr24_sgpr25, $sgpr34_sgpr35, $sgpr38_sgpr39, $sgpr40_sgpr41, $sgpr42_sgpr43, $sgpr44_sgpr45, $sgpr48_sgpr49, $sgpr50_sgpr51, $sgpr52_sgpr53, $sgpr54_sgpr55, $sgpr64_sgpr65, $sgpr66_sgpr67, $vgpr0_vgpr1:0x000000000000000F, $vgpr4_vgpr5:0x000000000000000F, $vgpr6_vgpr7:0x000000000000000F, $vgpr8_vgpr9:0x000000000000000F, $vgpr10_vgpr11:0x000000000000000C, $vgpr12_vgpr13:0x000000000000000C, $vgpr18_vgpr19:0x000000000000000C, $vgpr20_vgpr21:0x000000000000000C, $vgpr22_vgpr23:0x000000000000000C, $vgpr24_vgpr25:0x000000000000000C, $vgpr40_vgpr41:0x000000000000000F, $vgpr42_vgpr43:0x000000000000000F, $vgpr44_vgpr45:0x000000000000000F, $vgpr46_vgpr47:0x000000000000000F, $vgpr56_vgpr57:0x000000000000000F, $vgpr58_vgpr59:0x000000000000000F, $vgpr60_vgpr61:0x000000000000000F, $vgpr62_vgpr63:0x000000000000000F, $sgpr0_sgpr1_sgpr2_sgpr3
; GFX90A-NEXT: {{ $}}
; GFX90A-NEXT: renamable $sgpr36_sgpr37 = S_MOV_B64 0
- ; GFX90A-NEXT: S_BRANCH %bb.63
+ ; GFX90A-NEXT: S_BRANCH %bb.64
; GFX90A-NEXT: {{ $}}
- ; GFX90A-NEXT: bb.72.bb196:
- ; GFX90A-NEXT: successors: %bb.69(0x80000000)
+ ; GFX90A-NEXT: bb.73.bb196:
+ ; GFX90A-NEXT: successors: %bb.70(0x80000000)
; GFX90A-NEXT: liveins: $sgpr14, $sgpr15, $sgpr16, $vgpr15, $vgpr17, $vgpr30, $vgpr31, $vgpr52, $vgpr53, $sgpr4_sgpr5, $sgpr6_sgpr7, $sgpr8_sgpr9:0x000000000000000F, $sgpr10_sgpr11, $sgpr18_sgpr19, $sgpr24_sgpr25, $sgpr34_sgpr35, $sgpr38_sgpr39, $sgpr40_sgpr41, $sgpr42_sgpr43, $sgpr44_sgpr45, $sgpr48_sgpr49, $sgpr50_sgpr51, $sgpr52_sgpr53, $sgpr54_sgpr55, $sgpr64_sgpr65, $sgpr66_sgpr67, $sgpr20_sgpr21_sgpr22_sgpr23:0x000000000000003C, $sgpr24_sgpr25_sgpr26_sgpr27:0x00000000000000F0, $vgpr0_vgpr1:0x000000000000000F, $vgpr2_vgpr3:0x000000000000000F, $vgpr4_vgpr5:0x000000000000000F, $vgpr6_vgpr7:0x000000000000000F, $vgpr8_vgpr9:0x000000000000000F, $vgpr10_vgpr11:0x000000000000000C, $vgpr12_vgpr13:0x000000000000000C, $vgpr14_vgpr15:0x0000000000000003, $vgpr16_vgpr17:0x0000000000000003, $vgpr18_vgpr19:0x000000000000000C, $vgpr20_vgpr21:0x000000000000000C, $vgpr22_vgpr23:0x000000000000000C, $vgpr24_vgpr25:0x000000000000000C, $vgpr26_vgpr27:0x0000000000000003, $vgpr28_vgpr29:0x0000000000000003, $vgpr32_vgpr33:0x0000000000000003, $vgpr34_vgpr35:0x0000000000000003, $vgpr36_vgpr37:0x0000000000000003, $vgpr38_vgpr39:0x0000000000000003, $vgpr40_vgpr41:0x000000000000000F, $vgpr42_vgpr43:0x000000000000000F, $vgpr44_vgpr45:0x000000000000000F, $vgpr46_vgpr47:0x000000000000000F, $vgpr48_vgpr49:0x0000000000000003, $vgpr50_vgpr51:0x0000000000000003, $vgpr56_vgpr57:0x000000000000000F, $vgpr58_vgpr59:0x000000000000000F, $vgpr60_vgpr61:0x000000000000000F, $vgpr62_vgpr63:0x000000000000000F, $sgpr0_sgpr1_sgpr2_sgpr3
; GFX90A-NEXT: {{ $}}
; GFX90A-NEXT: renamable $vgpr10 = V_OR_B32_e32 $vgpr50, killed $vgpr16, implicit $exec
@@ -992,7 +1002,7 @@ define amdgpu_kernel void @f1(ptr addrspace(1) %arg, ptr addrspace(1) %arg1, i64
; GFX90A-NEXT: renamable $vgpr55 = V_MOV_B32_e32 0, implicit $exec
; GFX90A-NEXT: DS_WRITE_B64_gfx9 killed renamable $vgpr55, renamable $vgpr54_vgpr55, 0, 0, implicit $exec :: (store (s64) into `ptr addrspace(3) null`, addrspace 3)
; GFX90A-NEXT: renamable $sgpr12_sgpr13 = S_MOV_B64 0
- ; GFX90A-NEXT: S_BRANCH %bb.69
+ ; GFX90A-NEXT: S_BRANCH %bb.70
bb:
%i = tail call i32 @llvm.amdgcn.workitem.id.x()
%i11 = icmp eq i32 %i, 0
diff --git a/llvm/test/CodeGen/AMDGPU/buffer-fat-pointer-atomicrmw-fadd.ll b/llvm/test/CodeGen/AMDGPU/buffer-fat-pointer-atomicrmw-fadd.ll
index 3eba106b861c6..ed8af156e39f2 100644
--- a/llvm/test/CodeGen/AMDGPU/buffer-fat-pointer-atomicrmw-fadd.ll
+++ b/llvm/test/CodeGen/AMDGPU/buffer-fat-pointer-atomicrmw-fadd.ll
@@ -248,20 +248,20 @@ define void @buffer_fat_ptr_agent_atomic_fadd_noret_f32__offset__amdgpu_no_fine_
; GFX10-NEXT: s_add_i32 s4, s20, 0x400
; GFX10-NEXT: v_mov_b32_e32 v3, s4
; GFX10-NEXT: s_mov_b32 s4, 0
-; GFX10-NEXT: buffer_load_dword v2, v1, s[16:19], 0 offen offset:1024
+; GFX10-NEXT: buffer_load_dword v1, v1, s[16:19], 0 offen offset:1024
; GFX10-NEXT: .LBB1_1: ; %atomicrmw.start
; GFX10-NEXT: ; =>This Inner Loop Header: Depth=1
; GFX10-NEXT: s_waitcnt vmcnt(0)
-; GFX10-NEXT: v_add_f32_e32 v1, v2, v0
-; GFX10-NEXT: v_mov_b32_e32 v5, v2
+; GFX10-NEXT: v_mov_b32_e32 v5, v1
; GFX10-NEXT: s_waitcnt_vscnt null, 0x0
-; GFX10-NEXT: v_mov_b32_e32 v4, v1
-; GFX10-NEXT: buffer_atomic_cmpswap v[4:5], v3, s[16:19], 0 offen glc
+; GFX10-NEXT: v_add_f32_e32 v4, v5, v0
+; GFX10-NEXT: v_mov_b32_e32 v1, v4
+; GFX10-NEXT: v_mov_b32_e32 v2, v5
+; GFX10-NEXT: buffer_atomic_cmpswap v[1:2], v3, s[16:19], 0 offen glc
; GFX10-NEXT: s_waitcnt vmcnt(0)
; GFX10-NEXT: buffer_gl1_inv
; GFX10-NEXT: buffer_gl0_inv
-; GFX10-NEXT: v_cmp_eq_u32_e32 vcc_lo, v4, v2
-; GFX10-NEXT: v_mov_b32_e32 v2, v4
+; GFX10-NEXT: v_cmp_eq_u32_e32 vcc_lo, v1, v5
; GFX10-NEXT: s_or_b32 s4, vcc_lo, s4
; GFX10-NEXT: s_andn2_b32 exec_lo, exec_lo, s4
; GFX10-NEXT: s_cbranch_execnz .LBB1_1
@@ -291,22 +291,22 @@ define void @buffer_fat_ptr_agent_atomic_fadd_noret_f32__offset__amdgpu_no_fine_
; GFX8: ; %bb.0:
; GFX8-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX8-NEXT: v_mov_b32_e32 v1, s20
-; GFX8-NEXT: buffer_load_dword v2, v1, s[16:19], 0 offen offset:1024
+; GFX8-NEXT: buffer_load_dword v1, v1, s[16:19], 0 offen offset:1024
; GFX8-NEXT: s_add_i32 s6, s20, 0x400
; GFX8-NEXT: s_mov_b64 s[4:5], 0
; GFX8-NEXT: v_mov_b32_e32 v3, s6
; GFX8-NEXT: .LBB1_1: ; %atomicrmw.start
; GFX8-NEXT: ; =>This Inner Loop Header: Depth=1
; GFX8-NEXT: s_waitcnt vmcnt(0)
-; GFX8-NEXT: v_add_f32_e32 v1, v2, v0
-; GFX8-NEXT: v_mov_b32_e32 v5, v2
-; GFX8-NEXT: v_mov_b32_e32 v4, v1
-; GFX8-NEXT: buffer_atomic_cmpswap v[4:5], v3, s[16:19], 0 offen glc
+; GFX8-NEXT: v_mov_b32_e32 v5, v1
+; GFX8-NEXT: v_add_f32_e32 v4, v5, v0
+; GFX8-NEXT: v_mov_b32_e32 v1, v4
+; GFX8-NEXT: v_mov_b32_e32 v2, v5
+; GFX8-NEXT: buffer_atomic_cmpswap v[1:2], v3, s[16:19], 0 offen glc
; GFX8-NEXT: s_waitcnt vmcnt(0)
; GFX8-NEXT: buffer_wbinvl1
-; GFX8-NEXT: v_cmp_eq_u32_e32 vcc, v4, v2
+; GFX8-NEXT: v_cmp_eq_u32_e32 vcc, v1, v5
; GFX8-NEXT: s_or_b64 s[4:5], vcc, s[4:5]
-; GFX8-NEXT: v_mov_b32_e32 v2, v4
; GFX8-NEXT: s_andn2_b64 exec, exec, s[4:5]
; GFX8-NEXT: s_cbranch_execnz .LBB1_1
; GFX8-NEXT: ; %bb.2: ; %atomicrmw.end
@@ -317,22 +317,22 @@ define void @buffer_fat_ptr_agent_atomic_fadd_noret_f32__offset__amdgpu_no_fine_
; GFX7: ; %bb.0:
; GFX7-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX7-NEXT: v_mov_b32_e32 v1, s20
-; GFX7-NEXT: buffer_load_dword v2, v1, s[16:19], 0 offen offset:1024
+; GFX7-NEXT: buffer_load_dword v1, v1, s[16:19], 0 offen offset:1024
; GFX7-NEXT: s_add_i32 s6, s20, 0x400
; GFX7-NEXT: s_mov_b64 s[4:5], 0
; GFX7-NEXT: v_mov_b32_e32 v3, s6
; GFX7-NEXT: .LBB1_1: ; %atomicrmw.start
; GFX7-NEXT: ; =>This Inner Loop Header: Depth=1
; GFX7-NEXT: s_waitcnt vmcnt(0)
-; GFX7-NEXT: v_add_f32_e32 v1, v2, v0
-; GFX7-NEXT: v_mov_b32_e32 v5, v2
-; GFX7-NEXT: v_mov_b32_e32 v4, v1
-; GFX7-NEXT: buffer_atomic_cmpswap v[4:5], v3, s[16:19], 0 offen glc
+; GFX7-NEXT: v_mov_b32_e32 v5, v1
+; GFX7-NEXT: v_add_f32_e32 v4, v5, v0
+; GFX7-NEXT: v_mov_b32_e32 v1, v4
+; GFX7-NEXT: v_mov_b32_e32 v2, v5
+; GFX7-NEXT: buffer_atomic_cmpswap v[1:2], v3, s[16:19], 0 offen glc
; GFX7-NEXT: s_waitcnt vmcnt(0)
; GFX7-NEXT: buffer_wbinvl1
-; GFX7-NEXT: v_cmp_eq_u32_e32 vcc, v4, v2
+; GFX7-NEXT: v_cmp_eq_u32_e32 vcc, v1, v5
; GFX7-NEXT: s_or_b64 s[4:5], vcc, s[4:5]
-; GFX7-NEXT: v_mov_b32_e32 v2, v4
; GFX7-NEXT: s_andn2_b64 exec, exec, s[4:5]
; GFX7-NEXT: s_cbranch_execnz .LBB1_1
; GFX7-NEXT: ; %bb.2: ; %atomicrmw.end
@@ -343,23 +343,23 @@ define void @buffer_fat_ptr_agent_atomic_fadd_noret_f32__offset__amdgpu_no_fine_
; GFX6: ; %bb.0:
; GFX6-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX6-NEXT: v_mov_b32_e32 v1, s20
-; GFX6-NEXT: buffer_load_dword v2, v1, s[16:19], 0 offen offset:1024
+; GFX6-NEXT: buffer_load_dword v1, v1, s[16:19], 0 offen offset:1024
; GFX6-NEXT: s_add_i32 s6, s20, 0x400
; GFX6-NEXT: s_mov_b64 s[4:5], 0
; GFX6-NEXT: v_mov_b32_e32 v3, s6
; GFX6-NEXT: .LBB1_1: ; %atomicrmw.start
; GFX6-NEXT: ; =>This Inner Loop Header: Depth=1
; GFX6-NEXT: s_waitcnt vmcnt(0)
-; GFX6-NEXT: v_add_f32_e32 v1, v2, v0
+; GFX6-NEXT: v_mov_b32_e32 v5, v1
+; GFX6-NEXT: v_add_f32_e32 v4, v5, v0
; GFX6-NEXT: s_waitcnt expcnt(0)
-; GFX6-NEXT: v_mov_b32_e32 v5, v2
-; GFX6-NEXT: v_mov_b32_e32 v4, v1
-; GFX6-NEXT: buffer_atomic_cmpswap v[4:5], v3, s[16:19], 0 offen glc
+; GFX6-NEXT: v_mov_b32_e32 v1, v4
+; GFX6-NEXT: v_mov_b32_e32 v2, v5
+; GFX6-NEXT: buffer_atomic_cmpswap v[1:2], v3, s[16:19], 0 offen glc
; GFX6-NEXT: s_waitcnt vmcnt(0)
; GFX6-NEXT: buffer_wbinvl1
-; GFX6-NEXT: v_cmp_eq_u32_e32 vcc, v4, v2
+; GFX6-NEXT: v_cmp_eq_u32_e32 vcc, v1, v5
; GFX6-NEXT: s_or_b64 s[4:5], vcc, s[4:5]
-; GFX6-NEXT: v_mov_b32_e32 v2, v4
; GFX6-NEXT: s_andn2_b64 exec, exec, s[4:5]
; GFX6-NEXT: s_cbranch_execnz .LBB1_1
; GFX6-NEXT: ; %bb.2: ; %atomicrmw.end
@@ -1037,20 +1037,20 @@ define void @buffer_fat_ptr_agent_atomic_fadd_noret_f32__offset__amdgpu_no_fine_
; GFX10-NEXT: s_add_i32 s4, s20, 0x400
; GFX10-NEXT: v_mov_b32_e32 v3, s4
; GFX10-NEXT: s_mov_b32 s4, 0
-; GFX10-NEXT: buffer_load_dword v2, v1, s[16:19], 0 offen offset:1024
+; GFX10-NEXT: buffer_load_dword v1, v1, s[16:19], 0 offen offset:1024
; GFX10-NEXT: .LBB4_1: ; %atomicrmw.start
; GFX10-NEXT: ; =>This Inner Loop Header: Depth=1
; GFX10-NEXT: s_waitcnt vmcnt(0)
-; GFX10-NEXT: v_add_f32_e32 v1, v2, v0
-; GFX10-NEXT: v_mov_b32_e32 v5, v2
+; GFX10-NEXT: v_mov_b32_e32 v5, v1
; GFX10-NEXT: s_waitcnt_vscnt null, 0x0
-; GFX10-NEXT: v_mov_b32_e32 v4, v1
-; GFX10-NEXT: buffer_atomic_cmpswap v[4:5], v3, s[16:19], 0 offen glc
+; GFX10-NEXT: v_add_f32_e32 v4, v5, v0
+; GFX10-NEXT: v_mov_b32_e32 v1, v4
+; GFX10-NEXT: v_mov_b32_e32 v2, v5
+; GFX10-NEXT: buffer_atomic_cmpswap v[1:2], v3, s[16:19], 0 offen glc
; GFX10-NEXT: s_waitcnt vmcnt(0)
; GFX10-NEXT: buffer_gl1_inv
; GFX10-NEXT: buffer_gl0_inv
-; GFX10-NEXT: v_cmp_eq_u32_e32 vcc_lo, v4, v2
-; GFX10-NEXT: v_mov_b32_e32 v2, v4
+; GFX10-NEXT: v_cmp_eq_u32_e32 vcc_lo, v1, v5
; GFX10-NEXT: s_or_b32 s4, vcc_lo, s4
; GFX10-NEXT: s_andn2_b32 exec_lo, exec_lo, s4
; GFX10-NEXT: s_cbranch_execnz .LBB4_1
@@ -1062,21 +1062,21 @@ define void @buffer_fat_ptr_agent_atomic_fadd_noret_f32__offset__amdgpu_no_fine_
; GFX90A: ; %bb.0:
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX90A-NEXT: v_mov_b32_e32 v1, s20
-; GFX90A-NEXT: buffer_load_dword v3, v1, s[16:19], 0 offen offset:1024
+; GFX90A-NEXT: buffer_load_dword v2, v1, s[16:19], 0 offen offset:1024
; GFX90A-NEXT: s_add_i32 s6, s20, 0x400
; GFX90A-NEXT: s_mov_b64 s[4:5], 0
; GFX90A-NEXT: v_mov_b32_e32 v1, s6
; GFX90A-NEXT: .LBB4_1: ; %atomicrmw.start
; GFX90A-NEXT: ; =>This Inner Loop Header: Depth=1
; GFX90A-NEXT: s_waitcnt vmcnt(0)
-; GFX90A-NEXT: v_add_f32_e32 v2, v3, v0
-; GFX90A-NEXT: v_pk_mov_b32 v[4:5], v[2:3], v[2:3] op_sel:[0,1]
-; GFX90A-NEXT: buffer_atomic_cmpswap v[4:5], v1, s[16:19], 0 offen glc
+; GFX90A-NEXT: v_mov_b32_e32 v5, v2
+; GFX90A-NEXT: v_add_f32_e32 v4, v5, v0
+; GFX90A-NEXT: v_pk_mov_b32 v[2:3], v[4:5], v[4:5] op_sel:[0,1]
+; GFX90A-NEXT: buffer_atomic_cmpswap v[2:3], v1, s[16:19], 0 offen glc
; GFX90A-NEXT: s_waitcnt vmcnt(0)
; GFX90A-NEXT: buffer_wbinvl1
-; GFX90A-NEXT: v_cmp_eq_u32_e32 vcc, v4, v3
+; GFX90A-NEXT: v_cmp_eq_u32_e32 vcc, v2, v5
; GFX90A-NEXT: s_or_b64 s[4:5], vcc, s[4:5]
-; GFX90A-NEXT: v_mov_b32_e32 v3, v4
; GFX90A-NEXT: s_andn2_b64 exec, exec, s[4:5]
; GFX90A-NEXT: s_cbranch_execnz .LBB4_1
; GFX90A-NEXT: ; %bb.2: ; %atomicrmw.end
@@ -1087,22 +1087,22 @@ define void @buffer_fat_ptr_agent_atomic_fadd_noret_f32__offset__amdgpu_no_fine_
; GFX908: ; %bb.0:
; GFX908-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX908-NEXT: v_mov_b32_e32 v1, s20
-; GFX908-NEXT: buffer_load_dword v2, v1, s[16:19], 0 offen offset:1024
+; GFX908-NEXT: buffer_load_dword v1, v1, s[16:19], 0 offen offset:1024
; GFX908-NEXT: s_add_i32 s6, s20, 0x400
; GFX908-NEXT: s_mov_b64 s[4:5], 0
; GFX908-NEXT: v_mov_b32_e32 v3, s6
; GFX908-NEXT: .LBB4_1: ; %atomicrmw.start
; GFX908-NEXT: ; =>This Inner Loop Header: Depth=1
; GFX908-NEXT: s_waitcnt vmcnt(0)
-; GFX908-NEXT: v_add_f32_e32 v1, v2, v0
-; GFX908-NEXT: v_mov_b32_e32 v5, v2
-; GFX908-NEXT: v_mov_b32_e32 v4, v1
-; GFX908-NEXT: buffer_atomic_cmpswap v[4:5], v3, s[16:19], 0 offen glc
+; GFX908-NEXT: v_mov_b32_e32 v5, v1
+; GFX908-NEXT: v_add_f32_e32 v4, v5, v0
+; GFX908-NEXT: v_mov_b32_e32 v1, v4
+; GFX908-NEXT: v_mov_b32_e32 v2, v5
+; GFX908-NEXT: buffer_atomic_cmpswap v[1:2], v3, s[16:19], 0 offen glc
; GFX908-NEXT: s_waitcnt vmcnt(0)
; GFX908-NEXT: buffer_wbinvl1
-; GFX908-NEXT: v_cmp_eq_u32_e32 vcc, v4, v2
+; GFX908-NEXT: v_cmp_eq_u32_e32 vcc, v1, v5
; GFX908-NEXT: s_or_b64 s[4:5], vcc, s[4:5]
-; GFX908-NEXT: v_mov_b32_e32 v2, v4
; GFX908-NEXT: s_andn2_b64 exec, exec, s[4:5]
; GFX908-NEXT: s_cbranch_execnz .LBB4_1
; GFX908-NEXT: ; %bb.2: ; %atomicrmw.end
@@ -1113,22 +1113,22 @@ define void @buffer_fat_ptr_agent_atomic_fadd_noret_f32__offset__amdgpu_no_fine_
; GFX8: ; %bb.0:
; GFX8-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX8-NEXT: v_mov_b32_e32 v1, s20
-; GFX8-NEXT: buffer_load_dword v2, v1, s[16:19], 0 offen offset:1024
+; GFX8-NEXT: buffer_load_dword v1, v1, s[16:19], 0 offen offset:1024
; GFX8-NEXT: s_add_i32 s6, s20, 0x400
; GFX8-NEXT: s_mov_b64 s[4:5], 0
; GFX8-NEXT: v_mov_b32_e32 v3, s6
; GFX8-NEXT: .LBB4_1: ; %atomicrmw.start
; GFX8-NEXT: ; =>This Inner Loop Header: Depth=1
; GFX8-NEXT: s_waitcnt vmcnt(0)
-; GFX8-NEXT: v_add_f32_e32 v1, v2, v0
-; GFX8-NEXT: v_mov_b32_e32 v5, v2
-; GFX8-NEXT: v_mov_b32_e32 v4, v1
-; GFX8-NEXT: buffer_atomic_cmpswap v[4:5], v3, s[16:19], 0 offen glc
+; GFX8-NEXT: v_mov_b32_e32 v5, v1
+; GFX8-NEXT: v_add_f32_e32 v4, v5, v0
+; GFX8-NEXT: v_mov_b32_e32 v1, v4
+; GFX8-NEXT: v_mov_b32_e32 v2, v5
+; GFX8-NEXT: buffer_atomic_cmpswap v[1:2], v3, s[16:19], 0 offen glc
; GFX8-NEXT: s_waitcnt vmcnt(0)
; GFX8-NEXT: buffer_wbinvl1
-; GFX8-NEXT: v_cmp_eq_u32_e32 vcc, v4, v2
+; GFX8-NEXT: v_cmp_eq_u32_e32 vcc, v1, v5
; GFX8-NEXT: s_or_b64 s[4:5], vcc, s[4:5]
-; GFX8-NEXT: v_mov_b32_e32 v2, v4
; GFX8-NEXT: s_andn2_b64 exec, exec, s[4:5]
; GFX8-NEXT: s_cbranch_execnz .LBB4_1
; GFX8-NEXT: ; %bb.2: ; %atomicrmw.end
@@ -1139,22 +1139,22 @@ define void @buffer_fat_ptr_agent_atomic_fadd_noret_f32__offset__amdgpu_no_fine_
; GFX7: ; %bb.0:
; GFX7-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX7-NEXT: v_mov_b32_e32 v1, s20
-; GFX7-NEXT: buffer_load_dword v2, v1, s[16:19], 0 offen offset:1024
+; GFX7-NEXT: buffer_load_dword v1, v1, s[16:19], 0 offen offset:1024
; GFX7-NEXT: s_add_i32 s6, s20, 0x400
; GFX7-NEXT: s_mov_b64 s[4:5], 0
; GFX7-NEXT: v_mov_b32_e32 v3, s6
; GFX7-NEXT: .LBB4_1: ; %atomicrmw.start
; GFX7-NEXT: ; =>This Inner Loop Header: Depth=1
; GFX7-NEXT: s_waitcnt vmcnt(0)
-; GFX7-NEXT: v_add_f32_e32 v1, v2, v0
-; GFX7-NEXT: v_mov_b32_e32 v5, v2
-; GFX7-NEXT: v_mov_b32_e32 v4, v1
-; GFX7-NEXT: buffer_atomic_cmpswap v[4:5], v3, s[16:19], 0 offen glc
+; GFX7-NEXT: v_mov_b32_e32 v5, v1
+; GFX7-NEXT: v_add_f32_e32 v4, v5, v0
+; GFX7-NEXT: v_mov_b32_e32 v1, v4
+; GFX7-NEXT: v_mov_b32_e32 v2, v5
+; GFX7-NEXT: buffer_atomic_cmpswap v[1:2], v3, s[16:19], 0 offen glc
; GFX7-NEXT: s_waitcnt vmcnt(0)
; GFX7-NEXT: buffer_wbinvl1
-; GFX7-NEXT: v_cmp_eq_u32_e32 vcc, v4, v2
+; GFX7-NEXT: v_cmp_eq_u32_e32 vcc, v1, v5
; GFX7-NEXT: s_or_b64 s[4:5], vcc, s[4:5]
-; GFX7-NEXT: v_mov_b32_e32 v2, v4
; GFX7-NEXT: s_andn2_b64 exec, exec, s[4:5]
; GFX7-NEXT: s_cbranch_execnz .LBB4_1
; GFX7-NEXT: ; %bb.2: ; %atomicrmw.end
@@ -1165,23 +1165,23 @@ define void @buffer_fat_ptr_agent_atomic_fadd_noret_f32__offset__amdgpu_no_fine_
; GFX6: ; %bb.0:
; GFX6-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX6-NEXT: v_mov_b32_e32 v1, s20
-; GFX6-NEXT: buffer_load_dword v2, v1, s[16:19], 0 offen offset:1024
+; GFX6-NEXT: buffer_load_dword v1, v1, s[16:19], 0 offen offset:1024
; GFX6-NEXT: s_add_i32 s6, s20, 0x400
; GFX6-NEXT: s_mov_b64 s[4:5], 0
; GFX6-NEXT: v_mov_b32_e32 v3, s6
; GFX6-NEXT: .LBB4_1: ; %atomicrmw.start
; GFX6-NEXT: ; =>This Inner Loop Header: Depth=1
; GFX6-NEXT: s_waitcnt vmcnt(0)
-; GFX6-NEXT: v_add_f32_e32 v1, v2, v0
+; GFX6-NEXT: v_mov_b32_e32 v5, v1
+; GFX6-NEXT: v_add_f32_e32 v4, v5, v0
; GFX6-NEXT: s_waitcnt expcnt(0)
-; GFX6-NEXT: v_mov_b32_e32 v5, v2
-; GFX6-NEXT: v_mov_b32_e32 v4, v1
-; GFX6-NEXT: buffer_atomic_cmpswap v[4:5], v3, s[16:19], 0 offen glc
+; GFX6-NEXT: v_mov_b32_e32 v1, v4
+; GFX6-NEXT: v_mov_b32_e32 v2, v5
+; GFX6-NEXT: buffer_atomic_cmpswap v[1:2], v3, s[16:19], 0 offen glc
; GFX6-NEXT: s_waitcnt vmcnt(0)
; GFX6-NEXT: buffer_wbinvl1
-; GFX6-NEXT: v_cmp_eq_u32_e32 vcc, v4, v2
+; GFX6-NEXT: v_cmp_eq_u32_e32 vcc, v1, v5
; GFX6-NEXT: s_or_b64 s[4:5], vcc, s[4:5]
-; GFX6-NEXT: v_mov_b32_e32 v2, v4
; GFX6-NEXT: s_andn2_b64 exec, exec, s[4:5]
; GFX6-NEXT: s_cbranch_execnz .LBB4_1
; GFX6-NEXT: ; %bb.2: ; %atomicrmw.end
@@ -2136,20 +2136,20 @@ define void @buffer_fat_ptr_agent_atomic_fadd_noret_f64__offset__amdgpu_no_fine_
; GFX12-NEXT: s_wait_alu 0xfffe
; GFX12-NEXT: v_mov_b32_e32 v6, s4
; GFX12-NEXT: s_mov_b32 s4, 0
-; GFX12-NEXT: buffer_load_b64 v[4:5], v2, s[0:3], null offen offset:2048
+; GFX12-NEXT: buffer_load_b64 v[2:3], v2, s[0:3], null offen offset:2048
; GFX12-NEXT: .LBB9_1: ; %atomicrmw.start
; GFX12-NEXT: ; =>This Inner Loop Header: Depth=1
; GFX12-NEXT: s_wait_loadcnt 0x0
-; GFX12-NEXT: v_add_f64_e32 v[2:3], v[4:5], v[0:1]
-; GFX12-NEXT: v_dual_mov_b32 v10, v5 :: v_dual_mov_b32 v9, v4
+; GFX12-NEXT: v_dual_mov_b32 v10, v3 :: v_dual_mov_b32 v9, v2
; GFX12-NEXT: s_wait_storecnt 0x0
-; GFX12-NEXT: s_delay_alu instid0(VALU_DEP_2)
-; GFX12-NEXT: v_dual_mov_b32 v8, v3 :: v_dual_mov_b32 v7, v2
-; GFX12-NEXT: buffer_atomic_cmpswap_b64 v[7:10], v6, s[0:3], null offen th:TH_ATOMIC_RETURN
+; GFX12-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
+; GFX12-NEXT: v_add_f64_e32 v[7:8], v[9:10], v[0:1]
+; GFX12-NEXT: v_dual_mov_b32 v2, v7 :: v_dual_mov_b32 v3, v8
+; GFX12-NEXT: v_dual_mov_b32 v4, v9 :: v_dual_mov_b32 v5, v10
+; GFX12-NEXT: buffer_atomic_cmpswap_b64 v[2:5], v6, s[0:3], null offen th:TH_ATOMIC_RETURN
; GFX12-NEXT: s_wait_loadcnt 0x0
; GFX12-NEXT: global_inv scope:SCOPE_DEV
-; GFX12-NEXT: v_cmp_eq_u64_e32 vcc_lo, v[7:8], v[4:5]
-; GFX12-NEXT: v_dual_mov_b32 v4, v7 :: v_dual_mov_b32 v5, v8
+; GFX12-NEXT: v_cmp_eq_u64_e32 vcc_lo, v[2:3], v[9:10]
; GFX12-NEXT: s_or_b32 s4, vcc_lo, s4
; GFX12-NEXT: s_wait_alu 0xfffe
; GFX12-NEXT: s_and_not1_b32 exec_lo, exec_lo, s4
@@ -2177,21 +2177,21 @@ define void @buffer_fat_ptr_agent_atomic_fadd_noret_f64__offset__amdgpu_no_fine_
; GFX11-NEXT: s_delay_alu instid0(SALU_CYCLE_1)
; GFX11-NEXT: v_mov_b32_e32 v6, s4
; GFX11-NEXT: s_mov_b32 s4, 0
-; GFX11-NEXT: buffer_load_b64 v[4:5], v2, s[0:3], 0 offen offset:2048
+; GFX11-NEXT: buffer_load_b64 v[2:3], v2, s[0:3], 0 offen offset:2048
; GFX11-NEXT: .LBB9_1: ; %atomicrmw.start
; GFX11-NEXT: ; =>This Inner Loop Header: Depth=1
; GFX11-NEXT: s_waitcnt vmcnt(0)
-; GFX11-NEXT: v_add_f64 v[2:3], v[4:5], v[0:1]
-; GFX11-NEXT: v_dual_mov_b32 v10, v5 :: v_dual_mov_b32 v9, v4
+; GFX11-NEXT: v_dual_mov_b32 v10, v3 :: v_dual_mov_b32 v9, v2
; GFX11-NEXT: s_waitcnt_vscnt null, 0x0
-; GFX11-NEXT: s_delay_alu instid0(VALU_DEP_2)
-; GFX11-NEXT: v_dual_mov_b32 v8, v3 :: v_dual_mov_b32 v7, v2
-; GFX11-NEXT: buffer_atomic_cmpswap_b64 v[7:10], v6, s[0:3], 0 offen glc
+; GFX11-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
+; GFX11-NEXT: v_add_f64 v[7:8], v[9:10], v[0:1]
+; GFX11-NEXT: v_dual_mov_b32 v2, v7 :: v_dual_mov_b32 v3, v8
+; GFX11-NEXT: v_dual_mov_b32 v4, v9 :: v_dual_mov_b32 v5, v10
+; GFX11-NEXT: buffer_atomic_cmpswap_b64 v[2:5], v6, s[0:3], 0 offen glc
; GFX11-NEXT: s_waitcnt vmcnt(0)
; GFX11-NEXT: buffer_gl1_inv
; GFX11-NEXT: buffer_gl0_inv
-; GFX11-NEXT: v_cmp_eq_u64_e32 vcc_lo, v[7:8], v[4:5]
-; GFX11-NEXT: v_dual_mov_b32 v4, v7 :: v_dual_mov_b32 v5, v8
+; GFX11-NEXT: v_cmp_eq_u64_e32 vcc_lo, v[2:3], v[9:10]
; GFX11-NEXT: s_or_b32 s4, vcc_lo, s4
; GFX11-NEXT: s_delay_alu instid0(SALU_CYCLE_1)
; GFX11-NEXT: s_and_not1_b32 exec_lo, exec_lo, s4
@@ -2207,23 +2207,23 @@ define void @buffer_fat_ptr_agent_atomic_fadd_noret_f64__offset__amdgpu_no_fine_
; GFX10-NEXT: s_add_i32 s4, s20, 0x800
; GFX10-NEXT: v_mov_b32_e32 v6, s4
; GFX10-NEXT: s_mov_b32 s4, 0
-; GFX10-NEXT: buffer_load_dwordx2 v[4:5], v2, s[16:19], 0 offen offset:2048
+; GFX10-NEXT: buffer_load_dwordx2 v[2:3], v2, s[16:19], 0 offen offset:2048
; GFX10-NEXT: .LBB9_1: ; %atomicrmw.start
; GFX10-NEXT: ; =>This Inner Loop Header: Depth=1
; GFX10-NEXT: s_waitcnt vmcnt(0)
-; GFX10-NEXT: v_add_f64 v[2:3], v[4:5], v[0:1]
-; GFX10-NEXT: v_mov_b32_e32 v10, v5
-; GFX10-NEXT: v_mov_b32_e32 v9, v4
+; GFX10-NEXT: v_mov_b32_e32 v10, v3
+; GFX10-NEXT: v_mov_b32_e32 v9, v2
; GFX10-NEXT: s_waitcnt_vscnt null, 0x0
-; GFX10-NEXT: v_mov_b32_e32 v8, v3
-; GFX10-NEXT: v_mov_b32_e32 v7, v2
-; GFX10-NEXT: buffer_atomic_cmpswap_x2 v[7:10], v6, s[16:19], 0 offen glc
+; GFX10-NEXT: v_add_f64 v[7:8], v[9:10], v[0:1]
+; GFX10-NEXT: v_mov_b32_e32 v2, v7
+; GFX10-NEXT: v_mov_b32_e32 v3, v8
+; GFX10-NEXT: v_mov_b32_e32 v4, v9
+; GFX10-NEXT: v_mov_b32_e32 v5, v10
+; GFX10-NEXT: buffer_atomic_cmpswap_x2 v[2:5], v6, s[16:19], 0 offen glc
; GFX10-NEXT: s_waitcnt vmcnt(0)
; GFX10-NEXT: buffer_gl1_inv
; GFX10-NEXT: buffer_gl0_inv
-; GFX10-NEXT: v_cmp_eq_u64_e32 vcc_lo, v[7:8], v[4:5]
-; GFX10-NEXT: v_mov_b32_e32 v4, v7
-; GFX10-NEXT: v_mov_b32_e32 v5, v8
+; GFX10-NEXT: v_cmp_eq_u64_e32 vcc_lo, v[2:3], v[9:10]
; GFX10-NEXT: s_or_b32 s4, vcc_lo, s4
; GFX10-NEXT: s_andn2_b32 exec_lo, exec_lo, s4
; GFX10-NEXT: s_cbranch_execnz .LBB9_1
@@ -2244,25 +2244,25 @@ define void @buffer_fat_ptr_agent_atomic_fadd_noret_f64__offset__amdgpu_no_fine_
; GFX908: ; %bb.0:
; GFX908-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX908-NEXT: v_mov_b32_e32 v2, s20
-; GFX908-NEXT: buffer_load_dwordx2 v[4:5], v2, s[16:19], 0 offen offset:2048
+; GFX908-NEXT: buffer_load_dwordx2 v[2:3], v2, s[16:19], 0 offen offset:2048
; GFX908-NEXT: s_add_i32 s6, s20, 0x800
; GFX908-NEXT: s_mov_b64 s[4:5], 0
; GFX908-NEXT: v_mov_b32_e32 v6, s6
; GFX908-NEXT: .LBB9_1: ; %atomicrmw.start
; GFX908-NEXT: ; =>This Inner Loop Header: Depth=1
; GFX908-NEXT: s_waitcnt vmcnt(0)
-; GFX908-NEXT: v_add_f64 v[2:3], v[4:5], v[0:1]
-; GFX908-NEXT: v_mov_b32_e32 v10, v5
-; GFX908-NEXT: v_mov_b32_e32 v9, v4
-; GFX908-NEXT: v_mov_b32_e32 v8, v3
-; GFX908-NEXT: v_mov_b32_e32 v7, v2
-; GFX908-NEXT: buffer_atomic_cmpswap_x2 v[7:10], v6, s[16:19], 0 offen glc
+; GFX908-NEXT: v_mov_b32_e32 v10, v3
+; GFX908-NEXT: v_mov_b32_e32 v9, v2
+; GFX908-NEXT: v_add_f64 v[7:8], v[9:10], v[0:1]
+; GFX908-NEXT: v_mov_b32_e32 v2, v7
+; GFX908-NEXT: v_mov_b32_e32 v3, v8
+; GFX908-NEXT: v_mov_b32_e32 v4, v9
+; GFX908-NEXT: v_mov_b32_e32 v5, v10
+; GFX908-NEXT: buffer_atomic_cmpswap_x2 v[2:5], v6, s[16:19], 0 offen glc
; GFX908-NEXT: s_waitcnt vmcnt(0)
; GFX908-NEXT: buffer_wbinvl1
-; GFX908-NEXT: v_cmp_eq_u64_e32 vcc, v[7:8], v[4:5]
-; GFX908-NEXT: v_mov_b32_e32 v4, v7
+; GFX908-NEXT: v_cmp_eq_u64_e32 vcc, v[2:3], v[9:10]
; GFX908-NEXT: s_or_b64 s[4:5], vcc, s[4:5]
-; GFX908-NEXT: v_mov_b32_e32 v5, v8
; GFX908-NEXT: s_andn2_b64 exec, exec, s[4:5]
; GFX908-NEXT: s_cbranch_execnz .LBB9_1
; GFX908-NEXT: ; %bb.2: ; %atomicrmw.end
@@ -2273,25 +2273,25 @@ define void @buffer_fat_ptr_agent_atomic_fadd_noret_f64__offset__amdgpu_no_fine_
; GFX8: ; %bb.0:
; GFX8-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX8-NEXT: v_mov_b32_e32 v2, s20
-; GFX8-NEXT: buffer_load_dwordx2 v[4:5], v2, s[16:19], 0 offen offset:2048
+; GFX8-NEXT: buffer_load_dwordx2 v[2:3], v2, s[16:19], 0 offen offset:2048
; GFX8-NEXT: s_add_i32 s6, s20, 0x800
; GFX8-NEXT: s_mov_b64 s[4:5], 0
; GFX8-NEXT: v_mov_b32_e32 v6, s6
; GFX8-NEXT: .LBB9_1: ; %atomicrmw.start
; GFX8-NEXT: ; =>This Inner Loop Header: Depth=1
; GFX8-NEXT: s_waitcnt vmcnt(0)
-; GFX8-NEXT: v_add_f64 v[2:3], v[4:5], v[0:1]
-; GFX8-NEXT: v_mov_b32_e32 v10, v5
-; GFX8-NEXT: v_mov_b32_e32 v9, v4
-; GFX8-NEXT: v_mov_b32_e32 v8, v3
-; GFX8-NEXT: v_mov_b32_e32 v7, v2
-; GFX8-NEXT: buffer_atomic_cmpswap_x2 v[7:10], v6, s[16:19], 0 offen glc
+; GFX8-NEXT: v_mov_b32_e32 v10, v3
+; GFX8-NEXT: v_mov_b32_e32 v9, v2
+; GFX8-NEXT: v_add_f64 v[7:8], v[9:10], v[0:1]
+; GFX8-NEXT: v_mov_b32_e32 v2, v7
+; GFX8-NEXT: v_mov_b32_e32 v3, v8
+; GFX8-NEXT: v_mov_b32_e32 v4, v9
+; GFX8-NEXT: v_mov_b32_e32 v5, v10
+; GFX8-NEXT: buffer_atomic_cmpswap_x2 v[2:5], v6, s[16:19], 0 offen glc
; GFX8-NEXT: s_waitcnt vmcnt(0)
; GFX8-NEXT: buffer_wbinvl1
-; GFX8-NEXT: v_cmp_eq_u64_e32 vcc, v[7:8], v[4:5]
-; GFX8-NEXT: v_mov_b32_e32 v4, v7
+; GFX8-NEXT: v_cmp_eq_u64_e32 vcc, v[2:3], v[9:10]
; GFX8-NEXT: s_or_b64 s[4:5], vcc, s[4:5]
-; GFX8-NEXT: v_mov_b32_e32 v5, v8
; GFX8-NEXT: s_andn2_b64 exec, exec, s[4:5]
; GFX8-NEXT: s_cbranch_execnz .LBB9_1
; GFX8-NEXT: ; %bb.2: ; %atomicrmw.end
@@ -2302,25 +2302,25 @@ define void @buffer_fat_ptr_agent_atomic_fadd_noret_f64__offset__amdgpu_no_fine_
; GFX7: ; %bb.0:
; GFX7-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX7-NEXT: v_mov_b32_e32 v2, s20
-; GFX7-NEXT: buffer_load_dwordx2 v[4:5], v2, s[16:19], 0 offen offset:2048
+; GFX7-NEXT: buffer_load_dwordx2 v[2:3], v2, s[16:19], 0 offen offset:2048
; GFX7-NEXT: s_add_i32 s6, s20, 0x800
; GFX7-NEXT: s_mov_b64 s[4:5], 0
; GFX7-NEXT: v_mov_b32_e32 v6, s6
; GFX7-NEXT: .LBB9_1: ; %atomicrmw.start
; GFX7-NEXT: ; =>This Inner Loop Header: Depth=1
; GFX7-NEXT: s_waitcnt vmcnt(0)
-; GFX7-NEXT: v_add_f64 v[2:3], v[4:5], v[0:1]
-; GFX7-NEXT: v_mov_b32_e32 v10, v5
-; GFX7-NEXT: v_mov_b32_e32 v9, v4
-; GFX7-NEXT: v_mov_b32_e32 v8, v3
-; GFX7-NEXT: v_mov_b32_e32 v7, v2
-; GFX7-NEXT: buffer_atomic_cmpswap_x2 v[7:10], v6, s[16:19], 0 offen glc
+; GFX7-NEXT: v_mov_b32_e32 v10, v3
+; GFX7-NEXT: v_mov_b32_e32 v9, v2
+; GFX7-NEXT: v_add_f64 v[7:8], v[9:10], v[0:1]
+; GFX7-NEXT: v_mov_b32_e32 v2, v7
+; GFX7-NEXT: v_mov_b32_e32 v3, v8
+; GFX7-NEXT: v_mov_b32_e32 v4, v9
+; GFX7-NEXT: v_mov_b32_e32 v5, v10
+; GFX7-NEXT: buffer_atomic_cmpswap_x2 v[2:5], v6, s[16:19], 0 offen glc
; GFX7-NEXT: s_waitcnt vmcnt(0)
; GFX7-NEXT: buffer_wbinvl1
-; GFX7-NEXT: v_cmp_eq_u64_e32 vcc, v[7:8], v[4:5]
-; GFX7-NEXT: v_mov_b32_e32 v4, v7
+; GFX7-NEXT: v_cmp_eq_u64_e32 vcc, v[2:3], v[9:10]
; GFX7-NEXT: s_or_b64 s[4:5], vcc, s[4:5]
-; GFX7-NEXT: v_mov_b32_e32 v5, v8
; GFX7-NEXT: s_andn2_b64 exec, exec, s[4:5]
; GFX7-NEXT: s_cbranch_execnz .LBB9_1
; GFX7-NEXT: ; %bb.2: ; %atomicrmw.end
@@ -2331,26 +2331,26 @@ define void @buffer_fat_ptr_agent_atomic_fadd_noret_f64__offset__amdgpu_no_fine_
; GFX6: ; %bb.0:
; GFX6-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX6-NEXT: v_mov_b32_e32 v2, s20
-; GFX6-NEXT: buffer_load_dwordx2 v[4:5], v2, s[16:19], 0 offen offset:2048
+; GFX6-NEXT: buffer_load_dwordx2 v[2:3], v2, s[16:19], 0 offen offset:2048
; GFX6-NEXT: s_add_i32 s6, s20, 0x800
; GFX6-NEXT: s_mov_b64 s[4:5], 0
; GFX6-NEXT: v_mov_b32_e32 v6, s6
; GFX6-NEXT: .LBB9_1: ; %atomicrmw.start
; GFX6-NEXT: ; =>This Inner Loop Header: Depth=1
; GFX6-NEXT: s_waitcnt vmcnt(0)
-; GFX6-NEXT: v_add_f64 v[2:3], v[4:5], v[0:1]
+; GFX6-NEXT: v_mov_b32_e32 v10, v3
+; GFX6-NEXT: v_mov_b32_e32 v9, v2
+; GFX6-NEXT: v_add_f64 v[7:8], v[9:10], v[0:1]
; GFX6-NEXT: s_waitcnt expcnt(0)
-; GFX6-NEXT: v_mov_b32_e32 v10, v5
-; GFX6-NEXT: v_mov_b32_e32 v9, v4
-; GFX6-NEXT: v_mov_b32_e32 v8, v3
-; GFX6-NEXT: v_mov_b32_e32 v7, v2
-; GFX6-NEXT: buffer_atomic_cmpswap_x2 v[7:10], v6, s[16:19], 0 offen glc
+; GFX6-NEXT: v_mov_b32_e32 v2, v7
+; GFX6-NEXT: v_mov_b32_e32 v3, v8
+; GFX6-NEXT: v_mov_b32_e32 v4, v9
+; GFX6-NEXT: v_mov_b32_e32 v5, v10
+; GFX6-NEXT: buffer_atomic_cmpswap_x2 v[2:5], v6, s[16:19], 0 offen glc
; GFX6-NEXT: s_waitcnt vmcnt(0)
; GFX6-NEXT: buffer_wbinvl1
-; GFX6-NEXT: v_cmp_eq_u64_e32 vcc, v[7:8], v[4:5]
-; GFX6-NEXT: v_mov_b32_e32 v4, v7
+; GFX6-NEXT: v_cmp_eq_u64_e32 vcc, v[2:3], v[9:10]
; GFX6-NEXT: s_or_b64 s[4:5], vcc, s[4:5]
-; GFX6-NEXT: v_mov_b32_e32 v5, v8
; GFX6-NEXT: s_andn2_b64 exec, exec, s[4:5]
; GFX6-NEXT: s_cbranch_execnz .LBB9_1
; GFX6-NEXT: ; %bb.2: ; %atomicrmw.end
@@ -3427,41 +3427,41 @@ define half @buffer_fat_ptr_agent_atomic_fadd_ret_f16__offset__amdgpu_no_fine_gr
; GFX12-TRUE16-NEXT: s_wait_alu 0xfffe
; GFX12-TRUE16-NEXT: s_and_b32 s4, s16, -4
; GFX12-TRUE16-NEXT: s_wait_alu 0xfffe
-; GFX12-TRUE16-NEXT: v_mov_b32_e32 v5, s4
+; GFX12-TRUE16-NEXT: v_mov_b32_e32 v3, s4
; GFX12-TRUE16-NEXT: s_and_b32 s4, s16, 3
; GFX12-TRUE16-NEXT: s_wait_alu 0xfffe
; GFX12-TRUE16-NEXT: s_lshl_b32 s4, s4, 3
; GFX12-TRUE16-NEXT: s_wait_alu 0xfffe
; GFX12-TRUE16-NEXT: s_lshl_b32 s5, 0xffff, s4
-; GFX12-TRUE16-NEXT: buffer_load_b32 v2, v5, s[0:3], null offen
+; GFX12-TRUE16-NEXT: buffer_load_b32 v1, v3, s[0:3], null offen
; GFX12-TRUE16-NEXT: s_not_b32 s6, s5
; GFX12-TRUE16-NEXT: s_mov_b32 s5, 0
; GFX12-TRUE16-NEXT: .LBB13_1: ; %atomicrmw.start
; GFX12-TRUE16-NEXT: ; =>This Inner Loop Header: Depth=1
; GFX12-TRUE16-NEXT: s_wait_loadcnt 0x0
-; GFX12-TRUE16-NEXT: v_lshrrev_b32_e32 v1, s4, v2
+; GFX12-TRUE16-NEXT: v_mov_b32_e32 v5, v1
; GFX12-TRUE16-NEXT: s_wait_storecnt 0x0
; GFX12-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
+; GFX12-TRUE16-NEXT: v_lshrrev_b32_e32 v1, s4, v5
; GFX12-TRUE16-NEXT: v_add_f16_e32 v1.l, v1.l, v0.l
+; GFX12-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
; GFX12-TRUE16-NEXT: v_and_b32_e32 v1, 0xffff, v1
-; GFX12-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(SKIP_1) | instid1(VALU_DEP_1)
; GFX12-TRUE16-NEXT: v_lshlrev_b32_e32 v1, s4, v1
; GFX12-TRUE16-NEXT: s_wait_alu 0xfffe
-; GFX12-TRUE16-NEXT: v_and_or_b32 v1, v2, s6, v1
-; GFX12-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1)
-; GFX12-TRUE16-NEXT: v_dual_mov_b32 v4, v2 :: v_dual_mov_b32 v3, v1
-; GFX12-TRUE16-NEXT: buffer_atomic_cmpswap_b32 v[3:4], v5, s[0:3], null offen th:TH_ATOMIC_RETURN
+; GFX12-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
+; GFX12-TRUE16-NEXT: v_and_or_b32 v4, v5, s6, v1
+; GFX12-TRUE16-NEXT: v_dual_mov_b32 v1, v4 :: v_dual_mov_b32 v2, v5
+; GFX12-TRUE16-NEXT: buffer_atomic_cmpswap_b32 v[1:2], v3, s[0:3], null offen th:TH_ATOMIC_RETURN
; GFX12-TRUE16-NEXT: s_wait_loadcnt 0x0
; GFX12-TRUE16-NEXT: global_inv scope:SCOPE_DEV
-; GFX12-TRUE16-NEXT: v_cmp_eq_u32_e32 vcc_lo, v3, v2
-; GFX12-TRUE16-NEXT: v_mov_b32_e32 v2, v3
+; GFX12-TRUE16-NEXT: v_cmp_eq_u32_e32 vcc_lo, v1, v5
; GFX12-TRUE16-NEXT: s_or_b32 s5, vcc_lo, s5
; GFX12-TRUE16-NEXT: s_wait_alu 0xfffe
; GFX12-TRUE16-NEXT: s_and_not1_b32 exec_lo, exec_lo, s5
; GFX12-TRUE16-NEXT: s_cbranch_execnz .LBB13_1
; GFX12-TRUE16-NEXT: ; %bb.2: ; %atomicrmw.end
; GFX12-TRUE16-NEXT: s_or_b32 exec_lo, exec_lo, s5
-; GFX12-TRUE16-NEXT: v_lshrrev_b32_e32 v0, s4, v3
+; GFX12-TRUE16-NEXT: v_lshrrev_b32_e32 v0, s4, v1
; GFX12-TRUE16-NEXT: s_wait_loadcnt 0x0
; GFX12-TRUE16-NEXT: s_setpc_b64 s[30:31]
;
@@ -3476,41 +3476,41 @@ define half @buffer_fat_ptr_agent_atomic_fadd_ret_f16__offset__amdgpu_no_fine_gr
; GFX12-FAKE16-NEXT: s_wait_alu 0xfffe
; GFX12-FAKE16-NEXT: s_and_b32 s4, s16, -4
; GFX12-FAKE16-NEXT: s_wait_alu 0xfffe
-; GFX12-FAKE16-NEXT: v_mov_b32_e32 v5, s4
+; GFX12-FAKE16-NEXT: v_mov_b32_e32 v3, s4
; GFX12-FAKE16-NEXT: s_and_b32 s4, s16, 3
; GFX12-FAKE16-NEXT: s_wait_alu 0xfffe
; GFX12-FAKE16-NEXT: s_lshl_b32 s4, s4, 3
; GFX12-FAKE16-NEXT: s_wait_alu 0xfffe
; GFX12-FAKE16-NEXT: s_lshl_b32 s5, 0xffff, s4
-; GFX12-FAKE16-NEXT: buffer_load_b32 v2, v5, s[0:3], null offen
+; GFX12-FAKE16-NEXT: buffer_load_b32 v1, v3, s[0:3], null offen
; GFX12-FAKE16-NEXT: s_not_b32 s6, s5
; GFX12-FAKE16-NEXT: s_mov_b32 s5, 0
; GFX12-FAKE16-NEXT: .LBB13_1: ; %atomicrmw.start
; GFX12-FAKE16-NEXT: ; =>This Inner Loop Header: Depth=1
; GFX12-FAKE16-NEXT: s_wait_loadcnt 0x0
-; GFX12-FAKE16-NEXT: v_lshrrev_b32_e32 v1, s4, v2
+; GFX12-FAKE16-NEXT: v_mov_b32_e32 v5, v1
; GFX12-FAKE16-NEXT: s_wait_storecnt 0x0
; GFX12-FAKE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
+; GFX12-FAKE16-NEXT: v_lshrrev_b32_e32 v1, s4, v5
; GFX12-FAKE16-NEXT: v_add_f16_e32 v1, v1, v0
+; GFX12-FAKE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
; GFX12-FAKE16-NEXT: v_and_b32_e32 v1, 0xffff, v1
-; GFX12-FAKE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(SKIP_1) | instid1(VALU_DEP_1)
; GFX12-FAKE16-NEXT: v_lshlrev_b32_e32 v1, s4, v1
; GFX12-FAKE16-NEXT: s_wait_alu 0xfffe
-; GFX12-FAKE16-NEXT: v_and_or_b32 v1, v2, s6, v1
-; GFX12-FAKE16-NEXT: s_delay_alu instid0(VALU_DEP_1)
-; GFX12-FAKE16-NEXT: v_dual_mov_b32 v4, v2 :: v_dual_mov_b32 v3, v1
-; GFX12-FAKE16-NEXT: buffer_atomic_cmpswap_b32 v[3:4], v5, s[0:3], null offen th:TH_ATOMIC_RETURN
+; GFX12-FAKE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
+; GFX12-FAKE16-NEXT: v_and_or_b32 v4, v5, s6, v1
+; GFX12-FAKE16-NEXT: v_dual_mov_b32 v1, v4 :: v_dual_mov_b32 v2, v5
+; GFX12-FAKE16-NEXT: buffer_atomic_cmpswap_b32 v[1:2], v3, s[0:3], null offen th:TH_ATOMIC_RETURN
; GFX12-FAKE16-NEXT: s_wait_loadcnt 0x0
; GFX12-FAKE16-NEXT: global_inv scope:SCOPE_DEV
-; GFX12-FAKE16-NEXT: v_cmp_eq_u32_e32 vcc_lo, v3, v2
-; GFX12-FAKE16-NEXT: v_mov_b32_e32 v2, v3
+; GFX12-FAKE16-NEXT: v_cmp_eq_u32_e32 vcc_lo, v1, v5
; GFX12-FAKE16-NEXT: s_or_b32 s5, vcc_lo, s5
; GFX12-FAKE16-NEXT: s_wait_alu 0xfffe
; GFX12-FAKE16-NEXT: s_and_not1_b32 exec_lo, exec_lo, s5
; GFX12-FAKE16-NEXT: s_cbranch_execnz .LBB13_1
; GFX12-FAKE16-NEXT: ; %bb.2: ; %atomicrmw.end
; GFX12-FAKE16-NEXT: s_or_b32 exec_lo, exec_lo, s5
-; GFX12-FAKE16-NEXT: v_lshrrev_b32_e32 v0, s4, v3
+; GFX12-FAKE16-NEXT: v_lshrrev_b32_e32 v0, s4, v1
; GFX12-FAKE16-NEXT: s_wait_loadcnt 0x0
; GFX12-FAKE16-NEXT: s_setpc_b64 s[30:31]
;
@@ -3520,7 +3520,7 @@ define half @buffer_fat_ptr_agent_atomic_fadd_ret_f16__offset__amdgpu_no_fine_gr
; GFX942-NEXT: s_addk_i32 s16, 0x200
; GFX942-NEXT: s_and_b32 s4, s16, -4
; GFX942-NEXT: v_mov_b32_e32 v1, s4
-; GFX942-NEXT: buffer_load_dword v3, v1, s[0:3], 0 offen
+; GFX942-NEXT: buffer_load_dword v2, v1, s[0:3], 0 offen
; GFX942-NEXT: s_and_b32 s4, s16, 3
; GFX942-NEXT: s_lshl_b32 s6, s4, 3
; GFX942-NEXT: s_lshl_b32 s4, 0xffff, s6
@@ -3529,23 +3529,23 @@ define half @buffer_fat_ptr_agent_atomic_fadd_ret_f16__offset__amdgpu_no_fine_gr
; GFX942-NEXT: .LBB13_1: ; %atomicrmw.start
; GFX942-NEXT: ; =>This Inner Loop Header: Depth=1
; GFX942-NEXT: s_waitcnt vmcnt(0)
-; GFX942-NEXT: v_lshrrev_b32_e32 v2, s6, v3
+; GFX942-NEXT: v_mov_b32_e32 v5, v2
+; GFX942-NEXT: v_lshrrev_b32_e32 v2, s6, v5
; GFX942-NEXT: v_add_f16_e32 v2, v2, v0
; GFX942-NEXT: v_lshlrev_b32_e32 v2, s6, v2
-; GFX942-NEXT: v_and_or_b32 v2, v3, s7, v2
-; GFX942-NEXT: v_mov_b64_e32 v[4:5], v[2:3]
+; GFX942-NEXT: v_and_or_b32 v4, v5, s7, v2
+; GFX942-NEXT: v_mov_b64_e32 v[2:3], v[4:5]
; GFX942-NEXT: buffer_wbl2 sc1
-; GFX942-NEXT: buffer_atomic_cmpswap v[4:5], v1, s[0:3], 0 offen sc0
+; GFX942-NEXT: buffer_atomic_cmpswap v[2:3], v1, s[0:3], 0 offen sc0
; GFX942-NEXT: s_waitcnt vmcnt(0)
; GFX942-NEXT: buffer_inv sc1
-; GFX942-NEXT: v_cmp_eq_u32_e32 vcc, v4, v3
+; GFX942-NEXT: v_cmp_eq_u32_e32 vcc, v2, v5
; GFX942-NEXT: s_or_b64 s[4:5], vcc, s[4:5]
-; GFX942-NEXT: v_mov_b32_e32 v3, v4
; GFX942-NEXT: s_andn2_b64 exec, exec, s[4:5]
; GFX942-NEXT: s_cbranch_execnz .LBB13_1
; GFX942-NEXT: ; %bb.2: ; %atomicrmw.end
; GFX942-NEXT: s_or_b64 exec, exec, s[4:5]
-; GFX942-NEXT: v_lshrrev_b32_e32 v0, s6, v4
+; GFX942-NEXT: v_lshrrev_b32_e32 v0, s6, v2
; GFX942-NEXT: s_setpc_b64 s[30:31]
;
; GFX11-TRUE16-LABEL: buffer_fat_ptr_agent_atomic_fadd_ret_f16__offset__amdgpu_no_fine_grained_memory:
@@ -3554,40 +3554,40 @@ define half @buffer_fat_ptr_agent_atomic_fadd_ret_f16__offset__amdgpu_no_fine_gr
; GFX11-TRUE16-NEXT: s_addk_i32 s16, 0x200
; GFX11-TRUE16-NEXT: s_delay_alu instid0(SALU_CYCLE_1) | instskip(NEXT) | instid1(SALU_CYCLE_1)
; GFX11-TRUE16-NEXT: s_and_b32 s4, s16, -4
-; GFX11-TRUE16-NEXT: v_mov_b32_e32 v5, s4
+; GFX11-TRUE16-NEXT: v_mov_b32_e32 v3, s4
; GFX11-TRUE16-NEXT: s_and_b32 s4, s16, 3
; GFX11-TRUE16-NEXT: s_delay_alu instid0(SALU_CYCLE_1) | instskip(NEXT) | instid1(SALU_CYCLE_1)
; GFX11-TRUE16-NEXT: s_lshl_b32 s4, s4, 3
; GFX11-TRUE16-NEXT: s_lshl_b32 s5, 0xffff, s4
-; GFX11-TRUE16-NEXT: buffer_load_b32 v2, v5, s[0:3], 0 offen
+; GFX11-TRUE16-NEXT: buffer_load_b32 v1, v3, s[0:3], 0 offen
; GFX11-TRUE16-NEXT: s_not_b32 s6, s5
; GFX11-TRUE16-NEXT: s_mov_b32 s5, 0
; GFX11-TRUE16-NEXT: .LBB13_1: ; %atomicrmw.start
; GFX11-TRUE16-NEXT: ; =>This Inner Loop Header: Depth=1
; GFX11-TRUE16-NEXT: s_waitcnt vmcnt(0)
-; GFX11-TRUE16-NEXT: v_lshrrev_b32_e32 v1, s4, v2
+; GFX11-TRUE16-NEXT: v_mov_b32_e32 v5, v1
; GFX11-TRUE16-NEXT: s_waitcnt_vscnt null, 0x0
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
+; GFX11-TRUE16-NEXT: v_lshrrev_b32_e32 v1, s4, v5
; GFX11-TRUE16-NEXT: v_add_f16_e32 v1.l, v1.l, v0.l
-; GFX11-TRUE16-NEXT: v_and_b32_e32 v1, 0xffff, v1
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
+; GFX11-TRUE16-NEXT: v_and_b32_e32 v1, 0xffff, v1
; GFX11-TRUE16-NEXT: v_lshlrev_b32_e32 v1, s4, v1
-; GFX11-TRUE16-NEXT: v_and_or_b32 v1, v2, s6, v1
-; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1)
-; GFX11-TRUE16-NEXT: v_dual_mov_b32 v4, v2 :: v_dual_mov_b32 v3, v1
-; GFX11-TRUE16-NEXT: buffer_atomic_cmpswap_b32 v[3:4], v5, s[0:3], 0 offen glc
+; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
+; GFX11-TRUE16-NEXT: v_and_or_b32 v4, v5, s6, v1
+; GFX11-TRUE16-NEXT: v_dual_mov_b32 v1, v4 :: v_dual_mov_b32 v2, v5
+; GFX11-TRUE16-NEXT: buffer_atomic_cmpswap_b32 v[1:2], v3, s[0:3], 0 offen glc
; GFX11-TRUE16-NEXT: s_waitcnt vmcnt(0)
; GFX11-TRUE16-NEXT: buffer_gl1_inv
; GFX11-TRUE16-NEXT: buffer_gl0_inv
-; GFX11-TRUE16-NEXT: v_cmp_eq_u32_e32 vcc_lo, v3, v2
-; GFX11-TRUE16-NEXT: v_mov_b32_e32 v2, v3
+; GFX11-TRUE16-NEXT: v_cmp_eq_u32_e32 vcc_lo, v1, v5
; GFX11-TRUE16-NEXT: s_or_b32 s5, vcc_lo, s5
; GFX11-TRUE16-NEXT: s_delay_alu instid0(SALU_CYCLE_1)
; GFX11-TRUE16-NEXT: s_and_not1_b32 exec_lo, exec_lo, s5
; GFX11-TRUE16-NEXT: s_cbranch_execnz .LBB13_1
; GFX11-TRUE16-NEXT: ; %bb.2: ; %atomicrmw.end
; GFX11-TRUE16-NEXT: s_or_b32 exec_lo, exec_lo, s5
-; GFX11-TRUE16-NEXT: v_lshrrev_b32_e32 v0, s4, v3
+; GFX11-TRUE16-NEXT: v_lshrrev_b32_e32 v0, s4, v1
; GFX11-TRUE16-NEXT: s_setpc_b64 s[30:31]
;
; GFX11-FAKE16-LABEL: buffer_fat_ptr_agent_atomic_fadd_ret_f16__offset__amdgpu_no_fine_grained_memory:
@@ -3596,40 +3596,40 @@ define half @buffer_fat_ptr_agent_atomic_fadd_ret_f16__offset__amdgpu_no_fine_gr
; GFX11-FAKE16-NEXT: s_addk_i32 s16, 0x200
; GFX11-FAKE16-NEXT: s_delay_alu instid0(SALU_CYCLE_1) | instskip(NEXT) | instid1(SALU_CYCLE_1)
; GFX11-FAKE16-NEXT: s_and_b32 s4, s16, -4
-; GFX11-FAKE16-NEXT: v_mov_b32_e32 v5, s4
+; GFX11-FAKE16-NEXT: v_mov_b32_e32 v3, s4
; GFX11-FAKE16-NEXT: s_and_b32 s4, s16, 3
; GFX11-FAKE16-NEXT: s_delay_alu instid0(SALU_CYCLE_1) | instskip(NEXT) | instid1(SALU_CYCLE_1)
; GFX11-FAKE16-NEXT: s_lshl_b32 s4, s4, 3
; GFX11-FAKE16-NEXT: s_lshl_b32 s5, 0xffff, s4
-; GFX11-FAKE16-NEXT: buffer_load_b32 v2, v5, s[0:3], 0 offen
+; GFX11-FAKE16-NEXT: buffer_load_b32 v1, v3, s[0:3], 0 offen
; GFX11-FAKE16-NEXT: s_not_b32 s6, s5
; GFX11-FAKE16-NEXT: s_mov_b32 s5, 0
; GFX11-FAKE16-NEXT: .LBB13_1: ; %atomicrmw.start
; GFX11-FAKE16-NEXT: ; =>This Inner Loop Header: Depth=1
; GFX11-FAKE16-NEXT: s_waitcnt vmcnt(0)
-; GFX11-FAKE16-NEXT: v_lshrrev_b32_e32 v1, s4, v2
+; GFX11-FAKE16-NEXT: v_mov_b32_e32 v5, v1
; GFX11-FAKE16-NEXT: s_waitcnt_vscnt null, 0x0
; GFX11-FAKE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
+; GFX11-FAKE16-NEXT: v_lshrrev_b32_e32 v1, s4, v5
; GFX11-FAKE16-NEXT: v_add_f16_e32 v1, v1, v0
-; GFX11-FAKE16-NEXT: v_and_b32_e32 v1, 0xffff, v1
; GFX11-FAKE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
+; GFX11-FAKE16-NEXT: v_and_b32_e32 v1, 0xffff, v1
; GFX11-FAKE16-NEXT: v_lshlrev_b32_e32 v1, s4, v1
-; GFX11-FAKE16-NEXT: v_and_or_b32 v1, v2, s6, v1
-; GFX11-FAKE16-NEXT: s_delay_alu instid0(VALU_DEP_1)
-; GFX11-FAKE16-NEXT: v_dual_mov_b32 v4, v2 :: v_dual_mov_b32 v3, v1
-; GFX11-FAKE16-NEXT: buffer_atomic_cmpswap_b32 v[3:4], v5, s[0:3], 0 offen glc
+; GFX11-FAKE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
+; GFX11-FAKE16-NEXT: v_and_or_b32 v4, v5, s6, v1
+; GFX11-FAKE16-NEXT: v_dual_mov_b32 v1, v4 :: v_dual_mov_b32 v2, v5
+; GFX11-FAKE16-NEXT: buffer_atomic_cmpswap_b32 v[1:2], v3, s[0:3], 0 offen glc
; GFX11-FAKE16-NEXT: s_waitcnt vmcnt(0)
; GFX11-FAKE16-NEXT: buffer_gl1_inv
; GFX11-FAKE16-NEXT: buffer_gl0_inv
-; GFX11-FAKE16-NEXT: v_cmp_eq_u32_e32 vcc_lo, v3, v2
-; GFX11-FAKE16-NEXT: v_mov_b32_e32 v2, v3
+; GFX11-FAKE16-NEXT: v_cmp_eq_u32_e32 vcc_lo, v1, v5
; GFX11-FAKE16-NEXT: s_or_b32 s5, vcc_lo, s5
; GFX11-FAKE16-NEXT: s_delay_alu instid0(SALU_CYCLE_1)
; GFX11-FAKE16-NEXT: s_and_not1_b32 exec_lo, exec_lo, s5
; GFX11-FAKE16-NEXT: s_cbranch_execnz .LBB13_1
; GFX11-FAKE16-NEXT: ; %bb.2: ; %atomicrmw.end
; GFX11-FAKE16-NEXT: s_or_b32 exec_lo, exec_lo, s5
-; GFX11-FAKE16-NEXT: v_lshrrev_b32_e32 v0, s4, v3
+; GFX11-FAKE16-NEXT: v_lshrrev_b32_e32 v0, s4, v1
; GFX11-FAKE16-NEXT: s_setpc_b64 s[30:31]
;
; GFX10-LABEL: buffer_fat_ptr_agent_atomic_fadd_ret_f16__offset__amdgpu_no_fine_grained_memory:
@@ -3637,35 +3637,35 @@ define half @buffer_fat_ptr_agent_atomic_fadd_ret_f16__offset__amdgpu_no_fine_gr
; GFX10-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX10-NEXT: s_addk_i32 s20, 0x200
; GFX10-NEXT: s_and_b32 s4, s20, -4
-; GFX10-NEXT: v_mov_b32_e32 v5, s4
+; GFX10-NEXT: v_mov_b32_e32 v3, s4
; GFX10-NEXT: s_and_b32 s4, s20, 3
; GFX10-NEXT: s_lshl_b32 s4, s4, 3
; GFX10-NEXT: s_lshl_b32 s5, 0xffff, s4
-; GFX10-NEXT: buffer_load_dword v2, v5, s[16:19], 0 offen
+; GFX10-NEXT: buffer_load_dword v1, v3, s[16:19], 0 offen
; GFX10-NEXT: s_not_b32 s6, s5
; GFX10-NEXT: s_mov_b32 s5, 0
; GFX10-NEXT: .LBB13_1: ; %atomicrmw.start
; GFX10-NEXT: ; =>This Inner Loop Header: Depth=1
; GFX10-NEXT: s_waitcnt vmcnt(0)
-; GFX10-NEXT: v_lshrrev_b32_e32 v1, s4, v2
+; GFX10-NEXT: v_mov_b32_e32 v5, v1
; GFX10-NEXT: s_waitcnt_vscnt null, 0x0
+; GFX10-NEXT: v_lshrrev_b32_e32 v1, s4, v5
; GFX10-NEXT: v_add_f16_e32 v1, v1, v0
; GFX10-NEXT: v_lshlrev_b32_sdwa v1, s4, v1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:WORD_0
-; GFX10-NEXT: v_and_or_b32 v1, v2, s6, v1
-; GFX10-NEXT: v_mov_b32_e32 v4, v2
-; GFX10-NEXT: v_mov_b32_e32 v3, v1
-; GFX10-NEXT: buffer_atomic_cmpswap v[3:4], v5, s[16:19], 0 offen glc
+; GFX10-NEXT: v_and_or_b32 v4, v5, s6, v1
+; GFX10-NEXT: v_mov_b32_e32 v1, v4
+; GFX10-NEXT: v_mov_b32_e32 v2, v5
+; GFX10-NEXT: buffer_atomic_cmpswap v[1:2], v3, s[16:19], 0 offen glc
; GFX10-NEXT: s_waitcnt vmcnt(0)
; GFX10-NEXT: buffer_gl1_inv
; GFX10-NEXT: buffer_gl0_inv
-; GFX10-NEXT: v_cmp_eq_u32_e32 vcc_lo, v3, v2
-; GFX10-NEXT: v_mov_b32_e32 v2, v3
+; GFX10-NEXT: v_cmp_eq_u32_e32 vcc_lo, v1, v5
; GFX10-NEXT: s_or_b32 s5, vcc_lo, s5
; GFX10-NEXT: s_andn2_b32 exec_lo, exec_lo, s5
; GFX10-NEXT: s_cbranch_execnz .LBB13_1
; GFX10-NEXT: ; %bb.2: ; %atomicrmw.end
; GFX10-NEXT: s_or_b32 exec_lo, exec_lo, s5
-; GFX10-NEXT: v_lshrrev_b32_e32 v0, s4, v3
+; GFX10-NEXT: v_lshrrev_b32_e32 v0, s4, v1
; GFX10-NEXT: s_setpc_b64 s[30:31]
;
; GFX90A-LABEL: buffer_fat_ptr_agent_atomic_fadd_ret_f16__offset__amdgpu_no_fine_grained_memory:
@@ -3674,7 +3674,7 @@ define half @buffer_fat_ptr_agent_atomic_fadd_ret_f16__offset__amdgpu_no_fine_gr
; GFX90A-NEXT: s_addk_i32 s20, 0x200
; GFX90A-NEXT: s_and_b32 s4, s20, -4
; GFX90A-NEXT: v_mov_b32_e32 v1, s4
-; GFX90A-NEXT: buffer_load_dword v3, v1, s[16:19], 0 offen
+; GFX90A-NEXT: buffer_load_dword v2, v1, s[16:19], 0 offen
; GFX90A-NEXT: s_and_b32 s4, s20, 3
; GFX90A-NEXT: s_lshl_b32 s6, s4, 3
; GFX90A-NEXT: s_lshl_b32 s4, 0xffff, s6
@@ -3683,22 +3683,22 @@ define half @buffer_fat_ptr_agent_atomic_fadd_ret_f16__offset__amdgpu_no_fine_gr
; GFX90A-NEXT: .LBB13_1: ; %atomicrmw.start
; GFX90A-NEXT: ; =>This Inner Loop Header: Depth=1
; GFX90A-NEXT: s_waitcnt vmcnt(0)
-; GFX90A-NEXT: v_lshrrev_b32_e32 v2, s6, v3
+; GFX90A-NEXT: v_mov_b32_e32 v5, v2
+; GFX90A-NEXT: v_lshrrev_b32_e32 v2, s6, v5
; GFX90A-NEXT: v_add_f16_e32 v2, v2, v0
; GFX90A-NEXT: v_lshlrev_b32_e32 v2, s6, v2
-; GFX90A-NEXT: v_and_or_b32 v2, v3, s7, v2
-; GFX90A-NEXT: v_pk_mov_b32 v[4:5], v[2:3], v[2:3] op_sel:[0,1]
-; GFX90A-NEXT: buffer_atomic_cmpswap v[4:5], v1, s[16:19], 0 offen glc
+; GFX90A-NEXT: v_and_or_b32 v4, v5, s7, v2
+; GFX90A-NEXT: v_pk_mov_b32 v[2:3], v[4:5], v[4:5] op_sel:[0,1]
+; GFX90A-NEXT: buffer_atomic_cmpswap v[2:3], v1, s[16:19], 0 offen glc
; GFX90A-NEXT: s_waitcnt vmcnt(0)
; GFX90A-NEXT: buffer_wbinvl1
-; GFX90A-NEXT: v_cmp_eq_u32_e32 vcc, v4, v3
+; GFX90A-NEXT: v_cmp_eq_u32_e32 vcc, v2, v5
; GFX90A-NEXT: s_or_b64 s[4:5], vcc, s[4:5]
-; GFX90A-NEXT: v_mov_b32_e32 v3, v4
; GFX90A-NEXT: s_andn2_b64 exec, exec, s[4:5]
; GFX90A-NEXT: s_cbranch_execnz .LBB13_1
; GFX90A-NEXT: ; %bb.2: ; %atomicrmw.end
; GFX90A-NEXT: s_or_b64 exec, exec, s[4:5]
-; GFX90A-NEXT: v_lshrrev_b32_e32 v0, s6, v4
+; GFX90A-NEXT: v_lshrrev_b32_e32 v0, s6, v2
; GFX90A-NEXT: s_setpc_b64 s[30:31]
;
; GFX908-LABEL: buffer_fat_ptr_agent_atomic_fadd_ret_f16__offset__amdgpu_no_fine_grained_memory:
@@ -3706,8 +3706,8 @@ define half @buffer_fat_ptr_agent_atomic_fadd_ret_f16__offset__amdgpu_no_fine_gr
; GFX908-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX908-NEXT: s_addk_i32 s20, 0x200
; GFX908-NEXT: s_and_b32 s4, s20, -4
-; GFX908-NEXT: v_mov_b32_e32 v5, s4
-; GFX908-NEXT: buffer_load_dword v2, v5, s[16:19], 0 offen
+; GFX908-NEXT: v_mov_b32_e32 v3, s4
+; GFX908-NEXT: buffer_load_dword v1, v3, s[16:19], 0 offen
; GFX908-NEXT: s_and_b32 s4, s20, 3
; GFX908-NEXT: s_lshl_b32 s6, s4, 3
; GFX908-NEXT: s_lshl_b32 s4, 0xffff, s6
@@ -3716,23 +3716,23 @@ define half @buffer_fat_ptr_agent_atomic_fadd_ret_f16__offset__amdgpu_no_fine_gr
; GFX908-NEXT: .LBB13_1: ; %atomicrmw.start
; GFX908-NEXT: ; =>This Inner Loop Header: Depth=1
; GFX908-NEXT: s_waitcnt vmcnt(0)
-; GFX908-NEXT: v_lshrrev_b32_e32 v1, s6, v2
+; GFX908-NEXT: v_mov_b32_e32 v5, v1
+; GFX908-NEXT: v_lshrrev_b32_e32 v1, s6, v5
; GFX908-NEXT: v_add_f16_e32 v1, v1, v0
; GFX908-NEXT: v_lshlrev_b32_e32 v1, s6, v1
-; GFX908-NEXT: v_and_or_b32 v1, v2, s7, v1
-; GFX908-NEXT: v_mov_b32_e32 v4, v2
-; GFX908-NEXT: v_mov_b32_e32 v3, v1
-; GFX908-NEXT: buffer_atomic_cmpswap v[3:4], v5, s[16:19], 0 offen glc
+; GFX908-NEXT: v_and_or_b32 v4, v5, s7, v1
+; GFX908-NEXT: v_mov_b32_e32 v1, v4
+; GFX908-NEXT: v_mov_b32_e32 v2, v5
+; GFX908-NEXT: buffer_atomic_cmpswap v[1:2], v3, s[16:19], 0 offen glc
; GFX908-NEXT: s_waitcnt vmcnt(0)
; GFX908-NEXT: buffer_wbinvl1
-; GFX908-NEXT: v_cmp_eq_u32_e32 vcc, v3, v2
+; GFX908-NEXT: v_cmp_eq_u32_e32 vcc, v1, v5
; GFX908-NEXT: s_or_b64 s[4:5], vcc, s[4:5]
-; GFX908-NEXT: v_mov_b32_e32 v2, v3
; GFX908-NEXT: s_andn2_b64 exec, exec, s[4:5]
; GFX908-NEXT: s_cbranch_execnz .LBB13_1
; GFX908-NEXT: ; %bb.2: ; %atomicrmw.end
; GFX908-NEXT: s_or_b64 exec, exec, s[4:5]
-; GFX908-NEXT: v_lshrrev_b32_e32 v0, s6, v3
+; GFX908-NEXT: v_lshrrev_b32_e32 v0, s6, v1
; GFX908-NEXT: s_setpc_b64 s[30:31]
;
; GFX8-LABEL: buffer_fat_ptr_agent_atomic_fadd_ret_f16__offset__amdgpu_no_fine_grained_memory:
@@ -3740,8 +3740,8 @@ define half @buffer_fat_ptr_agent_atomic_fadd_ret_f16__offset__amdgpu_no_fine_gr
; GFX8-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX8-NEXT: s_addk_i32 s20, 0x200
; GFX8-NEXT: s_and_b32 s4, s20, -4
-; GFX8-NEXT: v_mov_b32_e32 v5, s4
-; GFX8-NEXT: buffer_load_dword v2, v5, s[16:19], 0 offen
+; GFX8-NEXT: v_mov_b32_e32 v3, s4
+; GFX8-NEXT: buffer_load_dword v1, v3, s[16:19], 0 offen
; GFX8-NEXT: s_and_b32 s4, s20, 3
; GFX8-NEXT: s_lshl_b32 s6, s4, 3
; GFX8-NEXT: s_lshl_b32 s4, 0xffff, s6
@@ -3750,24 +3750,24 @@ define half @buffer_fat_ptr_agent_atomic_fadd_ret_f16__offset__amdgpu_no_fine_gr
; GFX8-NEXT: .LBB13_1: ; %atomicrmw.start
; GFX8-NEXT: ; =>This Inner Loop Header: Depth=1
; GFX8-NEXT: s_waitcnt vmcnt(0)
-; GFX8-NEXT: v_lshrrev_b32_e32 v1, s6, v2
+; GFX8-NEXT: v_mov_b32_e32 v5, v1
+; GFX8-NEXT: v_lshrrev_b32_e32 v1, s6, v5
; GFX8-NEXT: v_add_f16_e32 v1, v1, v0
-; GFX8-NEXT: v_and_b32_e32 v3, s7, v2
+; GFX8-NEXT: v_and_b32_e32 v2, s7, v5
; GFX8-NEXT: v_lshlrev_b32_e32 v1, s6, v1
-; GFX8-NEXT: v_or_b32_e32 v1, v3, v1
-; GFX8-NEXT: v_mov_b32_e32 v4, v2
-; GFX8-NEXT: v_mov_b32_e32 v3, v1
-; GFX8-NEXT: buffer_atomic_cmpswap v[3:4], v5, s[16:19], 0 offen glc
+; GFX8-NEXT: v_or_b32_e32 v4, v2, v1
+; GFX8-NEXT: v_mov_b32_e32 v1, v4
+; GFX8-NEXT: v_mov_b32_e32 v2, v5
+; GFX8-NEXT: buffer_atomic_cmpswap v[1:2], v3, s[16:19], 0 offen glc
; GFX8-NEXT: s_waitcnt vmcnt(0)
; GFX8-NEXT: buffer_wbinvl1
-; GFX8-NEXT: v_cmp_eq_u32_e32 vcc, v3, v2
+; GFX8-NEXT: v_cmp_eq_u32_e32 vcc, v1, v5
; GFX8-NEXT: s_or_b64 s[4:5], vcc, s[4:5]
-; GFX8-NEXT: v_mov_b32_e32 v2, v3
; GFX8-NEXT: s_andn2_b64 exec, exec, s[4:5]
; GFX8-NEXT: s_cbranch_execnz .LBB13_1
; GFX8-NEXT: ; %bb.2: ; %atomicrmw.end
; GFX8-NEXT: s_or_b64 exec, exec, s[4:5]
-; GFX8-NEXT: v_lshrrev_b32_e32 v0, s6, v3
+; GFX8-NEXT: v_lshrrev_b32_e32 v0, s6, v1
; GFX8-NEXT: s_setpc_b64 s[30:31]
;
; GFX7-LABEL: buffer_fat_ptr_agent_atomic_fadd_ret_f16__offset__amdgpu_no_fine_grained_memory:
@@ -3775,38 +3775,38 @@ define half @buffer_fat_ptr_agent_atomic_fadd_ret_f16__offset__amdgpu_no_fine_gr
; GFX7-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX7-NEXT: s_addk_i32 s20, 0x200
; GFX7-NEXT: s_and_b32 s4, s20, -4
-; GFX7-NEXT: v_mov_b32_e32 v4, s4
-; GFX7-NEXT: buffer_load_dword v1, v4, s[16:19], 0 offen
+; GFX7-NEXT: v_mov_b32_e32 v3, s4
+; GFX7-NEXT: buffer_load_dword v1, v3, s[16:19], 0 offen
; GFX7-NEXT: v_cvt_f16_f32_e32 v0, v0
; GFX7-NEXT: s_and_b32 s4, s20, 3
; GFX7-NEXT: s_lshl_b32 s6, s4, 3
; GFX7-NEXT: s_lshl_b32 s4, 0xffff, s6
-; GFX7-NEXT: v_cvt_f32_f16_e32 v5, v0
+; GFX7-NEXT: v_cvt_f32_f16_e32 v0, v0
; GFX7-NEXT: s_not_b32 s7, s4
; GFX7-NEXT: s_mov_b64 s[4:5], 0
; GFX7-NEXT: .LBB13_1: ; %atomicrmw.start
; GFX7-NEXT: ; =>This Inner Loop Header: Depth=1
; GFX7-NEXT: s_waitcnt vmcnt(0)
-; GFX7-NEXT: v_lshrrev_b32_e32 v0, s6, v1
-; GFX7-NEXT: v_cvt_f32_f16_e32 v0, v0
-; GFX7-NEXT: v_and_b32_e32 v2, s7, v1
-; GFX7-NEXT: v_add_f32_e32 v0, v0, v5
-; GFX7-NEXT: v_cvt_f16_f32_e32 v0, v0
-; GFX7-NEXT: v_lshlrev_b32_e32 v0, s6, v0
-; GFX7-NEXT: v_or_b32_e32 v0, v2, v0
-; GFX7-NEXT: v_mov_b32_e32 v3, v1
-; GFX7-NEXT: v_mov_b32_e32 v2, v0
-; GFX7-NEXT: buffer_atomic_cmpswap v[2:3], v4, s[16:19], 0 offen glc
+; GFX7-NEXT: v_mov_b32_e32 v5, v1
+; GFX7-NEXT: v_lshrrev_b32_e32 v1, s6, v5
+; GFX7-NEXT: v_cvt_f32_f16_e32 v1, v1
+; GFX7-NEXT: v_and_b32_e32 v2, s7, v5
+; GFX7-NEXT: v_add_f32_e32 v1, v1, v0
+; GFX7-NEXT: v_cvt_f16_f32_e32 v1, v1
+; GFX7-NEXT: v_lshlrev_b32_e32 v1, s6, v1
+; GFX7-NEXT: v_or_b32_e32 v4, v2, v1
+; GFX7-NEXT: v_mov_b32_e32 v1, v4
+; GFX7-NEXT: v_mov_b32_e32 v2, v5
+; GFX7-NEXT: buffer_atomic_cmpswap v[1:2], v3, s[16:19], 0 offen glc
; GFX7-NEXT: s_waitcnt vmcnt(0)
; GFX7-NEXT: buffer_wbinvl1
-; GFX7-NEXT: v_cmp_eq_u32_e32 vcc, v2, v1
+; GFX7-NEXT: v_cmp_eq_u32_e32 vcc, v1, v5
; GFX7-NEXT: s_or_b64 s[4:5], vcc, s[4:5]
-; GFX7-NEXT: v_mov_b32_e32 v1, v2
; GFX7-NEXT: s_andn2_b64 exec, exec, s[4:5]
; GFX7-NEXT: s_cbranch_execnz .LBB13_1
; GFX7-NEXT: ; %bb.2: ; %atomicrmw.end
; GFX7-NEXT: s_or_b64 exec, exec, s[4:5]
-; GFX7-NEXT: v_lshrrev_b32_e32 v0, s6, v2
+; GFX7-NEXT: v_lshrrev_b32_e32 v0, s6, v1
; GFX7-NEXT: v_cvt_f32_f16_e32 v0, v0
; GFX7-NEXT: s_setpc_b64 s[30:31]
;
@@ -3815,39 +3815,39 @@ define half @buffer_fat_ptr_agent_atomic_fadd_ret_f16__offset__amdgpu_no_fine_gr
; GFX6-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX6-NEXT: s_addk_i32 s20, 0x200
; GFX6-NEXT: s_and_b32 s4, s20, -4
-; GFX6-NEXT: v_mov_b32_e32 v4, s4
-; GFX6-NEXT: buffer_load_dword v1, v4, s[16:19], 0 offen
+; GFX6-NEXT: v_mov_b32_e32 v3, s4
+; GFX6-NEXT: buffer_load_dword v1, v3, s[16:19], 0 offen
; GFX6-NEXT: v_cvt_f16_f32_e32 v0, v0
; GFX6-NEXT: s_and_b32 s4, s20, 3
; GFX6-NEXT: s_lshl_b32 s6, s4, 3
; GFX6-NEXT: s_lshl_b32 s4, 0xffff, s6
-; GFX6-NEXT: v_cvt_f32_f16_e32 v5, v0
+; GFX6-NEXT: v_cvt_f32_f16_e32 v0, v0
; GFX6-NEXT: s_not_b32 s7, s4
; GFX6-NEXT: s_mov_b64 s[4:5], 0
; GFX6-NEXT: .LBB13_1: ; %atomicrmw.start
; GFX6-NEXT: ; =>This Inner Loop Header: Depth=1
; GFX6-NEXT: s_waitcnt vmcnt(0)
-; GFX6-NEXT: v_lshrrev_b32_e32 v0, s6, v1
-; GFX6-NEXT: v_cvt_f32_f16_e32 v0, v0
+; GFX6-NEXT: v_mov_b32_e32 v5, v1
; GFX6-NEXT: s_waitcnt expcnt(0)
-; GFX6-NEXT: v_and_b32_e32 v2, s7, v1
-; GFX6-NEXT: v_add_f32_e32 v0, v0, v5
-; GFX6-NEXT: v_cvt_f16_f32_e32 v0, v0
-; GFX6-NEXT: v_lshlrev_b32_e32 v0, s6, v0
-; GFX6-NEXT: v_or_b32_e32 v0, v2, v0
-; GFX6-NEXT: v_mov_b32_e32 v3, v1
-; GFX6-NEXT: v_mov_b32_e32 v2, v0
-; GFX6-NEXT: buffer_atomic_cmpswap v[2:3], v4, s[16:19], 0 offen glc
+; GFX6-NEXT: v_lshrrev_b32_e32 v1, s6, v5
+; GFX6-NEXT: v_cvt_f32_f16_e32 v1, v1
+; GFX6-NEXT: v_and_b32_e32 v2, s7, v5
+; GFX6-NEXT: v_add_f32_e32 v1, v1, v0
+; GFX6-NEXT: v_cvt_f16_f32_e32 v1, v1
+; GFX6-NEXT: v_lshlrev_b32_e32 v1, s6, v1
+; GFX6-NEXT: v_or_b32_e32 v4, v2, v1
+; GFX6-NEXT: v_mov_b32_e32 v1, v4
+; GFX6-NEXT: v_mov_b32_e32 v2, v5
+; GFX6-NEXT: buffer_atomic_cmpswap v[1:2], v3, s[16:19], 0 offen glc
; GFX6-NEXT: s_waitcnt vmcnt(0)
; GFX6-NEXT: buffer_wbinvl1
-; GFX6-NEXT: v_cmp_eq_u32_e32 vcc, v2, v1
+; GFX6-NEXT: v_cmp_eq_u32_e32 vcc, v1, v5
; GFX6-NEXT: s_or_b64 s[4:5], vcc, s[4:5]
-; GFX6-NEXT: v_mov_b32_e32 v1, v2
; GFX6-NEXT: s_andn2_b64 exec, exec, s[4:5]
; GFX6-NEXT: s_cbranch_execnz .LBB13_1
; GFX6-NEXT: ; %bb.2: ; %atomicrmw.end
; GFX6-NEXT: s_or_b64 exec, exec, s[4:5]
-; GFX6-NEXT: v_lshrrev_b32_e32 v0, s6, v2
+; GFX6-NEXT: v_lshrrev_b32_e32 v0, s6, v1
; GFX6-NEXT: v_cvt_f32_f16_e32 v0, v0
; GFX6-NEXT: s_waitcnt expcnt(0)
; GFX6-NEXT: s_setpc_b64 s[30:31]
@@ -3874,28 +3874,28 @@ define void @buffer_fat_ptr_agent_atomic_fadd_noret_f16__offset__amdgpu_no_fine_
; GFX12-TRUE16-NEXT: s_lshl_b32 s4, s4, 3
; GFX12-TRUE16-NEXT: s_wait_alu 0xfffe
; GFX12-TRUE16-NEXT: s_lshl_b32 s5, 0xffff, s4
-; GFX12-TRUE16-NEXT: buffer_load_b32 v2, v3, s[0:3], null offen
+; GFX12-TRUE16-NEXT: buffer_load_b32 v1, v3, s[0:3], null offen
; GFX12-TRUE16-NEXT: s_not_b32 s6, s5
; GFX12-TRUE16-NEXT: s_mov_b32 s5, 0
; GFX12-TRUE16-NEXT: .LBB14_1: ; %atomicrmw.start
; GFX12-TRUE16-NEXT: ; =>This Inner Loop Header: Depth=1
; GFX12-TRUE16-NEXT: s_wait_loadcnt 0x0
-; GFX12-TRUE16-NEXT: v_lshrrev_b32_e32 v1, s4, v2
+; GFX12-TRUE16-NEXT: v_mov_b32_e32 v5, v1
; GFX12-TRUE16-NEXT: s_wait_storecnt 0x0
; GFX12-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
+; GFX12-TRUE16-NEXT: v_lshrrev_b32_e32 v1, s4, v5
; GFX12-TRUE16-NEXT: v_add_f16_e32 v1.l, v1.l, v0.l
+; GFX12-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
; GFX12-TRUE16-NEXT: v_and_b32_e32 v1, 0xffff, v1
-; GFX12-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(SKIP_1) | instid1(VALU_DEP_1)
; GFX12-TRUE16-NEXT: v_lshlrev_b32_e32 v1, s4, v1
; GFX12-TRUE16-NEXT: s_wait_alu 0xfffe
-; GFX12-TRUE16-NEXT: v_and_or_b32 v1, v2, s6, v1
-; GFX12-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1)
-; GFX12-TRUE16-NEXT: v_dual_mov_b32 v5, v2 :: v_dual_mov_b32 v4, v1
-; GFX12-TRUE16-NEXT: buffer_atomic_cmpswap_b32 v[4:5], v3, s[0:3], null offen th:TH_ATOMIC_RETURN
+; GFX12-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
+; GFX12-TRUE16-NEXT: v_and_or_b32 v4, v5, s6, v1
+; GFX12-TRUE16-NEXT: v_dual_mov_b32 v1, v4 :: v_dual_mov_b32 v2, v5
+; GFX12-TRUE16-NEXT: buffer_atomic_cmpswap_b32 v[1:2], v3, s[0:3], null offen th:TH_ATOMIC_RETURN
; GFX12-TRUE16-NEXT: s_wait_loadcnt 0x0
; GFX12-TRUE16-NEXT: global_inv scope:SCOPE_DEV
-; GFX12-TRUE16-NEXT: v_cmp_eq_u32_e32 vcc_lo, v4, v2
-; GFX12-TRUE16-NEXT: v_mov_b32_e32 v2, v4
+; GFX12-TRUE16-NEXT: v_cmp_eq_u32_e32 vcc_lo, v1, v5
; GFX12-TRUE16-NEXT: s_or_b32 s5, vcc_lo, s5
; GFX12-TRUE16-NEXT: s_wait_alu 0xfffe
; GFX12-TRUE16-NEXT: s_and_not1_b32 exec_lo, exec_lo, s5
@@ -3922,28 +3922,28 @@ define void @buffer_fat_ptr_agent_atomic_fadd_noret_f16__offset__amdgpu_no_fine_
; GFX12-FAKE16-NEXT: s_lshl_b32 s4, s4, 3
; GFX12-FAKE16-NEXT: s_wait_alu 0xfffe
; GFX12-FAKE16-NEXT: s_lshl_b32 s5, 0xffff, s4
-; GFX12-FAKE16-NEXT: buffer_load_b32 v2, v3, s[0:3], null offen
+; GFX12-FAKE16-NEXT: buffer_load_b32 v1, v3, s[0:3], null offen
; GFX12-FAKE16-NEXT: s_not_b32 s6, s5
; GFX12-FAKE16-NEXT: s_mov_b32 s5, 0
; GFX12-FAKE16-NEXT: .LBB14_1: ; %atomicrmw.start
; GFX12-FAKE16-NEXT: ; =>This Inner Loop Header: Depth=1
; GFX12-FAKE16-NEXT: s_wait_loadcnt 0x0
-; GFX12-FAKE16-NEXT: v_lshrrev_b32_e32 v1, s4, v2
+; GFX12-FAKE16-NEXT: v_mov_b32_e32 v5, v1
; GFX12-FAKE16-NEXT: s_wait_storecnt 0x0
; GFX12-FAKE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
+; GFX12-FAKE16-NEXT: v_lshrrev_b32_e32 v1, s4, v5
; GFX12-FAKE16-NEXT: v_add_f16_e32 v1, v1, v0
+; GFX12-FAKE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
; GFX12-FAKE16-NEXT: v_and_b32_e32 v1, 0xffff, v1
-; GFX12-FAKE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(SKIP_1) | instid1(VALU_DEP_1)
; GFX12-FAKE16-NEXT: v_lshlrev_b32_e32 v1, s4, v1
; GFX12-FAKE16-NEXT: s_wait_alu 0xfffe
-; GFX12-FAKE16-NEXT: v_and_or_b32 v1, v2, s6, v1
-; GFX12-FAKE16-NEXT: s_delay_alu instid0(VALU_DEP_1)
-; GFX12-FAKE16-NEXT: v_dual_mov_b32 v5, v2 :: v_dual_mov_b32 v4, v1
-; GFX12-FAKE16-NEXT: buffer_atomic_cmpswap_b32 v[4:5], v3, s[0:3], null offen th:TH_ATOMIC_RETURN
+; GFX12-FAKE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
+; GFX12-FAKE16-NEXT: v_and_or_b32 v4, v5, s6, v1
+; GFX12-FAKE16-NEXT: v_dual_mov_b32 v1, v4 :: v_dual_mov_b32 v2, v5
+; GFX12-FAKE16-NEXT: buffer_atomic_cmpswap_b32 v[1:2], v3, s[0:3], null offen th:TH_ATOMIC_RETURN
; GFX12-FAKE16-NEXT: s_wait_loadcnt 0x0
; GFX12-FAKE16-NEXT: global_inv scope:SCOPE_DEV
-; GFX12-FAKE16-NEXT: v_cmp_eq_u32_e32 vcc_lo, v4, v2
-; GFX12-FAKE16-NEXT: v_mov_b32_e32 v2, v4
+; GFX12-FAKE16-NEXT: v_cmp_eq_u32_e32 vcc_lo, v1, v5
; GFX12-FAKE16-NEXT: s_or_b32 s5, vcc_lo, s5
; GFX12-FAKE16-NEXT: s_wait_alu 0xfffe
; GFX12-FAKE16-NEXT: s_and_not1_b32 exec_lo, exec_lo, s5
@@ -3959,7 +3959,7 @@ define void @buffer_fat_ptr_agent_atomic_fadd_noret_f16__offset__amdgpu_no_fine_
; GFX942-NEXT: s_addk_i32 s16, 0x200
; GFX942-NEXT: s_and_b32 s4, s16, -4
; GFX942-NEXT: v_mov_b32_e32 v1, s4
-; GFX942-NEXT: buffer_load_dword v3, v1, s[0:3], 0 offen
+; GFX942-NEXT: buffer_load_dword v2, v1, s[0:3], 0 offen
; GFX942-NEXT: s_and_b32 s4, s16, 3
; GFX942-NEXT: s_lshl_b32 s6, s4, 3
; GFX942-NEXT: s_lshl_b32 s4, 0xffff, s6
@@ -3968,18 +3968,18 @@ define void @buffer_fat_ptr_agent_atomic_fadd_noret_f16__offset__amdgpu_no_fine_
; GFX942-NEXT: .LBB14_1: ; %atomicrmw.start
; GFX942-NEXT: ; =>This Inner Loop Header: Depth=1
; GFX942-NEXT: s_waitcnt vmcnt(0)
-; GFX942-NEXT: v_lshrrev_b32_e32 v2, s6, v3
+; GFX942-NEXT: v_mov_b32_e32 v5, v2
+; GFX942-NEXT: v_lshrrev_b32_e32 v2, s6, v5
; GFX942-NEXT: v_add_f16_e32 v2, v2, v0
; GFX942-NEXT: v_lshlrev_b32_e32 v2, s6, v2
-; GFX942-NEXT: v_and_or_b32 v2, v3, s7, v2
-; GFX942-NEXT: v_mov_b64_e32 v[4:5], v[2:3]
+; GFX942-NEXT: v_and_or_b32 v4, v5, s7, v2
+; GFX942-NEXT: v_mov_b64_e32 v[2:3], v[4:5]
; GFX942-NEXT: buffer_wbl2 sc1
-; GFX942-NEXT: buffer_atomic_cmpswap v[4:5], v1, s[0:3], 0 offen sc0
+; GFX942-NEXT: buffer_atomic_cmpswap v[2:3], v1, s[0:3], 0 offen sc0
; GFX942-NEXT: s_waitcnt vmcnt(0)
; GFX942-NEXT: buffer_inv sc1
-; GFX942-NEXT: v_cmp_eq_u32_e32 vcc, v4, v3
+; GFX942-NEXT: v_cmp_eq_u32_e32 vcc, v2, v5
; GFX942-NEXT: s_or_b64 s[4:5], vcc, s[4:5]
-; GFX942-NEXT: v_mov_b32_e32 v3, v4
; GFX942-NEXT: s_andn2_b64 exec, exec, s[4:5]
; GFX942-NEXT: s_cbranch_execnz .LBB14_1
; GFX942-NEXT: ; %bb.2: ; %atomicrmw.end
@@ -3997,28 +3997,28 @@ define void @buffer_fat_ptr_agent_atomic_fadd_noret_f16__offset__amdgpu_no_fine_
; GFX11-TRUE16-NEXT: s_delay_alu instid0(SALU_CYCLE_1) | instskip(NEXT) | instid1(SALU_CYCLE_1)
; GFX11-TRUE16-NEXT: s_lshl_b32 s4, s4, 3
; GFX11-TRUE16-NEXT: s_lshl_b32 s5, 0xffff, s4
-; GFX11-TRUE16-NEXT: buffer_load_b32 v2, v3, s[0:3], 0 offen
+; GFX11-TRUE16-NEXT: buffer_load_b32 v1, v3, s[0:3], 0 offen
; GFX11-TRUE16-NEXT: s_not_b32 s6, s5
; GFX11-TRUE16-NEXT: s_mov_b32 s5, 0
; GFX11-TRUE16-NEXT: .LBB14_1: ; %atomicrmw.start
; GFX11-TRUE16-NEXT: ; =>This Inner Loop Header: Depth=1
; GFX11-TRUE16-NEXT: s_waitcnt vmcnt(0)
-; GFX11-TRUE16-NEXT: v_lshrrev_b32_e32 v1, s4, v2
+; GFX11-TRUE16-NEXT: v_mov_b32_e32 v5, v1
; GFX11-TRUE16-NEXT: s_waitcnt_vscnt null, 0x0
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
+; GFX11-TRUE16-NEXT: v_lshrrev_b32_e32 v1, s4, v5
; GFX11-TRUE16-NEXT: v_add_f16_e32 v1.l, v1.l, v0.l
-; GFX11-TRUE16-NEXT: v_and_b32_e32 v1, 0xffff, v1
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
+; GFX11-TRUE16-NEXT: v_and_b32_e32 v1, 0xffff, v1
; GFX11-TRUE16-NEXT: v_lshlrev_b32_e32 v1, s4, v1
-; GFX11-TRUE16-NEXT: v_and_or_b32 v1, v2, s6, v1
-; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1)
-; GFX11-TRUE16-NEXT: v_dual_mov_b32 v5, v2 :: v_dual_mov_b32 v4, v1
-; GFX11-TRUE16-NEXT: buffer_atomic_cmpswap_b32 v[4:5], v3, s[0:3], 0 offen glc
+; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
+; GFX11-TRUE16-NEXT: v_and_or_b32 v4, v5, s6, v1
+; GFX11-TRUE16-NEXT: v_dual_mov_b32 v1, v4 :: v_dual_mov_b32 v2, v5
+; GFX11-TRUE16-NEXT: buffer_atomic_cmpswap_b32 v[1:2], v3, s[0:3], 0 offen glc
; GFX11-TRUE16-NEXT: s_waitcnt vmcnt(0)
; GFX11-TRUE16-NEXT: buffer_gl1_inv
; GFX11-TRUE16-NEXT: buffer_gl0_inv
-; GFX11-TRUE16-NEXT: v_cmp_eq_u32_e32 vcc_lo, v4, v2
-; GFX11-TRUE16-NEXT: v_mov_b32_e32 v2, v4
+; GFX11-TRUE16-NEXT: v_cmp_eq_u32_e32 vcc_lo, v1, v5
; GFX11-TRUE16-NEXT: s_or_b32 s5, vcc_lo, s5
; GFX11-TRUE16-NEXT: s_delay_alu instid0(SALU_CYCLE_1)
; GFX11-TRUE16-NEXT: s_and_not1_b32 exec_lo, exec_lo, s5
@@ -4038,28 +4038,28 @@ define void @buffer_fat_ptr_agent_atomic_fadd_noret_f16__offset__amdgpu_no_fine_
; GFX11-FAKE16-NEXT: s_delay_alu instid0(SALU_CYCLE_1) | instskip(NEXT) | instid1(SALU_CYCLE_1)
; GFX11-FAKE16-NEXT: s_lshl_b32 s4, s4, 3
; GFX11-FAKE16-NEXT: s_lshl_b32 s5, 0xffff, s4
-; GFX11-FAKE16-NEXT: buffer_load_b32 v2, v3, s[0:3], 0 offen
+; GFX11-FAKE16-NEXT: buffer_load_b32 v1, v3, s[0:3], 0 offen
; GFX11-FAKE16-NEXT: s_not_b32 s6, s5
; GFX11-FAKE16-NEXT: s_mov_b32 s5, 0
; GFX11-FAKE16-NEXT: .LBB14_1: ; %atomicrmw.start
; GFX11-FAKE16-NEXT: ; =>This Inner Loop Header: Depth=1
; GFX11-FAKE16-NEXT: s_waitcnt vmcnt(0)
-; GFX11-FAKE16-NEXT: v_lshrrev_b32_e32 v1, s4, v2
+; GFX11-FAKE16-NEXT: v_mov_b32_e32 v5, v1
; GFX11-FAKE16-NEXT: s_waitcnt_vscnt null, 0x0
; GFX11-FAKE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
+; GFX11-FAKE16-NEXT: v_lshrrev_b32_e32 v1, s4, v5
; GFX11-FAKE16-NEXT: v_add_f16_e32 v1, v1, v0
-; GFX11-FAKE16-NEXT: v_and_b32_e32 v1, 0xffff, v1
; GFX11-FAKE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
+; GFX11-FAKE16-NEXT: v_and_b32_e32 v1, 0xffff, v1
; GFX11-FAKE16-NEXT: v_lshlrev_b32_e32 v1, s4, v1
-; GFX11-FAKE16-NEXT: v_and_or_b32 v1, v2, s6, v1
-; GFX11-FAKE16-NEXT: s_delay_alu instid0(VALU_DEP_1)
-; GFX11-FAKE16-NEXT: v_dual_mov_b32 v5, v2 :: v_dual_mov_b32 v4, v1
-; GFX11-FAKE16-NEXT: buffer_atomic_cmpswap_b32 v[4:5], v3, s[0:3], 0 offen glc
+; GFX11-FAKE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
+; GFX11-FAKE16-NEXT: v_and_or_b32 v4, v5, s6, v1
+; GFX11-FAKE16-NEXT: v_dual_mov_b32 v1, v4 :: v_dual_mov_b32 v2, v5
+; GFX11-FAKE16-NEXT: buffer_atomic_cmpswap_b32 v[1:2], v3, s[0:3], 0 offen glc
; GFX11-FAKE16-NEXT: s_waitcnt vmcnt(0)
; GFX11-FAKE16-NEXT: buffer_gl1_inv
; GFX11-FAKE16-NEXT: buffer_gl0_inv
-; GFX11-FAKE16-NEXT: v_cmp_eq_u32_e32 vcc_lo, v4, v2
-; GFX11-FAKE16-NEXT: v_mov_b32_e32 v2, v4
+; GFX11-FAKE16-NEXT: v_cmp_eq_u32_e32 vcc_lo, v1, v5
; GFX11-FAKE16-NEXT: s_or_b32 s5, vcc_lo, s5
; GFX11-FAKE16-NEXT: s_delay_alu instid0(SALU_CYCLE_1)
; GFX11-FAKE16-NEXT: s_and_not1_b32 exec_lo, exec_lo, s5
@@ -4077,25 +4077,25 @@ define void @buffer_fat_ptr_agent_atomic_fadd_noret_f16__offset__amdgpu_no_fine_
; GFX10-NEXT: s_and_b32 s4, s20, 3
; GFX10-NEXT: s_lshl_b32 s4, s4, 3
; GFX10-NEXT: s_lshl_b32 s5, 0xffff, s4
-; GFX10-NEXT: buffer_load_dword v2, v3, s[16:19], 0 offen
+; GFX10-NEXT: buffer_load_dword v1, v3, s[16:19], 0 offen
; GFX10-NEXT: s_not_b32 s6, s5
; GFX10-NEXT: s_mov_b32 s5, 0
; GFX10-NEXT: .LBB14_1: ; %atomicrmw.start
; GFX10-NEXT: ; =>This Inner Loop Header: Depth=1
; GFX10-NEXT: s_waitcnt vmcnt(0)
-; GFX10-NEXT: v_lshrrev_b32_e32 v1, s4, v2
+; GFX10-NEXT: v_mov_b32_e32 v5, v1
; GFX10-NEXT: s_waitcnt_vscnt null, 0x0
+; GFX10-NEXT: v_lshrrev_b32_e32 v1, s4, v5
; GFX10-NEXT: v_add_f16_e32 v1, v1, v0
; GFX10-NEXT: v_lshlrev_b32_sdwa v1, s4, v1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:WORD_0
-; GFX10-NEXT: v_and_or_b32 v1, v2, s6, v1
-; GFX10-NEXT: v_mov_b32_e32 v5, v2
-; GFX10-NEXT: v_mov_b32_e32 v4, v1
-; GFX10-NEXT: buffer_atomic_cmpswap v[4:5], v3, s[16:19], 0 offen glc
+; GFX10-NEXT: v_and_or_b32 v4, v5, s6, v1
+; GFX10-NEXT: v_mov_b32_e32 v1, v4
+; GFX10-NEXT: v_mov_b32_e32 v2, v5
+; GFX10-NEXT: buffer_atomic_cmpswap v[1:2], v3, s[16:19], 0 offen glc
; GFX10-NEXT: s_waitcnt vmcnt(0)
; GFX10-NEXT: buffer_gl1_inv
; GFX10-NEXT: buffer_gl0_inv
-; GFX10-NEXT: v_cmp_eq_u32_e32 vcc_lo, v4, v2
-; GFX10-NEXT: v_mov_b32_e32 v2, v4
+; GFX10-NEXT: v_cmp_eq_u32_e32 vcc_lo, v1, v5
; GFX10-NEXT: s_or_b32 s5, vcc_lo, s5
; GFX10-NEXT: s_andn2_b32 exec_lo, exec_lo, s5
; GFX10-NEXT: s_cbranch_execnz .LBB14_1
@@ -4109,7 +4109,7 @@ define void @buffer_fat_ptr_agent_atomic_fadd_noret_f16__offset__amdgpu_no_fine_
; GFX90A-NEXT: s_addk_i32 s20, 0x200
; GFX90A-NEXT: s_and_b32 s4, s20, -4
; GFX90A-NEXT: v_mov_b32_e32 v1, s4
-; GFX90A-NEXT: buffer_load_dword v3, v1, s[16:19], 0 offen
+; GFX90A-NEXT: buffer_load_dword v2, v1, s[16:19], 0 offen
; GFX90A-NEXT: s_and_b32 s4, s20, 3
; GFX90A-NEXT: s_lshl_b32 s6, s4, 3
; GFX90A-NEXT: s_lshl_b32 s4, 0xffff, s6
@@ -4118,17 +4118,17 @@ define void @buffer_fat_ptr_agent_atomic_fadd_noret_f16__offset__amdgpu_no_fine_
; GFX90A-NEXT: .LBB14_1: ; %atomicrmw.start
; GFX90A-NEXT: ; =>This Inner Loop Header: Depth=1
; GFX90A-NEXT: s_waitcnt vmcnt(0)
-; GFX90A-NEXT: v_lshrrev_b32_e32 v2, s6, v3
+; GFX90A-NEXT: v_mov_b32_e32 v5, v2
+; GFX90A-NEXT: v_lshrrev_b32_e32 v2, s6, v5
; GFX90A-NEXT: v_add_f16_e32 v2, v2, v0
; GFX90A-NEXT: v_lshlrev_b32_e32 v2, s6, v2
-; GFX90A-NEXT: v_and_or_b32 v2, v3, s7, v2
-; GFX90A-NEXT: v_pk_mov_b32 v[4:5], v[2:3], v[2:3] op_sel:[0,1]
-; GFX90A-NEXT: buffer_atomic_cmpswap v[4:5], v1, s[16:19], 0 offen glc
+; GFX90A-NEXT: v_and_or_b32 v4, v5, s7, v2
+; GFX90A-NEXT: v_pk_mov_b32 v[2:3], v[4:5], v[4:5] op_sel:[0,1]
+; GFX90A-NEXT: buffer_atomic_cmpswap v[2:3], v1, s[16:19], 0 offen glc
; GFX90A-NEXT: s_waitcnt vmcnt(0)
; GFX90A-NEXT: buffer_wbinvl1
-; GFX90A-NEXT: v_cmp_eq_u32_e32 vcc, v4, v3
+; GFX90A-NEXT: v_cmp_eq_u32_e32 vcc, v2, v5
; GFX90A-NEXT: s_or_b64 s[4:5], vcc, s[4:5]
-; GFX90A-NEXT: v_mov_b32_e32 v3, v4
; GFX90A-NEXT: s_andn2_b64 exec, exec, s[4:5]
; GFX90A-NEXT: s_cbranch_execnz .LBB14_1
; GFX90A-NEXT: ; %bb.2: ; %atomicrmw.end
@@ -4141,7 +4141,7 @@ define void @buffer_fat_ptr_agent_atomic_fadd_noret_f16__offset__amdgpu_no_fine_
; GFX908-NEXT: s_addk_i32 s20, 0x200
; GFX908-NEXT: s_and_b32 s4, s20, -4
; GFX908-NEXT: v_mov_b32_e32 v3, s4
-; GFX908-NEXT: buffer_load_dword v2, v3, s[16:19], 0 offen
+; GFX908-NEXT: buffer_load_dword v1, v3, s[16:19], 0 offen
; GFX908-NEXT: s_and_b32 s4, s20, 3
; GFX908-NEXT: s_lshl_b32 s6, s4, 3
; GFX908-NEXT: s_lshl_b32 s4, 0xffff, s6
@@ -4150,18 +4150,18 @@ define void @buffer_fat_ptr_agent_atomic_fadd_noret_f16__offset__amdgpu_no_fine_
; GFX908-NEXT: .LBB14_1: ; %atomicrmw.start
; GFX908-NEXT: ; =>This Inner Loop Header: Depth=1
; GFX908-NEXT: s_waitcnt vmcnt(0)
-; GFX908-NEXT: v_lshrrev_b32_e32 v1, s6, v2
+; GFX908-NEXT: v_mov_b32_e32 v5, v1
+; GFX908-NEXT: v_lshrrev_b32_e32 v1, s6, v5
; GFX908-NEXT: v_add_f16_e32 v1, v1, v0
; GFX908-NEXT: v_lshlrev_b32_e32 v1, s6, v1
-; GFX908-NEXT: v_and_or_b32 v1, v2, s7, v1
-; GFX908-NEXT: v_mov_b32_e32 v5, v2
-; GFX908-NEXT: v_mov_b32_e32 v4, v1
-; GFX908-NEXT: buffer_atomic_cmpswap v[4:5], v3, s[16:19], 0 offen glc
+; GFX908-NEXT: v_and_or_b32 v4, v5, s7, v1
+; GFX908-NEXT: v_mov_b32_e32 v1, v4
+; GFX908-NEXT: v_mov_b32_e32 v2, v5
+; GFX908-NEXT: buffer_atomic_cmpswap v[1:2], v3, s[16:19], 0 offen glc
; GFX908-NEXT: s_waitcnt vmcnt(0)
; GFX908-NEXT: buffer_wbinvl1
-; GFX908-NEXT: v_cmp_eq_u32_e32 vcc, v4, v2
+; GFX908-NEXT: v_cmp_eq_u32_e32 vcc, v1, v5
; GFX908-NEXT: s_or_b64 s[4:5], vcc, s[4:5]
-; GFX908-NEXT: v_mov_b32_e32 v2, v4
; GFX908-NEXT: s_andn2_b64 exec, exec, s[4:5]
; GFX908-NEXT: s_cbranch_execnz .LBB14_1
; GFX908-NEXT: ; %bb.2: ; %atomicrmw.end
@@ -4174,7 +4174,7 @@ define void @buffer_fat_ptr_agent_atomic_fadd_noret_f16__offset__amdgpu_no_fine_
; GFX8-NEXT: s_addk_i32 s20, 0x200
; GFX8-NEXT: s_and_b32 s4, s20, -4
; GFX8-NEXT: v_mov_b32_e32 v3, s4
-; GFX8-NEXT: buffer_load_dword v2, v3, s[16:19], 0 offen
+; GFX8-NEXT: buffer_load_dword v1, v3, s[16:19], 0 offen
; GFX8-NEXT: s_and_b32 s4, s20, 3
; GFX8-NEXT: s_lshl_b32 s6, s4, 3
; GFX8-NEXT: s_lshl_b32 s4, 0xffff, s6
@@ -4183,19 +4183,19 @@ define void @buffer_fat_ptr_agent_atomic_fadd_noret_f16__offset__amdgpu_no_fine_
; GFX8-NEXT: .LBB14_1: ; %atomicrmw.start
; GFX8-NEXT: ; =>This Inner Loop Header: Depth=1
; GFX8-NEXT: s_waitcnt vmcnt(0)
-; GFX8-NEXT: v_lshrrev_b32_e32 v1, s6, v2
+; GFX8-NEXT: v_mov_b32_e32 v5, v1
+; GFX8-NEXT: v_lshrrev_b32_e32 v1, s6, v5
; GFX8-NEXT: v_add_f16_e32 v1, v1, v0
-; GFX8-NEXT: v_and_b32_e32 v4, s7, v2
+; GFX8-NEXT: v_and_b32_e32 v2, s7, v5
; GFX8-NEXT: v_lshlrev_b32_e32 v1, s6, v1
-; GFX8-NEXT: v_or_b32_e32 v1, v4, v1
-; GFX8-NEXT: v_mov_b32_e32 v5, v2
-; GFX8-NEXT: v_mov_b32_e32 v4, v1
-; GFX8-NEXT: buffer_atomic_cmpswap v[4:5], v3, s[16:19], 0 offen glc
+; GFX8-NEXT: v_or_b32_e32 v4, v2, v1
+; GFX8-NEXT: v_mov_b32_e32 v1, v4
+; GFX8-NEXT: v_mov_b32_e32 v2, v5
+; GFX8-NEXT: buffer_atomic_cmpswap v[1:2], v3, s[16:19], 0 offen glc
; GFX8-NEXT: s_waitcnt vmcnt(0)
; GFX8-NEXT: buffer_wbinvl1
-; GFX8-NEXT: v_cmp_eq_u32_e32 vcc, v4, v2
+; GFX8-NEXT: v_cmp_eq_u32_e32 vcc, v1, v5
; GFX8-NEXT: s_or_b64 s[4:5], vcc, s[4:5]
-; GFX8-NEXT: v_mov_b32_e32 v2, v4
; GFX8-NEXT: s_andn2_b64 exec, exec, s[4:5]
; GFX8-NEXT: s_cbranch_execnz .LBB14_1
; GFX8-NEXT: ; %bb.2: ; %atomicrmw.end
@@ -4207,33 +4207,33 @@ define void @buffer_fat_ptr_agent_atomic_fadd_noret_f16__offset__amdgpu_no_fine_
; GFX7-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX7-NEXT: s_addk_i32 s20, 0x200
; GFX7-NEXT: s_and_b32 s4, s20, -4
-; GFX7-NEXT: v_mov_b32_e32 v2, s4
-; GFX7-NEXT: buffer_load_dword v1, v2, s[16:19], 0 offen
+; GFX7-NEXT: v_mov_b32_e32 v3, s4
+; GFX7-NEXT: buffer_load_dword v1, v3, s[16:19], 0 offen
; GFX7-NEXT: v_cvt_f16_f32_e32 v0, v0
; GFX7-NEXT: s_and_b32 s4, s20, 3
; GFX7-NEXT: s_lshl_b32 s6, s4, 3
; GFX7-NEXT: s_lshl_b32 s4, 0xffff, s6
-; GFX7-NEXT: v_cvt_f32_f16_e32 v3, v0
+; GFX7-NEXT: v_cvt_f32_f16_e32 v0, v0
; GFX7-NEXT: s_not_b32 s7, s4
; GFX7-NEXT: s_mov_b64 s[4:5], 0
; GFX7-NEXT: .LBB14_1: ; %atomicrmw.start
; GFX7-NEXT: ; =>This Inner Loop Header: Depth=1
; GFX7-NEXT: s_waitcnt vmcnt(0)
-; GFX7-NEXT: v_lshrrev_b32_e32 v0, s6, v1
-; GFX7-NEXT: v_cvt_f32_f16_e32 v0, v0
-; GFX7-NEXT: v_and_b32_e32 v4, s7, v1
-; GFX7-NEXT: v_add_f32_e32 v0, v0, v3
-; GFX7-NEXT: v_cvt_f16_f32_e32 v0, v0
-; GFX7-NEXT: v_lshlrev_b32_e32 v0, s6, v0
-; GFX7-NEXT: v_or_b32_e32 v0, v4, v0
; GFX7-NEXT: v_mov_b32_e32 v5, v1
-; GFX7-NEXT: v_mov_b32_e32 v4, v0
-; GFX7-NEXT: buffer_atomic_cmpswap v[4:5], v2, s[16:19], 0 offen glc
+; GFX7-NEXT: v_lshrrev_b32_e32 v1, s6, v5
+; GFX7-NEXT: v_cvt_f32_f16_e32 v1, v1
+; GFX7-NEXT: v_and_b32_e32 v2, s7, v5
+; GFX7-NEXT: v_add_f32_e32 v1, v1, v0
+; GFX7-NEXT: v_cvt_f16_f32_e32 v1, v1
+; GFX7-NEXT: v_lshlrev_b32_e32 v1, s6, v1
+; GFX7-NEXT: v_or_b32_e32 v4, v2, v1
+; GFX7-NEXT: v_mov_b32_e32 v1, v4
+; GFX7-NEXT: v_mov_b32_e32 v2, v5
+; GFX7-NEXT: buffer_atomic_cmpswap v[1:2], v3, s[16:19], 0 offen glc
; GFX7-NEXT: s_waitcnt vmcnt(0)
; GFX7-NEXT: buffer_wbinvl1
-; GFX7-NEXT: v_cmp_eq_u32_e32 vcc, v4, v1
+; GFX7-NEXT: v_cmp_eq_u32_e32 vcc, v1, v5
; GFX7-NEXT: s_or_b64 s[4:5], vcc, s[4:5]
-; GFX7-NEXT: v_mov_b32_e32 v1, v4
; GFX7-NEXT: s_andn2_b64 exec, exec, s[4:5]
; GFX7-NEXT: s_cbranch_execnz .LBB14_1
; GFX7-NEXT: ; %bb.2: ; %atomicrmw.end
@@ -4245,34 +4245,34 @@ define void @buffer_fat_ptr_agent_atomic_fadd_noret_f16__offset__amdgpu_no_fine_
; GFX6-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX6-NEXT: s_addk_i32 s20, 0x200
; GFX6-NEXT: s_and_b32 s4, s20, -4
-; GFX6-NEXT: v_mov_b32_e32 v2, s4
-; GFX6-NEXT: buffer_load_dword v1, v2, s[16:19], 0 offen
+; GFX6-NEXT: v_mov_b32_e32 v3, s4
+; GFX6-NEXT: buffer_load_dword v1, v3, s[16:19], 0 offen
; GFX6-NEXT: v_cvt_f16_f32_e32 v0, v0
; GFX6-NEXT: s_and_b32 s4, s20, 3
; GFX6-NEXT: s_lshl_b32 s6, s4, 3
; GFX6-NEXT: s_lshl_b32 s4, 0xffff, s6
-; GFX6-NEXT: v_cvt_f32_f16_e32 v3, v0
+; GFX6-NEXT: v_cvt_f32_f16_e32 v0, v0
; GFX6-NEXT: s_not_b32 s7, s4
; GFX6-NEXT: s_mov_b64 s[4:5], 0
; GFX6-NEXT: .LBB14_1: ; %atomicrmw.start
; GFX6-NEXT: ; =>This Inner Loop Header: Depth=1
; GFX6-NEXT: s_waitcnt vmcnt(0)
-; GFX6-NEXT: v_lshrrev_b32_e32 v0, s6, v1
-; GFX6-NEXT: v_cvt_f32_f16_e32 v0, v0
-; GFX6-NEXT: s_waitcnt expcnt(0)
-; GFX6-NEXT: v_and_b32_e32 v4, s7, v1
-; GFX6-NEXT: v_add_f32_e32 v0, v0, v3
-; GFX6-NEXT: v_cvt_f16_f32_e32 v0, v0
-; GFX6-NEXT: v_lshlrev_b32_e32 v0, s6, v0
-; GFX6-NEXT: v_or_b32_e32 v0, v4, v0
; GFX6-NEXT: v_mov_b32_e32 v5, v1
-; GFX6-NEXT: v_mov_b32_e32 v4, v0
-; GFX6-NEXT: buffer_atomic_cmpswap v[4:5], v2, s[16:19], 0 offen glc
+; GFX6-NEXT: s_waitcnt expcnt(0)
+; GFX6-NEXT: v_lshrrev_b32_e32 v1, s6, v5
+; GFX6-NEXT: v_cvt_f32_f16_e32 v1, v1
+; GFX6-NEXT: v_and_b32_e32 v2, s7, v5
+; GFX6-NEXT: v_add_f32_e32 v1, v1, v0
+; GFX6-NEXT: v_cvt_f16_f32_e32 v1, v1
+; GFX6-NEXT: v_lshlrev_b32_e32 v1, s6, v1
+; GFX6-NEXT: v_or_b32_e32 v4, v2, v1
+; GFX6-NEXT: v_mov_b32_e32 v1, v4
+; GFX6-NEXT: v_mov_b32_e32 v2, v5
+; GFX6-NEXT: buffer_atomic_cmpswap v[1:2], v3, s[16:19], 0 offen glc
; GFX6-NEXT: s_waitcnt vmcnt(0)
; GFX6-NEXT: buffer_wbinvl1
-; GFX6-NEXT: v_cmp_eq_u32_e32 vcc, v4, v1
+; GFX6-NEXT: v_cmp_eq_u32_e32 vcc, v1, v5
; GFX6-NEXT: s_or_b64 s[4:5], vcc, s[4:5]
-; GFX6-NEXT: v_mov_b32_e32 v1, v4
; GFX6-NEXT: s_andn2_b64 exec, exec, s[4:5]
; GFX6-NEXT: s_cbranch_execnz .LBB14_1
; GFX6-NEXT: ; %bb.2: ; %atomicrmw.end
@@ -4296,11 +4296,11 @@ define half @buffer_fat_ptr_agent_atomic_fadd_ret_f16__offset__waterfall__amdgpu
; GFX12-TRUE16-NEXT: s_mov_b32 s1, exec_lo
; GFX12-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(SKIP_1) | instid1(VALU_DEP_2)
; GFX12-TRUE16-NEXT: v_and_b32_e32 v4, 3, v6
-; GFX12-TRUE16-NEXT: v_and_b32_e32 v10, -4, v6
+; GFX12-TRUE16-NEXT: v_and_b32_e32 v9, -4, v6
; GFX12-TRUE16-NEXT: v_lshlrev_b32_e32 v4, 3, v4
; GFX12-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
; GFX12-TRUE16-NEXT: v_lshlrev_b32_e64 v7, v4, 0xffff
-; GFX12-TRUE16-NEXT: v_not_b32_e32 v11, v7
+; GFX12-TRUE16-NEXT: v_not_b32_e32 v10, v7
; GFX12-TRUE16-NEXT: .LBB15_1: ; =>This Inner Loop Header: Depth=1
; GFX12-TRUE16-NEXT: v_readfirstlane_b32 s4, v0
; GFX12-TRUE16-NEXT: v_readfirstlane_b32 s5, v1
@@ -4313,7 +4313,7 @@ define half @buffer_fat_ptr_agent_atomic_fadd_ret_f16__offset__waterfall__amdgpu
; GFX12-TRUE16-NEXT: s_wait_alu 0xfffe
; GFX12-TRUE16-NEXT: s_and_saveexec_b32 s0, s0
; GFX12-TRUE16-NEXT: s_wait_loadcnt 0x0
-; GFX12-TRUE16-NEXT: buffer_load_b32 v7, v10, s[4:7], null offen
+; GFX12-TRUE16-NEXT: buffer_load_b32 v6, v9, s[4:7], null offen
; GFX12-TRUE16-NEXT: s_xor_b32 exec_lo, exec_lo, s0
; GFX12-TRUE16-NEXT: s_cbranch_execnz .LBB15_1
; GFX12-TRUE16-NEXT: ; %bb.2:
@@ -4323,17 +4323,19 @@ define half @buffer_fat_ptr_agent_atomic_fadd_ret_f16__offset__waterfall__amdgpu
; GFX12-TRUE16-NEXT: ; =>This Loop Header: Depth=1
; GFX12-TRUE16-NEXT: ; Child Loop BB15_4 Depth 2
; GFX12-TRUE16-NEXT: s_wait_loadcnt 0x0
-; GFX12-TRUE16-NEXT: v_lshrrev_b32_e32 v6, v4, v7
+; GFX12-TRUE16-NEXT: v_mov_b32_e32 v8, v6
; GFX12-TRUE16-NEXT: s_mov_b32 s2, exec_lo
; GFX12-TRUE16-NEXT: s_wait_storecnt 0x0
; GFX12-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
+; GFX12-TRUE16-NEXT: v_lshrrev_b32_e32 v6, v4, v8
; GFX12-TRUE16-NEXT: v_add_f16_e32 v6.l, v6.l, v5.l
-; GFX12-TRUE16-NEXT: v_and_b32_e32 v6, 0xffff, v6
; GFX12-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
+; GFX12-TRUE16-NEXT: v_and_b32_e32 v6, 0xffff, v6
; GFX12-TRUE16-NEXT: v_lshlrev_b32_e32 v6, v4, v6
-; GFX12-TRUE16-NEXT: v_and_or_b32 v6, v7, v11, v6
-; GFX12-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1)
-; GFX12-TRUE16-NEXT: v_dual_mov_b32 v9, v7 :: v_dual_mov_b32 v8, v6
+; GFX12-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
+; GFX12-TRUE16-NEXT: v_and_or_b32 v7, v8, v10, v6
+; GFX12-TRUE16-NEXT: v_mov_b32_e32 v6, v7
+; GFX12-TRUE16-NEXT: v_mov_b32_e32 v7, v8
; GFX12-TRUE16-NEXT: .LBB15_4: ; Parent Loop BB15_3 Depth=1
; GFX12-TRUE16-NEXT: ; => This Inner Loop Header: Depth=2
; GFX12-TRUE16-NEXT: v_readfirstlane_b32 s4, v0
@@ -4347,14 +4349,13 @@ define half @buffer_fat_ptr_agent_atomic_fadd_ret_f16__offset__waterfall__amdgpu
; GFX12-TRUE16-NEXT: s_wait_alu 0xfffe
; GFX12-TRUE16-NEXT: s_and_saveexec_b32 s0, s0
; GFX12-TRUE16-NEXT: s_wait_loadcnt 0x0
-; GFX12-TRUE16-NEXT: buffer_atomic_cmpswap_b32 v[8:9], v10, s[4:7], null offen th:TH_ATOMIC_RETURN
+; GFX12-TRUE16-NEXT: buffer_atomic_cmpswap_b32 v[6:7], v9, s[4:7], null offen th:TH_ATOMIC_RETURN
; GFX12-TRUE16-NEXT: s_xor_b32 exec_lo, exec_lo, s0
; GFX12-TRUE16-NEXT: s_cbranch_execnz .LBB15_4
; GFX12-TRUE16-NEXT: ; %bb.5: ; in Loop: Header=BB15_3 Depth=1
; GFX12-TRUE16-NEXT: s_mov_b32 exec_lo, s2
; GFX12-TRUE16-NEXT: s_wait_loadcnt 0x0
-; GFX12-TRUE16-NEXT: v_cmp_eq_u32_e32 vcc_lo, v8, v7
-; GFX12-TRUE16-NEXT: v_mov_b32_e32 v7, v8
+; GFX12-TRUE16-NEXT: v_cmp_eq_u32_e32 vcc_lo, v6, v8
; GFX12-TRUE16-NEXT: global_inv scope:SCOPE_DEV
; GFX12-TRUE16-NEXT: s_or_b32 s1, vcc_lo, s1
; GFX12-TRUE16-NEXT: s_wait_alu 0xfffe
@@ -4362,7 +4363,7 @@ define half @buffer_fat_ptr_agent_atomic_fadd_ret_f16__offset__waterfall__amdgpu
; GFX12-TRUE16-NEXT: s_cbranch_execnz .LBB15_3
; GFX12-TRUE16-NEXT: ; %bb.6: ; %atomicrmw.end
; GFX12-TRUE16-NEXT: s_or_b32 exec_lo, exec_lo, s1
-; GFX12-TRUE16-NEXT: v_lshrrev_b32_e32 v0, v4, v8
+; GFX12-TRUE16-NEXT: v_lshrrev_b32_e32 v0, v4, v6
; GFX12-TRUE16-NEXT: s_wait_loadcnt 0x0
; GFX12-TRUE16-NEXT: s_setpc_b64 s[30:31]
;
@@ -4377,11 +4378,11 @@ define half @buffer_fat_ptr_agent_atomic_fadd_ret_f16__offset__waterfall__amdgpu
; GFX12-FAKE16-NEXT: s_mov_b32 s1, exec_lo
; GFX12-FAKE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(SKIP_1) | instid1(VALU_DEP_2)
; GFX12-FAKE16-NEXT: v_and_b32_e32 v4, 3, v6
-; GFX12-FAKE16-NEXT: v_and_b32_e32 v10, -4, v6
+; GFX12-FAKE16-NEXT: v_and_b32_e32 v9, -4, v6
; GFX12-FAKE16-NEXT: v_lshlrev_b32_e32 v4, 3, v4
; GFX12-FAKE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
; GFX12-FAKE16-NEXT: v_lshlrev_b32_e64 v7, v4, 0xffff
-; GFX12-FAKE16-NEXT: v_not_b32_e32 v11, v7
+; GFX12-FAKE16-NEXT: v_not_b32_e32 v10, v7
; GFX12-FAKE16-NEXT: .LBB15_1: ; =>This Inner Loop Header: Depth=1
; GFX12-FAKE16-NEXT: v_readfirstlane_b32 s4, v0
; GFX12-FAKE16-NEXT: v_readfirstlane_b32 s5, v1
@@ -4394,7 +4395,7 @@ define half @buffer_fat_ptr_agent_atomic_fadd_ret_f16__offset__waterfall__amdgpu
; GFX12-FAKE16-NEXT: s_wait_alu 0xfffe
; GFX12-FAKE16-NEXT: s_and_saveexec_b32 s0, s0
; GFX12-FAKE16-NEXT: s_wait_loadcnt 0x0
-; GFX12-FAKE16-NEXT: buffer_load_b32 v7, v10, s[4:7], null offen
+; GFX12-FAKE16-NEXT: buffer_load_b32 v6, v9, s[4:7], null offen
; GFX12-FAKE16-NEXT: s_xor_b32 exec_lo, exec_lo, s0
; GFX12-FAKE16-NEXT: s_cbranch_execnz .LBB15_1
; GFX12-FAKE16-NEXT: ; %bb.2:
@@ -4404,17 +4405,19 @@ define half @buffer_fat_ptr_agent_atomic_fadd_ret_f16__offset__waterfall__amdgpu
; GFX12-FAKE16-NEXT: ; =>This Loop Header: Depth=1
; GFX12-FAKE16-NEXT: ; Child Loop BB15_4 Depth 2
; GFX12-FAKE16-NEXT: s_wait_loadcnt 0x0
-; GFX12-FAKE16-NEXT: v_lshrrev_b32_e32 v6, v4, v7
+; GFX12-FAKE16-NEXT: v_mov_b32_e32 v8, v6
; GFX12-FAKE16-NEXT: s_mov_b32 s2, exec_lo
; GFX12-FAKE16-NEXT: s_wait_storecnt 0x0
; GFX12-FAKE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
+; GFX12-FAKE16-NEXT: v_lshrrev_b32_e32 v6, v4, v8
; GFX12-FAKE16-NEXT: v_add_f16_e32 v6, v6, v5
-; GFX12-FAKE16-NEXT: v_and_b32_e32 v6, 0xffff, v6
; GFX12-FAKE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
+; GFX12-FAKE16-NEXT: v_and_b32_e32 v6, 0xffff, v6
; GFX12-FAKE16-NEXT: v_lshlrev_b32_e32 v6, v4, v6
-; GFX12-FAKE16-NEXT: v_and_or_b32 v6, v7, v11, v6
-; GFX12-FAKE16-NEXT: s_delay_alu instid0(VALU_DEP_1)
-; GFX12-FAKE16-NEXT: v_dual_mov_b32 v9, v7 :: v_dual_mov_b32 v8, v6
+; GFX12-FAKE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
+; GFX12-FAKE16-NEXT: v_and_or_b32 v7, v8, v10, v6
+; GFX12-FAKE16-NEXT: v_mov_b32_e32 v6, v7
+; GFX12-FAKE16-NEXT: v_mov_b32_e32 v7, v8
; GFX12-FAKE16-NEXT: .LBB15_4: ; Parent Loop BB15_3 Depth=1
; GFX12-FAKE16-NEXT: ; => This Inner Loop Header: Depth=2
; GFX12-FAKE16-NEXT: v_readfirstlane_b32 s4, v0
@@ -4428,14 +4431,13 @@ define half @buffer_fat_ptr_agent_atomic_fadd_ret_f16__offset__waterfall__amdgpu
; GFX12-FAKE16-NEXT: s_wait_alu 0xfffe
; GFX12-FAKE16-NEXT: s_and_saveexec_b32 s0, s0
; GFX12-FAKE16-NEXT: s_wait_loadcnt 0x0
-; GFX12-FAKE16-NEXT: buffer_atomic_cmpswap_b32 v[8:9], v10, s[4:7], null offen th:TH_ATOMIC_RETURN
+; GFX12-FAKE16-NEXT: buffer_atomic_cmpswap_b32 v[6:7], v9, s[4:7], null offen th:TH_ATOMIC_RETURN
; GFX12-FAKE16-NEXT: s_xor_b32 exec_lo, exec_lo, s0
; GFX12-FAKE16-NEXT: s_cbranch_execnz .LBB15_4
; GFX12-FAKE16-NEXT: ; %bb.5: ; in Loop: Header=BB15_3 Depth=1
; GFX12-FAKE16-NEXT: s_mov_b32 exec_lo, s2
; GFX12-FAKE16-NEXT: s_wait_loadcnt 0x0
-; GFX12-FAKE16-NEXT: v_cmp_eq_u32_e32 vcc_lo, v8, v7
-; GFX12-FAKE16-NEXT: v_mov_b32_e32 v7, v8
+; GFX12-FAKE16-NEXT: v_cmp_eq_u32_e32 vcc_lo, v6, v8
; GFX12-FAKE16-NEXT: global_inv scope:SCOPE_DEV
; GFX12-FAKE16-NEXT: s_or_b32 s1, vcc_lo, s1
; GFX12-FAKE16-NEXT: s_wait_alu 0xfffe
@@ -4443,7 +4445,7 @@ define half @buffer_fat_ptr_agent_atomic_fadd_ret_f16__offset__waterfall__amdgpu
; GFX12-FAKE16-NEXT: s_cbranch_execnz .LBB15_3
; GFX12-FAKE16-NEXT: ; %bb.6: ; %atomicrmw.end
; GFX12-FAKE16-NEXT: s_or_b32 exec_lo, exec_lo, s1
-; GFX12-FAKE16-NEXT: v_lshrrev_b32_e32 v0, v4, v8
+; GFX12-FAKE16-NEXT: v_lshrrev_b32_e32 v0, v4, v6
; GFX12-FAKE16-NEXT: s_wait_loadcnt 0x0
; GFX12-FAKE16-NEXT: s_setpc_b64 s[30:31]
;
@@ -4468,7 +4470,7 @@ define half @buffer_fat_ptr_agent_atomic_fadd_ret_f16__offset__waterfall__amdgpu
; GFX942-NEXT: v_cmp_eq_u64_e64 s[0:1], s[6:7], v[2:3]
; GFX942-NEXT: s_and_b64 s[0:1], vcc, s[0:1]
; GFX942-NEXT: s_and_saveexec_b64 s[0:1], s[0:1]
-; GFX942-NEXT: buffer_load_dword v7, v10, s[4:7], 0 offen
+; GFX942-NEXT: buffer_load_dword v6, v10, s[4:7], 0 offen
; GFX942-NEXT: s_xor_b64 exec, exec, s[0:1]
; GFX942-NEXT: s_cbranch_execnz .LBB15_1
; GFX942-NEXT: ; %bb.2:
@@ -4478,12 +4480,13 @@ define half @buffer_fat_ptr_agent_atomic_fadd_ret_f16__offset__waterfall__amdgpu
; GFX942-NEXT: ; =>This Loop Header: Depth=1
; GFX942-NEXT: ; Child Loop BB15_4 Depth 2
; GFX942-NEXT: s_waitcnt vmcnt(0)
-; GFX942-NEXT: v_lshrrev_b32_e32 v6, v4, v7
+; GFX942-NEXT: v_mov_b32_e32 v9, v6
+; GFX942-NEXT: v_lshrrev_b32_e32 v6, v4, v9
; GFX942-NEXT: v_add_f16_e32 v6, v6, v5
; GFX942-NEXT: v_lshlrev_b32_e32 v6, v4, v6
-; GFX942-NEXT: v_and_or_b32 v6, v7, v11, v6
+; GFX942-NEXT: v_and_or_b32 v8, v9, v11, v6
; GFX942-NEXT: s_mov_b64 s[8:9], exec
-; GFX942-NEXT: v_mov_b64_e32 v[8:9], v[6:7]
+; GFX942-NEXT: v_mov_b64_e32 v[6:7], v[8:9]
; GFX942-NEXT: buffer_wbl2 sc1
; GFX942-NEXT: .LBB15_4: ; Parent Loop BB15_3 Depth=1
; GFX942-NEXT: ; => This Inner Loop Header: Depth=2
@@ -4497,21 +4500,20 @@ define half @buffer_fat_ptr_agent_atomic_fadd_ret_f16__offset__waterfall__amdgpu
; GFX942-NEXT: s_and_b64 s[0:1], vcc, s[0:1]
; GFX942-NEXT: s_and_saveexec_b64 s[0:1], s[0:1]
; GFX942-NEXT: s_waitcnt vmcnt(0)
-; GFX942-NEXT: buffer_atomic_cmpswap v[8:9], v10, s[4:7], 0 offen sc0
+; GFX942-NEXT: buffer_atomic_cmpswap v[6:7], v10, s[4:7], 0 offen sc0
; GFX942-NEXT: s_xor_b64 exec, exec, s[0:1]
; GFX942-NEXT: s_cbranch_execnz .LBB15_4
; GFX942-NEXT: ; %bb.5: ; in Loop: Header=BB15_3 Depth=1
; GFX942-NEXT: s_mov_b64 exec, s[8:9]
; GFX942-NEXT: s_waitcnt vmcnt(0)
-; GFX942-NEXT: v_cmp_eq_u32_e32 vcc, v8, v7
+; GFX942-NEXT: v_cmp_eq_u32_e32 vcc, v6, v9
; GFX942-NEXT: s_or_b64 s[2:3], vcc, s[2:3]
-; GFX942-NEXT: v_mov_b32_e32 v7, v8
; GFX942-NEXT: buffer_inv sc1
; GFX942-NEXT: s_andn2_b64 exec, exec, s[2:3]
; GFX942-NEXT: s_cbranch_execnz .LBB15_3
; GFX942-NEXT: ; %bb.6: ; %atomicrmw.end
; GFX942-NEXT: s_or_b64 exec, exec, s[2:3]
-; GFX942-NEXT: v_lshrrev_b32_e32 v0, v4, v8
+; GFX942-NEXT: v_lshrrev_b32_e32 v0, v4, v6
; GFX942-NEXT: s_setpc_b64 s[30:31]
;
; GFX11-TRUE16-LABEL: buffer_fat_ptr_agent_atomic_fadd_ret_f16__offset__waterfall__amdgpu_no_fine_grained_memory:
@@ -4522,11 +4524,11 @@ define half @buffer_fat_ptr_agent_atomic_fadd_ret_f16__offset__waterfall__amdgpu
; GFX11-TRUE16-NEXT: s_mov_b32 s2, exec_lo
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(SKIP_1) | instid1(VALU_DEP_2)
; GFX11-TRUE16-NEXT: v_and_b32_e32 v4, 3, v6
-; GFX11-TRUE16-NEXT: v_and_b32_e32 v10, -4, v6
+; GFX11-TRUE16-NEXT: v_and_b32_e32 v9, -4, v6
; GFX11-TRUE16-NEXT: v_lshlrev_b32_e32 v4, 3, v4
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
; GFX11-TRUE16-NEXT: v_lshlrev_b32_e64 v7, v4, 0xffff
-; GFX11-TRUE16-NEXT: v_not_b32_e32 v11, v7
+; GFX11-TRUE16-NEXT: v_not_b32_e32 v10, v7
; GFX11-TRUE16-NEXT: .LBB15_1: ; =>This Inner Loop Header: Depth=1
; GFX11-TRUE16-NEXT: v_readfirstlane_b32 s4, v0
; GFX11-TRUE16-NEXT: v_readfirstlane_b32 s5, v1
@@ -4538,7 +4540,7 @@ define half @buffer_fat_ptr_agent_atomic_fadd_ret_f16__offset__waterfall__amdgpu
; GFX11-TRUE16-NEXT: s_and_b32 s0, vcc_lo, s0
; GFX11-TRUE16-NEXT: s_delay_alu instid0(SALU_CYCLE_1)
; GFX11-TRUE16-NEXT: s_and_saveexec_b32 s0, s0
-; GFX11-TRUE16-NEXT: buffer_load_b32 v7, v10, s[4:7], 0 offen
+; GFX11-TRUE16-NEXT: buffer_load_b32 v6, v9, s[4:7], 0 offen
; GFX11-TRUE16-NEXT: s_xor_b32 exec_lo, exec_lo, s0
; GFX11-TRUE16-NEXT: s_cbranch_execnz .LBB15_1
; GFX11-TRUE16-NEXT: ; %bb.2:
@@ -4548,17 +4550,19 @@ define half @buffer_fat_ptr_agent_atomic_fadd_ret_f16__offset__waterfall__amdgpu
; GFX11-TRUE16-NEXT: ; =>This Loop Header: Depth=1
; GFX11-TRUE16-NEXT: ; Child Loop BB15_4 Depth 2
; GFX11-TRUE16-NEXT: s_waitcnt vmcnt(0)
-; GFX11-TRUE16-NEXT: v_lshrrev_b32_e32 v6, v4, v7
+; GFX11-TRUE16-NEXT: v_mov_b32_e32 v8, v6
; GFX11-TRUE16-NEXT: s_mov_b32 s2, exec_lo
; GFX11-TRUE16-NEXT: s_waitcnt_vscnt null, 0x0
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
+; GFX11-TRUE16-NEXT: v_lshrrev_b32_e32 v6, v4, v8
; GFX11-TRUE16-NEXT: v_add_f16_e32 v6.l, v6.l, v5.l
-; GFX11-TRUE16-NEXT: v_and_b32_e32 v6, 0xffff, v6
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
+; GFX11-TRUE16-NEXT: v_and_b32_e32 v6, 0xffff, v6
; GFX11-TRUE16-NEXT: v_lshlrev_b32_e32 v6, v4, v6
-; GFX11-TRUE16-NEXT: v_and_or_b32 v6, v7, v11, v6
-; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1)
-; GFX11-TRUE16-NEXT: v_dual_mov_b32 v9, v7 :: v_dual_mov_b32 v8, v6
+; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
+; GFX11-TRUE16-NEXT: v_and_or_b32 v7, v8, v10, v6
+; GFX11-TRUE16-NEXT: v_mov_b32_e32 v6, v7
+; GFX11-TRUE16-NEXT: v_mov_b32_e32 v7, v8
; GFX11-TRUE16-NEXT: .LBB15_4: ; Parent Loop BB15_3 Depth=1
; GFX11-TRUE16-NEXT: ; => This Inner Loop Header: Depth=2
; GFX11-TRUE16-NEXT: v_readfirstlane_b32 s4, v0
@@ -4572,14 +4576,13 @@ define half @buffer_fat_ptr_agent_atomic_fadd_ret_f16__offset__waterfall__amdgpu
; GFX11-TRUE16-NEXT: s_delay_alu instid0(SALU_CYCLE_1)
; GFX11-TRUE16-NEXT: s_and_saveexec_b32 s0, s0
; GFX11-TRUE16-NEXT: s_waitcnt vmcnt(0)
-; GFX11-TRUE16-NEXT: buffer_atomic_cmpswap_b32 v[8:9], v10, s[4:7], 0 offen glc
+; GFX11-TRUE16-NEXT: buffer_atomic_cmpswap_b32 v[6:7], v9, s[4:7], 0 offen glc
; GFX11-TRUE16-NEXT: s_xor_b32 exec_lo, exec_lo, s0
; GFX11-TRUE16-NEXT: s_cbranch_execnz .LBB15_4
; GFX11-TRUE16-NEXT: ; %bb.5: ; in Loop: Header=BB15_3 Depth=1
; GFX11-TRUE16-NEXT: s_mov_b32 exec_lo, s2
; GFX11-TRUE16-NEXT: s_waitcnt vmcnt(0)
-; GFX11-TRUE16-NEXT: v_cmp_eq_u32_e32 vcc_lo, v8, v7
-; GFX11-TRUE16-NEXT: v_mov_b32_e32 v7, v8
+; GFX11-TRUE16-NEXT: v_cmp_eq_u32_e32 vcc_lo, v6, v8
; GFX11-TRUE16-NEXT: buffer_gl1_inv
; GFX11-TRUE16-NEXT: buffer_gl0_inv
; GFX11-TRUE16-NEXT: s_or_b32 s1, vcc_lo, s1
@@ -4588,7 +4591,7 @@ define half @buffer_fat_ptr_agent_atomic_fadd_ret_f16__offset__waterfall__amdgpu
; GFX11-TRUE16-NEXT: s_cbranch_execnz .LBB15_3
; GFX11-TRUE16-NEXT: ; %bb.6: ; %atomicrmw.end
; GFX11-TRUE16-NEXT: s_or_b32 exec_lo, exec_lo, s1
-; GFX11-TRUE16-NEXT: v_lshrrev_b32_e32 v0, v4, v8
+; GFX11-TRUE16-NEXT: v_lshrrev_b32_e32 v0, v4, v6
; GFX11-TRUE16-NEXT: s_setpc_b64 s[30:31]
;
; GFX11-FAKE16-LABEL: buffer_fat_ptr_agent_atomic_fadd_ret_f16__offset__waterfall__amdgpu_no_fine_grained_memory:
@@ -4599,11 +4602,11 @@ define half @buffer_fat_ptr_agent_atomic_fadd_ret_f16__offset__waterfall__amdgpu
; GFX11-FAKE16-NEXT: s_mov_b32 s2, exec_lo
; GFX11-FAKE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(SKIP_1) | instid1(VALU_DEP_2)
; GFX11-FAKE16-NEXT: v_and_b32_e32 v4, 3, v6
-; GFX11-FAKE16-NEXT: v_and_b32_e32 v10, -4, v6
+; GFX11-FAKE16-NEXT: v_and_b32_e32 v9, -4, v6
; GFX11-FAKE16-NEXT: v_lshlrev_b32_e32 v4, 3, v4
; GFX11-FAKE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
; GFX11-FAKE16-NEXT: v_lshlrev_b32_e64 v7, v4, 0xffff
-; GFX11-FAKE16-NEXT: v_not_b32_e32 v11, v7
+; GFX11-FAKE16-NEXT: v_not_b32_e32 v10, v7
; GFX11-FAKE16-NEXT: .LBB15_1: ; =>This Inner Loop Header: Depth=1
; GFX11-FAKE16-NEXT: v_readfirstlane_b32 s4, v0
; GFX11-FAKE16-NEXT: v_readfirstlane_b32 s5, v1
@@ -4615,7 +4618,7 @@ define half @buffer_fat_ptr_agent_atomic_fadd_ret_f16__offset__waterfall__amdgpu
; GFX11-FAKE16-NEXT: s_and_b32 s0, vcc_lo, s0
; GFX11-FAKE16-NEXT: s_delay_alu instid0(SALU_CYCLE_1)
; GFX11-FAKE16-NEXT: s_and_saveexec_b32 s0, s0
-; GFX11-FAKE16-NEXT: buffer_load_b32 v7, v10, s[4:7], 0 offen
+; GFX11-FAKE16-NEXT: buffer_load_b32 v6, v9, s[4:7], 0 offen
; GFX11-FAKE16-NEXT: s_xor_b32 exec_lo, exec_lo, s0
; GFX11-FAKE16-NEXT: s_cbranch_execnz .LBB15_1
; GFX11-FAKE16-NEXT: ; %bb.2:
@@ -4625,17 +4628,19 @@ define half @buffer_fat_ptr_agent_atomic_fadd_ret_f16__offset__waterfall__amdgpu
; GFX11-FAKE16-NEXT: ; =>This Loop Header: Depth=1
; GFX11-FAKE16-NEXT: ; Child Loop BB15_4 Depth 2
; GFX11-FAKE16-NEXT: s_waitcnt vmcnt(0)
-; GFX11-FAKE16-NEXT: v_lshrrev_b32_e32 v6, v4, v7
+; GFX11-FAKE16-NEXT: v_mov_b32_e32 v8, v6
; GFX11-FAKE16-NEXT: s_mov_b32 s2, exec_lo
; GFX11-FAKE16-NEXT: s_waitcnt_vscnt null, 0x0
; GFX11-FAKE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
+; GFX11-FAKE16-NEXT: v_lshrrev_b32_e32 v6, v4, v8
; GFX11-FAKE16-NEXT: v_add_f16_e32 v6, v6, v5
-; GFX11-FAKE16-NEXT: v_and_b32_e32 v6, 0xffff, v6
; GFX11-FAKE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
+; GFX11-FAKE16-NEXT: v_and_b32_e32 v6, 0xffff, v6
; GFX11-FAKE16-NEXT: v_lshlrev_b32_e32 v6, v4, v6
-; GFX11-FAKE16-NEXT: v_and_or_b32 v6, v7, v11, v6
-; GFX11-FAKE16-NEXT: s_delay_alu instid0(VALU_DEP_1)
-; GFX11-FAKE16-NEXT: v_dual_mov_b32 v9, v7 :: v_dual_mov_b32 v8, v6
+; GFX11-FAKE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
+; GFX11-FAKE16-NEXT: v_and_or_b32 v7, v8, v10, v6
+; GFX11-FAKE16-NEXT: v_mov_b32_e32 v6, v7
+; GFX11-FAKE16-NEXT: v_mov_b32_e32 v7, v8
; GFX11-FAKE16-NEXT: .LBB15_4: ; Parent Loop BB15_3 Depth=1
; GFX11-FAKE16-NEXT: ; => This Inner Loop Header: Depth=2
; GFX11-FAKE16-NEXT: v_readfirstlane_b32 s4, v0
@@ -4649,14 +4654,13 @@ define half @buffer_fat_ptr_agent_atomic_fadd_ret_f16__offset__waterfall__amdgpu
; GFX11-FAKE16-NEXT: s_delay_alu instid0(SALU_CYCLE_1)
; GFX11-FAKE16-NEXT: s_and_saveexec_b32 s0, s0
; GFX11-FAKE16-NEXT: s_waitcnt vmcnt(0)
-; GFX11-FAKE16-NEXT: buffer_atomic_cmpswap_b32 v[8:9], v10, s[4:7], 0 offen glc
+; GFX11-FAKE16-NEXT: buffer_atomic_cmpswap_b32 v[6:7], v9, s[4:7], 0 offen glc
; GFX11-FAKE16-NEXT: s_xor_b32 exec_lo, exec_lo, s0
; GFX11-FAKE16-NEXT: s_cbranch_execnz .LBB15_4
; GFX11-FAKE16-NEXT: ; %bb.5: ; in Loop: Header=BB15_3 Depth=1
; GFX11-FAKE16-NEXT: s_mov_b32 exec_lo, s2
; GFX11-FAKE16-NEXT: s_waitcnt vmcnt(0)
-; GFX11-FAKE16-NEXT: v_cmp_eq_u32_e32 vcc_lo, v8, v7
-; GFX11-FAKE16-NEXT: v_mov_b32_e32 v7, v8
+; GFX11-FAKE16-NEXT: v_cmp_eq_u32_e32 vcc_lo, v6, v8
; GFX11-FAKE16-NEXT: buffer_gl1_inv
; GFX11-FAKE16-NEXT: buffer_gl0_inv
; GFX11-FAKE16-NEXT: s_or_b32 s1, vcc_lo, s1
@@ -4665,7 +4669,7 @@ define half @buffer_fat_ptr_agent_atomic_fadd_ret_f16__offset__waterfall__amdgpu
; GFX11-FAKE16-NEXT: s_cbranch_execnz .LBB15_3
; GFX11-FAKE16-NEXT: ; %bb.6: ; %atomicrmw.end
; GFX11-FAKE16-NEXT: s_or_b32 exec_lo, exec_lo, s1
-; GFX11-FAKE16-NEXT: v_lshrrev_b32_e32 v0, v4, v8
+; GFX11-FAKE16-NEXT: v_lshrrev_b32_e32 v0, v4, v6
; GFX11-FAKE16-NEXT: s_setpc_b64 s[30:31]
;
; GFX10-LABEL: buffer_fat_ptr_agent_atomic_fadd_ret_f16__offset__waterfall__amdgpu_no_fine_grained_memory:
@@ -4675,10 +4679,10 @@ define half @buffer_fat_ptr_agent_atomic_fadd_ret_f16__offset__waterfall__amdgpu
; GFX10-NEXT: s_mov_b32 s5, 0
; GFX10-NEXT: s_mov_b32 s6, exec_lo
; GFX10-NEXT: v_and_b32_e32 v4, 3, v6
-; GFX10-NEXT: v_and_b32_e32 v10, -4, v6
+; GFX10-NEXT: v_and_b32_e32 v9, -4, v6
; GFX10-NEXT: v_lshlrev_b32_e32 v4, 3, v4
; GFX10-NEXT: v_lshlrev_b32_e64 v7, v4, 0xffff
-; GFX10-NEXT: v_not_b32_e32 v11, v7
+; GFX10-NEXT: v_not_b32_e32 v10, v7
; GFX10-NEXT: .LBB15_1: ; =>This Inner Loop Header: Depth=1
; GFX10-NEXT: v_readfirstlane_b32 s8, v0
; GFX10-NEXT: v_readfirstlane_b32 s9, v1
@@ -4688,7 +4692,7 @@ define half @buffer_fat_ptr_agent_atomic_fadd_ret_f16__offset__waterfall__amdgpu
; GFX10-NEXT: v_cmp_eq_u64_e64 s4, s[10:11], v[2:3]
; GFX10-NEXT: s_and_b32 s4, vcc_lo, s4
; GFX10-NEXT: s_and_saveexec_b32 s4, s4
-; GFX10-NEXT: buffer_load_dword v7, v10, s[8:11], 0 offen
+; GFX10-NEXT: buffer_load_dword v6, v9, s[8:11], 0 offen
; GFX10-NEXT: s_waitcnt_depctr 0xffe3
; GFX10-NEXT: s_xor_b32 exec_lo, exec_lo, s4
; GFX10-NEXT: s_cbranch_execnz .LBB15_1
@@ -4698,14 +4702,15 @@ define half @buffer_fat_ptr_agent_atomic_fadd_ret_f16__offset__waterfall__amdgpu
; GFX10-NEXT: ; =>This Loop Header: Depth=1
; GFX10-NEXT: ; Child Loop BB15_4 Depth 2
; GFX10-NEXT: s_waitcnt vmcnt(0)
-; GFX10-NEXT: v_lshrrev_b32_e32 v6, v4, v7
+; GFX10-NEXT: v_mov_b32_e32 v8, v6
; GFX10-NEXT: s_mov_b32 s6, exec_lo
; GFX10-NEXT: s_waitcnt_vscnt null, 0x0
+; GFX10-NEXT: v_lshrrev_b32_e32 v6, v4, v8
; GFX10-NEXT: v_add_f16_e32 v6, v6, v5
; GFX10-NEXT: v_lshlrev_b32_sdwa v6, v4, v6 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:WORD_0
-; GFX10-NEXT: v_and_or_b32 v6, v7, v11, v6
-; GFX10-NEXT: v_mov_b32_e32 v9, v7
-; GFX10-NEXT: v_mov_b32_e32 v8, v6
+; GFX10-NEXT: v_and_or_b32 v7, v8, v10, v6
+; GFX10-NEXT: v_mov_b32_e32 v6, v7
+; GFX10-NEXT: v_mov_b32_e32 v7, v8
; GFX10-NEXT: .LBB15_4: ; Parent Loop BB15_3 Depth=1
; GFX10-NEXT: ; => This Inner Loop Header: Depth=2
; GFX10-NEXT: v_readfirstlane_b32 s8, v0
@@ -4717,15 +4722,14 @@ define half @buffer_fat_ptr_agent_atomic_fadd_ret_f16__offset__waterfall__amdgpu
; GFX10-NEXT: s_and_b32 s4, vcc_lo, s4
; GFX10-NEXT: s_and_saveexec_b32 s4, s4
; GFX10-NEXT: s_waitcnt vmcnt(0)
-; GFX10-NEXT: buffer_atomic_cmpswap v[8:9], v10, s[8:11], 0 offen glc
+; GFX10-NEXT: buffer_atomic_cmpswap v[6:7], v9, s[8:11], 0 offen glc
; GFX10-NEXT: s_waitcnt_depctr 0xffe3
; GFX10-NEXT: s_xor_b32 exec_lo, exec_lo, s4
; GFX10-NEXT: s_cbranch_execnz .LBB15_4
; GFX10-NEXT: ; %bb.5: ; in Loop: Header=BB15_3 Depth=1
; GFX10-NEXT: s_mov_b32 exec_lo, s6
; GFX10-NEXT: s_waitcnt vmcnt(0)
-; GFX10-NEXT: v_cmp_eq_u32_e32 vcc_lo, v8, v7
-; GFX10-NEXT: v_mov_b32_e32 v7, v8
+; GFX10-NEXT: v_cmp_eq_u32_e32 vcc_lo, v6, v8
; GFX10-NEXT: buffer_gl1_inv
; GFX10-NEXT: buffer_gl0_inv
; GFX10-NEXT: s_or_b32 s5, vcc_lo, s5
@@ -4734,7 +4738,7 @@ define half @buffer_fat_ptr_agent_atomic_fadd_ret_f16__offset__waterfall__amdgpu
; GFX10-NEXT: s_cbranch_execnz .LBB15_3
; GFX10-NEXT: ; %bb.6: ; %atomicrmw.end
; GFX10-NEXT: s_or_b32 exec_lo, exec_lo, s5
-; GFX10-NEXT: v_lshrrev_b32_e32 v0, v4, v8
+; GFX10-NEXT: v_lshrrev_b32_e32 v0, v4, v6
; GFX10-NEXT: s_setpc_b64 s[30:31]
;
; GFX90A-LABEL: buffer_fat_ptr_agent_atomic_fadd_ret_f16__offset__waterfall__amdgpu_no_fine_grained_memory:
@@ -4758,7 +4762,7 @@ define half @buffer_fat_ptr_agent_atomic_fadd_ret_f16__offset__waterfall__amdgpu
; GFX90A-NEXT: s_and_b64 s[4:5], vcc, s[4:5]
; GFX90A-NEXT: s_and_saveexec_b64 s[4:5], s[4:5]
; GFX90A-NEXT: s_nop 0
-; GFX90A-NEXT: buffer_load_dword v7, v10, s[8:11], 0 offen
+; GFX90A-NEXT: buffer_load_dword v6, v10, s[8:11], 0 offen
; GFX90A-NEXT: s_xor_b64 exec, exec, s[4:5]
; GFX90A-NEXT: s_cbranch_execnz .LBB15_1
; GFX90A-NEXT: ; %bb.2:
@@ -4768,12 +4772,13 @@ define half @buffer_fat_ptr_agent_atomic_fadd_ret_f16__offset__waterfall__amdgpu
; GFX90A-NEXT: ; =>This Loop Header: Depth=1
; GFX90A-NEXT: ; Child Loop BB15_4 Depth 2
; GFX90A-NEXT: s_waitcnt vmcnt(0)
-; GFX90A-NEXT: v_lshrrev_b32_e32 v6, v4, v7
+; GFX90A-NEXT: v_mov_b32_e32 v9, v6
+; GFX90A-NEXT: v_lshrrev_b32_e32 v6, v4, v9
; GFX90A-NEXT: v_add_f16_e32 v6, v6, v5
; GFX90A-NEXT: v_lshlrev_b32_e32 v6, v4, v6
-; GFX90A-NEXT: v_and_or_b32 v6, v7, v11, v6
+; GFX90A-NEXT: v_and_or_b32 v8, v9, v11, v6
; GFX90A-NEXT: s_mov_b64 s[12:13], exec
-; GFX90A-NEXT: v_pk_mov_b32 v[8:9], v[6:7], v[6:7] op_sel:[0,1]
+; GFX90A-NEXT: v_pk_mov_b32 v[6:7], v[8:9], v[8:9] op_sel:[0,1]
; GFX90A-NEXT: .LBB15_4: ; Parent Loop BB15_3 Depth=1
; GFX90A-NEXT: ; => This Inner Loop Header: Depth=2
; GFX90A-NEXT: v_readfirstlane_b32 s8, v0
@@ -4785,33 +4790,32 @@ define half @buffer_fat_ptr_agent_atomic_fadd_ret_f16__offset__waterfall__amdgpu
; GFX90A-NEXT: s_and_b64 s[4:5], vcc, s[4:5]
; GFX90A-NEXT: s_and_saveexec_b64 s[4:5], s[4:5]
; GFX90A-NEXT: s_waitcnt vmcnt(0)
-; GFX90A-NEXT: buffer_atomic_cmpswap v[8:9], v10, s[8:11], 0 offen glc
+; GFX90A-NEXT: buffer_atomic_cmpswap v[6:7], v10, s[8:11], 0 offen glc
; GFX90A-NEXT: s_xor_b64 exec, exec, s[4:5]
; GFX90A-NEXT: s_cbranch_execnz .LBB15_4
; GFX90A-NEXT: ; %bb.5: ; in Loop: Header=BB15_3 Depth=1
; GFX90A-NEXT: s_mov_b64 exec, s[12:13]
; GFX90A-NEXT: s_waitcnt vmcnt(0)
-; GFX90A-NEXT: v_cmp_eq_u32_e32 vcc, v8, v7
+; GFX90A-NEXT: v_cmp_eq_u32_e32 vcc, v6, v9
; GFX90A-NEXT: s_or_b64 s[6:7], vcc, s[6:7]
-; GFX90A-NEXT: v_mov_b32_e32 v7, v8
; GFX90A-NEXT: buffer_wbinvl1
; GFX90A-NEXT: s_andn2_b64 exec, exec, s[6:7]
; GFX90A-NEXT: s_cbranch_execnz .LBB15_3
; GFX90A-NEXT: ; %bb.6: ; %atomicrmw.end
; GFX90A-NEXT: s_or_b64 exec, exec, s[6:7]
-; GFX90A-NEXT: v_lshrrev_b32_e32 v0, v4, v8
+; GFX90A-NEXT: v_lshrrev_b32_e32 v0, v4, v6
; GFX90A-NEXT: s_setpc_b64 s[30:31]
;
; GFX908-LABEL: buffer_fat_ptr_agent_atomic_fadd_ret_f16__offset__waterfall__amdgpu_no_fine_grained_memory:
; GFX908: ; %bb.0:
; GFX908-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX908-NEXT: v_add_u32_e32 v4, 0x200, v4
-; GFX908-NEXT: v_and_b32_e32 v10, -4, v4
+; GFX908-NEXT: v_and_b32_e32 v9, -4, v4
; GFX908-NEXT: v_and_b32_e32 v4, 3, v4
; GFX908-NEXT: v_lshlrev_b32_e32 v4, 3, v4
; GFX908-NEXT: s_mov_b32 s4, 0xffff
; GFX908-NEXT: v_lshlrev_b32_e64 v6, v4, s4
-; GFX908-NEXT: v_not_b32_e32 v11, v6
+; GFX908-NEXT: v_not_b32_e32 v10, v6
; GFX908-NEXT: s_mov_b64 s[6:7], exec
; GFX908-NEXT: .LBB15_1: ; =>This Inner Loop Header: Depth=1
; GFX908-NEXT: v_readfirstlane_b32 s8, v0
@@ -4823,7 +4827,7 @@ define half @buffer_fat_ptr_agent_atomic_fadd_ret_f16__offset__waterfall__amdgpu
; GFX908-NEXT: s_and_b64 s[4:5], vcc, s[4:5]
; GFX908-NEXT: s_and_saveexec_b64 s[4:5], s[4:5]
; GFX908-NEXT: s_nop 0
-; GFX908-NEXT: buffer_load_dword v7, v10, s[8:11], 0 offen
+; GFX908-NEXT: buffer_load_dword v6, v9, s[8:11], 0 offen
; GFX908-NEXT: s_xor_b64 exec, exec, s[4:5]
; GFX908-NEXT: s_cbranch_execnz .LBB15_1
; GFX908-NEXT: ; %bb.2:
@@ -4833,13 +4837,14 @@ define half @buffer_fat_ptr_agent_atomic_fadd_ret_f16__offset__waterfall__amdgpu
; GFX908-NEXT: ; =>This Loop Header: Depth=1
; GFX908-NEXT: ; Child Loop BB15_4 Depth 2
; GFX908-NEXT: s_waitcnt vmcnt(0)
-; GFX908-NEXT: v_lshrrev_b32_e32 v6, v4, v7
+; GFX908-NEXT: v_mov_b32_e32 v8, v6
+; GFX908-NEXT: v_lshrrev_b32_e32 v6, v4, v8
; GFX908-NEXT: v_add_f16_e32 v6, v6, v5
; GFX908-NEXT: v_lshlrev_b32_e32 v6, v4, v6
-; GFX908-NEXT: v_and_or_b32 v6, v7, v11, v6
-; GFX908-NEXT: v_mov_b32_e32 v9, v7
+; GFX908-NEXT: v_and_or_b32 v7, v8, v10, v6
+; GFX908-NEXT: v_mov_b32_e32 v6, v7
; GFX908-NEXT: s_mov_b64 s[12:13], exec
-; GFX908-NEXT: v_mov_b32_e32 v8, v6
+; GFX908-NEXT: v_mov_b32_e32 v7, v8
; GFX908-NEXT: .LBB15_4: ; Parent Loop BB15_3 Depth=1
; GFX908-NEXT: ; => This Inner Loop Header: Depth=2
; GFX908-NEXT: v_readfirstlane_b32 s8, v0
@@ -4851,33 +4856,32 @@ define half @buffer_fat_ptr_agent_atomic_fadd_ret_f16__offset__waterfall__amdgpu
; GFX908-NEXT: s_and_b64 s[4:5], vcc, s[4:5]
; GFX908-NEXT: s_and_saveexec_b64 s[4:5], s[4:5]
; GFX908-NEXT: s_waitcnt vmcnt(0)
-; GFX908-NEXT: buffer_atomic_cmpswap v[8:9], v10, s[8:11], 0 offen glc
+; GFX908-NEXT: buffer_atomic_cmpswap v[6:7], v9, s[8:11], 0 offen glc
; GFX908-NEXT: s_xor_b64 exec, exec, s[4:5]
; GFX908-NEXT: s_cbranch_execnz .LBB15_4
; GFX908-NEXT: ; %bb.5: ; in Loop: Header=BB15_3 Depth=1
; GFX908-NEXT: s_mov_b64 exec, s[12:13]
; GFX908-NEXT: s_waitcnt vmcnt(0)
-; GFX908-NEXT: v_cmp_eq_u32_e32 vcc, v8, v7
+; GFX908-NEXT: v_cmp_eq_u32_e32 vcc, v6, v8
; GFX908-NEXT: s_or_b64 s[6:7], vcc, s[6:7]
-; GFX908-NEXT: v_mov_b32_e32 v7, v8
; GFX908-NEXT: buffer_wbinvl1
; GFX908-NEXT: s_andn2_b64 exec, exec, s[6:7]
; GFX908-NEXT: s_cbranch_execnz .LBB15_3
; GFX908-NEXT: ; %bb.6: ; %atomicrmw.end
; GFX908-NEXT: s_or_b64 exec, exec, s[6:7]
-; GFX908-NEXT: v_lshrrev_b32_e32 v0, v4, v8
+; GFX908-NEXT: v_lshrrev_b32_e32 v0, v4, v6
; GFX908-NEXT: s_setpc_b64 s[30:31]
;
; GFX8-LABEL: buffer_fat_ptr_agent_atomic_fadd_ret_f16__offset__waterfall__amdgpu_no_fine_grained_memory:
; GFX8: ; %bb.0:
; GFX8-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX8-NEXT: v_add_u32_e32 v4, vcc, 0x200, v4
-; GFX8-NEXT: v_and_b32_e32 v10, -4, v4
+; GFX8-NEXT: v_and_b32_e32 v9, -4, v4
; GFX8-NEXT: v_and_b32_e32 v4, 3, v4
; GFX8-NEXT: v_lshlrev_b32_e32 v4, 3, v4
; GFX8-NEXT: s_mov_b32 s4, 0xffff
; GFX8-NEXT: v_lshlrev_b32_e64 v6, v4, s4
-; GFX8-NEXT: v_not_b32_e32 v11, v6
+; GFX8-NEXT: v_not_b32_e32 v10, v6
; GFX8-NEXT: s_mov_b64 s[6:7], exec
; GFX8-NEXT: .LBB15_1: ; =>This Inner Loop Header: Depth=1
; GFX8-NEXT: v_readfirstlane_b32 s8, v0
@@ -4889,7 +4893,7 @@ define half @buffer_fat_ptr_agent_atomic_fadd_ret_f16__offset__waterfall__amdgpu
; GFX8-NEXT: s_and_b64 s[4:5], vcc, s[4:5]
; GFX8-NEXT: s_and_saveexec_b64 s[4:5], s[4:5]
; GFX8-NEXT: s_nop 0
-; GFX8-NEXT: buffer_load_dword v7, v10, s[8:11], 0 offen
+; GFX8-NEXT: buffer_load_dword v6, v9, s[8:11], 0 offen
; GFX8-NEXT: s_xor_b64 exec, exec, s[4:5]
; GFX8-NEXT: s_cbranch_execnz .LBB15_1
; GFX8-NEXT: ; %bb.2:
@@ -4899,14 +4903,15 @@ define half @buffer_fat_ptr_agent_atomic_fadd_ret_f16__offset__waterfall__amdgpu
; GFX8-NEXT: ; =>This Loop Header: Depth=1
; GFX8-NEXT: ; Child Loop BB15_4 Depth 2
; GFX8-NEXT: s_waitcnt vmcnt(0)
-; GFX8-NEXT: v_lshrrev_b32_e32 v6, v4, v7
+; GFX8-NEXT: v_mov_b32_e32 v8, v6
+; GFX8-NEXT: v_lshrrev_b32_e32 v6, v4, v8
; GFX8-NEXT: v_add_f16_e32 v6, v6, v5
; GFX8-NEXT: v_lshlrev_b32_e32 v6, v4, v6
-; GFX8-NEXT: v_and_b32_e32 v8, v7, v11
-; GFX8-NEXT: v_or_b32_e32 v6, v8, v6
-; GFX8-NEXT: v_mov_b32_e32 v9, v7
+; GFX8-NEXT: v_and_b32_e32 v7, v8, v10
+; GFX8-NEXT: v_or_b32_e32 v7, v7, v6
+; GFX8-NEXT: v_mov_b32_e32 v6, v7
; GFX8-NEXT: s_mov_b64 s[12:13], exec
-; GFX8-NEXT: v_mov_b32_e32 v8, v6
+; GFX8-NEXT: v_mov_b32_e32 v7, v8
; GFX8-NEXT: .LBB15_4: ; Parent Loop BB15_3 Depth=1
; GFX8-NEXT: ; => This Inner Loop Header: Depth=2
; GFX8-NEXT: v_readfirstlane_b32 s8, v0
@@ -4918,21 +4923,20 @@ define half @buffer_fat_ptr_agent_atomic_fadd_ret_f16__offset__waterfall__amdgpu
; GFX8-NEXT: s_and_b64 s[4:5], vcc, s[4:5]
; GFX8-NEXT: s_and_saveexec_b64 s[4:5], s[4:5]
; GFX8-NEXT: s_waitcnt vmcnt(0)
-; GFX8-NEXT: buffer_atomic_cmpswap v[8:9], v10, s[8:11], 0 offen glc
+; GFX8-NEXT: buffer_atomic_cmpswap v[6:7], v9, s[8:11], 0 offen glc
; GFX8-NEXT: s_xor_b64 exec, exec, s[4:5]
; GFX8-NEXT: s_cbranch_execnz .LBB15_4
; GFX8-NEXT: ; %bb.5: ; in Loop: Header=BB15_3 Depth=1
; GFX8-NEXT: s_mov_b64 exec, s[12:13]
; GFX8-NEXT: s_waitcnt vmcnt(0)
-; GFX8-NEXT: v_cmp_eq_u32_e32 vcc, v8, v7
+; GFX8-NEXT: v_cmp_eq_u32_e32 vcc, v6, v8
; GFX8-NEXT: s_or_b64 s[6:7], vcc, s[6:7]
-; GFX8-NEXT: v_mov_b32_e32 v7, v8
; GFX8-NEXT: buffer_wbinvl1
; GFX8-NEXT: s_andn2_b64 exec, exec, s[6:7]
; GFX8-NEXT: s_cbranch_execnz .LBB15_3
; GFX8-NEXT: ; %bb.6: ; %atomicrmw.end
; GFX8-NEXT: s_or_b64 exec, exec, s[6:7]
-; GFX8-NEXT: v_lshrrev_b32_e32 v0, v4, v8
+; GFX8-NEXT: v_lshrrev_b32_e32 v0, v4, v6
; GFX8-NEXT: s_setpc_b64 s[30:31]
;
; GFX7-LABEL: buffer_fat_ptr_agent_atomic_fadd_ret_f16__offset__waterfall__amdgpu_no_fine_grained_memory:
@@ -4954,18 +4958,19 @@ define half @buffer_fat_ptr_agent_atomic_fadd_ret_f16__offset__waterfall__amdgpu
; GFX7-NEXT: v_cmp_eq_u64_e64 s[4:5], s[10:11], v[2:3]
; GFX7-NEXT: s_and_b64 s[4:5], vcc, s[4:5]
; GFX7-NEXT: s_and_saveexec_b64 s[4:5], s[4:5]
-; GFX7-NEXT: buffer_load_dword v6, v8, s[8:11], 0 offen
+; GFX7-NEXT: buffer_load_dword v4, v8, s[8:11], 0 offen
; GFX7-NEXT: s_xor_b64 exec, exec, s[4:5]
; GFX7-NEXT: s_cbranch_execnz .LBB15_1
; GFX7-NEXT: ; %bb.2:
; GFX7-NEXT: s_mov_b64 exec, s[6:7]
-; GFX7-NEXT: v_cvt_f16_f32_e32 v4, v5
+; GFX7-NEXT: v_cvt_f16_f32_e32 v5, v5
; GFX7-NEXT: s_mov_b64 s[6:7], 0
-; GFX7-NEXT: v_cvt_f32_f16_e32 v10, v4
+; GFX7-NEXT: v_cvt_f32_f16_e32 v10, v5
; GFX7-NEXT: .LBB15_3: ; %atomicrmw.start
; GFX7-NEXT: ; =>This Loop Header: Depth=1
; GFX7-NEXT: ; Child Loop BB15_4 Depth 2
; GFX7-NEXT: s_waitcnt vmcnt(0)
+; GFX7-NEXT: v_mov_b32_e32 v6, v4
; GFX7-NEXT: v_lshrrev_b32_e32 v4, v7, v6
; GFX7-NEXT: v_cvt_f32_f16_e32 v4, v4
; GFX7-NEXT: v_and_b32_e32 v5, v6, v9
@@ -4995,7 +5000,6 @@ define half @buffer_fat_ptr_agent_atomic_fadd_ret_f16__offset__waterfall__amdgpu
; GFX7-NEXT: s_waitcnt vmcnt(0)
; GFX7-NEXT: v_cmp_eq_u32_e32 vcc, v4, v6
; GFX7-NEXT: s_or_b64 s[6:7], vcc, s[6:7]
-; GFX7-NEXT: v_mov_b32_e32 v6, v4
; GFX7-NEXT: buffer_wbinvl1
; GFX7-NEXT: s_andn2_b64 exec, exec, s[6:7]
; GFX7-NEXT: s_cbranch_execnz .LBB15_3
@@ -5024,18 +5028,20 @@ define half @buffer_fat_ptr_agent_atomic_fadd_ret_f16__offset__waterfall__amdgpu
; GFX6-NEXT: v_cmp_eq_u64_e64 s[4:5], s[10:11], v[2:3]
; GFX6-NEXT: s_and_b64 s[4:5], vcc, s[4:5]
; GFX6-NEXT: s_and_saveexec_b64 s[4:5], s[4:5]
-; GFX6-NEXT: buffer_load_dword v6, v8, s[8:11], 0 offen
+; GFX6-NEXT: buffer_load_dword v4, v8, s[8:11], 0 offen
; GFX6-NEXT: s_xor_b64 exec, exec, s[4:5]
; GFX6-NEXT: s_cbranch_execnz .LBB15_1
; GFX6-NEXT: ; %bb.2:
; GFX6-NEXT: s_mov_b64 exec, s[6:7]
-; GFX6-NEXT: v_cvt_f16_f32_e32 v4, v5
+; GFX6-NEXT: v_cvt_f16_f32_e32 v5, v5
; GFX6-NEXT: s_mov_b64 s[6:7], 0
-; GFX6-NEXT: v_cvt_f32_f16_e32 v10, v4
+; GFX6-NEXT: v_cvt_f32_f16_e32 v10, v5
; GFX6-NEXT: .LBB15_3: ; %atomicrmw.start
; GFX6-NEXT: ; =>This Loop Header: Depth=1
; GFX6-NEXT: ; Child Loop BB15_4 Depth 2
-; GFX6-NEXT: s_waitcnt vmcnt(0) expcnt(0)
+; GFX6-NEXT: s_waitcnt vmcnt(0)
+; GFX6-NEXT: v_mov_b32_e32 v6, v4
+; GFX6-NEXT: s_waitcnt expcnt(0)
; GFX6-NEXT: v_lshrrev_b32_e32 v4, v7, v6
; GFX6-NEXT: v_cvt_f32_f16_e32 v4, v4
; GFX6-NEXT: v_and_b32_e32 v5, v6, v9
@@ -5065,7 +5071,6 @@ define half @buffer_fat_ptr_agent_atomic_fadd_ret_f16__offset__waterfall__amdgpu
; GFX6-NEXT: s_waitcnt vmcnt(0)
; GFX6-NEXT: v_cmp_eq_u32_e32 vcc, v4, v6
; GFX6-NEXT: s_or_b64 s[6:7], vcc, s[6:7]
-; GFX6-NEXT: v_mov_b32_e32 v6, v4
; GFX6-NEXT: buffer_wbinvl1
; GFX6-NEXT: s_andn2_b64 exec, exec, s[6:7]
; GFX6-NEXT: s_cbranch_execnz .LBB15_3
@@ -5093,55 +5098,55 @@ define bfloat @buffer_fat_ptr_agent_atomic_fadd_ret_bf16__offset__amdgpu_no_fine
; GFX12-TRUE16-NEXT: s_wait_bvhcnt 0x0
; GFX12-TRUE16-NEXT: s_wait_kmcnt 0x0
; GFX12-TRUE16-NEXT: s_addk_co_i32 s16, 0x200
-; GFX12-TRUE16-NEXT: v_lshlrev_b32_e32 v5, 16, v0
+; GFX12-TRUE16-NEXT: v_lshlrev_b32_e32 v0, 16, v0
; GFX12-TRUE16-NEXT: s_wait_alu 0xfffe
; GFX12-TRUE16-NEXT: s_and_b32 s4, s16, -4
; GFX12-TRUE16-NEXT: s_wait_alu 0xfffe
-; GFX12-TRUE16-NEXT: v_mov_b32_e32 v4, s4
+; GFX12-TRUE16-NEXT: v_mov_b32_e32 v3, s4
; GFX12-TRUE16-NEXT: s_and_b32 s4, s16, 3
; GFX12-TRUE16-NEXT: s_wait_alu 0xfffe
; GFX12-TRUE16-NEXT: s_lshl_b32 s4, s4, 3
; GFX12-TRUE16-NEXT: s_wait_alu 0xfffe
; GFX12-TRUE16-NEXT: s_lshl_b32 s5, 0xffff, s4
-; GFX12-TRUE16-NEXT: buffer_load_b32 v1, v4, s[0:3], null offen
+; GFX12-TRUE16-NEXT: buffer_load_b32 v1, v3, s[0:3], null offen
; GFX12-TRUE16-NEXT: s_not_b32 s6, s5
; GFX12-TRUE16-NEXT: s_mov_b32 s5, 0
; GFX12-TRUE16-NEXT: .LBB16_1: ; %atomicrmw.start
; GFX12-TRUE16-NEXT: ; =>This Inner Loop Header: Depth=1
; GFX12-TRUE16-NEXT: s_wait_loadcnt 0x0
-; GFX12-TRUE16-NEXT: v_lshrrev_b32_e32 v0, s4, v1
+; GFX12-TRUE16-NEXT: v_mov_b32_e32 v5, v1
; GFX12-TRUE16-NEXT: s_wait_storecnt 0x0
; GFX12-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
-; GFX12-TRUE16-NEXT: v_lshlrev_b32_e32 v0, 16, v0
-; GFX12-TRUE16-NEXT: v_add_f32_e32 v0, v0, v5
-; GFX12-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(SKIP_2) | instid1(VALU_DEP_3)
-; GFX12-TRUE16-NEXT: v_bfe_u32 v2, v0, 16, 1
-; GFX12-TRUE16-NEXT: v_or_b32_e32 v3, 0x400000, v0
-; GFX12-TRUE16-NEXT: v_cmp_u_f32_e32 vcc_lo, v0, v0
-; GFX12-TRUE16-NEXT: v_add3_u32 v2, v2, v0, 0x7fff
+; GFX12-TRUE16-NEXT: v_lshrrev_b32_e32 v1, s4, v5
+; GFX12-TRUE16-NEXT: v_lshlrev_b32_e32 v1, 16, v1
+; GFX12-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
+; GFX12-TRUE16-NEXT: v_add_f32_e32 v1, v1, v0
+; GFX12-TRUE16-NEXT: v_bfe_u32 v2, v1, 16, 1
+; GFX12-TRUE16-NEXT: v_or_b32_e32 v4, 0x400000, v1
+; GFX12-TRUE16-NEXT: v_cmp_u_f32_e32 vcc_lo, v1, v1
+; GFX12-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_3) | instskip(SKIP_1) | instid1(VALU_DEP_1)
+; GFX12-TRUE16-NEXT: v_add3_u32 v2, v2, v1, 0x7fff
; GFX12-TRUE16-NEXT: s_wait_alu 0xfffd
-; GFX12-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(SKIP_1) | instid1(VALU_DEP_2)
-; GFX12-TRUE16-NEXT: v_cndmask_b32_e32 v0, v2, v3, vcc_lo
+; GFX12-TRUE16-NEXT: v_cndmask_b32_e32 v1, v2, v4, vcc_lo
; GFX12-TRUE16-NEXT: v_mov_b16_e32 v2.h, 0
-; GFX12-TRUE16-NEXT: v_mov_b16_e32 v2.l, v0.h
-; GFX12-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(SKIP_1) | instid1(VALU_DEP_1)
-; GFX12-TRUE16-NEXT: v_lshlrev_b32_e32 v0, s4, v2
+; GFX12-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_2) | instskip(NEXT) | instid1(VALU_DEP_1)
+; GFX12-TRUE16-NEXT: v_mov_b16_e32 v2.l, v1.h
+; GFX12-TRUE16-NEXT: v_lshlrev_b32_e32 v1, s4, v2
; GFX12-TRUE16-NEXT: s_wait_alu 0xfffe
-; GFX12-TRUE16-NEXT: v_and_or_b32 v0, v1, s6, v0
-; GFX12-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1)
-; GFX12-TRUE16-NEXT: v_dual_mov_b32 v3, v1 :: v_dual_mov_b32 v2, v0
-; GFX12-TRUE16-NEXT: buffer_atomic_cmpswap_b32 v[2:3], v4, s[0:3], null offen th:TH_ATOMIC_RETURN
+; GFX12-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
+; GFX12-TRUE16-NEXT: v_and_or_b32 v4, v5, s6, v1
+; GFX12-TRUE16-NEXT: v_dual_mov_b32 v1, v4 :: v_dual_mov_b32 v2, v5
+; GFX12-TRUE16-NEXT: buffer_atomic_cmpswap_b32 v[1:2], v3, s[0:3], null offen th:TH_ATOMIC_RETURN
; GFX12-TRUE16-NEXT: s_wait_loadcnt 0x0
; GFX12-TRUE16-NEXT: global_inv scope:SCOPE_DEV
-; GFX12-TRUE16-NEXT: v_cmp_eq_u32_e32 vcc_lo, v2, v1
-; GFX12-TRUE16-NEXT: v_mov_b32_e32 v1, v2
+; GFX12-TRUE16-NEXT: v_cmp_eq_u32_e32 vcc_lo, v1, v5
; GFX12-TRUE16-NEXT: s_or_b32 s5, vcc_lo, s5
; GFX12-TRUE16-NEXT: s_wait_alu 0xfffe
; GFX12-TRUE16-NEXT: s_and_not1_b32 exec_lo, exec_lo, s5
; GFX12-TRUE16-NEXT: s_cbranch_execnz .LBB16_1
; GFX12-TRUE16-NEXT: ; %bb.2: ; %atomicrmw.end
; GFX12-TRUE16-NEXT: s_or_b32 exec_lo, exec_lo, s5
-; GFX12-TRUE16-NEXT: v_lshrrev_b32_e32 v0, s4, v2
+; GFX12-TRUE16-NEXT: v_lshrrev_b32_e32 v0, s4, v1
; GFX12-TRUE16-NEXT: s_wait_loadcnt 0x0
; GFX12-TRUE16-NEXT: s_setpc_b64 s[30:31]
;
@@ -5153,54 +5158,54 @@ define bfloat @buffer_fat_ptr_agent_atomic_fadd_ret_bf16__offset__amdgpu_no_fine
; GFX12-FAKE16-NEXT: s_wait_bvhcnt 0x0
; GFX12-FAKE16-NEXT: s_wait_kmcnt 0x0
; GFX12-FAKE16-NEXT: s_addk_co_i32 s16, 0x200
-; GFX12-FAKE16-NEXT: v_lshlrev_b32_e32 v5, 16, v0
+; GFX12-FAKE16-NEXT: v_lshlrev_b32_e32 v0, 16, v0
; GFX12-FAKE16-NEXT: s_wait_alu 0xfffe
; GFX12-FAKE16-NEXT: s_and_b32 s4, s16, -4
; GFX12-FAKE16-NEXT: s_wait_alu 0xfffe
-; GFX12-FAKE16-NEXT: v_mov_b32_e32 v4, s4
+; GFX12-FAKE16-NEXT: v_mov_b32_e32 v3, s4
; GFX12-FAKE16-NEXT: s_and_b32 s4, s16, 3
; GFX12-FAKE16-NEXT: s_wait_alu 0xfffe
; GFX12-FAKE16-NEXT: s_lshl_b32 s4, s4, 3
; GFX12-FAKE16-NEXT: s_wait_alu 0xfffe
; GFX12-FAKE16-NEXT: s_lshl_b32 s5, 0xffff, s4
-; GFX12-FAKE16-NEXT: buffer_load_b32 v1, v4, s[0:3], null offen
+; GFX12-FAKE16-NEXT: buffer_load_b32 v1, v3, s[0:3], null offen
; GFX12-FAKE16-NEXT: s_not_b32 s6, s5
; GFX12-FAKE16-NEXT: s_mov_b32 s5, 0
; GFX12-FAKE16-NEXT: .LBB16_1: ; %atomicrmw.start
; GFX12-FAKE16-NEXT: ; =>This Inner Loop Header: Depth=1
; GFX12-FAKE16-NEXT: s_wait_loadcnt 0x0
-; GFX12-FAKE16-NEXT: v_lshrrev_b32_e32 v0, s4, v1
+; GFX12-FAKE16-NEXT: v_mov_b32_e32 v5, v1
; GFX12-FAKE16-NEXT: s_wait_storecnt 0x0
; GFX12-FAKE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
-; GFX12-FAKE16-NEXT: v_lshlrev_b32_e32 v0, 16, v0
-; GFX12-FAKE16-NEXT: v_add_f32_e32 v0, v0, v5
-; GFX12-FAKE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(SKIP_2) | instid1(VALU_DEP_3)
-; GFX12-FAKE16-NEXT: v_bfe_u32 v2, v0, 16, 1
-; GFX12-FAKE16-NEXT: v_or_b32_e32 v3, 0x400000, v0
-; GFX12-FAKE16-NEXT: v_cmp_u_f32_e32 vcc_lo, v0, v0
-; GFX12-FAKE16-NEXT: v_add3_u32 v2, v2, v0, 0x7fff
+; GFX12-FAKE16-NEXT: v_lshrrev_b32_e32 v1, s4, v5
+; GFX12-FAKE16-NEXT: v_lshlrev_b32_e32 v1, 16, v1
+; GFX12-FAKE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
+; GFX12-FAKE16-NEXT: v_add_f32_e32 v1, v1, v0
+; GFX12-FAKE16-NEXT: v_bfe_u32 v2, v1, 16, 1
+; GFX12-FAKE16-NEXT: v_or_b32_e32 v4, 0x400000, v1
+; GFX12-FAKE16-NEXT: v_cmp_u_f32_e32 vcc_lo, v1, v1
+; GFX12-FAKE16-NEXT: s_delay_alu instid0(VALU_DEP_3) | instskip(SKIP_1) | instid1(VALU_DEP_1)
+; GFX12-FAKE16-NEXT: v_add3_u32 v2, v2, v1, 0x7fff
; GFX12-FAKE16-NEXT: s_wait_alu 0xfffd
+; GFX12-FAKE16-NEXT: v_cndmask_b32_e32 v1, v2, v4, vcc_lo
; GFX12-FAKE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
-; GFX12-FAKE16-NEXT: v_cndmask_b32_e32 v0, v2, v3, vcc_lo
-; GFX12-FAKE16-NEXT: v_lshrrev_b32_e32 v0, 16, v0
-; GFX12-FAKE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(SKIP_1) | instid1(VALU_DEP_1)
-; GFX12-FAKE16-NEXT: v_lshlrev_b32_e32 v0, s4, v0
+; GFX12-FAKE16-NEXT: v_lshrrev_b32_e32 v1, 16, v1
+; GFX12-FAKE16-NEXT: v_lshlrev_b32_e32 v1, s4, v1
; GFX12-FAKE16-NEXT: s_wait_alu 0xfffe
-; GFX12-FAKE16-NEXT: v_and_or_b32 v0, v1, s6, v0
-; GFX12-FAKE16-NEXT: s_delay_alu instid0(VALU_DEP_1)
-; GFX12-FAKE16-NEXT: v_dual_mov_b32 v3, v1 :: v_dual_mov_b32 v2, v0
-; GFX12-FAKE16-NEXT: buffer_atomic_cmpswap_b32 v[2:3], v4, s[0:3], null offen th:TH_ATOMIC_RETURN
+; GFX12-FAKE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
+; GFX12-FAKE16-NEXT: v_and_or_b32 v4, v5, s6, v1
+; GFX12-FAKE16-NEXT: v_dual_mov_b32 v1, v4 :: v_dual_mov_b32 v2, v5
+; GFX12-FAKE16-NEXT: buffer_atomic_cmpswap_b32 v[1:2], v3, s[0:3], null offen th:TH_ATOMIC_RETURN
; GFX12-FAKE16-NEXT: s_wait_loadcnt 0x0
; GFX12-FAKE16-NEXT: global_inv scope:SCOPE_DEV
-; GFX12-FAKE16-NEXT: v_cmp_eq_u32_e32 vcc_lo, v2, v1
-; GFX12-FAKE16-NEXT: v_mov_b32_e32 v1, v2
+; GFX12-FAKE16-NEXT: v_cmp_eq_u32_e32 vcc_lo, v1, v5
; GFX12-FAKE16-NEXT: s_or_b32 s5, vcc_lo, s5
; GFX12-FAKE16-NEXT: s_wait_alu 0xfffe
; GFX12-FAKE16-NEXT: s_and_not1_b32 exec_lo, exec_lo, s5
; GFX12-FAKE16-NEXT: s_cbranch_execnz .LBB16_1
; GFX12-FAKE16-NEXT: ; %bb.2: ; %atomicrmw.end
; GFX12-FAKE16-NEXT: s_or_b32 exec_lo, exec_lo, s5
-; GFX12-FAKE16-NEXT: v_lshrrev_b32_e32 v0, s4, v2
+; GFX12-FAKE16-NEXT: v_lshrrev_b32_e32 v0, s4, v1
; GFX12-FAKE16-NEXT: s_wait_loadcnt 0x0
; GFX12-FAKE16-NEXT: s_setpc_b64 s[30:31]
;
@@ -5209,36 +5214,36 @@ define bfloat @buffer_fat_ptr_agent_atomic_fadd_ret_bf16__offset__amdgpu_no_fine
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX942-NEXT: s_addk_i32 s16, 0x200
; GFX942-NEXT: s_and_b32 s4, s16, -4
-; GFX942-NEXT: v_mov_b32_e32 v4, s4
-; GFX942-NEXT: buffer_load_dword v1, v4, s[0:3], 0 offen
+; GFX942-NEXT: v_mov_b32_e32 v1, s4
+; GFX942-NEXT: buffer_load_dword v2, v1, s[0:3], 0 offen
; GFX942-NEXT: s_and_b32 s4, s16, 3
; GFX942-NEXT: s_lshl_b32 s6, s4, 3
; GFX942-NEXT: s_lshl_b32 s4, 0xffff, s6
; GFX942-NEXT: s_not_b32 s7, s4
; GFX942-NEXT: s_mov_b64 s[4:5], 0
-; GFX942-NEXT: v_lshlrev_b32_e32 v5, 16, v0
+; GFX942-NEXT: v_lshlrev_b32_e32 v0, 16, v0
; GFX942-NEXT: s_movk_i32 s8, 0x7fff
; GFX942-NEXT: .LBB16_1: ; %atomicrmw.start
; GFX942-NEXT: ; =>This Inner Loop Header: Depth=1
; GFX942-NEXT: s_waitcnt vmcnt(0)
-; GFX942-NEXT: v_lshrrev_b32_sdwa v0, s6, v1 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:DWORD
+; GFX942-NEXT: v_mov_b32_e32 v5, v2
+; GFX942-NEXT: v_lshrrev_b32_sdwa v2, s6, v5 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:DWORD
; GFX942-NEXT: buffer_wbl2 sc1
-; GFX942-NEXT: v_add_f32_e32 v0, v0, v5
-; GFX942-NEXT: v_bfe_u32 v2, v0, 16, 1
-; GFX942-NEXT: v_or_b32_e32 v3, 0x400000, v0
-; GFX942-NEXT: v_add3_u32 v2, v2, v0, s8
-; GFX942-NEXT: v_cmp_u_f32_e32 vcc, v0, v0
+; GFX942-NEXT: v_add_f32_e32 v2, v2, v0
+; GFX942-NEXT: v_bfe_u32 v3, v2, 16, 1
+; GFX942-NEXT: v_or_b32_e32 v4, 0x400000, v2
+; GFX942-NEXT: v_add3_u32 v3, v3, v2, s8
+; GFX942-NEXT: v_cmp_u_f32_e32 vcc, v2, v2
; GFX942-NEXT: s_nop 1
-; GFX942-NEXT: v_cndmask_b32_e32 v0, v2, v3, vcc
-; GFX942-NEXT: v_lshlrev_b32_sdwa v0, s6, v0 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:WORD_1
-; GFX942-NEXT: v_and_or_b32 v0, v1, s7, v0
-; GFX942-NEXT: v_mov_b64_e32 v[2:3], v[0:1]
-; GFX942-NEXT: buffer_atomic_cmpswap v[2:3], v4, s[0:3], 0 offen sc0
+; GFX942-NEXT: v_cndmask_b32_e32 v2, v3, v4, vcc
+; GFX942-NEXT: v_lshlrev_b32_sdwa v2, s6, v2 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:WORD_1
+; GFX942-NEXT: v_and_or_b32 v4, v5, s7, v2
+; GFX942-NEXT: v_mov_b64_e32 v[2:3], v[4:5]
+; GFX942-NEXT: buffer_atomic_cmpswap v[2:3], v1, s[0:3], 0 offen sc0
; GFX942-NEXT: s_waitcnt vmcnt(0)
; GFX942-NEXT: buffer_inv sc1
-; GFX942-NEXT: v_cmp_eq_u32_e32 vcc, v2, v1
+; GFX942-NEXT: v_cmp_eq_u32_e32 vcc, v2, v5
; GFX942-NEXT: s_or_b64 s[4:5], vcc, s[4:5]
-; GFX942-NEXT: v_mov_b32_e32 v1, v2
; GFX942-NEXT: s_andn2_b64 exec, exec, s[4:5]
; GFX942-NEXT: s_cbranch_execnz .LBB16_1
; GFX942-NEXT: ; %bb.2: ; %atomicrmw.end
@@ -5250,147 +5255,147 @@ define bfloat @buffer_fat_ptr_agent_atomic_fadd_ret_bf16__offset__amdgpu_no_fine
; GFX11-TRUE16: ; %bb.0:
; GFX11-TRUE16-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX11-TRUE16-NEXT: s_addk_i32 s16, 0x200
-; GFX11-TRUE16-NEXT: v_lshlrev_b32_e32 v5, 16, v0
+; GFX11-TRUE16-NEXT: v_lshlrev_b32_e32 v0, 16, v0
; GFX11-TRUE16-NEXT: s_and_b32 s4, s16, -4
; GFX11-TRUE16-NEXT: s_delay_alu instid0(SALU_CYCLE_1) | instskip(SKIP_1) | instid1(SALU_CYCLE_1)
-; GFX11-TRUE16-NEXT: v_mov_b32_e32 v4, s4
+; GFX11-TRUE16-NEXT: v_mov_b32_e32 v3, s4
; GFX11-TRUE16-NEXT: s_and_b32 s4, s16, 3
; GFX11-TRUE16-NEXT: s_lshl_b32 s4, s4, 3
; GFX11-TRUE16-NEXT: s_delay_alu instid0(SALU_CYCLE_1)
; GFX11-TRUE16-NEXT: s_lshl_b32 s5, 0xffff, s4
-; GFX11-TRUE16-NEXT: buffer_load_b32 v1, v4, s[0:3], 0 offen
+; GFX11-TRUE16-NEXT: buffer_load_b32 v1, v3, s[0:3], 0 offen
; GFX11-TRUE16-NEXT: s_not_b32 s6, s5
; GFX11-TRUE16-NEXT: s_mov_b32 s5, 0
; GFX11-TRUE16-NEXT: .p2align 6
; GFX11-TRUE16-NEXT: .LBB16_1: ; %atomicrmw.start
; GFX11-TRUE16-NEXT: ; =>This Inner Loop Header: Depth=1
; GFX11-TRUE16-NEXT: s_waitcnt vmcnt(0)
-; GFX11-TRUE16-NEXT: v_lshrrev_b32_e32 v0, s4, v1
+; GFX11-TRUE16-NEXT: v_mov_b32_e32 v5, v1
; GFX11-TRUE16-NEXT: s_waitcnt_vscnt null, 0x0
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
-; GFX11-TRUE16-NEXT: v_lshlrev_b32_e32 v0, 16, v0
-; GFX11-TRUE16-NEXT: v_add_f32_e32 v0, v0, v5
-; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(SKIP_2) | instid1(VALU_DEP_3)
-; GFX11-TRUE16-NEXT: v_bfe_u32 v2, v0, 16, 1
-; GFX11-TRUE16-NEXT: v_or_b32_e32 v3, 0x400000, v0
-; GFX11-TRUE16-NEXT: v_cmp_u_f32_e32 vcc_lo, v0, v0
-; GFX11-TRUE16-NEXT: v_add3_u32 v2, v2, v0, 0x7fff
-; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(SKIP_1) | instid1(VALU_DEP_2)
-; GFX11-TRUE16-NEXT: v_cndmask_b32_e32 v0, v2, v3, vcc_lo
+; GFX11-TRUE16-NEXT: v_lshrrev_b32_e32 v1, s4, v5
+; GFX11-TRUE16-NEXT: v_lshlrev_b32_e32 v1, 16, v1
+; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
+; GFX11-TRUE16-NEXT: v_add_f32_e32 v1, v1, v0
+; GFX11-TRUE16-NEXT: v_bfe_u32 v2, v1, 16, 1
+; GFX11-TRUE16-NEXT: v_or_b32_e32 v4, 0x400000, v1
+; GFX11-TRUE16-NEXT: v_cmp_u_f32_e32 vcc_lo, v1, v1
+; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_3) | instskip(NEXT) | instid1(VALU_DEP_1)
+; GFX11-TRUE16-NEXT: v_add3_u32 v2, v2, v1, 0x7fff
+; GFX11-TRUE16-NEXT: v_cndmask_b32_e32 v1, v2, v4, vcc_lo
; GFX11-TRUE16-NEXT: v_mov_b16_e32 v2.h, 0
-; GFX11-TRUE16-NEXT: v_mov_b16_e32 v2.l, v0.h
+; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_2) | instskip(NEXT) | instid1(VALU_DEP_1)
+; GFX11-TRUE16-NEXT: v_mov_b16_e32 v2.l, v1.h
+; GFX11-TRUE16-NEXT: v_lshlrev_b32_e32 v1, s4, v2
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
-; GFX11-TRUE16-NEXT: v_lshlrev_b32_e32 v0, s4, v2
-; GFX11-TRUE16-NEXT: v_and_or_b32 v0, v1, s6, v0
-; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1)
-; GFX11-TRUE16-NEXT: v_dual_mov_b32 v3, v1 :: v_dual_mov_b32 v2, v0
-; GFX11-TRUE16-NEXT: buffer_atomic_cmpswap_b32 v[2:3], v4, s[0:3], 0 offen glc
+; GFX11-TRUE16-NEXT: v_and_or_b32 v4, v5, s6, v1
+; GFX11-TRUE16-NEXT: v_dual_mov_b32 v1, v4 :: v_dual_mov_b32 v2, v5
+; GFX11-TRUE16-NEXT: buffer_atomic_cmpswap_b32 v[1:2], v3, s[0:3], 0 offen glc
; GFX11-TRUE16-NEXT: s_waitcnt vmcnt(0)
; GFX11-TRUE16-NEXT: buffer_gl1_inv
; GFX11-TRUE16-NEXT: buffer_gl0_inv
-; GFX11-TRUE16-NEXT: v_cmp_eq_u32_e32 vcc_lo, v2, v1
-; GFX11-TRUE16-NEXT: v_mov_b32_e32 v1, v2
+; GFX11-TRUE16-NEXT: v_cmp_eq_u32_e32 vcc_lo, v1, v5
; GFX11-TRUE16-NEXT: s_or_b32 s5, vcc_lo, s5
; GFX11-TRUE16-NEXT: s_delay_alu instid0(SALU_CYCLE_1)
; GFX11-TRUE16-NEXT: s_and_not1_b32 exec_lo, exec_lo, s5
; GFX11-TRUE16-NEXT: s_cbranch_execnz .LBB16_1
; GFX11-TRUE16-NEXT: ; %bb.2: ; %atomicrmw.end
; GFX11-TRUE16-NEXT: s_or_b32 exec_lo, exec_lo, s5
-; GFX11-TRUE16-NEXT: v_lshrrev_b32_e32 v0, s4, v2
+; GFX11-TRUE16-NEXT: v_lshrrev_b32_e32 v0, s4, v1
; GFX11-TRUE16-NEXT: s_setpc_b64 s[30:31]
;
; GFX11-FAKE16-LABEL: buffer_fat_ptr_agent_atomic_fadd_ret_bf16__offset__amdgpu_no_fine_grained_memory:
; GFX11-FAKE16: ; %bb.0:
; GFX11-FAKE16-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX11-FAKE16-NEXT: s_addk_i32 s16, 0x200
-; GFX11-FAKE16-NEXT: v_lshlrev_b32_e32 v5, 16, v0
+; GFX11-FAKE16-NEXT: v_lshlrev_b32_e32 v0, 16, v0
; GFX11-FAKE16-NEXT: s_and_b32 s4, s16, -4
; GFX11-FAKE16-NEXT: s_delay_alu instid0(SALU_CYCLE_1) | instskip(SKIP_1) | instid1(SALU_CYCLE_1)
-; GFX11-FAKE16-NEXT: v_mov_b32_e32 v4, s4
+; GFX11-FAKE16-NEXT: v_mov_b32_e32 v3, s4
; GFX11-FAKE16-NEXT: s_and_b32 s4, s16, 3
; GFX11-FAKE16-NEXT: s_lshl_b32 s4, s4, 3
; GFX11-FAKE16-NEXT: s_delay_alu instid0(SALU_CYCLE_1)
; GFX11-FAKE16-NEXT: s_lshl_b32 s5, 0xffff, s4
-; GFX11-FAKE16-NEXT: buffer_load_b32 v1, v4, s[0:3], 0 offen
+; GFX11-FAKE16-NEXT: buffer_load_b32 v1, v3, s[0:3], 0 offen
; GFX11-FAKE16-NEXT: s_not_b32 s6, s5
; GFX11-FAKE16-NEXT: s_mov_b32 s5, 0
; GFX11-FAKE16-NEXT: .p2align 6
; GFX11-FAKE16-NEXT: .LBB16_1: ; %atomicrmw.start
; GFX11-FAKE16-NEXT: ; =>This Inner Loop Header: Depth=1
; GFX11-FAKE16-NEXT: s_waitcnt vmcnt(0)
-; GFX11-FAKE16-NEXT: v_lshrrev_b32_e32 v0, s4, v1
+; GFX11-FAKE16-NEXT: v_mov_b32_e32 v5, v1
; GFX11-FAKE16-NEXT: s_waitcnt_vscnt null, 0x0
; GFX11-FAKE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
-; GFX11-FAKE16-NEXT: v_lshlrev_b32_e32 v0, 16, v0
-; GFX11-FAKE16-NEXT: v_add_f32_e32 v0, v0, v5
-; GFX11-FAKE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(SKIP_2) | instid1(VALU_DEP_3)
-; GFX11-FAKE16-NEXT: v_bfe_u32 v2, v0, 16, 1
-; GFX11-FAKE16-NEXT: v_or_b32_e32 v3, 0x400000, v0
-; GFX11-FAKE16-NEXT: v_cmp_u_f32_e32 vcc_lo, v0, v0
-; GFX11-FAKE16-NEXT: v_add3_u32 v2, v2, v0, 0x7fff
+; GFX11-FAKE16-NEXT: v_lshrrev_b32_e32 v1, s4, v5
+; GFX11-FAKE16-NEXT: v_lshlrev_b32_e32 v1, 16, v1
; GFX11-FAKE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
-; GFX11-FAKE16-NEXT: v_cndmask_b32_e32 v0, v2, v3, vcc_lo
-; GFX11-FAKE16-NEXT: v_lshrrev_b32_e32 v0, 16, v0
+; GFX11-FAKE16-NEXT: v_add_f32_e32 v1, v1, v0
+; GFX11-FAKE16-NEXT: v_bfe_u32 v2, v1, 16, 1
+; GFX11-FAKE16-NEXT: v_or_b32_e32 v4, 0x400000, v1
+; GFX11-FAKE16-NEXT: v_cmp_u_f32_e32 vcc_lo, v1, v1
+; GFX11-FAKE16-NEXT: s_delay_alu instid0(VALU_DEP_3) | instskip(NEXT) | instid1(VALU_DEP_1)
+; GFX11-FAKE16-NEXT: v_add3_u32 v2, v2, v1, 0x7fff
+; GFX11-FAKE16-NEXT: v_cndmask_b32_e32 v1, v2, v4, vcc_lo
; GFX11-FAKE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
-; GFX11-FAKE16-NEXT: v_lshlrev_b32_e32 v0, s4, v0
-; GFX11-FAKE16-NEXT: v_and_or_b32 v0, v1, s6, v0
-; GFX11-FAKE16-NEXT: s_delay_alu instid0(VALU_DEP_1)
-; GFX11-FAKE16-NEXT: v_dual_mov_b32 v3, v1 :: v_dual_mov_b32 v2, v0
-; GFX11-FAKE16-NEXT: buffer_atomic_cmpswap_b32 v[2:3], v4, s[0:3], 0 offen glc
+; GFX11-FAKE16-NEXT: v_lshrrev_b32_e32 v1, 16, v1
+; GFX11-FAKE16-NEXT: v_lshlrev_b32_e32 v1, s4, v1
+; GFX11-FAKE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
+; GFX11-FAKE16-NEXT: v_and_or_b32 v4, v5, s6, v1
+; GFX11-FAKE16-NEXT: v_dual_mov_b32 v1, v4 :: v_dual_mov_b32 v2, v5
+; GFX11-FAKE16-NEXT: buffer_atomic_cmpswap_b32 v[1:2], v3, s[0:3], 0 offen glc
; GFX11-FAKE16-NEXT: s_waitcnt vmcnt(0)
; GFX11-FAKE16-NEXT: buffer_gl1_inv
; GFX11-FAKE16-NEXT: buffer_gl0_inv
-; GFX11-FAKE16-NEXT: v_cmp_eq_u32_e32 vcc_lo, v2, v1
-; GFX11-FAKE16-NEXT: v_mov_b32_e32 v1, v2
+; GFX11-FAKE16-NEXT: v_cmp_eq_u32_e32 vcc_lo, v1, v5
; GFX11-FAKE16-NEXT: s_or_b32 s5, vcc_lo, s5
; GFX11-FAKE16-NEXT: s_delay_alu instid0(SALU_CYCLE_1)
; GFX11-FAKE16-NEXT: s_and_not1_b32 exec_lo, exec_lo, s5
; GFX11-FAKE16-NEXT: s_cbranch_execnz .LBB16_1
; GFX11-FAKE16-NEXT: ; %bb.2: ; %atomicrmw.end
; GFX11-FAKE16-NEXT: s_or_b32 exec_lo, exec_lo, s5
-; GFX11-FAKE16-NEXT: v_lshrrev_b32_e32 v0, s4, v2
+; GFX11-FAKE16-NEXT: v_lshrrev_b32_e32 v0, s4, v1
; GFX11-FAKE16-NEXT: s_setpc_b64 s[30:31]
;
; GFX10-LABEL: buffer_fat_ptr_agent_atomic_fadd_ret_bf16__offset__amdgpu_no_fine_grained_memory:
; GFX10: ; %bb.0:
; GFX10-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX10-NEXT: s_addk_i32 s20, 0x200
-; GFX10-NEXT: v_lshlrev_b32_e32 v5, 16, v0
+; GFX10-NEXT: v_lshlrev_b32_e32 v0, 16, v0
; GFX10-NEXT: s_and_b32 s4, s20, -4
-; GFX10-NEXT: v_mov_b32_e32 v4, s4
+; GFX10-NEXT: v_mov_b32_e32 v3, s4
; GFX10-NEXT: s_and_b32 s4, s20, 3
; GFX10-NEXT: s_lshl_b32 s4, s4, 3
; GFX10-NEXT: s_lshl_b32 s5, 0xffff, s4
-; GFX10-NEXT: buffer_load_dword v1, v4, s[16:19], 0 offen
+; GFX10-NEXT: buffer_load_dword v1, v3, s[16:19], 0 offen
; GFX10-NEXT: s_not_b32 s6, s5
; GFX10-NEXT: s_mov_b32 s5, 0
; GFX10-NEXT: .LBB16_1: ; %atomicrmw.start
; GFX10-NEXT: ; =>This Inner Loop Header: Depth=1
; GFX10-NEXT: s_waitcnt vmcnt(0)
-; GFX10-NEXT: v_lshrrev_b32_sdwa v0, s4, v1 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:DWORD
+; GFX10-NEXT: v_mov_b32_e32 v5, v1
; GFX10-NEXT: s_waitcnt_vscnt null, 0x0
-; GFX10-NEXT: v_add_f32_e32 v0, v0, v5
-; GFX10-NEXT: v_bfe_u32 v2, v0, 16, 1
-; GFX10-NEXT: v_or_b32_e32 v3, 0x400000, v0
-; GFX10-NEXT: v_cmp_u_f32_e32 vcc_lo, v0, v0
-; GFX10-NEXT: v_add3_u32 v2, v2, v0, 0x7fff
-; GFX10-NEXT: v_cndmask_b32_e32 v0, v2, v3, vcc_lo
-; GFX10-NEXT: v_lshlrev_b32_sdwa v0, s4, v0 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:WORD_1
-; GFX10-NEXT: v_and_or_b32 v0, v1, s6, v0
-; GFX10-NEXT: v_mov_b32_e32 v3, v1
-; GFX10-NEXT: v_mov_b32_e32 v2, v0
-; GFX10-NEXT: buffer_atomic_cmpswap v[2:3], v4, s[16:19], 0 offen glc
+; GFX10-NEXT: v_lshrrev_b32_sdwa v1, s4, v5 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:DWORD
+; GFX10-NEXT: v_add_f32_e32 v1, v1, v0
+; GFX10-NEXT: v_bfe_u32 v2, v1, 16, 1
+; GFX10-NEXT: v_or_b32_e32 v4, 0x400000, v1
+; GFX10-NEXT: v_cmp_u_f32_e32 vcc_lo, v1, v1
+; GFX10-NEXT: v_add3_u32 v2, v2, v1, 0x7fff
+; GFX10-NEXT: v_cndmask_b32_e32 v1, v2, v4, vcc_lo
+; GFX10-NEXT: v_lshlrev_b32_sdwa v1, s4, v1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:WORD_1
+; GFX10-NEXT: v_and_or_b32 v4, v5, s6, v1
+; GFX10-NEXT: v_mov_b32_e32 v1, v4
+; GFX10-NEXT: v_mov_b32_e32 v2, v5
+; GFX10-NEXT: buffer_atomic_cmpswap v[1:2], v3, s[16:19], 0 offen glc
; GFX10-NEXT: s_waitcnt vmcnt(0)
; GFX10-NEXT: buffer_gl1_inv
; GFX10-NEXT: buffer_gl0_inv
-; GFX10-NEXT: v_cmp_eq_u32_e32 vcc_lo, v2, v1
-; GFX10-NEXT: v_mov_b32_e32 v1, v2
+; GFX10-NEXT: v_cmp_eq_u32_e32 vcc_lo, v1, v5
; GFX10-NEXT: s_or_b32 s5, vcc_lo, s5
; GFX10-NEXT: s_andn2_b32 exec_lo, exec_lo, s5
; GFX10-NEXT: s_cbranch_execnz .LBB16_1
; GFX10-NEXT: ; %bb.2: ; %atomicrmw.end
; GFX10-NEXT: s_or_b32 exec_lo, exec_lo, s5
-; GFX10-NEXT: v_lshrrev_b32_e32 v0, s4, v2
+; GFX10-NEXT: v_lshrrev_b32_e32 v0, s4, v1
; GFX10-NEXT: s_setpc_b64 s[30:31]
;
; GFX90A-LABEL: buffer_fat_ptr_agent_atomic_fadd_ret_bf16__offset__amdgpu_no_fine_grained_memory:
@@ -5398,34 +5403,34 @@ define bfloat @buffer_fat_ptr_agent_atomic_fadd_ret_bf16__offset__amdgpu_no_fine
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX90A-NEXT: s_addk_i32 s20, 0x200
; GFX90A-NEXT: s_and_b32 s4, s20, -4
-; GFX90A-NEXT: v_mov_b32_e32 v4, s4
-; GFX90A-NEXT: buffer_load_dword v1, v4, s[16:19], 0 offen
+; GFX90A-NEXT: v_mov_b32_e32 v1, s4
+; GFX90A-NEXT: buffer_load_dword v2, v1, s[16:19], 0 offen
; GFX90A-NEXT: s_and_b32 s4, s20, 3
; GFX90A-NEXT: s_lshl_b32 s6, s4, 3
; GFX90A-NEXT: s_lshl_b32 s4, 0xffff, s6
; GFX90A-NEXT: s_not_b32 s7, s4
; GFX90A-NEXT: s_mov_b64 s[4:5], 0
-; GFX90A-NEXT: v_lshlrev_b32_e32 v5, 16, v0
+; GFX90A-NEXT: v_lshlrev_b32_e32 v0, 16, v0
; GFX90A-NEXT: s_movk_i32 s8, 0x7fff
; GFX90A-NEXT: .LBB16_1: ; %atomicrmw.start
; GFX90A-NEXT: ; =>This Inner Loop Header: Depth=1
; GFX90A-NEXT: s_waitcnt vmcnt(0)
-; GFX90A-NEXT: v_lshrrev_b32_sdwa v0, s6, v1 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:DWORD
-; GFX90A-NEXT: v_add_f32_e32 v0, v0, v5
-; GFX90A-NEXT: v_bfe_u32 v2, v0, 16, 1
-; GFX90A-NEXT: v_or_b32_e32 v3, 0x400000, v0
-; GFX90A-NEXT: v_add3_u32 v2, v2, v0, s8
-; GFX90A-NEXT: v_cmp_u_f32_e32 vcc, v0, v0
-; GFX90A-NEXT: v_cndmask_b32_e32 v0, v2, v3, vcc
-; GFX90A-NEXT: v_lshlrev_b32_sdwa v0, s6, v0 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:WORD_1
-; GFX90A-NEXT: v_and_or_b32 v0, v1, s7, v0
-; GFX90A-NEXT: v_pk_mov_b32 v[2:3], v[0:1], v[0:1] op_sel:[0,1]
-; GFX90A-NEXT: buffer_atomic_cmpswap v[2:3], v4, s[16:19], 0 offen glc
+; GFX90A-NEXT: v_mov_b32_e32 v5, v2
+; GFX90A-NEXT: v_lshrrev_b32_sdwa v2, s6, v5 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:DWORD
+; GFX90A-NEXT: v_add_f32_e32 v2, v2, v0
+; GFX90A-NEXT: v_bfe_u32 v3, v2, 16, 1
+; GFX90A-NEXT: v_or_b32_e32 v4, 0x400000, v2
+; GFX90A-NEXT: v_add3_u32 v3, v3, v2, s8
+; GFX90A-NEXT: v_cmp_u_f32_e32 vcc, v2, v2
+; GFX90A-NEXT: v_cndmask_b32_e32 v2, v3, v4, vcc
+; GFX90A-NEXT: v_lshlrev_b32_sdwa v2, s6, v2 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:WORD_1
+; GFX90A-NEXT: v_and_or_b32 v4, v5, s7, v2
+; GFX90A-NEXT: v_pk_mov_b32 v[2:3], v[4:5], v[4:5] op_sel:[0,1]
+; GFX90A-NEXT: buffer_atomic_cmpswap v[2:3], v1, s[16:19], 0 offen glc
; GFX90A-NEXT: s_waitcnt vmcnt(0)
; GFX90A-NEXT: buffer_wbinvl1
-; GFX90A-NEXT: v_cmp_eq_u32_e32 vcc, v2, v1
+; GFX90A-NEXT: v_cmp_eq_u32_e32 vcc, v2, v5
; GFX90A-NEXT: s_or_b64 s[4:5], vcc, s[4:5]
-; GFX90A-NEXT: v_mov_b32_e32 v1, v2
; GFX90A-NEXT: s_andn2_b64 exec, exec, s[4:5]
; GFX90A-NEXT: s_cbranch_execnz .LBB16_1
; GFX90A-NEXT: ; %bb.2: ; %atomicrmw.end
@@ -5438,40 +5443,40 @@ define bfloat @buffer_fat_ptr_agent_atomic_fadd_ret_bf16__offset__amdgpu_no_fine
; GFX908-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX908-NEXT: s_addk_i32 s20, 0x200
; GFX908-NEXT: s_and_b32 s4, s20, -4
-; GFX908-NEXT: v_mov_b32_e32 v4, s4
-; GFX908-NEXT: buffer_load_dword v1, v4, s[16:19], 0 offen
+; GFX908-NEXT: v_mov_b32_e32 v3, s4
+; GFX908-NEXT: buffer_load_dword v1, v3, s[16:19], 0 offen
; GFX908-NEXT: s_and_b32 s4, s20, 3
; GFX908-NEXT: s_lshl_b32 s6, s4, 3
; GFX908-NEXT: s_lshl_b32 s4, 0xffff, s6
; GFX908-NEXT: s_not_b32 s7, s4
; GFX908-NEXT: s_mov_b64 s[4:5], 0
-; GFX908-NEXT: v_lshlrev_b32_e32 v5, 16, v0
+; GFX908-NEXT: v_lshlrev_b32_e32 v0, 16, v0
; GFX908-NEXT: s_movk_i32 s8, 0x7fff
; GFX908-NEXT: .LBB16_1: ; %atomicrmw.start
; GFX908-NEXT: ; =>This Inner Loop Header: Depth=1
; GFX908-NEXT: s_waitcnt vmcnt(0)
-; GFX908-NEXT: v_lshrrev_b32_sdwa v0, s6, v1 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:DWORD
-; GFX908-NEXT: v_add_f32_e32 v0, v0, v5
-; GFX908-NEXT: v_bfe_u32 v2, v0, 16, 1
-; GFX908-NEXT: v_or_b32_e32 v3, 0x400000, v0
-; GFX908-NEXT: v_add3_u32 v2, v2, v0, s8
-; GFX908-NEXT: v_cmp_u_f32_e32 vcc, v0, v0
-; GFX908-NEXT: v_cndmask_b32_e32 v0, v2, v3, vcc
-; GFX908-NEXT: v_lshlrev_b32_sdwa v0, s6, v0 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:WORD_1
-; GFX908-NEXT: v_and_or_b32 v0, v1, s7, v0
-; GFX908-NEXT: v_mov_b32_e32 v3, v1
-; GFX908-NEXT: v_mov_b32_e32 v2, v0
-; GFX908-NEXT: buffer_atomic_cmpswap v[2:3], v4, s[16:19], 0 offen glc
+; GFX908-NEXT: v_mov_b32_e32 v5, v1
+; GFX908-NEXT: v_lshrrev_b32_sdwa v1, s6, v5 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:DWORD
+; GFX908-NEXT: v_add_f32_e32 v1, v1, v0
+; GFX908-NEXT: v_bfe_u32 v2, v1, 16, 1
+; GFX908-NEXT: v_or_b32_e32 v4, 0x400000, v1
+; GFX908-NEXT: v_add3_u32 v2, v2, v1, s8
+; GFX908-NEXT: v_cmp_u_f32_e32 vcc, v1, v1
+; GFX908-NEXT: v_cndmask_b32_e32 v1, v2, v4, vcc
+; GFX908-NEXT: v_lshlrev_b32_sdwa v1, s6, v1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:WORD_1
+; GFX908-NEXT: v_and_or_b32 v4, v5, s7, v1
+; GFX908-NEXT: v_mov_b32_e32 v1, v4
+; GFX908-NEXT: v_mov_b32_e32 v2, v5
+; GFX908-NEXT: buffer_atomic_cmpswap v[1:2], v3, s[16:19], 0 offen glc
; GFX908-NEXT: s_waitcnt vmcnt(0)
; GFX908-NEXT: buffer_wbinvl1
-; GFX908-NEXT: v_cmp_eq_u32_e32 vcc, v2, v1
+; GFX908-NEXT: v_cmp_eq_u32_e32 vcc, v1, v5
; GFX908-NEXT: s_or_b64 s[4:5], vcc, s[4:5]
-; GFX908-NEXT: v_mov_b32_e32 v1, v2
; GFX908-NEXT: s_andn2_b64 exec, exec, s[4:5]
; GFX908-NEXT: s_cbranch_execnz .LBB16_1
; GFX908-NEXT: ; %bb.2: ; %atomicrmw.end
; GFX908-NEXT: s_or_b64 exec, exec, s[4:5]
-; GFX908-NEXT: v_lshrrev_b32_e32 v0, s6, v2
+; GFX908-NEXT: v_lshrrev_b32_e32 v0, s6, v1
; GFX908-NEXT: s_setpc_b64 s[30:31]
;
; GFX8-LABEL: buffer_fat_ptr_agent_atomic_fadd_ret_bf16__offset__amdgpu_no_fine_grained_memory:
@@ -5479,42 +5484,42 @@ define bfloat @buffer_fat_ptr_agent_atomic_fadd_ret_bf16__offset__amdgpu_no_fine
; GFX8-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX8-NEXT: s_addk_i32 s20, 0x200
; GFX8-NEXT: s_and_b32 s4, s20, -4
-; GFX8-NEXT: v_mov_b32_e32 v4, s4
-; GFX8-NEXT: buffer_load_dword v1, v4, s[16:19], 0 offen
+; GFX8-NEXT: v_mov_b32_e32 v3, s4
+; GFX8-NEXT: buffer_load_dword v1, v3, s[16:19], 0 offen
; GFX8-NEXT: s_and_b32 s4, s20, 3
; GFX8-NEXT: s_lshl_b32 s6, s4, 3
; GFX8-NEXT: s_lshl_b32 s4, 0xffff, s6
; GFX8-NEXT: s_not_b32 s7, s4
; GFX8-NEXT: s_mov_b64 s[4:5], 0
-; GFX8-NEXT: v_lshlrev_b32_e32 v5, 16, v0
+; GFX8-NEXT: v_lshlrev_b32_e32 v0, 16, v0
; GFX8-NEXT: .LBB16_1: ; %atomicrmw.start
; GFX8-NEXT: ; =>This Inner Loop Header: Depth=1
-; GFX8-NEXT: v_mov_b32_e32 v0, s6
+; GFX8-NEXT: v_mov_b32_e32 v2, s6
; GFX8-NEXT: s_waitcnt vmcnt(0)
-; GFX8-NEXT: v_lshrrev_b32_sdwa v3, v0, v1 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:DWORD
-; GFX8-NEXT: v_add_f32_e32 v3, v3, v5
-; GFX8-NEXT: v_bfe_u32 v6, v3, 16, 1
-; GFX8-NEXT: v_add_u32_e32 v6, vcc, v6, v3
+; GFX8-NEXT: v_mov_b32_e32 v5, v1
+; GFX8-NEXT: v_lshrrev_b32_sdwa v1, v2, v5 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:DWORD
+; GFX8-NEXT: v_add_f32_e32 v1, v1, v0
+; GFX8-NEXT: v_bfe_u32 v6, v1, 16, 1
+; GFX8-NEXT: v_add_u32_e32 v6, vcc, v6, v1
; GFX8-NEXT: v_add_u32_e32 v6, vcc, 0x7fff, v6
-; GFX8-NEXT: v_or_b32_e32 v7, 0x400000, v3
-; GFX8-NEXT: v_cmp_u_f32_e32 vcc, v3, v3
-; GFX8-NEXT: v_cndmask_b32_e32 v3, v6, v7, vcc
-; GFX8-NEXT: v_and_b32_e32 v2, s7, v1
-; GFX8-NEXT: v_lshlrev_b32_sdwa v0, v0, v3 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:WORD_1
-; GFX8-NEXT: v_or_b32_e32 v0, v2, v0
-; GFX8-NEXT: v_mov_b32_e32 v3, v1
-; GFX8-NEXT: v_mov_b32_e32 v2, v0
-; GFX8-NEXT: buffer_atomic_cmpswap v[2:3], v4, s[16:19], 0 offen glc
+; GFX8-NEXT: v_or_b32_e32 v7, 0x400000, v1
+; GFX8-NEXT: v_cmp_u_f32_e32 vcc, v1, v1
+; GFX8-NEXT: v_cndmask_b32_e32 v1, v6, v7, vcc
+; GFX8-NEXT: v_and_b32_e32 v4, s7, v5
+; GFX8-NEXT: v_lshlrev_b32_sdwa v1, v2, v1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:WORD_1
+; GFX8-NEXT: v_or_b32_e32 v4, v4, v1
+; GFX8-NEXT: v_mov_b32_e32 v1, v4
+; GFX8-NEXT: v_mov_b32_e32 v2, v5
+; GFX8-NEXT: buffer_atomic_cmpswap v[1:2], v3, s[16:19], 0 offen glc
; GFX8-NEXT: s_waitcnt vmcnt(0)
; GFX8-NEXT: buffer_wbinvl1
-; GFX8-NEXT: v_cmp_eq_u32_e32 vcc, v2, v1
+; GFX8-NEXT: v_cmp_eq_u32_e32 vcc, v1, v5
; GFX8-NEXT: s_or_b64 s[4:5], vcc, s[4:5]
-; GFX8-NEXT: v_mov_b32_e32 v1, v2
; GFX8-NEXT: s_andn2_b64 exec, exec, s[4:5]
; GFX8-NEXT: s_cbranch_execnz .LBB16_1
; GFX8-NEXT: ; %bb.2: ; %atomicrmw.end
; GFX8-NEXT: s_or_b64 exec, exec, s[4:5]
-; GFX8-NEXT: v_lshrrev_b32_e32 v0, s6, v2
+; GFX8-NEXT: v_lshrrev_b32_e32 v0, s6, v1
; GFX8-NEXT: s_setpc_b64 s[30:31]
;
; GFX7-LABEL: buffer_fat_ptr_agent_atomic_fadd_ret_bf16__offset__amdgpu_no_fine_grained_memory:
@@ -5522,38 +5527,38 @@ define bfloat @buffer_fat_ptr_agent_atomic_fadd_ret_bf16__offset__amdgpu_no_fine
; GFX7-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX7-NEXT: s_addk_i32 s20, 0x200
; GFX7-NEXT: s_and_b32 s4, s20, -4
-; GFX7-NEXT: v_mov_b32_e32 v4, s4
-; GFX7-NEXT: buffer_load_dword v1, v4, s[16:19], 0 offen
+; GFX7-NEXT: v_mov_b32_e32 v3, s4
+; GFX7-NEXT: buffer_load_dword v1, v3, s[16:19], 0 offen
; GFX7-NEXT: s_and_b32 s4, s20, 3
; GFX7-NEXT: s_lshl_b32 s6, s4, 3
; GFX7-NEXT: s_lshl_b32 s4, 0xffff, s6
; GFX7-NEXT: v_mul_f32_e32 v0, 1.0, v0
; GFX7-NEXT: s_not_b32 s7, s4
; GFX7-NEXT: s_mov_b64 s[4:5], 0
-; GFX7-NEXT: v_and_b32_e32 v5, 0xffff0000, v0
+; GFX7-NEXT: v_and_b32_e32 v0, 0xffff0000, v0
; GFX7-NEXT: .LBB16_1: ; %atomicrmw.start
; GFX7-NEXT: ; =>This Inner Loop Header: Depth=1
; GFX7-NEXT: s_waitcnt vmcnt(0)
-; GFX7-NEXT: v_lshrrev_b32_e32 v0, s6, v1
-; GFX7-NEXT: v_lshlrev_b32_e32 v0, 16, v0
-; GFX7-NEXT: v_add_f32_e32 v0, v0, v5
-; GFX7-NEXT: v_lshrrev_b32_e32 v0, 16, v0
-; GFX7-NEXT: v_and_b32_e32 v2, s7, v1
-; GFX7-NEXT: v_lshlrev_b32_e32 v0, s6, v0
-; GFX7-NEXT: v_or_b32_e32 v0, v2, v0
-; GFX7-NEXT: v_mov_b32_e32 v3, v1
-; GFX7-NEXT: v_mov_b32_e32 v2, v0
-; GFX7-NEXT: buffer_atomic_cmpswap v[2:3], v4, s[16:19], 0 offen glc
+; GFX7-NEXT: v_mov_b32_e32 v5, v1
+; GFX7-NEXT: v_lshrrev_b32_e32 v1, s6, v5
+; GFX7-NEXT: v_lshlrev_b32_e32 v1, 16, v1
+; GFX7-NEXT: v_add_f32_e32 v1, v1, v0
+; GFX7-NEXT: v_lshrrev_b32_e32 v1, 16, v1
+; GFX7-NEXT: v_and_b32_e32 v2, s7, v5
+; GFX7-NEXT: v_lshlrev_b32_e32 v1, s6, v1
+; GFX7-NEXT: v_or_b32_e32 v4, v2, v1
+; GFX7-NEXT: v_mov_b32_e32 v1, v4
+; GFX7-NEXT: v_mov_b32_e32 v2, v5
+; GFX7-NEXT: buffer_atomic_cmpswap v[1:2], v3, s[16:19], 0 offen glc
; GFX7-NEXT: s_waitcnt vmcnt(0)
; GFX7-NEXT: buffer_wbinvl1
-; GFX7-NEXT: v_cmp_eq_u32_e32 vcc, v2, v1
+; GFX7-NEXT: v_cmp_eq_u32_e32 vcc, v1, v5
; GFX7-NEXT: s_or_b64 s[4:5], vcc, s[4:5]
-; GFX7-NEXT: v_mov_b32_e32 v1, v2
; GFX7-NEXT: s_andn2_b64 exec, exec, s[4:5]
; GFX7-NEXT: s_cbranch_execnz .LBB16_1
; GFX7-NEXT: ; %bb.2: ; %atomicrmw.end
; GFX7-NEXT: s_or_b64 exec, exec, s[4:5]
-; GFX7-NEXT: v_lshrrev_b32_e32 v0, s6, v2
+; GFX7-NEXT: v_lshrrev_b32_e32 v0, s6, v1
; GFX7-NEXT: v_lshlrev_b32_e32 v0, 16, v0
; GFX7-NEXT: s_setpc_b64 s[30:31]
;
@@ -5562,39 +5567,39 @@ define bfloat @buffer_fat_ptr_agent_atomic_fadd_ret_bf16__offset__amdgpu_no_fine
; GFX6-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX6-NEXT: s_addk_i32 s20, 0x200
; GFX6-NEXT: s_and_b32 s4, s20, -4
-; GFX6-NEXT: v_mov_b32_e32 v4, s4
-; GFX6-NEXT: buffer_load_dword v1, v4, s[16:19], 0 offen
+; GFX6-NEXT: v_mov_b32_e32 v3, s4
+; GFX6-NEXT: buffer_load_dword v1, v3, s[16:19], 0 offen
; GFX6-NEXT: s_and_b32 s4, s20, 3
; GFX6-NEXT: s_lshl_b32 s6, s4, 3
; GFX6-NEXT: s_lshl_b32 s4, 0xffff, s6
; GFX6-NEXT: v_mul_f32_e32 v0, 1.0, v0
; GFX6-NEXT: s_not_b32 s7, s4
; GFX6-NEXT: s_mov_b64 s[4:5], 0
-; GFX6-NEXT: v_and_b32_e32 v5, 0xffff0000, v0
+; GFX6-NEXT: v_and_b32_e32 v0, 0xffff0000, v0
; GFX6-NEXT: .LBB16_1: ; %atomicrmw.start
; GFX6-NEXT: ; =>This Inner Loop Header: Depth=1
; GFX6-NEXT: s_waitcnt vmcnt(0)
-; GFX6-NEXT: v_lshrrev_b32_e32 v0, s6, v1
-; GFX6-NEXT: v_lshlrev_b32_e32 v0, 16, v0
-; GFX6-NEXT: v_add_f32_e32 v0, v0, v5
-; GFX6-NEXT: v_lshrrev_b32_e32 v0, 16, v0
+; GFX6-NEXT: v_mov_b32_e32 v5, v1
; GFX6-NEXT: s_waitcnt expcnt(0)
-; GFX6-NEXT: v_and_b32_e32 v2, s7, v1
-; GFX6-NEXT: v_lshlrev_b32_e32 v0, s6, v0
-; GFX6-NEXT: v_or_b32_e32 v0, v2, v0
-; GFX6-NEXT: v_mov_b32_e32 v3, v1
-; GFX6-NEXT: v_mov_b32_e32 v2, v0
-; GFX6-NEXT: buffer_atomic_cmpswap v[2:3], v4, s[16:19], 0 offen glc
+; GFX6-NEXT: v_lshrrev_b32_e32 v1, s6, v5
+; GFX6-NEXT: v_lshlrev_b32_e32 v1, 16, v1
+; GFX6-NEXT: v_add_f32_e32 v1, v1, v0
+; GFX6-NEXT: v_lshrrev_b32_e32 v1, 16, v1
+; GFX6-NEXT: v_and_b32_e32 v2, s7, v5
+; GFX6-NEXT: v_lshlrev_b32_e32 v1, s6, v1
+; GFX6-NEXT: v_or_b32_e32 v4, v2, v1
+; GFX6-NEXT: v_mov_b32_e32 v1, v4
+; GFX6-NEXT: v_mov_b32_e32 v2, v5
+; GFX6-NEXT: buffer_atomic_cmpswap v[1:2], v3, s[16:19], 0 offen glc
; GFX6-NEXT: s_waitcnt vmcnt(0)
; GFX6-NEXT: buffer_wbinvl1
-; GFX6-NEXT: v_cmp_eq_u32_e32 vcc, v2, v1
+; GFX6-NEXT: v_cmp_eq_u32_e32 vcc, v1, v5
; GFX6-NEXT: s_or_b64 s[4:5], vcc, s[4:5]
-; GFX6-NEXT: v_mov_b32_e32 v1, v2
; GFX6-NEXT: s_andn2_b64 exec, exec, s[4:5]
; GFX6-NEXT: s_cbranch_execnz .LBB16_1
; GFX6-NEXT: ; %bb.2: ; %atomicrmw.end
; GFX6-NEXT: s_or_b64 exec, exec, s[4:5]
-; GFX6-NEXT: v_lshrrev_b32_e32 v0, s6, v2
+; GFX6-NEXT: v_lshrrev_b32_e32 v0, s6, v1
; GFX6-NEXT: v_lshlrev_b32_e32 v0, 16, v0
; GFX6-NEXT: s_waitcnt expcnt(0)
; GFX6-NEXT: s_setpc_b64 s[30:31]
@@ -5612,48 +5617,48 @@ define void @buffer_fat_ptr_agent_atomic_fadd_noret_bf16__offset__amdgpu_no_fine
; GFX12-TRUE16-NEXT: s_wait_bvhcnt 0x0
; GFX12-TRUE16-NEXT: s_wait_kmcnt 0x0
; GFX12-TRUE16-NEXT: s_addk_co_i32 s16, 0x200
-; GFX12-TRUE16-NEXT: v_lshlrev_b32_e32 v3, 16, v0
+; GFX12-TRUE16-NEXT: v_lshlrev_b32_e32 v0, 16, v0
; GFX12-TRUE16-NEXT: s_wait_alu 0xfffe
; GFX12-TRUE16-NEXT: s_and_b32 s4, s16, -4
; GFX12-TRUE16-NEXT: s_wait_alu 0xfffe
-; GFX12-TRUE16-NEXT: v_mov_b32_e32 v2, s4
+; GFX12-TRUE16-NEXT: v_mov_b32_e32 v3, s4
; GFX12-TRUE16-NEXT: s_and_b32 s4, s16, 3
; GFX12-TRUE16-NEXT: s_wait_alu 0xfffe
; GFX12-TRUE16-NEXT: s_lshl_b32 s4, s4, 3
; GFX12-TRUE16-NEXT: s_wait_alu 0xfffe
; GFX12-TRUE16-NEXT: s_lshl_b32 s5, 0xffff, s4
-; GFX12-TRUE16-NEXT: buffer_load_b32 v1, v2, s[0:3], null offen
+; GFX12-TRUE16-NEXT: buffer_load_b32 v1, v3, s[0:3], null offen
; GFX12-TRUE16-NEXT: s_not_b32 s6, s5
; GFX12-TRUE16-NEXT: s_mov_b32 s5, 0
; GFX12-TRUE16-NEXT: .LBB17_1: ; %atomicrmw.start
; GFX12-TRUE16-NEXT: ; =>This Inner Loop Header: Depth=1
; GFX12-TRUE16-NEXT: s_wait_loadcnt 0x0
-; GFX12-TRUE16-NEXT: v_lshrrev_b32_e32 v0, s4, v1
+; GFX12-TRUE16-NEXT: v_mov_b32_e32 v5, v1
; GFX12-TRUE16-NEXT: s_wait_storecnt 0x0
; GFX12-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
-; GFX12-TRUE16-NEXT: v_lshlrev_b32_e32 v0, 16, v0
-; GFX12-TRUE16-NEXT: v_add_f32_e32 v0, v0, v3
-; GFX12-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(SKIP_2) | instid1(VALU_DEP_3)
-; GFX12-TRUE16-NEXT: v_bfe_u32 v4, v0, 16, 1
-; GFX12-TRUE16-NEXT: v_or_b32_e32 v5, 0x400000, v0
-; GFX12-TRUE16-NEXT: v_cmp_u_f32_e32 vcc_lo, v0, v0
-; GFX12-TRUE16-NEXT: v_add3_u32 v4, v4, v0, 0x7fff
+; GFX12-TRUE16-NEXT: v_lshrrev_b32_e32 v1, s4, v5
+; GFX12-TRUE16-NEXT: v_lshlrev_b32_e32 v1, 16, v1
+; GFX12-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
+; GFX12-TRUE16-NEXT: v_add_f32_e32 v1, v1, v0
+; GFX12-TRUE16-NEXT: v_bfe_u32 v2, v1, 16, 1
+; GFX12-TRUE16-NEXT: v_or_b32_e32 v4, 0x400000, v1
+; GFX12-TRUE16-NEXT: v_cmp_u_f32_e32 vcc_lo, v1, v1
+; GFX12-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_3) | instskip(SKIP_1) | instid1(VALU_DEP_1)
+; GFX12-TRUE16-NEXT: v_add3_u32 v2, v2, v1, 0x7fff
; GFX12-TRUE16-NEXT: s_wait_alu 0xfffd
-; GFX12-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(SKIP_1) | instid1(VALU_DEP_2)
-; GFX12-TRUE16-NEXT: v_cndmask_b32_e32 v0, v4, v5, vcc_lo
-; GFX12-TRUE16-NEXT: v_mov_b16_e32 v4.h, 0
-; GFX12-TRUE16-NEXT: v_mov_b16_e32 v4.l, v0.h
-; GFX12-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(SKIP_1) | instid1(VALU_DEP_1)
-; GFX12-TRUE16-NEXT: v_lshlrev_b32_e32 v0, s4, v4
+; GFX12-TRUE16-NEXT: v_cndmask_b32_e32 v1, v2, v4, vcc_lo
+; GFX12-TRUE16-NEXT: v_mov_b16_e32 v2.h, 0
+; GFX12-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_2) | instskip(NEXT) | instid1(VALU_DEP_1)
+; GFX12-TRUE16-NEXT: v_mov_b16_e32 v2.l, v1.h
+; GFX12-TRUE16-NEXT: v_lshlrev_b32_e32 v1, s4, v2
; GFX12-TRUE16-NEXT: s_wait_alu 0xfffe
-; GFX12-TRUE16-NEXT: v_and_or_b32 v0, v1, s6, v0
-; GFX12-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1)
-; GFX12-TRUE16-NEXT: v_dual_mov_b32 v5, v1 :: v_dual_mov_b32 v4, v0
-; GFX12-TRUE16-NEXT: buffer_atomic_cmpswap_b32 v[4:5], v2, s[0:3], null offen th:TH_ATOMIC_RETURN
+; GFX12-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
+; GFX12-TRUE16-NEXT: v_and_or_b32 v4, v5, s6, v1
+; GFX12-TRUE16-NEXT: v_dual_mov_b32 v1, v4 :: v_dual_mov_b32 v2, v5
+; GFX12-TRUE16-NEXT: buffer_atomic_cmpswap_b32 v[1:2], v3, s[0:3], null offen th:TH_ATOMIC_RETURN
; GFX12-TRUE16-NEXT: s_wait_loadcnt 0x0
; GFX12-TRUE16-NEXT: global_inv scope:SCOPE_DEV
-; GFX12-TRUE16-NEXT: v_cmp_eq_u32_e32 vcc_lo, v4, v1
-; GFX12-TRUE16-NEXT: v_mov_b32_e32 v1, v4
+; GFX12-TRUE16-NEXT: v_cmp_eq_u32_e32 vcc_lo, v1, v5
; GFX12-TRUE16-NEXT: s_or_b32 s5, vcc_lo, s5
; GFX12-TRUE16-NEXT: s_wait_alu 0xfffe
; GFX12-TRUE16-NEXT: s_and_not1_b32 exec_lo, exec_lo, s5
@@ -5671,47 +5676,47 @@ define void @buffer_fat_ptr_agent_atomic_fadd_noret_bf16__offset__amdgpu_no_fine
; GFX12-FAKE16-NEXT: s_wait_bvhcnt 0x0
; GFX12-FAKE16-NEXT: s_wait_kmcnt 0x0
; GFX12-FAKE16-NEXT: s_addk_co_i32 s16, 0x200
-; GFX12-FAKE16-NEXT: v_lshlrev_b32_e32 v3, 16, v0
+; GFX12-FAKE16-NEXT: v_lshlrev_b32_e32 v0, 16, v0
; GFX12-FAKE16-NEXT: s_wait_alu 0xfffe
; GFX12-FAKE16-NEXT: s_and_b32 s4, s16, -4
; GFX12-FAKE16-NEXT: s_wait_alu 0xfffe
-; GFX12-FAKE16-NEXT: v_mov_b32_e32 v2, s4
+; GFX12-FAKE16-NEXT: v_mov_b32_e32 v3, s4
; GFX12-FAKE16-NEXT: s_and_b32 s4, s16, 3
; GFX12-FAKE16-NEXT: s_wait_alu 0xfffe
; GFX12-FAKE16-NEXT: s_lshl_b32 s4, s4, 3
; GFX12-FAKE16-NEXT: s_wait_alu 0xfffe
; GFX12-FAKE16-NEXT: s_lshl_b32 s5, 0xffff, s4
-; GFX12-FAKE16-NEXT: buffer_load_b32 v1, v2, s[0:3], null offen
+; GFX12-FAKE16-NEXT: buffer_load_b32 v1, v3, s[0:3], null offen
; GFX12-FAKE16-NEXT: s_not_b32 s6, s5
; GFX12-FAKE16-NEXT: s_mov_b32 s5, 0
; GFX12-FAKE16-NEXT: .LBB17_1: ; %atomicrmw.start
; GFX12-FAKE16-NEXT: ; =>This Inner Loop Header: Depth=1
; GFX12-FAKE16-NEXT: s_wait_loadcnt 0x0
-; GFX12-FAKE16-NEXT: v_lshrrev_b32_e32 v0, s4, v1
+; GFX12-FAKE16-NEXT: v_mov_b32_e32 v5, v1
; GFX12-FAKE16-NEXT: s_wait_storecnt 0x0
; GFX12-FAKE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
-; GFX12-FAKE16-NEXT: v_lshlrev_b32_e32 v0, 16, v0
-; GFX12-FAKE16-NEXT: v_add_f32_e32 v0, v0, v3
-; GFX12-FAKE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(SKIP_2) | instid1(VALU_DEP_3)
-; GFX12-FAKE16-NEXT: v_bfe_u32 v4, v0, 16, 1
-; GFX12-FAKE16-NEXT: v_or_b32_e32 v5, 0x400000, v0
-; GFX12-FAKE16-NEXT: v_cmp_u_f32_e32 vcc_lo, v0, v0
-; GFX12-FAKE16-NEXT: v_add3_u32 v4, v4, v0, 0x7fff
+; GFX12-FAKE16-NEXT: v_lshrrev_b32_e32 v1, s4, v5
+; GFX12-FAKE16-NEXT: v_lshlrev_b32_e32 v1, 16, v1
+; GFX12-FAKE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
+; GFX12-FAKE16-NEXT: v_add_f32_e32 v1, v1, v0
+; GFX12-FAKE16-NEXT: v_bfe_u32 v2, v1, 16, 1
+; GFX12-FAKE16-NEXT: v_or_b32_e32 v4, 0x400000, v1
+; GFX12-FAKE16-NEXT: v_cmp_u_f32_e32 vcc_lo, v1, v1
+; GFX12-FAKE16-NEXT: s_delay_alu instid0(VALU_DEP_3) | instskip(SKIP_1) | instid1(VALU_DEP_1)
+; GFX12-FAKE16-NEXT: v_add3_u32 v2, v2, v1, 0x7fff
; GFX12-FAKE16-NEXT: s_wait_alu 0xfffd
+; GFX12-FAKE16-NEXT: v_cndmask_b32_e32 v1, v2, v4, vcc_lo
; GFX12-FAKE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
-; GFX12-FAKE16-NEXT: v_cndmask_b32_e32 v0, v4, v5, vcc_lo
-; GFX12-FAKE16-NEXT: v_lshrrev_b32_e32 v0, 16, v0
-; GFX12-FAKE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(SKIP_1) | instid1(VALU_DEP_1)
-; GFX12-FAKE16-NEXT: v_lshlrev_b32_e32 v0, s4, v0
+; GFX12-FAKE16-NEXT: v_lshrrev_b32_e32 v1, 16, v1
+; GFX12-FAKE16-NEXT: v_lshlrev_b32_e32 v1, s4, v1
; GFX12-FAKE16-NEXT: s_wait_alu 0xfffe
-; GFX12-FAKE16-NEXT: v_and_or_b32 v0, v1, s6, v0
-; GFX12-FAKE16-NEXT: s_delay_alu instid0(VALU_DEP_1)
-; GFX12-FAKE16-NEXT: v_dual_mov_b32 v5, v1 :: v_dual_mov_b32 v4, v0
-; GFX12-FAKE16-NEXT: buffer_atomic_cmpswap_b32 v[4:5], v2, s[0:3], null offen th:TH_ATOMIC_RETURN
+; GFX12-FAKE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
+; GFX12-FAKE16-NEXT: v_and_or_b32 v4, v5, s6, v1
+; GFX12-FAKE16-NEXT: v_dual_mov_b32 v1, v4 :: v_dual_mov_b32 v2, v5
+; GFX12-FAKE16-NEXT: buffer_atomic_cmpswap_b32 v[1:2], v3, s[0:3], null offen th:TH_ATOMIC_RETURN
; GFX12-FAKE16-NEXT: s_wait_loadcnt 0x0
; GFX12-FAKE16-NEXT: global_inv scope:SCOPE_DEV
-; GFX12-FAKE16-NEXT: v_cmp_eq_u32_e32 vcc_lo, v4, v1
-; GFX12-FAKE16-NEXT: v_mov_b32_e32 v1, v4
+; GFX12-FAKE16-NEXT: v_cmp_eq_u32_e32 vcc_lo, v1, v5
; GFX12-FAKE16-NEXT: s_or_b32 s5, vcc_lo, s5
; GFX12-FAKE16-NEXT: s_wait_alu 0xfffe
; GFX12-FAKE16-NEXT: s_and_not1_b32 exec_lo, exec_lo, s5
@@ -5726,36 +5731,36 @@ define void @buffer_fat_ptr_agent_atomic_fadd_noret_bf16__offset__amdgpu_no_fine
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX942-NEXT: s_addk_i32 s16, 0x200
; GFX942-NEXT: s_and_b32 s4, s16, -4
-; GFX942-NEXT: v_mov_b32_e32 v2, s4
-; GFX942-NEXT: buffer_load_dword v1, v2, s[0:3], 0 offen
+; GFX942-NEXT: v_mov_b32_e32 v1, s4
+; GFX942-NEXT: buffer_load_dword v2, v1, s[0:3], 0 offen
; GFX942-NEXT: s_and_b32 s4, s16, 3
; GFX942-NEXT: s_lshl_b32 s6, s4, 3
; GFX942-NEXT: s_lshl_b32 s4, 0xffff, s6
; GFX942-NEXT: s_not_b32 s7, s4
; GFX942-NEXT: s_mov_b64 s[4:5], 0
-; GFX942-NEXT: v_lshlrev_b32_e32 v3, 16, v0
+; GFX942-NEXT: v_lshlrev_b32_e32 v0, 16, v0
; GFX942-NEXT: s_movk_i32 s8, 0x7fff
; GFX942-NEXT: .LBB17_1: ; %atomicrmw.start
; GFX942-NEXT: ; =>This Inner Loop Header: Depth=1
; GFX942-NEXT: s_waitcnt vmcnt(0)
-; GFX942-NEXT: v_lshrrev_b32_sdwa v0, s6, v1 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:DWORD
+; GFX942-NEXT: v_mov_b32_e32 v5, v2
+; GFX942-NEXT: v_lshrrev_b32_sdwa v2, s6, v5 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:DWORD
; GFX942-NEXT: buffer_wbl2 sc1
-; GFX942-NEXT: v_add_f32_e32 v0, v0, v3
-; GFX942-NEXT: v_bfe_u32 v4, v0, 16, 1
-; GFX942-NEXT: v_or_b32_e32 v5, 0x400000, v0
-; GFX942-NEXT: v_add3_u32 v4, v4, v0, s8
-; GFX942-NEXT: v_cmp_u_f32_e32 vcc, v0, v0
+; GFX942-NEXT: v_add_f32_e32 v2, v2, v0
+; GFX942-NEXT: v_bfe_u32 v3, v2, 16, 1
+; GFX942-NEXT: v_or_b32_e32 v4, 0x400000, v2
+; GFX942-NEXT: v_add3_u32 v3, v3, v2, s8
+; GFX942-NEXT: v_cmp_u_f32_e32 vcc, v2, v2
; GFX942-NEXT: s_nop 1
-; GFX942-NEXT: v_cndmask_b32_e32 v0, v4, v5, vcc
-; GFX942-NEXT: v_lshlrev_b32_sdwa v0, s6, v0 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:WORD_1
-; GFX942-NEXT: v_and_or_b32 v0, v1, s7, v0
-; GFX942-NEXT: v_mov_b64_e32 v[4:5], v[0:1]
-; GFX942-NEXT: buffer_atomic_cmpswap v[4:5], v2, s[0:3], 0 offen sc0
+; GFX942-NEXT: v_cndmask_b32_e32 v2, v3, v4, vcc
+; GFX942-NEXT: v_lshlrev_b32_sdwa v2, s6, v2 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:WORD_1
+; GFX942-NEXT: v_and_or_b32 v4, v5, s7, v2
+; GFX942-NEXT: v_mov_b64_e32 v[2:3], v[4:5]
+; GFX942-NEXT: buffer_atomic_cmpswap v[2:3], v1, s[0:3], 0 offen sc0
; GFX942-NEXT: s_waitcnt vmcnt(0)
; GFX942-NEXT: buffer_inv sc1
-; GFX942-NEXT: v_cmp_eq_u32_e32 vcc, v4, v1
+; GFX942-NEXT: v_cmp_eq_u32_e32 vcc, v2, v5
; GFX942-NEXT: s_or_b64 s[4:5], vcc, s[4:5]
-; GFX942-NEXT: v_mov_b32_e32 v1, v4
; GFX942-NEXT: s_andn2_b64 exec, exec, s[4:5]
; GFX942-NEXT: s_cbranch_execnz .LBB17_1
; GFX942-NEXT: ; %bb.2: ; %atomicrmw.end
@@ -5766,46 +5771,46 @@ define void @buffer_fat_ptr_agent_atomic_fadd_noret_bf16__offset__amdgpu_no_fine
; GFX11-TRUE16: ; %bb.0:
; GFX11-TRUE16-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX11-TRUE16-NEXT: s_addk_i32 s16, 0x200
-; GFX11-TRUE16-NEXT: v_lshlrev_b32_e32 v3, 16, v0
+; GFX11-TRUE16-NEXT: v_lshlrev_b32_e32 v0, 16, v0
; GFX11-TRUE16-NEXT: s_and_b32 s4, s16, -4
; GFX11-TRUE16-NEXT: s_delay_alu instid0(SALU_CYCLE_1) | instskip(SKIP_1) | instid1(SALU_CYCLE_1)
-; GFX11-TRUE16-NEXT: v_mov_b32_e32 v2, s4
+; GFX11-TRUE16-NEXT: v_mov_b32_e32 v3, s4
; GFX11-TRUE16-NEXT: s_and_b32 s4, s16, 3
; GFX11-TRUE16-NEXT: s_lshl_b32 s4, s4, 3
; GFX11-TRUE16-NEXT: s_delay_alu instid0(SALU_CYCLE_1)
; GFX11-TRUE16-NEXT: s_lshl_b32 s5, 0xffff, s4
-; GFX11-TRUE16-NEXT: buffer_load_b32 v1, v2, s[0:3], 0 offen
+; GFX11-TRUE16-NEXT: buffer_load_b32 v1, v3, s[0:3], 0 offen
; GFX11-TRUE16-NEXT: s_not_b32 s6, s5
; GFX11-TRUE16-NEXT: s_mov_b32 s5, 0
; GFX11-TRUE16-NEXT: .p2align 6
; GFX11-TRUE16-NEXT: .LBB17_1: ; %atomicrmw.start
; GFX11-TRUE16-NEXT: ; =>This Inner Loop Header: Depth=1
; GFX11-TRUE16-NEXT: s_waitcnt vmcnt(0)
-; GFX11-TRUE16-NEXT: v_lshrrev_b32_e32 v0, s4, v1
+; GFX11-TRUE16-NEXT: v_mov_b32_e32 v5, v1
; GFX11-TRUE16-NEXT: s_waitcnt_vscnt null, 0x0
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
-; GFX11-TRUE16-NEXT: v_lshlrev_b32_e32 v0, 16, v0
-; GFX11-TRUE16-NEXT: v_add_f32_e32 v0, v0, v3
-; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(SKIP_2) | instid1(VALU_DEP_3)
-; GFX11-TRUE16-NEXT: v_bfe_u32 v4, v0, 16, 1
-; GFX11-TRUE16-NEXT: v_or_b32_e32 v5, 0x400000, v0
-; GFX11-TRUE16-NEXT: v_cmp_u_f32_e32 vcc_lo, v0, v0
-; GFX11-TRUE16-NEXT: v_add3_u32 v4, v4, v0, 0x7fff
-; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(SKIP_1) | instid1(VALU_DEP_2)
-; GFX11-TRUE16-NEXT: v_cndmask_b32_e32 v0, v4, v5, vcc_lo
-; GFX11-TRUE16-NEXT: v_mov_b16_e32 v4.h, 0
-; GFX11-TRUE16-NEXT: v_mov_b16_e32 v4.l, v0.h
+; GFX11-TRUE16-NEXT: v_lshrrev_b32_e32 v1, s4, v5
+; GFX11-TRUE16-NEXT: v_lshlrev_b32_e32 v1, 16, v1
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
-; GFX11-TRUE16-NEXT: v_lshlrev_b32_e32 v0, s4, v4
-; GFX11-TRUE16-NEXT: v_and_or_b32 v0, v1, s6, v0
-; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1)
-; GFX11-TRUE16-NEXT: v_dual_mov_b32 v5, v1 :: v_dual_mov_b32 v4, v0
-; GFX11-TRUE16-NEXT: buffer_atomic_cmpswap_b32 v[4:5], v2, s[0:3], 0 offen glc
+; GFX11-TRUE16-NEXT: v_add_f32_e32 v1, v1, v0
+; GFX11-TRUE16-NEXT: v_bfe_u32 v2, v1, 16, 1
+; GFX11-TRUE16-NEXT: v_or_b32_e32 v4, 0x400000, v1
+; GFX11-TRUE16-NEXT: v_cmp_u_f32_e32 vcc_lo, v1, v1
+; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_3) | instskip(NEXT) | instid1(VALU_DEP_1)
+; GFX11-TRUE16-NEXT: v_add3_u32 v2, v2, v1, 0x7fff
+; GFX11-TRUE16-NEXT: v_cndmask_b32_e32 v1, v2, v4, vcc_lo
+; GFX11-TRUE16-NEXT: v_mov_b16_e32 v2.h, 0
+; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_2) | instskip(NEXT) | instid1(VALU_DEP_1)
+; GFX11-TRUE16-NEXT: v_mov_b16_e32 v2.l, v1.h
+; GFX11-TRUE16-NEXT: v_lshlrev_b32_e32 v1, s4, v2
+; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
+; GFX11-TRUE16-NEXT: v_and_or_b32 v4, v5, s6, v1
+; GFX11-TRUE16-NEXT: v_dual_mov_b32 v1, v4 :: v_dual_mov_b32 v2, v5
+; GFX11-TRUE16-NEXT: buffer_atomic_cmpswap_b32 v[1:2], v3, s[0:3], 0 offen glc
; GFX11-TRUE16-NEXT: s_waitcnt vmcnt(0)
; GFX11-TRUE16-NEXT: buffer_gl1_inv
; GFX11-TRUE16-NEXT: buffer_gl0_inv
-; GFX11-TRUE16-NEXT: v_cmp_eq_u32_e32 vcc_lo, v4, v1
-; GFX11-TRUE16-NEXT: v_mov_b32_e32 v1, v4
+; GFX11-TRUE16-NEXT: v_cmp_eq_u32_e32 vcc_lo, v1, v5
; GFX11-TRUE16-NEXT: s_or_b32 s5, vcc_lo, s5
; GFX11-TRUE16-NEXT: s_delay_alu instid0(SALU_CYCLE_1)
; GFX11-TRUE16-NEXT: s_and_not1_b32 exec_lo, exec_lo, s5
@@ -5818,45 +5823,45 @@ define void @buffer_fat_ptr_agent_atomic_fadd_noret_bf16__offset__amdgpu_no_fine
; GFX11-FAKE16: ; %bb.0:
; GFX11-FAKE16-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX11-FAKE16-NEXT: s_addk_i32 s16, 0x200
-; GFX11-FAKE16-NEXT: v_lshlrev_b32_e32 v3, 16, v0
+; GFX11-FAKE16-NEXT: v_lshlrev_b32_e32 v0, 16, v0
; GFX11-FAKE16-NEXT: s_and_b32 s4, s16, -4
; GFX11-FAKE16-NEXT: s_delay_alu instid0(SALU_CYCLE_1) | instskip(SKIP_1) | instid1(SALU_CYCLE_1)
-; GFX11-FAKE16-NEXT: v_mov_b32_e32 v2, s4
+; GFX11-FAKE16-NEXT: v_mov_b32_e32 v3, s4
; GFX11-FAKE16-NEXT: s_and_b32 s4, s16, 3
; GFX11-FAKE16-NEXT: s_lshl_b32 s4, s4, 3
; GFX11-FAKE16-NEXT: s_delay_alu instid0(SALU_CYCLE_1)
; GFX11-FAKE16-NEXT: s_lshl_b32 s5, 0xffff, s4
-; GFX11-FAKE16-NEXT: buffer_load_b32 v1, v2, s[0:3], 0 offen
+; GFX11-FAKE16-NEXT: buffer_load_b32 v1, v3, s[0:3], 0 offen
; GFX11-FAKE16-NEXT: s_not_b32 s6, s5
; GFX11-FAKE16-NEXT: s_mov_b32 s5, 0
; GFX11-FAKE16-NEXT: .p2align 6
; GFX11-FAKE16-NEXT: .LBB17_1: ; %atomicrmw.start
; GFX11-FAKE16-NEXT: ; =>This Inner Loop Header: Depth=1
; GFX11-FAKE16-NEXT: s_waitcnt vmcnt(0)
-; GFX11-FAKE16-NEXT: v_lshrrev_b32_e32 v0, s4, v1
+; GFX11-FAKE16-NEXT: v_mov_b32_e32 v5, v1
; GFX11-FAKE16-NEXT: s_waitcnt_vscnt null, 0x0
; GFX11-FAKE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
-; GFX11-FAKE16-NEXT: v_lshlrev_b32_e32 v0, 16, v0
-; GFX11-FAKE16-NEXT: v_add_f32_e32 v0, v0, v3
-; GFX11-FAKE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(SKIP_2) | instid1(VALU_DEP_3)
-; GFX11-FAKE16-NEXT: v_bfe_u32 v4, v0, 16, 1
-; GFX11-FAKE16-NEXT: v_or_b32_e32 v5, 0x400000, v0
-; GFX11-FAKE16-NEXT: v_cmp_u_f32_e32 vcc_lo, v0, v0
-; GFX11-FAKE16-NEXT: v_add3_u32 v4, v4, v0, 0x7fff
+; GFX11-FAKE16-NEXT: v_lshrrev_b32_e32 v1, s4, v5
+; GFX11-FAKE16-NEXT: v_lshlrev_b32_e32 v1, 16, v1
+; GFX11-FAKE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
+; GFX11-FAKE16-NEXT: v_add_f32_e32 v1, v1, v0
+; GFX11-FAKE16-NEXT: v_bfe_u32 v2, v1, 16, 1
+; GFX11-FAKE16-NEXT: v_or_b32_e32 v4, 0x400000, v1
+; GFX11-FAKE16-NEXT: v_cmp_u_f32_e32 vcc_lo, v1, v1
+; GFX11-FAKE16-NEXT: s_delay_alu instid0(VALU_DEP_3) | instskip(NEXT) | instid1(VALU_DEP_1)
+; GFX11-FAKE16-NEXT: v_add3_u32 v2, v2, v1, 0x7fff
+; GFX11-FAKE16-NEXT: v_cndmask_b32_e32 v1, v2, v4, vcc_lo
; GFX11-FAKE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
-; GFX11-FAKE16-NEXT: v_cndmask_b32_e32 v0, v4, v5, vcc_lo
-; GFX11-FAKE16-NEXT: v_lshrrev_b32_e32 v0, 16, v0
+; GFX11-FAKE16-NEXT: v_lshrrev_b32_e32 v1, 16, v1
+; GFX11-FAKE16-NEXT: v_lshlrev_b32_e32 v1, s4, v1
; GFX11-FAKE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
-; GFX11-FAKE16-NEXT: v_lshlrev_b32_e32 v0, s4, v0
-; GFX11-FAKE16-NEXT: v_and_or_b32 v0, v1, s6, v0
-; GFX11-FAKE16-NEXT: s_delay_alu instid0(VALU_DEP_1)
-; GFX11-FAKE16-NEXT: v_dual_mov_b32 v5, v1 :: v_dual_mov_b32 v4, v0
-; GFX11-FAKE16-NEXT: buffer_atomic_cmpswap_b32 v[4:5], v2, s[0:3], 0 offen glc
+; GFX11-FAKE16-NEXT: v_and_or_b32 v4, v5, s6, v1
+; GFX11-FAKE16-NEXT: v_dual_mov_b32 v1, v4 :: v_dual_mov_b32 v2, v5
+; GFX11-FAKE16-NEXT: buffer_atomic_cmpswap_b32 v[1:2], v3, s[0:3], 0 offen glc
; GFX11-FAKE16-NEXT: s_waitcnt vmcnt(0)
; GFX11-FAKE16-NEXT: buffer_gl1_inv
; GFX11-FAKE16-NEXT: buffer_gl0_inv
-; GFX11-FAKE16-NEXT: v_cmp_eq_u32_e32 vcc_lo, v4, v1
-; GFX11-FAKE16-NEXT: v_mov_b32_e32 v1, v4
+; GFX11-FAKE16-NEXT: v_cmp_eq_u32_e32 vcc_lo, v1, v5
; GFX11-FAKE16-NEXT: s_or_b32 s5, vcc_lo, s5
; GFX11-FAKE16-NEXT: s_delay_alu instid0(SALU_CYCLE_1)
; GFX11-FAKE16-NEXT: s_and_not1_b32 exec_lo, exec_lo, s5
@@ -5869,36 +5874,36 @@ define void @buffer_fat_ptr_agent_atomic_fadd_noret_bf16__offset__amdgpu_no_fine
; GFX10: ; %bb.0:
; GFX10-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX10-NEXT: s_addk_i32 s20, 0x200
-; GFX10-NEXT: v_lshlrev_b32_e32 v3, 16, v0
+; GFX10-NEXT: v_lshlrev_b32_e32 v0, 16, v0
; GFX10-NEXT: s_and_b32 s4, s20, -4
-; GFX10-NEXT: v_mov_b32_e32 v2, s4
+; GFX10-NEXT: v_mov_b32_e32 v3, s4
; GFX10-NEXT: s_and_b32 s4, s20, 3
; GFX10-NEXT: s_lshl_b32 s4, s4, 3
; GFX10-NEXT: s_lshl_b32 s5, 0xffff, s4
-; GFX10-NEXT: buffer_load_dword v1, v2, s[16:19], 0 offen
+; GFX10-NEXT: buffer_load_dword v1, v3, s[16:19], 0 offen
; GFX10-NEXT: s_not_b32 s6, s5
; GFX10-NEXT: s_mov_b32 s5, 0
; GFX10-NEXT: .LBB17_1: ; %atomicrmw.start
; GFX10-NEXT: ; =>This Inner Loop Header: Depth=1
; GFX10-NEXT: s_waitcnt vmcnt(0)
-; GFX10-NEXT: v_lshrrev_b32_sdwa v0, s4, v1 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:DWORD
-; GFX10-NEXT: s_waitcnt_vscnt null, 0x0
-; GFX10-NEXT: v_add_f32_e32 v0, v0, v3
-; GFX10-NEXT: v_bfe_u32 v4, v0, 16, 1
-; GFX10-NEXT: v_or_b32_e32 v5, 0x400000, v0
-; GFX10-NEXT: v_cmp_u_f32_e32 vcc_lo, v0, v0
-; GFX10-NEXT: v_add3_u32 v4, v4, v0, 0x7fff
-; GFX10-NEXT: v_cndmask_b32_e32 v0, v4, v5, vcc_lo
-; GFX10-NEXT: v_lshlrev_b32_sdwa v0, s4, v0 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:WORD_1
-; GFX10-NEXT: v_and_or_b32 v0, v1, s6, v0
; GFX10-NEXT: v_mov_b32_e32 v5, v1
-; GFX10-NEXT: v_mov_b32_e32 v4, v0
-; GFX10-NEXT: buffer_atomic_cmpswap v[4:5], v2, s[16:19], 0 offen glc
+; GFX10-NEXT: s_waitcnt_vscnt null, 0x0
+; GFX10-NEXT: v_lshrrev_b32_sdwa v1, s4, v5 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:DWORD
+; GFX10-NEXT: v_add_f32_e32 v1, v1, v0
+; GFX10-NEXT: v_bfe_u32 v2, v1, 16, 1
+; GFX10-NEXT: v_or_b32_e32 v4, 0x400000, v1
+; GFX10-NEXT: v_cmp_u_f32_e32 vcc_lo, v1, v1
+; GFX10-NEXT: v_add3_u32 v2, v2, v1, 0x7fff
+; GFX10-NEXT: v_cndmask_b32_e32 v1, v2, v4, vcc_lo
+; GFX10-NEXT: v_lshlrev_b32_sdwa v1, s4, v1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:WORD_1
+; GFX10-NEXT: v_and_or_b32 v4, v5, s6, v1
+; GFX10-NEXT: v_mov_b32_e32 v1, v4
+; GFX10-NEXT: v_mov_b32_e32 v2, v5
+; GFX10-NEXT: buffer_atomic_cmpswap v[1:2], v3, s[16:19], 0 offen glc
; GFX10-NEXT: s_waitcnt vmcnt(0)
; GFX10-NEXT: buffer_gl1_inv
; GFX10-NEXT: buffer_gl0_inv
-; GFX10-NEXT: v_cmp_eq_u32_e32 vcc_lo, v4, v1
-; GFX10-NEXT: v_mov_b32_e32 v1, v4
+; GFX10-NEXT: v_cmp_eq_u32_e32 vcc_lo, v1, v5
; GFX10-NEXT: s_or_b32 s5, vcc_lo, s5
; GFX10-NEXT: s_andn2_b32 exec_lo, exec_lo, s5
; GFX10-NEXT: s_cbranch_execnz .LBB17_1
@@ -5911,34 +5916,34 @@ define void @buffer_fat_ptr_agent_atomic_fadd_noret_bf16__offset__amdgpu_no_fine
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX90A-NEXT: s_addk_i32 s20, 0x200
; GFX90A-NEXT: s_and_b32 s4, s20, -4
-; GFX90A-NEXT: v_mov_b32_e32 v2, s4
-; GFX90A-NEXT: buffer_load_dword v1, v2, s[16:19], 0 offen
+; GFX90A-NEXT: v_mov_b32_e32 v1, s4
+; GFX90A-NEXT: buffer_load_dword v2, v1, s[16:19], 0 offen
; GFX90A-NEXT: s_and_b32 s4, s20, 3
; GFX90A-NEXT: s_lshl_b32 s6, s4, 3
; GFX90A-NEXT: s_lshl_b32 s4, 0xffff, s6
; GFX90A-NEXT: s_not_b32 s7, s4
; GFX90A-NEXT: s_mov_b64 s[4:5], 0
-; GFX90A-NEXT: v_lshlrev_b32_e32 v3, 16, v0
+; GFX90A-NEXT: v_lshlrev_b32_e32 v0, 16, v0
; GFX90A-NEXT: s_movk_i32 s8, 0x7fff
; GFX90A-NEXT: .LBB17_1: ; %atomicrmw.start
; GFX90A-NEXT: ; =>This Inner Loop Header: Depth=1
; GFX90A-NEXT: s_waitcnt vmcnt(0)
-; GFX90A-NEXT: v_lshrrev_b32_sdwa v0, s6, v1 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:DWORD
-; GFX90A-NEXT: v_add_f32_e32 v0, v0, v3
-; GFX90A-NEXT: v_bfe_u32 v4, v0, 16, 1
-; GFX90A-NEXT: v_or_b32_e32 v5, 0x400000, v0
-; GFX90A-NEXT: v_add3_u32 v4, v4, v0, s8
-; GFX90A-NEXT: v_cmp_u_f32_e32 vcc, v0, v0
-; GFX90A-NEXT: v_cndmask_b32_e32 v0, v4, v5, vcc
-; GFX90A-NEXT: v_lshlrev_b32_sdwa v0, s6, v0 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:WORD_1
-; GFX90A-NEXT: v_and_or_b32 v0, v1, s7, v0
-; GFX90A-NEXT: v_pk_mov_b32 v[4:5], v[0:1], v[0:1] op_sel:[0,1]
-; GFX90A-NEXT: buffer_atomic_cmpswap v[4:5], v2, s[16:19], 0 offen glc
+; GFX90A-NEXT: v_mov_b32_e32 v5, v2
+; GFX90A-NEXT: v_lshrrev_b32_sdwa v2, s6, v5 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:DWORD
+; GFX90A-NEXT: v_add_f32_e32 v2, v2, v0
+; GFX90A-NEXT: v_bfe_u32 v3, v2, 16, 1
+; GFX90A-NEXT: v_or_b32_e32 v4, 0x400000, v2
+; GFX90A-NEXT: v_add3_u32 v3, v3, v2, s8
+; GFX90A-NEXT: v_cmp_u_f32_e32 vcc, v2, v2
+; GFX90A-NEXT: v_cndmask_b32_e32 v2, v3, v4, vcc
+; GFX90A-NEXT: v_lshlrev_b32_sdwa v2, s6, v2 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:WORD_1
+; GFX90A-NEXT: v_and_or_b32 v4, v5, s7, v2
+; GFX90A-NEXT: v_pk_mov_b32 v[2:3], v[4:5], v[4:5] op_sel:[0,1]
+; GFX90A-NEXT: buffer_atomic_cmpswap v[2:3], v1, s[16:19], 0 offen glc
; GFX90A-NEXT: s_waitcnt vmcnt(0)
; GFX90A-NEXT: buffer_wbinvl1
-; GFX90A-NEXT: v_cmp_eq_u32_e32 vcc, v4, v1
+; GFX90A-NEXT: v_cmp_eq_u32_e32 vcc, v2, v5
; GFX90A-NEXT: s_or_b64 s[4:5], vcc, s[4:5]
-; GFX90A-NEXT: v_mov_b32_e32 v1, v4
; GFX90A-NEXT: s_andn2_b64 exec, exec, s[4:5]
; GFX90A-NEXT: s_cbranch_execnz .LBB17_1
; GFX90A-NEXT: ; %bb.2: ; %atomicrmw.end
@@ -5950,35 +5955,35 @@ define void @buffer_fat_ptr_agent_atomic_fadd_noret_bf16__offset__amdgpu_no_fine
; GFX908-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX908-NEXT: s_addk_i32 s20, 0x200
; GFX908-NEXT: s_and_b32 s4, s20, -4
-; GFX908-NEXT: v_mov_b32_e32 v2, s4
-; GFX908-NEXT: buffer_load_dword v1, v2, s[16:19], 0 offen
+; GFX908-NEXT: v_mov_b32_e32 v3, s4
+; GFX908-NEXT: buffer_load_dword v1, v3, s[16:19], 0 offen
; GFX908-NEXT: s_and_b32 s4, s20, 3
; GFX908-NEXT: s_lshl_b32 s6, s4, 3
; GFX908-NEXT: s_lshl_b32 s4, 0xffff, s6
; GFX908-NEXT: s_not_b32 s7, s4
; GFX908-NEXT: s_mov_b64 s[4:5], 0
-; GFX908-NEXT: v_lshlrev_b32_e32 v3, 16, v0
+; GFX908-NEXT: v_lshlrev_b32_e32 v0, 16, v0
; GFX908-NEXT: s_movk_i32 s8, 0x7fff
; GFX908-NEXT: .LBB17_1: ; %atomicrmw.start
; GFX908-NEXT: ; =>This Inner Loop Header: Depth=1
; GFX908-NEXT: s_waitcnt vmcnt(0)
-; GFX908-NEXT: v_lshrrev_b32_sdwa v0, s6, v1 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:DWORD
-; GFX908-NEXT: v_add_f32_e32 v0, v0, v3
-; GFX908-NEXT: v_bfe_u32 v4, v0, 16, 1
-; GFX908-NEXT: v_or_b32_e32 v5, 0x400000, v0
-; GFX908-NEXT: v_add3_u32 v4, v4, v0, s8
-; GFX908-NEXT: v_cmp_u_f32_e32 vcc, v0, v0
-; GFX908-NEXT: v_cndmask_b32_e32 v0, v4, v5, vcc
-; GFX908-NEXT: v_lshlrev_b32_sdwa v0, s6, v0 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:WORD_1
-; GFX908-NEXT: v_and_or_b32 v0, v1, s7, v0
; GFX908-NEXT: v_mov_b32_e32 v5, v1
-; GFX908-NEXT: v_mov_b32_e32 v4, v0
-; GFX908-NEXT: buffer_atomic_cmpswap v[4:5], v2, s[16:19], 0 offen glc
+; GFX908-NEXT: v_lshrrev_b32_sdwa v1, s6, v5 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:DWORD
+; GFX908-NEXT: v_add_f32_e32 v1, v1, v0
+; GFX908-NEXT: v_bfe_u32 v2, v1, 16, 1
+; GFX908-NEXT: v_or_b32_e32 v4, 0x400000, v1
+; GFX908-NEXT: v_add3_u32 v2, v2, v1, s8
+; GFX908-NEXT: v_cmp_u_f32_e32 vcc, v1, v1
+; GFX908-NEXT: v_cndmask_b32_e32 v1, v2, v4, vcc
+; GFX908-NEXT: v_lshlrev_b32_sdwa v1, s6, v1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:WORD_1
+; GFX908-NEXT: v_and_or_b32 v4, v5, s7, v1
+; GFX908-NEXT: v_mov_b32_e32 v1, v4
+; GFX908-NEXT: v_mov_b32_e32 v2, v5
+; GFX908-NEXT: buffer_atomic_cmpswap v[1:2], v3, s[16:19], 0 offen glc
; GFX908-NEXT: s_waitcnt vmcnt(0)
; GFX908-NEXT: buffer_wbinvl1
-; GFX908-NEXT: v_cmp_eq_u32_e32 vcc, v4, v1
+; GFX908-NEXT: v_cmp_eq_u32_e32 vcc, v1, v5
; GFX908-NEXT: s_or_b64 s[4:5], vcc, s[4:5]
-; GFX908-NEXT: v_mov_b32_e32 v1, v4
; GFX908-NEXT: s_andn2_b64 exec, exec, s[4:5]
; GFX908-NEXT: s_cbranch_execnz .LBB17_1
; GFX908-NEXT: ; %bb.2: ; %atomicrmw.end
@@ -5990,37 +5995,37 @@ define void @buffer_fat_ptr_agent_atomic_fadd_noret_bf16__offset__amdgpu_no_fine
; GFX8-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX8-NEXT: s_addk_i32 s20, 0x200
; GFX8-NEXT: s_and_b32 s4, s20, -4
-; GFX8-NEXT: v_mov_b32_e32 v2, s4
-; GFX8-NEXT: buffer_load_dword v1, v2, s[16:19], 0 offen
+; GFX8-NEXT: v_mov_b32_e32 v3, s4
+; GFX8-NEXT: buffer_load_dword v1, v3, s[16:19], 0 offen
; GFX8-NEXT: s_and_b32 s4, s20, 3
; GFX8-NEXT: s_lshl_b32 s6, s4, 3
; GFX8-NEXT: s_lshl_b32 s4, 0xffff, s6
; GFX8-NEXT: s_not_b32 s7, s4
; GFX8-NEXT: s_mov_b64 s[4:5], 0
-; GFX8-NEXT: v_lshlrev_b32_e32 v3, 16, v0
+; GFX8-NEXT: v_lshlrev_b32_e32 v0, 16, v0
; GFX8-NEXT: .LBB17_1: ; %atomicrmw.start
; GFX8-NEXT: ; =>This Inner Loop Header: Depth=1
-; GFX8-NEXT: v_mov_b32_e32 v0, s6
+; GFX8-NEXT: v_mov_b32_e32 v2, s6
; GFX8-NEXT: s_waitcnt vmcnt(0)
-; GFX8-NEXT: v_lshrrev_b32_sdwa v5, v0, v1 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:DWORD
-; GFX8-NEXT: v_add_f32_e32 v5, v5, v3
-; GFX8-NEXT: v_bfe_u32 v6, v5, 16, 1
-; GFX8-NEXT: v_add_u32_e32 v6, vcc, v6, v5
-; GFX8-NEXT: v_add_u32_e32 v6, vcc, 0x7fff, v6
-; GFX8-NEXT: v_or_b32_e32 v7, 0x400000, v5
-; GFX8-NEXT: v_cmp_u_f32_e32 vcc, v5, v5
-; GFX8-NEXT: v_cndmask_b32_e32 v5, v6, v7, vcc
-; GFX8-NEXT: v_and_b32_e32 v4, s7, v1
-; GFX8-NEXT: v_lshlrev_b32_sdwa v0, v0, v5 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:WORD_1
-; GFX8-NEXT: v_or_b32_e32 v0, v4, v0
; GFX8-NEXT: v_mov_b32_e32 v5, v1
-; GFX8-NEXT: v_mov_b32_e32 v4, v0
-; GFX8-NEXT: buffer_atomic_cmpswap v[4:5], v2, s[16:19], 0 offen glc
+; GFX8-NEXT: v_lshrrev_b32_sdwa v1, v2, v5 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:DWORD
+; GFX8-NEXT: v_add_f32_e32 v1, v1, v0
+; GFX8-NEXT: v_bfe_u32 v6, v1, 16, 1
+; GFX8-NEXT: v_add_u32_e32 v6, vcc, v6, v1
+; GFX8-NEXT: v_add_u32_e32 v6, vcc, 0x7fff, v6
+; GFX8-NEXT: v_or_b32_e32 v7, 0x400000, v1
+; GFX8-NEXT: v_cmp_u_f32_e32 vcc, v1, v1
+; GFX8-NEXT: v_cndmask_b32_e32 v1, v6, v7, vcc
+; GFX8-NEXT: v_and_b32_e32 v4, s7, v5
+; GFX8-NEXT: v_lshlrev_b32_sdwa v1, v2, v1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:WORD_1
+; GFX8-NEXT: v_or_b32_e32 v4, v4, v1
+; GFX8-NEXT: v_mov_b32_e32 v1, v4
+; GFX8-NEXT: v_mov_b32_e32 v2, v5
+; GFX8-NEXT: buffer_atomic_cmpswap v[1:2], v3, s[16:19], 0 offen glc
; GFX8-NEXT: s_waitcnt vmcnt(0)
; GFX8-NEXT: buffer_wbinvl1
-; GFX8-NEXT: v_cmp_eq_u32_e32 vcc, v4, v1
+; GFX8-NEXT: v_cmp_eq_u32_e32 vcc, v1, v5
; GFX8-NEXT: s_or_b64 s[4:5], vcc, s[4:5]
-; GFX8-NEXT: v_mov_b32_e32 v1, v4
; GFX8-NEXT: s_andn2_b64 exec, exec, s[4:5]
; GFX8-NEXT: s_cbranch_execnz .LBB17_1
; GFX8-NEXT: ; %bb.2: ; %atomicrmw.end
@@ -6032,33 +6037,33 @@ define void @buffer_fat_ptr_agent_atomic_fadd_noret_bf16__offset__amdgpu_no_fine
; GFX7-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX7-NEXT: s_addk_i32 s20, 0x200
; GFX7-NEXT: s_and_b32 s4, s20, -4
-; GFX7-NEXT: v_mov_b32_e32 v2, s4
-; GFX7-NEXT: buffer_load_dword v1, v2, s[16:19], 0 offen
+; GFX7-NEXT: v_mov_b32_e32 v3, s4
+; GFX7-NEXT: buffer_load_dword v1, v3, s[16:19], 0 offen
; GFX7-NEXT: s_and_b32 s4, s20, 3
; GFX7-NEXT: s_lshl_b32 s6, s4, 3
; GFX7-NEXT: s_lshl_b32 s4, 0xffff, s6
; GFX7-NEXT: v_mul_f32_e32 v0, 1.0, v0
; GFX7-NEXT: s_not_b32 s7, s4
; GFX7-NEXT: s_mov_b64 s[4:5], 0
-; GFX7-NEXT: v_and_b32_e32 v3, 0xffff0000, v0
+; GFX7-NEXT: v_and_b32_e32 v0, 0xffff0000, v0
; GFX7-NEXT: .LBB17_1: ; %atomicrmw.start
; GFX7-NEXT: ; =>This Inner Loop Header: Depth=1
; GFX7-NEXT: s_waitcnt vmcnt(0)
-; GFX7-NEXT: v_lshrrev_b32_e32 v0, s6, v1
-; GFX7-NEXT: v_lshlrev_b32_e32 v0, 16, v0
-; GFX7-NEXT: v_add_f32_e32 v0, v0, v3
-; GFX7-NEXT: v_lshrrev_b32_e32 v0, 16, v0
-; GFX7-NEXT: v_and_b32_e32 v4, s7, v1
-; GFX7-NEXT: v_lshlrev_b32_e32 v0, s6, v0
-; GFX7-NEXT: v_or_b32_e32 v0, v4, v0
; GFX7-NEXT: v_mov_b32_e32 v5, v1
-; GFX7-NEXT: v_mov_b32_e32 v4, v0
-; GFX7-NEXT: buffer_atomic_cmpswap v[4:5], v2, s[16:19], 0 offen glc
+; GFX7-NEXT: v_lshrrev_b32_e32 v1, s6, v5
+; GFX7-NEXT: v_lshlrev_b32_e32 v1, 16, v1
+; GFX7-NEXT: v_add_f32_e32 v1, v1, v0
+; GFX7-NEXT: v_lshrrev_b32_e32 v1, 16, v1
+; GFX7-NEXT: v_and_b32_e32 v2, s7, v5
+; GFX7-NEXT: v_lshlrev_b32_e32 v1, s6, v1
+; GFX7-NEXT: v_or_b32_e32 v4, v2, v1
+; GFX7-NEXT: v_mov_b32_e32 v1, v4
+; GFX7-NEXT: v_mov_b32_e32 v2, v5
+; GFX7-NEXT: buffer_atomic_cmpswap v[1:2], v3, s[16:19], 0 offen glc
; GFX7-NEXT: s_waitcnt vmcnt(0)
; GFX7-NEXT: buffer_wbinvl1
-; GFX7-NEXT: v_cmp_eq_u32_e32 vcc, v4, v1
+; GFX7-NEXT: v_cmp_eq_u32_e32 vcc, v1, v5
; GFX7-NEXT: s_or_b64 s[4:5], vcc, s[4:5]
-; GFX7-NEXT: v_mov_b32_e32 v1, v4
; GFX7-NEXT: s_andn2_b64 exec, exec, s[4:5]
; GFX7-NEXT: s_cbranch_execnz .LBB17_1
; GFX7-NEXT: ; %bb.2: ; %atomicrmw.end
@@ -6070,34 +6075,34 @@ define void @buffer_fat_ptr_agent_atomic_fadd_noret_bf16__offset__amdgpu_no_fine
; GFX6-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX6-NEXT: s_addk_i32 s20, 0x200
; GFX6-NEXT: s_and_b32 s4, s20, -4
-; GFX6-NEXT: v_mov_b32_e32 v2, s4
-; GFX6-NEXT: buffer_load_dword v1, v2, s[16:19], 0 offen
+; GFX6-NEXT: v_mov_b32_e32 v3, s4
+; GFX6-NEXT: buffer_load_dword v1, v3, s[16:19], 0 offen
; GFX6-NEXT: s_and_b32 s4, s20, 3
; GFX6-NEXT: s_lshl_b32 s6, s4, 3
; GFX6-NEXT: s_lshl_b32 s4, 0xffff, s6
; GFX6-NEXT: v_mul_f32_e32 v0, 1.0, v0
; GFX6-NEXT: s_not_b32 s7, s4
; GFX6-NEXT: s_mov_b64 s[4:5], 0
-; GFX6-NEXT: v_and_b32_e32 v3, 0xffff0000, v0
+; GFX6-NEXT: v_and_b32_e32 v0, 0xffff0000, v0
; GFX6-NEXT: .LBB17_1: ; %atomicrmw.start
; GFX6-NEXT: ; =>This Inner Loop Header: Depth=1
; GFX6-NEXT: s_waitcnt vmcnt(0)
-; GFX6-NEXT: v_lshrrev_b32_e32 v0, s6, v1
-; GFX6-NEXT: v_lshlrev_b32_e32 v0, 16, v0
-; GFX6-NEXT: v_add_f32_e32 v0, v0, v3
-; GFX6-NEXT: v_lshrrev_b32_e32 v0, 16, v0
-; GFX6-NEXT: s_waitcnt expcnt(0)
-; GFX6-NEXT: v_and_b32_e32 v4, s7, v1
-; GFX6-NEXT: v_lshlrev_b32_e32 v0, s6, v0
-; GFX6-NEXT: v_or_b32_e32 v0, v4, v0
; GFX6-NEXT: v_mov_b32_e32 v5, v1
-; GFX6-NEXT: v_mov_b32_e32 v4, v0
-; GFX6-NEXT: buffer_atomic_cmpswap v[4:5], v2, s[16:19], 0 offen glc
+; GFX6-NEXT: s_waitcnt expcnt(0)
+; GFX6-NEXT: v_lshrrev_b32_e32 v1, s6, v5
+; GFX6-NEXT: v_lshlrev_b32_e32 v1, 16, v1
+; GFX6-NEXT: v_add_f32_e32 v1, v1, v0
+; GFX6-NEXT: v_lshrrev_b32_e32 v1, 16, v1
+; GFX6-NEXT: v_and_b32_e32 v2, s7, v5
+; GFX6-NEXT: v_lshlrev_b32_e32 v1, s6, v1
+; GFX6-NEXT: v_or_b32_e32 v4, v2, v1
+; GFX6-NEXT: v_mov_b32_e32 v1, v4
+; GFX6-NEXT: v_mov_b32_e32 v2, v5
+; GFX6-NEXT: buffer_atomic_cmpswap v[1:2], v3, s[16:19], 0 offen glc
; GFX6-NEXT: s_waitcnt vmcnt(0)
; GFX6-NEXT: buffer_wbinvl1
-; GFX6-NEXT: v_cmp_eq_u32_e32 vcc, v4, v1
+; GFX6-NEXT: v_cmp_eq_u32_e32 vcc, v1, v5
; GFX6-NEXT: s_or_b64 s[4:5], vcc, s[4:5]
-; GFX6-NEXT: v_mov_b32_e32 v1, v4
; GFX6-NEXT: s_andn2_b64 exec, exec, s[4:5]
; GFX6-NEXT: s_cbranch_execnz .LBB17_1
; GFX6-NEXT: ; %bb.2: ; %atomicrmw.end
@@ -6138,7 +6143,7 @@ define bfloat @buffer_fat_ptr_agent_atomic_fadd_ret_bf16__offset__waterfall__amd
; GFX12-TRUE16-NEXT: s_wait_alu 0xfffe
; GFX12-TRUE16-NEXT: s_and_saveexec_b32 s0, s0
; GFX12-TRUE16-NEXT: s_wait_loadcnt 0x0
-; GFX12-TRUE16-NEXT: buffer_load_b32 v6, v8, s[4:7], null offen
+; GFX12-TRUE16-NEXT: buffer_load_b32 v4, v8, s[4:7], null offen
; GFX12-TRUE16-NEXT: s_xor_b32 exec_lo, exec_lo, s0
; GFX12-TRUE16-NEXT: s_cbranch_execnz .LBB18_1
; GFX12-TRUE16-NEXT: ; %bb.2:
@@ -6149,26 +6154,27 @@ define bfloat @buffer_fat_ptr_agent_atomic_fadd_ret_bf16__offset__waterfall__amd
; GFX12-TRUE16-NEXT: ; =>This Loop Header: Depth=1
; GFX12-TRUE16-NEXT: ; Child Loop BB18_4 Depth 2
; GFX12-TRUE16-NEXT: s_wait_loadcnt 0x0
-; GFX12-TRUE16-NEXT: v_lshrrev_b32_e32 v4, v7, v6
+; GFX12-TRUE16-NEXT: v_mov_b32_e32 v6, v4
; GFX12-TRUE16-NEXT: s_mov_b32 s2, exec_lo
; GFX12-TRUE16-NEXT: s_wait_storecnt 0x0
; GFX12-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
+; GFX12-TRUE16-NEXT: v_lshrrev_b32_e32 v4, v7, v6
; GFX12-TRUE16-NEXT: v_lshlrev_b32_e32 v4, 16, v4
+; GFX12-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
; GFX12-TRUE16-NEXT: v_add_f32_e32 v4, v4, v10
-; GFX12-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(SKIP_2) | instid1(VALU_DEP_3)
; GFX12-TRUE16-NEXT: v_bfe_u32 v5, v4, 16, 1
; GFX12-TRUE16-NEXT: v_or_b32_e32 v11, 0x400000, v4
; GFX12-TRUE16-NEXT: v_cmp_u_f32_e32 vcc_lo, v4, v4
+; GFX12-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_3) | instskip(SKIP_1) | instid1(VALU_DEP_1)
; GFX12-TRUE16-NEXT: v_add3_u32 v5, v5, v4, 0x7fff
; GFX12-TRUE16-NEXT: s_wait_alu 0xfffd
-; GFX12-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(SKIP_1) | instid1(VALU_DEP_2)
; GFX12-TRUE16-NEXT: v_cndmask_b32_e32 v4, v5, v11, vcc_lo
; GFX12-TRUE16-NEXT: v_mov_b16_e32 v5.h, 0
+; GFX12-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_2) | instskip(NEXT) | instid1(VALU_DEP_1)
; GFX12-TRUE16-NEXT: v_mov_b16_e32 v5.l, v4.h
-; GFX12-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
; GFX12-TRUE16-NEXT: v_lshlrev_b32_e32 v4, v7, v5
+; GFX12-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
; GFX12-TRUE16-NEXT: v_and_or_b32 v5, v6, v9, v4
-; GFX12-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1)
; GFX12-TRUE16-NEXT: v_mov_b32_e32 v4, v5
; GFX12-TRUE16-NEXT: v_mov_b32_e32 v5, v6
; GFX12-TRUE16-NEXT: .LBB18_4: ; Parent Loop BB18_3 Depth=1
@@ -6191,7 +6197,6 @@ define bfloat @buffer_fat_ptr_agent_atomic_fadd_ret_bf16__offset__waterfall__amd
; GFX12-TRUE16-NEXT: s_mov_b32 exec_lo, s2
; GFX12-TRUE16-NEXT: s_wait_loadcnt 0x0
; GFX12-TRUE16-NEXT: v_cmp_eq_u32_e32 vcc_lo, v4, v6
-; GFX12-TRUE16-NEXT: v_mov_b32_e32 v6, v4
; GFX12-TRUE16-NEXT: global_inv scope:SCOPE_DEV
; GFX12-TRUE16-NEXT: s_or_b32 s1, vcc_lo, s1
; GFX12-TRUE16-NEXT: s_wait_alu 0xfffe
@@ -6231,7 +6236,7 @@ define bfloat @buffer_fat_ptr_agent_atomic_fadd_ret_bf16__offset__waterfall__amd
; GFX12-FAKE16-NEXT: s_wait_alu 0xfffe
; GFX12-FAKE16-NEXT: s_and_saveexec_b32 s0, s0
; GFX12-FAKE16-NEXT: s_wait_loadcnt 0x0
-; GFX12-FAKE16-NEXT: buffer_load_b32 v6, v8, s[4:7], null offen
+; GFX12-FAKE16-NEXT: buffer_load_b32 v4, v8, s[4:7], null offen
; GFX12-FAKE16-NEXT: s_xor_b32 exec_lo, exec_lo, s0
; GFX12-FAKE16-NEXT: s_cbranch_execnz .LBB18_1
; GFX12-FAKE16-NEXT: ; %bb.2:
@@ -6242,25 +6247,26 @@ define bfloat @buffer_fat_ptr_agent_atomic_fadd_ret_bf16__offset__waterfall__amd
; GFX12-FAKE16-NEXT: ; =>This Loop Header: Depth=1
; GFX12-FAKE16-NEXT: ; Child Loop BB18_4 Depth 2
; GFX12-FAKE16-NEXT: s_wait_loadcnt 0x0
-; GFX12-FAKE16-NEXT: v_lshrrev_b32_e32 v4, v7, v6
+; GFX12-FAKE16-NEXT: v_mov_b32_e32 v6, v4
; GFX12-FAKE16-NEXT: s_mov_b32 s2, exec_lo
; GFX12-FAKE16-NEXT: s_wait_storecnt 0x0
; GFX12-FAKE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
+; GFX12-FAKE16-NEXT: v_lshrrev_b32_e32 v4, v7, v6
; GFX12-FAKE16-NEXT: v_lshlrev_b32_e32 v4, 16, v4
+; GFX12-FAKE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
; GFX12-FAKE16-NEXT: v_add_f32_e32 v4, v4, v10
-; GFX12-FAKE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(SKIP_2) | instid1(VALU_DEP_3)
; GFX12-FAKE16-NEXT: v_bfe_u32 v5, v4, 16, 1
; GFX12-FAKE16-NEXT: v_or_b32_e32 v11, 0x400000, v4
; GFX12-FAKE16-NEXT: v_cmp_u_f32_e32 vcc_lo, v4, v4
+; GFX12-FAKE16-NEXT: s_delay_alu instid0(VALU_DEP_3) | instskip(SKIP_1) | instid1(VALU_DEP_1)
; GFX12-FAKE16-NEXT: v_add3_u32 v5, v5, v4, 0x7fff
; GFX12-FAKE16-NEXT: s_wait_alu 0xfffd
-; GFX12-FAKE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
; GFX12-FAKE16-NEXT: v_cndmask_b32_e32 v4, v5, v11, vcc_lo
-; GFX12-FAKE16-NEXT: v_lshrrev_b32_e32 v4, 16, v4
; GFX12-FAKE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
+; GFX12-FAKE16-NEXT: v_lshrrev_b32_e32 v4, 16, v4
; GFX12-FAKE16-NEXT: v_lshlrev_b32_e32 v4, v7, v4
+; GFX12-FAKE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
; GFX12-FAKE16-NEXT: v_and_or_b32 v5, v6, v9, v4
-; GFX12-FAKE16-NEXT: s_delay_alu instid0(VALU_DEP_1)
; GFX12-FAKE16-NEXT: v_mov_b32_e32 v4, v5
; GFX12-FAKE16-NEXT: v_mov_b32_e32 v5, v6
; GFX12-FAKE16-NEXT: .LBB18_4: ; Parent Loop BB18_3 Depth=1
@@ -6283,7 +6289,6 @@ define bfloat @buffer_fat_ptr_agent_atomic_fadd_ret_bf16__offset__waterfall__amd
; GFX12-FAKE16-NEXT: s_mov_b32 exec_lo, s2
; GFX12-FAKE16-NEXT: s_wait_loadcnt 0x0
; GFX12-FAKE16-NEXT: v_cmp_eq_u32_e32 vcc_lo, v4, v6
-; GFX12-FAKE16-NEXT: v_mov_b32_e32 v6, v4
; GFX12-FAKE16-NEXT: global_inv scope:SCOPE_DEV
; GFX12-FAKE16-NEXT: s_or_b32 s1, vcc_lo, s1
; GFX12-FAKE16-NEXT: s_wait_alu 0xfffe
@@ -6316,7 +6321,7 @@ define bfloat @buffer_fat_ptr_agent_atomic_fadd_ret_bf16__offset__waterfall__amd
; GFX942-NEXT: v_cmp_eq_u64_e64 s[0:1], s[6:7], v[2:3]
; GFX942-NEXT: s_and_b64 s[0:1], vcc, s[0:1]
; GFX942-NEXT: s_and_saveexec_b64 s[0:1], s[0:1]
-; GFX942-NEXT: buffer_load_dword v7, v9, s[4:7], 0 offen
+; GFX942-NEXT: buffer_load_dword v4, v9, s[4:7], 0 offen
; GFX942-NEXT: s_xor_b64 exec, exec, s[0:1]
; GFX942-NEXT: s_cbranch_execnz .LBB18_1
; GFX942-NEXT: ; %bb.2:
@@ -6328,6 +6333,7 @@ define bfloat @buffer_fat_ptr_agent_atomic_fadd_ret_bf16__offset__waterfall__amd
; GFX942-NEXT: ; =>This Loop Header: Depth=1
; GFX942-NEXT: ; Child Loop BB18_4 Depth 2
; GFX942-NEXT: s_waitcnt vmcnt(0)
+; GFX942-NEXT: v_mov_b32_e32 v7, v4
; GFX942-NEXT: v_lshrrev_b32_sdwa v4, v8, v7 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:DWORD
; GFX942-NEXT: s_mov_b64 s[8:9], exec
; GFX942-NEXT: v_add_f32_e32 v4, v4, v11
@@ -6361,7 +6367,6 @@ define bfloat @buffer_fat_ptr_agent_atomic_fadd_ret_bf16__offset__waterfall__amd
; GFX942-NEXT: s_waitcnt vmcnt(0)
; GFX942-NEXT: v_cmp_eq_u32_e32 vcc, v4, v7
; GFX942-NEXT: s_or_b64 s[2:3], vcc, s[2:3]
-; GFX942-NEXT: v_mov_b32_e32 v7, v4
; GFX942-NEXT: buffer_inv sc1
; GFX942-NEXT: s_andn2_b64 exec, exec, s[2:3]
; GFX942-NEXT: s_cbranch_execnz .LBB18_3
@@ -6394,7 +6399,7 @@ define bfloat @buffer_fat_ptr_agent_atomic_fadd_ret_bf16__offset__waterfall__amd
; GFX11-TRUE16-NEXT: s_and_b32 s0, vcc_lo, s0
; GFX11-TRUE16-NEXT: s_delay_alu instid0(SALU_CYCLE_1)
; GFX11-TRUE16-NEXT: s_and_saveexec_b32 s0, s0
-; GFX11-TRUE16-NEXT: buffer_load_b32 v6, v8, s[4:7], 0 offen
+; GFX11-TRUE16-NEXT: buffer_load_b32 v4, v8, s[4:7], 0 offen
; GFX11-TRUE16-NEXT: s_xor_b32 exec_lo, exec_lo, s0
; GFX11-TRUE16-NEXT: s_cbranch_execnz .LBB18_1
; GFX11-TRUE16-NEXT: ; %bb.2:
@@ -6406,25 +6411,26 @@ define bfloat @buffer_fat_ptr_agent_atomic_fadd_ret_bf16__offset__waterfall__amd
; GFX11-TRUE16-NEXT: ; =>This Loop Header: Depth=1
; GFX11-TRUE16-NEXT: ; Child Loop BB18_4 Depth 2
; GFX11-TRUE16-NEXT: s_waitcnt vmcnt(0)
-; GFX11-TRUE16-NEXT: v_lshrrev_b32_e32 v4, v7, v6
+; GFX11-TRUE16-NEXT: v_mov_b32_e32 v6, v4
; GFX11-TRUE16-NEXT: s_mov_b32 s2, exec_lo
; GFX11-TRUE16-NEXT: s_waitcnt_vscnt null, 0x0
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
+; GFX11-TRUE16-NEXT: v_lshrrev_b32_e32 v4, v7, v6
; GFX11-TRUE16-NEXT: v_lshlrev_b32_e32 v4, 16, v4
+; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
; GFX11-TRUE16-NEXT: v_add_f32_e32 v4, v4, v10
-; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(SKIP_2) | instid1(VALU_DEP_3)
; GFX11-TRUE16-NEXT: v_bfe_u32 v5, v4, 16, 1
; GFX11-TRUE16-NEXT: v_or_b32_e32 v11, 0x400000, v4
; GFX11-TRUE16-NEXT: v_cmp_u_f32_e32 vcc_lo, v4, v4
+; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_3) | instskip(NEXT) | instid1(VALU_DEP_1)
; GFX11-TRUE16-NEXT: v_add3_u32 v5, v5, v4, 0x7fff
-; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(SKIP_1) | instid1(VALU_DEP_2)
; GFX11-TRUE16-NEXT: v_cndmask_b32_e32 v4, v5, v11, vcc_lo
; GFX11-TRUE16-NEXT: v_mov_b16_e32 v5.h, 0
+; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_2) | instskip(NEXT) | instid1(VALU_DEP_1)
; GFX11-TRUE16-NEXT: v_mov_b16_e32 v5.l, v4.h
-; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
; GFX11-TRUE16-NEXT: v_lshlrev_b32_e32 v4, v7, v5
+; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
; GFX11-TRUE16-NEXT: v_and_or_b32 v5, v6, v9, v4
-; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1)
; GFX11-TRUE16-NEXT: v_mov_b32_e32 v4, v5
; GFX11-TRUE16-NEXT: v_mov_b32_e32 v5, v6
; GFX11-TRUE16-NEXT: .LBB18_4: ; Parent Loop BB18_3 Depth=1
@@ -6447,7 +6453,6 @@ define bfloat @buffer_fat_ptr_agent_atomic_fadd_ret_bf16__offset__waterfall__amd
; GFX11-TRUE16-NEXT: s_mov_b32 exec_lo, s2
; GFX11-TRUE16-NEXT: s_waitcnt vmcnt(0)
; GFX11-TRUE16-NEXT: v_cmp_eq_u32_e32 vcc_lo, v4, v6
-; GFX11-TRUE16-NEXT: v_mov_b32_e32 v6, v4
; GFX11-TRUE16-NEXT: buffer_gl1_inv
; GFX11-TRUE16-NEXT: buffer_gl0_inv
; GFX11-TRUE16-NEXT: s_or_b32 s1, vcc_lo, s1
@@ -6484,7 +6489,7 @@ define bfloat @buffer_fat_ptr_agent_atomic_fadd_ret_bf16__offset__waterfall__amd
; GFX11-FAKE16-NEXT: s_and_b32 s0, vcc_lo, s0
; GFX11-FAKE16-NEXT: s_delay_alu instid0(SALU_CYCLE_1)
; GFX11-FAKE16-NEXT: s_and_saveexec_b32 s0, s0
-; GFX11-FAKE16-NEXT: buffer_load_b32 v6, v8, s[4:7], 0 offen
+; GFX11-FAKE16-NEXT: buffer_load_b32 v4, v8, s[4:7], 0 offen
; GFX11-FAKE16-NEXT: s_xor_b32 exec_lo, exec_lo, s0
; GFX11-FAKE16-NEXT: s_cbranch_execnz .LBB18_1
; GFX11-FAKE16-NEXT: ; %bb.2:
@@ -6496,24 +6501,25 @@ define bfloat @buffer_fat_ptr_agent_atomic_fadd_ret_bf16__offset__waterfall__amd
; GFX11-FAKE16-NEXT: ; =>This Loop Header: Depth=1
; GFX11-FAKE16-NEXT: ; Child Loop BB18_4 Depth 2
; GFX11-FAKE16-NEXT: s_waitcnt vmcnt(0)
-; GFX11-FAKE16-NEXT: v_lshrrev_b32_e32 v4, v7, v6
+; GFX11-FAKE16-NEXT: v_mov_b32_e32 v6, v4
; GFX11-FAKE16-NEXT: s_mov_b32 s2, exec_lo
; GFX11-FAKE16-NEXT: s_waitcnt_vscnt null, 0x0
; GFX11-FAKE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
+; GFX11-FAKE16-NEXT: v_lshrrev_b32_e32 v4, v7, v6
; GFX11-FAKE16-NEXT: v_lshlrev_b32_e32 v4, 16, v4
+; GFX11-FAKE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
; GFX11-FAKE16-NEXT: v_add_f32_e32 v4, v4, v10
-; GFX11-FAKE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(SKIP_2) | instid1(VALU_DEP_3)
; GFX11-FAKE16-NEXT: v_bfe_u32 v5, v4, 16, 1
; GFX11-FAKE16-NEXT: v_or_b32_e32 v11, 0x400000, v4
; GFX11-FAKE16-NEXT: v_cmp_u_f32_e32 vcc_lo, v4, v4
+; GFX11-FAKE16-NEXT: s_delay_alu instid0(VALU_DEP_3) | instskip(NEXT) | instid1(VALU_DEP_1)
; GFX11-FAKE16-NEXT: v_add3_u32 v5, v5, v4, 0x7fff
-; GFX11-FAKE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
; GFX11-FAKE16-NEXT: v_cndmask_b32_e32 v4, v5, v11, vcc_lo
-; GFX11-FAKE16-NEXT: v_lshrrev_b32_e32 v4, 16, v4
; GFX11-FAKE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
+; GFX11-FAKE16-NEXT: v_lshrrev_b32_e32 v4, 16, v4
; GFX11-FAKE16-NEXT: v_lshlrev_b32_e32 v4, v7, v4
+; GFX11-FAKE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
; GFX11-FAKE16-NEXT: v_and_or_b32 v5, v6, v9, v4
-; GFX11-FAKE16-NEXT: s_delay_alu instid0(VALU_DEP_1)
; GFX11-FAKE16-NEXT: v_mov_b32_e32 v4, v5
; GFX11-FAKE16-NEXT: v_mov_b32_e32 v5, v6
; GFX11-FAKE16-NEXT: .LBB18_4: ; Parent Loop BB18_3 Depth=1
@@ -6536,7 +6542,6 @@ define bfloat @buffer_fat_ptr_agent_atomic_fadd_ret_bf16__offset__waterfall__amd
; GFX11-FAKE16-NEXT: s_mov_b32 exec_lo, s2
; GFX11-FAKE16-NEXT: s_waitcnt vmcnt(0)
; GFX11-FAKE16-NEXT: v_cmp_eq_u32_e32 vcc_lo, v4, v6
-; GFX11-FAKE16-NEXT: v_mov_b32_e32 v6, v4
; GFX11-FAKE16-NEXT: buffer_gl1_inv
; GFX11-FAKE16-NEXT: buffer_gl0_inv
; GFX11-FAKE16-NEXT: s_or_b32 s1, vcc_lo, s1
@@ -6569,7 +6574,7 @@ define bfloat @buffer_fat_ptr_agent_atomic_fadd_ret_bf16__offset__waterfall__amd
; GFX10-NEXT: v_cmp_eq_u64_e64 s4, s[10:11], v[2:3]
; GFX10-NEXT: s_and_b32 s4, vcc_lo, s4
; GFX10-NEXT: s_and_saveexec_b32 s4, s4
-; GFX10-NEXT: buffer_load_dword v6, v8, s[8:11], 0 offen
+; GFX10-NEXT: buffer_load_dword v4, v8, s[8:11], 0 offen
; GFX10-NEXT: s_waitcnt_depctr 0xffe3
; GFX10-NEXT: s_xor_b32 exec_lo, exec_lo, s4
; GFX10-NEXT: s_cbranch_execnz .LBB18_1
@@ -6580,9 +6585,10 @@ define bfloat @buffer_fat_ptr_agent_atomic_fadd_ret_bf16__offset__waterfall__amd
; GFX10-NEXT: ; =>This Loop Header: Depth=1
; GFX10-NEXT: ; Child Loop BB18_4 Depth 2
; GFX10-NEXT: s_waitcnt vmcnt(0)
-; GFX10-NEXT: v_lshrrev_b32_sdwa v4, v7, v6 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:DWORD
+; GFX10-NEXT: v_mov_b32_e32 v6, v4
; GFX10-NEXT: s_mov_b32 s6, exec_lo
; GFX10-NEXT: s_waitcnt_vscnt null, 0x0
+; GFX10-NEXT: v_lshrrev_b32_sdwa v4, v7, v6 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:DWORD
; GFX10-NEXT: v_add_f32_e32 v4, v4, v10
; GFX10-NEXT: v_bfe_u32 v5, v4, 16, 1
; GFX10-NEXT: v_or_b32_e32 v11, 0x400000, v4
@@ -6612,7 +6618,6 @@ define bfloat @buffer_fat_ptr_agent_atomic_fadd_ret_bf16__offset__waterfall__amd
; GFX10-NEXT: s_mov_b32 exec_lo, s6
; GFX10-NEXT: s_waitcnt vmcnt(0)
; GFX10-NEXT: v_cmp_eq_u32_e32 vcc_lo, v4, v6
-; GFX10-NEXT: v_mov_b32_e32 v6, v4
; GFX10-NEXT: buffer_gl1_inv
; GFX10-NEXT: buffer_gl0_inv
; GFX10-NEXT: s_or_b32 s5, vcc_lo, s5
@@ -6645,7 +6650,7 @@ define bfloat @buffer_fat_ptr_agent_atomic_fadd_ret_bf16__offset__waterfall__amd
; GFX90A-NEXT: s_and_b64 s[4:5], vcc, s[4:5]
; GFX90A-NEXT: s_and_saveexec_b64 s[4:5], s[4:5]
; GFX90A-NEXT: s_nop 0
-; GFX90A-NEXT: buffer_load_dword v7, v9, s[8:11], 0 offen
+; GFX90A-NEXT: buffer_load_dword v4, v9, s[8:11], 0 offen
; GFX90A-NEXT: s_xor_b64 exec, exec, s[4:5]
; GFX90A-NEXT: s_cbranch_execnz .LBB18_1
; GFX90A-NEXT: ; %bb.2:
@@ -6657,6 +6662,7 @@ define bfloat @buffer_fat_ptr_agent_atomic_fadd_ret_bf16__offset__waterfall__amd
; GFX90A-NEXT: ; =>This Loop Header: Depth=1
; GFX90A-NEXT: ; Child Loop BB18_4 Depth 2
; GFX90A-NEXT: s_waitcnt vmcnt(0)
+; GFX90A-NEXT: v_mov_b32_e32 v7, v4
; GFX90A-NEXT: v_lshrrev_b32_sdwa v4, v8, v7 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:DWORD
; GFX90A-NEXT: v_add_f32_e32 v4, v4, v11
; GFX90A-NEXT: v_bfe_u32 v5, v4, 16, 1
@@ -6687,7 +6693,6 @@ define bfloat @buffer_fat_ptr_agent_atomic_fadd_ret_bf16__offset__waterfall__amd
; GFX90A-NEXT: s_waitcnt vmcnt(0)
; GFX90A-NEXT: v_cmp_eq_u32_e32 vcc, v4, v7
; GFX90A-NEXT: s_or_b64 s[6:7], vcc, s[6:7]
-; GFX90A-NEXT: v_mov_b32_e32 v7, v4
; GFX90A-NEXT: buffer_wbinvl1
; GFX90A-NEXT: s_andn2_b64 exec, exec, s[6:7]
; GFX90A-NEXT: s_cbranch_execnz .LBB18_3
@@ -6717,7 +6722,7 @@ define bfloat @buffer_fat_ptr_agent_atomic_fadd_ret_bf16__offset__waterfall__amd
; GFX908-NEXT: s_and_b64 s[4:5], vcc, s[4:5]
; GFX908-NEXT: s_and_saveexec_b64 s[4:5], s[4:5]
; GFX908-NEXT: s_nop 0
-; GFX908-NEXT: buffer_load_dword v6, v8, s[8:11], 0 offen
+; GFX908-NEXT: buffer_load_dword v4, v8, s[8:11], 0 offen
; GFX908-NEXT: s_xor_b64 exec, exec, s[4:5]
; GFX908-NEXT: s_cbranch_execnz .LBB18_1
; GFX908-NEXT: ; %bb.2:
@@ -6729,6 +6734,7 @@ define bfloat @buffer_fat_ptr_agent_atomic_fadd_ret_bf16__offset__waterfall__amd
; GFX908-NEXT: ; =>This Loop Header: Depth=1
; GFX908-NEXT: ; Child Loop BB18_4 Depth 2
; GFX908-NEXT: s_waitcnt vmcnt(0)
+; GFX908-NEXT: v_mov_b32_e32 v6, v4
; GFX908-NEXT: v_lshrrev_b32_sdwa v4, v7, v6 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:DWORD
; GFX908-NEXT: v_add_f32_e32 v4, v4, v10
; GFX908-NEXT: v_bfe_u32 v5, v4, 16, 1
@@ -6760,7 +6766,6 @@ define bfloat @buffer_fat_ptr_agent_atomic_fadd_ret_bf16__offset__waterfall__amd
; GFX908-NEXT: s_waitcnt vmcnt(0)
; GFX908-NEXT: v_cmp_eq_u32_e32 vcc, v4, v6
; GFX908-NEXT: s_or_b64 s[6:7], vcc, s[6:7]
-; GFX908-NEXT: v_mov_b32_e32 v6, v4
; GFX908-NEXT: buffer_wbinvl1
; GFX908-NEXT: s_andn2_b64 exec, exec, s[6:7]
; GFX908-NEXT: s_cbranch_execnz .LBB18_3
@@ -6790,7 +6795,7 @@ define bfloat @buffer_fat_ptr_agent_atomic_fadd_ret_bf16__offset__waterfall__amd
; GFX8-NEXT: s_and_b64 s[4:5], vcc, s[4:5]
; GFX8-NEXT: s_and_saveexec_b64 s[4:5], s[4:5]
; GFX8-NEXT: s_nop 0
-; GFX8-NEXT: buffer_load_dword v6, v8, s[8:11], 0 offen
+; GFX8-NEXT: buffer_load_dword v4, v8, s[8:11], 0 offen
; GFX8-NEXT: s_xor_b64 exec, exec, s[4:5]
; GFX8-NEXT: s_cbranch_execnz .LBB18_1
; GFX8-NEXT: ; %bb.2:
@@ -6801,6 +6806,7 @@ define bfloat @buffer_fat_ptr_agent_atomic_fadd_ret_bf16__offset__waterfall__amd
; GFX8-NEXT: ; =>This Loop Header: Depth=1
; GFX8-NEXT: ; Child Loop BB18_4 Depth 2
; GFX8-NEXT: s_waitcnt vmcnt(0)
+; GFX8-NEXT: v_mov_b32_e32 v6, v4
; GFX8-NEXT: v_lshrrev_b32_sdwa v4, v7, v6 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:DWORD
; GFX8-NEXT: v_add_f32_e32 v4, v4, v10
; GFX8-NEXT: v_bfe_u32 v5, v4, 16, 1
@@ -6834,7 +6840,6 @@ define bfloat @buffer_fat_ptr_agent_atomic_fadd_ret_bf16__offset__waterfall__amd
; GFX8-NEXT: s_waitcnt vmcnt(0)
; GFX8-NEXT: v_cmp_eq_u32_e32 vcc, v4, v6
; GFX8-NEXT: s_or_b64 s[6:7], vcc, s[6:7]
-; GFX8-NEXT: v_mov_b32_e32 v6, v4
; GFX8-NEXT: buffer_wbinvl1
; GFX8-NEXT: s_andn2_b64 exec, exec, s[6:7]
; GFX8-NEXT: s_cbranch_execnz .LBB18_3
@@ -6862,18 +6867,19 @@ define bfloat @buffer_fat_ptr_agent_atomic_fadd_ret_bf16__offset__waterfall__amd
; GFX7-NEXT: v_cmp_eq_u64_e64 s[4:5], s[10:11], v[2:3]
; GFX7-NEXT: s_and_b64 s[4:5], vcc, s[4:5]
; GFX7-NEXT: s_and_saveexec_b64 s[4:5], s[4:5]
-; GFX7-NEXT: buffer_load_dword v6, v8, s[8:11], 0 offen
+; GFX7-NEXT: buffer_load_dword v4, v8, s[8:11], 0 offen
; GFX7-NEXT: s_xor_b64 exec, exec, s[4:5]
; GFX7-NEXT: s_cbranch_execnz .LBB18_1
; GFX7-NEXT: ; %bb.2:
; GFX7-NEXT: s_mov_b64 exec, s[6:7]
-; GFX7-NEXT: v_mul_f32_e32 v4, 1.0, v5
+; GFX7-NEXT: v_mul_f32_e32 v5, 1.0, v5
; GFX7-NEXT: s_mov_b64 s[6:7], 0
-; GFX7-NEXT: v_and_b32_e32 v10, 0xffff0000, v4
+; GFX7-NEXT: v_and_b32_e32 v10, 0xffff0000, v5
; GFX7-NEXT: .LBB18_3: ; %atomicrmw.start
; GFX7-NEXT: ; =>This Loop Header: Depth=1
; GFX7-NEXT: ; Child Loop BB18_4 Depth 2
; GFX7-NEXT: s_waitcnt vmcnt(0)
+; GFX7-NEXT: v_mov_b32_e32 v6, v4
; GFX7-NEXT: v_lshrrev_b32_e32 v4, v7, v6
; GFX7-NEXT: v_lshlrev_b32_e32 v4, 16, v4
; GFX7-NEXT: v_add_f32_e32 v4, v4, v10
@@ -6903,7 +6909,6 @@ define bfloat @buffer_fat_ptr_agent_atomic_fadd_ret_bf16__offset__waterfall__amd
; GFX7-NEXT: s_waitcnt vmcnt(0)
; GFX7-NEXT: v_cmp_eq_u32_e32 vcc, v4, v6
; GFX7-NEXT: s_or_b64 s[6:7], vcc, s[6:7]
-; GFX7-NEXT: v_mov_b32_e32 v6, v4
; GFX7-NEXT: buffer_wbinvl1
; GFX7-NEXT: s_andn2_b64 exec, exec, s[6:7]
; GFX7-NEXT: s_cbranch_execnz .LBB18_3
@@ -6932,18 +6937,20 @@ define bfloat @buffer_fat_ptr_agent_atomic_fadd_ret_bf16__offset__waterfall__amd
; GFX6-NEXT: v_cmp_eq_u64_e64 s[4:5], s[10:11], v[2:3]
; GFX6-NEXT: s_and_b64 s[4:5], vcc, s[4:5]
; GFX6-NEXT: s_and_saveexec_b64 s[4:5], s[4:5]
-; GFX6-NEXT: buffer_load_dword v6, v8, s[8:11], 0 offen
+; GFX6-NEXT: buffer_load_dword v4, v8, s[8:11], 0 offen
; GFX6-NEXT: s_xor_b64 exec, exec, s[4:5]
; GFX6-NEXT: s_cbranch_execnz .LBB18_1
; GFX6-NEXT: ; %bb.2:
; GFX6-NEXT: s_mov_b64 exec, s[6:7]
-; GFX6-NEXT: v_mul_f32_e32 v4, 1.0, v5
+; GFX6-NEXT: v_mul_f32_e32 v5, 1.0, v5
; GFX6-NEXT: s_mov_b64 s[6:7], 0
-; GFX6-NEXT: v_and_b32_e32 v10, 0xffff0000, v4
+; GFX6-NEXT: v_and_b32_e32 v10, 0xffff0000, v5
; GFX6-NEXT: .LBB18_3: ; %atomicrmw.start
; GFX6-NEXT: ; =>This Loop Header: Depth=1
; GFX6-NEXT: ; Child Loop BB18_4 Depth 2
-; GFX6-NEXT: s_waitcnt vmcnt(0) expcnt(0)
+; GFX6-NEXT: s_waitcnt vmcnt(0)
+; GFX6-NEXT: v_mov_b32_e32 v6, v4
+; GFX6-NEXT: s_waitcnt expcnt(0)
; GFX6-NEXT: v_lshrrev_b32_e32 v4, v7, v6
; GFX6-NEXT: v_lshlrev_b32_e32 v4, 16, v4
; GFX6-NEXT: v_add_f32_e32 v4, v4, v10
@@ -6973,7 +6980,6 @@ define bfloat @buffer_fat_ptr_agent_atomic_fadd_ret_bf16__offset__waterfall__amd
; GFX6-NEXT: s_waitcnt vmcnt(0)
; GFX6-NEXT: v_cmp_eq_u32_e32 vcc, v4, v6
; GFX6-NEXT: s_or_b64 s[6:7], vcc, s[6:7]
-; GFX6-NEXT: v_mov_b32_e32 v6, v4
; GFX6-NEXT: buffer_wbinvl1
; GFX6-NEXT: s_andn2_b64 exec, exec, s[6:7]
; GFX6-NEXT: s_cbranch_execnz .LBB18_3
@@ -7274,21 +7280,20 @@ define void @buffer_fat_ptr_agent_atomic_fadd_noret_v2f16__offset__amdgpu_no_fin
; GFX11-NEXT: s_delay_alu instid0(SALU_CYCLE_1)
; GFX11-NEXT: v_mov_b32_e32 v3, s4
; GFX11-NEXT: s_mov_b32 s4, 0
-; GFX11-NEXT: buffer_load_b32 v2, v1, s[0:3], 0 offen offset:1024
+; GFX11-NEXT: buffer_load_b32 v1, v1, s[0:3], 0 offen offset:1024
; GFX11-NEXT: .LBB20_1: ; %atomicrmw.start
; GFX11-NEXT: ; =>This Inner Loop Header: Depth=1
; GFX11-NEXT: s_waitcnt vmcnt(0)
-; GFX11-NEXT: v_pk_add_f16 v1, v2, v0
-; GFX11-NEXT: v_mov_b32_e32 v5, v2
+; GFX11-NEXT: v_mov_b32_e32 v5, v1
; GFX11-NEXT: s_waitcnt_vscnt null, 0x0
-; GFX11-NEXT: s_delay_alu instid0(VALU_DEP_2)
-; GFX11-NEXT: v_mov_b32_e32 v4, v1
-; GFX11-NEXT: buffer_atomic_cmpswap_b32 v[4:5], v3, s[0:3], 0 offen glc
+; GFX11-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
+; GFX11-NEXT: v_pk_add_f16 v4, v5, v0
+; GFX11-NEXT: v_dual_mov_b32 v1, v4 :: v_dual_mov_b32 v2, v5
+; GFX11-NEXT: buffer_atomic_cmpswap_b32 v[1:2], v3, s[0:3], 0 offen glc
; GFX11-NEXT: s_waitcnt vmcnt(0)
; GFX11-NEXT: buffer_gl1_inv
; GFX11-NEXT: buffer_gl0_inv
-; GFX11-NEXT: v_cmp_eq_u32_e32 vcc_lo, v4, v2
-; GFX11-NEXT: v_mov_b32_e32 v2, v4
+; GFX11-NEXT: v_cmp_eq_u32_e32 vcc_lo, v1, v5
; GFX11-NEXT: s_or_b32 s4, vcc_lo, s4
; GFX11-NEXT: s_delay_alu instid0(SALU_CYCLE_1)
; GFX11-NEXT: s_and_not1_b32 exec_lo, exec_lo, s4
@@ -7304,20 +7309,20 @@ define void @buffer_fat_ptr_agent_atomic_fadd_noret_v2f16__offset__amdgpu_no_fin
; GFX10-NEXT: s_add_i32 s4, s20, 0x400
; GFX10-NEXT: v_mov_b32_e32 v3, s4
; GFX10-NEXT: s_mov_b32 s4, 0
-; GFX10-NEXT: buffer_load_dword v2, v1, s[16:19], 0 offen offset:1024
+; GFX10-NEXT: buffer_load_dword v1, v1, s[16:19], 0 offen offset:1024
; GFX10-NEXT: .LBB20_1: ; %atomicrmw.start
; GFX10-NEXT: ; =>This Inner Loop Header: Depth=1
; GFX10-NEXT: s_waitcnt vmcnt(0)
-; GFX10-NEXT: v_pk_add_f16 v1, v2, v0
-; GFX10-NEXT: v_mov_b32_e32 v5, v2
+; GFX10-NEXT: v_mov_b32_e32 v5, v1
; GFX10-NEXT: s_waitcnt_vscnt null, 0x0
-; GFX10-NEXT: v_mov_b32_e32 v4, v1
-; GFX10-NEXT: buffer_atomic_cmpswap v[4:5], v3, s[16:19], 0 offen glc
+; GFX10-NEXT: v_pk_add_f16 v4, v5, v0
+; GFX10-NEXT: v_mov_b32_e32 v1, v4
+; GFX10-NEXT: v_mov_b32_e32 v2, v5
+; GFX10-NEXT: buffer_atomic_cmpswap v[1:2], v3, s[16:19], 0 offen glc
; GFX10-NEXT: s_waitcnt vmcnt(0)
; GFX10-NEXT: buffer_gl1_inv
; GFX10-NEXT: buffer_gl0_inv
-; GFX10-NEXT: v_cmp_eq_u32_e32 vcc_lo, v4, v2
-; GFX10-NEXT: v_mov_b32_e32 v2, v4
+; GFX10-NEXT: v_cmp_eq_u32_e32 vcc_lo, v1, v5
; GFX10-NEXT: s_or_b32 s4, vcc_lo, s4
; GFX10-NEXT: s_andn2_b32 exec_lo, exec_lo, s4
; GFX10-NEXT: s_cbranch_execnz .LBB20_1
@@ -7347,24 +7352,24 @@ define void @buffer_fat_ptr_agent_atomic_fadd_noret_v2f16__offset__amdgpu_no_fin
; GFX8: ; %bb.0:
; GFX8-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX8-NEXT: v_mov_b32_e32 v1, s20
-; GFX8-NEXT: buffer_load_dword v2, v1, s[16:19], 0 offen offset:1024
+; GFX8-NEXT: buffer_load_dword v1, v1, s[16:19], 0 offen offset:1024
; GFX8-NEXT: s_add_i32 s6, s20, 0x400
; GFX8-NEXT: s_mov_b64 s[4:5], 0
; GFX8-NEXT: v_mov_b32_e32 v3, s6
; GFX8-NEXT: .LBB20_1: ; %atomicrmw.start
; GFX8-NEXT: ; =>This Inner Loop Header: Depth=1
; GFX8-NEXT: s_waitcnt vmcnt(0)
-; GFX8-NEXT: v_add_f16_sdwa v1, v2, v0 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:WORD_1 src1_sel:WORD_1
-; GFX8-NEXT: v_add_f16_e32 v4, v2, v0
-; GFX8-NEXT: v_or_b32_e32 v1, v4, v1
-; GFX8-NEXT: v_mov_b32_e32 v5, v2
-; GFX8-NEXT: v_mov_b32_e32 v4, v1
-; GFX8-NEXT: buffer_atomic_cmpswap v[4:5], v3, s[16:19], 0 offen glc
+; GFX8-NEXT: v_mov_b32_e32 v5, v1
+; GFX8-NEXT: v_add_f16_sdwa v1, v5, v0 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:WORD_1 src1_sel:WORD_1
+; GFX8-NEXT: v_add_f16_e32 v2, v5, v0
+; GFX8-NEXT: v_or_b32_e32 v4, v2, v1
+; GFX8-NEXT: v_mov_b32_e32 v1, v4
+; GFX8-NEXT: v_mov_b32_e32 v2, v5
+; GFX8-NEXT: buffer_atomic_cmpswap v[1:2], v3, s[16:19], 0 offen glc
; GFX8-NEXT: s_waitcnt vmcnt(0)
; GFX8-NEXT: buffer_wbinvl1
-; GFX8-NEXT: v_cmp_eq_u32_e32 vcc, v4, v2
+; GFX8-NEXT: v_cmp_eq_u32_e32 vcc, v1, v5
; GFX8-NEXT: s_or_b64 s[4:5], vcc, s[4:5]
-; GFX8-NEXT: v_mov_b32_e32 v2, v4
; GFX8-NEXT: s_andn2_b64 exec, exec, s[4:5]
; GFX8-NEXT: s_cbranch_execnz .LBB20_1
; GFX8-NEXT: ; %bb.2: ; %atomicrmw.end
@@ -8265,21 +8270,20 @@ define void @buffer_fat_ptr_agent_atomic_fadd_noret_v2f16__offset(ptr addrspace(
; GFX11-NEXT: s_delay_alu instid0(SALU_CYCLE_1)
; GFX11-NEXT: v_mov_b32_e32 v3, s4
; GFX11-NEXT: s_mov_b32 s4, 0
-; GFX11-NEXT: buffer_load_b32 v2, v1, s[0:3], 0 offen offset:1024
+; GFX11-NEXT: buffer_load_b32 v1, v1, s[0:3], 0 offen offset:1024
; GFX11-NEXT: .LBB23_1: ; %atomicrmw.start
; GFX11-NEXT: ; =>This Inner Loop Header: Depth=1
; GFX11-NEXT: s_waitcnt vmcnt(0)
-; GFX11-NEXT: v_pk_add_f16 v1, v2, v0
-; GFX11-NEXT: v_mov_b32_e32 v5, v2
+; GFX11-NEXT: v_mov_b32_e32 v5, v1
; GFX11-NEXT: s_waitcnt_vscnt null, 0x0
-; GFX11-NEXT: s_delay_alu instid0(VALU_DEP_2)
-; GFX11-NEXT: v_mov_b32_e32 v4, v1
-; GFX11-NEXT: buffer_atomic_cmpswap_b32 v[4:5], v3, s[0:3], 0 offen glc
+; GFX11-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
+; GFX11-NEXT: v_pk_add_f16 v4, v5, v0
+; GFX11-NEXT: v_dual_mov_b32 v1, v4 :: v_dual_mov_b32 v2, v5
+; GFX11-NEXT: buffer_atomic_cmpswap_b32 v[1:2], v3, s[0:3], 0 offen glc
; GFX11-NEXT: s_waitcnt vmcnt(0)
; GFX11-NEXT: buffer_gl1_inv
; GFX11-NEXT: buffer_gl0_inv
-; GFX11-NEXT: v_cmp_eq_u32_e32 vcc_lo, v4, v2
-; GFX11-NEXT: v_mov_b32_e32 v2, v4
+; GFX11-NEXT: v_cmp_eq_u32_e32 vcc_lo, v1, v5
; GFX11-NEXT: s_or_b32 s4, vcc_lo, s4
; GFX11-NEXT: s_delay_alu instid0(SALU_CYCLE_1)
; GFX11-NEXT: s_and_not1_b32 exec_lo, exec_lo, s4
@@ -8295,20 +8299,20 @@ define void @buffer_fat_ptr_agent_atomic_fadd_noret_v2f16__offset(ptr addrspace(
; GFX10-NEXT: s_add_i32 s4, s20, 0x400
; GFX10-NEXT: v_mov_b32_e32 v3, s4
; GFX10-NEXT: s_mov_b32 s4, 0
-; GFX10-NEXT: buffer_load_dword v2, v1, s[16:19], 0 offen offset:1024
+; GFX10-NEXT: buffer_load_dword v1, v1, s[16:19], 0 offen offset:1024
; GFX10-NEXT: .LBB23_1: ; %atomicrmw.start
; GFX10-NEXT: ; =>This Inner Loop Header: Depth=1
; GFX10-NEXT: s_waitcnt vmcnt(0)
-; GFX10-NEXT: v_pk_add_f16 v1, v2, v0
-; GFX10-NEXT: v_mov_b32_e32 v5, v2
+; GFX10-NEXT: v_mov_b32_e32 v5, v1
; GFX10-NEXT: s_waitcnt_vscnt null, 0x0
-; GFX10-NEXT: v_mov_b32_e32 v4, v1
-; GFX10-NEXT: buffer_atomic_cmpswap v[4:5], v3, s[16:19], 0 offen glc
+; GFX10-NEXT: v_pk_add_f16 v4, v5, v0
+; GFX10-NEXT: v_mov_b32_e32 v1, v4
+; GFX10-NEXT: v_mov_b32_e32 v2, v5
+; GFX10-NEXT: buffer_atomic_cmpswap v[1:2], v3, s[16:19], 0 offen glc
; GFX10-NEXT: s_waitcnt vmcnt(0)
; GFX10-NEXT: buffer_gl1_inv
; GFX10-NEXT: buffer_gl0_inv
-; GFX10-NEXT: v_cmp_eq_u32_e32 vcc_lo, v4, v2
-; GFX10-NEXT: v_mov_b32_e32 v2, v4
+; GFX10-NEXT: v_cmp_eq_u32_e32 vcc_lo, v1, v5
; GFX10-NEXT: s_or_b32 s4, vcc_lo, s4
; GFX10-NEXT: s_andn2_b32 exec_lo, exec_lo, s4
; GFX10-NEXT: s_cbranch_execnz .LBB23_1
@@ -8320,21 +8324,21 @@ define void @buffer_fat_ptr_agent_atomic_fadd_noret_v2f16__offset(ptr addrspace(
; GFX90A: ; %bb.0:
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX90A-NEXT: v_mov_b32_e32 v1, s20
-; GFX90A-NEXT: buffer_load_dword v3, v1, s[16:19], 0 offen offset:1024
+; GFX90A-NEXT: buffer_load_dword v2, v1, s[16:19], 0 offen offset:1024
; GFX90A-NEXT: s_add_i32 s6, s20, 0x400
; GFX90A-NEXT: s_mov_b64 s[4:5], 0
; GFX90A-NEXT: v_mov_b32_e32 v1, s6
; GFX90A-NEXT: .LBB23_1: ; %atomicrmw.start
; GFX90A-NEXT: ; =>This Inner Loop Header: Depth=1
; GFX90A-NEXT: s_waitcnt vmcnt(0)
-; GFX90A-NEXT: v_pk_add_f16 v2, v3, v0
-; GFX90A-NEXT: v_pk_mov_b32 v[4:5], v[2:3], v[2:3] op_sel:[0,1]
-; GFX90A-NEXT: buffer_atomic_cmpswap v[4:5], v1, s[16:19], 0 offen glc
+; GFX90A-NEXT: v_mov_b32_e32 v5, v2
+; GFX90A-NEXT: v_pk_add_f16 v4, v5, v0
+; GFX90A-NEXT: v_pk_mov_b32 v[2:3], v[4:5], v[4:5] op_sel:[0,1]
+; GFX90A-NEXT: buffer_atomic_cmpswap v[2:3], v1, s[16:19], 0 offen glc
; GFX90A-NEXT: s_waitcnt vmcnt(0)
; GFX90A-NEXT: buffer_wbinvl1
-; GFX90A-NEXT: v_cmp_eq_u32_e32 vcc, v4, v3
+; GFX90A-NEXT: v_cmp_eq_u32_e32 vcc, v2, v5
; GFX90A-NEXT: s_or_b64 s[4:5], vcc, s[4:5]
-; GFX90A-NEXT: v_mov_b32_e32 v3, v4
; GFX90A-NEXT: s_andn2_b64 exec, exec, s[4:5]
; GFX90A-NEXT: s_cbranch_execnz .LBB23_1
; GFX90A-NEXT: ; %bb.2: ; %atomicrmw.end
@@ -8345,22 +8349,22 @@ define void @buffer_fat_ptr_agent_atomic_fadd_noret_v2f16__offset(ptr addrspace(
; GFX908: ; %bb.0:
; GFX908-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX908-NEXT: v_mov_b32_e32 v1, s20
-; GFX908-NEXT: buffer_load_dword v2, v1, s[16:19], 0 offen offset:1024
+; GFX908-NEXT: buffer_load_dword v1, v1, s[16:19], 0 offen offset:1024
; GFX908-NEXT: s_add_i32 s6, s20, 0x400
; GFX908-NEXT: s_mov_b64 s[4:5], 0
; GFX908-NEXT: v_mov_b32_e32 v3, s6
; GFX908-NEXT: .LBB23_1: ; %atomicrmw.start
; GFX908-NEXT: ; =>This Inner Loop Header: Depth=1
; GFX908-NEXT: s_waitcnt vmcnt(0)
-; GFX908-NEXT: v_pk_add_f16 v1, v2, v0
-; GFX908-NEXT: v_mov_b32_e32 v5, v2
-; GFX908-NEXT: v_mov_b32_e32 v4, v1
-; GFX908-NEXT: buffer_atomic_cmpswap v[4:5], v3, s[16:19], 0 offen glc
+; GFX908-NEXT: v_mov_b32_e32 v5, v1
+; GFX908-NEXT: v_pk_add_f16 v4, v5, v0
+; GFX908-NEXT: v_mov_b32_e32 v1, v4
+; GFX908-NEXT: v_mov_b32_e32 v2, v5
+; GFX908-NEXT: buffer_atomic_cmpswap v[1:2], v3, s[16:19], 0 offen glc
; GFX908-NEXT: s_waitcnt vmcnt(0)
; GFX908-NEXT: buffer_wbinvl1
-; GFX908-NEXT: v_cmp_eq_u32_e32 vcc, v4, v2
+; GFX908-NEXT: v_cmp_eq_u32_e32 vcc, v1, v5
; GFX908-NEXT: s_or_b64 s[4:5], vcc, s[4:5]
-; GFX908-NEXT: v_mov_b32_e32 v2, v4
; GFX908-NEXT: s_andn2_b64 exec, exec, s[4:5]
; GFX908-NEXT: s_cbranch_execnz .LBB23_1
; GFX908-NEXT: ; %bb.2: ; %atomicrmw.end
@@ -8371,24 +8375,24 @@ define void @buffer_fat_ptr_agent_atomic_fadd_noret_v2f16__offset(ptr addrspace(
; GFX8: ; %bb.0:
; GFX8-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX8-NEXT: v_mov_b32_e32 v1, s20
-; GFX8-NEXT: buffer_load_dword v2, v1, s[16:19], 0 offen offset:1024
+; GFX8-NEXT: buffer_load_dword v1, v1, s[16:19], 0 offen offset:1024
; GFX8-NEXT: s_add_i32 s6, s20, 0x400
; GFX8-NEXT: s_mov_b64 s[4:5], 0
; GFX8-NEXT: v_mov_b32_e32 v3, s6
; GFX8-NEXT: .LBB23_1: ; %atomicrmw.start
; GFX8-NEXT: ; =>This Inner Loop Header: Depth=1
; GFX8-NEXT: s_waitcnt vmcnt(0)
-; GFX8-NEXT: v_add_f16_sdwa v1, v2, v0 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:WORD_1 src1_sel:WORD_1
-; GFX8-NEXT: v_add_f16_e32 v4, v2, v0
-; GFX8-NEXT: v_or_b32_e32 v1, v4, v1
-; GFX8-NEXT: v_mov_b32_e32 v5, v2
-; GFX8-NEXT: v_mov_b32_e32 v4, v1
-; GFX8-NEXT: buffer_atomic_cmpswap v[4:5], v3, s[16:19], 0 offen glc
+; GFX8-NEXT: v_mov_b32_e32 v5, v1
+; GFX8-NEXT: v_add_f16_sdwa v1, v5, v0 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:WORD_1 src1_sel:WORD_1
+; GFX8-NEXT: v_add_f16_e32 v2, v5, v0
+; GFX8-NEXT: v_or_b32_e32 v4, v2, v1
+; GFX8-NEXT: v_mov_b32_e32 v1, v4
+; GFX8-NEXT: v_mov_b32_e32 v2, v5
+; GFX8-NEXT: buffer_atomic_cmpswap v[1:2], v3, s[16:19], 0 offen glc
; GFX8-NEXT: s_waitcnt vmcnt(0)
; GFX8-NEXT: buffer_wbinvl1
-; GFX8-NEXT: v_cmp_eq_u32_e32 vcc, v4, v2
+; GFX8-NEXT: v_cmp_eq_u32_e32 vcc, v1, v5
; GFX8-NEXT: s_or_b64 s[4:5], vcc, s[4:5]
-; GFX8-NEXT: v_mov_b32_e32 v2, v4
; GFX8-NEXT: s_andn2_b64 exec, exec, s[4:5]
; GFX8-NEXT: s_cbranch_execnz .LBB23_1
; GFX8-NEXT: ; %bb.2: ; %atomicrmw.end
@@ -8792,21 +8796,20 @@ define void @buffer_fat_ptr_agent_atomic_fadd_noret_v2f16__offset__amdgpu_no_rem
; GFX11-NEXT: s_delay_alu instid0(SALU_CYCLE_1)
; GFX11-NEXT: v_mov_b32_e32 v3, s4
; GFX11-NEXT: s_mov_b32 s4, 0
-; GFX11-NEXT: buffer_load_b32 v2, v1, s[0:3], 0 offen offset:1024
+; GFX11-NEXT: buffer_load_b32 v1, v1, s[0:3], 0 offen offset:1024
; GFX11-NEXT: .LBB25_1: ; %atomicrmw.start
; GFX11-NEXT: ; =>This Inner Loop Header: Depth=1
; GFX11-NEXT: s_waitcnt vmcnt(0)
-; GFX11-NEXT: v_pk_add_f16 v1, v2, v0
-; GFX11-NEXT: v_mov_b32_e32 v5, v2
+; GFX11-NEXT: v_mov_b32_e32 v5, v1
; GFX11-NEXT: s_waitcnt_vscnt null, 0x0
-; GFX11-NEXT: s_delay_alu instid0(VALU_DEP_2)
-; GFX11-NEXT: v_mov_b32_e32 v4, v1
-; GFX11-NEXT: buffer_atomic_cmpswap_b32 v[4:5], v3, s[0:3], 0 offen glc
+; GFX11-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
+; GFX11-NEXT: v_pk_add_f16 v4, v5, v0
+; GFX11-NEXT: v_dual_mov_b32 v1, v4 :: v_dual_mov_b32 v2, v5
+; GFX11-NEXT: buffer_atomic_cmpswap_b32 v[1:2], v3, s[0:3], 0 offen glc
; GFX11-NEXT: s_waitcnt vmcnt(0)
; GFX11-NEXT: buffer_gl1_inv
; GFX11-NEXT: buffer_gl0_inv
-; GFX11-NEXT: v_cmp_eq_u32_e32 vcc_lo, v4, v2
-; GFX11-NEXT: v_mov_b32_e32 v2, v4
+; GFX11-NEXT: v_cmp_eq_u32_e32 vcc_lo, v1, v5
; GFX11-NEXT: s_or_b32 s4, vcc_lo, s4
; GFX11-NEXT: s_delay_alu instid0(SALU_CYCLE_1)
; GFX11-NEXT: s_and_not1_b32 exec_lo, exec_lo, s4
@@ -8822,20 +8825,20 @@ define void @buffer_fat_ptr_agent_atomic_fadd_noret_v2f16__offset__amdgpu_no_rem
; GFX10-NEXT: s_add_i32 s4, s20, 0x400
; GFX10-NEXT: v_mov_b32_e32 v3, s4
; GFX10-NEXT: s_mov_b32 s4, 0
-; GFX10-NEXT: buffer_load_dword v2, v1, s[16:19], 0 offen offset:1024
+; GFX10-NEXT: buffer_load_dword v1, v1, s[16:19], 0 offen offset:1024
; GFX10-NEXT: .LBB25_1: ; %atomicrmw.start
; GFX10-NEXT: ; =>This Inner Loop Header: Depth=1
; GFX10-NEXT: s_waitcnt vmcnt(0)
-; GFX10-NEXT: v_pk_add_f16 v1, v2, v0
-; GFX10-NEXT: v_mov_b32_e32 v5, v2
+; GFX10-NEXT: v_mov_b32_e32 v5, v1
; GFX10-NEXT: s_waitcnt_vscnt null, 0x0
-; GFX10-NEXT: v_mov_b32_e32 v4, v1
-; GFX10-NEXT: buffer_atomic_cmpswap v[4:5], v3, s[16:19], 0 offen glc
+; GFX10-NEXT: v_pk_add_f16 v4, v5, v0
+; GFX10-NEXT: v_mov_b32_e32 v1, v4
+; GFX10-NEXT: v_mov_b32_e32 v2, v5
+; GFX10-NEXT: buffer_atomic_cmpswap v[1:2], v3, s[16:19], 0 offen glc
; GFX10-NEXT: s_waitcnt vmcnt(0)
; GFX10-NEXT: buffer_gl1_inv
; GFX10-NEXT: buffer_gl0_inv
-; GFX10-NEXT: v_cmp_eq_u32_e32 vcc_lo, v4, v2
-; GFX10-NEXT: v_mov_b32_e32 v2, v4
+; GFX10-NEXT: v_cmp_eq_u32_e32 vcc_lo, v1, v5
; GFX10-NEXT: s_or_b32 s4, vcc_lo, s4
; GFX10-NEXT: s_andn2_b32 exec_lo, exec_lo, s4
; GFX10-NEXT: s_cbranch_execnz .LBB25_1
@@ -8847,21 +8850,21 @@ define void @buffer_fat_ptr_agent_atomic_fadd_noret_v2f16__offset__amdgpu_no_rem
; GFX90A: ; %bb.0:
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX90A-NEXT: v_mov_b32_e32 v1, s20
-; GFX90A-NEXT: buffer_load_dword v3, v1, s[16:19], 0 offen offset:1024
+; GFX90A-NEXT: buffer_load_dword v2, v1, s[16:19], 0 offen offset:1024
; GFX90A-NEXT: s_add_i32 s6, s20, 0x400
; GFX90A-NEXT: s_mov_b64 s[4:5], 0
; GFX90A-NEXT: v_mov_b32_e32 v1, s6
; GFX90A-NEXT: .LBB25_1: ; %atomicrmw.start
; GFX90A-NEXT: ; =>This Inner Loop Header: Depth=1
; GFX90A-NEXT: s_waitcnt vmcnt(0)
-; GFX90A-NEXT: v_pk_add_f16 v2, v3, v0
-; GFX90A-NEXT: v_pk_mov_b32 v[4:5], v[2:3], v[2:3] op_sel:[0,1]
-; GFX90A-NEXT: buffer_atomic_cmpswap v[4:5], v1, s[16:19], 0 offen glc
+; GFX90A-NEXT: v_mov_b32_e32 v5, v2
+; GFX90A-NEXT: v_pk_add_f16 v4, v5, v0
+; GFX90A-NEXT: v_pk_mov_b32 v[2:3], v[4:5], v[4:5] op_sel:[0,1]
+; GFX90A-NEXT: buffer_atomic_cmpswap v[2:3], v1, s[16:19], 0 offen glc
; GFX90A-NEXT: s_waitcnt vmcnt(0)
; GFX90A-NEXT: buffer_wbinvl1
-; GFX90A-NEXT: v_cmp_eq_u32_e32 vcc, v4, v3
+; GFX90A-NEXT: v_cmp_eq_u32_e32 vcc, v2, v5
; GFX90A-NEXT: s_or_b64 s[4:5], vcc, s[4:5]
-; GFX90A-NEXT: v_mov_b32_e32 v3, v4
; GFX90A-NEXT: s_andn2_b64 exec, exec, s[4:5]
; GFX90A-NEXT: s_cbranch_execnz .LBB25_1
; GFX90A-NEXT: ; %bb.2: ; %atomicrmw.end
@@ -8872,22 +8875,22 @@ define void @buffer_fat_ptr_agent_atomic_fadd_noret_v2f16__offset__amdgpu_no_rem
; GFX908: ; %bb.0:
; GFX908-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX908-NEXT: v_mov_b32_e32 v1, s20
-; GFX908-NEXT: buffer_load_dword v2, v1, s[16:19], 0 offen offset:1024
+; GFX908-NEXT: buffer_load_dword v1, v1, s[16:19], 0 offen offset:1024
; GFX908-NEXT: s_add_i32 s6, s20, 0x400
; GFX908-NEXT: s_mov_b64 s[4:5], 0
; GFX908-NEXT: v_mov_b32_e32 v3, s6
; GFX908-NEXT: .LBB25_1: ; %atomicrmw.start
; GFX908-NEXT: ; =>This Inner Loop Header: Depth=1
; GFX908-NEXT: s_waitcnt vmcnt(0)
-; GFX908-NEXT: v_pk_add_f16 v1, v2, v0
-; GFX908-NEXT: v_mov_b32_e32 v5, v2
-; GFX908-NEXT: v_mov_b32_e32 v4, v1
-; GFX908-NEXT: buffer_atomic_cmpswap v[4:5], v3, s[16:19], 0 offen glc
+; GFX908-NEXT: v_mov_b32_e32 v5, v1
+; GFX908-NEXT: v_pk_add_f16 v4, v5, v0
+; GFX908-NEXT: v_mov_b32_e32 v1, v4
+; GFX908-NEXT: v_mov_b32_e32 v2, v5
+; GFX908-NEXT: buffer_atomic_cmpswap v[1:2], v3, s[16:19], 0 offen glc
; GFX908-NEXT: s_waitcnt vmcnt(0)
; GFX908-NEXT: buffer_wbinvl1
-; GFX908-NEXT: v_cmp_eq_u32_e32 vcc, v4, v2
+; GFX908-NEXT: v_cmp_eq_u32_e32 vcc, v1, v5
; GFX908-NEXT: s_or_b64 s[4:5], vcc, s[4:5]
-; GFX908-NEXT: v_mov_b32_e32 v2, v4
; GFX908-NEXT: s_andn2_b64 exec, exec, s[4:5]
; GFX908-NEXT: s_cbranch_execnz .LBB25_1
; GFX908-NEXT: ; %bb.2: ; %atomicrmw.end
@@ -8898,24 +8901,24 @@ define void @buffer_fat_ptr_agent_atomic_fadd_noret_v2f16__offset__amdgpu_no_rem
; GFX8: ; %bb.0:
; GFX8-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX8-NEXT: v_mov_b32_e32 v1, s20
-; GFX8-NEXT: buffer_load_dword v2, v1, s[16:19], 0 offen offset:1024
+; GFX8-NEXT: buffer_load_dword v1, v1, s[16:19], 0 offen offset:1024
; GFX8-NEXT: s_add_i32 s6, s20, 0x400
; GFX8-NEXT: s_mov_b64 s[4:5], 0
; GFX8-NEXT: v_mov_b32_e32 v3, s6
; GFX8-NEXT: .LBB25_1: ; %atomicrmw.start
; GFX8-NEXT: ; =>This Inner Loop Header: Depth=1
; GFX8-NEXT: s_waitcnt vmcnt(0)
-; GFX8-NEXT: v_add_f16_sdwa v1, v2, v0 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:WORD_1 src1_sel:WORD_1
-; GFX8-NEXT: v_add_f16_e32 v4, v2, v0
-; GFX8-NEXT: v_or_b32_e32 v1, v4, v1
-; GFX8-NEXT: v_mov_b32_e32 v5, v2
-; GFX8-NEXT: v_mov_b32_e32 v4, v1
-; GFX8-NEXT: buffer_atomic_cmpswap v[4:5], v3, s[16:19], 0 offen glc
+; GFX8-NEXT: v_mov_b32_e32 v5, v1
+; GFX8-NEXT: v_add_f16_sdwa v1, v5, v0 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:WORD_1 src1_sel:WORD_1
+; GFX8-NEXT: v_add_f16_e32 v2, v5, v0
+; GFX8-NEXT: v_or_b32_e32 v4, v2, v1
+; GFX8-NEXT: v_mov_b32_e32 v1, v4
+; GFX8-NEXT: v_mov_b32_e32 v2, v5
+; GFX8-NEXT: buffer_atomic_cmpswap v[1:2], v3, s[16:19], 0 offen glc
; GFX8-NEXT: s_waitcnt vmcnt(0)
; GFX8-NEXT: buffer_wbinvl1
-; GFX8-NEXT: v_cmp_eq_u32_e32 vcc, v4, v2
+; GFX8-NEXT: v_cmp_eq_u32_e32 vcc, v1, v5
; GFX8-NEXT: s_or_b64 s[4:5], vcc, s[4:5]
-; GFX8-NEXT: v_mov_b32_e32 v2, v4
; GFX8-NEXT: s_andn2_b64 exec, exec, s[4:5]
; GFX8-NEXT: s_cbranch_execnz .LBB25_1
; GFX8-NEXT: ; %bb.2: ; %atomicrmw.end
@@ -9481,40 +9484,40 @@ define void @buffer_fat_ptr_agent_atomic_fadd_noret_v2bf16__offset__amdgpu_no_fi
; GFX942: ; %bb.0:
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX942-NEXT: v_mov_b32_e32 v1, s16
-; GFX942-NEXT: buffer_load_dword v1, v1, s[0:3], 0 offen offset:1024
+; GFX942-NEXT: buffer_load_dword v2, v1, s[0:3], 0 offen offset:1024
; GFX942-NEXT: s_add_i32 s4, s16, 0x400
; GFX942-NEXT: s_mov_b64 s[6:7], 0
-; GFX942-NEXT: v_lshlrev_b32_e32 v2, 16, v0
+; GFX942-NEXT: v_lshlrev_b32_e32 v1, 16, v0
; GFX942-NEXT: s_movk_i32 s8, 0x7fff
-; GFX942-NEXT: v_and_b32_e32 v3, 0xffff0000, v0
+; GFX942-NEXT: v_and_b32_e32 v0, 0xffff0000, v0
; GFX942-NEXT: s_mov_b32 s9, 0x7060302
; GFX942-NEXT: v_mov_b32_e32 v4, s4
; GFX942-NEXT: .LBB27_1: ; %atomicrmw.start
; GFX942-NEXT: ; =>This Inner Loop Header: Depth=1
; GFX942-NEXT: s_waitcnt vmcnt(0)
-; GFX942-NEXT: v_lshlrev_b32_e32 v0, 16, v1
-; GFX942-NEXT: v_and_b32_e32 v5, 0xffff0000, v1
-; GFX942-NEXT: v_add_f32_e32 v0, v0, v2
-; GFX942-NEXT: v_add_f32_e32 v5, v5, v3
-; GFX942-NEXT: v_bfe_u32 v6, v0, 16, 1
-; GFX942-NEXT: v_bfe_u32 v8, v5, 16, 1
-; GFX942-NEXT: v_or_b32_e32 v7, 0x400000, v0
-; GFX942-NEXT: v_or_b32_e32 v9, 0x400000, v5
-; GFX942-NEXT: v_add3_u32 v6, v6, v0, s8
-; GFX942-NEXT: v_add3_u32 v8, v8, v5, s8
-; GFX942-NEXT: v_cmp_u_f32_e32 vcc, v5, v5
-; GFX942-NEXT: v_cmp_u_f32_e64 s[4:5], v0, v0
+; GFX942-NEXT: v_mov_b32_e32 v7, v2
+; GFX942-NEXT: v_lshlrev_b32_e32 v2, 16, v7
+; GFX942-NEXT: v_and_b32_e32 v3, 0xffff0000, v7
+; GFX942-NEXT: v_add_f32_e32 v2, v2, v1
+; GFX942-NEXT: v_add_f32_e32 v3, v3, v0
+; GFX942-NEXT: v_bfe_u32 v5, v2, 16, 1
+; GFX942-NEXT: v_bfe_u32 v8, v3, 16, 1
+; GFX942-NEXT: v_or_b32_e32 v6, 0x400000, v2
+; GFX942-NEXT: v_or_b32_e32 v9, 0x400000, v3
+; GFX942-NEXT: v_add3_u32 v5, v5, v2, s8
+; GFX942-NEXT: v_add3_u32 v8, v8, v3, s8
+; GFX942-NEXT: v_cmp_u_f32_e32 vcc, v3, v3
+; GFX942-NEXT: v_cmp_u_f32_e64 s[4:5], v2, v2
; GFX942-NEXT: buffer_wbl2 sc1
-; GFX942-NEXT: v_cndmask_b32_e32 v5, v8, v9, vcc
-; GFX942-NEXT: v_cndmask_b32_e64 v0, v6, v7, s[4:5]
-; GFX942-NEXT: v_perm_b32 v0, v5, v0, s9
-; GFX942-NEXT: v_mov_b64_e32 v[6:7], v[0:1]
-; GFX942-NEXT: buffer_atomic_cmpswap v[6:7], v4, s[0:3], 0 offen sc0
+; GFX942-NEXT: v_cndmask_b32_e32 v3, v8, v9, vcc
+; GFX942-NEXT: v_cndmask_b32_e64 v2, v5, v6, s[4:5]
+; GFX942-NEXT: v_perm_b32 v6, v3, v2, s9
+; GFX942-NEXT: v_mov_b64_e32 v[2:3], v[6:7]
+; GFX942-NEXT: buffer_atomic_cmpswap v[2:3], v4, s[0:3], 0 offen sc0
; GFX942-NEXT: s_waitcnt vmcnt(0)
; GFX942-NEXT: buffer_inv sc1
-; GFX942-NEXT: v_cmp_eq_u32_e32 vcc, v6, v1
+; GFX942-NEXT: v_cmp_eq_u32_e32 vcc, v2, v7
; GFX942-NEXT: s_or_b64 s[6:7], vcc, s[6:7]
-; GFX942-NEXT: v_mov_b32_e32 v1, v6
; GFX942-NEXT: s_andn2_b64 exec, exec, s[6:7]
; GFX942-NEXT: s_cbranch_execnz .LBB27_1
; GFX942-NEXT: ; %bb.2: ; %atomicrmw.end
@@ -9524,10 +9527,11 @@ define void @buffer_fat_ptr_agent_atomic_fadd_noret_v2bf16__offset__amdgpu_no_fi
; GFX11-TRUE16-LABEL: buffer_fat_ptr_agent_atomic_fadd_noret_v2bf16__offset__amdgpu_no_fine_grained_memory:
; GFX11-TRUE16: ; %bb.0:
; GFX11-TRUE16-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
-; GFX11-TRUE16-NEXT: v_dual_mov_b32 v1, s16 :: v_dual_and_b32 v2, 0xffff0000, v0
+; GFX11-TRUE16-NEXT: v_mov_b32_e32 v1, s16
; GFX11-TRUE16-NEXT: s_add_i32 s4, s16, 0x400
; GFX11-TRUE16-NEXT: s_delay_alu instid0(SALU_CYCLE_1)
-; GFX11-TRUE16-NEXT: v_dual_mov_b32 v4, s4 :: v_dual_lshlrev_b32 v3, 16, v0
+; GFX11-TRUE16-NEXT: v_dual_mov_b32 v4, s4 :: v_dual_and_b32 v3, 0xffff0000, v0
+; GFX11-TRUE16-NEXT: v_lshlrev_b32_e32 v0, 16, v0
; GFX11-TRUE16-NEXT: buffer_load_b32 v1, v1, s[0:3], 0 offen offset:1024
; GFX11-TRUE16-NEXT: s_mov_b32 s4, 0
; GFX11-TRUE16-NEXT: s_set_inst_prefetch_distance 0x1
@@ -9535,34 +9539,35 @@ define void @buffer_fat_ptr_agent_atomic_fadd_noret_v2bf16__offset__amdgpu_no_fi
; GFX11-TRUE16-NEXT: .LBB27_1: ; %atomicrmw.start
; GFX11-TRUE16-NEXT: ; =>This Inner Loop Header: Depth=1
; GFX11-TRUE16-NEXT: s_waitcnt vmcnt(0)
-; GFX11-TRUE16-NEXT: v_and_b32_e32 v5, 0xffff0000, v1
-; GFX11-TRUE16-NEXT: v_lshlrev_b32_e32 v0, 16, v1
+; GFX11-TRUE16-NEXT: v_mov_b32_e32 v6, v1
; GFX11-TRUE16-NEXT: s_waitcnt_vscnt null, 0x0
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
-; GFX11-TRUE16-NEXT: v_dual_add_f32 v5, v5, v2 :: v_dual_add_f32 v0, v0, v3
-; GFX11-TRUE16-NEXT: v_bfe_u32 v7, v5, 16, 1
+; GFX11-TRUE16-NEXT: v_and_b32_e32 v2, 0xffff0000, v6
+; GFX11-TRUE16-NEXT: v_dual_add_f32 v2, v2, v3 :: v_dual_lshlrev_b32 v1, 16, v6
+; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_2)
+; GFX11-TRUE16-NEXT: v_add_f32_e32 v1, v1, v0
+; GFX11-TRUE16-NEXT: v_bfe_u32 v7, v2, 16, 1
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_2)
-; GFX11-TRUE16-NEXT: v_bfe_u32 v6, v0, 16, 1
-; GFX11-TRUE16-NEXT: v_or_b32_e32 v8, 0x400000, v0
-; GFX11-TRUE16-NEXT: v_cmp_u_f32_e32 vcc_lo, v0, v0
-; GFX11-TRUE16-NEXT: v_or_b32_e32 v9, 0x400000, v5
-; GFX11-TRUE16-NEXT: v_add3_u32 v7, v7, v5, 0x7fff
-; GFX11-TRUE16-NEXT: v_add3_u32 v6, v6, v0, 0x7fff
+; GFX11-TRUE16-NEXT: v_bfe_u32 v5, v1, 16, 1
+; GFX11-TRUE16-NEXT: v_or_b32_e32 v8, 0x400000, v1
+; GFX11-TRUE16-NEXT: v_cmp_u_f32_e32 vcc_lo, v1, v1
+; GFX11-TRUE16-NEXT: v_or_b32_e32 v9, 0x400000, v2
+; GFX11-TRUE16-NEXT: v_add3_u32 v7, v7, v2, 0x7fff
+; GFX11-TRUE16-NEXT: v_add3_u32 v5, v5, v1, 0x7fff
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(SKIP_1) | instid1(VALU_DEP_4)
-; GFX11-TRUE16-NEXT: v_cndmask_b32_e32 v0, v6, v8, vcc_lo
-; GFX11-TRUE16-NEXT: v_cmp_u_f32_e32 vcc_lo, v5, v5
-; GFX11-TRUE16-NEXT: v_cndmask_b32_e32 v5, v7, v9, vcc_lo
+; GFX11-TRUE16-NEXT: v_cndmask_b32_e32 v1, v5, v8, vcc_lo
+; GFX11-TRUE16-NEXT: v_cmp_u_f32_e32 vcc_lo, v2, v2
+; GFX11-TRUE16-NEXT: v_cndmask_b32_e32 v2, v7, v9, vcc_lo
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_3) | instskip(NEXT) | instid1(VALU_DEP_1)
-; GFX11-TRUE16-NEXT: v_mov_b16_e32 v0.l, v0.h
-; GFX11-TRUE16-NEXT: v_bfi_b32 v0, 0xffff, v0, v5
+; GFX11-TRUE16-NEXT: v_mov_b16_e32 v1.l, v1.h
+; GFX11-TRUE16-NEXT: v_bfi_b32 v5, 0xffff, v1, v2
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1)
-; GFX11-TRUE16-NEXT: v_dual_mov_b32 v6, v1 :: v_dual_mov_b32 v5, v0
-; GFX11-TRUE16-NEXT: buffer_atomic_cmpswap_b32 v[5:6], v4, s[0:3], 0 offen glc
+; GFX11-TRUE16-NEXT: v_dual_mov_b32 v1, v5 :: v_dual_mov_b32 v2, v6
+; GFX11-TRUE16-NEXT: buffer_atomic_cmpswap_b32 v[1:2], v4, s[0:3], 0 offen glc
; GFX11-TRUE16-NEXT: s_waitcnt vmcnt(0)
; GFX11-TRUE16-NEXT: buffer_gl1_inv
; GFX11-TRUE16-NEXT: buffer_gl0_inv
-; GFX11-TRUE16-NEXT: v_cmp_eq_u32_e32 vcc_lo, v5, v1
-; GFX11-TRUE16-NEXT: v_mov_b32_e32 v1, v5
+; GFX11-TRUE16-NEXT: v_cmp_eq_u32_e32 vcc_lo, v1, v6
; GFX11-TRUE16-NEXT: s_or_b32 s4, vcc_lo, s4
; GFX11-TRUE16-NEXT: s_delay_alu instid0(SALU_CYCLE_1)
; GFX11-TRUE16-NEXT: s_and_not1_b32 exec_lo, exec_lo, s4
@@ -9575,10 +9580,11 @@ define void @buffer_fat_ptr_agent_atomic_fadd_noret_v2bf16__offset__amdgpu_no_fi
; GFX11-FAKE16-LABEL: buffer_fat_ptr_agent_atomic_fadd_noret_v2bf16__offset__amdgpu_no_fine_grained_memory:
; GFX11-FAKE16: ; %bb.0:
; GFX11-FAKE16-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
-; GFX11-FAKE16-NEXT: v_dual_mov_b32 v1, s16 :: v_dual_lshlrev_b32 v2, 16, v0
+; GFX11-FAKE16-NEXT: v_mov_b32_e32 v1, s16
; GFX11-FAKE16-NEXT: s_add_i32 s4, s16, 0x400
; GFX11-FAKE16-NEXT: s_delay_alu instid0(SALU_CYCLE_1)
-; GFX11-FAKE16-NEXT: v_dual_mov_b32 v4, s4 :: v_dual_and_b32 v3, 0xffff0000, v0
+; GFX11-FAKE16-NEXT: v_dual_mov_b32 v4, s4 :: v_dual_lshlrev_b32 v3, 16, v0
+; GFX11-FAKE16-NEXT: v_and_b32_e32 v0, 0xffff0000, v0
; GFX11-FAKE16-NEXT: buffer_load_b32 v1, v1, s[0:3], 0 offen offset:1024
; GFX11-FAKE16-NEXT: s_mov_b32 s5, 0
; GFX11-FAKE16-NEXT: s_set_inst_prefetch_distance 0x1
@@ -9586,32 +9592,33 @@ define void @buffer_fat_ptr_agent_atomic_fadd_noret_v2bf16__offset__amdgpu_no_fi
; GFX11-FAKE16-NEXT: .LBB27_1: ; %atomicrmw.start
; GFX11-FAKE16-NEXT: ; =>This Inner Loop Header: Depth=1
; GFX11-FAKE16-NEXT: s_waitcnt vmcnt(0)
-; GFX11-FAKE16-NEXT: v_and_b32_e32 v5, 0xffff0000, v1
-; GFX11-FAKE16-NEXT: v_lshlrev_b32_e32 v0, 16, v1
+; GFX11-FAKE16-NEXT: v_mov_b32_e32 v6, v1
; GFX11-FAKE16-NEXT: s_waitcnt_vscnt null, 0x0
; GFX11-FAKE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
-; GFX11-FAKE16-NEXT: v_dual_add_f32 v5, v5, v3 :: v_dual_add_f32 v0, v0, v2
-; GFX11-FAKE16-NEXT: v_bfe_u32 v7, v5, 16, 1
+; GFX11-FAKE16-NEXT: v_and_b32_e32 v2, 0xffff0000, v6
+; GFX11-FAKE16-NEXT: v_dual_add_f32 v2, v2, v0 :: v_dual_lshlrev_b32 v1, 16, v6
+; GFX11-FAKE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_2)
+; GFX11-FAKE16-NEXT: v_add_f32_e32 v1, v1, v3
+; GFX11-FAKE16-NEXT: v_bfe_u32 v7, v2, 16, 1
; GFX11-FAKE16-NEXT: s_delay_alu instid0(VALU_DEP_2)
-; GFX11-FAKE16-NEXT: v_bfe_u32 v6, v0, 16, 1
-; GFX11-FAKE16-NEXT: v_or_b32_e32 v8, 0x400000, v0
-; GFX11-FAKE16-NEXT: v_or_b32_e32 v9, 0x400000, v5
-; GFX11-FAKE16-NEXT: v_cmp_u_f32_e32 vcc_lo, v5, v5
-; GFX11-FAKE16-NEXT: v_add3_u32 v7, v7, v5, 0x7fff
-; GFX11-FAKE16-NEXT: v_add3_u32 v6, v6, v0, 0x7fff
-; GFX11-FAKE16-NEXT: v_cmp_u_f32_e64 s4, v0, v0
+; GFX11-FAKE16-NEXT: v_bfe_u32 v5, v1, 16, 1
+; GFX11-FAKE16-NEXT: v_or_b32_e32 v8, 0x400000, v1
+; GFX11-FAKE16-NEXT: v_or_b32_e32 v9, 0x400000, v2
+; GFX11-FAKE16-NEXT: v_cmp_u_f32_e32 vcc_lo, v2, v2
+; GFX11-FAKE16-NEXT: v_add3_u32 v7, v7, v2, 0x7fff
+; GFX11-FAKE16-NEXT: v_add3_u32 v5, v5, v1, 0x7fff
+; GFX11-FAKE16-NEXT: v_cmp_u_f32_e64 s4, v1, v1
; GFX11-FAKE16-NEXT: s_delay_alu instid0(VALU_DEP_3) | instskip(NEXT) | instid1(VALU_DEP_2)
-; GFX11-FAKE16-NEXT: v_cndmask_b32_e32 v5, v7, v9, vcc_lo
-; GFX11-FAKE16-NEXT: v_cndmask_b32_e64 v0, v6, v8, s4
+; GFX11-FAKE16-NEXT: v_cndmask_b32_e32 v2, v7, v9, vcc_lo
+; GFX11-FAKE16-NEXT: v_cndmask_b32_e64 v1, v5, v8, s4
; GFX11-FAKE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
-; GFX11-FAKE16-NEXT: v_perm_b32 v0, v5, v0, 0x7060302
-; GFX11-FAKE16-NEXT: v_dual_mov_b32 v6, v1 :: v_dual_mov_b32 v5, v0
-; GFX11-FAKE16-NEXT: buffer_atomic_cmpswap_b32 v[5:6], v4, s[0:3], 0 offen glc
+; GFX11-FAKE16-NEXT: v_perm_b32 v5, v2, v1, 0x7060302
+; GFX11-FAKE16-NEXT: v_dual_mov_b32 v1, v5 :: v_dual_mov_b32 v2, v6
+; GFX11-FAKE16-NEXT: buffer_atomic_cmpswap_b32 v[1:2], v4, s[0:3], 0 offen glc
; GFX11-FAKE16-NEXT: s_waitcnt vmcnt(0)
; GFX11-FAKE16-NEXT: buffer_gl1_inv
; GFX11-FAKE16-NEXT: buffer_gl0_inv
-; GFX11-FAKE16-NEXT: v_cmp_eq_u32_e32 vcc_lo, v5, v1
-; GFX11-FAKE16-NEXT: v_mov_b32_e32 v1, v5
+; GFX11-FAKE16-NEXT: v_cmp_eq_u32_e32 vcc_lo, v1, v6
; GFX11-FAKE16-NEXT: s_or_b32 s5, vcc_lo, s5
; GFX11-FAKE16-NEXT: s_delay_alu instid0(SALU_CYCLE_1)
; GFX11-FAKE16-NEXT: s_and_not1_b32 exec_lo, exec_lo, s5
@@ -9626,38 +9633,38 @@ define void @buffer_fat_ptr_agent_atomic_fadd_noret_v2bf16__offset__amdgpu_no_fi
; GFX10-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX10-NEXT: v_mov_b32_e32 v1, s20
; GFX10-NEXT: s_add_i32 s4, s20, 0x400
-; GFX10-NEXT: v_lshlrev_b32_e32 v2, 16, v0
-; GFX10-NEXT: v_and_b32_e32 v3, 0xffff0000, v0
+; GFX10-NEXT: v_lshlrev_b32_e32 v3, 16, v0
+; GFX10-NEXT: v_and_b32_e32 v0, 0xffff0000, v0
; GFX10-NEXT: v_mov_b32_e32 v4, s4
; GFX10-NEXT: buffer_load_dword v1, v1, s[16:19], 0 offen offset:1024
; GFX10-NEXT: s_mov_b32 s5, 0
; GFX10-NEXT: .LBB27_1: ; %atomicrmw.start
; GFX10-NEXT: ; =>This Inner Loop Header: Depth=1
; GFX10-NEXT: s_waitcnt vmcnt(0)
-; GFX10-NEXT: v_lshlrev_b32_e32 v0, 16, v1
-; GFX10-NEXT: v_and_b32_e32 v5, 0xffff0000, v1
-; GFX10-NEXT: s_waitcnt_vscnt null, 0x0
-; GFX10-NEXT: v_add_f32_e32 v0, v0, v2
-; GFX10-NEXT: v_add_f32_e32 v5, v5, v3
-; GFX10-NEXT: v_bfe_u32 v6, v0, 16, 1
-; GFX10-NEXT: v_bfe_u32 v7, v5, 16, 1
-; GFX10-NEXT: v_or_b32_e32 v8, 0x400000, v0
-; GFX10-NEXT: v_or_b32_e32 v9, 0x400000, v5
-; GFX10-NEXT: v_cmp_u_f32_e32 vcc_lo, v5, v5
-; GFX10-NEXT: v_add3_u32 v6, v6, v0, 0x7fff
-; GFX10-NEXT: v_add3_u32 v7, v7, v5, 0x7fff
-; GFX10-NEXT: v_cmp_u_f32_e64 s4, v0, v0
-; GFX10-NEXT: v_cndmask_b32_e32 v5, v7, v9, vcc_lo
-; GFX10-NEXT: v_cndmask_b32_e64 v0, v6, v8, s4
-; GFX10-NEXT: v_perm_b32 v0, v5, v0, 0x7060302
; GFX10-NEXT: v_mov_b32_e32 v6, v1
-; GFX10-NEXT: v_mov_b32_e32 v5, v0
-; GFX10-NEXT: buffer_atomic_cmpswap v[5:6], v4, s[16:19], 0 offen glc
+; GFX10-NEXT: s_waitcnt_vscnt null, 0x0
+; GFX10-NEXT: v_lshlrev_b32_e32 v1, 16, v6
+; GFX10-NEXT: v_and_b32_e32 v2, 0xffff0000, v6
+; GFX10-NEXT: v_add_f32_e32 v1, v1, v3
+; GFX10-NEXT: v_add_f32_e32 v2, v2, v0
+; GFX10-NEXT: v_bfe_u32 v5, v1, 16, 1
+; GFX10-NEXT: v_bfe_u32 v7, v2, 16, 1
+; GFX10-NEXT: v_or_b32_e32 v8, 0x400000, v1
+; GFX10-NEXT: v_or_b32_e32 v9, 0x400000, v2
+; GFX10-NEXT: v_cmp_u_f32_e32 vcc_lo, v2, v2
+; GFX10-NEXT: v_add3_u32 v5, v5, v1, 0x7fff
+; GFX10-NEXT: v_add3_u32 v7, v7, v2, 0x7fff
+; GFX10-NEXT: v_cmp_u_f32_e64 s4, v1, v1
+; GFX10-NEXT: v_cndmask_b32_e32 v2, v7, v9, vcc_lo
+; GFX10-NEXT: v_cndmask_b32_e64 v1, v5, v8, s4
+; GFX10-NEXT: v_perm_b32 v5, v2, v1, 0x7060302
+; GFX10-NEXT: v_mov_b32_e32 v1, v5
+; GFX10-NEXT: v_mov_b32_e32 v2, v6
+; GFX10-NEXT: buffer_atomic_cmpswap v[1:2], v4, s[16:19], 0 offen glc
; GFX10-NEXT: s_waitcnt vmcnt(0)
; GFX10-NEXT: buffer_gl1_inv
; GFX10-NEXT: buffer_gl0_inv
-; GFX10-NEXT: v_cmp_eq_u32_e32 vcc_lo, v5, v1
-; GFX10-NEXT: v_mov_b32_e32 v1, v5
+; GFX10-NEXT: v_cmp_eq_u32_e32 vcc_lo, v1, v6
; GFX10-NEXT: s_or_b32 s5, vcc_lo, s5
; GFX10-NEXT: s_andn2_b32 exec_lo, exec_lo, s5
; GFX10-NEXT: s_cbranch_execnz .LBB27_1
@@ -9669,39 +9676,39 @@ define void @buffer_fat_ptr_agent_atomic_fadd_noret_v2bf16__offset__amdgpu_no_fi
; GFX90A: ; %bb.0:
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX90A-NEXT: v_mov_b32_e32 v1, s20
-; GFX90A-NEXT: buffer_load_dword v1, v1, s[16:19], 0 offen offset:1024
+; GFX90A-NEXT: buffer_load_dword v2, v1, s[16:19], 0 offen offset:1024
; GFX90A-NEXT: s_add_i32 s4, s20, 0x400
; GFX90A-NEXT: s_mov_b64 s[6:7], 0
-; GFX90A-NEXT: v_lshlrev_b32_e32 v2, 16, v0
+; GFX90A-NEXT: v_lshlrev_b32_e32 v1, 16, v0
; GFX90A-NEXT: s_movk_i32 s8, 0x7fff
-; GFX90A-NEXT: v_and_b32_e32 v3, 0xffff0000, v0
+; GFX90A-NEXT: v_and_b32_e32 v0, 0xffff0000, v0
; GFX90A-NEXT: s_mov_b32 s9, 0x7060302
; GFX90A-NEXT: v_mov_b32_e32 v4, s4
; GFX90A-NEXT: .LBB27_1: ; %atomicrmw.start
; GFX90A-NEXT: ; =>This Inner Loop Header: Depth=1
; GFX90A-NEXT: s_waitcnt vmcnt(0)
-; GFX90A-NEXT: v_lshlrev_b32_e32 v0, 16, v1
-; GFX90A-NEXT: v_and_b32_e32 v5, 0xffff0000, v1
-; GFX90A-NEXT: v_add_f32_e32 v0, v0, v2
-; GFX90A-NEXT: v_add_f32_e32 v5, v5, v3
-; GFX90A-NEXT: v_bfe_u32 v6, v0, 16, 1
-; GFX90A-NEXT: v_bfe_u32 v8, v5, 16, 1
-; GFX90A-NEXT: v_or_b32_e32 v7, 0x400000, v0
-; GFX90A-NEXT: v_or_b32_e32 v9, 0x400000, v5
-; GFX90A-NEXT: v_add3_u32 v6, v6, v0, s8
-; GFX90A-NEXT: v_add3_u32 v8, v8, v5, s8
-; GFX90A-NEXT: v_cmp_u_f32_e32 vcc, v5, v5
-; GFX90A-NEXT: v_cmp_u_f32_e64 s[4:5], v0, v0
-; GFX90A-NEXT: v_cndmask_b32_e64 v0, v6, v7, s[4:5]
-; GFX90A-NEXT: v_cndmask_b32_e32 v5, v8, v9, vcc
-; GFX90A-NEXT: v_perm_b32 v0, v5, v0, s9
-; GFX90A-NEXT: v_pk_mov_b32 v[6:7], v[0:1], v[0:1] op_sel:[0,1]
-; GFX90A-NEXT: buffer_atomic_cmpswap v[6:7], v4, s[16:19], 0 offen glc
+; GFX90A-NEXT: v_mov_b32_e32 v7, v2
+; GFX90A-NEXT: v_lshlrev_b32_e32 v2, 16, v7
+; GFX90A-NEXT: v_and_b32_e32 v3, 0xffff0000, v7
+; GFX90A-NEXT: v_add_f32_e32 v2, v2, v1
+; GFX90A-NEXT: v_add_f32_e32 v3, v3, v0
+; GFX90A-NEXT: v_bfe_u32 v5, v2, 16, 1
+; GFX90A-NEXT: v_bfe_u32 v8, v3, 16, 1
+; GFX90A-NEXT: v_or_b32_e32 v6, 0x400000, v2
+; GFX90A-NEXT: v_or_b32_e32 v9, 0x400000, v3
+; GFX90A-NEXT: v_add3_u32 v5, v5, v2, s8
+; GFX90A-NEXT: v_add3_u32 v8, v8, v3, s8
+; GFX90A-NEXT: v_cmp_u_f32_e32 vcc, v3, v3
+; GFX90A-NEXT: v_cmp_u_f32_e64 s[4:5], v2, v2
+; GFX90A-NEXT: v_cndmask_b32_e64 v2, v5, v6, s[4:5]
+; GFX90A-NEXT: v_cndmask_b32_e32 v3, v8, v9, vcc
+; GFX90A-NEXT: v_perm_b32 v6, v3, v2, s9
+; GFX90A-NEXT: v_pk_mov_b32 v[2:3], v[6:7], v[6:7] op_sel:[0,1]
+; GFX90A-NEXT: buffer_atomic_cmpswap v[2:3], v4, s[16:19], 0 offen glc
; GFX90A-NEXT: s_waitcnt vmcnt(0)
; GFX90A-NEXT: buffer_wbinvl1
-; GFX90A-NEXT: v_cmp_eq_u32_e32 vcc, v6, v1
+; GFX90A-NEXT: v_cmp_eq_u32_e32 vcc, v2, v7
; GFX90A-NEXT: s_or_b64 s[6:7], vcc, s[6:7]
-; GFX90A-NEXT: v_mov_b32_e32 v1, v6
; GFX90A-NEXT: s_andn2_b64 exec, exec, s[6:7]
; GFX90A-NEXT: s_cbranch_execnz .LBB27_1
; GFX90A-NEXT: ; %bb.2: ; %atomicrmw.end
@@ -9715,37 +9722,37 @@ define void @buffer_fat_ptr_agent_atomic_fadd_noret_v2bf16__offset__amdgpu_no_fi
; GFX908-NEXT: buffer_load_dword v1, v1, s[16:19], 0 offen offset:1024
; GFX908-NEXT: s_add_i32 s4, s20, 0x400
; GFX908-NEXT: s_mov_b64 s[6:7], 0
-; GFX908-NEXT: v_lshlrev_b32_e32 v2, 16, v0
+; GFX908-NEXT: v_lshlrev_b32_e32 v3, 16, v0
; GFX908-NEXT: s_movk_i32 s8, 0x7fff
-; GFX908-NEXT: v_and_b32_e32 v3, 0xffff0000, v0
+; GFX908-NEXT: v_and_b32_e32 v0, 0xffff0000, v0
; GFX908-NEXT: s_mov_b32 s9, 0x7060302
; GFX908-NEXT: v_mov_b32_e32 v4, s4
; GFX908-NEXT: .LBB27_1: ; %atomicrmw.start
; GFX908-NEXT: ; =>This Inner Loop Header: Depth=1
; GFX908-NEXT: s_waitcnt vmcnt(0)
-; GFX908-NEXT: v_lshlrev_b32_e32 v0, 16, v1
-; GFX908-NEXT: v_and_b32_e32 v5, 0xffff0000, v1
-; GFX908-NEXT: v_add_f32_e32 v0, v0, v2
-; GFX908-NEXT: v_add_f32_e32 v5, v5, v3
-; GFX908-NEXT: v_bfe_u32 v6, v0, 16, 1
-; GFX908-NEXT: v_bfe_u32 v8, v5, 16, 1
-; GFX908-NEXT: v_or_b32_e32 v7, 0x400000, v0
-; GFX908-NEXT: v_or_b32_e32 v9, 0x400000, v5
-; GFX908-NEXT: v_add3_u32 v6, v6, v0, s8
-; GFX908-NEXT: v_add3_u32 v8, v8, v5, s8
-; GFX908-NEXT: v_cmp_u_f32_e32 vcc, v5, v5
-; GFX908-NEXT: v_cmp_u_f32_e64 s[4:5], v0, v0
-; GFX908-NEXT: v_cndmask_b32_e64 v0, v6, v7, s[4:5]
-; GFX908-NEXT: v_cndmask_b32_e32 v5, v8, v9, vcc
-; GFX908-NEXT: v_perm_b32 v0, v5, v0, s9
; GFX908-NEXT: v_mov_b32_e32 v6, v1
-; GFX908-NEXT: v_mov_b32_e32 v5, v0
-; GFX908-NEXT: buffer_atomic_cmpswap v[5:6], v4, s[16:19], 0 offen glc
+; GFX908-NEXT: v_lshlrev_b32_e32 v1, 16, v6
+; GFX908-NEXT: v_and_b32_e32 v2, 0xffff0000, v6
+; GFX908-NEXT: v_add_f32_e32 v1, v1, v3
+; GFX908-NEXT: v_add_f32_e32 v2, v2, v0
+; GFX908-NEXT: v_bfe_u32 v5, v1, 16, 1
+; GFX908-NEXT: v_bfe_u32 v8, v2, 16, 1
+; GFX908-NEXT: v_or_b32_e32 v7, 0x400000, v1
+; GFX908-NEXT: v_or_b32_e32 v9, 0x400000, v2
+; GFX908-NEXT: v_add3_u32 v5, v5, v1, s8
+; GFX908-NEXT: v_add3_u32 v8, v8, v2, s8
+; GFX908-NEXT: v_cmp_u_f32_e32 vcc, v2, v2
+; GFX908-NEXT: v_cmp_u_f32_e64 s[4:5], v1, v1
+; GFX908-NEXT: v_cndmask_b32_e64 v1, v5, v7, s[4:5]
+; GFX908-NEXT: v_cndmask_b32_e32 v2, v8, v9, vcc
+; GFX908-NEXT: v_perm_b32 v5, v2, v1, s9
+; GFX908-NEXT: v_mov_b32_e32 v1, v5
+; GFX908-NEXT: v_mov_b32_e32 v2, v6
+; GFX908-NEXT: buffer_atomic_cmpswap v[1:2], v4, s[16:19], 0 offen glc
; GFX908-NEXT: s_waitcnt vmcnt(0)
; GFX908-NEXT: buffer_wbinvl1
-; GFX908-NEXT: v_cmp_eq_u32_e32 vcc, v5, v1
+; GFX908-NEXT: v_cmp_eq_u32_e32 vcc, v1, v6
; GFX908-NEXT: s_or_b64 s[6:7], vcc, s[6:7]
-; GFX908-NEXT: v_mov_b32_e32 v1, v5
; GFX908-NEXT: s_andn2_b64 exec, exec, s[6:7]
; GFX908-NEXT: s_cbranch_execnz .LBB27_1
; GFX908-NEXT: ; %bb.2: ; %atomicrmw.end
@@ -9759,38 +9766,38 @@ define void @buffer_fat_ptr_agent_atomic_fadd_noret_v2bf16__offset__amdgpu_no_fi
; GFX8-NEXT: buffer_load_dword v1, v1, s[16:19], 0 offen offset:1024
; GFX8-NEXT: s_add_i32 s4, s20, 0x400
; GFX8-NEXT: s_mov_b64 s[6:7], 0
-; GFX8-NEXT: v_lshlrev_b32_e32 v2, 16, v0
-; GFX8-NEXT: v_and_b32_e32 v3, 0xffff0000, v0
+; GFX8-NEXT: v_lshlrev_b32_e32 v3, 16, v0
+; GFX8-NEXT: v_and_b32_e32 v0, 0xffff0000, v0
; GFX8-NEXT: v_mov_b32_e32 v4, s4
; GFX8-NEXT: .LBB27_1: ; %atomicrmw.start
; GFX8-NEXT: ; =>This Inner Loop Header: Depth=1
; GFX8-NEXT: s_waitcnt vmcnt(0)
-; GFX8-NEXT: v_lshlrev_b32_e32 v0, 16, v1
-; GFX8-NEXT: v_and_b32_e32 v5, 0xffff0000, v1
-; GFX8-NEXT: v_add_f32_e32 v0, v0, v2
-; GFX8-NEXT: v_add_f32_e32 v5, v5, v3
-; GFX8-NEXT: v_bfe_u32 v6, v0, 16, 1
-; GFX8-NEXT: v_bfe_u32 v8, v5, 16, 1
-; GFX8-NEXT: v_add_u32_e32 v6, vcc, v6, v0
-; GFX8-NEXT: v_add_u32_e32 v8, vcc, v8, v5
-; GFX8-NEXT: v_add_u32_e32 v6, vcc, 0x7fff, v6
-; GFX8-NEXT: v_add_u32_e32 v8, vcc, 0x7fff, v8
-; GFX8-NEXT: v_or_b32_e32 v9, 0x400000, v5
-; GFX8-NEXT: v_cmp_u_f32_e32 vcc, v5, v5
-; GFX8-NEXT: v_or_b32_e32 v7, 0x400000, v0
-; GFX8-NEXT: v_cmp_u_f32_e64 s[4:5], v0, v0
-; GFX8-NEXT: v_cndmask_b32_e32 v5, v8, v9, vcc
-; GFX8-NEXT: v_cndmask_b32_e64 v0, v6, v7, s[4:5]
-; GFX8-NEXT: v_lshrrev_b32_e32 v5, 16, v5
-; GFX8-NEXT: v_alignbit_b32 v0, v5, v0, 16
; GFX8-NEXT: v_mov_b32_e32 v6, v1
-; GFX8-NEXT: v_mov_b32_e32 v5, v0
-; GFX8-NEXT: buffer_atomic_cmpswap v[5:6], v4, s[16:19], 0 offen glc
+; GFX8-NEXT: v_lshlrev_b32_e32 v1, 16, v6
+; GFX8-NEXT: v_and_b32_e32 v2, 0xffff0000, v6
+; GFX8-NEXT: v_add_f32_e32 v1, v1, v3
+; GFX8-NEXT: v_add_f32_e32 v2, v2, v0
+; GFX8-NEXT: v_bfe_u32 v5, v1, 16, 1
+; GFX8-NEXT: v_bfe_u32 v8, v2, 16, 1
+; GFX8-NEXT: v_add_u32_e32 v5, vcc, v5, v1
+; GFX8-NEXT: v_add_u32_e32 v8, vcc, v8, v2
+; GFX8-NEXT: v_add_u32_e32 v5, vcc, 0x7fff, v5
+; GFX8-NEXT: v_add_u32_e32 v8, vcc, 0x7fff, v8
+; GFX8-NEXT: v_or_b32_e32 v9, 0x400000, v2
+; GFX8-NEXT: v_cmp_u_f32_e32 vcc, v2, v2
+; GFX8-NEXT: v_or_b32_e32 v7, 0x400000, v1
+; GFX8-NEXT: v_cmp_u_f32_e64 s[4:5], v1, v1
+; GFX8-NEXT: v_cndmask_b32_e32 v2, v8, v9, vcc
+; GFX8-NEXT: v_cndmask_b32_e64 v1, v5, v7, s[4:5]
+; GFX8-NEXT: v_lshrrev_b32_e32 v2, 16, v2
+; GFX8-NEXT: v_alignbit_b32 v5, v2, v1, 16
+; GFX8-NEXT: v_mov_b32_e32 v1, v5
+; GFX8-NEXT: v_mov_b32_e32 v2, v6
+; GFX8-NEXT: buffer_atomic_cmpswap v[1:2], v4, s[16:19], 0 offen glc
; GFX8-NEXT: s_waitcnt vmcnt(0)
; GFX8-NEXT: buffer_wbinvl1
-; GFX8-NEXT: v_cmp_eq_u32_e32 vcc, v5, v1
+; GFX8-NEXT: v_cmp_eq_u32_e32 vcc, v1, v6
; GFX8-NEXT: s_or_b64 s[6:7], vcc, s[6:7]
-; GFX8-NEXT: v_mov_b32_e32 v1, v5
; GFX8-NEXT: s_andn2_b64 exec, exec, s[6:7]
; GFX8-NEXT: s_cbranch_execnz .LBB27_1
; GFX8-NEXT: ; %bb.2: ; %atomicrmw.end
@@ -11087,40 +11094,40 @@ define void @buffer_fat_ptr_agent_atomic_fadd_noret_v2bf16__offset(ptr addrspace
; GFX942: ; %bb.0:
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX942-NEXT: v_mov_b32_e32 v1, s16
-; GFX942-NEXT: buffer_load_dword v1, v1, s[0:3], 0 offen offset:1024
+; GFX942-NEXT: buffer_load_dword v2, v1, s[0:3], 0 offen offset:1024
; GFX942-NEXT: s_add_i32 s4, s16, 0x400
; GFX942-NEXT: s_mov_b64 s[6:7], 0
-; GFX942-NEXT: v_lshlrev_b32_e32 v2, 16, v0
+; GFX942-NEXT: v_lshlrev_b32_e32 v1, 16, v0
; GFX942-NEXT: s_movk_i32 s8, 0x7fff
-; GFX942-NEXT: v_and_b32_e32 v3, 0xffff0000, v0
+; GFX942-NEXT: v_and_b32_e32 v0, 0xffff0000, v0
; GFX942-NEXT: s_mov_b32 s9, 0x7060302
; GFX942-NEXT: v_mov_b32_e32 v4, s4
; GFX942-NEXT: .LBB30_1: ; %atomicrmw.start
; GFX942-NEXT: ; =>This Inner Loop Header: Depth=1
; GFX942-NEXT: s_waitcnt vmcnt(0)
-; GFX942-NEXT: v_lshlrev_b32_e32 v0, 16, v1
-; GFX942-NEXT: v_and_b32_e32 v5, 0xffff0000, v1
-; GFX942-NEXT: v_add_f32_e32 v0, v0, v2
-; GFX942-NEXT: v_add_f32_e32 v5, v5, v3
-; GFX942-NEXT: v_bfe_u32 v6, v0, 16, 1
-; GFX942-NEXT: v_bfe_u32 v8, v5, 16, 1
-; GFX942-NEXT: v_or_b32_e32 v7, 0x400000, v0
-; GFX942-NEXT: v_or_b32_e32 v9, 0x400000, v5
-; GFX942-NEXT: v_add3_u32 v6, v6, v0, s8
-; GFX942-NEXT: v_add3_u32 v8, v8, v5, s8
-; GFX942-NEXT: v_cmp_u_f32_e32 vcc, v5, v5
-; GFX942-NEXT: v_cmp_u_f32_e64 s[4:5], v0, v0
+; GFX942-NEXT: v_mov_b32_e32 v7, v2
+; GFX942-NEXT: v_lshlrev_b32_e32 v2, 16, v7
+; GFX942-NEXT: v_and_b32_e32 v3, 0xffff0000, v7
+; GFX942-NEXT: v_add_f32_e32 v2, v2, v1
+; GFX942-NEXT: v_add_f32_e32 v3, v3, v0
+; GFX942-NEXT: v_bfe_u32 v5, v2, 16, 1
+; GFX942-NEXT: v_bfe_u32 v8, v3, 16, 1
+; GFX942-NEXT: v_or_b32_e32 v6, 0x400000, v2
+; GFX942-NEXT: v_or_b32_e32 v9, 0x400000, v3
+; GFX942-NEXT: v_add3_u32 v5, v5, v2, s8
+; GFX942-NEXT: v_add3_u32 v8, v8, v3, s8
+; GFX942-NEXT: v_cmp_u_f32_e32 vcc, v3, v3
+; GFX942-NEXT: v_cmp_u_f32_e64 s[4:5], v2, v2
; GFX942-NEXT: buffer_wbl2 sc1
-; GFX942-NEXT: v_cndmask_b32_e32 v5, v8, v9, vcc
-; GFX942-NEXT: v_cndmask_b32_e64 v0, v6, v7, s[4:5]
-; GFX942-NEXT: v_perm_b32 v0, v5, v0, s9
-; GFX942-NEXT: v_mov_b64_e32 v[6:7], v[0:1]
-; GFX942-NEXT: buffer_atomic_cmpswap v[6:7], v4, s[0:3], 0 offen sc0
+; GFX942-NEXT: v_cndmask_b32_e32 v3, v8, v9, vcc
+; GFX942-NEXT: v_cndmask_b32_e64 v2, v5, v6, s[4:5]
+; GFX942-NEXT: v_perm_b32 v6, v3, v2, s9
+; GFX942-NEXT: v_mov_b64_e32 v[2:3], v[6:7]
+; GFX942-NEXT: buffer_atomic_cmpswap v[2:3], v4, s[0:3], 0 offen sc0
; GFX942-NEXT: s_waitcnt vmcnt(0)
; GFX942-NEXT: buffer_inv sc1
-; GFX942-NEXT: v_cmp_eq_u32_e32 vcc, v6, v1
+; GFX942-NEXT: v_cmp_eq_u32_e32 vcc, v2, v7
; GFX942-NEXT: s_or_b64 s[6:7], vcc, s[6:7]
-; GFX942-NEXT: v_mov_b32_e32 v1, v6
; GFX942-NEXT: s_andn2_b64 exec, exec, s[6:7]
; GFX942-NEXT: s_cbranch_execnz .LBB30_1
; GFX942-NEXT: ; %bb.2: ; %atomicrmw.end
@@ -11130,10 +11137,11 @@ define void @buffer_fat_ptr_agent_atomic_fadd_noret_v2bf16__offset(ptr addrspace
; GFX11-TRUE16-LABEL: buffer_fat_ptr_agent_atomic_fadd_noret_v2bf16__offset:
; GFX11-TRUE16: ; %bb.0:
; GFX11-TRUE16-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
-; GFX11-TRUE16-NEXT: v_dual_mov_b32 v1, s16 :: v_dual_and_b32 v2, 0xffff0000, v0
+; GFX11-TRUE16-NEXT: v_mov_b32_e32 v1, s16
; GFX11-TRUE16-NEXT: s_add_i32 s4, s16, 0x400
; GFX11-TRUE16-NEXT: s_delay_alu instid0(SALU_CYCLE_1)
-; GFX11-TRUE16-NEXT: v_dual_mov_b32 v4, s4 :: v_dual_lshlrev_b32 v3, 16, v0
+; GFX11-TRUE16-NEXT: v_dual_mov_b32 v4, s4 :: v_dual_and_b32 v3, 0xffff0000, v0
+; GFX11-TRUE16-NEXT: v_lshlrev_b32_e32 v0, 16, v0
; GFX11-TRUE16-NEXT: buffer_load_b32 v1, v1, s[0:3], 0 offen offset:1024
; GFX11-TRUE16-NEXT: s_mov_b32 s4, 0
; GFX11-TRUE16-NEXT: s_set_inst_prefetch_distance 0x1
@@ -11141,34 +11149,35 @@ define void @buffer_fat_ptr_agent_atomic_fadd_noret_v2bf16__offset(ptr addrspace
; GFX11-TRUE16-NEXT: .LBB30_1: ; %atomicrmw.start
; GFX11-TRUE16-NEXT: ; =>This Inner Loop Header: Depth=1
; GFX11-TRUE16-NEXT: s_waitcnt vmcnt(0)
-; GFX11-TRUE16-NEXT: v_and_b32_e32 v5, 0xffff0000, v1
-; GFX11-TRUE16-NEXT: v_lshlrev_b32_e32 v0, 16, v1
+; GFX11-TRUE16-NEXT: v_mov_b32_e32 v6, v1
; GFX11-TRUE16-NEXT: s_waitcnt_vscnt null, 0x0
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
-; GFX11-TRUE16-NEXT: v_dual_add_f32 v5, v5, v2 :: v_dual_add_f32 v0, v0, v3
-; GFX11-TRUE16-NEXT: v_bfe_u32 v7, v5, 16, 1
+; GFX11-TRUE16-NEXT: v_and_b32_e32 v2, 0xffff0000, v6
+; GFX11-TRUE16-NEXT: v_dual_add_f32 v2, v2, v3 :: v_dual_lshlrev_b32 v1, 16, v6
+; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_2)
+; GFX11-TRUE16-NEXT: v_add_f32_e32 v1, v1, v0
+; GFX11-TRUE16-NEXT: v_bfe_u32 v7, v2, 16, 1
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_2)
-; GFX11-TRUE16-NEXT: v_bfe_u32 v6, v0, 16, 1
-; GFX11-TRUE16-NEXT: v_or_b32_e32 v8, 0x400000, v0
-; GFX11-TRUE16-NEXT: v_cmp_u_f32_e32 vcc_lo, v0, v0
-; GFX11-TRUE16-NEXT: v_or_b32_e32 v9, 0x400000, v5
-; GFX11-TRUE16-NEXT: v_add3_u32 v7, v7, v5, 0x7fff
-; GFX11-TRUE16-NEXT: v_add3_u32 v6, v6, v0, 0x7fff
+; GFX11-TRUE16-NEXT: v_bfe_u32 v5, v1, 16, 1
+; GFX11-TRUE16-NEXT: v_or_b32_e32 v8, 0x400000, v1
+; GFX11-TRUE16-NEXT: v_cmp_u_f32_e32 vcc_lo, v1, v1
+; GFX11-TRUE16-NEXT: v_or_b32_e32 v9, 0x400000, v2
+; GFX11-TRUE16-NEXT: v_add3_u32 v7, v7, v2, 0x7fff
+; GFX11-TRUE16-NEXT: v_add3_u32 v5, v5, v1, 0x7fff
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(SKIP_1) | instid1(VALU_DEP_4)
-; GFX11-TRUE16-NEXT: v_cndmask_b32_e32 v0, v6, v8, vcc_lo
-; GFX11-TRUE16-NEXT: v_cmp_u_f32_e32 vcc_lo, v5, v5
-; GFX11-TRUE16-NEXT: v_cndmask_b32_e32 v5, v7, v9, vcc_lo
+; GFX11-TRUE16-NEXT: v_cndmask_b32_e32 v1, v5, v8, vcc_lo
+; GFX11-TRUE16-NEXT: v_cmp_u_f32_e32 vcc_lo, v2, v2
+; GFX11-TRUE16-NEXT: v_cndmask_b32_e32 v2, v7, v9, vcc_lo
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_3) | instskip(NEXT) | instid1(VALU_DEP_1)
-; GFX11-TRUE16-NEXT: v_mov_b16_e32 v0.l, v0.h
-; GFX11-TRUE16-NEXT: v_bfi_b32 v0, 0xffff, v0, v5
+; GFX11-TRUE16-NEXT: v_mov_b16_e32 v1.l, v1.h
+; GFX11-TRUE16-NEXT: v_bfi_b32 v5, 0xffff, v1, v2
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1)
-; GFX11-TRUE16-NEXT: v_dual_mov_b32 v6, v1 :: v_dual_mov_b32 v5, v0
-; GFX11-TRUE16-NEXT: buffer_atomic_cmpswap_b32 v[5:6], v4, s[0:3], 0 offen glc
+; GFX11-TRUE16-NEXT: v_dual_mov_b32 v1, v5 :: v_dual_mov_b32 v2, v6
+; GFX11-TRUE16-NEXT: buffer_atomic_cmpswap_b32 v[1:2], v4, s[0:3], 0 offen glc
; GFX11-TRUE16-NEXT: s_waitcnt vmcnt(0)
; GFX11-TRUE16-NEXT: buffer_gl1_inv
; GFX11-TRUE16-NEXT: buffer_gl0_inv
-; GFX11-TRUE16-NEXT: v_cmp_eq_u32_e32 vcc_lo, v5, v1
-; GFX11-TRUE16-NEXT: v_mov_b32_e32 v1, v5
+; GFX11-TRUE16-NEXT: v_cmp_eq_u32_e32 vcc_lo, v1, v6
; GFX11-TRUE16-NEXT: s_or_b32 s4, vcc_lo, s4
; GFX11-TRUE16-NEXT: s_delay_alu instid0(SALU_CYCLE_1)
; GFX11-TRUE16-NEXT: s_and_not1_b32 exec_lo, exec_lo, s4
@@ -11181,10 +11190,11 @@ define void @buffer_fat_ptr_agent_atomic_fadd_noret_v2bf16__offset(ptr addrspace
; GFX11-FAKE16-LABEL: buffer_fat_ptr_agent_atomic_fadd_noret_v2bf16__offset:
; GFX11-FAKE16: ; %bb.0:
; GFX11-FAKE16-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
-; GFX11-FAKE16-NEXT: v_dual_mov_b32 v1, s16 :: v_dual_lshlrev_b32 v2, 16, v0
+; GFX11-FAKE16-NEXT: v_mov_b32_e32 v1, s16
; GFX11-FAKE16-NEXT: s_add_i32 s4, s16, 0x400
; GFX11-FAKE16-NEXT: s_delay_alu instid0(SALU_CYCLE_1)
-; GFX11-FAKE16-NEXT: v_dual_mov_b32 v4, s4 :: v_dual_and_b32 v3, 0xffff0000, v0
+; GFX11-FAKE16-NEXT: v_dual_mov_b32 v4, s4 :: v_dual_lshlrev_b32 v3, 16, v0
+; GFX11-FAKE16-NEXT: v_and_b32_e32 v0, 0xffff0000, v0
; GFX11-FAKE16-NEXT: buffer_load_b32 v1, v1, s[0:3], 0 offen offset:1024
; GFX11-FAKE16-NEXT: s_mov_b32 s5, 0
; GFX11-FAKE16-NEXT: s_set_inst_prefetch_distance 0x1
@@ -11192,32 +11202,33 @@ define void @buffer_fat_ptr_agent_atomic_fadd_noret_v2bf16__offset(ptr addrspace
; GFX11-FAKE16-NEXT: .LBB30_1: ; %atomicrmw.start
; GFX11-FAKE16-NEXT: ; =>This Inner Loop Header: Depth=1
; GFX11-FAKE16-NEXT: s_waitcnt vmcnt(0)
-; GFX11-FAKE16-NEXT: v_and_b32_e32 v5, 0xffff0000, v1
-; GFX11-FAKE16-NEXT: v_lshlrev_b32_e32 v0, 16, v1
+; GFX11-FAKE16-NEXT: v_mov_b32_e32 v6, v1
; GFX11-FAKE16-NEXT: s_waitcnt_vscnt null, 0x0
; GFX11-FAKE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
-; GFX11-FAKE16-NEXT: v_dual_add_f32 v5, v5, v3 :: v_dual_add_f32 v0, v0, v2
-; GFX11-FAKE16-NEXT: v_bfe_u32 v7, v5, 16, 1
+; GFX11-FAKE16-NEXT: v_and_b32_e32 v2, 0xffff0000, v6
+; GFX11-FAKE16-NEXT: v_dual_add_f32 v2, v2, v0 :: v_dual_lshlrev_b32 v1, 16, v6
+; GFX11-FAKE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_2)
+; GFX11-FAKE16-NEXT: v_add_f32_e32 v1, v1, v3
+; GFX11-FAKE16-NEXT: v_bfe_u32 v7, v2, 16, 1
; GFX11-FAKE16-NEXT: s_delay_alu instid0(VALU_DEP_2)
-; GFX11-FAKE16-NEXT: v_bfe_u32 v6, v0, 16, 1
-; GFX11-FAKE16-NEXT: v_or_b32_e32 v8, 0x400000, v0
-; GFX11-FAKE16-NEXT: v_or_b32_e32 v9, 0x400000, v5
-; GFX11-FAKE16-NEXT: v_cmp_u_f32_e32 vcc_lo, v5, v5
-; GFX11-FAKE16-NEXT: v_add3_u32 v7, v7, v5, 0x7fff
-; GFX11-FAKE16-NEXT: v_add3_u32 v6, v6, v0, 0x7fff
-; GFX11-FAKE16-NEXT: v_cmp_u_f32_e64 s4, v0, v0
+; GFX11-FAKE16-NEXT: v_bfe_u32 v5, v1, 16, 1
+; GFX11-FAKE16-NEXT: v_or_b32_e32 v8, 0x400000, v1
+; GFX11-FAKE16-NEXT: v_or_b32_e32 v9, 0x400000, v2
+; GFX11-FAKE16-NEXT: v_cmp_u_f32_e32 vcc_lo, v2, v2
+; GFX11-FAKE16-NEXT: v_add3_u32 v7, v7, v2, 0x7fff
+; GFX11-FAKE16-NEXT: v_add3_u32 v5, v5, v1, 0x7fff
+; GFX11-FAKE16-NEXT: v_cmp_u_f32_e64 s4, v1, v1
; GFX11-FAKE16-NEXT: s_delay_alu instid0(VALU_DEP_3) | instskip(NEXT) | instid1(VALU_DEP_2)
-; GFX11-FAKE16-NEXT: v_cndmask_b32_e32 v5, v7, v9, vcc_lo
-; GFX11-FAKE16-NEXT: v_cndmask_b32_e64 v0, v6, v8, s4
+; GFX11-FAKE16-NEXT: v_cndmask_b32_e32 v2, v7, v9, vcc_lo
+; GFX11-FAKE16-NEXT: v_cndmask_b32_e64 v1, v5, v8, s4
; GFX11-FAKE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
-; GFX11-FAKE16-NEXT: v_perm_b32 v0, v5, v0, 0x7060302
-; GFX11-FAKE16-NEXT: v_dual_mov_b32 v6, v1 :: v_dual_mov_b32 v5, v0
-; GFX11-FAKE16-NEXT: buffer_atomic_cmpswap_b32 v[5:6], v4, s[0:3], 0 offen glc
+; GFX11-FAKE16-NEXT: v_perm_b32 v5, v2, v1, 0x7060302
+; GFX11-FAKE16-NEXT: v_dual_mov_b32 v1, v5 :: v_dual_mov_b32 v2, v6
+; GFX11-FAKE16-NEXT: buffer_atomic_cmpswap_b32 v[1:2], v4, s[0:3], 0 offen glc
; GFX11-FAKE16-NEXT: s_waitcnt vmcnt(0)
; GFX11-FAKE16-NEXT: buffer_gl1_inv
; GFX11-FAKE16-NEXT: buffer_gl0_inv
-; GFX11-FAKE16-NEXT: v_cmp_eq_u32_e32 vcc_lo, v5, v1
-; GFX11-FAKE16-NEXT: v_mov_b32_e32 v1, v5
+; GFX11-FAKE16-NEXT: v_cmp_eq_u32_e32 vcc_lo, v1, v6
; GFX11-FAKE16-NEXT: s_or_b32 s5, vcc_lo, s5
; GFX11-FAKE16-NEXT: s_delay_alu instid0(SALU_CYCLE_1)
; GFX11-FAKE16-NEXT: s_and_not1_b32 exec_lo, exec_lo, s5
@@ -11232,38 +11243,38 @@ define void @buffer_fat_ptr_agent_atomic_fadd_noret_v2bf16__offset(ptr addrspace
; GFX10-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX10-NEXT: v_mov_b32_e32 v1, s20
; GFX10-NEXT: s_add_i32 s4, s20, 0x400
-; GFX10-NEXT: v_lshlrev_b32_e32 v2, 16, v0
-; GFX10-NEXT: v_and_b32_e32 v3, 0xffff0000, v0
+; GFX10-NEXT: v_lshlrev_b32_e32 v3, 16, v0
+; GFX10-NEXT: v_and_b32_e32 v0, 0xffff0000, v0
; GFX10-NEXT: v_mov_b32_e32 v4, s4
; GFX10-NEXT: buffer_load_dword v1, v1, s[16:19], 0 offen offset:1024
; GFX10-NEXT: s_mov_b32 s5, 0
; GFX10-NEXT: .LBB30_1: ; %atomicrmw.start
; GFX10-NEXT: ; =>This Inner Loop Header: Depth=1
; GFX10-NEXT: s_waitcnt vmcnt(0)
-; GFX10-NEXT: v_lshlrev_b32_e32 v0, 16, v1
-; GFX10-NEXT: v_and_b32_e32 v5, 0xffff0000, v1
-; GFX10-NEXT: s_waitcnt_vscnt null, 0x0
-; GFX10-NEXT: v_add_f32_e32 v0, v0, v2
-; GFX10-NEXT: v_add_f32_e32 v5, v5, v3
-; GFX10-NEXT: v_bfe_u32 v6, v0, 16, 1
-; GFX10-NEXT: v_bfe_u32 v7, v5, 16, 1
-; GFX10-NEXT: v_or_b32_e32 v8, 0x400000, v0
-; GFX10-NEXT: v_or_b32_e32 v9, 0x400000, v5
-; GFX10-NEXT: v_cmp_u_f32_e32 vcc_lo, v5, v5
-; GFX10-NEXT: v_add3_u32 v6, v6, v0, 0x7fff
-; GFX10-NEXT: v_add3_u32 v7, v7, v5, 0x7fff
-; GFX10-NEXT: v_cmp_u_f32_e64 s4, v0, v0
-; GFX10-NEXT: v_cndmask_b32_e32 v5, v7, v9, vcc_lo
-; GFX10-NEXT: v_cndmask_b32_e64 v0, v6, v8, s4
-; GFX10-NEXT: v_perm_b32 v0, v5, v0, 0x7060302
; GFX10-NEXT: v_mov_b32_e32 v6, v1
-; GFX10-NEXT: v_mov_b32_e32 v5, v0
-; GFX10-NEXT: buffer_atomic_cmpswap v[5:6], v4, s[16:19], 0 offen glc
+; GFX10-NEXT: s_waitcnt_vscnt null, 0x0
+; GFX10-NEXT: v_lshlrev_b32_e32 v1, 16, v6
+; GFX10-NEXT: v_and_b32_e32 v2, 0xffff0000, v6
+; GFX10-NEXT: v_add_f32_e32 v1, v1, v3
+; GFX10-NEXT: v_add_f32_e32 v2, v2, v0
+; GFX10-NEXT: v_bfe_u32 v5, v1, 16, 1
+; GFX10-NEXT: v_bfe_u32 v7, v2, 16, 1
+; GFX10-NEXT: v_or_b32_e32 v8, 0x400000, v1
+; GFX10-NEXT: v_or_b32_e32 v9, 0x400000, v2
+; GFX10-NEXT: v_cmp_u_f32_e32 vcc_lo, v2, v2
+; GFX10-NEXT: v_add3_u32 v5, v5, v1, 0x7fff
+; GFX10-NEXT: v_add3_u32 v7, v7, v2, 0x7fff
+; GFX10-NEXT: v_cmp_u_f32_e64 s4, v1, v1
+; GFX10-NEXT: v_cndmask_b32_e32 v2, v7, v9, vcc_lo
+; GFX10-NEXT: v_cndmask_b32_e64 v1, v5, v8, s4
+; GFX10-NEXT: v_perm_b32 v5, v2, v1, 0x7060302
+; GFX10-NEXT: v_mov_b32_e32 v1, v5
+; GFX10-NEXT: v_mov_b32_e32 v2, v6
+; GFX10-NEXT: buffer_atomic_cmpswap v[1:2], v4, s[16:19], 0 offen glc
; GFX10-NEXT: s_waitcnt vmcnt(0)
; GFX10-NEXT: buffer_gl1_inv
; GFX10-NEXT: buffer_gl0_inv
-; GFX10-NEXT: v_cmp_eq_u32_e32 vcc_lo, v5, v1
-; GFX10-NEXT: v_mov_b32_e32 v1, v5
+; GFX10-NEXT: v_cmp_eq_u32_e32 vcc_lo, v1, v6
; GFX10-NEXT: s_or_b32 s5, vcc_lo, s5
; GFX10-NEXT: s_andn2_b32 exec_lo, exec_lo, s5
; GFX10-NEXT: s_cbranch_execnz .LBB30_1
@@ -11275,39 +11286,39 @@ define void @buffer_fat_ptr_agent_atomic_fadd_noret_v2bf16__offset(ptr addrspace
; GFX90A: ; %bb.0:
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX90A-NEXT: v_mov_b32_e32 v1, s20
-; GFX90A-NEXT: buffer_load_dword v1, v1, s[16:19], 0 offen offset:1024
+; GFX90A-NEXT: buffer_load_dword v2, v1, s[16:19], 0 offen offset:1024
; GFX90A-NEXT: s_add_i32 s4, s20, 0x400
; GFX90A-NEXT: s_mov_b64 s[6:7], 0
-; GFX90A-NEXT: v_lshlrev_b32_e32 v2, 16, v0
+; GFX90A-NEXT: v_lshlrev_b32_e32 v1, 16, v0
; GFX90A-NEXT: s_movk_i32 s8, 0x7fff
-; GFX90A-NEXT: v_and_b32_e32 v3, 0xffff0000, v0
+; GFX90A-NEXT: v_and_b32_e32 v0, 0xffff0000, v0
; GFX90A-NEXT: s_mov_b32 s9, 0x7060302
; GFX90A-NEXT: v_mov_b32_e32 v4, s4
; GFX90A-NEXT: .LBB30_1: ; %atomicrmw.start
; GFX90A-NEXT: ; =>This Inner Loop Header: Depth=1
; GFX90A-NEXT: s_waitcnt vmcnt(0)
-; GFX90A-NEXT: v_lshlrev_b32_e32 v0, 16, v1
-; GFX90A-NEXT: v_and_b32_e32 v5, 0xffff0000, v1
-; GFX90A-NEXT: v_add_f32_e32 v0, v0, v2
-; GFX90A-NEXT: v_add_f32_e32 v5, v5, v3
-; GFX90A-NEXT: v_bfe_u32 v6, v0, 16, 1
-; GFX90A-NEXT: v_bfe_u32 v8, v5, 16, 1
-; GFX90A-NEXT: v_or_b32_e32 v7, 0x400000, v0
-; GFX90A-NEXT: v_or_b32_e32 v9, 0x400000, v5
-; GFX90A-NEXT: v_add3_u32 v6, v6, v0, s8
-; GFX90A-NEXT: v_add3_u32 v8, v8, v5, s8
-; GFX90A-NEXT: v_cmp_u_f32_e32 vcc, v5, v5
-; GFX90A-NEXT: v_cmp_u_f32_e64 s[4:5], v0, v0
-; GFX90A-NEXT: v_cndmask_b32_e64 v0, v6, v7, s[4:5]
-; GFX90A-NEXT: v_cndmask_b32_e32 v5, v8, v9, vcc
-; GFX90A-NEXT: v_perm_b32 v0, v5, v0, s9
-; GFX90A-NEXT: v_pk_mov_b32 v[6:7], v[0:1], v[0:1] op_sel:[0,1]
-; GFX90A-NEXT: buffer_atomic_cmpswap v[6:7], v4, s[16:19], 0 offen glc
+; GFX90A-NEXT: v_mov_b32_e32 v7, v2
+; GFX90A-NEXT: v_lshlrev_b32_e32 v2, 16, v7
+; GFX90A-NEXT: v_and_b32_e32 v3, 0xffff0000, v7
+; GFX90A-NEXT: v_add_f32_e32 v2, v2, v1
+; GFX90A-NEXT: v_add_f32_e32 v3, v3, v0
+; GFX90A-NEXT: v_bfe_u32 v5, v2, 16, 1
+; GFX90A-NEXT: v_bfe_u32 v8, v3, 16, 1
+; GFX90A-NEXT: v_or_b32_e32 v6, 0x400000, v2
+; GFX90A-NEXT: v_or_b32_e32 v9, 0x400000, v3
+; GFX90A-NEXT: v_add3_u32 v5, v5, v2, s8
+; GFX90A-NEXT: v_add3_u32 v8, v8, v3, s8
+; GFX90A-NEXT: v_cmp_u_f32_e32 vcc, v3, v3
+; GFX90A-NEXT: v_cmp_u_f32_e64 s[4:5], v2, v2
+; GFX90A-NEXT: v_cndmask_b32_e64 v2, v5, v6, s[4:5]
+; GFX90A-NEXT: v_cndmask_b32_e32 v3, v8, v9, vcc
+; GFX90A-NEXT: v_perm_b32 v6, v3, v2, s9
+; GFX90A-NEXT: v_pk_mov_b32 v[2:3], v[6:7], v[6:7] op_sel:[0,1]
+; GFX90A-NEXT: buffer_atomic_cmpswap v[2:3], v4, s[16:19], 0 offen glc
; GFX90A-NEXT: s_waitcnt vmcnt(0)
; GFX90A-NEXT: buffer_wbinvl1
-; GFX90A-NEXT: v_cmp_eq_u32_e32 vcc, v6, v1
+; GFX90A-NEXT: v_cmp_eq_u32_e32 vcc, v2, v7
; GFX90A-NEXT: s_or_b64 s[6:7], vcc, s[6:7]
-; GFX90A-NEXT: v_mov_b32_e32 v1, v6
; GFX90A-NEXT: s_andn2_b64 exec, exec, s[6:7]
; GFX90A-NEXT: s_cbranch_execnz .LBB30_1
; GFX90A-NEXT: ; %bb.2: ; %atomicrmw.end
@@ -11321,37 +11332,37 @@ define void @buffer_fat_ptr_agent_atomic_fadd_noret_v2bf16__offset(ptr addrspace
; GFX908-NEXT: buffer_load_dword v1, v1, s[16:19], 0 offen offset:1024
; GFX908-NEXT: s_add_i32 s4, s20, 0x400
; GFX908-NEXT: s_mov_b64 s[6:7], 0
-; GFX908-NEXT: v_lshlrev_b32_e32 v2, 16, v0
+; GFX908-NEXT: v_lshlrev_b32_e32 v3, 16, v0
; GFX908-NEXT: s_movk_i32 s8, 0x7fff
-; GFX908-NEXT: v_and_b32_e32 v3, 0xffff0000, v0
+; GFX908-NEXT: v_and_b32_e32 v0, 0xffff0000, v0
; GFX908-NEXT: s_mov_b32 s9, 0x7060302
; GFX908-NEXT: v_mov_b32_e32 v4, s4
; GFX908-NEXT: .LBB30_1: ; %atomicrmw.start
; GFX908-NEXT: ; =>This Inner Loop Header: Depth=1
; GFX908-NEXT: s_waitcnt vmcnt(0)
-; GFX908-NEXT: v_lshlrev_b32_e32 v0, 16, v1
-; GFX908-NEXT: v_and_b32_e32 v5, 0xffff0000, v1
-; GFX908-NEXT: v_add_f32_e32 v0, v0, v2
-; GFX908-NEXT: v_add_f32_e32 v5, v5, v3
-; GFX908-NEXT: v_bfe_u32 v6, v0, 16, 1
-; GFX908-NEXT: v_bfe_u32 v8, v5, 16, 1
-; GFX908-NEXT: v_or_b32_e32 v7, 0x400000, v0
-; GFX908-NEXT: v_or_b32_e32 v9, 0x400000, v5
-; GFX908-NEXT: v_add3_u32 v6, v6, v0, s8
-; GFX908-NEXT: v_add3_u32 v8, v8, v5, s8
-; GFX908-NEXT: v_cmp_u_f32_e32 vcc, v5, v5
-; GFX908-NEXT: v_cmp_u_f32_e64 s[4:5], v0, v0
-; GFX908-NEXT: v_cndmask_b32_e64 v0, v6, v7, s[4:5]
-; GFX908-NEXT: v_cndmask_b32_e32 v5, v8, v9, vcc
-; GFX908-NEXT: v_perm_b32 v0, v5, v0, s9
; GFX908-NEXT: v_mov_b32_e32 v6, v1
-; GFX908-NEXT: v_mov_b32_e32 v5, v0
-; GFX908-NEXT: buffer_atomic_cmpswap v[5:6], v4, s[16:19], 0 offen glc
+; GFX908-NEXT: v_lshlrev_b32_e32 v1, 16, v6
+; GFX908-NEXT: v_and_b32_e32 v2, 0xffff0000, v6
+; GFX908-NEXT: v_add_f32_e32 v1, v1, v3
+; GFX908-NEXT: v_add_f32_e32 v2, v2, v0
+; GFX908-NEXT: v_bfe_u32 v5, v1, 16, 1
+; GFX908-NEXT: v_bfe_u32 v8, v2, 16, 1
+; GFX908-NEXT: v_or_b32_e32 v7, 0x400000, v1
+; GFX908-NEXT: v_or_b32_e32 v9, 0x400000, v2
+; GFX908-NEXT: v_add3_u32 v5, v5, v1, s8
+; GFX908-NEXT: v_add3_u32 v8, v8, v2, s8
+; GFX908-NEXT: v_cmp_u_f32_e32 vcc, v2, v2
+; GFX908-NEXT: v_cmp_u_f32_e64 s[4:5], v1, v1
+; GFX908-NEXT: v_cndmask_b32_e64 v1, v5, v7, s[4:5]
+; GFX908-NEXT: v_cndmask_b32_e32 v2, v8, v9, vcc
+; GFX908-NEXT: v_perm_b32 v5, v2, v1, s9
+; GFX908-NEXT: v_mov_b32_e32 v1, v5
+; GFX908-NEXT: v_mov_b32_e32 v2, v6
+; GFX908-NEXT: buffer_atomic_cmpswap v[1:2], v4, s[16:19], 0 offen glc
; GFX908-NEXT: s_waitcnt vmcnt(0)
; GFX908-NEXT: buffer_wbinvl1
-; GFX908-NEXT: v_cmp_eq_u32_e32 vcc, v5, v1
+; GFX908-NEXT: v_cmp_eq_u32_e32 vcc, v1, v6
; GFX908-NEXT: s_or_b64 s[6:7], vcc, s[6:7]
-; GFX908-NEXT: v_mov_b32_e32 v1, v5
; GFX908-NEXT: s_andn2_b64 exec, exec, s[6:7]
; GFX908-NEXT: s_cbranch_execnz .LBB30_1
; GFX908-NEXT: ; %bb.2: ; %atomicrmw.end
@@ -11365,38 +11376,38 @@ define void @buffer_fat_ptr_agent_atomic_fadd_noret_v2bf16__offset(ptr addrspace
; GFX8-NEXT: buffer_load_dword v1, v1, s[16:19], 0 offen offset:1024
; GFX8-NEXT: s_add_i32 s4, s20, 0x400
; GFX8-NEXT: s_mov_b64 s[6:7], 0
-; GFX8-NEXT: v_lshlrev_b32_e32 v2, 16, v0
-; GFX8-NEXT: v_and_b32_e32 v3, 0xffff0000, v0
+; GFX8-NEXT: v_lshlrev_b32_e32 v3, 16, v0
+; GFX8-NEXT: v_and_b32_e32 v0, 0xffff0000, v0
; GFX8-NEXT: v_mov_b32_e32 v4, s4
; GFX8-NEXT: .LBB30_1: ; %atomicrmw.start
; GFX8-NEXT: ; =>This Inner Loop Header: Depth=1
; GFX8-NEXT: s_waitcnt vmcnt(0)
-; GFX8-NEXT: v_lshlrev_b32_e32 v0, 16, v1
-; GFX8-NEXT: v_and_b32_e32 v5, 0xffff0000, v1
-; GFX8-NEXT: v_add_f32_e32 v0, v0, v2
-; GFX8-NEXT: v_add_f32_e32 v5, v5, v3
-; GFX8-NEXT: v_bfe_u32 v6, v0, 16, 1
-; GFX8-NEXT: v_bfe_u32 v8, v5, 16, 1
-; GFX8-NEXT: v_add_u32_e32 v6, vcc, v6, v0
-; GFX8-NEXT: v_add_u32_e32 v8, vcc, v8, v5
-; GFX8-NEXT: v_add_u32_e32 v6, vcc, 0x7fff, v6
-; GFX8-NEXT: v_add_u32_e32 v8, vcc, 0x7fff, v8
-; GFX8-NEXT: v_or_b32_e32 v9, 0x400000, v5
-; GFX8-NEXT: v_cmp_u_f32_e32 vcc, v5, v5
-; GFX8-NEXT: v_or_b32_e32 v7, 0x400000, v0
-; GFX8-NEXT: v_cmp_u_f32_e64 s[4:5], v0, v0
-; GFX8-NEXT: v_cndmask_b32_e32 v5, v8, v9, vcc
-; GFX8-NEXT: v_cndmask_b32_e64 v0, v6, v7, s[4:5]
-; GFX8-NEXT: v_lshrrev_b32_e32 v5, 16, v5
-; GFX8-NEXT: v_alignbit_b32 v0, v5, v0, 16
; GFX8-NEXT: v_mov_b32_e32 v6, v1
-; GFX8-NEXT: v_mov_b32_e32 v5, v0
-; GFX8-NEXT: buffer_atomic_cmpswap v[5:6], v4, s[16:19], 0 offen glc
+; GFX8-NEXT: v_lshlrev_b32_e32 v1, 16, v6
+; GFX8-NEXT: v_and_b32_e32 v2, 0xffff0000, v6
+; GFX8-NEXT: v_add_f32_e32 v1, v1, v3
+; GFX8-NEXT: v_add_f32_e32 v2, v2, v0
+; GFX8-NEXT: v_bfe_u32 v5, v1, 16, 1
+; GFX8-NEXT: v_bfe_u32 v8, v2, 16, 1
+; GFX8-NEXT: v_add_u32_e32 v5, vcc, v5, v1
+; GFX8-NEXT: v_add_u32_e32 v8, vcc, v8, v2
+; GFX8-NEXT: v_add_u32_e32 v5, vcc, 0x7fff, v5
+; GFX8-NEXT: v_add_u32_e32 v8, vcc, 0x7fff, v8
+; GFX8-NEXT: v_or_b32_e32 v9, 0x400000, v2
+; GFX8-NEXT: v_cmp_u_f32_e32 vcc, v2, v2
+; GFX8-NEXT: v_or_b32_e32 v7, 0x400000, v1
+; GFX8-NEXT: v_cmp_u_f32_e64 s[4:5], v1, v1
+; GFX8-NEXT: v_cndmask_b32_e32 v2, v8, v9, vcc
+; GFX8-NEXT: v_cndmask_b32_e64 v1, v5, v7, s[4:5]
+; GFX8-NEXT: v_lshrrev_b32_e32 v2, 16, v2
+; GFX8-NEXT: v_alignbit_b32 v5, v2, v1, 16
+; GFX8-NEXT: v_mov_b32_e32 v1, v5
+; GFX8-NEXT: v_mov_b32_e32 v2, v6
+; GFX8-NEXT: buffer_atomic_cmpswap v[1:2], v4, s[16:19], 0 offen glc
; GFX8-NEXT: s_waitcnt vmcnt(0)
; GFX8-NEXT: buffer_wbinvl1
-; GFX8-NEXT: v_cmp_eq_u32_e32 vcc, v5, v1
+; GFX8-NEXT: v_cmp_eq_u32_e32 vcc, v1, v6
; GFX8-NEXT: s_or_b64 s[6:7], vcc, s[6:7]
-; GFX8-NEXT: v_mov_b32_e32 v1, v5
; GFX8-NEXT: s_andn2_b64 exec, exec, s[6:7]
; GFX8-NEXT: s_cbranch_execnz .LBB30_1
; GFX8-NEXT: ; %bb.2: ; %atomicrmw.end
@@ -11950,40 +11961,40 @@ define void @buffer_fat_ptr_agent_atomic_fadd_noret_v2bf16__offset__amdgpu_no_re
; GFX942: ; %bb.0:
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX942-NEXT: v_mov_b32_e32 v1, s16
-; GFX942-NEXT: buffer_load_dword v1, v1, s[0:3], 0 offen offset:1024
+; GFX942-NEXT: buffer_load_dword v2, v1, s[0:3], 0 offen offset:1024
; GFX942-NEXT: s_add_i32 s4, s16, 0x400
; GFX942-NEXT: s_mov_b64 s[6:7], 0
-; GFX942-NEXT: v_lshlrev_b32_e32 v2, 16, v0
+; GFX942-NEXT: v_lshlrev_b32_e32 v1, 16, v0
; GFX942-NEXT: s_movk_i32 s8, 0x7fff
-; GFX942-NEXT: v_and_b32_e32 v3, 0xffff0000, v0
+; GFX942-NEXT: v_and_b32_e32 v0, 0xffff0000, v0
; GFX942-NEXT: s_mov_b32 s9, 0x7060302
; GFX942-NEXT: v_mov_b32_e32 v4, s4
; GFX942-NEXT: .LBB32_1: ; %atomicrmw.start
; GFX942-NEXT: ; =>This Inner Loop Header: Depth=1
; GFX942-NEXT: s_waitcnt vmcnt(0)
-; GFX942-NEXT: v_lshlrev_b32_e32 v0, 16, v1
-; GFX942-NEXT: v_and_b32_e32 v5, 0xffff0000, v1
-; GFX942-NEXT: v_add_f32_e32 v0, v0, v2
-; GFX942-NEXT: v_add_f32_e32 v5, v5, v3
-; GFX942-NEXT: v_bfe_u32 v6, v0, 16, 1
-; GFX942-NEXT: v_bfe_u32 v8, v5, 16, 1
-; GFX942-NEXT: v_or_b32_e32 v7, 0x400000, v0
-; GFX942-NEXT: v_or_b32_e32 v9, 0x400000, v5
-; GFX942-NEXT: v_add3_u32 v6, v6, v0, s8
-; GFX942-NEXT: v_add3_u32 v8, v8, v5, s8
-; GFX942-NEXT: v_cmp_u_f32_e32 vcc, v5, v5
-; GFX942-NEXT: v_cmp_u_f32_e64 s[4:5], v0, v0
+; GFX942-NEXT: v_mov_b32_e32 v7, v2
+; GFX942-NEXT: v_lshlrev_b32_e32 v2, 16, v7
+; GFX942-NEXT: v_and_b32_e32 v3, 0xffff0000, v7
+; GFX942-NEXT: v_add_f32_e32 v2, v2, v1
+; GFX942-NEXT: v_add_f32_e32 v3, v3, v0
+; GFX942-NEXT: v_bfe_u32 v5, v2, 16, 1
+; GFX942-NEXT: v_bfe_u32 v8, v3, 16, 1
+; GFX942-NEXT: v_or_b32_e32 v6, 0x400000, v2
+; GFX942-NEXT: v_or_b32_e32 v9, 0x400000, v3
+; GFX942-NEXT: v_add3_u32 v5, v5, v2, s8
+; GFX942-NEXT: v_add3_u32 v8, v8, v3, s8
+; GFX942-NEXT: v_cmp_u_f32_e32 vcc, v3, v3
+; GFX942-NEXT: v_cmp_u_f32_e64 s[4:5], v2, v2
; GFX942-NEXT: buffer_wbl2 sc1
-; GFX942-NEXT: v_cndmask_b32_e32 v5, v8, v9, vcc
-; GFX942-NEXT: v_cndmask_b32_e64 v0, v6, v7, s[4:5]
-; GFX942-NEXT: v_perm_b32 v0, v5, v0, s9
-; GFX942-NEXT: v_mov_b64_e32 v[6:7], v[0:1]
-; GFX942-NEXT: buffer_atomic_cmpswap v[6:7], v4, s[0:3], 0 offen sc0
+; GFX942-NEXT: v_cndmask_b32_e32 v3, v8, v9, vcc
+; GFX942-NEXT: v_cndmask_b32_e64 v2, v5, v6, s[4:5]
+; GFX942-NEXT: v_perm_b32 v6, v3, v2, s9
+; GFX942-NEXT: v_mov_b64_e32 v[2:3], v[6:7]
+; GFX942-NEXT: buffer_atomic_cmpswap v[2:3], v4, s[0:3], 0 offen sc0
; GFX942-NEXT: s_waitcnt vmcnt(0)
; GFX942-NEXT: buffer_inv sc1
-; GFX942-NEXT: v_cmp_eq_u32_e32 vcc, v6, v1
+; GFX942-NEXT: v_cmp_eq_u32_e32 vcc, v2, v7
; GFX942-NEXT: s_or_b64 s[6:7], vcc, s[6:7]
-; GFX942-NEXT: v_mov_b32_e32 v1, v6
; GFX942-NEXT: s_andn2_b64 exec, exec, s[6:7]
; GFX942-NEXT: s_cbranch_execnz .LBB32_1
; GFX942-NEXT: ; %bb.2: ; %atomicrmw.end
@@ -11993,10 +12004,11 @@ define void @buffer_fat_ptr_agent_atomic_fadd_noret_v2bf16__offset__amdgpu_no_re
; GFX11-TRUE16-LABEL: buffer_fat_ptr_agent_atomic_fadd_noret_v2bf16__offset__amdgpu_no_remote_memory:
; GFX11-TRUE16: ; %bb.0:
; GFX11-TRUE16-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
-; GFX11-TRUE16-NEXT: v_dual_mov_b32 v1, s16 :: v_dual_and_b32 v2, 0xffff0000, v0
+; GFX11-TRUE16-NEXT: v_mov_b32_e32 v1, s16
; GFX11-TRUE16-NEXT: s_add_i32 s4, s16, 0x400
; GFX11-TRUE16-NEXT: s_delay_alu instid0(SALU_CYCLE_1)
-; GFX11-TRUE16-NEXT: v_dual_mov_b32 v4, s4 :: v_dual_lshlrev_b32 v3, 16, v0
+; GFX11-TRUE16-NEXT: v_dual_mov_b32 v4, s4 :: v_dual_and_b32 v3, 0xffff0000, v0
+; GFX11-TRUE16-NEXT: v_lshlrev_b32_e32 v0, 16, v0
; GFX11-TRUE16-NEXT: buffer_load_b32 v1, v1, s[0:3], 0 offen offset:1024
; GFX11-TRUE16-NEXT: s_mov_b32 s4, 0
; GFX11-TRUE16-NEXT: s_set_inst_prefetch_distance 0x1
@@ -12004,34 +12016,35 @@ define void @buffer_fat_ptr_agent_atomic_fadd_noret_v2bf16__offset__amdgpu_no_re
; GFX11-TRUE16-NEXT: .LBB32_1: ; %atomicrmw.start
; GFX11-TRUE16-NEXT: ; =>This Inner Loop Header: Depth=1
; GFX11-TRUE16-NEXT: s_waitcnt vmcnt(0)
-; GFX11-TRUE16-NEXT: v_and_b32_e32 v5, 0xffff0000, v1
-; GFX11-TRUE16-NEXT: v_lshlrev_b32_e32 v0, 16, v1
+; GFX11-TRUE16-NEXT: v_mov_b32_e32 v6, v1
; GFX11-TRUE16-NEXT: s_waitcnt_vscnt null, 0x0
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
-; GFX11-TRUE16-NEXT: v_dual_add_f32 v5, v5, v2 :: v_dual_add_f32 v0, v0, v3
-; GFX11-TRUE16-NEXT: v_bfe_u32 v7, v5, 16, 1
+; GFX11-TRUE16-NEXT: v_and_b32_e32 v2, 0xffff0000, v6
+; GFX11-TRUE16-NEXT: v_dual_add_f32 v2, v2, v3 :: v_dual_lshlrev_b32 v1, 16, v6
+; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_2)
+; GFX11-TRUE16-NEXT: v_add_f32_e32 v1, v1, v0
+; GFX11-TRUE16-NEXT: v_bfe_u32 v7, v2, 16, 1
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_2)
-; GFX11-TRUE16-NEXT: v_bfe_u32 v6, v0, 16, 1
-; GFX11-TRUE16-NEXT: v_or_b32_e32 v8, 0x400000, v0
-; GFX11-TRUE16-NEXT: v_cmp_u_f32_e32 vcc_lo, v0, v0
-; GFX11-TRUE16-NEXT: v_or_b32_e32 v9, 0x400000, v5
-; GFX11-TRUE16-NEXT: v_add3_u32 v7, v7, v5, 0x7fff
-; GFX11-TRUE16-NEXT: v_add3_u32 v6, v6, v0, 0x7fff
+; GFX11-TRUE16-NEXT: v_bfe_u32 v5, v1, 16, 1
+; GFX11-TRUE16-NEXT: v_or_b32_e32 v8, 0x400000, v1
+; GFX11-TRUE16-NEXT: v_cmp_u_f32_e32 vcc_lo, v1, v1
+; GFX11-TRUE16-NEXT: v_or_b32_e32 v9, 0x400000, v2
+; GFX11-TRUE16-NEXT: v_add3_u32 v7, v7, v2, 0x7fff
+; GFX11-TRUE16-NEXT: v_add3_u32 v5, v5, v1, 0x7fff
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(SKIP_1) | instid1(VALU_DEP_4)
-; GFX11-TRUE16-NEXT: v_cndmask_b32_e32 v0, v6, v8, vcc_lo
-; GFX11-TRUE16-NEXT: v_cmp_u_f32_e32 vcc_lo, v5, v5
-; GFX11-TRUE16-NEXT: v_cndmask_b32_e32 v5, v7, v9, vcc_lo
+; GFX11-TRUE16-NEXT: v_cndmask_b32_e32 v1, v5, v8, vcc_lo
+; GFX11-TRUE16-NEXT: v_cmp_u_f32_e32 vcc_lo, v2, v2
+; GFX11-TRUE16-NEXT: v_cndmask_b32_e32 v2, v7, v9, vcc_lo
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_3) | instskip(NEXT) | instid1(VALU_DEP_1)
-; GFX11-TRUE16-NEXT: v_mov_b16_e32 v0.l, v0.h
-; GFX11-TRUE16-NEXT: v_bfi_b32 v0, 0xffff, v0, v5
+; GFX11-TRUE16-NEXT: v_mov_b16_e32 v1.l, v1.h
+; GFX11-TRUE16-NEXT: v_bfi_b32 v5, 0xffff, v1, v2
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1)
-; GFX11-TRUE16-NEXT: v_dual_mov_b32 v6, v1 :: v_dual_mov_b32 v5, v0
-; GFX11-TRUE16-NEXT: buffer_atomic_cmpswap_b32 v[5:6], v4, s[0:3], 0 offen glc
+; GFX11-TRUE16-NEXT: v_dual_mov_b32 v1, v5 :: v_dual_mov_b32 v2, v6
+; GFX11-TRUE16-NEXT: buffer_atomic_cmpswap_b32 v[1:2], v4, s[0:3], 0 offen glc
; GFX11-TRUE16-NEXT: s_waitcnt vmcnt(0)
; GFX11-TRUE16-NEXT: buffer_gl1_inv
; GFX11-TRUE16-NEXT: buffer_gl0_inv
-; GFX11-TRUE16-NEXT: v_cmp_eq_u32_e32 vcc_lo, v5, v1
-; GFX11-TRUE16-NEXT: v_mov_b32_e32 v1, v5
+; GFX11-TRUE16-NEXT: v_cmp_eq_u32_e32 vcc_lo, v1, v6
; GFX11-TRUE16-NEXT: s_or_b32 s4, vcc_lo, s4
; GFX11-TRUE16-NEXT: s_delay_alu instid0(SALU_CYCLE_1)
; GFX11-TRUE16-NEXT: s_and_not1_b32 exec_lo, exec_lo, s4
@@ -12044,10 +12057,11 @@ define void @buffer_fat_ptr_agent_atomic_fadd_noret_v2bf16__offset__amdgpu_no_re
; GFX11-FAKE16-LABEL: buffer_fat_ptr_agent_atomic_fadd_noret_v2bf16__offset__amdgpu_no_remote_memory:
; GFX11-FAKE16: ; %bb.0:
; GFX11-FAKE16-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
-; GFX11-FAKE16-NEXT: v_dual_mov_b32 v1, s16 :: v_dual_lshlrev_b32 v2, 16, v0
+; GFX11-FAKE16-NEXT: v_mov_b32_e32 v1, s16
; GFX11-FAKE16-NEXT: s_add_i32 s4, s16, 0x400
; GFX11-FAKE16-NEXT: s_delay_alu instid0(SALU_CYCLE_1)
-; GFX11-FAKE16-NEXT: v_dual_mov_b32 v4, s4 :: v_dual_and_b32 v3, 0xffff0000, v0
+; GFX11-FAKE16-NEXT: v_dual_mov_b32 v4, s4 :: v_dual_lshlrev_b32 v3, 16, v0
+; GFX11-FAKE16-NEXT: v_and_b32_e32 v0, 0xffff0000, v0
; GFX11-FAKE16-NEXT: buffer_load_b32 v1, v1, s[0:3], 0 offen offset:1024
; GFX11-FAKE16-NEXT: s_mov_b32 s5, 0
; GFX11-FAKE16-NEXT: s_set_inst_prefetch_distance 0x1
@@ -12055,32 +12069,33 @@ define void @buffer_fat_ptr_agent_atomic_fadd_noret_v2bf16__offset__amdgpu_no_re
; GFX11-FAKE16-NEXT: .LBB32_1: ; %atomicrmw.start
; GFX11-FAKE16-NEXT: ; =>This Inner Loop Header: Depth=1
; GFX11-FAKE16-NEXT: s_waitcnt vmcnt(0)
-; GFX11-FAKE16-NEXT: v_and_b32_e32 v5, 0xffff0000, v1
-; GFX11-FAKE16-NEXT: v_lshlrev_b32_e32 v0, 16, v1
+; GFX11-FAKE16-NEXT: v_mov_b32_e32 v6, v1
; GFX11-FAKE16-NEXT: s_waitcnt_vscnt null, 0x0
; GFX11-FAKE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
-; GFX11-FAKE16-NEXT: v_dual_add_f32 v5, v5, v3 :: v_dual_add_f32 v0, v0, v2
-; GFX11-FAKE16-NEXT: v_bfe_u32 v7, v5, 16, 1
+; GFX11-FAKE16-NEXT: v_and_b32_e32 v2, 0xffff0000, v6
+; GFX11-FAKE16-NEXT: v_dual_add_f32 v2, v2, v0 :: v_dual_lshlrev_b32 v1, 16, v6
+; GFX11-FAKE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_2)
+; GFX11-FAKE16-NEXT: v_add_f32_e32 v1, v1, v3
+; GFX11-FAKE16-NEXT: v_bfe_u32 v7, v2, 16, 1
; GFX11-FAKE16-NEXT: s_delay_alu instid0(VALU_DEP_2)
-; GFX11-FAKE16-NEXT: v_bfe_u32 v6, v0, 16, 1
-; GFX11-FAKE16-NEXT: v_or_b32_e32 v8, 0x400000, v0
-; GFX11-FAKE16-NEXT: v_or_b32_e32 v9, 0x400000, v5
-; GFX11-FAKE16-NEXT: v_cmp_u_f32_e32 vcc_lo, v5, v5
-; GFX11-FAKE16-NEXT: v_add3_u32 v7, v7, v5, 0x7fff
-; GFX11-FAKE16-NEXT: v_add3_u32 v6, v6, v0, 0x7fff
-; GFX11-FAKE16-NEXT: v_cmp_u_f32_e64 s4, v0, v0
+; GFX11-FAKE16-NEXT: v_bfe_u32 v5, v1, 16, 1
+; GFX11-FAKE16-NEXT: v_or_b32_e32 v8, 0x400000, v1
+; GFX11-FAKE16-NEXT: v_or_b32_e32 v9, 0x400000, v2
+; GFX11-FAKE16-NEXT: v_cmp_u_f32_e32 vcc_lo, v2, v2
+; GFX11-FAKE16-NEXT: v_add3_u32 v7, v7, v2, 0x7fff
+; GFX11-FAKE16-NEXT: v_add3_u32 v5, v5, v1, 0x7fff
+; GFX11-FAKE16-NEXT: v_cmp_u_f32_e64 s4, v1, v1
; GFX11-FAKE16-NEXT: s_delay_alu instid0(VALU_DEP_3) | instskip(NEXT) | instid1(VALU_DEP_2)
-; GFX11-FAKE16-NEXT: v_cndmask_b32_e32 v5, v7, v9, vcc_lo
-; GFX11-FAKE16-NEXT: v_cndmask_b32_e64 v0, v6, v8, s4
+; GFX11-FAKE16-NEXT: v_cndmask_b32_e32 v2, v7, v9, vcc_lo
+; GFX11-FAKE16-NEXT: v_cndmask_b32_e64 v1, v5, v8, s4
; GFX11-FAKE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
-; GFX11-FAKE16-NEXT: v_perm_b32 v0, v5, v0, 0x7060302
-; GFX11-FAKE16-NEXT: v_dual_mov_b32 v6, v1 :: v_dual_mov_b32 v5, v0
-; GFX11-FAKE16-NEXT: buffer_atomic_cmpswap_b32 v[5:6], v4, s[0:3], 0 offen glc
+; GFX11-FAKE16-NEXT: v_perm_b32 v5, v2, v1, 0x7060302
+; GFX11-FAKE16-NEXT: v_dual_mov_b32 v1, v5 :: v_dual_mov_b32 v2, v6
+; GFX11-FAKE16-NEXT: buffer_atomic_cmpswap_b32 v[1:2], v4, s[0:3], 0 offen glc
; GFX11-FAKE16-NEXT: s_waitcnt vmcnt(0)
; GFX11-FAKE16-NEXT: buffer_gl1_inv
; GFX11-FAKE16-NEXT: buffer_gl0_inv
-; GFX11-FAKE16-NEXT: v_cmp_eq_u32_e32 vcc_lo, v5, v1
-; GFX11-FAKE16-NEXT: v_mov_b32_e32 v1, v5
+; GFX11-FAKE16-NEXT: v_cmp_eq_u32_e32 vcc_lo, v1, v6
; GFX11-FAKE16-NEXT: s_or_b32 s5, vcc_lo, s5
; GFX11-FAKE16-NEXT: s_delay_alu instid0(SALU_CYCLE_1)
; GFX11-FAKE16-NEXT: s_and_not1_b32 exec_lo, exec_lo, s5
@@ -12095,38 +12110,38 @@ define void @buffer_fat_ptr_agent_atomic_fadd_noret_v2bf16__offset__amdgpu_no_re
; GFX10-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX10-NEXT: v_mov_b32_e32 v1, s20
; GFX10-NEXT: s_add_i32 s4, s20, 0x400
-; GFX10-NEXT: v_lshlrev_b32_e32 v2, 16, v0
-; GFX10-NEXT: v_and_b32_e32 v3, 0xffff0000, v0
+; GFX10-NEXT: v_lshlrev_b32_e32 v3, 16, v0
+; GFX10-NEXT: v_and_b32_e32 v0, 0xffff0000, v0
; GFX10-NEXT: v_mov_b32_e32 v4, s4
; GFX10-NEXT: buffer_load_dword v1, v1, s[16:19], 0 offen offset:1024
; GFX10-NEXT: s_mov_b32 s5, 0
; GFX10-NEXT: .LBB32_1: ; %atomicrmw.start
; GFX10-NEXT: ; =>This Inner Loop Header: Depth=1
; GFX10-NEXT: s_waitcnt vmcnt(0)
-; GFX10-NEXT: v_lshlrev_b32_e32 v0, 16, v1
-; GFX10-NEXT: v_and_b32_e32 v5, 0xffff0000, v1
-; GFX10-NEXT: s_waitcnt_vscnt null, 0x0
-; GFX10-NEXT: v_add_f32_e32 v0, v0, v2
-; GFX10-NEXT: v_add_f32_e32 v5, v5, v3
-; GFX10-NEXT: v_bfe_u32 v6, v0, 16, 1
-; GFX10-NEXT: v_bfe_u32 v7, v5, 16, 1
-; GFX10-NEXT: v_or_b32_e32 v8, 0x400000, v0
-; GFX10-NEXT: v_or_b32_e32 v9, 0x400000, v5
-; GFX10-NEXT: v_cmp_u_f32_e32 vcc_lo, v5, v5
-; GFX10-NEXT: v_add3_u32 v6, v6, v0, 0x7fff
-; GFX10-NEXT: v_add3_u32 v7, v7, v5, 0x7fff
-; GFX10-NEXT: v_cmp_u_f32_e64 s4, v0, v0
-; GFX10-NEXT: v_cndmask_b32_e32 v5, v7, v9, vcc_lo
-; GFX10-NEXT: v_cndmask_b32_e64 v0, v6, v8, s4
-; GFX10-NEXT: v_perm_b32 v0, v5, v0, 0x7060302
; GFX10-NEXT: v_mov_b32_e32 v6, v1
-; GFX10-NEXT: v_mov_b32_e32 v5, v0
-; GFX10-NEXT: buffer_atomic_cmpswap v[5:6], v4, s[16:19], 0 offen glc
+; GFX10-NEXT: s_waitcnt_vscnt null, 0x0
+; GFX10-NEXT: v_lshlrev_b32_e32 v1, 16, v6
+; GFX10-NEXT: v_and_b32_e32 v2, 0xffff0000, v6
+; GFX10-NEXT: v_add_f32_e32 v1, v1, v3
+; GFX10-NEXT: v_add_f32_e32 v2, v2, v0
+; GFX10-NEXT: v_bfe_u32 v5, v1, 16, 1
+; GFX10-NEXT: v_bfe_u32 v7, v2, 16, 1
+; GFX10-NEXT: v_or_b32_e32 v8, 0x400000, v1
+; GFX10-NEXT: v_or_b32_e32 v9, 0x400000, v2
+; GFX10-NEXT: v_cmp_u_f32_e32 vcc_lo, v2, v2
+; GFX10-NEXT: v_add3_u32 v5, v5, v1, 0x7fff
+; GFX10-NEXT: v_add3_u32 v7, v7, v2, 0x7fff
+; GFX10-NEXT: v_cmp_u_f32_e64 s4, v1, v1
+; GFX10-NEXT: v_cndmask_b32_e32 v2, v7, v9, vcc_lo
+; GFX10-NEXT: v_cndmask_b32_e64 v1, v5, v8, s4
+; GFX10-NEXT: v_perm_b32 v5, v2, v1, 0x7060302
+; GFX10-NEXT: v_mov_b32_e32 v1, v5
+; GFX10-NEXT: v_mov_b32_e32 v2, v6
+; GFX10-NEXT: buffer_atomic_cmpswap v[1:2], v4, s[16:19], 0 offen glc
; GFX10-NEXT: s_waitcnt vmcnt(0)
; GFX10-NEXT: buffer_gl1_inv
; GFX10-NEXT: buffer_gl0_inv
-; GFX10-NEXT: v_cmp_eq_u32_e32 vcc_lo, v5, v1
-; GFX10-NEXT: v_mov_b32_e32 v1, v5
+; GFX10-NEXT: v_cmp_eq_u32_e32 vcc_lo, v1, v6
; GFX10-NEXT: s_or_b32 s5, vcc_lo, s5
; GFX10-NEXT: s_andn2_b32 exec_lo, exec_lo, s5
; GFX10-NEXT: s_cbranch_execnz .LBB32_1
@@ -12138,39 +12153,39 @@ define void @buffer_fat_ptr_agent_atomic_fadd_noret_v2bf16__offset__amdgpu_no_re
; GFX90A: ; %bb.0:
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX90A-NEXT: v_mov_b32_e32 v1, s20
-; GFX90A-NEXT: buffer_load_dword v1, v1, s[16:19], 0 offen offset:1024
+; GFX90A-NEXT: buffer_load_dword v2, v1, s[16:19], 0 offen offset:1024
; GFX90A-NEXT: s_add_i32 s4, s20, 0x400
; GFX90A-NEXT: s_mov_b64 s[6:7], 0
-; GFX90A-NEXT: v_lshlrev_b32_e32 v2, 16, v0
+; GFX90A-NEXT: v_lshlrev_b32_e32 v1, 16, v0
; GFX90A-NEXT: s_movk_i32 s8, 0x7fff
-; GFX90A-NEXT: v_and_b32_e32 v3, 0xffff0000, v0
+; GFX90A-NEXT: v_and_b32_e32 v0, 0xffff0000, v0
; GFX90A-NEXT: s_mov_b32 s9, 0x7060302
; GFX90A-NEXT: v_mov_b32_e32 v4, s4
; GFX90A-NEXT: .LBB32_1: ; %atomicrmw.start
; GFX90A-NEXT: ; =>This Inner Loop Header: Depth=1
; GFX90A-NEXT: s_waitcnt vmcnt(0)
-; GFX90A-NEXT: v_lshlrev_b32_e32 v0, 16, v1
-; GFX90A-NEXT: v_and_b32_e32 v5, 0xffff0000, v1
-; GFX90A-NEXT: v_add_f32_e32 v0, v0, v2
-; GFX90A-NEXT: v_add_f32_e32 v5, v5, v3
-; GFX90A-NEXT: v_bfe_u32 v6, v0, 16, 1
-; GFX90A-NEXT: v_bfe_u32 v8, v5, 16, 1
-; GFX90A-NEXT: v_or_b32_e32 v7, 0x400000, v0
-; GFX90A-NEXT: v_or_b32_e32 v9, 0x400000, v5
-; GFX90A-NEXT: v_add3_u32 v6, v6, v0, s8
-; GFX90A-NEXT: v_add3_u32 v8, v8, v5, s8
-; GFX90A-NEXT: v_cmp_u_f32_e32 vcc, v5, v5
-; GFX90A-NEXT: v_cmp_u_f32_e64 s[4:5], v0, v0
-; GFX90A-NEXT: v_cndmask_b32_e64 v0, v6, v7, s[4:5]
-; GFX90A-NEXT: v_cndmask_b32_e32 v5, v8, v9, vcc
-; GFX90A-NEXT: v_perm_b32 v0, v5, v0, s9
-; GFX90A-NEXT: v_pk_mov_b32 v[6:7], v[0:1], v[0:1] op_sel:[0,1]
-; GFX90A-NEXT: buffer_atomic_cmpswap v[6:7], v4, s[16:19], 0 offen glc
+; GFX90A-NEXT: v_mov_b32_e32 v7, v2
+; GFX90A-NEXT: v_lshlrev_b32_e32 v2, 16, v7
+; GFX90A-NEXT: v_and_b32_e32 v3, 0xffff0000, v7
+; GFX90A-NEXT: v_add_f32_e32 v2, v2, v1
+; GFX90A-NEXT: v_add_f32_e32 v3, v3, v0
+; GFX90A-NEXT: v_bfe_u32 v5, v2, 16, 1
+; GFX90A-NEXT: v_bfe_u32 v8, v3, 16, 1
+; GFX90A-NEXT: v_or_b32_e32 v6, 0x400000, v2
+; GFX90A-NEXT: v_or_b32_e32 v9, 0x400000, v3
+; GFX90A-NEXT: v_add3_u32 v5, v5, v2, s8
+; GFX90A-NEXT: v_add3_u32 v8, v8, v3, s8
+; GFX90A-NEXT: v_cmp_u_f32_e32 vcc, v3, v3
+; GFX90A-NEXT: v_cmp_u_f32_e64 s[4:5], v2, v2
+; GFX90A-NEXT: v_cndmask_b32_e64 v2, v5, v6, s[4:5]
+; GFX90A-NEXT: v_cndmask_b32_e32 v3, v8, v9, vcc
+; GFX90A-NEXT: v_perm_b32 v6, v3, v2, s9
+; GFX90A-NEXT: v_pk_mov_b32 v[2:3], v[6:7], v[6:7] op_sel:[0,1]
+; GFX90A-NEXT: buffer_atomic_cmpswap v[2:3], v4, s[16:19], 0 offen glc
; GFX90A-NEXT: s_waitcnt vmcnt(0)
; GFX90A-NEXT: buffer_wbinvl1
-; GFX90A-NEXT: v_cmp_eq_u32_e32 vcc, v6, v1
+; GFX90A-NEXT: v_cmp_eq_u32_e32 vcc, v2, v7
; GFX90A-NEXT: s_or_b64 s[6:7], vcc, s[6:7]
-; GFX90A-NEXT: v_mov_b32_e32 v1, v6
; GFX90A-NEXT: s_andn2_b64 exec, exec, s[6:7]
; GFX90A-NEXT: s_cbranch_execnz .LBB32_1
; GFX90A-NEXT: ; %bb.2: ; %atomicrmw.end
@@ -12184,37 +12199,37 @@ define void @buffer_fat_ptr_agent_atomic_fadd_noret_v2bf16__offset__amdgpu_no_re
; GFX908-NEXT: buffer_load_dword v1, v1, s[16:19], 0 offen offset:1024
; GFX908-NEXT: s_add_i32 s4, s20, 0x400
; GFX908-NEXT: s_mov_b64 s[6:7], 0
-; GFX908-NEXT: v_lshlrev_b32_e32 v2, 16, v0
+; GFX908-NEXT: v_lshlrev_b32_e32 v3, 16, v0
; GFX908-NEXT: s_movk_i32 s8, 0x7fff
-; GFX908-NEXT: v_and_b32_e32 v3, 0xffff0000, v0
+; GFX908-NEXT: v_and_b32_e32 v0, 0xffff0000, v0
; GFX908-NEXT: s_mov_b32 s9, 0x7060302
; GFX908-NEXT: v_mov_b32_e32 v4, s4
; GFX908-NEXT: .LBB32_1: ; %atomicrmw.start
; GFX908-NEXT: ; =>This Inner Loop Header: Depth=1
; GFX908-NEXT: s_waitcnt vmcnt(0)
-; GFX908-NEXT: v_lshlrev_b32_e32 v0, 16, v1
-; GFX908-NEXT: v_and_b32_e32 v5, 0xffff0000, v1
-; GFX908-NEXT: v_add_f32_e32 v0, v0, v2
-; GFX908-NEXT: v_add_f32_e32 v5, v5, v3
-; GFX908-NEXT: v_bfe_u32 v6, v0, 16, 1
-; GFX908-NEXT: v_bfe_u32 v8, v5, 16, 1
-; GFX908-NEXT: v_or_b32_e32 v7, 0x400000, v0
-; GFX908-NEXT: v_or_b32_e32 v9, 0x400000, v5
-; GFX908-NEXT: v_add3_u32 v6, v6, v0, s8
-; GFX908-NEXT: v_add3_u32 v8, v8, v5, s8
-; GFX908-NEXT: v_cmp_u_f32_e32 vcc, v5, v5
-; GFX908-NEXT: v_cmp_u_f32_e64 s[4:5], v0, v0
-; GFX908-NEXT: v_cndmask_b32_e64 v0, v6, v7, s[4:5]
-; GFX908-NEXT: v_cndmask_b32_e32 v5, v8, v9, vcc
-; GFX908-NEXT: v_perm_b32 v0, v5, v0, s9
; GFX908-NEXT: v_mov_b32_e32 v6, v1
-; GFX908-NEXT: v_mov_b32_e32 v5, v0
-; GFX908-NEXT: buffer_atomic_cmpswap v[5:6], v4, s[16:19], 0 offen glc
+; GFX908-NEXT: v_lshlrev_b32_e32 v1, 16, v6
+; GFX908-NEXT: v_and_b32_e32 v2, 0xffff0000, v6
+; GFX908-NEXT: v_add_f32_e32 v1, v1, v3
+; GFX908-NEXT: v_add_f32_e32 v2, v2, v0
+; GFX908-NEXT: v_bfe_u32 v5, v1, 16, 1
+; GFX908-NEXT: v_bfe_u32 v8, v2, 16, 1
+; GFX908-NEXT: v_or_b32_e32 v7, 0x400000, v1
+; GFX908-NEXT: v_or_b32_e32 v9, 0x400000, v2
+; GFX908-NEXT: v_add3_u32 v5, v5, v1, s8
+; GFX908-NEXT: v_add3_u32 v8, v8, v2, s8
+; GFX908-NEXT: v_cmp_u_f32_e32 vcc, v2, v2
+; GFX908-NEXT: v_cmp_u_f32_e64 s[4:5], v1, v1
+; GFX908-NEXT: v_cndmask_b32_e64 v1, v5, v7, s[4:5]
+; GFX908-NEXT: v_cndmask_b32_e32 v2, v8, v9, vcc
+; GFX908-NEXT: v_perm_b32 v5, v2, v1, s9
+; GFX908-NEXT: v_mov_b32_e32 v1, v5
+; GFX908-NEXT: v_mov_b32_e32 v2, v6
+; GFX908-NEXT: buffer_atomic_cmpswap v[1:2], v4, s[16:19], 0 offen glc
; GFX908-NEXT: s_waitcnt vmcnt(0)
; GFX908-NEXT: buffer_wbinvl1
-; GFX908-NEXT: v_cmp_eq_u32_e32 vcc, v5, v1
+; GFX908-NEXT: v_cmp_eq_u32_e32 vcc, v1, v6
; GFX908-NEXT: s_or_b64 s[6:7], vcc, s[6:7]
-; GFX908-NEXT: v_mov_b32_e32 v1, v5
; GFX908-NEXT: s_andn2_b64 exec, exec, s[6:7]
; GFX908-NEXT: s_cbranch_execnz .LBB32_1
; GFX908-NEXT: ; %bb.2: ; %atomicrmw.end
@@ -12228,38 +12243,38 @@ define void @buffer_fat_ptr_agent_atomic_fadd_noret_v2bf16__offset__amdgpu_no_re
; GFX8-NEXT: buffer_load_dword v1, v1, s[16:19], 0 offen offset:1024
; GFX8-NEXT: s_add_i32 s4, s20, 0x400
; GFX8-NEXT: s_mov_b64 s[6:7], 0
-; GFX8-NEXT: v_lshlrev_b32_e32 v2, 16, v0
-; GFX8-NEXT: v_and_b32_e32 v3, 0xffff0000, v0
+; GFX8-NEXT: v_lshlrev_b32_e32 v3, 16, v0
+; GFX8-NEXT: v_and_b32_e32 v0, 0xffff0000, v0
; GFX8-NEXT: v_mov_b32_e32 v4, s4
; GFX8-NEXT: .LBB32_1: ; %atomicrmw.start
; GFX8-NEXT: ; =>This Inner Loop Header: Depth=1
; GFX8-NEXT: s_waitcnt vmcnt(0)
-; GFX8-NEXT: v_lshlrev_b32_e32 v0, 16, v1
-; GFX8-NEXT: v_and_b32_e32 v5, 0xffff0000, v1
-; GFX8-NEXT: v_add_f32_e32 v0, v0, v2
-; GFX8-NEXT: v_add_f32_e32 v5, v5, v3
-; GFX8-NEXT: v_bfe_u32 v6, v0, 16, 1
-; GFX8-NEXT: v_bfe_u32 v8, v5, 16, 1
-; GFX8-NEXT: v_add_u32_e32 v6, vcc, v6, v0
-; GFX8-NEXT: v_add_u32_e32 v8, vcc, v8, v5
-; GFX8-NEXT: v_add_u32_e32 v6, vcc, 0x7fff, v6
-; GFX8-NEXT: v_add_u32_e32 v8, vcc, 0x7fff, v8
-; GFX8-NEXT: v_or_b32_e32 v9, 0x400000, v5
-; GFX8-NEXT: v_cmp_u_f32_e32 vcc, v5, v5
-; GFX8-NEXT: v_or_b32_e32 v7, 0x400000, v0
-; GFX8-NEXT: v_cmp_u_f32_e64 s[4:5], v0, v0
-; GFX8-NEXT: v_cndmask_b32_e32 v5, v8, v9, vcc
-; GFX8-NEXT: v_cndmask_b32_e64 v0, v6, v7, s[4:5]
-; GFX8-NEXT: v_lshrrev_b32_e32 v5, 16, v5
-; GFX8-NEXT: v_alignbit_b32 v0, v5, v0, 16
; GFX8-NEXT: v_mov_b32_e32 v6, v1
-; GFX8-NEXT: v_mov_b32_e32 v5, v0
-; GFX8-NEXT: buffer_atomic_cmpswap v[5:6], v4, s[16:19], 0 offen glc
+; GFX8-NEXT: v_lshlrev_b32_e32 v1, 16, v6
+; GFX8-NEXT: v_and_b32_e32 v2, 0xffff0000, v6
+; GFX8-NEXT: v_add_f32_e32 v1, v1, v3
+; GFX8-NEXT: v_add_f32_e32 v2, v2, v0
+; GFX8-NEXT: v_bfe_u32 v5, v1, 16, 1
+; GFX8-NEXT: v_bfe_u32 v8, v2, 16, 1
+; GFX8-NEXT: v_add_u32_e32 v5, vcc, v5, v1
+; GFX8-NEXT: v_add_u32_e32 v8, vcc, v8, v2
+; GFX8-NEXT: v_add_u32_e32 v5, vcc, 0x7fff, v5
+; GFX8-NEXT: v_add_u32_e32 v8, vcc, 0x7fff, v8
+; GFX8-NEXT: v_or_b32_e32 v9, 0x400000, v2
+; GFX8-NEXT: v_cmp_u_f32_e32 vcc, v2, v2
+; GFX8-NEXT: v_or_b32_e32 v7, 0x400000, v1
+; GFX8-NEXT: v_cmp_u_f32_e64 s[4:5], v1, v1
+; GFX8-NEXT: v_cndmask_b32_e32 v2, v8, v9, vcc
+; GFX8-NEXT: v_cndmask_b32_e64 v1, v5, v7, s[4:5]
+; GFX8-NEXT: v_lshrrev_b32_e32 v2, 16, v2
+; GFX8-NEXT: v_alignbit_b32 v5, v2, v1, 16
+; GFX8-NEXT: v_mov_b32_e32 v1, v5
+; GFX8-NEXT: v_mov_b32_e32 v2, v6
+; GFX8-NEXT: buffer_atomic_cmpswap v[1:2], v4, s[16:19], 0 offen glc
; GFX8-NEXT: s_waitcnt vmcnt(0)
; GFX8-NEXT: buffer_wbinvl1
-; GFX8-NEXT: v_cmp_eq_u32_e32 vcc, v5, v1
+; GFX8-NEXT: v_cmp_eq_u32_e32 vcc, v1, v6
; GFX8-NEXT: s_or_b64 s[6:7], vcc, s[6:7]
-; GFX8-NEXT: v_mov_b32_e32 v1, v5
; GFX8-NEXT: s_andn2_b64 exec, exec, s[6:7]
; GFX8-NEXT: s_cbranch_execnz .LBB32_1
; GFX8-NEXT: ; %bb.2: ; %atomicrmw.end
@@ -12376,40 +12391,40 @@ define void @buffer_fat_ptr_agent_atomic_fadd_noret_v2bf16__offset__amdgpu_no_fi
; GFX942: ; %bb.0:
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX942-NEXT: v_mov_b32_e32 v1, s16
-; GFX942-NEXT: buffer_load_dword v1, v1, s[0:3], 0 offen offset:1024
+; GFX942-NEXT: buffer_load_dword v2, v1, s[0:3], 0 offen offset:1024
; GFX942-NEXT: s_add_i32 s4, s16, 0x400
; GFX942-NEXT: s_mov_b64 s[6:7], 0
-; GFX942-NEXT: v_lshlrev_b32_e32 v2, 16, v0
+; GFX942-NEXT: v_lshlrev_b32_e32 v1, 16, v0
; GFX942-NEXT: s_movk_i32 s8, 0x7fff
-; GFX942-NEXT: v_and_b32_e32 v3, 0xffff0000, v0
+; GFX942-NEXT: v_and_b32_e32 v0, 0xffff0000, v0
; GFX942-NEXT: s_mov_b32 s9, 0x7060302
; GFX942-NEXT: v_mov_b32_e32 v4, s4
; GFX942-NEXT: .LBB33_1: ; %atomicrmw.start
; GFX942-NEXT: ; =>This Inner Loop Header: Depth=1
; GFX942-NEXT: s_waitcnt vmcnt(0)
-; GFX942-NEXT: v_lshlrev_b32_e32 v0, 16, v1
-; GFX942-NEXT: v_and_b32_e32 v5, 0xffff0000, v1
-; GFX942-NEXT: v_add_f32_e32 v0, v0, v2
-; GFX942-NEXT: v_add_f32_e32 v5, v5, v3
-; GFX942-NEXT: v_bfe_u32 v6, v0, 16, 1
-; GFX942-NEXT: v_bfe_u32 v8, v5, 16, 1
-; GFX942-NEXT: v_or_b32_e32 v7, 0x400000, v0
-; GFX942-NEXT: v_or_b32_e32 v9, 0x400000, v5
-; GFX942-NEXT: v_add3_u32 v6, v6, v0, s8
-; GFX942-NEXT: v_add3_u32 v8, v8, v5, s8
-; GFX942-NEXT: v_cmp_u_f32_e32 vcc, v5, v5
-; GFX942-NEXT: v_cmp_u_f32_e64 s[4:5], v0, v0
+; GFX942-NEXT: v_mov_b32_e32 v7, v2
+; GFX942-NEXT: v_lshlrev_b32_e32 v2, 16, v7
+; GFX942-NEXT: v_and_b32_e32 v3, 0xffff0000, v7
+; GFX942-NEXT: v_add_f32_e32 v2, v2, v1
+; GFX942-NEXT: v_add_f32_e32 v3, v3, v0
+; GFX942-NEXT: v_bfe_u32 v5, v2, 16, 1
+; GFX942-NEXT: v_bfe_u32 v8, v3, 16, 1
+; GFX942-NEXT: v_or_b32_e32 v6, 0x400000, v2
+; GFX942-NEXT: v_or_b32_e32 v9, 0x400000, v3
+; GFX942-NEXT: v_add3_u32 v5, v5, v2, s8
+; GFX942-NEXT: v_add3_u32 v8, v8, v3, s8
+; GFX942-NEXT: v_cmp_u_f32_e32 vcc, v3, v3
+; GFX942-NEXT: v_cmp_u_f32_e64 s[4:5], v2, v2
; GFX942-NEXT: buffer_wbl2 sc1
-; GFX942-NEXT: v_cndmask_b32_e32 v5, v8, v9, vcc
-; GFX942-NEXT: v_cndmask_b32_e64 v0, v6, v7, s[4:5]
-; GFX942-NEXT: v_perm_b32 v0, v5, v0, s9
-; GFX942-NEXT: v_mov_b64_e32 v[6:7], v[0:1]
-; GFX942-NEXT: buffer_atomic_cmpswap v[6:7], v4, s[0:3], 0 offen sc0
+; GFX942-NEXT: v_cndmask_b32_e32 v3, v8, v9, vcc
+; GFX942-NEXT: v_cndmask_b32_e64 v2, v5, v6, s[4:5]
+; GFX942-NEXT: v_perm_b32 v6, v3, v2, s9
+; GFX942-NEXT: v_mov_b64_e32 v[2:3], v[6:7]
+; GFX942-NEXT: buffer_atomic_cmpswap v[2:3], v4, s[0:3], 0 offen sc0
; GFX942-NEXT: s_waitcnt vmcnt(0)
; GFX942-NEXT: buffer_inv sc1
-; GFX942-NEXT: v_cmp_eq_u32_e32 vcc, v6, v1
+; GFX942-NEXT: v_cmp_eq_u32_e32 vcc, v2, v7
; GFX942-NEXT: s_or_b64 s[6:7], vcc, s[6:7]
-; GFX942-NEXT: v_mov_b32_e32 v1, v6
; GFX942-NEXT: s_andn2_b64 exec, exec, s[6:7]
; GFX942-NEXT: s_cbranch_execnz .LBB33_1
; GFX942-NEXT: ; %bb.2: ; %atomicrmw.end
@@ -12419,10 +12434,11 @@ define void @buffer_fat_ptr_agent_atomic_fadd_noret_v2bf16__offset__amdgpu_no_fi
; GFX11-TRUE16-LABEL: buffer_fat_ptr_agent_atomic_fadd_noret_v2bf16__offset__amdgpu_no_fine_grained_memory__amdgpu_no_remote_memory:
; GFX11-TRUE16: ; %bb.0:
; GFX11-TRUE16-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
-; GFX11-TRUE16-NEXT: v_dual_mov_b32 v1, s16 :: v_dual_and_b32 v2, 0xffff0000, v0
+; GFX11-TRUE16-NEXT: v_mov_b32_e32 v1, s16
; GFX11-TRUE16-NEXT: s_add_i32 s4, s16, 0x400
; GFX11-TRUE16-NEXT: s_delay_alu instid0(SALU_CYCLE_1)
-; GFX11-TRUE16-NEXT: v_dual_mov_b32 v4, s4 :: v_dual_lshlrev_b32 v3, 16, v0
+; GFX11-TRUE16-NEXT: v_dual_mov_b32 v4, s4 :: v_dual_and_b32 v3, 0xffff0000, v0
+; GFX11-TRUE16-NEXT: v_lshlrev_b32_e32 v0, 16, v0
; GFX11-TRUE16-NEXT: buffer_load_b32 v1, v1, s[0:3], 0 offen offset:1024
; GFX11-TRUE16-NEXT: s_mov_b32 s4, 0
; GFX11-TRUE16-NEXT: s_set_inst_prefetch_distance 0x1
@@ -12430,34 +12446,35 @@ define void @buffer_fat_ptr_agent_atomic_fadd_noret_v2bf16__offset__amdgpu_no_fi
; GFX11-TRUE16-NEXT: .LBB33_1: ; %atomicrmw.start
; GFX11-TRUE16-NEXT: ; =>This Inner Loop Header: Depth=1
; GFX11-TRUE16-NEXT: s_waitcnt vmcnt(0)
-; GFX11-TRUE16-NEXT: v_and_b32_e32 v5, 0xffff0000, v1
-; GFX11-TRUE16-NEXT: v_lshlrev_b32_e32 v0, 16, v1
+; GFX11-TRUE16-NEXT: v_mov_b32_e32 v6, v1
; GFX11-TRUE16-NEXT: s_waitcnt_vscnt null, 0x0
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
-; GFX11-TRUE16-NEXT: v_dual_add_f32 v5, v5, v2 :: v_dual_add_f32 v0, v0, v3
-; GFX11-TRUE16-NEXT: v_bfe_u32 v7, v5, 16, 1
+; GFX11-TRUE16-NEXT: v_and_b32_e32 v2, 0xffff0000, v6
+; GFX11-TRUE16-NEXT: v_dual_add_f32 v2, v2, v3 :: v_dual_lshlrev_b32 v1, 16, v6
+; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_2)
+; GFX11-TRUE16-NEXT: v_add_f32_e32 v1, v1, v0
+; GFX11-TRUE16-NEXT: v_bfe_u32 v7, v2, 16, 1
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_2)
-; GFX11-TRUE16-NEXT: v_bfe_u32 v6, v0, 16, 1
-; GFX11-TRUE16-NEXT: v_or_b32_e32 v8, 0x400000, v0
-; GFX11-TRUE16-NEXT: v_cmp_u_f32_e32 vcc_lo, v0, v0
-; GFX11-TRUE16-NEXT: v_or_b32_e32 v9, 0x400000, v5
-; GFX11-TRUE16-NEXT: v_add3_u32 v7, v7, v5, 0x7fff
-; GFX11-TRUE16-NEXT: v_add3_u32 v6, v6, v0, 0x7fff
+; GFX11-TRUE16-NEXT: v_bfe_u32 v5, v1, 16, 1
+; GFX11-TRUE16-NEXT: v_or_b32_e32 v8, 0x400000, v1
+; GFX11-TRUE16-NEXT: v_cmp_u_f32_e32 vcc_lo, v1, v1
+; GFX11-TRUE16-NEXT: v_or_b32_e32 v9, 0x400000, v2
+; GFX11-TRUE16-NEXT: v_add3_u32 v7, v7, v2, 0x7fff
+; GFX11-TRUE16-NEXT: v_add3_u32 v5, v5, v1, 0x7fff
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(SKIP_1) | instid1(VALU_DEP_4)
-; GFX11-TRUE16-NEXT: v_cndmask_b32_e32 v0, v6, v8, vcc_lo
-; GFX11-TRUE16-NEXT: v_cmp_u_f32_e32 vcc_lo, v5, v5
-; GFX11-TRUE16-NEXT: v_cndmask_b32_e32 v5, v7, v9, vcc_lo
+; GFX11-TRUE16-NEXT: v_cndmask_b32_e32 v1, v5, v8, vcc_lo
+; GFX11-TRUE16-NEXT: v_cmp_u_f32_e32 vcc_lo, v2, v2
+; GFX11-TRUE16-NEXT: v_cndmask_b32_e32 v2, v7, v9, vcc_lo
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_3) | instskip(NEXT) | instid1(VALU_DEP_1)
-; GFX11-TRUE16-NEXT: v_mov_b16_e32 v0.l, v0.h
-; GFX11-TRUE16-NEXT: v_bfi_b32 v0, 0xffff, v0, v5
+; GFX11-TRUE16-NEXT: v_mov_b16_e32 v1.l, v1.h
+; GFX11-TRUE16-NEXT: v_bfi_b32 v5, 0xffff, v1, v2
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1)
-; GFX11-TRUE16-NEXT: v_dual_mov_b32 v6, v1 :: v_dual_mov_b32 v5, v0
-; GFX11-TRUE16-NEXT: buffer_atomic_cmpswap_b32 v[5:6], v4, s[0:3], 0 offen glc
+; GFX11-TRUE16-NEXT: v_dual_mov_b32 v1, v5 :: v_dual_mov_b32 v2, v6
+; GFX11-TRUE16-NEXT: buffer_atomic_cmpswap_b32 v[1:2], v4, s[0:3], 0 offen glc
; GFX11-TRUE16-NEXT: s_waitcnt vmcnt(0)
; GFX11-TRUE16-NEXT: buffer_gl1_inv
; GFX11-TRUE16-NEXT: buffer_gl0_inv
-; GFX11-TRUE16-NEXT: v_cmp_eq_u32_e32 vcc_lo, v5, v1
-; GFX11-TRUE16-NEXT: v_mov_b32_e32 v1, v5
+; GFX11-TRUE16-NEXT: v_cmp_eq_u32_e32 vcc_lo, v1, v6
; GFX11-TRUE16-NEXT: s_or_b32 s4, vcc_lo, s4
; GFX11-TRUE16-NEXT: s_delay_alu instid0(SALU_CYCLE_1)
; GFX11-TRUE16-NEXT: s_and_not1_b32 exec_lo, exec_lo, s4
@@ -12470,10 +12487,11 @@ define void @buffer_fat_ptr_agent_atomic_fadd_noret_v2bf16__offset__amdgpu_no_fi
; GFX11-FAKE16-LABEL: buffer_fat_ptr_agent_atomic_fadd_noret_v2bf16__offset__amdgpu_no_fine_grained_memory__amdgpu_no_remote_memory:
; GFX11-FAKE16: ; %bb.0:
; GFX11-FAKE16-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
-; GFX11-FAKE16-NEXT: v_dual_mov_b32 v1, s16 :: v_dual_lshlrev_b32 v2, 16, v0
+; GFX11-FAKE16-NEXT: v_mov_b32_e32 v1, s16
; GFX11-FAKE16-NEXT: s_add_i32 s4, s16, 0x400
; GFX11-FAKE16-NEXT: s_delay_alu instid0(SALU_CYCLE_1)
-; GFX11-FAKE16-NEXT: v_dual_mov_b32 v4, s4 :: v_dual_and_b32 v3, 0xffff0000, v0
+; GFX11-FAKE16-NEXT: v_dual_mov_b32 v4, s4 :: v_dual_lshlrev_b32 v3, 16, v0
+; GFX11-FAKE16-NEXT: v_and_b32_e32 v0, 0xffff0000, v0
; GFX11-FAKE16-NEXT: buffer_load_b32 v1, v1, s[0:3], 0 offen offset:1024
; GFX11-FAKE16-NEXT: s_mov_b32 s5, 0
; GFX11-FAKE16-NEXT: s_set_inst_prefetch_distance 0x1
@@ -12481,32 +12499,33 @@ define void @buffer_fat_ptr_agent_atomic_fadd_noret_v2bf16__offset__amdgpu_no_fi
; GFX11-FAKE16-NEXT: .LBB33_1: ; %atomicrmw.start
; GFX11-FAKE16-NEXT: ; =>This Inner Loop Header: Depth=1
; GFX11-FAKE16-NEXT: s_waitcnt vmcnt(0)
-; GFX11-FAKE16-NEXT: v_and_b32_e32 v5, 0xffff0000, v1
-; GFX11-FAKE16-NEXT: v_lshlrev_b32_e32 v0, 16, v1
+; GFX11-FAKE16-NEXT: v_mov_b32_e32 v6, v1
; GFX11-FAKE16-NEXT: s_waitcnt_vscnt null, 0x0
; GFX11-FAKE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
-; GFX11-FAKE16-NEXT: v_dual_add_f32 v5, v5, v3 :: v_dual_add_f32 v0, v0, v2
-; GFX11-FAKE16-NEXT: v_bfe_u32 v7, v5, 16, 1
+; GFX11-FAKE16-NEXT: v_and_b32_e32 v2, 0xffff0000, v6
+; GFX11-FAKE16-NEXT: v_dual_add_f32 v2, v2, v0 :: v_dual_lshlrev_b32 v1, 16, v6
+; GFX11-FAKE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_2)
+; GFX11-FAKE16-NEXT: v_add_f32_e32 v1, v1, v3
+; GFX11-FAKE16-NEXT: v_bfe_u32 v7, v2, 16, 1
; GFX11-FAKE16-NEXT: s_delay_alu instid0(VALU_DEP_2)
-; GFX11-FAKE16-NEXT: v_bfe_u32 v6, v0, 16, 1
-; GFX11-FAKE16-NEXT: v_or_b32_e32 v8, 0x400000, v0
-; GFX11-FAKE16-NEXT: v_or_b32_e32 v9, 0x400000, v5
-; GFX11-FAKE16-NEXT: v_cmp_u_f32_e32 vcc_lo, v5, v5
-; GFX11-FAKE16-NEXT: v_add3_u32 v7, v7, v5, 0x7fff
-; GFX11-FAKE16-NEXT: v_add3_u32 v6, v6, v0, 0x7fff
-; GFX11-FAKE16-NEXT: v_cmp_u_f32_e64 s4, v0, v0
+; GFX11-FAKE16-NEXT: v_bfe_u32 v5, v1, 16, 1
+; GFX11-FAKE16-NEXT: v_or_b32_e32 v8, 0x400000, v1
+; GFX11-FAKE16-NEXT: v_or_b32_e32 v9, 0x400000, v2
+; GFX11-FAKE16-NEXT: v_cmp_u_f32_e32 vcc_lo, v2, v2
+; GFX11-FAKE16-NEXT: v_add3_u32 v7, v7, v2, 0x7fff
+; GFX11-FAKE16-NEXT: v_add3_u32 v5, v5, v1, 0x7fff
+; GFX11-FAKE16-NEXT: v_cmp_u_f32_e64 s4, v1, v1
; GFX11-FAKE16-NEXT: s_delay_alu instid0(VALU_DEP_3) | instskip(NEXT) | instid1(VALU_DEP_2)
-; GFX11-FAKE16-NEXT: v_cndmask_b32_e32 v5, v7, v9, vcc_lo
-; GFX11-FAKE16-NEXT: v_cndmask_b32_e64 v0, v6, v8, s4
+; GFX11-FAKE16-NEXT: v_cndmask_b32_e32 v2, v7, v9, vcc_lo
+; GFX11-FAKE16-NEXT: v_cndmask_b32_e64 v1, v5, v8, s4
; GFX11-FAKE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
-; GFX11-FAKE16-NEXT: v_perm_b32 v0, v5, v0, 0x7060302
-; GFX11-FAKE16-NEXT: v_dual_mov_b32 v6, v1 :: v_dual_mov_b32 v5, v0
-; GFX11-FAKE16-NEXT: buffer_atomic_cmpswap_b32 v[5:6], v4, s[0:3], 0 offen glc
+; GFX11-FAKE16-NEXT: v_perm_b32 v5, v2, v1, 0x7060302
+; GFX11-FAKE16-NEXT: v_dual_mov_b32 v1, v5 :: v_dual_mov_b32 v2, v6
+; GFX11-FAKE16-NEXT: buffer_atomic_cmpswap_b32 v[1:2], v4, s[0:3], 0 offen glc
; GFX11-FAKE16-NEXT: s_waitcnt vmcnt(0)
; GFX11-FAKE16-NEXT: buffer_gl1_inv
; GFX11-FAKE16-NEXT: buffer_gl0_inv
-; GFX11-FAKE16-NEXT: v_cmp_eq_u32_e32 vcc_lo, v5, v1
-; GFX11-FAKE16-NEXT: v_mov_b32_e32 v1, v5
+; GFX11-FAKE16-NEXT: v_cmp_eq_u32_e32 vcc_lo, v1, v6
; GFX11-FAKE16-NEXT: s_or_b32 s5, vcc_lo, s5
; GFX11-FAKE16-NEXT: s_delay_alu instid0(SALU_CYCLE_1)
; GFX11-FAKE16-NEXT: s_and_not1_b32 exec_lo, exec_lo, s5
@@ -12521,38 +12540,38 @@ define void @buffer_fat_ptr_agent_atomic_fadd_noret_v2bf16__offset__amdgpu_no_fi
; GFX10-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX10-NEXT: v_mov_b32_e32 v1, s20
; GFX10-NEXT: s_add_i32 s4, s20, 0x400
-; GFX10-NEXT: v_lshlrev_b32_e32 v2, 16, v0
-; GFX10-NEXT: v_and_b32_e32 v3, 0xffff0000, v0
+; GFX10-NEXT: v_lshlrev_b32_e32 v3, 16, v0
+; GFX10-NEXT: v_and_b32_e32 v0, 0xffff0000, v0
; GFX10-NEXT: v_mov_b32_e32 v4, s4
; GFX10-NEXT: buffer_load_dword v1, v1, s[16:19], 0 offen offset:1024
; GFX10-NEXT: s_mov_b32 s5, 0
; GFX10-NEXT: .LBB33_1: ; %atomicrmw.start
; GFX10-NEXT: ; =>This Inner Loop Header: Depth=1
; GFX10-NEXT: s_waitcnt vmcnt(0)
-; GFX10-NEXT: v_lshlrev_b32_e32 v0, 16, v1
-; GFX10-NEXT: v_and_b32_e32 v5, 0xffff0000, v1
-; GFX10-NEXT: s_waitcnt_vscnt null, 0x0
-; GFX10-NEXT: v_add_f32_e32 v0, v0, v2
-; GFX10-NEXT: v_add_f32_e32 v5, v5, v3
-; GFX10-NEXT: v_bfe_u32 v6, v0, 16, 1
-; GFX10-NEXT: v_bfe_u32 v7, v5, 16, 1
-; GFX10-NEXT: v_or_b32_e32 v8, 0x400000, v0
-; GFX10-NEXT: v_or_b32_e32 v9, 0x400000, v5
-; GFX10-NEXT: v_cmp_u_f32_e32 vcc_lo, v5, v5
-; GFX10-NEXT: v_add3_u32 v6, v6, v0, 0x7fff
-; GFX10-NEXT: v_add3_u32 v7, v7, v5, 0x7fff
-; GFX10-NEXT: v_cmp_u_f32_e64 s4, v0, v0
-; GFX10-NEXT: v_cndmask_b32_e32 v5, v7, v9, vcc_lo
-; GFX10-NEXT: v_cndmask_b32_e64 v0, v6, v8, s4
-; GFX10-NEXT: v_perm_b32 v0, v5, v0, 0x7060302
; GFX10-NEXT: v_mov_b32_e32 v6, v1
-; GFX10-NEXT: v_mov_b32_e32 v5, v0
-; GFX10-NEXT: buffer_atomic_cmpswap v[5:6], v4, s[16:19], 0 offen glc
+; GFX10-NEXT: s_waitcnt_vscnt null, 0x0
+; GFX10-NEXT: v_lshlrev_b32_e32 v1, 16, v6
+; GFX10-NEXT: v_and_b32_e32 v2, 0xffff0000, v6
+; GFX10-NEXT: v_add_f32_e32 v1, v1, v3
+; GFX10-NEXT: v_add_f32_e32 v2, v2, v0
+; GFX10-NEXT: v_bfe_u32 v5, v1, 16, 1
+; GFX10-NEXT: v_bfe_u32 v7, v2, 16, 1
+; GFX10-NEXT: v_or_b32_e32 v8, 0x400000, v1
+; GFX10-NEXT: v_or_b32_e32 v9, 0x400000, v2
+; GFX10-NEXT: v_cmp_u_f32_e32 vcc_lo, v2, v2
+; GFX10-NEXT: v_add3_u32 v5, v5, v1, 0x7fff
+; GFX10-NEXT: v_add3_u32 v7, v7, v2, 0x7fff
+; GFX10-NEXT: v_cmp_u_f32_e64 s4, v1, v1
+; GFX10-NEXT: v_cndmask_b32_e32 v2, v7, v9, vcc_lo
+; GFX10-NEXT: v_cndmask_b32_e64 v1, v5, v8, s4
+; GFX10-NEXT: v_perm_b32 v5, v2, v1, 0x7060302
+; GFX10-NEXT: v_mov_b32_e32 v1, v5
+; GFX10-NEXT: v_mov_b32_e32 v2, v6
+; GFX10-NEXT: buffer_atomic_cmpswap v[1:2], v4, s[16:19], 0 offen glc
; GFX10-NEXT: s_waitcnt vmcnt(0)
; GFX10-NEXT: buffer_gl1_inv
; GFX10-NEXT: buffer_gl0_inv
-; GFX10-NEXT: v_cmp_eq_u32_e32 vcc_lo, v5, v1
-; GFX10-NEXT: v_mov_b32_e32 v1, v5
+; GFX10-NEXT: v_cmp_eq_u32_e32 vcc_lo, v1, v6
; GFX10-NEXT: s_or_b32 s5, vcc_lo, s5
; GFX10-NEXT: s_andn2_b32 exec_lo, exec_lo, s5
; GFX10-NEXT: s_cbranch_execnz .LBB33_1
@@ -12564,39 +12583,39 @@ define void @buffer_fat_ptr_agent_atomic_fadd_noret_v2bf16__offset__amdgpu_no_fi
; GFX90A: ; %bb.0:
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX90A-NEXT: v_mov_b32_e32 v1, s20
-; GFX90A-NEXT: buffer_load_dword v1, v1, s[16:19], 0 offen offset:1024
+; GFX90A-NEXT: buffer_load_dword v2, v1, s[16:19], 0 offen offset:1024
; GFX90A-NEXT: s_add_i32 s4, s20, 0x400
; GFX90A-NEXT: s_mov_b64 s[6:7], 0
-; GFX90A-NEXT: v_lshlrev_b32_e32 v2, 16, v0
+; GFX90A-NEXT: v_lshlrev_b32_e32 v1, 16, v0
; GFX90A-NEXT: s_movk_i32 s8, 0x7fff
-; GFX90A-NEXT: v_and_b32_e32 v3, 0xffff0000, v0
+; GFX90A-NEXT: v_and_b32_e32 v0, 0xffff0000, v0
; GFX90A-NEXT: s_mov_b32 s9, 0x7060302
; GFX90A-NEXT: v_mov_b32_e32 v4, s4
; GFX90A-NEXT: .LBB33_1: ; %atomicrmw.start
; GFX90A-NEXT: ; =>This Inner Loop Header: Depth=1
; GFX90A-NEXT: s_waitcnt vmcnt(0)
-; GFX90A-NEXT: v_lshlrev_b32_e32 v0, 16, v1
-; GFX90A-NEXT: v_and_b32_e32 v5, 0xffff0000, v1
-; GFX90A-NEXT: v_add_f32_e32 v0, v0, v2
-; GFX90A-NEXT: v_add_f32_e32 v5, v5, v3
-; GFX90A-NEXT: v_bfe_u32 v6, v0, 16, 1
-; GFX90A-NEXT: v_bfe_u32 v8, v5, 16, 1
-; GFX90A-NEXT: v_or_b32_e32 v7, 0x400000, v0
-; GFX90A-NEXT: v_or_b32_e32 v9, 0x400000, v5
-; GFX90A-NEXT: v_add3_u32 v6, v6, v0, s8
-; GFX90A-NEXT: v_add3_u32 v8, v8, v5, s8
-; GFX90A-NEXT: v_cmp_u_f32_e32 vcc, v5, v5
-; GFX90A-NEXT: v_cmp_u_f32_e64 s[4:5], v0, v0
-; GFX90A-NEXT: v_cndmask_b32_e64 v0, v6, v7, s[4:5]
-; GFX90A-NEXT: v_cndmask_b32_e32 v5, v8, v9, vcc
-; GFX90A-NEXT: v_perm_b32 v0, v5, v0, s9
-; GFX90A-NEXT: v_pk_mov_b32 v[6:7], v[0:1], v[0:1] op_sel:[0,1]
-; GFX90A-NEXT: buffer_atomic_cmpswap v[6:7], v4, s[16:19], 0 offen glc
+; GFX90A-NEXT: v_mov_b32_e32 v7, v2
+; GFX90A-NEXT: v_lshlrev_b32_e32 v2, 16, v7
+; GFX90A-NEXT: v_and_b32_e32 v3, 0xffff0000, v7
+; GFX90A-NEXT: v_add_f32_e32 v2, v2, v1
+; GFX90A-NEXT: v_add_f32_e32 v3, v3, v0
+; GFX90A-NEXT: v_bfe_u32 v5, v2, 16, 1
+; GFX90A-NEXT: v_bfe_u32 v8, v3, 16, 1
+; GFX90A-NEXT: v_or_b32_e32 v6, 0x400000, v2
+; GFX90A-NEXT: v_or_b32_e32 v9, 0x400000, v3
+; GFX90A-NEXT: v_add3_u32 v5, v5, v2, s8
+; GFX90A-NEXT: v_add3_u32 v8, v8, v3, s8
+; GFX90A-NEXT: v_cmp_u_f32_e32 vcc, v3, v3
+; GFX90A-NEXT: v_cmp_u_f32_e64 s[4:5], v2, v2
+; GFX90A-NEXT: v_cndmask_b32_e64 v2, v5, v6, s[4:5]
+; GFX90A-NEXT: v_cndmask_b32_e32 v3, v8, v9, vcc
+; GFX90A-NEXT: v_perm_b32 v6, v3, v2, s9
+; GFX90A-NEXT: v_pk_mov_b32 v[2:3], v[6:7], v[6:7] op_sel:[0,1]
+; GFX90A-NEXT: buffer_atomic_cmpswap v[2:3], v4, s[16:19], 0 offen glc
; GFX90A-NEXT: s_waitcnt vmcnt(0)
; GFX90A-NEXT: buffer_wbinvl1
-; GFX90A-NEXT: v_cmp_eq_u32_e32 vcc, v6, v1
+; GFX90A-NEXT: v_cmp_eq_u32_e32 vcc, v2, v7
; GFX90A-NEXT: s_or_b64 s[6:7], vcc, s[6:7]
-; GFX90A-NEXT: v_mov_b32_e32 v1, v6
; GFX90A-NEXT: s_andn2_b64 exec, exec, s[6:7]
; GFX90A-NEXT: s_cbranch_execnz .LBB33_1
; GFX90A-NEXT: ; %bb.2: ; %atomicrmw.end
@@ -12610,37 +12629,37 @@ define void @buffer_fat_ptr_agent_atomic_fadd_noret_v2bf16__offset__amdgpu_no_fi
; GFX908-NEXT: buffer_load_dword v1, v1, s[16:19], 0 offen offset:1024
; GFX908-NEXT: s_add_i32 s4, s20, 0x400
; GFX908-NEXT: s_mov_b64 s[6:7], 0
-; GFX908-NEXT: v_lshlrev_b32_e32 v2, 16, v0
+; GFX908-NEXT: v_lshlrev_b32_e32 v3, 16, v0
; GFX908-NEXT: s_movk_i32 s8, 0x7fff
-; GFX908-NEXT: v_and_b32_e32 v3, 0xffff0000, v0
+; GFX908-NEXT: v_and_b32_e32 v0, 0xffff0000, v0
; GFX908-NEXT: s_mov_b32 s9, 0x7060302
; GFX908-NEXT: v_mov_b32_e32 v4, s4
; GFX908-NEXT: .LBB33_1: ; %atomicrmw.start
; GFX908-NEXT: ; =>This Inner Loop Header: Depth=1
; GFX908-NEXT: s_waitcnt vmcnt(0)
-; GFX908-NEXT: v_lshlrev_b32_e32 v0, 16, v1
-; GFX908-NEXT: v_and_b32_e32 v5, 0xffff0000, v1
-; GFX908-NEXT: v_add_f32_e32 v0, v0, v2
-; GFX908-NEXT: v_add_f32_e32 v5, v5, v3
-; GFX908-NEXT: v_bfe_u32 v6, v0, 16, 1
-; GFX908-NEXT: v_bfe_u32 v8, v5, 16, 1
-; GFX908-NEXT: v_or_b32_e32 v7, 0x400000, v0
-; GFX908-NEXT: v_or_b32_e32 v9, 0x400000, v5
-; GFX908-NEXT: v_add3_u32 v6, v6, v0, s8
-; GFX908-NEXT: v_add3_u32 v8, v8, v5, s8
-; GFX908-NEXT: v_cmp_u_f32_e32 vcc, v5, v5
-; GFX908-NEXT: v_cmp_u_f32_e64 s[4:5], v0, v0
-; GFX908-NEXT: v_cndmask_b32_e64 v0, v6, v7, s[4:5]
-; GFX908-NEXT: v_cndmask_b32_e32 v5, v8, v9, vcc
-; GFX908-NEXT: v_perm_b32 v0, v5, v0, s9
; GFX908-NEXT: v_mov_b32_e32 v6, v1
-; GFX908-NEXT: v_mov_b32_e32 v5, v0
-; GFX908-NEXT: buffer_atomic_cmpswap v[5:6], v4, s[16:19], 0 offen glc
+; GFX908-NEXT: v_lshlrev_b32_e32 v1, 16, v6
+; GFX908-NEXT: v_and_b32_e32 v2, 0xffff0000, v6
+; GFX908-NEXT: v_add_f32_e32 v1, v1, v3
+; GFX908-NEXT: v_add_f32_e32 v2, v2, v0
+; GFX908-NEXT: v_bfe_u32 v5, v1, 16, 1
+; GFX908-NEXT: v_bfe_u32 v8, v2, 16, 1
+; GFX908-NEXT: v_or_b32_e32 v7, 0x400000, v1
+; GFX908-NEXT: v_or_b32_e32 v9, 0x400000, v2
+; GFX908-NEXT: v_add3_u32 v5, v5, v1, s8
+; GFX908-NEXT: v_add3_u32 v8, v8, v2, s8
+; GFX908-NEXT: v_cmp_u_f32_e32 vcc, v2, v2
+; GFX908-NEXT: v_cmp_u_f32_e64 s[4:5], v1, v1
+; GFX908-NEXT: v_cndmask_b32_e64 v1, v5, v7, s[4:5]
+; GFX908-NEXT: v_cndmask_b32_e32 v2, v8, v9, vcc
+; GFX908-NEXT: v_perm_b32 v5, v2, v1, s9
+; GFX908-NEXT: v_mov_b32_e32 v1, v5
+; GFX908-NEXT: v_mov_b32_e32 v2, v6
+; GFX908-NEXT: buffer_atomic_cmpswap v[1:2], v4, s[16:19], 0 offen glc
; GFX908-NEXT: s_waitcnt vmcnt(0)
; GFX908-NEXT: buffer_wbinvl1
-; GFX908-NEXT: v_cmp_eq_u32_e32 vcc, v5, v1
+; GFX908-NEXT: v_cmp_eq_u32_e32 vcc, v1, v6
; GFX908-NEXT: s_or_b64 s[6:7], vcc, s[6:7]
-; GFX908-NEXT: v_mov_b32_e32 v1, v5
; GFX908-NEXT: s_andn2_b64 exec, exec, s[6:7]
; GFX908-NEXT: s_cbranch_execnz .LBB33_1
; GFX908-NEXT: ; %bb.2: ; %atomicrmw.end
@@ -12654,38 +12673,38 @@ define void @buffer_fat_ptr_agent_atomic_fadd_noret_v2bf16__offset__amdgpu_no_fi
; GFX8-NEXT: buffer_load_dword v1, v1, s[16:19], 0 offen offset:1024
; GFX8-NEXT: s_add_i32 s4, s20, 0x400
; GFX8-NEXT: s_mov_b64 s[6:7], 0
-; GFX8-NEXT: v_lshlrev_b32_e32 v2, 16, v0
-; GFX8-NEXT: v_and_b32_e32 v3, 0xffff0000, v0
+; GFX8-NEXT: v_lshlrev_b32_e32 v3, 16, v0
+; GFX8-NEXT: v_and_b32_e32 v0, 0xffff0000, v0
; GFX8-NEXT: v_mov_b32_e32 v4, s4
; GFX8-NEXT: .LBB33_1: ; %atomicrmw.start
; GFX8-NEXT: ; =>This Inner Loop Header: Depth=1
; GFX8-NEXT: s_waitcnt vmcnt(0)
-; GFX8-NEXT: v_lshlrev_b32_e32 v0, 16, v1
-; GFX8-NEXT: v_and_b32_e32 v5, 0xffff0000, v1
-; GFX8-NEXT: v_add_f32_e32 v0, v0, v2
-; GFX8-NEXT: v_add_f32_e32 v5, v5, v3
-; GFX8-NEXT: v_bfe_u32 v6, v0, 16, 1
-; GFX8-NEXT: v_bfe_u32 v8, v5, 16, 1
-; GFX8-NEXT: v_add_u32_e32 v6, vcc, v6, v0
-; GFX8-NEXT: v_add_u32_e32 v8, vcc, v8, v5
-; GFX8-NEXT: v_add_u32_e32 v6, vcc, 0x7fff, v6
-; GFX8-NEXT: v_add_u32_e32 v8, vcc, 0x7fff, v8
-; GFX8-NEXT: v_or_b32_e32 v9, 0x400000, v5
-; GFX8-NEXT: v_cmp_u_f32_e32 vcc, v5, v5
-; GFX8-NEXT: v_or_b32_e32 v7, 0x400000, v0
-; GFX8-NEXT: v_cmp_u_f32_e64 s[4:5], v0, v0
-; GFX8-NEXT: v_cndmask_b32_e32 v5, v8, v9, vcc
-; GFX8-NEXT: v_cndmask_b32_e64 v0, v6, v7, s[4:5]
-; GFX8-NEXT: v_lshrrev_b32_e32 v5, 16, v5
-; GFX8-NEXT: v_alignbit_b32 v0, v5, v0, 16
; GFX8-NEXT: v_mov_b32_e32 v6, v1
-; GFX8-NEXT: v_mov_b32_e32 v5, v0
-; GFX8-NEXT: buffer_atomic_cmpswap v[5:6], v4, s[16:19], 0 offen glc
+; GFX8-NEXT: v_lshlrev_b32_e32 v1, 16, v6
+; GFX8-NEXT: v_and_b32_e32 v2, 0xffff0000, v6
+; GFX8-NEXT: v_add_f32_e32 v1, v1, v3
+; GFX8-NEXT: v_add_f32_e32 v2, v2, v0
+; GFX8-NEXT: v_bfe_u32 v5, v1, 16, 1
+; GFX8-NEXT: v_bfe_u32 v8, v2, 16, 1
+; GFX8-NEXT: v_add_u32_e32 v5, vcc, v5, v1
+; GFX8-NEXT: v_add_u32_e32 v8, vcc, v8, v2
+; GFX8-NEXT: v_add_u32_e32 v5, vcc, 0x7fff, v5
+; GFX8-NEXT: v_add_u32_e32 v8, vcc, 0x7fff, v8
+; GFX8-NEXT: v_or_b32_e32 v9, 0x400000, v2
+; GFX8-NEXT: v_cmp_u_f32_e32 vcc, v2, v2
+; GFX8-NEXT: v_or_b32_e32 v7, 0x400000, v1
+; GFX8-NEXT: v_cmp_u_f32_e64 s[4:5], v1, v1
+; GFX8-NEXT: v_cndmask_b32_e32 v2, v8, v9, vcc
+; GFX8-NEXT: v_cndmask_b32_e64 v1, v5, v7, s[4:5]
+; GFX8-NEXT: v_lshrrev_b32_e32 v2, 16, v2
+; GFX8-NEXT: v_alignbit_b32 v5, v2, v1, 16
+; GFX8-NEXT: v_mov_b32_e32 v1, v5
+; GFX8-NEXT: v_mov_b32_e32 v2, v6
+; GFX8-NEXT: buffer_atomic_cmpswap v[1:2], v4, s[16:19], 0 offen glc
; GFX8-NEXT: s_waitcnt vmcnt(0)
; GFX8-NEXT: buffer_wbinvl1
-; GFX8-NEXT: v_cmp_eq_u32_e32 vcc, v5, v1
+; GFX8-NEXT: v_cmp_eq_u32_e32 vcc, v1, v6
; GFX8-NEXT: s_or_b64 s[6:7], vcc, s[6:7]
-; GFX8-NEXT: v_mov_b32_e32 v1, v5
; GFX8-NEXT: s_andn2_b64 exec, exec, s[6:7]
; GFX8-NEXT: s_cbranch_execnz .LBB33_1
; GFX8-NEXT: ; %bb.2: ; %atomicrmw.end
diff --git a/llvm/test/CodeGen/AMDGPU/buffer-fat-pointer-atomicrmw-fmax.ll b/llvm/test/CodeGen/AMDGPU/buffer-fat-pointer-atomicrmw-fmax.ll
index c17225594164f..5150ee6ed1f1a 100644
--- a/llvm/test/CodeGen/AMDGPU/buffer-fat-pointer-atomicrmw-fmax.ll
+++ b/llvm/test/CodeGen/AMDGPU/buffer-fat-pointer-atomicrmw-fmax.ll
@@ -211,24 +211,24 @@ define void @buffer_fat_ptr_agent_atomic_fmax_noret_f32__offset__amdgpu_no_fine_
; GFX942: ; %bb.0:
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX942-NEXT: v_mov_b32_e32 v1, s16
-; GFX942-NEXT: buffer_load_dword v1, v1, s[0:3], 0 offen offset:1024
+; GFX942-NEXT: buffer_load_dword v2, v1, s[0:3], 0 offen offset:1024
; GFX942-NEXT: s_add_i32 s6, s16, 0x400
; GFX942-NEXT: s_mov_b64 s[4:5], 0
-; GFX942-NEXT: v_max_f32_e32 v2, v0, v0
-; GFX942-NEXT: v_mov_b32_e32 v3, s6
+; GFX942-NEXT: v_max_f32_e32 v0, v0, v0
+; GFX942-NEXT: v_mov_b32_e32 v1, s6
; GFX942-NEXT: .LBB1_1: ; %atomicrmw.start
; GFX942-NEXT: ; =>This Inner Loop Header: Depth=1
; GFX942-NEXT: s_waitcnt vmcnt(0)
-; GFX942-NEXT: v_max_f32_e32 v0, v1, v1
-; GFX942-NEXT: v_max_f32_e32 v0, v0, v2
-; GFX942-NEXT: v_mov_b64_e32 v[4:5], v[0:1]
+; GFX942-NEXT: v_mov_b32_e32 v5, v2
+; GFX942-NEXT: v_max_f32_e32 v2, v5, v5
+; GFX942-NEXT: v_max_f32_e32 v4, v2, v0
+; GFX942-NEXT: v_mov_b64_e32 v[2:3], v[4:5]
; GFX942-NEXT: buffer_wbl2 sc1
-; GFX942-NEXT: buffer_atomic_cmpswap v[4:5], v3, s[0:3], 0 offen sc0
+; GFX942-NEXT: buffer_atomic_cmpswap v[2:3], v1, s[0:3], 0 offen sc0
; GFX942-NEXT: s_waitcnt vmcnt(0)
; GFX942-NEXT: buffer_inv sc1
-; GFX942-NEXT: v_cmp_eq_u32_e32 vcc, v4, v1
+; GFX942-NEXT: v_cmp_eq_u32_e32 vcc, v2, v5
; GFX942-NEXT: s_or_b64 s[4:5], vcc, s[4:5]
-; GFX942-NEXT: v_mov_b32_e32 v1, v4
; GFX942-NEXT: s_andn2_b64 exec, exec, s[4:5]
; GFX942-NEXT: s_cbranch_execnz .LBB1_1
; GFX942-NEXT: ; %bb.2: ; %atomicrmw.end
@@ -261,23 +261,23 @@ define void @buffer_fat_ptr_agent_atomic_fmax_noret_f32__offset__amdgpu_no_fine_
; GFX90A: ; %bb.0:
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX90A-NEXT: v_mov_b32_e32 v1, s20
-; GFX90A-NEXT: buffer_load_dword v1, v1, s[16:19], 0 offen offset:1024
+; GFX90A-NEXT: buffer_load_dword v2, v1, s[16:19], 0 offen offset:1024
; GFX90A-NEXT: s_add_i32 s6, s20, 0x400
; GFX90A-NEXT: s_mov_b64 s[4:5], 0
-; GFX90A-NEXT: v_max_f32_e32 v2, v0, v0
-; GFX90A-NEXT: v_mov_b32_e32 v3, s6
+; GFX90A-NEXT: v_max_f32_e32 v0, v0, v0
+; GFX90A-NEXT: v_mov_b32_e32 v1, s6
; GFX90A-NEXT: .LBB1_1: ; %atomicrmw.start
; GFX90A-NEXT: ; =>This Inner Loop Header: Depth=1
; GFX90A-NEXT: s_waitcnt vmcnt(0)
-; GFX90A-NEXT: v_max_f32_e32 v0, v1, v1
-; GFX90A-NEXT: v_max_f32_e32 v0, v0, v2
-; GFX90A-NEXT: v_pk_mov_b32 v[4:5], v[0:1], v[0:1] op_sel:[0,1]
-; GFX90A-NEXT: buffer_atomic_cmpswap v[4:5], v3, s[16:19], 0 offen glc
+; GFX90A-NEXT: v_mov_b32_e32 v5, v2
+; GFX90A-NEXT: v_max_f32_e32 v2, v5, v5
+; GFX90A-NEXT: v_max_f32_e32 v4, v2, v0
+; GFX90A-NEXT: v_pk_mov_b32 v[2:3], v[4:5], v[4:5] op_sel:[0,1]
+; GFX90A-NEXT: buffer_atomic_cmpswap v[2:3], v1, s[16:19], 0 offen glc
; GFX90A-NEXT: s_waitcnt vmcnt(0)
; GFX90A-NEXT: buffer_wbinvl1
-; GFX90A-NEXT: v_cmp_eq_u32_e32 vcc, v4, v1
+; GFX90A-NEXT: v_cmp_eq_u32_e32 vcc, v2, v5
; GFX90A-NEXT: s_or_b64 s[4:5], vcc, s[4:5]
-; GFX90A-NEXT: v_mov_b32_e32 v1, v4
; GFX90A-NEXT: s_andn2_b64 exec, exec, s[4:5]
; GFX90A-NEXT: s_cbranch_execnz .LBB1_1
; GFX90A-NEXT: ; %bb.2: ; %atomicrmw.end
@@ -291,21 +291,21 @@ define void @buffer_fat_ptr_agent_atomic_fmax_noret_f32__offset__amdgpu_no_fine_
; GFX908-NEXT: buffer_load_dword v1, v1, s[16:19], 0 offen offset:1024
; GFX908-NEXT: s_add_i32 s6, s20, 0x400
; GFX908-NEXT: s_mov_b64 s[4:5], 0
-; GFX908-NEXT: v_max_f32_e32 v2, v0, v0
+; GFX908-NEXT: v_max_f32_e32 v0, v0, v0
; GFX908-NEXT: v_mov_b32_e32 v3, s6
; GFX908-NEXT: .LBB1_1: ; %atomicrmw.start
; GFX908-NEXT: ; =>This Inner Loop Header: Depth=1
; GFX908-NEXT: s_waitcnt vmcnt(0)
-; GFX908-NEXT: v_max_f32_e32 v0, v1, v1
-; GFX908-NEXT: v_max_f32_e32 v0, v0, v2
; GFX908-NEXT: v_mov_b32_e32 v5, v1
-; GFX908-NEXT: v_mov_b32_e32 v4, v0
-; GFX908-NEXT: buffer_atomic_cmpswap v[4:5], v3, s[16:19], 0 offen glc
+; GFX908-NEXT: v_max_f32_e32 v1, v5, v5
+; GFX908-NEXT: v_max_f32_e32 v4, v1, v0
+; GFX908-NEXT: v_mov_b32_e32 v1, v4
+; GFX908-NEXT: v_mov_b32_e32 v2, v5
+; GFX908-NEXT: buffer_atomic_cmpswap v[1:2], v3, s[16:19], 0 offen glc
; GFX908-NEXT: s_waitcnt vmcnt(0)
; GFX908-NEXT: buffer_wbinvl1
-; GFX908-NEXT: v_cmp_eq_u32_e32 vcc, v4, v1
+; GFX908-NEXT: v_cmp_eq_u32_e32 vcc, v1, v5
; GFX908-NEXT: s_or_b64 s[4:5], vcc, s[4:5]
-; GFX908-NEXT: v_mov_b32_e32 v1, v4
; GFX908-NEXT: s_andn2_b64 exec, exec, s[4:5]
; GFX908-NEXT: s_cbranch_execnz .LBB1_1
; GFX908-NEXT: ; %bb.2: ; %atomicrmw.end
@@ -319,21 +319,21 @@ define void @buffer_fat_ptr_agent_atomic_fmax_noret_f32__offset__amdgpu_no_fine_
; GFX8-NEXT: buffer_load_dword v1, v1, s[16:19], 0 offen offset:1024
; GFX8-NEXT: s_add_i32 s6, s20, 0x400
; GFX8-NEXT: s_mov_b64 s[4:5], 0
-; GFX8-NEXT: v_mul_f32_e32 v2, 1.0, v0
+; GFX8-NEXT: v_mul_f32_e32 v0, 1.0, v0
; GFX8-NEXT: v_mov_b32_e32 v3, s6
; GFX8-NEXT: .LBB1_1: ; %atomicrmw.start
; GFX8-NEXT: ; =>This Inner Loop Header: Depth=1
; GFX8-NEXT: s_waitcnt vmcnt(0)
-; GFX8-NEXT: v_mul_f32_e32 v0, 1.0, v1
-; GFX8-NEXT: v_max_f32_e32 v0, v0, v2
; GFX8-NEXT: v_mov_b32_e32 v5, v1
-; GFX8-NEXT: v_mov_b32_e32 v4, v0
-; GFX8-NEXT: buffer_atomic_cmpswap v[4:5], v3, s[16:19], 0 offen glc
+; GFX8-NEXT: v_mul_f32_e32 v1, 1.0, v5
+; GFX8-NEXT: v_max_f32_e32 v4, v1, v0
+; GFX8-NEXT: v_mov_b32_e32 v1, v4
+; GFX8-NEXT: v_mov_b32_e32 v2, v5
+; GFX8-NEXT: buffer_atomic_cmpswap v[1:2], v3, s[16:19], 0 offen glc
; GFX8-NEXT: s_waitcnt vmcnt(0)
; GFX8-NEXT: buffer_wbinvl1
-; GFX8-NEXT: v_cmp_eq_u32_e32 vcc, v4, v1
+; GFX8-NEXT: v_cmp_eq_u32_e32 vcc, v1, v5
; GFX8-NEXT: s_or_b64 s[4:5], vcc, s[4:5]
-; GFX8-NEXT: v_mov_b32_e32 v1, v4
; GFX8-NEXT: s_andn2_b64 exec, exec, s[4:5]
; GFX8-NEXT: s_cbranch_execnz .LBB1_1
; GFX8-NEXT: ; %bb.2: ; %atomicrmw.end
@@ -1395,7 +1395,7 @@ define void @buffer_fat_ptr_agent_atomic_fmax_noret_f64__offset__amdgpu_no_fine_
; GFX12-NEXT: s_wait_bvhcnt 0x0
; GFX12-NEXT: s_wait_kmcnt 0x0
; GFX12-NEXT: v_mov_b32_e32 v2, s16
-; GFX12-NEXT: v_max_num_f64_e32 v[4:5], v[0:1], v[0:1]
+; GFX12-NEXT: v_max_num_f64_e32 v[0:1], v[0:1], v[0:1]
; GFX12-NEXT: s_add_co_i32 s4, s16, 0x800
; GFX12-NEXT: s_wait_alu 0xfffe
; GFX12-NEXT: v_mov_b32_e32 v6, s4
@@ -1404,17 +1404,18 @@ define void @buffer_fat_ptr_agent_atomic_fmax_noret_f64__offset__amdgpu_no_fine_
; GFX12-NEXT: .LBB6_1: ; %atomicrmw.start
; GFX12-NEXT: ; =>This Inner Loop Header: Depth=1
; GFX12-NEXT: s_wait_loadcnt 0x0
-; GFX12-NEXT: v_max_num_f64_e32 v[0:1], v[2:3], v[2:3]
-; GFX12-NEXT: s_wait_storecnt 0x0
-; GFX12-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(SKIP_1) | instid1(VALU_DEP_2)
-; GFX12-NEXT: v_max_num_f64_e32 v[0:1], v[0:1], v[4:5]
; GFX12-NEXT: v_dual_mov_b32 v10, v3 :: v_dual_mov_b32 v9, v2
-; GFX12-NEXT: v_dual_mov_b32 v8, v1 :: v_dual_mov_b32 v7, v0
-; GFX12-NEXT: buffer_atomic_cmpswap_b64 v[7:10], v6, s[0:3], null offen th:TH_ATOMIC_RETURN
+; GFX12-NEXT: s_wait_storecnt 0x0
+; GFX12-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
+; GFX12-NEXT: v_max_num_f64_e32 v[2:3], v[9:10], v[9:10]
+; GFX12-NEXT: v_max_num_f64_e32 v[7:8], v[2:3], v[0:1]
+; GFX12-NEXT: s_delay_alu instid0(VALU_DEP_1)
+; GFX12-NEXT: v_dual_mov_b32 v2, v7 :: v_dual_mov_b32 v3, v8
+; GFX12-NEXT: v_dual_mov_b32 v4, v9 :: v_dual_mov_b32 v5, v10
+; GFX12-NEXT: buffer_atomic_cmpswap_b64 v[2:5], v6, s[0:3], null offen th:TH_ATOMIC_RETURN
; GFX12-NEXT: s_wait_loadcnt 0x0
; GFX12-NEXT: global_inv scope:SCOPE_DEV
-; GFX12-NEXT: v_cmp_eq_u64_e32 vcc_lo, v[7:8], v[2:3]
-; GFX12-NEXT: v_dual_mov_b32 v2, v7 :: v_dual_mov_b32 v3, v8
+; GFX12-NEXT: v_cmp_eq_u64_e32 vcc_lo, v[2:3], v[9:10]
; GFX12-NEXT: s_or_b32 s4, vcc_lo, s4
; GFX12-NEXT: s_wait_alu 0xfffe
; GFX12-NEXT: s_and_not1_b32 exec_lo, exec_lo, s4
@@ -1438,7 +1439,7 @@ define void @buffer_fat_ptr_agent_atomic_fmax_noret_f64__offset__amdgpu_no_fine_
; GFX11: ; %bb.0:
; GFX11-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX11-NEXT: v_mov_b32_e32 v2, s16
-; GFX11-NEXT: v_max_f64 v[4:5], v[0:1], v[0:1]
+; GFX11-NEXT: v_max_f64 v[0:1], v[0:1], v[0:1]
; GFX11-NEXT: s_add_i32 s4, s16, 0x800
; GFX11-NEXT: s_delay_alu instid0(SALU_CYCLE_1)
; GFX11-NEXT: v_mov_b32_e32 v6, s4
@@ -1447,18 +1448,19 @@ define void @buffer_fat_ptr_agent_atomic_fmax_noret_f64__offset__amdgpu_no_fine_
; GFX11-NEXT: .LBB6_1: ; %atomicrmw.start
; GFX11-NEXT: ; =>This Inner Loop Header: Depth=1
; GFX11-NEXT: s_waitcnt vmcnt(0)
-; GFX11-NEXT: v_max_f64 v[0:1], v[2:3], v[2:3]
-; GFX11-NEXT: s_waitcnt_vscnt null, 0x0
-; GFX11-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(SKIP_1) | instid1(VALU_DEP_2)
-; GFX11-NEXT: v_max_f64 v[0:1], v[0:1], v[4:5]
; GFX11-NEXT: v_dual_mov_b32 v10, v3 :: v_dual_mov_b32 v9, v2
-; GFX11-NEXT: v_dual_mov_b32 v8, v1 :: v_dual_mov_b32 v7, v0
-; GFX11-NEXT: buffer_atomic_cmpswap_b64 v[7:10], v6, s[0:3], 0 offen glc
+; GFX11-NEXT: s_waitcnt_vscnt null, 0x0
+; GFX11-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
+; GFX11-NEXT: v_max_f64 v[2:3], v[9:10], v[9:10]
+; GFX11-NEXT: v_max_f64 v[7:8], v[2:3], v[0:1]
+; GFX11-NEXT: s_delay_alu instid0(VALU_DEP_1)
+; GFX11-NEXT: v_dual_mov_b32 v2, v7 :: v_dual_mov_b32 v3, v8
+; GFX11-NEXT: v_dual_mov_b32 v4, v9 :: v_dual_mov_b32 v5, v10
+; GFX11-NEXT: buffer_atomic_cmpswap_b64 v[2:5], v6, s[0:3], 0 offen glc
; GFX11-NEXT: s_waitcnt vmcnt(0)
; GFX11-NEXT: buffer_gl1_inv
; GFX11-NEXT: buffer_gl0_inv
-; GFX11-NEXT: v_cmp_eq_u64_e32 vcc_lo, v[7:8], v[2:3]
-; GFX11-NEXT: v_dual_mov_b32 v2, v7 :: v_dual_mov_b32 v3, v8
+; GFX11-NEXT: v_cmp_eq_u64_e32 vcc_lo, v[2:3], v[9:10]
; GFX11-NEXT: s_or_b32 s4, vcc_lo, s4
; GFX11-NEXT: s_delay_alu instid0(SALU_CYCLE_1)
; GFX11-NEXT: s_and_not1_b32 exec_lo, exec_lo, s4
@@ -1492,26 +1494,26 @@ define void @buffer_fat_ptr_agent_atomic_fmax_noret_f64__offset__amdgpu_no_fine_
; GFX908-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX908-NEXT: v_mov_b32_e32 v2, s20
; GFX908-NEXT: buffer_load_dwordx2 v[2:3], v2, s[16:19], 0 offen offset:2048
-; GFX908-NEXT: v_max_f64 v[4:5], v[0:1], v[0:1]
+; GFX908-NEXT: v_max_f64 v[0:1], v[0:1], v[0:1]
; GFX908-NEXT: s_add_i32 s6, s20, 0x800
; GFX908-NEXT: s_mov_b64 s[4:5], 0
; GFX908-NEXT: v_mov_b32_e32 v6, s6
; GFX908-NEXT: .LBB6_1: ; %atomicrmw.start
; GFX908-NEXT: ; =>This Inner Loop Header: Depth=1
; GFX908-NEXT: s_waitcnt vmcnt(0)
-; GFX908-NEXT: v_max_f64 v[0:1], v[2:3], v[2:3]
-; GFX908-NEXT: v_max_f64 v[0:1], v[0:1], v[4:5]
; GFX908-NEXT: v_mov_b32_e32 v10, v3
; GFX908-NEXT: v_mov_b32_e32 v9, v2
-; GFX908-NEXT: v_mov_b32_e32 v8, v1
-; GFX908-NEXT: v_mov_b32_e32 v7, v0
-; GFX908-NEXT: buffer_atomic_cmpswap_x2 v[7:10], v6, s[16:19], 0 offen glc
+; GFX908-NEXT: v_max_f64 v[2:3], v[9:10], v[9:10]
+; GFX908-NEXT: v_max_f64 v[7:8], v[2:3], v[0:1]
+; GFX908-NEXT: v_mov_b32_e32 v2, v7
+; GFX908-NEXT: v_mov_b32_e32 v3, v8
+; GFX908-NEXT: v_mov_b32_e32 v4, v9
+; GFX908-NEXT: v_mov_b32_e32 v5, v10
+; GFX908-NEXT: buffer_atomic_cmpswap_x2 v[2:5], v6, s[16:19], 0 offen glc
; GFX908-NEXT: s_waitcnt vmcnt(0)
; GFX908-NEXT: buffer_wbinvl1
-; GFX908-NEXT: v_cmp_eq_u64_e32 vcc, v[7:8], v[2:3]
-; GFX908-NEXT: v_mov_b32_e32 v2, v7
+; GFX908-NEXT: v_cmp_eq_u64_e32 vcc, v[2:3], v[9:10]
; GFX908-NEXT: s_or_b64 s[4:5], vcc, s[4:5]
-; GFX908-NEXT: v_mov_b32_e32 v3, v8
; GFX908-NEXT: s_andn2_b64 exec, exec, s[4:5]
; GFX908-NEXT: s_cbranch_execnz .LBB6_1
; GFX908-NEXT: ; %bb.2: ; %atomicrmw.end
@@ -1523,26 +1525,26 @@ define void @buffer_fat_ptr_agent_atomic_fmax_noret_f64__offset__amdgpu_no_fine_
; GFX8-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX8-NEXT: v_mov_b32_e32 v2, s20
; GFX8-NEXT: buffer_load_dwordx2 v[2:3], v2, s[16:19], 0 offen offset:2048
-; GFX8-NEXT: v_max_f64 v[4:5], v[0:1], v[0:1]
+; GFX8-NEXT: v_max_f64 v[0:1], v[0:1], v[0:1]
; GFX8-NEXT: s_add_i32 s6, s20, 0x800
; GFX8-NEXT: s_mov_b64 s[4:5], 0
; GFX8-NEXT: v_mov_b32_e32 v6, s6
; GFX8-NEXT: .LBB6_1: ; %atomicrmw.start
; GFX8-NEXT: ; =>This Inner Loop Header: Depth=1
; GFX8-NEXT: s_waitcnt vmcnt(0)
-; GFX8-NEXT: v_max_f64 v[0:1], v[2:3], v[2:3]
-; GFX8-NEXT: v_max_f64 v[0:1], v[0:1], v[4:5]
; GFX8-NEXT: v_mov_b32_e32 v10, v3
; GFX8-NEXT: v_mov_b32_e32 v9, v2
-; GFX8-NEXT: v_mov_b32_e32 v8, v1
-; GFX8-NEXT: v_mov_b32_e32 v7, v0
-; GFX8-NEXT: buffer_atomic_cmpswap_x2 v[7:10], v6, s[16:19], 0 offen glc
+; GFX8-NEXT: v_max_f64 v[2:3], v[9:10], v[9:10]
+; GFX8-NEXT: v_max_f64 v[7:8], v[2:3], v[0:1]
+; GFX8-NEXT: v_mov_b32_e32 v2, v7
+; GFX8-NEXT: v_mov_b32_e32 v3, v8
+; GFX8-NEXT: v_mov_b32_e32 v4, v9
+; GFX8-NEXT: v_mov_b32_e32 v5, v10
+; GFX8-NEXT: buffer_atomic_cmpswap_x2 v[2:5], v6, s[16:19], 0 offen glc
; GFX8-NEXT: s_waitcnt vmcnt(0)
; GFX8-NEXT: buffer_wbinvl1
-; GFX8-NEXT: v_cmp_eq_u64_e32 vcc, v[7:8], v[2:3]
-; GFX8-NEXT: v_mov_b32_e32 v2, v7
+; GFX8-NEXT: v_cmp_eq_u64_e32 vcc, v[2:3], v[9:10]
; GFX8-NEXT: s_or_b64 s[4:5], vcc, s[4:5]
-; GFX8-NEXT: v_mov_b32_e32 v3, v8
; GFX8-NEXT: s_andn2_b64 exec, exec, s[4:5]
; GFX8-NEXT: s_cbranch_execnz .LBB6_1
; GFX8-NEXT: ; %bb.2: ; %atomicrmw.end
@@ -2496,42 +2498,43 @@ define half @buffer_fat_ptr_agent_atomic_fmax_ret_f16__offset__amdgpu_no_fine_gr
; GFX12-TRUE16-NEXT: s_wait_alu 0xfffe
; GFX12-TRUE16-NEXT: s_and_b32 s4, s16, -4
; GFX12-TRUE16-NEXT: s_wait_alu 0xfffe
-; GFX12-TRUE16-NEXT: v_mov_b32_e32 v5, s4
+; GFX12-TRUE16-NEXT: v_mov_b32_e32 v3, s4
; GFX12-TRUE16-NEXT: s_and_b32 s4, s16, 3
; GFX12-TRUE16-NEXT: s_wait_alu 0xfffe
; GFX12-TRUE16-NEXT: s_lshl_b32 s4, s4, 3
; GFX12-TRUE16-NEXT: s_wait_alu 0xfffe
; GFX12-TRUE16-NEXT: s_lshl_b32 s5, 0xffff, s4
-; GFX12-TRUE16-NEXT: buffer_load_b32 v2, v5, s[0:3], null offen
+; GFX12-TRUE16-NEXT: buffer_load_b32 v1, v3, s[0:3], null offen
; GFX12-TRUE16-NEXT: s_not_b32 s6, s5
; GFX12-TRUE16-NEXT: s_mov_b32 s5, 0
; GFX12-TRUE16-NEXT: .LBB10_1: ; %atomicrmw.start
; GFX12-TRUE16-NEXT: ; =>This Inner Loop Header: Depth=1
; GFX12-TRUE16-NEXT: s_wait_loadcnt 0x0
-; GFX12-TRUE16-NEXT: v_lshrrev_b32_e32 v1, s4, v2
+; GFX12-TRUE16-NEXT: v_mov_b32_e32 v5, v1
; GFX12-TRUE16-NEXT: s_wait_storecnt 0x0
; GFX12-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
+; GFX12-TRUE16-NEXT: v_lshrrev_b32_e32 v1, s4, v5
; GFX12-TRUE16-NEXT: v_max_num_f16_e32 v0.h, v1.l, v1.l
-; GFX12-TRUE16-NEXT: v_max_num_f16_e32 v1.l, v0.h, v0.l
; GFX12-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
+; GFX12-TRUE16-NEXT: v_max_num_f16_e32 v1.l, v0.h, v0.l
; GFX12-TRUE16-NEXT: v_and_b32_e32 v1, 0xffff, v1
+; GFX12-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(SKIP_1) | instid1(VALU_DEP_1)
; GFX12-TRUE16-NEXT: v_lshlrev_b32_e32 v1, s4, v1
; GFX12-TRUE16-NEXT: s_wait_alu 0xfffe
-; GFX12-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
-; GFX12-TRUE16-NEXT: v_and_or_b32 v1, v2, s6, v1
-; GFX12-TRUE16-NEXT: v_dual_mov_b32 v4, v2 :: v_dual_mov_b32 v3, v1
-; GFX12-TRUE16-NEXT: buffer_atomic_cmpswap_b32 v[3:4], v5, s[0:3], null offen th:TH_ATOMIC_RETURN
+; GFX12-TRUE16-NEXT: v_and_or_b32 v4, v5, s6, v1
+; GFX12-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1)
+; GFX12-TRUE16-NEXT: v_dual_mov_b32 v1, v4 :: v_dual_mov_b32 v2, v5
+; GFX12-TRUE16-NEXT: buffer_atomic_cmpswap_b32 v[1:2], v3, s[0:3], null offen th:TH_ATOMIC_RETURN
; GFX12-TRUE16-NEXT: s_wait_loadcnt 0x0
; GFX12-TRUE16-NEXT: global_inv scope:SCOPE_DEV
-; GFX12-TRUE16-NEXT: v_cmp_eq_u32_e32 vcc_lo, v3, v2
-; GFX12-TRUE16-NEXT: v_mov_b32_e32 v2, v3
+; GFX12-TRUE16-NEXT: v_cmp_eq_u32_e32 vcc_lo, v1, v5
; GFX12-TRUE16-NEXT: s_or_b32 s5, vcc_lo, s5
; GFX12-TRUE16-NEXT: s_wait_alu 0xfffe
; GFX12-TRUE16-NEXT: s_and_not1_b32 exec_lo, exec_lo, s5
; GFX12-TRUE16-NEXT: s_cbranch_execnz .LBB10_1
; GFX12-TRUE16-NEXT: ; %bb.2: ; %atomicrmw.end
; GFX12-TRUE16-NEXT: s_or_b32 exec_lo, exec_lo, s5
-; GFX12-TRUE16-NEXT: v_lshrrev_b32_e32 v0, s4, v3
+; GFX12-TRUE16-NEXT: v_lshrrev_b32_e32 v0, s4, v1
; GFX12-TRUE16-NEXT: s_wait_loadcnt 0x0
; GFX12-TRUE16-NEXT: s_setpc_b64 s[30:31]
;
@@ -2543,46 +2546,47 @@ define half @buffer_fat_ptr_agent_atomic_fmax_ret_f16__offset__amdgpu_no_fine_gr
; GFX12-FAKE16-NEXT: s_wait_bvhcnt 0x0
; GFX12-FAKE16-NEXT: s_wait_kmcnt 0x0
; GFX12-FAKE16-NEXT: s_addk_co_i32 s16, 0x200
-; GFX12-FAKE16-NEXT: v_max_num_f16_e32 v5, v0, v0
+; GFX12-FAKE16-NEXT: v_max_num_f16_e32 v0, v0, v0
; GFX12-FAKE16-NEXT: s_wait_alu 0xfffe
; GFX12-FAKE16-NEXT: s_and_b32 s4, s16, -4
; GFX12-FAKE16-NEXT: s_wait_alu 0xfffe
-; GFX12-FAKE16-NEXT: v_mov_b32_e32 v4, s4
+; GFX12-FAKE16-NEXT: v_mov_b32_e32 v3, s4
; GFX12-FAKE16-NEXT: s_and_b32 s4, s16, 3
; GFX12-FAKE16-NEXT: s_wait_alu 0xfffe
; GFX12-FAKE16-NEXT: s_lshl_b32 s4, s4, 3
; GFX12-FAKE16-NEXT: s_wait_alu 0xfffe
; GFX12-FAKE16-NEXT: s_lshl_b32 s5, 0xffff, s4
-; GFX12-FAKE16-NEXT: buffer_load_b32 v1, v4, s[0:3], null offen
+; GFX12-FAKE16-NEXT: buffer_load_b32 v1, v3, s[0:3], null offen
; GFX12-FAKE16-NEXT: s_not_b32 s6, s5
; GFX12-FAKE16-NEXT: s_mov_b32 s5, 0
; GFX12-FAKE16-NEXT: .LBB10_1: ; %atomicrmw.start
; GFX12-FAKE16-NEXT: ; =>This Inner Loop Header: Depth=1
; GFX12-FAKE16-NEXT: s_wait_loadcnt 0x0
-; GFX12-FAKE16-NEXT: v_lshrrev_b32_e32 v0, s4, v1
+; GFX12-FAKE16-NEXT: v_mov_b32_e32 v5, v1
; GFX12-FAKE16-NEXT: s_wait_storecnt 0x0
; GFX12-FAKE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
-; GFX12-FAKE16-NEXT: v_max_num_f16_e32 v0, v0, v0
-; GFX12-FAKE16-NEXT: v_max_num_f16_e32 v0, v0, v5
+; GFX12-FAKE16-NEXT: v_lshrrev_b32_e32 v1, s4, v5
+; GFX12-FAKE16-NEXT: v_max_num_f16_e32 v1, v1, v1
; GFX12-FAKE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
-; GFX12-FAKE16-NEXT: v_and_b32_e32 v0, 0xffff, v0
-; GFX12-FAKE16-NEXT: v_lshlrev_b32_e32 v0, s4, v0
+; GFX12-FAKE16-NEXT: v_max_num_f16_e32 v1, v1, v0
+; GFX12-FAKE16-NEXT: v_and_b32_e32 v1, 0xffff, v1
+; GFX12-FAKE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(SKIP_1) | instid1(VALU_DEP_1)
+; GFX12-FAKE16-NEXT: v_lshlrev_b32_e32 v1, s4, v1
; GFX12-FAKE16-NEXT: s_wait_alu 0xfffe
-; GFX12-FAKE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
-; GFX12-FAKE16-NEXT: v_and_or_b32 v0, v1, s6, v0
-; GFX12-FAKE16-NEXT: v_dual_mov_b32 v3, v1 :: v_dual_mov_b32 v2, v0
-; GFX12-FAKE16-NEXT: buffer_atomic_cmpswap_b32 v[2:3], v4, s[0:3], null offen th:TH_ATOMIC_RETURN
+; GFX12-FAKE16-NEXT: v_and_or_b32 v4, v5, s6, v1
+; GFX12-FAKE16-NEXT: s_delay_alu instid0(VALU_DEP_1)
+; GFX12-FAKE16-NEXT: v_dual_mov_b32 v1, v4 :: v_dual_mov_b32 v2, v5
+; GFX12-FAKE16-NEXT: buffer_atomic_cmpswap_b32 v[1:2], v3, s[0:3], null offen th:TH_ATOMIC_RETURN
; GFX12-FAKE16-NEXT: s_wait_loadcnt 0x0
; GFX12-FAKE16-NEXT: global_inv scope:SCOPE_DEV
-; GFX12-FAKE16-NEXT: v_cmp_eq_u32_e32 vcc_lo, v2, v1
-; GFX12-FAKE16-NEXT: v_mov_b32_e32 v1, v2
+; GFX12-FAKE16-NEXT: v_cmp_eq_u32_e32 vcc_lo, v1, v5
; GFX12-FAKE16-NEXT: s_or_b32 s5, vcc_lo, s5
; GFX12-FAKE16-NEXT: s_wait_alu 0xfffe
; GFX12-FAKE16-NEXT: s_and_not1_b32 exec_lo, exec_lo, s5
; GFX12-FAKE16-NEXT: s_cbranch_execnz .LBB10_1
; GFX12-FAKE16-NEXT: ; %bb.2: ; %atomicrmw.end
; GFX12-FAKE16-NEXT: s_or_b32 exec_lo, exec_lo, s5
-; GFX12-FAKE16-NEXT: v_lshrrev_b32_e32 v0, s4, v2
+; GFX12-FAKE16-NEXT: v_lshrrev_b32_e32 v0, s4, v1
; GFX12-FAKE16-NEXT: s_wait_loadcnt 0x0
; GFX12-FAKE16-NEXT: s_setpc_b64 s[30:31]
;
@@ -2591,30 +2595,30 @@ define half @buffer_fat_ptr_agent_atomic_fmax_ret_f16__offset__amdgpu_no_fine_gr
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX942-NEXT: s_addk_i32 s16, 0x200
; GFX942-NEXT: s_and_b32 s4, s16, -4
-; GFX942-NEXT: v_mov_b32_e32 v4, s4
-; GFX942-NEXT: buffer_load_dword v1, v4, s[0:3], 0 offen
+; GFX942-NEXT: v_mov_b32_e32 v1, s4
+; GFX942-NEXT: buffer_load_dword v2, v1, s[0:3], 0 offen
; GFX942-NEXT: s_and_b32 s4, s16, 3
; GFX942-NEXT: s_lshl_b32 s6, s4, 3
; GFX942-NEXT: s_lshl_b32 s4, 0xffff, s6
; GFX942-NEXT: s_not_b32 s7, s4
; GFX942-NEXT: s_mov_b64 s[4:5], 0
-; GFX942-NEXT: v_max_f16_e32 v5, v0, v0
+; GFX942-NEXT: v_max_f16_e32 v0, v0, v0
; GFX942-NEXT: .LBB10_1: ; %atomicrmw.start
; GFX942-NEXT: ; =>This Inner Loop Header: Depth=1
; GFX942-NEXT: s_waitcnt vmcnt(0)
-; GFX942-NEXT: v_lshrrev_b32_e32 v0, s6, v1
-; GFX942-NEXT: v_max_f16_e32 v0, v0, v0
-; GFX942-NEXT: v_max_f16_e32 v0, v0, v5
-; GFX942-NEXT: v_lshlrev_b32_e32 v0, s6, v0
-; GFX942-NEXT: v_and_or_b32 v0, v1, s7, v0
-; GFX942-NEXT: v_mov_b64_e32 v[2:3], v[0:1]
+; GFX942-NEXT: v_mov_b32_e32 v5, v2
+; GFX942-NEXT: v_lshrrev_b32_e32 v2, s6, v5
+; GFX942-NEXT: v_max_f16_e32 v2, v2, v2
+; GFX942-NEXT: v_max_f16_e32 v2, v2, v0
+; GFX942-NEXT: v_lshlrev_b32_e32 v2, s6, v2
+; GFX942-NEXT: v_and_or_b32 v4, v5, s7, v2
+; GFX942-NEXT: v_mov_b64_e32 v[2:3], v[4:5]
; GFX942-NEXT: buffer_wbl2 sc1
-; GFX942-NEXT: buffer_atomic_cmpswap v[2:3], v4, s[0:3], 0 offen sc0
+; GFX942-NEXT: buffer_atomic_cmpswap v[2:3], v1, s[0:3], 0 offen sc0
; GFX942-NEXT: s_waitcnt vmcnt(0)
; GFX942-NEXT: buffer_inv sc1
-; GFX942-NEXT: v_cmp_eq_u32_e32 vcc, v2, v1
+; GFX942-NEXT: v_cmp_eq_u32_e32 vcc, v2, v5
; GFX942-NEXT: s_or_b64 s[4:5], vcc, s[4:5]
-; GFX942-NEXT: v_mov_b32_e32 v1, v2
; GFX942-NEXT: s_andn2_b64 exec, exec, s[4:5]
; GFX942-NEXT: s_cbranch_execnz .LBB10_1
; GFX942-NEXT: ; %bb.2: ; %atomicrmw.end
@@ -2629,125 +2633,127 @@ define half @buffer_fat_ptr_agent_atomic_fmax_ret_f16__offset__amdgpu_no_fine_gr
; GFX11-TRUE16-NEXT: v_max_f16_e32 v0.l, v0.l, v0.l
; GFX11-TRUE16-NEXT: s_and_b32 s4, s16, -4
; GFX11-TRUE16-NEXT: s_delay_alu instid0(SALU_CYCLE_1) | instskip(SKIP_1) | instid1(SALU_CYCLE_1)
-; GFX11-TRUE16-NEXT: v_mov_b32_e32 v5, s4
+; GFX11-TRUE16-NEXT: v_mov_b32_e32 v3, s4
; GFX11-TRUE16-NEXT: s_and_b32 s4, s16, 3
; GFX11-TRUE16-NEXT: s_lshl_b32 s4, s4, 3
; GFX11-TRUE16-NEXT: s_delay_alu instid0(SALU_CYCLE_1)
; GFX11-TRUE16-NEXT: s_lshl_b32 s5, 0xffff, s4
-; GFX11-TRUE16-NEXT: buffer_load_b32 v2, v5, s[0:3], 0 offen
+; GFX11-TRUE16-NEXT: buffer_load_b32 v1, v3, s[0:3], 0 offen
; GFX11-TRUE16-NEXT: s_not_b32 s6, s5
; GFX11-TRUE16-NEXT: s_mov_b32 s5, 0
; GFX11-TRUE16-NEXT: .p2align 6
; GFX11-TRUE16-NEXT: .LBB10_1: ; %atomicrmw.start
; GFX11-TRUE16-NEXT: ; =>This Inner Loop Header: Depth=1
; GFX11-TRUE16-NEXT: s_waitcnt vmcnt(0)
-; GFX11-TRUE16-NEXT: v_lshrrev_b32_e32 v1, s4, v2
+; GFX11-TRUE16-NEXT: v_mov_b32_e32 v5, v1
; GFX11-TRUE16-NEXT: s_waitcnt_vscnt null, 0x0
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
+; GFX11-TRUE16-NEXT: v_lshrrev_b32_e32 v1, s4, v5
; GFX11-TRUE16-NEXT: v_max_f16_e32 v0.h, v1.l, v1.l
-; GFX11-TRUE16-NEXT: v_max_f16_e32 v1.l, v0.h, v0.l
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
+; GFX11-TRUE16-NEXT: v_max_f16_e32 v1.l, v0.h, v0.l
; GFX11-TRUE16-NEXT: v_and_b32_e32 v1, 0xffff, v1
-; GFX11-TRUE16-NEXT: v_lshlrev_b32_e32 v1, s4, v1
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
-; GFX11-TRUE16-NEXT: v_and_or_b32 v1, v2, s6, v1
-; GFX11-TRUE16-NEXT: v_dual_mov_b32 v4, v2 :: v_dual_mov_b32 v3, v1
-; GFX11-TRUE16-NEXT: buffer_atomic_cmpswap_b32 v[3:4], v5, s[0:3], 0 offen glc
+; GFX11-TRUE16-NEXT: v_lshlrev_b32_e32 v1, s4, v1
+; GFX11-TRUE16-NEXT: v_and_or_b32 v4, v5, s6, v1
+; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1)
+; GFX11-TRUE16-NEXT: v_dual_mov_b32 v1, v4 :: v_dual_mov_b32 v2, v5
+; GFX11-TRUE16-NEXT: buffer_atomic_cmpswap_b32 v[1:2], v3, s[0:3], 0 offen glc
; GFX11-TRUE16-NEXT: s_waitcnt vmcnt(0)
; GFX11-TRUE16-NEXT: buffer_gl1_inv
; GFX11-TRUE16-NEXT: buffer_gl0_inv
-; GFX11-TRUE16-NEXT: v_cmp_eq_u32_e32 vcc_lo, v3, v2
-; GFX11-TRUE16-NEXT: v_mov_b32_e32 v2, v3
+; GFX11-TRUE16-NEXT: v_cmp_eq_u32_e32 vcc_lo, v1, v5
; GFX11-TRUE16-NEXT: s_or_b32 s5, vcc_lo, s5
; GFX11-TRUE16-NEXT: s_delay_alu instid0(SALU_CYCLE_1)
; GFX11-TRUE16-NEXT: s_and_not1_b32 exec_lo, exec_lo, s5
; GFX11-TRUE16-NEXT: s_cbranch_execnz .LBB10_1
; GFX11-TRUE16-NEXT: ; %bb.2: ; %atomicrmw.end
; GFX11-TRUE16-NEXT: s_or_b32 exec_lo, exec_lo, s5
-; GFX11-TRUE16-NEXT: v_lshrrev_b32_e32 v0, s4, v3
+; GFX11-TRUE16-NEXT: v_lshrrev_b32_e32 v0, s4, v1
; GFX11-TRUE16-NEXT: s_setpc_b64 s[30:31]
;
; GFX11-FAKE16-LABEL: buffer_fat_ptr_agent_atomic_fmax_ret_f16__offset__amdgpu_no_fine_grained_memory:
; GFX11-FAKE16: ; %bb.0:
; GFX11-FAKE16-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX11-FAKE16-NEXT: s_addk_i32 s16, 0x200
-; GFX11-FAKE16-NEXT: v_max_f16_e32 v5, v0, v0
+; GFX11-FAKE16-NEXT: v_max_f16_e32 v0, v0, v0
; GFX11-FAKE16-NEXT: s_and_b32 s4, s16, -4
; GFX11-FAKE16-NEXT: s_delay_alu instid0(SALU_CYCLE_1) | instskip(SKIP_1) | instid1(SALU_CYCLE_1)
-; GFX11-FAKE16-NEXT: v_mov_b32_e32 v4, s4
+; GFX11-FAKE16-NEXT: v_mov_b32_e32 v3, s4
; GFX11-FAKE16-NEXT: s_and_b32 s4, s16, 3
; GFX11-FAKE16-NEXT: s_lshl_b32 s4, s4, 3
; GFX11-FAKE16-NEXT: s_delay_alu instid0(SALU_CYCLE_1)
; GFX11-FAKE16-NEXT: s_lshl_b32 s5, 0xffff, s4
-; GFX11-FAKE16-NEXT: buffer_load_b32 v1, v4, s[0:3], 0 offen
+; GFX11-FAKE16-NEXT: buffer_load_b32 v1, v3, s[0:3], 0 offen
; GFX11-FAKE16-NEXT: s_not_b32 s6, s5
; GFX11-FAKE16-NEXT: s_mov_b32 s5, 0
; GFX11-FAKE16-NEXT: .p2align 6
; GFX11-FAKE16-NEXT: .LBB10_1: ; %atomicrmw.start
; GFX11-FAKE16-NEXT: ; =>This Inner Loop Header: Depth=1
; GFX11-FAKE16-NEXT: s_waitcnt vmcnt(0)
-; GFX11-FAKE16-NEXT: v_lshrrev_b32_e32 v0, s4, v1
+; GFX11-FAKE16-NEXT: v_mov_b32_e32 v5, v1
; GFX11-FAKE16-NEXT: s_waitcnt_vscnt null, 0x0
; GFX11-FAKE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
-; GFX11-FAKE16-NEXT: v_max_f16_e32 v0, v0, v0
-; GFX11-FAKE16-NEXT: v_max_f16_e32 v0, v0, v5
+; GFX11-FAKE16-NEXT: v_lshrrev_b32_e32 v1, s4, v5
+; GFX11-FAKE16-NEXT: v_max_f16_e32 v1, v1, v1
; GFX11-FAKE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
-; GFX11-FAKE16-NEXT: v_and_b32_e32 v0, 0xffff, v0
-; GFX11-FAKE16-NEXT: v_lshlrev_b32_e32 v0, s4, v0
+; GFX11-FAKE16-NEXT: v_max_f16_e32 v1, v1, v0
+; GFX11-FAKE16-NEXT: v_and_b32_e32 v1, 0xffff, v1
; GFX11-FAKE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
-; GFX11-FAKE16-NEXT: v_and_or_b32 v0, v1, s6, v0
-; GFX11-FAKE16-NEXT: v_dual_mov_b32 v3, v1 :: v_dual_mov_b32 v2, v0
-; GFX11-FAKE16-NEXT: buffer_atomic_cmpswap_b32 v[2:3], v4, s[0:3], 0 offen glc
+; GFX11-FAKE16-NEXT: v_lshlrev_b32_e32 v1, s4, v1
+; GFX11-FAKE16-NEXT: v_and_or_b32 v4, v5, s6, v1
+; GFX11-FAKE16-NEXT: s_delay_alu instid0(VALU_DEP_1)
+; GFX11-FAKE16-NEXT: v_dual_mov_b32 v1, v4 :: v_dual_mov_b32 v2, v5
+; GFX11-FAKE16-NEXT: buffer_atomic_cmpswap_b32 v[1:2], v3, s[0:3], 0 offen glc
; GFX11-FAKE16-NEXT: s_waitcnt vmcnt(0)
; GFX11-FAKE16-NEXT: buffer_gl1_inv
; GFX11-FAKE16-NEXT: buffer_gl0_inv
-; GFX11-FAKE16-NEXT: v_cmp_eq_u32_e32 vcc_lo, v2, v1
-; GFX11-FAKE16-NEXT: v_mov_b32_e32 v1, v2
+; GFX11-FAKE16-NEXT: v_cmp_eq_u32_e32 vcc_lo, v1, v5
; GFX11-FAKE16-NEXT: s_or_b32 s5, vcc_lo, s5
; GFX11-FAKE16-NEXT: s_delay_alu instid0(SALU_CYCLE_1)
; GFX11-FAKE16-NEXT: s_and_not1_b32 exec_lo, exec_lo, s5
; GFX11-FAKE16-NEXT: s_cbranch_execnz .LBB10_1
; GFX11-FAKE16-NEXT: ; %bb.2: ; %atomicrmw.end
; GFX11-FAKE16-NEXT: s_or_b32 exec_lo, exec_lo, s5
-; GFX11-FAKE16-NEXT: v_lshrrev_b32_e32 v0, s4, v2
+; GFX11-FAKE16-NEXT: v_lshrrev_b32_e32 v0, s4, v1
; GFX11-FAKE16-NEXT: s_setpc_b64 s[30:31]
;
; GFX10-LABEL: buffer_fat_ptr_agent_atomic_fmax_ret_f16__offset__amdgpu_no_fine_grained_memory:
; GFX10: ; %bb.0:
; GFX10-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX10-NEXT: s_addk_i32 s20, 0x200
-; GFX10-NEXT: v_max_f16_e32 v5, v0, v0
+; GFX10-NEXT: v_max_f16_e32 v0, v0, v0
; GFX10-NEXT: s_and_b32 s4, s20, -4
-; GFX10-NEXT: v_mov_b32_e32 v4, s4
+; GFX10-NEXT: v_mov_b32_e32 v3, s4
; GFX10-NEXT: s_and_b32 s4, s20, 3
; GFX10-NEXT: s_lshl_b32 s4, s4, 3
; GFX10-NEXT: s_lshl_b32 s5, 0xffff, s4
-; GFX10-NEXT: buffer_load_dword v1, v4, s[16:19], 0 offen
+; GFX10-NEXT: buffer_load_dword v1, v3, s[16:19], 0 offen
; GFX10-NEXT: s_not_b32 s6, s5
; GFX10-NEXT: s_mov_b32 s5, 0
; GFX10-NEXT: .LBB10_1: ; %atomicrmw.start
; GFX10-NEXT: ; =>This Inner Loop Header: Depth=1
; GFX10-NEXT: s_waitcnt vmcnt(0)
-; GFX10-NEXT: v_lshrrev_b32_e32 v0, s4, v1
+; GFX10-NEXT: v_mov_b32_e32 v5, v1
; GFX10-NEXT: s_waitcnt_vscnt null, 0x0
-; GFX10-NEXT: v_max_f16_e32 v0, v0, v0
-; GFX10-NEXT: v_max_f16_e32 v0, v0, v5
-; GFX10-NEXT: v_lshlrev_b32_sdwa v0, s4, v0 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:WORD_0
-; GFX10-NEXT: v_and_or_b32 v0, v1, s6, v0
-; GFX10-NEXT: v_mov_b32_e32 v3, v1
-; GFX10-NEXT: v_mov_b32_e32 v2, v0
-; GFX10-NEXT: buffer_atomic_cmpswap v[2:3], v4, s[16:19], 0 offen glc
+; GFX10-NEXT: v_lshrrev_b32_e32 v1, s4, v5
+; GFX10-NEXT: v_max_f16_e32 v1, v1, v1
+; GFX10-NEXT: v_max_f16_e32 v1, v1, v0
+; GFX10-NEXT: v_lshlrev_b32_sdwa v1, s4, v1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:WORD_0
+; GFX10-NEXT: v_and_or_b32 v4, v5, s6, v1
+; GFX10-NEXT: v_mov_b32_e32 v1, v4
+; GFX10-NEXT: v_mov_b32_e32 v2, v5
+; GFX10-NEXT: buffer_atomic_cmpswap v[1:2], v3, s[16:19], 0 offen glc
; GFX10-NEXT: s_waitcnt vmcnt(0)
; GFX10-NEXT: buffer_gl1_inv
; GFX10-NEXT: buffer_gl0_inv
-; GFX10-NEXT: v_cmp_eq_u32_e32 vcc_lo, v2, v1
-; GFX10-NEXT: v_mov_b32_e32 v1, v2
+; GFX10-NEXT: v_cmp_eq_u32_e32 vcc_lo, v1, v5
; GFX10-NEXT: s_or_b32 s5, vcc_lo, s5
; GFX10-NEXT: s_andn2_b32 exec_lo, exec_lo, s5
; GFX10-NEXT: s_cbranch_execnz .LBB10_1
; GFX10-NEXT: ; %bb.2: ; %atomicrmw.end
; GFX10-NEXT: s_or_b32 exec_lo, exec_lo, s5
-; GFX10-NEXT: v_lshrrev_b32_e32 v0, s4, v2
+; GFX10-NEXT: v_lshrrev_b32_e32 v0, s4, v1
; GFX10-NEXT: s_setpc_b64 s[30:31]
;
; GFX90A-LABEL: buffer_fat_ptr_agent_atomic_fmax_ret_f16__offset__amdgpu_no_fine_grained_memory:
@@ -2755,29 +2761,29 @@ define half @buffer_fat_ptr_agent_atomic_fmax_ret_f16__offset__amdgpu_no_fine_gr
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX90A-NEXT: s_addk_i32 s20, 0x200
; GFX90A-NEXT: s_and_b32 s4, s20, -4
-; GFX90A-NEXT: v_mov_b32_e32 v4, s4
-; GFX90A-NEXT: buffer_load_dword v1, v4, s[16:19], 0 offen
+; GFX90A-NEXT: v_mov_b32_e32 v1, s4
+; GFX90A-NEXT: buffer_load_dword v2, v1, s[16:19], 0 offen
; GFX90A-NEXT: s_and_b32 s4, s20, 3
; GFX90A-NEXT: s_lshl_b32 s6, s4, 3
; GFX90A-NEXT: s_lshl_b32 s4, 0xffff, s6
; GFX90A-NEXT: s_not_b32 s7, s4
; GFX90A-NEXT: s_mov_b64 s[4:5], 0
-; GFX90A-NEXT: v_max_f16_e32 v5, v0, v0
+; GFX90A-NEXT: v_max_f16_e32 v0, v0, v0
; GFX90A-NEXT: .LBB10_1: ; %atomicrmw.start
; GFX90A-NEXT: ; =>This Inner Loop Header: Depth=1
; GFX90A-NEXT: s_waitcnt vmcnt(0)
-; GFX90A-NEXT: v_lshrrev_b32_e32 v0, s6, v1
-; GFX90A-NEXT: v_max_f16_e32 v0, v0, v0
-; GFX90A-NEXT: v_max_f16_e32 v0, v0, v5
-; GFX90A-NEXT: v_lshlrev_b32_e32 v0, s6, v0
-; GFX90A-NEXT: v_and_or_b32 v0, v1, s7, v0
-; GFX90A-NEXT: v_pk_mov_b32 v[2:3], v[0:1], v[0:1] op_sel:[0,1]
-; GFX90A-NEXT: buffer_atomic_cmpswap v[2:3], v4, s[16:19], 0 offen glc
+; GFX90A-NEXT: v_mov_b32_e32 v5, v2
+; GFX90A-NEXT: v_lshrrev_b32_e32 v2, s6, v5
+; GFX90A-NEXT: v_max_f16_e32 v2, v2, v2
+; GFX90A-NEXT: v_max_f16_e32 v2, v2, v0
+; GFX90A-NEXT: v_lshlrev_b32_e32 v2, s6, v2
+; GFX90A-NEXT: v_and_or_b32 v4, v5, s7, v2
+; GFX90A-NEXT: v_pk_mov_b32 v[2:3], v[4:5], v[4:5] op_sel:[0,1]
+; GFX90A-NEXT: buffer_atomic_cmpswap v[2:3], v1, s[16:19], 0 offen glc
; GFX90A-NEXT: s_waitcnt vmcnt(0)
; GFX90A-NEXT: buffer_wbinvl1
-; GFX90A-NEXT: v_cmp_eq_u32_e32 vcc, v2, v1
+; GFX90A-NEXT: v_cmp_eq_u32_e32 vcc, v2, v5
; GFX90A-NEXT: s_or_b64 s[4:5], vcc, s[4:5]
-; GFX90A-NEXT: v_mov_b32_e32 v1, v2
; GFX90A-NEXT: s_andn2_b64 exec, exec, s[4:5]
; GFX90A-NEXT: s_cbranch_execnz .LBB10_1
; GFX90A-NEXT: ; %bb.2: ; %atomicrmw.end
@@ -2790,35 +2796,35 @@ define half @buffer_fat_ptr_agent_atomic_fmax_ret_f16__offset__amdgpu_no_fine_gr
; GFX908-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX908-NEXT: s_addk_i32 s20, 0x200
; GFX908-NEXT: s_and_b32 s4, s20, -4
-; GFX908-NEXT: v_mov_b32_e32 v4, s4
-; GFX908-NEXT: buffer_load_dword v1, v4, s[16:19], 0 offen
+; GFX908-NEXT: v_mov_b32_e32 v3, s4
+; GFX908-NEXT: buffer_load_dword v1, v3, s[16:19], 0 offen
; GFX908-NEXT: s_and_b32 s4, s20, 3
; GFX908-NEXT: s_lshl_b32 s6, s4, 3
; GFX908-NEXT: s_lshl_b32 s4, 0xffff, s6
; GFX908-NEXT: s_not_b32 s7, s4
; GFX908-NEXT: s_mov_b64 s[4:5], 0
-; GFX908-NEXT: v_max_f16_e32 v5, v0, v0
+; GFX908-NEXT: v_max_f16_e32 v0, v0, v0
; GFX908-NEXT: .LBB10_1: ; %atomicrmw.start
; GFX908-NEXT: ; =>This Inner Loop Header: Depth=1
; GFX908-NEXT: s_waitcnt vmcnt(0)
-; GFX908-NEXT: v_lshrrev_b32_e32 v0, s6, v1
-; GFX908-NEXT: v_max_f16_e32 v0, v0, v0
-; GFX908-NEXT: v_max_f16_e32 v0, v0, v5
-; GFX908-NEXT: v_lshlrev_b32_e32 v0, s6, v0
-; GFX908-NEXT: v_and_or_b32 v0, v1, s7, v0
-; GFX908-NEXT: v_mov_b32_e32 v3, v1
-; GFX908-NEXT: v_mov_b32_e32 v2, v0
-; GFX908-NEXT: buffer_atomic_cmpswap v[2:3], v4, s[16:19], 0 offen glc
+; GFX908-NEXT: v_mov_b32_e32 v5, v1
+; GFX908-NEXT: v_lshrrev_b32_e32 v1, s6, v5
+; GFX908-NEXT: v_max_f16_e32 v1, v1, v1
+; GFX908-NEXT: v_max_f16_e32 v1, v1, v0
+; GFX908-NEXT: v_lshlrev_b32_e32 v1, s6, v1
+; GFX908-NEXT: v_and_or_b32 v4, v5, s7, v1
+; GFX908-NEXT: v_mov_b32_e32 v1, v4
+; GFX908-NEXT: v_mov_b32_e32 v2, v5
+; GFX908-NEXT: buffer_atomic_cmpswap v[1:2], v3, s[16:19], 0 offen glc
; GFX908-NEXT: s_waitcnt vmcnt(0)
; GFX908-NEXT: buffer_wbinvl1
-; GFX908-NEXT: v_cmp_eq_u32_e32 vcc, v2, v1
+; GFX908-NEXT: v_cmp_eq_u32_e32 vcc, v1, v5
; GFX908-NEXT: s_or_b64 s[4:5], vcc, s[4:5]
-; GFX908-NEXT: v_mov_b32_e32 v1, v2
; GFX908-NEXT: s_andn2_b64 exec, exec, s[4:5]
; GFX908-NEXT: s_cbranch_execnz .LBB10_1
; GFX908-NEXT: ; %bb.2: ; %atomicrmw.end
; GFX908-NEXT: s_or_b64 exec, exec, s[4:5]
-; GFX908-NEXT: v_lshrrev_b32_e32 v0, s6, v2
+; GFX908-NEXT: v_lshrrev_b32_e32 v0, s6, v1
; GFX908-NEXT: s_setpc_b64 s[30:31]
;
; GFX8-LABEL: buffer_fat_ptr_agent_atomic_fmax_ret_f16__offset__amdgpu_no_fine_grained_memory:
@@ -2826,36 +2832,36 @@ define half @buffer_fat_ptr_agent_atomic_fmax_ret_f16__offset__amdgpu_no_fine_gr
; GFX8-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX8-NEXT: s_addk_i32 s20, 0x200
; GFX8-NEXT: s_and_b32 s4, s20, -4
-; GFX8-NEXT: v_mov_b32_e32 v4, s4
-; GFX8-NEXT: buffer_load_dword v1, v4, s[16:19], 0 offen
+; GFX8-NEXT: v_mov_b32_e32 v3, s4
+; GFX8-NEXT: buffer_load_dword v1, v3, s[16:19], 0 offen
; GFX8-NEXT: s_and_b32 s4, s20, 3
; GFX8-NEXT: s_lshl_b32 s6, s4, 3
; GFX8-NEXT: s_lshl_b32 s4, 0xffff, s6
; GFX8-NEXT: s_not_b32 s7, s4
; GFX8-NEXT: s_mov_b64 s[4:5], 0
-; GFX8-NEXT: v_max_f16_e32 v5, v0, v0
+; GFX8-NEXT: v_max_f16_e32 v0, v0, v0
; GFX8-NEXT: .LBB10_1: ; %atomicrmw.start
; GFX8-NEXT: ; =>This Inner Loop Header: Depth=1
; GFX8-NEXT: s_waitcnt vmcnt(0)
-; GFX8-NEXT: v_lshrrev_b32_e32 v0, s6, v1
-; GFX8-NEXT: v_max_f16_e32 v0, v0, v0
-; GFX8-NEXT: v_max_f16_e32 v0, v0, v5
-; GFX8-NEXT: v_and_b32_e32 v2, s7, v1
-; GFX8-NEXT: v_lshlrev_b32_e32 v0, s6, v0
-; GFX8-NEXT: v_or_b32_e32 v0, v2, v0
-; GFX8-NEXT: v_mov_b32_e32 v3, v1
-; GFX8-NEXT: v_mov_b32_e32 v2, v0
-; GFX8-NEXT: buffer_atomic_cmpswap v[2:3], v4, s[16:19], 0 offen glc
+; GFX8-NEXT: v_mov_b32_e32 v5, v1
+; GFX8-NEXT: v_lshrrev_b32_e32 v1, s6, v5
+; GFX8-NEXT: v_max_f16_e32 v1, v1, v1
+; GFX8-NEXT: v_max_f16_e32 v1, v1, v0
+; GFX8-NEXT: v_and_b32_e32 v2, s7, v5
+; GFX8-NEXT: v_lshlrev_b32_e32 v1, s6, v1
+; GFX8-NEXT: v_or_b32_e32 v4, v2, v1
+; GFX8-NEXT: v_mov_b32_e32 v1, v4
+; GFX8-NEXT: v_mov_b32_e32 v2, v5
+; GFX8-NEXT: buffer_atomic_cmpswap v[1:2], v3, s[16:19], 0 offen glc
; GFX8-NEXT: s_waitcnt vmcnt(0)
; GFX8-NEXT: buffer_wbinvl1
-; GFX8-NEXT: v_cmp_eq_u32_e32 vcc, v2, v1
+; GFX8-NEXT: v_cmp_eq_u32_e32 vcc, v1, v5
; GFX8-NEXT: s_or_b64 s[4:5], vcc, s[4:5]
-; GFX8-NEXT: v_mov_b32_e32 v1, v2
; GFX8-NEXT: s_andn2_b64 exec, exec, s[4:5]
; GFX8-NEXT: s_cbranch_execnz .LBB10_1
; GFX8-NEXT: ; %bb.2: ; %atomicrmw.end
; GFX8-NEXT: s_or_b64 exec, exec, s[4:5]
-; GFX8-NEXT: v_lshrrev_b32_e32 v0, s6, v2
+; GFX8-NEXT: v_lshrrev_b32_e32 v0, s6, v1
; GFX8-NEXT: s_setpc_b64 s[30:31]
;
; GFX7-LABEL: buffer_fat_ptr_agent_atomic_fmax_ret_f16__offset__amdgpu_no_fine_grained_memory:
@@ -2863,38 +2869,38 @@ define half @buffer_fat_ptr_agent_atomic_fmax_ret_f16__offset__amdgpu_no_fine_gr
; GFX7-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX7-NEXT: s_addk_i32 s20, 0x200
; GFX7-NEXT: s_and_b32 s4, s20, -4
-; GFX7-NEXT: v_mov_b32_e32 v4, s4
-; GFX7-NEXT: buffer_load_dword v1, v4, s[16:19], 0 offen
+; GFX7-NEXT: v_mov_b32_e32 v3, s4
+; GFX7-NEXT: buffer_load_dword v1, v3, s[16:19], 0 offen
; GFX7-NEXT: v_cvt_f16_f32_e32 v0, v0
; GFX7-NEXT: s_and_b32 s4, s20, 3
; GFX7-NEXT: s_lshl_b32 s6, s4, 3
; GFX7-NEXT: s_lshl_b32 s4, 0xffff, s6
-; GFX7-NEXT: v_cvt_f32_f16_e32 v5, v0
+; GFX7-NEXT: v_cvt_f32_f16_e32 v0, v0
; GFX7-NEXT: s_not_b32 s7, s4
; GFX7-NEXT: s_mov_b64 s[4:5], 0
; GFX7-NEXT: .LBB10_1: ; %atomicrmw.start
; GFX7-NEXT: ; =>This Inner Loop Header: Depth=1
; GFX7-NEXT: s_waitcnt vmcnt(0)
-; GFX7-NEXT: v_lshrrev_b32_e32 v0, s6, v1
-; GFX7-NEXT: v_cvt_f32_f16_e32 v0, v0
-; GFX7-NEXT: v_and_b32_e32 v2, s7, v1
-; GFX7-NEXT: v_max_f32_e32 v0, v0, v5
-; GFX7-NEXT: v_cvt_f16_f32_e32 v0, v0
-; GFX7-NEXT: v_lshlrev_b32_e32 v0, s6, v0
-; GFX7-NEXT: v_or_b32_e32 v0, v2, v0
-; GFX7-NEXT: v_mov_b32_e32 v3, v1
-; GFX7-NEXT: v_mov_b32_e32 v2, v0
-; GFX7-NEXT: buffer_atomic_cmpswap v[2:3], v4, s[16:19], 0 offen glc
+; GFX7-NEXT: v_mov_b32_e32 v5, v1
+; GFX7-NEXT: v_lshrrev_b32_e32 v1, s6, v5
+; GFX7-NEXT: v_cvt_f32_f16_e32 v1, v1
+; GFX7-NEXT: v_and_b32_e32 v2, s7, v5
+; GFX7-NEXT: v_max_f32_e32 v1, v1, v0
+; GFX7-NEXT: v_cvt_f16_f32_e32 v1, v1
+; GFX7-NEXT: v_lshlrev_b32_e32 v1, s6, v1
+; GFX7-NEXT: v_or_b32_e32 v4, v2, v1
+; GFX7-NEXT: v_mov_b32_e32 v1, v4
+; GFX7-NEXT: v_mov_b32_e32 v2, v5
+; GFX7-NEXT: buffer_atomic_cmpswap v[1:2], v3, s[16:19], 0 offen glc
; GFX7-NEXT: s_waitcnt vmcnt(0)
; GFX7-NEXT: buffer_wbinvl1
-; GFX7-NEXT: v_cmp_eq_u32_e32 vcc, v2, v1
+; GFX7-NEXT: v_cmp_eq_u32_e32 vcc, v1, v5
; GFX7-NEXT: s_or_b64 s[4:5], vcc, s[4:5]
-; GFX7-NEXT: v_mov_b32_e32 v1, v2
; GFX7-NEXT: s_andn2_b64 exec, exec, s[4:5]
; GFX7-NEXT: s_cbranch_execnz .LBB10_1
; GFX7-NEXT: ; %bb.2: ; %atomicrmw.end
; GFX7-NEXT: s_or_b64 exec, exec, s[4:5]
-; GFX7-NEXT: v_lshrrev_b32_e32 v0, s6, v2
+; GFX7-NEXT: v_lshrrev_b32_e32 v0, s6, v1
; GFX7-NEXT: v_cvt_f32_f16_e32 v0, v0
; GFX7-NEXT: s_setpc_b64 s[30:31]
;
@@ -2903,39 +2909,39 @@ define half @buffer_fat_ptr_agent_atomic_fmax_ret_f16__offset__amdgpu_no_fine_gr
; GFX6-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX6-NEXT: s_addk_i32 s20, 0x200
; GFX6-NEXT: s_and_b32 s4, s20, -4
-; GFX6-NEXT: v_mov_b32_e32 v4, s4
-; GFX6-NEXT: buffer_load_dword v1, v4, s[16:19], 0 offen
+; GFX6-NEXT: v_mov_b32_e32 v3, s4
+; GFX6-NEXT: buffer_load_dword v1, v3, s[16:19], 0 offen
; GFX6-NEXT: v_cvt_f16_f32_e32 v0, v0
; GFX6-NEXT: s_and_b32 s4, s20, 3
; GFX6-NEXT: s_lshl_b32 s6, s4, 3
; GFX6-NEXT: s_lshl_b32 s4, 0xffff, s6
-; GFX6-NEXT: v_cvt_f32_f16_e32 v5, v0
+; GFX6-NEXT: v_cvt_f32_f16_e32 v0, v0
; GFX6-NEXT: s_not_b32 s7, s4
; GFX6-NEXT: s_mov_b64 s[4:5], 0
; GFX6-NEXT: .LBB10_1: ; %atomicrmw.start
; GFX6-NEXT: ; =>This Inner Loop Header: Depth=1
; GFX6-NEXT: s_waitcnt vmcnt(0)
-; GFX6-NEXT: v_lshrrev_b32_e32 v0, s6, v1
-; GFX6-NEXT: v_cvt_f32_f16_e32 v0, v0
+; GFX6-NEXT: v_mov_b32_e32 v5, v1
; GFX6-NEXT: s_waitcnt expcnt(0)
-; GFX6-NEXT: v_and_b32_e32 v2, s7, v1
-; GFX6-NEXT: v_max_f32_e32 v0, v0, v5
-; GFX6-NEXT: v_cvt_f16_f32_e32 v0, v0
-; GFX6-NEXT: v_lshlrev_b32_e32 v0, s6, v0
-; GFX6-NEXT: v_or_b32_e32 v0, v2, v0
-; GFX6-NEXT: v_mov_b32_e32 v3, v1
-; GFX6-NEXT: v_mov_b32_e32 v2, v0
-; GFX6-NEXT: buffer_atomic_cmpswap v[2:3], v4, s[16:19], 0 offen glc
+; GFX6-NEXT: v_lshrrev_b32_e32 v1, s6, v5
+; GFX6-NEXT: v_cvt_f32_f16_e32 v1, v1
+; GFX6-NEXT: v_and_b32_e32 v2, s7, v5
+; GFX6-NEXT: v_max_f32_e32 v1, v1, v0
+; GFX6-NEXT: v_cvt_f16_f32_e32 v1, v1
+; GFX6-NEXT: v_lshlrev_b32_e32 v1, s6, v1
+; GFX6-NEXT: v_or_b32_e32 v4, v2, v1
+; GFX6-NEXT: v_mov_b32_e32 v1, v4
+; GFX6-NEXT: v_mov_b32_e32 v2, v5
+; GFX6-NEXT: buffer_atomic_cmpswap v[1:2], v3, s[16:19], 0 offen glc
; GFX6-NEXT: s_waitcnt vmcnt(0)
; GFX6-NEXT: buffer_wbinvl1
-; GFX6-NEXT: v_cmp_eq_u32_e32 vcc, v2, v1
+; GFX6-NEXT: v_cmp_eq_u32_e32 vcc, v1, v5
; GFX6-NEXT: s_or_b64 s[4:5], vcc, s[4:5]
-; GFX6-NEXT: v_mov_b32_e32 v1, v2
; GFX6-NEXT: s_andn2_b64 exec, exec, s[4:5]
; GFX6-NEXT: s_cbranch_execnz .LBB10_1
; GFX6-NEXT: ; %bb.2: ; %atomicrmw.end
; GFX6-NEXT: s_or_b64 exec, exec, s[4:5]
-; GFX6-NEXT: v_lshrrev_b32_e32 v0, s6, v2
+; GFX6-NEXT: v_lshrrev_b32_e32 v0, s6, v1
; GFX6-NEXT: v_cvt_f32_f16_e32 v0, v0
; GFX6-NEXT: s_waitcnt expcnt(0)
; GFX6-NEXT: s_setpc_b64 s[30:31]
@@ -2963,29 +2969,30 @@ define void @buffer_fat_ptr_agent_atomic_fmax_noret_f16__offset__amdgpu_no_fine_
; GFX12-TRUE16-NEXT: s_lshl_b32 s4, s4, 3
; GFX12-TRUE16-NEXT: s_wait_alu 0xfffe
; GFX12-TRUE16-NEXT: s_lshl_b32 s5, 0xffff, s4
-; GFX12-TRUE16-NEXT: buffer_load_b32 v2, v3, s[0:3], null offen
+; GFX12-TRUE16-NEXT: buffer_load_b32 v1, v3, s[0:3], null offen
; GFX12-TRUE16-NEXT: s_not_b32 s6, s5
; GFX12-TRUE16-NEXT: s_mov_b32 s5, 0
; GFX12-TRUE16-NEXT: .LBB11_1: ; %atomicrmw.start
; GFX12-TRUE16-NEXT: ; =>This Inner Loop Header: Depth=1
; GFX12-TRUE16-NEXT: s_wait_loadcnt 0x0
-; GFX12-TRUE16-NEXT: v_lshrrev_b32_e32 v1, s4, v2
+; GFX12-TRUE16-NEXT: v_mov_b32_e32 v5, v1
; GFX12-TRUE16-NEXT: s_wait_storecnt 0x0
; GFX12-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
+; GFX12-TRUE16-NEXT: v_lshrrev_b32_e32 v1, s4, v5
; GFX12-TRUE16-NEXT: v_max_num_f16_e32 v0.h, v1.l, v1.l
-; GFX12-TRUE16-NEXT: v_max_num_f16_e32 v1.l, v0.h, v0.l
; GFX12-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
+; GFX12-TRUE16-NEXT: v_max_num_f16_e32 v1.l, v0.h, v0.l
; GFX12-TRUE16-NEXT: v_and_b32_e32 v1, 0xffff, v1
+; GFX12-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(SKIP_1) | instid1(VALU_DEP_1)
; GFX12-TRUE16-NEXT: v_lshlrev_b32_e32 v1, s4, v1
; GFX12-TRUE16-NEXT: s_wait_alu 0xfffe
-; GFX12-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
-; GFX12-TRUE16-NEXT: v_and_or_b32 v1, v2, s6, v1
-; GFX12-TRUE16-NEXT: v_dual_mov_b32 v5, v2 :: v_dual_mov_b32 v4, v1
-; GFX12-TRUE16-NEXT: buffer_atomic_cmpswap_b32 v[4:5], v3, s[0:3], null offen th:TH_ATOMIC_RETURN
+; GFX12-TRUE16-NEXT: v_and_or_b32 v4, v5, s6, v1
+; GFX12-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1)
+; GFX12-TRUE16-NEXT: v_dual_mov_b32 v1, v4 :: v_dual_mov_b32 v2, v5
+; GFX12-TRUE16-NEXT: buffer_atomic_cmpswap_b32 v[1:2], v3, s[0:3], null offen th:TH_ATOMIC_RETURN
; GFX12-TRUE16-NEXT: s_wait_loadcnt 0x0
; GFX12-TRUE16-NEXT: global_inv scope:SCOPE_DEV
-; GFX12-TRUE16-NEXT: v_cmp_eq_u32_e32 vcc_lo, v4, v2
-; GFX12-TRUE16-NEXT: v_mov_b32_e32 v2, v4
+; GFX12-TRUE16-NEXT: v_cmp_eq_u32_e32 vcc_lo, v1, v5
; GFX12-TRUE16-NEXT: s_or_b32 s5, vcc_lo, s5
; GFX12-TRUE16-NEXT: s_wait_alu 0xfffe
; GFX12-TRUE16-NEXT: s_and_not1_b32 exec_lo, exec_lo, s5
@@ -3003,39 +3010,40 @@ define void @buffer_fat_ptr_agent_atomic_fmax_noret_f16__offset__amdgpu_no_fine_
; GFX12-FAKE16-NEXT: s_wait_bvhcnt 0x0
; GFX12-FAKE16-NEXT: s_wait_kmcnt 0x0
; GFX12-FAKE16-NEXT: s_addk_co_i32 s16, 0x200
-; GFX12-FAKE16-NEXT: v_max_num_f16_e32 v3, v0, v0
+; GFX12-FAKE16-NEXT: v_max_num_f16_e32 v0, v0, v0
; GFX12-FAKE16-NEXT: s_wait_alu 0xfffe
; GFX12-FAKE16-NEXT: s_and_b32 s4, s16, -4
; GFX12-FAKE16-NEXT: s_wait_alu 0xfffe
-; GFX12-FAKE16-NEXT: v_mov_b32_e32 v2, s4
+; GFX12-FAKE16-NEXT: v_mov_b32_e32 v3, s4
; GFX12-FAKE16-NEXT: s_and_b32 s4, s16, 3
; GFX12-FAKE16-NEXT: s_wait_alu 0xfffe
; GFX12-FAKE16-NEXT: s_lshl_b32 s4, s4, 3
; GFX12-FAKE16-NEXT: s_wait_alu 0xfffe
; GFX12-FAKE16-NEXT: s_lshl_b32 s5, 0xffff, s4
-; GFX12-FAKE16-NEXT: buffer_load_b32 v1, v2, s[0:3], null offen
+; GFX12-FAKE16-NEXT: buffer_load_b32 v1, v3, s[0:3], null offen
; GFX12-FAKE16-NEXT: s_not_b32 s6, s5
; GFX12-FAKE16-NEXT: s_mov_b32 s5, 0
; GFX12-FAKE16-NEXT: .LBB11_1: ; %atomicrmw.start
; GFX12-FAKE16-NEXT: ; =>This Inner Loop Header: Depth=1
; GFX12-FAKE16-NEXT: s_wait_loadcnt 0x0
-; GFX12-FAKE16-NEXT: v_lshrrev_b32_e32 v0, s4, v1
+; GFX12-FAKE16-NEXT: v_mov_b32_e32 v5, v1
; GFX12-FAKE16-NEXT: s_wait_storecnt 0x0
; GFX12-FAKE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
-; GFX12-FAKE16-NEXT: v_max_num_f16_e32 v0, v0, v0
-; GFX12-FAKE16-NEXT: v_max_num_f16_e32 v0, v0, v3
+; GFX12-FAKE16-NEXT: v_lshrrev_b32_e32 v1, s4, v5
+; GFX12-FAKE16-NEXT: v_max_num_f16_e32 v1, v1, v1
; GFX12-FAKE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
-; GFX12-FAKE16-NEXT: v_and_b32_e32 v0, 0xffff, v0
-; GFX12-FAKE16-NEXT: v_lshlrev_b32_e32 v0, s4, v0
+; GFX12-FAKE16-NEXT: v_max_num_f16_e32 v1, v1, v0
+; GFX12-FAKE16-NEXT: v_and_b32_e32 v1, 0xffff, v1
+; GFX12-FAKE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(SKIP_1) | instid1(VALU_DEP_1)
+; GFX12-FAKE16-NEXT: v_lshlrev_b32_e32 v1, s4, v1
; GFX12-FAKE16-NEXT: s_wait_alu 0xfffe
-; GFX12-FAKE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
-; GFX12-FAKE16-NEXT: v_and_or_b32 v0, v1, s6, v0
-; GFX12-FAKE16-NEXT: v_dual_mov_b32 v5, v1 :: v_dual_mov_b32 v4, v0
-; GFX12-FAKE16-NEXT: buffer_atomic_cmpswap_b32 v[4:5], v2, s[0:3], null offen th:TH_ATOMIC_RETURN
+; GFX12-FAKE16-NEXT: v_and_or_b32 v4, v5, s6, v1
+; GFX12-FAKE16-NEXT: s_delay_alu instid0(VALU_DEP_1)
+; GFX12-FAKE16-NEXT: v_dual_mov_b32 v1, v4 :: v_dual_mov_b32 v2, v5
+; GFX12-FAKE16-NEXT: buffer_atomic_cmpswap_b32 v[1:2], v3, s[0:3], null offen th:TH_ATOMIC_RETURN
; GFX12-FAKE16-NEXT: s_wait_loadcnt 0x0
; GFX12-FAKE16-NEXT: global_inv scope:SCOPE_DEV
-; GFX12-FAKE16-NEXT: v_cmp_eq_u32_e32 vcc_lo, v4, v1
-; GFX12-FAKE16-NEXT: v_mov_b32_e32 v1, v4
+; GFX12-FAKE16-NEXT: v_cmp_eq_u32_e32 vcc_lo, v1, v5
; GFX12-FAKE16-NEXT: s_or_b32 s5, vcc_lo, s5
; GFX12-FAKE16-NEXT: s_wait_alu 0xfffe
; GFX12-FAKE16-NEXT: s_and_not1_b32 exec_lo, exec_lo, s5
@@ -3050,30 +3058,30 @@ define void @buffer_fat_ptr_agent_atomic_fmax_noret_f16__offset__amdgpu_no_fine_
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX942-NEXT: s_addk_i32 s16, 0x200
; GFX942-NEXT: s_and_b32 s4, s16, -4
-; GFX942-NEXT: v_mov_b32_e32 v2, s4
-; GFX942-NEXT: buffer_load_dword v1, v2, s[0:3], 0 offen
+; GFX942-NEXT: v_mov_b32_e32 v1, s4
+; GFX942-NEXT: buffer_load_dword v2, v1, s[0:3], 0 offen
; GFX942-NEXT: s_and_b32 s4, s16, 3
; GFX942-NEXT: s_lshl_b32 s6, s4, 3
; GFX942-NEXT: s_lshl_b32 s4, 0xffff, s6
; GFX942-NEXT: s_not_b32 s7, s4
; GFX942-NEXT: s_mov_b64 s[4:5], 0
-; GFX942-NEXT: v_max_f16_e32 v3, v0, v0
+; GFX942-NEXT: v_max_f16_e32 v0, v0, v0
; GFX942-NEXT: .LBB11_1: ; %atomicrmw.start
; GFX942-NEXT: ; =>This Inner Loop Header: Depth=1
; GFX942-NEXT: s_waitcnt vmcnt(0)
-; GFX942-NEXT: v_lshrrev_b32_e32 v0, s6, v1
-; GFX942-NEXT: v_max_f16_e32 v0, v0, v0
-; GFX942-NEXT: v_max_f16_e32 v0, v0, v3
-; GFX942-NEXT: v_lshlrev_b32_e32 v0, s6, v0
-; GFX942-NEXT: v_and_or_b32 v0, v1, s7, v0
-; GFX942-NEXT: v_mov_b64_e32 v[4:5], v[0:1]
+; GFX942-NEXT: v_mov_b32_e32 v5, v2
+; GFX942-NEXT: v_lshrrev_b32_e32 v2, s6, v5
+; GFX942-NEXT: v_max_f16_e32 v2, v2, v2
+; GFX942-NEXT: v_max_f16_e32 v2, v2, v0
+; GFX942-NEXT: v_lshlrev_b32_e32 v2, s6, v2
+; GFX942-NEXT: v_and_or_b32 v4, v5, s7, v2
+; GFX942-NEXT: v_mov_b64_e32 v[2:3], v[4:5]
; GFX942-NEXT: buffer_wbl2 sc1
-; GFX942-NEXT: buffer_atomic_cmpswap v[4:5], v2, s[0:3], 0 offen sc0
+; GFX942-NEXT: buffer_atomic_cmpswap v[2:3], v1, s[0:3], 0 offen sc0
; GFX942-NEXT: s_waitcnt vmcnt(0)
; GFX942-NEXT: buffer_inv sc1
-; GFX942-NEXT: v_cmp_eq_u32_e32 vcc, v4, v1
+; GFX942-NEXT: v_cmp_eq_u32_e32 vcc, v2, v5
; GFX942-NEXT: s_or_b64 s[4:5], vcc, s[4:5]
-; GFX942-NEXT: v_mov_b32_e32 v1, v4
; GFX942-NEXT: s_andn2_b64 exec, exec, s[4:5]
; GFX942-NEXT: s_cbranch_execnz .LBB11_1
; GFX942-NEXT: ; %bb.2: ; %atomicrmw.end
@@ -3092,30 +3100,31 @@ define void @buffer_fat_ptr_agent_atomic_fmax_noret_f16__offset__amdgpu_no_fine_
; GFX11-TRUE16-NEXT: s_lshl_b32 s4, s4, 3
; GFX11-TRUE16-NEXT: s_delay_alu instid0(SALU_CYCLE_1)
; GFX11-TRUE16-NEXT: s_lshl_b32 s5, 0xffff, s4
-; GFX11-TRUE16-NEXT: buffer_load_b32 v2, v3, s[0:3], 0 offen
+; GFX11-TRUE16-NEXT: buffer_load_b32 v1, v3, s[0:3], 0 offen
; GFX11-TRUE16-NEXT: s_not_b32 s6, s5
; GFX11-TRUE16-NEXT: s_mov_b32 s5, 0
; GFX11-TRUE16-NEXT: .p2align 6
; GFX11-TRUE16-NEXT: .LBB11_1: ; %atomicrmw.start
; GFX11-TRUE16-NEXT: ; =>This Inner Loop Header: Depth=1
; GFX11-TRUE16-NEXT: s_waitcnt vmcnt(0)
-; GFX11-TRUE16-NEXT: v_lshrrev_b32_e32 v1, s4, v2
+; GFX11-TRUE16-NEXT: v_mov_b32_e32 v5, v1
; GFX11-TRUE16-NEXT: s_waitcnt_vscnt null, 0x0
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
+; GFX11-TRUE16-NEXT: v_lshrrev_b32_e32 v1, s4, v5
; GFX11-TRUE16-NEXT: v_max_f16_e32 v0.h, v1.l, v1.l
-; GFX11-TRUE16-NEXT: v_max_f16_e32 v1.l, v0.h, v0.l
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
+; GFX11-TRUE16-NEXT: v_max_f16_e32 v1.l, v0.h, v0.l
; GFX11-TRUE16-NEXT: v_and_b32_e32 v1, 0xffff, v1
-; GFX11-TRUE16-NEXT: v_lshlrev_b32_e32 v1, s4, v1
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
-; GFX11-TRUE16-NEXT: v_and_or_b32 v1, v2, s6, v1
-; GFX11-TRUE16-NEXT: v_dual_mov_b32 v5, v2 :: v_dual_mov_b32 v4, v1
-; GFX11-TRUE16-NEXT: buffer_atomic_cmpswap_b32 v[4:5], v3, s[0:3], 0 offen glc
+; GFX11-TRUE16-NEXT: v_lshlrev_b32_e32 v1, s4, v1
+; GFX11-TRUE16-NEXT: v_and_or_b32 v4, v5, s6, v1
+; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1)
+; GFX11-TRUE16-NEXT: v_dual_mov_b32 v1, v4 :: v_dual_mov_b32 v2, v5
+; GFX11-TRUE16-NEXT: buffer_atomic_cmpswap_b32 v[1:2], v3, s[0:3], 0 offen glc
; GFX11-TRUE16-NEXT: s_waitcnt vmcnt(0)
; GFX11-TRUE16-NEXT: buffer_gl1_inv
; GFX11-TRUE16-NEXT: buffer_gl0_inv
-; GFX11-TRUE16-NEXT: v_cmp_eq_u32_e32 vcc_lo, v4, v2
-; GFX11-TRUE16-NEXT: v_mov_b32_e32 v2, v4
+; GFX11-TRUE16-NEXT: v_cmp_eq_u32_e32 vcc_lo, v1, v5
; GFX11-TRUE16-NEXT: s_or_b32 s5, vcc_lo, s5
; GFX11-TRUE16-NEXT: s_delay_alu instid0(SALU_CYCLE_1)
; GFX11-TRUE16-NEXT: s_and_not1_b32 exec_lo, exec_lo, s5
@@ -3128,38 +3137,39 @@ define void @buffer_fat_ptr_agent_atomic_fmax_noret_f16__offset__amdgpu_no_fine_
; GFX11-FAKE16: ; %bb.0:
; GFX11-FAKE16-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX11-FAKE16-NEXT: s_addk_i32 s16, 0x200
-; GFX11-FAKE16-NEXT: v_max_f16_e32 v3, v0, v0
+; GFX11-FAKE16-NEXT: v_max_f16_e32 v0, v0, v0
; GFX11-FAKE16-NEXT: s_and_b32 s4, s16, -4
; GFX11-FAKE16-NEXT: s_delay_alu instid0(SALU_CYCLE_1) | instskip(SKIP_1) | instid1(SALU_CYCLE_1)
-; GFX11-FAKE16-NEXT: v_mov_b32_e32 v2, s4
+; GFX11-FAKE16-NEXT: v_mov_b32_e32 v3, s4
; GFX11-FAKE16-NEXT: s_and_b32 s4, s16, 3
; GFX11-FAKE16-NEXT: s_lshl_b32 s4, s4, 3
; GFX11-FAKE16-NEXT: s_delay_alu instid0(SALU_CYCLE_1)
; GFX11-FAKE16-NEXT: s_lshl_b32 s5, 0xffff, s4
-; GFX11-FAKE16-NEXT: buffer_load_b32 v1, v2, s[0:3], 0 offen
+; GFX11-FAKE16-NEXT: buffer_load_b32 v1, v3, s[0:3], 0 offen
; GFX11-FAKE16-NEXT: s_not_b32 s6, s5
; GFX11-FAKE16-NEXT: s_mov_b32 s5, 0
; GFX11-FAKE16-NEXT: .p2align 6
; GFX11-FAKE16-NEXT: .LBB11_1: ; %atomicrmw.start
; GFX11-FAKE16-NEXT: ; =>This Inner Loop Header: Depth=1
; GFX11-FAKE16-NEXT: s_waitcnt vmcnt(0)
-; GFX11-FAKE16-NEXT: v_lshrrev_b32_e32 v0, s4, v1
+; GFX11-FAKE16-NEXT: v_mov_b32_e32 v5, v1
; GFX11-FAKE16-NEXT: s_waitcnt_vscnt null, 0x0
; GFX11-FAKE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
-; GFX11-FAKE16-NEXT: v_max_f16_e32 v0, v0, v0
-; GFX11-FAKE16-NEXT: v_max_f16_e32 v0, v0, v3
+; GFX11-FAKE16-NEXT: v_lshrrev_b32_e32 v1, s4, v5
+; GFX11-FAKE16-NEXT: v_max_f16_e32 v1, v1, v1
; GFX11-FAKE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
-; GFX11-FAKE16-NEXT: v_and_b32_e32 v0, 0xffff, v0
-; GFX11-FAKE16-NEXT: v_lshlrev_b32_e32 v0, s4, v0
+; GFX11-FAKE16-NEXT: v_max_f16_e32 v1, v1, v0
+; GFX11-FAKE16-NEXT: v_and_b32_e32 v1, 0xffff, v1
; GFX11-FAKE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
-; GFX11-FAKE16-NEXT: v_and_or_b32 v0, v1, s6, v0
-; GFX11-FAKE16-NEXT: v_dual_mov_b32 v5, v1 :: v_dual_mov_b32 v4, v0
-; GFX11-FAKE16-NEXT: buffer_atomic_cmpswap_b32 v[4:5], v2, s[0:3], 0 offen glc
+; GFX11-FAKE16-NEXT: v_lshlrev_b32_e32 v1, s4, v1
+; GFX11-FAKE16-NEXT: v_and_or_b32 v4, v5, s6, v1
+; GFX11-FAKE16-NEXT: s_delay_alu instid0(VALU_DEP_1)
+; GFX11-FAKE16-NEXT: v_dual_mov_b32 v1, v4 :: v_dual_mov_b32 v2, v5
+; GFX11-FAKE16-NEXT: buffer_atomic_cmpswap_b32 v[1:2], v3, s[0:3], 0 offen glc
; GFX11-FAKE16-NEXT: s_waitcnt vmcnt(0)
; GFX11-FAKE16-NEXT: buffer_gl1_inv
; GFX11-FAKE16-NEXT: buffer_gl0_inv
-; GFX11-FAKE16-NEXT: v_cmp_eq_u32_e32 vcc_lo, v4, v1
-; GFX11-FAKE16-NEXT: v_mov_b32_e32 v1, v4
+; GFX11-FAKE16-NEXT: v_cmp_eq_u32_e32 vcc_lo, v1, v5
; GFX11-FAKE16-NEXT: s_or_b32 s5, vcc_lo, s5
; GFX11-FAKE16-NEXT: s_delay_alu instid0(SALU_CYCLE_1)
; GFX11-FAKE16-NEXT: s_and_not1_b32 exec_lo, exec_lo, s5
@@ -3172,32 +3182,32 @@ define void @buffer_fat_ptr_agent_atomic_fmax_noret_f16__offset__amdgpu_no_fine_
; GFX10: ; %bb.0:
; GFX10-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX10-NEXT: s_addk_i32 s20, 0x200
-; GFX10-NEXT: v_max_f16_e32 v3, v0, v0
+; GFX10-NEXT: v_max_f16_e32 v0, v0, v0
; GFX10-NEXT: s_and_b32 s4, s20, -4
-; GFX10-NEXT: v_mov_b32_e32 v2, s4
+; GFX10-NEXT: v_mov_b32_e32 v3, s4
; GFX10-NEXT: s_and_b32 s4, s20, 3
; GFX10-NEXT: s_lshl_b32 s4, s4, 3
; GFX10-NEXT: s_lshl_b32 s5, 0xffff, s4
-; GFX10-NEXT: buffer_load_dword v1, v2, s[16:19], 0 offen
+; GFX10-NEXT: buffer_load_dword v1, v3, s[16:19], 0 offen
; GFX10-NEXT: s_not_b32 s6, s5
; GFX10-NEXT: s_mov_b32 s5, 0
; GFX10-NEXT: .LBB11_1: ; %atomicrmw.start
; GFX10-NEXT: ; =>This Inner Loop Header: Depth=1
; GFX10-NEXT: s_waitcnt vmcnt(0)
-; GFX10-NEXT: v_lshrrev_b32_e32 v0, s4, v1
-; GFX10-NEXT: s_waitcnt_vscnt null, 0x0
-; GFX10-NEXT: v_max_f16_e32 v0, v0, v0
-; GFX10-NEXT: v_max_f16_e32 v0, v0, v3
-; GFX10-NEXT: v_lshlrev_b32_sdwa v0, s4, v0 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:WORD_0
-; GFX10-NEXT: v_and_or_b32 v0, v1, s6, v0
; GFX10-NEXT: v_mov_b32_e32 v5, v1
-; GFX10-NEXT: v_mov_b32_e32 v4, v0
-; GFX10-NEXT: buffer_atomic_cmpswap v[4:5], v2, s[16:19], 0 offen glc
+; GFX10-NEXT: s_waitcnt_vscnt null, 0x0
+; GFX10-NEXT: v_lshrrev_b32_e32 v1, s4, v5
+; GFX10-NEXT: v_max_f16_e32 v1, v1, v1
+; GFX10-NEXT: v_max_f16_e32 v1, v1, v0
+; GFX10-NEXT: v_lshlrev_b32_sdwa v1, s4, v1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:WORD_0
+; GFX10-NEXT: v_and_or_b32 v4, v5, s6, v1
+; GFX10-NEXT: v_mov_b32_e32 v1, v4
+; GFX10-NEXT: v_mov_b32_e32 v2, v5
+; GFX10-NEXT: buffer_atomic_cmpswap v[1:2], v3, s[16:19], 0 offen glc
; GFX10-NEXT: s_waitcnt vmcnt(0)
; GFX10-NEXT: buffer_gl1_inv
; GFX10-NEXT: buffer_gl0_inv
-; GFX10-NEXT: v_cmp_eq_u32_e32 vcc_lo, v4, v1
-; GFX10-NEXT: v_mov_b32_e32 v1, v4
+; GFX10-NEXT: v_cmp_eq_u32_e32 vcc_lo, v1, v5
; GFX10-NEXT: s_or_b32 s5, vcc_lo, s5
; GFX10-NEXT: s_andn2_b32 exec_lo, exec_lo, s5
; GFX10-NEXT: s_cbranch_execnz .LBB11_1
@@ -3210,29 +3220,29 @@ define void @buffer_fat_ptr_agent_atomic_fmax_noret_f16__offset__amdgpu_no_fine_
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX90A-NEXT: s_addk_i32 s20, 0x200
; GFX90A-NEXT: s_and_b32 s4, s20, -4
-; GFX90A-NEXT: v_mov_b32_e32 v2, s4
-; GFX90A-NEXT: buffer_load_dword v1, v2, s[16:19], 0 offen
+; GFX90A-NEXT: v_mov_b32_e32 v1, s4
+; GFX90A-NEXT: buffer_load_dword v2, v1, s[16:19], 0 offen
; GFX90A-NEXT: s_and_b32 s4, s20, 3
; GFX90A-NEXT: s_lshl_b32 s6, s4, 3
; GFX90A-NEXT: s_lshl_b32 s4, 0xffff, s6
; GFX90A-NEXT: s_not_b32 s7, s4
; GFX90A-NEXT: s_mov_b64 s[4:5], 0
-; GFX90A-NEXT: v_max_f16_e32 v3, v0, v0
+; GFX90A-NEXT: v_max_f16_e32 v0, v0, v0
; GFX90A-NEXT: .LBB11_1: ; %atomicrmw.start
; GFX90A-NEXT: ; =>This Inner Loop Header: Depth=1
; GFX90A-NEXT: s_waitcnt vmcnt(0)
-; GFX90A-NEXT: v_lshrrev_b32_e32 v0, s6, v1
-; GFX90A-NEXT: v_max_f16_e32 v0, v0, v0
-; GFX90A-NEXT: v_max_f16_e32 v0, v0, v3
-; GFX90A-NEXT: v_lshlrev_b32_e32 v0, s6, v0
-; GFX90A-NEXT: v_and_or_b32 v0, v1, s7, v0
-; GFX90A-NEXT: v_pk_mov_b32 v[4:5], v[0:1], v[0:1] op_sel:[0,1]
-; GFX90A-NEXT: buffer_atomic_cmpswap v[4:5], v2, s[16:19], 0 offen glc
+; GFX90A-NEXT: v_mov_b32_e32 v5, v2
+; GFX90A-NEXT: v_lshrrev_b32_e32 v2, s6, v5
+; GFX90A-NEXT: v_max_f16_e32 v2, v2, v2
+; GFX90A-NEXT: v_max_f16_e32 v2, v2, v0
+; GFX90A-NEXT: v_lshlrev_b32_e32 v2, s6, v2
+; GFX90A-NEXT: v_and_or_b32 v4, v5, s7, v2
+; GFX90A-NEXT: v_pk_mov_b32 v[2:3], v[4:5], v[4:5] op_sel:[0,1]
+; GFX90A-NEXT: buffer_atomic_cmpswap v[2:3], v1, s[16:19], 0 offen glc
; GFX90A-NEXT: s_waitcnt vmcnt(0)
; GFX90A-NEXT: buffer_wbinvl1
-; GFX90A-NEXT: v_cmp_eq_u32_e32 vcc, v4, v1
+; GFX90A-NEXT: v_cmp_eq_u32_e32 vcc, v2, v5
; GFX90A-NEXT: s_or_b64 s[4:5], vcc, s[4:5]
-; GFX90A-NEXT: v_mov_b32_e32 v1, v4
; GFX90A-NEXT: s_andn2_b64 exec, exec, s[4:5]
; GFX90A-NEXT: s_cbranch_execnz .LBB11_1
; GFX90A-NEXT: ; %bb.2: ; %atomicrmw.end
@@ -3244,30 +3254,30 @@ define void @buffer_fat_ptr_agent_atomic_fmax_noret_f16__offset__amdgpu_no_fine_
; GFX908-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX908-NEXT: s_addk_i32 s20, 0x200
; GFX908-NEXT: s_and_b32 s4, s20, -4
-; GFX908-NEXT: v_mov_b32_e32 v2, s4
-; GFX908-NEXT: buffer_load_dword v1, v2, s[16:19], 0 offen
+; GFX908-NEXT: v_mov_b32_e32 v3, s4
+; GFX908-NEXT: buffer_load_dword v1, v3, s[16:19], 0 offen
; GFX908-NEXT: s_and_b32 s4, s20, 3
; GFX908-NEXT: s_lshl_b32 s6, s4, 3
; GFX908-NEXT: s_lshl_b32 s4, 0xffff, s6
; GFX908-NEXT: s_not_b32 s7, s4
; GFX908-NEXT: s_mov_b64 s[4:5], 0
-; GFX908-NEXT: v_max_f16_e32 v3, v0, v0
+; GFX908-NEXT: v_max_f16_e32 v0, v0, v0
; GFX908-NEXT: .LBB11_1: ; %atomicrmw.start
; GFX908-NEXT: ; =>This Inner Loop Header: Depth=1
; GFX908-NEXT: s_waitcnt vmcnt(0)
-; GFX908-NEXT: v_lshrrev_b32_e32 v0, s6, v1
-; GFX908-NEXT: v_max_f16_e32 v0, v0, v0
-; GFX908-NEXT: v_max_f16_e32 v0, v0, v3
-; GFX908-NEXT: v_lshlrev_b32_e32 v0, s6, v0
-; GFX908-NEXT: v_and_or_b32 v0, v1, s7, v0
; GFX908-NEXT: v_mov_b32_e32 v5, v1
-; GFX908-NEXT: v_mov_b32_e32 v4, v0
-; GFX908-NEXT: buffer_atomic_cmpswap v[4:5], v2, s[16:19], 0 offen glc
+; GFX908-NEXT: v_lshrrev_b32_e32 v1, s6, v5
+; GFX908-NEXT: v_max_f16_e32 v1, v1, v1
+; GFX908-NEXT: v_max_f16_e32 v1, v1, v0
+; GFX908-NEXT: v_lshlrev_b32_e32 v1, s6, v1
+; GFX908-NEXT: v_and_or_b32 v4, v5, s7, v1
+; GFX908-NEXT: v_mov_b32_e32 v1, v4
+; GFX908-NEXT: v_mov_b32_e32 v2, v5
+; GFX908-NEXT: buffer_atomic_cmpswap v[1:2], v3, s[16:19], 0 offen glc
; GFX908-NEXT: s_waitcnt vmcnt(0)
; GFX908-NEXT: buffer_wbinvl1
-; GFX908-NEXT: v_cmp_eq_u32_e32 vcc, v4, v1
+; GFX908-NEXT: v_cmp_eq_u32_e32 vcc, v1, v5
; GFX908-NEXT: s_or_b64 s[4:5], vcc, s[4:5]
-; GFX908-NEXT: v_mov_b32_e32 v1, v4
; GFX908-NEXT: s_andn2_b64 exec, exec, s[4:5]
; GFX908-NEXT: s_cbranch_execnz .LBB11_1
; GFX908-NEXT: ; %bb.2: ; %atomicrmw.end
@@ -3279,31 +3289,31 @@ define void @buffer_fat_ptr_agent_atomic_fmax_noret_f16__offset__amdgpu_no_fine_
; GFX8-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX8-NEXT: s_addk_i32 s20, 0x200
; GFX8-NEXT: s_and_b32 s4, s20, -4
-; GFX8-NEXT: v_mov_b32_e32 v2, s4
-; GFX8-NEXT: buffer_load_dword v1, v2, s[16:19], 0 offen
+; GFX8-NEXT: v_mov_b32_e32 v3, s4
+; GFX8-NEXT: buffer_load_dword v1, v3, s[16:19], 0 offen
; GFX8-NEXT: s_and_b32 s4, s20, 3
; GFX8-NEXT: s_lshl_b32 s6, s4, 3
; GFX8-NEXT: s_lshl_b32 s4, 0xffff, s6
; GFX8-NEXT: s_not_b32 s7, s4
; GFX8-NEXT: s_mov_b64 s[4:5], 0
-; GFX8-NEXT: v_max_f16_e32 v3, v0, v0
+; GFX8-NEXT: v_max_f16_e32 v0, v0, v0
; GFX8-NEXT: .LBB11_1: ; %atomicrmw.start
; GFX8-NEXT: ; =>This Inner Loop Header: Depth=1
; GFX8-NEXT: s_waitcnt vmcnt(0)
-; GFX8-NEXT: v_lshrrev_b32_e32 v0, s6, v1
-; GFX8-NEXT: v_max_f16_e32 v0, v0, v0
-; GFX8-NEXT: v_max_f16_e32 v0, v0, v3
-; GFX8-NEXT: v_and_b32_e32 v4, s7, v1
-; GFX8-NEXT: v_lshlrev_b32_e32 v0, s6, v0
-; GFX8-NEXT: v_or_b32_e32 v0, v4, v0
; GFX8-NEXT: v_mov_b32_e32 v5, v1
-; GFX8-NEXT: v_mov_b32_e32 v4, v0
-; GFX8-NEXT: buffer_atomic_cmpswap v[4:5], v2, s[16:19], 0 offen glc
+; GFX8-NEXT: v_lshrrev_b32_e32 v1, s6, v5
+; GFX8-NEXT: v_max_f16_e32 v1, v1, v1
+; GFX8-NEXT: v_max_f16_e32 v1, v1, v0
+; GFX8-NEXT: v_and_b32_e32 v2, s7, v5
+; GFX8-NEXT: v_lshlrev_b32_e32 v1, s6, v1
+; GFX8-NEXT: v_or_b32_e32 v4, v2, v1
+; GFX8-NEXT: v_mov_b32_e32 v1, v4
+; GFX8-NEXT: v_mov_b32_e32 v2, v5
+; GFX8-NEXT: buffer_atomic_cmpswap v[1:2], v3, s[16:19], 0 offen glc
; GFX8-NEXT: s_waitcnt vmcnt(0)
; GFX8-NEXT: buffer_wbinvl1
-; GFX8-NEXT: v_cmp_eq_u32_e32 vcc, v4, v1
+; GFX8-NEXT: v_cmp_eq_u32_e32 vcc, v1, v5
; GFX8-NEXT: s_or_b64 s[4:5], vcc, s[4:5]
-; GFX8-NEXT: v_mov_b32_e32 v1, v4
; GFX8-NEXT: s_andn2_b64 exec, exec, s[4:5]
; GFX8-NEXT: s_cbranch_execnz .LBB11_1
; GFX8-NEXT: ; %bb.2: ; %atomicrmw.end
@@ -3315,33 +3325,33 @@ define void @buffer_fat_ptr_agent_atomic_fmax_noret_f16__offset__amdgpu_no_fine_
; GFX7-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX7-NEXT: s_addk_i32 s20, 0x200
; GFX7-NEXT: s_and_b32 s4, s20, -4
-; GFX7-NEXT: v_mov_b32_e32 v2, s4
-; GFX7-NEXT: buffer_load_dword v1, v2, s[16:19], 0 offen
+; GFX7-NEXT: v_mov_b32_e32 v3, s4
+; GFX7-NEXT: buffer_load_dword v1, v3, s[16:19], 0 offen
; GFX7-NEXT: v_cvt_f16_f32_e32 v0, v0
; GFX7-NEXT: s_and_b32 s4, s20, 3
; GFX7-NEXT: s_lshl_b32 s6, s4, 3
; GFX7-NEXT: s_lshl_b32 s4, 0xffff, s6
-; GFX7-NEXT: v_cvt_f32_f16_e32 v3, v0
+; GFX7-NEXT: v_cvt_f32_f16_e32 v0, v0
; GFX7-NEXT: s_not_b32 s7, s4
; GFX7-NEXT: s_mov_b64 s[4:5], 0
; GFX7-NEXT: .LBB11_1: ; %atomicrmw.start
; GFX7-NEXT: ; =>This Inner Loop Header: Depth=1
; GFX7-NEXT: s_waitcnt vmcnt(0)
-; GFX7-NEXT: v_lshrrev_b32_e32 v0, s6, v1
-; GFX7-NEXT: v_cvt_f32_f16_e32 v0, v0
-; GFX7-NEXT: v_and_b32_e32 v4, s7, v1
-; GFX7-NEXT: v_max_f32_e32 v0, v0, v3
-; GFX7-NEXT: v_cvt_f16_f32_e32 v0, v0
-; GFX7-NEXT: v_lshlrev_b32_e32 v0, s6, v0
-; GFX7-NEXT: v_or_b32_e32 v0, v4, v0
; GFX7-NEXT: v_mov_b32_e32 v5, v1
-; GFX7-NEXT: v_mov_b32_e32 v4, v0
-; GFX7-NEXT: buffer_atomic_cmpswap v[4:5], v2, s[16:19], 0 offen glc
+; GFX7-NEXT: v_lshrrev_b32_e32 v1, s6, v5
+; GFX7-NEXT: v_cvt_f32_f16_e32 v1, v1
+; GFX7-NEXT: v_and_b32_e32 v2, s7, v5
+; GFX7-NEXT: v_max_f32_e32 v1, v1, v0
+; GFX7-NEXT: v_cvt_f16_f32_e32 v1, v1
+; GFX7-NEXT: v_lshlrev_b32_e32 v1, s6, v1
+; GFX7-NEXT: v_or_b32_e32 v4, v2, v1
+; GFX7-NEXT: v_mov_b32_e32 v1, v4
+; GFX7-NEXT: v_mov_b32_e32 v2, v5
+; GFX7-NEXT: buffer_atomic_cmpswap v[1:2], v3, s[16:19], 0 offen glc
; GFX7-NEXT: s_waitcnt vmcnt(0)
; GFX7-NEXT: buffer_wbinvl1
-; GFX7-NEXT: v_cmp_eq_u32_e32 vcc, v4, v1
+; GFX7-NEXT: v_cmp_eq_u32_e32 vcc, v1, v5
; GFX7-NEXT: s_or_b64 s[4:5], vcc, s[4:5]
-; GFX7-NEXT: v_mov_b32_e32 v1, v4
; GFX7-NEXT: s_andn2_b64 exec, exec, s[4:5]
; GFX7-NEXT: s_cbranch_execnz .LBB11_1
; GFX7-NEXT: ; %bb.2: ; %atomicrmw.end
@@ -3353,34 +3363,34 @@ define void @buffer_fat_ptr_agent_atomic_fmax_noret_f16__offset__amdgpu_no_fine_
; GFX6-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX6-NEXT: s_addk_i32 s20, 0x200
; GFX6-NEXT: s_and_b32 s4, s20, -4
-; GFX6-NEXT: v_mov_b32_e32 v2, s4
-; GFX6-NEXT: buffer_load_dword v1, v2, s[16:19], 0 offen
+; GFX6-NEXT: v_mov_b32_e32 v3, s4
+; GFX6-NEXT: buffer_load_dword v1, v3, s[16:19], 0 offen
; GFX6-NEXT: v_cvt_f16_f32_e32 v0, v0
; GFX6-NEXT: s_and_b32 s4, s20, 3
; GFX6-NEXT: s_lshl_b32 s6, s4, 3
; GFX6-NEXT: s_lshl_b32 s4, 0xffff, s6
-; GFX6-NEXT: v_cvt_f32_f16_e32 v3, v0
+; GFX6-NEXT: v_cvt_f32_f16_e32 v0, v0
; GFX6-NEXT: s_not_b32 s7, s4
; GFX6-NEXT: s_mov_b64 s[4:5], 0
; GFX6-NEXT: .LBB11_1: ; %atomicrmw.start
; GFX6-NEXT: ; =>This Inner Loop Header: Depth=1
; GFX6-NEXT: s_waitcnt vmcnt(0)
-; GFX6-NEXT: v_lshrrev_b32_e32 v0, s6, v1
-; GFX6-NEXT: v_cvt_f32_f16_e32 v0, v0
-; GFX6-NEXT: s_waitcnt expcnt(0)
-; GFX6-NEXT: v_and_b32_e32 v4, s7, v1
-; GFX6-NEXT: v_max_f32_e32 v0, v0, v3
-; GFX6-NEXT: v_cvt_f16_f32_e32 v0, v0
-; GFX6-NEXT: v_lshlrev_b32_e32 v0, s6, v0
-; GFX6-NEXT: v_or_b32_e32 v0, v4, v0
; GFX6-NEXT: v_mov_b32_e32 v5, v1
-; GFX6-NEXT: v_mov_b32_e32 v4, v0
-; GFX6-NEXT: buffer_atomic_cmpswap v[4:5], v2, s[16:19], 0 offen glc
+; GFX6-NEXT: s_waitcnt expcnt(0)
+; GFX6-NEXT: v_lshrrev_b32_e32 v1, s6, v5
+; GFX6-NEXT: v_cvt_f32_f16_e32 v1, v1
+; GFX6-NEXT: v_and_b32_e32 v2, s7, v5
+; GFX6-NEXT: v_max_f32_e32 v1, v1, v0
+; GFX6-NEXT: v_cvt_f16_f32_e32 v1, v1
+; GFX6-NEXT: v_lshlrev_b32_e32 v1, s6, v1
+; GFX6-NEXT: v_or_b32_e32 v4, v2, v1
+; GFX6-NEXT: v_mov_b32_e32 v1, v4
+; GFX6-NEXT: v_mov_b32_e32 v2, v5
+; GFX6-NEXT: buffer_atomic_cmpswap v[1:2], v3, s[16:19], 0 offen glc
; GFX6-NEXT: s_waitcnt vmcnt(0)
; GFX6-NEXT: buffer_wbinvl1
-; GFX6-NEXT: v_cmp_eq_u32_e32 vcc, v4, v1
+; GFX6-NEXT: v_cmp_eq_u32_e32 vcc, v1, v5
; GFX6-NEXT: s_or_b64 s[4:5], vcc, s[4:5]
-; GFX6-NEXT: v_mov_b32_e32 v1, v4
; GFX6-NEXT: s_andn2_b64 exec, exec, s[4:5]
; GFX6-NEXT: s_cbranch_execnz .LBB11_1
; GFX6-NEXT: ; %bb.2: ; %atomicrmw.end
@@ -3432,18 +3442,21 @@ define half @buffer_fat_ptr_agent_atomic_fmax_ret_f16__offset__waterfall__amdgpu
; GFX12-TRUE16-NEXT: ; =>This Loop Header: Depth=1
; GFX12-TRUE16-NEXT: ; Child Loop BB12_4 Depth 2
; GFX12-TRUE16-NEXT: s_wait_loadcnt 0x0
-; GFX12-TRUE16-NEXT: v_lshrrev_b32_e32 v5, v9, v6
+; GFX12-TRUE16-NEXT: v_mov_b32_e32 v8, v6
; GFX12-TRUE16-NEXT: s_mov_b32 s2, exec_lo
; GFX12-TRUE16-NEXT: s_wait_storecnt 0x0
; GFX12-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
+; GFX12-TRUE16-NEXT: v_lshrrev_b32_e32 v5, v9, v8
; GFX12-TRUE16-NEXT: v_max_num_f16_e32 v4.h, v5.l, v5.l
-; GFX12-TRUE16-NEXT: v_max_num_f16_e32 v5.l, v4.h, v4.l
; GFX12-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
+; GFX12-TRUE16-NEXT: v_max_num_f16_e32 v5.l, v4.h, v4.l
; GFX12-TRUE16-NEXT: v_and_b32_e32 v5, 0xffff, v5
-; GFX12-TRUE16-NEXT: v_lshlrev_b32_e32 v5, v9, v5
; GFX12-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
-; GFX12-TRUE16-NEXT: v_and_or_b32 v5, v6, v11, v5
-; GFX12-TRUE16-NEXT: v_dual_mov_b32 v8, v6 :: v_dual_mov_b32 v7, v5
+; GFX12-TRUE16-NEXT: v_lshlrev_b32_e32 v5, v9, v5
+; GFX12-TRUE16-NEXT: v_and_or_b32 v7, v8, v11, v5
+; GFX12-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1)
+; GFX12-TRUE16-NEXT: v_mov_b32_e32 v6, v7
+; GFX12-TRUE16-NEXT: v_mov_b32_e32 v7, v8
; GFX12-TRUE16-NEXT: .LBB12_4: ; Parent Loop BB12_3 Depth=1
; GFX12-TRUE16-NEXT: ; => This Inner Loop Header: Depth=2
; GFX12-TRUE16-NEXT: v_readfirstlane_b32 s4, v0
@@ -3457,14 +3470,13 @@ define half @buffer_fat_ptr_agent_atomic_fmax_ret_f16__offset__waterfall__amdgpu
; GFX12-TRUE16-NEXT: s_wait_alu 0xfffe
; GFX12-TRUE16-NEXT: s_and_saveexec_b32 s0, s0
; GFX12-TRUE16-NEXT: s_wait_loadcnt 0x0
-; GFX12-TRUE16-NEXT: buffer_atomic_cmpswap_b32 v[7:8], v10, s[4:7], null offen th:TH_ATOMIC_RETURN
+; GFX12-TRUE16-NEXT: buffer_atomic_cmpswap_b32 v[6:7], v10, s[4:7], null offen th:TH_ATOMIC_RETURN
; GFX12-TRUE16-NEXT: s_xor_b32 exec_lo, exec_lo, s0
; GFX12-TRUE16-NEXT: s_cbranch_execnz .LBB12_4
; GFX12-TRUE16-NEXT: ; %bb.5: ; in Loop: Header=BB12_3 Depth=1
; GFX12-TRUE16-NEXT: s_mov_b32 exec_lo, s2
; GFX12-TRUE16-NEXT: s_wait_loadcnt 0x0
-; GFX12-TRUE16-NEXT: v_cmp_eq_u32_e32 vcc_lo, v7, v6
-; GFX12-TRUE16-NEXT: v_mov_b32_e32 v6, v7
+; GFX12-TRUE16-NEXT: v_cmp_eq_u32_e32 vcc_lo, v6, v8
; GFX12-TRUE16-NEXT: global_inv scope:SCOPE_DEV
; GFX12-TRUE16-NEXT: s_or_b32 s1, vcc_lo, s1
; GFX12-TRUE16-NEXT: s_wait_alu 0xfffe
@@ -3472,7 +3484,7 @@ define half @buffer_fat_ptr_agent_atomic_fmax_ret_f16__offset__waterfall__amdgpu
; GFX12-TRUE16-NEXT: s_cbranch_execnz .LBB12_3
; GFX12-TRUE16-NEXT: ; %bb.6: ; %atomicrmw.end
; GFX12-TRUE16-NEXT: s_or_b32 exec_lo, exec_lo, s1
-; GFX12-TRUE16-NEXT: v_lshrrev_b32_e32 v0, v9, v7
+; GFX12-TRUE16-NEXT: v_lshrrev_b32_e32 v0, v9, v6
; GFX12-TRUE16-NEXT: s_wait_loadcnt 0x0
; GFX12-TRUE16-NEXT: s_setpc_b64 s[30:31]
;
@@ -3504,7 +3516,7 @@ define half @buffer_fat_ptr_agent_atomic_fmax_ret_f16__offset__waterfall__amdgpu
; GFX12-FAKE16-NEXT: s_wait_alu 0xfffe
; GFX12-FAKE16-NEXT: s_and_saveexec_b32 s0, s0
; GFX12-FAKE16-NEXT: s_wait_loadcnt 0x0
-; GFX12-FAKE16-NEXT: buffer_load_b32 v6, v8, s[4:7], null offen
+; GFX12-FAKE16-NEXT: buffer_load_b32 v4, v8, s[4:7], null offen
; GFX12-FAKE16-NEXT: s_xor_b32 exec_lo, exec_lo, s0
; GFX12-FAKE16-NEXT: s_cbranch_execnz .LBB12_1
; GFX12-FAKE16-NEXT: ; %bb.2:
@@ -3515,17 +3527,19 @@ define half @buffer_fat_ptr_agent_atomic_fmax_ret_f16__offset__waterfall__amdgpu
; GFX12-FAKE16-NEXT: ; =>This Loop Header: Depth=1
; GFX12-FAKE16-NEXT: ; Child Loop BB12_4 Depth 2
; GFX12-FAKE16-NEXT: s_wait_loadcnt 0x0
-; GFX12-FAKE16-NEXT: v_lshrrev_b32_e32 v4, v7, v6
+; GFX12-FAKE16-NEXT: v_mov_b32_e32 v6, v4
; GFX12-FAKE16-NEXT: s_mov_b32 s2, exec_lo
; GFX12-FAKE16-NEXT: s_wait_storecnt 0x0
; GFX12-FAKE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
+; GFX12-FAKE16-NEXT: v_lshrrev_b32_e32 v4, v7, v6
; GFX12-FAKE16-NEXT: v_max_num_f16_e32 v4, v4, v4
-; GFX12-FAKE16-NEXT: v_max_num_f16_e32 v4, v4, v10
; GFX12-FAKE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
+; GFX12-FAKE16-NEXT: v_max_num_f16_e32 v4, v4, v10
; GFX12-FAKE16-NEXT: v_and_b32_e32 v4, 0xffff, v4
-; GFX12-FAKE16-NEXT: v_lshlrev_b32_e32 v4, v7, v4
; GFX12-FAKE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
+; GFX12-FAKE16-NEXT: v_lshlrev_b32_e32 v4, v7, v4
; GFX12-FAKE16-NEXT: v_and_or_b32 v5, v6, v9, v4
+; GFX12-FAKE16-NEXT: s_delay_alu instid0(VALU_DEP_1)
; GFX12-FAKE16-NEXT: v_mov_b32_e32 v4, v5
; GFX12-FAKE16-NEXT: v_mov_b32_e32 v5, v6
; GFX12-FAKE16-NEXT: .LBB12_4: ; Parent Loop BB12_3 Depth=1
@@ -3548,7 +3562,6 @@ define half @buffer_fat_ptr_agent_atomic_fmax_ret_f16__offset__waterfall__amdgpu
; GFX12-FAKE16-NEXT: s_mov_b32 exec_lo, s2
; GFX12-FAKE16-NEXT: s_wait_loadcnt 0x0
; GFX12-FAKE16-NEXT: v_cmp_eq_u32_e32 vcc_lo, v4, v6
-; GFX12-FAKE16-NEXT: v_mov_b32_e32 v6, v4
; GFX12-FAKE16-NEXT: global_inv scope:SCOPE_DEV
; GFX12-FAKE16-NEXT: s_or_b32 s1, vcc_lo, s1
; GFX12-FAKE16-NEXT: s_wait_alu 0xfffe
@@ -3581,7 +3594,7 @@ define half @buffer_fat_ptr_agent_atomic_fmax_ret_f16__offset__waterfall__amdgpu
; GFX942-NEXT: v_cmp_eq_u64_e64 s[0:1], s[6:7], v[2:3]
; GFX942-NEXT: s_and_b64 s[0:1], vcc, s[0:1]
; GFX942-NEXT: s_and_saveexec_b64 s[0:1], s[0:1]
-; GFX942-NEXT: buffer_load_dword v7, v9, s[4:7], 0 offen
+; GFX942-NEXT: buffer_load_dword v4, v9, s[4:7], 0 offen
; GFX942-NEXT: s_xor_b64 exec, exec, s[0:1]
; GFX942-NEXT: s_cbranch_execnz .LBB12_1
; GFX942-NEXT: ; %bb.2:
@@ -3592,6 +3605,7 @@ define half @buffer_fat_ptr_agent_atomic_fmax_ret_f16__offset__waterfall__amdgpu
; GFX942-NEXT: ; =>This Loop Header: Depth=1
; GFX942-NEXT: ; Child Loop BB12_4 Depth 2
; GFX942-NEXT: s_waitcnt vmcnt(0)
+; GFX942-NEXT: v_mov_b32_e32 v7, v4
; GFX942-NEXT: v_lshrrev_b32_e32 v4, v8, v7
; GFX942-NEXT: v_max_f16_e32 v4, v4, v4
; GFX942-NEXT: v_max_f16_e32 v4, v4, v11
@@ -3620,7 +3634,6 @@ define half @buffer_fat_ptr_agent_atomic_fmax_ret_f16__offset__waterfall__amdgpu
; GFX942-NEXT: s_waitcnt vmcnt(0)
; GFX942-NEXT: v_cmp_eq_u32_e32 vcc, v4, v7
; GFX942-NEXT: s_or_b64 s[2:3], vcc, s[2:3]
-; GFX942-NEXT: v_mov_b32_e32 v7, v4
; GFX942-NEXT: buffer_inv sc1
; GFX942-NEXT: s_andn2_b64 exec, exec, s[2:3]
; GFX942-NEXT: s_cbranch_execnz .LBB12_3
@@ -3664,18 +3677,21 @@ define half @buffer_fat_ptr_agent_atomic_fmax_ret_f16__offset__waterfall__amdgpu
; GFX11-TRUE16-NEXT: ; =>This Loop Header: Depth=1
; GFX11-TRUE16-NEXT: ; Child Loop BB12_4 Depth 2
; GFX11-TRUE16-NEXT: s_waitcnt vmcnt(0)
-; GFX11-TRUE16-NEXT: v_lshrrev_b32_e32 v5, v9, v6
+; GFX11-TRUE16-NEXT: v_mov_b32_e32 v8, v6
; GFX11-TRUE16-NEXT: s_mov_b32 s2, exec_lo
; GFX11-TRUE16-NEXT: s_waitcnt_vscnt null, 0x0
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
+; GFX11-TRUE16-NEXT: v_lshrrev_b32_e32 v5, v9, v8
; GFX11-TRUE16-NEXT: v_max_f16_e32 v4.h, v5.l, v5.l
-; GFX11-TRUE16-NEXT: v_max_f16_e32 v5.l, v4.h, v4.l
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
+; GFX11-TRUE16-NEXT: v_max_f16_e32 v5.l, v4.h, v4.l
; GFX11-TRUE16-NEXT: v_and_b32_e32 v5, 0xffff, v5
-; GFX11-TRUE16-NEXT: v_lshlrev_b32_e32 v5, v9, v5
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
-; GFX11-TRUE16-NEXT: v_and_or_b32 v5, v6, v11, v5
-; GFX11-TRUE16-NEXT: v_dual_mov_b32 v8, v6 :: v_dual_mov_b32 v7, v5
+; GFX11-TRUE16-NEXT: v_lshlrev_b32_e32 v5, v9, v5
+; GFX11-TRUE16-NEXT: v_and_or_b32 v7, v8, v11, v5
+; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1)
+; GFX11-TRUE16-NEXT: v_mov_b32_e32 v6, v7
+; GFX11-TRUE16-NEXT: v_mov_b32_e32 v7, v8
; GFX11-TRUE16-NEXT: .LBB12_4: ; Parent Loop BB12_3 Depth=1
; GFX11-TRUE16-NEXT: ; => This Inner Loop Header: Depth=2
; GFX11-TRUE16-NEXT: v_readfirstlane_b32 s4, v0
@@ -3689,14 +3705,13 @@ define half @buffer_fat_ptr_agent_atomic_fmax_ret_f16__offset__waterfall__amdgpu
; GFX11-TRUE16-NEXT: s_delay_alu instid0(SALU_CYCLE_1)
; GFX11-TRUE16-NEXT: s_and_saveexec_b32 s0, s0
; GFX11-TRUE16-NEXT: s_waitcnt vmcnt(0)
-; GFX11-TRUE16-NEXT: buffer_atomic_cmpswap_b32 v[7:8], v10, s[4:7], 0 offen glc
+; GFX11-TRUE16-NEXT: buffer_atomic_cmpswap_b32 v[6:7], v10, s[4:7], 0 offen glc
; GFX11-TRUE16-NEXT: s_xor_b32 exec_lo, exec_lo, s0
; GFX11-TRUE16-NEXT: s_cbranch_execnz .LBB12_4
; GFX11-TRUE16-NEXT: ; %bb.5: ; in Loop: Header=BB12_3 Depth=1
; GFX11-TRUE16-NEXT: s_mov_b32 exec_lo, s2
; GFX11-TRUE16-NEXT: s_waitcnt vmcnt(0)
-; GFX11-TRUE16-NEXT: v_cmp_eq_u32_e32 vcc_lo, v7, v6
-; GFX11-TRUE16-NEXT: v_mov_b32_e32 v6, v7
+; GFX11-TRUE16-NEXT: v_cmp_eq_u32_e32 vcc_lo, v6, v8
; GFX11-TRUE16-NEXT: buffer_gl1_inv
; GFX11-TRUE16-NEXT: buffer_gl0_inv
; GFX11-TRUE16-NEXT: s_or_b32 s1, vcc_lo, s1
@@ -3705,7 +3720,7 @@ define half @buffer_fat_ptr_agent_atomic_fmax_ret_f16__offset__waterfall__amdgpu
; GFX11-TRUE16-NEXT: s_cbranch_execnz .LBB12_3
; GFX11-TRUE16-NEXT: ; %bb.6: ; %atomicrmw.end
; GFX11-TRUE16-NEXT: s_or_b32 exec_lo, exec_lo, s1
-; GFX11-TRUE16-NEXT: v_lshrrev_b32_e32 v0, v9, v7
+; GFX11-TRUE16-NEXT: v_lshrrev_b32_e32 v0, v9, v6
; GFX11-TRUE16-NEXT: s_setpc_b64 s[30:31]
;
; GFX11-FAKE16-LABEL: buffer_fat_ptr_agent_atomic_fmax_ret_f16__offset__waterfall__amdgpu_no_fine_grained_memory:
@@ -3732,7 +3747,7 @@ define half @buffer_fat_ptr_agent_atomic_fmax_ret_f16__offset__waterfall__amdgpu
; GFX11-FAKE16-NEXT: s_and_b32 s0, vcc_lo, s0
; GFX11-FAKE16-NEXT: s_delay_alu instid0(SALU_CYCLE_1)
; GFX11-FAKE16-NEXT: s_and_saveexec_b32 s0, s0
-; GFX11-FAKE16-NEXT: buffer_load_b32 v6, v8, s[4:7], 0 offen
+; GFX11-FAKE16-NEXT: buffer_load_b32 v4, v8, s[4:7], 0 offen
; GFX11-FAKE16-NEXT: s_xor_b32 exec_lo, exec_lo, s0
; GFX11-FAKE16-NEXT: s_cbranch_execnz .LBB12_1
; GFX11-FAKE16-NEXT: ; %bb.2:
@@ -3743,17 +3758,19 @@ define half @buffer_fat_ptr_agent_atomic_fmax_ret_f16__offset__waterfall__amdgpu
; GFX11-FAKE16-NEXT: ; =>This Loop Header: Depth=1
; GFX11-FAKE16-NEXT: ; Child Loop BB12_4 Depth 2
; GFX11-FAKE16-NEXT: s_waitcnt vmcnt(0)
-; GFX11-FAKE16-NEXT: v_lshrrev_b32_e32 v4, v7, v6
+; GFX11-FAKE16-NEXT: v_mov_b32_e32 v6, v4
; GFX11-FAKE16-NEXT: s_mov_b32 s2, exec_lo
; GFX11-FAKE16-NEXT: s_waitcnt_vscnt null, 0x0
; GFX11-FAKE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
+; GFX11-FAKE16-NEXT: v_lshrrev_b32_e32 v4, v7, v6
; GFX11-FAKE16-NEXT: v_max_f16_e32 v4, v4, v4
-; GFX11-FAKE16-NEXT: v_max_f16_e32 v4, v4, v10
; GFX11-FAKE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
+; GFX11-FAKE16-NEXT: v_max_f16_e32 v4, v4, v10
; GFX11-FAKE16-NEXT: v_and_b32_e32 v4, 0xffff, v4
-; GFX11-FAKE16-NEXT: v_lshlrev_b32_e32 v4, v7, v4
; GFX11-FAKE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
+; GFX11-FAKE16-NEXT: v_lshlrev_b32_e32 v4, v7, v4
; GFX11-FAKE16-NEXT: v_and_or_b32 v5, v6, v9, v4
+; GFX11-FAKE16-NEXT: s_delay_alu instid0(VALU_DEP_1)
; GFX11-FAKE16-NEXT: v_mov_b32_e32 v4, v5
; GFX11-FAKE16-NEXT: v_mov_b32_e32 v5, v6
; GFX11-FAKE16-NEXT: .LBB12_4: ; Parent Loop BB12_3 Depth=1
@@ -3776,7 +3793,6 @@ define half @buffer_fat_ptr_agent_atomic_fmax_ret_f16__offset__waterfall__amdgpu
; GFX11-FAKE16-NEXT: s_mov_b32 exec_lo, s2
; GFX11-FAKE16-NEXT: s_waitcnt vmcnt(0)
; GFX11-FAKE16-NEXT: v_cmp_eq_u32_e32 vcc_lo, v4, v6
-; GFX11-FAKE16-NEXT: v_mov_b32_e32 v6, v4
; GFX11-FAKE16-NEXT: buffer_gl1_inv
; GFX11-FAKE16-NEXT: buffer_gl0_inv
; GFX11-FAKE16-NEXT: s_or_b32 s1, vcc_lo, s1
@@ -3808,7 +3824,7 @@ define half @buffer_fat_ptr_agent_atomic_fmax_ret_f16__offset__waterfall__amdgpu
; GFX10-NEXT: v_cmp_eq_u64_e64 s4, s[10:11], v[2:3]
; GFX10-NEXT: s_and_b32 s4, vcc_lo, s4
; GFX10-NEXT: s_and_saveexec_b32 s4, s4
-; GFX10-NEXT: buffer_load_dword v6, v8, s[8:11], 0 offen
+; GFX10-NEXT: buffer_load_dword v4, v8, s[8:11], 0 offen
; GFX10-NEXT: s_waitcnt_depctr 0xffe3
; GFX10-NEXT: s_xor_b32 exec_lo, exec_lo, s4
; GFX10-NEXT: s_cbranch_execnz .LBB12_1
@@ -3819,9 +3835,10 @@ define half @buffer_fat_ptr_agent_atomic_fmax_ret_f16__offset__waterfall__amdgpu
; GFX10-NEXT: ; =>This Loop Header: Depth=1
; GFX10-NEXT: ; Child Loop BB12_4 Depth 2
; GFX10-NEXT: s_waitcnt vmcnt(0)
-; GFX10-NEXT: v_lshrrev_b32_e32 v4, v7, v6
+; GFX10-NEXT: v_mov_b32_e32 v6, v4
; GFX10-NEXT: s_mov_b32 s6, exec_lo
; GFX10-NEXT: s_waitcnt_vscnt null, 0x0
+; GFX10-NEXT: v_lshrrev_b32_e32 v4, v7, v6
; GFX10-NEXT: v_max_f16_e32 v4, v4, v4
; GFX10-NEXT: v_max_f16_e32 v4, v4, v10
; GFX10-NEXT: v_lshlrev_b32_sdwa v4, v7, v4 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:WORD_0
@@ -3847,7 +3864,6 @@ define half @buffer_fat_ptr_agent_atomic_fmax_ret_f16__offset__waterfall__amdgpu
; GFX10-NEXT: s_mov_b32 exec_lo, s6
; GFX10-NEXT: s_waitcnt vmcnt(0)
; GFX10-NEXT: v_cmp_eq_u32_e32 vcc_lo, v4, v6
-; GFX10-NEXT: v_mov_b32_e32 v6, v4
; GFX10-NEXT: buffer_gl1_inv
; GFX10-NEXT: buffer_gl0_inv
; GFX10-NEXT: s_or_b32 s5, vcc_lo, s5
@@ -3880,7 +3896,7 @@ define half @buffer_fat_ptr_agent_atomic_fmax_ret_f16__offset__waterfall__amdgpu
; GFX90A-NEXT: s_and_b64 s[4:5], vcc, s[4:5]
; GFX90A-NEXT: s_and_saveexec_b64 s[4:5], s[4:5]
; GFX90A-NEXT: s_nop 0
-; GFX90A-NEXT: buffer_load_dword v7, v9, s[8:11], 0 offen
+; GFX90A-NEXT: buffer_load_dword v4, v9, s[8:11], 0 offen
; GFX90A-NEXT: s_xor_b64 exec, exec, s[4:5]
; GFX90A-NEXT: s_cbranch_execnz .LBB12_1
; GFX90A-NEXT: ; %bb.2:
@@ -3891,6 +3907,7 @@ define half @buffer_fat_ptr_agent_atomic_fmax_ret_f16__offset__waterfall__amdgpu
; GFX90A-NEXT: ; =>This Loop Header: Depth=1
; GFX90A-NEXT: ; Child Loop BB12_4 Depth 2
; GFX90A-NEXT: s_waitcnt vmcnt(0)
+; GFX90A-NEXT: v_mov_b32_e32 v7, v4
; GFX90A-NEXT: v_lshrrev_b32_e32 v4, v8, v7
; GFX90A-NEXT: v_max_f16_e32 v4, v4, v4
; GFX90A-NEXT: v_max_f16_e32 v4, v4, v11
@@ -3917,7 +3934,6 @@ define half @buffer_fat_ptr_agent_atomic_fmax_ret_f16__offset__waterfall__amdgpu
; GFX90A-NEXT: s_waitcnt vmcnt(0)
; GFX90A-NEXT: v_cmp_eq_u32_e32 vcc, v4, v7
; GFX90A-NEXT: s_or_b64 s[6:7], vcc, s[6:7]
-; GFX90A-NEXT: v_mov_b32_e32 v7, v4
; GFX90A-NEXT: buffer_wbinvl1
; GFX90A-NEXT: s_andn2_b64 exec, exec, s[6:7]
; GFX90A-NEXT: s_cbranch_execnz .LBB12_3
@@ -3947,7 +3963,7 @@ define half @buffer_fat_ptr_agent_atomic_fmax_ret_f16__offset__waterfall__amdgpu
; GFX908-NEXT: s_and_b64 s[4:5], vcc, s[4:5]
; GFX908-NEXT: s_and_saveexec_b64 s[4:5], s[4:5]
; GFX908-NEXT: s_nop 0
-; GFX908-NEXT: buffer_load_dword v6, v8, s[8:11], 0 offen
+; GFX908-NEXT: buffer_load_dword v4, v8, s[8:11], 0 offen
; GFX908-NEXT: s_xor_b64 exec, exec, s[4:5]
; GFX908-NEXT: s_cbranch_execnz .LBB12_1
; GFX908-NEXT: ; %bb.2:
@@ -3958,6 +3974,7 @@ define half @buffer_fat_ptr_agent_atomic_fmax_ret_f16__offset__waterfall__amdgpu
; GFX908-NEXT: ; =>This Loop Header: Depth=1
; GFX908-NEXT: ; Child Loop BB12_4 Depth 2
; GFX908-NEXT: s_waitcnt vmcnt(0)
+; GFX908-NEXT: v_mov_b32_e32 v6, v4
; GFX908-NEXT: v_lshrrev_b32_e32 v4, v7, v6
; GFX908-NEXT: v_max_f16_e32 v4, v4, v4
; GFX908-NEXT: v_max_f16_e32 v4, v4, v10
@@ -3985,7 +4002,6 @@ define half @buffer_fat_ptr_agent_atomic_fmax_ret_f16__offset__waterfall__amdgpu
; GFX908-NEXT: s_waitcnt vmcnt(0)
; GFX908-NEXT: v_cmp_eq_u32_e32 vcc, v4, v6
; GFX908-NEXT: s_or_b64 s[6:7], vcc, s[6:7]
-; GFX908-NEXT: v_mov_b32_e32 v6, v4
; GFX908-NEXT: buffer_wbinvl1
; GFX908-NEXT: s_andn2_b64 exec, exec, s[6:7]
; GFX908-NEXT: s_cbranch_execnz .LBB12_3
@@ -4015,7 +4031,7 @@ define half @buffer_fat_ptr_agent_atomic_fmax_ret_f16__offset__waterfall__amdgpu
; GFX8-NEXT: s_and_b64 s[4:5], vcc, s[4:5]
; GFX8-NEXT: s_and_saveexec_b64 s[4:5], s[4:5]
; GFX8-NEXT: s_nop 0
-; GFX8-NEXT: buffer_load_dword v6, v8, s[8:11], 0 offen
+; GFX8-NEXT: buffer_load_dword v4, v8, s[8:11], 0 offen
; GFX8-NEXT: s_xor_b64 exec, exec, s[4:5]
; GFX8-NEXT: s_cbranch_execnz .LBB12_1
; GFX8-NEXT: ; %bb.2:
@@ -4026,6 +4042,7 @@ define half @buffer_fat_ptr_agent_atomic_fmax_ret_f16__offset__waterfall__amdgpu
; GFX8-NEXT: ; =>This Loop Header: Depth=1
; GFX8-NEXT: ; Child Loop BB12_4 Depth 2
; GFX8-NEXT: s_waitcnt vmcnt(0)
+; GFX8-NEXT: v_mov_b32_e32 v6, v4
; GFX8-NEXT: v_lshrrev_b32_e32 v4, v7, v6
; GFX8-NEXT: v_max_f16_e32 v4, v4, v4
; GFX8-NEXT: v_max_f16_e32 v4, v4, v10
@@ -4054,7 +4071,6 @@ define half @buffer_fat_ptr_agent_atomic_fmax_ret_f16__offset__waterfall__amdgpu
; GFX8-NEXT: s_waitcnt vmcnt(0)
; GFX8-NEXT: v_cmp_eq_u32_e32 vcc, v4, v6
; GFX8-NEXT: s_or_b64 s[6:7], vcc, s[6:7]
-; GFX8-NEXT: v_mov_b32_e32 v6, v4
; GFX8-NEXT: buffer_wbinvl1
; GFX8-NEXT: s_andn2_b64 exec, exec, s[6:7]
; GFX8-NEXT: s_cbranch_execnz .LBB12_3
@@ -4082,18 +4098,19 @@ define half @buffer_fat_ptr_agent_atomic_fmax_ret_f16__offset__waterfall__amdgpu
; GFX7-NEXT: v_cmp_eq_u64_e64 s[4:5], s[10:11], v[2:3]
; GFX7-NEXT: s_and_b64 s[4:5], vcc, s[4:5]
; GFX7-NEXT: s_and_saveexec_b64 s[4:5], s[4:5]
-; GFX7-NEXT: buffer_load_dword v6, v8, s[8:11], 0 offen
+; GFX7-NEXT: buffer_load_dword v4, v8, s[8:11], 0 offen
; GFX7-NEXT: s_xor_b64 exec, exec, s[4:5]
; GFX7-NEXT: s_cbranch_execnz .LBB12_1
; GFX7-NEXT: ; %bb.2:
; GFX7-NEXT: s_mov_b64 exec, s[6:7]
-; GFX7-NEXT: v_cvt_f16_f32_e32 v4, v5
+; GFX7-NEXT: v_cvt_f16_f32_e32 v5, v5
; GFX7-NEXT: s_mov_b64 s[6:7], 0
-; GFX7-NEXT: v_cvt_f32_f16_e32 v10, v4
+; GFX7-NEXT: v_cvt_f32_f16_e32 v10, v5
; GFX7-NEXT: .LBB12_3: ; %atomicrmw.start
; GFX7-NEXT: ; =>This Loop Header: Depth=1
; GFX7-NEXT: ; Child Loop BB12_4 Depth 2
; GFX7-NEXT: s_waitcnt vmcnt(0)
+; GFX7-NEXT: v_mov_b32_e32 v6, v4
; GFX7-NEXT: v_lshrrev_b32_e32 v4, v7, v6
; GFX7-NEXT: v_cvt_f32_f16_e32 v4, v4
; GFX7-NEXT: v_and_b32_e32 v5, v6, v9
@@ -4123,7 +4140,6 @@ define half @buffer_fat_ptr_agent_atomic_fmax_ret_f16__offset__waterfall__amdgpu
; GFX7-NEXT: s_waitcnt vmcnt(0)
; GFX7-NEXT: v_cmp_eq_u32_e32 vcc, v4, v6
; GFX7-NEXT: s_or_b64 s[6:7], vcc, s[6:7]
-; GFX7-NEXT: v_mov_b32_e32 v6, v4
; GFX7-NEXT: buffer_wbinvl1
; GFX7-NEXT: s_andn2_b64 exec, exec, s[6:7]
; GFX7-NEXT: s_cbranch_execnz .LBB12_3
@@ -4152,18 +4168,20 @@ define half @buffer_fat_ptr_agent_atomic_fmax_ret_f16__offset__waterfall__amdgpu
; GFX6-NEXT: v_cmp_eq_u64_e64 s[4:5], s[10:11], v[2:3]
; GFX6-NEXT: s_and_b64 s[4:5], vcc, s[4:5]
; GFX6-NEXT: s_and_saveexec_b64 s[4:5], s[4:5]
-; GFX6-NEXT: buffer_load_dword v6, v8, s[8:11], 0 offen
+; GFX6-NEXT: buffer_load_dword v4, v8, s[8:11], 0 offen
; GFX6-NEXT: s_xor_b64 exec, exec, s[4:5]
; GFX6-NEXT: s_cbranch_execnz .LBB12_1
; GFX6-NEXT: ; %bb.2:
; GFX6-NEXT: s_mov_b64 exec, s[6:7]
-; GFX6-NEXT: v_cvt_f16_f32_e32 v4, v5
+; GFX6-NEXT: v_cvt_f16_f32_e32 v5, v5
; GFX6-NEXT: s_mov_b64 s[6:7], 0
-; GFX6-NEXT: v_cvt_f32_f16_e32 v10, v4
+; GFX6-NEXT: v_cvt_f32_f16_e32 v10, v5
; GFX6-NEXT: .LBB12_3: ; %atomicrmw.start
; GFX6-NEXT: ; =>This Loop Header: Depth=1
; GFX6-NEXT: ; Child Loop BB12_4 Depth 2
-; GFX6-NEXT: s_waitcnt vmcnt(0) expcnt(0)
+; GFX6-NEXT: s_waitcnt vmcnt(0)
+; GFX6-NEXT: v_mov_b32_e32 v6, v4
+; GFX6-NEXT: s_waitcnt expcnt(0)
; GFX6-NEXT: v_lshrrev_b32_e32 v4, v7, v6
; GFX6-NEXT: v_cvt_f32_f16_e32 v4, v4
; GFX6-NEXT: v_and_b32_e32 v5, v6, v9
@@ -4193,7 +4211,6 @@ define half @buffer_fat_ptr_agent_atomic_fmax_ret_f16__offset__waterfall__amdgpu
; GFX6-NEXT: s_waitcnt vmcnt(0)
; GFX6-NEXT: v_cmp_eq_u32_e32 vcc, v4, v6
; GFX6-NEXT: s_or_b64 s[6:7], vcc, s[6:7]
-; GFX6-NEXT: v_mov_b32_e32 v6, v4
; GFX6-NEXT: buffer_wbinvl1
; GFX6-NEXT: s_andn2_b64 exec, exec, s[6:7]
; GFX6-NEXT: s_cbranch_execnz .LBB12_3
@@ -4221,55 +4238,55 @@ define bfloat @buffer_fat_ptr_agent_atomic_fmax_ret_bf16__offset__amdgpu_no_fine
; GFX12-TRUE16-NEXT: s_wait_bvhcnt 0x0
; GFX12-TRUE16-NEXT: s_wait_kmcnt 0x0
; GFX12-TRUE16-NEXT: s_addk_co_i32 s16, 0x200
-; GFX12-TRUE16-NEXT: v_lshlrev_b32_e32 v5, 16, v0
+; GFX12-TRUE16-NEXT: v_lshlrev_b32_e32 v0, 16, v0
; GFX12-TRUE16-NEXT: s_wait_alu 0xfffe
; GFX12-TRUE16-NEXT: s_and_b32 s4, s16, -4
; GFX12-TRUE16-NEXT: s_wait_alu 0xfffe
-; GFX12-TRUE16-NEXT: v_mov_b32_e32 v4, s4
+; GFX12-TRUE16-NEXT: v_mov_b32_e32 v3, s4
; GFX12-TRUE16-NEXT: s_and_b32 s4, s16, 3
; GFX12-TRUE16-NEXT: s_wait_alu 0xfffe
; GFX12-TRUE16-NEXT: s_lshl_b32 s4, s4, 3
; GFX12-TRUE16-NEXT: s_wait_alu 0xfffe
; GFX12-TRUE16-NEXT: s_lshl_b32 s5, 0xffff, s4
-; GFX12-TRUE16-NEXT: buffer_load_b32 v1, v4, s[0:3], null offen
+; GFX12-TRUE16-NEXT: buffer_load_b32 v1, v3, s[0:3], null offen
; GFX12-TRUE16-NEXT: s_not_b32 s6, s5
; GFX12-TRUE16-NEXT: s_mov_b32 s5, 0
; GFX12-TRUE16-NEXT: .LBB13_1: ; %atomicrmw.start
; GFX12-TRUE16-NEXT: ; =>This Inner Loop Header: Depth=1
; GFX12-TRUE16-NEXT: s_wait_loadcnt 0x0
-; GFX12-TRUE16-NEXT: v_lshrrev_b32_e32 v0, s4, v1
+; GFX12-TRUE16-NEXT: v_mov_b32_e32 v5, v1
; GFX12-TRUE16-NEXT: s_wait_storecnt 0x0
; GFX12-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
-; GFX12-TRUE16-NEXT: v_lshlrev_b32_e32 v0, 16, v0
-; GFX12-TRUE16-NEXT: v_max_num_f32_e32 v0, v0, v5
-; GFX12-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(SKIP_2) | instid1(VALU_DEP_3)
-; GFX12-TRUE16-NEXT: v_bfe_u32 v2, v0, 16, 1
-; GFX12-TRUE16-NEXT: v_or_b32_e32 v3, 0x400000, v0
-; GFX12-TRUE16-NEXT: v_cmp_u_f32_e32 vcc_lo, v0, v0
-; GFX12-TRUE16-NEXT: v_add3_u32 v2, v2, v0, 0x7fff
+; GFX12-TRUE16-NEXT: v_lshrrev_b32_e32 v1, s4, v5
+; GFX12-TRUE16-NEXT: v_lshlrev_b32_e32 v1, 16, v1
+; GFX12-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
+; GFX12-TRUE16-NEXT: v_max_num_f32_e32 v1, v1, v0
+; GFX12-TRUE16-NEXT: v_bfe_u32 v2, v1, 16, 1
+; GFX12-TRUE16-NEXT: v_or_b32_e32 v4, 0x400000, v1
+; GFX12-TRUE16-NEXT: v_cmp_u_f32_e32 vcc_lo, v1, v1
+; GFX12-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_3) | instskip(SKIP_1) | instid1(VALU_DEP_1)
+; GFX12-TRUE16-NEXT: v_add3_u32 v2, v2, v1, 0x7fff
; GFX12-TRUE16-NEXT: s_wait_alu 0xfffd
-; GFX12-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(SKIP_1) | instid1(VALU_DEP_2)
-; GFX12-TRUE16-NEXT: v_cndmask_b32_e32 v0, v2, v3, vcc_lo
+; GFX12-TRUE16-NEXT: v_cndmask_b32_e32 v1, v2, v4, vcc_lo
; GFX12-TRUE16-NEXT: v_mov_b16_e32 v2.h, 0
-; GFX12-TRUE16-NEXT: v_mov_b16_e32 v2.l, v0.h
-; GFX12-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(SKIP_1) | instid1(VALU_DEP_1)
-; GFX12-TRUE16-NEXT: v_lshlrev_b32_e32 v0, s4, v2
+; GFX12-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_2) | instskip(NEXT) | instid1(VALU_DEP_1)
+; GFX12-TRUE16-NEXT: v_mov_b16_e32 v2.l, v1.h
+; GFX12-TRUE16-NEXT: v_lshlrev_b32_e32 v1, s4, v2
; GFX12-TRUE16-NEXT: s_wait_alu 0xfffe
-; GFX12-TRUE16-NEXT: v_and_or_b32 v0, v1, s6, v0
-; GFX12-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1)
-; GFX12-TRUE16-NEXT: v_dual_mov_b32 v3, v1 :: v_dual_mov_b32 v2, v0
-; GFX12-TRUE16-NEXT: buffer_atomic_cmpswap_b32 v[2:3], v4, s[0:3], null offen th:TH_ATOMIC_RETURN
+; GFX12-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
+; GFX12-TRUE16-NEXT: v_and_or_b32 v4, v5, s6, v1
+; GFX12-TRUE16-NEXT: v_dual_mov_b32 v1, v4 :: v_dual_mov_b32 v2, v5
+; GFX12-TRUE16-NEXT: buffer_atomic_cmpswap_b32 v[1:2], v3, s[0:3], null offen th:TH_ATOMIC_RETURN
; GFX12-TRUE16-NEXT: s_wait_loadcnt 0x0
; GFX12-TRUE16-NEXT: global_inv scope:SCOPE_DEV
-; GFX12-TRUE16-NEXT: v_cmp_eq_u32_e32 vcc_lo, v2, v1
-; GFX12-TRUE16-NEXT: v_mov_b32_e32 v1, v2
+; GFX12-TRUE16-NEXT: v_cmp_eq_u32_e32 vcc_lo, v1, v5
; GFX12-TRUE16-NEXT: s_or_b32 s5, vcc_lo, s5
; GFX12-TRUE16-NEXT: s_wait_alu 0xfffe
; GFX12-TRUE16-NEXT: s_and_not1_b32 exec_lo, exec_lo, s5
; GFX12-TRUE16-NEXT: s_cbranch_execnz .LBB13_1
; GFX12-TRUE16-NEXT: ; %bb.2: ; %atomicrmw.end
; GFX12-TRUE16-NEXT: s_or_b32 exec_lo, exec_lo, s5
-; GFX12-TRUE16-NEXT: v_lshrrev_b32_e32 v0, s4, v2
+; GFX12-TRUE16-NEXT: v_lshrrev_b32_e32 v0, s4, v1
; GFX12-TRUE16-NEXT: s_wait_loadcnt 0x0
; GFX12-TRUE16-NEXT: s_setpc_b64 s[30:31]
;
@@ -4281,54 +4298,54 @@ define bfloat @buffer_fat_ptr_agent_atomic_fmax_ret_bf16__offset__amdgpu_no_fine
; GFX12-FAKE16-NEXT: s_wait_bvhcnt 0x0
; GFX12-FAKE16-NEXT: s_wait_kmcnt 0x0
; GFX12-FAKE16-NEXT: s_addk_co_i32 s16, 0x200
-; GFX12-FAKE16-NEXT: v_lshlrev_b32_e32 v5, 16, v0
+; GFX12-FAKE16-NEXT: v_lshlrev_b32_e32 v0, 16, v0
; GFX12-FAKE16-NEXT: s_wait_alu 0xfffe
; GFX12-FAKE16-NEXT: s_and_b32 s4, s16, -4
; GFX12-FAKE16-NEXT: s_wait_alu 0xfffe
-; GFX12-FAKE16-NEXT: v_mov_b32_e32 v4, s4
+; GFX12-FAKE16-NEXT: v_mov_b32_e32 v3, s4
; GFX12-FAKE16-NEXT: s_and_b32 s4, s16, 3
; GFX12-FAKE16-NEXT: s_wait_alu 0xfffe
; GFX12-FAKE16-NEXT: s_lshl_b32 s4, s4, 3
; GFX12-FAKE16-NEXT: s_wait_alu 0xfffe
; GFX12-FAKE16-NEXT: s_lshl_b32 s5, 0xffff, s4
-; GFX12-FAKE16-NEXT: buffer_load_b32 v1, v4, s[0:3], null offen
+; GFX12-FAKE16-NEXT: buffer_load_b32 v1, v3, s[0:3], null offen
; GFX12-FAKE16-NEXT: s_not_b32 s6, s5
; GFX12-FAKE16-NEXT: s_mov_b32 s5, 0
; GFX12-FAKE16-NEXT: .LBB13_1: ; %atomicrmw.start
; GFX12-FAKE16-NEXT: ; =>This Inner Loop Header: Depth=1
; GFX12-FAKE16-NEXT: s_wait_loadcnt 0x0
-; GFX12-FAKE16-NEXT: v_lshrrev_b32_e32 v0, s4, v1
+; GFX12-FAKE16-NEXT: v_mov_b32_e32 v5, v1
; GFX12-FAKE16-NEXT: s_wait_storecnt 0x0
; GFX12-FAKE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
-; GFX12-FAKE16-NEXT: v_lshlrev_b32_e32 v0, 16, v0
-; GFX12-FAKE16-NEXT: v_max_num_f32_e32 v0, v0, v5
-; GFX12-FAKE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(SKIP_2) | instid1(VALU_DEP_3)
-; GFX12-FAKE16-NEXT: v_bfe_u32 v2, v0, 16, 1
-; GFX12-FAKE16-NEXT: v_or_b32_e32 v3, 0x400000, v0
-; GFX12-FAKE16-NEXT: v_cmp_u_f32_e32 vcc_lo, v0, v0
-; GFX12-FAKE16-NEXT: v_add3_u32 v2, v2, v0, 0x7fff
+; GFX12-FAKE16-NEXT: v_lshrrev_b32_e32 v1, s4, v5
+; GFX12-FAKE16-NEXT: v_lshlrev_b32_e32 v1, 16, v1
+; GFX12-FAKE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
+; GFX12-FAKE16-NEXT: v_max_num_f32_e32 v1, v1, v0
+; GFX12-FAKE16-NEXT: v_bfe_u32 v2, v1, 16, 1
+; GFX12-FAKE16-NEXT: v_or_b32_e32 v4, 0x400000, v1
+; GFX12-FAKE16-NEXT: v_cmp_u_f32_e32 vcc_lo, v1, v1
+; GFX12-FAKE16-NEXT: s_delay_alu instid0(VALU_DEP_3) | instskip(SKIP_1) | instid1(VALU_DEP_1)
+; GFX12-FAKE16-NEXT: v_add3_u32 v2, v2, v1, 0x7fff
; GFX12-FAKE16-NEXT: s_wait_alu 0xfffd
+; GFX12-FAKE16-NEXT: v_cndmask_b32_e32 v1, v2, v4, vcc_lo
; GFX12-FAKE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
-; GFX12-FAKE16-NEXT: v_cndmask_b32_e32 v0, v2, v3, vcc_lo
-; GFX12-FAKE16-NEXT: v_lshrrev_b32_e32 v0, 16, v0
-; GFX12-FAKE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(SKIP_1) | instid1(VALU_DEP_1)
-; GFX12-FAKE16-NEXT: v_lshlrev_b32_e32 v0, s4, v0
+; GFX12-FAKE16-NEXT: v_lshrrev_b32_e32 v1, 16, v1
+; GFX12-FAKE16-NEXT: v_lshlrev_b32_e32 v1, s4, v1
; GFX12-FAKE16-NEXT: s_wait_alu 0xfffe
-; GFX12-FAKE16-NEXT: v_and_or_b32 v0, v1, s6, v0
-; GFX12-FAKE16-NEXT: s_delay_alu instid0(VALU_DEP_1)
-; GFX12-FAKE16-NEXT: v_dual_mov_b32 v3, v1 :: v_dual_mov_b32 v2, v0
-; GFX12-FAKE16-NEXT: buffer_atomic_cmpswap_b32 v[2:3], v4, s[0:3], null offen th:TH_ATOMIC_RETURN
+; GFX12-FAKE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
+; GFX12-FAKE16-NEXT: v_and_or_b32 v4, v5, s6, v1
+; GFX12-FAKE16-NEXT: v_dual_mov_b32 v1, v4 :: v_dual_mov_b32 v2, v5
+; GFX12-FAKE16-NEXT: buffer_atomic_cmpswap_b32 v[1:2], v3, s[0:3], null offen th:TH_ATOMIC_RETURN
; GFX12-FAKE16-NEXT: s_wait_loadcnt 0x0
; GFX12-FAKE16-NEXT: global_inv scope:SCOPE_DEV
-; GFX12-FAKE16-NEXT: v_cmp_eq_u32_e32 vcc_lo, v2, v1
-; GFX12-FAKE16-NEXT: v_mov_b32_e32 v1, v2
+; GFX12-FAKE16-NEXT: v_cmp_eq_u32_e32 vcc_lo, v1, v5
; GFX12-FAKE16-NEXT: s_or_b32 s5, vcc_lo, s5
; GFX12-FAKE16-NEXT: s_wait_alu 0xfffe
; GFX12-FAKE16-NEXT: s_and_not1_b32 exec_lo, exec_lo, s5
; GFX12-FAKE16-NEXT: s_cbranch_execnz .LBB13_1
; GFX12-FAKE16-NEXT: ; %bb.2: ; %atomicrmw.end
; GFX12-FAKE16-NEXT: s_or_b32 exec_lo, exec_lo, s5
-; GFX12-FAKE16-NEXT: v_lshrrev_b32_e32 v0, s4, v2
+; GFX12-FAKE16-NEXT: v_lshrrev_b32_e32 v0, s4, v1
; GFX12-FAKE16-NEXT: s_wait_loadcnt 0x0
; GFX12-FAKE16-NEXT: s_setpc_b64 s[30:31]
;
@@ -4337,36 +4354,36 @@ define bfloat @buffer_fat_ptr_agent_atomic_fmax_ret_bf16__offset__amdgpu_no_fine
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX942-NEXT: s_addk_i32 s16, 0x200
; GFX942-NEXT: s_and_b32 s4, s16, -4
-; GFX942-NEXT: v_mov_b32_e32 v4, s4
-; GFX942-NEXT: buffer_load_dword v1, v4, s[0:3], 0 offen
+; GFX942-NEXT: v_mov_b32_e32 v1, s4
+; GFX942-NEXT: buffer_load_dword v2, v1, s[0:3], 0 offen
; GFX942-NEXT: s_and_b32 s4, s16, 3
; GFX942-NEXT: s_lshl_b32 s6, s4, 3
; GFX942-NEXT: s_lshl_b32 s4, 0xffff, s6
; GFX942-NEXT: s_not_b32 s7, s4
; GFX942-NEXT: s_mov_b64 s[4:5], 0
-; GFX942-NEXT: v_lshlrev_b32_e32 v5, 16, v0
+; GFX942-NEXT: v_lshlrev_b32_e32 v0, 16, v0
; GFX942-NEXT: s_movk_i32 s8, 0x7fff
; GFX942-NEXT: .LBB13_1: ; %atomicrmw.start
; GFX942-NEXT: ; =>This Inner Loop Header: Depth=1
; GFX942-NEXT: s_waitcnt vmcnt(0)
-; GFX942-NEXT: v_lshrrev_b32_sdwa v0, s6, v1 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:DWORD
+; GFX942-NEXT: v_mov_b32_e32 v5, v2
+; GFX942-NEXT: v_lshrrev_b32_sdwa v2, s6, v5 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:DWORD
; GFX942-NEXT: buffer_wbl2 sc1
-; GFX942-NEXT: v_max_f32_e32 v0, v0, v5
-; GFX942-NEXT: v_bfe_u32 v2, v0, 16, 1
-; GFX942-NEXT: v_or_b32_e32 v3, 0x400000, v0
-; GFX942-NEXT: v_add3_u32 v2, v2, v0, s8
-; GFX942-NEXT: v_cmp_u_f32_e32 vcc, v0, v0
+; GFX942-NEXT: v_max_f32_e32 v2, v2, v0
+; GFX942-NEXT: v_bfe_u32 v3, v2, 16, 1
+; GFX942-NEXT: v_or_b32_e32 v4, 0x400000, v2
+; GFX942-NEXT: v_add3_u32 v3, v3, v2, s8
+; GFX942-NEXT: v_cmp_u_f32_e32 vcc, v2, v2
; GFX942-NEXT: s_nop 1
-; GFX942-NEXT: v_cndmask_b32_e32 v0, v2, v3, vcc
-; GFX942-NEXT: v_lshlrev_b32_sdwa v0, s6, v0 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:WORD_1
-; GFX942-NEXT: v_and_or_b32 v0, v1, s7, v0
-; GFX942-NEXT: v_mov_b64_e32 v[2:3], v[0:1]
-; GFX942-NEXT: buffer_atomic_cmpswap v[2:3], v4, s[0:3], 0 offen sc0
+; GFX942-NEXT: v_cndmask_b32_e32 v2, v3, v4, vcc
+; GFX942-NEXT: v_lshlrev_b32_sdwa v2, s6, v2 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:WORD_1
+; GFX942-NEXT: v_and_or_b32 v4, v5, s7, v2
+; GFX942-NEXT: v_mov_b64_e32 v[2:3], v[4:5]
+; GFX942-NEXT: buffer_atomic_cmpswap v[2:3], v1, s[0:3], 0 offen sc0
; GFX942-NEXT: s_waitcnt vmcnt(0)
; GFX942-NEXT: buffer_inv sc1
-; GFX942-NEXT: v_cmp_eq_u32_e32 vcc, v2, v1
+; GFX942-NEXT: v_cmp_eq_u32_e32 vcc, v2, v5
; GFX942-NEXT: s_or_b64 s[4:5], vcc, s[4:5]
-; GFX942-NEXT: v_mov_b32_e32 v1, v2
; GFX942-NEXT: s_andn2_b64 exec, exec, s[4:5]
; GFX942-NEXT: s_cbranch_execnz .LBB13_1
; GFX942-NEXT: ; %bb.2: ; %atomicrmw.end
@@ -4378,147 +4395,147 @@ define bfloat @buffer_fat_ptr_agent_atomic_fmax_ret_bf16__offset__amdgpu_no_fine
; GFX11-TRUE16: ; %bb.0:
; GFX11-TRUE16-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX11-TRUE16-NEXT: s_addk_i32 s16, 0x200
-; GFX11-TRUE16-NEXT: v_lshlrev_b32_e32 v5, 16, v0
+; GFX11-TRUE16-NEXT: v_lshlrev_b32_e32 v0, 16, v0
; GFX11-TRUE16-NEXT: s_and_b32 s4, s16, -4
; GFX11-TRUE16-NEXT: s_delay_alu instid0(SALU_CYCLE_1) | instskip(SKIP_1) | instid1(SALU_CYCLE_1)
-; GFX11-TRUE16-NEXT: v_mov_b32_e32 v4, s4
+; GFX11-TRUE16-NEXT: v_mov_b32_e32 v3, s4
; GFX11-TRUE16-NEXT: s_and_b32 s4, s16, 3
; GFX11-TRUE16-NEXT: s_lshl_b32 s4, s4, 3
; GFX11-TRUE16-NEXT: s_delay_alu instid0(SALU_CYCLE_1)
; GFX11-TRUE16-NEXT: s_lshl_b32 s5, 0xffff, s4
-; GFX11-TRUE16-NEXT: buffer_load_b32 v1, v4, s[0:3], 0 offen
+; GFX11-TRUE16-NEXT: buffer_load_b32 v1, v3, s[0:3], 0 offen
; GFX11-TRUE16-NEXT: s_not_b32 s6, s5
; GFX11-TRUE16-NEXT: s_mov_b32 s5, 0
; GFX11-TRUE16-NEXT: .p2align 6
; GFX11-TRUE16-NEXT: .LBB13_1: ; %atomicrmw.start
; GFX11-TRUE16-NEXT: ; =>This Inner Loop Header: Depth=1
; GFX11-TRUE16-NEXT: s_waitcnt vmcnt(0)
-; GFX11-TRUE16-NEXT: v_lshrrev_b32_e32 v0, s4, v1
+; GFX11-TRUE16-NEXT: v_mov_b32_e32 v5, v1
; GFX11-TRUE16-NEXT: s_waitcnt_vscnt null, 0x0
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
-; GFX11-TRUE16-NEXT: v_lshlrev_b32_e32 v0, 16, v0
-; GFX11-TRUE16-NEXT: v_max_f32_e32 v0, v0, v5
-; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(SKIP_2) | instid1(VALU_DEP_3)
-; GFX11-TRUE16-NEXT: v_bfe_u32 v2, v0, 16, 1
-; GFX11-TRUE16-NEXT: v_or_b32_e32 v3, 0x400000, v0
-; GFX11-TRUE16-NEXT: v_cmp_u_f32_e32 vcc_lo, v0, v0
-; GFX11-TRUE16-NEXT: v_add3_u32 v2, v2, v0, 0x7fff
-; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(SKIP_1) | instid1(VALU_DEP_2)
-; GFX11-TRUE16-NEXT: v_cndmask_b32_e32 v0, v2, v3, vcc_lo
+; GFX11-TRUE16-NEXT: v_lshrrev_b32_e32 v1, s4, v5
+; GFX11-TRUE16-NEXT: v_lshlrev_b32_e32 v1, 16, v1
+; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
+; GFX11-TRUE16-NEXT: v_max_f32_e32 v1, v1, v0
+; GFX11-TRUE16-NEXT: v_bfe_u32 v2, v1, 16, 1
+; GFX11-TRUE16-NEXT: v_or_b32_e32 v4, 0x400000, v1
+; GFX11-TRUE16-NEXT: v_cmp_u_f32_e32 vcc_lo, v1, v1
+; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_3) | instskip(NEXT) | instid1(VALU_DEP_1)
+; GFX11-TRUE16-NEXT: v_add3_u32 v2, v2, v1, 0x7fff
+; GFX11-TRUE16-NEXT: v_cndmask_b32_e32 v1, v2, v4, vcc_lo
; GFX11-TRUE16-NEXT: v_mov_b16_e32 v2.h, 0
-; GFX11-TRUE16-NEXT: v_mov_b16_e32 v2.l, v0.h
+; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_2) | instskip(NEXT) | instid1(VALU_DEP_1)
+; GFX11-TRUE16-NEXT: v_mov_b16_e32 v2.l, v1.h
+; GFX11-TRUE16-NEXT: v_lshlrev_b32_e32 v1, s4, v2
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
-; GFX11-TRUE16-NEXT: v_lshlrev_b32_e32 v0, s4, v2
-; GFX11-TRUE16-NEXT: v_and_or_b32 v0, v1, s6, v0
-; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1)
-; GFX11-TRUE16-NEXT: v_dual_mov_b32 v3, v1 :: v_dual_mov_b32 v2, v0
-; GFX11-TRUE16-NEXT: buffer_atomic_cmpswap_b32 v[2:3], v4, s[0:3], 0 offen glc
+; GFX11-TRUE16-NEXT: v_and_or_b32 v4, v5, s6, v1
+; GFX11-TRUE16-NEXT: v_dual_mov_b32 v1, v4 :: v_dual_mov_b32 v2, v5
+; GFX11-TRUE16-NEXT: buffer_atomic_cmpswap_b32 v[1:2], v3, s[0:3], 0 offen glc
; GFX11-TRUE16-NEXT: s_waitcnt vmcnt(0)
; GFX11-TRUE16-NEXT: buffer_gl1_inv
; GFX11-TRUE16-NEXT: buffer_gl0_inv
-; GFX11-TRUE16-NEXT: v_cmp_eq_u32_e32 vcc_lo, v2, v1
-; GFX11-TRUE16-NEXT: v_mov_b32_e32 v1, v2
+; GFX11-TRUE16-NEXT: v_cmp_eq_u32_e32 vcc_lo, v1, v5
; GFX11-TRUE16-NEXT: s_or_b32 s5, vcc_lo, s5
; GFX11-TRUE16-NEXT: s_delay_alu instid0(SALU_CYCLE_1)
; GFX11-TRUE16-NEXT: s_and_not1_b32 exec_lo, exec_lo, s5
; GFX11-TRUE16-NEXT: s_cbranch_execnz .LBB13_1
; GFX11-TRUE16-NEXT: ; %bb.2: ; %atomicrmw.end
; GFX11-TRUE16-NEXT: s_or_b32 exec_lo, exec_lo, s5
-; GFX11-TRUE16-NEXT: v_lshrrev_b32_e32 v0, s4, v2
+; GFX11-TRUE16-NEXT: v_lshrrev_b32_e32 v0, s4, v1
; GFX11-TRUE16-NEXT: s_setpc_b64 s[30:31]
;
; GFX11-FAKE16-LABEL: buffer_fat_ptr_agent_atomic_fmax_ret_bf16__offset__amdgpu_no_fine_grained_memory:
; GFX11-FAKE16: ; %bb.0:
; GFX11-FAKE16-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX11-FAKE16-NEXT: s_addk_i32 s16, 0x200
-; GFX11-FAKE16-NEXT: v_lshlrev_b32_e32 v5, 16, v0
+; GFX11-FAKE16-NEXT: v_lshlrev_b32_e32 v0, 16, v0
; GFX11-FAKE16-NEXT: s_and_b32 s4, s16, -4
; GFX11-FAKE16-NEXT: s_delay_alu instid0(SALU_CYCLE_1) | instskip(SKIP_1) | instid1(SALU_CYCLE_1)
-; GFX11-FAKE16-NEXT: v_mov_b32_e32 v4, s4
+; GFX11-FAKE16-NEXT: v_mov_b32_e32 v3, s4
; GFX11-FAKE16-NEXT: s_and_b32 s4, s16, 3
; GFX11-FAKE16-NEXT: s_lshl_b32 s4, s4, 3
; GFX11-FAKE16-NEXT: s_delay_alu instid0(SALU_CYCLE_1)
; GFX11-FAKE16-NEXT: s_lshl_b32 s5, 0xffff, s4
-; GFX11-FAKE16-NEXT: buffer_load_b32 v1, v4, s[0:3], 0 offen
+; GFX11-FAKE16-NEXT: buffer_load_b32 v1, v3, s[0:3], 0 offen
; GFX11-FAKE16-NEXT: s_not_b32 s6, s5
; GFX11-FAKE16-NEXT: s_mov_b32 s5, 0
; GFX11-FAKE16-NEXT: .p2align 6
; GFX11-FAKE16-NEXT: .LBB13_1: ; %atomicrmw.start
; GFX11-FAKE16-NEXT: ; =>This Inner Loop Header: Depth=1
; GFX11-FAKE16-NEXT: s_waitcnt vmcnt(0)
-; GFX11-FAKE16-NEXT: v_lshrrev_b32_e32 v0, s4, v1
+; GFX11-FAKE16-NEXT: v_mov_b32_e32 v5, v1
; GFX11-FAKE16-NEXT: s_waitcnt_vscnt null, 0x0
; GFX11-FAKE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
-; GFX11-FAKE16-NEXT: v_lshlrev_b32_e32 v0, 16, v0
-; GFX11-FAKE16-NEXT: v_max_f32_e32 v0, v0, v5
-; GFX11-FAKE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(SKIP_2) | instid1(VALU_DEP_3)
-; GFX11-FAKE16-NEXT: v_bfe_u32 v2, v0, 16, 1
-; GFX11-FAKE16-NEXT: v_or_b32_e32 v3, 0x400000, v0
-; GFX11-FAKE16-NEXT: v_cmp_u_f32_e32 vcc_lo, v0, v0
-; GFX11-FAKE16-NEXT: v_add3_u32 v2, v2, v0, 0x7fff
+; GFX11-FAKE16-NEXT: v_lshrrev_b32_e32 v1, s4, v5
+; GFX11-FAKE16-NEXT: v_lshlrev_b32_e32 v1, 16, v1
; GFX11-FAKE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
-; GFX11-FAKE16-NEXT: v_cndmask_b32_e32 v0, v2, v3, vcc_lo
-; GFX11-FAKE16-NEXT: v_lshrrev_b32_e32 v0, 16, v0
+; GFX11-FAKE16-NEXT: v_max_f32_e32 v1, v1, v0
+; GFX11-FAKE16-NEXT: v_bfe_u32 v2, v1, 16, 1
+; GFX11-FAKE16-NEXT: v_or_b32_e32 v4, 0x400000, v1
+; GFX11-FAKE16-NEXT: v_cmp_u_f32_e32 vcc_lo, v1, v1
+; GFX11-FAKE16-NEXT: s_delay_alu instid0(VALU_DEP_3) | instskip(NEXT) | instid1(VALU_DEP_1)
+; GFX11-FAKE16-NEXT: v_add3_u32 v2, v2, v1, 0x7fff
+; GFX11-FAKE16-NEXT: v_cndmask_b32_e32 v1, v2, v4, vcc_lo
; GFX11-FAKE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
-; GFX11-FAKE16-NEXT: v_lshlrev_b32_e32 v0, s4, v0
-; GFX11-FAKE16-NEXT: v_and_or_b32 v0, v1, s6, v0
-; GFX11-FAKE16-NEXT: s_delay_alu instid0(VALU_DEP_1)
-; GFX11-FAKE16-NEXT: v_dual_mov_b32 v3, v1 :: v_dual_mov_b32 v2, v0
-; GFX11-FAKE16-NEXT: buffer_atomic_cmpswap_b32 v[2:3], v4, s[0:3], 0 offen glc
+; GFX11-FAKE16-NEXT: v_lshrrev_b32_e32 v1, 16, v1
+; GFX11-FAKE16-NEXT: v_lshlrev_b32_e32 v1, s4, v1
+; GFX11-FAKE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
+; GFX11-FAKE16-NEXT: v_and_or_b32 v4, v5, s6, v1
+; GFX11-FAKE16-NEXT: v_dual_mov_b32 v1, v4 :: v_dual_mov_b32 v2, v5
+; GFX11-FAKE16-NEXT: buffer_atomic_cmpswap_b32 v[1:2], v3, s[0:3], 0 offen glc
; GFX11-FAKE16-NEXT: s_waitcnt vmcnt(0)
; GFX11-FAKE16-NEXT: buffer_gl1_inv
; GFX11-FAKE16-NEXT: buffer_gl0_inv
-; GFX11-FAKE16-NEXT: v_cmp_eq_u32_e32 vcc_lo, v2, v1
-; GFX11-FAKE16-NEXT: v_mov_b32_e32 v1, v2
+; GFX11-FAKE16-NEXT: v_cmp_eq_u32_e32 vcc_lo, v1, v5
; GFX11-FAKE16-NEXT: s_or_b32 s5, vcc_lo, s5
; GFX11-FAKE16-NEXT: s_delay_alu instid0(SALU_CYCLE_1)
; GFX11-FAKE16-NEXT: s_and_not1_b32 exec_lo, exec_lo, s5
; GFX11-FAKE16-NEXT: s_cbranch_execnz .LBB13_1
; GFX11-FAKE16-NEXT: ; %bb.2: ; %atomicrmw.end
; GFX11-FAKE16-NEXT: s_or_b32 exec_lo, exec_lo, s5
-; GFX11-FAKE16-NEXT: v_lshrrev_b32_e32 v0, s4, v2
+; GFX11-FAKE16-NEXT: v_lshrrev_b32_e32 v0, s4, v1
; GFX11-FAKE16-NEXT: s_setpc_b64 s[30:31]
;
; GFX10-LABEL: buffer_fat_ptr_agent_atomic_fmax_ret_bf16__offset__amdgpu_no_fine_grained_memory:
; GFX10: ; %bb.0:
; GFX10-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX10-NEXT: s_addk_i32 s20, 0x200
-; GFX10-NEXT: v_lshlrev_b32_e32 v5, 16, v0
+; GFX10-NEXT: v_lshlrev_b32_e32 v0, 16, v0
; GFX10-NEXT: s_and_b32 s4, s20, -4
-; GFX10-NEXT: v_mov_b32_e32 v4, s4
+; GFX10-NEXT: v_mov_b32_e32 v3, s4
; GFX10-NEXT: s_and_b32 s4, s20, 3
; GFX10-NEXT: s_lshl_b32 s4, s4, 3
; GFX10-NEXT: s_lshl_b32 s5, 0xffff, s4
-; GFX10-NEXT: buffer_load_dword v1, v4, s[16:19], 0 offen
+; GFX10-NEXT: buffer_load_dword v1, v3, s[16:19], 0 offen
; GFX10-NEXT: s_not_b32 s6, s5
; GFX10-NEXT: s_mov_b32 s5, 0
; GFX10-NEXT: .LBB13_1: ; %atomicrmw.start
; GFX10-NEXT: ; =>This Inner Loop Header: Depth=1
; GFX10-NEXT: s_waitcnt vmcnt(0)
-; GFX10-NEXT: v_lshrrev_b32_sdwa v0, s4, v1 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:DWORD
+; GFX10-NEXT: v_mov_b32_e32 v5, v1
; GFX10-NEXT: s_waitcnt_vscnt null, 0x0
-; GFX10-NEXT: v_max_f32_e32 v0, v0, v5
-; GFX10-NEXT: v_bfe_u32 v2, v0, 16, 1
-; GFX10-NEXT: v_or_b32_e32 v3, 0x400000, v0
-; GFX10-NEXT: v_cmp_u_f32_e32 vcc_lo, v0, v0
-; GFX10-NEXT: v_add3_u32 v2, v2, v0, 0x7fff
-; GFX10-NEXT: v_cndmask_b32_e32 v0, v2, v3, vcc_lo
-; GFX10-NEXT: v_lshlrev_b32_sdwa v0, s4, v0 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:WORD_1
-; GFX10-NEXT: v_and_or_b32 v0, v1, s6, v0
-; GFX10-NEXT: v_mov_b32_e32 v3, v1
-; GFX10-NEXT: v_mov_b32_e32 v2, v0
-; GFX10-NEXT: buffer_atomic_cmpswap v[2:3], v4, s[16:19], 0 offen glc
+; GFX10-NEXT: v_lshrrev_b32_sdwa v1, s4, v5 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:DWORD
+; GFX10-NEXT: v_max_f32_e32 v1, v1, v0
+; GFX10-NEXT: v_bfe_u32 v2, v1, 16, 1
+; GFX10-NEXT: v_or_b32_e32 v4, 0x400000, v1
+; GFX10-NEXT: v_cmp_u_f32_e32 vcc_lo, v1, v1
+; GFX10-NEXT: v_add3_u32 v2, v2, v1, 0x7fff
+; GFX10-NEXT: v_cndmask_b32_e32 v1, v2, v4, vcc_lo
+; GFX10-NEXT: v_lshlrev_b32_sdwa v1, s4, v1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:WORD_1
+; GFX10-NEXT: v_and_or_b32 v4, v5, s6, v1
+; GFX10-NEXT: v_mov_b32_e32 v1, v4
+; GFX10-NEXT: v_mov_b32_e32 v2, v5
+; GFX10-NEXT: buffer_atomic_cmpswap v[1:2], v3, s[16:19], 0 offen glc
; GFX10-NEXT: s_waitcnt vmcnt(0)
; GFX10-NEXT: buffer_gl1_inv
; GFX10-NEXT: buffer_gl0_inv
-; GFX10-NEXT: v_cmp_eq_u32_e32 vcc_lo, v2, v1
-; GFX10-NEXT: v_mov_b32_e32 v1, v2
+; GFX10-NEXT: v_cmp_eq_u32_e32 vcc_lo, v1, v5
; GFX10-NEXT: s_or_b32 s5, vcc_lo, s5
; GFX10-NEXT: s_andn2_b32 exec_lo, exec_lo, s5
; GFX10-NEXT: s_cbranch_execnz .LBB13_1
; GFX10-NEXT: ; %bb.2: ; %atomicrmw.end
; GFX10-NEXT: s_or_b32 exec_lo, exec_lo, s5
-; GFX10-NEXT: v_lshrrev_b32_e32 v0, s4, v2
+; GFX10-NEXT: v_lshrrev_b32_e32 v0, s4, v1
; GFX10-NEXT: s_setpc_b64 s[30:31]
;
; GFX90A-LABEL: buffer_fat_ptr_agent_atomic_fmax_ret_bf16__offset__amdgpu_no_fine_grained_memory:
@@ -4526,34 +4543,34 @@ define bfloat @buffer_fat_ptr_agent_atomic_fmax_ret_bf16__offset__amdgpu_no_fine
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX90A-NEXT: s_addk_i32 s20, 0x200
; GFX90A-NEXT: s_and_b32 s4, s20, -4
-; GFX90A-NEXT: v_mov_b32_e32 v4, s4
-; GFX90A-NEXT: buffer_load_dword v1, v4, s[16:19], 0 offen
+; GFX90A-NEXT: v_mov_b32_e32 v1, s4
+; GFX90A-NEXT: buffer_load_dword v2, v1, s[16:19], 0 offen
; GFX90A-NEXT: s_and_b32 s4, s20, 3
; GFX90A-NEXT: s_lshl_b32 s6, s4, 3
; GFX90A-NEXT: s_lshl_b32 s4, 0xffff, s6
; GFX90A-NEXT: s_not_b32 s7, s4
; GFX90A-NEXT: s_mov_b64 s[4:5], 0
-; GFX90A-NEXT: v_lshlrev_b32_e32 v5, 16, v0
+; GFX90A-NEXT: v_lshlrev_b32_e32 v0, 16, v0
; GFX90A-NEXT: s_movk_i32 s8, 0x7fff
; GFX90A-NEXT: .LBB13_1: ; %atomicrmw.start
; GFX90A-NEXT: ; =>This Inner Loop Header: Depth=1
; GFX90A-NEXT: s_waitcnt vmcnt(0)
-; GFX90A-NEXT: v_lshrrev_b32_sdwa v0, s6, v1 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:DWORD
-; GFX90A-NEXT: v_max_f32_e32 v0, v0, v5
-; GFX90A-NEXT: v_bfe_u32 v2, v0, 16, 1
-; GFX90A-NEXT: v_or_b32_e32 v3, 0x400000, v0
-; GFX90A-NEXT: v_add3_u32 v2, v2, v0, s8
-; GFX90A-NEXT: v_cmp_u_f32_e32 vcc, v0, v0
-; GFX90A-NEXT: v_cndmask_b32_e32 v0, v2, v3, vcc
-; GFX90A-NEXT: v_lshlrev_b32_sdwa v0, s6, v0 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:WORD_1
-; GFX90A-NEXT: v_and_or_b32 v0, v1, s7, v0
-; GFX90A-NEXT: v_pk_mov_b32 v[2:3], v[0:1], v[0:1] op_sel:[0,1]
-; GFX90A-NEXT: buffer_atomic_cmpswap v[2:3], v4, s[16:19], 0 offen glc
+; GFX90A-NEXT: v_mov_b32_e32 v5, v2
+; GFX90A-NEXT: v_lshrrev_b32_sdwa v2, s6, v5 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:DWORD
+; GFX90A-NEXT: v_max_f32_e32 v2, v2, v0
+; GFX90A-NEXT: v_bfe_u32 v3, v2, 16, 1
+; GFX90A-NEXT: v_or_b32_e32 v4, 0x400000, v2
+; GFX90A-NEXT: v_add3_u32 v3, v3, v2, s8
+; GFX90A-NEXT: v_cmp_u_f32_e32 vcc, v2, v2
+; GFX90A-NEXT: v_cndmask_b32_e32 v2, v3, v4, vcc
+; GFX90A-NEXT: v_lshlrev_b32_sdwa v2, s6, v2 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:WORD_1
+; GFX90A-NEXT: v_and_or_b32 v4, v5, s7, v2
+; GFX90A-NEXT: v_pk_mov_b32 v[2:3], v[4:5], v[4:5] op_sel:[0,1]
+; GFX90A-NEXT: buffer_atomic_cmpswap v[2:3], v1, s[16:19], 0 offen glc
; GFX90A-NEXT: s_waitcnt vmcnt(0)
; GFX90A-NEXT: buffer_wbinvl1
-; GFX90A-NEXT: v_cmp_eq_u32_e32 vcc, v2, v1
+; GFX90A-NEXT: v_cmp_eq_u32_e32 vcc, v2, v5
; GFX90A-NEXT: s_or_b64 s[4:5], vcc, s[4:5]
-; GFX90A-NEXT: v_mov_b32_e32 v1, v2
; GFX90A-NEXT: s_andn2_b64 exec, exec, s[4:5]
; GFX90A-NEXT: s_cbranch_execnz .LBB13_1
; GFX90A-NEXT: ; %bb.2: ; %atomicrmw.end
@@ -4566,40 +4583,40 @@ define bfloat @buffer_fat_ptr_agent_atomic_fmax_ret_bf16__offset__amdgpu_no_fine
; GFX908-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX908-NEXT: s_addk_i32 s20, 0x200
; GFX908-NEXT: s_and_b32 s4, s20, -4
-; GFX908-NEXT: v_mov_b32_e32 v4, s4
-; GFX908-NEXT: buffer_load_dword v1, v4, s[16:19], 0 offen
+; GFX908-NEXT: v_mov_b32_e32 v3, s4
+; GFX908-NEXT: buffer_load_dword v1, v3, s[16:19], 0 offen
; GFX908-NEXT: s_and_b32 s4, s20, 3
; GFX908-NEXT: s_lshl_b32 s6, s4, 3
; GFX908-NEXT: s_lshl_b32 s4, 0xffff, s6
; GFX908-NEXT: s_not_b32 s7, s4
; GFX908-NEXT: s_mov_b64 s[4:5], 0
-; GFX908-NEXT: v_lshlrev_b32_e32 v5, 16, v0
+; GFX908-NEXT: v_lshlrev_b32_e32 v0, 16, v0
; GFX908-NEXT: s_movk_i32 s8, 0x7fff
; GFX908-NEXT: .LBB13_1: ; %atomicrmw.start
; GFX908-NEXT: ; =>This Inner Loop Header: Depth=1
; GFX908-NEXT: s_waitcnt vmcnt(0)
-; GFX908-NEXT: v_lshrrev_b32_sdwa v0, s6, v1 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:DWORD
-; GFX908-NEXT: v_max_f32_e32 v0, v0, v5
-; GFX908-NEXT: v_bfe_u32 v2, v0, 16, 1
-; GFX908-NEXT: v_or_b32_e32 v3, 0x400000, v0
-; GFX908-NEXT: v_add3_u32 v2, v2, v0, s8
-; GFX908-NEXT: v_cmp_u_f32_e32 vcc, v0, v0
-; GFX908-NEXT: v_cndmask_b32_e32 v0, v2, v3, vcc
-; GFX908-NEXT: v_lshlrev_b32_sdwa v0, s6, v0 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:WORD_1
-; GFX908-NEXT: v_and_or_b32 v0, v1, s7, v0
-; GFX908-NEXT: v_mov_b32_e32 v3, v1
-; GFX908-NEXT: v_mov_b32_e32 v2, v0
-; GFX908-NEXT: buffer_atomic_cmpswap v[2:3], v4, s[16:19], 0 offen glc
+; GFX908-NEXT: v_mov_b32_e32 v5, v1
+; GFX908-NEXT: v_lshrrev_b32_sdwa v1, s6, v5 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:DWORD
+; GFX908-NEXT: v_max_f32_e32 v1, v1, v0
+; GFX908-NEXT: v_bfe_u32 v2, v1, 16, 1
+; GFX908-NEXT: v_or_b32_e32 v4, 0x400000, v1
+; GFX908-NEXT: v_add3_u32 v2, v2, v1, s8
+; GFX908-NEXT: v_cmp_u_f32_e32 vcc, v1, v1
+; GFX908-NEXT: v_cndmask_b32_e32 v1, v2, v4, vcc
+; GFX908-NEXT: v_lshlrev_b32_sdwa v1, s6, v1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:WORD_1
+; GFX908-NEXT: v_and_or_b32 v4, v5, s7, v1
+; GFX908-NEXT: v_mov_b32_e32 v1, v4
+; GFX908-NEXT: v_mov_b32_e32 v2, v5
+; GFX908-NEXT: buffer_atomic_cmpswap v[1:2], v3, s[16:19], 0 offen glc
; GFX908-NEXT: s_waitcnt vmcnt(0)
; GFX908-NEXT: buffer_wbinvl1
-; GFX908-NEXT: v_cmp_eq_u32_e32 vcc, v2, v1
+; GFX908-NEXT: v_cmp_eq_u32_e32 vcc, v1, v5
; GFX908-NEXT: s_or_b64 s[4:5], vcc, s[4:5]
-; GFX908-NEXT: v_mov_b32_e32 v1, v2
; GFX908-NEXT: s_andn2_b64 exec, exec, s[4:5]
; GFX908-NEXT: s_cbranch_execnz .LBB13_1
; GFX908-NEXT: ; %bb.2: ; %atomicrmw.end
; GFX908-NEXT: s_or_b64 exec, exec, s[4:5]
-; GFX908-NEXT: v_lshrrev_b32_e32 v0, s6, v2
+; GFX908-NEXT: v_lshrrev_b32_e32 v0, s6, v1
; GFX908-NEXT: s_setpc_b64 s[30:31]
;
; GFX8-LABEL: buffer_fat_ptr_agent_atomic_fmax_ret_bf16__offset__amdgpu_no_fine_grained_memory:
@@ -4607,42 +4624,42 @@ define bfloat @buffer_fat_ptr_agent_atomic_fmax_ret_bf16__offset__amdgpu_no_fine
; GFX8-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX8-NEXT: s_addk_i32 s20, 0x200
; GFX8-NEXT: s_and_b32 s4, s20, -4
-; GFX8-NEXT: v_mov_b32_e32 v4, s4
-; GFX8-NEXT: buffer_load_dword v1, v4, s[16:19], 0 offen
+; GFX8-NEXT: v_mov_b32_e32 v3, s4
+; GFX8-NEXT: buffer_load_dword v1, v3, s[16:19], 0 offen
; GFX8-NEXT: s_and_b32 s4, s20, 3
; GFX8-NEXT: s_lshl_b32 s6, s4, 3
; GFX8-NEXT: s_lshl_b32 s4, 0xffff, s6
; GFX8-NEXT: s_not_b32 s7, s4
; GFX8-NEXT: s_mov_b64 s[4:5], 0
-; GFX8-NEXT: v_lshlrev_b32_e32 v5, 16, v0
+; GFX8-NEXT: v_lshlrev_b32_e32 v0, 16, v0
; GFX8-NEXT: .LBB13_1: ; %atomicrmw.start
; GFX8-NEXT: ; =>This Inner Loop Header: Depth=1
-; GFX8-NEXT: v_mov_b32_e32 v0, s6
+; GFX8-NEXT: v_mov_b32_e32 v2, s6
; GFX8-NEXT: s_waitcnt vmcnt(0)
-; GFX8-NEXT: v_lshrrev_b32_sdwa v3, v0, v1 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:DWORD
-; GFX8-NEXT: v_max_f32_e32 v3, v3, v5
-; GFX8-NEXT: v_bfe_u32 v6, v3, 16, 1
-; GFX8-NEXT: v_add_u32_e32 v6, vcc, v6, v3
+; GFX8-NEXT: v_mov_b32_e32 v5, v1
+; GFX8-NEXT: v_lshrrev_b32_sdwa v1, v2, v5 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:DWORD
+; GFX8-NEXT: v_max_f32_e32 v1, v1, v0
+; GFX8-NEXT: v_bfe_u32 v6, v1, 16, 1
+; GFX8-NEXT: v_add_u32_e32 v6, vcc, v6, v1
; GFX8-NEXT: v_add_u32_e32 v6, vcc, 0x7fff, v6
-; GFX8-NEXT: v_or_b32_e32 v7, 0x400000, v3
-; GFX8-NEXT: v_cmp_u_f32_e32 vcc, v3, v3
-; GFX8-NEXT: v_cndmask_b32_e32 v3, v6, v7, vcc
-; GFX8-NEXT: v_and_b32_e32 v2, s7, v1
-; GFX8-NEXT: v_lshlrev_b32_sdwa v0, v0, v3 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:WORD_1
-; GFX8-NEXT: v_or_b32_e32 v0, v2, v0
-; GFX8-NEXT: v_mov_b32_e32 v3, v1
-; GFX8-NEXT: v_mov_b32_e32 v2, v0
-; GFX8-NEXT: buffer_atomic_cmpswap v[2:3], v4, s[16:19], 0 offen glc
+; GFX8-NEXT: v_or_b32_e32 v7, 0x400000, v1
+; GFX8-NEXT: v_cmp_u_f32_e32 vcc, v1, v1
+; GFX8-NEXT: v_cndmask_b32_e32 v1, v6, v7, vcc
+; GFX8-NEXT: v_and_b32_e32 v4, s7, v5
+; GFX8-NEXT: v_lshlrev_b32_sdwa v1, v2, v1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:WORD_1
+; GFX8-NEXT: v_or_b32_e32 v4, v4, v1
+; GFX8-NEXT: v_mov_b32_e32 v1, v4
+; GFX8-NEXT: v_mov_b32_e32 v2, v5
+; GFX8-NEXT: buffer_atomic_cmpswap v[1:2], v3, s[16:19], 0 offen glc
; GFX8-NEXT: s_waitcnt vmcnt(0)
; GFX8-NEXT: buffer_wbinvl1
-; GFX8-NEXT: v_cmp_eq_u32_e32 vcc, v2, v1
+; GFX8-NEXT: v_cmp_eq_u32_e32 vcc, v1, v5
; GFX8-NEXT: s_or_b64 s[4:5], vcc, s[4:5]
-; GFX8-NEXT: v_mov_b32_e32 v1, v2
; GFX8-NEXT: s_andn2_b64 exec, exec, s[4:5]
; GFX8-NEXT: s_cbranch_execnz .LBB13_1
; GFX8-NEXT: ; %bb.2: ; %atomicrmw.end
; GFX8-NEXT: s_or_b64 exec, exec, s[4:5]
-; GFX8-NEXT: v_lshrrev_b32_e32 v0, s6, v2
+; GFX8-NEXT: v_lshrrev_b32_e32 v0, s6, v1
; GFX8-NEXT: s_setpc_b64 s[30:31]
;
; GFX7-LABEL: buffer_fat_ptr_agent_atomic_fmax_ret_bf16__offset__amdgpu_no_fine_grained_memory:
@@ -4650,39 +4667,39 @@ define bfloat @buffer_fat_ptr_agent_atomic_fmax_ret_bf16__offset__amdgpu_no_fine
; GFX7-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX7-NEXT: s_addk_i32 s20, 0x200
; GFX7-NEXT: s_and_b32 s4, s20, -4
-; GFX7-NEXT: v_mov_b32_e32 v4, s4
-; GFX7-NEXT: buffer_load_dword v1, v4, s[16:19], 0 offen
+; GFX7-NEXT: v_mov_b32_e32 v3, s4
+; GFX7-NEXT: buffer_load_dword v1, v3, s[16:19], 0 offen
; GFX7-NEXT: s_and_b32 s4, s20, 3
; GFX7-NEXT: s_lshl_b32 s6, s4, 3
; GFX7-NEXT: s_lshl_b32 s4, 0xffff, s6
; GFX7-NEXT: v_mul_f32_e32 v0, 1.0, v0
; GFX7-NEXT: s_not_b32 s7, s4
; GFX7-NEXT: s_mov_b64 s[4:5], 0
-; GFX7-NEXT: v_and_b32_e32 v5, 0xffff0000, v0
+; GFX7-NEXT: v_and_b32_e32 v0, 0xffff0000, v0
; GFX7-NEXT: .LBB13_1: ; %atomicrmw.start
; GFX7-NEXT: ; =>This Inner Loop Header: Depth=1
; GFX7-NEXT: s_waitcnt vmcnt(0)
-; GFX7-NEXT: v_lshrrev_b32_e32 v0, s6, v1
-; GFX7-NEXT: v_lshlrev_b32_e32 v0, 16, v0
-; GFX7-NEXT: v_mul_f32_e32 v0, 1.0, v0
-; GFX7-NEXT: v_max_f32_e32 v0, v0, v5
-; GFX7-NEXT: v_lshrrev_b32_e32 v0, 16, v0
-; GFX7-NEXT: v_and_b32_e32 v2, s7, v1
-; GFX7-NEXT: v_lshlrev_b32_e32 v0, s6, v0
-; GFX7-NEXT: v_or_b32_e32 v0, v2, v0
-; GFX7-NEXT: v_mov_b32_e32 v3, v1
-; GFX7-NEXT: v_mov_b32_e32 v2, v0
-; GFX7-NEXT: buffer_atomic_cmpswap v[2:3], v4, s[16:19], 0 offen glc
+; GFX7-NEXT: v_mov_b32_e32 v5, v1
+; GFX7-NEXT: v_lshrrev_b32_e32 v1, s6, v5
+; GFX7-NEXT: v_lshlrev_b32_e32 v1, 16, v1
+; GFX7-NEXT: v_mul_f32_e32 v1, 1.0, v1
+; GFX7-NEXT: v_max_f32_e32 v1, v1, v0
+; GFX7-NEXT: v_lshrrev_b32_e32 v1, 16, v1
+; GFX7-NEXT: v_and_b32_e32 v2, s7, v5
+; GFX7-NEXT: v_lshlrev_b32_e32 v1, s6, v1
+; GFX7-NEXT: v_or_b32_e32 v4, v2, v1
+; GFX7-NEXT: v_mov_b32_e32 v1, v4
+; GFX7-NEXT: v_mov_b32_e32 v2, v5
+; GFX7-NEXT: buffer_atomic_cmpswap v[1:2], v3, s[16:19], 0 offen glc
; GFX7-NEXT: s_waitcnt vmcnt(0)
; GFX7-NEXT: buffer_wbinvl1
-; GFX7-NEXT: v_cmp_eq_u32_e32 vcc, v2, v1
+; GFX7-NEXT: v_cmp_eq_u32_e32 vcc, v1, v5
; GFX7-NEXT: s_or_b64 s[4:5], vcc, s[4:5]
-; GFX7-NEXT: v_mov_b32_e32 v1, v2
; GFX7-NEXT: s_andn2_b64 exec, exec, s[4:5]
; GFX7-NEXT: s_cbranch_execnz .LBB13_1
; GFX7-NEXT: ; %bb.2: ; %atomicrmw.end
; GFX7-NEXT: s_or_b64 exec, exec, s[4:5]
-; GFX7-NEXT: v_lshrrev_b32_e32 v0, s6, v2
+; GFX7-NEXT: v_lshrrev_b32_e32 v0, s6, v1
; GFX7-NEXT: v_lshlrev_b32_e32 v0, 16, v0
; GFX7-NEXT: s_setpc_b64 s[30:31]
;
@@ -4691,40 +4708,40 @@ define bfloat @buffer_fat_ptr_agent_atomic_fmax_ret_bf16__offset__amdgpu_no_fine
; GFX6-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX6-NEXT: s_addk_i32 s20, 0x200
; GFX6-NEXT: s_and_b32 s4, s20, -4
-; GFX6-NEXT: v_mov_b32_e32 v4, s4
-; GFX6-NEXT: buffer_load_dword v1, v4, s[16:19], 0 offen
+; GFX6-NEXT: v_mov_b32_e32 v3, s4
+; GFX6-NEXT: buffer_load_dword v1, v3, s[16:19], 0 offen
; GFX6-NEXT: s_and_b32 s4, s20, 3
; GFX6-NEXT: s_lshl_b32 s6, s4, 3
; GFX6-NEXT: s_lshl_b32 s4, 0xffff, s6
; GFX6-NEXT: v_mul_f32_e32 v0, 1.0, v0
; GFX6-NEXT: s_not_b32 s7, s4
; GFX6-NEXT: s_mov_b64 s[4:5], 0
-; GFX6-NEXT: v_and_b32_e32 v5, 0xffff0000, v0
+; GFX6-NEXT: v_and_b32_e32 v0, 0xffff0000, v0
; GFX6-NEXT: .LBB13_1: ; %atomicrmw.start
; GFX6-NEXT: ; =>This Inner Loop Header: Depth=1
; GFX6-NEXT: s_waitcnt vmcnt(0)
-; GFX6-NEXT: v_lshrrev_b32_e32 v0, s6, v1
-; GFX6-NEXT: v_lshlrev_b32_e32 v0, 16, v0
-; GFX6-NEXT: v_mul_f32_e32 v0, 1.0, v0
-; GFX6-NEXT: v_max_f32_e32 v0, v0, v5
-; GFX6-NEXT: v_lshrrev_b32_e32 v0, 16, v0
+; GFX6-NEXT: v_mov_b32_e32 v5, v1
; GFX6-NEXT: s_waitcnt expcnt(0)
-; GFX6-NEXT: v_and_b32_e32 v2, s7, v1
-; GFX6-NEXT: v_lshlrev_b32_e32 v0, s6, v0
-; GFX6-NEXT: v_or_b32_e32 v0, v2, v0
-; GFX6-NEXT: v_mov_b32_e32 v3, v1
-; GFX6-NEXT: v_mov_b32_e32 v2, v0
-; GFX6-NEXT: buffer_atomic_cmpswap v[2:3], v4, s[16:19], 0 offen glc
+; GFX6-NEXT: v_lshrrev_b32_e32 v1, s6, v5
+; GFX6-NEXT: v_lshlrev_b32_e32 v1, 16, v1
+; GFX6-NEXT: v_mul_f32_e32 v1, 1.0, v1
+; GFX6-NEXT: v_max_f32_e32 v1, v1, v0
+; GFX6-NEXT: v_lshrrev_b32_e32 v1, 16, v1
+; GFX6-NEXT: v_and_b32_e32 v2, s7, v5
+; GFX6-NEXT: v_lshlrev_b32_e32 v1, s6, v1
+; GFX6-NEXT: v_or_b32_e32 v4, v2, v1
+; GFX6-NEXT: v_mov_b32_e32 v1, v4
+; GFX6-NEXT: v_mov_b32_e32 v2, v5
+; GFX6-NEXT: buffer_atomic_cmpswap v[1:2], v3, s[16:19], 0 offen glc
; GFX6-NEXT: s_waitcnt vmcnt(0)
; GFX6-NEXT: buffer_wbinvl1
-; GFX6-NEXT: v_cmp_eq_u32_e32 vcc, v2, v1
+; GFX6-NEXT: v_cmp_eq_u32_e32 vcc, v1, v5
; GFX6-NEXT: s_or_b64 s[4:5], vcc, s[4:5]
-; GFX6-NEXT: v_mov_b32_e32 v1, v2
; GFX6-NEXT: s_andn2_b64 exec, exec, s[4:5]
; GFX6-NEXT: s_cbranch_execnz .LBB13_1
; GFX6-NEXT: ; %bb.2: ; %atomicrmw.end
; GFX6-NEXT: s_or_b64 exec, exec, s[4:5]
-; GFX6-NEXT: v_lshrrev_b32_e32 v0, s6, v2
+; GFX6-NEXT: v_lshrrev_b32_e32 v0, s6, v1
; GFX6-NEXT: v_lshlrev_b32_e32 v0, 16, v0
; GFX6-NEXT: s_waitcnt expcnt(0)
; GFX6-NEXT: s_setpc_b64 s[30:31]
@@ -4742,48 +4759,48 @@ define void @buffer_fat_ptr_agent_atomic_fmax_noret_bf16__offset__amdgpu_no_fine
; GFX12-TRUE16-NEXT: s_wait_bvhcnt 0x0
; GFX12-TRUE16-NEXT: s_wait_kmcnt 0x0
; GFX12-TRUE16-NEXT: s_addk_co_i32 s16, 0x200
-; GFX12-TRUE16-NEXT: v_lshlrev_b32_e32 v3, 16, v0
+; GFX12-TRUE16-NEXT: v_lshlrev_b32_e32 v0, 16, v0
; GFX12-TRUE16-NEXT: s_wait_alu 0xfffe
; GFX12-TRUE16-NEXT: s_and_b32 s4, s16, -4
; GFX12-TRUE16-NEXT: s_wait_alu 0xfffe
-; GFX12-TRUE16-NEXT: v_mov_b32_e32 v2, s4
+; GFX12-TRUE16-NEXT: v_mov_b32_e32 v3, s4
; GFX12-TRUE16-NEXT: s_and_b32 s4, s16, 3
; GFX12-TRUE16-NEXT: s_wait_alu 0xfffe
; GFX12-TRUE16-NEXT: s_lshl_b32 s4, s4, 3
; GFX12-TRUE16-NEXT: s_wait_alu 0xfffe
; GFX12-TRUE16-NEXT: s_lshl_b32 s5, 0xffff, s4
-; GFX12-TRUE16-NEXT: buffer_load_b32 v1, v2, s[0:3], null offen
+; GFX12-TRUE16-NEXT: buffer_load_b32 v1, v3, s[0:3], null offen
; GFX12-TRUE16-NEXT: s_not_b32 s6, s5
; GFX12-TRUE16-NEXT: s_mov_b32 s5, 0
; GFX12-TRUE16-NEXT: .LBB14_1: ; %atomicrmw.start
; GFX12-TRUE16-NEXT: ; =>This Inner Loop Header: Depth=1
; GFX12-TRUE16-NEXT: s_wait_loadcnt 0x0
-; GFX12-TRUE16-NEXT: v_lshrrev_b32_e32 v0, s4, v1
+; GFX12-TRUE16-NEXT: v_mov_b32_e32 v5, v1
; GFX12-TRUE16-NEXT: s_wait_storecnt 0x0
; GFX12-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
-; GFX12-TRUE16-NEXT: v_lshlrev_b32_e32 v0, 16, v0
-; GFX12-TRUE16-NEXT: v_max_num_f32_e32 v0, v0, v3
-; GFX12-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(SKIP_2) | instid1(VALU_DEP_3)
-; GFX12-TRUE16-NEXT: v_bfe_u32 v4, v0, 16, 1
-; GFX12-TRUE16-NEXT: v_or_b32_e32 v5, 0x400000, v0
-; GFX12-TRUE16-NEXT: v_cmp_u_f32_e32 vcc_lo, v0, v0
-; GFX12-TRUE16-NEXT: v_add3_u32 v4, v4, v0, 0x7fff
+; GFX12-TRUE16-NEXT: v_lshrrev_b32_e32 v1, s4, v5
+; GFX12-TRUE16-NEXT: v_lshlrev_b32_e32 v1, 16, v1
+; GFX12-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
+; GFX12-TRUE16-NEXT: v_max_num_f32_e32 v1, v1, v0
+; GFX12-TRUE16-NEXT: v_bfe_u32 v2, v1, 16, 1
+; GFX12-TRUE16-NEXT: v_or_b32_e32 v4, 0x400000, v1
+; GFX12-TRUE16-NEXT: v_cmp_u_f32_e32 vcc_lo, v1, v1
+; GFX12-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_3) | instskip(SKIP_1) | instid1(VALU_DEP_1)
+; GFX12-TRUE16-NEXT: v_add3_u32 v2, v2, v1, 0x7fff
; GFX12-TRUE16-NEXT: s_wait_alu 0xfffd
-; GFX12-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(SKIP_1) | instid1(VALU_DEP_2)
-; GFX12-TRUE16-NEXT: v_cndmask_b32_e32 v0, v4, v5, vcc_lo
-; GFX12-TRUE16-NEXT: v_mov_b16_e32 v4.h, 0
-; GFX12-TRUE16-NEXT: v_mov_b16_e32 v4.l, v0.h
-; GFX12-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(SKIP_1) | instid1(VALU_DEP_1)
-; GFX12-TRUE16-NEXT: v_lshlrev_b32_e32 v0, s4, v4
+; GFX12-TRUE16-NEXT: v_cndmask_b32_e32 v1, v2, v4, vcc_lo
+; GFX12-TRUE16-NEXT: v_mov_b16_e32 v2.h, 0
+; GFX12-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_2) | instskip(NEXT) | instid1(VALU_DEP_1)
+; GFX12-TRUE16-NEXT: v_mov_b16_e32 v2.l, v1.h
+; GFX12-TRUE16-NEXT: v_lshlrev_b32_e32 v1, s4, v2
; GFX12-TRUE16-NEXT: s_wait_alu 0xfffe
-; GFX12-TRUE16-NEXT: v_and_or_b32 v0, v1, s6, v0
-; GFX12-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1)
-; GFX12-TRUE16-NEXT: v_dual_mov_b32 v5, v1 :: v_dual_mov_b32 v4, v0
-; GFX12-TRUE16-NEXT: buffer_atomic_cmpswap_b32 v[4:5], v2, s[0:3], null offen th:TH_ATOMIC_RETURN
+; GFX12-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
+; GFX12-TRUE16-NEXT: v_and_or_b32 v4, v5, s6, v1
+; GFX12-TRUE16-NEXT: v_dual_mov_b32 v1, v4 :: v_dual_mov_b32 v2, v5
+; GFX12-TRUE16-NEXT: buffer_atomic_cmpswap_b32 v[1:2], v3, s[0:3], null offen th:TH_ATOMIC_RETURN
; GFX12-TRUE16-NEXT: s_wait_loadcnt 0x0
; GFX12-TRUE16-NEXT: global_inv scope:SCOPE_DEV
-; GFX12-TRUE16-NEXT: v_cmp_eq_u32_e32 vcc_lo, v4, v1
-; GFX12-TRUE16-NEXT: v_mov_b32_e32 v1, v4
+; GFX12-TRUE16-NEXT: v_cmp_eq_u32_e32 vcc_lo, v1, v5
; GFX12-TRUE16-NEXT: s_or_b32 s5, vcc_lo, s5
; GFX12-TRUE16-NEXT: s_wait_alu 0xfffe
; GFX12-TRUE16-NEXT: s_and_not1_b32 exec_lo, exec_lo, s5
@@ -4801,47 +4818,47 @@ define void @buffer_fat_ptr_agent_atomic_fmax_noret_bf16__offset__amdgpu_no_fine
; GFX12-FAKE16-NEXT: s_wait_bvhcnt 0x0
; GFX12-FAKE16-NEXT: s_wait_kmcnt 0x0
; GFX12-FAKE16-NEXT: s_addk_co_i32 s16, 0x200
-; GFX12-FAKE16-NEXT: v_lshlrev_b32_e32 v3, 16, v0
+; GFX12-FAKE16-NEXT: v_lshlrev_b32_e32 v0, 16, v0
; GFX12-FAKE16-NEXT: s_wait_alu 0xfffe
; GFX12-FAKE16-NEXT: s_and_b32 s4, s16, -4
; GFX12-FAKE16-NEXT: s_wait_alu 0xfffe
-; GFX12-FAKE16-NEXT: v_mov_b32_e32 v2, s4
+; GFX12-FAKE16-NEXT: v_mov_b32_e32 v3, s4
; GFX12-FAKE16-NEXT: s_and_b32 s4, s16, 3
; GFX12-FAKE16-NEXT: s_wait_alu 0xfffe
; GFX12-FAKE16-NEXT: s_lshl_b32 s4, s4, 3
; GFX12-FAKE16-NEXT: s_wait_alu 0xfffe
; GFX12-FAKE16-NEXT: s_lshl_b32 s5, 0xffff, s4
-; GFX12-FAKE16-NEXT: buffer_load_b32 v1, v2, s[0:3], null offen
+; GFX12-FAKE16-NEXT: buffer_load_b32 v1, v3, s[0:3], null offen
; GFX12-FAKE16-NEXT: s_not_b32 s6, s5
; GFX12-FAKE16-NEXT: s_mov_b32 s5, 0
; GFX12-FAKE16-NEXT: .LBB14_1: ; %atomicrmw.start
; GFX12-FAKE16-NEXT: ; =>This Inner Loop Header: Depth=1
; GFX12-FAKE16-NEXT: s_wait_loadcnt 0x0
-; GFX12-FAKE16-NEXT: v_lshrrev_b32_e32 v0, s4, v1
+; GFX12-FAKE16-NEXT: v_mov_b32_e32 v5, v1
; GFX12-FAKE16-NEXT: s_wait_storecnt 0x0
; GFX12-FAKE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
-; GFX12-FAKE16-NEXT: v_lshlrev_b32_e32 v0, 16, v0
-; GFX12-FAKE16-NEXT: v_max_num_f32_e32 v0, v0, v3
-; GFX12-FAKE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(SKIP_2) | instid1(VALU_DEP_3)
-; GFX12-FAKE16-NEXT: v_bfe_u32 v4, v0, 16, 1
-; GFX12-FAKE16-NEXT: v_or_b32_e32 v5, 0x400000, v0
-; GFX12-FAKE16-NEXT: v_cmp_u_f32_e32 vcc_lo, v0, v0
-; GFX12-FAKE16-NEXT: v_add3_u32 v4, v4, v0, 0x7fff
+; GFX12-FAKE16-NEXT: v_lshrrev_b32_e32 v1, s4, v5
+; GFX12-FAKE16-NEXT: v_lshlrev_b32_e32 v1, 16, v1
+; GFX12-FAKE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
+; GFX12-FAKE16-NEXT: v_max_num_f32_e32 v1, v1, v0
+; GFX12-FAKE16-NEXT: v_bfe_u32 v2, v1, 16, 1
+; GFX12-FAKE16-NEXT: v_or_b32_e32 v4, 0x400000, v1
+; GFX12-FAKE16-NEXT: v_cmp_u_f32_e32 vcc_lo, v1, v1
+; GFX12-FAKE16-NEXT: s_delay_alu instid0(VALU_DEP_3) | instskip(SKIP_1) | instid1(VALU_DEP_1)
+; GFX12-FAKE16-NEXT: v_add3_u32 v2, v2, v1, 0x7fff
; GFX12-FAKE16-NEXT: s_wait_alu 0xfffd
+; GFX12-FAKE16-NEXT: v_cndmask_b32_e32 v1, v2, v4, vcc_lo
; GFX12-FAKE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
-; GFX12-FAKE16-NEXT: v_cndmask_b32_e32 v0, v4, v5, vcc_lo
-; GFX12-FAKE16-NEXT: v_lshrrev_b32_e32 v0, 16, v0
-; GFX12-FAKE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(SKIP_1) | instid1(VALU_DEP_1)
-; GFX12-FAKE16-NEXT: v_lshlrev_b32_e32 v0, s4, v0
+; GFX12-FAKE16-NEXT: v_lshrrev_b32_e32 v1, 16, v1
+; GFX12-FAKE16-NEXT: v_lshlrev_b32_e32 v1, s4, v1
; GFX12-FAKE16-NEXT: s_wait_alu 0xfffe
-; GFX12-FAKE16-NEXT: v_and_or_b32 v0, v1, s6, v0
-; GFX12-FAKE16-NEXT: s_delay_alu instid0(VALU_DEP_1)
-; GFX12-FAKE16-NEXT: v_dual_mov_b32 v5, v1 :: v_dual_mov_b32 v4, v0
-; GFX12-FAKE16-NEXT: buffer_atomic_cmpswap_b32 v[4:5], v2, s[0:3], null offen th:TH_ATOMIC_RETURN
+; GFX12-FAKE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
+; GFX12-FAKE16-NEXT: v_and_or_b32 v4, v5, s6, v1
+; GFX12-FAKE16-NEXT: v_dual_mov_b32 v1, v4 :: v_dual_mov_b32 v2, v5
+; GFX12-FAKE16-NEXT: buffer_atomic_cmpswap_b32 v[1:2], v3, s[0:3], null offen th:TH_ATOMIC_RETURN
; GFX12-FAKE16-NEXT: s_wait_loadcnt 0x0
; GFX12-FAKE16-NEXT: global_inv scope:SCOPE_DEV
-; GFX12-FAKE16-NEXT: v_cmp_eq_u32_e32 vcc_lo, v4, v1
-; GFX12-FAKE16-NEXT: v_mov_b32_e32 v1, v4
+; GFX12-FAKE16-NEXT: v_cmp_eq_u32_e32 vcc_lo, v1, v5
; GFX12-FAKE16-NEXT: s_or_b32 s5, vcc_lo, s5
; GFX12-FAKE16-NEXT: s_wait_alu 0xfffe
; GFX12-FAKE16-NEXT: s_and_not1_b32 exec_lo, exec_lo, s5
@@ -4856,36 +4873,36 @@ define void @buffer_fat_ptr_agent_atomic_fmax_noret_bf16__offset__amdgpu_no_fine
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX942-NEXT: s_addk_i32 s16, 0x200
; GFX942-NEXT: s_and_b32 s4, s16, -4
-; GFX942-NEXT: v_mov_b32_e32 v2, s4
-; GFX942-NEXT: buffer_load_dword v1, v2, s[0:3], 0 offen
+; GFX942-NEXT: v_mov_b32_e32 v1, s4
+; GFX942-NEXT: buffer_load_dword v2, v1, s[0:3], 0 offen
; GFX942-NEXT: s_and_b32 s4, s16, 3
; GFX942-NEXT: s_lshl_b32 s6, s4, 3
; GFX942-NEXT: s_lshl_b32 s4, 0xffff, s6
; GFX942-NEXT: s_not_b32 s7, s4
; GFX942-NEXT: s_mov_b64 s[4:5], 0
-; GFX942-NEXT: v_lshlrev_b32_e32 v3, 16, v0
+; GFX942-NEXT: v_lshlrev_b32_e32 v0, 16, v0
; GFX942-NEXT: s_movk_i32 s8, 0x7fff
; GFX942-NEXT: .LBB14_1: ; %atomicrmw.start
; GFX942-NEXT: ; =>This Inner Loop Header: Depth=1
; GFX942-NEXT: s_waitcnt vmcnt(0)
-; GFX942-NEXT: v_lshrrev_b32_sdwa v0, s6, v1 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:DWORD
+; GFX942-NEXT: v_mov_b32_e32 v5, v2
+; GFX942-NEXT: v_lshrrev_b32_sdwa v2, s6, v5 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:DWORD
; GFX942-NEXT: buffer_wbl2 sc1
-; GFX942-NEXT: v_max_f32_e32 v0, v0, v3
-; GFX942-NEXT: v_bfe_u32 v4, v0, 16, 1
-; GFX942-NEXT: v_or_b32_e32 v5, 0x400000, v0
-; GFX942-NEXT: v_add3_u32 v4, v4, v0, s8
-; GFX942-NEXT: v_cmp_u_f32_e32 vcc, v0, v0
+; GFX942-NEXT: v_max_f32_e32 v2, v2, v0
+; GFX942-NEXT: v_bfe_u32 v3, v2, 16, 1
+; GFX942-NEXT: v_or_b32_e32 v4, 0x400000, v2
+; GFX942-NEXT: v_add3_u32 v3, v3, v2, s8
+; GFX942-NEXT: v_cmp_u_f32_e32 vcc, v2, v2
; GFX942-NEXT: s_nop 1
-; GFX942-NEXT: v_cndmask_b32_e32 v0, v4, v5, vcc
-; GFX942-NEXT: v_lshlrev_b32_sdwa v0, s6, v0 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:WORD_1
-; GFX942-NEXT: v_and_or_b32 v0, v1, s7, v0
-; GFX942-NEXT: v_mov_b64_e32 v[4:5], v[0:1]
-; GFX942-NEXT: buffer_atomic_cmpswap v[4:5], v2, s[0:3], 0 offen sc0
+; GFX942-NEXT: v_cndmask_b32_e32 v2, v3, v4, vcc
+; GFX942-NEXT: v_lshlrev_b32_sdwa v2, s6, v2 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:WORD_1
+; GFX942-NEXT: v_and_or_b32 v4, v5, s7, v2
+; GFX942-NEXT: v_mov_b64_e32 v[2:3], v[4:5]
+; GFX942-NEXT: buffer_atomic_cmpswap v[2:3], v1, s[0:3], 0 offen sc0
; GFX942-NEXT: s_waitcnt vmcnt(0)
; GFX942-NEXT: buffer_inv sc1
-; GFX942-NEXT: v_cmp_eq_u32_e32 vcc, v4, v1
+; GFX942-NEXT: v_cmp_eq_u32_e32 vcc, v2, v5
; GFX942-NEXT: s_or_b64 s[4:5], vcc, s[4:5]
-; GFX942-NEXT: v_mov_b32_e32 v1, v4
; GFX942-NEXT: s_andn2_b64 exec, exec, s[4:5]
; GFX942-NEXT: s_cbranch_execnz .LBB14_1
; GFX942-NEXT: ; %bb.2: ; %atomicrmw.end
@@ -4896,46 +4913,46 @@ define void @buffer_fat_ptr_agent_atomic_fmax_noret_bf16__offset__amdgpu_no_fine
; GFX11-TRUE16: ; %bb.0:
; GFX11-TRUE16-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX11-TRUE16-NEXT: s_addk_i32 s16, 0x200
-; GFX11-TRUE16-NEXT: v_lshlrev_b32_e32 v3, 16, v0
+; GFX11-TRUE16-NEXT: v_lshlrev_b32_e32 v0, 16, v0
; GFX11-TRUE16-NEXT: s_and_b32 s4, s16, -4
; GFX11-TRUE16-NEXT: s_delay_alu instid0(SALU_CYCLE_1) | instskip(SKIP_1) | instid1(SALU_CYCLE_1)
-; GFX11-TRUE16-NEXT: v_mov_b32_e32 v2, s4
+; GFX11-TRUE16-NEXT: v_mov_b32_e32 v3, s4
; GFX11-TRUE16-NEXT: s_and_b32 s4, s16, 3
; GFX11-TRUE16-NEXT: s_lshl_b32 s4, s4, 3
; GFX11-TRUE16-NEXT: s_delay_alu instid0(SALU_CYCLE_1)
; GFX11-TRUE16-NEXT: s_lshl_b32 s5, 0xffff, s4
-; GFX11-TRUE16-NEXT: buffer_load_b32 v1, v2, s[0:3], 0 offen
+; GFX11-TRUE16-NEXT: buffer_load_b32 v1, v3, s[0:3], 0 offen
; GFX11-TRUE16-NEXT: s_not_b32 s6, s5
; GFX11-TRUE16-NEXT: s_mov_b32 s5, 0
; GFX11-TRUE16-NEXT: .p2align 6
; GFX11-TRUE16-NEXT: .LBB14_1: ; %atomicrmw.start
; GFX11-TRUE16-NEXT: ; =>This Inner Loop Header: Depth=1
; GFX11-TRUE16-NEXT: s_waitcnt vmcnt(0)
-; GFX11-TRUE16-NEXT: v_lshrrev_b32_e32 v0, s4, v1
+; GFX11-TRUE16-NEXT: v_mov_b32_e32 v5, v1
; GFX11-TRUE16-NEXT: s_waitcnt_vscnt null, 0x0
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
-; GFX11-TRUE16-NEXT: v_lshlrev_b32_e32 v0, 16, v0
-; GFX11-TRUE16-NEXT: v_max_f32_e32 v0, v0, v3
-; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(SKIP_2) | instid1(VALU_DEP_3)
-; GFX11-TRUE16-NEXT: v_bfe_u32 v4, v0, 16, 1
-; GFX11-TRUE16-NEXT: v_or_b32_e32 v5, 0x400000, v0
-; GFX11-TRUE16-NEXT: v_cmp_u_f32_e32 vcc_lo, v0, v0
-; GFX11-TRUE16-NEXT: v_add3_u32 v4, v4, v0, 0x7fff
-; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(SKIP_1) | instid1(VALU_DEP_2)
-; GFX11-TRUE16-NEXT: v_cndmask_b32_e32 v0, v4, v5, vcc_lo
-; GFX11-TRUE16-NEXT: v_mov_b16_e32 v4.h, 0
-; GFX11-TRUE16-NEXT: v_mov_b16_e32 v4.l, v0.h
+; GFX11-TRUE16-NEXT: v_lshrrev_b32_e32 v1, s4, v5
+; GFX11-TRUE16-NEXT: v_lshlrev_b32_e32 v1, 16, v1
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
-; GFX11-TRUE16-NEXT: v_lshlrev_b32_e32 v0, s4, v4
-; GFX11-TRUE16-NEXT: v_and_or_b32 v0, v1, s6, v0
-; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1)
-; GFX11-TRUE16-NEXT: v_dual_mov_b32 v5, v1 :: v_dual_mov_b32 v4, v0
-; GFX11-TRUE16-NEXT: buffer_atomic_cmpswap_b32 v[4:5], v2, s[0:3], 0 offen glc
+; GFX11-TRUE16-NEXT: v_max_f32_e32 v1, v1, v0
+; GFX11-TRUE16-NEXT: v_bfe_u32 v2, v1, 16, 1
+; GFX11-TRUE16-NEXT: v_or_b32_e32 v4, 0x400000, v1
+; GFX11-TRUE16-NEXT: v_cmp_u_f32_e32 vcc_lo, v1, v1
+; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_3) | instskip(NEXT) | instid1(VALU_DEP_1)
+; GFX11-TRUE16-NEXT: v_add3_u32 v2, v2, v1, 0x7fff
+; GFX11-TRUE16-NEXT: v_cndmask_b32_e32 v1, v2, v4, vcc_lo
+; GFX11-TRUE16-NEXT: v_mov_b16_e32 v2.h, 0
+; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_2) | instskip(NEXT) | instid1(VALU_DEP_1)
+; GFX11-TRUE16-NEXT: v_mov_b16_e32 v2.l, v1.h
+; GFX11-TRUE16-NEXT: v_lshlrev_b32_e32 v1, s4, v2
+; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
+; GFX11-TRUE16-NEXT: v_and_or_b32 v4, v5, s6, v1
+; GFX11-TRUE16-NEXT: v_dual_mov_b32 v1, v4 :: v_dual_mov_b32 v2, v5
+; GFX11-TRUE16-NEXT: buffer_atomic_cmpswap_b32 v[1:2], v3, s[0:3], 0 offen glc
; GFX11-TRUE16-NEXT: s_waitcnt vmcnt(0)
; GFX11-TRUE16-NEXT: buffer_gl1_inv
; GFX11-TRUE16-NEXT: buffer_gl0_inv
-; GFX11-TRUE16-NEXT: v_cmp_eq_u32_e32 vcc_lo, v4, v1
-; GFX11-TRUE16-NEXT: v_mov_b32_e32 v1, v4
+; GFX11-TRUE16-NEXT: v_cmp_eq_u32_e32 vcc_lo, v1, v5
; GFX11-TRUE16-NEXT: s_or_b32 s5, vcc_lo, s5
; GFX11-TRUE16-NEXT: s_delay_alu instid0(SALU_CYCLE_1)
; GFX11-TRUE16-NEXT: s_and_not1_b32 exec_lo, exec_lo, s5
@@ -4948,45 +4965,45 @@ define void @buffer_fat_ptr_agent_atomic_fmax_noret_bf16__offset__amdgpu_no_fine
; GFX11-FAKE16: ; %bb.0:
; GFX11-FAKE16-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX11-FAKE16-NEXT: s_addk_i32 s16, 0x200
-; GFX11-FAKE16-NEXT: v_lshlrev_b32_e32 v3, 16, v0
+; GFX11-FAKE16-NEXT: v_lshlrev_b32_e32 v0, 16, v0
; GFX11-FAKE16-NEXT: s_and_b32 s4, s16, -4
; GFX11-FAKE16-NEXT: s_delay_alu instid0(SALU_CYCLE_1) | instskip(SKIP_1) | instid1(SALU_CYCLE_1)
-; GFX11-FAKE16-NEXT: v_mov_b32_e32 v2, s4
+; GFX11-FAKE16-NEXT: v_mov_b32_e32 v3, s4
; GFX11-FAKE16-NEXT: s_and_b32 s4, s16, 3
; GFX11-FAKE16-NEXT: s_lshl_b32 s4, s4, 3
; GFX11-FAKE16-NEXT: s_delay_alu instid0(SALU_CYCLE_1)
; GFX11-FAKE16-NEXT: s_lshl_b32 s5, 0xffff, s4
-; GFX11-FAKE16-NEXT: buffer_load_b32 v1, v2, s[0:3], 0 offen
+; GFX11-FAKE16-NEXT: buffer_load_b32 v1, v3, s[0:3], 0 offen
; GFX11-FAKE16-NEXT: s_not_b32 s6, s5
; GFX11-FAKE16-NEXT: s_mov_b32 s5, 0
; GFX11-FAKE16-NEXT: .p2align 6
; GFX11-FAKE16-NEXT: .LBB14_1: ; %atomicrmw.start
; GFX11-FAKE16-NEXT: ; =>This Inner Loop Header: Depth=1
; GFX11-FAKE16-NEXT: s_waitcnt vmcnt(0)
-; GFX11-FAKE16-NEXT: v_lshrrev_b32_e32 v0, s4, v1
+; GFX11-FAKE16-NEXT: v_mov_b32_e32 v5, v1
; GFX11-FAKE16-NEXT: s_waitcnt_vscnt null, 0x0
; GFX11-FAKE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
-; GFX11-FAKE16-NEXT: v_lshlrev_b32_e32 v0, 16, v0
-; GFX11-FAKE16-NEXT: v_max_f32_e32 v0, v0, v3
-; GFX11-FAKE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(SKIP_2) | instid1(VALU_DEP_3)
-; GFX11-FAKE16-NEXT: v_bfe_u32 v4, v0, 16, 1
-; GFX11-FAKE16-NEXT: v_or_b32_e32 v5, 0x400000, v0
-; GFX11-FAKE16-NEXT: v_cmp_u_f32_e32 vcc_lo, v0, v0
-; GFX11-FAKE16-NEXT: v_add3_u32 v4, v4, v0, 0x7fff
+; GFX11-FAKE16-NEXT: v_lshrrev_b32_e32 v1, s4, v5
+; GFX11-FAKE16-NEXT: v_lshlrev_b32_e32 v1, 16, v1
; GFX11-FAKE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
-; GFX11-FAKE16-NEXT: v_cndmask_b32_e32 v0, v4, v5, vcc_lo
-; GFX11-FAKE16-NEXT: v_lshrrev_b32_e32 v0, 16, v0
+; GFX11-FAKE16-NEXT: v_max_f32_e32 v1, v1, v0
+; GFX11-FAKE16-NEXT: v_bfe_u32 v2, v1, 16, 1
+; GFX11-FAKE16-NEXT: v_or_b32_e32 v4, 0x400000, v1
+; GFX11-FAKE16-NEXT: v_cmp_u_f32_e32 vcc_lo, v1, v1
+; GFX11-FAKE16-NEXT: s_delay_alu instid0(VALU_DEP_3) | instskip(NEXT) | instid1(VALU_DEP_1)
+; GFX11-FAKE16-NEXT: v_add3_u32 v2, v2, v1, 0x7fff
+; GFX11-FAKE16-NEXT: v_cndmask_b32_e32 v1, v2, v4, vcc_lo
; GFX11-FAKE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
-; GFX11-FAKE16-NEXT: v_lshlrev_b32_e32 v0, s4, v0
-; GFX11-FAKE16-NEXT: v_and_or_b32 v0, v1, s6, v0
-; GFX11-FAKE16-NEXT: s_delay_alu instid0(VALU_DEP_1)
-; GFX11-FAKE16-NEXT: v_dual_mov_b32 v5, v1 :: v_dual_mov_b32 v4, v0
-; GFX11-FAKE16-NEXT: buffer_atomic_cmpswap_b32 v[4:5], v2, s[0:3], 0 offen glc
+; GFX11-FAKE16-NEXT: v_lshrrev_b32_e32 v1, 16, v1
+; GFX11-FAKE16-NEXT: v_lshlrev_b32_e32 v1, s4, v1
+; GFX11-FAKE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
+; GFX11-FAKE16-NEXT: v_and_or_b32 v4, v5, s6, v1
+; GFX11-FAKE16-NEXT: v_dual_mov_b32 v1, v4 :: v_dual_mov_b32 v2, v5
+; GFX11-FAKE16-NEXT: buffer_atomic_cmpswap_b32 v[1:2], v3, s[0:3], 0 offen glc
; GFX11-FAKE16-NEXT: s_waitcnt vmcnt(0)
; GFX11-FAKE16-NEXT: buffer_gl1_inv
; GFX11-FAKE16-NEXT: buffer_gl0_inv
-; GFX11-FAKE16-NEXT: v_cmp_eq_u32_e32 vcc_lo, v4, v1
-; GFX11-FAKE16-NEXT: v_mov_b32_e32 v1, v4
+; GFX11-FAKE16-NEXT: v_cmp_eq_u32_e32 vcc_lo, v1, v5
; GFX11-FAKE16-NEXT: s_or_b32 s5, vcc_lo, s5
; GFX11-FAKE16-NEXT: s_delay_alu instid0(SALU_CYCLE_1)
; GFX11-FAKE16-NEXT: s_and_not1_b32 exec_lo, exec_lo, s5
@@ -4999,36 +5016,36 @@ define void @buffer_fat_ptr_agent_atomic_fmax_noret_bf16__offset__amdgpu_no_fine
; GFX10: ; %bb.0:
; GFX10-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX10-NEXT: s_addk_i32 s20, 0x200
-; GFX10-NEXT: v_lshlrev_b32_e32 v3, 16, v0
+; GFX10-NEXT: v_lshlrev_b32_e32 v0, 16, v0
; GFX10-NEXT: s_and_b32 s4, s20, -4
-; GFX10-NEXT: v_mov_b32_e32 v2, s4
+; GFX10-NEXT: v_mov_b32_e32 v3, s4
; GFX10-NEXT: s_and_b32 s4, s20, 3
; GFX10-NEXT: s_lshl_b32 s4, s4, 3
; GFX10-NEXT: s_lshl_b32 s5, 0xffff, s4
-; GFX10-NEXT: buffer_load_dword v1, v2, s[16:19], 0 offen
+; GFX10-NEXT: buffer_load_dword v1, v3, s[16:19], 0 offen
; GFX10-NEXT: s_not_b32 s6, s5
; GFX10-NEXT: s_mov_b32 s5, 0
; GFX10-NEXT: .LBB14_1: ; %atomicrmw.start
; GFX10-NEXT: ; =>This Inner Loop Header: Depth=1
; GFX10-NEXT: s_waitcnt vmcnt(0)
-; GFX10-NEXT: v_lshrrev_b32_sdwa v0, s4, v1 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:DWORD
-; GFX10-NEXT: s_waitcnt_vscnt null, 0x0
-; GFX10-NEXT: v_max_f32_e32 v0, v0, v3
-; GFX10-NEXT: v_bfe_u32 v4, v0, 16, 1
-; GFX10-NEXT: v_or_b32_e32 v5, 0x400000, v0
-; GFX10-NEXT: v_cmp_u_f32_e32 vcc_lo, v0, v0
-; GFX10-NEXT: v_add3_u32 v4, v4, v0, 0x7fff
-; GFX10-NEXT: v_cndmask_b32_e32 v0, v4, v5, vcc_lo
-; GFX10-NEXT: v_lshlrev_b32_sdwa v0, s4, v0 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:WORD_1
-; GFX10-NEXT: v_and_or_b32 v0, v1, s6, v0
; GFX10-NEXT: v_mov_b32_e32 v5, v1
-; GFX10-NEXT: v_mov_b32_e32 v4, v0
-; GFX10-NEXT: buffer_atomic_cmpswap v[4:5], v2, s[16:19], 0 offen glc
+; GFX10-NEXT: s_waitcnt_vscnt null, 0x0
+; GFX10-NEXT: v_lshrrev_b32_sdwa v1, s4, v5 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:DWORD
+; GFX10-NEXT: v_max_f32_e32 v1, v1, v0
+; GFX10-NEXT: v_bfe_u32 v2, v1, 16, 1
+; GFX10-NEXT: v_or_b32_e32 v4, 0x400000, v1
+; GFX10-NEXT: v_cmp_u_f32_e32 vcc_lo, v1, v1
+; GFX10-NEXT: v_add3_u32 v2, v2, v1, 0x7fff
+; GFX10-NEXT: v_cndmask_b32_e32 v1, v2, v4, vcc_lo
+; GFX10-NEXT: v_lshlrev_b32_sdwa v1, s4, v1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:WORD_1
+; GFX10-NEXT: v_and_or_b32 v4, v5, s6, v1
+; GFX10-NEXT: v_mov_b32_e32 v1, v4
+; GFX10-NEXT: v_mov_b32_e32 v2, v5
+; GFX10-NEXT: buffer_atomic_cmpswap v[1:2], v3, s[16:19], 0 offen glc
; GFX10-NEXT: s_waitcnt vmcnt(0)
; GFX10-NEXT: buffer_gl1_inv
; GFX10-NEXT: buffer_gl0_inv
-; GFX10-NEXT: v_cmp_eq_u32_e32 vcc_lo, v4, v1
-; GFX10-NEXT: v_mov_b32_e32 v1, v4
+; GFX10-NEXT: v_cmp_eq_u32_e32 vcc_lo, v1, v5
; GFX10-NEXT: s_or_b32 s5, vcc_lo, s5
; GFX10-NEXT: s_andn2_b32 exec_lo, exec_lo, s5
; GFX10-NEXT: s_cbranch_execnz .LBB14_1
@@ -5041,34 +5058,34 @@ define void @buffer_fat_ptr_agent_atomic_fmax_noret_bf16__offset__amdgpu_no_fine
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX90A-NEXT: s_addk_i32 s20, 0x200
; GFX90A-NEXT: s_and_b32 s4, s20, -4
-; GFX90A-NEXT: v_mov_b32_e32 v2, s4
-; GFX90A-NEXT: buffer_load_dword v1, v2, s[16:19], 0 offen
+; GFX90A-NEXT: v_mov_b32_e32 v1, s4
+; GFX90A-NEXT: buffer_load_dword v2, v1, s[16:19], 0 offen
; GFX90A-NEXT: s_and_b32 s4, s20, 3
; GFX90A-NEXT: s_lshl_b32 s6, s4, 3
; GFX90A-NEXT: s_lshl_b32 s4, 0xffff, s6
; GFX90A-NEXT: s_not_b32 s7, s4
; GFX90A-NEXT: s_mov_b64 s[4:5], 0
-; GFX90A-NEXT: v_lshlrev_b32_e32 v3, 16, v0
+; GFX90A-NEXT: v_lshlrev_b32_e32 v0, 16, v0
; GFX90A-NEXT: s_movk_i32 s8, 0x7fff
; GFX90A-NEXT: .LBB14_1: ; %atomicrmw.start
; GFX90A-NEXT: ; =>This Inner Loop Header: Depth=1
; GFX90A-NEXT: s_waitcnt vmcnt(0)
-; GFX90A-NEXT: v_lshrrev_b32_sdwa v0, s6, v1 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:DWORD
-; GFX90A-NEXT: v_max_f32_e32 v0, v0, v3
-; GFX90A-NEXT: v_bfe_u32 v4, v0, 16, 1
-; GFX90A-NEXT: v_or_b32_e32 v5, 0x400000, v0
-; GFX90A-NEXT: v_add3_u32 v4, v4, v0, s8
-; GFX90A-NEXT: v_cmp_u_f32_e32 vcc, v0, v0
-; GFX90A-NEXT: v_cndmask_b32_e32 v0, v4, v5, vcc
-; GFX90A-NEXT: v_lshlrev_b32_sdwa v0, s6, v0 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:WORD_1
-; GFX90A-NEXT: v_and_or_b32 v0, v1, s7, v0
-; GFX90A-NEXT: v_pk_mov_b32 v[4:5], v[0:1], v[0:1] op_sel:[0,1]
-; GFX90A-NEXT: buffer_atomic_cmpswap v[4:5], v2, s[16:19], 0 offen glc
+; GFX90A-NEXT: v_mov_b32_e32 v5, v2
+; GFX90A-NEXT: v_lshrrev_b32_sdwa v2, s6, v5 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:DWORD
+; GFX90A-NEXT: v_max_f32_e32 v2, v2, v0
+; GFX90A-NEXT: v_bfe_u32 v3, v2, 16, 1
+; GFX90A-NEXT: v_or_b32_e32 v4, 0x400000, v2
+; GFX90A-NEXT: v_add3_u32 v3, v3, v2, s8
+; GFX90A-NEXT: v_cmp_u_f32_e32 vcc, v2, v2
+; GFX90A-NEXT: v_cndmask_b32_e32 v2, v3, v4, vcc
+; GFX90A-NEXT: v_lshlrev_b32_sdwa v2, s6, v2 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:WORD_1
+; GFX90A-NEXT: v_and_or_b32 v4, v5, s7, v2
+; GFX90A-NEXT: v_pk_mov_b32 v[2:3], v[4:5], v[4:5] op_sel:[0,1]
+; GFX90A-NEXT: buffer_atomic_cmpswap v[2:3], v1, s[16:19], 0 offen glc
; GFX90A-NEXT: s_waitcnt vmcnt(0)
; GFX90A-NEXT: buffer_wbinvl1
-; GFX90A-NEXT: v_cmp_eq_u32_e32 vcc, v4, v1
+; GFX90A-NEXT: v_cmp_eq_u32_e32 vcc, v2, v5
; GFX90A-NEXT: s_or_b64 s[4:5], vcc, s[4:5]
-; GFX90A-NEXT: v_mov_b32_e32 v1, v4
; GFX90A-NEXT: s_andn2_b64 exec, exec, s[4:5]
; GFX90A-NEXT: s_cbranch_execnz .LBB14_1
; GFX90A-NEXT: ; %bb.2: ; %atomicrmw.end
@@ -5080,35 +5097,35 @@ define void @buffer_fat_ptr_agent_atomic_fmax_noret_bf16__offset__amdgpu_no_fine
; GFX908-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX908-NEXT: s_addk_i32 s20, 0x200
; GFX908-NEXT: s_and_b32 s4, s20, -4
-; GFX908-NEXT: v_mov_b32_e32 v2, s4
-; GFX908-NEXT: buffer_load_dword v1, v2, s[16:19], 0 offen
+; GFX908-NEXT: v_mov_b32_e32 v3, s4
+; GFX908-NEXT: buffer_load_dword v1, v3, s[16:19], 0 offen
; GFX908-NEXT: s_and_b32 s4, s20, 3
; GFX908-NEXT: s_lshl_b32 s6, s4, 3
; GFX908-NEXT: s_lshl_b32 s4, 0xffff, s6
; GFX908-NEXT: s_not_b32 s7, s4
; GFX908-NEXT: s_mov_b64 s[4:5], 0
-; GFX908-NEXT: v_lshlrev_b32_e32 v3, 16, v0
+; GFX908-NEXT: v_lshlrev_b32_e32 v0, 16, v0
; GFX908-NEXT: s_movk_i32 s8, 0x7fff
; GFX908-NEXT: .LBB14_1: ; %atomicrmw.start
; GFX908-NEXT: ; =>This Inner Loop Header: Depth=1
; GFX908-NEXT: s_waitcnt vmcnt(0)
-; GFX908-NEXT: v_lshrrev_b32_sdwa v0, s6, v1 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:DWORD
-; GFX908-NEXT: v_max_f32_e32 v0, v0, v3
-; GFX908-NEXT: v_bfe_u32 v4, v0, 16, 1
-; GFX908-NEXT: v_or_b32_e32 v5, 0x400000, v0
-; GFX908-NEXT: v_add3_u32 v4, v4, v0, s8
-; GFX908-NEXT: v_cmp_u_f32_e32 vcc, v0, v0
-; GFX908-NEXT: v_cndmask_b32_e32 v0, v4, v5, vcc
-; GFX908-NEXT: v_lshlrev_b32_sdwa v0, s6, v0 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:WORD_1
-; GFX908-NEXT: v_and_or_b32 v0, v1, s7, v0
; GFX908-NEXT: v_mov_b32_e32 v5, v1
-; GFX908-NEXT: v_mov_b32_e32 v4, v0
-; GFX908-NEXT: buffer_atomic_cmpswap v[4:5], v2, s[16:19], 0 offen glc
+; GFX908-NEXT: v_lshrrev_b32_sdwa v1, s6, v5 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:DWORD
+; GFX908-NEXT: v_max_f32_e32 v1, v1, v0
+; GFX908-NEXT: v_bfe_u32 v2, v1, 16, 1
+; GFX908-NEXT: v_or_b32_e32 v4, 0x400000, v1
+; GFX908-NEXT: v_add3_u32 v2, v2, v1, s8
+; GFX908-NEXT: v_cmp_u_f32_e32 vcc, v1, v1
+; GFX908-NEXT: v_cndmask_b32_e32 v1, v2, v4, vcc
+; GFX908-NEXT: v_lshlrev_b32_sdwa v1, s6, v1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:WORD_1
+; GFX908-NEXT: v_and_or_b32 v4, v5, s7, v1
+; GFX908-NEXT: v_mov_b32_e32 v1, v4
+; GFX908-NEXT: v_mov_b32_e32 v2, v5
+; GFX908-NEXT: buffer_atomic_cmpswap v[1:2], v3, s[16:19], 0 offen glc
; GFX908-NEXT: s_waitcnt vmcnt(0)
; GFX908-NEXT: buffer_wbinvl1
-; GFX908-NEXT: v_cmp_eq_u32_e32 vcc, v4, v1
+; GFX908-NEXT: v_cmp_eq_u32_e32 vcc, v1, v5
; GFX908-NEXT: s_or_b64 s[4:5], vcc, s[4:5]
-; GFX908-NEXT: v_mov_b32_e32 v1, v4
; GFX908-NEXT: s_andn2_b64 exec, exec, s[4:5]
; GFX908-NEXT: s_cbranch_execnz .LBB14_1
; GFX908-NEXT: ; %bb.2: ; %atomicrmw.end
@@ -5120,37 +5137,37 @@ define void @buffer_fat_ptr_agent_atomic_fmax_noret_bf16__offset__amdgpu_no_fine
; GFX8-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX8-NEXT: s_addk_i32 s20, 0x200
; GFX8-NEXT: s_and_b32 s4, s20, -4
-; GFX8-NEXT: v_mov_b32_e32 v2, s4
-; GFX8-NEXT: buffer_load_dword v1, v2, s[16:19], 0 offen
+; GFX8-NEXT: v_mov_b32_e32 v3, s4
+; GFX8-NEXT: buffer_load_dword v1, v3, s[16:19], 0 offen
; GFX8-NEXT: s_and_b32 s4, s20, 3
; GFX8-NEXT: s_lshl_b32 s6, s4, 3
; GFX8-NEXT: s_lshl_b32 s4, 0xffff, s6
; GFX8-NEXT: s_not_b32 s7, s4
; GFX8-NEXT: s_mov_b64 s[4:5], 0
-; GFX8-NEXT: v_lshlrev_b32_e32 v3, 16, v0
+; GFX8-NEXT: v_lshlrev_b32_e32 v0, 16, v0
; GFX8-NEXT: .LBB14_1: ; %atomicrmw.start
; GFX8-NEXT: ; =>This Inner Loop Header: Depth=1
-; GFX8-NEXT: v_mov_b32_e32 v0, s6
+; GFX8-NEXT: v_mov_b32_e32 v2, s6
; GFX8-NEXT: s_waitcnt vmcnt(0)
-; GFX8-NEXT: v_lshrrev_b32_sdwa v5, v0, v1 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:DWORD
-; GFX8-NEXT: v_max_f32_e32 v5, v5, v3
-; GFX8-NEXT: v_bfe_u32 v6, v5, 16, 1
-; GFX8-NEXT: v_add_u32_e32 v6, vcc, v6, v5
-; GFX8-NEXT: v_add_u32_e32 v6, vcc, 0x7fff, v6
-; GFX8-NEXT: v_or_b32_e32 v7, 0x400000, v5
-; GFX8-NEXT: v_cmp_u_f32_e32 vcc, v5, v5
-; GFX8-NEXT: v_cndmask_b32_e32 v5, v6, v7, vcc
-; GFX8-NEXT: v_and_b32_e32 v4, s7, v1
-; GFX8-NEXT: v_lshlrev_b32_sdwa v0, v0, v5 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:WORD_1
-; GFX8-NEXT: v_or_b32_e32 v0, v4, v0
; GFX8-NEXT: v_mov_b32_e32 v5, v1
-; GFX8-NEXT: v_mov_b32_e32 v4, v0
-; GFX8-NEXT: buffer_atomic_cmpswap v[4:5], v2, s[16:19], 0 offen glc
+; GFX8-NEXT: v_lshrrev_b32_sdwa v1, v2, v5 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:DWORD
+; GFX8-NEXT: v_max_f32_e32 v1, v1, v0
+; GFX8-NEXT: v_bfe_u32 v6, v1, 16, 1
+; GFX8-NEXT: v_add_u32_e32 v6, vcc, v6, v1
+; GFX8-NEXT: v_add_u32_e32 v6, vcc, 0x7fff, v6
+; GFX8-NEXT: v_or_b32_e32 v7, 0x400000, v1
+; GFX8-NEXT: v_cmp_u_f32_e32 vcc, v1, v1
+; GFX8-NEXT: v_cndmask_b32_e32 v1, v6, v7, vcc
+; GFX8-NEXT: v_and_b32_e32 v4, s7, v5
+; GFX8-NEXT: v_lshlrev_b32_sdwa v1, v2, v1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:WORD_1
+; GFX8-NEXT: v_or_b32_e32 v4, v4, v1
+; GFX8-NEXT: v_mov_b32_e32 v1, v4
+; GFX8-NEXT: v_mov_b32_e32 v2, v5
+; GFX8-NEXT: buffer_atomic_cmpswap v[1:2], v3, s[16:19], 0 offen glc
; GFX8-NEXT: s_waitcnt vmcnt(0)
; GFX8-NEXT: buffer_wbinvl1
-; GFX8-NEXT: v_cmp_eq_u32_e32 vcc, v4, v1
+; GFX8-NEXT: v_cmp_eq_u32_e32 vcc, v1, v5
; GFX8-NEXT: s_or_b64 s[4:5], vcc, s[4:5]
-; GFX8-NEXT: v_mov_b32_e32 v1, v4
; GFX8-NEXT: s_andn2_b64 exec, exec, s[4:5]
; GFX8-NEXT: s_cbranch_execnz .LBB14_1
; GFX8-NEXT: ; %bb.2: ; %atomicrmw.end
@@ -5162,34 +5179,34 @@ define void @buffer_fat_ptr_agent_atomic_fmax_noret_bf16__offset__amdgpu_no_fine
; GFX7-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX7-NEXT: s_addk_i32 s20, 0x200
; GFX7-NEXT: s_and_b32 s4, s20, -4
-; GFX7-NEXT: v_mov_b32_e32 v2, s4
-; GFX7-NEXT: buffer_load_dword v1, v2, s[16:19], 0 offen
+; GFX7-NEXT: v_mov_b32_e32 v3, s4
+; GFX7-NEXT: buffer_load_dword v1, v3, s[16:19], 0 offen
; GFX7-NEXT: s_and_b32 s4, s20, 3
; GFX7-NEXT: s_lshl_b32 s6, s4, 3
; GFX7-NEXT: s_lshl_b32 s4, 0xffff, s6
; GFX7-NEXT: v_mul_f32_e32 v0, 1.0, v0
; GFX7-NEXT: s_not_b32 s7, s4
; GFX7-NEXT: s_mov_b64 s[4:5], 0
-; GFX7-NEXT: v_and_b32_e32 v3, 0xffff0000, v0
+; GFX7-NEXT: v_and_b32_e32 v0, 0xffff0000, v0
; GFX7-NEXT: .LBB14_1: ; %atomicrmw.start
; GFX7-NEXT: ; =>This Inner Loop Header: Depth=1
; GFX7-NEXT: s_waitcnt vmcnt(0)
-; GFX7-NEXT: v_lshrrev_b32_e32 v0, s6, v1
-; GFX7-NEXT: v_lshlrev_b32_e32 v0, 16, v0
-; GFX7-NEXT: v_mul_f32_e32 v0, 1.0, v0
-; GFX7-NEXT: v_max_f32_e32 v0, v0, v3
-; GFX7-NEXT: v_lshrrev_b32_e32 v0, 16, v0
-; GFX7-NEXT: v_and_b32_e32 v4, s7, v1
-; GFX7-NEXT: v_lshlrev_b32_e32 v0, s6, v0
-; GFX7-NEXT: v_or_b32_e32 v0, v4, v0
; GFX7-NEXT: v_mov_b32_e32 v5, v1
-; GFX7-NEXT: v_mov_b32_e32 v4, v0
-; GFX7-NEXT: buffer_atomic_cmpswap v[4:5], v2, s[16:19], 0 offen glc
+; GFX7-NEXT: v_lshrrev_b32_e32 v1, s6, v5
+; GFX7-NEXT: v_lshlrev_b32_e32 v1, 16, v1
+; GFX7-NEXT: v_mul_f32_e32 v1, 1.0, v1
+; GFX7-NEXT: v_max_f32_e32 v1, v1, v0
+; GFX7-NEXT: v_lshrrev_b32_e32 v1, 16, v1
+; GFX7-NEXT: v_and_b32_e32 v2, s7, v5
+; GFX7-NEXT: v_lshlrev_b32_e32 v1, s6, v1
+; GFX7-NEXT: v_or_b32_e32 v4, v2, v1
+; GFX7-NEXT: v_mov_b32_e32 v1, v4
+; GFX7-NEXT: v_mov_b32_e32 v2, v5
+; GFX7-NEXT: buffer_atomic_cmpswap v[1:2], v3, s[16:19], 0 offen glc
; GFX7-NEXT: s_waitcnt vmcnt(0)
; GFX7-NEXT: buffer_wbinvl1
-; GFX7-NEXT: v_cmp_eq_u32_e32 vcc, v4, v1
+; GFX7-NEXT: v_cmp_eq_u32_e32 vcc, v1, v5
; GFX7-NEXT: s_or_b64 s[4:5], vcc, s[4:5]
-; GFX7-NEXT: v_mov_b32_e32 v1, v4
; GFX7-NEXT: s_andn2_b64 exec, exec, s[4:5]
; GFX7-NEXT: s_cbranch_execnz .LBB14_1
; GFX7-NEXT: ; %bb.2: ; %atomicrmw.end
@@ -5201,35 +5218,35 @@ define void @buffer_fat_ptr_agent_atomic_fmax_noret_bf16__offset__amdgpu_no_fine
; GFX6-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX6-NEXT: s_addk_i32 s20, 0x200
; GFX6-NEXT: s_and_b32 s4, s20, -4
-; GFX6-NEXT: v_mov_b32_e32 v2, s4
-; GFX6-NEXT: buffer_load_dword v1, v2, s[16:19], 0 offen
+; GFX6-NEXT: v_mov_b32_e32 v3, s4
+; GFX6-NEXT: buffer_load_dword v1, v3, s[16:19], 0 offen
; GFX6-NEXT: s_and_b32 s4, s20, 3
; GFX6-NEXT: s_lshl_b32 s6, s4, 3
; GFX6-NEXT: s_lshl_b32 s4, 0xffff, s6
; GFX6-NEXT: v_mul_f32_e32 v0, 1.0, v0
; GFX6-NEXT: s_not_b32 s7, s4
; GFX6-NEXT: s_mov_b64 s[4:5], 0
-; GFX6-NEXT: v_and_b32_e32 v3, 0xffff0000, v0
+; GFX6-NEXT: v_and_b32_e32 v0, 0xffff0000, v0
; GFX6-NEXT: .LBB14_1: ; %atomicrmw.start
; GFX6-NEXT: ; =>This Inner Loop Header: Depth=1
; GFX6-NEXT: s_waitcnt vmcnt(0)
-; GFX6-NEXT: v_lshrrev_b32_e32 v0, s6, v1
-; GFX6-NEXT: v_lshlrev_b32_e32 v0, 16, v0
-; GFX6-NEXT: v_mul_f32_e32 v0, 1.0, v0
-; GFX6-NEXT: v_max_f32_e32 v0, v0, v3
-; GFX6-NEXT: v_lshrrev_b32_e32 v0, 16, v0
-; GFX6-NEXT: s_waitcnt expcnt(0)
-; GFX6-NEXT: v_and_b32_e32 v4, s7, v1
-; GFX6-NEXT: v_lshlrev_b32_e32 v0, s6, v0
-; GFX6-NEXT: v_or_b32_e32 v0, v4, v0
; GFX6-NEXT: v_mov_b32_e32 v5, v1
-; GFX6-NEXT: v_mov_b32_e32 v4, v0
-; GFX6-NEXT: buffer_atomic_cmpswap v[4:5], v2, s[16:19], 0 offen glc
+; GFX6-NEXT: s_waitcnt expcnt(0)
+; GFX6-NEXT: v_lshrrev_b32_e32 v1, s6, v5
+; GFX6-NEXT: v_lshlrev_b32_e32 v1, 16, v1
+; GFX6-NEXT: v_mul_f32_e32 v1, 1.0, v1
+; GFX6-NEXT: v_max_f32_e32 v1, v1, v0
+; GFX6-NEXT: v_lshrrev_b32_e32 v1, 16, v1
+; GFX6-NEXT: v_and_b32_e32 v2, s7, v5
+; GFX6-NEXT: v_lshlrev_b32_e32 v1, s6, v1
+; GFX6-NEXT: v_or_b32_e32 v4, v2, v1
+; GFX6-NEXT: v_mov_b32_e32 v1, v4
+; GFX6-NEXT: v_mov_b32_e32 v2, v5
+; GFX6-NEXT: buffer_atomic_cmpswap v[1:2], v3, s[16:19], 0 offen glc
; GFX6-NEXT: s_waitcnt vmcnt(0)
; GFX6-NEXT: buffer_wbinvl1
-; GFX6-NEXT: v_cmp_eq_u32_e32 vcc, v4, v1
+; GFX6-NEXT: v_cmp_eq_u32_e32 vcc, v1, v5
; GFX6-NEXT: s_or_b64 s[4:5], vcc, s[4:5]
-; GFX6-NEXT: v_mov_b32_e32 v1, v4
; GFX6-NEXT: s_andn2_b64 exec, exec, s[4:5]
; GFX6-NEXT: s_cbranch_execnz .LBB14_1
; GFX6-NEXT: ; %bb.2: ; %atomicrmw.end
@@ -5270,7 +5287,7 @@ define bfloat @buffer_fat_ptr_agent_atomic_fmax_ret_bf16__offset__waterfall__amd
; GFX12-TRUE16-NEXT: s_wait_alu 0xfffe
; GFX12-TRUE16-NEXT: s_and_saveexec_b32 s0, s0
; GFX12-TRUE16-NEXT: s_wait_loadcnt 0x0
-; GFX12-TRUE16-NEXT: buffer_load_b32 v6, v8, s[4:7], null offen
+; GFX12-TRUE16-NEXT: buffer_load_b32 v4, v8, s[4:7], null offen
; GFX12-TRUE16-NEXT: s_xor_b32 exec_lo, exec_lo, s0
; GFX12-TRUE16-NEXT: s_cbranch_execnz .LBB15_1
; GFX12-TRUE16-NEXT: ; %bb.2:
@@ -5281,26 +5298,27 @@ define bfloat @buffer_fat_ptr_agent_atomic_fmax_ret_bf16__offset__waterfall__amd
; GFX12-TRUE16-NEXT: ; =>This Loop Header: Depth=1
; GFX12-TRUE16-NEXT: ; Child Loop BB15_4 Depth 2
; GFX12-TRUE16-NEXT: s_wait_loadcnt 0x0
-; GFX12-TRUE16-NEXT: v_lshrrev_b32_e32 v4, v7, v6
+; GFX12-TRUE16-NEXT: v_mov_b32_e32 v6, v4
; GFX12-TRUE16-NEXT: s_mov_b32 s2, exec_lo
; GFX12-TRUE16-NEXT: s_wait_storecnt 0x0
; GFX12-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
+; GFX12-TRUE16-NEXT: v_lshrrev_b32_e32 v4, v7, v6
; GFX12-TRUE16-NEXT: v_lshlrev_b32_e32 v4, 16, v4
+; GFX12-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
; GFX12-TRUE16-NEXT: v_max_num_f32_e32 v4, v4, v10
-; GFX12-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(SKIP_2) | instid1(VALU_DEP_3)
; GFX12-TRUE16-NEXT: v_bfe_u32 v5, v4, 16, 1
; GFX12-TRUE16-NEXT: v_or_b32_e32 v11, 0x400000, v4
; GFX12-TRUE16-NEXT: v_cmp_u_f32_e32 vcc_lo, v4, v4
+; GFX12-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_3) | instskip(SKIP_1) | instid1(VALU_DEP_1)
; GFX12-TRUE16-NEXT: v_add3_u32 v5, v5, v4, 0x7fff
; GFX12-TRUE16-NEXT: s_wait_alu 0xfffd
-; GFX12-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(SKIP_1) | instid1(VALU_DEP_2)
; GFX12-TRUE16-NEXT: v_cndmask_b32_e32 v4, v5, v11, vcc_lo
; GFX12-TRUE16-NEXT: v_mov_b16_e32 v5.h, 0
+; GFX12-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_2) | instskip(NEXT) | instid1(VALU_DEP_1)
; GFX12-TRUE16-NEXT: v_mov_b16_e32 v5.l, v4.h
-; GFX12-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
; GFX12-TRUE16-NEXT: v_lshlrev_b32_e32 v4, v7, v5
+; GFX12-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
; GFX12-TRUE16-NEXT: v_and_or_b32 v5, v6, v9, v4
-; GFX12-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1)
; GFX12-TRUE16-NEXT: v_mov_b32_e32 v4, v5
; GFX12-TRUE16-NEXT: v_mov_b32_e32 v5, v6
; GFX12-TRUE16-NEXT: .LBB15_4: ; Parent Loop BB15_3 Depth=1
@@ -5323,7 +5341,6 @@ define bfloat @buffer_fat_ptr_agent_atomic_fmax_ret_bf16__offset__waterfall__amd
; GFX12-TRUE16-NEXT: s_mov_b32 exec_lo, s2
; GFX12-TRUE16-NEXT: s_wait_loadcnt 0x0
; GFX12-TRUE16-NEXT: v_cmp_eq_u32_e32 vcc_lo, v4, v6
-; GFX12-TRUE16-NEXT: v_mov_b32_e32 v6, v4
; GFX12-TRUE16-NEXT: global_inv scope:SCOPE_DEV
; GFX12-TRUE16-NEXT: s_or_b32 s1, vcc_lo, s1
; GFX12-TRUE16-NEXT: s_wait_alu 0xfffe
@@ -5363,7 +5380,7 @@ define bfloat @buffer_fat_ptr_agent_atomic_fmax_ret_bf16__offset__waterfall__amd
; GFX12-FAKE16-NEXT: s_wait_alu 0xfffe
; GFX12-FAKE16-NEXT: s_and_saveexec_b32 s0, s0
; GFX12-FAKE16-NEXT: s_wait_loadcnt 0x0
-; GFX12-FAKE16-NEXT: buffer_load_b32 v6, v8, s[4:7], null offen
+; GFX12-FAKE16-NEXT: buffer_load_b32 v4, v8, s[4:7], null offen
; GFX12-FAKE16-NEXT: s_xor_b32 exec_lo, exec_lo, s0
; GFX12-FAKE16-NEXT: s_cbranch_execnz .LBB15_1
; GFX12-FAKE16-NEXT: ; %bb.2:
@@ -5374,25 +5391,26 @@ define bfloat @buffer_fat_ptr_agent_atomic_fmax_ret_bf16__offset__waterfall__amd
; GFX12-FAKE16-NEXT: ; =>This Loop Header: Depth=1
; GFX12-FAKE16-NEXT: ; Child Loop BB15_4 Depth 2
; GFX12-FAKE16-NEXT: s_wait_loadcnt 0x0
-; GFX12-FAKE16-NEXT: v_lshrrev_b32_e32 v4, v7, v6
+; GFX12-FAKE16-NEXT: v_mov_b32_e32 v6, v4
; GFX12-FAKE16-NEXT: s_mov_b32 s2, exec_lo
; GFX12-FAKE16-NEXT: s_wait_storecnt 0x0
; GFX12-FAKE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
+; GFX12-FAKE16-NEXT: v_lshrrev_b32_e32 v4, v7, v6
; GFX12-FAKE16-NEXT: v_lshlrev_b32_e32 v4, 16, v4
+; GFX12-FAKE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
; GFX12-FAKE16-NEXT: v_max_num_f32_e32 v4, v4, v10
-; GFX12-FAKE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(SKIP_2) | instid1(VALU_DEP_3)
; GFX12-FAKE16-NEXT: v_bfe_u32 v5, v4, 16, 1
; GFX12-FAKE16-NEXT: v_or_b32_e32 v11, 0x400000, v4
; GFX12-FAKE16-NEXT: v_cmp_u_f32_e32 vcc_lo, v4, v4
+; GFX12-FAKE16-NEXT: s_delay_alu instid0(VALU_DEP_3) | instskip(SKIP_1) | instid1(VALU_DEP_1)
; GFX12-FAKE16-NEXT: v_add3_u32 v5, v5, v4, 0x7fff
; GFX12-FAKE16-NEXT: s_wait_alu 0xfffd
-; GFX12-FAKE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
; GFX12-FAKE16-NEXT: v_cndmask_b32_e32 v4, v5, v11, vcc_lo
-; GFX12-FAKE16-NEXT: v_lshrrev_b32_e32 v4, 16, v4
; GFX12-FAKE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
+; GFX12-FAKE16-NEXT: v_lshrrev_b32_e32 v4, 16, v4
; GFX12-FAKE16-NEXT: v_lshlrev_b32_e32 v4, v7, v4
+; GFX12-FAKE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
; GFX12-FAKE16-NEXT: v_and_or_b32 v5, v6, v9, v4
-; GFX12-FAKE16-NEXT: s_delay_alu instid0(VALU_DEP_1)
; GFX12-FAKE16-NEXT: v_mov_b32_e32 v4, v5
; GFX12-FAKE16-NEXT: v_mov_b32_e32 v5, v6
; GFX12-FAKE16-NEXT: .LBB15_4: ; Parent Loop BB15_3 Depth=1
@@ -5415,7 +5433,6 @@ define bfloat @buffer_fat_ptr_agent_atomic_fmax_ret_bf16__offset__waterfall__amd
; GFX12-FAKE16-NEXT: s_mov_b32 exec_lo, s2
; GFX12-FAKE16-NEXT: s_wait_loadcnt 0x0
; GFX12-FAKE16-NEXT: v_cmp_eq_u32_e32 vcc_lo, v4, v6
-; GFX12-FAKE16-NEXT: v_mov_b32_e32 v6, v4
; GFX12-FAKE16-NEXT: global_inv scope:SCOPE_DEV
; GFX12-FAKE16-NEXT: s_or_b32 s1, vcc_lo, s1
; GFX12-FAKE16-NEXT: s_wait_alu 0xfffe
@@ -5448,7 +5465,7 @@ define bfloat @buffer_fat_ptr_agent_atomic_fmax_ret_bf16__offset__waterfall__amd
; GFX942-NEXT: v_cmp_eq_u64_e64 s[0:1], s[6:7], v[2:3]
; GFX942-NEXT: s_and_b64 s[0:1], vcc, s[0:1]
; GFX942-NEXT: s_and_saveexec_b64 s[0:1], s[0:1]
-; GFX942-NEXT: buffer_load_dword v7, v9, s[4:7], 0 offen
+; GFX942-NEXT: buffer_load_dword v4, v9, s[4:7], 0 offen
; GFX942-NEXT: s_xor_b64 exec, exec, s[0:1]
; GFX942-NEXT: s_cbranch_execnz .LBB15_1
; GFX942-NEXT: ; %bb.2:
@@ -5460,6 +5477,7 @@ define bfloat @buffer_fat_ptr_agent_atomic_fmax_ret_bf16__offset__waterfall__amd
; GFX942-NEXT: ; =>This Loop Header: Depth=1
; GFX942-NEXT: ; Child Loop BB15_4 Depth 2
; GFX942-NEXT: s_waitcnt vmcnt(0)
+; GFX942-NEXT: v_mov_b32_e32 v7, v4
; GFX942-NEXT: v_lshrrev_b32_sdwa v4, v8, v7 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:DWORD
; GFX942-NEXT: s_mov_b64 s[8:9], exec
; GFX942-NEXT: v_max_f32_e32 v4, v4, v11
@@ -5493,7 +5511,6 @@ define bfloat @buffer_fat_ptr_agent_atomic_fmax_ret_bf16__offset__waterfall__amd
; GFX942-NEXT: s_waitcnt vmcnt(0)
; GFX942-NEXT: v_cmp_eq_u32_e32 vcc, v4, v7
; GFX942-NEXT: s_or_b64 s[2:3], vcc, s[2:3]
-; GFX942-NEXT: v_mov_b32_e32 v7, v4
; GFX942-NEXT: buffer_inv sc1
; GFX942-NEXT: s_andn2_b64 exec, exec, s[2:3]
; GFX942-NEXT: s_cbranch_execnz .LBB15_3
@@ -5526,7 +5543,7 @@ define bfloat @buffer_fat_ptr_agent_atomic_fmax_ret_bf16__offset__waterfall__amd
; GFX11-TRUE16-NEXT: s_and_b32 s0, vcc_lo, s0
; GFX11-TRUE16-NEXT: s_delay_alu instid0(SALU_CYCLE_1)
; GFX11-TRUE16-NEXT: s_and_saveexec_b32 s0, s0
-; GFX11-TRUE16-NEXT: buffer_load_b32 v6, v8, s[4:7], 0 offen
+; GFX11-TRUE16-NEXT: buffer_load_b32 v4, v8, s[4:7], 0 offen
; GFX11-TRUE16-NEXT: s_xor_b32 exec_lo, exec_lo, s0
; GFX11-TRUE16-NEXT: s_cbranch_execnz .LBB15_1
; GFX11-TRUE16-NEXT: ; %bb.2:
@@ -5538,25 +5555,26 @@ define bfloat @buffer_fat_ptr_agent_atomic_fmax_ret_bf16__offset__waterfall__amd
; GFX11-TRUE16-NEXT: ; =>This Loop Header: Depth=1
; GFX11-TRUE16-NEXT: ; Child Loop BB15_4 Depth 2
; GFX11-TRUE16-NEXT: s_waitcnt vmcnt(0)
-; GFX11-TRUE16-NEXT: v_lshrrev_b32_e32 v4, v7, v6
+; GFX11-TRUE16-NEXT: v_mov_b32_e32 v6, v4
; GFX11-TRUE16-NEXT: s_mov_b32 s2, exec_lo
; GFX11-TRUE16-NEXT: s_waitcnt_vscnt null, 0x0
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
+; GFX11-TRUE16-NEXT: v_lshrrev_b32_e32 v4, v7, v6
; GFX11-TRUE16-NEXT: v_lshlrev_b32_e32 v4, 16, v4
+; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
; GFX11-TRUE16-NEXT: v_max_f32_e32 v4, v4, v10
-; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(SKIP_2) | instid1(VALU_DEP_3)
; GFX11-TRUE16-NEXT: v_bfe_u32 v5, v4, 16, 1
; GFX11-TRUE16-NEXT: v_or_b32_e32 v11, 0x400000, v4
; GFX11-TRUE16-NEXT: v_cmp_u_f32_e32 vcc_lo, v4, v4
+; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_3) | instskip(NEXT) | instid1(VALU_DEP_1)
; GFX11-TRUE16-NEXT: v_add3_u32 v5, v5, v4, 0x7fff
-; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(SKIP_1) | instid1(VALU_DEP_2)
; GFX11-TRUE16-NEXT: v_cndmask_b32_e32 v4, v5, v11, vcc_lo
; GFX11-TRUE16-NEXT: v_mov_b16_e32 v5.h, 0
+; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_2) | instskip(NEXT) | instid1(VALU_DEP_1)
; GFX11-TRUE16-NEXT: v_mov_b16_e32 v5.l, v4.h
-; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
; GFX11-TRUE16-NEXT: v_lshlrev_b32_e32 v4, v7, v5
+; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
; GFX11-TRUE16-NEXT: v_and_or_b32 v5, v6, v9, v4
-; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1)
; GFX11-TRUE16-NEXT: v_mov_b32_e32 v4, v5
; GFX11-TRUE16-NEXT: v_mov_b32_e32 v5, v6
; GFX11-TRUE16-NEXT: .LBB15_4: ; Parent Loop BB15_3 Depth=1
@@ -5579,7 +5597,6 @@ define bfloat @buffer_fat_ptr_agent_atomic_fmax_ret_bf16__offset__waterfall__amd
; GFX11-TRUE16-NEXT: s_mov_b32 exec_lo, s2
; GFX11-TRUE16-NEXT: s_waitcnt vmcnt(0)
; GFX11-TRUE16-NEXT: v_cmp_eq_u32_e32 vcc_lo, v4, v6
-; GFX11-TRUE16-NEXT: v_mov_b32_e32 v6, v4
; GFX11-TRUE16-NEXT: buffer_gl1_inv
; GFX11-TRUE16-NEXT: buffer_gl0_inv
; GFX11-TRUE16-NEXT: s_or_b32 s1, vcc_lo, s1
@@ -5616,7 +5633,7 @@ define bfloat @buffer_fat_ptr_agent_atomic_fmax_ret_bf16__offset__waterfall__amd
; GFX11-FAKE16-NEXT: s_and_b32 s0, vcc_lo, s0
; GFX11-FAKE16-NEXT: s_delay_alu instid0(SALU_CYCLE_1)
; GFX11-FAKE16-NEXT: s_and_saveexec_b32 s0, s0
-; GFX11-FAKE16-NEXT: buffer_load_b32 v6, v8, s[4:7], 0 offen
+; GFX11-FAKE16-NEXT: buffer_load_b32 v4, v8, s[4:7], 0 offen
; GFX11-FAKE16-NEXT: s_xor_b32 exec_lo, exec_lo, s0
; GFX11-FAKE16-NEXT: s_cbranch_execnz .LBB15_1
; GFX11-FAKE16-NEXT: ; %bb.2:
@@ -5628,24 +5645,25 @@ define bfloat @buffer_fat_ptr_agent_atomic_fmax_ret_bf16__offset__waterfall__amd
; GFX11-FAKE16-NEXT: ; =>This Loop Header: Depth=1
; GFX11-FAKE16-NEXT: ; Child Loop BB15_4 Depth 2
; GFX11-FAKE16-NEXT: s_waitcnt vmcnt(0)
-; GFX11-FAKE16-NEXT: v_lshrrev_b32_e32 v4, v7, v6
+; GFX11-FAKE16-NEXT: v_mov_b32_e32 v6, v4
; GFX11-FAKE16-NEXT: s_mov_b32 s2, exec_lo
; GFX11-FAKE16-NEXT: s_waitcnt_vscnt null, 0x0
; GFX11-FAKE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
+; GFX11-FAKE16-NEXT: v_lshrrev_b32_e32 v4, v7, v6
; GFX11-FAKE16-NEXT: v_lshlrev_b32_e32 v4, 16, v4
+; GFX11-FAKE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
; GFX11-FAKE16-NEXT: v_max_f32_e32 v4, v4, v10
-; GFX11-FAKE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(SKIP_2) | instid1(VALU_DEP_3)
; GFX11-FAKE16-NEXT: v_bfe_u32 v5, v4, 16, 1
; GFX11-FAKE16-NEXT: v_or_b32_e32 v11, 0x400000, v4
; GFX11-FAKE16-NEXT: v_cmp_u_f32_e32 vcc_lo, v4, v4
+; GFX11-FAKE16-NEXT: s_delay_alu instid0(VALU_DEP_3) | instskip(NEXT) | instid1(VALU_DEP_1)
; GFX11-FAKE16-NEXT: v_add3_u32 v5, v5, v4, 0x7fff
-; GFX11-FAKE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
; GFX11-FAKE16-NEXT: v_cndmask_b32_e32 v4, v5, v11, vcc_lo
-; GFX11-FAKE16-NEXT: v_lshrrev_b32_e32 v4, 16, v4
; GFX11-FAKE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
+; GFX11-FAKE16-NEXT: v_lshrrev_b32_e32 v4, 16, v4
; GFX11-FAKE16-NEXT: v_lshlrev_b32_e32 v4, v7, v4
+; GFX11-FAKE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
; GFX11-FAKE16-NEXT: v_and_or_b32 v5, v6, v9, v4
-; GFX11-FAKE16-NEXT: s_delay_alu instid0(VALU_DEP_1)
; GFX11-FAKE16-NEXT: v_mov_b32_e32 v4, v5
; GFX11-FAKE16-NEXT: v_mov_b32_e32 v5, v6
; GFX11-FAKE16-NEXT: .LBB15_4: ; Parent Loop BB15_3 Depth=1
@@ -5668,7 +5686,6 @@ define bfloat @buffer_fat_ptr_agent_atomic_fmax_ret_bf16__offset__waterfall__amd
; GFX11-FAKE16-NEXT: s_mov_b32 exec_lo, s2
; GFX11-FAKE16-NEXT: s_waitcnt vmcnt(0)
; GFX11-FAKE16-NEXT: v_cmp_eq_u32_e32 vcc_lo, v4, v6
-; GFX11-FAKE16-NEXT: v_mov_b32_e32 v6, v4
; GFX11-FAKE16-NEXT: buffer_gl1_inv
; GFX11-FAKE16-NEXT: buffer_gl0_inv
; GFX11-FAKE16-NEXT: s_or_b32 s1, vcc_lo, s1
@@ -5701,7 +5718,7 @@ define bfloat @buffer_fat_ptr_agent_atomic_fmax_ret_bf16__offset__waterfall__amd
; GFX10-NEXT: v_cmp_eq_u64_e64 s4, s[10:11], v[2:3]
; GFX10-NEXT: s_and_b32 s4, vcc_lo, s4
; GFX10-NEXT: s_and_saveexec_b32 s4, s4
-; GFX10-NEXT: buffer_load_dword v6, v8, s[8:11], 0 offen
+; GFX10-NEXT: buffer_load_dword v4, v8, s[8:11], 0 offen
; GFX10-NEXT: s_waitcnt_depctr 0xffe3
; GFX10-NEXT: s_xor_b32 exec_lo, exec_lo, s4
; GFX10-NEXT: s_cbranch_execnz .LBB15_1
@@ -5712,9 +5729,10 @@ define bfloat @buffer_fat_ptr_agent_atomic_fmax_ret_bf16__offset__waterfall__amd
; GFX10-NEXT: ; =>This Loop Header: Depth=1
; GFX10-NEXT: ; Child Loop BB15_4 Depth 2
; GFX10-NEXT: s_waitcnt vmcnt(0)
-; GFX10-NEXT: v_lshrrev_b32_sdwa v4, v7, v6 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:DWORD
+; GFX10-NEXT: v_mov_b32_e32 v6, v4
; GFX10-NEXT: s_mov_b32 s6, exec_lo
; GFX10-NEXT: s_waitcnt_vscnt null, 0x0
+; GFX10-NEXT: v_lshrrev_b32_sdwa v4, v7, v6 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:DWORD
; GFX10-NEXT: v_max_f32_e32 v4, v4, v10
; GFX10-NEXT: v_bfe_u32 v5, v4, 16, 1
; GFX10-NEXT: v_or_b32_e32 v11, 0x400000, v4
@@ -5744,7 +5762,6 @@ define bfloat @buffer_fat_ptr_agent_atomic_fmax_ret_bf16__offset__waterfall__amd
; GFX10-NEXT: s_mov_b32 exec_lo, s6
; GFX10-NEXT: s_waitcnt vmcnt(0)
; GFX10-NEXT: v_cmp_eq_u32_e32 vcc_lo, v4, v6
-; GFX10-NEXT: v_mov_b32_e32 v6, v4
; GFX10-NEXT: buffer_gl1_inv
; GFX10-NEXT: buffer_gl0_inv
; GFX10-NEXT: s_or_b32 s5, vcc_lo, s5
@@ -5777,7 +5794,7 @@ define bfloat @buffer_fat_ptr_agent_atomic_fmax_ret_bf16__offset__waterfall__amd
; GFX90A-NEXT: s_and_b64 s[4:5], vcc, s[4:5]
; GFX90A-NEXT: s_and_saveexec_b64 s[4:5], s[4:5]
; GFX90A-NEXT: s_nop 0
-; GFX90A-NEXT: buffer_load_dword v7, v9, s[8:11], 0 offen
+; GFX90A-NEXT: buffer_load_dword v4, v9, s[8:11], 0 offen
; GFX90A-NEXT: s_xor_b64 exec, exec, s[4:5]
; GFX90A-NEXT: s_cbranch_execnz .LBB15_1
; GFX90A-NEXT: ; %bb.2:
@@ -5789,6 +5806,7 @@ define bfloat @buffer_fat_ptr_agent_atomic_fmax_ret_bf16__offset__waterfall__amd
; GFX90A-NEXT: ; =>This Loop Header: Depth=1
; GFX90A-NEXT: ; Child Loop BB15_4 Depth 2
; GFX90A-NEXT: s_waitcnt vmcnt(0)
+; GFX90A-NEXT: v_mov_b32_e32 v7, v4
; GFX90A-NEXT: v_lshrrev_b32_sdwa v4, v8, v7 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:DWORD
; GFX90A-NEXT: v_max_f32_e32 v4, v4, v11
; GFX90A-NEXT: v_bfe_u32 v5, v4, 16, 1
@@ -5819,7 +5837,6 @@ define bfloat @buffer_fat_ptr_agent_atomic_fmax_ret_bf16__offset__waterfall__amd
; GFX90A-NEXT: s_waitcnt vmcnt(0)
; GFX90A-NEXT: v_cmp_eq_u32_e32 vcc, v4, v7
; GFX90A-NEXT: s_or_b64 s[6:7], vcc, s[6:7]
-; GFX90A-NEXT: v_mov_b32_e32 v7, v4
; GFX90A-NEXT: buffer_wbinvl1
; GFX90A-NEXT: s_andn2_b64 exec, exec, s[6:7]
; GFX90A-NEXT: s_cbranch_execnz .LBB15_3
@@ -5849,7 +5866,7 @@ define bfloat @buffer_fat_ptr_agent_atomic_fmax_ret_bf16__offset__waterfall__amd
; GFX908-NEXT: s_and_b64 s[4:5], vcc, s[4:5]
; GFX908-NEXT: s_and_saveexec_b64 s[4:5], s[4:5]
; GFX908-NEXT: s_nop 0
-; GFX908-NEXT: buffer_load_dword v6, v8, s[8:11], 0 offen
+; GFX908-NEXT: buffer_load_dword v4, v8, s[8:11], 0 offen
; GFX908-NEXT: s_xor_b64 exec, exec, s[4:5]
; GFX908-NEXT: s_cbranch_execnz .LBB15_1
; GFX908-NEXT: ; %bb.2:
@@ -5861,6 +5878,7 @@ define bfloat @buffer_fat_ptr_agent_atomic_fmax_ret_bf16__offset__waterfall__amd
; GFX908-NEXT: ; =>This Loop Header: Depth=1
; GFX908-NEXT: ; Child Loop BB15_4 Depth 2
; GFX908-NEXT: s_waitcnt vmcnt(0)
+; GFX908-NEXT: v_mov_b32_e32 v6, v4
; GFX908-NEXT: v_lshrrev_b32_sdwa v4, v7, v6 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:DWORD
; GFX908-NEXT: v_max_f32_e32 v4, v4, v10
; GFX908-NEXT: v_bfe_u32 v5, v4, 16, 1
@@ -5892,7 +5910,6 @@ define bfloat @buffer_fat_ptr_agent_atomic_fmax_ret_bf16__offset__waterfall__amd
; GFX908-NEXT: s_waitcnt vmcnt(0)
; GFX908-NEXT: v_cmp_eq_u32_e32 vcc, v4, v6
; GFX908-NEXT: s_or_b64 s[6:7], vcc, s[6:7]
-; GFX908-NEXT: v_mov_b32_e32 v6, v4
; GFX908-NEXT: buffer_wbinvl1
; GFX908-NEXT: s_andn2_b64 exec, exec, s[6:7]
; GFX908-NEXT: s_cbranch_execnz .LBB15_3
@@ -5922,7 +5939,7 @@ define bfloat @buffer_fat_ptr_agent_atomic_fmax_ret_bf16__offset__waterfall__amd
; GFX8-NEXT: s_and_b64 s[4:5], vcc, s[4:5]
; GFX8-NEXT: s_and_saveexec_b64 s[4:5], s[4:5]
; GFX8-NEXT: s_nop 0
-; GFX8-NEXT: buffer_load_dword v6, v8, s[8:11], 0 offen
+; GFX8-NEXT: buffer_load_dword v4, v8, s[8:11], 0 offen
; GFX8-NEXT: s_xor_b64 exec, exec, s[4:5]
; GFX8-NEXT: s_cbranch_execnz .LBB15_1
; GFX8-NEXT: ; %bb.2:
@@ -5933,6 +5950,7 @@ define bfloat @buffer_fat_ptr_agent_atomic_fmax_ret_bf16__offset__waterfall__amd
; GFX8-NEXT: ; =>This Loop Header: Depth=1
; GFX8-NEXT: ; Child Loop BB15_4 Depth 2
; GFX8-NEXT: s_waitcnt vmcnt(0)
+; GFX8-NEXT: v_mov_b32_e32 v6, v4
; GFX8-NEXT: v_lshrrev_b32_sdwa v4, v7, v6 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:DWORD
; GFX8-NEXT: v_max_f32_e32 v4, v4, v10
; GFX8-NEXT: v_bfe_u32 v5, v4, 16, 1
@@ -5966,7 +5984,6 @@ define bfloat @buffer_fat_ptr_agent_atomic_fmax_ret_bf16__offset__waterfall__amd
; GFX8-NEXT: s_waitcnt vmcnt(0)
; GFX8-NEXT: v_cmp_eq_u32_e32 vcc, v4, v6
; GFX8-NEXT: s_or_b64 s[6:7], vcc, s[6:7]
-; GFX8-NEXT: v_mov_b32_e32 v6, v4
; GFX8-NEXT: buffer_wbinvl1
; GFX8-NEXT: s_andn2_b64 exec, exec, s[6:7]
; GFX8-NEXT: s_cbranch_execnz .LBB15_3
@@ -5994,18 +6011,19 @@ define bfloat @buffer_fat_ptr_agent_atomic_fmax_ret_bf16__offset__waterfall__amd
; GFX7-NEXT: v_cmp_eq_u64_e64 s[4:5], s[10:11], v[2:3]
; GFX7-NEXT: s_and_b64 s[4:5], vcc, s[4:5]
; GFX7-NEXT: s_and_saveexec_b64 s[4:5], s[4:5]
-; GFX7-NEXT: buffer_load_dword v6, v8, s[8:11], 0 offen
+; GFX7-NEXT: buffer_load_dword v4, v8, s[8:11], 0 offen
; GFX7-NEXT: s_xor_b64 exec, exec, s[4:5]
; GFX7-NEXT: s_cbranch_execnz .LBB15_1
; GFX7-NEXT: ; %bb.2:
; GFX7-NEXT: s_mov_b64 exec, s[6:7]
-; GFX7-NEXT: v_mul_f32_e32 v4, 1.0, v5
+; GFX7-NEXT: v_mul_f32_e32 v5, 1.0, v5
; GFX7-NEXT: s_mov_b64 s[6:7], 0
-; GFX7-NEXT: v_and_b32_e32 v10, 0xffff0000, v4
+; GFX7-NEXT: v_and_b32_e32 v10, 0xffff0000, v5
; GFX7-NEXT: .LBB15_3: ; %atomicrmw.start
; GFX7-NEXT: ; =>This Loop Header: Depth=1
; GFX7-NEXT: ; Child Loop BB15_4 Depth 2
; GFX7-NEXT: s_waitcnt vmcnt(0)
+; GFX7-NEXT: v_mov_b32_e32 v6, v4
; GFX7-NEXT: v_lshrrev_b32_e32 v4, v7, v6
; GFX7-NEXT: v_lshlrev_b32_e32 v4, 16, v4
; GFX7-NEXT: v_mul_f32_e32 v4, 1.0, v4
@@ -6036,7 +6054,6 @@ define bfloat @buffer_fat_ptr_agent_atomic_fmax_ret_bf16__offset__waterfall__amd
; GFX7-NEXT: s_waitcnt vmcnt(0)
; GFX7-NEXT: v_cmp_eq_u32_e32 vcc, v4, v6
; GFX7-NEXT: s_or_b64 s[6:7], vcc, s[6:7]
-; GFX7-NEXT: v_mov_b32_e32 v6, v4
; GFX7-NEXT: buffer_wbinvl1
; GFX7-NEXT: s_andn2_b64 exec, exec, s[6:7]
; GFX7-NEXT: s_cbranch_execnz .LBB15_3
@@ -6065,18 +6082,20 @@ define bfloat @buffer_fat_ptr_agent_atomic_fmax_ret_bf16__offset__waterfall__amd
; GFX6-NEXT: v_cmp_eq_u64_e64 s[4:5], s[10:11], v[2:3]
; GFX6-NEXT: s_and_b64 s[4:5], vcc, s[4:5]
; GFX6-NEXT: s_and_saveexec_b64 s[4:5], s[4:5]
-; GFX6-NEXT: buffer_load_dword v6, v8, s[8:11], 0 offen
+; GFX6-NEXT: buffer_load_dword v4, v8, s[8:11], 0 offen
; GFX6-NEXT: s_xor_b64 exec, exec, s[4:5]
; GFX6-NEXT: s_cbranch_execnz .LBB15_1
; GFX6-NEXT: ; %bb.2:
; GFX6-NEXT: s_mov_b64 exec, s[6:7]
-; GFX6-NEXT: v_mul_f32_e32 v4, 1.0, v5
+; GFX6-NEXT: v_mul_f32_e32 v5, 1.0, v5
; GFX6-NEXT: s_mov_b64 s[6:7], 0
-; GFX6-NEXT: v_and_b32_e32 v10, 0xffff0000, v4
+; GFX6-NEXT: v_and_b32_e32 v10, 0xffff0000, v5
; GFX6-NEXT: .LBB15_3: ; %atomicrmw.start
; GFX6-NEXT: ; =>This Loop Header: Depth=1
; GFX6-NEXT: ; Child Loop BB15_4 Depth 2
-; GFX6-NEXT: s_waitcnt vmcnt(0) expcnt(0)
+; GFX6-NEXT: s_waitcnt vmcnt(0)
+; GFX6-NEXT: v_mov_b32_e32 v6, v4
+; GFX6-NEXT: s_waitcnt expcnt(0)
; GFX6-NEXT: v_lshrrev_b32_e32 v4, v7, v6
; GFX6-NEXT: v_lshlrev_b32_e32 v4, 16, v4
; GFX6-NEXT: v_mul_f32_e32 v4, 1.0, v4
@@ -6107,7 +6126,6 @@ define bfloat @buffer_fat_ptr_agent_atomic_fmax_ret_bf16__offset__waterfall__amd
; GFX6-NEXT: s_waitcnt vmcnt(0)
; GFX6-NEXT: v_cmp_eq_u32_e32 vcc, v4, v6
; GFX6-NEXT: s_or_b64 s[6:7], vcc, s[6:7]
-; GFX6-NEXT: v_mov_b32_e32 v6, v4
; GFX6-NEXT: buffer_wbinvl1
; GFX6-NEXT: s_andn2_b64 exec, exec, s[6:7]
; GFX6-NEXT: s_cbranch_execnz .LBB15_3
@@ -6457,7 +6475,7 @@ define void @buffer_fat_ptr_agent_atomic_fmax_noret_v2f16__offset__amdgpu_no_fin
; GFX12-NEXT: s_wait_kmcnt 0x0
; GFX12-NEXT: v_mov_b32_e32 v1, s16
; GFX12-NEXT: s_add_co_i32 s4, s16, 0x400
-; GFX12-NEXT: v_pk_max_num_f16 v2, v0, v0
+; GFX12-NEXT: v_pk_max_num_f16 v0, v0, v0
; GFX12-NEXT: s_wait_alu 0xfffe
; GFX12-NEXT: v_mov_b32_e32 v3, s4
; GFX12-NEXT: s_mov_b32 s4, 0
@@ -6465,16 +6483,17 @@ define void @buffer_fat_ptr_agent_atomic_fmax_noret_v2f16__offset__amdgpu_no_fin
; GFX12-NEXT: .LBB17_1: ; %atomicrmw.start
; GFX12-NEXT: ; =>This Inner Loop Header: Depth=1
; GFX12-NEXT: s_wait_loadcnt 0x0
-; GFX12-NEXT: v_pk_max_num_f16 v0, v1, v1
+; GFX12-NEXT: v_mov_b32_e32 v5, v1
; GFX12-NEXT: s_wait_storecnt 0x0
; GFX12-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
-; GFX12-NEXT: v_pk_max_num_f16 v0, v0, v2
-; GFX12-NEXT: v_dual_mov_b32 v5, v1 :: v_dual_mov_b32 v4, v0
-; GFX12-NEXT: buffer_atomic_cmpswap_b32 v[4:5], v3, s[0:3], null offen th:TH_ATOMIC_RETURN
+; GFX12-NEXT: v_pk_max_num_f16 v1, v5, v5
+; GFX12-NEXT: v_pk_max_num_f16 v4, v1, v0
+; GFX12-NEXT: s_delay_alu instid0(VALU_DEP_1)
+; GFX12-NEXT: v_dual_mov_b32 v1, v4 :: v_dual_mov_b32 v2, v5
+; GFX12-NEXT: buffer_atomic_cmpswap_b32 v[1:2], v3, s[0:3], null offen th:TH_ATOMIC_RETURN
; GFX12-NEXT: s_wait_loadcnt 0x0
; GFX12-NEXT: global_inv scope:SCOPE_DEV
-; GFX12-NEXT: v_cmp_eq_u32_e32 vcc_lo, v4, v1
-; GFX12-NEXT: v_mov_b32_e32 v1, v4
+; GFX12-NEXT: v_cmp_eq_u32_e32 vcc_lo, v1, v5
; GFX12-NEXT: s_or_b32 s4, vcc_lo, s4
; GFX12-NEXT: s_wait_alu 0xfffe
; GFX12-NEXT: s_and_not1_b32 exec_lo, exec_lo, s4
@@ -6488,25 +6507,25 @@ define void @buffer_fat_ptr_agent_atomic_fmax_noret_v2f16__offset__amdgpu_no_fin
; GFX942: ; %bb.0:
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX942-NEXT: v_mov_b32_e32 v1, s16
-; GFX942-NEXT: buffer_load_dword v1, v1, s[0:3], 0 offen offset:1024
+; GFX942-NEXT: buffer_load_dword v2, v1, s[0:3], 0 offen offset:1024
; GFX942-NEXT: s_add_i32 s6, s16, 0x400
; GFX942-NEXT: s_mov_b64 s[4:5], 0
-; GFX942-NEXT: v_pk_max_f16 v2, v0, v0
-; GFX942-NEXT: v_mov_b32_e32 v3, s6
+; GFX942-NEXT: v_pk_max_f16 v0, v0, v0
+; GFX942-NEXT: v_mov_b32_e32 v1, s6
; GFX942-NEXT: .LBB17_1: ; %atomicrmw.start
; GFX942-NEXT: ; =>This Inner Loop Header: Depth=1
; GFX942-NEXT: s_waitcnt vmcnt(0)
-; GFX942-NEXT: v_pk_max_f16 v0, v1, v1
+; GFX942-NEXT: v_mov_b32_e32 v5, v2
+; GFX942-NEXT: v_pk_max_f16 v2, v5, v5
; GFX942-NEXT: buffer_wbl2 sc1
-; GFX942-NEXT: v_pk_max_f16 v0, v0, v2
+; GFX942-NEXT: v_pk_max_f16 v4, v2, v0
; GFX942-NEXT: s_nop 0
-; GFX942-NEXT: v_mov_b64_e32 v[4:5], v[0:1]
-; GFX942-NEXT: buffer_atomic_cmpswap v[4:5], v3, s[0:3], 0 offen sc0
+; GFX942-NEXT: v_mov_b64_e32 v[2:3], v[4:5]
+; GFX942-NEXT: buffer_atomic_cmpswap v[2:3], v1, s[0:3], 0 offen sc0
; GFX942-NEXT: s_waitcnt vmcnt(0)
; GFX942-NEXT: buffer_inv sc1
-; GFX942-NEXT: v_cmp_eq_u32_e32 vcc, v4, v1
+; GFX942-NEXT: v_cmp_eq_u32_e32 vcc, v2, v5
; GFX942-NEXT: s_or_b64 s[4:5], vcc, s[4:5]
-; GFX942-NEXT: v_mov_b32_e32 v1, v4
; GFX942-NEXT: s_andn2_b64 exec, exec, s[4:5]
; GFX942-NEXT: s_cbranch_execnz .LBB17_1
; GFX942-NEXT: ; %bb.2: ; %atomicrmw.end
@@ -6518,24 +6537,25 @@ define void @buffer_fat_ptr_agent_atomic_fmax_noret_v2f16__offset__amdgpu_no_fin
; GFX11-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX11-NEXT: v_mov_b32_e32 v1, s16
; GFX11-NEXT: s_add_i32 s4, s16, 0x400
-; GFX11-NEXT: v_pk_max_f16 v2, v0, v0
+; GFX11-NEXT: v_pk_max_f16 v0, v0, v0
; GFX11-NEXT: v_mov_b32_e32 v3, s4
; GFX11-NEXT: s_mov_b32 s4, 0
; GFX11-NEXT: buffer_load_b32 v1, v1, s[0:3], 0 offen offset:1024
; GFX11-NEXT: .LBB17_1: ; %atomicrmw.start
; GFX11-NEXT: ; =>This Inner Loop Header: Depth=1
; GFX11-NEXT: s_waitcnt vmcnt(0)
-; GFX11-NEXT: v_pk_max_f16 v0, v1, v1
+; GFX11-NEXT: v_mov_b32_e32 v5, v1
; GFX11-NEXT: s_waitcnt_vscnt null, 0x0
; GFX11-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
-; GFX11-NEXT: v_pk_max_f16 v0, v0, v2
-; GFX11-NEXT: v_dual_mov_b32 v5, v1 :: v_dual_mov_b32 v4, v0
-; GFX11-NEXT: buffer_atomic_cmpswap_b32 v[4:5], v3, s[0:3], 0 offen glc
+; GFX11-NEXT: v_pk_max_f16 v1, v5, v5
+; GFX11-NEXT: v_pk_max_f16 v4, v1, v0
+; GFX11-NEXT: s_delay_alu instid0(VALU_DEP_1)
+; GFX11-NEXT: v_dual_mov_b32 v1, v4 :: v_dual_mov_b32 v2, v5
+; GFX11-NEXT: buffer_atomic_cmpswap_b32 v[1:2], v3, s[0:3], 0 offen glc
; GFX11-NEXT: s_waitcnt vmcnt(0)
; GFX11-NEXT: buffer_gl1_inv
; GFX11-NEXT: buffer_gl0_inv
-; GFX11-NEXT: v_cmp_eq_u32_e32 vcc_lo, v4, v1
-; GFX11-NEXT: v_mov_b32_e32 v1, v4
+; GFX11-NEXT: v_cmp_eq_u32_e32 vcc_lo, v1, v5
; GFX11-NEXT: s_or_b32 s4, vcc_lo, s4
; GFX11-NEXT: s_delay_alu instid0(SALU_CYCLE_1)
; GFX11-NEXT: s_and_not1_b32 exec_lo, exec_lo, s4
@@ -6549,24 +6569,24 @@ define void @buffer_fat_ptr_agent_atomic_fmax_noret_v2f16__offset__amdgpu_no_fin
; GFX10-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX10-NEXT: v_mov_b32_e32 v1, s20
; GFX10-NEXT: s_add_i32 s4, s20, 0x400
-; GFX10-NEXT: v_pk_max_f16 v2, v0, v0
+; GFX10-NEXT: v_pk_max_f16 v0, v0, v0
; GFX10-NEXT: v_mov_b32_e32 v3, s4
; GFX10-NEXT: s_mov_b32 s4, 0
; GFX10-NEXT: buffer_load_dword v1, v1, s[16:19], 0 offen offset:1024
; GFX10-NEXT: .LBB17_1: ; %atomicrmw.start
; GFX10-NEXT: ; =>This Inner Loop Header: Depth=1
; GFX10-NEXT: s_waitcnt vmcnt(0)
-; GFX10-NEXT: v_pk_max_f16 v0, v1, v1
-; GFX10-NEXT: s_waitcnt_vscnt null, 0x0
-; GFX10-NEXT: v_pk_max_f16 v0, v0, v2
; GFX10-NEXT: v_mov_b32_e32 v5, v1
-; GFX10-NEXT: v_mov_b32_e32 v4, v0
-; GFX10-NEXT: buffer_atomic_cmpswap v[4:5], v3, s[16:19], 0 offen glc
+; GFX10-NEXT: s_waitcnt_vscnt null, 0x0
+; GFX10-NEXT: v_pk_max_f16 v1, v5, v5
+; GFX10-NEXT: v_pk_max_f16 v4, v1, v0
+; GFX10-NEXT: v_mov_b32_e32 v1, v4
+; GFX10-NEXT: v_mov_b32_e32 v2, v5
+; GFX10-NEXT: buffer_atomic_cmpswap v[1:2], v3, s[16:19], 0 offen glc
; GFX10-NEXT: s_waitcnt vmcnt(0)
; GFX10-NEXT: buffer_gl1_inv
; GFX10-NEXT: buffer_gl0_inv
-; GFX10-NEXT: v_cmp_eq_u32_e32 vcc_lo, v4, v1
-; GFX10-NEXT: v_mov_b32_e32 v1, v4
+; GFX10-NEXT: v_cmp_eq_u32_e32 vcc_lo, v1, v5
; GFX10-NEXT: s_or_b32 s4, vcc_lo, s4
; GFX10-NEXT: s_andn2_b32 exec_lo, exec_lo, s4
; GFX10-NEXT: s_cbranch_execnz .LBB17_1
@@ -6578,23 +6598,23 @@ define void @buffer_fat_ptr_agent_atomic_fmax_noret_v2f16__offset__amdgpu_no_fin
; GFX90A: ; %bb.0:
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX90A-NEXT: v_mov_b32_e32 v1, s20
-; GFX90A-NEXT: buffer_load_dword v1, v1, s[16:19], 0 offen offset:1024
+; GFX90A-NEXT: buffer_load_dword v2, v1, s[16:19], 0 offen offset:1024
; GFX90A-NEXT: s_add_i32 s6, s20, 0x400
; GFX90A-NEXT: s_mov_b64 s[4:5], 0
-; GFX90A-NEXT: v_pk_max_f16 v2, v0, v0
-; GFX90A-NEXT: v_mov_b32_e32 v3, s6
+; GFX90A-NEXT: v_pk_max_f16 v0, v0, v0
+; GFX90A-NEXT: v_mov_b32_e32 v1, s6
; GFX90A-NEXT: .LBB17_1: ; %atomicrmw.start
; GFX90A-NEXT: ; =>This Inner Loop Header: Depth=1
; GFX90A-NEXT: s_waitcnt vmcnt(0)
-; GFX90A-NEXT: v_pk_max_f16 v0, v1, v1
-; GFX90A-NEXT: v_pk_max_f16 v0, v0, v2
-; GFX90A-NEXT: v_pk_mov_b32 v[4:5], v[0:1], v[0:1] op_sel:[0,1]
-; GFX90A-NEXT: buffer_atomic_cmpswap v[4:5], v3, s[16:19], 0 offen glc
+; GFX90A-NEXT: v_mov_b32_e32 v5, v2
+; GFX90A-NEXT: v_pk_max_f16 v2, v5, v5
+; GFX90A-NEXT: v_pk_max_f16 v4, v2, v0
+; GFX90A-NEXT: v_pk_mov_b32 v[2:3], v[4:5], v[4:5] op_sel:[0,1]
+; GFX90A-NEXT: buffer_atomic_cmpswap v[2:3], v1, s[16:19], 0 offen glc
; GFX90A-NEXT: s_waitcnt vmcnt(0)
; GFX90A-NEXT: buffer_wbinvl1
-; GFX90A-NEXT: v_cmp_eq_u32_e32 vcc, v4, v1
+; GFX90A-NEXT: v_cmp_eq_u32_e32 vcc, v2, v5
; GFX90A-NEXT: s_or_b64 s[4:5], vcc, s[4:5]
-; GFX90A-NEXT: v_mov_b32_e32 v1, v4
; GFX90A-NEXT: s_andn2_b64 exec, exec, s[4:5]
; GFX90A-NEXT: s_cbranch_execnz .LBB17_1
; GFX90A-NEXT: ; %bb.2: ; %atomicrmw.end
@@ -6608,21 +6628,21 @@ define void @buffer_fat_ptr_agent_atomic_fmax_noret_v2f16__offset__amdgpu_no_fin
; GFX908-NEXT: buffer_load_dword v1, v1, s[16:19], 0 offen offset:1024
; GFX908-NEXT: s_add_i32 s6, s20, 0x400
; GFX908-NEXT: s_mov_b64 s[4:5], 0
-; GFX908-NEXT: v_pk_max_f16 v2, v0, v0
+; GFX908-NEXT: v_pk_max_f16 v0, v0, v0
; GFX908-NEXT: v_mov_b32_e32 v3, s6
; GFX908-NEXT: .LBB17_1: ; %atomicrmw.start
; GFX908-NEXT: ; =>This Inner Loop Header: Depth=1
; GFX908-NEXT: s_waitcnt vmcnt(0)
-; GFX908-NEXT: v_pk_max_f16 v0, v1, v1
-; GFX908-NEXT: v_pk_max_f16 v0, v0, v2
; GFX908-NEXT: v_mov_b32_e32 v5, v1
-; GFX908-NEXT: v_mov_b32_e32 v4, v0
-; GFX908-NEXT: buffer_atomic_cmpswap v[4:5], v3, s[16:19], 0 offen glc
+; GFX908-NEXT: v_pk_max_f16 v1, v5, v5
+; GFX908-NEXT: v_pk_max_f16 v4, v1, v0
+; GFX908-NEXT: v_mov_b32_e32 v1, v4
+; GFX908-NEXT: v_mov_b32_e32 v2, v5
+; GFX908-NEXT: buffer_atomic_cmpswap v[1:2], v3, s[16:19], 0 offen glc
; GFX908-NEXT: s_waitcnt vmcnt(0)
; GFX908-NEXT: buffer_wbinvl1
-; GFX908-NEXT: v_cmp_eq_u32_e32 vcc, v4, v1
+; GFX908-NEXT: v_cmp_eq_u32_e32 vcc, v1, v5
; GFX908-NEXT: s_or_b64 s[4:5], vcc, s[4:5]
-; GFX908-NEXT: v_mov_b32_e32 v1, v4
; GFX908-NEXT: s_andn2_b64 exec, exec, s[4:5]
; GFX908-NEXT: s_cbranch_execnz .LBB17_1
; GFX908-NEXT: ; %bb.2: ; %atomicrmw.end
@@ -6636,25 +6656,25 @@ define void @buffer_fat_ptr_agent_atomic_fmax_noret_v2f16__offset__amdgpu_no_fin
; GFX8-NEXT: buffer_load_dword v1, v1, s[16:19], 0 offen offset:1024
; GFX8-NEXT: s_add_i32 s6, s20, 0x400
; GFX8-NEXT: s_mov_b64 s[4:5], 0
-; GFX8-NEXT: v_max_f16_sdwa v2, v0, v0 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_1 src1_sel:WORD_1
-; GFX8-NEXT: v_max_f16_e32 v3, v0, v0
+; GFX8-NEXT: v_max_f16_sdwa v3, v0, v0 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_1 src1_sel:WORD_1
+; GFX8-NEXT: v_max_f16_e32 v0, v0, v0
; GFX8-NEXT: v_mov_b32_e32 v4, s6
; GFX8-NEXT: .LBB17_1: ; %atomicrmw.start
; GFX8-NEXT: ; =>This Inner Loop Header: Depth=1
; GFX8-NEXT: s_waitcnt vmcnt(0)
-; GFX8-NEXT: v_max_f16_sdwa v0, v1, v1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_1 src1_sel:WORD_1
-; GFX8-NEXT: v_max_f16_e32 v5, v1, v1
-; GFX8-NEXT: v_max_f16_sdwa v0, v0, v2 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:DWORD
-; GFX8-NEXT: v_max_f16_e32 v5, v5, v3
-; GFX8-NEXT: v_or_b32_e32 v0, v5, v0
; GFX8-NEXT: v_mov_b32_e32 v6, v1
-; GFX8-NEXT: v_mov_b32_e32 v5, v0
-; GFX8-NEXT: buffer_atomic_cmpswap v[5:6], v4, s[16:19], 0 offen glc
+; GFX8-NEXT: v_max_f16_sdwa v1, v6, v6 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_1 src1_sel:WORD_1
+; GFX8-NEXT: v_max_f16_e32 v2, v6, v6
+; GFX8-NEXT: v_max_f16_sdwa v1, v1, v3 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:DWORD
+; GFX8-NEXT: v_max_f16_e32 v2, v2, v0
+; GFX8-NEXT: v_or_b32_e32 v5, v2, v1
+; GFX8-NEXT: v_mov_b32_e32 v1, v5
+; GFX8-NEXT: v_mov_b32_e32 v2, v6
+; GFX8-NEXT: buffer_atomic_cmpswap v[1:2], v4, s[16:19], 0 offen glc
; GFX8-NEXT: s_waitcnt vmcnt(0)
; GFX8-NEXT: buffer_wbinvl1
-; GFX8-NEXT: v_cmp_eq_u32_e32 vcc, v5, v1
+; GFX8-NEXT: v_cmp_eq_u32_e32 vcc, v1, v6
; GFX8-NEXT: s_or_b64 s[4:5], vcc, s[4:5]
-; GFX8-NEXT: v_mov_b32_e32 v1, v5
; GFX8-NEXT: s_andn2_b64 exec, exec, s[4:5]
; GFX8-NEXT: s_cbranch_execnz .LBB17_1
; GFX8-NEXT: ; %bb.2: ; %atomicrmw.end
@@ -7918,43 +7938,45 @@ define void @buffer_fat_ptr_agent_atomic_fmax_noret_v2bf16__offset__amdgpu_no_fi
; GFX12-TRUE16-NEXT: s_wait_samplecnt 0x0
; GFX12-TRUE16-NEXT: s_wait_bvhcnt 0x0
; GFX12-TRUE16-NEXT: s_wait_kmcnt 0x0
-; GFX12-TRUE16-NEXT: v_dual_mov_b32 v1, s16 :: v_dual_and_b32 v2, 0xffff0000, v0
+; GFX12-TRUE16-NEXT: v_mov_b32_e32 v1, s16
; GFX12-TRUE16-NEXT: s_add_co_i32 s4, s16, 0x400
; GFX12-TRUE16-NEXT: s_wait_alu 0xfffe
-; GFX12-TRUE16-NEXT: v_dual_mov_b32 v4, s4 :: v_dual_lshlrev_b32 v3, 16, v0
+; GFX12-TRUE16-NEXT: v_dual_mov_b32 v4, s4 :: v_dual_and_b32 v3, 0xffff0000, v0
+; GFX12-TRUE16-NEXT: v_lshlrev_b32_e32 v0, 16, v0
; GFX12-TRUE16-NEXT: buffer_load_b32 v1, v1, s[0:3], null offen offset:1024
; GFX12-TRUE16-NEXT: s_mov_b32 s4, 0
; GFX12-TRUE16-NEXT: .LBB20_1: ; %atomicrmw.start
; GFX12-TRUE16-NEXT: ; =>This Inner Loop Header: Depth=1
; GFX12-TRUE16-NEXT: s_wait_loadcnt 0x0
-; GFX12-TRUE16-NEXT: v_and_b32_e32 v5, 0xffff0000, v1
-; GFX12-TRUE16-NEXT: v_lshlrev_b32_e32 v0, 16, v1
+; GFX12-TRUE16-NEXT: v_mov_b32_e32 v6, v1
; GFX12-TRUE16-NEXT: s_wait_storecnt 0x0
; GFX12-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
-; GFX12-TRUE16-NEXT: v_dual_max_num_f32 v5, v5, v2 :: v_dual_max_num_f32 v0, v0, v3
-; GFX12-TRUE16-NEXT: v_bfe_u32 v7, v5, 16, 1
+; GFX12-TRUE16-NEXT: v_and_b32_e32 v2, 0xffff0000, v6
+; GFX12-TRUE16-NEXT: v_dual_max_num_f32 v2, v2, v3 :: v_dual_lshlrev_b32 v1, 16, v6
+; GFX12-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_2)
+; GFX12-TRUE16-NEXT: v_max_num_f32_e32 v1, v1, v0
+; GFX12-TRUE16-NEXT: v_bfe_u32 v7, v2, 16, 1
; GFX12-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_2)
-; GFX12-TRUE16-NEXT: v_bfe_u32 v6, v0, 16, 1
-; GFX12-TRUE16-NEXT: v_or_b32_e32 v8, 0x400000, v0
-; GFX12-TRUE16-NEXT: v_cmp_u_f32_e32 vcc_lo, v0, v0
-; GFX12-TRUE16-NEXT: v_or_b32_e32 v9, 0x400000, v5
-; GFX12-TRUE16-NEXT: v_add3_u32 v7, v7, v5, 0x7fff
-; GFX12-TRUE16-NEXT: v_add3_u32 v6, v6, v0, 0x7fff
+; GFX12-TRUE16-NEXT: v_bfe_u32 v5, v1, 16, 1
+; GFX12-TRUE16-NEXT: v_or_b32_e32 v8, 0x400000, v1
+; GFX12-TRUE16-NEXT: v_cmp_u_f32_e32 vcc_lo, v1, v1
+; GFX12-TRUE16-NEXT: v_or_b32_e32 v9, 0x400000, v2
+; GFX12-TRUE16-NEXT: v_add3_u32 v7, v7, v2, 0x7fff
+; GFX12-TRUE16-NEXT: v_add3_u32 v5, v5, v1, 0x7fff
; GFX12-TRUE16-NEXT: s_wait_alu 0xfffd
; GFX12-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(SKIP_3) | instid1(VALU_DEP_3)
-; GFX12-TRUE16-NEXT: v_cndmask_b32_e32 v0, v6, v8, vcc_lo
-; GFX12-TRUE16-NEXT: v_cmp_u_f32_e32 vcc_lo, v5, v5
+; GFX12-TRUE16-NEXT: v_cndmask_b32_e32 v1, v5, v8, vcc_lo
+; GFX12-TRUE16-NEXT: v_cmp_u_f32_e32 vcc_lo, v2, v2
; GFX12-TRUE16-NEXT: s_wait_alu 0xfffd
-; GFX12-TRUE16-NEXT: v_cndmask_b32_e32 v5, v7, v9, vcc_lo
-; GFX12-TRUE16-NEXT: v_mov_b16_e32 v0.l, v0.h
+; GFX12-TRUE16-NEXT: v_cndmask_b32_e32 v2, v7, v9, vcc_lo
+; GFX12-TRUE16-NEXT: v_mov_b16_e32 v1.l, v1.h
; GFX12-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
-; GFX12-TRUE16-NEXT: v_bfi_b32 v0, 0xffff, v0, v5
-; GFX12-TRUE16-NEXT: v_dual_mov_b32 v6, v1 :: v_dual_mov_b32 v5, v0
-; GFX12-TRUE16-NEXT: buffer_atomic_cmpswap_b32 v[5:6], v4, s[0:3], null offen th:TH_ATOMIC_RETURN
+; GFX12-TRUE16-NEXT: v_bfi_b32 v5, 0xffff, v1, v2
+; GFX12-TRUE16-NEXT: v_dual_mov_b32 v1, v5 :: v_dual_mov_b32 v2, v6
+; GFX12-TRUE16-NEXT: buffer_atomic_cmpswap_b32 v[1:2], v4, s[0:3], null offen th:TH_ATOMIC_RETURN
; GFX12-TRUE16-NEXT: s_wait_loadcnt 0x0
; GFX12-TRUE16-NEXT: global_inv scope:SCOPE_DEV
-; GFX12-TRUE16-NEXT: v_cmp_eq_u32_e32 vcc_lo, v5, v1
-; GFX12-TRUE16-NEXT: v_mov_b32_e32 v1, v5
+; GFX12-TRUE16-NEXT: v_cmp_eq_u32_e32 vcc_lo, v1, v6
; GFX12-TRUE16-NEXT: s_or_b32 s4, vcc_lo, s4
; GFX12-TRUE16-NEXT: s_wait_alu 0xfffe
; GFX12-TRUE16-NEXT: s_and_not1_b32 exec_lo, exec_lo, s4
@@ -7971,41 +7993,43 @@ define void @buffer_fat_ptr_agent_atomic_fmax_noret_v2bf16__offset__amdgpu_no_fi
; GFX12-FAKE16-NEXT: s_wait_samplecnt 0x0
; GFX12-FAKE16-NEXT: s_wait_bvhcnt 0x0
; GFX12-FAKE16-NEXT: s_wait_kmcnt 0x0
-; GFX12-FAKE16-NEXT: v_dual_mov_b32 v1, s16 :: v_dual_lshlrev_b32 v2, 16, v0
+; GFX12-FAKE16-NEXT: v_mov_b32_e32 v1, s16
; GFX12-FAKE16-NEXT: s_add_co_i32 s4, s16, 0x400
; GFX12-FAKE16-NEXT: s_wait_alu 0xfffe
-; GFX12-FAKE16-NEXT: v_dual_mov_b32 v4, s4 :: v_dual_and_b32 v3, 0xffff0000, v0
+; GFX12-FAKE16-NEXT: v_dual_mov_b32 v4, s4 :: v_dual_lshlrev_b32 v3, 16, v0
+; GFX12-FAKE16-NEXT: v_and_b32_e32 v0, 0xffff0000, v0
; GFX12-FAKE16-NEXT: buffer_load_b32 v1, v1, s[0:3], null offen offset:1024
; GFX12-FAKE16-NEXT: s_mov_b32 s5, 0
; GFX12-FAKE16-NEXT: .LBB20_1: ; %atomicrmw.start
; GFX12-FAKE16-NEXT: ; =>This Inner Loop Header: Depth=1
; GFX12-FAKE16-NEXT: s_wait_loadcnt 0x0
-; GFX12-FAKE16-NEXT: v_and_b32_e32 v5, 0xffff0000, v1
-; GFX12-FAKE16-NEXT: v_lshlrev_b32_e32 v0, 16, v1
+; GFX12-FAKE16-NEXT: v_mov_b32_e32 v6, v1
; GFX12-FAKE16-NEXT: s_wait_storecnt 0x0
; GFX12-FAKE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
-; GFX12-FAKE16-NEXT: v_dual_max_num_f32 v5, v5, v3 :: v_dual_max_num_f32 v0, v0, v2
-; GFX12-FAKE16-NEXT: v_bfe_u32 v7, v5, 16, 1
+; GFX12-FAKE16-NEXT: v_and_b32_e32 v2, 0xffff0000, v6
+; GFX12-FAKE16-NEXT: v_dual_max_num_f32 v2, v2, v0 :: v_dual_lshlrev_b32 v1, 16, v6
+; GFX12-FAKE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_2)
+; GFX12-FAKE16-NEXT: v_max_num_f32_e32 v1, v1, v3
+; GFX12-FAKE16-NEXT: v_bfe_u32 v7, v2, 16, 1
; GFX12-FAKE16-NEXT: s_delay_alu instid0(VALU_DEP_2)
-; GFX12-FAKE16-NEXT: v_bfe_u32 v6, v0, 16, 1
-; GFX12-FAKE16-NEXT: v_or_b32_e32 v8, 0x400000, v0
-; GFX12-FAKE16-NEXT: v_or_b32_e32 v9, 0x400000, v5
-; GFX12-FAKE16-NEXT: v_cmp_u_f32_e32 vcc_lo, v5, v5
-; GFX12-FAKE16-NEXT: v_add3_u32 v7, v7, v5, 0x7fff
-; GFX12-FAKE16-NEXT: v_add3_u32 v6, v6, v0, 0x7fff
-; GFX12-FAKE16-NEXT: v_cmp_u_f32_e64 s4, v0, v0
+; GFX12-FAKE16-NEXT: v_bfe_u32 v5, v1, 16, 1
+; GFX12-FAKE16-NEXT: v_or_b32_e32 v8, 0x400000, v1
+; GFX12-FAKE16-NEXT: v_or_b32_e32 v9, 0x400000, v2
+; GFX12-FAKE16-NEXT: v_cmp_u_f32_e32 vcc_lo, v2, v2
+; GFX12-FAKE16-NEXT: v_add3_u32 v7, v7, v2, 0x7fff
+; GFX12-FAKE16-NEXT: v_add3_u32 v5, v5, v1, 0x7fff
+; GFX12-FAKE16-NEXT: v_cmp_u_f32_e64 s4, v1, v1
; GFX12-FAKE16-NEXT: s_wait_alu 0xfffd
-; GFX12-FAKE16-NEXT: v_cndmask_b32_e32 v5, v7, v9, vcc_lo
+; GFX12-FAKE16-NEXT: v_cndmask_b32_e32 v2, v7, v9, vcc_lo
; GFX12-FAKE16-NEXT: s_wait_alu 0xf1ff
-; GFX12-FAKE16-NEXT: v_cndmask_b32_e64 v0, v6, v8, s4
+; GFX12-FAKE16-NEXT: v_cndmask_b32_e64 v1, v5, v8, s4
; GFX12-FAKE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
-; GFX12-FAKE16-NEXT: v_perm_b32 v0, v5, v0, 0x7060302
-; GFX12-FAKE16-NEXT: v_dual_mov_b32 v6, v1 :: v_dual_mov_b32 v5, v0
-; GFX12-FAKE16-NEXT: buffer_atomic_cmpswap_b32 v[5:6], v4, s[0:3], null offen th:TH_ATOMIC_RETURN
+; GFX12-FAKE16-NEXT: v_perm_b32 v5, v2, v1, 0x7060302
+; GFX12-FAKE16-NEXT: v_dual_mov_b32 v1, v5 :: v_dual_mov_b32 v2, v6
+; GFX12-FAKE16-NEXT: buffer_atomic_cmpswap_b32 v[1:2], v4, s[0:3], null offen th:TH_ATOMIC_RETURN
; GFX12-FAKE16-NEXT: s_wait_loadcnt 0x0
; GFX12-FAKE16-NEXT: global_inv scope:SCOPE_DEV
-; GFX12-FAKE16-NEXT: v_cmp_eq_u32_e32 vcc_lo, v5, v1
-; GFX12-FAKE16-NEXT: v_mov_b32_e32 v1, v5
+; GFX12-FAKE16-NEXT: v_cmp_eq_u32_e32 vcc_lo, v1, v6
; GFX12-FAKE16-NEXT: s_or_b32 s5, vcc_lo, s5
; GFX12-FAKE16-NEXT: s_wait_alu 0xfffe
; GFX12-FAKE16-NEXT: s_and_not1_b32 exec_lo, exec_lo, s5
@@ -8019,40 +8043,40 @@ define void @buffer_fat_ptr_agent_atomic_fmax_noret_v2bf16__offset__amdgpu_no_fi
; GFX942: ; %bb.0:
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX942-NEXT: v_mov_b32_e32 v1, s16
-; GFX942-NEXT: buffer_load_dword v1, v1, s[0:3], 0 offen offset:1024
+; GFX942-NEXT: buffer_load_dword v2, v1, s[0:3], 0 offen offset:1024
; GFX942-NEXT: s_add_i32 s4, s16, 0x400
; GFX942-NEXT: s_mov_b64 s[6:7], 0
-; GFX942-NEXT: v_lshlrev_b32_e32 v2, 16, v0
+; GFX942-NEXT: v_lshlrev_b32_e32 v1, 16, v0
; GFX942-NEXT: s_movk_i32 s8, 0x7fff
-; GFX942-NEXT: v_and_b32_e32 v3, 0xffff0000, v0
+; GFX942-NEXT: v_and_b32_e32 v0, 0xffff0000, v0
; GFX942-NEXT: s_mov_b32 s9, 0x7060302
; GFX942-NEXT: v_mov_b32_e32 v4, s4
; GFX942-NEXT: .LBB20_1: ; %atomicrmw.start
; GFX942-NEXT: ; =>This Inner Loop Header: Depth=1
; GFX942-NEXT: s_waitcnt vmcnt(0)
-; GFX942-NEXT: v_lshlrev_b32_e32 v0, 16, v1
-; GFX942-NEXT: v_and_b32_e32 v5, 0xffff0000, v1
-; GFX942-NEXT: v_max_f32_e32 v0, v0, v2
-; GFX942-NEXT: v_max_f32_e32 v5, v5, v3
-; GFX942-NEXT: v_bfe_u32 v6, v0, 16, 1
-; GFX942-NEXT: v_bfe_u32 v8, v5, 16, 1
-; GFX942-NEXT: v_or_b32_e32 v7, 0x400000, v0
-; GFX942-NEXT: v_or_b32_e32 v9, 0x400000, v5
-; GFX942-NEXT: v_add3_u32 v6, v6, v0, s8
-; GFX942-NEXT: v_add3_u32 v8, v8, v5, s8
-; GFX942-NEXT: v_cmp_u_f32_e32 vcc, v5, v5
-; GFX942-NEXT: v_cmp_u_f32_e64 s[4:5], v0, v0
+; GFX942-NEXT: v_mov_b32_e32 v7, v2
+; GFX942-NEXT: v_lshlrev_b32_e32 v2, 16, v7
+; GFX942-NEXT: v_and_b32_e32 v3, 0xffff0000, v7
+; GFX942-NEXT: v_max_f32_e32 v2, v2, v1
+; GFX942-NEXT: v_max_f32_e32 v3, v3, v0
+; GFX942-NEXT: v_bfe_u32 v5, v2, 16, 1
+; GFX942-NEXT: v_bfe_u32 v8, v3, 16, 1
+; GFX942-NEXT: v_or_b32_e32 v6, 0x400000, v2
+; GFX942-NEXT: v_or_b32_e32 v9, 0x400000, v3
+; GFX942-NEXT: v_add3_u32 v5, v5, v2, s8
+; GFX942-NEXT: v_add3_u32 v8, v8, v3, s8
+; GFX942-NEXT: v_cmp_u_f32_e32 vcc, v3, v3
+; GFX942-NEXT: v_cmp_u_f32_e64 s[4:5], v2, v2
; GFX942-NEXT: buffer_wbl2 sc1
-; GFX942-NEXT: v_cndmask_b32_e32 v5, v8, v9, vcc
-; GFX942-NEXT: v_cndmask_b32_e64 v0, v6, v7, s[4:5]
-; GFX942-NEXT: v_perm_b32 v0, v5, v0, s9
-; GFX942-NEXT: v_mov_b64_e32 v[6:7], v[0:1]
-; GFX942-NEXT: buffer_atomic_cmpswap v[6:7], v4, s[0:3], 0 offen sc0
+; GFX942-NEXT: v_cndmask_b32_e32 v3, v8, v9, vcc
+; GFX942-NEXT: v_cndmask_b32_e64 v2, v5, v6, s[4:5]
+; GFX942-NEXT: v_perm_b32 v6, v3, v2, s9
+; GFX942-NEXT: v_mov_b64_e32 v[2:3], v[6:7]
+; GFX942-NEXT: buffer_atomic_cmpswap v[2:3], v4, s[0:3], 0 offen sc0
; GFX942-NEXT: s_waitcnt vmcnt(0)
; GFX942-NEXT: buffer_inv sc1
-; GFX942-NEXT: v_cmp_eq_u32_e32 vcc, v6, v1
+; GFX942-NEXT: v_cmp_eq_u32_e32 vcc, v2, v7
; GFX942-NEXT: s_or_b64 s[6:7], vcc, s[6:7]
-; GFX942-NEXT: v_mov_b32_e32 v1, v6
; GFX942-NEXT: s_andn2_b64 exec, exec, s[6:7]
; GFX942-NEXT: s_cbranch_execnz .LBB20_1
; GFX942-NEXT: ; %bb.2: ; %atomicrmw.end
@@ -8062,10 +8086,11 @@ define void @buffer_fat_ptr_agent_atomic_fmax_noret_v2bf16__offset__amdgpu_no_fi
; GFX11-TRUE16-LABEL: buffer_fat_ptr_agent_atomic_fmax_noret_v2bf16__offset__amdgpu_no_fine_grained_memory:
; GFX11-TRUE16: ; %bb.0:
; GFX11-TRUE16-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
-; GFX11-TRUE16-NEXT: v_dual_mov_b32 v1, s16 :: v_dual_and_b32 v2, 0xffff0000, v0
+; GFX11-TRUE16-NEXT: v_mov_b32_e32 v1, s16
; GFX11-TRUE16-NEXT: s_add_i32 s4, s16, 0x400
; GFX11-TRUE16-NEXT: s_delay_alu instid0(SALU_CYCLE_1)
-; GFX11-TRUE16-NEXT: v_dual_mov_b32 v4, s4 :: v_dual_lshlrev_b32 v3, 16, v0
+; GFX11-TRUE16-NEXT: v_dual_mov_b32 v4, s4 :: v_dual_and_b32 v3, 0xffff0000, v0
+; GFX11-TRUE16-NEXT: v_lshlrev_b32_e32 v0, 16, v0
; GFX11-TRUE16-NEXT: buffer_load_b32 v1, v1, s[0:3], 0 offen offset:1024
; GFX11-TRUE16-NEXT: s_mov_b32 s4, 0
; GFX11-TRUE16-NEXT: s_set_inst_prefetch_distance 0x1
@@ -8073,34 +8098,35 @@ define void @buffer_fat_ptr_agent_atomic_fmax_noret_v2bf16__offset__amdgpu_no_fi
; GFX11-TRUE16-NEXT: .LBB20_1: ; %atomicrmw.start
; GFX11-TRUE16-NEXT: ; =>This Inner Loop Header: Depth=1
; GFX11-TRUE16-NEXT: s_waitcnt vmcnt(0)
-; GFX11-TRUE16-NEXT: v_and_b32_e32 v5, 0xffff0000, v1
-; GFX11-TRUE16-NEXT: v_lshlrev_b32_e32 v0, 16, v1
+; GFX11-TRUE16-NEXT: v_mov_b32_e32 v6, v1
; GFX11-TRUE16-NEXT: s_waitcnt_vscnt null, 0x0
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
-; GFX11-TRUE16-NEXT: v_dual_max_f32 v5, v5, v2 :: v_dual_max_f32 v0, v0, v3
-; GFX11-TRUE16-NEXT: v_bfe_u32 v7, v5, 16, 1
+; GFX11-TRUE16-NEXT: v_and_b32_e32 v2, 0xffff0000, v6
+; GFX11-TRUE16-NEXT: v_dual_max_f32 v2, v2, v3 :: v_dual_lshlrev_b32 v1, 16, v6
+; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_2)
+; GFX11-TRUE16-NEXT: v_max_f32_e32 v1, v1, v0
+; GFX11-TRUE16-NEXT: v_bfe_u32 v7, v2, 16, 1
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_2)
-; GFX11-TRUE16-NEXT: v_bfe_u32 v6, v0, 16, 1
-; GFX11-TRUE16-NEXT: v_or_b32_e32 v8, 0x400000, v0
-; GFX11-TRUE16-NEXT: v_cmp_u_f32_e32 vcc_lo, v0, v0
-; GFX11-TRUE16-NEXT: v_or_b32_e32 v9, 0x400000, v5
-; GFX11-TRUE16-NEXT: v_add3_u32 v7, v7, v5, 0x7fff
-; GFX11-TRUE16-NEXT: v_add3_u32 v6, v6, v0, 0x7fff
+; GFX11-TRUE16-NEXT: v_bfe_u32 v5, v1, 16, 1
+; GFX11-TRUE16-NEXT: v_or_b32_e32 v8, 0x400000, v1
+; GFX11-TRUE16-NEXT: v_cmp_u_f32_e32 vcc_lo, v1, v1
+; GFX11-TRUE16-NEXT: v_or_b32_e32 v9, 0x400000, v2
+; GFX11-TRUE16-NEXT: v_add3_u32 v7, v7, v2, 0x7fff
+; GFX11-TRUE16-NEXT: v_add3_u32 v5, v5, v1, 0x7fff
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(SKIP_1) | instid1(VALU_DEP_4)
-; GFX11-TRUE16-NEXT: v_cndmask_b32_e32 v0, v6, v8, vcc_lo
-; GFX11-TRUE16-NEXT: v_cmp_u_f32_e32 vcc_lo, v5, v5
-; GFX11-TRUE16-NEXT: v_cndmask_b32_e32 v5, v7, v9, vcc_lo
+; GFX11-TRUE16-NEXT: v_cndmask_b32_e32 v1, v5, v8, vcc_lo
+; GFX11-TRUE16-NEXT: v_cmp_u_f32_e32 vcc_lo, v2, v2
+; GFX11-TRUE16-NEXT: v_cndmask_b32_e32 v2, v7, v9, vcc_lo
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_3) | instskip(NEXT) | instid1(VALU_DEP_1)
-; GFX11-TRUE16-NEXT: v_mov_b16_e32 v0.l, v0.h
-; GFX11-TRUE16-NEXT: v_bfi_b32 v0, 0xffff, v0, v5
+; GFX11-TRUE16-NEXT: v_mov_b16_e32 v1.l, v1.h
+; GFX11-TRUE16-NEXT: v_bfi_b32 v5, 0xffff, v1, v2
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1)
-; GFX11-TRUE16-NEXT: v_dual_mov_b32 v6, v1 :: v_dual_mov_b32 v5, v0
-; GFX11-TRUE16-NEXT: buffer_atomic_cmpswap_b32 v[5:6], v4, s[0:3], 0 offen glc
+; GFX11-TRUE16-NEXT: v_dual_mov_b32 v1, v5 :: v_dual_mov_b32 v2, v6
+; GFX11-TRUE16-NEXT: buffer_atomic_cmpswap_b32 v[1:2], v4, s[0:3], 0 offen glc
; GFX11-TRUE16-NEXT: s_waitcnt vmcnt(0)
; GFX11-TRUE16-NEXT: buffer_gl1_inv
; GFX11-TRUE16-NEXT: buffer_gl0_inv
-; GFX11-TRUE16-NEXT: v_cmp_eq_u32_e32 vcc_lo, v5, v1
-; GFX11-TRUE16-NEXT: v_mov_b32_e32 v1, v5
+; GFX11-TRUE16-NEXT: v_cmp_eq_u32_e32 vcc_lo, v1, v6
; GFX11-TRUE16-NEXT: s_or_b32 s4, vcc_lo, s4
; GFX11-TRUE16-NEXT: s_delay_alu instid0(SALU_CYCLE_1)
; GFX11-TRUE16-NEXT: s_and_not1_b32 exec_lo, exec_lo, s4
@@ -8113,10 +8139,11 @@ define void @buffer_fat_ptr_agent_atomic_fmax_noret_v2bf16__offset__amdgpu_no_fi
; GFX11-FAKE16-LABEL: buffer_fat_ptr_agent_atomic_fmax_noret_v2bf16__offset__amdgpu_no_fine_grained_memory:
; GFX11-FAKE16: ; %bb.0:
; GFX11-FAKE16-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
-; GFX11-FAKE16-NEXT: v_dual_mov_b32 v1, s16 :: v_dual_lshlrev_b32 v2, 16, v0
+; GFX11-FAKE16-NEXT: v_mov_b32_e32 v1, s16
; GFX11-FAKE16-NEXT: s_add_i32 s4, s16, 0x400
; GFX11-FAKE16-NEXT: s_delay_alu instid0(SALU_CYCLE_1)
-; GFX11-FAKE16-NEXT: v_dual_mov_b32 v4, s4 :: v_dual_and_b32 v3, 0xffff0000, v0
+; GFX11-FAKE16-NEXT: v_dual_mov_b32 v4, s4 :: v_dual_lshlrev_b32 v3, 16, v0
+; GFX11-FAKE16-NEXT: v_and_b32_e32 v0, 0xffff0000, v0
; GFX11-FAKE16-NEXT: buffer_load_b32 v1, v1, s[0:3], 0 offen offset:1024
; GFX11-FAKE16-NEXT: s_mov_b32 s5, 0
; GFX11-FAKE16-NEXT: s_set_inst_prefetch_distance 0x1
@@ -8124,32 +8151,33 @@ define void @buffer_fat_ptr_agent_atomic_fmax_noret_v2bf16__offset__amdgpu_no_fi
; GFX11-FAKE16-NEXT: .LBB20_1: ; %atomicrmw.start
; GFX11-FAKE16-NEXT: ; =>This Inner Loop Header: Depth=1
; GFX11-FAKE16-NEXT: s_waitcnt vmcnt(0)
-; GFX11-FAKE16-NEXT: v_and_b32_e32 v5, 0xffff0000, v1
-; GFX11-FAKE16-NEXT: v_lshlrev_b32_e32 v0, 16, v1
+; GFX11-FAKE16-NEXT: v_mov_b32_e32 v6, v1
; GFX11-FAKE16-NEXT: s_waitcnt_vscnt null, 0x0
; GFX11-FAKE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
-; GFX11-FAKE16-NEXT: v_dual_max_f32 v5, v5, v3 :: v_dual_max_f32 v0, v0, v2
-; GFX11-FAKE16-NEXT: v_bfe_u32 v7, v5, 16, 1
+; GFX11-FAKE16-NEXT: v_and_b32_e32 v2, 0xffff0000, v6
+; GFX11-FAKE16-NEXT: v_dual_max_f32 v2, v2, v0 :: v_dual_lshlrev_b32 v1, 16, v6
+; GFX11-FAKE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_2)
+; GFX11-FAKE16-NEXT: v_max_f32_e32 v1, v1, v3
+; GFX11-FAKE16-NEXT: v_bfe_u32 v7, v2, 16, 1
; GFX11-FAKE16-NEXT: s_delay_alu instid0(VALU_DEP_2)
-; GFX11-FAKE16-NEXT: v_bfe_u32 v6, v0, 16, 1
-; GFX11-FAKE16-NEXT: v_or_b32_e32 v8, 0x400000, v0
-; GFX11-FAKE16-NEXT: v_or_b32_e32 v9, 0x400000, v5
-; GFX11-FAKE16-NEXT: v_cmp_u_f32_e32 vcc_lo, v5, v5
-; GFX11-FAKE16-NEXT: v_add3_u32 v7, v7, v5, 0x7fff
-; GFX11-FAKE16-NEXT: v_add3_u32 v6, v6, v0, 0x7fff
-; GFX11-FAKE16-NEXT: v_cmp_u_f32_e64 s4, v0, v0
+; GFX11-FAKE16-NEXT: v_bfe_u32 v5, v1, 16, 1
+; GFX11-FAKE16-NEXT: v_or_b32_e32 v8, 0x400000, v1
+; GFX11-FAKE16-NEXT: v_or_b32_e32 v9, 0x400000, v2
+; GFX11-FAKE16-NEXT: v_cmp_u_f32_e32 vcc_lo, v2, v2
+; GFX11-FAKE16-NEXT: v_add3_u32 v7, v7, v2, 0x7fff
+; GFX11-FAKE16-NEXT: v_add3_u32 v5, v5, v1, 0x7fff
+; GFX11-FAKE16-NEXT: v_cmp_u_f32_e64 s4, v1, v1
; GFX11-FAKE16-NEXT: s_delay_alu instid0(VALU_DEP_3) | instskip(NEXT) | instid1(VALU_DEP_2)
-; GFX11-FAKE16-NEXT: v_cndmask_b32_e32 v5, v7, v9, vcc_lo
-; GFX11-FAKE16-NEXT: v_cndmask_b32_e64 v0, v6, v8, s4
+; GFX11-FAKE16-NEXT: v_cndmask_b32_e32 v2, v7, v9, vcc_lo
+; GFX11-FAKE16-NEXT: v_cndmask_b32_e64 v1, v5, v8, s4
; GFX11-FAKE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
-; GFX11-FAKE16-NEXT: v_perm_b32 v0, v5, v0, 0x7060302
-; GFX11-FAKE16-NEXT: v_dual_mov_b32 v6, v1 :: v_dual_mov_b32 v5, v0
-; GFX11-FAKE16-NEXT: buffer_atomic_cmpswap_b32 v[5:6], v4, s[0:3], 0 offen glc
+; GFX11-FAKE16-NEXT: v_perm_b32 v5, v2, v1, 0x7060302
+; GFX11-FAKE16-NEXT: v_dual_mov_b32 v1, v5 :: v_dual_mov_b32 v2, v6
+; GFX11-FAKE16-NEXT: buffer_atomic_cmpswap_b32 v[1:2], v4, s[0:3], 0 offen glc
; GFX11-FAKE16-NEXT: s_waitcnt vmcnt(0)
; GFX11-FAKE16-NEXT: buffer_gl1_inv
; GFX11-FAKE16-NEXT: buffer_gl0_inv
-; GFX11-FAKE16-NEXT: v_cmp_eq_u32_e32 vcc_lo, v5, v1
-; GFX11-FAKE16-NEXT: v_mov_b32_e32 v1, v5
+; GFX11-FAKE16-NEXT: v_cmp_eq_u32_e32 vcc_lo, v1, v6
; GFX11-FAKE16-NEXT: s_or_b32 s5, vcc_lo, s5
; GFX11-FAKE16-NEXT: s_delay_alu instid0(SALU_CYCLE_1)
; GFX11-FAKE16-NEXT: s_and_not1_b32 exec_lo, exec_lo, s5
@@ -8164,38 +8192,38 @@ define void @buffer_fat_ptr_agent_atomic_fmax_noret_v2bf16__offset__amdgpu_no_fi
; GFX10-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX10-NEXT: v_mov_b32_e32 v1, s20
; GFX10-NEXT: s_add_i32 s4, s20, 0x400
-; GFX10-NEXT: v_lshlrev_b32_e32 v2, 16, v0
-; GFX10-NEXT: v_and_b32_e32 v3, 0xffff0000, v0
+; GFX10-NEXT: v_lshlrev_b32_e32 v3, 16, v0
+; GFX10-NEXT: v_and_b32_e32 v0, 0xffff0000, v0
; GFX10-NEXT: v_mov_b32_e32 v4, s4
; GFX10-NEXT: buffer_load_dword v1, v1, s[16:19], 0 offen offset:1024
; GFX10-NEXT: s_mov_b32 s5, 0
; GFX10-NEXT: .LBB20_1: ; %atomicrmw.start
; GFX10-NEXT: ; =>This Inner Loop Header: Depth=1
; GFX10-NEXT: s_waitcnt vmcnt(0)
-; GFX10-NEXT: v_lshlrev_b32_e32 v0, 16, v1
-; GFX10-NEXT: v_and_b32_e32 v5, 0xffff0000, v1
-; GFX10-NEXT: s_waitcnt_vscnt null, 0x0
-; GFX10-NEXT: v_max_f32_e32 v0, v0, v2
-; GFX10-NEXT: v_max_f32_e32 v5, v5, v3
-; GFX10-NEXT: v_bfe_u32 v6, v0, 16, 1
-; GFX10-NEXT: v_bfe_u32 v7, v5, 16, 1
-; GFX10-NEXT: v_or_b32_e32 v8, 0x400000, v0
-; GFX10-NEXT: v_or_b32_e32 v9, 0x400000, v5
-; GFX10-NEXT: v_cmp_u_f32_e32 vcc_lo, v5, v5
-; GFX10-NEXT: v_add3_u32 v6, v6, v0, 0x7fff
-; GFX10-NEXT: v_add3_u32 v7, v7, v5, 0x7fff
-; GFX10-NEXT: v_cmp_u_f32_e64 s4, v0, v0
-; GFX10-NEXT: v_cndmask_b32_e32 v5, v7, v9, vcc_lo
-; GFX10-NEXT: v_cndmask_b32_e64 v0, v6, v8, s4
-; GFX10-NEXT: v_perm_b32 v0, v5, v0, 0x7060302
; GFX10-NEXT: v_mov_b32_e32 v6, v1
-; GFX10-NEXT: v_mov_b32_e32 v5, v0
-; GFX10-NEXT: buffer_atomic_cmpswap v[5:6], v4, s[16:19], 0 offen glc
+; GFX10-NEXT: s_waitcnt_vscnt null, 0x0
+; GFX10-NEXT: v_lshlrev_b32_e32 v1, 16, v6
+; GFX10-NEXT: v_and_b32_e32 v2, 0xffff0000, v6
+; GFX10-NEXT: v_max_f32_e32 v1, v1, v3
+; GFX10-NEXT: v_max_f32_e32 v2, v2, v0
+; GFX10-NEXT: v_bfe_u32 v5, v1, 16, 1
+; GFX10-NEXT: v_bfe_u32 v7, v2, 16, 1
+; GFX10-NEXT: v_or_b32_e32 v8, 0x400000, v1
+; GFX10-NEXT: v_or_b32_e32 v9, 0x400000, v2
+; GFX10-NEXT: v_cmp_u_f32_e32 vcc_lo, v2, v2
+; GFX10-NEXT: v_add3_u32 v5, v5, v1, 0x7fff
+; GFX10-NEXT: v_add3_u32 v7, v7, v2, 0x7fff
+; GFX10-NEXT: v_cmp_u_f32_e64 s4, v1, v1
+; GFX10-NEXT: v_cndmask_b32_e32 v2, v7, v9, vcc_lo
+; GFX10-NEXT: v_cndmask_b32_e64 v1, v5, v8, s4
+; GFX10-NEXT: v_perm_b32 v5, v2, v1, 0x7060302
+; GFX10-NEXT: v_mov_b32_e32 v1, v5
+; GFX10-NEXT: v_mov_b32_e32 v2, v6
+; GFX10-NEXT: buffer_atomic_cmpswap v[1:2], v4, s[16:19], 0 offen glc
; GFX10-NEXT: s_waitcnt vmcnt(0)
; GFX10-NEXT: buffer_gl1_inv
; GFX10-NEXT: buffer_gl0_inv
-; GFX10-NEXT: v_cmp_eq_u32_e32 vcc_lo, v5, v1
-; GFX10-NEXT: v_mov_b32_e32 v1, v5
+; GFX10-NEXT: v_cmp_eq_u32_e32 vcc_lo, v1, v6
; GFX10-NEXT: s_or_b32 s5, vcc_lo, s5
; GFX10-NEXT: s_andn2_b32 exec_lo, exec_lo, s5
; GFX10-NEXT: s_cbranch_execnz .LBB20_1
@@ -8207,39 +8235,39 @@ define void @buffer_fat_ptr_agent_atomic_fmax_noret_v2bf16__offset__amdgpu_no_fi
; GFX90A: ; %bb.0:
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX90A-NEXT: v_mov_b32_e32 v1, s20
-; GFX90A-NEXT: buffer_load_dword v1, v1, s[16:19], 0 offen offset:1024
+; GFX90A-NEXT: buffer_load_dword v2, v1, s[16:19], 0 offen offset:1024
; GFX90A-NEXT: s_add_i32 s4, s20, 0x400
; GFX90A-NEXT: s_mov_b64 s[6:7], 0
-; GFX90A-NEXT: v_lshlrev_b32_e32 v2, 16, v0
+; GFX90A-NEXT: v_lshlrev_b32_e32 v1, 16, v0
; GFX90A-NEXT: s_movk_i32 s8, 0x7fff
-; GFX90A-NEXT: v_and_b32_e32 v3, 0xffff0000, v0
+; GFX90A-NEXT: v_and_b32_e32 v0, 0xffff0000, v0
; GFX90A-NEXT: s_mov_b32 s9, 0x7060302
; GFX90A-NEXT: v_mov_b32_e32 v4, s4
; GFX90A-NEXT: .LBB20_1: ; %atomicrmw.start
; GFX90A-NEXT: ; =>This Inner Loop Header: Depth=1
; GFX90A-NEXT: s_waitcnt vmcnt(0)
-; GFX90A-NEXT: v_lshlrev_b32_e32 v0, 16, v1
-; GFX90A-NEXT: v_and_b32_e32 v5, 0xffff0000, v1
-; GFX90A-NEXT: v_max_f32_e32 v0, v0, v2
-; GFX90A-NEXT: v_max_f32_e32 v5, v5, v3
-; GFX90A-NEXT: v_bfe_u32 v6, v0, 16, 1
-; GFX90A-NEXT: v_bfe_u32 v8, v5, 16, 1
-; GFX90A-NEXT: v_or_b32_e32 v7, 0x400000, v0
-; GFX90A-NEXT: v_or_b32_e32 v9, 0x400000, v5
-; GFX90A-NEXT: v_add3_u32 v6, v6, v0, s8
-; GFX90A-NEXT: v_add3_u32 v8, v8, v5, s8
-; GFX90A-NEXT: v_cmp_u_f32_e32 vcc, v5, v5
-; GFX90A-NEXT: v_cmp_u_f32_e64 s[4:5], v0, v0
-; GFX90A-NEXT: v_cndmask_b32_e64 v0, v6, v7, s[4:5]
-; GFX90A-NEXT: v_cndmask_b32_e32 v5, v8, v9, vcc
-; GFX90A-NEXT: v_perm_b32 v0, v5, v0, s9
-; GFX90A-NEXT: v_pk_mov_b32 v[6:7], v[0:1], v[0:1] op_sel:[0,1]
-; GFX90A-NEXT: buffer_atomic_cmpswap v[6:7], v4, s[16:19], 0 offen glc
+; GFX90A-NEXT: v_mov_b32_e32 v7, v2
+; GFX90A-NEXT: v_lshlrev_b32_e32 v2, 16, v7
+; GFX90A-NEXT: v_and_b32_e32 v3, 0xffff0000, v7
+; GFX90A-NEXT: v_max_f32_e32 v2, v2, v1
+; GFX90A-NEXT: v_max_f32_e32 v3, v3, v0
+; GFX90A-NEXT: v_bfe_u32 v5, v2, 16, 1
+; GFX90A-NEXT: v_bfe_u32 v8, v3, 16, 1
+; GFX90A-NEXT: v_or_b32_e32 v6, 0x400000, v2
+; GFX90A-NEXT: v_or_b32_e32 v9, 0x400000, v3
+; GFX90A-NEXT: v_add3_u32 v5, v5, v2, s8
+; GFX90A-NEXT: v_add3_u32 v8, v8, v3, s8
+; GFX90A-NEXT: v_cmp_u_f32_e32 vcc, v3, v3
+; GFX90A-NEXT: v_cmp_u_f32_e64 s[4:5], v2, v2
+; GFX90A-NEXT: v_cndmask_b32_e64 v2, v5, v6, s[4:5]
+; GFX90A-NEXT: v_cndmask_b32_e32 v3, v8, v9, vcc
+; GFX90A-NEXT: v_perm_b32 v6, v3, v2, s9
+; GFX90A-NEXT: v_pk_mov_b32 v[2:3], v[6:7], v[6:7] op_sel:[0,1]
+; GFX90A-NEXT: buffer_atomic_cmpswap v[2:3], v4, s[16:19], 0 offen glc
; GFX90A-NEXT: s_waitcnt vmcnt(0)
; GFX90A-NEXT: buffer_wbinvl1
-; GFX90A-NEXT: v_cmp_eq_u32_e32 vcc, v6, v1
+; GFX90A-NEXT: v_cmp_eq_u32_e32 vcc, v2, v7
; GFX90A-NEXT: s_or_b64 s[6:7], vcc, s[6:7]
-; GFX90A-NEXT: v_mov_b32_e32 v1, v6
; GFX90A-NEXT: s_andn2_b64 exec, exec, s[6:7]
; GFX90A-NEXT: s_cbranch_execnz .LBB20_1
; GFX90A-NEXT: ; %bb.2: ; %atomicrmw.end
@@ -8253,37 +8281,37 @@ define void @buffer_fat_ptr_agent_atomic_fmax_noret_v2bf16__offset__amdgpu_no_fi
; GFX908-NEXT: buffer_load_dword v1, v1, s[16:19], 0 offen offset:1024
; GFX908-NEXT: s_add_i32 s4, s20, 0x400
; GFX908-NEXT: s_mov_b64 s[6:7], 0
-; GFX908-NEXT: v_lshlrev_b32_e32 v2, 16, v0
+; GFX908-NEXT: v_lshlrev_b32_e32 v3, 16, v0
; GFX908-NEXT: s_movk_i32 s8, 0x7fff
-; GFX908-NEXT: v_and_b32_e32 v3, 0xffff0000, v0
+; GFX908-NEXT: v_and_b32_e32 v0, 0xffff0000, v0
; GFX908-NEXT: s_mov_b32 s9, 0x7060302
; GFX908-NEXT: v_mov_b32_e32 v4, s4
; GFX908-NEXT: .LBB20_1: ; %atomicrmw.start
; GFX908-NEXT: ; =>This Inner Loop Header: Depth=1
; GFX908-NEXT: s_waitcnt vmcnt(0)
-; GFX908-NEXT: v_lshlrev_b32_e32 v0, 16, v1
-; GFX908-NEXT: v_and_b32_e32 v5, 0xffff0000, v1
-; GFX908-NEXT: v_max_f32_e32 v0, v0, v2
-; GFX908-NEXT: v_max_f32_e32 v5, v5, v3
-; GFX908-NEXT: v_bfe_u32 v6, v0, 16, 1
-; GFX908-NEXT: v_bfe_u32 v8, v5, 16, 1
-; GFX908-NEXT: v_or_b32_e32 v7, 0x400000, v0
-; GFX908-NEXT: v_or_b32_e32 v9, 0x400000, v5
-; GFX908-NEXT: v_add3_u32 v6, v6, v0, s8
-; GFX908-NEXT: v_add3_u32 v8, v8, v5, s8
-; GFX908-NEXT: v_cmp_u_f32_e32 vcc, v5, v5
-; GFX908-NEXT: v_cmp_u_f32_e64 s[4:5], v0, v0
-; GFX908-NEXT: v_cndmask_b32_e64 v0, v6, v7, s[4:5]
-; GFX908-NEXT: v_cndmask_b32_e32 v5, v8, v9, vcc
-; GFX908-NEXT: v_perm_b32 v0, v5, v0, s9
; GFX908-NEXT: v_mov_b32_e32 v6, v1
-; GFX908-NEXT: v_mov_b32_e32 v5, v0
-; GFX908-NEXT: buffer_atomic_cmpswap v[5:6], v4, s[16:19], 0 offen glc
+; GFX908-NEXT: v_lshlrev_b32_e32 v1, 16, v6
+; GFX908-NEXT: v_and_b32_e32 v2, 0xffff0000, v6
+; GFX908-NEXT: v_max_f32_e32 v1, v1, v3
+; GFX908-NEXT: v_max_f32_e32 v2, v2, v0
+; GFX908-NEXT: v_bfe_u32 v5, v1, 16, 1
+; GFX908-NEXT: v_bfe_u32 v8, v2, 16, 1
+; GFX908-NEXT: v_or_b32_e32 v7, 0x400000, v1
+; GFX908-NEXT: v_or_b32_e32 v9, 0x400000, v2
+; GFX908-NEXT: v_add3_u32 v5, v5, v1, s8
+; GFX908-NEXT: v_add3_u32 v8, v8, v2, s8
+; GFX908-NEXT: v_cmp_u_f32_e32 vcc, v2, v2
+; GFX908-NEXT: v_cmp_u_f32_e64 s[4:5], v1, v1
+; GFX908-NEXT: v_cndmask_b32_e64 v1, v5, v7, s[4:5]
+; GFX908-NEXT: v_cndmask_b32_e32 v2, v8, v9, vcc
+; GFX908-NEXT: v_perm_b32 v5, v2, v1, s9
+; GFX908-NEXT: v_mov_b32_e32 v1, v5
+; GFX908-NEXT: v_mov_b32_e32 v2, v6
+; GFX908-NEXT: buffer_atomic_cmpswap v[1:2], v4, s[16:19], 0 offen glc
; GFX908-NEXT: s_waitcnt vmcnt(0)
; GFX908-NEXT: buffer_wbinvl1
-; GFX908-NEXT: v_cmp_eq_u32_e32 vcc, v5, v1
+; GFX908-NEXT: v_cmp_eq_u32_e32 vcc, v1, v6
; GFX908-NEXT: s_or_b64 s[6:7], vcc, s[6:7]
-; GFX908-NEXT: v_mov_b32_e32 v1, v5
; GFX908-NEXT: s_andn2_b64 exec, exec, s[6:7]
; GFX908-NEXT: s_cbranch_execnz .LBB20_1
; GFX908-NEXT: ; %bb.2: ; %atomicrmw.end
@@ -8297,38 +8325,38 @@ define void @buffer_fat_ptr_agent_atomic_fmax_noret_v2bf16__offset__amdgpu_no_fi
; GFX8-NEXT: buffer_load_dword v1, v1, s[16:19], 0 offen offset:1024
; GFX8-NEXT: s_add_i32 s4, s20, 0x400
; GFX8-NEXT: s_mov_b64 s[6:7], 0
-; GFX8-NEXT: v_lshlrev_b32_e32 v2, 16, v0
-; GFX8-NEXT: v_and_b32_e32 v3, 0xffff0000, v0
+; GFX8-NEXT: v_lshlrev_b32_e32 v3, 16, v0
+; GFX8-NEXT: v_and_b32_e32 v0, 0xffff0000, v0
; GFX8-NEXT: v_mov_b32_e32 v4, s4
; GFX8-NEXT: .LBB20_1: ; %atomicrmw.start
; GFX8-NEXT: ; =>This Inner Loop Header: Depth=1
; GFX8-NEXT: s_waitcnt vmcnt(0)
-; GFX8-NEXT: v_lshlrev_b32_e32 v0, 16, v1
-; GFX8-NEXT: v_and_b32_e32 v5, 0xffff0000, v1
-; GFX8-NEXT: v_max_f32_e32 v0, v0, v2
-; GFX8-NEXT: v_max_f32_e32 v5, v5, v3
-; GFX8-NEXT: v_bfe_u32 v6, v0, 16, 1
-; GFX8-NEXT: v_bfe_u32 v8, v5, 16, 1
-; GFX8-NEXT: v_add_u32_e32 v6, vcc, v6, v0
-; GFX8-NEXT: v_add_u32_e32 v8, vcc, v8, v5
-; GFX8-NEXT: v_add_u32_e32 v6, vcc, 0x7fff, v6
-; GFX8-NEXT: v_add_u32_e32 v8, vcc, 0x7fff, v8
-; GFX8-NEXT: v_or_b32_e32 v9, 0x400000, v5
-; GFX8-NEXT: v_cmp_u_f32_e32 vcc, v5, v5
-; GFX8-NEXT: v_or_b32_e32 v7, 0x400000, v0
-; GFX8-NEXT: v_cmp_u_f32_e64 s[4:5], v0, v0
-; GFX8-NEXT: v_cndmask_b32_e32 v5, v8, v9, vcc
-; GFX8-NEXT: v_cndmask_b32_e64 v0, v6, v7, s[4:5]
-; GFX8-NEXT: v_lshrrev_b32_e32 v5, 16, v5
-; GFX8-NEXT: v_alignbit_b32 v0, v5, v0, 16
; GFX8-NEXT: v_mov_b32_e32 v6, v1
-; GFX8-NEXT: v_mov_b32_e32 v5, v0
-; GFX8-NEXT: buffer_atomic_cmpswap v[5:6], v4, s[16:19], 0 offen glc
+; GFX8-NEXT: v_lshlrev_b32_e32 v1, 16, v6
+; GFX8-NEXT: v_and_b32_e32 v2, 0xffff0000, v6
+; GFX8-NEXT: v_max_f32_e32 v1, v1, v3
+; GFX8-NEXT: v_max_f32_e32 v2, v2, v0
+; GFX8-NEXT: v_bfe_u32 v5, v1, 16, 1
+; GFX8-NEXT: v_bfe_u32 v8, v2, 16, 1
+; GFX8-NEXT: v_add_u32_e32 v5, vcc, v5, v1
+; GFX8-NEXT: v_add_u32_e32 v8, vcc, v8, v2
+; GFX8-NEXT: v_add_u32_e32 v5, vcc, 0x7fff, v5
+; GFX8-NEXT: v_add_u32_e32 v8, vcc, 0x7fff, v8
+; GFX8-NEXT: v_or_b32_e32 v9, 0x400000, v2
+; GFX8-NEXT: v_cmp_u_f32_e32 vcc, v2, v2
+; GFX8-NEXT: v_or_b32_e32 v7, 0x400000, v1
+; GFX8-NEXT: v_cmp_u_f32_e64 s[4:5], v1, v1
+; GFX8-NEXT: v_cndmask_b32_e32 v2, v8, v9, vcc
+; GFX8-NEXT: v_cndmask_b32_e64 v1, v5, v7, s[4:5]
+; GFX8-NEXT: v_lshrrev_b32_e32 v2, 16, v2
+; GFX8-NEXT: v_alignbit_b32 v5, v2, v1, 16
+; GFX8-NEXT: v_mov_b32_e32 v1, v5
+; GFX8-NEXT: v_mov_b32_e32 v2, v6
+; GFX8-NEXT: buffer_atomic_cmpswap v[1:2], v4, s[16:19], 0 offen glc
; GFX8-NEXT: s_waitcnt vmcnt(0)
; GFX8-NEXT: buffer_wbinvl1
-; GFX8-NEXT: v_cmp_eq_u32_e32 vcc, v5, v1
+; GFX8-NEXT: v_cmp_eq_u32_e32 vcc, v1, v6
; GFX8-NEXT: s_or_b64 s[6:7], vcc, s[6:7]
-; GFX8-NEXT: v_mov_b32_e32 v1, v5
; GFX8-NEXT: s_andn2_b64 exec, exec, s[6:7]
; GFX8-NEXT: s_cbranch_execnz .LBB20_1
; GFX8-NEXT: ; %bb.2: ; %atomicrmw.end
diff --git a/llvm/test/CodeGen/AMDGPU/buffer-fat-pointer-atomicrmw-fmin.ll b/llvm/test/CodeGen/AMDGPU/buffer-fat-pointer-atomicrmw-fmin.ll
index 56719dccbd08a..c0c4ddc67cd3e 100644
--- a/llvm/test/CodeGen/AMDGPU/buffer-fat-pointer-atomicrmw-fmin.ll
+++ b/llvm/test/CodeGen/AMDGPU/buffer-fat-pointer-atomicrmw-fmin.ll
@@ -211,24 +211,24 @@ define void @buffer_fat_ptr_agent_atomic_fmin_noret_f32__offset__amdgpu_no_fine_
; GFX942: ; %bb.0:
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX942-NEXT: v_mov_b32_e32 v1, s16
-; GFX942-NEXT: buffer_load_dword v1, v1, s[0:3], 0 offen offset:1024
+; GFX942-NEXT: buffer_load_dword v2, v1, s[0:3], 0 offen offset:1024
; GFX942-NEXT: s_add_i32 s6, s16, 0x400
; GFX942-NEXT: s_mov_b64 s[4:5], 0
-; GFX942-NEXT: v_max_f32_e32 v2, v0, v0
-; GFX942-NEXT: v_mov_b32_e32 v3, s6
+; GFX942-NEXT: v_max_f32_e32 v0, v0, v0
+; GFX942-NEXT: v_mov_b32_e32 v1, s6
; GFX942-NEXT: .LBB1_1: ; %atomicrmw.start
; GFX942-NEXT: ; =>This Inner Loop Header: Depth=1
; GFX942-NEXT: s_waitcnt vmcnt(0)
-; GFX942-NEXT: v_max_f32_e32 v0, v1, v1
-; GFX942-NEXT: v_min_f32_e32 v0, v0, v2
-; GFX942-NEXT: v_mov_b64_e32 v[4:5], v[0:1]
+; GFX942-NEXT: v_mov_b32_e32 v5, v2
+; GFX942-NEXT: v_max_f32_e32 v2, v5, v5
+; GFX942-NEXT: v_min_f32_e32 v4, v2, v0
+; GFX942-NEXT: v_mov_b64_e32 v[2:3], v[4:5]
; GFX942-NEXT: buffer_wbl2 sc1
-; GFX942-NEXT: buffer_atomic_cmpswap v[4:5], v3, s[0:3], 0 offen sc0
+; GFX942-NEXT: buffer_atomic_cmpswap v[2:3], v1, s[0:3], 0 offen sc0
; GFX942-NEXT: s_waitcnt vmcnt(0)
; GFX942-NEXT: buffer_inv sc1
-; GFX942-NEXT: v_cmp_eq_u32_e32 vcc, v4, v1
+; GFX942-NEXT: v_cmp_eq_u32_e32 vcc, v2, v5
; GFX942-NEXT: s_or_b64 s[4:5], vcc, s[4:5]
-; GFX942-NEXT: v_mov_b32_e32 v1, v4
; GFX942-NEXT: s_andn2_b64 exec, exec, s[4:5]
; GFX942-NEXT: s_cbranch_execnz .LBB1_1
; GFX942-NEXT: ; %bb.2: ; %atomicrmw.end
@@ -261,23 +261,23 @@ define void @buffer_fat_ptr_agent_atomic_fmin_noret_f32__offset__amdgpu_no_fine_
; GFX90A: ; %bb.0:
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX90A-NEXT: v_mov_b32_e32 v1, s20
-; GFX90A-NEXT: buffer_load_dword v1, v1, s[16:19], 0 offen offset:1024
+; GFX90A-NEXT: buffer_load_dword v2, v1, s[16:19], 0 offen offset:1024
; GFX90A-NEXT: s_add_i32 s6, s20, 0x400
; GFX90A-NEXT: s_mov_b64 s[4:5], 0
-; GFX90A-NEXT: v_max_f32_e32 v2, v0, v0
-; GFX90A-NEXT: v_mov_b32_e32 v3, s6
+; GFX90A-NEXT: v_max_f32_e32 v0, v0, v0
+; GFX90A-NEXT: v_mov_b32_e32 v1, s6
; GFX90A-NEXT: .LBB1_1: ; %atomicrmw.start
; GFX90A-NEXT: ; =>This Inner Loop Header: Depth=1
; GFX90A-NEXT: s_waitcnt vmcnt(0)
-; GFX90A-NEXT: v_max_f32_e32 v0, v1, v1
-; GFX90A-NEXT: v_min_f32_e32 v0, v0, v2
-; GFX90A-NEXT: v_pk_mov_b32 v[4:5], v[0:1], v[0:1] op_sel:[0,1]
-; GFX90A-NEXT: buffer_atomic_cmpswap v[4:5], v3, s[16:19], 0 offen glc
+; GFX90A-NEXT: v_mov_b32_e32 v5, v2
+; GFX90A-NEXT: v_max_f32_e32 v2, v5, v5
+; GFX90A-NEXT: v_min_f32_e32 v4, v2, v0
+; GFX90A-NEXT: v_pk_mov_b32 v[2:3], v[4:5], v[4:5] op_sel:[0,1]
+; GFX90A-NEXT: buffer_atomic_cmpswap v[2:3], v1, s[16:19], 0 offen glc
; GFX90A-NEXT: s_waitcnt vmcnt(0)
; GFX90A-NEXT: buffer_wbinvl1
-; GFX90A-NEXT: v_cmp_eq_u32_e32 vcc, v4, v1
+; GFX90A-NEXT: v_cmp_eq_u32_e32 vcc, v2, v5
; GFX90A-NEXT: s_or_b64 s[4:5], vcc, s[4:5]
-; GFX90A-NEXT: v_mov_b32_e32 v1, v4
; GFX90A-NEXT: s_andn2_b64 exec, exec, s[4:5]
; GFX90A-NEXT: s_cbranch_execnz .LBB1_1
; GFX90A-NEXT: ; %bb.2: ; %atomicrmw.end
@@ -291,21 +291,21 @@ define void @buffer_fat_ptr_agent_atomic_fmin_noret_f32__offset__amdgpu_no_fine_
; GFX908-NEXT: buffer_load_dword v1, v1, s[16:19], 0 offen offset:1024
; GFX908-NEXT: s_add_i32 s6, s20, 0x400
; GFX908-NEXT: s_mov_b64 s[4:5], 0
-; GFX908-NEXT: v_max_f32_e32 v2, v0, v0
+; GFX908-NEXT: v_max_f32_e32 v0, v0, v0
; GFX908-NEXT: v_mov_b32_e32 v3, s6
; GFX908-NEXT: .LBB1_1: ; %atomicrmw.start
; GFX908-NEXT: ; =>This Inner Loop Header: Depth=1
; GFX908-NEXT: s_waitcnt vmcnt(0)
-; GFX908-NEXT: v_max_f32_e32 v0, v1, v1
-; GFX908-NEXT: v_min_f32_e32 v0, v0, v2
; GFX908-NEXT: v_mov_b32_e32 v5, v1
-; GFX908-NEXT: v_mov_b32_e32 v4, v0
-; GFX908-NEXT: buffer_atomic_cmpswap v[4:5], v3, s[16:19], 0 offen glc
+; GFX908-NEXT: v_max_f32_e32 v1, v5, v5
+; GFX908-NEXT: v_min_f32_e32 v4, v1, v0
+; GFX908-NEXT: v_mov_b32_e32 v1, v4
+; GFX908-NEXT: v_mov_b32_e32 v2, v5
+; GFX908-NEXT: buffer_atomic_cmpswap v[1:2], v3, s[16:19], 0 offen glc
; GFX908-NEXT: s_waitcnt vmcnt(0)
; GFX908-NEXT: buffer_wbinvl1
-; GFX908-NEXT: v_cmp_eq_u32_e32 vcc, v4, v1
+; GFX908-NEXT: v_cmp_eq_u32_e32 vcc, v1, v5
; GFX908-NEXT: s_or_b64 s[4:5], vcc, s[4:5]
-; GFX908-NEXT: v_mov_b32_e32 v1, v4
; GFX908-NEXT: s_andn2_b64 exec, exec, s[4:5]
; GFX908-NEXT: s_cbranch_execnz .LBB1_1
; GFX908-NEXT: ; %bb.2: ; %atomicrmw.end
@@ -319,21 +319,21 @@ define void @buffer_fat_ptr_agent_atomic_fmin_noret_f32__offset__amdgpu_no_fine_
; GFX8-NEXT: buffer_load_dword v1, v1, s[16:19], 0 offen offset:1024
; GFX8-NEXT: s_add_i32 s6, s20, 0x400
; GFX8-NEXT: s_mov_b64 s[4:5], 0
-; GFX8-NEXT: v_mul_f32_e32 v2, 1.0, v0
+; GFX8-NEXT: v_mul_f32_e32 v0, 1.0, v0
; GFX8-NEXT: v_mov_b32_e32 v3, s6
; GFX8-NEXT: .LBB1_1: ; %atomicrmw.start
; GFX8-NEXT: ; =>This Inner Loop Header: Depth=1
; GFX8-NEXT: s_waitcnt vmcnt(0)
-; GFX8-NEXT: v_mul_f32_e32 v0, 1.0, v1
-; GFX8-NEXT: v_min_f32_e32 v0, v0, v2
; GFX8-NEXT: v_mov_b32_e32 v5, v1
-; GFX8-NEXT: v_mov_b32_e32 v4, v0
-; GFX8-NEXT: buffer_atomic_cmpswap v[4:5], v3, s[16:19], 0 offen glc
+; GFX8-NEXT: v_mul_f32_e32 v1, 1.0, v5
+; GFX8-NEXT: v_min_f32_e32 v4, v1, v0
+; GFX8-NEXT: v_mov_b32_e32 v1, v4
+; GFX8-NEXT: v_mov_b32_e32 v2, v5
+; GFX8-NEXT: buffer_atomic_cmpswap v[1:2], v3, s[16:19], 0 offen glc
; GFX8-NEXT: s_waitcnt vmcnt(0)
; GFX8-NEXT: buffer_wbinvl1
-; GFX8-NEXT: v_cmp_eq_u32_e32 vcc, v4, v1
+; GFX8-NEXT: v_cmp_eq_u32_e32 vcc, v1, v5
; GFX8-NEXT: s_or_b64 s[4:5], vcc, s[4:5]
-; GFX8-NEXT: v_mov_b32_e32 v1, v4
; GFX8-NEXT: s_andn2_b64 exec, exec, s[4:5]
; GFX8-NEXT: s_cbranch_execnz .LBB1_1
; GFX8-NEXT: ; %bb.2: ; %atomicrmw.end
@@ -1395,7 +1395,7 @@ define void @buffer_fat_ptr_agent_atomic_fmin_noret_f64__offset__amdgpu_no_fine_
; GFX12-NEXT: s_wait_bvhcnt 0x0
; GFX12-NEXT: s_wait_kmcnt 0x0
; GFX12-NEXT: v_mov_b32_e32 v2, s16
-; GFX12-NEXT: v_max_num_f64_e32 v[4:5], v[0:1], v[0:1]
+; GFX12-NEXT: v_max_num_f64_e32 v[0:1], v[0:1], v[0:1]
; GFX12-NEXT: s_add_co_i32 s4, s16, 0x800
; GFX12-NEXT: s_wait_alu 0xfffe
; GFX12-NEXT: v_mov_b32_e32 v6, s4
@@ -1404,17 +1404,18 @@ define void @buffer_fat_ptr_agent_atomic_fmin_noret_f64__offset__amdgpu_no_fine_
; GFX12-NEXT: .LBB6_1: ; %atomicrmw.start
; GFX12-NEXT: ; =>This Inner Loop Header: Depth=1
; GFX12-NEXT: s_wait_loadcnt 0x0
-; GFX12-NEXT: v_max_num_f64_e32 v[0:1], v[2:3], v[2:3]
-; GFX12-NEXT: s_wait_storecnt 0x0
-; GFX12-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(SKIP_1) | instid1(VALU_DEP_2)
-; GFX12-NEXT: v_min_num_f64_e32 v[0:1], v[0:1], v[4:5]
; GFX12-NEXT: v_dual_mov_b32 v10, v3 :: v_dual_mov_b32 v9, v2
-; GFX12-NEXT: v_dual_mov_b32 v8, v1 :: v_dual_mov_b32 v7, v0
-; GFX12-NEXT: buffer_atomic_cmpswap_b64 v[7:10], v6, s[0:3], null offen th:TH_ATOMIC_RETURN
+; GFX12-NEXT: s_wait_storecnt 0x0
+; GFX12-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
+; GFX12-NEXT: v_max_num_f64_e32 v[2:3], v[9:10], v[9:10]
+; GFX12-NEXT: v_min_num_f64_e32 v[7:8], v[2:3], v[0:1]
+; GFX12-NEXT: s_delay_alu instid0(VALU_DEP_1)
+; GFX12-NEXT: v_dual_mov_b32 v2, v7 :: v_dual_mov_b32 v3, v8
+; GFX12-NEXT: v_dual_mov_b32 v4, v9 :: v_dual_mov_b32 v5, v10
+; GFX12-NEXT: buffer_atomic_cmpswap_b64 v[2:5], v6, s[0:3], null offen th:TH_ATOMIC_RETURN
; GFX12-NEXT: s_wait_loadcnt 0x0
; GFX12-NEXT: global_inv scope:SCOPE_DEV
-; GFX12-NEXT: v_cmp_eq_u64_e32 vcc_lo, v[7:8], v[2:3]
-; GFX12-NEXT: v_dual_mov_b32 v2, v7 :: v_dual_mov_b32 v3, v8
+; GFX12-NEXT: v_cmp_eq_u64_e32 vcc_lo, v[2:3], v[9:10]
; GFX12-NEXT: s_or_b32 s4, vcc_lo, s4
; GFX12-NEXT: s_wait_alu 0xfffe
; GFX12-NEXT: s_and_not1_b32 exec_lo, exec_lo, s4
@@ -1438,7 +1439,7 @@ define void @buffer_fat_ptr_agent_atomic_fmin_noret_f64__offset__amdgpu_no_fine_
; GFX11: ; %bb.0:
; GFX11-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX11-NEXT: v_mov_b32_e32 v2, s16
-; GFX11-NEXT: v_max_f64 v[4:5], v[0:1], v[0:1]
+; GFX11-NEXT: v_max_f64 v[0:1], v[0:1], v[0:1]
; GFX11-NEXT: s_add_i32 s4, s16, 0x800
; GFX11-NEXT: s_delay_alu instid0(SALU_CYCLE_1)
; GFX11-NEXT: v_mov_b32_e32 v6, s4
@@ -1447,18 +1448,19 @@ define void @buffer_fat_ptr_agent_atomic_fmin_noret_f64__offset__amdgpu_no_fine_
; GFX11-NEXT: .LBB6_1: ; %atomicrmw.start
; GFX11-NEXT: ; =>This Inner Loop Header: Depth=1
; GFX11-NEXT: s_waitcnt vmcnt(0)
-; GFX11-NEXT: v_max_f64 v[0:1], v[2:3], v[2:3]
-; GFX11-NEXT: s_waitcnt_vscnt null, 0x0
-; GFX11-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(SKIP_1) | instid1(VALU_DEP_2)
-; GFX11-NEXT: v_min_f64 v[0:1], v[0:1], v[4:5]
; GFX11-NEXT: v_dual_mov_b32 v10, v3 :: v_dual_mov_b32 v9, v2
-; GFX11-NEXT: v_dual_mov_b32 v8, v1 :: v_dual_mov_b32 v7, v0
-; GFX11-NEXT: buffer_atomic_cmpswap_b64 v[7:10], v6, s[0:3], 0 offen glc
+; GFX11-NEXT: s_waitcnt_vscnt null, 0x0
+; GFX11-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
+; GFX11-NEXT: v_max_f64 v[2:3], v[9:10], v[9:10]
+; GFX11-NEXT: v_min_f64 v[7:8], v[2:3], v[0:1]
+; GFX11-NEXT: s_delay_alu instid0(VALU_DEP_1)
+; GFX11-NEXT: v_dual_mov_b32 v2, v7 :: v_dual_mov_b32 v3, v8
+; GFX11-NEXT: v_dual_mov_b32 v4, v9 :: v_dual_mov_b32 v5, v10
+; GFX11-NEXT: buffer_atomic_cmpswap_b64 v[2:5], v6, s[0:3], 0 offen glc
; GFX11-NEXT: s_waitcnt vmcnt(0)
; GFX11-NEXT: buffer_gl1_inv
; GFX11-NEXT: buffer_gl0_inv
-; GFX11-NEXT: v_cmp_eq_u64_e32 vcc_lo, v[7:8], v[2:3]
-; GFX11-NEXT: v_dual_mov_b32 v2, v7 :: v_dual_mov_b32 v3, v8
+; GFX11-NEXT: v_cmp_eq_u64_e32 vcc_lo, v[2:3], v[9:10]
; GFX11-NEXT: s_or_b32 s4, vcc_lo, s4
; GFX11-NEXT: s_delay_alu instid0(SALU_CYCLE_1)
; GFX11-NEXT: s_and_not1_b32 exec_lo, exec_lo, s4
@@ -1492,26 +1494,26 @@ define void @buffer_fat_ptr_agent_atomic_fmin_noret_f64__offset__amdgpu_no_fine_
; GFX908-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX908-NEXT: v_mov_b32_e32 v2, s20
; GFX908-NEXT: buffer_load_dwordx2 v[2:3], v2, s[16:19], 0 offen offset:2048
-; GFX908-NEXT: v_max_f64 v[4:5], v[0:1], v[0:1]
+; GFX908-NEXT: v_max_f64 v[0:1], v[0:1], v[0:1]
; GFX908-NEXT: s_add_i32 s6, s20, 0x800
; GFX908-NEXT: s_mov_b64 s[4:5], 0
; GFX908-NEXT: v_mov_b32_e32 v6, s6
; GFX908-NEXT: .LBB6_1: ; %atomicrmw.start
; GFX908-NEXT: ; =>This Inner Loop Header: Depth=1
; GFX908-NEXT: s_waitcnt vmcnt(0)
-; GFX908-NEXT: v_max_f64 v[0:1], v[2:3], v[2:3]
-; GFX908-NEXT: v_min_f64 v[0:1], v[0:1], v[4:5]
; GFX908-NEXT: v_mov_b32_e32 v10, v3
; GFX908-NEXT: v_mov_b32_e32 v9, v2
-; GFX908-NEXT: v_mov_b32_e32 v8, v1
-; GFX908-NEXT: v_mov_b32_e32 v7, v0
-; GFX908-NEXT: buffer_atomic_cmpswap_x2 v[7:10], v6, s[16:19], 0 offen glc
+; GFX908-NEXT: v_max_f64 v[2:3], v[9:10], v[9:10]
+; GFX908-NEXT: v_min_f64 v[7:8], v[2:3], v[0:1]
+; GFX908-NEXT: v_mov_b32_e32 v2, v7
+; GFX908-NEXT: v_mov_b32_e32 v3, v8
+; GFX908-NEXT: v_mov_b32_e32 v4, v9
+; GFX908-NEXT: v_mov_b32_e32 v5, v10
+; GFX908-NEXT: buffer_atomic_cmpswap_x2 v[2:5], v6, s[16:19], 0 offen glc
; GFX908-NEXT: s_waitcnt vmcnt(0)
; GFX908-NEXT: buffer_wbinvl1
-; GFX908-NEXT: v_cmp_eq_u64_e32 vcc, v[7:8], v[2:3]
-; GFX908-NEXT: v_mov_b32_e32 v2, v7
+; GFX908-NEXT: v_cmp_eq_u64_e32 vcc, v[2:3], v[9:10]
; GFX908-NEXT: s_or_b64 s[4:5], vcc, s[4:5]
-; GFX908-NEXT: v_mov_b32_e32 v3, v8
; GFX908-NEXT: s_andn2_b64 exec, exec, s[4:5]
; GFX908-NEXT: s_cbranch_execnz .LBB6_1
; GFX908-NEXT: ; %bb.2: ; %atomicrmw.end
@@ -1523,26 +1525,26 @@ define void @buffer_fat_ptr_agent_atomic_fmin_noret_f64__offset__amdgpu_no_fine_
; GFX8-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX8-NEXT: v_mov_b32_e32 v2, s20
; GFX8-NEXT: buffer_load_dwordx2 v[2:3], v2, s[16:19], 0 offen offset:2048
-; GFX8-NEXT: v_max_f64 v[4:5], v[0:1], v[0:1]
+; GFX8-NEXT: v_max_f64 v[0:1], v[0:1], v[0:1]
; GFX8-NEXT: s_add_i32 s6, s20, 0x800
; GFX8-NEXT: s_mov_b64 s[4:5], 0
; GFX8-NEXT: v_mov_b32_e32 v6, s6
; GFX8-NEXT: .LBB6_1: ; %atomicrmw.start
; GFX8-NEXT: ; =>This Inner Loop Header: Depth=1
; GFX8-NEXT: s_waitcnt vmcnt(0)
-; GFX8-NEXT: v_max_f64 v[0:1], v[2:3], v[2:3]
-; GFX8-NEXT: v_min_f64 v[0:1], v[0:1], v[4:5]
; GFX8-NEXT: v_mov_b32_e32 v10, v3
; GFX8-NEXT: v_mov_b32_e32 v9, v2
-; GFX8-NEXT: v_mov_b32_e32 v8, v1
-; GFX8-NEXT: v_mov_b32_e32 v7, v0
-; GFX8-NEXT: buffer_atomic_cmpswap_x2 v[7:10], v6, s[16:19], 0 offen glc
+; GFX8-NEXT: v_max_f64 v[2:3], v[9:10], v[9:10]
+; GFX8-NEXT: v_min_f64 v[7:8], v[2:3], v[0:1]
+; GFX8-NEXT: v_mov_b32_e32 v2, v7
+; GFX8-NEXT: v_mov_b32_e32 v3, v8
+; GFX8-NEXT: v_mov_b32_e32 v4, v9
+; GFX8-NEXT: v_mov_b32_e32 v5, v10
+; GFX8-NEXT: buffer_atomic_cmpswap_x2 v[2:5], v6, s[16:19], 0 offen glc
; GFX8-NEXT: s_waitcnt vmcnt(0)
; GFX8-NEXT: buffer_wbinvl1
-; GFX8-NEXT: v_cmp_eq_u64_e32 vcc, v[7:8], v[2:3]
-; GFX8-NEXT: v_mov_b32_e32 v2, v7
+; GFX8-NEXT: v_cmp_eq_u64_e32 vcc, v[2:3], v[9:10]
; GFX8-NEXT: s_or_b64 s[4:5], vcc, s[4:5]
-; GFX8-NEXT: v_mov_b32_e32 v3, v8
; GFX8-NEXT: s_andn2_b64 exec, exec, s[4:5]
; GFX8-NEXT: s_cbranch_execnz .LBB6_1
; GFX8-NEXT: ; %bb.2: ; %atomicrmw.end
@@ -2496,42 +2498,43 @@ define half @buffer_fat_ptr_agent_atomic_fmin_ret_f16__offset__amdgpu_no_fine_gr
; GFX12-TRUE16-NEXT: s_wait_alu 0xfffe
; GFX12-TRUE16-NEXT: s_and_b32 s4, s16, -4
; GFX12-TRUE16-NEXT: s_wait_alu 0xfffe
-; GFX12-TRUE16-NEXT: v_mov_b32_e32 v5, s4
+; GFX12-TRUE16-NEXT: v_mov_b32_e32 v3, s4
; GFX12-TRUE16-NEXT: s_and_b32 s4, s16, 3
; GFX12-TRUE16-NEXT: s_wait_alu 0xfffe
; GFX12-TRUE16-NEXT: s_lshl_b32 s4, s4, 3
; GFX12-TRUE16-NEXT: s_wait_alu 0xfffe
; GFX12-TRUE16-NEXT: s_lshl_b32 s5, 0xffff, s4
-; GFX12-TRUE16-NEXT: buffer_load_b32 v2, v5, s[0:3], null offen
+; GFX12-TRUE16-NEXT: buffer_load_b32 v1, v3, s[0:3], null offen
; GFX12-TRUE16-NEXT: s_not_b32 s6, s5
; GFX12-TRUE16-NEXT: s_mov_b32 s5, 0
; GFX12-TRUE16-NEXT: .LBB10_1: ; %atomicrmw.start
; GFX12-TRUE16-NEXT: ; =>This Inner Loop Header: Depth=1
; GFX12-TRUE16-NEXT: s_wait_loadcnt 0x0
-; GFX12-TRUE16-NEXT: v_lshrrev_b32_e32 v1, s4, v2
+; GFX12-TRUE16-NEXT: v_mov_b32_e32 v5, v1
; GFX12-TRUE16-NEXT: s_wait_storecnt 0x0
; GFX12-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
+; GFX12-TRUE16-NEXT: v_lshrrev_b32_e32 v1, s4, v5
; GFX12-TRUE16-NEXT: v_max_num_f16_e32 v0.h, v1.l, v1.l
-; GFX12-TRUE16-NEXT: v_min_num_f16_e32 v1.l, v0.h, v0.l
; GFX12-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
+; GFX12-TRUE16-NEXT: v_min_num_f16_e32 v1.l, v0.h, v0.l
; GFX12-TRUE16-NEXT: v_and_b32_e32 v1, 0xffff, v1
+; GFX12-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(SKIP_1) | instid1(VALU_DEP_1)
; GFX12-TRUE16-NEXT: v_lshlrev_b32_e32 v1, s4, v1
; GFX12-TRUE16-NEXT: s_wait_alu 0xfffe
-; GFX12-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
-; GFX12-TRUE16-NEXT: v_and_or_b32 v1, v2, s6, v1
-; GFX12-TRUE16-NEXT: v_dual_mov_b32 v4, v2 :: v_dual_mov_b32 v3, v1
-; GFX12-TRUE16-NEXT: buffer_atomic_cmpswap_b32 v[3:4], v5, s[0:3], null offen th:TH_ATOMIC_RETURN
+; GFX12-TRUE16-NEXT: v_and_or_b32 v4, v5, s6, v1
+; GFX12-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1)
+; GFX12-TRUE16-NEXT: v_dual_mov_b32 v1, v4 :: v_dual_mov_b32 v2, v5
+; GFX12-TRUE16-NEXT: buffer_atomic_cmpswap_b32 v[1:2], v3, s[0:3], null offen th:TH_ATOMIC_RETURN
; GFX12-TRUE16-NEXT: s_wait_loadcnt 0x0
; GFX12-TRUE16-NEXT: global_inv scope:SCOPE_DEV
-; GFX12-TRUE16-NEXT: v_cmp_eq_u32_e32 vcc_lo, v3, v2
-; GFX12-TRUE16-NEXT: v_mov_b32_e32 v2, v3
+; GFX12-TRUE16-NEXT: v_cmp_eq_u32_e32 vcc_lo, v1, v5
; GFX12-TRUE16-NEXT: s_or_b32 s5, vcc_lo, s5
; GFX12-TRUE16-NEXT: s_wait_alu 0xfffe
; GFX12-TRUE16-NEXT: s_and_not1_b32 exec_lo, exec_lo, s5
; GFX12-TRUE16-NEXT: s_cbranch_execnz .LBB10_1
; GFX12-TRUE16-NEXT: ; %bb.2: ; %atomicrmw.end
; GFX12-TRUE16-NEXT: s_or_b32 exec_lo, exec_lo, s5
-; GFX12-TRUE16-NEXT: v_lshrrev_b32_e32 v0, s4, v3
+; GFX12-TRUE16-NEXT: v_lshrrev_b32_e32 v0, s4, v1
; GFX12-TRUE16-NEXT: s_wait_loadcnt 0x0
; GFX12-TRUE16-NEXT: s_setpc_b64 s[30:31]
;
@@ -2543,46 +2546,47 @@ define half @buffer_fat_ptr_agent_atomic_fmin_ret_f16__offset__amdgpu_no_fine_gr
; GFX12-FAKE16-NEXT: s_wait_bvhcnt 0x0
; GFX12-FAKE16-NEXT: s_wait_kmcnt 0x0
; GFX12-FAKE16-NEXT: s_addk_co_i32 s16, 0x200
-; GFX12-FAKE16-NEXT: v_max_num_f16_e32 v5, v0, v0
+; GFX12-FAKE16-NEXT: v_max_num_f16_e32 v0, v0, v0
; GFX12-FAKE16-NEXT: s_wait_alu 0xfffe
; GFX12-FAKE16-NEXT: s_and_b32 s4, s16, -4
; GFX12-FAKE16-NEXT: s_wait_alu 0xfffe
-; GFX12-FAKE16-NEXT: v_mov_b32_e32 v4, s4
+; GFX12-FAKE16-NEXT: v_mov_b32_e32 v3, s4
; GFX12-FAKE16-NEXT: s_and_b32 s4, s16, 3
; GFX12-FAKE16-NEXT: s_wait_alu 0xfffe
; GFX12-FAKE16-NEXT: s_lshl_b32 s4, s4, 3
; GFX12-FAKE16-NEXT: s_wait_alu 0xfffe
; GFX12-FAKE16-NEXT: s_lshl_b32 s5, 0xffff, s4
-; GFX12-FAKE16-NEXT: buffer_load_b32 v1, v4, s[0:3], null offen
+; GFX12-FAKE16-NEXT: buffer_load_b32 v1, v3, s[0:3], null offen
; GFX12-FAKE16-NEXT: s_not_b32 s6, s5
; GFX12-FAKE16-NEXT: s_mov_b32 s5, 0
; GFX12-FAKE16-NEXT: .LBB10_1: ; %atomicrmw.start
; GFX12-FAKE16-NEXT: ; =>This Inner Loop Header: Depth=1
; GFX12-FAKE16-NEXT: s_wait_loadcnt 0x0
-; GFX12-FAKE16-NEXT: v_lshrrev_b32_e32 v0, s4, v1
+; GFX12-FAKE16-NEXT: v_mov_b32_e32 v5, v1
; GFX12-FAKE16-NEXT: s_wait_storecnt 0x0
; GFX12-FAKE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
-; GFX12-FAKE16-NEXT: v_max_num_f16_e32 v0, v0, v0
-; GFX12-FAKE16-NEXT: v_min_num_f16_e32 v0, v0, v5
+; GFX12-FAKE16-NEXT: v_lshrrev_b32_e32 v1, s4, v5
+; GFX12-FAKE16-NEXT: v_max_num_f16_e32 v1, v1, v1
; GFX12-FAKE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
-; GFX12-FAKE16-NEXT: v_and_b32_e32 v0, 0xffff, v0
-; GFX12-FAKE16-NEXT: v_lshlrev_b32_e32 v0, s4, v0
+; GFX12-FAKE16-NEXT: v_min_num_f16_e32 v1, v1, v0
+; GFX12-FAKE16-NEXT: v_and_b32_e32 v1, 0xffff, v1
+; GFX12-FAKE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(SKIP_1) | instid1(VALU_DEP_1)
+; GFX12-FAKE16-NEXT: v_lshlrev_b32_e32 v1, s4, v1
; GFX12-FAKE16-NEXT: s_wait_alu 0xfffe
-; GFX12-FAKE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
-; GFX12-FAKE16-NEXT: v_and_or_b32 v0, v1, s6, v0
-; GFX12-FAKE16-NEXT: v_dual_mov_b32 v3, v1 :: v_dual_mov_b32 v2, v0
-; GFX12-FAKE16-NEXT: buffer_atomic_cmpswap_b32 v[2:3], v4, s[0:3], null offen th:TH_ATOMIC_RETURN
+; GFX12-FAKE16-NEXT: v_and_or_b32 v4, v5, s6, v1
+; GFX12-FAKE16-NEXT: s_delay_alu instid0(VALU_DEP_1)
+; GFX12-FAKE16-NEXT: v_dual_mov_b32 v1, v4 :: v_dual_mov_b32 v2, v5
+; GFX12-FAKE16-NEXT: buffer_atomic_cmpswap_b32 v[1:2], v3, s[0:3], null offen th:TH_ATOMIC_RETURN
; GFX12-FAKE16-NEXT: s_wait_loadcnt 0x0
; GFX12-FAKE16-NEXT: global_inv scope:SCOPE_DEV
-; GFX12-FAKE16-NEXT: v_cmp_eq_u32_e32 vcc_lo, v2, v1
-; GFX12-FAKE16-NEXT: v_mov_b32_e32 v1, v2
+; GFX12-FAKE16-NEXT: v_cmp_eq_u32_e32 vcc_lo, v1, v5
; GFX12-FAKE16-NEXT: s_or_b32 s5, vcc_lo, s5
; GFX12-FAKE16-NEXT: s_wait_alu 0xfffe
; GFX12-FAKE16-NEXT: s_and_not1_b32 exec_lo, exec_lo, s5
; GFX12-FAKE16-NEXT: s_cbranch_execnz .LBB10_1
; GFX12-FAKE16-NEXT: ; %bb.2: ; %atomicrmw.end
; GFX12-FAKE16-NEXT: s_or_b32 exec_lo, exec_lo, s5
-; GFX12-FAKE16-NEXT: v_lshrrev_b32_e32 v0, s4, v2
+; GFX12-FAKE16-NEXT: v_lshrrev_b32_e32 v0, s4, v1
; GFX12-FAKE16-NEXT: s_wait_loadcnt 0x0
; GFX12-FAKE16-NEXT: s_setpc_b64 s[30:31]
;
@@ -2591,30 +2595,30 @@ define half @buffer_fat_ptr_agent_atomic_fmin_ret_f16__offset__amdgpu_no_fine_gr
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX942-NEXT: s_addk_i32 s16, 0x200
; GFX942-NEXT: s_and_b32 s4, s16, -4
-; GFX942-NEXT: v_mov_b32_e32 v4, s4
-; GFX942-NEXT: buffer_load_dword v1, v4, s[0:3], 0 offen
+; GFX942-NEXT: v_mov_b32_e32 v1, s4
+; GFX942-NEXT: buffer_load_dword v2, v1, s[0:3], 0 offen
; GFX942-NEXT: s_and_b32 s4, s16, 3
; GFX942-NEXT: s_lshl_b32 s6, s4, 3
; GFX942-NEXT: s_lshl_b32 s4, 0xffff, s6
; GFX942-NEXT: s_not_b32 s7, s4
; GFX942-NEXT: s_mov_b64 s[4:5], 0
-; GFX942-NEXT: v_max_f16_e32 v5, v0, v0
+; GFX942-NEXT: v_max_f16_e32 v0, v0, v0
; GFX942-NEXT: .LBB10_1: ; %atomicrmw.start
; GFX942-NEXT: ; =>This Inner Loop Header: Depth=1
; GFX942-NEXT: s_waitcnt vmcnt(0)
-; GFX942-NEXT: v_lshrrev_b32_e32 v0, s6, v1
-; GFX942-NEXT: v_max_f16_e32 v0, v0, v0
-; GFX942-NEXT: v_min_f16_e32 v0, v0, v5
-; GFX942-NEXT: v_lshlrev_b32_e32 v0, s6, v0
-; GFX942-NEXT: v_and_or_b32 v0, v1, s7, v0
-; GFX942-NEXT: v_mov_b64_e32 v[2:3], v[0:1]
+; GFX942-NEXT: v_mov_b32_e32 v5, v2
+; GFX942-NEXT: v_lshrrev_b32_e32 v2, s6, v5
+; GFX942-NEXT: v_max_f16_e32 v2, v2, v2
+; GFX942-NEXT: v_min_f16_e32 v2, v2, v0
+; GFX942-NEXT: v_lshlrev_b32_e32 v2, s6, v2
+; GFX942-NEXT: v_and_or_b32 v4, v5, s7, v2
+; GFX942-NEXT: v_mov_b64_e32 v[2:3], v[4:5]
; GFX942-NEXT: buffer_wbl2 sc1
-; GFX942-NEXT: buffer_atomic_cmpswap v[2:3], v4, s[0:3], 0 offen sc0
+; GFX942-NEXT: buffer_atomic_cmpswap v[2:3], v1, s[0:3], 0 offen sc0
; GFX942-NEXT: s_waitcnt vmcnt(0)
; GFX942-NEXT: buffer_inv sc1
-; GFX942-NEXT: v_cmp_eq_u32_e32 vcc, v2, v1
+; GFX942-NEXT: v_cmp_eq_u32_e32 vcc, v2, v5
; GFX942-NEXT: s_or_b64 s[4:5], vcc, s[4:5]
-; GFX942-NEXT: v_mov_b32_e32 v1, v2
; GFX942-NEXT: s_andn2_b64 exec, exec, s[4:5]
; GFX942-NEXT: s_cbranch_execnz .LBB10_1
; GFX942-NEXT: ; %bb.2: ; %atomicrmw.end
@@ -2629,125 +2633,127 @@ define half @buffer_fat_ptr_agent_atomic_fmin_ret_f16__offset__amdgpu_no_fine_gr
; GFX11-TRUE16-NEXT: v_max_f16_e32 v0.l, v0.l, v0.l
; GFX11-TRUE16-NEXT: s_and_b32 s4, s16, -4
; GFX11-TRUE16-NEXT: s_delay_alu instid0(SALU_CYCLE_1) | instskip(SKIP_1) | instid1(SALU_CYCLE_1)
-; GFX11-TRUE16-NEXT: v_mov_b32_e32 v5, s4
+; GFX11-TRUE16-NEXT: v_mov_b32_e32 v3, s4
; GFX11-TRUE16-NEXT: s_and_b32 s4, s16, 3
; GFX11-TRUE16-NEXT: s_lshl_b32 s4, s4, 3
; GFX11-TRUE16-NEXT: s_delay_alu instid0(SALU_CYCLE_1)
; GFX11-TRUE16-NEXT: s_lshl_b32 s5, 0xffff, s4
-; GFX11-TRUE16-NEXT: buffer_load_b32 v2, v5, s[0:3], 0 offen
+; GFX11-TRUE16-NEXT: buffer_load_b32 v1, v3, s[0:3], 0 offen
; GFX11-TRUE16-NEXT: s_not_b32 s6, s5
; GFX11-TRUE16-NEXT: s_mov_b32 s5, 0
; GFX11-TRUE16-NEXT: .p2align 6
; GFX11-TRUE16-NEXT: .LBB10_1: ; %atomicrmw.start
; GFX11-TRUE16-NEXT: ; =>This Inner Loop Header: Depth=1
; GFX11-TRUE16-NEXT: s_waitcnt vmcnt(0)
-; GFX11-TRUE16-NEXT: v_lshrrev_b32_e32 v1, s4, v2
+; GFX11-TRUE16-NEXT: v_mov_b32_e32 v5, v1
; GFX11-TRUE16-NEXT: s_waitcnt_vscnt null, 0x0
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
+; GFX11-TRUE16-NEXT: v_lshrrev_b32_e32 v1, s4, v5
; GFX11-TRUE16-NEXT: v_max_f16_e32 v0.h, v1.l, v1.l
-; GFX11-TRUE16-NEXT: v_min_f16_e32 v1.l, v0.h, v0.l
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
+; GFX11-TRUE16-NEXT: v_min_f16_e32 v1.l, v0.h, v0.l
; GFX11-TRUE16-NEXT: v_and_b32_e32 v1, 0xffff, v1
-; GFX11-TRUE16-NEXT: v_lshlrev_b32_e32 v1, s4, v1
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
-; GFX11-TRUE16-NEXT: v_and_or_b32 v1, v2, s6, v1
-; GFX11-TRUE16-NEXT: v_dual_mov_b32 v4, v2 :: v_dual_mov_b32 v3, v1
-; GFX11-TRUE16-NEXT: buffer_atomic_cmpswap_b32 v[3:4], v5, s[0:3], 0 offen glc
+; GFX11-TRUE16-NEXT: v_lshlrev_b32_e32 v1, s4, v1
+; GFX11-TRUE16-NEXT: v_and_or_b32 v4, v5, s6, v1
+; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1)
+; GFX11-TRUE16-NEXT: v_dual_mov_b32 v1, v4 :: v_dual_mov_b32 v2, v5
+; GFX11-TRUE16-NEXT: buffer_atomic_cmpswap_b32 v[1:2], v3, s[0:3], 0 offen glc
; GFX11-TRUE16-NEXT: s_waitcnt vmcnt(0)
; GFX11-TRUE16-NEXT: buffer_gl1_inv
; GFX11-TRUE16-NEXT: buffer_gl0_inv
-; GFX11-TRUE16-NEXT: v_cmp_eq_u32_e32 vcc_lo, v3, v2
-; GFX11-TRUE16-NEXT: v_mov_b32_e32 v2, v3
+; GFX11-TRUE16-NEXT: v_cmp_eq_u32_e32 vcc_lo, v1, v5
; GFX11-TRUE16-NEXT: s_or_b32 s5, vcc_lo, s5
; GFX11-TRUE16-NEXT: s_delay_alu instid0(SALU_CYCLE_1)
; GFX11-TRUE16-NEXT: s_and_not1_b32 exec_lo, exec_lo, s5
; GFX11-TRUE16-NEXT: s_cbranch_execnz .LBB10_1
; GFX11-TRUE16-NEXT: ; %bb.2: ; %atomicrmw.end
; GFX11-TRUE16-NEXT: s_or_b32 exec_lo, exec_lo, s5
-; GFX11-TRUE16-NEXT: v_lshrrev_b32_e32 v0, s4, v3
+; GFX11-TRUE16-NEXT: v_lshrrev_b32_e32 v0, s4, v1
; GFX11-TRUE16-NEXT: s_setpc_b64 s[30:31]
;
; GFX11-FAKE16-LABEL: buffer_fat_ptr_agent_atomic_fmin_ret_f16__offset__amdgpu_no_fine_grained_memory:
; GFX11-FAKE16: ; %bb.0:
; GFX11-FAKE16-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX11-FAKE16-NEXT: s_addk_i32 s16, 0x200
-; GFX11-FAKE16-NEXT: v_max_f16_e32 v5, v0, v0
+; GFX11-FAKE16-NEXT: v_max_f16_e32 v0, v0, v0
; GFX11-FAKE16-NEXT: s_and_b32 s4, s16, -4
; GFX11-FAKE16-NEXT: s_delay_alu instid0(SALU_CYCLE_1) | instskip(SKIP_1) | instid1(SALU_CYCLE_1)
-; GFX11-FAKE16-NEXT: v_mov_b32_e32 v4, s4
+; GFX11-FAKE16-NEXT: v_mov_b32_e32 v3, s4
; GFX11-FAKE16-NEXT: s_and_b32 s4, s16, 3
; GFX11-FAKE16-NEXT: s_lshl_b32 s4, s4, 3
; GFX11-FAKE16-NEXT: s_delay_alu instid0(SALU_CYCLE_1)
; GFX11-FAKE16-NEXT: s_lshl_b32 s5, 0xffff, s4
-; GFX11-FAKE16-NEXT: buffer_load_b32 v1, v4, s[0:3], 0 offen
+; GFX11-FAKE16-NEXT: buffer_load_b32 v1, v3, s[0:3], 0 offen
; GFX11-FAKE16-NEXT: s_not_b32 s6, s5
; GFX11-FAKE16-NEXT: s_mov_b32 s5, 0
; GFX11-FAKE16-NEXT: .p2align 6
; GFX11-FAKE16-NEXT: .LBB10_1: ; %atomicrmw.start
; GFX11-FAKE16-NEXT: ; =>This Inner Loop Header: Depth=1
; GFX11-FAKE16-NEXT: s_waitcnt vmcnt(0)
-; GFX11-FAKE16-NEXT: v_lshrrev_b32_e32 v0, s4, v1
+; GFX11-FAKE16-NEXT: v_mov_b32_e32 v5, v1
; GFX11-FAKE16-NEXT: s_waitcnt_vscnt null, 0x0
; GFX11-FAKE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
-; GFX11-FAKE16-NEXT: v_max_f16_e32 v0, v0, v0
-; GFX11-FAKE16-NEXT: v_min_f16_e32 v0, v0, v5
+; GFX11-FAKE16-NEXT: v_lshrrev_b32_e32 v1, s4, v5
+; GFX11-FAKE16-NEXT: v_max_f16_e32 v1, v1, v1
; GFX11-FAKE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
-; GFX11-FAKE16-NEXT: v_and_b32_e32 v0, 0xffff, v0
-; GFX11-FAKE16-NEXT: v_lshlrev_b32_e32 v0, s4, v0
+; GFX11-FAKE16-NEXT: v_min_f16_e32 v1, v1, v0
+; GFX11-FAKE16-NEXT: v_and_b32_e32 v1, 0xffff, v1
; GFX11-FAKE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
-; GFX11-FAKE16-NEXT: v_and_or_b32 v0, v1, s6, v0
-; GFX11-FAKE16-NEXT: v_dual_mov_b32 v3, v1 :: v_dual_mov_b32 v2, v0
-; GFX11-FAKE16-NEXT: buffer_atomic_cmpswap_b32 v[2:3], v4, s[0:3], 0 offen glc
+; GFX11-FAKE16-NEXT: v_lshlrev_b32_e32 v1, s4, v1
+; GFX11-FAKE16-NEXT: v_and_or_b32 v4, v5, s6, v1
+; GFX11-FAKE16-NEXT: s_delay_alu instid0(VALU_DEP_1)
+; GFX11-FAKE16-NEXT: v_dual_mov_b32 v1, v4 :: v_dual_mov_b32 v2, v5
+; GFX11-FAKE16-NEXT: buffer_atomic_cmpswap_b32 v[1:2], v3, s[0:3], 0 offen glc
; GFX11-FAKE16-NEXT: s_waitcnt vmcnt(0)
; GFX11-FAKE16-NEXT: buffer_gl1_inv
; GFX11-FAKE16-NEXT: buffer_gl0_inv
-; GFX11-FAKE16-NEXT: v_cmp_eq_u32_e32 vcc_lo, v2, v1
-; GFX11-FAKE16-NEXT: v_mov_b32_e32 v1, v2
+; GFX11-FAKE16-NEXT: v_cmp_eq_u32_e32 vcc_lo, v1, v5
; GFX11-FAKE16-NEXT: s_or_b32 s5, vcc_lo, s5
; GFX11-FAKE16-NEXT: s_delay_alu instid0(SALU_CYCLE_1)
; GFX11-FAKE16-NEXT: s_and_not1_b32 exec_lo, exec_lo, s5
; GFX11-FAKE16-NEXT: s_cbranch_execnz .LBB10_1
; GFX11-FAKE16-NEXT: ; %bb.2: ; %atomicrmw.end
; GFX11-FAKE16-NEXT: s_or_b32 exec_lo, exec_lo, s5
-; GFX11-FAKE16-NEXT: v_lshrrev_b32_e32 v0, s4, v2
+; GFX11-FAKE16-NEXT: v_lshrrev_b32_e32 v0, s4, v1
; GFX11-FAKE16-NEXT: s_setpc_b64 s[30:31]
;
; GFX10-LABEL: buffer_fat_ptr_agent_atomic_fmin_ret_f16__offset__amdgpu_no_fine_grained_memory:
; GFX10: ; %bb.0:
; GFX10-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX10-NEXT: s_addk_i32 s20, 0x200
-; GFX10-NEXT: v_max_f16_e32 v5, v0, v0
+; GFX10-NEXT: v_max_f16_e32 v0, v0, v0
; GFX10-NEXT: s_and_b32 s4, s20, -4
-; GFX10-NEXT: v_mov_b32_e32 v4, s4
+; GFX10-NEXT: v_mov_b32_e32 v3, s4
; GFX10-NEXT: s_and_b32 s4, s20, 3
; GFX10-NEXT: s_lshl_b32 s4, s4, 3
; GFX10-NEXT: s_lshl_b32 s5, 0xffff, s4
-; GFX10-NEXT: buffer_load_dword v1, v4, s[16:19], 0 offen
+; GFX10-NEXT: buffer_load_dword v1, v3, s[16:19], 0 offen
; GFX10-NEXT: s_not_b32 s6, s5
; GFX10-NEXT: s_mov_b32 s5, 0
; GFX10-NEXT: .LBB10_1: ; %atomicrmw.start
; GFX10-NEXT: ; =>This Inner Loop Header: Depth=1
; GFX10-NEXT: s_waitcnt vmcnt(0)
-; GFX10-NEXT: v_lshrrev_b32_e32 v0, s4, v1
+; GFX10-NEXT: v_mov_b32_e32 v5, v1
; GFX10-NEXT: s_waitcnt_vscnt null, 0x0
-; GFX10-NEXT: v_max_f16_e32 v0, v0, v0
-; GFX10-NEXT: v_min_f16_e32 v0, v0, v5
-; GFX10-NEXT: v_lshlrev_b32_sdwa v0, s4, v0 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:WORD_0
-; GFX10-NEXT: v_and_or_b32 v0, v1, s6, v0
-; GFX10-NEXT: v_mov_b32_e32 v3, v1
-; GFX10-NEXT: v_mov_b32_e32 v2, v0
-; GFX10-NEXT: buffer_atomic_cmpswap v[2:3], v4, s[16:19], 0 offen glc
+; GFX10-NEXT: v_lshrrev_b32_e32 v1, s4, v5
+; GFX10-NEXT: v_max_f16_e32 v1, v1, v1
+; GFX10-NEXT: v_min_f16_e32 v1, v1, v0
+; GFX10-NEXT: v_lshlrev_b32_sdwa v1, s4, v1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:WORD_0
+; GFX10-NEXT: v_and_or_b32 v4, v5, s6, v1
+; GFX10-NEXT: v_mov_b32_e32 v1, v4
+; GFX10-NEXT: v_mov_b32_e32 v2, v5
+; GFX10-NEXT: buffer_atomic_cmpswap v[1:2], v3, s[16:19], 0 offen glc
; GFX10-NEXT: s_waitcnt vmcnt(0)
; GFX10-NEXT: buffer_gl1_inv
; GFX10-NEXT: buffer_gl0_inv
-; GFX10-NEXT: v_cmp_eq_u32_e32 vcc_lo, v2, v1
-; GFX10-NEXT: v_mov_b32_e32 v1, v2
+; GFX10-NEXT: v_cmp_eq_u32_e32 vcc_lo, v1, v5
; GFX10-NEXT: s_or_b32 s5, vcc_lo, s5
; GFX10-NEXT: s_andn2_b32 exec_lo, exec_lo, s5
; GFX10-NEXT: s_cbranch_execnz .LBB10_1
; GFX10-NEXT: ; %bb.2: ; %atomicrmw.end
; GFX10-NEXT: s_or_b32 exec_lo, exec_lo, s5
-; GFX10-NEXT: v_lshrrev_b32_e32 v0, s4, v2
+; GFX10-NEXT: v_lshrrev_b32_e32 v0, s4, v1
; GFX10-NEXT: s_setpc_b64 s[30:31]
;
; GFX90A-LABEL: buffer_fat_ptr_agent_atomic_fmin_ret_f16__offset__amdgpu_no_fine_grained_memory:
@@ -2755,29 +2761,29 @@ define half @buffer_fat_ptr_agent_atomic_fmin_ret_f16__offset__amdgpu_no_fine_gr
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX90A-NEXT: s_addk_i32 s20, 0x200
; GFX90A-NEXT: s_and_b32 s4, s20, -4
-; GFX90A-NEXT: v_mov_b32_e32 v4, s4
-; GFX90A-NEXT: buffer_load_dword v1, v4, s[16:19], 0 offen
+; GFX90A-NEXT: v_mov_b32_e32 v1, s4
+; GFX90A-NEXT: buffer_load_dword v2, v1, s[16:19], 0 offen
; GFX90A-NEXT: s_and_b32 s4, s20, 3
; GFX90A-NEXT: s_lshl_b32 s6, s4, 3
; GFX90A-NEXT: s_lshl_b32 s4, 0xffff, s6
; GFX90A-NEXT: s_not_b32 s7, s4
; GFX90A-NEXT: s_mov_b64 s[4:5], 0
-; GFX90A-NEXT: v_max_f16_e32 v5, v0, v0
+; GFX90A-NEXT: v_max_f16_e32 v0, v0, v0
; GFX90A-NEXT: .LBB10_1: ; %atomicrmw.start
; GFX90A-NEXT: ; =>This Inner Loop Header: Depth=1
; GFX90A-NEXT: s_waitcnt vmcnt(0)
-; GFX90A-NEXT: v_lshrrev_b32_e32 v0, s6, v1
-; GFX90A-NEXT: v_max_f16_e32 v0, v0, v0
-; GFX90A-NEXT: v_min_f16_e32 v0, v0, v5
-; GFX90A-NEXT: v_lshlrev_b32_e32 v0, s6, v0
-; GFX90A-NEXT: v_and_or_b32 v0, v1, s7, v0
-; GFX90A-NEXT: v_pk_mov_b32 v[2:3], v[0:1], v[0:1] op_sel:[0,1]
-; GFX90A-NEXT: buffer_atomic_cmpswap v[2:3], v4, s[16:19], 0 offen glc
+; GFX90A-NEXT: v_mov_b32_e32 v5, v2
+; GFX90A-NEXT: v_lshrrev_b32_e32 v2, s6, v5
+; GFX90A-NEXT: v_max_f16_e32 v2, v2, v2
+; GFX90A-NEXT: v_min_f16_e32 v2, v2, v0
+; GFX90A-NEXT: v_lshlrev_b32_e32 v2, s6, v2
+; GFX90A-NEXT: v_and_or_b32 v4, v5, s7, v2
+; GFX90A-NEXT: v_pk_mov_b32 v[2:3], v[4:5], v[4:5] op_sel:[0,1]
+; GFX90A-NEXT: buffer_atomic_cmpswap v[2:3], v1, s[16:19], 0 offen glc
; GFX90A-NEXT: s_waitcnt vmcnt(0)
; GFX90A-NEXT: buffer_wbinvl1
-; GFX90A-NEXT: v_cmp_eq_u32_e32 vcc, v2, v1
+; GFX90A-NEXT: v_cmp_eq_u32_e32 vcc, v2, v5
; GFX90A-NEXT: s_or_b64 s[4:5], vcc, s[4:5]
-; GFX90A-NEXT: v_mov_b32_e32 v1, v2
; GFX90A-NEXT: s_andn2_b64 exec, exec, s[4:5]
; GFX90A-NEXT: s_cbranch_execnz .LBB10_1
; GFX90A-NEXT: ; %bb.2: ; %atomicrmw.end
@@ -2790,35 +2796,35 @@ define half @buffer_fat_ptr_agent_atomic_fmin_ret_f16__offset__amdgpu_no_fine_gr
; GFX908-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX908-NEXT: s_addk_i32 s20, 0x200
; GFX908-NEXT: s_and_b32 s4, s20, -4
-; GFX908-NEXT: v_mov_b32_e32 v4, s4
-; GFX908-NEXT: buffer_load_dword v1, v4, s[16:19], 0 offen
+; GFX908-NEXT: v_mov_b32_e32 v3, s4
+; GFX908-NEXT: buffer_load_dword v1, v3, s[16:19], 0 offen
; GFX908-NEXT: s_and_b32 s4, s20, 3
; GFX908-NEXT: s_lshl_b32 s6, s4, 3
; GFX908-NEXT: s_lshl_b32 s4, 0xffff, s6
; GFX908-NEXT: s_not_b32 s7, s4
; GFX908-NEXT: s_mov_b64 s[4:5], 0
-; GFX908-NEXT: v_max_f16_e32 v5, v0, v0
+; GFX908-NEXT: v_max_f16_e32 v0, v0, v0
; GFX908-NEXT: .LBB10_1: ; %atomicrmw.start
; GFX908-NEXT: ; =>This Inner Loop Header: Depth=1
; GFX908-NEXT: s_waitcnt vmcnt(0)
-; GFX908-NEXT: v_lshrrev_b32_e32 v0, s6, v1
-; GFX908-NEXT: v_max_f16_e32 v0, v0, v0
-; GFX908-NEXT: v_min_f16_e32 v0, v0, v5
-; GFX908-NEXT: v_lshlrev_b32_e32 v0, s6, v0
-; GFX908-NEXT: v_and_or_b32 v0, v1, s7, v0
-; GFX908-NEXT: v_mov_b32_e32 v3, v1
-; GFX908-NEXT: v_mov_b32_e32 v2, v0
-; GFX908-NEXT: buffer_atomic_cmpswap v[2:3], v4, s[16:19], 0 offen glc
+; GFX908-NEXT: v_mov_b32_e32 v5, v1
+; GFX908-NEXT: v_lshrrev_b32_e32 v1, s6, v5
+; GFX908-NEXT: v_max_f16_e32 v1, v1, v1
+; GFX908-NEXT: v_min_f16_e32 v1, v1, v0
+; GFX908-NEXT: v_lshlrev_b32_e32 v1, s6, v1
+; GFX908-NEXT: v_and_or_b32 v4, v5, s7, v1
+; GFX908-NEXT: v_mov_b32_e32 v1, v4
+; GFX908-NEXT: v_mov_b32_e32 v2, v5
+; GFX908-NEXT: buffer_atomic_cmpswap v[1:2], v3, s[16:19], 0 offen glc
; GFX908-NEXT: s_waitcnt vmcnt(0)
; GFX908-NEXT: buffer_wbinvl1
-; GFX908-NEXT: v_cmp_eq_u32_e32 vcc, v2, v1
+; GFX908-NEXT: v_cmp_eq_u32_e32 vcc, v1, v5
; GFX908-NEXT: s_or_b64 s[4:5], vcc, s[4:5]
-; GFX908-NEXT: v_mov_b32_e32 v1, v2
; GFX908-NEXT: s_andn2_b64 exec, exec, s[4:5]
; GFX908-NEXT: s_cbranch_execnz .LBB10_1
; GFX908-NEXT: ; %bb.2: ; %atomicrmw.end
; GFX908-NEXT: s_or_b64 exec, exec, s[4:5]
-; GFX908-NEXT: v_lshrrev_b32_e32 v0, s6, v2
+; GFX908-NEXT: v_lshrrev_b32_e32 v0, s6, v1
; GFX908-NEXT: s_setpc_b64 s[30:31]
;
; GFX8-LABEL: buffer_fat_ptr_agent_atomic_fmin_ret_f16__offset__amdgpu_no_fine_grained_memory:
@@ -2826,36 +2832,36 @@ define half @buffer_fat_ptr_agent_atomic_fmin_ret_f16__offset__amdgpu_no_fine_gr
; GFX8-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX8-NEXT: s_addk_i32 s20, 0x200
; GFX8-NEXT: s_and_b32 s4, s20, -4
-; GFX8-NEXT: v_mov_b32_e32 v4, s4
-; GFX8-NEXT: buffer_load_dword v1, v4, s[16:19], 0 offen
+; GFX8-NEXT: v_mov_b32_e32 v3, s4
+; GFX8-NEXT: buffer_load_dword v1, v3, s[16:19], 0 offen
; GFX8-NEXT: s_and_b32 s4, s20, 3
; GFX8-NEXT: s_lshl_b32 s6, s4, 3
; GFX8-NEXT: s_lshl_b32 s4, 0xffff, s6
; GFX8-NEXT: s_not_b32 s7, s4
; GFX8-NEXT: s_mov_b64 s[4:5], 0
-; GFX8-NEXT: v_max_f16_e32 v5, v0, v0
+; GFX8-NEXT: v_max_f16_e32 v0, v0, v0
; GFX8-NEXT: .LBB10_1: ; %atomicrmw.start
; GFX8-NEXT: ; =>This Inner Loop Header: Depth=1
; GFX8-NEXT: s_waitcnt vmcnt(0)
-; GFX8-NEXT: v_lshrrev_b32_e32 v0, s6, v1
-; GFX8-NEXT: v_max_f16_e32 v0, v0, v0
-; GFX8-NEXT: v_min_f16_e32 v0, v0, v5
-; GFX8-NEXT: v_and_b32_e32 v2, s7, v1
-; GFX8-NEXT: v_lshlrev_b32_e32 v0, s6, v0
-; GFX8-NEXT: v_or_b32_e32 v0, v2, v0
-; GFX8-NEXT: v_mov_b32_e32 v3, v1
-; GFX8-NEXT: v_mov_b32_e32 v2, v0
-; GFX8-NEXT: buffer_atomic_cmpswap v[2:3], v4, s[16:19], 0 offen glc
+; GFX8-NEXT: v_mov_b32_e32 v5, v1
+; GFX8-NEXT: v_lshrrev_b32_e32 v1, s6, v5
+; GFX8-NEXT: v_max_f16_e32 v1, v1, v1
+; GFX8-NEXT: v_min_f16_e32 v1, v1, v0
+; GFX8-NEXT: v_and_b32_e32 v2, s7, v5
+; GFX8-NEXT: v_lshlrev_b32_e32 v1, s6, v1
+; GFX8-NEXT: v_or_b32_e32 v4, v2, v1
+; GFX8-NEXT: v_mov_b32_e32 v1, v4
+; GFX8-NEXT: v_mov_b32_e32 v2, v5
+; GFX8-NEXT: buffer_atomic_cmpswap v[1:2], v3, s[16:19], 0 offen glc
; GFX8-NEXT: s_waitcnt vmcnt(0)
; GFX8-NEXT: buffer_wbinvl1
-; GFX8-NEXT: v_cmp_eq_u32_e32 vcc, v2, v1
+; GFX8-NEXT: v_cmp_eq_u32_e32 vcc, v1, v5
; GFX8-NEXT: s_or_b64 s[4:5], vcc, s[4:5]
-; GFX8-NEXT: v_mov_b32_e32 v1, v2
; GFX8-NEXT: s_andn2_b64 exec, exec, s[4:5]
; GFX8-NEXT: s_cbranch_execnz .LBB10_1
; GFX8-NEXT: ; %bb.2: ; %atomicrmw.end
; GFX8-NEXT: s_or_b64 exec, exec, s[4:5]
-; GFX8-NEXT: v_lshrrev_b32_e32 v0, s6, v2
+; GFX8-NEXT: v_lshrrev_b32_e32 v0, s6, v1
; GFX8-NEXT: s_setpc_b64 s[30:31]
;
; GFX7-LABEL: buffer_fat_ptr_agent_atomic_fmin_ret_f16__offset__amdgpu_no_fine_grained_memory:
@@ -2863,38 +2869,38 @@ define half @buffer_fat_ptr_agent_atomic_fmin_ret_f16__offset__amdgpu_no_fine_gr
; GFX7-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX7-NEXT: s_addk_i32 s20, 0x200
; GFX7-NEXT: s_and_b32 s4, s20, -4
-; GFX7-NEXT: v_mov_b32_e32 v4, s4
-; GFX7-NEXT: buffer_load_dword v1, v4, s[16:19], 0 offen
+; GFX7-NEXT: v_mov_b32_e32 v3, s4
+; GFX7-NEXT: buffer_load_dword v1, v3, s[16:19], 0 offen
; GFX7-NEXT: v_cvt_f16_f32_e32 v0, v0
; GFX7-NEXT: s_and_b32 s4, s20, 3
; GFX7-NEXT: s_lshl_b32 s6, s4, 3
; GFX7-NEXT: s_lshl_b32 s4, 0xffff, s6
-; GFX7-NEXT: v_cvt_f32_f16_e32 v5, v0
+; GFX7-NEXT: v_cvt_f32_f16_e32 v0, v0
; GFX7-NEXT: s_not_b32 s7, s4
; GFX7-NEXT: s_mov_b64 s[4:5], 0
; GFX7-NEXT: .LBB10_1: ; %atomicrmw.start
; GFX7-NEXT: ; =>This Inner Loop Header: Depth=1
; GFX7-NEXT: s_waitcnt vmcnt(0)
-; GFX7-NEXT: v_lshrrev_b32_e32 v0, s6, v1
-; GFX7-NEXT: v_cvt_f32_f16_e32 v0, v0
-; GFX7-NEXT: v_and_b32_e32 v2, s7, v1
-; GFX7-NEXT: v_min_f32_e32 v0, v0, v5
-; GFX7-NEXT: v_cvt_f16_f32_e32 v0, v0
-; GFX7-NEXT: v_lshlrev_b32_e32 v0, s6, v0
-; GFX7-NEXT: v_or_b32_e32 v0, v2, v0
-; GFX7-NEXT: v_mov_b32_e32 v3, v1
-; GFX7-NEXT: v_mov_b32_e32 v2, v0
-; GFX7-NEXT: buffer_atomic_cmpswap v[2:3], v4, s[16:19], 0 offen glc
+; GFX7-NEXT: v_mov_b32_e32 v5, v1
+; GFX7-NEXT: v_lshrrev_b32_e32 v1, s6, v5
+; GFX7-NEXT: v_cvt_f32_f16_e32 v1, v1
+; GFX7-NEXT: v_and_b32_e32 v2, s7, v5
+; GFX7-NEXT: v_min_f32_e32 v1, v1, v0
+; GFX7-NEXT: v_cvt_f16_f32_e32 v1, v1
+; GFX7-NEXT: v_lshlrev_b32_e32 v1, s6, v1
+; GFX7-NEXT: v_or_b32_e32 v4, v2, v1
+; GFX7-NEXT: v_mov_b32_e32 v1, v4
+; GFX7-NEXT: v_mov_b32_e32 v2, v5
+; GFX7-NEXT: buffer_atomic_cmpswap v[1:2], v3, s[16:19], 0 offen glc
; GFX7-NEXT: s_waitcnt vmcnt(0)
; GFX7-NEXT: buffer_wbinvl1
-; GFX7-NEXT: v_cmp_eq_u32_e32 vcc, v2, v1
+; GFX7-NEXT: v_cmp_eq_u32_e32 vcc, v1, v5
; GFX7-NEXT: s_or_b64 s[4:5], vcc, s[4:5]
-; GFX7-NEXT: v_mov_b32_e32 v1, v2
; GFX7-NEXT: s_andn2_b64 exec, exec, s[4:5]
; GFX7-NEXT: s_cbranch_execnz .LBB10_1
; GFX7-NEXT: ; %bb.2: ; %atomicrmw.end
; GFX7-NEXT: s_or_b64 exec, exec, s[4:5]
-; GFX7-NEXT: v_lshrrev_b32_e32 v0, s6, v2
+; GFX7-NEXT: v_lshrrev_b32_e32 v0, s6, v1
; GFX7-NEXT: v_cvt_f32_f16_e32 v0, v0
; GFX7-NEXT: s_setpc_b64 s[30:31]
;
@@ -2903,39 +2909,39 @@ define half @buffer_fat_ptr_agent_atomic_fmin_ret_f16__offset__amdgpu_no_fine_gr
; GFX6-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX6-NEXT: s_addk_i32 s20, 0x200
; GFX6-NEXT: s_and_b32 s4, s20, -4
-; GFX6-NEXT: v_mov_b32_e32 v4, s4
-; GFX6-NEXT: buffer_load_dword v1, v4, s[16:19], 0 offen
+; GFX6-NEXT: v_mov_b32_e32 v3, s4
+; GFX6-NEXT: buffer_load_dword v1, v3, s[16:19], 0 offen
; GFX6-NEXT: v_cvt_f16_f32_e32 v0, v0
; GFX6-NEXT: s_and_b32 s4, s20, 3
; GFX6-NEXT: s_lshl_b32 s6, s4, 3
; GFX6-NEXT: s_lshl_b32 s4, 0xffff, s6
-; GFX6-NEXT: v_cvt_f32_f16_e32 v5, v0
+; GFX6-NEXT: v_cvt_f32_f16_e32 v0, v0
; GFX6-NEXT: s_not_b32 s7, s4
; GFX6-NEXT: s_mov_b64 s[4:5], 0
; GFX6-NEXT: .LBB10_1: ; %atomicrmw.start
; GFX6-NEXT: ; =>This Inner Loop Header: Depth=1
; GFX6-NEXT: s_waitcnt vmcnt(0)
-; GFX6-NEXT: v_lshrrev_b32_e32 v0, s6, v1
-; GFX6-NEXT: v_cvt_f32_f16_e32 v0, v0
+; GFX6-NEXT: v_mov_b32_e32 v5, v1
; GFX6-NEXT: s_waitcnt expcnt(0)
-; GFX6-NEXT: v_and_b32_e32 v2, s7, v1
-; GFX6-NEXT: v_min_f32_e32 v0, v0, v5
-; GFX6-NEXT: v_cvt_f16_f32_e32 v0, v0
-; GFX6-NEXT: v_lshlrev_b32_e32 v0, s6, v0
-; GFX6-NEXT: v_or_b32_e32 v0, v2, v0
-; GFX6-NEXT: v_mov_b32_e32 v3, v1
-; GFX6-NEXT: v_mov_b32_e32 v2, v0
-; GFX6-NEXT: buffer_atomic_cmpswap v[2:3], v4, s[16:19], 0 offen glc
+; GFX6-NEXT: v_lshrrev_b32_e32 v1, s6, v5
+; GFX6-NEXT: v_cvt_f32_f16_e32 v1, v1
+; GFX6-NEXT: v_and_b32_e32 v2, s7, v5
+; GFX6-NEXT: v_min_f32_e32 v1, v1, v0
+; GFX6-NEXT: v_cvt_f16_f32_e32 v1, v1
+; GFX6-NEXT: v_lshlrev_b32_e32 v1, s6, v1
+; GFX6-NEXT: v_or_b32_e32 v4, v2, v1
+; GFX6-NEXT: v_mov_b32_e32 v1, v4
+; GFX6-NEXT: v_mov_b32_e32 v2, v5
+; GFX6-NEXT: buffer_atomic_cmpswap v[1:2], v3, s[16:19], 0 offen glc
; GFX6-NEXT: s_waitcnt vmcnt(0)
; GFX6-NEXT: buffer_wbinvl1
-; GFX6-NEXT: v_cmp_eq_u32_e32 vcc, v2, v1
+; GFX6-NEXT: v_cmp_eq_u32_e32 vcc, v1, v5
; GFX6-NEXT: s_or_b64 s[4:5], vcc, s[4:5]
-; GFX6-NEXT: v_mov_b32_e32 v1, v2
; GFX6-NEXT: s_andn2_b64 exec, exec, s[4:5]
; GFX6-NEXT: s_cbranch_execnz .LBB10_1
; GFX6-NEXT: ; %bb.2: ; %atomicrmw.end
; GFX6-NEXT: s_or_b64 exec, exec, s[4:5]
-; GFX6-NEXT: v_lshrrev_b32_e32 v0, s6, v2
+; GFX6-NEXT: v_lshrrev_b32_e32 v0, s6, v1
; GFX6-NEXT: v_cvt_f32_f16_e32 v0, v0
; GFX6-NEXT: s_waitcnt expcnt(0)
; GFX6-NEXT: s_setpc_b64 s[30:31]
@@ -2963,29 +2969,30 @@ define void @buffer_fat_ptr_agent_atomic_fmin_noret_f16__offset__amdgpu_no_fine_
; GFX12-TRUE16-NEXT: s_lshl_b32 s4, s4, 3
; GFX12-TRUE16-NEXT: s_wait_alu 0xfffe
; GFX12-TRUE16-NEXT: s_lshl_b32 s5, 0xffff, s4
-; GFX12-TRUE16-NEXT: buffer_load_b32 v2, v3, s[0:3], null offen
+; GFX12-TRUE16-NEXT: buffer_load_b32 v1, v3, s[0:3], null offen
; GFX12-TRUE16-NEXT: s_not_b32 s6, s5
; GFX12-TRUE16-NEXT: s_mov_b32 s5, 0
; GFX12-TRUE16-NEXT: .LBB11_1: ; %atomicrmw.start
; GFX12-TRUE16-NEXT: ; =>This Inner Loop Header: Depth=1
; GFX12-TRUE16-NEXT: s_wait_loadcnt 0x0
-; GFX12-TRUE16-NEXT: v_lshrrev_b32_e32 v1, s4, v2
+; GFX12-TRUE16-NEXT: v_mov_b32_e32 v5, v1
; GFX12-TRUE16-NEXT: s_wait_storecnt 0x0
; GFX12-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
+; GFX12-TRUE16-NEXT: v_lshrrev_b32_e32 v1, s4, v5
; GFX12-TRUE16-NEXT: v_max_num_f16_e32 v0.h, v1.l, v1.l
-; GFX12-TRUE16-NEXT: v_min_num_f16_e32 v1.l, v0.h, v0.l
; GFX12-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
+; GFX12-TRUE16-NEXT: v_min_num_f16_e32 v1.l, v0.h, v0.l
; GFX12-TRUE16-NEXT: v_and_b32_e32 v1, 0xffff, v1
+; GFX12-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(SKIP_1) | instid1(VALU_DEP_1)
; GFX12-TRUE16-NEXT: v_lshlrev_b32_e32 v1, s4, v1
; GFX12-TRUE16-NEXT: s_wait_alu 0xfffe
-; GFX12-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
-; GFX12-TRUE16-NEXT: v_and_or_b32 v1, v2, s6, v1
-; GFX12-TRUE16-NEXT: v_dual_mov_b32 v5, v2 :: v_dual_mov_b32 v4, v1
-; GFX12-TRUE16-NEXT: buffer_atomic_cmpswap_b32 v[4:5], v3, s[0:3], null offen th:TH_ATOMIC_RETURN
+; GFX12-TRUE16-NEXT: v_and_or_b32 v4, v5, s6, v1
+; GFX12-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1)
+; GFX12-TRUE16-NEXT: v_dual_mov_b32 v1, v4 :: v_dual_mov_b32 v2, v5
+; GFX12-TRUE16-NEXT: buffer_atomic_cmpswap_b32 v[1:2], v3, s[0:3], null offen th:TH_ATOMIC_RETURN
; GFX12-TRUE16-NEXT: s_wait_loadcnt 0x0
; GFX12-TRUE16-NEXT: global_inv scope:SCOPE_DEV
-; GFX12-TRUE16-NEXT: v_cmp_eq_u32_e32 vcc_lo, v4, v2
-; GFX12-TRUE16-NEXT: v_mov_b32_e32 v2, v4
+; GFX12-TRUE16-NEXT: v_cmp_eq_u32_e32 vcc_lo, v1, v5
; GFX12-TRUE16-NEXT: s_or_b32 s5, vcc_lo, s5
; GFX12-TRUE16-NEXT: s_wait_alu 0xfffe
; GFX12-TRUE16-NEXT: s_and_not1_b32 exec_lo, exec_lo, s5
@@ -3003,39 +3010,40 @@ define void @buffer_fat_ptr_agent_atomic_fmin_noret_f16__offset__amdgpu_no_fine_
; GFX12-FAKE16-NEXT: s_wait_bvhcnt 0x0
; GFX12-FAKE16-NEXT: s_wait_kmcnt 0x0
; GFX12-FAKE16-NEXT: s_addk_co_i32 s16, 0x200
-; GFX12-FAKE16-NEXT: v_max_num_f16_e32 v3, v0, v0
+; GFX12-FAKE16-NEXT: v_max_num_f16_e32 v0, v0, v0
; GFX12-FAKE16-NEXT: s_wait_alu 0xfffe
; GFX12-FAKE16-NEXT: s_and_b32 s4, s16, -4
; GFX12-FAKE16-NEXT: s_wait_alu 0xfffe
-; GFX12-FAKE16-NEXT: v_mov_b32_e32 v2, s4
+; GFX12-FAKE16-NEXT: v_mov_b32_e32 v3, s4
; GFX12-FAKE16-NEXT: s_and_b32 s4, s16, 3
; GFX12-FAKE16-NEXT: s_wait_alu 0xfffe
; GFX12-FAKE16-NEXT: s_lshl_b32 s4, s4, 3
; GFX12-FAKE16-NEXT: s_wait_alu 0xfffe
; GFX12-FAKE16-NEXT: s_lshl_b32 s5, 0xffff, s4
-; GFX12-FAKE16-NEXT: buffer_load_b32 v1, v2, s[0:3], null offen
+; GFX12-FAKE16-NEXT: buffer_load_b32 v1, v3, s[0:3], null offen
; GFX12-FAKE16-NEXT: s_not_b32 s6, s5
; GFX12-FAKE16-NEXT: s_mov_b32 s5, 0
; GFX12-FAKE16-NEXT: .LBB11_1: ; %atomicrmw.start
; GFX12-FAKE16-NEXT: ; =>This Inner Loop Header: Depth=1
; GFX12-FAKE16-NEXT: s_wait_loadcnt 0x0
-; GFX12-FAKE16-NEXT: v_lshrrev_b32_e32 v0, s4, v1
+; GFX12-FAKE16-NEXT: v_mov_b32_e32 v5, v1
; GFX12-FAKE16-NEXT: s_wait_storecnt 0x0
; GFX12-FAKE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
-; GFX12-FAKE16-NEXT: v_max_num_f16_e32 v0, v0, v0
-; GFX12-FAKE16-NEXT: v_min_num_f16_e32 v0, v0, v3
+; GFX12-FAKE16-NEXT: v_lshrrev_b32_e32 v1, s4, v5
+; GFX12-FAKE16-NEXT: v_max_num_f16_e32 v1, v1, v1
; GFX12-FAKE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
-; GFX12-FAKE16-NEXT: v_and_b32_e32 v0, 0xffff, v0
-; GFX12-FAKE16-NEXT: v_lshlrev_b32_e32 v0, s4, v0
+; GFX12-FAKE16-NEXT: v_min_num_f16_e32 v1, v1, v0
+; GFX12-FAKE16-NEXT: v_and_b32_e32 v1, 0xffff, v1
+; GFX12-FAKE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(SKIP_1) | instid1(VALU_DEP_1)
+; GFX12-FAKE16-NEXT: v_lshlrev_b32_e32 v1, s4, v1
; GFX12-FAKE16-NEXT: s_wait_alu 0xfffe
-; GFX12-FAKE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
-; GFX12-FAKE16-NEXT: v_and_or_b32 v0, v1, s6, v0
-; GFX12-FAKE16-NEXT: v_dual_mov_b32 v5, v1 :: v_dual_mov_b32 v4, v0
-; GFX12-FAKE16-NEXT: buffer_atomic_cmpswap_b32 v[4:5], v2, s[0:3], null offen th:TH_ATOMIC_RETURN
+; GFX12-FAKE16-NEXT: v_and_or_b32 v4, v5, s6, v1
+; GFX12-FAKE16-NEXT: s_delay_alu instid0(VALU_DEP_1)
+; GFX12-FAKE16-NEXT: v_dual_mov_b32 v1, v4 :: v_dual_mov_b32 v2, v5
+; GFX12-FAKE16-NEXT: buffer_atomic_cmpswap_b32 v[1:2], v3, s[0:3], null offen th:TH_ATOMIC_RETURN
; GFX12-FAKE16-NEXT: s_wait_loadcnt 0x0
; GFX12-FAKE16-NEXT: global_inv scope:SCOPE_DEV
-; GFX12-FAKE16-NEXT: v_cmp_eq_u32_e32 vcc_lo, v4, v1
-; GFX12-FAKE16-NEXT: v_mov_b32_e32 v1, v4
+; GFX12-FAKE16-NEXT: v_cmp_eq_u32_e32 vcc_lo, v1, v5
; GFX12-FAKE16-NEXT: s_or_b32 s5, vcc_lo, s5
; GFX12-FAKE16-NEXT: s_wait_alu 0xfffe
; GFX12-FAKE16-NEXT: s_and_not1_b32 exec_lo, exec_lo, s5
@@ -3050,30 +3058,30 @@ define void @buffer_fat_ptr_agent_atomic_fmin_noret_f16__offset__amdgpu_no_fine_
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX942-NEXT: s_addk_i32 s16, 0x200
; GFX942-NEXT: s_and_b32 s4, s16, -4
-; GFX942-NEXT: v_mov_b32_e32 v2, s4
-; GFX942-NEXT: buffer_load_dword v1, v2, s[0:3], 0 offen
+; GFX942-NEXT: v_mov_b32_e32 v1, s4
+; GFX942-NEXT: buffer_load_dword v2, v1, s[0:3], 0 offen
; GFX942-NEXT: s_and_b32 s4, s16, 3
; GFX942-NEXT: s_lshl_b32 s6, s4, 3
; GFX942-NEXT: s_lshl_b32 s4, 0xffff, s6
; GFX942-NEXT: s_not_b32 s7, s4
; GFX942-NEXT: s_mov_b64 s[4:5], 0
-; GFX942-NEXT: v_max_f16_e32 v3, v0, v0
+; GFX942-NEXT: v_max_f16_e32 v0, v0, v0
; GFX942-NEXT: .LBB11_1: ; %atomicrmw.start
; GFX942-NEXT: ; =>This Inner Loop Header: Depth=1
; GFX942-NEXT: s_waitcnt vmcnt(0)
-; GFX942-NEXT: v_lshrrev_b32_e32 v0, s6, v1
-; GFX942-NEXT: v_max_f16_e32 v0, v0, v0
-; GFX942-NEXT: v_min_f16_e32 v0, v0, v3
-; GFX942-NEXT: v_lshlrev_b32_e32 v0, s6, v0
-; GFX942-NEXT: v_and_or_b32 v0, v1, s7, v0
-; GFX942-NEXT: v_mov_b64_e32 v[4:5], v[0:1]
+; GFX942-NEXT: v_mov_b32_e32 v5, v2
+; GFX942-NEXT: v_lshrrev_b32_e32 v2, s6, v5
+; GFX942-NEXT: v_max_f16_e32 v2, v2, v2
+; GFX942-NEXT: v_min_f16_e32 v2, v2, v0
+; GFX942-NEXT: v_lshlrev_b32_e32 v2, s6, v2
+; GFX942-NEXT: v_and_or_b32 v4, v5, s7, v2
+; GFX942-NEXT: v_mov_b64_e32 v[2:3], v[4:5]
; GFX942-NEXT: buffer_wbl2 sc1
-; GFX942-NEXT: buffer_atomic_cmpswap v[4:5], v2, s[0:3], 0 offen sc0
+; GFX942-NEXT: buffer_atomic_cmpswap v[2:3], v1, s[0:3], 0 offen sc0
; GFX942-NEXT: s_waitcnt vmcnt(0)
; GFX942-NEXT: buffer_inv sc1
-; GFX942-NEXT: v_cmp_eq_u32_e32 vcc, v4, v1
+; GFX942-NEXT: v_cmp_eq_u32_e32 vcc, v2, v5
; GFX942-NEXT: s_or_b64 s[4:5], vcc, s[4:5]
-; GFX942-NEXT: v_mov_b32_e32 v1, v4
; GFX942-NEXT: s_andn2_b64 exec, exec, s[4:5]
; GFX942-NEXT: s_cbranch_execnz .LBB11_1
; GFX942-NEXT: ; %bb.2: ; %atomicrmw.end
@@ -3092,30 +3100,31 @@ define void @buffer_fat_ptr_agent_atomic_fmin_noret_f16__offset__amdgpu_no_fine_
; GFX11-TRUE16-NEXT: s_lshl_b32 s4, s4, 3
; GFX11-TRUE16-NEXT: s_delay_alu instid0(SALU_CYCLE_1)
; GFX11-TRUE16-NEXT: s_lshl_b32 s5, 0xffff, s4
-; GFX11-TRUE16-NEXT: buffer_load_b32 v2, v3, s[0:3], 0 offen
+; GFX11-TRUE16-NEXT: buffer_load_b32 v1, v3, s[0:3], 0 offen
; GFX11-TRUE16-NEXT: s_not_b32 s6, s5
; GFX11-TRUE16-NEXT: s_mov_b32 s5, 0
; GFX11-TRUE16-NEXT: .p2align 6
; GFX11-TRUE16-NEXT: .LBB11_1: ; %atomicrmw.start
; GFX11-TRUE16-NEXT: ; =>This Inner Loop Header: Depth=1
; GFX11-TRUE16-NEXT: s_waitcnt vmcnt(0)
-; GFX11-TRUE16-NEXT: v_lshrrev_b32_e32 v1, s4, v2
+; GFX11-TRUE16-NEXT: v_mov_b32_e32 v5, v1
; GFX11-TRUE16-NEXT: s_waitcnt_vscnt null, 0x0
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
+; GFX11-TRUE16-NEXT: v_lshrrev_b32_e32 v1, s4, v5
; GFX11-TRUE16-NEXT: v_max_f16_e32 v0.h, v1.l, v1.l
-; GFX11-TRUE16-NEXT: v_min_f16_e32 v1.l, v0.h, v0.l
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
+; GFX11-TRUE16-NEXT: v_min_f16_e32 v1.l, v0.h, v0.l
; GFX11-TRUE16-NEXT: v_and_b32_e32 v1, 0xffff, v1
-; GFX11-TRUE16-NEXT: v_lshlrev_b32_e32 v1, s4, v1
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
-; GFX11-TRUE16-NEXT: v_and_or_b32 v1, v2, s6, v1
-; GFX11-TRUE16-NEXT: v_dual_mov_b32 v5, v2 :: v_dual_mov_b32 v4, v1
-; GFX11-TRUE16-NEXT: buffer_atomic_cmpswap_b32 v[4:5], v3, s[0:3], 0 offen glc
+; GFX11-TRUE16-NEXT: v_lshlrev_b32_e32 v1, s4, v1
+; GFX11-TRUE16-NEXT: v_and_or_b32 v4, v5, s6, v1
+; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1)
+; GFX11-TRUE16-NEXT: v_dual_mov_b32 v1, v4 :: v_dual_mov_b32 v2, v5
+; GFX11-TRUE16-NEXT: buffer_atomic_cmpswap_b32 v[1:2], v3, s[0:3], 0 offen glc
; GFX11-TRUE16-NEXT: s_waitcnt vmcnt(0)
; GFX11-TRUE16-NEXT: buffer_gl1_inv
; GFX11-TRUE16-NEXT: buffer_gl0_inv
-; GFX11-TRUE16-NEXT: v_cmp_eq_u32_e32 vcc_lo, v4, v2
-; GFX11-TRUE16-NEXT: v_mov_b32_e32 v2, v4
+; GFX11-TRUE16-NEXT: v_cmp_eq_u32_e32 vcc_lo, v1, v5
; GFX11-TRUE16-NEXT: s_or_b32 s5, vcc_lo, s5
; GFX11-TRUE16-NEXT: s_delay_alu instid0(SALU_CYCLE_1)
; GFX11-TRUE16-NEXT: s_and_not1_b32 exec_lo, exec_lo, s5
@@ -3128,38 +3137,39 @@ define void @buffer_fat_ptr_agent_atomic_fmin_noret_f16__offset__amdgpu_no_fine_
; GFX11-FAKE16: ; %bb.0:
; GFX11-FAKE16-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX11-FAKE16-NEXT: s_addk_i32 s16, 0x200
-; GFX11-FAKE16-NEXT: v_max_f16_e32 v3, v0, v0
+; GFX11-FAKE16-NEXT: v_max_f16_e32 v0, v0, v0
; GFX11-FAKE16-NEXT: s_and_b32 s4, s16, -4
; GFX11-FAKE16-NEXT: s_delay_alu instid0(SALU_CYCLE_1) | instskip(SKIP_1) | instid1(SALU_CYCLE_1)
-; GFX11-FAKE16-NEXT: v_mov_b32_e32 v2, s4
+; GFX11-FAKE16-NEXT: v_mov_b32_e32 v3, s4
; GFX11-FAKE16-NEXT: s_and_b32 s4, s16, 3
; GFX11-FAKE16-NEXT: s_lshl_b32 s4, s4, 3
; GFX11-FAKE16-NEXT: s_delay_alu instid0(SALU_CYCLE_1)
; GFX11-FAKE16-NEXT: s_lshl_b32 s5, 0xffff, s4
-; GFX11-FAKE16-NEXT: buffer_load_b32 v1, v2, s[0:3], 0 offen
+; GFX11-FAKE16-NEXT: buffer_load_b32 v1, v3, s[0:3], 0 offen
; GFX11-FAKE16-NEXT: s_not_b32 s6, s5
; GFX11-FAKE16-NEXT: s_mov_b32 s5, 0
; GFX11-FAKE16-NEXT: .p2align 6
; GFX11-FAKE16-NEXT: .LBB11_1: ; %atomicrmw.start
; GFX11-FAKE16-NEXT: ; =>This Inner Loop Header: Depth=1
; GFX11-FAKE16-NEXT: s_waitcnt vmcnt(0)
-; GFX11-FAKE16-NEXT: v_lshrrev_b32_e32 v0, s4, v1
+; GFX11-FAKE16-NEXT: v_mov_b32_e32 v5, v1
; GFX11-FAKE16-NEXT: s_waitcnt_vscnt null, 0x0
; GFX11-FAKE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
-; GFX11-FAKE16-NEXT: v_max_f16_e32 v0, v0, v0
-; GFX11-FAKE16-NEXT: v_min_f16_e32 v0, v0, v3
+; GFX11-FAKE16-NEXT: v_lshrrev_b32_e32 v1, s4, v5
+; GFX11-FAKE16-NEXT: v_max_f16_e32 v1, v1, v1
; GFX11-FAKE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
-; GFX11-FAKE16-NEXT: v_and_b32_e32 v0, 0xffff, v0
-; GFX11-FAKE16-NEXT: v_lshlrev_b32_e32 v0, s4, v0
+; GFX11-FAKE16-NEXT: v_min_f16_e32 v1, v1, v0
+; GFX11-FAKE16-NEXT: v_and_b32_e32 v1, 0xffff, v1
; GFX11-FAKE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
-; GFX11-FAKE16-NEXT: v_and_or_b32 v0, v1, s6, v0
-; GFX11-FAKE16-NEXT: v_dual_mov_b32 v5, v1 :: v_dual_mov_b32 v4, v0
-; GFX11-FAKE16-NEXT: buffer_atomic_cmpswap_b32 v[4:5], v2, s[0:3], 0 offen glc
+; GFX11-FAKE16-NEXT: v_lshlrev_b32_e32 v1, s4, v1
+; GFX11-FAKE16-NEXT: v_and_or_b32 v4, v5, s6, v1
+; GFX11-FAKE16-NEXT: s_delay_alu instid0(VALU_DEP_1)
+; GFX11-FAKE16-NEXT: v_dual_mov_b32 v1, v4 :: v_dual_mov_b32 v2, v5
+; GFX11-FAKE16-NEXT: buffer_atomic_cmpswap_b32 v[1:2], v3, s[0:3], 0 offen glc
; GFX11-FAKE16-NEXT: s_waitcnt vmcnt(0)
; GFX11-FAKE16-NEXT: buffer_gl1_inv
; GFX11-FAKE16-NEXT: buffer_gl0_inv
-; GFX11-FAKE16-NEXT: v_cmp_eq_u32_e32 vcc_lo, v4, v1
-; GFX11-FAKE16-NEXT: v_mov_b32_e32 v1, v4
+; GFX11-FAKE16-NEXT: v_cmp_eq_u32_e32 vcc_lo, v1, v5
; GFX11-FAKE16-NEXT: s_or_b32 s5, vcc_lo, s5
; GFX11-FAKE16-NEXT: s_delay_alu instid0(SALU_CYCLE_1)
; GFX11-FAKE16-NEXT: s_and_not1_b32 exec_lo, exec_lo, s5
@@ -3172,32 +3182,32 @@ define void @buffer_fat_ptr_agent_atomic_fmin_noret_f16__offset__amdgpu_no_fine_
; GFX10: ; %bb.0:
; GFX10-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX10-NEXT: s_addk_i32 s20, 0x200
-; GFX10-NEXT: v_max_f16_e32 v3, v0, v0
+; GFX10-NEXT: v_max_f16_e32 v0, v0, v0
; GFX10-NEXT: s_and_b32 s4, s20, -4
-; GFX10-NEXT: v_mov_b32_e32 v2, s4
+; GFX10-NEXT: v_mov_b32_e32 v3, s4
; GFX10-NEXT: s_and_b32 s4, s20, 3
; GFX10-NEXT: s_lshl_b32 s4, s4, 3
; GFX10-NEXT: s_lshl_b32 s5, 0xffff, s4
-; GFX10-NEXT: buffer_load_dword v1, v2, s[16:19], 0 offen
+; GFX10-NEXT: buffer_load_dword v1, v3, s[16:19], 0 offen
; GFX10-NEXT: s_not_b32 s6, s5
; GFX10-NEXT: s_mov_b32 s5, 0
; GFX10-NEXT: .LBB11_1: ; %atomicrmw.start
; GFX10-NEXT: ; =>This Inner Loop Header: Depth=1
; GFX10-NEXT: s_waitcnt vmcnt(0)
-; GFX10-NEXT: v_lshrrev_b32_e32 v0, s4, v1
-; GFX10-NEXT: s_waitcnt_vscnt null, 0x0
-; GFX10-NEXT: v_max_f16_e32 v0, v0, v0
-; GFX10-NEXT: v_min_f16_e32 v0, v0, v3
-; GFX10-NEXT: v_lshlrev_b32_sdwa v0, s4, v0 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:WORD_0
-; GFX10-NEXT: v_and_or_b32 v0, v1, s6, v0
; GFX10-NEXT: v_mov_b32_e32 v5, v1
-; GFX10-NEXT: v_mov_b32_e32 v4, v0
-; GFX10-NEXT: buffer_atomic_cmpswap v[4:5], v2, s[16:19], 0 offen glc
+; GFX10-NEXT: s_waitcnt_vscnt null, 0x0
+; GFX10-NEXT: v_lshrrev_b32_e32 v1, s4, v5
+; GFX10-NEXT: v_max_f16_e32 v1, v1, v1
+; GFX10-NEXT: v_min_f16_e32 v1, v1, v0
+; GFX10-NEXT: v_lshlrev_b32_sdwa v1, s4, v1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:WORD_0
+; GFX10-NEXT: v_and_or_b32 v4, v5, s6, v1
+; GFX10-NEXT: v_mov_b32_e32 v1, v4
+; GFX10-NEXT: v_mov_b32_e32 v2, v5
+; GFX10-NEXT: buffer_atomic_cmpswap v[1:2], v3, s[16:19], 0 offen glc
; GFX10-NEXT: s_waitcnt vmcnt(0)
; GFX10-NEXT: buffer_gl1_inv
; GFX10-NEXT: buffer_gl0_inv
-; GFX10-NEXT: v_cmp_eq_u32_e32 vcc_lo, v4, v1
-; GFX10-NEXT: v_mov_b32_e32 v1, v4
+; GFX10-NEXT: v_cmp_eq_u32_e32 vcc_lo, v1, v5
; GFX10-NEXT: s_or_b32 s5, vcc_lo, s5
; GFX10-NEXT: s_andn2_b32 exec_lo, exec_lo, s5
; GFX10-NEXT: s_cbranch_execnz .LBB11_1
@@ -3210,29 +3220,29 @@ define void @buffer_fat_ptr_agent_atomic_fmin_noret_f16__offset__amdgpu_no_fine_
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX90A-NEXT: s_addk_i32 s20, 0x200
; GFX90A-NEXT: s_and_b32 s4, s20, -4
-; GFX90A-NEXT: v_mov_b32_e32 v2, s4
-; GFX90A-NEXT: buffer_load_dword v1, v2, s[16:19], 0 offen
+; GFX90A-NEXT: v_mov_b32_e32 v1, s4
+; GFX90A-NEXT: buffer_load_dword v2, v1, s[16:19], 0 offen
; GFX90A-NEXT: s_and_b32 s4, s20, 3
; GFX90A-NEXT: s_lshl_b32 s6, s4, 3
; GFX90A-NEXT: s_lshl_b32 s4, 0xffff, s6
; GFX90A-NEXT: s_not_b32 s7, s4
; GFX90A-NEXT: s_mov_b64 s[4:5], 0
-; GFX90A-NEXT: v_max_f16_e32 v3, v0, v0
+; GFX90A-NEXT: v_max_f16_e32 v0, v0, v0
; GFX90A-NEXT: .LBB11_1: ; %atomicrmw.start
; GFX90A-NEXT: ; =>This Inner Loop Header: Depth=1
; GFX90A-NEXT: s_waitcnt vmcnt(0)
-; GFX90A-NEXT: v_lshrrev_b32_e32 v0, s6, v1
-; GFX90A-NEXT: v_max_f16_e32 v0, v0, v0
-; GFX90A-NEXT: v_min_f16_e32 v0, v0, v3
-; GFX90A-NEXT: v_lshlrev_b32_e32 v0, s6, v0
-; GFX90A-NEXT: v_and_or_b32 v0, v1, s7, v0
-; GFX90A-NEXT: v_pk_mov_b32 v[4:5], v[0:1], v[0:1] op_sel:[0,1]
-; GFX90A-NEXT: buffer_atomic_cmpswap v[4:5], v2, s[16:19], 0 offen glc
+; GFX90A-NEXT: v_mov_b32_e32 v5, v2
+; GFX90A-NEXT: v_lshrrev_b32_e32 v2, s6, v5
+; GFX90A-NEXT: v_max_f16_e32 v2, v2, v2
+; GFX90A-NEXT: v_min_f16_e32 v2, v2, v0
+; GFX90A-NEXT: v_lshlrev_b32_e32 v2, s6, v2
+; GFX90A-NEXT: v_and_or_b32 v4, v5, s7, v2
+; GFX90A-NEXT: v_pk_mov_b32 v[2:3], v[4:5], v[4:5] op_sel:[0,1]
+; GFX90A-NEXT: buffer_atomic_cmpswap v[2:3], v1, s[16:19], 0 offen glc
; GFX90A-NEXT: s_waitcnt vmcnt(0)
; GFX90A-NEXT: buffer_wbinvl1
-; GFX90A-NEXT: v_cmp_eq_u32_e32 vcc, v4, v1
+; GFX90A-NEXT: v_cmp_eq_u32_e32 vcc, v2, v5
; GFX90A-NEXT: s_or_b64 s[4:5], vcc, s[4:5]
-; GFX90A-NEXT: v_mov_b32_e32 v1, v4
; GFX90A-NEXT: s_andn2_b64 exec, exec, s[4:5]
; GFX90A-NEXT: s_cbranch_execnz .LBB11_1
; GFX90A-NEXT: ; %bb.2: ; %atomicrmw.end
@@ -3244,30 +3254,30 @@ define void @buffer_fat_ptr_agent_atomic_fmin_noret_f16__offset__amdgpu_no_fine_
; GFX908-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX908-NEXT: s_addk_i32 s20, 0x200
; GFX908-NEXT: s_and_b32 s4, s20, -4
-; GFX908-NEXT: v_mov_b32_e32 v2, s4
-; GFX908-NEXT: buffer_load_dword v1, v2, s[16:19], 0 offen
+; GFX908-NEXT: v_mov_b32_e32 v3, s4
+; GFX908-NEXT: buffer_load_dword v1, v3, s[16:19], 0 offen
; GFX908-NEXT: s_and_b32 s4, s20, 3
; GFX908-NEXT: s_lshl_b32 s6, s4, 3
; GFX908-NEXT: s_lshl_b32 s4, 0xffff, s6
; GFX908-NEXT: s_not_b32 s7, s4
; GFX908-NEXT: s_mov_b64 s[4:5], 0
-; GFX908-NEXT: v_max_f16_e32 v3, v0, v0
+; GFX908-NEXT: v_max_f16_e32 v0, v0, v0
; GFX908-NEXT: .LBB11_1: ; %atomicrmw.start
; GFX908-NEXT: ; =>This Inner Loop Header: Depth=1
; GFX908-NEXT: s_waitcnt vmcnt(0)
-; GFX908-NEXT: v_lshrrev_b32_e32 v0, s6, v1
-; GFX908-NEXT: v_max_f16_e32 v0, v0, v0
-; GFX908-NEXT: v_min_f16_e32 v0, v0, v3
-; GFX908-NEXT: v_lshlrev_b32_e32 v0, s6, v0
-; GFX908-NEXT: v_and_or_b32 v0, v1, s7, v0
; GFX908-NEXT: v_mov_b32_e32 v5, v1
-; GFX908-NEXT: v_mov_b32_e32 v4, v0
-; GFX908-NEXT: buffer_atomic_cmpswap v[4:5], v2, s[16:19], 0 offen glc
+; GFX908-NEXT: v_lshrrev_b32_e32 v1, s6, v5
+; GFX908-NEXT: v_max_f16_e32 v1, v1, v1
+; GFX908-NEXT: v_min_f16_e32 v1, v1, v0
+; GFX908-NEXT: v_lshlrev_b32_e32 v1, s6, v1
+; GFX908-NEXT: v_and_or_b32 v4, v5, s7, v1
+; GFX908-NEXT: v_mov_b32_e32 v1, v4
+; GFX908-NEXT: v_mov_b32_e32 v2, v5
+; GFX908-NEXT: buffer_atomic_cmpswap v[1:2], v3, s[16:19], 0 offen glc
; GFX908-NEXT: s_waitcnt vmcnt(0)
; GFX908-NEXT: buffer_wbinvl1
-; GFX908-NEXT: v_cmp_eq_u32_e32 vcc, v4, v1
+; GFX908-NEXT: v_cmp_eq_u32_e32 vcc, v1, v5
; GFX908-NEXT: s_or_b64 s[4:5], vcc, s[4:5]
-; GFX908-NEXT: v_mov_b32_e32 v1, v4
; GFX908-NEXT: s_andn2_b64 exec, exec, s[4:5]
; GFX908-NEXT: s_cbranch_execnz .LBB11_1
; GFX908-NEXT: ; %bb.2: ; %atomicrmw.end
@@ -3279,31 +3289,31 @@ define void @buffer_fat_ptr_agent_atomic_fmin_noret_f16__offset__amdgpu_no_fine_
; GFX8-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX8-NEXT: s_addk_i32 s20, 0x200
; GFX8-NEXT: s_and_b32 s4, s20, -4
-; GFX8-NEXT: v_mov_b32_e32 v2, s4
-; GFX8-NEXT: buffer_load_dword v1, v2, s[16:19], 0 offen
+; GFX8-NEXT: v_mov_b32_e32 v3, s4
+; GFX8-NEXT: buffer_load_dword v1, v3, s[16:19], 0 offen
; GFX8-NEXT: s_and_b32 s4, s20, 3
; GFX8-NEXT: s_lshl_b32 s6, s4, 3
; GFX8-NEXT: s_lshl_b32 s4, 0xffff, s6
; GFX8-NEXT: s_not_b32 s7, s4
; GFX8-NEXT: s_mov_b64 s[4:5], 0
-; GFX8-NEXT: v_max_f16_e32 v3, v0, v0
+; GFX8-NEXT: v_max_f16_e32 v0, v0, v0
; GFX8-NEXT: .LBB11_1: ; %atomicrmw.start
; GFX8-NEXT: ; =>This Inner Loop Header: Depth=1
; GFX8-NEXT: s_waitcnt vmcnt(0)
-; GFX8-NEXT: v_lshrrev_b32_e32 v0, s6, v1
-; GFX8-NEXT: v_max_f16_e32 v0, v0, v0
-; GFX8-NEXT: v_min_f16_e32 v0, v0, v3
-; GFX8-NEXT: v_and_b32_e32 v4, s7, v1
-; GFX8-NEXT: v_lshlrev_b32_e32 v0, s6, v0
-; GFX8-NEXT: v_or_b32_e32 v0, v4, v0
; GFX8-NEXT: v_mov_b32_e32 v5, v1
-; GFX8-NEXT: v_mov_b32_e32 v4, v0
-; GFX8-NEXT: buffer_atomic_cmpswap v[4:5], v2, s[16:19], 0 offen glc
+; GFX8-NEXT: v_lshrrev_b32_e32 v1, s6, v5
+; GFX8-NEXT: v_max_f16_e32 v1, v1, v1
+; GFX8-NEXT: v_min_f16_e32 v1, v1, v0
+; GFX8-NEXT: v_and_b32_e32 v2, s7, v5
+; GFX8-NEXT: v_lshlrev_b32_e32 v1, s6, v1
+; GFX8-NEXT: v_or_b32_e32 v4, v2, v1
+; GFX8-NEXT: v_mov_b32_e32 v1, v4
+; GFX8-NEXT: v_mov_b32_e32 v2, v5
+; GFX8-NEXT: buffer_atomic_cmpswap v[1:2], v3, s[16:19], 0 offen glc
; GFX8-NEXT: s_waitcnt vmcnt(0)
; GFX8-NEXT: buffer_wbinvl1
-; GFX8-NEXT: v_cmp_eq_u32_e32 vcc, v4, v1
+; GFX8-NEXT: v_cmp_eq_u32_e32 vcc, v1, v5
; GFX8-NEXT: s_or_b64 s[4:5], vcc, s[4:5]
-; GFX8-NEXT: v_mov_b32_e32 v1, v4
; GFX8-NEXT: s_andn2_b64 exec, exec, s[4:5]
; GFX8-NEXT: s_cbranch_execnz .LBB11_1
; GFX8-NEXT: ; %bb.2: ; %atomicrmw.end
@@ -3315,33 +3325,33 @@ define void @buffer_fat_ptr_agent_atomic_fmin_noret_f16__offset__amdgpu_no_fine_
; GFX7-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX7-NEXT: s_addk_i32 s20, 0x200
; GFX7-NEXT: s_and_b32 s4, s20, -4
-; GFX7-NEXT: v_mov_b32_e32 v2, s4
-; GFX7-NEXT: buffer_load_dword v1, v2, s[16:19], 0 offen
+; GFX7-NEXT: v_mov_b32_e32 v3, s4
+; GFX7-NEXT: buffer_load_dword v1, v3, s[16:19], 0 offen
; GFX7-NEXT: v_cvt_f16_f32_e32 v0, v0
; GFX7-NEXT: s_and_b32 s4, s20, 3
; GFX7-NEXT: s_lshl_b32 s6, s4, 3
; GFX7-NEXT: s_lshl_b32 s4, 0xffff, s6
-; GFX7-NEXT: v_cvt_f32_f16_e32 v3, v0
+; GFX7-NEXT: v_cvt_f32_f16_e32 v0, v0
; GFX7-NEXT: s_not_b32 s7, s4
; GFX7-NEXT: s_mov_b64 s[4:5], 0
; GFX7-NEXT: .LBB11_1: ; %atomicrmw.start
; GFX7-NEXT: ; =>This Inner Loop Header: Depth=1
; GFX7-NEXT: s_waitcnt vmcnt(0)
-; GFX7-NEXT: v_lshrrev_b32_e32 v0, s6, v1
-; GFX7-NEXT: v_cvt_f32_f16_e32 v0, v0
-; GFX7-NEXT: v_and_b32_e32 v4, s7, v1
-; GFX7-NEXT: v_min_f32_e32 v0, v0, v3
-; GFX7-NEXT: v_cvt_f16_f32_e32 v0, v0
-; GFX7-NEXT: v_lshlrev_b32_e32 v0, s6, v0
-; GFX7-NEXT: v_or_b32_e32 v0, v4, v0
; GFX7-NEXT: v_mov_b32_e32 v5, v1
-; GFX7-NEXT: v_mov_b32_e32 v4, v0
-; GFX7-NEXT: buffer_atomic_cmpswap v[4:5], v2, s[16:19], 0 offen glc
+; GFX7-NEXT: v_lshrrev_b32_e32 v1, s6, v5
+; GFX7-NEXT: v_cvt_f32_f16_e32 v1, v1
+; GFX7-NEXT: v_and_b32_e32 v2, s7, v5
+; GFX7-NEXT: v_min_f32_e32 v1, v1, v0
+; GFX7-NEXT: v_cvt_f16_f32_e32 v1, v1
+; GFX7-NEXT: v_lshlrev_b32_e32 v1, s6, v1
+; GFX7-NEXT: v_or_b32_e32 v4, v2, v1
+; GFX7-NEXT: v_mov_b32_e32 v1, v4
+; GFX7-NEXT: v_mov_b32_e32 v2, v5
+; GFX7-NEXT: buffer_atomic_cmpswap v[1:2], v3, s[16:19], 0 offen glc
; GFX7-NEXT: s_waitcnt vmcnt(0)
; GFX7-NEXT: buffer_wbinvl1
-; GFX7-NEXT: v_cmp_eq_u32_e32 vcc, v4, v1
+; GFX7-NEXT: v_cmp_eq_u32_e32 vcc, v1, v5
; GFX7-NEXT: s_or_b64 s[4:5], vcc, s[4:5]
-; GFX7-NEXT: v_mov_b32_e32 v1, v4
; GFX7-NEXT: s_andn2_b64 exec, exec, s[4:5]
; GFX7-NEXT: s_cbranch_execnz .LBB11_1
; GFX7-NEXT: ; %bb.2: ; %atomicrmw.end
@@ -3353,34 +3363,34 @@ define void @buffer_fat_ptr_agent_atomic_fmin_noret_f16__offset__amdgpu_no_fine_
; GFX6-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX6-NEXT: s_addk_i32 s20, 0x200
; GFX6-NEXT: s_and_b32 s4, s20, -4
-; GFX6-NEXT: v_mov_b32_e32 v2, s4
-; GFX6-NEXT: buffer_load_dword v1, v2, s[16:19], 0 offen
+; GFX6-NEXT: v_mov_b32_e32 v3, s4
+; GFX6-NEXT: buffer_load_dword v1, v3, s[16:19], 0 offen
; GFX6-NEXT: v_cvt_f16_f32_e32 v0, v0
; GFX6-NEXT: s_and_b32 s4, s20, 3
; GFX6-NEXT: s_lshl_b32 s6, s4, 3
; GFX6-NEXT: s_lshl_b32 s4, 0xffff, s6
-; GFX6-NEXT: v_cvt_f32_f16_e32 v3, v0
+; GFX6-NEXT: v_cvt_f32_f16_e32 v0, v0
; GFX6-NEXT: s_not_b32 s7, s4
; GFX6-NEXT: s_mov_b64 s[4:5], 0
; GFX6-NEXT: .LBB11_1: ; %atomicrmw.start
; GFX6-NEXT: ; =>This Inner Loop Header: Depth=1
; GFX6-NEXT: s_waitcnt vmcnt(0)
-; GFX6-NEXT: v_lshrrev_b32_e32 v0, s6, v1
-; GFX6-NEXT: v_cvt_f32_f16_e32 v0, v0
-; GFX6-NEXT: s_waitcnt expcnt(0)
-; GFX6-NEXT: v_and_b32_e32 v4, s7, v1
-; GFX6-NEXT: v_min_f32_e32 v0, v0, v3
-; GFX6-NEXT: v_cvt_f16_f32_e32 v0, v0
-; GFX6-NEXT: v_lshlrev_b32_e32 v0, s6, v0
-; GFX6-NEXT: v_or_b32_e32 v0, v4, v0
; GFX6-NEXT: v_mov_b32_e32 v5, v1
-; GFX6-NEXT: v_mov_b32_e32 v4, v0
-; GFX6-NEXT: buffer_atomic_cmpswap v[4:5], v2, s[16:19], 0 offen glc
+; GFX6-NEXT: s_waitcnt expcnt(0)
+; GFX6-NEXT: v_lshrrev_b32_e32 v1, s6, v5
+; GFX6-NEXT: v_cvt_f32_f16_e32 v1, v1
+; GFX6-NEXT: v_and_b32_e32 v2, s7, v5
+; GFX6-NEXT: v_min_f32_e32 v1, v1, v0
+; GFX6-NEXT: v_cvt_f16_f32_e32 v1, v1
+; GFX6-NEXT: v_lshlrev_b32_e32 v1, s6, v1
+; GFX6-NEXT: v_or_b32_e32 v4, v2, v1
+; GFX6-NEXT: v_mov_b32_e32 v1, v4
+; GFX6-NEXT: v_mov_b32_e32 v2, v5
+; GFX6-NEXT: buffer_atomic_cmpswap v[1:2], v3, s[16:19], 0 offen glc
; GFX6-NEXT: s_waitcnt vmcnt(0)
; GFX6-NEXT: buffer_wbinvl1
-; GFX6-NEXT: v_cmp_eq_u32_e32 vcc, v4, v1
+; GFX6-NEXT: v_cmp_eq_u32_e32 vcc, v1, v5
; GFX6-NEXT: s_or_b64 s[4:5], vcc, s[4:5]
-; GFX6-NEXT: v_mov_b32_e32 v1, v4
; GFX6-NEXT: s_andn2_b64 exec, exec, s[4:5]
; GFX6-NEXT: s_cbranch_execnz .LBB11_1
; GFX6-NEXT: ; %bb.2: ; %atomicrmw.end
@@ -3432,18 +3442,21 @@ define half @buffer_fat_ptr_agent_atomic_fmin_ret_f16__offset__waterfall__amdgpu
; GFX12-TRUE16-NEXT: ; =>This Loop Header: Depth=1
; GFX12-TRUE16-NEXT: ; Child Loop BB12_4 Depth 2
; GFX12-TRUE16-NEXT: s_wait_loadcnt 0x0
-; GFX12-TRUE16-NEXT: v_lshrrev_b32_e32 v5, v9, v6
+; GFX12-TRUE16-NEXT: v_mov_b32_e32 v8, v6
; GFX12-TRUE16-NEXT: s_mov_b32 s2, exec_lo
; GFX12-TRUE16-NEXT: s_wait_storecnt 0x0
; GFX12-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
+; GFX12-TRUE16-NEXT: v_lshrrev_b32_e32 v5, v9, v8
; GFX12-TRUE16-NEXT: v_max_num_f16_e32 v4.h, v5.l, v5.l
-; GFX12-TRUE16-NEXT: v_min_num_f16_e32 v5.l, v4.h, v4.l
; GFX12-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
+; GFX12-TRUE16-NEXT: v_min_num_f16_e32 v5.l, v4.h, v4.l
; GFX12-TRUE16-NEXT: v_and_b32_e32 v5, 0xffff, v5
-; GFX12-TRUE16-NEXT: v_lshlrev_b32_e32 v5, v9, v5
; GFX12-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
-; GFX12-TRUE16-NEXT: v_and_or_b32 v5, v6, v11, v5
-; GFX12-TRUE16-NEXT: v_dual_mov_b32 v8, v6 :: v_dual_mov_b32 v7, v5
+; GFX12-TRUE16-NEXT: v_lshlrev_b32_e32 v5, v9, v5
+; GFX12-TRUE16-NEXT: v_and_or_b32 v7, v8, v11, v5
+; GFX12-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1)
+; GFX12-TRUE16-NEXT: v_mov_b32_e32 v6, v7
+; GFX12-TRUE16-NEXT: v_mov_b32_e32 v7, v8
; GFX12-TRUE16-NEXT: .LBB12_4: ; Parent Loop BB12_3 Depth=1
; GFX12-TRUE16-NEXT: ; => This Inner Loop Header: Depth=2
; GFX12-TRUE16-NEXT: v_readfirstlane_b32 s4, v0
@@ -3457,14 +3470,13 @@ define half @buffer_fat_ptr_agent_atomic_fmin_ret_f16__offset__waterfall__amdgpu
; GFX12-TRUE16-NEXT: s_wait_alu 0xfffe
; GFX12-TRUE16-NEXT: s_and_saveexec_b32 s0, s0
; GFX12-TRUE16-NEXT: s_wait_loadcnt 0x0
-; GFX12-TRUE16-NEXT: buffer_atomic_cmpswap_b32 v[7:8], v10, s[4:7], null offen th:TH_ATOMIC_RETURN
+; GFX12-TRUE16-NEXT: buffer_atomic_cmpswap_b32 v[6:7], v10, s[4:7], null offen th:TH_ATOMIC_RETURN
; GFX12-TRUE16-NEXT: s_xor_b32 exec_lo, exec_lo, s0
; GFX12-TRUE16-NEXT: s_cbranch_execnz .LBB12_4
; GFX12-TRUE16-NEXT: ; %bb.5: ; in Loop: Header=BB12_3 Depth=1
; GFX12-TRUE16-NEXT: s_mov_b32 exec_lo, s2
; GFX12-TRUE16-NEXT: s_wait_loadcnt 0x0
-; GFX12-TRUE16-NEXT: v_cmp_eq_u32_e32 vcc_lo, v7, v6
-; GFX12-TRUE16-NEXT: v_mov_b32_e32 v6, v7
+; GFX12-TRUE16-NEXT: v_cmp_eq_u32_e32 vcc_lo, v6, v8
; GFX12-TRUE16-NEXT: global_inv scope:SCOPE_DEV
; GFX12-TRUE16-NEXT: s_or_b32 s1, vcc_lo, s1
; GFX12-TRUE16-NEXT: s_wait_alu 0xfffe
@@ -3472,7 +3484,7 @@ define half @buffer_fat_ptr_agent_atomic_fmin_ret_f16__offset__waterfall__amdgpu
; GFX12-TRUE16-NEXT: s_cbranch_execnz .LBB12_3
; GFX12-TRUE16-NEXT: ; %bb.6: ; %atomicrmw.end
; GFX12-TRUE16-NEXT: s_or_b32 exec_lo, exec_lo, s1
-; GFX12-TRUE16-NEXT: v_lshrrev_b32_e32 v0, v9, v7
+; GFX12-TRUE16-NEXT: v_lshrrev_b32_e32 v0, v9, v6
; GFX12-TRUE16-NEXT: s_wait_loadcnt 0x0
; GFX12-TRUE16-NEXT: s_setpc_b64 s[30:31]
;
@@ -3504,7 +3516,7 @@ define half @buffer_fat_ptr_agent_atomic_fmin_ret_f16__offset__waterfall__amdgpu
; GFX12-FAKE16-NEXT: s_wait_alu 0xfffe
; GFX12-FAKE16-NEXT: s_and_saveexec_b32 s0, s0
; GFX12-FAKE16-NEXT: s_wait_loadcnt 0x0
-; GFX12-FAKE16-NEXT: buffer_load_b32 v6, v8, s[4:7], null offen
+; GFX12-FAKE16-NEXT: buffer_load_b32 v4, v8, s[4:7], null offen
; GFX12-FAKE16-NEXT: s_xor_b32 exec_lo, exec_lo, s0
; GFX12-FAKE16-NEXT: s_cbranch_execnz .LBB12_1
; GFX12-FAKE16-NEXT: ; %bb.2:
@@ -3515,17 +3527,19 @@ define half @buffer_fat_ptr_agent_atomic_fmin_ret_f16__offset__waterfall__amdgpu
; GFX12-FAKE16-NEXT: ; =>This Loop Header: Depth=1
; GFX12-FAKE16-NEXT: ; Child Loop BB12_4 Depth 2
; GFX12-FAKE16-NEXT: s_wait_loadcnt 0x0
-; GFX12-FAKE16-NEXT: v_lshrrev_b32_e32 v4, v7, v6
+; GFX12-FAKE16-NEXT: v_mov_b32_e32 v6, v4
; GFX12-FAKE16-NEXT: s_mov_b32 s2, exec_lo
; GFX12-FAKE16-NEXT: s_wait_storecnt 0x0
; GFX12-FAKE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
+; GFX12-FAKE16-NEXT: v_lshrrev_b32_e32 v4, v7, v6
; GFX12-FAKE16-NEXT: v_max_num_f16_e32 v4, v4, v4
-; GFX12-FAKE16-NEXT: v_min_num_f16_e32 v4, v4, v10
; GFX12-FAKE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
+; GFX12-FAKE16-NEXT: v_min_num_f16_e32 v4, v4, v10
; GFX12-FAKE16-NEXT: v_and_b32_e32 v4, 0xffff, v4
-; GFX12-FAKE16-NEXT: v_lshlrev_b32_e32 v4, v7, v4
; GFX12-FAKE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
+; GFX12-FAKE16-NEXT: v_lshlrev_b32_e32 v4, v7, v4
; GFX12-FAKE16-NEXT: v_and_or_b32 v5, v6, v9, v4
+; GFX12-FAKE16-NEXT: s_delay_alu instid0(VALU_DEP_1)
; GFX12-FAKE16-NEXT: v_mov_b32_e32 v4, v5
; GFX12-FAKE16-NEXT: v_mov_b32_e32 v5, v6
; GFX12-FAKE16-NEXT: .LBB12_4: ; Parent Loop BB12_3 Depth=1
@@ -3548,7 +3562,6 @@ define half @buffer_fat_ptr_agent_atomic_fmin_ret_f16__offset__waterfall__amdgpu
; GFX12-FAKE16-NEXT: s_mov_b32 exec_lo, s2
; GFX12-FAKE16-NEXT: s_wait_loadcnt 0x0
; GFX12-FAKE16-NEXT: v_cmp_eq_u32_e32 vcc_lo, v4, v6
-; GFX12-FAKE16-NEXT: v_mov_b32_e32 v6, v4
; GFX12-FAKE16-NEXT: global_inv scope:SCOPE_DEV
; GFX12-FAKE16-NEXT: s_or_b32 s1, vcc_lo, s1
; GFX12-FAKE16-NEXT: s_wait_alu 0xfffe
@@ -3581,7 +3594,7 @@ define half @buffer_fat_ptr_agent_atomic_fmin_ret_f16__offset__waterfall__amdgpu
; GFX942-NEXT: v_cmp_eq_u64_e64 s[0:1], s[6:7], v[2:3]
; GFX942-NEXT: s_and_b64 s[0:1], vcc, s[0:1]
; GFX942-NEXT: s_and_saveexec_b64 s[0:1], s[0:1]
-; GFX942-NEXT: buffer_load_dword v7, v9, s[4:7], 0 offen
+; GFX942-NEXT: buffer_load_dword v4, v9, s[4:7], 0 offen
; GFX942-NEXT: s_xor_b64 exec, exec, s[0:1]
; GFX942-NEXT: s_cbranch_execnz .LBB12_1
; GFX942-NEXT: ; %bb.2:
@@ -3592,6 +3605,7 @@ define half @buffer_fat_ptr_agent_atomic_fmin_ret_f16__offset__waterfall__amdgpu
; GFX942-NEXT: ; =>This Loop Header: Depth=1
; GFX942-NEXT: ; Child Loop BB12_4 Depth 2
; GFX942-NEXT: s_waitcnt vmcnt(0)
+; GFX942-NEXT: v_mov_b32_e32 v7, v4
; GFX942-NEXT: v_lshrrev_b32_e32 v4, v8, v7
; GFX942-NEXT: v_max_f16_e32 v4, v4, v4
; GFX942-NEXT: v_min_f16_e32 v4, v4, v11
@@ -3620,7 +3634,6 @@ define half @buffer_fat_ptr_agent_atomic_fmin_ret_f16__offset__waterfall__amdgpu
; GFX942-NEXT: s_waitcnt vmcnt(0)
; GFX942-NEXT: v_cmp_eq_u32_e32 vcc, v4, v7
; GFX942-NEXT: s_or_b64 s[2:3], vcc, s[2:3]
-; GFX942-NEXT: v_mov_b32_e32 v7, v4
; GFX942-NEXT: buffer_inv sc1
; GFX942-NEXT: s_andn2_b64 exec, exec, s[2:3]
; GFX942-NEXT: s_cbranch_execnz .LBB12_3
@@ -3664,18 +3677,21 @@ define half @buffer_fat_ptr_agent_atomic_fmin_ret_f16__offset__waterfall__amdgpu
; GFX11-TRUE16-NEXT: ; =>This Loop Header: Depth=1
; GFX11-TRUE16-NEXT: ; Child Loop BB12_4 Depth 2
; GFX11-TRUE16-NEXT: s_waitcnt vmcnt(0)
-; GFX11-TRUE16-NEXT: v_lshrrev_b32_e32 v5, v9, v6
+; GFX11-TRUE16-NEXT: v_mov_b32_e32 v8, v6
; GFX11-TRUE16-NEXT: s_mov_b32 s2, exec_lo
; GFX11-TRUE16-NEXT: s_waitcnt_vscnt null, 0x0
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
+; GFX11-TRUE16-NEXT: v_lshrrev_b32_e32 v5, v9, v8
; GFX11-TRUE16-NEXT: v_max_f16_e32 v4.h, v5.l, v5.l
-; GFX11-TRUE16-NEXT: v_min_f16_e32 v5.l, v4.h, v4.l
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
+; GFX11-TRUE16-NEXT: v_min_f16_e32 v5.l, v4.h, v4.l
; GFX11-TRUE16-NEXT: v_and_b32_e32 v5, 0xffff, v5
-; GFX11-TRUE16-NEXT: v_lshlrev_b32_e32 v5, v9, v5
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
-; GFX11-TRUE16-NEXT: v_and_or_b32 v5, v6, v11, v5
-; GFX11-TRUE16-NEXT: v_dual_mov_b32 v8, v6 :: v_dual_mov_b32 v7, v5
+; GFX11-TRUE16-NEXT: v_lshlrev_b32_e32 v5, v9, v5
+; GFX11-TRUE16-NEXT: v_and_or_b32 v7, v8, v11, v5
+; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1)
+; GFX11-TRUE16-NEXT: v_mov_b32_e32 v6, v7
+; GFX11-TRUE16-NEXT: v_mov_b32_e32 v7, v8
; GFX11-TRUE16-NEXT: .LBB12_4: ; Parent Loop BB12_3 Depth=1
; GFX11-TRUE16-NEXT: ; => This Inner Loop Header: Depth=2
; GFX11-TRUE16-NEXT: v_readfirstlane_b32 s4, v0
@@ -3689,14 +3705,13 @@ define half @buffer_fat_ptr_agent_atomic_fmin_ret_f16__offset__waterfall__amdgpu
; GFX11-TRUE16-NEXT: s_delay_alu instid0(SALU_CYCLE_1)
; GFX11-TRUE16-NEXT: s_and_saveexec_b32 s0, s0
; GFX11-TRUE16-NEXT: s_waitcnt vmcnt(0)
-; GFX11-TRUE16-NEXT: buffer_atomic_cmpswap_b32 v[7:8], v10, s[4:7], 0 offen glc
+; GFX11-TRUE16-NEXT: buffer_atomic_cmpswap_b32 v[6:7], v10, s[4:7], 0 offen glc
; GFX11-TRUE16-NEXT: s_xor_b32 exec_lo, exec_lo, s0
; GFX11-TRUE16-NEXT: s_cbranch_execnz .LBB12_4
; GFX11-TRUE16-NEXT: ; %bb.5: ; in Loop: Header=BB12_3 Depth=1
; GFX11-TRUE16-NEXT: s_mov_b32 exec_lo, s2
; GFX11-TRUE16-NEXT: s_waitcnt vmcnt(0)
-; GFX11-TRUE16-NEXT: v_cmp_eq_u32_e32 vcc_lo, v7, v6
-; GFX11-TRUE16-NEXT: v_mov_b32_e32 v6, v7
+; GFX11-TRUE16-NEXT: v_cmp_eq_u32_e32 vcc_lo, v6, v8
; GFX11-TRUE16-NEXT: buffer_gl1_inv
; GFX11-TRUE16-NEXT: buffer_gl0_inv
; GFX11-TRUE16-NEXT: s_or_b32 s1, vcc_lo, s1
@@ -3705,7 +3720,7 @@ define half @buffer_fat_ptr_agent_atomic_fmin_ret_f16__offset__waterfall__amdgpu
; GFX11-TRUE16-NEXT: s_cbranch_execnz .LBB12_3
; GFX11-TRUE16-NEXT: ; %bb.6: ; %atomicrmw.end
; GFX11-TRUE16-NEXT: s_or_b32 exec_lo, exec_lo, s1
-; GFX11-TRUE16-NEXT: v_lshrrev_b32_e32 v0, v9, v7
+; GFX11-TRUE16-NEXT: v_lshrrev_b32_e32 v0, v9, v6
; GFX11-TRUE16-NEXT: s_setpc_b64 s[30:31]
;
; GFX11-FAKE16-LABEL: buffer_fat_ptr_agent_atomic_fmin_ret_f16__offset__waterfall__amdgpu_no_fine_grained_memory:
@@ -3732,7 +3747,7 @@ define half @buffer_fat_ptr_agent_atomic_fmin_ret_f16__offset__waterfall__amdgpu
; GFX11-FAKE16-NEXT: s_and_b32 s0, vcc_lo, s0
; GFX11-FAKE16-NEXT: s_delay_alu instid0(SALU_CYCLE_1)
; GFX11-FAKE16-NEXT: s_and_saveexec_b32 s0, s0
-; GFX11-FAKE16-NEXT: buffer_load_b32 v6, v8, s[4:7], 0 offen
+; GFX11-FAKE16-NEXT: buffer_load_b32 v4, v8, s[4:7], 0 offen
; GFX11-FAKE16-NEXT: s_xor_b32 exec_lo, exec_lo, s0
; GFX11-FAKE16-NEXT: s_cbranch_execnz .LBB12_1
; GFX11-FAKE16-NEXT: ; %bb.2:
@@ -3743,17 +3758,19 @@ define half @buffer_fat_ptr_agent_atomic_fmin_ret_f16__offset__waterfall__amdgpu
; GFX11-FAKE16-NEXT: ; =>This Loop Header: Depth=1
; GFX11-FAKE16-NEXT: ; Child Loop BB12_4 Depth 2
; GFX11-FAKE16-NEXT: s_waitcnt vmcnt(0)
-; GFX11-FAKE16-NEXT: v_lshrrev_b32_e32 v4, v7, v6
+; GFX11-FAKE16-NEXT: v_mov_b32_e32 v6, v4
; GFX11-FAKE16-NEXT: s_mov_b32 s2, exec_lo
; GFX11-FAKE16-NEXT: s_waitcnt_vscnt null, 0x0
; GFX11-FAKE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
+; GFX11-FAKE16-NEXT: v_lshrrev_b32_e32 v4, v7, v6
; GFX11-FAKE16-NEXT: v_max_f16_e32 v4, v4, v4
-; GFX11-FAKE16-NEXT: v_min_f16_e32 v4, v4, v10
; GFX11-FAKE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
+; GFX11-FAKE16-NEXT: v_min_f16_e32 v4, v4, v10
; GFX11-FAKE16-NEXT: v_and_b32_e32 v4, 0xffff, v4
-; GFX11-FAKE16-NEXT: v_lshlrev_b32_e32 v4, v7, v4
; GFX11-FAKE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
+; GFX11-FAKE16-NEXT: v_lshlrev_b32_e32 v4, v7, v4
; GFX11-FAKE16-NEXT: v_and_or_b32 v5, v6, v9, v4
+; GFX11-FAKE16-NEXT: s_delay_alu instid0(VALU_DEP_1)
; GFX11-FAKE16-NEXT: v_mov_b32_e32 v4, v5
; GFX11-FAKE16-NEXT: v_mov_b32_e32 v5, v6
; GFX11-FAKE16-NEXT: .LBB12_4: ; Parent Loop BB12_3 Depth=1
@@ -3776,7 +3793,6 @@ define half @buffer_fat_ptr_agent_atomic_fmin_ret_f16__offset__waterfall__amdgpu
; GFX11-FAKE16-NEXT: s_mov_b32 exec_lo, s2
; GFX11-FAKE16-NEXT: s_waitcnt vmcnt(0)
; GFX11-FAKE16-NEXT: v_cmp_eq_u32_e32 vcc_lo, v4, v6
-; GFX11-FAKE16-NEXT: v_mov_b32_e32 v6, v4
; GFX11-FAKE16-NEXT: buffer_gl1_inv
; GFX11-FAKE16-NEXT: buffer_gl0_inv
; GFX11-FAKE16-NEXT: s_or_b32 s1, vcc_lo, s1
@@ -3808,7 +3824,7 @@ define half @buffer_fat_ptr_agent_atomic_fmin_ret_f16__offset__waterfall__amdgpu
; GFX10-NEXT: v_cmp_eq_u64_e64 s4, s[10:11], v[2:3]
; GFX10-NEXT: s_and_b32 s4, vcc_lo, s4
; GFX10-NEXT: s_and_saveexec_b32 s4, s4
-; GFX10-NEXT: buffer_load_dword v6, v8, s[8:11], 0 offen
+; GFX10-NEXT: buffer_load_dword v4, v8, s[8:11], 0 offen
; GFX10-NEXT: s_waitcnt_depctr 0xffe3
; GFX10-NEXT: s_xor_b32 exec_lo, exec_lo, s4
; GFX10-NEXT: s_cbranch_execnz .LBB12_1
@@ -3819,9 +3835,10 @@ define half @buffer_fat_ptr_agent_atomic_fmin_ret_f16__offset__waterfall__amdgpu
; GFX10-NEXT: ; =>This Loop Header: Depth=1
; GFX10-NEXT: ; Child Loop BB12_4 Depth 2
; GFX10-NEXT: s_waitcnt vmcnt(0)
-; GFX10-NEXT: v_lshrrev_b32_e32 v4, v7, v6
+; GFX10-NEXT: v_mov_b32_e32 v6, v4
; GFX10-NEXT: s_mov_b32 s6, exec_lo
; GFX10-NEXT: s_waitcnt_vscnt null, 0x0
+; GFX10-NEXT: v_lshrrev_b32_e32 v4, v7, v6
; GFX10-NEXT: v_max_f16_e32 v4, v4, v4
; GFX10-NEXT: v_min_f16_e32 v4, v4, v10
; GFX10-NEXT: v_lshlrev_b32_sdwa v4, v7, v4 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:WORD_0
@@ -3847,7 +3864,6 @@ define half @buffer_fat_ptr_agent_atomic_fmin_ret_f16__offset__waterfall__amdgpu
; GFX10-NEXT: s_mov_b32 exec_lo, s6
; GFX10-NEXT: s_waitcnt vmcnt(0)
; GFX10-NEXT: v_cmp_eq_u32_e32 vcc_lo, v4, v6
-; GFX10-NEXT: v_mov_b32_e32 v6, v4
; GFX10-NEXT: buffer_gl1_inv
; GFX10-NEXT: buffer_gl0_inv
; GFX10-NEXT: s_or_b32 s5, vcc_lo, s5
@@ -3880,7 +3896,7 @@ define half @buffer_fat_ptr_agent_atomic_fmin_ret_f16__offset__waterfall__amdgpu
; GFX90A-NEXT: s_and_b64 s[4:5], vcc, s[4:5]
; GFX90A-NEXT: s_and_saveexec_b64 s[4:5], s[4:5]
; GFX90A-NEXT: s_nop 0
-; GFX90A-NEXT: buffer_load_dword v7, v9, s[8:11], 0 offen
+; GFX90A-NEXT: buffer_load_dword v4, v9, s[8:11], 0 offen
; GFX90A-NEXT: s_xor_b64 exec, exec, s[4:5]
; GFX90A-NEXT: s_cbranch_execnz .LBB12_1
; GFX90A-NEXT: ; %bb.2:
@@ -3891,6 +3907,7 @@ define half @buffer_fat_ptr_agent_atomic_fmin_ret_f16__offset__waterfall__amdgpu
; GFX90A-NEXT: ; =>This Loop Header: Depth=1
; GFX90A-NEXT: ; Child Loop BB12_4 Depth 2
; GFX90A-NEXT: s_waitcnt vmcnt(0)
+; GFX90A-NEXT: v_mov_b32_e32 v7, v4
; GFX90A-NEXT: v_lshrrev_b32_e32 v4, v8, v7
; GFX90A-NEXT: v_max_f16_e32 v4, v4, v4
; GFX90A-NEXT: v_min_f16_e32 v4, v4, v11
@@ -3917,7 +3934,6 @@ define half @buffer_fat_ptr_agent_atomic_fmin_ret_f16__offset__waterfall__amdgpu
; GFX90A-NEXT: s_waitcnt vmcnt(0)
; GFX90A-NEXT: v_cmp_eq_u32_e32 vcc, v4, v7
; GFX90A-NEXT: s_or_b64 s[6:7], vcc, s[6:7]
-; GFX90A-NEXT: v_mov_b32_e32 v7, v4
; GFX90A-NEXT: buffer_wbinvl1
; GFX90A-NEXT: s_andn2_b64 exec, exec, s[6:7]
; GFX90A-NEXT: s_cbranch_execnz .LBB12_3
@@ -3947,7 +3963,7 @@ define half @buffer_fat_ptr_agent_atomic_fmin_ret_f16__offset__waterfall__amdgpu
; GFX908-NEXT: s_and_b64 s[4:5], vcc, s[4:5]
; GFX908-NEXT: s_and_saveexec_b64 s[4:5], s[4:5]
; GFX908-NEXT: s_nop 0
-; GFX908-NEXT: buffer_load_dword v6, v8, s[8:11], 0 offen
+; GFX908-NEXT: buffer_load_dword v4, v8, s[8:11], 0 offen
; GFX908-NEXT: s_xor_b64 exec, exec, s[4:5]
; GFX908-NEXT: s_cbranch_execnz .LBB12_1
; GFX908-NEXT: ; %bb.2:
@@ -3958,6 +3974,7 @@ define half @buffer_fat_ptr_agent_atomic_fmin_ret_f16__offset__waterfall__amdgpu
; GFX908-NEXT: ; =>This Loop Header: Depth=1
; GFX908-NEXT: ; Child Loop BB12_4 Depth 2
; GFX908-NEXT: s_waitcnt vmcnt(0)
+; GFX908-NEXT: v_mov_b32_e32 v6, v4
; GFX908-NEXT: v_lshrrev_b32_e32 v4, v7, v6
; GFX908-NEXT: v_max_f16_e32 v4, v4, v4
; GFX908-NEXT: v_min_f16_e32 v4, v4, v10
@@ -3985,7 +4002,6 @@ define half @buffer_fat_ptr_agent_atomic_fmin_ret_f16__offset__waterfall__amdgpu
; GFX908-NEXT: s_waitcnt vmcnt(0)
; GFX908-NEXT: v_cmp_eq_u32_e32 vcc, v4, v6
; GFX908-NEXT: s_or_b64 s[6:7], vcc, s[6:7]
-; GFX908-NEXT: v_mov_b32_e32 v6, v4
; GFX908-NEXT: buffer_wbinvl1
; GFX908-NEXT: s_andn2_b64 exec, exec, s[6:7]
; GFX908-NEXT: s_cbranch_execnz .LBB12_3
@@ -4015,7 +4031,7 @@ define half @buffer_fat_ptr_agent_atomic_fmin_ret_f16__offset__waterfall__amdgpu
; GFX8-NEXT: s_and_b64 s[4:5], vcc, s[4:5]
; GFX8-NEXT: s_and_saveexec_b64 s[4:5], s[4:5]
; GFX8-NEXT: s_nop 0
-; GFX8-NEXT: buffer_load_dword v6, v8, s[8:11], 0 offen
+; GFX8-NEXT: buffer_load_dword v4, v8, s[8:11], 0 offen
; GFX8-NEXT: s_xor_b64 exec, exec, s[4:5]
; GFX8-NEXT: s_cbranch_execnz .LBB12_1
; GFX8-NEXT: ; %bb.2:
@@ -4026,6 +4042,7 @@ define half @buffer_fat_ptr_agent_atomic_fmin_ret_f16__offset__waterfall__amdgpu
; GFX8-NEXT: ; =>This Loop Header: Depth=1
; GFX8-NEXT: ; Child Loop BB12_4 Depth 2
; GFX8-NEXT: s_waitcnt vmcnt(0)
+; GFX8-NEXT: v_mov_b32_e32 v6, v4
; GFX8-NEXT: v_lshrrev_b32_e32 v4, v7, v6
; GFX8-NEXT: v_max_f16_e32 v4, v4, v4
; GFX8-NEXT: v_min_f16_e32 v4, v4, v10
@@ -4054,7 +4071,6 @@ define half @buffer_fat_ptr_agent_atomic_fmin_ret_f16__offset__waterfall__amdgpu
; GFX8-NEXT: s_waitcnt vmcnt(0)
; GFX8-NEXT: v_cmp_eq_u32_e32 vcc, v4, v6
; GFX8-NEXT: s_or_b64 s[6:7], vcc, s[6:7]
-; GFX8-NEXT: v_mov_b32_e32 v6, v4
; GFX8-NEXT: buffer_wbinvl1
; GFX8-NEXT: s_andn2_b64 exec, exec, s[6:7]
; GFX8-NEXT: s_cbranch_execnz .LBB12_3
@@ -4082,18 +4098,19 @@ define half @buffer_fat_ptr_agent_atomic_fmin_ret_f16__offset__waterfall__amdgpu
; GFX7-NEXT: v_cmp_eq_u64_e64 s[4:5], s[10:11], v[2:3]
; GFX7-NEXT: s_and_b64 s[4:5], vcc, s[4:5]
; GFX7-NEXT: s_and_saveexec_b64 s[4:5], s[4:5]
-; GFX7-NEXT: buffer_load_dword v6, v8, s[8:11], 0 offen
+; GFX7-NEXT: buffer_load_dword v4, v8, s[8:11], 0 offen
; GFX7-NEXT: s_xor_b64 exec, exec, s[4:5]
; GFX7-NEXT: s_cbranch_execnz .LBB12_1
; GFX7-NEXT: ; %bb.2:
; GFX7-NEXT: s_mov_b64 exec, s[6:7]
-; GFX7-NEXT: v_cvt_f16_f32_e32 v4, v5
+; GFX7-NEXT: v_cvt_f16_f32_e32 v5, v5
; GFX7-NEXT: s_mov_b64 s[6:7], 0
-; GFX7-NEXT: v_cvt_f32_f16_e32 v10, v4
+; GFX7-NEXT: v_cvt_f32_f16_e32 v10, v5
; GFX7-NEXT: .LBB12_3: ; %atomicrmw.start
; GFX7-NEXT: ; =>This Loop Header: Depth=1
; GFX7-NEXT: ; Child Loop BB12_4 Depth 2
; GFX7-NEXT: s_waitcnt vmcnt(0)
+; GFX7-NEXT: v_mov_b32_e32 v6, v4
; GFX7-NEXT: v_lshrrev_b32_e32 v4, v7, v6
; GFX7-NEXT: v_cvt_f32_f16_e32 v4, v4
; GFX7-NEXT: v_and_b32_e32 v5, v6, v9
@@ -4123,7 +4140,6 @@ define half @buffer_fat_ptr_agent_atomic_fmin_ret_f16__offset__waterfall__amdgpu
; GFX7-NEXT: s_waitcnt vmcnt(0)
; GFX7-NEXT: v_cmp_eq_u32_e32 vcc, v4, v6
; GFX7-NEXT: s_or_b64 s[6:7], vcc, s[6:7]
-; GFX7-NEXT: v_mov_b32_e32 v6, v4
; GFX7-NEXT: buffer_wbinvl1
; GFX7-NEXT: s_andn2_b64 exec, exec, s[6:7]
; GFX7-NEXT: s_cbranch_execnz .LBB12_3
@@ -4152,18 +4168,20 @@ define half @buffer_fat_ptr_agent_atomic_fmin_ret_f16__offset__waterfall__amdgpu
; GFX6-NEXT: v_cmp_eq_u64_e64 s[4:5], s[10:11], v[2:3]
; GFX6-NEXT: s_and_b64 s[4:5], vcc, s[4:5]
; GFX6-NEXT: s_and_saveexec_b64 s[4:5], s[4:5]
-; GFX6-NEXT: buffer_load_dword v6, v8, s[8:11], 0 offen
+; GFX6-NEXT: buffer_load_dword v4, v8, s[8:11], 0 offen
; GFX6-NEXT: s_xor_b64 exec, exec, s[4:5]
; GFX6-NEXT: s_cbranch_execnz .LBB12_1
; GFX6-NEXT: ; %bb.2:
; GFX6-NEXT: s_mov_b64 exec, s[6:7]
-; GFX6-NEXT: v_cvt_f16_f32_e32 v4, v5
+; GFX6-NEXT: v_cvt_f16_f32_e32 v5, v5
; GFX6-NEXT: s_mov_b64 s[6:7], 0
-; GFX6-NEXT: v_cvt_f32_f16_e32 v10, v4
+; GFX6-NEXT: v_cvt_f32_f16_e32 v10, v5
; GFX6-NEXT: .LBB12_3: ; %atomicrmw.start
; GFX6-NEXT: ; =>This Loop Header: Depth=1
; GFX6-NEXT: ; Child Loop BB12_4 Depth 2
-; GFX6-NEXT: s_waitcnt vmcnt(0) expcnt(0)
+; GFX6-NEXT: s_waitcnt vmcnt(0)
+; GFX6-NEXT: v_mov_b32_e32 v6, v4
+; GFX6-NEXT: s_waitcnt expcnt(0)
; GFX6-NEXT: v_lshrrev_b32_e32 v4, v7, v6
; GFX6-NEXT: v_cvt_f32_f16_e32 v4, v4
; GFX6-NEXT: v_and_b32_e32 v5, v6, v9
@@ -4193,7 +4211,6 @@ define half @buffer_fat_ptr_agent_atomic_fmin_ret_f16__offset__waterfall__amdgpu
; GFX6-NEXT: s_waitcnt vmcnt(0)
; GFX6-NEXT: v_cmp_eq_u32_e32 vcc, v4, v6
; GFX6-NEXT: s_or_b64 s[6:7], vcc, s[6:7]
-; GFX6-NEXT: v_mov_b32_e32 v6, v4
; GFX6-NEXT: buffer_wbinvl1
; GFX6-NEXT: s_andn2_b64 exec, exec, s[6:7]
; GFX6-NEXT: s_cbranch_execnz .LBB12_3
@@ -4221,55 +4238,55 @@ define bfloat @buffer_fat_ptr_agent_atomic_fmin_ret_bf16__offset__amdgpu_no_fine
; GFX12-TRUE16-NEXT: s_wait_bvhcnt 0x0
; GFX12-TRUE16-NEXT: s_wait_kmcnt 0x0
; GFX12-TRUE16-NEXT: s_addk_co_i32 s16, 0x200
-; GFX12-TRUE16-NEXT: v_lshlrev_b32_e32 v5, 16, v0
+; GFX12-TRUE16-NEXT: v_lshlrev_b32_e32 v0, 16, v0
; GFX12-TRUE16-NEXT: s_wait_alu 0xfffe
; GFX12-TRUE16-NEXT: s_and_b32 s4, s16, -4
; GFX12-TRUE16-NEXT: s_wait_alu 0xfffe
-; GFX12-TRUE16-NEXT: v_mov_b32_e32 v4, s4
+; GFX12-TRUE16-NEXT: v_mov_b32_e32 v3, s4
; GFX12-TRUE16-NEXT: s_and_b32 s4, s16, 3
; GFX12-TRUE16-NEXT: s_wait_alu 0xfffe
; GFX12-TRUE16-NEXT: s_lshl_b32 s4, s4, 3
; GFX12-TRUE16-NEXT: s_wait_alu 0xfffe
; GFX12-TRUE16-NEXT: s_lshl_b32 s5, 0xffff, s4
-; GFX12-TRUE16-NEXT: buffer_load_b32 v1, v4, s[0:3], null offen
+; GFX12-TRUE16-NEXT: buffer_load_b32 v1, v3, s[0:3], null offen
; GFX12-TRUE16-NEXT: s_not_b32 s6, s5
; GFX12-TRUE16-NEXT: s_mov_b32 s5, 0
; GFX12-TRUE16-NEXT: .LBB13_1: ; %atomicrmw.start
; GFX12-TRUE16-NEXT: ; =>This Inner Loop Header: Depth=1
; GFX12-TRUE16-NEXT: s_wait_loadcnt 0x0
-; GFX12-TRUE16-NEXT: v_lshrrev_b32_e32 v0, s4, v1
+; GFX12-TRUE16-NEXT: v_mov_b32_e32 v5, v1
; GFX12-TRUE16-NEXT: s_wait_storecnt 0x0
; GFX12-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
-; GFX12-TRUE16-NEXT: v_lshlrev_b32_e32 v0, 16, v0
-; GFX12-TRUE16-NEXT: v_min_num_f32_e32 v0, v0, v5
-; GFX12-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(SKIP_2) | instid1(VALU_DEP_3)
-; GFX12-TRUE16-NEXT: v_bfe_u32 v2, v0, 16, 1
-; GFX12-TRUE16-NEXT: v_or_b32_e32 v3, 0x400000, v0
-; GFX12-TRUE16-NEXT: v_cmp_u_f32_e32 vcc_lo, v0, v0
-; GFX12-TRUE16-NEXT: v_add3_u32 v2, v2, v0, 0x7fff
+; GFX12-TRUE16-NEXT: v_lshrrev_b32_e32 v1, s4, v5
+; GFX12-TRUE16-NEXT: v_lshlrev_b32_e32 v1, 16, v1
+; GFX12-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
+; GFX12-TRUE16-NEXT: v_min_num_f32_e32 v1, v1, v0
+; GFX12-TRUE16-NEXT: v_bfe_u32 v2, v1, 16, 1
+; GFX12-TRUE16-NEXT: v_or_b32_e32 v4, 0x400000, v1
+; GFX12-TRUE16-NEXT: v_cmp_u_f32_e32 vcc_lo, v1, v1
+; GFX12-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_3) | instskip(SKIP_1) | instid1(VALU_DEP_1)
+; GFX12-TRUE16-NEXT: v_add3_u32 v2, v2, v1, 0x7fff
; GFX12-TRUE16-NEXT: s_wait_alu 0xfffd
-; GFX12-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(SKIP_1) | instid1(VALU_DEP_2)
-; GFX12-TRUE16-NEXT: v_cndmask_b32_e32 v0, v2, v3, vcc_lo
+; GFX12-TRUE16-NEXT: v_cndmask_b32_e32 v1, v2, v4, vcc_lo
; GFX12-TRUE16-NEXT: v_mov_b16_e32 v2.h, 0
-; GFX12-TRUE16-NEXT: v_mov_b16_e32 v2.l, v0.h
-; GFX12-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(SKIP_1) | instid1(VALU_DEP_1)
-; GFX12-TRUE16-NEXT: v_lshlrev_b32_e32 v0, s4, v2
+; GFX12-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_2) | instskip(NEXT) | instid1(VALU_DEP_1)
+; GFX12-TRUE16-NEXT: v_mov_b16_e32 v2.l, v1.h
+; GFX12-TRUE16-NEXT: v_lshlrev_b32_e32 v1, s4, v2
; GFX12-TRUE16-NEXT: s_wait_alu 0xfffe
-; GFX12-TRUE16-NEXT: v_and_or_b32 v0, v1, s6, v0
-; GFX12-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1)
-; GFX12-TRUE16-NEXT: v_dual_mov_b32 v3, v1 :: v_dual_mov_b32 v2, v0
-; GFX12-TRUE16-NEXT: buffer_atomic_cmpswap_b32 v[2:3], v4, s[0:3], null offen th:TH_ATOMIC_RETURN
+; GFX12-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
+; GFX12-TRUE16-NEXT: v_and_or_b32 v4, v5, s6, v1
+; GFX12-TRUE16-NEXT: v_dual_mov_b32 v1, v4 :: v_dual_mov_b32 v2, v5
+; GFX12-TRUE16-NEXT: buffer_atomic_cmpswap_b32 v[1:2], v3, s[0:3], null offen th:TH_ATOMIC_RETURN
; GFX12-TRUE16-NEXT: s_wait_loadcnt 0x0
; GFX12-TRUE16-NEXT: global_inv scope:SCOPE_DEV
-; GFX12-TRUE16-NEXT: v_cmp_eq_u32_e32 vcc_lo, v2, v1
-; GFX12-TRUE16-NEXT: v_mov_b32_e32 v1, v2
+; GFX12-TRUE16-NEXT: v_cmp_eq_u32_e32 vcc_lo, v1, v5
; GFX12-TRUE16-NEXT: s_or_b32 s5, vcc_lo, s5
; GFX12-TRUE16-NEXT: s_wait_alu 0xfffe
; GFX12-TRUE16-NEXT: s_and_not1_b32 exec_lo, exec_lo, s5
; GFX12-TRUE16-NEXT: s_cbranch_execnz .LBB13_1
; GFX12-TRUE16-NEXT: ; %bb.2: ; %atomicrmw.end
; GFX12-TRUE16-NEXT: s_or_b32 exec_lo, exec_lo, s5
-; GFX12-TRUE16-NEXT: v_lshrrev_b32_e32 v0, s4, v2
+; GFX12-TRUE16-NEXT: v_lshrrev_b32_e32 v0, s4, v1
; GFX12-TRUE16-NEXT: s_wait_loadcnt 0x0
; GFX12-TRUE16-NEXT: s_setpc_b64 s[30:31]
;
@@ -4281,54 +4298,54 @@ define bfloat @buffer_fat_ptr_agent_atomic_fmin_ret_bf16__offset__amdgpu_no_fine
; GFX12-FAKE16-NEXT: s_wait_bvhcnt 0x0
; GFX12-FAKE16-NEXT: s_wait_kmcnt 0x0
; GFX12-FAKE16-NEXT: s_addk_co_i32 s16, 0x200
-; GFX12-FAKE16-NEXT: v_lshlrev_b32_e32 v5, 16, v0
+; GFX12-FAKE16-NEXT: v_lshlrev_b32_e32 v0, 16, v0
; GFX12-FAKE16-NEXT: s_wait_alu 0xfffe
; GFX12-FAKE16-NEXT: s_and_b32 s4, s16, -4
; GFX12-FAKE16-NEXT: s_wait_alu 0xfffe
-; GFX12-FAKE16-NEXT: v_mov_b32_e32 v4, s4
+; GFX12-FAKE16-NEXT: v_mov_b32_e32 v3, s4
; GFX12-FAKE16-NEXT: s_and_b32 s4, s16, 3
; GFX12-FAKE16-NEXT: s_wait_alu 0xfffe
; GFX12-FAKE16-NEXT: s_lshl_b32 s4, s4, 3
; GFX12-FAKE16-NEXT: s_wait_alu 0xfffe
; GFX12-FAKE16-NEXT: s_lshl_b32 s5, 0xffff, s4
-; GFX12-FAKE16-NEXT: buffer_load_b32 v1, v4, s[0:3], null offen
+; GFX12-FAKE16-NEXT: buffer_load_b32 v1, v3, s[0:3], null offen
; GFX12-FAKE16-NEXT: s_not_b32 s6, s5
; GFX12-FAKE16-NEXT: s_mov_b32 s5, 0
; GFX12-FAKE16-NEXT: .LBB13_1: ; %atomicrmw.start
; GFX12-FAKE16-NEXT: ; =>This Inner Loop Header: Depth=1
; GFX12-FAKE16-NEXT: s_wait_loadcnt 0x0
-; GFX12-FAKE16-NEXT: v_lshrrev_b32_e32 v0, s4, v1
+; GFX12-FAKE16-NEXT: v_mov_b32_e32 v5, v1
; GFX12-FAKE16-NEXT: s_wait_storecnt 0x0
; GFX12-FAKE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
-; GFX12-FAKE16-NEXT: v_lshlrev_b32_e32 v0, 16, v0
-; GFX12-FAKE16-NEXT: v_min_num_f32_e32 v0, v0, v5
-; GFX12-FAKE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(SKIP_2) | instid1(VALU_DEP_3)
-; GFX12-FAKE16-NEXT: v_bfe_u32 v2, v0, 16, 1
-; GFX12-FAKE16-NEXT: v_or_b32_e32 v3, 0x400000, v0
-; GFX12-FAKE16-NEXT: v_cmp_u_f32_e32 vcc_lo, v0, v0
-; GFX12-FAKE16-NEXT: v_add3_u32 v2, v2, v0, 0x7fff
+; GFX12-FAKE16-NEXT: v_lshrrev_b32_e32 v1, s4, v5
+; GFX12-FAKE16-NEXT: v_lshlrev_b32_e32 v1, 16, v1
+; GFX12-FAKE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
+; GFX12-FAKE16-NEXT: v_min_num_f32_e32 v1, v1, v0
+; GFX12-FAKE16-NEXT: v_bfe_u32 v2, v1, 16, 1
+; GFX12-FAKE16-NEXT: v_or_b32_e32 v4, 0x400000, v1
+; GFX12-FAKE16-NEXT: v_cmp_u_f32_e32 vcc_lo, v1, v1
+; GFX12-FAKE16-NEXT: s_delay_alu instid0(VALU_DEP_3) | instskip(SKIP_1) | instid1(VALU_DEP_1)
+; GFX12-FAKE16-NEXT: v_add3_u32 v2, v2, v1, 0x7fff
; GFX12-FAKE16-NEXT: s_wait_alu 0xfffd
+; GFX12-FAKE16-NEXT: v_cndmask_b32_e32 v1, v2, v4, vcc_lo
; GFX12-FAKE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
-; GFX12-FAKE16-NEXT: v_cndmask_b32_e32 v0, v2, v3, vcc_lo
-; GFX12-FAKE16-NEXT: v_lshrrev_b32_e32 v0, 16, v0
-; GFX12-FAKE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(SKIP_1) | instid1(VALU_DEP_1)
-; GFX12-FAKE16-NEXT: v_lshlrev_b32_e32 v0, s4, v0
+; GFX12-FAKE16-NEXT: v_lshrrev_b32_e32 v1, 16, v1
+; GFX12-FAKE16-NEXT: v_lshlrev_b32_e32 v1, s4, v1
; GFX12-FAKE16-NEXT: s_wait_alu 0xfffe
-; GFX12-FAKE16-NEXT: v_and_or_b32 v0, v1, s6, v0
-; GFX12-FAKE16-NEXT: s_delay_alu instid0(VALU_DEP_1)
-; GFX12-FAKE16-NEXT: v_dual_mov_b32 v3, v1 :: v_dual_mov_b32 v2, v0
-; GFX12-FAKE16-NEXT: buffer_atomic_cmpswap_b32 v[2:3], v4, s[0:3], null offen th:TH_ATOMIC_RETURN
+; GFX12-FAKE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
+; GFX12-FAKE16-NEXT: v_and_or_b32 v4, v5, s6, v1
+; GFX12-FAKE16-NEXT: v_dual_mov_b32 v1, v4 :: v_dual_mov_b32 v2, v5
+; GFX12-FAKE16-NEXT: buffer_atomic_cmpswap_b32 v[1:2], v3, s[0:3], null offen th:TH_ATOMIC_RETURN
; GFX12-FAKE16-NEXT: s_wait_loadcnt 0x0
; GFX12-FAKE16-NEXT: global_inv scope:SCOPE_DEV
-; GFX12-FAKE16-NEXT: v_cmp_eq_u32_e32 vcc_lo, v2, v1
-; GFX12-FAKE16-NEXT: v_mov_b32_e32 v1, v2
+; GFX12-FAKE16-NEXT: v_cmp_eq_u32_e32 vcc_lo, v1, v5
; GFX12-FAKE16-NEXT: s_or_b32 s5, vcc_lo, s5
; GFX12-FAKE16-NEXT: s_wait_alu 0xfffe
; GFX12-FAKE16-NEXT: s_and_not1_b32 exec_lo, exec_lo, s5
; GFX12-FAKE16-NEXT: s_cbranch_execnz .LBB13_1
; GFX12-FAKE16-NEXT: ; %bb.2: ; %atomicrmw.end
; GFX12-FAKE16-NEXT: s_or_b32 exec_lo, exec_lo, s5
-; GFX12-FAKE16-NEXT: v_lshrrev_b32_e32 v0, s4, v2
+; GFX12-FAKE16-NEXT: v_lshrrev_b32_e32 v0, s4, v1
; GFX12-FAKE16-NEXT: s_wait_loadcnt 0x0
; GFX12-FAKE16-NEXT: s_setpc_b64 s[30:31]
;
@@ -4337,36 +4354,36 @@ define bfloat @buffer_fat_ptr_agent_atomic_fmin_ret_bf16__offset__amdgpu_no_fine
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX942-NEXT: s_addk_i32 s16, 0x200
; GFX942-NEXT: s_and_b32 s4, s16, -4
-; GFX942-NEXT: v_mov_b32_e32 v4, s4
-; GFX942-NEXT: buffer_load_dword v1, v4, s[0:3], 0 offen
+; GFX942-NEXT: v_mov_b32_e32 v1, s4
+; GFX942-NEXT: buffer_load_dword v2, v1, s[0:3], 0 offen
; GFX942-NEXT: s_and_b32 s4, s16, 3
; GFX942-NEXT: s_lshl_b32 s6, s4, 3
; GFX942-NEXT: s_lshl_b32 s4, 0xffff, s6
; GFX942-NEXT: s_not_b32 s7, s4
; GFX942-NEXT: s_mov_b64 s[4:5], 0
-; GFX942-NEXT: v_lshlrev_b32_e32 v5, 16, v0
+; GFX942-NEXT: v_lshlrev_b32_e32 v0, 16, v0
; GFX942-NEXT: s_movk_i32 s8, 0x7fff
; GFX942-NEXT: .LBB13_1: ; %atomicrmw.start
; GFX942-NEXT: ; =>This Inner Loop Header: Depth=1
; GFX942-NEXT: s_waitcnt vmcnt(0)
-; GFX942-NEXT: v_lshrrev_b32_sdwa v0, s6, v1 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:DWORD
+; GFX942-NEXT: v_mov_b32_e32 v5, v2
+; GFX942-NEXT: v_lshrrev_b32_sdwa v2, s6, v5 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:DWORD
; GFX942-NEXT: buffer_wbl2 sc1
-; GFX942-NEXT: v_min_f32_e32 v0, v0, v5
-; GFX942-NEXT: v_bfe_u32 v2, v0, 16, 1
-; GFX942-NEXT: v_or_b32_e32 v3, 0x400000, v0
-; GFX942-NEXT: v_add3_u32 v2, v2, v0, s8
-; GFX942-NEXT: v_cmp_u_f32_e32 vcc, v0, v0
+; GFX942-NEXT: v_min_f32_e32 v2, v2, v0
+; GFX942-NEXT: v_bfe_u32 v3, v2, 16, 1
+; GFX942-NEXT: v_or_b32_e32 v4, 0x400000, v2
+; GFX942-NEXT: v_add3_u32 v3, v3, v2, s8
+; GFX942-NEXT: v_cmp_u_f32_e32 vcc, v2, v2
; GFX942-NEXT: s_nop 1
-; GFX942-NEXT: v_cndmask_b32_e32 v0, v2, v3, vcc
-; GFX942-NEXT: v_lshlrev_b32_sdwa v0, s6, v0 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:WORD_1
-; GFX942-NEXT: v_and_or_b32 v0, v1, s7, v0
-; GFX942-NEXT: v_mov_b64_e32 v[2:3], v[0:1]
-; GFX942-NEXT: buffer_atomic_cmpswap v[2:3], v4, s[0:3], 0 offen sc0
+; GFX942-NEXT: v_cndmask_b32_e32 v2, v3, v4, vcc
+; GFX942-NEXT: v_lshlrev_b32_sdwa v2, s6, v2 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:WORD_1
+; GFX942-NEXT: v_and_or_b32 v4, v5, s7, v2
+; GFX942-NEXT: v_mov_b64_e32 v[2:3], v[4:5]
+; GFX942-NEXT: buffer_atomic_cmpswap v[2:3], v1, s[0:3], 0 offen sc0
; GFX942-NEXT: s_waitcnt vmcnt(0)
; GFX942-NEXT: buffer_inv sc1
-; GFX942-NEXT: v_cmp_eq_u32_e32 vcc, v2, v1
+; GFX942-NEXT: v_cmp_eq_u32_e32 vcc, v2, v5
; GFX942-NEXT: s_or_b64 s[4:5], vcc, s[4:5]
-; GFX942-NEXT: v_mov_b32_e32 v1, v2
; GFX942-NEXT: s_andn2_b64 exec, exec, s[4:5]
; GFX942-NEXT: s_cbranch_execnz .LBB13_1
; GFX942-NEXT: ; %bb.2: ; %atomicrmw.end
@@ -4378,147 +4395,147 @@ define bfloat @buffer_fat_ptr_agent_atomic_fmin_ret_bf16__offset__amdgpu_no_fine
; GFX11-TRUE16: ; %bb.0:
; GFX11-TRUE16-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX11-TRUE16-NEXT: s_addk_i32 s16, 0x200
-; GFX11-TRUE16-NEXT: v_lshlrev_b32_e32 v5, 16, v0
+; GFX11-TRUE16-NEXT: v_lshlrev_b32_e32 v0, 16, v0
; GFX11-TRUE16-NEXT: s_and_b32 s4, s16, -4
; GFX11-TRUE16-NEXT: s_delay_alu instid0(SALU_CYCLE_1) | instskip(SKIP_1) | instid1(SALU_CYCLE_1)
-; GFX11-TRUE16-NEXT: v_mov_b32_e32 v4, s4
+; GFX11-TRUE16-NEXT: v_mov_b32_e32 v3, s4
; GFX11-TRUE16-NEXT: s_and_b32 s4, s16, 3
; GFX11-TRUE16-NEXT: s_lshl_b32 s4, s4, 3
; GFX11-TRUE16-NEXT: s_delay_alu instid0(SALU_CYCLE_1)
; GFX11-TRUE16-NEXT: s_lshl_b32 s5, 0xffff, s4
-; GFX11-TRUE16-NEXT: buffer_load_b32 v1, v4, s[0:3], 0 offen
+; GFX11-TRUE16-NEXT: buffer_load_b32 v1, v3, s[0:3], 0 offen
; GFX11-TRUE16-NEXT: s_not_b32 s6, s5
; GFX11-TRUE16-NEXT: s_mov_b32 s5, 0
; GFX11-TRUE16-NEXT: .p2align 6
; GFX11-TRUE16-NEXT: .LBB13_1: ; %atomicrmw.start
; GFX11-TRUE16-NEXT: ; =>This Inner Loop Header: Depth=1
; GFX11-TRUE16-NEXT: s_waitcnt vmcnt(0)
-; GFX11-TRUE16-NEXT: v_lshrrev_b32_e32 v0, s4, v1
+; GFX11-TRUE16-NEXT: v_mov_b32_e32 v5, v1
; GFX11-TRUE16-NEXT: s_waitcnt_vscnt null, 0x0
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
-; GFX11-TRUE16-NEXT: v_lshlrev_b32_e32 v0, 16, v0
-; GFX11-TRUE16-NEXT: v_min_f32_e32 v0, v0, v5
-; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(SKIP_2) | instid1(VALU_DEP_3)
-; GFX11-TRUE16-NEXT: v_bfe_u32 v2, v0, 16, 1
-; GFX11-TRUE16-NEXT: v_or_b32_e32 v3, 0x400000, v0
-; GFX11-TRUE16-NEXT: v_cmp_u_f32_e32 vcc_lo, v0, v0
-; GFX11-TRUE16-NEXT: v_add3_u32 v2, v2, v0, 0x7fff
-; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(SKIP_1) | instid1(VALU_DEP_2)
-; GFX11-TRUE16-NEXT: v_cndmask_b32_e32 v0, v2, v3, vcc_lo
+; GFX11-TRUE16-NEXT: v_lshrrev_b32_e32 v1, s4, v5
+; GFX11-TRUE16-NEXT: v_lshlrev_b32_e32 v1, 16, v1
+; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
+; GFX11-TRUE16-NEXT: v_min_f32_e32 v1, v1, v0
+; GFX11-TRUE16-NEXT: v_bfe_u32 v2, v1, 16, 1
+; GFX11-TRUE16-NEXT: v_or_b32_e32 v4, 0x400000, v1
+; GFX11-TRUE16-NEXT: v_cmp_u_f32_e32 vcc_lo, v1, v1
+; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_3) | instskip(NEXT) | instid1(VALU_DEP_1)
+; GFX11-TRUE16-NEXT: v_add3_u32 v2, v2, v1, 0x7fff
+; GFX11-TRUE16-NEXT: v_cndmask_b32_e32 v1, v2, v4, vcc_lo
; GFX11-TRUE16-NEXT: v_mov_b16_e32 v2.h, 0
-; GFX11-TRUE16-NEXT: v_mov_b16_e32 v2.l, v0.h
+; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_2) | instskip(NEXT) | instid1(VALU_DEP_1)
+; GFX11-TRUE16-NEXT: v_mov_b16_e32 v2.l, v1.h
+; GFX11-TRUE16-NEXT: v_lshlrev_b32_e32 v1, s4, v2
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
-; GFX11-TRUE16-NEXT: v_lshlrev_b32_e32 v0, s4, v2
-; GFX11-TRUE16-NEXT: v_and_or_b32 v0, v1, s6, v0
-; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1)
-; GFX11-TRUE16-NEXT: v_dual_mov_b32 v3, v1 :: v_dual_mov_b32 v2, v0
-; GFX11-TRUE16-NEXT: buffer_atomic_cmpswap_b32 v[2:3], v4, s[0:3], 0 offen glc
+; GFX11-TRUE16-NEXT: v_and_or_b32 v4, v5, s6, v1
+; GFX11-TRUE16-NEXT: v_dual_mov_b32 v1, v4 :: v_dual_mov_b32 v2, v5
+; GFX11-TRUE16-NEXT: buffer_atomic_cmpswap_b32 v[1:2], v3, s[0:3], 0 offen glc
; GFX11-TRUE16-NEXT: s_waitcnt vmcnt(0)
; GFX11-TRUE16-NEXT: buffer_gl1_inv
; GFX11-TRUE16-NEXT: buffer_gl0_inv
-; GFX11-TRUE16-NEXT: v_cmp_eq_u32_e32 vcc_lo, v2, v1
-; GFX11-TRUE16-NEXT: v_mov_b32_e32 v1, v2
+; GFX11-TRUE16-NEXT: v_cmp_eq_u32_e32 vcc_lo, v1, v5
; GFX11-TRUE16-NEXT: s_or_b32 s5, vcc_lo, s5
; GFX11-TRUE16-NEXT: s_delay_alu instid0(SALU_CYCLE_1)
; GFX11-TRUE16-NEXT: s_and_not1_b32 exec_lo, exec_lo, s5
; GFX11-TRUE16-NEXT: s_cbranch_execnz .LBB13_1
; GFX11-TRUE16-NEXT: ; %bb.2: ; %atomicrmw.end
; GFX11-TRUE16-NEXT: s_or_b32 exec_lo, exec_lo, s5
-; GFX11-TRUE16-NEXT: v_lshrrev_b32_e32 v0, s4, v2
+; GFX11-TRUE16-NEXT: v_lshrrev_b32_e32 v0, s4, v1
; GFX11-TRUE16-NEXT: s_setpc_b64 s[30:31]
;
; GFX11-FAKE16-LABEL: buffer_fat_ptr_agent_atomic_fmin_ret_bf16__offset__amdgpu_no_fine_grained_memory:
; GFX11-FAKE16: ; %bb.0:
; GFX11-FAKE16-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX11-FAKE16-NEXT: s_addk_i32 s16, 0x200
-; GFX11-FAKE16-NEXT: v_lshlrev_b32_e32 v5, 16, v0
+; GFX11-FAKE16-NEXT: v_lshlrev_b32_e32 v0, 16, v0
; GFX11-FAKE16-NEXT: s_and_b32 s4, s16, -4
; GFX11-FAKE16-NEXT: s_delay_alu instid0(SALU_CYCLE_1) | instskip(SKIP_1) | instid1(SALU_CYCLE_1)
-; GFX11-FAKE16-NEXT: v_mov_b32_e32 v4, s4
+; GFX11-FAKE16-NEXT: v_mov_b32_e32 v3, s4
; GFX11-FAKE16-NEXT: s_and_b32 s4, s16, 3
; GFX11-FAKE16-NEXT: s_lshl_b32 s4, s4, 3
; GFX11-FAKE16-NEXT: s_delay_alu instid0(SALU_CYCLE_1)
; GFX11-FAKE16-NEXT: s_lshl_b32 s5, 0xffff, s4
-; GFX11-FAKE16-NEXT: buffer_load_b32 v1, v4, s[0:3], 0 offen
+; GFX11-FAKE16-NEXT: buffer_load_b32 v1, v3, s[0:3], 0 offen
; GFX11-FAKE16-NEXT: s_not_b32 s6, s5
; GFX11-FAKE16-NEXT: s_mov_b32 s5, 0
; GFX11-FAKE16-NEXT: .p2align 6
; GFX11-FAKE16-NEXT: .LBB13_1: ; %atomicrmw.start
; GFX11-FAKE16-NEXT: ; =>This Inner Loop Header: Depth=1
; GFX11-FAKE16-NEXT: s_waitcnt vmcnt(0)
-; GFX11-FAKE16-NEXT: v_lshrrev_b32_e32 v0, s4, v1
+; GFX11-FAKE16-NEXT: v_mov_b32_e32 v5, v1
; GFX11-FAKE16-NEXT: s_waitcnt_vscnt null, 0x0
; GFX11-FAKE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
-; GFX11-FAKE16-NEXT: v_lshlrev_b32_e32 v0, 16, v0
-; GFX11-FAKE16-NEXT: v_min_f32_e32 v0, v0, v5
-; GFX11-FAKE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(SKIP_2) | instid1(VALU_DEP_3)
-; GFX11-FAKE16-NEXT: v_bfe_u32 v2, v0, 16, 1
-; GFX11-FAKE16-NEXT: v_or_b32_e32 v3, 0x400000, v0
-; GFX11-FAKE16-NEXT: v_cmp_u_f32_e32 vcc_lo, v0, v0
-; GFX11-FAKE16-NEXT: v_add3_u32 v2, v2, v0, 0x7fff
+; GFX11-FAKE16-NEXT: v_lshrrev_b32_e32 v1, s4, v5
+; GFX11-FAKE16-NEXT: v_lshlrev_b32_e32 v1, 16, v1
; GFX11-FAKE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
-; GFX11-FAKE16-NEXT: v_cndmask_b32_e32 v0, v2, v3, vcc_lo
-; GFX11-FAKE16-NEXT: v_lshrrev_b32_e32 v0, 16, v0
+; GFX11-FAKE16-NEXT: v_min_f32_e32 v1, v1, v0
+; GFX11-FAKE16-NEXT: v_bfe_u32 v2, v1, 16, 1
+; GFX11-FAKE16-NEXT: v_or_b32_e32 v4, 0x400000, v1
+; GFX11-FAKE16-NEXT: v_cmp_u_f32_e32 vcc_lo, v1, v1
+; GFX11-FAKE16-NEXT: s_delay_alu instid0(VALU_DEP_3) | instskip(NEXT) | instid1(VALU_DEP_1)
+; GFX11-FAKE16-NEXT: v_add3_u32 v2, v2, v1, 0x7fff
+; GFX11-FAKE16-NEXT: v_cndmask_b32_e32 v1, v2, v4, vcc_lo
; GFX11-FAKE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
-; GFX11-FAKE16-NEXT: v_lshlrev_b32_e32 v0, s4, v0
-; GFX11-FAKE16-NEXT: v_and_or_b32 v0, v1, s6, v0
-; GFX11-FAKE16-NEXT: s_delay_alu instid0(VALU_DEP_1)
-; GFX11-FAKE16-NEXT: v_dual_mov_b32 v3, v1 :: v_dual_mov_b32 v2, v0
-; GFX11-FAKE16-NEXT: buffer_atomic_cmpswap_b32 v[2:3], v4, s[0:3], 0 offen glc
+; GFX11-FAKE16-NEXT: v_lshrrev_b32_e32 v1, 16, v1
+; GFX11-FAKE16-NEXT: v_lshlrev_b32_e32 v1, s4, v1
+; GFX11-FAKE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
+; GFX11-FAKE16-NEXT: v_and_or_b32 v4, v5, s6, v1
+; GFX11-FAKE16-NEXT: v_dual_mov_b32 v1, v4 :: v_dual_mov_b32 v2, v5
+; GFX11-FAKE16-NEXT: buffer_atomic_cmpswap_b32 v[1:2], v3, s[0:3], 0 offen glc
; GFX11-FAKE16-NEXT: s_waitcnt vmcnt(0)
; GFX11-FAKE16-NEXT: buffer_gl1_inv
; GFX11-FAKE16-NEXT: buffer_gl0_inv
-; GFX11-FAKE16-NEXT: v_cmp_eq_u32_e32 vcc_lo, v2, v1
-; GFX11-FAKE16-NEXT: v_mov_b32_e32 v1, v2
+; GFX11-FAKE16-NEXT: v_cmp_eq_u32_e32 vcc_lo, v1, v5
; GFX11-FAKE16-NEXT: s_or_b32 s5, vcc_lo, s5
; GFX11-FAKE16-NEXT: s_delay_alu instid0(SALU_CYCLE_1)
; GFX11-FAKE16-NEXT: s_and_not1_b32 exec_lo, exec_lo, s5
; GFX11-FAKE16-NEXT: s_cbranch_execnz .LBB13_1
; GFX11-FAKE16-NEXT: ; %bb.2: ; %atomicrmw.end
; GFX11-FAKE16-NEXT: s_or_b32 exec_lo, exec_lo, s5
-; GFX11-FAKE16-NEXT: v_lshrrev_b32_e32 v0, s4, v2
+; GFX11-FAKE16-NEXT: v_lshrrev_b32_e32 v0, s4, v1
; GFX11-FAKE16-NEXT: s_setpc_b64 s[30:31]
;
; GFX10-LABEL: buffer_fat_ptr_agent_atomic_fmin_ret_bf16__offset__amdgpu_no_fine_grained_memory:
; GFX10: ; %bb.0:
; GFX10-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX10-NEXT: s_addk_i32 s20, 0x200
-; GFX10-NEXT: v_lshlrev_b32_e32 v5, 16, v0
+; GFX10-NEXT: v_lshlrev_b32_e32 v0, 16, v0
; GFX10-NEXT: s_and_b32 s4, s20, -4
-; GFX10-NEXT: v_mov_b32_e32 v4, s4
+; GFX10-NEXT: v_mov_b32_e32 v3, s4
; GFX10-NEXT: s_and_b32 s4, s20, 3
; GFX10-NEXT: s_lshl_b32 s4, s4, 3
; GFX10-NEXT: s_lshl_b32 s5, 0xffff, s4
-; GFX10-NEXT: buffer_load_dword v1, v4, s[16:19], 0 offen
+; GFX10-NEXT: buffer_load_dword v1, v3, s[16:19], 0 offen
; GFX10-NEXT: s_not_b32 s6, s5
; GFX10-NEXT: s_mov_b32 s5, 0
; GFX10-NEXT: .LBB13_1: ; %atomicrmw.start
; GFX10-NEXT: ; =>This Inner Loop Header: Depth=1
; GFX10-NEXT: s_waitcnt vmcnt(0)
-; GFX10-NEXT: v_lshrrev_b32_sdwa v0, s4, v1 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:DWORD
+; GFX10-NEXT: v_mov_b32_e32 v5, v1
; GFX10-NEXT: s_waitcnt_vscnt null, 0x0
-; GFX10-NEXT: v_min_f32_e32 v0, v0, v5
-; GFX10-NEXT: v_bfe_u32 v2, v0, 16, 1
-; GFX10-NEXT: v_or_b32_e32 v3, 0x400000, v0
-; GFX10-NEXT: v_cmp_u_f32_e32 vcc_lo, v0, v0
-; GFX10-NEXT: v_add3_u32 v2, v2, v0, 0x7fff
-; GFX10-NEXT: v_cndmask_b32_e32 v0, v2, v3, vcc_lo
-; GFX10-NEXT: v_lshlrev_b32_sdwa v0, s4, v0 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:WORD_1
-; GFX10-NEXT: v_and_or_b32 v0, v1, s6, v0
-; GFX10-NEXT: v_mov_b32_e32 v3, v1
-; GFX10-NEXT: v_mov_b32_e32 v2, v0
-; GFX10-NEXT: buffer_atomic_cmpswap v[2:3], v4, s[16:19], 0 offen glc
+; GFX10-NEXT: v_lshrrev_b32_sdwa v1, s4, v5 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:DWORD
+; GFX10-NEXT: v_min_f32_e32 v1, v1, v0
+; GFX10-NEXT: v_bfe_u32 v2, v1, 16, 1
+; GFX10-NEXT: v_or_b32_e32 v4, 0x400000, v1
+; GFX10-NEXT: v_cmp_u_f32_e32 vcc_lo, v1, v1
+; GFX10-NEXT: v_add3_u32 v2, v2, v1, 0x7fff
+; GFX10-NEXT: v_cndmask_b32_e32 v1, v2, v4, vcc_lo
+; GFX10-NEXT: v_lshlrev_b32_sdwa v1, s4, v1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:WORD_1
+; GFX10-NEXT: v_and_or_b32 v4, v5, s6, v1
+; GFX10-NEXT: v_mov_b32_e32 v1, v4
+; GFX10-NEXT: v_mov_b32_e32 v2, v5
+; GFX10-NEXT: buffer_atomic_cmpswap v[1:2], v3, s[16:19], 0 offen glc
; GFX10-NEXT: s_waitcnt vmcnt(0)
; GFX10-NEXT: buffer_gl1_inv
; GFX10-NEXT: buffer_gl0_inv
-; GFX10-NEXT: v_cmp_eq_u32_e32 vcc_lo, v2, v1
-; GFX10-NEXT: v_mov_b32_e32 v1, v2
+; GFX10-NEXT: v_cmp_eq_u32_e32 vcc_lo, v1, v5
; GFX10-NEXT: s_or_b32 s5, vcc_lo, s5
; GFX10-NEXT: s_andn2_b32 exec_lo, exec_lo, s5
; GFX10-NEXT: s_cbranch_execnz .LBB13_1
; GFX10-NEXT: ; %bb.2: ; %atomicrmw.end
; GFX10-NEXT: s_or_b32 exec_lo, exec_lo, s5
-; GFX10-NEXT: v_lshrrev_b32_e32 v0, s4, v2
+; GFX10-NEXT: v_lshrrev_b32_e32 v0, s4, v1
; GFX10-NEXT: s_setpc_b64 s[30:31]
;
; GFX90A-LABEL: buffer_fat_ptr_agent_atomic_fmin_ret_bf16__offset__amdgpu_no_fine_grained_memory:
@@ -4526,34 +4543,34 @@ define bfloat @buffer_fat_ptr_agent_atomic_fmin_ret_bf16__offset__amdgpu_no_fine
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX90A-NEXT: s_addk_i32 s20, 0x200
; GFX90A-NEXT: s_and_b32 s4, s20, -4
-; GFX90A-NEXT: v_mov_b32_e32 v4, s4
-; GFX90A-NEXT: buffer_load_dword v1, v4, s[16:19], 0 offen
+; GFX90A-NEXT: v_mov_b32_e32 v1, s4
+; GFX90A-NEXT: buffer_load_dword v2, v1, s[16:19], 0 offen
; GFX90A-NEXT: s_and_b32 s4, s20, 3
; GFX90A-NEXT: s_lshl_b32 s6, s4, 3
; GFX90A-NEXT: s_lshl_b32 s4, 0xffff, s6
; GFX90A-NEXT: s_not_b32 s7, s4
; GFX90A-NEXT: s_mov_b64 s[4:5], 0
-; GFX90A-NEXT: v_lshlrev_b32_e32 v5, 16, v0
+; GFX90A-NEXT: v_lshlrev_b32_e32 v0, 16, v0
; GFX90A-NEXT: s_movk_i32 s8, 0x7fff
; GFX90A-NEXT: .LBB13_1: ; %atomicrmw.start
; GFX90A-NEXT: ; =>This Inner Loop Header: Depth=1
; GFX90A-NEXT: s_waitcnt vmcnt(0)
-; GFX90A-NEXT: v_lshrrev_b32_sdwa v0, s6, v1 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:DWORD
-; GFX90A-NEXT: v_min_f32_e32 v0, v0, v5
-; GFX90A-NEXT: v_bfe_u32 v2, v0, 16, 1
-; GFX90A-NEXT: v_or_b32_e32 v3, 0x400000, v0
-; GFX90A-NEXT: v_add3_u32 v2, v2, v0, s8
-; GFX90A-NEXT: v_cmp_u_f32_e32 vcc, v0, v0
-; GFX90A-NEXT: v_cndmask_b32_e32 v0, v2, v3, vcc
-; GFX90A-NEXT: v_lshlrev_b32_sdwa v0, s6, v0 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:WORD_1
-; GFX90A-NEXT: v_and_or_b32 v0, v1, s7, v0
-; GFX90A-NEXT: v_pk_mov_b32 v[2:3], v[0:1], v[0:1] op_sel:[0,1]
-; GFX90A-NEXT: buffer_atomic_cmpswap v[2:3], v4, s[16:19], 0 offen glc
+; GFX90A-NEXT: v_mov_b32_e32 v5, v2
+; GFX90A-NEXT: v_lshrrev_b32_sdwa v2, s6, v5 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:DWORD
+; GFX90A-NEXT: v_min_f32_e32 v2, v2, v0
+; GFX90A-NEXT: v_bfe_u32 v3, v2, 16, 1
+; GFX90A-NEXT: v_or_b32_e32 v4, 0x400000, v2
+; GFX90A-NEXT: v_add3_u32 v3, v3, v2, s8
+; GFX90A-NEXT: v_cmp_u_f32_e32 vcc, v2, v2
+; GFX90A-NEXT: v_cndmask_b32_e32 v2, v3, v4, vcc
+; GFX90A-NEXT: v_lshlrev_b32_sdwa v2, s6, v2 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:WORD_1
+; GFX90A-NEXT: v_and_or_b32 v4, v5, s7, v2
+; GFX90A-NEXT: v_pk_mov_b32 v[2:3], v[4:5], v[4:5] op_sel:[0,1]
+; GFX90A-NEXT: buffer_atomic_cmpswap v[2:3], v1, s[16:19], 0 offen glc
; GFX90A-NEXT: s_waitcnt vmcnt(0)
; GFX90A-NEXT: buffer_wbinvl1
-; GFX90A-NEXT: v_cmp_eq_u32_e32 vcc, v2, v1
+; GFX90A-NEXT: v_cmp_eq_u32_e32 vcc, v2, v5
; GFX90A-NEXT: s_or_b64 s[4:5], vcc, s[4:5]
-; GFX90A-NEXT: v_mov_b32_e32 v1, v2
; GFX90A-NEXT: s_andn2_b64 exec, exec, s[4:5]
; GFX90A-NEXT: s_cbranch_execnz .LBB13_1
; GFX90A-NEXT: ; %bb.2: ; %atomicrmw.end
@@ -4566,40 +4583,40 @@ define bfloat @buffer_fat_ptr_agent_atomic_fmin_ret_bf16__offset__amdgpu_no_fine
; GFX908-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX908-NEXT: s_addk_i32 s20, 0x200
; GFX908-NEXT: s_and_b32 s4, s20, -4
-; GFX908-NEXT: v_mov_b32_e32 v4, s4
-; GFX908-NEXT: buffer_load_dword v1, v4, s[16:19], 0 offen
+; GFX908-NEXT: v_mov_b32_e32 v3, s4
+; GFX908-NEXT: buffer_load_dword v1, v3, s[16:19], 0 offen
; GFX908-NEXT: s_and_b32 s4, s20, 3
; GFX908-NEXT: s_lshl_b32 s6, s4, 3
; GFX908-NEXT: s_lshl_b32 s4, 0xffff, s6
; GFX908-NEXT: s_not_b32 s7, s4
; GFX908-NEXT: s_mov_b64 s[4:5], 0
-; GFX908-NEXT: v_lshlrev_b32_e32 v5, 16, v0
+; GFX908-NEXT: v_lshlrev_b32_e32 v0, 16, v0
; GFX908-NEXT: s_movk_i32 s8, 0x7fff
; GFX908-NEXT: .LBB13_1: ; %atomicrmw.start
; GFX908-NEXT: ; =>This Inner Loop Header: Depth=1
; GFX908-NEXT: s_waitcnt vmcnt(0)
-; GFX908-NEXT: v_lshrrev_b32_sdwa v0, s6, v1 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:DWORD
-; GFX908-NEXT: v_min_f32_e32 v0, v0, v5
-; GFX908-NEXT: v_bfe_u32 v2, v0, 16, 1
-; GFX908-NEXT: v_or_b32_e32 v3, 0x400000, v0
-; GFX908-NEXT: v_add3_u32 v2, v2, v0, s8
-; GFX908-NEXT: v_cmp_u_f32_e32 vcc, v0, v0
-; GFX908-NEXT: v_cndmask_b32_e32 v0, v2, v3, vcc
-; GFX908-NEXT: v_lshlrev_b32_sdwa v0, s6, v0 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:WORD_1
-; GFX908-NEXT: v_and_or_b32 v0, v1, s7, v0
-; GFX908-NEXT: v_mov_b32_e32 v3, v1
-; GFX908-NEXT: v_mov_b32_e32 v2, v0
-; GFX908-NEXT: buffer_atomic_cmpswap v[2:3], v4, s[16:19], 0 offen glc
+; GFX908-NEXT: v_mov_b32_e32 v5, v1
+; GFX908-NEXT: v_lshrrev_b32_sdwa v1, s6, v5 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:DWORD
+; GFX908-NEXT: v_min_f32_e32 v1, v1, v0
+; GFX908-NEXT: v_bfe_u32 v2, v1, 16, 1
+; GFX908-NEXT: v_or_b32_e32 v4, 0x400000, v1
+; GFX908-NEXT: v_add3_u32 v2, v2, v1, s8
+; GFX908-NEXT: v_cmp_u_f32_e32 vcc, v1, v1
+; GFX908-NEXT: v_cndmask_b32_e32 v1, v2, v4, vcc
+; GFX908-NEXT: v_lshlrev_b32_sdwa v1, s6, v1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:WORD_1
+; GFX908-NEXT: v_and_or_b32 v4, v5, s7, v1
+; GFX908-NEXT: v_mov_b32_e32 v1, v4
+; GFX908-NEXT: v_mov_b32_e32 v2, v5
+; GFX908-NEXT: buffer_atomic_cmpswap v[1:2], v3, s[16:19], 0 offen glc
; GFX908-NEXT: s_waitcnt vmcnt(0)
; GFX908-NEXT: buffer_wbinvl1
-; GFX908-NEXT: v_cmp_eq_u32_e32 vcc, v2, v1
+; GFX908-NEXT: v_cmp_eq_u32_e32 vcc, v1, v5
; GFX908-NEXT: s_or_b64 s[4:5], vcc, s[4:5]
-; GFX908-NEXT: v_mov_b32_e32 v1, v2
; GFX908-NEXT: s_andn2_b64 exec, exec, s[4:5]
; GFX908-NEXT: s_cbranch_execnz .LBB13_1
; GFX908-NEXT: ; %bb.2: ; %atomicrmw.end
; GFX908-NEXT: s_or_b64 exec, exec, s[4:5]
-; GFX908-NEXT: v_lshrrev_b32_e32 v0, s6, v2
+; GFX908-NEXT: v_lshrrev_b32_e32 v0, s6, v1
; GFX908-NEXT: s_setpc_b64 s[30:31]
;
; GFX8-LABEL: buffer_fat_ptr_agent_atomic_fmin_ret_bf16__offset__amdgpu_no_fine_grained_memory:
@@ -4607,42 +4624,42 @@ define bfloat @buffer_fat_ptr_agent_atomic_fmin_ret_bf16__offset__amdgpu_no_fine
; GFX8-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX8-NEXT: s_addk_i32 s20, 0x200
; GFX8-NEXT: s_and_b32 s4, s20, -4
-; GFX8-NEXT: v_mov_b32_e32 v4, s4
-; GFX8-NEXT: buffer_load_dword v1, v4, s[16:19], 0 offen
+; GFX8-NEXT: v_mov_b32_e32 v3, s4
+; GFX8-NEXT: buffer_load_dword v1, v3, s[16:19], 0 offen
; GFX8-NEXT: s_and_b32 s4, s20, 3
; GFX8-NEXT: s_lshl_b32 s6, s4, 3
; GFX8-NEXT: s_lshl_b32 s4, 0xffff, s6
; GFX8-NEXT: s_not_b32 s7, s4
; GFX8-NEXT: s_mov_b64 s[4:5], 0
-; GFX8-NEXT: v_lshlrev_b32_e32 v5, 16, v0
+; GFX8-NEXT: v_lshlrev_b32_e32 v0, 16, v0
; GFX8-NEXT: .LBB13_1: ; %atomicrmw.start
; GFX8-NEXT: ; =>This Inner Loop Header: Depth=1
-; GFX8-NEXT: v_mov_b32_e32 v0, s6
+; GFX8-NEXT: v_mov_b32_e32 v2, s6
; GFX8-NEXT: s_waitcnt vmcnt(0)
-; GFX8-NEXT: v_lshrrev_b32_sdwa v3, v0, v1 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:DWORD
-; GFX8-NEXT: v_min_f32_e32 v3, v3, v5
-; GFX8-NEXT: v_bfe_u32 v6, v3, 16, 1
-; GFX8-NEXT: v_add_u32_e32 v6, vcc, v6, v3
+; GFX8-NEXT: v_mov_b32_e32 v5, v1
+; GFX8-NEXT: v_lshrrev_b32_sdwa v1, v2, v5 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:DWORD
+; GFX8-NEXT: v_min_f32_e32 v1, v1, v0
+; GFX8-NEXT: v_bfe_u32 v6, v1, 16, 1
+; GFX8-NEXT: v_add_u32_e32 v6, vcc, v6, v1
; GFX8-NEXT: v_add_u32_e32 v6, vcc, 0x7fff, v6
-; GFX8-NEXT: v_or_b32_e32 v7, 0x400000, v3
-; GFX8-NEXT: v_cmp_u_f32_e32 vcc, v3, v3
-; GFX8-NEXT: v_cndmask_b32_e32 v3, v6, v7, vcc
-; GFX8-NEXT: v_and_b32_e32 v2, s7, v1
-; GFX8-NEXT: v_lshlrev_b32_sdwa v0, v0, v3 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:WORD_1
-; GFX8-NEXT: v_or_b32_e32 v0, v2, v0
-; GFX8-NEXT: v_mov_b32_e32 v3, v1
-; GFX8-NEXT: v_mov_b32_e32 v2, v0
-; GFX8-NEXT: buffer_atomic_cmpswap v[2:3], v4, s[16:19], 0 offen glc
+; GFX8-NEXT: v_or_b32_e32 v7, 0x400000, v1
+; GFX8-NEXT: v_cmp_u_f32_e32 vcc, v1, v1
+; GFX8-NEXT: v_cndmask_b32_e32 v1, v6, v7, vcc
+; GFX8-NEXT: v_and_b32_e32 v4, s7, v5
+; GFX8-NEXT: v_lshlrev_b32_sdwa v1, v2, v1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:WORD_1
+; GFX8-NEXT: v_or_b32_e32 v4, v4, v1
+; GFX8-NEXT: v_mov_b32_e32 v1, v4
+; GFX8-NEXT: v_mov_b32_e32 v2, v5
+; GFX8-NEXT: buffer_atomic_cmpswap v[1:2], v3, s[16:19], 0 offen glc
; GFX8-NEXT: s_waitcnt vmcnt(0)
; GFX8-NEXT: buffer_wbinvl1
-; GFX8-NEXT: v_cmp_eq_u32_e32 vcc, v2, v1
+; GFX8-NEXT: v_cmp_eq_u32_e32 vcc, v1, v5
; GFX8-NEXT: s_or_b64 s[4:5], vcc, s[4:5]
-; GFX8-NEXT: v_mov_b32_e32 v1, v2
; GFX8-NEXT: s_andn2_b64 exec, exec, s[4:5]
; GFX8-NEXT: s_cbranch_execnz .LBB13_1
; GFX8-NEXT: ; %bb.2: ; %atomicrmw.end
; GFX8-NEXT: s_or_b64 exec, exec, s[4:5]
-; GFX8-NEXT: v_lshrrev_b32_e32 v0, s6, v2
+; GFX8-NEXT: v_lshrrev_b32_e32 v0, s6, v1
; GFX8-NEXT: s_setpc_b64 s[30:31]
;
; GFX7-LABEL: buffer_fat_ptr_agent_atomic_fmin_ret_bf16__offset__amdgpu_no_fine_grained_memory:
@@ -4650,39 +4667,39 @@ define bfloat @buffer_fat_ptr_agent_atomic_fmin_ret_bf16__offset__amdgpu_no_fine
; GFX7-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX7-NEXT: s_addk_i32 s20, 0x200
; GFX7-NEXT: s_and_b32 s4, s20, -4
-; GFX7-NEXT: v_mov_b32_e32 v4, s4
-; GFX7-NEXT: buffer_load_dword v1, v4, s[16:19], 0 offen
+; GFX7-NEXT: v_mov_b32_e32 v3, s4
+; GFX7-NEXT: buffer_load_dword v1, v3, s[16:19], 0 offen
; GFX7-NEXT: s_and_b32 s4, s20, 3
; GFX7-NEXT: s_lshl_b32 s6, s4, 3
; GFX7-NEXT: s_lshl_b32 s4, 0xffff, s6
; GFX7-NEXT: v_mul_f32_e32 v0, 1.0, v0
; GFX7-NEXT: s_not_b32 s7, s4
; GFX7-NEXT: s_mov_b64 s[4:5], 0
-; GFX7-NEXT: v_and_b32_e32 v5, 0xffff0000, v0
+; GFX7-NEXT: v_and_b32_e32 v0, 0xffff0000, v0
; GFX7-NEXT: .LBB13_1: ; %atomicrmw.start
; GFX7-NEXT: ; =>This Inner Loop Header: Depth=1
; GFX7-NEXT: s_waitcnt vmcnt(0)
-; GFX7-NEXT: v_lshrrev_b32_e32 v0, s6, v1
-; GFX7-NEXT: v_lshlrev_b32_e32 v0, 16, v0
-; GFX7-NEXT: v_mul_f32_e32 v0, 1.0, v0
-; GFX7-NEXT: v_min_f32_e32 v0, v0, v5
-; GFX7-NEXT: v_lshrrev_b32_e32 v0, 16, v0
-; GFX7-NEXT: v_and_b32_e32 v2, s7, v1
-; GFX7-NEXT: v_lshlrev_b32_e32 v0, s6, v0
-; GFX7-NEXT: v_or_b32_e32 v0, v2, v0
-; GFX7-NEXT: v_mov_b32_e32 v3, v1
-; GFX7-NEXT: v_mov_b32_e32 v2, v0
-; GFX7-NEXT: buffer_atomic_cmpswap v[2:3], v4, s[16:19], 0 offen glc
+; GFX7-NEXT: v_mov_b32_e32 v5, v1
+; GFX7-NEXT: v_lshrrev_b32_e32 v1, s6, v5
+; GFX7-NEXT: v_lshlrev_b32_e32 v1, 16, v1
+; GFX7-NEXT: v_mul_f32_e32 v1, 1.0, v1
+; GFX7-NEXT: v_min_f32_e32 v1, v1, v0
+; GFX7-NEXT: v_lshrrev_b32_e32 v1, 16, v1
+; GFX7-NEXT: v_and_b32_e32 v2, s7, v5
+; GFX7-NEXT: v_lshlrev_b32_e32 v1, s6, v1
+; GFX7-NEXT: v_or_b32_e32 v4, v2, v1
+; GFX7-NEXT: v_mov_b32_e32 v1, v4
+; GFX7-NEXT: v_mov_b32_e32 v2, v5
+; GFX7-NEXT: buffer_atomic_cmpswap v[1:2], v3, s[16:19], 0 offen glc
; GFX7-NEXT: s_waitcnt vmcnt(0)
; GFX7-NEXT: buffer_wbinvl1
-; GFX7-NEXT: v_cmp_eq_u32_e32 vcc, v2, v1
+; GFX7-NEXT: v_cmp_eq_u32_e32 vcc, v1, v5
; GFX7-NEXT: s_or_b64 s[4:5], vcc, s[4:5]
-; GFX7-NEXT: v_mov_b32_e32 v1, v2
; GFX7-NEXT: s_andn2_b64 exec, exec, s[4:5]
; GFX7-NEXT: s_cbranch_execnz .LBB13_1
; GFX7-NEXT: ; %bb.2: ; %atomicrmw.end
; GFX7-NEXT: s_or_b64 exec, exec, s[4:5]
-; GFX7-NEXT: v_lshrrev_b32_e32 v0, s6, v2
+; GFX7-NEXT: v_lshrrev_b32_e32 v0, s6, v1
; GFX7-NEXT: v_lshlrev_b32_e32 v0, 16, v0
; GFX7-NEXT: s_setpc_b64 s[30:31]
;
@@ -4691,40 +4708,40 @@ define bfloat @buffer_fat_ptr_agent_atomic_fmin_ret_bf16__offset__amdgpu_no_fine
; GFX6-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX6-NEXT: s_addk_i32 s20, 0x200
; GFX6-NEXT: s_and_b32 s4, s20, -4
-; GFX6-NEXT: v_mov_b32_e32 v4, s4
-; GFX6-NEXT: buffer_load_dword v1, v4, s[16:19], 0 offen
+; GFX6-NEXT: v_mov_b32_e32 v3, s4
+; GFX6-NEXT: buffer_load_dword v1, v3, s[16:19], 0 offen
; GFX6-NEXT: s_and_b32 s4, s20, 3
; GFX6-NEXT: s_lshl_b32 s6, s4, 3
; GFX6-NEXT: s_lshl_b32 s4, 0xffff, s6
; GFX6-NEXT: v_mul_f32_e32 v0, 1.0, v0
; GFX6-NEXT: s_not_b32 s7, s4
; GFX6-NEXT: s_mov_b64 s[4:5], 0
-; GFX6-NEXT: v_and_b32_e32 v5, 0xffff0000, v0
+; GFX6-NEXT: v_and_b32_e32 v0, 0xffff0000, v0
; GFX6-NEXT: .LBB13_1: ; %atomicrmw.start
; GFX6-NEXT: ; =>This Inner Loop Header: Depth=1
; GFX6-NEXT: s_waitcnt vmcnt(0)
-; GFX6-NEXT: v_lshrrev_b32_e32 v0, s6, v1
-; GFX6-NEXT: v_lshlrev_b32_e32 v0, 16, v0
-; GFX6-NEXT: v_mul_f32_e32 v0, 1.0, v0
-; GFX6-NEXT: v_min_f32_e32 v0, v0, v5
-; GFX6-NEXT: v_lshrrev_b32_e32 v0, 16, v0
+; GFX6-NEXT: v_mov_b32_e32 v5, v1
; GFX6-NEXT: s_waitcnt expcnt(0)
-; GFX6-NEXT: v_and_b32_e32 v2, s7, v1
-; GFX6-NEXT: v_lshlrev_b32_e32 v0, s6, v0
-; GFX6-NEXT: v_or_b32_e32 v0, v2, v0
-; GFX6-NEXT: v_mov_b32_e32 v3, v1
-; GFX6-NEXT: v_mov_b32_e32 v2, v0
-; GFX6-NEXT: buffer_atomic_cmpswap v[2:3], v4, s[16:19], 0 offen glc
+; GFX6-NEXT: v_lshrrev_b32_e32 v1, s6, v5
+; GFX6-NEXT: v_lshlrev_b32_e32 v1, 16, v1
+; GFX6-NEXT: v_mul_f32_e32 v1, 1.0, v1
+; GFX6-NEXT: v_min_f32_e32 v1, v1, v0
+; GFX6-NEXT: v_lshrrev_b32_e32 v1, 16, v1
+; GFX6-NEXT: v_and_b32_e32 v2, s7, v5
+; GFX6-NEXT: v_lshlrev_b32_e32 v1, s6, v1
+; GFX6-NEXT: v_or_b32_e32 v4, v2, v1
+; GFX6-NEXT: v_mov_b32_e32 v1, v4
+; GFX6-NEXT: v_mov_b32_e32 v2, v5
+; GFX6-NEXT: buffer_atomic_cmpswap v[1:2], v3, s[16:19], 0 offen glc
; GFX6-NEXT: s_waitcnt vmcnt(0)
; GFX6-NEXT: buffer_wbinvl1
-; GFX6-NEXT: v_cmp_eq_u32_e32 vcc, v2, v1
+; GFX6-NEXT: v_cmp_eq_u32_e32 vcc, v1, v5
; GFX6-NEXT: s_or_b64 s[4:5], vcc, s[4:5]
-; GFX6-NEXT: v_mov_b32_e32 v1, v2
; GFX6-NEXT: s_andn2_b64 exec, exec, s[4:5]
; GFX6-NEXT: s_cbranch_execnz .LBB13_1
; GFX6-NEXT: ; %bb.2: ; %atomicrmw.end
; GFX6-NEXT: s_or_b64 exec, exec, s[4:5]
-; GFX6-NEXT: v_lshrrev_b32_e32 v0, s6, v2
+; GFX6-NEXT: v_lshrrev_b32_e32 v0, s6, v1
; GFX6-NEXT: v_lshlrev_b32_e32 v0, 16, v0
; GFX6-NEXT: s_waitcnt expcnt(0)
; GFX6-NEXT: s_setpc_b64 s[30:31]
@@ -4742,48 +4759,48 @@ define void @buffer_fat_ptr_agent_atomic_fmin_noret_bf16__offset__amdgpu_no_fine
; GFX12-TRUE16-NEXT: s_wait_bvhcnt 0x0
; GFX12-TRUE16-NEXT: s_wait_kmcnt 0x0
; GFX12-TRUE16-NEXT: s_addk_co_i32 s16, 0x200
-; GFX12-TRUE16-NEXT: v_lshlrev_b32_e32 v3, 16, v0
+; GFX12-TRUE16-NEXT: v_lshlrev_b32_e32 v0, 16, v0
; GFX12-TRUE16-NEXT: s_wait_alu 0xfffe
; GFX12-TRUE16-NEXT: s_and_b32 s4, s16, -4
; GFX12-TRUE16-NEXT: s_wait_alu 0xfffe
-; GFX12-TRUE16-NEXT: v_mov_b32_e32 v2, s4
+; GFX12-TRUE16-NEXT: v_mov_b32_e32 v3, s4
; GFX12-TRUE16-NEXT: s_and_b32 s4, s16, 3
; GFX12-TRUE16-NEXT: s_wait_alu 0xfffe
; GFX12-TRUE16-NEXT: s_lshl_b32 s4, s4, 3
; GFX12-TRUE16-NEXT: s_wait_alu 0xfffe
; GFX12-TRUE16-NEXT: s_lshl_b32 s5, 0xffff, s4
-; GFX12-TRUE16-NEXT: buffer_load_b32 v1, v2, s[0:3], null offen
+; GFX12-TRUE16-NEXT: buffer_load_b32 v1, v3, s[0:3], null offen
; GFX12-TRUE16-NEXT: s_not_b32 s6, s5
; GFX12-TRUE16-NEXT: s_mov_b32 s5, 0
; GFX12-TRUE16-NEXT: .LBB14_1: ; %atomicrmw.start
; GFX12-TRUE16-NEXT: ; =>This Inner Loop Header: Depth=1
; GFX12-TRUE16-NEXT: s_wait_loadcnt 0x0
-; GFX12-TRUE16-NEXT: v_lshrrev_b32_e32 v0, s4, v1
+; GFX12-TRUE16-NEXT: v_mov_b32_e32 v5, v1
; GFX12-TRUE16-NEXT: s_wait_storecnt 0x0
; GFX12-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
-; GFX12-TRUE16-NEXT: v_lshlrev_b32_e32 v0, 16, v0
-; GFX12-TRUE16-NEXT: v_min_num_f32_e32 v0, v0, v3
-; GFX12-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(SKIP_2) | instid1(VALU_DEP_3)
-; GFX12-TRUE16-NEXT: v_bfe_u32 v4, v0, 16, 1
-; GFX12-TRUE16-NEXT: v_or_b32_e32 v5, 0x400000, v0
-; GFX12-TRUE16-NEXT: v_cmp_u_f32_e32 vcc_lo, v0, v0
-; GFX12-TRUE16-NEXT: v_add3_u32 v4, v4, v0, 0x7fff
+; GFX12-TRUE16-NEXT: v_lshrrev_b32_e32 v1, s4, v5
+; GFX12-TRUE16-NEXT: v_lshlrev_b32_e32 v1, 16, v1
+; GFX12-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
+; GFX12-TRUE16-NEXT: v_min_num_f32_e32 v1, v1, v0
+; GFX12-TRUE16-NEXT: v_bfe_u32 v2, v1, 16, 1
+; GFX12-TRUE16-NEXT: v_or_b32_e32 v4, 0x400000, v1
+; GFX12-TRUE16-NEXT: v_cmp_u_f32_e32 vcc_lo, v1, v1
+; GFX12-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_3) | instskip(SKIP_1) | instid1(VALU_DEP_1)
+; GFX12-TRUE16-NEXT: v_add3_u32 v2, v2, v1, 0x7fff
; GFX12-TRUE16-NEXT: s_wait_alu 0xfffd
-; GFX12-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(SKIP_1) | instid1(VALU_DEP_2)
-; GFX12-TRUE16-NEXT: v_cndmask_b32_e32 v0, v4, v5, vcc_lo
-; GFX12-TRUE16-NEXT: v_mov_b16_e32 v4.h, 0
-; GFX12-TRUE16-NEXT: v_mov_b16_e32 v4.l, v0.h
-; GFX12-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(SKIP_1) | instid1(VALU_DEP_1)
-; GFX12-TRUE16-NEXT: v_lshlrev_b32_e32 v0, s4, v4
+; GFX12-TRUE16-NEXT: v_cndmask_b32_e32 v1, v2, v4, vcc_lo
+; GFX12-TRUE16-NEXT: v_mov_b16_e32 v2.h, 0
+; GFX12-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_2) | instskip(NEXT) | instid1(VALU_DEP_1)
+; GFX12-TRUE16-NEXT: v_mov_b16_e32 v2.l, v1.h
+; GFX12-TRUE16-NEXT: v_lshlrev_b32_e32 v1, s4, v2
; GFX12-TRUE16-NEXT: s_wait_alu 0xfffe
-; GFX12-TRUE16-NEXT: v_and_or_b32 v0, v1, s6, v0
-; GFX12-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1)
-; GFX12-TRUE16-NEXT: v_dual_mov_b32 v5, v1 :: v_dual_mov_b32 v4, v0
-; GFX12-TRUE16-NEXT: buffer_atomic_cmpswap_b32 v[4:5], v2, s[0:3], null offen th:TH_ATOMIC_RETURN
+; GFX12-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
+; GFX12-TRUE16-NEXT: v_and_or_b32 v4, v5, s6, v1
+; GFX12-TRUE16-NEXT: v_dual_mov_b32 v1, v4 :: v_dual_mov_b32 v2, v5
+; GFX12-TRUE16-NEXT: buffer_atomic_cmpswap_b32 v[1:2], v3, s[0:3], null offen th:TH_ATOMIC_RETURN
; GFX12-TRUE16-NEXT: s_wait_loadcnt 0x0
; GFX12-TRUE16-NEXT: global_inv scope:SCOPE_DEV
-; GFX12-TRUE16-NEXT: v_cmp_eq_u32_e32 vcc_lo, v4, v1
-; GFX12-TRUE16-NEXT: v_mov_b32_e32 v1, v4
+; GFX12-TRUE16-NEXT: v_cmp_eq_u32_e32 vcc_lo, v1, v5
; GFX12-TRUE16-NEXT: s_or_b32 s5, vcc_lo, s5
; GFX12-TRUE16-NEXT: s_wait_alu 0xfffe
; GFX12-TRUE16-NEXT: s_and_not1_b32 exec_lo, exec_lo, s5
@@ -4801,47 +4818,47 @@ define void @buffer_fat_ptr_agent_atomic_fmin_noret_bf16__offset__amdgpu_no_fine
; GFX12-FAKE16-NEXT: s_wait_bvhcnt 0x0
; GFX12-FAKE16-NEXT: s_wait_kmcnt 0x0
; GFX12-FAKE16-NEXT: s_addk_co_i32 s16, 0x200
-; GFX12-FAKE16-NEXT: v_lshlrev_b32_e32 v3, 16, v0
+; GFX12-FAKE16-NEXT: v_lshlrev_b32_e32 v0, 16, v0
; GFX12-FAKE16-NEXT: s_wait_alu 0xfffe
; GFX12-FAKE16-NEXT: s_and_b32 s4, s16, -4
; GFX12-FAKE16-NEXT: s_wait_alu 0xfffe
-; GFX12-FAKE16-NEXT: v_mov_b32_e32 v2, s4
+; GFX12-FAKE16-NEXT: v_mov_b32_e32 v3, s4
; GFX12-FAKE16-NEXT: s_and_b32 s4, s16, 3
; GFX12-FAKE16-NEXT: s_wait_alu 0xfffe
; GFX12-FAKE16-NEXT: s_lshl_b32 s4, s4, 3
; GFX12-FAKE16-NEXT: s_wait_alu 0xfffe
; GFX12-FAKE16-NEXT: s_lshl_b32 s5, 0xffff, s4
-; GFX12-FAKE16-NEXT: buffer_load_b32 v1, v2, s[0:3], null offen
+; GFX12-FAKE16-NEXT: buffer_load_b32 v1, v3, s[0:3], null offen
; GFX12-FAKE16-NEXT: s_not_b32 s6, s5
; GFX12-FAKE16-NEXT: s_mov_b32 s5, 0
; GFX12-FAKE16-NEXT: .LBB14_1: ; %atomicrmw.start
; GFX12-FAKE16-NEXT: ; =>This Inner Loop Header: Depth=1
; GFX12-FAKE16-NEXT: s_wait_loadcnt 0x0
-; GFX12-FAKE16-NEXT: v_lshrrev_b32_e32 v0, s4, v1
+; GFX12-FAKE16-NEXT: v_mov_b32_e32 v5, v1
; GFX12-FAKE16-NEXT: s_wait_storecnt 0x0
; GFX12-FAKE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
-; GFX12-FAKE16-NEXT: v_lshlrev_b32_e32 v0, 16, v0
-; GFX12-FAKE16-NEXT: v_min_num_f32_e32 v0, v0, v3
-; GFX12-FAKE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(SKIP_2) | instid1(VALU_DEP_3)
-; GFX12-FAKE16-NEXT: v_bfe_u32 v4, v0, 16, 1
-; GFX12-FAKE16-NEXT: v_or_b32_e32 v5, 0x400000, v0
-; GFX12-FAKE16-NEXT: v_cmp_u_f32_e32 vcc_lo, v0, v0
-; GFX12-FAKE16-NEXT: v_add3_u32 v4, v4, v0, 0x7fff
+; GFX12-FAKE16-NEXT: v_lshrrev_b32_e32 v1, s4, v5
+; GFX12-FAKE16-NEXT: v_lshlrev_b32_e32 v1, 16, v1
+; GFX12-FAKE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
+; GFX12-FAKE16-NEXT: v_min_num_f32_e32 v1, v1, v0
+; GFX12-FAKE16-NEXT: v_bfe_u32 v2, v1, 16, 1
+; GFX12-FAKE16-NEXT: v_or_b32_e32 v4, 0x400000, v1
+; GFX12-FAKE16-NEXT: v_cmp_u_f32_e32 vcc_lo, v1, v1
+; GFX12-FAKE16-NEXT: s_delay_alu instid0(VALU_DEP_3) | instskip(SKIP_1) | instid1(VALU_DEP_1)
+; GFX12-FAKE16-NEXT: v_add3_u32 v2, v2, v1, 0x7fff
; GFX12-FAKE16-NEXT: s_wait_alu 0xfffd
+; GFX12-FAKE16-NEXT: v_cndmask_b32_e32 v1, v2, v4, vcc_lo
; GFX12-FAKE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
-; GFX12-FAKE16-NEXT: v_cndmask_b32_e32 v0, v4, v5, vcc_lo
-; GFX12-FAKE16-NEXT: v_lshrrev_b32_e32 v0, 16, v0
-; GFX12-FAKE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(SKIP_1) | instid1(VALU_DEP_1)
-; GFX12-FAKE16-NEXT: v_lshlrev_b32_e32 v0, s4, v0
+; GFX12-FAKE16-NEXT: v_lshrrev_b32_e32 v1, 16, v1
+; GFX12-FAKE16-NEXT: v_lshlrev_b32_e32 v1, s4, v1
; GFX12-FAKE16-NEXT: s_wait_alu 0xfffe
-; GFX12-FAKE16-NEXT: v_and_or_b32 v0, v1, s6, v0
-; GFX12-FAKE16-NEXT: s_delay_alu instid0(VALU_DEP_1)
-; GFX12-FAKE16-NEXT: v_dual_mov_b32 v5, v1 :: v_dual_mov_b32 v4, v0
-; GFX12-FAKE16-NEXT: buffer_atomic_cmpswap_b32 v[4:5], v2, s[0:3], null offen th:TH_ATOMIC_RETURN
+; GFX12-FAKE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
+; GFX12-FAKE16-NEXT: v_and_or_b32 v4, v5, s6, v1
+; GFX12-FAKE16-NEXT: v_dual_mov_b32 v1, v4 :: v_dual_mov_b32 v2, v5
+; GFX12-FAKE16-NEXT: buffer_atomic_cmpswap_b32 v[1:2], v3, s[0:3], null offen th:TH_ATOMIC_RETURN
; GFX12-FAKE16-NEXT: s_wait_loadcnt 0x0
; GFX12-FAKE16-NEXT: global_inv scope:SCOPE_DEV
-; GFX12-FAKE16-NEXT: v_cmp_eq_u32_e32 vcc_lo, v4, v1
-; GFX12-FAKE16-NEXT: v_mov_b32_e32 v1, v4
+; GFX12-FAKE16-NEXT: v_cmp_eq_u32_e32 vcc_lo, v1, v5
; GFX12-FAKE16-NEXT: s_or_b32 s5, vcc_lo, s5
; GFX12-FAKE16-NEXT: s_wait_alu 0xfffe
; GFX12-FAKE16-NEXT: s_and_not1_b32 exec_lo, exec_lo, s5
@@ -4856,36 +4873,36 @@ define void @buffer_fat_ptr_agent_atomic_fmin_noret_bf16__offset__amdgpu_no_fine
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX942-NEXT: s_addk_i32 s16, 0x200
; GFX942-NEXT: s_and_b32 s4, s16, -4
-; GFX942-NEXT: v_mov_b32_e32 v2, s4
-; GFX942-NEXT: buffer_load_dword v1, v2, s[0:3], 0 offen
+; GFX942-NEXT: v_mov_b32_e32 v1, s4
+; GFX942-NEXT: buffer_load_dword v2, v1, s[0:3], 0 offen
; GFX942-NEXT: s_and_b32 s4, s16, 3
; GFX942-NEXT: s_lshl_b32 s6, s4, 3
; GFX942-NEXT: s_lshl_b32 s4, 0xffff, s6
; GFX942-NEXT: s_not_b32 s7, s4
; GFX942-NEXT: s_mov_b64 s[4:5], 0
-; GFX942-NEXT: v_lshlrev_b32_e32 v3, 16, v0
+; GFX942-NEXT: v_lshlrev_b32_e32 v0, 16, v0
; GFX942-NEXT: s_movk_i32 s8, 0x7fff
; GFX942-NEXT: .LBB14_1: ; %atomicrmw.start
; GFX942-NEXT: ; =>This Inner Loop Header: Depth=1
; GFX942-NEXT: s_waitcnt vmcnt(0)
-; GFX942-NEXT: v_lshrrev_b32_sdwa v0, s6, v1 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:DWORD
+; GFX942-NEXT: v_mov_b32_e32 v5, v2
+; GFX942-NEXT: v_lshrrev_b32_sdwa v2, s6, v5 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:DWORD
; GFX942-NEXT: buffer_wbl2 sc1
-; GFX942-NEXT: v_min_f32_e32 v0, v0, v3
-; GFX942-NEXT: v_bfe_u32 v4, v0, 16, 1
-; GFX942-NEXT: v_or_b32_e32 v5, 0x400000, v0
-; GFX942-NEXT: v_add3_u32 v4, v4, v0, s8
-; GFX942-NEXT: v_cmp_u_f32_e32 vcc, v0, v0
+; GFX942-NEXT: v_min_f32_e32 v2, v2, v0
+; GFX942-NEXT: v_bfe_u32 v3, v2, 16, 1
+; GFX942-NEXT: v_or_b32_e32 v4, 0x400000, v2
+; GFX942-NEXT: v_add3_u32 v3, v3, v2, s8
+; GFX942-NEXT: v_cmp_u_f32_e32 vcc, v2, v2
; GFX942-NEXT: s_nop 1
-; GFX942-NEXT: v_cndmask_b32_e32 v0, v4, v5, vcc
-; GFX942-NEXT: v_lshlrev_b32_sdwa v0, s6, v0 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:WORD_1
-; GFX942-NEXT: v_and_or_b32 v0, v1, s7, v0
-; GFX942-NEXT: v_mov_b64_e32 v[4:5], v[0:1]
-; GFX942-NEXT: buffer_atomic_cmpswap v[4:5], v2, s[0:3], 0 offen sc0
+; GFX942-NEXT: v_cndmask_b32_e32 v2, v3, v4, vcc
+; GFX942-NEXT: v_lshlrev_b32_sdwa v2, s6, v2 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:WORD_1
+; GFX942-NEXT: v_and_or_b32 v4, v5, s7, v2
+; GFX942-NEXT: v_mov_b64_e32 v[2:3], v[4:5]
+; GFX942-NEXT: buffer_atomic_cmpswap v[2:3], v1, s[0:3], 0 offen sc0
; GFX942-NEXT: s_waitcnt vmcnt(0)
; GFX942-NEXT: buffer_inv sc1
-; GFX942-NEXT: v_cmp_eq_u32_e32 vcc, v4, v1
+; GFX942-NEXT: v_cmp_eq_u32_e32 vcc, v2, v5
; GFX942-NEXT: s_or_b64 s[4:5], vcc, s[4:5]
-; GFX942-NEXT: v_mov_b32_e32 v1, v4
; GFX942-NEXT: s_andn2_b64 exec, exec, s[4:5]
; GFX942-NEXT: s_cbranch_execnz .LBB14_1
; GFX942-NEXT: ; %bb.2: ; %atomicrmw.end
@@ -4896,46 +4913,46 @@ define void @buffer_fat_ptr_agent_atomic_fmin_noret_bf16__offset__amdgpu_no_fine
; GFX11-TRUE16: ; %bb.0:
; GFX11-TRUE16-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX11-TRUE16-NEXT: s_addk_i32 s16, 0x200
-; GFX11-TRUE16-NEXT: v_lshlrev_b32_e32 v3, 16, v0
+; GFX11-TRUE16-NEXT: v_lshlrev_b32_e32 v0, 16, v0
; GFX11-TRUE16-NEXT: s_and_b32 s4, s16, -4
; GFX11-TRUE16-NEXT: s_delay_alu instid0(SALU_CYCLE_1) | instskip(SKIP_1) | instid1(SALU_CYCLE_1)
-; GFX11-TRUE16-NEXT: v_mov_b32_e32 v2, s4
+; GFX11-TRUE16-NEXT: v_mov_b32_e32 v3, s4
; GFX11-TRUE16-NEXT: s_and_b32 s4, s16, 3
; GFX11-TRUE16-NEXT: s_lshl_b32 s4, s4, 3
; GFX11-TRUE16-NEXT: s_delay_alu instid0(SALU_CYCLE_1)
; GFX11-TRUE16-NEXT: s_lshl_b32 s5, 0xffff, s4
-; GFX11-TRUE16-NEXT: buffer_load_b32 v1, v2, s[0:3], 0 offen
+; GFX11-TRUE16-NEXT: buffer_load_b32 v1, v3, s[0:3], 0 offen
; GFX11-TRUE16-NEXT: s_not_b32 s6, s5
; GFX11-TRUE16-NEXT: s_mov_b32 s5, 0
; GFX11-TRUE16-NEXT: .p2align 6
; GFX11-TRUE16-NEXT: .LBB14_1: ; %atomicrmw.start
; GFX11-TRUE16-NEXT: ; =>This Inner Loop Header: Depth=1
; GFX11-TRUE16-NEXT: s_waitcnt vmcnt(0)
-; GFX11-TRUE16-NEXT: v_lshrrev_b32_e32 v0, s4, v1
+; GFX11-TRUE16-NEXT: v_mov_b32_e32 v5, v1
; GFX11-TRUE16-NEXT: s_waitcnt_vscnt null, 0x0
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
-; GFX11-TRUE16-NEXT: v_lshlrev_b32_e32 v0, 16, v0
-; GFX11-TRUE16-NEXT: v_min_f32_e32 v0, v0, v3
-; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(SKIP_2) | instid1(VALU_DEP_3)
-; GFX11-TRUE16-NEXT: v_bfe_u32 v4, v0, 16, 1
-; GFX11-TRUE16-NEXT: v_or_b32_e32 v5, 0x400000, v0
-; GFX11-TRUE16-NEXT: v_cmp_u_f32_e32 vcc_lo, v0, v0
-; GFX11-TRUE16-NEXT: v_add3_u32 v4, v4, v0, 0x7fff
-; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(SKIP_1) | instid1(VALU_DEP_2)
-; GFX11-TRUE16-NEXT: v_cndmask_b32_e32 v0, v4, v5, vcc_lo
-; GFX11-TRUE16-NEXT: v_mov_b16_e32 v4.h, 0
-; GFX11-TRUE16-NEXT: v_mov_b16_e32 v4.l, v0.h
+; GFX11-TRUE16-NEXT: v_lshrrev_b32_e32 v1, s4, v5
+; GFX11-TRUE16-NEXT: v_lshlrev_b32_e32 v1, 16, v1
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
-; GFX11-TRUE16-NEXT: v_lshlrev_b32_e32 v0, s4, v4
-; GFX11-TRUE16-NEXT: v_and_or_b32 v0, v1, s6, v0
-; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1)
-; GFX11-TRUE16-NEXT: v_dual_mov_b32 v5, v1 :: v_dual_mov_b32 v4, v0
-; GFX11-TRUE16-NEXT: buffer_atomic_cmpswap_b32 v[4:5], v2, s[0:3], 0 offen glc
+; GFX11-TRUE16-NEXT: v_min_f32_e32 v1, v1, v0
+; GFX11-TRUE16-NEXT: v_bfe_u32 v2, v1, 16, 1
+; GFX11-TRUE16-NEXT: v_or_b32_e32 v4, 0x400000, v1
+; GFX11-TRUE16-NEXT: v_cmp_u_f32_e32 vcc_lo, v1, v1
+; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_3) | instskip(NEXT) | instid1(VALU_DEP_1)
+; GFX11-TRUE16-NEXT: v_add3_u32 v2, v2, v1, 0x7fff
+; GFX11-TRUE16-NEXT: v_cndmask_b32_e32 v1, v2, v4, vcc_lo
+; GFX11-TRUE16-NEXT: v_mov_b16_e32 v2.h, 0
+; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_2) | instskip(NEXT) | instid1(VALU_DEP_1)
+; GFX11-TRUE16-NEXT: v_mov_b16_e32 v2.l, v1.h
+; GFX11-TRUE16-NEXT: v_lshlrev_b32_e32 v1, s4, v2
+; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
+; GFX11-TRUE16-NEXT: v_and_or_b32 v4, v5, s6, v1
+; GFX11-TRUE16-NEXT: v_dual_mov_b32 v1, v4 :: v_dual_mov_b32 v2, v5
+; GFX11-TRUE16-NEXT: buffer_atomic_cmpswap_b32 v[1:2], v3, s[0:3], 0 offen glc
; GFX11-TRUE16-NEXT: s_waitcnt vmcnt(0)
; GFX11-TRUE16-NEXT: buffer_gl1_inv
; GFX11-TRUE16-NEXT: buffer_gl0_inv
-; GFX11-TRUE16-NEXT: v_cmp_eq_u32_e32 vcc_lo, v4, v1
-; GFX11-TRUE16-NEXT: v_mov_b32_e32 v1, v4
+; GFX11-TRUE16-NEXT: v_cmp_eq_u32_e32 vcc_lo, v1, v5
; GFX11-TRUE16-NEXT: s_or_b32 s5, vcc_lo, s5
; GFX11-TRUE16-NEXT: s_delay_alu instid0(SALU_CYCLE_1)
; GFX11-TRUE16-NEXT: s_and_not1_b32 exec_lo, exec_lo, s5
@@ -4948,45 +4965,45 @@ define void @buffer_fat_ptr_agent_atomic_fmin_noret_bf16__offset__amdgpu_no_fine
; GFX11-FAKE16: ; %bb.0:
; GFX11-FAKE16-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX11-FAKE16-NEXT: s_addk_i32 s16, 0x200
-; GFX11-FAKE16-NEXT: v_lshlrev_b32_e32 v3, 16, v0
+; GFX11-FAKE16-NEXT: v_lshlrev_b32_e32 v0, 16, v0
; GFX11-FAKE16-NEXT: s_and_b32 s4, s16, -4
; GFX11-FAKE16-NEXT: s_delay_alu instid0(SALU_CYCLE_1) | instskip(SKIP_1) | instid1(SALU_CYCLE_1)
-; GFX11-FAKE16-NEXT: v_mov_b32_e32 v2, s4
+; GFX11-FAKE16-NEXT: v_mov_b32_e32 v3, s4
; GFX11-FAKE16-NEXT: s_and_b32 s4, s16, 3
; GFX11-FAKE16-NEXT: s_lshl_b32 s4, s4, 3
; GFX11-FAKE16-NEXT: s_delay_alu instid0(SALU_CYCLE_1)
; GFX11-FAKE16-NEXT: s_lshl_b32 s5, 0xffff, s4
-; GFX11-FAKE16-NEXT: buffer_load_b32 v1, v2, s[0:3], 0 offen
+; GFX11-FAKE16-NEXT: buffer_load_b32 v1, v3, s[0:3], 0 offen
; GFX11-FAKE16-NEXT: s_not_b32 s6, s5
; GFX11-FAKE16-NEXT: s_mov_b32 s5, 0
; GFX11-FAKE16-NEXT: .p2align 6
; GFX11-FAKE16-NEXT: .LBB14_1: ; %atomicrmw.start
; GFX11-FAKE16-NEXT: ; =>This Inner Loop Header: Depth=1
; GFX11-FAKE16-NEXT: s_waitcnt vmcnt(0)
-; GFX11-FAKE16-NEXT: v_lshrrev_b32_e32 v0, s4, v1
+; GFX11-FAKE16-NEXT: v_mov_b32_e32 v5, v1
; GFX11-FAKE16-NEXT: s_waitcnt_vscnt null, 0x0
; GFX11-FAKE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
-; GFX11-FAKE16-NEXT: v_lshlrev_b32_e32 v0, 16, v0
-; GFX11-FAKE16-NEXT: v_min_f32_e32 v0, v0, v3
-; GFX11-FAKE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(SKIP_2) | instid1(VALU_DEP_3)
-; GFX11-FAKE16-NEXT: v_bfe_u32 v4, v0, 16, 1
-; GFX11-FAKE16-NEXT: v_or_b32_e32 v5, 0x400000, v0
-; GFX11-FAKE16-NEXT: v_cmp_u_f32_e32 vcc_lo, v0, v0
-; GFX11-FAKE16-NEXT: v_add3_u32 v4, v4, v0, 0x7fff
+; GFX11-FAKE16-NEXT: v_lshrrev_b32_e32 v1, s4, v5
+; GFX11-FAKE16-NEXT: v_lshlrev_b32_e32 v1, 16, v1
; GFX11-FAKE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
-; GFX11-FAKE16-NEXT: v_cndmask_b32_e32 v0, v4, v5, vcc_lo
-; GFX11-FAKE16-NEXT: v_lshrrev_b32_e32 v0, 16, v0
+; GFX11-FAKE16-NEXT: v_min_f32_e32 v1, v1, v0
+; GFX11-FAKE16-NEXT: v_bfe_u32 v2, v1, 16, 1
+; GFX11-FAKE16-NEXT: v_or_b32_e32 v4, 0x400000, v1
+; GFX11-FAKE16-NEXT: v_cmp_u_f32_e32 vcc_lo, v1, v1
+; GFX11-FAKE16-NEXT: s_delay_alu instid0(VALU_DEP_3) | instskip(NEXT) | instid1(VALU_DEP_1)
+; GFX11-FAKE16-NEXT: v_add3_u32 v2, v2, v1, 0x7fff
+; GFX11-FAKE16-NEXT: v_cndmask_b32_e32 v1, v2, v4, vcc_lo
; GFX11-FAKE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
-; GFX11-FAKE16-NEXT: v_lshlrev_b32_e32 v0, s4, v0
-; GFX11-FAKE16-NEXT: v_and_or_b32 v0, v1, s6, v0
-; GFX11-FAKE16-NEXT: s_delay_alu instid0(VALU_DEP_1)
-; GFX11-FAKE16-NEXT: v_dual_mov_b32 v5, v1 :: v_dual_mov_b32 v4, v0
-; GFX11-FAKE16-NEXT: buffer_atomic_cmpswap_b32 v[4:5], v2, s[0:3], 0 offen glc
+; GFX11-FAKE16-NEXT: v_lshrrev_b32_e32 v1, 16, v1
+; GFX11-FAKE16-NEXT: v_lshlrev_b32_e32 v1, s4, v1
+; GFX11-FAKE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
+; GFX11-FAKE16-NEXT: v_and_or_b32 v4, v5, s6, v1
+; GFX11-FAKE16-NEXT: v_dual_mov_b32 v1, v4 :: v_dual_mov_b32 v2, v5
+; GFX11-FAKE16-NEXT: buffer_atomic_cmpswap_b32 v[1:2], v3, s[0:3], 0 offen glc
; GFX11-FAKE16-NEXT: s_waitcnt vmcnt(0)
; GFX11-FAKE16-NEXT: buffer_gl1_inv
; GFX11-FAKE16-NEXT: buffer_gl0_inv
-; GFX11-FAKE16-NEXT: v_cmp_eq_u32_e32 vcc_lo, v4, v1
-; GFX11-FAKE16-NEXT: v_mov_b32_e32 v1, v4
+; GFX11-FAKE16-NEXT: v_cmp_eq_u32_e32 vcc_lo, v1, v5
; GFX11-FAKE16-NEXT: s_or_b32 s5, vcc_lo, s5
; GFX11-FAKE16-NEXT: s_delay_alu instid0(SALU_CYCLE_1)
; GFX11-FAKE16-NEXT: s_and_not1_b32 exec_lo, exec_lo, s5
@@ -4999,36 +5016,36 @@ define void @buffer_fat_ptr_agent_atomic_fmin_noret_bf16__offset__amdgpu_no_fine
; GFX10: ; %bb.0:
; GFX10-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX10-NEXT: s_addk_i32 s20, 0x200
-; GFX10-NEXT: v_lshlrev_b32_e32 v3, 16, v0
+; GFX10-NEXT: v_lshlrev_b32_e32 v0, 16, v0
; GFX10-NEXT: s_and_b32 s4, s20, -4
-; GFX10-NEXT: v_mov_b32_e32 v2, s4
+; GFX10-NEXT: v_mov_b32_e32 v3, s4
; GFX10-NEXT: s_and_b32 s4, s20, 3
; GFX10-NEXT: s_lshl_b32 s4, s4, 3
; GFX10-NEXT: s_lshl_b32 s5, 0xffff, s4
-; GFX10-NEXT: buffer_load_dword v1, v2, s[16:19], 0 offen
+; GFX10-NEXT: buffer_load_dword v1, v3, s[16:19], 0 offen
; GFX10-NEXT: s_not_b32 s6, s5
; GFX10-NEXT: s_mov_b32 s5, 0
; GFX10-NEXT: .LBB14_1: ; %atomicrmw.start
; GFX10-NEXT: ; =>This Inner Loop Header: Depth=1
; GFX10-NEXT: s_waitcnt vmcnt(0)
-; GFX10-NEXT: v_lshrrev_b32_sdwa v0, s4, v1 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:DWORD
-; GFX10-NEXT: s_waitcnt_vscnt null, 0x0
-; GFX10-NEXT: v_min_f32_e32 v0, v0, v3
-; GFX10-NEXT: v_bfe_u32 v4, v0, 16, 1
-; GFX10-NEXT: v_or_b32_e32 v5, 0x400000, v0
-; GFX10-NEXT: v_cmp_u_f32_e32 vcc_lo, v0, v0
-; GFX10-NEXT: v_add3_u32 v4, v4, v0, 0x7fff
-; GFX10-NEXT: v_cndmask_b32_e32 v0, v4, v5, vcc_lo
-; GFX10-NEXT: v_lshlrev_b32_sdwa v0, s4, v0 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:WORD_1
-; GFX10-NEXT: v_and_or_b32 v0, v1, s6, v0
; GFX10-NEXT: v_mov_b32_e32 v5, v1
-; GFX10-NEXT: v_mov_b32_e32 v4, v0
-; GFX10-NEXT: buffer_atomic_cmpswap v[4:5], v2, s[16:19], 0 offen glc
+; GFX10-NEXT: s_waitcnt_vscnt null, 0x0
+; GFX10-NEXT: v_lshrrev_b32_sdwa v1, s4, v5 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:DWORD
+; GFX10-NEXT: v_min_f32_e32 v1, v1, v0
+; GFX10-NEXT: v_bfe_u32 v2, v1, 16, 1
+; GFX10-NEXT: v_or_b32_e32 v4, 0x400000, v1
+; GFX10-NEXT: v_cmp_u_f32_e32 vcc_lo, v1, v1
+; GFX10-NEXT: v_add3_u32 v2, v2, v1, 0x7fff
+; GFX10-NEXT: v_cndmask_b32_e32 v1, v2, v4, vcc_lo
+; GFX10-NEXT: v_lshlrev_b32_sdwa v1, s4, v1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:WORD_1
+; GFX10-NEXT: v_and_or_b32 v4, v5, s6, v1
+; GFX10-NEXT: v_mov_b32_e32 v1, v4
+; GFX10-NEXT: v_mov_b32_e32 v2, v5
+; GFX10-NEXT: buffer_atomic_cmpswap v[1:2], v3, s[16:19], 0 offen glc
; GFX10-NEXT: s_waitcnt vmcnt(0)
; GFX10-NEXT: buffer_gl1_inv
; GFX10-NEXT: buffer_gl0_inv
-; GFX10-NEXT: v_cmp_eq_u32_e32 vcc_lo, v4, v1
-; GFX10-NEXT: v_mov_b32_e32 v1, v4
+; GFX10-NEXT: v_cmp_eq_u32_e32 vcc_lo, v1, v5
; GFX10-NEXT: s_or_b32 s5, vcc_lo, s5
; GFX10-NEXT: s_andn2_b32 exec_lo, exec_lo, s5
; GFX10-NEXT: s_cbranch_execnz .LBB14_1
@@ -5041,34 +5058,34 @@ define void @buffer_fat_ptr_agent_atomic_fmin_noret_bf16__offset__amdgpu_no_fine
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX90A-NEXT: s_addk_i32 s20, 0x200
; GFX90A-NEXT: s_and_b32 s4, s20, -4
-; GFX90A-NEXT: v_mov_b32_e32 v2, s4
-; GFX90A-NEXT: buffer_load_dword v1, v2, s[16:19], 0 offen
+; GFX90A-NEXT: v_mov_b32_e32 v1, s4
+; GFX90A-NEXT: buffer_load_dword v2, v1, s[16:19], 0 offen
; GFX90A-NEXT: s_and_b32 s4, s20, 3
; GFX90A-NEXT: s_lshl_b32 s6, s4, 3
; GFX90A-NEXT: s_lshl_b32 s4, 0xffff, s6
; GFX90A-NEXT: s_not_b32 s7, s4
; GFX90A-NEXT: s_mov_b64 s[4:5], 0
-; GFX90A-NEXT: v_lshlrev_b32_e32 v3, 16, v0
+; GFX90A-NEXT: v_lshlrev_b32_e32 v0, 16, v0
; GFX90A-NEXT: s_movk_i32 s8, 0x7fff
; GFX90A-NEXT: .LBB14_1: ; %atomicrmw.start
; GFX90A-NEXT: ; =>This Inner Loop Header: Depth=1
; GFX90A-NEXT: s_waitcnt vmcnt(0)
-; GFX90A-NEXT: v_lshrrev_b32_sdwa v0, s6, v1 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:DWORD
-; GFX90A-NEXT: v_min_f32_e32 v0, v0, v3
-; GFX90A-NEXT: v_bfe_u32 v4, v0, 16, 1
-; GFX90A-NEXT: v_or_b32_e32 v5, 0x400000, v0
-; GFX90A-NEXT: v_add3_u32 v4, v4, v0, s8
-; GFX90A-NEXT: v_cmp_u_f32_e32 vcc, v0, v0
-; GFX90A-NEXT: v_cndmask_b32_e32 v0, v4, v5, vcc
-; GFX90A-NEXT: v_lshlrev_b32_sdwa v0, s6, v0 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:WORD_1
-; GFX90A-NEXT: v_and_or_b32 v0, v1, s7, v0
-; GFX90A-NEXT: v_pk_mov_b32 v[4:5], v[0:1], v[0:1] op_sel:[0,1]
-; GFX90A-NEXT: buffer_atomic_cmpswap v[4:5], v2, s[16:19], 0 offen glc
+; GFX90A-NEXT: v_mov_b32_e32 v5, v2
+; GFX90A-NEXT: v_lshrrev_b32_sdwa v2, s6, v5 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:DWORD
+; GFX90A-NEXT: v_min_f32_e32 v2, v2, v0
+; GFX90A-NEXT: v_bfe_u32 v3, v2, 16, 1
+; GFX90A-NEXT: v_or_b32_e32 v4, 0x400000, v2
+; GFX90A-NEXT: v_add3_u32 v3, v3, v2, s8
+; GFX90A-NEXT: v_cmp_u_f32_e32 vcc, v2, v2
+; GFX90A-NEXT: v_cndmask_b32_e32 v2, v3, v4, vcc
+; GFX90A-NEXT: v_lshlrev_b32_sdwa v2, s6, v2 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:WORD_1
+; GFX90A-NEXT: v_and_or_b32 v4, v5, s7, v2
+; GFX90A-NEXT: v_pk_mov_b32 v[2:3], v[4:5], v[4:5] op_sel:[0,1]
+; GFX90A-NEXT: buffer_atomic_cmpswap v[2:3], v1, s[16:19], 0 offen glc
; GFX90A-NEXT: s_waitcnt vmcnt(0)
; GFX90A-NEXT: buffer_wbinvl1
-; GFX90A-NEXT: v_cmp_eq_u32_e32 vcc, v4, v1
+; GFX90A-NEXT: v_cmp_eq_u32_e32 vcc, v2, v5
; GFX90A-NEXT: s_or_b64 s[4:5], vcc, s[4:5]
-; GFX90A-NEXT: v_mov_b32_e32 v1, v4
; GFX90A-NEXT: s_andn2_b64 exec, exec, s[4:5]
; GFX90A-NEXT: s_cbranch_execnz .LBB14_1
; GFX90A-NEXT: ; %bb.2: ; %atomicrmw.end
@@ -5080,35 +5097,35 @@ define void @buffer_fat_ptr_agent_atomic_fmin_noret_bf16__offset__amdgpu_no_fine
; GFX908-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX908-NEXT: s_addk_i32 s20, 0x200
; GFX908-NEXT: s_and_b32 s4, s20, -4
-; GFX908-NEXT: v_mov_b32_e32 v2, s4
-; GFX908-NEXT: buffer_load_dword v1, v2, s[16:19], 0 offen
+; GFX908-NEXT: v_mov_b32_e32 v3, s4
+; GFX908-NEXT: buffer_load_dword v1, v3, s[16:19], 0 offen
; GFX908-NEXT: s_and_b32 s4, s20, 3
; GFX908-NEXT: s_lshl_b32 s6, s4, 3
; GFX908-NEXT: s_lshl_b32 s4, 0xffff, s6
; GFX908-NEXT: s_not_b32 s7, s4
; GFX908-NEXT: s_mov_b64 s[4:5], 0
-; GFX908-NEXT: v_lshlrev_b32_e32 v3, 16, v0
+; GFX908-NEXT: v_lshlrev_b32_e32 v0, 16, v0
; GFX908-NEXT: s_movk_i32 s8, 0x7fff
; GFX908-NEXT: .LBB14_1: ; %atomicrmw.start
; GFX908-NEXT: ; =>This Inner Loop Header: Depth=1
; GFX908-NEXT: s_waitcnt vmcnt(0)
-; GFX908-NEXT: v_lshrrev_b32_sdwa v0, s6, v1 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:DWORD
-; GFX908-NEXT: v_min_f32_e32 v0, v0, v3
-; GFX908-NEXT: v_bfe_u32 v4, v0, 16, 1
-; GFX908-NEXT: v_or_b32_e32 v5, 0x400000, v0
-; GFX908-NEXT: v_add3_u32 v4, v4, v0, s8
-; GFX908-NEXT: v_cmp_u_f32_e32 vcc, v0, v0
-; GFX908-NEXT: v_cndmask_b32_e32 v0, v4, v5, vcc
-; GFX908-NEXT: v_lshlrev_b32_sdwa v0, s6, v0 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:WORD_1
-; GFX908-NEXT: v_and_or_b32 v0, v1, s7, v0
; GFX908-NEXT: v_mov_b32_e32 v5, v1
-; GFX908-NEXT: v_mov_b32_e32 v4, v0
-; GFX908-NEXT: buffer_atomic_cmpswap v[4:5], v2, s[16:19], 0 offen glc
+; GFX908-NEXT: v_lshrrev_b32_sdwa v1, s6, v5 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:DWORD
+; GFX908-NEXT: v_min_f32_e32 v1, v1, v0
+; GFX908-NEXT: v_bfe_u32 v2, v1, 16, 1
+; GFX908-NEXT: v_or_b32_e32 v4, 0x400000, v1
+; GFX908-NEXT: v_add3_u32 v2, v2, v1, s8
+; GFX908-NEXT: v_cmp_u_f32_e32 vcc, v1, v1
+; GFX908-NEXT: v_cndmask_b32_e32 v1, v2, v4, vcc
+; GFX908-NEXT: v_lshlrev_b32_sdwa v1, s6, v1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:WORD_1
+; GFX908-NEXT: v_and_or_b32 v4, v5, s7, v1
+; GFX908-NEXT: v_mov_b32_e32 v1, v4
+; GFX908-NEXT: v_mov_b32_e32 v2, v5
+; GFX908-NEXT: buffer_atomic_cmpswap v[1:2], v3, s[16:19], 0 offen glc
; GFX908-NEXT: s_waitcnt vmcnt(0)
; GFX908-NEXT: buffer_wbinvl1
-; GFX908-NEXT: v_cmp_eq_u32_e32 vcc, v4, v1
+; GFX908-NEXT: v_cmp_eq_u32_e32 vcc, v1, v5
; GFX908-NEXT: s_or_b64 s[4:5], vcc, s[4:5]
-; GFX908-NEXT: v_mov_b32_e32 v1, v4
; GFX908-NEXT: s_andn2_b64 exec, exec, s[4:5]
; GFX908-NEXT: s_cbranch_execnz .LBB14_1
; GFX908-NEXT: ; %bb.2: ; %atomicrmw.end
@@ -5120,37 +5137,37 @@ define void @buffer_fat_ptr_agent_atomic_fmin_noret_bf16__offset__amdgpu_no_fine
; GFX8-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX8-NEXT: s_addk_i32 s20, 0x200
; GFX8-NEXT: s_and_b32 s4, s20, -4
-; GFX8-NEXT: v_mov_b32_e32 v2, s4
-; GFX8-NEXT: buffer_load_dword v1, v2, s[16:19], 0 offen
+; GFX8-NEXT: v_mov_b32_e32 v3, s4
+; GFX8-NEXT: buffer_load_dword v1, v3, s[16:19], 0 offen
; GFX8-NEXT: s_and_b32 s4, s20, 3
; GFX8-NEXT: s_lshl_b32 s6, s4, 3
; GFX8-NEXT: s_lshl_b32 s4, 0xffff, s6
; GFX8-NEXT: s_not_b32 s7, s4
; GFX8-NEXT: s_mov_b64 s[4:5], 0
-; GFX8-NEXT: v_lshlrev_b32_e32 v3, 16, v0
+; GFX8-NEXT: v_lshlrev_b32_e32 v0, 16, v0
; GFX8-NEXT: .LBB14_1: ; %atomicrmw.start
; GFX8-NEXT: ; =>This Inner Loop Header: Depth=1
-; GFX8-NEXT: v_mov_b32_e32 v0, s6
+; GFX8-NEXT: v_mov_b32_e32 v2, s6
; GFX8-NEXT: s_waitcnt vmcnt(0)
-; GFX8-NEXT: v_lshrrev_b32_sdwa v5, v0, v1 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:DWORD
-; GFX8-NEXT: v_min_f32_e32 v5, v5, v3
-; GFX8-NEXT: v_bfe_u32 v6, v5, 16, 1
-; GFX8-NEXT: v_add_u32_e32 v6, vcc, v6, v5
-; GFX8-NEXT: v_add_u32_e32 v6, vcc, 0x7fff, v6
-; GFX8-NEXT: v_or_b32_e32 v7, 0x400000, v5
-; GFX8-NEXT: v_cmp_u_f32_e32 vcc, v5, v5
-; GFX8-NEXT: v_cndmask_b32_e32 v5, v6, v7, vcc
-; GFX8-NEXT: v_and_b32_e32 v4, s7, v1
-; GFX8-NEXT: v_lshlrev_b32_sdwa v0, v0, v5 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:WORD_1
-; GFX8-NEXT: v_or_b32_e32 v0, v4, v0
; GFX8-NEXT: v_mov_b32_e32 v5, v1
-; GFX8-NEXT: v_mov_b32_e32 v4, v0
-; GFX8-NEXT: buffer_atomic_cmpswap v[4:5], v2, s[16:19], 0 offen glc
+; GFX8-NEXT: v_lshrrev_b32_sdwa v1, v2, v5 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:DWORD
+; GFX8-NEXT: v_min_f32_e32 v1, v1, v0
+; GFX8-NEXT: v_bfe_u32 v6, v1, 16, 1
+; GFX8-NEXT: v_add_u32_e32 v6, vcc, v6, v1
+; GFX8-NEXT: v_add_u32_e32 v6, vcc, 0x7fff, v6
+; GFX8-NEXT: v_or_b32_e32 v7, 0x400000, v1
+; GFX8-NEXT: v_cmp_u_f32_e32 vcc, v1, v1
+; GFX8-NEXT: v_cndmask_b32_e32 v1, v6, v7, vcc
+; GFX8-NEXT: v_and_b32_e32 v4, s7, v5
+; GFX8-NEXT: v_lshlrev_b32_sdwa v1, v2, v1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:WORD_1
+; GFX8-NEXT: v_or_b32_e32 v4, v4, v1
+; GFX8-NEXT: v_mov_b32_e32 v1, v4
+; GFX8-NEXT: v_mov_b32_e32 v2, v5
+; GFX8-NEXT: buffer_atomic_cmpswap v[1:2], v3, s[16:19], 0 offen glc
; GFX8-NEXT: s_waitcnt vmcnt(0)
; GFX8-NEXT: buffer_wbinvl1
-; GFX8-NEXT: v_cmp_eq_u32_e32 vcc, v4, v1
+; GFX8-NEXT: v_cmp_eq_u32_e32 vcc, v1, v5
; GFX8-NEXT: s_or_b64 s[4:5], vcc, s[4:5]
-; GFX8-NEXT: v_mov_b32_e32 v1, v4
; GFX8-NEXT: s_andn2_b64 exec, exec, s[4:5]
; GFX8-NEXT: s_cbranch_execnz .LBB14_1
; GFX8-NEXT: ; %bb.2: ; %atomicrmw.end
@@ -5162,34 +5179,34 @@ define void @buffer_fat_ptr_agent_atomic_fmin_noret_bf16__offset__amdgpu_no_fine
; GFX7-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX7-NEXT: s_addk_i32 s20, 0x200
; GFX7-NEXT: s_and_b32 s4, s20, -4
-; GFX7-NEXT: v_mov_b32_e32 v2, s4
-; GFX7-NEXT: buffer_load_dword v1, v2, s[16:19], 0 offen
+; GFX7-NEXT: v_mov_b32_e32 v3, s4
+; GFX7-NEXT: buffer_load_dword v1, v3, s[16:19], 0 offen
; GFX7-NEXT: s_and_b32 s4, s20, 3
; GFX7-NEXT: s_lshl_b32 s6, s4, 3
; GFX7-NEXT: s_lshl_b32 s4, 0xffff, s6
; GFX7-NEXT: v_mul_f32_e32 v0, 1.0, v0
; GFX7-NEXT: s_not_b32 s7, s4
; GFX7-NEXT: s_mov_b64 s[4:5], 0
-; GFX7-NEXT: v_and_b32_e32 v3, 0xffff0000, v0
+; GFX7-NEXT: v_and_b32_e32 v0, 0xffff0000, v0
; GFX7-NEXT: .LBB14_1: ; %atomicrmw.start
; GFX7-NEXT: ; =>This Inner Loop Header: Depth=1
; GFX7-NEXT: s_waitcnt vmcnt(0)
-; GFX7-NEXT: v_lshrrev_b32_e32 v0, s6, v1
-; GFX7-NEXT: v_lshlrev_b32_e32 v0, 16, v0
-; GFX7-NEXT: v_mul_f32_e32 v0, 1.0, v0
-; GFX7-NEXT: v_min_f32_e32 v0, v0, v3
-; GFX7-NEXT: v_lshrrev_b32_e32 v0, 16, v0
-; GFX7-NEXT: v_and_b32_e32 v4, s7, v1
-; GFX7-NEXT: v_lshlrev_b32_e32 v0, s6, v0
-; GFX7-NEXT: v_or_b32_e32 v0, v4, v0
; GFX7-NEXT: v_mov_b32_e32 v5, v1
-; GFX7-NEXT: v_mov_b32_e32 v4, v0
-; GFX7-NEXT: buffer_atomic_cmpswap v[4:5], v2, s[16:19], 0 offen glc
+; GFX7-NEXT: v_lshrrev_b32_e32 v1, s6, v5
+; GFX7-NEXT: v_lshlrev_b32_e32 v1, 16, v1
+; GFX7-NEXT: v_mul_f32_e32 v1, 1.0, v1
+; GFX7-NEXT: v_min_f32_e32 v1, v1, v0
+; GFX7-NEXT: v_lshrrev_b32_e32 v1, 16, v1
+; GFX7-NEXT: v_and_b32_e32 v2, s7, v5
+; GFX7-NEXT: v_lshlrev_b32_e32 v1, s6, v1
+; GFX7-NEXT: v_or_b32_e32 v4, v2, v1
+; GFX7-NEXT: v_mov_b32_e32 v1, v4
+; GFX7-NEXT: v_mov_b32_e32 v2, v5
+; GFX7-NEXT: buffer_atomic_cmpswap v[1:2], v3, s[16:19], 0 offen glc
; GFX7-NEXT: s_waitcnt vmcnt(0)
; GFX7-NEXT: buffer_wbinvl1
-; GFX7-NEXT: v_cmp_eq_u32_e32 vcc, v4, v1
+; GFX7-NEXT: v_cmp_eq_u32_e32 vcc, v1, v5
; GFX7-NEXT: s_or_b64 s[4:5], vcc, s[4:5]
-; GFX7-NEXT: v_mov_b32_e32 v1, v4
; GFX7-NEXT: s_andn2_b64 exec, exec, s[4:5]
; GFX7-NEXT: s_cbranch_execnz .LBB14_1
; GFX7-NEXT: ; %bb.2: ; %atomicrmw.end
@@ -5201,35 +5218,35 @@ define void @buffer_fat_ptr_agent_atomic_fmin_noret_bf16__offset__amdgpu_no_fine
; GFX6-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX6-NEXT: s_addk_i32 s20, 0x200
; GFX6-NEXT: s_and_b32 s4, s20, -4
-; GFX6-NEXT: v_mov_b32_e32 v2, s4
-; GFX6-NEXT: buffer_load_dword v1, v2, s[16:19], 0 offen
+; GFX6-NEXT: v_mov_b32_e32 v3, s4
+; GFX6-NEXT: buffer_load_dword v1, v3, s[16:19], 0 offen
; GFX6-NEXT: s_and_b32 s4, s20, 3
; GFX6-NEXT: s_lshl_b32 s6, s4, 3
; GFX6-NEXT: s_lshl_b32 s4, 0xffff, s6
; GFX6-NEXT: v_mul_f32_e32 v0, 1.0, v0
; GFX6-NEXT: s_not_b32 s7, s4
; GFX6-NEXT: s_mov_b64 s[4:5], 0
-; GFX6-NEXT: v_and_b32_e32 v3, 0xffff0000, v0
+; GFX6-NEXT: v_and_b32_e32 v0, 0xffff0000, v0
; GFX6-NEXT: .LBB14_1: ; %atomicrmw.start
; GFX6-NEXT: ; =>This Inner Loop Header: Depth=1
; GFX6-NEXT: s_waitcnt vmcnt(0)
-; GFX6-NEXT: v_lshrrev_b32_e32 v0, s6, v1
-; GFX6-NEXT: v_lshlrev_b32_e32 v0, 16, v0
-; GFX6-NEXT: v_mul_f32_e32 v0, 1.0, v0
-; GFX6-NEXT: v_min_f32_e32 v0, v0, v3
-; GFX6-NEXT: v_lshrrev_b32_e32 v0, 16, v0
-; GFX6-NEXT: s_waitcnt expcnt(0)
-; GFX6-NEXT: v_and_b32_e32 v4, s7, v1
-; GFX6-NEXT: v_lshlrev_b32_e32 v0, s6, v0
-; GFX6-NEXT: v_or_b32_e32 v0, v4, v0
; GFX6-NEXT: v_mov_b32_e32 v5, v1
-; GFX6-NEXT: v_mov_b32_e32 v4, v0
-; GFX6-NEXT: buffer_atomic_cmpswap v[4:5], v2, s[16:19], 0 offen glc
+; GFX6-NEXT: s_waitcnt expcnt(0)
+; GFX6-NEXT: v_lshrrev_b32_e32 v1, s6, v5
+; GFX6-NEXT: v_lshlrev_b32_e32 v1, 16, v1
+; GFX6-NEXT: v_mul_f32_e32 v1, 1.0, v1
+; GFX6-NEXT: v_min_f32_e32 v1, v1, v0
+; GFX6-NEXT: v_lshrrev_b32_e32 v1, 16, v1
+; GFX6-NEXT: v_and_b32_e32 v2, s7, v5
+; GFX6-NEXT: v_lshlrev_b32_e32 v1, s6, v1
+; GFX6-NEXT: v_or_b32_e32 v4, v2, v1
+; GFX6-NEXT: v_mov_b32_e32 v1, v4
+; GFX6-NEXT: v_mov_b32_e32 v2, v5
+; GFX6-NEXT: buffer_atomic_cmpswap v[1:2], v3, s[16:19], 0 offen glc
; GFX6-NEXT: s_waitcnt vmcnt(0)
; GFX6-NEXT: buffer_wbinvl1
-; GFX6-NEXT: v_cmp_eq_u32_e32 vcc, v4, v1
+; GFX6-NEXT: v_cmp_eq_u32_e32 vcc, v1, v5
; GFX6-NEXT: s_or_b64 s[4:5], vcc, s[4:5]
-; GFX6-NEXT: v_mov_b32_e32 v1, v4
; GFX6-NEXT: s_andn2_b64 exec, exec, s[4:5]
; GFX6-NEXT: s_cbranch_execnz .LBB14_1
; GFX6-NEXT: ; %bb.2: ; %atomicrmw.end
@@ -5270,7 +5287,7 @@ define bfloat @buffer_fat_ptr_agent_atomic_fmin_ret_bf16__offset__waterfall__amd
; GFX12-TRUE16-NEXT: s_wait_alu 0xfffe
; GFX12-TRUE16-NEXT: s_and_saveexec_b32 s0, s0
; GFX12-TRUE16-NEXT: s_wait_loadcnt 0x0
-; GFX12-TRUE16-NEXT: buffer_load_b32 v6, v8, s[4:7], null offen
+; GFX12-TRUE16-NEXT: buffer_load_b32 v4, v8, s[4:7], null offen
; GFX12-TRUE16-NEXT: s_xor_b32 exec_lo, exec_lo, s0
; GFX12-TRUE16-NEXT: s_cbranch_execnz .LBB15_1
; GFX12-TRUE16-NEXT: ; %bb.2:
@@ -5281,26 +5298,27 @@ define bfloat @buffer_fat_ptr_agent_atomic_fmin_ret_bf16__offset__waterfall__amd
; GFX12-TRUE16-NEXT: ; =>This Loop Header: Depth=1
; GFX12-TRUE16-NEXT: ; Child Loop BB15_4 Depth 2
; GFX12-TRUE16-NEXT: s_wait_loadcnt 0x0
-; GFX12-TRUE16-NEXT: v_lshrrev_b32_e32 v4, v7, v6
+; GFX12-TRUE16-NEXT: v_mov_b32_e32 v6, v4
; GFX12-TRUE16-NEXT: s_mov_b32 s2, exec_lo
; GFX12-TRUE16-NEXT: s_wait_storecnt 0x0
; GFX12-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
+; GFX12-TRUE16-NEXT: v_lshrrev_b32_e32 v4, v7, v6
; GFX12-TRUE16-NEXT: v_lshlrev_b32_e32 v4, 16, v4
+; GFX12-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
; GFX12-TRUE16-NEXT: v_min_num_f32_e32 v4, v4, v10
-; GFX12-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(SKIP_2) | instid1(VALU_DEP_3)
; GFX12-TRUE16-NEXT: v_bfe_u32 v5, v4, 16, 1
; GFX12-TRUE16-NEXT: v_or_b32_e32 v11, 0x400000, v4
; GFX12-TRUE16-NEXT: v_cmp_u_f32_e32 vcc_lo, v4, v4
+; GFX12-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_3) | instskip(SKIP_1) | instid1(VALU_DEP_1)
; GFX12-TRUE16-NEXT: v_add3_u32 v5, v5, v4, 0x7fff
; GFX12-TRUE16-NEXT: s_wait_alu 0xfffd
-; GFX12-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(SKIP_1) | instid1(VALU_DEP_2)
; GFX12-TRUE16-NEXT: v_cndmask_b32_e32 v4, v5, v11, vcc_lo
; GFX12-TRUE16-NEXT: v_mov_b16_e32 v5.h, 0
+; GFX12-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_2) | instskip(NEXT) | instid1(VALU_DEP_1)
; GFX12-TRUE16-NEXT: v_mov_b16_e32 v5.l, v4.h
-; GFX12-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
; GFX12-TRUE16-NEXT: v_lshlrev_b32_e32 v4, v7, v5
+; GFX12-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
; GFX12-TRUE16-NEXT: v_and_or_b32 v5, v6, v9, v4
-; GFX12-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1)
; GFX12-TRUE16-NEXT: v_mov_b32_e32 v4, v5
; GFX12-TRUE16-NEXT: v_mov_b32_e32 v5, v6
; GFX12-TRUE16-NEXT: .LBB15_4: ; Parent Loop BB15_3 Depth=1
@@ -5323,7 +5341,6 @@ define bfloat @buffer_fat_ptr_agent_atomic_fmin_ret_bf16__offset__waterfall__amd
; GFX12-TRUE16-NEXT: s_mov_b32 exec_lo, s2
; GFX12-TRUE16-NEXT: s_wait_loadcnt 0x0
; GFX12-TRUE16-NEXT: v_cmp_eq_u32_e32 vcc_lo, v4, v6
-; GFX12-TRUE16-NEXT: v_mov_b32_e32 v6, v4
; GFX12-TRUE16-NEXT: global_inv scope:SCOPE_DEV
; GFX12-TRUE16-NEXT: s_or_b32 s1, vcc_lo, s1
; GFX12-TRUE16-NEXT: s_wait_alu 0xfffe
@@ -5363,7 +5380,7 @@ define bfloat @buffer_fat_ptr_agent_atomic_fmin_ret_bf16__offset__waterfall__amd
; GFX12-FAKE16-NEXT: s_wait_alu 0xfffe
; GFX12-FAKE16-NEXT: s_and_saveexec_b32 s0, s0
; GFX12-FAKE16-NEXT: s_wait_loadcnt 0x0
-; GFX12-FAKE16-NEXT: buffer_load_b32 v6, v8, s[4:7], null offen
+; GFX12-FAKE16-NEXT: buffer_load_b32 v4, v8, s[4:7], null offen
; GFX12-FAKE16-NEXT: s_xor_b32 exec_lo, exec_lo, s0
; GFX12-FAKE16-NEXT: s_cbranch_execnz .LBB15_1
; GFX12-FAKE16-NEXT: ; %bb.2:
@@ -5374,25 +5391,26 @@ define bfloat @buffer_fat_ptr_agent_atomic_fmin_ret_bf16__offset__waterfall__amd
; GFX12-FAKE16-NEXT: ; =>This Loop Header: Depth=1
; GFX12-FAKE16-NEXT: ; Child Loop BB15_4 Depth 2
; GFX12-FAKE16-NEXT: s_wait_loadcnt 0x0
-; GFX12-FAKE16-NEXT: v_lshrrev_b32_e32 v4, v7, v6
+; GFX12-FAKE16-NEXT: v_mov_b32_e32 v6, v4
; GFX12-FAKE16-NEXT: s_mov_b32 s2, exec_lo
; GFX12-FAKE16-NEXT: s_wait_storecnt 0x0
; GFX12-FAKE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
+; GFX12-FAKE16-NEXT: v_lshrrev_b32_e32 v4, v7, v6
; GFX12-FAKE16-NEXT: v_lshlrev_b32_e32 v4, 16, v4
+; GFX12-FAKE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
; GFX12-FAKE16-NEXT: v_min_num_f32_e32 v4, v4, v10
-; GFX12-FAKE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(SKIP_2) | instid1(VALU_DEP_3)
; GFX12-FAKE16-NEXT: v_bfe_u32 v5, v4, 16, 1
; GFX12-FAKE16-NEXT: v_or_b32_e32 v11, 0x400000, v4
; GFX12-FAKE16-NEXT: v_cmp_u_f32_e32 vcc_lo, v4, v4
+; GFX12-FAKE16-NEXT: s_delay_alu instid0(VALU_DEP_3) | instskip(SKIP_1) | instid1(VALU_DEP_1)
; GFX12-FAKE16-NEXT: v_add3_u32 v5, v5, v4, 0x7fff
; GFX12-FAKE16-NEXT: s_wait_alu 0xfffd
-; GFX12-FAKE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
; GFX12-FAKE16-NEXT: v_cndmask_b32_e32 v4, v5, v11, vcc_lo
-; GFX12-FAKE16-NEXT: v_lshrrev_b32_e32 v4, 16, v4
; GFX12-FAKE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
+; GFX12-FAKE16-NEXT: v_lshrrev_b32_e32 v4, 16, v4
; GFX12-FAKE16-NEXT: v_lshlrev_b32_e32 v4, v7, v4
+; GFX12-FAKE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
; GFX12-FAKE16-NEXT: v_and_or_b32 v5, v6, v9, v4
-; GFX12-FAKE16-NEXT: s_delay_alu instid0(VALU_DEP_1)
; GFX12-FAKE16-NEXT: v_mov_b32_e32 v4, v5
; GFX12-FAKE16-NEXT: v_mov_b32_e32 v5, v6
; GFX12-FAKE16-NEXT: .LBB15_4: ; Parent Loop BB15_3 Depth=1
@@ -5415,7 +5433,6 @@ define bfloat @buffer_fat_ptr_agent_atomic_fmin_ret_bf16__offset__waterfall__amd
; GFX12-FAKE16-NEXT: s_mov_b32 exec_lo, s2
; GFX12-FAKE16-NEXT: s_wait_loadcnt 0x0
; GFX12-FAKE16-NEXT: v_cmp_eq_u32_e32 vcc_lo, v4, v6
-; GFX12-FAKE16-NEXT: v_mov_b32_e32 v6, v4
; GFX12-FAKE16-NEXT: global_inv scope:SCOPE_DEV
; GFX12-FAKE16-NEXT: s_or_b32 s1, vcc_lo, s1
; GFX12-FAKE16-NEXT: s_wait_alu 0xfffe
@@ -5448,7 +5465,7 @@ define bfloat @buffer_fat_ptr_agent_atomic_fmin_ret_bf16__offset__waterfall__amd
; GFX942-NEXT: v_cmp_eq_u64_e64 s[0:1], s[6:7], v[2:3]
; GFX942-NEXT: s_and_b64 s[0:1], vcc, s[0:1]
; GFX942-NEXT: s_and_saveexec_b64 s[0:1], s[0:1]
-; GFX942-NEXT: buffer_load_dword v7, v9, s[4:7], 0 offen
+; GFX942-NEXT: buffer_load_dword v4, v9, s[4:7], 0 offen
; GFX942-NEXT: s_xor_b64 exec, exec, s[0:1]
; GFX942-NEXT: s_cbranch_execnz .LBB15_1
; GFX942-NEXT: ; %bb.2:
@@ -5460,6 +5477,7 @@ define bfloat @buffer_fat_ptr_agent_atomic_fmin_ret_bf16__offset__waterfall__amd
; GFX942-NEXT: ; =>This Loop Header: Depth=1
; GFX942-NEXT: ; Child Loop BB15_4 Depth 2
; GFX942-NEXT: s_waitcnt vmcnt(0)
+; GFX942-NEXT: v_mov_b32_e32 v7, v4
; GFX942-NEXT: v_lshrrev_b32_sdwa v4, v8, v7 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:DWORD
; GFX942-NEXT: s_mov_b64 s[8:9], exec
; GFX942-NEXT: v_min_f32_e32 v4, v4, v11
@@ -5493,7 +5511,6 @@ define bfloat @buffer_fat_ptr_agent_atomic_fmin_ret_bf16__offset__waterfall__amd
; GFX942-NEXT: s_waitcnt vmcnt(0)
; GFX942-NEXT: v_cmp_eq_u32_e32 vcc, v4, v7
; GFX942-NEXT: s_or_b64 s[2:3], vcc, s[2:3]
-; GFX942-NEXT: v_mov_b32_e32 v7, v4
; GFX942-NEXT: buffer_inv sc1
; GFX942-NEXT: s_andn2_b64 exec, exec, s[2:3]
; GFX942-NEXT: s_cbranch_execnz .LBB15_3
@@ -5526,7 +5543,7 @@ define bfloat @buffer_fat_ptr_agent_atomic_fmin_ret_bf16__offset__waterfall__amd
; GFX11-TRUE16-NEXT: s_and_b32 s0, vcc_lo, s0
; GFX11-TRUE16-NEXT: s_delay_alu instid0(SALU_CYCLE_1)
; GFX11-TRUE16-NEXT: s_and_saveexec_b32 s0, s0
-; GFX11-TRUE16-NEXT: buffer_load_b32 v6, v8, s[4:7], 0 offen
+; GFX11-TRUE16-NEXT: buffer_load_b32 v4, v8, s[4:7], 0 offen
; GFX11-TRUE16-NEXT: s_xor_b32 exec_lo, exec_lo, s0
; GFX11-TRUE16-NEXT: s_cbranch_execnz .LBB15_1
; GFX11-TRUE16-NEXT: ; %bb.2:
@@ -5538,25 +5555,26 @@ define bfloat @buffer_fat_ptr_agent_atomic_fmin_ret_bf16__offset__waterfall__amd
; GFX11-TRUE16-NEXT: ; =>This Loop Header: Depth=1
; GFX11-TRUE16-NEXT: ; Child Loop BB15_4 Depth 2
; GFX11-TRUE16-NEXT: s_waitcnt vmcnt(0)
-; GFX11-TRUE16-NEXT: v_lshrrev_b32_e32 v4, v7, v6
+; GFX11-TRUE16-NEXT: v_mov_b32_e32 v6, v4
; GFX11-TRUE16-NEXT: s_mov_b32 s2, exec_lo
; GFX11-TRUE16-NEXT: s_waitcnt_vscnt null, 0x0
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
+; GFX11-TRUE16-NEXT: v_lshrrev_b32_e32 v4, v7, v6
; GFX11-TRUE16-NEXT: v_lshlrev_b32_e32 v4, 16, v4
+; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
; GFX11-TRUE16-NEXT: v_min_f32_e32 v4, v4, v10
-; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(SKIP_2) | instid1(VALU_DEP_3)
; GFX11-TRUE16-NEXT: v_bfe_u32 v5, v4, 16, 1
; GFX11-TRUE16-NEXT: v_or_b32_e32 v11, 0x400000, v4
; GFX11-TRUE16-NEXT: v_cmp_u_f32_e32 vcc_lo, v4, v4
+; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_3) | instskip(NEXT) | instid1(VALU_DEP_1)
; GFX11-TRUE16-NEXT: v_add3_u32 v5, v5, v4, 0x7fff
-; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(SKIP_1) | instid1(VALU_DEP_2)
; GFX11-TRUE16-NEXT: v_cndmask_b32_e32 v4, v5, v11, vcc_lo
; GFX11-TRUE16-NEXT: v_mov_b16_e32 v5.h, 0
+; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_2) | instskip(NEXT) | instid1(VALU_DEP_1)
; GFX11-TRUE16-NEXT: v_mov_b16_e32 v5.l, v4.h
-; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
; GFX11-TRUE16-NEXT: v_lshlrev_b32_e32 v4, v7, v5
+; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
; GFX11-TRUE16-NEXT: v_and_or_b32 v5, v6, v9, v4
-; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1)
; GFX11-TRUE16-NEXT: v_mov_b32_e32 v4, v5
; GFX11-TRUE16-NEXT: v_mov_b32_e32 v5, v6
; GFX11-TRUE16-NEXT: .LBB15_4: ; Parent Loop BB15_3 Depth=1
@@ -5579,7 +5597,6 @@ define bfloat @buffer_fat_ptr_agent_atomic_fmin_ret_bf16__offset__waterfall__amd
; GFX11-TRUE16-NEXT: s_mov_b32 exec_lo, s2
; GFX11-TRUE16-NEXT: s_waitcnt vmcnt(0)
; GFX11-TRUE16-NEXT: v_cmp_eq_u32_e32 vcc_lo, v4, v6
-; GFX11-TRUE16-NEXT: v_mov_b32_e32 v6, v4
; GFX11-TRUE16-NEXT: buffer_gl1_inv
; GFX11-TRUE16-NEXT: buffer_gl0_inv
; GFX11-TRUE16-NEXT: s_or_b32 s1, vcc_lo, s1
@@ -5616,7 +5633,7 @@ define bfloat @buffer_fat_ptr_agent_atomic_fmin_ret_bf16__offset__waterfall__amd
; GFX11-FAKE16-NEXT: s_and_b32 s0, vcc_lo, s0
; GFX11-FAKE16-NEXT: s_delay_alu instid0(SALU_CYCLE_1)
; GFX11-FAKE16-NEXT: s_and_saveexec_b32 s0, s0
-; GFX11-FAKE16-NEXT: buffer_load_b32 v6, v8, s[4:7], 0 offen
+; GFX11-FAKE16-NEXT: buffer_load_b32 v4, v8, s[4:7], 0 offen
; GFX11-FAKE16-NEXT: s_xor_b32 exec_lo, exec_lo, s0
; GFX11-FAKE16-NEXT: s_cbranch_execnz .LBB15_1
; GFX11-FAKE16-NEXT: ; %bb.2:
@@ -5628,24 +5645,25 @@ define bfloat @buffer_fat_ptr_agent_atomic_fmin_ret_bf16__offset__waterfall__amd
; GFX11-FAKE16-NEXT: ; =>This Loop Header: Depth=1
; GFX11-FAKE16-NEXT: ; Child Loop BB15_4 Depth 2
; GFX11-FAKE16-NEXT: s_waitcnt vmcnt(0)
-; GFX11-FAKE16-NEXT: v_lshrrev_b32_e32 v4, v7, v6
+; GFX11-FAKE16-NEXT: v_mov_b32_e32 v6, v4
; GFX11-FAKE16-NEXT: s_mov_b32 s2, exec_lo
; GFX11-FAKE16-NEXT: s_waitcnt_vscnt null, 0x0
; GFX11-FAKE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
+; GFX11-FAKE16-NEXT: v_lshrrev_b32_e32 v4, v7, v6
; GFX11-FAKE16-NEXT: v_lshlrev_b32_e32 v4, 16, v4
+; GFX11-FAKE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
; GFX11-FAKE16-NEXT: v_min_f32_e32 v4, v4, v10
-; GFX11-FAKE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(SKIP_2) | instid1(VALU_DEP_3)
; GFX11-FAKE16-NEXT: v_bfe_u32 v5, v4, 16, 1
; GFX11-FAKE16-NEXT: v_or_b32_e32 v11, 0x400000, v4
; GFX11-FAKE16-NEXT: v_cmp_u_f32_e32 vcc_lo, v4, v4
+; GFX11-FAKE16-NEXT: s_delay_alu instid0(VALU_DEP_3) | instskip(NEXT) | instid1(VALU_DEP_1)
; GFX11-FAKE16-NEXT: v_add3_u32 v5, v5, v4, 0x7fff
-; GFX11-FAKE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
; GFX11-FAKE16-NEXT: v_cndmask_b32_e32 v4, v5, v11, vcc_lo
-; GFX11-FAKE16-NEXT: v_lshrrev_b32_e32 v4, 16, v4
; GFX11-FAKE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
+; GFX11-FAKE16-NEXT: v_lshrrev_b32_e32 v4, 16, v4
; GFX11-FAKE16-NEXT: v_lshlrev_b32_e32 v4, v7, v4
+; GFX11-FAKE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
; GFX11-FAKE16-NEXT: v_and_or_b32 v5, v6, v9, v4
-; GFX11-FAKE16-NEXT: s_delay_alu instid0(VALU_DEP_1)
; GFX11-FAKE16-NEXT: v_mov_b32_e32 v4, v5
; GFX11-FAKE16-NEXT: v_mov_b32_e32 v5, v6
; GFX11-FAKE16-NEXT: .LBB15_4: ; Parent Loop BB15_3 Depth=1
@@ -5668,7 +5686,6 @@ define bfloat @buffer_fat_ptr_agent_atomic_fmin_ret_bf16__offset__waterfall__amd
; GFX11-FAKE16-NEXT: s_mov_b32 exec_lo, s2
; GFX11-FAKE16-NEXT: s_waitcnt vmcnt(0)
; GFX11-FAKE16-NEXT: v_cmp_eq_u32_e32 vcc_lo, v4, v6
-; GFX11-FAKE16-NEXT: v_mov_b32_e32 v6, v4
; GFX11-FAKE16-NEXT: buffer_gl1_inv
; GFX11-FAKE16-NEXT: buffer_gl0_inv
; GFX11-FAKE16-NEXT: s_or_b32 s1, vcc_lo, s1
@@ -5701,7 +5718,7 @@ define bfloat @buffer_fat_ptr_agent_atomic_fmin_ret_bf16__offset__waterfall__amd
; GFX10-NEXT: v_cmp_eq_u64_e64 s4, s[10:11], v[2:3]
; GFX10-NEXT: s_and_b32 s4, vcc_lo, s4
; GFX10-NEXT: s_and_saveexec_b32 s4, s4
-; GFX10-NEXT: buffer_load_dword v6, v8, s[8:11], 0 offen
+; GFX10-NEXT: buffer_load_dword v4, v8, s[8:11], 0 offen
; GFX10-NEXT: s_waitcnt_depctr 0xffe3
; GFX10-NEXT: s_xor_b32 exec_lo, exec_lo, s4
; GFX10-NEXT: s_cbranch_execnz .LBB15_1
@@ -5712,9 +5729,10 @@ define bfloat @buffer_fat_ptr_agent_atomic_fmin_ret_bf16__offset__waterfall__amd
; GFX10-NEXT: ; =>This Loop Header: Depth=1
; GFX10-NEXT: ; Child Loop BB15_4 Depth 2
; GFX10-NEXT: s_waitcnt vmcnt(0)
-; GFX10-NEXT: v_lshrrev_b32_sdwa v4, v7, v6 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:DWORD
+; GFX10-NEXT: v_mov_b32_e32 v6, v4
; GFX10-NEXT: s_mov_b32 s6, exec_lo
; GFX10-NEXT: s_waitcnt_vscnt null, 0x0
+; GFX10-NEXT: v_lshrrev_b32_sdwa v4, v7, v6 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:DWORD
; GFX10-NEXT: v_min_f32_e32 v4, v4, v10
; GFX10-NEXT: v_bfe_u32 v5, v4, 16, 1
; GFX10-NEXT: v_or_b32_e32 v11, 0x400000, v4
@@ -5744,7 +5762,6 @@ define bfloat @buffer_fat_ptr_agent_atomic_fmin_ret_bf16__offset__waterfall__amd
; GFX10-NEXT: s_mov_b32 exec_lo, s6
; GFX10-NEXT: s_waitcnt vmcnt(0)
; GFX10-NEXT: v_cmp_eq_u32_e32 vcc_lo, v4, v6
-; GFX10-NEXT: v_mov_b32_e32 v6, v4
; GFX10-NEXT: buffer_gl1_inv
; GFX10-NEXT: buffer_gl0_inv
; GFX10-NEXT: s_or_b32 s5, vcc_lo, s5
@@ -5777,7 +5794,7 @@ define bfloat @buffer_fat_ptr_agent_atomic_fmin_ret_bf16__offset__waterfall__amd
; GFX90A-NEXT: s_and_b64 s[4:5], vcc, s[4:5]
; GFX90A-NEXT: s_and_saveexec_b64 s[4:5], s[4:5]
; GFX90A-NEXT: s_nop 0
-; GFX90A-NEXT: buffer_load_dword v7, v9, s[8:11], 0 offen
+; GFX90A-NEXT: buffer_load_dword v4, v9, s[8:11], 0 offen
; GFX90A-NEXT: s_xor_b64 exec, exec, s[4:5]
; GFX90A-NEXT: s_cbranch_execnz .LBB15_1
; GFX90A-NEXT: ; %bb.2:
@@ -5789,6 +5806,7 @@ define bfloat @buffer_fat_ptr_agent_atomic_fmin_ret_bf16__offset__waterfall__amd
; GFX90A-NEXT: ; =>This Loop Header: Depth=1
; GFX90A-NEXT: ; Child Loop BB15_4 Depth 2
; GFX90A-NEXT: s_waitcnt vmcnt(0)
+; GFX90A-NEXT: v_mov_b32_e32 v7, v4
; GFX90A-NEXT: v_lshrrev_b32_sdwa v4, v8, v7 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:DWORD
; GFX90A-NEXT: v_min_f32_e32 v4, v4, v11
; GFX90A-NEXT: v_bfe_u32 v5, v4, 16, 1
@@ -5819,7 +5837,6 @@ define bfloat @buffer_fat_ptr_agent_atomic_fmin_ret_bf16__offset__waterfall__amd
; GFX90A-NEXT: s_waitcnt vmcnt(0)
; GFX90A-NEXT: v_cmp_eq_u32_e32 vcc, v4, v7
; GFX90A-NEXT: s_or_b64 s[6:7], vcc, s[6:7]
-; GFX90A-NEXT: v_mov_b32_e32 v7, v4
; GFX90A-NEXT: buffer_wbinvl1
; GFX90A-NEXT: s_andn2_b64 exec, exec, s[6:7]
; GFX90A-NEXT: s_cbranch_execnz .LBB15_3
@@ -5849,7 +5866,7 @@ define bfloat @buffer_fat_ptr_agent_atomic_fmin_ret_bf16__offset__waterfall__amd
; GFX908-NEXT: s_and_b64 s[4:5], vcc, s[4:5]
; GFX908-NEXT: s_and_saveexec_b64 s[4:5], s[4:5]
; GFX908-NEXT: s_nop 0
-; GFX908-NEXT: buffer_load_dword v6, v8, s[8:11], 0 offen
+; GFX908-NEXT: buffer_load_dword v4, v8, s[8:11], 0 offen
; GFX908-NEXT: s_xor_b64 exec, exec, s[4:5]
; GFX908-NEXT: s_cbranch_execnz .LBB15_1
; GFX908-NEXT: ; %bb.2:
@@ -5861,6 +5878,7 @@ define bfloat @buffer_fat_ptr_agent_atomic_fmin_ret_bf16__offset__waterfall__amd
; GFX908-NEXT: ; =>This Loop Header: Depth=1
; GFX908-NEXT: ; Child Loop BB15_4 Depth 2
; GFX908-NEXT: s_waitcnt vmcnt(0)
+; GFX908-NEXT: v_mov_b32_e32 v6, v4
; GFX908-NEXT: v_lshrrev_b32_sdwa v4, v7, v6 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:DWORD
; GFX908-NEXT: v_min_f32_e32 v4, v4, v10
; GFX908-NEXT: v_bfe_u32 v5, v4, 16, 1
@@ -5892,7 +5910,6 @@ define bfloat @buffer_fat_ptr_agent_atomic_fmin_ret_bf16__offset__waterfall__amd
; GFX908-NEXT: s_waitcnt vmcnt(0)
; GFX908-NEXT: v_cmp_eq_u32_e32 vcc, v4, v6
; GFX908-NEXT: s_or_b64 s[6:7], vcc, s[6:7]
-; GFX908-NEXT: v_mov_b32_e32 v6, v4
; GFX908-NEXT: buffer_wbinvl1
; GFX908-NEXT: s_andn2_b64 exec, exec, s[6:7]
; GFX908-NEXT: s_cbranch_execnz .LBB15_3
@@ -5922,7 +5939,7 @@ define bfloat @buffer_fat_ptr_agent_atomic_fmin_ret_bf16__offset__waterfall__amd
; GFX8-NEXT: s_and_b64 s[4:5], vcc, s[4:5]
; GFX8-NEXT: s_and_saveexec_b64 s[4:5], s[4:5]
; GFX8-NEXT: s_nop 0
-; GFX8-NEXT: buffer_load_dword v6, v8, s[8:11], 0 offen
+; GFX8-NEXT: buffer_load_dword v4, v8, s[8:11], 0 offen
; GFX8-NEXT: s_xor_b64 exec, exec, s[4:5]
; GFX8-NEXT: s_cbranch_execnz .LBB15_1
; GFX8-NEXT: ; %bb.2:
@@ -5933,6 +5950,7 @@ define bfloat @buffer_fat_ptr_agent_atomic_fmin_ret_bf16__offset__waterfall__amd
; GFX8-NEXT: ; =>This Loop Header: Depth=1
; GFX8-NEXT: ; Child Loop BB15_4 Depth 2
; GFX8-NEXT: s_waitcnt vmcnt(0)
+; GFX8-NEXT: v_mov_b32_e32 v6, v4
; GFX8-NEXT: v_lshrrev_b32_sdwa v4, v7, v6 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:DWORD
; GFX8-NEXT: v_min_f32_e32 v4, v4, v10
; GFX8-NEXT: v_bfe_u32 v5, v4, 16, 1
@@ -5966,7 +5984,6 @@ define bfloat @buffer_fat_ptr_agent_atomic_fmin_ret_bf16__offset__waterfall__amd
; GFX8-NEXT: s_waitcnt vmcnt(0)
; GFX8-NEXT: v_cmp_eq_u32_e32 vcc, v4, v6
; GFX8-NEXT: s_or_b64 s[6:7], vcc, s[6:7]
-; GFX8-NEXT: v_mov_b32_e32 v6, v4
; GFX8-NEXT: buffer_wbinvl1
; GFX8-NEXT: s_andn2_b64 exec, exec, s[6:7]
; GFX8-NEXT: s_cbranch_execnz .LBB15_3
@@ -5994,18 +6011,19 @@ define bfloat @buffer_fat_ptr_agent_atomic_fmin_ret_bf16__offset__waterfall__amd
; GFX7-NEXT: v_cmp_eq_u64_e64 s[4:5], s[10:11], v[2:3]
; GFX7-NEXT: s_and_b64 s[4:5], vcc, s[4:5]
; GFX7-NEXT: s_and_saveexec_b64 s[4:5], s[4:5]
-; GFX7-NEXT: buffer_load_dword v6, v8, s[8:11], 0 offen
+; GFX7-NEXT: buffer_load_dword v4, v8, s[8:11], 0 offen
; GFX7-NEXT: s_xor_b64 exec, exec, s[4:5]
; GFX7-NEXT: s_cbranch_execnz .LBB15_1
; GFX7-NEXT: ; %bb.2:
; GFX7-NEXT: s_mov_b64 exec, s[6:7]
-; GFX7-NEXT: v_mul_f32_e32 v4, 1.0, v5
+; GFX7-NEXT: v_mul_f32_e32 v5, 1.0, v5
; GFX7-NEXT: s_mov_b64 s[6:7], 0
-; GFX7-NEXT: v_and_b32_e32 v10, 0xffff0000, v4
+; GFX7-NEXT: v_and_b32_e32 v10, 0xffff0000, v5
; GFX7-NEXT: .LBB15_3: ; %atomicrmw.start
; GFX7-NEXT: ; =>This Loop Header: Depth=1
; GFX7-NEXT: ; Child Loop BB15_4 Depth 2
; GFX7-NEXT: s_waitcnt vmcnt(0)
+; GFX7-NEXT: v_mov_b32_e32 v6, v4
; GFX7-NEXT: v_lshrrev_b32_e32 v4, v7, v6
; GFX7-NEXT: v_lshlrev_b32_e32 v4, 16, v4
; GFX7-NEXT: v_mul_f32_e32 v4, 1.0, v4
@@ -6036,7 +6054,6 @@ define bfloat @buffer_fat_ptr_agent_atomic_fmin_ret_bf16__offset__waterfall__amd
; GFX7-NEXT: s_waitcnt vmcnt(0)
; GFX7-NEXT: v_cmp_eq_u32_e32 vcc, v4, v6
; GFX7-NEXT: s_or_b64 s[6:7], vcc, s[6:7]
-; GFX7-NEXT: v_mov_b32_e32 v6, v4
; GFX7-NEXT: buffer_wbinvl1
; GFX7-NEXT: s_andn2_b64 exec, exec, s[6:7]
; GFX7-NEXT: s_cbranch_execnz .LBB15_3
@@ -6065,18 +6082,20 @@ define bfloat @buffer_fat_ptr_agent_atomic_fmin_ret_bf16__offset__waterfall__amd
; GFX6-NEXT: v_cmp_eq_u64_e64 s[4:5], s[10:11], v[2:3]
; GFX6-NEXT: s_and_b64 s[4:5], vcc, s[4:5]
; GFX6-NEXT: s_and_saveexec_b64 s[4:5], s[4:5]
-; GFX6-NEXT: buffer_load_dword v6, v8, s[8:11], 0 offen
+; GFX6-NEXT: buffer_load_dword v4, v8, s[8:11], 0 offen
; GFX6-NEXT: s_xor_b64 exec, exec, s[4:5]
; GFX6-NEXT: s_cbranch_execnz .LBB15_1
; GFX6-NEXT: ; %bb.2:
; GFX6-NEXT: s_mov_b64 exec, s[6:7]
-; GFX6-NEXT: v_mul_f32_e32 v4, 1.0, v5
+; GFX6-NEXT: v_mul_f32_e32 v5, 1.0, v5
; GFX6-NEXT: s_mov_b64 s[6:7], 0
-; GFX6-NEXT: v_and_b32_e32 v10, 0xffff0000, v4
+; GFX6-NEXT: v_and_b32_e32 v10, 0xffff0000, v5
; GFX6-NEXT: .LBB15_3: ; %atomicrmw.start
; GFX6-NEXT: ; =>This Loop Header: Depth=1
; GFX6-NEXT: ; Child Loop BB15_4 Depth 2
-; GFX6-NEXT: s_waitcnt vmcnt(0) expcnt(0)
+; GFX6-NEXT: s_waitcnt vmcnt(0)
+; GFX6-NEXT: v_mov_b32_e32 v6, v4
+; GFX6-NEXT: s_waitcnt expcnt(0)
; GFX6-NEXT: v_lshrrev_b32_e32 v4, v7, v6
; GFX6-NEXT: v_lshlrev_b32_e32 v4, 16, v4
; GFX6-NEXT: v_mul_f32_e32 v4, 1.0, v4
@@ -6107,7 +6126,6 @@ define bfloat @buffer_fat_ptr_agent_atomic_fmin_ret_bf16__offset__waterfall__amd
; GFX6-NEXT: s_waitcnt vmcnt(0)
; GFX6-NEXT: v_cmp_eq_u32_e32 vcc, v4, v6
; GFX6-NEXT: s_or_b64 s[6:7], vcc, s[6:7]
-; GFX6-NEXT: v_mov_b32_e32 v6, v4
; GFX6-NEXT: buffer_wbinvl1
; GFX6-NEXT: s_andn2_b64 exec, exec, s[6:7]
; GFX6-NEXT: s_cbranch_execnz .LBB15_3
@@ -6457,7 +6475,7 @@ define void @buffer_fat_ptr_agent_atomic_fmin_noret_v2f16__offset__amdgpu_no_fin
; GFX12-NEXT: s_wait_kmcnt 0x0
; GFX12-NEXT: v_mov_b32_e32 v1, s16
; GFX12-NEXT: s_add_co_i32 s4, s16, 0x400
-; GFX12-NEXT: v_pk_max_num_f16 v2, v0, v0
+; GFX12-NEXT: v_pk_max_num_f16 v0, v0, v0
; GFX12-NEXT: s_wait_alu 0xfffe
; GFX12-NEXT: v_mov_b32_e32 v3, s4
; GFX12-NEXT: s_mov_b32 s4, 0
@@ -6465,16 +6483,17 @@ define void @buffer_fat_ptr_agent_atomic_fmin_noret_v2f16__offset__amdgpu_no_fin
; GFX12-NEXT: .LBB17_1: ; %atomicrmw.start
; GFX12-NEXT: ; =>This Inner Loop Header: Depth=1
; GFX12-NEXT: s_wait_loadcnt 0x0
-; GFX12-NEXT: v_pk_max_num_f16 v0, v1, v1
+; GFX12-NEXT: v_mov_b32_e32 v5, v1
; GFX12-NEXT: s_wait_storecnt 0x0
; GFX12-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
-; GFX12-NEXT: v_pk_min_num_f16 v0, v0, v2
-; GFX12-NEXT: v_dual_mov_b32 v5, v1 :: v_dual_mov_b32 v4, v0
-; GFX12-NEXT: buffer_atomic_cmpswap_b32 v[4:5], v3, s[0:3], null offen th:TH_ATOMIC_RETURN
+; GFX12-NEXT: v_pk_max_num_f16 v1, v5, v5
+; GFX12-NEXT: v_pk_min_num_f16 v4, v1, v0
+; GFX12-NEXT: s_delay_alu instid0(VALU_DEP_1)
+; GFX12-NEXT: v_dual_mov_b32 v1, v4 :: v_dual_mov_b32 v2, v5
+; GFX12-NEXT: buffer_atomic_cmpswap_b32 v[1:2], v3, s[0:3], null offen th:TH_ATOMIC_RETURN
; GFX12-NEXT: s_wait_loadcnt 0x0
; GFX12-NEXT: global_inv scope:SCOPE_DEV
-; GFX12-NEXT: v_cmp_eq_u32_e32 vcc_lo, v4, v1
-; GFX12-NEXT: v_mov_b32_e32 v1, v4
+; GFX12-NEXT: v_cmp_eq_u32_e32 vcc_lo, v1, v5
; GFX12-NEXT: s_or_b32 s4, vcc_lo, s4
; GFX12-NEXT: s_wait_alu 0xfffe
; GFX12-NEXT: s_and_not1_b32 exec_lo, exec_lo, s4
@@ -6488,25 +6507,25 @@ define void @buffer_fat_ptr_agent_atomic_fmin_noret_v2f16__offset__amdgpu_no_fin
; GFX942: ; %bb.0:
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX942-NEXT: v_mov_b32_e32 v1, s16
-; GFX942-NEXT: buffer_load_dword v1, v1, s[0:3], 0 offen offset:1024
+; GFX942-NEXT: buffer_load_dword v2, v1, s[0:3], 0 offen offset:1024
; GFX942-NEXT: s_add_i32 s6, s16, 0x400
; GFX942-NEXT: s_mov_b64 s[4:5], 0
-; GFX942-NEXT: v_pk_max_f16 v2, v0, v0
-; GFX942-NEXT: v_mov_b32_e32 v3, s6
+; GFX942-NEXT: v_pk_max_f16 v0, v0, v0
+; GFX942-NEXT: v_mov_b32_e32 v1, s6
; GFX942-NEXT: .LBB17_1: ; %atomicrmw.start
; GFX942-NEXT: ; =>This Inner Loop Header: Depth=1
; GFX942-NEXT: s_waitcnt vmcnt(0)
-; GFX942-NEXT: v_pk_max_f16 v0, v1, v1
+; GFX942-NEXT: v_mov_b32_e32 v5, v2
+; GFX942-NEXT: v_pk_max_f16 v2, v5, v5
; GFX942-NEXT: buffer_wbl2 sc1
-; GFX942-NEXT: v_pk_min_f16 v0, v0, v2
+; GFX942-NEXT: v_pk_min_f16 v4, v2, v0
; GFX942-NEXT: s_nop 0
-; GFX942-NEXT: v_mov_b64_e32 v[4:5], v[0:1]
-; GFX942-NEXT: buffer_atomic_cmpswap v[4:5], v3, s[0:3], 0 offen sc0
+; GFX942-NEXT: v_mov_b64_e32 v[2:3], v[4:5]
+; GFX942-NEXT: buffer_atomic_cmpswap v[2:3], v1, s[0:3], 0 offen sc0
; GFX942-NEXT: s_waitcnt vmcnt(0)
; GFX942-NEXT: buffer_inv sc1
-; GFX942-NEXT: v_cmp_eq_u32_e32 vcc, v4, v1
+; GFX942-NEXT: v_cmp_eq_u32_e32 vcc, v2, v5
; GFX942-NEXT: s_or_b64 s[4:5], vcc, s[4:5]
-; GFX942-NEXT: v_mov_b32_e32 v1, v4
; GFX942-NEXT: s_andn2_b64 exec, exec, s[4:5]
; GFX942-NEXT: s_cbranch_execnz .LBB17_1
; GFX942-NEXT: ; %bb.2: ; %atomicrmw.end
@@ -6518,24 +6537,25 @@ define void @buffer_fat_ptr_agent_atomic_fmin_noret_v2f16__offset__amdgpu_no_fin
; GFX11-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX11-NEXT: v_mov_b32_e32 v1, s16
; GFX11-NEXT: s_add_i32 s4, s16, 0x400
-; GFX11-NEXT: v_pk_max_f16 v2, v0, v0
+; GFX11-NEXT: v_pk_max_f16 v0, v0, v0
; GFX11-NEXT: v_mov_b32_e32 v3, s4
; GFX11-NEXT: s_mov_b32 s4, 0
; GFX11-NEXT: buffer_load_b32 v1, v1, s[0:3], 0 offen offset:1024
; GFX11-NEXT: .LBB17_1: ; %atomicrmw.start
; GFX11-NEXT: ; =>This Inner Loop Header: Depth=1
; GFX11-NEXT: s_waitcnt vmcnt(0)
-; GFX11-NEXT: v_pk_max_f16 v0, v1, v1
+; GFX11-NEXT: v_mov_b32_e32 v5, v1
; GFX11-NEXT: s_waitcnt_vscnt null, 0x0
; GFX11-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
-; GFX11-NEXT: v_pk_min_f16 v0, v0, v2
-; GFX11-NEXT: v_dual_mov_b32 v5, v1 :: v_dual_mov_b32 v4, v0
-; GFX11-NEXT: buffer_atomic_cmpswap_b32 v[4:5], v3, s[0:3], 0 offen glc
+; GFX11-NEXT: v_pk_max_f16 v1, v5, v5
+; GFX11-NEXT: v_pk_min_f16 v4, v1, v0
+; GFX11-NEXT: s_delay_alu instid0(VALU_DEP_1)
+; GFX11-NEXT: v_dual_mov_b32 v1, v4 :: v_dual_mov_b32 v2, v5
+; GFX11-NEXT: buffer_atomic_cmpswap_b32 v[1:2], v3, s[0:3], 0 offen glc
; GFX11-NEXT: s_waitcnt vmcnt(0)
; GFX11-NEXT: buffer_gl1_inv
; GFX11-NEXT: buffer_gl0_inv
-; GFX11-NEXT: v_cmp_eq_u32_e32 vcc_lo, v4, v1
-; GFX11-NEXT: v_mov_b32_e32 v1, v4
+; GFX11-NEXT: v_cmp_eq_u32_e32 vcc_lo, v1, v5
; GFX11-NEXT: s_or_b32 s4, vcc_lo, s4
; GFX11-NEXT: s_delay_alu instid0(SALU_CYCLE_1)
; GFX11-NEXT: s_and_not1_b32 exec_lo, exec_lo, s4
@@ -6549,24 +6569,24 @@ define void @buffer_fat_ptr_agent_atomic_fmin_noret_v2f16__offset__amdgpu_no_fin
; GFX10-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX10-NEXT: v_mov_b32_e32 v1, s20
; GFX10-NEXT: s_add_i32 s4, s20, 0x400
-; GFX10-NEXT: v_pk_max_f16 v2, v0, v0
+; GFX10-NEXT: v_pk_max_f16 v0, v0, v0
; GFX10-NEXT: v_mov_b32_e32 v3, s4
; GFX10-NEXT: s_mov_b32 s4, 0
; GFX10-NEXT: buffer_load_dword v1, v1, s[16:19], 0 offen offset:1024
; GFX10-NEXT: .LBB17_1: ; %atomicrmw.start
; GFX10-NEXT: ; =>This Inner Loop Header: Depth=1
; GFX10-NEXT: s_waitcnt vmcnt(0)
-; GFX10-NEXT: v_pk_max_f16 v0, v1, v1
-; GFX10-NEXT: s_waitcnt_vscnt null, 0x0
-; GFX10-NEXT: v_pk_min_f16 v0, v0, v2
; GFX10-NEXT: v_mov_b32_e32 v5, v1
-; GFX10-NEXT: v_mov_b32_e32 v4, v0
-; GFX10-NEXT: buffer_atomic_cmpswap v[4:5], v3, s[16:19], 0 offen glc
+; GFX10-NEXT: s_waitcnt_vscnt null, 0x0
+; GFX10-NEXT: v_pk_max_f16 v1, v5, v5
+; GFX10-NEXT: v_pk_min_f16 v4, v1, v0
+; GFX10-NEXT: v_mov_b32_e32 v1, v4
+; GFX10-NEXT: v_mov_b32_e32 v2, v5
+; GFX10-NEXT: buffer_atomic_cmpswap v[1:2], v3, s[16:19], 0 offen glc
; GFX10-NEXT: s_waitcnt vmcnt(0)
; GFX10-NEXT: buffer_gl1_inv
; GFX10-NEXT: buffer_gl0_inv
-; GFX10-NEXT: v_cmp_eq_u32_e32 vcc_lo, v4, v1
-; GFX10-NEXT: v_mov_b32_e32 v1, v4
+; GFX10-NEXT: v_cmp_eq_u32_e32 vcc_lo, v1, v5
; GFX10-NEXT: s_or_b32 s4, vcc_lo, s4
; GFX10-NEXT: s_andn2_b32 exec_lo, exec_lo, s4
; GFX10-NEXT: s_cbranch_execnz .LBB17_1
@@ -6578,23 +6598,23 @@ define void @buffer_fat_ptr_agent_atomic_fmin_noret_v2f16__offset__amdgpu_no_fin
; GFX90A: ; %bb.0:
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX90A-NEXT: v_mov_b32_e32 v1, s20
-; GFX90A-NEXT: buffer_load_dword v1, v1, s[16:19], 0 offen offset:1024
+; GFX90A-NEXT: buffer_load_dword v2, v1, s[16:19], 0 offen offset:1024
; GFX90A-NEXT: s_add_i32 s6, s20, 0x400
; GFX90A-NEXT: s_mov_b64 s[4:5], 0
-; GFX90A-NEXT: v_pk_max_f16 v2, v0, v0
-; GFX90A-NEXT: v_mov_b32_e32 v3, s6
+; GFX90A-NEXT: v_pk_max_f16 v0, v0, v0
+; GFX90A-NEXT: v_mov_b32_e32 v1, s6
; GFX90A-NEXT: .LBB17_1: ; %atomicrmw.start
; GFX90A-NEXT: ; =>This Inner Loop Header: Depth=1
; GFX90A-NEXT: s_waitcnt vmcnt(0)
-; GFX90A-NEXT: v_pk_max_f16 v0, v1, v1
-; GFX90A-NEXT: v_pk_min_f16 v0, v0, v2
-; GFX90A-NEXT: v_pk_mov_b32 v[4:5], v[0:1], v[0:1] op_sel:[0,1]
-; GFX90A-NEXT: buffer_atomic_cmpswap v[4:5], v3, s[16:19], 0 offen glc
+; GFX90A-NEXT: v_mov_b32_e32 v5, v2
+; GFX90A-NEXT: v_pk_max_f16 v2, v5, v5
+; GFX90A-NEXT: v_pk_min_f16 v4, v2, v0
+; GFX90A-NEXT: v_pk_mov_b32 v[2:3], v[4:5], v[4:5] op_sel:[0,1]
+; GFX90A-NEXT: buffer_atomic_cmpswap v[2:3], v1, s[16:19], 0 offen glc
; GFX90A-NEXT: s_waitcnt vmcnt(0)
; GFX90A-NEXT: buffer_wbinvl1
-; GFX90A-NEXT: v_cmp_eq_u32_e32 vcc, v4, v1
+; GFX90A-NEXT: v_cmp_eq_u32_e32 vcc, v2, v5
; GFX90A-NEXT: s_or_b64 s[4:5], vcc, s[4:5]
-; GFX90A-NEXT: v_mov_b32_e32 v1, v4
; GFX90A-NEXT: s_andn2_b64 exec, exec, s[4:5]
; GFX90A-NEXT: s_cbranch_execnz .LBB17_1
; GFX90A-NEXT: ; %bb.2: ; %atomicrmw.end
@@ -6608,21 +6628,21 @@ define void @buffer_fat_ptr_agent_atomic_fmin_noret_v2f16__offset__amdgpu_no_fin
; GFX908-NEXT: buffer_load_dword v1, v1, s[16:19], 0 offen offset:1024
; GFX908-NEXT: s_add_i32 s6, s20, 0x400
; GFX908-NEXT: s_mov_b64 s[4:5], 0
-; GFX908-NEXT: v_pk_max_f16 v2, v0, v0
+; GFX908-NEXT: v_pk_max_f16 v0, v0, v0
; GFX908-NEXT: v_mov_b32_e32 v3, s6
; GFX908-NEXT: .LBB17_1: ; %atomicrmw.start
; GFX908-NEXT: ; =>This Inner Loop Header: Depth=1
; GFX908-NEXT: s_waitcnt vmcnt(0)
-; GFX908-NEXT: v_pk_max_f16 v0, v1, v1
-; GFX908-NEXT: v_pk_min_f16 v0, v0, v2
; GFX908-NEXT: v_mov_b32_e32 v5, v1
-; GFX908-NEXT: v_mov_b32_e32 v4, v0
-; GFX908-NEXT: buffer_atomic_cmpswap v[4:5], v3, s[16:19], 0 offen glc
+; GFX908-NEXT: v_pk_max_f16 v1, v5, v5
+; GFX908-NEXT: v_pk_min_f16 v4, v1, v0
+; GFX908-NEXT: v_mov_b32_e32 v1, v4
+; GFX908-NEXT: v_mov_b32_e32 v2, v5
+; GFX908-NEXT: buffer_atomic_cmpswap v[1:2], v3, s[16:19], 0 offen glc
; GFX908-NEXT: s_waitcnt vmcnt(0)
; GFX908-NEXT: buffer_wbinvl1
-; GFX908-NEXT: v_cmp_eq_u32_e32 vcc, v4, v1
+; GFX908-NEXT: v_cmp_eq_u32_e32 vcc, v1, v5
; GFX908-NEXT: s_or_b64 s[4:5], vcc, s[4:5]
-; GFX908-NEXT: v_mov_b32_e32 v1, v4
; GFX908-NEXT: s_andn2_b64 exec, exec, s[4:5]
; GFX908-NEXT: s_cbranch_execnz .LBB17_1
; GFX908-NEXT: ; %bb.2: ; %atomicrmw.end
@@ -6636,25 +6656,25 @@ define void @buffer_fat_ptr_agent_atomic_fmin_noret_v2f16__offset__amdgpu_no_fin
; GFX8-NEXT: buffer_load_dword v1, v1, s[16:19], 0 offen offset:1024
; GFX8-NEXT: s_add_i32 s6, s20, 0x400
; GFX8-NEXT: s_mov_b64 s[4:5], 0
-; GFX8-NEXT: v_max_f16_sdwa v2, v0, v0 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_1 src1_sel:WORD_1
-; GFX8-NEXT: v_max_f16_e32 v3, v0, v0
+; GFX8-NEXT: v_max_f16_sdwa v3, v0, v0 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_1 src1_sel:WORD_1
+; GFX8-NEXT: v_max_f16_e32 v0, v0, v0
; GFX8-NEXT: v_mov_b32_e32 v4, s6
; GFX8-NEXT: .LBB17_1: ; %atomicrmw.start
; GFX8-NEXT: ; =>This Inner Loop Header: Depth=1
; GFX8-NEXT: s_waitcnt vmcnt(0)
-; GFX8-NEXT: v_max_f16_sdwa v0, v1, v1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_1 src1_sel:WORD_1
-; GFX8-NEXT: v_max_f16_e32 v5, v1, v1
-; GFX8-NEXT: v_min_f16_sdwa v0, v0, v2 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:DWORD
-; GFX8-NEXT: v_min_f16_e32 v5, v5, v3
-; GFX8-NEXT: v_or_b32_e32 v0, v5, v0
; GFX8-NEXT: v_mov_b32_e32 v6, v1
-; GFX8-NEXT: v_mov_b32_e32 v5, v0
-; GFX8-NEXT: buffer_atomic_cmpswap v[5:6], v4, s[16:19], 0 offen glc
+; GFX8-NEXT: v_max_f16_sdwa v1, v6, v6 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_1 src1_sel:WORD_1
+; GFX8-NEXT: v_max_f16_e32 v2, v6, v6
+; GFX8-NEXT: v_min_f16_sdwa v1, v1, v3 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:DWORD
+; GFX8-NEXT: v_min_f16_e32 v2, v2, v0
+; GFX8-NEXT: v_or_b32_e32 v5, v2, v1
+; GFX8-NEXT: v_mov_b32_e32 v1, v5
+; GFX8-NEXT: v_mov_b32_e32 v2, v6
+; GFX8-NEXT: buffer_atomic_cmpswap v[1:2], v4, s[16:19], 0 offen glc
; GFX8-NEXT: s_waitcnt vmcnt(0)
; GFX8-NEXT: buffer_wbinvl1
-; GFX8-NEXT: v_cmp_eq_u32_e32 vcc, v5, v1
+; GFX8-NEXT: v_cmp_eq_u32_e32 vcc, v1, v6
; GFX8-NEXT: s_or_b64 s[4:5], vcc, s[4:5]
-; GFX8-NEXT: v_mov_b32_e32 v1, v5
; GFX8-NEXT: s_andn2_b64 exec, exec, s[4:5]
; GFX8-NEXT: s_cbranch_execnz .LBB17_1
; GFX8-NEXT: ; %bb.2: ; %atomicrmw.end
@@ -7918,43 +7938,45 @@ define void @buffer_fat_ptr_agent_atomic_fmin_noret_v2bf16__offset__amdgpu_no_fi
; GFX12-TRUE16-NEXT: s_wait_samplecnt 0x0
; GFX12-TRUE16-NEXT: s_wait_bvhcnt 0x0
; GFX12-TRUE16-NEXT: s_wait_kmcnt 0x0
-; GFX12-TRUE16-NEXT: v_dual_mov_b32 v1, s16 :: v_dual_and_b32 v2, 0xffff0000, v0
+; GFX12-TRUE16-NEXT: v_mov_b32_e32 v1, s16
; GFX12-TRUE16-NEXT: s_add_co_i32 s4, s16, 0x400
; GFX12-TRUE16-NEXT: s_wait_alu 0xfffe
-; GFX12-TRUE16-NEXT: v_dual_mov_b32 v4, s4 :: v_dual_lshlrev_b32 v3, 16, v0
+; GFX12-TRUE16-NEXT: v_dual_mov_b32 v4, s4 :: v_dual_and_b32 v3, 0xffff0000, v0
+; GFX12-TRUE16-NEXT: v_lshlrev_b32_e32 v0, 16, v0
; GFX12-TRUE16-NEXT: buffer_load_b32 v1, v1, s[0:3], null offen offset:1024
; GFX12-TRUE16-NEXT: s_mov_b32 s4, 0
; GFX12-TRUE16-NEXT: .LBB20_1: ; %atomicrmw.start
; GFX12-TRUE16-NEXT: ; =>This Inner Loop Header: Depth=1
; GFX12-TRUE16-NEXT: s_wait_loadcnt 0x0
-; GFX12-TRUE16-NEXT: v_and_b32_e32 v5, 0xffff0000, v1
-; GFX12-TRUE16-NEXT: v_lshlrev_b32_e32 v0, 16, v1
+; GFX12-TRUE16-NEXT: v_mov_b32_e32 v6, v1
; GFX12-TRUE16-NEXT: s_wait_storecnt 0x0
; GFX12-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
-; GFX12-TRUE16-NEXT: v_dual_min_num_f32 v5, v5, v2 :: v_dual_min_num_f32 v0, v0, v3
-; GFX12-TRUE16-NEXT: v_bfe_u32 v7, v5, 16, 1
+; GFX12-TRUE16-NEXT: v_and_b32_e32 v2, 0xffff0000, v6
+; GFX12-TRUE16-NEXT: v_dual_min_num_f32 v2, v2, v3 :: v_dual_lshlrev_b32 v1, 16, v6
+; GFX12-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_2)
+; GFX12-TRUE16-NEXT: v_min_num_f32_e32 v1, v1, v0
+; GFX12-TRUE16-NEXT: v_bfe_u32 v7, v2, 16, 1
; GFX12-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_2)
-; GFX12-TRUE16-NEXT: v_bfe_u32 v6, v0, 16, 1
-; GFX12-TRUE16-NEXT: v_or_b32_e32 v8, 0x400000, v0
-; GFX12-TRUE16-NEXT: v_cmp_u_f32_e32 vcc_lo, v0, v0
-; GFX12-TRUE16-NEXT: v_or_b32_e32 v9, 0x400000, v5
-; GFX12-TRUE16-NEXT: v_add3_u32 v7, v7, v5, 0x7fff
-; GFX12-TRUE16-NEXT: v_add3_u32 v6, v6, v0, 0x7fff
+; GFX12-TRUE16-NEXT: v_bfe_u32 v5, v1, 16, 1
+; GFX12-TRUE16-NEXT: v_or_b32_e32 v8, 0x400000, v1
+; GFX12-TRUE16-NEXT: v_cmp_u_f32_e32 vcc_lo, v1, v1
+; GFX12-TRUE16-NEXT: v_or_b32_e32 v9, 0x400000, v2
+; GFX12-TRUE16-NEXT: v_add3_u32 v7, v7, v2, 0x7fff
+; GFX12-TRUE16-NEXT: v_add3_u32 v5, v5, v1, 0x7fff
; GFX12-TRUE16-NEXT: s_wait_alu 0xfffd
; GFX12-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(SKIP_3) | instid1(VALU_DEP_3)
-; GFX12-TRUE16-NEXT: v_cndmask_b32_e32 v0, v6, v8, vcc_lo
-; GFX12-TRUE16-NEXT: v_cmp_u_f32_e32 vcc_lo, v5, v5
+; GFX12-TRUE16-NEXT: v_cndmask_b32_e32 v1, v5, v8, vcc_lo
+; GFX12-TRUE16-NEXT: v_cmp_u_f32_e32 vcc_lo, v2, v2
; GFX12-TRUE16-NEXT: s_wait_alu 0xfffd
-; GFX12-TRUE16-NEXT: v_cndmask_b32_e32 v5, v7, v9, vcc_lo
-; GFX12-TRUE16-NEXT: v_mov_b16_e32 v0.l, v0.h
+; GFX12-TRUE16-NEXT: v_cndmask_b32_e32 v2, v7, v9, vcc_lo
+; GFX12-TRUE16-NEXT: v_mov_b16_e32 v1.l, v1.h
; GFX12-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
-; GFX12-TRUE16-NEXT: v_bfi_b32 v0, 0xffff, v0, v5
-; GFX12-TRUE16-NEXT: v_dual_mov_b32 v6, v1 :: v_dual_mov_b32 v5, v0
-; GFX12-TRUE16-NEXT: buffer_atomic_cmpswap_b32 v[5:6], v4, s[0:3], null offen th:TH_ATOMIC_RETURN
+; GFX12-TRUE16-NEXT: v_bfi_b32 v5, 0xffff, v1, v2
+; GFX12-TRUE16-NEXT: v_dual_mov_b32 v1, v5 :: v_dual_mov_b32 v2, v6
+; GFX12-TRUE16-NEXT: buffer_atomic_cmpswap_b32 v[1:2], v4, s[0:3], null offen th:TH_ATOMIC_RETURN
; GFX12-TRUE16-NEXT: s_wait_loadcnt 0x0
; GFX12-TRUE16-NEXT: global_inv scope:SCOPE_DEV
-; GFX12-TRUE16-NEXT: v_cmp_eq_u32_e32 vcc_lo, v5, v1
-; GFX12-TRUE16-NEXT: v_mov_b32_e32 v1, v5
+; GFX12-TRUE16-NEXT: v_cmp_eq_u32_e32 vcc_lo, v1, v6
; GFX12-TRUE16-NEXT: s_or_b32 s4, vcc_lo, s4
; GFX12-TRUE16-NEXT: s_wait_alu 0xfffe
; GFX12-TRUE16-NEXT: s_and_not1_b32 exec_lo, exec_lo, s4
@@ -7971,41 +7993,43 @@ define void @buffer_fat_ptr_agent_atomic_fmin_noret_v2bf16__offset__amdgpu_no_fi
; GFX12-FAKE16-NEXT: s_wait_samplecnt 0x0
; GFX12-FAKE16-NEXT: s_wait_bvhcnt 0x0
; GFX12-FAKE16-NEXT: s_wait_kmcnt 0x0
-; GFX12-FAKE16-NEXT: v_dual_mov_b32 v1, s16 :: v_dual_lshlrev_b32 v2, 16, v0
+; GFX12-FAKE16-NEXT: v_mov_b32_e32 v1, s16
; GFX12-FAKE16-NEXT: s_add_co_i32 s4, s16, 0x400
; GFX12-FAKE16-NEXT: s_wait_alu 0xfffe
-; GFX12-FAKE16-NEXT: v_dual_mov_b32 v4, s4 :: v_dual_and_b32 v3, 0xffff0000, v0
+; GFX12-FAKE16-NEXT: v_dual_mov_b32 v4, s4 :: v_dual_lshlrev_b32 v3, 16, v0
+; GFX12-FAKE16-NEXT: v_and_b32_e32 v0, 0xffff0000, v0
; GFX12-FAKE16-NEXT: buffer_load_b32 v1, v1, s[0:3], null offen offset:1024
; GFX12-FAKE16-NEXT: s_mov_b32 s5, 0
; GFX12-FAKE16-NEXT: .LBB20_1: ; %atomicrmw.start
; GFX12-FAKE16-NEXT: ; =>This Inner Loop Header: Depth=1
; GFX12-FAKE16-NEXT: s_wait_loadcnt 0x0
-; GFX12-FAKE16-NEXT: v_and_b32_e32 v5, 0xffff0000, v1
-; GFX12-FAKE16-NEXT: v_lshlrev_b32_e32 v0, 16, v1
+; GFX12-FAKE16-NEXT: v_mov_b32_e32 v6, v1
; GFX12-FAKE16-NEXT: s_wait_storecnt 0x0
; GFX12-FAKE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
-; GFX12-FAKE16-NEXT: v_dual_min_num_f32 v5, v5, v3 :: v_dual_min_num_f32 v0, v0, v2
-; GFX12-FAKE16-NEXT: v_bfe_u32 v7, v5, 16, 1
+; GFX12-FAKE16-NEXT: v_and_b32_e32 v2, 0xffff0000, v6
+; GFX12-FAKE16-NEXT: v_dual_min_num_f32 v2, v2, v0 :: v_dual_lshlrev_b32 v1, 16, v6
+; GFX12-FAKE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_2)
+; GFX12-FAKE16-NEXT: v_min_num_f32_e32 v1, v1, v3
+; GFX12-FAKE16-NEXT: v_bfe_u32 v7, v2, 16, 1
; GFX12-FAKE16-NEXT: s_delay_alu instid0(VALU_DEP_2)
-; GFX12-FAKE16-NEXT: v_bfe_u32 v6, v0, 16, 1
-; GFX12-FAKE16-NEXT: v_or_b32_e32 v8, 0x400000, v0
-; GFX12-FAKE16-NEXT: v_or_b32_e32 v9, 0x400000, v5
-; GFX12-FAKE16-NEXT: v_cmp_u_f32_e32 vcc_lo, v5, v5
-; GFX12-FAKE16-NEXT: v_add3_u32 v7, v7, v5, 0x7fff
-; GFX12-FAKE16-NEXT: v_add3_u32 v6, v6, v0, 0x7fff
-; GFX12-FAKE16-NEXT: v_cmp_u_f32_e64 s4, v0, v0
+; GFX12-FAKE16-NEXT: v_bfe_u32 v5, v1, 16, 1
+; GFX12-FAKE16-NEXT: v_or_b32_e32 v8, 0x400000, v1
+; GFX12-FAKE16-NEXT: v_or_b32_e32 v9, 0x400000, v2
+; GFX12-FAKE16-NEXT: v_cmp_u_f32_e32 vcc_lo, v2, v2
+; GFX12-FAKE16-NEXT: v_add3_u32 v7, v7, v2, 0x7fff
+; GFX12-FAKE16-NEXT: v_add3_u32 v5, v5, v1, 0x7fff
+; GFX12-FAKE16-NEXT: v_cmp_u_f32_e64 s4, v1, v1
; GFX12-FAKE16-NEXT: s_wait_alu 0xfffd
-; GFX12-FAKE16-NEXT: v_cndmask_b32_e32 v5, v7, v9, vcc_lo
+; GFX12-FAKE16-NEXT: v_cndmask_b32_e32 v2, v7, v9, vcc_lo
; GFX12-FAKE16-NEXT: s_wait_alu 0xf1ff
-; GFX12-FAKE16-NEXT: v_cndmask_b32_e64 v0, v6, v8, s4
+; GFX12-FAKE16-NEXT: v_cndmask_b32_e64 v1, v5, v8, s4
; GFX12-FAKE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
-; GFX12-FAKE16-NEXT: v_perm_b32 v0, v5, v0, 0x7060302
-; GFX12-FAKE16-NEXT: v_dual_mov_b32 v6, v1 :: v_dual_mov_b32 v5, v0
-; GFX12-FAKE16-NEXT: buffer_atomic_cmpswap_b32 v[5:6], v4, s[0:3], null offen th:TH_ATOMIC_RETURN
+; GFX12-FAKE16-NEXT: v_perm_b32 v5, v2, v1, 0x7060302
+; GFX12-FAKE16-NEXT: v_dual_mov_b32 v1, v5 :: v_dual_mov_b32 v2, v6
+; GFX12-FAKE16-NEXT: buffer_atomic_cmpswap_b32 v[1:2], v4, s[0:3], null offen th:TH_ATOMIC_RETURN
; GFX12-FAKE16-NEXT: s_wait_loadcnt 0x0
; GFX12-FAKE16-NEXT: global_inv scope:SCOPE_DEV
-; GFX12-FAKE16-NEXT: v_cmp_eq_u32_e32 vcc_lo, v5, v1
-; GFX12-FAKE16-NEXT: v_mov_b32_e32 v1, v5
+; GFX12-FAKE16-NEXT: v_cmp_eq_u32_e32 vcc_lo, v1, v6
; GFX12-FAKE16-NEXT: s_or_b32 s5, vcc_lo, s5
; GFX12-FAKE16-NEXT: s_wait_alu 0xfffe
; GFX12-FAKE16-NEXT: s_and_not1_b32 exec_lo, exec_lo, s5
@@ -8019,40 +8043,40 @@ define void @buffer_fat_ptr_agent_atomic_fmin_noret_v2bf16__offset__amdgpu_no_fi
; GFX942: ; %bb.0:
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX942-NEXT: v_mov_b32_e32 v1, s16
-; GFX942-NEXT: buffer_load_dword v1, v1, s[0:3], 0 offen offset:1024
+; GFX942-NEXT: buffer_load_dword v2, v1, s[0:3], 0 offen offset:1024
; GFX942-NEXT: s_add_i32 s4, s16, 0x400
; GFX942-NEXT: s_mov_b64 s[6:7], 0
-; GFX942-NEXT: v_lshlrev_b32_e32 v2, 16, v0
+; GFX942-NEXT: v_lshlrev_b32_e32 v1, 16, v0
; GFX942-NEXT: s_movk_i32 s8, 0x7fff
-; GFX942-NEXT: v_and_b32_e32 v3, 0xffff0000, v0
+; GFX942-NEXT: v_and_b32_e32 v0, 0xffff0000, v0
; GFX942-NEXT: s_mov_b32 s9, 0x7060302
; GFX942-NEXT: v_mov_b32_e32 v4, s4
; GFX942-NEXT: .LBB20_1: ; %atomicrmw.start
; GFX942-NEXT: ; =>This Inner Loop Header: Depth=1
; GFX942-NEXT: s_waitcnt vmcnt(0)
-; GFX942-NEXT: v_lshlrev_b32_e32 v0, 16, v1
-; GFX942-NEXT: v_and_b32_e32 v5, 0xffff0000, v1
-; GFX942-NEXT: v_min_f32_e32 v0, v0, v2
-; GFX942-NEXT: v_min_f32_e32 v5, v5, v3
-; GFX942-NEXT: v_bfe_u32 v6, v0, 16, 1
-; GFX942-NEXT: v_bfe_u32 v8, v5, 16, 1
-; GFX942-NEXT: v_or_b32_e32 v7, 0x400000, v0
-; GFX942-NEXT: v_or_b32_e32 v9, 0x400000, v5
-; GFX942-NEXT: v_add3_u32 v6, v6, v0, s8
-; GFX942-NEXT: v_add3_u32 v8, v8, v5, s8
-; GFX942-NEXT: v_cmp_u_f32_e32 vcc, v5, v5
-; GFX942-NEXT: v_cmp_u_f32_e64 s[4:5], v0, v0
+; GFX942-NEXT: v_mov_b32_e32 v7, v2
+; GFX942-NEXT: v_lshlrev_b32_e32 v2, 16, v7
+; GFX942-NEXT: v_and_b32_e32 v3, 0xffff0000, v7
+; GFX942-NEXT: v_min_f32_e32 v2, v2, v1
+; GFX942-NEXT: v_min_f32_e32 v3, v3, v0
+; GFX942-NEXT: v_bfe_u32 v5, v2, 16, 1
+; GFX942-NEXT: v_bfe_u32 v8, v3, 16, 1
+; GFX942-NEXT: v_or_b32_e32 v6, 0x400000, v2
+; GFX942-NEXT: v_or_b32_e32 v9, 0x400000, v3
+; GFX942-NEXT: v_add3_u32 v5, v5, v2, s8
+; GFX942-NEXT: v_add3_u32 v8, v8, v3, s8
+; GFX942-NEXT: v_cmp_u_f32_e32 vcc, v3, v3
+; GFX942-NEXT: v_cmp_u_f32_e64 s[4:5], v2, v2
; GFX942-NEXT: buffer_wbl2 sc1
-; GFX942-NEXT: v_cndmask_b32_e32 v5, v8, v9, vcc
-; GFX942-NEXT: v_cndmask_b32_e64 v0, v6, v7, s[4:5]
-; GFX942-NEXT: v_perm_b32 v0, v5, v0, s9
-; GFX942-NEXT: v_mov_b64_e32 v[6:7], v[0:1]
-; GFX942-NEXT: buffer_atomic_cmpswap v[6:7], v4, s[0:3], 0 offen sc0
+; GFX942-NEXT: v_cndmask_b32_e32 v3, v8, v9, vcc
+; GFX942-NEXT: v_cndmask_b32_e64 v2, v5, v6, s[4:5]
+; GFX942-NEXT: v_perm_b32 v6, v3, v2, s9
+; GFX942-NEXT: v_mov_b64_e32 v[2:3], v[6:7]
+; GFX942-NEXT: buffer_atomic_cmpswap v[2:3], v4, s[0:3], 0 offen sc0
; GFX942-NEXT: s_waitcnt vmcnt(0)
; GFX942-NEXT: buffer_inv sc1
-; GFX942-NEXT: v_cmp_eq_u32_e32 vcc, v6, v1
+; GFX942-NEXT: v_cmp_eq_u32_e32 vcc, v2, v7
; GFX942-NEXT: s_or_b64 s[6:7], vcc, s[6:7]
-; GFX942-NEXT: v_mov_b32_e32 v1, v6
; GFX942-NEXT: s_andn2_b64 exec, exec, s[6:7]
; GFX942-NEXT: s_cbranch_execnz .LBB20_1
; GFX942-NEXT: ; %bb.2: ; %atomicrmw.end
@@ -8062,10 +8086,11 @@ define void @buffer_fat_ptr_agent_atomic_fmin_noret_v2bf16__offset__amdgpu_no_fi
; GFX11-TRUE16-LABEL: buffer_fat_ptr_agent_atomic_fmin_noret_v2bf16__offset__amdgpu_no_fine_grained_memory:
; GFX11-TRUE16: ; %bb.0:
; GFX11-TRUE16-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
-; GFX11-TRUE16-NEXT: v_dual_mov_b32 v1, s16 :: v_dual_and_b32 v2, 0xffff0000, v0
+; GFX11-TRUE16-NEXT: v_mov_b32_e32 v1, s16
; GFX11-TRUE16-NEXT: s_add_i32 s4, s16, 0x400
; GFX11-TRUE16-NEXT: s_delay_alu instid0(SALU_CYCLE_1)
-; GFX11-TRUE16-NEXT: v_dual_mov_b32 v4, s4 :: v_dual_lshlrev_b32 v3, 16, v0
+; GFX11-TRUE16-NEXT: v_dual_mov_b32 v4, s4 :: v_dual_and_b32 v3, 0xffff0000, v0
+; GFX11-TRUE16-NEXT: v_lshlrev_b32_e32 v0, 16, v0
; GFX11-TRUE16-NEXT: buffer_load_b32 v1, v1, s[0:3], 0 offen offset:1024
; GFX11-TRUE16-NEXT: s_mov_b32 s4, 0
; GFX11-TRUE16-NEXT: s_set_inst_prefetch_distance 0x1
@@ -8073,34 +8098,35 @@ define void @buffer_fat_ptr_agent_atomic_fmin_noret_v2bf16__offset__amdgpu_no_fi
; GFX11-TRUE16-NEXT: .LBB20_1: ; %atomicrmw.start
; GFX11-TRUE16-NEXT: ; =>This Inner Loop Header: Depth=1
; GFX11-TRUE16-NEXT: s_waitcnt vmcnt(0)
-; GFX11-TRUE16-NEXT: v_and_b32_e32 v5, 0xffff0000, v1
-; GFX11-TRUE16-NEXT: v_lshlrev_b32_e32 v0, 16, v1
+; GFX11-TRUE16-NEXT: v_mov_b32_e32 v6, v1
; GFX11-TRUE16-NEXT: s_waitcnt_vscnt null, 0x0
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
-; GFX11-TRUE16-NEXT: v_dual_min_f32 v5, v5, v2 :: v_dual_min_f32 v0, v0, v3
-; GFX11-TRUE16-NEXT: v_bfe_u32 v7, v5, 16, 1
+; GFX11-TRUE16-NEXT: v_and_b32_e32 v2, 0xffff0000, v6
+; GFX11-TRUE16-NEXT: v_dual_min_f32 v2, v2, v3 :: v_dual_lshlrev_b32 v1, 16, v6
+; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_2)
+; GFX11-TRUE16-NEXT: v_min_f32_e32 v1, v1, v0
+; GFX11-TRUE16-NEXT: v_bfe_u32 v7, v2, 16, 1
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_2)
-; GFX11-TRUE16-NEXT: v_bfe_u32 v6, v0, 16, 1
-; GFX11-TRUE16-NEXT: v_or_b32_e32 v8, 0x400000, v0
-; GFX11-TRUE16-NEXT: v_cmp_u_f32_e32 vcc_lo, v0, v0
-; GFX11-TRUE16-NEXT: v_or_b32_e32 v9, 0x400000, v5
-; GFX11-TRUE16-NEXT: v_add3_u32 v7, v7, v5, 0x7fff
-; GFX11-TRUE16-NEXT: v_add3_u32 v6, v6, v0, 0x7fff
+; GFX11-TRUE16-NEXT: v_bfe_u32 v5, v1, 16, 1
+; GFX11-TRUE16-NEXT: v_or_b32_e32 v8, 0x400000, v1
+; GFX11-TRUE16-NEXT: v_cmp_u_f32_e32 vcc_lo, v1, v1
+; GFX11-TRUE16-NEXT: v_or_b32_e32 v9, 0x400000, v2
+; GFX11-TRUE16-NEXT: v_add3_u32 v7, v7, v2, 0x7fff
+; GFX11-TRUE16-NEXT: v_add3_u32 v5, v5, v1, 0x7fff
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(SKIP_1) | instid1(VALU_DEP_4)
-; GFX11-TRUE16-NEXT: v_cndmask_b32_e32 v0, v6, v8, vcc_lo
-; GFX11-TRUE16-NEXT: v_cmp_u_f32_e32 vcc_lo, v5, v5
-; GFX11-TRUE16-NEXT: v_cndmask_b32_e32 v5, v7, v9, vcc_lo
+; GFX11-TRUE16-NEXT: v_cndmask_b32_e32 v1, v5, v8, vcc_lo
+; GFX11-TRUE16-NEXT: v_cmp_u_f32_e32 vcc_lo, v2, v2
+; GFX11-TRUE16-NEXT: v_cndmask_b32_e32 v2, v7, v9, vcc_lo
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_3) | instskip(NEXT) | instid1(VALU_DEP_1)
-; GFX11-TRUE16-NEXT: v_mov_b16_e32 v0.l, v0.h
-; GFX11-TRUE16-NEXT: v_bfi_b32 v0, 0xffff, v0, v5
+; GFX11-TRUE16-NEXT: v_mov_b16_e32 v1.l, v1.h
+; GFX11-TRUE16-NEXT: v_bfi_b32 v5, 0xffff, v1, v2
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1)
-; GFX11-TRUE16-NEXT: v_dual_mov_b32 v6, v1 :: v_dual_mov_b32 v5, v0
-; GFX11-TRUE16-NEXT: buffer_atomic_cmpswap_b32 v[5:6], v4, s[0:3], 0 offen glc
+; GFX11-TRUE16-NEXT: v_dual_mov_b32 v1, v5 :: v_dual_mov_b32 v2, v6
+; GFX11-TRUE16-NEXT: buffer_atomic_cmpswap_b32 v[1:2], v4, s[0:3], 0 offen glc
; GFX11-TRUE16-NEXT: s_waitcnt vmcnt(0)
; GFX11-TRUE16-NEXT: buffer_gl1_inv
; GFX11-TRUE16-NEXT: buffer_gl0_inv
-; GFX11-TRUE16-NEXT: v_cmp_eq_u32_e32 vcc_lo, v5, v1
-; GFX11-TRUE16-NEXT: v_mov_b32_e32 v1, v5
+; GFX11-TRUE16-NEXT: v_cmp_eq_u32_e32 vcc_lo, v1, v6
; GFX11-TRUE16-NEXT: s_or_b32 s4, vcc_lo, s4
; GFX11-TRUE16-NEXT: s_delay_alu instid0(SALU_CYCLE_1)
; GFX11-TRUE16-NEXT: s_and_not1_b32 exec_lo, exec_lo, s4
@@ -8113,10 +8139,11 @@ define void @buffer_fat_ptr_agent_atomic_fmin_noret_v2bf16__offset__amdgpu_no_fi
; GFX11-FAKE16-LABEL: buffer_fat_ptr_agent_atomic_fmin_noret_v2bf16__offset__amdgpu_no_fine_grained_memory:
; GFX11-FAKE16: ; %bb.0:
; GFX11-FAKE16-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
-; GFX11-FAKE16-NEXT: v_dual_mov_b32 v1, s16 :: v_dual_lshlrev_b32 v2, 16, v0
+; GFX11-FAKE16-NEXT: v_mov_b32_e32 v1, s16
; GFX11-FAKE16-NEXT: s_add_i32 s4, s16, 0x400
; GFX11-FAKE16-NEXT: s_delay_alu instid0(SALU_CYCLE_1)
-; GFX11-FAKE16-NEXT: v_dual_mov_b32 v4, s4 :: v_dual_and_b32 v3, 0xffff0000, v0
+; GFX11-FAKE16-NEXT: v_dual_mov_b32 v4, s4 :: v_dual_lshlrev_b32 v3, 16, v0
+; GFX11-FAKE16-NEXT: v_and_b32_e32 v0, 0xffff0000, v0
; GFX11-FAKE16-NEXT: buffer_load_b32 v1, v1, s[0:3], 0 offen offset:1024
; GFX11-FAKE16-NEXT: s_mov_b32 s5, 0
; GFX11-FAKE16-NEXT: s_set_inst_prefetch_distance 0x1
@@ -8124,32 +8151,33 @@ define void @buffer_fat_ptr_agent_atomic_fmin_noret_v2bf16__offset__amdgpu_no_fi
; GFX11-FAKE16-NEXT: .LBB20_1: ; %atomicrmw.start
; GFX11-FAKE16-NEXT: ; =>This Inner Loop Header: Depth=1
; GFX11-FAKE16-NEXT: s_waitcnt vmcnt(0)
-; GFX11-FAKE16-NEXT: v_and_b32_e32 v5, 0xffff0000, v1
-; GFX11-FAKE16-NEXT: v_lshlrev_b32_e32 v0, 16, v1
+; GFX11-FAKE16-NEXT: v_mov_b32_e32 v6, v1
; GFX11-FAKE16-NEXT: s_waitcnt_vscnt null, 0x0
; GFX11-FAKE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
-; GFX11-FAKE16-NEXT: v_dual_min_f32 v5, v5, v3 :: v_dual_min_f32 v0, v0, v2
-; GFX11-FAKE16-NEXT: v_bfe_u32 v7, v5, 16, 1
+; GFX11-FAKE16-NEXT: v_and_b32_e32 v2, 0xffff0000, v6
+; GFX11-FAKE16-NEXT: v_dual_min_f32 v2, v2, v0 :: v_dual_lshlrev_b32 v1, 16, v6
+; GFX11-FAKE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_2)
+; GFX11-FAKE16-NEXT: v_min_f32_e32 v1, v1, v3
+; GFX11-FAKE16-NEXT: v_bfe_u32 v7, v2, 16, 1
; GFX11-FAKE16-NEXT: s_delay_alu instid0(VALU_DEP_2)
-; GFX11-FAKE16-NEXT: v_bfe_u32 v6, v0, 16, 1
-; GFX11-FAKE16-NEXT: v_or_b32_e32 v8, 0x400000, v0
-; GFX11-FAKE16-NEXT: v_or_b32_e32 v9, 0x400000, v5
-; GFX11-FAKE16-NEXT: v_cmp_u_f32_e32 vcc_lo, v5, v5
-; GFX11-FAKE16-NEXT: v_add3_u32 v7, v7, v5, 0x7fff
-; GFX11-FAKE16-NEXT: v_add3_u32 v6, v6, v0, 0x7fff
-; GFX11-FAKE16-NEXT: v_cmp_u_f32_e64 s4, v0, v0
+; GFX11-FAKE16-NEXT: v_bfe_u32 v5, v1, 16, 1
+; GFX11-FAKE16-NEXT: v_or_b32_e32 v8, 0x400000, v1
+; GFX11-FAKE16-NEXT: v_or_b32_e32 v9, 0x400000, v2
+; GFX11-FAKE16-NEXT: v_cmp_u_f32_e32 vcc_lo, v2, v2
+; GFX11-FAKE16-NEXT: v_add3_u32 v7, v7, v2, 0x7fff
+; GFX11-FAKE16-NEXT: v_add3_u32 v5, v5, v1, 0x7fff
+; GFX11-FAKE16-NEXT: v_cmp_u_f32_e64 s4, v1, v1
; GFX11-FAKE16-NEXT: s_delay_alu instid0(VALU_DEP_3) | instskip(NEXT) | instid1(VALU_DEP_2)
-; GFX11-FAKE16-NEXT: v_cndmask_b32_e32 v5, v7, v9, vcc_lo
-; GFX11-FAKE16-NEXT: v_cndmask_b32_e64 v0, v6, v8, s4
+; GFX11-FAKE16-NEXT: v_cndmask_b32_e32 v2, v7, v9, vcc_lo
+; GFX11-FAKE16-NEXT: v_cndmask_b32_e64 v1, v5, v8, s4
; GFX11-FAKE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
-; GFX11-FAKE16-NEXT: v_perm_b32 v0, v5, v0, 0x7060302
-; GFX11-FAKE16-NEXT: v_dual_mov_b32 v6, v1 :: v_dual_mov_b32 v5, v0
-; GFX11-FAKE16-NEXT: buffer_atomic_cmpswap_b32 v[5:6], v4, s[0:3], 0 offen glc
+; GFX11-FAKE16-NEXT: v_perm_b32 v5, v2, v1, 0x7060302
+; GFX11-FAKE16-NEXT: v_dual_mov_b32 v1, v5 :: v_dual_mov_b32 v2, v6
+; GFX11-FAKE16-NEXT: buffer_atomic_cmpswap_b32 v[1:2], v4, s[0:3], 0 offen glc
; GFX11-FAKE16-NEXT: s_waitcnt vmcnt(0)
; GFX11-FAKE16-NEXT: buffer_gl1_inv
; GFX11-FAKE16-NEXT: buffer_gl0_inv
-; GFX11-FAKE16-NEXT: v_cmp_eq_u32_e32 vcc_lo, v5, v1
-; GFX11-FAKE16-NEXT: v_mov_b32_e32 v1, v5
+; GFX11-FAKE16-NEXT: v_cmp_eq_u32_e32 vcc_lo, v1, v6
; GFX11-FAKE16-NEXT: s_or_b32 s5, vcc_lo, s5
; GFX11-FAKE16-NEXT: s_delay_alu instid0(SALU_CYCLE_1)
; GFX11-FAKE16-NEXT: s_and_not1_b32 exec_lo, exec_lo, s5
@@ -8164,38 +8192,38 @@ define void @buffer_fat_ptr_agent_atomic_fmin_noret_v2bf16__offset__amdgpu_no_fi
; GFX10-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX10-NEXT: v_mov_b32_e32 v1, s20
; GFX10-NEXT: s_add_i32 s4, s20, 0x400
-; GFX10-NEXT: v_lshlrev_b32_e32 v2, 16, v0
-; GFX10-NEXT: v_and_b32_e32 v3, 0xffff0000, v0
+; GFX10-NEXT: v_lshlrev_b32_e32 v3, 16, v0
+; GFX10-NEXT: v_and_b32_e32 v0, 0xffff0000, v0
; GFX10-NEXT: v_mov_b32_e32 v4, s4
; GFX10-NEXT: buffer_load_dword v1, v1, s[16:19], 0 offen offset:1024
; GFX10-NEXT: s_mov_b32 s5, 0
; GFX10-NEXT: .LBB20_1: ; %atomicrmw.start
; GFX10-NEXT: ; =>This Inner Loop Header: Depth=1
; GFX10-NEXT: s_waitcnt vmcnt(0)
-; GFX10-NEXT: v_lshlrev_b32_e32 v0, 16, v1
-; GFX10-NEXT: v_and_b32_e32 v5, 0xffff0000, v1
-; GFX10-NEXT: s_waitcnt_vscnt null, 0x0
-; GFX10-NEXT: v_min_f32_e32 v0, v0, v2
-; GFX10-NEXT: v_min_f32_e32 v5, v5, v3
-; GFX10-NEXT: v_bfe_u32 v6, v0, 16, 1
-; GFX10-NEXT: v_bfe_u32 v7, v5, 16, 1
-; GFX10-NEXT: v_or_b32_e32 v8, 0x400000, v0
-; GFX10-NEXT: v_or_b32_e32 v9, 0x400000, v5
-; GFX10-NEXT: v_cmp_u_f32_e32 vcc_lo, v5, v5
-; GFX10-NEXT: v_add3_u32 v6, v6, v0, 0x7fff
-; GFX10-NEXT: v_add3_u32 v7, v7, v5, 0x7fff
-; GFX10-NEXT: v_cmp_u_f32_e64 s4, v0, v0
-; GFX10-NEXT: v_cndmask_b32_e32 v5, v7, v9, vcc_lo
-; GFX10-NEXT: v_cndmask_b32_e64 v0, v6, v8, s4
-; GFX10-NEXT: v_perm_b32 v0, v5, v0, 0x7060302
; GFX10-NEXT: v_mov_b32_e32 v6, v1
-; GFX10-NEXT: v_mov_b32_e32 v5, v0
-; GFX10-NEXT: buffer_atomic_cmpswap v[5:6], v4, s[16:19], 0 offen glc
+; GFX10-NEXT: s_waitcnt_vscnt null, 0x0
+; GFX10-NEXT: v_lshlrev_b32_e32 v1, 16, v6
+; GFX10-NEXT: v_and_b32_e32 v2, 0xffff0000, v6
+; GFX10-NEXT: v_min_f32_e32 v1, v1, v3
+; GFX10-NEXT: v_min_f32_e32 v2, v2, v0
+; GFX10-NEXT: v_bfe_u32 v5, v1, 16, 1
+; GFX10-NEXT: v_bfe_u32 v7, v2, 16, 1
+; GFX10-NEXT: v_or_b32_e32 v8, 0x400000, v1
+; GFX10-NEXT: v_or_b32_e32 v9, 0x400000, v2
+; GFX10-NEXT: v_cmp_u_f32_e32 vcc_lo, v2, v2
+; GFX10-NEXT: v_add3_u32 v5, v5, v1, 0x7fff
+; GFX10-NEXT: v_add3_u32 v7, v7, v2, 0x7fff
+; GFX10-NEXT: v_cmp_u_f32_e64 s4, v1, v1
+; GFX10-NEXT: v_cndmask_b32_e32 v2, v7, v9, vcc_lo
+; GFX10-NEXT: v_cndmask_b32_e64 v1, v5, v8, s4
+; GFX10-NEXT: v_perm_b32 v5, v2, v1, 0x7060302
+; GFX10-NEXT: v_mov_b32_e32 v1, v5
+; GFX10-NEXT: v_mov_b32_e32 v2, v6
+; GFX10-NEXT: buffer_atomic_cmpswap v[1:2], v4, s[16:19], 0 offen glc
; GFX10-NEXT: s_waitcnt vmcnt(0)
; GFX10-NEXT: buffer_gl1_inv
; GFX10-NEXT: buffer_gl0_inv
-; GFX10-NEXT: v_cmp_eq_u32_e32 vcc_lo, v5, v1
-; GFX10-NEXT: v_mov_b32_e32 v1, v5
+; GFX10-NEXT: v_cmp_eq_u32_e32 vcc_lo, v1, v6
; GFX10-NEXT: s_or_b32 s5, vcc_lo, s5
; GFX10-NEXT: s_andn2_b32 exec_lo, exec_lo, s5
; GFX10-NEXT: s_cbranch_execnz .LBB20_1
@@ -8207,39 +8235,39 @@ define void @buffer_fat_ptr_agent_atomic_fmin_noret_v2bf16__offset__amdgpu_no_fi
; GFX90A: ; %bb.0:
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX90A-NEXT: v_mov_b32_e32 v1, s20
-; GFX90A-NEXT: buffer_load_dword v1, v1, s[16:19], 0 offen offset:1024
+; GFX90A-NEXT: buffer_load_dword v2, v1, s[16:19], 0 offen offset:1024
; GFX90A-NEXT: s_add_i32 s4, s20, 0x400
; GFX90A-NEXT: s_mov_b64 s[6:7], 0
-; GFX90A-NEXT: v_lshlrev_b32_e32 v2, 16, v0
+; GFX90A-NEXT: v_lshlrev_b32_e32 v1, 16, v0
; GFX90A-NEXT: s_movk_i32 s8, 0x7fff
-; GFX90A-NEXT: v_and_b32_e32 v3, 0xffff0000, v0
+; GFX90A-NEXT: v_and_b32_e32 v0, 0xffff0000, v0
; GFX90A-NEXT: s_mov_b32 s9, 0x7060302
; GFX90A-NEXT: v_mov_b32_e32 v4, s4
; GFX90A-NEXT: .LBB20_1: ; %atomicrmw.start
; GFX90A-NEXT: ; =>This Inner Loop Header: Depth=1
; GFX90A-NEXT: s_waitcnt vmcnt(0)
-; GFX90A-NEXT: v_lshlrev_b32_e32 v0, 16, v1
-; GFX90A-NEXT: v_and_b32_e32 v5, 0xffff0000, v1
-; GFX90A-NEXT: v_min_f32_e32 v0, v0, v2
-; GFX90A-NEXT: v_min_f32_e32 v5, v5, v3
-; GFX90A-NEXT: v_bfe_u32 v6, v0, 16, 1
-; GFX90A-NEXT: v_bfe_u32 v8, v5, 16, 1
-; GFX90A-NEXT: v_or_b32_e32 v7, 0x400000, v0
-; GFX90A-NEXT: v_or_b32_e32 v9, 0x400000, v5
-; GFX90A-NEXT: v_add3_u32 v6, v6, v0, s8
-; GFX90A-NEXT: v_add3_u32 v8, v8, v5, s8
-; GFX90A-NEXT: v_cmp_u_f32_e32 vcc, v5, v5
-; GFX90A-NEXT: v_cmp_u_f32_e64 s[4:5], v0, v0
-; GFX90A-NEXT: v_cndmask_b32_e64 v0, v6, v7, s[4:5]
-; GFX90A-NEXT: v_cndmask_b32_e32 v5, v8, v9, vcc
-; GFX90A-NEXT: v_perm_b32 v0, v5, v0, s9
-; GFX90A-NEXT: v_pk_mov_b32 v[6:7], v[0:1], v[0:1] op_sel:[0,1]
-; GFX90A-NEXT: buffer_atomic_cmpswap v[6:7], v4, s[16:19], 0 offen glc
+; GFX90A-NEXT: v_mov_b32_e32 v7, v2
+; GFX90A-NEXT: v_lshlrev_b32_e32 v2, 16, v7
+; GFX90A-NEXT: v_and_b32_e32 v3, 0xffff0000, v7
+; GFX90A-NEXT: v_min_f32_e32 v2, v2, v1
+; GFX90A-NEXT: v_min_f32_e32 v3, v3, v0
+; GFX90A-NEXT: v_bfe_u32 v5, v2, 16, 1
+; GFX90A-NEXT: v_bfe_u32 v8, v3, 16, 1
+; GFX90A-NEXT: v_or_b32_e32 v6, 0x400000, v2
+; GFX90A-NEXT: v_or_b32_e32 v9, 0x400000, v3
+; GFX90A-NEXT: v_add3_u32 v5, v5, v2, s8
+; GFX90A-NEXT: v_add3_u32 v8, v8, v3, s8
+; GFX90A-NEXT: v_cmp_u_f32_e32 vcc, v3, v3
+; GFX90A-NEXT: v_cmp_u_f32_e64 s[4:5], v2, v2
+; GFX90A-NEXT: v_cndmask_b32_e64 v2, v5, v6, s[4:5]
+; GFX90A-NEXT: v_cndmask_b32_e32 v3, v8, v9, vcc
+; GFX90A-NEXT: v_perm_b32 v6, v3, v2, s9
+; GFX90A-NEXT: v_pk_mov_b32 v[2:3], v[6:7], v[6:7] op_sel:[0,1]
+; GFX90A-NEXT: buffer_atomic_cmpswap v[2:3], v4, s[16:19], 0 offen glc
; GFX90A-NEXT: s_waitcnt vmcnt(0)
; GFX90A-NEXT: buffer_wbinvl1
-; GFX90A-NEXT: v_cmp_eq_u32_e32 vcc, v6, v1
+; GFX90A-NEXT: v_cmp_eq_u32_e32 vcc, v2, v7
; GFX90A-NEXT: s_or_b64 s[6:7], vcc, s[6:7]
-; GFX90A-NEXT: v_mov_b32_e32 v1, v6
; GFX90A-NEXT: s_andn2_b64 exec, exec, s[6:7]
; GFX90A-NEXT: s_cbranch_execnz .LBB20_1
; GFX90A-NEXT: ; %bb.2: ; %atomicrmw.end
@@ -8253,37 +8281,37 @@ define void @buffer_fat_ptr_agent_atomic_fmin_noret_v2bf16__offset__amdgpu_no_fi
; GFX908-NEXT: buffer_load_dword v1, v1, s[16:19], 0 offen offset:1024
; GFX908-NEXT: s_add_i32 s4, s20, 0x400
; GFX908-NEXT: s_mov_b64 s[6:7], 0
-; GFX908-NEXT: v_lshlrev_b32_e32 v2, 16, v0
+; GFX908-NEXT: v_lshlrev_b32_e32 v3, 16, v0
; GFX908-NEXT: s_movk_i32 s8, 0x7fff
-; GFX908-NEXT: v_and_b32_e32 v3, 0xffff0000, v0
+; GFX908-NEXT: v_and_b32_e32 v0, 0xffff0000, v0
; GFX908-NEXT: s_mov_b32 s9, 0x7060302
; GFX908-NEXT: v_mov_b32_e32 v4, s4
; GFX908-NEXT: .LBB20_1: ; %atomicrmw.start
; GFX908-NEXT: ; =>This Inner Loop Header: Depth=1
; GFX908-NEXT: s_waitcnt vmcnt(0)
-; GFX908-NEXT: v_lshlrev_b32_e32 v0, 16, v1
-; GFX908-NEXT: v_and_b32_e32 v5, 0xffff0000, v1
-; GFX908-NEXT: v_min_f32_e32 v0, v0, v2
-; GFX908-NEXT: v_min_f32_e32 v5, v5, v3
-; GFX908-NEXT: v_bfe_u32 v6, v0, 16, 1
-; GFX908-NEXT: v_bfe_u32 v8, v5, 16, 1
-; GFX908-NEXT: v_or_b32_e32 v7, 0x400000, v0
-; GFX908-NEXT: v_or_b32_e32 v9, 0x400000, v5
-; GFX908-NEXT: v_add3_u32 v6, v6, v0, s8
-; GFX908-NEXT: v_add3_u32 v8, v8, v5, s8
-; GFX908-NEXT: v_cmp_u_f32_e32 vcc, v5, v5
-; GFX908-NEXT: v_cmp_u_f32_e64 s[4:5], v0, v0
-; GFX908-NEXT: v_cndmask_b32_e64 v0, v6, v7, s[4:5]
-; GFX908-NEXT: v_cndmask_b32_e32 v5, v8, v9, vcc
-; GFX908-NEXT: v_perm_b32 v0, v5, v0, s9
; GFX908-NEXT: v_mov_b32_e32 v6, v1
-; GFX908-NEXT: v_mov_b32_e32 v5, v0
-; GFX908-NEXT: buffer_atomic_cmpswap v[5:6], v4, s[16:19], 0 offen glc
+; GFX908-NEXT: v_lshlrev_b32_e32 v1, 16, v6
+; GFX908-NEXT: v_and_b32_e32 v2, 0xffff0000, v6
+; GFX908-NEXT: v_min_f32_e32 v1, v1, v3
+; GFX908-NEXT: v_min_f32_e32 v2, v2, v0
+; GFX908-NEXT: v_bfe_u32 v5, v1, 16, 1
+; GFX908-NEXT: v_bfe_u32 v8, v2, 16, 1
+; GFX908-NEXT: v_or_b32_e32 v7, 0x400000, v1
+; GFX908-NEXT: v_or_b32_e32 v9, 0x400000, v2
+; GFX908-NEXT: v_add3_u32 v5, v5, v1, s8
+; GFX908-NEXT: v_add3_u32 v8, v8, v2, s8
+; GFX908-NEXT: v_cmp_u_f32_e32 vcc, v2, v2
+; GFX908-NEXT: v_cmp_u_f32_e64 s[4:5], v1, v1
+; GFX908-NEXT: v_cndmask_b32_e64 v1, v5, v7, s[4:5]
+; GFX908-NEXT: v_cndmask_b32_e32 v2, v8, v9, vcc
+; GFX908-NEXT: v_perm_b32 v5, v2, v1, s9
+; GFX908-NEXT: v_mov_b32_e32 v1, v5
+; GFX908-NEXT: v_mov_b32_e32 v2, v6
+; GFX908-NEXT: buffer_atomic_cmpswap v[1:2], v4, s[16:19], 0 offen glc
; GFX908-NEXT: s_waitcnt vmcnt(0)
; GFX908-NEXT: buffer_wbinvl1
-; GFX908-NEXT: v_cmp_eq_u32_e32 vcc, v5, v1
+; GFX908-NEXT: v_cmp_eq_u32_e32 vcc, v1, v6
; GFX908-NEXT: s_or_b64 s[6:7], vcc, s[6:7]
-; GFX908-NEXT: v_mov_b32_e32 v1, v5
; GFX908-NEXT: s_andn2_b64 exec, exec, s[6:7]
; GFX908-NEXT: s_cbranch_execnz .LBB20_1
; GFX908-NEXT: ; %bb.2: ; %atomicrmw.end
@@ -8297,38 +8325,38 @@ define void @buffer_fat_ptr_agent_atomic_fmin_noret_v2bf16__offset__amdgpu_no_fi
; GFX8-NEXT: buffer_load_dword v1, v1, s[16:19], 0 offen offset:1024
; GFX8-NEXT: s_add_i32 s4, s20, 0x400
; GFX8-NEXT: s_mov_b64 s[6:7], 0
-; GFX8-NEXT: v_lshlrev_b32_e32 v2, 16, v0
-; GFX8-NEXT: v_and_b32_e32 v3, 0xffff0000, v0
+; GFX8-NEXT: v_lshlrev_b32_e32 v3, 16, v0
+; GFX8-NEXT: v_and_b32_e32 v0, 0xffff0000, v0
; GFX8-NEXT: v_mov_b32_e32 v4, s4
; GFX8-NEXT: .LBB20_1: ; %atomicrmw.start
; GFX8-NEXT: ; =>This Inner Loop Header: Depth=1
; GFX8-NEXT: s_waitcnt vmcnt(0)
-; GFX8-NEXT: v_lshlrev_b32_e32 v0, 16, v1
-; GFX8-NEXT: v_and_b32_e32 v5, 0xffff0000, v1
-; GFX8-NEXT: v_min_f32_e32 v0, v0, v2
-; GFX8-NEXT: v_min_f32_e32 v5, v5, v3
-; GFX8-NEXT: v_bfe_u32 v6, v0, 16, 1
-; GFX8-NEXT: v_bfe_u32 v8, v5, 16, 1
-; GFX8-NEXT: v_add_u32_e32 v6, vcc, v6, v0
-; GFX8-NEXT: v_add_u32_e32 v8, vcc, v8, v5
-; GFX8-NEXT: v_add_u32_e32 v6, vcc, 0x7fff, v6
-; GFX8-NEXT: v_add_u32_e32 v8, vcc, 0x7fff, v8
-; GFX8-NEXT: v_or_b32_e32 v9, 0x400000, v5
-; GFX8-NEXT: v_cmp_u_f32_e32 vcc, v5, v5
-; GFX8-NEXT: v_or_b32_e32 v7, 0x400000, v0
-; GFX8-NEXT: v_cmp_u_f32_e64 s[4:5], v0, v0
-; GFX8-NEXT: v_cndmask_b32_e32 v5, v8, v9, vcc
-; GFX8-NEXT: v_cndmask_b32_e64 v0, v6, v7, s[4:5]
-; GFX8-NEXT: v_lshrrev_b32_e32 v5, 16, v5
-; GFX8-NEXT: v_alignbit_b32 v0, v5, v0, 16
; GFX8-NEXT: v_mov_b32_e32 v6, v1
-; GFX8-NEXT: v_mov_b32_e32 v5, v0
-; GFX8-NEXT: buffer_atomic_cmpswap v[5:6], v4, s[16:19], 0 offen glc
+; GFX8-NEXT: v_lshlrev_b32_e32 v1, 16, v6
+; GFX8-NEXT: v_and_b32_e32 v2, 0xffff0000, v6
+; GFX8-NEXT: v_min_f32_e32 v1, v1, v3
+; GFX8-NEXT: v_min_f32_e32 v2, v2, v0
+; GFX8-NEXT: v_bfe_u32 v5, v1, 16, 1
+; GFX8-NEXT: v_bfe_u32 v8, v2, 16, 1
+; GFX8-NEXT: v_add_u32_e32 v5, vcc, v5, v1
+; GFX8-NEXT: v_add_u32_e32 v8, vcc, v8, v2
+; GFX8-NEXT: v_add_u32_e32 v5, vcc, 0x7fff, v5
+; GFX8-NEXT: v_add_u32_e32 v8, vcc, 0x7fff, v8
+; GFX8-NEXT: v_or_b32_e32 v9, 0x400000, v2
+; GFX8-NEXT: v_cmp_u_f32_e32 vcc, v2, v2
+; GFX8-NEXT: v_or_b32_e32 v7, 0x400000, v1
+; GFX8-NEXT: v_cmp_u_f32_e64 s[4:5], v1, v1
+; GFX8-NEXT: v_cndmask_b32_e32 v2, v8, v9, vcc
+; GFX8-NEXT: v_cndmask_b32_e64 v1, v5, v7, s[4:5]
+; GFX8-NEXT: v_lshrrev_b32_e32 v2, 16, v2
+; GFX8-NEXT: v_alignbit_b32 v5, v2, v1, 16
+; GFX8-NEXT: v_mov_b32_e32 v1, v5
+; GFX8-NEXT: v_mov_b32_e32 v2, v6
+; GFX8-NEXT: buffer_atomic_cmpswap v[1:2], v4, s[16:19], 0 offen glc
; GFX8-NEXT: s_waitcnt vmcnt(0)
; GFX8-NEXT: buffer_wbinvl1
-; GFX8-NEXT: v_cmp_eq_u32_e32 vcc, v5, v1
+; GFX8-NEXT: v_cmp_eq_u32_e32 vcc, v1, v6
; GFX8-NEXT: s_or_b64 s[6:7], vcc, s[6:7]
-; GFX8-NEXT: v_mov_b32_e32 v1, v5
; GFX8-NEXT: s_andn2_b64 exec, exec, s[6:7]
; GFX8-NEXT: s_cbranch_execnz .LBB20_1
; GFX8-NEXT: ; %bb.2: ; %atomicrmw.end
diff --git a/llvm/test/CodeGen/AMDGPU/div_i128.ll b/llvm/test/CodeGen/AMDGPU/div_i128.ll
index 51398a45055eb..1396099dbfa6a 100644
--- a/llvm/test/CodeGen/AMDGPU/div_i128.ll
+++ b/llvm/test/CodeGen/AMDGPU/div_i128.ll
@@ -220,19 +220,22 @@ define i128 @v_sdiv_i128_vv(i128 %lhs, i128 %rhs) {
; GFX9-O0: ; %bb.0: ; %_udiv-special-cases
; GFX9-O0-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX9-O0-NEXT: s_xor_saveexec_b64 s[4:5], -1
-; GFX9-O0-NEXT: buffer_store_dword v30, off, s[0:3], s32 offset:312 ; 4-byte Folded Spill
+; GFX9-O0-NEXT: buffer_store_dword v29, off, s[0:3], s32 offset:316 ; 4-byte Folded Spill
; GFX9-O0-NEXT: s_mov_b64 exec, s[4:5]
-; GFX9-O0-NEXT: buffer_store_dword v7, off, s[0:3], s32 offset:84 ; 4-byte Folded Spill
+; GFX9-O0-NEXT: buffer_store_dword v7, off, s[0:3], s32 offset:88 ; 4-byte Folded Spill
; GFX9-O0-NEXT: v_mov_b32_e32 v20, v6
; GFX9-O0-NEXT: v_mov_b32_e32 v7, v4
; GFX9-O0-NEXT: v_mov_b32_e32 v9, v2
-; GFX9-O0-NEXT: v_mov_b32_e32 v2, v0
+; GFX9-O0-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:88 ; 4-byte Folded Reload
+; GFX9-O0-NEXT: s_nop 0
+; GFX9-O0-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:84 ; 4-byte Folded Spill
+; GFX9-O0-NEXT: v_mov_b32_e32 v1, v0
; GFX9-O0-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:84 ; 4-byte Folded Reload
; GFX9-O0-NEXT: ; implicit-def: $sgpr4
; GFX9-O0-NEXT: ; implicit-def: $sgpr4
; GFX9-O0-NEXT: ; kill: def $vgpr20 killed $vgpr20 def $vgpr20_vgpr21 killed $exec
-; GFX9-O0-NEXT: s_waitcnt vmcnt(0)
-; GFX9-O0-NEXT: v_mov_b32_e32 v21, v0
+; GFX9-O0-NEXT: s_waitcnt vmcnt(2)
+; GFX9-O0-NEXT: v_mov_b32_e32 v21, v2
; GFX9-O0-NEXT: ; implicit-def: $sgpr4
; GFX9-O0-NEXT: ; implicit-def: $sgpr4
; GFX9-O0-NEXT: ; kill: def $vgpr7 killed $vgpr7 def $vgpr7_vgpr8 killed $exec
@@ -243,31 +246,32 @@ define i128 @v_sdiv_i128_vv(i128 %lhs, i128 %rhs) {
; GFX9-O0-NEXT: v_mov_b32_e32 v10, v3
; GFX9-O0-NEXT: ; implicit-def: $sgpr4
; GFX9-O0-NEXT: ; implicit-def: $sgpr4
-; GFX9-O0-NEXT: ; kill: def $vgpr2 killed $vgpr2 def $vgpr2_vgpr3 killed $exec
-; GFX9-O0-NEXT: v_mov_b32_e32 v3, v1
+; GFX9-O0-NEXT: ; kill: def $vgpr1 killed $vgpr1 def $vgpr1_vgpr2 killed $exec
+; GFX9-O0-NEXT: s_waitcnt vmcnt(0)
+; GFX9-O0-NEXT: v_mov_b32_e32 v2, v0
; GFX9-O0-NEXT: ; implicit-def: $sgpr4_sgpr5
; GFX9-O0-NEXT: ; implicit-def: $sgpr4_sgpr5
; GFX9-O0-NEXT: ; implicit-def: $sgpr4_sgpr5
; GFX9-O0-NEXT: ; implicit-def: $sgpr4_sgpr5
+; GFX9-O0-NEXT: v_mov_b32_e32 v0, v1
; GFX9-O0-NEXT: v_mov_b32_e32 v1, v2
-; GFX9-O0-NEXT: v_mov_b32_e32 v2, v3
; GFX9-O0-NEXT: v_mov_b32_e32 v13, v9
; GFX9-O0-NEXT: v_mov_b32_e32 v19, v10
; GFX9-O0-NEXT: s_mov_b64 s[6:7], 0
-; GFX9-O0-NEXT: ; implicit-def: $vgpr30 : SGPR spill to VGPR lane
-; GFX9-O0-NEXT: v_writelane_b32 v30, s6, 0
-; GFX9-O0-NEXT: v_writelane_b32 v30, s7, 1
+; GFX9-O0-NEXT: ; implicit-def: $vgpr29 : SGPR spill to VGPR lane
+; GFX9-O0-NEXT: v_writelane_b32 v29, s6, 0
+; GFX9-O0-NEXT: v_writelane_b32 v29, s7, 1
; GFX9-O0-NEXT: s_mov_b32 s10, s6
-; GFX9-O0-NEXT: v_writelane_b32 v30, s10, 2
+; GFX9-O0-NEXT: v_writelane_b32 v29, s10, 2
; GFX9-O0-NEXT: s_mov_b32 s11, s7
-; GFX9-O0-NEXT: v_writelane_b32 v30, s11, 3
-; GFX9-O0-NEXT: v_sub_co_u32_e32 v5, vcc, s10, v1
-; GFX9-O0-NEXT: v_mov_b32_e32 v0, s11
-; GFX9-O0-NEXT: v_subb_co_u32_e32 v3, vcc, v0, v2, vcc
-; GFX9-O0-NEXT: v_mov_b32_e32 v0, s10
-; GFX9-O0-NEXT: v_subb_co_u32_e32 v4, vcc, v0, v13, vcc
-; GFX9-O0-NEXT: v_mov_b32_e32 v0, s11
-; GFX9-O0-NEXT: v_subb_co_u32_e32 v0, vcc, v0, v19, vcc
+; GFX9-O0-NEXT: v_writelane_b32 v29, s11, 3
+; GFX9-O0-NEXT: v_sub_co_u32_e32 v5, vcc, s10, v0
+; GFX9-O0-NEXT: v_mov_b32_e32 v2, s11
+; GFX9-O0-NEXT: v_subb_co_u32_e32 v3, vcc, v2, v1, vcc
+; GFX9-O0-NEXT: v_mov_b32_e32 v2, s10
+; GFX9-O0-NEXT: v_subb_co_u32_e32 v4, vcc, v2, v13, vcc
+; GFX9-O0-NEXT: v_mov_b32_e32 v2, s11
+; GFX9-O0-NEXT: v_subb_co_u32_e32 v2, vcc, v2, v19, vcc
; GFX9-O0-NEXT: ; implicit-def: $sgpr4
; GFX9-O0-NEXT: ; implicit-def: $sgpr4
; GFX9-O0-NEXT: ; kill: def $vgpr5 killed $vgpr5 def $vgpr5_vgpr6 killed $exec
@@ -275,25 +279,25 @@ define i128 @v_sdiv_i128_vv(i128 %lhs, i128 %rhs) {
; GFX9-O0-NEXT: v_mov_b32_e32 v3, v6
; GFX9-O0-NEXT: s_mov_b64 s[4:5], s[6:7]
; GFX9-O0-NEXT: v_cmp_lt_i64_e64 s[4:5], v[9:10], s[4:5]
-; GFX9-O0-NEXT: v_cndmask_b32_e64 v3, v2, v3, s[4:5]
-; GFX9-O0-NEXT: v_mov_b32_e32 v2, v5
-; GFX9-O0-NEXT: v_cndmask_b32_e64 v2, v1, v2, s[4:5]
+; GFX9-O0-NEXT: v_cndmask_b32_e64 v1, v1, v3, s[4:5]
+; GFX9-O0-NEXT: v_mov_b32_e32 v3, v5
+; GFX9-O0-NEXT: v_cndmask_b32_e64 v0, v0, v3, s[4:5]
; GFX9-O0-NEXT: ; implicit-def: $sgpr8
; GFX9-O0-NEXT: ; implicit-def: $sgpr8
-; GFX9-O0-NEXT: v_mov_b32_e32 v15, v2
-; GFX9-O0-NEXT: v_mov_b32_e32 v16, v3
+; GFX9-O0-NEXT: v_mov_b32_e32 v15, v0
+; GFX9-O0-NEXT: v_mov_b32_e32 v16, v1
; GFX9-O0-NEXT: ; implicit-def: $sgpr8
; GFX9-O0-NEXT: ; implicit-def: $sgpr8
; GFX9-O0-NEXT: ; kill: def $vgpr4 killed $vgpr4 def $vgpr4_vgpr5 killed $exec
-; GFX9-O0-NEXT: v_mov_b32_e32 v5, v0
-; GFX9-O0-NEXT: v_mov_b32_e32 v0, v5
-; GFX9-O0-NEXT: v_cndmask_b32_e64 v1, v19, v0, s[4:5]
-; GFX9-O0-NEXT: v_mov_b32_e32 v0, v4
-; GFX9-O0-NEXT: v_cndmask_b32_e64 v0, v13, v0, s[4:5]
+; GFX9-O0-NEXT: v_mov_b32_e32 v5, v2
+; GFX9-O0-NEXT: v_mov_b32_e32 v2, v5
+; GFX9-O0-NEXT: v_cndmask_b32_e64 v3, v19, v2, s[4:5]
+; GFX9-O0-NEXT: v_mov_b32_e32 v2, v4
+; GFX9-O0-NEXT: v_cndmask_b32_e64 v2, v13, v2, s[4:5]
; GFX9-O0-NEXT: ; implicit-def: $sgpr4
; GFX9-O0-NEXT: ; implicit-def: $sgpr4
-; GFX9-O0-NEXT: v_mov_b32_e32 v4, v0
-; GFX9-O0-NEXT: v_mov_b32_e32 v5, v1
+; GFX9-O0-NEXT: v_mov_b32_e32 v4, v2
+; GFX9-O0-NEXT: v_mov_b32_e32 v5, v3
; GFX9-O0-NEXT: v_mov_b32_e32 v6, v7
; GFX9-O0-NEXT: ; kill: def $vgpr8 killed $vgpr8 killed $vgpr7_vgpr8 killed $exec
; GFX9-O0-NEXT: v_mov_b32_e32 v12, v20
@@ -424,18 +428,18 @@ define i128 @v_sdiv_i128_vv(i128 %lhs, i128 %rhs) {
; GFX9-O0-NEXT: v_mov_b32_e32 v9, v6
; GFX9-O0-NEXT: s_mov_b64 s[8:9], s[6:7]
; GFX9-O0-NEXT: v_cmp_ne_u64_e64 s[8:9], v[4:5], s[8:9]
-; GFX9-O0-NEXT: v_ffbh_u32_e64 v4, v0
+; GFX9-O0-NEXT: v_ffbh_u32_e64 v4, v2
; GFX9-O0-NEXT: v_add_u32_e64 v4, v4, s13
-; GFX9-O0-NEXT: v_ffbh_u32_e64 v5, v1
+; GFX9-O0-NEXT: v_ffbh_u32_e64 v5, v3
; GFX9-O0-NEXT: v_min_u32_e64 v5, v4, v5
; GFX9-O0-NEXT: ; implicit-def: $sgpr16
; GFX9-O0-NEXT: v_mov_b32_e32 v4, s12
; GFX9-O0-NEXT: ; kill: def $vgpr5 killed $vgpr5 def $vgpr5_vgpr6 killed $exec
; GFX9-O0-NEXT: v_mov_b32_e32 v6, v4
; GFX9-O0-NEXT: v_mov_b32_e32 v7, v6
-; GFX9-O0-NEXT: v_ffbh_u32_e64 v4, v2
+; GFX9-O0-NEXT: v_ffbh_u32_e64 v4, v0
; GFX9-O0-NEXT: v_add_u32_e64 v4, v4, s13
-; GFX9-O0-NEXT: v_ffbh_u32_e64 v10, v3
+; GFX9-O0-NEXT: v_ffbh_u32_e64 v10, v1
; GFX9-O0-NEXT: v_min_u32_e64 v11, v4, v10
; GFX9-O0-NEXT: ; implicit-def: $sgpr13
; GFX9-O0-NEXT: v_mov_b32_e32 v4, s12
@@ -514,33 +518,33 @@ define i128 @v_sdiv_i128_vv(i128 %lhs, i128 %rhs) {
; GFX9-O0-NEXT: v_mov_b32_e32 v5, v6
; GFX9-O0-NEXT: v_cmp_ne_u64_e64 s[6:7], v[4:5], s[6:7]
; GFX9-O0-NEXT: v_mov_b32_e32 v4, s11
-; GFX9-O0-NEXT: v_cndmask_b32_e64 v4, v1, v4, s[8:9]
-; GFX9-O0-NEXT: v_mov_b32_e32 v1, s10
-; GFX9-O0-NEXT: v_cndmask_b32_e64 v0, v0, v1, s[8:9]
-; GFX9-O0-NEXT: ; implicit-def: $sgpr12
-; GFX9-O0-NEXT: ; implicit-def: $sgpr12
-; GFX9-O0-NEXT: ; kill: def $vgpr0 killed $vgpr0 def $vgpr0_vgpr1 killed $exec
-; GFX9-O0-NEXT: v_mov_b32_e32 v1, v4
-; GFX9-O0-NEXT: v_mov_b32_e32 v4, s11
; GFX9-O0-NEXT: v_cndmask_b32_e64 v4, v3, v4, s[8:9]
; GFX9-O0-NEXT: v_mov_b32_e32 v3, s10
; GFX9-O0-NEXT: v_cndmask_b32_e64 v2, v2, v3, s[8:9]
-; GFX9-O0-NEXT: ; implicit-def: $sgpr8
-; GFX9-O0-NEXT: ; implicit-def: $sgpr8
+; GFX9-O0-NEXT: ; implicit-def: $sgpr12
+; GFX9-O0-NEXT: ; implicit-def: $sgpr12
; GFX9-O0-NEXT: ; kill: def $vgpr2 killed $vgpr2 def $vgpr2_vgpr3 killed $exec
; GFX9-O0-NEXT: v_mov_b32_e32 v3, v4
-; GFX9-O0-NEXT: s_and_b64 s[6:7], s[4:5], s[6:7]
+; GFX9-O0-NEXT: v_mov_b32_e32 v4, s11
+; GFX9-O0-NEXT: v_cndmask_b32_e64 v4, v1, v4, s[8:9]
+; GFX9-O0-NEXT: v_mov_b32_e32 v1, s10
+; GFX9-O0-NEXT: v_cndmask_b32_e64 v0, v0, v1, s[8:9]
+; GFX9-O0-NEXT: ; implicit-def: $sgpr8
+; GFX9-O0-NEXT: ; implicit-def: $sgpr8
+; GFX9-O0-NEXT: ; kill: def $vgpr0 killed $vgpr0 def $vgpr0_vgpr1 killed $exec
+; GFX9-O0-NEXT: v_mov_b32_e32 v1, v4
; GFX9-O0-NEXT: buffer_store_dword v2, off, s[0:3], s32 offset:12 ; 4-byte Folded Spill
; GFX9-O0-NEXT: s_nop 0
; GFX9-O0-NEXT: buffer_store_dword v3, off, s[0:3], s32 offset:16 ; 4-byte Folded Spill
; GFX9-O0-NEXT: buffer_store_dword v0, off, s[0:3], s32 offset:4 ; 4-byte Folded Spill
; GFX9-O0-NEXT: s_nop 0
; GFX9-O0-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:8 ; 4-byte Folded Spill
+; GFX9-O0-NEXT: s_and_b64 s[6:7], s[4:5], s[6:7]
; GFX9-O0-NEXT: s_mov_b64 s[4:5], exec
-; GFX9-O0-NEXT: v_writelane_b32 v30, s4, 4
-; GFX9-O0-NEXT: v_writelane_b32 v30, s5, 5
+; GFX9-O0-NEXT: v_writelane_b32 v29, s4, 4
+; GFX9-O0-NEXT: v_writelane_b32 v29, s5, 5
; GFX9-O0-NEXT: s_or_saveexec_b64 s[22:23], -1
-; GFX9-O0-NEXT: buffer_store_dword v30, off, s[0:3], s32 ; 4-byte Folded Spill
+; GFX9-O0-NEXT: buffer_store_dword v29, off, s[0:3], s32 ; 4-byte Folded Spill
; GFX9-O0-NEXT: s_mov_b64 exec, s[22:23]
; GFX9-O0-NEXT: s_and_b64 s[4:5], s[4:5], s[6:7]
; GFX9-O0-NEXT: s_mov_b64 exec, s[4:5]
@@ -548,67 +552,64 @@ define i128 @v_sdiv_i128_vv(i128 %lhs, i128 %rhs) {
; GFX9-O0-NEXT: s_branch .LBB0_8
; GFX9-O0-NEXT: .LBB0_1: ; %Flow
; GFX9-O0-NEXT: s_or_saveexec_b64 s[22:23], -1
-; GFX9-O0-NEXT: buffer_load_dword v30, off, s[0:3], s32 ; 4-byte Folded Reload
+; GFX9-O0-NEXT: buffer_load_dword v29, off, s[0:3], s32 ; 4-byte Folded Reload
; GFX9-O0-NEXT: s_mov_b64 exec, s[22:23]
; GFX9-O0-NEXT: s_waitcnt vmcnt(0)
-; GFX9-O0-NEXT: v_readlane_b32 s4, v30, 6
-; GFX9-O0-NEXT: v_readlane_b32 s5, v30, 7
+; GFX9-O0-NEXT: v_readlane_b32 s4, v29, 6
+; GFX9-O0-NEXT: v_readlane_b32 s5, v29, 7
; GFX9-O0-NEXT: s_or_b64 exec, exec, s[4:5]
; GFX9-O0-NEXT: ; %bb.2: ; %Flow
-; GFX9-O0-NEXT: buffer_load_dword v6, off, s[0:3], s32 offset:120 ; 4-byte Folded Reload
-; GFX9-O0-NEXT: buffer_load_dword v7, off, s[0:3], s32 offset:124 ; 4-byte Folded Reload
-; GFX9-O0-NEXT: buffer_load_dword v4, off, s[0:3], s32 offset:128 ; 4-byte Folded Reload
-; GFX9-O0-NEXT: buffer_load_dword v5, off, s[0:3], s32 offset:132 ; 4-byte Folded Reload
-; GFX9-O0-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:136 ; 4-byte Folded Reload
-; GFX9-O0-NEXT: buffer_load_dword v3, off, s[0:3], s32 offset:140 ; 4-byte Folded Reload
-; GFX9-O0-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:144 ; 4-byte Folded Reload
-; GFX9-O0-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:148 ; 4-byte Folded Reload
-; GFX9-O0-NEXT: s_waitcnt vmcnt(7)
-; GFX9-O0-NEXT: buffer_store_dword v6, off, s[0:3], s32 offset:112 ; 4-byte Folded Spill
-; GFX9-O0-NEXT: s_waitcnt vmcnt(7)
-; GFX9-O0-NEXT: buffer_store_dword v7, off, s[0:3], s32 offset:116 ; 4-byte Folded Spill
-; GFX9-O0-NEXT: s_waitcnt vmcnt(7)
-; GFX9-O0-NEXT: buffer_store_dword v4, off, s[0:3], s32 offset:104 ; 4-byte Folded Spill
-; GFX9-O0-NEXT: s_waitcnt vmcnt(7)
-; GFX9-O0-NEXT: buffer_store_dword v5, off, s[0:3], s32 offset:108 ; 4-byte Folded Spill
-; GFX9-O0-NEXT: s_waitcnt vmcnt(7)
-; GFX9-O0-NEXT: buffer_store_dword v2, off, s[0:3], s32 offset:96 ; 4-byte Folded Spill
-; GFX9-O0-NEXT: s_waitcnt vmcnt(7)
-; GFX9-O0-NEXT: buffer_store_dword v3, off, s[0:3], s32 offset:100 ; 4-byte Folded Spill
-; GFX9-O0-NEXT: s_waitcnt vmcnt(7)
-; GFX9-O0-NEXT: buffer_store_dword v0, off, s[0:3], s32 offset:88 ; 4-byte Folded Spill
-; GFX9-O0-NEXT: s_waitcnt vmcnt(7)
-; GFX9-O0-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:92 ; 4-byte Folded Spill
+; GFX9-O0-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:124 ; 4-byte Folded Reload
+; GFX9-O0-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:128 ; 4-byte Folded Reload
+; GFX9-O0-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:132 ; 4-byte Folded Reload
+; GFX9-O0-NEXT: buffer_load_dword v3, off, s[0:3], s32 offset:136 ; 4-byte Folded Reload
+; GFX9-O0-NEXT: buffer_load_dword v4, off, s[0:3], s32 offset:140 ; 4-byte Folded Reload
+; GFX9-O0-NEXT: buffer_load_dword v5, off, s[0:3], s32 offset:144 ; 4-byte Folded Reload
+; GFX9-O0-NEXT: buffer_load_dword v6, off, s[0:3], s32 offset:148 ; 4-byte Folded Reload
+; GFX9-O0-NEXT: buffer_load_dword v7, off, s[0:3], s32 offset:152 ; 4-byte Folded Reload
+; GFX9-O0-NEXT: s_waitcnt vmcnt(1)
+; GFX9-O0-NEXT: buffer_store_dword v6, off, s[0:3], s32 offset:116 ; 4-byte Folded Spill
+; GFX9-O0-NEXT: s_waitcnt vmcnt(1)
+; GFX9-O0-NEXT: buffer_store_dword v7, off, s[0:3], s32 offset:120 ; 4-byte Folded Spill
+; GFX9-O0-NEXT: buffer_store_dword v4, off, s[0:3], s32 offset:108 ; 4-byte Folded Spill
+; GFX9-O0-NEXT: s_nop 0
+; GFX9-O0-NEXT: buffer_store_dword v5, off, s[0:3], s32 offset:112 ; 4-byte Folded Spill
+; GFX9-O0-NEXT: buffer_store_dword v2, off, s[0:3], s32 offset:100 ; 4-byte Folded Spill
+; GFX9-O0-NEXT: s_nop 0
+; GFX9-O0-NEXT: buffer_store_dword v3, off, s[0:3], s32 offset:104 ; 4-byte Folded Spill
+; GFX9-O0-NEXT: buffer_store_dword v0, off, s[0:3], s32 offset:92 ; 4-byte Folded Spill
+; GFX9-O0-NEXT: s_nop 0
+; GFX9-O0-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:96 ; 4-byte Folded Spill
; GFX9-O0-NEXT: s_branch .LBB0_5
; GFX9-O0-NEXT: .LBB0_3: ; %Flow2
; GFX9-O0-NEXT: s_or_saveexec_b64 s[22:23], -1
-; GFX9-O0-NEXT: buffer_load_dword v30, off, s[0:3], s32 ; 4-byte Folded Reload
+; GFX9-O0-NEXT: buffer_load_dword v29, off, s[0:3], s32 ; 4-byte Folded Reload
; GFX9-O0-NEXT: s_mov_b64 exec, s[22:23]
; GFX9-O0-NEXT: s_waitcnt vmcnt(0)
-; GFX9-O0-NEXT: v_readlane_b32 s4, v30, 4
-; GFX9-O0-NEXT: v_readlane_b32 s5, v30, 5
+; GFX9-O0-NEXT: v_readlane_b32 s4, v29, 4
+; GFX9-O0-NEXT: v_readlane_b32 s5, v29, 5
; GFX9-O0-NEXT: s_or_b64 exec, exec, s[4:5]
-; GFX9-O0-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:12 ; 4-byte Folded Reload
-; GFX9-O0-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:16 ; 4-byte Folded Reload
-; GFX9-O0-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:4 ; 4-byte Folded Reload
-; GFX9-O0-NEXT: buffer_load_dword v3, off, s[0:3], s32 offset:8 ; 4-byte Folded Reload
+; GFX9-O0-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:4 ; 4-byte Folded Reload
+; GFX9-O0-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:8 ; 4-byte Folded Reload
+; GFX9-O0-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:12 ; 4-byte Folded Reload
+; GFX9-O0-NEXT: buffer_load_dword v3, off, s[0:3], s32 offset:16 ; 4-byte Folded Reload
; GFX9-O0-NEXT: s_waitcnt vmcnt(1)
-; GFX9-O0-NEXT: buffer_store_dword v2, off, s[0:3], s32 offset:160 ; 4-byte Folded Spill
+; GFX9-O0-NEXT: buffer_store_dword v2, off, s[0:3], s32 offset:164 ; 4-byte Folded Spill
; GFX9-O0-NEXT: s_waitcnt vmcnt(1)
-; GFX9-O0-NEXT: buffer_store_dword v3, off, s[0:3], s32 offset:164 ; 4-byte Folded Spill
-; GFX9-O0-NEXT: buffer_store_dword v0, off, s[0:3], s32 offset:152 ; 4-byte Folded Spill
+; GFX9-O0-NEXT: buffer_store_dword v3, off, s[0:3], s32 offset:168 ; 4-byte Folded Spill
+; GFX9-O0-NEXT: buffer_store_dword v0, off, s[0:3], s32 offset:156 ; 4-byte Folded Spill
; GFX9-O0-NEXT: s_nop 0
-; GFX9-O0-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:156 ; 4-byte Folded Spill
+; GFX9-O0-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:160 ; 4-byte Folded Spill
; GFX9-O0-NEXT: s_branch .LBB0_9
; GFX9-O0-NEXT: .LBB0_4: ; %udiv-loop-exit
-; GFX9-O0-NEXT: buffer_load_dword v5, off, s[0:3], s32 offset:168 ; 4-byte Folded Reload
-; GFX9-O0-NEXT: buffer_load_dword v6, off, s[0:3], s32 offset:172 ; 4-byte Folded Reload
-; GFX9-O0-NEXT: buffer_load_dword v7, off, s[0:3], s32 offset:176 ; 4-byte Folded Reload
-; GFX9-O0-NEXT: buffer_load_dword v8, off, s[0:3], s32 offset:180 ; 4-byte Folded Reload
-; GFX9-O0-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:184 ; 4-byte Folded Reload
-; GFX9-O0-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:188 ; 4-byte Folded Reload
-; GFX9-O0-NEXT: buffer_load_dword v9, off, s[0:3], s32 offset:192 ; 4-byte Folded Reload
-; GFX9-O0-NEXT: buffer_load_dword v10, off, s[0:3], s32 offset:196 ; 4-byte Folded Reload
+; GFX9-O0-NEXT: buffer_load_dword v5, off, s[0:3], s32 offset:172 ; 4-byte Folded Reload
+; GFX9-O0-NEXT: buffer_load_dword v6, off, s[0:3], s32 offset:176 ; 4-byte Folded Reload
+; GFX9-O0-NEXT: buffer_load_dword v7, off, s[0:3], s32 offset:180 ; 4-byte Folded Reload
+; GFX9-O0-NEXT: buffer_load_dword v8, off, s[0:3], s32 offset:184 ; 4-byte Folded Reload
+; GFX9-O0-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:188 ; 4-byte Folded Reload
+; GFX9-O0-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:192 ; 4-byte Folded Reload
+; GFX9-O0-NEXT: buffer_load_dword v9, off, s[0:3], s32 offset:196 ; 4-byte Folded Reload
+; GFX9-O0-NEXT: buffer_load_dword v10, off, s[0:3], s32 offset:200 ; 4-byte Folded Reload
; GFX9-O0-NEXT: s_mov_b32 s4, 1
; GFX9-O0-NEXT: s_waitcnt vmcnt(2)
; GFX9-O0-NEXT: v_lshlrev_b64 v[2:3], s4, v[0:1]
@@ -633,408 +634,408 @@ define i128 @v_sdiv_i128_vv(i128 %lhs, i128 %rhs) {
; GFX9-O0-NEXT: v_or_b32_e64 v2, v2, v3
; GFX9-O0-NEXT: ; kill: def $vgpr2 killed $vgpr2 def $vgpr2_vgpr3 killed $exec
; GFX9-O0-NEXT: v_mov_b32_e32 v3, v4
-; GFX9-O0-NEXT: buffer_store_dword v2, off, s[0:3], s32 offset:12 ; 4-byte Folded Spill
+; GFX9-O0-NEXT: buffer_store_dword v2, off, s[0:3], s32 offset:4 ; 4-byte Folded Spill
; GFX9-O0-NEXT: s_nop 0
-; GFX9-O0-NEXT: buffer_store_dword v3, off, s[0:3], s32 offset:16 ; 4-byte Folded Spill
-; GFX9-O0-NEXT: buffer_store_dword v0, off, s[0:3], s32 offset:4 ; 4-byte Folded Spill
+; GFX9-O0-NEXT: buffer_store_dword v3, off, s[0:3], s32 offset:8 ; 4-byte Folded Spill
+; GFX9-O0-NEXT: buffer_store_dword v0, off, s[0:3], s32 offset:12 ; 4-byte Folded Spill
; GFX9-O0-NEXT: s_nop 0
-; GFX9-O0-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:8 ; 4-byte Folded Spill
+; GFX9-O0-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:16 ; 4-byte Folded Spill
; GFX9-O0-NEXT: s_branch .LBB0_3
; GFX9-O0-NEXT: .LBB0_5: ; %Flow1
; GFX9-O0-NEXT: s_or_saveexec_b64 s[22:23], -1
-; GFX9-O0-NEXT: buffer_load_dword v30, off, s[0:3], s32 ; 4-byte Folded Reload
+; GFX9-O0-NEXT: buffer_load_dword v29, off, s[0:3], s32 ; 4-byte Folded Reload
; GFX9-O0-NEXT: s_mov_b64 exec, s[22:23]
; GFX9-O0-NEXT: s_waitcnt vmcnt(0)
-; GFX9-O0-NEXT: v_readlane_b32 s4, v30, 8
-; GFX9-O0-NEXT: v_readlane_b32 s5, v30, 9
+; GFX9-O0-NEXT: v_readlane_b32 s4, v29, 8
+; GFX9-O0-NEXT: v_readlane_b32 s5, v29, 9
; GFX9-O0-NEXT: s_or_b64 exec, exec, s[4:5]
-; GFX9-O0-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:112 ; 4-byte Folded Reload
-; GFX9-O0-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:116 ; 4-byte Folded Reload
-; GFX9-O0-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:104 ; 4-byte Folded Reload
-; GFX9-O0-NEXT: buffer_load_dword v3, off, s[0:3], s32 offset:108 ; 4-byte Folded Reload
-; GFX9-O0-NEXT: buffer_load_dword v4, off, s[0:3], s32 offset:96 ; 4-byte Folded Reload
-; GFX9-O0-NEXT: buffer_load_dword v5, off, s[0:3], s32 offset:100 ; 4-byte Folded Reload
-; GFX9-O0-NEXT: buffer_load_dword v6, off, s[0:3], s32 offset:88 ; 4-byte Folded Reload
-; GFX9-O0-NEXT: buffer_load_dword v7, off, s[0:3], s32 offset:92 ; 4-byte Folded Reload
+; GFX9-O0-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:92 ; 4-byte Folded Reload
+; GFX9-O0-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:96 ; 4-byte Folded Reload
+; GFX9-O0-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:100 ; 4-byte Folded Reload
+; GFX9-O0-NEXT: buffer_load_dword v3, off, s[0:3], s32 offset:104 ; 4-byte Folded Reload
+; GFX9-O0-NEXT: buffer_load_dword v4, off, s[0:3], s32 offset:108 ; 4-byte Folded Reload
+; GFX9-O0-NEXT: buffer_load_dword v5, off, s[0:3], s32 offset:112 ; 4-byte Folded Reload
+; GFX9-O0-NEXT: buffer_load_dword v6, off, s[0:3], s32 offset:116 ; 4-byte Folded Reload
+; GFX9-O0-NEXT: buffer_load_dword v7, off, s[0:3], s32 offset:120 ; 4-byte Folded Reload
; GFX9-O0-NEXT: s_waitcnt vmcnt(1)
-; GFX9-O0-NEXT: buffer_store_dword v6, off, s[0:3], s32 offset:176 ; 4-byte Folded Spill
+; GFX9-O0-NEXT: buffer_store_dword v6, off, s[0:3], s32 offset:180 ; 4-byte Folded Spill
; GFX9-O0-NEXT: s_waitcnt vmcnt(1)
-; GFX9-O0-NEXT: buffer_store_dword v7, off, s[0:3], s32 offset:180 ; 4-byte Folded Spill
-; GFX9-O0-NEXT: buffer_store_dword v4, off, s[0:3], s32 offset:168 ; 4-byte Folded Spill
+; GFX9-O0-NEXT: buffer_store_dword v7, off, s[0:3], s32 offset:184 ; 4-byte Folded Spill
+; GFX9-O0-NEXT: buffer_store_dword v4, off, s[0:3], s32 offset:172 ; 4-byte Folded Spill
; GFX9-O0-NEXT: s_nop 0
-; GFX9-O0-NEXT: buffer_store_dword v5, off, s[0:3], s32 offset:172 ; 4-byte Folded Spill
-; GFX9-O0-NEXT: buffer_store_dword v2, off, s[0:3], s32 offset:192 ; 4-byte Folded Spill
+; GFX9-O0-NEXT: buffer_store_dword v5, off, s[0:3], s32 offset:176 ; 4-byte Folded Spill
+; GFX9-O0-NEXT: buffer_store_dword v2, off, s[0:3], s32 offset:196 ; 4-byte Folded Spill
; GFX9-O0-NEXT: s_nop 0
-; GFX9-O0-NEXT: buffer_store_dword v3, off, s[0:3], s32 offset:196 ; 4-byte Folded Spill
-; GFX9-O0-NEXT: buffer_store_dword v0, off, s[0:3], s32 offset:184 ; 4-byte Folded Spill
+; GFX9-O0-NEXT: buffer_store_dword v3, off, s[0:3], s32 offset:200 ; 4-byte Folded Spill
+; GFX9-O0-NEXT: buffer_store_dword v0, off, s[0:3], s32 offset:188 ; 4-byte Folded Spill
; GFX9-O0-NEXT: s_nop 0
-; GFX9-O0-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:188 ; 4-byte Folded Spill
+; GFX9-O0-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:192 ; 4-byte Folded Spill
; GFX9-O0-NEXT: s_branch .LBB0_4
; GFX9-O0-NEXT: .LBB0_6: ; %udiv-do-while
; GFX9-O0-NEXT: ; =>This Inner Loop Header: Depth=1
; GFX9-O0-NEXT: s_or_saveexec_b64 s[22:23], -1
-; GFX9-O0-NEXT: buffer_load_dword v30, off, s[0:3], s32 ; 4-byte Folded Reload
+; GFX9-O0-NEXT: buffer_load_dword v29, off, s[0:3], s32 ; 4-byte Folded Reload
; GFX9-O0-NEXT: s_mov_b64 exec, s[22:23]
; GFX9-O0-NEXT: s_waitcnt vmcnt(0)
-; GFX9-O0-NEXT: v_readlane_b32 s6, v30, 10
-; GFX9-O0-NEXT: v_readlane_b32 s7, v30, 11
-; GFX9-O0-NEXT: buffer_load_dword v6, off, s[0:3], s32 offset:200 ; 4-byte Folded Reload
-; GFX9-O0-NEXT: buffer_load_dword v7, off, s[0:3], s32 offset:204 ; 4-byte Folded Reload
-; GFX9-O0-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:208 ; 4-byte Folded Reload
-; GFX9-O0-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:212 ; 4-byte Folded Reload
-; GFX9-O0-NEXT: buffer_load_dword v22, off, s[0:3], s32 offset:216 ; 4-byte Folded Reload
-; GFX9-O0-NEXT: buffer_load_dword v23, off, s[0:3], s32 offset:220 ; 4-byte Folded Reload
-; GFX9-O0-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:224 ; 4-byte Folded Reload
-; GFX9-O0-NEXT: buffer_load_dword v3, off, s[0:3], s32 offset:228 ; 4-byte Folded Reload
-; GFX9-O0-NEXT: buffer_load_dword v16, off, s[0:3], s32 offset:232 ; 4-byte Folded Reload
-; GFX9-O0-NEXT: buffer_load_dword v17, off, s[0:3], s32 offset:236 ; 4-byte Folded Reload
-; GFX9-O0-NEXT: buffer_load_dword v8, off, s[0:3], s32 offset:240 ; 4-byte Folded Reload
-; GFX9-O0-NEXT: buffer_load_dword v9, off, s[0:3], s32 offset:244 ; 4-byte Folded Reload
-; GFX9-O0-NEXT: buffer_load_dword v26, off, s[0:3], s32 offset:248 ; 4-byte Folded Reload
-; GFX9-O0-NEXT: buffer_load_dword v27, off, s[0:3], s32 offset:252 ; 4-byte Folded Reload
-; GFX9-O0-NEXT: buffer_load_dword v24, off, s[0:3], s32 offset:256 ; 4-byte Folded Reload
-; GFX9-O0-NEXT: buffer_load_dword v25, off, s[0:3], s32 offset:260 ; 4-byte Folded Reload
+; GFX9-O0-NEXT: v_readlane_b32 s6, v29, 10
+; GFX9-O0-NEXT: v_readlane_b32 s7, v29, 11
+; GFX9-O0-NEXT: buffer_load_dword v10, off, s[0:3], s32 offset:204 ; 4-byte Folded Reload
+; GFX9-O0-NEXT: buffer_load_dword v11, off, s[0:3], s32 offset:208 ; 4-byte Folded Reload
+; GFX9-O0-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:212 ; 4-byte Folded Reload
+; GFX9-O0-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:216 ; 4-byte Folded Reload
+; GFX9-O0-NEXT: buffer_load_dword v24, off, s[0:3], s32 offset:220 ; 4-byte Folded Reload
+; GFX9-O0-NEXT: buffer_load_dword v25, off, s[0:3], s32 offset:224 ; 4-byte Folded Reload
+; GFX9-O0-NEXT: buffer_load_dword v16, off, s[0:3], s32 offset:228 ; 4-byte Folded Reload
+; GFX9-O0-NEXT: buffer_load_dword v17, off, s[0:3], s32 offset:232 ; 4-byte Folded Reload
+; GFX9-O0-NEXT: buffer_load_dword v8, off, s[0:3], s32 offset:236 ; 4-byte Folded Reload
+; GFX9-O0-NEXT: buffer_load_dword v9, off, s[0:3], s32 offset:240 ; 4-byte Folded Reload
+; GFX9-O0-NEXT: buffer_load_dword v12, off, s[0:3], s32 offset:244 ; 4-byte Folded Reload
+; GFX9-O0-NEXT: buffer_load_dword v13, off, s[0:3], s32 offset:248 ; 4-byte Folded Reload
+; GFX9-O0-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:252 ; 4-byte Folded Reload
+; GFX9-O0-NEXT: buffer_load_dword v3, off, s[0:3], s32 offset:256 ; 4-byte Folded Reload
+; GFX9-O0-NEXT: buffer_load_dword v26, off, s[0:3], s32 offset:260 ; 4-byte Folded Reload
+; GFX9-O0-NEXT: buffer_load_dword v27, off, s[0:3], s32 offset:264 ; 4-byte Folded Reload
; GFX9-O0-NEXT: buffer_load_dword v18, off, s[0:3], s32 offset:52 ; 4-byte Folded Reload
; GFX9-O0-NEXT: buffer_load_dword v19, off, s[0:3], s32 offset:56 ; 4-byte Folded Reload
; GFX9-O0-NEXT: buffer_load_dword v20, off, s[0:3], s32 offset:60 ; 4-byte Folded Reload
; GFX9-O0-NEXT: buffer_load_dword v21, off, s[0:3], s32 offset:64 ; 4-byte Folded Reload
-; GFX9-O0-NEXT: buffer_load_dword v14, off, s[0:3], s32 offset:264 ; 4-byte Folded Reload
-; GFX9-O0-NEXT: buffer_load_dword v15, off, s[0:3], s32 offset:268 ; 4-byte Folded Reload
-; GFX9-O0-NEXT: buffer_load_dword v11, off, s[0:3], s32 offset:272 ; 4-byte Folded Reload
-; GFX9-O0-NEXT: buffer_load_dword v12, off, s[0:3], s32 offset:276 ; 4-byte Folded Reload
+; GFX9-O0-NEXT: buffer_load_dword v22, off, s[0:3], s32 offset:268 ; 4-byte Folded Reload
+; GFX9-O0-NEXT: buffer_load_dword v23, off, s[0:3], s32 offset:272 ; 4-byte Folded Reload
+; GFX9-O0-NEXT: buffer_load_dword v4, off, s[0:3], s32 offset:276 ; 4-byte Folded Reload
+; GFX9-O0-NEXT: buffer_load_dword v5, off, s[0:3], s32 offset:280 ; 4-byte Folded Reload
; GFX9-O0-NEXT: s_mov_b32 s4, 63
-; GFX9-O0-NEXT: s_waitcnt vmcnt(16)
-; GFX9-O0-NEXT: v_lshrrev_b64 v[28:29], s4, v[2:3]
-; GFX9-O0-NEXT: v_mov_b32_e32 v5, v29
+; GFX9-O0-NEXT: s_waitcnt vmcnt(10)
+; GFX9-O0-NEXT: v_lshrrev_b64 v[6:7], s4, v[2:3]
+; GFX9-O0-NEXT: v_mov_b32_e32 v15, v7
; GFX9-O0-NEXT: s_mov_b32 s5, 1
-; GFX9-O0-NEXT: v_lshlrev_b64 v[22:23], s5, v[22:23]
-; GFX9-O0-NEXT: v_mov_b32_e32 v4, v23
-; GFX9-O0-NEXT: v_or_b32_e64 v4, v4, v5
-; GFX9-O0-NEXT: v_mov_b32_e32 v10, v28
-; GFX9-O0-NEXT: v_mov_b32_e32 v5, v22
-; GFX9-O0-NEXT: v_or_b32_e64 v22, v5, v10
-; GFX9-O0-NEXT: ; kill: def $vgpr22 killed $vgpr22 def $vgpr22_vgpr23 killed $exec
-; GFX9-O0-NEXT: v_mov_b32_e32 v23, v4
-; GFX9-O0-NEXT: v_lshlrev_b64 v[28:29], s5, v[2:3]
-; GFX9-O0-NEXT: v_lshrrev_b64 v[4:5], s4, v[6:7]
-; GFX9-O0-NEXT: v_mov_b32_e32 v2, v29
-; GFX9-O0-NEXT: v_mov_b32_e32 v3, v5
+; GFX9-O0-NEXT: s_waitcnt vmcnt(8)
+; GFX9-O0-NEXT: v_lshlrev_b64 v[26:27], s5, v[26:27]
+; GFX9-O0-NEXT: v_mov_b32_e32 v14, v27
+; GFX9-O0-NEXT: v_or_b32_e64 v14, v14, v15
+; GFX9-O0-NEXT: v_mov_b32_e32 v7, v6
+; GFX9-O0-NEXT: v_mov_b32_e32 v6, v26
+; GFX9-O0-NEXT: v_or_b32_e64 v6, v6, v7
+; GFX9-O0-NEXT: ; kill: def $vgpr6 killed $vgpr6 def $vgpr6_vgpr7 killed $exec
+; GFX9-O0-NEXT: v_mov_b32_e32 v7, v14
+; GFX9-O0-NEXT: v_lshlrev_b64 v[26:27], s5, v[2:3]
+; GFX9-O0-NEXT: v_lshrrev_b64 v[14:15], s4, v[10:11]
+; GFX9-O0-NEXT: v_mov_b32_e32 v2, v27
+; GFX9-O0-NEXT: v_mov_b32_e32 v3, v15
; GFX9-O0-NEXT: v_or_b32_e64 v2, v2, v3
-; GFX9-O0-NEXT: v_mov_b32_e32 v3, v28
-; GFX9-O0-NEXT: ; kill: def $vgpr4 killed $vgpr4 killed $vgpr4_vgpr5 killed $exec
-; GFX9-O0-NEXT: v_or_b32_e64 v4, v3, v4
-; GFX9-O0-NEXT: ; kill: def $vgpr4 killed $vgpr4 def $vgpr4_vgpr5 killed $exec
-; GFX9-O0-NEXT: v_mov_b32_e32 v5, v2
+; GFX9-O0-NEXT: v_mov_b32_e32 v3, v26
+; GFX9-O0-NEXT: ; kill: def $vgpr14 killed $vgpr14 killed $vgpr14_vgpr15 killed $exec
+; GFX9-O0-NEXT: v_or_b32_e64 v14, v3, v14
+; GFX9-O0-NEXT: ; kill: def $vgpr14 killed $vgpr14 def $vgpr14_vgpr15 killed $exec
+; GFX9-O0-NEXT: v_mov_b32_e32 v15, v2
; GFX9-O0-NEXT: v_lshlrev_b64 v[2:3], s5, v[0:1]
-; GFX9-O0-NEXT: v_lshlrev_b64 v[28:29], s5, v[6:7]
+; GFX9-O0-NEXT: v_lshlrev_b64 v[26:27], s5, v[10:11]
; GFX9-O0-NEXT: v_lshrrev_b64 v[0:1], s4, v[0:1]
-; GFX9-O0-NEXT: v_mov_b32_e32 v7, v1
-; GFX9-O0-NEXT: v_mov_b32_e32 v6, v29
-; GFX9-O0-NEXT: s_waitcnt vmcnt(10)
+; GFX9-O0-NEXT: v_mov_b32_e32 v11, v1
; GFX9-O0-NEXT: v_mov_b32_e32 v10, v27
-; GFX9-O0-NEXT: v_or3_b32 v6, v6, v7, v10
+; GFX9-O0-NEXT: v_mov_b32_e32 v28, v25
+; GFX9-O0-NEXT: v_or3_b32 v10, v10, v11, v28
; GFX9-O0-NEXT: v_mov_b32_e32 v1, v0
-; GFX9-O0-NEXT: v_mov_b32_e32 v0, v28
-; GFX9-O0-NEXT: v_mov_b32_e32 v7, v26
-; GFX9-O0-NEXT: v_or3_b32 v0, v0, v1, v7
+; GFX9-O0-NEXT: v_mov_b32_e32 v0, v26
+; GFX9-O0-NEXT: v_mov_b32_e32 v11, v24
+; GFX9-O0-NEXT: v_or3_b32 v0, v0, v1, v11
; GFX9-O0-NEXT: ; kill: def $vgpr0 killed $vgpr0 def $vgpr0_vgpr1 killed $exec
-; GFX9-O0-NEXT: v_mov_b32_e32 v1, v6
-; GFX9-O0-NEXT: v_mov_b32_e32 v7, v3
-; GFX9-O0-NEXT: s_waitcnt vmcnt(8)
-; GFX9-O0-NEXT: v_mov_b32_e32 v6, v25
-; GFX9-O0-NEXT: v_or_b32_e64 v6, v6, v7
+; GFX9-O0-NEXT: v_mov_b32_e32 v1, v10
+; GFX9-O0-NEXT: v_mov_b32_e32 v11, v3
+; GFX9-O0-NEXT: v_mov_b32_e32 v10, v17
+; GFX9-O0-NEXT: v_or_b32_e64 v10, v10, v11
; GFX9-O0-NEXT: v_mov_b32_e32 v3, v2
-; GFX9-O0-NEXT: v_mov_b32_e32 v2, v24
+; GFX9-O0-NEXT: v_mov_b32_e32 v2, v16
; GFX9-O0-NEXT: v_or_b32_e64 v2, v2, v3
; GFX9-O0-NEXT: ; kill: def $vgpr2 killed $vgpr2 def $vgpr2_vgpr3 killed $exec
-; GFX9-O0-NEXT: v_mov_b32_e32 v3, v6
-; GFX9-O0-NEXT: v_mov_b32_e32 v6, v4
-; GFX9-O0-NEXT: v_mov_b32_e32 v10, v5
-; GFX9-O0-NEXT: v_mov_b32_e32 v4, v22
-; GFX9-O0-NEXT: v_mov_b32_e32 v5, v23
+; GFX9-O0-NEXT: v_mov_b32_e32 v3, v10
+; GFX9-O0-NEXT: v_mov_b32_e32 v10, v14
+; GFX9-O0-NEXT: v_mov_b32_e32 v16, v15
+; GFX9-O0-NEXT: v_mov_b32_e32 v14, v6
+; GFX9-O0-NEXT: v_mov_b32_e32 v11, v7
; GFX9-O0-NEXT: s_waitcnt vmcnt(1)
-; GFX9-O0-NEXT: v_mov_b32_e32 v13, v11
-; GFX9-O0-NEXT: v_mov_b32_e32 v11, v14
-; GFX9-O0-NEXT: v_mov_b32_e32 v7, v15
-; GFX9-O0-NEXT: v_sub_co_u32_e32 v13, vcc, v13, v6
+; GFX9-O0-NEXT: v_mov_b32_e32 v7, v4
; GFX9-O0-NEXT: s_waitcnt vmcnt(0)
-; GFX9-O0-NEXT: v_subb_co_u32_e32 v12, vcc, v12, v10, vcc
-; GFX9-O0-NEXT: v_subb_co_u32_e32 v11, vcc, v11, v4, vcc
-; GFX9-O0-NEXT: v_subb_co_u32_e32 v7, vcc, v7, v5, vcc
+; GFX9-O0-NEXT: v_mov_b32_e32 v6, v5
+; GFX9-O0-NEXT: v_mov_b32_e32 v4, v22
+; GFX9-O0-NEXT: v_mov_b32_e32 v5, v23
+; GFX9-O0-NEXT: v_sub_co_u32_e32 v7, vcc, v7, v10
+; GFX9-O0-NEXT: v_subb_co_u32_e32 v6, vcc, v6, v16, vcc
+; GFX9-O0-NEXT: v_subb_co_u32_e32 v4, vcc, v4, v14, vcc
+; GFX9-O0-NEXT: v_subb_co_u32_e32 v6, vcc, v5, v11, vcc
; GFX9-O0-NEXT: ; implicit-def: $sgpr5
; GFX9-O0-NEXT: ; implicit-def: $sgpr5
-; GFX9-O0-NEXT: ; kill: def $vgpr11 killed $vgpr11 def $vgpr11_vgpr12 killed $exec
-; GFX9-O0-NEXT: v_mov_b32_e32 v12, v7
-; GFX9-O0-NEXT: v_ashrrev_i64 v[13:14], s4, v[11:12]
-; GFX9-O0-NEXT: v_mov_b32_e32 v7, v14
+; GFX9-O0-NEXT: ; kill: def $vgpr4 killed $vgpr4 def $vgpr4_vgpr5 killed $exec
+; GFX9-O0-NEXT: v_mov_b32_e32 v5, v6
+; GFX9-O0-NEXT: v_ashrrev_i64 v[5:6], s4, v[4:5]
+; GFX9-O0-NEXT: v_mov_b32_e32 v15, v6
; GFX9-O0-NEXT: s_mov_b64 s[4:5], 1
; GFX9-O0-NEXT: s_mov_b32 s8, s5
-; GFX9-O0-NEXT: v_and_b32_e64 v12, v7, s8
-; GFX9-O0-NEXT: v_mov_b32_e32 v11, v13
+; GFX9-O0-NEXT: v_and_b32_e64 v4, v15, s8
+; GFX9-O0-NEXT: v_mov_b32_e32 v17, v5
; GFX9-O0-NEXT: ; kill: def $sgpr4 killed $sgpr4 killed $sgpr4_sgpr5
-; GFX9-O0-NEXT: v_and_b32_e64 v14, v11, s4
-; GFX9-O0-NEXT: ; kill: def $vgpr14 killed $vgpr14 def $vgpr14_vgpr15 killed $exec
-; GFX9-O0-NEXT: v_mov_b32_e32 v15, v12
-; GFX9-O0-NEXT: v_mov_b32_e32 v12, 0
-; GFX9-O0-NEXT: v_mov_b32_e32 v13, 0
+; GFX9-O0-NEXT: v_and_b32_e64 v6, v17, s4
+; GFX9-O0-NEXT: ; kill: def $vgpr6 killed $vgpr6 def $vgpr6_vgpr7 killed $exec
+; GFX9-O0-NEXT: v_mov_b32_e32 v7, v4
+; GFX9-O0-NEXT: v_mov_b32_e32 v4, 0
+; GFX9-O0-NEXT: v_mov_b32_e32 v5, 0
; GFX9-O0-NEXT: v_mov_b32_e32 v22, v21
-; GFX9-O0-NEXT: v_and_b32_e64 v22, v7, v22
-; GFX9-O0-NEXT: v_and_b32_e64 v20, v11, v20
+; GFX9-O0-NEXT: v_and_b32_e64 v22, v15, v22
+; GFX9-O0-NEXT: v_and_b32_e64 v20, v17, v20
; GFX9-O0-NEXT: ; kill: def $vgpr20 killed $vgpr20 def $vgpr20_vgpr21 killed $exec
; GFX9-O0-NEXT: v_mov_b32_e32 v21, v22
; GFX9-O0-NEXT: v_mov_b32_e32 v22, v19
-; GFX9-O0-NEXT: v_and_b32_e64 v7, v7, v22
-; GFX9-O0-NEXT: v_and_b32_e64 v22, v11, v18
-; GFX9-O0-NEXT: ; kill: def $vgpr22 killed $vgpr22 def $vgpr22_vgpr23 killed $exec
-; GFX9-O0-NEXT: v_mov_b32_e32 v23, v7
-; GFX9-O0-NEXT: v_mov_b32_e32 v19, v22
-; GFX9-O0-NEXT: v_mov_b32_e32 v18, v23
-; GFX9-O0-NEXT: v_mov_b32_e32 v11, v20
-; GFX9-O0-NEXT: v_mov_b32_e32 v7, v21
-; GFX9-O0-NEXT: v_sub_co_u32_e32 v6, vcc, v6, v19
-; GFX9-O0-NEXT: v_subb_co_u32_e32 v10, vcc, v10, v18, vcc
-; GFX9-O0-NEXT: v_subb_co_u32_e32 v4, vcc, v4, v11, vcc
-; GFX9-O0-NEXT: v_subb_co_u32_e32 v7, vcc, v5, v7, vcc
+; GFX9-O0-NEXT: v_and_b32_e64 v15, v15, v22
+; GFX9-O0-NEXT: v_and_b32_e64 v17, v17, v18
+; GFX9-O0-NEXT: ; kill: def $vgpr17 killed $vgpr17 def $vgpr17_vgpr18 killed $exec
+; GFX9-O0-NEXT: v_mov_b32_e32 v18, v15
+; GFX9-O0-NEXT: v_mov_b32_e32 v19, v17
+; GFX9-O0-NEXT: ; kill: def $vgpr18 killed $vgpr18 killed $vgpr17_vgpr18 killed $exec
+; GFX9-O0-NEXT: v_mov_b32_e32 v17, v20
+; GFX9-O0-NEXT: v_mov_b32_e32 v15, v21
+; GFX9-O0-NEXT: v_sub_co_u32_e32 v10, vcc, v10, v19
+; GFX9-O0-NEXT: v_subb_co_u32_e32 v16, vcc, v16, v18, vcc
+; GFX9-O0-NEXT: v_subb_co_u32_e32 v14, vcc, v14, v17, vcc
+; GFX9-O0-NEXT: v_subb_co_u32_e32 v11, vcc, v11, v15, vcc
; GFX9-O0-NEXT: ; implicit-def: $sgpr4
; GFX9-O0-NEXT: ; implicit-def: $sgpr4
-; GFX9-O0-NEXT: ; kill: def $vgpr4 killed $vgpr4 def $vgpr4_vgpr5 killed $exec
-; GFX9-O0-NEXT: v_mov_b32_e32 v5, v7
+; GFX9-O0-NEXT: ; kill: def $vgpr14 killed $vgpr14 def $vgpr14_vgpr15 killed $exec
+; GFX9-O0-NEXT: v_mov_b32_e32 v15, v11
; GFX9-O0-NEXT: ; implicit-def: $sgpr4
; GFX9-O0-NEXT: ; implicit-def: $sgpr4
-; GFX9-O0-NEXT: ; kill: def $vgpr6 killed $vgpr6 def $vgpr6_vgpr7 killed $exec
-; GFX9-O0-NEXT: v_mov_b32_e32 v7, v10
+; GFX9-O0-NEXT: ; kill: def $vgpr10 killed $vgpr10 def $vgpr10_vgpr11 killed $exec
+; GFX9-O0-NEXT: v_mov_b32_e32 v11, v16
+; GFX9-O0-NEXT: buffer_store_dword v14, off, s[0:3], s32 offset:260 ; 4-byte Folded Spill
+; GFX9-O0-NEXT: s_nop 0
+; GFX9-O0-NEXT: buffer_store_dword v15, off, s[0:3], s32 offset:264 ; 4-byte Folded Spill
+; GFX9-O0-NEXT: buffer_store_dword v10, off, s[0:3], s32 offset:252 ; 4-byte Folded Spill
+; GFX9-O0-NEXT: s_nop 0
+; GFX9-O0-NEXT: buffer_store_dword v11, off, s[0:3], s32 offset:256 ; 4-byte Folded Spill
; GFX9-O0-NEXT: v_mov_b32_e32 v11, v8
; GFX9-O0-NEXT: ; kill: def $vgpr9 killed $vgpr9 killed $vgpr8_vgpr9 killed $exec
; GFX9-O0-NEXT: s_mov_b64 s[8:9], -1
; GFX9-O0-NEXT: s_mov_b32 s5, s8
; GFX9-O0-NEXT: s_mov_b32 s4, s9
-; GFX9-O0-NEXT: v_mov_b32_e32 v10, v16
-; GFX9-O0-NEXT: v_mov_b32_e32 v8, v17
-; GFX9-O0-NEXT: v_mov_b32_e32 v16, s5
-; GFX9-O0-NEXT: v_add_co_u32_e32 v19, vcc, v11, v16
-; GFX9-O0-NEXT: v_mov_b32_e32 v11, s4
-; GFX9-O0-NEXT: v_addc_co_u32_e32 v9, vcc, v9, v11, vcc
-; GFX9-O0-NEXT: v_mov_b32_e32 v11, s5
-; GFX9-O0-NEXT: v_addc_co_u32_e32 v16, vcc, v10, v11, vcc
-; GFX9-O0-NEXT: v_mov_b32_e32 v10, s4
-; GFX9-O0-NEXT: v_addc_co_u32_e32 v8, vcc, v8, v10, vcc
+; GFX9-O0-NEXT: v_mov_b32_e32 v8, v12
+; GFX9-O0-NEXT: v_mov_b32_e32 v10, v13
+; GFX9-O0-NEXT: v_mov_b32_e32 v12, s5
+; GFX9-O0-NEXT: v_add_co_u32_e32 v11, vcc, v11, v12
+; GFX9-O0-NEXT: v_mov_b32_e32 v12, s4
+; GFX9-O0-NEXT: v_addc_co_u32_e32 v9, vcc, v9, v12, vcc
+; GFX9-O0-NEXT: v_mov_b32_e32 v12, s5
+; GFX9-O0-NEXT: v_addc_co_u32_e32 v8, vcc, v8, v12, vcc
+; GFX9-O0-NEXT: v_mov_b32_e32 v12, s4
+; GFX9-O0-NEXT: v_addc_co_u32_e32 v10, vcc, v10, v12, vcc
; GFX9-O0-NEXT: ; implicit-def: $sgpr4
; GFX9-O0-NEXT: ; implicit-def: $sgpr4
-; GFX9-O0-NEXT: ; kill: def $vgpr19 killed $vgpr19 def $vgpr19_vgpr20 killed $exec
-; GFX9-O0-NEXT: v_mov_b32_e32 v20, v9
+; GFX9-O0-NEXT: ; kill: def $vgpr11 killed $vgpr11 def $vgpr11_vgpr12 killed $exec
+; GFX9-O0-NEXT: v_mov_b32_e32 v12, v9
; GFX9-O0-NEXT: ; implicit-def: $sgpr4
; GFX9-O0-NEXT: ; implicit-def: $sgpr4
-; GFX9-O0-NEXT: ; kill: def $vgpr16 killed $vgpr16 def $vgpr16_vgpr17 killed $exec
-; GFX9-O0-NEXT: v_mov_b32_e32 v17, v8
-; GFX9-O0-NEXT: v_mov_b32_e32 v8, v16
-; GFX9-O0-NEXT: v_mov_b32_e32 v9, v17
-; GFX9-O0-NEXT: v_mov_b32_e32 v10, v19
-; GFX9-O0-NEXT: v_mov_b32_e32 v11, v20
-; GFX9-O0-NEXT: v_mov_b32_e32 v21, v17
-; GFX9-O0-NEXT: v_mov_b32_e32 v18, v20
-; GFX9-O0-NEXT: v_or_b32_e64 v18, v18, v21
-; GFX9-O0-NEXT: v_mov_b32_e32 v17, v16
-; GFX9-O0-NEXT: v_mov_b32_e32 v16, v19
-; GFX9-O0-NEXT: v_or_b32_e64 v16, v16, v17
-; GFX9-O0-NEXT: ; kill: def $vgpr16 killed $vgpr16 def $vgpr16_vgpr17 killed $exec
-; GFX9-O0-NEXT: v_mov_b32_e32 v17, v18
-; GFX9-O0-NEXT: v_cmp_eq_u64_e64 s[4:5], v[16:17], v[12:13]
+; GFX9-O0-NEXT: ; kill: def $vgpr8 killed $vgpr8 def $vgpr8_vgpr9 killed $exec
+; GFX9-O0-NEXT: v_mov_b32_e32 v9, v10
+; GFX9-O0-NEXT: v_mov_b32_e32 v14, v9
+; GFX9-O0-NEXT: v_mov_b32_e32 v13, v8
+; GFX9-O0-NEXT: buffer_store_dword v13, off, s[0:3], s32 offset:244 ; 4-byte Folded Spill
+; GFX9-O0-NEXT: s_nop 0
+; GFX9-O0-NEXT: buffer_store_dword v14, off, s[0:3], s32 offset:248 ; 4-byte Folded Spill
+; GFX9-O0-NEXT: v_mov_b32_e32 v14, v12
+; GFX9-O0-NEXT: v_mov_b32_e32 v13, v11
+; GFX9-O0-NEXT: buffer_store_dword v13, off, s[0:3], s32 offset:236 ; 4-byte Folded Spill
+; GFX9-O0-NEXT: s_nop 0
+; GFX9-O0-NEXT: buffer_store_dword v14, off, s[0:3], s32 offset:240 ; 4-byte Folded Spill
+; GFX9-O0-NEXT: v_mov_b32_e32 v13, v9
+; GFX9-O0-NEXT: v_mov_b32_e32 v10, v12
+; GFX9-O0-NEXT: v_or_b32_e64 v10, v10, v13
+; GFX9-O0-NEXT: v_mov_b32_e32 v9, v8
+; GFX9-O0-NEXT: v_mov_b32_e32 v8, v11
+; GFX9-O0-NEXT: v_or_b32_e64 v8, v8, v9
+; GFX9-O0-NEXT: ; kill: def $vgpr8 killed $vgpr8 def $vgpr8_vgpr9 killed $exec
+; GFX9-O0-NEXT: v_mov_b32_e32 v9, v10
+; GFX9-O0-NEXT: v_cmp_eq_u64_e64 s[4:5], v[8:9], v[4:5]
; GFX9-O0-NEXT: s_or_b64 s[4:5], s[4:5], s[6:7]
-; GFX9-O0-NEXT: v_mov_b32_e32 v17, v3
-; GFX9-O0-NEXT: v_mov_b32_e32 v16, v2
-; GFX9-O0-NEXT: buffer_store_dword v16, off, s[0:3], s32 offset:120 ; 4-byte Folded Spill
+; GFX9-O0-NEXT: v_mov_b32_e32 v9, v3
+; GFX9-O0-NEXT: v_mov_b32_e32 v8, v2
+; GFX9-O0-NEXT: buffer_store_dword v8, off, s[0:3], s32 offset:124 ; 4-byte Folded Spill
; GFX9-O0-NEXT: s_nop 0
-; GFX9-O0-NEXT: buffer_store_dword v17, off, s[0:3], s32 offset:124 ; 4-byte Folded Spill
-; GFX9-O0-NEXT: v_mov_b32_e32 v17, v1
-; GFX9-O0-NEXT: v_mov_b32_e32 v16, v0
-; GFX9-O0-NEXT: buffer_store_dword v16, off, s[0:3], s32 offset:128 ; 4-byte Folded Spill
+; GFX9-O0-NEXT: buffer_store_dword v9, off, s[0:3], s32 offset:128 ; 4-byte Folded Spill
+; GFX9-O0-NEXT: v_mov_b32_e32 v9, v1
+; GFX9-O0-NEXT: v_mov_b32_e32 v8, v0
+; GFX9-O0-NEXT: buffer_store_dword v8, off, s[0:3], s32 offset:132 ; 4-byte Folded Spill
; GFX9-O0-NEXT: s_nop 0
-; GFX9-O0-NEXT: buffer_store_dword v17, off, s[0:3], s32 offset:132 ; 4-byte Folded Spill
-; GFX9-O0-NEXT: v_mov_b32_e32 v17, v15
-; GFX9-O0-NEXT: v_mov_b32_e32 v16, v14
-; GFX9-O0-NEXT: buffer_store_dword v16, off, s[0:3], s32 offset:136 ; 4-byte Folded Spill
+; GFX9-O0-NEXT: buffer_store_dword v9, off, s[0:3], s32 offset:136 ; 4-byte Folded Spill
+; GFX9-O0-NEXT: v_mov_b32_e32 v9, v7
+; GFX9-O0-NEXT: v_mov_b32_e32 v8, v6
+; GFX9-O0-NEXT: buffer_store_dword v8, off, s[0:3], s32 offset:140 ; 4-byte Folded Spill
; GFX9-O0-NEXT: s_nop 0
-; GFX9-O0-NEXT: buffer_store_dword v17, off, s[0:3], s32 offset:140 ; 4-byte Folded Spill
-; GFX9-O0-NEXT: v_mov_b32_e32 v17, v13
-; GFX9-O0-NEXT: v_mov_b32_e32 v16, v12
-; GFX9-O0-NEXT: buffer_store_dword v16, off, s[0:3], s32 offset:144 ; 4-byte Folded Spill
+; GFX9-O0-NEXT: buffer_store_dword v9, off, s[0:3], s32 offset:144 ; 4-byte Folded Spill
+; GFX9-O0-NEXT: v_mov_b32_e32 v9, v5
+; GFX9-O0-NEXT: v_mov_b32_e32 v8, v4
+; GFX9-O0-NEXT: buffer_store_dword v8, off, s[0:3], s32 offset:148 ; 4-byte Folded Spill
; GFX9-O0-NEXT: s_nop 0
-; GFX9-O0-NEXT: buffer_store_dword v17, off, s[0:3], s32 offset:148 ; 4-byte Folded Spill
+; GFX9-O0-NEXT: buffer_store_dword v9, off, s[0:3], s32 offset:152 ; 4-byte Folded Spill
; GFX9-O0-NEXT: s_mov_b64 s[6:7], s[4:5]
-; GFX9-O0-NEXT: v_writelane_b32 v30, s6, 6
-; GFX9-O0-NEXT: v_writelane_b32 v30, s7, 7
+; GFX9-O0-NEXT: v_writelane_b32 v29, s6, 6
+; GFX9-O0-NEXT: v_writelane_b32 v29, s7, 7
; GFX9-O0-NEXT: s_mov_b64 s[6:7], s[4:5]
-; GFX9-O0-NEXT: v_writelane_b32 v30, s6, 10
-; GFX9-O0-NEXT: v_writelane_b32 v30, s7, 11
+; GFX9-O0-NEXT: v_writelane_b32 v29, s6, 10
+; GFX9-O0-NEXT: v_writelane_b32 v29, s7, 11
; GFX9-O0-NEXT: s_or_saveexec_b64 s[22:23], -1
-; GFX9-O0-NEXT: buffer_store_dword v30, off, s[0:3], s32 ; 4-byte Folded Spill
+; GFX9-O0-NEXT: buffer_store_dword v29, off, s[0:3], s32 ; 4-byte Folded Spill
; GFX9-O0-NEXT: s_mov_b64 exec, s[22:23]
-; GFX9-O0-NEXT: buffer_store_dword v14, off, s[0:3], s32 offset:256 ; 4-byte Folded Spill
+; GFX9-O0-NEXT: buffer_store_dword v6, off, s[0:3], s32 offset:228 ; 4-byte Folded Spill
; GFX9-O0-NEXT: s_nop 0
-; GFX9-O0-NEXT: buffer_store_dword v15, off, s[0:3], s32 offset:260 ; 4-byte Folded Spill
-; GFX9-O0-NEXT: buffer_store_dword v12, off, s[0:3], s32 offset:248 ; 4-byte Folded Spill
+; GFX9-O0-NEXT: buffer_store_dword v7, off, s[0:3], s32 offset:232 ; 4-byte Folded Spill
+; GFX9-O0-NEXT: buffer_store_dword v4, off, s[0:3], s32 offset:220 ; 4-byte Folded Spill
; GFX9-O0-NEXT: s_nop 0
-; GFX9-O0-NEXT: buffer_store_dword v13, off, s[0:3], s32 offset:252 ; 4-byte Folded Spill
-; GFX9-O0-NEXT: buffer_store_dword v10, off, s[0:3], s32 offset:240 ; 4-byte Folded Spill
+; GFX9-O0-NEXT: buffer_store_dword v5, off, s[0:3], s32 offset:224 ; 4-byte Folded Spill
+; GFX9-O0-NEXT: buffer_store_dword v2, off, s[0:3], s32 offset:212 ; 4-byte Folded Spill
; GFX9-O0-NEXT: s_nop 0
-; GFX9-O0-NEXT: buffer_store_dword v11, off, s[0:3], s32 offset:244 ; 4-byte Folded Spill
-; GFX9-O0-NEXT: buffer_store_dword v8, off, s[0:3], s32 offset:232 ; 4-byte Folded Spill
+; GFX9-O0-NEXT: buffer_store_dword v3, off, s[0:3], s32 offset:216 ; 4-byte Folded Spill
+; GFX9-O0-NEXT: buffer_store_dword v0, off, s[0:3], s32 offset:204 ; 4-byte Folded Spill
; GFX9-O0-NEXT: s_nop 0
-; GFX9-O0-NEXT: buffer_store_dword v9, off, s[0:3], s32 offset:236 ; 4-byte Folded Spill
-; GFX9-O0-NEXT: buffer_store_dword v6, off, s[0:3], s32 offset:224 ; 4-byte Folded Spill
-; GFX9-O0-NEXT: s_nop 0
-; GFX9-O0-NEXT: buffer_store_dword v7, off, s[0:3], s32 offset:228 ; 4-byte Folded Spill
-; GFX9-O0-NEXT: buffer_store_dword v4, off, s[0:3], s32 offset:216 ; 4-byte Folded Spill
-; GFX9-O0-NEXT: s_nop 0
-; GFX9-O0-NEXT: buffer_store_dword v5, off, s[0:3], s32 offset:220 ; 4-byte Folded Spill
-; GFX9-O0-NEXT: buffer_store_dword v2, off, s[0:3], s32 offset:208 ; 4-byte Folded Spill
-; GFX9-O0-NEXT: s_nop 0
-; GFX9-O0-NEXT: buffer_store_dword v3, off, s[0:3], s32 offset:212 ; 4-byte Folded Spill
-; GFX9-O0-NEXT: buffer_store_dword v0, off, s[0:3], s32 offset:200 ; 4-byte Folded Spill
-; GFX9-O0-NEXT: s_nop 0
-; GFX9-O0-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:204 ; 4-byte Folded Spill
+; GFX9-O0-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:208 ; 4-byte Folded Spill
; GFX9-O0-NEXT: s_andn2_b64 exec, exec, s[4:5]
; GFX9-O0-NEXT: s_cbranch_execnz .LBB0_6
; GFX9-O0-NEXT: s_branch .LBB0_1
; GFX9-O0-NEXT: .LBB0_7: ; %udiv-preheader
; GFX9-O0-NEXT: s_or_saveexec_b64 s[22:23], -1
-; GFX9-O0-NEXT: buffer_load_dword v30, off, s[0:3], s32 ; 4-byte Folded Reload
+; GFX9-O0-NEXT: buffer_load_dword v29, off, s[0:3], s32 ; 4-byte Folded Reload
; GFX9-O0-NEXT: s_mov_b64 exec, s[22:23]
-; GFX9-O0-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:280 ; 4-byte Folded Reload
-; GFX9-O0-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:284 ; 4-byte Folded Reload
-; GFX9-O0-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:288 ; 4-byte Folded Reload
-; GFX9-O0-NEXT: buffer_load_dword v3, off, s[0:3], s32 offset:292 ; 4-byte Folded Reload
-; GFX9-O0-NEXT: buffer_load_dword v8, off, s[0:3], s32 offset:296 ; 4-byte Folded Reload
-; GFX9-O0-NEXT: buffer_load_dword v9, off, s[0:3], s32 offset:300 ; 4-byte Folded Reload
-; GFX9-O0-NEXT: buffer_load_dword v10, off, s[0:3], s32 offset:304 ; 4-byte Folded Reload
-; GFX9-O0-NEXT: buffer_load_dword v11, off, s[0:3], s32 offset:308 ; 4-byte Folded Reload
-; GFX9-O0-NEXT: buffer_load_dword v16, off, s[0:3], s32 offset:60 ; 4-byte Folded Reload
-; GFX9-O0-NEXT: buffer_load_dword v17, off, s[0:3], s32 offset:64 ; 4-byte Folded Reload
-; GFX9-O0-NEXT: buffer_load_dword v12, off, s[0:3], s32 offset:52 ; 4-byte Folded Reload
-; GFX9-O0-NEXT: buffer_load_dword v13, off, s[0:3], s32 offset:56 ; 4-byte Folded Reload
-; GFX9-O0-NEXT: buffer_load_dword v14, off, s[0:3], s32 offset:44 ; 4-byte Folded Reload
-; GFX9-O0-NEXT: buffer_load_dword v15, off, s[0:3], s32 offset:48 ; 4-byte Folded Reload
+; GFX9-O0-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:284 ; 4-byte Folded Reload
+; GFX9-O0-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:288 ; 4-byte Folded Reload
+; GFX9-O0-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:292 ; 4-byte Folded Reload
+; GFX9-O0-NEXT: buffer_load_dword v3, off, s[0:3], s32 offset:296 ; 4-byte Folded Reload
+; GFX9-O0-NEXT: buffer_load_dword v4, off, s[0:3], s32 offset:300 ; 4-byte Folded Reload
+; GFX9-O0-NEXT: buffer_load_dword v5, off, s[0:3], s32 offset:304 ; 4-byte Folded Reload
+; GFX9-O0-NEXT: buffer_load_dword v6, off, s[0:3], s32 offset:308 ; 4-byte Folded Reload
+; GFX9-O0-NEXT: buffer_load_dword v7, off, s[0:3], s32 offset:312 ; 4-byte Folded Reload
+; GFX9-O0-NEXT: buffer_load_dword v12, off, s[0:3], s32 offset:60 ; 4-byte Folded Reload
+; GFX9-O0-NEXT: buffer_load_dword v13, off, s[0:3], s32 offset:64 ; 4-byte Folded Reload
+; GFX9-O0-NEXT: buffer_load_dword v9, off, s[0:3], s32 offset:52 ; 4-byte Folded Reload
+; GFX9-O0-NEXT: buffer_load_dword v10, off, s[0:3], s32 offset:56 ; 4-byte Folded Reload
+; GFX9-O0-NEXT: buffer_load_dword v16, off, s[0:3], s32 offset:44 ; 4-byte Folded Reload
+; GFX9-O0-NEXT: buffer_load_dword v17, off, s[0:3], s32 offset:48 ; 4-byte Folded Reload
; GFX9-O0-NEXT: buffer_load_dword v18, off, s[0:3], s32 offset:36 ; 4-byte Folded Reload
; GFX9-O0-NEXT: buffer_load_dword v19, off, s[0:3], s32 offset:40 ; 4-byte Folded Reload
; GFX9-O0-NEXT: s_waitcnt vmcnt(9)
-; GFX9-O0-NEXT: v_mov_b32_e32 v4, v10
+; GFX9-O0-NEXT: v_mov_b32_e32 v8, v6
; GFX9-O0-NEXT: s_waitcnt vmcnt(0)
-; GFX9-O0-NEXT: v_lshrrev_b64 v[6:7], v4, v[18:19]
-; GFX9-O0-NEXT: v_mov_b32_e32 v5, v7
+; GFX9-O0-NEXT: v_lshrrev_b64 v[14:15], v8, v[18:19]
+; GFX9-O0-NEXT: v_mov_b32_e32 v11, v15
; GFX9-O0-NEXT: s_mov_b32 s4, 64
-; GFX9-O0-NEXT: v_sub_u32_e64 v20, s4, v4
-; GFX9-O0-NEXT: v_lshlrev_b64 v[20:21], v20, v[14:15]
+; GFX9-O0-NEXT: v_sub_u32_e64 v20, s4, v8
+; GFX9-O0-NEXT: v_lshlrev_b64 v[20:21], v20, v[16:17]
; GFX9-O0-NEXT: v_mov_b32_e32 v22, v21
-; GFX9-O0-NEXT: v_or_b32_e64 v5, v5, v22
-; GFX9-O0-NEXT: ; kill: def $vgpr6 killed $vgpr6 killed $vgpr6_vgpr7 killed $exec
-; GFX9-O0-NEXT: v_mov_b32_e32 v7, v20
-; GFX9-O0-NEXT: v_or_b32_e64 v6, v6, v7
-; GFX9-O0-NEXT: ; kill: def $vgpr6 killed $vgpr6 def $vgpr6_vgpr7 killed $exec
-; GFX9-O0-NEXT: v_mov_b32_e32 v7, v5
-; GFX9-O0-NEXT: v_mov_b32_e32 v22, v7
-; GFX9-O0-NEXT: v_cmp_lt_u32_e64 s[6:7], v4, s4
-; GFX9-O0-NEXT: v_sub_u32_e64 v5, v4, s4
-; GFX9-O0-NEXT: v_lshrrev_b64 v[20:21], v5, v[14:15]
-; GFX9-O0-NEXT: v_mov_b32_e32 v5, v21
-; GFX9-O0-NEXT: v_cndmask_b32_e64 v5, v5, v22, s[6:7]
+; GFX9-O0-NEXT: v_or_b32_e64 v11, v11, v22
+; GFX9-O0-NEXT: ; kill: def $vgpr14 killed $vgpr14 killed $vgpr14_vgpr15 killed $exec
+; GFX9-O0-NEXT: v_mov_b32_e32 v15, v20
+; GFX9-O0-NEXT: v_or_b32_e64 v14, v14, v15
+; GFX9-O0-NEXT: ; kill: def $vgpr14 killed $vgpr14 def $vgpr14_vgpr15 killed $exec
+; GFX9-O0-NEXT: v_mov_b32_e32 v15, v11
+; GFX9-O0-NEXT: v_mov_b32_e32 v22, v15
+; GFX9-O0-NEXT: v_cmp_lt_u32_e64 s[6:7], v8, s4
+; GFX9-O0-NEXT: v_sub_u32_e64 v11, v8, s4
+; GFX9-O0-NEXT: v_lshrrev_b64 v[20:21], v11, v[16:17]
+; GFX9-O0-NEXT: v_mov_b32_e32 v11, v21
+; GFX9-O0-NEXT: v_cndmask_b32_e64 v11, v11, v22, s[6:7]
; GFX9-O0-NEXT: s_mov_b32 s4, 0
-; GFX9-O0-NEXT: v_cmp_eq_u32_e64 s[4:5], v4, s4
+; GFX9-O0-NEXT: v_cmp_eq_u32_e64 s[4:5], v8, s4
; GFX9-O0-NEXT: v_mov_b32_e32 v22, v19
-; GFX9-O0-NEXT: v_cndmask_b32_e64 v5, v5, v22, s[4:5]
-; GFX9-O0-NEXT: v_mov_b32_e32 v7, v6
-; GFX9-O0-NEXT: v_mov_b32_e32 v6, v20
-; GFX9-O0-NEXT: v_cndmask_b32_e64 v6, v6, v7, s[6:7]
-; GFX9-O0-NEXT: v_mov_b32_e32 v7, v18
-; GFX9-O0-NEXT: v_cndmask_b32_e64 v6, v6, v7, s[4:5]
+; GFX9-O0-NEXT: v_cndmask_b32_e64 v11, v11, v22, s[4:5]
+; GFX9-O0-NEXT: v_mov_b32_e32 v15, v14
+; GFX9-O0-NEXT: v_mov_b32_e32 v14, v20
+; GFX9-O0-NEXT: v_cndmask_b32_e64 v14, v14, v15, s[6:7]
+; GFX9-O0-NEXT: v_mov_b32_e32 v15, v18
+; GFX9-O0-NEXT: v_cndmask_b32_e64 v14, v14, v15, s[4:5]
; GFX9-O0-NEXT: ; implicit-def: $sgpr4
; GFX9-O0-NEXT: ; implicit-def: $sgpr4
-; GFX9-O0-NEXT: ; kill: def $vgpr6 killed $vgpr6 def $vgpr6_vgpr7 killed $exec
-; GFX9-O0-NEXT: v_mov_b32_e32 v7, v5
-; GFX9-O0-NEXT: v_lshrrev_b64 v[4:5], v4, v[14:15]
-; GFX9-O0-NEXT: v_mov_b32_e32 v15, v5
+; GFX9-O0-NEXT: ; kill: def $vgpr14 killed $vgpr14 def $vgpr14_vgpr15 killed $exec
+; GFX9-O0-NEXT: v_mov_b32_e32 v15, v11
+; GFX9-O0-NEXT: v_lshrrev_b64 v[16:17], v8, v[16:17]
+; GFX9-O0-NEXT: v_mov_b32_e32 v11, v17
; GFX9-O0-NEXT: s_mov_b64 s[4:5], 0
; GFX9-O0-NEXT: s_mov_b32 s8, s5
-; GFX9-O0-NEXT: v_mov_b32_e32 v14, s8
-; GFX9-O0-NEXT: v_cndmask_b32_e64 v14, v14, v15, s[6:7]
-; GFX9-O0-NEXT: v_mov_b32_e32 v5, v4
+; GFX9-O0-NEXT: v_mov_b32_e32 v8, s8
+; GFX9-O0-NEXT: v_cndmask_b32_e64 v8, v8, v11, s[6:7]
+; GFX9-O0-NEXT: ; kill: def $vgpr16 killed $vgpr16 killed $vgpr16_vgpr17 killed $exec
; GFX9-O0-NEXT: s_mov_b32 s8, s4
-; GFX9-O0-NEXT: v_mov_b32_e32 v4, s8
-; GFX9-O0-NEXT: v_cndmask_b32_e64 v4, v4, v5, s[6:7]
+; GFX9-O0-NEXT: v_mov_b32_e32 v11, s8
+; GFX9-O0-NEXT: v_cndmask_b32_e64 v16, v11, v16, s[6:7]
; GFX9-O0-NEXT: ; implicit-def: $sgpr6
; GFX9-O0-NEXT: ; implicit-def: $sgpr6
-; GFX9-O0-NEXT: ; kill: def $vgpr4 killed $vgpr4 def $vgpr4_vgpr5 killed $exec
-; GFX9-O0-NEXT: v_mov_b32_e32 v5, v14
-; GFX9-O0-NEXT: v_mov_b32_e32 v15, v12
-; GFX9-O0-NEXT: v_mov_b32_e32 v12, v13
+; GFX9-O0-NEXT: ; kill: def $vgpr16 killed $vgpr16 def $vgpr16_vgpr17 killed $exec
+; GFX9-O0-NEXT: v_mov_b32_e32 v17, v8
+; GFX9-O0-NEXT: buffer_store_dword v16, off, s[0:3], s32 offset:260 ; 4-byte Folded Spill
+; GFX9-O0-NEXT: s_nop 0
+; GFX9-O0-NEXT: buffer_store_dword v17, off, s[0:3], s32 offset:264 ; 4-byte Folded Spill
+; GFX9-O0-NEXT: buffer_store_dword v14, off, s[0:3], s32 offset:252 ; 4-byte Folded Spill
+; GFX9-O0-NEXT: s_nop 0
+; GFX9-O0-NEXT: buffer_store_dword v15, off, s[0:3], s32 offset:256 ; 4-byte Folded Spill
+; GFX9-O0-NEXT: v_mov_b32_e32 v8, v9
+; GFX9-O0-NEXT: v_mov_b32_e32 v11, v10
; GFX9-O0-NEXT: s_mov_b64 s[8:9], -1
; GFX9-O0-NEXT: s_mov_b32 s7, s8
; GFX9-O0-NEXT: s_mov_b32 s6, s9
-; GFX9-O0-NEXT: v_mov_b32_e32 v14, v16
-; GFX9-O0-NEXT: v_mov_b32_e32 v13, v17
-; GFX9-O0-NEXT: v_mov_b32_e32 v16, s7
-; GFX9-O0-NEXT: v_add_co_u32_e32 v16, vcc, v15, v16
-; GFX9-O0-NEXT: v_mov_b32_e32 v15, s6
-; GFX9-O0-NEXT: v_addc_co_u32_e32 v12, vcc, v12, v15, vcc
-; GFX9-O0-NEXT: v_mov_b32_e32 v15, s7
-; GFX9-O0-NEXT: v_addc_co_u32_e32 v18, vcc, v14, v15, vcc
-; GFX9-O0-NEXT: v_mov_b32_e32 v14, s6
-; GFX9-O0-NEXT: v_addc_co_u32_e32 v13, vcc, v13, v14, vcc
+; GFX9-O0-NEXT: v_mov_b32_e32 v10, v12
+; GFX9-O0-NEXT: v_mov_b32_e32 v9, v13
+; GFX9-O0-NEXT: v_mov_b32_e32 v12, s7
+; GFX9-O0-NEXT: v_add_co_u32_e32 v8, vcc, v8, v12
+; GFX9-O0-NEXT: v_mov_b32_e32 v12, s6
+; GFX9-O0-NEXT: v_addc_co_u32_e32 v12, vcc, v11, v12, vcc
+; GFX9-O0-NEXT: v_mov_b32_e32 v11, s7
+; GFX9-O0-NEXT: v_addc_co_u32_e32 v10, vcc, v10, v11, vcc
+; GFX9-O0-NEXT: v_mov_b32_e32 v11, s6
+; GFX9-O0-NEXT: v_addc_co_u32_e32 v9, vcc, v9, v11, vcc
; GFX9-O0-NEXT: ; implicit-def: $sgpr6
; GFX9-O0-NEXT: ; implicit-def: $sgpr6
-; GFX9-O0-NEXT: ; kill: def $vgpr18 killed $vgpr18 def $vgpr18_vgpr19 killed $exec
-; GFX9-O0-NEXT: v_mov_b32_e32 v19, v13
+; GFX9-O0-NEXT: ; kill: def $vgpr10 killed $vgpr10 def $vgpr10_vgpr11 killed $exec
+; GFX9-O0-NEXT: v_mov_b32_e32 v11, v9
; GFX9-O0-NEXT: ; implicit-def: $sgpr6
; GFX9-O0-NEXT: ; implicit-def: $sgpr6
-; GFX9-O0-NEXT: ; kill: def $vgpr16 killed $vgpr16 def $vgpr16_vgpr17 killed $exec
-; GFX9-O0-NEXT: v_mov_b32_e32 v17, v12
+; GFX9-O0-NEXT: ; kill: def $vgpr8 killed $vgpr8 def $vgpr8_vgpr9 killed $exec
+; GFX9-O0-NEXT: v_mov_b32_e32 v9, v12
; GFX9-O0-NEXT: v_mov_b32_e32 v13, s5
; GFX9-O0-NEXT: v_mov_b32_e32 v12, s4
-; GFX9-O0-NEXT: v_mov_b32_e32 v15, s5
-; GFX9-O0-NEXT: v_mov_b32_e32 v14, s4
-; GFX9-O0-NEXT: buffer_store_dword v18, off, s[0:3], s32 offset:264 ; 4-byte Folded Spill
-; GFX9-O0-NEXT: s_nop 0
-; GFX9-O0-NEXT: buffer_store_dword v19, off, s[0:3], s32 offset:268 ; 4-byte Folded Spill
-; GFX9-O0-NEXT: buffer_store_dword v16, off, s[0:3], s32 offset:272 ; 4-byte Folded Spill
+; GFX9-O0-NEXT: buffer_store_dword v12, off, s[0:3], s32 offset:220 ; 4-byte Folded Spill
; GFX9-O0-NEXT: s_nop 0
-; GFX9-O0-NEXT: buffer_store_dword v17, off, s[0:3], s32 offset:276 ; 4-byte Folded Spill
-; GFX9-O0-NEXT: v_writelane_b32 v30, s4, 10
-; GFX9-O0-NEXT: v_writelane_b32 v30, s5, 11
-; GFX9-O0-NEXT: s_or_saveexec_b64 s[22:23], -1
-; GFX9-O0-NEXT: buffer_store_dword v30, off, s[0:3], s32 ; 4-byte Folded Spill
-; GFX9-O0-NEXT: s_mov_b64 exec, s[22:23]
-; GFX9-O0-NEXT: buffer_store_dword v14, off, s[0:3], s32 offset:256 ; 4-byte Folded Spill
-; GFX9-O0-NEXT: s_nop 0
-; GFX9-O0-NEXT: buffer_store_dword v15, off, s[0:3], s32 offset:260 ; 4-byte Folded Spill
-; GFX9-O0-NEXT: buffer_store_dword v12, off, s[0:3], s32 offset:248 ; 4-byte Folded Spill
+; GFX9-O0-NEXT: buffer_store_dword v13, off, s[0:3], s32 offset:224 ; 4-byte Folded Spill
+; GFX9-O0-NEXT: v_mov_b32_e32 v13, s5
+; GFX9-O0-NEXT: v_mov_b32_e32 v12, s4
+; GFX9-O0-NEXT: buffer_store_dword v12, off, s[0:3], s32 offset:228 ; 4-byte Folded Spill
; GFX9-O0-NEXT: s_nop 0
-; GFX9-O0-NEXT: buffer_store_dword v13, off, s[0:3], s32 offset:252 ; 4-byte Folded Spill
-; GFX9-O0-NEXT: buffer_store_dword v10, off, s[0:3], s32 offset:240 ; 4-byte Folded Spill
+; GFX9-O0-NEXT: buffer_store_dword v13, off, s[0:3], s32 offset:232 ; 4-byte Folded Spill
+; GFX9-O0-NEXT: buffer_store_dword v10, off, s[0:3], s32 offset:268 ; 4-byte Folded Spill
; GFX9-O0-NEXT: s_nop 0
-; GFX9-O0-NEXT: buffer_store_dword v11, off, s[0:3], s32 offset:244 ; 4-byte Folded Spill
-; GFX9-O0-NEXT: buffer_store_dword v8, off, s[0:3], s32 offset:232 ; 4-byte Folded Spill
+; GFX9-O0-NEXT: buffer_store_dword v11, off, s[0:3], s32 offset:272 ; 4-byte Folded Spill
+; GFX9-O0-NEXT: buffer_store_dword v8, off, s[0:3], s32 offset:276 ; 4-byte Folded Spill
; GFX9-O0-NEXT: s_nop 0
-; GFX9-O0-NEXT: buffer_store_dword v9, off, s[0:3], s32 offset:236 ; 4-byte Folded Spill
-; GFX9-O0-NEXT: buffer_store_dword v6, off, s[0:3], s32 offset:224 ; 4-byte Folded Spill
+; GFX9-O0-NEXT: buffer_store_dword v9, off, s[0:3], s32 offset:280 ; 4-byte Folded Spill
+; GFX9-O0-NEXT: v_writelane_b32 v29, s4, 10
+; GFX9-O0-NEXT: v_writelane_b32 v29, s5, 11
+; GFX9-O0-NEXT: s_or_saveexec_b64 s[22:23], -1
+; GFX9-O0-NEXT: buffer_store_dword v29, off, s[0:3], s32 ; 4-byte Folded Spill
+; GFX9-O0-NEXT: s_mov_b64 exec, s[22:23]
+; GFX9-O0-NEXT: buffer_store_dword v6, off, s[0:3], s32 offset:236 ; 4-byte Folded Spill
; GFX9-O0-NEXT: s_nop 0
-; GFX9-O0-NEXT: buffer_store_dword v7, off, s[0:3], s32 offset:228 ; 4-byte Folded Spill
-; GFX9-O0-NEXT: buffer_store_dword v4, off, s[0:3], s32 offset:216 ; 4-byte Folded Spill
+; GFX9-O0-NEXT: buffer_store_dword v7, off, s[0:3], s32 offset:240 ; 4-byte Folded Spill
+; GFX9-O0-NEXT: buffer_store_dword v4, off, s[0:3], s32 offset:244 ; 4-byte Folded Spill
; GFX9-O0-NEXT: s_nop 0
-; GFX9-O0-NEXT: buffer_store_dword v5, off, s[0:3], s32 offset:220 ; 4-byte Folded Spill
-; GFX9-O0-NEXT: buffer_store_dword v2, off, s[0:3], s32 offset:208 ; 4-byte Folded Spill
+; GFX9-O0-NEXT: buffer_store_dword v5, off, s[0:3], s32 offset:248 ; 4-byte Folded Spill
+; GFX9-O0-NEXT: buffer_store_dword v2, off, s[0:3], s32 offset:212 ; 4-byte Folded Spill
; GFX9-O0-NEXT: s_nop 0
-; GFX9-O0-NEXT: buffer_store_dword v3, off, s[0:3], s32 offset:212 ; 4-byte Folded Spill
-; GFX9-O0-NEXT: buffer_store_dword v0, off, s[0:3], s32 offset:200 ; 4-byte Folded Spill
+; GFX9-O0-NEXT: buffer_store_dword v3, off, s[0:3], s32 offset:216 ; 4-byte Folded Spill
+; GFX9-O0-NEXT: buffer_store_dword v0, off, s[0:3], s32 offset:204 ; 4-byte Folded Spill
; GFX9-O0-NEXT: s_nop 0
-; GFX9-O0-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:204 ; 4-byte Folded Spill
+; GFX9-O0-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:208 ; 4-byte Folded Spill
; GFX9-O0-NEXT: s_branch .LBB0_6
; GFX9-O0-NEXT: .LBB0_8: ; %udiv-bb1
; GFX9-O0-NEXT: s_or_saveexec_b64 s[22:23], -1
-; GFX9-O0-NEXT: buffer_load_dword v30, off, s[0:3], s32 ; 4-byte Folded Reload
+; GFX9-O0-NEXT: buffer_load_dword v29, off, s[0:3], s32 ; 4-byte Folded Reload
; GFX9-O0-NEXT: s_mov_b64 exec, s[22:23]
-; GFX9-O0-NEXT: buffer_load_dword v6, off, s[0:3], s32 offset:36 ; 4-byte Folded Reload
-; GFX9-O0-NEXT: buffer_load_dword v7, off, s[0:3], s32 offset:40 ; 4-byte Folded Reload
-; GFX9-O0-NEXT: buffer_load_dword v10, off, s[0:3], s32 offset:44 ; 4-byte Folded Reload
-; GFX9-O0-NEXT: buffer_load_dword v11, off, s[0:3], s32 offset:48 ; 4-byte Folded Reload
+; GFX9-O0-NEXT: buffer_load_dword v9, off, s[0:3], s32 offset:36 ; 4-byte Folded Reload
+; GFX9-O0-NEXT: buffer_load_dword v10, off, s[0:3], s32 offset:40 ; 4-byte Folded Reload
+; GFX9-O0-NEXT: buffer_load_dword v11, off, s[0:3], s32 offset:44 ; 4-byte Folded Reload
+; GFX9-O0-NEXT: buffer_load_dword v12, off, s[0:3], s32 offset:48 ; 4-byte Folded Reload
; GFX9-O0-NEXT: buffer_load_dword v4, off, s[0:3], s32 offset:20 ; 4-byte Folded Reload
; GFX9-O0-NEXT: buffer_load_dword v5, off, s[0:3], s32 offset:24 ; 4-byte Folded Reload
; GFX9-O0-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:28 ; 4-byte Folded Reload
@@ -1047,129 +1048,129 @@ define i128 @v_sdiv_i128_vv(i128 %lhs, i128 %rhs) {
; GFX9-O0-NEXT: s_mov_b64 s[6:7], 0
; GFX9-O0-NEXT: s_mov_b32 s8, s6
; GFX9-O0-NEXT: s_mov_b32 s9, s7
-; GFX9-O0-NEXT: v_mov_b32_e32 v0, v4
-; GFX9-O0-NEXT: v_mov_b32_e32 v2, v5
+; GFX9-O0-NEXT: v_mov_b32_e32 v2, v4
+; GFX9-O0-NEXT: v_mov_b32_e32 v0, v5
; GFX9-O0-NEXT: v_mov_b32_e32 v4, s5
-; GFX9-O0-NEXT: v_add_co_u32_e32 v8, vcc, v3, v4
+; GFX9-O0-NEXT: v_add_co_u32_e32 v7, vcc, v3, v4
; GFX9-O0-NEXT: v_mov_b32_e32 v4, s4
; GFX9-O0-NEXT: s_waitcnt vmcnt(0)
; GFX9-O0-NEXT: v_addc_co_u32_e32 v1, vcc, v1, v4, vcc
; GFX9-O0-NEXT: v_mov_b32_e32 v4, s8
-; GFX9-O0-NEXT: v_addc_co_u32_e32 v0, vcc, v0, v4, vcc
-; GFX9-O0-NEXT: v_mov_b32_e32 v4, s9
-; GFX9-O0-NEXT: v_addc_co_u32_e32 v2, vcc, v2, v4, vcc
+; GFX9-O0-NEXT: v_addc_co_u32_e32 v4, vcc, v2, v4, vcc
+; GFX9-O0-NEXT: v_mov_b32_e32 v2, s9
+; GFX9-O0-NEXT: v_addc_co_u32_e32 v0, vcc, v0, v2, vcc
; GFX9-O0-NEXT: ; implicit-def: $sgpr4
; GFX9-O0-NEXT: ; implicit-def: $sgpr4
-; GFX9-O0-NEXT: ; kill: def $vgpr8 killed $vgpr8 def $vgpr8_vgpr9 killed $exec
-; GFX9-O0-NEXT: v_mov_b32_e32 v9, v1
+; GFX9-O0-NEXT: ; kill: def $vgpr7 killed $vgpr7 def $vgpr7_vgpr8 killed $exec
+; GFX9-O0-NEXT: v_mov_b32_e32 v8, v1
; GFX9-O0-NEXT: ; implicit-def: $sgpr4
; GFX9-O0-NEXT: ; implicit-def: $sgpr4
-; GFX9-O0-NEXT: ; kill: def $vgpr0 killed $vgpr0 def $vgpr0_vgpr1 killed $exec
-; GFX9-O0-NEXT: v_mov_b32_e32 v1, v2
-; GFX9-O0-NEXT: v_mov_b32_e32 v5, v1
-; GFX9-O0-NEXT: v_mov_b32_e32 v4, v0
-; GFX9-O0-NEXT: buffer_store_dword v4, off, s[0:3], s32 offset:296 ; 4-byte Folded Spill
+; GFX9-O0-NEXT: ; kill: def $vgpr4 killed $vgpr4 def $vgpr4_vgpr5 killed $exec
+; GFX9-O0-NEXT: v_mov_b32_e32 v5, v0
+; GFX9-O0-NEXT: v_mov_b32_e32 v0, v4
+; GFX9-O0-NEXT: v_mov_b32_e32 v1, v5
+; GFX9-O0-NEXT: buffer_store_dword v0, off, s[0:3], s32 offset:300 ; 4-byte Folded Spill
; GFX9-O0-NEXT: s_nop 0
-; GFX9-O0-NEXT: buffer_store_dword v5, off, s[0:3], s32 offset:300 ; 4-byte Folded Spill
-; GFX9-O0-NEXT: v_mov_b32_e32 v4, v8
-; GFX9-O0-NEXT: v_mov_b32_e32 v5, v9
-; GFX9-O0-NEXT: buffer_store_dword v4, off, s[0:3], s32 offset:304 ; 4-byte Folded Spill
+; GFX9-O0-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:304 ; 4-byte Folded Spill
+; GFX9-O0-NEXT: v_mov_b32_e32 v0, v7
+; GFX9-O0-NEXT: v_mov_b32_e32 v1, v8
+; GFX9-O0-NEXT: buffer_store_dword v0, off, s[0:3], s32 offset:308 ; 4-byte Folded Spill
; GFX9-O0-NEXT: s_nop 0
-; GFX9-O0-NEXT: buffer_store_dword v5, off, s[0:3], s32 offset:308 ; 4-byte Folded Spill
+; GFX9-O0-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:312 ; 4-byte Folded Spill
; GFX9-O0-NEXT: s_mov_b32 s4, 0x7f
; GFX9-O0-NEXT: v_sub_u32_e64 v2, s4, v3
-; GFX9-O0-NEXT: v_lshlrev_b64 v[4:5], v2, v[10:11]
-; GFX9-O0-NEXT: v_mov_b32_e32 v12, v5
+; GFX9-O0-NEXT: v_lshlrev_b64 v[0:1], v2, v[11:12]
+; GFX9-O0-NEXT: v_mov_b32_e32 v6, v1
; GFX9-O0-NEXT: s_mov_b32 s4, 64
; GFX9-O0-NEXT: v_sub_u32_e64 v13, s4, v2
-; GFX9-O0-NEXT: v_lshrrev_b64 v[13:14], v13, v[6:7]
+; GFX9-O0-NEXT: v_lshrrev_b64 v[13:14], v13, v[9:10]
; GFX9-O0-NEXT: v_mov_b32_e32 v15, v14
-; GFX9-O0-NEXT: v_or_b32_e64 v12, v12, v15
-; GFX9-O0-NEXT: ; kill: def $vgpr4 killed $vgpr4 killed $vgpr4_vgpr5 killed $exec
-; GFX9-O0-NEXT: v_mov_b32_e32 v5, v13
-; GFX9-O0-NEXT: v_or_b32_e64 v4, v4, v5
-; GFX9-O0-NEXT: ; kill: def $vgpr4 killed $vgpr4 def $vgpr4_vgpr5 killed $exec
-; GFX9-O0-NEXT: v_mov_b32_e32 v5, v12
-; GFX9-O0-NEXT: v_mov_b32_e32 v14, v5
+; GFX9-O0-NEXT: v_or_b32_e64 v6, v6, v15
+; GFX9-O0-NEXT: ; kill: def $vgpr0 killed $vgpr0 killed $vgpr0_vgpr1 killed $exec
+; GFX9-O0-NEXT: v_mov_b32_e32 v1, v13
+; GFX9-O0-NEXT: v_or_b32_e64 v0, v0, v1
+; GFX9-O0-NEXT: ; kill: def $vgpr0 killed $vgpr0 def $vgpr0_vgpr1 killed $exec
+; GFX9-O0-NEXT: v_mov_b32_e32 v1, v6
+; GFX9-O0-NEXT: v_mov_b32_e32 v6, v1
; GFX9-O0-NEXT: v_cmp_lt_u32_e64 s[4:5], v2, s4
; GFX9-O0-NEXT: s_mov_b32 s10, 63
; GFX9-O0-NEXT: v_sub_u32_e64 v3, s10, v3
-; GFX9-O0-NEXT: v_lshlrev_b64 v[12:13], v3, v[6:7]
-; GFX9-O0-NEXT: v_mov_b32_e32 v3, v13
-; GFX9-O0-NEXT: v_cndmask_b32_e64 v3, v3, v14, s[4:5]
+; GFX9-O0-NEXT: v_lshlrev_b64 v[13:14], v3, v[9:10]
+; GFX9-O0-NEXT: v_mov_b32_e32 v3, v14
+; GFX9-O0-NEXT: v_cndmask_b32_e64 v3, v3, v6, s[4:5]
; GFX9-O0-NEXT: s_mov_b32 s10, 0
; GFX9-O0-NEXT: v_cmp_eq_u32_e64 s[10:11], v2, s10
-; GFX9-O0-NEXT: v_mov_b32_e32 v14, v11
-; GFX9-O0-NEXT: v_cndmask_b32_e64 v3, v3, v14, s[10:11]
-; GFX9-O0-NEXT: v_mov_b32_e32 v5, v4
-; GFX9-O0-NEXT: v_mov_b32_e32 v4, v12
-; GFX9-O0-NEXT: v_cndmask_b32_e64 v4, v4, v5, s[4:5]
-; GFX9-O0-NEXT: v_mov_b32_e32 v5, v10
-; GFX9-O0-NEXT: v_cndmask_b32_e64 v4, v4, v5, s[10:11]
+; GFX9-O0-NEXT: v_mov_b32_e32 v6, v12
+; GFX9-O0-NEXT: v_cndmask_b32_e64 v3, v3, v6, s[10:11]
+; GFX9-O0-NEXT: v_mov_b32_e32 v1, v0
+; GFX9-O0-NEXT: v_mov_b32_e32 v0, v13
+; GFX9-O0-NEXT: v_cndmask_b32_e64 v0, v0, v1, s[4:5]
+; GFX9-O0-NEXT: v_mov_b32_e32 v1, v11
+; GFX9-O0-NEXT: v_cndmask_b32_e64 v0, v0, v1, s[10:11]
; GFX9-O0-NEXT: ; implicit-def: $sgpr10
; GFX9-O0-NEXT: ; implicit-def: $sgpr10
-; GFX9-O0-NEXT: ; kill: def $vgpr4 killed $vgpr4 def $vgpr4_vgpr5 killed $exec
-; GFX9-O0-NEXT: v_mov_b32_e32 v5, v3
-; GFX9-O0-NEXT: v_lshlrev_b64 v[6:7], v2, v[6:7]
-; GFX9-O0-NEXT: v_mov_b32_e32 v3, v7
-; GFX9-O0-NEXT: v_mov_b32_e32 v2, s9
+; GFX9-O0-NEXT: ; kill: def $vgpr0 killed $vgpr0 def $vgpr0_vgpr1 killed $exec
+; GFX9-O0-NEXT: v_mov_b32_e32 v1, v3
+; GFX9-O0-NEXT: v_lshlrev_b64 v[2:3], v2, v[9:10]
+; GFX9-O0-NEXT: v_mov_b32_e32 v9, v3
+; GFX9-O0-NEXT: v_mov_b32_e32 v6, s9
+; GFX9-O0-NEXT: v_cndmask_b32_e64 v6, v6, v9, s[4:5]
+; GFX9-O0-NEXT: v_mov_b32_e32 v3, v2
+; GFX9-O0-NEXT: v_mov_b32_e32 v2, s8
; GFX9-O0-NEXT: v_cndmask_b32_e64 v2, v2, v3, s[4:5]
-; GFX9-O0-NEXT: ; kill: def $vgpr6 killed $vgpr6 killed $vgpr6_vgpr7 killed $exec
-; GFX9-O0-NEXT: v_mov_b32_e32 v3, s8
-; GFX9-O0-NEXT: v_cndmask_b32_e64 v6, v3, v6, s[4:5]
; GFX9-O0-NEXT: ; implicit-def: $sgpr4
; GFX9-O0-NEXT: ; implicit-def: $sgpr4
-; GFX9-O0-NEXT: ; kill: def $vgpr6 killed $vgpr6 def $vgpr6_vgpr7 killed $exec
-; GFX9-O0-NEXT: v_mov_b32_e32 v7, v2
-; GFX9-O0-NEXT: buffer_store_dword v6, off, s[0:3], s32 offset:288 ; 4-byte Folded Spill
+; GFX9-O0-NEXT: ; kill: def $vgpr2 killed $vgpr2 def $vgpr2_vgpr3 killed $exec
+; GFX9-O0-NEXT: v_mov_b32_e32 v3, v6
+; GFX9-O0-NEXT: buffer_store_dword v2, off, s[0:3], s32 offset:292 ; 4-byte Folded Spill
; GFX9-O0-NEXT: s_nop 0
-; GFX9-O0-NEXT: buffer_store_dword v7, off, s[0:3], s32 offset:292 ; 4-byte Folded Spill
-; GFX9-O0-NEXT: buffer_store_dword v4, off, s[0:3], s32 offset:280 ; 4-byte Folded Spill
+; GFX9-O0-NEXT: buffer_store_dword v3, off, s[0:3], s32 offset:296 ; 4-byte Folded Spill
+; GFX9-O0-NEXT: buffer_store_dword v0, off, s[0:3], s32 offset:284 ; 4-byte Folded Spill
; GFX9-O0-NEXT: s_nop 0
-; GFX9-O0-NEXT: buffer_store_dword v5, off, s[0:3], s32 offset:284 ; 4-byte Folded Spill
-; GFX9-O0-NEXT: v_mov_b32_e32 v3, v1
-; GFX9-O0-NEXT: v_mov_b32_e32 v2, v9
-; GFX9-O0-NEXT: v_or_b32_e64 v2, v2, v3
-; GFX9-O0-NEXT: v_mov_b32_e32 v1, v0
-; GFX9-O0-NEXT: v_mov_b32_e32 v0, v8
-; GFX9-O0-NEXT: v_or_b32_e64 v0, v0, v1
-; GFX9-O0-NEXT: ; kill: def $vgpr0 killed $vgpr0 def $vgpr0_vgpr1 killed $exec
-; GFX9-O0-NEXT: v_mov_b32_e32 v1, v2
-; GFX9-O0-NEXT: v_cmp_ne_u64_e64 s[4:5], v[0:1], s[6:7]
-; GFX9-O0-NEXT: v_mov_b32_e32 v0, s6
-; GFX9-O0-NEXT: v_mov_b32_e32 v1, s7
-; GFX9-O0-NEXT: v_mov_b32_e32 v2, s6
-; GFX9-O0-NEXT: v_mov_b32_e32 v3, s7
-; GFX9-O0-NEXT: buffer_store_dword v6, off, s[0:3], s32 offset:112 ; 4-byte Folded Spill
+; GFX9-O0-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:288 ; 4-byte Folded Spill
+; GFX9-O0-NEXT: v_mov_b32_e32 v9, v5
+; GFX9-O0-NEXT: v_mov_b32_e32 v6, v8
+; GFX9-O0-NEXT: v_or_b32_e64 v6, v6, v9
+; GFX9-O0-NEXT: v_mov_b32_e32 v5, v4
+; GFX9-O0-NEXT: v_mov_b32_e32 v4, v7
+; GFX9-O0-NEXT: v_or_b32_e64 v4, v4, v5
+; GFX9-O0-NEXT: ; kill: def $vgpr4 killed $vgpr4 def $vgpr4_vgpr5 killed $exec
+; GFX9-O0-NEXT: v_mov_b32_e32 v5, v6
+; GFX9-O0-NEXT: v_cmp_ne_u64_e64 s[4:5], v[4:5], s[6:7]
+; GFX9-O0-NEXT: v_mov_b32_e32 v4, s6
+; GFX9-O0-NEXT: v_mov_b32_e32 v5, s7
+; GFX9-O0-NEXT: buffer_store_dword v4, off, s[0:3], s32 offset:116 ; 4-byte Folded Spill
; GFX9-O0-NEXT: s_nop 0
-; GFX9-O0-NEXT: buffer_store_dword v7, off, s[0:3], s32 offset:116 ; 4-byte Folded Spill
-; GFX9-O0-NEXT: buffer_store_dword v4, off, s[0:3], s32 offset:104 ; 4-byte Folded Spill
+; GFX9-O0-NEXT: buffer_store_dword v5, off, s[0:3], s32 offset:120 ; 4-byte Folded Spill
+; GFX9-O0-NEXT: v_mov_b32_e32 v4, s6
+; GFX9-O0-NEXT: v_mov_b32_e32 v5, s7
+; GFX9-O0-NEXT: buffer_store_dword v4, off, s[0:3], s32 offset:108 ; 4-byte Folded Spill
; GFX9-O0-NEXT: s_nop 0
-; GFX9-O0-NEXT: buffer_store_dword v5, off, s[0:3], s32 offset:108 ; 4-byte Folded Spill
-; GFX9-O0-NEXT: buffer_store_dword v2, off, s[0:3], s32 offset:96 ; 4-byte Folded Spill
+; GFX9-O0-NEXT: buffer_store_dword v5, off, s[0:3], s32 offset:112 ; 4-byte Folded Spill
+; GFX9-O0-NEXT: buffer_store_dword v2, off, s[0:3], s32 offset:92 ; 4-byte Folded Spill
; GFX9-O0-NEXT: s_nop 0
-; GFX9-O0-NEXT: buffer_store_dword v3, off, s[0:3], s32 offset:100 ; 4-byte Folded Spill
-; GFX9-O0-NEXT: buffer_store_dword v0, off, s[0:3], s32 offset:88 ; 4-byte Folded Spill
+; GFX9-O0-NEXT: buffer_store_dword v3, off, s[0:3], s32 offset:96 ; 4-byte Folded Spill
+; GFX9-O0-NEXT: buffer_store_dword v0, off, s[0:3], s32 offset:100 ; 4-byte Folded Spill
; GFX9-O0-NEXT: s_nop 0
-; GFX9-O0-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:92 ; 4-byte Folded Spill
+; GFX9-O0-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:104 ; 4-byte Folded Spill
; GFX9-O0-NEXT: s_mov_b64 s[6:7], exec
; GFX9-O0-NEXT: s_and_b64 s[4:5], s[6:7], s[4:5]
; GFX9-O0-NEXT: s_xor_b64 s[6:7], s[4:5], s[6:7]
-; GFX9-O0-NEXT: v_writelane_b32 v30, s6, 8
-; GFX9-O0-NEXT: v_writelane_b32 v30, s7, 9
+; GFX9-O0-NEXT: v_writelane_b32 v29, s6, 8
+; GFX9-O0-NEXT: v_writelane_b32 v29, s7, 9
; GFX9-O0-NEXT: s_or_saveexec_b64 s[22:23], -1
-; GFX9-O0-NEXT: buffer_store_dword v30, off, s[0:3], s32 ; 4-byte Folded Spill
+; GFX9-O0-NEXT: buffer_store_dword v29, off, s[0:3], s32 ; 4-byte Folded Spill
; GFX9-O0-NEXT: s_mov_b64 exec, s[22:23]
; GFX9-O0-NEXT: s_mov_b64 exec, s[4:5]
; GFX9-O0-NEXT: s_cbranch_execz .LBB0_5
; GFX9-O0-NEXT: s_branch .LBB0_7
; GFX9-O0-NEXT: .LBB0_9: ; %udiv-end
-; GFX9-O0-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:152 ; 4-byte Folded Reload
-; GFX9-O0-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:156 ; 4-byte Folded Reload
+; GFX9-O0-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:156 ; 4-byte Folded Reload
+; GFX9-O0-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:160 ; 4-byte Folded Reload
; GFX9-O0-NEXT: buffer_load_dword v10, off, s[0:3], s32 offset:68 ; 4-byte Folded Reload
; GFX9-O0-NEXT: buffer_load_dword v11, off, s[0:3], s32 offset:72 ; 4-byte Folded Reload
-; GFX9-O0-NEXT: buffer_load_dword v5, off, s[0:3], s32 offset:160 ; 4-byte Folded Reload
-; GFX9-O0-NEXT: buffer_load_dword v6, off, s[0:3], s32 offset:164 ; 4-byte Folded Reload
+; GFX9-O0-NEXT: buffer_load_dword v5, off, s[0:3], s32 offset:164 ; 4-byte Folded Reload
+; GFX9-O0-NEXT: buffer_load_dword v6, off, s[0:3], s32 offset:168 ; 4-byte Folded Reload
; GFX9-O0-NEXT: buffer_load_dword v7, off, s[0:3], s32 offset:76 ; 4-byte Folded Reload
; GFX9-O0-NEXT: buffer_load_dword v8, off, s[0:3], s32 offset:80 ; 4-byte Folded Reload
; GFX9-O0-NEXT: s_waitcnt vmcnt(0)
@@ -1213,7 +1214,7 @@ define i128 @v_sdiv_i128_vv(i128 %lhs, i128 %rhs) {
; GFX9-O0-NEXT: v_lshrrev_b64 v[3:4], s4, v[3:4]
; GFX9-O0-NEXT: ; kill: def $vgpr3 killed $vgpr3 killed $vgpr3_vgpr4 killed $exec
; GFX9-O0-NEXT: s_xor_saveexec_b64 s[4:5], -1
-; GFX9-O0-NEXT: buffer_load_dword v30, off, s[0:3], s32 offset:312 ; 4-byte Folded Reload
+; GFX9-O0-NEXT: buffer_load_dword v29, off, s[0:3], s32 offset:316 ; 4-byte Folded Reload
; GFX9-O0-NEXT: s_mov_b64 exec, s[4:5]
; GFX9-O0-NEXT: s_waitcnt vmcnt(0)
; GFX9-O0-NEXT: s_setpc_b64 s[30:31]
@@ -1684,29 +1685,26 @@ define i128 @v_sdiv_i128_vv(i128 %lhs, i128 %rhs) {
; GFX9-G-O0-NEXT: v_readlane_b32 s5, v31, 3
; GFX9-G-O0-NEXT: s_or_b64 exec, exec, s[4:5]
; GFX9-G-O0-NEXT: ; %bb.2: ; %Flow
-; GFX9-G-O0-NEXT: buffer_load_dword v4, off, s[0:3], s32 offset:132 ; 4-byte Folded Reload
-; GFX9-G-O0-NEXT: buffer_load_dword v5, off, s[0:3], s32 offset:136 ; 4-byte Folded Reload
-; GFX9-G-O0-NEXT: buffer_load_dword v6, off, s[0:3], s32 offset:140 ; 4-byte Folded Reload
-; GFX9-G-O0-NEXT: buffer_load_dword v7, off, s[0:3], s32 offset:144 ; 4-byte Folded Reload
-; GFX9-G-O0-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:148 ; 4-byte Folded Reload
-; GFX9-G-O0-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:152 ; 4-byte Folded Reload
-; GFX9-G-O0-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:156 ; 4-byte Folded Reload
-; GFX9-G-O0-NEXT: buffer_load_dword v3, off, s[0:3], s32 offset:160 ; 4-byte Folded Reload
-; GFX9-G-O0-NEXT: s_waitcnt vmcnt(7)
+; GFX9-G-O0-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:132 ; 4-byte Folded Reload
+; GFX9-G-O0-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:136 ; 4-byte Folded Reload
+; GFX9-G-O0-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:140 ; 4-byte Folded Reload
+; GFX9-G-O0-NEXT: buffer_load_dword v3, off, s[0:3], s32 offset:144 ; 4-byte Folded Reload
+; GFX9-G-O0-NEXT: buffer_load_dword v4, off, s[0:3], s32 offset:148 ; 4-byte Folded Reload
+; GFX9-G-O0-NEXT: buffer_load_dword v5, off, s[0:3], s32 offset:152 ; 4-byte Folded Reload
+; GFX9-G-O0-NEXT: buffer_load_dword v6, off, s[0:3], s32 offset:156 ; 4-byte Folded Reload
+; GFX9-G-O0-NEXT: buffer_load_dword v7, off, s[0:3], s32 offset:160 ; 4-byte Folded Reload
+; GFX9-G-O0-NEXT: s_waitcnt vmcnt(3)
; GFX9-G-O0-NEXT: buffer_store_dword v4, off, s[0:3], s32 offset:116 ; 4-byte Folded Spill
-; GFX9-G-O0-NEXT: s_waitcnt vmcnt(7)
+; GFX9-G-O0-NEXT: s_waitcnt vmcnt(3)
; GFX9-G-O0-NEXT: buffer_store_dword v5, off, s[0:3], s32 offset:120 ; 4-byte Folded Spill
-; GFX9-G-O0-NEXT: s_waitcnt vmcnt(7)
+; GFX9-G-O0-NEXT: s_waitcnt vmcnt(3)
; GFX9-G-O0-NEXT: buffer_store_dword v6, off, s[0:3], s32 offset:124 ; 4-byte Folded Spill
-; GFX9-G-O0-NEXT: s_waitcnt vmcnt(7)
+; GFX9-G-O0-NEXT: s_waitcnt vmcnt(3)
; GFX9-G-O0-NEXT: buffer_store_dword v7, off, s[0:3], s32 offset:128 ; 4-byte Folded Spill
-; GFX9-G-O0-NEXT: s_waitcnt vmcnt(7)
; GFX9-G-O0-NEXT: buffer_store_dword v0, off, s[0:3], s32 offset:100 ; 4-byte Folded Spill
-; GFX9-G-O0-NEXT: s_waitcnt vmcnt(7)
+; GFX9-G-O0-NEXT: s_nop 0
; GFX9-G-O0-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:104 ; 4-byte Folded Spill
-; GFX9-G-O0-NEXT: s_waitcnt vmcnt(7)
; GFX9-G-O0-NEXT: buffer_store_dword v2, off, s[0:3], s32 offset:108 ; 4-byte Folded Spill
-; GFX9-G-O0-NEXT: s_waitcnt vmcnt(7)
; GFX9-G-O0-NEXT: buffer_store_dword v3, off, s[0:3], s32 offset:112 ; 4-byte Folded Spill
; GFX9-G-O0-NEXT: s_branch .LBB0_5
; GFX9-G-O0-NEXT: .LBB0_3: ; %Flow2
@@ -1792,14 +1790,14 @@ define i128 @v_sdiv_i128_vv(i128 %lhs, i128 %rhs) {
; GFX9-G-O0-NEXT: v_readlane_b32 s4, v31, 4
; GFX9-G-O0-NEXT: v_readlane_b32 s5, v31, 5
; GFX9-G-O0-NEXT: s_or_b64 exec, exec, s[4:5]
-; GFX9-G-O0-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:116 ; 4-byte Folded Reload
-; GFX9-G-O0-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:120 ; 4-byte Folded Reload
-; GFX9-G-O0-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:124 ; 4-byte Folded Reload
-; GFX9-G-O0-NEXT: buffer_load_dword v3, off, s[0:3], s32 offset:128 ; 4-byte Folded Reload
-; GFX9-G-O0-NEXT: buffer_load_dword v4, off, s[0:3], s32 offset:100 ; 4-byte Folded Reload
-; GFX9-G-O0-NEXT: buffer_load_dword v5, off, s[0:3], s32 offset:104 ; 4-byte Folded Reload
-; GFX9-G-O0-NEXT: buffer_load_dword v6, off, s[0:3], s32 offset:108 ; 4-byte Folded Reload
-; GFX9-G-O0-NEXT: buffer_load_dword v7, off, s[0:3], s32 offset:112 ; 4-byte Folded Reload
+; GFX9-G-O0-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:100 ; 4-byte Folded Reload
+; GFX9-G-O0-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:104 ; 4-byte Folded Reload
+; GFX9-G-O0-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:108 ; 4-byte Folded Reload
+; GFX9-G-O0-NEXT: buffer_load_dword v3, off, s[0:3], s32 offset:112 ; 4-byte Folded Reload
+; GFX9-G-O0-NEXT: buffer_load_dword v4, off, s[0:3], s32 offset:116 ; 4-byte Folded Reload
+; GFX9-G-O0-NEXT: buffer_load_dword v5, off, s[0:3], s32 offset:120 ; 4-byte Folded Reload
+; GFX9-G-O0-NEXT: buffer_load_dword v6, off, s[0:3], s32 offset:124 ; 4-byte Folded Reload
+; GFX9-G-O0-NEXT: buffer_load_dword v7, off, s[0:3], s32 offset:128 ; 4-byte Folded Reload
; GFX9-G-O0-NEXT: s_waitcnt vmcnt(3)
; GFX9-G-O0-NEXT: buffer_store_dword v4, off, s[0:3], s32 offset:180 ; 4-byte Folded Spill
; GFX9-G-O0-NEXT: s_waitcnt vmcnt(3)
@@ -2242,20 +2240,20 @@ define i128 @v_sdiv_i128_vv(i128 %lhs, i128 %rhs) {
; GFX9-G-O0-NEXT: v_mov_b32_e32 v7, s5
; GFX9-G-O0-NEXT: v_mov_b32_e32 v6, s4
; GFX9-G-O0-NEXT: v_cmp_ne_u64_e64 s[4:5], v[4:5], v[6:7]
-; GFX9-G-O0-NEXT: buffer_store_dword v0, off, s[0:3], s32 offset:116 ; 4-byte Folded Spill
+; GFX9-G-O0-NEXT: buffer_store_dword v0, off, s[0:3], s32 offset:100 ; 4-byte Folded Spill
; GFX9-G-O0-NEXT: s_nop 0
-; GFX9-G-O0-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:120 ; 4-byte Folded Spill
-; GFX9-G-O0-NEXT: buffer_store_dword v2, off, s[0:3], s32 offset:124 ; 4-byte Folded Spill
-; GFX9-G-O0-NEXT: buffer_store_dword v3, off, s[0:3], s32 offset:128 ; 4-byte Folded Spill
+; GFX9-G-O0-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:104 ; 4-byte Folded Spill
+; GFX9-G-O0-NEXT: buffer_store_dword v2, off, s[0:3], s32 offset:108 ; 4-byte Folded Spill
+; GFX9-G-O0-NEXT: buffer_store_dword v3, off, s[0:3], s32 offset:112 ; 4-byte Folded Spill
; GFX9-G-O0-NEXT: v_mov_b32_e32 v0, s8
; GFX9-G-O0-NEXT: v_mov_b32_e32 v1, s9
; GFX9-G-O0-NEXT: v_mov_b32_e32 v2, s10
; GFX9-G-O0-NEXT: v_mov_b32_e32 v3, s11
-; GFX9-G-O0-NEXT: buffer_store_dword v0, off, s[0:3], s32 offset:100 ; 4-byte Folded Spill
+; GFX9-G-O0-NEXT: buffer_store_dword v0, off, s[0:3], s32 offset:116 ; 4-byte Folded Spill
; GFX9-G-O0-NEXT: s_nop 0
-; GFX9-G-O0-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:104 ; 4-byte Folded Spill
-; GFX9-G-O0-NEXT: buffer_store_dword v2, off, s[0:3], s32 offset:108 ; 4-byte Folded Spill
-; GFX9-G-O0-NEXT: buffer_store_dword v3, off, s[0:3], s32 offset:112 ; 4-byte Folded Spill
+; GFX9-G-O0-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:120 ; 4-byte Folded Spill
+; GFX9-G-O0-NEXT: buffer_store_dword v2, off, s[0:3], s32 offset:124 ; 4-byte Folded Spill
+; GFX9-G-O0-NEXT: buffer_store_dword v3, off, s[0:3], s32 offset:128 ; 4-byte Folded Spill
; GFX9-G-O0-NEXT: s_mov_b64 s[6:7], exec
; GFX9-G-O0-NEXT: s_and_b64 s[4:5], s[6:7], s[4:5]
; GFX9-G-O0-NEXT: s_xor_b64 s[6:7], s[4:5], s[6:7]
@@ -2490,7 +2488,7 @@ define i128 @v_udiv_i128_vv(i128 %lhs, i128 %rhs) {
; GFX9-O0: ; %bb.0: ; %_udiv-special-cases
; GFX9-O0-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX9-O0-NEXT: s_xor_saveexec_b64 s[4:5], -1
-; GFX9-O0-NEXT: buffer_store_dword v30, off, s[0:3], s32 offset:296 ; 4-byte Folded Spill
+; GFX9-O0-NEXT: buffer_store_dword v29, off, s[0:3], s32 offset:296 ; 4-byte Folded Spill
; GFX9-O0-NEXT: s_mov_b64 exec, s[4:5]
; GFX9-O0-NEXT: v_mov_b32_e32 v10, v6
; GFX9-O0-NEXT: buffer_store_dword v4, off, s[0:3], s32 offset:68 ; 4-byte Folded Spill
@@ -2547,16 +2545,16 @@ define i128 @v_udiv_i128_vv(i128 %lhs, i128 %rhs) {
; GFX9-O0-NEXT: ; kill: def $vgpr0 killed $vgpr0 def $vgpr0_vgpr1 killed $exec
; GFX9-O0-NEXT: v_mov_b32_e32 v1, v2
; GFX9-O0-NEXT: s_mov_b64 s[6:7], 0
-; GFX9-O0-NEXT: ; implicit-def: $vgpr30 : SGPR spill to VGPR lane
-; GFX9-O0-NEXT: v_writelane_b32 v30, s6, 0
-; GFX9-O0-NEXT: v_writelane_b32 v30, s7, 1
+; GFX9-O0-NEXT: ; implicit-def: $vgpr29 : SGPR spill to VGPR lane
+; GFX9-O0-NEXT: v_writelane_b32 v29, s6, 0
+; GFX9-O0-NEXT: v_writelane_b32 v29, s7, 1
; GFX9-O0-NEXT: v_cmp_eq_u64_e64 s[4:5], v[0:1], s[6:7]
-; GFX9-O0-NEXT: v_mov_b32_e32 v1, v13
-; GFX9-O0-NEXT: v_mov_b32_e32 v3, v9
-; GFX9-O0-NEXT: v_or_b32_e64 v14, v3, v1
-; GFX9-O0-NEXT: v_mov_b32_e32 v0, v12
-; GFX9-O0-NEXT: v_mov_b32_e32 v2, v8
-; GFX9-O0-NEXT: v_or_b32_e64 v8, v2, v0
+; GFX9-O0-NEXT: v_mov_b32_e32 v3, v13
+; GFX9-O0-NEXT: v_mov_b32_e32 v1, v9
+; GFX9-O0-NEXT: v_or_b32_e64 v14, v1, v3
+; GFX9-O0-NEXT: v_mov_b32_e32 v2, v12
+; GFX9-O0-NEXT: v_mov_b32_e32 v0, v8
+; GFX9-O0-NEXT: v_or_b32_e64 v8, v0, v2
; GFX9-O0-NEXT: ; kill: def $vgpr8 killed $vgpr8 def $vgpr8_vgpr9 killed $exec
; GFX9-O0-NEXT: v_mov_b32_e32 v9, v14
; GFX9-O0-NEXT: v_cmp_eq_u64_e64 s[8:9], v[8:9], s[6:7]
@@ -2601,18 +2599,18 @@ define i128 @v_udiv_i128_vv(i128 %lhs, i128 %rhs) {
; GFX9-O0-NEXT: ; implicit-def: $sgpr12
; GFX9-O0-NEXT: ; kill: def $vgpr8 killed $vgpr8 def $vgpr8_vgpr9 killed $exec
; GFX9-O0-NEXT: v_mov_b32_e32 v9, v4
-; GFX9-O0-NEXT: v_ffbh_u32_e64 v4, v0
+; GFX9-O0-NEXT: v_ffbh_u32_e64 v4, v2
; GFX9-O0-NEXT: v_add_u32_e64 v4, v4, s9
-; GFX9-O0-NEXT: v_ffbh_u32_e64 v5, v1
+; GFX9-O0-NEXT: v_ffbh_u32_e64 v5, v3
; GFX9-O0-NEXT: v_min_u32_e64 v5, v4, v5
; GFX9-O0-NEXT: ; implicit-def: $sgpr12
; GFX9-O0-NEXT: v_mov_b32_e32 v4, s8
; GFX9-O0-NEXT: ; kill: def $vgpr5 killed $vgpr5 def $vgpr5_vgpr6 killed $exec
; GFX9-O0-NEXT: v_mov_b32_e32 v6, v4
; GFX9-O0-NEXT: v_mov_b32_e32 v7, v6
-; GFX9-O0-NEXT: v_ffbh_u32_e64 v4, v2
+; GFX9-O0-NEXT: v_ffbh_u32_e64 v4, v0
; GFX9-O0-NEXT: v_add_u32_e64 v4, v4, s9
-; GFX9-O0-NEXT: v_ffbh_u32_e64 v10, v3
+; GFX9-O0-NEXT: v_ffbh_u32_e64 v10, v1
; GFX9-O0-NEXT: v_min_u32_e64 v14, v4, v10
; GFX9-O0-NEXT: ; implicit-def: $sgpr9
; GFX9-O0-NEXT: v_mov_b32_e32 v4, s8
@@ -2695,33 +2693,33 @@ define i128 @v_udiv_i128_vv(i128 %lhs, i128 %rhs) {
; GFX9-O0-NEXT: v_mov_b32_e32 v5, v6
; GFX9-O0-NEXT: v_cmp_ne_u64_e64 s[6:7], v[4:5], s[6:7]
; GFX9-O0-NEXT: v_mov_b32_e32 v4, s11
-; GFX9-O0-NEXT: v_cndmask_b32_e64 v4, v1, v4, s[8:9]
-; GFX9-O0-NEXT: v_mov_b32_e32 v1, s10
-; GFX9-O0-NEXT: v_cndmask_b32_e64 v0, v0, v1, s[8:9]
-; GFX9-O0-NEXT: ; implicit-def: $sgpr12
-; GFX9-O0-NEXT: ; implicit-def: $sgpr12
-; GFX9-O0-NEXT: ; kill: def $vgpr0 killed $vgpr0 def $vgpr0_vgpr1 killed $exec
-; GFX9-O0-NEXT: v_mov_b32_e32 v1, v4
-; GFX9-O0-NEXT: v_mov_b32_e32 v4, s11
; GFX9-O0-NEXT: v_cndmask_b32_e64 v4, v3, v4, s[8:9]
; GFX9-O0-NEXT: v_mov_b32_e32 v3, s10
; GFX9-O0-NEXT: v_cndmask_b32_e64 v2, v2, v3, s[8:9]
-; GFX9-O0-NEXT: ; implicit-def: $sgpr8
-; GFX9-O0-NEXT: ; implicit-def: $sgpr8
+; GFX9-O0-NEXT: ; implicit-def: $sgpr12
+; GFX9-O0-NEXT: ; implicit-def: $sgpr12
; GFX9-O0-NEXT: ; kill: def $vgpr2 killed $vgpr2 def $vgpr2_vgpr3 killed $exec
; GFX9-O0-NEXT: v_mov_b32_e32 v3, v4
-; GFX9-O0-NEXT: s_and_b64 s[6:7], s[4:5], s[6:7]
+; GFX9-O0-NEXT: v_mov_b32_e32 v4, s11
+; GFX9-O0-NEXT: v_cndmask_b32_e64 v4, v1, v4, s[8:9]
+; GFX9-O0-NEXT: v_mov_b32_e32 v1, s10
+; GFX9-O0-NEXT: v_cndmask_b32_e64 v0, v0, v1, s[8:9]
+; GFX9-O0-NEXT: ; implicit-def: $sgpr8
+; GFX9-O0-NEXT: ; implicit-def: $sgpr8
+; GFX9-O0-NEXT: ; kill: def $vgpr0 killed $vgpr0 def $vgpr0_vgpr1 killed $exec
+; GFX9-O0-NEXT: v_mov_b32_e32 v1, v4
; GFX9-O0-NEXT: buffer_store_dword v2, off, s[0:3], s32 offset:12 ; 4-byte Folded Spill
; GFX9-O0-NEXT: s_nop 0
; GFX9-O0-NEXT: buffer_store_dword v3, off, s[0:3], s32 offset:16 ; 4-byte Folded Spill
; GFX9-O0-NEXT: buffer_store_dword v0, off, s[0:3], s32 offset:4 ; 4-byte Folded Spill
; GFX9-O0-NEXT: s_nop 0
; GFX9-O0-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:8 ; 4-byte Folded Spill
+; GFX9-O0-NEXT: s_and_b64 s[6:7], s[4:5], s[6:7]
; GFX9-O0-NEXT: s_mov_b64 s[4:5], exec
-; GFX9-O0-NEXT: v_writelane_b32 v30, s4, 2
-; GFX9-O0-NEXT: v_writelane_b32 v30, s5, 3
+; GFX9-O0-NEXT: v_writelane_b32 v29, s4, 2
+; GFX9-O0-NEXT: v_writelane_b32 v29, s5, 3
; GFX9-O0-NEXT: s_or_saveexec_b64 s[18:19], -1
-; GFX9-O0-NEXT: buffer_store_dword v30, off, s[0:3], s32 ; 4-byte Folded Spill
+; GFX9-O0-NEXT: buffer_store_dword v29, off, s[0:3], s32 ; 4-byte Folded Spill
; GFX9-O0-NEXT: s_mov_b64 exec, s[18:19]
; GFX9-O0-NEXT: s_and_b64 s[4:5], s[4:5], s[6:7]
; GFX9-O0-NEXT: s_mov_b64 exec, s[4:5]
@@ -2729,50 +2727,47 @@ define i128 @v_udiv_i128_vv(i128 %lhs, i128 %rhs) {
; GFX9-O0-NEXT: s_branch .LBB1_8
; GFX9-O0-NEXT: .LBB1_1: ; %Flow
; GFX9-O0-NEXT: s_or_saveexec_b64 s[18:19], -1
-; GFX9-O0-NEXT: buffer_load_dword v30, off, s[0:3], s32 ; 4-byte Folded Reload
+; GFX9-O0-NEXT: buffer_load_dword v29, off, s[0:3], s32 ; 4-byte Folded Reload
; GFX9-O0-NEXT: s_mov_b64 exec, s[18:19]
; GFX9-O0-NEXT: s_waitcnt vmcnt(0)
-; GFX9-O0-NEXT: v_readlane_b32 s4, v30, 4
-; GFX9-O0-NEXT: v_readlane_b32 s5, v30, 5
+; GFX9-O0-NEXT: v_readlane_b32 s4, v29, 4
+; GFX9-O0-NEXT: v_readlane_b32 s5, v29, 5
; GFX9-O0-NEXT: s_or_b64 exec, exec, s[4:5]
; GFX9-O0-NEXT: ; %bb.2: ; %Flow
-; GFX9-O0-NEXT: buffer_load_dword v6, off, s[0:3], s32 offset:104 ; 4-byte Folded Reload
-; GFX9-O0-NEXT: buffer_load_dword v7, off, s[0:3], s32 offset:108 ; 4-byte Folded Reload
-; GFX9-O0-NEXT: buffer_load_dword v4, off, s[0:3], s32 offset:112 ; 4-byte Folded Reload
-; GFX9-O0-NEXT: buffer_load_dword v5, off, s[0:3], s32 offset:116 ; 4-byte Folded Reload
-; GFX9-O0-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:120 ; 4-byte Folded Reload
-; GFX9-O0-NEXT: buffer_load_dword v3, off, s[0:3], s32 offset:124 ; 4-byte Folded Reload
-; GFX9-O0-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:128 ; 4-byte Folded Reload
-; GFX9-O0-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:132 ; 4-byte Folded Reload
-; GFX9-O0-NEXT: s_waitcnt vmcnt(7)
+; GFX9-O0-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:104 ; 4-byte Folded Reload
+; GFX9-O0-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:108 ; 4-byte Folded Reload
+; GFX9-O0-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:112 ; 4-byte Folded Reload
+; GFX9-O0-NEXT: buffer_load_dword v3, off, s[0:3], s32 offset:116 ; 4-byte Folded Reload
+; GFX9-O0-NEXT: buffer_load_dword v4, off, s[0:3], s32 offset:120 ; 4-byte Folded Reload
+; GFX9-O0-NEXT: buffer_load_dword v5, off, s[0:3], s32 offset:124 ; 4-byte Folded Reload
+; GFX9-O0-NEXT: buffer_load_dword v6, off, s[0:3], s32 offset:128 ; 4-byte Folded Reload
+; GFX9-O0-NEXT: buffer_load_dword v7, off, s[0:3], s32 offset:132 ; 4-byte Folded Reload
+; GFX9-O0-NEXT: s_waitcnt vmcnt(1)
; GFX9-O0-NEXT: buffer_store_dword v6, off, s[0:3], s32 offset:96 ; 4-byte Folded Spill
-; GFX9-O0-NEXT: s_waitcnt vmcnt(7)
+; GFX9-O0-NEXT: s_waitcnt vmcnt(1)
; GFX9-O0-NEXT: buffer_store_dword v7, off, s[0:3], s32 offset:100 ; 4-byte Folded Spill
-; GFX9-O0-NEXT: s_waitcnt vmcnt(7)
; GFX9-O0-NEXT: buffer_store_dword v4, off, s[0:3], s32 offset:88 ; 4-byte Folded Spill
-; GFX9-O0-NEXT: s_waitcnt vmcnt(7)
+; GFX9-O0-NEXT: s_nop 0
; GFX9-O0-NEXT: buffer_store_dword v5, off, s[0:3], s32 offset:92 ; 4-byte Folded Spill
-; GFX9-O0-NEXT: s_waitcnt vmcnt(7)
; GFX9-O0-NEXT: buffer_store_dword v2, off, s[0:3], s32 offset:80 ; 4-byte Folded Spill
-; GFX9-O0-NEXT: s_waitcnt vmcnt(7)
+; GFX9-O0-NEXT: s_nop 0
; GFX9-O0-NEXT: buffer_store_dword v3, off, s[0:3], s32 offset:84 ; 4-byte Folded Spill
-; GFX9-O0-NEXT: s_waitcnt vmcnt(7)
; GFX9-O0-NEXT: buffer_store_dword v0, off, s[0:3], s32 offset:72 ; 4-byte Folded Spill
-; GFX9-O0-NEXT: s_waitcnt vmcnt(7)
+; GFX9-O0-NEXT: s_nop 0
; GFX9-O0-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:76 ; 4-byte Folded Spill
; GFX9-O0-NEXT: s_branch .LBB1_5
; GFX9-O0-NEXT: .LBB1_3: ; %Flow2
; GFX9-O0-NEXT: s_or_saveexec_b64 s[18:19], -1
-; GFX9-O0-NEXT: buffer_load_dword v30, off, s[0:3], s32 ; 4-byte Folded Reload
+; GFX9-O0-NEXT: buffer_load_dword v29, off, s[0:3], s32 ; 4-byte Folded Reload
; GFX9-O0-NEXT: s_mov_b64 exec, s[18:19]
; GFX9-O0-NEXT: s_waitcnt vmcnt(0)
-; GFX9-O0-NEXT: v_readlane_b32 s4, v30, 2
-; GFX9-O0-NEXT: v_readlane_b32 s5, v30, 3
+; GFX9-O0-NEXT: v_readlane_b32 s4, v29, 2
+; GFX9-O0-NEXT: v_readlane_b32 s5, v29, 3
; GFX9-O0-NEXT: s_or_b64 exec, exec, s[4:5]
-; GFX9-O0-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:12 ; 4-byte Folded Reload
-; GFX9-O0-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:16 ; 4-byte Folded Reload
-; GFX9-O0-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:4 ; 4-byte Folded Reload
-; GFX9-O0-NEXT: buffer_load_dword v3, off, s[0:3], s32 offset:8 ; 4-byte Folded Reload
+; GFX9-O0-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:4 ; 4-byte Folded Reload
+; GFX9-O0-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:8 ; 4-byte Folded Reload
+; GFX9-O0-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:12 ; 4-byte Folded Reload
+; GFX9-O0-NEXT: buffer_load_dword v3, off, s[0:3], s32 offset:16 ; 4-byte Folded Reload
; GFX9-O0-NEXT: s_waitcnt vmcnt(1)
; GFX9-O0-NEXT: buffer_store_dword v2, off, s[0:3], s32 offset:144 ; 4-byte Folded Spill
; GFX9-O0-NEXT: s_waitcnt vmcnt(1)
@@ -2814,29 +2809,29 @@ define i128 @v_udiv_i128_vv(i128 %lhs, i128 %rhs) {
; GFX9-O0-NEXT: v_or_b32_e64 v2, v2, v3
; GFX9-O0-NEXT: ; kill: def $vgpr2 killed $vgpr2 def $vgpr2_vgpr3 killed $exec
; GFX9-O0-NEXT: v_mov_b32_e32 v3, v4
-; GFX9-O0-NEXT: buffer_store_dword v2, off, s[0:3], s32 offset:12 ; 4-byte Folded Spill
+; GFX9-O0-NEXT: buffer_store_dword v2, off, s[0:3], s32 offset:4 ; 4-byte Folded Spill
; GFX9-O0-NEXT: s_nop 0
-; GFX9-O0-NEXT: buffer_store_dword v3, off, s[0:3], s32 offset:16 ; 4-byte Folded Spill
-; GFX9-O0-NEXT: buffer_store_dword v0, off, s[0:3], s32 offset:4 ; 4-byte Folded Spill
+; GFX9-O0-NEXT: buffer_store_dword v3, off, s[0:3], s32 offset:8 ; 4-byte Folded Spill
+; GFX9-O0-NEXT: buffer_store_dword v0, off, s[0:3], s32 offset:12 ; 4-byte Folded Spill
; GFX9-O0-NEXT: s_nop 0
-; GFX9-O0-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:8 ; 4-byte Folded Spill
+; GFX9-O0-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:16 ; 4-byte Folded Spill
; GFX9-O0-NEXT: s_branch .LBB1_3
; GFX9-O0-NEXT: .LBB1_5: ; %Flow1
; GFX9-O0-NEXT: s_or_saveexec_b64 s[18:19], -1
-; GFX9-O0-NEXT: buffer_load_dword v30, off, s[0:3], s32 ; 4-byte Folded Reload
+; GFX9-O0-NEXT: buffer_load_dword v29, off, s[0:3], s32 ; 4-byte Folded Reload
; GFX9-O0-NEXT: s_mov_b64 exec, s[18:19]
; GFX9-O0-NEXT: s_waitcnt vmcnt(0)
-; GFX9-O0-NEXT: v_readlane_b32 s4, v30, 6
-; GFX9-O0-NEXT: v_readlane_b32 s5, v30, 7
+; GFX9-O0-NEXT: v_readlane_b32 s4, v29, 6
+; GFX9-O0-NEXT: v_readlane_b32 s5, v29, 7
; GFX9-O0-NEXT: s_or_b64 exec, exec, s[4:5]
-; GFX9-O0-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:96 ; 4-byte Folded Reload
-; GFX9-O0-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:100 ; 4-byte Folded Reload
-; GFX9-O0-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:88 ; 4-byte Folded Reload
-; GFX9-O0-NEXT: buffer_load_dword v3, off, s[0:3], s32 offset:92 ; 4-byte Folded Reload
-; GFX9-O0-NEXT: buffer_load_dword v4, off, s[0:3], s32 offset:80 ; 4-byte Folded Reload
-; GFX9-O0-NEXT: buffer_load_dword v5, off, s[0:3], s32 offset:84 ; 4-byte Folded Reload
-; GFX9-O0-NEXT: buffer_load_dword v6, off, s[0:3], s32 offset:72 ; 4-byte Folded Reload
-; GFX9-O0-NEXT: buffer_load_dword v7, off, s[0:3], s32 offset:76 ; 4-byte Folded Reload
+; GFX9-O0-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:72 ; 4-byte Folded Reload
+; GFX9-O0-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:76 ; 4-byte Folded Reload
+; GFX9-O0-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:80 ; 4-byte Folded Reload
+; GFX9-O0-NEXT: buffer_load_dword v3, off, s[0:3], s32 offset:84 ; 4-byte Folded Reload
+; GFX9-O0-NEXT: buffer_load_dword v4, off, s[0:3], s32 offset:88 ; 4-byte Folded Reload
+; GFX9-O0-NEXT: buffer_load_dword v5, off, s[0:3], s32 offset:92 ; 4-byte Folded Reload
+; GFX9-O0-NEXT: buffer_load_dword v6, off, s[0:3], s32 offset:96 ; 4-byte Folded Reload
+; GFX9-O0-NEXT: buffer_load_dword v7, off, s[0:3], s32 offset:100 ; 4-byte Folded Reload
; GFX9-O0-NEXT: s_waitcnt vmcnt(1)
; GFX9-O0-NEXT: buffer_store_dword v6, off, s[0:3], s32 offset:160 ; 4-byte Folded Spill
; GFX9-O0-NEXT: s_waitcnt vmcnt(1)
@@ -2854,214 +2849,214 @@ define i128 @v_udiv_i128_vv(i128 %lhs, i128 %rhs) {
; GFX9-O0-NEXT: .LBB1_6: ; %udiv-do-while
; GFX9-O0-NEXT: ; =>This Inner Loop Header: Depth=1
; GFX9-O0-NEXT: s_or_saveexec_b64 s[18:19], -1
-; GFX9-O0-NEXT: buffer_load_dword v30, off, s[0:3], s32 ; 4-byte Folded Reload
+; GFX9-O0-NEXT: buffer_load_dword v29, off, s[0:3], s32 ; 4-byte Folded Reload
; GFX9-O0-NEXT: s_mov_b64 exec, s[18:19]
; GFX9-O0-NEXT: s_waitcnt vmcnt(0)
-; GFX9-O0-NEXT: v_readlane_b32 s6, v30, 8
-; GFX9-O0-NEXT: v_readlane_b32 s7, v30, 9
-; GFX9-O0-NEXT: buffer_load_dword v6, off, s[0:3], s32 offset:184 ; 4-byte Folded Reload
-; GFX9-O0-NEXT: buffer_load_dword v7, off, s[0:3], s32 offset:188 ; 4-byte Folded Reload
+; GFX9-O0-NEXT: v_readlane_b32 s6, v29, 8
+; GFX9-O0-NEXT: v_readlane_b32 s7, v29, 9
+; GFX9-O0-NEXT: buffer_load_dword v10, off, s[0:3], s32 offset:184 ; 4-byte Folded Reload
+; GFX9-O0-NEXT: buffer_load_dword v11, off, s[0:3], s32 offset:188 ; 4-byte Folded Reload
; GFX9-O0-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:192 ; 4-byte Folded Reload
; GFX9-O0-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:196 ; 4-byte Folded Reload
-; GFX9-O0-NEXT: buffer_load_dword v22, off, s[0:3], s32 offset:200 ; 4-byte Folded Reload
-; GFX9-O0-NEXT: buffer_load_dword v23, off, s[0:3], s32 offset:204 ; 4-byte Folded Reload
-; GFX9-O0-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:208 ; 4-byte Folded Reload
-; GFX9-O0-NEXT: buffer_load_dword v3, off, s[0:3], s32 offset:212 ; 4-byte Folded Reload
-; GFX9-O0-NEXT: buffer_load_dword v16, off, s[0:3], s32 offset:216 ; 4-byte Folded Reload
-; GFX9-O0-NEXT: buffer_load_dword v17, off, s[0:3], s32 offset:220 ; 4-byte Folded Reload
-; GFX9-O0-NEXT: buffer_load_dword v8, off, s[0:3], s32 offset:224 ; 4-byte Folded Reload
-; GFX9-O0-NEXT: buffer_load_dword v9, off, s[0:3], s32 offset:228 ; 4-byte Folded Reload
-; GFX9-O0-NEXT: buffer_load_dword v26, off, s[0:3], s32 offset:232 ; 4-byte Folded Reload
-; GFX9-O0-NEXT: buffer_load_dword v27, off, s[0:3], s32 offset:236 ; 4-byte Folded Reload
-; GFX9-O0-NEXT: buffer_load_dword v24, off, s[0:3], s32 offset:240 ; 4-byte Folded Reload
-; GFX9-O0-NEXT: buffer_load_dword v25, off, s[0:3], s32 offset:244 ; 4-byte Folded Reload
+; GFX9-O0-NEXT: buffer_load_dword v24, off, s[0:3], s32 offset:200 ; 4-byte Folded Reload
+; GFX9-O0-NEXT: buffer_load_dword v25, off, s[0:3], s32 offset:204 ; 4-byte Folded Reload
+; GFX9-O0-NEXT: buffer_load_dword v16, off, s[0:3], s32 offset:208 ; 4-byte Folded Reload
+; GFX9-O0-NEXT: buffer_load_dword v17, off, s[0:3], s32 offset:212 ; 4-byte Folded Reload
+; GFX9-O0-NEXT: buffer_load_dword v8, off, s[0:3], s32 offset:216 ; 4-byte Folded Reload
+; GFX9-O0-NEXT: buffer_load_dword v9, off, s[0:3], s32 offset:220 ; 4-byte Folded Reload
+; GFX9-O0-NEXT: buffer_load_dword v12, off, s[0:3], s32 offset:224 ; 4-byte Folded Reload
+; GFX9-O0-NEXT: buffer_load_dword v13, off, s[0:3], s32 offset:228 ; 4-byte Folded Reload
+; GFX9-O0-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:232 ; 4-byte Folded Reload
+; GFX9-O0-NEXT: buffer_load_dword v3, off, s[0:3], s32 offset:236 ; 4-byte Folded Reload
+; GFX9-O0-NEXT: buffer_load_dword v26, off, s[0:3], s32 offset:240 ; 4-byte Folded Reload
+; GFX9-O0-NEXT: buffer_load_dword v27, off, s[0:3], s32 offset:244 ; 4-byte Folded Reload
; GFX9-O0-NEXT: buffer_load_dword v18, off, s[0:3], s32 offset:52 ; 4-byte Folded Reload
; GFX9-O0-NEXT: buffer_load_dword v19, off, s[0:3], s32 offset:56 ; 4-byte Folded Reload
; GFX9-O0-NEXT: buffer_load_dword v20, off, s[0:3], s32 offset:60 ; 4-byte Folded Reload
; GFX9-O0-NEXT: buffer_load_dword v21, off, s[0:3], s32 offset:64 ; 4-byte Folded Reload
-; GFX9-O0-NEXT: buffer_load_dword v14, off, s[0:3], s32 offset:248 ; 4-byte Folded Reload
-; GFX9-O0-NEXT: buffer_load_dword v15, off, s[0:3], s32 offset:252 ; 4-byte Folded Reload
-; GFX9-O0-NEXT: buffer_load_dword v11, off, s[0:3], s32 offset:256 ; 4-byte Folded Reload
-; GFX9-O0-NEXT: buffer_load_dword v12, off, s[0:3], s32 offset:260 ; 4-byte Folded Reload
+; GFX9-O0-NEXT: buffer_load_dword v22, off, s[0:3], s32 offset:248 ; 4-byte Folded Reload
+; GFX9-O0-NEXT: buffer_load_dword v23, off, s[0:3], s32 offset:252 ; 4-byte Folded Reload
+; GFX9-O0-NEXT: buffer_load_dword v4, off, s[0:3], s32 offset:256 ; 4-byte Folded Reload
+; GFX9-O0-NEXT: buffer_load_dword v5, off, s[0:3], s32 offset:260 ; 4-byte Folded Reload
; GFX9-O0-NEXT: s_mov_b32 s4, 63
-; GFX9-O0-NEXT: s_waitcnt vmcnt(16)
-; GFX9-O0-NEXT: v_lshrrev_b64 v[28:29], s4, v[2:3]
-; GFX9-O0-NEXT: v_mov_b32_e32 v5, v29
+; GFX9-O0-NEXT: s_waitcnt vmcnt(10)
+; GFX9-O0-NEXT: v_lshrrev_b64 v[6:7], s4, v[2:3]
+; GFX9-O0-NEXT: v_mov_b32_e32 v15, v7
; GFX9-O0-NEXT: s_mov_b32 s5, 1
-; GFX9-O0-NEXT: v_lshlrev_b64 v[22:23], s5, v[22:23]
-; GFX9-O0-NEXT: v_mov_b32_e32 v4, v23
-; GFX9-O0-NEXT: v_or_b32_e64 v4, v4, v5
-; GFX9-O0-NEXT: v_mov_b32_e32 v10, v28
-; GFX9-O0-NEXT: v_mov_b32_e32 v5, v22
-; GFX9-O0-NEXT: v_or_b32_e64 v22, v5, v10
-; GFX9-O0-NEXT: ; kill: def $vgpr22 killed $vgpr22 def $vgpr22_vgpr23 killed $exec
-; GFX9-O0-NEXT: v_mov_b32_e32 v23, v4
-; GFX9-O0-NEXT: v_lshlrev_b64 v[28:29], s5, v[2:3]
-; GFX9-O0-NEXT: v_lshrrev_b64 v[4:5], s4, v[6:7]
-; GFX9-O0-NEXT: v_mov_b32_e32 v2, v29
-; GFX9-O0-NEXT: v_mov_b32_e32 v3, v5
+; GFX9-O0-NEXT: s_waitcnt vmcnt(8)
+; GFX9-O0-NEXT: v_lshlrev_b64 v[26:27], s5, v[26:27]
+; GFX9-O0-NEXT: v_mov_b32_e32 v14, v27
+; GFX9-O0-NEXT: v_or_b32_e64 v14, v14, v15
+; GFX9-O0-NEXT: v_mov_b32_e32 v7, v6
+; GFX9-O0-NEXT: v_mov_b32_e32 v6, v26
+; GFX9-O0-NEXT: v_or_b32_e64 v6, v6, v7
+; GFX9-O0-NEXT: ; kill: def $vgpr6 killed $vgpr6 def $vgpr6_vgpr7 killed $exec
+; GFX9-O0-NEXT: v_mov_b32_e32 v7, v14
+; GFX9-O0-NEXT: v_lshlrev_b64 v[26:27], s5, v[2:3]
+; GFX9-O0-NEXT: v_lshrrev_b64 v[14:15], s4, v[10:11]
+; GFX9-O0-NEXT: v_mov_b32_e32 v2, v27
+; GFX9-O0-NEXT: v_mov_b32_e32 v3, v15
; GFX9-O0-NEXT: v_or_b32_e64 v2, v2, v3
-; GFX9-O0-NEXT: v_mov_b32_e32 v3, v28
-; GFX9-O0-NEXT: ; kill: def $vgpr4 killed $vgpr4 killed $vgpr4_vgpr5 killed $exec
-; GFX9-O0-NEXT: v_or_b32_e64 v4, v3, v4
-; GFX9-O0-NEXT: ; kill: def $vgpr4 killed $vgpr4 def $vgpr4_vgpr5 killed $exec
-; GFX9-O0-NEXT: v_mov_b32_e32 v5, v2
+; GFX9-O0-NEXT: v_mov_b32_e32 v3, v26
+; GFX9-O0-NEXT: ; kill: def $vgpr14 killed $vgpr14 killed $vgpr14_vgpr15 killed $exec
+; GFX9-O0-NEXT: v_or_b32_e64 v14, v3, v14
+; GFX9-O0-NEXT: ; kill: def $vgpr14 killed $vgpr14 def $vgpr14_vgpr15 killed $exec
+; GFX9-O0-NEXT: v_mov_b32_e32 v15, v2
; GFX9-O0-NEXT: v_lshlrev_b64 v[2:3], s5, v[0:1]
-; GFX9-O0-NEXT: v_lshlrev_b64 v[28:29], s5, v[6:7]
+; GFX9-O0-NEXT: v_lshlrev_b64 v[26:27], s5, v[10:11]
; GFX9-O0-NEXT: v_lshrrev_b64 v[0:1], s4, v[0:1]
-; GFX9-O0-NEXT: v_mov_b32_e32 v7, v1
-; GFX9-O0-NEXT: v_mov_b32_e32 v6, v29
-; GFX9-O0-NEXT: s_waitcnt vmcnt(10)
+; GFX9-O0-NEXT: v_mov_b32_e32 v11, v1
; GFX9-O0-NEXT: v_mov_b32_e32 v10, v27
-; GFX9-O0-NEXT: v_or3_b32 v6, v6, v7, v10
+; GFX9-O0-NEXT: v_mov_b32_e32 v28, v25
+; GFX9-O0-NEXT: v_or3_b32 v10, v10, v11, v28
; GFX9-O0-NEXT: v_mov_b32_e32 v1, v0
-; GFX9-O0-NEXT: v_mov_b32_e32 v0, v28
-; GFX9-O0-NEXT: v_mov_b32_e32 v7, v26
-; GFX9-O0-NEXT: v_or3_b32 v0, v0, v1, v7
+; GFX9-O0-NEXT: v_mov_b32_e32 v0, v26
+; GFX9-O0-NEXT: v_mov_b32_e32 v11, v24
+; GFX9-O0-NEXT: v_or3_b32 v0, v0, v1, v11
; GFX9-O0-NEXT: ; kill: def $vgpr0 killed $vgpr0 def $vgpr0_vgpr1 killed $exec
-; GFX9-O0-NEXT: v_mov_b32_e32 v1, v6
-; GFX9-O0-NEXT: v_mov_b32_e32 v7, v3
-; GFX9-O0-NEXT: s_waitcnt vmcnt(8)
-; GFX9-O0-NEXT: v_mov_b32_e32 v6, v25
-; GFX9-O0-NEXT: v_or_b32_e64 v6, v6, v7
+; GFX9-O0-NEXT: v_mov_b32_e32 v1, v10
+; GFX9-O0-NEXT: v_mov_b32_e32 v11, v3
+; GFX9-O0-NEXT: v_mov_b32_e32 v10, v17
+; GFX9-O0-NEXT: v_or_b32_e64 v10, v10, v11
; GFX9-O0-NEXT: v_mov_b32_e32 v3, v2
-; GFX9-O0-NEXT: v_mov_b32_e32 v2, v24
+; GFX9-O0-NEXT: v_mov_b32_e32 v2, v16
; GFX9-O0-NEXT: v_or_b32_e64 v2, v2, v3
; GFX9-O0-NEXT: ; kill: def $vgpr2 killed $vgpr2 def $vgpr2_vgpr3 killed $exec
-; GFX9-O0-NEXT: v_mov_b32_e32 v3, v6
-; GFX9-O0-NEXT: v_mov_b32_e32 v6, v4
-; GFX9-O0-NEXT: v_mov_b32_e32 v10, v5
-; GFX9-O0-NEXT: v_mov_b32_e32 v4, v22
-; GFX9-O0-NEXT: v_mov_b32_e32 v5, v23
+; GFX9-O0-NEXT: v_mov_b32_e32 v3, v10
+; GFX9-O0-NEXT: v_mov_b32_e32 v10, v14
+; GFX9-O0-NEXT: v_mov_b32_e32 v16, v15
+; GFX9-O0-NEXT: v_mov_b32_e32 v14, v6
+; GFX9-O0-NEXT: v_mov_b32_e32 v11, v7
; GFX9-O0-NEXT: s_waitcnt vmcnt(1)
-; GFX9-O0-NEXT: v_mov_b32_e32 v13, v11
-; GFX9-O0-NEXT: v_mov_b32_e32 v11, v14
-; GFX9-O0-NEXT: v_mov_b32_e32 v7, v15
-; GFX9-O0-NEXT: v_sub_co_u32_e32 v13, vcc, v13, v6
+; GFX9-O0-NEXT: v_mov_b32_e32 v7, v4
; GFX9-O0-NEXT: s_waitcnt vmcnt(0)
-; GFX9-O0-NEXT: v_subb_co_u32_e32 v12, vcc, v12, v10, vcc
-; GFX9-O0-NEXT: v_subb_co_u32_e32 v11, vcc, v11, v4, vcc
-; GFX9-O0-NEXT: v_subb_co_u32_e32 v7, vcc, v7, v5, vcc
+; GFX9-O0-NEXT: v_mov_b32_e32 v6, v5
+; GFX9-O0-NEXT: v_mov_b32_e32 v4, v22
+; GFX9-O0-NEXT: v_mov_b32_e32 v5, v23
+; GFX9-O0-NEXT: v_sub_co_u32_e32 v7, vcc, v7, v10
+; GFX9-O0-NEXT: v_subb_co_u32_e32 v6, vcc, v6, v16, vcc
+; GFX9-O0-NEXT: v_subb_co_u32_e32 v4, vcc, v4, v14, vcc
+; GFX9-O0-NEXT: v_subb_co_u32_e32 v6, vcc, v5, v11, vcc
; GFX9-O0-NEXT: ; implicit-def: $sgpr5
; GFX9-O0-NEXT: ; implicit-def: $sgpr5
-; GFX9-O0-NEXT: ; kill: def $vgpr11 killed $vgpr11 def $vgpr11_vgpr12 killed $exec
-; GFX9-O0-NEXT: v_mov_b32_e32 v12, v7
-; GFX9-O0-NEXT: v_ashrrev_i64 v[13:14], s4, v[11:12]
-; GFX9-O0-NEXT: v_mov_b32_e32 v7, v14
+; GFX9-O0-NEXT: ; kill: def $vgpr4 killed $vgpr4 def $vgpr4_vgpr5 killed $exec
+; GFX9-O0-NEXT: v_mov_b32_e32 v5, v6
+; GFX9-O0-NEXT: v_ashrrev_i64 v[5:6], s4, v[4:5]
+; GFX9-O0-NEXT: v_mov_b32_e32 v15, v6
; GFX9-O0-NEXT: s_mov_b64 s[4:5], 1
; GFX9-O0-NEXT: s_mov_b32 s8, s5
-; GFX9-O0-NEXT: v_and_b32_e64 v12, v7, s8
-; GFX9-O0-NEXT: v_mov_b32_e32 v11, v13
+; GFX9-O0-NEXT: v_and_b32_e64 v4, v15, s8
+; GFX9-O0-NEXT: v_mov_b32_e32 v17, v5
; GFX9-O0-NEXT: ; kill: def $sgpr4 killed $sgpr4 killed $sgpr4_sgpr5
-; GFX9-O0-NEXT: v_and_b32_e64 v14, v11, s4
-; GFX9-O0-NEXT: ; kill: def $vgpr14 killed $vgpr14 def $vgpr14_vgpr15 killed $exec
-; GFX9-O0-NEXT: v_mov_b32_e32 v15, v12
-; GFX9-O0-NEXT: v_mov_b32_e32 v12, 0
-; GFX9-O0-NEXT: v_mov_b32_e32 v13, 0
+; GFX9-O0-NEXT: v_and_b32_e64 v6, v17, s4
+; GFX9-O0-NEXT: ; kill: def $vgpr6 killed $vgpr6 def $vgpr6_vgpr7 killed $exec
+; GFX9-O0-NEXT: v_mov_b32_e32 v7, v4
+; GFX9-O0-NEXT: v_mov_b32_e32 v4, 0
+; GFX9-O0-NEXT: v_mov_b32_e32 v5, 0
; GFX9-O0-NEXT: v_mov_b32_e32 v22, v21
-; GFX9-O0-NEXT: v_and_b32_e64 v22, v7, v22
-; GFX9-O0-NEXT: v_and_b32_e64 v20, v11, v20
+; GFX9-O0-NEXT: v_and_b32_e64 v22, v15, v22
+; GFX9-O0-NEXT: v_and_b32_e64 v20, v17, v20
; GFX9-O0-NEXT: ; kill: def $vgpr20 killed $vgpr20 def $vgpr20_vgpr21 killed $exec
; GFX9-O0-NEXT: v_mov_b32_e32 v21, v22
; GFX9-O0-NEXT: v_mov_b32_e32 v22, v19
-; GFX9-O0-NEXT: v_and_b32_e64 v7, v7, v22
-; GFX9-O0-NEXT: v_and_b32_e64 v22, v11, v18
-; GFX9-O0-NEXT: ; kill: def $vgpr22 killed $vgpr22 def $vgpr22_vgpr23 killed $exec
-; GFX9-O0-NEXT: v_mov_b32_e32 v23, v7
-; GFX9-O0-NEXT: v_mov_b32_e32 v19, v22
-; GFX9-O0-NEXT: v_mov_b32_e32 v18, v23
-; GFX9-O0-NEXT: v_mov_b32_e32 v11, v20
-; GFX9-O0-NEXT: v_mov_b32_e32 v7, v21
-; GFX9-O0-NEXT: v_sub_co_u32_e32 v6, vcc, v6, v19
-; GFX9-O0-NEXT: v_subb_co_u32_e32 v10, vcc, v10, v18, vcc
-; GFX9-O0-NEXT: v_subb_co_u32_e32 v4, vcc, v4, v11, vcc
-; GFX9-O0-NEXT: v_subb_co_u32_e32 v7, vcc, v5, v7, vcc
+; GFX9-O0-NEXT: v_and_b32_e64 v15, v15, v22
+; GFX9-O0-NEXT: v_and_b32_e64 v17, v17, v18
+; GFX9-O0-NEXT: ; kill: def $vgpr17 killed $vgpr17 def $vgpr17_vgpr18 killed $exec
+; GFX9-O0-NEXT: v_mov_b32_e32 v18, v15
+; GFX9-O0-NEXT: v_mov_b32_e32 v19, v17
+; GFX9-O0-NEXT: ; kill: def $vgpr18 killed $vgpr18 killed $vgpr17_vgpr18 killed $exec
+; GFX9-O0-NEXT: v_mov_b32_e32 v17, v20
+; GFX9-O0-NEXT: v_mov_b32_e32 v15, v21
+; GFX9-O0-NEXT: v_sub_co_u32_e32 v10, vcc, v10, v19
+; GFX9-O0-NEXT: v_subb_co_u32_e32 v16, vcc, v16, v18, vcc
+; GFX9-O0-NEXT: v_subb_co_u32_e32 v14, vcc, v14, v17, vcc
+; GFX9-O0-NEXT: v_subb_co_u32_e32 v11, vcc, v11, v15, vcc
; GFX9-O0-NEXT: ; implicit-def: $sgpr4
; GFX9-O0-NEXT: ; implicit-def: $sgpr4
-; GFX9-O0-NEXT: ; kill: def $vgpr4 killed $vgpr4 def $vgpr4_vgpr5 killed $exec
-; GFX9-O0-NEXT: v_mov_b32_e32 v5, v7
+; GFX9-O0-NEXT: ; kill: def $vgpr14 killed $vgpr14 def $vgpr14_vgpr15 killed $exec
+; GFX9-O0-NEXT: v_mov_b32_e32 v15, v11
; GFX9-O0-NEXT: ; implicit-def: $sgpr4
; GFX9-O0-NEXT: ; implicit-def: $sgpr4
-; GFX9-O0-NEXT: ; kill: def $vgpr6 killed $vgpr6 def $vgpr6_vgpr7 killed $exec
-; GFX9-O0-NEXT: v_mov_b32_e32 v7, v10
+; GFX9-O0-NEXT: ; kill: def $vgpr10 killed $vgpr10 def $vgpr10_vgpr11 killed $exec
+; GFX9-O0-NEXT: v_mov_b32_e32 v11, v16
+; GFX9-O0-NEXT: buffer_store_dword v14, off, s[0:3], s32 offset:240 ; 4-byte Folded Spill
+; GFX9-O0-NEXT: s_nop 0
+; GFX9-O0-NEXT: buffer_store_dword v15, off, s[0:3], s32 offset:244 ; 4-byte Folded Spill
+; GFX9-O0-NEXT: buffer_store_dword v10, off, s[0:3], s32 offset:232 ; 4-byte Folded Spill
+; GFX9-O0-NEXT: s_nop 0
+; GFX9-O0-NEXT: buffer_store_dword v11, off, s[0:3], s32 offset:236 ; 4-byte Folded Spill
; GFX9-O0-NEXT: v_mov_b32_e32 v11, v8
; GFX9-O0-NEXT: ; kill: def $vgpr9 killed $vgpr9 killed $vgpr8_vgpr9 killed $exec
; GFX9-O0-NEXT: s_mov_b64 s[8:9], -1
; GFX9-O0-NEXT: s_mov_b32 s5, s8
; GFX9-O0-NEXT: s_mov_b32 s4, s9
-; GFX9-O0-NEXT: v_mov_b32_e32 v10, v16
-; GFX9-O0-NEXT: v_mov_b32_e32 v8, v17
-; GFX9-O0-NEXT: v_mov_b32_e32 v16, s5
-; GFX9-O0-NEXT: v_add_co_u32_e32 v19, vcc, v11, v16
-; GFX9-O0-NEXT: v_mov_b32_e32 v11, s4
-; GFX9-O0-NEXT: v_addc_co_u32_e32 v9, vcc, v9, v11, vcc
-; GFX9-O0-NEXT: v_mov_b32_e32 v11, s5
-; GFX9-O0-NEXT: v_addc_co_u32_e32 v16, vcc, v10, v11, vcc
-; GFX9-O0-NEXT: v_mov_b32_e32 v10, s4
-; GFX9-O0-NEXT: v_addc_co_u32_e32 v8, vcc, v8, v10, vcc
+; GFX9-O0-NEXT: v_mov_b32_e32 v8, v12
+; GFX9-O0-NEXT: v_mov_b32_e32 v10, v13
+; GFX9-O0-NEXT: v_mov_b32_e32 v12, s5
+; GFX9-O0-NEXT: v_add_co_u32_e32 v11, vcc, v11, v12
+; GFX9-O0-NEXT: v_mov_b32_e32 v12, s4
+; GFX9-O0-NEXT: v_addc_co_u32_e32 v9, vcc, v9, v12, vcc
+; GFX9-O0-NEXT: v_mov_b32_e32 v12, s5
+; GFX9-O0-NEXT: v_addc_co_u32_e32 v8, vcc, v8, v12, vcc
+; GFX9-O0-NEXT: v_mov_b32_e32 v12, s4
+; GFX9-O0-NEXT: v_addc_co_u32_e32 v10, vcc, v10, v12, vcc
; GFX9-O0-NEXT: ; implicit-def: $sgpr4
; GFX9-O0-NEXT: ; implicit-def: $sgpr4
-; GFX9-O0-NEXT: ; kill: def $vgpr19 killed $vgpr19 def $vgpr19_vgpr20 killed $exec
-; GFX9-O0-NEXT: v_mov_b32_e32 v20, v9
+; GFX9-O0-NEXT: ; kill: def $vgpr11 killed $vgpr11 def $vgpr11_vgpr12 killed $exec
+; GFX9-O0-NEXT: v_mov_b32_e32 v12, v9
; GFX9-O0-NEXT: ; implicit-def: $sgpr4
; GFX9-O0-NEXT: ; implicit-def: $sgpr4
-; GFX9-O0-NEXT: ; kill: def $vgpr16 killed $vgpr16 def $vgpr16_vgpr17 killed $exec
-; GFX9-O0-NEXT: v_mov_b32_e32 v17, v8
-; GFX9-O0-NEXT: v_mov_b32_e32 v8, v16
-; GFX9-O0-NEXT: v_mov_b32_e32 v9, v17
-; GFX9-O0-NEXT: v_mov_b32_e32 v10, v19
-; GFX9-O0-NEXT: v_mov_b32_e32 v11, v20
-; GFX9-O0-NEXT: v_mov_b32_e32 v21, v17
-; GFX9-O0-NEXT: v_mov_b32_e32 v18, v20
-; GFX9-O0-NEXT: v_or_b32_e64 v18, v18, v21
-; GFX9-O0-NEXT: v_mov_b32_e32 v17, v16
-; GFX9-O0-NEXT: v_mov_b32_e32 v16, v19
-; GFX9-O0-NEXT: v_or_b32_e64 v16, v16, v17
-; GFX9-O0-NEXT: ; kill: def $vgpr16 killed $vgpr16 def $vgpr16_vgpr17 killed $exec
-; GFX9-O0-NEXT: v_mov_b32_e32 v17, v18
-; GFX9-O0-NEXT: v_cmp_eq_u64_e64 s[4:5], v[16:17], v[12:13]
+; GFX9-O0-NEXT: ; kill: def $vgpr8 killed $vgpr8 def $vgpr8_vgpr9 killed $exec
+; GFX9-O0-NEXT: v_mov_b32_e32 v9, v10
+; GFX9-O0-NEXT: v_mov_b32_e32 v14, v9
+; GFX9-O0-NEXT: v_mov_b32_e32 v13, v8
+; GFX9-O0-NEXT: buffer_store_dword v13, off, s[0:3], s32 offset:224 ; 4-byte Folded Spill
+; GFX9-O0-NEXT: s_nop 0
+; GFX9-O0-NEXT: buffer_store_dword v14, off, s[0:3], s32 offset:228 ; 4-byte Folded Spill
+; GFX9-O0-NEXT: v_mov_b32_e32 v14, v12
+; GFX9-O0-NEXT: v_mov_b32_e32 v13, v11
+; GFX9-O0-NEXT: buffer_store_dword v13, off, s[0:3], s32 offset:216 ; 4-byte Folded Spill
+; GFX9-O0-NEXT: s_nop 0
+; GFX9-O0-NEXT: buffer_store_dword v14, off, s[0:3], s32 offset:220 ; 4-byte Folded Spill
+; GFX9-O0-NEXT: v_mov_b32_e32 v13, v9
+; GFX9-O0-NEXT: v_mov_b32_e32 v10, v12
+; GFX9-O0-NEXT: v_or_b32_e64 v10, v10, v13
+; GFX9-O0-NEXT: v_mov_b32_e32 v9, v8
+; GFX9-O0-NEXT: v_mov_b32_e32 v8, v11
+; GFX9-O0-NEXT: v_or_b32_e64 v8, v8, v9
+; GFX9-O0-NEXT: ; kill: def $vgpr8 killed $vgpr8 def $vgpr8_vgpr9 killed $exec
+; GFX9-O0-NEXT: v_mov_b32_e32 v9, v10
+; GFX9-O0-NEXT: v_cmp_eq_u64_e64 s[4:5], v[8:9], v[4:5]
; GFX9-O0-NEXT: s_or_b64 s[4:5], s[4:5], s[6:7]
-; GFX9-O0-NEXT: v_mov_b32_e32 v17, v3
-; GFX9-O0-NEXT: v_mov_b32_e32 v16, v2
-; GFX9-O0-NEXT: buffer_store_dword v16, off, s[0:3], s32 offset:104 ; 4-byte Folded Spill
+; GFX9-O0-NEXT: v_mov_b32_e32 v9, v3
+; GFX9-O0-NEXT: v_mov_b32_e32 v8, v2
+; GFX9-O0-NEXT: buffer_store_dword v8, off, s[0:3], s32 offset:104 ; 4-byte Folded Spill
; GFX9-O0-NEXT: s_nop 0
-; GFX9-O0-NEXT: buffer_store_dword v17, off, s[0:3], s32 offset:108 ; 4-byte Folded Spill
-; GFX9-O0-NEXT: v_mov_b32_e32 v17, v1
-; GFX9-O0-NEXT: v_mov_b32_e32 v16, v0
-; GFX9-O0-NEXT: buffer_store_dword v16, off, s[0:3], s32 offset:112 ; 4-byte Folded Spill
+; GFX9-O0-NEXT: buffer_store_dword v9, off, s[0:3], s32 offset:108 ; 4-byte Folded Spill
+; GFX9-O0-NEXT: v_mov_b32_e32 v9, v1
+; GFX9-O0-NEXT: v_mov_b32_e32 v8, v0
+; GFX9-O0-NEXT: buffer_store_dword v8, off, s[0:3], s32 offset:112 ; 4-byte Folded Spill
; GFX9-O0-NEXT: s_nop 0
-; GFX9-O0-NEXT: buffer_store_dword v17, off, s[0:3], s32 offset:116 ; 4-byte Folded Spill
-; GFX9-O0-NEXT: v_mov_b32_e32 v17, v15
-; GFX9-O0-NEXT: v_mov_b32_e32 v16, v14
-; GFX9-O0-NEXT: buffer_store_dword v16, off, s[0:3], s32 offset:120 ; 4-byte Folded Spill
+; GFX9-O0-NEXT: buffer_store_dword v9, off, s[0:3], s32 offset:116 ; 4-byte Folded Spill
+; GFX9-O0-NEXT: v_mov_b32_e32 v9, v7
+; GFX9-O0-NEXT: v_mov_b32_e32 v8, v6
+; GFX9-O0-NEXT: buffer_store_dword v8, off, s[0:3], s32 offset:120 ; 4-byte Folded Spill
; GFX9-O0-NEXT: s_nop 0
-; GFX9-O0-NEXT: buffer_store_dword v17, off, s[0:3], s32 offset:124 ; 4-byte Folded Spill
-; GFX9-O0-NEXT: v_mov_b32_e32 v17, v13
-; GFX9-O0-NEXT: v_mov_b32_e32 v16, v12
-; GFX9-O0-NEXT: buffer_store_dword v16, off, s[0:3], s32 offset:128 ; 4-byte Folded Spill
+; GFX9-O0-NEXT: buffer_store_dword v9, off, s[0:3], s32 offset:124 ; 4-byte Folded Spill
+; GFX9-O0-NEXT: v_mov_b32_e32 v9, v5
+; GFX9-O0-NEXT: v_mov_b32_e32 v8, v4
+; GFX9-O0-NEXT: buffer_store_dword v8, off, s[0:3], s32 offset:128 ; 4-byte Folded Spill
; GFX9-O0-NEXT: s_nop 0
-; GFX9-O0-NEXT: buffer_store_dword v17, off, s[0:3], s32 offset:132 ; 4-byte Folded Spill
+; GFX9-O0-NEXT: buffer_store_dword v9, off, s[0:3], s32 offset:132 ; 4-byte Folded Spill
; GFX9-O0-NEXT: s_mov_b64 s[6:7], s[4:5]
-; GFX9-O0-NEXT: v_writelane_b32 v30, s6, 4
-; GFX9-O0-NEXT: v_writelane_b32 v30, s7, 5
+; GFX9-O0-NEXT: v_writelane_b32 v29, s6, 4
+; GFX9-O0-NEXT: v_writelane_b32 v29, s7, 5
; GFX9-O0-NEXT: s_mov_b64 s[6:7], s[4:5]
-; GFX9-O0-NEXT: v_writelane_b32 v30, s6, 8
-; GFX9-O0-NEXT: v_writelane_b32 v30, s7, 9
+; GFX9-O0-NEXT: v_writelane_b32 v29, s6, 8
+; GFX9-O0-NEXT: v_writelane_b32 v29, s7, 9
; GFX9-O0-NEXT: s_or_saveexec_b64 s[18:19], -1
-; GFX9-O0-NEXT: buffer_store_dword v30, off, s[0:3], s32 ; 4-byte Folded Spill
+; GFX9-O0-NEXT: buffer_store_dword v29, off, s[0:3], s32 ; 4-byte Folded Spill
; GFX9-O0-NEXT: s_mov_b64 exec, s[18:19]
-; GFX9-O0-NEXT: buffer_store_dword v14, off, s[0:3], s32 offset:240 ; 4-byte Folded Spill
-; GFX9-O0-NEXT: s_nop 0
-; GFX9-O0-NEXT: buffer_store_dword v15, off, s[0:3], s32 offset:244 ; 4-byte Folded Spill
-; GFX9-O0-NEXT: buffer_store_dword v12, off, s[0:3], s32 offset:232 ; 4-byte Folded Spill
-; GFX9-O0-NEXT: s_nop 0
-; GFX9-O0-NEXT: buffer_store_dword v13, off, s[0:3], s32 offset:236 ; 4-byte Folded Spill
-; GFX9-O0-NEXT: buffer_store_dword v10, off, s[0:3], s32 offset:224 ; 4-byte Folded Spill
-; GFX9-O0-NEXT: s_nop 0
-; GFX9-O0-NEXT: buffer_store_dword v11, off, s[0:3], s32 offset:228 ; 4-byte Folded Spill
-; GFX9-O0-NEXT: buffer_store_dword v8, off, s[0:3], s32 offset:216 ; 4-byte Folded Spill
-; GFX9-O0-NEXT: s_nop 0
-; GFX9-O0-NEXT: buffer_store_dword v9, off, s[0:3], s32 offset:220 ; 4-byte Folded Spill
; GFX9-O0-NEXT: buffer_store_dword v6, off, s[0:3], s32 offset:208 ; 4-byte Folded Spill
; GFX9-O0-NEXT: s_nop 0
; GFX9-O0-NEXT: buffer_store_dword v7, off, s[0:3], s32 offset:212 ; 4-byte Folded Spill
@@ -3079,128 +3074,128 @@ define i128 @v_udiv_i128_vv(i128 %lhs, i128 %rhs) {
; GFX9-O0-NEXT: s_branch .LBB1_1
; GFX9-O0-NEXT: .LBB1_7: ; %udiv-preheader
; GFX9-O0-NEXT: s_or_saveexec_b64 s[18:19], -1
-; GFX9-O0-NEXT: buffer_load_dword v30, off, s[0:3], s32 ; 4-byte Folded Reload
+; GFX9-O0-NEXT: buffer_load_dword v29, off, s[0:3], s32 ; 4-byte Folded Reload
; GFX9-O0-NEXT: s_mov_b64 exec, s[18:19]
; GFX9-O0-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:264 ; 4-byte Folded Reload
; GFX9-O0-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:268 ; 4-byte Folded Reload
; GFX9-O0-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:272 ; 4-byte Folded Reload
; GFX9-O0-NEXT: buffer_load_dword v3, off, s[0:3], s32 offset:276 ; 4-byte Folded Reload
-; GFX9-O0-NEXT: buffer_load_dword v8, off, s[0:3], s32 offset:280 ; 4-byte Folded Reload
-; GFX9-O0-NEXT: buffer_load_dword v9, off, s[0:3], s32 offset:284 ; 4-byte Folded Reload
-; GFX9-O0-NEXT: buffer_load_dword v10, off, s[0:3], s32 offset:288 ; 4-byte Folded Reload
-; GFX9-O0-NEXT: buffer_load_dword v11, off, s[0:3], s32 offset:292 ; 4-byte Folded Reload
-; GFX9-O0-NEXT: buffer_load_dword v16, off, s[0:3], s32 offset:60 ; 4-byte Folded Reload
-; GFX9-O0-NEXT: buffer_load_dword v17, off, s[0:3], s32 offset:64 ; 4-byte Folded Reload
-; GFX9-O0-NEXT: buffer_load_dword v12, off, s[0:3], s32 offset:52 ; 4-byte Folded Reload
-; GFX9-O0-NEXT: buffer_load_dword v13, off, s[0:3], s32 offset:56 ; 4-byte Folded Reload
-; GFX9-O0-NEXT: buffer_load_dword v14, off, s[0:3], s32 offset:44 ; 4-byte Folded Reload
-; GFX9-O0-NEXT: buffer_load_dword v15, off, s[0:3], s32 offset:48 ; 4-byte Folded Reload
+; GFX9-O0-NEXT: buffer_load_dword v4, off, s[0:3], s32 offset:280 ; 4-byte Folded Reload
+; GFX9-O0-NEXT: buffer_load_dword v5, off, s[0:3], s32 offset:284 ; 4-byte Folded Reload
+; GFX9-O0-NEXT: buffer_load_dword v6, off, s[0:3], s32 offset:288 ; 4-byte Folded Reload
+; GFX9-O0-NEXT: buffer_load_dword v7, off, s[0:3], s32 offset:292 ; 4-byte Folded Reload
+; GFX9-O0-NEXT: buffer_load_dword v12, off, s[0:3], s32 offset:60 ; 4-byte Folded Reload
+; GFX9-O0-NEXT: buffer_load_dword v13, off, s[0:3], s32 offset:64 ; 4-byte Folded Reload
+; GFX9-O0-NEXT: buffer_load_dword v9, off, s[0:3], s32 offset:52 ; 4-byte Folded Reload
+; GFX9-O0-NEXT: buffer_load_dword v10, off, s[0:3], s32 offset:56 ; 4-byte Folded Reload
+; GFX9-O0-NEXT: buffer_load_dword v16, off, s[0:3], s32 offset:44 ; 4-byte Folded Reload
+; GFX9-O0-NEXT: buffer_load_dword v17, off, s[0:3], s32 offset:48 ; 4-byte Folded Reload
; GFX9-O0-NEXT: buffer_load_dword v18, off, s[0:3], s32 offset:36 ; 4-byte Folded Reload
; GFX9-O0-NEXT: buffer_load_dword v19, off, s[0:3], s32 offset:40 ; 4-byte Folded Reload
; GFX9-O0-NEXT: s_waitcnt vmcnt(9)
-; GFX9-O0-NEXT: v_mov_b32_e32 v4, v10
+; GFX9-O0-NEXT: v_mov_b32_e32 v8, v6
; GFX9-O0-NEXT: s_waitcnt vmcnt(0)
-; GFX9-O0-NEXT: v_lshrrev_b64 v[6:7], v4, v[18:19]
-; GFX9-O0-NEXT: v_mov_b32_e32 v5, v7
+; GFX9-O0-NEXT: v_lshrrev_b64 v[14:15], v8, v[18:19]
+; GFX9-O0-NEXT: v_mov_b32_e32 v11, v15
; GFX9-O0-NEXT: s_mov_b32 s4, 64
-; GFX9-O0-NEXT: v_sub_u32_e64 v20, s4, v4
-; GFX9-O0-NEXT: v_lshlrev_b64 v[20:21], v20, v[14:15]
+; GFX9-O0-NEXT: v_sub_u32_e64 v20, s4, v8
+; GFX9-O0-NEXT: v_lshlrev_b64 v[20:21], v20, v[16:17]
; GFX9-O0-NEXT: v_mov_b32_e32 v22, v21
-; GFX9-O0-NEXT: v_or_b32_e64 v5, v5, v22
-; GFX9-O0-NEXT: ; kill: def $vgpr6 killed $vgpr6 killed $vgpr6_vgpr7 killed $exec
-; GFX9-O0-NEXT: v_mov_b32_e32 v7, v20
-; GFX9-O0-NEXT: v_or_b32_e64 v6, v6, v7
-; GFX9-O0-NEXT: ; kill: def $vgpr6 killed $vgpr6 def $vgpr6_vgpr7 killed $exec
-; GFX9-O0-NEXT: v_mov_b32_e32 v7, v5
-; GFX9-O0-NEXT: v_mov_b32_e32 v22, v7
-; GFX9-O0-NEXT: v_cmp_lt_u32_e64 s[6:7], v4, s4
-; GFX9-O0-NEXT: v_sub_u32_e64 v5, v4, s4
-; GFX9-O0-NEXT: v_lshrrev_b64 v[20:21], v5, v[14:15]
-; GFX9-O0-NEXT: v_mov_b32_e32 v5, v21
-; GFX9-O0-NEXT: v_cndmask_b32_e64 v5, v5, v22, s[6:7]
+; GFX9-O0-NEXT: v_or_b32_e64 v11, v11, v22
+; GFX9-O0-NEXT: ; kill: def $vgpr14 killed $vgpr14 killed $vgpr14_vgpr15 killed $exec
+; GFX9-O0-NEXT: v_mov_b32_e32 v15, v20
+; GFX9-O0-NEXT: v_or_b32_e64 v14, v14, v15
+; GFX9-O0-NEXT: ; kill: def $vgpr14 killed $vgpr14 def $vgpr14_vgpr15 killed $exec
+; GFX9-O0-NEXT: v_mov_b32_e32 v15, v11
+; GFX9-O0-NEXT: v_mov_b32_e32 v22, v15
+; GFX9-O0-NEXT: v_cmp_lt_u32_e64 s[6:7], v8, s4
+; GFX9-O0-NEXT: v_sub_u32_e64 v11, v8, s4
+; GFX9-O0-NEXT: v_lshrrev_b64 v[20:21], v11, v[16:17]
+; GFX9-O0-NEXT: v_mov_b32_e32 v11, v21
+; GFX9-O0-NEXT: v_cndmask_b32_e64 v11, v11, v22, s[6:7]
; GFX9-O0-NEXT: s_mov_b32 s4, 0
-; GFX9-O0-NEXT: v_cmp_eq_u32_e64 s[4:5], v4, s4
+; GFX9-O0-NEXT: v_cmp_eq_u32_e64 s[4:5], v8, s4
; GFX9-O0-NEXT: v_mov_b32_e32 v22, v19
-; GFX9-O0-NEXT: v_cndmask_b32_e64 v5, v5, v22, s[4:5]
-; GFX9-O0-NEXT: v_mov_b32_e32 v7, v6
-; GFX9-O0-NEXT: v_mov_b32_e32 v6, v20
-; GFX9-O0-NEXT: v_cndmask_b32_e64 v6, v6, v7, s[6:7]
-; GFX9-O0-NEXT: v_mov_b32_e32 v7, v18
-; GFX9-O0-NEXT: v_cndmask_b32_e64 v6, v6, v7, s[4:5]
+; GFX9-O0-NEXT: v_cndmask_b32_e64 v11, v11, v22, s[4:5]
+; GFX9-O0-NEXT: v_mov_b32_e32 v15, v14
+; GFX9-O0-NEXT: v_mov_b32_e32 v14, v20
+; GFX9-O0-NEXT: v_cndmask_b32_e64 v14, v14, v15, s[6:7]
+; GFX9-O0-NEXT: v_mov_b32_e32 v15, v18
+; GFX9-O0-NEXT: v_cndmask_b32_e64 v14, v14, v15, s[4:5]
; GFX9-O0-NEXT: ; implicit-def: $sgpr4
; GFX9-O0-NEXT: ; implicit-def: $sgpr4
-; GFX9-O0-NEXT: ; kill: def $vgpr6 killed $vgpr6 def $vgpr6_vgpr7 killed $exec
-; GFX9-O0-NEXT: v_mov_b32_e32 v7, v5
-; GFX9-O0-NEXT: v_lshrrev_b64 v[4:5], v4, v[14:15]
-; GFX9-O0-NEXT: v_mov_b32_e32 v15, v5
+; GFX9-O0-NEXT: ; kill: def $vgpr14 killed $vgpr14 def $vgpr14_vgpr15 killed $exec
+; GFX9-O0-NEXT: v_mov_b32_e32 v15, v11
+; GFX9-O0-NEXT: v_lshrrev_b64 v[16:17], v8, v[16:17]
+; GFX9-O0-NEXT: v_mov_b32_e32 v11, v17
; GFX9-O0-NEXT: s_mov_b64 s[4:5], 0
; GFX9-O0-NEXT: s_mov_b32 s8, s5
-; GFX9-O0-NEXT: v_mov_b32_e32 v14, s8
-; GFX9-O0-NEXT: v_cndmask_b32_e64 v14, v14, v15, s[6:7]
-; GFX9-O0-NEXT: v_mov_b32_e32 v5, v4
+; GFX9-O0-NEXT: v_mov_b32_e32 v8, s8
+; GFX9-O0-NEXT: v_cndmask_b32_e64 v8, v8, v11, s[6:7]
+; GFX9-O0-NEXT: ; kill: def $vgpr16 killed $vgpr16 killed $vgpr16_vgpr17 killed $exec
; GFX9-O0-NEXT: s_mov_b32 s8, s4
-; GFX9-O0-NEXT: v_mov_b32_e32 v4, s8
-; GFX9-O0-NEXT: v_cndmask_b32_e64 v4, v4, v5, s[6:7]
+; GFX9-O0-NEXT: v_mov_b32_e32 v11, s8
+; GFX9-O0-NEXT: v_cndmask_b32_e64 v16, v11, v16, s[6:7]
; GFX9-O0-NEXT: ; implicit-def: $sgpr6
; GFX9-O0-NEXT: ; implicit-def: $sgpr6
-; GFX9-O0-NEXT: ; kill: def $vgpr4 killed $vgpr4 def $vgpr4_vgpr5 killed $exec
-; GFX9-O0-NEXT: v_mov_b32_e32 v5, v14
-; GFX9-O0-NEXT: v_mov_b32_e32 v15, v12
-; GFX9-O0-NEXT: v_mov_b32_e32 v12, v13
+; GFX9-O0-NEXT: ; kill: def $vgpr16 killed $vgpr16 def $vgpr16_vgpr17 killed $exec
+; GFX9-O0-NEXT: v_mov_b32_e32 v17, v8
+; GFX9-O0-NEXT: buffer_store_dword v16, off, s[0:3], s32 offset:240 ; 4-byte Folded Spill
+; GFX9-O0-NEXT: s_nop 0
+; GFX9-O0-NEXT: buffer_store_dword v17, off, s[0:3], s32 offset:244 ; 4-byte Folded Spill
+; GFX9-O0-NEXT: buffer_store_dword v14, off, s[0:3], s32 offset:232 ; 4-byte Folded Spill
+; GFX9-O0-NEXT: s_nop 0
+; GFX9-O0-NEXT: buffer_store_dword v15, off, s[0:3], s32 offset:236 ; 4-byte Folded Spill
+; GFX9-O0-NEXT: v_mov_b32_e32 v8, v9
+; GFX9-O0-NEXT: v_mov_b32_e32 v11, v10
; GFX9-O0-NEXT: s_mov_b64 s[8:9], -1
; GFX9-O0-NEXT: s_mov_b32 s7, s8
; GFX9-O0-NEXT: s_mov_b32 s6, s9
-; GFX9-O0-NEXT: v_mov_b32_e32 v14, v16
-; GFX9-O0-NEXT: v_mov_b32_e32 v13, v17
-; GFX9-O0-NEXT: v_mov_b32_e32 v16, s7
-; GFX9-O0-NEXT: v_add_co_u32_e32 v16, vcc, v15, v16
-; GFX9-O0-NEXT: v_mov_b32_e32 v15, s6
-; GFX9-O0-NEXT: v_addc_co_u32_e32 v12, vcc, v12, v15, vcc
-; GFX9-O0-NEXT: v_mov_b32_e32 v15, s7
-; GFX9-O0-NEXT: v_addc_co_u32_e32 v18, vcc, v14, v15, vcc
-; GFX9-O0-NEXT: v_mov_b32_e32 v14, s6
-; GFX9-O0-NEXT: v_addc_co_u32_e32 v13, vcc, v13, v14, vcc
+; GFX9-O0-NEXT: v_mov_b32_e32 v10, v12
+; GFX9-O0-NEXT: v_mov_b32_e32 v9, v13
+; GFX9-O0-NEXT: v_mov_b32_e32 v12, s7
+; GFX9-O0-NEXT: v_add_co_u32_e32 v8, vcc, v8, v12
+; GFX9-O0-NEXT: v_mov_b32_e32 v12, s6
+; GFX9-O0-NEXT: v_addc_co_u32_e32 v12, vcc, v11, v12, vcc
+; GFX9-O0-NEXT: v_mov_b32_e32 v11, s7
+; GFX9-O0-NEXT: v_addc_co_u32_e32 v10, vcc, v10, v11, vcc
+; GFX9-O0-NEXT: v_mov_b32_e32 v11, s6
+; GFX9-O0-NEXT: v_addc_co_u32_e32 v9, vcc, v9, v11, vcc
; GFX9-O0-NEXT: ; implicit-def: $sgpr6
; GFX9-O0-NEXT: ; implicit-def: $sgpr6
-; GFX9-O0-NEXT: ; kill: def $vgpr18 killed $vgpr18 def $vgpr18_vgpr19 killed $exec
-; GFX9-O0-NEXT: v_mov_b32_e32 v19, v13
+; GFX9-O0-NEXT: ; kill: def $vgpr10 killed $vgpr10 def $vgpr10_vgpr11 killed $exec
+; GFX9-O0-NEXT: v_mov_b32_e32 v11, v9
; GFX9-O0-NEXT: ; implicit-def: $sgpr6
; GFX9-O0-NEXT: ; implicit-def: $sgpr6
-; GFX9-O0-NEXT: ; kill: def $vgpr16 killed $vgpr16 def $vgpr16_vgpr17 killed $exec
-; GFX9-O0-NEXT: v_mov_b32_e32 v17, v12
+; GFX9-O0-NEXT: ; kill: def $vgpr8 killed $vgpr8 def $vgpr8_vgpr9 killed $exec
+; GFX9-O0-NEXT: v_mov_b32_e32 v9, v12
; GFX9-O0-NEXT: v_mov_b32_e32 v13, s5
; GFX9-O0-NEXT: v_mov_b32_e32 v12, s4
-; GFX9-O0-NEXT: v_mov_b32_e32 v15, s5
-; GFX9-O0-NEXT: v_mov_b32_e32 v14, s4
-; GFX9-O0-NEXT: buffer_store_dword v18, off, s[0:3], s32 offset:248 ; 4-byte Folded Spill
+; GFX9-O0-NEXT: buffer_store_dword v12, off, s[0:3], s32 offset:200 ; 4-byte Folded Spill
; GFX9-O0-NEXT: s_nop 0
-; GFX9-O0-NEXT: buffer_store_dword v19, off, s[0:3], s32 offset:252 ; 4-byte Folded Spill
-; GFX9-O0-NEXT: buffer_store_dword v16, off, s[0:3], s32 offset:256 ; 4-byte Folded Spill
-; GFX9-O0-NEXT: s_nop 0
-; GFX9-O0-NEXT: buffer_store_dword v17, off, s[0:3], s32 offset:260 ; 4-byte Folded Spill
-; GFX9-O0-NEXT: v_writelane_b32 v30, s4, 8
-; GFX9-O0-NEXT: v_writelane_b32 v30, s5, 9
-; GFX9-O0-NEXT: s_or_saveexec_b64 s[18:19], -1
-; GFX9-O0-NEXT: buffer_store_dword v30, off, s[0:3], s32 ; 4-byte Folded Spill
-; GFX9-O0-NEXT: s_mov_b64 exec, s[18:19]
-; GFX9-O0-NEXT: buffer_store_dword v14, off, s[0:3], s32 offset:240 ; 4-byte Folded Spill
+; GFX9-O0-NEXT: buffer_store_dword v13, off, s[0:3], s32 offset:204 ; 4-byte Folded Spill
+; GFX9-O0-NEXT: v_mov_b32_e32 v13, s5
+; GFX9-O0-NEXT: v_mov_b32_e32 v12, s4
+; GFX9-O0-NEXT: buffer_store_dword v12, off, s[0:3], s32 offset:208 ; 4-byte Folded Spill
; GFX9-O0-NEXT: s_nop 0
-; GFX9-O0-NEXT: buffer_store_dword v15, off, s[0:3], s32 offset:244 ; 4-byte Folded Spill
-; GFX9-O0-NEXT: buffer_store_dword v12, off, s[0:3], s32 offset:232 ; 4-byte Folded Spill
+; GFX9-O0-NEXT: buffer_store_dword v13, off, s[0:3], s32 offset:212 ; 4-byte Folded Spill
+; GFX9-O0-NEXT: buffer_store_dword v10, off, s[0:3], s32 offset:248 ; 4-byte Folded Spill
; GFX9-O0-NEXT: s_nop 0
-; GFX9-O0-NEXT: buffer_store_dword v13, off, s[0:3], s32 offset:236 ; 4-byte Folded Spill
-; GFX9-O0-NEXT: buffer_store_dword v10, off, s[0:3], s32 offset:224 ; 4-byte Folded Spill
+; GFX9-O0-NEXT: buffer_store_dword v11, off, s[0:3], s32 offset:252 ; 4-byte Folded Spill
+; GFX9-O0-NEXT: buffer_store_dword v8, off, s[0:3], s32 offset:256 ; 4-byte Folded Spill
; GFX9-O0-NEXT: s_nop 0
-; GFX9-O0-NEXT: buffer_store_dword v11, off, s[0:3], s32 offset:228 ; 4-byte Folded Spill
-; GFX9-O0-NEXT: buffer_store_dword v8, off, s[0:3], s32 offset:216 ; 4-byte Folded Spill
-; GFX9-O0-NEXT: s_nop 0
-; GFX9-O0-NEXT: buffer_store_dword v9, off, s[0:3], s32 offset:220 ; 4-byte Folded Spill
-; GFX9-O0-NEXT: buffer_store_dword v6, off, s[0:3], s32 offset:208 ; 4-byte Folded Spill
+; GFX9-O0-NEXT: buffer_store_dword v9, off, s[0:3], s32 offset:260 ; 4-byte Folded Spill
+; GFX9-O0-NEXT: v_writelane_b32 v29, s4, 8
+; GFX9-O0-NEXT: v_writelane_b32 v29, s5, 9
+; GFX9-O0-NEXT: s_or_saveexec_b64 s[18:19], -1
+; GFX9-O0-NEXT: buffer_store_dword v29, off, s[0:3], s32 ; 4-byte Folded Spill
+; GFX9-O0-NEXT: s_mov_b64 exec, s[18:19]
+; GFX9-O0-NEXT: buffer_store_dword v6, off, s[0:3], s32 offset:216 ; 4-byte Folded Spill
; GFX9-O0-NEXT: s_nop 0
-; GFX9-O0-NEXT: buffer_store_dword v7, off, s[0:3], s32 offset:212 ; 4-byte Folded Spill
-; GFX9-O0-NEXT: buffer_store_dword v4, off, s[0:3], s32 offset:200 ; 4-byte Folded Spill
+; GFX9-O0-NEXT: buffer_store_dword v7, off, s[0:3], s32 offset:220 ; 4-byte Folded Spill
+; GFX9-O0-NEXT: buffer_store_dword v4, off, s[0:3], s32 offset:224 ; 4-byte Folded Spill
; GFX9-O0-NEXT: s_nop 0
-; GFX9-O0-NEXT: buffer_store_dword v5, off, s[0:3], s32 offset:204 ; 4-byte Folded Spill
+; GFX9-O0-NEXT: buffer_store_dword v5, off, s[0:3], s32 offset:228 ; 4-byte Folded Spill
; GFX9-O0-NEXT: buffer_store_dword v2, off, s[0:3], s32 offset:192 ; 4-byte Folded Spill
; GFX9-O0-NEXT: s_nop 0
; GFX9-O0-NEXT: buffer_store_dword v3, off, s[0:3], s32 offset:196 ; 4-byte Folded Spill
@@ -3210,12 +3205,12 @@ define i128 @v_udiv_i128_vv(i128 %lhs, i128 %rhs) {
; GFX9-O0-NEXT: s_branch .LBB1_6
; GFX9-O0-NEXT: .LBB1_8: ; %udiv-bb1
; GFX9-O0-NEXT: s_or_saveexec_b64 s[18:19], -1
-; GFX9-O0-NEXT: buffer_load_dword v30, off, s[0:3], s32 ; 4-byte Folded Reload
+; GFX9-O0-NEXT: buffer_load_dword v29, off, s[0:3], s32 ; 4-byte Folded Reload
; GFX9-O0-NEXT: s_mov_b64 exec, s[18:19]
-; GFX9-O0-NEXT: buffer_load_dword v6, off, s[0:3], s32 offset:36 ; 4-byte Folded Reload
-; GFX9-O0-NEXT: buffer_load_dword v7, off, s[0:3], s32 offset:40 ; 4-byte Folded Reload
-; GFX9-O0-NEXT: buffer_load_dword v10, off, s[0:3], s32 offset:44 ; 4-byte Folded Reload
-; GFX9-O0-NEXT: buffer_load_dword v11, off, s[0:3], s32 offset:48 ; 4-byte Folded Reload
+; GFX9-O0-NEXT: buffer_load_dword v9, off, s[0:3], s32 offset:36 ; 4-byte Folded Reload
+; GFX9-O0-NEXT: buffer_load_dword v10, off, s[0:3], s32 offset:40 ; 4-byte Folded Reload
+; GFX9-O0-NEXT: buffer_load_dword v11, off, s[0:3], s32 offset:44 ; 4-byte Folded Reload
+; GFX9-O0-NEXT: buffer_load_dword v12, off, s[0:3], s32 offset:48 ; 4-byte Folded Reload
; GFX9-O0-NEXT: buffer_load_dword v4, off, s[0:3], s32 offset:20 ; 4-byte Folded Reload
; GFX9-O0-NEXT: buffer_load_dword v5, off, s[0:3], s32 offset:24 ; 4-byte Folded Reload
; GFX9-O0-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:28 ; 4-byte Folded Reload
@@ -3228,118 +3223,118 @@ define i128 @v_udiv_i128_vv(i128 %lhs, i128 %rhs) {
; GFX9-O0-NEXT: s_mov_b64 s[6:7], 0
; GFX9-O0-NEXT: s_mov_b32 s8, s6
; GFX9-O0-NEXT: s_mov_b32 s9, s7
-; GFX9-O0-NEXT: v_mov_b32_e32 v0, v4
-; GFX9-O0-NEXT: v_mov_b32_e32 v2, v5
+; GFX9-O0-NEXT: v_mov_b32_e32 v2, v4
+; GFX9-O0-NEXT: v_mov_b32_e32 v0, v5
; GFX9-O0-NEXT: v_mov_b32_e32 v4, s5
-; GFX9-O0-NEXT: v_add_co_u32_e32 v8, vcc, v3, v4
+; GFX9-O0-NEXT: v_add_co_u32_e32 v7, vcc, v3, v4
; GFX9-O0-NEXT: v_mov_b32_e32 v4, s4
; GFX9-O0-NEXT: s_waitcnt vmcnt(0)
; GFX9-O0-NEXT: v_addc_co_u32_e32 v1, vcc, v1, v4, vcc
; GFX9-O0-NEXT: v_mov_b32_e32 v4, s8
-; GFX9-O0-NEXT: v_addc_co_u32_e32 v0, vcc, v0, v4, vcc
-; GFX9-O0-NEXT: v_mov_b32_e32 v4, s9
-; GFX9-O0-NEXT: v_addc_co_u32_e32 v2, vcc, v2, v4, vcc
+; GFX9-O0-NEXT: v_addc_co_u32_e32 v4, vcc, v2, v4, vcc
+; GFX9-O0-NEXT: v_mov_b32_e32 v2, s9
+; GFX9-O0-NEXT: v_addc_co_u32_e32 v0, vcc, v0, v2, vcc
; GFX9-O0-NEXT: ; implicit-def: $sgpr4
; GFX9-O0-NEXT: ; implicit-def: $sgpr4
-; GFX9-O0-NEXT: ; kill: def $vgpr8 killed $vgpr8 def $vgpr8_vgpr9 killed $exec
-; GFX9-O0-NEXT: v_mov_b32_e32 v9, v1
+; GFX9-O0-NEXT: ; kill: def $vgpr7 killed $vgpr7 def $vgpr7_vgpr8 killed $exec
+; GFX9-O0-NEXT: v_mov_b32_e32 v8, v1
; GFX9-O0-NEXT: ; implicit-def: $sgpr4
; GFX9-O0-NEXT: ; implicit-def: $sgpr4
-; GFX9-O0-NEXT: ; kill: def $vgpr0 killed $vgpr0 def $vgpr0_vgpr1 killed $exec
-; GFX9-O0-NEXT: v_mov_b32_e32 v1, v2
-; GFX9-O0-NEXT: v_mov_b32_e32 v5, v1
-; GFX9-O0-NEXT: v_mov_b32_e32 v4, v0
-; GFX9-O0-NEXT: buffer_store_dword v4, off, s[0:3], s32 offset:280 ; 4-byte Folded Spill
+; GFX9-O0-NEXT: ; kill: def $vgpr4 killed $vgpr4 def $vgpr4_vgpr5 killed $exec
+; GFX9-O0-NEXT: v_mov_b32_e32 v5, v0
+; GFX9-O0-NEXT: v_mov_b32_e32 v0, v4
+; GFX9-O0-NEXT: v_mov_b32_e32 v1, v5
+; GFX9-O0-NEXT: buffer_store_dword v0, off, s[0:3], s32 offset:280 ; 4-byte Folded Spill
; GFX9-O0-NEXT: s_nop 0
-; GFX9-O0-NEXT: buffer_store_dword v5, off, s[0:3], s32 offset:284 ; 4-byte Folded Spill
-; GFX9-O0-NEXT: v_mov_b32_e32 v4, v8
-; GFX9-O0-NEXT: v_mov_b32_e32 v5, v9
-; GFX9-O0-NEXT: buffer_store_dword v4, off, s[0:3], s32 offset:288 ; 4-byte Folded Spill
+; GFX9-O0-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:284 ; 4-byte Folded Spill
+; GFX9-O0-NEXT: v_mov_b32_e32 v0, v7
+; GFX9-O0-NEXT: v_mov_b32_e32 v1, v8
+; GFX9-O0-NEXT: buffer_store_dword v0, off, s[0:3], s32 offset:288 ; 4-byte Folded Spill
; GFX9-O0-NEXT: s_nop 0
-; GFX9-O0-NEXT: buffer_store_dword v5, off, s[0:3], s32 offset:292 ; 4-byte Folded Spill
+; GFX9-O0-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:292 ; 4-byte Folded Spill
; GFX9-O0-NEXT: s_mov_b32 s4, 0x7f
; GFX9-O0-NEXT: v_sub_u32_e64 v2, s4, v3
-; GFX9-O0-NEXT: v_lshlrev_b64 v[4:5], v2, v[10:11]
-; GFX9-O0-NEXT: v_mov_b32_e32 v12, v5
+; GFX9-O0-NEXT: v_lshlrev_b64 v[0:1], v2, v[11:12]
+; GFX9-O0-NEXT: v_mov_b32_e32 v6, v1
; GFX9-O0-NEXT: s_mov_b32 s4, 64
; GFX9-O0-NEXT: v_sub_u32_e64 v13, s4, v2
-; GFX9-O0-NEXT: v_lshrrev_b64 v[13:14], v13, v[6:7]
+; GFX9-O0-NEXT: v_lshrrev_b64 v[13:14], v13, v[9:10]
; GFX9-O0-NEXT: v_mov_b32_e32 v15, v14
-; GFX9-O0-NEXT: v_or_b32_e64 v12, v12, v15
-; GFX9-O0-NEXT: ; kill: def $vgpr4 killed $vgpr4 killed $vgpr4_vgpr5 killed $exec
-; GFX9-O0-NEXT: v_mov_b32_e32 v5, v13
-; GFX9-O0-NEXT: v_or_b32_e64 v4, v4, v5
-; GFX9-O0-NEXT: ; kill: def $vgpr4 killed $vgpr4 def $vgpr4_vgpr5 killed $exec
-; GFX9-O0-NEXT: v_mov_b32_e32 v5, v12
-; GFX9-O0-NEXT: v_mov_b32_e32 v14, v5
+; GFX9-O0-NEXT: v_or_b32_e64 v6, v6, v15
+; GFX9-O0-NEXT: ; kill: def $vgpr0 killed $vgpr0 killed $vgpr0_vgpr1 killed $exec
+; GFX9-O0-NEXT: v_mov_b32_e32 v1, v13
+; GFX9-O0-NEXT: v_or_b32_e64 v0, v0, v1
+; GFX9-O0-NEXT: ; kill: def $vgpr0 killed $vgpr0 def $vgpr0_vgpr1 killed $exec
+; GFX9-O0-NEXT: v_mov_b32_e32 v1, v6
+; GFX9-O0-NEXT: v_mov_b32_e32 v6, v1
; GFX9-O0-NEXT: v_cmp_lt_u32_e64 s[4:5], v2, s4
; GFX9-O0-NEXT: s_mov_b32 s10, 63
; GFX9-O0-NEXT: v_sub_u32_e64 v3, s10, v3
-; GFX9-O0-NEXT: v_lshlrev_b64 v[12:13], v3, v[6:7]
-; GFX9-O0-NEXT: v_mov_b32_e32 v3, v13
-; GFX9-O0-NEXT: v_cndmask_b32_e64 v3, v3, v14, s[4:5]
+; GFX9-O0-NEXT: v_lshlrev_b64 v[13:14], v3, v[9:10]
+; GFX9-O0-NEXT: v_mov_b32_e32 v3, v14
+; GFX9-O0-NEXT: v_cndmask_b32_e64 v3, v3, v6, s[4:5]
; GFX9-O0-NEXT: s_mov_b32 s10, 0
; GFX9-O0-NEXT: v_cmp_eq_u32_e64 s[10:11], v2, s10
-; GFX9-O0-NEXT: v_mov_b32_e32 v14, v11
-; GFX9-O0-NEXT: v_cndmask_b32_e64 v3, v3, v14, s[10:11]
-; GFX9-O0-NEXT: v_mov_b32_e32 v5, v4
-; GFX9-O0-NEXT: v_mov_b32_e32 v4, v12
-; GFX9-O0-NEXT: v_cndmask_b32_e64 v4, v4, v5, s[4:5]
-; GFX9-O0-NEXT: v_mov_b32_e32 v5, v10
-; GFX9-O0-NEXT: v_cndmask_b32_e64 v4, v4, v5, s[10:11]
+; GFX9-O0-NEXT: v_mov_b32_e32 v6, v12
+; GFX9-O0-NEXT: v_cndmask_b32_e64 v3, v3, v6, s[10:11]
+; GFX9-O0-NEXT: v_mov_b32_e32 v1, v0
+; GFX9-O0-NEXT: v_mov_b32_e32 v0, v13
+; GFX9-O0-NEXT: v_cndmask_b32_e64 v0, v0, v1, s[4:5]
+; GFX9-O0-NEXT: v_mov_b32_e32 v1, v11
+; GFX9-O0-NEXT: v_cndmask_b32_e64 v0, v0, v1, s[10:11]
; GFX9-O0-NEXT: ; implicit-def: $sgpr10
; GFX9-O0-NEXT: ; implicit-def: $sgpr10
-; GFX9-O0-NEXT: ; kill: def $vgpr4 killed $vgpr4 def $vgpr4_vgpr5 killed $exec
-; GFX9-O0-NEXT: v_mov_b32_e32 v5, v3
-; GFX9-O0-NEXT: v_lshlrev_b64 v[6:7], v2, v[6:7]
-; GFX9-O0-NEXT: v_mov_b32_e32 v3, v7
-; GFX9-O0-NEXT: v_mov_b32_e32 v2, s9
+; GFX9-O0-NEXT: ; kill: def $vgpr0 killed $vgpr0 def $vgpr0_vgpr1 killed $exec
+; GFX9-O0-NEXT: v_mov_b32_e32 v1, v3
+; GFX9-O0-NEXT: v_lshlrev_b64 v[2:3], v2, v[9:10]
+; GFX9-O0-NEXT: v_mov_b32_e32 v9, v3
+; GFX9-O0-NEXT: v_mov_b32_e32 v6, s9
+; GFX9-O0-NEXT: v_cndmask_b32_e64 v6, v6, v9, s[4:5]
+; GFX9-O0-NEXT: v_mov_b32_e32 v3, v2
+; GFX9-O0-NEXT: v_mov_b32_e32 v2, s8
; GFX9-O0-NEXT: v_cndmask_b32_e64 v2, v2, v3, s[4:5]
-; GFX9-O0-NEXT: ; kill: def $vgpr6 killed $vgpr6 killed $vgpr6_vgpr7 killed $exec
-; GFX9-O0-NEXT: v_mov_b32_e32 v3, s8
-; GFX9-O0-NEXT: v_cndmask_b32_e64 v6, v3, v6, s[4:5]
; GFX9-O0-NEXT: ; implicit-def: $sgpr4
; GFX9-O0-NEXT: ; implicit-def: $sgpr4
-; GFX9-O0-NEXT: ; kill: def $vgpr6 killed $vgpr6 def $vgpr6_vgpr7 killed $exec
-; GFX9-O0-NEXT: v_mov_b32_e32 v7, v2
-; GFX9-O0-NEXT: buffer_store_dword v6, off, s[0:3], s32 offset:272 ; 4-byte Folded Spill
+; GFX9-O0-NEXT: ; kill: def $vgpr2 killed $vgpr2 def $vgpr2_vgpr3 killed $exec
+; GFX9-O0-NEXT: v_mov_b32_e32 v3, v6
+; GFX9-O0-NEXT: buffer_store_dword v2, off, s[0:3], s32 offset:272 ; 4-byte Folded Spill
; GFX9-O0-NEXT: s_nop 0
-; GFX9-O0-NEXT: buffer_store_dword v7, off, s[0:3], s32 offset:276 ; 4-byte Folded Spill
-; GFX9-O0-NEXT: buffer_store_dword v4, off, s[0:3], s32 offset:264 ; 4-byte Folded Spill
+; GFX9-O0-NEXT: buffer_store_dword v3, off, s[0:3], s32 offset:276 ; 4-byte Folded Spill
+; GFX9-O0-NEXT: buffer_store_dword v0, off, s[0:3], s32 offset:264 ; 4-byte Folded Spill
; GFX9-O0-NEXT: s_nop 0
-; GFX9-O0-NEXT: buffer_store_dword v5, off, s[0:3], s32 offset:268 ; 4-byte Folded Spill
-; GFX9-O0-NEXT: v_mov_b32_e32 v3, v1
-; GFX9-O0-NEXT: v_mov_b32_e32 v2, v9
-; GFX9-O0-NEXT: v_or_b32_e64 v2, v2, v3
-; GFX9-O0-NEXT: v_mov_b32_e32 v1, v0
-; GFX9-O0-NEXT: v_mov_b32_e32 v0, v8
-; GFX9-O0-NEXT: v_or_b32_e64 v0, v0, v1
-; GFX9-O0-NEXT: ; kill: def $vgpr0 killed $vgpr0 def $vgpr0_vgpr1 killed $exec
-; GFX9-O0-NEXT: v_mov_b32_e32 v1, v2
-; GFX9-O0-NEXT: v_cmp_ne_u64_e64 s[4:5], v[0:1], s[6:7]
-; GFX9-O0-NEXT: v_mov_b32_e32 v0, s6
-; GFX9-O0-NEXT: v_mov_b32_e32 v1, s7
-; GFX9-O0-NEXT: v_mov_b32_e32 v2, s6
-; GFX9-O0-NEXT: v_mov_b32_e32 v3, s7
-; GFX9-O0-NEXT: buffer_store_dword v6, off, s[0:3], s32 offset:96 ; 4-byte Folded Spill
+; GFX9-O0-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:268 ; 4-byte Folded Spill
+; GFX9-O0-NEXT: v_mov_b32_e32 v9, v5
+; GFX9-O0-NEXT: v_mov_b32_e32 v6, v8
+; GFX9-O0-NEXT: v_or_b32_e64 v6, v6, v9
+; GFX9-O0-NEXT: v_mov_b32_e32 v5, v4
+; GFX9-O0-NEXT: v_mov_b32_e32 v4, v7
+; GFX9-O0-NEXT: v_or_b32_e64 v4, v4, v5
+; GFX9-O0-NEXT: ; kill: def $vgpr4 killed $vgpr4 def $vgpr4_vgpr5 killed $exec
+; GFX9-O0-NEXT: v_mov_b32_e32 v5, v6
+; GFX9-O0-NEXT: v_cmp_ne_u64_e64 s[4:5], v[4:5], s[6:7]
+; GFX9-O0-NEXT: v_mov_b32_e32 v4, s6
+; GFX9-O0-NEXT: v_mov_b32_e32 v5, s7
+; GFX9-O0-NEXT: buffer_store_dword v4, off, s[0:3], s32 offset:96 ; 4-byte Folded Spill
; GFX9-O0-NEXT: s_nop 0
-; GFX9-O0-NEXT: buffer_store_dword v7, off, s[0:3], s32 offset:100 ; 4-byte Folded Spill
+; GFX9-O0-NEXT: buffer_store_dword v5, off, s[0:3], s32 offset:100 ; 4-byte Folded Spill
+; GFX9-O0-NEXT: v_mov_b32_e32 v4, s6
+; GFX9-O0-NEXT: v_mov_b32_e32 v5, s7
; GFX9-O0-NEXT: buffer_store_dword v4, off, s[0:3], s32 offset:88 ; 4-byte Folded Spill
; GFX9-O0-NEXT: s_nop 0
; GFX9-O0-NEXT: buffer_store_dword v5, off, s[0:3], s32 offset:92 ; 4-byte Folded Spill
-; GFX9-O0-NEXT: buffer_store_dword v2, off, s[0:3], s32 offset:80 ; 4-byte Folded Spill
+; GFX9-O0-NEXT: buffer_store_dword v2, off, s[0:3], s32 offset:72 ; 4-byte Folded Spill
; GFX9-O0-NEXT: s_nop 0
-; GFX9-O0-NEXT: buffer_store_dword v3, off, s[0:3], s32 offset:84 ; 4-byte Folded Spill
-; GFX9-O0-NEXT: buffer_store_dword v0, off, s[0:3], s32 offset:72 ; 4-byte Folded Spill
+; GFX9-O0-NEXT: buffer_store_dword v3, off, s[0:3], s32 offset:76 ; 4-byte Folded Spill
+; GFX9-O0-NEXT: buffer_store_dword v0, off, s[0:3], s32 offset:80 ; 4-byte Folded Spill
; GFX9-O0-NEXT: s_nop 0
-; GFX9-O0-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:76 ; 4-byte Folded Spill
+; GFX9-O0-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:84 ; 4-byte Folded Spill
; GFX9-O0-NEXT: s_mov_b64 s[6:7], exec
; GFX9-O0-NEXT: s_and_b64 s[4:5], s[6:7], s[4:5]
; GFX9-O0-NEXT: s_xor_b64 s[6:7], s[4:5], s[6:7]
-; GFX9-O0-NEXT: v_writelane_b32 v30, s6, 6
-; GFX9-O0-NEXT: v_writelane_b32 v30, s7, 7
+; GFX9-O0-NEXT: v_writelane_b32 v29, s6, 6
+; GFX9-O0-NEXT: v_writelane_b32 v29, s7, 7
; GFX9-O0-NEXT: s_or_saveexec_b64 s[18:19], -1
-; GFX9-O0-NEXT: buffer_store_dword v30, off, s[0:3], s32 ; 4-byte Folded Spill
+; GFX9-O0-NEXT: buffer_store_dword v29, off, s[0:3], s32 ; 4-byte Folded Spill
; GFX9-O0-NEXT: s_mov_b64 exec, s[18:19]
; GFX9-O0-NEXT: s_mov_b64 exec, s[4:5]
; GFX9-O0-NEXT: s_cbranch_execz .LBB1_5
@@ -3358,7 +3353,7 @@ define i128 @v_udiv_i128_vv(i128 %lhs, i128 %rhs) {
; GFX9-O0-NEXT: v_mov_b32_e32 v0, v6
; GFX9-O0-NEXT: v_mov_b32_e32 v2, v4
; GFX9-O0-NEXT: s_xor_saveexec_b64 s[4:5], -1
-; GFX9-O0-NEXT: buffer_load_dword v30, off, s[0:3], s32 offset:296 ; 4-byte Folded Reload
+; GFX9-O0-NEXT: buffer_load_dword v29, off, s[0:3], s32 offset:296 ; 4-byte Folded Reload
; GFX9-O0-NEXT: s_mov_b64 exec, s[4:5]
; GFX9-O0-NEXT: s_waitcnt vmcnt(0)
; GFX9-O0-NEXT: s_setpc_b64 s[30:31]
@@ -3768,29 +3763,26 @@ define i128 @v_udiv_i128_vv(i128 %lhs, i128 %rhs) {
; GFX9-G-O0-NEXT: v_readlane_b32 s5, v32, 3
; GFX9-G-O0-NEXT: s_or_b64 exec, exec, s[4:5]
; GFX9-G-O0-NEXT: ; %bb.2: ; %Flow
-; GFX9-G-O0-NEXT: buffer_load_dword v4, off, s[0:3], s32 offset:100 ; 4-byte Folded Reload
-; GFX9-G-O0-NEXT: buffer_load_dword v5, off, s[0:3], s32 offset:104 ; 4-byte Folded Reload
-; GFX9-G-O0-NEXT: buffer_load_dword v6, off, s[0:3], s32 offset:108 ; 4-byte Folded Reload
-; GFX9-G-O0-NEXT: buffer_load_dword v7, off, s[0:3], s32 offset:112 ; 4-byte Folded Reload
-; GFX9-G-O0-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:116 ; 4-byte Folded Reload
-; GFX9-G-O0-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:120 ; 4-byte Folded Reload
-; GFX9-G-O0-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:124 ; 4-byte Folded Reload
-; GFX9-G-O0-NEXT: buffer_load_dword v3, off, s[0:3], s32 offset:128 ; 4-byte Folded Reload
-; GFX9-G-O0-NEXT: s_waitcnt vmcnt(7)
+; GFX9-G-O0-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:100 ; 4-byte Folded Reload
+; GFX9-G-O0-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:104 ; 4-byte Folded Reload
+; GFX9-G-O0-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:108 ; 4-byte Folded Reload
+; GFX9-G-O0-NEXT: buffer_load_dword v3, off, s[0:3], s32 offset:112 ; 4-byte Folded Reload
+; GFX9-G-O0-NEXT: buffer_load_dword v4, off, s[0:3], s32 offset:116 ; 4-byte Folded Reload
+; GFX9-G-O0-NEXT: buffer_load_dword v5, off, s[0:3], s32 offset:120 ; 4-byte Folded Reload
+; GFX9-G-O0-NEXT: buffer_load_dword v6, off, s[0:3], s32 offset:124 ; 4-byte Folded Reload
+; GFX9-G-O0-NEXT: buffer_load_dword v7, off, s[0:3], s32 offset:128 ; 4-byte Folded Reload
+; GFX9-G-O0-NEXT: s_waitcnt vmcnt(3)
; GFX9-G-O0-NEXT: buffer_store_dword v4, off, s[0:3], s32 offset:84 ; 4-byte Folded Spill
-; GFX9-G-O0-NEXT: s_waitcnt vmcnt(7)
+; GFX9-G-O0-NEXT: s_waitcnt vmcnt(3)
; GFX9-G-O0-NEXT: buffer_store_dword v5, off, s[0:3], s32 offset:88 ; 4-byte Folded Spill
-; GFX9-G-O0-NEXT: s_waitcnt vmcnt(7)
+; GFX9-G-O0-NEXT: s_waitcnt vmcnt(3)
; GFX9-G-O0-NEXT: buffer_store_dword v6, off, s[0:3], s32 offset:92 ; 4-byte Folded Spill
-; GFX9-G-O0-NEXT: s_waitcnt vmcnt(7)
+; GFX9-G-O0-NEXT: s_waitcnt vmcnt(3)
; GFX9-G-O0-NEXT: buffer_store_dword v7, off, s[0:3], s32 offset:96 ; 4-byte Folded Spill
-; GFX9-G-O0-NEXT: s_waitcnt vmcnt(7)
; GFX9-G-O0-NEXT: buffer_store_dword v0, off, s[0:3], s32 offset:68 ; 4-byte Folded Spill
-; GFX9-G-O0-NEXT: s_waitcnt vmcnt(7)
+; GFX9-G-O0-NEXT: s_nop 0
; GFX9-G-O0-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:72 ; 4-byte Folded Spill
-; GFX9-G-O0-NEXT: s_waitcnt vmcnt(7)
; GFX9-G-O0-NEXT: buffer_store_dword v2, off, s[0:3], s32 offset:76 ; 4-byte Folded Spill
-; GFX9-G-O0-NEXT: s_waitcnt vmcnt(7)
; GFX9-G-O0-NEXT: buffer_store_dword v3, off, s[0:3], s32 offset:80 ; 4-byte Folded Spill
; GFX9-G-O0-NEXT: s_branch .LBB1_5
; GFX9-G-O0-NEXT: .LBB1_3: ; %Flow2
@@ -3876,14 +3868,14 @@ define i128 @v_udiv_i128_vv(i128 %lhs, i128 %rhs) {
; GFX9-G-O0-NEXT: v_readlane_b32 s4, v32, 4
; GFX9-G-O0-NEXT: v_readlane_b32 s5, v32, 5
; GFX9-G-O0-NEXT: s_or_b64 exec, exec, s[4:5]
-; GFX9-G-O0-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:84 ; 4-byte Folded Reload
-; GFX9-G-O0-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:88 ; 4-byte Folded Reload
-; GFX9-G-O0-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:92 ; 4-byte Folded Reload
-; GFX9-G-O0-NEXT: buffer_load_dword v3, off, s[0:3], s32 offset:96 ; 4-byte Folded Reload
-; GFX9-G-O0-NEXT: buffer_load_dword v4, off, s[0:3], s32 offset:68 ; 4-byte Folded Reload
-; GFX9-G-O0-NEXT: buffer_load_dword v5, off, s[0:3], s32 offset:72 ; 4-byte Folded Reload
-; GFX9-G-O0-NEXT: buffer_load_dword v6, off, s[0:3], s32 offset:76 ; 4-byte Folded Reload
-; GFX9-G-O0-NEXT: buffer_load_dword v7, off, s[0:3], s32 offset:80 ; 4-byte Folded Reload
+; GFX9-G-O0-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:68 ; 4-byte Folded Reload
+; GFX9-G-O0-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:72 ; 4-byte Folded Reload
+; GFX9-G-O0-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:76 ; 4-byte Folded Reload
+; GFX9-G-O0-NEXT: buffer_load_dword v3, off, s[0:3], s32 offset:80 ; 4-byte Folded Reload
+; GFX9-G-O0-NEXT: buffer_load_dword v4, off, s[0:3], s32 offset:84 ; 4-byte Folded Reload
+; GFX9-G-O0-NEXT: buffer_load_dword v5, off, s[0:3], s32 offset:88 ; 4-byte Folded Reload
+; GFX9-G-O0-NEXT: buffer_load_dword v6, off, s[0:3], s32 offset:92 ; 4-byte Folded Reload
+; GFX9-G-O0-NEXT: buffer_load_dword v7, off, s[0:3], s32 offset:96 ; 4-byte Folded Reload
; GFX9-G-O0-NEXT: s_waitcnt vmcnt(3)
; GFX9-G-O0-NEXT: buffer_store_dword v4, off, s[0:3], s32 offset:148 ; 4-byte Folded Spill
; GFX9-G-O0-NEXT: s_waitcnt vmcnt(3)
@@ -4339,20 +4331,20 @@ define i128 @v_udiv_i128_vv(i128 %lhs, i128 %rhs) {
; GFX9-G-O0-NEXT: v_mov_b32_e32 v7, s5
; GFX9-G-O0-NEXT: v_mov_b32_e32 v6, s4
; GFX9-G-O0-NEXT: v_cmp_ne_u64_e64 s[4:5], v[4:5], v[6:7]
-; GFX9-G-O0-NEXT: buffer_store_dword v0, off, s[0:3], s32 offset:84 ; 4-byte Folded Spill
+; GFX9-G-O0-NEXT: buffer_store_dword v0, off, s[0:3], s32 offset:68 ; 4-byte Folded Spill
; GFX9-G-O0-NEXT: s_nop 0
-; GFX9-G-O0-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:88 ; 4-byte Folded Spill
-; GFX9-G-O0-NEXT: buffer_store_dword v2, off, s[0:3], s32 offset:92 ; 4-byte Folded Spill
-; GFX9-G-O0-NEXT: buffer_store_dword v3, off, s[0:3], s32 offset:96 ; 4-byte Folded Spill
+; GFX9-G-O0-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:72 ; 4-byte Folded Spill
+; GFX9-G-O0-NEXT: buffer_store_dword v2, off, s[0:3], s32 offset:76 ; 4-byte Folded Spill
+; GFX9-G-O0-NEXT: buffer_store_dword v3, off, s[0:3], s32 offset:80 ; 4-byte Folded Spill
; GFX9-G-O0-NEXT: v_mov_b32_e32 v0, s8
; GFX9-G-O0-NEXT: v_mov_b32_e32 v1, s9
; GFX9-G-O0-NEXT: v_mov_b32_e32 v2, s10
; GFX9-G-O0-NEXT: v_mov_b32_e32 v3, s11
-; GFX9-G-O0-NEXT: buffer_store_dword v0, off, s[0:3], s32 offset:68 ; 4-byte Folded Spill
+; GFX9-G-O0-NEXT: buffer_store_dword v0, off, s[0:3], s32 offset:84 ; 4-byte Folded Spill
; GFX9-G-O0-NEXT: s_nop 0
-; GFX9-G-O0-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:72 ; 4-byte Folded Spill
-; GFX9-G-O0-NEXT: buffer_store_dword v2, off, s[0:3], s32 offset:76 ; 4-byte Folded Spill
-; GFX9-G-O0-NEXT: buffer_store_dword v3, off, s[0:3], s32 offset:80 ; 4-byte Folded Spill
+; GFX9-G-O0-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:88 ; 4-byte Folded Spill
+; GFX9-G-O0-NEXT: buffer_store_dword v2, off, s[0:3], s32 offset:92 ; 4-byte Folded Spill
+; GFX9-G-O0-NEXT: buffer_store_dword v3, off, s[0:3], s32 offset:96 ; 4-byte Folded Spill
; GFX9-G-O0-NEXT: s_mov_b64 s[6:7], exec
; GFX9-G-O0-NEXT: s_and_b64 s[4:5], s[6:7], s[4:5]
; GFX9-G-O0-NEXT: s_xor_b64 s[6:7], s[4:5], s[6:7]
diff --git a/llvm/test/CodeGen/AMDGPU/div_v2i128.ll b/llvm/test/CodeGen/AMDGPU/div_v2i128.ll
index 77b78f1f8a333..de439c6f46c6e 100644
--- a/llvm/test/CodeGen/AMDGPU/div_v2i128.ll
+++ b/llvm/test/CodeGen/AMDGPU/div_v2i128.ll
@@ -7,37 +7,37 @@ define <2 x i128> @v_sdiv_v2i128_vv(<2 x i128> %lhs, <2 x i128> %rhs) {
; SDAG: ; %bb.0: ; %_udiv-special-cases_udiv-special-cases
; SDAG-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; SDAG-NEXT: v_sub_i32_e32 v16, vcc, 0, v0
-; SDAG-NEXT: v_mov_b32_e32 v18, 0
+; SDAG-NEXT: v_mov_b32_e32 v20, 0
; SDAG-NEXT: v_ashrrev_i32_e32 v24, 31, v3
; SDAG-NEXT: v_ashrrev_i32_e32 v25, 31, v11
; SDAG-NEXT: s_mov_b64 s[10:11], 0x7f
; SDAG-NEXT: v_subb_u32_e32 v17, vcc, 0, v1, vcc
; SDAG-NEXT: v_mov_b32_e32 v26, v24
; SDAG-NEXT: v_mov_b32_e32 v27, v25
-; SDAG-NEXT: v_subb_u32_e32 v19, vcc, 0, v2, vcc
+; SDAG-NEXT: v_subb_u32_e32 v21, vcc, 0, v2, vcc
; SDAG-NEXT: v_cmp_gt_i64_e64 s[4:5], 0, v[2:3]
-; SDAG-NEXT: v_cndmask_b32_e64 v21, v1, v17, s[4:5]
-; SDAG-NEXT: v_cndmask_b32_e64 v20, v0, v16, s[4:5]
+; SDAG-NEXT: v_cndmask_b32_e64 v19, v1, v17, s[4:5]
+; SDAG-NEXT: v_cndmask_b32_e64 v18, v0, v16, s[4:5]
; SDAG-NEXT: v_subb_u32_e32 v0, vcc, 0, v3, vcc
-; SDAG-NEXT: v_cndmask_b32_e64 v16, v2, v19, s[4:5]
-; SDAG-NEXT: v_ffbh_u32_e32 v1, v20
-; SDAG-NEXT: v_ffbh_u32_e32 v2, v21
+; SDAG-NEXT: v_cndmask_b32_e64 v16, v2, v21, s[4:5]
+; SDAG-NEXT: v_ffbh_u32_e32 v1, v18
+; SDAG-NEXT: v_ffbh_u32_e32 v2, v19
; SDAG-NEXT: v_cndmask_b32_e64 v17, v3, v0, s[4:5]
-; SDAG-NEXT: v_or_b32_e32 v0, v20, v16
+; SDAG-NEXT: v_or_b32_e32 v0, v18, v16
; SDAG-NEXT: v_sub_i32_e32 v3, vcc, 0, v8
-; SDAG-NEXT: v_add_i32_e64 v19, s[4:5], 32, v1
+; SDAG-NEXT: v_add_i32_e64 v21, s[4:5], 32, v1
; SDAG-NEXT: v_ffbh_u32_e32 v22, v16
-; SDAG-NEXT: v_or_b32_e32 v1, v21, v17
+; SDAG-NEXT: v_or_b32_e32 v1, v19, v17
; SDAG-NEXT: v_subb_u32_e32 v23, vcc, 0, v9, vcc
-; SDAG-NEXT: v_min_u32_e32 v2, v19, v2
-; SDAG-NEXT: v_add_i32_e64 v19, s[4:5], 32, v22
+; SDAG-NEXT: v_min_u32_e32 v2, v21, v2
+; SDAG-NEXT: v_add_i32_e64 v21, s[4:5], 32, v22
; SDAG-NEXT: v_ffbh_u32_e32 v22, v17
; SDAG-NEXT: v_cmp_eq_u64_e64 s[4:5], 0, v[0:1]
; SDAG-NEXT: v_cmp_gt_i64_e64 s[6:7], 0, v[10:11]
; SDAG-NEXT: v_cndmask_b32_e64 v28, v9, v23, s[6:7]
; SDAG-NEXT: v_subb_u32_e32 v0, vcc, 0, v10, vcc
; SDAG-NEXT: v_cndmask_b32_e64 v29, v8, v3, s[6:7]
-; SDAG-NEXT: v_min_u32_e32 v1, v19, v22
+; SDAG-NEXT: v_min_u32_e32 v1, v21, v22
; SDAG-NEXT: v_add_i32_e64 v2, s[8:9], 64, v2
; SDAG-NEXT: v_addc_u32_e64 v3, s[8:9], 0, 0, s[8:9]
; SDAG-NEXT: v_subb_u32_e32 v8, vcc, 0, v11, vcc
@@ -46,17 +46,17 @@ define <2 x i128> @v_sdiv_v2i128_vv(<2 x i128> %lhs, <2 x i128> %rhs) {
; SDAG-NEXT: v_cndmask_b32_e64 v9, v3, 0, vcc
; SDAG-NEXT: v_cndmask_b32_e32 v10, v2, v1, vcc
; SDAG-NEXT: v_ffbh_u32_e32 v3, v29
-; SDAG-NEXT: v_ffbh_u32_e32 v19, v28
+; SDAG-NEXT: v_ffbh_u32_e32 v21, v28
; SDAG-NEXT: v_cndmask_b32_e64 v1, v11, v8, s[6:7]
; SDAG-NEXT: v_or_b32_e32 v2, v29, v0
; SDAG-NEXT: v_add_i32_e32 v8, vcc, 32, v3
; SDAG-NEXT: v_ffbh_u32_e32 v11, v0
; SDAG-NEXT: v_or_b32_e32 v3, v28, v1
-; SDAG-NEXT: v_min_u32_e32 v8, v8, v19
+; SDAG-NEXT: v_min_u32_e32 v8, v8, v21
; SDAG-NEXT: v_add_i32_e32 v11, vcc, 32, v11
-; SDAG-NEXT: v_ffbh_u32_e32 v19, v1
+; SDAG-NEXT: v_ffbh_u32_e32 v21, v1
; SDAG-NEXT: v_cmp_eq_u64_e32 vcc, 0, v[2:3]
-; SDAG-NEXT: v_min_u32_e32 v2, v11, v19
+; SDAG-NEXT: v_min_u32_e32 v2, v11, v21
; SDAG-NEXT: v_add_i32_e64 v3, s[6:7], 64, v8
; SDAG-NEXT: v_addc_u32_e64 v8, s[6:7], 0, 0, s[6:7]
; SDAG-NEXT: v_cmp_ne_u64_e64 s[6:7], 0, v[0:1]
@@ -66,35 +66,35 @@ define <2 x i128> @v_sdiv_v2i128_vv(<2 x i128> %lhs, <2 x i128> %rhs) {
; SDAG-NEXT: v_sub_i32_e32 v2, vcc, v2, v10
; SDAG-NEXT: v_subb_u32_e32 v3, vcc, v8, v9, vcc
; SDAG-NEXT: v_xor_b32_e32 v8, 0x7f, v2
-; SDAG-NEXT: v_subb_u32_e32 v10, vcc, 0, v18, vcc
+; SDAG-NEXT: v_subb_u32_e32 v10, vcc, 0, v20, vcc
; SDAG-NEXT: v_cmp_lt_u64_e64 s[4:5], s[10:11], v[2:3]
-; SDAG-NEXT: v_cndmask_b32_e64 v19, 0, 1, s[4:5]
-; SDAG-NEXT: v_subb_u32_e32 v11, vcc, 0, v18, vcc
+; SDAG-NEXT: v_cndmask_b32_e64 v21, 0, 1, s[4:5]
+; SDAG-NEXT: v_subb_u32_e32 v11, vcc, 0, v20, vcc
; SDAG-NEXT: v_or_b32_e32 v8, v8, v10
; SDAG-NEXT: v_or_b32_e32 v9, v3, v11
; SDAG-NEXT: v_cmp_ne_u64_e32 vcc, 0, v[10:11]
-; SDAG-NEXT: v_cndmask_b32_e64 v18, 0, 1, vcc
+; SDAG-NEXT: v_cndmask_b32_e64 v20, 0, 1, vcc
; SDAG-NEXT: v_cmp_ne_u64_e32 vcc, 0, v[8:9]
; SDAG-NEXT: v_cmp_eq_u64_e64 s[4:5], 0, v[10:11]
-; SDAG-NEXT: v_cndmask_b32_e64 v8, v18, v19, s[4:5]
+; SDAG-NEXT: v_cndmask_b32_e64 v8, v20, v21, s[4:5]
; SDAG-NEXT: v_and_b32_e32 v8, 1, v8
; SDAG-NEXT: v_cmp_eq_u32_e64 s[4:5], 1, v8
; SDAG-NEXT: s_or_b64 s[4:5], s[6:7], s[4:5]
-; SDAG-NEXT: v_cndmask_b32_e64 v18, v17, 0, s[4:5]
+; SDAG-NEXT: v_cndmask_b32_e64 v20, v17, 0, s[4:5]
; SDAG-NEXT: s_xor_b64 s[6:7], s[4:5], -1
; SDAG-NEXT: v_cndmask_b32_e64 v22, v16, 0, s[4:5]
-; SDAG-NEXT: v_cndmask_b32_e64 v19, v21, 0, s[4:5]
+; SDAG-NEXT: v_cndmask_b32_e64 v21, v19, 0, s[4:5]
; SDAG-NEXT: s_and_b64 s[8:9], s[6:7], vcc
-; SDAG-NEXT: v_cndmask_b32_e64 v23, v20, 0, s[4:5]
+; SDAG-NEXT: v_cndmask_b32_e64 v23, v18, 0, s[4:5]
; SDAG-NEXT: s_and_saveexec_b64 s[6:7], s[8:9]
; SDAG-NEXT: s_cbranch_execz .LBB0_6
; SDAG-NEXT: ; %bb.1: ; %udiv-bb15
; SDAG-NEXT: v_add_i32_e32 v30, vcc, 1, v2
-; SDAG-NEXT: v_sub_i32_e64 v18, s[4:5], 63, v2
+; SDAG-NEXT: v_sub_i32_e64 v20, s[4:5], 63, v2
; SDAG-NEXT: v_mov_b32_e32 v8, 0
; SDAG-NEXT: v_mov_b32_e32 v9, 0
; SDAG-NEXT: v_addc_u32_e32 v31, vcc, 0, v3, vcc
-; SDAG-NEXT: v_lshl_b64 v[18:19], v[20:21], v18
+; SDAG-NEXT: v_lshl_b64 v[20:21], v[18:19], v20
; SDAG-NEXT: v_addc_u32_e32 v32, vcc, 0, v10, vcc
; SDAG-NEXT: v_addc_u32_e32 v33, vcc, 0, v11, vcc
; SDAG-NEXT: v_or_b32_e32 v10, v30, v32
@@ -102,16 +102,16 @@ define <2 x i128> @v_sdiv_v2i128_vv(<2 x i128> %lhs, <2 x i128> %rhs) {
; SDAG-NEXT: v_or_b32_e32 v11, v31, v33
; SDAG-NEXT: v_lshl_b64 v[2:3], v[16:17], v34
; SDAG-NEXT: v_sub_i32_e32 v35, vcc, 64, v34
-; SDAG-NEXT: v_lshl_b64 v[22:23], v[20:21], v34
+; SDAG-NEXT: v_lshl_b64 v[22:23], v[18:19], v34
; SDAG-NEXT: v_cmp_ne_u64_e32 vcc, 0, v[10:11]
-; SDAG-NEXT: v_lshr_b64 v[10:11], v[20:21], v35
+; SDAG-NEXT: v_lshr_b64 v[10:11], v[18:19], v35
; SDAG-NEXT: v_or_b32_e32 v3, v3, v11
; SDAG-NEXT: v_or_b32_e32 v2, v2, v10
; SDAG-NEXT: v_cmp_gt_u32_e64 s[4:5], 64, v34
-; SDAG-NEXT: v_cndmask_b32_e64 v3, v19, v3, s[4:5]
-; SDAG-NEXT: v_cndmask_b32_e64 v2, v18, v2, s[4:5]
-; SDAG-NEXT: v_cndmask_b32_e64 v19, 0, v23, s[4:5]
-; SDAG-NEXT: v_cndmask_b32_e64 v18, 0, v22, s[4:5]
+; SDAG-NEXT: v_cndmask_b32_e64 v3, v21, v3, s[4:5]
+; SDAG-NEXT: v_cndmask_b32_e64 v2, v20, v2, s[4:5]
+; SDAG-NEXT: v_cndmask_b32_e64 v21, 0, v23, s[4:5]
+; SDAG-NEXT: v_cndmask_b32_e64 v20, 0, v22, s[4:5]
; SDAG-NEXT: v_cmp_eq_u32_e64 s[4:5], 0, v34
; SDAG-NEXT: v_cndmask_b32_e64 v3, v3, v17, s[4:5]
; SDAG-NEXT: v_cndmask_b32_e64 v2, v2, v16, s[4:5]
@@ -121,7 +121,7 @@ define <2 x i128> @v_sdiv_v2i128_vv(<2 x i128> %lhs, <2 x i128> %rhs) {
; SDAG-NEXT: s_xor_b64 s[8:9], exec, s[4:5]
; SDAG-NEXT: s_cbranch_execz .LBB0_5
; SDAG-NEXT: ; %bb.2: ; %udiv-preheader4
-; SDAG-NEXT: v_lshr_b64 v[8:9], v[20:21], v30
+; SDAG-NEXT: v_lshr_b64 v[8:9], v[18:19], v30
; SDAG-NEXT: v_sub_i32_e32 v10, vcc, 64, v30
; SDAG-NEXT: v_lshl_b64 v[10:11], v[16:17], v10
; SDAG-NEXT: v_or_b32_e32 v11, v9, v11
@@ -131,9 +131,9 @@ define <2 x i128> @v_sdiv_v2i128_vv(<2 x i128> %lhs, <2 x i128> %rhs) {
; SDAG-NEXT: v_lshr_b64 v[8:9], v[16:17], v8
; SDAG-NEXT: v_cndmask_b32_e32 v9, v9, v11, vcc
; SDAG-NEXT: v_cmp_eq_u32_e64 s[4:5], 0, v30
-; SDAG-NEXT: v_cndmask_b32_e64 v21, v9, v21, s[4:5]
+; SDAG-NEXT: v_cndmask_b32_e64 v19, v9, v19, s[4:5]
; SDAG-NEXT: v_cndmask_b32_e32 v8, v8, v10, vcc
-; SDAG-NEXT: v_cndmask_b32_e64 v20, v8, v20, s[4:5]
+; SDAG-NEXT: v_cndmask_b32_e64 v18, v8, v18, s[4:5]
; SDAG-NEXT: v_lshr_b64 v[8:9], v[16:17], v30
; SDAG-NEXT: v_cndmask_b32_e32 v23, 0, v9, vcc
; SDAG-NEXT: v_cndmask_b32_e32 v22, 0, v8, vcc
@@ -149,30 +149,30 @@ define <2 x i128> @v_sdiv_v2i128_vv(<2 x i128> %lhs, <2 x i128> %rhs) {
; SDAG-NEXT: v_mov_b32_e32 v9, 0
; SDAG-NEXT: .LBB0_3: ; %udiv-do-while3
; SDAG-NEXT: ; =>This Inner Loop Header: Depth=1
-; SDAG-NEXT: v_lshrrev_b32_e32 v8, 31, v19
-; SDAG-NEXT: v_lshl_b64 v[18:19], v[18:19], 1
-; SDAG-NEXT: v_lshl_b64 v[22:23], v[22:23], 1
-; SDAG-NEXT: v_lshrrev_b32_e32 v38, 31, v21
+; SDAG-NEXT: v_lshrrev_b32_e32 v8, 31, v21
; SDAG-NEXT: v_lshl_b64 v[20:21], v[20:21], 1
+; SDAG-NEXT: v_lshl_b64 v[22:23], v[22:23], 1
+; SDAG-NEXT: v_lshrrev_b32_e32 v38, 31, v19
+; SDAG-NEXT: v_lshl_b64 v[18:19], v[18:19], 1
; SDAG-NEXT: v_lshrrev_b32_e32 v39, 31, v3
; SDAG-NEXT: v_lshl_b64 v[2:3], v[2:3], 1
-; SDAG-NEXT: v_or_b32_e32 v19, v17, v19
-; SDAG-NEXT: v_or_b32_e32 v18, v16, v18
+; SDAG-NEXT: v_or_b32_e32 v21, v17, v21
+; SDAG-NEXT: v_or_b32_e32 v20, v16, v20
; SDAG-NEXT: v_or_b32_e32 v16, v22, v38
-; SDAG-NEXT: v_or_b32_e32 v17, v20, v39
+; SDAG-NEXT: v_or_b32_e32 v17, v18, v39
; SDAG-NEXT: v_or_b32_e32 v2, v2, v8
; SDAG-NEXT: v_sub_i32_e32 v8, vcc, v34, v17
-; SDAG-NEXT: v_subb_u32_e32 v8, vcc, v35, v21, vcc
+; SDAG-NEXT: v_subb_u32_e32 v8, vcc, v35, v19, vcc
; SDAG-NEXT: v_subb_u32_e32 v8, vcc, v36, v16, vcc
; SDAG-NEXT: v_subb_u32_e32 v8, vcc, v37, v23, vcc
; SDAG-NEXT: v_ashrrev_i32_e32 v8, 31, v8
-; SDAG-NEXT: v_and_b32_e32 v20, v8, v29
+; SDAG-NEXT: v_and_b32_e32 v18, v8, v29
; SDAG-NEXT: v_and_b32_e32 v22, v8, v28
; SDAG-NEXT: v_and_b32_e32 v38, v8, v0
; SDAG-NEXT: v_and_b32_e32 v39, v8, v1
; SDAG-NEXT: v_and_b32_e32 v8, 1, v8
-; SDAG-NEXT: v_sub_i32_e32 v20, vcc, v17, v20
-; SDAG-NEXT: v_subb_u32_e32 v21, vcc, v21, v22, vcc
+; SDAG-NEXT: v_sub_i32_e32 v18, vcc, v17, v18
+; SDAG-NEXT: v_subb_u32_e32 v19, vcc, v19, v22, vcc
; SDAG-NEXT: v_subb_u32_e32 v22, vcc, v16, v38, vcc
; SDAG-NEXT: v_subb_u32_e32 v23, vcc, v23, v39, vcc
; SDAG-NEXT: v_add_i32_e32 v30, vcc, -1, v30
@@ -194,11 +194,11 @@ define <2 x i128> @v_sdiv_v2i128_vv(<2 x i128> %lhs, <2 x i128> %rhs) {
; SDAG-NEXT: .LBB0_5: ; %Flow14
; SDAG-NEXT: s_or_b64 exec, exec, s[8:9]
; SDAG-NEXT: v_lshl_b64 v[0:1], v[2:3], 1
-; SDAG-NEXT: v_lshrrev_b32_e32 v16, 31, v19
-; SDAG-NEXT: v_lshl_b64 v[2:3], v[18:19], 1
+; SDAG-NEXT: v_lshrrev_b32_e32 v16, 31, v21
+; SDAG-NEXT: v_lshl_b64 v[2:3], v[20:21], 1
; SDAG-NEXT: v_or_b32_e32 v0, v0, v16
-; SDAG-NEXT: v_or_b32_e32 v18, v11, v1
-; SDAG-NEXT: v_or_b32_e32 v19, v9, v3
+; SDAG-NEXT: v_or_b32_e32 v20, v11, v1
+; SDAG-NEXT: v_or_b32_e32 v21, v9, v3
; SDAG-NEXT: v_or_b32_e32 v22, v10, v0
; SDAG-NEXT: v_or_b32_e32 v23, v8, v2
; SDAG-NEXT: .LBB0_6: ; %Flow16
@@ -208,110 +208,110 @@ define <2 x i128> @v_sdiv_v2i128_vv(<2 x i128> %lhs, <2 x i128> %rhs) {
; SDAG-NEXT: v_sub_i32_e32 v0, vcc, 0, v4
; SDAG-NEXT: v_mov_b32_e32 v8, 0
; SDAG-NEXT: s_mov_b64 s[10:11], 0x7f
-; SDAG-NEXT: v_mov_b32_e32 v20, v16
-; SDAG-NEXT: v_mov_b32_e32 v21, v17
+; SDAG-NEXT: v_mov_b32_e32 v18, v16
+; SDAG-NEXT: v_mov_b32_e32 v19, v17
; SDAG-NEXT: v_subb_u32_e32 v1, vcc, 0, v5, vcc
; SDAG-NEXT: v_subb_u32_e32 v9, vcc, 0, v6, vcc
; SDAG-NEXT: v_cmp_gt_i64_e64 s[4:5], 0, v[6:7]
; SDAG-NEXT: v_cndmask_b32_e64 v3, v5, v1, s[4:5]
; SDAG-NEXT: v_cndmask_b32_e64 v2, v4, v0, s[4:5]
; SDAG-NEXT: v_subb_u32_e32 v0, vcc, 0, v7, vcc
-; SDAG-NEXT: v_cndmask_b32_e64 v6, v6, v9, s[4:5]
+; SDAG-NEXT: v_cndmask_b32_e64 v4, v6, v9, s[4:5]
; SDAG-NEXT: v_ffbh_u32_e32 v1, v2
-; SDAG-NEXT: v_ffbh_u32_e32 v4, v3
-; SDAG-NEXT: v_cndmask_b32_e64 v7, v7, v0, s[4:5]
-; SDAG-NEXT: v_sub_i32_e32 v5, vcc, 0, v12
-; SDAG-NEXT: v_or_b32_e32 v0, v2, v6
-; SDAG-NEXT: v_ffbh_u32_e32 v9, v6
+; SDAG-NEXT: v_ffbh_u32_e32 v6, v3
+; SDAG-NEXT: v_cndmask_b32_e64 v5, v7, v0, s[4:5]
+; SDAG-NEXT: v_sub_i32_e32 v7, vcc, 0, v12
+; SDAG-NEXT: v_or_b32_e32 v0, v2, v4
+; SDAG-NEXT: v_ffbh_u32_e32 v9, v4
; SDAG-NEXT: v_add_i32_e64 v10, s[4:5], 32, v1
; SDAG-NEXT: v_subb_u32_e32 v11, vcc, 0, v13, vcc
-; SDAG-NEXT: v_or_b32_e32 v1, v3, v7
+; SDAG-NEXT: v_or_b32_e32 v1, v3, v5
; SDAG-NEXT: v_add_i32_e64 v9, s[4:5], 32, v9
-; SDAG-NEXT: v_ffbh_u32_e32 v30, v7
-; SDAG-NEXT: v_min_u32_e32 v4, v10, v4
+; SDAG-NEXT: v_ffbh_u32_e32 v30, v5
+; SDAG-NEXT: v_min_u32_e32 v6, v10, v6
; SDAG-NEXT: v_subb_u32_e32 v10, vcc, 0, v14, vcc
; SDAG-NEXT: v_cmp_gt_i64_e64 s[4:5], 0, v[14:15]
; SDAG-NEXT: v_cndmask_b32_e64 v28, v13, v11, s[4:5]
-; SDAG-NEXT: v_cndmask_b32_e64 v29, v12, v5, s[4:5]
+; SDAG-NEXT: v_cndmask_b32_e64 v29, v12, v7, s[4:5]
; SDAG-NEXT: v_cmp_eq_u64_e64 s[6:7], 0, v[0:1]
; SDAG-NEXT: v_min_u32_e32 v1, v9, v30
-; SDAG-NEXT: v_add_i32_e64 v4, s[8:9], 64, v4
-; SDAG-NEXT: v_addc_u32_e64 v5, s[8:9], 0, 0, s[8:9]
+; SDAG-NEXT: v_add_i32_e64 v6, s[8:9], 64, v6
+; SDAG-NEXT: v_addc_u32_e64 v7, s[8:9], 0, 0, s[8:9]
; SDAG-NEXT: v_subb_u32_e32 v9, vcc, 0, v15, vcc
; SDAG-NEXT: v_cndmask_b32_e64 v0, v14, v10, s[4:5]
; SDAG-NEXT: v_ffbh_u32_e32 v10, v29
; SDAG-NEXT: v_ffbh_u32_e32 v11, v28
-; SDAG-NEXT: v_cmp_ne_u64_e32 vcc, 0, v[6:7]
-; SDAG-NEXT: v_cndmask_b32_e64 v12, v5, 0, vcc
-; SDAG-NEXT: v_cndmask_b32_e32 v13, v4, v1, vcc
+; SDAG-NEXT: v_cmp_ne_u64_e32 vcc, 0, v[4:5]
+; SDAG-NEXT: v_cndmask_b32_e64 v12, v7, 0, vcc
+; SDAG-NEXT: v_cndmask_b32_e32 v13, v6, v1, vcc
; SDAG-NEXT: v_cndmask_b32_e64 v1, v15, v9, s[4:5]
-; SDAG-NEXT: v_or_b32_e32 v4, v29, v0
+; SDAG-NEXT: v_or_b32_e32 v6, v29, v0
; SDAG-NEXT: v_ffbh_u32_e32 v9, v0
; SDAG-NEXT: v_add_i32_e32 v10, vcc, 32, v10
-; SDAG-NEXT: v_or_b32_e32 v5, v28, v1
+; SDAG-NEXT: v_or_b32_e32 v7, v28, v1
; SDAG-NEXT: v_add_i32_e32 v9, vcc, 32, v9
; SDAG-NEXT: v_ffbh_u32_e32 v14, v1
; SDAG-NEXT: v_min_u32_e32 v10, v10, v11
-; SDAG-NEXT: v_cmp_eq_u64_e32 vcc, 0, v[4:5]
-; SDAG-NEXT: v_min_u32_e32 v4, v9, v14
-; SDAG-NEXT: v_add_i32_e64 v5, s[4:5], 64, v10
+; SDAG-NEXT: v_cmp_eq_u64_e32 vcc, 0, v[6:7]
+; SDAG-NEXT: v_min_u32_e32 v6, v9, v14
+; SDAG-NEXT: v_add_i32_e64 v7, s[4:5], 64, v10
; SDAG-NEXT: v_addc_u32_e64 v9, s[4:5], 0, 0, s[4:5]
; SDAG-NEXT: s_or_b64 s[6:7], vcc, s[6:7]
; SDAG-NEXT: v_cmp_ne_u64_e32 vcc, 0, v[0:1]
; SDAG-NEXT: v_cndmask_b32_e64 v9, v9, 0, vcc
-; SDAG-NEXT: v_cndmask_b32_e32 v4, v5, v4, vcc
-; SDAG-NEXT: v_sub_i32_e32 v4, vcc, v4, v13
-; SDAG-NEXT: v_subb_u32_e32 v5, vcc, v9, v12, vcc
-; SDAG-NEXT: v_xor_b32_e32 v9, 0x7f, v4
+; SDAG-NEXT: v_cndmask_b32_e32 v6, v7, v6, vcc
+; SDAG-NEXT: v_sub_i32_e32 v6, vcc, v6, v13
+; SDAG-NEXT: v_subb_u32_e32 v7, vcc, v9, v12, vcc
+; SDAG-NEXT: v_xor_b32_e32 v9, 0x7f, v6
; SDAG-NEXT: v_subb_u32_e32 v10, vcc, 0, v8, vcc
-; SDAG-NEXT: v_cmp_lt_u64_e64 s[4:5], s[10:11], v[4:5]
+; SDAG-NEXT: v_cmp_lt_u64_e64 s[4:5], s[10:11], v[6:7]
; SDAG-NEXT: v_cndmask_b32_e64 v12, 0, 1, s[4:5]
; SDAG-NEXT: v_subb_u32_e32 v11, vcc, 0, v8, vcc
; SDAG-NEXT: v_or_b32_e32 v8, v9, v10
; SDAG-NEXT: v_cmp_ne_u64_e32 vcc, 0, v[10:11]
; SDAG-NEXT: v_cndmask_b32_e64 v13, 0, 1, vcc
-; SDAG-NEXT: v_or_b32_e32 v9, v5, v11
+; SDAG-NEXT: v_or_b32_e32 v9, v7, v11
; SDAG-NEXT: v_cmp_eq_u64_e32 vcc, 0, v[10:11]
; SDAG-NEXT: v_cndmask_b32_e32 v12, v13, v12, vcc
; SDAG-NEXT: v_cmp_ne_u64_e32 vcc, 0, v[8:9]
; SDAG-NEXT: v_and_b32_e32 v8, 1, v12
; SDAG-NEXT: v_cmp_eq_u32_e64 s[4:5], 1, v8
; SDAG-NEXT: s_or_b64 s[4:5], s[6:7], s[4:5]
-; SDAG-NEXT: v_cndmask_b32_e64 v13, v7, 0, s[4:5]
+; SDAG-NEXT: v_cndmask_b32_e64 v13, v5, 0, s[4:5]
; SDAG-NEXT: s_xor_b64 s[6:7], s[4:5], -1
-; SDAG-NEXT: v_cndmask_b32_e64 v9, v6, 0, s[4:5]
+; SDAG-NEXT: v_cndmask_b32_e64 v9, v4, 0, s[4:5]
; SDAG-NEXT: v_cndmask_b32_e64 v14, v3, 0, s[4:5]
; SDAG-NEXT: v_cndmask_b32_e64 v8, v2, 0, s[4:5]
; SDAG-NEXT: s_and_b64 s[4:5], s[6:7], vcc
; SDAG-NEXT: s_and_saveexec_b64 s[6:7], s[4:5]
; SDAG-NEXT: s_cbranch_execz .LBB0_12
; SDAG-NEXT: ; %bb.7: ; %udiv-bb1
-; SDAG-NEXT: v_add_i32_e32 v30, vcc, 1, v4
-; SDAG-NEXT: v_sub_i32_e64 v12, s[4:5], 63, v4
+; SDAG-NEXT: v_add_i32_e32 v30, vcc, 1, v6
+; SDAG-NEXT: v_sub_i32_e64 v12, s[4:5], 63, v6
; SDAG-NEXT: v_mov_b32_e32 v8, 0
; SDAG-NEXT: v_mov_b32_e32 v9, 0
-; SDAG-NEXT: v_addc_u32_e32 v31, vcc, 0, v5, vcc
+; SDAG-NEXT: v_addc_u32_e32 v31, vcc, 0, v7, vcc
; SDAG-NEXT: v_lshl_b64 v[12:13], v[2:3], v12
; SDAG-NEXT: v_addc_u32_e32 v32, vcc, 0, v10, vcc
; SDAG-NEXT: v_addc_u32_e32 v33, vcc, 0, v11, vcc
; SDAG-NEXT: v_or_b32_e32 v10, v30, v32
-; SDAG-NEXT: v_sub_i32_e32 v34, vcc, 0x7f, v4
+; SDAG-NEXT: v_sub_i32_e32 v34, vcc, 0x7f, v6
; SDAG-NEXT: v_or_b32_e32 v11, v31, v33
-; SDAG-NEXT: v_lshl_b64 v[4:5], v[6:7], v34
+; SDAG-NEXT: v_lshl_b64 v[6:7], v[4:5], v34
; SDAG-NEXT: v_sub_i32_e32 v35, vcc, 64, v34
; SDAG-NEXT: v_lshl_b64 v[14:15], v[2:3], v34
; SDAG-NEXT: v_cmp_ne_u64_e32 vcc, 0, v[10:11]
; SDAG-NEXT: v_lshr_b64 v[10:11], v[2:3], v35
-; SDAG-NEXT: v_or_b32_e32 v5, v5, v11
-; SDAG-NEXT: v_or_b32_e32 v4, v4, v10
+; SDAG-NEXT: v_or_b32_e32 v7, v7, v11
+; SDAG-NEXT: v_or_b32_e32 v6, v6, v10
; SDAG-NEXT: v_cmp_gt_u32_e64 s[4:5], 64, v34
-; SDAG-NEXT: v_cndmask_b32_e64 v5, v13, v5, s[4:5]
-; SDAG-NEXT: v_cndmask_b32_e64 v4, v12, v4, s[4:5]
+; SDAG-NEXT: v_cndmask_b32_e64 v7, v13, v7, s[4:5]
+; SDAG-NEXT: v_cndmask_b32_e64 v6, v12, v6, s[4:5]
; SDAG-NEXT: v_cndmask_b32_e64 v11, 0, v15, s[4:5]
; SDAG-NEXT: v_cndmask_b32_e64 v10, 0, v14, s[4:5]
; SDAG-NEXT: v_cmp_eq_u32_e64 s[4:5], 0, v34
-; SDAG-NEXT: v_cndmask_b32_e64 v5, v5, v7, s[4:5]
-; SDAG-NEXT: v_cndmask_b32_e64 v4, v4, v6, s[4:5]
+; SDAG-NEXT: v_cndmask_b32_e64 v7, v7, v5, s[4:5]
+; SDAG-NEXT: v_cndmask_b32_e64 v6, v6, v4, s[4:5]
; SDAG-NEXT: v_mov_b32_e32 v12, 0
; SDAG-NEXT: v_mov_b32_e32 v13, 0
; SDAG-NEXT: s_and_saveexec_b64 s[4:5], vcc
@@ -321,24 +321,24 @@ define <2 x i128> @v_sdiv_v2i128_vv(<2 x i128> %lhs, <2 x i128> %rhs) {
; SDAG-NEXT: v_lshr_b64 v[8:9], v[2:3], v30
; SDAG-NEXT: v_sub_i32_e32 v35, vcc, 64, v30
; SDAG-NEXT: v_subrev_i32_e32 v36, vcc, 64, v30
-; SDAG-NEXT: v_lshr_b64 v[37:38], v[6:7], v30
+; SDAG-NEXT: v_lshr_b64 v[37:38], v[4:5], v30
; SDAG-NEXT: v_add_i32_e32 v34, vcc, -1, v29
; SDAG-NEXT: v_mov_b32_e32 v12, 0
; SDAG-NEXT: v_mov_b32_e32 v13, 0
; SDAG-NEXT: v_mov_b32_e32 v14, 0
; SDAG-NEXT: v_mov_b32_e32 v15, 0
; SDAG-NEXT: s_mov_b64 s[10:11], 0
-; SDAG-NEXT: v_lshl_b64 v[48:49], v[6:7], v35
-; SDAG-NEXT: v_lshr_b64 v[6:7], v[6:7], v36
+; SDAG-NEXT: v_lshl_b64 v[48:49], v[4:5], v35
+; SDAG-NEXT: v_lshr_b64 v[4:5], v[4:5], v36
; SDAG-NEXT: v_addc_u32_e32 v35, vcc, -1, v28, vcc
; SDAG-NEXT: v_or_b32_e32 v9, v9, v49
; SDAG-NEXT: v_or_b32_e32 v8, v8, v48
; SDAG-NEXT: v_addc_u32_e32 v36, vcc, -1, v0, vcc
; SDAG-NEXT: v_cmp_gt_u32_e64 s[4:5], 64, v30
-; SDAG-NEXT: v_cndmask_b32_e64 v9, v7, v9, s[4:5]
-; SDAG-NEXT: v_cndmask_b32_e64 v8, v6, v8, s[4:5]
-; SDAG-NEXT: v_cndmask_b32_e64 v7, 0, v38, s[4:5]
-; SDAG-NEXT: v_cndmask_b32_e64 v6, 0, v37, s[4:5]
+; SDAG-NEXT: v_cndmask_b32_e64 v9, v5, v9, s[4:5]
+; SDAG-NEXT: v_cndmask_b32_e64 v8, v4, v8, s[4:5]
+; SDAG-NEXT: v_cndmask_b32_e64 v5, 0, v38, s[4:5]
+; SDAG-NEXT: v_cndmask_b32_e64 v4, 0, v37, s[4:5]
; SDAG-NEXT: v_addc_u32_e32 v37, vcc, -1, v1, vcc
; SDAG-NEXT: v_cmp_eq_u32_e32 vcc, 0, v30
; SDAG-NEXT: v_cndmask_b32_e32 v3, v9, v3, vcc
@@ -346,23 +346,23 @@ define <2 x i128> @v_sdiv_v2i128_vv(<2 x i128> %lhs, <2 x i128> %rhs) {
; SDAG-NEXT: v_mov_b32_e32 v9, 0
; SDAG-NEXT: .LBB0_9: ; %udiv-do-while
; SDAG-NEXT: ; =>This Inner Loop Header: Depth=1
-; SDAG-NEXT: v_lshl_b64 v[6:7], v[6:7], 1
+; SDAG-NEXT: v_lshl_b64 v[4:5], v[4:5], 1
; SDAG-NEXT: v_lshrrev_b32_e32 v8, 31, v3
; SDAG-NEXT: v_lshl_b64 v[2:3], v[2:3], 1
-; SDAG-NEXT: v_lshrrev_b32_e32 v38, 31, v5
-; SDAG-NEXT: v_lshl_b64 v[4:5], v[4:5], 1
+; SDAG-NEXT: v_lshrrev_b32_e32 v38, 31, v7
+; SDAG-NEXT: v_lshl_b64 v[6:7], v[6:7], 1
; SDAG-NEXT: v_lshrrev_b32_e32 v39, 31, v11
; SDAG-NEXT: v_lshl_b64 v[10:11], v[10:11], 1
-; SDAG-NEXT: v_or_b32_e32 v6, v6, v8
+; SDAG-NEXT: v_or_b32_e32 v4, v4, v8
; SDAG-NEXT: v_or_b32_e32 v2, v2, v38
-; SDAG-NEXT: v_or_b32_e32 v4, v4, v39
-; SDAG-NEXT: v_or_b32_e32 v5, v13, v5
+; SDAG-NEXT: v_or_b32_e32 v6, v6, v39
+; SDAG-NEXT: v_or_b32_e32 v7, v13, v7
; SDAG-NEXT: v_or_b32_e32 v11, v15, v11
; SDAG-NEXT: v_sub_i32_e32 v8, vcc, v34, v2
-; SDAG-NEXT: v_or_b32_e32 v4, v12, v4
+; SDAG-NEXT: v_or_b32_e32 v6, v12, v6
; SDAG-NEXT: v_subb_u32_e32 v8, vcc, v35, v3, vcc
-; SDAG-NEXT: v_subb_u32_e32 v8, vcc, v36, v6, vcc
-; SDAG-NEXT: v_subb_u32_e32 v8, vcc, v37, v7, vcc
+; SDAG-NEXT: v_subb_u32_e32 v8, vcc, v36, v4, vcc
+; SDAG-NEXT: v_subb_u32_e32 v8, vcc, v37, v5, vcc
; SDAG-NEXT: v_ashrrev_i32_e32 v8, 31, v8
; SDAG-NEXT: v_and_b32_e32 v15, v8, v29
; SDAG-NEXT: v_and_b32_e32 v38, v8, v28
@@ -370,8 +370,8 @@ define <2 x i128> @v_sdiv_v2i128_vv(<2 x i128> %lhs, <2 x i128> %rhs) {
; SDAG-NEXT: v_and_b32_e32 v48, v8, v1
; SDAG-NEXT: v_sub_i32_e32 v2, vcc, v2, v15
; SDAG-NEXT: v_subb_u32_e32 v3, vcc, v3, v38, vcc
-; SDAG-NEXT: v_subb_u32_e32 v6, vcc, v6, v39, vcc
-; SDAG-NEXT: v_subb_u32_e32 v7, vcc, v7, v48, vcc
+; SDAG-NEXT: v_subb_u32_e32 v4, vcc, v4, v39, vcc
+; SDAG-NEXT: v_subb_u32_e32 v5, vcc, v5, v48, vcc
; SDAG-NEXT: v_add_i32_e32 v30, vcc, -1, v30
; SDAG-NEXT: v_addc_u32_e32 v31, vcc, -1, v31, vcc
; SDAG-NEXT: v_addc_u32_e32 v32, vcc, -1, v32, vcc
@@ -390,7 +390,7 @@ define <2 x i128> @v_sdiv_v2i128_vv(<2 x i128> %lhs, <2 x i128> %rhs) {
; SDAG-NEXT: s_or_b64 exec, exec, s[10:11]
; SDAG-NEXT: .LBB0_11: ; %Flow11
; SDAG-NEXT: s_or_b64 exec, exec, s[8:9]
-; SDAG-NEXT: v_lshl_b64 v[0:1], v[4:5], 1
+; SDAG-NEXT: v_lshl_b64 v[0:1], v[6:7], 1
; SDAG-NEXT: v_lshrrev_b32_e32 v4, 31, v11
; SDAG-NEXT: v_lshl_b64 v[2:3], v[10:11], 1
; SDAG-NEXT: v_or_b32_e32 v0, v0, v4
@@ -402,11 +402,11 @@ define <2 x i128> @v_sdiv_v2i128_vv(<2 x i128> %lhs, <2 x i128> %rhs) {
; SDAG-NEXT: s_or_b64 exec, exec, s[6:7]
; SDAG-NEXT: v_xor_b32_e32 v3, v27, v26
; SDAG-NEXT: v_xor_b32_e32 v2, v25, v24
-; SDAG-NEXT: v_xor_b32_e32 v7, v21, v20
+; SDAG-NEXT: v_xor_b32_e32 v7, v19, v18
; SDAG-NEXT: v_xor_b32_e32 v6, v17, v16
-; SDAG-NEXT: v_xor_b32_e32 v4, v18, v3
+; SDAG-NEXT: v_xor_b32_e32 v4, v20, v3
; SDAG-NEXT: v_xor_b32_e32 v5, v22, v2
-; SDAG-NEXT: v_xor_b32_e32 v1, v19, v3
+; SDAG-NEXT: v_xor_b32_e32 v1, v21, v3
; SDAG-NEXT: v_xor_b32_e32 v0, v23, v2
; SDAG-NEXT: v_xor_b32_e32 v10, v13, v7
; SDAG-NEXT: v_xor_b32_e32 v9, v9, v6
diff --git a/llvm/test/CodeGen/AMDGPU/divergent-branch-uniform-condition.ll b/llvm/test/CodeGen/AMDGPU/divergent-branch-uniform-condition.ll
index 827cb4ac2589a..279c34722e272 100644
--- a/llvm/test/CodeGen/AMDGPU/divergent-branch-uniform-condition.ll
+++ b/llvm/test/CodeGen/AMDGPU/divergent-branch-uniform-condition.ll
@@ -30,8 +30,8 @@ define amdgpu_ps void @main(i32 %0, float %1) {
; ISA-NEXT: .LBB0_1: ; %Flow1
; ISA-NEXT: ; in Loop: Header=BB0_3 Depth=1
; ISA-NEXT: s_or_b64 exec, exec, s[4:5]
-; ISA-NEXT: s_mov_b64 s[8:9], 0
; ISA-NEXT: s_mov_b64 s[4:5], s[6:7]
+; ISA-NEXT: s_mov_b64 s[8:9], 0
; ISA-NEXT: .LBB0_2: ; %Flow
; ISA-NEXT: ; in Loop: Header=BB0_3 Depth=1
; ISA-NEXT: s_and_b64 s[6:7], exec, s[4:5]
diff --git a/llvm/test/CodeGen/AMDGPU/flat-atomicrmw-fmax.ll b/llvm/test/CodeGen/AMDGPU/flat-atomicrmw-fmax.ll
index d2cbc25bf7e04..7555c660c8206 100644
--- a/llvm/test/CodeGen/AMDGPU/flat-atomicrmw-fmax.ll
+++ b/llvm/test/CodeGen/AMDGPU/flat-atomicrmw-fmax.ll
@@ -5407,54 +5407,50 @@ define double @flat_agent_atomic_fmax_ret_f64__amdgpu_no_remote_memory(ptr %ptr,
; GFX90A-LABEL: flat_agent_atomic_fmax_ret_f64__amdgpu_no_remote_memory:
; GFX90A: ; %bb.0:
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
-; GFX90A-NEXT: v_mov_b32_e32 v5, v1
; GFX90A-NEXT: s_mov_b64 s[4:5], src_private_base
-; GFX90A-NEXT: v_mov_b32_e32 v4, v0
-; GFX90A-NEXT: v_cmp_ne_u32_e32 vcc, s5, v5
-; GFX90A-NEXT: v_max_f64 v[2:3], v[2:3], v[2:3]
-; GFX90A-NEXT: ; implicit-def: $vgpr0_vgpr1
+; GFX90A-NEXT: v_cmp_ne_u32_e32 vcc, s5, v1
+; GFX90A-NEXT: v_max_f64 v[4:5], v[2:3], v[2:3]
+; GFX90A-NEXT: ; implicit-def: $vgpr2_vgpr3
; GFX90A-NEXT: s_and_saveexec_b64 s[4:5], vcc
; GFX90A-NEXT: s_xor_b64 s[4:5], exec, s[4:5]
-; GFX90A-NEXT: s_cbranch_execnz .LBB24_3
-; GFX90A-NEXT: ; %bb.1: ; %Flow2
-; GFX90A-NEXT: s_andn2_saveexec_b64 s[4:5], s[4:5]
-; GFX90A-NEXT: s_cbranch_execnz .LBB24_6
-; GFX90A-NEXT: .LBB24_2: ; %atomicrmw.phi
-; GFX90A-NEXT: s_or_b64 exec, exec, s[4:5]
-; GFX90A-NEXT: s_setpc_b64 s[30:31]
-; GFX90A-NEXT: .LBB24_3: ; %atomicrmw.global
-; GFX90A-NEXT: flat_load_dwordx2 v[0:1], v[4:5]
+; GFX90A-NEXT: s_cbranch_execz .LBB24_4
+; GFX90A-NEXT: ; %bb.1: ; %atomicrmw.global
+; GFX90A-NEXT: flat_load_dwordx2 v[2:3], v[0:1]
; GFX90A-NEXT: s_mov_b64 s[6:7], 0
-; GFX90A-NEXT: .LBB24_4: ; %atomicrmw.start
+; GFX90A-NEXT: .LBB24_2: ; %atomicrmw.start
; GFX90A-NEXT: ; =>This Inner Loop Header: Depth=1
; GFX90A-NEXT: s_waitcnt vmcnt(0) lgkmcnt(0)
-; GFX90A-NEXT: v_pk_mov_b32 v[8:9], v[0:1], v[0:1] op_sel:[0,1]
-; GFX90A-NEXT: v_max_f64 v[0:1], v[8:9], v[8:9]
-; GFX90A-NEXT: v_max_f64 v[6:7], v[0:1], v[2:3]
-; GFX90A-NEXT: flat_atomic_cmpswap_x2 v[0:1], v[4:5], v[6:9] glc
+; GFX90A-NEXT: v_pk_mov_b32 v[8:9], v[2:3], v[2:3] op_sel:[0,1]
+; GFX90A-NEXT: v_max_f64 v[2:3], v[8:9], v[8:9]
+; GFX90A-NEXT: v_max_f64 v[6:7], v[2:3], v[4:5]
+; GFX90A-NEXT: flat_atomic_cmpswap_x2 v[2:3], v[0:1], v[6:9] glc
; GFX90A-NEXT: s_waitcnt vmcnt(0) lgkmcnt(0)
; GFX90A-NEXT: buffer_wbinvl1
-; GFX90A-NEXT: v_cmp_eq_u64_e32 vcc, v[0:1], v[8:9]
+; GFX90A-NEXT: v_cmp_eq_u64_e32 vcc, v[2:3], v[8:9]
; GFX90A-NEXT: s_or_b64 s[6:7], vcc, s[6:7]
; GFX90A-NEXT: s_andn2_b64 exec, exec, s[6:7]
-; GFX90A-NEXT: s_cbranch_execnz .LBB24_4
-; GFX90A-NEXT: ; %bb.5: ; %Flow
+; GFX90A-NEXT: s_cbranch_execnz .LBB24_2
+; GFX90A-NEXT: ; %bb.3: ; %Flow
; GFX90A-NEXT: s_or_b64 exec, exec, s[6:7]
+; GFX90A-NEXT: ; implicit-def: $vgpr0_vgpr1
; GFX90A-NEXT: ; implicit-def: $vgpr4_vgpr5
-; GFX90A-NEXT: ; implicit-def: $vgpr2_vgpr3
+; GFX90A-NEXT: .LBB24_4: ; %Flow2
; GFX90A-NEXT: s_andn2_saveexec_b64 s[4:5], s[4:5]
-; GFX90A-NEXT: s_cbranch_execz .LBB24_2
-; GFX90A-NEXT: .LBB24_6: ; %atomicrmw.private
-; GFX90A-NEXT: v_cmp_ne_u64_e32 vcc, 0, v[4:5]
-; GFX90A-NEXT: v_cndmask_b32_e32 v6, -1, v4, vcc
-; GFX90A-NEXT: buffer_load_dword v0, v6, s[0:3], 0 offen
-; GFX90A-NEXT: buffer_load_dword v1, v6, s[0:3], 0 offen offset:4
+; GFX90A-NEXT: s_cbranch_execz .LBB24_6
+; GFX90A-NEXT: ; %bb.5: ; %atomicrmw.private
+; GFX90A-NEXT: v_cmp_ne_u64_e32 vcc, 0, v[0:1]
+; GFX90A-NEXT: v_cndmask_b32_e32 v6, -1, v0, vcc
+; GFX90A-NEXT: buffer_load_dword v2, v6, s[0:3], 0 offen
+; GFX90A-NEXT: buffer_load_dword v3, v6, s[0:3], 0 offen offset:4
; GFX90A-NEXT: s_waitcnt vmcnt(0)
-; GFX90A-NEXT: v_max_f64 v[4:5], v[0:1], v[0:1]
-; GFX90A-NEXT: v_max_f64 v[2:3], v[4:5], v[2:3]
-; GFX90A-NEXT: buffer_store_dword v2, v6, s[0:3], 0 offen
-; GFX90A-NEXT: buffer_store_dword v3, v6, s[0:3], 0 offen offset:4
+; GFX90A-NEXT: v_max_f64 v[0:1], v[2:3], v[2:3]
+; GFX90A-NEXT: v_max_f64 v[0:1], v[0:1], v[4:5]
+; GFX90A-NEXT: buffer_store_dword v0, v6, s[0:3], 0 offen
+; GFX90A-NEXT: buffer_store_dword v1, v6, s[0:3], 0 offen offset:4
+; GFX90A-NEXT: .LBB24_6: ; %atomicrmw.phi
; GFX90A-NEXT: s_or_b64 exec, exec, s[4:5]
+; GFX90A-NEXT: v_mov_b32_e32 v0, v2
+; GFX90A-NEXT: v_mov_b32_e32 v1, v3
; GFX90A-NEXT: s_waitcnt vmcnt(0)
; GFX90A-NEXT: s_setpc_b64 s[30:31]
;
diff --git a/llvm/test/CodeGen/AMDGPU/flat-atomicrmw-fmin.ll b/llvm/test/CodeGen/AMDGPU/flat-atomicrmw-fmin.ll
index 805848fc3e1cc..8e97f2e509092 100644
--- a/llvm/test/CodeGen/AMDGPU/flat-atomicrmw-fmin.ll
+++ b/llvm/test/CodeGen/AMDGPU/flat-atomicrmw-fmin.ll
@@ -5407,54 +5407,50 @@ define double @flat_agent_atomic_fmin_ret_f64__amdgpu_no_remote_memory(ptr %ptr,
; GFX90A-LABEL: flat_agent_atomic_fmin_ret_f64__amdgpu_no_remote_memory:
; GFX90A: ; %bb.0:
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
-; GFX90A-NEXT: v_mov_b32_e32 v5, v1
; GFX90A-NEXT: s_mov_b64 s[4:5], src_private_base
-; GFX90A-NEXT: v_mov_b32_e32 v4, v0
-; GFX90A-NEXT: v_cmp_ne_u32_e32 vcc, s5, v5
-; GFX90A-NEXT: v_max_f64 v[2:3], v[2:3], v[2:3]
-; GFX90A-NEXT: ; implicit-def: $vgpr0_vgpr1
+; GFX90A-NEXT: v_cmp_ne_u32_e32 vcc, s5, v1
+; GFX90A-NEXT: v_max_f64 v[4:5], v[2:3], v[2:3]
+; GFX90A-NEXT: ; implicit-def: $vgpr2_vgpr3
; GFX90A-NEXT: s_and_saveexec_b64 s[4:5], vcc
; GFX90A-NEXT: s_xor_b64 s[4:5], exec, s[4:5]
-; GFX90A-NEXT: s_cbranch_execnz .LBB24_3
-; GFX90A-NEXT: ; %bb.1: ; %Flow2
-; GFX90A-NEXT: s_andn2_saveexec_b64 s[4:5], s[4:5]
-; GFX90A-NEXT: s_cbranch_execnz .LBB24_6
-; GFX90A-NEXT: .LBB24_2: ; %atomicrmw.phi
-; GFX90A-NEXT: s_or_b64 exec, exec, s[4:5]
-; GFX90A-NEXT: s_setpc_b64 s[30:31]
-; GFX90A-NEXT: .LBB24_3: ; %atomicrmw.global
-; GFX90A-NEXT: flat_load_dwordx2 v[0:1], v[4:5]
+; GFX90A-NEXT: s_cbranch_execz .LBB24_4
+; GFX90A-NEXT: ; %bb.1: ; %atomicrmw.global
+; GFX90A-NEXT: flat_load_dwordx2 v[2:3], v[0:1]
; GFX90A-NEXT: s_mov_b64 s[6:7], 0
-; GFX90A-NEXT: .LBB24_4: ; %atomicrmw.start
+; GFX90A-NEXT: .LBB24_2: ; %atomicrmw.start
; GFX90A-NEXT: ; =>This Inner Loop Header: Depth=1
; GFX90A-NEXT: s_waitcnt vmcnt(0) lgkmcnt(0)
-; GFX90A-NEXT: v_pk_mov_b32 v[8:9], v[0:1], v[0:1] op_sel:[0,1]
-; GFX90A-NEXT: v_max_f64 v[0:1], v[8:9], v[8:9]
-; GFX90A-NEXT: v_min_f64 v[6:7], v[0:1], v[2:3]
-; GFX90A-NEXT: flat_atomic_cmpswap_x2 v[0:1], v[4:5], v[6:9] glc
+; GFX90A-NEXT: v_pk_mov_b32 v[8:9], v[2:3], v[2:3] op_sel:[0,1]
+; GFX90A-NEXT: v_max_f64 v[2:3], v[8:9], v[8:9]
+; GFX90A-NEXT: v_min_f64 v[6:7], v[2:3], v[4:5]
+; GFX90A-NEXT: flat_atomic_cmpswap_x2 v[2:3], v[0:1], v[6:9] glc
; GFX90A-NEXT: s_waitcnt vmcnt(0) lgkmcnt(0)
; GFX90A-NEXT: buffer_wbinvl1
-; GFX90A-NEXT: v_cmp_eq_u64_e32 vcc, v[0:1], v[8:9]
+; GFX90A-NEXT: v_cmp_eq_u64_e32 vcc, v[2:3], v[8:9]
; GFX90A-NEXT: s_or_b64 s[6:7], vcc, s[6:7]
; GFX90A-NEXT: s_andn2_b64 exec, exec, s[6:7]
-; GFX90A-NEXT: s_cbranch_execnz .LBB24_4
-; GFX90A-NEXT: ; %bb.5: ; %Flow
+; GFX90A-NEXT: s_cbranch_execnz .LBB24_2
+; GFX90A-NEXT: ; %bb.3: ; %Flow
; GFX90A-NEXT: s_or_b64 exec, exec, s[6:7]
+; GFX90A-NEXT: ; implicit-def: $vgpr0_vgpr1
; GFX90A-NEXT: ; implicit-def: $vgpr4_vgpr5
-; GFX90A-NEXT: ; implicit-def: $vgpr2_vgpr3
+; GFX90A-NEXT: .LBB24_4: ; %Flow2
; GFX90A-NEXT: s_andn2_saveexec_b64 s[4:5], s[4:5]
-; GFX90A-NEXT: s_cbranch_execz .LBB24_2
-; GFX90A-NEXT: .LBB24_6: ; %atomicrmw.private
-; GFX90A-NEXT: v_cmp_ne_u64_e32 vcc, 0, v[4:5]
-; GFX90A-NEXT: v_cndmask_b32_e32 v6, -1, v4, vcc
-; GFX90A-NEXT: buffer_load_dword v0, v6, s[0:3], 0 offen
-; GFX90A-NEXT: buffer_load_dword v1, v6, s[0:3], 0 offen offset:4
+; GFX90A-NEXT: s_cbranch_execz .LBB24_6
+; GFX90A-NEXT: ; %bb.5: ; %atomicrmw.private
+; GFX90A-NEXT: v_cmp_ne_u64_e32 vcc, 0, v[0:1]
+; GFX90A-NEXT: v_cndmask_b32_e32 v6, -1, v0, vcc
+; GFX90A-NEXT: buffer_load_dword v2, v6, s[0:3], 0 offen
+; GFX90A-NEXT: buffer_load_dword v3, v6, s[0:3], 0 offen offset:4
; GFX90A-NEXT: s_waitcnt vmcnt(0)
-; GFX90A-NEXT: v_max_f64 v[4:5], v[0:1], v[0:1]
-; GFX90A-NEXT: v_min_f64 v[2:3], v[4:5], v[2:3]
-; GFX90A-NEXT: buffer_store_dword v2, v6, s[0:3], 0 offen
-; GFX90A-NEXT: buffer_store_dword v3, v6, s[0:3], 0 offen offset:4
+; GFX90A-NEXT: v_max_f64 v[0:1], v[2:3], v[2:3]
+; GFX90A-NEXT: v_min_f64 v[0:1], v[0:1], v[4:5]
+; GFX90A-NEXT: buffer_store_dword v0, v6, s[0:3], 0 offen
+; GFX90A-NEXT: buffer_store_dword v1, v6, s[0:3], 0 offen offset:4
+; GFX90A-NEXT: .LBB24_6: ; %atomicrmw.phi
; GFX90A-NEXT: s_or_b64 exec, exec, s[4:5]
+; GFX90A-NEXT: v_mov_b32_e32 v0, v2
+; GFX90A-NEXT: v_mov_b32_e32 v1, v3
; GFX90A-NEXT: s_waitcnt vmcnt(0)
; GFX90A-NEXT: s_setpc_b64 s[30:31]
;
diff --git a/llvm/test/CodeGen/AMDGPU/global-atomicrmw-fadd.ll b/llvm/test/CodeGen/AMDGPU/global-atomicrmw-fadd.ll
index 0ddb181decb94..32285798eaca9 100644
--- a/llvm/test/CodeGen/AMDGPU/global-atomicrmw-fadd.ll
+++ b/llvm/test/CodeGen/AMDGPU/global-atomicrmw-fadd.ll
@@ -708,20 +708,20 @@ define void @global_agent_atomic_fadd_noret_f32__amdgpu_no_fine_grained_memory(p
; GFX7-NEXT: s_mov_b32 s7, 0xf000
; GFX7-NEXT: s_mov_b32 s4, s6
; GFX7-NEXT: s_mov_b32 s5, s6
-; GFX7-NEXT: buffer_load_dword v4, v[0:1], s[4:7], 0 addr64
+; GFX7-NEXT: buffer_load_dword v3, v[0:1], s[4:7], 0 addr64
; GFX7-NEXT: s_mov_b64 s[8:9], 0
; GFX7-NEXT: .LBB3_1: ; %atomicrmw.start
; GFX7-NEXT: ; =>This Inner Loop Header: Depth=1
; GFX7-NEXT: s_waitcnt vmcnt(0)
-; GFX7-NEXT: v_add_f32_e32 v3, v4, v2
-; GFX7-NEXT: v_mov_b32_e32 v6, v4
; GFX7-NEXT: v_mov_b32_e32 v5, v3
-; GFX7-NEXT: buffer_atomic_cmpswap v[5:6], v[0:1], s[4:7], 0 addr64 glc
+; GFX7-NEXT: v_add_f32_e32 v4, v5, v2
+; GFX7-NEXT: v_mov_b32_e32 v3, v4
+; GFX7-NEXT: v_mov_b32_e32 v4, v5
+; GFX7-NEXT: buffer_atomic_cmpswap v[3:4], v[0:1], s[4:7], 0 addr64 glc
; GFX7-NEXT: s_waitcnt vmcnt(0)
; GFX7-NEXT: buffer_wbinvl1
-; GFX7-NEXT: v_cmp_eq_u32_e32 vcc, v5, v4
+; GFX7-NEXT: v_cmp_eq_u32_e32 vcc, v3, v5
; GFX7-NEXT: s_or_b64 s[8:9], vcc, s[8:9]
-; GFX7-NEXT: v_mov_b32_e32 v4, v5
; GFX7-NEXT: s_andn2_b64 exec, exec, s[8:9]
; GFX7-NEXT: s_cbranch_execnz .LBB3_1
; GFX7-NEXT: ; %bb.2: ; %atomicrmw.end
@@ -734,21 +734,21 @@ define void @global_agent_atomic_fadd_noret_f32__amdgpu_no_fine_grained_memory(p
; GFX6-NEXT: s_mov_b32 s7, 0xf000
; GFX6-NEXT: s_mov_b32 s4, s6
; GFX6-NEXT: s_mov_b32 s5, s6
-; GFX6-NEXT: buffer_load_dword v4, v[0:1], s[4:7], 0 addr64
+; GFX6-NEXT: buffer_load_dword v3, v[0:1], s[4:7], 0 addr64
; GFX6-NEXT: s_mov_b64 s[8:9], 0
; GFX6-NEXT: .LBB3_1: ; %atomicrmw.start
; GFX6-NEXT: ; =>This Inner Loop Header: Depth=1
; GFX6-NEXT: s_waitcnt vmcnt(0)
-; GFX6-NEXT: v_add_f32_e32 v3, v4, v2
-; GFX6-NEXT: s_waitcnt expcnt(0)
-; GFX6-NEXT: v_mov_b32_e32 v6, v4
; GFX6-NEXT: v_mov_b32_e32 v5, v3
-; GFX6-NEXT: buffer_atomic_cmpswap v[5:6], v[0:1], s[4:7], 0 addr64 glc
+; GFX6-NEXT: s_waitcnt expcnt(0)
+; GFX6-NEXT: v_add_f32_e32 v4, v5, v2
+; GFX6-NEXT: v_mov_b32_e32 v3, v4
+; GFX6-NEXT: v_mov_b32_e32 v4, v5
+; GFX6-NEXT: buffer_atomic_cmpswap v[3:4], v[0:1], s[4:7], 0 addr64 glc
; GFX6-NEXT: s_waitcnt vmcnt(0)
; GFX6-NEXT: buffer_wbinvl1
-; GFX6-NEXT: v_cmp_eq_u32_e32 vcc, v5, v4
+; GFX6-NEXT: v_cmp_eq_u32_e32 vcc, v3, v5
; GFX6-NEXT: s_or_b64 s[8:9], vcc, s[8:9]
-; GFX6-NEXT: v_mov_b32_e32 v4, v5
; GFX6-NEXT: s_andn2_b64 exec, exec, s[8:9]
; GFX6-NEXT: s_cbranch_execnz .LBB3_1
; GFX6-NEXT: ; %bb.2: ; %atomicrmw.end
@@ -888,20 +888,20 @@ define void @global_agent_atomic_fadd_noret_f32__offset12b_pos__amdgpu_no_fine_g
; GFX7-NEXT: s_mov_b32 s7, 0xf000
; GFX7-NEXT: s_mov_b32 s4, s6
; GFX7-NEXT: s_mov_b32 s5, s6
-; GFX7-NEXT: buffer_load_dword v4, v[0:1], s[4:7], 0 addr64 offset:2044
+; GFX7-NEXT: buffer_load_dword v3, v[0:1], s[4:7], 0 addr64 offset:2044
; GFX7-NEXT: s_mov_b64 s[8:9], 0
; GFX7-NEXT: .LBB4_1: ; %atomicrmw.start
; GFX7-NEXT: ; =>This Inner Loop Header: Depth=1
; GFX7-NEXT: s_waitcnt vmcnt(0)
-; GFX7-NEXT: v_add_f32_e32 v3, v4, v2
-; GFX7-NEXT: v_mov_b32_e32 v6, v4
; GFX7-NEXT: v_mov_b32_e32 v5, v3
-; GFX7-NEXT: buffer_atomic_cmpswap v[5:6], v[0:1], s[4:7], 0 addr64 offset:2044 glc
+; GFX7-NEXT: v_add_f32_e32 v4, v5, v2
+; GFX7-NEXT: v_mov_b32_e32 v3, v4
+; GFX7-NEXT: v_mov_b32_e32 v4, v5
+; GFX7-NEXT: buffer_atomic_cmpswap v[3:4], v[0:1], s[4:7], 0 addr64 offset:2044 glc
; GFX7-NEXT: s_waitcnt vmcnt(0)
; GFX7-NEXT: buffer_wbinvl1
-; GFX7-NEXT: v_cmp_eq_u32_e32 vcc, v5, v4
+; GFX7-NEXT: v_cmp_eq_u32_e32 vcc, v3, v5
; GFX7-NEXT: s_or_b64 s[8:9], vcc, s[8:9]
-; GFX7-NEXT: v_mov_b32_e32 v4, v5
; GFX7-NEXT: s_andn2_b64 exec, exec, s[8:9]
; GFX7-NEXT: s_cbranch_execnz .LBB4_1
; GFX7-NEXT: ; %bb.2: ; %atomicrmw.end
@@ -914,21 +914,21 @@ define void @global_agent_atomic_fadd_noret_f32__offset12b_pos__amdgpu_no_fine_g
; GFX6-NEXT: s_mov_b32 s7, 0xf000
; GFX6-NEXT: s_mov_b32 s4, s6
; GFX6-NEXT: s_mov_b32 s5, s6
-; GFX6-NEXT: buffer_load_dword v4, v[0:1], s[4:7], 0 addr64 offset:2044
+; GFX6-NEXT: buffer_load_dword v3, v[0:1], s[4:7], 0 addr64 offset:2044
; GFX6-NEXT: s_mov_b64 s[8:9], 0
; GFX6-NEXT: .LBB4_1: ; %atomicrmw.start
; GFX6-NEXT: ; =>This Inner Loop Header: Depth=1
; GFX6-NEXT: s_waitcnt vmcnt(0)
-; GFX6-NEXT: v_add_f32_e32 v3, v4, v2
-; GFX6-NEXT: s_waitcnt expcnt(0)
-; GFX6-NEXT: v_mov_b32_e32 v6, v4
; GFX6-NEXT: v_mov_b32_e32 v5, v3
-; GFX6-NEXT: buffer_atomic_cmpswap v[5:6], v[0:1], s[4:7], 0 addr64 offset:2044 glc
+; GFX6-NEXT: s_waitcnt expcnt(0)
+; GFX6-NEXT: v_add_f32_e32 v4, v5, v2
+; GFX6-NEXT: v_mov_b32_e32 v3, v4
+; GFX6-NEXT: v_mov_b32_e32 v4, v5
+; GFX6-NEXT: buffer_atomic_cmpswap v[3:4], v[0:1], s[4:7], 0 addr64 offset:2044 glc
; GFX6-NEXT: s_waitcnt vmcnt(0)
; GFX6-NEXT: buffer_wbinvl1
-; GFX6-NEXT: v_cmp_eq_u32_e32 vcc, v5, v4
+; GFX6-NEXT: v_cmp_eq_u32_e32 vcc, v3, v5
; GFX6-NEXT: s_or_b64 s[8:9], vcc, s[8:9]
-; GFX6-NEXT: v_mov_b32_e32 v4, v5
; GFX6-NEXT: s_andn2_b64 exec, exec, s[8:9]
; GFX6-NEXT: s_cbranch_execnz .LBB4_1
; GFX6-NEXT: ; %bb.2: ; %atomicrmw.end
@@ -1069,7 +1069,7 @@ define void @global_agent_atomic_fadd_noret_f32__offset12b_neg__amdgpu_no_fine_g
; GFX7-NEXT: s_mov_b32 s5, -1
; GFX7-NEXT: s_mov_b32 s7, 0xf000
; GFX7-NEXT: s_mov_b32 s6, 0
-; GFX7-NEXT: buffer_load_dword v4, v[0:1], s[4:7], 0 addr64
+; GFX7-NEXT: buffer_load_dword v3, v[0:1], s[4:7], 0 addr64
; GFX7-NEXT: v_add_i32_e32 v0, vcc, 0xfffff800, v0
; GFX7-NEXT: v_addc_u32_e32 v1, vcc, -1, v1, vcc
; GFX7-NEXT: s_mov_b64 s[8:9], 0
@@ -1078,15 +1078,15 @@ define void @global_agent_atomic_fadd_noret_f32__offset12b_neg__amdgpu_no_fine_g
; GFX7-NEXT: .LBB5_1: ; %atomicrmw.start
; GFX7-NEXT: ; =>This Inner Loop Header: Depth=1
; GFX7-NEXT: s_waitcnt vmcnt(0)
-; GFX7-NEXT: v_add_f32_e32 v3, v4, v2
-; GFX7-NEXT: v_mov_b32_e32 v6, v4
; GFX7-NEXT: v_mov_b32_e32 v5, v3
-; GFX7-NEXT: buffer_atomic_cmpswap v[5:6], v[0:1], s[4:7], 0 addr64 glc
+; GFX7-NEXT: v_add_f32_e32 v4, v5, v2
+; GFX7-NEXT: v_mov_b32_e32 v3, v4
+; GFX7-NEXT: v_mov_b32_e32 v4, v5
+; GFX7-NEXT: buffer_atomic_cmpswap v[3:4], v[0:1], s[4:7], 0 addr64 glc
; GFX7-NEXT: s_waitcnt vmcnt(0)
; GFX7-NEXT: buffer_wbinvl1
-; GFX7-NEXT: v_cmp_eq_u32_e32 vcc, v5, v4
+; GFX7-NEXT: v_cmp_eq_u32_e32 vcc, v3, v5
; GFX7-NEXT: s_or_b64 s[8:9], vcc, s[8:9]
-; GFX7-NEXT: v_mov_b32_e32 v4, v5
; GFX7-NEXT: s_andn2_b64 exec, exec, s[8:9]
; GFX7-NEXT: s_cbranch_execnz .LBB5_1
; GFX7-NEXT: ; %bb.2: ; %atomicrmw.end
@@ -1099,7 +1099,7 @@ define void @global_agent_atomic_fadd_noret_f32__offset12b_neg__amdgpu_no_fine_g
; GFX6-NEXT: s_mov_b32 s5, -1
; GFX6-NEXT: s_mov_b32 s7, 0xf000
; GFX6-NEXT: s_mov_b32 s6, 0
-; GFX6-NEXT: buffer_load_dword v4, v[0:1], s[4:7], 0 addr64
+; GFX6-NEXT: buffer_load_dword v3, v[0:1], s[4:7], 0 addr64
; GFX6-NEXT: v_add_i32_e32 v0, vcc, 0xfffff800, v0
; GFX6-NEXT: v_addc_u32_e32 v1, vcc, -1, v1, vcc
; GFX6-NEXT: s_mov_b64 s[8:9], 0
@@ -1108,16 +1108,16 @@ define void @global_agent_atomic_fadd_noret_f32__offset12b_neg__amdgpu_no_fine_g
; GFX6-NEXT: .LBB5_1: ; %atomicrmw.start
; GFX6-NEXT: ; =>This Inner Loop Header: Depth=1
; GFX6-NEXT: s_waitcnt vmcnt(0)
-; GFX6-NEXT: v_add_f32_e32 v3, v4, v2
-; GFX6-NEXT: s_waitcnt expcnt(0)
-; GFX6-NEXT: v_mov_b32_e32 v6, v4
; GFX6-NEXT: v_mov_b32_e32 v5, v3
-; GFX6-NEXT: buffer_atomic_cmpswap v[5:6], v[0:1], s[4:7], 0 addr64 glc
+; GFX6-NEXT: s_waitcnt expcnt(0)
+; GFX6-NEXT: v_add_f32_e32 v4, v5, v2
+; GFX6-NEXT: v_mov_b32_e32 v3, v4
+; GFX6-NEXT: v_mov_b32_e32 v4, v5
+; GFX6-NEXT: buffer_atomic_cmpswap v[3:4], v[0:1], s[4:7], 0 addr64 glc
; GFX6-NEXT: s_waitcnt vmcnt(0)
; GFX6-NEXT: buffer_wbinvl1
-; GFX6-NEXT: v_cmp_eq_u32_e32 vcc, v5, v4
+; GFX6-NEXT: v_cmp_eq_u32_e32 vcc, v3, v5
; GFX6-NEXT: s_or_b64 s[8:9], vcc, s[8:9]
-; GFX6-NEXT: v_mov_b32_e32 v4, v5
; GFX6-NEXT: s_andn2_b64 exec, exec, s[8:9]
; GFX6-NEXT: s_cbranch_execnz .LBB5_1
; GFX6-NEXT: ; %bb.2: ; %atomicrmw.end
@@ -1450,20 +1450,20 @@ define void @global_system_atomic_fadd_noret_f32__offset12b_pos__amdgpu_no_fine_
; GFX7-NEXT: s_mov_b32 s7, 0xf000
; GFX7-NEXT: s_mov_b32 s4, s6
; GFX7-NEXT: s_mov_b32 s5, s6
-; GFX7-NEXT: buffer_load_dword v4, v[0:1], s[4:7], 0 addr64 offset:2044
+; GFX7-NEXT: buffer_load_dword v3, v[0:1], s[4:7], 0 addr64 offset:2044
; GFX7-NEXT: s_mov_b64 s[8:9], 0
; GFX7-NEXT: .LBB7_1: ; %atomicrmw.start
; GFX7-NEXT: ; =>This Inner Loop Header: Depth=1
; GFX7-NEXT: s_waitcnt vmcnt(0)
-; GFX7-NEXT: v_add_f32_e32 v3, v4, v2
-; GFX7-NEXT: v_mov_b32_e32 v6, v4
; GFX7-NEXT: v_mov_b32_e32 v5, v3
-; GFX7-NEXT: buffer_atomic_cmpswap v[5:6], v[0:1], s[4:7], 0 addr64 offset:2044 glc
+; GFX7-NEXT: v_add_f32_e32 v4, v5, v2
+; GFX7-NEXT: v_mov_b32_e32 v3, v4
+; GFX7-NEXT: v_mov_b32_e32 v4, v5
+; GFX7-NEXT: buffer_atomic_cmpswap v[3:4], v[0:1], s[4:7], 0 addr64 offset:2044 glc
; GFX7-NEXT: s_waitcnt vmcnt(0)
; GFX7-NEXT: buffer_wbinvl1
-; GFX7-NEXT: v_cmp_eq_u32_e32 vcc, v5, v4
+; GFX7-NEXT: v_cmp_eq_u32_e32 vcc, v3, v5
; GFX7-NEXT: s_or_b64 s[8:9], vcc, s[8:9]
-; GFX7-NEXT: v_mov_b32_e32 v4, v5
; GFX7-NEXT: s_andn2_b64 exec, exec, s[8:9]
; GFX7-NEXT: s_cbranch_execnz .LBB7_1
; GFX7-NEXT: ; %bb.2: ; %atomicrmw.end
@@ -1476,21 +1476,21 @@ define void @global_system_atomic_fadd_noret_f32__offset12b_pos__amdgpu_no_fine_
; GFX6-NEXT: s_mov_b32 s7, 0xf000
; GFX6-NEXT: s_mov_b32 s4, s6
; GFX6-NEXT: s_mov_b32 s5, s6
-; GFX6-NEXT: buffer_load_dword v4, v[0:1], s[4:7], 0 addr64 offset:2044
+; GFX6-NEXT: buffer_load_dword v3, v[0:1], s[4:7], 0 addr64 offset:2044
; GFX6-NEXT: s_mov_b64 s[8:9], 0
; GFX6-NEXT: .LBB7_1: ; %atomicrmw.start
; GFX6-NEXT: ; =>This Inner Loop Header: Depth=1
; GFX6-NEXT: s_waitcnt vmcnt(0)
-; GFX6-NEXT: v_add_f32_e32 v3, v4, v2
-; GFX6-NEXT: s_waitcnt expcnt(0)
-; GFX6-NEXT: v_mov_b32_e32 v6, v4
; GFX6-NEXT: v_mov_b32_e32 v5, v3
-; GFX6-NEXT: buffer_atomic_cmpswap v[5:6], v[0:1], s[4:7], 0 addr64 offset:2044 glc
+; GFX6-NEXT: s_waitcnt expcnt(0)
+; GFX6-NEXT: v_add_f32_e32 v4, v5, v2
+; GFX6-NEXT: v_mov_b32_e32 v3, v4
+; GFX6-NEXT: v_mov_b32_e32 v4, v5
+; GFX6-NEXT: buffer_atomic_cmpswap v[3:4], v[0:1], s[4:7], 0 addr64 offset:2044 glc
; GFX6-NEXT: s_waitcnt vmcnt(0)
; GFX6-NEXT: buffer_wbinvl1
-; GFX6-NEXT: v_cmp_eq_u32_e32 vcc, v5, v4
+; GFX6-NEXT: v_cmp_eq_u32_e32 vcc, v3, v5
; GFX6-NEXT: s_or_b64 s[8:9], vcc, s[8:9]
-; GFX6-NEXT: v_mov_b32_e32 v4, v5
; GFX6-NEXT: s_andn2_b64 exec, exec, s[8:9]
; GFX6-NEXT: s_cbranch_execnz .LBB7_1
; GFX6-NEXT: ; %bb.2: ; %atomicrmw.end
@@ -2049,20 +2049,20 @@ define void @global_agent_atomic_fadd_noret_f32_maybe_remote__amdgpu_ignore_deno
; GFX7-NEXT: s_mov_b32 s7, 0xf000
; GFX7-NEXT: s_mov_b32 s4, s6
; GFX7-NEXT: s_mov_b32 s5, s6
-; GFX7-NEXT: buffer_load_dword v4, v[0:1], s[4:7], 0 addr64 offset:2044
+; GFX7-NEXT: buffer_load_dword v3, v[0:1], s[4:7], 0 addr64 offset:2044
; GFX7-NEXT: s_mov_b64 s[8:9], 0
; GFX7-NEXT: .LBB10_1: ; %atomicrmw.start
; GFX7-NEXT: ; =>This Inner Loop Header: Depth=1
; GFX7-NEXT: s_waitcnt vmcnt(0)
-; GFX7-NEXT: v_add_f32_e32 v3, v4, v2
-; GFX7-NEXT: v_mov_b32_e32 v6, v4
; GFX7-NEXT: v_mov_b32_e32 v5, v3
-; GFX7-NEXT: buffer_atomic_cmpswap v[5:6], v[0:1], s[4:7], 0 addr64 offset:2044 glc
+; GFX7-NEXT: v_add_f32_e32 v4, v5, v2
+; GFX7-NEXT: v_mov_b32_e32 v3, v4
+; GFX7-NEXT: v_mov_b32_e32 v4, v5
+; GFX7-NEXT: buffer_atomic_cmpswap v[3:4], v[0:1], s[4:7], 0 addr64 offset:2044 glc
; GFX7-NEXT: s_waitcnt vmcnt(0)
; GFX7-NEXT: buffer_wbinvl1
-; GFX7-NEXT: v_cmp_eq_u32_e32 vcc, v5, v4
+; GFX7-NEXT: v_cmp_eq_u32_e32 vcc, v3, v5
; GFX7-NEXT: s_or_b64 s[8:9], vcc, s[8:9]
-; GFX7-NEXT: v_mov_b32_e32 v4, v5
; GFX7-NEXT: s_andn2_b64 exec, exec, s[8:9]
; GFX7-NEXT: s_cbranch_execnz .LBB10_1
; GFX7-NEXT: ; %bb.2: ; %atomicrmw.end
@@ -2075,21 +2075,21 @@ define void @global_agent_atomic_fadd_noret_f32_maybe_remote__amdgpu_ignore_deno
; GFX6-NEXT: s_mov_b32 s7, 0xf000
; GFX6-NEXT: s_mov_b32 s4, s6
; GFX6-NEXT: s_mov_b32 s5, s6
-; GFX6-NEXT: buffer_load_dword v4, v[0:1], s[4:7], 0 addr64 offset:2044
+; GFX6-NEXT: buffer_load_dword v3, v[0:1], s[4:7], 0 addr64 offset:2044
; GFX6-NEXT: s_mov_b64 s[8:9], 0
; GFX6-NEXT: .LBB10_1: ; %atomicrmw.start
; GFX6-NEXT: ; =>This Inner Loop Header: Depth=1
; GFX6-NEXT: s_waitcnt vmcnt(0)
-; GFX6-NEXT: v_add_f32_e32 v3, v4, v2
-; GFX6-NEXT: s_waitcnt expcnt(0)
-; GFX6-NEXT: v_mov_b32_e32 v6, v4
; GFX6-NEXT: v_mov_b32_e32 v5, v3
-; GFX6-NEXT: buffer_atomic_cmpswap v[5:6], v[0:1], s[4:7], 0 addr64 offset:2044 glc
+; GFX6-NEXT: s_waitcnt expcnt(0)
+; GFX6-NEXT: v_add_f32_e32 v4, v5, v2
+; GFX6-NEXT: v_mov_b32_e32 v3, v4
+; GFX6-NEXT: v_mov_b32_e32 v4, v5
+; GFX6-NEXT: buffer_atomic_cmpswap v[3:4], v[0:1], s[4:7], 0 addr64 offset:2044 glc
; GFX6-NEXT: s_waitcnt vmcnt(0)
; GFX6-NEXT: buffer_wbinvl1
-; GFX6-NEXT: v_cmp_eq_u32_e32 vcc, v5, v4
+; GFX6-NEXT: v_cmp_eq_u32_e32 vcc, v3, v5
; GFX6-NEXT: s_or_b64 s[8:9], vcc, s[8:9]
-; GFX6-NEXT: v_mov_b32_e32 v4, v5
; GFX6-NEXT: s_andn2_b64 exec, exec, s[8:9]
; GFX6-NEXT: s_cbranch_execnz .LBB10_1
; GFX6-NEXT: ; %bb.2: ; %atomicrmw.end
@@ -2804,20 +2804,20 @@ define void @global_agent_atomic_fadd_noret_f32_maybe_remote(ptr addrspace(1) %p
; GFX7-NEXT: s_mov_b32 s7, 0xf000
; GFX7-NEXT: s_mov_b32 s4, s6
; GFX7-NEXT: s_mov_b32 s5, s6
-; GFX7-NEXT: buffer_load_dword v4, v[0:1], s[4:7], 0 addr64 offset:2044
+; GFX7-NEXT: buffer_load_dword v3, v[0:1], s[4:7], 0 addr64 offset:2044
; GFX7-NEXT: s_mov_b64 s[8:9], 0
; GFX7-NEXT: .LBB14_1: ; %atomicrmw.start
; GFX7-NEXT: ; =>This Inner Loop Header: Depth=1
; GFX7-NEXT: s_waitcnt vmcnt(0)
-; GFX7-NEXT: v_add_f32_e32 v3, v4, v2
-; GFX7-NEXT: v_mov_b32_e32 v6, v4
; GFX7-NEXT: v_mov_b32_e32 v5, v3
-; GFX7-NEXT: buffer_atomic_cmpswap v[5:6], v[0:1], s[4:7], 0 addr64 offset:2044 glc
+; GFX7-NEXT: v_add_f32_e32 v4, v5, v2
+; GFX7-NEXT: v_mov_b32_e32 v3, v4
+; GFX7-NEXT: v_mov_b32_e32 v4, v5
+; GFX7-NEXT: buffer_atomic_cmpswap v[3:4], v[0:1], s[4:7], 0 addr64 offset:2044 glc
; GFX7-NEXT: s_waitcnt vmcnt(0)
; GFX7-NEXT: buffer_wbinvl1
-; GFX7-NEXT: v_cmp_eq_u32_e32 vcc, v5, v4
+; GFX7-NEXT: v_cmp_eq_u32_e32 vcc, v3, v5
; GFX7-NEXT: s_or_b64 s[8:9], vcc, s[8:9]
-; GFX7-NEXT: v_mov_b32_e32 v4, v5
; GFX7-NEXT: s_andn2_b64 exec, exec, s[8:9]
; GFX7-NEXT: s_cbranch_execnz .LBB14_1
; GFX7-NEXT: ; %bb.2: ; %atomicrmw.end
@@ -2830,21 +2830,21 @@ define void @global_agent_atomic_fadd_noret_f32_maybe_remote(ptr addrspace(1) %p
; GFX6-NEXT: s_mov_b32 s7, 0xf000
; GFX6-NEXT: s_mov_b32 s4, s6
; GFX6-NEXT: s_mov_b32 s5, s6
-; GFX6-NEXT: buffer_load_dword v4, v[0:1], s[4:7], 0 addr64 offset:2044
+; GFX6-NEXT: buffer_load_dword v3, v[0:1], s[4:7], 0 addr64 offset:2044
; GFX6-NEXT: s_mov_b64 s[8:9], 0
; GFX6-NEXT: .LBB14_1: ; %atomicrmw.start
; GFX6-NEXT: ; =>This Inner Loop Header: Depth=1
; GFX6-NEXT: s_waitcnt vmcnt(0)
-; GFX6-NEXT: v_add_f32_e32 v3, v4, v2
-; GFX6-NEXT: s_waitcnt expcnt(0)
-; GFX6-NEXT: v_mov_b32_e32 v6, v4
; GFX6-NEXT: v_mov_b32_e32 v5, v3
-; GFX6-NEXT: buffer_atomic_cmpswap v[5:6], v[0:1], s[4:7], 0 addr64 offset:2044 glc
+; GFX6-NEXT: s_waitcnt expcnt(0)
+; GFX6-NEXT: v_add_f32_e32 v4, v5, v2
+; GFX6-NEXT: v_mov_b32_e32 v3, v4
+; GFX6-NEXT: v_mov_b32_e32 v4, v5
+; GFX6-NEXT: buffer_atomic_cmpswap v[3:4], v[0:1], s[4:7], 0 addr64 offset:2044 glc
; GFX6-NEXT: s_waitcnt vmcnt(0)
; GFX6-NEXT: buffer_wbinvl1
-; GFX6-NEXT: v_cmp_eq_u32_e32 vcc, v5, v4
+; GFX6-NEXT: v_cmp_eq_u32_e32 vcc, v3, v5
; GFX6-NEXT: s_or_b64 s[8:9], vcc, s[8:9]
-; GFX6-NEXT: v_mov_b32_e32 v4, v5
; GFX6-NEXT: s_andn2_b64 exec, exec, s[8:9]
; GFX6-NEXT: s_cbranch_execnz .LBB14_1
; GFX6-NEXT: ; %bb.2: ; %atomicrmw.end
@@ -2985,20 +2985,20 @@ define void @global_agent_atomic_fadd_noret_f32___amdgpu_no_fine_grained_memory(
; GFX7-NEXT: s_mov_b32 s7, 0xf000
; GFX7-NEXT: s_mov_b32 s4, s6
; GFX7-NEXT: s_mov_b32 s5, s6
-; GFX7-NEXT: buffer_load_dword v4, v[0:1], s[4:7], 0 addr64 offset:2044
+; GFX7-NEXT: buffer_load_dword v3, v[0:1], s[4:7], 0 addr64 offset:2044
; GFX7-NEXT: s_mov_b64 s[8:9], 0
; GFX7-NEXT: .LBB15_1: ; %atomicrmw.start
; GFX7-NEXT: ; =>This Inner Loop Header: Depth=1
; GFX7-NEXT: s_waitcnt vmcnt(0)
-; GFX7-NEXT: v_add_f32_e32 v3, v4, v2
-; GFX7-NEXT: v_mov_b32_e32 v6, v4
; GFX7-NEXT: v_mov_b32_e32 v5, v3
-; GFX7-NEXT: buffer_atomic_cmpswap v[5:6], v[0:1], s[4:7], 0 addr64 offset:2044 glc
+; GFX7-NEXT: v_add_f32_e32 v4, v5, v2
+; GFX7-NEXT: v_mov_b32_e32 v3, v4
+; GFX7-NEXT: v_mov_b32_e32 v4, v5
+; GFX7-NEXT: buffer_atomic_cmpswap v[3:4], v[0:1], s[4:7], 0 addr64 offset:2044 glc
; GFX7-NEXT: s_waitcnt vmcnt(0)
; GFX7-NEXT: buffer_wbinvl1
-; GFX7-NEXT: v_cmp_eq_u32_e32 vcc, v5, v4
+; GFX7-NEXT: v_cmp_eq_u32_e32 vcc, v3, v5
; GFX7-NEXT: s_or_b64 s[8:9], vcc, s[8:9]
-; GFX7-NEXT: v_mov_b32_e32 v4, v5
; GFX7-NEXT: s_andn2_b64 exec, exec, s[8:9]
; GFX7-NEXT: s_cbranch_execnz .LBB15_1
; GFX7-NEXT: ; %bb.2: ; %atomicrmw.end
@@ -3011,21 +3011,21 @@ define void @global_agent_atomic_fadd_noret_f32___amdgpu_no_fine_grained_memory(
; GFX6-NEXT: s_mov_b32 s7, 0xf000
; GFX6-NEXT: s_mov_b32 s4, s6
; GFX6-NEXT: s_mov_b32 s5, s6
-; GFX6-NEXT: buffer_load_dword v4, v[0:1], s[4:7], 0 addr64 offset:2044
+; GFX6-NEXT: buffer_load_dword v3, v[0:1], s[4:7], 0 addr64 offset:2044
; GFX6-NEXT: s_mov_b64 s[8:9], 0
; GFX6-NEXT: .LBB15_1: ; %atomicrmw.start
; GFX6-NEXT: ; =>This Inner Loop Header: Depth=1
; GFX6-NEXT: s_waitcnt vmcnt(0)
-; GFX6-NEXT: v_add_f32_e32 v3, v4, v2
-; GFX6-NEXT: s_waitcnt expcnt(0)
-; GFX6-NEXT: v_mov_b32_e32 v6, v4
; GFX6-NEXT: v_mov_b32_e32 v5, v3
-; GFX6-NEXT: buffer_atomic_cmpswap v[5:6], v[0:1], s[4:7], 0 addr64 offset:2044 glc
+; GFX6-NEXT: s_waitcnt expcnt(0)
+; GFX6-NEXT: v_add_f32_e32 v4, v5, v2
+; GFX6-NEXT: v_mov_b32_e32 v3, v4
+; GFX6-NEXT: v_mov_b32_e32 v4, v5
+; GFX6-NEXT: buffer_atomic_cmpswap v[3:4], v[0:1], s[4:7], 0 addr64 offset:2044 glc
; GFX6-NEXT: s_waitcnt vmcnt(0)
; GFX6-NEXT: buffer_wbinvl1
-; GFX6-NEXT: v_cmp_eq_u32_e32 vcc, v5, v4
+; GFX6-NEXT: v_cmp_eq_u32_e32 vcc, v3, v5
; GFX6-NEXT: s_or_b64 s[8:9], vcc, s[8:9]
-; GFX6-NEXT: v_mov_b32_e32 v4, v5
; GFX6-NEXT: s_andn2_b64 exec, exec, s[8:9]
; GFX6-NEXT: s_cbranch_execnz .LBB15_1
; GFX6-NEXT: ; %bb.2: ; %atomicrmw.end
@@ -3140,20 +3140,20 @@ define void @global_agent_atomic_fadd_noret_f32___amdgpu_no_fine_grained_memory_
; GFX7-NEXT: s_mov_b32 s7, 0xf000
; GFX7-NEXT: s_mov_b32 s4, s6
; GFX7-NEXT: s_mov_b32 s5, s6
-; GFX7-NEXT: buffer_load_dword v4, v[0:1], s[4:7], 0 addr64 offset:2044
+; GFX7-NEXT: buffer_load_dword v3, v[0:1], s[4:7], 0 addr64 offset:2044
; GFX7-NEXT: s_mov_b64 s[8:9], 0
; GFX7-NEXT: .LBB16_1: ; %atomicrmw.start
; GFX7-NEXT: ; =>This Inner Loop Header: Depth=1
; GFX7-NEXT: s_waitcnt vmcnt(0)
-; GFX7-NEXT: v_add_f32_e32 v3, v4, v2
-; GFX7-NEXT: v_mov_b32_e32 v6, v4
; GFX7-NEXT: v_mov_b32_e32 v5, v3
-; GFX7-NEXT: buffer_atomic_cmpswap v[5:6], v[0:1], s[4:7], 0 addr64 offset:2044 glc
+; GFX7-NEXT: v_add_f32_e32 v4, v5, v2
+; GFX7-NEXT: v_mov_b32_e32 v3, v4
+; GFX7-NEXT: v_mov_b32_e32 v4, v5
+; GFX7-NEXT: buffer_atomic_cmpswap v[3:4], v[0:1], s[4:7], 0 addr64 offset:2044 glc
; GFX7-NEXT: s_waitcnt vmcnt(0)
; GFX7-NEXT: buffer_wbinvl1
-; GFX7-NEXT: v_cmp_eq_u32_e32 vcc, v5, v4
+; GFX7-NEXT: v_cmp_eq_u32_e32 vcc, v3, v5
; GFX7-NEXT: s_or_b64 s[8:9], vcc, s[8:9]
-; GFX7-NEXT: v_mov_b32_e32 v4, v5
; GFX7-NEXT: s_andn2_b64 exec, exec, s[8:9]
; GFX7-NEXT: s_cbranch_execnz .LBB16_1
; GFX7-NEXT: ; %bb.2: ; %atomicrmw.end
@@ -3166,21 +3166,21 @@ define void @global_agent_atomic_fadd_noret_f32___amdgpu_no_fine_grained_memory_
; GFX6-NEXT: s_mov_b32 s7, 0xf000
; GFX6-NEXT: s_mov_b32 s4, s6
; GFX6-NEXT: s_mov_b32 s5, s6
-; GFX6-NEXT: buffer_load_dword v4, v[0:1], s[4:7], 0 addr64 offset:2044
+; GFX6-NEXT: buffer_load_dword v3, v[0:1], s[4:7], 0 addr64 offset:2044
; GFX6-NEXT: s_mov_b64 s[8:9], 0
; GFX6-NEXT: .LBB16_1: ; %atomicrmw.start
; GFX6-NEXT: ; =>This Inner Loop Header: Depth=1
; GFX6-NEXT: s_waitcnt vmcnt(0)
-; GFX6-NEXT: v_add_f32_e32 v3, v4, v2
-; GFX6-NEXT: s_waitcnt expcnt(0)
-; GFX6-NEXT: v_mov_b32_e32 v6, v4
; GFX6-NEXT: v_mov_b32_e32 v5, v3
-; GFX6-NEXT: buffer_atomic_cmpswap v[5:6], v[0:1], s[4:7], 0 addr64 offset:2044 glc
+; GFX6-NEXT: s_waitcnt expcnt(0)
+; GFX6-NEXT: v_add_f32_e32 v4, v5, v2
+; GFX6-NEXT: v_mov_b32_e32 v3, v4
+; GFX6-NEXT: v_mov_b32_e32 v4, v5
+; GFX6-NEXT: buffer_atomic_cmpswap v[3:4], v[0:1], s[4:7], 0 addr64 offset:2044 glc
; GFX6-NEXT: s_waitcnt vmcnt(0)
; GFX6-NEXT: buffer_wbinvl1
-; GFX6-NEXT: v_cmp_eq_u32_e32 vcc, v5, v4
+; GFX6-NEXT: v_cmp_eq_u32_e32 vcc, v3, v5
; GFX6-NEXT: s_or_b64 s[8:9], vcc, s[8:9]
-; GFX6-NEXT: v_mov_b32_e32 v4, v5
; GFX6-NEXT: s_andn2_b64 exec, exec, s[8:9]
; GFX6-NEXT: s_cbranch_execnz .LBB16_1
; GFX6-NEXT: ; %bb.2: ; %atomicrmw.end
@@ -3335,20 +3335,20 @@ define void @global_agent_atomic_fadd_noret_f32_amdgpu_ignore_denormal_mode(ptr
; GFX7-NEXT: s_mov_b32 s7, 0xf000
; GFX7-NEXT: s_mov_b32 s4, s6
; GFX7-NEXT: s_mov_b32 s5, s6
-; GFX7-NEXT: buffer_load_dword v4, v[0:1], s[4:7], 0 addr64 offset:2044
+; GFX7-NEXT: buffer_load_dword v3, v[0:1], s[4:7], 0 addr64 offset:2044
; GFX7-NEXT: s_mov_b64 s[8:9], 0
; GFX7-NEXT: .LBB17_1: ; %atomicrmw.start
; GFX7-NEXT: ; =>This Inner Loop Header: Depth=1
; GFX7-NEXT: s_waitcnt vmcnt(0)
-; GFX7-NEXT: v_add_f32_e32 v3, v4, v2
-; GFX7-NEXT: v_mov_b32_e32 v6, v4
; GFX7-NEXT: v_mov_b32_e32 v5, v3
-; GFX7-NEXT: buffer_atomic_cmpswap v[5:6], v[0:1], s[4:7], 0 addr64 offset:2044 glc
+; GFX7-NEXT: v_add_f32_e32 v4, v5, v2
+; GFX7-NEXT: v_mov_b32_e32 v3, v4
+; GFX7-NEXT: v_mov_b32_e32 v4, v5
+; GFX7-NEXT: buffer_atomic_cmpswap v[3:4], v[0:1], s[4:7], 0 addr64 offset:2044 glc
; GFX7-NEXT: s_waitcnt vmcnt(0)
; GFX7-NEXT: buffer_wbinvl1
-; GFX7-NEXT: v_cmp_eq_u32_e32 vcc, v5, v4
+; GFX7-NEXT: v_cmp_eq_u32_e32 vcc, v3, v5
; GFX7-NEXT: s_or_b64 s[8:9], vcc, s[8:9]
-; GFX7-NEXT: v_mov_b32_e32 v4, v5
; GFX7-NEXT: s_andn2_b64 exec, exec, s[8:9]
; GFX7-NEXT: s_cbranch_execnz .LBB17_1
; GFX7-NEXT: ; %bb.2: ; %atomicrmw.end
@@ -3361,21 +3361,21 @@ define void @global_agent_atomic_fadd_noret_f32_amdgpu_ignore_denormal_mode(ptr
; GFX6-NEXT: s_mov_b32 s7, 0xf000
; GFX6-NEXT: s_mov_b32 s4, s6
; GFX6-NEXT: s_mov_b32 s5, s6
-; GFX6-NEXT: buffer_load_dword v4, v[0:1], s[4:7], 0 addr64 offset:2044
+; GFX6-NEXT: buffer_load_dword v3, v[0:1], s[4:7], 0 addr64 offset:2044
; GFX6-NEXT: s_mov_b64 s[8:9], 0
; GFX6-NEXT: .LBB17_1: ; %atomicrmw.start
; GFX6-NEXT: ; =>This Inner Loop Header: Depth=1
; GFX6-NEXT: s_waitcnt vmcnt(0)
-; GFX6-NEXT: v_add_f32_e32 v3, v4, v2
-; GFX6-NEXT: s_waitcnt expcnt(0)
-; GFX6-NEXT: v_mov_b32_e32 v6, v4
; GFX6-NEXT: v_mov_b32_e32 v5, v3
-; GFX6-NEXT: buffer_atomic_cmpswap v[5:6], v[0:1], s[4:7], 0 addr64 offset:2044 glc
+; GFX6-NEXT: s_waitcnt expcnt(0)
+; GFX6-NEXT: v_add_f32_e32 v4, v5, v2
+; GFX6-NEXT: v_mov_b32_e32 v3, v4
+; GFX6-NEXT: v_mov_b32_e32 v4, v5
+; GFX6-NEXT: buffer_atomic_cmpswap v[3:4], v[0:1], s[4:7], 0 addr64 offset:2044 glc
; GFX6-NEXT: s_waitcnt vmcnt(0)
; GFX6-NEXT: buffer_wbinvl1
-; GFX6-NEXT: v_cmp_eq_u32_e32 vcc, v5, v4
+; GFX6-NEXT: v_cmp_eq_u32_e32 vcc, v3, v5
; GFX6-NEXT: s_or_b64 s[8:9], vcc, s[8:9]
-; GFX6-NEXT: v_mov_b32_e32 v4, v5
; GFX6-NEXT: s_andn2_b64 exec, exec, s[8:9]
; GFX6-NEXT: s_cbranch_execnz .LBB17_1
; GFX6-NEXT: ; %bb.2: ; %atomicrmw.end
@@ -3728,20 +3728,20 @@ define void @global_agent_atomic_fadd_noret_f32__amdgpu_no_remote_memory(ptr add
; GFX7-NEXT: s_mov_b32 s7, 0xf000
; GFX7-NEXT: s_mov_b32 s4, s6
; GFX7-NEXT: s_mov_b32 s5, s6
-; GFX7-NEXT: buffer_load_dword v4, v[0:1], s[4:7], 0 addr64
+; GFX7-NEXT: buffer_load_dword v3, v[0:1], s[4:7], 0 addr64
; GFX7-NEXT: s_mov_b64 s[8:9], 0
; GFX7-NEXT: .LBB19_1: ; %atomicrmw.start
; GFX7-NEXT: ; =>This Inner Loop Header: Depth=1
; GFX7-NEXT: s_waitcnt vmcnt(0)
-; GFX7-NEXT: v_add_f32_e32 v3, v4, v2
-; GFX7-NEXT: v_mov_b32_e32 v6, v4
; GFX7-NEXT: v_mov_b32_e32 v5, v3
-; GFX7-NEXT: buffer_atomic_cmpswap v[5:6], v[0:1], s[4:7], 0 addr64 glc
+; GFX7-NEXT: v_add_f32_e32 v4, v5, v2
+; GFX7-NEXT: v_mov_b32_e32 v3, v4
+; GFX7-NEXT: v_mov_b32_e32 v4, v5
+; GFX7-NEXT: buffer_atomic_cmpswap v[3:4], v[0:1], s[4:7], 0 addr64 glc
; GFX7-NEXT: s_waitcnt vmcnt(0)
; GFX7-NEXT: buffer_wbinvl1
-; GFX7-NEXT: v_cmp_eq_u32_e32 vcc, v5, v4
+; GFX7-NEXT: v_cmp_eq_u32_e32 vcc, v3, v5
; GFX7-NEXT: s_or_b64 s[8:9], vcc, s[8:9]
-; GFX7-NEXT: v_mov_b32_e32 v4, v5
; GFX7-NEXT: s_andn2_b64 exec, exec, s[8:9]
; GFX7-NEXT: s_cbranch_execnz .LBB19_1
; GFX7-NEXT: ; %bb.2: ; %atomicrmw.end
@@ -3754,21 +3754,21 @@ define void @global_agent_atomic_fadd_noret_f32__amdgpu_no_remote_memory(ptr add
; GFX6-NEXT: s_mov_b32 s7, 0xf000
; GFX6-NEXT: s_mov_b32 s4, s6
; GFX6-NEXT: s_mov_b32 s5, s6
-; GFX6-NEXT: buffer_load_dword v4, v[0:1], s[4:7], 0 addr64
+; GFX6-NEXT: buffer_load_dword v3, v[0:1], s[4:7], 0 addr64
; GFX6-NEXT: s_mov_b64 s[8:9], 0
; GFX6-NEXT: .LBB19_1: ; %atomicrmw.start
; GFX6-NEXT: ; =>This Inner Loop Header: Depth=1
; GFX6-NEXT: s_waitcnt vmcnt(0)
-; GFX6-NEXT: v_add_f32_e32 v3, v4, v2
-; GFX6-NEXT: s_waitcnt expcnt(0)
-; GFX6-NEXT: v_mov_b32_e32 v6, v4
; GFX6-NEXT: v_mov_b32_e32 v5, v3
-; GFX6-NEXT: buffer_atomic_cmpswap v[5:6], v[0:1], s[4:7], 0 addr64 glc
+; GFX6-NEXT: s_waitcnt expcnt(0)
+; GFX6-NEXT: v_add_f32_e32 v4, v5, v2
+; GFX6-NEXT: v_mov_b32_e32 v3, v4
+; GFX6-NEXT: v_mov_b32_e32 v4, v5
+; GFX6-NEXT: buffer_atomic_cmpswap v[3:4], v[0:1], s[4:7], 0 addr64 glc
; GFX6-NEXT: s_waitcnt vmcnt(0)
; GFX6-NEXT: buffer_wbinvl1
-; GFX6-NEXT: v_cmp_eq_u32_e32 vcc, v5, v4
+; GFX6-NEXT: v_cmp_eq_u32_e32 vcc, v3, v5
; GFX6-NEXT: s_or_b64 s[8:9], vcc, s[8:9]
-; GFX6-NEXT: v_mov_b32_e32 v4, v5
; GFX6-NEXT: s_andn2_b64 exec, exec, s[8:9]
; GFX6-NEXT: s_cbranch_execnz .LBB19_1
; GFX6-NEXT: ; %bb.2: ; %atomicrmw.end
@@ -4120,20 +4120,20 @@ define void @global_agent_atomic_fadd_noret_f32__amdgpu_no_remote_memory__amdgpu
; GFX7-NEXT: s_mov_b32 s7, 0xf000
; GFX7-NEXT: s_mov_b32 s4, s6
; GFX7-NEXT: s_mov_b32 s5, s6
-; GFX7-NEXT: buffer_load_dword v4, v[0:1], s[4:7], 0 addr64
+; GFX7-NEXT: buffer_load_dword v3, v[0:1], s[4:7], 0 addr64
; GFX7-NEXT: s_mov_b64 s[8:9], 0
; GFX7-NEXT: .LBB21_1: ; %atomicrmw.start
; GFX7-NEXT: ; =>This Inner Loop Header: Depth=1
; GFX7-NEXT: s_waitcnt vmcnt(0)
-; GFX7-NEXT: v_add_f32_e32 v3, v4, v2
-; GFX7-NEXT: v_mov_b32_e32 v6, v4
; GFX7-NEXT: v_mov_b32_e32 v5, v3
-; GFX7-NEXT: buffer_atomic_cmpswap v[5:6], v[0:1], s[4:7], 0 addr64 glc
+; GFX7-NEXT: v_add_f32_e32 v4, v5, v2
+; GFX7-NEXT: v_mov_b32_e32 v3, v4
+; GFX7-NEXT: v_mov_b32_e32 v4, v5
+; GFX7-NEXT: buffer_atomic_cmpswap v[3:4], v[0:1], s[4:7], 0 addr64 glc
; GFX7-NEXT: s_waitcnt vmcnt(0)
; GFX7-NEXT: buffer_wbinvl1
-; GFX7-NEXT: v_cmp_eq_u32_e32 vcc, v5, v4
+; GFX7-NEXT: v_cmp_eq_u32_e32 vcc, v3, v5
; GFX7-NEXT: s_or_b64 s[8:9], vcc, s[8:9]
-; GFX7-NEXT: v_mov_b32_e32 v4, v5
; GFX7-NEXT: s_andn2_b64 exec, exec, s[8:9]
; GFX7-NEXT: s_cbranch_execnz .LBB21_1
; GFX7-NEXT: ; %bb.2: ; %atomicrmw.end
@@ -4146,21 +4146,21 @@ define void @global_agent_atomic_fadd_noret_f32__amdgpu_no_remote_memory__amdgpu
; GFX6-NEXT: s_mov_b32 s7, 0xf000
; GFX6-NEXT: s_mov_b32 s4, s6
; GFX6-NEXT: s_mov_b32 s5, s6
-; GFX6-NEXT: buffer_load_dword v4, v[0:1], s[4:7], 0 addr64
+; GFX6-NEXT: buffer_load_dword v3, v[0:1], s[4:7], 0 addr64
; GFX6-NEXT: s_mov_b64 s[8:9], 0
; GFX6-NEXT: .LBB21_1: ; %atomicrmw.start
; GFX6-NEXT: ; =>This Inner Loop Header: Depth=1
; GFX6-NEXT: s_waitcnt vmcnt(0)
-; GFX6-NEXT: v_add_f32_e32 v3, v4, v2
-; GFX6-NEXT: s_waitcnt expcnt(0)
-; GFX6-NEXT: v_mov_b32_e32 v6, v4
; GFX6-NEXT: v_mov_b32_e32 v5, v3
-; GFX6-NEXT: buffer_atomic_cmpswap v[5:6], v[0:1], s[4:7], 0 addr64 glc
+; GFX6-NEXT: s_waitcnt expcnt(0)
+; GFX6-NEXT: v_add_f32_e32 v4, v5, v2
+; GFX6-NEXT: v_mov_b32_e32 v3, v4
+; GFX6-NEXT: v_mov_b32_e32 v4, v5
+; GFX6-NEXT: buffer_atomic_cmpswap v[3:4], v[0:1], s[4:7], 0 addr64 glc
; GFX6-NEXT: s_waitcnt vmcnt(0)
; GFX6-NEXT: buffer_wbinvl1
-; GFX6-NEXT: v_cmp_eq_u32_e32 vcc, v5, v4
+; GFX6-NEXT: v_cmp_eq_u32_e32 vcc, v3, v5
; GFX6-NEXT: s_or_b64 s[8:9], vcc, s[8:9]
-; GFX6-NEXT: v_mov_b32_e32 v4, v5
; GFX6-NEXT: s_andn2_b64 exec, exec, s[8:9]
; GFX6-NEXT: s_cbranch_execnz .LBB21_1
; GFX6-NEXT: ; %bb.2: ; %atomicrmw.end
@@ -4482,20 +4482,20 @@ define void @global_agent_atomic_fadd_noret_f32__amdgpu_no_remote_memory__amdgpu
; GFX7-NEXT: s_mov_b32 s7, 0xf000
; GFX7-NEXT: s_mov_b32 s4, s6
; GFX7-NEXT: s_mov_b32 s5, s6
-; GFX7-NEXT: buffer_load_dword v4, v[0:1], s[4:7], 0 addr64
+; GFX7-NEXT: buffer_load_dword v3, v[0:1], s[4:7], 0 addr64
; GFX7-NEXT: s_mov_b64 s[8:9], 0
; GFX7-NEXT: .LBB23_1: ; %atomicrmw.start
; GFX7-NEXT: ; =>This Inner Loop Header: Depth=1
; GFX7-NEXT: s_waitcnt vmcnt(0)
-; GFX7-NEXT: v_add_f32_e32 v3, v4, v2
-; GFX7-NEXT: v_mov_b32_e32 v6, v4
; GFX7-NEXT: v_mov_b32_e32 v5, v3
-; GFX7-NEXT: buffer_atomic_cmpswap v[5:6], v[0:1], s[4:7], 0 addr64 glc
+; GFX7-NEXT: v_add_f32_e32 v4, v5, v2
+; GFX7-NEXT: v_mov_b32_e32 v3, v4
+; GFX7-NEXT: v_mov_b32_e32 v4, v5
+; GFX7-NEXT: buffer_atomic_cmpswap v[3:4], v[0:1], s[4:7], 0 addr64 glc
; GFX7-NEXT: s_waitcnt vmcnt(0)
; GFX7-NEXT: buffer_wbinvl1
-; GFX7-NEXT: v_cmp_eq_u32_e32 vcc, v5, v4
+; GFX7-NEXT: v_cmp_eq_u32_e32 vcc, v3, v5
; GFX7-NEXT: s_or_b64 s[8:9], vcc, s[8:9]
-; GFX7-NEXT: v_mov_b32_e32 v4, v5
; GFX7-NEXT: s_andn2_b64 exec, exec, s[8:9]
; GFX7-NEXT: s_cbranch_execnz .LBB23_1
; GFX7-NEXT: ; %bb.2: ; %atomicrmw.end
@@ -4508,21 +4508,21 @@ define void @global_agent_atomic_fadd_noret_f32__amdgpu_no_remote_memory__amdgpu
; GFX6-NEXT: s_mov_b32 s7, 0xf000
; GFX6-NEXT: s_mov_b32 s4, s6
; GFX6-NEXT: s_mov_b32 s5, s6
-; GFX6-NEXT: buffer_load_dword v4, v[0:1], s[4:7], 0 addr64
+; GFX6-NEXT: buffer_load_dword v3, v[0:1], s[4:7], 0 addr64
; GFX6-NEXT: s_mov_b64 s[8:9], 0
; GFX6-NEXT: .LBB23_1: ; %atomicrmw.start
; GFX6-NEXT: ; =>This Inner Loop Header: Depth=1
; GFX6-NEXT: s_waitcnt vmcnt(0)
-; GFX6-NEXT: v_add_f32_e32 v3, v4, v2
-; GFX6-NEXT: s_waitcnt expcnt(0)
-; GFX6-NEXT: v_mov_b32_e32 v6, v4
; GFX6-NEXT: v_mov_b32_e32 v5, v3
-; GFX6-NEXT: buffer_atomic_cmpswap v[5:6], v[0:1], s[4:7], 0 addr64 glc
+; GFX6-NEXT: s_waitcnt expcnt(0)
+; GFX6-NEXT: v_add_f32_e32 v4, v5, v2
+; GFX6-NEXT: v_mov_b32_e32 v3, v4
+; GFX6-NEXT: v_mov_b32_e32 v4, v5
+; GFX6-NEXT: buffer_atomic_cmpswap v[3:4], v[0:1], s[4:7], 0 addr64 glc
; GFX6-NEXT: s_waitcnt vmcnt(0)
; GFX6-NEXT: buffer_wbinvl1
-; GFX6-NEXT: v_cmp_eq_u32_e32 vcc, v5, v4
+; GFX6-NEXT: v_cmp_eq_u32_e32 vcc, v3, v5
; GFX6-NEXT: s_or_b64 s[8:9], vcc, s[8:9]
-; GFX6-NEXT: v_mov_b32_e32 v4, v5
; GFX6-NEXT: s_andn2_b64 exec, exec, s[8:9]
; GFX6-NEXT: s_cbranch_execnz .LBB23_1
; GFX6-NEXT: ; %bb.2: ; %atomicrmw.end
@@ -5162,20 +5162,20 @@ define void @global_agent_atomic_fadd_noret_f32__ftz__amdgpu_no_fine_grained_mem
; GFX7-NEXT: s_mov_b32 s7, 0xf000
; GFX7-NEXT: s_mov_b32 s4, s6
; GFX7-NEXT: s_mov_b32 s5, s6
-; GFX7-NEXT: buffer_load_dword v4, v[0:1], s[4:7], 0 addr64
+; GFX7-NEXT: buffer_load_dword v3, v[0:1], s[4:7], 0 addr64
; GFX7-NEXT: s_mov_b64 s[8:9], 0
; GFX7-NEXT: .LBB27_1: ; %atomicrmw.start
; GFX7-NEXT: ; =>This Inner Loop Header: Depth=1
; GFX7-NEXT: s_waitcnt vmcnt(0)
-; GFX7-NEXT: v_add_f32_e32 v3, v4, v2
-; GFX7-NEXT: v_mov_b32_e32 v6, v4
; GFX7-NEXT: v_mov_b32_e32 v5, v3
-; GFX7-NEXT: buffer_atomic_cmpswap v[5:6], v[0:1], s[4:7], 0 addr64 glc
+; GFX7-NEXT: v_add_f32_e32 v4, v5, v2
+; GFX7-NEXT: v_mov_b32_e32 v3, v4
+; GFX7-NEXT: v_mov_b32_e32 v4, v5
+; GFX7-NEXT: buffer_atomic_cmpswap v[3:4], v[0:1], s[4:7], 0 addr64 glc
; GFX7-NEXT: s_waitcnt vmcnt(0)
; GFX7-NEXT: buffer_wbinvl1
-; GFX7-NEXT: v_cmp_eq_u32_e32 vcc, v5, v4
+; GFX7-NEXT: v_cmp_eq_u32_e32 vcc, v3, v5
; GFX7-NEXT: s_or_b64 s[8:9], vcc, s[8:9]
-; GFX7-NEXT: v_mov_b32_e32 v4, v5
; GFX7-NEXT: s_andn2_b64 exec, exec, s[8:9]
; GFX7-NEXT: s_cbranch_execnz .LBB27_1
; GFX7-NEXT: ; %bb.2: ; %atomicrmw.end
@@ -5188,21 +5188,21 @@ define void @global_agent_atomic_fadd_noret_f32__ftz__amdgpu_no_fine_grained_mem
; GFX6-NEXT: s_mov_b32 s7, 0xf000
; GFX6-NEXT: s_mov_b32 s4, s6
; GFX6-NEXT: s_mov_b32 s5, s6
-; GFX6-NEXT: buffer_load_dword v4, v[0:1], s[4:7], 0 addr64
+; GFX6-NEXT: buffer_load_dword v3, v[0:1], s[4:7], 0 addr64
; GFX6-NEXT: s_mov_b64 s[8:9], 0
; GFX6-NEXT: .LBB27_1: ; %atomicrmw.start
; GFX6-NEXT: ; =>This Inner Loop Header: Depth=1
; GFX6-NEXT: s_waitcnt vmcnt(0)
-; GFX6-NEXT: v_add_f32_e32 v3, v4, v2
-; GFX6-NEXT: s_waitcnt expcnt(0)
-; GFX6-NEXT: v_mov_b32_e32 v6, v4
; GFX6-NEXT: v_mov_b32_e32 v5, v3
-; GFX6-NEXT: buffer_atomic_cmpswap v[5:6], v[0:1], s[4:7], 0 addr64 glc
+; GFX6-NEXT: s_waitcnt expcnt(0)
+; GFX6-NEXT: v_add_f32_e32 v4, v5, v2
+; GFX6-NEXT: v_mov_b32_e32 v3, v4
+; GFX6-NEXT: v_mov_b32_e32 v4, v5
+; GFX6-NEXT: buffer_atomic_cmpswap v[3:4], v[0:1], s[4:7], 0 addr64 glc
; GFX6-NEXT: s_waitcnt vmcnt(0)
; GFX6-NEXT: buffer_wbinvl1
-; GFX6-NEXT: v_cmp_eq_u32_e32 vcc, v5, v4
+; GFX6-NEXT: v_cmp_eq_u32_e32 vcc, v3, v5
; GFX6-NEXT: s_or_b64 s[8:9], vcc, s[8:9]
-; GFX6-NEXT: v_mov_b32_e32 v4, v5
; GFX6-NEXT: s_andn2_b64 exec, exec, s[8:9]
; GFX6-NEXT: s_cbranch_execnz .LBB27_1
; GFX6-NEXT: ; %bb.2: ; %atomicrmw.end
@@ -5316,20 +5316,20 @@ define void @global_agent_atomic_fadd_noret_f32__offset12b_pos__ftz__amdgpu_no_f
; GFX7-NEXT: s_mov_b32 s7, 0xf000
; GFX7-NEXT: s_mov_b32 s4, s6
; GFX7-NEXT: s_mov_b32 s5, s6
-; GFX7-NEXT: buffer_load_dword v4, v[0:1], s[4:7], 0 addr64 offset:2044
+; GFX7-NEXT: buffer_load_dword v3, v[0:1], s[4:7], 0 addr64 offset:2044
; GFX7-NEXT: s_mov_b64 s[8:9], 0
; GFX7-NEXT: .LBB28_1: ; %atomicrmw.start
; GFX7-NEXT: ; =>This Inner Loop Header: Depth=1
; GFX7-NEXT: s_waitcnt vmcnt(0)
-; GFX7-NEXT: v_add_f32_e32 v3, v4, v2
-; GFX7-NEXT: v_mov_b32_e32 v6, v4
; GFX7-NEXT: v_mov_b32_e32 v5, v3
-; GFX7-NEXT: buffer_atomic_cmpswap v[5:6], v[0:1], s[4:7], 0 addr64 offset:2044 glc
+; GFX7-NEXT: v_add_f32_e32 v4, v5, v2
+; GFX7-NEXT: v_mov_b32_e32 v3, v4
+; GFX7-NEXT: v_mov_b32_e32 v4, v5
+; GFX7-NEXT: buffer_atomic_cmpswap v[3:4], v[0:1], s[4:7], 0 addr64 offset:2044 glc
; GFX7-NEXT: s_waitcnt vmcnt(0)
; GFX7-NEXT: buffer_wbinvl1
-; GFX7-NEXT: v_cmp_eq_u32_e32 vcc, v5, v4
+; GFX7-NEXT: v_cmp_eq_u32_e32 vcc, v3, v5
; GFX7-NEXT: s_or_b64 s[8:9], vcc, s[8:9]
-; GFX7-NEXT: v_mov_b32_e32 v4, v5
; GFX7-NEXT: s_andn2_b64 exec, exec, s[8:9]
; GFX7-NEXT: s_cbranch_execnz .LBB28_1
; GFX7-NEXT: ; %bb.2: ; %atomicrmw.end
@@ -5342,21 +5342,21 @@ define void @global_agent_atomic_fadd_noret_f32__offset12b_pos__ftz__amdgpu_no_f
; GFX6-NEXT: s_mov_b32 s7, 0xf000
; GFX6-NEXT: s_mov_b32 s4, s6
; GFX6-NEXT: s_mov_b32 s5, s6
-; GFX6-NEXT: buffer_load_dword v4, v[0:1], s[4:7], 0 addr64 offset:2044
+; GFX6-NEXT: buffer_load_dword v3, v[0:1], s[4:7], 0 addr64 offset:2044
; GFX6-NEXT: s_mov_b64 s[8:9], 0
; GFX6-NEXT: .LBB28_1: ; %atomicrmw.start
; GFX6-NEXT: ; =>This Inner Loop Header: Depth=1
; GFX6-NEXT: s_waitcnt vmcnt(0)
-; GFX6-NEXT: v_add_f32_e32 v3, v4, v2
+; GFX6-NEXT: v_mov_b32_e32 v5, v3
; GFX6-NEXT: s_waitcnt expcnt(0)
-; GFX6-NEXT: v_mov_b32_e32 v6, v4
-; GFX6-NEXT: v_mov_b32_e32 v5, v3
-; GFX6-NEXT: buffer_atomic_cmpswap v[5:6], v[0:1], s[4:7], 0 addr64 offset:2044 glc
+; GFX6-NEXT: v_add_f32_e32 v4, v5, v2
+; GFX6-NEXT: v_mov_b32_e32 v3, v4
+; GFX6-NEXT: v_mov_b32_e32 v4, v5
+; GFX6-NEXT: buffer_atomic_cmpswap v[3:4], v[0:1], s[4:7], 0 addr64 offset:2044 glc
; GFX6-NEXT: s_waitcnt vmcnt(0)
; GFX6-NEXT: buffer_wbinvl1
-; GFX6-NEXT: v_cmp_eq_u32_e32 vcc, v5, v4
+; GFX6-NEXT: v_cmp_eq_u32_e32 vcc, v3, v5
; GFX6-NEXT: s_or_b64 s[8:9], vcc, s[8:9]
-; GFX6-NEXT: v_mov_b32_e32 v4, v5
; GFX6-NEXT: s_andn2_b64 exec, exec, s[8:9]
; GFX6-NEXT: s_cbranch_execnz .LBB28_1
; GFX6-NEXT: ; %bb.2: ; %atomicrmw.end
@@ -5471,7 +5471,7 @@ define void @global_agent_atomic_fadd_noret_f32__offset12b_neg__ftz__amdgpu_no_f
; GFX7-NEXT: s_mov_b32 s5, -1
; GFX7-NEXT: s_mov_b32 s7, 0xf000
; GFX7-NEXT: s_mov_b32 s6, 0
-; GFX7-NEXT: buffer_load_dword v4, v[0:1], s[4:7], 0 addr64
+; GFX7-NEXT: buffer_load_dword v3, v[0:1], s[4:7], 0 addr64
; GFX7-NEXT: v_add_i32_e32 v0, vcc, 0xfffff800, v0
; GFX7-NEXT: v_addc_u32_e32 v1, vcc, -1, v1, vcc
; GFX7-NEXT: s_mov_b64 s[8:9], 0
@@ -5480,15 +5480,15 @@ define void @global_agent_atomic_fadd_noret_f32__offset12b_neg__ftz__amdgpu_no_f
; GFX7-NEXT: .LBB29_1: ; %atomicrmw.start
; GFX7-NEXT: ; =>This Inner Loop Header: Depth=1
; GFX7-NEXT: s_waitcnt vmcnt(0)
-; GFX7-NEXT: v_add_f32_e32 v3, v4, v2
-; GFX7-NEXT: v_mov_b32_e32 v6, v4
; GFX7-NEXT: v_mov_b32_e32 v5, v3
-; GFX7-NEXT: buffer_atomic_cmpswap v[5:6], v[0:1], s[4:7], 0 addr64 glc
+; GFX7-NEXT: v_add_f32_e32 v4, v5, v2
+; GFX7-NEXT: v_mov_b32_e32 v3, v4
+; GFX7-NEXT: v_mov_b32_e32 v4, v5
+; GFX7-NEXT: buffer_atomic_cmpswap v[3:4], v[0:1], s[4:7], 0 addr64 glc
; GFX7-NEXT: s_waitcnt vmcnt(0)
; GFX7-NEXT: buffer_wbinvl1
-; GFX7-NEXT: v_cmp_eq_u32_e32 vcc, v5, v4
+; GFX7-NEXT: v_cmp_eq_u32_e32 vcc, v3, v5
; GFX7-NEXT: s_or_b64 s[8:9], vcc, s[8:9]
-; GFX7-NEXT: v_mov_b32_e32 v4, v5
; GFX7-NEXT: s_andn2_b64 exec, exec, s[8:9]
; GFX7-NEXT: s_cbranch_execnz .LBB29_1
; GFX7-NEXT: ; %bb.2: ; %atomicrmw.end
@@ -5501,7 +5501,7 @@ define void @global_agent_atomic_fadd_noret_f32__offset12b_neg__ftz__amdgpu_no_f
; GFX6-NEXT: s_mov_b32 s5, -1
; GFX6-NEXT: s_mov_b32 s7, 0xf000
; GFX6-NEXT: s_mov_b32 s6, 0
-; GFX6-NEXT: buffer_load_dword v4, v[0:1], s[4:7], 0 addr64
+; GFX6-NEXT: buffer_load_dword v3, v[0:1], s[4:7], 0 addr64
; GFX6-NEXT: v_add_i32_e32 v0, vcc, 0xfffff800, v0
; GFX6-NEXT: v_addc_u32_e32 v1, vcc, -1, v1, vcc
; GFX6-NEXT: s_mov_b64 s[8:9], 0
@@ -5510,16 +5510,16 @@ define void @global_agent_atomic_fadd_noret_f32__offset12b_neg__ftz__amdgpu_no_f
; GFX6-NEXT: .LBB29_1: ; %atomicrmw.start
; GFX6-NEXT: ; =>This Inner Loop Header: Depth=1
; GFX6-NEXT: s_waitcnt vmcnt(0)
-; GFX6-NEXT: v_add_f32_e32 v3, v4, v2
-; GFX6-NEXT: s_waitcnt expcnt(0)
-; GFX6-NEXT: v_mov_b32_e32 v6, v4
; GFX6-NEXT: v_mov_b32_e32 v5, v3
-; GFX6-NEXT: buffer_atomic_cmpswap v[5:6], v[0:1], s[4:7], 0 addr64 glc
+; GFX6-NEXT: s_waitcnt expcnt(0)
+; GFX6-NEXT: v_add_f32_e32 v4, v5, v2
+; GFX6-NEXT: v_mov_b32_e32 v3, v4
+; GFX6-NEXT: v_mov_b32_e32 v4, v5
+; GFX6-NEXT: buffer_atomic_cmpswap v[3:4], v[0:1], s[4:7], 0 addr64 glc
; GFX6-NEXT: s_waitcnt vmcnt(0)
; GFX6-NEXT: buffer_wbinvl1
-; GFX6-NEXT: v_cmp_eq_u32_e32 vcc, v5, v4
+; GFX6-NEXT: v_cmp_eq_u32_e32 vcc, v3, v5
; GFX6-NEXT: s_or_b64 s[8:9], vcc, s[8:9]
-; GFX6-NEXT: v_mov_b32_e32 v4, v5
; GFX6-NEXT: s_andn2_b64 exec, exec, s[8:9]
; GFX6-NEXT: s_cbranch_execnz .LBB29_1
; GFX6-NEXT: ; %bb.2: ; %atomicrmw.end
@@ -5812,20 +5812,20 @@ define void @global_system_atomic_fadd_noret_f32__offset12b_pos__ftz__amdgpu_no_
; GFX7-NEXT: s_mov_b32 s7, 0xf000
; GFX7-NEXT: s_mov_b32 s4, s6
; GFX7-NEXT: s_mov_b32 s5, s6
-; GFX7-NEXT: buffer_load_dword v4, v[0:1], s[4:7], 0 addr64 offset:2044
+; GFX7-NEXT: buffer_load_dword v3, v[0:1], s[4:7], 0 addr64 offset:2044
; GFX7-NEXT: s_mov_b64 s[8:9], 0
; GFX7-NEXT: .LBB31_1: ; %atomicrmw.start
; GFX7-NEXT: ; =>This Inner Loop Header: Depth=1
; GFX7-NEXT: s_waitcnt vmcnt(0)
-; GFX7-NEXT: v_add_f32_e32 v3, v4, v2
-; GFX7-NEXT: v_mov_b32_e32 v6, v4
; GFX7-NEXT: v_mov_b32_e32 v5, v3
-; GFX7-NEXT: buffer_atomic_cmpswap v[5:6], v[0:1], s[4:7], 0 addr64 offset:2044 glc
+; GFX7-NEXT: v_add_f32_e32 v4, v5, v2
+; GFX7-NEXT: v_mov_b32_e32 v3, v4
+; GFX7-NEXT: v_mov_b32_e32 v4, v5
+; GFX7-NEXT: buffer_atomic_cmpswap v[3:4], v[0:1], s[4:7], 0 addr64 offset:2044 glc
; GFX7-NEXT: s_waitcnt vmcnt(0)
; GFX7-NEXT: buffer_wbinvl1
-; GFX7-NEXT: v_cmp_eq_u32_e32 vcc, v5, v4
+; GFX7-NEXT: v_cmp_eq_u32_e32 vcc, v3, v5
; GFX7-NEXT: s_or_b64 s[8:9], vcc, s[8:9]
-; GFX7-NEXT: v_mov_b32_e32 v4, v5
; GFX7-NEXT: s_andn2_b64 exec, exec, s[8:9]
; GFX7-NEXT: s_cbranch_execnz .LBB31_1
; GFX7-NEXT: ; %bb.2: ; %atomicrmw.end
@@ -5838,21 +5838,21 @@ define void @global_system_atomic_fadd_noret_f32__offset12b_pos__ftz__amdgpu_no_
; GFX6-NEXT: s_mov_b32 s7, 0xf000
; GFX6-NEXT: s_mov_b32 s4, s6
; GFX6-NEXT: s_mov_b32 s5, s6
-; GFX6-NEXT: buffer_load_dword v4, v[0:1], s[4:7], 0 addr64 offset:2044
+; GFX6-NEXT: buffer_load_dword v3, v[0:1], s[4:7], 0 addr64 offset:2044
; GFX6-NEXT: s_mov_b64 s[8:9], 0
; GFX6-NEXT: .LBB31_1: ; %atomicrmw.start
; GFX6-NEXT: ; =>This Inner Loop Header: Depth=1
; GFX6-NEXT: s_waitcnt vmcnt(0)
-; GFX6-NEXT: v_add_f32_e32 v3, v4, v2
-; GFX6-NEXT: s_waitcnt expcnt(0)
-; GFX6-NEXT: v_mov_b32_e32 v6, v4
; GFX6-NEXT: v_mov_b32_e32 v5, v3
-; GFX6-NEXT: buffer_atomic_cmpswap v[5:6], v[0:1], s[4:7], 0 addr64 offset:2044 glc
+; GFX6-NEXT: s_waitcnt expcnt(0)
+; GFX6-NEXT: v_add_f32_e32 v4, v5, v2
+; GFX6-NEXT: v_mov_b32_e32 v3, v4
+; GFX6-NEXT: v_mov_b32_e32 v4, v5
+; GFX6-NEXT: buffer_atomic_cmpswap v[3:4], v[0:1], s[4:7], 0 addr64 offset:2044 glc
; GFX6-NEXT: s_waitcnt vmcnt(0)
; GFX6-NEXT: buffer_wbinvl1
-; GFX6-NEXT: v_cmp_eq_u32_e32 vcc, v5, v4
+; GFX6-NEXT: v_cmp_eq_u32_e32 vcc, v3, v5
; GFX6-NEXT: s_or_b64 s[8:9], vcc, s[8:9]
-; GFX6-NEXT: v_mov_b32_e32 v4, v5
; GFX6-NEXT: s_andn2_b64 exec, exec, s[8:9]
; GFX6-NEXT: s_cbranch_execnz .LBB31_1
; GFX6-NEXT: ; %bb.2: ; %atomicrmw.end
@@ -6139,20 +6139,20 @@ define void @global_agent_atomic_fadd_noret_f32__offset12b_pos__ieee__amdgpu_no_
; GFX7-NEXT: s_mov_b32 s7, 0xf000
; GFX7-NEXT: s_mov_b32 s4, s6
; GFX7-NEXT: s_mov_b32 s5, s6
-; GFX7-NEXT: buffer_load_dword v4, v[0:1], s[4:7], 0 addr64 offset:2044
+; GFX7-NEXT: buffer_load_dword v3, v[0:1], s[4:7], 0 addr64 offset:2044
; GFX7-NEXT: s_mov_b64 s[8:9], 0
; GFX7-NEXT: .LBB33_1: ; %atomicrmw.start
; GFX7-NEXT: ; =>This Inner Loop Header: Depth=1
; GFX7-NEXT: s_waitcnt vmcnt(0)
-; GFX7-NEXT: v_add_f32_e32 v3, v4, v2
-; GFX7-NEXT: v_mov_b32_e32 v6, v4
; GFX7-NEXT: v_mov_b32_e32 v5, v3
-; GFX7-NEXT: buffer_atomic_cmpswap v[5:6], v[0:1], s[4:7], 0 addr64 offset:2044 glc
+; GFX7-NEXT: v_add_f32_e32 v4, v5, v2
+; GFX7-NEXT: v_mov_b32_e32 v3, v4
+; GFX7-NEXT: v_mov_b32_e32 v4, v5
+; GFX7-NEXT: buffer_atomic_cmpswap v[3:4], v[0:1], s[4:7], 0 addr64 offset:2044 glc
; GFX7-NEXT: s_waitcnt vmcnt(0)
; GFX7-NEXT: buffer_wbinvl1
-; GFX7-NEXT: v_cmp_eq_u32_e32 vcc, v5, v4
+; GFX7-NEXT: v_cmp_eq_u32_e32 vcc, v3, v5
; GFX7-NEXT: s_or_b64 s[8:9], vcc, s[8:9]
-; GFX7-NEXT: v_mov_b32_e32 v4, v5
; GFX7-NEXT: s_andn2_b64 exec, exec, s[8:9]
; GFX7-NEXT: s_cbranch_execnz .LBB33_1
; GFX7-NEXT: ; %bb.2: ; %atomicrmw.end
@@ -6165,21 +6165,21 @@ define void @global_agent_atomic_fadd_noret_f32__offset12b_pos__ieee__amdgpu_no_
; GFX6-NEXT: s_mov_b32 s7, 0xf000
; GFX6-NEXT: s_mov_b32 s4, s6
; GFX6-NEXT: s_mov_b32 s5, s6
-; GFX6-NEXT: buffer_load_dword v4, v[0:1], s[4:7], 0 addr64 offset:2044
+; GFX6-NEXT: buffer_load_dword v3, v[0:1], s[4:7], 0 addr64 offset:2044
; GFX6-NEXT: s_mov_b64 s[8:9], 0
; GFX6-NEXT: .LBB33_1: ; %atomicrmw.start
; GFX6-NEXT: ; =>This Inner Loop Header: Depth=1
; GFX6-NEXT: s_waitcnt vmcnt(0)
-; GFX6-NEXT: v_add_f32_e32 v3, v4, v2
-; GFX6-NEXT: s_waitcnt expcnt(0)
-; GFX6-NEXT: v_mov_b32_e32 v6, v4
; GFX6-NEXT: v_mov_b32_e32 v5, v3
-; GFX6-NEXT: buffer_atomic_cmpswap v[5:6], v[0:1], s[4:7], 0 addr64 offset:2044 glc
+; GFX6-NEXT: s_waitcnt expcnt(0)
+; GFX6-NEXT: v_add_f32_e32 v4, v5, v2
+; GFX6-NEXT: v_mov_b32_e32 v3, v4
+; GFX6-NEXT: v_mov_b32_e32 v4, v5
+; GFX6-NEXT: buffer_atomic_cmpswap v[3:4], v[0:1], s[4:7], 0 addr64 offset:2044 glc
; GFX6-NEXT: s_waitcnt vmcnt(0)
; GFX6-NEXT: buffer_wbinvl1
-; GFX6-NEXT: v_cmp_eq_u32_e32 vcc, v5, v4
+; GFX6-NEXT: v_cmp_eq_u32_e32 vcc, v3, v5
; GFX6-NEXT: s_or_b64 s[8:9], vcc, s[8:9]
-; GFX6-NEXT: v_mov_b32_e32 v4, v5
; GFX6-NEXT: s_andn2_b64 exec, exec, s[8:9]
; GFX6-NEXT: s_cbranch_execnz .LBB33_1
; GFX6-NEXT: ; %bb.2: ; %atomicrmw.end
@@ -6532,20 +6532,20 @@ define void @global_agent_atomic_fadd_noret_f32__ftz__amdgpu_no_remote_memory(pt
; GFX7-NEXT: s_mov_b32 s7, 0xf000
; GFX7-NEXT: s_mov_b32 s4, s6
; GFX7-NEXT: s_mov_b32 s5, s6
-; GFX7-NEXT: buffer_load_dword v4, v[0:1], s[4:7], 0 addr64
+; GFX7-NEXT: buffer_load_dword v3, v[0:1], s[4:7], 0 addr64
; GFX7-NEXT: s_mov_b64 s[8:9], 0
; GFX7-NEXT: .LBB35_1: ; %atomicrmw.start
; GFX7-NEXT: ; =>This Inner Loop Header: Depth=1
; GFX7-NEXT: s_waitcnt vmcnt(0)
-; GFX7-NEXT: v_add_f32_e32 v3, v4, v2
-; GFX7-NEXT: v_mov_b32_e32 v6, v4
; GFX7-NEXT: v_mov_b32_e32 v5, v3
-; GFX7-NEXT: buffer_atomic_cmpswap v[5:6], v[0:1], s[4:7], 0 addr64 glc
+; GFX7-NEXT: v_add_f32_e32 v4, v5, v2
+; GFX7-NEXT: v_mov_b32_e32 v3, v4
+; GFX7-NEXT: v_mov_b32_e32 v4, v5
+; GFX7-NEXT: buffer_atomic_cmpswap v[3:4], v[0:1], s[4:7], 0 addr64 glc
; GFX7-NEXT: s_waitcnt vmcnt(0)
; GFX7-NEXT: buffer_wbinvl1
-; GFX7-NEXT: v_cmp_eq_u32_e32 vcc, v5, v4
+; GFX7-NEXT: v_cmp_eq_u32_e32 vcc, v3, v5
; GFX7-NEXT: s_or_b64 s[8:9], vcc, s[8:9]
-; GFX7-NEXT: v_mov_b32_e32 v4, v5
; GFX7-NEXT: s_andn2_b64 exec, exec, s[8:9]
; GFX7-NEXT: s_cbranch_execnz .LBB35_1
; GFX7-NEXT: ; %bb.2: ; %atomicrmw.end
@@ -6558,21 +6558,21 @@ define void @global_agent_atomic_fadd_noret_f32__ftz__amdgpu_no_remote_memory(pt
; GFX6-NEXT: s_mov_b32 s7, 0xf000
; GFX6-NEXT: s_mov_b32 s4, s6
; GFX6-NEXT: s_mov_b32 s5, s6
-; GFX6-NEXT: buffer_load_dword v4, v[0:1], s[4:7], 0 addr64
+; GFX6-NEXT: buffer_load_dword v3, v[0:1], s[4:7], 0 addr64
; GFX6-NEXT: s_mov_b64 s[8:9], 0
; GFX6-NEXT: .LBB35_1: ; %atomicrmw.start
; GFX6-NEXT: ; =>This Inner Loop Header: Depth=1
; GFX6-NEXT: s_waitcnt vmcnt(0)
-; GFX6-NEXT: v_add_f32_e32 v3, v4, v2
-; GFX6-NEXT: s_waitcnt expcnt(0)
-; GFX6-NEXT: v_mov_b32_e32 v6, v4
; GFX6-NEXT: v_mov_b32_e32 v5, v3
-; GFX6-NEXT: buffer_atomic_cmpswap v[5:6], v[0:1], s[4:7], 0 addr64 glc
+; GFX6-NEXT: s_waitcnt expcnt(0)
+; GFX6-NEXT: v_add_f32_e32 v4, v5, v2
+; GFX6-NEXT: v_mov_b32_e32 v3, v4
+; GFX6-NEXT: v_mov_b32_e32 v4, v5
+; GFX6-NEXT: buffer_atomic_cmpswap v[3:4], v[0:1], s[4:7], 0 addr64 glc
; GFX6-NEXT: s_waitcnt vmcnt(0)
; GFX6-NEXT: buffer_wbinvl1
-; GFX6-NEXT: v_cmp_eq_u32_e32 vcc, v5, v4
+; GFX6-NEXT: v_cmp_eq_u32_e32 vcc, v3, v5
; GFX6-NEXT: s_or_b64 s[8:9], vcc, s[8:9]
-; GFX6-NEXT: v_mov_b32_e32 v4, v5
; GFX6-NEXT: s_andn2_b64 exec, exec, s[8:9]
; GFX6-NEXT: s_cbranch_execnz .LBB35_1
; GFX6-NEXT: ; %bb.2: ; %atomicrmw.end
@@ -6854,20 +6854,20 @@ define void @global_agent_atomic_fadd_noret_f32__ftz__amdgpu_no_fine_grained_mem
; GFX7-NEXT: s_mov_b32 s7, 0xf000
; GFX7-NEXT: s_mov_b32 s4, s6
; GFX7-NEXT: s_mov_b32 s5, s6
-; GFX7-NEXT: buffer_load_dword v4, v[0:1], s[4:7], 0 addr64
+; GFX7-NEXT: buffer_load_dword v3, v[0:1], s[4:7], 0 addr64
; GFX7-NEXT: s_mov_b64 s[8:9], 0
; GFX7-NEXT: .LBB37_1: ; %atomicrmw.start
; GFX7-NEXT: ; =>This Inner Loop Header: Depth=1
; GFX7-NEXT: s_waitcnt vmcnt(0)
-; GFX7-NEXT: v_add_f32_e32 v3, v4, v2
-; GFX7-NEXT: v_mov_b32_e32 v6, v4
; GFX7-NEXT: v_mov_b32_e32 v5, v3
-; GFX7-NEXT: buffer_atomic_cmpswap v[5:6], v[0:1], s[4:7], 0 addr64 glc
+; GFX7-NEXT: v_add_f32_e32 v4, v5, v2
+; GFX7-NEXT: v_mov_b32_e32 v3, v4
+; GFX7-NEXT: v_mov_b32_e32 v4, v5
+; GFX7-NEXT: buffer_atomic_cmpswap v[3:4], v[0:1], s[4:7], 0 addr64 glc
; GFX7-NEXT: s_waitcnt vmcnt(0)
; GFX7-NEXT: buffer_wbinvl1
-; GFX7-NEXT: v_cmp_eq_u32_e32 vcc, v5, v4
+; GFX7-NEXT: v_cmp_eq_u32_e32 vcc, v3, v5
; GFX7-NEXT: s_or_b64 s[8:9], vcc, s[8:9]
-; GFX7-NEXT: v_mov_b32_e32 v4, v5
; GFX7-NEXT: s_andn2_b64 exec, exec, s[8:9]
; GFX7-NEXT: s_cbranch_execnz .LBB37_1
; GFX7-NEXT: ; %bb.2: ; %atomicrmw.end
@@ -6880,21 +6880,21 @@ define void @global_agent_atomic_fadd_noret_f32__ftz__amdgpu_no_fine_grained_mem
; GFX6-NEXT: s_mov_b32 s7, 0xf000
; GFX6-NEXT: s_mov_b32 s4, s6
; GFX6-NEXT: s_mov_b32 s5, s6
-; GFX6-NEXT: buffer_load_dword v4, v[0:1], s[4:7], 0 addr64
+; GFX6-NEXT: buffer_load_dword v3, v[0:1], s[4:7], 0 addr64
; GFX6-NEXT: s_mov_b64 s[8:9], 0
; GFX6-NEXT: .LBB37_1: ; %atomicrmw.start
; GFX6-NEXT: ; =>This Inner Loop Header: Depth=1
; GFX6-NEXT: s_waitcnt vmcnt(0)
-; GFX6-NEXT: v_add_f32_e32 v3, v4, v2
-; GFX6-NEXT: s_waitcnt expcnt(0)
-; GFX6-NEXT: v_mov_b32_e32 v6, v4
; GFX6-NEXT: v_mov_b32_e32 v5, v3
-; GFX6-NEXT: buffer_atomic_cmpswap v[5:6], v[0:1], s[4:7], 0 addr64 glc
+; GFX6-NEXT: s_waitcnt expcnt(0)
+; GFX6-NEXT: v_add_f32_e32 v4, v5, v2
+; GFX6-NEXT: v_mov_b32_e32 v3, v4
+; GFX6-NEXT: v_mov_b32_e32 v4, v5
+; GFX6-NEXT: buffer_atomic_cmpswap v[3:4], v[0:1], s[4:7], 0 addr64 glc
; GFX6-NEXT: s_waitcnt vmcnt(0)
; GFX6-NEXT: buffer_wbinvl1
-; GFX6-NEXT: v_cmp_eq_u32_e32 vcc, v5, v4
+; GFX6-NEXT: v_cmp_eq_u32_e32 vcc, v3, v5
; GFX6-NEXT: s_or_b64 s[8:9], vcc, s[8:9]
-; GFX6-NEXT: v_mov_b32_e32 v4, v5
; GFX6-NEXT: s_andn2_b64 exec, exec, s[8:9]
; GFX6-NEXT: s_cbranch_execnz .LBB37_1
; GFX6-NEXT: ; %bb.2: ; %atomicrmw.end
@@ -7728,23 +7728,23 @@ define void @global_agent_atomic_fadd_noret_f64__amdgpu_no_fine_grained_memory(p
; GFX7-NEXT: s_mov_b32 s7, 0xf000
; GFX7-NEXT: s_mov_b32 s4, s6
; GFX7-NEXT: s_mov_b32 s5, s6
-; GFX7-NEXT: buffer_load_dwordx2 v[6:7], v[0:1], s[4:7], 0 addr64
+; GFX7-NEXT: buffer_load_dwordx2 v[4:5], v[0:1], s[4:7], 0 addr64
; GFX7-NEXT: s_mov_b64 s[8:9], 0
; GFX7-NEXT: .LBB41_1: ; %atomicrmw.start
; GFX7-NEXT: ; =>This Inner Loop Header: Depth=1
; GFX7-NEXT: s_waitcnt vmcnt(0)
-; GFX7-NEXT: v_add_f64 v[4:5], v[6:7], v[2:3]
-; GFX7-NEXT: v_mov_b32_e32 v11, v7
-; GFX7-NEXT: v_mov_b32_e32 v10, v6
; GFX7-NEXT: v_mov_b32_e32 v9, v5
; GFX7-NEXT: v_mov_b32_e32 v8, v4
-; GFX7-NEXT: buffer_atomic_cmpswap_x2 v[8:11], v[0:1], s[4:7], 0 addr64 glc
+; GFX7-NEXT: v_add_f64 v[6:7], v[8:9], v[2:3]
+; GFX7-NEXT: v_mov_b32_e32 v4, v6
+; GFX7-NEXT: v_mov_b32_e32 v5, v7
+; GFX7-NEXT: v_mov_b32_e32 v6, v8
+; GFX7-NEXT: v_mov_b32_e32 v7, v9
+; GFX7-NEXT: buffer_atomic_cmpswap_x2 v[4:7], v[0:1], s[4:7], 0 addr64 glc
; GFX7-NEXT: s_waitcnt vmcnt(0)
; GFX7-NEXT: buffer_wbinvl1
-; GFX7-NEXT: v_cmp_eq_u64_e32 vcc, v[8:9], v[6:7]
-; GFX7-NEXT: v_mov_b32_e32 v6, v8
+; GFX7-NEXT: v_cmp_eq_u64_e32 vcc, v[4:5], v[8:9]
; GFX7-NEXT: s_or_b64 s[8:9], vcc, s[8:9]
-; GFX7-NEXT: v_mov_b32_e32 v7, v9
; GFX7-NEXT: s_andn2_b64 exec, exec, s[8:9]
; GFX7-NEXT: s_cbranch_execnz .LBB41_1
; GFX7-NEXT: ; %bb.2: ; %atomicrmw.end
@@ -7757,24 +7757,24 @@ define void @global_agent_atomic_fadd_noret_f64__amdgpu_no_fine_grained_memory(p
; GFX6-NEXT: s_mov_b32 s7, 0xf000
; GFX6-NEXT: s_mov_b32 s4, s6
; GFX6-NEXT: s_mov_b32 s5, s6
-; GFX6-NEXT: buffer_load_dwordx2 v[6:7], v[0:1], s[4:7], 0 addr64
+; GFX6-NEXT: buffer_load_dwordx2 v[4:5], v[0:1], s[4:7], 0 addr64
; GFX6-NEXT: s_mov_b64 s[8:9], 0
; GFX6-NEXT: .LBB41_1: ; %atomicrmw.start
; GFX6-NEXT: ; =>This Inner Loop Header: Depth=1
; GFX6-NEXT: s_waitcnt vmcnt(0)
-; GFX6-NEXT: v_add_f64 v[4:5], v[6:7], v[2:3]
-; GFX6-NEXT: s_waitcnt expcnt(0)
-; GFX6-NEXT: v_mov_b32_e32 v11, v7
-; GFX6-NEXT: v_mov_b32_e32 v10, v6
; GFX6-NEXT: v_mov_b32_e32 v9, v5
; GFX6-NEXT: v_mov_b32_e32 v8, v4
-; GFX6-NEXT: buffer_atomic_cmpswap_x2 v[8:11], v[0:1], s[4:7], 0 addr64 glc
+; GFX6-NEXT: s_waitcnt expcnt(0)
+; GFX6-NEXT: v_add_f64 v[6:7], v[8:9], v[2:3]
+; GFX6-NEXT: v_mov_b32_e32 v4, v6
+; GFX6-NEXT: v_mov_b32_e32 v5, v7
+; GFX6-NEXT: v_mov_b32_e32 v6, v8
+; GFX6-NEXT: v_mov_b32_e32 v7, v9
+; GFX6-NEXT: buffer_atomic_cmpswap_x2 v[4:7], v[0:1], s[4:7], 0 addr64 glc
; GFX6-NEXT: s_waitcnt vmcnt(0)
; GFX6-NEXT: buffer_wbinvl1
-; GFX6-NEXT: v_cmp_eq_u64_e32 vcc, v[8:9], v[6:7]
-; GFX6-NEXT: v_mov_b32_e32 v6, v8
+; GFX6-NEXT: v_cmp_eq_u64_e32 vcc, v[4:5], v[8:9]
; GFX6-NEXT: s_or_b64 s[8:9], vcc, s[8:9]
-; GFX6-NEXT: v_mov_b32_e32 v7, v9
; GFX6-NEXT: s_andn2_b64 exec, exec, s[8:9]
; GFX6-NEXT: s_cbranch_execnz .LBB41_1
; GFX6-NEXT: ; %bb.2: ; %atomicrmw.end
@@ -7933,23 +7933,23 @@ define void @global_agent_atomic_fadd_noret_f64__offset12b_pos__amdgpu_no_fine_g
; GFX7-NEXT: s_mov_b32 s7, 0xf000
; GFX7-NEXT: s_mov_b32 s4, s6
; GFX7-NEXT: s_mov_b32 s5, s6
-; GFX7-NEXT: buffer_load_dwordx2 v[6:7], v[0:1], s[4:7], 0 addr64 offset:2040
+; GFX7-NEXT: buffer_load_dwordx2 v[4:5], v[0:1], s[4:7], 0 addr64 offset:2040
; GFX7-NEXT: s_mov_b64 s[8:9], 0
; GFX7-NEXT: .LBB42_1: ; %atomicrmw.start
; GFX7-NEXT: ; =>This Inner Loop Header: Depth=1
; GFX7-NEXT: s_waitcnt vmcnt(0)
-; GFX7-NEXT: v_add_f64 v[4:5], v[6:7], v[2:3]
-; GFX7-NEXT: v_mov_b32_e32 v11, v7
-; GFX7-NEXT: v_mov_b32_e32 v10, v6
; GFX7-NEXT: v_mov_b32_e32 v9, v5
; GFX7-NEXT: v_mov_b32_e32 v8, v4
-; GFX7-NEXT: buffer_atomic_cmpswap_x2 v[8:11], v[0:1], s[4:7], 0 addr64 offset:2040 glc
+; GFX7-NEXT: v_add_f64 v[6:7], v[8:9], v[2:3]
+; GFX7-NEXT: v_mov_b32_e32 v4, v6
+; GFX7-NEXT: v_mov_b32_e32 v5, v7
+; GFX7-NEXT: v_mov_b32_e32 v6, v8
+; GFX7-NEXT: v_mov_b32_e32 v7, v9
+; GFX7-NEXT: buffer_atomic_cmpswap_x2 v[4:7], v[0:1], s[4:7], 0 addr64 offset:2040 glc
; GFX7-NEXT: s_waitcnt vmcnt(0)
; GFX7-NEXT: buffer_wbinvl1
-; GFX7-NEXT: v_cmp_eq_u64_e32 vcc, v[8:9], v[6:7]
-; GFX7-NEXT: v_mov_b32_e32 v6, v8
+; GFX7-NEXT: v_cmp_eq_u64_e32 vcc, v[4:5], v[8:9]
; GFX7-NEXT: s_or_b64 s[8:9], vcc, s[8:9]
-; GFX7-NEXT: v_mov_b32_e32 v7, v9
; GFX7-NEXT: s_andn2_b64 exec, exec, s[8:9]
; GFX7-NEXT: s_cbranch_execnz .LBB42_1
; GFX7-NEXT: ; %bb.2: ; %atomicrmw.end
@@ -7962,24 +7962,24 @@ define void @global_agent_atomic_fadd_noret_f64__offset12b_pos__amdgpu_no_fine_g
; GFX6-NEXT: s_mov_b32 s7, 0xf000
; GFX6-NEXT: s_mov_b32 s4, s6
; GFX6-NEXT: s_mov_b32 s5, s6
-; GFX6-NEXT: buffer_load_dwordx2 v[6:7], v[0:1], s[4:7], 0 addr64 offset:2040
+; GFX6-NEXT: buffer_load_dwordx2 v[4:5], v[0:1], s[4:7], 0 addr64 offset:2040
; GFX6-NEXT: s_mov_b64 s[8:9], 0
; GFX6-NEXT: .LBB42_1: ; %atomicrmw.start
; GFX6-NEXT: ; =>This Inner Loop Header: Depth=1
; GFX6-NEXT: s_waitcnt vmcnt(0)
-; GFX6-NEXT: v_add_f64 v[4:5], v[6:7], v[2:3]
-; GFX6-NEXT: s_waitcnt expcnt(0)
-; GFX6-NEXT: v_mov_b32_e32 v11, v7
-; GFX6-NEXT: v_mov_b32_e32 v10, v6
; GFX6-NEXT: v_mov_b32_e32 v9, v5
; GFX6-NEXT: v_mov_b32_e32 v8, v4
-; GFX6-NEXT: buffer_atomic_cmpswap_x2 v[8:11], v[0:1], s[4:7], 0 addr64 offset:2040 glc
+; GFX6-NEXT: s_waitcnt expcnt(0)
+; GFX6-NEXT: v_add_f64 v[6:7], v[8:9], v[2:3]
+; GFX6-NEXT: v_mov_b32_e32 v4, v6
+; GFX6-NEXT: v_mov_b32_e32 v5, v7
+; GFX6-NEXT: v_mov_b32_e32 v6, v8
+; GFX6-NEXT: v_mov_b32_e32 v7, v9
+; GFX6-NEXT: buffer_atomic_cmpswap_x2 v[4:7], v[0:1], s[4:7], 0 addr64 offset:2040 glc
; GFX6-NEXT: s_waitcnt vmcnt(0)
; GFX6-NEXT: buffer_wbinvl1
-; GFX6-NEXT: v_cmp_eq_u64_e32 vcc, v[8:9], v[6:7]
-; GFX6-NEXT: v_mov_b32_e32 v6, v8
+; GFX6-NEXT: v_cmp_eq_u64_e32 vcc, v[4:5], v[8:9]
; GFX6-NEXT: s_or_b64 s[8:9], vcc, s[8:9]
-; GFX6-NEXT: v_mov_b32_e32 v7, v9
; GFX6-NEXT: s_andn2_b64 exec, exec, s[8:9]
; GFX6-NEXT: s_cbranch_execnz .LBB42_1
; GFX6-NEXT: ; %bb.2: ; %atomicrmw.end
@@ -8139,7 +8139,7 @@ define void @global_agent_atomic_fadd_noret_f64__offset12b_neg__amdgpu_no_fine_g
; GFX7-NEXT: s_mov_b32 s5, -1
; GFX7-NEXT: s_mov_b32 s7, 0xf000
; GFX7-NEXT: s_mov_b32 s6, 0
-; GFX7-NEXT: buffer_load_dwordx2 v[6:7], v[0:1], s[4:7], 0 addr64
+; GFX7-NEXT: buffer_load_dwordx2 v[4:5], v[0:1], s[4:7], 0 addr64
; GFX7-NEXT: v_add_i32_e32 v0, vcc, 0xfffff800, v0
; GFX7-NEXT: v_addc_u32_e32 v1, vcc, -1, v1, vcc
; GFX7-NEXT: s_mov_b64 s[8:9], 0
@@ -8148,18 +8148,18 @@ define void @global_agent_atomic_fadd_noret_f64__offset12b_neg__amdgpu_no_fine_g
; GFX7-NEXT: .LBB43_1: ; %atomicrmw.start
; GFX7-NEXT: ; =>This Inner Loop Header: Depth=1
; GFX7-NEXT: s_waitcnt vmcnt(0)
-; GFX7-NEXT: v_add_f64 v[4:5], v[6:7], v[2:3]
-; GFX7-NEXT: v_mov_b32_e32 v11, v7
-; GFX7-NEXT: v_mov_b32_e32 v10, v6
; GFX7-NEXT: v_mov_b32_e32 v9, v5
; GFX7-NEXT: v_mov_b32_e32 v8, v4
-; GFX7-NEXT: buffer_atomic_cmpswap_x2 v[8:11], v[0:1], s[4:7], 0 addr64 glc
+; GFX7-NEXT: v_add_f64 v[6:7], v[8:9], v[2:3]
+; GFX7-NEXT: v_mov_b32_e32 v4, v6
+; GFX7-NEXT: v_mov_b32_e32 v5, v7
+; GFX7-NEXT: v_mov_b32_e32 v6, v8
+; GFX7-NEXT: v_mov_b32_e32 v7, v9
+; GFX7-NEXT: buffer_atomic_cmpswap_x2 v[4:7], v[0:1], s[4:7], 0 addr64 glc
; GFX7-NEXT: s_waitcnt vmcnt(0)
; GFX7-NEXT: buffer_wbinvl1
-; GFX7-NEXT: v_cmp_eq_u64_e32 vcc, v[8:9], v[6:7]
-; GFX7-NEXT: v_mov_b32_e32 v6, v8
+; GFX7-NEXT: v_cmp_eq_u64_e32 vcc, v[4:5], v[8:9]
; GFX7-NEXT: s_or_b64 s[8:9], vcc, s[8:9]
-; GFX7-NEXT: v_mov_b32_e32 v7, v9
; GFX7-NEXT: s_andn2_b64 exec, exec, s[8:9]
; GFX7-NEXT: s_cbranch_execnz .LBB43_1
; GFX7-NEXT: ; %bb.2: ; %atomicrmw.end
@@ -8172,7 +8172,7 @@ define void @global_agent_atomic_fadd_noret_f64__offset12b_neg__amdgpu_no_fine_g
; GFX6-NEXT: s_mov_b32 s5, -1
; GFX6-NEXT: s_mov_b32 s7, 0xf000
; GFX6-NEXT: s_mov_b32 s6, 0
-; GFX6-NEXT: buffer_load_dwordx2 v[6:7], v[0:1], s[4:7], 0 addr64
+; GFX6-NEXT: buffer_load_dwordx2 v[4:5], v[0:1], s[4:7], 0 addr64
; GFX6-NEXT: v_add_i32_e32 v0, vcc, 0xfffff800, v0
; GFX6-NEXT: v_addc_u32_e32 v1, vcc, -1, v1, vcc
; GFX6-NEXT: s_mov_b64 s[8:9], 0
@@ -8181,19 +8181,19 @@ define void @global_agent_atomic_fadd_noret_f64__offset12b_neg__amdgpu_no_fine_g
; GFX6-NEXT: .LBB43_1: ; %atomicrmw.start
; GFX6-NEXT: ; =>This Inner Loop Header: Depth=1
; GFX6-NEXT: s_waitcnt vmcnt(0)
-; GFX6-NEXT: v_add_f64 v[4:5], v[6:7], v[2:3]
-; GFX6-NEXT: s_waitcnt expcnt(0)
-; GFX6-NEXT: v_mov_b32_e32 v11, v7
-; GFX6-NEXT: v_mov_b32_e32 v10, v6
; GFX6-NEXT: v_mov_b32_e32 v9, v5
; GFX6-NEXT: v_mov_b32_e32 v8, v4
-; GFX6-NEXT: buffer_atomic_cmpswap_x2 v[8:11], v[0:1], s[4:7], 0 addr64 glc
+; GFX6-NEXT: s_waitcnt expcnt(0)
+; GFX6-NEXT: v_add_f64 v[6:7], v[8:9], v[2:3]
+; GFX6-NEXT: v_mov_b32_e32 v4, v6
+; GFX6-NEXT: v_mov_b32_e32 v5, v7
+; GFX6-NEXT: v_mov_b32_e32 v6, v8
+; GFX6-NEXT: v_mov_b32_e32 v7, v9
+; GFX6-NEXT: buffer_atomic_cmpswap_x2 v[4:7], v[0:1], s[4:7], 0 addr64 glc
; GFX6-NEXT: s_waitcnt vmcnt(0)
; GFX6-NEXT: buffer_wbinvl1
-; GFX6-NEXT: v_cmp_eq_u64_e32 vcc, v[8:9], v[6:7]
-; GFX6-NEXT: v_mov_b32_e32 v6, v8
+; GFX6-NEXT: v_cmp_eq_u64_e32 vcc, v[4:5], v[8:9]
; GFX6-NEXT: s_or_b64 s[8:9], vcc, s[8:9]
-; GFX6-NEXT: v_mov_b32_e32 v7, v9
; GFX6-NEXT: s_andn2_b64 exec, exec, s[8:9]
; GFX6-NEXT: s_cbranch_execnz .LBB43_1
; GFX6-NEXT: ; %bb.2: ; %atomicrmw.end
@@ -8553,36 +8553,36 @@ define half @global_agent_atomic_fadd_ret_f16__amdgpu_no_fine_grained_memory(ptr
; GFX7-NEXT: s_mov_b32 s4, s6
; GFX7-NEXT: s_mov_b32 s5, s6
; GFX7-NEXT: buffer_load_dword v4, v[0:1], s[4:7], 0 addr64
-; GFX7-NEXT: v_cvt_f16_f32_e32 v2, v2
-; GFX7-NEXT: v_and_b32_e32 v3, 3, v3
-; GFX7-NEXT: v_lshlrev_b32_e32 v5, 3, v3
+; GFX7-NEXT: v_cvt_f16_f32_e32 v5, v2
+; GFX7-NEXT: v_and_b32_e32 v2, 3, v3
+; GFX7-NEXT: v_lshlrev_b32_e32 v2, 3, v2
; GFX7-NEXT: s_mov_b64 s[8:9], 0
-; GFX7-NEXT: v_cvt_f32_f16_e32 v6, v2
-; GFX7-NEXT: v_lshl_b32_e32 v2, 0xffff, v5
-; GFX7-NEXT: v_not_b32_e32 v7, v2
+; GFX7-NEXT: v_cvt_f32_f16_e32 v3, v5
+; GFX7-NEXT: v_lshl_b32_e32 v5, 0xffff, v2
+; GFX7-NEXT: v_not_b32_e32 v6, v5
; GFX7-NEXT: .LBB44_1: ; %atomicrmw.start
; GFX7-NEXT: ; =>This Inner Loop Header: Depth=1
; GFX7-NEXT: s_waitcnt vmcnt(0)
-; GFX7-NEXT: v_lshrrev_b32_e32 v2, v5, v4
-; GFX7-NEXT: v_cvt_f32_f16_e32 v2, v2
-; GFX7-NEXT: v_and_b32_e32 v3, v4, v7
-; GFX7-NEXT: v_add_f32_e32 v2, v2, v6
-; GFX7-NEXT: v_cvt_f16_f32_e32 v2, v2
-; GFX7-NEXT: v_lshlrev_b32_e32 v2, v5, v2
-; GFX7-NEXT: v_or_b32_e32 v3, v3, v2
-; GFX7-NEXT: v_mov_b32_e32 v2, v3
-; GFX7-NEXT: v_mov_b32_e32 v3, v4
-; GFX7-NEXT: buffer_atomic_cmpswap v[2:3], v[0:1], s[4:7], 0 addr64 glc
+; GFX7-NEXT: v_mov_b32_e32 v8, v4
+; GFX7-NEXT: v_lshrrev_b32_e32 v4, v2, v8
+; GFX7-NEXT: v_cvt_f32_f16_e32 v4, v4
+; GFX7-NEXT: v_and_b32_e32 v5, v8, v6
+; GFX7-NEXT: v_add_f32_e32 v4, v4, v3
+; GFX7-NEXT: v_cvt_f16_f32_e32 v4, v4
+; GFX7-NEXT: v_lshlrev_b32_e32 v4, v2, v4
+; GFX7-NEXT: v_or_b32_e32 v7, v5, v4
+; GFX7-NEXT: v_mov_b32_e32 v4, v7
+; GFX7-NEXT: v_mov_b32_e32 v5, v8
+; GFX7-NEXT: buffer_atomic_cmpswap v[4:5], v[0:1], s[4:7], 0 addr64 glc
; GFX7-NEXT: s_waitcnt vmcnt(0)
; GFX7-NEXT: buffer_wbinvl1
-; GFX7-NEXT: v_cmp_eq_u32_e32 vcc, v2, v4
+; GFX7-NEXT: v_cmp_eq_u32_e32 vcc, v4, v8
; GFX7-NEXT: s_or_b64 s[8:9], vcc, s[8:9]
-; GFX7-NEXT: v_mov_b32_e32 v4, v2
; GFX7-NEXT: s_andn2_b64 exec, exec, s[8:9]
; GFX7-NEXT: s_cbranch_execnz .LBB44_1
; GFX7-NEXT: ; %bb.2: ; %atomicrmw.end
; GFX7-NEXT: s_or_b64 exec, exec, s[8:9]
-; GFX7-NEXT: v_lshrrev_b32_e32 v0, v5, v2
+; GFX7-NEXT: v_lshrrev_b32_e32 v0, v2, v4
; GFX7-NEXT: v_cvt_f32_f16_e32 v0, v0
; GFX7-NEXT: s_setpc_b64 s[30:31]
; GFX6-LABEL: global_agent_atomic_fadd_ret_f16__amdgpu_no_fine_grained_memory:
@@ -8595,36 +8595,37 @@ define half @global_agent_atomic_fadd_ret_f16__amdgpu_no_fine_grained_memory(ptr
; GFX6-NEXT: s_mov_b32 s4, s6
; GFX6-NEXT: s_mov_b32 s5, s6
; GFX6-NEXT: buffer_load_dword v4, v[0:1], s[4:7], 0 addr64
-; GFX6-NEXT: v_cvt_f16_f32_e32 v2, v2
-; GFX6-NEXT: v_and_b32_e32 v3, 3, v3
-; GFX6-NEXT: v_lshlrev_b32_e32 v5, 3, v3
+; GFX6-NEXT: v_cvt_f16_f32_e32 v5, v2
+; GFX6-NEXT: v_and_b32_e32 v2, 3, v3
+; GFX6-NEXT: v_lshlrev_b32_e32 v2, 3, v2
; GFX6-NEXT: s_mov_b64 s[8:9], 0
-; GFX6-NEXT: v_cvt_f32_f16_e32 v6, v2
-; GFX6-NEXT: v_lshl_b32_e32 v2, 0xffff, v5
-; GFX6-NEXT: v_not_b32_e32 v7, v2
+; GFX6-NEXT: v_cvt_f32_f16_e32 v3, v5
+; GFX6-NEXT: v_lshl_b32_e32 v5, 0xffff, v2
+; GFX6-NEXT: v_not_b32_e32 v6, v5
; GFX6-NEXT: .LBB44_1: ; %atomicrmw.start
; GFX6-NEXT: ; =>This Inner Loop Header: Depth=1
-; GFX6-NEXT: s_waitcnt vmcnt(0) expcnt(0)
-; GFX6-NEXT: v_lshrrev_b32_e32 v2, v5, v4
-; GFX6-NEXT: v_cvt_f32_f16_e32 v2, v2
-; GFX6-NEXT: v_and_b32_e32 v3, v4, v7
-; GFX6-NEXT: v_add_f32_e32 v2, v2, v6
-; GFX6-NEXT: v_cvt_f16_f32_e32 v2, v2
-; GFX6-NEXT: v_lshlrev_b32_e32 v2, v5, v2
-; GFX6-NEXT: v_or_b32_e32 v3, v3, v2
-; GFX6-NEXT: v_mov_b32_e32 v2, v3
-; GFX6-NEXT: v_mov_b32_e32 v3, v4
-; GFX6-NEXT: buffer_atomic_cmpswap v[2:3], v[0:1], s[4:7], 0 addr64 glc
+; GFX6-NEXT: s_waitcnt vmcnt(0)
+; GFX6-NEXT: v_mov_b32_e32 v8, v4
+; GFX6-NEXT: s_waitcnt expcnt(0)
+; GFX6-NEXT: v_lshrrev_b32_e32 v4, v2, v8
+; GFX6-NEXT: v_cvt_f32_f16_e32 v4, v4
+; GFX6-NEXT: v_and_b32_e32 v5, v8, v6
+; GFX6-NEXT: v_add_f32_e32 v4, v4, v3
+; GFX6-NEXT: v_cvt_f16_f32_e32 v4, v4
+; GFX6-NEXT: v_lshlrev_b32_e32 v4, v2, v4
+; GFX6-NEXT: v_or_b32_e32 v7, v5, v4
+; GFX6-NEXT: v_mov_b32_e32 v4, v7
+; GFX6-NEXT: v_mov_b32_e32 v5, v8
+; GFX6-NEXT: buffer_atomic_cmpswap v[4:5], v[0:1], s[4:7], 0 addr64 glc
; GFX6-NEXT: s_waitcnt vmcnt(0)
; GFX6-NEXT: buffer_wbinvl1
-; GFX6-NEXT: v_cmp_eq_u32_e32 vcc, v2, v4
+; GFX6-NEXT: v_cmp_eq_u32_e32 vcc, v4, v8
; GFX6-NEXT: s_or_b64 s[8:9], vcc, s[8:9]
-; GFX6-NEXT: v_mov_b32_e32 v4, v2
; GFX6-NEXT: s_andn2_b64 exec, exec, s[8:9]
; GFX6-NEXT: s_cbranch_execnz .LBB44_1
; GFX6-NEXT: ; %bb.2: ; %atomicrmw.end
; GFX6-NEXT: s_or_b64 exec, exec, s[8:9]
-; GFX6-NEXT: v_lshrrev_b32_e32 v0, v5, v2
+; GFX6-NEXT: v_lshrrev_b32_e32 v0, v2, v4
; GFX6-NEXT: v_cvt_f32_f16_e32 v0, v0
; GFX6-NEXT: s_waitcnt expcnt(0)
; GFX6-NEXT: s_setpc_b64 s[30:31]
@@ -8987,36 +8988,36 @@ define half @global_agent_atomic_fadd_ret_f16__offset12b_pos__amdgpu_no_fine_gra
; GFX7-NEXT: s_mov_b32 s4, s6
; GFX7-NEXT: s_mov_b32 s5, s6
; GFX7-NEXT: buffer_load_dword v3, v[0:1], s[4:7], 0 addr64
-; GFX7-NEXT: v_cvt_f16_f32_e32 v2, v2
-; GFX7-NEXT: v_and_b32_e32 v4, 3, v4
-; GFX7-NEXT: v_lshlrev_b32_e32 v6, 3, v4
+; GFX7-NEXT: v_cvt_f16_f32_e32 v5, v2
+; GFX7-NEXT: v_and_b32_e32 v2, 3, v4
+; GFX7-NEXT: v_lshlrev_b32_e32 v2, 3, v2
+; GFX7-NEXT: v_lshl_b32_e32 v4, 0xffff, v2
+; GFX7-NEXT: v_cvt_f32_f16_e32 v5, v5
+; GFX7-NEXT: v_not_b32_e32 v6, v4
; GFX7-NEXT: s_mov_b64 s[8:9], 0
-; GFX7-NEXT: v_cvt_f32_f16_e32 v7, v2
-; GFX7-NEXT: v_lshl_b32_e32 v2, 0xffff, v6
-; GFX7-NEXT: v_not_b32_e32 v8, v2
; GFX7-NEXT: .LBB45_1: ; %atomicrmw.start
; GFX7-NEXT: ; =>This Inner Loop Header: Depth=1
; GFX7-NEXT: s_waitcnt vmcnt(0)
-; GFX7-NEXT: v_lshrrev_b32_e32 v2, v6, v3
-; GFX7-NEXT: v_cvt_f32_f16_e32 v2, v2
-; GFX7-NEXT: v_and_b32_e32 v4, v3, v8
-; GFX7-NEXT: v_add_f32_e32 v2, v2, v7
-; GFX7-NEXT: v_cvt_f16_f32_e32 v2, v2
-; GFX7-NEXT: v_lshlrev_b32_e32 v2, v6, v2
-; GFX7-NEXT: v_or_b32_e32 v2, v4, v2
-; GFX7-NEXT: v_mov_b32_e32 v5, v3
-; GFX7-NEXT: v_mov_b32_e32 v4, v2
-; GFX7-NEXT: buffer_atomic_cmpswap v[4:5], v[0:1], s[4:7], 0 addr64 glc
+; GFX7-NEXT: v_mov_b32_e32 v8, v3
+; GFX7-NEXT: v_lshrrev_b32_e32 v3, v2, v8
+; GFX7-NEXT: v_cvt_f32_f16_e32 v3, v3
+; GFX7-NEXT: v_and_b32_e32 v4, v8, v6
+; GFX7-NEXT: v_add_f32_e32 v3, v3, v5
+; GFX7-NEXT: v_cvt_f16_f32_e32 v3, v3
+; GFX7-NEXT: v_lshlrev_b32_e32 v3, v2, v3
+; GFX7-NEXT: v_or_b32_e32 v7, v4, v3
+; GFX7-NEXT: v_mov_b32_e32 v3, v7
+; GFX7-NEXT: v_mov_b32_e32 v4, v8
+; GFX7-NEXT: buffer_atomic_cmpswap v[3:4], v[0:1], s[4:7], 0 addr64 glc
; GFX7-NEXT: s_waitcnt vmcnt(0)
; GFX7-NEXT: buffer_wbinvl1
-; GFX7-NEXT: v_cmp_eq_u32_e32 vcc, v4, v3
+; GFX7-NEXT: v_cmp_eq_u32_e32 vcc, v3, v8
; GFX7-NEXT: s_or_b64 s[8:9], vcc, s[8:9]
-; GFX7-NEXT: v_mov_b32_e32 v3, v4
; GFX7-NEXT: s_andn2_b64 exec, exec, s[8:9]
; GFX7-NEXT: s_cbranch_execnz .LBB45_1
; GFX7-NEXT: ; %bb.2: ; %atomicrmw.end
; GFX7-NEXT: s_or_b64 exec, exec, s[8:9]
-; GFX7-NEXT: v_lshrrev_b32_e32 v0, v6, v4
+; GFX7-NEXT: v_lshrrev_b32_e32 v0, v2, v3
; GFX7-NEXT: v_cvt_f32_f16_e32 v0, v0
; GFX7-NEXT: s_setpc_b64 s[30:31]
; GFX6-LABEL: global_agent_atomic_fadd_ret_f16__offset12b_pos__amdgpu_no_fine_grained_memory:
@@ -9030,37 +9031,37 @@ define half @global_agent_atomic_fadd_ret_f16__offset12b_pos__amdgpu_no_fine_gra
; GFX6-NEXT: s_mov_b32 s4, s6
; GFX6-NEXT: s_mov_b32 s5, s6
; GFX6-NEXT: buffer_load_dword v3, v[0:1], s[4:7], 0 addr64
-; GFX6-NEXT: v_cvt_f16_f32_e32 v2, v2
-; GFX6-NEXT: v_and_b32_e32 v4, 3, v4
-; GFX6-NEXT: v_lshlrev_b32_e32 v6, 3, v4
+; GFX6-NEXT: v_cvt_f16_f32_e32 v5, v2
+; GFX6-NEXT: v_and_b32_e32 v2, 3, v4
+; GFX6-NEXT: v_lshlrev_b32_e32 v2, 3, v2
+; GFX6-NEXT: v_lshl_b32_e32 v4, 0xffff, v2
+; GFX6-NEXT: v_cvt_f32_f16_e32 v5, v5
+; GFX6-NEXT: v_not_b32_e32 v6, v4
; GFX6-NEXT: s_mov_b64 s[8:9], 0
-; GFX6-NEXT: v_cvt_f32_f16_e32 v7, v2
-; GFX6-NEXT: v_lshl_b32_e32 v2, 0xffff, v6
-; GFX6-NEXT: v_not_b32_e32 v8, v2
; GFX6-NEXT: .LBB45_1: ; %atomicrmw.start
; GFX6-NEXT: ; =>This Inner Loop Header: Depth=1
; GFX6-NEXT: s_waitcnt vmcnt(0)
-; GFX6-NEXT: v_lshrrev_b32_e32 v2, v6, v3
-; GFX6-NEXT: v_cvt_f32_f16_e32 v2, v2
+; GFX6-NEXT: v_mov_b32_e32 v8, v3
; GFX6-NEXT: s_waitcnt expcnt(0)
-; GFX6-NEXT: v_and_b32_e32 v4, v3, v8
-; GFX6-NEXT: v_add_f32_e32 v2, v2, v7
-; GFX6-NEXT: v_cvt_f16_f32_e32 v2, v2
-; GFX6-NEXT: v_lshlrev_b32_e32 v2, v6, v2
-; GFX6-NEXT: v_or_b32_e32 v2, v4, v2
-; GFX6-NEXT: v_mov_b32_e32 v5, v3
-; GFX6-NEXT: v_mov_b32_e32 v4, v2
-; GFX6-NEXT: buffer_atomic_cmpswap v[4:5], v[0:1], s[4:7], 0 addr64 glc
+; GFX6-NEXT: v_lshrrev_b32_e32 v3, v2, v8
+; GFX6-NEXT: v_cvt_f32_f16_e32 v3, v3
+; GFX6-NEXT: v_and_b32_e32 v4, v8, v6
+; GFX6-NEXT: v_add_f32_e32 v3, v3, v5
+; GFX6-NEXT: v_cvt_f16_f32_e32 v3, v3
+; GFX6-NEXT: v_lshlrev_b32_e32 v3, v2, v3
+; GFX6-NEXT: v_or_b32_e32 v7, v4, v3
+; GFX6-NEXT: v_mov_b32_e32 v3, v7
+; GFX6-NEXT: v_mov_b32_e32 v4, v8
+; GFX6-NEXT: buffer_atomic_cmpswap v[3:4], v[0:1], s[4:7], 0 addr64 glc
; GFX6-NEXT: s_waitcnt vmcnt(0)
; GFX6-NEXT: buffer_wbinvl1
-; GFX6-NEXT: v_cmp_eq_u32_e32 vcc, v4, v3
+; GFX6-NEXT: v_cmp_eq_u32_e32 vcc, v3, v8
; GFX6-NEXT: s_or_b64 s[8:9], vcc, s[8:9]
-; GFX6-NEXT: v_mov_b32_e32 v3, v4
; GFX6-NEXT: s_andn2_b64 exec, exec, s[8:9]
; GFX6-NEXT: s_cbranch_execnz .LBB45_1
; GFX6-NEXT: ; %bb.2: ; %atomicrmw.end
; GFX6-NEXT: s_or_b64 exec, exec, s[8:9]
-; GFX6-NEXT: v_lshrrev_b32_e32 v0, v6, v4
+; GFX6-NEXT: v_lshrrev_b32_e32 v0, v2, v3
; GFX6-NEXT: v_cvt_f32_f16_e32 v0, v0
; GFX6-NEXT: s_waitcnt expcnt(0)
; GFX6-NEXT: s_setpc_b64 s[30:31]
@@ -9425,36 +9426,36 @@ define half @global_agent_atomic_fadd_ret_f16__offset12b_neg__amdgpu_no_fine_gra
; GFX7-NEXT: s_mov_b32 s4, s6
; GFX7-NEXT: s_mov_b32 s5, s6
; GFX7-NEXT: buffer_load_dword v3, v[0:1], s[4:7], 0 addr64
-; GFX7-NEXT: v_cvt_f16_f32_e32 v2, v2
-; GFX7-NEXT: v_and_b32_e32 v4, 3, v4
-; GFX7-NEXT: v_lshlrev_b32_e32 v6, 3, v4
+; GFX7-NEXT: v_cvt_f16_f32_e32 v5, v2
+; GFX7-NEXT: v_and_b32_e32 v2, 3, v4
+; GFX7-NEXT: v_lshlrev_b32_e32 v2, 3, v2
+; GFX7-NEXT: v_lshl_b32_e32 v4, 0xffff, v2
+; GFX7-NEXT: v_cvt_f32_f16_e32 v5, v5
+; GFX7-NEXT: v_not_b32_e32 v6, v4
; GFX7-NEXT: s_mov_b64 s[8:9], 0
-; GFX7-NEXT: v_cvt_f32_f16_e32 v7, v2
-; GFX7-NEXT: v_lshl_b32_e32 v2, 0xffff, v6
-; GFX7-NEXT: v_not_b32_e32 v8, v2
; GFX7-NEXT: .LBB46_1: ; %atomicrmw.start
; GFX7-NEXT: ; =>This Inner Loop Header: Depth=1
; GFX7-NEXT: s_waitcnt vmcnt(0)
-; GFX7-NEXT: v_lshrrev_b32_e32 v2, v6, v3
-; GFX7-NEXT: v_cvt_f32_f16_e32 v2, v2
-; GFX7-NEXT: v_and_b32_e32 v4, v3, v8
-; GFX7-NEXT: v_add_f32_e32 v2, v2, v7
-; GFX7-NEXT: v_cvt_f16_f32_e32 v2, v2
-; GFX7-NEXT: v_lshlrev_b32_e32 v2, v6, v2
-; GFX7-NEXT: v_or_b32_e32 v2, v4, v2
-; GFX7-NEXT: v_mov_b32_e32 v5, v3
-; GFX7-NEXT: v_mov_b32_e32 v4, v2
-; GFX7-NEXT: buffer_atomic_cmpswap v[4:5], v[0:1], s[4:7], 0 addr64 glc
+; GFX7-NEXT: v_mov_b32_e32 v8, v3
+; GFX7-NEXT: v_lshrrev_b32_e32 v3, v2, v8
+; GFX7-NEXT: v_cvt_f32_f16_e32 v3, v3
+; GFX7-NEXT: v_and_b32_e32 v4, v8, v6
+; GFX7-NEXT: v_add_f32_e32 v3, v3, v5
+; GFX7-NEXT: v_cvt_f16_f32_e32 v3, v3
+; GFX7-NEXT: v_lshlrev_b32_e32 v3, v2, v3
+; GFX7-NEXT: v_or_b32_e32 v7, v4, v3
+; GFX7-NEXT: v_mov_b32_e32 v3, v7
+; GFX7-NEXT: v_mov_b32_e32 v4, v8
+; GFX7-NEXT: buffer_atomic_cmpswap v[3:4], v[0:1], s[4:7], 0 addr64 glc
; GFX7-NEXT: s_waitcnt vmcnt(0)
; GFX7-NEXT: buffer_wbinvl1
-; GFX7-NEXT: v_cmp_eq_u32_e32 vcc, v4, v3
+; GFX7-NEXT: v_cmp_eq_u32_e32 vcc, v3, v8
; GFX7-NEXT: s_or_b64 s[8:9], vcc, s[8:9]
-; GFX7-NEXT: v_mov_b32_e32 v3, v4
; GFX7-NEXT: s_andn2_b64 exec, exec, s[8:9]
; GFX7-NEXT: s_cbranch_execnz .LBB46_1
; GFX7-NEXT: ; %bb.2: ; %atomicrmw.end
; GFX7-NEXT: s_or_b64 exec, exec, s[8:9]
-; GFX7-NEXT: v_lshrrev_b32_e32 v0, v6, v4
+; GFX7-NEXT: v_lshrrev_b32_e32 v0, v2, v3
; GFX7-NEXT: v_cvt_f32_f16_e32 v0, v0
; GFX7-NEXT: s_setpc_b64 s[30:31]
; GFX6-LABEL: global_agent_atomic_fadd_ret_f16__offset12b_neg__amdgpu_no_fine_grained_memory:
@@ -9468,37 +9469,37 @@ define half @global_agent_atomic_fadd_ret_f16__offset12b_neg__amdgpu_no_fine_gra
; GFX6-NEXT: s_mov_b32 s4, s6
; GFX6-NEXT: s_mov_b32 s5, s6
; GFX6-NEXT: buffer_load_dword v3, v[0:1], s[4:7], 0 addr64
-; GFX6-NEXT: v_cvt_f16_f32_e32 v2, v2
-; GFX6-NEXT: v_and_b32_e32 v4, 3, v4
-; GFX6-NEXT: v_lshlrev_b32_e32 v6, 3, v4
+; GFX6-NEXT: v_cvt_f16_f32_e32 v5, v2
+; GFX6-NEXT: v_and_b32_e32 v2, 3, v4
+; GFX6-NEXT: v_lshlrev_b32_e32 v2, 3, v2
+; GFX6-NEXT: v_lshl_b32_e32 v4, 0xffff, v2
+; GFX6-NEXT: v_cvt_f32_f16_e32 v5, v5
+; GFX6-NEXT: v_not_b32_e32 v6, v4
; GFX6-NEXT: s_mov_b64 s[8:9], 0
-; GFX6-NEXT: v_cvt_f32_f16_e32 v7, v2
-; GFX6-NEXT: v_lshl_b32_e32 v2, 0xffff, v6
-; GFX6-NEXT: v_not_b32_e32 v8, v2
; GFX6-NEXT: .LBB46_1: ; %atomicrmw.start
; GFX6-NEXT: ; =>This Inner Loop Header: Depth=1
; GFX6-NEXT: s_waitcnt vmcnt(0)
-; GFX6-NEXT: v_lshrrev_b32_e32 v2, v6, v3
-; GFX6-NEXT: v_cvt_f32_f16_e32 v2, v2
+; GFX6-NEXT: v_mov_b32_e32 v8, v3
; GFX6-NEXT: s_waitcnt expcnt(0)
-; GFX6-NEXT: v_and_b32_e32 v4, v3, v8
-; GFX6-NEXT: v_add_f32_e32 v2, v2, v7
-; GFX6-NEXT: v_cvt_f16_f32_e32 v2, v2
-; GFX6-NEXT: v_lshlrev_b32_e32 v2, v6, v2
-; GFX6-NEXT: v_or_b32_e32 v2, v4, v2
-; GFX6-NEXT: v_mov_b32_e32 v5, v3
-; GFX6-NEXT: v_mov_b32_e32 v4, v2
-; GFX6-NEXT: buffer_atomic_cmpswap v[4:5], v[0:1], s[4:7], 0 addr64 glc
+; GFX6-NEXT: v_lshrrev_b32_e32 v3, v2, v8
+; GFX6-NEXT: v_cvt_f32_f16_e32 v3, v3
+; GFX6-NEXT: v_and_b32_e32 v4, v8, v6
+; GFX6-NEXT: v_add_f32_e32 v3, v3, v5
+; GFX6-NEXT: v_cvt_f16_f32_e32 v3, v3
+; GFX6-NEXT: v_lshlrev_b32_e32 v3, v2, v3
+; GFX6-NEXT: v_or_b32_e32 v7, v4, v3
+; GFX6-NEXT: v_mov_b32_e32 v3, v7
+; GFX6-NEXT: v_mov_b32_e32 v4, v8
+; GFX6-NEXT: buffer_atomic_cmpswap v[3:4], v[0:1], s[4:7], 0 addr64 glc
; GFX6-NEXT: s_waitcnt vmcnt(0)
; GFX6-NEXT: buffer_wbinvl1
-; GFX6-NEXT: v_cmp_eq_u32_e32 vcc, v4, v3
+; GFX6-NEXT: v_cmp_eq_u32_e32 vcc, v3, v8
; GFX6-NEXT: s_or_b64 s[8:9], vcc, s[8:9]
-; GFX6-NEXT: v_mov_b32_e32 v3, v4
; GFX6-NEXT: s_andn2_b64 exec, exec, s[8:9]
; GFX6-NEXT: s_cbranch_execnz .LBB46_1
; GFX6-NEXT: ; %bb.2: ; %atomicrmw.end
; GFX6-NEXT: s_or_b64 exec, exec, s[8:9]
-; GFX6-NEXT: v_lshrrev_b32_e32 v0, v6, v4
+; GFX6-NEXT: v_lshrrev_b32_e32 v0, v2, v3
; GFX6-NEXT: v_cvt_f32_f16_e32 v0, v0
; GFX6-NEXT: s_waitcnt expcnt(0)
; GFX6-NEXT: s_setpc_b64 s[30:31]
@@ -9841,28 +9842,28 @@ define void @global_agent_atomic_fadd_noret_f16__amdgpu_no_fine_grained_memory(p
; GFX7-NEXT: v_cvt_f16_f32_e32 v5, v2
; GFX7-NEXT: v_and_b32_e32 v2, 3, v3
; GFX7-NEXT: v_lshlrev_b32_e32 v2, 3, v2
-; GFX7-NEXT: v_lshl_b32_e32 v3, 0xffff, v2
-; GFX7-NEXT: v_cvt_f32_f16_e32 v5, v5
-; GFX7-NEXT: v_not_b32_e32 v6, v3
; GFX7-NEXT: s_mov_b64 s[8:9], 0
+; GFX7-NEXT: v_cvt_f32_f16_e32 v3, v5
+; GFX7-NEXT: v_lshl_b32_e32 v5, 0xffff, v2
+; GFX7-NEXT: v_not_b32_e32 v6, v5
; GFX7-NEXT: .LBB47_1: ; %atomicrmw.start
; GFX7-NEXT: ; =>This Inner Loop Header: Depth=1
; GFX7-NEXT: s_waitcnt vmcnt(0)
-; GFX7-NEXT: v_lshrrev_b32_e32 v3, v2, v4
-; GFX7-NEXT: v_cvt_f32_f16_e32 v3, v3
-; GFX7-NEXT: v_and_b32_e32 v7, v4, v6
-; GFX7-NEXT: v_add_f32_e32 v3, v3, v5
-; GFX7-NEXT: v_cvt_f16_f32_e32 v3, v3
-; GFX7-NEXT: v_lshlrev_b32_e32 v3, v2, v3
-; GFX7-NEXT: v_or_b32_e32 v3, v7, v3
; GFX7-NEXT: v_mov_b32_e32 v8, v4
-; GFX7-NEXT: v_mov_b32_e32 v7, v3
-; GFX7-NEXT: buffer_atomic_cmpswap v[7:8], v[0:1], s[4:7], 0 addr64 glc
+; GFX7-NEXT: v_lshrrev_b32_e32 v4, v2, v8
+; GFX7-NEXT: v_cvt_f32_f16_e32 v4, v4
+; GFX7-NEXT: v_and_b32_e32 v5, v8, v6
+; GFX7-NEXT: v_add_f32_e32 v4, v4, v3
+; GFX7-NEXT: v_cvt_f16_f32_e32 v4, v4
+; GFX7-NEXT: v_lshlrev_b32_e32 v4, v2, v4
+; GFX7-NEXT: v_or_b32_e32 v7, v5, v4
+; GFX7-NEXT: v_mov_b32_e32 v4, v7
+; GFX7-NEXT: v_mov_b32_e32 v5, v8
+; GFX7-NEXT: buffer_atomic_cmpswap v[4:5], v[0:1], s[4:7], 0 addr64 glc
; GFX7-NEXT: s_waitcnt vmcnt(0)
; GFX7-NEXT: buffer_wbinvl1
-; GFX7-NEXT: v_cmp_eq_u32_e32 vcc, v7, v4
+; GFX7-NEXT: v_cmp_eq_u32_e32 vcc, v4, v8
; GFX7-NEXT: s_or_b64 s[8:9], vcc, s[8:9]
-; GFX7-NEXT: v_mov_b32_e32 v4, v7
; GFX7-NEXT: s_andn2_b64 exec, exec, s[8:9]
; GFX7-NEXT: s_cbranch_execnz .LBB47_1
; GFX7-NEXT: ; %bb.2: ; %atomicrmw.end
@@ -9881,29 +9882,29 @@ define void @global_agent_atomic_fadd_noret_f16__amdgpu_no_fine_grained_memory(p
; GFX6-NEXT: v_cvt_f16_f32_e32 v5, v2
; GFX6-NEXT: v_and_b32_e32 v2, 3, v3
; GFX6-NEXT: v_lshlrev_b32_e32 v2, 3, v2
-; GFX6-NEXT: v_lshl_b32_e32 v3, 0xffff, v2
-; GFX6-NEXT: v_cvt_f32_f16_e32 v5, v5
-; GFX6-NEXT: v_not_b32_e32 v6, v3
; GFX6-NEXT: s_mov_b64 s[8:9], 0
+; GFX6-NEXT: v_cvt_f32_f16_e32 v3, v5
+; GFX6-NEXT: v_lshl_b32_e32 v5, 0xffff, v2
+; GFX6-NEXT: v_not_b32_e32 v6, v5
; GFX6-NEXT: .LBB47_1: ; %atomicrmw.start
; GFX6-NEXT: ; =>This Inner Loop Header: Depth=1
; GFX6-NEXT: s_waitcnt vmcnt(0)
-; GFX6-NEXT: v_lshrrev_b32_e32 v3, v2, v4
-; GFX6-NEXT: v_cvt_f32_f16_e32 v3, v3
-; GFX6-NEXT: s_waitcnt expcnt(0)
-; GFX6-NEXT: v_and_b32_e32 v7, v4, v6
-; GFX6-NEXT: v_add_f32_e32 v3, v3, v5
-; GFX6-NEXT: v_cvt_f16_f32_e32 v3, v3
-; GFX6-NEXT: v_lshlrev_b32_e32 v3, v2, v3
-; GFX6-NEXT: v_or_b32_e32 v3, v7, v3
; GFX6-NEXT: v_mov_b32_e32 v8, v4
-; GFX6-NEXT: v_mov_b32_e32 v7, v3
-; GFX6-NEXT: buffer_atomic_cmpswap v[7:8], v[0:1], s[4:7], 0 addr64 glc
+; GFX6-NEXT: s_waitcnt expcnt(0)
+; GFX6-NEXT: v_lshrrev_b32_e32 v4, v2, v8
+; GFX6-NEXT: v_cvt_f32_f16_e32 v4, v4
+; GFX6-NEXT: v_and_b32_e32 v5, v8, v6
+; GFX6-NEXT: v_add_f32_e32 v4, v4, v3
+; GFX6-NEXT: v_cvt_f16_f32_e32 v4, v4
+; GFX6-NEXT: v_lshlrev_b32_e32 v4, v2, v4
+; GFX6-NEXT: v_or_b32_e32 v7, v5, v4
+; GFX6-NEXT: v_mov_b32_e32 v4, v7
+; GFX6-NEXT: v_mov_b32_e32 v5, v8
+; GFX6-NEXT: buffer_atomic_cmpswap v[4:5], v[0:1], s[4:7], 0 addr64 glc
; GFX6-NEXT: s_waitcnt vmcnt(0)
; GFX6-NEXT: buffer_wbinvl1
-; GFX6-NEXT: v_cmp_eq_u32_e32 vcc, v7, v4
+; GFX6-NEXT: v_cmp_eq_u32_e32 vcc, v4, v8
; GFX6-NEXT: s_or_b64 s[8:9], vcc, s[8:9]
-; GFX6-NEXT: v_mov_b32_e32 v4, v7
; GFX6-NEXT: s_andn2_b64 exec, exec, s[8:9]
; GFX6-NEXT: s_cbranch_execnz .LBB47_1
; GFX6-NEXT: ; %bb.2: ; %atomicrmw.end
@@ -10256,31 +10257,31 @@ define void @global_agent_atomic_fadd_noret_f16__offset12b_pos__amdgpu_no_fine_g
; GFX7-NEXT: s_mov_b32 s4, s6
; GFX7-NEXT: s_mov_b32 s5, s6
; GFX7-NEXT: buffer_load_dword v3, v[0:1], s[4:7], 0 addr64
-; GFX7-NEXT: v_cvt_f16_f32_e32 v2, v2
-; GFX7-NEXT: v_and_b32_e32 v4, 3, v4
-; GFX7-NEXT: v_lshlrev_b32_e32 v4, 3, v4
+; GFX7-NEXT: v_cvt_f16_f32_e32 v5, v2
+; GFX7-NEXT: v_and_b32_e32 v2, 3, v4
+; GFX7-NEXT: v_lshlrev_b32_e32 v2, 3, v2
+; GFX7-NEXT: v_lshl_b32_e32 v4, 0xffff, v2
+; GFX7-NEXT: v_cvt_f32_f16_e32 v5, v5
+; GFX7-NEXT: v_not_b32_e32 v6, v4
; GFX7-NEXT: s_mov_b64 s[8:9], 0
-; GFX7-NEXT: v_cvt_f32_f16_e32 v5, v2
-; GFX7-NEXT: v_lshl_b32_e32 v2, 0xffff, v4
-; GFX7-NEXT: v_not_b32_e32 v6, v2
; GFX7-NEXT: .LBB48_1: ; %atomicrmw.start
; GFX7-NEXT: ; =>This Inner Loop Header: Depth=1
; GFX7-NEXT: s_waitcnt vmcnt(0)
-; GFX7-NEXT: v_lshrrev_b32_e32 v2, v4, v3
-; GFX7-NEXT: v_cvt_f32_f16_e32 v2, v2
-; GFX7-NEXT: v_and_b32_e32 v7, v3, v6
-; GFX7-NEXT: v_add_f32_e32 v2, v2, v5
-; GFX7-NEXT: v_cvt_f16_f32_e32 v2, v2
-; GFX7-NEXT: v_lshlrev_b32_e32 v2, v4, v2
-; GFX7-NEXT: v_or_b32_e32 v2, v7, v2
; GFX7-NEXT: v_mov_b32_e32 v8, v3
-; GFX7-NEXT: v_mov_b32_e32 v7, v2
-; GFX7-NEXT: buffer_atomic_cmpswap v[7:8], v[0:1], s[4:7], 0 addr64 glc
+; GFX7-NEXT: v_lshrrev_b32_e32 v3, v2, v8
+; GFX7-NEXT: v_cvt_f32_f16_e32 v3, v3
+; GFX7-NEXT: v_and_b32_e32 v4, v8, v6
+; GFX7-NEXT: v_add_f32_e32 v3, v3, v5
+; GFX7-NEXT: v_cvt_f16_f32_e32 v3, v3
+; GFX7-NEXT: v_lshlrev_b32_e32 v3, v2, v3
+; GFX7-NEXT: v_or_b32_e32 v7, v4, v3
+; GFX7-NEXT: v_mov_b32_e32 v3, v7
+; GFX7-NEXT: v_mov_b32_e32 v4, v8
+; GFX7-NEXT: buffer_atomic_cmpswap v[3:4], v[0:1], s[4:7], 0 addr64 glc
; GFX7-NEXT: s_waitcnt vmcnt(0)
; GFX7-NEXT: buffer_wbinvl1
-; GFX7-NEXT: v_cmp_eq_u32_e32 vcc, v7, v3
+; GFX7-NEXT: v_cmp_eq_u32_e32 vcc, v3, v8
; GFX7-NEXT: s_or_b64 s[8:9], vcc, s[8:9]
-; GFX7-NEXT: v_mov_b32_e32 v3, v7
; GFX7-NEXT: s_andn2_b64 exec, exec, s[8:9]
; GFX7-NEXT: s_cbranch_execnz .LBB48_1
; GFX7-NEXT: ; %bb.2: ; %atomicrmw.end
@@ -10297,32 +10298,32 @@ define void @global_agent_atomic_fadd_noret_f16__offset12b_pos__amdgpu_no_fine_g
; GFX6-NEXT: s_mov_b32 s4, s6
; GFX6-NEXT: s_mov_b32 s5, s6
; GFX6-NEXT: buffer_load_dword v3, v[0:1], s[4:7], 0 addr64
-; GFX6-NEXT: v_cvt_f16_f32_e32 v2, v2
-; GFX6-NEXT: v_and_b32_e32 v4, 3, v4
-; GFX6-NEXT: v_lshlrev_b32_e32 v4, 3, v4
+; GFX6-NEXT: v_cvt_f16_f32_e32 v5, v2
+; GFX6-NEXT: v_and_b32_e32 v2, 3, v4
+; GFX6-NEXT: v_lshlrev_b32_e32 v2, 3, v2
+; GFX6-NEXT: v_lshl_b32_e32 v4, 0xffff, v2
+; GFX6-NEXT: v_cvt_f32_f16_e32 v5, v5
+; GFX6-NEXT: v_not_b32_e32 v6, v4
; GFX6-NEXT: s_mov_b64 s[8:9], 0
-; GFX6-NEXT: v_cvt_f32_f16_e32 v5, v2
-; GFX6-NEXT: v_lshl_b32_e32 v2, 0xffff, v4
-; GFX6-NEXT: v_not_b32_e32 v6, v2
; GFX6-NEXT: .LBB48_1: ; %atomicrmw.start
; GFX6-NEXT: ; =>This Inner Loop Header: Depth=1
; GFX6-NEXT: s_waitcnt vmcnt(0)
-; GFX6-NEXT: v_lshrrev_b32_e32 v2, v4, v3
-; GFX6-NEXT: v_cvt_f32_f16_e32 v2, v2
-; GFX6-NEXT: s_waitcnt expcnt(0)
-; GFX6-NEXT: v_and_b32_e32 v7, v3, v6
-; GFX6-NEXT: v_add_f32_e32 v2, v2, v5
-; GFX6-NEXT: v_cvt_f16_f32_e32 v2, v2
-; GFX6-NEXT: v_lshlrev_b32_e32 v2, v4, v2
-; GFX6-NEXT: v_or_b32_e32 v2, v7, v2
; GFX6-NEXT: v_mov_b32_e32 v8, v3
-; GFX6-NEXT: v_mov_b32_e32 v7, v2
-; GFX6-NEXT: buffer_atomic_cmpswap v[7:8], v[0:1], s[4:7], 0 addr64 glc
+; GFX6-NEXT: s_waitcnt expcnt(0)
+; GFX6-NEXT: v_lshrrev_b32_e32 v3, v2, v8
+; GFX6-NEXT: v_cvt_f32_f16_e32 v3, v3
+; GFX6-NEXT: v_and_b32_e32 v4, v8, v6
+; GFX6-NEXT: v_add_f32_e32 v3, v3, v5
+; GFX6-NEXT: v_cvt_f16_f32_e32 v3, v3
+; GFX6-NEXT: v_lshlrev_b32_e32 v3, v2, v3
+; GFX6-NEXT: v_or_b32_e32 v7, v4, v3
+; GFX6-NEXT: v_mov_b32_e32 v3, v7
+; GFX6-NEXT: v_mov_b32_e32 v4, v8
+; GFX6-NEXT: buffer_atomic_cmpswap v[3:4], v[0:1], s[4:7], 0 addr64 glc
; GFX6-NEXT: s_waitcnt vmcnt(0)
; GFX6-NEXT: buffer_wbinvl1
-; GFX6-NEXT: v_cmp_eq_u32_e32 vcc, v7, v3
+; GFX6-NEXT: v_cmp_eq_u32_e32 vcc, v3, v8
; GFX6-NEXT: s_or_b64 s[8:9], vcc, s[8:9]
-; GFX6-NEXT: v_mov_b32_e32 v3, v7
; GFX6-NEXT: s_andn2_b64 exec, exec, s[8:9]
; GFX6-NEXT: s_cbranch_execnz .LBB48_1
; GFX6-NEXT: ; %bb.2: ; %atomicrmw.end
@@ -10677,31 +10678,31 @@ define void @global_agent_atomic_fadd_noret_f16__offset12b_neg__amdgpu_no_fine_g
; GFX7-NEXT: s_mov_b32 s4, s6
; GFX7-NEXT: s_mov_b32 s5, s6
; GFX7-NEXT: buffer_load_dword v3, v[0:1], s[4:7], 0 addr64
-; GFX7-NEXT: v_cvt_f16_f32_e32 v2, v2
-; GFX7-NEXT: v_and_b32_e32 v4, 3, v4
-; GFX7-NEXT: v_lshlrev_b32_e32 v4, 3, v4
+; GFX7-NEXT: v_cvt_f16_f32_e32 v5, v2
+; GFX7-NEXT: v_and_b32_e32 v2, 3, v4
+; GFX7-NEXT: v_lshlrev_b32_e32 v2, 3, v2
+; GFX7-NEXT: v_lshl_b32_e32 v4, 0xffff, v2
+; GFX7-NEXT: v_cvt_f32_f16_e32 v5, v5
+; GFX7-NEXT: v_not_b32_e32 v6, v4
; GFX7-NEXT: s_mov_b64 s[8:9], 0
-; GFX7-NEXT: v_cvt_f32_f16_e32 v5, v2
-; GFX7-NEXT: v_lshl_b32_e32 v2, 0xffff, v4
-; GFX7-NEXT: v_not_b32_e32 v6, v2
; GFX7-NEXT: .LBB49_1: ; %atomicrmw.start
; GFX7-NEXT: ; =>This Inner Loop Header: Depth=1
; GFX7-NEXT: s_waitcnt vmcnt(0)
-; GFX7-NEXT: v_lshrrev_b32_e32 v2, v4, v3
-; GFX7-NEXT: v_cvt_f32_f16_e32 v2, v2
-; GFX7-NEXT: v_and_b32_e32 v7, v3, v6
-; GFX7-NEXT: v_add_f32_e32 v2, v2, v5
-; GFX7-NEXT: v_cvt_f16_f32_e32 v2, v2
-; GFX7-NEXT: v_lshlrev_b32_e32 v2, v4, v2
-; GFX7-NEXT: v_or_b32_e32 v2, v7, v2
; GFX7-NEXT: v_mov_b32_e32 v8, v3
-; GFX7-NEXT: v_mov_b32_e32 v7, v2
-; GFX7-NEXT: buffer_atomic_cmpswap v[7:8], v[0:1], s[4:7], 0 addr64 glc
+; GFX7-NEXT: v_lshrrev_b32_e32 v3, v2, v8
+; GFX7-NEXT: v_cvt_f32_f16_e32 v3, v3
+; GFX7-NEXT: v_and_b32_e32 v4, v8, v6
+; GFX7-NEXT: v_add_f32_e32 v3, v3, v5
+; GFX7-NEXT: v_cvt_f16_f32_e32 v3, v3
+; GFX7-NEXT: v_lshlrev_b32_e32 v3, v2, v3
+; GFX7-NEXT: v_or_b32_e32 v7, v4, v3
+; GFX7-NEXT: v_mov_b32_e32 v3, v7
+; GFX7-NEXT: v_mov_b32_e32 v4, v8
+; GFX7-NEXT: buffer_atomic_cmpswap v[3:4], v[0:1], s[4:7], 0 addr64 glc
; GFX7-NEXT: s_waitcnt vmcnt(0)
; GFX7-NEXT: buffer_wbinvl1
-; GFX7-NEXT: v_cmp_eq_u32_e32 vcc, v7, v3
+; GFX7-NEXT: v_cmp_eq_u32_e32 vcc, v3, v8
; GFX7-NEXT: s_or_b64 s[8:9], vcc, s[8:9]
-; GFX7-NEXT: v_mov_b32_e32 v3, v7
; GFX7-NEXT: s_andn2_b64 exec, exec, s[8:9]
; GFX7-NEXT: s_cbranch_execnz .LBB49_1
; GFX7-NEXT: ; %bb.2: ; %atomicrmw.end
@@ -10718,32 +10719,32 @@ define void @global_agent_atomic_fadd_noret_f16__offset12b_neg__amdgpu_no_fine_g
; GFX6-NEXT: s_mov_b32 s4, s6
; GFX6-NEXT: s_mov_b32 s5, s6
; GFX6-NEXT: buffer_load_dword v3, v[0:1], s[4:7], 0 addr64
-; GFX6-NEXT: v_cvt_f16_f32_e32 v2, v2
-; GFX6-NEXT: v_and_b32_e32 v4, 3, v4
-; GFX6-NEXT: v_lshlrev_b32_e32 v4, 3, v4
+; GFX6-NEXT: v_cvt_f16_f32_e32 v5, v2
+; GFX6-NEXT: v_and_b32_e32 v2, 3, v4
+; GFX6-NEXT: v_lshlrev_b32_e32 v2, 3, v2
+; GFX6-NEXT: v_lshl_b32_e32 v4, 0xffff, v2
+; GFX6-NEXT: v_cvt_f32_f16_e32 v5, v5
+; GFX6-NEXT: v_not_b32_e32 v6, v4
; GFX6-NEXT: s_mov_b64 s[8:9], 0
-; GFX6-NEXT: v_cvt_f32_f16_e32 v5, v2
-; GFX6-NEXT: v_lshl_b32_e32 v2, 0xffff, v4
-; GFX6-NEXT: v_not_b32_e32 v6, v2
; GFX6-NEXT: .LBB49_1: ; %atomicrmw.start
; GFX6-NEXT: ; =>This Inner Loop Header: Depth=1
; GFX6-NEXT: s_waitcnt vmcnt(0)
-; GFX6-NEXT: v_lshrrev_b32_e32 v2, v4, v3
-; GFX6-NEXT: v_cvt_f32_f16_e32 v2, v2
-; GFX6-NEXT: s_waitcnt expcnt(0)
-; GFX6-NEXT: v_and_b32_e32 v7, v3, v6
-; GFX6-NEXT: v_add_f32_e32 v2, v2, v5
-; GFX6-NEXT: v_cvt_f16_f32_e32 v2, v2
-; GFX6-NEXT: v_lshlrev_b32_e32 v2, v4, v2
-; GFX6-NEXT: v_or_b32_e32 v2, v7, v2
; GFX6-NEXT: v_mov_b32_e32 v8, v3
-; GFX6-NEXT: v_mov_b32_e32 v7, v2
-; GFX6-NEXT: buffer_atomic_cmpswap v[7:8], v[0:1], s[4:7], 0 addr64 glc
+; GFX6-NEXT: s_waitcnt expcnt(0)
+; GFX6-NEXT: v_lshrrev_b32_e32 v3, v2, v8
+; GFX6-NEXT: v_cvt_f32_f16_e32 v3, v3
+; GFX6-NEXT: v_and_b32_e32 v4, v8, v6
+; GFX6-NEXT: v_add_f32_e32 v3, v3, v5
+; GFX6-NEXT: v_cvt_f16_f32_e32 v3, v3
+; GFX6-NEXT: v_lshlrev_b32_e32 v3, v2, v3
+; GFX6-NEXT: v_or_b32_e32 v7, v4, v3
+; GFX6-NEXT: v_mov_b32_e32 v3, v7
+; GFX6-NEXT: v_mov_b32_e32 v4, v8
+; GFX6-NEXT: buffer_atomic_cmpswap v[3:4], v[0:1], s[4:7], 0 addr64 glc
; GFX6-NEXT: s_waitcnt vmcnt(0)
; GFX6-NEXT: buffer_wbinvl1
-; GFX6-NEXT: v_cmp_eq_u32_e32 vcc, v7, v3
+; GFX6-NEXT: v_cmp_eq_u32_e32 vcc, v3, v8
; GFX6-NEXT: s_or_b64 s[8:9], vcc, s[8:9]
-; GFX6-NEXT: v_mov_b32_e32 v3, v7
; GFX6-NEXT: s_andn2_b64 exec, exec, s[8:9]
; GFX6-NEXT: s_cbranch_execnz .LBB49_1
; GFX6-NEXT: ; %bb.2: ; %atomicrmw.end
@@ -11016,28 +11017,28 @@ define half @global_agent_atomic_fadd_ret_f16__offset12b_pos__align4__amdgpu_no_
; GFX7-NEXT: buffer_load_dword v3, v[0:1], s[4:7], 0 addr64 offset:2046
; GFX7-NEXT: v_cvt_f16_f32_e32 v2, v2
; GFX7-NEXT: s_mov_b64 s[8:9], 0
-; GFX7-NEXT: v_cvt_f32_f16_e32 v6, v2
+; GFX7-NEXT: v_cvt_f32_f16_e32 v2, v2
; GFX7-NEXT: .LBB50_1: ; %atomicrmw.start
; GFX7-NEXT: ; =>This Inner Loop Header: Depth=1
; GFX7-NEXT: s_waitcnt vmcnt(0)
-; GFX7-NEXT: v_cvt_f32_f16_e32 v2, v3
-; GFX7-NEXT: v_and_b32_e32 v4, 0xffff0000, v3
-; GFX7-NEXT: v_add_f32_e32 v2, v2, v6
-; GFX7-NEXT: v_cvt_f16_f32_e32 v2, v2
-; GFX7-NEXT: v_or_b32_e32 v2, v4, v2
; GFX7-NEXT: v_mov_b32_e32 v5, v3
-; GFX7-NEXT: v_mov_b32_e32 v4, v2
-; GFX7-NEXT: buffer_atomic_cmpswap v[4:5], v[0:1], s[4:7], 0 addr64 offset:2046 glc
+; GFX7-NEXT: v_cvt_f32_f16_e32 v3, v5
+; GFX7-NEXT: v_and_b32_e32 v4, 0xffff0000, v5
+; GFX7-NEXT: v_add_f32_e32 v3, v3, v2
+; GFX7-NEXT: v_cvt_f16_f32_e32 v3, v3
+; GFX7-NEXT: v_or_b32_e32 v4, v4, v3
+; GFX7-NEXT: v_mov_b32_e32 v3, v4
+; GFX7-NEXT: v_mov_b32_e32 v4, v5
+; GFX7-NEXT: buffer_atomic_cmpswap v[3:4], v[0:1], s[4:7], 0 addr64 offset:2046 glc
; GFX7-NEXT: s_waitcnt vmcnt(0)
; GFX7-NEXT: buffer_wbinvl1
-; GFX7-NEXT: v_cmp_eq_u32_e32 vcc, v4, v3
+; GFX7-NEXT: v_cmp_eq_u32_e32 vcc, v3, v5
; GFX7-NEXT: s_or_b64 s[8:9], vcc, s[8:9]
-; GFX7-NEXT: v_mov_b32_e32 v3, v4
; GFX7-NEXT: s_andn2_b64 exec, exec, s[8:9]
; GFX7-NEXT: s_cbranch_execnz .LBB50_1
; GFX7-NEXT: ; %bb.2: ; %atomicrmw.end
; GFX7-NEXT: s_or_b64 exec, exec, s[8:9]
-; GFX7-NEXT: v_cvt_f32_f16_e32 v0, v4
+; GFX7-NEXT: v_cvt_f32_f16_e32 v0, v3
; GFX7-NEXT: s_setpc_b64 s[30:31]
; GFX6-LABEL: global_agent_atomic_fadd_ret_f16__offset12b_pos__align4__amdgpu_no_fine_grained_memory:
; GFX6: ; %bb.0:
@@ -11049,29 +11050,29 @@ define half @global_agent_atomic_fadd_ret_f16__offset12b_pos__align4__amdgpu_no_
; GFX6-NEXT: buffer_load_dword v3, v[0:1], s[4:7], 0 addr64 offset:2046
; GFX6-NEXT: v_cvt_f16_f32_e32 v2, v2
; GFX6-NEXT: s_mov_b64 s[8:9], 0
-; GFX6-NEXT: v_cvt_f32_f16_e32 v6, v2
+; GFX6-NEXT: v_cvt_f32_f16_e32 v2, v2
; GFX6-NEXT: .LBB50_1: ; %atomicrmw.start
; GFX6-NEXT: ; =>This Inner Loop Header: Depth=1
; GFX6-NEXT: s_waitcnt vmcnt(0)
-; GFX6-NEXT: v_cvt_f32_f16_e32 v2, v3
-; GFX6-NEXT: s_waitcnt expcnt(0)
-; GFX6-NEXT: v_and_b32_e32 v4, 0xffff0000, v3
-; GFX6-NEXT: v_add_f32_e32 v2, v2, v6
-; GFX6-NEXT: v_cvt_f16_f32_e32 v2, v2
-; GFX6-NEXT: v_or_b32_e32 v2, v4, v2
; GFX6-NEXT: v_mov_b32_e32 v5, v3
-; GFX6-NEXT: v_mov_b32_e32 v4, v2
-; GFX6-NEXT: buffer_atomic_cmpswap v[4:5], v[0:1], s[4:7], 0 addr64 offset:2046 glc
+; GFX6-NEXT: s_waitcnt expcnt(0)
+; GFX6-NEXT: v_cvt_f32_f16_e32 v3, v5
+; GFX6-NEXT: v_and_b32_e32 v4, 0xffff0000, v5
+; GFX6-NEXT: v_add_f32_e32 v3, v3, v2
+; GFX6-NEXT: v_cvt_f16_f32_e32 v3, v3
+; GFX6-NEXT: v_or_b32_e32 v4, v4, v3
+; GFX6-NEXT: v_mov_b32_e32 v3, v4
+; GFX6-NEXT: v_mov_b32_e32 v4, v5
+; GFX6-NEXT: buffer_atomic_cmpswap v[3:4], v[0:1], s[4:7], 0 addr64 offset:2046 glc
; GFX6-NEXT: s_waitcnt vmcnt(0)
; GFX6-NEXT: buffer_wbinvl1
-; GFX6-NEXT: v_cmp_eq_u32_e32 vcc, v4, v3
+; GFX6-NEXT: v_cmp_eq_u32_e32 vcc, v3, v5
; GFX6-NEXT: s_or_b64 s[8:9], vcc, s[8:9]
-; GFX6-NEXT: v_mov_b32_e32 v3, v4
; GFX6-NEXT: s_andn2_b64 exec, exec, s[8:9]
; GFX6-NEXT: s_cbranch_execnz .LBB50_1
; GFX6-NEXT: ; %bb.2: ; %atomicrmw.end
; GFX6-NEXT: s_or_b64 exec, exec, s[8:9]
-; GFX6-NEXT: v_cvt_f32_f16_e32 v0, v4
+; GFX6-NEXT: v_cvt_f32_f16_e32 v0, v3
; GFX6-NEXT: s_waitcnt expcnt(0)
; GFX6-NEXT: s_setpc_b64 s[30:31]
%gep = getelementptr half, ptr addrspace(1) %ptr, i64 1023
@@ -11328,23 +11329,23 @@ define void @global_agent_atomic_fadd_noret_f16__offset12b__align4_pos__amdgpu_n
; GFX7-NEXT: buffer_load_dword v3, v[0:1], s[4:7], 0 addr64 offset:2046
; GFX7-NEXT: v_cvt_f16_f32_e32 v2, v2
; GFX7-NEXT: s_mov_b64 s[8:9], 0
-; GFX7-NEXT: v_cvt_f32_f16_e32 v4, v2
+; GFX7-NEXT: v_cvt_f32_f16_e32 v2, v2
; GFX7-NEXT: .LBB51_1: ; %atomicrmw.start
; GFX7-NEXT: ; =>This Inner Loop Header: Depth=1
; GFX7-NEXT: s_waitcnt vmcnt(0)
-; GFX7-NEXT: v_cvt_f32_f16_e32 v2, v3
-; GFX7-NEXT: v_and_b32_e32 v5, 0xffff0000, v3
-; GFX7-NEXT: v_add_f32_e32 v2, v2, v4
-; GFX7-NEXT: v_cvt_f16_f32_e32 v2, v2
-; GFX7-NEXT: v_or_b32_e32 v2, v5, v2
-; GFX7-NEXT: v_mov_b32_e32 v6, v3
-; GFX7-NEXT: v_mov_b32_e32 v5, v2
-; GFX7-NEXT: buffer_atomic_cmpswap v[5:6], v[0:1], s[4:7], 0 addr64 offset:2046 glc
+; GFX7-NEXT: v_mov_b32_e32 v5, v3
+; GFX7-NEXT: v_cvt_f32_f16_e32 v3, v5
+; GFX7-NEXT: v_and_b32_e32 v4, 0xffff0000, v5
+; GFX7-NEXT: v_add_f32_e32 v3, v3, v2
+; GFX7-NEXT: v_cvt_f16_f32_e32 v3, v3
+; GFX7-NEXT: v_or_b32_e32 v4, v4, v3
+; GFX7-NEXT: v_mov_b32_e32 v3, v4
+; GFX7-NEXT: v_mov_b32_e32 v4, v5
+; GFX7-NEXT: buffer_atomic_cmpswap v[3:4], v[0:1], s[4:7], 0 addr64 offset:2046 glc
; GFX7-NEXT: s_waitcnt vmcnt(0)
; GFX7-NEXT: buffer_wbinvl1
-; GFX7-NEXT: v_cmp_eq_u32_e32 vcc, v5, v3
+; GFX7-NEXT: v_cmp_eq_u32_e32 vcc, v3, v5
; GFX7-NEXT: s_or_b64 s[8:9], vcc, s[8:9]
-; GFX7-NEXT: v_mov_b32_e32 v3, v5
; GFX7-NEXT: s_andn2_b64 exec, exec, s[8:9]
; GFX7-NEXT: s_cbranch_execnz .LBB51_1
; GFX7-NEXT: ; %bb.2: ; %atomicrmw.end
@@ -11360,24 +11361,24 @@ define void @global_agent_atomic_fadd_noret_f16__offset12b__align4_pos__amdgpu_n
; GFX6-NEXT: buffer_load_dword v3, v[0:1], s[4:7], 0 addr64 offset:2046
; GFX6-NEXT: v_cvt_f16_f32_e32 v2, v2
; GFX6-NEXT: s_mov_b64 s[8:9], 0
-; GFX6-NEXT: v_cvt_f32_f16_e32 v4, v2
+; GFX6-NEXT: v_cvt_f32_f16_e32 v2, v2
; GFX6-NEXT: .LBB51_1: ; %atomicrmw.start
; GFX6-NEXT: ; =>This Inner Loop Header: Depth=1
; GFX6-NEXT: s_waitcnt vmcnt(0)
-; GFX6-NEXT: v_cvt_f32_f16_e32 v2, v3
+; GFX6-NEXT: v_mov_b32_e32 v5, v3
; GFX6-NEXT: s_waitcnt expcnt(0)
-; GFX6-NEXT: v_and_b32_e32 v5, 0xffff0000, v3
-; GFX6-NEXT: v_add_f32_e32 v2, v2, v4
-; GFX6-NEXT: v_cvt_f16_f32_e32 v2, v2
-; GFX6-NEXT: v_or_b32_e32 v2, v5, v2
-; GFX6-NEXT: v_mov_b32_e32 v6, v3
-; GFX6-NEXT: v_mov_b32_e32 v5, v2
-; GFX6-NEXT: buffer_atomic_cmpswap v[5:6], v[0:1], s[4:7], 0 addr64 offset:2046 glc
+; GFX6-NEXT: v_cvt_f32_f16_e32 v3, v5
+; GFX6-NEXT: v_and_b32_e32 v4, 0xffff0000, v5
+; GFX6-NEXT: v_add_f32_e32 v3, v3, v2
+; GFX6-NEXT: v_cvt_f16_f32_e32 v3, v3
+; GFX6-NEXT: v_or_b32_e32 v4, v4, v3
+; GFX6-NEXT: v_mov_b32_e32 v3, v4
+; GFX6-NEXT: v_mov_b32_e32 v4, v5
+; GFX6-NEXT: buffer_atomic_cmpswap v[3:4], v[0:1], s[4:7], 0 addr64 offset:2046 glc
; GFX6-NEXT: s_waitcnt vmcnt(0)
; GFX6-NEXT: buffer_wbinvl1
-; GFX6-NEXT: v_cmp_eq_u32_e32 vcc, v5, v3
+; GFX6-NEXT: v_cmp_eq_u32_e32 vcc, v3, v5
; GFX6-NEXT: s_or_b64 s[8:9], vcc, s[8:9]
-; GFX6-NEXT: v_mov_b32_e32 v3, v5
; GFX6-NEXT: s_andn2_b64 exec, exec, s[8:9]
; GFX6-NEXT: s_cbranch_execnz .LBB51_1
; GFX6-NEXT: ; %bb.2: ; %atomicrmw.end
@@ -11748,36 +11749,36 @@ define half @global_system_atomic_fadd_ret_f16__offset12b_pos__amdgpu_no_fine_gr
; GFX7-NEXT: s_mov_b32 s4, s6
; GFX7-NEXT: s_mov_b32 s5, s6
; GFX7-NEXT: buffer_load_dword v3, v[0:1], s[4:7], 0 addr64
-; GFX7-NEXT: v_cvt_f16_f32_e32 v2, v2
-; GFX7-NEXT: v_and_b32_e32 v4, 3, v4
-; GFX7-NEXT: v_lshlrev_b32_e32 v6, 3, v4
+; GFX7-NEXT: v_cvt_f16_f32_e32 v5, v2
+; GFX7-NEXT: v_and_b32_e32 v2, 3, v4
+; GFX7-NEXT: v_lshlrev_b32_e32 v2, 3, v2
+; GFX7-NEXT: v_lshl_b32_e32 v4, 0xffff, v2
+; GFX7-NEXT: v_cvt_f32_f16_e32 v5, v5
+; GFX7-NEXT: v_not_b32_e32 v6, v4
; GFX7-NEXT: s_mov_b64 s[8:9], 0
-; GFX7-NEXT: v_cvt_f32_f16_e32 v7, v2
-; GFX7-NEXT: v_lshl_b32_e32 v2, 0xffff, v6
-; GFX7-NEXT: v_not_b32_e32 v8, v2
; GFX7-NEXT: .LBB52_1: ; %atomicrmw.start
; GFX7-NEXT: ; =>This Inner Loop Header: Depth=1
; GFX7-NEXT: s_waitcnt vmcnt(0)
-; GFX7-NEXT: v_lshrrev_b32_e32 v2, v6, v3
-; GFX7-NEXT: v_cvt_f32_f16_e32 v2, v2
-; GFX7-NEXT: v_and_b32_e32 v4, v3, v8
-; GFX7-NEXT: v_add_f32_e32 v2, v2, v7
-; GFX7-NEXT: v_cvt_f16_f32_e32 v2, v2
-; GFX7-NEXT: v_lshlrev_b32_e32 v2, v6, v2
-; GFX7-NEXT: v_or_b32_e32 v2, v4, v2
-; GFX7-NEXT: v_mov_b32_e32 v5, v3
-; GFX7-NEXT: v_mov_b32_e32 v4, v2
-; GFX7-NEXT: buffer_atomic_cmpswap v[4:5], v[0:1], s[4:7], 0 addr64 glc
+; GFX7-NEXT: v_mov_b32_e32 v8, v3
+; GFX7-NEXT: v_lshrrev_b32_e32 v3, v2, v8
+; GFX7-NEXT: v_cvt_f32_f16_e32 v3, v3
+; GFX7-NEXT: v_and_b32_e32 v4, v8, v6
+; GFX7-NEXT: v_add_f32_e32 v3, v3, v5
+; GFX7-NEXT: v_cvt_f16_f32_e32 v3, v3
+; GFX7-NEXT: v_lshlrev_b32_e32 v3, v2, v3
+; GFX7-NEXT: v_or_b32_e32 v7, v4, v3
+; GFX7-NEXT: v_mov_b32_e32 v3, v7
+; GFX7-NEXT: v_mov_b32_e32 v4, v8
+; GFX7-NEXT: buffer_atomic_cmpswap v[3:4], v[0:1], s[4:7], 0 addr64 glc
; GFX7-NEXT: s_waitcnt vmcnt(0)
; GFX7-NEXT: buffer_wbinvl1
-; GFX7-NEXT: v_cmp_eq_u32_e32 vcc, v4, v3
+; GFX7-NEXT: v_cmp_eq_u32_e32 vcc, v3, v8
; GFX7-NEXT: s_or_b64 s[8:9], vcc, s[8:9]
-; GFX7-NEXT: v_mov_b32_e32 v3, v4
; GFX7-NEXT: s_andn2_b64 exec, exec, s[8:9]
; GFX7-NEXT: s_cbranch_execnz .LBB52_1
; GFX7-NEXT: ; %bb.2: ; %atomicrmw.end
; GFX7-NEXT: s_or_b64 exec, exec, s[8:9]
-; GFX7-NEXT: v_lshrrev_b32_e32 v0, v6, v4
+; GFX7-NEXT: v_lshrrev_b32_e32 v0, v2, v3
; GFX7-NEXT: v_cvt_f32_f16_e32 v0, v0
; GFX7-NEXT: s_setpc_b64 s[30:31]
; GFX6-LABEL: global_system_atomic_fadd_ret_f16__offset12b_pos__amdgpu_no_fine_grained_memory:
@@ -11791,37 +11792,37 @@ define half @global_system_atomic_fadd_ret_f16__offset12b_pos__amdgpu_no_fine_gr
; GFX6-NEXT: s_mov_b32 s4, s6
; GFX6-NEXT: s_mov_b32 s5, s6
; GFX6-NEXT: buffer_load_dword v3, v[0:1], s[4:7], 0 addr64
-; GFX6-NEXT: v_cvt_f16_f32_e32 v2, v2
-; GFX6-NEXT: v_and_b32_e32 v4, 3, v4
-; GFX6-NEXT: v_lshlrev_b32_e32 v6, 3, v4
+; GFX6-NEXT: v_cvt_f16_f32_e32 v5, v2
+; GFX6-NEXT: v_and_b32_e32 v2, 3, v4
+; GFX6-NEXT: v_lshlrev_b32_e32 v2, 3, v2
+; GFX6-NEXT: v_lshl_b32_e32 v4, 0xffff, v2
+; GFX6-NEXT: v_cvt_f32_f16_e32 v5, v5
+; GFX6-NEXT: v_not_b32_e32 v6, v4
; GFX6-NEXT: s_mov_b64 s[8:9], 0
-; GFX6-NEXT: v_cvt_f32_f16_e32 v7, v2
-; GFX6-NEXT: v_lshl_b32_e32 v2, 0xffff, v6
-; GFX6-NEXT: v_not_b32_e32 v8, v2
; GFX6-NEXT: .LBB52_1: ; %atomicrmw.start
; GFX6-NEXT: ; =>This Inner Loop Header: Depth=1
; GFX6-NEXT: s_waitcnt vmcnt(0)
-; GFX6-NEXT: v_lshrrev_b32_e32 v2, v6, v3
-; GFX6-NEXT: v_cvt_f32_f16_e32 v2, v2
+; GFX6-NEXT: v_mov_b32_e32 v8, v3
; GFX6-NEXT: s_waitcnt expcnt(0)
-; GFX6-NEXT: v_and_b32_e32 v4, v3, v8
-; GFX6-NEXT: v_add_f32_e32 v2, v2, v7
-; GFX6-NEXT: v_cvt_f16_f32_e32 v2, v2
-; GFX6-NEXT: v_lshlrev_b32_e32 v2, v6, v2
-; GFX6-NEXT: v_or_b32_e32 v2, v4, v2
-; GFX6-NEXT: v_mov_b32_e32 v5, v3
-; GFX6-NEXT: v_mov_b32_e32 v4, v2
-; GFX6-NEXT: buffer_atomic_cmpswap v[4:5], v[0:1], s[4:7], 0 addr64 glc
+; GFX6-NEXT: v_lshrrev_b32_e32 v3, v2, v8
+; GFX6-NEXT: v_cvt_f32_f16_e32 v3, v3
+; GFX6-NEXT: v_and_b32_e32 v4, v8, v6
+; GFX6-NEXT: v_add_f32_e32 v3, v3, v5
+; GFX6-NEXT: v_cvt_f16_f32_e32 v3, v3
+; GFX6-NEXT: v_lshlrev_b32_e32 v3, v2, v3
+; GFX6-NEXT: v_or_b32_e32 v7, v4, v3
+; GFX6-NEXT: v_mov_b32_e32 v3, v7
+; GFX6-NEXT: v_mov_b32_e32 v4, v8
+; GFX6-NEXT: buffer_atomic_cmpswap v[3:4], v[0:1], s[4:7], 0 addr64 glc
; GFX6-NEXT: s_waitcnt vmcnt(0)
; GFX6-NEXT: buffer_wbinvl1
-; GFX6-NEXT: v_cmp_eq_u32_e32 vcc, v4, v3
+; GFX6-NEXT: v_cmp_eq_u32_e32 vcc, v3, v8
; GFX6-NEXT: s_or_b64 s[8:9], vcc, s[8:9]
-; GFX6-NEXT: v_mov_b32_e32 v3, v4
; GFX6-NEXT: s_andn2_b64 exec, exec, s[8:9]
; GFX6-NEXT: s_cbranch_execnz .LBB52_1
; GFX6-NEXT: ; %bb.2: ; %atomicrmw.end
; GFX6-NEXT: s_or_b64 exec, exec, s[8:9]
-; GFX6-NEXT: v_lshrrev_b32_e32 v0, v6, v4
+; GFX6-NEXT: v_lshrrev_b32_e32 v0, v2, v3
; GFX6-NEXT: v_cvt_f32_f16_e32 v0, v0
; GFX6-NEXT: s_waitcnt expcnt(0)
; GFX6-NEXT: s_setpc_b64 s[30:31]
@@ -12176,31 +12177,31 @@ define void @global_system_atomic_fadd_noret_f16__offset12b_pos__amdgpu_no_fine_
; GFX7-NEXT: s_mov_b32 s4, s6
; GFX7-NEXT: s_mov_b32 s5, s6
; GFX7-NEXT: buffer_load_dword v3, v[0:1], s[4:7], 0 addr64
-; GFX7-NEXT: v_cvt_f16_f32_e32 v2, v2
-; GFX7-NEXT: v_and_b32_e32 v4, 3, v4
-; GFX7-NEXT: v_lshlrev_b32_e32 v4, 3, v4
+; GFX7-NEXT: v_cvt_f16_f32_e32 v5, v2
+; GFX7-NEXT: v_and_b32_e32 v2, 3, v4
+; GFX7-NEXT: v_lshlrev_b32_e32 v2, 3, v2
+; GFX7-NEXT: v_lshl_b32_e32 v4, 0xffff, v2
+; GFX7-NEXT: v_cvt_f32_f16_e32 v5, v5
+; GFX7-NEXT: v_not_b32_e32 v6, v4
; GFX7-NEXT: s_mov_b64 s[8:9], 0
-; GFX7-NEXT: v_cvt_f32_f16_e32 v5, v2
-; GFX7-NEXT: v_lshl_b32_e32 v2, 0xffff, v4
-; GFX7-NEXT: v_not_b32_e32 v6, v2
; GFX7-NEXT: .LBB53_1: ; %atomicrmw.start
; GFX7-NEXT: ; =>This Inner Loop Header: Depth=1
; GFX7-NEXT: s_waitcnt vmcnt(0)
-; GFX7-NEXT: v_lshrrev_b32_e32 v2, v4, v3
-; GFX7-NEXT: v_cvt_f32_f16_e32 v2, v2
-; GFX7-NEXT: v_and_b32_e32 v7, v3, v6
-; GFX7-NEXT: v_add_f32_e32 v2, v2, v5
-; GFX7-NEXT: v_cvt_f16_f32_e32 v2, v2
-; GFX7-NEXT: v_lshlrev_b32_e32 v2, v4, v2
-; GFX7-NEXT: v_or_b32_e32 v2, v7, v2
; GFX7-NEXT: v_mov_b32_e32 v8, v3
-; GFX7-NEXT: v_mov_b32_e32 v7, v2
-; GFX7-NEXT: buffer_atomic_cmpswap v[7:8], v[0:1], s[4:7], 0 addr64 glc
+; GFX7-NEXT: v_lshrrev_b32_e32 v3, v2, v8
+; GFX7-NEXT: v_cvt_f32_f16_e32 v3, v3
+; GFX7-NEXT: v_and_b32_e32 v4, v8, v6
+; GFX7-NEXT: v_add_f32_e32 v3, v3, v5
+; GFX7-NEXT: v_cvt_f16_f32_e32 v3, v3
+; GFX7-NEXT: v_lshlrev_b32_e32 v3, v2, v3
+; GFX7-NEXT: v_or_b32_e32 v7, v4, v3
+; GFX7-NEXT: v_mov_b32_e32 v3, v7
+; GFX7-NEXT: v_mov_b32_e32 v4, v8
+; GFX7-NEXT: buffer_atomic_cmpswap v[3:4], v[0:1], s[4:7], 0 addr64 glc
; GFX7-NEXT: s_waitcnt vmcnt(0)
; GFX7-NEXT: buffer_wbinvl1
-; GFX7-NEXT: v_cmp_eq_u32_e32 vcc, v7, v3
+; GFX7-NEXT: v_cmp_eq_u32_e32 vcc, v3, v8
; GFX7-NEXT: s_or_b64 s[8:9], vcc, s[8:9]
-; GFX7-NEXT: v_mov_b32_e32 v3, v7
; GFX7-NEXT: s_andn2_b64 exec, exec, s[8:9]
; GFX7-NEXT: s_cbranch_execnz .LBB53_1
; GFX7-NEXT: ; %bb.2: ; %atomicrmw.end
@@ -12217,32 +12218,32 @@ define void @global_system_atomic_fadd_noret_f16__offset12b_pos__amdgpu_no_fine_
; GFX6-NEXT: s_mov_b32 s4, s6
; GFX6-NEXT: s_mov_b32 s5, s6
; GFX6-NEXT: buffer_load_dword v3, v[0:1], s[4:7], 0 addr64
-; GFX6-NEXT: v_cvt_f16_f32_e32 v2, v2
-; GFX6-NEXT: v_and_b32_e32 v4, 3, v4
-; GFX6-NEXT: v_lshlrev_b32_e32 v4, 3, v4
+; GFX6-NEXT: v_cvt_f16_f32_e32 v5, v2
+; GFX6-NEXT: v_and_b32_e32 v2, 3, v4
+; GFX6-NEXT: v_lshlrev_b32_e32 v2, 3, v2
+; GFX6-NEXT: v_lshl_b32_e32 v4, 0xffff, v2
+; GFX6-NEXT: v_cvt_f32_f16_e32 v5, v5
+; GFX6-NEXT: v_not_b32_e32 v6, v4
; GFX6-NEXT: s_mov_b64 s[8:9], 0
-; GFX6-NEXT: v_cvt_f32_f16_e32 v5, v2
-; GFX6-NEXT: v_lshl_b32_e32 v2, 0xffff, v4
-; GFX6-NEXT: v_not_b32_e32 v6, v2
; GFX6-NEXT: .LBB53_1: ; %atomicrmw.start
; GFX6-NEXT: ; =>This Inner Loop Header: Depth=1
; GFX6-NEXT: s_waitcnt vmcnt(0)
-; GFX6-NEXT: v_lshrrev_b32_e32 v2, v4, v3
-; GFX6-NEXT: v_cvt_f32_f16_e32 v2, v2
-; GFX6-NEXT: s_waitcnt expcnt(0)
-; GFX6-NEXT: v_and_b32_e32 v7, v3, v6
-; GFX6-NEXT: v_add_f32_e32 v2, v2, v5
-; GFX6-NEXT: v_cvt_f16_f32_e32 v2, v2
-; GFX6-NEXT: v_lshlrev_b32_e32 v2, v4, v2
-; GFX6-NEXT: v_or_b32_e32 v2, v7, v2
; GFX6-NEXT: v_mov_b32_e32 v8, v3
-; GFX6-NEXT: v_mov_b32_e32 v7, v2
-; GFX6-NEXT: buffer_atomic_cmpswap v[7:8], v[0:1], s[4:7], 0 addr64 glc
+; GFX6-NEXT: s_waitcnt expcnt(0)
+; GFX6-NEXT: v_lshrrev_b32_e32 v3, v2, v8
+; GFX6-NEXT: v_cvt_f32_f16_e32 v3, v3
+; GFX6-NEXT: v_and_b32_e32 v4, v8, v6
+; GFX6-NEXT: v_add_f32_e32 v3, v3, v5
+; GFX6-NEXT: v_cvt_f16_f32_e32 v3, v3
+; GFX6-NEXT: v_lshlrev_b32_e32 v3, v2, v3
+; GFX6-NEXT: v_or_b32_e32 v7, v4, v3
+; GFX6-NEXT: v_mov_b32_e32 v3, v7
+; GFX6-NEXT: v_mov_b32_e32 v4, v8
+; GFX6-NEXT: buffer_atomic_cmpswap v[3:4], v[0:1], s[4:7], 0 addr64 glc
; GFX6-NEXT: s_waitcnt vmcnt(0)
; GFX6-NEXT: buffer_wbinvl1
-; GFX6-NEXT: v_cmp_eq_u32_e32 vcc, v7, v3
+; GFX6-NEXT: v_cmp_eq_u32_e32 vcc, v3, v8
; GFX6-NEXT: s_or_b64 s[8:9], vcc, s[8:9]
-; GFX6-NEXT: v_mov_b32_e32 v3, v7
; GFX6-NEXT: s_andn2_b64 exec, exec, s[8:9]
; GFX6-NEXT: s_cbranch_execnz .LBB53_1
; GFX6-NEXT: ; %bb.2: ; %atomicrmw.end
@@ -12677,35 +12678,35 @@ define bfloat @global_agent_atomic_fadd_ret_bf16__amdgpu_no_fine_grained_memory(
; GFX7-NEXT: s_mov_b32 s5, s6
; GFX7-NEXT: buffer_load_dword v4, v[0:1], s[4:7], 0 addr64
; GFX7-NEXT: v_and_b32_e32 v3, 3, v3
-; GFX7-NEXT: v_lshlrev_b32_e32 v5, 3, v3
-; GFX7-NEXT: v_lshl_b32_e32 v3, 0xffff, v5
+; GFX7-NEXT: v_lshlrev_b32_e32 v3, 3, v3
+; GFX7-NEXT: v_lshl_b32_e32 v5, 0xffff, v3
; GFX7-NEXT: v_mul_f32_e32 v2, 1.0, v2
-; GFX7-NEXT: v_not_b32_e32 v6, v3
+; GFX7-NEXT: v_not_b32_e32 v6, v5
; GFX7-NEXT: s_mov_b64 s[8:9], 0
-; GFX7-NEXT: v_and_b32_e32 v7, 0xffff0000, v2
+; GFX7-NEXT: v_and_b32_e32 v2, 0xffff0000, v2
; GFX7-NEXT: .LBB54_1: ; %atomicrmw.start
; GFX7-NEXT: ; =>This Inner Loop Header: Depth=1
; GFX7-NEXT: s_waitcnt vmcnt(0)
-; GFX7-NEXT: v_lshrrev_b32_e32 v2, v5, v4
-; GFX7-NEXT: v_lshlrev_b32_e32 v2, 16, v2
-; GFX7-NEXT: v_add_f32_e32 v2, v2, v7
-; GFX7-NEXT: v_lshrrev_b32_e32 v2, 16, v2
-; GFX7-NEXT: v_and_b32_e32 v3, v4, v6
-; GFX7-NEXT: v_lshlrev_b32_e32 v2, v5, v2
-; GFX7-NEXT: v_or_b32_e32 v3, v3, v2
-; GFX7-NEXT: v_mov_b32_e32 v2, v3
-; GFX7-NEXT: v_mov_b32_e32 v3, v4
-; GFX7-NEXT: buffer_atomic_cmpswap v[2:3], v[0:1], s[4:7], 0 addr64 glc
+; GFX7-NEXT: v_mov_b32_e32 v8, v4
+; GFX7-NEXT: v_lshrrev_b32_e32 v4, v3, v8
+; GFX7-NEXT: v_lshlrev_b32_e32 v4, 16, v4
+; GFX7-NEXT: v_add_f32_e32 v4, v4, v2
+; GFX7-NEXT: v_lshrrev_b32_e32 v4, 16, v4
+; GFX7-NEXT: v_and_b32_e32 v5, v8, v6
+; GFX7-NEXT: v_lshlrev_b32_e32 v4, v3, v4
+; GFX7-NEXT: v_or_b32_e32 v7, v5, v4
+; GFX7-NEXT: v_mov_b32_e32 v4, v7
+; GFX7-NEXT: v_mov_b32_e32 v5, v8
+; GFX7-NEXT: buffer_atomic_cmpswap v[4:5], v[0:1], s[4:7], 0 addr64 glc
; GFX7-NEXT: s_waitcnt vmcnt(0)
; GFX7-NEXT: buffer_wbinvl1
-; GFX7-NEXT: v_cmp_eq_u32_e32 vcc, v2, v4
+; GFX7-NEXT: v_cmp_eq_u32_e32 vcc, v4, v8
; GFX7-NEXT: s_or_b64 s[8:9], vcc, s[8:9]
-; GFX7-NEXT: v_mov_b32_e32 v4, v2
; GFX7-NEXT: s_andn2_b64 exec, exec, s[8:9]
; GFX7-NEXT: s_cbranch_execnz .LBB54_1
; GFX7-NEXT: ; %bb.2: ; %atomicrmw.end
; GFX7-NEXT: s_or_b64 exec, exec, s[8:9]
-; GFX7-NEXT: v_lshrrev_b32_e32 v0, v5, v2
+; GFX7-NEXT: v_lshrrev_b32_e32 v0, v3, v4
; GFX7-NEXT: v_lshlrev_b32_e32 v0, 16, v0
; GFX7-NEXT: s_setpc_b64 s[30:31]
; GFX6-LABEL: global_agent_atomic_fadd_ret_bf16__amdgpu_no_fine_grained_memory:
@@ -12719,35 +12720,36 @@ define bfloat @global_agent_atomic_fadd_ret_bf16__amdgpu_no_fine_grained_memory(
; GFX6-NEXT: s_mov_b32 s5, s6
; GFX6-NEXT: buffer_load_dword v4, v[0:1], s[4:7], 0 addr64
; GFX6-NEXT: v_and_b32_e32 v3, 3, v3
-; GFX6-NEXT: v_lshlrev_b32_e32 v5, 3, v3
-; GFX6-NEXT: v_lshl_b32_e32 v3, 0xffff, v5
+; GFX6-NEXT: v_lshlrev_b32_e32 v3, 3, v3
+; GFX6-NEXT: v_lshl_b32_e32 v5, 0xffff, v3
; GFX6-NEXT: v_mul_f32_e32 v2, 1.0, v2
-; GFX6-NEXT: v_not_b32_e32 v6, v3
+; GFX6-NEXT: v_not_b32_e32 v6, v5
; GFX6-NEXT: s_mov_b64 s[8:9], 0
-; GFX6-NEXT: v_and_b32_e32 v7, 0xffff0000, v2
+; GFX6-NEXT: v_and_b32_e32 v2, 0xffff0000, v2
; GFX6-NEXT: .LBB54_1: ; %atomicrmw.start
; GFX6-NEXT: ; =>This Inner Loop Header: Depth=1
-; GFX6-NEXT: s_waitcnt vmcnt(0) expcnt(0)
-; GFX6-NEXT: v_lshrrev_b32_e32 v2, v5, v4
-; GFX6-NEXT: v_lshlrev_b32_e32 v2, 16, v2
-; GFX6-NEXT: v_add_f32_e32 v2, v2, v7
-; GFX6-NEXT: v_lshrrev_b32_e32 v2, 16, v2
-; GFX6-NEXT: v_and_b32_e32 v3, v4, v6
-; GFX6-NEXT: v_lshlrev_b32_e32 v2, v5, v2
-; GFX6-NEXT: v_or_b32_e32 v3, v3, v2
-; GFX6-NEXT: v_mov_b32_e32 v2, v3
-; GFX6-NEXT: v_mov_b32_e32 v3, v4
-; GFX6-NEXT: buffer_atomic_cmpswap v[2:3], v[0:1], s[4:7], 0 addr64 glc
+; GFX6-NEXT: s_waitcnt vmcnt(0)
+; GFX6-NEXT: v_mov_b32_e32 v8, v4
+; GFX6-NEXT: s_waitcnt expcnt(0)
+; GFX6-NEXT: v_lshrrev_b32_e32 v4, v3, v8
+; GFX6-NEXT: v_lshlrev_b32_e32 v4, 16, v4
+; GFX6-NEXT: v_add_f32_e32 v4, v4, v2
+; GFX6-NEXT: v_lshrrev_b32_e32 v4, 16, v4
+; GFX6-NEXT: v_and_b32_e32 v5, v8, v6
+; GFX6-NEXT: v_lshlrev_b32_e32 v4, v3, v4
+; GFX6-NEXT: v_or_b32_e32 v7, v5, v4
+; GFX6-NEXT: v_mov_b32_e32 v4, v7
+; GFX6-NEXT: v_mov_b32_e32 v5, v8
+; GFX6-NEXT: buffer_atomic_cmpswap v[4:5], v[0:1], s[4:7], 0 addr64 glc
; GFX6-NEXT: s_waitcnt vmcnt(0)
; GFX6-NEXT: buffer_wbinvl1
-; GFX6-NEXT: v_cmp_eq_u32_e32 vcc, v2, v4
+; GFX6-NEXT: v_cmp_eq_u32_e32 vcc, v4, v8
; GFX6-NEXT: s_or_b64 s[8:9], vcc, s[8:9]
-; GFX6-NEXT: v_mov_b32_e32 v4, v2
; GFX6-NEXT: s_andn2_b64 exec, exec, s[8:9]
; GFX6-NEXT: s_cbranch_execnz .LBB54_1
; GFX6-NEXT: ; %bb.2: ; %atomicrmw.end
; GFX6-NEXT: s_or_b64 exec, exec, s[8:9]
-; GFX6-NEXT: v_lshrrev_b32_e32 v0, v5, v2
+; GFX6-NEXT: v_lshrrev_b32_e32 v0, v3, v4
; GFX6-NEXT: v_lshlrev_b32_e32 v0, 16, v0
; GFX6-NEXT: s_waitcnt expcnt(0)
; GFX6-NEXT: s_setpc_b64 s[30:31]
@@ -13189,35 +13191,35 @@ define bfloat @global_agent_atomic_fadd_ret_bf16__offset12b_pos__amdgpu_no_fine_
; GFX7-NEXT: s_mov_b32 s5, s6
; GFX7-NEXT: buffer_load_dword v3, v[0:1], s[4:7], 0 addr64
; GFX7-NEXT: v_and_b32_e32 v4, 3, v4
-; GFX7-NEXT: v_lshlrev_b32_e32 v6, 3, v4
-; GFX7-NEXT: v_lshl_b32_e32 v4, 0xffff, v6
+; GFX7-NEXT: v_lshlrev_b32_e32 v5, 3, v4
+; GFX7-NEXT: v_lshl_b32_e32 v4, 0xffff, v5
; GFX7-NEXT: v_mul_f32_e32 v2, 1.0, v2
-; GFX7-NEXT: v_not_b32_e32 v7, v4
+; GFX7-NEXT: v_not_b32_e32 v6, v4
; GFX7-NEXT: s_mov_b64 s[8:9], 0
-; GFX7-NEXT: v_and_b32_e32 v8, 0xffff0000, v2
+; GFX7-NEXT: v_and_b32_e32 v2, 0xffff0000, v2
; GFX7-NEXT: .LBB55_1: ; %atomicrmw.start
; GFX7-NEXT: ; =>This Inner Loop Header: Depth=1
; GFX7-NEXT: s_waitcnt vmcnt(0)
-; GFX7-NEXT: v_lshrrev_b32_e32 v2, v6, v3
-; GFX7-NEXT: v_lshlrev_b32_e32 v2, 16, v2
-; GFX7-NEXT: v_add_f32_e32 v2, v2, v8
-; GFX7-NEXT: v_lshrrev_b32_e32 v2, 16, v2
-; GFX7-NEXT: v_and_b32_e32 v4, v3, v7
-; GFX7-NEXT: v_lshlrev_b32_e32 v2, v6, v2
-; GFX7-NEXT: v_or_b32_e32 v2, v4, v2
-; GFX7-NEXT: v_mov_b32_e32 v5, v3
-; GFX7-NEXT: v_mov_b32_e32 v4, v2
-; GFX7-NEXT: buffer_atomic_cmpswap v[4:5], v[0:1], s[4:7], 0 addr64 glc
+; GFX7-NEXT: v_mov_b32_e32 v8, v3
+; GFX7-NEXT: v_lshrrev_b32_e32 v3, v5, v8
+; GFX7-NEXT: v_lshlrev_b32_e32 v3, 16, v3
+; GFX7-NEXT: v_add_f32_e32 v3, v3, v2
+; GFX7-NEXT: v_lshrrev_b32_e32 v3, 16, v3
+; GFX7-NEXT: v_and_b32_e32 v4, v8, v6
+; GFX7-NEXT: v_lshlrev_b32_e32 v3, v5, v3
+; GFX7-NEXT: v_or_b32_e32 v7, v4, v3
+; GFX7-NEXT: v_mov_b32_e32 v3, v7
+; GFX7-NEXT: v_mov_b32_e32 v4, v8
+; GFX7-NEXT: buffer_atomic_cmpswap v[3:4], v[0:1], s[4:7], 0 addr64 glc
; GFX7-NEXT: s_waitcnt vmcnt(0)
; GFX7-NEXT: buffer_wbinvl1
-; GFX7-NEXT: v_cmp_eq_u32_e32 vcc, v4, v3
+; GFX7-NEXT: v_cmp_eq_u32_e32 vcc, v3, v8
; GFX7-NEXT: s_or_b64 s[8:9], vcc, s[8:9]
-; GFX7-NEXT: v_mov_b32_e32 v3, v4
; GFX7-NEXT: s_andn2_b64 exec, exec, s[8:9]
; GFX7-NEXT: s_cbranch_execnz .LBB55_1
; GFX7-NEXT: ; %bb.2: ; %atomicrmw.end
; GFX7-NEXT: s_or_b64 exec, exec, s[8:9]
-; GFX7-NEXT: v_lshrrev_b32_e32 v0, v6, v4
+; GFX7-NEXT: v_lshrrev_b32_e32 v0, v5, v3
; GFX7-NEXT: v_lshlrev_b32_e32 v0, 16, v0
; GFX7-NEXT: s_setpc_b64 s[30:31]
; GFX6-LABEL: global_agent_atomic_fadd_ret_bf16__offset12b_pos__amdgpu_no_fine_grained_memory:
@@ -13232,36 +13234,36 @@ define bfloat @global_agent_atomic_fadd_ret_bf16__offset12b_pos__amdgpu_no_fine_
; GFX6-NEXT: s_mov_b32 s5, s6
; GFX6-NEXT: buffer_load_dword v3, v[0:1], s[4:7], 0 addr64
; GFX6-NEXT: v_and_b32_e32 v4, 3, v4
-; GFX6-NEXT: v_lshlrev_b32_e32 v6, 3, v4
-; GFX6-NEXT: v_lshl_b32_e32 v4, 0xffff, v6
+; GFX6-NEXT: v_lshlrev_b32_e32 v5, 3, v4
+; GFX6-NEXT: v_lshl_b32_e32 v4, 0xffff, v5
; GFX6-NEXT: v_mul_f32_e32 v2, 1.0, v2
-; GFX6-NEXT: v_not_b32_e32 v7, v4
+; GFX6-NEXT: v_not_b32_e32 v6, v4
; GFX6-NEXT: s_mov_b64 s[8:9], 0
-; GFX6-NEXT: v_and_b32_e32 v8, 0xffff0000, v2
+; GFX6-NEXT: v_and_b32_e32 v2, 0xffff0000, v2
; GFX6-NEXT: .LBB55_1: ; %atomicrmw.start
; GFX6-NEXT: ; =>This Inner Loop Header: Depth=1
; GFX6-NEXT: s_waitcnt vmcnt(0)
-; GFX6-NEXT: v_lshrrev_b32_e32 v2, v6, v3
-; GFX6-NEXT: v_lshlrev_b32_e32 v2, 16, v2
-; GFX6-NEXT: v_add_f32_e32 v2, v2, v8
-; GFX6-NEXT: v_lshrrev_b32_e32 v2, 16, v2
+; GFX6-NEXT: v_mov_b32_e32 v8, v3
; GFX6-NEXT: s_waitcnt expcnt(0)
-; GFX6-NEXT: v_and_b32_e32 v4, v3, v7
-; GFX6-NEXT: v_lshlrev_b32_e32 v2, v6, v2
-; GFX6-NEXT: v_or_b32_e32 v2, v4, v2
-; GFX6-NEXT: v_mov_b32_e32 v5, v3
-; GFX6-NEXT: v_mov_b32_e32 v4, v2
-; GFX6-NEXT: buffer_atomic_cmpswap v[4:5], v[0:1], s[4:7], 0 addr64 glc
+; GFX6-NEXT: v_lshrrev_b32_e32 v3, v5, v8
+; GFX6-NEXT: v_lshlrev_b32_e32 v3, 16, v3
+; GFX6-NEXT: v_add_f32_e32 v3, v3, v2
+; GFX6-NEXT: v_lshrrev_b32_e32 v3, 16, v3
+; GFX6-NEXT: v_and_b32_e32 v4, v8, v6
+; GFX6-NEXT: v_lshlrev_b32_e32 v3, v5, v3
+; GFX6-NEXT: v_or_b32_e32 v7, v4, v3
+; GFX6-NEXT: v_mov_b32_e32 v3, v7
+; GFX6-NEXT: v_mov_b32_e32 v4, v8
+; GFX6-NEXT: buffer_atomic_cmpswap v[3:4], v[0:1], s[4:7], 0 addr64 glc
; GFX6-NEXT: s_waitcnt vmcnt(0)
; GFX6-NEXT: buffer_wbinvl1
-; GFX6-NEXT: v_cmp_eq_u32_e32 vcc, v4, v3
+; GFX6-NEXT: v_cmp_eq_u32_e32 vcc, v3, v8
; GFX6-NEXT: s_or_b64 s[8:9], vcc, s[8:9]
-; GFX6-NEXT: v_mov_b32_e32 v3, v4
; GFX6-NEXT: s_andn2_b64 exec, exec, s[8:9]
; GFX6-NEXT: s_cbranch_execnz .LBB55_1
; GFX6-NEXT: ; %bb.2: ; %atomicrmw.end
; GFX6-NEXT: s_or_b64 exec, exec, s[8:9]
-; GFX6-NEXT: v_lshrrev_b32_e32 v0, v6, v4
+; GFX6-NEXT: v_lshrrev_b32_e32 v0, v5, v3
; GFX6-NEXT: v_lshlrev_b32_e32 v0, 16, v0
; GFX6-NEXT: s_waitcnt expcnt(0)
; GFX6-NEXT: s_setpc_b64 s[30:31]
@@ -13705,35 +13707,35 @@ define bfloat @global_agent_atomic_fadd_ret_bf16__offset12b_neg__amdgpu_no_fine_
; GFX7-NEXT: s_mov_b32 s5, s6
; GFX7-NEXT: buffer_load_dword v3, v[0:1], s[4:7], 0 addr64
; GFX7-NEXT: v_and_b32_e32 v4, 3, v4
-; GFX7-NEXT: v_lshlrev_b32_e32 v6, 3, v4
-; GFX7-NEXT: v_lshl_b32_e32 v4, 0xffff, v6
+; GFX7-NEXT: v_lshlrev_b32_e32 v5, 3, v4
+; GFX7-NEXT: v_lshl_b32_e32 v4, 0xffff, v5
; GFX7-NEXT: v_mul_f32_e32 v2, 1.0, v2
-; GFX7-NEXT: v_not_b32_e32 v7, v4
+; GFX7-NEXT: v_not_b32_e32 v6, v4
; GFX7-NEXT: s_mov_b64 s[8:9], 0
-; GFX7-NEXT: v_and_b32_e32 v8, 0xffff0000, v2
+; GFX7-NEXT: v_and_b32_e32 v2, 0xffff0000, v2
; GFX7-NEXT: .LBB56_1: ; %atomicrmw.start
; GFX7-NEXT: ; =>This Inner Loop Header: Depth=1
; GFX7-NEXT: s_waitcnt vmcnt(0)
-; GFX7-NEXT: v_lshrrev_b32_e32 v2, v6, v3
-; GFX7-NEXT: v_lshlrev_b32_e32 v2, 16, v2
-; GFX7-NEXT: v_add_f32_e32 v2, v2, v8
-; GFX7-NEXT: v_lshrrev_b32_e32 v2, 16, v2
-; GFX7-NEXT: v_and_b32_e32 v4, v3, v7
-; GFX7-NEXT: v_lshlrev_b32_e32 v2, v6, v2
-; GFX7-NEXT: v_or_b32_e32 v2, v4, v2
-; GFX7-NEXT: v_mov_b32_e32 v5, v3
-; GFX7-NEXT: v_mov_b32_e32 v4, v2
-; GFX7-NEXT: buffer_atomic_cmpswap v[4:5], v[0:1], s[4:7], 0 addr64 glc
+; GFX7-NEXT: v_mov_b32_e32 v8, v3
+; GFX7-NEXT: v_lshrrev_b32_e32 v3, v5, v8
+; GFX7-NEXT: v_lshlrev_b32_e32 v3, 16, v3
+; GFX7-NEXT: v_add_f32_e32 v3, v3, v2
+; GFX7-NEXT: v_lshrrev_b32_e32 v3, 16, v3
+; GFX7-NEXT: v_and_b32_e32 v4, v8, v6
+; GFX7-NEXT: v_lshlrev_b32_e32 v3, v5, v3
+; GFX7-NEXT: v_or_b32_e32 v7, v4, v3
+; GFX7-NEXT: v_mov_b32_e32 v3, v7
+; GFX7-NEXT: v_mov_b32_e32 v4, v8
+; GFX7-NEXT: buffer_atomic_cmpswap v[3:4], v[0:1], s[4:7], 0 addr64 glc
; GFX7-NEXT: s_waitcnt vmcnt(0)
; GFX7-NEXT: buffer_wbinvl1
-; GFX7-NEXT: v_cmp_eq_u32_e32 vcc, v4, v3
+; GFX7-NEXT: v_cmp_eq_u32_e32 vcc, v3, v8
; GFX7-NEXT: s_or_b64 s[8:9], vcc, s[8:9]
-; GFX7-NEXT: v_mov_b32_e32 v3, v4
; GFX7-NEXT: s_andn2_b64 exec, exec, s[8:9]
; GFX7-NEXT: s_cbranch_execnz .LBB56_1
; GFX7-NEXT: ; %bb.2: ; %atomicrmw.end
; GFX7-NEXT: s_or_b64 exec, exec, s[8:9]
-; GFX7-NEXT: v_lshrrev_b32_e32 v0, v6, v4
+; GFX7-NEXT: v_lshrrev_b32_e32 v0, v5, v3
; GFX7-NEXT: v_lshlrev_b32_e32 v0, 16, v0
; GFX7-NEXT: s_setpc_b64 s[30:31]
; GFX6-LABEL: global_agent_atomic_fadd_ret_bf16__offset12b_neg__amdgpu_no_fine_grained_memory:
@@ -13748,36 +13750,36 @@ define bfloat @global_agent_atomic_fadd_ret_bf16__offset12b_neg__amdgpu_no_fine_
; GFX6-NEXT: s_mov_b32 s5, s6
; GFX6-NEXT: buffer_load_dword v3, v[0:1], s[4:7], 0 addr64
; GFX6-NEXT: v_and_b32_e32 v4, 3, v4
-; GFX6-NEXT: v_lshlrev_b32_e32 v6, 3, v4
-; GFX6-NEXT: v_lshl_b32_e32 v4, 0xffff, v6
+; GFX6-NEXT: v_lshlrev_b32_e32 v5, 3, v4
+; GFX6-NEXT: v_lshl_b32_e32 v4, 0xffff, v5
; GFX6-NEXT: v_mul_f32_e32 v2, 1.0, v2
-; GFX6-NEXT: v_not_b32_e32 v7, v4
+; GFX6-NEXT: v_not_b32_e32 v6, v4
; GFX6-NEXT: s_mov_b64 s[8:9], 0
-; GFX6-NEXT: v_and_b32_e32 v8, 0xffff0000, v2
+; GFX6-NEXT: v_and_b32_e32 v2, 0xffff0000, v2
; GFX6-NEXT: .LBB56_1: ; %atomicrmw.start
; GFX6-NEXT: ; =>This Inner Loop Header: Depth=1
; GFX6-NEXT: s_waitcnt vmcnt(0)
-; GFX6-NEXT: v_lshrrev_b32_e32 v2, v6, v3
-; GFX6-NEXT: v_lshlrev_b32_e32 v2, 16, v2
-; GFX6-NEXT: v_add_f32_e32 v2, v2, v8
-; GFX6-NEXT: v_lshrrev_b32_e32 v2, 16, v2
+; GFX6-NEXT: v_mov_b32_e32 v8, v3
; GFX6-NEXT: s_waitcnt expcnt(0)
-; GFX6-NEXT: v_and_b32_e32 v4, v3, v7
-; GFX6-NEXT: v_lshlrev_b32_e32 v2, v6, v2
-; GFX6-NEXT: v_or_b32_e32 v2, v4, v2
-; GFX6-NEXT: v_mov_b32_e32 v5, v3
-; GFX6-NEXT: v_mov_b32_e32 v4, v2
-; GFX6-NEXT: buffer_atomic_cmpswap v[4:5], v[0:1], s[4:7], 0 addr64 glc
+; GFX6-NEXT: v_lshrrev_b32_e32 v3, v5, v8
+; GFX6-NEXT: v_lshlrev_b32_e32 v3, 16, v3
+; GFX6-NEXT: v_add_f32_e32 v3, v3, v2
+; GFX6-NEXT: v_lshrrev_b32_e32 v3, 16, v3
+; GFX6-NEXT: v_and_b32_e32 v4, v8, v6
+; GFX6-NEXT: v_lshlrev_b32_e32 v3, v5, v3
+; GFX6-NEXT: v_or_b32_e32 v7, v4, v3
+; GFX6-NEXT: v_mov_b32_e32 v3, v7
+; GFX6-NEXT: v_mov_b32_e32 v4, v8
+; GFX6-NEXT: buffer_atomic_cmpswap v[3:4], v[0:1], s[4:7], 0 addr64 glc
; GFX6-NEXT: s_waitcnt vmcnt(0)
; GFX6-NEXT: buffer_wbinvl1
-; GFX6-NEXT: v_cmp_eq_u32_e32 vcc, v4, v3
+; GFX6-NEXT: v_cmp_eq_u32_e32 vcc, v3, v8
; GFX6-NEXT: s_or_b64 s[8:9], vcc, s[8:9]
-; GFX6-NEXT: v_mov_b32_e32 v3, v4
; GFX6-NEXT: s_andn2_b64 exec, exec, s[8:9]
; GFX6-NEXT: s_cbranch_execnz .LBB56_1
; GFX6-NEXT: ; %bb.2: ; %atomicrmw.end
; GFX6-NEXT: s_or_b64 exec, exec, s[8:9]
-; GFX6-NEXT: v_lshrrev_b32_e32 v0, v6, v4
+; GFX6-NEXT: v_lshrrev_b32_e32 v0, v5, v3
; GFX6-NEXT: v_lshlrev_b32_e32 v0, 16, v0
; GFX6-NEXT: s_waitcnt expcnt(0)
; GFX6-NEXT: s_setpc_b64 s[30:31]
@@ -14192,30 +14194,30 @@ define void @global_agent_atomic_fadd_noret_bf16__amdgpu_no_fine_grained_memory(
; GFX7-NEXT: s_mov_b32 s5, s6
; GFX7-NEXT: buffer_load_dword v4, v[0:1], s[4:7], 0 addr64
; GFX7-NEXT: v_and_b32_e32 v3, 3, v3
-; GFX7-NEXT: v_lshlrev_b32_e32 v5, 3, v3
-; GFX7-NEXT: v_lshl_b32_e32 v3, 0xffff, v5
+; GFX7-NEXT: v_lshlrev_b32_e32 v3, 3, v3
+; GFX7-NEXT: v_lshl_b32_e32 v5, 0xffff, v3
; GFX7-NEXT: v_mul_f32_e32 v2, 1.0, v2
-; GFX7-NEXT: v_not_b32_e32 v6, v3
+; GFX7-NEXT: v_not_b32_e32 v6, v5
; GFX7-NEXT: s_mov_b64 s[8:9], 0
; GFX7-NEXT: v_and_b32_e32 v2, 0xffff0000, v2
; GFX7-NEXT: .LBB57_1: ; %atomicrmw.start
; GFX7-NEXT: ; =>This Inner Loop Header: Depth=1
; GFX7-NEXT: s_waitcnt vmcnt(0)
-; GFX7-NEXT: v_lshrrev_b32_e32 v3, v5, v4
-; GFX7-NEXT: v_lshlrev_b32_e32 v3, 16, v3
-; GFX7-NEXT: v_add_f32_e32 v3, v3, v2
-; GFX7-NEXT: v_lshrrev_b32_e32 v3, 16, v3
-; GFX7-NEXT: v_and_b32_e32 v7, v4, v6
-; GFX7-NEXT: v_lshlrev_b32_e32 v3, v5, v3
-; GFX7-NEXT: v_or_b32_e32 v3, v7, v3
; GFX7-NEXT: v_mov_b32_e32 v8, v4
-; GFX7-NEXT: v_mov_b32_e32 v7, v3
-; GFX7-NEXT: buffer_atomic_cmpswap v[7:8], v[0:1], s[4:7], 0 addr64 glc
+; GFX7-NEXT: v_lshrrev_b32_e32 v4, v3, v8
+; GFX7-NEXT: v_lshlrev_b32_e32 v4, 16, v4
+; GFX7-NEXT: v_add_f32_e32 v4, v4, v2
+; GFX7-NEXT: v_lshrrev_b32_e32 v4, 16, v4
+; GFX7-NEXT: v_and_b32_e32 v5, v8, v6
+; GFX7-NEXT: v_lshlrev_b32_e32 v4, v3, v4
+; GFX7-NEXT: v_or_b32_e32 v7, v5, v4
+; GFX7-NEXT: v_mov_b32_e32 v4, v7
+; GFX7-NEXT: v_mov_b32_e32 v5, v8
+; GFX7-NEXT: buffer_atomic_cmpswap v[4:5], v[0:1], s[4:7], 0 addr64 glc
; GFX7-NEXT: s_waitcnt vmcnt(0)
; GFX7-NEXT: buffer_wbinvl1
-; GFX7-NEXT: v_cmp_eq_u32_e32 vcc, v7, v4
+; GFX7-NEXT: v_cmp_eq_u32_e32 vcc, v4, v8
; GFX7-NEXT: s_or_b64 s[8:9], vcc, s[8:9]
-; GFX7-NEXT: v_mov_b32_e32 v4, v7
; GFX7-NEXT: s_andn2_b64 exec, exec, s[8:9]
; GFX7-NEXT: s_cbranch_execnz .LBB57_1
; GFX7-NEXT: ; %bb.2: ; %atomicrmw.end
@@ -14232,31 +14234,31 @@ define void @global_agent_atomic_fadd_noret_bf16__amdgpu_no_fine_grained_memory(
; GFX6-NEXT: s_mov_b32 s5, s6
; GFX6-NEXT: buffer_load_dword v4, v[0:1], s[4:7], 0 addr64
; GFX6-NEXT: v_and_b32_e32 v3, 3, v3
-; GFX6-NEXT: v_lshlrev_b32_e32 v5, 3, v3
-; GFX6-NEXT: v_lshl_b32_e32 v3, 0xffff, v5
+; GFX6-NEXT: v_lshlrev_b32_e32 v3, 3, v3
+; GFX6-NEXT: v_lshl_b32_e32 v5, 0xffff, v3
; GFX6-NEXT: v_mul_f32_e32 v2, 1.0, v2
-; GFX6-NEXT: v_not_b32_e32 v6, v3
+; GFX6-NEXT: v_not_b32_e32 v6, v5
; GFX6-NEXT: s_mov_b64 s[8:9], 0
; GFX6-NEXT: v_and_b32_e32 v2, 0xffff0000, v2
; GFX6-NEXT: .LBB57_1: ; %atomicrmw.start
; GFX6-NEXT: ; =>This Inner Loop Header: Depth=1
; GFX6-NEXT: s_waitcnt vmcnt(0)
-; GFX6-NEXT: v_lshrrev_b32_e32 v3, v5, v4
-; GFX6-NEXT: v_lshlrev_b32_e32 v3, 16, v3
-; GFX6-NEXT: v_add_f32_e32 v3, v3, v2
-; GFX6-NEXT: v_lshrrev_b32_e32 v3, 16, v3
-; GFX6-NEXT: s_waitcnt expcnt(0)
-; GFX6-NEXT: v_and_b32_e32 v7, v4, v6
-; GFX6-NEXT: v_lshlrev_b32_e32 v3, v5, v3
-; GFX6-NEXT: v_or_b32_e32 v3, v7, v3
; GFX6-NEXT: v_mov_b32_e32 v8, v4
-; GFX6-NEXT: v_mov_b32_e32 v7, v3
-; GFX6-NEXT: buffer_atomic_cmpswap v[7:8], v[0:1], s[4:7], 0 addr64 glc
+; GFX6-NEXT: s_waitcnt expcnt(0)
+; GFX6-NEXT: v_lshrrev_b32_e32 v4, v3, v8
+; GFX6-NEXT: v_lshlrev_b32_e32 v4, 16, v4
+; GFX6-NEXT: v_add_f32_e32 v4, v4, v2
+; GFX6-NEXT: v_lshrrev_b32_e32 v4, 16, v4
+; GFX6-NEXT: v_and_b32_e32 v5, v8, v6
+; GFX6-NEXT: v_lshlrev_b32_e32 v4, v3, v4
+; GFX6-NEXT: v_or_b32_e32 v7, v5, v4
+; GFX6-NEXT: v_mov_b32_e32 v4, v7
+; GFX6-NEXT: v_mov_b32_e32 v5, v8
+; GFX6-NEXT: buffer_atomic_cmpswap v[4:5], v[0:1], s[4:7], 0 addr64 glc
; GFX6-NEXT: s_waitcnt vmcnt(0)
; GFX6-NEXT: buffer_wbinvl1
-; GFX6-NEXT: v_cmp_eq_u32_e32 vcc, v7, v4
+; GFX6-NEXT: v_cmp_eq_u32_e32 vcc, v4, v8
; GFX6-NEXT: s_or_b64 s[8:9], vcc, s[8:9]
-; GFX6-NEXT: v_mov_b32_e32 v4, v7
; GFX6-NEXT: s_andn2_b64 exec, exec, s[8:9]
; GFX6-NEXT: s_cbranch_execnz .LBB57_1
; GFX6-NEXT: ; %bb.2: ; %atomicrmw.end
@@ -14688,30 +14690,30 @@ define void @global_agent_atomic_fadd_noret_bf16__offset12b_pos__amdgpu_no_fine_
; GFX7-NEXT: s_mov_b32 s5, s6
; GFX7-NEXT: buffer_load_dword v3, v[0:1], s[4:7], 0 addr64
; GFX7-NEXT: v_and_b32_e32 v4, 3, v4
-; GFX7-NEXT: v_lshlrev_b32_e32 v4, 3, v4
-; GFX7-NEXT: v_lshl_b32_e32 v5, 0xffff, v4
+; GFX7-NEXT: v_lshlrev_b32_e32 v5, 3, v4
+; GFX7-NEXT: v_lshl_b32_e32 v4, 0xffff, v5
; GFX7-NEXT: v_mul_f32_e32 v2, 1.0, v2
-; GFX7-NEXT: v_not_b32_e32 v5, v5
+; GFX7-NEXT: v_not_b32_e32 v6, v4
; GFX7-NEXT: s_mov_b64 s[8:9], 0
-; GFX7-NEXT: v_and_b32_e32 v6, 0xffff0000, v2
+; GFX7-NEXT: v_and_b32_e32 v2, 0xffff0000, v2
; GFX7-NEXT: .LBB58_1: ; %atomicrmw.start
; GFX7-NEXT: ; =>This Inner Loop Header: Depth=1
; GFX7-NEXT: s_waitcnt vmcnt(0)
-; GFX7-NEXT: v_lshrrev_b32_e32 v2, v4, v3
-; GFX7-NEXT: v_lshlrev_b32_e32 v2, 16, v2
-; GFX7-NEXT: v_add_f32_e32 v2, v2, v6
-; GFX7-NEXT: v_lshrrev_b32_e32 v2, 16, v2
-; GFX7-NEXT: v_and_b32_e32 v7, v3, v5
-; GFX7-NEXT: v_lshlrev_b32_e32 v2, v4, v2
-; GFX7-NEXT: v_or_b32_e32 v2, v7, v2
; GFX7-NEXT: v_mov_b32_e32 v8, v3
-; GFX7-NEXT: v_mov_b32_e32 v7, v2
-; GFX7-NEXT: buffer_atomic_cmpswap v[7:8], v[0:1], s[4:7], 0 addr64 glc
+; GFX7-NEXT: v_lshrrev_b32_e32 v3, v5, v8
+; GFX7-NEXT: v_lshlrev_b32_e32 v3, 16, v3
+; GFX7-NEXT: v_add_f32_e32 v3, v3, v2
+; GFX7-NEXT: v_lshrrev_b32_e32 v3, 16, v3
+; GFX7-NEXT: v_and_b32_e32 v4, v8, v6
+; GFX7-NEXT: v_lshlrev_b32_e32 v3, v5, v3
+; GFX7-NEXT: v_or_b32_e32 v7, v4, v3
+; GFX7-NEXT: v_mov_b32_e32 v3, v7
+; GFX7-NEXT: v_mov_b32_e32 v4, v8
+; GFX7-NEXT: buffer_atomic_cmpswap v[3:4], v[0:1], s[4:7], 0 addr64 glc
; GFX7-NEXT: s_waitcnt vmcnt(0)
; GFX7-NEXT: buffer_wbinvl1
-; GFX7-NEXT: v_cmp_eq_u32_e32 vcc, v7, v3
+; GFX7-NEXT: v_cmp_eq_u32_e32 vcc, v3, v8
; GFX7-NEXT: s_or_b64 s[8:9], vcc, s[8:9]
-; GFX7-NEXT: v_mov_b32_e32 v3, v7
; GFX7-NEXT: s_andn2_b64 exec, exec, s[8:9]
; GFX7-NEXT: s_cbranch_execnz .LBB58_1
; GFX7-NEXT: ; %bb.2: ; %atomicrmw.end
@@ -14729,31 +14731,31 @@ define void @global_agent_atomic_fadd_noret_bf16__offset12b_pos__amdgpu_no_fine_
; GFX6-NEXT: s_mov_b32 s5, s6
; GFX6-NEXT: buffer_load_dword v3, v[0:1], s[4:7], 0 addr64
; GFX6-NEXT: v_and_b32_e32 v4, 3, v4
-; GFX6-NEXT: v_lshlrev_b32_e32 v4, 3, v4
-; GFX6-NEXT: v_lshl_b32_e32 v5, 0xffff, v4
+; GFX6-NEXT: v_lshlrev_b32_e32 v5, 3, v4
+; GFX6-NEXT: v_lshl_b32_e32 v4, 0xffff, v5
; GFX6-NEXT: v_mul_f32_e32 v2, 1.0, v2
-; GFX6-NEXT: v_not_b32_e32 v5, v5
+; GFX6-NEXT: v_not_b32_e32 v6, v4
; GFX6-NEXT: s_mov_b64 s[8:9], 0
-; GFX6-NEXT: v_and_b32_e32 v6, 0xffff0000, v2
+; GFX6-NEXT: v_and_b32_e32 v2, 0xffff0000, v2
; GFX6-NEXT: .LBB58_1: ; %atomicrmw.start
; GFX6-NEXT: ; =>This Inner Loop Header: Depth=1
; GFX6-NEXT: s_waitcnt vmcnt(0)
-; GFX6-NEXT: v_lshrrev_b32_e32 v2, v4, v3
-; GFX6-NEXT: v_lshlrev_b32_e32 v2, 16, v2
-; GFX6-NEXT: v_add_f32_e32 v2, v2, v6
-; GFX6-NEXT: v_lshrrev_b32_e32 v2, 16, v2
-; GFX6-NEXT: s_waitcnt expcnt(0)
-; GFX6-NEXT: v_and_b32_e32 v7, v3, v5
-; GFX6-NEXT: v_lshlrev_b32_e32 v2, v4, v2
-; GFX6-NEXT: v_or_b32_e32 v2, v7, v2
; GFX6-NEXT: v_mov_b32_e32 v8, v3
-; GFX6-NEXT: v_mov_b32_e32 v7, v2
-; GFX6-NEXT: buffer_atomic_cmpswap v[7:8], v[0:1], s[4:7], 0 addr64 glc
+; GFX6-NEXT: s_waitcnt expcnt(0)
+; GFX6-NEXT: v_lshrrev_b32_e32 v3, v5, v8
+; GFX6-NEXT: v_lshlrev_b32_e32 v3, 16, v3
+; GFX6-NEXT: v_add_f32_e32 v3, v3, v2
+; GFX6-NEXT: v_lshrrev_b32_e32 v3, 16, v3
+; GFX6-NEXT: v_and_b32_e32 v4, v8, v6
+; GFX6-NEXT: v_lshlrev_b32_e32 v3, v5, v3
+; GFX6-NEXT: v_or_b32_e32 v7, v4, v3
+; GFX6-NEXT: v_mov_b32_e32 v3, v7
+; GFX6-NEXT: v_mov_b32_e32 v4, v8
+; GFX6-NEXT: buffer_atomic_cmpswap v[3:4], v[0:1], s[4:7], 0 addr64 glc
; GFX6-NEXT: s_waitcnt vmcnt(0)
; GFX6-NEXT: buffer_wbinvl1
-; GFX6-NEXT: v_cmp_eq_u32_e32 vcc, v7, v3
+; GFX6-NEXT: v_cmp_eq_u32_e32 vcc, v3, v8
; GFX6-NEXT: s_or_b64 s[8:9], vcc, s[8:9]
-; GFX6-NEXT: v_mov_b32_e32 v3, v7
; GFX6-NEXT: s_andn2_b64 exec, exec, s[8:9]
; GFX6-NEXT: s_cbranch_execnz .LBB58_1
; GFX6-NEXT: ; %bb.2: ; %atomicrmw.end
@@ -15187,30 +15189,30 @@ define void @global_agent_atomic_fadd_noret_bf16__offset12b_neg__amdgpu_no_fine_
; GFX7-NEXT: s_mov_b32 s5, s6
; GFX7-NEXT: buffer_load_dword v3, v[0:1], s[4:7], 0 addr64
; GFX7-NEXT: v_and_b32_e32 v4, 3, v4
-; GFX7-NEXT: v_lshlrev_b32_e32 v4, 3, v4
-; GFX7-NEXT: v_lshl_b32_e32 v5, 0xffff, v4
+; GFX7-NEXT: v_lshlrev_b32_e32 v5, 3, v4
+; GFX7-NEXT: v_lshl_b32_e32 v4, 0xffff, v5
; GFX7-NEXT: v_mul_f32_e32 v2, 1.0, v2
-; GFX7-NEXT: v_not_b32_e32 v5, v5
+; GFX7-NEXT: v_not_b32_e32 v6, v4
; GFX7-NEXT: s_mov_b64 s[8:9], 0
-; GFX7-NEXT: v_and_b32_e32 v6, 0xffff0000, v2
+; GFX7-NEXT: v_and_b32_e32 v2, 0xffff0000, v2
; GFX7-NEXT: .LBB59_1: ; %atomicrmw.start
; GFX7-NEXT: ; =>This Inner Loop Header: Depth=1
; GFX7-NEXT: s_waitcnt vmcnt(0)
-; GFX7-NEXT: v_lshrrev_b32_e32 v2, v4, v3
-; GFX7-NEXT: v_lshlrev_b32_e32 v2, 16, v2
-; GFX7-NEXT: v_add_f32_e32 v2, v2, v6
-; GFX7-NEXT: v_lshrrev_b32_e32 v2, 16, v2
-; GFX7-NEXT: v_and_b32_e32 v7, v3, v5
-; GFX7-NEXT: v_lshlrev_b32_e32 v2, v4, v2
-; GFX7-NEXT: v_or_b32_e32 v2, v7, v2
; GFX7-NEXT: v_mov_b32_e32 v8, v3
-; GFX7-NEXT: v_mov_b32_e32 v7, v2
-; GFX7-NEXT: buffer_atomic_cmpswap v[7:8], v[0:1], s[4:7], 0 addr64 glc
+; GFX7-NEXT: v_lshrrev_b32_e32 v3, v5, v8
+; GFX7-NEXT: v_lshlrev_b32_e32 v3, 16, v3
+; GFX7-NEXT: v_add_f32_e32 v3, v3, v2
+; GFX7-NEXT: v_lshrrev_b32_e32 v3, 16, v3
+; GFX7-NEXT: v_and_b32_e32 v4, v8, v6
+; GFX7-NEXT: v_lshlrev_b32_e32 v3, v5, v3
+; GFX7-NEXT: v_or_b32_e32 v7, v4, v3
+; GFX7-NEXT: v_mov_b32_e32 v3, v7
+; GFX7-NEXT: v_mov_b32_e32 v4, v8
+; GFX7-NEXT: buffer_atomic_cmpswap v[3:4], v[0:1], s[4:7], 0 addr64 glc
; GFX7-NEXT: s_waitcnt vmcnt(0)
; GFX7-NEXT: buffer_wbinvl1
-; GFX7-NEXT: v_cmp_eq_u32_e32 vcc, v7, v3
+; GFX7-NEXT: v_cmp_eq_u32_e32 vcc, v3, v8
; GFX7-NEXT: s_or_b64 s[8:9], vcc, s[8:9]
-; GFX7-NEXT: v_mov_b32_e32 v3, v7
; GFX7-NEXT: s_andn2_b64 exec, exec, s[8:9]
; GFX7-NEXT: s_cbranch_execnz .LBB59_1
; GFX7-NEXT: ; %bb.2: ; %atomicrmw.end
@@ -15228,31 +15230,31 @@ define void @global_agent_atomic_fadd_noret_bf16__offset12b_neg__amdgpu_no_fine_
; GFX6-NEXT: s_mov_b32 s5, s6
; GFX6-NEXT: buffer_load_dword v3, v[0:1], s[4:7], 0 addr64
; GFX6-NEXT: v_and_b32_e32 v4, 3, v4
-; GFX6-NEXT: v_lshlrev_b32_e32 v4, 3, v4
-; GFX6-NEXT: v_lshl_b32_e32 v5, 0xffff, v4
+; GFX6-NEXT: v_lshlrev_b32_e32 v5, 3, v4
+; GFX6-NEXT: v_lshl_b32_e32 v4, 0xffff, v5
; GFX6-NEXT: v_mul_f32_e32 v2, 1.0, v2
-; GFX6-NEXT: v_not_b32_e32 v5, v5
+; GFX6-NEXT: v_not_b32_e32 v6, v4
; GFX6-NEXT: s_mov_b64 s[8:9], 0
-; GFX6-NEXT: v_and_b32_e32 v6, 0xffff0000, v2
+; GFX6-NEXT: v_and_b32_e32 v2, 0xffff0000, v2
; GFX6-NEXT: .LBB59_1: ; %atomicrmw.start
; GFX6-NEXT: ; =>This Inner Loop Header: Depth=1
; GFX6-NEXT: s_waitcnt vmcnt(0)
-; GFX6-NEXT: v_lshrrev_b32_e32 v2, v4, v3
-; GFX6-NEXT: v_lshlrev_b32_e32 v2, 16, v2
-; GFX6-NEXT: v_add_f32_e32 v2, v2, v6
-; GFX6-NEXT: v_lshrrev_b32_e32 v2, 16, v2
-; GFX6-NEXT: s_waitcnt expcnt(0)
-; GFX6-NEXT: v_and_b32_e32 v7, v3, v5
-; GFX6-NEXT: v_lshlrev_b32_e32 v2, v4, v2
-; GFX6-NEXT: v_or_b32_e32 v2, v7, v2
; GFX6-NEXT: v_mov_b32_e32 v8, v3
-; GFX6-NEXT: v_mov_b32_e32 v7, v2
-; GFX6-NEXT: buffer_atomic_cmpswap v[7:8], v[0:1], s[4:7], 0 addr64 glc
+; GFX6-NEXT: s_waitcnt expcnt(0)
+; GFX6-NEXT: v_lshrrev_b32_e32 v3, v5, v8
+; GFX6-NEXT: v_lshlrev_b32_e32 v3, 16, v3
+; GFX6-NEXT: v_add_f32_e32 v3, v3, v2
+; GFX6-NEXT: v_lshrrev_b32_e32 v3, 16, v3
+; GFX6-NEXT: v_and_b32_e32 v4, v8, v6
+; GFX6-NEXT: v_lshlrev_b32_e32 v3, v5, v3
+; GFX6-NEXT: v_or_b32_e32 v7, v4, v3
+; GFX6-NEXT: v_mov_b32_e32 v3, v7
+; GFX6-NEXT: v_mov_b32_e32 v4, v8
+; GFX6-NEXT: buffer_atomic_cmpswap v[3:4], v[0:1], s[4:7], 0 addr64 glc
; GFX6-NEXT: s_waitcnt vmcnt(0)
; GFX6-NEXT: buffer_wbinvl1
-; GFX6-NEXT: v_cmp_eq_u32_e32 vcc, v7, v3
+; GFX6-NEXT: v_cmp_eq_u32_e32 vcc, v3, v8
; GFX6-NEXT: s_or_b64 s[8:9], vcc, s[8:9]
-; GFX6-NEXT: v_mov_b32_e32 v3, v7
; GFX6-NEXT: s_andn2_b64 exec, exec, s[8:9]
; GFX6-NEXT: s_cbranch_execnz .LBB59_1
; GFX6-NEXT: ; %bb.2: ; %atomicrmw.end
@@ -15610,28 +15612,28 @@ define bfloat @global_agent_atomic_fadd_ret_bf16__offset12b_pos__align4__amdgpu_
; GFX7-NEXT: buffer_load_dword v3, v[0:1], s[4:7], 0 addr64 offset:2046
; GFX7-NEXT: v_mul_f32_e32 v2, 1.0, v2
; GFX7-NEXT: s_mov_b64 s[8:9], 0
-; GFX7-NEXT: v_and_b32_e32 v6, 0xffff0000, v2
+; GFX7-NEXT: v_and_b32_e32 v2, 0xffff0000, v2
; GFX7-NEXT: .LBB60_1: ; %atomicrmw.start
; GFX7-NEXT: ; =>This Inner Loop Header: Depth=1
; GFX7-NEXT: s_waitcnt vmcnt(0)
-; GFX7-NEXT: v_lshlrev_b32_e32 v2, 16, v3
-; GFX7-NEXT: v_add_f32_e32 v2, v2, v6
-; GFX7-NEXT: v_and_b32_e32 v4, 0xffff0000, v3
-; GFX7-NEXT: v_lshrrev_b32_e32 v2, 16, v2
-; GFX7-NEXT: v_or_b32_e32 v2, v4, v2
; GFX7-NEXT: v_mov_b32_e32 v5, v3
-; GFX7-NEXT: v_mov_b32_e32 v4, v2
-; GFX7-NEXT: buffer_atomic_cmpswap v[4:5], v[0:1], s[4:7], 0 addr64 offset:2046 glc
+; GFX7-NEXT: v_lshlrev_b32_e32 v3, 16, v5
+; GFX7-NEXT: v_add_f32_e32 v3, v3, v2
+; GFX7-NEXT: v_and_b32_e32 v4, 0xffff0000, v5
+; GFX7-NEXT: v_lshrrev_b32_e32 v3, 16, v3
+; GFX7-NEXT: v_or_b32_e32 v4, v4, v3
+; GFX7-NEXT: v_mov_b32_e32 v3, v4
+; GFX7-NEXT: v_mov_b32_e32 v4, v5
+; GFX7-NEXT: buffer_atomic_cmpswap v[3:4], v[0:1], s[4:7], 0 addr64 offset:2046 glc
; GFX7-NEXT: s_waitcnt vmcnt(0)
; GFX7-NEXT: buffer_wbinvl1
-; GFX7-NEXT: v_cmp_eq_u32_e32 vcc, v4, v3
+; GFX7-NEXT: v_cmp_eq_u32_e32 vcc, v3, v5
; GFX7-NEXT: s_or_b64 s[8:9], vcc, s[8:9]
-; GFX7-NEXT: v_mov_b32_e32 v3, v4
; GFX7-NEXT: s_andn2_b64 exec, exec, s[8:9]
; GFX7-NEXT: s_cbranch_execnz .LBB60_1
; GFX7-NEXT: ; %bb.2: ; %atomicrmw.end
; GFX7-NEXT: s_or_b64 exec, exec, s[8:9]
-; GFX7-NEXT: v_lshlrev_b32_e32 v0, 16, v4
+; GFX7-NEXT: v_lshlrev_b32_e32 v0, 16, v3
; GFX7-NEXT: s_setpc_b64 s[30:31]
; GFX6-LABEL: global_agent_atomic_fadd_ret_bf16__offset12b_pos__align4__amdgpu_no_fine_grained_memory:
; GFX6: ; %bb.0:
@@ -15643,29 +15645,29 @@ define bfloat @global_agent_atomic_fadd_ret_bf16__offset12b_pos__align4__amdgpu_
; GFX6-NEXT: buffer_load_dword v3, v[0:1], s[4:7], 0 addr64 offset:2046
; GFX6-NEXT: v_mul_f32_e32 v2, 1.0, v2
; GFX6-NEXT: s_mov_b64 s[8:9], 0
-; GFX6-NEXT: v_and_b32_e32 v6, 0xffff0000, v2
+; GFX6-NEXT: v_and_b32_e32 v2, 0xffff0000, v2
; GFX6-NEXT: .LBB60_1: ; %atomicrmw.start
; GFX6-NEXT: ; =>This Inner Loop Header: Depth=1
; GFX6-NEXT: s_waitcnt vmcnt(0)
-; GFX6-NEXT: v_lshlrev_b32_e32 v2, 16, v3
-; GFX6-NEXT: v_add_f32_e32 v2, v2, v6
-; GFX6-NEXT: s_waitcnt expcnt(0)
-; GFX6-NEXT: v_and_b32_e32 v4, 0xffff0000, v3
-; GFX6-NEXT: v_lshrrev_b32_e32 v2, 16, v2
-; GFX6-NEXT: v_or_b32_e32 v2, v4, v2
; GFX6-NEXT: v_mov_b32_e32 v5, v3
-; GFX6-NEXT: v_mov_b32_e32 v4, v2
-; GFX6-NEXT: buffer_atomic_cmpswap v[4:5], v[0:1], s[4:7], 0 addr64 offset:2046 glc
+; GFX6-NEXT: s_waitcnt expcnt(0)
+; GFX6-NEXT: v_lshlrev_b32_e32 v3, 16, v5
+; GFX6-NEXT: v_add_f32_e32 v3, v3, v2
+; GFX6-NEXT: v_and_b32_e32 v4, 0xffff0000, v5
+; GFX6-NEXT: v_lshrrev_b32_e32 v3, 16, v3
+; GFX6-NEXT: v_or_b32_e32 v4, v4, v3
+; GFX6-NEXT: v_mov_b32_e32 v3, v4
+; GFX6-NEXT: v_mov_b32_e32 v4, v5
+; GFX6-NEXT: buffer_atomic_cmpswap v[3:4], v[0:1], s[4:7], 0 addr64 offset:2046 glc
; GFX6-NEXT: s_waitcnt vmcnt(0)
; GFX6-NEXT: buffer_wbinvl1
-; GFX6-NEXT: v_cmp_eq_u32_e32 vcc, v4, v3
+; GFX6-NEXT: v_cmp_eq_u32_e32 vcc, v3, v5
; GFX6-NEXT: s_or_b64 s[8:9], vcc, s[8:9]
-; GFX6-NEXT: v_mov_b32_e32 v3, v4
; GFX6-NEXT: s_andn2_b64 exec, exec, s[8:9]
; GFX6-NEXT: s_cbranch_execnz .LBB60_1
; GFX6-NEXT: ; %bb.2: ; %atomicrmw.end
; GFX6-NEXT: s_or_b64 exec, exec, s[8:9]
-; GFX6-NEXT: v_lshlrev_b32_e32 v0, 16, v4
+; GFX6-NEXT: v_lshlrev_b32_e32 v0, 16, v3
; GFX6-NEXT: s_waitcnt expcnt(0)
; GFX6-NEXT: s_setpc_b64 s[30:31]
%gep = getelementptr bfloat, ptr addrspace(1) %ptr, i64 1023
@@ -16007,23 +16009,23 @@ define void @global_agent_atomic_fadd_noret_bf16__offset12b__align4_pos__amdgpu_
; GFX7-NEXT: buffer_load_dword v3, v[0:1], s[4:7], 0 addr64 offset:2046
; GFX7-NEXT: v_mul_f32_e32 v2, 1.0, v2
; GFX7-NEXT: s_mov_b64 s[8:9], 0
-; GFX7-NEXT: v_and_b32_e32 v4, 0xffff0000, v2
+; GFX7-NEXT: v_and_b32_e32 v2, 0xffff0000, v2
; GFX7-NEXT: .LBB61_1: ; %atomicrmw.start
; GFX7-NEXT: ; =>This Inner Loop Header: Depth=1
; GFX7-NEXT: s_waitcnt vmcnt(0)
-; GFX7-NEXT: v_lshlrev_b32_e32 v2, 16, v3
-; GFX7-NEXT: v_add_f32_e32 v2, v2, v4
-; GFX7-NEXT: v_and_b32_e32 v5, 0xffff0000, v3
-; GFX7-NEXT: v_lshrrev_b32_e32 v2, 16, v2
-; GFX7-NEXT: v_or_b32_e32 v2, v5, v2
-; GFX7-NEXT: v_mov_b32_e32 v6, v3
-; GFX7-NEXT: v_mov_b32_e32 v5, v2
-; GFX7-NEXT: buffer_atomic_cmpswap v[5:6], v[0:1], s[4:7], 0 addr64 offset:2046 glc
+; GFX7-NEXT: v_mov_b32_e32 v5, v3
+; GFX7-NEXT: v_lshlrev_b32_e32 v3, 16, v5
+; GFX7-NEXT: v_add_f32_e32 v3, v3, v2
+; GFX7-NEXT: v_and_b32_e32 v4, 0xffff0000, v5
+; GFX7-NEXT: v_lshrrev_b32_e32 v3, 16, v3
+; GFX7-NEXT: v_or_b32_e32 v4, v4, v3
+; GFX7-NEXT: v_mov_b32_e32 v3, v4
+; GFX7-NEXT: v_mov_b32_e32 v4, v5
+; GFX7-NEXT: buffer_atomic_cmpswap v[3:4], v[0:1], s[4:7], 0 addr64 offset:2046 glc
; GFX7-NEXT: s_waitcnt vmcnt(0)
; GFX7-NEXT: buffer_wbinvl1
-; GFX7-NEXT: v_cmp_eq_u32_e32 vcc, v5, v3
+; GFX7-NEXT: v_cmp_eq_u32_e32 vcc, v3, v5
; GFX7-NEXT: s_or_b64 s[8:9], vcc, s[8:9]
-; GFX7-NEXT: v_mov_b32_e32 v3, v5
; GFX7-NEXT: s_andn2_b64 exec, exec, s[8:9]
; GFX7-NEXT: s_cbranch_execnz .LBB61_1
; GFX7-NEXT: ; %bb.2: ; %atomicrmw.end
@@ -16039,24 +16041,24 @@ define void @global_agent_atomic_fadd_noret_bf16__offset12b__align4_pos__amdgpu_
; GFX6-NEXT: buffer_load_dword v3, v[0:1], s[4:7], 0 addr64 offset:2046
; GFX6-NEXT: v_mul_f32_e32 v2, 1.0, v2
; GFX6-NEXT: s_mov_b64 s[8:9], 0
-; GFX6-NEXT: v_and_b32_e32 v4, 0xffff0000, v2
+; GFX6-NEXT: v_and_b32_e32 v2, 0xffff0000, v2
; GFX6-NEXT: .LBB61_1: ; %atomicrmw.start
; GFX6-NEXT: ; =>This Inner Loop Header: Depth=1
; GFX6-NEXT: s_waitcnt vmcnt(0)
-; GFX6-NEXT: v_lshlrev_b32_e32 v2, 16, v3
-; GFX6-NEXT: v_add_f32_e32 v2, v2, v4
+; GFX6-NEXT: v_mov_b32_e32 v5, v3
; GFX6-NEXT: s_waitcnt expcnt(0)
-; GFX6-NEXT: v_and_b32_e32 v5, 0xffff0000, v3
-; GFX6-NEXT: v_lshrrev_b32_e32 v2, 16, v2
-; GFX6-NEXT: v_or_b32_e32 v2, v5, v2
-; GFX6-NEXT: v_mov_b32_e32 v6, v3
-; GFX6-NEXT: v_mov_b32_e32 v5, v2
-; GFX6-NEXT: buffer_atomic_cmpswap v[5:6], v[0:1], s[4:7], 0 addr64 offset:2046 glc
+; GFX6-NEXT: v_lshlrev_b32_e32 v3, 16, v5
+; GFX6-NEXT: v_add_f32_e32 v3, v3, v2
+; GFX6-NEXT: v_and_b32_e32 v4, 0xffff0000, v5
+; GFX6-NEXT: v_lshrrev_b32_e32 v3, 16, v3
+; GFX6-NEXT: v_or_b32_e32 v4, v4, v3
+; GFX6-NEXT: v_mov_b32_e32 v3, v4
+; GFX6-NEXT: v_mov_b32_e32 v4, v5
+; GFX6-NEXT: buffer_atomic_cmpswap v[3:4], v[0:1], s[4:7], 0 addr64 offset:2046 glc
; GFX6-NEXT: s_waitcnt vmcnt(0)
; GFX6-NEXT: buffer_wbinvl1
-; GFX6-NEXT: v_cmp_eq_u32_e32 vcc, v5, v3
+; GFX6-NEXT: v_cmp_eq_u32_e32 vcc, v3, v5
; GFX6-NEXT: s_or_b64 s[8:9], vcc, s[8:9]
-; GFX6-NEXT: v_mov_b32_e32 v3, v5
; GFX6-NEXT: s_andn2_b64 exec, exec, s[8:9]
; GFX6-NEXT: s_cbranch_execnz .LBB61_1
; GFX6-NEXT: ; %bb.2: ; %atomicrmw.end
@@ -16506,35 +16508,35 @@ define bfloat @global_system_atomic_fadd_ret_bf16__offset12b_pos__amdgpu_no_fine
; GFX7-NEXT: s_mov_b32 s5, s6
; GFX7-NEXT: buffer_load_dword v3, v[0:1], s[4:7], 0 addr64
; GFX7-NEXT: v_and_b32_e32 v4, 3, v4
-; GFX7-NEXT: v_lshlrev_b32_e32 v6, 3, v4
-; GFX7-NEXT: v_lshl_b32_e32 v4, 0xffff, v6
+; GFX7-NEXT: v_lshlrev_b32_e32 v5, 3, v4
+; GFX7-NEXT: v_lshl_b32_e32 v4, 0xffff, v5
; GFX7-NEXT: v_mul_f32_e32 v2, 1.0, v2
-; GFX7-NEXT: v_not_b32_e32 v7, v4
+; GFX7-NEXT: v_not_b32_e32 v6, v4
; GFX7-NEXT: s_mov_b64 s[8:9], 0
-; GFX7-NEXT: v_and_b32_e32 v8, 0xffff0000, v2
+; GFX7-NEXT: v_and_b32_e32 v2, 0xffff0000, v2
; GFX7-NEXT: .LBB62_1: ; %atomicrmw.start
; GFX7-NEXT: ; =>This Inner Loop Header: Depth=1
; GFX7-NEXT: s_waitcnt vmcnt(0)
-; GFX7-NEXT: v_lshrrev_b32_e32 v2, v6, v3
-; GFX7-NEXT: v_lshlrev_b32_e32 v2, 16, v2
-; GFX7-NEXT: v_add_f32_e32 v2, v2, v8
-; GFX7-NEXT: v_lshrrev_b32_e32 v2, 16, v2
-; GFX7-NEXT: v_and_b32_e32 v4, v3, v7
-; GFX7-NEXT: v_lshlrev_b32_e32 v2, v6, v2
-; GFX7-NEXT: v_or_b32_e32 v2, v4, v2
-; GFX7-NEXT: v_mov_b32_e32 v5, v3
-; GFX7-NEXT: v_mov_b32_e32 v4, v2
-; GFX7-NEXT: buffer_atomic_cmpswap v[4:5], v[0:1], s[4:7], 0 addr64 glc
+; GFX7-NEXT: v_mov_b32_e32 v8, v3
+; GFX7-NEXT: v_lshrrev_b32_e32 v3, v5, v8
+; GFX7-NEXT: v_lshlrev_b32_e32 v3, 16, v3
+; GFX7-NEXT: v_add_f32_e32 v3, v3, v2
+; GFX7-NEXT: v_lshrrev_b32_e32 v3, 16, v3
+; GFX7-NEXT: v_and_b32_e32 v4, v8, v6
+; GFX7-NEXT: v_lshlrev_b32_e32 v3, v5, v3
+; GFX7-NEXT: v_or_b32_e32 v7, v4, v3
+; GFX7-NEXT: v_mov_b32_e32 v3, v7
+; GFX7-NEXT: v_mov_b32_e32 v4, v8
+; GFX7-NEXT: buffer_atomic_cmpswap v[3:4], v[0:1], s[4:7], 0 addr64 glc
; GFX7-NEXT: s_waitcnt vmcnt(0)
; GFX7-NEXT: buffer_wbinvl1
-; GFX7-NEXT: v_cmp_eq_u32_e32 vcc, v4, v3
+; GFX7-NEXT: v_cmp_eq_u32_e32 vcc, v3, v8
; GFX7-NEXT: s_or_b64 s[8:9], vcc, s[8:9]
-; GFX7-NEXT: v_mov_b32_e32 v3, v4
; GFX7-NEXT: s_andn2_b64 exec, exec, s[8:9]
; GFX7-NEXT: s_cbranch_execnz .LBB62_1
; GFX7-NEXT: ; %bb.2: ; %atomicrmw.end
; GFX7-NEXT: s_or_b64 exec, exec, s[8:9]
-; GFX7-NEXT: v_lshrrev_b32_e32 v0, v6, v4
+; GFX7-NEXT: v_lshrrev_b32_e32 v0, v5, v3
; GFX7-NEXT: v_lshlrev_b32_e32 v0, 16, v0
; GFX7-NEXT: s_setpc_b64 s[30:31]
; GFX6-LABEL: global_system_atomic_fadd_ret_bf16__offset12b_pos__amdgpu_no_fine_grained_memory:
@@ -16549,36 +16551,36 @@ define bfloat @global_system_atomic_fadd_ret_bf16__offset12b_pos__amdgpu_no_fine
; GFX6-NEXT: s_mov_b32 s5, s6
; GFX6-NEXT: buffer_load_dword v3, v[0:1], s[4:7], 0 addr64
; GFX6-NEXT: v_and_b32_e32 v4, 3, v4
-; GFX6-NEXT: v_lshlrev_b32_e32 v6, 3, v4
-; GFX6-NEXT: v_lshl_b32_e32 v4, 0xffff, v6
+; GFX6-NEXT: v_lshlrev_b32_e32 v5, 3, v4
+; GFX6-NEXT: v_lshl_b32_e32 v4, 0xffff, v5
; GFX6-NEXT: v_mul_f32_e32 v2, 1.0, v2
-; GFX6-NEXT: v_not_b32_e32 v7, v4
+; GFX6-NEXT: v_not_b32_e32 v6, v4
; GFX6-NEXT: s_mov_b64 s[8:9], 0
-; GFX6-NEXT: v_and_b32_e32 v8, 0xffff0000, v2
+; GFX6-NEXT: v_and_b32_e32 v2, 0xffff0000, v2
; GFX6-NEXT: .LBB62_1: ; %atomicrmw.start
; GFX6-NEXT: ; =>This Inner Loop Header: Depth=1
; GFX6-NEXT: s_waitcnt vmcnt(0)
-; GFX6-NEXT: v_lshrrev_b32_e32 v2, v6, v3
-; GFX6-NEXT: v_lshlrev_b32_e32 v2, 16, v2
-; GFX6-NEXT: v_add_f32_e32 v2, v2, v8
-; GFX6-NEXT: v_lshrrev_b32_e32 v2, 16, v2
+; GFX6-NEXT: v_mov_b32_e32 v8, v3
; GFX6-NEXT: s_waitcnt expcnt(0)
-; GFX6-NEXT: v_and_b32_e32 v4, v3, v7
-; GFX6-NEXT: v_lshlrev_b32_e32 v2, v6, v2
-; GFX6-NEXT: v_or_b32_e32 v2, v4, v2
-; GFX6-NEXT: v_mov_b32_e32 v5, v3
-; GFX6-NEXT: v_mov_b32_e32 v4, v2
-; GFX6-NEXT: buffer_atomic_cmpswap v[4:5], v[0:1], s[4:7], 0 addr64 glc
+; GFX6-NEXT: v_lshrrev_b32_e32 v3, v5, v8
+; GFX6-NEXT: v_lshlrev_b32_e32 v3, 16, v3
+; GFX6-NEXT: v_add_f32_e32 v3, v3, v2
+; GFX6-NEXT: v_lshrrev_b32_e32 v3, 16, v3
+; GFX6-NEXT: v_and_b32_e32 v4, v8, v6
+; GFX6-NEXT: v_lshlrev_b32_e32 v3, v5, v3
+; GFX6-NEXT: v_or_b32_e32 v7, v4, v3
+; GFX6-NEXT: v_mov_b32_e32 v3, v7
+; GFX6-NEXT: v_mov_b32_e32 v4, v8
+; GFX6-NEXT: buffer_atomic_cmpswap v[3:4], v[0:1], s[4:7], 0 addr64 glc
; GFX6-NEXT: s_waitcnt vmcnt(0)
; GFX6-NEXT: buffer_wbinvl1
-; GFX6-NEXT: v_cmp_eq_u32_e32 vcc, v4, v3
+; GFX6-NEXT: v_cmp_eq_u32_e32 vcc, v3, v8
; GFX6-NEXT: s_or_b64 s[8:9], vcc, s[8:9]
-; GFX6-NEXT: v_mov_b32_e32 v3, v4
; GFX6-NEXT: s_andn2_b64 exec, exec, s[8:9]
; GFX6-NEXT: s_cbranch_execnz .LBB62_1
; GFX6-NEXT: ; %bb.2: ; %atomicrmw.end
; GFX6-NEXT: s_or_b64 exec, exec, s[8:9]
-; GFX6-NEXT: v_lshrrev_b32_e32 v0, v6, v4
+; GFX6-NEXT: v_lshrrev_b32_e32 v0, v5, v3
; GFX6-NEXT: v_lshlrev_b32_e32 v0, 16, v0
; GFX6-NEXT: s_waitcnt expcnt(0)
; GFX6-NEXT: s_setpc_b64 s[30:31]
@@ -17012,30 +17014,30 @@ define void @global_system_atomic_fadd_noret_bf16__offset12b_pos__amdgpu_no_fine
; GFX7-NEXT: s_mov_b32 s5, s6
; GFX7-NEXT: buffer_load_dword v3, v[0:1], s[4:7], 0 addr64
; GFX7-NEXT: v_and_b32_e32 v4, 3, v4
-; GFX7-NEXT: v_lshlrev_b32_e32 v4, 3, v4
-; GFX7-NEXT: v_lshl_b32_e32 v5, 0xffff, v4
+; GFX7-NEXT: v_lshlrev_b32_e32 v5, 3, v4
+; GFX7-NEXT: v_lshl_b32_e32 v4, 0xffff, v5
; GFX7-NEXT: v_mul_f32_e32 v2, 1.0, v2
-; GFX7-NEXT: v_not_b32_e32 v5, v5
+; GFX7-NEXT: v_not_b32_e32 v6, v4
; GFX7-NEXT: s_mov_b64 s[8:9], 0
-; GFX7-NEXT: v_and_b32_e32 v6, 0xffff0000, v2
+; GFX7-NEXT: v_and_b32_e32 v2, 0xffff0000, v2
; GFX7-NEXT: .LBB63_1: ; %atomicrmw.start
; GFX7-NEXT: ; =>This Inner Loop Header: Depth=1
; GFX7-NEXT: s_waitcnt vmcnt(0)
-; GFX7-NEXT: v_lshrrev_b32_e32 v2, v4, v3
-; GFX7-NEXT: v_lshlrev_b32_e32 v2, 16, v2
-; GFX7-NEXT: v_add_f32_e32 v2, v2, v6
-; GFX7-NEXT: v_lshrrev_b32_e32 v2, 16, v2
-; GFX7-NEXT: v_and_b32_e32 v7, v3, v5
-; GFX7-NEXT: v_lshlrev_b32_e32 v2, v4, v2
-; GFX7-NEXT: v_or_b32_e32 v2, v7, v2
; GFX7-NEXT: v_mov_b32_e32 v8, v3
-; GFX7-NEXT: v_mov_b32_e32 v7, v2
-; GFX7-NEXT: buffer_atomic_cmpswap v[7:8], v[0:1], s[4:7], 0 addr64 glc
+; GFX7-NEXT: v_lshrrev_b32_e32 v3, v5, v8
+; GFX7-NEXT: v_lshlrev_b32_e32 v3, 16, v3
+; GFX7-NEXT: v_add_f32_e32 v3, v3, v2
+; GFX7-NEXT: v_lshrrev_b32_e32 v3, 16, v3
+; GFX7-NEXT: v_and_b32_e32 v4, v8, v6
+; GFX7-NEXT: v_lshlrev_b32_e32 v3, v5, v3
+; GFX7-NEXT: v_or_b32_e32 v7, v4, v3
+; GFX7-NEXT: v_mov_b32_e32 v3, v7
+; GFX7-NEXT: v_mov_b32_e32 v4, v8
+; GFX7-NEXT: buffer_atomic_cmpswap v[3:4], v[0:1], s[4:7], 0 addr64 glc
; GFX7-NEXT: s_waitcnt vmcnt(0)
; GFX7-NEXT: buffer_wbinvl1
-; GFX7-NEXT: v_cmp_eq_u32_e32 vcc, v7, v3
+; GFX7-NEXT: v_cmp_eq_u32_e32 vcc, v3, v8
; GFX7-NEXT: s_or_b64 s[8:9], vcc, s[8:9]
-; GFX7-NEXT: v_mov_b32_e32 v3, v7
; GFX7-NEXT: s_andn2_b64 exec, exec, s[8:9]
; GFX7-NEXT: s_cbranch_execnz .LBB63_1
; GFX7-NEXT: ; %bb.2: ; %atomicrmw.end
@@ -17053,31 +17055,31 @@ define void @global_system_atomic_fadd_noret_bf16__offset12b_pos__amdgpu_no_fine
; GFX6-NEXT: s_mov_b32 s5, s6
; GFX6-NEXT: buffer_load_dword v3, v[0:1], s[4:7], 0 addr64
; GFX6-NEXT: v_and_b32_e32 v4, 3, v4
-; GFX6-NEXT: v_lshlrev_b32_e32 v4, 3, v4
-; GFX6-NEXT: v_lshl_b32_e32 v5, 0xffff, v4
+; GFX6-NEXT: v_lshlrev_b32_e32 v5, 3, v4
+; GFX6-NEXT: v_lshl_b32_e32 v4, 0xffff, v5
; GFX6-NEXT: v_mul_f32_e32 v2, 1.0, v2
-; GFX6-NEXT: v_not_b32_e32 v5, v5
+; GFX6-NEXT: v_not_b32_e32 v6, v4
; GFX6-NEXT: s_mov_b64 s[8:9], 0
-; GFX6-NEXT: v_and_b32_e32 v6, 0xffff0000, v2
+; GFX6-NEXT: v_and_b32_e32 v2, 0xffff0000, v2
; GFX6-NEXT: .LBB63_1: ; %atomicrmw.start
; GFX6-NEXT: ; =>This Inner Loop Header: Depth=1
; GFX6-NEXT: s_waitcnt vmcnt(0)
-; GFX6-NEXT: v_lshrrev_b32_e32 v2, v4, v3
-; GFX6-NEXT: v_lshlrev_b32_e32 v2, 16, v2
-; GFX6-NEXT: v_add_f32_e32 v2, v2, v6
-; GFX6-NEXT: v_lshrrev_b32_e32 v2, 16, v2
-; GFX6-NEXT: s_waitcnt expcnt(0)
-; GFX6-NEXT: v_and_b32_e32 v7, v3, v5
-; GFX6-NEXT: v_lshlrev_b32_e32 v2, v4, v2
-; GFX6-NEXT: v_or_b32_e32 v2, v7, v2
; GFX6-NEXT: v_mov_b32_e32 v8, v3
-; GFX6-NEXT: v_mov_b32_e32 v7, v2
-; GFX6-NEXT: buffer_atomic_cmpswap v[7:8], v[0:1], s[4:7], 0 addr64 glc
+; GFX6-NEXT: s_waitcnt expcnt(0)
+; GFX6-NEXT: v_lshrrev_b32_e32 v3, v5, v8
+; GFX6-NEXT: v_lshlrev_b32_e32 v3, 16, v3
+; GFX6-NEXT: v_add_f32_e32 v3, v3, v2
+; GFX6-NEXT: v_lshrrev_b32_e32 v3, 16, v3
+; GFX6-NEXT: v_and_b32_e32 v4, v8, v6
+; GFX6-NEXT: v_lshlrev_b32_e32 v3, v5, v3
+; GFX6-NEXT: v_or_b32_e32 v7, v4, v3
+; GFX6-NEXT: v_mov_b32_e32 v3, v7
+; GFX6-NEXT: v_mov_b32_e32 v4, v8
+; GFX6-NEXT: buffer_atomic_cmpswap v[3:4], v[0:1], s[4:7], 0 addr64 glc
; GFX6-NEXT: s_waitcnt vmcnt(0)
; GFX6-NEXT: buffer_wbinvl1
-; GFX6-NEXT: v_cmp_eq_u32_e32 vcc, v7, v3
+; GFX6-NEXT: v_cmp_eq_u32_e32 vcc, v3, v8
; GFX6-NEXT: s_or_b64 s[8:9], vcc, s[8:9]
-; GFX6-NEXT: v_mov_b32_e32 v3, v7
; GFX6-NEXT: s_andn2_b64 exec, exec, s[8:9]
; GFX6-NEXT: s_cbranch_execnz .LBB63_1
; GFX6-NEXT: ; %bb.2: ; %atomicrmw.end
diff --git a/llvm/test/CodeGen/AMDGPU/global-atomicrmw-fmax.ll b/llvm/test/CodeGen/AMDGPU/global-atomicrmw-fmax.ll
index a24d6c5ff2222..c8f85a39eac98 100644
--- a/llvm/test/CodeGen/AMDGPU/global-atomicrmw-fmax.ll
+++ b/llvm/test/CodeGen/AMDGPU/global-atomicrmw-fmax.ll
@@ -4806,36 +4806,36 @@ define half @global_agent_atomic_fmax_ret_f16__amdgpu_no_fine_grained_memory(ptr
; GFX7-NEXT: s_mov_b32 s4, s6
; GFX7-NEXT: s_mov_b32 s5, s6
; GFX7-NEXT: buffer_load_dword v4, v[0:1], s[4:7], 0 addr64
-; GFX7-NEXT: v_cvt_f16_f32_e32 v2, v2
-; GFX7-NEXT: v_and_b32_e32 v3, 3, v3
-; GFX7-NEXT: v_lshlrev_b32_e32 v5, 3, v3
+; GFX7-NEXT: v_cvt_f16_f32_e32 v5, v2
+; GFX7-NEXT: v_and_b32_e32 v2, 3, v3
+; GFX7-NEXT: v_lshlrev_b32_e32 v2, 3, v2
; GFX7-NEXT: s_mov_b64 s[8:9], 0
-; GFX7-NEXT: v_cvt_f32_f16_e32 v6, v2
-; GFX7-NEXT: v_lshl_b32_e32 v2, 0xffff, v5
-; GFX7-NEXT: v_not_b32_e32 v7, v2
+; GFX7-NEXT: v_cvt_f32_f16_e32 v3, v5
+; GFX7-NEXT: v_lshl_b32_e32 v5, 0xffff, v2
+; GFX7-NEXT: v_not_b32_e32 v6, v5
; GFX7-NEXT: .LBB26_1: ; %atomicrmw.start
; GFX7-NEXT: ; =>This Inner Loop Header: Depth=1
; GFX7-NEXT: s_waitcnt vmcnt(0)
-; GFX7-NEXT: v_lshrrev_b32_e32 v2, v5, v4
-; GFX7-NEXT: v_cvt_f32_f16_e32 v2, v2
-; GFX7-NEXT: v_and_b32_e32 v3, v4, v7
-; GFX7-NEXT: v_max_f32_e32 v2, v2, v6
-; GFX7-NEXT: v_cvt_f16_f32_e32 v2, v2
-; GFX7-NEXT: v_lshlrev_b32_e32 v2, v5, v2
-; GFX7-NEXT: v_or_b32_e32 v3, v3, v2
-; GFX7-NEXT: v_mov_b32_e32 v2, v3
-; GFX7-NEXT: v_mov_b32_e32 v3, v4
-; GFX7-NEXT: buffer_atomic_cmpswap v[2:3], v[0:1], s[4:7], 0 addr64 glc
+; GFX7-NEXT: v_mov_b32_e32 v8, v4
+; GFX7-NEXT: v_lshrrev_b32_e32 v4, v2, v8
+; GFX7-NEXT: v_cvt_f32_f16_e32 v4, v4
+; GFX7-NEXT: v_and_b32_e32 v5, v8, v6
+; GFX7-NEXT: v_max_f32_e32 v4, v4, v3
+; GFX7-NEXT: v_cvt_f16_f32_e32 v4, v4
+; GFX7-NEXT: v_lshlrev_b32_e32 v4, v2, v4
+; GFX7-NEXT: v_or_b32_e32 v7, v5, v4
+; GFX7-NEXT: v_mov_b32_e32 v4, v7
+; GFX7-NEXT: v_mov_b32_e32 v5, v8
+; GFX7-NEXT: buffer_atomic_cmpswap v[4:5], v[0:1], s[4:7], 0 addr64 glc
; GFX7-NEXT: s_waitcnt vmcnt(0)
; GFX7-NEXT: buffer_wbinvl1
-; GFX7-NEXT: v_cmp_eq_u32_e32 vcc, v2, v4
+; GFX7-NEXT: v_cmp_eq_u32_e32 vcc, v4, v8
; GFX7-NEXT: s_or_b64 s[8:9], vcc, s[8:9]
-; GFX7-NEXT: v_mov_b32_e32 v4, v2
; GFX7-NEXT: s_andn2_b64 exec, exec, s[8:9]
; GFX7-NEXT: s_cbranch_execnz .LBB26_1
; GFX7-NEXT: ; %bb.2: ; %atomicrmw.end
; GFX7-NEXT: s_or_b64 exec, exec, s[8:9]
-; GFX7-NEXT: v_lshrrev_b32_e32 v0, v5, v2
+; GFX7-NEXT: v_lshrrev_b32_e32 v0, v2, v4
; GFX7-NEXT: v_cvt_f32_f16_e32 v0, v0
; GFX7-NEXT: s_setpc_b64 s[30:31]
;
@@ -4849,36 +4849,37 @@ define half @global_agent_atomic_fmax_ret_f16__amdgpu_no_fine_grained_memory(ptr
; GFX6-NEXT: s_mov_b32 s4, s6
; GFX6-NEXT: s_mov_b32 s5, s6
; GFX6-NEXT: buffer_load_dword v4, v[0:1], s[4:7], 0 addr64
-; GFX6-NEXT: v_cvt_f16_f32_e32 v2, v2
-; GFX6-NEXT: v_and_b32_e32 v3, 3, v3
-; GFX6-NEXT: v_lshlrev_b32_e32 v5, 3, v3
+; GFX6-NEXT: v_cvt_f16_f32_e32 v5, v2
+; GFX6-NEXT: v_and_b32_e32 v2, 3, v3
+; GFX6-NEXT: v_lshlrev_b32_e32 v2, 3, v2
; GFX6-NEXT: s_mov_b64 s[8:9], 0
-; GFX6-NEXT: v_cvt_f32_f16_e32 v6, v2
-; GFX6-NEXT: v_lshl_b32_e32 v2, 0xffff, v5
-; GFX6-NEXT: v_not_b32_e32 v7, v2
+; GFX6-NEXT: v_cvt_f32_f16_e32 v3, v5
+; GFX6-NEXT: v_lshl_b32_e32 v5, 0xffff, v2
+; GFX6-NEXT: v_not_b32_e32 v6, v5
; GFX6-NEXT: .LBB26_1: ; %atomicrmw.start
; GFX6-NEXT: ; =>This Inner Loop Header: Depth=1
-; GFX6-NEXT: s_waitcnt vmcnt(0) expcnt(0)
-; GFX6-NEXT: v_lshrrev_b32_e32 v2, v5, v4
-; GFX6-NEXT: v_cvt_f32_f16_e32 v2, v2
-; GFX6-NEXT: v_and_b32_e32 v3, v4, v7
-; GFX6-NEXT: v_max_f32_e32 v2, v2, v6
-; GFX6-NEXT: v_cvt_f16_f32_e32 v2, v2
-; GFX6-NEXT: v_lshlrev_b32_e32 v2, v5, v2
-; GFX6-NEXT: v_or_b32_e32 v3, v3, v2
-; GFX6-NEXT: v_mov_b32_e32 v2, v3
-; GFX6-NEXT: v_mov_b32_e32 v3, v4
-; GFX6-NEXT: buffer_atomic_cmpswap v[2:3], v[0:1], s[4:7], 0 addr64 glc
+; GFX6-NEXT: s_waitcnt vmcnt(0)
+; GFX6-NEXT: v_mov_b32_e32 v8, v4
+; GFX6-NEXT: s_waitcnt expcnt(0)
+; GFX6-NEXT: v_lshrrev_b32_e32 v4, v2, v8
+; GFX6-NEXT: v_cvt_f32_f16_e32 v4, v4
+; GFX6-NEXT: v_and_b32_e32 v5, v8, v6
+; GFX6-NEXT: v_max_f32_e32 v4, v4, v3
+; GFX6-NEXT: v_cvt_f16_f32_e32 v4, v4
+; GFX6-NEXT: v_lshlrev_b32_e32 v4, v2, v4
+; GFX6-NEXT: v_or_b32_e32 v7, v5, v4
+; GFX6-NEXT: v_mov_b32_e32 v4, v7
+; GFX6-NEXT: v_mov_b32_e32 v5, v8
+; GFX6-NEXT: buffer_atomic_cmpswap v[4:5], v[0:1], s[4:7], 0 addr64 glc
; GFX6-NEXT: s_waitcnt vmcnt(0)
; GFX6-NEXT: buffer_wbinvl1
-; GFX6-NEXT: v_cmp_eq_u32_e32 vcc, v2, v4
+; GFX6-NEXT: v_cmp_eq_u32_e32 vcc, v4, v8
; GFX6-NEXT: s_or_b64 s[8:9], vcc, s[8:9]
-; GFX6-NEXT: v_mov_b32_e32 v4, v2
; GFX6-NEXT: s_andn2_b64 exec, exec, s[8:9]
; GFX6-NEXT: s_cbranch_execnz .LBB26_1
; GFX6-NEXT: ; %bb.2: ; %atomicrmw.end
; GFX6-NEXT: s_or_b64 exec, exec, s[8:9]
-; GFX6-NEXT: v_lshrrev_b32_e32 v0, v5, v2
+; GFX6-NEXT: v_lshrrev_b32_e32 v0, v2, v4
; GFX6-NEXT: v_cvt_f32_f16_e32 v0, v0
; GFX6-NEXT: s_waitcnt expcnt(0)
; GFX6-NEXT: s_setpc_b64 s[30:31]
@@ -5263,36 +5264,36 @@ define half @global_agent_atomic_fmax_ret_f16__offset12b_pos__amdgpu_no_fine_gra
; GFX7-NEXT: s_mov_b32 s4, s6
; GFX7-NEXT: s_mov_b32 s5, s6
; GFX7-NEXT: buffer_load_dword v3, v[0:1], s[4:7], 0 addr64
-; GFX7-NEXT: v_cvt_f16_f32_e32 v2, v2
-; GFX7-NEXT: v_and_b32_e32 v4, 3, v4
-; GFX7-NEXT: v_lshlrev_b32_e32 v6, 3, v4
+; GFX7-NEXT: v_cvt_f16_f32_e32 v5, v2
+; GFX7-NEXT: v_and_b32_e32 v2, 3, v4
+; GFX7-NEXT: v_lshlrev_b32_e32 v2, 3, v2
+; GFX7-NEXT: v_lshl_b32_e32 v4, 0xffff, v2
+; GFX7-NEXT: v_cvt_f32_f16_e32 v5, v5
+; GFX7-NEXT: v_not_b32_e32 v6, v4
; GFX7-NEXT: s_mov_b64 s[8:9], 0
-; GFX7-NEXT: v_cvt_f32_f16_e32 v7, v2
-; GFX7-NEXT: v_lshl_b32_e32 v2, 0xffff, v6
-; GFX7-NEXT: v_not_b32_e32 v8, v2
; GFX7-NEXT: .LBB27_1: ; %atomicrmw.start
; GFX7-NEXT: ; =>This Inner Loop Header: Depth=1
; GFX7-NEXT: s_waitcnt vmcnt(0)
-; GFX7-NEXT: v_lshrrev_b32_e32 v2, v6, v3
-; GFX7-NEXT: v_cvt_f32_f16_e32 v2, v2
-; GFX7-NEXT: v_and_b32_e32 v4, v3, v8
-; GFX7-NEXT: v_max_f32_e32 v2, v2, v7
-; GFX7-NEXT: v_cvt_f16_f32_e32 v2, v2
-; GFX7-NEXT: v_lshlrev_b32_e32 v2, v6, v2
-; GFX7-NEXT: v_or_b32_e32 v2, v4, v2
-; GFX7-NEXT: v_mov_b32_e32 v5, v3
-; GFX7-NEXT: v_mov_b32_e32 v4, v2
-; GFX7-NEXT: buffer_atomic_cmpswap v[4:5], v[0:1], s[4:7], 0 addr64 glc
+; GFX7-NEXT: v_mov_b32_e32 v8, v3
+; GFX7-NEXT: v_lshrrev_b32_e32 v3, v2, v8
+; GFX7-NEXT: v_cvt_f32_f16_e32 v3, v3
+; GFX7-NEXT: v_and_b32_e32 v4, v8, v6
+; GFX7-NEXT: v_max_f32_e32 v3, v3, v5
+; GFX7-NEXT: v_cvt_f16_f32_e32 v3, v3
+; GFX7-NEXT: v_lshlrev_b32_e32 v3, v2, v3
+; GFX7-NEXT: v_or_b32_e32 v7, v4, v3
+; GFX7-NEXT: v_mov_b32_e32 v3, v7
+; GFX7-NEXT: v_mov_b32_e32 v4, v8
+; GFX7-NEXT: buffer_atomic_cmpswap v[3:4], v[0:1], s[4:7], 0 addr64 glc
; GFX7-NEXT: s_waitcnt vmcnt(0)
; GFX7-NEXT: buffer_wbinvl1
-; GFX7-NEXT: v_cmp_eq_u32_e32 vcc, v4, v3
+; GFX7-NEXT: v_cmp_eq_u32_e32 vcc, v3, v8
; GFX7-NEXT: s_or_b64 s[8:9], vcc, s[8:9]
-; GFX7-NEXT: v_mov_b32_e32 v3, v4
; GFX7-NEXT: s_andn2_b64 exec, exec, s[8:9]
; GFX7-NEXT: s_cbranch_execnz .LBB27_1
; GFX7-NEXT: ; %bb.2: ; %atomicrmw.end
; GFX7-NEXT: s_or_b64 exec, exec, s[8:9]
-; GFX7-NEXT: v_lshrrev_b32_e32 v0, v6, v4
+; GFX7-NEXT: v_lshrrev_b32_e32 v0, v2, v3
; GFX7-NEXT: v_cvt_f32_f16_e32 v0, v0
; GFX7-NEXT: s_setpc_b64 s[30:31]
;
@@ -5307,37 +5308,37 @@ define half @global_agent_atomic_fmax_ret_f16__offset12b_pos__amdgpu_no_fine_gra
; GFX6-NEXT: s_mov_b32 s4, s6
; GFX6-NEXT: s_mov_b32 s5, s6
; GFX6-NEXT: buffer_load_dword v3, v[0:1], s[4:7], 0 addr64
-; GFX6-NEXT: v_cvt_f16_f32_e32 v2, v2
-; GFX6-NEXT: v_and_b32_e32 v4, 3, v4
-; GFX6-NEXT: v_lshlrev_b32_e32 v6, 3, v4
+; GFX6-NEXT: v_cvt_f16_f32_e32 v5, v2
+; GFX6-NEXT: v_and_b32_e32 v2, 3, v4
+; GFX6-NEXT: v_lshlrev_b32_e32 v2, 3, v2
+; GFX6-NEXT: v_lshl_b32_e32 v4, 0xffff, v2
+; GFX6-NEXT: v_cvt_f32_f16_e32 v5, v5
+; GFX6-NEXT: v_not_b32_e32 v6, v4
; GFX6-NEXT: s_mov_b64 s[8:9], 0
-; GFX6-NEXT: v_cvt_f32_f16_e32 v7, v2
-; GFX6-NEXT: v_lshl_b32_e32 v2, 0xffff, v6
-; GFX6-NEXT: v_not_b32_e32 v8, v2
; GFX6-NEXT: .LBB27_1: ; %atomicrmw.start
; GFX6-NEXT: ; =>This Inner Loop Header: Depth=1
; GFX6-NEXT: s_waitcnt vmcnt(0)
-; GFX6-NEXT: v_lshrrev_b32_e32 v2, v6, v3
-; GFX6-NEXT: v_cvt_f32_f16_e32 v2, v2
+; GFX6-NEXT: v_mov_b32_e32 v8, v3
; GFX6-NEXT: s_waitcnt expcnt(0)
-; GFX6-NEXT: v_and_b32_e32 v4, v3, v8
-; GFX6-NEXT: v_max_f32_e32 v2, v2, v7
-; GFX6-NEXT: v_cvt_f16_f32_e32 v2, v2
-; GFX6-NEXT: v_lshlrev_b32_e32 v2, v6, v2
-; GFX6-NEXT: v_or_b32_e32 v2, v4, v2
-; GFX6-NEXT: v_mov_b32_e32 v5, v3
-; GFX6-NEXT: v_mov_b32_e32 v4, v2
-; GFX6-NEXT: buffer_atomic_cmpswap v[4:5], v[0:1], s[4:7], 0 addr64 glc
+; GFX6-NEXT: v_lshrrev_b32_e32 v3, v2, v8
+; GFX6-NEXT: v_cvt_f32_f16_e32 v3, v3
+; GFX6-NEXT: v_and_b32_e32 v4, v8, v6
+; GFX6-NEXT: v_max_f32_e32 v3, v3, v5
+; GFX6-NEXT: v_cvt_f16_f32_e32 v3, v3
+; GFX6-NEXT: v_lshlrev_b32_e32 v3, v2, v3
+; GFX6-NEXT: v_or_b32_e32 v7, v4, v3
+; GFX6-NEXT: v_mov_b32_e32 v3, v7
+; GFX6-NEXT: v_mov_b32_e32 v4, v8
+; GFX6-NEXT: buffer_atomic_cmpswap v[3:4], v[0:1], s[4:7], 0 addr64 glc
; GFX6-NEXT: s_waitcnt vmcnt(0)
; GFX6-NEXT: buffer_wbinvl1
-; GFX6-NEXT: v_cmp_eq_u32_e32 vcc, v4, v3
+; GFX6-NEXT: v_cmp_eq_u32_e32 vcc, v3, v8
; GFX6-NEXT: s_or_b64 s[8:9], vcc, s[8:9]
-; GFX6-NEXT: v_mov_b32_e32 v3, v4
; GFX6-NEXT: s_andn2_b64 exec, exec, s[8:9]
; GFX6-NEXT: s_cbranch_execnz .LBB27_1
; GFX6-NEXT: ; %bb.2: ; %atomicrmw.end
; GFX6-NEXT: s_or_b64 exec, exec, s[8:9]
-; GFX6-NEXT: v_lshrrev_b32_e32 v0, v6, v4
+; GFX6-NEXT: v_lshrrev_b32_e32 v0, v2, v3
; GFX6-NEXT: v_cvt_f32_f16_e32 v0, v0
; GFX6-NEXT: s_waitcnt expcnt(0)
; GFX6-NEXT: s_setpc_b64 s[30:31]
@@ -5724,36 +5725,36 @@ define half @global_agent_atomic_fmax_ret_f16__offset12b_neg__amdgpu_no_fine_gra
; GFX7-NEXT: s_mov_b32 s4, s6
; GFX7-NEXT: s_mov_b32 s5, s6
; GFX7-NEXT: buffer_load_dword v3, v[0:1], s[4:7], 0 addr64
-; GFX7-NEXT: v_cvt_f16_f32_e32 v2, v2
-; GFX7-NEXT: v_and_b32_e32 v4, 3, v4
-; GFX7-NEXT: v_lshlrev_b32_e32 v6, 3, v4
+; GFX7-NEXT: v_cvt_f16_f32_e32 v5, v2
+; GFX7-NEXT: v_and_b32_e32 v2, 3, v4
+; GFX7-NEXT: v_lshlrev_b32_e32 v2, 3, v2
+; GFX7-NEXT: v_lshl_b32_e32 v4, 0xffff, v2
+; GFX7-NEXT: v_cvt_f32_f16_e32 v5, v5
+; GFX7-NEXT: v_not_b32_e32 v6, v4
; GFX7-NEXT: s_mov_b64 s[8:9], 0
-; GFX7-NEXT: v_cvt_f32_f16_e32 v7, v2
-; GFX7-NEXT: v_lshl_b32_e32 v2, 0xffff, v6
-; GFX7-NEXT: v_not_b32_e32 v8, v2
; GFX7-NEXT: .LBB28_1: ; %atomicrmw.start
; GFX7-NEXT: ; =>This Inner Loop Header: Depth=1
; GFX7-NEXT: s_waitcnt vmcnt(0)
-; GFX7-NEXT: v_lshrrev_b32_e32 v2, v6, v3
-; GFX7-NEXT: v_cvt_f32_f16_e32 v2, v2
-; GFX7-NEXT: v_and_b32_e32 v4, v3, v8
-; GFX7-NEXT: v_max_f32_e32 v2, v2, v7
-; GFX7-NEXT: v_cvt_f16_f32_e32 v2, v2
-; GFX7-NEXT: v_lshlrev_b32_e32 v2, v6, v2
-; GFX7-NEXT: v_or_b32_e32 v2, v4, v2
-; GFX7-NEXT: v_mov_b32_e32 v5, v3
-; GFX7-NEXT: v_mov_b32_e32 v4, v2
-; GFX7-NEXT: buffer_atomic_cmpswap v[4:5], v[0:1], s[4:7], 0 addr64 glc
+; GFX7-NEXT: v_mov_b32_e32 v8, v3
+; GFX7-NEXT: v_lshrrev_b32_e32 v3, v2, v8
+; GFX7-NEXT: v_cvt_f32_f16_e32 v3, v3
+; GFX7-NEXT: v_and_b32_e32 v4, v8, v6
+; GFX7-NEXT: v_max_f32_e32 v3, v3, v5
+; GFX7-NEXT: v_cvt_f16_f32_e32 v3, v3
+; GFX7-NEXT: v_lshlrev_b32_e32 v3, v2, v3
+; GFX7-NEXT: v_or_b32_e32 v7, v4, v3
+; GFX7-NEXT: v_mov_b32_e32 v3, v7
+; GFX7-NEXT: v_mov_b32_e32 v4, v8
+; GFX7-NEXT: buffer_atomic_cmpswap v[3:4], v[0:1], s[4:7], 0 addr64 glc
; GFX7-NEXT: s_waitcnt vmcnt(0)
; GFX7-NEXT: buffer_wbinvl1
-; GFX7-NEXT: v_cmp_eq_u32_e32 vcc, v4, v3
+; GFX7-NEXT: v_cmp_eq_u32_e32 vcc, v3, v8
; GFX7-NEXT: s_or_b64 s[8:9], vcc, s[8:9]
-; GFX7-NEXT: v_mov_b32_e32 v3, v4
; GFX7-NEXT: s_andn2_b64 exec, exec, s[8:9]
; GFX7-NEXT: s_cbranch_execnz .LBB28_1
; GFX7-NEXT: ; %bb.2: ; %atomicrmw.end
; GFX7-NEXT: s_or_b64 exec, exec, s[8:9]
-; GFX7-NEXT: v_lshrrev_b32_e32 v0, v6, v4
+; GFX7-NEXT: v_lshrrev_b32_e32 v0, v2, v3
; GFX7-NEXT: v_cvt_f32_f16_e32 v0, v0
; GFX7-NEXT: s_setpc_b64 s[30:31]
;
@@ -5768,37 +5769,37 @@ define half @global_agent_atomic_fmax_ret_f16__offset12b_neg__amdgpu_no_fine_gra
; GFX6-NEXT: s_mov_b32 s4, s6
; GFX6-NEXT: s_mov_b32 s5, s6
; GFX6-NEXT: buffer_load_dword v3, v[0:1], s[4:7], 0 addr64
-; GFX6-NEXT: v_cvt_f16_f32_e32 v2, v2
-; GFX6-NEXT: v_and_b32_e32 v4, 3, v4
-; GFX6-NEXT: v_lshlrev_b32_e32 v6, 3, v4
+; GFX6-NEXT: v_cvt_f16_f32_e32 v5, v2
+; GFX6-NEXT: v_and_b32_e32 v2, 3, v4
+; GFX6-NEXT: v_lshlrev_b32_e32 v2, 3, v2
+; GFX6-NEXT: v_lshl_b32_e32 v4, 0xffff, v2
+; GFX6-NEXT: v_cvt_f32_f16_e32 v5, v5
+; GFX6-NEXT: v_not_b32_e32 v6, v4
; GFX6-NEXT: s_mov_b64 s[8:9], 0
-; GFX6-NEXT: v_cvt_f32_f16_e32 v7, v2
-; GFX6-NEXT: v_lshl_b32_e32 v2, 0xffff, v6
-; GFX6-NEXT: v_not_b32_e32 v8, v2
; GFX6-NEXT: .LBB28_1: ; %atomicrmw.start
; GFX6-NEXT: ; =>This Inner Loop Header: Depth=1
; GFX6-NEXT: s_waitcnt vmcnt(0)
-; GFX6-NEXT: v_lshrrev_b32_e32 v2, v6, v3
-; GFX6-NEXT: v_cvt_f32_f16_e32 v2, v2
+; GFX6-NEXT: v_mov_b32_e32 v8, v3
; GFX6-NEXT: s_waitcnt expcnt(0)
-; GFX6-NEXT: v_and_b32_e32 v4, v3, v8
-; GFX6-NEXT: v_max_f32_e32 v2, v2, v7
-; GFX6-NEXT: v_cvt_f16_f32_e32 v2, v2
-; GFX6-NEXT: v_lshlrev_b32_e32 v2, v6, v2
-; GFX6-NEXT: v_or_b32_e32 v2, v4, v2
-; GFX6-NEXT: v_mov_b32_e32 v5, v3
-; GFX6-NEXT: v_mov_b32_e32 v4, v2
-; GFX6-NEXT: buffer_atomic_cmpswap v[4:5], v[0:1], s[4:7], 0 addr64 glc
+; GFX6-NEXT: v_lshrrev_b32_e32 v3, v2, v8
+; GFX6-NEXT: v_cvt_f32_f16_e32 v3, v3
+; GFX6-NEXT: v_and_b32_e32 v4, v8, v6
+; GFX6-NEXT: v_max_f32_e32 v3, v3, v5
+; GFX6-NEXT: v_cvt_f16_f32_e32 v3, v3
+; GFX6-NEXT: v_lshlrev_b32_e32 v3, v2, v3
+; GFX6-NEXT: v_or_b32_e32 v7, v4, v3
+; GFX6-NEXT: v_mov_b32_e32 v3, v7
+; GFX6-NEXT: v_mov_b32_e32 v4, v8
+; GFX6-NEXT: buffer_atomic_cmpswap v[3:4], v[0:1], s[4:7], 0 addr64 glc
; GFX6-NEXT: s_waitcnt vmcnt(0)
; GFX6-NEXT: buffer_wbinvl1
-; GFX6-NEXT: v_cmp_eq_u32_e32 vcc, v4, v3
+; GFX6-NEXT: v_cmp_eq_u32_e32 vcc, v3, v8
; GFX6-NEXT: s_or_b64 s[8:9], vcc, s[8:9]
-; GFX6-NEXT: v_mov_b32_e32 v3, v4
; GFX6-NEXT: s_andn2_b64 exec, exec, s[8:9]
; GFX6-NEXT: s_cbranch_execnz .LBB28_1
; GFX6-NEXT: ; %bb.2: ; %atomicrmw.end
; GFX6-NEXT: s_or_b64 exec, exec, s[8:9]
-; GFX6-NEXT: v_lshrrev_b32_e32 v0, v6, v4
+; GFX6-NEXT: v_lshrrev_b32_e32 v0, v2, v3
; GFX6-NEXT: v_cvt_f32_f16_e32 v0, v0
; GFX6-NEXT: s_waitcnt expcnt(0)
; GFX6-NEXT: s_setpc_b64 s[30:31]
@@ -6163,28 +6164,28 @@ define void @global_agent_atomic_fmax_noret_f16__amdgpu_no_fine_grained_memory(p
; GFX7-NEXT: v_cvt_f16_f32_e32 v5, v2
; GFX7-NEXT: v_and_b32_e32 v2, 3, v3
; GFX7-NEXT: v_lshlrev_b32_e32 v2, 3, v2
-; GFX7-NEXT: v_lshl_b32_e32 v3, 0xffff, v2
-; GFX7-NEXT: v_cvt_f32_f16_e32 v5, v5
-; GFX7-NEXT: v_not_b32_e32 v6, v3
; GFX7-NEXT: s_mov_b64 s[8:9], 0
+; GFX7-NEXT: v_cvt_f32_f16_e32 v3, v5
+; GFX7-NEXT: v_lshl_b32_e32 v5, 0xffff, v2
+; GFX7-NEXT: v_not_b32_e32 v6, v5
; GFX7-NEXT: .LBB29_1: ; %atomicrmw.start
; GFX7-NEXT: ; =>This Inner Loop Header: Depth=1
; GFX7-NEXT: s_waitcnt vmcnt(0)
-; GFX7-NEXT: v_lshrrev_b32_e32 v3, v2, v4
-; GFX7-NEXT: v_cvt_f32_f16_e32 v3, v3
-; GFX7-NEXT: v_and_b32_e32 v7, v4, v6
-; GFX7-NEXT: v_max_f32_e32 v3, v3, v5
-; GFX7-NEXT: v_cvt_f16_f32_e32 v3, v3
-; GFX7-NEXT: v_lshlrev_b32_e32 v3, v2, v3
-; GFX7-NEXT: v_or_b32_e32 v3, v7, v3
; GFX7-NEXT: v_mov_b32_e32 v8, v4
-; GFX7-NEXT: v_mov_b32_e32 v7, v3
-; GFX7-NEXT: buffer_atomic_cmpswap v[7:8], v[0:1], s[4:7], 0 addr64 glc
+; GFX7-NEXT: v_lshrrev_b32_e32 v4, v2, v8
+; GFX7-NEXT: v_cvt_f32_f16_e32 v4, v4
+; GFX7-NEXT: v_and_b32_e32 v5, v8, v6
+; GFX7-NEXT: v_max_f32_e32 v4, v4, v3
+; GFX7-NEXT: v_cvt_f16_f32_e32 v4, v4
+; GFX7-NEXT: v_lshlrev_b32_e32 v4, v2, v4
+; GFX7-NEXT: v_or_b32_e32 v7, v5, v4
+; GFX7-NEXT: v_mov_b32_e32 v4, v7
+; GFX7-NEXT: v_mov_b32_e32 v5, v8
+; GFX7-NEXT: buffer_atomic_cmpswap v[4:5], v[0:1], s[4:7], 0 addr64 glc
; GFX7-NEXT: s_waitcnt vmcnt(0)
; GFX7-NEXT: buffer_wbinvl1
-; GFX7-NEXT: v_cmp_eq_u32_e32 vcc, v7, v4
+; GFX7-NEXT: v_cmp_eq_u32_e32 vcc, v4, v8
; GFX7-NEXT: s_or_b64 s[8:9], vcc, s[8:9]
-; GFX7-NEXT: v_mov_b32_e32 v4, v7
; GFX7-NEXT: s_andn2_b64 exec, exec, s[8:9]
; GFX7-NEXT: s_cbranch_execnz .LBB29_1
; GFX7-NEXT: ; %bb.2: ; %atomicrmw.end
@@ -6204,29 +6205,29 @@ define void @global_agent_atomic_fmax_noret_f16__amdgpu_no_fine_grained_memory(p
; GFX6-NEXT: v_cvt_f16_f32_e32 v5, v2
; GFX6-NEXT: v_and_b32_e32 v2, 3, v3
; GFX6-NEXT: v_lshlrev_b32_e32 v2, 3, v2
-; GFX6-NEXT: v_lshl_b32_e32 v3, 0xffff, v2
-; GFX6-NEXT: v_cvt_f32_f16_e32 v5, v5
-; GFX6-NEXT: v_not_b32_e32 v6, v3
; GFX6-NEXT: s_mov_b64 s[8:9], 0
+; GFX6-NEXT: v_cvt_f32_f16_e32 v3, v5
+; GFX6-NEXT: v_lshl_b32_e32 v5, 0xffff, v2
+; GFX6-NEXT: v_not_b32_e32 v6, v5
; GFX6-NEXT: .LBB29_1: ; %atomicrmw.start
; GFX6-NEXT: ; =>This Inner Loop Header: Depth=1
; GFX6-NEXT: s_waitcnt vmcnt(0)
-; GFX6-NEXT: v_lshrrev_b32_e32 v3, v2, v4
-; GFX6-NEXT: v_cvt_f32_f16_e32 v3, v3
-; GFX6-NEXT: s_waitcnt expcnt(0)
-; GFX6-NEXT: v_and_b32_e32 v7, v4, v6
-; GFX6-NEXT: v_max_f32_e32 v3, v3, v5
-; GFX6-NEXT: v_cvt_f16_f32_e32 v3, v3
-; GFX6-NEXT: v_lshlrev_b32_e32 v3, v2, v3
-; GFX6-NEXT: v_or_b32_e32 v3, v7, v3
; GFX6-NEXT: v_mov_b32_e32 v8, v4
-; GFX6-NEXT: v_mov_b32_e32 v7, v3
-; GFX6-NEXT: buffer_atomic_cmpswap v[7:8], v[0:1], s[4:7], 0 addr64 glc
+; GFX6-NEXT: s_waitcnt expcnt(0)
+; GFX6-NEXT: v_lshrrev_b32_e32 v4, v2, v8
+; GFX6-NEXT: v_cvt_f32_f16_e32 v4, v4
+; GFX6-NEXT: v_and_b32_e32 v5, v8, v6
+; GFX6-NEXT: v_max_f32_e32 v4, v4, v3
+; GFX6-NEXT: v_cvt_f16_f32_e32 v4, v4
+; GFX6-NEXT: v_lshlrev_b32_e32 v4, v2, v4
+; GFX6-NEXT: v_or_b32_e32 v7, v5, v4
+; GFX6-NEXT: v_mov_b32_e32 v4, v7
+; GFX6-NEXT: v_mov_b32_e32 v5, v8
+; GFX6-NEXT: buffer_atomic_cmpswap v[4:5], v[0:1], s[4:7], 0 addr64 glc
; GFX6-NEXT: s_waitcnt vmcnt(0)
; GFX6-NEXT: buffer_wbinvl1
-; GFX6-NEXT: v_cmp_eq_u32_e32 vcc, v7, v4
+; GFX6-NEXT: v_cmp_eq_u32_e32 vcc, v4, v8
; GFX6-NEXT: s_or_b64 s[8:9], vcc, s[8:9]
-; GFX6-NEXT: v_mov_b32_e32 v4, v7
; GFX6-NEXT: s_andn2_b64 exec, exec, s[8:9]
; GFX6-NEXT: s_cbranch_execnz .LBB29_1
; GFX6-NEXT: ; %bb.2: ; %atomicrmw.end
@@ -6605,31 +6606,31 @@ define void @global_agent_atomic_fmax_noret_f16__offset12b_pos__amdgpu_no_fine_g
; GFX7-NEXT: s_mov_b32 s4, s6
; GFX7-NEXT: s_mov_b32 s5, s6
; GFX7-NEXT: buffer_load_dword v3, v[0:1], s[4:7], 0 addr64
-; GFX7-NEXT: v_cvt_f16_f32_e32 v2, v2
-; GFX7-NEXT: v_and_b32_e32 v4, 3, v4
-; GFX7-NEXT: v_lshlrev_b32_e32 v4, 3, v4
+; GFX7-NEXT: v_cvt_f16_f32_e32 v5, v2
+; GFX7-NEXT: v_and_b32_e32 v2, 3, v4
+; GFX7-NEXT: v_lshlrev_b32_e32 v2, 3, v2
+; GFX7-NEXT: v_lshl_b32_e32 v4, 0xffff, v2
+; GFX7-NEXT: v_cvt_f32_f16_e32 v5, v5
+; GFX7-NEXT: v_not_b32_e32 v6, v4
; GFX7-NEXT: s_mov_b64 s[8:9], 0
-; GFX7-NEXT: v_cvt_f32_f16_e32 v5, v2
-; GFX7-NEXT: v_lshl_b32_e32 v2, 0xffff, v4
-; GFX7-NEXT: v_not_b32_e32 v6, v2
; GFX7-NEXT: .LBB30_1: ; %atomicrmw.start
; GFX7-NEXT: ; =>This Inner Loop Header: Depth=1
; GFX7-NEXT: s_waitcnt vmcnt(0)
-; GFX7-NEXT: v_lshrrev_b32_e32 v2, v4, v3
-; GFX7-NEXT: v_cvt_f32_f16_e32 v2, v2
-; GFX7-NEXT: v_and_b32_e32 v7, v3, v6
-; GFX7-NEXT: v_max_f32_e32 v2, v2, v5
-; GFX7-NEXT: v_cvt_f16_f32_e32 v2, v2
-; GFX7-NEXT: v_lshlrev_b32_e32 v2, v4, v2
-; GFX7-NEXT: v_or_b32_e32 v2, v7, v2
; GFX7-NEXT: v_mov_b32_e32 v8, v3
-; GFX7-NEXT: v_mov_b32_e32 v7, v2
-; GFX7-NEXT: buffer_atomic_cmpswap v[7:8], v[0:1], s[4:7], 0 addr64 glc
+; GFX7-NEXT: v_lshrrev_b32_e32 v3, v2, v8
+; GFX7-NEXT: v_cvt_f32_f16_e32 v3, v3
+; GFX7-NEXT: v_and_b32_e32 v4, v8, v6
+; GFX7-NEXT: v_max_f32_e32 v3, v3, v5
+; GFX7-NEXT: v_cvt_f16_f32_e32 v3, v3
+; GFX7-NEXT: v_lshlrev_b32_e32 v3, v2, v3
+; GFX7-NEXT: v_or_b32_e32 v7, v4, v3
+; GFX7-NEXT: v_mov_b32_e32 v3, v7
+; GFX7-NEXT: v_mov_b32_e32 v4, v8
+; GFX7-NEXT: buffer_atomic_cmpswap v[3:4], v[0:1], s[4:7], 0 addr64 glc
; GFX7-NEXT: s_waitcnt vmcnt(0)
; GFX7-NEXT: buffer_wbinvl1
-; GFX7-NEXT: v_cmp_eq_u32_e32 vcc, v7, v3
+; GFX7-NEXT: v_cmp_eq_u32_e32 vcc, v3, v8
; GFX7-NEXT: s_or_b64 s[8:9], vcc, s[8:9]
-; GFX7-NEXT: v_mov_b32_e32 v3, v7
; GFX7-NEXT: s_andn2_b64 exec, exec, s[8:9]
; GFX7-NEXT: s_cbranch_execnz .LBB30_1
; GFX7-NEXT: ; %bb.2: ; %atomicrmw.end
@@ -6647,32 +6648,32 @@ define void @global_agent_atomic_fmax_noret_f16__offset12b_pos__amdgpu_no_fine_g
; GFX6-NEXT: s_mov_b32 s4, s6
; GFX6-NEXT: s_mov_b32 s5, s6
; GFX6-NEXT: buffer_load_dword v3, v[0:1], s[4:7], 0 addr64
-; GFX6-NEXT: v_cvt_f16_f32_e32 v2, v2
-; GFX6-NEXT: v_and_b32_e32 v4, 3, v4
-; GFX6-NEXT: v_lshlrev_b32_e32 v4, 3, v4
+; GFX6-NEXT: v_cvt_f16_f32_e32 v5, v2
+; GFX6-NEXT: v_and_b32_e32 v2, 3, v4
+; GFX6-NEXT: v_lshlrev_b32_e32 v2, 3, v2
+; GFX6-NEXT: v_lshl_b32_e32 v4, 0xffff, v2
+; GFX6-NEXT: v_cvt_f32_f16_e32 v5, v5
+; GFX6-NEXT: v_not_b32_e32 v6, v4
; GFX6-NEXT: s_mov_b64 s[8:9], 0
-; GFX6-NEXT: v_cvt_f32_f16_e32 v5, v2
-; GFX6-NEXT: v_lshl_b32_e32 v2, 0xffff, v4
-; GFX6-NEXT: v_not_b32_e32 v6, v2
; GFX6-NEXT: .LBB30_1: ; %atomicrmw.start
; GFX6-NEXT: ; =>This Inner Loop Header: Depth=1
; GFX6-NEXT: s_waitcnt vmcnt(0)
-; GFX6-NEXT: v_lshrrev_b32_e32 v2, v4, v3
-; GFX6-NEXT: v_cvt_f32_f16_e32 v2, v2
-; GFX6-NEXT: s_waitcnt expcnt(0)
-; GFX6-NEXT: v_and_b32_e32 v7, v3, v6
-; GFX6-NEXT: v_max_f32_e32 v2, v2, v5
-; GFX6-NEXT: v_cvt_f16_f32_e32 v2, v2
-; GFX6-NEXT: v_lshlrev_b32_e32 v2, v4, v2
-; GFX6-NEXT: v_or_b32_e32 v2, v7, v2
; GFX6-NEXT: v_mov_b32_e32 v8, v3
-; GFX6-NEXT: v_mov_b32_e32 v7, v2
-; GFX6-NEXT: buffer_atomic_cmpswap v[7:8], v[0:1], s[4:7], 0 addr64 glc
+; GFX6-NEXT: s_waitcnt expcnt(0)
+; GFX6-NEXT: v_lshrrev_b32_e32 v3, v2, v8
+; GFX6-NEXT: v_cvt_f32_f16_e32 v3, v3
+; GFX6-NEXT: v_and_b32_e32 v4, v8, v6
+; GFX6-NEXT: v_max_f32_e32 v3, v3, v5
+; GFX6-NEXT: v_cvt_f16_f32_e32 v3, v3
+; GFX6-NEXT: v_lshlrev_b32_e32 v3, v2, v3
+; GFX6-NEXT: v_or_b32_e32 v7, v4, v3
+; GFX6-NEXT: v_mov_b32_e32 v3, v7
+; GFX6-NEXT: v_mov_b32_e32 v4, v8
+; GFX6-NEXT: buffer_atomic_cmpswap v[3:4], v[0:1], s[4:7], 0 addr64 glc
; GFX6-NEXT: s_waitcnt vmcnt(0)
; GFX6-NEXT: buffer_wbinvl1
-; GFX6-NEXT: v_cmp_eq_u32_e32 vcc, v7, v3
+; GFX6-NEXT: v_cmp_eq_u32_e32 vcc, v3, v8
; GFX6-NEXT: s_or_b64 s[8:9], vcc, s[8:9]
-; GFX6-NEXT: v_mov_b32_e32 v3, v7
; GFX6-NEXT: s_andn2_b64 exec, exec, s[8:9]
; GFX6-NEXT: s_cbranch_execnz .LBB30_1
; GFX6-NEXT: ; %bb.2: ; %atomicrmw.end
@@ -7053,31 +7054,31 @@ define void @global_agent_atomic_fmax_noret_f16__offset12b_neg__amdgpu_no_fine_g
; GFX7-NEXT: s_mov_b32 s4, s6
; GFX7-NEXT: s_mov_b32 s5, s6
; GFX7-NEXT: buffer_load_dword v3, v[0:1], s[4:7], 0 addr64
-; GFX7-NEXT: v_cvt_f16_f32_e32 v2, v2
-; GFX7-NEXT: v_and_b32_e32 v4, 3, v4
-; GFX7-NEXT: v_lshlrev_b32_e32 v4, 3, v4
+; GFX7-NEXT: v_cvt_f16_f32_e32 v5, v2
+; GFX7-NEXT: v_and_b32_e32 v2, 3, v4
+; GFX7-NEXT: v_lshlrev_b32_e32 v2, 3, v2
+; GFX7-NEXT: v_lshl_b32_e32 v4, 0xffff, v2
+; GFX7-NEXT: v_cvt_f32_f16_e32 v5, v5
+; GFX7-NEXT: v_not_b32_e32 v6, v4
; GFX7-NEXT: s_mov_b64 s[8:9], 0
-; GFX7-NEXT: v_cvt_f32_f16_e32 v5, v2
-; GFX7-NEXT: v_lshl_b32_e32 v2, 0xffff, v4
-; GFX7-NEXT: v_not_b32_e32 v6, v2
; GFX7-NEXT: .LBB31_1: ; %atomicrmw.start
; GFX7-NEXT: ; =>This Inner Loop Header: Depth=1
; GFX7-NEXT: s_waitcnt vmcnt(0)
-; GFX7-NEXT: v_lshrrev_b32_e32 v2, v4, v3
-; GFX7-NEXT: v_cvt_f32_f16_e32 v2, v2
-; GFX7-NEXT: v_and_b32_e32 v7, v3, v6
-; GFX7-NEXT: v_max_f32_e32 v2, v2, v5
-; GFX7-NEXT: v_cvt_f16_f32_e32 v2, v2
-; GFX7-NEXT: v_lshlrev_b32_e32 v2, v4, v2
-; GFX7-NEXT: v_or_b32_e32 v2, v7, v2
; GFX7-NEXT: v_mov_b32_e32 v8, v3
-; GFX7-NEXT: v_mov_b32_e32 v7, v2
-; GFX7-NEXT: buffer_atomic_cmpswap v[7:8], v[0:1], s[4:7], 0 addr64 glc
+; GFX7-NEXT: v_lshrrev_b32_e32 v3, v2, v8
+; GFX7-NEXT: v_cvt_f32_f16_e32 v3, v3
+; GFX7-NEXT: v_and_b32_e32 v4, v8, v6
+; GFX7-NEXT: v_max_f32_e32 v3, v3, v5
+; GFX7-NEXT: v_cvt_f16_f32_e32 v3, v3
+; GFX7-NEXT: v_lshlrev_b32_e32 v3, v2, v3
+; GFX7-NEXT: v_or_b32_e32 v7, v4, v3
+; GFX7-NEXT: v_mov_b32_e32 v3, v7
+; GFX7-NEXT: v_mov_b32_e32 v4, v8
+; GFX7-NEXT: buffer_atomic_cmpswap v[3:4], v[0:1], s[4:7], 0 addr64 glc
; GFX7-NEXT: s_waitcnt vmcnt(0)
; GFX7-NEXT: buffer_wbinvl1
-; GFX7-NEXT: v_cmp_eq_u32_e32 vcc, v7, v3
+; GFX7-NEXT: v_cmp_eq_u32_e32 vcc, v3, v8
; GFX7-NEXT: s_or_b64 s[8:9], vcc, s[8:9]
-; GFX7-NEXT: v_mov_b32_e32 v3, v7
; GFX7-NEXT: s_andn2_b64 exec, exec, s[8:9]
; GFX7-NEXT: s_cbranch_execnz .LBB31_1
; GFX7-NEXT: ; %bb.2: ; %atomicrmw.end
@@ -7095,32 +7096,32 @@ define void @global_agent_atomic_fmax_noret_f16__offset12b_neg__amdgpu_no_fine_g
; GFX6-NEXT: s_mov_b32 s4, s6
; GFX6-NEXT: s_mov_b32 s5, s6
; GFX6-NEXT: buffer_load_dword v3, v[0:1], s[4:7], 0 addr64
-; GFX6-NEXT: v_cvt_f16_f32_e32 v2, v2
-; GFX6-NEXT: v_and_b32_e32 v4, 3, v4
-; GFX6-NEXT: v_lshlrev_b32_e32 v4, 3, v4
+; GFX6-NEXT: v_cvt_f16_f32_e32 v5, v2
+; GFX6-NEXT: v_and_b32_e32 v2, 3, v4
+; GFX6-NEXT: v_lshlrev_b32_e32 v2, 3, v2
+; GFX6-NEXT: v_lshl_b32_e32 v4, 0xffff, v2
+; GFX6-NEXT: v_cvt_f32_f16_e32 v5, v5
+; GFX6-NEXT: v_not_b32_e32 v6, v4
; GFX6-NEXT: s_mov_b64 s[8:9], 0
-; GFX6-NEXT: v_cvt_f32_f16_e32 v5, v2
-; GFX6-NEXT: v_lshl_b32_e32 v2, 0xffff, v4
-; GFX6-NEXT: v_not_b32_e32 v6, v2
; GFX6-NEXT: .LBB31_1: ; %atomicrmw.start
; GFX6-NEXT: ; =>This Inner Loop Header: Depth=1
; GFX6-NEXT: s_waitcnt vmcnt(0)
-; GFX6-NEXT: v_lshrrev_b32_e32 v2, v4, v3
-; GFX6-NEXT: v_cvt_f32_f16_e32 v2, v2
-; GFX6-NEXT: s_waitcnt expcnt(0)
-; GFX6-NEXT: v_and_b32_e32 v7, v3, v6
-; GFX6-NEXT: v_max_f32_e32 v2, v2, v5
-; GFX6-NEXT: v_cvt_f16_f32_e32 v2, v2
-; GFX6-NEXT: v_lshlrev_b32_e32 v2, v4, v2
-; GFX6-NEXT: v_or_b32_e32 v2, v7, v2
; GFX6-NEXT: v_mov_b32_e32 v8, v3
-; GFX6-NEXT: v_mov_b32_e32 v7, v2
-; GFX6-NEXT: buffer_atomic_cmpswap v[7:8], v[0:1], s[4:7], 0 addr64 glc
+; GFX6-NEXT: s_waitcnt expcnt(0)
+; GFX6-NEXT: v_lshrrev_b32_e32 v3, v2, v8
+; GFX6-NEXT: v_cvt_f32_f16_e32 v3, v3
+; GFX6-NEXT: v_and_b32_e32 v4, v8, v6
+; GFX6-NEXT: v_max_f32_e32 v3, v3, v5
+; GFX6-NEXT: v_cvt_f16_f32_e32 v3, v3
+; GFX6-NEXT: v_lshlrev_b32_e32 v3, v2, v3
+; GFX6-NEXT: v_or_b32_e32 v7, v4, v3
+; GFX6-NEXT: v_mov_b32_e32 v3, v7
+; GFX6-NEXT: v_mov_b32_e32 v4, v8
+; GFX6-NEXT: buffer_atomic_cmpswap v[3:4], v[0:1], s[4:7], 0 addr64 glc
; GFX6-NEXT: s_waitcnt vmcnt(0)
; GFX6-NEXT: buffer_wbinvl1
-; GFX6-NEXT: v_cmp_eq_u32_e32 vcc, v7, v3
+; GFX6-NEXT: v_cmp_eq_u32_e32 vcc, v3, v8
; GFX6-NEXT: s_or_b64 s[8:9], vcc, s[8:9]
-; GFX6-NEXT: v_mov_b32_e32 v3, v7
; GFX6-NEXT: s_andn2_b64 exec, exec, s[8:9]
; GFX6-NEXT: s_cbranch_execnz .LBB31_1
; GFX6-NEXT: ; %bb.2: ; %atomicrmw.end
@@ -7411,28 +7412,28 @@ define half @global_agent_atomic_fmax_ret_f16__offset12b_pos__align4__amdgpu_no_
; GFX7-NEXT: buffer_load_dword v3, v[0:1], s[4:7], 0 addr64 offset:2046
; GFX7-NEXT: v_cvt_f16_f32_e32 v2, v2
; GFX7-NEXT: s_mov_b64 s[8:9], 0
-; GFX7-NEXT: v_cvt_f32_f16_e32 v6, v2
+; GFX7-NEXT: v_cvt_f32_f16_e32 v2, v2
; GFX7-NEXT: .LBB32_1: ; %atomicrmw.start
; GFX7-NEXT: ; =>This Inner Loop Header: Depth=1
; GFX7-NEXT: s_waitcnt vmcnt(0)
-; GFX7-NEXT: v_cvt_f32_f16_e32 v2, v3
-; GFX7-NEXT: v_and_b32_e32 v4, 0xffff0000, v3
-; GFX7-NEXT: v_max_f32_e32 v2, v2, v6
-; GFX7-NEXT: v_cvt_f16_f32_e32 v2, v2
-; GFX7-NEXT: v_or_b32_e32 v2, v4, v2
; GFX7-NEXT: v_mov_b32_e32 v5, v3
-; GFX7-NEXT: v_mov_b32_e32 v4, v2
-; GFX7-NEXT: buffer_atomic_cmpswap v[4:5], v[0:1], s[4:7], 0 addr64 offset:2046 glc
+; GFX7-NEXT: v_cvt_f32_f16_e32 v3, v5
+; GFX7-NEXT: v_and_b32_e32 v4, 0xffff0000, v5
+; GFX7-NEXT: v_max_f32_e32 v3, v3, v2
+; GFX7-NEXT: v_cvt_f16_f32_e32 v3, v3
+; GFX7-NEXT: v_or_b32_e32 v4, v4, v3
+; GFX7-NEXT: v_mov_b32_e32 v3, v4
+; GFX7-NEXT: v_mov_b32_e32 v4, v5
+; GFX7-NEXT: buffer_atomic_cmpswap v[3:4], v[0:1], s[4:7], 0 addr64 offset:2046 glc
; GFX7-NEXT: s_waitcnt vmcnt(0)
; GFX7-NEXT: buffer_wbinvl1
-; GFX7-NEXT: v_cmp_eq_u32_e32 vcc, v4, v3
+; GFX7-NEXT: v_cmp_eq_u32_e32 vcc, v3, v5
; GFX7-NEXT: s_or_b64 s[8:9], vcc, s[8:9]
-; GFX7-NEXT: v_mov_b32_e32 v3, v4
; GFX7-NEXT: s_andn2_b64 exec, exec, s[8:9]
; GFX7-NEXT: s_cbranch_execnz .LBB32_1
; GFX7-NEXT: ; %bb.2: ; %atomicrmw.end
; GFX7-NEXT: s_or_b64 exec, exec, s[8:9]
-; GFX7-NEXT: v_cvt_f32_f16_e32 v0, v4
+; GFX7-NEXT: v_cvt_f32_f16_e32 v0, v3
; GFX7-NEXT: s_setpc_b64 s[30:31]
;
; GFX6-LABEL: global_agent_atomic_fmax_ret_f16__offset12b_pos__align4__amdgpu_no_fine_grained_memory:
@@ -7445,29 +7446,29 @@ define half @global_agent_atomic_fmax_ret_f16__offset12b_pos__align4__amdgpu_no_
; GFX6-NEXT: buffer_load_dword v3, v[0:1], s[4:7], 0 addr64 offset:2046
; GFX6-NEXT: v_cvt_f16_f32_e32 v2, v2
; GFX6-NEXT: s_mov_b64 s[8:9], 0
-; GFX6-NEXT: v_cvt_f32_f16_e32 v6, v2
+; GFX6-NEXT: v_cvt_f32_f16_e32 v2, v2
; GFX6-NEXT: .LBB32_1: ; %atomicrmw.start
; GFX6-NEXT: ; =>This Inner Loop Header: Depth=1
; GFX6-NEXT: s_waitcnt vmcnt(0)
-; GFX6-NEXT: v_cvt_f32_f16_e32 v2, v3
-; GFX6-NEXT: s_waitcnt expcnt(0)
-; GFX6-NEXT: v_and_b32_e32 v4, 0xffff0000, v3
-; GFX6-NEXT: v_max_f32_e32 v2, v2, v6
-; GFX6-NEXT: v_cvt_f16_f32_e32 v2, v2
-; GFX6-NEXT: v_or_b32_e32 v2, v4, v2
; GFX6-NEXT: v_mov_b32_e32 v5, v3
-; GFX6-NEXT: v_mov_b32_e32 v4, v2
-; GFX6-NEXT: buffer_atomic_cmpswap v[4:5], v[0:1], s[4:7], 0 addr64 offset:2046 glc
+; GFX6-NEXT: s_waitcnt expcnt(0)
+; GFX6-NEXT: v_cvt_f32_f16_e32 v3, v5
+; GFX6-NEXT: v_and_b32_e32 v4, 0xffff0000, v5
+; GFX6-NEXT: v_max_f32_e32 v3, v3, v2
+; GFX6-NEXT: v_cvt_f16_f32_e32 v3, v3
+; GFX6-NEXT: v_or_b32_e32 v4, v4, v3
+; GFX6-NEXT: v_mov_b32_e32 v3, v4
+; GFX6-NEXT: v_mov_b32_e32 v4, v5
+; GFX6-NEXT: buffer_atomic_cmpswap v[3:4], v[0:1], s[4:7], 0 addr64 offset:2046 glc
; GFX6-NEXT: s_waitcnt vmcnt(0)
; GFX6-NEXT: buffer_wbinvl1
-; GFX6-NEXT: v_cmp_eq_u32_e32 vcc, v4, v3
+; GFX6-NEXT: v_cmp_eq_u32_e32 vcc, v3, v5
; GFX6-NEXT: s_or_b64 s[8:9], vcc, s[8:9]
-; GFX6-NEXT: v_mov_b32_e32 v3, v4
; GFX6-NEXT: s_andn2_b64 exec, exec, s[8:9]
; GFX6-NEXT: s_cbranch_execnz .LBB32_1
; GFX6-NEXT: ; %bb.2: ; %atomicrmw.end
; GFX6-NEXT: s_or_b64 exec, exec, s[8:9]
-; GFX6-NEXT: v_cvt_f32_f16_e32 v0, v4
+; GFX6-NEXT: v_cvt_f32_f16_e32 v0, v3
; GFX6-NEXT: s_waitcnt expcnt(0)
; GFX6-NEXT: s_setpc_b64 s[30:31]
%gep = getelementptr half, ptr addrspace(1) %ptr, i64 1023
@@ -7746,23 +7747,23 @@ define void @global_agent_atomic_fmax_noret_f16__offset12b__align4_pos__amdgpu_n
; GFX7-NEXT: buffer_load_dword v3, v[0:1], s[4:7], 0 addr64 offset:2046
; GFX7-NEXT: v_cvt_f16_f32_e32 v2, v2
; GFX7-NEXT: s_mov_b64 s[8:9], 0
-; GFX7-NEXT: v_cvt_f32_f16_e32 v4, v2
+; GFX7-NEXT: v_cvt_f32_f16_e32 v2, v2
; GFX7-NEXT: .LBB33_1: ; %atomicrmw.start
; GFX7-NEXT: ; =>This Inner Loop Header: Depth=1
; GFX7-NEXT: s_waitcnt vmcnt(0)
-; GFX7-NEXT: v_cvt_f32_f16_e32 v2, v3
-; GFX7-NEXT: v_and_b32_e32 v5, 0xffff0000, v3
-; GFX7-NEXT: v_max_f32_e32 v2, v2, v4
-; GFX7-NEXT: v_cvt_f16_f32_e32 v2, v2
-; GFX7-NEXT: v_or_b32_e32 v2, v5, v2
-; GFX7-NEXT: v_mov_b32_e32 v6, v3
-; GFX7-NEXT: v_mov_b32_e32 v5, v2
-; GFX7-NEXT: buffer_atomic_cmpswap v[5:6], v[0:1], s[4:7], 0 addr64 offset:2046 glc
+; GFX7-NEXT: v_mov_b32_e32 v5, v3
+; GFX7-NEXT: v_cvt_f32_f16_e32 v3, v5
+; GFX7-NEXT: v_and_b32_e32 v4, 0xffff0000, v5
+; GFX7-NEXT: v_max_f32_e32 v3, v3, v2
+; GFX7-NEXT: v_cvt_f16_f32_e32 v3, v3
+; GFX7-NEXT: v_or_b32_e32 v4, v4, v3
+; GFX7-NEXT: v_mov_b32_e32 v3, v4
+; GFX7-NEXT: v_mov_b32_e32 v4, v5
+; GFX7-NEXT: buffer_atomic_cmpswap v[3:4], v[0:1], s[4:7], 0 addr64 offset:2046 glc
; GFX7-NEXT: s_waitcnt vmcnt(0)
; GFX7-NEXT: buffer_wbinvl1
-; GFX7-NEXT: v_cmp_eq_u32_e32 vcc, v5, v3
+; GFX7-NEXT: v_cmp_eq_u32_e32 vcc, v3, v5
; GFX7-NEXT: s_or_b64 s[8:9], vcc, s[8:9]
-; GFX7-NEXT: v_mov_b32_e32 v3, v5
; GFX7-NEXT: s_andn2_b64 exec, exec, s[8:9]
; GFX7-NEXT: s_cbranch_execnz .LBB33_1
; GFX7-NEXT: ; %bb.2: ; %atomicrmw.end
@@ -7779,24 +7780,24 @@ define void @global_agent_atomic_fmax_noret_f16__offset12b__align4_pos__amdgpu_n
; GFX6-NEXT: buffer_load_dword v3, v[0:1], s[4:7], 0 addr64 offset:2046
; GFX6-NEXT: v_cvt_f16_f32_e32 v2, v2
; GFX6-NEXT: s_mov_b64 s[8:9], 0
-; GFX6-NEXT: v_cvt_f32_f16_e32 v4, v2
+; GFX6-NEXT: v_cvt_f32_f16_e32 v2, v2
; GFX6-NEXT: .LBB33_1: ; %atomicrmw.start
; GFX6-NEXT: ; =>This Inner Loop Header: Depth=1
; GFX6-NEXT: s_waitcnt vmcnt(0)
-; GFX6-NEXT: v_cvt_f32_f16_e32 v2, v3
+; GFX6-NEXT: v_mov_b32_e32 v5, v3
; GFX6-NEXT: s_waitcnt expcnt(0)
-; GFX6-NEXT: v_and_b32_e32 v5, 0xffff0000, v3
-; GFX6-NEXT: v_max_f32_e32 v2, v2, v4
-; GFX6-NEXT: v_cvt_f16_f32_e32 v2, v2
-; GFX6-NEXT: v_or_b32_e32 v2, v5, v2
-; GFX6-NEXT: v_mov_b32_e32 v6, v3
-; GFX6-NEXT: v_mov_b32_e32 v5, v2
-; GFX6-NEXT: buffer_atomic_cmpswap v[5:6], v[0:1], s[4:7], 0 addr64 offset:2046 glc
+; GFX6-NEXT: v_cvt_f32_f16_e32 v3, v5
+; GFX6-NEXT: v_and_b32_e32 v4, 0xffff0000, v5
+; GFX6-NEXT: v_max_f32_e32 v3, v3, v2
+; GFX6-NEXT: v_cvt_f16_f32_e32 v3, v3
+; GFX6-NEXT: v_or_b32_e32 v4, v4, v3
+; GFX6-NEXT: v_mov_b32_e32 v3, v4
+; GFX6-NEXT: v_mov_b32_e32 v4, v5
+; GFX6-NEXT: buffer_atomic_cmpswap v[3:4], v[0:1], s[4:7], 0 addr64 offset:2046 glc
; GFX6-NEXT: s_waitcnt vmcnt(0)
; GFX6-NEXT: buffer_wbinvl1
-; GFX6-NEXT: v_cmp_eq_u32_e32 vcc, v5, v3
+; GFX6-NEXT: v_cmp_eq_u32_e32 vcc, v3, v5
; GFX6-NEXT: s_or_b64 s[8:9], vcc, s[8:9]
-; GFX6-NEXT: v_mov_b32_e32 v3, v5
; GFX6-NEXT: s_andn2_b64 exec, exec, s[8:9]
; GFX6-NEXT: s_cbranch_execnz .LBB33_1
; GFX6-NEXT: ; %bb.2: ; %atomicrmw.end
@@ -8189,36 +8190,36 @@ define half @global_system_atomic_fmax_ret_f16__offset12b_pos__amdgpu_no_fine_gr
; GFX7-NEXT: s_mov_b32 s4, s6
; GFX7-NEXT: s_mov_b32 s5, s6
; GFX7-NEXT: buffer_load_dword v3, v[0:1], s[4:7], 0 addr64
-; GFX7-NEXT: v_cvt_f16_f32_e32 v2, v2
-; GFX7-NEXT: v_and_b32_e32 v4, 3, v4
-; GFX7-NEXT: v_lshlrev_b32_e32 v6, 3, v4
+; GFX7-NEXT: v_cvt_f16_f32_e32 v5, v2
+; GFX7-NEXT: v_and_b32_e32 v2, 3, v4
+; GFX7-NEXT: v_lshlrev_b32_e32 v2, 3, v2
+; GFX7-NEXT: v_lshl_b32_e32 v4, 0xffff, v2
+; GFX7-NEXT: v_cvt_f32_f16_e32 v5, v5
+; GFX7-NEXT: v_not_b32_e32 v6, v4
; GFX7-NEXT: s_mov_b64 s[8:9], 0
-; GFX7-NEXT: v_cvt_f32_f16_e32 v7, v2
-; GFX7-NEXT: v_lshl_b32_e32 v2, 0xffff, v6
-; GFX7-NEXT: v_not_b32_e32 v8, v2
; GFX7-NEXT: .LBB34_1: ; %atomicrmw.start
; GFX7-NEXT: ; =>This Inner Loop Header: Depth=1
; GFX7-NEXT: s_waitcnt vmcnt(0)
-; GFX7-NEXT: v_lshrrev_b32_e32 v2, v6, v3
-; GFX7-NEXT: v_cvt_f32_f16_e32 v2, v2
-; GFX7-NEXT: v_and_b32_e32 v4, v3, v8
-; GFX7-NEXT: v_max_f32_e32 v2, v2, v7
-; GFX7-NEXT: v_cvt_f16_f32_e32 v2, v2
-; GFX7-NEXT: v_lshlrev_b32_e32 v2, v6, v2
-; GFX7-NEXT: v_or_b32_e32 v2, v4, v2
-; GFX7-NEXT: v_mov_b32_e32 v5, v3
-; GFX7-NEXT: v_mov_b32_e32 v4, v2
-; GFX7-NEXT: buffer_atomic_cmpswap v[4:5], v[0:1], s[4:7], 0 addr64 glc
+; GFX7-NEXT: v_mov_b32_e32 v8, v3
+; GFX7-NEXT: v_lshrrev_b32_e32 v3, v2, v8
+; GFX7-NEXT: v_cvt_f32_f16_e32 v3, v3
+; GFX7-NEXT: v_and_b32_e32 v4, v8, v6
+; GFX7-NEXT: v_max_f32_e32 v3, v3, v5
+; GFX7-NEXT: v_cvt_f16_f32_e32 v3, v3
+; GFX7-NEXT: v_lshlrev_b32_e32 v3, v2, v3
+; GFX7-NEXT: v_or_b32_e32 v7, v4, v3
+; GFX7-NEXT: v_mov_b32_e32 v3, v7
+; GFX7-NEXT: v_mov_b32_e32 v4, v8
+; GFX7-NEXT: buffer_atomic_cmpswap v[3:4], v[0:1], s[4:7], 0 addr64 glc
; GFX7-NEXT: s_waitcnt vmcnt(0)
; GFX7-NEXT: buffer_wbinvl1
-; GFX7-NEXT: v_cmp_eq_u32_e32 vcc, v4, v3
+; GFX7-NEXT: v_cmp_eq_u32_e32 vcc, v3, v8
; GFX7-NEXT: s_or_b64 s[8:9], vcc, s[8:9]
-; GFX7-NEXT: v_mov_b32_e32 v3, v4
; GFX7-NEXT: s_andn2_b64 exec, exec, s[8:9]
; GFX7-NEXT: s_cbranch_execnz .LBB34_1
; GFX7-NEXT: ; %bb.2: ; %atomicrmw.end
; GFX7-NEXT: s_or_b64 exec, exec, s[8:9]
-; GFX7-NEXT: v_lshrrev_b32_e32 v0, v6, v4
+; GFX7-NEXT: v_lshrrev_b32_e32 v0, v2, v3
; GFX7-NEXT: v_cvt_f32_f16_e32 v0, v0
; GFX7-NEXT: s_setpc_b64 s[30:31]
;
@@ -8233,37 +8234,37 @@ define half @global_system_atomic_fmax_ret_f16__offset12b_pos__amdgpu_no_fine_gr
; GFX6-NEXT: s_mov_b32 s4, s6
; GFX6-NEXT: s_mov_b32 s5, s6
; GFX6-NEXT: buffer_load_dword v3, v[0:1], s[4:7], 0 addr64
-; GFX6-NEXT: v_cvt_f16_f32_e32 v2, v2
-; GFX6-NEXT: v_and_b32_e32 v4, 3, v4
-; GFX6-NEXT: v_lshlrev_b32_e32 v6, 3, v4
+; GFX6-NEXT: v_cvt_f16_f32_e32 v5, v2
+; GFX6-NEXT: v_and_b32_e32 v2, 3, v4
+; GFX6-NEXT: v_lshlrev_b32_e32 v2, 3, v2
+; GFX6-NEXT: v_lshl_b32_e32 v4, 0xffff, v2
+; GFX6-NEXT: v_cvt_f32_f16_e32 v5, v5
+; GFX6-NEXT: v_not_b32_e32 v6, v4
; GFX6-NEXT: s_mov_b64 s[8:9], 0
-; GFX6-NEXT: v_cvt_f32_f16_e32 v7, v2
-; GFX6-NEXT: v_lshl_b32_e32 v2, 0xffff, v6
-; GFX6-NEXT: v_not_b32_e32 v8, v2
; GFX6-NEXT: .LBB34_1: ; %atomicrmw.start
; GFX6-NEXT: ; =>This Inner Loop Header: Depth=1
; GFX6-NEXT: s_waitcnt vmcnt(0)
-; GFX6-NEXT: v_lshrrev_b32_e32 v2, v6, v3
-; GFX6-NEXT: v_cvt_f32_f16_e32 v2, v2
+; GFX6-NEXT: v_mov_b32_e32 v8, v3
; GFX6-NEXT: s_waitcnt expcnt(0)
-; GFX6-NEXT: v_and_b32_e32 v4, v3, v8
-; GFX6-NEXT: v_max_f32_e32 v2, v2, v7
-; GFX6-NEXT: v_cvt_f16_f32_e32 v2, v2
-; GFX6-NEXT: v_lshlrev_b32_e32 v2, v6, v2
-; GFX6-NEXT: v_or_b32_e32 v2, v4, v2
-; GFX6-NEXT: v_mov_b32_e32 v5, v3
-; GFX6-NEXT: v_mov_b32_e32 v4, v2
-; GFX6-NEXT: buffer_atomic_cmpswap v[4:5], v[0:1], s[4:7], 0 addr64 glc
+; GFX6-NEXT: v_lshrrev_b32_e32 v3, v2, v8
+; GFX6-NEXT: v_cvt_f32_f16_e32 v3, v3
+; GFX6-NEXT: v_and_b32_e32 v4, v8, v6
+; GFX6-NEXT: v_max_f32_e32 v3, v3, v5
+; GFX6-NEXT: v_cvt_f16_f32_e32 v3, v3
+; GFX6-NEXT: v_lshlrev_b32_e32 v3, v2, v3
+; GFX6-NEXT: v_or_b32_e32 v7, v4, v3
+; GFX6-NEXT: v_mov_b32_e32 v3, v7
+; GFX6-NEXT: v_mov_b32_e32 v4, v8
+; GFX6-NEXT: buffer_atomic_cmpswap v[3:4], v[0:1], s[4:7], 0 addr64 glc
; GFX6-NEXT: s_waitcnt vmcnt(0)
; GFX6-NEXT: buffer_wbinvl1
-; GFX6-NEXT: v_cmp_eq_u32_e32 vcc, v4, v3
+; GFX6-NEXT: v_cmp_eq_u32_e32 vcc, v3, v8
; GFX6-NEXT: s_or_b64 s[8:9], vcc, s[8:9]
-; GFX6-NEXT: v_mov_b32_e32 v3, v4
; GFX6-NEXT: s_andn2_b64 exec, exec, s[8:9]
; GFX6-NEXT: s_cbranch_execnz .LBB34_1
; GFX6-NEXT: ; %bb.2: ; %atomicrmw.end
; GFX6-NEXT: s_or_b64 exec, exec, s[8:9]
-; GFX6-NEXT: v_lshrrev_b32_e32 v0, v6, v4
+; GFX6-NEXT: v_lshrrev_b32_e32 v0, v2, v3
; GFX6-NEXT: v_cvt_f32_f16_e32 v0, v0
; GFX6-NEXT: s_waitcnt expcnt(0)
; GFX6-NEXT: s_setpc_b64 s[30:31]
@@ -8644,31 +8645,31 @@ define void @global_system_atomic_fmax_noret_f16__offset12b_pos__amdgpu_no_fine_
; GFX7-NEXT: s_mov_b32 s4, s6
; GFX7-NEXT: s_mov_b32 s5, s6
; GFX7-NEXT: buffer_load_dword v3, v[0:1], s[4:7], 0 addr64
-; GFX7-NEXT: v_cvt_f16_f32_e32 v2, v2
-; GFX7-NEXT: v_and_b32_e32 v4, 3, v4
-; GFX7-NEXT: v_lshlrev_b32_e32 v4, 3, v4
+; GFX7-NEXT: v_cvt_f16_f32_e32 v5, v2
+; GFX7-NEXT: v_and_b32_e32 v2, 3, v4
+; GFX7-NEXT: v_lshlrev_b32_e32 v2, 3, v2
+; GFX7-NEXT: v_lshl_b32_e32 v4, 0xffff, v2
+; GFX7-NEXT: v_cvt_f32_f16_e32 v5, v5
+; GFX7-NEXT: v_not_b32_e32 v6, v4
; GFX7-NEXT: s_mov_b64 s[8:9], 0
-; GFX7-NEXT: v_cvt_f32_f16_e32 v5, v2
-; GFX7-NEXT: v_lshl_b32_e32 v2, 0xffff, v4
-; GFX7-NEXT: v_not_b32_e32 v6, v2
; GFX7-NEXT: .LBB35_1: ; %atomicrmw.start
; GFX7-NEXT: ; =>This Inner Loop Header: Depth=1
; GFX7-NEXT: s_waitcnt vmcnt(0)
-; GFX7-NEXT: v_lshrrev_b32_e32 v2, v4, v3
-; GFX7-NEXT: v_cvt_f32_f16_e32 v2, v2
-; GFX7-NEXT: v_and_b32_e32 v7, v3, v6
-; GFX7-NEXT: v_max_f32_e32 v2, v2, v5
-; GFX7-NEXT: v_cvt_f16_f32_e32 v2, v2
-; GFX7-NEXT: v_lshlrev_b32_e32 v2, v4, v2
-; GFX7-NEXT: v_or_b32_e32 v2, v7, v2
; GFX7-NEXT: v_mov_b32_e32 v8, v3
-; GFX7-NEXT: v_mov_b32_e32 v7, v2
-; GFX7-NEXT: buffer_atomic_cmpswap v[7:8], v[0:1], s[4:7], 0 addr64 glc
+; GFX7-NEXT: v_lshrrev_b32_e32 v3, v2, v8
+; GFX7-NEXT: v_cvt_f32_f16_e32 v3, v3
+; GFX7-NEXT: v_and_b32_e32 v4, v8, v6
+; GFX7-NEXT: v_max_f32_e32 v3, v3, v5
+; GFX7-NEXT: v_cvt_f16_f32_e32 v3, v3
+; GFX7-NEXT: v_lshlrev_b32_e32 v3, v2, v3
+; GFX7-NEXT: v_or_b32_e32 v7, v4, v3
+; GFX7-NEXT: v_mov_b32_e32 v3, v7
+; GFX7-NEXT: v_mov_b32_e32 v4, v8
+; GFX7-NEXT: buffer_atomic_cmpswap v[3:4], v[0:1], s[4:7], 0 addr64 glc
; GFX7-NEXT: s_waitcnt vmcnt(0)
; GFX7-NEXT: buffer_wbinvl1
-; GFX7-NEXT: v_cmp_eq_u32_e32 vcc, v7, v3
+; GFX7-NEXT: v_cmp_eq_u32_e32 vcc, v3, v8
; GFX7-NEXT: s_or_b64 s[8:9], vcc, s[8:9]
-; GFX7-NEXT: v_mov_b32_e32 v3, v7
; GFX7-NEXT: s_andn2_b64 exec, exec, s[8:9]
; GFX7-NEXT: s_cbranch_execnz .LBB35_1
; GFX7-NEXT: ; %bb.2: ; %atomicrmw.end
@@ -8686,32 +8687,32 @@ define void @global_system_atomic_fmax_noret_f16__offset12b_pos__amdgpu_no_fine_
; GFX6-NEXT: s_mov_b32 s4, s6
; GFX6-NEXT: s_mov_b32 s5, s6
; GFX6-NEXT: buffer_load_dword v3, v[0:1], s[4:7], 0 addr64
-; GFX6-NEXT: v_cvt_f16_f32_e32 v2, v2
-; GFX6-NEXT: v_and_b32_e32 v4, 3, v4
-; GFX6-NEXT: v_lshlrev_b32_e32 v4, 3, v4
+; GFX6-NEXT: v_cvt_f16_f32_e32 v5, v2
+; GFX6-NEXT: v_and_b32_e32 v2, 3, v4
+; GFX6-NEXT: v_lshlrev_b32_e32 v2, 3, v2
+; GFX6-NEXT: v_lshl_b32_e32 v4, 0xffff, v2
+; GFX6-NEXT: v_cvt_f32_f16_e32 v5, v5
+; GFX6-NEXT: v_not_b32_e32 v6, v4
; GFX6-NEXT: s_mov_b64 s[8:9], 0
-; GFX6-NEXT: v_cvt_f32_f16_e32 v5, v2
-; GFX6-NEXT: v_lshl_b32_e32 v2, 0xffff, v4
-; GFX6-NEXT: v_not_b32_e32 v6, v2
; GFX6-NEXT: .LBB35_1: ; %atomicrmw.start
; GFX6-NEXT: ; =>This Inner Loop Header: Depth=1
; GFX6-NEXT: s_waitcnt vmcnt(0)
-; GFX6-NEXT: v_lshrrev_b32_e32 v2, v4, v3
-; GFX6-NEXT: v_cvt_f32_f16_e32 v2, v2
-; GFX6-NEXT: s_waitcnt expcnt(0)
-; GFX6-NEXT: v_and_b32_e32 v7, v3, v6
-; GFX6-NEXT: v_max_f32_e32 v2, v2, v5
-; GFX6-NEXT: v_cvt_f16_f32_e32 v2, v2
-; GFX6-NEXT: v_lshlrev_b32_e32 v2, v4, v2
-; GFX6-NEXT: v_or_b32_e32 v2, v7, v2
; GFX6-NEXT: v_mov_b32_e32 v8, v3
-; GFX6-NEXT: v_mov_b32_e32 v7, v2
-; GFX6-NEXT: buffer_atomic_cmpswap v[7:8], v[0:1], s[4:7], 0 addr64 glc
+; GFX6-NEXT: s_waitcnt expcnt(0)
+; GFX6-NEXT: v_lshrrev_b32_e32 v3, v2, v8
+; GFX6-NEXT: v_cvt_f32_f16_e32 v3, v3
+; GFX6-NEXT: v_and_b32_e32 v4, v8, v6
+; GFX6-NEXT: v_max_f32_e32 v3, v3, v5
+; GFX6-NEXT: v_cvt_f16_f32_e32 v3, v3
+; GFX6-NEXT: v_lshlrev_b32_e32 v3, v2, v3
+; GFX6-NEXT: v_or_b32_e32 v7, v4, v3
+; GFX6-NEXT: v_mov_b32_e32 v3, v7
+; GFX6-NEXT: v_mov_b32_e32 v4, v8
+; GFX6-NEXT: buffer_atomic_cmpswap v[3:4], v[0:1], s[4:7], 0 addr64 glc
; GFX6-NEXT: s_waitcnt vmcnt(0)
; GFX6-NEXT: buffer_wbinvl1
-; GFX6-NEXT: v_cmp_eq_u32_e32 vcc, v7, v3
+; GFX6-NEXT: v_cmp_eq_u32_e32 vcc, v3, v8
; GFX6-NEXT: s_or_b64 s[8:9], vcc, s[8:9]
-; GFX6-NEXT: v_mov_b32_e32 v3, v7
; GFX6-NEXT: s_andn2_b64 exec, exec, s[8:9]
; GFX6-NEXT: s_cbranch_execnz .LBB35_1
; GFX6-NEXT: ; %bb.2: ; %atomicrmw.end
@@ -9146,36 +9147,36 @@ define bfloat @global_agent_atomic_fmax_ret_bf16__amdgpu_no_fine_grained_memory(
; GFX7-NEXT: s_mov_b32 s5, s6
; GFX7-NEXT: buffer_load_dword v4, v[0:1], s[4:7], 0 addr64
; GFX7-NEXT: v_and_b32_e32 v3, 3, v3
-; GFX7-NEXT: v_lshlrev_b32_e32 v5, 3, v3
-; GFX7-NEXT: v_lshl_b32_e32 v3, 0xffff, v5
+; GFX7-NEXT: v_lshlrev_b32_e32 v3, 3, v3
+; GFX7-NEXT: v_lshl_b32_e32 v5, 0xffff, v3
; GFX7-NEXT: v_mul_f32_e32 v2, 1.0, v2
-; GFX7-NEXT: v_not_b32_e32 v6, v3
+; GFX7-NEXT: v_not_b32_e32 v6, v5
; GFX7-NEXT: s_mov_b64 s[8:9], 0
-; GFX7-NEXT: v_and_b32_e32 v7, 0xffff0000, v2
+; GFX7-NEXT: v_and_b32_e32 v2, 0xffff0000, v2
; GFX7-NEXT: .LBB36_1: ; %atomicrmw.start
; GFX7-NEXT: ; =>This Inner Loop Header: Depth=1
; GFX7-NEXT: s_waitcnt vmcnt(0)
-; GFX7-NEXT: v_lshrrev_b32_e32 v2, v5, v4
-; GFX7-NEXT: v_lshlrev_b32_e32 v2, 16, v2
-; GFX7-NEXT: v_mul_f32_e32 v2, 1.0, v2
-; GFX7-NEXT: v_max_f32_e32 v2, v2, v7
-; GFX7-NEXT: v_lshrrev_b32_e32 v2, 16, v2
-; GFX7-NEXT: v_and_b32_e32 v3, v4, v6
-; GFX7-NEXT: v_lshlrev_b32_e32 v2, v5, v2
-; GFX7-NEXT: v_or_b32_e32 v3, v3, v2
-; GFX7-NEXT: v_mov_b32_e32 v2, v3
-; GFX7-NEXT: v_mov_b32_e32 v3, v4
-; GFX7-NEXT: buffer_atomic_cmpswap v[2:3], v[0:1], s[4:7], 0 addr64 glc
+; GFX7-NEXT: v_mov_b32_e32 v8, v4
+; GFX7-NEXT: v_lshrrev_b32_e32 v4, v3, v8
+; GFX7-NEXT: v_lshlrev_b32_e32 v4, 16, v4
+; GFX7-NEXT: v_mul_f32_e32 v4, 1.0, v4
+; GFX7-NEXT: v_max_f32_e32 v4, v4, v2
+; GFX7-NEXT: v_lshrrev_b32_e32 v4, 16, v4
+; GFX7-NEXT: v_and_b32_e32 v5, v8, v6
+; GFX7-NEXT: v_lshlrev_b32_e32 v4, v3, v4
+; GFX7-NEXT: v_or_b32_e32 v7, v5, v4
+; GFX7-NEXT: v_mov_b32_e32 v4, v7
+; GFX7-NEXT: v_mov_b32_e32 v5, v8
+; GFX7-NEXT: buffer_atomic_cmpswap v[4:5], v[0:1], s[4:7], 0 addr64 glc
; GFX7-NEXT: s_waitcnt vmcnt(0)
; GFX7-NEXT: buffer_wbinvl1
-; GFX7-NEXT: v_cmp_eq_u32_e32 vcc, v2, v4
+; GFX7-NEXT: v_cmp_eq_u32_e32 vcc, v4, v8
; GFX7-NEXT: s_or_b64 s[8:9], vcc, s[8:9]
-; GFX7-NEXT: v_mov_b32_e32 v4, v2
; GFX7-NEXT: s_andn2_b64 exec, exec, s[8:9]
; GFX7-NEXT: s_cbranch_execnz .LBB36_1
; GFX7-NEXT: ; %bb.2: ; %atomicrmw.end
; GFX7-NEXT: s_or_b64 exec, exec, s[8:9]
-; GFX7-NEXT: v_lshrrev_b32_e32 v0, v5, v2
+; GFX7-NEXT: v_lshrrev_b32_e32 v0, v3, v4
; GFX7-NEXT: v_lshlrev_b32_e32 v0, 16, v0
; GFX7-NEXT: s_setpc_b64 s[30:31]
;
@@ -9190,36 +9191,37 @@ define bfloat @global_agent_atomic_fmax_ret_bf16__amdgpu_no_fine_grained_memory(
; GFX6-NEXT: s_mov_b32 s5, s6
; GFX6-NEXT: buffer_load_dword v4, v[0:1], s[4:7], 0 addr64
; GFX6-NEXT: v_and_b32_e32 v3, 3, v3
-; GFX6-NEXT: v_lshlrev_b32_e32 v5, 3, v3
-; GFX6-NEXT: v_lshl_b32_e32 v3, 0xffff, v5
+; GFX6-NEXT: v_lshlrev_b32_e32 v3, 3, v3
+; GFX6-NEXT: v_lshl_b32_e32 v5, 0xffff, v3
; GFX6-NEXT: v_mul_f32_e32 v2, 1.0, v2
-; GFX6-NEXT: v_not_b32_e32 v6, v3
+; GFX6-NEXT: v_not_b32_e32 v6, v5
; GFX6-NEXT: s_mov_b64 s[8:9], 0
-; GFX6-NEXT: v_and_b32_e32 v7, 0xffff0000, v2
+; GFX6-NEXT: v_and_b32_e32 v2, 0xffff0000, v2
; GFX6-NEXT: .LBB36_1: ; %atomicrmw.start
; GFX6-NEXT: ; =>This Inner Loop Header: Depth=1
-; GFX6-NEXT: s_waitcnt vmcnt(0) expcnt(0)
-; GFX6-NEXT: v_lshrrev_b32_e32 v2, v5, v4
-; GFX6-NEXT: v_lshlrev_b32_e32 v2, 16, v2
-; GFX6-NEXT: v_mul_f32_e32 v2, 1.0, v2
-; GFX6-NEXT: v_max_f32_e32 v2, v2, v7
-; GFX6-NEXT: v_lshrrev_b32_e32 v2, 16, v2
-; GFX6-NEXT: v_and_b32_e32 v3, v4, v6
-; GFX6-NEXT: v_lshlrev_b32_e32 v2, v5, v2
-; GFX6-NEXT: v_or_b32_e32 v3, v3, v2
-; GFX6-NEXT: v_mov_b32_e32 v2, v3
-; GFX6-NEXT: v_mov_b32_e32 v3, v4
-; GFX6-NEXT: buffer_atomic_cmpswap v[2:3], v[0:1], s[4:7], 0 addr64 glc
+; GFX6-NEXT: s_waitcnt vmcnt(0)
+; GFX6-NEXT: v_mov_b32_e32 v8, v4
+; GFX6-NEXT: s_waitcnt expcnt(0)
+; GFX6-NEXT: v_lshrrev_b32_e32 v4, v3, v8
+; GFX6-NEXT: v_lshlrev_b32_e32 v4, 16, v4
+; GFX6-NEXT: v_mul_f32_e32 v4, 1.0, v4
+; GFX6-NEXT: v_max_f32_e32 v4, v4, v2
+; GFX6-NEXT: v_lshrrev_b32_e32 v4, 16, v4
+; GFX6-NEXT: v_and_b32_e32 v5, v8, v6
+; GFX6-NEXT: v_lshlrev_b32_e32 v4, v3, v4
+; GFX6-NEXT: v_or_b32_e32 v7, v5, v4
+; GFX6-NEXT: v_mov_b32_e32 v4, v7
+; GFX6-NEXT: v_mov_b32_e32 v5, v8
+; GFX6-NEXT: buffer_atomic_cmpswap v[4:5], v[0:1], s[4:7], 0 addr64 glc
; GFX6-NEXT: s_waitcnt vmcnt(0)
; GFX6-NEXT: buffer_wbinvl1
-; GFX6-NEXT: v_cmp_eq_u32_e32 vcc, v2, v4
+; GFX6-NEXT: v_cmp_eq_u32_e32 vcc, v4, v8
; GFX6-NEXT: s_or_b64 s[8:9], vcc, s[8:9]
-; GFX6-NEXT: v_mov_b32_e32 v4, v2
; GFX6-NEXT: s_andn2_b64 exec, exec, s[8:9]
; GFX6-NEXT: s_cbranch_execnz .LBB36_1
; GFX6-NEXT: ; %bb.2: ; %atomicrmw.end
; GFX6-NEXT: s_or_b64 exec, exec, s[8:9]
-; GFX6-NEXT: v_lshrrev_b32_e32 v0, v5, v2
+; GFX6-NEXT: v_lshrrev_b32_e32 v0, v3, v4
; GFX6-NEXT: v_lshlrev_b32_e32 v0, 16, v0
; GFX6-NEXT: s_waitcnt expcnt(0)
; GFX6-NEXT: s_setpc_b64 s[30:31]
@@ -9661,36 +9663,36 @@ define bfloat @global_agent_atomic_fmax_ret_bf16__offset12b_pos__amdgpu_no_fine_
; GFX7-NEXT: s_mov_b32 s5, s6
; GFX7-NEXT: buffer_load_dword v3, v[0:1], s[4:7], 0 addr64
; GFX7-NEXT: v_and_b32_e32 v4, 3, v4
-; GFX7-NEXT: v_lshlrev_b32_e32 v6, 3, v4
-; GFX7-NEXT: v_lshl_b32_e32 v4, 0xffff, v6
+; GFX7-NEXT: v_lshlrev_b32_e32 v5, 3, v4
+; GFX7-NEXT: v_lshl_b32_e32 v4, 0xffff, v5
; GFX7-NEXT: v_mul_f32_e32 v2, 1.0, v2
-; GFX7-NEXT: v_not_b32_e32 v7, v4
+; GFX7-NEXT: v_not_b32_e32 v6, v4
; GFX7-NEXT: s_mov_b64 s[8:9], 0
-; GFX7-NEXT: v_and_b32_e32 v8, 0xffff0000, v2
+; GFX7-NEXT: v_and_b32_e32 v2, 0xffff0000, v2
; GFX7-NEXT: .LBB37_1: ; %atomicrmw.start
; GFX7-NEXT: ; =>This Inner Loop Header: Depth=1
; GFX7-NEXT: s_waitcnt vmcnt(0)
-; GFX7-NEXT: v_lshrrev_b32_e32 v2, v6, v3
-; GFX7-NEXT: v_lshlrev_b32_e32 v2, 16, v2
-; GFX7-NEXT: v_mul_f32_e32 v2, 1.0, v2
-; GFX7-NEXT: v_max_f32_e32 v2, v2, v8
-; GFX7-NEXT: v_lshrrev_b32_e32 v2, 16, v2
-; GFX7-NEXT: v_and_b32_e32 v4, v3, v7
-; GFX7-NEXT: v_lshlrev_b32_e32 v2, v6, v2
-; GFX7-NEXT: v_or_b32_e32 v2, v4, v2
-; GFX7-NEXT: v_mov_b32_e32 v5, v3
-; GFX7-NEXT: v_mov_b32_e32 v4, v2
-; GFX7-NEXT: buffer_atomic_cmpswap v[4:5], v[0:1], s[4:7], 0 addr64 glc
+; GFX7-NEXT: v_mov_b32_e32 v8, v3
+; GFX7-NEXT: v_lshrrev_b32_e32 v3, v5, v8
+; GFX7-NEXT: v_lshlrev_b32_e32 v3, 16, v3
+; GFX7-NEXT: v_mul_f32_e32 v3, 1.0, v3
+; GFX7-NEXT: v_max_f32_e32 v3, v3, v2
+; GFX7-NEXT: v_lshrrev_b32_e32 v3, 16, v3
+; GFX7-NEXT: v_and_b32_e32 v4, v8, v6
+; GFX7-NEXT: v_lshlrev_b32_e32 v3, v5, v3
+; GFX7-NEXT: v_or_b32_e32 v7, v4, v3
+; GFX7-NEXT: v_mov_b32_e32 v3, v7
+; GFX7-NEXT: v_mov_b32_e32 v4, v8
+; GFX7-NEXT: buffer_atomic_cmpswap v[3:4], v[0:1], s[4:7], 0 addr64 glc
; GFX7-NEXT: s_waitcnt vmcnt(0)
; GFX7-NEXT: buffer_wbinvl1
-; GFX7-NEXT: v_cmp_eq_u32_e32 vcc, v4, v3
+; GFX7-NEXT: v_cmp_eq_u32_e32 vcc, v3, v8
; GFX7-NEXT: s_or_b64 s[8:9], vcc, s[8:9]
-; GFX7-NEXT: v_mov_b32_e32 v3, v4
; GFX7-NEXT: s_andn2_b64 exec, exec, s[8:9]
; GFX7-NEXT: s_cbranch_execnz .LBB37_1
; GFX7-NEXT: ; %bb.2: ; %atomicrmw.end
; GFX7-NEXT: s_or_b64 exec, exec, s[8:9]
-; GFX7-NEXT: v_lshrrev_b32_e32 v0, v6, v4
+; GFX7-NEXT: v_lshrrev_b32_e32 v0, v5, v3
; GFX7-NEXT: v_lshlrev_b32_e32 v0, 16, v0
; GFX7-NEXT: s_setpc_b64 s[30:31]
;
@@ -9706,37 +9708,37 @@ define bfloat @global_agent_atomic_fmax_ret_bf16__offset12b_pos__amdgpu_no_fine_
; GFX6-NEXT: s_mov_b32 s5, s6
; GFX6-NEXT: buffer_load_dword v3, v[0:1], s[4:7], 0 addr64
; GFX6-NEXT: v_and_b32_e32 v4, 3, v4
-; GFX6-NEXT: v_lshlrev_b32_e32 v6, 3, v4
-; GFX6-NEXT: v_lshl_b32_e32 v4, 0xffff, v6
+; GFX6-NEXT: v_lshlrev_b32_e32 v5, 3, v4
+; GFX6-NEXT: v_lshl_b32_e32 v4, 0xffff, v5
; GFX6-NEXT: v_mul_f32_e32 v2, 1.0, v2
-; GFX6-NEXT: v_not_b32_e32 v7, v4
+; GFX6-NEXT: v_not_b32_e32 v6, v4
; GFX6-NEXT: s_mov_b64 s[8:9], 0
-; GFX6-NEXT: v_and_b32_e32 v8, 0xffff0000, v2
+; GFX6-NEXT: v_and_b32_e32 v2, 0xffff0000, v2
; GFX6-NEXT: .LBB37_1: ; %atomicrmw.start
; GFX6-NEXT: ; =>This Inner Loop Header: Depth=1
; GFX6-NEXT: s_waitcnt vmcnt(0)
-; GFX6-NEXT: v_lshrrev_b32_e32 v2, v6, v3
-; GFX6-NEXT: v_lshlrev_b32_e32 v2, 16, v2
-; GFX6-NEXT: v_mul_f32_e32 v2, 1.0, v2
-; GFX6-NEXT: v_max_f32_e32 v2, v2, v8
-; GFX6-NEXT: v_lshrrev_b32_e32 v2, 16, v2
+; GFX6-NEXT: v_mov_b32_e32 v8, v3
; GFX6-NEXT: s_waitcnt expcnt(0)
-; GFX6-NEXT: v_and_b32_e32 v4, v3, v7
-; GFX6-NEXT: v_lshlrev_b32_e32 v2, v6, v2
-; GFX6-NEXT: v_or_b32_e32 v2, v4, v2
-; GFX6-NEXT: v_mov_b32_e32 v5, v3
-; GFX6-NEXT: v_mov_b32_e32 v4, v2
-; GFX6-NEXT: buffer_atomic_cmpswap v[4:5], v[0:1], s[4:7], 0 addr64 glc
+; GFX6-NEXT: v_lshrrev_b32_e32 v3, v5, v8
+; GFX6-NEXT: v_lshlrev_b32_e32 v3, 16, v3
+; GFX6-NEXT: v_mul_f32_e32 v3, 1.0, v3
+; GFX6-NEXT: v_max_f32_e32 v3, v3, v2
+; GFX6-NEXT: v_lshrrev_b32_e32 v3, 16, v3
+; GFX6-NEXT: v_and_b32_e32 v4, v8, v6
+; GFX6-NEXT: v_lshlrev_b32_e32 v3, v5, v3
+; GFX6-NEXT: v_or_b32_e32 v7, v4, v3
+; GFX6-NEXT: v_mov_b32_e32 v3, v7
+; GFX6-NEXT: v_mov_b32_e32 v4, v8
+; GFX6-NEXT: buffer_atomic_cmpswap v[3:4], v[0:1], s[4:7], 0 addr64 glc
; GFX6-NEXT: s_waitcnt vmcnt(0)
; GFX6-NEXT: buffer_wbinvl1
-; GFX6-NEXT: v_cmp_eq_u32_e32 vcc, v4, v3
+; GFX6-NEXT: v_cmp_eq_u32_e32 vcc, v3, v8
; GFX6-NEXT: s_or_b64 s[8:9], vcc, s[8:9]
-; GFX6-NEXT: v_mov_b32_e32 v3, v4
; GFX6-NEXT: s_andn2_b64 exec, exec, s[8:9]
; GFX6-NEXT: s_cbranch_execnz .LBB37_1
; GFX6-NEXT: ; %bb.2: ; %atomicrmw.end
; GFX6-NEXT: s_or_b64 exec, exec, s[8:9]
-; GFX6-NEXT: v_lshrrev_b32_e32 v0, v6, v4
+; GFX6-NEXT: v_lshrrev_b32_e32 v0, v5, v3
; GFX6-NEXT: v_lshlrev_b32_e32 v0, 16, v0
; GFX6-NEXT: s_waitcnt expcnt(0)
; GFX6-NEXT: s_setpc_b64 s[30:31]
@@ -10180,36 +10182,36 @@ define bfloat @global_agent_atomic_fmax_ret_bf16__offset12b_neg__amdgpu_no_fine_
; GFX7-NEXT: s_mov_b32 s5, s6
; GFX7-NEXT: buffer_load_dword v3, v[0:1], s[4:7], 0 addr64
; GFX7-NEXT: v_and_b32_e32 v4, 3, v4
-; GFX7-NEXT: v_lshlrev_b32_e32 v6, 3, v4
-; GFX7-NEXT: v_lshl_b32_e32 v4, 0xffff, v6
+; GFX7-NEXT: v_lshlrev_b32_e32 v5, 3, v4
+; GFX7-NEXT: v_lshl_b32_e32 v4, 0xffff, v5
; GFX7-NEXT: v_mul_f32_e32 v2, 1.0, v2
-; GFX7-NEXT: v_not_b32_e32 v7, v4
+; GFX7-NEXT: v_not_b32_e32 v6, v4
; GFX7-NEXT: s_mov_b64 s[8:9], 0
-; GFX7-NEXT: v_and_b32_e32 v8, 0xffff0000, v2
+; GFX7-NEXT: v_and_b32_e32 v2, 0xffff0000, v2
; GFX7-NEXT: .LBB38_1: ; %atomicrmw.start
; GFX7-NEXT: ; =>This Inner Loop Header: Depth=1
; GFX7-NEXT: s_waitcnt vmcnt(0)
-; GFX7-NEXT: v_lshrrev_b32_e32 v2, v6, v3
-; GFX7-NEXT: v_lshlrev_b32_e32 v2, 16, v2
-; GFX7-NEXT: v_mul_f32_e32 v2, 1.0, v2
-; GFX7-NEXT: v_max_f32_e32 v2, v2, v8
-; GFX7-NEXT: v_lshrrev_b32_e32 v2, 16, v2
-; GFX7-NEXT: v_and_b32_e32 v4, v3, v7
-; GFX7-NEXT: v_lshlrev_b32_e32 v2, v6, v2
-; GFX7-NEXT: v_or_b32_e32 v2, v4, v2
-; GFX7-NEXT: v_mov_b32_e32 v5, v3
-; GFX7-NEXT: v_mov_b32_e32 v4, v2
-; GFX7-NEXT: buffer_atomic_cmpswap v[4:5], v[0:1], s[4:7], 0 addr64 glc
+; GFX7-NEXT: v_mov_b32_e32 v8, v3
+; GFX7-NEXT: v_lshrrev_b32_e32 v3, v5, v8
+; GFX7-NEXT: v_lshlrev_b32_e32 v3, 16, v3
+; GFX7-NEXT: v_mul_f32_e32 v3, 1.0, v3
+; GFX7-NEXT: v_max_f32_e32 v3, v3, v2
+; GFX7-NEXT: v_lshrrev_b32_e32 v3, 16, v3
+; GFX7-NEXT: v_and_b32_e32 v4, v8, v6
+; GFX7-NEXT: v_lshlrev_b32_e32 v3, v5, v3
+; GFX7-NEXT: v_or_b32_e32 v7, v4, v3
+; GFX7-NEXT: v_mov_b32_e32 v3, v7
+; GFX7-NEXT: v_mov_b32_e32 v4, v8
+; GFX7-NEXT: buffer_atomic_cmpswap v[3:4], v[0:1], s[4:7], 0 addr64 glc
; GFX7-NEXT: s_waitcnt vmcnt(0)
; GFX7-NEXT: buffer_wbinvl1
-; GFX7-NEXT: v_cmp_eq_u32_e32 vcc, v4, v3
+; GFX7-NEXT: v_cmp_eq_u32_e32 vcc, v3, v8
; GFX7-NEXT: s_or_b64 s[8:9], vcc, s[8:9]
-; GFX7-NEXT: v_mov_b32_e32 v3, v4
; GFX7-NEXT: s_andn2_b64 exec, exec, s[8:9]
; GFX7-NEXT: s_cbranch_execnz .LBB38_1
; GFX7-NEXT: ; %bb.2: ; %atomicrmw.end
; GFX7-NEXT: s_or_b64 exec, exec, s[8:9]
-; GFX7-NEXT: v_lshrrev_b32_e32 v0, v6, v4
+; GFX7-NEXT: v_lshrrev_b32_e32 v0, v5, v3
; GFX7-NEXT: v_lshlrev_b32_e32 v0, 16, v0
; GFX7-NEXT: s_setpc_b64 s[30:31]
;
@@ -10225,37 +10227,37 @@ define bfloat @global_agent_atomic_fmax_ret_bf16__offset12b_neg__amdgpu_no_fine_
; GFX6-NEXT: s_mov_b32 s5, s6
; GFX6-NEXT: buffer_load_dword v3, v[0:1], s[4:7], 0 addr64
; GFX6-NEXT: v_and_b32_e32 v4, 3, v4
-; GFX6-NEXT: v_lshlrev_b32_e32 v6, 3, v4
-; GFX6-NEXT: v_lshl_b32_e32 v4, 0xffff, v6
+; GFX6-NEXT: v_lshlrev_b32_e32 v5, 3, v4
+; GFX6-NEXT: v_lshl_b32_e32 v4, 0xffff, v5
; GFX6-NEXT: v_mul_f32_e32 v2, 1.0, v2
-; GFX6-NEXT: v_not_b32_e32 v7, v4
+; GFX6-NEXT: v_not_b32_e32 v6, v4
; GFX6-NEXT: s_mov_b64 s[8:9], 0
-; GFX6-NEXT: v_and_b32_e32 v8, 0xffff0000, v2
+; GFX6-NEXT: v_and_b32_e32 v2, 0xffff0000, v2
; GFX6-NEXT: .LBB38_1: ; %atomicrmw.start
; GFX6-NEXT: ; =>This Inner Loop Header: Depth=1
; GFX6-NEXT: s_waitcnt vmcnt(0)
-; GFX6-NEXT: v_lshrrev_b32_e32 v2, v6, v3
-; GFX6-NEXT: v_lshlrev_b32_e32 v2, 16, v2
-; GFX6-NEXT: v_mul_f32_e32 v2, 1.0, v2
-; GFX6-NEXT: v_max_f32_e32 v2, v2, v8
-; GFX6-NEXT: v_lshrrev_b32_e32 v2, 16, v2
+; GFX6-NEXT: v_mov_b32_e32 v8, v3
; GFX6-NEXT: s_waitcnt expcnt(0)
-; GFX6-NEXT: v_and_b32_e32 v4, v3, v7
-; GFX6-NEXT: v_lshlrev_b32_e32 v2, v6, v2
-; GFX6-NEXT: v_or_b32_e32 v2, v4, v2
-; GFX6-NEXT: v_mov_b32_e32 v5, v3
-; GFX6-NEXT: v_mov_b32_e32 v4, v2
-; GFX6-NEXT: buffer_atomic_cmpswap v[4:5], v[0:1], s[4:7], 0 addr64 glc
+; GFX6-NEXT: v_lshrrev_b32_e32 v3, v5, v8
+; GFX6-NEXT: v_lshlrev_b32_e32 v3, 16, v3
+; GFX6-NEXT: v_mul_f32_e32 v3, 1.0, v3
+; GFX6-NEXT: v_max_f32_e32 v3, v3, v2
+; GFX6-NEXT: v_lshrrev_b32_e32 v3, 16, v3
+; GFX6-NEXT: v_and_b32_e32 v4, v8, v6
+; GFX6-NEXT: v_lshlrev_b32_e32 v3, v5, v3
+; GFX6-NEXT: v_or_b32_e32 v7, v4, v3
+; GFX6-NEXT: v_mov_b32_e32 v3, v7
+; GFX6-NEXT: v_mov_b32_e32 v4, v8
+; GFX6-NEXT: buffer_atomic_cmpswap v[3:4], v[0:1], s[4:7], 0 addr64 glc
; GFX6-NEXT: s_waitcnt vmcnt(0)
; GFX6-NEXT: buffer_wbinvl1
-; GFX6-NEXT: v_cmp_eq_u32_e32 vcc, v4, v3
+; GFX6-NEXT: v_cmp_eq_u32_e32 vcc, v3, v8
; GFX6-NEXT: s_or_b64 s[8:9], vcc, s[8:9]
-; GFX6-NEXT: v_mov_b32_e32 v3, v4
; GFX6-NEXT: s_andn2_b64 exec, exec, s[8:9]
; GFX6-NEXT: s_cbranch_execnz .LBB38_1
; GFX6-NEXT: ; %bb.2: ; %atomicrmw.end
; GFX6-NEXT: s_or_b64 exec, exec, s[8:9]
-; GFX6-NEXT: v_lshrrev_b32_e32 v0, v6, v4
+; GFX6-NEXT: v_lshrrev_b32_e32 v0, v5, v3
; GFX6-NEXT: v_lshlrev_b32_e32 v0, 16, v0
; GFX6-NEXT: s_waitcnt expcnt(0)
; GFX6-NEXT: s_setpc_b64 s[30:31]
@@ -10670,31 +10672,31 @@ define void @global_agent_atomic_fmax_noret_bf16__amdgpu_no_fine_grained_memory(
; GFX7-NEXT: s_mov_b32 s5, s6
; GFX7-NEXT: buffer_load_dword v4, v[0:1], s[4:7], 0 addr64
; GFX7-NEXT: v_and_b32_e32 v3, 3, v3
-; GFX7-NEXT: v_lshlrev_b32_e32 v5, 3, v3
-; GFX7-NEXT: v_lshl_b32_e32 v3, 0xffff, v5
+; GFX7-NEXT: v_lshlrev_b32_e32 v3, 3, v3
+; GFX7-NEXT: v_lshl_b32_e32 v5, 0xffff, v3
; GFX7-NEXT: v_mul_f32_e32 v2, 1.0, v2
-; GFX7-NEXT: v_not_b32_e32 v6, v3
+; GFX7-NEXT: v_not_b32_e32 v6, v5
; GFX7-NEXT: s_mov_b64 s[8:9], 0
; GFX7-NEXT: v_and_b32_e32 v2, 0xffff0000, v2
; GFX7-NEXT: .LBB39_1: ; %atomicrmw.start
; GFX7-NEXT: ; =>This Inner Loop Header: Depth=1
; GFX7-NEXT: s_waitcnt vmcnt(0)
-; GFX7-NEXT: v_lshrrev_b32_e32 v3, v5, v4
-; GFX7-NEXT: v_lshlrev_b32_e32 v3, 16, v3
-; GFX7-NEXT: v_mul_f32_e32 v3, 1.0, v3
-; GFX7-NEXT: v_max_f32_e32 v3, v3, v2
-; GFX7-NEXT: v_lshrrev_b32_e32 v3, 16, v3
-; GFX7-NEXT: v_and_b32_e32 v7, v4, v6
-; GFX7-NEXT: v_lshlrev_b32_e32 v3, v5, v3
-; GFX7-NEXT: v_or_b32_e32 v3, v7, v3
; GFX7-NEXT: v_mov_b32_e32 v8, v4
-; GFX7-NEXT: v_mov_b32_e32 v7, v3
-; GFX7-NEXT: buffer_atomic_cmpswap v[7:8], v[0:1], s[4:7], 0 addr64 glc
+; GFX7-NEXT: v_lshrrev_b32_e32 v4, v3, v8
+; GFX7-NEXT: v_lshlrev_b32_e32 v4, 16, v4
+; GFX7-NEXT: v_mul_f32_e32 v4, 1.0, v4
+; GFX7-NEXT: v_max_f32_e32 v4, v4, v2
+; GFX7-NEXT: v_lshrrev_b32_e32 v4, 16, v4
+; GFX7-NEXT: v_and_b32_e32 v5, v8, v6
+; GFX7-NEXT: v_lshlrev_b32_e32 v4, v3, v4
+; GFX7-NEXT: v_or_b32_e32 v7, v5, v4
+; GFX7-NEXT: v_mov_b32_e32 v4, v7
+; GFX7-NEXT: v_mov_b32_e32 v5, v8
+; GFX7-NEXT: buffer_atomic_cmpswap v[4:5], v[0:1], s[4:7], 0 addr64 glc
; GFX7-NEXT: s_waitcnt vmcnt(0)
; GFX7-NEXT: buffer_wbinvl1
-; GFX7-NEXT: v_cmp_eq_u32_e32 vcc, v7, v4
+; GFX7-NEXT: v_cmp_eq_u32_e32 vcc, v4, v8
; GFX7-NEXT: s_or_b64 s[8:9], vcc, s[8:9]
-; GFX7-NEXT: v_mov_b32_e32 v4, v7
; GFX7-NEXT: s_andn2_b64 exec, exec, s[8:9]
; GFX7-NEXT: s_cbranch_execnz .LBB39_1
; GFX7-NEXT: ; %bb.2: ; %atomicrmw.end
@@ -10712,32 +10714,32 @@ define void @global_agent_atomic_fmax_noret_bf16__amdgpu_no_fine_grained_memory(
; GFX6-NEXT: s_mov_b32 s5, s6
; GFX6-NEXT: buffer_load_dword v4, v[0:1], s[4:7], 0 addr64
; GFX6-NEXT: v_and_b32_e32 v3, 3, v3
-; GFX6-NEXT: v_lshlrev_b32_e32 v5, 3, v3
-; GFX6-NEXT: v_lshl_b32_e32 v3, 0xffff, v5
+; GFX6-NEXT: v_lshlrev_b32_e32 v3, 3, v3
+; GFX6-NEXT: v_lshl_b32_e32 v5, 0xffff, v3
; GFX6-NEXT: v_mul_f32_e32 v2, 1.0, v2
-; GFX6-NEXT: v_not_b32_e32 v6, v3
+; GFX6-NEXT: v_not_b32_e32 v6, v5
; GFX6-NEXT: s_mov_b64 s[8:9], 0
; GFX6-NEXT: v_and_b32_e32 v2, 0xffff0000, v2
; GFX6-NEXT: .LBB39_1: ; %atomicrmw.start
; GFX6-NEXT: ; =>This Inner Loop Header: Depth=1
; GFX6-NEXT: s_waitcnt vmcnt(0)
-; GFX6-NEXT: v_lshrrev_b32_e32 v3, v5, v4
-; GFX6-NEXT: v_lshlrev_b32_e32 v3, 16, v3
-; GFX6-NEXT: v_mul_f32_e32 v3, 1.0, v3
-; GFX6-NEXT: v_max_f32_e32 v3, v3, v2
-; GFX6-NEXT: v_lshrrev_b32_e32 v3, 16, v3
-; GFX6-NEXT: s_waitcnt expcnt(0)
-; GFX6-NEXT: v_and_b32_e32 v7, v4, v6
-; GFX6-NEXT: v_lshlrev_b32_e32 v3, v5, v3
-; GFX6-NEXT: v_or_b32_e32 v3, v7, v3
; GFX6-NEXT: v_mov_b32_e32 v8, v4
-; GFX6-NEXT: v_mov_b32_e32 v7, v3
-; GFX6-NEXT: buffer_atomic_cmpswap v[7:8], v[0:1], s[4:7], 0 addr64 glc
+; GFX6-NEXT: s_waitcnt expcnt(0)
+; GFX6-NEXT: v_lshrrev_b32_e32 v4, v3, v8
+; GFX6-NEXT: v_lshlrev_b32_e32 v4, 16, v4
+; GFX6-NEXT: v_mul_f32_e32 v4, 1.0, v4
+; GFX6-NEXT: v_max_f32_e32 v4, v4, v2
+; GFX6-NEXT: v_lshrrev_b32_e32 v4, 16, v4
+; GFX6-NEXT: v_and_b32_e32 v5, v8, v6
+; GFX6-NEXT: v_lshlrev_b32_e32 v4, v3, v4
+; GFX6-NEXT: v_or_b32_e32 v7, v5, v4
+; GFX6-NEXT: v_mov_b32_e32 v4, v7
+; GFX6-NEXT: v_mov_b32_e32 v5, v8
+; GFX6-NEXT: buffer_atomic_cmpswap v[4:5], v[0:1], s[4:7], 0 addr64 glc
; GFX6-NEXT: s_waitcnt vmcnt(0)
; GFX6-NEXT: buffer_wbinvl1
-; GFX6-NEXT: v_cmp_eq_u32_e32 vcc, v7, v4
+; GFX6-NEXT: v_cmp_eq_u32_e32 vcc, v4, v8
; GFX6-NEXT: s_or_b64 s[8:9], vcc, s[8:9]
-; GFX6-NEXT: v_mov_b32_e32 v4, v7
; GFX6-NEXT: s_andn2_b64 exec, exec, s[8:9]
; GFX6-NEXT: s_cbranch_execnz .LBB39_1
; GFX6-NEXT: ; %bb.2: ; %atomicrmw.end
@@ -11169,31 +11171,31 @@ define void @global_agent_atomic_fmax_noret_bf16__offset12b_pos__amdgpu_no_fine_
; GFX7-NEXT: s_mov_b32 s5, s6
; GFX7-NEXT: buffer_load_dword v3, v[0:1], s[4:7], 0 addr64
; GFX7-NEXT: v_and_b32_e32 v4, 3, v4
-; GFX7-NEXT: v_lshlrev_b32_e32 v4, 3, v4
-; GFX7-NEXT: v_lshl_b32_e32 v5, 0xffff, v4
+; GFX7-NEXT: v_lshlrev_b32_e32 v5, 3, v4
+; GFX7-NEXT: v_lshl_b32_e32 v4, 0xffff, v5
; GFX7-NEXT: v_mul_f32_e32 v2, 1.0, v2
-; GFX7-NEXT: v_not_b32_e32 v5, v5
+; GFX7-NEXT: v_not_b32_e32 v6, v4
; GFX7-NEXT: s_mov_b64 s[8:9], 0
-; GFX7-NEXT: v_and_b32_e32 v6, 0xffff0000, v2
+; GFX7-NEXT: v_and_b32_e32 v2, 0xffff0000, v2
; GFX7-NEXT: .LBB40_1: ; %atomicrmw.start
; GFX7-NEXT: ; =>This Inner Loop Header: Depth=1
; GFX7-NEXT: s_waitcnt vmcnt(0)
-; GFX7-NEXT: v_lshrrev_b32_e32 v2, v4, v3
-; GFX7-NEXT: v_lshlrev_b32_e32 v2, 16, v2
-; GFX7-NEXT: v_mul_f32_e32 v2, 1.0, v2
-; GFX7-NEXT: v_max_f32_e32 v2, v2, v6
-; GFX7-NEXT: v_lshrrev_b32_e32 v2, 16, v2
-; GFX7-NEXT: v_and_b32_e32 v7, v3, v5
-; GFX7-NEXT: v_lshlrev_b32_e32 v2, v4, v2
-; GFX7-NEXT: v_or_b32_e32 v2, v7, v2
; GFX7-NEXT: v_mov_b32_e32 v8, v3
-; GFX7-NEXT: v_mov_b32_e32 v7, v2
-; GFX7-NEXT: buffer_atomic_cmpswap v[7:8], v[0:1], s[4:7], 0 addr64 glc
+; GFX7-NEXT: v_lshrrev_b32_e32 v3, v5, v8
+; GFX7-NEXT: v_lshlrev_b32_e32 v3, 16, v3
+; GFX7-NEXT: v_mul_f32_e32 v3, 1.0, v3
+; GFX7-NEXT: v_max_f32_e32 v3, v3, v2
+; GFX7-NEXT: v_lshrrev_b32_e32 v3, 16, v3
+; GFX7-NEXT: v_and_b32_e32 v4, v8, v6
+; GFX7-NEXT: v_lshlrev_b32_e32 v3, v5, v3
+; GFX7-NEXT: v_or_b32_e32 v7, v4, v3
+; GFX7-NEXT: v_mov_b32_e32 v3, v7
+; GFX7-NEXT: v_mov_b32_e32 v4, v8
+; GFX7-NEXT: buffer_atomic_cmpswap v[3:4], v[0:1], s[4:7], 0 addr64 glc
; GFX7-NEXT: s_waitcnt vmcnt(0)
; GFX7-NEXT: buffer_wbinvl1
-; GFX7-NEXT: v_cmp_eq_u32_e32 vcc, v7, v3
+; GFX7-NEXT: v_cmp_eq_u32_e32 vcc, v3, v8
; GFX7-NEXT: s_or_b64 s[8:9], vcc, s[8:9]
-; GFX7-NEXT: v_mov_b32_e32 v3, v7
; GFX7-NEXT: s_andn2_b64 exec, exec, s[8:9]
; GFX7-NEXT: s_cbranch_execnz .LBB40_1
; GFX7-NEXT: ; %bb.2: ; %atomicrmw.end
@@ -11212,32 +11214,32 @@ define void @global_agent_atomic_fmax_noret_bf16__offset12b_pos__amdgpu_no_fine_
; GFX6-NEXT: s_mov_b32 s5, s6
; GFX6-NEXT: buffer_load_dword v3, v[0:1], s[4:7], 0 addr64
; GFX6-NEXT: v_and_b32_e32 v4, 3, v4
-; GFX6-NEXT: v_lshlrev_b32_e32 v4, 3, v4
-; GFX6-NEXT: v_lshl_b32_e32 v5, 0xffff, v4
+; GFX6-NEXT: v_lshlrev_b32_e32 v5, 3, v4
+; GFX6-NEXT: v_lshl_b32_e32 v4, 0xffff, v5
; GFX6-NEXT: v_mul_f32_e32 v2, 1.0, v2
-; GFX6-NEXT: v_not_b32_e32 v5, v5
+; GFX6-NEXT: v_not_b32_e32 v6, v4
; GFX6-NEXT: s_mov_b64 s[8:9], 0
-; GFX6-NEXT: v_and_b32_e32 v6, 0xffff0000, v2
+; GFX6-NEXT: v_and_b32_e32 v2, 0xffff0000, v2
; GFX6-NEXT: .LBB40_1: ; %atomicrmw.start
; GFX6-NEXT: ; =>This Inner Loop Header: Depth=1
; GFX6-NEXT: s_waitcnt vmcnt(0)
-; GFX6-NEXT: v_lshrrev_b32_e32 v2, v4, v3
-; GFX6-NEXT: v_lshlrev_b32_e32 v2, 16, v2
-; GFX6-NEXT: v_mul_f32_e32 v2, 1.0, v2
-; GFX6-NEXT: v_max_f32_e32 v2, v2, v6
-; GFX6-NEXT: v_lshrrev_b32_e32 v2, 16, v2
-; GFX6-NEXT: s_waitcnt expcnt(0)
-; GFX6-NEXT: v_and_b32_e32 v7, v3, v5
-; GFX6-NEXT: v_lshlrev_b32_e32 v2, v4, v2
-; GFX6-NEXT: v_or_b32_e32 v2, v7, v2
; GFX6-NEXT: v_mov_b32_e32 v8, v3
-; GFX6-NEXT: v_mov_b32_e32 v7, v2
-; GFX6-NEXT: buffer_atomic_cmpswap v[7:8], v[0:1], s[4:7], 0 addr64 glc
+; GFX6-NEXT: s_waitcnt expcnt(0)
+; GFX6-NEXT: v_lshrrev_b32_e32 v3, v5, v8
+; GFX6-NEXT: v_lshlrev_b32_e32 v3, 16, v3
+; GFX6-NEXT: v_mul_f32_e32 v3, 1.0, v3
+; GFX6-NEXT: v_max_f32_e32 v3, v3, v2
+; GFX6-NEXT: v_lshrrev_b32_e32 v3, 16, v3
+; GFX6-NEXT: v_and_b32_e32 v4, v8, v6
+; GFX6-NEXT: v_lshlrev_b32_e32 v3, v5, v3
+; GFX6-NEXT: v_or_b32_e32 v7, v4, v3
+; GFX6-NEXT: v_mov_b32_e32 v3, v7
+; GFX6-NEXT: v_mov_b32_e32 v4, v8
+; GFX6-NEXT: buffer_atomic_cmpswap v[3:4], v[0:1], s[4:7], 0 addr64 glc
; GFX6-NEXT: s_waitcnt vmcnt(0)
; GFX6-NEXT: buffer_wbinvl1
-; GFX6-NEXT: v_cmp_eq_u32_e32 vcc, v7, v3
+; GFX6-NEXT: v_cmp_eq_u32_e32 vcc, v3, v8
; GFX6-NEXT: s_or_b64 s[8:9], vcc, s[8:9]
-; GFX6-NEXT: v_mov_b32_e32 v3, v7
; GFX6-NEXT: s_andn2_b64 exec, exec, s[8:9]
; GFX6-NEXT: s_cbranch_execnz .LBB40_1
; GFX6-NEXT: ; %bb.2: ; %atomicrmw.end
@@ -11671,31 +11673,31 @@ define void @global_agent_atomic_fmax_noret_bf16__offset12b_neg__amdgpu_no_fine_
; GFX7-NEXT: s_mov_b32 s5, s6
; GFX7-NEXT: buffer_load_dword v3, v[0:1], s[4:7], 0 addr64
; GFX7-NEXT: v_and_b32_e32 v4, 3, v4
-; GFX7-NEXT: v_lshlrev_b32_e32 v4, 3, v4
-; GFX7-NEXT: v_lshl_b32_e32 v5, 0xffff, v4
+; GFX7-NEXT: v_lshlrev_b32_e32 v5, 3, v4
+; GFX7-NEXT: v_lshl_b32_e32 v4, 0xffff, v5
; GFX7-NEXT: v_mul_f32_e32 v2, 1.0, v2
-; GFX7-NEXT: v_not_b32_e32 v5, v5
+; GFX7-NEXT: v_not_b32_e32 v6, v4
; GFX7-NEXT: s_mov_b64 s[8:9], 0
-; GFX7-NEXT: v_and_b32_e32 v6, 0xffff0000, v2
+; GFX7-NEXT: v_and_b32_e32 v2, 0xffff0000, v2
; GFX7-NEXT: .LBB41_1: ; %atomicrmw.start
; GFX7-NEXT: ; =>This Inner Loop Header: Depth=1
; GFX7-NEXT: s_waitcnt vmcnt(0)
-; GFX7-NEXT: v_lshrrev_b32_e32 v2, v4, v3
-; GFX7-NEXT: v_lshlrev_b32_e32 v2, 16, v2
-; GFX7-NEXT: v_mul_f32_e32 v2, 1.0, v2
-; GFX7-NEXT: v_max_f32_e32 v2, v2, v6
-; GFX7-NEXT: v_lshrrev_b32_e32 v2, 16, v2
-; GFX7-NEXT: v_and_b32_e32 v7, v3, v5
-; GFX7-NEXT: v_lshlrev_b32_e32 v2, v4, v2
-; GFX7-NEXT: v_or_b32_e32 v2, v7, v2
; GFX7-NEXT: v_mov_b32_e32 v8, v3
-; GFX7-NEXT: v_mov_b32_e32 v7, v2
-; GFX7-NEXT: buffer_atomic_cmpswap v[7:8], v[0:1], s[4:7], 0 addr64 glc
+; GFX7-NEXT: v_lshrrev_b32_e32 v3, v5, v8
+; GFX7-NEXT: v_lshlrev_b32_e32 v3, 16, v3
+; GFX7-NEXT: v_mul_f32_e32 v3, 1.0, v3
+; GFX7-NEXT: v_max_f32_e32 v3, v3, v2
+; GFX7-NEXT: v_lshrrev_b32_e32 v3, 16, v3
+; GFX7-NEXT: v_and_b32_e32 v4, v8, v6
+; GFX7-NEXT: v_lshlrev_b32_e32 v3, v5, v3
+; GFX7-NEXT: v_or_b32_e32 v7, v4, v3
+; GFX7-NEXT: v_mov_b32_e32 v3, v7
+; GFX7-NEXT: v_mov_b32_e32 v4, v8
+; GFX7-NEXT: buffer_atomic_cmpswap v[3:4], v[0:1], s[4:7], 0 addr64 glc
; GFX7-NEXT: s_waitcnt vmcnt(0)
; GFX7-NEXT: buffer_wbinvl1
-; GFX7-NEXT: v_cmp_eq_u32_e32 vcc, v7, v3
+; GFX7-NEXT: v_cmp_eq_u32_e32 vcc, v3, v8
; GFX7-NEXT: s_or_b64 s[8:9], vcc, s[8:9]
-; GFX7-NEXT: v_mov_b32_e32 v3, v7
; GFX7-NEXT: s_andn2_b64 exec, exec, s[8:9]
; GFX7-NEXT: s_cbranch_execnz .LBB41_1
; GFX7-NEXT: ; %bb.2: ; %atomicrmw.end
@@ -11714,32 +11716,32 @@ define void @global_agent_atomic_fmax_noret_bf16__offset12b_neg__amdgpu_no_fine_
; GFX6-NEXT: s_mov_b32 s5, s6
; GFX6-NEXT: buffer_load_dword v3, v[0:1], s[4:7], 0 addr64
; GFX6-NEXT: v_and_b32_e32 v4, 3, v4
-; GFX6-NEXT: v_lshlrev_b32_e32 v4, 3, v4
-; GFX6-NEXT: v_lshl_b32_e32 v5, 0xffff, v4
+; GFX6-NEXT: v_lshlrev_b32_e32 v5, 3, v4
+; GFX6-NEXT: v_lshl_b32_e32 v4, 0xffff, v5
; GFX6-NEXT: v_mul_f32_e32 v2, 1.0, v2
-; GFX6-NEXT: v_not_b32_e32 v5, v5
+; GFX6-NEXT: v_not_b32_e32 v6, v4
; GFX6-NEXT: s_mov_b64 s[8:9], 0
-; GFX6-NEXT: v_and_b32_e32 v6, 0xffff0000, v2
+; GFX6-NEXT: v_and_b32_e32 v2, 0xffff0000, v2
; GFX6-NEXT: .LBB41_1: ; %atomicrmw.start
; GFX6-NEXT: ; =>This Inner Loop Header: Depth=1
; GFX6-NEXT: s_waitcnt vmcnt(0)
-; GFX6-NEXT: v_lshrrev_b32_e32 v2, v4, v3
-; GFX6-NEXT: v_lshlrev_b32_e32 v2, 16, v2
-; GFX6-NEXT: v_mul_f32_e32 v2, 1.0, v2
-; GFX6-NEXT: v_max_f32_e32 v2, v2, v6
-; GFX6-NEXT: v_lshrrev_b32_e32 v2, 16, v2
-; GFX6-NEXT: s_waitcnt expcnt(0)
-; GFX6-NEXT: v_and_b32_e32 v7, v3, v5
-; GFX6-NEXT: v_lshlrev_b32_e32 v2, v4, v2
-; GFX6-NEXT: v_or_b32_e32 v2, v7, v2
; GFX6-NEXT: v_mov_b32_e32 v8, v3
-; GFX6-NEXT: v_mov_b32_e32 v7, v2
-; GFX6-NEXT: buffer_atomic_cmpswap v[7:8], v[0:1], s[4:7], 0 addr64 glc
+; GFX6-NEXT: s_waitcnt expcnt(0)
+; GFX6-NEXT: v_lshrrev_b32_e32 v3, v5, v8
+; GFX6-NEXT: v_lshlrev_b32_e32 v3, 16, v3
+; GFX6-NEXT: v_mul_f32_e32 v3, 1.0, v3
+; GFX6-NEXT: v_max_f32_e32 v3, v3, v2
+; GFX6-NEXT: v_lshrrev_b32_e32 v3, 16, v3
+; GFX6-NEXT: v_and_b32_e32 v4, v8, v6
+; GFX6-NEXT: v_lshlrev_b32_e32 v3, v5, v3
+; GFX6-NEXT: v_or_b32_e32 v7, v4, v3
+; GFX6-NEXT: v_mov_b32_e32 v3, v7
+; GFX6-NEXT: v_mov_b32_e32 v4, v8
+; GFX6-NEXT: buffer_atomic_cmpswap v[3:4], v[0:1], s[4:7], 0 addr64 glc
; GFX6-NEXT: s_waitcnt vmcnt(0)
; GFX6-NEXT: buffer_wbinvl1
-; GFX6-NEXT: v_cmp_eq_u32_e32 vcc, v7, v3
+; GFX6-NEXT: v_cmp_eq_u32_e32 vcc, v3, v8
; GFX6-NEXT: s_or_b64 s[8:9], vcc, s[8:9]
-; GFX6-NEXT: v_mov_b32_e32 v3, v7
; GFX6-NEXT: s_andn2_b64 exec, exec, s[8:9]
; GFX6-NEXT: s_cbranch_execnz .LBB41_1
; GFX6-NEXT: ; %bb.2: ; %atomicrmw.end
@@ -12097,29 +12099,29 @@ define bfloat @global_agent_atomic_fmax_ret_bf16__offset12b_pos__align4__amdgpu_
; GFX7-NEXT: buffer_load_dword v3, v[0:1], s[4:7], 0 addr64 offset:2046
; GFX7-NEXT: v_mul_f32_e32 v2, 1.0, v2
; GFX7-NEXT: s_mov_b64 s[8:9], 0
-; GFX7-NEXT: v_and_b32_e32 v6, 0xffff0000, v2
+; GFX7-NEXT: v_and_b32_e32 v2, 0xffff0000, v2
; GFX7-NEXT: .LBB42_1: ; %atomicrmw.start
; GFX7-NEXT: ; =>This Inner Loop Header: Depth=1
; GFX7-NEXT: s_waitcnt vmcnt(0)
-; GFX7-NEXT: v_lshlrev_b32_e32 v2, 16, v3
-; GFX7-NEXT: v_mul_f32_e32 v2, 1.0, v2
-; GFX7-NEXT: v_max_f32_e32 v2, v2, v6
-; GFX7-NEXT: v_and_b32_e32 v4, 0xffff0000, v3
-; GFX7-NEXT: v_lshrrev_b32_e32 v2, 16, v2
-; GFX7-NEXT: v_or_b32_e32 v2, v4, v2
; GFX7-NEXT: v_mov_b32_e32 v5, v3
-; GFX7-NEXT: v_mov_b32_e32 v4, v2
-; GFX7-NEXT: buffer_atomic_cmpswap v[4:5], v[0:1], s[4:7], 0 addr64 offset:2046 glc
+; GFX7-NEXT: v_lshlrev_b32_e32 v3, 16, v5
+; GFX7-NEXT: v_mul_f32_e32 v3, 1.0, v3
+; GFX7-NEXT: v_max_f32_e32 v3, v3, v2
+; GFX7-NEXT: v_and_b32_e32 v4, 0xffff0000, v5
+; GFX7-NEXT: v_lshrrev_b32_e32 v3, 16, v3
+; GFX7-NEXT: v_or_b32_e32 v4, v4, v3
+; GFX7-NEXT: v_mov_b32_e32 v3, v4
+; GFX7-NEXT: v_mov_b32_e32 v4, v5
+; GFX7-NEXT: buffer_atomic_cmpswap v[3:4], v[0:1], s[4:7], 0 addr64 offset:2046 glc
; GFX7-NEXT: s_waitcnt vmcnt(0)
; GFX7-NEXT: buffer_wbinvl1
-; GFX7-NEXT: v_cmp_eq_u32_e32 vcc, v4, v3
+; GFX7-NEXT: v_cmp_eq_u32_e32 vcc, v3, v5
; GFX7-NEXT: s_or_b64 s[8:9], vcc, s[8:9]
-; GFX7-NEXT: v_mov_b32_e32 v3, v4
; GFX7-NEXT: s_andn2_b64 exec, exec, s[8:9]
; GFX7-NEXT: s_cbranch_execnz .LBB42_1
; GFX7-NEXT: ; %bb.2: ; %atomicrmw.end
; GFX7-NEXT: s_or_b64 exec, exec, s[8:9]
-; GFX7-NEXT: v_lshlrev_b32_e32 v0, 16, v4
+; GFX7-NEXT: v_lshlrev_b32_e32 v0, 16, v3
; GFX7-NEXT: s_setpc_b64 s[30:31]
;
; GFX6-LABEL: global_agent_atomic_fmax_ret_bf16__offset12b_pos__align4__amdgpu_no_fine_grained_memory:
@@ -12132,30 +12134,30 @@ define bfloat @global_agent_atomic_fmax_ret_bf16__offset12b_pos__align4__amdgpu_
; GFX6-NEXT: buffer_load_dword v3, v[0:1], s[4:7], 0 addr64 offset:2046
; GFX6-NEXT: v_mul_f32_e32 v2, 1.0, v2
; GFX6-NEXT: s_mov_b64 s[8:9], 0
-; GFX6-NEXT: v_and_b32_e32 v6, 0xffff0000, v2
+; GFX6-NEXT: v_and_b32_e32 v2, 0xffff0000, v2
; GFX6-NEXT: .LBB42_1: ; %atomicrmw.start
; GFX6-NEXT: ; =>This Inner Loop Header: Depth=1
; GFX6-NEXT: s_waitcnt vmcnt(0)
-; GFX6-NEXT: v_lshlrev_b32_e32 v2, 16, v3
-; GFX6-NEXT: v_mul_f32_e32 v2, 1.0, v2
-; GFX6-NEXT: v_max_f32_e32 v2, v2, v6
-; GFX6-NEXT: s_waitcnt expcnt(0)
-; GFX6-NEXT: v_and_b32_e32 v4, 0xffff0000, v3
-; GFX6-NEXT: v_lshrrev_b32_e32 v2, 16, v2
-; GFX6-NEXT: v_or_b32_e32 v2, v4, v2
; GFX6-NEXT: v_mov_b32_e32 v5, v3
-; GFX6-NEXT: v_mov_b32_e32 v4, v2
-; GFX6-NEXT: buffer_atomic_cmpswap v[4:5], v[0:1], s[4:7], 0 addr64 offset:2046 glc
+; GFX6-NEXT: s_waitcnt expcnt(0)
+; GFX6-NEXT: v_lshlrev_b32_e32 v3, 16, v5
+; GFX6-NEXT: v_mul_f32_e32 v3, 1.0, v3
+; GFX6-NEXT: v_max_f32_e32 v3, v3, v2
+; GFX6-NEXT: v_and_b32_e32 v4, 0xffff0000, v5
+; GFX6-NEXT: v_lshrrev_b32_e32 v3, 16, v3
+; GFX6-NEXT: v_or_b32_e32 v4, v4, v3
+; GFX6-NEXT: v_mov_b32_e32 v3, v4
+; GFX6-NEXT: v_mov_b32_e32 v4, v5
+; GFX6-NEXT: buffer_atomic_cmpswap v[3:4], v[0:1], s[4:7], 0 addr64 offset:2046 glc
; GFX6-NEXT: s_waitcnt vmcnt(0)
; GFX6-NEXT: buffer_wbinvl1
-; GFX6-NEXT: v_cmp_eq_u32_e32 vcc, v4, v3
+; GFX6-NEXT: v_cmp_eq_u32_e32 vcc, v3, v5
; GFX6-NEXT: s_or_b64 s[8:9], vcc, s[8:9]
-; GFX6-NEXT: v_mov_b32_e32 v3, v4
; GFX6-NEXT: s_andn2_b64 exec, exec, s[8:9]
; GFX6-NEXT: s_cbranch_execnz .LBB42_1
; GFX6-NEXT: ; %bb.2: ; %atomicrmw.end
; GFX6-NEXT: s_or_b64 exec, exec, s[8:9]
-; GFX6-NEXT: v_lshlrev_b32_e32 v0, 16, v4
+; GFX6-NEXT: v_lshlrev_b32_e32 v0, 16, v3
; GFX6-NEXT: s_waitcnt expcnt(0)
; GFX6-NEXT: s_setpc_b64 s[30:31]
%gep = getelementptr bfloat, ptr addrspace(1) %ptr, i64 1023
@@ -12497,24 +12499,24 @@ define void @global_agent_atomic_fmax_noret_bf16__offset12b__align4_pos__amdgpu_
; GFX7-NEXT: buffer_load_dword v3, v[0:1], s[4:7], 0 addr64 offset:2046
; GFX7-NEXT: v_mul_f32_e32 v2, 1.0, v2
; GFX7-NEXT: s_mov_b64 s[8:9], 0
-; GFX7-NEXT: v_and_b32_e32 v4, 0xffff0000, v2
+; GFX7-NEXT: v_and_b32_e32 v2, 0xffff0000, v2
; GFX7-NEXT: .LBB43_1: ; %atomicrmw.start
; GFX7-NEXT: ; =>This Inner Loop Header: Depth=1
; GFX7-NEXT: s_waitcnt vmcnt(0)
-; GFX7-NEXT: v_lshlrev_b32_e32 v2, 16, v3
-; GFX7-NEXT: v_mul_f32_e32 v2, 1.0, v2
-; GFX7-NEXT: v_max_f32_e32 v2, v2, v4
-; GFX7-NEXT: v_and_b32_e32 v5, 0xffff0000, v3
-; GFX7-NEXT: v_lshrrev_b32_e32 v2, 16, v2
-; GFX7-NEXT: v_or_b32_e32 v2, v5, v2
-; GFX7-NEXT: v_mov_b32_e32 v6, v3
-; GFX7-NEXT: v_mov_b32_e32 v5, v2
-; GFX7-NEXT: buffer_atomic_cmpswap v[5:6], v[0:1], s[4:7], 0 addr64 offset:2046 glc
+; GFX7-NEXT: v_mov_b32_e32 v5, v3
+; GFX7-NEXT: v_lshlrev_b32_e32 v3, 16, v5
+; GFX7-NEXT: v_mul_f32_e32 v3, 1.0, v3
+; GFX7-NEXT: v_max_f32_e32 v3, v3, v2
+; GFX7-NEXT: v_and_b32_e32 v4, 0xffff0000, v5
+; GFX7-NEXT: v_lshrrev_b32_e32 v3, 16, v3
+; GFX7-NEXT: v_or_b32_e32 v4, v4, v3
+; GFX7-NEXT: v_mov_b32_e32 v3, v4
+; GFX7-NEXT: v_mov_b32_e32 v4, v5
+; GFX7-NEXT: buffer_atomic_cmpswap v[3:4], v[0:1], s[4:7], 0 addr64 offset:2046 glc
; GFX7-NEXT: s_waitcnt vmcnt(0)
; GFX7-NEXT: buffer_wbinvl1
-; GFX7-NEXT: v_cmp_eq_u32_e32 vcc, v5, v3
+; GFX7-NEXT: v_cmp_eq_u32_e32 vcc, v3, v5
; GFX7-NEXT: s_or_b64 s[8:9], vcc, s[8:9]
-; GFX7-NEXT: v_mov_b32_e32 v3, v5
; GFX7-NEXT: s_andn2_b64 exec, exec, s[8:9]
; GFX7-NEXT: s_cbranch_execnz .LBB43_1
; GFX7-NEXT: ; %bb.2: ; %atomicrmw.end
@@ -12531,25 +12533,25 @@ define void @global_agent_atomic_fmax_noret_bf16__offset12b__align4_pos__amdgpu_
; GFX6-NEXT: buffer_load_dword v3, v[0:1], s[4:7], 0 addr64 offset:2046
; GFX6-NEXT: v_mul_f32_e32 v2, 1.0, v2
; GFX6-NEXT: s_mov_b64 s[8:9], 0
-; GFX6-NEXT: v_and_b32_e32 v4, 0xffff0000, v2
+; GFX6-NEXT: v_and_b32_e32 v2, 0xffff0000, v2
; GFX6-NEXT: .LBB43_1: ; %atomicrmw.start
; GFX6-NEXT: ; =>This Inner Loop Header: Depth=1
; GFX6-NEXT: s_waitcnt vmcnt(0)
-; GFX6-NEXT: v_lshlrev_b32_e32 v2, 16, v3
-; GFX6-NEXT: v_mul_f32_e32 v2, 1.0, v2
-; GFX6-NEXT: v_max_f32_e32 v2, v2, v4
+; GFX6-NEXT: v_mov_b32_e32 v5, v3
; GFX6-NEXT: s_waitcnt expcnt(0)
-; GFX6-NEXT: v_and_b32_e32 v5, 0xffff0000, v3
-; GFX6-NEXT: v_lshrrev_b32_e32 v2, 16, v2
-; GFX6-NEXT: v_or_b32_e32 v2, v5, v2
-; GFX6-NEXT: v_mov_b32_e32 v6, v3
-; GFX6-NEXT: v_mov_b32_e32 v5, v2
-; GFX6-NEXT: buffer_atomic_cmpswap v[5:6], v[0:1], s[4:7], 0 addr64 offset:2046 glc
+; GFX6-NEXT: v_lshlrev_b32_e32 v3, 16, v5
+; GFX6-NEXT: v_mul_f32_e32 v3, 1.0, v3
+; GFX6-NEXT: v_max_f32_e32 v3, v3, v2
+; GFX6-NEXT: v_and_b32_e32 v4, 0xffff0000, v5
+; GFX6-NEXT: v_lshrrev_b32_e32 v3, 16, v3
+; GFX6-NEXT: v_or_b32_e32 v4, v4, v3
+; GFX6-NEXT: v_mov_b32_e32 v3, v4
+; GFX6-NEXT: v_mov_b32_e32 v4, v5
+; GFX6-NEXT: buffer_atomic_cmpswap v[3:4], v[0:1], s[4:7], 0 addr64 offset:2046 glc
; GFX6-NEXT: s_waitcnt vmcnt(0)
; GFX6-NEXT: buffer_wbinvl1
-; GFX6-NEXT: v_cmp_eq_u32_e32 vcc, v5, v3
+; GFX6-NEXT: v_cmp_eq_u32_e32 vcc, v3, v5
; GFX6-NEXT: s_or_b64 s[8:9], vcc, s[8:9]
-; GFX6-NEXT: v_mov_b32_e32 v3, v5
; GFX6-NEXT: s_andn2_b64 exec, exec, s[8:9]
; GFX6-NEXT: s_cbranch_execnz .LBB43_1
; GFX6-NEXT: ; %bb.2: ; %atomicrmw.end
@@ -12999,36 +13001,36 @@ define bfloat @global_system_atomic_fmax_ret_bf16__offset12b_pos__amdgpu_no_fine
; GFX7-NEXT: s_mov_b32 s5, s6
; GFX7-NEXT: buffer_load_dword v3, v[0:1], s[4:7], 0 addr64
; GFX7-NEXT: v_and_b32_e32 v4, 3, v4
-; GFX7-NEXT: v_lshlrev_b32_e32 v6, 3, v4
-; GFX7-NEXT: v_lshl_b32_e32 v4, 0xffff, v6
+; GFX7-NEXT: v_lshlrev_b32_e32 v5, 3, v4
+; GFX7-NEXT: v_lshl_b32_e32 v4, 0xffff, v5
; GFX7-NEXT: v_mul_f32_e32 v2, 1.0, v2
-; GFX7-NEXT: v_not_b32_e32 v7, v4
+; GFX7-NEXT: v_not_b32_e32 v6, v4
; GFX7-NEXT: s_mov_b64 s[8:9], 0
-; GFX7-NEXT: v_and_b32_e32 v8, 0xffff0000, v2
+; GFX7-NEXT: v_and_b32_e32 v2, 0xffff0000, v2
; GFX7-NEXT: .LBB44_1: ; %atomicrmw.start
; GFX7-NEXT: ; =>This Inner Loop Header: Depth=1
; GFX7-NEXT: s_waitcnt vmcnt(0)
-; GFX7-NEXT: v_lshrrev_b32_e32 v2, v6, v3
-; GFX7-NEXT: v_lshlrev_b32_e32 v2, 16, v2
-; GFX7-NEXT: v_mul_f32_e32 v2, 1.0, v2
-; GFX7-NEXT: v_max_f32_e32 v2, v2, v8
-; GFX7-NEXT: v_lshrrev_b32_e32 v2, 16, v2
-; GFX7-NEXT: v_and_b32_e32 v4, v3, v7
-; GFX7-NEXT: v_lshlrev_b32_e32 v2, v6, v2
-; GFX7-NEXT: v_or_b32_e32 v2, v4, v2
-; GFX7-NEXT: v_mov_b32_e32 v5, v3
-; GFX7-NEXT: v_mov_b32_e32 v4, v2
-; GFX7-NEXT: buffer_atomic_cmpswap v[4:5], v[0:1], s[4:7], 0 addr64 glc
+; GFX7-NEXT: v_mov_b32_e32 v8, v3
+; GFX7-NEXT: v_lshrrev_b32_e32 v3, v5, v8
+; GFX7-NEXT: v_lshlrev_b32_e32 v3, 16, v3
+; GFX7-NEXT: v_mul_f32_e32 v3, 1.0, v3
+; GFX7-NEXT: v_max_f32_e32 v3, v3, v2
+; GFX7-NEXT: v_lshrrev_b32_e32 v3, 16, v3
+; GFX7-NEXT: v_and_b32_e32 v4, v8, v6
+; GFX7-NEXT: v_lshlrev_b32_e32 v3, v5, v3
+; GFX7-NEXT: v_or_b32_e32 v7, v4, v3
+; GFX7-NEXT: v_mov_b32_e32 v3, v7
+; GFX7-NEXT: v_mov_b32_e32 v4, v8
+; GFX7-NEXT: buffer_atomic_cmpswap v[3:4], v[0:1], s[4:7], 0 addr64 glc
; GFX7-NEXT: s_waitcnt vmcnt(0)
; GFX7-NEXT: buffer_wbinvl1
-; GFX7-NEXT: v_cmp_eq_u32_e32 vcc, v4, v3
+; GFX7-NEXT: v_cmp_eq_u32_e32 vcc, v3, v8
; GFX7-NEXT: s_or_b64 s[8:9], vcc, s[8:9]
-; GFX7-NEXT: v_mov_b32_e32 v3, v4
; GFX7-NEXT: s_andn2_b64 exec, exec, s[8:9]
; GFX7-NEXT: s_cbranch_execnz .LBB44_1
; GFX7-NEXT: ; %bb.2: ; %atomicrmw.end
; GFX7-NEXT: s_or_b64 exec, exec, s[8:9]
-; GFX7-NEXT: v_lshrrev_b32_e32 v0, v6, v4
+; GFX7-NEXT: v_lshrrev_b32_e32 v0, v5, v3
; GFX7-NEXT: v_lshlrev_b32_e32 v0, 16, v0
; GFX7-NEXT: s_setpc_b64 s[30:31]
;
@@ -13044,37 +13046,37 @@ define bfloat @global_system_atomic_fmax_ret_bf16__offset12b_pos__amdgpu_no_fine
; GFX6-NEXT: s_mov_b32 s5, s6
; GFX6-NEXT: buffer_load_dword v3, v[0:1], s[4:7], 0 addr64
; GFX6-NEXT: v_and_b32_e32 v4, 3, v4
-; GFX6-NEXT: v_lshlrev_b32_e32 v6, 3, v4
-; GFX6-NEXT: v_lshl_b32_e32 v4, 0xffff, v6
+; GFX6-NEXT: v_lshlrev_b32_e32 v5, 3, v4
+; GFX6-NEXT: v_lshl_b32_e32 v4, 0xffff, v5
; GFX6-NEXT: v_mul_f32_e32 v2, 1.0, v2
-; GFX6-NEXT: v_not_b32_e32 v7, v4
+; GFX6-NEXT: v_not_b32_e32 v6, v4
; GFX6-NEXT: s_mov_b64 s[8:9], 0
-; GFX6-NEXT: v_and_b32_e32 v8, 0xffff0000, v2
+; GFX6-NEXT: v_and_b32_e32 v2, 0xffff0000, v2
; GFX6-NEXT: .LBB44_1: ; %atomicrmw.start
; GFX6-NEXT: ; =>This Inner Loop Header: Depth=1
; GFX6-NEXT: s_waitcnt vmcnt(0)
-; GFX6-NEXT: v_lshrrev_b32_e32 v2, v6, v3
-; GFX6-NEXT: v_lshlrev_b32_e32 v2, 16, v2
-; GFX6-NEXT: v_mul_f32_e32 v2, 1.0, v2
-; GFX6-NEXT: v_max_f32_e32 v2, v2, v8
-; GFX6-NEXT: v_lshrrev_b32_e32 v2, 16, v2
+; GFX6-NEXT: v_mov_b32_e32 v8, v3
; GFX6-NEXT: s_waitcnt expcnt(0)
-; GFX6-NEXT: v_and_b32_e32 v4, v3, v7
-; GFX6-NEXT: v_lshlrev_b32_e32 v2, v6, v2
-; GFX6-NEXT: v_or_b32_e32 v2, v4, v2
-; GFX6-NEXT: v_mov_b32_e32 v5, v3
-; GFX6-NEXT: v_mov_b32_e32 v4, v2
-; GFX6-NEXT: buffer_atomic_cmpswap v[4:5], v[0:1], s[4:7], 0 addr64 glc
+; GFX6-NEXT: v_lshrrev_b32_e32 v3, v5, v8
+; GFX6-NEXT: v_lshlrev_b32_e32 v3, 16, v3
+; GFX6-NEXT: v_mul_f32_e32 v3, 1.0, v3
+; GFX6-NEXT: v_max_f32_e32 v3, v3, v2
+; GFX6-NEXT: v_lshrrev_b32_e32 v3, 16, v3
+; GFX6-NEXT: v_and_b32_e32 v4, v8, v6
+; GFX6-NEXT: v_lshlrev_b32_e32 v3, v5, v3
+; GFX6-NEXT: v_or_b32_e32 v7, v4, v3
+; GFX6-NEXT: v_mov_b32_e32 v3, v7
+; GFX6-NEXT: v_mov_b32_e32 v4, v8
+; GFX6-NEXT: buffer_atomic_cmpswap v[3:4], v[0:1], s[4:7], 0 addr64 glc
; GFX6-NEXT: s_waitcnt vmcnt(0)
; GFX6-NEXT: buffer_wbinvl1
-; GFX6-NEXT: v_cmp_eq_u32_e32 vcc, v4, v3
+; GFX6-NEXT: v_cmp_eq_u32_e32 vcc, v3, v8
; GFX6-NEXT: s_or_b64 s[8:9], vcc, s[8:9]
-; GFX6-NEXT: v_mov_b32_e32 v3, v4
; GFX6-NEXT: s_andn2_b64 exec, exec, s[8:9]
; GFX6-NEXT: s_cbranch_execnz .LBB44_1
; GFX6-NEXT: ; %bb.2: ; %atomicrmw.end
; GFX6-NEXT: s_or_b64 exec, exec, s[8:9]
-; GFX6-NEXT: v_lshrrev_b32_e32 v0, v6, v4
+; GFX6-NEXT: v_lshrrev_b32_e32 v0, v5, v3
; GFX6-NEXT: v_lshlrev_b32_e32 v0, 16, v0
; GFX6-NEXT: s_waitcnt expcnt(0)
; GFX6-NEXT: s_setpc_b64 s[30:31]
@@ -13508,31 +13510,31 @@ define void @global_system_atomic_fmax_noret_bf16__offset12b_pos__amdgpu_no_fine
; GFX7-NEXT: s_mov_b32 s5, s6
; GFX7-NEXT: buffer_load_dword v3, v[0:1], s[4:7], 0 addr64
; GFX7-NEXT: v_and_b32_e32 v4, 3, v4
-; GFX7-NEXT: v_lshlrev_b32_e32 v4, 3, v4
-; GFX7-NEXT: v_lshl_b32_e32 v5, 0xffff, v4
+; GFX7-NEXT: v_lshlrev_b32_e32 v5, 3, v4
+; GFX7-NEXT: v_lshl_b32_e32 v4, 0xffff, v5
; GFX7-NEXT: v_mul_f32_e32 v2, 1.0, v2
-; GFX7-NEXT: v_not_b32_e32 v5, v5
+; GFX7-NEXT: v_not_b32_e32 v6, v4
; GFX7-NEXT: s_mov_b64 s[8:9], 0
-; GFX7-NEXT: v_and_b32_e32 v6, 0xffff0000, v2
+; GFX7-NEXT: v_and_b32_e32 v2, 0xffff0000, v2
; GFX7-NEXT: .LBB45_1: ; %atomicrmw.start
; GFX7-NEXT: ; =>This Inner Loop Header: Depth=1
; GFX7-NEXT: s_waitcnt vmcnt(0)
-; GFX7-NEXT: v_lshrrev_b32_e32 v2, v4, v3
-; GFX7-NEXT: v_lshlrev_b32_e32 v2, 16, v2
-; GFX7-NEXT: v_mul_f32_e32 v2, 1.0, v2
-; GFX7-NEXT: v_max_f32_e32 v2, v2, v6
-; GFX7-NEXT: v_lshrrev_b32_e32 v2, 16, v2
-; GFX7-NEXT: v_and_b32_e32 v7, v3, v5
-; GFX7-NEXT: v_lshlrev_b32_e32 v2, v4, v2
-; GFX7-NEXT: v_or_b32_e32 v2, v7, v2
; GFX7-NEXT: v_mov_b32_e32 v8, v3
-; GFX7-NEXT: v_mov_b32_e32 v7, v2
-; GFX7-NEXT: buffer_atomic_cmpswap v[7:8], v[0:1], s[4:7], 0 addr64 glc
+; GFX7-NEXT: v_lshrrev_b32_e32 v3, v5, v8
+; GFX7-NEXT: v_lshlrev_b32_e32 v3, 16, v3
+; GFX7-NEXT: v_mul_f32_e32 v3, 1.0, v3
+; GFX7-NEXT: v_max_f32_e32 v3, v3, v2
+; GFX7-NEXT: v_lshrrev_b32_e32 v3, 16, v3
+; GFX7-NEXT: v_and_b32_e32 v4, v8, v6
+; GFX7-NEXT: v_lshlrev_b32_e32 v3, v5, v3
+; GFX7-NEXT: v_or_b32_e32 v7, v4, v3
+; GFX7-NEXT: v_mov_b32_e32 v3, v7
+; GFX7-NEXT: v_mov_b32_e32 v4, v8
+; GFX7-NEXT: buffer_atomic_cmpswap v[3:4], v[0:1], s[4:7], 0 addr64 glc
; GFX7-NEXT: s_waitcnt vmcnt(0)
; GFX7-NEXT: buffer_wbinvl1
-; GFX7-NEXT: v_cmp_eq_u32_e32 vcc, v7, v3
+; GFX7-NEXT: v_cmp_eq_u32_e32 vcc, v3, v8
; GFX7-NEXT: s_or_b64 s[8:9], vcc, s[8:9]
-; GFX7-NEXT: v_mov_b32_e32 v3, v7
; GFX7-NEXT: s_andn2_b64 exec, exec, s[8:9]
; GFX7-NEXT: s_cbranch_execnz .LBB45_1
; GFX7-NEXT: ; %bb.2: ; %atomicrmw.end
@@ -13551,32 +13553,32 @@ define void @global_system_atomic_fmax_noret_bf16__offset12b_pos__amdgpu_no_fine
; GFX6-NEXT: s_mov_b32 s5, s6
; GFX6-NEXT: buffer_load_dword v3, v[0:1], s[4:7], 0 addr64
; GFX6-NEXT: v_and_b32_e32 v4, 3, v4
-; GFX6-NEXT: v_lshlrev_b32_e32 v4, 3, v4
-; GFX6-NEXT: v_lshl_b32_e32 v5, 0xffff, v4
+; GFX6-NEXT: v_lshlrev_b32_e32 v5, 3, v4
+; GFX6-NEXT: v_lshl_b32_e32 v4, 0xffff, v5
; GFX6-NEXT: v_mul_f32_e32 v2, 1.0, v2
-; GFX6-NEXT: v_not_b32_e32 v5, v5
+; GFX6-NEXT: v_not_b32_e32 v6, v4
; GFX6-NEXT: s_mov_b64 s[8:9], 0
-; GFX6-NEXT: v_and_b32_e32 v6, 0xffff0000, v2
+; GFX6-NEXT: v_and_b32_e32 v2, 0xffff0000, v2
; GFX6-NEXT: .LBB45_1: ; %atomicrmw.start
; GFX6-NEXT: ; =>This Inner Loop Header: Depth=1
; GFX6-NEXT: s_waitcnt vmcnt(0)
-; GFX6-NEXT: v_lshrrev_b32_e32 v2, v4, v3
-; GFX6-NEXT: v_lshlrev_b32_e32 v2, 16, v2
-; GFX6-NEXT: v_mul_f32_e32 v2, 1.0, v2
-; GFX6-NEXT: v_max_f32_e32 v2, v2, v6
-; GFX6-NEXT: v_lshrrev_b32_e32 v2, 16, v2
-; GFX6-NEXT: s_waitcnt expcnt(0)
-; GFX6-NEXT: v_and_b32_e32 v7, v3, v5
-; GFX6-NEXT: v_lshlrev_b32_e32 v2, v4, v2
-; GFX6-NEXT: v_or_b32_e32 v2, v7, v2
; GFX6-NEXT: v_mov_b32_e32 v8, v3
-; GFX6-NEXT: v_mov_b32_e32 v7, v2
-; GFX6-NEXT: buffer_atomic_cmpswap v[7:8], v[0:1], s[4:7], 0 addr64 glc
+; GFX6-NEXT: s_waitcnt expcnt(0)
+; GFX6-NEXT: v_lshrrev_b32_e32 v3, v5, v8
+; GFX6-NEXT: v_lshlrev_b32_e32 v3, 16, v3
+; GFX6-NEXT: v_mul_f32_e32 v3, 1.0, v3
+; GFX6-NEXT: v_max_f32_e32 v3, v3, v2
+; GFX6-NEXT: v_lshrrev_b32_e32 v3, 16, v3
+; GFX6-NEXT: v_and_b32_e32 v4, v8, v6
+; GFX6-NEXT: v_lshlrev_b32_e32 v3, v5, v3
+; GFX6-NEXT: v_or_b32_e32 v7, v4, v3
+; GFX6-NEXT: v_mov_b32_e32 v3, v7
+; GFX6-NEXT: v_mov_b32_e32 v4, v8
+; GFX6-NEXT: buffer_atomic_cmpswap v[3:4], v[0:1], s[4:7], 0 addr64 glc
; GFX6-NEXT: s_waitcnt vmcnt(0)
; GFX6-NEXT: buffer_wbinvl1
-; GFX6-NEXT: v_cmp_eq_u32_e32 vcc, v7, v3
+; GFX6-NEXT: v_cmp_eq_u32_e32 vcc, v3, v8
; GFX6-NEXT: s_or_b64 s[8:9], vcc, s[8:9]
-; GFX6-NEXT: v_mov_b32_e32 v3, v7
; GFX6-NEXT: s_andn2_b64 exec, exec, s[8:9]
; GFX6-NEXT: s_cbranch_execnz .LBB45_1
; GFX6-NEXT: ; %bb.2: ; %atomicrmw.end
diff --git a/llvm/test/CodeGen/AMDGPU/global-atomicrmw-fmin.ll b/llvm/test/CodeGen/AMDGPU/global-atomicrmw-fmin.ll
index 5834d4ab4d8e7..865da45dbedea 100644
--- a/llvm/test/CodeGen/AMDGPU/global-atomicrmw-fmin.ll
+++ b/llvm/test/CodeGen/AMDGPU/global-atomicrmw-fmin.ll
@@ -4806,36 +4806,36 @@ define half @global_agent_atomic_fmin_ret_f16__amdgpu_no_fine_grained_memory(ptr
; GFX7-NEXT: s_mov_b32 s4, s6
; GFX7-NEXT: s_mov_b32 s5, s6
; GFX7-NEXT: buffer_load_dword v4, v[0:1], s[4:7], 0 addr64
-; GFX7-NEXT: v_cvt_f16_f32_e32 v2, v2
-; GFX7-NEXT: v_and_b32_e32 v3, 3, v3
-; GFX7-NEXT: v_lshlrev_b32_e32 v5, 3, v3
+; GFX7-NEXT: v_cvt_f16_f32_e32 v5, v2
+; GFX7-NEXT: v_and_b32_e32 v2, 3, v3
+; GFX7-NEXT: v_lshlrev_b32_e32 v2, 3, v2
; GFX7-NEXT: s_mov_b64 s[8:9], 0
-; GFX7-NEXT: v_cvt_f32_f16_e32 v6, v2
-; GFX7-NEXT: v_lshl_b32_e32 v2, 0xffff, v5
-; GFX7-NEXT: v_not_b32_e32 v7, v2
+; GFX7-NEXT: v_cvt_f32_f16_e32 v3, v5
+; GFX7-NEXT: v_lshl_b32_e32 v5, 0xffff, v2
+; GFX7-NEXT: v_not_b32_e32 v6, v5
; GFX7-NEXT: .LBB26_1: ; %atomicrmw.start
; GFX7-NEXT: ; =>This Inner Loop Header: Depth=1
; GFX7-NEXT: s_waitcnt vmcnt(0)
-; GFX7-NEXT: v_lshrrev_b32_e32 v2, v5, v4
-; GFX7-NEXT: v_cvt_f32_f16_e32 v2, v2
-; GFX7-NEXT: v_and_b32_e32 v3, v4, v7
-; GFX7-NEXT: v_min_f32_e32 v2, v2, v6
-; GFX7-NEXT: v_cvt_f16_f32_e32 v2, v2
-; GFX7-NEXT: v_lshlrev_b32_e32 v2, v5, v2
-; GFX7-NEXT: v_or_b32_e32 v3, v3, v2
-; GFX7-NEXT: v_mov_b32_e32 v2, v3
-; GFX7-NEXT: v_mov_b32_e32 v3, v4
-; GFX7-NEXT: buffer_atomic_cmpswap v[2:3], v[0:1], s[4:7], 0 addr64 glc
+; GFX7-NEXT: v_mov_b32_e32 v8, v4
+; GFX7-NEXT: v_lshrrev_b32_e32 v4, v2, v8
+; GFX7-NEXT: v_cvt_f32_f16_e32 v4, v4
+; GFX7-NEXT: v_and_b32_e32 v5, v8, v6
+; GFX7-NEXT: v_min_f32_e32 v4, v4, v3
+; GFX7-NEXT: v_cvt_f16_f32_e32 v4, v4
+; GFX7-NEXT: v_lshlrev_b32_e32 v4, v2, v4
+; GFX7-NEXT: v_or_b32_e32 v7, v5, v4
+; GFX7-NEXT: v_mov_b32_e32 v4, v7
+; GFX7-NEXT: v_mov_b32_e32 v5, v8
+; GFX7-NEXT: buffer_atomic_cmpswap v[4:5], v[0:1], s[4:7], 0 addr64 glc
; GFX7-NEXT: s_waitcnt vmcnt(0)
; GFX7-NEXT: buffer_wbinvl1
-; GFX7-NEXT: v_cmp_eq_u32_e32 vcc, v2, v4
+; GFX7-NEXT: v_cmp_eq_u32_e32 vcc, v4, v8
; GFX7-NEXT: s_or_b64 s[8:9], vcc, s[8:9]
-; GFX7-NEXT: v_mov_b32_e32 v4, v2
; GFX7-NEXT: s_andn2_b64 exec, exec, s[8:9]
; GFX7-NEXT: s_cbranch_execnz .LBB26_1
; GFX7-NEXT: ; %bb.2: ; %atomicrmw.end
; GFX7-NEXT: s_or_b64 exec, exec, s[8:9]
-; GFX7-NEXT: v_lshrrev_b32_e32 v0, v5, v2
+; GFX7-NEXT: v_lshrrev_b32_e32 v0, v2, v4
; GFX7-NEXT: v_cvt_f32_f16_e32 v0, v0
; GFX7-NEXT: s_setpc_b64 s[30:31]
;
@@ -4849,36 +4849,37 @@ define half @global_agent_atomic_fmin_ret_f16__amdgpu_no_fine_grained_memory(ptr
; GFX6-NEXT: s_mov_b32 s4, s6
; GFX6-NEXT: s_mov_b32 s5, s6
; GFX6-NEXT: buffer_load_dword v4, v[0:1], s[4:7], 0 addr64
-; GFX6-NEXT: v_cvt_f16_f32_e32 v2, v2
-; GFX6-NEXT: v_and_b32_e32 v3, 3, v3
-; GFX6-NEXT: v_lshlrev_b32_e32 v5, 3, v3
+; GFX6-NEXT: v_cvt_f16_f32_e32 v5, v2
+; GFX6-NEXT: v_and_b32_e32 v2, 3, v3
+; GFX6-NEXT: v_lshlrev_b32_e32 v2, 3, v2
; GFX6-NEXT: s_mov_b64 s[8:9], 0
-; GFX6-NEXT: v_cvt_f32_f16_e32 v6, v2
-; GFX6-NEXT: v_lshl_b32_e32 v2, 0xffff, v5
-; GFX6-NEXT: v_not_b32_e32 v7, v2
+; GFX6-NEXT: v_cvt_f32_f16_e32 v3, v5
+; GFX6-NEXT: v_lshl_b32_e32 v5, 0xffff, v2
+; GFX6-NEXT: v_not_b32_e32 v6, v5
; GFX6-NEXT: .LBB26_1: ; %atomicrmw.start
; GFX6-NEXT: ; =>This Inner Loop Header: Depth=1
-; GFX6-NEXT: s_waitcnt vmcnt(0) expcnt(0)
-; GFX6-NEXT: v_lshrrev_b32_e32 v2, v5, v4
-; GFX6-NEXT: v_cvt_f32_f16_e32 v2, v2
-; GFX6-NEXT: v_and_b32_e32 v3, v4, v7
-; GFX6-NEXT: v_min_f32_e32 v2, v2, v6
-; GFX6-NEXT: v_cvt_f16_f32_e32 v2, v2
-; GFX6-NEXT: v_lshlrev_b32_e32 v2, v5, v2
-; GFX6-NEXT: v_or_b32_e32 v3, v3, v2
-; GFX6-NEXT: v_mov_b32_e32 v2, v3
-; GFX6-NEXT: v_mov_b32_e32 v3, v4
-; GFX6-NEXT: buffer_atomic_cmpswap v[2:3], v[0:1], s[4:7], 0 addr64 glc
+; GFX6-NEXT: s_waitcnt vmcnt(0)
+; GFX6-NEXT: v_mov_b32_e32 v8, v4
+; GFX6-NEXT: s_waitcnt expcnt(0)
+; GFX6-NEXT: v_lshrrev_b32_e32 v4, v2, v8
+; GFX6-NEXT: v_cvt_f32_f16_e32 v4, v4
+; GFX6-NEXT: v_and_b32_e32 v5, v8, v6
+; GFX6-NEXT: v_min_f32_e32 v4, v4, v3
+; GFX6-NEXT: v_cvt_f16_f32_e32 v4, v4
+; GFX6-NEXT: v_lshlrev_b32_e32 v4, v2, v4
+; GFX6-NEXT: v_or_b32_e32 v7, v5, v4
+; GFX6-NEXT: v_mov_b32_e32 v4, v7
+; GFX6-NEXT: v_mov_b32_e32 v5, v8
+; GFX6-NEXT: buffer_atomic_cmpswap v[4:5], v[0:1], s[4:7], 0 addr64 glc
; GFX6-NEXT: s_waitcnt vmcnt(0)
; GFX6-NEXT: buffer_wbinvl1
-; GFX6-NEXT: v_cmp_eq_u32_e32 vcc, v2, v4
+; GFX6-NEXT: v_cmp_eq_u32_e32 vcc, v4, v8
; GFX6-NEXT: s_or_b64 s[8:9], vcc, s[8:9]
-; GFX6-NEXT: v_mov_b32_e32 v4, v2
; GFX6-NEXT: s_andn2_b64 exec, exec, s[8:9]
; GFX6-NEXT: s_cbranch_execnz .LBB26_1
; GFX6-NEXT: ; %bb.2: ; %atomicrmw.end
; GFX6-NEXT: s_or_b64 exec, exec, s[8:9]
-; GFX6-NEXT: v_lshrrev_b32_e32 v0, v5, v2
+; GFX6-NEXT: v_lshrrev_b32_e32 v0, v2, v4
; GFX6-NEXT: v_cvt_f32_f16_e32 v0, v0
; GFX6-NEXT: s_waitcnt expcnt(0)
; GFX6-NEXT: s_setpc_b64 s[30:31]
@@ -5263,36 +5264,36 @@ define half @global_agent_atomic_fmin_ret_f16__offset12b_pos__amdgpu_no_fine_gra
; GFX7-NEXT: s_mov_b32 s4, s6
; GFX7-NEXT: s_mov_b32 s5, s6
; GFX7-NEXT: buffer_load_dword v3, v[0:1], s[4:7], 0 addr64
-; GFX7-NEXT: v_cvt_f16_f32_e32 v2, v2
-; GFX7-NEXT: v_and_b32_e32 v4, 3, v4
-; GFX7-NEXT: v_lshlrev_b32_e32 v6, 3, v4
+; GFX7-NEXT: v_cvt_f16_f32_e32 v5, v2
+; GFX7-NEXT: v_and_b32_e32 v2, 3, v4
+; GFX7-NEXT: v_lshlrev_b32_e32 v2, 3, v2
+; GFX7-NEXT: v_lshl_b32_e32 v4, 0xffff, v2
+; GFX7-NEXT: v_cvt_f32_f16_e32 v5, v5
+; GFX7-NEXT: v_not_b32_e32 v6, v4
; GFX7-NEXT: s_mov_b64 s[8:9], 0
-; GFX7-NEXT: v_cvt_f32_f16_e32 v7, v2
-; GFX7-NEXT: v_lshl_b32_e32 v2, 0xffff, v6
-; GFX7-NEXT: v_not_b32_e32 v8, v2
; GFX7-NEXT: .LBB27_1: ; %atomicrmw.start
; GFX7-NEXT: ; =>This Inner Loop Header: Depth=1
; GFX7-NEXT: s_waitcnt vmcnt(0)
-; GFX7-NEXT: v_lshrrev_b32_e32 v2, v6, v3
-; GFX7-NEXT: v_cvt_f32_f16_e32 v2, v2
-; GFX7-NEXT: v_and_b32_e32 v4, v3, v8
-; GFX7-NEXT: v_min_f32_e32 v2, v2, v7
-; GFX7-NEXT: v_cvt_f16_f32_e32 v2, v2
-; GFX7-NEXT: v_lshlrev_b32_e32 v2, v6, v2
-; GFX7-NEXT: v_or_b32_e32 v2, v4, v2
-; GFX7-NEXT: v_mov_b32_e32 v5, v3
-; GFX7-NEXT: v_mov_b32_e32 v4, v2
-; GFX7-NEXT: buffer_atomic_cmpswap v[4:5], v[0:1], s[4:7], 0 addr64 glc
+; GFX7-NEXT: v_mov_b32_e32 v8, v3
+; GFX7-NEXT: v_lshrrev_b32_e32 v3, v2, v8
+; GFX7-NEXT: v_cvt_f32_f16_e32 v3, v3
+; GFX7-NEXT: v_and_b32_e32 v4, v8, v6
+; GFX7-NEXT: v_min_f32_e32 v3, v3, v5
+; GFX7-NEXT: v_cvt_f16_f32_e32 v3, v3
+; GFX7-NEXT: v_lshlrev_b32_e32 v3, v2, v3
+; GFX7-NEXT: v_or_b32_e32 v7, v4, v3
+; GFX7-NEXT: v_mov_b32_e32 v3, v7
+; GFX7-NEXT: v_mov_b32_e32 v4, v8
+; GFX7-NEXT: buffer_atomic_cmpswap v[3:4], v[0:1], s[4:7], 0 addr64 glc
; GFX7-NEXT: s_waitcnt vmcnt(0)
; GFX7-NEXT: buffer_wbinvl1
-; GFX7-NEXT: v_cmp_eq_u32_e32 vcc, v4, v3
+; GFX7-NEXT: v_cmp_eq_u32_e32 vcc, v3, v8
; GFX7-NEXT: s_or_b64 s[8:9], vcc, s[8:9]
-; GFX7-NEXT: v_mov_b32_e32 v3, v4
; GFX7-NEXT: s_andn2_b64 exec, exec, s[8:9]
; GFX7-NEXT: s_cbranch_execnz .LBB27_1
; GFX7-NEXT: ; %bb.2: ; %atomicrmw.end
; GFX7-NEXT: s_or_b64 exec, exec, s[8:9]
-; GFX7-NEXT: v_lshrrev_b32_e32 v0, v6, v4
+; GFX7-NEXT: v_lshrrev_b32_e32 v0, v2, v3
; GFX7-NEXT: v_cvt_f32_f16_e32 v0, v0
; GFX7-NEXT: s_setpc_b64 s[30:31]
;
@@ -5307,37 +5308,37 @@ define half @global_agent_atomic_fmin_ret_f16__offset12b_pos__amdgpu_no_fine_gra
; GFX6-NEXT: s_mov_b32 s4, s6
; GFX6-NEXT: s_mov_b32 s5, s6
; GFX6-NEXT: buffer_load_dword v3, v[0:1], s[4:7], 0 addr64
-; GFX6-NEXT: v_cvt_f16_f32_e32 v2, v2
-; GFX6-NEXT: v_and_b32_e32 v4, 3, v4
-; GFX6-NEXT: v_lshlrev_b32_e32 v6, 3, v4
+; GFX6-NEXT: v_cvt_f16_f32_e32 v5, v2
+; GFX6-NEXT: v_and_b32_e32 v2, 3, v4
+; GFX6-NEXT: v_lshlrev_b32_e32 v2, 3, v2
+; GFX6-NEXT: v_lshl_b32_e32 v4, 0xffff, v2
+; GFX6-NEXT: v_cvt_f32_f16_e32 v5, v5
+; GFX6-NEXT: v_not_b32_e32 v6, v4
; GFX6-NEXT: s_mov_b64 s[8:9], 0
-; GFX6-NEXT: v_cvt_f32_f16_e32 v7, v2
-; GFX6-NEXT: v_lshl_b32_e32 v2, 0xffff, v6
-; GFX6-NEXT: v_not_b32_e32 v8, v2
; GFX6-NEXT: .LBB27_1: ; %atomicrmw.start
; GFX6-NEXT: ; =>This Inner Loop Header: Depth=1
; GFX6-NEXT: s_waitcnt vmcnt(0)
-; GFX6-NEXT: v_lshrrev_b32_e32 v2, v6, v3
-; GFX6-NEXT: v_cvt_f32_f16_e32 v2, v2
+; GFX6-NEXT: v_mov_b32_e32 v8, v3
; GFX6-NEXT: s_waitcnt expcnt(0)
-; GFX6-NEXT: v_and_b32_e32 v4, v3, v8
-; GFX6-NEXT: v_min_f32_e32 v2, v2, v7
-; GFX6-NEXT: v_cvt_f16_f32_e32 v2, v2
-; GFX6-NEXT: v_lshlrev_b32_e32 v2, v6, v2
-; GFX6-NEXT: v_or_b32_e32 v2, v4, v2
-; GFX6-NEXT: v_mov_b32_e32 v5, v3
-; GFX6-NEXT: v_mov_b32_e32 v4, v2
-; GFX6-NEXT: buffer_atomic_cmpswap v[4:5], v[0:1], s[4:7], 0 addr64 glc
+; GFX6-NEXT: v_lshrrev_b32_e32 v3, v2, v8
+; GFX6-NEXT: v_cvt_f32_f16_e32 v3, v3
+; GFX6-NEXT: v_and_b32_e32 v4, v8, v6
+; GFX6-NEXT: v_min_f32_e32 v3, v3, v5
+; GFX6-NEXT: v_cvt_f16_f32_e32 v3, v3
+; GFX6-NEXT: v_lshlrev_b32_e32 v3, v2, v3
+; GFX6-NEXT: v_or_b32_e32 v7, v4, v3
+; GFX6-NEXT: v_mov_b32_e32 v3, v7
+; GFX6-NEXT: v_mov_b32_e32 v4, v8
+; GFX6-NEXT: buffer_atomic_cmpswap v[3:4], v[0:1], s[4:7], 0 addr64 glc
; GFX6-NEXT: s_waitcnt vmcnt(0)
; GFX6-NEXT: buffer_wbinvl1
-; GFX6-NEXT: v_cmp_eq_u32_e32 vcc, v4, v3
+; GFX6-NEXT: v_cmp_eq_u32_e32 vcc, v3, v8
; GFX6-NEXT: s_or_b64 s[8:9], vcc, s[8:9]
-; GFX6-NEXT: v_mov_b32_e32 v3, v4
; GFX6-NEXT: s_andn2_b64 exec, exec, s[8:9]
; GFX6-NEXT: s_cbranch_execnz .LBB27_1
; GFX6-NEXT: ; %bb.2: ; %atomicrmw.end
; GFX6-NEXT: s_or_b64 exec, exec, s[8:9]
-; GFX6-NEXT: v_lshrrev_b32_e32 v0, v6, v4
+; GFX6-NEXT: v_lshrrev_b32_e32 v0, v2, v3
; GFX6-NEXT: v_cvt_f32_f16_e32 v0, v0
; GFX6-NEXT: s_waitcnt expcnt(0)
; GFX6-NEXT: s_setpc_b64 s[30:31]
@@ -5724,36 +5725,36 @@ define half @global_agent_atomic_fmin_ret_f16__offset12b_neg__amdgpu_no_fine_gra
; GFX7-NEXT: s_mov_b32 s4, s6
; GFX7-NEXT: s_mov_b32 s5, s6
; GFX7-NEXT: buffer_load_dword v3, v[0:1], s[4:7], 0 addr64
-; GFX7-NEXT: v_cvt_f16_f32_e32 v2, v2
-; GFX7-NEXT: v_and_b32_e32 v4, 3, v4
-; GFX7-NEXT: v_lshlrev_b32_e32 v6, 3, v4
+; GFX7-NEXT: v_cvt_f16_f32_e32 v5, v2
+; GFX7-NEXT: v_and_b32_e32 v2, 3, v4
+; GFX7-NEXT: v_lshlrev_b32_e32 v2, 3, v2
+; GFX7-NEXT: v_lshl_b32_e32 v4, 0xffff, v2
+; GFX7-NEXT: v_cvt_f32_f16_e32 v5, v5
+; GFX7-NEXT: v_not_b32_e32 v6, v4
; GFX7-NEXT: s_mov_b64 s[8:9], 0
-; GFX7-NEXT: v_cvt_f32_f16_e32 v7, v2
-; GFX7-NEXT: v_lshl_b32_e32 v2, 0xffff, v6
-; GFX7-NEXT: v_not_b32_e32 v8, v2
; GFX7-NEXT: .LBB28_1: ; %atomicrmw.start
; GFX7-NEXT: ; =>This Inner Loop Header: Depth=1
; GFX7-NEXT: s_waitcnt vmcnt(0)
-; GFX7-NEXT: v_lshrrev_b32_e32 v2, v6, v3
-; GFX7-NEXT: v_cvt_f32_f16_e32 v2, v2
-; GFX7-NEXT: v_and_b32_e32 v4, v3, v8
-; GFX7-NEXT: v_min_f32_e32 v2, v2, v7
-; GFX7-NEXT: v_cvt_f16_f32_e32 v2, v2
-; GFX7-NEXT: v_lshlrev_b32_e32 v2, v6, v2
-; GFX7-NEXT: v_or_b32_e32 v2, v4, v2
-; GFX7-NEXT: v_mov_b32_e32 v5, v3
-; GFX7-NEXT: v_mov_b32_e32 v4, v2
-; GFX7-NEXT: buffer_atomic_cmpswap v[4:5], v[0:1], s[4:7], 0 addr64 glc
+; GFX7-NEXT: v_mov_b32_e32 v8, v3
+; GFX7-NEXT: v_lshrrev_b32_e32 v3, v2, v8
+; GFX7-NEXT: v_cvt_f32_f16_e32 v3, v3
+; GFX7-NEXT: v_and_b32_e32 v4, v8, v6
+; GFX7-NEXT: v_min_f32_e32 v3, v3, v5
+; GFX7-NEXT: v_cvt_f16_f32_e32 v3, v3
+; GFX7-NEXT: v_lshlrev_b32_e32 v3, v2, v3
+; GFX7-NEXT: v_or_b32_e32 v7, v4, v3
+; GFX7-NEXT: v_mov_b32_e32 v3, v7
+; GFX7-NEXT: v_mov_b32_e32 v4, v8
+; GFX7-NEXT: buffer_atomic_cmpswap v[3:4], v[0:1], s[4:7], 0 addr64 glc
; GFX7-NEXT: s_waitcnt vmcnt(0)
; GFX7-NEXT: buffer_wbinvl1
-; GFX7-NEXT: v_cmp_eq_u32_e32 vcc, v4, v3
+; GFX7-NEXT: v_cmp_eq_u32_e32 vcc, v3, v8
; GFX7-NEXT: s_or_b64 s[8:9], vcc, s[8:9]
-; GFX7-NEXT: v_mov_b32_e32 v3, v4
; GFX7-NEXT: s_andn2_b64 exec, exec, s[8:9]
; GFX7-NEXT: s_cbranch_execnz .LBB28_1
; GFX7-NEXT: ; %bb.2: ; %atomicrmw.end
; GFX7-NEXT: s_or_b64 exec, exec, s[8:9]
-; GFX7-NEXT: v_lshrrev_b32_e32 v0, v6, v4
+; GFX7-NEXT: v_lshrrev_b32_e32 v0, v2, v3
; GFX7-NEXT: v_cvt_f32_f16_e32 v0, v0
; GFX7-NEXT: s_setpc_b64 s[30:31]
;
@@ -5768,37 +5769,37 @@ define half @global_agent_atomic_fmin_ret_f16__offset12b_neg__amdgpu_no_fine_gra
; GFX6-NEXT: s_mov_b32 s4, s6
; GFX6-NEXT: s_mov_b32 s5, s6
; GFX6-NEXT: buffer_load_dword v3, v[0:1], s[4:7], 0 addr64
-; GFX6-NEXT: v_cvt_f16_f32_e32 v2, v2
-; GFX6-NEXT: v_and_b32_e32 v4, 3, v4
-; GFX6-NEXT: v_lshlrev_b32_e32 v6, 3, v4
+; GFX6-NEXT: v_cvt_f16_f32_e32 v5, v2
+; GFX6-NEXT: v_and_b32_e32 v2, 3, v4
+; GFX6-NEXT: v_lshlrev_b32_e32 v2, 3, v2
+; GFX6-NEXT: v_lshl_b32_e32 v4, 0xffff, v2
+; GFX6-NEXT: v_cvt_f32_f16_e32 v5, v5
+; GFX6-NEXT: v_not_b32_e32 v6, v4
; GFX6-NEXT: s_mov_b64 s[8:9], 0
-; GFX6-NEXT: v_cvt_f32_f16_e32 v7, v2
-; GFX6-NEXT: v_lshl_b32_e32 v2, 0xffff, v6
-; GFX6-NEXT: v_not_b32_e32 v8, v2
; GFX6-NEXT: .LBB28_1: ; %atomicrmw.start
; GFX6-NEXT: ; =>This Inner Loop Header: Depth=1
; GFX6-NEXT: s_waitcnt vmcnt(0)
-; GFX6-NEXT: v_lshrrev_b32_e32 v2, v6, v3
-; GFX6-NEXT: v_cvt_f32_f16_e32 v2, v2
+; GFX6-NEXT: v_mov_b32_e32 v8, v3
; GFX6-NEXT: s_waitcnt expcnt(0)
-; GFX6-NEXT: v_and_b32_e32 v4, v3, v8
-; GFX6-NEXT: v_min_f32_e32 v2, v2, v7
-; GFX6-NEXT: v_cvt_f16_f32_e32 v2, v2
-; GFX6-NEXT: v_lshlrev_b32_e32 v2, v6, v2
-; GFX6-NEXT: v_or_b32_e32 v2, v4, v2
-; GFX6-NEXT: v_mov_b32_e32 v5, v3
-; GFX6-NEXT: v_mov_b32_e32 v4, v2
-; GFX6-NEXT: buffer_atomic_cmpswap v[4:5], v[0:1], s[4:7], 0 addr64 glc
+; GFX6-NEXT: v_lshrrev_b32_e32 v3, v2, v8
+; GFX6-NEXT: v_cvt_f32_f16_e32 v3, v3
+; GFX6-NEXT: v_and_b32_e32 v4, v8, v6
+; GFX6-NEXT: v_min_f32_e32 v3, v3, v5
+; GFX6-NEXT: v_cvt_f16_f32_e32 v3, v3
+; GFX6-NEXT: v_lshlrev_b32_e32 v3, v2, v3
+; GFX6-NEXT: v_or_b32_e32 v7, v4, v3
+; GFX6-NEXT: v_mov_b32_e32 v3, v7
+; GFX6-NEXT: v_mov_b32_e32 v4, v8
+; GFX6-NEXT: buffer_atomic_cmpswap v[3:4], v[0:1], s[4:7], 0 addr64 glc
; GFX6-NEXT: s_waitcnt vmcnt(0)
; GFX6-NEXT: buffer_wbinvl1
-; GFX6-NEXT: v_cmp_eq_u32_e32 vcc, v4, v3
+; GFX6-NEXT: v_cmp_eq_u32_e32 vcc, v3, v8
; GFX6-NEXT: s_or_b64 s[8:9], vcc, s[8:9]
-; GFX6-NEXT: v_mov_b32_e32 v3, v4
; GFX6-NEXT: s_andn2_b64 exec, exec, s[8:9]
; GFX6-NEXT: s_cbranch_execnz .LBB28_1
; GFX6-NEXT: ; %bb.2: ; %atomicrmw.end
; GFX6-NEXT: s_or_b64 exec, exec, s[8:9]
-; GFX6-NEXT: v_lshrrev_b32_e32 v0, v6, v4
+; GFX6-NEXT: v_lshrrev_b32_e32 v0, v2, v3
; GFX6-NEXT: v_cvt_f32_f16_e32 v0, v0
; GFX6-NEXT: s_waitcnt expcnt(0)
; GFX6-NEXT: s_setpc_b64 s[30:31]
@@ -6163,28 +6164,28 @@ define void @global_agent_atomic_fmin_noret_f16__amdgpu_no_fine_grained_memory(p
; GFX7-NEXT: v_cvt_f16_f32_e32 v5, v2
; GFX7-NEXT: v_and_b32_e32 v2, 3, v3
; GFX7-NEXT: v_lshlrev_b32_e32 v2, 3, v2
-; GFX7-NEXT: v_lshl_b32_e32 v3, 0xffff, v2
-; GFX7-NEXT: v_cvt_f32_f16_e32 v5, v5
-; GFX7-NEXT: v_not_b32_e32 v6, v3
; GFX7-NEXT: s_mov_b64 s[8:9], 0
+; GFX7-NEXT: v_cvt_f32_f16_e32 v3, v5
+; GFX7-NEXT: v_lshl_b32_e32 v5, 0xffff, v2
+; GFX7-NEXT: v_not_b32_e32 v6, v5
; GFX7-NEXT: .LBB29_1: ; %atomicrmw.start
; GFX7-NEXT: ; =>This Inner Loop Header: Depth=1
; GFX7-NEXT: s_waitcnt vmcnt(0)
-; GFX7-NEXT: v_lshrrev_b32_e32 v3, v2, v4
-; GFX7-NEXT: v_cvt_f32_f16_e32 v3, v3
-; GFX7-NEXT: v_and_b32_e32 v7, v4, v6
-; GFX7-NEXT: v_min_f32_e32 v3, v3, v5
-; GFX7-NEXT: v_cvt_f16_f32_e32 v3, v3
-; GFX7-NEXT: v_lshlrev_b32_e32 v3, v2, v3
-; GFX7-NEXT: v_or_b32_e32 v3, v7, v3
; GFX7-NEXT: v_mov_b32_e32 v8, v4
-; GFX7-NEXT: v_mov_b32_e32 v7, v3
-; GFX7-NEXT: buffer_atomic_cmpswap v[7:8], v[0:1], s[4:7], 0 addr64 glc
+; GFX7-NEXT: v_lshrrev_b32_e32 v4, v2, v8
+; GFX7-NEXT: v_cvt_f32_f16_e32 v4, v4
+; GFX7-NEXT: v_and_b32_e32 v5, v8, v6
+; GFX7-NEXT: v_min_f32_e32 v4, v4, v3
+; GFX7-NEXT: v_cvt_f16_f32_e32 v4, v4
+; GFX7-NEXT: v_lshlrev_b32_e32 v4, v2, v4
+; GFX7-NEXT: v_or_b32_e32 v7, v5, v4
+; GFX7-NEXT: v_mov_b32_e32 v4, v7
+; GFX7-NEXT: v_mov_b32_e32 v5, v8
+; GFX7-NEXT: buffer_atomic_cmpswap v[4:5], v[0:1], s[4:7], 0 addr64 glc
; GFX7-NEXT: s_waitcnt vmcnt(0)
; GFX7-NEXT: buffer_wbinvl1
-; GFX7-NEXT: v_cmp_eq_u32_e32 vcc, v7, v4
+; GFX7-NEXT: v_cmp_eq_u32_e32 vcc, v4, v8
; GFX7-NEXT: s_or_b64 s[8:9], vcc, s[8:9]
-; GFX7-NEXT: v_mov_b32_e32 v4, v7
; GFX7-NEXT: s_andn2_b64 exec, exec, s[8:9]
; GFX7-NEXT: s_cbranch_execnz .LBB29_1
; GFX7-NEXT: ; %bb.2: ; %atomicrmw.end
@@ -6204,29 +6205,29 @@ define void @global_agent_atomic_fmin_noret_f16__amdgpu_no_fine_grained_memory(p
; GFX6-NEXT: v_cvt_f16_f32_e32 v5, v2
; GFX6-NEXT: v_and_b32_e32 v2, 3, v3
; GFX6-NEXT: v_lshlrev_b32_e32 v2, 3, v2
-; GFX6-NEXT: v_lshl_b32_e32 v3, 0xffff, v2
-; GFX6-NEXT: v_cvt_f32_f16_e32 v5, v5
-; GFX6-NEXT: v_not_b32_e32 v6, v3
; GFX6-NEXT: s_mov_b64 s[8:9], 0
+; GFX6-NEXT: v_cvt_f32_f16_e32 v3, v5
+; GFX6-NEXT: v_lshl_b32_e32 v5, 0xffff, v2
+; GFX6-NEXT: v_not_b32_e32 v6, v5
; GFX6-NEXT: .LBB29_1: ; %atomicrmw.start
; GFX6-NEXT: ; =>This Inner Loop Header: Depth=1
; GFX6-NEXT: s_waitcnt vmcnt(0)
-; GFX6-NEXT: v_lshrrev_b32_e32 v3, v2, v4
-; GFX6-NEXT: v_cvt_f32_f16_e32 v3, v3
-; GFX6-NEXT: s_waitcnt expcnt(0)
-; GFX6-NEXT: v_and_b32_e32 v7, v4, v6
-; GFX6-NEXT: v_min_f32_e32 v3, v3, v5
-; GFX6-NEXT: v_cvt_f16_f32_e32 v3, v3
-; GFX6-NEXT: v_lshlrev_b32_e32 v3, v2, v3
-; GFX6-NEXT: v_or_b32_e32 v3, v7, v3
; GFX6-NEXT: v_mov_b32_e32 v8, v4
-; GFX6-NEXT: v_mov_b32_e32 v7, v3
-; GFX6-NEXT: buffer_atomic_cmpswap v[7:8], v[0:1], s[4:7], 0 addr64 glc
+; GFX6-NEXT: s_waitcnt expcnt(0)
+; GFX6-NEXT: v_lshrrev_b32_e32 v4, v2, v8
+; GFX6-NEXT: v_cvt_f32_f16_e32 v4, v4
+; GFX6-NEXT: v_and_b32_e32 v5, v8, v6
+; GFX6-NEXT: v_min_f32_e32 v4, v4, v3
+; GFX6-NEXT: v_cvt_f16_f32_e32 v4, v4
+; GFX6-NEXT: v_lshlrev_b32_e32 v4, v2, v4
+; GFX6-NEXT: v_or_b32_e32 v7, v5, v4
+; GFX6-NEXT: v_mov_b32_e32 v4, v7
+; GFX6-NEXT: v_mov_b32_e32 v5, v8
+; GFX6-NEXT: buffer_atomic_cmpswap v[4:5], v[0:1], s[4:7], 0 addr64 glc
; GFX6-NEXT: s_waitcnt vmcnt(0)
; GFX6-NEXT: buffer_wbinvl1
-; GFX6-NEXT: v_cmp_eq_u32_e32 vcc, v7, v4
+; GFX6-NEXT: v_cmp_eq_u32_e32 vcc, v4, v8
; GFX6-NEXT: s_or_b64 s[8:9], vcc, s[8:9]
-; GFX6-NEXT: v_mov_b32_e32 v4, v7
; GFX6-NEXT: s_andn2_b64 exec, exec, s[8:9]
; GFX6-NEXT: s_cbranch_execnz .LBB29_1
; GFX6-NEXT: ; %bb.2: ; %atomicrmw.end
@@ -6605,31 +6606,31 @@ define void @global_agent_atomic_fmin_noret_f16__offset12b_pos__amdgpu_no_fine_g
; GFX7-NEXT: s_mov_b32 s4, s6
; GFX7-NEXT: s_mov_b32 s5, s6
; GFX7-NEXT: buffer_load_dword v3, v[0:1], s[4:7], 0 addr64
-; GFX7-NEXT: v_cvt_f16_f32_e32 v2, v2
-; GFX7-NEXT: v_and_b32_e32 v4, 3, v4
-; GFX7-NEXT: v_lshlrev_b32_e32 v4, 3, v4
+; GFX7-NEXT: v_cvt_f16_f32_e32 v5, v2
+; GFX7-NEXT: v_and_b32_e32 v2, 3, v4
+; GFX7-NEXT: v_lshlrev_b32_e32 v2, 3, v2
+; GFX7-NEXT: v_lshl_b32_e32 v4, 0xffff, v2
+; GFX7-NEXT: v_cvt_f32_f16_e32 v5, v5
+; GFX7-NEXT: v_not_b32_e32 v6, v4
; GFX7-NEXT: s_mov_b64 s[8:9], 0
-; GFX7-NEXT: v_cvt_f32_f16_e32 v5, v2
-; GFX7-NEXT: v_lshl_b32_e32 v2, 0xffff, v4
-; GFX7-NEXT: v_not_b32_e32 v6, v2
; GFX7-NEXT: .LBB30_1: ; %atomicrmw.start
; GFX7-NEXT: ; =>This Inner Loop Header: Depth=1
; GFX7-NEXT: s_waitcnt vmcnt(0)
-; GFX7-NEXT: v_lshrrev_b32_e32 v2, v4, v3
-; GFX7-NEXT: v_cvt_f32_f16_e32 v2, v2
-; GFX7-NEXT: v_and_b32_e32 v7, v3, v6
-; GFX7-NEXT: v_min_f32_e32 v2, v2, v5
-; GFX7-NEXT: v_cvt_f16_f32_e32 v2, v2
-; GFX7-NEXT: v_lshlrev_b32_e32 v2, v4, v2
-; GFX7-NEXT: v_or_b32_e32 v2, v7, v2
; GFX7-NEXT: v_mov_b32_e32 v8, v3
-; GFX7-NEXT: v_mov_b32_e32 v7, v2
-; GFX7-NEXT: buffer_atomic_cmpswap v[7:8], v[0:1], s[4:7], 0 addr64 glc
+; GFX7-NEXT: v_lshrrev_b32_e32 v3, v2, v8
+; GFX7-NEXT: v_cvt_f32_f16_e32 v3, v3
+; GFX7-NEXT: v_and_b32_e32 v4, v8, v6
+; GFX7-NEXT: v_min_f32_e32 v3, v3, v5
+; GFX7-NEXT: v_cvt_f16_f32_e32 v3, v3
+; GFX7-NEXT: v_lshlrev_b32_e32 v3, v2, v3
+; GFX7-NEXT: v_or_b32_e32 v7, v4, v3
+; GFX7-NEXT: v_mov_b32_e32 v3, v7
+; GFX7-NEXT: v_mov_b32_e32 v4, v8
+; GFX7-NEXT: buffer_atomic_cmpswap v[3:4], v[0:1], s[4:7], 0 addr64 glc
; GFX7-NEXT: s_waitcnt vmcnt(0)
; GFX7-NEXT: buffer_wbinvl1
-; GFX7-NEXT: v_cmp_eq_u32_e32 vcc, v7, v3
+; GFX7-NEXT: v_cmp_eq_u32_e32 vcc, v3, v8
; GFX7-NEXT: s_or_b64 s[8:9], vcc, s[8:9]
-; GFX7-NEXT: v_mov_b32_e32 v3, v7
; GFX7-NEXT: s_andn2_b64 exec, exec, s[8:9]
; GFX7-NEXT: s_cbranch_execnz .LBB30_1
; GFX7-NEXT: ; %bb.2: ; %atomicrmw.end
@@ -6647,32 +6648,32 @@ define void @global_agent_atomic_fmin_noret_f16__offset12b_pos__amdgpu_no_fine_g
; GFX6-NEXT: s_mov_b32 s4, s6
; GFX6-NEXT: s_mov_b32 s5, s6
; GFX6-NEXT: buffer_load_dword v3, v[0:1], s[4:7], 0 addr64
-; GFX6-NEXT: v_cvt_f16_f32_e32 v2, v2
-; GFX6-NEXT: v_and_b32_e32 v4, 3, v4
-; GFX6-NEXT: v_lshlrev_b32_e32 v4, 3, v4
+; GFX6-NEXT: v_cvt_f16_f32_e32 v5, v2
+; GFX6-NEXT: v_and_b32_e32 v2, 3, v4
+; GFX6-NEXT: v_lshlrev_b32_e32 v2, 3, v2
+; GFX6-NEXT: v_lshl_b32_e32 v4, 0xffff, v2
+; GFX6-NEXT: v_cvt_f32_f16_e32 v5, v5
+; GFX6-NEXT: v_not_b32_e32 v6, v4
; GFX6-NEXT: s_mov_b64 s[8:9], 0
-; GFX6-NEXT: v_cvt_f32_f16_e32 v5, v2
-; GFX6-NEXT: v_lshl_b32_e32 v2, 0xffff, v4
-; GFX6-NEXT: v_not_b32_e32 v6, v2
; GFX6-NEXT: .LBB30_1: ; %atomicrmw.start
; GFX6-NEXT: ; =>This Inner Loop Header: Depth=1
; GFX6-NEXT: s_waitcnt vmcnt(0)
-; GFX6-NEXT: v_lshrrev_b32_e32 v2, v4, v3
-; GFX6-NEXT: v_cvt_f32_f16_e32 v2, v2
-; GFX6-NEXT: s_waitcnt expcnt(0)
-; GFX6-NEXT: v_and_b32_e32 v7, v3, v6
-; GFX6-NEXT: v_min_f32_e32 v2, v2, v5
-; GFX6-NEXT: v_cvt_f16_f32_e32 v2, v2
-; GFX6-NEXT: v_lshlrev_b32_e32 v2, v4, v2
-; GFX6-NEXT: v_or_b32_e32 v2, v7, v2
; GFX6-NEXT: v_mov_b32_e32 v8, v3
-; GFX6-NEXT: v_mov_b32_e32 v7, v2
-; GFX6-NEXT: buffer_atomic_cmpswap v[7:8], v[0:1], s[4:7], 0 addr64 glc
+; GFX6-NEXT: s_waitcnt expcnt(0)
+; GFX6-NEXT: v_lshrrev_b32_e32 v3, v2, v8
+; GFX6-NEXT: v_cvt_f32_f16_e32 v3, v3
+; GFX6-NEXT: v_and_b32_e32 v4, v8, v6
+; GFX6-NEXT: v_min_f32_e32 v3, v3, v5
+; GFX6-NEXT: v_cvt_f16_f32_e32 v3, v3
+; GFX6-NEXT: v_lshlrev_b32_e32 v3, v2, v3
+; GFX6-NEXT: v_or_b32_e32 v7, v4, v3
+; GFX6-NEXT: v_mov_b32_e32 v3, v7
+; GFX6-NEXT: v_mov_b32_e32 v4, v8
+; GFX6-NEXT: buffer_atomic_cmpswap v[3:4], v[0:1], s[4:7], 0 addr64 glc
; GFX6-NEXT: s_waitcnt vmcnt(0)
; GFX6-NEXT: buffer_wbinvl1
-; GFX6-NEXT: v_cmp_eq_u32_e32 vcc, v7, v3
+; GFX6-NEXT: v_cmp_eq_u32_e32 vcc, v3, v8
; GFX6-NEXT: s_or_b64 s[8:9], vcc, s[8:9]
-; GFX6-NEXT: v_mov_b32_e32 v3, v7
; GFX6-NEXT: s_andn2_b64 exec, exec, s[8:9]
; GFX6-NEXT: s_cbranch_execnz .LBB30_1
; GFX6-NEXT: ; %bb.2: ; %atomicrmw.end
@@ -7053,31 +7054,31 @@ define void @global_agent_atomic_fmin_noret_f16__offset12b_neg__amdgpu_no_fine_g
; GFX7-NEXT: s_mov_b32 s4, s6
; GFX7-NEXT: s_mov_b32 s5, s6
; GFX7-NEXT: buffer_load_dword v3, v[0:1], s[4:7], 0 addr64
-; GFX7-NEXT: v_cvt_f16_f32_e32 v2, v2
-; GFX7-NEXT: v_and_b32_e32 v4, 3, v4
-; GFX7-NEXT: v_lshlrev_b32_e32 v4, 3, v4
+; GFX7-NEXT: v_cvt_f16_f32_e32 v5, v2
+; GFX7-NEXT: v_and_b32_e32 v2, 3, v4
+; GFX7-NEXT: v_lshlrev_b32_e32 v2, 3, v2
+; GFX7-NEXT: v_lshl_b32_e32 v4, 0xffff, v2
+; GFX7-NEXT: v_cvt_f32_f16_e32 v5, v5
+; GFX7-NEXT: v_not_b32_e32 v6, v4
; GFX7-NEXT: s_mov_b64 s[8:9], 0
-; GFX7-NEXT: v_cvt_f32_f16_e32 v5, v2
-; GFX7-NEXT: v_lshl_b32_e32 v2, 0xffff, v4
-; GFX7-NEXT: v_not_b32_e32 v6, v2
; GFX7-NEXT: .LBB31_1: ; %atomicrmw.start
; GFX7-NEXT: ; =>This Inner Loop Header: Depth=1
; GFX7-NEXT: s_waitcnt vmcnt(0)
-; GFX7-NEXT: v_lshrrev_b32_e32 v2, v4, v3
-; GFX7-NEXT: v_cvt_f32_f16_e32 v2, v2
-; GFX7-NEXT: v_and_b32_e32 v7, v3, v6
-; GFX7-NEXT: v_min_f32_e32 v2, v2, v5
-; GFX7-NEXT: v_cvt_f16_f32_e32 v2, v2
-; GFX7-NEXT: v_lshlrev_b32_e32 v2, v4, v2
-; GFX7-NEXT: v_or_b32_e32 v2, v7, v2
; GFX7-NEXT: v_mov_b32_e32 v8, v3
-; GFX7-NEXT: v_mov_b32_e32 v7, v2
-; GFX7-NEXT: buffer_atomic_cmpswap v[7:8], v[0:1], s[4:7], 0 addr64 glc
+; GFX7-NEXT: v_lshrrev_b32_e32 v3, v2, v8
+; GFX7-NEXT: v_cvt_f32_f16_e32 v3, v3
+; GFX7-NEXT: v_and_b32_e32 v4, v8, v6
+; GFX7-NEXT: v_min_f32_e32 v3, v3, v5
+; GFX7-NEXT: v_cvt_f16_f32_e32 v3, v3
+; GFX7-NEXT: v_lshlrev_b32_e32 v3, v2, v3
+; GFX7-NEXT: v_or_b32_e32 v7, v4, v3
+; GFX7-NEXT: v_mov_b32_e32 v3, v7
+; GFX7-NEXT: v_mov_b32_e32 v4, v8
+; GFX7-NEXT: buffer_atomic_cmpswap v[3:4], v[0:1], s[4:7], 0 addr64 glc
; GFX7-NEXT: s_waitcnt vmcnt(0)
; GFX7-NEXT: buffer_wbinvl1
-; GFX7-NEXT: v_cmp_eq_u32_e32 vcc, v7, v3
+; GFX7-NEXT: v_cmp_eq_u32_e32 vcc, v3, v8
; GFX7-NEXT: s_or_b64 s[8:9], vcc, s[8:9]
-; GFX7-NEXT: v_mov_b32_e32 v3, v7
; GFX7-NEXT: s_andn2_b64 exec, exec, s[8:9]
; GFX7-NEXT: s_cbranch_execnz .LBB31_1
; GFX7-NEXT: ; %bb.2: ; %atomicrmw.end
@@ -7095,32 +7096,32 @@ define void @global_agent_atomic_fmin_noret_f16__offset12b_neg__amdgpu_no_fine_g
; GFX6-NEXT: s_mov_b32 s4, s6
; GFX6-NEXT: s_mov_b32 s5, s6
; GFX6-NEXT: buffer_load_dword v3, v[0:1], s[4:7], 0 addr64
-; GFX6-NEXT: v_cvt_f16_f32_e32 v2, v2
-; GFX6-NEXT: v_and_b32_e32 v4, 3, v4
-; GFX6-NEXT: v_lshlrev_b32_e32 v4, 3, v4
+; GFX6-NEXT: v_cvt_f16_f32_e32 v5, v2
+; GFX6-NEXT: v_and_b32_e32 v2, 3, v4
+; GFX6-NEXT: v_lshlrev_b32_e32 v2, 3, v2
+; GFX6-NEXT: v_lshl_b32_e32 v4, 0xffff, v2
+; GFX6-NEXT: v_cvt_f32_f16_e32 v5, v5
+; GFX6-NEXT: v_not_b32_e32 v6, v4
; GFX6-NEXT: s_mov_b64 s[8:9], 0
-; GFX6-NEXT: v_cvt_f32_f16_e32 v5, v2
-; GFX6-NEXT: v_lshl_b32_e32 v2, 0xffff, v4
-; GFX6-NEXT: v_not_b32_e32 v6, v2
; GFX6-NEXT: .LBB31_1: ; %atomicrmw.start
; GFX6-NEXT: ; =>This Inner Loop Header: Depth=1
; GFX6-NEXT: s_waitcnt vmcnt(0)
-; GFX6-NEXT: v_lshrrev_b32_e32 v2, v4, v3
-; GFX6-NEXT: v_cvt_f32_f16_e32 v2, v2
-; GFX6-NEXT: s_waitcnt expcnt(0)
-; GFX6-NEXT: v_and_b32_e32 v7, v3, v6
-; GFX6-NEXT: v_min_f32_e32 v2, v2, v5
-; GFX6-NEXT: v_cvt_f16_f32_e32 v2, v2
-; GFX6-NEXT: v_lshlrev_b32_e32 v2, v4, v2
-; GFX6-NEXT: v_or_b32_e32 v2, v7, v2
; GFX6-NEXT: v_mov_b32_e32 v8, v3
-; GFX6-NEXT: v_mov_b32_e32 v7, v2
-; GFX6-NEXT: buffer_atomic_cmpswap v[7:8], v[0:1], s[4:7], 0 addr64 glc
+; GFX6-NEXT: s_waitcnt expcnt(0)
+; GFX6-NEXT: v_lshrrev_b32_e32 v3, v2, v8
+; GFX6-NEXT: v_cvt_f32_f16_e32 v3, v3
+; GFX6-NEXT: v_and_b32_e32 v4, v8, v6
+; GFX6-NEXT: v_min_f32_e32 v3, v3, v5
+; GFX6-NEXT: v_cvt_f16_f32_e32 v3, v3
+; GFX6-NEXT: v_lshlrev_b32_e32 v3, v2, v3
+; GFX6-NEXT: v_or_b32_e32 v7, v4, v3
+; GFX6-NEXT: v_mov_b32_e32 v3, v7
+; GFX6-NEXT: v_mov_b32_e32 v4, v8
+; GFX6-NEXT: buffer_atomic_cmpswap v[3:4], v[0:1], s[4:7], 0 addr64 glc
; GFX6-NEXT: s_waitcnt vmcnt(0)
; GFX6-NEXT: buffer_wbinvl1
-; GFX6-NEXT: v_cmp_eq_u32_e32 vcc, v7, v3
+; GFX6-NEXT: v_cmp_eq_u32_e32 vcc, v3, v8
; GFX6-NEXT: s_or_b64 s[8:9], vcc, s[8:9]
-; GFX6-NEXT: v_mov_b32_e32 v3, v7
; GFX6-NEXT: s_andn2_b64 exec, exec, s[8:9]
; GFX6-NEXT: s_cbranch_execnz .LBB31_1
; GFX6-NEXT: ; %bb.2: ; %atomicrmw.end
@@ -7411,28 +7412,28 @@ define half @global_agent_atomic_fmin_ret_f16__offset12b_pos__align4__amdgpu_no_
; GFX7-NEXT: buffer_load_dword v3, v[0:1], s[4:7], 0 addr64 offset:2046
; GFX7-NEXT: v_cvt_f16_f32_e32 v2, v2
; GFX7-NEXT: s_mov_b64 s[8:9], 0
-; GFX7-NEXT: v_cvt_f32_f16_e32 v6, v2
+; GFX7-NEXT: v_cvt_f32_f16_e32 v2, v2
; GFX7-NEXT: .LBB32_1: ; %atomicrmw.start
; GFX7-NEXT: ; =>This Inner Loop Header: Depth=1
; GFX7-NEXT: s_waitcnt vmcnt(0)
-; GFX7-NEXT: v_cvt_f32_f16_e32 v2, v3
-; GFX7-NEXT: v_and_b32_e32 v4, 0xffff0000, v3
-; GFX7-NEXT: v_min_f32_e32 v2, v2, v6
-; GFX7-NEXT: v_cvt_f16_f32_e32 v2, v2
-; GFX7-NEXT: v_or_b32_e32 v2, v4, v2
; GFX7-NEXT: v_mov_b32_e32 v5, v3
-; GFX7-NEXT: v_mov_b32_e32 v4, v2
-; GFX7-NEXT: buffer_atomic_cmpswap v[4:5], v[0:1], s[4:7], 0 addr64 offset:2046 glc
+; GFX7-NEXT: v_cvt_f32_f16_e32 v3, v5
+; GFX7-NEXT: v_and_b32_e32 v4, 0xffff0000, v5
+; GFX7-NEXT: v_min_f32_e32 v3, v3, v2
+; GFX7-NEXT: v_cvt_f16_f32_e32 v3, v3
+; GFX7-NEXT: v_or_b32_e32 v4, v4, v3
+; GFX7-NEXT: v_mov_b32_e32 v3, v4
+; GFX7-NEXT: v_mov_b32_e32 v4, v5
+; GFX7-NEXT: buffer_atomic_cmpswap v[3:4], v[0:1], s[4:7], 0 addr64 offset:2046 glc
; GFX7-NEXT: s_waitcnt vmcnt(0)
; GFX7-NEXT: buffer_wbinvl1
-; GFX7-NEXT: v_cmp_eq_u32_e32 vcc, v4, v3
+; GFX7-NEXT: v_cmp_eq_u32_e32 vcc, v3, v5
; GFX7-NEXT: s_or_b64 s[8:9], vcc, s[8:9]
-; GFX7-NEXT: v_mov_b32_e32 v3, v4
; GFX7-NEXT: s_andn2_b64 exec, exec, s[8:9]
; GFX7-NEXT: s_cbranch_execnz .LBB32_1
; GFX7-NEXT: ; %bb.2: ; %atomicrmw.end
; GFX7-NEXT: s_or_b64 exec, exec, s[8:9]
-; GFX7-NEXT: v_cvt_f32_f16_e32 v0, v4
+; GFX7-NEXT: v_cvt_f32_f16_e32 v0, v3
; GFX7-NEXT: s_setpc_b64 s[30:31]
;
; GFX6-LABEL: global_agent_atomic_fmin_ret_f16__offset12b_pos__align4__amdgpu_no_fine_grained_memory:
@@ -7445,29 +7446,29 @@ define half @global_agent_atomic_fmin_ret_f16__offset12b_pos__align4__amdgpu_no_
; GFX6-NEXT: buffer_load_dword v3, v[0:1], s[4:7], 0 addr64 offset:2046
; GFX6-NEXT: v_cvt_f16_f32_e32 v2, v2
; GFX6-NEXT: s_mov_b64 s[8:9], 0
-; GFX6-NEXT: v_cvt_f32_f16_e32 v6, v2
+; GFX6-NEXT: v_cvt_f32_f16_e32 v2, v2
; GFX6-NEXT: .LBB32_1: ; %atomicrmw.start
; GFX6-NEXT: ; =>This Inner Loop Header: Depth=1
; GFX6-NEXT: s_waitcnt vmcnt(0)
-; GFX6-NEXT: v_cvt_f32_f16_e32 v2, v3
-; GFX6-NEXT: s_waitcnt expcnt(0)
-; GFX6-NEXT: v_and_b32_e32 v4, 0xffff0000, v3
-; GFX6-NEXT: v_min_f32_e32 v2, v2, v6
-; GFX6-NEXT: v_cvt_f16_f32_e32 v2, v2
-; GFX6-NEXT: v_or_b32_e32 v2, v4, v2
; GFX6-NEXT: v_mov_b32_e32 v5, v3
-; GFX6-NEXT: v_mov_b32_e32 v4, v2
-; GFX6-NEXT: buffer_atomic_cmpswap v[4:5], v[0:1], s[4:7], 0 addr64 offset:2046 glc
+; GFX6-NEXT: s_waitcnt expcnt(0)
+; GFX6-NEXT: v_cvt_f32_f16_e32 v3, v5
+; GFX6-NEXT: v_and_b32_e32 v4, 0xffff0000, v5
+; GFX6-NEXT: v_min_f32_e32 v3, v3, v2
+; GFX6-NEXT: v_cvt_f16_f32_e32 v3, v3
+; GFX6-NEXT: v_or_b32_e32 v4, v4, v3
+; GFX6-NEXT: v_mov_b32_e32 v3, v4
+; GFX6-NEXT: v_mov_b32_e32 v4, v5
+; GFX6-NEXT: buffer_atomic_cmpswap v[3:4], v[0:1], s[4:7], 0 addr64 offset:2046 glc
; GFX6-NEXT: s_waitcnt vmcnt(0)
; GFX6-NEXT: buffer_wbinvl1
-; GFX6-NEXT: v_cmp_eq_u32_e32 vcc, v4, v3
+; GFX6-NEXT: v_cmp_eq_u32_e32 vcc, v3, v5
; GFX6-NEXT: s_or_b64 s[8:9], vcc, s[8:9]
-; GFX6-NEXT: v_mov_b32_e32 v3, v4
; GFX6-NEXT: s_andn2_b64 exec, exec, s[8:9]
; GFX6-NEXT: s_cbranch_execnz .LBB32_1
; GFX6-NEXT: ; %bb.2: ; %atomicrmw.end
; GFX6-NEXT: s_or_b64 exec, exec, s[8:9]
-; GFX6-NEXT: v_cvt_f32_f16_e32 v0, v4
+; GFX6-NEXT: v_cvt_f32_f16_e32 v0, v3
; GFX6-NEXT: s_waitcnt expcnt(0)
; GFX6-NEXT: s_setpc_b64 s[30:31]
%gep = getelementptr half, ptr addrspace(1) %ptr, i64 1023
@@ -7746,23 +7747,23 @@ define void @global_agent_atomic_fmin_noret_f16__offset12b__align4_pos__amdgpu_n
; GFX7-NEXT: buffer_load_dword v3, v[0:1], s[4:7], 0 addr64 offset:2046
; GFX7-NEXT: v_cvt_f16_f32_e32 v2, v2
; GFX7-NEXT: s_mov_b64 s[8:9], 0
-; GFX7-NEXT: v_cvt_f32_f16_e32 v4, v2
+; GFX7-NEXT: v_cvt_f32_f16_e32 v2, v2
; GFX7-NEXT: .LBB33_1: ; %atomicrmw.start
; GFX7-NEXT: ; =>This Inner Loop Header: Depth=1
; GFX7-NEXT: s_waitcnt vmcnt(0)
-; GFX7-NEXT: v_cvt_f32_f16_e32 v2, v3
-; GFX7-NEXT: v_and_b32_e32 v5, 0xffff0000, v3
-; GFX7-NEXT: v_min_f32_e32 v2, v2, v4
-; GFX7-NEXT: v_cvt_f16_f32_e32 v2, v2
-; GFX7-NEXT: v_or_b32_e32 v2, v5, v2
-; GFX7-NEXT: v_mov_b32_e32 v6, v3
-; GFX7-NEXT: v_mov_b32_e32 v5, v2
-; GFX7-NEXT: buffer_atomic_cmpswap v[5:6], v[0:1], s[4:7], 0 addr64 offset:2046 glc
+; GFX7-NEXT: v_mov_b32_e32 v5, v3
+; GFX7-NEXT: v_cvt_f32_f16_e32 v3, v5
+; GFX7-NEXT: v_and_b32_e32 v4, 0xffff0000, v5
+; GFX7-NEXT: v_min_f32_e32 v3, v3, v2
+; GFX7-NEXT: v_cvt_f16_f32_e32 v3, v3
+; GFX7-NEXT: v_or_b32_e32 v4, v4, v3
+; GFX7-NEXT: v_mov_b32_e32 v3, v4
+; GFX7-NEXT: v_mov_b32_e32 v4, v5
+; GFX7-NEXT: buffer_atomic_cmpswap v[3:4], v[0:1], s[4:7], 0 addr64 offset:2046 glc
; GFX7-NEXT: s_waitcnt vmcnt(0)
; GFX7-NEXT: buffer_wbinvl1
-; GFX7-NEXT: v_cmp_eq_u32_e32 vcc, v5, v3
+; GFX7-NEXT: v_cmp_eq_u32_e32 vcc, v3, v5
; GFX7-NEXT: s_or_b64 s[8:9], vcc, s[8:9]
-; GFX7-NEXT: v_mov_b32_e32 v3, v5
; GFX7-NEXT: s_andn2_b64 exec, exec, s[8:9]
; GFX7-NEXT: s_cbranch_execnz .LBB33_1
; GFX7-NEXT: ; %bb.2: ; %atomicrmw.end
@@ -7779,24 +7780,24 @@ define void @global_agent_atomic_fmin_noret_f16__offset12b__align4_pos__amdgpu_n
; GFX6-NEXT: buffer_load_dword v3, v[0:1], s[4:7], 0 addr64 offset:2046
; GFX6-NEXT: v_cvt_f16_f32_e32 v2, v2
; GFX6-NEXT: s_mov_b64 s[8:9], 0
-; GFX6-NEXT: v_cvt_f32_f16_e32 v4, v2
+; GFX6-NEXT: v_cvt_f32_f16_e32 v2, v2
; GFX6-NEXT: .LBB33_1: ; %atomicrmw.start
; GFX6-NEXT: ; =>This Inner Loop Header: Depth=1
; GFX6-NEXT: s_waitcnt vmcnt(0)
-; GFX6-NEXT: v_cvt_f32_f16_e32 v2, v3
+; GFX6-NEXT: v_mov_b32_e32 v5, v3
; GFX6-NEXT: s_waitcnt expcnt(0)
-; GFX6-NEXT: v_and_b32_e32 v5, 0xffff0000, v3
-; GFX6-NEXT: v_min_f32_e32 v2, v2, v4
-; GFX6-NEXT: v_cvt_f16_f32_e32 v2, v2
-; GFX6-NEXT: v_or_b32_e32 v2, v5, v2
-; GFX6-NEXT: v_mov_b32_e32 v6, v3
-; GFX6-NEXT: v_mov_b32_e32 v5, v2
-; GFX6-NEXT: buffer_atomic_cmpswap v[5:6], v[0:1], s[4:7], 0 addr64 offset:2046 glc
+; GFX6-NEXT: v_cvt_f32_f16_e32 v3, v5
+; GFX6-NEXT: v_and_b32_e32 v4, 0xffff0000, v5
+; GFX6-NEXT: v_min_f32_e32 v3, v3, v2
+; GFX6-NEXT: v_cvt_f16_f32_e32 v3, v3
+; GFX6-NEXT: v_or_b32_e32 v4, v4, v3
+; GFX6-NEXT: v_mov_b32_e32 v3, v4
+; GFX6-NEXT: v_mov_b32_e32 v4, v5
+; GFX6-NEXT: buffer_atomic_cmpswap v[3:4], v[0:1], s[4:7], 0 addr64 offset:2046 glc
; GFX6-NEXT: s_waitcnt vmcnt(0)
; GFX6-NEXT: buffer_wbinvl1
-; GFX6-NEXT: v_cmp_eq_u32_e32 vcc, v5, v3
+; GFX6-NEXT: v_cmp_eq_u32_e32 vcc, v3, v5
; GFX6-NEXT: s_or_b64 s[8:9], vcc, s[8:9]
-; GFX6-NEXT: v_mov_b32_e32 v3, v5
; GFX6-NEXT: s_andn2_b64 exec, exec, s[8:9]
; GFX6-NEXT: s_cbranch_execnz .LBB33_1
; GFX6-NEXT: ; %bb.2: ; %atomicrmw.end
@@ -8189,36 +8190,36 @@ define half @global_system_atomic_fmin_ret_f16__offset12b_pos__amdgpu_no_fine_gr
; GFX7-NEXT: s_mov_b32 s4, s6
; GFX7-NEXT: s_mov_b32 s5, s6
; GFX7-NEXT: buffer_load_dword v3, v[0:1], s[4:7], 0 addr64
-; GFX7-NEXT: v_cvt_f16_f32_e32 v2, v2
-; GFX7-NEXT: v_and_b32_e32 v4, 3, v4
-; GFX7-NEXT: v_lshlrev_b32_e32 v6, 3, v4
+; GFX7-NEXT: v_cvt_f16_f32_e32 v5, v2
+; GFX7-NEXT: v_and_b32_e32 v2, 3, v4
+; GFX7-NEXT: v_lshlrev_b32_e32 v2, 3, v2
+; GFX7-NEXT: v_lshl_b32_e32 v4, 0xffff, v2
+; GFX7-NEXT: v_cvt_f32_f16_e32 v5, v5
+; GFX7-NEXT: v_not_b32_e32 v6, v4
; GFX7-NEXT: s_mov_b64 s[8:9], 0
-; GFX7-NEXT: v_cvt_f32_f16_e32 v7, v2
-; GFX7-NEXT: v_lshl_b32_e32 v2, 0xffff, v6
-; GFX7-NEXT: v_not_b32_e32 v8, v2
; GFX7-NEXT: .LBB34_1: ; %atomicrmw.start
; GFX7-NEXT: ; =>This Inner Loop Header: Depth=1
; GFX7-NEXT: s_waitcnt vmcnt(0)
-; GFX7-NEXT: v_lshrrev_b32_e32 v2, v6, v3
-; GFX7-NEXT: v_cvt_f32_f16_e32 v2, v2
-; GFX7-NEXT: v_and_b32_e32 v4, v3, v8
-; GFX7-NEXT: v_min_f32_e32 v2, v2, v7
-; GFX7-NEXT: v_cvt_f16_f32_e32 v2, v2
-; GFX7-NEXT: v_lshlrev_b32_e32 v2, v6, v2
-; GFX7-NEXT: v_or_b32_e32 v2, v4, v2
-; GFX7-NEXT: v_mov_b32_e32 v5, v3
-; GFX7-NEXT: v_mov_b32_e32 v4, v2
-; GFX7-NEXT: buffer_atomic_cmpswap v[4:5], v[0:1], s[4:7], 0 addr64 glc
+; GFX7-NEXT: v_mov_b32_e32 v8, v3
+; GFX7-NEXT: v_lshrrev_b32_e32 v3, v2, v8
+; GFX7-NEXT: v_cvt_f32_f16_e32 v3, v3
+; GFX7-NEXT: v_and_b32_e32 v4, v8, v6
+; GFX7-NEXT: v_min_f32_e32 v3, v3, v5
+; GFX7-NEXT: v_cvt_f16_f32_e32 v3, v3
+; GFX7-NEXT: v_lshlrev_b32_e32 v3, v2, v3
+; GFX7-NEXT: v_or_b32_e32 v7, v4, v3
+; GFX7-NEXT: v_mov_b32_e32 v3, v7
+; GFX7-NEXT: v_mov_b32_e32 v4, v8
+; GFX7-NEXT: buffer_atomic_cmpswap v[3:4], v[0:1], s[4:7], 0 addr64 glc
; GFX7-NEXT: s_waitcnt vmcnt(0)
; GFX7-NEXT: buffer_wbinvl1
-; GFX7-NEXT: v_cmp_eq_u32_e32 vcc, v4, v3
+; GFX7-NEXT: v_cmp_eq_u32_e32 vcc, v3, v8
; GFX7-NEXT: s_or_b64 s[8:9], vcc, s[8:9]
-; GFX7-NEXT: v_mov_b32_e32 v3, v4
; GFX7-NEXT: s_andn2_b64 exec, exec, s[8:9]
; GFX7-NEXT: s_cbranch_execnz .LBB34_1
; GFX7-NEXT: ; %bb.2: ; %atomicrmw.end
; GFX7-NEXT: s_or_b64 exec, exec, s[8:9]
-; GFX7-NEXT: v_lshrrev_b32_e32 v0, v6, v4
+; GFX7-NEXT: v_lshrrev_b32_e32 v0, v2, v3
; GFX7-NEXT: v_cvt_f32_f16_e32 v0, v0
; GFX7-NEXT: s_setpc_b64 s[30:31]
;
@@ -8233,37 +8234,37 @@ define half @global_system_atomic_fmin_ret_f16__offset12b_pos__amdgpu_no_fine_gr
; GFX6-NEXT: s_mov_b32 s4, s6
; GFX6-NEXT: s_mov_b32 s5, s6
; GFX6-NEXT: buffer_load_dword v3, v[0:1], s[4:7], 0 addr64
-; GFX6-NEXT: v_cvt_f16_f32_e32 v2, v2
-; GFX6-NEXT: v_and_b32_e32 v4, 3, v4
-; GFX6-NEXT: v_lshlrev_b32_e32 v6, 3, v4
+; GFX6-NEXT: v_cvt_f16_f32_e32 v5, v2
+; GFX6-NEXT: v_and_b32_e32 v2, 3, v4
+; GFX6-NEXT: v_lshlrev_b32_e32 v2, 3, v2
+; GFX6-NEXT: v_lshl_b32_e32 v4, 0xffff, v2
+; GFX6-NEXT: v_cvt_f32_f16_e32 v5, v5
+; GFX6-NEXT: v_not_b32_e32 v6, v4
; GFX6-NEXT: s_mov_b64 s[8:9], 0
-; GFX6-NEXT: v_cvt_f32_f16_e32 v7, v2
-; GFX6-NEXT: v_lshl_b32_e32 v2, 0xffff, v6
-; GFX6-NEXT: v_not_b32_e32 v8, v2
; GFX6-NEXT: .LBB34_1: ; %atomicrmw.start
; GFX6-NEXT: ; =>This Inner Loop Header: Depth=1
; GFX6-NEXT: s_waitcnt vmcnt(0)
-; GFX6-NEXT: v_lshrrev_b32_e32 v2, v6, v3
-; GFX6-NEXT: v_cvt_f32_f16_e32 v2, v2
+; GFX6-NEXT: v_mov_b32_e32 v8, v3
; GFX6-NEXT: s_waitcnt expcnt(0)
-; GFX6-NEXT: v_and_b32_e32 v4, v3, v8
-; GFX6-NEXT: v_min_f32_e32 v2, v2, v7
-; GFX6-NEXT: v_cvt_f16_f32_e32 v2, v2
-; GFX6-NEXT: v_lshlrev_b32_e32 v2, v6, v2
-; GFX6-NEXT: v_or_b32_e32 v2, v4, v2
-; GFX6-NEXT: v_mov_b32_e32 v5, v3
-; GFX6-NEXT: v_mov_b32_e32 v4, v2
-; GFX6-NEXT: buffer_atomic_cmpswap v[4:5], v[0:1], s[4:7], 0 addr64 glc
+; GFX6-NEXT: v_lshrrev_b32_e32 v3, v2, v8
+; GFX6-NEXT: v_cvt_f32_f16_e32 v3, v3
+; GFX6-NEXT: v_and_b32_e32 v4, v8, v6
+; GFX6-NEXT: v_min_f32_e32 v3, v3, v5
+; GFX6-NEXT: v_cvt_f16_f32_e32 v3, v3
+; GFX6-NEXT: v_lshlrev_b32_e32 v3, v2, v3
+; GFX6-NEXT: v_or_b32_e32 v7, v4, v3
+; GFX6-NEXT: v_mov_b32_e32 v3, v7
+; GFX6-NEXT: v_mov_b32_e32 v4, v8
+; GFX6-NEXT: buffer_atomic_cmpswap v[3:4], v[0:1], s[4:7], 0 addr64 glc
; GFX6-NEXT: s_waitcnt vmcnt(0)
; GFX6-NEXT: buffer_wbinvl1
-; GFX6-NEXT: v_cmp_eq_u32_e32 vcc, v4, v3
+; GFX6-NEXT: v_cmp_eq_u32_e32 vcc, v3, v8
; GFX6-NEXT: s_or_b64 s[8:9], vcc, s[8:9]
-; GFX6-NEXT: v_mov_b32_e32 v3, v4
; GFX6-NEXT: s_andn2_b64 exec, exec, s[8:9]
; GFX6-NEXT: s_cbranch_execnz .LBB34_1
; GFX6-NEXT: ; %bb.2: ; %atomicrmw.end
; GFX6-NEXT: s_or_b64 exec, exec, s[8:9]
-; GFX6-NEXT: v_lshrrev_b32_e32 v0, v6, v4
+; GFX6-NEXT: v_lshrrev_b32_e32 v0, v2, v3
; GFX6-NEXT: v_cvt_f32_f16_e32 v0, v0
; GFX6-NEXT: s_waitcnt expcnt(0)
; GFX6-NEXT: s_setpc_b64 s[30:31]
@@ -8644,31 +8645,31 @@ define void @global_system_atomic_fmin_noret_f16__offset12b_pos__amdgpu_no_fine_
; GFX7-NEXT: s_mov_b32 s4, s6
; GFX7-NEXT: s_mov_b32 s5, s6
; GFX7-NEXT: buffer_load_dword v3, v[0:1], s[4:7], 0 addr64
-; GFX7-NEXT: v_cvt_f16_f32_e32 v2, v2
-; GFX7-NEXT: v_and_b32_e32 v4, 3, v4
-; GFX7-NEXT: v_lshlrev_b32_e32 v4, 3, v4
+; GFX7-NEXT: v_cvt_f16_f32_e32 v5, v2
+; GFX7-NEXT: v_and_b32_e32 v2, 3, v4
+; GFX7-NEXT: v_lshlrev_b32_e32 v2, 3, v2
+; GFX7-NEXT: v_lshl_b32_e32 v4, 0xffff, v2
+; GFX7-NEXT: v_cvt_f32_f16_e32 v5, v5
+; GFX7-NEXT: v_not_b32_e32 v6, v4
; GFX7-NEXT: s_mov_b64 s[8:9], 0
-; GFX7-NEXT: v_cvt_f32_f16_e32 v5, v2
-; GFX7-NEXT: v_lshl_b32_e32 v2, 0xffff, v4
-; GFX7-NEXT: v_not_b32_e32 v6, v2
; GFX7-NEXT: .LBB35_1: ; %atomicrmw.start
; GFX7-NEXT: ; =>This Inner Loop Header: Depth=1
; GFX7-NEXT: s_waitcnt vmcnt(0)
-; GFX7-NEXT: v_lshrrev_b32_e32 v2, v4, v3
-; GFX7-NEXT: v_cvt_f32_f16_e32 v2, v2
-; GFX7-NEXT: v_and_b32_e32 v7, v3, v6
-; GFX7-NEXT: v_min_f32_e32 v2, v2, v5
-; GFX7-NEXT: v_cvt_f16_f32_e32 v2, v2
-; GFX7-NEXT: v_lshlrev_b32_e32 v2, v4, v2
-; GFX7-NEXT: v_or_b32_e32 v2, v7, v2
; GFX7-NEXT: v_mov_b32_e32 v8, v3
-; GFX7-NEXT: v_mov_b32_e32 v7, v2
-; GFX7-NEXT: buffer_atomic_cmpswap v[7:8], v[0:1], s[4:7], 0 addr64 glc
+; GFX7-NEXT: v_lshrrev_b32_e32 v3, v2, v8
+; GFX7-NEXT: v_cvt_f32_f16_e32 v3, v3
+; GFX7-NEXT: v_and_b32_e32 v4, v8, v6
+; GFX7-NEXT: v_min_f32_e32 v3, v3, v5
+; GFX7-NEXT: v_cvt_f16_f32_e32 v3, v3
+; GFX7-NEXT: v_lshlrev_b32_e32 v3, v2, v3
+; GFX7-NEXT: v_or_b32_e32 v7, v4, v3
+; GFX7-NEXT: v_mov_b32_e32 v3, v7
+; GFX7-NEXT: v_mov_b32_e32 v4, v8
+; GFX7-NEXT: buffer_atomic_cmpswap v[3:4], v[0:1], s[4:7], 0 addr64 glc
; GFX7-NEXT: s_waitcnt vmcnt(0)
; GFX7-NEXT: buffer_wbinvl1
-; GFX7-NEXT: v_cmp_eq_u32_e32 vcc, v7, v3
+; GFX7-NEXT: v_cmp_eq_u32_e32 vcc, v3, v8
; GFX7-NEXT: s_or_b64 s[8:9], vcc, s[8:9]
-; GFX7-NEXT: v_mov_b32_e32 v3, v7
; GFX7-NEXT: s_andn2_b64 exec, exec, s[8:9]
; GFX7-NEXT: s_cbranch_execnz .LBB35_1
; GFX7-NEXT: ; %bb.2: ; %atomicrmw.end
@@ -8686,32 +8687,32 @@ define void @global_system_atomic_fmin_noret_f16__offset12b_pos__amdgpu_no_fine_
; GFX6-NEXT: s_mov_b32 s4, s6
; GFX6-NEXT: s_mov_b32 s5, s6
; GFX6-NEXT: buffer_load_dword v3, v[0:1], s[4:7], 0 addr64
-; GFX6-NEXT: v_cvt_f16_f32_e32 v2, v2
-; GFX6-NEXT: v_and_b32_e32 v4, 3, v4
-; GFX6-NEXT: v_lshlrev_b32_e32 v4, 3, v4
+; GFX6-NEXT: v_cvt_f16_f32_e32 v5, v2
+; GFX6-NEXT: v_and_b32_e32 v2, 3, v4
+; GFX6-NEXT: v_lshlrev_b32_e32 v2, 3, v2
+; GFX6-NEXT: v_lshl_b32_e32 v4, 0xffff, v2
+; GFX6-NEXT: v_cvt_f32_f16_e32 v5, v5
+; GFX6-NEXT: v_not_b32_e32 v6, v4
; GFX6-NEXT: s_mov_b64 s[8:9], 0
-; GFX6-NEXT: v_cvt_f32_f16_e32 v5, v2
-; GFX6-NEXT: v_lshl_b32_e32 v2, 0xffff, v4
-; GFX6-NEXT: v_not_b32_e32 v6, v2
; GFX6-NEXT: .LBB35_1: ; %atomicrmw.start
; GFX6-NEXT: ; =>This Inner Loop Header: Depth=1
; GFX6-NEXT: s_waitcnt vmcnt(0)
-; GFX6-NEXT: v_lshrrev_b32_e32 v2, v4, v3
-; GFX6-NEXT: v_cvt_f32_f16_e32 v2, v2
-; GFX6-NEXT: s_waitcnt expcnt(0)
-; GFX6-NEXT: v_and_b32_e32 v7, v3, v6
-; GFX6-NEXT: v_min_f32_e32 v2, v2, v5
-; GFX6-NEXT: v_cvt_f16_f32_e32 v2, v2
-; GFX6-NEXT: v_lshlrev_b32_e32 v2, v4, v2
-; GFX6-NEXT: v_or_b32_e32 v2, v7, v2
; GFX6-NEXT: v_mov_b32_e32 v8, v3
-; GFX6-NEXT: v_mov_b32_e32 v7, v2
-; GFX6-NEXT: buffer_atomic_cmpswap v[7:8], v[0:1], s[4:7], 0 addr64 glc
+; GFX6-NEXT: s_waitcnt expcnt(0)
+; GFX6-NEXT: v_lshrrev_b32_e32 v3, v2, v8
+; GFX6-NEXT: v_cvt_f32_f16_e32 v3, v3
+; GFX6-NEXT: v_and_b32_e32 v4, v8, v6
+; GFX6-NEXT: v_min_f32_e32 v3, v3, v5
+; GFX6-NEXT: v_cvt_f16_f32_e32 v3, v3
+; GFX6-NEXT: v_lshlrev_b32_e32 v3, v2, v3
+; GFX6-NEXT: v_or_b32_e32 v7, v4, v3
+; GFX6-NEXT: v_mov_b32_e32 v3, v7
+; GFX6-NEXT: v_mov_b32_e32 v4, v8
+; GFX6-NEXT: buffer_atomic_cmpswap v[3:4], v[0:1], s[4:7], 0 addr64 glc
; GFX6-NEXT: s_waitcnt vmcnt(0)
; GFX6-NEXT: buffer_wbinvl1
-; GFX6-NEXT: v_cmp_eq_u32_e32 vcc, v7, v3
+; GFX6-NEXT: v_cmp_eq_u32_e32 vcc, v3, v8
; GFX6-NEXT: s_or_b64 s[8:9], vcc, s[8:9]
-; GFX6-NEXT: v_mov_b32_e32 v3, v7
; GFX6-NEXT: s_andn2_b64 exec, exec, s[8:9]
; GFX6-NEXT: s_cbranch_execnz .LBB35_1
; GFX6-NEXT: ; %bb.2: ; %atomicrmw.end
@@ -9146,36 +9147,36 @@ define bfloat @global_agent_atomic_fmin_ret_bf16__amdgpu_no_fine_grained_memory(
; GFX7-NEXT: s_mov_b32 s5, s6
; GFX7-NEXT: buffer_load_dword v4, v[0:1], s[4:7], 0 addr64
; GFX7-NEXT: v_and_b32_e32 v3, 3, v3
-; GFX7-NEXT: v_lshlrev_b32_e32 v5, 3, v3
-; GFX7-NEXT: v_lshl_b32_e32 v3, 0xffff, v5
+; GFX7-NEXT: v_lshlrev_b32_e32 v3, 3, v3
+; GFX7-NEXT: v_lshl_b32_e32 v5, 0xffff, v3
; GFX7-NEXT: v_mul_f32_e32 v2, 1.0, v2
-; GFX7-NEXT: v_not_b32_e32 v6, v3
+; GFX7-NEXT: v_not_b32_e32 v6, v5
; GFX7-NEXT: s_mov_b64 s[8:9], 0
-; GFX7-NEXT: v_and_b32_e32 v7, 0xffff0000, v2
+; GFX7-NEXT: v_and_b32_e32 v2, 0xffff0000, v2
; GFX7-NEXT: .LBB36_1: ; %atomicrmw.start
; GFX7-NEXT: ; =>This Inner Loop Header: Depth=1
; GFX7-NEXT: s_waitcnt vmcnt(0)
-; GFX7-NEXT: v_lshrrev_b32_e32 v2, v5, v4
-; GFX7-NEXT: v_lshlrev_b32_e32 v2, 16, v2
-; GFX7-NEXT: v_mul_f32_e32 v2, 1.0, v2
-; GFX7-NEXT: v_min_f32_e32 v2, v2, v7
-; GFX7-NEXT: v_lshrrev_b32_e32 v2, 16, v2
-; GFX7-NEXT: v_and_b32_e32 v3, v4, v6
-; GFX7-NEXT: v_lshlrev_b32_e32 v2, v5, v2
-; GFX7-NEXT: v_or_b32_e32 v3, v3, v2
-; GFX7-NEXT: v_mov_b32_e32 v2, v3
-; GFX7-NEXT: v_mov_b32_e32 v3, v4
-; GFX7-NEXT: buffer_atomic_cmpswap v[2:3], v[0:1], s[4:7], 0 addr64 glc
+; GFX7-NEXT: v_mov_b32_e32 v8, v4
+; GFX7-NEXT: v_lshrrev_b32_e32 v4, v3, v8
+; GFX7-NEXT: v_lshlrev_b32_e32 v4, 16, v4
+; GFX7-NEXT: v_mul_f32_e32 v4, 1.0, v4
+; GFX7-NEXT: v_min_f32_e32 v4, v4, v2
+; GFX7-NEXT: v_lshrrev_b32_e32 v4, 16, v4
+; GFX7-NEXT: v_and_b32_e32 v5, v8, v6
+; GFX7-NEXT: v_lshlrev_b32_e32 v4, v3, v4
+; GFX7-NEXT: v_or_b32_e32 v7, v5, v4
+; GFX7-NEXT: v_mov_b32_e32 v4, v7
+; GFX7-NEXT: v_mov_b32_e32 v5, v8
+; GFX7-NEXT: buffer_atomic_cmpswap v[4:5], v[0:1], s[4:7], 0 addr64 glc
; GFX7-NEXT: s_waitcnt vmcnt(0)
; GFX7-NEXT: buffer_wbinvl1
-; GFX7-NEXT: v_cmp_eq_u32_e32 vcc, v2, v4
+; GFX7-NEXT: v_cmp_eq_u32_e32 vcc, v4, v8
; GFX7-NEXT: s_or_b64 s[8:9], vcc, s[8:9]
-; GFX7-NEXT: v_mov_b32_e32 v4, v2
; GFX7-NEXT: s_andn2_b64 exec, exec, s[8:9]
; GFX7-NEXT: s_cbranch_execnz .LBB36_1
; GFX7-NEXT: ; %bb.2: ; %atomicrmw.end
; GFX7-NEXT: s_or_b64 exec, exec, s[8:9]
-; GFX7-NEXT: v_lshrrev_b32_e32 v0, v5, v2
+; GFX7-NEXT: v_lshrrev_b32_e32 v0, v3, v4
; GFX7-NEXT: v_lshlrev_b32_e32 v0, 16, v0
; GFX7-NEXT: s_setpc_b64 s[30:31]
;
@@ -9190,36 +9191,37 @@ define bfloat @global_agent_atomic_fmin_ret_bf16__amdgpu_no_fine_grained_memory(
; GFX6-NEXT: s_mov_b32 s5, s6
; GFX6-NEXT: buffer_load_dword v4, v[0:1], s[4:7], 0 addr64
; GFX6-NEXT: v_and_b32_e32 v3, 3, v3
-; GFX6-NEXT: v_lshlrev_b32_e32 v5, 3, v3
-; GFX6-NEXT: v_lshl_b32_e32 v3, 0xffff, v5
+; GFX6-NEXT: v_lshlrev_b32_e32 v3, 3, v3
+; GFX6-NEXT: v_lshl_b32_e32 v5, 0xffff, v3
; GFX6-NEXT: v_mul_f32_e32 v2, 1.0, v2
-; GFX6-NEXT: v_not_b32_e32 v6, v3
+; GFX6-NEXT: v_not_b32_e32 v6, v5
; GFX6-NEXT: s_mov_b64 s[8:9], 0
-; GFX6-NEXT: v_and_b32_e32 v7, 0xffff0000, v2
+; GFX6-NEXT: v_and_b32_e32 v2, 0xffff0000, v2
; GFX6-NEXT: .LBB36_1: ; %atomicrmw.start
; GFX6-NEXT: ; =>This Inner Loop Header: Depth=1
-; GFX6-NEXT: s_waitcnt vmcnt(0) expcnt(0)
-; GFX6-NEXT: v_lshrrev_b32_e32 v2, v5, v4
-; GFX6-NEXT: v_lshlrev_b32_e32 v2, 16, v2
-; GFX6-NEXT: v_mul_f32_e32 v2, 1.0, v2
-; GFX6-NEXT: v_min_f32_e32 v2, v2, v7
-; GFX6-NEXT: v_lshrrev_b32_e32 v2, 16, v2
-; GFX6-NEXT: v_and_b32_e32 v3, v4, v6
-; GFX6-NEXT: v_lshlrev_b32_e32 v2, v5, v2
-; GFX6-NEXT: v_or_b32_e32 v3, v3, v2
-; GFX6-NEXT: v_mov_b32_e32 v2, v3
-; GFX6-NEXT: v_mov_b32_e32 v3, v4
-; GFX6-NEXT: buffer_atomic_cmpswap v[2:3], v[0:1], s[4:7], 0 addr64 glc
+; GFX6-NEXT: s_waitcnt vmcnt(0)
+; GFX6-NEXT: v_mov_b32_e32 v8, v4
+; GFX6-NEXT: s_waitcnt expcnt(0)
+; GFX6-NEXT: v_lshrrev_b32_e32 v4, v3, v8
+; GFX6-NEXT: v_lshlrev_b32_e32 v4, 16, v4
+; GFX6-NEXT: v_mul_f32_e32 v4, 1.0, v4
+; GFX6-NEXT: v_min_f32_e32 v4, v4, v2
+; GFX6-NEXT: v_lshrrev_b32_e32 v4, 16, v4
+; GFX6-NEXT: v_and_b32_e32 v5, v8, v6
+; GFX6-NEXT: v_lshlrev_b32_e32 v4, v3, v4
+; GFX6-NEXT: v_or_b32_e32 v7, v5, v4
+; GFX6-NEXT: v_mov_b32_e32 v4, v7
+; GFX6-NEXT: v_mov_b32_e32 v5, v8
+; GFX6-NEXT: buffer_atomic_cmpswap v[4:5], v[0:1], s[4:7], 0 addr64 glc
; GFX6-NEXT: s_waitcnt vmcnt(0)
; GFX6-NEXT: buffer_wbinvl1
-; GFX6-NEXT: v_cmp_eq_u32_e32 vcc, v2, v4
+; GFX6-NEXT: v_cmp_eq_u32_e32 vcc, v4, v8
; GFX6-NEXT: s_or_b64 s[8:9], vcc, s[8:9]
-; GFX6-NEXT: v_mov_b32_e32 v4, v2
; GFX6-NEXT: s_andn2_b64 exec, exec, s[8:9]
; GFX6-NEXT: s_cbranch_execnz .LBB36_1
; GFX6-NEXT: ; %bb.2: ; %atomicrmw.end
; GFX6-NEXT: s_or_b64 exec, exec, s[8:9]
-; GFX6-NEXT: v_lshrrev_b32_e32 v0, v5, v2
+; GFX6-NEXT: v_lshrrev_b32_e32 v0, v3, v4
; GFX6-NEXT: v_lshlrev_b32_e32 v0, 16, v0
; GFX6-NEXT: s_waitcnt expcnt(0)
; GFX6-NEXT: s_setpc_b64 s[30:31]
@@ -9661,36 +9663,36 @@ define bfloat @global_agent_atomic_fmin_ret_bf16__offset12b_pos__amdgpu_no_fine_
; GFX7-NEXT: s_mov_b32 s5, s6
; GFX7-NEXT: buffer_load_dword v3, v[0:1], s[4:7], 0 addr64
; GFX7-NEXT: v_and_b32_e32 v4, 3, v4
-; GFX7-NEXT: v_lshlrev_b32_e32 v6, 3, v4
-; GFX7-NEXT: v_lshl_b32_e32 v4, 0xffff, v6
+; GFX7-NEXT: v_lshlrev_b32_e32 v5, 3, v4
+; GFX7-NEXT: v_lshl_b32_e32 v4, 0xffff, v5
; GFX7-NEXT: v_mul_f32_e32 v2, 1.0, v2
-; GFX7-NEXT: v_not_b32_e32 v7, v4
+; GFX7-NEXT: v_not_b32_e32 v6, v4
; GFX7-NEXT: s_mov_b64 s[8:9], 0
-; GFX7-NEXT: v_and_b32_e32 v8, 0xffff0000, v2
+; GFX7-NEXT: v_and_b32_e32 v2, 0xffff0000, v2
; GFX7-NEXT: .LBB37_1: ; %atomicrmw.start
; GFX7-NEXT: ; =>This Inner Loop Header: Depth=1
; GFX7-NEXT: s_waitcnt vmcnt(0)
-; GFX7-NEXT: v_lshrrev_b32_e32 v2, v6, v3
-; GFX7-NEXT: v_lshlrev_b32_e32 v2, 16, v2
-; GFX7-NEXT: v_mul_f32_e32 v2, 1.0, v2
-; GFX7-NEXT: v_min_f32_e32 v2, v2, v8
-; GFX7-NEXT: v_lshrrev_b32_e32 v2, 16, v2
-; GFX7-NEXT: v_and_b32_e32 v4, v3, v7
-; GFX7-NEXT: v_lshlrev_b32_e32 v2, v6, v2
-; GFX7-NEXT: v_or_b32_e32 v2, v4, v2
-; GFX7-NEXT: v_mov_b32_e32 v5, v3
-; GFX7-NEXT: v_mov_b32_e32 v4, v2
-; GFX7-NEXT: buffer_atomic_cmpswap v[4:5], v[0:1], s[4:7], 0 addr64 glc
+; GFX7-NEXT: v_mov_b32_e32 v8, v3
+; GFX7-NEXT: v_lshrrev_b32_e32 v3, v5, v8
+; GFX7-NEXT: v_lshlrev_b32_e32 v3, 16, v3
+; GFX7-NEXT: v_mul_f32_e32 v3, 1.0, v3
+; GFX7-NEXT: v_min_f32_e32 v3, v3, v2
+; GFX7-NEXT: v_lshrrev_b32_e32 v3, 16, v3
+; GFX7-NEXT: v_and_b32_e32 v4, v8, v6
+; GFX7-NEXT: v_lshlrev_b32_e32 v3, v5, v3
+; GFX7-NEXT: v_or_b32_e32 v7, v4, v3
+; GFX7-NEXT: v_mov_b32_e32 v3, v7
+; GFX7-NEXT: v_mov_b32_e32 v4, v8
+; GFX7-NEXT: buffer_atomic_cmpswap v[3:4], v[0:1], s[4:7], 0 addr64 glc
; GFX7-NEXT: s_waitcnt vmcnt(0)
; GFX7-NEXT: buffer_wbinvl1
-; GFX7-NEXT: v_cmp_eq_u32_e32 vcc, v4, v3
+; GFX7-NEXT: v_cmp_eq_u32_e32 vcc, v3, v8
; GFX7-NEXT: s_or_b64 s[8:9], vcc, s[8:9]
-; GFX7-NEXT: v_mov_b32_e32 v3, v4
; GFX7-NEXT: s_andn2_b64 exec, exec, s[8:9]
; GFX7-NEXT: s_cbranch_execnz .LBB37_1
; GFX7-NEXT: ; %bb.2: ; %atomicrmw.end
; GFX7-NEXT: s_or_b64 exec, exec, s[8:9]
-; GFX7-NEXT: v_lshrrev_b32_e32 v0, v6, v4
+; GFX7-NEXT: v_lshrrev_b32_e32 v0, v5, v3
; GFX7-NEXT: v_lshlrev_b32_e32 v0, 16, v0
; GFX7-NEXT: s_setpc_b64 s[30:31]
;
@@ -9706,37 +9708,37 @@ define bfloat @global_agent_atomic_fmin_ret_bf16__offset12b_pos__amdgpu_no_fine_
; GFX6-NEXT: s_mov_b32 s5, s6
; GFX6-NEXT: buffer_load_dword v3, v[0:1], s[4:7], 0 addr64
; GFX6-NEXT: v_and_b32_e32 v4, 3, v4
-; GFX6-NEXT: v_lshlrev_b32_e32 v6, 3, v4
-; GFX6-NEXT: v_lshl_b32_e32 v4, 0xffff, v6
+; GFX6-NEXT: v_lshlrev_b32_e32 v5, 3, v4
+; GFX6-NEXT: v_lshl_b32_e32 v4, 0xffff, v5
; GFX6-NEXT: v_mul_f32_e32 v2, 1.0, v2
-; GFX6-NEXT: v_not_b32_e32 v7, v4
+; GFX6-NEXT: v_not_b32_e32 v6, v4
; GFX6-NEXT: s_mov_b64 s[8:9], 0
-; GFX6-NEXT: v_and_b32_e32 v8, 0xffff0000, v2
+; GFX6-NEXT: v_and_b32_e32 v2, 0xffff0000, v2
; GFX6-NEXT: .LBB37_1: ; %atomicrmw.start
; GFX6-NEXT: ; =>This Inner Loop Header: Depth=1
; GFX6-NEXT: s_waitcnt vmcnt(0)
-; GFX6-NEXT: v_lshrrev_b32_e32 v2, v6, v3
-; GFX6-NEXT: v_lshlrev_b32_e32 v2, 16, v2
-; GFX6-NEXT: v_mul_f32_e32 v2, 1.0, v2
-; GFX6-NEXT: v_min_f32_e32 v2, v2, v8
-; GFX6-NEXT: v_lshrrev_b32_e32 v2, 16, v2
+; GFX6-NEXT: v_mov_b32_e32 v8, v3
; GFX6-NEXT: s_waitcnt expcnt(0)
-; GFX6-NEXT: v_and_b32_e32 v4, v3, v7
-; GFX6-NEXT: v_lshlrev_b32_e32 v2, v6, v2
-; GFX6-NEXT: v_or_b32_e32 v2, v4, v2
-; GFX6-NEXT: v_mov_b32_e32 v5, v3
-; GFX6-NEXT: v_mov_b32_e32 v4, v2
-; GFX6-NEXT: buffer_atomic_cmpswap v[4:5], v[0:1], s[4:7], 0 addr64 glc
+; GFX6-NEXT: v_lshrrev_b32_e32 v3, v5, v8
+; GFX6-NEXT: v_lshlrev_b32_e32 v3, 16, v3
+; GFX6-NEXT: v_mul_f32_e32 v3, 1.0, v3
+; GFX6-NEXT: v_min_f32_e32 v3, v3, v2
+; GFX6-NEXT: v_lshrrev_b32_e32 v3, 16, v3
+; GFX6-NEXT: v_and_b32_e32 v4, v8, v6
+; GFX6-NEXT: v_lshlrev_b32_e32 v3, v5, v3
+; GFX6-NEXT: v_or_b32_e32 v7, v4, v3
+; GFX6-NEXT: v_mov_b32_e32 v3, v7
+; GFX6-NEXT: v_mov_b32_e32 v4, v8
+; GFX6-NEXT: buffer_atomic_cmpswap v[3:4], v[0:1], s[4:7], 0 addr64 glc
; GFX6-NEXT: s_waitcnt vmcnt(0)
; GFX6-NEXT: buffer_wbinvl1
-; GFX6-NEXT: v_cmp_eq_u32_e32 vcc, v4, v3
+; GFX6-NEXT: v_cmp_eq_u32_e32 vcc, v3, v8
; GFX6-NEXT: s_or_b64 s[8:9], vcc, s[8:9]
-; GFX6-NEXT: v_mov_b32_e32 v3, v4
; GFX6-NEXT: s_andn2_b64 exec, exec, s[8:9]
; GFX6-NEXT: s_cbranch_execnz .LBB37_1
; GFX6-NEXT: ; %bb.2: ; %atomicrmw.end
; GFX6-NEXT: s_or_b64 exec, exec, s[8:9]
-; GFX6-NEXT: v_lshrrev_b32_e32 v0, v6, v4
+; GFX6-NEXT: v_lshrrev_b32_e32 v0, v5, v3
; GFX6-NEXT: v_lshlrev_b32_e32 v0, 16, v0
; GFX6-NEXT: s_waitcnt expcnt(0)
; GFX6-NEXT: s_setpc_b64 s[30:31]
@@ -10180,36 +10182,36 @@ define bfloat @global_agent_atomic_fmin_ret_bf16__offset12b_neg__amdgpu_no_fine_
; GFX7-NEXT: s_mov_b32 s5, s6
; GFX7-NEXT: buffer_load_dword v3, v[0:1], s[4:7], 0 addr64
; GFX7-NEXT: v_and_b32_e32 v4, 3, v4
-; GFX7-NEXT: v_lshlrev_b32_e32 v6, 3, v4
-; GFX7-NEXT: v_lshl_b32_e32 v4, 0xffff, v6
+; GFX7-NEXT: v_lshlrev_b32_e32 v5, 3, v4
+; GFX7-NEXT: v_lshl_b32_e32 v4, 0xffff, v5
; GFX7-NEXT: v_mul_f32_e32 v2, 1.0, v2
-; GFX7-NEXT: v_not_b32_e32 v7, v4
+; GFX7-NEXT: v_not_b32_e32 v6, v4
; GFX7-NEXT: s_mov_b64 s[8:9], 0
-; GFX7-NEXT: v_and_b32_e32 v8, 0xffff0000, v2
+; GFX7-NEXT: v_and_b32_e32 v2, 0xffff0000, v2
; GFX7-NEXT: .LBB38_1: ; %atomicrmw.start
; GFX7-NEXT: ; =>This Inner Loop Header: Depth=1
; GFX7-NEXT: s_waitcnt vmcnt(0)
-; GFX7-NEXT: v_lshrrev_b32_e32 v2, v6, v3
-; GFX7-NEXT: v_lshlrev_b32_e32 v2, 16, v2
-; GFX7-NEXT: v_mul_f32_e32 v2, 1.0, v2
-; GFX7-NEXT: v_min_f32_e32 v2, v2, v8
-; GFX7-NEXT: v_lshrrev_b32_e32 v2, 16, v2
-; GFX7-NEXT: v_and_b32_e32 v4, v3, v7
-; GFX7-NEXT: v_lshlrev_b32_e32 v2, v6, v2
-; GFX7-NEXT: v_or_b32_e32 v2, v4, v2
-; GFX7-NEXT: v_mov_b32_e32 v5, v3
-; GFX7-NEXT: v_mov_b32_e32 v4, v2
-; GFX7-NEXT: buffer_atomic_cmpswap v[4:5], v[0:1], s[4:7], 0 addr64 glc
+; GFX7-NEXT: v_mov_b32_e32 v8, v3
+; GFX7-NEXT: v_lshrrev_b32_e32 v3, v5, v8
+; GFX7-NEXT: v_lshlrev_b32_e32 v3, 16, v3
+; GFX7-NEXT: v_mul_f32_e32 v3, 1.0, v3
+; GFX7-NEXT: v_min_f32_e32 v3, v3, v2
+; GFX7-NEXT: v_lshrrev_b32_e32 v3, 16, v3
+; GFX7-NEXT: v_and_b32_e32 v4, v8, v6
+; GFX7-NEXT: v_lshlrev_b32_e32 v3, v5, v3
+; GFX7-NEXT: v_or_b32_e32 v7, v4, v3
+; GFX7-NEXT: v_mov_b32_e32 v3, v7
+; GFX7-NEXT: v_mov_b32_e32 v4, v8
+; GFX7-NEXT: buffer_atomic_cmpswap v[3:4], v[0:1], s[4:7], 0 addr64 glc
; GFX7-NEXT: s_waitcnt vmcnt(0)
; GFX7-NEXT: buffer_wbinvl1
-; GFX7-NEXT: v_cmp_eq_u32_e32 vcc, v4, v3
+; GFX7-NEXT: v_cmp_eq_u32_e32 vcc, v3, v8
; GFX7-NEXT: s_or_b64 s[8:9], vcc, s[8:9]
-; GFX7-NEXT: v_mov_b32_e32 v3, v4
; GFX7-NEXT: s_andn2_b64 exec, exec, s[8:9]
; GFX7-NEXT: s_cbranch_execnz .LBB38_1
; GFX7-NEXT: ; %bb.2: ; %atomicrmw.end
; GFX7-NEXT: s_or_b64 exec, exec, s[8:9]
-; GFX7-NEXT: v_lshrrev_b32_e32 v0, v6, v4
+; GFX7-NEXT: v_lshrrev_b32_e32 v0, v5, v3
; GFX7-NEXT: v_lshlrev_b32_e32 v0, 16, v0
; GFX7-NEXT: s_setpc_b64 s[30:31]
;
@@ -10225,37 +10227,37 @@ define bfloat @global_agent_atomic_fmin_ret_bf16__offset12b_neg__amdgpu_no_fine_
; GFX6-NEXT: s_mov_b32 s5, s6
; GFX6-NEXT: buffer_load_dword v3, v[0:1], s[4:7], 0 addr64
; GFX6-NEXT: v_and_b32_e32 v4, 3, v4
-; GFX6-NEXT: v_lshlrev_b32_e32 v6, 3, v4
-; GFX6-NEXT: v_lshl_b32_e32 v4, 0xffff, v6
+; GFX6-NEXT: v_lshlrev_b32_e32 v5, 3, v4
+; GFX6-NEXT: v_lshl_b32_e32 v4, 0xffff, v5
; GFX6-NEXT: v_mul_f32_e32 v2, 1.0, v2
-; GFX6-NEXT: v_not_b32_e32 v7, v4
+; GFX6-NEXT: v_not_b32_e32 v6, v4
; GFX6-NEXT: s_mov_b64 s[8:9], 0
-; GFX6-NEXT: v_and_b32_e32 v8, 0xffff0000, v2
+; GFX6-NEXT: v_and_b32_e32 v2, 0xffff0000, v2
; GFX6-NEXT: .LBB38_1: ; %atomicrmw.start
; GFX6-NEXT: ; =>This Inner Loop Header: Depth=1
; GFX6-NEXT: s_waitcnt vmcnt(0)
-; GFX6-NEXT: v_lshrrev_b32_e32 v2, v6, v3
-; GFX6-NEXT: v_lshlrev_b32_e32 v2, 16, v2
-; GFX6-NEXT: v_mul_f32_e32 v2, 1.0, v2
-; GFX6-NEXT: v_min_f32_e32 v2, v2, v8
-; GFX6-NEXT: v_lshrrev_b32_e32 v2, 16, v2
+; GFX6-NEXT: v_mov_b32_e32 v8, v3
; GFX6-NEXT: s_waitcnt expcnt(0)
-; GFX6-NEXT: v_and_b32_e32 v4, v3, v7
-; GFX6-NEXT: v_lshlrev_b32_e32 v2, v6, v2
-; GFX6-NEXT: v_or_b32_e32 v2, v4, v2
-; GFX6-NEXT: v_mov_b32_e32 v5, v3
-; GFX6-NEXT: v_mov_b32_e32 v4, v2
-; GFX6-NEXT: buffer_atomic_cmpswap v[4:5], v[0:1], s[4:7], 0 addr64 glc
+; GFX6-NEXT: v_lshrrev_b32_e32 v3, v5, v8
+; GFX6-NEXT: v_lshlrev_b32_e32 v3, 16, v3
+; GFX6-NEXT: v_mul_f32_e32 v3, 1.0, v3
+; GFX6-NEXT: v_min_f32_e32 v3, v3, v2
+; GFX6-NEXT: v_lshrrev_b32_e32 v3, 16, v3
+; GFX6-NEXT: v_and_b32_e32 v4, v8, v6
+; GFX6-NEXT: v_lshlrev_b32_e32 v3, v5, v3
+; GFX6-NEXT: v_or_b32_e32 v7, v4, v3
+; GFX6-NEXT: v_mov_b32_e32 v3, v7
+; GFX6-NEXT: v_mov_b32_e32 v4, v8
+; GFX6-NEXT: buffer_atomic_cmpswap v[3:4], v[0:1], s[4:7], 0 addr64 glc
; GFX6-NEXT: s_waitcnt vmcnt(0)
; GFX6-NEXT: buffer_wbinvl1
-; GFX6-NEXT: v_cmp_eq_u32_e32 vcc, v4, v3
+; GFX6-NEXT: v_cmp_eq_u32_e32 vcc, v3, v8
; GFX6-NEXT: s_or_b64 s[8:9], vcc, s[8:9]
-; GFX6-NEXT: v_mov_b32_e32 v3, v4
; GFX6-NEXT: s_andn2_b64 exec, exec, s[8:9]
; GFX6-NEXT: s_cbranch_execnz .LBB38_1
; GFX6-NEXT: ; %bb.2: ; %atomicrmw.end
; GFX6-NEXT: s_or_b64 exec, exec, s[8:9]
-; GFX6-NEXT: v_lshrrev_b32_e32 v0, v6, v4
+; GFX6-NEXT: v_lshrrev_b32_e32 v0, v5, v3
; GFX6-NEXT: v_lshlrev_b32_e32 v0, 16, v0
; GFX6-NEXT: s_waitcnt expcnt(0)
; GFX6-NEXT: s_setpc_b64 s[30:31]
@@ -10670,31 +10672,31 @@ define void @global_agent_atomic_fmin_noret_bf16__amdgpu_no_fine_grained_memory(
; GFX7-NEXT: s_mov_b32 s5, s6
; GFX7-NEXT: buffer_load_dword v4, v[0:1], s[4:7], 0 addr64
; GFX7-NEXT: v_and_b32_e32 v3, 3, v3
-; GFX7-NEXT: v_lshlrev_b32_e32 v5, 3, v3
-; GFX7-NEXT: v_lshl_b32_e32 v3, 0xffff, v5
+; GFX7-NEXT: v_lshlrev_b32_e32 v3, 3, v3
+; GFX7-NEXT: v_lshl_b32_e32 v5, 0xffff, v3
; GFX7-NEXT: v_mul_f32_e32 v2, 1.0, v2
-; GFX7-NEXT: v_not_b32_e32 v6, v3
+; GFX7-NEXT: v_not_b32_e32 v6, v5
; GFX7-NEXT: s_mov_b64 s[8:9], 0
; GFX7-NEXT: v_and_b32_e32 v2, 0xffff0000, v2
; GFX7-NEXT: .LBB39_1: ; %atomicrmw.start
; GFX7-NEXT: ; =>This Inner Loop Header: Depth=1
; GFX7-NEXT: s_waitcnt vmcnt(0)
-; GFX7-NEXT: v_lshrrev_b32_e32 v3, v5, v4
-; GFX7-NEXT: v_lshlrev_b32_e32 v3, 16, v3
-; GFX7-NEXT: v_mul_f32_e32 v3, 1.0, v3
-; GFX7-NEXT: v_min_f32_e32 v3, v3, v2
-; GFX7-NEXT: v_lshrrev_b32_e32 v3, 16, v3
-; GFX7-NEXT: v_and_b32_e32 v7, v4, v6
-; GFX7-NEXT: v_lshlrev_b32_e32 v3, v5, v3
-; GFX7-NEXT: v_or_b32_e32 v3, v7, v3
; GFX7-NEXT: v_mov_b32_e32 v8, v4
-; GFX7-NEXT: v_mov_b32_e32 v7, v3
-; GFX7-NEXT: buffer_atomic_cmpswap v[7:8], v[0:1], s[4:7], 0 addr64 glc
+; GFX7-NEXT: v_lshrrev_b32_e32 v4, v3, v8
+; GFX7-NEXT: v_lshlrev_b32_e32 v4, 16, v4
+; GFX7-NEXT: v_mul_f32_e32 v4, 1.0, v4
+; GFX7-NEXT: v_min_f32_e32 v4, v4, v2
+; GFX7-NEXT: v_lshrrev_b32_e32 v4, 16, v4
+; GFX7-NEXT: v_and_b32_e32 v5, v8, v6
+; GFX7-NEXT: v_lshlrev_b32_e32 v4, v3, v4
+; GFX7-NEXT: v_or_b32_e32 v7, v5, v4
+; GFX7-NEXT: v_mov_b32_e32 v4, v7
+; GFX7-NEXT: v_mov_b32_e32 v5, v8
+; GFX7-NEXT: buffer_atomic_cmpswap v[4:5], v[0:1], s[4:7], 0 addr64 glc
; GFX7-NEXT: s_waitcnt vmcnt(0)
; GFX7-NEXT: buffer_wbinvl1
-; GFX7-NEXT: v_cmp_eq_u32_e32 vcc, v7, v4
+; GFX7-NEXT: v_cmp_eq_u32_e32 vcc, v4, v8
; GFX7-NEXT: s_or_b64 s[8:9], vcc, s[8:9]
-; GFX7-NEXT: v_mov_b32_e32 v4, v7
; GFX7-NEXT: s_andn2_b64 exec, exec, s[8:9]
; GFX7-NEXT: s_cbranch_execnz .LBB39_1
; GFX7-NEXT: ; %bb.2: ; %atomicrmw.end
@@ -10712,32 +10714,32 @@ define void @global_agent_atomic_fmin_noret_bf16__amdgpu_no_fine_grained_memory(
; GFX6-NEXT: s_mov_b32 s5, s6
; GFX6-NEXT: buffer_load_dword v4, v[0:1], s[4:7], 0 addr64
; GFX6-NEXT: v_and_b32_e32 v3, 3, v3
-; GFX6-NEXT: v_lshlrev_b32_e32 v5, 3, v3
-; GFX6-NEXT: v_lshl_b32_e32 v3, 0xffff, v5
+; GFX6-NEXT: v_lshlrev_b32_e32 v3, 3, v3
+; GFX6-NEXT: v_lshl_b32_e32 v5, 0xffff, v3
; GFX6-NEXT: v_mul_f32_e32 v2, 1.0, v2
-; GFX6-NEXT: v_not_b32_e32 v6, v3
+; GFX6-NEXT: v_not_b32_e32 v6, v5
; GFX6-NEXT: s_mov_b64 s[8:9], 0
; GFX6-NEXT: v_and_b32_e32 v2, 0xffff0000, v2
; GFX6-NEXT: .LBB39_1: ; %atomicrmw.start
; GFX6-NEXT: ; =>This Inner Loop Header: Depth=1
; GFX6-NEXT: s_waitcnt vmcnt(0)
-; GFX6-NEXT: v_lshrrev_b32_e32 v3, v5, v4
-; GFX6-NEXT: v_lshlrev_b32_e32 v3, 16, v3
-; GFX6-NEXT: v_mul_f32_e32 v3, 1.0, v3
-; GFX6-NEXT: v_min_f32_e32 v3, v3, v2
-; GFX6-NEXT: v_lshrrev_b32_e32 v3, 16, v3
-; GFX6-NEXT: s_waitcnt expcnt(0)
-; GFX6-NEXT: v_and_b32_e32 v7, v4, v6
-; GFX6-NEXT: v_lshlrev_b32_e32 v3, v5, v3
-; GFX6-NEXT: v_or_b32_e32 v3, v7, v3
; GFX6-NEXT: v_mov_b32_e32 v8, v4
-; GFX6-NEXT: v_mov_b32_e32 v7, v3
-; GFX6-NEXT: buffer_atomic_cmpswap v[7:8], v[0:1], s[4:7], 0 addr64 glc
+; GFX6-NEXT: s_waitcnt expcnt(0)
+; GFX6-NEXT: v_lshrrev_b32_e32 v4, v3, v8
+; GFX6-NEXT: v_lshlrev_b32_e32 v4, 16, v4
+; GFX6-NEXT: v_mul_f32_e32 v4, 1.0, v4
+; GFX6-NEXT: v_min_f32_e32 v4, v4, v2
+; GFX6-NEXT: v_lshrrev_b32_e32 v4, 16, v4
+; GFX6-NEXT: v_and_b32_e32 v5, v8, v6
+; GFX6-NEXT: v_lshlrev_b32_e32 v4, v3, v4
+; GFX6-NEXT: v_or_b32_e32 v7, v5, v4
+; GFX6-NEXT: v_mov_b32_e32 v4, v7
+; GFX6-NEXT: v_mov_b32_e32 v5, v8
+; GFX6-NEXT: buffer_atomic_cmpswap v[4:5], v[0:1], s[4:7], 0 addr64 glc
; GFX6-NEXT: s_waitcnt vmcnt(0)
; GFX6-NEXT: buffer_wbinvl1
-; GFX6-NEXT: v_cmp_eq_u32_e32 vcc, v7, v4
+; GFX6-NEXT: v_cmp_eq_u32_e32 vcc, v4, v8
; GFX6-NEXT: s_or_b64 s[8:9], vcc, s[8:9]
-; GFX6-NEXT: v_mov_b32_e32 v4, v7
; GFX6-NEXT: s_andn2_b64 exec, exec, s[8:9]
; GFX6-NEXT: s_cbranch_execnz .LBB39_1
; GFX6-NEXT: ; %bb.2: ; %atomicrmw.end
@@ -11169,31 +11171,31 @@ define void @global_agent_atomic_fmin_noret_bf16__offset12b_pos__amdgpu_no_fine_
; GFX7-NEXT: s_mov_b32 s5, s6
; GFX7-NEXT: buffer_load_dword v3, v[0:1], s[4:7], 0 addr64
; GFX7-NEXT: v_and_b32_e32 v4, 3, v4
-; GFX7-NEXT: v_lshlrev_b32_e32 v4, 3, v4
-; GFX7-NEXT: v_lshl_b32_e32 v5, 0xffff, v4
+; GFX7-NEXT: v_lshlrev_b32_e32 v5, 3, v4
+; GFX7-NEXT: v_lshl_b32_e32 v4, 0xffff, v5
; GFX7-NEXT: v_mul_f32_e32 v2, 1.0, v2
-; GFX7-NEXT: v_not_b32_e32 v5, v5
+; GFX7-NEXT: v_not_b32_e32 v6, v4
; GFX7-NEXT: s_mov_b64 s[8:9], 0
-; GFX7-NEXT: v_and_b32_e32 v6, 0xffff0000, v2
+; GFX7-NEXT: v_and_b32_e32 v2, 0xffff0000, v2
; GFX7-NEXT: .LBB40_1: ; %atomicrmw.start
; GFX7-NEXT: ; =>This Inner Loop Header: Depth=1
; GFX7-NEXT: s_waitcnt vmcnt(0)
-; GFX7-NEXT: v_lshrrev_b32_e32 v2, v4, v3
-; GFX7-NEXT: v_lshlrev_b32_e32 v2, 16, v2
-; GFX7-NEXT: v_mul_f32_e32 v2, 1.0, v2
-; GFX7-NEXT: v_min_f32_e32 v2, v2, v6
-; GFX7-NEXT: v_lshrrev_b32_e32 v2, 16, v2
-; GFX7-NEXT: v_and_b32_e32 v7, v3, v5
-; GFX7-NEXT: v_lshlrev_b32_e32 v2, v4, v2
-; GFX7-NEXT: v_or_b32_e32 v2, v7, v2
; GFX7-NEXT: v_mov_b32_e32 v8, v3
-; GFX7-NEXT: v_mov_b32_e32 v7, v2
-; GFX7-NEXT: buffer_atomic_cmpswap v[7:8], v[0:1], s[4:7], 0 addr64 glc
+; GFX7-NEXT: v_lshrrev_b32_e32 v3, v5, v8
+; GFX7-NEXT: v_lshlrev_b32_e32 v3, 16, v3
+; GFX7-NEXT: v_mul_f32_e32 v3, 1.0, v3
+; GFX7-NEXT: v_min_f32_e32 v3, v3, v2
+; GFX7-NEXT: v_lshrrev_b32_e32 v3, 16, v3
+; GFX7-NEXT: v_and_b32_e32 v4, v8, v6
+; GFX7-NEXT: v_lshlrev_b32_e32 v3, v5, v3
+; GFX7-NEXT: v_or_b32_e32 v7, v4, v3
+; GFX7-NEXT: v_mov_b32_e32 v3, v7
+; GFX7-NEXT: v_mov_b32_e32 v4, v8
+; GFX7-NEXT: buffer_atomic_cmpswap v[3:4], v[0:1], s[4:7], 0 addr64 glc
; GFX7-NEXT: s_waitcnt vmcnt(0)
; GFX7-NEXT: buffer_wbinvl1
-; GFX7-NEXT: v_cmp_eq_u32_e32 vcc, v7, v3
+; GFX7-NEXT: v_cmp_eq_u32_e32 vcc, v3, v8
; GFX7-NEXT: s_or_b64 s[8:9], vcc, s[8:9]
-; GFX7-NEXT: v_mov_b32_e32 v3, v7
; GFX7-NEXT: s_andn2_b64 exec, exec, s[8:9]
; GFX7-NEXT: s_cbranch_execnz .LBB40_1
; GFX7-NEXT: ; %bb.2: ; %atomicrmw.end
@@ -11212,32 +11214,32 @@ define void @global_agent_atomic_fmin_noret_bf16__offset12b_pos__amdgpu_no_fine_
; GFX6-NEXT: s_mov_b32 s5, s6
; GFX6-NEXT: buffer_load_dword v3, v[0:1], s[4:7], 0 addr64
; GFX6-NEXT: v_and_b32_e32 v4, 3, v4
-; GFX6-NEXT: v_lshlrev_b32_e32 v4, 3, v4
-; GFX6-NEXT: v_lshl_b32_e32 v5, 0xffff, v4
+; GFX6-NEXT: v_lshlrev_b32_e32 v5, 3, v4
+; GFX6-NEXT: v_lshl_b32_e32 v4, 0xffff, v5
; GFX6-NEXT: v_mul_f32_e32 v2, 1.0, v2
-; GFX6-NEXT: v_not_b32_e32 v5, v5
+; GFX6-NEXT: v_not_b32_e32 v6, v4
; GFX6-NEXT: s_mov_b64 s[8:9], 0
-; GFX6-NEXT: v_and_b32_e32 v6, 0xffff0000, v2
+; GFX6-NEXT: v_and_b32_e32 v2, 0xffff0000, v2
; GFX6-NEXT: .LBB40_1: ; %atomicrmw.start
; GFX6-NEXT: ; =>This Inner Loop Header: Depth=1
; GFX6-NEXT: s_waitcnt vmcnt(0)
-; GFX6-NEXT: v_lshrrev_b32_e32 v2, v4, v3
-; GFX6-NEXT: v_lshlrev_b32_e32 v2, 16, v2
-; GFX6-NEXT: v_mul_f32_e32 v2, 1.0, v2
-; GFX6-NEXT: v_min_f32_e32 v2, v2, v6
-; GFX6-NEXT: v_lshrrev_b32_e32 v2, 16, v2
-; GFX6-NEXT: s_waitcnt expcnt(0)
-; GFX6-NEXT: v_and_b32_e32 v7, v3, v5
-; GFX6-NEXT: v_lshlrev_b32_e32 v2, v4, v2
-; GFX6-NEXT: v_or_b32_e32 v2, v7, v2
; GFX6-NEXT: v_mov_b32_e32 v8, v3
-; GFX6-NEXT: v_mov_b32_e32 v7, v2
-; GFX6-NEXT: buffer_atomic_cmpswap v[7:8], v[0:1], s[4:7], 0 addr64 glc
+; GFX6-NEXT: s_waitcnt expcnt(0)
+; GFX6-NEXT: v_lshrrev_b32_e32 v3, v5, v8
+; GFX6-NEXT: v_lshlrev_b32_e32 v3, 16, v3
+; GFX6-NEXT: v_mul_f32_e32 v3, 1.0, v3
+; GFX6-NEXT: v_min_f32_e32 v3, v3, v2
+; GFX6-NEXT: v_lshrrev_b32_e32 v3, 16, v3
+; GFX6-NEXT: v_and_b32_e32 v4, v8, v6
+; GFX6-NEXT: v_lshlrev_b32_e32 v3, v5, v3
+; GFX6-NEXT: v_or_b32_e32 v7, v4, v3
+; GFX6-NEXT: v_mov_b32_e32 v3, v7
+; GFX6-NEXT: v_mov_b32_e32 v4, v8
+; GFX6-NEXT: buffer_atomic_cmpswap v[3:4], v[0:1], s[4:7], 0 addr64 glc
; GFX6-NEXT: s_waitcnt vmcnt(0)
; GFX6-NEXT: buffer_wbinvl1
-; GFX6-NEXT: v_cmp_eq_u32_e32 vcc, v7, v3
+; GFX6-NEXT: v_cmp_eq_u32_e32 vcc, v3, v8
; GFX6-NEXT: s_or_b64 s[8:9], vcc, s[8:9]
-; GFX6-NEXT: v_mov_b32_e32 v3, v7
; GFX6-NEXT: s_andn2_b64 exec, exec, s[8:9]
; GFX6-NEXT: s_cbranch_execnz .LBB40_1
; GFX6-NEXT: ; %bb.2: ; %atomicrmw.end
@@ -11671,31 +11673,31 @@ define void @global_agent_atomic_fmin_noret_bf16__offset12b_neg__amdgpu_no_fine_
; GFX7-NEXT: s_mov_b32 s5, s6
; GFX7-NEXT: buffer_load_dword v3, v[0:1], s[4:7], 0 addr64
; GFX7-NEXT: v_and_b32_e32 v4, 3, v4
-; GFX7-NEXT: v_lshlrev_b32_e32 v4, 3, v4
-; GFX7-NEXT: v_lshl_b32_e32 v5, 0xffff, v4
+; GFX7-NEXT: v_lshlrev_b32_e32 v5, 3, v4
+; GFX7-NEXT: v_lshl_b32_e32 v4, 0xffff, v5
; GFX7-NEXT: v_mul_f32_e32 v2, 1.0, v2
-; GFX7-NEXT: v_not_b32_e32 v5, v5
+; GFX7-NEXT: v_not_b32_e32 v6, v4
; GFX7-NEXT: s_mov_b64 s[8:9], 0
-; GFX7-NEXT: v_and_b32_e32 v6, 0xffff0000, v2
+; GFX7-NEXT: v_and_b32_e32 v2, 0xffff0000, v2
; GFX7-NEXT: .LBB41_1: ; %atomicrmw.start
; GFX7-NEXT: ; =>This Inner Loop Header: Depth=1
; GFX7-NEXT: s_waitcnt vmcnt(0)
-; GFX7-NEXT: v_lshrrev_b32_e32 v2, v4, v3
-; GFX7-NEXT: v_lshlrev_b32_e32 v2, 16, v2
-; GFX7-NEXT: v_mul_f32_e32 v2, 1.0, v2
-; GFX7-NEXT: v_min_f32_e32 v2, v2, v6
-; GFX7-NEXT: v_lshrrev_b32_e32 v2, 16, v2
-; GFX7-NEXT: v_and_b32_e32 v7, v3, v5
-; GFX7-NEXT: v_lshlrev_b32_e32 v2, v4, v2
-; GFX7-NEXT: v_or_b32_e32 v2, v7, v2
; GFX7-NEXT: v_mov_b32_e32 v8, v3
-; GFX7-NEXT: v_mov_b32_e32 v7, v2
-; GFX7-NEXT: buffer_atomic_cmpswap v[7:8], v[0:1], s[4:7], 0 addr64 glc
+; GFX7-NEXT: v_lshrrev_b32_e32 v3, v5, v8
+; GFX7-NEXT: v_lshlrev_b32_e32 v3, 16, v3
+; GFX7-NEXT: v_mul_f32_e32 v3, 1.0, v3
+; GFX7-NEXT: v_min_f32_e32 v3, v3, v2
+; GFX7-NEXT: v_lshrrev_b32_e32 v3, 16, v3
+; GFX7-NEXT: v_and_b32_e32 v4, v8, v6
+; GFX7-NEXT: v_lshlrev_b32_e32 v3, v5, v3
+; GFX7-NEXT: v_or_b32_e32 v7, v4, v3
+; GFX7-NEXT: v_mov_b32_e32 v3, v7
+; GFX7-NEXT: v_mov_b32_e32 v4, v8
+; GFX7-NEXT: buffer_atomic_cmpswap v[3:4], v[0:1], s[4:7], 0 addr64 glc
; GFX7-NEXT: s_waitcnt vmcnt(0)
; GFX7-NEXT: buffer_wbinvl1
-; GFX7-NEXT: v_cmp_eq_u32_e32 vcc, v7, v3
+; GFX7-NEXT: v_cmp_eq_u32_e32 vcc, v3, v8
; GFX7-NEXT: s_or_b64 s[8:9], vcc, s[8:9]
-; GFX7-NEXT: v_mov_b32_e32 v3, v7
; GFX7-NEXT: s_andn2_b64 exec, exec, s[8:9]
; GFX7-NEXT: s_cbranch_execnz .LBB41_1
; GFX7-NEXT: ; %bb.2: ; %atomicrmw.end
@@ -11714,32 +11716,32 @@ define void @global_agent_atomic_fmin_noret_bf16__offset12b_neg__amdgpu_no_fine_
; GFX6-NEXT: s_mov_b32 s5, s6
; GFX6-NEXT: buffer_load_dword v3, v[0:1], s[4:7], 0 addr64
; GFX6-NEXT: v_and_b32_e32 v4, 3, v4
-; GFX6-NEXT: v_lshlrev_b32_e32 v4, 3, v4
-; GFX6-NEXT: v_lshl_b32_e32 v5, 0xffff, v4
+; GFX6-NEXT: v_lshlrev_b32_e32 v5, 3, v4
+; GFX6-NEXT: v_lshl_b32_e32 v4, 0xffff, v5
; GFX6-NEXT: v_mul_f32_e32 v2, 1.0, v2
-; GFX6-NEXT: v_not_b32_e32 v5, v5
+; GFX6-NEXT: v_not_b32_e32 v6, v4
; GFX6-NEXT: s_mov_b64 s[8:9], 0
-; GFX6-NEXT: v_and_b32_e32 v6, 0xffff0000, v2
+; GFX6-NEXT: v_and_b32_e32 v2, 0xffff0000, v2
; GFX6-NEXT: .LBB41_1: ; %atomicrmw.start
; GFX6-NEXT: ; =>This Inner Loop Header: Depth=1
; GFX6-NEXT: s_waitcnt vmcnt(0)
-; GFX6-NEXT: v_lshrrev_b32_e32 v2, v4, v3
-; GFX6-NEXT: v_lshlrev_b32_e32 v2, 16, v2
-; GFX6-NEXT: v_mul_f32_e32 v2, 1.0, v2
-; GFX6-NEXT: v_min_f32_e32 v2, v2, v6
-; GFX6-NEXT: v_lshrrev_b32_e32 v2, 16, v2
-; GFX6-NEXT: s_waitcnt expcnt(0)
-; GFX6-NEXT: v_and_b32_e32 v7, v3, v5
-; GFX6-NEXT: v_lshlrev_b32_e32 v2, v4, v2
-; GFX6-NEXT: v_or_b32_e32 v2, v7, v2
; GFX6-NEXT: v_mov_b32_e32 v8, v3
-; GFX6-NEXT: v_mov_b32_e32 v7, v2
-; GFX6-NEXT: buffer_atomic_cmpswap v[7:8], v[0:1], s[4:7], 0 addr64 glc
+; GFX6-NEXT: s_waitcnt expcnt(0)
+; GFX6-NEXT: v_lshrrev_b32_e32 v3, v5, v8
+; GFX6-NEXT: v_lshlrev_b32_e32 v3, 16, v3
+; GFX6-NEXT: v_mul_f32_e32 v3, 1.0, v3
+; GFX6-NEXT: v_min_f32_e32 v3, v3, v2
+; GFX6-NEXT: v_lshrrev_b32_e32 v3, 16, v3
+; GFX6-NEXT: v_and_b32_e32 v4, v8, v6
+; GFX6-NEXT: v_lshlrev_b32_e32 v3, v5, v3
+; GFX6-NEXT: v_or_b32_e32 v7, v4, v3
+; GFX6-NEXT: v_mov_b32_e32 v3, v7
+; GFX6-NEXT: v_mov_b32_e32 v4, v8
+; GFX6-NEXT: buffer_atomic_cmpswap v[3:4], v[0:1], s[4:7], 0 addr64 glc
; GFX6-NEXT: s_waitcnt vmcnt(0)
; GFX6-NEXT: buffer_wbinvl1
-; GFX6-NEXT: v_cmp_eq_u32_e32 vcc, v7, v3
+; GFX6-NEXT: v_cmp_eq_u32_e32 vcc, v3, v8
; GFX6-NEXT: s_or_b64 s[8:9], vcc, s[8:9]
-; GFX6-NEXT: v_mov_b32_e32 v3, v7
; GFX6-NEXT: s_andn2_b64 exec, exec, s[8:9]
; GFX6-NEXT: s_cbranch_execnz .LBB41_1
; GFX6-NEXT: ; %bb.2: ; %atomicrmw.end
@@ -12097,29 +12099,29 @@ define bfloat @global_agent_atomic_fmin_ret_bf16__offset12b_pos__align4__amdgpu_
; GFX7-NEXT: buffer_load_dword v3, v[0:1], s[4:7], 0 addr64 offset:2046
; GFX7-NEXT: v_mul_f32_e32 v2, 1.0, v2
; GFX7-NEXT: s_mov_b64 s[8:9], 0
-; GFX7-NEXT: v_and_b32_e32 v6, 0xffff0000, v2
+; GFX7-NEXT: v_and_b32_e32 v2, 0xffff0000, v2
; GFX7-NEXT: .LBB42_1: ; %atomicrmw.start
; GFX7-NEXT: ; =>This Inner Loop Header: Depth=1
; GFX7-NEXT: s_waitcnt vmcnt(0)
-; GFX7-NEXT: v_lshlrev_b32_e32 v2, 16, v3
-; GFX7-NEXT: v_mul_f32_e32 v2, 1.0, v2
-; GFX7-NEXT: v_min_f32_e32 v2, v2, v6
-; GFX7-NEXT: v_and_b32_e32 v4, 0xffff0000, v3
-; GFX7-NEXT: v_lshrrev_b32_e32 v2, 16, v2
-; GFX7-NEXT: v_or_b32_e32 v2, v4, v2
; GFX7-NEXT: v_mov_b32_e32 v5, v3
-; GFX7-NEXT: v_mov_b32_e32 v4, v2
-; GFX7-NEXT: buffer_atomic_cmpswap v[4:5], v[0:1], s[4:7], 0 addr64 offset:2046 glc
+; GFX7-NEXT: v_lshlrev_b32_e32 v3, 16, v5
+; GFX7-NEXT: v_mul_f32_e32 v3, 1.0, v3
+; GFX7-NEXT: v_min_f32_e32 v3, v3, v2
+; GFX7-NEXT: v_and_b32_e32 v4, 0xffff0000, v5
+; GFX7-NEXT: v_lshrrev_b32_e32 v3, 16, v3
+; GFX7-NEXT: v_or_b32_e32 v4, v4, v3
+; GFX7-NEXT: v_mov_b32_e32 v3, v4
+; GFX7-NEXT: v_mov_b32_e32 v4, v5
+; GFX7-NEXT: buffer_atomic_cmpswap v[3:4], v[0:1], s[4:7], 0 addr64 offset:2046 glc
; GFX7-NEXT: s_waitcnt vmcnt(0)
; GFX7-NEXT: buffer_wbinvl1
-; GFX7-NEXT: v_cmp_eq_u32_e32 vcc, v4, v3
+; GFX7-NEXT: v_cmp_eq_u32_e32 vcc, v3, v5
; GFX7-NEXT: s_or_b64 s[8:9], vcc, s[8:9]
-; GFX7-NEXT: v_mov_b32_e32 v3, v4
; GFX7-NEXT: s_andn2_b64 exec, exec, s[8:9]
; GFX7-NEXT: s_cbranch_execnz .LBB42_1
; GFX7-NEXT: ; %bb.2: ; %atomicrmw.end
; GFX7-NEXT: s_or_b64 exec, exec, s[8:9]
-; GFX7-NEXT: v_lshlrev_b32_e32 v0, 16, v4
+; GFX7-NEXT: v_lshlrev_b32_e32 v0, 16, v3
; GFX7-NEXT: s_setpc_b64 s[30:31]
;
; GFX6-LABEL: global_agent_atomic_fmin_ret_bf16__offset12b_pos__align4__amdgpu_no_fine_grained_memory:
@@ -12132,30 +12134,30 @@ define bfloat @global_agent_atomic_fmin_ret_bf16__offset12b_pos__align4__amdgpu_
; GFX6-NEXT: buffer_load_dword v3, v[0:1], s[4:7], 0 addr64 offset:2046
; GFX6-NEXT: v_mul_f32_e32 v2, 1.0, v2
; GFX6-NEXT: s_mov_b64 s[8:9], 0
-; GFX6-NEXT: v_and_b32_e32 v6, 0xffff0000, v2
+; GFX6-NEXT: v_and_b32_e32 v2, 0xffff0000, v2
; GFX6-NEXT: .LBB42_1: ; %atomicrmw.start
; GFX6-NEXT: ; =>This Inner Loop Header: Depth=1
; GFX6-NEXT: s_waitcnt vmcnt(0)
-; GFX6-NEXT: v_lshlrev_b32_e32 v2, 16, v3
-; GFX6-NEXT: v_mul_f32_e32 v2, 1.0, v2
-; GFX6-NEXT: v_min_f32_e32 v2, v2, v6
-; GFX6-NEXT: s_waitcnt expcnt(0)
-; GFX6-NEXT: v_and_b32_e32 v4, 0xffff0000, v3
-; GFX6-NEXT: v_lshrrev_b32_e32 v2, 16, v2
-; GFX6-NEXT: v_or_b32_e32 v2, v4, v2
; GFX6-NEXT: v_mov_b32_e32 v5, v3
-; GFX6-NEXT: v_mov_b32_e32 v4, v2
-; GFX6-NEXT: buffer_atomic_cmpswap v[4:5], v[0:1], s[4:7], 0 addr64 offset:2046 glc
+; GFX6-NEXT: s_waitcnt expcnt(0)
+; GFX6-NEXT: v_lshlrev_b32_e32 v3, 16, v5
+; GFX6-NEXT: v_mul_f32_e32 v3, 1.0, v3
+; GFX6-NEXT: v_min_f32_e32 v3, v3, v2
+; GFX6-NEXT: v_and_b32_e32 v4, 0xffff0000, v5
+; GFX6-NEXT: v_lshrrev_b32_e32 v3, 16, v3
+; GFX6-NEXT: v_or_b32_e32 v4, v4, v3
+; GFX6-NEXT: v_mov_b32_e32 v3, v4
+; GFX6-NEXT: v_mov_b32_e32 v4, v5
+; GFX6-NEXT: buffer_atomic_cmpswap v[3:4], v[0:1], s[4:7], 0 addr64 offset:2046 glc
; GFX6-NEXT: s_waitcnt vmcnt(0)
; GFX6-NEXT: buffer_wbinvl1
-; GFX6-NEXT: v_cmp_eq_u32_e32 vcc, v4, v3
+; GFX6-NEXT: v_cmp_eq_u32_e32 vcc, v3, v5
; GFX6-NEXT: s_or_b64 s[8:9], vcc, s[8:9]
-; GFX6-NEXT: v_mov_b32_e32 v3, v4
; GFX6-NEXT: s_andn2_b64 exec, exec, s[8:9]
; GFX6-NEXT: s_cbranch_execnz .LBB42_1
; GFX6-NEXT: ; %bb.2: ; %atomicrmw.end
; GFX6-NEXT: s_or_b64 exec, exec, s[8:9]
-; GFX6-NEXT: v_lshlrev_b32_e32 v0, 16, v4
+; GFX6-NEXT: v_lshlrev_b32_e32 v0, 16, v3
; GFX6-NEXT: s_waitcnt expcnt(0)
; GFX6-NEXT: s_setpc_b64 s[30:31]
%gep = getelementptr bfloat, ptr addrspace(1) %ptr, i64 1023
@@ -12497,24 +12499,24 @@ define void @global_agent_atomic_fmin_noret_bf16__offset12b__align4_pos__amdgpu_
; GFX7-NEXT: buffer_load_dword v3, v[0:1], s[4:7], 0 addr64 offset:2046
; GFX7-NEXT: v_mul_f32_e32 v2, 1.0, v2
; GFX7-NEXT: s_mov_b64 s[8:9], 0
-; GFX7-NEXT: v_and_b32_e32 v4, 0xffff0000, v2
+; GFX7-NEXT: v_and_b32_e32 v2, 0xffff0000, v2
; GFX7-NEXT: .LBB43_1: ; %atomicrmw.start
; GFX7-NEXT: ; =>This Inner Loop Header: Depth=1
; GFX7-NEXT: s_waitcnt vmcnt(0)
-; GFX7-NEXT: v_lshlrev_b32_e32 v2, 16, v3
-; GFX7-NEXT: v_mul_f32_e32 v2, 1.0, v2
-; GFX7-NEXT: v_min_f32_e32 v2, v2, v4
-; GFX7-NEXT: v_and_b32_e32 v5, 0xffff0000, v3
-; GFX7-NEXT: v_lshrrev_b32_e32 v2, 16, v2
-; GFX7-NEXT: v_or_b32_e32 v2, v5, v2
-; GFX7-NEXT: v_mov_b32_e32 v6, v3
-; GFX7-NEXT: v_mov_b32_e32 v5, v2
-; GFX7-NEXT: buffer_atomic_cmpswap v[5:6], v[0:1], s[4:7], 0 addr64 offset:2046 glc
+; GFX7-NEXT: v_mov_b32_e32 v5, v3
+; GFX7-NEXT: v_lshlrev_b32_e32 v3, 16, v5
+; GFX7-NEXT: v_mul_f32_e32 v3, 1.0, v3
+; GFX7-NEXT: v_min_f32_e32 v3, v3, v2
+; GFX7-NEXT: v_and_b32_e32 v4, 0xffff0000, v5
+; GFX7-NEXT: v_lshrrev_b32_e32 v3, 16, v3
+; GFX7-NEXT: v_or_b32_e32 v4, v4, v3
+; GFX7-NEXT: v_mov_b32_e32 v3, v4
+; GFX7-NEXT: v_mov_b32_e32 v4, v5
+; GFX7-NEXT: buffer_atomic_cmpswap v[3:4], v[0:1], s[4:7], 0 addr64 offset:2046 glc
; GFX7-NEXT: s_waitcnt vmcnt(0)
; GFX7-NEXT: buffer_wbinvl1
-; GFX7-NEXT: v_cmp_eq_u32_e32 vcc, v5, v3
+; GFX7-NEXT: v_cmp_eq_u32_e32 vcc, v3, v5
; GFX7-NEXT: s_or_b64 s[8:9], vcc, s[8:9]
-; GFX7-NEXT: v_mov_b32_e32 v3, v5
; GFX7-NEXT: s_andn2_b64 exec, exec, s[8:9]
; GFX7-NEXT: s_cbranch_execnz .LBB43_1
; GFX7-NEXT: ; %bb.2: ; %atomicrmw.end
@@ -12531,25 +12533,25 @@ define void @global_agent_atomic_fmin_noret_bf16__offset12b__align4_pos__amdgpu_
; GFX6-NEXT: buffer_load_dword v3, v[0:1], s[4:7], 0 addr64 offset:2046
; GFX6-NEXT: v_mul_f32_e32 v2, 1.0, v2
; GFX6-NEXT: s_mov_b64 s[8:9], 0
-; GFX6-NEXT: v_and_b32_e32 v4, 0xffff0000, v2
+; GFX6-NEXT: v_and_b32_e32 v2, 0xffff0000, v2
; GFX6-NEXT: .LBB43_1: ; %atomicrmw.start
; GFX6-NEXT: ; =>This Inner Loop Header: Depth=1
; GFX6-NEXT: s_waitcnt vmcnt(0)
-; GFX6-NEXT: v_lshlrev_b32_e32 v2, 16, v3
-; GFX6-NEXT: v_mul_f32_e32 v2, 1.0, v2
-; GFX6-NEXT: v_min_f32_e32 v2, v2, v4
+; GFX6-NEXT: v_mov_b32_e32 v5, v3
; GFX6-NEXT: s_waitcnt expcnt(0)
-; GFX6-NEXT: v_and_b32_e32 v5, 0xffff0000, v3
-; GFX6-NEXT: v_lshrrev_b32_e32 v2, 16, v2
-; GFX6-NEXT: v_or_b32_e32 v2, v5, v2
-; GFX6-NEXT: v_mov_b32_e32 v6, v3
-; GFX6-NEXT: v_mov_b32_e32 v5, v2
-; GFX6-NEXT: buffer_atomic_cmpswap v[5:6], v[0:1], s[4:7], 0 addr64 offset:2046 glc
+; GFX6-NEXT: v_lshlrev_b32_e32 v3, 16, v5
+; GFX6-NEXT: v_mul_f32_e32 v3, 1.0, v3
+; GFX6-NEXT: v_min_f32_e32 v3, v3, v2
+; GFX6-NEXT: v_and_b32_e32 v4, 0xffff0000, v5
+; GFX6-NEXT: v_lshrrev_b32_e32 v3, 16, v3
+; GFX6-NEXT: v_or_b32_e32 v4, v4, v3
+; GFX6-NEXT: v_mov_b32_e32 v3, v4
+; GFX6-NEXT: v_mov_b32_e32 v4, v5
+; GFX6-NEXT: buffer_atomic_cmpswap v[3:4], v[0:1], s[4:7], 0 addr64 offset:2046 glc
; GFX6-NEXT: s_waitcnt vmcnt(0)
; GFX6-NEXT: buffer_wbinvl1
-; GFX6-NEXT: v_cmp_eq_u32_e32 vcc, v5, v3
+; GFX6-NEXT: v_cmp_eq_u32_e32 vcc, v3, v5
; GFX6-NEXT: s_or_b64 s[8:9], vcc, s[8:9]
-; GFX6-NEXT: v_mov_b32_e32 v3, v5
; GFX6-NEXT: s_andn2_b64 exec, exec, s[8:9]
; GFX6-NEXT: s_cbranch_execnz .LBB43_1
; GFX6-NEXT: ; %bb.2: ; %atomicrmw.end
@@ -12999,36 +13001,36 @@ define bfloat @global_system_atomic_fmin_ret_bf16__offset12b_pos__amdgpu_no_fine
; GFX7-NEXT: s_mov_b32 s5, s6
; GFX7-NEXT: buffer_load_dword v3, v[0:1], s[4:7], 0 addr64
; GFX7-NEXT: v_and_b32_e32 v4, 3, v4
-; GFX7-NEXT: v_lshlrev_b32_e32 v6, 3, v4
-; GFX7-NEXT: v_lshl_b32_e32 v4, 0xffff, v6
+; GFX7-NEXT: v_lshlrev_b32_e32 v5, 3, v4
+; GFX7-NEXT: v_lshl_b32_e32 v4, 0xffff, v5
; GFX7-NEXT: v_mul_f32_e32 v2, 1.0, v2
-; GFX7-NEXT: v_not_b32_e32 v7, v4
+; GFX7-NEXT: v_not_b32_e32 v6, v4
; GFX7-NEXT: s_mov_b64 s[8:9], 0
-; GFX7-NEXT: v_and_b32_e32 v8, 0xffff0000, v2
+; GFX7-NEXT: v_and_b32_e32 v2, 0xffff0000, v2
; GFX7-NEXT: .LBB44_1: ; %atomicrmw.start
; GFX7-NEXT: ; =>This Inner Loop Header: Depth=1
; GFX7-NEXT: s_waitcnt vmcnt(0)
-; GFX7-NEXT: v_lshrrev_b32_e32 v2, v6, v3
-; GFX7-NEXT: v_lshlrev_b32_e32 v2, 16, v2
-; GFX7-NEXT: v_mul_f32_e32 v2, 1.0, v2
-; GFX7-NEXT: v_min_f32_e32 v2, v2, v8
-; GFX7-NEXT: v_lshrrev_b32_e32 v2, 16, v2
-; GFX7-NEXT: v_and_b32_e32 v4, v3, v7
-; GFX7-NEXT: v_lshlrev_b32_e32 v2, v6, v2
-; GFX7-NEXT: v_or_b32_e32 v2, v4, v2
-; GFX7-NEXT: v_mov_b32_e32 v5, v3
-; GFX7-NEXT: v_mov_b32_e32 v4, v2
-; GFX7-NEXT: buffer_atomic_cmpswap v[4:5], v[0:1], s[4:7], 0 addr64 glc
+; GFX7-NEXT: v_mov_b32_e32 v8, v3
+; GFX7-NEXT: v_lshrrev_b32_e32 v3, v5, v8
+; GFX7-NEXT: v_lshlrev_b32_e32 v3, 16, v3
+; GFX7-NEXT: v_mul_f32_e32 v3, 1.0, v3
+; GFX7-NEXT: v_min_f32_e32 v3, v3, v2
+; GFX7-NEXT: v_lshrrev_b32_e32 v3, 16, v3
+; GFX7-NEXT: v_and_b32_e32 v4, v8, v6
+; GFX7-NEXT: v_lshlrev_b32_e32 v3, v5, v3
+; GFX7-NEXT: v_or_b32_e32 v7, v4, v3
+; GFX7-NEXT: v_mov_b32_e32 v3, v7
+; GFX7-NEXT: v_mov_b32_e32 v4, v8
+; GFX7-NEXT: buffer_atomic_cmpswap v[3:4], v[0:1], s[4:7], 0 addr64 glc
; GFX7-NEXT: s_waitcnt vmcnt(0)
; GFX7-NEXT: buffer_wbinvl1
-; GFX7-NEXT: v_cmp_eq_u32_e32 vcc, v4, v3
+; GFX7-NEXT: v_cmp_eq_u32_e32 vcc, v3, v8
; GFX7-NEXT: s_or_b64 s[8:9], vcc, s[8:9]
-; GFX7-NEXT: v_mov_b32_e32 v3, v4
; GFX7-NEXT: s_andn2_b64 exec, exec, s[8:9]
; GFX7-NEXT: s_cbranch_execnz .LBB44_1
; GFX7-NEXT: ; %bb.2: ; %atomicrmw.end
; GFX7-NEXT: s_or_b64 exec, exec, s[8:9]
-; GFX7-NEXT: v_lshrrev_b32_e32 v0, v6, v4
+; GFX7-NEXT: v_lshrrev_b32_e32 v0, v5, v3
; GFX7-NEXT: v_lshlrev_b32_e32 v0, 16, v0
; GFX7-NEXT: s_setpc_b64 s[30:31]
;
@@ -13044,37 +13046,37 @@ define bfloat @global_system_atomic_fmin_ret_bf16__offset12b_pos__amdgpu_no_fine
; GFX6-NEXT: s_mov_b32 s5, s6
; GFX6-NEXT: buffer_load_dword v3, v[0:1], s[4:7], 0 addr64
; GFX6-NEXT: v_and_b32_e32 v4, 3, v4
-; GFX6-NEXT: v_lshlrev_b32_e32 v6, 3, v4
-; GFX6-NEXT: v_lshl_b32_e32 v4, 0xffff, v6
+; GFX6-NEXT: v_lshlrev_b32_e32 v5, 3, v4
+; GFX6-NEXT: v_lshl_b32_e32 v4, 0xffff, v5
; GFX6-NEXT: v_mul_f32_e32 v2, 1.0, v2
-; GFX6-NEXT: v_not_b32_e32 v7, v4
+; GFX6-NEXT: v_not_b32_e32 v6, v4
; GFX6-NEXT: s_mov_b64 s[8:9], 0
-; GFX6-NEXT: v_and_b32_e32 v8, 0xffff0000, v2
+; GFX6-NEXT: v_and_b32_e32 v2, 0xffff0000, v2
; GFX6-NEXT: .LBB44_1: ; %atomicrmw.start
; GFX6-NEXT: ; =>This Inner Loop Header: Depth=1
; GFX6-NEXT: s_waitcnt vmcnt(0)
-; GFX6-NEXT: v_lshrrev_b32_e32 v2, v6, v3
-; GFX6-NEXT: v_lshlrev_b32_e32 v2, 16, v2
-; GFX6-NEXT: v_mul_f32_e32 v2, 1.0, v2
-; GFX6-NEXT: v_min_f32_e32 v2, v2, v8
-; GFX6-NEXT: v_lshrrev_b32_e32 v2, 16, v2
+; GFX6-NEXT: v_mov_b32_e32 v8, v3
; GFX6-NEXT: s_waitcnt expcnt(0)
-; GFX6-NEXT: v_and_b32_e32 v4, v3, v7
-; GFX6-NEXT: v_lshlrev_b32_e32 v2, v6, v2
-; GFX6-NEXT: v_or_b32_e32 v2, v4, v2
-; GFX6-NEXT: v_mov_b32_e32 v5, v3
-; GFX6-NEXT: v_mov_b32_e32 v4, v2
-; GFX6-NEXT: buffer_atomic_cmpswap v[4:5], v[0:1], s[4:7], 0 addr64 glc
+; GFX6-NEXT: v_lshrrev_b32_e32 v3, v5, v8
+; GFX6-NEXT: v_lshlrev_b32_e32 v3, 16, v3
+; GFX6-NEXT: v_mul_f32_e32 v3, 1.0, v3
+; GFX6-NEXT: v_min_f32_e32 v3, v3, v2
+; GFX6-NEXT: v_lshrrev_b32_e32 v3, 16, v3
+; GFX6-NEXT: v_and_b32_e32 v4, v8, v6
+; GFX6-NEXT: v_lshlrev_b32_e32 v3, v5, v3
+; GFX6-NEXT: v_or_b32_e32 v7, v4, v3
+; GFX6-NEXT: v_mov_b32_e32 v3, v7
+; GFX6-NEXT: v_mov_b32_e32 v4, v8
+; GFX6-NEXT: buffer_atomic_cmpswap v[3:4], v[0:1], s[4:7], 0 addr64 glc
; GFX6-NEXT: s_waitcnt vmcnt(0)
; GFX6-NEXT: buffer_wbinvl1
-; GFX6-NEXT: v_cmp_eq_u32_e32 vcc, v4, v3
+; GFX6-NEXT: v_cmp_eq_u32_e32 vcc, v3, v8
; GFX6-NEXT: s_or_b64 s[8:9], vcc, s[8:9]
-; GFX6-NEXT: v_mov_b32_e32 v3, v4
; GFX6-NEXT: s_andn2_b64 exec, exec, s[8:9]
; GFX6-NEXT: s_cbranch_execnz .LBB44_1
; GFX6-NEXT: ; %bb.2: ; %atomicrmw.end
; GFX6-NEXT: s_or_b64 exec, exec, s[8:9]
-; GFX6-NEXT: v_lshrrev_b32_e32 v0, v6, v4
+; GFX6-NEXT: v_lshrrev_b32_e32 v0, v5, v3
; GFX6-NEXT: v_lshlrev_b32_e32 v0, 16, v0
; GFX6-NEXT: s_waitcnt expcnt(0)
; GFX6-NEXT: s_setpc_b64 s[30:31]
@@ -13508,31 +13510,31 @@ define void @global_system_atomic_fmin_noret_bf16__offset12b_pos__amdgpu_no_fine
; GFX7-NEXT: s_mov_b32 s5, s6
; GFX7-NEXT: buffer_load_dword v3, v[0:1], s[4:7], 0 addr64
; GFX7-NEXT: v_and_b32_e32 v4, 3, v4
-; GFX7-NEXT: v_lshlrev_b32_e32 v4, 3, v4
-; GFX7-NEXT: v_lshl_b32_e32 v5, 0xffff, v4
+; GFX7-NEXT: v_lshlrev_b32_e32 v5, 3, v4
+; GFX7-NEXT: v_lshl_b32_e32 v4, 0xffff, v5
; GFX7-NEXT: v_mul_f32_e32 v2, 1.0, v2
-; GFX7-NEXT: v_not_b32_e32 v5, v5
+; GFX7-NEXT: v_not_b32_e32 v6, v4
; GFX7-NEXT: s_mov_b64 s[8:9], 0
-; GFX7-NEXT: v_and_b32_e32 v6, 0xffff0000, v2
+; GFX7-NEXT: v_and_b32_e32 v2, 0xffff0000, v2
; GFX7-NEXT: .LBB45_1: ; %atomicrmw.start
; GFX7-NEXT: ; =>This Inner Loop Header: Depth=1
; GFX7-NEXT: s_waitcnt vmcnt(0)
-; GFX7-NEXT: v_lshrrev_b32_e32 v2, v4, v3
-; GFX7-NEXT: v_lshlrev_b32_e32 v2, 16, v2
-; GFX7-NEXT: v_mul_f32_e32 v2, 1.0, v2
-; GFX7-NEXT: v_min_f32_e32 v2, v2, v6
-; GFX7-NEXT: v_lshrrev_b32_e32 v2, 16, v2
-; GFX7-NEXT: v_and_b32_e32 v7, v3, v5
-; GFX7-NEXT: v_lshlrev_b32_e32 v2, v4, v2
-; GFX7-NEXT: v_or_b32_e32 v2, v7, v2
; GFX7-NEXT: v_mov_b32_e32 v8, v3
-; GFX7-NEXT: v_mov_b32_e32 v7, v2
-; GFX7-NEXT: buffer_atomic_cmpswap v[7:8], v[0:1], s[4:7], 0 addr64 glc
+; GFX7-NEXT: v_lshrrev_b32_e32 v3, v5, v8
+; GFX7-NEXT: v_lshlrev_b32_e32 v3, 16, v3
+; GFX7-NEXT: v_mul_f32_e32 v3, 1.0, v3
+; GFX7-NEXT: v_min_f32_e32 v3, v3, v2
+; GFX7-NEXT: v_lshrrev_b32_e32 v3, 16, v3
+; GFX7-NEXT: v_and_b32_e32 v4, v8, v6
+; GFX7-NEXT: v_lshlrev_b32_e32 v3, v5, v3
+; GFX7-NEXT: v_or_b32_e32 v7, v4, v3
+; GFX7-NEXT: v_mov_b32_e32 v3, v7
+; GFX7-NEXT: v_mov_b32_e32 v4, v8
+; GFX7-NEXT: buffer_atomic_cmpswap v[3:4], v[0:1], s[4:7], 0 addr64 glc
; GFX7-NEXT: s_waitcnt vmcnt(0)
; GFX7-NEXT: buffer_wbinvl1
-; GFX7-NEXT: v_cmp_eq_u32_e32 vcc, v7, v3
+; GFX7-NEXT: v_cmp_eq_u32_e32 vcc, v3, v8
; GFX7-NEXT: s_or_b64 s[8:9], vcc, s[8:9]
-; GFX7-NEXT: v_mov_b32_e32 v3, v7
; GFX7-NEXT: s_andn2_b64 exec, exec, s[8:9]
; GFX7-NEXT: s_cbranch_execnz .LBB45_1
; GFX7-NEXT: ; %bb.2: ; %atomicrmw.end
@@ -13551,32 +13553,32 @@ define void @global_system_atomic_fmin_noret_bf16__offset12b_pos__amdgpu_no_fine
; GFX6-NEXT: s_mov_b32 s5, s6
; GFX6-NEXT: buffer_load_dword v3, v[0:1], s[4:7], 0 addr64
; GFX6-NEXT: v_and_b32_e32 v4, 3, v4
-; GFX6-NEXT: v_lshlrev_b32_e32 v4, 3, v4
-; GFX6-NEXT: v_lshl_b32_e32 v5, 0xffff, v4
+; GFX6-NEXT: v_lshlrev_b32_e32 v5, 3, v4
+; GFX6-NEXT: v_lshl_b32_e32 v4, 0xffff, v5
; GFX6-NEXT: v_mul_f32_e32 v2, 1.0, v2
-; GFX6-NEXT: v_not_b32_e32 v5, v5
+; GFX6-NEXT: v_not_b32_e32 v6, v4
; GFX6-NEXT: s_mov_b64 s[8:9], 0
-; GFX6-NEXT: v_and_b32_e32 v6, 0xffff0000, v2
+; GFX6-NEXT: v_and_b32_e32 v2, 0xffff0000, v2
; GFX6-NEXT: .LBB45_1: ; %atomicrmw.start
; GFX6-NEXT: ; =>This Inner Loop Header: Depth=1
; GFX6-NEXT: s_waitcnt vmcnt(0)
-; GFX6-NEXT: v_lshrrev_b32_e32 v2, v4, v3
-; GFX6-NEXT: v_lshlrev_b32_e32 v2, 16, v2
-; GFX6-NEXT: v_mul_f32_e32 v2, 1.0, v2
-; GFX6-NEXT: v_min_f32_e32 v2, v2, v6
-; GFX6-NEXT: v_lshrrev_b32_e32 v2, 16, v2
-; GFX6-NEXT: s_waitcnt expcnt(0)
-; GFX6-NEXT: v_and_b32_e32 v7, v3, v5
-; GFX6-NEXT: v_lshlrev_b32_e32 v2, v4, v2
-; GFX6-NEXT: v_or_b32_e32 v2, v7, v2
; GFX6-NEXT: v_mov_b32_e32 v8, v3
-; GFX6-NEXT: v_mov_b32_e32 v7, v2
-; GFX6-NEXT: buffer_atomic_cmpswap v[7:8], v[0:1], s[4:7], 0 addr64 glc
+; GFX6-NEXT: s_waitcnt expcnt(0)
+; GFX6-NEXT: v_lshrrev_b32_e32 v3, v5, v8
+; GFX6-NEXT: v_lshlrev_b32_e32 v3, 16, v3
+; GFX6-NEXT: v_mul_f32_e32 v3, 1.0, v3
+; GFX6-NEXT: v_min_f32_e32 v3, v3, v2
+; GFX6-NEXT: v_lshrrev_b32_e32 v3, 16, v3
+; GFX6-NEXT: v_and_b32_e32 v4, v8, v6
+; GFX6-NEXT: v_lshlrev_b32_e32 v3, v5, v3
+; GFX6-NEXT: v_or_b32_e32 v7, v4, v3
+; GFX6-NEXT: v_mov_b32_e32 v3, v7
+; GFX6-NEXT: v_mov_b32_e32 v4, v8
+; GFX6-NEXT: buffer_atomic_cmpswap v[3:4], v[0:1], s[4:7], 0 addr64 glc
; GFX6-NEXT: s_waitcnt vmcnt(0)
; GFX6-NEXT: buffer_wbinvl1
-; GFX6-NEXT: v_cmp_eq_u32_e32 vcc, v7, v3
+; GFX6-NEXT: v_cmp_eq_u32_e32 vcc, v3, v8
; GFX6-NEXT: s_or_b64 s[8:9], vcc, s[8:9]
-; GFX6-NEXT: v_mov_b32_e32 v3, v7
; GFX6-NEXT: s_andn2_b64 exec, exec, s[8:9]
; GFX6-NEXT: s_cbranch_execnz .LBB45_1
; GFX6-NEXT: ; %bb.2: ; %atomicrmw.end
diff --git a/llvm/test/CodeGen/AMDGPU/global-atomicrmw-fsub.ll b/llvm/test/CodeGen/AMDGPU/global-atomicrmw-fsub.ll
index 765185327a03e..b8944dff58bff 100644
--- a/llvm/test/CodeGen/AMDGPU/global-atomicrmw-fsub.ll
+++ b/llvm/test/CodeGen/AMDGPU/global-atomicrmw-fsub.ll
@@ -894,20 +894,20 @@ define void @global_agent_atomic_fsub_noret_f32(ptr addrspace(1) %ptr, float %va
; GFX7-NEXT: s_mov_b32 s7, 0xf000
; GFX7-NEXT: s_mov_b32 s4, s6
; GFX7-NEXT: s_mov_b32 s5, s6
-; GFX7-NEXT: buffer_load_dword v4, v[0:1], s[4:7], 0 addr64
+; GFX7-NEXT: buffer_load_dword v3, v[0:1], s[4:7], 0 addr64
; GFX7-NEXT: s_mov_b64 s[8:9], 0
; GFX7-NEXT: .LBB3_1: ; %atomicrmw.start
; GFX7-NEXT: ; =>This Inner Loop Header: Depth=1
; GFX7-NEXT: s_waitcnt vmcnt(0)
-; GFX7-NEXT: v_sub_f32_e32 v3, v4, v2
-; GFX7-NEXT: v_mov_b32_e32 v6, v4
; GFX7-NEXT: v_mov_b32_e32 v5, v3
-; GFX7-NEXT: buffer_atomic_cmpswap v[5:6], v[0:1], s[4:7], 0 addr64 glc
+; GFX7-NEXT: v_sub_f32_e32 v4, v5, v2
+; GFX7-NEXT: v_mov_b32_e32 v3, v4
+; GFX7-NEXT: v_mov_b32_e32 v4, v5
+; GFX7-NEXT: buffer_atomic_cmpswap v[3:4], v[0:1], s[4:7], 0 addr64 glc
; GFX7-NEXT: s_waitcnt vmcnt(0)
; GFX7-NEXT: buffer_wbinvl1
-; GFX7-NEXT: v_cmp_eq_u32_e32 vcc, v5, v4
+; GFX7-NEXT: v_cmp_eq_u32_e32 vcc, v3, v5
; GFX7-NEXT: s_or_b64 s[8:9], vcc, s[8:9]
-; GFX7-NEXT: v_mov_b32_e32 v4, v5
; GFX7-NEXT: s_andn2_b64 exec, exec, s[8:9]
; GFX7-NEXT: s_cbranch_execnz .LBB3_1
; GFX7-NEXT: ; %bb.2: ; %atomicrmw.end
@@ -921,21 +921,21 @@ define void @global_agent_atomic_fsub_noret_f32(ptr addrspace(1) %ptr, float %va
; GFX6-NEXT: s_mov_b32 s7, 0xf000
; GFX6-NEXT: s_mov_b32 s4, s6
; GFX6-NEXT: s_mov_b32 s5, s6
-; GFX6-NEXT: buffer_load_dword v4, v[0:1], s[4:7], 0 addr64
+; GFX6-NEXT: buffer_load_dword v3, v[0:1], s[4:7], 0 addr64
; GFX6-NEXT: s_mov_b64 s[8:9], 0
; GFX6-NEXT: .LBB3_1: ; %atomicrmw.start
; GFX6-NEXT: ; =>This Inner Loop Header: Depth=1
; GFX6-NEXT: s_waitcnt vmcnt(0)
-; GFX6-NEXT: v_sub_f32_e32 v3, v4, v2
-; GFX6-NEXT: s_waitcnt expcnt(0)
-; GFX6-NEXT: v_mov_b32_e32 v6, v4
; GFX6-NEXT: v_mov_b32_e32 v5, v3
-; GFX6-NEXT: buffer_atomic_cmpswap v[5:6], v[0:1], s[4:7], 0 addr64 glc
+; GFX6-NEXT: s_waitcnt expcnt(0)
+; GFX6-NEXT: v_sub_f32_e32 v4, v5, v2
+; GFX6-NEXT: v_mov_b32_e32 v3, v4
+; GFX6-NEXT: v_mov_b32_e32 v4, v5
+; GFX6-NEXT: buffer_atomic_cmpswap v[3:4], v[0:1], s[4:7], 0 addr64 glc
; GFX6-NEXT: s_waitcnt vmcnt(0)
; GFX6-NEXT: buffer_wbinvl1
-; GFX6-NEXT: v_cmp_eq_u32_e32 vcc, v5, v4
+; GFX6-NEXT: v_cmp_eq_u32_e32 vcc, v3, v5
; GFX6-NEXT: s_or_b64 s[8:9], vcc, s[8:9]
-; GFX6-NEXT: v_mov_b32_e32 v4, v5
; GFX6-NEXT: s_andn2_b64 exec, exec, s[8:9]
; GFX6-NEXT: s_cbranch_execnz .LBB3_1
; GFX6-NEXT: ; %bb.2: ; %atomicrmw.end
@@ -1117,20 +1117,20 @@ define void @global_agent_atomic_fsub_noret_f32__offset12b_pos(ptr addrspace(1)
; GFX7-NEXT: s_mov_b32 s7, 0xf000
; GFX7-NEXT: s_mov_b32 s4, s6
; GFX7-NEXT: s_mov_b32 s5, s6
-; GFX7-NEXT: buffer_load_dword v4, v[0:1], s[4:7], 0 addr64 offset:2044
+; GFX7-NEXT: buffer_load_dword v3, v[0:1], s[4:7], 0 addr64 offset:2044
; GFX7-NEXT: s_mov_b64 s[8:9], 0
; GFX7-NEXT: .LBB4_1: ; %atomicrmw.start
; GFX7-NEXT: ; =>This Inner Loop Header: Depth=1
; GFX7-NEXT: s_waitcnt vmcnt(0)
-; GFX7-NEXT: v_sub_f32_e32 v3, v4, v2
-; GFX7-NEXT: v_mov_b32_e32 v6, v4
; GFX7-NEXT: v_mov_b32_e32 v5, v3
-; GFX7-NEXT: buffer_atomic_cmpswap v[5:6], v[0:1], s[4:7], 0 addr64 offset:2044 glc
+; GFX7-NEXT: v_sub_f32_e32 v4, v5, v2
+; GFX7-NEXT: v_mov_b32_e32 v3, v4
+; GFX7-NEXT: v_mov_b32_e32 v4, v5
+; GFX7-NEXT: buffer_atomic_cmpswap v[3:4], v[0:1], s[4:7], 0 addr64 offset:2044 glc
; GFX7-NEXT: s_waitcnt vmcnt(0)
; GFX7-NEXT: buffer_wbinvl1
-; GFX7-NEXT: v_cmp_eq_u32_e32 vcc, v5, v4
+; GFX7-NEXT: v_cmp_eq_u32_e32 vcc, v3, v5
; GFX7-NEXT: s_or_b64 s[8:9], vcc, s[8:9]
-; GFX7-NEXT: v_mov_b32_e32 v4, v5
; GFX7-NEXT: s_andn2_b64 exec, exec, s[8:9]
; GFX7-NEXT: s_cbranch_execnz .LBB4_1
; GFX7-NEXT: ; %bb.2: ; %atomicrmw.end
@@ -1144,21 +1144,21 @@ define void @global_agent_atomic_fsub_noret_f32__offset12b_pos(ptr addrspace(1)
; GFX6-NEXT: s_mov_b32 s7, 0xf000
; GFX6-NEXT: s_mov_b32 s4, s6
; GFX6-NEXT: s_mov_b32 s5, s6
-; GFX6-NEXT: buffer_load_dword v4, v[0:1], s[4:7], 0 addr64 offset:2044
+; GFX6-NEXT: buffer_load_dword v3, v[0:1], s[4:7], 0 addr64 offset:2044
; GFX6-NEXT: s_mov_b64 s[8:9], 0
; GFX6-NEXT: .LBB4_1: ; %atomicrmw.start
; GFX6-NEXT: ; =>This Inner Loop Header: Depth=1
; GFX6-NEXT: s_waitcnt vmcnt(0)
-; GFX6-NEXT: v_sub_f32_e32 v3, v4, v2
-; GFX6-NEXT: s_waitcnt expcnt(0)
-; GFX6-NEXT: v_mov_b32_e32 v6, v4
; GFX6-NEXT: v_mov_b32_e32 v5, v3
-; GFX6-NEXT: buffer_atomic_cmpswap v[5:6], v[0:1], s[4:7], 0 addr64 offset:2044 glc
+; GFX6-NEXT: s_waitcnt expcnt(0)
+; GFX6-NEXT: v_sub_f32_e32 v4, v5, v2
+; GFX6-NEXT: v_mov_b32_e32 v3, v4
+; GFX6-NEXT: v_mov_b32_e32 v4, v5
+; GFX6-NEXT: buffer_atomic_cmpswap v[3:4], v[0:1], s[4:7], 0 addr64 offset:2044 glc
; GFX6-NEXT: s_waitcnt vmcnt(0)
; GFX6-NEXT: buffer_wbinvl1
-; GFX6-NEXT: v_cmp_eq_u32_e32 vcc, v5, v4
+; GFX6-NEXT: v_cmp_eq_u32_e32 vcc, v3, v5
; GFX6-NEXT: s_or_b64 s[8:9], vcc, s[8:9]
-; GFX6-NEXT: v_mov_b32_e32 v4, v5
; GFX6-NEXT: s_andn2_b64 exec, exec, s[8:9]
; GFX6-NEXT: s_cbranch_execnz .LBB4_1
; GFX6-NEXT: ; %bb.2: ; %atomicrmw.end
@@ -1341,7 +1341,7 @@ define void @global_agent_atomic_fsub_noret_f32__offset12b_neg(ptr addrspace(1)
; GFX7-NEXT: s_mov_b32 s5, -1
; GFX7-NEXT: s_mov_b32 s7, 0xf000
; GFX7-NEXT: s_mov_b32 s6, 0
-; GFX7-NEXT: buffer_load_dword v4, v[0:1], s[4:7], 0 addr64
+; GFX7-NEXT: buffer_load_dword v3, v[0:1], s[4:7], 0 addr64
; GFX7-NEXT: v_add_i32_e32 v0, vcc, 0xfffff800, v0
; GFX7-NEXT: v_addc_u32_e32 v1, vcc, -1, v1, vcc
; GFX7-NEXT: s_mov_b64 s[8:9], 0
@@ -1350,15 +1350,15 @@ define void @global_agent_atomic_fsub_noret_f32__offset12b_neg(ptr addrspace(1)
; GFX7-NEXT: .LBB5_1: ; %atomicrmw.start
; GFX7-NEXT: ; =>This Inner Loop Header: Depth=1
; GFX7-NEXT: s_waitcnt vmcnt(0)
-; GFX7-NEXT: v_sub_f32_e32 v3, v4, v2
-; GFX7-NEXT: v_mov_b32_e32 v6, v4
; GFX7-NEXT: v_mov_b32_e32 v5, v3
-; GFX7-NEXT: buffer_atomic_cmpswap v[5:6], v[0:1], s[4:7], 0 addr64 glc
+; GFX7-NEXT: v_sub_f32_e32 v4, v5, v2
+; GFX7-NEXT: v_mov_b32_e32 v3, v4
+; GFX7-NEXT: v_mov_b32_e32 v4, v5
+; GFX7-NEXT: buffer_atomic_cmpswap v[3:4], v[0:1], s[4:7], 0 addr64 glc
; GFX7-NEXT: s_waitcnt vmcnt(0)
; GFX7-NEXT: buffer_wbinvl1
-; GFX7-NEXT: v_cmp_eq_u32_e32 vcc, v5, v4
+; GFX7-NEXT: v_cmp_eq_u32_e32 vcc, v3, v5
; GFX7-NEXT: s_or_b64 s[8:9], vcc, s[8:9]
-; GFX7-NEXT: v_mov_b32_e32 v4, v5
; GFX7-NEXT: s_andn2_b64 exec, exec, s[8:9]
; GFX7-NEXT: s_cbranch_execnz .LBB5_1
; GFX7-NEXT: ; %bb.2: ; %atomicrmw.end
@@ -1372,7 +1372,7 @@ define void @global_agent_atomic_fsub_noret_f32__offset12b_neg(ptr addrspace(1)
; GFX6-NEXT: s_mov_b32 s5, -1
; GFX6-NEXT: s_mov_b32 s7, 0xf000
; GFX6-NEXT: s_mov_b32 s6, 0
-; GFX6-NEXT: buffer_load_dword v4, v[0:1], s[4:7], 0 addr64
+; GFX6-NEXT: buffer_load_dword v3, v[0:1], s[4:7], 0 addr64
; GFX6-NEXT: v_add_i32_e32 v0, vcc, 0xfffff800, v0
; GFX6-NEXT: v_addc_u32_e32 v1, vcc, -1, v1, vcc
; GFX6-NEXT: s_mov_b64 s[8:9], 0
@@ -1381,16 +1381,16 @@ define void @global_agent_atomic_fsub_noret_f32__offset12b_neg(ptr addrspace(1)
; GFX6-NEXT: .LBB5_1: ; %atomicrmw.start
; GFX6-NEXT: ; =>This Inner Loop Header: Depth=1
; GFX6-NEXT: s_waitcnt vmcnt(0)
-; GFX6-NEXT: v_sub_f32_e32 v3, v4, v2
-; GFX6-NEXT: s_waitcnt expcnt(0)
-; GFX6-NEXT: v_mov_b32_e32 v6, v4
; GFX6-NEXT: v_mov_b32_e32 v5, v3
-; GFX6-NEXT: buffer_atomic_cmpswap v[5:6], v[0:1], s[4:7], 0 addr64 glc
+; GFX6-NEXT: s_waitcnt expcnt(0)
+; GFX6-NEXT: v_sub_f32_e32 v4, v5, v2
+; GFX6-NEXT: v_mov_b32_e32 v3, v4
+; GFX6-NEXT: v_mov_b32_e32 v4, v5
+; GFX6-NEXT: buffer_atomic_cmpswap v[3:4], v[0:1], s[4:7], 0 addr64 glc
; GFX6-NEXT: s_waitcnt vmcnt(0)
; GFX6-NEXT: buffer_wbinvl1
-; GFX6-NEXT: v_cmp_eq_u32_e32 vcc, v5, v4
+; GFX6-NEXT: v_cmp_eq_u32_e32 vcc, v3, v5
; GFX6-NEXT: s_or_b64 s[8:9], vcc, s[8:9]
-; GFX6-NEXT: v_mov_b32_e32 v4, v5
; GFX6-NEXT: s_andn2_b64 exec, exec, s[8:9]
; GFX6-NEXT: s_cbranch_execnz .LBB5_1
; GFX6-NEXT: ; %bb.2: ; %atomicrmw.end
@@ -1813,20 +1813,20 @@ define void @global_system_atomic_fsub_noret_f32__offset12b_pos(ptr addrspace(1)
; GFX7-NEXT: s_mov_b32 s7, 0xf000
; GFX7-NEXT: s_mov_b32 s4, s6
; GFX7-NEXT: s_mov_b32 s5, s6
-; GFX7-NEXT: buffer_load_dword v4, v[0:1], s[4:7], 0 addr64 offset:2044
+; GFX7-NEXT: buffer_load_dword v3, v[0:1], s[4:7], 0 addr64 offset:2044
; GFX7-NEXT: s_mov_b64 s[8:9], 0
; GFX7-NEXT: .LBB7_1: ; %atomicrmw.start
; GFX7-NEXT: ; =>This Inner Loop Header: Depth=1
; GFX7-NEXT: s_waitcnt vmcnt(0)
-; GFX7-NEXT: v_sub_f32_e32 v3, v4, v2
-; GFX7-NEXT: v_mov_b32_e32 v6, v4
; GFX7-NEXT: v_mov_b32_e32 v5, v3
-; GFX7-NEXT: buffer_atomic_cmpswap v[5:6], v[0:1], s[4:7], 0 addr64 offset:2044 glc
+; GFX7-NEXT: v_sub_f32_e32 v4, v5, v2
+; GFX7-NEXT: v_mov_b32_e32 v3, v4
+; GFX7-NEXT: v_mov_b32_e32 v4, v5
+; GFX7-NEXT: buffer_atomic_cmpswap v[3:4], v[0:1], s[4:7], 0 addr64 offset:2044 glc
; GFX7-NEXT: s_waitcnt vmcnt(0)
; GFX7-NEXT: buffer_wbinvl1
-; GFX7-NEXT: v_cmp_eq_u32_e32 vcc, v5, v4
+; GFX7-NEXT: v_cmp_eq_u32_e32 vcc, v3, v5
; GFX7-NEXT: s_or_b64 s[8:9], vcc, s[8:9]
-; GFX7-NEXT: v_mov_b32_e32 v4, v5
; GFX7-NEXT: s_andn2_b64 exec, exec, s[8:9]
; GFX7-NEXT: s_cbranch_execnz .LBB7_1
; GFX7-NEXT: ; %bb.2: ; %atomicrmw.end
@@ -1840,21 +1840,21 @@ define void @global_system_atomic_fsub_noret_f32__offset12b_pos(ptr addrspace(1)
; GFX6-NEXT: s_mov_b32 s7, 0xf000
; GFX6-NEXT: s_mov_b32 s4, s6
; GFX6-NEXT: s_mov_b32 s5, s6
-; GFX6-NEXT: buffer_load_dword v4, v[0:1], s[4:7], 0 addr64 offset:2044
+; GFX6-NEXT: buffer_load_dword v3, v[0:1], s[4:7], 0 addr64 offset:2044
; GFX6-NEXT: s_mov_b64 s[8:9], 0
; GFX6-NEXT: .LBB7_1: ; %atomicrmw.start
; GFX6-NEXT: ; =>This Inner Loop Header: Depth=1
; GFX6-NEXT: s_waitcnt vmcnt(0)
-; GFX6-NEXT: v_sub_f32_e32 v3, v4, v2
-; GFX6-NEXT: s_waitcnt expcnt(0)
-; GFX6-NEXT: v_mov_b32_e32 v6, v4
; GFX6-NEXT: v_mov_b32_e32 v5, v3
-; GFX6-NEXT: buffer_atomic_cmpswap v[5:6], v[0:1], s[4:7], 0 addr64 offset:2044 glc
+; GFX6-NEXT: s_waitcnt expcnt(0)
+; GFX6-NEXT: v_sub_f32_e32 v4, v5, v2
+; GFX6-NEXT: v_mov_b32_e32 v3, v4
+; GFX6-NEXT: v_mov_b32_e32 v4, v5
+; GFX6-NEXT: buffer_atomic_cmpswap v[3:4], v[0:1], s[4:7], 0 addr64 offset:2044 glc
; GFX6-NEXT: s_waitcnt vmcnt(0)
; GFX6-NEXT: buffer_wbinvl1
-; GFX6-NEXT: v_cmp_eq_u32_e32 vcc, v5, v4
+; GFX6-NEXT: v_cmp_eq_u32_e32 vcc, v3, v5
; GFX6-NEXT: s_or_b64 s[8:9], vcc, s[8:9]
-; GFX6-NEXT: v_mov_b32_e32 v4, v5
; GFX6-NEXT: s_andn2_b64 exec, exec, s[8:9]
; GFX6-NEXT: s_cbranch_execnz .LBB7_1
; GFX6-NEXT: ; %bb.2: ; %atomicrmw.end
@@ -2749,20 +2749,20 @@ define void @global_agent_atomic_fsub_noret_f32__ftz(ptr addrspace(1) %ptr, floa
; GFX7-NEXT: s_mov_b32 s7, 0xf000
; GFX7-NEXT: s_mov_b32 s4, s6
; GFX7-NEXT: s_mov_b32 s5, s6
-; GFX7-NEXT: buffer_load_dword v4, v[0:1], s[4:7], 0 addr64
+; GFX7-NEXT: buffer_load_dword v3, v[0:1], s[4:7], 0 addr64
; GFX7-NEXT: s_mov_b64 s[8:9], 0
; GFX7-NEXT: .LBB11_1: ; %atomicrmw.start
; GFX7-NEXT: ; =>This Inner Loop Header: Depth=1
; GFX7-NEXT: s_waitcnt vmcnt(0)
-; GFX7-NEXT: v_sub_f32_e32 v3, v4, v2
-; GFX7-NEXT: v_mov_b32_e32 v6, v4
; GFX7-NEXT: v_mov_b32_e32 v5, v3
-; GFX7-NEXT: buffer_atomic_cmpswap v[5:6], v[0:1], s[4:7], 0 addr64 glc
+; GFX7-NEXT: v_sub_f32_e32 v4, v5, v2
+; GFX7-NEXT: v_mov_b32_e32 v3, v4
+; GFX7-NEXT: v_mov_b32_e32 v4, v5
+; GFX7-NEXT: buffer_atomic_cmpswap v[3:4], v[0:1], s[4:7], 0 addr64 glc
; GFX7-NEXT: s_waitcnt vmcnt(0)
; GFX7-NEXT: buffer_wbinvl1
-; GFX7-NEXT: v_cmp_eq_u32_e32 vcc, v5, v4
+; GFX7-NEXT: v_cmp_eq_u32_e32 vcc, v3, v5
; GFX7-NEXT: s_or_b64 s[8:9], vcc, s[8:9]
-; GFX7-NEXT: v_mov_b32_e32 v4, v5
; GFX7-NEXT: s_andn2_b64 exec, exec, s[8:9]
; GFX7-NEXT: s_cbranch_execnz .LBB11_1
; GFX7-NEXT: ; %bb.2: ; %atomicrmw.end
@@ -2776,21 +2776,21 @@ define void @global_agent_atomic_fsub_noret_f32__ftz(ptr addrspace(1) %ptr, floa
; GFX6-NEXT: s_mov_b32 s7, 0xf000
; GFX6-NEXT: s_mov_b32 s4, s6
; GFX6-NEXT: s_mov_b32 s5, s6
-; GFX6-NEXT: buffer_load_dword v4, v[0:1], s[4:7], 0 addr64
+; GFX6-NEXT: buffer_load_dword v3, v[0:1], s[4:7], 0 addr64
; GFX6-NEXT: s_mov_b64 s[8:9], 0
; GFX6-NEXT: .LBB11_1: ; %atomicrmw.start
; GFX6-NEXT: ; =>This Inner Loop Header: Depth=1
; GFX6-NEXT: s_waitcnt vmcnt(0)
-; GFX6-NEXT: v_sub_f32_e32 v3, v4, v2
-; GFX6-NEXT: s_waitcnt expcnt(0)
-; GFX6-NEXT: v_mov_b32_e32 v6, v4
; GFX6-NEXT: v_mov_b32_e32 v5, v3
-; GFX6-NEXT: buffer_atomic_cmpswap v[5:6], v[0:1], s[4:7], 0 addr64 glc
+; GFX6-NEXT: s_waitcnt expcnt(0)
+; GFX6-NEXT: v_sub_f32_e32 v4, v5, v2
+; GFX6-NEXT: v_mov_b32_e32 v3, v4
+; GFX6-NEXT: v_mov_b32_e32 v4, v5
+; GFX6-NEXT: buffer_atomic_cmpswap v[3:4], v[0:1], s[4:7], 0 addr64 glc
; GFX6-NEXT: s_waitcnt vmcnt(0)
; GFX6-NEXT: buffer_wbinvl1
-; GFX6-NEXT: v_cmp_eq_u32_e32 vcc, v5, v4
+; GFX6-NEXT: v_cmp_eq_u32_e32 vcc, v3, v5
; GFX6-NEXT: s_or_b64 s[8:9], vcc, s[8:9]
-; GFX6-NEXT: v_mov_b32_e32 v4, v5
; GFX6-NEXT: s_andn2_b64 exec, exec, s[8:9]
; GFX6-NEXT: s_cbranch_execnz .LBB11_1
; GFX6-NEXT: ; %bb.2: ; %atomicrmw.end
@@ -2972,20 +2972,20 @@ define void @global_agent_atomic_fsub_noret_f32__offset12b_pos__ftz(ptr addrspac
; GFX7-NEXT: s_mov_b32 s7, 0xf000
; GFX7-NEXT: s_mov_b32 s4, s6
; GFX7-NEXT: s_mov_b32 s5, s6
-; GFX7-NEXT: buffer_load_dword v4, v[0:1], s[4:7], 0 addr64 offset:2044
+; GFX7-NEXT: buffer_load_dword v3, v[0:1], s[4:7], 0 addr64 offset:2044
; GFX7-NEXT: s_mov_b64 s[8:9], 0
; GFX7-NEXT: .LBB12_1: ; %atomicrmw.start
; GFX7-NEXT: ; =>This Inner Loop Header: Depth=1
; GFX7-NEXT: s_waitcnt vmcnt(0)
-; GFX7-NEXT: v_sub_f32_e32 v3, v4, v2
-; GFX7-NEXT: v_mov_b32_e32 v6, v4
; GFX7-NEXT: v_mov_b32_e32 v5, v3
-; GFX7-NEXT: buffer_atomic_cmpswap v[5:6], v[0:1], s[4:7], 0 addr64 offset:2044 glc
+; GFX7-NEXT: v_sub_f32_e32 v4, v5, v2
+; GFX7-NEXT: v_mov_b32_e32 v3, v4
+; GFX7-NEXT: v_mov_b32_e32 v4, v5
+; GFX7-NEXT: buffer_atomic_cmpswap v[3:4], v[0:1], s[4:7], 0 addr64 offset:2044 glc
; GFX7-NEXT: s_waitcnt vmcnt(0)
; GFX7-NEXT: buffer_wbinvl1
-; GFX7-NEXT: v_cmp_eq_u32_e32 vcc, v5, v4
+; GFX7-NEXT: v_cmp_eq_u32_e32 vcc, v3, v5
; GFX7-NEXT: s_or_b64 s[8:9], vcc, s[8:9]
-; GFX7-NEXT: v_mov_b32_e32 v4, v5
; GFX7-NEXT: s_andn2_b64 exec, exec, s[8:9]
; GFX7-NEXT: s_cbranch_execnz .LBB12_1
; GFX7-NEXT: ; %bb.2: ; %atomicrmw.end
@@ -2999,21 +2999,21 @@ define void @global_agent_atomic_fsub_noret_f32__offset12b_pos__ftz(ptr addrspac
; GFX6-NEXT: s_mov_b32 s7, 0xf000
; GFX6-NEXT: s_mov_b32 s4, s6
; GFX6-NEXT: s_mov_b32 s5, s6
-; GFX6-NEXT: buffer_load_dword v4, v[0:1], s[4:7], 0 addr64 offset:2044
+; GFX6-NEXT: buffer_load_dword v3, v[0:1], s[4:7], 0 addr64 offset:2044
; GFX6-NEXT: s_mov_b64 s[8:9], 0
; GFX6-NEXT: .LBB12_1: ; %atomicrmw.start
; GFX6-NEXT: ; =>This Inner Loop Header: Depth=1
; GFX6-NEXT: s_waitcnt vmcnt(0)
-; GFX6-NEXT: v_sub_f32_e32 v3, v4, v2
-; GFX6-NEXT: s_waitcnt expcnt(0)
-; GFX6-NEXT: v_mov_b32_e32 v6, v4
; GFX6-NEXT: v_mov_b32_e32 v5, v3
-; GFX6-NEXT: buffer_atomic_cmpswap v[5:6], v[0:1], s[4:7], 0 addr64 offset:2044 glc
+; GFX6-NEXT: s_waitcnt expcnt(0)
+; GFX6-NEXT: v_sub_f32_e32 v4, v5, v2
+; GFX6-NEXT: v_mov_b32_e32 v3, v4
+; GFX6-NEXT: v_mov_b32_e32 v4, v5
+; GFX6-NEXT: buffer_atomic_cmpswap v[3:4], v[0:1], s[4:7], 0 addr64 offset:2044 glc
; GFX6-NEXT: s_waitcnt vmcnt(0)
; GFX6-NEXT: buffer_wbinvl1
-; GFX6-NEXT: v_cmp_eq_u32_e32 vcc, v5, v4
+; GFX6-NEXT: v_cmp_eq_u32_e32 vcc, v3, v5
; GFX6-NEXT: s_or_b64 s[8:9], vcc, s[8:9]
-; GFX6-NEXT: v_mov_b32_e32 v4, v5
; GFX6-NEXT: s_andn2_b64 exec, exec, s[8:9]
; GFX6-NEXT: s_cbranch_execnz .LBB12_1
; GFX6-NEXT: ; %bb.2: ; %atomicrmw.end
@@ -3196,7 +3196,7 @@ define void @global_agent_atomic_fsub_noret_f32__offset12b_neg__ftz(ptr addrspac
; GFX7-NEXT: s_mov_b32 s5, -1
; GFX7-NEXT: s_mov_b32 s7, 0xf000
; GFX7-NEXT: s_mov_b32 s6, 0
-; GFX7-NEXT: buffer_load_dword v4, v[0:1], s[4:7], 0 addr64
+; GFX7-NEXT: buffer_load_dword v3, v[0:1], s[4:7], 0 addr64
; GFX7-NEXT: v_add_i32_e32 v0, vcc, 0xfffff800, v0
; GFX7-NEXT: v_addc_u32_e32 v1, vcc, -1, v1, vcc
; GFX7-NEXT: s_mov_b64 s[8:9], 0
@@ -3205,15 +3205,15 @@ define void @global_agent_atomic_fsub_noret_f32__offset12b_neg__ftz(ptr addrspac
; GFX7-NEXT: .LBB13_1: ; %atomicrmw.start
; GFX7-NEXT: ; =>This Inner Loop Header: Depth=1
; GFX7-NEXT: s_waitcnt vmcnt(0)
-; GFX7-NEXT: v_sub_f32_e32 v3, v4, v2
-; GFX7-NEXT: v_mov_b32_e32 v6, v4
; GFX7-NEXT: v_mov_b32_e32 v5, v3
-; GFX7-NEXT: buffer_atomic_cmpswap v[5:6], v[0:1], s[4:7], 0 addr64 glc
+; GFX7-NEXT: v_sub_f32_e32 v4, v5, v2
+; GFX7-NEXT: v_mov_b32_e32 v3, v4
+; GFX7-NEXT: v_mov_b32_e32 v4, v5
+; GFX7-NEXT: buffer_atomic_cmpswap v[3:4], v[0:1], s[4:7], 0 addr64 glc
; GFX7-NEXT: s_waitcnt vmcnt(0)
; GFX7-NEXT: buffer_wbinvl1
-; GFX7-NEXT: v_cmp_eq_u32_e32 vcc, v5, v4
+; GFX7-NEXT: v_cmp_eq_u32_e32 vcc, v3, v5
; GFX7-NEXT: s_or_b64 s[8:9], vcc, s[8:9]
-; GFX7-NEXT: v_mov_b32_e32 v4, v5
; GFX7-NEXT: s_andn2_b64 exec, exec, s[8:9]
; GFX7-NEXT: s_cbranch_execnz .LBB13_1
; GFX7-NEXT: ; %bb.2: ; %atomicrmw.end
@@ -3227,7 +3227,7 @@ define void @global_agent_atomic_fsub_noret_f32__offset12b_neg__ftz(ptr addrspac
; GFX6-NEXT: s_mov_b32 s5, -1
; GFX6-NEXT: s_mov_b32 s7, 0xf000
; GFX6-NEXT: s_mov_b32 s6, 0
-; GFX6-NEXT: buffer_load_dword v4, v[0:1], s[4:7], 0 addr64
+; GFX6-NEXT: buffer_load_dword v3, v[0:1], s[4:7], 0 addr64
; GFX6-NEXT: v_add_i32_e32 v0, vcc, 0xfffff800, v0
; GFX6-NEXT: v_addc_u32_e32 v1, vcc, -1, v1, vcc
; GFX6-NEXT: s_mov_b64 s[8:9], 0
@@ -3236,16 +3236,16 @@ define void @global_agent_atomic_fsub_noret_f32__offset12b_neg__ftz(ptr addrspac
; GFX6-NEXT: .LBB13_1: ; %atomicrmw.start
; GFX6-NEXT: ; =>This Inner Loop Header: Depth=1
; GFX6-NEXT: s_waitcnt vmcnt(0)
-; GFX6-NEXT: v_sub_f32_e32 v3, v4, v2
-; GFX6-NEXT: s_waitcnt expcnt(0)
-; GFX6-NEXT: v_mov_b32_e32 v6, v4
; GFX6-NEXT: v_mov_b32_e32 v5, v3
-; GFX6-NEXT: buffer_atomic_cmpswap v[5:6], v[0:1], s[4:7], 0 addr64 glc
+; GFX6-NEXT: s_waitcnt expcnt(0)
+; GFX6-NEXT: v_sub_f32_e32 v4, v5, v2
+; GFX6-NEXT: v_mov_b32_e32 v3, v4
+; GFX6-NEXT: v_mov_b32_e32 v4, v5
+; GFX6-NEXT: buffer_atomic_cmpswap v[3:4], v[0:1], s[4:7], 0 addr64 glc
; GFX6-NEXT: s_waitcnt vmcnt(0)
; GFX6-NEXT: buffer_wbinvl1
-; GFX6-NEXT: v_cmp_eq_u32_e32 vcc, v5, v4
+; GFX6-NEXT: v_cmp_eq_u32_e32 vcc, v3, v5
; GFX6-NEXT: s_or_b64 s[8:9], vcc, s[8:9]
-; GFX6-NEXT: v_mov_b32_e32 v4, v5
; GFX6-NEXT: s_andn2_b64 exec, exec, s[8:9]
; GFX6-NEXT: s_cbranch_execnz .LBB13_1
; GFX6-NEXT: ; %bb.2: ; %atomicrmw.end
@@ -3668,20 +3668,20 @@ define void @global_system_atomic_fsub_noret_f32__offset12b_pos__ftz(ptr addrspa
; GFX7-NEXT: s_mov_b32 s7, 0xf000
; GFX7-NEXT: s_mov_b32 s4, s6
; GFX7-NEXT: s_mov_b32 s5, s6
-; GFX7-NEXT: buffer_load_dword v4, v[0:1], s[4:7], 0 addr64 offset:2044
+; GFX7-NEXT: buffer_load_dword v3, v[0:1], s[4:7], 0 addr64 offset:2044
; GFX7-NEXT: s_mov_b64 s[8:9], 0
; GFX7-NEXT: .LBB15_1: ; %atomicrmw.start
; GFX7-NEXT: ; =>This Inner Loop Header: Depth=1
; GFX7-NEXT: s_waitcnt vmcnt(0)
-; GFX7-NEXT: v_sub_f32_e32 v3, v4, v2
-; GFX7-NEXT: v_mov_b32_e32 v6, v4
; GFX7-NEXT: v_mov_b32_e32 v5, v3
-; GFX7-NEXT: buffer_atomic_cmpswap v[5:6], v[0:1], s[4:7], 0 addr64 offset:2044 glc
+; GFX7-NEXT: v_sub_f32_e32 v4, v5, v2
+; GFX7-NEXT: v_mov_b32_e32 v3, v4
+; GFX7-NEXT: v_mov_b32_e32 v4, v5
+; GFX7-NEXT: buffer_atomic_cmpswap v[3:4], v[0:1], s[4:7], 0 addr64 offset:2044 glc
; GFX7-NEXT: s_waitcnt vmcnt(0)
; GFX7-NEXT: buffer_wbinvl1
-; GFX7-NEXT: v_cmp_eq_u32_e32 vcc, v5, v4
+; GFX7-NEXT: v_cmp_eq_u32_e32 vcc, v3, v5
; GFX7-NEXT: s_or_b64 s[8:9], vcc, s[8:9]
-; GFX7-NEXT: v_mov_b32_e32 v4, v5
; GFX7-NEXT: s_andn2_b64 exec, exec, s[8:9]
; GFX7-NEXT: s_cbranch_execnz .LBB15_1
; GFX7-NEXT: ; %bb.2: ; %atomicrmw.end
@@ -3695,21 +3695,21 @@ define void @global_system_atomic_fsub_noret_f32__offset12b_pos__ftz(ptr addrspa
; GFX6-NEXT: s_mov_b32 s7, 0xf000
; GFX6-NEXT: s_mov_b32 s4, s6
; GFX6-NEXT: s_mov_b32 s5, s6
-; GFX6-NEXT: buffer_load_dword v4, v[0:1], s[4:7], 0 addr64 offset:2044
+; GFX6-NEXT: buffer_load_dword v3, v[0:1], s[4:7], 0 addr64 offset:2044
; GFX6-NEXT: s_mov_b64 s[8:9], 0
; GFX6-NEXT: .LBB15_1: ; %atomicrmw.start
; GFX6-NEXT: ; =>This Inner Loop Header: Depth=1
; GFX6-NEXT: s_waitcnt vmcnt(0)
-; GFX6-NEXT: v_sub_f32_e32 v3, v4, v2
-; GFX6-NEXT: s_waitcnt expcnt(0)
-; GFX6-NEXT: v_mov_b32_e32 v6, v4
; GFX6-NEXT: v_mov_b32_e32 v5, v3
-; GFX6-NEXT: buffer_atomic_cmpswap v[5:6], v[0:1], s[4:7], 0 addr64 offset:2044 glc
+; GFX6-NEXT: s_waitcnt expcnt(0)
+; GFX6-NEXT: v_sub_f32_e32 v4, v5, v2
+; GFX6-NEXT: v_mov_b32_e32 v3, v4
+; GFX6-NEXT: v_mov_b32_e32 v4, v5
+; GFX6-NEXT: buffer_atomic_cmpswap v[3:4], v[0:1], s[4:7], 0 addr64 offset:2044 glc
; GFX6-NEXT: s_waitcnt vmcnt(0)
; GFX6-NEXT: buffer_wbinvl1
-; GFX6-NEXT: v_cmp_eq_u32_e32 vcc, v5, v4
+; GFX6-NEXT: v_cmp_eq_u32_e32 vcc, v3, v5
; GFX6-NEXT: s_or_b64 s[8:9], vcc, s[8:9]
-; GFX6-NEXT: v_mov_b32_e32 v4, v5
; GFX6-NEXT: s_andn2_b64 exec, exec, s[8:9]
; GFX6-NEXT: s_cbranch_execnz .LBB15_1
; GFX6-NEXT: ; %bb.2: ; %atomicrmw.end
@@ -4663,23 +4663,23 @@ define void @global_agent_atomic_fsub_noret_f64(ptr addrspace(1) %ptr, double %v
; GFX7-NEXT: s_mov_b32 s7, 0xf000
; GFX7-NEXT: s_mov_b32 s4, s6
; GFX7-NEXT: s_mov_b32 s5, s6
-; GFX7-NEXT: buffer_load_dwordx2 v[6:7], v[0:1], s[4:7], 0 addr64
+; GFX7-NEXT: buffer_load_dwordx2 v[4:5], v[0:1], s[4:7], 0 addr64
; GFX7-NEXT: s_mov_b64 s[8:9], 0
; GFX7-NEXT: .LBB19_1: ; %atomicrmw.start
; GFX7-NEXT: ; =>This Inner Loop Header: Depth=1
; GFX7-NEXT: s_waitcnt vmcnt(0)
-; GFX7-NEXT: v_add_f64 v[4:5], v[6:7], -v[2:3]
-; GFX7-NEXT: v_mov_b32_e32 v11, v7
-; GFX7-NEXT: v_mov_b32_e32 v10, v6
; GFX7-NEXT: v_mov_b32_e32 v9, v5
; GFX7-NEXT: v_mov_b32_e32 v8, v4
-; GFX7-NEXT: buffer_atomic_cmpswap_x2 v[8:11], v[0:1], s[4:7], 0 addr64 glc
+; GFX7-NEXT: v_add_f64 v[6:7], v[8:9], -v[2:3]
+; GFX7-NEXT: v_mov_b32_e32 v4, v6
+; GFX7-NEXT: v_mov_b32_e32 v5, v7
+; GFX7-NEXT: v_mov_b32_e32 v6, v8
+; GFX7-NEXT: v_mov_b32_e32 v7, v9
+; GFX7-NEXT: buffer_atomic_cmpswap_x2 v[4:7], v[0:1], s[4:7], 0 addr64 glc
; GFX7-NEXT: s_waitcnt vmcnt(0)
; GFX7-NEXT: buffer_wbinvl1
-; GFX7-NEXT: v_cmp_eq_u64_e32 vcc, v[8:9], v[6:7]
-; GFX7-NEXT: v_mov_b32_e32 v6, v8
+; GFX7-NEXT: v_cmp_eq_u64_e32 vcc, v[4:5], v[8:9]
; GFX7-NEXT: s_or_b64 s[8:9], vcc, s[8:9]
-; GFX7-NEXT: v_mov_b32_e32 v7, v9
; GFX7-NEXT: s_andn2_b64 exec, exec, s[8:9]
; GFX7-NEXT: s_cbranch_execnz .LBB19_1
; GFX7-NEXT: ; %bb.2: ; %atomicrmw.end
@@ -4693,24 +4693,24 @@ define void @global_agent_atomic_fsub_noret_f64(ptr addrspace(1) %ptr, double %v
; GFX6-NEXT: s_mov_b32 s7, 0xf000
; GFX6-NEXT: s_mov_b32 s4, s6
; GFX6-NEXT: s_mov_b32 s5, s6
-; GFX6-NEXT: buffer_load_dwordx2 v[6:7], v[0:1], s[4:7], 0 addr64
+; GFX6-NEXT: buffer_load_dwordx2 v[4:5], v[0:1], s[4:7], 0 addr64
; GFX6-NEXT: s_mov_b64 s[8:9], 0
; GFX6-NEXT: .LBB19_1: ; %atomicrmw.start
; GFX6-NEXT: ; =>This Inner Loop Header: Depth=1
; GFX6-NEXT: s_waitcnt vmcnt(0)
-; GFX6-NEXT: v_add_f64 v[4:5], v[6:7], -v[2:3]
-; GFX6-NEXT: s_waitcnt expcnt(0)
-; GFX6-NEXT: v_mov_b32_e32 v11, v7
-; GFX6-NEXT: v_mov_b32_e32 v10, v6
; GFX6-NEXT: v_mov_b32_e32 v9, v5
; GFX6-NEXT: v_mov_b32_e32 v8, v4
-; GFX6-NEXT: buffer_atomic_cmpswap_x2 v[8:11], v[0:1], s[4:7], 0 addr64 glc
+; GFX6-NEXT: s_waitcnt expcnt(0)
+; GFX6-NEXT: v_add_f64 v[6:7], v[8:9], -v[2:3]
+; GFX6-NEXT: v_mov_b32_e32 v4, v6
+; GFX6-NEXT: v_mov_b32_e32 v5, v7
+; GFX6-NEXT: v_mov_b32_e32 v6, v8
+; GFX6-NEXT: v_mov_b32_e32 v7, v9
+; GFX6-NEXT: buffer_atomic_cmpswap_x2 v[4:7], v[0:1], s[4:7], 0 addr64 glc
; GFX6-NEXT: s_waitcnt vmcnt(0)
; GFX6-NEXT: buffer_wbinvl1
-; GFX6-NEXT: v_cmp_eq_u64_e32 vcc, v[8:9], v[6:7]
-; GFX6-NEXT: v_mov_b32_e32 v6, v8
+; GFX6-NEXT: v_cmp_eq_u64_e32 vcc, v[4:5], v[8:9]
; GFX6-NEXT: s_or_b64 s[8:9], vcc, s[8:9]
-; GFX6-NEXT: v_mov_b32_e32 v7, v9
; GFX6-NEXT: s_andn2_b64 exec, exec, s[8:9]
; GFX6-NEXT: s_cbranch_execnz .LBB19_1
; GFX6-NEXT: ; %bb.2: ; %atomicrmw.end
@@ -4895,23 +4895,23 @@ define void @global_agent_atomic_fsub_noret_f64__offset12b_pos(ptr addrspace(1)
; GFX7-NEXT: s_mov_b32 s7, 0xf000
; GFX7-NEXT: s_mov_b32 s4, s6
; GFX7-NEXT: s_mov_b32 s5, s6
-; GFX7-NEXT: buffer_load_dwordx2 v[6:7], v[0:1], s[4:7], 0 addr64 offset:2040
+; GFX7-NEXT: buffer_load_dwordx2 v[4:5], v[0:1], s[4:7], 0 addr64 offset:2040
; GFX7-NEXT: s_mov_b64 s[8:9], 0
; GFX7-NEXT: .LBB20_1: ; %atomicrmw.start
; GFX7-NEXT: ; =>This Inner Loop Header: Depth=1
; GFX7-NEXT: s_waitcnt vmcnt(0)
-; GFX7-NEXT: v_add_f64 v[4:5], v[6:7], -v[2:3]
-; GFX7-NEXT: v_mov_b32_e32 v11, v7
-; GFX7-NEXT: v_mov_b32_e32 v10, v6
; GFX7-NEXT: v_mov_b32_e32 v9, v5
; GFX7-NEXT: v_mov_b32_e32 v8, v4
-; GFX7-NEXT: buffer_atomic_cmpswap_x2 v[8:11], v[0:1], s[4:7], 0 addr64 offset:2040 glc
+; GFX7-NEXT: v_add_f64 v[6:7], v[8:9], -v[2:3]
+; GFX7-NEXT: v_mov_b32_e32 v4, v6
+; GFX7-NEXT: v_mov_b32_e32 v5, v7
+; GFX7-NEXT: v_mov_b32_e32 v6, v8
+; GFX7-NEXT: v_mov_b32_e32 v7, v9
+; GFX7-NEXT: buffer_atomic_cmpswap_x2 v[4:7], v[0:1], s[4:7], 0 addr64 offset:2040 glc
; GFX7-NEXT: s_waitcnt vmcnt(0)
; GFX7-NEXT: buffer_wbinvl1
-; GFX7-NEXT: v_cmp_eq_u64_e32 vcc, v[8:9], v[6:7]
-; GFX7-NEXT: v_mov_b32_e32 v6, v8
+; GFX7-NEXT: v_cmp_eq_u64_e32 vcc, v[4:5], v[8:9]
; GFX7-NEXT: s_or_b64 s[8:9], vcc, s[8:9]
-; GFX7-NEXT: v_mov_b32_e32 v7, v9
; GFX7-NEXT: s_andn2_b64 exec, exec, s[8:9]
; GFX7-NEXT: s_cbranch_execnz .LBB20_1
; GFX7-NEXT: ; %bb.2: ; %atomicrmw.end
@@ -4925,24 +4925,24 @@ define void @global_agent_atomic_fsub_noret_f64__offset12b_pos(ptr addrspace(1)
; GFX6-NEXT: s_mov_b32 s7, 0xf000
; GFX6-NEXT: s_mov_b32 s4, s6
; GFX6-NEXT: s_mov_b32 s5, s6
-; GFX6-NEXT: buffer_load_dwordx2 v[6:7], v[0:1], s[4:7], 0 addr64 offset:2040
+; GFX6-NEXT: buffer_load_dwordx2 v[4:5], v[0:1], s[4:7], 0 addr64 offset:2040
; GFX6-NEXT: s_mov_b64 s[8:9], 0
; GFX6-NEXT: .LBB20_1: ; %atomicrmw.start
; GFX6-NEXT: ; =>This Inner Loop Header: Depth=1
; GFX6-NEXT: s_waitcnt vmcnt(0)
-; GFX6-NEXT: v_add_f64 v[4:5], v[6:7], -v[2:3]
-; GFX6-NEXT: s_waitcnt expcnt(0)
-; GFX6-NEXT: v_mov_b32_e32 v11, v7
-; GFX6-NEXT: v_mov_b32_e32 v10, v6
; GFX6-NEXT: v_mov_b32_e32 v9, v5
; GFX6-NEXT: v_mov_b32_e32 v8, v4
-; GFX6-NEXT: buffer_atomic_cmpswap_x2 v[8:11], v[0:1], s[4:7], 0 addr64 offset:2040 glc
+; GFX6-NEXT: s_waitcnt expcnt(0)
+; GFX6-NEXT: v_add_f64 v[6:7], v[8:9], -v[2:3]
+; GFX6-NEXT: v_mov_b32_e32 v4, v6
+; GFX6-NEXT: v_mov_b32_e32 v5, v7
+; GFX6-NEXT: v_mov_b32_e32 v6, v8
+; GFX6-NEXT: v_mov_b32_e32 v7, v9
+; GFX6-NEXT: buffer_atomic_cmpswap_x2 v[4:7], v[0:1], s[4:7], 0 addr64 offset:2040 glc
; GFX6-NEXT: s_waitcnt vmcnt(0)
; GFX6-NEXT: buffer_wbinvl1
-; GFX6-NEXT: v_cmp_eq_u64_e32 vcc, v[8:9], v[6:7]
-; GFX6-NEXT: v_mov_b32_e32 v6, v8
+; GFX6-NEXT: v_cmp_eq_u64_e32 vcc, v[4:5], v[8:9]
; GFX6-NEXT: s_or_b64 s[8:9], vcc, s[8:9]
-; GFX6-NEXT: v_mov_b32_e32 v7, v9
; GFX6-NEXT: s_andn2_b64 exec, exec, s[8:9]
; GFX6-NEXT: s_cbranch_execnz .LBB20_1
; GFX6-NEXT: ; %bb.2: ; %atomicrmw.end
@@ -5128,7 +5128,7 @@ define void @global_agent_atomic_fsub_noret_f64__offset12b_neg(ptr addrspace(1)
; GFX7-NEXT: s_mov_b32 s5, -1
; GFX7-NEXT: s_mov_b32 s7, 0xf000
; GFX7-NEXT: s_mov_b32 s6, 0
-; GFX7-NEXT: buffer_load_dwordx2 v[6:7], v[0:1], s[4:7], 0 addr64
+; GFX7-NEXT: buffer_load_dwordx2 v[4:5], v[0:1], s[4:7], 0 addr64
; GFX7-NEXT: v_add_i32_e32 v0, vcc, 0xfffff800, v0
; GFX7-NEXT: v_addc_u32_e32 v1, vcc, -1, v1, vcc
; GFX7-NEXT: s_mov_b64 s[8:9], 0
@@ -5137,18 +5137,18 @@ define void @global_agent_atomic_fsub_noret_f64__offset12b_neg(ptr addrspace(1)
; GFX7-NEXT: .LBB21_1: ; %atomicrmw.start
; GFX7-NEXT: ; =>This Inner Loop Header: Depth=1
; GFX7-NEXT: s_waitcnt vmcnt(0)
-; GFX7-NEXT: v_add_f64 v[4:5], v[6:7], -v[2:3]
-; GFX7-NEXT: v_mov_b32_e32 v11, v7
-; GFX7-NEXT: v_mov_b32_e32 v10, v6
; GFX7-NEXT: v_mov_b32_e32 v9, v5
; GFX7-NEXT: v_mov_b32_e32 v8, v4
-; GFX7-NEXT: buffer_atomic_cmpswap_x2 v[8:11], v[0:1], s[4:7], 0 addr64 glc
+; GFX7-NEXT: v_add_f64 v[6:7], v[8:9], -v[2:3]
+; GFX7-NEXT: v_mov_b32_e32 v4, v6
+; GFX7-NEXT: v_mov_b32_e32 v5, v7
+; GFX7-NEXT: v_mov_b32_e32 v6, v8
+; GFX7-NEXT: v_mov_b32_e32 v7, v9
+; GFX7-NEXT: buffer_atomic_cmpswap_x2 v[4:7], v[0:1], s[4:7], 0 addr64 glc
; GFX7-NEXT: s_waitcnt vmcnt(0)
; GFX7-NEXT: buffer_wbinvl1
-; GFX7-NEXT: v_cmp_eq_u64_e32 vcc, v[8:9], v[6:7]
-; GFX7-NEXT: v_mov_b32_e32 v6, v8
+; GFX7-NEXT: v_cmp_eq_u64_e32 vcc, v[4:5], v[8:9]
; GFX7-NEXT: s_or_b64 s[8:9], vcc, s[8:9]
-; GFX7-NEXT: v_mov_b32_e32 v7, v9
; GFX7-NEXT: s_andn2_b64 exec, exec, s[8:9]
; GFX7-NEXT: s_cbranch_execnz .LBB21_1
; GFX7-NEXT: ; %bb.2: ; %atomicrmw.end
@@ -5162,7 +5162,7 @@ define void @global_agent_atomic_fsub_noret_f64__offset12b_neg(ptr addrspace(1)
; GFX6-NEXT: s_mov_b32 s5, -1
; GFX6-NEXT: s_mov_b32 s7, 0xf000
; GFX6-NEXT: s_mov_b32 s6, 0
-; GFX6-NEXT: buffer_load_dwordx2 v[6:7], v[0:1], s[4:7], 0 addr64
+; GFX6-NEXT: buffer_load_dwordx2 v[4:5], v[0:1], s[4:7], 0 addr64
; GFX6-NEXT: v_add_i32_e32 v0, vcc, 0xfffff800, v0
; GFX6-NEXT: v_addc_u32_e32 v1, vcc, -1, v1, vcc
; GFX6-NEXT: s_mov_b64 s[8:9], 0
@@ -5171,19 +5171,19 @@ define void @global_agent_atomic_fsub_noret_f64__offset12b_neg(ptr addrspace(1)
; GFX6-NEXT: .LBB21_1: ; %atomicrmw.start
; GFX6-NEXT: ; =>This Inner Loop Header: Depth=1
; GFX6-NEXT: s_waitcnt vmcnt(0)
-; GFX6-NEXT: v_add_f64 v[4:5], v[6:7], -v[2:3]
-; GFX6-NEXT: s_waitcnt expcnt(0)
-; GFX6-NEXT: v_mov_b32_e32 v11, v7
-; GFX6-NEXT: v_mov_b32_e32 v10, v6
; GFX6-NEXT: v_mov_b32_e32 v9, v5
; GFX6-NEXT: v_mov_b32_e32 v8, v4
-; GFX6-NEXT: buffer_atomic_cmpswap_x2 v[8:11], v[0:1], s[4:7], 0 addr64 glc
+; GFX6-NEXT: s_waitcnt expcnt(0)
+; GFX6-NEXT: v_add_f64 v[6:7], v[8:9], -v[2:3]
+; GFX6-NEXT: v_mov_b32_e32 v4, v6
+; GFX6-NEXT: v_mov_b32_e32 v5, v7
+; GFX6-NEXT: v_mov_b32_e32 v6, v8
+; GFX6-NEXT: v_mov_b32_e32 v7, v9
+; GFX6-NEXT: buffer_atomic_cmpswap_x2 v[4:7], v[0:1], s[4:7], 0 addr64 glc
; GFX6-NEXT: s_waitcnt vmcnt(0)
; GFX6-NEXT: buffer_wbinvl1
-; GFX6-NEXT: v_cmp_eq_u64_e32 vcc, v[8:9], v[6:7]
-; GFX6-NEXT: v_mov_b32_e32 v6, v8
+; GFX6-NEXT: v_cmp_eq_u64_e32 vcc, v[4:5], v[8:9]
; GFX6-NEXT: s_or_b64 s[8:9], vcc, s[8:9]
-; GFX6-NEXT: v_mov_b32_e32 v7, v9
; GFX6-NEXT: s_andn2_b64 exec, exec, s[8:9]
; GFX6-NEXT: s_cbranch_execnz .LBB21_1
; GFX6-NEXT: ; %bb.2: ; %atomicrmw.end
@@ -5543,36 +5543,36 @@ define half @global_agent_atomic_fsub_ret_f16(ptr addrspace(1) %ptr, half %val)
; GFX7-NEXT: s_mov_b32 s4, s6
; GFX7-NEXT: s_mov_b32 s5, s6
; GFX7-NEXT: buffer_load_dword v4, v[0:1], s[4:7], 0 addr64
-; GFX7-NEXT: v_cvt_f16_f32_e32 v2, v2
-; GFX7-NEXT: v_and_b32_e32 v3, 3, v3
-; GFX7-NEXT: v_lshlrev_b32_e32 v5, 3, v3
+; GFX7-NEXT: v_cvt_f16_f32_e32 v5, v2
+; GFX7-NEXT: v_and_b32_e32 v2, 3, v3
+; GFX7-NEXT: v_lshlrev_b32_e32 v2, 3, v2
; GFX7-NEXT: s_mov_b64 s[8:9], 0
-; GFX7-NEXT: v_cvt_f32_f16_e32 v6, v2
-; GFX7-NEXT: v_lshl_b32_e32 v2, 0xffff, v5
-; GFX7-NEXT: v_not_b32_e32 v7, v2
+; GFX7-NEXT: v_cvt_f32_f16_e32 v3, v5
+; GFX7-NEXT: v_lshl_b32_e32 v5, 0xffff, v2
+; GFX7-NEXT: v_not_b32_e32 v6, v5
; GFX7-NEXT: .LBB22_1: ; %atomicrmw.start
; GFX7-NEXT: ; =>This Inner Loop Header: Depth=1
; GFX7-NEXT: s_waitcnt vmcnt(0)
-; GFX7-NEXT: v_lshrrev_b32_e32 v2, v5, v4
-; GFX7-NEXT: v_cvt_f32_f16_e32 v2, v2
-; GFX7-NEXT: v_and_b32_e32 v3, v4, v7
-; GFX7-NEXT: v_sub_f32_e32 v2, v2, v6
-; GFX7-NEXT: v_cvt_f16_f32_e32 v2, v2
-; GFX7-NEXT: v_lshlrev_b32_e32 v2, v5, v2
-; GFX7-NEXT: v_or_b32_e32 v3, v3, v2
-; GFX7-NEXT: v_mov_b32_e32 v2, v3
-; GFX7-NEXT: v_mov_b32_e32 v3, v4
-; GFX7-NEXT: buffer_atomic_cmpswap v[2:3], v[0:1], s[4:7], 0 addr64 glc
+; GFX7-NEXT: v_mov_b32_e32 v8, v4
+; GFX7-NEXT: v_lshrrev_b32_e32 v4, v2, v8
+; GFX7-NEXT: v_cvt_f32_f16_e32 v4, v4
+; GFX7-NEXT: v_and_b32_e32 v5, v8, v6
+; GFX7-NEXT: v_sub_f32_e32 v4, v4, v3
+; GFX7-NEXT: v_cvt_f16_f32_e32 v4, v4
+; GFX7-NEXT: v_lshlrev_b32_e32 v4, v2, v4
+; GFX7-NEXT: v_or_b32_e32 v7, v5, v4
+; GFX7-NEXT: v_mov_b32_e32 v4, v7
+; GFX7-NEXT: v_mov_b32_e32 v5, v8
+; GFX7-NEXT: buffer_atomic_cmpswap v[4:5], v[0:1], s[4:7], 0 addr64 glc
; GFX7-NEXT: s_waitcnt vmcnt(0)
; GFX7-NEXT: buffer_wbinvl1
-; GFX7-NEXT: v_cmp_eq_u32_e32 vcc, v2, v4
+; GFX7-NEXT: v_cmp_eq_u32_e32 vcc, v4, v8
; GFX7-NEXT: s_or_b64 s[8:9], vcc, s[8:9]
-; GFX7-NEXT: v_mov_b32_e32 v4, v2
; GFX7-NEXT: s_andn2_b64 exec, exec, s[8:9]
; GFX7-NEXT: s_cbranch_execnz .LBB22_1
; GFX7-NEXT: ; %bb.2: ; %atomicrmw.end
; GFX7-NEXT: s_or_b64 exec, exec, s[8:9]
-; GFX7-NEXT: v_lshrrev_b32_e32 v0, v5, v2
+; GFX7-NEXT: v_lshrrev_b32_e32 v0, v2, v4
; GFX7-NEXT: v_cvt_f32_f16_e32 v0, v0
; GFX7-NEXT: s_setpc_b64 s[30:31]
;
@@ -5586,36 +5586,37 @@ define half @global_agent_atomic_fsub_ret_f16(ptr addrspace(1) %ptr, half %val)
; GFX6-NEXT: s_mov_b32 s4, s6
; GFX6-NEXT: s_mov_b32 s5, s6
; GFX6-NEXT: buffer_load_dword v4, v[0:1], s[4:7], 0 addr64
-; GFX6-NEXT: v_cvt_f16_f32_e32 v2, v2
-; GFX6-NEXT: v_and_b32_e32 v3, 3, v3
-; GFX6-NEXT: v_lshlrev_b32_e32 v5, 3, v3
+; GFX6-NEXT: v_cvt_f16_f32_e32 v5, v2
+; GFX6-NEXT: v_and_b32_e32 v2, 3, v3
+; GFX6-NEXT: v_lshlrev_b32_e32 v2, 3, v2
; GFX6-NEXT: s_mov_b64 s[8:9], 0
-; GFX6-NEXT: v_cvt_f32_f16_e32 v6, v2
-; GFX6-NEXT: v_lshl_b32_e32 v2, 0xffff, v5
-; GFX6-NEXT: v_not_b32_e32 v7, v2
+; GFX6-NEXT: v_cvt_f32_f16_e32 v3, v5
+; GFX6-NEXT: v_lshl_b32_e32 v5, 0xffff, v2
+; GFX6-NEXT: v_not_b32_e32 v6, v5
; GFX6-NEXT: .LBB22_1: ; %atomicrmw.start
; GFX6-NEXT: ; =>This Inner Loop Header: Depth=1
-; GFX6-NEXT: s_waitcnt vmcnt(0) expcnt(0)
-; GFX6-NEXT: v_lshrrev_b32_e32 v2, v5, v4
-; GFX6-NEXT: v_cvt_f32_f16_e32 v2, v2
-; GFX6-NEXT: v_and_b32_e32 v3, v4, v7
-; GFX6-NEXT: v_sub_f32_e32 v2, v2, v6
-; GFX6-NEXT: v_cvt_f16_f32_e32 v2, v2
-; GFX6-NEXT: v_lshlrev_b32_e32 v2, v5, v2
-; GFX6-NEXT: v_or_b32_e32 v3, v3, v2
-; GFX6-NEXT: v_mov_b32_e32 v2, v3
-; GFX6-NEXT: v_mov_b32_e32 v3, v4
-; GFX6-NEXT: buffer_atomic_cmpswap v[2:3], v[0:1], s[4:7], 0 addr64 glc
+; GFX6-NEXT: s_waitcnt vmcnt(0)
+; GFX6-NEXT: v_mov_b32_e32 v8, v4
+; GFX6-NEXT: s_waitcnt expcnt(0)
+; GFX6-NEXT: v_lshrrev_b32_e32 v4, v2, v8
+; GFX6-NEXT: v_cvt_f32_f16_e32 v4, v4
+; GFX6-NEXT: v_and_b32_e32 v5, v8, v6
+; GFX6-NEXT: v_sub_f32_e32 v4, v4, v3
+; GFX6-NEXT: v_cvt_f16_f32_e32 v4, v4
+; GFX6-NEXT: v_lshlrev_b32_e32 v4, v2, v4
+; GFX6-NEXT: v_or_b32_e32 v7, v5, v4
+; GFX6-NEXT: v_mov_b32_e32 v4, v7
+; GFX6-NEXT: v_mov_b32_e32 v5, v8
+; GFX6-NEXT: buffer_atomic_cmpswap v[4:5], v[0:1], s[4:7], 0 addr64 glc
; GFX6-NEXT: s_waitcnt vmcnt(0)
; GFX6-NEXT: buffer_wbinvl1
-; GFX6-NEXT: v_cmp_eq_u32_e32 vcc, v2, v4
+; GFX6-NEXT: v_cmp_eq_u32_e32 vcc, v4, v8
; GFX6-NEXT: s_or_b64 s[8:9], vcc, s[8:9]
-; GFX6-NEXT: v_mov_b32_e32 v4, v2
; GFX6-NEXT: s_andn2_b64 exec, exec, s[8:9]
; GFX6-NEXT: s_cbranch_execnz .LBB22_1
; GFX6-NEXT: ; %bb.2: ; %atomicrmw.end
; GFX6-NEXT: s_or_b64 exec, exec, s[8:9]
-; GFX6-NEXT: v_lshrrev_b32_e32 v0, v5, v2
+; GFX6-NEXT: v_lshrrev_b32_e32 v0, v2, v4
; GFX6-NEXT: v_cvt_f32_f16_e32 v0, v0
; GFX6-NEXT: s_waitcnt expcnt(0)
; GFX6-NEXT: s_setpc_b64 s[30:31]
@@ -5978,36 +5979,36 @@ define half @global_agent_atomic_fsub_ret_f16__offset12b_pos(ptr addrspace(1) %p
; GFX7-NEXT: s_mov_b32 s4, s6
; GFX7-NEXT: s_mov_b32 s5, s6
; GFX7-NEXT: buffer_load_dword v3, v[0:1], s[4:7], 0 addr64
-; GFX7-NEXT: v_cvt_f16_f32_e32 v2, v2
-; GFX7-NEXT: v_and_b32_e32 v4, 3, v4
-; GFX7-NEXT: v_lshlrev_b32_e32 v6, 3, v4
+; GFX7-NEXT: v_cvt_f16_f32_e32 v5, v2
+; GFX7-NEXT: v_and_b32_e32 v2, 3, v4
+; GFX7-NEXT: v_lshlrev_b32_e32 v2, 3, v2
+; GFX7-NEXT: v_lshl_b32_e32 v4, 0xffff, v2
+; GFX7-NEXT: v_cvt_f32_f16_e32 v5, v5
+; GFX7-NEXT: v_not_b32_e32 v6, v4
; GFX7-NEXT: s_mov_b64 s[8:9], 0
-; GFX7-NEXT: v_cvt_f32_f16_e32 v7, v2
-; GFX7-NEXT: v_lshl_b32_e32 v2, 0xffff, v6
-; GFX7-NEXT: v_not_b32_e32 v8, v2
; GFX7-NEXT: .LBB23_1: ; %atomicrmw.start
; GFX7-NEXT: ; =>This Inner Loop Header: Depth=1
; GFX7-NEXT: s_waitcnt vmcnt(0)
-; GFX7-NEXT: v_lshrrev_b32_e32 v2, v6, v3
-; GFX7-NEXT: v_cvt_f32_f16_e32 v2, v2
-; GFX7-NEXT: v_and_b32_e32 v4, v3, v8
-; GFX7-NEXT: v_sub_f32_e32 v2, v2, v7
-; GFX7-NEXT: v_cvt_f16_f32_e32 v2, v2
-; GFX7-NEXT: v_lshlrev_b32_e32 v2, v6, v2
-; GFX7-NEXT: v_or_b32_e32 v2, v4, v2
-; GFX7-NEXT: v_mov_b32_e32 v5, v3
-; GFX7-NEXT: v_mov_b32_e32 v4, v2
-; GFX7-NEXT: buffer_atomic_cmpswap v[4:5], v[0:1], s[4:7], 0 addr64 glc
+; GFX7-NEXT: v_mov_b32_e32 v8, v3
+; GFX7-NEXT: v_lshrrev_b32_e32 v3, v2, v8
+; GFX7-NEXT: v_cvt_f32_f16_e32 v3, v3
+; GFX7-NEXT: v_and_b32_e32 v4, v8, v6
+; GFX7-NEXT: v_sub_f32_e32 v3, v3, v5
+; GFX7-NEXT: v_cvt_f16_f32_e32 v3, v3
+; GFX7-NEXT: v_lshlrev_b32_e32 v3, v2, v3
+; GFX7-NEXT: v_or_b32_e32 v7, v4, v3
+; GFX7-NEXT: v_mov_b32_e32 v3, v7
+; GFX7-NEXT: v_mov_b32_e32 v4, v8
+; GFX7-NEXT: buffer_atomic_cmpswap v[3:4], v[0:1], s[4:7], 0 addr64 glc
; GFX7-NEXT: s_waitcnt vmcnt(0)
; GFX7-NEXT: buffer_wbinvl1
-; GFX7-NEXT: v_cmp_eq_u32_e32 vcc, v4, v3
+; GFX7-NEXT: v_cmp_eq_u32_e32 vcc, v3, v8
; GFX7-NEXT: s_or_b64 s[8:9], vcc, s[8:9]
-; GFX7-NEXT: v_mov_b32_e32 v3, v4
; GFX7-NEXT: s_andn2_b64 exec, exec, s[8:9]
; GFX7-NEXT: s_cbranch_execnz .LBB23_1
; GFX7-NEXT: ; %bb.2: ; %atomicrmw.end
; GFX7-NEXT: s_or_b64 exec, exec, s[8:9]
-; GFX7-NEXT: v_lshrrev_b32_e32 v0, v6, v4
+; GFX7-NEXT: v_lshrrev_b32_e32 v0, v2, v3
; GFX7-NEXT: v_cvt_f32_f16_e32 v0, v0
; GFX7-NEXT: s_setpc_b64 s[30:31]
;
@@ -6022,37 +6023,37 @@ define half @global_agent_atomic_fsub_ret_f16__offset12b_pos(ptr addrspace(1) %p
; GFX6-NEXT: s_mov_b32 s4, s6
; GFX6-NEXT: s_mov_b32 s5, s6
; GFX6-NEXT: buffer_load_dword v3, v[0:1], s[4:7], 0 addr64
-; GFX6-NEXT: v_cvt_f16_f32_e32 v2, v2
-; GFX6-NEXT: v_and_b32_e32 v4, 3, v4
-; GFX6-NEXT: v_lshlrev_b32_e32 v6, 3, v4
+; GFX6-NEXT: v_cvt_f16_f32_e32 v5, v2
+; GFX6-NEXT: v_and_b32_e32 v2, 3, v4
+; GFX6-NEXT: v_lshlrev_b32_e32 v2, 3, v2
+; GFX6-NEXT: v_lshl_b32_e32 v4, 0xffff, v2
+; GFX6-NEXT: v_cvt_f32_f16_e32 v5, v5
+; GFX6-NEXT: v_not_b32_e32 v6, v4
; GFX6-NEXT: s_mov_b64 s[8:9], 0
-; GFX6-NEXT: v_cvt_f32_f16_e32 v7, v2
-; GFX6-NEXT: v_lshl_b32_e32 v2, 0xffff, v6
-; GFX6-NEXT: v_not_b32_e32 v8, v2
; GFX6-NEXT: .LBB23_1: ; %atomicrmw.start
; GFX6-NEXT: ; =>This Inner Loop Header: Depth=1
; GFX6-NEXT: s_waitcnt vmcnt(0)
-; GFX6-NEXT: v_lshrrev_b32_e32 v2, v6, v3
-; GFX6-NEXT: v_cvt_f32_f16_e32 v2, v2
+; GFX6-NEXT: v_mov_b32_e32 v8, v3
; GFX6-NEXT: s_waitcnt expcnt(0)
-; GFX6-NEXT: v_and_b32_e32 v4, v3, v8
-; GFX6-NEXT: v_sub_f32_e32 v2, v2, v7
-; GFX6-NEXT: v_cvt_f16_f32_e32 v2, v2
-; GFX6-NEXT: v_lshlrev_b32_e32 v2, v6, v2
-; GFX6-NEXT: v_or_b32_e32 v2, v4, v2
-; GFX6-NEXT: v_mov_b32_e32 v5, v3
-; GFX6-NEXT: v_mov_b32_e32 v4, v2
-; GFX6-NEXT: buffer_atomic_cmpswap v[4:5], v[0:1], s[4:7], 0 addr64 glc
+; GFX6-NEXT: v_lshrrev_b32_e32 v3, v2, v8
+; GFX6-NEXT: v_cvt_f32_f16_e32 v3, v3
+; GFX6-NEXT: v_and_b32_e32 v4, v8, v6
+; GFX6-NEXT: v_sub_f32_e32 v3, v3, v5
+; GFX6-NEXT: v_cvt_f16_f32_e32 v3, v3
+; GFX6-NEXT: v_lshlrev_b32_e32 v3, v2, v3
+; GFX6-NEXT: v_or_b32_e32 v7, v4, v3
+; GFX6-NEXT: v_mov_b32_e32 v3, v7
+; GFX6-NEXT: v_mov_b32_e32 v4, v8
+; GFX6-NEXT: buffer_atomic_cmpswap v[3:4], v[0:1], s[4:7], 0 addr64 glc
; GFX6-NEXT: s_waitcnt vmcnt(0)
; GFX6-NEXT: buffer_wbinvl1
-; GFX6-NEXT: v_cmp_eq_u32_e32 vcc, v4, v3
+; GFX6-NEXT: v_cmp_eq_u32_e32 vcc, v3, v8
; GFX6-NEXT: s_or_b64 s[8:9], vcc, s[8:9]
-; GFX6-NEXT: v_mov_b32_e32 v3, v4
; GFX6-NEXT: s_andn2_b64 exec, exec, s[8:9]
; GFX6-NEXT: s_cbranch_execnz .LBB23_1
; GFX6-NEXT: ; %bb.2: ; %atomicrmw.end
; GFX6-NEXT: s_or_b64 exec, exec, s[8:9]
-; GFX6-NEXT: v_lshrrev_b32_e32 v0, v6, v4
+; GFX6-NEXT: v_lshrrev_b32_e32 v0, v2, v3
; GFX6-NEXT: v_cvt_f32_f16_e32 v0, v0
; GFX6-NEXT: s_waitcnt expcnt(0)
; GFX6-NEXT: s_setpc_b64 s[30:31]
@@ -6417,36 +6418,36 @@ define half @global_agent_atomic_fsub_ret_f16__offset12b_neg(ptr addrspace(1) %p
; GFX7-NEXT: s_mov_b32 s4, s6
; GFX7-NEXT: s_mov_b32 s5, s6
; GFX7-NEXT: buffer_load_dword v3, v[0:1], s[4:7], 0 addr64
-; GFX7-NEXT: v_cvt_f16_f32_e32 v2, v2
-; GFX7-NEXT: v_and_b32_e32 v4, 3, v4
-; GFX7-NEXT: v_lshlrev_b32_e32 v6, 3, v4
+; GFX7-NEXT: v_cvt_f16_f32_e32 v5, v2
+; GFX7-NEXT: v_and_b32_e32 v2, 3, v4
+; GFX7-NEXT: v_lshlrev_b32_e32 v2, 3, v2
+; GFX7-NEXT: v_lshl_b32_e32 v4, 0xffff, v2
+; GFX7-NEXT: v_cvt_f32_f16_e32 v5, v5
+; GFX7-NEXT: v_not_b32_e32 v6, v4
; GFX7-NEXT: s_mov_b64 s[8:9], 0
-; GFX7-NEXT: v_cvt_f32_f16_e32 v7, v2
-; GFX7-NEXT: v_lshl_b32_e32 v2, 0xffff, v6
-; GFX7-NEXT: v_not_b32_e32 v8, v2
; GFX7-NEXT: .LBB24_1: ; %atomicrmw.start
; GFX7-NEXT: ; =>This Inner Loop Header: Depth=1
; GFX7-NEXT: s_waitcnt vmcnt(0)
-; GFX7-NEXT: v_lshrrev_b32_e32 v2, v6, v3
-; GFX7-NEXT: v_cvt_f32_f16_e32 v2, v2
-; GFX7-NEXT: v_and_b32_e32 v4, v3, v8
-; GFX7-NEXT: v_sub_f32_e32 v2, v2, v7
-; GFX7-NEXT: v_cvt_f16_f32_e32 v2, v2
-; GFX7-NEXT: v_lshlrev_b32_e32 v2, v6, v2
-; GFX7-NEXT: v_or_b32_e32 v2, v4, v2
-; GFX7-NEXT: v_mov_b32_e32 v5, v3
-; GFX7-NEXT: v_mov_b32_e32 v4, v2
-; GFX7-NEXT: buffer_atomic_cmpswap v[4:5], v[0:1], s[4:7], 0 addr64 glc
+; GFX7-NEXT: v_mov_b32_e32 v8, v3
+; GFX7-NEXT: v_lshrrev_b32_e32 v3, v2, v8
+; GFX7-NEXT: v_cvt_f32_f16_e32 v3, v3
+; GFX7-NEXT: v_and_b32_e32 v4, v8, v6
+; GFX7-NEXT: v_sub_f32_e32 v3, v3, v5
+; GFX7-NEXT: v_cvt_f16_f32_e32 v3, v3
+; GFX7-NEXT: v_lshlrev_b32_e32 v3, v2, v3
+; GFX7-NEXT: v_or_b32_e32 v7, v4, v3
+; GFX7-NEXT: v_mov_b32_e32 v3, v7
+; GFX7-NEXT: v_mov_b32_e32 v4, v8
+; GFX7-NEXT: buffer_atomic_cmpswap v[3:4], v[0:1], s[4:7], 0 addr64 glc
; GFX7-NEXT: s_waitcnt vmcnt(0)
; GFX7-NEXT: buffer_wbinvl1
-; GFX7-NEXT: v_cmp_eq_u32_e32 vcc, v4, v3
+; GFX7-NEXT: v_cmp_eq_u32_e32 vcc, v3, v8
; GFX7-NEXT: s_or_b64 s[8:9], vcc, s[8:9]
-; GFX7-NEXT: v_mov_b32_e32 v3, v4
; GFX7-NEXT: s_andn2_b64 exec, exec, s[8:9]
; GFX7-NEXT: s_cbranch_execnz .LBB24_1
; GFX7-NEXT: ; %bb.2: ; %atomicrmw.end
; GFX7-NEXT: s_or_b64 exec, exec, s[8:9]
-; GFX7-NEXT: v_lshrrev_b32_e32 v0, v6, v4
+; GFX7-NEXT: v_lshrrev_b32_e32 v0, v2, v3
; GFX7-NEXT: v_cvt_f32_f16_e32 v0, v0
; GFX7-NEXT: s_setpc_b64 s[30:31]
;
@@ -6461,37 +6462,37 @@ define half @global_agent_atomic_fsub_ret_f16__offset12b_neg(ptr addrspace(1) %p
; GFX6-NEXT: s_mov_b32 s4, s6
; GFX6-NEXT: s_mov_b32 s5, s6
; GFX6-NEXT: buffer_load_dword v3, v[0:1], s[4:7], 0 addr64
-; GFX6-NEXT: v_cvt_f16_f32_e32 v2, v2
-; GFX6-NEXT: v_and_b32_e32 v4, 3, v4
-; GFX6-NEXT: v_lshlrev_b32_e32 v6, 3, v4
+; GFX6-NEXT: v_cvt_f16_f32_e32 v5, v2
+; GFX6-NEXT: v_and_b32_e32 v2, 3, v4
+; GFX6-NEXT: v_lshlrev_b32_e32 v2, 3, v2
+; GFX6-NEXT: v_lshl_b32_e32 v4, 0xffff, v2
+; GFX6-NEXT: v_cvt_f32_f16_e32 v5, v5
+; GFX6-NEXT: v_not_b32_e32 v6, v4
; GFX6-NEXT: s_mov_b64 s[8:9], 0
-; GFX6-NEXT: v_cvt_f32_f16_e32 v7, v2
-; GFX6-NEXT: v_lshl_b32_e32 v2, 0xffff, v6
-; GFX6-NEXT: v_not_b32_e32 v8, v2
; GFX6-NEXT: .LBB24_1: ; %atomicrmw.start
; GFX6-NEXT: ; =>This Inner Loop Header: Depth=1
; GFX6-NEXT: s_waitcnt vmcnt(0)
-; GFX6-NEXT: v_lshrrev_b32_e32 v2, v6, v3
-; GFX6-NEXT: v_cvt_f32_f16_e32 v2, v2
+; GFX6-NEXT: v_mov_b32_e32 v8, v3
; GFX6-NEXT: s_waitcnt expcnt(0)
-; GFX6-NEXT: v_and_b32_e32 v4, v3, v8
-; GFX6-NEXT: v_sub_f32_e32 v2, v2, v7
-; GFX6-NEXT: v_cvt_f16_f32_e32 v2, v2
-; GFX6-NEXT: v_lshlrev_b32_e32 v2, v6, v2
-; GFX6-NEXT: v_or_b32_e32 v2, v4, v2
-; GFX6-NEXT: v_mov_b32_e32 v5, v3
-; GFX6-NEXT: v_mov_b32_e32 v4, v2
-; GFX6-NEXT: buffer_atomic_cmpswap v[4:5], v[0:1], s[4:7], 0 addr64 glc
+; GFX6-NEXT: v_lshrrev_b32_e32 v3, v2, v8
+; GFX6-NEXT: v_cvt_f32_f16_e32 v3, v3
+; GFX6-NEXT: v_and_b32_e32 v4, v8, v6
+; GFX6-NEXT: v_sub_f32_e32 v3, v3, v5
+; GFX6-NEXT: v_cvt_f16_f32_e32 v3, v3
+; GFX6-NEXT: v_lshlrev_b32_e32 v3, v2, v3
+; GFX6-NEXT: v_or_b32_e32 v7, v4, v3
+; GFX6-NEXT: v_mov_b32_e32 v3, v7
+; GFX6-NEXT: v_mov_b32_e32 v4, v8
+; GFX6-NEXT: buffer_atomic_cmpswap v[3:4], v[0:1], s[4:7], 0 addr64 glc
; GFX6-NEXT: s_waitcnt vmcnt(0)
; GFX6-NEXT: buffer_wbinvl1
-; GFX6-NEXT: v_cmp_eq_u32_e32 vcc, v4, v3
+; GFX6-NEXT: v_cmp_eq_u32_e32 vcc, v3, v8
; GFX6-NEXT: s_or_b64 s[8:9], vcc, s[8:9]
-; GFX6-NEXT: v_mov_b32_e32 v3, v4
; GFX6-NEXT: s_andn2_b64 exec, exec, s[8:9]
; GFX6-NEXT: s_cbranch_execnz .LBB24_1
; GFX6-NEXT: ; %bb.2: ; %atomicrmw.end
; GFX6-NEXT: s_or_b64 exec, exec, s[8:9]
-; GFX6-NEXT: v_lshrrev_b32_e32 v0, v6, v4
+; GFX6-NEXT: v_lshrrev_b32_e32 v0, v2, v3
; GFX6-NEXT: v_cvt_f32_f16_e32 v0, v0
; GFX6-NEXT: s_waitcnt expcnt(0)
; GFX6-NEXT: s_setpc_b64 s[30:31]
@@ -6834,28 +6835,28 @@ define void @global_agent_atomic_fsub_noret_f16(ptr addrspace(1) %ptr, half %val
; GFX7-NEXT: v_cvt_f16_f32_e32 v5, v2
; GFX7-NEXT: v_and_b32_e32 v2, 3, v3
; GFX7-NEXT: v_lshlrev_b32_e32 v2, 3, v2
-; GFX7-NEXT: v_lshl_b32_e32 v3, 0xffff, v2
-; GFX7-NEXT: v_cvt_f32_f16_e32 v5, v5
-; GFX7-NEXT: v_not_b32_e32 v6, v3
; GFX7-NEXT: s_mov_b64 s[8:9], 0
+; GFX7-NEXT: v_cvt_f32_f16_e32 v3, v5
+; GFX7-NEXT: v_lshl_b32_e32 v5, 0xffff, v2
+; GFX7-NEXT: v_not_b32_e32 v6, v5
; GFX7-NEXT: .LBB25_1: ; %atomicrmw.start
; GFX7-NEXT: ; =>This Inner Loop Header: Depth=1
; GFX7-NEXT: s_waitcnt vmcnt(0)
-; GFX7-NEXT: v_lshrrev_b32_e32 v3, v2, v4
-; GFX7-NEXT: v_cvt_f32_f16_e32 v3, v3
-; GFX7-NEXT: v_and_b32_e32 v7, v4, v6
-; GFX7-NEXT: v_sub_f32_e32 v3, v3, v5
-; GFX7-NEXT: v_cvt_f16_f32_e32 v3, v3
-; GFX7-NEXT: v_lshlrev_b32_e32 v3, v2, v3
-; GFX7-NEXT: v_or_b32_e32 v3, v7, v3
; GFX7-NEXT: v_mov_b32_e32 v8, v4
-; GFX7-NEXT: v_mov_b32_e32 v7, v3
-; GFX7-NEXT: buffer_atomic_cmpswap v[7:8], v[0:1], s[4:7], 0 addr64 glc
+; GFX7-NEXT: v_lshrrev_b32_e32 v4, v2, v8
+; GFX7-NEXT: v_cvt_f32_f16_e32 v4, v4
+; GFX7-NEXT: v_and_b32_e32 v5, v8, v6
+; GFX7-NEXT: v_sub_f32_e32 v4, v4, v3
+; GFX7-NEXT: v_cvt_f16_f32_e32 v4, v4
+; GFX7-NEXT: v_lshlrev_b32_e32 v4, v2, v4
+; GFX7-NEXT: v_or_b32_e32 v7, v5, v4
+; GFX7-NEXT: v_mov_b32_e32 v4, v7
+; GFX7-NEXT: v_mov_b32_e32 v5, v8
+; GFX7-NEXT: buffer_atomic_cmpswap v[4:5], v[0:1], s[4:7], 0 addr64 glc
; GFX7-NEXT: s_waitcnt vmcnt(0)
; GFX7-NEXT: buffer_wbinvl1
-; GFX7-NEXT: v_cmp_eq_u32_e32 vcc, v7, v4
+; GFX7-NEXT: v_cmp_eq_u32_e32 vcc, v4, v8
; GFX7-NEXT: s_or_b64 s[8:9], vcc, s[8:9]
-; GFX7-NEXT: v_mov_b32_e32 v4, v7
; GFX7-NEXT: s_andn2_b64 exec, exec, s[8:9]
; GFX7-NEXT: s_cbranch_execnz .LBB25_1
; GFX7-NEXT: ; %bb.2: ; %atomicrmw.end
@@ -6875,29 +6876,29 @@ define void @global_agent_atomic_fsub_noret_f16(ptr addrspace(1) %ptr, half %val
; GFX6-NEXT: v_cvt_f16_f32_e32 v5, v2
; GFX6-NEXT: v_and_b32_e32 v2, 3, v3
; GFX6-NEXT: v_lshlrev_b32_e32 v2, 3, v2
-; GFX6-NEXT: v_lshl_b32_e32 v3, 0xffff, v2
-; GFX6-NEXT: v_cvt_f32_f16_e32 v5, v5
-; GFX6-NEXT: v_not_b32_e32 v6, v3
; GFX6-NEXT: s_mov_b64 s[8:9], 0
+; GFX6-NEXT: v_cvt_f32_f16_e32 v3, v5
+; GFX6-NEXT: v_lshl_b32_e32 v5, 0xffff, v2
+; GFX6-NEXT: v_not_b32_e32 v6, v5
; GFX6-NEXT: .LBB25_1: ; %atomicrmw.start
; GFX6-NEXT: ; =>This Inner Loop Header: Depth=1
; GFX6-NEXT: s_waitcnt vmcnt(0)
-; GFX6-NEXT: v_lshrrev_b32_e32 v3, v2, v4
-; GFX6-NEXT: v_cvt_f32_f16_e32 v3, v3
-; GFX6-NEXT: s_waitcnt expcnt(0)
-; GFX6-NEXT: v_and_b32_e32 v7, v4, v6
-; GFX6-NEXT: v_sub_f32_e32 v3, v3, v5
-; GFX6-NEXT: v_cvt_f16_f32_e32 v3, v3
-; GFX6-NEXT: v_lshlrev_b32_e32 v3, v2, v3
-; GFX6-NEXT: v_or_b32_e32 v3, v7, v3
; GFX6-NEXT: v_mov_b32_e32 v8, v4
-; GFX6-NEXT: v_mov_b32_e32 v7, v3
-; GFX6-NEXT: buffer_atomic_cmpswap v[7:8], v[0:1], s[4:7], 0 addr64 glc
+; GFX6-NEXT: s_waitcnt expcnt(0)
+; GFX6-NEXT: v_lshrrev_b32_e32 v4, v2, v8
+; GFX6-NEXT: v_cvt_f32_f16_e32 v4, v4
+; GFX6-NEXT: v_and_b32_e32 v5, v8, v6
+; GFX6-NEXT: v_sub_f32_e32 v4, v4, v3
+; GFX6-NEXT: v_cvt_f16_f32_e32 v4, v4
+; GFX6-NEXT: v_lshlrev_b32_e32 v4, v2, v4
+; GFX6-NEXT: v_or_b32_e32 v7, v5, v4
+; GFX6-NEXT: v_mov_b32_e32 v4, v7
+; GFX6-NEXT: v_mov_b32_e32 v5, v8
+; GFX6-NEXT: buffer_atomic_cmpswap v[4:5], v[0:1], s[4:7], 0 addr64 glc
; GFX6-NEXT: s_waitcnt vmcnt(0)
; GFX6-NEXT: buffer_wbinvl1
-; GFX6-NEXT: v_cmp_eq_u32_e32 vcc, v7, v4
+; GFX6-NEXT: v_cmp_eq_u32_e32 vcc, v4, v8
; GFX6-NEXT: s_or_b64 s[8:9], vcc, s[8:9]
-; GFX6-NEXT: v_mov_b32_e32 v4, v7
; GFX6-NEXT: s_andn2_b64 exec, exec, s[8:9]
; GFX6-NEXT: s_cbranch_execnz .LBB25_1
; GFX6-NEXT: ; %bb.2: ; %atomicrmw.end
@@ -7250,31 +7251,31 @@ define void @global_agent_atomic_fsub_noret_f16__offset12b_pos(ptr addrspace(1)
; GFX7-NEXT: s_mov_b32 s4, s6
; GFX7-NEXT: s_mov_b32 s5, s6
; GFX7-NEXT: buffer_load_dword v3, v[0:1], s[4:7], 0 addr64
-; GFX7-NEXT: v_cvt_f16_f32_e32 v2, v2
-; GFX7-NEXT: v_and_b32_e32 v4, 3, v4
-; GFX7-NEXT: v_lshlrev_b32_e32 v4, 3, v4
+; GFX7-NEXT: v_cvt_f16_f32_e32 v5, v2
+; GFX7-NEXT: v_and_b32_e32 v2, 3, v4
+; GFX7-NEXT: v_lshlrev_b32_e32 v2, 3, v2
+; GFX7-NEXT: v_lshl_b32_e32 v4, 0xffff, v2
+; GFX7-NEXT: v_cvt_f32_f16_e32 v5, v5
+; GFX7-NEXT: v_not_b32_e32 v6, v4
; GFX7-NEXT: s_mov_b64 s[8:9], 0
-; GFX7-NEXT: v_cvt_f32_f16_e32 v5, v2
-; GFX7-NEXT: v_lshl_b32_e32 v2, 0xffff, v4
-; GFX7-NEXT: v_not_b32_e32 v6, v2
; GFX7-NEXT: .LBB26_1: ; %atomicrmw.start
; GFX7-NEXT: ; =>This Inner Loop Header: Depth=1
; GFX7-NEXT: s_waitcnt vmcnt(0)
-; GFX7-NEXT: v_lshrrev_b32_e32 v2, v4, v3
-; GFX7-NEXT: v_cvt_f32_f16_e32 v2, v2
-; GFX7-NEXT: v_and_b32_e32 v7, v3, v6
-; GFX7-NEXT: v_sub_f32_e32 v2, v2, v5
-; GFX7-NEXT: v_cvt_f16_f32_e32 v2, v2
-; GFX7-NEXT: v_lshlrev_b32_e32 v2, v4, v2
-; GFX7-NEXT: v_or_b32_e32 v2, v7, v2
; GFX7-NEXT: v_mov_b32_e32 v8, v3
-; GFX7-NEXT: v_mov_b32_e32 v7, v2
-; GFX7-NEXT: buffer_atomic_cmpswap v[7:8], v[0:1], s[4:7], 0 addr64 glc
+; GFX7-NEXT: v_lshrrev_b32_e32 v3, v2, v8
+; GFX7-NEXT: v_cvt_f32_f16_e32 v3, v3
+; GFX7-NEXT: v_and_b32_e32 v4, v8, v6
+; GFX7-NEXT: v_sub_f32_e32 v3, v3, v5
+; GFX7-NEXT: v_cvt_f16_f32_e32 v3, v3
+; GFX7-NEXT: v_lshlrev_b32_e32 v3, v2, v3
+; GFX7-NEXT: v_or_b32_e32 v7, v4, v3
+; GFX7-NEXT: v_mov_b32_e32 v3, v7
+; GFX7-NEXT: v_mov_b32_e32 v4, v8
+; GFX7-NEXT: buffer_atomic_cmpswap v[3:4], v[0:1], s[4:7], 0 addr64 glc
; GFX7-NEXT: s_waitcnt vmcnt(0)
; GFX7-NEXT: buffer_wbinvl1
-; GFX7-NEXT: v_cmp_eq_u32_e32 vcc, v7, v3
+; GFX7-NEXT: v_cmp_eq_u32_e32 vcc, v3, v8
; GFX7-NEXT: s_or_b64 s[8:9], vcc, s[8:9]
-; GFX7-NEXT: v_mov_b32_e32 v3, v7
; GFX7-NEXT: s_andn2_b64 exec, exec, s[8:9]
; GFX7-NEXT: s_cbranch_execnz .LBB26_1
; GFX7-NEXT: ; %bb.2: ; %atomicrmw.end
@@ -7292,32 +7293,32 @@ define void @global_agent_atomic_fsub_noret_f16__offset12b_pos(ptr addrspace(1)
; GFX6-NEXT: s_mov_b32 s4, s6
; GFX6-NEXT: s_mov_b32 s5, s6
; GFX6-NEXT: buffer_load_dword v3, v[0:1], s[4:7], 0 addr64
-; GFX6-NEXT: v_cvt_f16_f32_e32 v2, v2
-; GFX6-NEXT: v_and_b32_e32 v4, 3, v4
-; GFX6-NEXT: v_lshlrev_b32_e32 v4, 3, v4
+; GFX6-NEXT: v_cvt_f16_f32_e32 v5, v2
+; GFX6-NEXT: v_and_b32_e32 v2, 3, v4
+; GFX6-NEXT: v_lshlrev_b32_e32 v2, 3, v2
+; GFX6-NEXT: v_lshl_b32_e32 v4, 0xffff, v2
+; GFX6-NEXT: v_cvt_f32_f16_e32 v5, v5
+; GFX6-NEXT: v_not_b32_e32 v6, v4
; GFX6-NEXT: s_mov_b64 s[8:9], 0
-; GFX6-NEXT: v_cvt_f32_f16_e32 v5, v2
-; GFX6-NEXT: v_lshl_b32_e32 v2, 0xffff, v4
-; GFX6-NEXT: v_not_b32_e32 v6, v2
; GFX6-NEXT: .LBB26_1: ; %atomicrmw.start
; GFX6-NEXT: ; =>This Inner Loop Header: Depth=1
; GFX6-NEXT: s_waitcnt vmcnt(0)
-; GFX6-NEXT: v_lshrrev_b32_e32 v2, v4, v3
-; GFX6-NEXT: v_cvt_f32_f16_e32 v2, v2
-; GFX6-NEXT: s_waitcnt expcnt(0)
-; GFX6-NEXT: v_and_b32_e32 v7, v3, v6
-; GFX6-NEXT: v_sub_f32_e32 v2, v2, v5
-; GFX6-NEXT: v_cvt_f16_f32_e32 v2, v2
-; GFX6-NEXT: v_lshlrev_b32_e32 v2, v4, v2
-; GFX6-NEXT: v_or_b32_e32 v2, v7, v2
; GFX6-NEXT: v_mov_b32_e32 v8, v3
-; GFX6-NEXT: v_mov_b32_e32 v7, v2
-; GFX6-NEXT: buffer_atomic_cmpswap v[7:8], v[0:1], s[4:7], 0 addr64 glc
+; GFX6-NEXT: s_waitcnt expcnt(0)
+; GFX6-NEXT: v_lshrrev_b32_e32 v3, v2, v8
+; GFX6-NEXT: v_cvt_f32_f16_e32 v3, v3
+; GFX6-NEXT: v_and_b32_e32 v4, v8, v6
+; GFX6-NEXT: v_sub_f32_e32 v3, v3, v5
+; GFX6-NEXT: v_cvt_f16_f32_e32 v3, v3
+; GFX6-NEXT: v_lshlrev_b32_e32 v3, v2, v3
+; GFX6-NEXT: v_or_b32_e32 v7, v4, v3
+; GFX6-NEXT: v_mov_b32_e32 v3, v7
+; GFX6-NEXT: v_mov_b32_e32 v4, v8
+; GFX6-NEXT: buffer_atomic_cmpswap v[3:4], v[0:1], s[4:7], 0 addr64 glc
; GFX6-NEXT: s_waitcnt vmcnt(0)
; GFX6-NEXT: buffer_wbinvl1
-; GFX6-NEXT: v_cmp_eq_u32_e32 vcc, v7, v3
+; GFX6-NEXT: v_cmp_eq_u32_e32 vcc, v3, v8
; GFX6-NEXT: s_or_b64 s[8:9], vcc, s[8:9]
-; GFX6-NEXT: v_mov_b32_e32 v3, v7
; GFX6-NEXT: s_andn2_b64 exec, exec, s[8:9]
; GFX6-NEXT: s_cbranch_execnz .LBB26_1
; GFX6-NEXT: ; %bb.2: ; %atomicrmw.end
@@ -7672,31 +7673,31 @@ define void @global_agent_atomic_fsub_noret_f16__offset12b_neg(ptr addrspace(1)
; GFX7-NEXT: s_mov_b32 s4, s6
; GFX7-NEXT: s_mov_b32 s5, s6
; GFX7-NEXT: buffer_load_dword v3, v[0:1], s[4:7], 0 addr64
-; GFX7-NEXT: v_cvt_f16_f32_e32 v2, v2
-; GFX7-NEXT: v_and_b32_e32 v4, 3, v4
-; GFX7-NEXT: v_lshlrev_b32_e32 v4, 3, v4
+; GFX7-NEXT: v_cvt_f16_f32_e32 v5, v2
+; GFX7-NEXT: v_and_b32_e32 v2, 3, v4
+; GFX7-NEXT: v_lshlrev_b32_e32 v2, 3, v2
+; GFX7-NEXT: v_lshl_b32_e32 v4, 0xffff, v2
+; GFX7-NEXT: v_cvt_f32_f16_e32 v5, v5
+; GFX7-NEXT: v_not_b32_e32 v6, v4
; GFX7-NEXT: s_mov_b64 s[8:9], 0
-; GFX7-NEXT: v_cvt_f32_f16_e32 v5, v2
-; GFX7-NEXT: v_lshl_b32_e32 v2, 0xffff, v4
-; GFX7-NEXT: v_not_b32_e32 v6, v2
; GFX7-NEXT: .LBB27_1: ; %atomicrmw.start
; GFX7-NEXT: ; =>This Inner Loop Header: Depth=1
; GFX7-NEXT: s_waitcnt vmcnt(0)
-; GFX7-NEXT: v_lshrrev_b32_e32 v2, v4, v3
-; GFX7-NEXT: v_cvt_f32_f16_e32 v2, v2
-; GFX7-NEXT: v_and_b32_e32 v7, v3, v6
-; GFX7-NEXT: v_sub_f32_e32 v2, v2, v5
-; GFX7-NEXT: v_cvt_f16_f32_e32 v2, v2
-; GFX7-NEXT: v_lshlrev_b32_e32 v2, v4, v2
-; GFX7-NEXT: v_or_b32_e32 v2, v7, v2
; GFX7-NEXT: v_mov_b32_e32 v8, v3
-; GFX7-NEXT: v_mov_b32_e32 v7, v2
-; GFX7-NEXT: buffer_atomic_cmpswap v[7:8], v[0:1], s[4:7], 0 addr64 glc
+; GFX7-NEXT: v_lshrrev_b32_e32 v3, v2, v8
+; GFX7-NEXT: v_cvt_f32_f16_e32 v3, v3
+; GFX7-NEXT: v_and_b32_e32 v4, v8, v6
+; GFX7-NEXT: v_sub_f32_e32 v3, v3, v5
+; GFX7-NEXT: v_cvt_f16_f32_e32 v3, v3
+; GFX7-NEXT: v_lshlrev_b32_e32 v3, v2, v3
+; GFX7-NEXT: v_or_b32_e32 v7, v4, v3
+; GFX7-NEXT: v_mov_b32_e32 v3, v7
+; GFX7-NEXT: v_mov_b32_e32 v4, v8
+; GFX7-NEXT: buffer_atomic_cmpswap v[3:4], v[0:1], s[4:7], 0 addr64 glc
; GFX7-NEXT: s_waitcnt vmcnt(0)
; GFX7-NEXT: buffer_wbinvl1
-; GFX7-NEXT: v_cmp_eq_u32_e32 vcc, v7, v3
+; GFX7-NEXT: v_cmp_eq_u32_e32 vcc, v3, v8
; GFX7-NEXT: s_or_b64 s[8:9], vcc, s[8:9]
-; GFX7-NEXT: v_mov_b32_e32 v3, v7
; GFX7-NEXT: s_andn2_b64 exec, exec, s[8:9]
; GFX7-NEXT: s_cbranch_execnz .LBB27_1
; GFX7-NEXT: ; %bb.2: ; %atomicrmw.end
@@ -7714,32 +7715,32 @@ define void @global_agent_atomic_fsub_noret_f16__offset12b_neg(ptr addrspace(1)
; GFX6-NEXT: s_mov_b32 s4, s6
; GFX6-NEXT: s_mov_b32 s5, s6
; GFX6-NEXT: buffer_load_dword v3, v[0:1], s[4:7], 0 addr64
-; GFX6-NEXT: v_cvt_f16_f32_e32 v2, v2
-; GFX6-NEXT: v_and_b32_e32 v4, 3, v4
-; GFX6-NEXT: v_lshlrev_b32_e32 v4, 3, v4
+; GFX6-NEXT: v_cvt_f16_f32_e32 v5, v2
+; GFX6-NEXT: v_and_b32_e32 v2, 3, v4
+; GFX6-NEXT: v_lshlrev_b32_e32 v2, 3, v2
+; GFX6-NEXT: v_lshl_b32_e32 v4, 0xffff, v2
+; GFX6-NEXT: v_cvt_f32_f16_e32 v5, v5
+; GFX6-NEXT: v_not_b32_e32 v6, v4
; GFX6-NEXT: s_mov_b64 s[8:9], 0
-; GFX6-NEXT: v_cvt_f32_f16_e32 v5, v2
-; GFX6-NEXT: v_lshl_b32_e32 v2, 0xffff, v4
-; GFX6-NEXT: v_not_b32_e32 v6, v2
; GFX6-NEXT: .LBB27_1: ; %atomicrmw.start
; GFX6-NEXT: ; =>This Inner Loop Header: Depth=1
; GFX6-NEXT: s_waitcnt vmcnt(0)
-; GFX6-NEXT: v_lshrrev_b32_e32 v2, v4, v3
-; GFX6-NEXT: v_cvt_f32_f16_e32 v2, v2
-; GFX6-NEXT: s_waitcnt expcnt(0)
-; GFX6-NEXT: v_and_b32_e32 v7, v3, v6
-; GFX6-NEXT: v_sub_f32_e32 v2, v2, v5
-; GFX6-NEXT: v_cvt_f16_f32_e32 v2, v2
-; GFX6-NEXT: v_lshlrev_b32_e32 v2, v4, v2
-; GFX6-NEXT: v_or_b32_e32 v2, v7, v2
; GFX6-NEXT: v_mov_b32_e32 v8, v3
-; GFX6-NEXT: v_mov_b32_e32 v7, v2
-; GFX6-NEXT: buffer_atomic_cmpswap v[7:8], v[0:1], s[4:7], 0 addr64 glc
+; GFX6-NEXT: s_waitcnt expcnt(0)
+; GFX6-NEXT: v_lshrrev_b32_e32 v3, v2, v8
+; GFX6-NEXT: v_cvt_f32_f16_e32 v3, v3
+; GFX6-NEXT: v_and_b32_e32 v4, v8, v6
+; GFX6-NEXT: v_sub_f32_e32 v3, v3, v5
+; GFX6-NEXT: v_cvt_f16_f32_e32 v3, v3
+; GFX6-NEXT: v_lshlrev_b32_e32 v3, v2, v3
+; GFX6-NEXT: v_or_b32_e32 v7, v4, v3
+; GFX6-NEXT: v_mov_b32_e32 v3, v7
+; GFX6-NEXT: v_mov_b32_e32 v4, v8
+; GFX6-NEXT: buffer_atomic_cmpswap v[3:4], v[0:1], s[4:7], 0 addr64 glc
; GFX6-NEXT: s_waitcnt vmcnt(0)
; GFX6-NEXT: buffer_wbinvl1
-; GFX6-NEXT: v_cmp_eq_u32_e32 vcc, v7, v3
+; GFX6-NEXT: v_cmp_eq_u32_e32 vcc, v3, v8
; GFX6-NEXT: s_or_b64 s[8:9], vcc, s[8:9]
-; GFX6-NEXT: v_mov_b32_e32 v3, v7
; GFX6-NEXT: s_andn2_b64 exec, exec, s[8:9]
; GFX6-NEXT: s_cbranch_execnz .LBB27_1
; GFX6-NEXT: ; %bb.2: ; %atomicrmw.end
@@ -8012,28 +8013,28 @@ define half @global_agent_atomic_fsub_ret_f16__offset12b_pos__align4(ptr addrspa
; GFX7-NEXT: buffer_load_dword v3, v[0:1], s[4:7], 0 addr64 offset:2046
; GFX7-NEXT: v_cvt_f16_f32_e32 v2, v2
; GFX7-NEXT: s_mov_b64 s[8:9], 0
-; GFX7-NEXT: v_cvt_f32_f16_e32 v6, v2
+; GFX7-NEXT: v_cvt_f32_f16_e32 v2, v2
; GFX7-NEXT: .LBB28_1: ; %atomicrmw.start
; GFX7-NEXT: ; =>This Inner Loop Header: Depth=1
; GFX7-NEXT: s_waitcnt vmcnt(0)
-; GFX7-NEXT: v_cvt_f32_f16_e32 v2, v3
-; GFX7-NEXT: v_and_b32_e32 v4, 0xffff0000, v3
-; GFX7-NEXT: v_sub_f32_e32 v2, v2, v6
-; GFX7-NEXT: v_cvt_f16_f32_e32 v2, v2
-; GFX7-NEXT: v_or_b32_e32 v2, v4, v2
; GFX7-NEXT: v_mov_b32_e32 v5, v3
-; GFX7-NEXT: v_mov_b32_e32 v4, v2
-; GFX7-NEXT: buffer_atomic_cmpswap v[4:5], v[0:1], s[4:7], 0 addr64 offset:2046 glc
+; GFX7-NEXT: v_cvt_f32_f16_e32 v3, v5
+; GFX7-NEXT: v_and_b32_e32 v4, 0xffff0000, v5
+; GFX7-NEXT: v_sub_f32_e32 v3, v3, v2
+; GFX7-NEXT: v_cvt_f16_f32_e32 v3, v3
+; GFX7-NEXT: v_or_b32_e32 v4, v4, v3
+; GFX7-NEXT: v_mov_b32_e32 v3, v4
+; GFX7-NEXT: v_mov_b32_e32 v4, v5
+; GFX7-NEXT: buffer_atomic_cmpswap v[3:4], v[0:1], s[4:7], 0 addr64 offset:2046 glc
; GFX7-NEXT: s_waitcnt vmcnt(0)
; GFX7-NEXT: buffer_wbinvl1
-; GFX7-NEXT: v_cmp_eq_u32_e32 vcc, v4, v3
+; GFX7-NEXT: v_cmp_eq_u32_e32 vcc, v3, v5
; GFX7-NEXT: s_or_b64 s[8:9], vcc, s[8:9]
-; GFX7-NEXT: v_mov_b32_e32 v3, v4
; GFX7-NEXT: s_andn2_b64 exec, exec, s[8:9]
; GFX7-NEXT: s_cbranch_execnz .LBB28_1
; GFX7-NEXT: ; %bb.2: ; %atomicrmw.end
; GFX7-NEXT: s_or_b64 exec, exec, s[8:9]
-; GFX7-NEXT: v_cvt_f32_f16_e32 v0, v4
+; GFX7-NEXT: v_cvt_f32_f16_e32 v0, v3
; GFX7-NEXT: s_setpc_b64 s[30:31]
;
; GFX6-LABEL: global_agent_atomic_fsub_ret_f16__offset12b_pos__align4:
@@ -8046,29 +8047,29 @@ define half @global_agent_atomic_fsub_ret_f16__offset12b_pos__align4(ptr addrspa
; GFX6-NEXT: buffer_load_dword v3, v[0:1], s[4:7], 0 addr64 offset:2046
; GFX6-NEXT: v_cvt_f16_f32_e32 v2, v2
; GFX6-NEXT: s_mov_b64 s[8:9], 0
-; GFX6-NEXT: v_cvt_f32_f16_e32 v6, v2
+; GFX6-NEXT: v_cvt_f32_f16_e32 v2, v2
; GFX6-NEXT: .LBB28_1: ; %atomicrmw.start
; GFX6-NEXT: ; =>This Inner Loop Header: Depth=1
; GFX6-NEXT: s_waitcnt vmcnt(0)
-; GFX6-NEXT: v_cvt_f32_f16_e32 v2, v3
-; GFX6-NEXT: s_waitcnt expcnt(0)
-; GFX6-NEXT: v_and_b32_e32 v4, 0xffff0000, v3
-; GFX6-NEXT: v_sub_f32_e32 v2, v2, v6
-; GFX6-NEXT: v_cvt_f16_f32_e32 v2, v2
-; GFX6-NEXT: v_or_b32_e32 v2, v4, v2
; GFX6-NEXT: v_mov_b32_e32 v5, v3
-; GFX6-NEXT: v_mov_b32_e32 v4, v2
-; GFX6-NEXT: buffer_atomic_cmpswap v[4:5], v[0:1], s[4:7], 0 addr64 offset:2046 glc
+; GFX6-NEXT: s_waitcnt expcnt(0)
+; GFX6-NEXT: v_cvt_f32_f16_e32 v3, v5
+; GFX6-NEXT: v_and_b32_e32 v4, 0xffff0000, v5
+; GFX6-NEXT: v_sub_f32_e32 v3, v3, v2
+; GFX6-NEXT: v_cvt_f16_f32_e32 v3, v3
+; GFX6-NEXT: v_or_b32_e32 v4, v4, v3
+; GFX6-NEXT: v_mov_b32_e32 v3, v4
+; GFX6-NEXT: v_mov_b32_e32 v4, v5
+; GFX6-NEXT: buffer_atomic_cmpswap v[3:4], v[0:1], s[4:7], 0 addr64 offset:2046 glc
; GFX6-NEXT: s_waitcnt vmcnt(0)
; GFX6-NEXT: buffer_wbinvl1
-; GFX6-NEXT: v_cmp_eq_u32_e32 vcc, v4, v3
+; GFX6-NEXT: v_cmp_eq_u32_e32 vcc, v3, v5
; GFX6-NEXT: s_or_b64 s[8:9], vcc, s[8:9]
-; GFX6-NEXT: v_mov_b32_e32 v3, v4
; GFX6-NEXT: s_andn2_b64 exec, exec, s[8:9]
; GFX6-NEXT: s_cbranch_execnz .LBB28_1
; GFX6-NEXT: ; %bb.2: ; %atomicrmw.end
; GFX6-NEXT: s_or_b64 exec, exec, s[8:9]
-; GFX6-NEXT: v_cvt_f32_f16_e32 v0, v4
+; GFX6-NEXT: v_cvt_f32_f16_e32 v0, v3
; GFX6-NEXT: s_waitcnt expcnt(0)
; GFX6-NEXT: s_setpc_b64 s[30:31]
%gep = getelementptr half, ptr addrspace(1) %ptr, i64 1023
@@ -8325,23 +8326,23 @@ define void @global_agent_atomic_fsub_noret_f16__offset12b__align4_pos(ptr addrs
; GFX7-NEXT: buffer_load_dword v3, v[0:1], s[4:7], 0 addr64 offset:2046
; GFX7-NEXT: v_cvt_f16_f32_e32 v2, v2
; GFX7-NEXT: s_mov_b64 s[8:9], 0
-; GFX7-NEXT: v_cvt_f32_f16_e32 v4, v2
+; GFX7-NEXT: v_cvt_f32_f16_e32 v2, v2
; GFX7-NEXT: .LBB29_1: ; %atomicrmw.start
; GFX7-NEXT: ; =>This Inner Loop Header: Depth=1
; GFX7-NEXT: s_waitcnt vmcnt(0)
-; GFX7-NEXT: v_cvt_f32_f16_e32 v2, v3
-; GFX7-NEXT: v_and_b32_e32 v5, 0xffff0000, v3
-; GFX7-NEXT: v_sub_f32_e32 v2, v2, v4
-; GFX7-NEXT: v_cvt_f16_f32_e32 v2, v2
-; GFX7-NEXT: v_or_b32_e32 v2, v5, v2
-; GFX7-NEXT: v_mov_b32_e32 v6, v3
-; GFX7-NEXT: v_mov_b32_e32 v5, v2
-; GFX7-NEXT: buffer_atomic_cmpswap v[5:6], v[0:1], s[4:7], 0 addr64 offset:2046 glc
+; GFX7-NEXT: v_mov_b32_e32 v5, v3
+; GFX7-NEXT: v_cvt_f32_f16_e32 v3, v5
+; GFX7-NEXT: v_and_b32_e32 v4, 0xffff0000, v5
+; GFX7-NEXT: v_sub_f32_e32 v3, v3, v2
+; GFX7-NEXT: v_cvt_f16_f32_e32 v3, v3
+; GFX7-NEXT: v_or_b32_e32 v4, v4, v3
+; GFX7-NEXT: v_mov_b32_e32 v3, v4
+; GFX7-NEXT: v_mov_b32_e32 v4, v5
+; GFX7-NEXT: buffer_atomic_cmpswap v[3:4], v[0:1], s[4:7], 0 addr64 offset:2046 glc
; GFX7-NEXT: s_waitcnt vmcnt(0)
; GFX7-NEXT: buffer_wbinvl1
-; GFX7-NEXT: v_cmp_eq_u32_e32 vcc, v5, v3
+; GFX7-NEXT: v_cmp_eq_u32_e32 vcc, v3, v5
; GFX7-NEXT: s_or_b64 s[8:9], vcc, s[8:9]
-; GFX7-NEXT: v_mov_b32_e32 v3, v5
; GFX7-NEXT: s_andn2_b64 exec, exec, s[8:9]
; GFX7-NEXT: s_cbranch_execnz .LBB29_1
; GFX7-NEXT: ; %bb.2: ; %atomicrmw.end
@@ -8358,24 +8359,24 @@ define void @global_agent_atomic_fsub_noret_f16__offset12b__align4_pos(ptr addrs
; GFX6-NEXT: buffer_load_dword v3, v[0:1], s[4:7], 0 addr64 offset:2046
; GFX6-NEXT: v_cvt_f16_f32_e32 v2, v2
; GFX6-NEXT: s_mov_b64 s[8:9], 0
-; GFX6-NEXT: v_cvt_f32_f16_e32 v4, v2
+; GFX6-NEXT: v_cvt_f32_f16_e32 v2, v2
; GFX6-NEXT: .LBB29_1: ; %atomicrmw.start
; GFX6-NEXT: ; =>This Inner Loop Header: Depth=1
; GFX6-NEXT: s_waitcnt vmcnt(0)
-; GFX6-NEXT: v_cvt_f32_f16_e32 v2, v3
+; GFX6-NEXT: v_mov_b32_e32 v5, v3
; GFX6-NEXT: s_waitcnt expcnt(0)
-; GFX6-NEXT: v_and_b32_e32 v5, 0xffff0000, v3
-; GFX6-NEXT: v_sub_f32_e32 v2, v2, v4
-; GFX6-NEXT: v_cvt_f16_f32_e32 v2, v2
-; GFX6-NEXT: v_or_b32_e32 v2, v5, v2
-; GFX6-NEXT: v_mov_b32_e32 v6, v3
-; GFX6-NEXT: v_mov_b32_e32 v5, v2
-; GFX6-NEXT: buffer_atomic_cmpswap v[5:6], v[0:1], s[4:7], 0 addr64 offset:2046 glc
+; GFX6-NEXT: v_cvt_f32_f16_e32 v3, v5
+; GFX6-NEXT: v_and_b32_e32 v4, 0xffff0000, v5
+; GFX6-NEXT: v_sub_f32_e32 v3, v3, v2
+; GFX6-NEXT: v_cvt_f16_f32_e32 v3, v3
+; GFX6-NEXT: v_or_b32_e32 v4, v4, v3
+; GFX6-NEXT: v_mov_b32_e32 v3, v4
+; GFX6-NEXT: v_mov_b32_e32 v4, v5
+; GFX6-NEXT: buffer_atomic_cmpswap v[3:4], v[0:1], s[4:7], 0 addr64 offset:2046 glc
; GFX6-NEXT: s_waitcnt vmcnt(0)
; GFX6-NEXT: buffer_wbinvl1
-; GFX6-NEXT: v_cmp_eq_u32_e32 vcc, v5, v3
+; GFX6-NEXT: v_cmp_eq_u32_e32 vcc, v3, v5
; GFX6-NEXT: s_or_b64 s[8:9], vcc, s[8:9]
-; GFX6-NEXT: v_mov_b32_e32 v3, v5
; GFX6-NEXT: s_andn2_b64 exec, exec, s[8:9]
; GFX6-NEXT: s_cbranch_execnz .LBB29_1
; GFX6-NEXT: ; %bb.2: ; %atomicrmw.end
@@ -8746,36 +8747,36 @@ define half @global_system_atomic_fsub_ret_f16__offset12b_pos(ptr addrspace(1) %
; GFX7-NEXT: s_mov_b32 s4, s6
; GFX7-NEXT: s_mov_b32 s5, s6
; GFX7-NEXT: buffer_load_dword v3, v[0:1], s[4:7], 0 addr64
-; GFX7-NEXT: v_cvt_f16_f32_e32 v2, v2
-; GFX7-NEXT: v_and_b32_e32 v4, 3, v4
-; GFX7-NEXT: v_lshlrev_b32_e32 v6, 3, v4
+; GFX7-NEXT: v_cvt_f16_f32_e32 v5, v2
+; GFX7-NEXT: v_and_b32_e32 v2, 3, v4
+; GFX7-NEXT: v_lshlrev_b32_e32 v2, 3, v2
+; GFX7-NEXT: v_lshl_b32_e32 v4, 0xffff, v2
+; GFX7-NEXT: v_cvt_f32_f16_e32 v5, v5
+; GFX7-NEXT: v_not_b32_e32 v6, v4
; GFX7-NEXT: s_mov_b64 s[8:9], 0
-; GFX7-NEXT: v_cvt_f32_f16_e32 v7, v2
-; GFX7-NEXT: v_lshl_b32_e32 v2, 0xffff, v6
-; GFX7-NEXT: v_not_b32_e32 v8, v2
; GFX7-NEXT: .LBB30_1: ; %atomicrmw.start
; GFX7-NEXT: ; =>This Inner Loop Header: Depth=1
; GFX7-NEXT: s_waitcnt vmcnt(0)
-; GFX7-NEXT: v_lshrrev_b32_e32 v2, v6, v3
-; GFX7-NEXT: v_cvt_f32_f16_e32 v2, v2
-; GFX7-NEXT: v_and_b32_e32 v4, v3, v8
-; GFX7-NEXT: v_sub_f32_e32 v2, v2, v7
-; GFX7-NEXT: v_cvt_f16_f32_e32 v2, v2
-; GFX7-NEXT: v_lshlrev_b32_e32 v2, v6, v2
-; GFX7-NEXT: v_or_b32_e32 v2, v4, v2
-; GFX7-NEXT: v_mov_b32_e32 v5, v3
-; GFX7-NEXT: v_mov_b32_e32 v4, v2
-; GFX7-NEXT: buffer_atomic_cmpswap v[4:5], v[0:1], s[4:7], 0 addr64 glc
+; GFX7-NEXT: v_mov_b32_e32 v8, v3
+; GFX7-NEXT: v_lshrrev_b32_e32 v3, v2, v8
+; GFX7-NEXT: v_cvt_f32_f16_e32 v3, v3
+; GFX7-NEXT: v_and_b32_e32 v4, v8, v6
+; GFX7-NEXT: v_sub_f32_e32 v3, v3, v5
+; GFX7-NEXT: v_cvt_f16_f32_e32 v3, v3
+; GFX7-NEXT: v_lshlrev_b32_e32 v3, v2, v3
+; GFX7-NEXT: v_or_b32_e32 v7, v4, v3
+; GFX7-NEXT: v_mov_b32_e32 v3, v7
+; GFX7-NEXT: v_mov_b32_e32 v4, v8
+; GFX7-NEXT: buffer_atomic_cmpswap v[3:4], v[0:1], s[4:7], 0 addr64 glc
; GFX7-NEXT: s_waitcnt vmcnt(0)
; GFX7-NEXT: buffer_wbinvl1
-; GFX7-NEXT: v_cmp_eq_u32_e32 vcc, v4, v3
+; GFX7-NEXT: v_cmp_eq_u32_e32 vcc, v3, v8
; GFX7-NEXT: s_or_b64 s[8:9], vcc, s[8:9]
-; GFX7-NEXT: v_mov_b32_e32 v3, v4
; GFX7-NEXT: s_andn2_b64 exec, exec, s[8:9]
; GFX7-NEXT: s_cbranch_execnz .LBB30_1
; GFX7-NEXT: ; %bb.2: ; %atomicrmw.end
; GFX7-NEXT: s_or_b64 exec, exec, s[8:9]
-; GFX7-NEXT: v_lshrrev_b32_e32 v0, v6, v4
+; GFX7-NEXT: v_lshrrev_b32_e32 v0, v2, v3
; GFX7-NEXT: v_cvt_f32_f16_e32 v0, v0
; GFX7-NEXT: s_setpc_b64 s[30:31]
;
@@ -8790,37 +8791,37 @@ define half @global_system_atomic_fsub_ret_f16__offset12b_pos(ptr addrspace(1) %
; GFX6-NEXT: s_mov_b32 s4, s6
; GFX6-NEXT: s_mov_b32 s5, s6
; GFX6-NEXT: buffer_load_dword v3, v[0:1], s[4:7], 0 addr64
-; GFX6-NEXT: v_cvt_f16_f32_e32 v2, v2
-; GFX6-NEXT: v_and_b32_e32 v4, 3, v4
-; GFX6-NEXT: v_lshlrev_b32_e32 v6, 3, v4
+; GFX6-NEXT: v_cvt_f16_f32_e32 v5, v2
+; GFX6-NEXT: v_and_b32_e32 v2, 3, v4
+; GFX6-NEXT: v_lshlrev_b32_e32 v2, 3, v2
+; GFX6-NEXT: v_lshl_b32_e32 v4, 0xffff, v2
+; GFX6-NEXT: v_cvt_f32_f16_e32 v5, v5
+; GFX6-NEXT: v_not_b32_e32 v6, v4
; GFX6-NEXT: s_mov_b64 s[8:9], 0
-; GFX6-NEXT: v_cvt_f32_f16_e32 v7, v2
-; GFX6-NEXT: v_lshl_b32_e32 v2, 0xffff, v6
-; GFX6-NEXT: v_not_b32_e32 v8, v2
; GFX6-NEXT: .LBB30_1: ; %atomicrmw.start
; GFX6-NEXT: ; =>This Inner Loop Header: Depth=1
; GFX6-NEXT: s_waitcnt vmcnt(0)
-; GFX6-NEXT: v_lshrrev_b32_e32 v2, v6, v3
-; GFX6-NEXT: v_cvt_f32_f16_e32 v2, v2
+; GFX6-NEXT: v_mov_b32_e32 v8, v3
; GFX6-NEXT: s_waitcnt expcnt(0)
-; GFX6-NEXT: v_and_b32_e32 v4, v3, v8
-; GFX6-NEXT: v_sub_f32_e32 v2, v2, v7
-; GFX6-NEXT: v_cvt_f16_f32_e32 v2, v2
-; GFX6-NEXT: v_lshlrev_b32_e32 v2, v6, v2
-; GFX6-NEXT: v_or_b32_e32 v2, v4, v2
-; GFX6-NEXT: v_mov_b32_e32 v5, v3
-; GFX6-NEXT: v_mov_b32_e32 v4, v2
-; GFX6-NEXT: buffer_atomic_cmpswap v[4:5], v[0:1], s[4:7], 0 addr64 glc
+; GFX6-NEXT: v_lshrrev_b32_e32 v3, v2, v8
+; GFX6-NEXT: v_cvt_f32_f16_e32 v3, v3
+; GFX6-NEXT: v_and_b32_e32 v4, v8, v6
+; GFX6-NEXT: v_sub_f32_e32 v3, v3, v5
+; GFX6-NEXT: v_cvt_f16_f32_e32 v3, v3
+; GFX6-NEXT: v_lshlrev_b32_e32 v3, v2, v3
+; GFX6-NEXT: v_or_b32_e32 v7, v4, v3
+; GFX6-NEXT: v_mov_b32_e32 v3, v7
+; GFX6-NEXT: v_mov_b32_e32 v4, v8
+; GFX6-NEXT: buffer_atomic_cmpswap v[3:4], v[0:1], s[4:7], 0 addr64 glc
; GFX6-NEXT: s_waitcnt vmcnt(0)
; GFX6-NEXT: buffer_wbinvl1
-; GFX6-NEXT: v_cmp_eq_u32_e32 vcc, v4, v3
+; GFX6-NEXT: v_cmp_eq_u32_e32 vcc, v3, v8
; GFX6-NEXT: s_or_b64 s[8:9], vcc, s[8:9]
-; GFX6-NEXT: v_mov_b32_e32 v3, v4
; GFX6-NEXT: s_andn2_b64 exec, exec, s[8:9]
; GFX6-NEXT: s_cbranch_execnz .LBB30_1
; GFX6-NEXT: ; %bb.2: ; %atomicrmw.end
; GFX6-NEXT: s_or_b64 exec, exec, s[8:9]
-; GFX6-NEXT: v_lshrrev_b32_e32 v0, v6, v4
+; GFX6-NEXT: v_lshrrev_b32_e32 v0, v2, v3
; GFX6-NEXT: v_cvt_f32_f16_e32 v0, v0
; GFX6-NEXT: s_waitcnt expcnt(0)
; GFX6-NEXT: s_setpc_b64 s[30:31]
@@ -9175,31 +9176,31 @@ define void @global_system_atomic_fsub_noret_f16__offset12b_pos(ptr addrspace(1)
; GFX7-NEXT: s_mov_b32 s4, s6
; GFX7-NEXT: s_mov_b32 s5, s6
; GFX7-NEXT: buffer_load_dword v3, v[0:1], s[4:7], 0 addr64
-; GFX7-NEXT: v_cvt_f16_f32_e32 v2, v2
-; GFX7-NEXT: v_and_b32_e32 v4, 3, v4
-; GFX7-NEXT: v_lshlrev_b32_e32 v4, 3, v4
+; GFX7-NEXT: v_cvt_f16_f32_e32 v5, v2
+; GFX7-NEXT: v_and_b32_e32 v2, 3, v4
+; GFX7-NEXT: v_lshlrev_b32_e32 v2, 3, v2
+; GFX7-NEXT: v_lshl_b32_e32 v4, 0xffff, v2
+; GFX7-NEXT: v_cvt_f32_f16_e32 v5, v5
+; GFX7-NEXT: v_not_b32_e32 v6, v4
; GFX7-NEXT: s_mov_b64 s[8:9], 0
-; GFX7-NEXT: v_cvt_f32_f16_e32 v5, v2
-; GFX7-NEXT: v_lshl_b32_e32 v2, 0xffff, v4
-; GFX7-NEXT: v_not_b32_e32 v6, v2
; GFX7-NEXT: .LBB31_1: ; %atomicrmw.start
; GFX7-NEXT: ; =>This Inner Loop Header: Depth=1
; GFX7-NEXT: s_waitcnt vmcnt(0)
-; GFX7-NEXT: v_lshrrev_b32_e32 v2, v4, v3
-; GFX7-NEXT: v_cvt_f32_f16_e32 v2, v2
-; GFX7-NEXT: v_and_b32_e32 v7, v3, v6
-; GFX7-NEXT: v_sub_f32_e32 v2, v2, v5
-; GFX7-NEXT: v_cvt_f16_f32_e32 v2, v2
-; GFX7-NEXT: v_lshlrev_b32_e32 v2, v4, v2
-; GFX7-NEXT: v_or_b32_e32 v2, v7, v2
; GFX7-NEXT: v_mov_b32_e32 v8, v3
-; GFX7-NEXT: v_mov_b32_e32 v7, v2
-; GFX7-NEXT: buffer_atomic_cmpswap v[7:8], v[0:1], s[4:7], 0 addr64 glc
+; GFX7-NEXT: v_lshrrev_b32_e32 v3, v2, v8
+; GFX7-NEXT: v_cvt_f32_f16_e32 v3, v3
+; GFX7-NEXT: v_and_b32_e32 v4, v8, v6
+; GFX7-NEXT: v_sub_f32_e32 v3, v3, v5
+; GFX7-NEXT: v_cvt_f16_f32_e32 v3, v3
+; GFX7-NEXT: v_lshlrev_b32_e32 v3, v2, v3
+; GFX7-NEXT: v_or_b32_e32 v7, v4, v3
+; GFX7-NEXT: v_mov_b32_e32 v3, v7
+; GFX7-NEXT: v_mov_b32_e32 v4, v8
+; GFX7-NEXT: buffer_atomic_cmpswap v[3:4], v[0:1], s[4:7], 0 addr64 glc
; GFX7-NEXT: s_waitcnt vmcnt(0)
; GFX7-NEXT: buffer_wbinvl1
-; GFX7-NEXT: v_cmp_eq_u32_e32 vcc, v7, v3
+; GFX7-NEXT: v_cmp_eq_u32_e32 vcc, v3, v8
; GFX7-NEXT: s_or_b64 s[8:9], vcc, s[8:9]
-; GFX7-NEXT: v_mov_b32_e32 v3, v7
; GFX7-NEXT: s_andn2_b64 exec, exec, s[8:9]
; GFX7-NEXT: s_cbranch_execnz .LBB31_1
; GFX7-NEXT: ; %bb.2: ; %atomicrmw.end
@@ -9217,32 +9218,32 @@ define void @global_system_atomic_fsub_noret_f16__offset12b_pos(ptr addrspace(1)
; GFX6-NEXT: s_mov_b32 s4, s6
; GFX6-NEXT: s_mov_b32 s5, s6
; GFX6-NEXT: buffer_load_dword v3, v[0:1], s[4:7], 0 addr64
-; GFX6-NEXT: v_cvt_f16_f32_e32 v2, v2
-; GFX6-NEXT: v_and_b32_e32 v4, 3, v4
-; GFX6-NEXT: v_lshlrev_b32_e32 v4, 3, v4
+; GFX6-NEXT: v_cvt_f16_f32_e32 v5, v2
+; GFX6-NEXT: v_and_b32_e32 v2, 3, v4
+; GFX6-NEXT: v_lshlrev_b32_e32 v2, 3, v2
+; GFX6-NEXT: v_lshl_b32_e32 v4, 0xffff, v2
+; GFX6-NEXT: v_cvt_f32_f16_e32 v5, v5
+; GFX6-NEXT: v_not_b32_e32 v6, v4
; GFX6-NEXT: s_mov_b64 s[8:9], 0
-; GFX6-NEXT: v_cvt_f32_f16_e32 v5, v2
-; GFX6-NEXT: v_lshl_b32_e32 v2, 0xffff, v4
-; GFX6-NEXT: v_not_b32_e32 v6, v2
; GFX6-NEXT: .LBB31_1: ; %atomicrmw.start
; GFX6-NEXT: ; =>This Inner Loop Header: Depth=1
; GFX6-NEXT: s_waitcnt vmcnt(0)
-; GFX6-NEXT: v_lshrrev_b32_e32 v2, v4, v3
-; GFX6-NEXT: v_cvt_f32_f16_e32 v2, v2
-; GFX6-NEXT: s_waitcnt expcnt(0)
-; GFX6-NEXT: v_and_b32_e32 v7, v3, v6
-; GFX6-NEXT: v_sub_f32_e32 v2, v2, v5
-; GFX6-NEXT: v_cvt_f16_f32_e32 v2, v2
-; GFX6-NEXT: v_lshlrev_b32_e32 v2, v4, v2
-; GFX6-NEXT: v_or_b32_e32 v2, v7, v2
; GFX6-NEXT: v_mov_b32_e32 v8, v3
-; GFX6-NEXT: v_mov_b32_e32 v7, v2
-; GFX6-NEXT: buffer_atomic_cmpswap v[7:8], v[0:1], s[4:7], 0 addr64 glc
+; GFX6-NEXT: s_waitcnt expcnt(0)
+; GFX6-NEXT: v_lshrrev_b32_e32 v3, v2, v8
+; GFX6-NEXT: v_cvt_f32_f16_e32 v3, v3
+; GFX6-NEXT: v_and_b32_e32 v4, v8, v6
+; GFX6-NEXT: v_sub_f32_e32 v3, v3, v5
+; GFX6-NEXT: v_cvt_f16_f32_e32 v3, v3
+; GFX6-NEXT: v_lshlrev_b32_e32 v3, v2, v3
+; GFX6-NEXT: v_or_b32_e32 v7, v4, v3
+; GFX6-NEXT: v_mov_b32_e32 v3, v7
+; GFX6-NEXT: v_mov_b32_e32 v4, v8
+; GFX6-NEXT: buffer_atomic_cmpswap v[3:4], v[0:1], s[4:7], 0 addr64 glc
; GFX6-NEXT: s_waitcnt vmcnt(0)
; GFX6-NEXT: buffer_wbinvl1
-; GFX6-NEXT: v_cmp_eq_u32_e32 vcc, v7, v3
+; GFX6-NEXT: v_cmp_eq_u32_e32 vcc, v3, v8
; GFX6-NEXT: s_or_b64 s[8:9], vcc, s[8:9]
-; GFX6-NEXT: v_mov_b32_e32 v3, v7
; GFX6-NEXT: s_andn2_b64 exec, exec, s[8:9]
; GFX6-NEXT: s_cbranch_execnz .LBB31_1
; GFX6-NEXT: ; %bb.2: ; %atomicrmw.end
@@ -9677,35 +9678,35 @@ define bfloat @global_agent_atomic_fsub_ret_bf16(ptr addrspace(1) %ptr, bfloat %
; GFX7-NEXT: s_mov_b32 s5, s6
; GFX7-NEXT: buffer_load_dword v4, v[0:1], s[4:7], 0 addr64
; GFX7-NEXT: v_and_b32_e32 v3, 3, v3
-; GFX7-NEXT: v_lshlrev_b32_e32 v5, 3, v3
-; GFX7-NEXT: v_lshl_b32_e32 v3, 0xffff, v5
+; GFX7-NEXT: v_lshlrev_b32_e32 v3, 3, v3
+; GFX7-NEXT: v_lshl_b32_e32 v5, 0xffff, v3
; GFX7-NEXT: v_mul_f32_e32 v2, 1.0, v2
-; GFX7-NEXT: v_not_b32_e32 v6, v3
+; GFX7-NEXT: v_not_b32_e32 v6, v5
; GFX7-NEXT: s_mov_b64 s[8:9], 0
-; GFX7-NEXT: v_and_b32_e32 v7, 0xffff0000, v2
+; GFX7-NEXT: v_and_b32_e32 v2, 0xffff0000, v2
; GFX7-NEXT: .LBB32_1: ; %atomicrmw.start
; GFX7-NEXT: ; =>This Inner Loop Header: Depth=1
; GFX7-NEXT: s_waitcnt vmcnt(0)
-; GFX7-NEXT: v_lshrrev_b32_e32 v2, v5, v4
-; GFX7-NEXT: v_lshlrev_b32_e32 v2, 16, v2
-; GFX7-NEXT: v_sub_f32_e32 v2, v2, v7
-; GFX7-NEXT: v_lshrrev_b32_e32 v2, 16, v2
-; GFX7-NEXT: v_and_b32_e32 v3, v4, v6
-; GFX7-NEXT: v_lshlrev_b32_e32 v2, v5, v2
-; GFX7-NEXT: v_or_b32_e32 v3, v3, v2
-; GFX7-NEXT: v_mov_b32_e32 v2, v3
-; GFX7-NEXT: v_mov_b32_e32 v3, v4
-; GFX7-NEXT: buffer_atomic_cmpswap v[2:3], v[0:1], s[4:7], 0 addr64 glc
+; GFX7-NEXT: v_mov_b32_e32 v8, v4
+; GFX7-NEXT: v_lshrrev_b32_e32 v4, v3, v8
+; GFX7-NEXT: v_lshlrev_b32_e32 v4, 16, v4
+; GFX7-NEXT: v_sub_f32_e32 v4, v4, v2
+; GFX7-NEXT: v_lshrrev_b32_e32 v4, 16, v4
+; GFX7-NEXT: v_and_b32_e32 v5, v8, v6
+; GFX7-NEXT: v_lshlrev_b32_e32 v4, v3, v4
+; GFX7-NEXT: v_or_b32_e32 v7, v5, v4
+; GFX7-NEXT: v_mov_b32_e32 v4, v7
+; GFX7-NEXT: v_mov_b32_e32 v5, v8
+; GFX7-NEXT: buffer_atomic_cmpswap v[4:5], v[0:1], s[4:7], 0 addr64 glc
; GFX7-NEXT: s_waitcnt vmcnt(0)
; GFX7-NEXT: buffer_wbinvl1
-; GFX7-NEXT: v_cmp_eq_u32_e32 vcc, v2, v4
+; GFX7-NEXT: v_cmp_eq_u32_e32 vcc, v4, v8
; GFX7-NEXT: s_or_b64 s[8:9], vcc, s[8:9]
-; GFX7-NEXT: v_mov_b32_e32 v4, v2
; GFX7-NEXT: s_andn2_b64 exec, exec, s[8:9]
; GFX7-NEXT: s_cbranch_execnz .LBB32_1
; GFX7-NEXT: ; %bb.2: ; %atomicrmw.end
; GFX7-NEXT: s_or_b64 exec, exec, s[8:9]
-; GFX7-NEXT: v_lshrrev_b32_e32 v0, v5, v2
+; GFX7-NEXT: v_lshrrev_b32_e32 v0, v3, v4
; GFX7-NEXT: v_lshlrev_b32_e32 v0, 16, v0
; GFX7-NEXT: s_setpc_b64 s[30:31]
;
@@ -9720,35 +9721,36 @@ define bfloat @global_agent_atomic_fsub_ret_bf16(ptr addrspace(1) %ptr, bfloat %
; GFX6-NEXT: s_mov_b32 s5, s6
; GFX6-NEXT: buffer_load_dword v4, v[0:1], s[4:7], 0 addr64
; GFX6-NEXT: v_and_b32_e32 v3, 3, v3
-; GFX6-NEXT: v_lshlrev_b32_e32 v5, 3, v3
-; GFX6-NEXT: v_lshl_b32_e32 v3, 0xffff, v5
+; GFX6-NEXT: v_lshlrev_b32_e32 v3, 3, v3
+; GFX6-NEXT: v_lshl_b32_e32 v5, 0xffff, v3
; GFX6-NEXT: v_mul_f32_e32 v2, 1.0, v2
-; GFX6-NEXT: v_not_b32_e32 v6, v3
+; GFX6-NEXT: v_not_b32_e32 v6, v5
; GFX6-NEXT: s_mov_b64 s[8:9], 0
-; GFX6-NEXT: v_and_b32_e32 v7, 0xffff0000, v2
+; GFX6-NEXT: v_and_b32_e32 v2, 0xffff0000, v2
; GFX6-NEXT: .LBB32_1: ; %atomicrmw.start
; GFX6-NEXT: ; =>This Inner Loop Header: Depth=1
-; GFX6-NEXT: s_waitcnt vmcnt(0) expcnt(0)
-; GFX6-NEXT: v_lshrrev_b32_e32 v2, v5, v4
-; GFX6-NEXT: v_lshlrev_b32_e32 v2, 16, v2
-; GFX6-NEXT: v_sub_f32_e32 v2, v2, v7
-; GFX6-NEXT: v_lshrrev_b32_e32 v2, 16, v2
-; GFX6-NEXT: v_and_b32_e32 v3, v4, v6
-; GFX6-NEXT: v_lshlrev_b32_e32 v2, v5, v2
-; GFX6-NEXT: v_or_b32_e32 v3, v3, v2
-; GFX6-NEXT: v_mov_b32_e32 v2, v3
-; GFX6-NEXT: v_mov_b32_e32 v3, v4
-; GFX6-NEXT: buffer_atomic_cmpswap v[2:3], v[0:1], s[4:7], 0 addr64 glc
+; GFX6-NEXT: s_waitcnt vmcnt(0)
+; GFX6-NEXT: v_mov_b32_e32 v8, v4
+; GFX6-NEXT: s_waitcnt expcnt(0)
+; GFX6-NEXT: v_lshrrev_b32_e32 v4, v3, v8
+; GFX6-NEXT: v_lshlrev_b32_e32 v4, 16, v4
+; GFX6-NEXT: v_sub_f32_e32 v4, v4, v2
+; GFX6-NEXT: v_lshrrev_b32_e32 v4, 16, v4
+; GFX6-NEXT: v_and_b32_e32 v5, v8, v6
+; GFX6-NEXT: v_lshlrev_b32_e32 v4, v3, v4
+; GFX6-NEXT: v_or_b32_e32 v7, v5, v4
+; GFX6-NEXT: v_mov_b32_e32 v4, v7
+; GFX6-NEXT: v_mov_b32_e32 v5, v8
+; GFX6-NEXT: buffer_atomic_cmpswap v[4:5], v[0:1], s[4:7], 0 addr64 glc
; GFX6-NEXT: s_waitcnt vmcnt(0)
; GFX6-NEXT: buffer_wbinvl1
-; GFX6-NEXT: v_cmp_eq_u32_e32 vcc, v2, v4
+; GFX6-NEXT: v_cmp_eq_u32_e32 vcc, v4, v8
; GFX6-NEXT: s_or_b64 s[8:9], vcc, s[8:9]
-; GFX6-NEXT: v_mov_b32_e32 v4, v2
; GFX6-NEXT: s_andn2_b64 exec, exec, s[8:9]
; GFX6-NEXT: s_cbranch_execnz .LBB32_1
; GFX6-NEXT: ; %bb.2: ; %atomicrmw.end
; GFX6-NEXT: s_or_b64 exec, exec, s[8:9]
-; GFX6-NEXT: v_lshrrev_b32_e32 v0, v5, v2
+; GFX6-NEXT: v_lshrrev_b32_e32 v0, v3, v4
; GFX6-NEXT: v_lshlrev_b32_e32 v0, 16, v0
; GFX6-NEXT: s_waitcnt expcnt(0)
; GFX6-NEXT: s_setpc_b64 s[30:31]
@@ -10190,35 +10192,35 @@ define bfloat @global_agent_atomic_fsub_ret_bf16__offset12b_pos(ptr addrspace(1)
; GFX7-NEXT: s_mov_b32 s5, s6
; GFX7-NEXT: buffer_load_dword v3, v[0:1], s[4:7], 0 addr64
; GFX7-NEXT: v_and_b32_e32 v4, 3, v4
-; GFX7-NEXT: v_lshlrev_b32_e32 v6, 3, v4
-; GFX7-NEXT: v_lshl_b32_e32 v4, 0xffff, v6
+; GFX7-NEXT: v_lshlrev_b32_e32 v5, 3, v4
+; GFX7-NEXT: v_lshl_b32_e32 v4, 0xffff, v5
; GFX7-NEXT: v_mul_f32_e32 v2, 1.0, v2
-; GFX7-NEXT: v_not_b32_e32 v7, v4
+; GFX7-NEXT: v_not_b32_e32 v6, v4
; GFX7-NEXT: s_mov_b64 s[8:9], 0
-; GFX7-NEXT: v_and_b32_e32 v8, 0xffff0000, v2
+; GFX7-NEXT: v_and_b32_e32 v2, 0xffff0000, v2
; GFX7-NEXT: .LBB33_1: ; %atomicrmw.start
; GFX7-NEXT: ; =>This Inner Loop Header: Depth=1
; GFX7-NEXT: s_waitcnt vmcnt(0)
-; GFX7-NEXT: v_lshrrev_b32_e32 v2, v6, v3
-; GFX7-NEXT: v_lshlrev_b32_e32 v2, 16, v2
-; GFX7-NEXT: v_sub_f32_e32 v2, v2, v8
-; GFX7-NEXT: v_lshrrev_b32_e32 v2, 16, v2
-; GFX7-NEXT: v_and_b32_e32 v4, v3, v7
-; GFX7-NEXT: v_lshlrev_b32_e32 v2, v6, v2
-; GFX7-NEXT: v_or_b32_e32 v2, v4, v2
-; GFX7-NEXT: v_mov_b32_e32 v5, v3
-; GFX7-NEXT: v_mov_b32_e32 v4, v2
-; GFX7-NEXT: buffer_atomic_cmpswap v[4:5], v[0:1], s[4:7], 0 addr64 glc
+; GFX7-NEXT: v_mov_b32_e32 v8, v3
+; GFX7-NEXT: v_lshrrev_b32_e32 v3, v5, v8
+; GFX7-NEXT: v_lshlrev_b32_e32 v3, 16, v3
+; GFX7-NEXT: v_sub_f32_e32 v3, v3, v2
+; GFX7-NEXT: v_lshrrev_b32_e32 v3, 16, v3
+; GFX7-NEXT: v_and_b32_e32 v4, v8, v6
+; GFX7-NEXT: v_lshlrev_b32_e32 v3, v5, v3
+; GFX7-NEXT: v_or_b32_e32 v7, v4, v3
+; GFX7-NEXT: v_mov_b32_e32 v3, v7
+; GFX7-NEXT: v_mov_b32_e32 v4, v8
+; GFX7-NEXT: buffer_atomic_cmpswap v[3:4], v[0:1], s[4:7], 0 addr64 glc
; GFX7-NEXT: s_waitcnt vmcnt(0)
; GFX7-NEXT: buffer_wbinvl1
-; GFX7-NEXT: v_cmp_eq_u32_e32 vcc, v4, v3
+; GFX7-NEXT: v_cmp_eq_u32_e32 vcc, v3, v8
; GFX7-NEXT: s_or_b64 s[8:9], vcc, s[8:9]
-; GFX7-NEXT: v_mov_b32_e32 v3, v4
; GFX7-NEXT: s_andn2_b64 exec, exec, s[8:9]
; GFX7-NEXT: s_cbranch_execnz .LBB33_1
; GFX7-NEXT: ; %bb.2: ; %atomicrmw.end
; GFX7-NEXT: s_or_b64 exec, exec, s[8:9]
-; GFX7-NEXT: v_lshrrev_b32_e32 v0, v6, v4
+; GFX7-NEXT: v_lshrrev_b32_e32 v0, v5, v3
; GFX7-NEXT: v_lshlrev_b32_e32 v0, 16, v0
; GFX7-NEXT: s_setpc_b64 s[30:31]
;
@@ -10234,36 +10236,36 @@ define bfloat @global_agent_atomic_fsub_ret_bf16__offset12b_pos(ptr addrspace(1)
; GFX6-NEXT: s_mov_b32 s5, s6
; GFX6-NEXT: buffer_load_dword v3, v[0:1], s[4:7], 0 addr64
; GFX6-NEXT: v_and_b32_e32 v4, 3, v4
-; GFX6-NEXT: v_lshlrev_b32_e32 v6, 3, v4
-; GFX6-NEXT: v_lshl_b32_e32 v4, 0xffff, v6
+; GFX6-NEXT: v_lshlrev_b32_e32 v5, 3, v4
+; GFX6-NEXT: v_lshl_b32_e32 v4, 0xffff, v5
; GFX6-NEXT: v_mul_f32_e32 v2, 1.0, v2
-; GFX6-NEXT: v_not_b32_e32 v7, v4
+; GFX6-NEXT: v_not_b32_e32 v6, v4
; GFX6-NEXT: s_mov_b64 s[8:9], 0
-; GFX6-NEXT: v_and_b32_e32 v8, 0xffff0000, v2
+; GFX6-NEXT: v_and_b32_e32 v2, 0xffff0000, v2
; GFX6-NEXT: .LBB33_1: ; %atomicrmw.start
; GFX6-NEXT: ; =>This Inner Loop Header: Depth=1
; GFX6-NEXT: s_waitcnt vmcnt(0)
-; GFX6-NEXT: v_lshrrev_b32_e32 v2, v6, v3
-; GFX6-NEXT: v_lshlrev_b32_e32 v2, 16, v2
-; GFX6-NEXT: v_sub_f32_e32 v2, v2, v8
-; GFX6-NEXT: v_lshrrev_b32_e32 v2, 16, v2
+; GFX6-NEXT: v_mov_b32_e32 v8, v3
; GFX6-NEXT: s_waitcnt expcnt(0)
-; GFX6-NEXT: v_and_b32_e32 v4, v3, v7
-; GFX6-NEXT: v_lshlrev_b32_e32 v2, v6, v2
-; GFX6-NEXT: v_or_b32_e32 v2, v4, v2
-; GFX6-NEXT: v_mov_b32_e32 v5, v3
-; GFX6-NEXT: v_mov_b32_e32 v4, v2
-; GFX6-NEXT: buffer_atomic_cmpswap v[4:5], v[0:1], s[4:7], 0 addr64 glc
+; GFX6-NEXT: v_lshrrev_b32_e32 v3, v5, v8
+; GFX6-NEXT: v_lshlrev_b32_e32 v3, 16, v3
+; GFX6-NEXT: v_sub_f32_e32 v3, v3, v2
+; GFX6-NEXT: v_lshrrev_b32_e32 v3, 16, v3
+; GFX6-NEXT: v_and_b32_e32 v4, v8, v6
+; GFX6-NEXT: v_lshlrev_b32_e32 v3, v5, v3
+; GFX6-NEXT: v_or_b32_e32 v7, v4, v3
+; GFX6-NEXT: v_mov_b32_e32 v3, v7
+; GFX6-NEXT: v_mov_b32_e32 v4, v8
+; GFX6-NEXT: buffer_atomic_cmpswap v[3:4], v[0:1], s[4:7], 0 addr64 glc
; GFX6-NEXT: s_waitcnt vmcnt(0)
; GFX6-NEXT: buffer_wbinvl1
-; GFX6-NEXT: v_cmp_eq_u32_e32 vcc, v4, v3
+; GFX6-NEXT: v_cmp_eq_u32_e32 vcc, v3, v8
; GFX6-NEXT: s_or_b64 s[8:9], vcc, s[8:9]
-; GFX6-NEXT: v_mov_b32_e32 v3, v4
; GFX6-NEXT: s_andn2_b64 exec, exec, s[8:9]
; GFX6-NEXT: s_cbranch_execnz .LBB33_1
; GFX6-NEXT: ; %bb.2: ; %atomicrmw.end
; GFX6-NEXT: s_or_b64 exec, exec, s[8:9]
-; GFX6-NEXT: v_lshrrev_b32_e32 v0, v6, v4
+; GFX6-NEXT: v_lshrrev_b32_e32 v0, v5, v3
; GFX6-NEXT: v_lshlrev_b32_e32 v0, 16, v0
; GFX6-NEXT: s_waitcnt expcnt(0)
; GFX6-NEXT: s_setpc_b64 s[30:31]
@@ -10707,35 +10709,35 @@ define bfloat @global_agent_atomic_fsub_ret_bf16__offset12b_neg(ptr addrspace(1)
; GFX7-NEXT: s_mov_b32 s5, s6
; GFX7-NEXT: buffer_load_dword v3, v[0:1], s[4:7], 0 addr64
; GFX7-NEXT: v_and_b32_e32 v4, 3, v4
-; GFX7-NEXT: v_lshlrev_b32_e32 v6, 3, v4
-; GFX7-NEXT: v_lshl_b32_e32 v4, 0xffff, v6
+; GFX7-NEXT: v_lshlrev_b32_e32 v5, 3, v4
+; GFX7-NEXT: v_lshl_b32_e32 v4, 0xffff, v5
; GFX7-NEXT: v_mul_f32_e32 v2, 1.0, v2
-; GFX7-NEXT: v_not_b32_e32 v7, v4
+; GFX7-NEXT: v_not_b32_e32 v6, v4
; GFX7-NEXT: s_mov_b64 s[8:9], 0
-; GFX7-NEXT: v_and_b32_e32 v8, 0xffff0000, v2
+; GFX7-NEXT: v_and_b32_e32 v2, 0xffff0000, v2
; GFX7-NEXT: .LBB34_1: ; %atomicrmw.start
; GFX7-NEXT: ; =>This Inner Loop Header: Depth=1
; GFX7-NEXT: s_waitcnt vmcnt(0)
-; GFX7-NEXT: v_lshrrev_b32_e32 v2, v6, v3
-; GFX7-NEXT: v_lshlrev_b32_e32 v2, 16, v2
-; GFX7-NEXT: v_sub_f32_e32 v2, v2, v8
-; GFX7-NEXT: v_lshrrev_b32_e32 v2, 16, v2
-; GFX7-NEXT: v_and_b32_e32 v4, v3, v7
-; GFX7-NEXT: v_lshlrev_b32_e32 v2, v6, v2
-; GFX7-NEXT: v_or_b32_e32 v2, v4, v2
-; GFX7-NEXT: v_mov_b32_e32 v5, v3
-; GFX7-NEXT: v_mov_b32_e32 v4, v2
-; GFX7-NEXT: buffer_atomic_cmpswap v[4:5], v[0:1], s[4:7], 0 addr64 glc
+; GFX7-NEXT: v_mov_b32_e32 v8, v3
+; GFX7-NEXT: v_lshrrev_b32_e32 v3, v5, v8
+; GFX7-NEXT: v_lshlrev_b32_e32 v3, 16, v3
+; GFX7-NEXT: v_sub_f32_e32 v3, v3, v2
+; GFX7-NEXT: v_lshrrev_b32_e32 v3, 16, v3
+; GFX7-NEXT: v_and_b32_e32 v4, v8, v6
+; GFX7-NEXT: v_lshlrev_b32_e32 v3, v5, v3
+; GFX7-NEXT: v_or_b32_e32 v7, v4, v3
+; GFX7-NEXT: v_mov_b32_e32 v3, v7
+; GFX7-NEXT: v_mov_b32_e32 v4, v8
+; GFX7-NEXT: buffer_atomic_cmpswap v[3:4], v[0:1], s[4:7], 0 addr64 glc
; GFX7-NEXT: s_waitcnt vmcnt(0)
; GFX7-NEXT: buffer_wbinvl1
-; GFX7-NEXT: v_cmp_eq_u32_e32 vcc, v4, v3
+; GFX7-NEXT: v_cmp_eq_u32_e32 vcc, v3, v8
; GFX7-NEXT: s_or_b64 s[8:9], vcc, s[8:9]
-; GFX7-NEXT: v_mov_b32_e32 v3, v4
; GFX7-NEXT: s_andn2_b64 exec, exec, s[8:9]
; GFX7-NEXT: s_cbranch_execnz .LBB34_1
; GFX7-NEXT: ; %bb.2: ; %atomicrmw.end
; GFX7-NEXT: s_or_b64 exec, exec, s[8:9]
-; GFX7-NEXT: v_lshrrev_b32_e32 v0, v6, v4
+; GFX7-NEXT: v_lshrrev_b32_e32 v0, v5, v3
; GFX7-NEXT: v_lshlrev_b32_e32 v0, 16, v0
; GFX7-NEXT: s_setpc_b64 s[30:31]
;
@@ -10751,36 +10753,36 @@ define bfloat @global_agent_atomic_fsub_ret_bf16__offset12b_neg(ptr addrspace(1)
; GFX6-NEXT: s_mov_b32 s5, s6
; GFX6-NEXT: buffer_load_dword v3, v[0:1], s[4:7], 0 addr64
; GFX6-NEXT: v_and_b32_e32 v4, 3, v4
-; GFX6-NEXT: v_lshlrev_b32_e32 v6, 3, v4
-; GFX6-NEXT: v_lshl_b32_e32 v4, 0xffff, v6
+; GFX6-NEXT: v_lshlrev_b32_e32 v5, 3, v4
+; GFX6-NEXT: v_lshl_b32_e32 v4, 0xffff, v5
; GFX6-NEXT: v_mul_f32_e32 v2, 1.0, v2
-; GFX6-NEXT: v_not_b32_e32 v7, v4
+; GFX6-NEXT: v_not_b32_e32 v6, v4
; GFX6-NEXT: s_mov_b64 s[8:9], 0
-; GFX6-NEXT: v_and_b32_e32 v8, 0xffff0000, v2
+; GFX6-NEXT: v_and_b32_e32 v2, 0xffff0000, v2
; GFX6-NEXT: .LBB34_1: ; %atomicrmw.start
; GFX6-NEXT: ; =>This Inner Loop Header: Depth=1
; GFX6-NEXT: s_waitcnt vmcnt(0)
-; GFX6-NEXT: v_lshrrev_b32_e32 v2, v6, v3
-; GFX6-NEXT: v_lshlrev_b32_e32 v2, 16, v2
-; GFX6-NEXT: v_sub_f32_e32 v2, v2, v8
-; GFX6-NEXT: v_lshrrev_b32_e32 v2, 16, v2
+; GFX6-NEXT: v_mov_b32_e32 v8, v3
; GFX6-NEXT: s_waitcnt expcnt(0)
-; GFX6-NEXT: v_and_b32_e32 v4, v3, v7
-; GFX6-NEXT: v_lshlrev_b32_e32 v2, v6, v2
-; GFX6-NEXT: v_or_b32_e32 v2, v4, v2
-; GFX6-NEXT: v_mov_b32_e32 v5, v3
-; GFX6-NEXT: v_mov_b32_e32 v4, v2
-; GFX6-NEXT: buffer_atomic_cmpswap v[4:5], v[0:1], s[4:7], 0 addr64 glc
+; GFX6-NEXT: v_lshrrev_b32_e32 v3, v5, v8
+; GFX6-NEXT: v_lshlrev_b32_e32 v3, 16, v3
+; GFX6-NEXT: v_sub_f32_e32 v3, v3, v2
+; GFX6-NEXT: v_lshrrev_b32_e32 v3, 16, v3
+; GFX6-NEXT: v_and_b32_e32 v4, v8, v6
+; GFX6-NEXT: v_lshlrev_b32_e32 v3, v5, v3
+; GFX6-NEXT: v_or_b32_e32 v7, v4, v3
+; GFX6-NEXT: v_mov_b32_e32 v3, v7
+; GFX6-NEXT: v_mov_b32_e32 v4, v8
+; GFX6-NEXT: buffer_atomic_cmpswap v[3:4], v[0:1], s[4:7], 0 addr64 glc
; GFX6-NEXT: s_waitcnt vmcnt(0)
; GFX6-NEXT: buffer_wbinvl1
-; GFX6-NEXT: v_cmp_eq_u32_e32 vcc, v4, v3
+; GFX6-NEXT: v_cmp_eq_u32_e32 vcc, v3, v8
; GFX6-NEXT: s_or_b64 s[8:9], vcc, s[8:9]
-; GFX6-NEXT: v_mov_b32_e32 v3, v4
; GFX6-NEXT: s_andn2_b64 exec, exec, s[8:9]
; GFX6-NEXT: s_cbranch_execnz .LBB34_1
; GFX6-NEXT: ; %bb.2: ; %atomicrmw.end
; GFX6-NEXT: s_or_b64 exec, exec, s[8:9]
-; GFX6-NEXT: v_lshrrev_b32_e32 v0, v6, v4
+; GFX6-NEXT: v_lshrrev_b32_e32 v0, v5, v3
; GFX6-NEXT: v_lshlrev_b32_e32 v0, 16, v0
; GFX6-NEXT: s_waitcnt expcnt(0)
; GFX6-NEXT: s_setpc_b64 s[30:31]
@@ -11195,30 +11197,30 @@ define void @global_agent_atomic_fsub_noret_bf16(ptr addrspace(1) %ptr, bfloat %
; GFX7-NEXT: s_mov_b32 s5, s6
; GFX7-NEXT: buffer_load_dword v4, v[0:1], s[4:7], 0 addr64
; GFX7-NEXT: v_and_b32_e32 v3, 3, v3
-; GFX7-NEXT: v_lshlrev_b32_e32 v5, 3, v3
-; GFX7-NEXT: v_lshl_b32_e32 v3, 0xffff, v5
+; GFX7-NEXT: v_lshlrev_b32_e32 v3, 3, v3
+; GFX7-NEXT: v_lshl_b32_e32 v5, 0xffff, v3
; GFX7-NEXT: v_mul_f32_e32 v2, 1.0, v2
-; GFX7-NEXT: v_not_b32_e32 v6, v3
+; GFX7-NEXT: v_not_b32_e32 v6, v5
; GFX7-NEXT: s_mov_b64 s[8:9], 0
; GFX7-NEXT: v_and_b32_e32 v2, 0xffff0000, v2
; GFX7-NEXT: .LBB35_1: ; %atomicrmw.start
; GFX7-NEXT: ; =>This Inner Loop Header: Depth=1
; GFX7-NEXT: s_waitcnt vmcnt(0)
-; GFX7-NEXT: v_lshrrev_b32_e32 v3, v5, v4
-; GFX7-NEXT: v_lshlrev_b32_e32 v3, 16, v3
-; GFX7-NEXT: v_sub_f32_e32 v3, v3, v2
-; GFX7-NEXT: v_lshrrev_b32_e32 v3, 16, v3
-; GFX7-NEXT: v_and_b32_e32 v7, v4, v6
-; GFX7-NEXT: v_lshlrev_b32_e32 v3, v5, v3
-; GFX7-NEXT: v_or_b32_e32 v3, v7, v3
; GFX7-NEXT: v_mov_b32_e32 v8, v4
-; GFX7-NEXT: v_mov_b32_e32 v7, v3
-; GFX7-NEXT: buffer_atomic_cmpswap v[7:8], v[0:1], s[4:7], 0 addr64 glc
+; GFX7-NEXT: v_lshrrev_b32_e32 v4, v3, v8
+; GFX7-NEXT: v_lshlrev_b32_e32 v4, 16, v4
+; GFX7-NEXT: v_sub_f32_e32 v4, v4, v2
+; GFX7-NEXT: v_lshrrev_b32_e32 v4, 16, v4
+; GFX7-NEXT: v_and_b32_e32 v5, v8, v6
+; GFX7-NEXT: v_lshlrev_b32_e32 v4, v3, v4
+; GFX7-NEXT: v_or_b32_e32 v7, v5, v4
+; GFX7-NEXT: v_mov_b32_e32 v4, v7
+; GFX7-NEXT: v_mov_b32_e32 v5, v8
+; GFX7-NEXT: buffer_atomic_cmpswap v[4:5], v[0:1], s[4:7], 0 addr64 glc
; GFX7-NEXT: s_waitcnt vmcnt(0)
; GFX7-NEXT: buffer_wbinvl1
-; GFX7-NEXT: v_cmp_eq_u32_e32 vcc, v7, v4
+; GFX7-NEXT: v_cmp_eq_u32_e32 vcc, v4, v8
; GFX7-NEXT: s_or_b64 s[8:9], vcc, s[8:9]
-; GFX7-NEXT: v_mov_b32_e32 v4, v7
; GFX7-NEXT: s_andn2_b64 exec, exec, s[8:9]
; GFX7-NEXT: s_cbranch_execnz .LBB35_1
; GFX7-NEXT: ; %bb.2: ; %atomicrmw.end
@@ -11236,31 +11238,31 @@ define void @global_agent_atomic_fsub_noret_bf16(ptr addrspace(1) %ptr, bfloat %
; GFX6-NEXT: s_mov_b32 s5, s6
; GFX6-NEXT: buffer_load_dword v4, v[0:1], s[4:7], 0 addr64
; GFX6-NEXT: v_and_b32_e32 v3, 3, v3
-; GFX6-NEXT: v_lshlrev_b32_e32 v5, 3, v3
-; GFX6-NEXT: v_lshl_b32_e32 v3, 0xffff, v5
+; GFX6-NEXT: v_lshlrev_b32_e32 v3, 3, v3
+; GFX6-NEXT: v_lshl_b32_e32 v5, 0xffff, v3
; GFX6-NEXT: v_mul_f32_e32 v2, 1.0, v2
-; GFX6-NEXT: v_not_b32_e32 v6, v3
+; GFX6-NEXT: v_not_b32_e32 v6, v5
; GFX6-NEXT: s_mov_b64 s[8:9], 0
; GFX6-NEXT: v_and_b32_e32 v2, 0xffff0000, v2
; GFX6-NEXT: .LBB35_1: ; %atomicrmw.start
; GFX6-NEXT: ; =>This Inner Loop Header: Depth=1
; GFX6-NEXT: s_waitcnt vmcnt(0)
-; GFX6-NEXT: v_lshrrev_b32_e32 v3, v5, v4
-; GFX6-NEXT: v_lshlrev_b32_e32 v3, 16, v3
-; GFX6-NEXT: v_sub_f32_e32 v3, v3, v2
-; GFX6-NEXT: v_lshrrev_b32_e32 v3, 16, v3
-; GFX6-NEXT: s_waitcnt expcnt(0)
-; GFX6-NEXT: v_and_b32_e32 v7, v4, v6
-; GFX6-NEXT: v_lshlrev_b32_e32 v3, v5, v3
-; GFX6-NEXT: v_or_b32_e32 v3, v7, v3
; GFX6-NEXT: v_mov_b32_e32 v8, v4
-; GFX6-NEXT: v_mov_b32_e32 v7, v3
-; GFX6-NEXT: buffer_atomic_cmpswap v[7:8], v[0:1], s[4:7], 0 addr64 glc
+; GFX6-NEXT: s_waitcnt expcnt(0)
+; GFX6-NEXT: v_lshrrev_b32_e32 v4, v3, v8
+; GFX6-NEXT: v_lshlrev_b32_e32 v4, 16, v4
+; GFX6-NEXT: v_sub_f32_e32 v4, v4, v2
+; GFX6-NEXT: v_lshrrev_b32_e32 v4, 16, v4
+; GFX6-NEXT: v_and_b32_e32 v5, v8, v6
+; GFX6-NEXT: v_lshlrev_b32_e32 v4, v3, v4
+; GFX6-NEXT: v_or_b32_e32 v7, v5, v4
+; GFX6-NEXT: v_mov_b32_e32 v4, v7
+; GFX6-NEXT: v_mov_b32_e32 v5, v8
+; GFX6-NEXT: buffer_atomic_cmpswap v[4:5], v[0:1], s[4:7], 0 addr64 glc
; GFX6-NEXT: s_waitcnt vmcnt(0)
; GFX6-NEXT: buffer_wbinvl1
-; GFX6-NEXT: v_cmp_eq_u32_e32 vcc, v7, v4
+; GFX6-NEXT: v_cmp_eq_u32_e32 vcc, v4, v8
; GFX6-NEXT: s_or_b64 s[8:9], vcc, s[8:9]
-; GFX6-NEXT: v_mov_b32_e32 v4, v7
; GFX6-NEXT: s_andn2_b64 exec, exec, s[8:9]
; GFX6-NEXT: s_cbranch_execnz .LBB35_1
; GFX6-NEXT: ; %bb.2: ; %atomicrmw.end
@@ -11692,30 +11694,30 @@ define void @global_agent_atomic_fsub_noret_bf16__offset12b_pos(ptr addrspace(1)
; GFX7-NEXT: s_mov_b32 s5, s6
; GFX7-NEXT: buffer_load_dword v3, v[0:1], s[4:7], 0 addr64
; GFX7-NEXT: v_and_b32_e32 v4, 3, v4
-; GFX7-NEXT: v_lshlrev_b32_e32 v4, 3, v4
-; GFX7-NEXT: v_lshl_b32_e32 v5, 0xffff, v4
+; GFX7-NEXT: v_lshlrev_b32_e32 v5, 3, v4
+; GFX7-NEXT: v_lshl_b32_e32 v4, 0xffff, v5
; GFX7-NEXT: v_mul_f32_e32 v2, 1.0, v2
-; GFX7-NEXT: v_not_b32_e32 v5, v5
+; GFX7-NEXT: v_not_b32_e32 v6, v4
; GFX7-NEXT: s_mov_b64 s[8:9], 0
-; GFX7-NEXT: v_and_b32_e32 v6, 0xffff0000, v2
+; GFX7-NEXT: v_and_b32_e32 v2, 0xffff0000, v2
; GFX7-NEXT: .LBB36_1: ; %atomicrmw.start
; GFX7-NEXT: ; =>This Inner Loop Header: Depth=1
; GFX7-NEXT: s_waitcnt vmcnt(0)
-; GFX7-NEXT: v_lshrrev_b32_e32 v2, v4, v3
-; GFX7-NEXT: v_lshlrev_b32_e32 v2, 16, v2
-; GFX7-NEXT: v_sub_f32_e32 v2, v2, v6
-; GFX7-NEXT: v_lshrrev_b32_e32 v2, 16, v2
-; GFX7-NEXT: v_and_b32_e32 v7, v3, v5
-; GFX7-NEXT: v_lshlrev_b32_e32 v2, v4, v2
-; GFX7-NEXT: v_or_b32_e32 v2, v7, v2
; GFX7-NEXT: v_mov_b32_e32 v8, v3
-; GFX7-NEXT: v_mov_b32_e32 v7, v2
-; GFX7-NEXT: buffer_atomic_cmpswap v[7:8], v[0:1], s[4:7], 0 addr64 glc
+; GFX7-NEXT: v_lshrrev_b32_e32 v3, v5, v8
+; GFX7-NEXT: v_lshlrev_b32_e32 v3, 16, v3
+; GFX7-NEXT: v_sub_f32_e32 v3, v3, v2
+; GFX7-NEXT: v_lshrrev_b32_e32 v3, 16, v3
+; GFX7-NEXT: v_and_b32_e32 v4, v8, v6
+; GFX7-NEXT: v_lshlrev_b32_e32 v3, v5, v3
+; GFX7-NEXT: v_or_b32_e32 v7, v4, v3
+; GFX7-NEXT: v_mov_b32_e32 v3, v7
+; GFX7-NEXT: v_mov_b32_e32 v4, v8
+; GFX7-NEXT: buffer_atomic_cmpswap v[3:4], v[0:1], s[4:7], 0 addr64 glc
; GFX7-NEXT: s_waitcnt vmcnt(0)
; GFX7-NEXT: buffer_wbinvl1
-; GFX7-NEXT: v_cmp_eq_u32_e32 vcc, v7, v3
+; GFX7-NEXT: v_cmp_eq_u32_e32 vcc, v3, v8
; GFX7-NEXT: s_or_b64 s[8:9], vcc, s[8:9]
-; GFX7-NEXT: v_mov_b32_e32 v3, v7
; GFX7-NEXT: s_andn2_b64 exec, exec, s[8:9]
; GFX7-NEXT: s_cbranch_execnz .LBB36_1
; GFX7-NEXT: ; %bb.2: ; %atomicrmw.end
@@ -11734,31 +11736,31 @@ define void @global_agent_atomic_fsub_noret_bf16__offset12b_pos(ptr addrspace(1)
; GFX6-NEXT: s_mov_b32 s5, s6
; GFX6-NEXT: buffer_load_dword v3, v[0:1], s[4:7], 0 addr64
; GFX6-NEXT: v_and_b32_e32 v4, 3, v4
-; GFX6-NEXT: v_lshlrev_b32_e32 v4, 3, v4
-; GFX6-NEXT: v_lshl_b32_e32 v5, 0xffff, v4
+; GFX6-NEXT: v_lshlrev_b32_e32 v5, 3, v4
+; GFX6-NEXT: v_lshl_b32_e32 v4, 0xffff, v5
; GFX6-NEXT: v_mul_f32_e32 v2, 1.0, v2
-; GFX6-NEXT: v_not_b32_e32 v5, v5
+; GFX6-NEXT: v_not_b32_e32 v6, v4
; GFX6-NEXT: s_mov_b64 s[8:9], 0
-; GFX6-NEXT: v_and_b32_e32 v6, 0xffff0000, v2
+; GFX6-NEXT: v_and_b32_e32 v2, 0xffff0000, v2
; GFX6-NEXT: .LBB36_1: ; %atomicrmw.start
; GFX6-NEXT: ; =>This Inner Loop Header: Depth=1
; GFX6-NEXT: s_waitcnt vmcnt(0)
-; GFX6-NEXT: v_lshrrev_b32_e32 v2, v4, v3
-; GFX6-NEXT: v_lshlrev_b32_e32 v2, 16, v2
-; GFX6-NEXT: v_sub_f32_e32 v2, v2, v6
-; GFX6-NEXT: v_lshrrev_b32_e32 v2, 16, v2
-; GFX6-NEXT: s_waitcnt expcnt(0)
-; GFX6-NEXT: v_and_b32_e32 v7, v3, v5
-; GFX6-NEXT: v_lshlrev_b32_e32 v2, v4, v2
-; GFX6-NEXT: v_or_b32_e32 v2, v7, v2
; GFX6-NEXT: v_mov_b32_e32 v8, v3
-; GFX6-NEXT: v_mov_b32_e32 v7, v2
-; GFX6-NEXT: buffer_atomic_cmpswap v[7:8], v[0:1], s[4:7], 0 addr64 glc
+; GFX6-NEXT: s_waitcnt expcnt(0)
+; GFX6-NEXT: v_lshrrev_b32_e32 v3, v5, v8
+; GFX6-NEXT: v_lshlrev_b32_e32 v3, 16, v3
+; GFX6-NEXT: v_sub_f32_e32 v3, v3, v2
+; GFX6-NEXT: v_lshrrev_b32_e32 v3, 16, v3
+; GFX6-NEXT: v_and_b32_e32 v4, v8, v6
+; GFX6-NEXT: v_lshlrev_b32_e32 v3, v5, v3
+; GFX6-NEXT: v_or_b32_e32 v7, v4, v3
+; GFX6-NEXT: v_mov_b32_e32 v3, v7
+; GFX6-NEXT: v_mov_b32_e32 v4, v8
+; GFX6-NEXT: buffer_atomic_cmpswap v[3:4], v[0:1], s[4:7], 0 addr64 glc
; GFX6-NEXT: s_waitcnt vmcnt(0)
; GFX6-NEXT: buffer_wbinvl1
-; GFX6-NEXT: v_cmp_eq_u32_e32 vcc, v7, v3
+; GFX6-NEXT: v_cmp_eq_u32_e32 vcc, v3, v8
; GFX6-NEXT: s_or_b64 s[8:9], vcc, s[8:9]
-; GFX6-NEXT: v_mov_b32_e32 v3, v7
; GFX6-NEXT: s_andn2_b64 exec, exec, s[8:9]
; GFX6-NEXT: s_cbranch_execnz .LBB36_1
; GFX6-NEXT: ; %bb.2: ; %atomicrmw.end
@@ -12192,30 +12194,30 @@ define void @global_agent_atomic_fsub_noret_bf16__offset12b_neg(ptr addrspace(1)
; GFX7-NEXT: s_mov_b32 s5, s6
; GFX7-NEXT: buffer_load_dword v3, v[0:1], s[4:7], 0 addr64
; GFX7-NEXT: v_and_b32_e32 v4, 3, v4
-; GFX7-NEXT: v_lshlrev_b32_e32 v4, 3, v4
-; GFX7-NEXT: v_lshl_b32_e32 v5, 0xffff, v4
+; GFX7-NEXT: v_lshlrev_b32_e32 v5, 3, v4
+; GFX7-NEXT: v_lshl_b32_e32 v4, 0xffff, v5
; GFX7-NEXT: v_mul_f32_e32 v2, 1.0, v2
-; GFX7-NEXT: v_not_b32_e32 v5, v5
+; GFX7-NEXT: v_not_b32_e32 v6, v4
; GFX7-NEXT: s_mov_b64 s[8:9], 0
-; GFX7-NEXT: v_and_b32_e32 v6, 0xffff0000, v2
+; GFX7-NEXT: v_and_b32_e32 v2, 0xffff0000, v2
; GFX7-NEXT: .LBB37_1: ; %atomicrmw.start
; GFX7-NEXT: ; =>This Inner Loop Header: Depth=1
; GFX7-NEXT: s_waitcnt vmcnt(0)
-; GFX7-NEXT: v_lshrrev_b32_e32 v2, v4, v3
-; GFX7-NEXT: v_lshlrev_b32_e32 v2, 16, v2
-; GFX7-NEXT: v_sub_f32_e32 v2, v2, v6
-; GFX7-NEXT: v_lshrrev_b32_e32 v2, 16, v2
-; GFX7-NEXT: v_and_b32_e32 v7, v3, v5
-; GFX7-NEXT: v_lshlrev_b32_e32 v2, v4, v2
-; GFX7-NEXT: v_or_b32_e32 v2, v7, v2
; GFX7-NEXT: v_mov_b32_e32 v8, v3
-; GFX7-NEXT: v_mov_b32_e32 v7, v2
-; GFX7-NEXT: buffer_atomic_cmpswap v[7:8], v[0:1], s[4:7], 0 addr64 glc
+; GFX7-NEXT: v_lshrrev_b32_e32 v3, v5, v8
+; GFX7-NEXT: v_lshlrev_b32_e32 v3, 16, v3
+; GFX7-NEXT: v_sub_f32_e32 v3, v3, v2
+; GFX7-NEXT: v_lshrrev_b32_e32 v3, 16, v3
+; GFX7-NEXT: v_and_b32_e32 v4, v8, v6
+; GFX7-NEXT: v_lshlrev_b32_e32 v3, v5, v3
+; GFX7-NEXT: v_or_b32_e32 v7, v4, v3
+; GFX7-NEXT: v_mov_b32_e32 v3, v7
+; GFX7-NEXT: v_mov_b32_e32 v4, v8
+; GFX7-NEXT: buffer_atomic_cmpswap v[3:4], v[0:1], s[4:7], 0 addr64 glc
; GFX7-NEXT: s_waitcnt vmcnt(0)
; GFX7-NEXT: buffer_wbinvl1
-; GFX7-NEXT: v_cmp_eq_u32_e32 vcc, v7, v3
+; GFX7-NEXT: v_cmp_eq_u32_e32 vcc, v3, v8
; GFX7-NEXT: s_or_b64 s[8:9], vcc, s[8:9]
-; GFX7-NEXT: v_mov_b32_e32 v3, v7
; GFX7-NEXT: s_andn2_b64 exec, exec, s[8:9]
; GFX7-NEXT: s_cbranch_execnz .LBB37_1
; GFX7-NEXT: ; %bb.2: ; %atomicrmw.end
@@ -12234,31 +12236,31 @@ define void @global_agent_atomic_fsub_noret_bf16__offset12b_neg(ptr addrspace(1)
; GFX6-NEXT: s_mov_b32 s5, s6
; GFX6-NEXT: buffer_load_dword v3, v[0:1], s[4:7], 0 addr64
; GFX6-NEXT: v_and_b32_e32 v4, 3, v4
-; GFX6-NEXT: v_lshlrev_b32_e32 v4, 3, v4
-; GFX6-NEXT: v_lshl_b32_e32 v5, 0xffff, v4
+; GFX6-NEXT: v_lshlrev_b32_e32 v5, 3, v4
+; GFX6-NEXT: v_lshl_b32_e32 v4, 0xffff, v5
; GFX6-NEXT: v_mul_f32_e32 v2, 1.0, v2
-; GFX6-NEXT: v_not_b32_e32 v5, v5
+; GFX6-NEXT: v_not_b32_e32 v6, v4
; GFX6-NEXT: s_mov_b64 s[8:9], 0
-; GFX6-NEXT: v_and_b32_e32 v6, 0xffff0000, v2
+; GFX6-NEXT: v_and_b32_e32 v2, 0xffff0000, v2
; GFX6-NEXT: .LBB37_1: ; %atomicrmw.start
; GFX6-NEXT: ; =>This Inner Loop Header: Depth=1
; GFX6-NEXT: s_waitcnt vmcnt(0)
-; GFX6-NEXT: v_lshrrev_b32_e32 v2, v4, v3
-; GFX6-NEXT: v_lshlrev_b32_e32 v2, 16, v2
-; GFX6-NEXT: v_sub_f32_e32 v2, v2, v6
-; GFX6-NEXT: v_lshrrev_b32_e32 v2, 16, v2
-; GFX6-NEXT: s_waitcnt expcnt(0)
-; GFX6-NEXT: v_and_b32_e32 v7, v3, v5
-; GFX6-NEXT: v_lshlrev_b32_e32 v2, v4, v2
-; GFX6-NEXT: v_or_b32_e32 v2, v7, v2
; GFX6-NEXT: v_mov_b32_e32 v8, v3
-; GFX6-NEXT: v_mov_b32_e32 v7, v2
-; GFX6-NEXT: buffer_atomic_cmpswap v[7:8], v[0:1], s[4:7], 0 addr64 glc
+; GFX6-NEXT: s_waitcnt expcnt(0)
+; GFX6-NEXT: v_lshrrev_b32_e32 v3, v5, v8
+; GFX6-NEXT: v_lshlrev_b32_e32 v3, 16, v3
+; GFX6-NEXT: v_sub_f32_e32 v3, v3, v2
+; GFX6-NEXT: v_lshrrev_b32_e32 v3, 16, v3
+; GFX6-NEXT: v_and_b32_e32 v4, v8, v6
+; GFX6-NEXT: v_lshlrev_b32_e32 v3, v5, v3
+; GFX6-NEXT: v_or_b32_e32 v7, v4, v3
+; GFX6-NEXT: v_mov_b32_e32 v3, v7
+; GFX6-NEXT: v_mov_b32_e32 v4, v8
+; GFX6-NEXT: buffer_atomic_cmpswap v[3:4], v[0:1], s[4:7], 0 addr64 glc
; GFX6-NEXT: s_waitcnt vmcnt(0)
; GFX6-NEXT: buffer_wbinvl1
-; GFX6-NEXT: v_cmp_eq_u32_e32 vcc, v7, v3
+; GFX6-NEXT: v_cmp_eq_u32_e32 vcc, v3, v8
; GFX6-NEXT: s_or_b64 s[8:9], vcc, s[8:9]
-; GFX6-NEXT: v_mov_b32_e32 v3, v7
; GFX6-NEXT: s_andn2_b64 exec, exec, s[8:9]
; GFX6-NEXT: s_cbranch_execnz .LBB37_1
; GFX6-NEXT: ; %bb.2: ; %atomicrmw.end
@@ -12616,28 +12618,28 @@ define bfloat @global_agent_atomic_fsub_ret_bf16__offset12b_pos__align4(ptr addr
; GFX7-NEXT: buffer_load_dword v3, v[0:1], s[4:7], 0 addr64 offset:2046
; GFX7-NEXT: v_mul_f32_e32 v2, 1.0, v2
; GFX7-NEXT: s_mov_b64 s[8:9], 0
-; GFX7-NEXT: v_and_b32_e32 v6, 0xffff0000, v2
+; GFX7-NEXT: v_and_b32_e32 v2, 0xffff0000, v2
; GFX7-NEXT: .LBB38_1: ; %atomicrmw.start
; GFX7-NEXT: ; =>This Inner Loop Header: Depth=1
; GFX7-NEXT: s_waitcnt vmcnt(0)
-; GFX7-NEXT: v_lshlrev_b32_e32 v2, 16, v3
-; GFX7-NEXT: v_sub_f32_e32 v2, v2, v6
-; GFX7-NEXT: v_and_b32_e32 v4, 0xffff0000, v3
-; GFX7-NEXT: v_lshrrev_b32_e32 v2, 16, v2
-; GFX7-NEXT: v_or_b32_e32 v2, v4, v2
; GFX7-NEXT: v_mov_b32_e32 v5, v3
-; GFX7-NEXT: v_mov_b32_e32 v4, v2
-; GFX7-NEXT: buffer_atomic_cmpswap v[4:5], v[0:1], s[4:7], 0 addr64 offset:2046 glc
+; GFX7-NEXT: v_lshlrev_b32_e32 v3, 16, v5
+; GFX7-NEXT: v_sub_f32_e32 v3, v3, v2
+; GFX7-NEXT: v_and_b32_e32 v4, 0xffff0000, v5
+; GFX7-NEXT: v_lshrrev_b32_e32 v3, 16, v3
+; GFX7-NEXT: v_or_b32_e32 v4, v4, v3
+; GFX7-NEXT: v_mov_b32_e32 v3, v4
+; GFX7-NEXT: v_mov_b32_e32 v4, v5
+; GFX7-NEXT: buffer_atomic_cmpswap v[3:4], v[0:1], s[4:7], 0 addr64 offset:2046 glc
; GFX7-NEXT: s_waitcnt vmcnt(0)
; GFX7-NEXT: buffer_wbinvl1
-; GFX7-NEXT: v_cmp_eq_u32_e32 vcc, v4, v3
+; GFX7-NEXT: v_cmp_eq_u32_e32 vcc, v3, v5
; GFX7-NEXT: s_or_b64 s[8:9], vcc, s[8:9]
-; GFX7-NEXT: v_mov_b32_e32 v3, v4
; GFX7-NEXT: s_andn2_b64 exec, exec, s[8:9]
; GFX7-NEXT: s_cbranch_execnz .LBB38_1
; GFX7-NEXT: ; %bb.2: ; %atomicrmw.end
; GFX7-NEXT: s_or_b64 exec, exec, s[8:9]
-; GFX7-NEXT: v_lshlrev_b32_e32 v0, 16, v4
+; GFX7-NEXT: v_lshlrev_b32_e32 v0, 16, v3
; GFX7-NEXT: s_setpc_b64 s[30:31]
;
; GFX6-LABEL: global_agent_atomic_fsub_ret_bf16__offset12b_pos__align4:
@@ -12650,29 +12652,29 @@ define bfloat @global_agent_atomic_fsub_ret_bf16__offset12b_pos__align4(ptr addr
; GFX6-NEXT: buffer_load_dword v3, v[0:1], s[4:7], 0 addr64 offset:2046
; GFX6-NEXT: v_mul_f32_e32 v2, 1.0, v2
; GFX6-NEXT: s_mov_b64 s[8:9], 0
-; GFX6-NEXT: v_and_b32_e32 v6, 0xffff0000, v2
+; GFX6-NEXT: v_and_b32_e32 v2, 0xffff0000, v2
; GFX6-NEXT: .LBB38_1: ; %atomicrmw.start
; GFX6-NEXT: ; =>This Inner Loop Header: Depth=1
; GFX6-NEXT: s_waitcnt vmcnt(0)
-; GFX6-NEXT: v_lshlrev_b32_e32 v2, 16, v3
-; GFX6-NEXT: v_sub_f32_e32 v2, v2, v6
-; GFX6-NEXT: s_waitcnt expcnt(0)
-; GFX6-NEXT: v_and_b32_e32 v4, 0xffff0000, v3
-; GFX6-NEXT: v_lshrrev_b32_e32 v2, 16, v2
-; GFX6-NEXT: v_or_b32_e32 v2, v4, v2
; GFX6-NEXT: v_mov_b32_e32 v5, v3
-; GFX6-NEXT: v_mov_b32_e32 v4, v2
-; GFX6-NEXT: buffer_atomic_cmpswap v[4:5], v[0:1], s[4:7], 0 addr64 offset:2046 glc
+; GFX6-NEXT: s_waitcnt expcnt(0)
+; GFX6-NEXT: v_lshlrev_b32_e32 v3, 16, v5
+; GFX6-NEXT: v_sub_f32_e32 v3, v3, v2
+; GFX6-NEXT: v_and_b32_e32 v4, 0xffff0000, v5
+; GFX6-NEXT: v_lshrrev_b32_e32 v3, 16, v3
+; GFX6-NEXT: v_or_b32_e32 v4, v4, v3
+; GFX6-NEXT: v_mov_b32_e32 v3, v4
+; GFX6-NEXT: v_mov_b32_e32 v4, v5
+; GFX6-NEXT: buffer_atomic_cmpswap v[3:4], v[0:1], s[4:7], 0 addr64 offset:2046 glc
; GFX6-NEXT: s_waitcnt vmcnt(0)
; GFX6-NEXT: buffer_wbinvl1
-; GFX6-NEXT: v_cmp_eq_u32_e32 vcc, v4, v3
+; GFX6-NEXT: v_cmp_eq_u32_e32 vcc, v3, v5
; GFX6-NEXT: s_or_b64 s[8:9], vcc, s[8:9]
-; GFX6-NEXT: v_mov_b32_e32 v3, v4
; GFX6-NEXT: s_andn2_b64 exec, exec, s[8:9]
; GFX6-NEXT: s_cbranch_execnz .LBB38_1
; GFX6-NEXT: ; %bb.2: ; %atomicrmw.end
; GFX6-NEXT: s_or_b64 exec, exec, s[8:9]
-; GFX6-NEXT: v_lshlrev_b32_e32 v0, 16, v4
+; GFX6-NEXT: v_lshlrev_b32_e32 v0, 16, v3
; GFX6-NEXT: s_waitcnt expcnt(0)
; GFX6-NEXT: s_setpc_b64 s[30:31]
%gep = getelementptr bfloat, ptr addrspace(1) %ptr, i64 1023
@@ -13014,23 +13016,23 @@ define void @global_agent_atomic_fsub_noret_bf16__offset12b__align4_pos(ptr addr
; GFX7-NEXT: buffer_load_dword v3, v[0:1], s[4:7], 0 addr64 offset:2046
; GFX7-NEXT: v_mul_f32_e32 v2, 1.0, v2
; GFX7-NEXT: s_mov_b64 s[8:9], 0
-; GFX7-NEXT: v_and_b32_e32 v4, 0xffff0000, v2
+; GFX7-NEXT: v_and_b32_e32 v2, 0xffff0000, v2
; GFX7-NEXT: .LBB39_1: ; %atomicrmw.start
; GFX7-NEXT: ; =>This Inner Loop Header: Depth=1
; GFX7-NEXT: s_waitcnt vmcnt(0)
-; GFX7-NEXT: v_lshlrev_b32_e32 v2, 16, v3
-; GFX7-NEXT: v_sub_f32_e32 v2, v2, v4
-; GFX7-NEXT: v_and_b32_e32 v5, 0xffff0000, v3
-; GFX7-NEXT: v_lshrrev_b32_e32 v2, 16, v2
-; GFX7-NEXT: v_or_b32_e32 v2, v5, v2
-; GFX7-NEXT: v_mov_b32_e32 v6, v3
-; GFX7-NEXT: v_mov_b32_e32 v5, v2
-; GFX7-NEXT: buffer_atomic_cmpswap v[5:6], v[0:1], s[4:7], 0 addr64 offset:2046 glc
+; GFX7-NEXT: v_mov_b32_e32 v5, v3
+; GFX7-NEXT: v_lshlrev_b32_e32 v3, 16, v5
+; GFX7-NEXT: v_sub_f32_e32 v3, v3, v2
+; GFX7-NEXT: v_and_b32_e32 v4, 0xffff0000, v5
+; GFX7-NEXT: v_lshrrev_b32_e32 v3, 16, v3
+; GFX7-NEXT: v_or_b32_e32 v4, v4, v3
+; GFX7-NEXT: v_mov_b32_e32 v3, v4
+; GFX7-NEXT: v_mov_b32_e32 v4, v5
+; GFX7-NEXT: buffer_atomic_cmpswap v[3:4], v[0:1], s[4:7], 0 addr64 offset:2046 glc
; GFX7-NEXT: s_waitcnt vmcnt(0)
; GFX7-NEXT: buffer_wbinvl1
-; GFX7-NEXT: v_cmp_eq_u32_e32 vcc, v5, v3
+; GFX7-NEXT: v_cmp_eq_u32_e32 vcc, v3, v5
; GFX7-NEXT: s_or_b64 s[8:9], vcc, s[8:9]
-; GFX7-NEXT: v_mov_b32_e32 v3, v5
; GFX7-NEXT: s_andn2_b64 exec, exec, s[8:9]
; GFX7-NEXT: s_cbranch_execnz .LBB39_1
; GFX7-NEXT: ; %bb.2: ; %atomicrmw.end
@@ -13047,24 +13049,24 @@ define void @global_agent_atomic_fsub_noret_bf16__offset12b__align4_pos(ptr addr
; GFX6-NEXT: buffer_load_dword v3, v[0:1], s[4:7], 0 addr64 offset:2046
; GFX6-NEXT: v_mul_f32_e32 v2, 1.0, v2
; GFX6-NEXT: s_mov_b64 s[8:9], 0
-; GFX6-NEXT: v_and_b32_e32 v4, 0xffff0000, v2
+; GFX6-NEXT: v_and_b32_e32 v2, 0xffff0000, v2
; GFX6-NEXT: .LBB39_1: ; %atomicrmw.start
; GFX6-NEXT: ; =>This Inner Loop Header: Depth=1
; GFX6-NEXT: s_waitcnt vmcnt(0)
-; GFX6-NEXT: v_lshlrev_b32_e32 v2, 16, v3
-; GFX6-NEXT: v_sub_f32_e32 v2, v2, v4
+; GFX6-NEXT: v_mov_b32_e32 v5, v3
; GFX6-NEXT: s_waitcnt expcnt(0)
-; GFX6-NEXT: v_and_b32_e32 v5, 0xffff0000, v3
-; GFX6-NEXT: v_lshrrev_b32_e32 v2, 16, v2
-; GFX6-NEXT: v_or_b32_e32 v2, v5, v2
-; GFX6-NEXT: v_mov_b32_e32 v6, v3
-; GFX6-NEXT: v_mov_b32_e32 v5, v2
-; GFX6-NEXT: buffer_atomic_cmpswap v[5:6], v[0:1], s[4:7], 0 addr64 offset:2046 glc
+; GFX6-NEXT: v_lshlrev_b32_e32 v3, 16, v5
+; GFX6-NEXT: v_sub_f32_e32 v3, v3, v2
+; GFX6-NEXT: v_and_b32_e32 v4, 0xffff0000, v5
+; GFX6-NEXT: v_lshrrev_b32_e32 v3, 16, v3
+; GFX6-NEXT: v_or_b32_e32 v4, v4, v3
+; GFX6-NEXT: v_mov_b32_e32 v3, v4
+; GFX6-NEXT: v_mov_b32_e32 v4, v5
+; GFX6-NEXT: buffer_atomic_cmpswap v[3:4], v[0:1], s[4:7], 0 addr64 offset:2046 glc
; GFX6-NEXT: s_waitcnt vmcnt(0)
; GFX6-NEXT: buffer_wbinvl1
-; GFX6-NEXT: v_cmp_eq_u32_e32 vcc, v5, v3
+; GFX6-NEXT: v_cmp_eq_u32_e32 vcc, v3, v5
; GFX6-NEXT: s_or_b64 s[8:9], vcc, s[8:9]
-; GFX6-NEXT: v_mov_b32_e32 v3, v5
; GFX6-NEXT: s_andn2_b64 exec, exec, s[8:9]
; GFX6-NEXT: s_cbranch_execnz .LBB39_1
; GFX6-NEXT: ; %bb.2: ; %atomicrmw.end
@@ -13514,35 +13516,35 @@ define bfloat @global_system_atomic_fsub_ret_bf16__offset12b_pos(ptr addrspace(1
; GFX7-NEXT: s_mov_b32 s5, s6
; GFX7-NEXT: buffer_load_dword v3, v[0:1], s[4:7], 0 addr64
; GFX7-NEXT: v_and_b32_e32 v4, 3, v4
-; GFX7-NEXT: v_lshlrev_b32_e32 v6, 3, v4
-; GFX7-NEXT: v_lshl_b32_e32 v4, 0xffff, v6
+; GFX7-NEXT: v_lshlrev_b32_e32 v5, 3, v4
+; GFX7-NEXT: v_lshl_b32_e32 v4, 0xffff, v5
; GFX7-NEXT: v_mul_f32_e32 v2, 1.0, v2
-; GFX7-NEXT: v_not_b32_e32 v7, v4
+; GFX7-NEXT: v_not_b32_e32 v6, v4
; GFX7-NEXT: s_mov_b64 s[8:9], 0
-; GFX7-NEXT: v_and_b32_e32 v8, 0xffff0000, v2
+; GFX7-NEXT: v_and_b32_e32 v2, 0xffff0000, v2
; GFX7-NEXT: .LBB40_1: ; %atomicrmw.start
; GFX7-NEXT: ; =>This Inner Loop Header: Depth=1
; GFX7-NEXT: s_waitcnt vmcnt(0)
-; GFX7-NEXT: v_lshrrev_b32_e32 v2, v6, v3
-; GFX7-NEXT: v_lshlrev_b32_e32 v2, 16, v2
-; GFX7-NEXT: v_sub_f32_e32 v2, v2, v8
-; GFX7-NEXT: v_lshrrev_b32_e32 v2, 16, v2
-; GFX7-NEXT: v_and_b32_e32 v4, v3, v7
-; GFX7-NEXT: v_lshlrev_b32_e32 v2, v6, v2
-; GFX7-NEXT: v_or_b32_e32 v2, v4, v2
-; GFX7-NEXT: v_mov_b32_e32 v5, v3
-; GFX7-NEXT: v_mov_b32_e32 v4, v2
-; GFX7-NEXT: buffer_atomic_cmpswap v[4:5], v[0:1], s[4:7], 0 addr64 glc
+; GFX7-NEXT: v_mov_b32_e32 v8, v3
+; GFX7-NEXT: v_lshrrev_b32_e32 v3, v5, v8
+; GFX7-NEXT: v_lshlrev_b32_e32 v3, 16, v3
+; GFX7-NEXT: v_sub_f32_e32 v3, v3, v2
+; GFX7-NEXT: v_lshrrev_b32_e32 v3, 16, v3
+; GFX7-NEXT: v_and_b32_e32 v4, v8, v6
+; GFX7-NEXT: v_lshlrev_b32_e32 v3, v5, v3
+; GFX7-NEXT: v_or_b32_e32 v7, v4, v3
+; GFX7-NEXT: v_mov_b32_e32 v3, v7
+; GFX7-NEXT: v_mov_b32_e32 v4, v8
+; GFX7-NEXT: buffer_atomic_cmpswap v[3:4], v[0:1], s[4:7], 0 addr64 glc
; GFX7-NEXT: s_waitcnt vmcnt(0)
; GFX7-NEXT: buffer_wbinvl1
-; GFX7-NEXT: v_cmp_eq_u32_e32 vcc, v4, v3
+; GFX7-NEXT: v_cmp_eq_u32_e32 vcc, v3, v8
; GFX7-NEXT: s_or_b64 s[8:9], vcc, s[8:9]
-; GFX7-NEXT: v_mov_b32_e32 v3, v4
; GFX7-NEXT: s_andn2_b64 exec, exec, s[8:9]
; GFX7-NEXT: s_cbranch_execnz .LBB40_1
; GFX7-NEXT: ; %bb.2: ; %atomicrmw.end
; GFX7-NEXT: s_or_b64 exec, exec, s[8:9]
-; GFX7-NEXT: v_lshrrev_b32_e32 v0, v6, v4
+; GFX7-NEXT: v_lshrrev_b32_e32 v0, v5, v3
; GFX7-NEXT: v_lshlrev_b32_e32 v0, 16, v0
; GFX7-NEXT: s_setpc_b64 s[30:31]
;
@@ -13558,36 +13560,36 @@ define bfloat @global_system_atomic_fsub_ret_bf16__offset12b_pos(ptr addrspace(1
; GFX6-NEXT: s_mov_b32 s5, s6
; GFX6-NEXT: buffer_load_dword v3, v[0:1], s[4:7], 0 addr64
; GFX6-NEXT: v_and_b32_e32 v4, 3, v4
-; GFX6-NEXT: v_lshlrev_b32_e32 v6, 3, v4
-; GFX6-NEXT: v_lshl_b32_e32 v4, 0xffff, v6
+; GFX6-NEXT: v_lshlrev_b32_e32 v5, 3, v4
+; GFX6-NEXT: v_lshl_b32_e32 v4, 0xffff, v5
; GFX6-NEXT: v_mul_f32_e32 v2, 1.0, v2
-; GFX6-NEXT: v_not_b32_e32 v7, v4
+; GFX6-NEXT: v_not_b32_e32 v6, v4
; GFX6-NEXT: s_mov_b64 s[8:9], 0
-; GFX6-NEXT: v_and_b32_e32 v8, 0xffff0000, v2
+; GFX6-NEXT: v_and_b32_e32 v2, 0xffff0000, v2
; GFX6-NEXT: .LBB40_1: ; %atomicrmw.start
; GFX6-NEXT: ; =>This Inner Loop Header: Depth=1
; GFX6-NEXT: s_waitcnt vmcnt(0)
-; GFX6-NEXT: v_lshrrev_b32_e32 v2, v6, v3
-; GFX6-NEXT: v_lshlrev_b32_e32 v2, 16, v2
-; GFX6-NEXT: v_sub_f32_e32 v2, v2, v8
-; GFX6-NEXT: v_lshrrev_b32_e32 v2, 16, v2
+; GFX6-NEXT: v_mov_b32_e32 v8, v3
; GFX6-NEXT: s_waitcnt expcnt(0)
-; GFX6-NEXT: v_and_b32_e32 v4, v3, v7
-; GFX6-NEXT: v_lshlrev_b32_e32 v2, v6, v2
-; GFX6-NEXT: v_or_b32_e32 v2, v4, v2
-; GFX6-NEXT: v_mov_b32_e32 v5, v3
-; GFX6-NEXT: v_mov_b32_e32 v4, v2
-; GFX6-NEXT: buffer_atomic_cmpswap v[4:5], v[0:1], s[4:7], 0 addr64 glc
+; GFX6-NEXT: v_lshrrev_b32_e32 v3, v5, v8
+; GFX6-NEXT: v_lshlrev_b32_e32 v3, 16, v3
+; GFX6-NEXT: v_sub_f32_e32 v3, v3, v2
+; GFX6-NEXT: v_lshrrev_b32_e32 v3, 16, v3
+; GFX6-NEXT: v_and_b32_e32 v4, v8, v6
+; GFX6-NEXT: v_lshlrev_b32_e32 v3, v5, v3
+; GFX6-NEXT: v_or_b32_e32 v7, v4, v3
+; GFX6-NEXT: v_mov_b32_e32 v3, v7
+; GFX6-NEXT: v_mov_b32_e32 v4, v8
+; GFX6-NEXT: buffer_atomic_cmpswap v[3:4], v[0:1], s[4:7], 0 addr64 glc
; GFX6-NEXT: s_waitcnt vmcnt(0)
; GFX6-NEXT: buffer_wbinvl1
-; GFX6-NEXT: v_cmp_eq_u32_e32 vcc, v4, v3
+; GFX6-NEXT: v_cmp_eq_u32_e32 vcc, v3, v8
; GFX6-NEXT: s_or_b64 s[8:9], vcc, s[8:9]
-; GFX6-NEXT: v_mov_b32_e32 v3, v4
; GFX6-NEXT: s_andn2_b64 exec, exec, s[8:9]
; GFX6-NEXT: s_cbranch_execnz .LBB40_1
; GFX6-NEXT: ; %bb.2: ; %atomicrmw.end
; GFX6-NEXT: s_or_b64 exec, exec, s[8:9]
-; GFX6-NEXT: v_lshrrev_b32_e32 v0, v6, v4
+; GFX6-NEXT: v_lshrrev_b32_e32 v0, v5, v3
; GFX6-NEXT: v_lshlrev_b32_e32 v0, 16, v0
; GFX6-NEXT: s_waitcnt expcnt(0)
; GFX6-NEXT: s_setpc_b64 s[30:31]
@@ -14021,30 +14023,30 @@ define void @global_system_atomic_fsub_noret_bf16__offset12b_pos(ptr addrspace(1
; GFX7-NEXT: s_mov_b32 s5, s6
; GFX7-NEXT: buffer_load_dword v3, v[0:1], s[4:7], 0 addr64
; GFX7-NEXT: v_and_b32_e32 v4, 3, v4
-; GFX7-NEXT: v_lshlrev_b32_e32 v4, 3, v4
-; GFX7-NEXT: v_lshl_b32_e32 v5, 0xffff, v4
+; GFX7-NEXT: v_lshlrev_b32_e32 v5, 3, v4
+; GFX7-NEXT: v_lshl_b32_e32 v4, 0xffff, v5
; GFX7-NEXT: v_mul_f32_e32 v2, 1.0, v2
-; GFX7-NEXT: v_not_b32_e32 v5, v5
+; GFX7-NEXT: v_not_b32_e32 v6, v4
; GFX7-NEXT: s_mov_b64 s[8:9], 0
-; GFX7-NEXT: v_and_b32_e32 v6, 0xffff0000, v2
+; GFX7-NEXT: v_and_b32_e32 v2, 0xffff0000, v2
; GFX7-NEXT: .LBB41_1: ; %atomicrmw.start
; GFX7-NEXT: ; =>This Inner Loop Header: Depth=1
; GFX7-NEXT: s_waitcnt vmcnt(0)
-; GFX7-NEXT: v_lshrrev_b32_e32 v2, v4, v3
-; GFX7-NEXT: v_lshlrev_b32_e32 v2, 16, v2
-; GFX7-NEXT: v_sub_f32_e32 v2, v2, v6
-; GFX7-NEXT: v_lshrrev_b32_e32 v2, 16, v2
-; GFX7-NEXT: v_and_b32_e32 v7, v3, v5
-; GFX7-NEXT: v_lshlrev_b32_e32 v2, v4, v2
-; GFX7-NEXT: v_or_b32_e32 v2, v7, v2
; GFX7-NEXT: v_mov_b32_e32 v8, v3
-; GFX7-NEXT: v_mov_b32_e32 v7, v2
-; GFX7-NEXT: buffer_atomic_cmpswap v[7:8], v[0:1], s[4:7], 0 addr64 glc
+; GFX7-NEXT: v_lshrrev_b32_e32 v3, v5, v8
+; GFX7-NEXT: v_lshlrev_b32_e32 v3, 16, v3
+; GFX7-NEXT: v_sub_f32_e32 v3, v3, v2
+; GFX7-NEXT: v_lshrrev_b32_e32 v3, 16, v3
+; GFX7-NEXT: v_and_b32_e32 v4, v8, v6
+; GFX7-NEXT: v_lshlrev_b32_e32 v3, v5, v3
+; GFX7-NEXT: v_or_b32_e32 v7, v4, v3
+; GFX7-NEXT: v_mov_b32_e32 v3, v7
+; GFX7-NEXT: v_mov_b32_e32 v4, v8
+; GFX7-NEXT: buffer_atomic_cmpswap v[3:4], v[0:1], s[4:7], 0 addr64 glc
; GFX7-NEXT: s_waitcnt vmcnt(0)
; GFX7-NEXT: buffer_wbinvl1
-; GFX7-NEXT: v_cmp_eq_u32_e32 vcc, v7, v3
+; GFX7-NEXT: v_cmp_eq_u32_e32 vcc, v3, v8
; GFX7-NEXT: s_or_b64 s[8:9], vcc, s[8:9]
-; GFX7-NEXT: v_mov_b32_e32 v3, v7
; GFX7-NEXT: s_andn2_b64 exec, exec, s[8:9]
; GFX7-NEXT: s_cbranch_execnz .LBB41_1
; GFX7-NEXT: ; %bb.2: ; %atomicrmw.end
@@ -14063,31 +14065,31 @@ define void @global_system_atomic_fsub_noret_bf16__offset12b_pos(ptr addrspace(1
; GFX6-NEXT: s_mov_b32 s5, s6
; GFX6-NEXT: buffer_load_dword v3, v[0:1], s[4:7], 0 addr64
; GFX6-NEXT: v_and_b32_e32 v4, 3, v4
-; GFX6-NEXT: v_lshlrev_b32_e32 v4, 3, v4
-; GFX6-NEXT: v_lshl_b32_e32 v5, 0xffff, v4
+; GFX6-NEXT: v_lshlrev_b32_e32 v5, 3, v4
+; GFX6-NEXT: v_lshl_b32_e32 v4, 0xffff, v5
; GFX6-NEXT: v_mul_f32_e32 v2, 1.0, v2
-; GFX6-NEXT: v_not_b32_e32 v5, v5
+; GFX6-NEXT: v_not_b32_e32 v6, v4
; GFX6-NEXT: s_mov_b64 s[8:9], 0
-; GFX6-NEXT: v_and_b32_e32 v6, 0xffff0000, v2
+; GFX6-NEXT: v_and_b32_e32 v2, 0xffff0000, v2
; GFX6-NEXT: .LBB41_1: ; %atomicrmw.start
; GFX6-NEXT: ; =>This Inner Loop Header: Depth=1
; GFX6-NEXT: s_waitcnt vmcnt(0)
-; GFX6-NEXT: v_lshrrev_b32_e32 v2, v4, v3
-; GFX6-NEXT: v_lshlrev_b32_e32 v2, 16, v2
-; GFX6-NEXT: v_sub_f32_e32 v2, v2, v6
-; GFX6-NEXT: v_lshrrev_b32_e32 v2, 16, v2
-; GFX6-NEXT: s_waitcnt expcnt(0)
-; GFX6-NEXT: v_and_b32_e32 v7, v3, v5
-; GFX6-NEXT: v_lshlrev_b32_e32 v2, v4, v2
-; GFX6-NEXT: v_or_b32_e32 v2, v7, v2
; GFX6-NEXT: v_mov_b32_e32 v8, v3
-; GFX6-NEXT: v_mov_b32_e32 v7, v2
-; GFX6-NEXT: buffer_atomic_cmpswap v[7:8], v[0:1], s[4:7], 0 addr64 glc
+; GFX6-NEXT: s_waitcnt expcnt(0)
+; GFX6-NEXT: v_lshrrev_b32_e32 v3, v5, v8
+; GFX6-NEXT: v_lshlrev_b32_e32 v3, 16, v3
+; GFX6-NEXT: v_sub_f32_e32 v3, v3, v2
+; GFX6-NEXT: v_lshrrev_b32_e32 v3, 16, v3
+; GFX6-NEXT: v_and_b32_e32 v4, v8, v6
+; GFX6-NEXT: v_lshlrev_b32_e32 v3, v5, v3
+; GFX6-NEXT: v_or_b32_e32 v7, v4, v3
+; GFX6-NEXT: v_mov_b32_e32 v3, v7
+; GFX6-NEXT: v_mov_b32_e32 v4, v8
+; GFX6-NEXT: buffer_atomic_cmpswap v[3:4], v[0:1], s[4:7], 0 addr64 glc
; GFX6-NEXT: s_waitcnt vmcnt(0)
; GFX6-NEXT: buffer_wbinvl1
-; GFX6-NEXT: v_cmp_eq_u32_e32 vcc, v7, v3
+; GFX6-NEXT: v_cmp_eq_u32_e32 vcc, v3, v8
; GFX6-NEXT: s_or_b64 s[8:9], vcc, s[8:9]
-; GFX6-NEXT: v_mov_b32_e32 v3, v7
; GFX6-NEXT: s_andn2_b64 exec, exec, s[8:9]
; GFX6-NEXT: s_cbranch_execnz .LBB41_1
; GFX6-NEXT: ; %bb.2: ; %atomicrmw.end
diff --git a/llvm/test/CodeGen/AMDGPU/global_atomics_i32_system.ll b/llvm/test/CodeGen/AMDGPU/global_atomics_i32_system.ll
index f7882e6f12022..b9a37c4b98a80 100644
--- a/llvm/test/CodeGen/AMDGPU/global_atomics_i32_system.ll
+++ b/llvm/test/CodeGen/AMDGPU/global_atomics_i32_system.ll
@@ -2167,22 +2167,22 @@ define void @global_atomic_nand_i32_noret(ptr addrspace(1) %ptr, i32 %in) {
; SI-NEXT: s_mov_b32 s7, 0xf000
; SI-NEXT: s_mov_b32 s4, s6
; SI-NEXT: s_mov_b32 s5, s6
-; SI-NEXT: buffer_load_dword v4, v[0:1], s[4:7], 0 addr64
+; SI-NEXT: buffer_load_dword v3, v[0:1], s[4:7], 0 addr64
; SI-NEXT: s_mov_b64 s[8:9], 0
; SI-NEXT: .LBB51_1: ; %atomicrmw.start
; SI-NEXT: ; =>This Inner Loop Header: Depth=1
; SI-NEXT: s_waitcnt vmcnt(0)
-; SI-NEXT: v_and_b32_e32 v3, v4, v2
-; SI-NEXT: v_not_b32_e32 v3, v3
-; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_mov_b32_e32 v6, v4
; SI-NEXT: v_mov_b32_e32 v5, v3
-; SI-NEXT: buffer_atomic_cmpswap v[5:6], v[0:1], s[4:7], 0 addr64 glc
+; SI-NEXT: s_waitcnt expcnt(0)
+; SI-NEXT: v_and_b32_e32 v3, v5, v2
+; SI-NEXT: v_not_b32_e32 v4, v3
+; SI-NEXT: v_mov_b32_e32 v3, v4
+; SI-NEXT: v_mov_b32_e32 v4, v5
+; SI-NEXT: buffer_atomic_cmpswap v[3:4], v[0:1], s[4:7], 0 addr64 glc
; SI-NEXT: s_waitcnt vmcnt(0)
; SI-NEXT: buffer_wbinvl1
-; SI-NEXT: v_cmp_eq_u32_e32 vcc, v5, v4
+; SI-NEXT: v_cmp_eq_u32_e32 vcc, v3, v5
; SI-NEXT: s_or_b64 s[8:9], vcc, s[8:9]
-; SI-NEXT: v_mov_b32_e32 v4, v5
; SI-NEXT: s_andn2_b64 exec, exec, s[8:9]
; SI-NEXT: s_cbranch_execnz .LBB51_1
; SI-NEXT: ; %bb.2: ; %atomicrmw.end
@@ -2245,22 +2245,22 @@ define void @global_atomic_nand_i32_noret_offset(ptr addrspace(1) %out, i32 %in)
; SI-NEXT: s_mov_b32 s7, 0xf000
; SI-NEXT: s_mov_b32 s4, s6
; SI-NEXT: s_mov_b32 s5, s6
-; SI-NEXT: buffer_load_dword v4, v[0:1], s[4:7], 0 addr64 offset:16
+; SI-NEXT: buffer_load_dword v3, v[0:1], s[4:7], 0 addr64 offset:16
; SI-NEXT: s_mov_b64 s[8:9], 0
; SI-NEXT: .LBB52_1: ; %atomicrmw.start
; SI-NEXT: ; =>This Inner Loop Header: Depth=1
; SI-NEXT: s_waitcnt vmcnt(0)
-; SI-NEXT: v_and_b32_e32 v3, v4, v2
-; SI-NEXT: v_not_b32_e32 v3, v3
-; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_mov_b32_e32 v6, v4
; SI-NEXT: v_mov_b32_e32 v5, v3
-; SI-NEXT: buffer_atomic_cmpswap v[5:6], v[0:1], s[4:7], 0 addr64 offset:16 glc
+; SI-NEXT: s_waitcnt expcnt(0)
+; SI-NEXT: v_and_b32_e32 v3, v5, v2
+; SI-NEXT: v_not_b32_e32 v4, v3
+; SI-NEXT: v_mov_b32_e32 v3, v4
+; SI-NEXT: v_mov_b32_e32 v4, v5
+; SI-NEXT: buffer_atomic_cmpswap v[3:4], v[0:1], s[4:7], 0 addr64 offset:16 glc
; SI-NEXT: s_waitcnt vmcnt(0)
; SI-NEXT: buffer_wbinvl1
-; SI-NEXT: v_cmp_eq_u32_e32 vcc, v5, v4
+; SI-NEXT: v_cmp_eq_u32_e32 vcc, v3, v5
; SI-NEXT: s_or_b64 s[8:9], vcc, s[8:9]
-; SI-NEXT: v_mov_b32_e32 v4, v5
; SI-NEXT: s_andn2_b64 exec, exec, s[8:9]
; SI-NEXT: s_cbranch_execnz .LBB52_1
; SI-NEXT: ; %bb.2: ; %atomicrmw.end
@@ -2487,38 +2487,38 @@ define amdgpu_gfx void @global_atomic_nand_i32_noret_scalar(ptr addrspace(1) inr
; SI: ; %bb.0:
; SI-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; SI-NEXT: s_xor_saveexec_b64 s[34:35], -1
-; SI-NEXT: buffer_store_dword v4, off, s[0:3], s32 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v3, off, s[0:3], s32 ; 4-byte Folded Spill
; SI-NEXT: s_mov_b64 exec, s[34:35]
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_writelane_b32 v4, s6, 0
-; SI-NEXT: v_writelane_b32 v4, s7, 1
+; SI-NEXT: v_writelane_b32 v3, s6, 0
+; SI-NEXT: v_writelane_b32 v3, s7, 1
; SI-NEXT: s_mov_b32 s34, s6
; SI-NEXT: s_mov_b32 s7, 0xf000
; SI-NEXT: s_mov_b32 s6, -1
-; SI-NEXT: buffer_load_dword v1, off, s[4:7], 0
+; SI-NEXT: buffer_load_dword v0, off, s[4:7], 0
; SI-NEXT: s_mov_b64 s[36:37], 0
; SI-NEXT: .LBB55_1: ; %atomicrmw.start
; SI-NEXT: ; =>This Inner Loop Header: Depth=1
; SI-NEXT: s_waitcnt vmcnt(0)
-; SI-NEXT: v_and_b32_e32 v0, s34, v1
-; SI-NEXT: v_not_b32_e32 v0, v0
-; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_mov_b32_e32 v3, v1
; SI-NEXT: v_mov_b32_e32 v2, v0
-; SI-NEXT: buffer_atomic_cmpswap v[2:3], off, s[4:7], 0 glc
+; SI-NEXT: s_waitcnt expcnt(0)
+; SI-NEXT: v_and_b32_e32 v0, s34, v2
+; SI-NEXT: v_not_b32_e32 v1, v0
+; SI-NEXT: v_mov_b32_e32 v0, v1
+; SI-NEXT: v_mov_b32_e32 v1, v2
+; SI-NEXT: buffer_atomic_cmpswap v[0:1], off, s[4:7], 0 glc
; SI-NEXT: s_waitcnt vmcnt(0)
; SI-NEXT: buffer_wbinvl1
-; SI-NEXT: v_cmp_eq_u32_e32 vcc, v2, v1
+; SI-NEXT: v_cmp_eq_u32_e32 vcc, v0, v2
; SI-NEXT: s_or_b64 s[36:37], vcc, s[36:37]
-; SI-NEXT: v_mov_b32_e32 v1, v2
; SI-NEXT: s_andn2_b64 exec, exec, s[36:37]
; SI-NEXT: s_cbranch_execnz .LBB55_1
; SI-NEXT: ; %bb.2: ; %atomicrmw.end
; SI-NEXT: s_or_b64 exec, exec, s[36:37]
-; SI-NEXT: v_readlane_b32 s7, v4, 1
-; SI-NEXT: v_readlane_b32 s6, v4, 0
+; SI-NEXT: v_readlane_b32 s7, v3, 1
+; SI-NEXT: v_readlane_b32 s6, v3, 0
; SI-NEXT: s_xor_saveexec_b64 s[34:35], -1
-; SI-NEXT: buffer_load_dword v4, off, s[0:3], s32 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v3, off, s[0:3], s32 ; 4-byte Folded Reload
; SI-NEXT: s_mov_b64 exec, s[34:35]
; SI-NEXT: s_waitcnt vmcnt(0) expcnt(0)
; SI-NEXT: s_setpc_b64 s[30:31]
@@ -2578,38 +2578,38 @@ define amdgpu_gfx void @global_atomic_nand_i32_noret_offset_scalar(ptr addrspace
; SI: ; %bb.0:
; SI-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; SI-NEXT: s_xor_saveexec_b64 s[34:35], -1
-; SI-NEXT: buffer_store_dword v4, off, s[0:3], s32 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v3, off, s[0:3], s32 ; 4-byte Folded Spill
; SI-NEXT: s_mov_b64 exec, s[34:35]
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_writelane_b32 v4, s6, 0
-; SI-NEXT: v_writelane_b32 v4, s7, 1
+; SI-NEXT: v_writelane_b32 v3, s6, 0
+; SI-NEXT: v_writelane_b32 v3, s7, 1
; SI-NEXT: s_mov_b32 s34, s6
; SI-NEXT: s_mov_b32 s7, 0xf000
; SI-NEXT: s_mov_b32 s6, -1
-; SI-NEXT: buffer_load_dword v1, off, s[4:7], 0 offset:16
+; SI-NEXT: buffer_load_dword v0, off, s[4:7], 0 offset:16
; SI-NEXT: s_mov_b64 s[36:37], 0
; SI-NEXT: .LBB56_1: ; %atomicrmw.start
; SI-NEXT: ; =>This Inner Loop Header: Depth=1
; SI-NEXT: s_waitcnt vmcnt(0)
-; SI-NEXT: v_and_b32_e32 v0, s34, v1
-; SI-NEXT: v_not_b32_e32 v0, v0
-; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_mov_b32_e32 v3, v1
; SI-NEXT: v_mov_b32_e32 v2, v0
-; SI-NEXT: buffer_atomic_cmpswap v[2:3], off, s[4:7], 0 offset:16 glc
+; SI-NEXT: s_waitcnt expcnt(0)
+; SI-NEXT: v_and_b32_e32 v0, s34, v2
+; SI-NEXT: v_not_b32_e32 v1, v0
+; SI-NEXT: v_mov_b32_e32 v0, v1
+; SI-NEXT: v_mov_b32_e32 v1, v2
+; SI-NEXT: buffer_atomic_cmpswap v[0:1], off, s[4:7], 0 offset:16 glc
; SI-NEXT: s_waitcnt vmcnt(0)
; SI-NEXT: buffer_wbinvl1
-; SI-NEXT: v_cmp_eq_u32_e32 vcc, v2, v1
+; SI-NEXT: v_cmp_eq_u32_e32 vcc, v0, v2
; SI-NEXT: s_or_b64 s[36:37], vcc, s[36:37]
-; SI-NEXT: v_mov_b32_e32 v1, v2
; SI-NEXT: s_andn2_b64 exec, exec, s[36:37]
; SI-NEXT: s_cbranch_execnz .LBB56_1
; SI-NEXT: ; %bb.2: ; %atomicrmw.end
; SI-NEXT: s_or_b64 exec, exec, s[36:37]
-; SI-NEXT: v_readlane_b32 s7, v4, 1
-; SI-NEXT: v_readlane_b32 s6, v4, 0
+; SI-NEXT: v_readlane_b32 s7, v3, 1
+; SI-NEXT: v_readlane_b32 s6, v3, 0
; SI-NEXT: s_xor_saveexec_b64 s[34:35], -1
-; SI-NEXT: buffer_load_dword v4, off, s[0:3], s32 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v3, off, s[0:3], s32 ; 4-byte Folded Reload
; SI-NEXT: s_mov_b64 exec, s[34:35]
; SI-NEXT: s_waitcnt vmcnt(0) expcnt(0)
; SI-NEXT: s_setpc_b64 s[30:31]
@@ -2862,22 +2862,22 @@ define void @global_atomic_nand_i32_noret_offset__amdgpu_no_remote_memory(ptr ad
; SI-NEXT: s_mov_b32 s7, 0xf000
; SI-NEXT: s_mov_b32 s4, s6
; SI-NEXT: s_mov_b32 s5, s6
-; SI-NEXT: buffer_load_dword v4, v[0:1], s[4:7], 0 addr64 offset:16
+; SI-NEXT: buffer_load_dword v3, v[0:1], s[4:7], 0 addr64 offset:16
; SI-NEXT: s_mov_b64 s[8:9], 0
; SI-NEXT: .LBB59_1: ; %atomicrmw.start
; SI-NEXT: ; =>This Inner Loop Header: Depth=1
; SI-NEXT: s_waitcnt vmcnt(0)
-; SI-NEXT: v_and_b32_e32 v3, v4, v2
-; SI-NEXT: v_not_b32_e32 v3, v3
-; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_mov_b32_e32 v6, v4
; SI-NEXT: v_mov_b32_e32 v5, v3
-; SI-NEXT: buffer_atomic_cmpswap v[5:6], v[0:1], s[4:7], 0 addr64 offset:16 glc
+; SI-NEXT: s_waitcnt expcnt(0)
+; SI-NEXT: v_and_b32_e32 v3, v5, v2
+; SI-NEXT: v_not_b32_e32 v4, v3
+; SI-NEXT: v_mov_b32_e32 v3, v4
+; SI-NEXT: v_mov_b32_e32 v4, v5
+; SI-NEXT: buffer_atomic_cmpswap v[3:4], v[0:1], s[4:7], 0 addr64 offset:16 glc
; SI-NEXT: s_waitcnt vmcnt(0)
; SI-NEXT: buffer_wbinvl1
-; SI-NEXT: v_cmp_eq_u32_e32 vcc, v5, v4
+; SI-NEXT: v_cmp_eq_u32_e32 vcc, v3, v5
; SI-NEXT: s_or_b64 s[8:9], vcc, s[8:9]
-; SI-NEXT: v_mov_b32_e32 v4, v5
; SI-NEXT: s_andn2_b64 exec, exec, s[8:9]
; SI-NEXT: s_cbranch_execnz .LBB59_1
; SI-NEXT: ; %bb.2: ; %atomicrmw.end
@@ -3950,21 +3950,21 @@ define void @global_atomic_max_i32_noret(ptr addrspace(1) %ptr, i32 %in) {
; SI-NEXT: s_mov_b32 s7, 0xf000
; SI-NEXT: s_mov_b32 s4, s6
; SI-NEXT: s_mov_b32 s5, s6
-; SI-NEXT: buffer_load_dword v4, v[0:1], s[4:7], 0 addr64
+; SI-NEXT: buffer_load_dword v3, v[0:1], s[4:7], 0 addr64
; SI-NEXT: s_mov_b64 s[8:9], 0
; SI-NEXT: .LBB83_1: ; %atomicrmw.start
; SI-NEXT: ; =>This Inner Loop Header: Depth=1
; SI-NEXT: s_waitcnt vmcnt(0)
-; SI-NEXT: v_max_i32_e32 v3, v4, v2
-; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_mov_b32_e32 v6, v4
; SI-NEXT: v_mov_b32_e32 v5, v3
-; SI-NEXT: buffer_atomic_cmpswap v[5:6], v[0:1], s[4:7], 0 addr64 glc
+; SI-NEXT: s_waitcnt expcnt(0)
+; SI-NEXT: v_max_i32_e32 v4, v5, v2
+; SI-NEXT: v_mov_b32_e32 v3, v4
+; SI-NEXT: v_mov_b32_e32 v4, v5
+; SI-NEXT: buffer_atomic_cmpswap v[3:4], v[0:1], s[4:7], 0 addr64 glc
; SI-NEXT: s_waitcnt vmcnt(0)
; SI-NEXT: buffer_wbinvl1
-; SI-NEXT: v_cmp_eq_u32_e32 vcc, v5, v4
+; SI-NEXT: v_cmp_eq_u32_e32 vcc, v3, v5
; SI-NEXT: s_or_b64 s[8:9], vcc, s[8:9]
-; SI-NEXT: v_mov_b32_e32 v4, v5
; SI-NEXT: s_andn2_b64 exec, exec, s[8:9]
; SI-NEXT: s_cbranch_execnz .LBB83_1
; SI-NEXT: ; %bb.2: ; %atomicrmw.end
@@ -4025,21 +4025,21 @@ define void @global_atomic_max_i32_noret_offset(ptr addrspace(1) %out, i32 %in)
; SI-NEXT: s_mov_b32 s7, 0xf000
; SI-NEXT: s_mov_b32 s4, s6
; SI-NEXT: s_mov_b32 s5, s6
-; SI-NEXT: buffer_load_dword v4, v[0:1], s[4:7], 0 addr64 offset:16
+; SI-NEXT: buffer_load_dword v3, v[0:1], s[4:7], 0 addr64 offset:16
; SI-NEXT: s_mov_b64 s[8:9], 0
; SI-NEXT: .LBB84_1: ; %atomicrmw.start
; SI-NEXT: ; =>This Inner Loop Header: Depth=1
; SI-NEXT: s_waitcnt vmcnt(0)
-; SI-NEXT: v_max_i32_e32 v3, v4, v2
-; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_mov_b32_e32 v6, v4
; SI-NEXT: v_mov_b32_e32 v5, v3
-; SI-NEXT: buffer_atomic_cmpswap v[5:6], v[0:1], s[4:7], 0 addr64 offset:16 glc
+; SI-NEXT: s_waitcnt expcnt(0)
+; SI-NEXT: v_max_i32_e32 v4, v5, v2
+; SI-NEXT: v_mov_b32_e32 v3, v4
+; SI-NEXT: v_mov_b32_e32 v4, v5
+; SI-NEXT: buffer_atomic_cmpswap v[3:4], v[0:1], s[4:7], 0 addr64 offset:16 glc
; SI-NEXT: s_waitcnt vmcnt(0)
; SI-NEXT: buffer_wbinvl1
-; SI-NEXT: v_cmp_eq_u32_e32 vcc, v5, v4
+; SI-NEXT: v_cmp_eq_u32_e32 vcc, v3, v5
; SI-NEXT: s_or_b64 s[8:9], vcc, s[8:9]
-; SI-NEXT: v_mov_b32_e32 v4, v5
; SI-NEXT: s_andn2_b64 exec, exec, s[8:9]
; SI-NEXT: s_cbranch_execnz .LBB84_1
; SI-NEXT: ; %bb.2: ; %atomicrmw.end
@@ -4258,37 +4258,37 @@ define amdgpu_gfx void @global_atomic_max_i32_noret_scalar(ptr addrspace(1) inre
; SI: ; %bb.0:
; SI-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; SI-NEXT: s_xor_saveexec_b64 s[34:35], -1
-; SI-NEXT: buffer_store_dword v4, off, s[0:3], s32 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v3, off, s[0:3], s32 ; 4-byte Folded Spill
; SI-NEXT: s_mov_b64 exec, s[34:35]
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_writelane_b32 v4, s6, 0
-; SI-NEXT: v_writelane_b32 v4, s7, 1
+; SI-NEXT: v_writelane_b32 v3, s6, 0
+; SI-NEXT: v_writelane_b32 v3, s7, 1
; SI-NEXT: s_mov_b32 s34, s6
; SI-NEXT: s_mov_b32 s7, 0xf000
; SI-NEXT: s_mov_b32 s6, -1
-; SI-NEXT: buffer_load_dword v1, off, s[4:7], 0
+; SI-NEXT: buffer_load_dword v0, off, s[4:7], 0
; SI-NEXT: s_mov_b64 s[36:37], 0
; SI-NEXT: .LBB87_1: ; %atomicrmw.start
; SI-NEXT: ; =>This Inner Loop Header: Depth=1
; SI-NEXT: s_waitcnt vmcnt(0)
-; SI-NEXT: v_max_i32_e32 v0, s34, v1
-; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_mov_b32_e32 v3, v1
; SI-NEXT: v_mov_b32_e32 v2, v0
-; SI-NEXT: buffer_atomic_cmpswap v[2:3], off, s[4:7], 0 glc
+; SI-NEXT: s_waitcnt expcnt(0)
+; SI-NEXT: v_max_i32_e32 v1, s34, v2
+; SI-NEXT: v_mov_b32_e32 v0, v1
+; SI-NEXT: v_mov_b32_e32 v1, v2
+; SI-NEXT: buffer_atomic_cmpswap v[0:1], off, s[4:7], 0 glc
; SI-NEXT: s_waitcnt vmcnt(0)
; SI-NEXT: buffer_wbinvl1
-; SI-NEXT: v_cmp_eq_u32_e32 vcc, v2, v1
+; SI-NEXT: v_cmp_eq_u32_e32 vcc, v0, v2
; SI-NEXT: s_or_b64 s[36:37], vcc, s[36:37]
-; SI-NEXT: v_mov_b32_e32 v1, v2
; SI-NEXT: s_andn2_b64 exec, exec, s[36:37]
; SI-NEXT: s_cbranch_execnz .LBB87_1
; SI-NEXT: ; %bb.2: ; %atomicrmw.end
; SI-NEXT: s_or_b64 exec, exec, s[36:37]
-; SI-NEXT: v_readlane_b32 s7, v4, 1
-; SI-NEXT: v_readlane_b32 s6, v4, 0
+; SI-NEXT: v_readlane_b32 s7, v3, 1
+; SI-NEXT: v_readlane_b32 s6, v3, 0
; SI-NEXT: s_xor_saveexec_b64 s[34:35], -1
-; SI-NEXT: buffer_load_dword v4, off, s[0:3], s32 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v3, off, s[0:3], s32 ; 4-byte Folded Reload
; SI-NEXT: s_mov_b64 exec, s[34:35]
; SI-NEXT: s_waitcnt vmcnt(0) expcnt(0)
; SI-NEXT: s_setpc_b64 s[30:31]
@@ -4346,37 +4346,37 @@ define amdgpu_gfx void @global_atomic_max_i32_noret_offset_scalar(ptr addrspace(
; SI: ; %bb.0:
; SI-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; SI-NEXT: s_xor_saveexec_b64 s[34:35], -1
-; SI-NEXT: buffer_store_dword v4, off, s[0:3], s32 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v3, off, s[0:3], s32 ; 4-byte Folded Spill
; SI-NEXT: s_mov_b64 exec, s[34:35]
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_writelane_b32 v4, s6, 0
-; SI-NEXT: v_writelane_b32 v4, s7, 1
+; SI-NEXT: v_writelane_b32 v3, s6, 0
+; SI-NEXT: v_writelane_b32 v3, s7, 1
; SI-NEXT: s_mov_b32 s34, s6
; SI-NEXT: s_mov_b32 s7, 0xf000
; SI-NEXT: s_mov_b32 s6, -1
-; SI-NEXT: buffer_load_dword v1, off, s[4:7], 0 offset:16
+; SI-NEXT: buffer_load_dword v0, off, s[4:7], 0 offset:16
; SI-NEXT: s_mov_b64 s[36:37], 0
; SI-NEXT: .LBB88_1: ; %atomicrmw.start
; SI-NEXT: ; =>This Inner Loop Header: Depth=1
; SI-NEXT: s_waitcnt vmcnt(0)
-; SI-NEXT: v_max_i32_e32 v0, s34, v1
-; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_mov_b32_e32 v3, v1
; SI-NEXT: v_mov_b32_e32 v2, v0
-; SI-NEXT: buffer_atomic_cmpswap v[2:3], off, s[4:7], 0 offset:16 glc
+; SI-NEXT: s_waitcnt expcnt(0)
+; SI-NEXT: v_max_i32_e32 v1, s34, v2
+; SI-NEXT: v_mov_b32_e32 v0, v1
+; SI-NEXT: v_mov_b32_e32 v1, v2
+; SI-NEXT: buffer_atomic_cmpswap v[0:1], off, s[4:7], 0 offset:16 glc
; SI-NEXT: s_waitcnt vmcnt(0)
; SI-NEXT: buffer_wbinvl1
-; SI-NEXT: v_cmp_eq_u32_e32 vcc, v2, v1
+; SI-NEXT: v_cmp_eq_u32_e32 vcc, v0, v2
; SI-NEXT: s_or_b64 s[36:37], vcc, s[36:37]
-; SI-NEXT: v_mov_b32_e32 v1, v2
; SI-NEXT: s_andn2_b64 exec, exec, s[36:37]
; SI-NEXT: s_cbranch_execnz .LBB88_1
; SI-NEXT: ; %bb.2: ; %atomicrmw.end
; SI-NEXT: s_or_b64 exec, exec, s[36:37]
-; SI-NEXT: v_readlane_b32 s7, v4, 1
-; SI-NEXT: v_readlane_b32 s6, v4, 0
+; SI-NEXT: v_readlane_b32 s7, v3, 1
+; SI-NEXT: v_readlane_b32 s6, v3, 0
; SI-NEXT: s_xor_saveexec_b64 s[34:35], -1
-; SI-NEXT: buffer_load_dword v4, off, s[0:3], s32 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v3, off, s[0:3], s32 ; 4-byte Folded Reload
; SI-NEXT: s_mov_b64 exec, s[34:35]
; SI-NEXT: s_waitcnt vmcnt(0) expcnt(0)
; SI-NEXT: s_setpc_b64 s[30:31]
@@ -4627,20 +4627,20 @@ define amdgpu_kernel void @atomic_max_i32_addr64_offset(ptr addrspace(1) %out, i
; SI-NEXT: s_mov_b64 s[0:1], 0
; SI-NEXT: s_mov_b32 s7, 0xf000
; SI-NEXT: s_waitcnt lgkmcnt(0)
-; SI-NEXT: v_mov_b32_e32 v1, s3
+; SI-NEXT: v_mov_b32_e32 v0, s3
; SI-NEXT: s_mov_b32 s6, -1
; SI-NEXT: .LBB91_1: ; %atomicrmw.start
; SI-NEXT: ; =>This Inner Loop Header: Depth=1
-; SI-NEXT: v_max_i32_e32 v0, s2, v1
-; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_mov_b32_e32 v3, v1
; SI-NEXT: v_mov_b32_e32 v2, v0
-; SI-NEXT: buffer_atomic_cmpswap v[2:3], off, s[4:7], 0 offset:16 glc
+; SI-NEXT: s_waitcnt expcnt(0)
+; SI-NEXT: v_max_i32_e32 v1, s2, v2
+; SI-NEXT: v_mov_b32_e32 v0, v1
+; SI-NEXT: v_mov_b32_e32 v1, v2
+; SI-NEXT: buffer_atomic_cmpswap v[0:1], off, s[4:7], 0 offset:16 glc
; SI-NEXT: s_waitcnt vmcnt(0)
; SI-NEXT: buffer_wbinvl1
-; SI-NEXT: v_cmp_eq_u32_e32 vcc, v2, v1
+; SI-NEXT: v_cmp_eq_u32_e32 vcc, v0, v2
; SI-NEXT: s_or_b64 s[0:1], vcc, s[0:1]
-; SI-NEXT: v_mov_b32_e32 v1, v2
; SI-NEXT: s_andn2_b64 exec, exec, s[0:1]
; SI-NEXT: s_cbranch_execnz .LBB91_1
; SI-NEXT: ; %bb.2: ; %atomicrmw.end
@@ -4726,20 +4726,20 @@ define amdgpu_kernel void @atomic_max_i32_ret_addr64_offset(ptr addrspace(1) %ou
; SI-NEXT: s_mov_b64 s[0:1], 0
; SI-NEXT: s_mov_b32 s7, 0xf000
; SI-NEXT: s_waitcnt lgkmcnt(0)
-; SI-NEXT: v_mov_b32_e32 v1, s6
+; SI-NEXT: v_mov_b32_e32 v0, s6
; SI-NEXT: s_mov_b32 s6, -1
; SI-NEXT: .LBB92_1: ; %atomicrmw.start
; SI-NEXT: ; =>This Inner Loop Header: Depth=1
-; SI-NEXT: v_max_i32_e32 v0, s8, v1
-; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_mov_b32_e32 v3, v1
; SI-NEXT: v_mov_b32_e32 v2, v0
-; SI-NEXT: buffer_atomic_cmpswap v[2:3], off, s[4:7], 0 offset:16 glc
+; SI-NEXT: s_waitcnt expcnt(0)
+; SI-NEXT: v_max_i32_e32 v1, s8, v2
+; SI-NEXT: v_mov_b32_e32 v0, v1
+; SI-NEXT: v_mov_b32_e32 v1, v2
+; SI-NEXT: buffer_atomic_cmpswap v[0:1], off, s[4:7], 0 offset:16 glc
; SI-NEXT: s_waitcnt vmcnt(0)
; SI-NEXT: buffer_wbinvl1
-; SI-NEXT: v_cmp_eq_u32_e32 vcc, v2, v1
+; SI-NEXT: v_cmp_eq_u32_e32 vcc, v0, v2
; SI-NEXT: s_or_b64 s[0:1], vcc, s[0:1]
-; SI-NEXT: v_mov_b32_e32 v1, v2
; SI-NEXT: s_andn2_b64 exec, exec, s[0:1]
; SI-NEXT: s_cbranch_execnz .LBB92_1
; SI-NEXT: ; %bb.2: ; %atomicrmw.end
@@ -4748,7 +4748,7 @@ define amdgpu_kernel void @atomic_max_i32_ret_addr64_offset(ptr addrspace(1) %ou
; SI-NEXT: s_mov_b32 s6, -1
; SI-NEXT: s_mov_b32 s4, s2
; SI-NEXT: s_mov_b32 s5, s3
-; SI-NEXT: buffer_store_dword v2, off, s[4:7], 0
+; SI-NEXT: buffer_store_dword v0, off, s[4:7], 0
; SI-NEXT: s_endpgm
;
; VI-LABEL: atomic_max_i32_ret_addr64_offset:
@@ -4840,20 +4840,20 @@ define amdgpu_kernel void @atomic_max_i32_addr64(ptr addrspace(1) %out, i32 %in,
; SI-NEXT: s_mov_b64 s[0:1], 0
; SI-NEXT: s_mov_b32 s7, 0xf000
; SI-NEXT: s_waitcnt lgkmcnt(0)
-; SI-NEXT: v_mov_b32_e32 v1, s3
+; SI-NEXT: v_mov_b32_e32 v0, s3
; SI-NEXT: s_mov_b32 s6, -1
; SI-NEXT: .LBB93_1: ; %atomicrmw.start
; SI-NEXT: ; =>This Inner Loop Header: Depth=1
-; SI-NEXT: v_max_i32_e32 v0, s2, v1
-; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_mov_b32_e32 v3, v1
; SI-NEXT: v_mov_b32_e32 v2, v0
-; SI-NEXT: buffer_atomic_cmpswap v[2:3], off, s[4:7], 0 glc
+; SI-NEXT: s_waitcnt expcnt(0)
+; SI-NEXT: v_max_i32_e32 v1, s2, v2
+; SI-NEXT: v_mov_b32_e32 v0, v1
+; SI-NEXT: v_mov_b32_e32 v1, v2
+; SI-NEXT: buffer_atomic_cmpswap v[0:1], off, s[4:7], 0 glc
; SI-NEXT: s_waitcnt vmcnt(0)
; SI-NEXT: buffer_wbinvl1
-; SI-NEXT: v_cmp_eq_u32_e32 vcc, v2, v1
+; SI-NEXT: v_cmp_eq_u32_e32 vcc, v0, v2
; SI-NEXT: s_or_b64 s[0:1], vcc, s[0:1]
-; SI-NEXT: v_mov_b32_e32 v1, v2
; SI-NEXT: s_andn2_b64 exec, exec, s[0:1]
; SI-NEXT: s_cbranch_execnz .LBB93_1
; SI-NEXT: ; %bb.2: ; %atomicrmw.end
@@ -4936,20 +4936,20 @@ define amdgpu_kernel void @atomic_max_i32_ret_addr64(ptr addrspace(1) %out, ptr
; SI-NEXT: s_mov_b64 s[0:1], 0
; SI-NEXT: s_mov_b32 s7, 0xf000
; SI-NEXT: s_waitcnt lgkmcnt(0)
-; SI-NEXT: v_mov_b32_e32 v1, s6
+; SI-NEXT: v_mov_b32_e32 v0, s6
; SI-NEXT: s_mov_b32 s6, -1
; SI-NEXT: .LBB94_1: ; %atomicrmw.start
; SI-NEXT: ; =>This Inner Loop Header: Depth=1
-; SI-NEXT: v_max_i32_e32 v0, s8, v1
-; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_mov_b32_e32 v3, v1
; SI-NEXT: v_mov_b32_e32 v2, v0
-; SI-NEXT: buffer_atomic_cmpswap v[2:3], off, s[4:7], 0 glc
+; SI-NEXT: s_waitcnt expcnt(0)
+; SI-NEXT: v_max_i32_e32 v1, s8, v2
+; SI-NEXT: v_mov_b32_e32 v0, v1
+; SI-NEXT: v_mov_b32_e32 v1, v2
+; SI-NEXT: buffer_atomic_cmpswap v[0:1], off, s[4:7], 0 glc
; SI-NEXT: s_waitcnt vmcnt(0)
; SI-NEXT: buffer_wbinvl1
-; SI-NEXT: v_cmp_eq_u32_e32 vcc, v2, v1
+; SI-NEXT: v_cmp_eq_u32_e32 vcc, v0, v2
; SI-NEXT: s_or_b64 s[0:1], vcc, s[0:1]
-; SI-NEXT: v_mov_b32_e32 v1, v2
; SI-NEXT: s_andn2_b64 exec, exec, s[0:1]
; SI-NEXT: s_cbranch_execnz .LBB94_1
; SI-NEXT: ; %bb.2: ; %atomicrmw.end
@@ -4958,7 +4958,7 @@ define amdgpu_kernel void @atomic_max_i32_ret_addr64(ptr addrspace(1) %out, ptr
; SI-NEXT: s_mov_b32 s6, -1
; SI-NEXT: s_mov_b32 s4, s2
; SI-NEXT: s_mov_b32 s5, s3
-; SI-NEXT: buffer_store_dword v2, off, s[4:7], 0
+; SI-NEXT: buffer_store_dword v0, off, s[4:7], 0
; SI-NEXT: s_endpgm
;
; VI-LABEL: atomic_max_i32_ret_addr64:
@@ -5041,21 +5041,21 @@ define void @global_atomic_max_i32_noret_offset__amdgpu_no_remote_memory(ptr add
; SI-NEXT: s_mov_b32 s7, 0xf000
; SI-NEXT: s_mov_b32 s4, s6
; SI-NEXT: s_mov_b32 s5, s6
-; SI-NEXT: buffer_load_dword v4, v[0:1], s[4:7], 0 addr64 offset:16
+; SI-NEXT: buffer_load_dword v3, v[0:1], s[4:7], 0 addr64 offset:16
; SI-NEXT: s_mov_b64 s[8:9], 0
; SI-NEXT: .LBB95_1: ; %atomicrmw.start
; SI-NEXT: ; =>This Inner Loop Header: Depth=1
; SI-NEXT: s_waitcnt vmcnt(0)
-; SI-NEXT: v_max_i32_e32 v3, v4, v2
-; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_mov_b32_e32 v6, v4
; SI-NEXT: v_mov_b32_e32 v5, v3
-; SI-NEXT: buffer_atomic_cmpswap v[5:6], v[0:1], s[4:7], 0 addr64 offset:16 glc
+; SI-NEXT: s_waitcnt expcnt(0)
+; SI-NEXT: v_max_i32_e32 v4, v5, v2
+; SI-NEXT: v_mov_b32_e32 v3, v4
+; SI-NEXT: v_mov_b32_e32 v4, v5
+; SI-NEXT: buffer_atomic_cmpswap v[3:4], v[0:1], s[4:7], 0 addr64 offset:16 glc
; SI-NEXT: s_waitcnt vmcnt(0)
; SI-NEXT: buffer_wbinvl1
-; SI-NEXT: v_cmp_eq_u32_e32 vcc, v5, v4
+; SI-NEXT: v_cmp_eq_u32_e32 vcc, v3, v5
; SI-NEXT: s_or_b64 s[8:9], vcc, s[8:9]
-; SI-NEXT: v_mov_b32_e32 v4, v5
; SI-NEXT: s_andn2_b64 exec, exec, s[8:9]
; SI-NEXT: s_cbranch_execnz .LBB95_1
; SI-NEXT: ; %bb.2: ; %atomicrmw.end
@@ -5203,21 +5203,21 @@ define void @global_atomic_umax_i32_noret(ptr addrspace(1) %ptr, i32 %in) {
; SI-NEXT: s_mov_b32 s7, 0xf000
; SI-NEXT: s_mov_b32 s4, s6
; SI-NEXT: s_mov_b32 s5, s6
-; SI-NEXT: buffer_load_dword v4, v[0:1], s[4:7], 0 addr64
+; SI-NEXT: buffer_load_dword v3, v[0:1], s[4:7], 0 addr64
; SI-NEXT: s_mov_b64 s[8:9], 0
; SI-NEXT: .LBB97_1: ; %atomicrmw.start
; SI-NEXT: ; =>This Inner Loop Header: Depth=1
; SI-NEXT: s_waitcnt vmcnt(0)
-; SI-NEXT: v_max_u32_e32 v3, v4, v2
-; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_mov_b32_e32 v6, v4
; SI-NEXT: v_mov_b32_e32 v5, v3
-; SI-NEXT: buffer_atomic_cmpswap v[5:6], v[0:1], s[4:7], 0 addr64 glc
+; SI-NEXT: s_waitcnt expcnt(0)
+; SI-NEXT: v_max_u32_e32 v4, v5, v2
+; SI-NEXT: v_mov_b32_e32 v3, v4
+; SI-NEXT: v_mov_b32_e32 v4, v5
+; SI-NEXT: buffer_atomic_cmpswap v[3:4], v[0:1], s[4:7], 0 addr64 glc
; SI-NEXT: s_waitcnt vmcnt(0)
; SI-NEXT: buffer_wbinvl1
-; SI-NEXT: v_cmp_eq_u32_e32 vcc, v5, v4
+; SI-NEXT: v_cmp_eq_u32_e32 vcc, v3, v5
; SI-NEXT: s_or_b64 s[8:9], vcc, s[8:9]
-; SI-NEXT: v_mov_b32_e32 v4, v5
; SI-NEXT: s_andn2_b64 exec, exec, s[8:9]
; SI-NEXT: s_cbranch_execnz .LBB97_1
; SI-NEXT: ; %bb.2: ; %atomicrmw.end
@@ -5278,21 +5278,21 @@ define void @global_atomic_umax_i32_noret_offset(ptr addrspace(1) %out, i32 %in)
; SI-NEXT: s_mov_b32 s7, 0xf000
; SI-NEXT: s_mov_b32 s4, s6
; SI-NEXT: s_mov_b32 s5, s6
-; SI-NEXT: buffer_load_dword v4, v[0:1], s[4:7], 0 addr64 offset:16
+; SI-NEXT: buffer_load_dword v3, v[0:1], s[4:7], 0 addr64 offset:16
; SI-NEXT: s_mov_b64 s[8:9], 0
; SI-NEXT: .LBB98_1: ; %atomicrmw.start
; SI-NEXT: ; =>This Inner Loop Header: Depth=1
; SI-NEXT: s_waitcnt vmcnt(0)
-; SI-NEXT: v_max_u32_e32 v3, v4, v2
-; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_mov_b32_e32 v6, v4
; SI-NEXT: v_mov_b32_e32 v5, v3
-; SI-NEXT: buffer_atomic_cmpswap v[5:6], v[0:1], s[4:7], 0 addr64 offset:16 glc
+; SI-NEXT: s_waitcnt expcnt(0)
+; SI-NEXT: v_max_u32_e32 v4, v5, v2
+; SI-NEXT: v_mov_b32_e32 v3, v4
+; SI-NEXT: v_mov_b32_e32 v4, v5
+; SI-NEXT: buffer_atomic_cmpswap v[3:4], v[0:1], s[4:7], 0 addr64 offset:16 glc
; SI-NEXT: s_waitcnt vmcnt(0)
; SI-NEXT: buffer_wbinvl1
-; SI-NEXT: v_cmp_eq_u32_e32 vcc, v5, v4
+; SI-NEXT: v_cmp_eq_u32_e32 vcc, v3, v5
; SI-NEXT: s_or_b64 s[8:9], vcc, s[8:9]
-; SI-NEXT: v_mov_b32_e32 v4, v5
; SI-NEXT: s_andn2_b64 exec, exec, s[8:9]
; SI-NEXT: s_cbranch_execnz .LBB98_1
; SI-NEXT: ; %bb.2: ; %atomicrmw.end
@@ -5511,37 +5511,37 @@ define amdgpu_gfx void @global_atomic_umax_i32_noret_scalar(ptr addrspace(1) inr
; SI: ; %bb.0:
; SI-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; SI-NEXT: s_xor_saveexec_b64 s[34:35], -1
-; SI-NEXT: buffer_store_dword v4, off, s[0:3], s32 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v3, off, s[0:3], s32 ; 4-byte Folded Spill
; SI-NEXT: s_mov_b64 exec, s[34:35]
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_writelane_b32 v4, s6, 0
-; SI-NEXT: v_writelane_b32 v4, s7, 1
+; SI-NEXT: v_writelane_b32 v3, s6, 0
+; SI-NEXT: v_writelane_b32 v3, s7, 1
; SI-NEXT: s_mov_b32 s34, s6
; SI-NEXT: s_mov_b32 s7, 0xf000
; SI-NEXT: s_mov_b32 s6, -1
-; SI-NEXT: buffer_load_dword v1, off, s[4:7], 0
+; SI-NEXT: buffer_load_dword v0, off, s[4:7], 0
; SI-NEXT: s_mov_b64 s[36:37], 0
; SI-NEXT: .LBB101_1: ; %atomicrmw.start
; SI-NEXT: ; =>This Inner Loop Header: Depth=1
; SI-NEXT: s_waitcnt vmcnt(0)
-; SI-NEXT: v_max_u32_e32 v0, s34, v1
-; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_mov_b32_e32 v3, v1
; SI-NEXT: v_mov_b32_e32 v2, v0
-; SI-NEXT: buffer_atomic_cmpswap v[2:3], off, s[4:7], 0 glc
+; SI-NEXT: s_waitcnt expcnt(0)
+; SI-NEXT: v_max_u32_e32 v1, s34, v2
+; SI-NEXT: v_mov_b32_e32 v0, v1
+; SI-NEXT: v_mov_b32_e32 v1, v2
+; SI-NEXT: buffer_atomic_cmpswap v[0:1], off, s[4:7], 0 glc
; SI-NEXT: s_waitcnt vmcnt(0)
; SI-NEXT: buffer_wbinvl1
-; SI-NEXT: v_cmp_eq_u32_e32 vcc, v2, v1
+; SI-NEXT: v_cmp_eq_u32_e32 vcc, v0, v2
; SI-NEXT: s_or_b64 s[36:37], vcc, s[36:37]
-; SI-NEXT: v_mov_b32_e32 v1, v2
; SI-NEXT: s_andn2_b64 exec, exec, s[36:37]
; SI-NEXT: s_cbranch_execnz .LBB101_1
; SI-NEXT: ; %bb.2: ; %atomicrmw.end
; SI-NEXT: s_or_b64 exec, exec, s[36:37]
-; SI-NEXT: v_readlane_b32 s7, v4, 1
-; SI-NEXT: v_readlane_b32 s6, v4, 0
+; SI-NEXT: v_readlane_b32 s7, v3, 1
+; SI-NEXT: v_readlane_b32 s6, v3, 0
; SI-NEXT: s_xor_saveexec_b64 s[34:35], -1
-; SI-NEXT: buffer_load_dword v4, off, s[0:3], s32 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v3, off, s[0:3], s32 ; 4-byte Folded Reload
; SI-NEXT: s_mov_b64 exec, s[34:35]
; SI-NEXT: s_waitcnt vmcnt(0) expcnt(0)
; SI-NEXT: s_setpc_b64 s[30:31]
@@ -5599,37 +5599,37 @@ define amdgpu_gfx void @global_atomic_umax_i32_noret_offset_scalar(ptr addrspace
; SI: ; %bb.0:
; SI-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; SI-NEXT: s_xor_saveexec_b64 s[34:35], -1
-; SI-NEXT: buffer_store_dword v4, off, s[0:3], s32 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v3, off, s[0:3], s32 ; 4-byte Folded Spill
; SI-NEXT: s_mov_b64 exec, s[34:35]
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_writelane_b32 v4, s6, 0
-; SI-NEXT: v_writelane_b32 v4, s7, 1
+; SI-NEXT: v_writelane_b32 v3, s6, 0
+; SI-NEXT: v_writelane_b32 v3, s7, 1
; SI-NEXT: s_mov_b32 s34, s6
; SI-NEXT: s_mov_b32 s7, 0xf000
; SI-NEXT: s_mov_b32 s6, -1
-; SI-NEXT: buffer_load_dword v1, off, s[4:7], 0 offset:16
+; SI-NEXT: buffer_load_dword v0, off, s[4:7], 0 offset:16
; SI-NEXT: s_mov_b64 s[36:37], 0
; SI-NEXT: .LBB102_1: ; %atomicrmw.start
; SI-NEXT: ; =>This Inner Loop Header: Depth=1
; SI-NEXT: s_waitcnt vmcnt(0)
-; SI-NEXT: v_max_u32_e32 v0, s34, v1
-; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_mov_b32_e32 v3, v1
; SI-NEXT: v_mov_b32_e32 v2, v0
-; SI-NEXT: buffer_atomic_cmpswap v[2:3], off, s[4:7], 0 offset:16 glc
+; SI-NEXT: s_waitcnt expcnt(0)
+; SI-NEXT: v_max_u32_e32 v1, s34, v2
+; SI-NEXT: v_mov_b32_e32 v0, v1
+; SI-NEXT: v_mov_b32_e32 v1, v2
+; SI-NEXT: buffer_atomic_cmpswap v[0:1], off, s[4:7], 0 offset:16 glc
; SI-NEXT: s_waitcnt vmcnt(0)
; SI-NEXT: buffer_wbinvl1
-; SI-NEXT: v_cmp_eq_u32_e32 vcc, v2, v1
+; SI-NEXT: v_cmp_eq_u32_e32 vcc, v0, v2
; SI-NEXT: s_or_b64 s[36:37], vcc, s[36:37]
-; SI-NEXT: v_mov_b32_e32 v1, v2
; SI-NEXT: s_andn2_b64 exec, exec, s[36:37]
; SI-NEXT: s_cbranch_execnz .LBB102_1
; SI-NEXT: ; %bb.2: ; %atomicrmw.end
; SI-NEXT: s_or_b64 exec, exec, s[36:37]
-; SI-NEXT: v_readlane_b32 s7, v4, 1
-; SI-NEXT: v_readlane_b32 s6, v4, 0
+; SI-NEXT: v_readlane_b32 s7, v3, 1
+; SI-NEXT: v_readlane_b32 s6, v3, 0
; SI-NEXT: s_xor_saveexec_b64 s[34:35], -1
-; SI-NEXT: buffer_load_dword v4, off, s[0:3], s32 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v3, off, s[0:3], s32 ; 4-byte Folded Reload
; SI-NEXT: s_mov_b64 exec, s[34:35]
; SI-NEXT: s_waitcnt vmcnt(0) expcnt(0)
; SI-NEXT: s_setpc_b64 s[30:31]
@@ -5880,20 +5880,20 @@ define amdgpu_kernel void @atomic_umax_i32_addr64_offset(ptr addrspace(1) %out,
; SI-NEXT: s_mov_b64 s[0:1], 0
; SI-NEXT: s_mov_b32 s7, 0xf000
; SI-NEXT: s_waitcnt lgkmcnt(0)
-; SI-NEXT: v_mov_b32_e32 v1, s3
+; SI-NEXT: v_mov_b32_e32 v0, s3
; SI-NEXT: s_mov_b32 s6, -1
; SI-NEXT: .LBB105_1: ; %atomicrmw.start
; SI-NEXT: ; =>This Inner Loop Header: Depth=1
-; SI-NEXT: v_max_u32_e32 v0, s2, v1
-; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_mov_b32_e32 v3, v1
; SI-NEXT: v_mov_b32_e32 v2, v0
-; SI-NEXT: buffer_atomic_cmpswap v[2:3], off, s[4:7], 0 offset:16 glc
+; SI-NEXT: s_waitcnt expcnt(0)
+; SI-NEXT: v_max_u32_e32 v1, s2, v2
+; SI-NEXT: v_mov_b32_e32 v0, v1
+; SI-NEXT: v_mov_b32_e32 v1, v2
+; SI-NEXT: buffer_atomic_cmpswap v[0:1], off, s[4:7], 0 offset:16 glc
; SI-NEXT: s_waitcnt vmcnt(0)
; SI-NEXT: buffer_wbinvl1
-; SI-NEXT: v_cmp_eq_u32_e32 vcc, v2, v1
+; SI-NEXT: v_cmp_eq_u32_e32 vcc, v0, v2
; SI-NEXT: s_or_b64 s[0:1], vcc, s[0:1]
-; SI-NEXT: v_mov_b32_e32 v1, v2
; SI-NEXT: s_andn2_b64 exec, exec, s[0:1]
; SI-NEXT: s_cbranch_execnz .LBB105_1
; SI-NEXT: ; %bb.2: ; %atomicrmw.end
@@ -5979,20 +5979,20 @@ define amdgpu_kernel void @atomic_umax_i32_ret_addr64_offset(ptr addrspace(1) %o
; SI-NEXT: s_mov_b64 s[0:1], 0
; SI-NEXT: s_mov_b32 s7, 0xf000
; SI-NEXT: s_waitcnt lgkmcnt(0)
-; SI-NEXT: v_mov_b32_e32 v1, s6
+; SI-NEXT: v_mov_b32_e32 v0, s6
; SI-NEXT: s_mov_b32 s6, -1
; SI-NEXT: .LBB106_1: ; %atomicrmw.start
; SI-NEXT: ; =>This Inner Loop Header: Depth=1
-; SI-NEXT: v_max_u32_e32 v0, s8, v1
-; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_mov_b32_e32 v3, v1
; SI-NEXT: v_mov_b32_e32 v2, v0
-; SI-NEXT: buffer_atomic_cmpswap v[2:3], off, s[4:7], 0 offset:16 glc
+; SI-NEXT: s_waitcnt expcnt(0)
+; SI-NEXT: v_max_u32_e32 v1, s8, v2
+; SI-NEXT: v_mov_b32_e32 v0, v1
+; SI-NEXT: v_mov_b32_e32 v1, v2
+; SI-NEXT: buffer_atomic_cmpswap v[0:1], off, s[4:7], 0 offset:16 glc
; SI-NEXT: s_waitcnt vmcnt(0)
; SI-NEXT: buffer_wbinvl1
-; SI-NEXT: v_cmp_eq_u32_e32 vcc, v2, v1
+; SI-NEXT: v_cmp_eq_u32_e32 vcc, v0, v2
; SI-NEXT: s_or_b64 s[0:1], vcc, s[0:1]
-; SI-NEXT: v_mov_b32_e32 v1, v2
; SI-NEXT: s_andn2_b64 exec, exec, s[0:1]
; SI-NEXT: s_cbranch_execnz .LBB106_1
; SI-NEXT: ; %bb.2: ; %atomicrmw.end
@@ -6001,7 +6001,7 @@ define amdgpu_kernel void @atomic_umax_i32_ret_addr64_offset(ptr addrspace(1) %o
; SI-NEXT: s_mov_b32 s6, -1
; SI-NEXT: s_mov_b32 s4, s2
; SI-NEXT: s_mov_b32 s5, s3
-; SI-NEXT: buffer_store_dword v2, off, s[4:7], 0
+; SI-NEXT: buffer_store_dword v0, off, s[4:7], 0
; SI-NEXT: s_endpgm
;
; VI-LABEL: atomic_umax_i32_ret_addr64_offset:
@@ -6094,20 +6094,20 @@ define amdgpu_kernel void @atomic_umax_i32_ret_addr64(ptr addrspace(1) %out, ptr
; SI-NEXT: s_mov_b64 s[0:1], 0
; SI-NEXT: s_mov_b32 s7, 0xf000
; SI-NEXT: s_waitcnt lgkmcnt(0)
-; SI-NEXT: v_mov_b32_e32 v1, s6
+; SI-NEXT: v_mov_b32_e32 v0, s6
; SI-NEXT: s_mov_b32 s6, -1
; SI-NEXT: .LBB107_1: ; %atomicrmw.start
; SI-NEXT: ; =>This Inner Loop Header: Depth=1
-; SI-NEXT: v_max_u32_e32 v0, s8, v1
-; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_mov_b32_e32 v3, v1
; SI-NEXT: v_mov_b32_e32 v2, v0
-; SI-NEXT: buffer_atomic_cmpswap v[2:3], off, s[4:7], 0 glc
+; SI-NEXT: s_waitcnt expcnt(0)
+; SI-NEXT: v_max_u32_e32 v1, s8, v2
+; SI-NEXT: v_mov_b32_e32 v0, v1
+; SI-NEXT: v_mov_b32_e32 v1, v2
+; SI-NEXT: buffer_atomic_cmpswap v[0:1], off, s[4:7], 0 glc
; SI-NEXT: s_waitcnt vmcnt(0)
; SI-NEXT: buffer_wbinvl1
-; SI-NEXT: v_cmp_eq_u32_e32 vcc, v2, v1
+; SI-NEXT: v_cmp_eq_u32_e32 vcc, v0, v2
; SI-NEXT: s_or_b64 s[0:1], vcc, s[0:1]
-; SI-NEXT: v_mov_b32_e32 v1, v2
; SI-NEXT: s_andn2_b64 exec, exec, s[0:1]
; SI-NEXT: s_cbranch_execnz .LBB107_1
; SI-NEXT: ; %bb.2: ; %atomicrmw.end
@@ -6116,7 +6116,7 @@ define amdgpu_kernel void @atomic_umax_i32_ret_addr64(ptr addrspace(1) %out, ptr
; SI-NEXT: s_mov_b32 s6, -1
; SI-NEXT: s_mov_b32 s4, s2
; SI-NEXT: s_mov_b32 s5, s3
-; SI-NEXT: buffer_store_dword v2, off, s[4:7], 0
+; SI-NEXT: buffer_store_dword v0, off, s[4:7], 0
; SI-NEXT: s_endpgm
;
; VI-LABEL: atomic_umax_i32_ret_addr64:
@@ -6199,21 +6199,21 @@ define void @global_atomic_umax_i32_noret_offset__amdgpu_no_remote_memory(ptr ad
; SI-NEXT: s_mov_b32 s7, 0xf000
; SI-NEXT: s_mov_b32 s4, s6
; SI-NEXT: s_mov_b32 s5, s6
-; SI-NEXT: buffer_load_dword v4, v[0:1], s[4:7], 0 addr64 offset:16
+; SI-NEXT: buffer_load_dword v3, v[0:1], s[4:7], 0 addr64 offset:16
; SI-NEXT: s_mov_b64 s[8:9], 0
; SI-NEXT: .LBB108_1: ; %atomicrmw.start
; SI-NEXT: ; =>This Inner Loop Header: Depth=1
; SI-NEXT: s_waitcnt vmcnt(0)
-; SI-NEXT: v_max_u32_e32 v3, v4, v2
-; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_mov_b32_e32 v6, v4
; SI-NEXT: v_mov_b32_e32 v5, v3
-; SI-NEXT: buffer_atomic_cmpswap v[5:6], v[0:1], s[4:7], 0 addr64 offset:16 glc
+; SI-NEXT: s_waitcnt expcnt(0)
+; SI-NEXT: v_max_u32_e32 v4, v5, v2
+; SI-NEXT: v_mov_b32_e32 v3, v4
+; SI-NEXT: v_mov_b32_e32 v4, v5
+; SI-NEXT: buffer_atomic_cmpswap v[3:4], v[0:1], s[4:7], 0 addr64 offset:16 glc
; SI-NEXT: s_waitcnt vmcnt(0)
; SI-NEXT: buffer_wbinvl1
-; SI-NEXT: v_cmp_eq_u32_e32 vcc, v5, v4
+; SI-NEXT: v_cmp_eq_u32_e32 vcc, v3, v5
; SI-NEXT: s_or_b64 s[8:9], vcc, s[8:9]
-; SI-NEXT: v_mov_b32_e32 v4, v5
; SI-NEXT: s_andn2_b64 exec, exec, s[8:9]
; SI-NEXT: s_cbranch_execnz .LBB108_1
; SI-NEXT: ; %bb.2: ; %atomicrmw.end
@@ -6361,21 +6361,21 @@ define void @global_atomic_umin_i32_noret(ptr addrspace(1) %ptr, i32 %in) {
; SI-NEXT: s_mov_b32 s7, 0xf000
; SI-NEXT: s_mov_b32 s4, s6
; SI-NEXT: s_mov_b32 s5, s6
-; SI-NEXT: buffer_load_dword v4, v[0:1], s[4:7], 0 addr64
+; SI-NEXT: buffer_load_dword v3, v[0:1], s[4:7], 0 addr64
; SI-NEXT: s_mov_b64 s[8:9], 0
; SI-NEXT: .LBB110_1: ; %atomicrmw.start
; SI-NEXT: ; =>This Inner Loop Header: Depth=1
; SI-NEXT: s_waitcnt vmcnt(0)
-; SI-NEXT: v_min_u32_e32 v3, v4, v2
-; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_mov_b32_e32 v6, v4
; SI-NEXT: v_mov_b32_e32 v5, v3
-; SI-NEXT: buffer_atomic_cmpswap v[5:6], v[0:1], s[4:7], 0 addr64 glc
+; SI-NEXT: s_waitcnt expcnt(0)
+; SI-NEXT: v_min_u32_e32 v4, v5, v2
+; SI-NEXT: v_mov_b32_e32 v3, v4
+; SI-NEXT: v_mov_b32_e32 v4, v5
+; SI-NEXT: buffer_atomic_cmpswap v[3:4], v[0:1], s[4:7], 0 addr64 glc
; SI-NEXT: s_waitcnt vmcnt(0)
; SI-NEXT: buffer_wbinvl1
-; SI-NEXT: v_cmp_eq_u32_e32 vcc, v5, v4
+; SI-NEXT: v_cmp_eq_u32_e32 vcc, v3, v5
; SI-NEXT: s_or_b64 s[8:9], vcc, s[8:9]
-; SI-NEXT: v_mov_b32_e32 v4, v5
; SI-NEXT: s_andn2_b64 exec, exec, s[8:9]
; SI-NEXT: s_cbranch_execnz .LBB110_1
; SI-NEXT: ; %bb.2: ; %atomicrmw.end
@@ -6436,21 +6436,21 @@ define void @global_atomic_umin_i32_noret_offset(ptr addrspace(1) %out, i32 %in)
; SI-NEXT: s_mov_b32 s7, 0xf000
; SI-NEXT: s_mov_b32 s4, s6
; SI-NEXT: s_mov_b32 s5, s6
-; SI-NEXT: buffer_load_dword v4, v[0:1], s[4:7], 0 addr64 offset:16
+; SI-NEXT: buffer_load_dword v3, v[0:1], s[4:7], 0 addr64 offset:16
; SI-NEXT: s_mov_b64 s[8:9], 0
; SI-NEXT: .LBB111_1: ; %atomicrmw.start
; SI-NEXT: ; =>This Inner Loop Header: Depth=1
; SI-NEXT: s_waitcnt vmcnt(0)
-; SI-NEXT: v_min_u32_e32 v3, v4, v2
-; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_mov_b32_e32 v6, v4
; SI-NEXT: v_mov_b32_e32 v5, v3
-; SI-NEXT: buffer_atomic_cmpswap v[5:6], v[0:1], s[4:7], 0 addr64 offset:16 glc
+; SI-NEXT: s_waitcnt expcnt(0)
+; SI-NEXT: v_min_u32_e32 v4, v5, v2
+; SI-NEXT: v_mov_b32_e32 v3, v4
+; SI-NEXT: v_mov_b32_e32 v4, v5
+; SI-NEXT: buffer_atomic_cmpswap v[3:4], v[0:1], s[4:7], 0 addr64 offset:16 glc
; SI-NEXT: s_waitcnt vmcnt(0)
; SI-NEXT: buffer_wbinvl1
-; SI-NEXT: v_cmp_eq_u32_e32 vcc, v5, v4
+; SI-NEXT: v_cmp_eq_u32_e32 vcc, v3, v5
; SI-NEXT: s_or_b64 s[8:9], vcc, s[8:9]
-; SI-NEXT: v_mov_b32_e32 v4, v5
; SI-NEXT: s_andn2_b64 exec, exec, s[8:9]
; SI-NEXT: s_cbranch_execnz .LBB111_1
; SI-NEXT: ; %bb.2: ; %atomicrmw.end
@@ -6669,37 +6669,37 @@ define amdgpu_gfx void @global_atomic_umin_i32_noret_scalar(ptr addrspace(1) inr
; SI: ; %bb.0:
; SI-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; SI-NEXT: s_xor_saveexec_b64 s[34:35], -1
-; SI-NEXT: buffer_store_dword v4, off, s[0:3], s32 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v3, off, s[0:3], s32 ; 4-byte Folded Spill
; SI-NEXT: s_mov_b64 exec, s[34:35]
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_writelane_b32 v4, s6, 0
-; SI-NEXT: v_writelane_b32 v4, s7, 1
+; SI-NEXT: v_writelane_b32 v3, s6, 0
+; SI-NEXT: v_writelane_b32 v3, s7, 1
; SI-NEXT: s_mov_b32 s34, s6
; SI-NEXT: s_mov_b32 s7, 0xf000
; SI-NEXT: s_mov_b32 s6, -1
-; SI-NEXT: buffer_load_dword v1, off, s[4:7], 0
+; SI-NEXT: buffer_load_dword v0, off, s[4:7], 0
; SI-NEXT: s_mov_b64 s[36:37], 0
; SI-NEXT: .LBB114_1: ; %atomicrmw.start
; SI-NEXT: ; =>This Inner Loop Header: Depth=1
; SI-NEXT: s_waitcnt vmcnt(0)
-; SI-NEXT: v_min_u32_e32 v0, s34, v1
-; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_mov_b32_e32 v3, v1
; SI-NEXT: v_mov_b32_e32 v2, v0
-; SI-NEXT: buffer_atomic_cmpswap v[2:3], off, s[4:7], 0 glc
+; SI-NEXT: s_waitcnt expcnt(0)
+; SI-NEXT: v_min_u32_e32 v1, s34, v2
+; SI-NEXT: v_mov_b32_e32 v0, v1
+; SI-NEXT: v_mov_b32_e32 v1, v2
+; SI-NEXT: buffer_atomic_cmpswap v[0:1], off, s[4:7], 0 glc
; SI-NEXT: s_waitcnt vmcnt(0)
; SI-NEXT: buffer_wbinvl1
-; SI-NEXT: v_cmp_eq_u32_e32 vcc, v2, v1
+; SI-NEXT: v_cmp_eq_u32_e32 vcc, v0, v2
; SI-NEXT: s_or_b64 s[36:37], vcc, s[36:37]
-; SI-NEXT: v_mov_b32_e32 v1, v2
; SI-NEXT: s_andn2_b64 exec, exec, s[36:37]
; SI-NEXT: s_cbranch_execnz .LBB114_1
; SI-NEXT: ; %bb.2: ; %atomicrmw.end
; SI-NEXT: s_or_b64 exec, exec, s[36:37]
-; SI-NEXT: v_readlane_b32 s7, v4, 1
-; SI-NEXT: v_readlane_b32 s6, v4, 0
+; SI-NEXT: v_readlane_b32 s7, v3, 1
+; SI-NEXT: v_readlane_b32 s6, v3, 0
; SI-NEXT: s_xor_saveexec_b64 s[34:35], -1
-; SI-NEXT: buffer_load_dword v4, off, s[0:3], s32 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v3, off, s[0:3], s32 ; 4-byte Folded Reload
; SI-NEXT: s_mov_b64 exec, s[34:35]
; SI-NEXT: s_waitcnt vmcnt(0) expcnt(0)
; SI-NEXT: s_setpc_b64 s[30:31]
@@ -6757,37 +6757,37 @@ define amdgpu_gfx void @global_atomic_umin_i32_noret_offset_scalar(ptr addrspace
; SI: ; %bb.0:
; SI-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; SI-NEXT: s_xor_saveexec_b64 s[34:35], -1
-; SI-NEXT: buffer_store_dword v4, off, s[0:3], s32 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v3, off, s[0:3], s32 ; 4-byte Folded Spill
; SI-NEXT: s_mov_b64 exec, s[34:35]
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_writelane_b32 v4, s6, 0
-; SI-NEXT: v_writelane_b32 v4, s7, 1
+; SI-NEXT: v_writelane_b32 v3, s6, 0
+; SI-NEXT: v_writelane_b32 v3, s7, 1
; SI-NEXT: s_mov_b32 s34, s6
; SI-NEXT: s_mov_b32 s7, 0xf000
; SI-NEXT: s_mov_b32 s6, -1
-; SI-NEXT: buffer_load_dword v1, off, s[4:7], 0 offset:16
+; SI-NEXT: buffer_load_dword v0, off, s[4:7], 0 offset:16
; SI-NEXT: s_mov_b64 s[36:37], 0
; SI-NEXT: .LBB115_1: ; %atomicrmw.start
; SI-NEXT: ; =>This Inner Loop Header: Depth=1
; SI-NEXT: s_waitcnt vmcnt(0)
-; SI-NEXT: v_min_u32_e32 v0, s34, v1
-; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_mov_b32_e32 v3, v1
; SI-NEXT: v_mov_b32_e32 v2, v0
-; SI-NEXT: buffer_atomic_cmpswap v[2:3], off, s[4:7], 0 offset:16 glc
+; SI-NEXT: s_waitcnt expcnt(0)
+; SI-NEXT: v_min_u32_e32 v1, s34, v2
+; SI-NEXT: v_mov_b32_e32 v0, v1
+; SI-NEXT: v_mov_b32_e32 v1, v2
+; SI-NEXT: buffer_atomic_cmpswap v[0:1], off, s[4:7], 0 offset:16 glc
; SI-NEXT: s_waitcnt vmcnt(0)
; SI-NEXT: buffer_wbinvl1
-; SI-NEXT: v_cmp_eq_u32_e32 vcc, v2, v1
+; SI-NEXT: v_cmp_eq_u32_e32 vcc, v0, v2
; SI-NEXT: s_or_b64 s[36:37], vcc, s[36:37]
-; SI-NEXT: v_mov_b32_e32 v1, v2
; SI-NEXT: s_andn2_b64 exec, exec, s[36:37]
; SI-NEXT: s_cbranch_execnz .LBB115_1
; SI-NEXT: ; %bb.2: ; %atomicrmw.end
; SI-NEXT: s_or_b64 exec, exec, s[36:37]
-; SI-NEXT: v_readlane_b32 s7, v4, 1
-; SI-NEXT: v_readlane_b32 s6, v4, 0
+; SI-NEXT: v_readlane_b32 s7, v3, 1
+; SI-NEXT: v_readlane_b32 s6, v3, 0
; SI-NEXT: s_xor_saveexec_b64 s[34:35], -1
-; SI-NEXT: buffer_load_dword v4, off, s[0:3], s32 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v3, off, s[0:3], s32 ; 4-byte Folded Reload
; SI-NEXT: s_mov_b64 exec, s[34:35]
; SI-NEXT: s_waitcnt vmcnt(0) expcnt(0)
; SI-NEXT: s_setpc_b64 s[30:31]
@@ -7032,21 +7032,21 @@ define void @global_atomic_umin_i32_noret_offset__amdgpu_no_remote_memory(ptr ad
; SI-NEXT: s_mov_b32 s7, 0xf000
; SI-NEXT: s_mov_b32 s4, s6
; SI-NEXT: s_mov_b32 s5, s6
-; SI-NEXT: buffer_load_dword v4, v[0:1], s[4:7], 0 addr64 offset:16
+; SI-NEXT: buffer_load_dword v3, v[0:1], s[4:7], 0 addr64 offset:16
; SI-NEXT: s_mov_b64 s[8:9], 0
; SI-NEXT: .LBB118_1: ; %atomicrmw.start
; SI-NEXT: ; =>This Inner Loop Header: Depth=1
; SI-NEXT: s_waitcnt vmcnt(0)
-; SI-NEXT: v_min_u32_e32 v3, v4, v2
-; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_mov_b32_e32 v6, v4
; SI-NEXT: v_mov_b32_e32 v5, v3
-; SI-NEXT: buffer_atomic_cmpswap v[5:6], v[0:1], s[4:7], 0 addr64 offset:16 glc
+; SI-NEXT: s_waitcnt expcnt(0)
+; SI-NEXT: v_min_u32_e32 v4, v5, v2
+; SI-NEXT: v_mov_b32_e32 v3, v4
+; SI-NEXT: v_mov_b32_e32 v4, v5
+; SI-NEXT: buffer_atomic_cmpswap v[3:4], v[0:1], s[4:7], 0 addr64 offset:16 glc
; SI-NEXT: s_waitcnt vmcnt(0)
; SI-NEXT: buffer_wbinvl1
-; SI-NEXT: v_cmp_eq_u32_e32 vcc, v5, v4
+; SI-NEXT: v_cmp_eq_u32_e32 vcc, v3, v5
; SI-NEXT: s_or_b64 s[8:9], vcc, s[8:9]
-; SI-NEXT: v_mov_b32_e32 v4, v5
; SI-NEXT: s_andn2_b64 exec, exec, s[8:9]
; SI-NEXT: s_cbranch_execnz .LBB118_1
; SI-NEXT: ; %bb.2: ; %atomicrmw.end
@@ -7194,21 +7194,21 @@ define void @global_atomic_min_i32_noret(ptr addrspace(1) %ptr, i32 %in) {
; SI-NEXT: s_mov_b32 s7, 0xf000
; SI-NEXT: s_mov_b32 s4, s6
; SI-NEXT: s_mov_b32 s5, s6
-; SI-NEXT: buffer_load_dword v4, v[0:1], s[4:7], 0 addr64
+; SI-NEXT: buffer_load_dword v3, v[0:1], s[4:7], 0 addr64
; SI-NEXT: s_mov_b64 s[8:9], 0
; SI-NEXT: .LBB120_1: ; %atomicrmw.start
; SI-NEXT: ; =>This Inner Loop Header: Depth=1
; SI-NEXT: s_waitcnt vmcnt(0)
-; SI-NEXT: v_min_i32_e32 v3, v4, v2
-; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_mov_b32_e32 v6, v4
; SI-NEXT: v_mov_b32_e32 v5, v3
-; SI-NEXT: buffer_atomic_cmpswap v[5:6], v[0:1], s[4:7], 0 addr64 glc
+; SI-NEXT: s_waitcnt expcnt(0)
+; SI-NEXT: v_min_i32_e32 v4, v5, v2
+; SI-NEXT: v_mov_b32_e32 v3, v4
+; SI-NEXT: v_mov_b32_e32 v4, v5
+; SI-NEXT: buffer_atomic_cmpswap v[3:4], v[0:1], s[4:7], 0 addr64 glc
; SI-NEXT: s_waitcnt vmcnt(0)
; SI-NEXT: buffer_wbinvl1
-; SI-NEXT: v_cmp_eq_u32_e32 vcc, v5, v4
+; SI-NEXT: v_cmp_eq_u32_e32 vcc, v3, v5
; SI-NEXT: s_or_b64 s[8:9], vcc, s[8:9]
-; SI-NEXT: v_mov_b32_e32 v4, v5
; SI-NEXT: s_andn2_b64 exec, exec, s[8:9]
; SI-NEXT: s_cbranch_execnz .LBB120_1
; SI-NEXT: ; %bb.2: ; %atomicrmw.end
@@ -7269,21 +7269,21 @@ define void @global_atomic_min_i32_noret_offset(ptr addrspace(1) %out, i32 %in)
; SI-NEXT: s_mov_b32 s7, 0xf000
; SI-NEXT: s_mov_b32 s4, s6
; SI-NEXT: s_mov_b32 s5, s6
-; SI-NEXT: buffer_load_dword v4, v[0:1], s[4:7], 0 addr64 offset:16
+; SI-NEXT: buffer_load_dword v3, v[0:1], s[4:7], 0 addr64 offset:16
; SI-NEXT: s_mov_b64 s[8:9], 0
; SI-NEXT: .LBB121_1: ; %atomicrmw.start
; SI-NEXT: ; =>This Inner Loop Header: Depth=1
; SI-NEXT: s_waitcnt vmcnt(0)
-; SI-NEXT: v_min_i32_e32 v3, v4, v2
-; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_mov_b32_e32 v6, v4
; SI-NEXT: v_mov_b32_e32 v5, v3
-; SI-NEXT: buffer_atomic_cmpswap v[5:6], v[0:1], s[4:7], 0 addr64 offset:16 glc
+; SI-NEXT: s_waitcnt expcnt(0)
+; SI-NEXT: v_min_i32_e32 v4, v5, v2
+; SI-NEXT: v_mov_b32_e32 v3, v4
+; SI-NEXT: v_mov_b32_e32 v4, v5
+; SI-NEXT: buffer_atomic_cmpswap v[3:4], v[0:1], s[4:7], 0 addr64 offset:16 glc
; SI-NEXT: s_waitcnt vmcnt(0)
; SI-NEXT: buffer_wbinvl1
-; SI-NEXT: v_cmp_eq_u32_e32 vcc, v5, v4
+; SI-NEXT: v_cmp_eq_u32_e32 vcc, v3, v5
; SI-NEXT: s_or_b64 s[8:9], vcc, s[8:9]
-; SI-NEXT: v_mov_b32_e32 v4, v5
; SI-NEXT: s_andn2_b64 exec, exec, s[8:9]
; SI-NEXT: s_cbranch_execnz .LBB121_1
; SI-NEXT: ; %bb.2: ; %atomicrmw.end
@@ -7502,37 +7502,37 @@ define amdgpu_gfx void @global_atomic_min_i32_noret_scalar(ptr addrspace(1) inre
; SI: ; %bb.0:
; SI-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; SI-NEXT: s_xor_saveexec_b64 s[34:35], -1
-; SI-NEXT: buffer_store_dword v4, off, s[0:3], s32 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v3, off, s[0:3], s32 ; 4-byte Folded Spill
; SI-NEXT: s_mov_b64 exec, s[34:35]
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_writelane_b32 v4, s6, 0
-; SI-NEXT: v_writelane_b32 v4, s7, 1
+; SI-NEXT: v_writelane_b32 v3, s6, 0
+; SI-NEXT: v_writelane_b32 v3, s7, 1
; SI-NEXT: s_mov_b32 s34, s6
; SI-NEXT: s_mov_b32 s7, 0xf000
; SI-NEXT: s_mov_b32 s6, -1
-; SI-NEXT: buffer_load_dword v1, off, s[4:7], 0
+; SI-NEXT: buffer_load_dword v0, off, s[4:7], 0
; SI-NEXT: s_mov_b64 s[36:37], 0
; SI-NEXT: .LBB124_1: ; %atomicrmw.start
; SI-NEXT: ; =>This Inner Loop Header: Depth=1
; SI-NEXT: s_waitcnt vmcnt(0)
-; SI-NEXT: v_min_i32_e32 v0, s34, v1
-; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_mov_b32_e32 v3, v1
; SI-NEXT: v_mov_b32_e32 v2, v0
-; SI-NEXT: buffer_atomic_cmpswap v[2:3], off, s[4:7], 0 glc
+; SI-NEXT: s_waitcnt expcnt(0)
+; SI-NEXT: v_min_i32_e32 v1, s34, v2
+; SI-NEXT: v_mov_b32_e32 v0, v1
+; SI-NEXT: v_mov_b32_e32 v1, v2
+; SI-NEXT: buffer_atomic_cmpswap v[0:1], off, s[4:7], 0 glc
; SI-NEXT: s_waitcnt vmcnt(0)
; SI-NEXT: buffer_wbinvl1
-; SI-NEXT: v_cmp_eq_u32_e32 vcc, v2, v1
+; SI-NEXT: v_cmp_eq_u32_e32 vcc, v0, v2
; SI-NEXT: s_or_b64 s[36:37], vcc, s[36:37]
-; SI-NEXT: v_mov_b32_e32 v1, v2
; SI-NEXT: s_andn2_b64 exec, exec, s[36:37]
; SI-NEXT: s_cbranch_execnz .LBB124_1
; SI-NEXT: ; %bb.2: ; %atomicrmw.end
; SI-NEXT: s_or_b64 exec, exec, s[36:37]
-; SI-NEXT: v_readlane_b32 s7, v4, 1
-; SI-NEXT: v_readlane_b32 s6, v4, 0
+; SI-NEXT: v_readlane_b32 s7, v3, 1
+; SI-NEXT: v_readlane_b32 s6, v3, 0
; SI-NEXT: s_xor_saveexec_b64 s[34:35], -1
-; SI-NEXT: buffer_load_dword v4, off, s[0:3], s32 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v3, off, s[0:3], s32 ; 4-byte Folded Reload
; SI-NEXT: s_mov_b64 exec, s[34:35]
; SI-NEXT: s_waitcnt vmcnt(0) expcnt(0)
; SI-NEXT: s_setpc_b64 s[30:31]
@@ -7590,37 +7590,37 @@ define amdgpu_gfx void @global_atomic_min_i32_noret_offset_scalar(ptr addrspace(
; SI: ; %bb.0:
; SI-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; SI-NEXT: s_xor_saveexec_b64 s[34:35], -1
-; SI-NEXT: buffer_store_dword v4, off, s[0:3], s32 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v3, off, s[0:3], s32 ; 4-byte Folded Spill
; SI-NEXT: s_mov_b64 exec, s[34:35]
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_writelane_b32 v4, s6, 0
-; SI-NEXT: v_writelane_b32 v4, s7, 1
+; SI-NEXT: v_writelane_b32 v3, s6, 0
+; SI-NEXT: v_writelane_b32 v3, s7, 1
; SI-NEXT: s_mov_b32 s34, s6
; SI-NEXT: s_mov_b32 s7, 0xf000
; SI-NEXT: s_mov_b32 s6, -1
-; SI-NEXT: buffer_load_dword v1, off, s[4:7], 0 offset:16
+; SI-NEXT: buffer_load_dword v0, off, s[4:7], 0 offset:16
; SI-NEXT: s_mov_b64 s[36:37], 0
; SI-NEXT: .LBB125_1: ; %atomicrmw.start
; SI-NEXT: ; =>This Inner Loop Header: Depth=1
; SI-NEXT: s_waitcnt vmcnt(0)
-; SI-NEXT: v_min_i32_e32 v0, s34, v1
-; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_mov_b32_e32 v3, v1
; SI-NEXT: v_mov_b32_e32 v2, v0
-; SI-NEXT: buffer_atomic_cmpswap v[2:3], off, s[4:7], 0 offset:16 glc
+; SI-NEXT: s_waitcnt expcnt(0)
+; SI-NEXT: v_min_i32_e32 v1, s34, v2
+; SI-NEXT: v_mov_b32_e32 v0, v1
+; SI-NEXT: v_mov_b32_e32 v1, v2
+; SI-NEXT: buffer_atomic_cmpswap v[0:1], off, s[4:7], 0 offset:16 glc
; SI-NEXT: s_waitcnt vmcnt(0)
; SI-NEXT: buffer_wbinvl1
-; SI-NEXT: v_cmp_eq_u32_e32 vcc, v2, v1
+; SI-NEXT: v_cmp_eq_u32_e32 vcc, v0, v2
; SI-NEXT: s_or_b64 s[36:37], vcc, s[36:37]
-; SI-NEXT: v_mov_b32_e32 v1, v2
; SI-NEXT: s_andn2_b64 exec, exec, s[36:37]
; SI-NEXT: s_cbranch_execnz .LBB125_1
; SI-NEXT: ; %bb.2: ; %atomicrmw.end
; SI-NEXT: s_or_b64 exec, exec, s[36:37]
-; SI-NEXT: v_readlane_b32 s7, v4, 1
-; SI-NEXT: v_readlane_b32 s6, v4, 0
+; SI-NEXT: v_readlane_b32 s7, v3, 1
+; SI-NEXT: v_readlane_b32 s6, v3, 0
; SI-NEXT: s_xor_saveexec_b64 s[34:35], -1
-; SI-NEXT: buffer_load_dword v4, off, s[0:3], s32 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v3, off, s[0:3], s32 ; 4-byte Folded Reload
; SI-NEXT: s_mov_b64 exec, s[34:35]
; SI-NEXT: s_waitcnt vmcnt(0) expcnt(0)
; SI-NEXT: s_setpc_b64 s[30:31]
@@ -7871,20 +7871,20 @@ define amdgpu_kernel void @atomic_min_i32_addr64_offset(ptr addrspace(1) %out, i
; SI-NEXT: s_mov_b64 s[0:1], 0
; SI-NEXT: s_mov_b32 s7, 0xf000
; SI-NEXT: s_waitcnt lgkmcnt(0)
-; SI-NEXT: v_mov_b32_e32 v1, s3
+; SI-NEXT: v_mov_b32_e32 v0, s3
; SI-NEXT: s_mov_b32 s6, -1
; SI-NEXT: .LBB128_1: ; %atomicrmw.start
; SI-NEXT: ; =>This Inner Loop Header: Depth=1
-; SI-NEXT: v_min_i32_e32 v0, s2, v1
-; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_mov_b32_e32 v3, v1
; SI-NEXT: v_mov_b32_e32 v2, v0
-; SI-NEXT: buffer_atomic_cmpswap v[2:3], off, s[4:7], 0 offset:16 glc
+; SI-NEXT: s_waitcnt expcnt(0)
+; SI-NEXT: v_min_i32_e32 v1, s2, v2
+; SI-NEXT: v_mov_b32_e32 v0, v1
+; SI-NEXT: v_mov_b32_e32 v1, v2
+; SI-NEXT: buffer_atomic_cmpswap v[0:1], off, s[4:7], 0 offset:16 glc
; SI-NEXT: s_waitcnt vmcnt(0)
; SI-NEXT: buffer_wbinvl1
-; SI-NEXT: v_cmp_eq_u32_e32 vcc, v2, v1
+; SI-NEXT: v_cmp_eq_u32_e32 vcc, v0, v2
; SI-NEXT: s_or_b64 s[0:1], vcc, s[0:1]
-; SI-NEXT: v_mov_b32_e32 v1, v2
; SI-NEXT: s_andn2_b64 exec, exec, s[0:1]
; SI-NEXT: s_cbranch_execnz .LBB128_1
; SI-NEXT: ; %bb.2: ; %atomicrmw.end
@@ -7970,20 +7970,20 @@ define amdgpu_kernel void @atomic_min_i32_ret_addr64_offset(ptr addrspace(1) %ou
; SI-NEXT: s_mov_b64 s[0:1], 0
; SI-NEXT: s_mov_b32 s7, 0xf000
; SI-NEXT: s_waitcnt lgkmcnt(0)
-; SI-NEXT: v_mov_b32_e32 v1, s6
+; SI-NEXT: v_mov_b32_e32 v0, s6
; SI-NEXT: s_mov_b32 s6, -1
; SI-NEXT: .LBB129_1: ; %atomicrmw.start
; SI-NEXT: ; =>This Inner Loop Header: Depth=1
-; SI-NEXT: v_min_i32_e32 v0, s8, v1
-; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_mov_b32_e32 v3, v1
; SI-NEXT: v_mov_b32_e32 v2, v0
-; SI-NEXT: buffer_atomic_cmpswap v[2:3], off, s[4:7], 0 offset:16 glc
+; SI-NEXT: s_waitcnt expcnt(0)
+; SI-NEXT: v_min_i32_e32 v1, s8, v2
+; SI-NEXT: v_mov_b32_e32 v0, v1
+; SI-NEXT: v_mov_b32_e32 v1, v2
+; SI-NEXT: buffer_atomic_cmpswap v[0:1], off, s[4:7], 0 offset:16 glc
; SI-NEXT: s_waitcnt vmcnt(0)
; SI-NEXT: buffer_wbinvl1
-; SI-NEXT: v_cmp_eq_u32_e32 vcc, v2, v1
+; SI-NEXT: v_cmp_eq_u32_e32 vcc, v0, v2
; SI-NEXT: s_or_b64 s[0:1], vcc, s[0:1]
-; SI-NEXT: v_mov_b32_e32 v1, v2
; SI-NEXT: s_andn2_b64 exec, exec, s[0:1]
; SI-NEXT: s_cbranch_execnz .LBB129_1
; SI-NEXT: ; %bb.2: ; %atomicrmw.end
@@ -7992,7 +7992,7 @@ define amdgpu_kernel void @atomic_min_i32_ret_addr64_offset(ptr addrspace(1) %ou
; SI-NEXT: s_mov_b32 s6, -1
; SI-NEXT: s_mov_b32 s4, s2
; SI-NEXT: s_mov_b32 s5, s3
-; SI-NEXT: buffer_store_dword v2, off, s[4:7], 0
+; SI-NEXT: buffer_store_dword v0, off, s[4:7], 0
; SI-NEXT: s_endpgm
;
; VI-LABEL: atomic_min_i32_ret_addr64_offset:
@@ -8080,20 +8080,20 @@ define amdgpu_kernel void @atomic_min_i32(ptr addrspace(1) %out, i32 %in) {
; SI-NEXT: s_mov_b64 s[4:5], 0
; SI-NEXT: s_mov_b32 s3, 0xf000
; SI-NEXT: s_waitcnt lgkmcnt(0)
-; SI-NEXT: v_mov_b32_e32 v1, s2
+; SI-NEXT: v_mov_b32_e32 v0, s2
; SI-NEXT: s_mov_b32 s2, -1
; SI-NEXT: .LBB130_1: ; %atomicrmw.start
; SI-NEXT: ; =>This Inner Loop Header: Depth=1
-; SI-NEXT: v_min_i32_e32 v0, s6, v1
-; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_mov_b32_e32 v3, v1
; SI-NEXT: v_mov_b32_e32 v2, v0
-; SI-NEXT: buffer_atomic_cmpswap v[2:3], off, s[0:3], 0 glc
+; SI-NEXT: s_waitcnt expcnt(0)
+; SI-NEXT: v_min_i32_e32 v1, s6, v2
+; SI-NEXT: v_mov_b32_e32 v0, v1
+; SI-NEXT: v_mov_b32_e32 v1, v2
+; SI-NEXT: buffer_atomic_cmpswap v[0:1], off, s[0:3], 0 glc
; SI-NEXT: s_waitcnt vmcnt(0)
; SI-NEXT: buffer_wbinvl1
-; SI-NEXT: v_cmp_eq_u32_e32 vcc, v2, v1
+; SI-NEXT: v_cmp_eq_u32_e32 vcc, v0, v2
; SI-NEXT: s_or_b64 s[4:5], vcc, s[4:5]
-; SI-NEXT: v_mov_b32_e32 v1, v2
; SI-NEXT: s_andn2_b64 exec, exec, s[4:5]
; SI-NEXT: s_cbranch_execnz .LBB130_1
; SI-NEXT: ; %bb.2: ; %atomicrmw.end
@@ -8167,20 +8167,20 @@ define amdgpu_kernel void @atomic_min_i32_ret_addr64(ptr addrspace(1) %out, ptr
; SI-NEXT: s_mov_b64 s[0:1], 0
; SI-NEXT: s_mov_b32 s7, 0xf000
; SI-NEXT: s_waitcnt lgkmcnt(0)
-; SI-NEXT: v_mov_b32_e32 v1, s6
+; SI-NEXT: v_mov_b32_e32 v0, s6
; SI-NEXT: s_mov_b32 s6, -1
; SI-NEXT: .LBB131_1: ; %atomicrmw.start
; SI-NEXT: ; =>This Inner Loop Header: Depth=1
-; SI-NEXT: v_min_i32_e32 v0, s8, v1
-; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_mov_b32_e32 v3, v1
; SI-NEXT: v_mov_b32_e32 v2, v0
-; SI-NEXT: buffer_atomic_cmpswap v[2:3], off, s[4:7], 0 glc
+; SI-NEXT: s_waitcnt expcnt(0)
+; SI-NEXT: v_min_i32_e32 v1, s8, v2
+; SI-NEXT: v_mov_b32_e32 v0, v1
+; SI-NEXT: v_mov_b32_e32 v1, v2
+; SI-NEXT: buffer_atomic_cmpswap v[0:1], off, s[4:7], 0 glc
; SI-NEXT: s_waitcnt vmcnt(0)
; SI-NEXT: buffer_wbinvl1
-; SI-NEXT: v_cmp_eq_u32_e32 vcc, v2, v1
+; SI-NEXT: v_cmp_eq_u32_e32 vcc, v0, v2
; SI-NEXT: s_or_b64 s[0:1], vcc, s[0:1]
-; SI-NEXT: v_mov_b32_e32 v1, v2
; SI-NEXT: s_andn2_b64 exec, exec, s[0:1]
; SI-NEXT: s_cbranch_execnz .LBB131_1
; SI-NEXT: ; %bb.2: ; %atomicrmw.end
@@ -8189,7 +8189,7 @@ define amdgpu_kernel void @atomic_min_i32_ret_addr64(ptr addrspace(1) %out, ptr
; SI-NEXT: s_mov_b32 s6, -1
; SI-NEXT: s_mov_b32 s4, s2
; SI-NEXT: s_mov_b32 s5, s3
-; SI-NEXT: buffer_store_dword v2, off, s[4:7], 0
+; SI-NEXT: buffer_store_dword v0, off, s[4:7], 0
; SI-NEXT: s_endpgm
;
; VI-LABEL: atomic_min_i32_ret_addr64:
@@ -8272,21 +8272,21 @@ define void @global_atomic_min_i32_noret_offset__amdgpu_no_remote_memory(ptr add
; SI-NEXT: s_mov_b32 s7, 0xf000
; SI-NEXT: s_mov_b32 s4, s6
; SI-NEXT: s_mov_b32 s5, s6
-; SI-NEXT: buffer_load_dword v4, v[0:1], s[4:7], 0 addr64 offset:16
+; SI-NEXT: buffer_load_dword v3, v[0:1], s[4:7], 0 addr64 offset:16
; SI-NEXT: s_mov_b64 s[8:9], 0
; SI-NEXT: .LBB132_1: ; %atomicrmw.start
; SI-NEXT: ; =>This Inner Loop Header: Depth=1
; SI-NEXT: s_waitcnt vmcnt(0)
-; SI-NEXT: v_min_i32_e32 v3, v4, v2
-; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_mov_b32_e32 v6, v4
; SI-NEXT: v_mov_b32_e32 v5, v3
-; SI-NEXT: buffer_atomic_cmpswap v[5:6], v[0:1], s[4:7], 0 addr64 offset:16 glc
+; SI-NEXT: s_waitcnt expcnt(0)
+; SI-NEXT: v_min_i32_e32 v4, v5, v2
+; SI-NEXT: v_mov_b32_e32 v3, v4
+; SI-NEXT: v_mov_b32_e32 v4, v5
+; SI-NEXT: buffer_atomic_cmpswap v[3:4], v[0:1], s[4:7], 0 addr64 offset:16 glc
; SI-NEXT: s_waitcnt vmcnt(0)
; SI-NEXT: buffer_wbinvl1
-; SI-NEXT: v_cmp_eq_u32_e32 vcc, v5, v4
+; SI-NEXT: v_cmp_eq_u32_e32 vcc, v3, v5
; SI-NEXT: s_or_b64 s[8:9], vcc, s[8:9]
-; SI-NEXT: v_mov_b32_e32 v4, v5
; SI-NEXT: s_andn2_b64 exec, exec, s[8:9]
; SI-NEXT: s_cbranch_execnz .LBB132_1
; SI-NEXT: ; %bb.2: ; %atomicrmw.end
diff --git a/llvm/test/CodeGen/AMDGPU/global_atomics_i64_system.ll b/llvm/test/CodeGen/AMDGPU/global_atomics_i64_system.ll
index 59a99a6a0328d..9845064604bc1 100644
--- a/llvm/test/CodeGen/AMDGPU/global_atomics_i64_system.ll
+++ b/llvm/test/CodeGen/AMDGPU/global_atomics_i64_system.ll
@@ -2205,27 +2205,27 @@ define void @global_atomic_nand_i64_noret(ptr addrspace(1) %ptr, i64 %in) {
; SI-NEXT: s_mov_b32 s7, 0xf000
; SI-NEXT: s_mov_b32 s4, s6
; SI-NEXT: s_mov_b32 s5, s6
-; SI-NEXT: buffer_load_dwordx2 v[6:7], v[0:1], s[4:7], 0 addr64
+; SI-NEXT: buffer_load_dwordx2 v[4:5], v[0:1], s[4:7], 0 addr64
; SI-NEXT: s_mov_b64 s[8:9], 0
; SI-NEXT: .LBB50_1: ; %atomicrmw.start
; SI-NEXT: ; =>This Inner Loop Header: Depth=1
; SI-NEXT: s_waitcnt vmcnt(0)
-; SI-NEXT: v_and_b32_e32 v4, v7, v3
-; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_and_b32_e32 v8, v6, v2
-; SI-NEXT: v_not_b32_e32 v5, v4
-; SI-NEXT: v_not_b32_e32 v4, v8
-; SI-NEXT: v_mov_b32_e32 v11, v7
-; SI-NEXT: v_mov_b32_e32 v10, v6
; SI-NEXT: v_mov_b32_e32 v9, v5
; SI-NEXT: v_mov_b32_e32 v8, v4
-; SI-NEXT: buffer_atomic_cmpswap_x2 v[8:11], v[0:1], s[4:7], 0 addr64 glc
+; SI-NEXT: s_waitcnt expcnt(0)
+; SI-NEXT: v_and_b32_e32 v4, v9, v3
+; SI-NEXT: v_and_b32_e32 v5, v8, v2
+; SI-NEXT: v_not_b32_e32 v7, v4
+; SI-NEXT: v_not_b32_e32 v6, v5
+; SI-NEXT: v_mov_b32_e32 v4, v6
+; SI-NEXT: v_mov_b32_e32 v5, v7
+; SI-NEXT: v_mov_b32_e32 v6, v8
+; SI-NEXT: v_mov_b32_e32 v7, v9
+; SI-NEXT: buffer_atomic_cmpswap_x2 v[4:7], v[0:1], s[4:7], 0 addr64 glc
; SI-NEXT: s_waitcnt vmcnt(0)
; SI-NEXT: buffer_wbinvl1
-; SI-NEXT: v_cmp_eq_u64_e32 vcc, v[8:9], v[6:7]
+; SI-NEXT: v_cmp_eq_u64_e32 vcc, v[4:5], v[8:9]
; SI-NEXT: s_or_b64 s[8:9], vcc, s[8:9]
-; SI-NEXT: v_mov_b32_e32 v6, v8
-; SI-NEXT: v_mov_b32_e32 v7, v9
; SI-NEXT: s_andn2_b64 exec, exec, s[8:9]
; SI-NEXT: s_cbranch_execnz .LBB50_1
; SI-NEXT: ; %bb.2: ; %atomicrmw.end
@@ -2294,27 +2294,27 @@ define void @global_atomic_nand_i64_noret_offset(ptr addrspace(1) %out, i64 %in)
; SI-NEXT: s_mov_b32 s7, 0xf000
; SI-NEXT: s_mov_b32 s4, s6
; SI-NEXT: s_mov_b32 s5, s6
-; SI-NEXT: buffer_load_dwordx2 v[6:7], v[0:1], s[4:7], 0 addr64 offset:32
+; SI-NEXT: buffer_load_dwordx2 v[4:5], v[0:1], s[4:7], 0 addr64 offset:32
; SI-NEXT: s_mov_b64 s[8:9], 0
; SI-NEXT: .LBB51_1: ; %atomicrmw.start
; SI-NEXT: ; =>This Inner Loop Header: Depth=1
; SI-NEXT: s_waitcnt vmcnt(0)
-; SI-NEXT: v_and_b32_e32 v4, v7, v3
-; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_and_b32_e32 v8, v6, v2
-; SI-NEXT: v_not_b32_e32 v5, v4
-; SI-NEXT: v_not_b32_e32 v4, v8
-; SI-NEXT: v_mov_b32_e32 v11, v7
-; SI-NEXT: v_mov_b32_e32 v10, v6
; SI-NEXT: v_mov_b32_e32 v9, v5
; SI-NEXT: v_mov_b32_e32 v8, v4
-; SI-NEXT: buffer_atomic_cmpswap_x2 v[8:11], v[0:1], s[4:7], 0 addr64 offset:32 glc
+; SI-NEXT: s_waitcnt expcnt(0)
+; SI-NEXT: v_and_b32_e32 v4, v9, v3
+; SI-NEXT: v_and_b32_e32 v5, v8, v2
+; SI-NEXT: v_not_b32_e32 v7, v4
+; SI-NEXT: v_not_b32_e32 v6, v5
+; SI-NEXT: v_mov_b32_e32 v4, v6
+; SI-NEXT: v_mov_b32_e32 v5, v7
+; SI-NEXT: v_mov_b32_e32 v6, v8
+; SI-NEXT: v_mov_b32_e32 v7, v9
+; SI-NEXT: buffer_atomic_cmpswap_x2 v[4:7], v[0:1], s[4:7], 0 addr64 offset:32 glc
; SI-NEXT: s_waitcnt vmcnt(0)
; SI-NEXT: buffer_wbinvl1
-; SI-NEXT: v_cmp_eq_u64_e32 vcc, v[8:9], v[6:7]
+; SI-NEXT: v_cmp_eq_u64_e32 vcc, v[4:5], v[8:9]
; SI-NEXT: s_or_b64 s[8:9], vcc, s[8:9]
-; SI-NEXT: v_mov_b32_e32 v6, v8
-; SI-NEXT: v_mov_b32_e32 v7, v9
; SI-NEXT: s_andn2_b64 exec, exec, s[8:9]
; SI-NEXT: s_cbranch_execnz .LBB51_1
; SI-NEXT: ; %bb.2: ; %atomicrmw.end
@@ -2578,44 +2578,44 @@ define amdgpu_gfx void @global_atomic_nand_i64_noret_scalar(ptr addrspace(1) inr
; SI: ; %bb.0:
; SI-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; SI-NEXT: s_xor_saveexec_b64 s[34:35], -1
-; SI-NEXT: buffer_store_dword v8, off, s[0:3], s32 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v6, off, s[0:3], s32 ; 4-byte Folded Spill
; SI-NEXT: s_mov_b64 exec, s[34:35]
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_writelane_b32 v8, s6, 0
-; SI-NEXT: v_writelane_b32 v8, s7, 1
+; SI-NEXT: v_writelane_b32 v6, s6, 0
+; SI-NEXT: v_writelane_b32 v6, s7, 1
; SI-NEXT: s_mov_b32 s34, s7
; SI-NEXT: s_mov_b32 s35, s6
; SI-NEXT: s_mov_b32 s7, 0xf000
; SI-NEXT: s_mov_b32 s6, -1
-; SI-NEXT: buffer_load_dwordx2 v[2:3], off, s[4:7], 0
+; SI-NEXT: buffer_load_dwordx2 v[0:1], off, s[4:7], 0
; SI-NEXT: s_mov_b64 s[36:37], 0
; SI-NEXT: .LBB54_1: ; %atomicrmw.start
; SI-NEXT: ; =>This Inner Loop Header: Depth=1
; SI-NEXT: s_waitcnt vmcnt(0)
-; SI-NEXT: v_and_b32_e32 v0, s34, v3
-; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_and_b32_e32 v4, s35, v2
-; SI-NEXT: v_not_b32_e32 v1, v0
-; SI-NEXT: v_not_b32_e32 v0, v4
-; SI-NEXT: v_mov_b32_e32 v7, v3
-; SI-NEXT: v_mov_b32_e32 v6, v2
; SI-NEXT: v_mov_b32_e32 v5, v1
; SI-NEXT: v_mov_b32_e32 v4, v0
-; SI-NEXT: buffer_atomic_cmpswap_x2 v[4:7], off, s[4:7], 0 glc
+; SI-NEXT: s_waitcnt expcnt(0)
+; SI-NEXT: v_and_b32_e32 v0, s34, v5
+; SI-NEXT: v_and_b32_e32 v1, s35, v4
+; SI-NEXT: v_not_b32_e32 v3, v0
+; SI-NEXT: v_not_b32_e32 v2, v1
+; SI-NEXT: v_mov_b32_e32 v0, v2
+; SI-NEXT: v_mov_b32_e32 v1, v3
+; SI-NEXT: v_mov_b32_e32 v2, v4
+; SI-NEXT: v_mov_b32_e32 v3, v5
+; SI-NEXT: buffer_atomic_cmpswap_x2 v[0:3], off, s[4:7], 0 glc
; SI-NEXT: s_waitcnt vmcnt(0)
; SI-NEXT: buffer_wbinvl1
-; SI-NEXT: v_cmp_eq_u64_e32 vcc, v[4:5], v[2:3]
+; SI-NEXT: v_cmp_eq_u64_e32 vcc, v[0:1], v[4:5]
; SI-NEXT: s_or_b64 s[36:37], vcc, s[36:37]
-; SI-NEXT: v_mov_b32_e32 v2, v4
-; SI-NEXT: v_mov_b32_e32 v3, v5
; SI-NEXT: s_andn2_b64 exec, exec, s[36:37]
; SI-NEXT: s_cbranch_execnz .LBB54_1
; SI-NEXT: ; %bb.2: ; %atomicrmw.end
; SI-NEXT: s_or_b64 exec, exec, s[36:37]
-; SI-NEXT: v_readlane_b32 s7, v8, 1
-; SI-NEXT: v_readlane_b32 s6, v8, 0
+; SI-NEXT: v_readlane_b32 s7, v6, 1
+; SI-NEXT: v_readlane_b32 s6, v6, 0
; SI-NEXT: s_xor_saveexec_b64 s[34:35], -1
-; SI-NEXT: buffer_load_dword v8, off, s[0:3], s32 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v6, off, s[0:3], s32 ; 4-byte Folded Reload
; SI-NEXT: s_mov_b64 exec, s[34:35]
; SI-NEXT: s_waitcnt vmcnt(0) expcnt(0)
; SI-NEXT: s_setpc_b64 s[30:31]
@@ -2683,44 +2683,44 @@ define amdgpu_gfx void @global_atomic_nand_i64_noret_offset_scalar(ptr addrspace
; SI: ; %bb.0:
; SI-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; SI-NEXT: s_xor_saveexec_b64 s[34:35], -1
-; SI-NEXT: buffer_store_dword v8, off, s[0:3], s32 ; 4-byte Folded Spill
+; SI-NEXT: buffer_store_dword v6, off, s[0:3], s32 ; 4-byte Folded Spill
; SI-NEXT: s_mov_b64 exec, s[34:35]
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_writelane_b32 v8, s6, 0
-; SI-NEXT: v_writelane_b32 v8, s7, 1
+; SI-NEXT: v_writelane_b32 v6, s6, 0
+; SI-NEXT: v_writelane_b32 v6, s7, 1
; SI-NEXT: s_mov_b32 s34, s7
; SI-NEXT: s_mov_b32 s35, s6
; SI-NEXT: s_mov_b32 s7, 0xf000
; SI-NEXT: s_mov_b32 s6, -1
-; SI-NEXT: buffer_load_dwordx2 v[2:3], off, s[4:7], 0 offset:32
+; SI-NEXT: buffer_load_dwordx2 v[0:1], off, s[4:7], 0 offset:32
; SI-NEXT: s_mov_b64 s[36:37], 0
; SI-NEXT: .LBB55_1: ; %atomicrmw.start
; SI-NEXT: ; =>This Inner Loop Header: Depth=1
; SI-NEXT: s_waitcnt vmcnt(0)
-; SI-NEXT: v_and_b32_e32 v0, s34, v3
-; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_and_b32_e32 v4, s35, v2
-; SI-NEXT: v_not_b32_e32 v1, v0
-; SI-NEXT: v_not_b32_e32 v0, v4
-; SI-NEXT: v_mov_b32_e32 v7, v3
-; SI-NEXT: v_mov_b32_e32 v6, v2
; SI-NEXT: v_mov_b32_e32 v5, v1
; SI-NEXT: v_mov_b32_e32 v4, v0
-; SI-NEXT: buffer_atomic_cmpswap_x2 v[4:7], off, s[4:7], 0 offset:32 glc
+; SI-NEXT: s_waitcnt expcnt(0)
+; SI-NEXT: v_and_b32_e32 v0, s34, v5
+; SI-NEXT: v_and_b32_e32 v1, s35, v4
+; SI-NEXT: v_not_b32_e32 v3, v0
+; SI-NEXT: v_not_b32_e32 v2, v1
+; SI-NEXT: v_mov_b32_e32 v0, v2
+; SI-NEXT: v_mov_b32_e32 v1, v3
+; SI-NEXT: v_mov_b32_e32 v2, v4
+; SI-NEXT: v_mov_b32_e32 v3, v5
+; SI-NEXT: buffer_atomic_cmpswap_x2 v[0:3], off, s[4:7], 0 offset:32 glc
; SI-NEXT: s_waitcnt vmcnt(0)
; SI-NEXT: buffer_wbinvl1
-; SI-NEXT: v_cmp_eq_u64_e32 vcc, v[4:5], v[2:3]
+; SI-NEXT: v_cmp_eq_u64_e32 vcc, v[0:1], v[4:5]
; SI-NEXT: s_or_b64 s[36:37], vcc, s[36:37]
-; SI-NEXT: v_mov_b32_e32 v2, v4
-; SI-NEXT: v_mov_b32_e32 v3, v5
; SI-NEXT: s_andn2_b64 exec, exec, s[36:37]
; SI-NEXT: s_cbranch_execnz .LBB55_1
; SI-NEXT: ; %bb.2: ; %atomicrmw.end
; SI-NEXT: s_or_b64 exec, exec, s[36:37]
-; SI-NEXT: v_readlane_b32 s7, v8, 1
-; SI-NEXT: v_readlane_b32 s6, v8, 0
+; SI-NEXT: v_readlane_b32 s7, v6, 1
+; SI-NEXT: v_readlane_b32 s6, v6, 0
; SI-NEXT: s_xor_saveexec_b64 s[34:35], -1
-; SI-NEXT: buffer_load_dword v8, off, s[0:3], s32 ; 4-byte Folded Reload
+; SI-NEXT: buffer_load_dword v6, off, s[0:3], s32 ; 4-byte Folded Reload
; SI-NEXT: s_mov_b64 exec, s[34:35]
; SI-NEXT: s_waitcnt vmcnt(0) expcnt(0)
; SI-NEXT: s_setpc_b64 s[30:31]
@@ -3003,27 +3003,27 @@ define void @global_atomic_nand_i64_noret_offset__amdgpu_no_remote_memory(ptr ad
; SI-NEXT: s_mov_b32 s7, 0xf000
; SI-NEXT: s_mov_b32 s4, s6
; SI-NEXT: s_mov_b32 s5, s6
-; SI-NEXT: buffer_load_dwordx2 v[6:7], v[0:1], s[4:7], 0 addr64 offset:32
+; SI-NEXT: buffer_load_dwordx2 v[4:5], v[0:1], s[4:7], 0 addr64 offset:32
; SI-NEXT: s_mov_b64 s[8:9], 0
; SI-NEXT: .LBB58_1: ; %atomicrmw.start
; SI-NEXT: ; =>This Inner Loop Header: Depth=1
; SI-NEXT: s_waitcnt vmcnt(0)
-; SI-NEXT: v_and_b32_e32 v4, v7, v3
-; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_and_b32_e32 v8, v6, v2
-; SI-NEXT: v_not_b32_e32 v5, v4
-; SI-NEXT: v_not_b32_e32 v4, v8
-; SI-NEXT: v_mov_b32_e32 v11, v7
-; SI-NEXT: v_mov_b32_e32 v10, v6
; SI-NEXT: v_mov_b32_e32 v9, v5
; SI-NEXT: v_mov_b32_e32 v8, v4
-; SI-NEXT: buffer_atomic_cmpswap_x2 v[8:11], v[0:1], s[4:7], 0 addr64 offset:32 glc
+; SI-NEXT: s_waitcnt expcnt(0)
+; SI-NEXT: v_and_b32_e32 v4, v9, v3
+; SI-NEXT: v_and_b32_e32 v5, v8, v2
+; SI-NEXT: v_not_b32_e32 v7, v4
+; SI-NEXT: v_not_b32_e32 v6, v5
+; SI-NEXT: v_mov_b32_e32 v4, v6
+; SI-NEXT: v_mov_b32_e32 v5, v7
+; SI-NEXT: v_mov_b32_e32 v6, v8
+; SI-NEXT: v_mov_b32_e32 v7, v9
+; SI-NEXT: buffer_atomic_cmpswap_x2 v[4:7], v[0:1], s[4:7], 0 addr64 offset:32 glc
; SI-NEXT: s_waitcnt vmcnt(0)
; SI-NEXT: buffer_wbinvl1
-; SI-NEXT: v_cmp_eq_u64_e32 vcc, v[8:9], v[6:7]
+; SI-NEXT: v_cmp_eq_u64_e32 vcc, v[4:5], v[8:9]
; SI-NEXT: s_or_b64 s[8:9], vcc, s[8:9]
-; SI-NEXT: v_mov_b32_e32 v6, v8
-; SI-NEXT: v_mov_b32_e32 v7, v9
; SI-NEXT: s_andn2_b64 exec, exec, s[8:9]
; SI-NEXT: s_cbranch_execnz .LBB58_1
; SI-NEXT: ; %bb.2: ; %atomicrmw.end
@@ -4073,26 +4073,26 @@ define void @global_atomic_max_i64_noret(ptr addrspace(1) %ptr, i64 %in) {
; SI-NEXT: s_mov_b32 s7, 0xf000
; SI-NEXT: s_mov_b32 s4, s6
; SI-NEXT: s_mov_b32 s5, s6
-; SI-NEXT: buffer_load_dwordx2 v[6:7], v[0:1], s[4:7], 0 addr64
+; SI-NEXT: buffer_load_dwordx2 v[4:5], v[0:1], s[4:7], 0 addr64
; SI-NEXT: s_mov_b64 s[8:9], 0
; SI-NEXT: .LBB80_1: ; %atomicrmw.start
; SI-NEXT: ; =>This Inner Loop Header: Depth=1
; SI-NEXT: s_waitcnt vmcnt(0)
-; SI-NEXT: v_cmp_gt_i64_e32 vcc, v[6:7], v[2:3]
-; SI-NEXT: v_cndmask_b32_e32 v5, v3, v7, vcc
-; SI-NEXT: v_cndmask_b32_e32 v4, v2, v6, vcc
-; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_mov_b32_e32 v11, v7
-; SI-NEXT: v_mov_b32_e32 v10, v6
; SI-NEXT: v_mov_b32_e32 v9, v5
; SI-NEXT: v_mov_b32_e32 v8, v4
-; SI-NEXT: buffer_atomic_cmpswap_x2 v[8:11], v[0:1], s[4:7], 0 addr64 glc
+; SI-NEXT: v_cmp_gt_i64_e32 vcc, v[8:9], v[2:3]
+; SI-NEXT: s_waitcnt expcnt(0)
+; SI-NEXT: v_cndmask_b32_e32 v7, v3, v9, vcc
+; SI-NEXT: v_cndmask_b32_e32 v6, v2, v8, vcc
+; SI-NEXT: v_mov_b32_e32 v4, v6
+; SI-NEXT: v_mov_b32_e32 v5, v7
+; SI-NEXT: v_mov_b32_e32 v6, v8
+; SI-NEXT: v_mov_b32_e32 v7, v9
+; SI-NEXT: buffer_atomic_cmpswap_x2 v[4:7], v[0:1], s[4:7], 0 addr64 glc
; SI-NEXT: s_waitcnt vmcnt(0)
; SI-NEXT: buffer_wbinvl1
-; SI-NEXT: v_cmp_eq_u64_e32 vcc, v[8:9], v[6:7]
+; SI-NEXT: v_cmp_eq_u64_e32 vcc, v[4:5], v[8:9]
; SI-NEXT: s_or_b64 s[8:9], vcc, s[8:9]
-; SI-NEXT: v_mov_b32_e32 v6, v8
-; SI-NEXT: v_mov_b32_e32 v7, v9
; SI-NEXT: s_andn2_b64 exec, exec, s[8:9]
; SI-NEXT: s_cbranch_execnz .LBB80_1
; SI-NEXT: ; %bb.2: ; %atomicrmw.end
@@ -4159,26 +4159,26 @@ define void @global_atomic_max_i64_noret_offset(ptr addrspace(1) %out, i64 %in)
; SI-NEXT: s_mov_b32 s7, 0xf000
; SI-NEXT: s_mov_b32 s4, s6
; SI-NEXT: s_mov_b32 s5, s6
-; SI-NEXT: buffer_load_dwordx2 v[6:7], v[0:1], s[4:7], 0 addr64 offset:32
+; SI-NEXT: buffer_load_dwordx2 v[4:5], v[0:1], s[4:7], 0 addr64 offset:32
; SI-NEXT: s_mov_b64 s[8:9], 0
; SI-NEXT: .LBB81_1: ; %atomicrmw.start
; SI-NEXT: ; =>This Inner Loop Header: Depth=1
; SI-NEXT: s_waitcnt vmcnt(0)
-; SI-NEXT: v_cmp_gt_i64_e32 vcc, v[6:7], v[2:3]
-; SI-NEXT: v_cndmask_b32_e32 v5, v3, v7, vcc
-; SI-NEXT: v_cndmask_b32_e32 v4, v2, v6, vcc
-; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_mov_b32_e32 v11, v7
-; SI-NEXT: v_mov_b32_e32 v10, v6
; SI-NEXT: v_mov_b32_e32 v9, v5
; SI-NEXT: v_mov_b32_e32 v8, v4
-; SI-NEXT: buffer_atomic_cmpswap_x2 v[8:11], v[0:1], s[4:7], 0 addr64 offset:32 glc
+; SI-NEXT: v_cmp_gt_i64_e32 vcc, v[8:9], v[2:3]
+; SI-NEXT: s_waitcnt expcnt(0)
+; SI-NEXT: v_cndmask_b32_e32 v7, v3, v9, vcc
+; SI-NEXT: v_cndmask_b32_e32 v6, v2, v8, vcc
+; SI-NEXT: v_mov_b32_e32 v4, v6
+; SI-NEXT: v_mov_b32_e32 v5, v7
+; SI-NEXT: v_mov_b32_e32 v6, v8
+; SI-NEXT: v_mov_b32_e32 v7, v9
+; SI-NEXT: buffer_atomic_cmpswap_x2 v[4:7], v[0:1], s[4:7], 0 addr64 offset:32 glc
; SI-NEXT: s_waitcnt vmcnt(0)
; SI-NEXT: buffer_wbinvl1
-; SI-NEXT: v_cmp_eq_u64_e32 vcc, v[8:9], v[6:7]
+; SI-NEXT: v_cmp_eq_u64_e32 vcc, v[4:5], v[8:9]
; SI-NEXT: s_or_b64 s[8:9], vcc, s[8:9]
-; SI-NEXT: v_mov_b32_e32 v6, v8
-; SI-NEXT: v_mov_b32_e32 v7, v9
; SI-NEXT: s_andn2_b64 exec, exec, s[8:9]
; SI-NEXT: s_cbranch_execnz .LBB81_1
; SI-NEXT: ; %bb.2: ; %atomicrmw.end
@@ -4443,28 +4443,28 @@ define amdgpu_gfx void @global_atomic_max_i64_noret_scalar(ptr addrspace(1) inre
; SI-NEXT: s_mov_b32 s34, s6
; SI-NEXT: s_mov_b32 s7, 0xf000
; SI-NEXT: s_mov_b32 s6, -1
-; SI-NEXT: buffer_load_dwordx2 v[2:3], off, s[4:7], 0
+; SI-NEXT: buffer_load_dwordx2 v[0:1], off, s[4:7], 0
; SI-NEXT: s_mov_b64 s[36:37], 0
; SI-NEXT: v_mov_b32_e32 v4, s35
; SI-NEXT: v_mov_b32_e32 v5, s34
; SI-NEXT: .LBB84_1: ; %atomicrmw.start
; SI-NEXT: ; =>This Inner Loop Header: Depth=1
; SI-NEXT: s_waitcnt vmcnt(0)
-; SI-NEXT: v_cmp_lt_i64_e32 vcc, s[34:35], v[2:3]
-; SI-NEXT: v_cndmask_b32_e32 v1, v4, v3, vcc
-; SI-NEXT: v_cndmask_b32_e32 v0, v5, v2, vcc
+; SI-NEXT: v_mov_b32_e32 v9, v1
+; SI-NEXT: v_mov_b32_e32 v8, v0
+; SI-NEXT: v_cmp_lt_i64_e32 vcc, s[34:35], v[8:9]
+; SI-NEXT: v_cndmask_b32_e32 v7, v4, v9, vcc
+; SI-NEXT: v_cndmask_b32_e32 v6, v5, v8, vcc
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_mov_b32_e32 v9, v3
-; SI-NEXT: v_mov_b32_e32 v8, v2
-; SI-NEXT: v_mov_b32_e32 v7, v1
-; SI-NEXT: v_mov_b32_e32 v6, v0
-; SI-NEXT: buffer_atomic_cmpswap_x2 v[6:9], off, s[4:7], 0 glc
+; SI-NEXT: v_mov_b32_e32 v0, v6
+; SI-NEXT: v_mov_b32_e32 v1, v7
+; SI-NEXT: v_mov_b32_e32 v2, v8
+; SI-NEXT: v_mov_b32_e32 v3, v9
+; SI-NEXT: buffer_atomic_cmpswap_x2 v[0:3], off, s[4:7], 0 glc
; SI-NEXT: s_waitcnt vmcnt(0)
; SI-NEXT: buffer_wbinvl1
-; SI-NEXT: v_cmp_eq_u64_e32 vcc, v[6:7], v[2:3]
+; SI-NEXT: v_cmp_eq_u64_e32 vcc, v[0:1], v[8:9]
; SI-NEXT: s_or_b64 s[36:37], vcc, s[36:37]
-; SI-NEXT: v_mov_b32_e32 v2, v6
-; SI-NEXT: v_mov_b32_e32 v3, v7
; SI-NEXT: s_andn2_b64 exec, exec, s[36:37]
; SI-NEXT: s_cbranch_execnz .LBB84_1
; SI-NEXT: ; %bb.2: ; %atomicrmw.end
@@ -4551,28 +4551,28 @@ define amdgpu_gfx void @global_atomic_max_i64_noret_offset_scalar(ptr addrspace(
; SI-NEXT: s_mov_b32 s34, s6
; SI-NEXT: s_mov_b32 s7, 0xf000
; SI-NEXT: s_mov_b32 s6, -1
-; SI-NEXT: buffer_load_dwordx2 v[2:3], off, s[4:7], 0 offset:32
+; SI-NEXT: buffer_load_dwordx2 v[0:1], off, s[4:7], 0 offset:32
; SI-NEXT: s_mov_b64 s[36:37], 0
; SI-NEXT: v_mov_b32_e32 v4, s35
; SI-NEXT: v_mov_b32_e32 v5, s34
; SI-NEXT: .LBB85_1: ; %atomicrmw.start
; SI-NEXT: ; =>This Inner Loop Header: Depth=1
; SI-NEXT: s_waitcnt vmcnt(0)
-; SI-NEXT: v_cmp_lt_i64_e32 vcc, s[34:35], v[2:3]
-; SI-NEXT: v_cndmask_b32_e32 v1, v4, v3, vcc
-; SI-NEXT: v_cndmask_b32_e32 v0, v5, v2, vcc
+; SI-NEXT: v_mov_b32_e32 v9, v1
+; SI-NEXT: v_mov_b32_e32 v8, v0
+; SI-NEXT: v_cmp_lt_i64_e32 vcc, s[34:35], v[8:9]
+; SI-NEXT: v_cndmask_b32_e32 v7, v4, v9, vcc
+; SI-NEXT: v_cndmask_b32_e32 v6, v5, v8, vcc
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_mov_b32_e32 v9, v3
-; SI-NEXT: v_mov_b32_e32 v8, v2
-; SI-NEXT: v_mov_b32_e32 v7, v1
-; SI-NEXT: v_mov_b32_e32 v6, v0
-; SI-NEXT: buffer_atomic_cmpswap_x2 v[6:9], off, s[4:7], 0 offset:32 glc
+; SI-NEXT: v_mov_b32_e32 v0, v6
+; SI-NEXT: v_mov_b32_e32 v1, v7
+; SI-NEXT: v_mov_b32_e32 v2, v8
+; SI-NEXT: v_mov_b32_e32 v3, v9
+; SI-NEXT: buffer_atomic_cmpswap_x2 v[0:3], off, s[4:7], 0 offset:32 glc
; SI-NEXT: s_waitcnt vmcnt(0)
; SI-NEXT: buffer_wbinvl1
-; SI-NEXT: v_cmp_eq_u64_e32 vcc, v[6:7], v[2:3]
+; SI-NEXT: v_cmp_eq_u64_e32 vcc, v[0:1], v[8:9]
; SI-NEXT: s_or_b64 s[36:37], vcc, s[36:37]
-; SI-NEXT: v_mov_b32_e32 v2, v6
-; SI-NEXT: v_mov_b32_e32 v3, v7
; SI-NEXT: s_andn2_b64 exec, exec, s[36:37]
; SI-NEXT: s_cbranch_execnz .LBB85_1
; SI-NEXT: ; %bb.2: ; %atomicrmw.end
@@ -4878,26 +4878,26 @@ define amdgpu_kernel void @atomic_max_i64_addr64_offset(ptr addrspace(1) %out, i
; SI-NEXT: v_mov_b32_e32 v4, s3
; SI-NEXT: v_mov_b32_e32 v5, s2
; SI-NEXT: s_waitcnt lgkmcnt(0)
-; SI-NEXT: v_mov_b32_e32 v2, s8
-; SI-NEXT: v_mov_b32_e32 v3, s9
+; SI-NEXT: v_mov_b32_e32 v0, s8
+; SI-NEXT: v_mov_b32_e32 v1, s9
; SI-NEXT: s_mov_b32 s6, -1
; SI-NEXT: .LBB88_1: ; %atomicrmw.start
; SI-NEXT: ; =>This Inner Loop Header: Depth=1
-; SI-NEXT: v_cmp_lt_i64_e32 vcc, s[2:3], v[2:3]
-; SI-NEXT: v_cndmask_b32_e32 v1, v4, v3, vcc
-; SI-NEXT: v_cndmask_b32_e32 v0, v5, v2, vcc
+; SI-NEXT: v_mov_b32_e32 v9, v1
+; SI-NEXT: v_mov_b32_e32 v8, v0
+; SI-NEXT: v_cmp_lt_i64_e32 vcc, s[2:3], v[8:9]
+; SI-NEXT: v_cndmask_b32_e32 v7, v4, v9, vcc
+; SI-NEXT: v_cndmask_b32_e32 v6, v5, v8, vcc
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_mov_b32_e32 v9, v3
-; SI-NEXT: v_mov_b32_e32 v8, v2
-; SI-NEXT: v_mov_b32_e32 v7, v1
-; SI-NEXT: v_mov_b32_e32 v6, v0
-; SI-NEXT: buffer_atomic_cmpswap_x2 v[6:9], off, s[4:7], 0 offset:32 glc
+; SI-NEXT: v_mov_b32_e32 v0, v6
+; SI-NEXT: v_mov_b32_e32 v1, v7
+; SI-NEXT: v_mov_b32_e32 v2, v8
+; SI-NEXT: v_mov_b32_e32 v3, v9
+; SI-NEXT: buffer_atomic_cmpswap_x2 v[0:3], off, s[4:7], 0 offset:32 glc
; SI-NEXT: s_waitcnt vmcnt(0)
; SI-NEXT: buffer_wbinvl1
-; SI-NEXT: v_cmp_eq_u64_e32 vcc, v[6:7], v[2:3]
+; SI-NEXT: v_cmp_eq_u64_e32 vcc, v[0:1], v[8:9]
; SI-NEXT: s_or_b64 s[0:1], vcc, s[0:1]
-; SI-NEXT: v_mov_b32_e32 v2, v6
-; SI-NEXT: v_mov_b32_e32 v3, v7
; SI-NEXT: s_andn2_b64 exec, exec, s[0:1]
; SI-NEXT: s_cbranch_execnz .LBB88_1
; SI-NEXT: ; %bb.2: ; %atomicrmw.end
@@ -4989,29 +4989,29 @@ define amdgpu_kernel void @atomic_max_i64_ret_addr64_offset(ptr addrspace(1) %ou
; SI-NEXT: s_load_dwordx2 s[6:7], s[8:9], 0x8
; SI-NEXT: s_mov_b64 s[0:1], 0
; SI-NEXT: s_mov_b32 s11, 0xf000
-; SI-NEXT: v_mov_b32_e32 v8, s5
-; SI-NEXT: v_mov_b32_e32 v9, s4
+; SI-NEXT: v_mov_b32_e32 v4, s5
+; SI-NEXT: v_mov_b32_e32 v5, s4
; SI-NEXT: s_waitcnt lgkmcnt(0)
-; SI-NEXT: v_mov_b32_e32 v2, s6
-; SI-NEXT: v_mov_b32_e32 v3, s7
+; SI-NEXT: v_mov_b32_e32 v0, s6
+; SI-NEXT: v_mov_b32_e32 v1, s7
; SI-NEXT: s_mov_b32 s10, -1
; SI-NEXT: .LBB89_1: ; %atomicrmw.start
; SI-NEXT: ; =>This Inner Loop Header: Depth=1
-; SI-NEXT: v_cmp_lt_i64_e32 vcc, s[4:5], v[2:3]
-; SI-NEXT: v_cndmask_b32_e32 v1, v8, v3, vcc
-; SI-NEXT: v_cndmask_b32_e32 v0, v9, v2, vcc
+; SI-NEXT: v_mov_b32_e32 v9, v1
+; SI-NEXT: v_mov_b32_e32 v8, v0
+; SI-NEXT: v_cmp_lt_i64_e32 vcc, s[4:5], v[8:9]
+; SI-NEXT: v_cndmask_b32_e32 v7, v4, v9, vcc
+; SI-NEXT: v_cndmask_b32_e32 v6, v5, v8, vcc
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_mov_b32_e32 v7, v3
-; SI-NEXT: v_mov_b32_e32 v6, v2
-; SI-NEXT: v_mov_b32_e32 v5, v1
-; SI-NEXT: v_mov_b32_e32 v4, v0
-; SI-NEXT: buffer_atomic_cmpswap_x2 v[4:7], off, s[8:11], 0 offset:32 glc
+; SI-NEXT: v_mov_b32_e32 v0, v6
+; SI-NEXT: v_mov_b32_e32 v1, v7
+; SI-NEXT: v_mov_b32_e32 v2, v8
+; SI-NEXT: v_mov_b32_e32 v3, v9
+; SI-NEXT: buffer_atomic_cmpswap_x2 v[0:3], off, s[8:11], 0 offset:32 glc
; SI-NEXT: s_waitcnt vmcnt(0)
; SI-NEXT: buffer_wbinvl1
-; SI-NEXT: v_cmp_eq_u64_e32 vcc, v[4:5], v[2:3]
+; SI-NEXT: v_cmp_eq_u64_e32 vcc, v[0:1], v[8:9]
; SI-NEXT: s_or_b64 s[0:1], vcc, s[0:1]
-; SI-NEXT: v_mov_b32_e32 v2, v4
-; SI-NEXT: v_mov_b32_e32 v3, v5
; SI-NEXT: s_andn2_b64 exec, exec, s[0:1]
; SI-NEXT: s_cbranch_execnz .LBB89_1
; SI-NEXT: ; %bb.2: ; %atomicrmw.end
@@ -5020,7 +5020,7 @@ define amdgpu_kernel void @atomic_max_i64_ret_addr64_offset(ptr addrspace(1) %ou
; SI-NEXT: s_mov_b32 s6, -1
; SI-NEXT: s_mov_b32 s4, s2
; SI-NEXT: s_mov_b32 s5, s3
-; SI-NEXT: buffer_store_dwordx2 v[4:5], off, s[4:7], 0
+; SI-NEXT: buffer_store_dwordx2 v[0:1], off, s[4:7], 0
; SI-NEXT: s_endpgm
;
; VI-LABEL: atomic_max_i64_ret_addr64_offset:
@@ -5119,26 +5119,26 @@ define amdgpu_kernel void @atomic_max_i64_addr64(ptr addrspace(1) %out, i64 %in,
; SI-NEXT: v_mov_b32_e32 v4, s3
; SI-NEXT: v_mov_b32_e32 v5, s2
; SI-NEXT: s_waitcnt lgkmcnt(0)
-; SI-NEXT: v_mov_b32_e32 v2, s8
-; SI-NEXT: v_mov_b32_e32 v3, s9
+; SI-NEXT: v_mov_b32_e32 v0, s8
+; SI-NEXT: v_mov_b32_e32 v1, s9
; SI-NEXT: s_mov_b32 s6, -1
; SI-NEXT: .LBB90_1: ; %atomicrmw.start
; SI-NEXT: ; =>This Inner Loop Header: Depth=1
-; SI-NEXT: v_cmp_lt_i64_e32 vcc, s[2:3], v[2:3]
-; SI-NEXT: v_cndmask_b32_e32 v1, v4, v3, vcc
-; SI-NEXT: v_cndmask_b32_e32 v0, v5, v2, vcc
+; SI-NEXT: v_mov_b32_e32 v9, v1
+; SI-NEXT: v_mov_b32_e32 v8, v0
+; SI-NEXT: v_cmp_lt_i64_e32 vcc, s[2:3], v[8:9]
+; SI-NEXT: v_cndmask_b32_e32 v7, v4, v9, vcc
+; SI-NEXT: v_cndmask_b32_e32 v6, v5, v8, vcc
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_mov_b32_e32 v9, v3
-; SI-NEXT: v_mov_b32_e32 v8, v2
-; SI-NEXT: v_mov_b32_e32 v7, v1
-; SI-NEXT: v_mov_b32_e32 v6, v0
-; SI-NEXT: buffer_atomic_cmpswap_x2 v[6:9], off, s[4:7], 0 glc
+; SI-NEXT: v_mov_b32_e32 v0, v6
+; SI-NEXT: v_mov_b32_e32 v1, v7
+; SI-NEXT: v_mov_b32_e32 v2, v8
+; SI-NEXT: v_mov_b32_e32 v3, v9
+; SI-NEXT: buffer_atomic_cmpswap_x2 v[0:3], off, s[4:7], 0 glc
; SI-NEXT: s_waitcnt vmcnt(0)
; SI-NEXT: buffer_wbinvl1
-; SI-NEXT: v_cmp_eq_u64_e32 vcc, v[6:7], v[2:3]
+; SI-NEXT: v_cmp_eq_u64_e32 vcc, v[0:1], v[8:9]
; SI-NEXT: s_or_b64 s[0:1], vcc, s[0:1]
-; SI-NEXT: v_mov_b32_e32 v2, v6
-; SI-NEXT: v_mov_b32_e32 v3, v7
; SI-NEXT: s_andn2_b64 exec, exec, s[0:1]
; SI-NEXT: s_cbranch_execnz .LBB90_1
; SI-NEXT: ; %bb.2: ; %atomicrmw.end
@@ -5227,29 +5227,29 @@ define amdgpu_kernel void @atomic_max_i64_ret_addr64(ptr addrspace(1) %out, ptr
; SI-NEXT: s_load_dwordx2 s[6:7], s[8:9], 0x0
; SI-NEXT: s_mov_b64 s[0:1], 0
; SI-NEXT: s_mov_b32 s11, 0xf000
-; SI-NEXT: v_mov_b32_e32 v8, s5
-; SI-NEXT: v_mov_b32_e32 v9, s4
+; SI-NEXT: v_mov_b32_e32 v4, s5
+; SI-NEXT: v_mov_b32_e32 v5, s4
; SI-NEXT: s_waitcnt lgkmcnt(0)
-; SI-NEXT: v_mov_b32_e32 v2, s6
-; SI-NEXT: v_mov_b32_e32 v3, s7
+; SI-NEXT: v_mov_b32_e32 v0, s6
+; SI-NEXT: v_mov_b32_e32 v1, s7
; SI-NEXT: s_mov_b32 s10, -1
; SI-NEXT: .LBB91_1: ; %atomicrmw.start
; SI-NEXT: ; =>This Inner Loop Header: Depth=1
-; SI-NEXT: v_cmp_lt_i64_e32 vcc, s[4:5], v[2:3]
-; SI-NEXT: v_cndmask_b32_e32 v1, v8, v3, vcc
-; SI-NEXT: v_cndmask_b32_e32 v0, v9, v2, vcc
+; SI-NEXT: v_mov_b32_e32 v9, v1
+; SI-NEXT: v_mov_b32_e32 v8, v0
+; SI-NEXT: v_cmp_lt_i64_e32 vcc, s[4:5], v[8:9]
+; SI-NEXT: v_cndmask_b32_e32 v7, v4, v9, vcc
+; SI-NEXT: v_cndmask_b32_e32 v6, v5, v8, vcc
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_mov_b32_e32 v7, v3
-; SI-NEXT: v_mov_b32_e32 v6, v2
-; SI-NEXT: v_mov_b32_e32 v5, v1
-; SI-NEXT: v_mov_b32_e32 v4, v0
-; SI-NEXT: buffer_atomic_cmpswap_x2 v[4:7], off, s[8:11], 0 glc
+; SI-NEXT: v_mov_b32_e32 v0, v6
+; SI-NEXT: v_mov_b32_e32 v1, v7
+; SI-NEXT: v_mov_b32_e32 v2, v8
+; SI-NEXT: v_mov_b32_e32 v3, v9
+; SI-NEXT: buffer_atomic_cmpswap_x2 v[0:3], off, s[8:11], 0 glc
; SI-NEXT: s_waitcnt vmcnt(0)
; SI-NEXT: buffer_wbinvl1
-; SI-NEXT: v_cmp_eq_u64_e32 vcc, v[4:5], v[2:3]
+; SI-NEXT: v_cmp_eq_u64_e32 vcc, v[0:1], v[8:9]
; SI-NEXT: s_or_b64 s[0:1], vcc, s[0:1]
-; SI-NEXT: v_mov_b32_e32 v2, v4
-; SI-NEXT: v_mov_b32_e32 v3, v5
; SI-NEXT: s_andn2_b64 exec, exec, s[0:1]
; SI-NEXT: s_cbranch_execnz .LBB91_1
; SI-NEXT: ; %bb.2: ; %atomicrmw.end
@@ -5258,7 +5258,7 @@ define amdgpu_kernel void @atomic_max_i64_ret_addr64(ptr addrspace(1) %out, ptr
; SI-NEXT: s_mov_b32 s6, -1
; SI-NEXT: s_mov_b32 s4, s2
; SI-NEXT: s_mov_b32 s5, s3
-; SI-NEXT: buffer_store_dwordx2 v[4:5], off, s[4:7], 0
+; SI-NEXT: buffer_store_dwordx2 v[0:1], off, s[4:7], 0
; SI-NEXT: s_endpgm
;
; VI-LABEL: atomic_max_i64_ret_addr64:
@@ -5347,26 +5347,26 @@ define void @global_atomic_max_i64_noret_offset__amdgpu_no_remote_memory(ptr add
; SI-NEXT: s_mov_b32 s7, 0xf000
; SI-NEXT: s_mov_b32 s4, s6
; SI-NEXT: s_mov_b32 s5, s6
-; SI-NEXT: buffer_load_dwordx2 v[6:7], v[0:1], s[4:7], 0 addr64 offset:32
+; SI-NEXT: buffer_load_dwordx2 v[4:5], v[0:1], s[4:7], 0 addr64 offset:32
; SI-NEXT: s_mov_b64 s[8:9], 0
; SI-NEXT: .LBB92_1: ; %atomicrmw.start
; SI-NEXT: ; =>This Inner Loop Header: Depth=1
; SI-NEXT: s_waitcnt vmcnt(0)
-; SI-NEXT: v_cmp_gt_i64_e32 vcc, v[6:7], v[2:3]
-; SI-NEXT: v_cndmask_b32_e32 v5, v3, v7, vcc
-; SI-NEXT: v_cndmask_b32_e32 v4, v2, v6, vcc
-; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_mov_b32_e32 v11, v7
-; SI-NEXT: v_mov_b32_e32 v10, v6
; SI-NEXT: v_mov_b32_e32 v9, v5
; SI-NEXT: v_mov_b32_e32 v8, v4
-; SI-NEXT: buffer_atomic_cmpswap_x2 v[8:11], v[0:1], s[4:7], 0 addr64 offset:32 glc
+; SI-NEXT: v_cmp_gt_i64_e32 vcc, v[8:9], v[2:3]
+; SI-NEXT: s_waitcnt expcnt(0)
+; SI-NEXT: v_cndmask_b32_e32 v7, v3, v9, vcc
+; SI-NEXT: v_cndmask_b32_e32 v6, v2, v8, vcc
+; SI-NEXT: v_mov_b32_e32 v4, v6
+; SI-NEXT: v_mov_b32_e32 v5, v7
+; SI-NEXT: v_mov_b32_e32 v6, v8
+; SI-NEXT: v_mov_b32_e32 v7, v9
+; SI-NEXT: buffer_atomic_cmpswap_x2 v[4:7], v[0:1], s[4:7], 0 addr64 offset:32 glc
; SI-NEXT: s_waitcnt vmcnt(0)
; SI-NEXT: buffer_wbinvl1
-; SI-NEXT: v_cmp_eq_u64_e32 vcc, v[8:9], v[6:7]
+; SI-NEXT: v_cmp_eq_u64_e32 vcc, v[4:5], v[8:9]
; SI-NEXT: s_or_b64 s[8:9], vcc, s[8:9]
-; SI-NEXT: v_mov_b32_e32 v6, v8
-; SI-NEXT: v_mov_b32_e32 v7, v9
; SI-NEXT: s_andn2_b64 exec, exec, s[8:9]
; SI-NEXT: s_cbranch_execnz .LBB92_1
; SI-NEXT: ; %bb.2: ; %atomicrmw.end
@@ -5535,26 +5535,26 @@ define void @global_atomic_umax_i64_noret(ptr addrspace(1) %ptr, i64 %in) {
; SI-NEXT: s_mov_b32 s7, 0xf000
; SI-NEXT: s_mov_b32 s4, s6
; SI-NEXT: s_mov_b32 s5, s6
-; SI-NEXT: buffer_load_dwordx2 v[6:7], v[0:1], s[4:7], 0 addr64
+; SI-NEXT: buffer_load_dwordx2 v[4:5], v[0:1], s[4:7], 0 addr64
; SI-NEXT: s_mov_b64 s[8:9], 0
; SI-NEXT: .LBB94_1: ; %atomicrmw.start
; SI-NEXT: ; =>This Inner Loop Header: Depth=1
; SI-NEXT: s_waitcnt vmcnt(0)
-; SI-NEXT: v_cmp_gt_u64_e32 vcc, v[6:7], v[2:3]
-; SI-NEXT: v_cndmask_b32_e32 v5, v3, v7, vcc
-; SI-NEXT: v_cndmask_b32_e32 v4, v2, v6, vcc
-; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_mov_b32_e32 v11, v7
-; SI-NEXT: v_mov_b32_e32 v10, v6
; SI-NEXT: v_mov_b32_e32 v9, v5
; SI-NEXT: v_mov_b32_e32 v8, v4
-; SI-NEXT: buffer_atomic_cmpswap_x2 v[8:11], v[0:1], s[4:7], 0 addr64 glc
+; SI-NEXT: v_cmp_gt_u64_e32 vcc, v[8:9], v[2:3]
+; SI-NEXT: s_waitcnt expcnt(0)
+; SI-NEXT: v_cndmask_b32_e32 v7, v3, v9, vcc
+; SI-NEXT: v_cndmask_b32_e32 v6, v2, v8, vcc
+; SI-NEXT: v_mov_b32_e32 v4, v6
+; SI-NEXT: v_mov_b32_e32 v5, v7
+; SI-NEXT: v_mov_b32_e32 v6, v8
+; SI-NEXT: v_mov_b32_e32 v7, v9
+; SI-NEXT: buffer_atomic_cmpswap_x2 v[4:7], v[0:1], s[4:7], 0 addr64 glc
; SI-NEXT: s_waitcnt vmcnt(0)
; SI-NEXT: buffer_wbinvl1
-; SI-NEXT: v_cmp_eq_u64_e32 vcc, v[8:9], v[6:7]
+; SI-NEXT: v_cmp_eq_u64_e32 vcc, v[4:5], v[8:9]
; SI-NEXT: s_or_b64 s[8:9], vcc, s[8:9]
-; SI-NEXT: v_mov_b32_e32 v6, v8
-; SI-NEXT: v_mov_b32_e32 v7, v9
; SI-NEXT: s_andn2_b64 exec, exec, s[8:9]
; SI-NEXT: s_cbranch_execnz .LBB94_1
; SI-NEXT: ; %bb.2: ; %atomicrmw.end
@@ -5621,26 +5621,26 @@ define void @global_atomic_umax_i64_noret_offset(ptr addrspace(1) %out, i64 %in)
; SI-NEXT: s_mov_b32 s7, 0xf000
; SI-NEXT: s_mov_b32 s4, s6
; SI-NEXT: s_mov_b32 s5, s6
-; SI-NEXT: buffer_load_dwordx2 v[6:7], v[0:1], s[4:7], 0 addr64 offset:32
+; SI-NEXT: buffer_load_dwordx2 v[4:5], v[0:1], s[4:7], 0 addr64 offset:32
; SI-NEXT: s_mov_b64 s[8:9], 0
; SI-NEXT: .LBB95_1: ; %atomicrmw.start
; SI-NEXT: ; =>This Inner Loop Header: Depth=1
; SI-NEXT: s_waitcnt vmcnt(0)
-; SI-NEXT: v_cmp_gt_u64_e32 vcc, v[6:7], v[2:3]
-; SI-NEXT: v_cndmask_b32_e32 v5, v3, v7, vcc
-; SI-NEXT: v_cndmask_b32_e32 v4, v2, v6, vcc
-; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_mov_b32_e32 v11, v7
-; SI-NEXT: v_mov_b32_e32 v10, v6
; SI-NEXT: v_mov_b32_e32 v9, v5
; SI-NEXT: v_mov_b32_e32 v8, v4
-; SI-NEXT: buffer_atomic_cmpswap_x2 v[8:11], v[0:1], s[4:7], 0 addr64 offset:32 glc
+; SI-NEXT: v_cmp_gt_u64_e32 vcc, v[8:9], v[2:3]
+; SI-NEXT: s_waitcnt expcnt(0)
+; SI-NEXT: v_cndmask_b32_e32 v7, v3, v9, vcc
+; SI-NEXT: v_cndmask_b32_e32 v6, v2, v8, vcc
+; SI-NEXT: v_mov_b32_e32 v4, v6
+; SI-NEXT: v_mov_b32_e32 v5, v7
+; SI-NEXT: v_mov_b32_e32 v6, v8
+; SI-NEXT: v_mov_b32_e32 v7, v9
+; SI-NEXT: buffer_atomic_cmpswap_x2 v[4:7], v[0:1], s[4:7], 0 addr64 offset:32 glc
; SI-NEXT: s_waitcnt vmcnt(0)
; SI-NEXT: buffer_wbinvl1
-; SI-NEXT: v_cmp_eq_u64_e32 vcc, v[8:9], v[6:7]
+; SI-NEXT: v_cmp_eq_u64_e32 vcc, v[4:5], v[8:9]
; SI-NEXT: s_or_b64 s[8:9], vcc, s[8:9]
-; SI-NEXT: v_mov_b32_e32 v6, v8
-; SI-NEXT: v_mov_b32_e32 v7, v9
; SI-NEXT: s_andn2_b64 exec, exec, s[8:9]
; SI-NEXT: s_cbranch_execnz .LBB95_1
; SI-NEXT: ; %bb.2: ; %atomicrmw.end
@@ -5905,28 +5905,28 @@ define amdgpu_gfx void @global_atomic_umax_i64_noret_scalar(ptr addrspace(1) inr
; SI-NEXT: s_mov_b32 s34, s6
; SI-NEXT: s_mov_b32 s7, 0xf000
; SI-NEXT: s_mov_b32 s6, -1
-; SI-NEXT: buffer_load_dwordx2 v[2:3], off, s[4:7], 0
+; SI-NEXT: buffer_load_dwordx2 v[0:1], off, s[4:7], 0
; SI-NEXT: s_mov_b64 s[36:37], 0
; SI-NEXT: v_mov_b32_e32 v4, s35
; SI-NEXT: v_mov_b32_e32 v5, s34
; SI-NEXT: .LBB98_1: ; %atomicrmw.start
; SI-NEXT: ; =>This Inner Loop Header: Depth=1
; SI-NEXT: s_waitcnt vmcnt(0)
-; SI-NEXT: v_cmp_lt_u64_e32 vcc, s[34:35], v[2:3]
-; SI-NEXT: v_cndmask_b32_e32 v1, v4, v3, vcc
-; SI-NEXT: v_cndmask_b32_e32 v0, v5, v2, vcc
+; SI-NEXT: v_mov_b32_e32 v9, v1
+; SI-NEXT: v_mov_b32_e32 v8, v0
+; SI-NEXT: v_cmp_lt_u64_e32 vcc, s[34:35], v[8:9]
+; SI-NEXT: v_cndmask_b32_e32 v7, v4, v9, vcc
+; SI-NEXT: v_cndmask_b32_e32 v6, v5, v8, vcc
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_mov_b32_e32 v9, v3
-; SI-NEXT: v_mov_b32_e32 v8, v2
-; SI-NEXT: v_mov_b32_e32 v7, v1
-; SI-NEXT: v_mov_b32_e32 v6, v0
-; SI-NEXT: buffer_atomic_cmpswap_x2 v[6:9], off, s[4:7], 0 glc
+; SI-NEXT: v_mov_b32_e32 v0, v6
+; SI-NEXT: v_mov_b32_e32 v1, v7
+; SI-NEXT: v_mov_b32_e32 v2, v8
+; SI-NEXT: v_mov_b32_e32 v3, v9
+; SI-NEXT: buffer_atomic_cmpswap_x2 v[0:3], off, s[4:7], 0 glc
; SI-NEXT: s_waitcnt vmcnt(0)
; SI-NEXT: buffer_wbinvl1
-; SI-NEXT: v_cmp_eq_u64_e32 vcc, v[6:7], v[2:3]
+; SI-NEXT: v_cmp_eq_u64_e32 vcc, v[0:1], v[8:9]
; SI-NEXT: s_or_b64 s[36:37], vcc, s[36:37]
-; SI-NEXT: v_mov_b32_e32 v2, v6
-; SI-NEXT: v_mov_b32_e32 v3, v7
; SI-NEXT: s_andn2_b64 exec, exec, s[36:37]
; SI-NEXT: s_cbranch_execnz .LBB98_1
; SI-NEXT: ; %bb.2: ; %atomicrmw.end
@@ -6013,28 +6013,28 @@ define amdgpu_gfx void @global_atomic_umax_i64_noret_offset_scalar(ptr addrspace
; SI-NEXT: s_mov_b32 s34, s6
; SI-NEXT: s_mov_b32 s7, 0xf000
; SI-NEXT: s_mov_b32 s6, -1
-; SI-NEXT: buffer_load_dwordx2 v[2:3], off, s[4:7], 0 offset:32
+; SI-NEXT: buffer_load_dwordx2 v[0:1], off, s[4:7], 0 offset:32
; SI-NEXT: s_mov_b64 s[36:37], 0
; SI-NEXT: v_mov_b32_e32 v4, s35
; SI-NEXT: v_mov_b32_e32 v5, s34
; SI-NEXT: .LBB99_1: ; %atomicrmw.start
; SI-NEXT: ; =>This Inner Loop Header: Depth=1
; SI-NEXT: s_waitcnt vmcnt(0)
-; SI-NEXT: v_cmp_lt_u64_e32 vcc, s[34:35], v[2:3]
-; SI-NEXT: v_cndmask_b32_e32 v1, v4, v3, vcc
-; SI-NEXT: v_cndmask_b32_e32 v0, v5, v2, vcc
+; SI-NEXT: v_mov_b32_e32 v9, v1
+; SI-NEXT: v_mov_b32_e32 v8, v0
+; SI-NEXT: v_cmp_lt_u64_e32 vcc, s[34:35], v[8:9]
+; SI-NEXT: v_cndmask_b32_e32 v7, v4, v9, vcc
+; SI-NEXT: v_cndmask_b32_e32 v6, v5, v8, vcc
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_mov_b32_e32 v9, v3
-; SI-NEXT: v_mov_b32_e32 v8, v2
-; SI-NEXT: v_mov_b32_e32 v7, v1
-; SI-NEXT: v_mov_b32_e32 v6, v0
-; SI-NEXT: buffer_atomic_cmpswap_x2 v[6:9], off, s[4:7], 0 offset:32 glc
+; SI-NEXT: v_mov_b32_e32 v0, v6
+; SI-NEXT: v_mov_b32_e32 v1, v7
+; SI-NEXT: v_mov_b32_e32 v2, v8
+; SI-NEXT: v_mov_b32_e32 v3, v9
+; SI-NEXT: buffer_atomic_cmpswap_x2 v[0:3], off, s[4:7], 0 offset:32 glc
; SI-NEXT: s_waitcnt vmcnt(0)
; SI-NEXT: buffer_wbinvl1
-; SI-NEXT: v_cmp_eq_u64_e32 vcc, v[6:7], v[2:3]
+; SI-NEXT: v_cmp_eq_u64_e32 vcc, v[0:1], v[8:9]
; SI-NEXT: s_or_b64 s[36:37], vcc, s[36:37]
-; SI-NEXT: v_mov_b32_e32 v2, v6
-; SI-NEXT: v_mov_b32_e32 v3, v7
; SI-NEXT: s_andn2_b64 exec, exec, s[36:37]
; SI-NEXT: s_cbranch_execnz .LBB99_1
; SI-NEXT: ; %bb.2: ; %atomicrmw.end
@@ -6340,26 +6340,26 @@ define amdgpu_kernel void @atomic_umax_i64_addr64_offset(ptr addrspace(1) %out,
; SI-NEXT: v_mov_b32_e32 v4, s3
; SI-NEXT: v_mov_b32_e32 v5, s2
; SI-NEXT: s_waitcnt lgkmcnt(0)
-; SI-NEXT: v_mov_b32_e32 v2, s8
-; SI-NEXT: v_mov_b32_e32 v3, s9
+; SI-NEXT: v_mov_b32_e32 v0, s8
+; SI-NEXT: v_mov_b32_e32 v1, s9
; SI-NEXT: s_mov_b32 s6, -1
; SI-NEXT: .LBB102_1: ; %atomicrmw.start
; SI-NEXT: ; =>This Inner Loop Header: Depth=1
-; SI-NEXT: v_cmp_lt_u64_e32 vcc, s[2:3], v[2:3]
-; SI-NEXT: v_cndmask_b32_e32 v1, v4, v3, vcc
-; SI-NEXT: v_cndmask_b32_e32 v0, v5, v2, vcc
+; SI-NEXT: v_mov_b32_e32 v9, v1
+; SI-NEXT: v_mov_b32_e32 v8, v0
+; SI-NEXT: v_cmp_lt_u64_e32 vcc, s[2:3], v[8:9]
+; SI-NEXT: v_cndmask_b32_e32 v7, v4, v9, vcc
+; SI-NEXT: v_cndmask_b32_e32 v6, v5, v8, vcc
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_mov_b32_e32 v9, v3
-; SI-NEXT: v_mov_b32_e32 v8, v2
-; SI-NEXT: v_mov_b32_e32 v7, v1
-; SI-NEXT: v_mov_b32_e32 v6, v0
-; SI-NEXT: buffer_atomic_cmpswap_x2 v[6:9], off, s[4:7], 0 offset:32 glc
+; SI-NEXT: v_mov_b32_e32 v0, v6
+; SI-NEXT: v_mov_b32_e32 v1, v7
+; SI-NEXT: v_mov_b32_e32 v2, v8
+; SI-NEXT: v_mov_b32_e32 v3, v9
+; SI-NEXT: buffer_atomic_cmpswap_x2 v[0:3], off, s[4:7], 0 offset:32 glc
; SI-NEXT: s_waitcnt vmcnt(0)
; SI-NEXT: buffer_wbinvl1
-; SI-NEXT: v_cmp_eq_u64_e32 vcc, v[6:7], v[2:3]
+; SI-NEXT: v_cmp_eq_u64_e32 vcc, v[0:1], v[8:9]
; SI-NEXT: s_or_b64 s[0:1], vcc, s[0:1]
-; SI-NEXT: v_mov_b32_e32 v2, v6
-; SI-NEXT: v_mov_b32_e32 v3, v7
; SI-NEXT: s_andn2_b64 exec, exec, s[0:1]
; SI-NEXT: s_cbranch_execnz .LBB102_1
; SI-NEXT: ; %bb.2: ; %atomicrmw.end
@@ -6451,29 +6451,29 @@ define amdgpu_kernel void @atomic_umax_i64_ret_addr64_offset(ptr addrspace(1) %o
; SI-NEXT: s_load_dwordx2 s[6:7], s[8:9], 0x8
; SI-NEXT: s_mov_b64 s[0:1], 0
; SI-NEXT: s_mov_b32 s11, 0xf000
-; SI-NEXT: v_mov_b32_e32 v8, s5
-; SI-NEXT: v_mov_b32_e32 v9, s4
+; SI-NEXT: v_mov_b32_e32 v4, s5
+; SI-NEXT: v_mov_b32_e32 v5, s4
; SI-NEXT: s_waitcnt lgkmcnt(0)
-; SI-NEXT: v_mov_b32_e32 v2, s6
-; SI-NEXT: v_mov_b32_e32 v3, s7
+; SI-NEXT: v_mov_b32_e32 v0, s6
+; SI-NEXT: v_mov_b32_e32 v1, s7
; SI-NEXT: s_mov_b32 s10, -1
; SI-NEXT: .LBB103_1: ; %atomicrmw.start
; SI-NEXT: ; =>This Inner Loop Header: Depth=1
-; SI-NEXT: v_cmp_lt_u64_e32 vcc, s[4:5], v[2:3]
-; SI-NEXT: v_cndmask_b32_e32 v1, v8, v3, vcc
-; SI-NEXT: v_cndmask_b32_e32 v0, v9, v2, vcc
+; SI-NEXT: v_mov_b32_e32 v9, v1
+; SI-NEXT: v_mov_b32_e32 v8, v0
+; SI-NEXT: v_cmp_lt_u64_e32 vcc, s[4:5], v[8:9]
+; SI-NEXT: v_cndmask_b32_e32 v7, v4, v9, vcc
+; SI-NEXT: v_cndmask_b32_e32 v6, v5, v8, vcc
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_mov_b32_e32 v7, v3
-; SI-NEXT: v_mov_b32_e32 v6, v2
-; SI-NEXT: v_mov_b32_e32 v5, v1
-; SI-NEXT: v_mov_b32_e32 v4, v0
-; SI-NEXT: buffer_atomic_cmpswap_x2 v[4:7], off, s[8:11], 0 offset:32 glc
+; SI-NEXT: v_mov_b32_e32 v0, v6
+; SI-NEXT: v_mov_b32_e32 v1, v7
+; SI-NEXT: v_mov_b32_e32 v2, v8
+; SI-NEXT: v_mov_b32_e32 v3, v9
+; SI-NEXT: buffer_atomic_cmpswap_x2 v[0:3], off, s[8:11], 0 offset:32 glc
; SI-NEXT: s_waitcnt vmcnt(0)
; SI-NEXT: buffer_wbinvl1
-; SI-NEXT: v_cmp_eq_u64_e32 vcc, v[4:5], v[2:3]
+; SI-NEXT: v_cmp_eq_u64_e32 vcc, v[0:1], v[8:9]
; SI-NEXT: s_or_b64 s[0:1], vcc, s[0:1]
-; SI-NEXT: v_mov_b32_e32 v2, v4
-; SI-NEXT: v_mov_b32_e32 v3, v5
; SI-NEXT: s_andn2_b64 exec, exec, s[0:1]
; SI-NEXT: s_cbranch_execnz .LBB103_1
; SI-NEXT: ; %bb.2: ; %atomicrmw.end
@@ -6482,7 +6482,7 @@ define amdgpu_kernel void @atomic_umax_i64_ret_addr64_offset(ptr addrspace(1) %o
; SI-NEXT: s_mov_b32 s6, -1
; SI-NEXT: s_mov_b32 s4, s2
; SI-NEXT: s_mov_b32 s5, s3
-; SI-NEXT: buffer_store_dwordx2 v[4:5], off, s[4:7], 0
+; SI-NEXT: buffer_store_dwordx2 v[0:1], off, s[4:7], 0
; SI-NEXT: s_endpgm
;
; VI-LABEL: atomic_umax_i64_ret_addr64_offset:
@@ -6577,29 +6577,29 @@ define amdgpu_kernel void @atomic_umax_i64_ret_addr64(ptr addrspace(1) %out, ptr
; SI-NEXT: s_load_dwordx2 s[6:7], s[8:9], 0x0
; SI-NEXT: s_mov_b64 s[0:1], 0
; SI-NEXT: s_mov_b32 s11, 0xf000
-; SI-NEXT: v_mov_b32_e32 v8, s5
-; SI-NEXT: v_mov_b32_e32 v9, s4
+; SI-NEXT: v_mov_b32_e32 v4, s5
+; SI-NEXT: v_mov_b32_e32 v5, s4
; SI-NEXT: s_waitcnt lgkmcnt(0)
-; SI-NEXT: v_mov_b32_e32 v2, s6
-; SI-NEXT: v_mov_b32_e32 v3, s7
+; SI-NEXT: v_mov_b32_e32 v0, s6
+; SI-NEXT: v_mov_b32_e32 v1, s7
; SI-NEXT: s_mov_b32 s10, -1
; SI-NEXT: .LBB104_1: ; %atomicrmw.start
; SI-NEXT: ; =>This Inner Loop Header: Depth=1
-; SI-NEXT: v_cmp_lt_u64_e32 vcc, s[4:5], v[2:3]
-; SI-NEXT: v_cndmask_b32_e32 v1, v8, v3, vcc
-; SI-NEXT: v_cndmask_b32_e32 v0, v9, v2, vcc
+; SI-NEXT: v_mov_b32_e32 v9, v1
+; SI-NEXT: v_mov_b32_e32 v8, v0
+; SI-NEXT: v_cmp_lt_u64_e32 vcc, s[4:5], v[8:9]
+; SI-NEXT: v_cndmask_b32_e32 v7, v4, v9, vcc
+; SI-NEXT: v_cndmask_b32_e32 v6, v5, v8, vcc
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_mov_b32_e32 v7, v3
-; SI-NEXT: v_mov_b32_e32 v6, v2
-; SI-NEXT: v_mov_b32_e32 v5, v1
-; SI-NEXT: v_mov_b32_e32 v4, v0
-; SI-NEXT: buffer_atomic_cmpswap_x2 v[4:7], off, s[8:11], 0 glc
+; SI-NEXT: v_mov_b32_e32 v0, v6
+; SI-NEXT: v_mov_b32_e32 v1, v7
+; SI-NEXT: v_mov_b32_e32 v2, v8
+; SI-NEXT: v_mov_b32_e32 v3, v9
+; SI-NEXT: buffer_atomic_cmpswap_x2 v[0:3], off, s[8:11], 0 glc
; SI-NEXT: s_waitcnt vmcnt(0)
; SI-NEXT: buffer_wbinvl1
-; SI-NEXT: v_cmp_eq_u64_e32 vcc, v[4:5], v[2:3]
+; SI-NEXT: v_cmp_eq_u64_e32 vcc, v[0:1], v[8:9]
; SI-NEXT: s_or_b64 s[0:1], vcc, s[0:1]
-; SI-NEXT: v_mov_b32_e32 v2, v4
-; SI-NEXT: v_mov_b32_e32 v3, v5
; SI-NEXT: s_andn2_b64 exec, exec, s[0:1]
; SI-NEXT: s_cbranch_execnz .LBB104_1
; SI-NEXT: ; %bb.2: ; %atomicrmw.end
@@ -6608,7 +6608,7 @@ define amdgpu_kernel void @atomic_umax_i64_ret_addr64(ptr addrspace(1) %out, ptr
; SI-NEXT: s_mov_b32 s6, -1
; SI-NEXT: s_mov_b32 s4, s2
; SI-NEXT: s_mov_b32 s5, s3
-; SI-NEXT: buffer_store_dwordx2 v[4:5], off, s[4:7], 0
+; SI-NEXT: buffer_store_dwordx2 v[0:1], off, s[4:7], 0
; SI-NEXT: s_endpgm
;
; VI-LABEL: atomic_umax_i64_ret_addr64:
@@ -6697,26 +6697,26 @@ define void @global_atomic_umax_i64_noret_offset__amdgpu_no_remote_memory(ptr ad
; SI-NEXT: s_mov_b32 s7, 0xf000
; SI-NEXT: s_mov_b32 s4, s6
; SI-NEXT: s_mov_b32 s5, s6
-; SI-NEXT: buffer_load_dwordx2 v[6:7], v[0:1], s[4:7], 0 addr64 offset:32
+; SI-NEXT: buffer_load_dwordx2 v[4:5], v[0:1], s[4:7], 0 addr64 offset:32
; SI-NEXT: s_mov_b64 s[8:9], 0
; SI-NEXT: .LBB105_1: ; %atomicrmw.start
; SI-NEXT: ; =>This Inner Loop Header: Depth=1
; SI-NEXT: s_waitcnt vmcnt(0)
-; SI-NEXT: v_cmp_gt_u64_e32 vcc, v[6:7], v[2:3]
-; SI-NEXT: v_cndmask_b32_e32 v5, v3, v7, vcc
-; SI-NEXT: v_cndmask_b32_e32 v4, v2, v6, vcc
-; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_mov_b32_e32 v11, v7
-; SI-NEXT: v_mov_b32_e32 v10, v6
; SI-NEXT: v_mov_b32_e32 v9, v5
; SI-NEXT: v_mov_b32_e32 v8, v4
-; SI-NEXT: buffer_atomic_cmpswap_x2 v[8:11], v[0:1], s[4:7], 0 addr64 offset:32 glc
+; SI-NEXT: v_cmp_gt_u64_e32 vcc, v[8:9], v[2:3]
+; SI-NEXT: s_waitcnt expcnt(0)
+; SI-NEXT: v_cndmask_b32_e32 v7, v3, v9, vcc
+; SI-NEXT: v_cndmask_b32_e32 v6, v2, v8, vcc
+; SI-NEXT: v_mov_b32_e32 v4, v6
+; SI-NEXT: v_mov_b32_e32 v5, v7
+; SI-NEXT: v_mov_b32_e32 v6, v8
+; SI-NEXT: v_mov_b32_e32 v7, v9
+; SI-NEXT: buffer_atomic_cmpswap_x2 v[4:7], v[0:1], s[4:7], 0 addr64 offset:32 glc
; SI-NEXT: s_waitcnt vmcnt(0)
; SI-NEXT: buffer_wbinvl1
-; SI-NEXT: v_cmp_eq_u64_e32 vcc, v[8:9], v[6:7]
+; SI-NEXT: v_cmp_eq_u64_e32 vcc, v[4:5], v[8:9]
; SI-NEXT: s_or_b64 s[8:9], vcc, s[8:9]
-; SI-NEXT: v_mov_b32_e32 v6, v8
-; SI-NEXT: v_mov_b32_e32 v7, v9
; SI-NEXT: s_andn2_b64 exec, exec, s[8:9]
; SI-NEXT: s_cbranch_execnz .LBB105_1
; SI-NEXT: ; %bb.2: ; %atomicrmw.end
@@ -6885,26 +6885,26 @@ define void @global_atomic_umin_i64_noret(ptr addrspace(1) %ptr, i64 %in) {
; SI-NEXT: s_mov_b32 s7, 0xf000
; SI-NEXT: s_mov_b32 s4, s6
; SI-NEXT: s_mov_b32 s5, s6
-; SI-NEXT: buffer_load_dwordx2 v[6:7], v[0:1], s[4:7], 0 addr64
+; SI-NEXT: buffer_load_dwordx2 v[4:5], v[0:1], s[4:7], 0 addr64
; SI-NEXT: s_mov_b64 s[8:9], 0
; SI-NEXT: .LBB107_1: ; %atomicrmw.start
; SI-NEXT: ; =>This Inner Loop Header: Depth=1
; SI-NEXT: s_waitcnt vmcnt(0)
-; SI-NEXT: v_cmp_le_u64_e32 vcc, v[6:7], v[2:3]
-; SI-NEXT: v_cndmask_b32_e32 v5, v3, v7, vcc
-; SI-NEXT: v_cndmask_b32_e32 v4, v2, v6, vcc
-; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_mov_b32_e32 v11, v7
-; SI-NEXT: v_mov_b32_e32 v10, v6
; SI-NEXT: v_mov_b32_e32 v9, v5
; SI-NEXT: v_mov_b32_e32 v8, v4
-; SI-NEXT: buffer_atomic_cmpswap_x2 v[8:11], v[0:1], s[4:7], 0 addr64 glc
+; SI-NEXT: v_cmp_le_u64_e32 vcc, v[8:9], v[2:3]
+; SI-NEXT: s_waitcnt expcnt(0)
+; SI-NEXT: v_cndmask_b32_e32 v7, v3, v9, vcc
+; SI-NEXT: v_cndmask_b32_e32 v6, v2, v8, vcc
+; SI-NEXT: v_mov_b32_e32 v4, v6
+; SI-NEXT: v_mov_b32_e32 v5, v7
+; SI-NEXT: v_mov_b32_e32 v6, v8
+; SI-NEXT: v_mov_b32_e32 v7, v9
+; SI-NEXT: buffer_atomic_cmpswap_x2 v[4:7], v[0:1], s[4:7], 0 addr64 glc
; SI-NEXT: s_waitcnt vmcnt(0)
; SI-NEXT: buffer_wbinvl1
-; SI-NEXT: v_cmp_eq_u64_e32 vcc, v[8:9], v[6:7]
+; SI-NEXT: v_cmp_eq_u64_e32 vcc, v[4:5], v[8:9]
; SI-NEXT: s_or_b64 s[8:9], vcc, s[8:9]
-; SI-NEXT: v_mov_b32_e32 v6, v8
-; SI-NEXT: v_mov_b32_e32 v7, v9
; SI-NEXT: s_andn2_b64 exec, exec, s[8:9]
; SI-NEXT: s_cbranch_execnz .LBB107_1
; SI-NEXT: ; %bb.2: ; %atomicrmw.end
@@ -6971,26 +6971,26 @@ define void @global_atomic_umin_i64_noret_offset(ptr addrspace(1) %out, i64 %in)
; SI-NEXT: s_mov_b32 s7, 0xf000
; SI-NEXT: s_mov_b32 s4, s6
; SI-NEXT: s_mov_b32 s5, s6
-; SI-NEXT: buffer_load_dwordx2 v[6:7], v[0:1], s[4:7], 0 addr64 offset:32
+; SI-NEXT: buffer_load_dwordx2 v[4:5], v[0:1], s[4:7], 0 addr64 offset:32
; SI-NEXT: s_mov_b64 s[8:9], 0
; SI-NEXT: .LBB108_1: ; %atomicrmw.start
; SI-NEXT: ; =>This Inner Loop Header: Depth=1
; SI-NEXT: s_waitcnt vmcnt(0)
-; SI-NEXT: v_cmp_le_u64_e32 vcc, v[6:7], v[2:3]
-; SI-NEXT: v_cndmask_b32_e32 v5, v3, v7, vcc
-; SI-NEXT: v_cndmask_b32_e32 v4, v2, v6, vcc
-; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_mov_b32_e32 v11, v7
-; SI-NEXT: v_mov_b32_e32 v10, v6
; SI-NEXT: v_mov_b32_e32 v9, v5
; SI-NEXT: v_mov_b32_e32 v8, v4
-; SI-NEXT: buffer_atomic_cmpswap_x2 v[8:11], v[0:1], s[4:7], 0 addr64 offset:32 glc
+; SI-NEXT: v_cmp_le_u64_e32 vcc, v[8:9], v[2:3]
+; SI-NEXT: s_waitcnt expcnt(0)
+; SI-NEXT: v_cndmask_b32_e32 v7, v3, v9, vcc
+; SI-NEXT: v_cndmask_b32_e32 v6, v2, v8, vcc
+; SI-NEXT: v_mov_b32_e32 v4, v6
+; SI-NEXT: v_mov_b32_e32 v5, v7
+; SI-NEXT: v_mov_b32_e32 v6, v8
+; SI-NEXT: v_mov_b32_e32 v7, v9
+; SI-NEXT: buffer_atomic_cmpswap_x2 v[4:7], v[0:1], s[4:7], 0 addr64 offset:32 glc
; SI-NEXT: s_waitcnt vmcnt(0)
; SI-NEXT: buffer_wbinvl1
-; SI-NEXT: v_cmp_eq_u64_e32 vcc, v[8:9], v[6:7]
+; SI-NEXT: v_cmp_eq_u64_e32 vcc, v[4:5], v[8:9]
; SI-NEXT: s_or_b64 s[8:9], vcc, s[8:9]
-; SI-NEXT: v_mov_b32_e32 v6, v8
-; SI-NEXT: v_mov_b32_e32 v7, v9
; SI-NEXT: s_andn2_b64 exec, exec, s[8:9]
; SI-NEXT: s_cbranch_execnz .LBB108_1
; SI-NEXT: ; %bb.2: ; %atomicrmw.end
@@ -7255,28 +7255,28 @@ define amdgpu_gfx void @global_atomic_umin_i64_noret_scalar(ptr addrspace(1) inr
; SI-NEXT: s_mov_b32 s34, s6
; SI-NEXT: s_mov_b32 s7, 0xf000
; SI-NEXT: s_mov_b32 s6, -1
-; SI-NEXT: buffer_load_dwordx2 v[2:3], off, s[4:7], 0
+; SI-NEXT: buffer_load_dwordx2 v[0:1], off, s[4:7], 0
; SI-NEXT: s_mov_b64 s[36:37], 0
; SI-NEXT: v_mov_b32_e32 v4, s35
; SI-NEXT: v_mov_b32_e32 v5, s34
; SI-NEXT: .LBB111_1: ; %atomicrmw.start
; SI-NEXT: ; =>This Inner Loop Header: Depth=1
; SI-NEXT: s_waitcnt vmcnt(0)
-; SI-NEXT: v_cmp_ge_u64_e32 vcc, s[34:35], v[2:3]
-; SI-NEXT: v_cndmask_b32_e32 v1, v4, v3, vcc
-; SI-NEXT: v_cndmask_b32_e32 v0, v5, v2, vcc
+; SI-NEXT: v_mov_b32_e32 v9, v1
+; SI-NEXT: v_mov_b32_e32 v8, v0
+; SI-NEXT: v_cmp_ge_u64_e32 vcc, s[34:35], v[8:9]
+; SI-NEXT: v_cndmask_b32_e32 v7, v4, v9, vcc
+; SI-NEXT: v_cndmask_b32_e32 v6, v5, v8, vcc
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_mov_b32_e32 v9, v3
-; SI-NEXT: v_mov_b32_e32 v8, v2
-; SI-NEXT: v_mov_b32_e32 v7, v1
-; SI-NEXT: v_mov_b32_e32 v6, v0
-; SI-NEXT: buffer_atomic_cmpswap_x2 v[6:9], off, s[4:7], 0 glc
+; SI-NEXT: v_mov_b32_e32 v0, v6
+; SI-NEXT: v_mov_b32_e32 v1, v7
+; SI-NEXT: v_mov_b32_e32 v2, v8
+; SI-NEXT: v_mov_b32_e32 v3, v9
+; SI-NEXT: buffer_atomic_cmpswap_x2 v[0:3], off, s[4:7], 0 glc
; SI-NEXT: s_waitcnt vmcnt(0)
; SI-NEXT: buffer_wbinvl1
-; SI-NEXT: v_cmp_eq_u64_e32 vcc, v[6:7], v[2:3]
+; SI-NEXT: v_cmp_eq_u64_e32 vcc, v[0:1], v[8:9]
; SI-NEXT: s_or_b64 s[36:37], vcc, s[36:37]
-; SI-NEXT: v_mov_b32_e32 v2, v6
-; SI-NEXT: v_mov_b32_e32 v3, v7
; SI-NEXT: s_andn2_b64 exec, exec, s[36:37]
; SI-NEXT: s_cbranch_execnz .LBB111_1
; SI-NEXT: ; %bb.2: ; %atomicrmw.end
@@ -7363,28 +7363,28 @@ define amdgpu_gfx void @global_atomic_umin_i64_noret_offset_scalar(ptr addrspace
; SI-NEXT: s_mov_b32 s34, s6
; SI-NEXT: s_mov_b32 s7, 0xf000
; SI-NEXT: s_mov_b32 s6, -1
-; SI-NEXT: buffer_load_dwordx2 v[2:3], off, s[4:7], 0 offset:32
+; SI-NEXT: buffer_load_dwordx2 v[0:1], off, s[4:7], 0 offset:32
; SI-NEXT: s_mov_b64 s[36:37], 0
; SI-NEXT: v_mov_b32_e32 v4, s35
; SI-NEXT: v_mov_b32_e32 v5, s34
; SI-NEXT: .LBB112_1: ; %atomicrmw.start
; SI-NEXT: ; =>This Inner Loop Header: Depth=1
; SI-NEXT: s_waitcnt vmcnt(0)
-; SI-NEXT: v_cmp_ge_u64_e32 vcc, s[34:35], v[2:3]
-; SI-NEXT: v_cndmask_b32_e32 v1, v4, v3, vcc
-; SI-NEXT: v_cndmask_b32_e32 v0, v5, v2, vcc
+; SI-NEXT: v_mov_b32_e32 v9, v1
+; SI-NEXT: v_mov_b32_e32 v8, v0
+; SI-NEXT: v_cmp_ge_u64_e32 vcc, s[34:35], v[8:9]
+; SI-NEXT: v_cndmask_b32_e32 v7, v4, v9, vcc
+; SI-NEXT: v_cndmask_b32_e32 v6, v5, v8, vcc
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_mov_b32_e32 v9, v3
-; SI-NEXT: v_mov_b32_e32 v8, v2
-; SI-NEXT: v_mov_b32_e32 v7, v1
-; SI-NEXT: v_mov_b32_e32 v6, v0
-; SI-NEXT: buffer_atomic_cmpswap_x2 v[6:9], off, s[4:7], 0 offset:32 glc
+; SI-NEXT: v_mov_b32_e32 v0, v6
+; SI-NEXT: v_mov_b32_e32 v1, v7
+; SI-NEXT: v_mov_b32_e32 v2, v8
+; SI-NEXT: v_mov_b32_e32 v3, v9
+; SI-NEXT: buffer_atomic_cmpswap_x2 v[0:3], off, s[4:7], 0 offset:32 glc
; SI-NEXT: s_waitcnt vmcnt(0)
; SI-NEXT: buffer_wbinvl1
-; SI-NEXT: v_cmp_eq_u64_e32 vcc, v[6:7], v[2:3]
+; SI-NEXT: v_cmp_eq_u64_e32 vcc, v[0:1], v[8:9]
; SI-NEXT: s_or_b64 s[36:37], vcc, s[36:37]
-; SI-NEXT: v_mov_b32_e32 v2, v6
-; SI-NEXT: v_mov_b32_e32 v3, v7
; SI-NEXT: s_andn2_b64 exec, exec, s[36:37]
; SI-NEXT: s_cbranch_execnz .LBB112_1
; SI-NEXT: ; %bb.2: ; %atomicrmw.end
@@ -7683,26 +7683,26 @@ define void @global_atomic_umin_i64_noret_offset__amdgpu_no_remote_memory(ptr ad
; SI-NEXT: s_mov_b32 s7, 0xf000
; SI-NEXT: s_mov_b32 s4, s6
; SI-NEXT: s_mov_b32 s5, s6
-; SI-NEXT: buffer_load_dwordx2 v[6:7], v[0:1], s[4:7], 0 addr64 offset:32
+; SI-NEXT: buffer_load_dwordx2 v[4:5], v[0:1], s[4:7], 0 addr64 offset:32
; SI-NEXT: s_mov_b64 s[8:9], 0
; SI-NEXT: .LBB115_1: ; %atomicrmw.start
; SI-NEXT: ; =>This Inner Loop Header: Depth=1
; SI-NEXT: s_waitcnt vmcnt(0)
-; SI-NEXT: v_cmp_le_u64_e32 vcc, v[6:7], v[2:3]
-; SI-NEXT: v_cndmask_b32_e32 v5, v3, v7, vcc
-; SI-NEXT: v_cndmask_b32_e32 v4, v2, v6, vcc
-; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_mov_b32_e32 v11, v7
-; SI-NEXT: v_mov_b32_e32 v10, v6
; SI-NEXT: v_mov_b32_e32 v9, v5
; SI-NEXT: v_mov_b32_e32 v8, v4
-; SI-NEXT: buffer_atomic_cmpswap_x2 v[8:11], v[0:1], s[4:7], 0 addr64 offset:32 glc
+; SI-NEXT: v_cmp_le_u64_e32 vcc, v[8:9], v[2:3]
+; SI-NEXT: s_waitcnt expcnt(0)
+; SI-NEXT: v_cndmask_b32_e32 v7, v3, v9, vcc
+; SI-NEXT: v_cndmask_b32_e32 v6, v2, v8, vcc
+; SI-NEXT: v_mov_b32_e32 v4, v6
+; SI-NEXT: v_mov_b32_e32 v5, v7
+; SI-NEXT: v_mov_b32_e32 v6, v8
+; SI-NEXT: v_mov_b32_e32 v7, v9
+; SI-NEXT: buffer_atomic_cmpswap_x2 v[4:7], v[0:1], s[4:7], 0 addr64 offset:32 glc
; SI-NEXT: s_waitcnt vmcnt(0)
; SI-NEXT: buffer_wbinvl1
-; SI-NEXT: v_cmp_eq_u64_e32 vcc, v[8:9], v[6:7]
+; SI-NEXT: v_cmp_eq_u64_e32 vcc, v[4:5], v[8:9]
; SI-NEXT: s_or_b64 s[8:9], vcc, s[8:9]
-; SI-NEXT: v_mov_b32_e32 v6, v8
-; SI-NEXT: v_mov_b32_e32 v7, v9
; SI-NEXT: s_andn2_b64 exec, exec, s[8:9]
; SI-NEXT: s_cbranch_execnz .LBB115_1
; SI-NEXT: ; %bb.2: ; %atomicrmw.end
@@ -7871,26 +7871,26 @@ define void @global_atomic_min_i64_noret(ptr addrspace(1) %ptr, i64 %in) {
; SI-NEXT: s_mov_b32 s7, 0xf000
; SI-NEXT: s_mov_b32 s4, s6
; SI-NEXT: s_mov_b32 s5, s6
-; SI-NEXT: buffer_load_dwordx2 v[6:7], v[0:1], s[4:7], 0 addr64
+; SI-NEXT: buffer_load_dwordx2 v[4:5], v[0:1], s[4:7], 0 addr64
; SI-NEXT: s_mov_b64 s[8:9], 0
; SI-NEXT: .LBB117_1: ; %atomicrmw.start
; SI-NEXT: ; =>This Inner Loop Header: Depth=1
; SI-NEXT: s_waitcnt vmcnt(0)
-; SI-NEXT: v_cmp_le_i64_e32 vcc, v[6:7], v[2:3]
-; SI-NEXT: v_cndmask_b32_e32 v5, v3, v7, vcc
-; SI-NEXT: v_cndmask_b32_e32 v4, v2, v6, vcc
-; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_mov_b32_e32 v11, v7
-; SI-NEXT: v_mov_b32_e32 v10, v6
; SI-NEXT: v_mov_b32_e32 v9, v5
; SI-NEXT: v_mov_b32_e32 v8, v4
-; SI-NEXT: buffer_atomic_cmpswap_x2 v[8:11], v[0:1], s[4:7], 0 addr64 glc
+; SI-NEXT: v_cmp_le_i64_e32 vcc, v[8:9], v[2:3]
+; SI-NEXT: s_waitcnt expcnt(0)
+; SI-NEXT: v_cndmask_b32_e32 v7, v3, v9, vcc
+; SI-NEXT: v_cndmask_b32_e32 v6, v2, v8, vcc
+; SI-NEXT: v_mov_b32_e32 v4, v6
+; SI-NEXT: v_mov_b32_e32 v5, v7
+; SI-NEXT: v_mov_b32_e32 v6, v8
+; SI-NEXT: v_mov_b32_e32 v7, v9
+; SI-NEXT: buffer_atomic_cmpswap_x2 v[4:7], v[0:1], s[4:7], 0 addr64 glc
; SI-NEXT: s_waitcnt vmcnt(0)
; SI-NEXT: buffer_wbinvl1
-; SI-NEXT: v_cmp_eq_u64_e32 vcc, v[8:9], v[6:7]
+; SI-NEXT: v_cmp_eq_u64_e32 vcc, v[4:5], v[8:9]
; SI-NEXT: s_or_b64 s[8:9], vcc, s[8:9]
-; SI-NEXT: v_mov_b32_e32 v6, v8
-; SI-NEXT: v_mov_b32_e32 v7, v9
; SI-NEXT: s_andn2_b64 exec, exec, s[8:9]
; SI-NEXT: s_cbranch_execnz .LBB117_1
; SI-NEXT: ; %bb.2: ; %atomicrmw.end
@@ -7957,26 +7957,26 @@ define void @global_atomic_min_i64_noret_offset(ptr addrspace(1) %out, i64 %in)
; SI-NEXT: s_mov_b32 s7, 0xf000
; SI-NEXT: s_mov_b32 s4, s6
; SI-NEXT: s_mov_b32 s5, s6
-; SI-NEXT: buffer_load_dwordx2 v[6:7], v[0:1], s[4:7], 0 addr64 offset:32
+; SI-NEXT: buffer_load_dwordx2 v[4:5], v[0:1], s[4:7], 0 addr64 offset:32
; SI-NEXT: s_mov_b64 s[8:9], 0
; SI-NEXT: .LBB118_1: ; %atomicrmw.start
; SI-NEXT: ; =>This Inner Loop Header: Depth=1
; SI-NEXT: s_waitcnt vmcnt(0)
-; SI-NEXT: v_cmp_le_i64_e32 vcc, v[6:7], v[2:3]
-; SI-NEXT: v_cndmask_b32_e32 v5, v3, v7, vcc
-; SI-NEXT: v_cndmask_b32_e32 v4, v2, v6, vcc
-; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_mov_b32_e32 v11, v7
-; SI-NEXT: v_mov_b32_e32 v10, v6
; SI-NEXT: v_mov_b32_e32 v9, v5
; SI-NEXT: v_mov_b32_e32 v8, v4
-; SI-NEXT: buffer_atomic_cmpswap_x2 v[8:11], v[0:1], s[4:7], 0 addr64 offset:32 glc
+; SI-NEXT: v_cmp_le_i64_e32 vcc, v[8:9], v[2:3]
+; SI-NEXT: s_waitcnt expcnt(0)
+; SI-NEXT: v_cndmask_b32_e32 v7, v3, v9, vcc
+; SI-NEXT: v_cndmask_b32_e32 v6, v2, v8, vcc
+; SI-NEXT: v_mov_b32_e32 v4, v6
+; SI-NEXT: v_mov_b32_e32 v5, v7
+; SI-NEXT: v_mov_b32_e32 v6, v8
+; SI-NEXT: v_mov_b32_e32 v7, v9
+; SI-NEXT: buffer_atomic_cmpswap_x2 v[4:7], v[0:1], s[4:7], 0 addr64 offset:32 glc
; SI-NEXT: s_waitcnt vmcnt(0)
; SI-NEXT: buffer_wbinvl1
-; SI-NEXT: v_cmp_eq_u64_e32 vcc, v[8:9], v[6:7]
+; SI-NEXT: v_cmp_eq_u64_e32 vcc, v[4:5], v[8:9]
; SI-NEXT: s_or_b64 s[8:9], vcc, s[8:9]
-; SI-NEXT: v_mov_b32_e32 v6, v8
-; SI-NEXT: v_mov_b32_e32 v7, v9
; SI-NEXT: s_andn2_b64 exec, exec, s[8:9]
; SI-NEXT: s_cbranch_execnz .LBB118_1
; SI-NEXT: ; %bb.2: ; %atomicrmw.end
@@ -8241,28 +8241,28 @@ define amdgpu_gfx void @global_atomic_min_i64_noret_scalar(ptr addrspace(1) inre
; SI-NEXT: s_mov_b32 s34, s6
; SI-NEXT: s_mov_b32 s7, 0xf000
; SI-NEXT: s_mov_b32 s6, -1
-; SI-NEXT: buffer_load_dwordx2 v[2:3], off, s[4:7], 0
+; SI-NEXT: buffer_load_dwordx2 v[0:1], off, s[4:7], 0
; SI-NEXT: s_mov_b64 s[36:37], 0
; SI-NEXT: v_mov_b32_e32 v4, s35
; SI-NEXT: v_mov_b32_e32 v5, s34
; SI-NEXT: .LBB121_1: ; %atomicrmw.start
; SI-NEXT: ; =>This Inner Loop Header: Depth=1
; SI-NEXT: s_waitcnt vmcnt(0)
-; SI-NEXT: v_cmp_ge_i64_e32 vcc, s[34:35], v[2:3]
-; SI-NEXT: v_cndmask_b32_e32 v1, v4, v3, vcc
-; SI-NEXT: v_cndmask_b32_e32 v0, v5, v2, vcc
+; SI-NEXT: v_mov_b32_e32 v9, v1
+; SI-NEXT: v_mov_b32_e32 v8, v0
+; SI-NEXT: v_cmp_ge_i64_e32 vcc, s[34:35], v[8:9]
+; SI-NEXT: v_cndmask_b32_e32 v7, v4, v9, vcc
+; SI-NEXT: v_cndmask_b32_e32 v6, v5, v8, vcc
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_mov_b32_e32 v9, v3
-; SI-NEXT: v_mov_b32_e32 v8, v2
-; SI-NEXT: v_mov_b32_e32 v7, v1
-; SI-NEXT: v_mov_b32_e32 v6, v0
-; SI-NEXT: buffer_atomic_cmpswap_x2 v[6:9], off, s[4:7], 0 glc
+; SI-NEXT: v_mov_b32_e32 v0, v6
+; SI-NEXT: v_mov_b32_e32 v1, v7
+; SI-NEXT: v_mov_b32_e32 v2, v8
+; SI-NEXT: v_mov_b32_e32 v3, v9
+; SI-NEXT: buffer_atomic_cmpswap_x2 v[0:3], off, s[4:7], 0 glc
; SI-NEXT: s_waitcnt vmcnt(0)
; SI-NEXT: buffer_wbinvl1
-; SI-NEXT: v_cmp_eq_u64_e32 vcc, v[6:7], v[2:3]
+; SI-NEXT: v_cmp_eq_u64_e32 vcc, v[0:1], v[8:9]
; SI-NEXT: s_or_b64 s[36:37], vcc, s[36:37]
-; SI-NEXT: v_mov_b32_e32 v2, v6
-; SI-NEXT: v_mov_b32_e32 v3, v7
; SI-NEXT: s_andn2_b64 exec, exec, s[36:37]
; SI-NEXT: s_cbranch_execnz .LBB121_1
; SI-NEXT: ; %bb.2: ; %atomicrmw.end
@@ -8349,28 +8349,28 @@ define amdgpu_gfx void @global_atomic_min_i64_noret_offset_scalar(ptr addrspace(
; SI-NEXT: s_mov_b32 s34, s6
; SI-NEXT: s_mov_b32 s7, 0xf000
; SI-NEXT: s_mov_b32 s6, -1
-; SI-NEXT: buffer_load_dwordx2 v[2:3], off, s[4:7], 0 offset:32
+; SI-NEXT: buffer_load_dwordx2 v[0:1], off, s[4:7], 0 offset:32
; SI-NEXT: s_mov_b64 s[36:37], 0
; SI-NEXT: v_mov_b32_e32 v4, s35
; SI-NEXT: v_mov_b32_e32 v5, s34
; SI-NEXT: .LBB122_1: ; %atomicrmw.start
; SI-NEXT: ; =>This Inner Loop Header: Depth=1
; SI-NEXT: s_waitcnt vmcnt(0)
-; SI-NEXT: v_cmp_ge_i64_e32 vcc, s[34:35], v[2:3]
-; SI-NEXT: v_cndmask_b32_e32 v1, v4, v3, vcc
-; SI-NEXT: v_cndmask_b32_e32 v0, v5, v2, vcc
+; SI-NEXT: v_mov_b32_e32 v9, v1
+; SI-NEXT: v_mov_b32_e32 v8, v0
+; SI-NEXT: v_cmp_ge_i64_e32 vcc, s[34:35], v[8:9]
+; SI-NEXT: v_cndmask_b32_e32 v7, v4, v9, vcc
+; SI-NEXT: v_cndmask_b32_e32 v6, v5, v8, vcc
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_mov_b32_e32 v9, v3
-; SI-NEXT: v_mov_b32_e32 v8, v2
-; SI-NEXT: v_mov_b32_e32 v7, v1
-; SI-NEXT: v_mov_b32_e32 v6, v0
-; SI-NEXT: buffer_atomic_cmpswap_x2 v[6:9], off, s[4:7], 0 offset:32 glc
+; SI-NEXT: v_mov_b32_e32 v0, v6
+; SI-NEXT: v_mov_b32_e32 v1, v7
+; SI-NEXT: v_mov_b32_e32 v2, v8
+; SI-NEXT: v_mov_b32_e32 v3, v9
+; SI-NEXT: buffer_atomic_cmpswap_x2 v[0:3], off, s[4:7], 0 offset:32 glc
; SI-NEXT: s_waitcnt vmcnt(0)
; SI-NEXT: buffer_wbinvl1
-; SI-NEXT: v_cmp_eq_u64_e32 vcc, v[6:7], v[2:3]
+; SI-NEXT: v_cmp_eq_u64_e32 vcc, v[0:1], v[8:9]
; SI-NEXT: s_or_b64 s[36:37], vcc, s[36:37]
-; SI-NEXT: v_mov_b32_e32 v2, v6
-; SI-NEXT: v_mov_b32_e32 v3, v7
; SI-NEXT: s_andn2_b64 exec, exec, s[36:37]
; SI-NEXT: s_cbranch_execnz .LBB122_1
; SI-NEXT: ; %bb.2: ; %atomicrmw.end
@@ -8676,26 +8676,26 @@ define amdgpu_kernel void @atomic_min_i64_addr64_offset(ptr addrspace(1) %out, i
; SI-NEXT: v_mov_b32_e32 v4, s3
; SI-NEXT: v_mov_b32_e32 v5, s2
; SI-NEXT: s_waitcnt lgkmcnt(0)
-; SI-NEXT: v_mov_b32_e32 v2, s8
-; SI-NEXT: v_mov_b32_e32 v3, s9
+; SI-NEXT: v_mov_b32_e32 v0, s8
+; SI-NEXT: v_mov_b32_e32 v1, s9
; SI-NEXT: s_mov_b32 s6, -1
; SI-NEXT: .LBB125_1: ; %atomicrmw.start
; SI-NEXT: ; =>This Inner Loop Header: Depth=1
-; SI-NEXT: v_cmp_ge_i64_e32 vcc, s[2:3], v[2:3]
-; SI-NEXT: v_cndmask_b32_e32 v1, v4, v3, vcc
-; SI-NEXT: v_cndmask_b32_e32 v0, v5, v2, vcc
+; SI-NEXT: v_mov_b32_e32 v9, v1
+; SI-NEXT: v_mov_b32_e32 v8, v0
+; SI-NEXT: v_cmp_ge_i64_e32 vcc, s[2:3], v[8:9]
+; SI-NEXT: v_cndmask_b32_e32 v7, v4, v9, vcc
+; SI-NEXT: v_cndmask_b32_e32 v6, v5, v8, vcc
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_mov_b32_e32 v9, v3
-; SI-NEXT: v_mov_b32_e32 v8, v2
-; SI-NEXT: v_mov_b32_e32 v7, v1
-; SI-NEXT: v_mov_b32_e32 v6, v0
-; SI-NEXT: buffer_atomic_cmpswap_x2 v[6:9], off, s[4:7], 0 offset:32 glc
+; SI-NEXT: v_mov_b32_e32 v0, v6
+; SI-NEXT: v_mov_b32_e32 v1, v7
+; SI-NEXT: v_mov_b32_e32 v2, v8
+; SI-NEXT: v_mov_b32_e32 v3, v9
+; SI-NEXT: buffer_atomic_cmpswap_x2 v[0:3], off, s[4:7], 0 offset:32 glc
; SI-NEXT: s_waitcnt vmcnt(0)
; SI-NEXT: buffer_wbinvl1
-; SI-NEXT: v_cmp_eq_u64_e32 vcc, v[6:7], v[2:3]
+; SI-NEXT: v_cmp_eq_u64_e32 vcc, v[0:1], v[8:9]
; SI-NEXT: s_or_b64 s[0:1], vcc, s[0:1]
-; SI-NEXT: v_mov_b32_e32 v2, v6
-; SI-NEXT: v_mov_b32_e32 v3, v7
; SI-NEXT: s_andn2_b64 exec, exec, s[0:1]
; SI-NEXT: s_cbranch_execnz .LBB125_1
; SI-NEXT: ; %bb.2: ; %atomicrmw.end
@@ -8787,29 +8787,29 @@ define amdgpu_kernel void @atomic_min_i64_ret_addr64_offset(ptr addrspace(1) %ou
; SI-NEXT: s_load_dwordx2 s[6:7], s[8:9], 0x8
; SI-NEXT: s_mov_b64 s[0:1], 0
; SI-NEXT: s_mov_b32 s11, 0xf000
-; SI-NEXT: v_mov_b32_e32 v8, s5
-; SI-NEXT: v_mov_b32_e32 v9, s4
+; SI-NEXT: v_mov_b32_e32 v4, s5
+; SI-NEXT: v_mov_b32_e32 v5, s4
; SI-NEXT: s_waitcnt lgkmcnt(0)
-; SI-NEXT: v_mov_b32_e32 v2, s6
-; SI-NEXT: v_mov_b32_e32 v3, s7
+; SI-NEXT: v_mov_b32_e32 v0, s6
+; SI-NEXT: v_mov_b32_e32 v1, s7
; SI-NEXT: s_mov_b32 s10, -1
; SI-NEXT: .LBB126_1: ; %atomicrmw.start
; SI-NEXT: ; =>This Inner Loop Header: Depth=1
-; SI-NEXT: v_cmp_ge_i64_e32 vcc, s[4:5], v[2:3]
-; SI-NEXT: v_cndmask_b32_e32 v1, v8, v3, vcc
-; SI-NEXT: v_cndmask_b32_e32 v0, v9, v2, vcc
+; SI-NEXT: v_mov_b32_e32 v9, v1
+; SI-NEXT: v_mov_b32_e32 v8, v0
+; SI-NEXT: v_cmp_ge_i64_e32 vcc, s[4:5], v[8:9]
+; SI-NEXT: v_cndmask_b32_e32 v7, v4, v9, vcc
+; SI-NEXT: v_cndmask_b32_e32 v6, v5, v8, vcc
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_mov_b32_e32 v7, v3
-; SI-NEXT: v_mov_b32_e32 v6, v2
-; SI-NEXT: v_mov_b32_e32 v5, v1
-; SI-NEXT: v_mov_b32_e32 v4, v0
-; SI-NEXT: buffer_atomic_cmpswap_x2 v[4:7], off, s[8:11], 0 offset:32 glc
+; SI-NEXT: v_mov_b32_e32 v0, v6
+; SI-NEXT: v_mov_b32_e32 v1, v7
+; SI-NEXT: v_mov_b32_e32 v2, v8
+; SI-NEXT: v_mov_b32_e32 v3, v9
+; SI-NEXT: buffer_atomic_cmpswap_x2 v[0:3], off, s[8:11], 0 offset:32 glc
; SI-NEXT: s_waitcnt vmcnt(0)
; SI-NEXT: buffer_wbinvl1
-; SI-NEXT: v_cmp_eq_u64_e32 vcc, v[4:5], v[2:3]
+; SI-NEXT: v_cmp_eq_u64_e32 vcc, v[0:1], v[8:9]
; SI-NEXT: s_or_b64 s[0:1], vcc, s[0:1]
-; SI-NEXT: v_mov_b32_e32 v2, v4
-; SI-NEXT: v_mov_b32_e32 v3, v5
; SI-NEXT: s_andn2_b64 exec, exec, s[0:1]
; SI-NEXT: s_cbranch_execnz .LBB126_1
; SI-NEXT: ; %bb.2: ; %atomicrmw.end
@@ -8818,7 +8818,7 @@ define amdgpu_kernel void @atomic_min_i64_ret_addr64_offset(ptr addrspace(1) %ou
; SI-NEXT: s_mov_b32 s6, -1
; SI-NEXT: s_mov_b32 s4, s2
; SI-NEXT: s_mov_b32 s5, s3
-; SI-NEXT: buffer_store_dwordx2 v[4:5], off, s[4:7], 0
+; SI-NEXT: buffer_store_dwordx2 v[0:1], off, s[4:7], 0
; SI-NEXT: s_endpgm
;
; VI-LABEL: atomic_min_i64_ret_addr64_offset:
@@ -8913,28 +8913,28 @@ define amdgpu_kernel void @atomic_min_i64(ptr addrspace(1) %out, i64 %in) {
; SI-NEXT: v_mov_b32_e32 v4, s3
; SI-NEXT: v_mov_b32_e32 v5, s2
; SI-NEXT: s_waitcnt lgkmcnt(0)
-; SI-NEXT: v_mov_b32_e32 v2, s4
-; SI-NEXT: v_mov_b32_e32 v3, s5
+; SI-NEXT: v_mov_b32_e32 v0, s4
+; SI-NEXT: v_mov_b32_e32 v1, s5
; SI-NEXT: s_mov_b32 s6, -1
; SI-NEXT: s_mov_b32 s4, s0
; SI-NEXT: s_mov_b32 s5, s1
; SI-NEXT: .LBB127_1: ; %atomicrmw.start
; SI-NEXT: ; =>This Inner Loop Header: Depth=1
-; SI-NEXT: v_cmp_ge_i64_e32 vcc, s[2:3], v[2:3]
-; SI-NEXT: v_cndmask_b32_e32 v1, v4, v3, vcc
-; SI-NEXT: v_cndmask_b32_e32 v0, v5, v2, vcc
+; SI-NEXT: v_mov_b32_e32 v9, v1
+; SI-NEXT: v_mov_b32_e32 v8, v0
+; SI-NEXT: v_cmp_ge_i64_e32 vcc, s[2:3], v[8:9]
+; SI-NEXT: v_cndmask_b32_e32 v7, v4, v9, vcc
+; SI-NEXT: v_cndmask_b32_e32 v6, v5, v8, vcc
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_mov_b32_e32 v9, v3
-; SI-NEXT: v_mov_b32_e32 v8, v2
-; SI-NEXT: v_mov_b32_e32 v7, v1
-; SI-NEXT: v_mov_b32_e32 v6, v0
-; SI-NEXT: buffer_atomic_cmpswap_x2 v[6:9], off, s[4:7], 0 glc
+; SI-NEXT: v_mov_b32_e32 v0, v6
+; SI-NEXT: v_mov_b32_e32 v1, v7
+; SI-NEXT: v_mov_b32_e32 v2, v8
+; SI-NEXT: v_mov_b32_e32 v3, v9
+; SI-NEXT: buffer_atomic_cmpswap_x2 v[0:3], off, s[4:7], 0 glc
; SI-NEXT: s_waitcnt vmcnt(0)
; SI-NEXT: buffer_wbinvl1
-; SI-NEXT: v_cmp_eq_u64_e32 vcc, v[6:7], v[2:3]
+; SI-NEXT: v_cmp_eq_u64_e32 vcc, v[0:1], v[8:9]
; SI-NEXT: s_or_b64 s[8:9], vcc, s[8:9]
-; SI-NEXT: v_mov_b32_e32 v2, v6
-; SI-NEXT: v_mov_b32_e32 v3, v7
; SI-NEXT: s_andn2_b64 exec, exec, s[8:9]
; SI-NEXT: s_cbranch_execnz .LBB127_1
; SI-NEXT: ; %bb.2: ; %atomicrmw.end
@@ -9014,29 +9014,29 @@ define amdgpu_kernel void @atomic_min_i64_ret_addr64(ptr addrspace(1) %out, ptr
; SI-NEXT: s_load_dwordx2 s[6:7], s[8:9], 0x0
; SI-NEXT: s_mov_b64 s[0:1], 0
; SI-NEXT: s_mov_b32 s11, 0xf000
-; SI-NEXT: v_mov_b32_e32 v8, s5
-; SI-NEXT: v_mov_b32_e32 v9, s4
+; SI-NEXT: v_mov_b32_e32 v4, s5
+; SI-NEXT: v_mov_b32_e32 v5, s4
; SI-NEXT: s_waitcnt lgkmcnt(0)
-; SI-NEXT: v_mov_b32_e32 v2, s6
-; SI-NEXT: v_mov_b32_e32 v3, s7
+; SI-NEXT: v_mov_b32_e32 v0, s6
+; SI-NEXT: v_mov_b32_e32 v1, s7
; SI-NEXT: s_mov_b32 s10, -1
; SI-NEXT: .LBB128_1: ; %atomicrmw.start
; SI-NEXT: ; =>This Inner Loop Header: Depth=1
-; SI-NEXT: v_cmp_ge_i64_e32 vcc, s[4:5], v[2:3]
-; SI-NEXT: v_cndmask_b32_e32 v1, v8, v3, vcc
-; SI-NEXT: v_cndmask_b32_e32 v0, v9, v2, vcc
+; SI-NEXT: v_mov_b32_e32 v9, v1
+; SI-NEXT: v_mov_b32_e32 v8, v0
+; SI-NEXT: v_cmp_ge_i64_e32 vcc, s[4:5], v[8:9]
+; SI-NEXT: v_cndmask_b32_e32 v7, v4, v9, vcc
+; SI-NEXT: v_cndmask_b32_e32 v6, v5, v8, vcc
; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_mov_b32_e32 v7, v3
-; SI-NEXT: v_mov_b32_e32 v6, v2
-; SI-NEXT: v_mov_b32_e32 v5, v1
-; SI-NEXT: v_mov_b32_e32 v4, v0
-; SI-NEXT: buffer_atomic_cmpswap_x2 v[4:7], off, s[8:11], 0 glc
+; SI-NEXT: v_mov_b32_e32 v0, v6
+; SI-NEXT: v_mov_b32_e32 v1, v7
+; SI-NEXT: v_mov_b32_e32 v2, v8
+; SI-NEXT: v_mov_b32_e32 v3, v9
+; SI-NEXT: buffer_atomic_cmpswap_x2 v[0:3], off, s[8:11], 0 glc
; SI-NEXT: s_waitcnt vmcnt(0)
; SI-NEXT: buffer_wbinvl1
-; SI-NEXT: v_cmp_eq_u64_e32 vcc, v[4:5], v[2:3]
+; SI-NEXT: v_cmp_eq_u64_e32 vcc, v[0:1], v[8:9]
; SI-NEXT: s_or_b64 s[0:1], vcc, s[0:1]
-; SI-NEXT: v_mov_b32_e32 v2, v4
-; SI-NEXT: v_mov_b32_e32 v3, v5
; SI-NEXT: s_andn2_b64 exec, exec, s[0:1]
; SI-NEXT: s_cbranch_execnz .LBB128_1
; SI-NEXT: ; %bb.2: ; %atomicrmw.end
@@ -9045,7 +9045,7 @@ define amdgpu_kernel void @atomic_min_i64_ret_addr64(ptr addrspace(1) %out, ptr
; SI-NEXT: s_mov_b32 s6, -1
; SI-NEXT: s_mov_b32 s4, s2
; SI-NEXT: s_mov_b32 s5, s3
-; SI-NEXT: buffer_store_dwordx2 v[4:5], off, s[4:7], 0
+; SI-NEXT: buffer_store_dwordx2 v[0:1], off, s[4:7], 0
; SI-NEXT: s_endpgm
;
; VI-LABEL: atomic_min_i64_ret_addr64:
@@ -9134,26 +9134,26 @@ define void @global_atomic_min_i64_noret_offset__amdgpu_no_remote_memory(ptr add
; SI-NEXT: s_mov_b32 s7, 0xf000
; SI-NEXT: s_mov_b32 s4, s6
; SI-NEXT: s_mov_b32 s5, s6
-; SI-NEXT: buffer_load_dwordx2 v[6:7], v[0:1], s[4:7], 0 addr64 offset:32
+; SI-NEXT: buffer_load_dwordx2 v[4:5], v[0:1], s[4:7], 0 addr64 offset:32
; SI-NEXT: s_mov_b64 s[8:9], 0
; SI-NEXT: .LBB129_1: ; %atomicrmw.start
; SI-NEXT: ; =>This Inner Loop Header: Depth=1
; SI-NEXT: s_waitcnt vmcnt(0)
-; SI-NEXT: v_cmp_le_i64_e32 vcc, v[6:7], v[2:3]
-; SI-NEXT: v_cndmask_b32_e32 v5, v3, v7, vcc
-; SI-NEXT: v_cndmask_b32_e32 v4, v2, v6, vcc
-; SI-NEXT: s_waitcnt expcnt(0)
-; SI-NEXT: v_mov_b32_e32 v11, v7
-; SI-NEXT: v_mov_b32_e32 v10, v6
; SI-NEXT: v_mov_b32_e32 v9, v5
; SI-NEXT: v_mov_b32_e32 v8, v4
-; SI-NEXT: buffer_atomic_cmpswap_x2 v[8:11], v[0:1], s[4:7], 0 addr64 offset:32 glc
+; SI-NEXT: v_cmp_le_i64_e32 vcc, v[8:9], v[2:3]
+; SI-NEXT: s_waitcnt expcnt(0)
+; SI-NEXT: v_cndmask_b32_e32 v7, v3, v9, vcc
+; SI-NEXT: v_cndmask_b32_e32 v6, v2, v8, vcc
+; SI-NEXT: v_mov_b32_e32 v4, v6
+; SI-NEXT: v_mov_b32_e32 v5, v7
+; SI-NEXT: v_mov_b32_e32 v6, v8
+; SI-NEXT: v_mov_b32_e32 v7, v9
+; SI-NEXT: buffer_atomic_cmpswap_x2 v[4:7], v[0:1], s[4:7], 0 addr64 offset:32 glc
; SI-NEXT: s_waitcnt vmcnt(0)
; SI-NEXT: buffer_wbinvl1
-; SI-NEXT: v_cmp_eq_u64_e32 vcc, v[8:9], v[6:7]
+; SI-NEXT: v_cmp_eq_u64_e32 vcc, v[4:5], v[8:9]
; SI-NEXT: s_or_b64 s[8:9], vcc, s[8:9]
-; SI-NEXT: v_mov_b32_e32 v6, v8
-; SI-NEXT: v_mov_b32_e32 v7, v9
; SI-NEXT: s_andn2_b64 exec, exec, s[8:9]
; SI-NEXT: s_cbranch_execnz .LBB129_1
; SI-NEXT: ; %bb.2: ; %atomicrmw.end
diff --git a/llvm/test/CodeGen/AMDGPU/global_atomics_scan_fadd.ll b/llvm/test/CodeGen/AMDGPU/global_atomics_scan_fadd.ll
index 82c58394c03bb..d17498a933be3 100644
--- a/llvm/test/CodeGen/AMDGPU/global_atomics_scan_fadd.ll
+++ b/llvm/test/CodeGen/AMDGPU/global_atomics_scan_fadd.ll
@@ -34,19 +34,19 @@ define amdgpu_kernel void @global_atomic_fadd_uni_address_uni_value_agent_scope_
; GFX7LESS-NEXT: v_cvt_f32_ubyte0_e32 v0, s2
; GFX7LESS-NEXT: v_mul_f32_e32 v2, 4.0, v0
; GFX7LESS-NEXT: s_waitcnt lgkmcnt(0)
-; GFX7LESS-NEXT: v_mov_b32_e32 v1, s6
+; GFX7LESS-NEXT: v_mov_b32_e32 v0, s6
; GFX7LESS-NEXT: s_mov_b32 s2, -1
; GFX7LESS-NEXT: .LBB0_2: ; %atomicrmw.start
; GFX7LESS-NEXT: ; =>This Inner Loop Header: Depth=1
-; GFX7LESS-NEXT: v_add_f32_e32 v0, v1, v2
+; GFX7LESS-NEXT: v_mov_b32_e32 v4, v0
+; GFX7LESS-NEXT: v_add_f32_e32 v3, v4, v2
; GFX7LESS-NEXT: s_waitcnt expcnt(0)
-; GFX7LESS-NEXT: v_mov_b32_e32 v4, v1
-; GFX7LESS-NEXT: v_mov_b32_e32 v3, v0
-; GFX7LESS-NEXT: buffer_atomic_cmpswap v[3:4], off, s[0:3], 0 glc
+; GFX7LESS-NEXT: v_mov_b32_e32 v0, v3
+; GFX7LESS-NEXT: v_mov_b32_e32 v1, v4
+; GFX7LESS-NEXT: buffer_atomic_cmpswap v[0:1], off, s[0:3], 0 glc
; GFX7LESS-NEXT: s_waitcnt vmcnt(0)
-; GFX7LESS-NEXT: v_cmp_eq_u32_e32 vcc, v3, v1
+; GFX7LESS-NEXT: v_cmp_eq_u32_e32 vcc, v0, v4
; GFX7LESS-NEXT: s_or_b64 s[4:5], vcc, s[4:5]
-; GFX7LESS-NEXT: v_mov_b32_e32 v1, v3
; GFX7LESS-NEXT: s_andn2_b64 exec, exec, s[4:5]
; GFX7LESS-NEXT: s_cbranch_execnz .LBB0_2
; GFX7LESS-NEXT: .LBB0_3:
@@ -205,19 +205,19 @@ define amdgpu_kernel void @global_atomic_fadd_uni_address_uni_value_agent_scope_
; GFX7LESS-DPP-NEXT: v_cvt_f32_ubyte0_e32 v0, s2
; GFX7LESS-DPP-NEXT: v_mul_f32_e32 v2, 4.0, v0
; GFX7LESS-DPP-NEXT: s_waitcnt lgkmcnt(0)
-; GFX7LESS-DPP-NEXT: v_mov_b32_e32 v1, s6
+; GFX7LESS-DPP-NEXT: v_mov_b32_e32 v0, s6
; GFX7LESS-DPP-NEXT: s_mov_b32 s2, -1
; GFX7LESS-DPP-NEXT: .LBB0_2: ; %atomicrmw.start
; GFX7LESS-DPP-NEXT: ; =>This Inner Loop Header: Depth=1
-; GFX7LESS-DPP-NEXT: v_add_f32_e32 v0, v1, v2
+; GFX7LESS-DPP-NEXT: v_mov_b32_e32 v4, v0
+; GFX7LESS-DPP-NEXT: v_add_f32_e32 v3, v4, v2
; GFX7LESS-DPP-NEXT: s_waitcnt expcnt(0)
-; GFX7LESS-DPP-NEXT: v_mov_b32_e32 v4, v1
-; GFX7LESS-DPP-NEXT: v_mov_b32_e32 v3, v0
-; GFX7LESS-DPP-NEXT: buffer_atomic_cmpswap v[3:4], off, s[0:3], 0 glc
+; GFX7LESS-DPP-NEXT: v_mov_b32_e32 v0, v3
+; GFX7LESS-DPP-NEXT: v_mov_b32_e32 v1, v4
+; GFX7LESS-DPP-NEXT: buffer_atomic_cmpswap v[0:1], off, s[0:3], 0 glc
; GFX7LESS-DPP-NEXT: s_waitcnt vmcnt(0)
-; GFX7LESS-DPP-NEXT: v_cmp_eq_u32_e32 vcc, v3, v1
+; GFX7LESS-DPP-NEXT: v_cmp_eq_u32_e32 vcc, v0, v4
; GFX7LESS-DPP-NEXT: s_or_b64 s[4:5], vcc, s[4:5]
-; GFX7LESS-DPP-NEXT: v_mov_b32_e32 v1, v3
; GFX7LESS-DPP-NEXT: s_andn2_b64 exec, exec, s[4:5]
; GFX7LESS-DPP-NEXT: s_cbranch_execnz .LBB0_2
; GFX7LESS-DPP-NEXT: .LBB0_3:
@@ -416,20 +416,20 @@ define amdgpu_kernel void @global_atomic_fadd_uni_address_div_value_agent_scope_
; GFX7LESS-NEXT: s_mov_b32 s3, 0xf000
; GFX7LESS-NEXT: s_mov_b32 s2, -1
; GFX7LESS-NEXT: s_waitcnt lgkmcnt(0)
-; GFX7LESS-NEXT: buffer_load_dword v1, off, s[0:3], 0
+; GFX7LESS-NEXT: buffer_load_dword v0, off, s[0:3], 0
; GFX7LESS-NEXT: s_mov_b64 s[4:5], 0
; GFX7LESS-NEXT: .LBB1_4: ; %atomicrmw.start
; GFX7LESS-NEXT: ; =>This Inner Loop Header: Depth=1
; GFX7LESS-NEXT: s_waitcnt vmcnt(0)
-; GFX7LESS-NEXT: v_add_f32_e32 v0, v1, v2
+; GFX7LESS-NEXT: v_mov_b32_e32 v4, v0
+; GFX7LESS-NEXT: v_add_f32_e32 v3, v4, v2
; GFX7LESS-NEXT: s_waitcnt expcnt(0)
-; GFX7LESS-NEXT: v_mov_b32_e32 v4, v1
-; GFX7LESS-NEXT: v_mov_b32_e32 v3, v0
-; GFX7LESS-NEXT: buffer_atomic_cmpswap v[3:4], off, s[0:3], 0 glc
+; GFX7LESS-NEXT: v_mov_b32_e32 v0, v3
+; GFX7LESS-NEXT: v_mov_b32_e32 v1, v4
+; GFX7LESS-NEXT: buffer_atomic_cmpswap v[0:1], off, s[0:3], 0 glc
; GFX7LESS-NEXT: s_waitcnt vmcnt(0)
-; GFX7LESS-NEXT: v_cmp_eq_u32_e32 vcc, v3, v1
+; GFX7LESS-NEXT: v_cmp_eq_u32_e32 vcc, v0, v4
; GFX7LESS-NEXT: s_or_b64 s[4:5], vcc, s[4:5]
-; GFX7LESS-NEXT: v_mov_b32_e32 v1, v3
; GFX7LESS-NEXT: s_andn2_b64 exec, exec, s[4:5]
; GFX7LESS-NEXT: s_cbranch_execnz .LBB1_4
; GFX7LESS-NEXT: .LBB1_5:
@@ -762,20 +762,20 @@ define amdgpu_kernel void @global_atomic_fadd_uni_address_div_value_agent_scope_
; GFX7LESS-DPP-NEXT: s_mov_b64 s[2:3], s[50:51]
; GFX7LESS-DPP-NEXT: s_waitcnt lgkmcnt(0)
; GFX7LESS-DPP-NEXT: s_swappc_b64 s[30:31], s[16:17]
-; GFX7LESS-DPP-NEXT: buffer_load_dword v2, off, s[36:39], 0
+; GFX7LESS-DPP-NEXT: buffer_load_dword v1, off, s[36:39], 0
; GFX7LESS-DPP-NEXT: s_mov_b64 s[0:1], 0
; GFX7LESS-DPP-NEXT: .LBB1_1: ; %atomicrmw.start
; GFX7LESS-DPP-NEXT: ; =>This Inner Loop Header: Depth=1
; GFX7LESS-DPP-NEXT: s_waitcnt vmcnt(0)
-; GFX7LESS-DPP-NEXT: v_add_f32_e32 v1, v2, v0
-; GFX7LESS-DPP-NEXT: s_waitcnt expcnt(0)
-; GFX7LESS-DPP-NEXT: v_mov_b32_e32 v4, v2
; GFX7LESS-DPP-NEXT: v_mov_b32_e32 v3, v1
-; GFX7LESS-DPP-NEXT: buffer_atomic_cmpswap v[3:4], off, s[36:39], 0 glc
+; GFX7LESS-DPP-NEXT: s_waitcnt expcnt(0)
+; GFX7LESS-DPP-NEXT: v_add_f32_e32 v2, v3, v0
+; GFX7LESS-DPP-NEXT: v_mov_b32_e32 v1, v2
+; GFX7LESS-DPP-NEXT: v_mov_b32_e32 v2, v3
+; GFX7LESS-DPP-NEXT: buffer_atomic_cmpswap v[1:2], off, s[36:39], 0 glc
; GFX7LESS-DPP-NEXT: s_waitcnt vmcnt(0)
-; GFX7LESS-DPP-NEXT: v_cmp_eq_u32_e32 vcc, v3, v2
+; GFX7LESS-DPP-NEXT: v_cmp_eq_u32_e32 vcc, v1, v3
; GFX7LESS-DPP-NEXT: s_or_b64 s[0:1], vcc, s[0:1]
-; GFX7LESS-DPP-NEXT: v_mov_b32_e32 v2, v3
; GFX7LESS-DPP-NEXT: s_andn2_b64 exec, exec, s[0:1]
; GFX7LESS-DPP-NEXT: s_cbranch_execnz .LBB1_1
; GFX7LESS-DPP-NEXT: ; %bb.2: ; %atomicrmw.end
@@ -1163,19 +1163,19 @@ define amdgpu_kernel void @global_atomic_fadd_uni_address_uni_value_one_as_scope
; GFX7LESS-NEXT: v_cvt_f32_f64_e32 v0, v[0:1]
; GFX7LESS-NEXT: v_mul_f32_e32 v2, 4.0, v0
; GFX7LESS-NEXT: s_waitcnt lgkmcnt(0)
-; GFX7LESS-NEXT: v_mov_b32_e32 v1, s2
+; GFX7LESS-NEXT: v_mov_b32_e32 v0, s2
; GFX7LESS-NEXT: s_mov_b32 s2, -1
; GFX7LESS-NEXT: .LBB2_2: ; %atomicrmw.start
; GFX7LESS-NEXT: ; =>This Inner Loop Header: Depth=1
-; GFX7LESS-NEXT: v_add_f32_e32 v0, v1, v2
+; GFX7LESS-NEXT: v_mov_b32_e32 v4, v0
+; GFX7LESS-NEXT: v_add_f32_e32 v3, v4, v2
; GFX7LESS-NEXT: s_waitcnt expcnt(0)
-; GFX7LESS-NEXT: v_mov_b32_e32 v4, v1
-; GFX7LESS-NEXT: v_mov_b32_e32 v3, v0
-; GFX7LESS-NEXT: buffer_atomic_cmpswap v[3:4], off, s[0:3], 0 glc
+; GFX7LESS-NEXT: v_mov_b32_e32 v0, v3
+; GFX7LESS-NEXT: v_mov_b32_e32 v1, v4
+; GFX7LESS-NEXT: buffer_atomic_cmpswap v[0:1], off, s[0:3], 0 glc
; GFX7LESS-NEXT: s_waitcnt vmcnt(0)
-; GFX7LESS-NEXT: v_cmp_eq_u32_e32 vcc, v3, v1
+; GFX7LESS-NEXT: v_cmp_eq_u32_e32 vcc, v0, v4
; GFX7LESS-NEXT: s_or_b64 s[4:5], vcc, s[4:5]
-; GFX7LESS-NEXT: v_mov_b32_e32 v1, v3
; GFX7LESS-NEXT: s_andn2_b64 exec, exec, s[4:5]
; GFX7LESS-NEXT: s_cbranch_execnz .LBB2_2
; GFX7LESS-NEXT: .LBB2_3:
@@ -1383,19 +1383,19 @@ define amdgpu_kernel void @global_atomic_fadd_uni_address_uni_value_one_as_scope
; GFX7LESS-DPP-NEXT: v_cvt_f32_f64_e32 v0, v[0:1]
; GFX7LESS-DPP-NEXT: v_mul_f32_e32 v2, 4.0, v0
; GFX7LESS-DPP-NEXT: s_waitcnt lgkmcnt(0)
-; GFX7LESS-DPP-NEXT: v_mov_b32_e32 v1, s2
+; GFX7LESS-DPP-NEXT: v_mov_b32_e32 v0, s2
; GFX7LESS-DPP-NEXT: s_mov_b32 s2, -1
; GFX7LESS-DPP-NEXT: .LBB2_2: ; %atomicrmw.start
; GFX7LESS-DPP-NEXT: ; =>This Inner Loop Header: Depth=1
-; GFX7LESS-DPP-NEXT: v_add_f32_e32 v0, v1, v2
+; GFX7LESS-DPP-NEXT: v_mov_b32_e32 v4, v0
+; GFX7LESS-DPP-NEXT: v_add_f32_e32 v3, v4, v2
; GFX7LESS-DPP-NEXT: s_waitcnt expcnt(0)
-; GFX7LESS-DPP-NEXT: v_mov_b32_e32 v4, v1
-; GFX7LESS-DPP-NEXT: v_mov_b32_e32 v3, v0
-; GFX7LESS-DPP-NEXT: buffer_atomic_cmpswap v[3:4], off, s[0:3], 0 glc
+; GFX7LESS-DPP-NEXT: v_mov_b32_e32 v0, v3
+; GFX7LESS-DPP-NEXT: v_mov_b32_e32 v1, v4
+; GFX7LESS-DPP-NEXT: buffer_atomic_cmpswap v[0:1], off, s[0:3], 0 glc
; GFX7LESS-DPP-NEXT: s_waitcnt vmcnt(0)
-; GFX7LESS-DPP-NEXT: v_cmp_eq_u32_e32 vcc, v3, v1
+; GFX7LESS-DPP-NEXT: v_cmp_eq_u32_e32 vcc, v0, v4
; GFX7LESS-DPP-NEXT: s_or_b64 s[4:5], vcc, s[4:5]
-; GFX7LESS-DPP-NEXT: v_mov_b32_e32 v1, v3
; GFX7LESS-DPP-NEXT: s_andn2_b64 exec, exec, s[4:5]
; GFX7LESS-DPP-NEXT: s_cbranch_execnz .LBB2_2
; GFX7LESS-DPP-NEXT: .LBB2_3:
@@ -1634,20 +1634,20 @@ define amdgpu_kernel void @global_atomic_fadd_uni_address_div_value_one_as_scope
; GFX7LESS-NEXT: s_mov_b32 s3, 0xf000
; GFX7LESS-NEXT: s_mov_b32 s2, -1
; GFX7LESS-NEXT: s_waitcnt lgkmcnt(0)
-; GFX7LESS-NEXT: buffer_load_dword v1, off, s[0:3], 0
+; GFX7LESS-NEXT: buffer_load_dword v0, off, s[0:3], 0
; GFX7LESS-NEXT: s_mov_b64 s[4:5], 0
; GFX7LESS-NEXT: .LBB3_4: ; %atomicrmw.start
; GFX7LESS-NEXT: ; =>This Inner Loop Header: Depth=1
; GFX7LESS-NEXT: s_waitcnt vmcnt(0)
-; GFX7LESS-NEXT: v_add_f32_e32 v0, v1, v2
+; GFX7LESS-NEXT: v_mov_b32_e32 v4, v0
+; GFX7LESS-NEXT: v_add_f32_e32 v3, v4, v2
; GFX7LESS-NEXT: s_waitcnt expcnt(0)
-; GFX7LESS-NEXT: v_mov_b32_e32 v4, v1
-; GFX7LESS-NEXT: v_mov_b32_e32 v3, v0
-; GFX7LESS-NEXT: buffer_atomic_cmpswap v[3:4], off, s[0:3], 0 glc
+; GFX7LESS-NEXT: v_mov_b32_e32 v0, v3
+; GFX7LESS-NEXT: v_mov_b32_e32 v1, v4
+; GFX7LESS-NEXT: buffer_atomic_cmpswap v[0:1], off, s[0:3], 0 glc
; GFX7LESS-NEXT: s_waitcnt vmcnt(0)
-; GFX7LESS-NEXT: v_cmp_eq_u32_e32 vcc, v3, v1
+; GFX7LESS-NEXT: v_cmp_eq_u32_e32 vcc, v0, v4
; GFX7LESS-NEXT: s_or_b64 s[4:5], vcc, s[4:5]
-; GFX7LESS-NEXT: v_mov_b32_e32 v1, v3
; GFX7LESS-NEXT: s_andn2_b64 exec, exec, s[4:5]
; GFX7LESS-NEXT: s_cbranch_execnz .LBB3_4
; GFX7LESS-NEXT: .LBB3_5:
@@ -1980,20 +1980,20 @@ define amdgpu_kernel void @global_atomic_fadd_uni_address_div_value_one_as_scope
; GFX7LESS-DPP-NEXT: s_mov_b64 s[2:3], s[50:51]
; GFX7LESS-DPP-NEXT: s_waitcnt lgkmcnt(0)
; GFX7LESS-DPP-NEXT: s_swappc_b64 s[30:31], s[16:17]
-; GFX7LESS-DPP-NEXT: buffer_load_dword v2, off, s[36:39], 0
+; GFX7LESS-DPP-NEXT: buffer_load_dword v1, off, s[36:39], 0
; GFX7LESS-DPP-NEXT: s_mov_b64 s[0:1], 0
; GFX7LESS-DPP-NEXT: .LBB3_1: ; %atomicrmw.start
; GFX7LESS-DPP-NEXT: ; =>This Inner Loop Header: Depth=1
; GFX7LESS-DPP-NEXT: s_waitcnt vmcnt(0)
-; GFX7LESS-DPP-NEXT: v_add_f32_e32 v1, v2, v0
-; GFX7LESS-DPP-NEXT: s_waitcnt expcnt(0)
-; GFX7LESS-DPP-NEXT: v_mov_b32_e32 v4, v2
; GFX7LESS-DPP-NEXT: v_mov_b32_e32 v3, v1
-; GFX7LESS-DPP-NEXT: buffer_atomic_cmpswap v[3:4], off, s[36:39], 0 glc
+; GFX7LESS-DPP-NEXT: s_waitcnt expcnt(0)
+; GFX7LESS-DPP-NEXT: v_add_f32_e32 v2, v3, v0
+; GFX7LESS-DPP-NEXT: v_mov_b32_e32 v1, v2
+; GFX7LESS-DPP-NEXT: v_mov_b32_e32 v2, v3
+; GFX7LESS-DPP-NEXT: buffer_atomic_cmpswap v[1:2], off, s[36:39], 0 glc
; GFX7LESS-DPP-NEXT: s_waitcnt vmcnt(0)
-; GFX7LESS-DPP-NEXT: v_cmp_eq_u32_e32 vcc, v3, v2
+; GFX7LESS-DPP-NEXT: v_cmp_eq_u32_e32 vcc, v1, v3
; GFX7LESS-DPP-NEXT: s_or_b64 s[0:1], vcc, s[0:1]
-; GFX7LESS-DPP-NEXT: v_mov_b32_e32 v2, v3
; GFX7LESS-DPP-NEXT: s_andn2_b64 exec, exec, s[0:1]
; GFX7LESS-DPP-NEXT: s_cbranch_execnz .LBB3_1
; GFX7LESS-DPP-NEXT: ; %bb.2: ; %atomicrmw.end
@@ -2381,19 +2381,19 @@ define amdgpu_kernel void @global_atomic_fadd_uni_address_uni_value_agent_scope_
; GFX7LESS-NEXT: v_cvt_f32_f64_e32 v0, v[0:1]
; GFX7LESS-NEXT: v_mul_f32_e32 v2, 4.0, v0
; GFX7LESS-NEXT: s_waitcnt lgkmcnt(0)
-; GFX7LESS-NEXT: v_mov_b32_e32 v1, s2
+; GFX7LESS-NEXT: v_mov_b32_e32 v0, s2
; GFX7LESS-NEXT: s_mov_b32 s2, -1
; GFX7LESS-NEXT: .LBB4_2: ; %atomicrmw.start
; GFX7LESS-NEXT: ; =>This Inner Loop Header: Depth=1
-; GFX7LESS-NEXT: v_add_f32_e32 v0, v1, v2
+; GFX7LESS-NEXT: v_mov_b32_e32 v4, v0
+; GFX7LESS-NEXT: v_add_f32_e32 v3, v4, v2
; GFX7LESS-NEXT: s_waitcnt expcnt(0)
-; GFX7LESS-NEXT: v_mov_b32_e32 v4, v1
-; GFX7LESS-NEXT: v_mov_b32_e32 v3, v0
-; GFX7LESS-NEXT: buffer_atomic_cmpswap v[3:4], off, s[0:3], 0 glc
+; GFX7LESS-NEXT: v_mov_b32_e32 v0, v3
+; GFX7LESS-NEXT: v_mov_b32_e32 v1, v4
+; GFX7LESS-NEXT: buffer_atomic_cmpswap v[0:1], off, s[0:3], 0 glc
; GFX7LESS-NEXT: s_waitcnt vmcnt(0)
-; GFX7LESS-NEXT: v_cmp_eq_u32_e32 vcc, v3, v1
+; GFX7LESS-NEXT: v_cmp_eq_u32_e32 vcc, v0, v4
; GFX7LESS-NEXT: s_or_b64 s[4:5], vcc, s[4:5]
-; GFX7LESS-NEXT: v_mov_b32_e32 v1, v3
; GFX7LESS-NEXT: s_andn2_b64 exec, exec, s[4:5]
; GFX7LESS-NEXT: s_cbranch_execnz .LBB4_2
; GFX7LESS-NEXT: .LBB4_3:
@@ -2631,19 +2631,19 @@ define amdgpu_kernel void @global_atomic_fadd_uni_address_uni_value_agent_scope_
; GFX7LESS-DPP-NEXT: v_cvt_f32_f64_e32 v0, v[0:1]
; GFX7LESS-DPP-NEXT: v_mul_f32_e32 v2, 4.0, v0
; GFX7LESS-DPP-NEXT: s_waitcnt lgkmcnt(0)
-; GFX7LESS-DPP-NEXT: v_mov_b32_e32 v1, s2
+; GFX7LESS-DPP-NEXT: v_mov_b32_e32 v0, s2
; GFX7LESS-DPP-NEXT: s_mov_b32 s2, -1
; GFX7LESS-DPP-NEXT: .LBB4_2: ; %atomicrmw.start
; GFX7LESS-DPP-NEXT: ; =>This Inner Loop Header: Depth=1
-; GFX7LESS-DPP-NEXT: v_add_f32_e32 v0, v1, v2
+; GFX7LESS-DPP-NEXT: v_mov_b32_e32 v4, v0
+; GFX7LESS-DPP-NEXT: v_add_f32_e32 v3, v4, v2
; GFX7LESS-DPP-NEXT: s_waitcnt expcnt(0)
-; GFX7LESS-DPP-NEXT: v_mov_b32_e32 v4, v1
-; GFX7LESS-DPP-NEXT: v_mov_b32_e32 v3, v0
-; GFX7LESS-DPP-NEXT: buffer_atomic_cmpswap v[3:4], off, s[0:3], 0 glc
+; GFX7LESS-DPP-NEXT: v_mov_b32_e32 v0, v3
+; GFX7LESS-DPP-NEXT: v_mov_b32_e32 v1, v4
+; GFX7LESS-DPP-NEXT: buffer_atomic_cmpswap v[0:1], off, s[0:3], 0 glc
; GFX7LESS-DPP-NEXT: s_waitcnt vmcnt(0)
-; GFX7LESS-DPP-NEXT: v_cmp_eq_u32_e32 vcc, v3, v1
+; GFX7LESS-DPP-NEXT: v_cmp_eq_u32_e32 vcc, v0, v4
; GFX7LESS-DPP-NEXT: s_or_b64 s[4:5], vcc, s[4:5]
-; GFX7LESS-DPP-NEXT: v_mov_b32_e32 v1, v3
; GFX7LESS-DPP-NEXT: s_andn2_b64 exec, exec, s[4:5]
; GFX7LESS-DPP-NEXT: s_cbranch_execnz .LBB4_2
; GFX7LESS-DPP-NEXT: .LBB4_3:
@@ -2912,20 +2912,20 @@ define amdgpu_kernel void @global_atomic_fadd_uni_address_div_value_agent_scope_
; GFX7LESS-NEXT: s_mov_b32 s3, 0xf000
; GFX7LESS-NEXT: s_mov_b32 s2, -1
; GFX7LESS-NEXT: s_waitcnt lgkmcnt(0)
-; GFX7LESS-NEXT: buffer_load_dword v1, off, s[0:3], 0
+; GFX7LESS-NEXT: buffer_load_dword v0, off, s[0:3], 0
; GFX7LESS-NEXT: s_mov_b64 s[4:5], 0
; GFX7LESS-NEXT: .LBB5_4: ; %atomicrmw.start
; GFX7LESS-NEXT: ; =>This Inner Loop Header: Depth=1
; GFX7LESS-NEXT: s_waitcnt vmcnt(0)
-; GFX7LESS-NEXT: v_add_f32_e32 v0, v1, v2
+; GFX7LESS-NEXT: v_mov_b32_e32 v4, v0
+; GFX7LESS-NEXT: v_add_f32_e32 v3, v4, v2
; GFX7LESS-NEXT: s_waitcnt expcnt(0)
-; GFX7LESS-NEXT: v_mov_b32_e32 v4, v1
-; GFX7LESS-NEXT: v_mov_b32_e32 v3, v0
-; GFX7LESS-NEXT: buffer_atomic_cmpswap v[3:4], off, s[0:3], 0 glc
+; GFX7LESS-NEXT: v_mov_b32_e32 v0, v3
+; GFX7LESS-NEXT: v_mov_b32_e32 v1, v4
+; GFX7LESS-NEXT: buffer_atomic_cmpswap v[0:1], off, s[0:3], 0 glc
; GFX7LESS-NEXT: s_waitcnt vmcnt(0)
-; GFX7LESS-NEXT: v_cmp_eq_u32_e32 vcc, v3, v1
+; GFX7LESS-NEXT: v_cmp_eq_u32_e32 vcc, v0, v4
; GFX7LESS-NEXT: s_or_b64 s[4:5], vcc, s[4:5]
-; GFX7LESS-NEXT: v_mov_b32_e32 v1, v3
; GFX7LESS-NEXT: s_andn2_b64 exec, exec, s[4:5]
; GFX7LESS-NEXT: s_cbranch_execnz .LBB5_4
; GFX7LESS-NEXT: .LBB5_5:
@@ -3258,20 +3258,20 @@ define amdgpu_kernel void @global_atomic_fadd_uni_address_div_value_agent_scope_
; GFX7LESS-DPP-NEXT: s_mov_b64 s[2:3], s[50:51]
; GFX7LESS-DPP-NEXT: s_waitcnt lgkmcnt(0)
; GFX7LESS-DPP-NEXT: s_swappc_b64 s[30:31], s[16:17]
-; GFX7LESS-DPP-NEXT: buffer_load_dword v2, off, s[36:39], 0
+; GFX7LESS-DPP-NEXT: buffer_load_dword v1, off, s[36:39], 0
; GFX7LESS-DPP-NEXT: s_mov_b64 s[0:1], 0
; GFX7LESS-DPP-NEXT: .LBB5_1: ; %atomicrmw.start
; GFX7LESS-DPP-NEXT: ; =>This Inner Loop Header: Depth=1
; GFX7LESS-DPP-NEXT: s_waitcnt vmcnt(0)
-; GFX7LESS-DPP-NEXT: v_add_f32_e32 v1, v2, v0
-; GFX7LESS-DPP-NEXT: s_waitcnt expcnt(0)
-; GFX7LESS-DPP-NEXT: v_mov_b32_e32 v4, v2
; GFX7LESS-DPP-NEXT: v_mov_b32_e32 v3, v1
-; GFX7LESS-DPP-NEXT: buffer_atomic_cmpswap v[3:4], off, s[36:39], 0 glc
+; GFX7LESS-DPP-NEXT: s_waitcnt expcnt(0)
+; GFX7LESS-DPP-NEXT: v_add_f32_e32 v2, v3, v0
+; GFX7LESS-DPP-NEXT: v_mov_b32_e32 v1, v2
+; GFX7LESS-DPP-NEXT: v_mov_b32_e32 v2, v3
+; GFX7LESS-DPP-NEXT: buffer_atomic_cmpswap v[1:2], off, s[36:39], 0 glc
; GFX7LESS-DPP-NEXT: s_waitcnt vmcnt(0)
-; GFX7LESS-DPP-NEXT: v_cmp_eq_u32_e32 vcc, v3, v2
+; GFX7LESS-DPP-NEXT: v_cmp_eq_u32_e32 vcc, v1, v3
; GFX7LESS-DPP-NEXT: s_or_b64 s[0:1], vcc, s[0:1]
-; GFX7LESS-DPP-NEXT: v_mov_b32_e32 v2, v3
; GFX7LESS-DPP-NEXT: s_andn2_b64 exec, exec, s[0:1]
; GFX7LESS-DPP-NEXT: s_cbranch_execnz .LBB5_1
; GFX7LESS-DPP-NEXT: ; %bb.2: ; %atomicrmw.end
@@ -3686,20 +3686,20 @@ define amdgpu_kernel void @global_atomic_fadd_uni_address_div_value_agent_scope_
; GFX7LESS-NEXT: s_mov_b32 s3, 0xf000
; GFX7LESS-NEXT: s_mov_b32 s2, -1
; GFX7LESS-NEXT: s_waitcnt lgkmcnt(0)
-; GFX7LESS-NEXT: buffer_load_dword v1, off, s[0:3], 0
+; GFX7LESS-NEXT: buffer_load_dword v0, off, s[0:3], 0
; GFX7LESS-NEXT: s_mov_b64 s[4:5], 0
; GFX7LESS-NEXT: .LBB6_4: ; %atomicrmw.start
; GFX7LESS-NEXT: ; =>This Inner Loop Header: Depth=1
; GFX7LESS-NEXT: s_waitcnt vmcnt(0)
-; GFX7LESS-NEXT: v_add_f32_e32 v0, v1, v2
+; GFX7LESS-NEXT: v_mov_b32_e32 v4, v0
+; GFX7LESS-NEXT: v_add_f32_e32 v3, v4, v2
; GFX7LESS-NEXT: s_waitcnt expcnt(0)
-; GFX7LESS-NEXT: v_mov_b32_e32 v4, v1
-; GFX7LESS-NEXT: v_mov_b32_e32 v3, v0
-; GFX7LESS-NEXT: buffer_atomic_cmpswap v[3:4], off, s[0:3], 0 glc
+; GFX7LESS-NEXT: v_mov_b32_e32 v0, v3
+; GFX7LESS-NEXT: v_mov_b32_e32 v1, v4
+; GFX7LESS-NEXT: buffer_atomic_cmpswap v[0:1], off, s[0:3], 0 glc
; GFX7LESS-NEXT: s_waitcnt vmcnt(0)
-; GFX7LESS-NEXT: v_cmp_eq_u32_e32 vcc, v3, v1
+; GFX7LESS-NEXT: v_cmp_eq_u32_e32 vcc, v0, v4
; GFX7LESS-NEXT: s_or_b64 s[4:5], vcc, s[4:5]
-; GFX7LESS-NEXT: v_mov_b32_e32 v1, v3
; GFX7LESS-NEXT: s_andn2_b64 exec, exec, s[4:5]
; GFX7LESS-NEXT: s_cbranch_execnz .LBB6_4
; GFX7LESS-NEXT: .LBB6_5:
@@ -4032,20 +4032,20 @@ define amdgpu_kernel void @global_atomic_fadd_uni_address_div_value_agent_scope_
; GFX7LESS-DPP-NEXT: s_mov_b64 s[2:3], s[50:51]
; GFX7LESS-DPP-NEXT: s_waitcnt lgkmcnt(0)
; GFX7LESS-DPP-NEXT: s_swappc_b64 s[30:31], s[16:17]
-; GFX7LESS-DPP-NEXT: buffer_load_dword v2, off, s[36:39], 0
+; GFX7LESS-DPP-NEXT: buffer_load_dword v1, off, s[36:39], 0
; GFX7LESS-DPP-NEXT: s_mov_b64 s[0:1], 0
; GFX7LESS-DPP-NEXT: .LBB6_1: ; %atomicrmw.start
; GFX7LESS-DPP-NEXT: ; =>This Inner Loop Header: Depth=1
; GFX7LESS-DPP-NEXT: s_waitcnt vmcnt(0)
-; GFX7LESS-DPP-NEXT: v_add_f32_e32 v1, v2, v0
-; GFX7LESS-DPP-NEXT: s_waitcnt expcnt(0)
-; GFX7LESS-DPP-NEXT: v_mov_b32_e32 v4, v2
; GFX7LESS-DPP-NEXT: v_mov_b32_e32 v3, v1
-; GFX7LESS-DPP-NEXT: buffer_atomic_cmpswap v[3:4], off, s[36:39], 0 glc
+; GFX7LESS-DPP-NEXT: s_waitcnt expcnt(0)
+; GFX7LESS-DPP-NEXT: v_add_f32_e32 v2, v3, v0
+; GFX7LESS-DPP-NEXT: v_mov_b32_e32 v1, v2
+; GFX7LESS-DPP-NEXT: v_mov_b32_e32 v2, v3
+; GFX7LESS-DPP-NEXT: buffer_atomic_cmpswap v[1:2], off, s[36:39], 0 glc
; GFX7LESS-DPP-NEXT: s_waitcnt vmcnt(0)
-; GFX7LESS-DPP-NEXT: v_cmp_eq_u32_e32 vcc, v3, v2
+; GFX7LESS-DPP-NEXT: v_cmp_eq_u32_e32 vcc, v1, v3
; GFX7LESS-DPP-NEXT: s_or_b64 s[0:1], vcc, s[0:1]
-; GFX7LESS-DPP-NEXT: v_mov_b32_e32 v2, v3
; GFX7LESS-DPP-NEXT: s_andn2_b64 exec, exec, s[0:1]
; GFX7LESS-DPP-NEXT: s_cbranch_execnz .LBB6_1
; GFX7LESS-DPP-NEXT: ; %bb.2: ; %atomicrmw.end
@@ -4433,19 +4433,19 @@ define amdgpu_kernel void @global_atomic_fadd_uni_address_uni_value_default_scop
; GFX7LESS-NEXT: v_cvt_f32_f64_e32 v0, v[0:1]
; GFX7LESS-NEXT: v_mul_f32_e32 v2, 4.0, v0
; GFX7LESS-NEXT: s_waitcnt lgkmcnt(0)
-; GFX7LESS-NEXT: v_mov_b32_e32 v1, s2
+; GFX7LESS-NEXT: v_mov_b32_e32 v0, s2
; GFX7LESS-NEXT: s_mov_b32 s2, -1
; GFX7LESS-NEXT: .LBB7_2: ; %atomicrmw.start
; GFX7LESS-NEXT: ; =>This Inner Loop Header: Depth=1
-; GFX7LESS-NEXT: v_add_f32_e32 v0, v1, v2
+; GFX7LESS-NEXT: v_mov_b32_e32 v4, v0
+; GFX7LESS-NEXT: v_add_f32_e32 v3, v4, v2
; GFX7LESS-NEXT: s_waitcnt expcnt(0)
-; GFX7LESS-NEXT: v_mov_b32_e32 v4, v1
-; GFX7LESS-NEXT: v_mov_b32_e32 v3, v0
-; GFX7LESS-NEXT: buffer_atomic_cmpswap v[3:4], off, s[0:3], 0 glc
+; GFX7LESS-NEXT: v_mov_b32_e32 v0, v3
+; GFX7LESS-NEXT: v_mov_b32_e32 v1, v4
+; GFX7LESS-NEXT: buffer_atomic_cmpswap v[0:1], off, s[0:3], 0 glc
; GFX7LESS-NEXT: s_waitcnt vmcnt(0)
-; GFX7LESS-NEXT: v_cmp_eq_u32_e32 vcc, v3, v1
+; GFX7LESS-NEXT: v_cmp_eq_u32_e32 vcc, v0, v4
; GFX7LESS-NEXT: s_or_b64 s[4:5], vcc, s[4:5]
-; GFX7LESS-NEXT: v_mov_b32_e32 v1, v3
; GFX7LESS-NEXT: s_andn2_b64 exec, exec, s[4:5]
; GFX7LESS-NEXT: s_cbranch_execnz .LBB7_2
; GFX7LESS-NEXT: .LBB7_3:
@@ -4683,19 +4683,19 @@ define amdgpu_kernel void @global_atomic_fadd_uni_address_uni_value_default_scop
; GFX7LESS-DPP-NEXT: v_cvt_f32_f64_e32 v0, v[0:1]
; GFX7LESS-DPP-NEXT: v_mul_f32_e32 v2, 4.0, v0
; GFX7LESS-DPP-NEXT: s_waitcnt lgkmcnt(0)
-; GFX7LESS-DPP-NEXT: v_mov_b32_e32 v1, s2
+; GFX7LESS-DPP-NEXT: v_mov_b32_e32 v0, s2
; GFX7LESS-DPP-NEXT: s_mov_b32 s2, -1
; GFX7LESS-DPP-NEXT: .LBB7_2: ; %atomicrmw.start
; GFX7LESS-DPP-NEXT: ; =>This Inner Loop Header: Depth=1
-; GFX7LESS-DPP-NEXT: v_add_f32_e32 v0, v1, v2
+; GFX7LESS-DPP-NEXT: v_mov_b32_e32 v4, v0
+; GFX7LESS-DPP-NEXT: v_add_f32_e32 v3, v4, v2
; GFX7LESS-DPP-NEXT: s_waitcnt expcnt(0)
-; GFX7LESS-DPP-NEXT: v_mov_b32_e32 v4, v1
-; GFX7LESS-DPP-NEXT: v_mov_b32_e32 v3, v0
-; GFX7LESS-DPP-NEXT: buffer_atomic_cmpswap v[3:4], off, s[0:3], 0 glc
+; GFX7LESS-DPP-NEXT: v_mov_b32_e32 v0, v3
+; GFX7LESS-DPP-NEXT: v_mov_b32_e32 v1, v4
+; GFX7LESS-DPP-NEXT: buffer_atomic_cmpswap v[0:1], off, s[0:3], 0 glc
; GFX7LESS-DPP-NEXT: s_waitcnt vmcnt(0)
-; GFX7LESS-DPP-NEXT: v_cmp_eq_u32_e32 vcc, v3, v1
+; GFX7LESS-DPP-NEXT: v_cmp_eq_u32_e32 vcc, v0, v4
; GFX7LESS-DPP-NEXT: s_or_b64 s[4:5], vcc, s[4:5]
-; GFX7LESS-DPP-NEXT: v_mov_b32_e32 v1, v3
; GFX7LESS-DPP-NEXT: s_andn2_b64 exec, exec, s[4:5]
; GFX7LESS-DPP-NEXT: s_cbranch_execnz .LBB7_2
; GFX7LESS-DPP-NEXT: .LBB7_3:
@@ -4963,20 +4963,20 @@ define amdgpu_kernel void @global_atomic_fadd_uni_address_div_value_default_scop
; GFX7LESS-NEXT: s_mov_b32 s3, 0xf000
; GFX7LESS-NEXT: s_mov_b32 s2, -1
; GFX7LESS-NEXT: s_waitcnt lgkmcnt(0)
-; GFX7LESS-NEXT: buffer_load_dword v1, off, s[0:3], 0
+; GFX7LESS-NEXT: buffer_load_dword v0, off, s[0:3], 0
; GFX7LESS-NEXT: s_mov_b64 s[4:5], 0
; GFX7LESS-NEXT: .LBB8_4: ; %atomicrmw.start
; GFX7LESS-NEXT: ; =>This Inner Loop Header: Depth=1
; GFX7LESS-NEXT: s_waitcnt vmcnt(0)
-; GFX7LESS-NEXT: v_add_f32_e32 v0, v1, v2
+; GFX7LESS-NEXT: v_mov_b32_e32 v4, v0
+; GFX7LESS-NEXT: v_add_f32_e32 v3, v4, v2
; GFX7LESS-NEXT: s_waitcnt expcnt(0)
-; GFX7LESS-NEXT: v_mov_b32_e32 v4, v1
-; GFX7LESS-NEXT: v_mov_b32_e32 v3, v0
-; GFX7LESS-NEXT: buffer_atomic_cmpswap v[3:4], off, s[0:3], 0 glc
+; GFX7LESS-NEXT: v_mov_b32_e32 v0, v3
+; GFX7LESS-NEXT: v_mov_b32_e32 v1, v4
+; GFX7LESS-NEXT: buffer_atomic_cmpswap v[0:1], off, s[0:3], 0 glc
; GFX7LESS-NEXT: s_waitcnt vmcnt(0)
-; GFX7LESS-NEXT: v_cmp_eq_u32_e32 vcc, v3, v1
+; GFX7LESS-NEXT: v_cmp_eq_u32_e32 vcc, v0, v4
; GFX7LESS-NEXT: s_or_b64 s[4:5], vcc, s[4:5]
-; GFX7LESS-NEXT: v_mov_b32_e32 v1, v3
; GFX7LESS-NEXT: s_andn2_b64 exec, exec, s[4:5]
; GFX7LESS-NEXT: s_cbranch_execnz .LBB8_4
; GFX7LESS-NEXT: .LBB8_5:
@@ -5335,20 +5335,20 @@ define amdgpu_kernel void @global_atomic_fadd_uni_address_div_value_default_scop
; GFX7LESS-DPP-NEXT: s_mov_b64 s[2:3], s[50:51]
; GFX7LESS-DPP-NEXT: s_waitcnt lgkmcnt(0)
; GFX7LESS-DPP-NEXT: s_swappc_b64 s[30:31], s[16:17]
-; GFX7LESS-DPP-NEXT: buffer_load_dword v2, off, s[36:39], 0
+; GFX7LESS-DPP-NEXT: buffer_load_dword v1, off, s[36:39], 0
; GFX7LESS-DPP-NEXT: s_mov_b64 s[0:1], 0
; GFX7LESS-DPP-NEXT: .LBB8_1: ; %atomicrmw.start
; GFX7LESS-DPP-NEXT: ; =>This Inner Loop Header: Depth=1
; GFX7LESS-DPP-NEXT: s_waitcnt vmcnt(0)
-; GFX7LESS-DPP-NEXT: v_add_f32_e32 v1, v2, v0
-; GFX7LESS-DPP-NEXT: s_waitcnt expcnt(0)
-; GFX7LESS-DPP-NEXT: v_mov_b32_e32 v4, v2
; GFX7LESS-DPP-NEXT: v_mov_b32_e32 v3, v1
-; GFX7LESS-DPP-NEXT: buffer_atomic_cmpswap v[3:4], off, s[36:39], 0 glc
+; GFX7LESS-DPP-NEXT: s_waitcnt expcnt(0)
+; GFX7LESS-DPP-NEXT: v_add_f32_e32 v2, v3, v0
+; GFX7LESS-DPP-NEXT: v_mov_b32_e32 v1, v2
+; GFX7LESS-DPP-NEXT: v_mov_b32_e32 v2, v3
+; GFX7LESS-DPP-NEXT: buffer_atomic_cmpswap v[1:2], off, s[36:39], 0 glc
; GFX7LESS-DPP-NEXT: s_waitcnt vmcnt(0)
-; GFX7LESS-DPP-NEXT: v_cmp_eq_u32_e32 vcc, v3, v2
+; GFX7LESS-DPP-NEXT: v_cmp_eq_u32_e32 vcc, v1, v3
; GFX7LESS-DPP-NEXT: s_or_b64 s[0:1], vcc, s[0:1]
-; GFX7LESS-DPP-NEXT: v_mov_b32_e32 v2, v3
; GFX7LESS-DPP-NEXT: s_andn2_b64 exec, exec, s[0:1]
; GFX7LESS-DPP-NEXT: s_cbranch_execnz .LBB8_1
; GFX7LESS-DPP-NEXT: ; %bb.2: ; %atomicrmw.end
@@ -8004,23 +8004,23 @@ define amdgpu_kernel void @global_atomic_fadd_double_uni_address_uni_value_one_a
; GFX7LESS-NEXT: v_add_f64 v[0:1], s[6:7], v[0:1]
; GFX7LESS-NEXT: v_mul_f64 v[4:5], 4.0, v[0:1]
; GFX7LESS-NEXT: s_waitcnt lgkmcnt(0)
-; GFX7LESS-NEXT: v_mov_b32_e32 v2, s8
-; GFX7LESS-NEXT: v_mov_b32_e32 v3, s9
+; GFX7LESS-NEXT: v_mov_b32_e32 v0, s8
+; GFX7LESS-NEXT: v_mov_b32_e32 v1, s9
; GFX7LESS-NEXT: s_mov_b32 s2, -1
; GFX7LESS-NEXT: .LBB11_2: ; %atomicrmw.start
; GFX7LESS-NEXT: ; =>This Inner Loop Header: Depth=1
-; GFX7LESS-NEXT: v_add_f64 v[0:1], v[2:3], v[4:5]
+; GFX7LESS-NEXT: v_mov_b32_e32 v9, v1
+; GFX7LESS-NEXT: v_mov_b32_e32 v8, v0
+; GFX7LESS-NEXT: v_add_f64 v[6:7], v[8:9], v[4:5]
; GFX7LESS-NEXT: s_waitcnt expcnt(0)
-; GFX7LESS-NEXT: v_mov_b32_e32 v9, v3
-; GFX7LESS-NEXT: v_mov_b32_e32 v8, v2
-; GFX7LESS-NEXT: v_mov_b32_e32 v7, v1
-; GFX7LESS-NEXT: v_mov_b32_e32 v6, v0
-; GFX7LESS-NEXT: buffer_atomic_cmpswap_x2 v[6:9], off, s[0:3], 0 glc
+; GFX7LESS-NEXT: v_mov_b32_e32 v0, v6
+; GFX7LESS-NEXT: v_mov_b32_e32 v1, v7
+; GFX7LESS-NEXT: v_mov_b32_e32 v2, v8
+; GFX7LESS-NEXT: v_mov_b32_e32 v3, v9
+; GFX7LESS-NEXT: buffer_atomic_cmpswap_x2 v[0:3], off, s[0:3], 0 glc
; GFX7LESS-NEXT: s_waitcnt vmcnt(0)
-; GFX7LESS-NEXT: v_cmp_eq_u64_e32 vcc, v[6:7], v[2:3]
+; GFX7LESS-NEXT: v_cmp_eq_u64_e32 vcc, v[0:1], v[8:9]
; GFX7LESS-NEXT: s_or_b64 s[4:5], vcc, s[4:5]
-; GFX7LESS-NEXT: v_mov_b32_e32 v2, v6
-; GFX7LESS-NEXT: v_mov_b32_e32 v3, v7
; GFX7LESS-NEXT: s_andn2_b64 exec, exec, s[4:5]
; GFX7LESS-NEXT: s_cbranch_execnz .LBB11_2
; GFX7LESS-NEXT: .LBB11_3:
@@ -8261,23 +8261,23 @@ define amdgpu_kernel void @global_atomic_fadd_double_uni_address_uni_value_one_a
; GFX7LESS-DPP-NEXT: v_add_f64 v[0:1], s[6:7], v[0:1]
; GFX7LESS-DPP-NEXT: v_mul_f64 v[4:5], 4.0, v[0:1]
; GFX7LESS-DPP-NEXT: s_waitcnt lgkmcnt(0)
-; GFX7LESS-DPP-NEXT: v_mov_b32_e32 v2, s8
-; GFX7LESS-DPP-NEXT: v_mov_b32_e32 v3, s9
+; GFX7LESS-DPP-NEXT: v_mov_b32_e32 v0, s8
+; GFX7LESS-DPP-NEXT: v_mov_b32_e32 v1, s9
; GFX7LESS-DPP-NEXT: s_mov_b32 s2, -1
; GFX7LESS-DPP-NEXT: .LBB11_2: ; %atomicrmw.start
; GFX7LESS-DPP-NEXT: ; =>This Inner Loop Header: Depth=1
-; GFX7LESS-DPP-NEXT: v_add_f64 v[0:1], v[2:3], v[4:5]
+; GFX7LESS-DPP-NEXT: v_mov_b32_e32 v9, v1
+; GFX7LESS-DPP-NEXT: v_mov_b32_e32 v8, v0
+; GFX7LESS-DPP-NEXT: v_add_f64 v[6:7], v[8:9], v[4:5]
; GFX7LESS-DPP-NEXT: s_waitcnt expcnt(0)
-; GFX7LESS-DPP-NEXT: v_mov_b32_e32 v9, v3
-; GFX7LESS-DPP-NEXT: v_mov_b32_e32 v8, v2
-; GFX7LESS-DPP-NEXT: v_mov_b32_e32 v7, v1
-; GFX7LESS-DPP-NEXT: v_mov_b32_e32 v6, v0
-; GFX7LESS-DPP-NEXT: buffer_atomic_cmpswap_x2 v[6:9], off, s[0:3], 0 glc
+; GFX7LESS-DPP-NEXT: v_mov_b32_e32 v0, v6
+; GFX7LESS-DPP-NEXT: v_mov_b32_e32 v1, v7
+; GFX7LESS-DPP-NEXT: v_mov_b32_e32 v2, v8
+; GFX7LESS-DPP-NEXT: v_mov_b32_e32 v3, v9
+; GFX7LESS-DPP-NEXT: buffer_atomic_cmpswap_x2 v[0:3], off, s[0:3], 0 glc
; GFX7LESS-DPP-NEXT: s_waitcnt vmcnt(0)
-; GFX7LESS-DPP-NEXT: v_cmp_eq_u64_e32 vcc, v[6:7], v[2:3]
+; GFX7LESS-DPP-NEXT: v_cmp_eq_u64_e32 vcc, v[0:1], v[8:9]
; GFX7LESS-DPP-NEXT: s_or_b64 s[4:5], vcc, s[4:5]
-; GFX7LESS-DPP-NEXT: v_mov_b32_e32 v2, v6
-; GFX7LESS-DPP-NEXT: v_mov_b32_e32 v3, v7
; GFX7LESS-DPP-NEXT: s_andn2_b64 exec, exec, s[4:5]
; GFX7LESS-DPP-NEXT: s_cbranch_execnz .LBB11_2
; GFX7LESS-DPP-NEXT: .LBB11_3:
@@ -8551,23 +8551,23 @@ define amdgpu_kernel void @global_atomic_fadd_double_uni_address_div_value_one_a
; GFX7LESS-NEXT: s_mov_b32 s3, 0xf000
; GFX7LESS-NEXT: s_mov_b32 s2, -1
; GFX7LESS-NEXT: s_waitcnt lgkmcnt(0)
-; GFX7LESS-NEXT: buffer_load_dwordx2 v[2:3], off, s[0:3], 0
+; GFX7LESS-NEXT: buffer_load_dwordx2 v[0:1], off, s[0:3], 0
; GFX7LESS-NEXT: s_mov_b64 s[4:5], 0
; GFX7LESS-NEXT: .LBB12_4: ; %atomicrmw.start
; GFX7LESS-NEXT: ; =>This Inner Loop Header: Depth=1
; GFX7LESS-NEXT: s_waitcnt vmcnt(0)
-; GFX7LESS-NEXT: v_add_f64 v[0:1], v[2:3], v[4:5]
+; GFX7LESS-NEXT: v_mov_b32_e32 v9, v1
+; GFX7LESS-NEXT: v_mov_b32_e32 v8, v0
+; GFX7LESS-NEXT: v_add_f64 v[6:7], v[8:9], v[4:5]
; GFX7LESS-NEXT: s_waitcnt expcnt(0)
-; GFX7LESS-NEXT: v_mov_b32_e32 v9, v3
-; GFX7LESS-NEXT: v_mov_b32_e32 v8, v2
-; GFX7LESS-NEXT: v_mov_b32_e32 v7, v1
-; GFX7LESS-NEXT: v_mov_b32_e32 v6, v0
-; GFX7LESS-NEXT: buffer_atomic_cmpswap_x2 v[6:9], off, s[0:3], 0 glc
+; GFX7LESS-NEXT: v_mov_b32_e32 v0, v6
+; GFX7LESS-NEXT: v_mov_b32_e32 v1, v7
+; GFX7LESS-NEXT: v_mov_b32_e32 v2, v8
+; GFX7LESS-NEXT: v_mov_b32_e32 v3, v9
+; GFX7LESS-NEXT: buffer_atomic_cmpswap_x2 v[0:3], off, s[0:3], 0 glc
; GFX7LESS-NEXT: s_waitcnt vmcnt(0)
-; GFX7LESS-NEXT: v_cmp_eq_u64_e32 vcc, v[6:7], v[2:3]
+; GFX7LESS-NEXT: v_cmp_eq_u64_e32 vcc, v[0:1], v[8:9]
; GFX7LESS-NEXT: s_or_b64 s[4:5], vcc, s[4:5]
-; GFX7LESS-NEXT: v_mov_b32_e32 v2, v6
-; GFX7LESS-NEXT: v_mov_b32_e32 v3, v7
; GFX7LESS-NEXT: s_andn2_b64 exec, exec, s[4:5]
; GFX7LESS-NEXT: s_cbranch_execnz .LBB12_4
; GFX7LESS-NEXT: .LBB12_5:
@@ -8941,23 +8941,23 @@ define amdgpu_kernel void @global_atomic_fadd_double_uni_address_div_value_one_a
; GFX7LESS-DPP-NEXT: s_mov_b64 s[2:3], s[50:51]
; GFX7LESS-DPP-NEXT: s_waitcnt lgkmcnt(0)
; GFX7LESS-DPP-NEXT: s_swappc_b64 s[30:31], s[16:17]
-; GFX7LESS-DPP-NEXT: buffer_load_dwordx2 v[4:5], off, s[36:39], 0
+; GFX7LESS-DPP-NEXT: buffer_load_dwordx2 v[2:3], off, s[36:39], 0
; GFX7LESS-DPP-NEXT: s_mov_b64 s[0:1], 0
; GFX7LESS-DPP-NEXT: .LBB12_1: ; %atomicrmw.start
; GFX7LESS-DPP-NEXT: ; =>This Inner Loop Header: Depth=1
; GFX7LESS-DPP-NEXT: s_waitcnt vmcnt(0)
-; GFX7LESS-DPP-NEXT: v_add_f64 v[2:3], v[4:5], v[0:1]
-; GFX7LESS-DPP-NEXT: s_waitcnt expcnt(0)
-; GFX7LESS-DPP-NEXT: v_mov_b32_e32 v9, v5
-; GFX7LESS-DPP-NEXT: v_mov_b32_e32 v8, v4
; GFX7LESS-DPP-NEXT: v_mov_b32_e32 v7, v3
; GFX7LESS-DPP-NEXT: v_mov_b32_e32 v6, v2
-; GFX7LESS-DPP-NEXT: buffer_atomic_cmpswap_x2 v[6:9], off, s[36:39], 0 glc
-; GFX7LESS-DPP-NEXT: s_waitcnt vmcnt(0)
-; GFX7LESS-DPP-NEXT: v_cmp_eq_u64_e32 vcc, v[6:7], v[4:5]
-; GFX7LESS-DPP-NEXT: s_or_b64 s[0:1], vcc, s[0:1]
+; GFX7LESS-DPP-NEXT: s_waitcnt expcnt(0)
+; GFX7LESS-DPP-NEXT: v_add_f64 v[4:5], v[6:7], v[0:1]
+; GFX7LESS-DPP-NEXT: v_mov_b32_e32 v2, v4
+; GFX7LESS-DPP-NEXT: v_mov_b32_e32 v3, v5
; GFX7LESS-DPP-NEXT: v_mov_b32_e32 v4, v6
; GFX7LESS-DPP-NEXT: v_mov_b32_e32 v5, v7
+; GFX7LESS-DPP-NEXT: buffer_atomic_cmpswap_x2 v[2:5], off, s[36:39], 0 glc
+; GFX7LESS-DPP-NEXT: s_waitcnt vmcnt(0)
+; GFX7LESS-DPP-NEXT: v_cmp_eq_u64_e32 vcc, v[2:3], v[6:7]
+; GFX7LESS-DPP-NEXT: s_or_b64 s[0:1], vcc, s[0:1]
; GFX7LESS-DPP-NEXT: s_andn2_b64 exec, exec, s[0:1]
; GFX7LESS-DPP-NEXT: s_cbranch_execnz .LBB12_1
; GFX7LESS-DPP-NEXT: ; %bb.2: ; %atomicrmw.end
@@ -9437,23 +9437,23 @@ define amdgpu_kernel void @global_atomic_fadd_double_uni_address_uni_value_agent
; GFX7LESS-NEXT: v_add_f64 v[0:1], s[6:7], v[0:1]
; GFX7LESS-NEXT: v_mul_f64 v[4:5], 4.0, v[0:1]
; GFX7LESS-NEXT: s_waitcnt lgkmcnt(0)
-; GFX7LESS-NEXT: v_mov_b32_e32 v2, s8
-; GFX7LESS-NEXT: v_mov_b32_e32 v3, s9
+; GFX7LESS-NEXT: v_mov_b32_e32 v0, s8
+; GFX7LESS-NEXT: v_mov_b32_e32 v1, s9
; GFX7LESS-NEXT: s_mov_b32 s2, -1
; GFX7LESS-NEXT: .LBB13_2: ; %atomicrmw.start
; GFX7LESS-NEXT: ; =>This Inner Loop Header: Depth=1
-; GFX7LESS-NEXT: v_add_f64 v[0:1], v[2:3], v[4:5]
+; GFX7LESS-NEXT: v_mov_b32_e32 v9, v1
+; GFX7LESS-NEXT: v_mov_b32_e32 v8, v0
+; GFX7LESS-NEXT: v_add_f64 v[6:7], v[8:9], v[4:5]
; GFX7LESS-NEXT: s_waitcnt expcnt(0)
-; GFX7LESS-NEXT: v_mov_b32_e32 v9, v3
-; GFX7LESS-NEXT: v_mov_b32_e32 v8, v2
-; GFX7LESS-NEXT: v_mov_b32_e32 v7, v1
-; GFX7LESS-NEXT: v_mov_b32_e32 v6, v0
-; GFX7LESS-NEXT: buffer_atomic_cmpswap_x2 v[6:9], off, s[0:3], 0 glc
+; GFX7LESS-NEXT: v_mov_b32_e32 v0, v6
+; GFX7LESS-NEXT: v_mov_b32_e32 v1, v7
+; GFX7LESS-NEXT: v_mov_b32_e32 v2, v8
+; GFX7LESS-NEXT: v_mov_b32_e32 v3, v9
+; GFX7LESS-NEXT: buffer_atomic_cmpswap_x2 v[0:3], off, s[0:3], 0 glc
; GFX7LESS-NEXT: s_waitcnt vmcnt(0)
-; GFX7LESS-NEXT: v_cmp_eq_u64_e32 vcc, v[6:7], v[2:3]
+; GFX7LESS-NEXT: v_cmp_eq_u64_e32 vcc, v[0:1], v[8:9]
; GFX7LESS-NEXT: s_or_b64 s[4:5], vcc, s[4:5]
-; GFX7LESS-NEXT: v_mov_b32_e32 v2, v6
-; GFX7LESS-NEXT: v_mov_b32_e32 v3, v7
; GFX7LESS-NEXT: s_andn2_b64 exec, exec, s[4:5]
; GFX7LESS-NEXT: s_cbranch_execnz .LBB13_2
; GFX7LESS-NEXT: .LBB13_3:
@@ -9694,23 +9694,23 @@ define amdgpu_kernel void @global_atomic_fadd_double_uni_address_uni_value_agent
; GFX7LESS-DPP-NEXT: v_add_f64 v[0:1], s[6:7], v[0:1]
; GFX7LESS-DPP-NEXT: v_mul_f64 v[4:5], 4.0, v[0:1]
; GFX7LESS-DPP-NEXT: s_waitcnt lgkmcnt(0)
-; GFX7LESS-DPP-NEXT: v_mov_b32_e32 v2, s8
-; GFX7LESS-DPP-NEXT: v_mov_b32_e32 v3, s9
+; GFX7LESS-DPP-NEXT: v_mov_b32_e32 v0, s8
+; GFX7LESS-DPP-NEXT: v_mov_b32_e32 v1, s9
; GFX7LESS-DPP-NEXT: s_mov_b32 s2, -1
; GFX7LESS-DPP-NEXT: .LBB13_2: ; %atomicrmw.start
; GFX7LESS-DPP-NEXT: ; =>This Inner Loop Header: Depth=1
-; GFX7LESS-DPP-NEXT: v_add_f64 v[0:1], v[2:3], v[4:5]
+; GFX7LESS-DPP-NEXT: v_mov_b32_e32 v9, v1
+; GFX7LESS-DPP-NEXT: v_mov_b32_e32 v8, v0
+; GFX7LESS-DPP-NEXT: v_add_f64 v[6:7], v[8:9], v[4:5]
; GFX7LESS-DPP-NEXT: s_waitcnt expcnt(0)
-; GFX7LESS-DPP-NEXT: v_mov_b32_e32 v9, v3
-; GFX7LESS-DPP-NEXT: v_mov_b32_e32 v8, v2
-; GFX7LESS-DPP-NEXT: v_mov_b32_e32 v7, v1
-; GFX7LESS-DPP-NEXT: v_mov_b32_e32 v6, v0
-; GFX7LESS-DPP-NEXT: buffer_atomic_cmpswap_x2 v[6:9], off, s[0:3], 0 glc
+; GFX7LESS-DPP-NEXT: v_mov_b32_e32 v0, v6
+; GFX7LESS-DPP-NEXT: v_mov_b32_e32 v1, v7
+; GFX7LESS-DPP-NEXT: v_mov_b32_e32 v2, v8
+; GFX7LESS-DPP-NEXT: v_mov_b32_e32 v3, v9
+; GFX7LESS-DPP-NEXT: buffer_atomic_cmpswap_x2 v[0:3], off, s[0:3], 0 glc
; GFX7LESS-DPP-NEXT: s_waitcnt vmcnt(0)
-; GFX7LESS-DPP-NEXT: v_cmp_eq_u64_e32 vcc, v[6:7], v[2:3]
+; GFX7LESS-DPP-NEXT: v_cmp_eq_u64_e32 vcc, v[0:1], v[8:9]
; GFX7LESS-DPP-NEXT: s_or_b64 s[4:5], vcc, s[4:5]
-; GFX7LESS-DPP-NEXT: v_mov_b32_e32 v2, v6
-; GFX7LESS-DPP-NEXT: v_mov_b32_e32 v3, v7
; GFX7LESS-DPP-NEXT: s_andn2_b64 exec, exec, s[4:5]
; GFX7LESS-DPP-NEXT: s_cbranch_execnz .LBB13_2
; GFX7LESS-DPP-NEXT: .LBB13_3:
@@ -9984,23 +9984,23 @@ define amdgpu_kernel void @global_atomic_fadd_double_uni_address_div_value_agent
; GFX7LESS-NEXT: s_mov_b32 s3, 0xf000
; GFX7LESS-NEXT: s_mov_b32 s2, -1
; GFX7LESS-NEXT: s_waitcnt lgkmcnt(0)
-; GFX7LESS-NEXT: buffer_load_dwordx2 v[2:3], off, s[0:3], 0
+; GFX7LESS-NEXT: buffer_load_dwordx2 v[0:1], off, s[0:3], 0
; GFX7LESS-NEXT: s_mov_b64 s[4:5], 0
; GFX7LESS-NEXT: .LBB14_4: ; %atomicrmw.start
; GFX7LESS-NEXT: ; =>This Inner Loop Header: Depth=1
; GFX7LESS-NEXT: s_waitcnt vmcnt(0)
-; GFX7LESS-NEXT: v_add_f64 v[0:1], v[2:3], v[4:5]
+; GFX7LESS-NEXT: v_mov_b32_e32 v9, v1
+; GFX7LESS-NEXT: v_mov_b32_e32 v8, v0
+; GFX7LESS-NEXT: v_add_f64 v[6:7], v[8:9], v[4:5]
; GFX7LESS-NEXT: s_waitcnt expcnt(0)
-; GFX7LESS-NEXT: v_mov_b32_e32 v9, v3
-; GFX7LESS-NEXT: v_mov_b32_e32 v8, v2
-; GFX7LESS-NEXT: v_mov_b32_e32 v7, v1
-; GFX7LESS-NEXT: v_mov_b32_e32 v6, v0
-; GFX7LESS-NEXT: buffer_atomic_cmpswap_x2 v[6:9], off, s[0:3], 0 glc
+; GFX7LESS-NEXT: v_mov_b32_e32 v0, v6
+; GFX7LESS-NEXT: v_mov_b32_e32 v1, v7
+; GFX7LESS-NEXT: v_mov_b32_e32 v2, v8
+; GFX7LESS-NEXT: v_mov_b32_e32 v3, v9
+; GFX7LESS-NEXT: buffer_atomic_cmpswap_x2 v[0:3], off, s[0:3], 0 glc
; GFX7LESS-NEXT: s_waitcnt vmcnt(0)
-; GFX7LESS-NEXT: v_cmp_eq_u64_e32 vcc, v[6:7], v[2:3]
+; GFX7LESS-NEXT: v_cmp_eq_u64_e32 vcc, v[0:1], v[8:9]
; GFX7LESS-NEXT: s_or_b64 s[4:5], vcc, s[4:5]
-; GFX7LESS-NEXT: v_mov_b32_e32 v2, v6
-; GFX7LESS-NEXT: v_mov_b32_e32 v3, v7
; GFX7LESS-NEXT: s_andn2_b64 exec, exec, s[4:5]
; GFX7LESS-NEXT: s_cbranch_execnz .LBB14_4
; GFX7LESS-NEXT: .LBB14_5:
@@ -10374,23 +10374,23 @@ define amdgpu_kernel void @global_atomic_fadd_double_uni_address_div_value_agent
; GFX7LESS-DPP-NEXT: s_mov_b64 s[2:3], s[50:51]
; GFX7LESS-DPP-NEXT: s_waitcnt lgkmcnt(0)
; GFX7LESS-DPP-NEXT: s_swappc_b64 s[30:31], s[16:17]
-; GFX7LESS-DPP-NEXT: buffer_load_dwordx2 v[4:5], off, s[36:39], 0
+; GFX7LESS-DPP-NEXT: buffer_load_dwordx2 v[2:3], off, s[36:39], 0
; GFX7LESS-DPP-NEXT: s_mov_b64 s[0:1], 0
; GFX7LESS-DPP-NEXT: .LBB14_1: ; %atomicrmw.start
; GFX7LESS-DPP-NEXT: ; =>This Inner Loop Header: Depth=1
; GFX7LESS-DPP-NEXT: s_waitcnt vmcnt(0)
-; GFX7LESS-DPP-NEXT: v_add_f64 v[2:3], v[4:5], v[0:1]
-; GFX7LESS-DPP-NEXT: s_waitcnt expcnt(0)
-; GFX7LESS-DPP-NEXT: v_mov_b32_e32 v9, v5
-; GFX7LESS-DPP-NEXT: v_mov_b32_e32 v8, v4
; GFX7LESS-DPP-NEXT: v_mov_b32_e32 v7, v3
; GFX7LESS-DPP-NEXT: v_mov_b32_e32 v6, v2
-; GFX7LESS-DPP-NEXT: buffer_atomic_cmpswap_x2 v[6:9], off, s[36:39], 0 glc
-; GFX7LESS-DPP-NEXT: s_waitcnt vmcnt(0)
-; GFX7LESS-DPP-NEXT: v_cmp_eq_u64_e32 vcc, v[6:7], v[4:5]
-; GFX7LESS-DPP-NEXT: s_or_b64 s[0:1], vcc, s[0:1]
+; GFX7LESS-DPP-NEXT: s_waitcnt expcnt(0)
+; GFX7LESS-DPP-NEXT: v_add_f64 v[4:5], v[6:7], v[0:1]
+; GFX7LESS-DPP-NEXT: v_mov_b32_e32 v2, v4
+; GFX7LESS-DPP-NEXT: v_mov_b32_e32 v3, v5
; GFX7LESS-DPP-NEXT: v_mov_b32_e32 v4, v6
; GFX7LESS-DPP-NEXT: v_mov_b32_e32 v5, v7
+; GFX7LESS-DPP-NEXT: buffer_atomic_cmpswap_x2 v[2:5], off, s[36:39], 0 glc
+; GFX7LESS-DPP-NEXT: s_waitcnt vmcnt(0)
+; GFX7LESS-DPP-NEXT: v_cmp_eq_u64_e32 vcc, v[2:3], v[6:7]
+; GFX7LESS-DPP-NEXT: s_or_b64 s[0:1], vcc, s[0:1]
; GFX7LESS-DPP-NEXT: s_andn2_b64 exec, exec, s[0:1]
; GFX7LESS-DPP-NEXT: s_cbranch_execnz .LBB14_1
; GFX7LESS-DPP-NEXT: ; %bb.2: ; %atomicrmw.end
@@ -10899,23 +10899,23 @@ define amdgpu_kernel void @global_atomic_fadd_double_uni_address_div_value_agent
; GFX7LESS-NEXT: s_mov_b32 s3, 0xf000
; GFX7LESS-NEXT: s_mov_b32 s2, -1
; GFX7LESS-NEXT: s_waitcnt lgkmcnt(0)
-; GFX7LESS-NEXT: buffer_load_dwordx2 v[2:3], off, s[0:3], 0
+; GFX7LESS-NEXT: buffer_load_dwordx2 v[0:1], off, s[0:3], 0
; GFX7LESS-NEXT: s_mov_b64 s[4:5], 0
; GFX7LESS-NEXT: .LBB15_4: ; %atomicrmw.start
; GFX7LESS-NEXT: ; =>This Inner Loop Header: Depth=1
; GFX7LESS-NEXT: s_waitcnt vmcnt(0)
-; GFX7LESS-NEXT: v_add_f64 v[0:1], v[2:3], v[4:5]
+; GFX7LESS-NEXT: v_mov_b32_e32 v9, v1
+; GFX7LESS-NEXT: v_mov_b32_e32 v8, v0
+; GFX7LESS-NEXT: v_add_f64 v[6:7], v[8:9], v[4:5]
; GFX7LESS-NEXT: s_waitcnt expcnt(0)
-; GFX7LESS-NEXT: v_mov_b32_e32 v9, v3
-; GFX7LESS-NEXT: v_mov_b32_e32 v8, v2
-; GFX7LESS-NEXT: v_mov_b32_e32 v7, v1
-; GFX7LESS-NEXT: v_mov_b32_e32 v6, v0
-; GFX7LESS-NEXT: buffer_atomic_cmpswap_x2 v[6:9], off, s[0:3], 0 glc
+; GFX7LESS-NEXT: v_mov_b32_e32 v0, v6
+; GFX7LESS-NEXT: v_mov_b32_e32 v1, v7
+; GFX7LESS-NEXT: v_mov_b32_e32 v2, v8
+; GFX7LESS-NEXT: v_mov_b32_e32 v3, v9
+; GFX7LESS-NEXT: buffer_atomic_cmpswap_x2 v[0:3], off, s[0:3], 0 glc
; GFX7LESS-NEXT: s_waitcnt vmcnt(0)
-; GFX7LESS-NEXT: v_cmp_eq_u64_e32 vcc, v[6:7], v[2:3]
+; GFX7LESS-NEXT: v_cmp_eq_u64_e32 vcc, v[0:1], v[8:9]
; GFX7LESS-NEXT: s_or_b64 s[4:5], vcc, s[4:5]
-; GFX7LESS-NEXT: v_mov_b32_e32 v2, v6
-; GFX7LESS-NEXT: v_mov_b32_e32 v3, v7
; GFX7LESS-NEXT: s_andn2_b64 exec, exec, s[4:5]
; GFX7LESS-NEXT: s_cbranch_execnz .LBB15_4
; GFX7LESS-NEXT: .LBB15_5:
@@ -11289,23 +11289,23 @@ define amdgpu_kernel void @global_atomic_fadd_double_uni_address_div_value_agent
; GFX7LESS-DPP-NEXT: s_mov_b64 s[2:3], s[50:51]
; GFX7LESS-DPP-NEXT: s_waitcnt lgkmcnt(0)
; GFX7LESS-DPP-NEXT: s_swappc_b64 s[30:31], s[16:17]
-; GFX7LESS-DPP-NEXT: buffer_load_dwordx2 v[4:5], off, s[36:39], 0
+; GFX7LESS-DPP-NEXT: buffer_load_dwordx2 v[2:3], off, s[36:39], 0
; GFX7LESS-DPP-NEXT: s_mov_b64 s[0:1], 0
; GFX7LESS-DPP-NEXT: .LBB15_1: ; %atomicrmw.start
; GFX7LESS-DPP-NEXT: ; =>This Inner Loop Header: Depth=1
; GFX7LESS-DPP-NEXT: s_waitcnt vmcnt(0)
-; GFX7LESS-DPP-NEXT: v_add_f64 v[2:3], v[4:5], v[0:1]
-; GFX7LESS-DPP-NEXT: s_waitcnt expcnt(0)
-; GFX7LESS-DPP-NEXT: v_mov_b32_e32 v9, v5
-; GFX7LESS-DPP-NEXT: v_mov_b32_e32 v8, v4
; GFX7LESS-DPP-NEXT: v_mov_b32_e32 v7, v3
; GFX7LESS-DPP-NEXT: v_mov_b32_e32 v6, v2
-; GFX7LESS-DPP-NEXT: buffer_atomic_cmpswap_x2 v[6:9], off, s[36:39], 0 glc
-; GFX7LESS-DPP-NEXT: s_waitcnt vmcnt(0)
-; GFX7LESS-DPP-NEXT: v_cmp_eq_u64_e32 vcc, v[6:7], v[4:5]
-; GFX7LESS-DPP-NEXT: s_or_b64 s[0:1], vcc, s[0:1]
+; GFX7LESS-DPP-NEXT: s_waitcnt expcnt(0)
+; GFX7LESS-DPP-NEXT: v_add_f64 v[4:5], v[6:7], v[0:1]
+; GFX7LESS-DPP-NEXT: v_mov_b32_e32 v2, v4
+; GFX7LESS-DPP-NEXT: v_mov_b32_e32 v3, v5
; GFX7LESS-DPP-NEXT: v_mov_b32_e32 v4, v6
; GFX7LESS-DPP-NEXT: v_mov_b32_e32 v5, v7
+; GFX7LESS-DPP-NEXT: buffer_atomic_cmpswap_x2 v[2:5], off, s[36:39], 0 glc
+; GFX7LESS-DPP-NEXT: s_waitcnt vmcnt(0)
+; GFX7LESS-DPP-NEXT: v_cmp_eq_u64_e32 vcc, v[2:3], v[6:7]
+; GFX7LESS-DPP-NEXT: s_or_b64 s[0:1], vcc, s[0:1]
; GFX7LESS-DPP-NEXT: s_andn2_b64 exec, exec, s[0:1]
; GFX7LESS-DPP-NEXT: s_cbranch_execnz .LBB15_1
; GFX7LESS-DPP-NEXT: ; %bb.2: ; %atomicrmw.end
@@ -14051,19 +14051,19 @@ define amdgpu_kernel void @global_atomic_fadd_uni_address_uni_value_system_scope
; GFX7LESS-NEXT: v_cvt_f32_ubyte0_e32 v0, s2
; GFX7LESS-NEXT: v_mul_f32_e32 v2, 4.0, v0
; GFX7LESS-NEXT: s_waitcnt lgkmcnt(0)
-; GFX7LESS-NEXT: v_mov_b32_e32 v1, s6
+; GFX7LESS-NEXT: v_mov_b32_e32 v0, s6
; GFX7LESS-NEXT: s_mov_b32 s2, -1
; GFX7LESS-NEXT: .LBB18_2: ; %atomicrmw.start
; GFX7LESS-NEXT: ; =>This Inner Loop Header: Depth=1
-; GFX7LESS-NEXT: v_add_f32_e32 v0, v1, v2
+; GFX7LESS-NEXT: v_mov_b32_e32 v4, v0
+; GFX7LESS-NEXT: v_add_f32_e32 v3, v4, v2
; GFX7LESS-NEXT: s_waitcnt expcnt(0)
-; GFX7LESS-NEXT: v_mov_b32_e32 v4, v1
-; GFX7LESS-NEXT: v_mov_b32_e32 v3, v0
-; GFX7LESS-NEXT: buffer_atomic_cmpswap v[3:4], off, s[0:3], 0 glc
+; GFX7LESS-NEXT: v_mov_b32_e32 v0, v3
+; GFX7LESS-NEXT: v_mov_b32_e32 v1, v4
+; GFX7LESS-NEXT: buffer_atomic_cmpswap v[0:1], off, s[0:3], 0 glc
; GFX7LESS-NEXT: s_waitcnt vmcnt(0)
-; GFX7LESS-NEXT: v_cmp_eq_u32_e32 vcc, v3, v1
+; GFX7LESS-NEXT: v_cmp_eq_u32_e32 vcc, v0, v4
; GFX7LESS-NEXT: s_or_b64 s[4:5], vcc, s[4:5]
-; GFX7LESS-NEXT: v_mov_b32_e32 v1, v3
; GFX7LESS-NEXT: s_andn2_b64 exec, exec, s[4:5]
; GFX7LESS-NEXT: s_cbranch_execnz .LBB18_2
; GFX7LESS-NEXT: .LBB18_3:
@@ -14222,19 +14222,19 @@ define amdgpu_kernel void @global_atomic_fadd_uni_address_uni_value_system_scope
; GFX7LESS-DPP-NEXT: v_cvt_f32_ubyte0_e32 v0, s2
; GFX7LESS-DPP-NEXT: v_mul_f32_e32 v2, 4.0, v0
; GFX7LESS-DPP-NEXT: s_waitcnt lgkmcnt(0)
-; GFX7LESS-DPP-NEXT: v_mov_b32_e32 v1, s6
+; GFX7LESS-DPP-NEXT: v_mov_b32_e32 v0, s6
; GFX7LESS-DPP-NEXT: s_mov_b32 s2, -1
; GFX7LESS-DPP-NEXT: .LBB18_2: ; %atomicrmw.start
; GFX7LESS-DPP-NEXT: ; =>This Inner Loop Header: Depth=1
-; GFX7LESS-DPP-NEXT: v_add_f32_e32 v0, v1, v2
+; GFX7LESS-DPP-NEXT: v_mov_b32_e32 v4, v0
+; GFX7LESS-DPP-NEXT: v_add_f32_e32 v3, v4, v2
; GFX7LESS-DPP-NEXT: s_waitcnt expcnt(0)
-; GFX7LESS-DPP-NEXT: v_mov_b32_e32 v4, v1
-; GFX7LESS-DPP-NEXT: v_mov_b32_e32 v3, v0
-; GFX7LESS-DPP-NEXT: buffer_atomic_cmpswap v[3:4], off, s[0:3], 0 glc
+; GFX7LESS-DPP-NEXT: v_mov_b32_e32 v0, v3
+; GFX7LESS-DPP-NEXT: v_mov_b32_e32 v1, v4
+; GFX7LESS-DPP-NEXT: buffer_atomic_cmpswap v[0:1], off, s[0:3], 0 glc
; GFX7LESS-DPP-NEXT: s_waitcnt vmcnt(0)
-; GFX7LESS-DPP-NEXT: v_cmp_eq_u32_e32 vcc, v3, v1
+; GFX7LESS-DPP-NEXT: v_cmp_eq_u32_e32 vcc, v0, v4
; GFX7LESS-DPP-NEXT: s_or_b64 s[4:5], vcc, s[4:5]
-; GFX7LESS-DPP-NEXT: v_mov_b32_e32 v1, v3
; GFX7LESS-DPP-NEXT: s_andn2_b64 exec, exec, s[4:5]
; GFX7LESS-DPP-NEXT: s_cbranch_execnz .LBB18_2
; GFX7LESS-DPP-NEXT: .LBB18_3:
@@ -14397,19 +14397,19 @@ define amdgpu_kernel void @global_atomic_fadd_uni_address_uni_value_system_scope
; GFX7LESS-NEXT: v_cvt_f32_ubyte0_e32 v0, s2
; GFX7LESS-NEXT: v_mul_f32_e32 v2, 4.0, v0
; GFX7LESS-NEXT: s_waitcnt lgkmcnt(0)
-; GFX7LESS-NEXT: v_mov_b32_e32 v1, s6
+; GFX7LESS-NEXT: v_mov_b32_e32 v0, s6
; GFX7LESS-NEXT: s_mov_b32 s2, -1
; GFX7LESS-NEXT: .LBB19_2: ; %atomicrmw.start
; GFX7LESS-NEXT: ; =>This Inner Loop Header: Depth=1
-; GFX7LESS-NEXT: v_add_f32_e32 v0, v1, v2
+; GFX7LESS-NEXT: v_mov_b32_e32 v4, v0
+; GFX7LESS-NEXT: v_add_f32_e32 v3, v4, v2
; GFX7LESS-NEXT: s_waitcnt expcnt(0)
-; GFX7LESS-NEXT: v_mov_b32_e32 v4, v1
-; GFX7LESS-NEXT: v_mov_b32_e32 v3, v0
-; GFX7LESS-NEXT: buffer_atomic_cmpswap v[3:4], off, s[0:3], 0 glc
+; GFX7LESS-NEXT: v_mov_b32_e32 v0, v3
+; GFX7LESS-NEXT: v_mov_b32_e32 v1, v4
+; GFX7LESS-NEXT: buffer_atomic_cmpswap v[0:1], off, s[0:3], 0 glc
; GFX7LESS-NEXT: s_waitcnt vmcnt(0)
-; GFX7LESS-NEXT: v_cmp_eq_u32_e32 vcc, v3, v1
+; GFX7LESS-NEXT: v_cmp_eq_u32_e32 vcc, v0, v4
; GFX7LESS-NEXT: s_or_b64 s[4:5], vcc, s[4:5]
-; GFX7LESS-NEXT: v_mov_b32_e32 v1, v3
; GFX7LESS-NEXT: s_andn2_b64 exec, exec, s[4:5]
; GFX7LESS-NEXT: s_cbranch_execnz .LBB19_2
; GFX7LESS-NEXT: .LBB19_3:
@@ -14568,19 +14568,19 @@ define amdgpu_kernel void @global_atomic_fadd_uni_address_uni_value_system_scope
; GFX7LESS-DPP-NEXT: v_cvt_f32_ubyte0_e32 v0, s2
; GFX7LESS-DPP-NEXT: v_mul_f32_e32 v2, 4.0, v0
; GFX7LESS-DPP-NEXT: s_waitcnt lgkmcnt(0)
-; GFX7LESS-DPP-NEXT: v_mov_b32_e32 v1, s6
+; GFX7LESS-DPP-NEXT: v_mov_b32_e32 v0, s6
; GFX7LESS-DPP-NEXT: s_mov_b32 s2, -1
; GFX7LESS-DPP-NEXT: .LBB19_2: ; %atomicrmw.start
; GFX7LESS-DPP-NEXT: ; =>This Inner Loop Header: Depth=1
-; GFX7LESS-DPP-NEXT: v_add_f32_e32 v0, v1, v2
+; GFX7LESS-DPP-NEXT: v_mov_b32_e32 v4, v0
+; GFX7LESS-DPP-NEXT: v_add_f32_e32 v3, v4, v2
; GFX7LESS-DPP-NEXT: s_waitcnt expcnt(0)
-; GFX7LESS-DPP-NEXT: v_mov_b32_e32 v4, v1
-; GFX7LESS-DPP-NEXT: v_mov_b32_e32 v3, v0
-; GFX7LESS-DPP-NEXT: buffer_atomic_cmpswap v[3:4], off, s[0:3], 0 glc
+; GFX7LESS-DPP-NEXT: v_mov_b32_e32 v0, v3
+; GFX7LESS-DPP-NEXT: v_mov_b32_e32 v1, v4
+; GFX7LESS-DPP-NEXT: buffer_atomic_cmpswap v[0:1], off, s[0:3], 0 glc
; GFX7LESS-DPP-NEXT: s_waitcnt vmcnt(0)
-; GFX7LESS-DPP-NEXT: v_cmp_eq_u32_e32 vcc, v3, v1
+; GFX7LESS-DPP-NEXT: v_cmp_eq_u32_e32 vcc, v0, v4
; GFX7LESS-DPP-NEXT: s_or_b64 s[4:5], vcc, s[4:5]
-; GFX7LESS-DPP-NEXT: v_mov_b32_e32 v1, v3
; GFX7LESS-DPP-NEXT: s_andn2_b64 exec, exec, s[4:5]
; GFX7LESS-DPP-NEXT: s_cbranch_execnz .LBB19_2
; GFX7LESS-DPP-NEXT: .LBB19_3:
diff --git a/llvm/test/CodeGen/AMDGPU/global_atomics_scan_fmax.ll b/llvm/test/CodeGen/AMDGPU/global_atomics_scan_fmax.ll
index f8f911b693e09..cc2f490cc1ff0 100644
--- a/llvm/test/CodeGen/AMDGPU/global_atomics_scan_fmax.ll
+++ b/llvm/test/CodeGen/AMDGPU/global_atomics_scan_fmax.ll
@@ -30,20 +30,20 @@ define amdgpu_kernel void @global_atomic_fmax_uni_address_uni_value_agent_scope_
; GFX7LESS-NEXT: s_mov_b64 s[4:5], 0
; GFX7LESS-NEXT: s_mov_b32 s3, 0xf000
; GFX7LESS-NEXT: s_waitcnt lgkmcnt(0)
-; GFX7LESS-NEXT: v_mov_b32_e32 v1, s2
+; GFX7LESS-NEXT: v_mov_b32_e32 v0, s2
; GFX7LESS-NEXT: s_mov_b32 s2, -1
; GFX7LESS-NEXT: .LBB0_2: ; %atomicrmw.start
; GFX7LESS-NEXT: ; =>This Inner Loop Header: Depth=1
-; GFX7LESS-NEXT: v_mul_f32_e32 v0, 1.0, v1
-; GFX7LESS-NEXT: v_max_f32_e32 v0, 4.0, v0
-; GFX7LESS-NEXT: s_waitcnt expcnt(0)
-; GFX7LESS-NEXT: v_mov_b32_e32 v3, v1
; GFX7LESS-NEXT: v_mov_b32_e32 v2, v0
-; GFX7LESS-NEXT: buffer_atomic_cmpswap v[2:3], off, s[0:3], 0 glc
+; GFX7LESS-NEXT: s_waitcnt expcnt(0)
+; GFX7LESS-NEXT: v_mul_f32_e32 v0, 1.0, v2
+; GFX7LESS-NEXT: v_max_f32_e32 v1, 4.0, v0
+; GFX7LESS-NEXT: v_mov_b32_e32 v0, v1
+; GFX7LESS-NEXT: v_mov_b32_e32 v1, v2
+; GFX7LESS-NEXT: buffer_atomic_cmpswap v[0:1], off, s[0:3], 0 glc
; GFX7LESS-NEXT: s_waitcnt vmcnt(0)
-; GFX7LESS-NEXT: v_cmp_eq_u32_e32 vcc, v2, v1
+; GFX7LESS-NEXT: v_cmp_eq_u32_e32 vcc, v0, v2
; GFX7LESS-NEXT: s_or_b64 s[4:5], vcc, s[4:5]
-; GFX7LESS-NEXT: v_mov_b32_e32 v1, v2
; GFX7LESS-NEXT: s_andn2_b64 exec, exec, s[4:5]
; GFX7LESS-NEXT: s_cbranch_execnz .LBB0_2
; GFX7LESS-NEXT: .LBB0_3:
@@ -155,20 +155,20 @@ define amdgpu_kernel void @global_atomic_fmax_uni_address_uni_value_agent_scope_
; GFX7LESS-DPP-NEXT: s_mov_b64 s[4:5], 0
; GFX7LESS-DPP-NEXT: s_mov_b32 s3, 0xf000
; GFX7LESS-DPP-NEXT: s_waitcnt lgkmcnt(0)
-; GFX7LESS-DPP-NEXT: v_mov_b32_e32 v1, s2
+; GFX7LESS-DPP-NEXT: v_mov_b32_e32 v0, s2
; GFX7LESS-DPP-NEXT: s_mov_b32 s2, -1
; GFX7LESS-DPP-NEXT: .LBB0_2: ; %atomicrmw.start
; GFX7LESS-DPP-NEXT: ; =>This Inner Loop Header: Depth=1
-; GFX7LESS-DPP-NEXT: v_mul_f32_e32 v0, 1.0, v1
-; GFX7LESS-DPP-NEXT: v_max_f32_e32 v0, 4.0, v0
-; GFX7LESS-DPP-NEXT: s_waitcnt expcnt(0)
-; GFX7LESS-DPP-NEXT: v_mov_b32_e32 v3, v1
; GFX7LESS-DPP-NEXT: v_mov_b32_e32 v2, v0
-; GFX7LESS-DPP-NEXT: buffer_atomic_cmpswap v[2:3], off, s[0:3], 0 glc
+; GFX7LESS-DPP-NEXT: s_waitcnt expcnt(0)
+; GFX7LESS-DPP-NEXT: v_mul_f32_e32 v0, 1.0, v2
+; GFX7LESS-DPP-NEXT: v_max_f32_e32 v1, 4.0, v0
+; GFX7LESS-DPP-NEXT: v_mov_b32_e32 v0, v1
+; GFX7LESS-DPP-NEXT: v_mov_b32_e32 v1, v2
+; GFX7LESS-DPP-NEXT: buffer_atomic_cmpswap v[0:1], off, s[0:3], 0 glc
; GFX7LESS-DPP-NEXT: s_waitcnt vmcnt(0)
-; GFX7LESS-DPP-NEXT: v_cmp_eq_u32_e32 vcc, v2, v1
+; GFX7LESS-DPP-NEXT: v_cmp_eq_u32_e32 vcc, v0, v2
; GFX7LESS-DPP-NEXT: s_or_b64 s[4:5], vcc, s[4:5]
-; GFX7LESS-DPP-NEXT: v_mov_b32_e32 v1, v2
; GFX7LESS-DPP-NEXT: s_andn2_b64 exec, exec, s[4:5]
; GFX7LESS-DPP-NEXT: s_cbranch_execnz .LBB0_2
; GFX7LESS-DPP-NEXT: .LBB0_3:
@@ -301,18 +301,18 @@ define amdgpu_kernel void @global_atomic_fmax_uni_address_div_value_agent_scope_
; GFX7LESS-NEXT: s_waitcnt lgkmcnt(0)
; GFX7LESS-NEXT: s_swappc_b64 s[30:31], s[16:17]
; GFX7LESS-NEXT: s_mov_b64 s[0:1], exec
-; GFX7LESS-NEXT: v_mov_b32_e32 v2, 0x7fc00000
+; GFX7LESS-NEXT: v_mov_b32_e32 v1, 0x7fc00000
; GFX7LESS-NEXT: .LBB1_1: ; %ComputeLoop
; GFX7LESS-NEXT: ; =>This Inner Loop Header: Depth=1
; GFX7LESS-NEXT: s_ff1_i32_b64 s2, s[0:1]
-; GFX7LESS-NEXT: v_mul_f32_e32 v1, 1.0, v2
+; GFX7LESS-NEXT: v_mul_f32_e32 v1, 1.0, v1
; GFX7LESS-NEXT: v_readlane_b32 s4, v0, s2
; GFX7LESS-NEXT: s_lshl_b64 s[2:3], 1, s2
; GFX7LESS-NEXT: v_mul_f32_e64 v2, 1.0, s4
; GFX7LESS-NEXT: s_andn2_b64 s[0:1], s[0:1], s[2:3]
; GFX7LESS-NEXT: v_cmp_ne_u64_e64 s[2:3], s[0:1], 0
; GFX7LESS-NEXT: s_and_b64 vcc, exec, s[2:3]
-; GFX7LESS-NEXT: v_max_f32_e32 v2, v1, v2
+; GFX7LESS-NEXT: v_max_f32_e32 v1, v1, v2
; GFX7LESS-NEXT: s_cbranch_vccnz .LBB1_1
; GFX7LESS-NEXT: ; %bb.2: ; %ComputeEnd
; GFX7LESS-NEXT: v_mbcnt_lo_u32_b32_e64 v0, exec_lo, 0
@@ -326,22 +326,22 @@ define amdgpu_kernel void @global_atomic_fmax_uni_address_div_value_agent_scope_
; GFX7LESS-NEXT: s_mov_b32 s3, 0xf000
; GFX7LESS-NEXT: s_mov_b32 s2, -1
; GFX7LESS-NEXT: s_waitcnt lgkmcnt(0)
-; GFX7LESS-NEXT: buffer_load_dword v1, off, s[0:3], 0
+; GFX7LESS-NEXT: buffer_load_dword v0, off, s[0:3], 0
; GFX7LESS-NEXT: s_mov_b64 s[4:5], 0
-; GFX7LESS-NEXT: v_mul_f32_e32 v2, 1.0, v2
+; GFX7LESS-NEXT: v_mul_f32_e32 v2, 1.0, v1
; GFX7LESS-NEXT: .LBB1_4: ; %atomicrmw.start
; GFX7LESS-NEXT: ; =>This Inner Loop Header: Depth=1
; GFX7LESS-NEXT: s_waitcnt vmcnt(0)
-; GFX7LESS-NEXT: v_mul_f32_e32 v0, 1.0, v1
-; GFX7LESS-NEXT: v_max_f32_e32 v0, v0, v2
+; GFX7LESS-NEXT: v_mov_b32_e32 v4, v0
; GFX7LESS-NEXT: s_waitcnt expcnt(0)
-; GFX7LESS-NEXT: v_mov_b32_e32 v4, v1
-; GFX7LESS-NEXT: v_mov_b32_e32 v3, v0
-; GFX7LESS-NEXT: buffer_atomic_cmpswap v[3:4], off, s[0:3], 0 glc
+; GFX7LESS-NEXT: v_mul_f32_e32 v0, 1.0, v4
+; GFX7LESS-NEXT: v_max_f32_e32 v3, v0, v2
+; GFX7LESS-NEXT: v_mov_b32_e32 v0, v3
+; GFX7LESS-NEXT: v_mov_b32_e32 v1, v4
+; GFX7LESS-NEXT: buffer_atomic_cmpswap v[0:1], off, s[0:3], 0 glc
; GFX7LESS-NEXT: s_waitcnt vmcnt(0)
-; GFX7LESS-NEXT: v_cmp_eq_u32_e32 vcc, v3, v1
+; GFX7LESS-NEXT: v_cmp_eq_u32_e32 vcc, v0, v4
; GFX7LESS-NEXT: s_or_b64 s[4:5], vcc, s[4:5]
-; GFX7LESS-NEXT: v_mov_b32_e32 v1, v3
; GFX7LESS-NEXT: s_andn2_b64 exec, exec, s[4:5]
; GFX7LESS-NEXT: s_cbranch_execnz .LBB1_4
; GFX7LESS-NEXT: .LBB1_5:
@@ -666,20 +666,20 @@ define amdgpu_kernel void @global_atomic_fmax_uni_address_div_value_agent_scope_
; GFX7LESS-DPP-NEXT: s_swappc_b64 s[30:31], s[16:17]
; GFX7LESS-DPP-NEXT: buffer_load_dword v1, off, s[36:39], 0
; GFX7LESS-DPP-NEXT: s_mov_b64 s[0:1], 0
-; GFX7LESS-DPP-NEXT: v_mul_f32_e32 v2, 1.0, v0
+; GFX7LESS-DPP-NEXT: v_mul_f32_e32 v0, 1.0, v0
; GFX7LESS-DPP-NEXT: .LBB1_1: ; %atomicrmw.start
; GFX7LESS-DPP-NEXT: ; =>This Inner Loop Header: Depth=1
; GFX7LESS-DPP-NEXT: s_waitcnt vmcnt(0)
-; GFX7LESS-DPP-NEXT: v_mul_f32_e32 v0, 1.0, v1
-; GFX7LESS-DPP-NEXT: v_max_f32_e32 v0, v0, v2
+; GFX7LESS-DPP-NEXT: v_mov_b32_e32 v3, v1
; GFX7LESS-DPP-NEXT: s_waitcnt expcnt(0)
-; GFX7LESS-DPP-NEXT: v_mov_b32_e32 v4, v1
-; GFX7LESS-DPP-NEXT: v_mov_b32_e32 v3, v0
-; GFX7LESS-DPP-NEXT: buffer_atomic_cmpswap v[3:4], off, s[36:39], 0 glc
+; GFX7LESS-DPP-NEXT: v_mul_f32_e32 v1, 1.0, v3
+; GFX7LESS-DPP-NEXT: v_max_f32_e32 v2, v1, v0
+; GFX7LESS-DPP-NEXT: v_mov_b32_e32 v1, v2
+; GFX7LESS-DPP-NEXT: v_mov_b32_e32 v2, v3
+; GFX7LESS-DPP-NEXT: buffer_atomic_cmpswap v[1:2], off, s[36:39], 0 glc
; GFX7LESS-DPP-NEXT: s_waitcnt vmcnt(0)
-; GFX7LESS-DPP-NEXT: v_cmp_eq_u32_e32 vcc, v3, v1
+; GFX7LESS-DPP-NEXT: v_cmp_eq_u32_e32 vcc, v1, v3
; GFX7LESS-DPP-NEXT: s_or_b64 s[0:1], vcc, s[0:1]
-; GFX7LESS-DPP-NEXT: v_mov_b32_e32 v1, v3
; GFX7LESS-DPP-NEXT: s_andn2_b64 exec, exec, s[0:1]
; GFX7LESS-DPP-NEXT: s_cbranch_execnz .LBB1_1
; GFX7LESS-DPP-NEXT: ; %bb.2: ; %atomicrmw.end
@@ -1068,20 +1068,20 @@ define amdgpu_kernel void @global_atomic_fmax_uni_address_uni_value_one_as_scope
; GFX7LESS-NEXT: s_mov_b64 s[4:5], 0
; GFX7LESS-NEXT: s_mov_b32 s3, 0xf000
; GFX7LESS-NEXT: s_waitcnt lgkmcnt(0)
-; GFX7LESS-NEXT: v_mov_b32_e32 v1, s2
+; GFX7LESS-NEXT: v_mov_b32_e32 v0, s2
; GFX7LESS-NEXT: s_mov_b32 s2, -1
; GFX7LESS-NEXT: .LBB2_2: ; %atomicrmw.start
; GFX7LESS-NEXT: ; =>This Inner Loop Header: Depth=1
-; GFX7LESS-NEXT: v_mul_f32_e32 v0, 1.0, v1
-; GFX7LESS-NEXT: v_max_f32_e32 v0, 4.0, v0
-; GFX7LESS-NEXT: s_waitcnt expcnt(0)
-; GFX7LESS-NEXT: v_mov_b32_e32 v3, v1
; GFX7LESS-NEXT: v_mov_b32_e32 v2, v0
-; GFX7LESS-NEXT: buffer_atomic_cmpswap v[2:3], off, s[0:3], 0 glc
+; GFX7LESS-NEXT: s_waitcnt expcnt(0)
+; GFX7LESS-NEXT: v_mul_f32_e32 v0, 1.0, v2
+; GFX7LESS-NEXT: v_max_f32_e32 v1, 4.0, v0
+; GFX7LESS-NEXT: v_mov_b32_e32 v0, v1
+; GFX7LESS-NEXT: v_mov_b32_e32 v1, v2
+; GFX7LESS-NEXT: buffer_atomic_cmpswap v[0:1], off, s[0:3], 0 glc
; GFX7LESS-NEXT: s_waitcnt vmcnt(0)
-; GFX7LESS-NEXT: v_cmp_eq_u32_e32 vcc, v2, v1
+; GFX7LESS-NEXT: v_cmp_eq_u32_e32 vcc, v0, v2
; GFX7LESS-NEXT: s_or_b64 s[4:5], vcc, s[4:5]
-; GFX7LESS-NEXT: v_mov_b32_e32 v1, v2
; GFX7LESS-NEXT: s_andn2_b64 exec, exec, s[4:5]
; GFX7LESS-NEXT: s_cbranch_execnz .LBB2_2
; GFX7LESS-NEXT: .LBB2_3:
@@ -1193,20 +1193,20 @@ define amdgpu_kernel void @global_atomic_fmax_uni_address_uni_value_one_as_scope
; GFX7LESS-DPP-NEXT: s_mov_b64 s[4:5], 0
; GFX7LESS-DPP-NEXT: s_mov_b32 s3, 0xf000
; GFX7LESS-DPP-NEXT: s_waitcnt lgkmcnt(0)
-; GFX7LESS-DPP-NEXT: v_mov_b32_e32 v1, s2
+; GFX7LESS-DPP-NEXT: v_mov_b32_e32 v0, s2
; GFX7LESS-DPP-NEXT: s_mov_b32 s2, -1
; GFX7LESS-DPP-NEXT: .LBB2_2: ; %atomicrmw.start
; GFX7LESS-DPP-NEXT: ; =>This Inner Loop Header: Depth=1
-; GFX7LESS-DPP-NEXT: v_mul_f32_e32 v0, 1.0, v1
-; GFX7LESS-DPP-NEXT: v_max_f32_e32 v0, 4.0, v0
-; GFX7LESS-DPP-NEXT: s_waitcnt expcnt(0)
-; GFX7LESS-DPP-NEXT: v_mov_b32_e32 v3, v1
; GFX7LESS-DPP-NEXT: v_mov_b32_e32 v2, v0
-; GFX7LESS-DPP-NEXT: buffer_atomic_cmpswap v[2:3], off, s[0:3], 0 glc
+; GFX7LESS-DPP-NEXT: s_waitcnt expcnt(0)
+; GFX7LESS-DPP-NEXT: v_mul_f32_e32 v0, 1.0, v2
+; GFX7LESS-DPP-NEXT: v_max_f32_e32 v1, 4.0, v0
+; GFX7LESS-DPP-NEXT: v_mov_b32_e32 v0, v1
+; GFX7LESS-DPP-NEXT: v_mov_b32_e32 v1, v2
+; GFX7LESS-DPP-NEXT: buffer_atomic_cmpswap v[0:1], off, s[0:3], 0 glc
; GFX7LESS-DPP-NEXT: s_waitcnt vmcnt(0)
-; GFX7LESS-DPP-NEXT: v_cmp_eq_u32_e32 vcc, v2, v1
+; GFX7LESS-DPP-NEXT: v_cmp_eq_u32_e32 vcc, v0, v2
; GFX7LESS-DPP-NEXT: s_or_b64 s[4:5], vcc, s[4:5]
-; GFX7LESS-DPP-NEXT: v_mov_b32_e32 v1, v2
; GFX7LESS-DPP-NEXT: s_andn2_b64 exec, exec, s[4:5]
; GFX7LESS-DPP-NEXT: s_cbranch_execnz .LBB2_2
; GFX7LESS-DPP-NEXT: .LBB2_3:
@@ -1340,18 +1340,18 @@ define amdgpu_kernel void @global_atomic_fmax_uni_address_div_value_one_as_scope
; GFX7LESS-NEXT: s_waitcnt lgkmcnt(0)
; GFX7LESS-NEXT: s_swappc_b64 s[30:31], s[16:17]
; GFX7LESS-NEXT: s_mov_b64 s[0:1], exec
-; GFX7LESS-NEXT: v_mov_b32_e32 v2, 0x7fc00000
+; GFX7LESS-NEXT: v_mov_b32_e32 v1, 0x7fc00000
; GFX7LESS-NEXT: .LBB3_1: ; %ComputeLoop
; GFX7LESS-NEXT: ; =>This Inner Loop Header: Depth=1
; GFX7LESS-NEXT: s_ff1_i32_b64 s2, s[0:1]
-; GFX7LESS-NEXT: v_mul_f32_e32 v1, 1.0, v2
+; GFX7LESS-NEXT: v_mul_f32_e32 v1, 1.0, v1
; GFX7LESS-NEXT: v_readlane_b32 s4, v0, s2
; GFX7LESS-NEXT: s_lshl_b64 s[2:3], 1, s2
; GFX7LESS-NEXT: v_mul_f32_e64 v2, 1.0, s4
; GFX7LESS-NEXT: s_andn2_b64 s[0:1], s[0:1], s[2:3]
; GFX7LESS-NEXT: v_cmp_ne_u64_e64 s[2:3], s[0:1], 0
; GFX7LESS-NEXT: s_and_b64 vcc, exec, s[2:3]
-; GFX7LESS-NEXT: v_max_f32_e32 v2, v1, v2
+; GFX7LESS-NEXT: v_max_f32_e32 v1, v1, v2
; GFX7LESS-NEXT: s_cbranch_vccnz .LBB3_1
; GFX7LESS-NEXT: ; %bb.2: ; %ComputeEnd
; GFX7LESS-NEXT: v_mbcnt_lo_u32_b32_e64 v0, exec_lo, 0
@@ -1365,22 +1365,22 @@ define amdgpu_kernel void @global_atomic_fmax_uni_address_div_value_one_as_scope
; GFX7LESS-NEXT: s_mov_b32 s3, 0xf000
; GFX7LESS-NEXT: s_mov_b32 s2, -1
; GFX7LESS-NEXT: s_waitcnt lgkmcnt(0)
-; GFX7LESS-NEXT: buffer_load_dword v1, off, s[0:3], 0
+; GFX7LESS-NEXT: buffer_load_dword v0, off, s[0:3], 0
; GFX7LESS-NEXT: s_mov_b64 s[4:5], 0
-; GFX7LESS-NEXT: v_mul_f32_e32 v2, 1.0, v2
+; GFX7LESS-NEXT: v_mul_f32_e32 v2, 1.0, v1
; GFX7LESS-NEXT: .LBB3_4: ; %atomicrmw.start
; GFX7LESS-NEXT: ; =>This Inner Loop Header: Depth=1
; GFX7LESS-NEXT: s_waitcnt vmcnt(0)
-; GFX7LESS-NEXT: v_mul_f32_e32 v0, 1.0, v1
-; GFX7LESS-NEXT: v_max_f32_e32 v0, v0, v2
+; GFX7LESS-NEXT: v_mov_b32_e32 v4, v0
; GFX7LESS-NEXT: s_waitcnt expcnt(0)
-; GFX7LESS-NEXT: v_mov_b32_e32 v4, v1
-; GFX7LESS-NEXT: v_mov_b32_e32 v3, v0
-; GFX7LESS-NEXT: buffer_atomic_cmpswap v[3:4], off, s[0:3], 0 glc
+; GFX7LESS-NEXT: v_mul_f32_e32 v0, 1.0, v4
+; GFX7LESS-NEXT: v_max_f32_e32 v3, v0, v2
+; GFX7LESS-NEXT: v_mov_b32_e32 v0, v3
+; GFX7LESS-NEXT: v_mov_b32_e32 v1, v4
+; GFX7LESS-NEXT: buffer_atomic_cmpswap v[0:1], off, s[0:3], 0 glc
; GFX7LESS-NEXT: s_waitcnt vmcnt(0)
-; GFX7LESS-NEXT: v_cmp_eq_u32_e32 vcc, v3, v1
+; GFX7LESS-NEXT: v_cmp_eq_u32_e32 vcc, v0, v4
; GFX7LESS-NEXT: s_or_b64 s[4:5], vcc, s[4:5]
-; GFX7LESS-NEXT: v_mov_b32_e32 v1, v3
; GFX7LESS-NEXT: s_andn2_b64 exec, exec, s[4:5]
; GFX7LESS-NEXT: s_cbranch_execnz .LBB3_4
; GFX7LESS-NEXT: .LBB3_5:
@@ -1705,20 +1705,20 @@ define amdgpu_kernel void @global_atomic_fmax_uni_address_div_value_one_as_scope
; GFX7LESS-DPP-NEXT: s_swappc_b64 s[30:31], s[16:17]
; GFX7LESS-DPP-NEXT: buffer_load_dword v1, off, s[36:39], 0
; GFX7LESS-DPP-NEXT: s_mov_b64 s[0:1], 0
-; GFX7LESS-DPP-NEXT: v_mul_f32_e32 v2, 1.0, v0
+; GFX7LESS-DPP-NEXT: v_mul_f32_e32 v0, 1.0, v0
; GFX7LESS-DPP-NEXT: .LBB3_1: ; %atomicrmw.start
; GFX7LESS-DPP-NEXT: ; =>This Inner Loop Header: Depth=1
; GFX7LESS-DPP-NEXT: s_waitcnt vmcnt(0)
-; GFX7LESS-DPP-NEXT: v_mul_f32_e32 v0, 1.0, v1
-; GFX7LESS-DPP-NEXT: v_max_f32_e32 v0, v0, v2
+; GFX7LESS-DPP-NEXT: v_mov_b32_e32 v3, v1
; GFX7LESS-DPP-NEXT: s_waitcnt expcnt(0)
-; GFX7LESS-DPP-NEXT: v_mov_b32_e32 v4, v1
-; GFX7LESS-DPP-NEXT: v_mov_b32_e32 v3, v0
-; GFX7LESS-DPP-NEXT: buffer_atomic_cmpswap v[3:4], off, s[36:39], 0 glc
+; GFX7LESS-DPP-NEXT: v_mul_f32_e32 v1, 1.0, v3
+; GFX7LESS-DPP-NEXT: v_max_f32_e32 v2, v1, v0
+; GFX7LESS-DPP-NEXT: v_mov_b32_e32 v1, v2
+; GFX7LESS-DPP-NEXT: v_mov_b32_e32 v2, v3
+; GFX7LESS-DPP-NEXT: buffer_atomic_cmpswap v[1:2], off, s[36:39], 0 glc
; GFX7LESS-DPP-NEXT: s_waitcnt vmcnt(0)
-; GFX7LESS-DPP-NEXT: v_cmp_eq_u32_e32 vcc, v3, v1
+; GFX7LESS-DPP-NEXT: v_cmp_eq_u32_e32 vcc, v1, v3
; GFX7LESS-DPP-NEXT: s_or_b64 s[0:1], vcc, s[0:1]
-; GFX7LESS-DPP-NEXT: v_mov_b32_e32 v1, v3
; GFX7LESS-DPP-NEXT: s_andn2_b64 exec, exec, s[0:1]
; GFX7LESS-DPP-NEXT: s_cbranch_execnz .LBB3_1
; GFX7LESS-DPP-NEXT: ; %bb.2: ; %atomicrmw.end
@@ -2108,20 +2108,20 @@ define amdgpu_kernel void @global_atomic_fmax_uni_address_uni_value_default_scop
; GFX7LESS-NEXT: s_mov_b64 s[4:5], 0
; GFX7LESS-NEXT: s_mov_b32 s3, 0xf000
; GFX7LESS-NEXT: s_waitcnt lgkmcnt(0)
-; GFX7LESS-NEXT: v_mov_b32_e32 v1, s2
+; GFX7LESS-NEXT: v_mov_b32_e32 v0, s2
; GFX7LESS-NEXT: s_mov_b32 s2, -1
; GFX7LESS-NEXT: .LBB4_2: ; %atomicrmw.start
; GFX7LESS-NEXT: ; =>This Inner Loop Header: Depth=1
-; GFX7LESS-NEXT: v_mul_f32_e32 v0, 1.0, v1
-; GFX7LESS-NEXT: v_max_f32_e32 v0, 4.0, v0
-; GFX7LESS-NEXT: s_waitcnt expcnt(0)
-; GFX7LESS-NEXT: v_mov_b32_e32 v3, v1
; GFX7LESS-NEXT: v_mov_b32_e32 v2, v0
-; GFX7LESS-NEXT: buffer_atomic_cmpswap v[2:3], off, s[0:3], 0 glc
+; GFX7LESS-NEXT: s_waitcnt expcnt(0)
+; GFX7LESS-NEXT: v_mul_f32_e32 v0, 1.0, v2
+; GFX7LESS-NEXT: v_max_f32_e32 v1, 4.0, v0
+; GFX7LESS-NEXT: v_mov_b32_e32 v0, v1
+; GFX7LESS-NEXT: v_mov_b32_e32 v1, v2
+; GFX7LESS-NEXT: buffer_atomic_cmpswap v[0:1], off, s[0:3], 0 glc
; GFX7LESS-NEXT: s_waitcnt vmcnt(0)
-; GFX7LESS-NEXT: v_cmp_eq_u32_e32 vcc, v2, v1
+; GFX7LESS-NEXT: v_cmp_eq_u32_e32 vcc, v0, v2
; GFX7LESS-NEXT: s_or_b64 s[4:5], vcc, s[4:5]
-; GFX7LESS-NEXT: v_mov_b32_e32 v1, v2
; GFX7LESS-NEXT: s_andn2_b64 exec, exec, s[4:5]
; GFX7LESS-NEXT: s_cbranch_execnz .LBB4_2
; GFX7LESS-NEXT: .LBB4_3:
@@ -2233,20 +2233,20 @@ define amdgpu_kernel void @global_atomic_fmax_uni_address_uni_value_default_scop
; GFX7LESS-DPP-NEXT: s_mov_b64 s[4:5], 0
; GFX7LESS-DPP-NEXT: s_mov_b32 s3, 0xf000
; GFX7LESS-DPP-NEXT: s_waitcnt lgkmcnt(0)
-; GFX7LESS-DPP-NEXT: v_mov_b32_e32 v1, s2
+; GFX7LESS-DPP-NEXT: v_mov_b32_e32 v0, s2
; GFX7LESS-DPP-NEXT: s_mov_b32 s2, -1
; GFX7LESS-DPP-NEXT: .LBB4_2: ; %atomicrmw.start
; GFX7LESS-DPP-NEXT: ; =>This Inner Loop Header: Depth=1
-; GFX7LESS-DPP-NEXT: v_mul_f32_e32 v0, 1.0, v1
-; GFX7LESS-DPP-NEXT: v_max_f32_e32 v0, 4.0, v0
-; GFX7LESS-DPP-NEXT: s_waitcnt expcnt(0)
-; GFX7LESS-DPP-NEXT: v_mov_b32_e32 v3, v1
; GFX7LESS-DPP-NEXT: v_mov_b32_e32 v2, v0
-; GFX7LESS-DPP-NEXT: buffer_atomic_cmpswap v[2:3], off, s[0:3], 0 glc
+; GFX7LESS-DPP-NEXT: s_waitcnt expcnt(0)
+; GFX7LESS-DPP-NEXT: v_mul_f32_e32 v0, 1.0, v2
+; GFX7LESS-DPP-NEXT: v_max_f32_e32 v1, 4.0, v0
+; GFX7LESS-DPP-NEXT: v_mov_b32_e32 v0, v1
+; GFX7LESS-DPP-NEXT: v_mov_b32_e32 v1, v2
+; GFX7LESS-DPP-NEXT: buffer_atomic_cmpswap v[0:1], off, s[0:3], 0 glc
; GFX7LESS-DPP-NEXT: s_waitcnt vmcnt(0)
-; GFX7LESS-DPP-NEXT: v_cmp_eq_u32_e32 vcc, v2, v1
+; GFX7LESS-DPP-NEXT: v_cmp_eq_u32_e32 vcc, v0, v2
; GFX7LESS-DPP-NEXT: s_or_b64 s[4:5], vcc, s[4:5]
-; GFX7LESS-DPP-NEXT: v_mov_b32_e32 v1, v2
; GFX7LESS-DPP-NEXT: s_andn2_b64 exec, exec, s[4:5]
; GFX7LESS-DPP-NEXT: s_cbranch_execnz .LBB4_2
; GFX7LESS-DPP-NEXT: .LBB4_3:
@@ -2379,18 +2379,18 @@ define amdgpu_kernel void @global_atomic_fmax_uni_address_div_value_default_scop
; GFX7LESS-NEXT: s_waitcnt lgkmcnt(0)
; GFX7LESS-NEXT: s_swappc_b64 s[30:31], s[16:17]
; GFX7LESS-NEXT: s_mov_b64 s[0:1], exec
-; GFX7LESS-NEXT: v_mov_b32_e32 v2, 0x7fc00000
+; GFX7LESS-NEXT: v_mov_b32_e32 v1, 0x7fc00000
; GFX7LESS-NEXT: .LBB5_1: ; %ComputeLoop
; GFX7LESS-NEXT: ; =>This Inner Loop Header: Depth=1
; GFX7LESS-NEXT: s_ff1_i32_b64 s2, s[0:1]
-; GFX7LESS-NEXT: v_mul_f32_e32 v1, 1.0, v2
+; GFX7LESS-NEXT: v_mul_f32_e32 v1, 1.0, v1
; GFX7LESS-NEXT: v_readlane_b32 s4, v0, s2
; GFX7LESS-NEXT: s_lshl_b64 s[2:3], 1, s2
; GFX7LESS-NEXT: v_mul_f32_e64 v2, 1.0, s4
; GFX7LESS-NEXT: s_andn2_b64 s[0:1], s[0:1], s[2:3]
; GFX7LESS-NEXT: v_cmp_ne_u64_e64 s[2:3], s[0:1], 0
; GFX7LESS-NEXT: s_and_b64 vcc, exec, s[2:3]
-; GFX7LESS-NEXT: v_max_f32_e32 v2, v1, v2
+; GFX7LESS-NEXT: v_max_f32_e32 v1, v1, v2
; GFX7LESS-NEXT: s_cbranch_vccnz .LBB5_1
; GFX7LESS-NEXT: ; %bb.2: ; %ComputeEnd
; GFX7LESS-NEXT: v_mbcnt_lo_u32_b32_e64 v0, exec_lo, 0
@@ -2404,22 +2404,22 @@ define amdgpu_kernel void @global_atomic_fmax_uni_address_div_value_default_scop
; GFX7LESS-NEXT: s_mov_b32 s3, 0xf000
; GFX7LESS-NEXT: s_mov_b32 s2, -1
; GFX7LESS-NEXT: s_waitcnt lgkmcnt(0)
-; GFX7LESS-NEXT: buffer_load_dword v1, off, s[0:3], 0
+; GFX7LESS-NEXT: buffer_load_dword v0, off, s[0:3], 0
; GFX7LESS-NEXT: s_mov_b64 s[4:5], 0
-; GFX7LESS-NEXT: v_mul_f32_e32 v2, 1.0, v2
+; GFX7LESS-NEXT: v_mul_f32_e32 v2, 1.0, v1
; GFX7LESS-NEXT: .LBB5_4: ; %atomicrmw.start
; GFX7LESS-NEXT: ; =>This Inner Loop Header: Depth=1
; GFX7LESS-NEXT: s_waitcnt vmcnt(0)
-; GFX7LESS-NEXT: v_mul_f32_e32 v0, 1.0, v1
-; GFX7LESS-NEXT: v_max_f32_e32 v0, v0, v2
+; GFX7LESS-NEXT: v_mov_b32_e32 v4, v0
; GFX7LESS-NEXT: s_waitcnt expcnt(0)
-; GFX7LESS-NEXT: v_mov_b32_e32 v4, v1
-; GFX7LESS-NEXT: v_mov_b32_e32 v3, v0
-; GFX7LESS-NEXT: buffer_atomic_cmpswap v[3:4], off, s[0:3], 0 glc
+; GFX7LESS-NEXT: v_mul_f32_e32 v0, 1.0, v4
+; GFX7LESS-NEXT: v_max_f32_e32 v3, v0, v2
+; GFX7LESS-NEXT: v_mov_b32_e32 v0, v3
+; GFX7LESS-NEXT: v_mov_b32_e32 v1, v4
+; GFX7LESS-NEXT: buffer_atomic_cmpswap v[0:1], off, s[0:3], 0 glc
; GFX7LESS-NEXT: s_waitcnt vmcnt(0)
-; GFX7LESS-NEXT: v_cmp_eq_u32_e32 vcc, v3, v1
+; GFX7LESS-NEXT: v_cmp_eq_u32_e32 vcc, v0, v4
; GFX7LESS-NEXT: s_or_b64 s[4:5], vcc, s[4:5]
-; GFX7LESS-NEXT: v_mov_b32_e32 v1, v3
; GFX7LESS-NEXT: s_andn2_b64 exec, exec, s[4:5]
; GFX7LESS-NEXT: s_cbranch_execnz .LBB5_4
; GFX7LESS-NEXT: .LBB5_5:
@@ -2744,20 +2744,20 @@ define amdgpu_kernel void @global_atomic_fmax_uni_address_div_value_default_scop
; GFX7LESS-DPP-NEXT: s_swappc_b64 s[30:31], s[16:17]
; GFX7LESS-DPP-NEXT: buffer_load_dword v1, off, s[36:39], 0
; GFX7LESS-DPP-NEXT: s_mov_b64 s[0:1], 0
-; GFX7LESS-DPP-NEXT: v_mul_f32_e32 v2, 1.0, v0
+; GFX7LESS-DPP-NEXT: v_mul_f32_e32 v0, 1.0, v0
; GFX7LESS-DPP-NEXT: .LBB5_1: ; %atomicrmw.start
; GFX7LESS-DPP-NEXT: ; =>This Inner Loop Header: Depth=1
; GFX7LESS-DPP-NEXT: s_waitcnt vmcnt(0)
-; GFX7LESS-DPP-NEXT: v_mul_f32_e32 v0, 1.0, v1
-; GFX7LESS-DPP-NEXT: v_max_f32_e32 v0, v0, v2
+; GFX7LESS-DPP-NEXT: v_mov_b32_e32 v3, v1
; GFX7LESS-DPP-NEXT: s_waitcnt expcnt(0)
-; GFX7LESS-DPP-NEXT: v_mov_b32_e32 v4, v1
-; GFX7LESS-DPP-NEXT: v_mov_b32_e32 v3, v0
-; GFX7LESS-DPP-NEXT: buffer_atomic_cmpswap v[3:4], off, s[36:39], 0 glc
+; GFX7LESS-DPP-NEXT: v_mul_f32_e32 v1, 1.0, v3
+; GFX7LESS-DPP-NEXT: v_max_f32_e32 v2, v1, v0
+; GFX7LESS-DPP-NEXT: v_mov_b32_e32 v1, v2
+; GFX7LESS-DPP-NEXT: v_mov_b32_e32 v2, v3
+; GFX7LESS-DPP-NEXT: buffer_atomic_cmpswap v[1:2], off, s[36:39], 0 glc
; GFX7LESS-DPP-NEXT: s_waitcnt vmcnt(0)
-; GFX7LESS-DPP-NEXT: v_cmp_eq_u32_e32 vcc, v3, v1
+; GFX7LESS-DPP-NEXT: v_cmp_eq_u32_e32 vcc, v1, v3
; GFX7LESS-DPP-NEXT: s_or_b64 s[0:1], vcc, s[0:1]
-; GFX7LESS-DPP-NEXT: v_mov_b32_e32 v1, v3
; GFX7LESS-DPP-NEXT: s_andn2_b64 exec, exec, s[0:1]
; GFX7LESS-DPP-NEXT: s_cbranch_execnz .LBB5_1
; GFX7LESS-DPP-NEXT: ; %bb.2: ; %atomicrmw.end
@@ -4275,20 +4275,20 @@ define amdgpu_kernel void @global_atomic_fmax_double_uni_address_div_value_agent
; GFX1064-NEXT: s_movk_i32 s32, 0x800
; GFX1064-NEXT: s_waitcnt lgkmcnt(0)
; GFX1064-NEXT: s_swappc_b64 s[30:31], s[16:17]
-; GFX1064-NEXT: v_mov_b32_e32 v2, 0
-; GFX1064-NEXT: v_mov_b32_e32 v3, 0x7ff80000
+; GFX1064-NEXT: v_mov_b32_e32 v3, 0
+; GFX1064-NEXT: v_mov_b32_e32 v4, 0x7ff80000
; GFX1064-NEXT: s_mov_b64 s[0:1], exec
; GFX1064-NEXT: .LBB7_1: ; %ComputeLoop
; GFX1064-NEXT: ; =>This Inner Loop Header: Depth=1
; GFX1064-NEXT: s_ff1_i32_b64 s4, s[0:1]
-; GFX1064-NEXT: v_max_f64 v[2:3], v[2:3], v[2:3]
+; GFX1064-NEXT: v_max_f64 v[2:3], v[3:4], v[3:4]
; GFX1064-NEXT: v_readlane_b32 s3, v1, s4
; GFX1064-NEXT: v_readlane_b32 s2, v0, s4
; GFX1064-NEXT: v_max_f64 v[4:5], s[2:3], s[2:3]
; GFX1064-NEXT: s_lshl_b64 s[2:3], 1, s4
; GFX1064-NEXT: s_andn2_b64 s[0:1], s[0:1], s[2:3]
; GFX1064-NEXT: s_cmp_lg_u64 s[0:1], 0
-; GFX1064-NEXT: v_max_f64 v[2:3], v[2:3], v[4:5]
+; GFX1064-NEXT: v_max_f64 v[3:4], v[2:3], v[4:5]
; GFX1064-NEXT: s_cbranch_scc1 .LBB7_1
; GFX1064-NEXT: ; %bb.2: ; %ComputeEnd
; GFX1064-NEXT: v_mbcnt_lo_u32_b32 v0, exec_lo, 0
@@ -4300,26 +4300,26 @@ define amdgpu_kernel void @global_atomic_fmax_double_uni_address_div_value_agent
; GFX1064-NEXT: ; %bb.3:
; GFX1064-NEXT: s_load_dwordx2 s[52:53], s[34:35], 0x24
; GFX1064-NEXT: v_mov_b32_e32 v0, 0
-; GFX1064-NEXT: v_max_f64 v[41:42], v[2:3], v[2:3]
+; GFX1064-NEXT: v_max_f64 v[41:42], v[3:4], v[3:4]
; GFX1064-NEXT: s_mov_b64 s[54:55], 0
; GFX1064-NEXT: s_waitcnt lgkmcnt(0)
-; GFX1064-NEXT: global_load_dwordx2 v[4:5], v0, s[52:53]
+; GFX1064-NEXT: global_load_dwordx2 v[1:2], v0, s[52:53]
; GFX1064-NEXT: .LBB7_4: ; %atomicrmw.start
; GFX1064-NEXT: ; =>This Inner Loop Header: Depth=1
; GFX1064-NEXT: s_waitcnt vmcnt(0)
-; GFX1064-NEXT: v_max_f64 v[0:1], v[4:5], v[4:5]
+; GFX1064-NEXT: v_max_f64 v[3:4], v[1:2], v[1:2]
; GFX1064-NEXT: s_add_u32 s8, s34, 44
; GFX1064-NEXT: s_addc_u32 s9, s35, 0
; GFX1064-NEXT: s_getpc_b64 s[0:1]
; GFX1064-NEXT: s_add_u32 s0, s0, __atomic_compare_exchange at gotpcrel32@lo+4
; GFX1064-NEXT: s_addc_u32 s1, s1, __atomic_compare_exchange at gotpcrel32@hi+12
-; GFX1064-NEXT: buffer_store_dword v5, off, s[64:67], 0 offset:4
-; GFX1064-NEXT: buffer_store_dword v4, off, s[64:67], 0
+; GFX1064-NEXT: buffer_store_dword v2, off, s[64:67], 0 offset:4
+; GFX1064-NEXT: buffer_store_dword v1, off, s[64:67], 0
; GFX1064-NEXT: s_load_dwordx2 s[16:17], s[0:1], 0x0
; GFX1064-NEXT: v_mov_b32_e32 v31, v40
+; GFX1064-NEXT: v_mov_b32_e32 v0, 8
+; GFX1064-NEXT: v_mov_b32_e32 v1, 0
; GFX1064-NEXT: v_mov_b32_e32 v2, s52
-; GFX1064-NEXT: v_mov_b32_e32 v3, s53
-; GFX1064-NEXT: v_mov_b32_e32 v4, 0
; GFX1064-NEXT: v_mov_b32_e32 v5, 8
; GFX1064-NEXT: v_mov_b32_e32 v6, 0
; GFX1064-NEXT: v_mov_b32_e32 v7, 0
@@ -4331,16 +4331,16 @@ define amdgpu_kernel void @global_atomic_fmax_double_uni_address_div_value_agent
; GFX1064-NEXT: s_mov_b32 s13, s50
; GFX1064-NEXT: s_mov_b32 s14, s33
; GFX1064-NEXT: s_mov_b64 s[2:3], s[66:67]
-; GFX1064-NEXT: v_max_f64 v[0:1], v[0:1], v[41:42]
-; GFX1064-NEXT: buffer_store_dword v1, off, s[64:67], 0 offset:12
-; GFX1064-NEXT: buffer_store_dword v0, off, s[64:67], 0 offset:8
-; GFX1064-NEXT: v_mov_b32_e32 v0, 8
-; GFX1064-NEXT: v_mov_b32_e32 v1, 0
+; GFX1064-NEXT: v_max_f64 v[3:4], v[3:4], v[41:42]
+; GFX1064-NEXT: buffer_store_dword v4, off, s[64:67], 0 offset:12
+; GFX1064-NEXT: buffer_store_dword v3, off, s[64:67], 0 offset:8
+; GFX1064-NEXT: v_mov_b32_e32 v3, s53
+; GFX1064-NEXT: v_mov_b32_e32 v4, 0
; GFX1064-NEXT: s_waitcnt lgkmcnt(0)
; GFX1064-NEXT: s_swappc_b64 s[30:31], s[16:17]
; GFX1064-NEXT: s_clause 0x1
-; GFX1064-NEXT: buffer_load_dword v4, off, s[64:67], 0
-; GFX1064-NEXT: buffer_load_dword v5, off, s[64:67], 0 offset:4
+; GFX1064-NEXT: buffer_load_dword v1, off, s[64:67], 0
+; GFX1064-NEXT: buffer_load_dword v2, off, s[64:67], 0 offset:4
; GFX1064-NEXT: v_and_b32_e32 v0, 1, v0
; GFX1064-NEXT: v_cmp_eq_u32_e32 vcc, 1, v0
; GFX1064-NEXT: s_or_b64 s[54:55], vcc, s[54:55]
@@ -5424,24 +5424,24 @@ define amdgpu_kernel void @global_atomic_fmax_double_uni_address_uni_value_one_a
; GFX7LESS-NEXT: s_mov_b64 s[4:5], 0
; GFX7LESS-NEXT: s_mov_b32 s3, 0xf000
; GFX7LESS-NEXT: s_waitcnt lgkmcnt(0)
-; GFX7LESS-NEXT: v_mov_b32_e32 v2, s6
-; GFX7LESS-NEXT: v_mov_b32_e32 v3, s7
+; GFX7LESS-NEXT: v_mov_b32_e32 v0, s6
+; GFX7LESS-NEXT: v_mov_b32_e32 v1, s7
; GFX7LESS-NEXT: s_mov_b32 s2, -1
; GFX7LESS-NEXT: .LBB8_2: ; %atomicrmw.start
; GFX7LESS-NEXT: ; =>This Inner Loop Header: Depth=1
-; GFX7LESS-NEXT: v_max_f64 v[0:1], v[2:3], v[2:3]
-; GFX7LESS-NEXT: v_max_f64 v[0:1], v[0:1], 4.0
-; GFX7LESS-NEXT: s_waitcnt expcnt(0)
-; GFX7LESS-NEXT: v_mov_b32_e32 v7, v3
-; GFX7LESS-NEXT: v_mov_b32_e32 v6, v2
; GFX7LESS-NEXT: v_mov_b32_e32 v5, v1
; GFX7LESS-NEXT: v_mov_b32_e32 v4, v0
-; GFX7LESS-NEXT: buffer_atomic_cmpswap_x2 v[4:7], off, s[0:3], 0 glc
-; GFX7LESS-NEXT: s_waitcnt vmcnt(0)
-; GFX7LESS-NEXT: v_cmp_eq_u64_e32 vcc, v[4:5], v[2:3]
-; GFX7LESS-NEXT: s_or_b64 s[4:5], vcc, s[4:5]
+; GFX7LESS-NEXT: s_waitcnt expcnt(0)
+; GFX7LESS-NEXT: v_max_f64 v[0:1], v[4:5], v[4:5]
+; GFX7LESS-NEXT: v_max_f64 v[2:3], v[0:1], 4.0
+; GFX7LESS-NEXT: v_mov_b32_e32 v0, v2
+; GFX7LESS-NEXT: v_mov_b32_e32 v1, v3
; GFX7LESS-NEXT: v_mov_b32_e32 v2, v4
; GFX7LESS-NEXT: v_mov_b32_e32 v3, v5
+; GFX7LESS-NEXT: buffer_atomic_cmpswap_x2 v[0:3], off, s[0:3], 0 glc
+; GFX7LESS-NEXT: s_waitcnt vmcnt(0)
+; GFX7LESS-NEXT: v_cmp_eq_u64_e32 vcc, v[0:1], v[4:5]
+; GFX7LESS-NEXT: s_or_b64 s[4:5], vcc, s[4:5]
; GFX7LESS-NEXT: s_andn2_b64 exec, exec, s[4:5]
; GFX7LESS-NEXT: s_cbranch_execnz .LBB8_2
; GFX7LESS-NEXT: .LBB8_3:
@@ -5590,24 +5590,24 @@ define amdgpu_kernel void @global_atomic_fmax_double_uni_address_uni_value_one_a
; GFX7LESS-DPP-NEXT: s_mov_b64 s[4:5], 0
; GFX7LESS-DPP-NEXT: s_mov_b32 s3, 0xf000
; GFX7LESS-DPP-NEXT: s_waitcnt lgkmcnt(0)
-; GFX7LESS-DPP-NEXT: v_mov_b32_e32 v2, s6
-; GFX7LESS-DPP-NEXT: v_mov_b32_e32 v3, s7
+; GFX7LESS-DPP-NEXT: v_mov_b32_e32 v0, s6
+; GFX7LESS-DPP-NEXT: v_mov_b32_e32 v1, s7
; GFX7LESS-DPP-NEXT: s_mov_b32 s2, -1
; GFX7LESS-DPP-NEXT: .LBB8_2: ; %atomicrmw.start
; GFX7LESS-DPP-NEXT: ; =>This Inner Loop Header: Depth=1
-; GFX7LESS-DPP-NEXT: v_max_f64 v[0:1], v[2:3], v[2:3]
-; GFX7LESS-DPP-NEXT: v_max_f64 v[0:1], v[0:1], 4.0
-; GFX7LESS-DPP-NEXT: s_waitcnt expcnt(0)
-; GFX7LESS-DPP-NEXT: v_mov_b32_e32 v7, v3
-; GFX7LESS-DPP-NEXT: v_mov_b32_e32 v6, v2
; GFX7LESS-DPP-NEXT: v_mov_b32_e32 v5, v1
; GFX7LESS-DPP-NEXT: v_mov_b32_e32 v4, v0
-; GFX7LESS-DPP-NEXT: buffer_atomic_cmpswap_x2 v[4:7], off, s[0:3], 0 glc
-; GFX7LESS-DPP-NEXT: s_waitcnt vmcnt(0)
-; GFX7LESS-DPP-NEXT: v_cmp_eq_u64_e32 vcc, v[4:5], v[2:3]
-; GFX7LESS-DPP-NEXT: s_or_b64 s[4:5], vcc, s[4:5]
+; GFX7LESS-DPP-NEXT: s_waitcnt expcnt(0)
+; GFX7LESS-DPP-NEXT: v_max_f64 v[0:1], v[4:5], v[4:5]
+; GFX7LESS-DPP-NEXT: v_max_f64 v[2:3], v[0:1], 4.0
+; GFX7LESS-DPP-NEXT: v_mov_b32_e32 v0, v2
+; GFX7LESS-DPP-NEXT: v_mov_b32_e32 v1, v3
; GFX7LESS-DPP-NEXT: v_mov_b32_e32 v2, v4
; GFX7LESS-DPP-NEXT: v_mov_b32_e32 v3, v5
+; GFX7LESS-DPP-NEXT: buffer_atomic_cmpswap_x2 v[0:3], off, s[0:3], 0 glc
+; GFX7LESS-DPP-NEXT: s_waitcnt vmcnt(0)
+; GFX7LESS-DPP-NEXT: v_cmp_eq_u64_e32 vcc, v[0:1], v[4:5]
+; GFX7LESS-DPP-NEXT: s_or_b64 s[4:5], vcc, s[4:5]
; GFX7LESS-DPP-NEXT: s_andn2_b64 exec, exec, s[4:5]
; GFX7LESS-DPP-NEXT: s_cbranch_execnz .LBB8_2
; GFX7LESS-DPP-NEXT: .LBB8_3:
@@ -5777,12 +5777,12 @@ define amdgpu_kernel void @global_atomic_fmax_double_uni_address_div_value_one_a
; GFX7LESS-NEXT: s_waitcnt lgkmcnt(0)
; GFX7LESS-NEXT: s_swappc_b64 s[30:31], s[16:17]
; GFX7LESS-NEXT: s_mov_b64 s[0:1], exec
-; GFX7LESS-NEXT: v_mov_b32_e32 v4, 0
-; GFX7LESS-NEXT: v_mov_b32_e32 v5, 0x7ff80000
+; GFX7LESS-NEXT: v_mov_b32_e32 v2, 0
+; GFX7LESS-NEXT: v_mov_b32_e32 v3, 0x7ff80000
; GFX7LESS-NEXT: .LBB9_1: ; %ComputeLoop
; GFX7LESS-NEXT: ; =>This Inner Loop Header: Depth=1
; GFX7LESS-NEXT: s_ff1_i32_b64 s4, s[0:1]
-; GFX7LESS-NEXT: v_max_f64 v[2:3], v[4:5], v[4:5]
+; GFX7LESS-NEXT: v_max_f64 v[2:3], v[2:3], v[2:3]
; GFX7LESS-NEXT: v_readlane_b32 s3, v1, s4
; GFX7LESS-NEXT: v_readlane_b32 s2, v0, s4
; GFX7LESS-NEXT: s_lshl_b64 s[4:5], 1, s4
@@ -5790,7 +5790,7 @@ define amdgpu_kernel void @global_atomic_fmax_double_uni_address_div_value_one_a
; GFX7LESS-NEXT: s_andn2_b64 s[0:1], s[0:1], s[4:5]
; GFX7LESS-NEXT: v_cmp_ne_u64_e64 s[2:3], s[0:1], 0
; GFX7LESS-NEXT: s_and_b64 vcc, exec, s[2:3]
-; GFX7LESS-NEXT: v_max_f64 v[4:5], v[2:3], v[4:5]
+; GFX7LESS-NEXT: v_max_f64 v[2:3], v[2:3], v[4:5]
; GFX7LESS-NEXT: s_cbranch_vccnz .LBB9_1
; GFX7LESS-NEXT: ; %bb.2: ; %ComputeEnd
; GFX7LESS-NEXT: v_mbcnt_lo_u32_b32_e64 v0, exec_lo, 0
@@ -5804,25 +5804,25 @@ define amdgpu_kernel void @global_atomic_fmax_double_uni_address_div_value_one_a
; GFX7LESS-NEXT: s_mov_b32 s3, 0xf000
; GFX7LESS-NEXT: s_mov_b32 s2, -1
; GFX7LESS-NEXT: s_waitcnt lgkmcnt(0)
-; GFX7LESS-NEXT: buffer_load_dwordx2 v[2:3], off, s[0:3], 0
+; GFX7LESS-NEXT: buffer_load_dwordx2 v[0:1], off, s[0:3], 0
; GFX7LESS-NEXT: s_mov_b64 s[4:5], 0
-; GFX7LESS-NEXT: v_max_f64 v[4:5], v[4:5], v[4:5]
+; GFX7LESS-NEXT: v_max_f64 v[4:5], v[2:3], v[2:3]
; GFX7LESS-NEXT: .LBB9_4: ; %atomicrmw.start
; GFX7LESS-NEXT: ; =>This Inner Loop Header: Depth=1
; GFX7LESS-NEXT: s_waitcnt vmcnt(0)
-; GFX7LESS-NEXT: v_max_f64 v[0:1], v[2:3], v[2:3]
-; GFX7LESS-NEXT: v_max_f64 v[0:1], v[0:1], v[4:5]
+; GFX7LESS-NEXT: v_mov_b32_e32 v9, v1
+; GFX7LESS-NEXT: v_mov_b32_e32 v8, v0
; GFX7LESS-NEXT: s_waitcnt expcnt(0)
-; GFX7LESS-NEXT: v_mov_b32_e32 v9, v3
-; GFX7LESS-NEXT: v_mov_b32_e32 v8, v2
-; GFX7LESS-NEXT: v_mov_b32_e32 v7, v1
-; GFX7LESS-NEXT: v_mov_b32_e32 v6, v0
-; GFX7LESS-NEXT: buffer_atomic_cmpswap_x2 v[6:9], off, s[0:3], 0 glc
+; GFX7LESS-NEXT: v_max_f64 v[0:1], v[8:9], v[8:9]
+; GFX7LESS-NEXT: v_max_f64 v[6:7], v[0:1], v[4:5]
+; GFX7LESS-NEXT: v_mov_b32_e32 v0, v6
+; GFX7LESS-NEXT: v_mov_b32_e32 v1, v7
+; GFX7LESS-NEXT: v_mov_b32_e32 v2, v8
+; GFX7LESS-NEXT: v_mov_b32_e32 v3, v9
+; GFX7LESS-NEXT: buffer_atomic_cmpswap_x2 v[0:3], off, s[0:3], 0 glc
; GFX7LESS-NEXT: s_waitcnt vmcnt(0)
-; GFX7LESS-NEXT: v_cmp_eq_u64_e32 vcc, v[6:7], v[2:3]
+; GFX7LESS-NEXT: v_cmp_eq_u64_e32 vcc, v[0:1], v[8:9]
; GFX7LESS-NEXT: s_or_b64 s[4:5], vcc, s[4:5]
-; GFX7LESS-NEXT: v_mov_b32_e32 v2, v6
-; GFX7LESS-NEXT: v_mov_b32_e32 v3, v7
; GFX7LESS-NEXT: s_andn2_b64 exec, exec, s[4:5]
; GFX7LESS-NEXT: s_cbranch_execnz .LBB9_4
; GFX7LESS-NEXT: .LBB9_5:
@@ -6191,23 +6191,23 @@ define amdgpu_kernel void @global_atomic_fmax_double_uni_address_div_value_one_a
; GFX7LESS-DPP-NEXT: s_swappc_b64 s[30:31], s[16:17]
; GFX7LESS-DPP-NEXT: buffer_load_dwordx2 v[2:3], off, s[36:39], 0
; GFX7LESS-DPP-NEXT: s_mov_b64 s[0:1], 0
-; GFX7LESS-DPP-NEXT: v_max_f64 v[4:5], v[0:1], v[0:1]
+; GFX7LESS-DPP-NEXT: v_max_f64 v[0:1], v[0:1], v[0:1]
; GFX7LESS-DPP-NEXT: .LBB9_1: ; %atomicrmw.start
; GFX7LESS-DPP-NEXT: ; =>This Inner Loop Header: Depth=1
; GFX7LESS-DPP-NEXT: s_waitcnt vmcnt(0)
-; GFX7LESS-DPP-NEXT: v_max_f64 v[0:1], v[2:3], v[2:3]
-; GFX7LESS-DPP-NEXT: v_max_f64 v[0:1], v[0:1], v[4:5]
+; GFX7LESS-DPP-NEXT: v_mov_b32_e32 v7, v3
+; GFX7LESS-DPP-NEXT: v_mov_b32_e32 v6, v2
; GFX7LESS-DPP-NEXT: s_waitcnt expcnt(0)
-; GFX7LESS-DPP-NEXT: v_mov_b32_e32 v9, v3
-; GFX7LESS-DPP-NEXT: v_mov_b32_e32 v8, v2
-; GFX7LESS-DPP-NEXT: v_mov_b32_e32 v7, v1
-; GFX7LESS-DPP-NEXT: v_mov_b32_e32 v6, v0
-; GFX7LESS-DPP-NEXT: buffer_atomic_cmpswap_x2 v[6:9], off, s[36:39], 0 glc
+; GFX7LESS-DPP-NEXT: v_max_f64 v[2:3], v[6:7], v[6:7]
+; GFX7LESS-DPP-NEXT: v_max_f64 v[4:5], v[2:3], v[0:1]
+; GFX7LESS-DPP-NEXT: v_mov_b32_e32 v2, v4
+; GFX7LESS-DPP-NEXT: v_mov_b32_e32 v3, v5
+; GFX7LESS-DPP-NEXT: v_mov_b32_e32 v4, v6
+; GFX7LESS-DPP-NEXT: v_mov_b32_e32 v5, v7
+; GFX7LESS-DPP-NEXT: buffer_atomic_cmpswap_x2 v[2:5], off, s[36:39], 0 glc
; GFX7LESS-DPP-NEXT: s_waitcnt vmcnt(0)
-; GFX7LESS-DPP-NEXT: v_cmp_eq_u64_e32 vcc, v[6:7], v[2:3]
+; GFX7LESS-DPP-NEXT: v_cmp_eq_u64_e32 vcc, v[2:3], v[6:7]
; GFX7LESS-DPP-NEXT: s_or_b64 s[0:1], vcc, s[0:1]
-; GFX7LESS-DPP-NEXT: v_mov_b32_e32 v2, v6
-; GFX7LESS-DPP-NEXT: v_mov_b32_e32 v3, v7
; GFX7LESS-DPP-NEXT: s_andn2_b64 exec, exec, s[0:1]
; GFX7LESS-DPP-NEXT: s_cbranch_execnz .LBB9_1
; GFX7LESS-DPP-NEXT: ; %bb.2: ; %atomicrmw.end
@@ -7821,20 +7821,20 @@ define amdgpu_kernel void @global_atomic_fmax_double_uni_address_div_value_defau
; GFX1064-NEXT: s_movk_i32 s32, 0x800
; GFX1064-NEXT: s_waitcnt lgkmcnt(0)
; GFX1064-NEXT: s_swappc_b64 s[30:31], s[16:17]
-; GFX1064-NEXT: v_mov_b32_e32 v2, 0
-; GFX1064-NEXT: v_mov_b32_e32 v3, 0x7ff80000
+; GFX1064-NEXT: v_mov_b32_e32 v3, 0
+; GFX1064-NEXT: v_mov_b32_e32 v4, 0x7ff80000
; GFX1064-NEXT: s_mov_b64 s[0:1], exec
; GFX1064-NEXT: .LBB11_1: ; %ComputeLoop
; GFX1064-NEXT: ; =>This Inner Loop Header: Depth=1
; GFX1064-NEXT: s_ff1_i32_b64 s4, s[0:1]
-; GFX1064-NEXT: v_max_f64 v[2:3], v[2:3], v[2:3]
+; GFX1064-NEXT: v_max_f64 v[2:3], v[3:4], v[3:4]
; GFX1064-NEXT: v_readlane_b32 s3, v1, s4
; GFX1064-NEXT: v_readlane_b32 s2, v0, s4
; GFX1064-NEXT: v_max_f64 v[4:5], s[2:3], s[2:3]
; GFX1064-NEXT: s_lshl_b64 s[2:3], 1, s4
; GFX1064-NEXT: s_andn2_b64 s[0:1], s[0:1], s[2:3]
; GFX1064-NEXT: s_cmp_lg_u64 s[0:1], 0
-; GFX1064-NEXT: v_max_f64 v[2:3], v[2:3], v[4:5]
+; GFX1064-NEXT: v_max_f64 v[3:4], v[2:3], v[4:5]
; GFX1064-NEXT: s_cbranch_scc1 .LBB11_1
; GFX1064-NEXT: ; %bb.2: ; %ComputeEnd
; GFX1064-NEXT: v_mbcnt_lo_u32_b32 v0, exec_lo, 0
@@ -7846,26 +7846,26 @@ define amdgpu_kernel void @global_atomic_fmax_double_uni_address_div_value_defau
; GFX1064-NEXT: ; %bb.3:
; GFX1064-NEXT: s_load_dwordx2 s[52:53], s[34:35], 0x24
; GFX1064-NEXT: v_mov_b32_e32 v0, 0
-; GFX1064-NEXT: v_max_f64 v[41:42], v[2:3], v[2:3]
+; GFX1064-NEXT: v_max_f64 v[41:42], v[3:4], v[3:4]
; GFX1064-NEXT: s_mov_b64 s[54:55], 0
; GFX1064-NEXT: s_waitcnt lgkmcnt(0)
-; GFX1064-NEXT: global_load_dwordx2 v[4:5], v0, s[52:53]
+; GFX1064-NEXT: global_load_dwordx2 v[1:2], v0, s[52:53]
; GFX1064-NEXT: .LBB11_4: ; %atomicrmw.start
; GFX1064-NEXT: ; =>This Inner Loop Header: Depth=1
; GFX1064-NEXT: s_waitcnt vmcnt(0)
-; GFX1064-NEXT: v_max_f64 v[0:1], v[4:5], v[4:5]
+; GFX1064-NEXT: v_max_f64 v[3:4], v[1:2], v[1:2]
; GFX1064-NEXT: s_add_u32 s8, s34, 44
; GFX1064-NEXT: s_addc_u32 s9, s35, 0
; GFX1064-NEXT: s_getpc_b64 s[0:1]
; GFX1064-NEXT: s_add_u32 s0, s0, __atomic_compare_exchange at gotpcrel32@lo+4
; GFX1064-NEXT: s_addc_u32 s1, s1, __atomic_compare_exchange at gotpcrel32@hi+12
-; GFX1064-NEXT: buffer_store_dword v5, off, s[64:67], 0 offset:4
-; GFX1064-NEXT: buffer_store_dword v4, off, s[64:67], 0
+; GFX1064-NEXT: buffer_store_dword v2, off, s[64:67], 0 offset:4
+; GFX1064-NEXT: buffer_store_dword v1, off, s[64:67], 0
; GFX1064-NEXT: s_load_dwordx2 s[16:17], s[0:1], 0x0
; GFX1064-NEXT: v_mov_b32_e32 v31, v40
+; GFX1064-NEXT: v_mov_b32_e32 v0, 8
+; GFX1064-NEXT: v_mov_b32_e32 v1, 0
; GFX1064-NEXT: v_mov_b32_e32 v2, s52
-; GFX1064-NEXT: v_mov_b32_e32 v3, s53
-; GFX1064-NEXT: v_mov_b32_e32 v4, 0
; GFX1064-NEXT: v_mov_b32_e32 v5, 8
; GFX1064-NEXT: v_mov_b32_e32 v6, 0
; GFX1064-NEXT: v_mov_b32_e32 v7, 0
@@ -7877,16 +7877,16 @@ define amdgpu_kernel void @global_atomic_fmax_double_uni_address_div_value_defau
; GFX1064-NEXT: s_mov_b32 s13, s50
; GFX1064-NEXT: s_mov_b32 s14, s33
; GFX1064-NEXT: s_mov_b64 s[2:3], s[66:67]
-; GFX1064-NEXT: v_max_f64 v[0:1], v[0:1], v[41:42]
-; GFX1064-NEXT: buffer_store_dword v1, off, s[64:67], 0 offset:12
-; GFX1064-NEXT: buffer_store_dword v0, off, s[64:67], 0 offset:8
-; GFX1064-NEXT: v_mov_b32_e32 v0, 8
-; GFX1064-NEXT: v_mov_b32_e32 v1, 0
+; GFX1064-NEXT: v_max_f64 v[3:4], v[3:4], v[41:42]
+; GFX1064-NEXT: buffer_store_dword v4, off, s[64:67], 0 offset:12
+; GFX1064-NEXT: buffer_store_dword v3, off, s[64:67], 0 offset:8
+; GFX1064-NEXT: v_mov_b32_e32 v3, s53
+; GFX1064-NEXT: v_mov_b32_e32 v4, 0
; GFX1064-NEXT: s_waitcnt lgkmcnt(0)
; GFX1064-NEXT: s_swappc_b64 s[30:31], s[16:17]
; GFX1064-NEXT: s_clause 0x1
-; GFX1064-NEXT: buffer_load_dword v4, off, s[64:67], 0
-; GFX1064-NEXT: buffer_load_dword v5, off, s[64:67], 0 offset:4
+; GFX1064-NEXT: buffer_load_dword v1, off, s[64:67], 0
+; GFX1064-NEXT: buffer_load_dword v2, off, s[64:67], 0 offset:4
; GFX1064-NEXT: v_and_b32_e32 v0, 1, v0
; GFX1064-NEXT: v_cmp_eq_u32_e32 vcc, 1, v0
; GFX1064-NEXT: s_or_b64 s[54:55], vcc, s[54:55]
@@ -8970,20 +8970,20 @@ define amdgpu_kernel void @global_atomic_fmax_uni_address_uni_value_system_scope
; GFX7LESS-NEXT: s_mov_b64 s[4:5], 0
; GFX7LESS-NEXT: s_mov_b32 s3, 0xf000
; GFX7LESS-NEXT: s_waitcnt lgkmcnt(0)
-; GFX7LESS-NEXT: v_mov_b32_e32 v1, s2
+; GFX7LESS-NEXT: v_mov_b32_e32 v0, s2
; GFX7LESS-NEXT: s_mov_b32 s2, -1
; GFX7LESS-NEXT: .LBB12_2: ; %atomicrmw.start
; GFX7LESS-NEXT: ; =>This Inner Loop Header: Depth=1
-; GFX7LESS-NEXT: v_mul_f32_e32 v0, 1.0, v1
-; GFX7LESS-NEXT: v_max_f32_e32 v0, 4.0, v0
-; GFX7LESS-NEXT: s_waitcnt expcnt(0)
-; GFX7LESS-NEXT: v_mov_b32_e32 v3, v1
; GFX7LESS-NEXT: v_mov_b32_e32 v2, v0
-; GFX7LESS-NEXT: buffer_atomic_cmpswap v[2:3], off, s[0:3], 0 glc
+; GFX7LESS-NEXT: s_waitcnt expcnt(0)
+; GFX7LESS-NEXT: v_mul_f32_e32 v0, 1.0, v2
+; GFX7LESS-NEXT: v_max_f32_e32 v1, 4.0, v0
+; GFX7LESS-NEXT: v_mov_b32_e32 v0, v1
+; GFX7LESS-NEXT: v_mov_b32_e32 v1, v2
+; GFX7LESS-NEXT: buffer_atomic_cmpswap v[0:1], off, s[0:3], 0 glc
; GFX7LESS-NEXT: s_waitcnt vmcnt(0)
-; GFX7LESS-NEXT: v_cmp_eq_u32_e32 vcc, v2, v1
+; GFX7LESS-NEXT: v_cmp_eq_u32_e32 vcc, v0, v2
; GFX7LESS-NEXT: s_or_b64 s[4:5], vcc, s[4:5]
-; GFX7LESS-NEXT: v_mov_b32_e32 v1, v2
; GFX7LESS-NEXT: s_andn2_b64 exec, exec, s[4:5]
; GFX7LESS-NEXT: s_cbranch_execnz .LBB12_2
; GFX7LESS-NEXT: .LBB12_3:
@@ -9095,20 +9095,20 @@ define amdgpu_kernel void @global_atomic_fmax_uni_address_uni_value_system_scope
; GFX7LESS-DPP-NEXT: s_mov_b64 s[4:5], 0
; GFX7LESS-DPP-NEXT: s_mov_b32 s3, 0xf000
; GFX7LESS-DPP-NEXT: s_waitcnt lgkmcnt(0)
-; GFX7LESS-DPP-NEXT: v_mov_b32_e32 v1, s2
+; GFX7LESS-DPP-NEXT: v_mov_b32_e32 v0, s2
; GFX7LESS-DPP-NEXT: s_mov_b32 s2, -1
; GFX7LESS-DPP-NEXT: .LBB12_2: ; %atomicrmw.start
; GFX7LESS-DPP-NEXT: ; =>This Inner Loop Header: Depth=1
-; GFX7LESS-DPP-NEXT: v_mul_f32_e32 v0, 1.0, v1
-; GFX7LESS-DPP-NEXT: v_max_f32_e32 v0, 4.0, v0
-; GFX7LESS-DPP-NEXT: s_waitcnt expcnt(0)
-; GFX7LESS-DPP-NEXT: v_mov_b32_e32 v3, v1
; GFX7LESS-DPP-NEXT: v_mov_b32_e32 v2, v0
-; GFX7LESS-DPP-NEXT: buffer_atomic_cmpswap v[2:3], off, s[0:3], 0 glc
+; GFX7LESS-DPP-NEXT: s_waitcnt expcnt(0)
+; GFX7LESS-DPP-NEXT: v_mul_f32_e32 v0, 1.0, v2
+; GFX7LESS-DPP-NEXT: v_max_f32_e32 v1, 4.0, v0
+; GFX7LESS-DPP-NEXT: v_mov_b32_e32 v0, v1
+; GFX7LESS-DPP-NEXT: v_mov_b32_e32 v1, v2
+; GFX7LESS-DPP-NEXT: buffer_atomic_cmpswap v[0:1], off, s[0:3], 0 glc
; GFX7LESS-DPP-NEXT: s_waitcnt vmcnt(0)
-; GFX7LESS-DPP-NEXT: v_cmp_eq_u32_e32 vcc, v2, v1
+; GFX7LESS-DPP-NEXT: v_cmp_eq_u32_e32 vcc, v0, v2
; GFX7LESS-DPP-NEXT: s_or_b64 s[4:5], vcc, s[4:5]
-; GFX7LESS-DPP-NEXT: v_mov_b32_e32 v1, v2
; GFX7LESS-DPP-NEXT: s_andn2_b64 exec, exec, s[4:5]
; GFX7LESS-DPP-NEXT: s_cbranch_execnz .LBB12_2
; GFX7LESS-DPP-NEXT: .LBB12_3:
@@ -9224,20 +9224,20 @@ define amdgpu_kernel void @global_atomic_fmax_uni_address_uni_value_system_scope
; GFX7LESS-NEXT: s_mov_b64 s[4:5], 0
; GFX7LESS-NEXT: s_mov_b32 s3, 0xf000
; GFX7LESS-NEXT: s_waitcnt lgkmcnt(0)
-; GFX7LESS-NEXT: v_mov_b32_e32 v1, s2
+; GFX7LESS-NEXT: v_mov_b32_e32 v0, s2
; GFX7LESS-NEXT: s_mov_b32 s2, -1
; GFX7LESS-NEXT: .LBB13_2: ; %atomicrmw.start
; GFX7LESS-NEXT: ; =>This Inner Loop Header: Depth=1
-; GFX7LESS-NEXT: v_mul_f32_e32 v0, 1.0, v1
-; GFX7LESS-NEXT: v_max_f32_e32 v0, 4.0, v0
-; GFX7LESS-NEXT: s_waitcnt expcnt(0)
-; GFX7LESS-NEXT: v_mov_b32_e32 v3, v1
; GFX7LESS-NEXT: v_mov_b32_e32 v2, v0
-; GFX7LESS-NEXT: buffer_atomic_cmpswap v[2:3], off, s[0:3], 0 glc
+; GFX7LESS-NEXT: s_waitcnt expcnt(0)
+; GFX7LESS-NEXT: v_mul_f32_e32 v0, 1.0, v2
+; GFX7LESS-NEXT: v_max_f32_e32 v1, 4.0, v0
+; GFX7LESS-NEXT: v_mov_b32_e32 v0, v1
+; GFX7LESS-NEXT: v_mov_b32_e32 v1, v2
+; GFX7LESS-NEXT: buffer_atomic_cmpswap v[0:1], off, s[0:3], 0 glc
; GFX7LESS-NEXT: s_waitcnt vmcnt(0)
-; GFX7LESS-NEXT: v_cmp_eq_u32_e32 vcc, v2, v1
+; GFX7LESS-NEXT: v_cmp_eq_u32_e32 vcc, v0, v2
; GFX7LESS-NEXT: s_or_b64 s[4:5], vcc, s[4:5]
-; GFX7LESS-NEXT: v_mov_b32_e32 v1, v2
; GFX7LESS-NEXT: s_andn2_b64 exec, exec, s[4:5]
; GFX7LESS-NEXT: s_cbranch_execnz .LBB13_2
; GFX7LESS-NEXT: .LBB13_3:
@@ -9349,20 +9349,20 @@ define amdgpu_kernel void @global_atomic_fmax_uni_address_uni_value_system_scope
; GFX7LESS-DPP-NEXT: s_mov_b64 s[4:5], 0
; GFX7LESS-DPP-NEXT: s_mov_b32 s3, 0xf000
; GFX7LESS-DPP-NEXT: s_waitcnt lgkmcnt(0)
-; GFX7LESS-DPP-NEXT: v_mov_b32_e32 v1, s2
+; GFX7LESS-DPP-NEXT: v_mov_b32_e32 v0, s2
; GFX7LESS-DPP-NEXT: s_mov_b32 s2, -1
; GFX7LESS-DPP-NEXT: .LBB13_2: ; %atomicrmw.start
; GFX7LESS-DPP-NEXT: ; =>This Inner Loop Header: Depth=1
-; GFX7LESS-DPP-NEXT: v_mul_f32_e32 v0, 1.0, v1
-; GFX7LESS-DPP-NEXT: v_max_f32_e32 v0, 4.0, v0
-; GFX7LESS-DPP-NEXT: s_waitcnt expcnt(0)
-; GFX7LESS-DPP-NEXT: v_mov_b32_e32 v3, v1
; GFX7LESS-DPP-NEXT: v_mov_b32_e32 v2, v0
-; GFX7LESS-DPP-NEXT: buffer_atomic_cmpswap v[2:3], off, s[0:3], 0 glc
+; GFX7LESS-DPP-NEXT: s_waitcnt expcnt(0)
+; GFX7LESS-DPP-NEXT: v_mul_f32_e32 v0, 1.0, v2
+; GFX7LESS-DPP-NEXT: v_max_f32_e32 v1, 4.0, v0
+; GFX7LESS-DPP-NEXT: v_mov_b32_e32 v0, v1
+; GFX7LESS-DPP-NEXT: v_mov_b32_e32 v1, v2
+; GFX7LESS-DPP-NEXT: buffer_atomic_cmpswap v[0:1], off, s[0:3], 0 glc
; GFX7LESS-DPP-NEXT: s_waitcnt vmcnt(0)
-; GFX7LESS-DPP-NEXT: v_cmp_eq_u32_e32 vcc, v2, v1
+; GFX7LESS-DPP-NEXT: v_cmp_eq_u32_e32 vcc, v0, v2
; GFX7LESS-DPP-NEXT: s_or_b64 s[4:5], vcc, s[4:5]
-; GFX7LESS-DPP-NEXT: v_mov_b32_e32 v1, v2
; GFX7LESS-DPP-NEXT: s_andn2_b64 exec, exec, s[4:5]
; GFX7LESS-DPP-NEXT: s_cbranch_execnz .LBB13_2
; GFX7LESS-DPP-NEXT: .LBB13_3:
diff --git a/llvm/test/CodeGen/AMDGPU/global_atomics_scan_fmin.ll b/llvm/test/CodeGen/AMDGPU/global_atomics_scan_fmin.ll
index 1f76a476107a3..b3d81b5e9aec8 100644
--- a/llvm/test/CodeGen/AMDGPU/global_atomics_scan_fmin.ll
+++ b/llvm/test/CodeGen/AMDGPU/global_atomics_scan_fmin.ll
@@ -30,20 +30,20 @@ define amdgpu_kernel void @global_atomic_fmin_uni_address_uni_value_agent_scope_
; GFX7LESS-NEXT: s_mov_b64 s[4:5], 0
; GFX7LESS-NEXT: s_mov_b32 s3, 0xf000
; GFX7LESS-NEXT: s_waitcnt lgkmcnt(0)
-; GFX7LESS-NEXT: v_mov_b32_e32 v1, s2
+; GFX7LESS-NEXT: v_mov_b32_e32 v0, s2
; GFX7LESS-NEXT: s_mov_b32 s2, -1
; GFX7LESS-NEXT: .LBB0_2: ; %atomicrmw.start
; GFX7LESS-NEXT: ; =>This Inner Loop Header: Depth=1
-; GFX7LESS-NEXT: v_mul_f32_e32 v0, 1.0, v1
-; GFX7LESS-NEXT: v_min_f32_e32 v0, 4.0, v0
-; GFX7LESS-NEXT: s_waitcnt expcnt(0)
-; GFX7LESS-NEXT: v_mov_b32_e32 v3, v1
; GFX7LESS-NEXT: v_mov_b32_e32 v2, v0
-; GFX7LESS-NEXT: buffer_atomic_cmpswap v[2:3], off, s[0:3], 0 glc
+; GFX7LESS-NEXT: s_waitcnt expcnt(0)
+; GFX7LESS-NEXT: v_mul_f32_e32 v0, 1.0, v2
+; GFX7LESS-NEXT: v_min_f32_e32 v1, 4.0, v0
+; GFX7LESS-NEXT: v_mov_b32_e32 v0, v1
+; GFX7LESS-NEXT: v_mov_b32_e32 v1, v2
+; GFX7LESS-NEXT: buffer_atomic_cmpswap v[0:1], off, s[0:3], 0 glc
; GFX7LESS-NEXT: s_waitcnt vmcnt(0)
-; GFX7LESS-NEXT: v_cmp_eq_u32_e32 vcc, v2, v1
+; GFX7LESS-NEXT: v_cmp_eq_u32_e32 vcc, v0, v2
; GFX7LESS-NEXT: s_or_b64 s[4:5], vcc, s[4:5]
-; GFX7LESS-NEXT: v_mov_b32_e32 v1, v2
; GFX7LESS-NEXT: s_andn2_b64 exec, exec, s[4:5]
; GFX7LESS-NEXT: s_cbranch_execnz .LBB0_2
; GFX7LESS-NEXT: .LBB0_3:
@@ -155,20 +155,20 @@ define amdgpu_kernel void @global_atomic_fmin_uni_address_uni_value_agent_scope_
; GFX7LESS-DPP-NEXT: s_mov_b64 s[4:5], 0
; GFX7LESS-DPP-NEXT: s_mov_b32 s3, 0xf000
; GFX7LESS-DPP-NEXT: s_waitcnt lgkmcnt(0)
-; GFX7LESS-DPP-NEXT: v_mov_b32_e32 v1, s2
+; GFX7LESS-DPP-NEXT: v_mov_b32_e32 v0, s2
; GFX7LESS-DPP-NEXT: s_mov_b32 s2, -1
; GFX7LESS-DPP-NEXT: .LBB0_2: ; %atomicrmw.start
; GFX7LESS-DPP-NEXT: ; =>This Inner Loop Header: Depth=1
-; GFX7LESS-DPP-NEXT: v_mul_f32_e32 v0, 1.0, v1
-; GFX7LESS-DPP-NEXT: v_min_f32_e32 v0, 4.0, v0
-; GFX7LESS-DPP-NEXT: s_waitcnt expcnt(0)
-; GFX7LESS-DPP-NEXT: v_mov_b32_e32 v3, v1
; GFX7LESS-DPP-NEXT: v_mov_b32_e32 v2, v0
-; GFX7LESS-DPP-NEXT: buffer_atomic_cmpswap v[2:3], off, s[0:3], 0 glc
+; GFX7LESS-DPP-NEXT: s_waitcnt expcnt(0)
+; GFX7LESS-DPP-NEXT: v_mul_f32_e32 v0, 1.0, v2
+; GFX7LESS-DPP-NEXT: v_min_f32_e32 v1, 4.0, v0
+; GFX7LESS-DPP-NEXT: v_mov_b32_e32 v0, v1
+; GFX7LESS-DPP-NEXT: v_mov_b32_e32 v1, v2
+; GFX7LESS-DPP-NEXT: buffer_atomic_cmpswap v[0:1], off, s[0:3], 0 glc
; GFX7LESS-DPP-NEXT: s_waitcnt vmcnt(0)
-; GFX7LESS-DPP-NEXT: v_cmp_eq_u32_e32 vcc, v2, v1
+; GFX7LESS-DPP-NEXT: v_cmp_eq_u32_e32 vcc, v0, v2
; GFX7LESS-DPP-NEXT: s_or_b64 s[4:5], vcc, s[4:5]
-; GFX7LESS-DPP-NEXT: v_mov_b32_e32 v1, v2
; GFX7LESS-DPP-NEXT: s_andn2_b64 exec, exec, s[4:5]
; GFX7LESS-DPP-NEXT: s_cbranch_execnz .LBB0_2
; GFX7LESS-DPP-NEXT: .LBB0_3:
@@ -301,18 +301,18 @@ define amdgpu_kernel void @global_atomic_fmin_uni_address_div_value_agent_scope_
; GFX7LESS-NEXT: s_waitcnt lgkmcnt(0)
; GFX7LESS-NEXT: s_swappc_b64 s[30:31], s[16:17]
; GFX7LESS-NEXT: s_mov_b64 s[0:1], exec
-; GFX7LESS-NEXT: v_mov_b32_e32 v2, 0x7fc00000
+; GFX7LESS-NEXT: v_mov_b32_e32 v1, 0x7fc00000
; GFX7LESS-NEXT: .LBB1_1: ; %ComputeLoop
; GFX7LESS-NEXT: ; =>This Inner Loop Header: Depth=1
; GFX7LESS-NEXT: s_ff1_i32_b64 s2, s[0:1]
-; GFX7LESS-NEXT: v_mul_f32_e32 v1, 1.0, v2
+; GFX7LESS-NEXT: v_mul_f32_e32 v1, 1.0, v1
; GFX7LESS-NEXT: v_readlane_b32 s4, v0, s2
; GFX7LESS-NEXT: s_lshl_b64 s[2:3], 1, s2
; GFX7LESS-NEXT: v_mul_f32_e64 v2, 1.0, s4
; GFX7LESS-NEXT: s_andn2_b64 s[0:1], s[0:1], s[2:3]
; GFX7LESS-NEXT: v_cmp_ne_u64_e64 s[2:3], s[0:1], 0
; GFX7LESS-NEXT: s_and_b64 vcc, exec, s[2:3]
-; GFX7LESS-NEXT: v_min_f32_e32 v2, v1, v2
+; GFX7LESS-NEXT: v_min_f32_e32 v1, v1, v2
; GFX7LESS-NEXT: s_cbranch_vccnz .LBB1_1
; GFX7LESS-NEXT: ; %bb.2: ; %ComputeEnd
; GFX7LESS-NEXT: v_mbcnt_lo_u32_b32_e64 v0, exec_lo, 0
@@ -326,22 +326,22 @@ define amdgpu_kernel void @global_atomic_fmin_uni_address_div_value_agent_scope_
; GFX7LESS-NEXT: s_mov_b32 s3, 0xf000
; GFX7LESS-NEXT: s_mov_b32 s2, -1
; GFX7LESS-NEXT: s_waitcnt lgkmcnt(0)
-; GFX7LESS-NEXT: buffer_load_dword v1, off, s[0:3], 0
+; GFX7LESS-NEXT: buffer_load_dword v0, off, s[0:3], 0
; GFX7LESS-NEXT: s_mov_b64 s[4:5], 0
-; GFX7LESS-NEXT: v_mul_f32_e32 v2, 1.0, v2
+; GFX7LESS-NEXT: v_mul_f32_e32 v2, 1.0, v1
; GFX7LESS-NEXT: .LBB1_4: ; %atomicrmw.start
; GFX7LESS-NEXT: ; =>This Inner Loop Header: Depth=1
; GFX7LESS-NEXT: s_waitcnt vmcnt(0)
-; GFX7LESS-NEXT: v_mul_f32_e32 v0, 1.0, v1
-; GFX7LESS-NEXT: v_min_f32_e32 v0, v0, v2
+; GFX7LESS-NEXT: v_mov_b32_e32 v4, v0
; GFX7LESS-NEXT: s_waitcnt expcnt(0)
-; GFX7LESS-NEXT: v_mov_b32_e32 v4, v1
-; GFX7LESS-NEXT: v_mov_b32_e32 v3, v0
-; GFX7LESS-NEXT: buffer_atomic_cmpswap v[3:4], off, s[0:3], 0 glc
+; GFX7LESS-NEXT: v_mul_f32_e32 v0, 1.0, v4
+; GFX7LESS-NEXT: v_min_f32_e32 v3, v0, v2
+; GFX7LESS-NEXT: v_mov_b32_e32 v0, v3
+; GFX7LESS-NEXT: v_mov_b32_e32 v1, v4
+; GFX7LESS-NEXT: buffer_atomic_cmpswap v[0:1], off, s[0:3], 0 glc
; GFX7LESS-NEXT: s_waitcnt vmcnt(0)
-; GFX7LESS-NEXT: v_cmp_eq_u32_e32 vcc, v3, v1
+; GFX7LESS-NEXT: v_cmp_eq_u32_e32 vcc, v0, v4
; GFX7LESS-NEXT: s_or_b64 s[4:5], vcc, s[4:5]
-; GFX7LESS-NEXT: v_mov_b32_e32 v1, v3
; GFX7LESS-NEXT: s_andn2_b64 exec, exec, s[4:5]
; GFX7LESS-NEXT: s_cbranch_execnz .LBB1_4
; GFX7LESS-NEXT: .LBB1_5:
@@ -666,20 +666,20 @@ define amdgpu_kernel void @global_atomic_fmin_uni_address_div_value_agent_scope_
; GFX7LESS-DPP-NEXT: s_swappc_b64 s[30:31], s[16:17]
; GFX7LESS-DPP-NEXT: buffer_load_dword v1, off, s[36:39], 0
; GFX7LESS-DPP-NEXT: s_mov_b64 s[0:1], 0
-; GFX7LESS-DPP-NEXT: v_mul_f32_e32 v2, 1.0, v0
+; GFX7LESS-DPP-NEXT: v_mul_f32_e32 v0, 1.0, v0
; GFX7LESS-DPP-NEXT: .LBB1_1: ; %atomicrmw.start
; GFX7LESS-DPP-NEXT: ; =>This Inner Loop Header: Depth=1
; GFX7LESS-DPP-NEXT: s_waitcnt vmcnt(0)
-; GFX7LESS-DPP-NEXT: v_mul_f32_e32 v0, 1.0, v1
-; GFX7LESS-DPP-NEXT: v_min_f32_e32 v0, v0, v2
+; GFX7LESS-DPP-NEXT: v_mov_b32_e32 v3, v1
; GFX7LESS-DPP-NEXT: s_waitcnt expcnt(0)
-; GFX7LESS-DPP-NEXT: v_mov_b32_e32 v4, v1
-; GFX7LESS-DPP-NEXT: v_mov_b32_e32 v3, v0
-; GFX7LESS-DPP-NEXT: buffer_atomic_cmpswap v[3:4], off, s[36:39], 0 glc
+; GFX7LESS-DPP-NEXT: v_mul_f32_e32 v1, 1.0, v3
+; GFX7LESS-DPP-NEXT: v_min_f32_e32 v2, v1, v0
+; GFX7LESS-DPP-NEXT: v_mov_b32_e32 v1, v2
+; GFX7LESS-DPP-NEXT: v_mov_b32_e32 v2, v3
+; GFX7LESS-DPP-NEXT: buffer_atomic_cmpswap v[1:2], off, s[36:39], 0 glc
; GFX7LESS-DPP-NEXT: s_waitcnt vmcnt(0)
-; GFX7LESS-DPP-NEXT: v_cmp_eq_u32_e32 vcc, v3, v1
+; GFX7LESS-DPP-NEXT: v_cmp_eq_u32_e32 vcc, v1, v3
; GFX7LESS-DPP-NEXT: s_or_b64 s[0:1], vcc, s[0:1]
-; GFX7LESS-DPP-NEXT: v_mov_b32_e32 v1, v3
; GFX7LESS-DPP-NEXT: s_andn2_b64 exec, exec, s[0:1]
; GFX7LESS-DPP-NEXT: s_cbranch_execnz .LBB1_1
; GFX7LESS-DPP-NEXT: ; %bb.2: ; %atomicrmw.end
@@ -1068,20 +1068,20 @@ define amdgpu_kernel void @global_atomic_fmin_uni_address_uni_value_one_as_scope
; GFX7LESS-NEXT: s_mov_b64 s[4:5], 0
; GFX7LESS-NEXT: s_mov_b32 s3, 0xf000
; GFX7LESS-NEXT: s_waitcnt lgkmcnt(0)
-; GFX7LESS-NEXT: v_mov_b32_e32 v1, s2
+; GFX7LESS-NEXT: v_mov_b32_e32 v0, s2
; GFX7LESS-NEXT: s_mov_b32 s2, -1
; GFX7LESS-NEXT: .LBB2_2: ; %atomicrmw.start
; GFX7LESS-NEXT: ; =>This Inner Loop Header: Depth=1
-; GFX7LESS-NEXT: v_mul_f32_e32 v0, 1.0, v1
-; GFX7LESS-NEXT: v_min_f32_e32 v0, 4.0, v0
-; GFX7LESS-NEXT: s_waitcnt expcnt(0)
-; GFX7LESS-NEXT: v_mov_b32_e32 v3, v1
; GFX7LESS-NEXT: v_mov_b32_e32 v2, v0
-; GFX7LESS-NEXT: buffer_atomic_cmpswap v[2:3], off, s[0:3], 0 glc
+; GFX7LESS-NEXT: s_waitcnt expcnt(0)
+; GFX7LESS-NEXT: v_mul_f32_e32 v0, 1.0, v2
+; GFX7LESS-NEXT: v_min_f32_e32 v1, 4.0, v0
+; GFX7LESS-NEXT: v_mov_b32_e32 v0, v1
+; GFX7LESS-NEXT: v_mov_b32_e32 v1, v2
+; GFX7LESS-NEXT: buffer_atomic_cmpswap v[0:1], off, s[0:3], 0 glc
; GFX7LESS-NEXT: s_waitcnt vmcnt(0)
-; GFX7LESS-NEXT: v_cmp_eq_u32_e32 vcc, v2, v1
+; GFX7LESS-NEXT: v_cmp_eq_u32_e32 vcc, v0, v2
; GFX7LESS-NEXT: s_or_b64 s[4:5], vcc, s[4:5]
-; GFX7LESS-NEXT: v_mov_b32_e32 v1, v2
; GFX7LESS-NEXT: s_andn2_b64 exec, exec, s[4:5]
; GFX7LESS-NEXT: s_cbranch_execnz .LBB2_2
; GFX7LESS-NEXT: .LBB2_3:
@@ -1193,20 +1193,20 @@ define amdgpu_kernel void @global_atomic_fmin_uni_address_uni_value_one_as_scope
; GFX7LESS-DPP-NEXT: s_mov_b64 s[4:5], 0
; GFX7LESS-DPP-NEXT: s_mov_b32 s3, 0xf000
; GFX7LESS-DPP-NEXT: s_waitcnt lgkmcnt(0)
-; GFX7LESS-DPP-NEXT: v_mov_b32_e32 v1, s2
+; GFX7LESS-DPP-NEXT: v_mov_b32_e32 v0, s2
; GFX7LESS-DPP-NEXT: s_mov_b32 s2, -1
; GFX7LESS-DPP-NEXT: .LBB2_2: ; %atomicrmw.start
; GFX7LESS-DPP-NEXT: ; =>This Inner Loop Header: Depth=1
-; GFX7LESS-DPP-NEXT: v_mul_f32_e32 v0, 1.0, v1
-; GFX7LESS-DPP-NEXT: v_min_f32_e32 v0, 4.0, v0
-; GFX7LESS-DPP-NEXT: s_waitcnt expcnt(0)
-; GFX7LESS-DPP-NEXT: v_mov_b32_e32 v3, v1
; GFX7LESS-DPP-NEXT: v_mov_b32_e32 v2, v0
-; GFX7LESS-DPP-NEXT: buffer_atomic_cmpswap v[2:3], off, s[0:3], 0 glc
+; GFX7LESS-DPP-NEXT: s_waitcnt expcnt(0)
+; GFX7LESS-DPP-NEXT: v_mul_f32_e32 v0, 1.0, v2
+; GFX7LESS-DPP-NEXT: v_min_f32_e32 v1, 4.0, v0
+; GFX7LESS-DPP-NEXT: v_mov_b32_e32 v0, v1
+; GFX7LESS-DPP-NEXT: v_mov_b32_e32 v1, v2
+; GFX7LESS-DPP-NEXT: buffer_atomic_cmpswap v[0:1], off, s[0:3], 0 glc
; GFX7LESS-DPP-NEXT: s_waitcnt vmcnt(0)
-; GFX7LESS-DPP-NEXT: v_cmp_eq_u32_e32 vcc, v2, v1
+; GFX7LESS-DPP-NEXT: v_cmp_eq_u32_e32 vcc, v0, v2
; GFX7LESS-DPP-NEXT: s_or_b64 s[4:5], vcc, s[4:5]
-; GFX7LESS-DPP-NEXT: v_mov_b32_e32 v1, v2
; GFX7LESS-DPP-NEXT: s_andn2_b64 exec, exec, s[4:5]
; GFX7LESS-DPP-NEXT: s_cbranch_execnz .LBB2_2
; GFX7LESS-DPP-NEXT: .LBB2_3:
@@ -1340,18 +1340,18 @@ define amdgpu_kernel void @global_atomic_fmin_uni_address_div_value_one_as_scope
; GFX7LESS-NEXT: s_waitcnt lgkmcnt(0)
; GFX7LESS-NEXT: s_swappc_b64 s[30:31], s[16:17]
; GFX7LESS-NEXT: s_mov_b64 s[0:1], exec
-; GFX7LESS-NEXT: v_mov_b32_e32 v2, 0x7fc00000
+; GFX7LESS-NEXT: v_mov_b32_e32 v1, 0x7fc00000
; GFX7LESS-NEXT: .LBB3_1: ; %ComputeLoop
; GFX7LESS-NEXT: ; =>This Inner Loop Header: Depth=1
; GFX7LESS-NEXT: s_ff1_i32_b64 s2, s[0:1]
-; GFX7LESS-NEXT: v_mul_f32_e32 v1, 1.0, v2
+; GFX7LESS-NEXT: v_mul_f32_e32 v1, 1.0, v1
; GFX7LESS-NEXT: v_readlane_b32 s4, v0, s2
; GFX7LESS-NEXT: s_lshl_b64 s[2:3], 1, s2
; GFX7LESS-NEXT: v_mul_f32_e64 v2, 1.0, s4
; GFX7LESS-NEXT: s_andn2_b64 s[0:1], s[0:1], s[2:3]
; GFX7LESS-NEXT: v_cmp_ne_u64_e64 s[2:3], s[0:1], 0
; GFX7LESS-NEXT: s_and_b64 vcc, exec, s[2:3]
-; GFX7LESS-NEXT: v_min_f32_e32 v2, v1, v2
+; GFX7LESS-NEXT: v_min_f32_e32 v1, v1, v2
; GFX7LESS-NEXT: s_cbranch_vccnz .LBB3_1
; GFX7LESS-NEXT: ; %bb.2: ; %ComputeEnd
; GFX7LESS-NEXT: v_mbcnt_lo_u32_b32_e64 v0, exec_lo, 0
@@ -1365,22 +1365,22 @@ define amdgpu_kernel void @global_atomic_fmin_uni_address_div_value_one_as_scope
; GFX7LESS-NEXT: s_mov_b32 s3, 0xf000
; GFX7LESS-NEXT: s_mov_b32 s2, -1
; GFX7LESS-NEXT: s_waitcnt lgkmcnt(0)
-; GFX7LESS-NEXT: buffer_load_dword v1, off, s[0:3], 0
+; GFX7LESS-NEXT: buffer_load_dword v0, off, s[0:3], 0
; GFX7LESS-NEXT: s_mov_b64 s[4:5], 0
-; GFX7LESS-NEXT: v_mul_f32_e32 v2, 1.0, v2
+; GFX7LESS-NEXT: v_mul_f32_e32 v2, 1.0, v1
; GFX7LESS-NEXT: .LBB3_4: ; %atomicrmw.start
; GFX7LESS-NEXT: ; =>This Inner Loop Header: Depth=1
; GFX7LESS-NEXT: s_waitcnt vmcnt(0)
-; GFX7LESS-NEXT: v_mul_f32_e32 v0, 1.0, v1
-; GFX7LESS-NEXT: v_min_f32_e32 v0, v0, v2
+; GFX7LESS-NEXT: v_mov_b32_e32 v4, v0
; GFX7LESS-NEXT: s_waitcnt expcnt(0)
-; GFX7LESS-NEXT: v_mov_b32_e32 v4, v1
-; GFX7LESS-NEXT: v_mov_b32_e32 v3, v0
-; GFX7LESS-NEXT: buffer_atomic_cmpswap v[3:4], off, s[0:3], 0 glc
+; GFX7LESS-NEXT: v_mul_f32_e32 v0, 1.0, v4
+; GFX7LESS-NEXT: v_min_f32_e32 v3, v0, v2
+; GFX7LESS-NEXT: v_mov_b32_e32 v0, v3
+; GFX7LESS-NEXT: v_mov_b32_e32 v1, v4
+; GFX7LESS-NEXT: buffer_atomic_cmpswap v[0:1], off, s[0:3], 0 glc
; GFX7LESS-NEXT: s_waitcnt vmcnt(0)
-; GFX7LESS-NEXT: v_cmp_eq_u32_e32 vcc, v3, v1
+; GFX7LESS-NEXT: v_cmp_eq_u32_e32 vcc, v0, v4
; GFX7LESS-NEXT: s_or_b64 s[4:5], vcc, s[4:5]
-; GFX7LESS-NEXT: v_mov_b32_e32 v1, v3
; GFX7LESS-NEXT: s_andn2_b64 exec, exec, s[4:5]
; GFX7LESS-NEXT: s_cbranch_execnz .LBB3_4
; GFX7LESS-NEXT: .LBB3_5:
@@ -1705,20 +1705,20 @@ define amdgpu_kernel void @global_atomic_fmin_uni_address_div_value_one_as_scope
; GFX7LESS-DPP-NEXT: s_swappc_b64 s[30:31], s[16:17]
; GFX7LESS-DPP-NEXT: buffer_load_dword v1, off, s[36:39], 0
; GFX7LESS-DPP-NEXT: s_mov_b64 s[0:1], 0
-; GFX7LESS-DPP-NEXT: v_mul_f32_e32 v2, 1.0, v0
+; GFX7LESS-DPP-NEXT: v_mul_f32_e32 v0, 1.0, v0
; GFX7LESS-DPP-NEXT: .LBB3_1: ; %atomicrmw.start
; GFX7LESS-DPP-NEXT: ; =>This Inner Loop Header: Depth=1
; GFX7LESS-DPP-NEXT: s_waitcnt vmcnt(0)
-; GFX7LESS-DPP-NEXT: v_mul_f32_e32 v0, 1.0, v1
-; GFX7LESS-DPP-NEXT: v_min_f32_e32 v0, v0, v2
+; GFX7LESS-DPP-NEXT: v_mov_b32_e32 v3, v1
; GFX7LESS-DPP-NEXT: s_waitcnt expcnt(0)
-; GFX7LESS-DPP-NEXT: v_mov_b32_e32 v4, v1
-; GFX7LESS-DPP-NEXT: v_mov_b32_e32 v3, v0
-; GFX7LESS-DPP-NEXT: buffer_atomic_cmpswap v[3:4], off, s[36:39], 0 glc
+; GFX7LESS-DPP-NEXT: v_mul_f32_e32 v1, 1.0, v3
+; GFX7LESS-DPP-NEXT: v_min_f32_e32 v2, v1, v0
+; GFX7LESS-DPP-NEXT: v_mov_b32_e32 v1, v2
+; GFX7LESS-DPP-NEXT: v_mov_b32_e32 v2, v3
+; GFX7LESS-DPP-NEXT: buffer_atomic_cmpswap v[1:2], off, s[36:39], 0 glc
; GFX7LESS-DPP-NEXT: s_waitcnt vmcnt(0)
-; GFX7LESS-DPP-NEXT: v_cmp_eq_u32_e32 vcc, v3, v1
+; GFX7LESS-DPP-NEXT: v_cmp_eq_u32_e32 vcc, v1, v3
; GFX7LESS-DPP-NEXT: s_or_b64 s[0:1], vcc, s[0:1]
-; GFX7LESS-DPP-NEXT: v_mov_b32_e32 v1, v3
; GFX7LESS-DPP-NEXT: s_andn2_b64 exec, exec, s[0:1]
; GFX7LESS-DPP-NEXT: s_cbranch_execnz .LBB3_1
; GFX7LESS-DPP-NEXT: ; %bb.2: ; %atomicrmw.end
@@ -2108,20 +2108,20 @@ define amdgpu_kernel void @global_atomic_fmin_uni_address_uni_value_default_scop
; GFX7LESS-NEXT: s_mov_b64 s[4:5], 0
; GFX7LESS-NEXT: s_mov_b32 s3, 0xf000
; GFX7LESS-NEXT: s_waitcnt lgkmcnt(0)
-; GFX7LESS-NEXT: v_mov_b32_e32 v1, s2
+; GFX7LESS-NEXT: v_mov_b32_e32 v0, s2
; GFX7LESS-NEXT: s_mov_b32 s2, -1
; GFX7LESS-NEXT: .LBB4_2: ; %atomicrmw.start
; GFX7LESS-NEXT: ; =>This Inner Loop Header: Depth=1
-; GFX7LESS-NEXT: v_mul_f32_e32 v0, 1.0, v1
-; GFX7LESS-NEXT: v_min_f32_e32 v0, 4.0, v0
-; GFX7LESS-NEXT: s_waitcnt expcnt(0)
-; GFX7LESS-NEXT: v_mov_b32_e32 v3, v1
; GFX7LESS-NEXT: v_mov_b32_e32 v2, v0
-; GFX7LESS-NEXT: buffer_atomic_cmpswap v[2:3], off, s[0:3], 0 glc
+; GFX7LESS-NEXT: s_waitcnt expcnt(0)
+; GFX7LESS-NEXT: v_mul_f32_e32 v0, 1.0, v2
+; GFX7LESS-NEXT: v_min_f32_e32 v1, 4.0, v0
+; GFX7LESS-NEXT: v_mov_b32_e32 v0, v1
+; GFX7LESS-NEXT: v_mov_b32_e32 v1, v2
+; GFX7LESS-NEXT: buffer_atomic_cmpswap v[0:1], off, s[0:3], 0 glc
; GFX7LESS-NEXT: s_waitcnt vmcnt(0)
-; GFX7LESS-NEXT: v_cmp_eq_u32_e32 vcc, v2, v1
+; GFX7LESS-NEXT: v_cmp_eq_u32_e32 vcc, v0, v2
; GFX7LESS-NEXT: s_or_b64 s[4:5], vcc, s[4:5]
-; GFX7LESS-NEXT: v_mov_b32_e32 v1, v2
; GFX7LESS-NEXT: s_andn2_b64 exec, exec, s[4:5]
; GFX7LESS-NEXT: s_cbranch_execnz .LBB4_2
; GFX7LESS-NEXT: .LBB4_3:
@@ -2233,20 +2233,20 @@ define amdgpu_kernel void @global_atomic_fmin_uni_address_uni_value_default_scop
; GFX7LESS-DPP-NEXT: s_mov_b64 s[4:5], 0
; GFX7LESS-DPP-NEXT: s_mov_b32 s3, 0xf000
; GFX7LESS-DPP-NEXT: s_waitcnt lgkmcnt(0)
-; GFX7LESS-DPP-NEXT: v_mov_b32_e32 v1, s2
+; GFX7LESS-DPP-NEXT: v_mov_b32_e32 v0, s2
; GFX7LESS-DPP-NEXT: s_mov_b32 s2, -1
; GFX7LESS-DPP-NEXT: .LBB4_2: ; %atomicrmw.start
; GFX7LESS-DPP-NEXT: ; =>This Inner Loop Header: Depth=1
-; GFX7LESS-DPP-NEXT: v_mul_f32_e32 v0, 1.0, v1
-; GFX7LESS-DPP-NEXT: v_min_f32_e32 v0, 4.0, v0
-; GFX7LESS-DPP-NEXT: s_waitcnt expcnt(0)
-; GFX7LESS-DPP-NEXT: v_mov_b32_e32 v3, v1
; GFX7LESS-DPP-NEXT: v_mov_b32_e32 v2, v0
-; GFX7LESS-DPP-NEXT: buffer_atomic_cmpswap v[2:3], off, s[0:3], 0 glc
+; GFX7LESS-DPP-NEXT: s_waitcnt expcnt(0)
+; GFX7LESS-DPP-NEXT: v_mul_f32_e32 v0, 1.0, v2
+; GFX7LESS-DPP-NEXT: v_min_f32_e32 v1, 4.0, v0
+; GFX7LESS-DPP-NEXT: v_mov_b32_e32 v0, v1
+; GFX7LESS-DPP-NEXT: v_mov_b32_e32 v1, v2
+; GFX7LESS-DPP-NEXT: buffer_atomic_cmpswap v[0:1], off, s[0:3], 0 glc
; GFX7LESS-DPP-NEXT: s_waitcnt vmcnt(0)
-; GFX7LESS-DPP-NEXT: v_cmp_eq_u32_e32 vcc, v2, v1
+; GFX7LESS-DPP-NEXT: v_cmp_eq_u32_e32 vcc, v0, v2
; GFX7LESS-DPP-NEXT: s_or_b64 s[4:5], vcc, s[4:5]
-; GFX7LESS-DPP-NEXT: v_mov_b32_e32 v1, v2
; GFX7LESS-DPP-NEXT: s_andn2_b64 exec, exec, s[4:5]
; GFX7LESS-DPP-NEXT: s_cbranch_execnz .LBB4_2
; GFX7LESS-DPP-NEXT: .LBB4_3:
@@ -2379,18 +2379,18 @@ define amdgpu_kernel void @global_atomic_fmin_uni_address_div_value_default_scop
; GFX7LESS-NEXT: s_waitcnt lgkmcnt(0)
; GFX7LESS-NEXT: s_swappc_b64 s[30:31], s[16:17]
; GFX7LESS-NEXT: s_mov_b64 s[0:1], exec
-; GFX7LESS-NEXT: v_mov_b32_e32 v2, 0x7fc00000
+; GFX7LESS-NEXT: v_mov_b32_e32 v1, 0x7fc00000
; GFX7LESS-NEXT: .LBB5_1: ; %ComputeLoop
; GFX7LESS-NEXT: ; =>This Inner Loop Header: Depth=1
; GFX7LESS-NEXT: s_ff1_i32_b64 s2, s[0:1]
-; GFX7LESS-NEXT: v_mul_f32_e32 v1, 1.0, v2
+; GFX7LESS-NEXT: v_mul_f32_e32 v1, 1.0, v1
; GFX7LESS-NEXT: v_readlane_b32 s4, v0, s2
; GFX7LESS-NEXT: s_lshl_b64 s[2:3], 1, s2
; GFX7LESS-NEXT: v_mul_f32_e64 v2, 1.0, s4
; GFX7LESS-NEXT: s_andn2_b64 s[0:1], s[0:1], s[2:3]
; GFX7LESS-NEXT: v_cmp_ne_u64_e64 s[2:3], s[0:1], 0
; GFX7LESS-NEXT: s_and_b64 vcc, exec, s[2:3]
-; GFX7LESS-NEXT: v_min_f32_e32 v2, v1, v2
+; GFX7LESS-NEXT: v_min_f32_e32 v1, v1, v2
; GFX7LESS-NEXT: s_cbranch_vccnz .LBB5_1
; GFX7LESS-NEXT: ; %bb.2: ; %ComputeEnd
; GFX7LESS-NEXT: v_mbcnt_lo_u32_b32_e64 v0, exec_lo, 0
@@ -2404,22 +2404,22 @@ define amdgpu_kernel void @global_atomic_fmin_uni_address_div_value_default_scop
; GFX7LESS-NEXT: s_mov_b32 s3, 0xf000
; GFX7LESS-NEXT: s_mov_b32 s2, -1
; GFX7LESS-NEXT: s_waitcnt lgkmcnt(0)
-; GFX7LESS-NEXT: buffer_load_dword v1, off, s[0:3], 0
+; GFX7LESS-NEXT: buffer_load_dword v0, off, s[0:3], 0
; GFX7LESS-NEXT: s_mov_b64 s[4:5], 0
-; GFX7LESS-NEXT: v_mul_f32_e32 v2, 1.0, v2
+; GFX7LESS-NEXT: v_mul_f32_e32 v2, 1.0, v1
; GFX7LESS-NEXT: .LBB5_4: ; %atomicrmw.start
; GFX7LESS-NEXT: ; =>This Inner Loop Header: Depth=1
; GFX7LESS-NEXT: s_waitcnt vmcnt(0)
-; GFX7LESS-NEXT: v_mul_f32_e32 v0, 1.0, v1
-; GFX7LESS-NEXT: v_min_f32_e32 v0, v0, v2
+; GFX7LESS-NEXT: v_mov_b32_e32 v4, v0
; GFX7LESS-NEXT: s_waitcnt expcnt(0)
-; GFX7LESS-NEXT: v_mov_b32_e32 v4, v1
-; GFX7LESS-NEXT: v_mov_b32_e32 v3, v0
-; GFX7LESS-NEXT: buffer_atomic_cmpswap v[3:4], off, s[0:3], 0 glc
+; GFX7LESS-NEXT: v_mul_f32_e32 v0, 1.0, v4
+; GFX7LESS-NEXT: v_min_f32_e32 v3, v0, v2
+; GFX7LESS-NEXT: v_mov_b32_e32 v0, v3
+; GFX7LESS-NEXT: v_mov_b32_e32 v1, v4
+; GFX7LESS-NEXT: buffer_atomic_cmpswap v[0:1], off, s[0:3], 0 glc
; GFX7LESS-NEXT: s_waitcnt vmcnt(0)
-; GFX7LESS-NEXT: v_cmp_eq_u32_e32 vcc, v3, v1
+; GFX7LESS-NEXT: v_cmp_eq_u32_e32 vcc, v0, v4
; GFX7LESS-NEXT: s_or_b64 s[4:5], vcc, s[4:5]
-; GFX7LESS-NEXT: v_mov_b32_e32 v1, v3
; GFX7LESS-NEXT: s_andn2_b64 exec, exec, s[4:5]
; GFX7LESS-NEXT: s_cbranch_execnz .LBB5_4
; GFX7LESS-NEXT: .LBB5_5:
@@ -2744,20 +2744,20 @@ define amdgpu_kernel void @global_atomic_fmin_uni_address_div_value_default_scop
; GFX7LESS-DPP-NEXT: s_swappc_b64 s[30:31], s[16:17]
; GFX7LESS-DPP-NEXT: buffer_load_dword v1, off, s[36:39], 0
; GFX7LESS-DPP-NEXT: s_mov_b64 s[0:1], 0
-; GFX7LESS-DPP-NEXT: v_mul_f32_e32 v2, 1.0, v0
+; GFX7LESS-DPP-NEXT: v_mul_f32_e32 v0, 1.0, v0
; GFX7LESS-DPP-NEXT: .LBB5_1: ; %atomicrmw.start
; GFX7LESS-DPP-NEXT: ; =>This Inner Loop Header: Depth=1
; GFX7LESS-DPP-NEXT: s_waitcnt vmcnt(0)
-; GFX7LESS-DPP-NEXT: v_mul_f32_e32 v0, 1.0, v1
-; GFX7LESS-DPP-NEXT: v_min_f32_e32 v0, v0, v2
+; GFX7LESS-DPP-NEXT: v_mov_b32_e32 v3, v1
; GFX7LESS-DPP-NEXT: s_waitcnt expcnt(0)
-; GFX7LESS-DPP-NEXT: v_mov_b32_e32 v4, v1
-; GFX7LESS-DPP-NEXT: v_mov_b32_e32 v3, v0
-; GFX7LESS-DPP-NEXT: buffer_atomic_cmpswap v[3:4], off, s[36:39], 0 glc
+; GFX7LESS-DPP-NEXT: v_mul_f32_e32 v1, 1.0, v3
+; GFX7LESS-DPP-NEXT: v_min_f32_e32 v2, v1, v0
+; GFX7LESS-DPP-NEXT: v_mov_b32_e32 v1, v2
+; GFX7LESS-DPP-NEXT: v_mov_b32_e32 v2, v3
+; GFX7LESS-DPP-NEXT: buffer_atomic_cmpswap v[1:2], off, s[36:39], 0 glc
; GFX7LESS-DPP-NEXT: s_waitcnt vmcnt(0)
-; GFX7LESS-DPP-NEXT: v_cmp_eq_u32_e32 vcc, v3, v1
+; GFX7LESS-DPP-NEXT: v_cmp_eq_u32_e32 vcc, v1, v3
; GFX7LESS-DPP-NEXT: s_or_b64 s[0:1], vcc, s[0:1]
-; GFX7LESS-DPP-NEXT: v_mov_b32_e32 v1, v3
; GFX7LESS-DPP-NEXT: s_andn2_b64 exec, exec, s[0:1]
; GFX7LESS-DPP-NEXT: s_cbranch_execnz .LBB5_1
; GFX7LESS-DPP-NEXT: ; %bb.2: ; %atomicrmw.end
@@ -4275,20 +4275,20 @@ define amdgpu_kernel void @global_atomic_fmin_double_uni_address_div_value_agent
; GFX1064-NEXT: s_movk_i32 s32, 0x800
; GFX1064-NEXT: s_waitcnt lgkmcnt(0)
; GFX1064-NEXT: s_swappc_b64 s[30:31], s[16:17]
-; GFX1064-NEXT: v_mov_b32_e32 v2, 0
-; GFX1064-NEXT: v_mov_b32_e32 v3, 0x7ff80000
+; GFX1064-NEXT: v_mov_b32_e32 v3, 0
+; GFX1064-NEXT: v_mov_b32_e32 v4, 0x7ff80000
; GFX1064-NEXT: s_mov_b64 s[0:1], exec
; GFX1064-NEXT: .LBB7_1: ; %ComputeLoop
; GFX1064-NEXT: ; =>This Inner Loop Header: Depth=1
; GFX1064-NEXT: s_ff1_i32_b64 s4, s[0:1]
-; GFX1064-NEXT: v_max_f64 v[2:3], v[2:3], v[2:3]
+; GFX1064-NEXT: v_max_f64 v[2:3], v[3:4], v[3:4]
; GFX1064-NEXT: v_readlane_b32 s3, v1, s4
; GFX1064-NEXT: v_readlane_b32 s2, v0, s4
; GFX1064-NEXT: v_max_f64 v[4:5], s[2:3], s[2:3]
; GFX1064-NEXT: s_lshl_b64 s[2:3], 1, s4
; GFX1064-NEXT: s_andn2_b64 s[0:1], s[0:1], s[2:3]
; GFX1064-NEXT: s_cmp_lg_u64 s[0:1], 0
-; GFX1064-NEXT: v_min_f64 v[2:3], v[2:3], v[4:5]
+; GFX1064-NEXT: v_min_f64 v[3:4], v[2:3], v[4:5]
; GFX1064-NEXT: s_cbranch_scc1 .LBB7_1
; GFX1064-NEXT: ; %bb.2: ; %ComputeEnd
; GFX1064-NEXT: v_mbcnt_lo_u32_b32 v0, exec_lo, 0
@@ -4300,26 +4300,26 @@ define amdgpu_kernel void @global_atomic_fmin_double_uni_address_div_value_agent
; GFX1064-NEXT: ; %bb.3:
; GFX1064-NEXT: s_load_dwordx2 s[52:53], s[34:35], 0x24
; GFX1064-NEXT: v_mov_b32_e32 v0, 0
-; GFX1064-NEXT: v_max_f64 v[41:42], v[2:3], v[2:3]
+; GFX1064-NEXT: v_max_f64 v[41:42], v[3:4], v[3:4]
; GFX1064-NEXT: s_mov_b64 s[54:55], 0
; GFX1064-NEXT: s_waitcnt lgkmcnt(0)
-; GFX1064-NEXT: global_load_dwordx2 v[4:5], v0, s[52:53]
+; GFX1064-NEXT: global_load_dwordx2 v[1:2], v0, s[52:53]
; GFX1064-NEXT: .LBB7_4: ; %atomicrmw.start
; GFX1064-NEXT: ; =>This Inner Loop Header: Depth=1
; GFX1064-NEXT: s_waitcnt vmcnt(0)
-; GFX1064-NEXT: v_max_f64 v[0:1], v[4:5], v[4:5]
+; GFX1064-NEXT: v_max_f64 v[3:4], v[1:2], v[1:2]
; GFX1064-NEXT: s_add_u32 s8, s34, 44
; GFX1064-NEXT: s_addc_u32 s9, s35, 0
; GFX1064-NEXT: s_getpc_b64 s[0:1]
; GFX1064-NEXT: s_add_u32 s0, s0, __atomic_compare_exchange at gotpcrel32@lo+4
; GFX1064-NEXT: s_addc_u32 s1, s1, __atomic_compare_exchange at gotpcrel32@hi+12
-; GFX1064-NEXT: buffer_store_dword v5, off, s[64:67], 0 offset:4
-; GFX1064-NEXT: buffer_store_dword v4, off, s[64:67], 0
+; GFX1064-NEXT: buffer_store_dword v2, off, s[64:67], 0 offset:4
+; GFX1064-NEXT: buffer_store_dword v1, off, s[64:67], 0
; GFX1064-NEXT: s_load_dwordx2 s[16:17], s[0:1], 0x0
; GFX1064-NEXT: v_mov_b32_e32 v31, v40
+; GFX1064-NEXT: v_mov_b32_e32 v0, 8
+; GFX1064-NEXT: v_mov_b32_e32 v1, 0
; GFX1064-NEXT: v_mov_b32_e32 v2, s52
-; GFX1064-NEXT: v_mov_b32_e32 v3, s53
-; GFX1064-NEXT: v_mov_b32_e32 v4, 0
; GFX1064-NEXT: v_mov_b32_e32 v5, 8
; GFX1064-NEXT: v_mov_b32_e32 v6, 0
; GFX1064-NEXT: v_mov_b32_e32 v7, 0
@@ -4331,16 +4331,16 @@ define amdgpu_kernel void @global_atomic_fmin_double_uni_address_div_value_agent
; GFX1064-NEXT: s_mov_b32 s13, s50
; GFX1064-NEXT: s_mov_b32 s14, s33
; GFX1064-NEXT: s_mov_b64 s[2:3], s[66:67]
-; GFX1064-NEXT: v_min_f64 v[0:1], v[0:1], v[41:42]
-; GFX1064-NEXT: buffer_store_dword v1, off, s[64:67], 0 offset:12
-; GFX1064-NEXT: buffer_store_dword v0, off, s[64:67], 0 offset:8
-; GFX1064-NEXT: v_mov_b32_e32 v0, 8
-; GFX1064-NEXT: v_mov_b32_e32 v1, 0
+; GFX1064-NEXT: v_min_f64 v[3:4], v[3:4], v[41:42]
+; GFX1064-NEXT: buffer_store_dword v4, off, s[64:67], 0 offset:12
+; GFX1064-NEXT: buffer_store_dword v3, off, s[64:67], 0 offset:8
+; GFX1064-NEXT: v_mov_b32_e32 v3, s53
+; GFX1064-NEXT: v_mov_b32_e32 v4, 0
; GFX1064-NEXT: s_waitcnt lgkmcnt(0)
; GFX1064-NEXT: s_swappc_b64 s[30:31], s[16:17]
; GFX1064-NEXT: s_clause 0x1
-; GFX1064-NEXT: buffer_load_dword v4, off, s[64:67], 0
-; GFX1064-NEXT: buffer_load_dword v5, off, s[64:67], 0 offset:4
+; GFX1064-NEXT: buffer_load_dword v1, off, s[64:67], 0
+; GFX1064-NEXT: buffer_load_dword v2, off, s[64:67], 0 offset:4
; GFX1064-NEXT: v_and_b32_e32 v0, 1, v0
; GFX1064-NEXT: v_cmp_eq_u32_e32 vcc, 1, v0
; GFX1064-NEXT: s_or_b64 s[54:55], vcc, s[54:55]
@@ -5424,24 +5424,24 @@ define amdgpu_kernel void @global_atomic_fmin_double_uni_address_uni_value_one_a
; GFX7LESS-NEXT: s_mov_b64 s[4:5], 0
; GFX7LESS-NEXT: s_mov_b32 s3, 0xf000
; GFX7LESS-NEXT: s_waitcnt lgkmcnt(0)
-; GFX7LESS-NEXT: v_mov_b32_e32 v2, s6
-; GFX7LESS-NEXT: v_mov_b32_e32 v3, s7
+; GFX7LESS-NEXT: v_mov_b32_e32 v0, s6
+; GFX7LESS-NEXT: v_mov_b32_e32 v1, s7
; GFX7LESS-NEXT: s_mov_b32 s2, -1
; GFX7LESS-NEXT: .LBB8_2: ; %atomicrmw.start
; GFX7LESS-NEXT: ; =>This Inner Loop Header: Depth=1
-; GFX7LESS-NEXT: v_max_f64 v[0:1], v[2:3], v[2:3]
-; GFX7LESS-NEXT: v_min_f64 v[0:1], v[0:1], 4.0
-; GFX7LESS-NEXT: s_waitcnt expcnt(0)
-; GFX7LESS-NEXT: v_mov_b32_e32 v7, v3
-; GFX7LESS-NEXT: v_mov_b32_e32 v6, v2
; GFX7LESS-NEXT: v_mov_b32_e32 v5, v1
; GFX7LESS-NEXT: v_mov_b32_e32 v4, v0
-; GFX7LESS-NEXT: buffer_atomic_cmpswap_x2 v[4:7], off, s[0:3], 0 glc
-; GFX7LESS-NEXT: s_waitcnt vmcnt(0)
-; GFX7LESS-NEXT: v_cmp_eq_u64_e32 vcc, v[4:5], v[2:3]
-; GFX7LESS-NEXT: s_or_b64 s[4:5], vcc, s[4:5]
+; GFX7LESS-NEXT: s_waitcnt expcnt(0)
+; GFX7LESS-NEXT: v_max_f64 v[0:1], v[4:5], v[4:5]
+; GFX7LESS-NEXT: v_min_f64 v[2:3], v[0:1], 4.0
+; GFX7LESS-NEXT: v_mov_b32_e32 v0, v2
+; GFX7LESS-NEXT: v_mov_b32_e32 v1, v3
; GFX7LESS-NEXT: v_mov_b32_e32 v2, v4
; GFX7LESS-NEXT: v_mov_b32_e32 v3, v5
+; GFX7LESS-NEXT: buffer_atomic_cmpswap_x2 v[0:3], off, s[0:3], 0 glc
+; GFX7LESS-NEXT: s_waitcnt vmcnt(0)
+; GFX7LESS-NEXT: v_cmp_eq_u64_e32 vcc, v[0:1], v[4:5]
+; GFX7LESS-NEXT: s_or_b64 s[4:5], vcc, s[4:5]
; GFX7LESS-NEXT: s_andn2_b64 exec, exec, s[4:5]
; GFX7LESS-NEXT: s_cbranch_execnz .LBB8_2
; GFX7LESS-NEXT: .LBB8_3:
@@ -5590,24 +5590,24 @@ define amdgpu_kernel void @global_atomic_fmin_double_uni_address_uni_value_one_a
; GFX7LESS-DPP-NEXT: s_mov_b64 s[4:5], 0
; GFX7LESS-DPP-NEXT: s_mov_b32 s3, 0xf000
; GFX7LESS-DPP-NEXT: s_waitcnt lgkmcnt(0)
-; GFX7LESS-DPP-NEXT: v_mov_b32_e32 v2, s6
-; GFX7LESS-DPP-NEXT: v_mov_b32_e32 v3, s7
+; GFX7LESS-DPP-NEXT: v_mov_b32_e32 v0, s6
+; GFX7LESS-DPP-NEXT: v_mov_b32_e32 v1, s7
; GFX7LESS-DPP-NEXT: s_mov_b32 s2, -1
; GFX7LESS-DPP-NEXT: .LBB8_2: ; %atomicrmw.start
; GFX7LESS-DPP-NEXT: ; =>This Inner Loop Header: Depth=1
-; GFX7LESS-DPP-NEXT: v_max_f64 v[0:1], v[2:3], v[2:3]
-; GFX7LESS-DPP-NEXT: v_min_f64 v[0:1], v[0:1], 4.0
-; GFX7LESS-DPP-NEXT: s_waitcnt expcnt(0)
-; GFX7LESS-DPP-NEXT: v_mov_b32_e32 v7, v3
-; GFX7LESS-DPP-NEXT: v_mov_b32_e32 v6, v2
; GFX7LESS-DPP-NEXT: v_mov_b32_e32 v5, v1
; GFX7LESS-DPP-NEXT: v_mov_b32_e32 v4, v0
-; GFX7LESS-DPP-NEXT: buffer_atomic_cmpswap_x2 v[4:7], off, s[0:3], 0 glc
-; GFX7LESS-DPP-NEXT: s_waitcnt vmcnt(0)
-; GFX7LESS-DPP-NEXT: v_cmp_eq_u64_e32 vcc, v[4:5], v[2:3]
-; GFX7LESS-DPP-NEXT: s_or_b64 s[4:5], vcc, s[4:5]
+; GFX7LESS-DPP-NEXT: s_waitcnt expcnt(0)
+; GFX7LESS-DPP-NEXT: v_max_f64 v[0:1], v[4:5], v[4:5]
+; GFX7LESS-DPP-NEXT: v_min_f64 v[2:3], v[0:1], 4.0
+; GFX7LESS-DPP-NEXT: v_mov_b32_e32 v0, v2
+; GFX7LESS-DPP-NEXT: v_mov_b32_e32 v1, v3
; GFX7LESS-DPP-NEXT: v_mov_b32_e32 v2, v4
; GFX7LESS-DPP-NEXT: v_mov_b32_e32 v3, v5
+; GFX7LESS-DPP-NEXT: buffer_atomic_cmpswap_x2 v[0:3], off, s[0:3], 0 glc
+; GFX7LESS-DPP-NEXT: s_waitcnt vmcnt(0)
+; GFX7LESS-DPP-NEXT: v_cmp_eq_u64_e32 vcc, v[0:1], v[4:5]
+; GFX7LESS-DPP-NEXT: s_or_b64 s[4:5], vcc, s[4:5]
; GFX7LESS-DPP-NEXT: s_andn2_b64 exec, exec, s[4:5]
; GFX7LESS-DPP-NEXT: s_cbranch_execnz .LBB8_2
; GFX7LESS-DPP-NEXT: .LBB8_3:
@@ -5777,12 +5777,12 @@ define amdgpu_kernel void @global_atomic_fmin_double_uni_address_div_value_one_a
; GFX7LESS-NEXT: s_waitcnt lgkmcnt(0)
; GFX7LESS-NEXT: s_swappc_b64 s[30:31], s[16:17]
; GFX7LESS-NEXT: s_mov_b64 s[0:1], exec
-; GFX7LESS-NEXT: v_mov_b32_e32 v4, 0
-; GFX7LESS-NEXT: v_mov_b32_e32 v5, 0x7ff80000
+; GFX7LESS-NEXT: v_mov_b32_e32 v2, 0
+; GFX7LESS-NEXT: v_mov_b32_e32 v3, 0x7ff80000
; GFX7LESS-NEXT: .LBB9_1: ; %ComputeLoop
; GFX7LESS-NEXT: ; =>This Inner Loop Header: Depth=1
; GFX7LESS-NEXT: s_ff1_i32_b64 s4, s[0:1]
-; GFX7LESS-NEXT: v_max_f64 v[2:3], v[4:5], v[4:5]
+; GFX7LESS-NEXT: v_max_f64 v[2:3], v[2:3], v[2:3]
; GFX7LESS-NEXT: v_readlane_b32 s3, v1, s4
; GFX7LESS-NEXT: v_readlane_b32 s2, v0, s4
; GFX7LESS-NEXT: s_lshl_b64 s[4:5], 1, s4
@@ -5790,7 +5790,7 @@ define amdgpu_kernel void @global_atomic_fmin_double_uni_address_div_value_one_a
; GFX7LESS-NEXT: s_andn2_b64 s[0:1], s[0:1], s[4:5]
; GFX7LESS-NEXT: v_cmp_ne_u64_e64 s[2:3], s[0:1], 0
; GFX7LESS-NEXT: s_and_b64 vcc, exec, s[2:3]
-; GFX7LESS-NEXT: v_min_f64 v[4:5], v[2:3], v[4:5]
+; GFX7LESS-NEXT: v_min_f64 v[2:3], v[2:3], v[4:5]
; GFX7LESS-NEXT: s_cbranch_vccnz .LBB9_1
; GFX7LESS-NEXT: ; %bb.2: ; %ComputeEnd
; GFX7LESS-NEXT: v_mbcnt_lo_u32_b32_e64 v0, exec_lo, 0
@@ -5804,25 +5804,25 @@ define amdgpu_kernel void @global_atomic_fmin_double_uni_address_div_value_one_a
; GFX7LESS-NEXT: s_mov_b32 s3, 0xf000
; GFX7LESS-NEXT: s_mov_b32 s2, -1
; GFX7LESS-NEXT: s_waitcnt lgkmcnt(0)
-; GFX7LESS-NEXT: buffer_load_dwordx2 v[2:3], off, s[0:3], 0
+; GFX7LESS-NEXT: buffer_load_dwordx2 v[0:1], off, s[0:3], 0
; GFX7LESS-NEXT: s_mov_b64 s[4:5], 0
-; GFX7LESS-NEXT: v_max_f64 v[4:5], v[4:5], v[4:5]
+; GFX7LESS-NEXT: v_max_f64 v[4:5], v[2:3], v[2:3]
; GFX7LESS-NEXT: .LBB9_4: ; %atomicrmw.start
; GFX7LESS-NEXT: ; =>This Inner Loop Header: Depth=1
; GFX7LESS-NEXT: s_waitcnt vmcnt(0)
-; GFX7LESS-NEXT: v_max_f64 v[0:1], v[2:3], v[2:3]
-; GFX7LESS-NEXT: v_min_f64 v[0:1], v[0:1], v[4:5]
+; GFX7LESS-NEXT: v_mov_b32_e32 v9, v1
+; GFX7LESS-NEXT: v_mov_b32_e32 v8, v0
; GFX7LESS-NEXT: s_waitcnt expcnt(0)
-; GFX7LESS-NEXT: v_mov_b32_e32 v9, v3
-; GFX7LESS-NEXT: v_mov_b32_e32 v8, v2
-; GFX7LESS-NEXT: v_mov_b32_e32 v7, v1
-; GFX7LESS-NEXT: v_mov_b32_e32 v6, v0
-; GFX7LESS-NEXT: buffer_atomic_cmpswap_x2 v[6:9], off, s[0:3], 0 glc
+; GFX7LESS-NEXT: v_max_f64 v[0:1], v[8:9], v[8:9]
+; GFX7LESS-NEXT: v_min_f64 v[6:7], v[0:1], v[4:5]
+; GFX7LESS-NEXT: v_mov_b32_e32 v0, v6
+; GFX7LESS-NEXT: v_mov_b32_e32 v1, v7
+; GFX7LESS-NEXT: v_mov_b32_e32 v2, v8
+; GFX7LESS-NEXT: v_mov_b32_e32 v3, v9
+; GFX7LESS-NEXT: buffer_atomic_cmpswap_x2 v[0:3], off, s[0:3], 0 glc
; GFX7LESS-NEXT: s_waitcnt vmcnt(0)
-; GFX7LESS-NEXT: v_cmp_eq_u64_e32 vcc, v[6:7], v[2:3]
+; GFX7LESS-NEXT: v_cmp_eq_u64_e32 vcc, v[0:1], v[8:9]
; GFX7LESS-NEXT: s_or_b64 s[4:5], vcc, s[4:5]
-; GFX7LESS-NEXT: v_mov_b32_e32 v2, v6
-; GFX7LESS-NEXT: v_mov_b32_e32 v3, v7
; GFX7LESS-NEXT: s_andn2_b64 exec, exec, s[4:5]
; GFX7LESS-NEXT: s_cbranch_execnz .LBB9_4
; GFX7LESS-NEXT: .LBB9_5:
@@ -6191,23 +6191,23 @@ define amdgpu_kernel void @global_atomic_fmin_double_uni_address_div_value_one_a
; GFX7LESS-DPP-NEXT: s_swappc_b64 s[30:31], s[16:17]
; GFX7LESS-DPP-NEXT: buffer_load_dwordx2 v[2:3], off, s[36:39], 0
; GFX7LESS-DPP-NEXT: s_mov_b64 s[0:1], 0
-; GFX7LESS-DPP-NEXT: v_max_f64 v[4:5], v[0:1], v[0:1]
+; GFX7LESS-DPP-NEXT: v_max_f64 v[0:1], v[0:1], v[0:1]
; GFX7LESS-DPP-NEXT: .LBB9_1: ; %atomicrmw.start
; GFX7LESS-DPP-NEXT: ; =>This Inner Loop Header: Depth=1
; GFX7LESS-DPP-NEXT: s_waitcnt vmcnt(0)
-; GFX7LESS-DPP-NEXT: v_max_f64 v[0:1], v[2:3], v[2:3]
-; GFX7LESS-DPP-NEXT: v_min_f64 v[0:1], v[0:1], v[4:5]
+; GFX7LESS-DPP-NEXT: v_mov_b32_e32 v7, v3
+; GFX7LESS-DPP-NEXT: v_mov_b32_e32 v6, v2
; GFX7LESS-DPP-NEXT: s_waitcnt expcnt(0)
-; GFX7LESS-DPP-NEXT: v_mov_b32_e32 v9, v3
-; GFX7LESS-DPP-NEXT: v_mov_b32_e32 v8, v2
-; GFX7LESS-DPP-NEXT: v_mov_b32_e32 v7, v1
-; GFX7LESS-DPP-NEXT: v_mov_b32_e32 v6, v0
-; GFX7LESS-DPP-NEXT: buffer_atomic_cmpswap_x2 v[6:9], off, s[36:39], 0 glc
+; GFX7LESS-DPP-NEXT: v_max_f64 v[2:3], v[6:7], v[6:7]
+; GFX7LESS-DPP-NEXT: v_min_f64 v[4:5], v[2:3], v[0:1]
+; GFX7LESS-DPP-NEXT: v_mov_b32_e32 v2, v4
+; GFX7LESS-DPP-NEXT: v_mov_b32_e32 v3, v5
+; GFX7LESS-DPP-NEXT: v_mov_b32_e32 v4, v6
+; GFX7LESS-DPP-NEXT: v_mov_b32_e32 v5, v7
+; GFX7LESS-DPP-NEXT: buffer_atomic_cmpswap_x2 v[2:5], off, s[36:39], 0 glc
; GFX7LESS-DPP-NEXT: s_waitcnt vmcnt(0)
-; GFX7LESS-DPP-NEXT: v_cmp_eq_u64_e32 vcc, v[6:7], v[2:3]
+; GFX7LESS-DPP-NEXT: v_cmp_eq_u64_e32 vcc, v[2:3], v[6:7]
; GFX7LESS-DPP-NEXT: s_or_b64 s[0:1], vcc, s[0:1]
-; GFX7LESS-DPP-NEXT: v_mov_b32_e32 v2, v6
-; GFX7LESS-DPP-NEXT: v_mov_b32_e32 v3, v7
; GFX7LESS-DPP-NEXT: s_andn2_b64 exec, exec, s[0:1]
; GFX7LESS-DPP-NEXT: s_cbranch_execnz .LBB9_1
; GFX7LESS-DPP-NEXT: ; %bb.2: ; %atomicrmw.end
@@ -7821,20 +7821,20 @@ define amdgpu_kernel void @global_atomic_fmin_double_uni_address_div_value_defau
; GFX1064-NEXT: s_movk_i32 s32, 0x800
; GFX1064-NEXT: s_waitcnt lgkmcnt(0)
; GFX1064-NEXT: s_swappc_b64 s[30:31], s[16:17]
-; GFX1064-NEXT: v_mov_b32_e32 v2, 0
-; GFX1064-NEXT: v_mov_b32_e32 v3, 0x7ff80000
+; GFX1064-NEXT: v_mov_b32_e32 v3, 0
+; GFX1064-NEXT: v_mov_b32_e32 v4, 0x7ff80000
; GFX1064-NEXT: s_mov_b64 s[0:1], exec
; GFX1064-NEXT: .LBB11_1: ; %ComputeLoop
; GFX1064-NEXT: ; =>This Inner Loop Header: Depth=1
; GFX1064-NEXT: s_ff1_i32_b64 s4, s[0:1]
-; GFX1064-NEXT: v_max_f64 v[2:3], v[2:3], v[2:3]
+; GFX1064-NEXT: v_max_f64 v[2:3], v[3:4], v[3:4]
; GFX1064-NEXT: v_readlane_b32 s3, v1, s4
; GFX1064-NEXT: v_readlane_b32 s2, v0, s4
; GFX1064-NEXT: v_max_f64 v[4:5], s[2:3], s[2:3]
; GFX1064-NEXT: s_lshl_b64 s[2:3], 1, s4
; GFX1064-NEXT: s_andn2_b64 s[0:1], s[0:1], s[2:3]
; GFX1064-NEXT: s_cmp_lg_u64 s[0:1], 0
-; GFX1064-NEXT: v_min_f64 v[2:3], v[2:3], v[4:5]
+; GFX1064-NEXT: v_min_f64 v[3:4], v[2:3], v[4:5]
; GFX1064-NEXT: s_cbranch_scc1 .LBB11_1
; GFX1064-NEXT: ; %bb.2: ; %ComputeEnd
; GFX1064-NEXT: v_mbcnt_lo_u32_b32 v0, exec_lo, 0
@@ -7846,26 +7846,26 @@ define amdgpu_kernel void @global_atomic_fmin_double_uni_address_div_value_defau
; GFX1064-NEXT: ; %bb.3:
; GFX1064-NEXT: s_load_dwordx2 s[52:53], s[34:35], 0x24
; GFX1064-NEXT: v_mov_b32_e32 v0, 0
-; GFX1064-NEXT: v_max_f64 v[41:42], v[2:3], v[2:3]
+; GFX1064-NEXT: v_max_f64 v[41:42], v[3:4], v[3:4]
; GFX1064-NEXT: s_mov_b64 s[54:55], 0
; GFX1064-NEXT: s_waitcnt lgkmcnt(0)
-; GFX1064-NEXT: global_load_dwordx2 v[4:5], v0, s[52:53]
+; GFX1064-NEXT: global_load_dwordx2 v[1:2], v0, s[52:53]
; GFX1064-NEXT: .LBB11_4: ; %atomicrmw.start
; GFX1064-NEXT: ; =>This Inner Loop Header: Depth=1
; GFX1064-NEXT: s_waitcnt vmcnt(0)
-; GFX1064-NEXT: v_max_f64 v[0:1], v[4:5], v[4:5]
+; GFX1064-NEXT: v_max_f64 v[3:4], v[1:2], v[1:2]
; GFX1064-NEXT: s_add_u32 s8, s34, 44
; GFX1064-NEXT: s_addc_u32 s9, s35, 0
; GFX1064-NEXT: s_getpc_b64 s[0:1]
; GFX1064-NEXT: s_add_u32 s0, s0, __atomic_compare_exchange at gotpcrel32@lo+4
; GFX1064-NEXT: s_addc_u32 s1, s1, __atomic_compare_exchange at gotpcrel32@hi+12
-; GFX1064-NEXT: buffer_store_dword v5, off, s[64:67], 0 offset:4
-; GFX1064-NEXT: buffer_store_dword v4, off, s[64:67], 0
+; GFX1064-NEXT: buffer_store_dword v2, off, s[64:67], 0 offset:4
+; GFX1064-NEXT: buffer_store_dword v1, off, s[64:67], 0
; GFX1064-NEXT: s_load_dwordx2 s[16:17], s[0:1], 0x0
; GFX1064-NEXT: v_mov_b32_e32 v31, v40
+; GFX1064-NEXT: v_mov_b32_e32 v0, 8
+; GFX1064-NEXT: v_mov_b32_e32 v1, 0
; GFX1064-NEXT: v_mov_b32_e32 v2, s52
-; GFX1064-NEXT: v_mov_b32_e32 v3, s53
-; GFX1064-NEXT: v_mov_b32_e32 v4, 0
; GFX1064-NEXT: v_mov_b32_e32 v5, 8
; GFX1064-NEXT: v_mov_b32_e32 v6, 0
; GFX1064-NEXT: v_mov_b32_e32 v7, 0
@@ -7877,16 +7877,16 @@ define amdgpu_kernel void @global_atomic_fmin_double_uni_address_div_value_defau
; GFX1064-NEXT: s_mov_b32 s13, s50
; GFX1064-NEXT: s_mov_b32 s14, s33
; GFX1064-NEXT: s_mov_b64 s[2:3], s[66:67]
-; GFX1064-NEXT: v_min_f64 v[0:1], v[0:1], v[41:42]
-; GFX1064-NEXT: buffer_store_dword v1, off, s[64:67], 0 offset:12
-; GFX1064-NEXT: buffer_store_dword v0, off, s[64:67], 0 offset:8
-; GFX1064-NEXT: v_mov_b32_e32 v0, 8
-; GFX1064-NEXT: v_mov_b32_e32 v1, 0
+; GFX1064-NEXT: v_min_f64 v[3:4], v[3:4], v[41:42]
+; GFX1064-NEXT: buffer_store_dword v4, off, s[64:67], 0 offset:12
+; GFX1064-NEXT: buffer_store_dword v3, off, s[64:67], 0 offset:8
+; GFX1064-NEXT: v_mov_b32_e32 v3, s53
+; GFX1064-NEXT: v_mov_b32_e32 v4, 0
; GFX1064-NEXT: s_waitcnt lgkmcnt(0)
; GFX1064-NEXT: s_swappc_b64 s[30:31], s[16:17]
; GFX1064-NEXT: s_clause 0x1
-; GFX1064-NEXT: buffer_load_dword v4, off, s[64:67], 0
-; GFX1064-NEXT: buffer_load_dword v5, off, s[64:67], 0 offset:4
+; GFX1064-NEXT: buffer_load_dword v1, off, s[64:67], 0
+; GFX1064-NEXT: buffer_load_dword v2, off, s[64:67], 0 offset:4
; GFX1064-NEXT: v_and_b32_e32 v0, 1, v0
; GFX1064-NEXT: v_cmp_eq_u32_e32 vcc, 1, v0
; GFX1064-NEXT: s_or_b64 s[54:55], vcc, s[54:55]
@@ -8970,20 +8970,20 @@ define amdgpu_kernel void @global_atomic_fmin_uni_address_uni_value_system_scope
; GFX7LESS-NEXT: s_mov_b64 s[4:5], 0
; GFX7LESS-NEXT: s_mov_b32 s3, 0xf000
; GFX7LESS-NEXT: s_waitcnt lgkmcnt(0)
-; GFX7LESS-NEXT: v_mov_b32_e32 v1, s2
+; GFX7LESS-NEXT: v_mov_b32_e32 v0, s2
; GFX7LESS-NEXT: s_mov_b32 s2, -1
; GFX7LESS-NEXT: .LBB12_2: ; %atomicrmw.start
; GFX7LESS-NEXT: ; =>This Inner Loop Header: Depth=1
-; GFX7LESS-NEXT: v_mul_f32_e32 v0, 1.0, v1
-; GFX7LESS-NEXT: v_min_f32_e32 v0, 4.0, v0
-; GFX7LESS-NEXT: s_waitcnt expcnt(0)
-; GFX7LESS-NEXT: v_mov_b32_e32 v3, v1
; GFX7LESS-NEXT: v_mov_b32_e32 v2, v0
-; GFX7LESS-NEXT: buffer_atomic_cmpswap v[2:3], off, s[0:3], 0 glc
+; GFX7LESS-NEXT: s_waitcnt expcnt(0)
+; GFX7LESS-NEXT: v_mul_f32_e32 v0, 1.0, v2
+; GFX7LESS-NEXT: v_min_f32_e32 v1, 4.0, v0
+; GFX7LESS-NEXT: v_mov_b32_e32 v0, v1
+; GFX7LESS-NEXT: v_mov_b32_e32 v1, v2
+; GFX7LESS-NEXT: buffer_atomic_cmpswap v[0:1], off, s[0:3], 0 glc
; GFX7LESS-NEXT: s_waitcnt vmcnt(0)
-; GFX7LESS-NEXT: v_cmp_eq_u32_e32 vcc, v2, v1
+; GFX7LESS-NEXT: v_cmp_eq_u32_e32 vcc, v0, v2
; GFX7LESS-NEXT: s_or_b64 s[4:5], vcc, s[4:5]
-; GFX7LESS-NEXT: v_mov_b32_e32 v1, v2
; GFX7LESS-NEXT: s_andn2_b64 exec, exec, s[4:5]
; GFX7LESS-NEXT: s_cbranch_execnz .LBB12_2
; GFX7LESS-NEXT: .LBB12_3:
@@ -9095,20 +9095,20 @@ define amdgpu_kernel void @global_atomic_fmin_uni_address_uni_value_system_scope
; GFX7LESS-DPP-NEXT: s_mov_b64 s[4:5], 0
; GFX7LESS-DPP-NEXT: s_mov_b32 s3, 0xf000
; GFX7LESS-DPP-NEXT: s_waitcnt lgkmcnt(0)
-; GFX7LESS-DPP-NEXT: v_mov_b32_e32 v1, s2
+; GFX7LESS-DPP-NEXT: v_mov_b32_e32 v0, s2
; GFX7LESS-DPP-NEXT: s_mov_b32 s2, -1
; GFX7LESS-DPP-NEXT: .LBB12_2: ; %atomicrmw.start
; GFX7LESS-DPP-NEXT: ; =>This Inner Loop Header: Depth=1
-; GFX7LESS-DPP-NEXT: v_mul_f32_e32 v0, 1.0, v1
-; GFX7LESS-DPP-NEXT: v_min_f32_e32 v0, 4.0, v0
-; GFX7LESS-DPP-NEXT: s_waitcnt expcnt(0)
-; GFX7LESS-DPP-NEXT: v_mov_b32_e32 v3, v1
; GFX7LESS-DPP-NEXT: v_mov_b32_e32 v2, v0
-; GFX7LESS-DPP-NEXT: buffer_atomic_cmpswap v[2:3], off, s[0:3], 0 glc
+; GFX7LESS-DPP-NEXT: s_waitcnt expcnt(0)
+; GFX7LESS-DPP-NEXT: v_mul_f32_e32 v0, 1.0, v2
+; GFX7LESS-DPP-NEXT: v_min_f32_e32 v1, 4.0, v0
+; GFX7LESS-DPP-NEXT: v_mov_b32_e32 v0, v1
+; GFX7LESS-DPP-NEXT: v_mov_b32_e32 v1, v2
+; GFX7LESS-DPP-NEXT: buffer_atomic_cmpswap v[0:1], off, s[0:3], 0 glc
; GFX7LESS-DPP-NEXT: s_waitcnt vmcnt(0)
-; GFX7LESS-DPP-NEXT: v_cmp_eq_u32_e32 vcc, v2, v1
+; GFX7LESS-DPP-NEXT: v_cmp_eq_u32_e32 vcc, v0, v2
; GFX7LESS-DPP-NEXT: s_or_b64 s[4:5], vcc, s[4:5]
-; GFX7LESS-DPP-NEXT: v_mov_b32_e32 v1, v2
; GFX7LESS-DPP-NEXT: s_andn2_b64 exec, exec, s[4:5]
; GFX7LESS-DPP-NEXT: s_cbranch_execnz .LBB12_2
; GFX7LESS-DPP-NEXT: .LBB12_3:
@@ -9224,20 +9224,20 @@ define amdgpu_kernel void @global_atomic_fmin_uni_address_uni_value_system_scope
; GFX7LESS-NEXT: s_mov_b64 s[4:5], 0
; GFX7LESS-NEXT: s_mov_b32 s3, 0xf000
; GFX7LESS-NEXT: s_waitcnt lgkmcnt(0)
-; GFX7LESS-NEXT: v_mov_b32_e32 v1, s2
+; GFX7LESS-NEXT: v_mov_b32_e32 v0, s2
; GFX7LESS-NEXT: s_mov_b32 s2, -1
; GFX7LESS-NEXT: .LBB13_2: ; %atomicrmw.start
; GFX7LESS-NEXT: ; =>This Inner Loop Header: Depth=1
-; GFX7LESS-NEXT: v_mul_f32_e32 v0, 1.0, v1
-; GFX7LESS-NEXT: v_min_f32_e32 v0, 4.0, v0
-; GFX7LESS-NEXT: s_waitcnt expcnt(0)
-; GFX7LESS-NEXT: v_mov_b32_e32 v3, v1
; GFX7LESS-NEXT: v_mov_b32_e32 v2, v0
-; GFX7LESS-NEXT: buffer_atomic_cmpswap v[2:3], off, s[0:3], 0 glc
+; GFX7LESS-NEXT: s_waitcnt expcnt(0)
+; GFX7LESS-NEXT: v_mul_f32_e32 v0, 1.0, v2
+; GFX7LESS-NEXT: v_min_f32_e32 v1, 4.0, v0
+; GFX7LESS-NEXT: v_mov_b32_e32 v0, v1
+; GFX7LESS-NEXT: v_mov_b32_e32 v1, v2
+; GFX7LESS-NEXT: buffer_atomic_cmpswap v[0:1], off, s[0:3], 0 glc
; GFX7LESS-NEXT: s_waitcnt vmcnt(0)
-; GFX7LESS-NEXT: v_cmp_eq_u32_e32 vcc, v2, v1
+; GFX7LESS-NEXT: v_cmp_eq_u32_e32 vcc, v0, v2
; GFX7LESS-NEXT: s_or_b64 s[4:5], vcc, s[4:5]
-; GFX7LESS-NEXT: v_mov_b32_e32 v1, v2
; GFX7LESS-NEXT: s_andn2_b64 exec, exec, s[4:5]
; GFX7LESS-NEXT: s_cbranch_execnz .LBB13_2
; GFX7LESS-NEXT: .LBB13_3:
@@ -9349,20 +9349,20 @@ define amdgpu_kernel void @global_atomic_fmin_uni_address_uni_value_system_scope
; GFX7LESS-DPP-NEXT: s_mov_b64 s[4:5], 0
; GFX7LESS-DPP-NEXT: s_mov_b32 s3, 0xf000
; GFX7LESS-DPP-NEXT: s_waitcnt lgkmcnt(0)
-; GFX7LESS-DPP-NEXT: v_mov_b32_e32 v1, s2
+; GFX7LESS-DPP-NEXT: v_mov_b32_e32 v0, s2
; GFX7LESS-DPP-NEXT: s_mov_b32 s2, -1
; GFX7LESS-DPP-NEXT: .LBB13_2: ; %atomicrmw.start
; GFX7LESS-DPP-NEXT: ; =>This Inner Loop Header: Depth=1
-; GFX7LESS-DPP-NEXT: v_mul_f32_e32 v0, 1.0, v1
-; GFX7LESS-DPP-NEXT: v_min_f32_e32 v0, 4.0, v0
-; GFX7LESS-DPP-NEXT: s_waitcnt expcnt(0)
-; GFX7LESS-DPP-NEXT: v_mov_b32_e32 v3, v1
; GFX7LESS-DPP-NEXT: v_mov_b32_e32 v2, v0
-; GFX7LESS-DPP-NEXT: buffer_atomic_cmpswap v[2:3], off, s[0:3], 0 glc
+; GFX7LESS-DPP-NEXT: s_waitcnt expcnt(0)
+; GFX7LESS-DPP-NEXT: v_mul_f32_e32 v0, 1.0, v2
+; GFX7LESS-DPP-NEXT: v_min_f32_e32 v1, 4.0, v0
+; GFX7LESS-DPP-NEXT: v_mov_b32_e32 v0, v1
+; GFX7LESS-DPP-NEXT: v_mov_b32_e32 v1, v2
+; GFX7LESS-DPP-NEXT: buffer_atomic_cmpswap v[0:1], off, s[0:3], 0 glc
; GFX7LESS-DPP-NEXT: s_waitcnt vmcnt(0)
-; GFX7LESS-DPP-NEXT: v_cmp_eq_u32_e32 vcc, v2, v1
+; GFX7LESS-DPP-NEXT: v_cmp_eq_u32_e32 vcc, v0, v2
; GFX7LESS-DPP-NEXT: s_or_b64 s[4:5], vcc, s[4:5]
-; GFX7LESS-DPP-NEXT: v_mov_b32_e32 v1, v2
; GFX7LESS-DPP-NEXT: s_andn2_b64 exec, exec, s[4:5]
; GFX7LESS-DPP-NEXT: s_cbranch_execnz .LBB13_2
; GFX7LESS-DPP-NEXT: .LBB13_3:
diff --git a/llvm/test/CodeGen/AMDGPU/global_atomics_scan_fsub.ll b/llvm/test/CodeGen/AMDGPU/global_atomics_scan_fsub.ll
index 9db3c37045ccf..b6fa8d7e32199 100644
--- a/llvm/test/CodeGen/AMDGPU/global_atomics_scan_fsub.ll
+++ b/llvm/test/CodeGen/AMDGPU/global_atomics_scan_fsub.ll
@@ -34,19 +34,19 @@ define amdgpu_kernel void @global_atomic_fsub_uni_address_uni_value_agent_scope_
; GFX7LESS-NEXT: v_cvt_f32_ubyte0_e32 v0, s2
; GFX7LESS-NEXT: v_mul_f32_e32 v2, 4.0, v0
; GFX7LESS-NEXT: s_waitcnt lgkmcnt(0)
-; GFX7LESS-NEXT: v_mov_b32_e32 v1, s6
+; GFX7LESS-NEXT: v_mov_b32_e32 v0, s6
; GFX7LESS-NEXT: s_mov_b32 s2, -1
; GFX7LESS-NEXT: .LBB0_2: ; %atomicrmw.start
; GFX7LESS-NEXT: ; =>This Inner Loop Header: Depth=1
-; GFX7LESS-NEXT: v_sub_f32_e32 v0, v1, v2
+; GFX7LESS-NEXT: v_mov_b32_e32 v4, v0
+; GFX7LESS-NEXT: v_sub_f32_e32 v3, v4, v2
; GFX7LESS-NEXT: s_waitcnt expcnt(0)
-; GFX7LESS-NEXT: v_mov_b32_e32 v4, v1
-; GFX7LESS-NEXT: v_mov_b32_e32 v3, v0
-; GFX7LESS-NEXT: buffer_atomic_cmpswap v[3:4], off, s[0:3], 0 glc
+; GFX7LESS-NEXT: v_mov_b32_e32 v0, v3
+; GFX7LESS-NEXT: v_mov_b32_e32 v1, v4
+; GFX7LESS-NEXT: buffer_atomic_cmpswap v[0:1], off, s[0:3], 0 glc
; GFX7LESS-NEXT: s_waitcnt vmcnt(0)
-; GFX7LESS-NEXT: v_cmp_eq_u32_e32 vcc, v3, v1
+; GFX7LESS-NEXT: v_cmp_eq_u32_e32 vcc, v0, v4
; GFX7LESS-NEXT: s_or_b64 s[4:5], vcc, s[4:5]
-; GFX7LESS-NEXT: v_mov_b32_e32 v1, v3
; GFX7LESS-NEXT: s_andn2_b64 exec, exec, s[4:5]
; GFX7LESS-NEXT: s_cbranch_execnz .LBB0_2
; GFX7LESS-NEXT: .LBB0_3:
@@ -235,19 +235,19 @@ define amdgpu_kernel void @global_atomic_fsub_uni_address_uni_value_agent_scope_
; GFX7LESS-DPP-NEXT: v_cvt_f32_ubyte0_e32 v0, s2
; GFX7LESS-DPP-NEXT: v_mul_f32_e32 v2, 4.0, v0
; GFX7LESS-DPP-NEXT: s_waitcnt lgkmcnt(0)
-; GFX7LESS-DPP-NEXT: v_mov_b32_e32 v1, s6
+; GFX7LESS-DPP-NEXT: v_mov_b32_e32 v0, s6
; GFX7LESS-DPP-NEXT: s_mov_b32 s2, -1
; GFX7LESS-DPP-NEXT: .LBB0_2: ; %atomicrmw.start
; GFX7LESS-DPP-NEXT: ; =>This Inner Loop Header: Depth=1
-; GFX7LESS-DPP-NEXT: v_sub_f32_e32 v0, v1, v2
+; GFX7LESS-DPP-NEXT: v_mov_b32_e32 v4, v0
+; GFX7LESS-DPP-NEXT: v_sub_f32_e32 v3, v4, v2
; GFX7LESS-DPP-NEXT: s_waitcnt expcnt(0)
-; GFX7LESS-DPP-NEXT: v_mov_b32_e32 v4, v1
-; GFX7LESS-DPP-NEXT: v_mov_b32_e32 v3, v0
-; GFX7LESS-DPP-NEXT: buffer_atomic_cmpswap v[3:4], off, s[0:3], 0 glc
+; GFX7LESS-DPP-NEXT: v_mov_b32_e32 v0, v3
+; GFX7LESS-DPP-NEXT: v_mov_b32_e32 v1, v4
+; GFX7LESS-DPP-NEXT: buffer_atomic_cmpswap v[0:1], off, s[0:3], 0 glc
; GFX7LESS-DPP-NEXT: s_waitcnt vmcnt(0)
-; GFX7LESS-DPP-NEXT: v_cmp_eq_u32_e32 vcc, v3, v1
+; GFX7LESS-DPP-NEXT: v_cmp_eq_u32_e32 vcc, v0, v4
; GFX7LESS-DPP-NEXT: s_or_b64 s[4:5], vcc, s[4:5]
-; GFX7LESS-DPP-NEXT: v_mov_b32_e32 v1, v3
; GFX7LESS-DPP-NEXT: s_andn2_b64 exec, exec, s[4:5]
; GFX7LESS-DPP-NEXT: s_cbranch_execnz .LBB0_2
; GFX7LESS-DPP-NEXT: .LBB0_3:
@@ -476,20 +476,20 @@ define amdgpu_kernel void @global_atomic_fsub_uni_address_div_value_agent_scope_
; GFX7LESS-NEXT: s_mov_b32 s3, 0xf000
; GFX7LESS-NEXT: s_mov_b32 s2, -1
; GFX7LESS-NEXT: s_waitcnt lgkmcnt(0)
-; GFX7LESS-NEXT: buffer_load_dword v1, off, s[0:3], 0
+; GFX7LESS-NEXT: buffer_load_dword v0, off, s[0:3], 0
; GFX7LESS-NEXT: s_mov_b64 s[4:5], 0
; GFX7LESS-NEXT: .LBB1_4: ; %atomicrmw.start
; GFX7LESS-NEXT: ; =>This Inner Loop Header: Depth=1
; GFX7LESS-NEXT: s_waitcnt vmcnt(0)
-; GFX7LESS-NEXT: v_sub_f32_e32 v0, v1, v2
+; GFX7LESS-NEXT: v_mov_b32_e32 v4, v0
+; GFX7LESS-NEXT: v_sub_f32_e32 v3, v4, v2
; GFX7LESS-NEXT: s_waitcnt expcnt(0)
-; GFX7LESS-NEXT: v_mov_b32_e32 v4, v1
-; GFX7LESS-NEXT: v_mov_b32_e32 v3, v0
-; GFX7LESS-NEXT: buffer_atomic_cmpswap v[3:4], off, s[0:3], 0 glc
+; GFX7LESS-NEXT: v_mov_b32_e32 v0, v3
+; GFX7LESS-NEXT: v_mov_b32_e32 v1, v4
+; GFX7LESS-NEXT: buffer_atomic_cmpswap v[0:1], off, s[0:3], 0 glc
; GFX7LESS-NEXT: s_waitcnt vmcnt(0)
-; GFX7LESS-NEXT: v_cmp_eq_u32_e32 vcc, v3, v1
+; GFX7LESS-NEXT: v_cmp_eq_u32_e32 vcc, v0, v4
; GFX7LESS-NEXT: s_or_b64 s[4:5], vcc, s[4:5]
-; GFX7LESS-NEXT: v_mov_b32_e32 v1, v3
; GFX7LESS-NEXT: s_andn2_b64 exec, exec, s[4:5]
; GFX7LESS-NEXT: s_cbranch_execnz .LBB1_4
; GFX7LESS-NEXT: .LBB1_5:
@@ -848,20 +848,20 @@ define amdgpu_kernel void @global_atomic_fsub_uni_address_div_value_agent_scope_
; GFX7LESS-DPP-NEXT: s_mov_b64 s[2:3], s[50:51]
; GFX7LESS-DPP-NEXT: s_waitcnt lgkmcnt(0)
; GFX7LESS-DPP-NEXT: s_swappc_b64 s[30:31], s[16:17]
-; GFX7LESS-DPP-NEXT: buffer_load_dword v2, off, s[36:39], 0
+; GFX7LESS-DPP-NEXT: buffer_load_dword v1, off, s[36:39], 0
; GFX7LESS-DPP-NEXT: s_mov_b64 s[0:1], 0
; GFX7LESS-DPP-NEXT: .LBB1_1: ; %atomicrmw.start
; GFX7LESS-DPP-NEXT: ; =>This Inner Loop Header: Depth=1
; GFX7LESS-DPP-NEXT: s_waitcnt vmcnt(0)
-; GFX7LESS-DPP-NEXT: v_sub_f32_e32 v1, v2, v0
-; GFX7LESS-DPP-NEXT: s_waitcnt expcnt(0)
-; GFX7LESS-DPP-NEXT: v_mov_b32_e32 v4, v2
; GFX7LESS-DPP-NEXT: v_mov_b32_e32 v3, v1
-; GFX7LESS-DPP-NEXT: buffer_atomic_cmpswap v[3:4], off, s[36:39], 0 glc
+; GFX7LESS-DPP-NEXT: s_waitcnt expcnt(0)
+; GFX7LESS-DPP-NEXT: v_sub_f32_e32 v2, v3, v0
+; GFX7LESS-DPP-NEXT: v_mov_b32_e32 v1, v2
+; GFX7LESS-DPP-NEXT: v_mov_b32_e32 v2, v3
+; GFX7LESS-DPP-NEXT: buffer_atomic_cmpswap v[1:2], off, s[36:39], 0 glc
; GFX7LESS-DPP-NEXT: s_waitcnt vmcnt(0)
-; GFX7LESS-DPP-NEXT: v_cmp_eq_u32_e32 vcc, v3, v2
+; GFX7LESS-DPP-NEXT: v_cmp_eq_u32_e32 vcc, v1, v3
; GFX7LESS-DPP-NEXT: s_or_b64 s[0:1], vcc, s[0:1]
-; GFX7LESS-DPP-NEXT: v_mov_b32_e32 v2, v3
; GFX7LESS-DPP-NEXT: s_andn2_b64 exec, exec, s[0:1]
; GFX7LESS-DPP-NEXT: s_cbranch_execnz .LBB1_1
; GFX7LESS-DPP-NEXT: ; %bb.2: ; %atomicrmw.end
@@ -1275,19 +1275,19 @@ define amdgpu_kernel void @global_atomic_fsub_uni_address_uni_value_one_as_scope
; GFX7LESS-NEXT: v_cvt_f32_f64_e32 v0, v[0:1]
; GFX7LESS-NEXT: v_mul_f32_e32 v2, 4.0, v0
; GFX7LESS-NEXT: s_waitcnt lgkmcnt(0)
-; GFX7LESS-NEXT: v_mov_b32_e32 v1, s2
+; GFX7LESS-NEXT: v_mov_b32_e32 v0, s2
; GFX7LESS-NEXT: s_mov_b32 s2, -1
; GFX7LESS-NEXT: .LBB2_2: ; %atomicrmw.start
; GFX7LESS-NEXT: ; =>This Inner Loop Header: Depth=1
-; GFX7LESS-NEXT: v_sub_f32_e32 v0, v1, v2
+; GFX7LESS-NEXT: v_mov_b32_e32 v4, v0
+; GFX7LESS-NEXT: v_sub_f32_e32 v3, v4, v2
; GFX7LESS-NEXT: s_waitcnt expcnt(0)
-; GFX7LESS-NEXT: v_mov_b32_e32 v4, v1
-; GFX7LESS-NEXT: v_mov_b32_e32 v3, v0
-; GFX7LESS-NEXT: buffer_atomic_cmpswap v[3:4], off, s[0:3], 0 glc
+; GFX7LESS-NEXT: v_mov_b32_e32 v0, v3
+; GFX7LESS-NEXT: v_mov_b32_e32 v1, v4
+; GFX7LESS-NEXT: buffer_atomic_cmpswap v[0:1], off, s[0:3], 0 glc
; GFX7LESS-NEXT: s_waitcnt vmcnt(0)
-; GFX7LESS-NEXT: v_cmp_eq_u32_e32 vcc, v3, v1
+; GFX7LESS-NEXT: v_cmp_eq_u32_e32 vcc, v0, v4
; GFX7LESS-NEXT: s_or_b64 s[4:5], vcc, s[4:5]
-; GFX7LESS-NEXT: v_mov_b32_e32 v1, v3
; GFX7LESS-NEXT: s_andn2_b64 exec, exec, s[4:5]
; GFX7LESS-NEXT: s_cbranch_execnz .LBB2_2
; GFX7LESS-NEXT: .LBB2_3:
@@ -1525,19 +1525,19 @@ define amdgpu_kernel void @global_atomic_fsub_uni_address_uni_value_one_as_scope
; GFX7LESS-DPP-NEXT: v_cvt_f32_f64_e32 v0, v[0:1]
; GFX7LESS-DPP-NEXT: v_mul_f32_e32 v2, 4.0, v0
; GFX7LESS-DPP-NEXT: s_waitcnt lgkmcnt(0)
-; GFX7LESS-DPP-NEXT: v_mov_b32_e32 v1, s2
+; GFX7LESS-DPP-NEXT: v_mov_b32_e32 v0, s2
; GFX7LESS-DPP-NEXT: s_mov_b32 s2, -1
; GFX7LESS-DPP-NEXT: .LBB2_2: ; %atomicrmw.start
; GFX7LESS-DPP-NEXT: ; =>This Inner Loop Header: Depth=1
-; GFX7LESS-DPP-NEXT: v_sub_f32_e32 v0, v1, v2
+; GFX7LESS-DPP-NEXT: v_mov_b32_e32 v4, v0
+; GFX7LESS-DPP-NEXT: v_sub_f32_e32 v3, v4, v2
; GFX7LESS-DPP-NEXT: s_waitcnt expcnt(0)
-; GFX7LESS-DPP-NEXT: v_mov_b32_e32 v4, v1
-; GFX7LESS-DPP-NEXT: v_mov_b32_e32 v3, v0
-; GFX7LESS-DPP-NEXT: buffer_atomic_cmpswap v[3:4], off, s[0:3], 0 glc
+; GFX7LESS-DPP-NEXT: v_mov_b32_e32 v0, v3
+; GFX7LESS-DPP-NEXT: v_mov_b32_e32 v1, v4
+; GFX7LESS-DPP-NEXT: buffer_atomic_cmpswap v[0:1], off, s[0:3], 0 glc
; GFX7LESS-DPP-NEXT: s_waitcnt vmcnt(0)
-; GFX7LESS-DPP-NEXT: v_cmp_eq_u32_e32 vcc, v3, v1
+; GFX7LESS-DPP-NEXT: v_cmp_eq_u32_e32 vcc, v0, v4
; GFX7LESS-DPP-NEXT: s_or_b64 s[4:5], vcc, s[4:5]
-; GFX7LESS-DPP-NEXT: v_mov_b32_e32 v1, v3
; GFX7LESS-DPP-NEXT: s_andn2_b64 exec, exec, s[4:5]
; GFX7LESS-DPP-NEXT: s_cbranch_execnz .LBB2_2
; GFX7LESS-DPP-NEXT: .LBB2_3:
@@ -1806,20 +1806,20 @@ define amdgpu_kernel void @global_atomic_fsub_uni_address_div_value_one_as_scope
; GFX7LESS-NEXT: s_mov_b32 s3, 0xf000
; GFX7LESS-NEXT: s_mov_b32 s2, -1
; GFX7LESS-NEXT: s_waitcnt lgkmcnt(0)
-; GFX7LESS-NEXT: buffer_load_dword v1, off, s[0:3], 0
+; GFX7LESS-NEXT: buffer_load_dword v0, off, s[0:3], 0
; GFX7LESS-NEXT: s_mov_b64 s[4:5], 0
; GFX7LESS-NEXT: .LBB3_4: ; %atomicrmw.start
; GFX7LESS-NEXT: ; =>This Inner Loop Header: Depth=1
; GFX7LESS-NEXT: s_waitcnt vmcnt(0)
-; GFX7LESS-NEXT: v_sub_f32_e32 v0, v1, v2
+; GFX7LESS-NEXT: v_mov_b32_e32 v4, v0
+; GFX7LESS-NEXT: v_sub_f32_e32 v3, v4, v2
; GFX7LESS-NEXT: s_waitcnt expcnt(0)
-; GFX7LESS-NEXT: v_mov_b32_e32 v4, v1
-; GFX7LESS-NEXT: v_mov_b32_e32 v3, v0
-; GFX7LESS-NEXT: buffer_atomic_cmpswap v[3:4], off, s[0:3], 0 glc
+; GFX7LESS-NEXT: v_mov_b32_e32 v0, v3
+; GFX7LESS-NEXT: v_mov_b32_e32 v1, v4
+; GFX7LESS-NEXT: buffer_atomic_cmpswap v[0:1], off, s[0:3], 0 glc
; GFX7LESS-NEXT: s_waitcnt vmcnt(0)
-; GFX7LESS-NEXT: v_cmp_eq_u32_e32 vcc, v3, v1
+; GFX7LESS-NEXT: v_cmp_eq_u32_e32 vcc, v0, v4
; GFX7LESS-NEXT: s_or_b64 s[4:5], vcc, s[4:5]
-; GFX7LESS-NEXT: v_mov_b32_e32 v1, v3
; GFX7LESS-NEXT: s_andn2_b64 exec, exec, s[4:5]
; GFX7LESS-NEXT: s_cbranch_execnz .LBB3_4
; GFX7LESS-NEXT: .LBB3_5:
@@ -2178,20 +2178,20 @@ define amdgpu_kernel void @global_atomic_fsub_uni_address_div_value_one_as_scope
; GFX7LESS-DPP-NEXT: s_mov_b64 s[2:3], s[50:51]
; GFX7LESS-DPP-NEXT: s_waitcnt lgkmcnt(0)
; GFX7LESS-DPP-NEXT: s_swappc_b64 s[30:31], s[16:17]
-; GFX7LESS-DPP-NEXT: buffer_load_dword v2, off, s[36:39], 0
+; GFX7LESS-DPP-NEXT: buffer_load_dword v1, off, s[36:39], 0
; GFX7LESS-DPP-NEXT: s_mov_b64 s[0:1], 0
; GFX7LESS-DPP-NEXT: .LBB3_1: ; %atomicrmw.start
; GFX7LESS-DPP-NEXT: ; =>This Inner Loop Header: Depth=1
; GFX7LESS-DPP-NEXT: s_waitcnt vmcnt(0)
-; GFX7LESS-DPP-NEXT: v_sub_f32_e32 v1, v2, v0
-; GFX7LESS-DPP-NEXT: s_waitcnt expcnt(0)
-; GFX7LESS-DPP-NEXT: v_mov_b32_e32 v4, v2
; GFX7LESS-DPP-NEXT: v_mov_b32_e32 v3, v1
-; GFX7LESS-DPP-NEXT: buffer_atomic_cmpswap v[3:4], off, s[36:39], 0 glc
+; GFX7LESS-DPP-NEXT: s_waitcnt expcnt(0)
+; GFX7LESS-DPP-NEXT: v_sub_f32_e32 v2, v3, v0
+; GFX7LESS-DPP-NEXT: v_mov_b32_e32 v1, v2
+; GFX7LESS-DPP-NEXT: v_mov_b32_e32 v2, v3
+; GFX7LESS-DPP-NEXT: buffer_atomic_cmpswap v[1:2], off, s[36:39], 0 glc
; GFX7LESS-DPP-NEXT: s_waitcnt vmcnt(0)
-; GFX7LESS-DPP-NEXT: v_cmp_eq_u32_e32 vcc, v3, v2
+; GFX7LESS-DPP-NEXT: v_cmp_eq_u32_e32 vcc, v1, v3
; GFX7LESS-DPP-NEXT: s_or_b64 s[0:1], vcc, s[0:1]
-; GFX7LESS-DPP-NEXT: v_mov_b32_e32 v2, v3
; GFX7LESS-DPP-NEXT: s_andn2_b64 exec, exec, s[0:1]
; GFX7LESS-DPP-NEXT: s_cbranch_execnz .LBB3_1
; GFX7LESS-DPP-NEXT: ; %bb.2: ; %atomicrmw.end
@@ -2605,19 +2605,19 @@ define amdgpu_kernel void @global_atomic_fsub_uni_address_uni_value_agent_scope_
; GFX7LESS-NEXT: v_cvt_f32_f64_e32 v0, v[0:1]
; GFX7LESS-NEXT: v_mul_f32_e32 v2, 4.0, v0
; GFX7LESS-NEXT: s_waitcnt lgkmcnt(0)
-; GFX7LESS-NEXT: v_mov_b32_e32 v1, s2
+; GFX7LESS-NEXT: v_mov_b32_e32 v0, s2
; GFX7LESS-NEXT: s_mov_b32 s2, -1
; GFX7LESS-NEXT: .LBB4_2: ; %atomicrmw.start
; GFX7LESS-NEXT: ; =>This Inner Loop Header: Depth=1
-; GFX7LESS-NEXT: v_sub_f32_e32 v0, v1, v2
+; GFX7LESS-NEXT: v_mov_b32_e32 v4, v0
+; GFX7LESS-NEXT: v_sub_f32_e32 v3, v4, v2
; GFX7LESS-NEXT: s_waitcnt expcnt(0)
-; GFX7LESS-NEXT: v_mov_b32_e32 v4, v1
-; GFX7LESS-NEXT: v_mov_b32_e32 v3, v0
-; GFX7LESS-NEXT: buffer_atomic_cmpswap v[3:4], off, s[0:3], 0 glc
+; GFX7LESS-NEXT: v_mov_b32_e32 v0, v3
+; GFX7LESS-NEXT: v_mov_b32_e32 v1, v4
+; GFX7LESS-NEXT: buffer_atomic_cmpswap v[0:1], off, s[0:3], 0 glc
; GFX7LESS-NEXT: s_waitcnt vmcnt(0)
-; GFX7LESS-NEXT: v_cmp_eq_u32_e32 vcc, v3, v1
+; GFX7LESS-NEXT: v_cmp_eq_u32_e32 vcc, v0, v4
; GFX7LESS-NEXT: s_or_b64 s[4:5], vcc, s[4:5]
-; GFX7LESS-NEXT: v_mov_b32_e32 v1, v3
; GFX7LESS-NEXT: s_andn2_b64 exec, exec, s[4:5]
; GFX7LESS-NEXT: s_cbranch_execnz .LBB4_2
; GFX7LESS-NEXT: .LBB4_3:
@@ -2855,19 +2855,19 @@ define amdgpu_kernel void @global_atomic_fsub_uni_address_uni_value_agent_scope_
; GFX7LESS-DPP-NEXT: v_cvt_f32_f64_e32 v0, v[0:1]
; GFX7LESS-DPP-NEXT: v_mul_f32_e32 v2, 4.0, v0
; GFX7LESS-DPP-NEXT: s_waitcnt lgkmcnt(0)
-; GFX7LESS-DPP-NEXT: v_mov_b32_e32 v1, s2
+; GFX7LESS-DPP-NEXT: v_mov_b32_e32 v0, s2
; GFX7LESS-DPP-NEXT: s_mov_b32 s2, -1
; GFX7LESS-DPP-NEXT: .LBB4_2: ; %atomicrmw.start
; GFX7LESS-DPP-NEXT: ; =>This Inner Loop Header: Depth=1
-; GFX7LESS-DPP-NEXT: v_sub_f32_e32 v0, v1, v2
+; GFX7LESS-DPP-NEXT: v_mov_b32_e32 v4, v0
+; GFX7LESS-DPP-NEXT: v_sub_f32_e32 v3, v4, v2
; GFX7LESS-DPP-NEXT: s_waitcnt expcnt(0)
-; GFX7LESS-DPP-NEXT: v_mov_b32_e32 v4, v1
-; GFX7LESS-DPP-NEXT: v_mov_b32_e32 v3, v0
-; GFX7LESS-DPP-NEXT: buffer_atomic_cmpswap v[3:4], off, s[0:3], 0 glc
+; GFX7LESS-DPP-NEXT: v_mov_b32_e32 v0, v3
+; GFX7LESS-DPP-NEXT: v_mov_b32_e32 v1, v4
+; GFX7LESS-DPP-NEXT: buffer_atomic_cmpswap v[0:1], off, s[0:3], 0 glc
; GFX7LESS-DPP-NEXT: s_waitcnt vmcnt(0)
-; GFX7LESS-DPP-NEXT: v_cmp_eq_u32_e32 vcc, v3, v1
+; GFX7LESS-DPP-NEXT: v_cmp_eq_u32_e32 vcc, v0, v4
; GFX7LESS-DPP-NEXT: s_or_b64 s[4:5], vcc, s[4:5]
-; GFX7LESS-DPP-NEXT: v_mov_b32_e32 v1, v3
; GFX7LESS-DPP-NEXT: s_andn2_b64 exec, exec, s[4:5]
; GFX7LESS-DPP-NEXT: s_cbranch_execnz .LBB4_2
; GFX7LESS-DPP-NEXT: .LBB4_3:
@@ -3136,20 +3136,20 @@ define amdgpu_kernel void @global_atomic_fsub_uni_address_div_value_agent_scope_
; GFX7LESS-NEXT: s_mov_b32 s3, 0xf000
; GFX7LESS-NEXT: s_mov_b32 s2, -1
; GFX7LESS-NEXT: s_waitcnt lgkmcnt(0)
-; GFX7LESS-NEXT: buffer_load_dword v1, off, s[0:3], 0
+; GFX7LESS-NEXT: buffer_load_dword v0, off, s[0:3], 0
; GFX7LESS-NEXT: s_mov_b64 s[4:5], 0
; GFX7LESS-NEXT: .LBB5_4: ; %atomicrmw.start
; GFX7LESS-NEXT: ; =>This Inner Loop Header: Depth=1
; GFX7LESS-NEXT: s_waitcnt vmcnt(0)
-; GFX7LESS-NEXT: v_sub_f32_e32 v0, v1, v2
+; GFX7LESS-NEXT: v_mov_b32_e32 v4, v0
+; GFX7LESS-NEXT: v_sub_f32_e32 v3, v4, v2
; GFX7LESS-NEXT: s_waitcnt expcnt(0)
-; GFX7LESS-NEXT: v_mov_b32_e32 v4, v1
-; GFX7LESS-NEXT: v_mov_b32_e32 v3, v0
-; GFX7LESS-NEXT: buffer_atomic_cmpswap v[3:4], off, s[0:3], 0 glc
+; GFX7LESS-NEXT: v_mov_b32_e32 v0, v3
+; GFX7LESS-NEXT: v_mov_b32_e32 v1, v4
+; GFX7LESS-NEXT: buffer_atomic_cmpswap v[0:1], off, s[0:3], 0 glc
; GFX7LESS-NEXT: s_waitcnt vmcnt(0)
-; GFX7LESS-NEXT: v_cmp_eq_u32_e32 vcc, v3, v1
+; GFX7LESS-NEXT: v_cmp_eq_u32_e32 vcc, v0, v4
; GFX7LESS-NEXT: s_or_b64 s[4:5], vcc, s[4:5]
-; GFX7LESS-NEXT: v_mov_b32_e32 v1, v3
; GFX7LESS-NEXT: s_andn2_b64 exec, exec, s[4:5]
; GFX7LESS-NEXT: s_cbranch_execnz .LBB5_4
; GFX7LESS-NEXT: .LBB5_5:
@@ -3508,20 +3508,20 @@ define amdgpu_kernel void @global_atomic_fsub_uni_address_div_value_agent_scope_
; GFX7LESS-DPP-NEXT: s_mov_b64 s[2:3], s[50:51]
; GFX7LESS-DPP-NEXT: s_waitcnt lgkmcnt(0)
; GFX7LESS-DPP-NEXT: s_swappc_b64 s[30:31], s[16:17]
-; GFX7LESS-DPP-NEXT: buffer_load_dword v2, off, s[36:39], 0
+; GFX7LESS-DPP-NEXT: buffer_load_dword v1, off, s[36:39], 0
; GFX7LESS-DPP-NEXT: s_mov_b64 s[0:1], 0
; GFX7LESS-DPP-NEXT: .LBB5_1: ; %atomicrmw.start
; GFX7LESS-DPP-NEXT: ; =>This Inner Loop Header: Depth=1
; GFX7LESS-DPP-NEXT: s_waitcnt vmcnt(0)
-; GFX7LESS-DPP-NEXT: v_sub_f32_e32 v1, v2, v0
-; GFX7LESS-DPP-NEXT: s_waitcnt expcnt(0)
-; GFX7LESS-DPP-NEXT: v_mov_b32_e32 v4, v2
; GFX7LESS-DPP-NEXT: v_mov_b32_e32 v3, v1
-; GFX7LESS-DPP-NEXT: buffer_atomic_cmpswap v[3:4], off, s[36:39], 0 glc
+; GFX7LESS-DPP-NEXT: s_waitcnt expcnt(0)
+; GFX7LESS-DPP-NEXT: v_sub_f32_e32 v2, v3, v0
+; GFX7LESS-DPP-NEXT: v_mov_b32_e32 v1, v2
+; GFX7LESS-DPP-NEXT: v_mov_b32_e32 v2, v3
+; GFX7LESS-DPP-NEXT: buffer_atomic_cmpswap v[1:2], off, s[36:39], 0 glc
; GFX7LESS-DPP-NEXT: s_waitcnt vmcnt(0)
-; GFX7LESS-DPP-NEXT: v_cmp_eq_u32_e32 vcc, v3, v2
+; GFX7LESS-DPP-NEXT: v_cmp_eq_u32_e32 vcc, v1, v3
; GFX7LESS-DPP-NEXT: s_or_b64 s[0:1], vcc, s[0:1]
-; GFX7LESS-DPP-NEXT: v_mov_b32_e32 v2, v3
; GFX7LESS-DPP-NEXT: s_andn2_b64 exec, exec, s[0:1]
; GFX7LESS-DPP-NEXT: s_cbranch_execnz .LBB5_1
; GFX7LESS-DPP-NEXT: ; %bb.2: ; %atomicrmw.end
@@ -3962,20 +3962,20 @@ define amdgpu_kernel void @global_atomic_fsub_uni_address_div_value_agent_scope_
; GFX7LESS-NEXT: s_mov_b32 s3, 0xf000
; GFX7LESS-NEXT: s_mov_b32 s2, -1
; GFX7LESS-NEXT: s_waitcnt lgkmcnt(0)
-; GFX7LESS-NEXT: buffer_load_dword v1, off, s[0:3], 0
+; GFX7LESS-NEXT: buffer_load_dword v0, off, s[0:3], 0
; GFX7LESS-NEXT: s_mov_b64 s[4:5], 0
; GFX7LESS-NEXT: .LBB6_4: ; %atomicrmw.start
; GFX7LESS-NEXT: ; =>This Inner Loop Header: Depth=1
; GFX7LESS-NEXT: s_waitcnt vmcnt(0)
-; GFX7LESS-NEXT: v_sub_f32_e32 v0, v1, v2
+; GFX7LESS-NEXT: v_mov_b32_e32 v4, v0
+; GFX7LESS-NEXT: v_sub_f32_e32 v3, v4, v2
; GFX7LESS-NEXT: s_waitcnt expcnt(0)
-; GFX7LESS-NEXT: v_mov_b32_e32 v4, v1
-; GFX7LESS-NEXT: v_mov_b32_e32 v3, v0
-; GFX7LESS-NEXT: buffer_atomic_cmpswap v[3:4], off, s[0:3], 0 glc
+; GFX7LESS-NEXT: v_mov_b32_e32 v0, v3
+; GFX7LESS-NEXT: v_mov_b32_e32 v1, v4
+; GFX7LESS-NEXT: buffer_atomic_cmpswap v[0:1], off, s[0:3], 0 glc
; GFX7LESS-NEXT: s_waitcnt vmcnt(0)
-; GFX7LESS-NEXT: v_cmp_eq_u32_e32 vcc, v3, v1
+; GFX7LESS-NEXT: v_cmp_eq_u32_e32 vcc, v0, v4
; GFX7LESS-NEXT: s_or_b64 s[4:5], vcc, s[4:5]
-; GFX7LESS-NEXT: v_mov_b32_e32 v1, v3
; GFX7LESS-NEXT: s_andn2_b64 exec, exec, s[4:5]
; GFX7LESS-NEXT: s_cbranch_execnz .LBB6_4
; GFX7LESS-NEXT: .LBB6_5:
@@ -4334,20 +4334,20 @@ define amdgpu_kernel void @global_atomic_fsub_uni_address_div_value_agent_scope_
; GFX7LESS-DPP-NEXT: s_mov_b64 s[2:3], s[50:51]
; GFX7LESS-DPP-NEXT: s_waitcnt lgkmcnt(0)
; GFX7LESS-DPP-NEXT: s_swappc_b64 s[30:31], s[16:17]
-; GFX7LESS-DPP-NEXT: buffer_load_dword v2, off, s[36:39], 0
+; GFX7LESS-DPP-NEXT: buffer_load_dword v1, off, s[36:39], 0
; GFX7LESS-DPP-NEXT: s_mov_b64 s[0:1], 0
; GFX7LESS-DPP-NEXT: .LBB6_1: ; %atomicrmw.start
; GFX7LESS-DPP-NEXT: ; =>This Inner Loop Header: Depth=1
; GFX7LESS-DPP-NEXT: s_waitcnt vmcnt(0)
-; GFX7LESS-DPP-NEXT: v_sub_f32_e32 v1, v2, v0
-; GFX7LESS-DPP-NEXT: s_waitcnt expcnt(0)
-; GFX7LESS-DPP-NEXT: v_mov_b32_e32 v4, v2
; GFX7LESS-DPP-NEXT: v_mov_b32_e32 v3, v1
-; GFX7LESS-DPP-NEXT: buffer_atomic_cmpswap v[3:4], off, s[36:39], 0 glc
+; GFX7LESS-DPP-NEXT: s_waitcnt expcnt(0)
+; GFX7LESS-DPP-NEXT: v_sub_f32_e32 v2, v3, v0
+; GFX7LESS-DPP-NEXT: v_mov_b32_e32 v1, v2
+; GFX7LESS-DPP-NEXT: v_mov_b32_e32 v2, v3
+; GFX7LESS-DPP-NEXT: buffer_atomic_cmpswap v[1:2], off, s[36:39], 0 glc
; GFX7LESS-DPP-NEXT: s_waitcnt vmcnt(0)
-; GFX7LESS-DPP-NEXT: v_cmp_eq_u32_e32 vcc, v3, v2
+; GFX7LESS-DPP-NEXT: v_cmp_eq_u32_e32 vcc, v1, v3
; GFX7LESS-DPP-NEXT: s_or_b64 s[0:1], vcc, s[0:1]
-; GFX7LESS-DPP-NEXT: v_mov_b32_e32 v2, v3
; GFX7LESS-DPP-NEXT: s_andn2_b64 exec, exec, s[0:1]
; GFX7LESS-DPP-NEXT: s_cbranch_execnz .LBB6_1
; GFX7LESS-DPP-NEXT: ; %bb.2: ; %atomicrmw.end
@@ -4761,19 +4761,19 @@ define amdgpu_kernel void @global_atomic_fsub_uni_address_uni_value_default_scop
; GFX7LESS-NEXT: v_cvt_f32_f64_e32 v0, v[0:1]
; GFX7LESS-NEXT: v_mul_f32_e32 v2, 4.0, v0
; GFX7LESS-NEXT: s_waitcnt lgkmcnt(0)
-; GFX7LESS-NEXT: v_mov_b32_e32 v1, s2
+; GFX7LESS-NEXT: v_mov_b32_e32 v0, s2
; GFX7LESS-NEXT: s_mov_b32 s2, -1
; GFX7LESS-NEXT: .LBB7_2: ; %atomicrmw.start
; GFX7LESS-NEXT: ; =>This Inner Loop Header: Depth=1
-; GFX7LESS-NEXT: v_sub_f32_e32 v0, v1, v2
+; GFX7LESS-NEXT: v_mov_b32_e32 v4, v0
+; GFX7LESS-NEXT: v_sub_f32_e32 v3, v4, v2
; GFX7LESS-NEXT: s_waitcnt expcnt(0)
-; GFX7LESS-NEXT: v_mov_b32_e32 v4, v1
-; GFX7LESS-NEXT: v_mov_b32_e32 v3, v0
-; GFX7LESS-NEXT: buffer_atomic_cmpswap v[3:4], off, s[0:3], 0 glc
+; GFX7LESS-NEXT: v_mov_b32_e32 v0, v3
+; GFX7LESS-NEXT: v_mov_b32_e32 v1, v4
+; GFX7LESS-NEXT: buffer_atomic_cmpswap v[0:1], off, s[0:3], 0 glc
; GFX7LESS-NEXT: s_waitcnt vmcnt(0)
-; GFX7LESS-NEXT: v_cmp_eq_u32_e32 vcc, v3, v1
+; GFX7LESS-NEXT: v_cmp_eq_u32_e32 vcc, v0, v4
; GFX7LESS-NEXT: s_or_b64 s[4:5], vcc, s[4:5]
-; GFX7LESS-NEXT: v_mov_b32_e32 v1, v3
; GFX7LESS-NEXT: s_andn2_b64 exec, exec, s[4:5]
; GFX7LESS-NEXT: s_cbranch_execnz .LBB7_2
; GFX7LESS-NEXT: .LBB7_3:
@@ -5011,19 +5011,19 @@ define amdgpu_kernel void @global_atomic_fsub_uni_address_uni_value_default_scop
; GFX7LESS-DPP-NEXT: v_cvt_f32_f64_e32 v0, v[0:1]
; GFX7LESS-DPP-NEXT: v_mul_f32_e32 v2, 4.0, v0
; GFX7LESS-DPP-NEXT: s_waitcnt lgkmcnt(0)
-; GFX7LESS-DPP-NEXT: v_mov_b32_e32 v1, s2
+; GFX7LESS-DPP-NEXT: v_mov_b32_e32 v0, s2
; GFX7LESS-DPP-NEXT: s_mov_b32 s2, -1
; GFX7LESS-DPP-NEXT: .LBB7_2: ; %atomicrmw.start
; GFX7LESS-DPP-NEXT: ; =>This Inner Loop Header: Depth=1
-; GFX7LESS-DPP-NEXT: v_sub_f32_e32 v0, v1, v2
+; GFX7LESS-DPP-NEXT: v_mov_b32_e32 v4, v0
+; GFX7LESS-DPP-NEXT: v_sub_f32_e32 v3, v4, v2
; GFX7LESS-DPP-NEXT: s_waitcnt expcnt(0)
-; GFX7LESS-DPP-NEXT: v_mov_b32_e32 v4, v1
-; GFX7LESS-DPP-NEXT: v_mov_b32_e32 v3, v0
-; GFX7LESS-DPP-NEXT: buffer_atomic_cmpswap v[3:4], off, s[0:3], 0 glc
+; GFX7LESS-DPP-NEXT: v_mov_b32_e32 v0, v3
+; GFX7LESS-DPP-NEXT: v_mov_b32_e32 v1, v4
+; GFX7LESS-DPP-NEXT: buffer_atomic_cmpswap v[0:1], off, s[0:3], 0 glc
; GFX7LESS-DPP-NEXT: s_waitcnt vmcnt(0)
-; GFX7LESS-DPP-NEXT: v_cmp_eq_u32_e32 vcc, v3, v1
+; GFX7LESS-DPP-NEXT: v_cmp_eq_u32_e32 vcc, v0, v4
; GFX7LESS-DPP-NEXT: s_or_b64 s[4:5], vcc, s[4:5]
-; GFX7LESS-DPP-NEXT: v_mov_b32_e32 v1, v3
; GFX7LESS-DPP-NEXT: s_andn2_b64 exec, exec, s[4:5]
; GFX7LESS-DPP-NEXT: s_cbranch_execnz .LBB7_2
; GFX7LESS-DPP-NEXT: .LBB7_3:
@@ -5291,20 +5291,20 @@ define amdgpu_kernel void @global_atomic_fsub_uni_address_div_value_default_scop
; GFX7LESS-NEXT: s_mov_b32 s3, 0xf000
; GFX7LESS-NEXT: s_mov_b32 s2, -1
; GFX7LESS-NEXT: s_waitcnt lgkmcnt(0)
-; GFX7LESS-NEXT: buffer_load_dword v1, off, s[0:3], 0
+; GFX7LESS-NEXT: buffer_load_dword v0, off, s[0:3], 0
; GFX7LESS-NEXT: s_mov_b64 s[4:5], 0
; GFX7LESS-NEXT: .LBB8_4: ; %atomicrmw.start
; GFX7LESS-NEXT: ; =>This Inner Loop Header: Depth=1
; GFX7LESS-NEXT: s_waitcnt vmcnt(0)
-; GFX7LESS-NEXT: v_sub_f32_e32 v0, v1, v2
+; GFX7LESS-NEXT: v_mov_b32_e32 v4, v0
+; GFX7LESS-NEXT: v_sub_f32_e32 v3, v4, v2
; GFX7LESS-NEXT: s_waitcnt expcnt(0)
-; GFX7LESS-NEXT: v_mov_b32_e32 v4, v1
-; GFX7LESS-NEXT: v_mov_b32_e32 v3, v0
-; GFX7LESS-NEXT: buffer_atomic_cmpswap v[3:4], off, s[0:3], 0 glc
+; GFX7LESS-NEXT: v_mov_b32_e32 v0, v3
+; GFX7LESS-NEXT: v_mov_b32_e32 v1, v4
+; GFX7LESS-NEXT: buffer_atomic_cmpswap v[0:1], off, s[0:3], 0 glc
; GFX7LESS-NEXT: s_waitcnt vmcnt(0)
-; GFX7LESS-NEXT: v_cmp_eq_u32_e32 vcc, v3, v1
+; GFX7LESS-NEXT: v_cmp_eq_u32_e32 vcc, v0, v4
; GFX7LESS-NEXT: s_or_b64 s[4:5], vcc, s[4:5]
-; GFX7LESS-NEXT: v_mov_b32_e32 v1, v3
; GFX7LESS-NEXT: s_andn2_b64 exec, exec, s[4:5]
; GFX7LESS-NEXT: s_cbranch_execnz .LBB8_4
; GFX7LESS-NEXT: .LBB8_5:
@@ -5663,20 +5663,20 @@ define amdgpu_kernel void @global_atomic_fsub_uni_address_div_value_default_scop
; GFX7LESS-DPP-NEXT: s_mov_b64 s[2:3], s[50:51]
; GFX7LESS-DPP-NEXT: s_waitcnt lgkmcnt(0)
; GFX7LESS-DPP-NEXT: s_swappc_b64 s[30:31], s[16:17]
-; GFX7LESS-DPP-NEXT: buffer_load_dword v2, off, s[36:39], 0
+; GFX7LESS-DPP-NEXT: buffer_load_dword v1, off, s[36:39], 0
; GFX7LESS-DPP-NEXT: s_mov_b64 s[0:1], 0
; GFX7LESS-DPP-NEXT: .LBB8_1: ; %atomicrmw.start
; GFX7LESS-DPP-NEXT: ; =>This Inner Loop Header: Depth=1
; GFX7LESS-DPP-NEXT: s_waitcnt vmcnt(0)
-; GFX7LESS-DPP-NEXT: v_sub_f32_e32 v1, v2, v0
-; GFX7LESS-DPP-NEXT: s_waitcnt expcnt(0)
-; GFX7LESS-DPP-NEXT: v_mov_b32_e32 v4, v2
; GFX7LESS-DPP-NEXT: v_mov_b32_e32 v3, v1
-; GFX7LESS-DPP-NEXT: buffer_atomic_cmpswap v[3:4], off, s[36:39], 0 glc
+; GFX7LESS-DPP-NEXT: s_waitcnt expcnt(0)
+; GFX7LESS-DPP-NEXT: v_sub_f32_e32 v2, v3, v0
+; GFX7LESS-DPP-NEXT: v_mov_b32_e32 v1, v2
+; GFX7LESS-DPP-NEXT: v_mov_b32_e32 v2, v3
+; GFX7LESS-DPP-NEXT: buffer_atomic_cmpswap v[1:2], off, s[36:39], 0 glc
; GFX7LESS-DPP-NEXT: s_waitcnt vmcnt(0)
-; GFX7LESS-DPP-NEXT: v_cmp_eq_u32_e32 vcc, v3, v2
+; GFX7LESS-DPP-NEXT: v_cmp_eq_u32_e32 vcc, v1, v3
; GFX7LESS-DPP-NEXT: s_or_b64 s[0:1], vcc, s[0:1]
-; GFX7LESS-DPP-NEXT: v_mov_b32_e32 v2, v3
; GFX7LESS-DPP-NEXT: s_andn2_b64 exec, exec, s[0:1]
; GFX7LESS-DPP-NEXT: s_cbranch_execnz .LBB8_1
; GFX7LESS-DPP-NEXT: ; %bb.2: ; %atomicrmw.end
@@ -8332,23 +8332,23 @@ define amdgpu_kernel void @global_atomic_fsub_double_uni_address_uni_value_one_a
; GFX7LESS-NEXT: v_add_f64 v[0:1], s[6:7], v[0:1]
; GFX7LESS-NEXT: v_mul_f64 v[4:5], 4.0, v[0:1]
; GFX7LESS-NEXT: s_waitcnt lgkmcnt(0)
-; GFX7LESS-NEXT: v_mov_b32_e32 v2, s8
-; GFX7LESS-NEXT: v_mov_b32_e32 v3, s9
+; GFX7LESS-NEXT: v_mov_b32_e32 v0, s8
+; GFX7LESS-NEXT: v_mov_b32_e32 v1, s9
; GFX7LESS-NEXT: s_mov_b32 s2, -1
; GFX7LESS-NEXT: .LBB11_2: ; %atomicrmw.start
; GFX7LESS-NEXT: ; =>This Inner Loop Header: Depth=1
-; GFX7LESS-NEXT: v_add_f64 v[0:1], v[2:3], -v[4:5]
+; GFX7LESS-NEXT: v_mov_b32_e32 v9, v1
+; GFX7LESS-NEXT: v_mov_b32_e32 v8, v0
+; GFX7LESS-NEXT: v_add_f64 v[6:7], v[8:9], -v[4:5]
; GFX7LESS-NEXT: s_waitcnt expcnt(0)
-; GFX7LESS-NEXT: v_mov_b32_e32 v9, v3
-; GFX7LESS-NEXT: v_mov_b32_e32 v8, v2
-; GFX7LESS-NEXT: v_mov_b32_e32 v7, v1
-; GFX7LESS-NEXT: v_mov_b32_e32 v6, v0
-; GFX7LESS-NEXT: buffer_atomic_cmpswap_x2 v[6:9], off, s[0:3], 0 glc
+; GFX7LESS-NEXT: v_mov_b32_e32 v0, v6
+; GFX7LESS-NEXT: v_mov_b32_e32 v1, v7
+; GFX7LESS-NEXT: v_mov_b32_e32 v2, v8
+; GFX7LESS-NEXT: v_mov_b32_e32 v3, v9
+; GFX7LESS-NEXT: buffer_atomic_cmpswap_x2 v[0:3], off, s[0:3], 0 glc
; GFX7LESS-NEXT: s_waitcnt vmcnt(0)
-; GFX7LESS-NEXT: v_cmp_eq_u64_e32 vcc, v[6:7], v[2:3]
+; GFX7LESS-NEXT: v_cmp_eq_u64_e32 vcc, v[0:1], v[8:9]
; GFX7LESS-NEXT: s_or_b64 s[4:5], vcc, s[4:5]
-; GFX7LESS-NEXT: v_mov_b32_e32 v2, v6
-; GFX7LESS-NEXT: v_mov_b32_e32 v3, v7
; GFX7LESS-NEXT: s_andn2_b64 exec, exec, s[4:5]
; GFX7LESS-NEXT: s_cbranch_execnz .LBB11_2
; GFX7LESS-NEXT: .LBB11_3:
@@ -8589,23 +8589,23 @@ define amdgpu_kernel void @global_atomic_fsub_double_uni_address_uni_value_one_a
; GFX7LESS-DPP-NEXT: v_add_f64 v[0:1], s[6:7], v[0:1]
; GFX7LESS-DPP-NEXT: v_mul_f64 v[4:5], 4.0, v[0:1]
; GFX7LESS-DPP-NEXT: s_waitcnt lgkmcnt(0)
-; GFX7LESS-DPP-NEXT: v_mov_b32_e32 v2, s8
-; GFX7LESS-DPP-NEXT: v_mov_b32_e32 v3, s9
+; GFX7LESS-DPP-NEXT: v_mov_b32_e32 v0, s8
+; GFX7LESS-DPP-NEXT: v_mov_b32_e32 v1, s9
; GFX7LESS-DPP-NEXT: s_mov_b32 s2, -1
; GFX7LESS-DPP-NEXT: .LBB11_2: ; %atomicrmw.start
; GFX7LESS-DPP-NEXT: ; =>This Inner Loop Header: Depth=1
-; GFX7LESS-DPP-NEXT: v_add_f64 v[0:1], v[2:3], -v[4:5]
+; GFX7LESS-DPP-NEXT: v_mov_b32_e32 v9, v1
+; GFX7LESS-DPP-NEXT: v_mov_b32_e32 v8, v0
+; GFX7LESS-DPP-NEXT: v_add_f64 v[6:7], v[8:9], -v[4:5]
; GFX7LESS-DPP-NEXT: s_waitcnt expcnt(0)
-; GFX7LESS-DPP-NEXT: v_mov_b32_e32 v9, v3
-; GFX7LESS-DPP-NEXT: v_mov_b32_e32 v8, v2
-; GFX7LESS-DPP-NEXT: v_mov_b32_e32 v7, v1
-; GFX7LESS-DPP-NEXT: v_mov_b32_e32 v6, v0
-; GFX7LESS-DPP-NEXT: buffer_atomic_cmpswap_x2 v[6:9], off, s[0:3], 0 glc
+; GFX7LESS-DPP-NEXT: v_mov_b32_e32 v0, v6
+; GFX7LESS-DPP-NEXT: v_mov_b32_e32 v1, v7
+; GFX7LESS-DPP-NEXT: v_mov_b32_e32 v2, v8
+; GFX7LESS-DPP-NEXT: v_mov_b32_e32 v3, v9
+; GFX7LESS-DPP-NEXT: buffer_atomic_cmpswap_x2 v[0:3], off, s[0:3], 0 glc
; GFX7LESS-DPP-NEXT: s_waitcnt vmcnt(0)
-; GFX7LESS-DPP-NEXT: v_cmp_eq_u64_e32 vcc, v[6:7], v[2:3]
+; GFX7LESS-DPP-NEXT: v_cmp_eq_u64_e32 vcc, v[0:1], v[8:9]
; GFX7LESS-DPP-NEXT: s_or_b64 s[4:5], vcc, s[4:5]
-; GFX7LESS-DPP-NEXT: v_mov_b32_e32 v2, v6
-; GFX7LESS-DPP-NEXT: v_mov_b32_e32 v3, v7
; GFX7LESS-DPP-NEXT: s_andn2_b64 exec, exec, s[4:5]
; GFX7LESS-DPP-NEXT: s_cbranch_execnz .LBB11_2
; GFX7LESS-DPP-NEXT: .LBB11_3:
@@ -8878,23 +8878,23 @@ define amdgpu_kernel void @global_atomic_fsub_double_uni_address_div_value_one_a
; GFX7LESS-NEXT: s_mov_b32 s3, 0xf000
; GFX7LESS-NEXT: s_mov_b32 s2, -1
; GFX7LESS-NEXT: s_waitcnt lgkmcnt(0)
-; GFX7LESS-NEXT: buffer_load_dwordx2 v[2:3], off, s[0:3], 0
+; GFX7LESS-NEXT: buffer_load_dwordx2 v[0:1], off, s[0:3], 0
; GFX7LESS-NEXT: s_mov_b64 s[4:5], 0
; GFX7LESS-NEXT: .LBB12_4: ; %atomicrmw.start
; GFX7LESS-NEXT: ; =>This Inner Loop Header: Depth=1
; GFX7LESS-NEXT: s_waitcnt vmcnt(0)
-; GFX7LESS-NEXT: v_add_f64 v[0:1], v[2:3], -v[4:5]
+; GFX7LESS-NEXT: v_mov_b32_e32 v9, v1
+; GFX7LESS-NEXT: v_mov_b32_e32 v8, v0
+; GFX7LESS-NEXT: v_add_f64 v[6:7], v[8:9], -v[4:5]
; GFX7LESS-NEXT: s_waitcnt expcnt(0)
-; GFX7LESS-NEXT: v_mov_b32_e32 v9, v3
-; GFX7LESS-NEXT: v_mov_b32_e32 v8, v2
-; GFX7LESS-NEXT: v_mov_b32_e32 v7, v1
-; GFX7LESS-NEXT: v_mov_b32_e32 v6, v0
-; GFX7LESS-NEXT: buffer_atomic_cmpswap_x2 v[6:9], off, s[0:3], 0 glc
+; GFX7LESS-NEXT: v_mov_b32_e32 v0, v6
+; GFX7LESS-NEXT: v_mov_b32_e32 v1, v7
+; GFX7LESS-NEXT: v_mov_b32_e32 v2, v8
+; GFX7LESS-NEXT: v_mov_b32_e32 v3, v9
+; GFX7LESS-NEXT: buffer_atomic_cmpswap_x2 v[0:3], off, s[0:3], 0 glc
; GFX7LESS-NEXT: s_waitcnt vmcnt(0)
-; GFX7LESS-NEXT: v_cmp_eq_u64_e32 vcc, v[6:7], v[2:3]
+; GFX7LESS-NEXT: v_cmp_eq_u64_e32 vcc, v[0:1], v[8:9]
; GFX7LESS-NEXT: s_or_b64 s[4:5], vcc, s[4:5]
-; GFX7LESS-NEXT: v_mov_b32_e32 v2, v6
-; GFX7LESS-NEXT: v_mov_b32_e32 v3, v7
; GFX7LESS-NEXT: s_andn2_b64 exec, exec, s[4:5]
; GFX7LESS-NEXT: s_cbranch_execnz .LBB12_4
; GFX7LESS-NEXT: .LBB12_5:
@@ -9268,23 +9268,23 @@ define amdgpu_kernel void @global_atomic_fsub_double_uni_address_div_value_one_a
; GFX7LESS-DPP-NEXT: s_mov_b64 s[2:3], s[50:51]
; GFX7LESS-DPP-NEXT: s_waitcnt lgkmcnt(0)
; GFX7LESS-DPP-NEXT: s_swappc_b64 s[30:31], s[16:17]
-; GFX7LESS-DPP-NEXT: buffer_load_dwordx2 v[4:5], off, s[36:39], 0
+; GFX7LESS-DPP-NEXT: buffer_load_dwordx2 v[2:3], off, s[36:39], 0
; GFX7LESS-DPP-NEXT: s_mov_b64 s[0:1], 0
; GFX7LESS-DPP-NEXT: .LBB12_1: ; %atomicrmw.start
; GFX7LESS-DPP-NEXT: ; =>This Inner Loop Header: Depth=1
; GFX7LESS-DPP-NEXT: s_waitcnt vmcnt(0)
-; GFX7LESS-DPP-NEXT: v_add_f64 v[2:3], v[4:5], -v[0:1]
-; GFX7LESS-DPP-NEXT: s_waitcnt expcnt(0)
-; GFX7LESS-DPP-NEXT: v_mov_b32_e32 v9, v5
-; GFX7LESS-DPP-NEXT: v_mov_b32_e32 v8, v4
; GFX7LESS-DPP-NEXT: v_mov_b32_e32 v7, v3
; GFX7LESS-DPP-NEXT: v_mov_b32_e32 v6, v2
-; GFX7LESS-DPP-NEXT: buffer_atomic_cmpswap_x2 v[6:9], off, s[36:39], 0 glc
-; GFX7LESS-DPP-NEXT: s_waitcnt vmcnt(0)
-; GFX7LESS-DPP-NEXT: v_cmp_eq_u64_e32 vcc, v[6:7], v[4:5]
-; GFX7LESS-DPP-NEXT: s_or_b64 s[0:1], vcc, s[0:1]
+; GFX7LESS-DPP-NEXT: s_waitcnt expcnt(0)
+; GFX7LESS-DPP-NEXT: v_add_f64 v[4:5], v[6:7], -v[0:1]
+; GFX7LESS-DPP-NEXT: v_mov_b32_e32 v2, v4
+; GFX7LESS-DPP-NEXT: v_mov_b32_e32 v3, v5
; GFX7LESS-DPP-NEXT: v_mov_b32_e32 v4, v6
; GFX7LESS-DPP-NEXT: v_mov_b32_e32 v5, v7
+; GFX7LESS-DPP-NEXT: buffer_atomic_cmpswap_x2 v[2:5], off, s[36:39], 0 glc
+; GFX7LESS-DPP-NEXT: s_waitcnt vmcnt(0)
+; GFX7LESS-DPP-NEXT: v_cmp_eq_u64_e32 vcc, v[2:3], v[6:7]
+; GFX7LESS-DPP-NEXT: s_or_b64 s[0:1], vcc, s[0:1]
; GFX7LESS-DPP-NEXT: s_andn2_b64 exec, exec, s[0:1]
; GFX7LESS-DPP-NEXT: s_cbranch_execnz .LBB12_1
; GFX7LESS-DPP-NEXT: ; %bb.2: ; %atomicrmw.end
@@ -9764,23 +9764,23 @@ define amdgpu_kernel void @global_atomic_fsub_double_uni_address_uni_value_agent
; GFX7LESS-NEXT: v_add_f64 v[0:1], s[6:7], v[0:1]
; GFX7LESS-NEXT: v_mul_f64 v[4:5], 4.0, v[0:1]
; GFX7LESS-NEXT: s_waitcnt lgkmcnt(0)
-; GFX7LESS-NEXT: v_mov_b32_e32 v2, s8
-; GFX7LESS-NEXT: v_mov_b32_e32 v3, s9
+; GFX7LESS-NEXT: v_mov_b32_e32 v0, s8
+; GFX7LESS-NEXT: v_mov_b32_e32 v1, s9
; GFX7LESS-NEXT: s_mov_b32 s2, -1
; GFX7LESS-NEXT: .LBB13_2: ; %atomicrmw.start
; GFX7LESS-NEXT: ; =>This Inner Loop Header: Depth=1
-; GFX7LESS-NEXT: v_add_f64 v[0:1], v[2:3], -v[4:5]
+; GFX7LESS-NEXT: v_mov_b32_e32 v9, v1
+; GFX7LESS-NEXT: v_mov_b32_e32 v8, v0
+; GFX7LESS-NEXT: v_add_f64 v[6:7], v[8:9], -v[4:5]
; GFX7LESS-NEXT: s_waitcnt expcnt(0)
-; GFX7LESS-NEXT: v_mov_b32_e32 v9, v3
-; GFX7LESS-NEXT: v_mov_b32_e32 v8, v2
-; GFX7LESS-NEXT: v_mov_b32_e32 v7, v1
-; GFX7LESS-NEXT: v_mov_b32_e32 v6, v0
-; GFX7LESS-NEXT: buffer_atomic_cmpswap_x2 v[6:9], off, s[0:3], 0 glc
+; GFX7LESS-NEXT: v_mov_b32_e32 v0, v6
+; GFX7LESS-NEXT: v_mov_b32_e32 v1, v7
+; GFX7LESS-NEXT: v_mov_b32_e32 v2, v8
+; GFX7LESS-NEXT: v_mov_b32_e32 v3, v9
+; GFX7LESS-NEXT: buffer_atomic_cmpswap_x2 v[0:3], off, s[0:3], 0 glc
; GFX7LESS-NEXT: s_waitcnt vmcnt(0)
-; GFX7LESS-NEXT: v_cmp_eq_u64_e32 vcc, v[6:7], v[2:3]
+; GFX7LESS-NEXT: v_cmp_eq_u64_e32 vcc, v[0:1], v[8:9]
; GFX7LESS-NEXT: s_or_b64 s[4:5], vcc, s[4:5]
-; GFX7LESS-NEXT: v_mov_b32_e32 v2, v6
-; GFX7LESS-NEXT: v_mov_b32_e32 v3, v7
; GFX7LESS-NEXT: s_andn2_b64 exec, exec, s[4:5]
; GFX7LESS-NEXT: s_cbranch_execnz .LBB13_2
; GFX7LESS-NEXT: .LBB13_3:
@@ -10021,23 +10021,23 @@ define amdgpu_kernel void @global_atomic_fsub_double_uni_address_uni_value_agent
; GFX7LESS-DPP-NEXT: v_add_f64 v[0:1], s[6:7], v[0:1]
; GFX7LESS-DPP-NEXT: v_mul_f64 v[4:5], 4.0, v[0:1]
; GFX7LESS-DPP-NEXT: s_waitcnt lgkmcnt(0)
-; GFX7LESS-DPP-NEXT: v_mov_b32_e32 v2, s8
-; GFX7LESS-DPP-NEXT: v_mov_b32_e32 v3, s9
+; GFX7LESS-DPP-NEXT: v_mov_b32_e32 v0, s8
+; GFX7LESS-DPP-NEXT: v_mov_b32_e32 v1, s9
; GFX7LESS-DPP-NEXT: s_mov_b32 s2, -1
; GFX7LESS-DPP-NEXT: .LBB13_2: ; %atomicrmw.start
; GFX7LESS-DPP-NEXT: ; =>This Inner Loop Header: Depth=1
-; GFX7LESS-DPP-NEXT: v_add_f64 v[0:1], v[2:3], -v[4:5]
+; GFX7LESS-DPP-NEXT: v_mov_b32_e32 v9, v1
+; GFX7LESS-DPP-NEXT: v_mov_b32_e32 v8, v0
+; GFX7LESS-DPP-NEXT: v_add_f64 v[6:7], v[8:9], -v[4:5]
; GFX7LESS-DPP-NEXT: s_waitcnt expcnt(0)
-; GFX7LESS-DPP-NEXT: v_mov_b32_e32 v9, v3
-; GFX7LESS-DPP-NEXT: v_mov_b32_e32 v8, v2
-; GFX7LESS-DPP-NEXT: v_mov_b32_e32 v7, v1
-; GFX7LESS-DPP-NEXT: v_mov_b32_e32 v6, v0
-; GFX7LESS-DPP-NEXT: buffer_atomic_cmpswap_x2 v[6:9], off, s[0:3], 0 glc
+; GFX7LESS-DPP-NEXT: v_mov_b32_e32 v0, v6
+; GFX7LESS-DPP-NEXT: v_mov_b32_e32 v1, v7
+; GFX7LESS-DPP-NEXT: v_mov_b32_e32 v2, v8
+; GFX7LESS-DPP-NEXT: v_mov_b32_e32 v3, v9
+; GFX7LESS-DPP-NEXT: buffer_atomic_cmpswap_x2 v[0:3], off, s[0:3], 0 glc
; GFX7LESS-DPP-NEXT: s_waitcnt vmcnt(0)
-; GFX7LESS-DPP-NEXT: v_cmp_eq_u64_e32 vcc, v[6:7], v[2:3]
+; GFX7LESS-DPP-NEXT: v_cmp_eq_u64_e32 vcc, v[0:1], v[8:9]
; GFX7LESS-DPP-NEXT: s_or_b64 s[4:5], vcc, s[4:5]
-; GFX7LESS-DPP-NEXT: v_mov_b32_e32 v2, v6
-; GFX7LESS-DPP-NEXT: v_mov_b32_e32 v3, v7
; GFX7LESS-DPP-NEXT: s_andn2_b64 exec, exec, s[4:5]
; GFX7LESS-DPP-NEXT: s_cbranch_execnz .LBB13_2
; GFX7LESS-DPP-NEXT: .LBB13_3:
@@ -10311,23 +10311,23 @@ define amdgpu_kernel void @global_atomic_fsub_double_uni_address_div_value_agent
; GFX7LESS-NEXT: s_mov_b32 s3, 0xf000
; GFX7LESS-NEXT: s_mov_b32 s2, -1
; GFX7LESS-NEXT: s_waitcnt lgkmcnt(0)
-; GFX7LESS-NEXT: buffer_load_dwordx2 v[2:3], off, s[0:3], 0
+; GFX7LESS-NEXT: buffer_load_dwordx2 v[0:1], off, s[0:3], 0
; GFX7LESS-NEXT: s_mov_b64 s[4:5], 0
; GFX7LESS-NEXT: .LBB14_4: ; %atomicrmw.start
; GFX7LESS-NEXT: ; =>This Inner Loop Header: Depth=1
; GFX7LESS-NEXT: s_waitcnt vmcnt(0)
-; GFX7LESS-NEXT: v_add_f64 v[0:1], v[2:3], -v[4:5]
+; GFX7LESS-NEXT: v_mov_b32_e32 v9, v1
+; GFX7LESS-NEXT: v_mov_b32_e32 v8, v0
+; GFX7LESS-NEXT: v_add_f64 v[6:7], v[8:9], -v[4:5]
; GFX7LESS-NEXT: s_waitcnt expcnt(0)
-; GFX7LESS-NEXT: v_mov_b32_e32 v9, v3
-; GFX7LESS-NEXT: v_mov_b32_e32 v8, v2
-; GFX7LESS-NEXT: v_mov_b32_e32 v7, v1
-; GFX7LESS-NEXT: v_mov_b32_e32 v6, v0
-; GFX7LESS-NEXT: buffer_atomic_cmpswap_x2 v[6:9], off, s[0:3], 0 glc
+; GFX7LESS-NEXT: v_mov_b32_e32 v0, v6
+; GFX7LESS-NEXT: v_mov_b32_e32 v1, v7
+; GFX7LESS-NEXT: v_mov_b32_e32 v2, v8
+; GFX7LESS-NEXT: v_mov_b32_e32 v3, v9
+; GFX7LESS-NEXT: buffer_atomic_cmpswap_x2 v[0:3], off, s[0:3], 0 glc
; GFX7LESS-NEXT: s_waitcnt vmcnt(0)
-; GFX7LESS-NEXT: v_cmp_eq_u64_e32 vcc, v[6:7], v[2:3]
+; GFX7LESS-NEXT: v_cmp_eq_u64_e32 vcc, v[0:1], v[8:9]
; GFX7LESS-NEXT: s_or_b64 s[4:5], vcc, s[4:5]
-; GFX7LESS-NEXT: v_mov_b32_e32 v2, v6
-; GFX7LESS-NEXT: v_mov_b32_e32 v3, v7
; GFX7LESS-NEXT: s_andn2_b64 exec, exec, s[4:5]
; GFX7LESS-NEXT: s_cbranch_execnz .LBB14_4
; GFX7LESS-NEXT: .LBB14_5:
@@ -10701,23 +10701,23 @@ define amdgpu_kernel void @global_atomic_fsub_double_uni_address_div_value_agent
; GFX7LESS-DPP-NEXT: s_mov_b64 s[2:3], s[50:51]
; GFX7LESS-DPP-NEXT: s_waitcnt lgkmcnt(0)
; GFX7LESS-DPP-NEXT: s_swappc_b64 s[30:31], s[16:17]
-; GFX7LESS-DPP-NEXT: buffer_load_dwordx2 v[4:5], off, s[36:39], 0
+; GFX7LESS-DPP-NEXT: buffer_load_dwordx2 v[2:3], off, s[36:39], 0
; GFX7LESS-DPP-NEXT: s_mov_b64 s[0:1], 0
; GFX7LESS-DPP-NEXT: .LBB14_1: ; %atomicrmw.start
; GFX7LESS-DPP-NEXT: ; =>This Inner Loop Header: Depth=1
; GFX7LESS-DPP-NEXT: s_waitcnt vmcnt(0)
-; GFX7LESS-DPP-NEXT: v_add_f64 v[2:3], v[4:5], -v[0:1]
-; GFX7LESS-DPP-NEXT: s_waitcnt expcnt(0)
-; GFX7LESS-DPP-NEXT: v_mov_b32_e32 v9, v5
-; GFX7LESS-DPP-NEXT: v_mov_b32_e32 v8, v4
; GFX7LESS-DPP-NEXT: v_mov_b32_e32 v7, v3
; GFX7LESS-DPP-NEXT: v_mov_b32_e32 v6, v2
-; GFX7LESS-DPP-NEXT: buffer_atomic_cmpswap_x2 v[6:9], off, s[36:39], 0 glc
-; GFX7LESS-DPP-NEXT: s_waitcnt vmcnt(0)
-; GFX7LESS-DPP-NEXT: v_cmp_eq_u64_e32 vcc, v[6:7], v[4:5]
-; GFX7LESS-DPP-NEXT: s_or_b64 s[0:1], vcc, s[0:1]
+; GFX7LESS-DPP-NEXT: s_waitcnt expcnt(0)
+; GFX7LESS-DPP-NEXT: v_add_f64 v[4:5], v[6:7], -v[0:1]
+; GFX7LESS-DPP-NEXT: v_mov_b32_e32 v2, v4
+; GFX7LESS-DPP-NEXT: v_mov_b32_e32 v3, v5
; GFX7LESS-DPP-NEXT: v_mov_b32_e32 v4, v6
; GFX7LESS-DPP-NEXT: v_mov_b32_e32 v5, v7
+; GFX7LESS-DPP-NEXT: buffer_atomic_cmpswap_x2 v[2:5], off, s[36:39], 0 glc
+; GFX7LESS-DPP-NEXT: s_waitcnt vmcnt(0)
+; GFX7LESS-DPP-NEXT: v_cmp_eq_u64_e32 vcc, v[2:3], v[6:7]
+; GFX7LESS-DPP-NEXT: s_or_b64 s[0:1], vcc, s[0:1]
; GFX7LESS-DPP-NEXT: s_andn2_b64 exec, exec, s[0:1]
; GFX7LESS-DPP-NEXT: s_cbranch_execnz .LBB14_1
; GFX7LESS-DPP-NEXT: ; %bb.2: ; %atomicrmw.end
@@ -11226,23 +11226,23 @@ define amdgpu_kernel void @global_atomic_fsub_double_uni_address_div_value_agent
; GFX7LESS-NEXT: s_mov_b32 s3, 0xf000
; GFX7LESS-NEXT: s_mov_b32 s2, -1
; GFX7LESS-NEXT: s_waitcnt lgkmcnt(0)
-; GFX7LESS-NEXT: buffer_load_dwordx2 v[2:3], off, s[0:3], 0
+; GFX7LESS-NEXT: buffer_load_dwordx2 v[0:1], off, s[0:3], 0
; GFX7LESS-NEXT: s_mov_b64 s[4:5], 0
; GFX7LESS-NEXT: .LBB15_4: ; %atomicrmw.start
; GFX7LESS-NEXT: ; =>This Inner Loop Header: Depth=1
; GFX7LESS-NEXT: s_waitcnt vmcnt(0)
-; GFX7LESS-NEXT: v_add_f64 v[0:1], v[2:3], -v[4:5]
+; GFX7LESS-NEXT: v_mov_b32_e32 v9, v1
+; GFX7LESS-NEXT: v_mov_b32_e32 v8, v0
+; GFX7LESS-NEXT: v_add_f64 v[6:7], v[8:9], -v[4:5]
; GFX7LESS-NEXT: s_waitcnt expcnt(0)
-; GFX7LESS-NEXT: v_mov_b32_e32 v9, v3
-; GFX7LESS-NEXT: v_mov_b32_e32 v8, v2
-; GFX7LESS-NEXT: v_mov_b32_e32 v7, v1
-; GFX7LESS-NEXT: v_mov_b32_e32 v6, v0
-; GFX7LESS-NEXT: buffer_atomic_cmpswap_x2 v[6:9], off, s[0:3], 0 glc
+; GFX7LESS-NEXT: v_mov_b32_e32 v0, v6
+; GFX7LESS-NEXT: v_mov_b32_e32 v1, v7
+; GFX7LESS-NEXT: v_mov_b32_e32 v2, v8
+; GFX7LESS-NEXT: v_mov_b32_e32 v3, v9
+; GFX7LESS-NEXT: buffer_atomic_cmpswap_x2 v[0:3], off, s[0:3], 0 glc
; GFX7LESS-NEXT: s_waitcnt vmcnt(0)
-; GFX7LESS-NEXT: v_cmp_eq_u64_e32 vcc, v[6:7], v[2:3]
+; GFX7LESS-NEXT: v_cmp_eq_u64_e32 vcc, v[0:1], v[8:9]
; GFX7LESS-NEXT: s_or_b64 s[4:5], vcc, s[4:5]
-; GFX7LESS-NEXT: v_mov_b32_e32 v2, v6
-; GFX7LESS-NEXT: v_mov_b32_e32 v3, v7
; GFX7LESS-NEXT: s_andn2_b64 exec, exec, s[4:5]
; GFX7LESS-NEXT: s_cbranch_execnz .LBB15_4
; GFX7LESS-NEXT: .LBB15_5:
@@ -11616,23 +11616,23 @@ define amdgpu_kernel void @global_atomic_fsub_double_uni_address_div_value_agent
; GFX7LESS-DPP-NEXT: s_mov_b64 s[2:3], s[50:51]
; GFX7LESS-DPP-NEXT: s_waitcnt lgkmcnt(0)
; GFX7LESS-DPP-NEXT: s_swappc_b64 s[30:31], s[16:17]
-; GFX7LESS-DPP-NEXT: buffer_load_dwordx2 v[4:5], off, s[36:39], 0
+; GFX7LESS-DPP-NEXT: buffer_load_dwordx2 v[2:3], off, s[36:39], 0
; GFX7LESS-DPP-NEXT: s_mov_b64 s[0:1], 0
; GFX7LESS-DPP-NEXT: .LBB15_1: ; %atomicrmw.start
; GFX7LESS-DPP-NEXT: ; =>This Inner Loop Header: Depth=1
; GFX7LESS-DPP-NEXT: s_waitcnt vmcnt(0)
-; GFX7LESS-DPP-NEXT: v_add_f64 v[2:3], v[4:5], -v[0:1]
-; GFX7LESS-DPP-NEXT: s_waitcnt expcnt(0)
-; GFX7LESS-DPP-NEXT: v_mov_b32_e32 v9, v5
-; GFX7LESS-DPP-NEXT: v_mov_b32_e32 v8, v4
; GFX7LESS-DPP-NEXT: v_mov_b32_e32 v7, v3
; GFX7LESS-DPP-NEXT: v_mov_b32_e32 v6, v2
-; GFX7LESS-DPP-NEXT: buffer_atomic_cmpswap_x2 v[6:9], off, s[36:39], 0 glc
-; GFX7LESS-DPP-NEXT: s_waitcnt vmcnt(0)
-; GFX7LESS-DPP-NEXT: v_cmp_eq_u64_e32 vcc, v[6:7], v[4:5]
-; GFX7LESS-DPP-NEXT: s_or_b64 s[0:1], vcc, s[0:1]
+; GFX7LESS-DPP-NEXT: s_waitcnt expcnt(0)
+; GFX7LESS-DPP-NEXT: v_add_f64 v[4:5], v[6:7], -v[0:1]
+; GFX7LESS-DPP-NEXT: v_mov_b32_e32 v2, v4
+; GFX7LESS-DPP-NEXT: v_mov_b32_e32 v3, v5
; GFX7LESS-DPP-NEXT: v_mov_b32_e32 v4, v6
; GFX7LESS-DPP-NEXT: v_mov_b32_e32 v5, v7
+; GFX7LESS-DPP-NEXT: buffer_atomic_cmpswap_x2 v[2:5], off, s[36:39], 0 glc
+; GFX7LESS-DPP-NEXT: s_waitcnt vmcnt(0)
+; GFX7LESS-DPP-NEXT: v_cmp_eq_u64_e32 vcc, v[2:3], v[6:7]
+; GFX7LESS-DPP-NEXT: s_or_b64 s[0:1], vcc, s[0:1]
; GFX7LESS-DPP-NEXT: s_andn2_b64 exec, exec, s[0:1]
; GFX7LESS-DPP-NEXT: s_cbranch_execnz .LBB15_1
; GFX7LESS-DPP-NEXT: ; %bb.2: ; %atomicrmw.end
diff --git a/llvm/test/CodeGen/AMDGPU/indirect-addressing-si.ll b/llvm/test/CodeGen/AMDGPU/indirect-addressing-si.ll
index b5665835eaf7a..3d9fff23107b0 100644
--- a/llvm/test/CodeGen/AMDGPU/indirect-addressing-si.ll
+++ b/llvm/test/CodeGen/AMDGPU/indirect-addressing-si.ll
@@ -5863,12 +5863,6 @@ define amdgpu_kernel void @insert_vgpr_offset_multiple_in_block(ptr addrspace(1)
; NOOPT-NEXT: v_mov_b32_e32 v13, s13
; NOOPT-NEXT: v_mov_b32_e32 v14, s14
; NOOPT-NEXT: v_mov_b32_e32 v15, s15
-; NOOPT-NEXT: s_mov_b64 s[0:1], exec
-; NOOPT-NEXT: v_writelane_b32 v32, s0, 5
-; NOOPT-NEXT: v_writelane_b32 v32, s1, 6
-; NOOPT-NEXT: s_or_saveexec_b64 s[26:27], -1
-; NOOPT-NEXT: buffer_store_dword v32, off, s[28:31], 0 ; 4-byte Folded Spill
-; NOOPT-NEXT: s_mov_b64 exec, s[26:27]
; NOOPT-NEXT: buffer_store_dword v0, off, s[28:31], 0 offset:4 ; 4-byte Folded Spill
; NOOPT-NEXT: buffer_store_dword v1, off, s[28:31], 0 offset:8 ; 4-byte Folded Spill
; NOOPT-NEXT: buffer_store_dword v2, off, s[28:31], 0 offset:12 ; 4-byte Folded Spill
@@ -5885,6 +5879,12 @@ define amdgpu_kernel void @insert_vgpr_offset_multiple_in_block(ptr addrspace(1)
; NOOPT-NEXT: buffer_store_dword v13, off, s[28:31], 0 offset:56 ; 4-byte Folded Spill
; NOOPT-NEXT: buffer_store_dword v14, off, s[28:31], 0 offset:60 ; 4-byte Folded Spill
; NOOPT-NEXT: buffer_store_dword v15, off, s[28:31], 0 offset:64 ; 4-byte Folded Spill
+; NOOPT-NEXT: s_mov_b64 s[0:1], exec
+; NOOPT-NEXT: v_writelane_b32 v32, s0, 5
+; NOOPT-NEXT: v_writelane_b32 v32, s1, 6
+; NOOPT-NEXT: s_or_saveexec_b64 s[26:27], -1
+; NOOPT-NEXT: buffer_store_dword v32, off, s[28:31], 0 ; 4-byte Folded Spill
+; NOOPT-NEXT: s_mov_b64 exec, s[26:27]
; NOOPT-NEXT: ; implicit-def: $sgpr0_sgpr1
; NOOPT-NEXT: .LBB17_1: ; =>This Inner Loop Header: Depth=1
; NOOPT-NEXT: s_or_saveexec_b64 s[26:27], -1
@@ -5903,19 +5903,12 @@ define amdgpu_kernel void @insert_vgpr_offset_multiple_in_block(ptr addrspace(1)
; NOOPT-NEXT: buffer_load_dword v6, off, s[28:31], 0 offset:28 ; 4-byte Folded Reload
; NOOPT-NEXT: buffer_load_dword v7, off, s[28:31], 0 offset:32 ; 4-byte Folded Reload
; NOOPT-NEXT: buffer_load_dword v8, off, s[28:31], 0 offset:36 ; 4-byte Folded Reload
-; NOOPT-NEXT: s_waitcnt expcnt(6)
; NOOPT-NEXT: buffer_load_dword v9, off, s[28:31], 0 offset:40 ; 4-byte Folded Reload
-; NOOPT-NEXT: s_waitcnt expcnt(5)
; NOOPT-NEXT: buffer_load_dword v10, off, s[28:31], 0 offset:44 ; 4-byte Folded Reload
-; NOOPT-NEXT: s_waitcnt expcnt(4)
; NOOPT-NEXT: buffer_load_dword v11, off, s[28:31], 0 offset:48 ; 4-byte Folded Reload
-; NOOPT-NEXT: s_waitcnt expcnt(3)
; NOOPT-NEXT: buffer_load_dword v12, off, s[28:31], 0 offset:52 ; 4-byte Folded Reload
-; NOOPT-NEXT: s_waitcnt expcnt(2)
; NOOPT-NEXT: buffer_load_dword v13, off, s[28:31], 0 offset:56 ; 4-byte Folded Reload
-; NOOPT-NEXT: s_waitcnt expcnt(1)
; NOOPT-NEXT: buffer_load_dword v14, off, s[28:31], 0 offset:60 ; 4-byte Folded Reload
-; NOOPT-NEXT: s_waitcnt expcnt(0)
; NOOPT-NEXT: buffer_load_dword v15, off, s[28:31], 0 offset:64 ; 4-byte Folded Reload
; NOOPT-NEXT: buffer_load_dword v16, off, s[28:31], 0 offset:72 ; 4-byte Folded Reload
; NOOPT-NEXT: buffer_load_dword v17, off, s[28:31], 0 offset:80 ; 4-byte Folded Reload
@@ -9009,27 +9002,26 @@ define amdgpu_kernel void @broken_phi_bb(i32 %arg, i32 %arg1) {
; GENERIC-LABEL: broken_phi_bb:
; GENERIC: ; %bb.0: ; %bb
; GENERIC-NEXT: s_load_dwordx2 s[0:1], s[4:5], 0x9
-; GENERIC-NEXT: s_mov_b32 s6, 8
+; GENERIC-NEXT: s_mov_b32 s4, 8
; GENERIC-NEXT: s_mov_b32 s3, 0xf000
; GENERIC-NEXT: s_mov_b32 s2, -1
; GENERIC-NEXT: s_branch .LBB26_2
; GENERIC-NEXT: .LBB26_1: ; %Flow
; GENERIC-NEXT: ; in Loop: Header=BB26_2 Depth=1
; GENERIC-NEXT: s_andn2_b64 vcc, exec, s[4:5]
+; GENERIC-NEXT: s_mov_b32 s4, s1
; GENERIC-NEXT: s_cbranch_vccz .LBB26_4
; GENERIC-NEXT: .LBB26_2: ; %bb2
; GENERIC-NEXT: ; =>This Inner Loop Header: Depth=1
; GENERIC-NEXT: s_waitcnt lgkmcnt(0)
-; GENERIC-NEXT: s_cmp_ge_i32 s6, s0
+; GENERIC-NEXT: s_cmp_ge_i32 s4, s0
; GENERIC-NEXT: s_mov_b64 s[4:5], -1
-; GENERIC-NEXT: ; implicit-def: $sgpr6
; GENERIC-NEXT: s_cbranch_scc1 .LBB26_1
; GENERIC-NEXT: ; %bb.3: ; %bb4
; GENERIC-NEXT: ; in Loop: Header=BB26_2 Depth=1
; GENERIC-NEXT: buffer_load_dword v0, off, s[0:3], 0 glc
; GENERIC-NEXT: s_waitcnt vmcnt(0)
; GENERIC-NEXT: s_mov_b64 s[4:5], 0
-; GENERIC-NEXT: s_mov_b32 s6, s1
; GENERIC-NEXT: s_branch .LBB26_1
; GENERIC-NEXT: .LBB26_4: ; %bb8
; GENERIC-NEXT: s_endpgm
@@ -9070,8 +9062,8 @@ define amdgpu_kernel void @broken_phi_bb(i32 %arg, i32 %arg1) {
; NOOPT-NEXT: s_waitcnt vmcnt(0)
; NOOPT-NEXT: v_cmp_ge_i32_e64 s[2:3], v0, s2
; NOOPT-NEXT: v_mov_b32_e32 v0, s4
-; NOOPT-NEXT: s_and_b64 vcc, exec, s[2:3]
; NOOPT-NEXT: buffer_store_dword v0, off, s[24:27], 0 offset:8 ; 4-byte Folded Spill
+; NOOPT-NEXT: s_and_b64 vcc, exec, s[2:3]
; NOOPT-NEXT: v_writelane_b32 v18, s0, 2
; NOOPT-NEXT: v_writelane_b32 v18, s1, 3
; NOOPT-NEXT: s_or_saveexec_b64 s[20:21], -1
@@ -9117,30 +9109,30 @@ define amdgpu_kernel void @broken_phi_bb(i32 %arg, i32 %arg1) {
; NOOPT-NEXT: v_mov_b32_e32 v13, s17
; NOOPT-NEXT: v_mov_b32_e32 v14, s18
; NOOPT-NEXT: v_mov_b32_e32 v15, s19
-; NOOPT-NEXT: v_mov_b32_e32 v16, s0
-; NOOPT-NEXT: buffer_store_dword v16, off, s[24:27], 0 offset:76 ; 4-byte Folded Spill
+; NOOPT-NEXT: buffer_store_dword v0, off, s[24:27], 0 offset:16 ; 4-byte Folded Spill
+; NOOPT-NEXT: buffer_store_dword v1, off, s[24:27], 0 offset:20 ; 4-byte Folded Spill
+; NOOPT-NEXT: buffer_store_dword v2, off, s[24:27], 0 offset:24 ; 4-byte Folded Spill
+; NOOPT-NEXT: buffer_store_dword v3, off, s[24:27], 0 offset:28 ; 4-byte Folded Spill
+; NOOPT-NEXT: buffer_store_dword v4, off, s[24:27], 0 offset:32 ; 4-byte Folded Spill
+; NOOPT-NEXT: buffer_store_dword v5, off, s[24:27], 0 offset:36 ; 4-byte Folded Spill
+; NOOPT-NEXT: buffer_store_dword v6, off, s[24:27], 0 offset:40 ; 4-byte Folded Spill
+; NOOPT-NEXT: buffer_store_dword v7, off, s[24:27], 0 offset:44 ; 4-byte Folded Spill
+; NOOPT-NEXT: buffer_store_dword v8, off, s[24:27], 0 offset:48 ; 4-byte Folded Spill
+; NOOPT-NEXT: buffer_store_dword v9, off, s[24:27], 0 offset:52 ; 4-byte Folded Spill
+; NOOPT-NEXT: buffer_store_dword v10, off, s[24:27], 0 offset:56 ; 4-byte Folded Spill
+; NOOPT-NEXT: buffer_store_dword v11, off, s[24:27], 0 offset:60 ; 4-byte Folded Spill
+; NOOPT-NEXT: buffer_store_dword v12, off, s[24:27], 0 offset:64 ; 4-byte Folded Spill
+; NOOPT-NEXT: buffer_store_dword v13, off, s[24:27], 0 offset:68 ; 4-byte Folded Spill
+; NOOPT-NEXT: buffer_store_dword v14, off, s[24:27], 0 offset:72 ; 4-byte Folded Spill
+; NOOPT-NEXT: buffer_store_dword v15, off, s[24:27], 0 offset:76 ; 4-byte Folded Spill
+; NOOPT-NEXT: v_mov_b32_e32 v0, s0
+; NOOPT-NEXT: buffer_store_dword v0, off, s[24:27], 0 offset:12 ; 4-byte Folded Spill
; NOOPT-NEXT: s_mov_b64 s[0:1], exec
; NOOPT-NEXT: v_writelane_b32 v18, s0, 4
; NOOPT-NEXT: v_writelane_b32 v18, s1, 5
; NOOPT-NEXT: s_or_saveexec_b64 s[20:21], -1
; NOOPT-NEXT: buffer_store_dword v18, off, s[24:27], 0 ; 4-byte Folded Spill
; NOOPT-NEXT: s_mov_b64 exec, s[20:21]
-; NOOPT-NEXT: buffer_store_dword v0, off, s[24:27], 0 offset:12 ; 4-byte Folded Spill
-; NOOPT-NEXT: buffer_store_dword v1, off, s[24:27], 0 offset:16 ; 4-byte Folded Spill
-; NOOPT-NEXT: buffer_store_dword v2, off, s[24:27], 0 offset:20 ; 4-byte Folded Spill
-; NOOPT-NEXT: buffer_store_dword v3, off, s[24:27], 0 offset:24 ; 4-byte Folded Spill
-; NOOPT-NEXT: buffer_store_dword v4, off, s[24:27], 0 offset:28 ; 4-byte Folded Spill
-; NOOPT-NEXT: buffer_store_dword v5, off, s[24:27], 0 offset:32 ; 4-byte Folded Spill
-; NOOPT-NEXT: buffer_store_dword v6, off, s[24:27], 0 offset:36 ; 4-byte Folded Spill
-; NOOPT-NEXT: buffer_store_dword v7, off, s[24:27], 0 offset:40 ; 4-byte Folded Spill
-; NOOPT-NEXT: buffer_store_dword v8, off, s[24:27], 0 offset:44 ; 4-byte Folded Spill
-; NOOPT-NEXT: buffer_store_dword v9, off, s[24:27], 0 offset:48 ; 4-byte Folded Spill
-; NOOPT-NEXT: buffer_store_dword v10, off, s[24:27], 0 offset:52 ; 4-byte Folded Spill
-; NOOPT-NEXT: buffer_store_dword v11, off, s[24:27], 0 offset:56 ; 4-byte Folded Spill
-; NOOPT-NEXT: buffer_store_dword v12, off, s[24:27], 0 offset:60 ; 4-byte Folded Spill
-; NOOPT-NEXT: buffer_store_dword v13, off, s[24:27], 0 offset:64 ; 4-byte Folded Spill
-; NOOPT-NEXT: buffer_store_dword v14, off, s[24:27], 0 offset:68 ; 4-byte Folded Spill
-; NOOPT-NEXT: buffer_store_dword v15, off, s[24:27], 0 offset:72 ; 4-byte Folded Spill
; NOOPT-NEXT: ; implicit-def: $sgpr0_sgpr1
; NOOPT-NEXT: .LBB26_3: ; Parent Loop BB26_1 Depth=1
; NOOPT-NEXT: ; => This Inner Loop Header: Depth=2
@@ -9151,30 +9143,23 @@ define amdgpu_kernel void @broken_phi_bb(i32 %arg, i32 %arg1) {
; NOOPT-NEXT: s_waitcnt vmcnt(0)
; NOOPT-NEXT: v_readlane_b32 s0, v18, 6
; NOOPT-NEXT: v_readlane_b32 s1, v18, 7
-; NOOPT-NEXT: buffer_load_dword v0, off, s[24:27], 0 offset:12 ; 4-byte Folded Reload
-; NOOPT-NEXT: buffer_load_dword v1, off, s[24:27], 0 offset:16 ; 4-byte Folded Reload
-; NOOPT-NEXT: buffer_load_dword v2, off, s[24:27], 0 offset:20 ; 4-byte Folded Reload
-; NOOPT-NEXT: buffer_load_dword v3, off, s[24:27], 0 offset:24 ; 4-byte Folded Reload
-; NOOPT-NEXT: buffer_load_dword v4, off, s[24:27], 0 offset:28 ; 4-byte Folded Reload
-; NOOPT-NEXT: buffer_load_dword v5, off, s[24:27], 0 offset:32 ; 4-byte Folded Reload
-; NOOPT-NEXT: buffer_load_dword v6, off, s[24:27], 0 offset:36 ; 4-byte Folded Reload
-; NOOPT-NEXT: buffer_load_dword v7, off, s[24:27], 0 offset:40 ; 4-byte Folded Reload
-; NOOPT-NEXT: buffer_load_dword v8, off, s[24:27], 0 offset:44 ; 4-byte Folded Reload
-; NOOPT-NEXT: s_waitcnt expcnt(6)
-; NOOPT-NEXT: buffer_load_dword v9, off, s[24:27], 0 offset:48 ; 4-byte Folded Reload
-; NOOPT-NEXT: s_waitcnt expcnt(5)
-; NOOPT-NEXT: buffer_load_dword v10, off, s[24:27], 0 offset:52 ; 4-byte Folded Reload
-; NOOPT-NEXT: s_waitcnt expcnt(4)
-; NOOPT-NEXT: buffer_load_dword v11, off, s[24:27], 0 offset:56 ; 4-byte Folded Reload
-; NOOPT-NEXT: s_waitcnt expcnt(3)
-; NOOPT-NEXT: buffer_load_dword v12, off, s[24:27], 0 offset:60 ; 4-byte Folded Reload
-; NOOPT-NEXT: s_waitcnt expcnt(2)
-; NOOPT-NEXT: buffer_load_dword v13, off, s[24:27], 0 offset:64 ; 4-byte Folded Reload
-; NOOPT-NEXT: s_waitcnt expcnt(1)
-; NOOPT-NEXT: buffer_load_dword v14, off, s[24:27], 0 offset:68 ; 4-byte Folded Reload
-; NOOPT-NEXT: s_waitcnt expcnt(0)
-; NOOPT-NEXT: buffer_load_dword v15, off, s[24:27], 0 offset:72 ; 4-byte Folded Reload
-; NOOPT-NEXT: buffer_load_dword v16, off, s[24:27], 0 offset:76 ; 4-byte Folded Reload
+; NOOPT-NEXT: buffer_load_dword v0, off, s[24:27], 0 offset:16 ; 4-byte Folded Reload
+; NOOPT-NEXT: buffer_load_dword v1, off, s[24:27], 0 offset:20 ; 4-byte Folded Reload
+; NOOPT-NEXT: buffer_load_dword v2, off, s[24:27], 0 offset:24 ; 4-byte Folded Reload
+; NOOPT-NEXT: buffer_load_dword v3, off, s[24:27], 0 offset:28 ; 4-byte Folded Reload
+; NOOPT-NEXT: buffer_load_dword v4, off, s[24:27], 0 offset:32 ; 4-byte Folded Reload
+; NOOPT-NEXT: buffer_load_dword v5, off, s[24:27], 0 offset:36 ; 4-byte Folded Reload
+; NOOPT-NEXT: buffer_load_dword v6, off, s[24:27], 0 offset:40 ; 4-byte Folded Reload
+; NOOPT-NEXT: buffer_load_dword v7, off, s[24:27], 0 offset:44 ; 4-byte Folded Reload
+; NOOPT-NEXT: buffer_load_dword v8, off, s[24:27], 0 offset:48 ; 4-byte Folded Reload
+; NOOPT-NEXT: buffer_load_dword v9, off, s[24:27], 0 offset:52 ; 4-byte Folded Reload
+; NOOPT-NEXT: buffer_load_dword v10, off, s[24:27], 0 offset:56 ; 4-byte Folded Reload
+; NOOPT-NEXT: buffer_load_dword v11, off, s[24:27], 0 offset:60 ; 4-byte Folded Reload
+; NOOPT-NEXT: buffer_load_dword v12, off, s[24:27], 0 offset:64 ; 4-byte Folded Reload
+; NOOPT-NEXT: buffer_load_dword v13, off, s[24:27], 0 offset:68 ; 4-byte Folded Reload
+; NOOPT-NEXT: buffer_load_dword v14, off, s[24:27], 0 offset:72 ; 4-byte Folded Reload
+; NOOPT-NEXT: buffer_load_dword v15, off, s[24:27], 0 offset:76 ; 4-byte Folded Reload
+; NOOPT-NEXT: buffer_load_dword v16, off, s[24:27], 0 offset:12 ; 4-byte Folded Reload
; NOOPT-NEXT: buffer_load_dword v17, off, s[24:27], 0 offset:80 ; 4-byte Folded Reload
; NOOPT-NEXT: s_waitcnt vmcnt(0)
; NOOPT-NEXT: v_readfirstlane_b32 s2, v17
@@ -9198,22 +9183,22 @@ define amdgpu_kernel void @broken_phi_bb(i32 %arg, i32 %arg1) {
; NOOPT-NEXT: buffer_store_dword v13, off, s[24:27], 0 offset:136 ; 4-byte Folded Spill
; NOOPT-NEXT: buffer_store_dword v14, off, s[24:27], 0 offset:140 ; 4-byte Folded Spill
; NOOPT-NEXT: buffer_store_dword v15, off, s[24:27], 0 offset:144 ; 4-byte Folded Spill
-; NOOPT-NEXT: buffer_store_dword v0, off, s[24:27], 0 offset:12 ; 4-byte Folded Spill
-; NOOPT-NEXT: buffer_store_dword v1, off, s[24:27], 0 offset:16 ; 4-byte Folded Spill
-; NOOPT-NEXT: buffer_store_dword v2, off, s[24:27], 0 offset:20 ; 4-byte Folded Spill
-; NOOPT-NEXT: buffer_store_dword v3, off, s[24:27], 0 offset:24 ; 4-byte Folded Spill
-; NOOPT-NEXT: buffer_store_dword v4, off, s[24:27], 0 offset:28 ; 4-byte Folded Spill
-; NOOPT-NEXT: buffer_store_dword v5, off, s[24:27], 0 offset:32 ; 4-byte Folded Spill
-; NOOPT-NEXT: buffer_store_dword v6, off, s[24:27], 0 offset:36 ; 4-byte Folded Spill
-; NOOPT-NEXT: buffer_store_dword v7, off, s[24:27], 0 offset:40 ; 4-byte Folded Spill
-; NOOPT-NEXT: buffer_store_dword v8, off, s[24:27], 0 offset:44 ; 4-byte Folded Spill
-; NOOPT-NEXT: buffer_store_dword v9, off, s[24:27], 0 offset:48 ; 4-byte Folded Spill
-; NOOPT-NEXT: buffer_store_dword v10, off, s[24:27], 0 offset:52 ; 4-byte Folded Spill
-; NOOPT-NEXT: buffer_store_dword v11, off, s[24:27], 0 offset:56 ; 4-byte Folded Spill
-; NOOPT-NEXT: buffer_store_dword v12, off, s[24:27], 0 offset:60 ; 4-byte Folded Spill
-; NOOPT-NEXT: buffer_store_dword v13, off, s[24:27], 0 offset:64 ; 4-byte Folded Spill
-; NOOPT-NEXT: buffer_store_dword v14, off, s[24:27], 0 offset:68 ; 4-byte Folded Spill
-; NOOPT-NEXT: buffer_store_dword v15, off, s[24:27], 0 offset:72 ; 4-byte Folded Spill
+; NOOPT-NEXT: buffer_store_dword v0, off, s[24:27], 0 offset:16 ; 4-byte Folded Spill
+; NOOPT-NEXT: buffer_store_dword v1, off, s[24:27], 0 offset:20 ; 4-byte Folded Spill
+; NOOPT-NEXT: buffer_store_dword v2, off, s[24:27], 0 offset:24 ; 4-byte Folded Spill
+; NOOPT-NEXT: buffer_store_dword v3, off, s[24:27], 0 offset:28 ; 4-byte Folded Spill
+; NOOPT-NEXT: buffer_store_dword v4, off, s[24:27], 0 offset:32 ; 4-byte Folded Spill
+; NOOPT-NEXT: buffer_store_dword v5, off, s[24:27], 0 offset:36 ; 4-byte Folded Spill
+; NOOPT-NEXT: buffer_store_dword v6, off, s[24:27], 0 offset:40 ; 4-byte Folded Spill
+; NOOPT-NEXT: buffer_store_dword v7, off, s[24:27], 0 offset:44 ; 4-byte Folded Spill
+; NOOPT-NEXT: buffer_store_dword v8, off, s[24:27], 0 offset:48 ; 4-byte Folded Spill
+; NOOPT-NEXT: buffer_store_dword v9, off, s[24:27], 0 offset:52 ; 4-byte Folded Spill
+; NOOPT-NEXT: buffer_store_dword v10, off, s[24:27], 0 offset:56 ; 4-byte Folded Spill
+; NOOPT-NEXT: buffer_store_dword v11, off, s[24:27], 0 offset:60 ; 4-byte Folded Spill
+; NOOPT-NEXT: buffer_store_dword v12, off, s[24:27], 0 offset:64 ; 4-byte Folded Spill
+; NOOPT-NEXT: buffer_store_dword v13, off, s[24:27], 0 offset:68 ; 4-byte Folded Spill
+; NOOPT-NEXT: buffer_store_dword v14, off, s[24:27], 0 offset:72 ; 4-byte Folded Spill
+; NOOPT-NEXT: buffer_store_dword v15, off, s[24:27], 0 offset:76 ; 4-byte Folded Spill
; NOOPT-NEXT: s_mov_b64 s[2:3], s[0:1]
; NOOPT-NEXT: v_writelane_b32 v18, s2, 6
; NOOPT-NEXT: v_writelane_b32 v18, s3, 7
@@ -9251,9 +9236,9 @@ define amdgpu_kernel void @broken_phi_bb(i32 %arg, i32 %arg1) {
; NOOPT-NEXT: buffer_load_dword v13, off, s[24:27], 0 offset:136 ; 4-byte Folded Reload
; NOOPT-NEXT: buffer_load_dword v14, off, s[24:27], 0 offset:140 ; 4-byte Folded Reload
; NOOPT-NEXT: buffer_load_dword v15, off, s[24:27], 0 offset:144 ; 4-byte Folded Reload
-; NOOPT-NEXT: s_mov_b64 s[0:1], 0
; NOOPT-NEXT: s_waitcnt vmcnt(14)
; NOOPT-NEXT: buffer_store_dword v0, off, s[24:27], 0 offset:8 ; 4-byte Folded Spill
+; NOOPT-NEXT: s_mov_b64 s[0:1], 0
; NOOPT-NEXT: v_writelane_b32 v18, s0, 2
; NOOPT-NEXT: v_writelane_b32 v18, s1, 3
; NOOPT-NEXT: s_or_saveexec_b64 s[20:21], -1
diff --git a/llvm/test/CodeGen/AMDGPU/move-to-valu-atomicrmw-system.ll b/llvm/test/CodeGen/AMDGPU/move-to-valu-atomicrmw-system.ll
index fab5d386446d3..044fb2abd5cc1 100644
--- a/llvm/test/CodeGen/AMDGPU/move-to-valu-atomicrmw-system.ll
+++ b/llvm/test/CodeGen/AMDGPU/move-to-valu-atomicrmw-system.ll
@@ -28,29 +28,29 @@ define amdgpu_kernel void @atomic_max_i32(ptr addrspace(1) %out, ptr addrspace(1
; GCN-NEXT: ; %bb.1: ; %atomic
; GCN-NEXT: s_mov_b32 s8, s10
; GCN-NEXT: s_mov_b32 s9, s10
-; GCN-NEXT: buffer_load_dword v4, v[1:2], s[8:11], 0 addr64 offset:400
+; GCN-NEXT: buffer_load_dword v3, v[1:2], s[8:11], 0 addr64 offset:400
; GCN-NEXT: s_load_dword s4, s[4:5], 0xf
; GCN-NEXT: s_mov_b64 s[2:3], 0
; GCN-NEXT: .LBB0_2: ; %atomicrmw.start
; GCN-NEXT: ; =>This Inner Loop Header: Depth=1
-; GCN-NEXT: s_waitcnt vmcnt(0) lgkmcnt(0)
-; GCN-NEXT: v_max_i32_e32 v3, s4, v4
-; GCN-NEXT: s_waitcnt expcnt(0)
-; GCN-NEXT: v_mov_b32_e32 v6, v4
+; GCN-NEXT: s_waitcnt vmcnt(0)
; GCN-NEXT: v_mov_b32_e32 v5, v3
-; GCN-NEXT: buffer_atomic_cmpswap v[5:6], v[1:2], s[8:11], 0 addr64 offset:400 glc
+; GCN-NEXT: s_waitcnt expcnt(0) lgkmcnt(0)
+; GCN-NEXT: v_max_i32_e32 v4, s4, v5
+; GCN-NEXT: v_mov_b32_e32 v3, v4
+; GCN-NEXT: v_mov_b32_e32 v4, v5
+; GCN-NEXT: buffer_atomic_cmpswap v[3:4], v[1:2], s[8:11], 0 addr64 offset:400 glc
; GCN-NEXT: s_waitcnt vmcnt(0)
; GCN-NEXT: buffer_wbinvl1
-; GCN-NEXT: v_cmp_eq_u32_e32 vcc, v5, v4
+; GCN-NEXT: v_cmp_eq_u32_e32 vcc, v3, v5
; GCN-NEXT: s_or_b64 s[2:3], vcc, s[2:3]
-; GCN-NEXT: v_mov_b32_e32 v4, v5
; GCN-NEXT: s_andn2_b64 exec, exec, s[2:3]
; GCN-NEXT: s_cbranch_execnz .LBB0_2
; GCN-NEXT: ; %bb.3: ; %atomicrmw.end
; GCN-NEXT: s_or_b64 exec, exec, s[2:3]
; GCN-NEXT: s_mov_b32 s3, 0xf000
; GCN-NEXT: s_mov_b32 s2, -1
-; GCN-NEXT: buffer_store_dword v5, off, s[0:3], 0
+; GCN-NEXT: buffer_store_dword v3, off, s[0:3], 0
; GCN-NEXT: .LBB0_4: ; %exit
; GCN-NEXT: s_endpgm
%tid = call i32 @llvm.amdgcn.workitem.id.x()
@@ -87,22 +87,22 @@ define amdgpu_kernel void @atomic_max_i32_noret(ptr addrspace(1) %out, ptr addrs
; GCN-NEXT: ; %bb.1: ; %atomic
; GCN-NEXT: s_mov_b32 s0, s2
; GCN-NEXT: s_mov_b32 s1, s2
-; GCN-NEXT: buffer_load_dword v4, v[1:2], s[0:3], 0 addr64 offset:400
+; GCN-NEXT: buffer_load_dword v3, v[1:2], s[0:3], 0 addr64 offset:400
; GCN-NEXT: s_load_dword s6, s[4:5], 0xf
; GCN-NEXT: s_mov_b64 s[4:5], 0
; GCN-NEXT: .LBB1_2: ; %atomicrmw.start
; GCN-NEXT: ; =>This Inner Loop Header: Depth=1
-; GCN-NEXT: s_waitcnt vmcnt(0) lgkmcnt(0)
-; GCN-NEXT: v_max_i32_e32 v3, s6, v4
-; GCN-NEXT: s_waitcnt expcnt(0)
-; GCN-NEXT: v_mov_b32_e32 v6, v4
+; GCN-NEXT: s_waitcnt vmcnt(0)
; GCN-NEXT: v_mov_b32_e32 v5, v3
-; GCN-NEXT: buffer_atomic_cmpswap v[5:6], v[1:2], s[0:3], 0 addr64 offset:400 glc
+; GCN-NEXT: s_waitcnt expcnt(0) lgkmcnt(0)
+; GCN-NEXT: v_max_i32_e32 v4, s6, v5
+; GCN-NEXT: v_mov_b32_e32 v3, v4
+; GCN-NEXT: v_mov_b32_e32 v4, v5
+; GCN-NEXT: buffer_atomic_cmpswap v[3:4], v[1:2], s[0:3], 0 addr64 offset:400 glc
; GCN-NEXT: s_waitcnt vmcnt(0)
; GCN-NEXT: buffer_wbinvl1
-; GCN-NEXT: v_cmp_eq_u32_e32 vcc, v5, v4
+; GCN-NEXT: v_cmp_eq_u32_e32 vcc, v3, v5
; GCN-NEXT: s_or_b64 s[4:5], vcc, s[4:5]
-; GCN-NEXT: v_mov_b32_e32 v4, v5
; GCN-NEXT: s_andn2_b64 exec, exec, s[4:5]
; GCN-NEXT: s_cbranch_execnz .LBB1_2
; GCN-NEXT: .LBB1_3: ; %exit
diff --git a/llvm/test/CodeGen/AMDGPU/mul.ll b/llvm/test/CodeGen/AMDGPU/mul.ll
index b5e7589cbd134..b1e6f47fdf2c7 100644
--- a/llvm/test/CodeGen/AMDGPU/mul.ll
+++ b/llvm/test/CodeGen/AMDGPU/mul.ll
@@ -2556,32 +2556,32 @@ define amdgpu_kernel void @mul64_in_branch(ptr addrspace(1) %out, ptr addrspace(
; EG-NEXT: Fetch clause starting at 12:
; EG-NEXT: VTX_READ_64 T0.XY, T0.X, 0, #1
; EG-NEXT: ALU clause starting at 14:
-; EG-NEXT: OR_INT T0.W, KC0[2].W, KC0[3].X,
-; EG-NEXT: MOV * T1.W, literal.x,
+; EG-NEXT: OR_INT T1.W, KC0[2].W, KC0[3].X,
+; EG-NEXT: MOV * T0.W, literal.x,
; EG-NEXT: 1(1.401298e-45), 0(0.000000e+00)
-; EG-NEXT: SETNE_INT * T0.W, PV.W, 0.0,
+; EG-NEXT: SETNE_INT * T1.W, PV.W, 0.0,
; EG-NEXT: PRED_SETNE_INT * ExecMask,PredicateBit (MASKED), PV.W, 0.0,
; EG-NEXT: ALU clause starting at 19:
-; EG-NEXT: MOV T0.W, KC0[2].W,
-; EG-NEXT: MOV * T1.W, KC0[3].Z,
+; EG-NEXT: MOV T1.W, KC0[2].W,
+; EG-NEXT: MOV * T0.W, KC0[3].Z,
; EG-NEXT: MOV T2.W, KC0[3].Y,
; EG-NEXT: MULLO_INT * T0.X, PV.W, PS,
-; EG-NEXT: MOV T1.W, KC0[3].X,
-; EG-NEXT: MULHI * T0.Y, T0.W, PV.W,
+; EG-NEXT: MOV T0.W, KC0[3].X,
+; EG-NEXT: MULHI * T0.Y, T1.W, PV.W,
; EG-NEXT: ADD_INT T3.W, PS, T0.X,
; EG-NEXT: MULLO_INT * T0.X, PV.W, T2.W,
; EG-NEXT: ADD_INT T0.Y, PV.W, PS,
-; EG-NEXT: MOV T1.W, literal.x,
-; EG-NEXT: MULLO_INT * T0.X, T0.W, T2.W,
+; EG-NEXT: MOV T0.W, literal.x,
+; EG-NEXT: MULLO_INT * T0.X, T1.W, T2.W,
; EG-NEXT: 0(0.000000e+00), 0(0.000000e+00)
; EG-NEXT: ALU clause starting at 31:
-; EG-NEXT: MOV T0.W, KC0[2].Y,
-; EG-NEXT: SETE_INT * T1.W, T1.W, 0.0,
+; EG-NEXT: MOV T1.W, KC0[2].Y,
+; EG-NEXT: SETE_INT * T0.W, T0.W, 0.0,
; EG-NEXT: PRED_SETE_INT * ExecMask,PredicateBit (MASKED), PS, 0.0,
; EG-NEXT: ALU clause starting at 34:
; EG-NEXT: MOV * T0.X, KC0[2].Z,
; EG-NEXT: ALU clause starting at 35:
-; EG-NEXT: LSHR * T1.X, T0.W, literal.x,
+; EG-NEXT: LSHR * T1.X, T1.W, literal.x,
; EG-NEXT: 2(2.802597e-45), 0(0.000000e+00)
entry:
%0 = icmp eq i64 %a, 0
diff --git a/llvm/test/CodeGen/AMDGPU/rem_i128.ll b/llvm/test/CodeGen/AMDGPU/rem_i128.ll
index 6512bee36e88b..d29a7a2dc5656 100644
--- a/llvm/test/CodeGen/AMDGPU/rem_i128.ll
+++ b/llvm/test/CodeGen/AMDGPU/rem_i128.ll
@@ -240,27 +240,31 @@ define i128 @v_srem_i128_vv(i128 %lhs, i128 %rhs) {
; GFX9-O0: ; %bb.0: ; %_udiv-special-cases
; GFX9-O0-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX9-O0-NEXT: s_xor_saveexec_b64 s[4:5], -1
-; GFX9-O0-NEXT: buffer_store_dword v30, off, s[0:3], s32 offset:344 ; 4-byte Folded Spill
+; GFX9-O0-NEXT: buffer_store_dword v29, off, s[0:3], s32 offset:348 ; 4-byte Folded Spill
; GFX9-O0-NEXT: s_mov_b64 exec, s[4:5]
-; GFX9-O0-NEXT: buffer_store_dword v7, off, s[0:3], s32 offset:116 ; 4-byte Folded Spill
+; GFX9-O0-NEXT: buffer_store_dword v7, off, s[0:3], s32 offset:120 ; 4-byte Folded Spill
; GFX9-O0-NEXT: v_mov_b32_e32 v17, v6
; GFX9-O0-NEXT: v_mov_b32_e32 v7, v4
; GFX9-O0-NEXT: v_mov_b32_e32 v11, v2
-; GFX9-O0-NEXT: v_mov_b32_e32 v2, v1
-; GFX9-O0-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:116 ; 4-byte Folded Reload
+; GFX9-O0-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:120 ; 4-byte Folded Reload
+; GFX9-O0-NEXT: s_nop 0
+; GFX9-O0-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:116 ; 4-byte Folded Spill
+; GFX9-O0-NEXT: v_mov_b32_e32 v1, v0
+; GFX9-O0-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:116 ; 4-byte Folded Reload
; GFX9-O0-NEXT: ; implicit-def: $sgpr4
; GFX9-O0-NEXT: ; implicit-def: $sgpr4
; GFX9-O0-NEXT: ; kill: def $vgpr17 killed $vgpr17 def $vgpr17_vgpr18 killed $exec
-; GFX9-O0-NEXT: s_waitcnt vmcnt(0)
-; GFX9-O0-NEXT: v_mov_b32_e32 v18, v1
+; GFX9-O0-NEXT: s_waitcnt vmcnt(2)
+; GFX9-O0-NEXT: v_mov_b32_e32 v18, v2
; GFX9-O0-NEXT: ; implicit-def: $sgpr4
; GFX9-O0-NEXT: ; implicit-def: $sgpr4
; GFX9-O0-NEXT: ; kill: def $vgpr7 killed $vgpr7 def $vgpr7_vgpr8 killed $exec
; GFX9-O0-NEXT: v_mov_b32_e32 v8, v5
; GFX9-O0-NEXT: ; implicit-def: $sgpr4
; GFX9-O0-NEXT: ; implicit-def: $sgpr4
-; GFX9-O0-NEXT: ; kill: def $vgpr0 killed $vgpr0 def $vgpr0_vgpr1 killed $exec
-; GFX9-O0-NEXT: v_mov_b32_e32 v1, v2
+; GFX9-O0-NEXT: ; kill: def $vgpr1 killed $vgpr1 def $vgpr1_vgpr2 killed $exec
+; GFX9-O0-NEXT: s_waitcnt vmcnt(0)
+; GFX9-O0-NEXT: v_mov_b32_e32 v2, v0
; GFX9-O0-NEXT: ; implicit-def: $sgpr4
; GFX9-O0-NEXT: ; implicit-def: $sgpr4
; GFX9-O0-NEXT: ; kill: def $vgpr11 killed $vgpr11 def $vgpr11_vgpr12 killed $exec
@@ -270,34 +274,34 @@ define i128 @v_srem_i128_vv(i128 %lhs, i128 %rhs) {
; GFX9-O0-NEXT: ; implicit-def: $sgpr4_sgpr5
; GFX9-O0-NEXT: ; implicit-def: $sgpr4_sgpr5
; GFX9-O0-NEXT: s_mov_b32 s4, 63
-; GFX9-O0-NEXT: v_mov_b32_e32 v2, v11
-; GFX9-O0-NEXT: v_mov_b32_e32 v3, v12
-; GFX9-O0-NEXT: v_ashrrev_i64 v[2:3], s4, v[2:3]
-; GFX9-O0-NEXT: buffer_store_dword v2, off, s[0:3], s32 offset:108 ; 4-byte Folded Spill
+; GFX9-O0-NEXT: v_mov_b32_e32 v3, v11
+; GFX9-O0-NEXT: v_mov_b32_e32 v4, v12
+; GFX9-O0-NEXT: v_ashrrev_i64 v[3:4], s4, v[3:4]
+; GFX9-O0-NEXT: buffer_store_dword v3, off, s[0:3], s32 offset:108 ; 4-byte Folded Spill
; GFX9-O0-NEXT: s_nop 0
-; GFX9-O0-NEXT: buffer_store_dword v3, off, s[0:3], s32 offset:112 ; 4-byte Folded Spill
-; GFX9-O0-NEXT: buffer_store_dword v2, off, s[0:3], s32 offset:100 ; 4-byte Folded Spill
+; GFX9-O0-NEXT: buffer_store_dword v4, off, s[0:3], s32 offset:112 ; 4-byte Folded Spill
+; GFX9-O0-NEXT: buffer_store_dword v3, off, s[0:3], s32 offset:100 ; 4-byte Folded Spill
; GFX9-O0-NEXT: s_nop 0
-; GFX9-O0-NEXT: buffer_store_dword v3, off, s[0:3], s32 offset:104 ; 4-byte Folded Spill
-; GFX9-O0-NEXT: v_mov_b32_e32 v2, v0
-; GFX9-O0-NEXT: v_mov_b32_e32 v3, v1
-; GFX9-O0-NEXT: v_mov_b32_e32 v0, v11
-; GFX9-O0-NEXT: v_mov_b32_e32 v1, v12
+; GFX9-O0-NEXT: buffer_store_dword v4, off, s[0:3], s32 offset:104 ; 4-byte Folded Spill
+; GFX9-O0-NEXT: v_mov_b32_e32 v0, v1
+; GFX9-O0-NEXT: v_mov_b32_e32 v1, v2
+; GFX9-O0-NEXT: v_mov_b32_e32 v2, v11
+; GFX9-O0-NEXT: v_mov_b32_e32 v3, v12
; GFX9-O0-NEXT: s_mov_b64 s[6:7], 0
-; GFX9-O0-NEXT: ; implicit-def: $vgpr30 : SGPR spill to VGPR lane
-; GFX9-O0-NEXT: v_writelane_b32 v30, s6, 0
-; GFX9-O0-NEXT: v_writelane_b32 v30, s7, 1
+; GFX9-O0-NEXT: ; implicit-def: $vgpr29 : SGPR spill to VGPR lane
+; GFX9-O0-NEXT: v_writelane_b32 v29, s6, 0
+; GFX9-O0-NEXT: v_writelane_b32 v29, s7, 1
; GFX9-O0-NEXT: s_mov_b32 s10, s6
-; GFX9-O0-NEXT: v_writelane_b32 v30, s10, 2
+; GFX9-O0-NEXT: v_writelane_b32 v29, s10, 2
; GFX9-O0-NEXT: s_mov_b32 s11, s7
-; GFX9-O0-NEXT: v_writelane_b32 v30, s11, 3
-; GFX9-O0-NEXT: v_sub_co_u32_e32 v9, vcc, s10, v2
+; GFX9-O0-NEXT: v_writelane_b32 v29, s11, 3
+; GFX9-O0-NEXT: v_sub_co_u32_e32 v9, vcc, s10, v0
; GFX9-O0-NEXT: v_mov_b32_e32 v4, s11
-; GFX9-O0-NEXT: v_subb_co_u32_e32 v5, vcc, v4, v3, vcc
+; GFX9-O0-NEXT: v_subb_co_u32_e32 v5, vcc, v4, v1, vcc
; GFX9-O0-NEXT: v_mov_b32_e32 v4, s10
-; GFX9-O0-NEXT: v_subb_co_u32_e32 v4, vcc, v4, v0, vcc
+; GFX9-O0-NEXT: v_subb_co_u32_e32 v4, vcc, v4, v2, vcc
; GFX9-O0-NEXT: v_mov_b32_e32 v6, s11
-; GFX9-O0-NEXT: v_subb_co_u32_e32 v6, vcc, v6, v1, vcc
+; GFX9-O0-NEXT: v_subb_co_u32_e32 v6, vcc, v6, v3, vcc
; GFX9-O0-NEXT: ; implicit-def: $sgpr4
; GFX9-O0-NEXT: ; implicit-def: $sgpr4
; GFX9-O0-NEXT: ; kill: def $vgpr9 killed $vgpr9 def $vgpr9_vgpr10 killed $exec
@@ -305,25 +309,25 @@ define i128 @v_srem_i128_vv(i128 %lhs, i128 %rhs) {
; GFX9-O0-NEXT: v_mov_b32_e32 v5, v10
; GFX9-O0-NEXT: s_mov_b64 s[4:5], s[6:7]
; GFX9-O0-NEXT: v_cmp_lt_i64_e64 s[4:5], v[11:12], s[4:5]
-; GFX9-O0-NEXT: v_cndmask_b32_e64 v3, v3, v5, s[4:5]
+; GFX9-O0-NEXT: v_cndmask_b32_e64 v1, v1, v5, s[4:5]
; GFX9-O0-NEXT: v_mov_b32_e32 v5, v9
-; GFX9-O0-NEXT: v_cndmask_b32_e64 v2, v2, v5, s[4:5]
+; GFX9-O0-NEXT: v_cndmask_b32_e64 v0, v0, v5, s[4:5]
; GFX9-O0-NEXT: ; implicit-def: $sgpr8
; GFX9-O0-NEXT: ; implicit-def: $sgpr8
-; GFX9-O0-NEXT: v_mov_b32_e32 v15, v2
-; GFX9-O0-NEXT: v_mov_b32_e32 v16, v3
+; GFX9-O0-NEXT: v_mov_b32_e32 v15, v0
+; GFX9-O0-NEXT: v_mov_b32_e32 v16, v1
; GFX9-O0-NEXT: ; implicit-def: $sgpr8
; GFX9-O0-NEXT: ; implicit-def: $sgpr8
; GFX9-O0-NEXT: ; kill: def $vgpr4 killed $vgpr4 def $vgpr4_vgpr5 killed $exec
; GFX9-O0-NEXT: v_mov_b32_e32 v5, v6
; GFX9-O0-NEXT: v_mov_b32_e32 v6, v5
-; GFX9-O0-NEXT: v_cndmask_b32_e64 v1, v1, v6, s[4:5]
+; GFX9-O0-NEXT: v_cndmask_b32_e64 v3, v3, v6, s[4:5]
; GFX9-O0-NEXT: ; kill: def $vgpr4 killed $vgpr4 killed $vgpr4_vgpr5 killed $exec
-; GFX9-O0-NEXT: v_cndmask_b32_e64 v0, v0, v4, s[4:5]
+; GFX9-O0-NEXT: v_cndmask_b32_e64 v2, v2, v4, s[4:5]
; GFX9-O0-NEXT: ; implicit-def: $sgpr4
; GFX9-O0-NEXT: ; implicit-def: $sgpr4
-; GFX9-O0-NEXT: v_mov_b32_e32 v4, v0
-; GFX9-O0-NEXT: v_mov_b32_e32 v5, v1
+; GFX9-O0-NEXT: v_mov_b32_e32 v4, v2
+; GFX9-O0-NEXT: v_mov_b32_e32 v5, v3
; GFX9-O0-NEXT: v_mov_b32_e32 v6, v7
; GFX9-O0-NEXT: v_mov_b32_e32 v9, v8
; GFX9-O0-NEXT: v_mov_b32_e32 v7, v17
@@ -462,18 +466,18 @@ define i128 @v_srem_i128_vv(i128 %lhs, i128 %rhs) {
; GFX9-O0-NEXT: v_mov_b32_e32 v9, v6
; GFX9-O0-NEXT: s_mov_b64 s[8:9], s[6:7]
; GFX9-O0-NEXT: v_cmp_ne_u64_e64 s[8:9], v[4:5], s[8:9]
-; GFX9-O0-NEXT: v_ffbh_u32_e64 v4, v0
+; GFX9-O0-NEXT: v_ffbh_u32_e64 v4, v2
; GFX9-O0-NEXT: v_add_u32_e64 v4, v4, s13
-; GFX9-O0-NEXT: v_ffbh_u32_e64 v5, v1
+; GFX9-O0-NEXT: v_ffbh_u32_e64 v5, v3
; GFX9-O0-NEXT: v_min_u32_e64 v5, v4, v5
; GFX9-O0-NEXT: ; implicit-def: $sgpr16
; GFX9-O0-NEXT: v_mov_b32_e32 v4, s12
; GFX9-O0-NEXT: ; kill: def $vgpr5 killed $vgpr5 def $vgpr5_vgpr6 killed $exec
; GFX9-O0-NEXT: v_mov_b32_e32 v6, v4
; GFX9-O0-NEXT: v_mov_b32_e32 v7, v6
-; GFX9-O0-NEXT: v_ffbh_u32_e64 v4, v2
+; GFX9-O0-NEXT: v_ffbh_u32_e64 v4, v0
; GFX9-O0-NEXT: v_add_u32_e64 v4, v4, s13
-; GFX9-O0-NEXT: v_ffbh_u32_e64 v10, v3
+; GFX9-O0-NEXT: v_ffbh_u32_e64 v10, v1
; GFX9-O0-NEXT: v_min_u32_e64 v11, v4, v10
; GFX9-O0-NEXT: ; implicit-def: $sgpr13
; GFX9-O0-NEXT: v_mov_b32_e32 v4, s12
@@ -552,33 +556,33 @@ define i128 @v_srem_i128_vv(i128 %lhs, i128 %rhs) {
; GFX9-O0-NEXT: v_mov_b32_e32 v5, v6
; GFX9-O0-NEXT: v_cmp_ne_u64_e64 s[6:7], v[4:5], s[6:7]
; GFX9-O0-NEXT: v_mov_b32_e32 v4, s11
-; GFX9-O0-NEXT: v_cndmask_b32_e64 v4, v1, v4, s[8:9]
-; GFX9-O0-NEXT: v_mov_b32_e32 v1, s10
-; GFX9-O0-NEXT: v_cndmask_b32_e64 v0, v0, v1, s[8:9]
-; GFX9-O0-NEXT: ; implicit-def: $sgpr12
-; GFX9-O0-NEXT: ; implicit-def: $sgpr12
-; GFX9-O0-NEXT: ; kill: def $vgpr0 killed $vgpr0 def $vgpr0_vgpr1 killed $exec
-; GFX9-O0-NEXT: v_mov_b32_e32 v1, v4
-; GFX9-O0-NEXT: v_mov_b32_e32 v4, s11
; GFX9-O0-NEXT: v_cndmask_b32_e64 v4, v3, v4, s[8:9]
; GFX9-O0-NEXT: v_mov_b32_e32 v3, s10
; GFX9-O0-NEXT: v_cndmask_b32_e64 v2, v2, v3, s[8:9]
-; GFX9-O0-NEXT: ; implicit-def: $sgpr8
-; GFX9-O0-NEXT: ; implicit-def: $sgpr8
+; GFX9-O0-NEXT: ; implicit-def: $sgpr12
+; GFX9-O0-NEXT: ; implicit-def: $sgpr12
; GFX9-O0-NEXT: ; kill: def $vgpr2 killed $vgpr2 def $vgpr2_vgpr3 killed $exec
; GFX9-O0-NEXT: v_mov_b32_e32 v3, v4
-; GFX9-O0-NEXT: s_and_b64 s[6:7], s[4:5], s[6:7]
+; GFX9-O0-NEXT: v_mov_b32_e32 v4, s11
+; GFX9-O0-NEXT: v_cndmask_b32_e64 v4, v1, v4, s[8:9]
+; GFX9-O0-NEXT: v_mov_b32_e32 v1, s10
+; GFX9-O0-NEXT: v_cndmask_b32_e64 v0, v0, v1, s[8:9]
+; GFX9-O0-NEXT: ; implicit-def: $sgpr8
+; GFX9-O0-NEXT: ; implicit-def: $sgpr8
+; GFX9-O0-NEXT: ; kill: def $vgpr0 killed $vgpr0 def $vgpr0_vgpr1 killed $exec
+; GFX9-O0-NEXT: v_mov_b32_e32 v1, v4
; GFX9-O0-NEXT: buffer_store_dword v2, off, s[0:3], s32 offset:12 ; 4-byte Folded Spill
; GFX9-O0-NEXT: s_nop 0
; GFX9-O0-NEXT: buffer_store_dword v3, off, s[0:3], s32 offset:16 ; 4-byte Folded Spill
; GFX9-O0-NEXT: buffer_store_dword v0, off, s[0:3], s32 offset:4 ; 4-byte Folded Spill
; GFX9-O0-NEXT: s_nop 0
; GFX9-O0-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:8 ; 4-byte Folded Spill
+; GFX9-O0-NEXT: s_and_b64 s[6:7], s[4:5], s[6:7]
; GFX9-O0-NEXT: s_mov_b64 s[4:5], exec
-; GFX9-O0-NEXT: v_writelane_b32 v30, s4, 4
-; GFX9-O0-NEXT: v_writelane_b32 v30, s5, 5
+; GFX9-O0-NEXT: v_writelane_b32 v29, s4, 4
+; GFX9-O0-NEXT: v_writelane_b32 v29, s5, 5
; GFX9-O0-NEXT: s_or_saveexec_b64 s[22:23], -1
-; GFX9-O0-NEXT: buffer_store_dword v30, off, s[0:3], s32 ; 4-byte Folded Spill
+; GFX9-O0-NEXT: buffer_store_dword v29, off, s[0:3], s32 ; 4-byte Folded Spill
; GFX9-O0-NEXT: s_mov_b64 exec, s[22:23]
; GFX9-O0-NEXT: s_and_b64 s[4:5], s[4:5], s[6:7]
; GFX9-O0-NEXT: s_mov_b64 exec, s[4:5]
@@ -586,67 +590,64 @@ define i128 @v_srem_i128_vv(i128 %lhs, i128 %rhs) {
; GFX9-O0-NEXT: s_branch .LBB0_8
; GFX9-O0-NEXT: .LBB0_1: ; %Flow
; GFX9-O0-NEXT: s_or_saveexec_b64 s[22:23], -1
-; GFX9-O0-NEXT: buffer_load_dword v30, off, s[0:3], s32 ; 4-byte Folded Reload
+; GFX9-O0-NEXT: buffer_load_dword v29, off, s[0:3], s32 ; 4-byte Folded Reload
; GFX9-O0-NEXT: s_mov_b64 exec, s[22:23]
; GFX9-O0-NEXT: s_waitcnt vmcnt(0)
-; GFX9-O0-NEXT: v_readlane_b32 s4, v30, 6
-; GFX9-O0-NEXT: v_readlane_b32 s5, v30, 7
+; GFX9-O0-NEXT: v_readlane_b32 s4, v29, 6
+; GFX9-O0-NEXT: v_readlane_b32 s5, v29, 7
; GFX9-O0-NEXT: s_or_b64 exec, exec, s[4:5]
; GFX9-O0-NEXT: ; %bb.2: ; %Flow
-; GFX9-O0-NEXT: buffer_load_dword v6, off, s[0:3], s32 offset:152 ; 4-byte Folded Reload
-; GFX9-O0-NEXT: buffer_load_dword v7, off, s[0:3], s32 offset:156 ; 4-byte Folded Reload
-; GFX9-O0-NEXT: buffer_load_dword v4, off, s[0:3], s32 offset:160 ; 4-byte Folded Reload
-; GFX9-O0-NEXT: buffer_load_dword v5, off, s[0:3], s32 offset:164 ; 4-byte Folded Reload
-; GFX9-O0-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:168 ; 4-byte Folded Reload
-; GFX9-O0-NEXT: buffer_load_dword v3, off, s[0:3], s32 offset:172 ; 4-byte Folded Reload
-; GFX9-O0-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:176 ; 4-byte Folded Reload
-; GFX9-O0-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:180 ; 4-byte Folded Reload
-; GFX9-O0-NEXT: s_waitcnt vmcnt(7)
-; GFX9-O0-NEXT: buffer_store_dword v6, off, s[0:3], s32 offset:144 ; 4-byte Folded Spill
-; GFX9-O0-NEXT: s_waitcnt vmcnt(7)
-; GFX9-O0-NEXT: buffer_store_dword v7, off, s[0:3], s32 offset:148 ; 4-byte Folded Spill
-; GFX9-O0-NEXT: s_waitcnt vmcnt(7)
-; GFX9-O0-NEXT: buffer_store_dword v4, off, s[0:3], s32 offset:136 ; 4-byte Folded Spill
-; GFX9-O0-NEXT: s_waitcnt vmcnt(7)
-; GFX9-O0-NEXT: buffer_store_dword v5, off, s[0:3], s32 offset:140 ; 4-byte Folded Spill
-; GFX9-O0-NEXT: s_waitcnt vmcnt(7)
-; GFX9-O0-NEXT: buffer_store_dword v2, off, s[0:3], s32 offset:128 ; 4-byte Folded Spill
-; GFX9-O0-NEXT: s_waitcnt vmcnt(7)
-; GFX9-O0-NEXT: buffer_store_dword v3, off, s[0:3], s32 offset:132 ; 4-byte Folded Spill
-; GFX9-O0-NEXT: s_waitcnt vmcnt(7)
-; GFX9-O0-NEXT: buffer_store_dword v0, off, s[0:3], s32 offset:120 ; 4-byte Folded Spill
-; GFX9-O0-NEXT: s_waitcnt vmcnt(7)
-; GFX9-O0-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:124 ; 4-byte Folded Spill
+; GFX9-O0-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:156 ; 4-byte Folded Reload
+; GFX9-O0-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:160 ; 4-byte Folded Reload
+; GFX9-O0-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:164 ; 4-byte Folded Reload
+; GFX9-O0-NEXT: buffer_load_dword v3, off, s[0:3], s32 offset:168 ; 4-byte Folded Reload
+; GFX9-O0-NEXT: buffer_load_dword v4, off, s[0:3], s32 offset:172 ; 4-byte Folded Reload
+; GFX9-O0-NEXT: buffer_load_dword v5, off, s[0:3], s32 offset:176 ; 4-byte Folded Reload
+; GFX9-O0-NEXT: buffer_load_dword v6, off, s[0:3], s32 offset:180 ; 4-byte Folded Reload
+; GFX9-O0-NEXT: buffer_load_dword v7, off, s[0:3], s32 offset:184 ; 4-byte Folded Reload
+; GFX9-O0-NEXT: s_waitcnt vmcnt(1)
+; GFX9-O0-NEXT: buffer_store_dword v6, off, s[0:3], s32 offset:148 ; 4-byte Folded Spill
+; GFX9-O0-NEXT: s_waitcnt vmcnt(1)
+; GFX9-O0-NEXT: buffer_store_dword v7, off, s[0:3], s32 offset:152 ; 4-byte Folded Spill
+; GFX9-O0-NEXT: buffer_store_dword v4, off, s[0:3], s32 offset:140 ; 4-byte Folded Spill
+; GFX9-O0-NEXT: s_nop 0
+; GFX9-O0-NEXT: buffer_store_dword v5, off, s[0:3], s32 offset:144 ; 4-byte Folded Spill
+; GFX9-O0-NEXT: buffer_store_dword v2, off, s[0:3], s32 offset:132 ; 4-byte Folded Spill
+; GFX9-O0-NEXT: s_nop 0
+; GFX9-O0-NEXT: buffer_store_dword v3, off, s[0:3], s32 offset:136 ; 4-byte Folded Spill
+; GFX9-O0-NEXT: buffer_store_dword v0, off, s[0:3], s32 offset:124 ; 4-byte Folded Spill
+; GFX9-O0-NEXT: s_nop 0
+; GFX9-O0-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:128 ; 4-byte Folded Spill
; GFX9-O0-NEXT: s_branch .LBB0_5
; GFX9-O0-NEXT: .LBB0_3: ; %Flow2
; GFX9-O0-NEXT: s_or_saveexec_b64 s[22:23], -1
-; GFX9-O0-NEXT: buffer_load_dword v30, off, s[0:3], s32 ; 4-byte Folded Reload
+; GFX9-O0-NEXT: buffer_load_dword v29, off, s[0:3], s32 ; 4-byte Folded Reload
; GFX9-O0-NEXT: s_mov_b64 exec, s[22:23]
; GFX9-O0-NEXT: s_waitcnt vmcnt(0)
-; GFX9-O0-NEXT: v_readlane_b32 s4, v30, 4
-; GFX9-O0-NEXT: v_readlane_b32 s5, v30, 5
+; GFX9-O0-NEXT: v_readlane_b32 s4, v29, 4
+; GFX9-O0-NEXT: v_readlane_b32 s5, v29, 5
; GFX9-O0-NEXT: s_or_b64 exec, exec, s[4:5]
-; GFX9-O0-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:12 ; 4-byte Folded Reload
-; GFX9-O0-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:16 ; 4-byte Folded Reload
-; GFX9-O0-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:4 ; 4-byte Folded Reload
-; GFX9-O0-NEXT: buffer_load_dword v3, off, s[0:3], s32 offset:8 ; 4-byte Folded Reload
+; GFX9-O0-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:4 ; 4-byte Folded Reload
+; GFX9-O0-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:8 ; 4-byte Folded Reload
+; GFX9-O0-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:12 ; 4-byte Folded Reload
+; GFX9-O0-NEXT: buffer_load_dword v3, off, s[0:3], s32 offset:16 ; 4-byte Folded Reload
; GFX9-O0-NEXT: s_waitcnt vmcnt(1)
-; GFX9-O0-NEXT: buffer_store_dword v2, off, s[0:3], s32 offset:192 ; 4-byte Folded Spill
+; GFX9-O0-NEXT: buffer_store_dword v2, off, s[0:3], s32 offset:196 ; 4-byte Folded Spill
; GFX9-O0-NEXT: s_waitcnt vmcnt(1)
-; GFX9-O0-NEXT: buffer_store_dword v3, off, s[0:3], s32 offset:196 ; 4-byte Folded Spill
-; GFX9-O0-NEXT: buffer_store_dword v0, off, s[0:3], s32 offset:184 ; 4-byte Folded Spill
+; GFX9-O0-NEXT: buffer_store_dword v3, off, s[0:3], s32 offset:200 ; 4-byte Folded Spill
+; GFX9-O0-NEXT: buffer_store_dword v0, off, s[0:3], s32 offset:188 ; 4-byte Folded Spill
; GFX9-O0-NEXT: s_nop 0
-; GFX9-O0-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:188 ; 4-byte Folded Spill
+; GFX9-O0-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:192 ; 4-byte Folded Spill
; GFX9-O0-NEXT: s_branch .LBB0_9
; GFX9-O0-NEXT: .LBB0_4: ; %udiv-loop-exit
-; GFX9-O0-NEXT: buffer_load_dword v5, off, s[0:3], s32 offset:200 ; 4-byte Folded Reload
-; GFX9-O0-NEXT: buffer_load_dword v6, off, s[0:3], s32 offset:204 ; 4-byte Folded Reload
-; GFX9-O0-NEXT: buffer_load_dword v7, off, s[0:3], s32 offset:208 ; 4-byte Folded Reload
-; GFX9-O0-NEXT: buffer_load_dword v8, off, s[0:3], s32 offset:212 ; 4-byte Folded Reload
-; GFX9-O0-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:216 ; 4-byte Folded Reload
-; GFX9-O0-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:220 ; 4-byte Folded Reload
-; GFX9-O0-NEXT: buffer_load_dword v9, off, s[0:3], s32 offset:224 ; 4-byte Folded Reload
-; GFX9-O0-NEXT: buffer_load_dword v10, off, s[0:3], s32 offset:228 ; 4-byte Folded Reload
+; GFX9-O0-NEXT: buffer_load_dword v5, off, s[0:3], s32 offset:204 ; 4-byte Folded Reload
+; GFX9-O0-NEXT: buffer_load_dword v6, off, s[0:3], s32 offset:208 ; 4-byte Folded Reload
+; GFX9-O0-NEXT: buffer_load_dword v7, off, s[0:3], s32 offset:212 ; 4-byte Folded Reload
+; GFX9-O0-NEXT: buffer_load_dword v8, off, s[0:3], s32 offset:216 ; 4-byte Folded Reload
+; GFX9-O0-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:220 ; 4-byte Folded Reload
+; GFX9-O0-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:224 ; 4-byte Folded Reload
+; GFX9-O0-NEXT: buffer_load_dword v9, off, s[0:3], s32 offset:228 ; 4-byte Folded Reload
+; GFX9-O0-NEXT: buffer_load_dword v10, off, s[0:3], s32 offset:232 ; 4-byte Folded Reload
; GFX9-O0-NEXT: s_mov_b32 s4, 1
; GFX9-O0-NEXT: s_waitcnt vmcnt(2)
; GFX9-O0-NEXT: v_lshlrev_b64 v[2:3], s4, v[0:1]
@@ -671,408 +672,408 @@ define i128 @v_srem_i128_vv(i128 %lhs, i128 %rhs) {
; GFX9-O0-NEXT: v_or_b32_e64 v2, v2, v3
; GFX9-O0-NEXT: ; kill: def $vgpr2 killed $vgpr2 def $vgpr2_vgpr3 killed $exec
; GFX9-O0-NEXT: v_mov_b32_e32 v3, v4
-; GFX9-O0-NEXT: buffer_store_dword v2, off, s[0:3], s32 offset:12 ; 4-byte Folded Spill
+; GFX9-O0-NEXT: buffer_store_dword v2, off, s[0:3], s32 offset:4 ; 4-byte Folded Spill
; GFX9-O0-NEXT: s_nop 0
-; GFX9-O0-NEXT: buffer_store_dword v3, off, s[0:3], s32 offset:16 ; 4-byte Folded Spill
-; GFX9-O0-NEXT: buffer_store_dword v0, off, s[0:3], s32 offset:4 ; 4-byte Folded Spill
+; GFX9-O0-NEXT: buffer_store_dword v3, off, s[0:3], s32 offset:8 ; 4-byte Folded Spill
+; GFX9-O0-NEXT: buffer_store_dword v0, off, s[0:3], s32 offset:12 ; 4-byte Folded Spill
; GFX9-O0-NEXT: s_nop 0
-; GFX9-O0-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:8 ; 4-byte Folded Spill
+; GFX9-O0-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:16 ; 4-byte Folded Spill
; GFX9-O0-NEXT: s_branch .LBB0_3
; GFX9-O0-NEXT: .LBB0_5: ; %Flow1
; GFX9-O0-NEXT: s_or_saveexec_b64 s[22:23], -1
-; GFX9-O0-NEXT: buffer_load_dword v30, off, s[0:3], s32 ; 4-byte Folded Reload
+; GFX9-O0-NEXT: buffer_load_dword v29, off, s[0:3], s32 ; 4-byte Folded Reload
; GFX9-O0-NEXT: s_mov_b64 exec, s[22:23]
; GFX9-O0-NEXT: s_waitcnt vmcnt(0)
-; GFX9-O0-NEXT: v_readlane_b32 s4, v30, 8
-; GFX9-O0-NEXT: v_readlane_b32 s5, v30, 9
+; GFX9-O0-NEXT: v_readlane_b32 s4, v29, 8
+; GFX9-O0-NEXT: v_readlane_b32 s5, v29, 9
; GFX9-O0-NEXT: s_or_b64 exec, exec, s[4:5]
-; GFX9-O0-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:144 ; 4-byte Folded Reload
-; GFX9-O0-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:148 ; 4-byte Folded Reload
-; GFX9-O0-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:136 ; 4-byte Folded Reload
-; GFX9-O0-NEXT: buffer_load_dword v3, off, s[0:3], s32 offset:140 ; 4-byte Folded Reload
-; GFX9-O0-NEXT: buffer_load_dword v4, off, s[0:3], s32 offset:128 ; 4-byte Folded Reload
-; GFX9-O0-NEXT: buffer_load_dword v5, off, s[0:3], s32 offset:132 ; 4-byte Folded Reload
-; GFX9-O0-NEXT: buffer_load_dword v6, off, s[0:3], s32 offset:120 ; 4-byte Folded Reload
-; GFX9-O0-NEXT: buffer_load_dword v7, off, s[0:3], s32 offset:124 ; 4-byte Folded Reload
+; GFX9-O0-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:124 ; 4-byte Folded Reload
+; GFX9-O0-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:128 ; 4-byte Folded Reload
+; GFX9-O0-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:132 ; 4-byte Folded Reload
+; GFX9-O0-NEXT: buffer_load_dword v3, off, s[0:3], s32 offset:136 ; 4-byte Folded Reload
+; GFX9-O0-NEXT: buffer_load_dword v4, off, s[0:3], s32 offset:140 ; 4-byte Folded Reload
+; GFX9-O0-NEXT: buffer_load_dword v5, off, s[0:3], s32 offset:144 ; 4-byte Folded Reload
+; GFX9-O0-NEXT: buffer_load_dword v6, off, s[0:3], s32 offset:148 ; 4-byte Folded Reload
+; GFX9-O0-NEXT: buffer_load_dword v7, off, s[0:3], s32 offset:152 ; 4-byte Folded Reload
; GFX9-O0-NEXT: s_waitcnt vmcnt(1)
-; GFX9-O0-NEXT: buffer_store_dword v6, off, s[0:3], s32 offset:208 ; 4-byte Folded Spill
+; GFX9-O0-NEXT: buffer_store_dword v6, off, s[0:3], s32 offset:212 ; 4-byte Folded Spill
; GFX9-O0-NEXT: s_waitcnt vmcnt(1)
-; GFX9-O0-NEXT: buffer_store_dword v7, off, s[0:3], s32 offset:212 ; 4-byte Folded Spill
-; GFX9-O0-NEXT: buffer_store_dword v4, off, s[0:3], s32 offset:200 ; 4-byte Folded Spill
+; GFX9-O0-NEXT: buffer_store_dword v7, off, s[0:3], s32 offset:216 ; 4-byte Folded Spill
+; GFX9-O0-NEXT: buffer_store_dword v4, off, s[0:3], s32 offset:204 ; 4-byte Folded Spill
; GFX9-O0-NEXT: s_nop 0
-; GFX9-O0-NEXT: buffer_store_dword v5, off, s[0:3], s32 offset:204 ; 4-byte Folded Spill
-; GFX9-O0-NEXT: buffer_store_dword v2, off, s[0:3], s32 offset:224 ; 4-byte Folded Spill
+; GFX9-O0-NEXT: buffer_store_dword v5, off, s[0:3], s32 offset:208 ; 4-byte Folded Spill
+; GFX9-O0-NEXT: buffer_store_dword v2, off, s[0:3], s32 offset:228 ; 4-byte Folded Spill
; GFX9-O0-NEXT: s_nop 0
-; GFX9-O0-NEXT: buffer_store_dword v3, off, s[0:3], s32 offset:228 ; 4-byte Folded Spill
-; GFX9-O0-NEXT: buffer_store_dword v0, off, s[0:3], s32 offset:216 ; 4-byte Folded Spill
+; GFX9-O0-NEXT: buffer_store_dword v3, off, s[0:3], s32 offset:232 ; 4-byte Folded Spill
+; GFX9-O0-NEXT: buffer_store_dword v0, off, s[0:3], s32 offset:220 ; 4-byte Folded Spill
; GFX9-O0-NEXT: s_nop 0
-; GFX9-O0-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:220 ; 4-byte Folded Spill
+; GFX9-O0-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:224 ; 4-byte Folded Spill
; GFX9-O0-NEXT: s_branch .LBB0_4
; GFX9-O0-NEXT: .LBB0_6: ; %udiv-do-while
; GFX9-O0-NEXT: ; =>This Inner Loop Header: Depth=1
; GFX9-O0-NEXT: s_or_saveexec_b64 s[22:23], -1
-; GFX9-O0-NEXT: buffer_load_dword v30, off, s[0:3], s32 ; 4-byte Folded Reload
+; GFX9-O0-NEXT: buffer_load_dword v29, off, s[0:3], s32 ; 4-byte Folded Reload
; GFX9-O0-NEXT: s_mov_b64 exec, s[22:23]
; GFX9-O0-NEXT: s_waitcnt vmcnt(0)
-; GFX9-O0-NEXT: v_readlane_b32 s6, v30, 10
-; GFX9-O0-NEXT: v_readlane_b32 s7, v30, 11
-; GFX9-O0-NEXT: buffer_load_dword v6, off, s[0:3], s32 offset:232 ; 4-byte Folded Reload
-; GFX9-O0-NEXT: buffer_load_dword v7, off, s[0:3], s32 offset:236 ; 4-byte Folded Reload
-; GFX9-O0-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:240 ; 4-byte Folded Reload
-; GFX9-O0-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:244 ; 4-byte Folded Reload
-; GFX9-O0-NEXT: buffer_load_dword v22, off, s[0:3], s32 offset:248 ; 4-byte Folded Reload
-; GFX9-O0-NEXT: buffer_load_dword v23, off, s[0:3], s32 offset:252 ; 4-byte Folded Reload
-; GFX9-O0-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:256 ; 4-byte Folded Reload
-; GFX9-O0-NEXT: buffer_load_dword v3, off, s[0:3], s32 offset:260 ; 4-byte Folded Reload
-; GFX9-O0-NEXT: buffer_load_dword v16, off, s[0:3], s32 offset:264 ; 4-byte Folded Reload
-; GFX9-O0-NEXT: buffer_load_dword v17, off, s[0:3], s32 offset:268 ; 4-byte Folded Reload
-; GFX9-O0-NEXT: buffer_load_dword v8, off, s[0:3], s32 offset:272 ; 4-byte Folded Reload
-; GFX9-O0-NEXT: buffer_load_dword v9, off, s[0:3], s32 offset:276 ; 4-byte Folded Reload
-; GFX9-O0-NEXT: buffer_load_dword v26, off, s[0:3], s32 offset:280 ; 4-byte Folded Reload
-; GFX9-O0-NEXT: buffer_load_dword v27, off, s[0:3], s32 offset:284 ; 4-byte Folded Reload
-; GFX9-O0-NEXT: buffer_load_dword v24, off, s[0:3], s32 offset:288 ; 4-byte Folded Reload
-; GFX9-O0-NEXT: buffer_load_dword v25, off, s[0:3], s32 offset:292 ; 4-byte Folded Reload
+; GFX9-O0-NEXT: v_readlane_b32 s6, v29, 10
+; GFX9-O0-NEXT: v_readlane_b32 s7, v29, 11
+; GFX9-O0-NEXT: buffer_load_dword v10, off, s[0:3], s32 offset:236 ; 4-byte Folded Reload
+; GFX9-O0-NEXT: buffer_load_dword v11, off, s[0:3], s32 offset:240 ; 4-byte Folded Reload
+; GFX9-O0-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:244 ; 4-byte Folded Reload
+; GFX9-O0-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:248 ; 4-byte Folded Reload
+; GFX9-O0-NEXT: buffer_load_dword v24, off, s[0:3], s32 offset:252 ; 4-byte Folded Reload
+; GFX9-O0-NEXT: buffer_load_dword v25, off, s[0:3], s32 offset:256 ; 4-byte Folded Reload
+; GFX9-O0-NEXT: buffer_load_dword v16, off, s[0:3], s32 offset:260 ; 4-byte Folded Reload
+; GFX9-O0-NEXT: buffer_load_dword v17, off, s[0:3], s32 offset:264 ; 4-byte Folded Reload
+; GFX9-O0-NEXT: buffer_load_dword v8, off, s[0:3], s32 offset:268 ; 4-byte Folded Reload
+; GFX9-O0-NEXT: buffer_load_dword v9, off, s[0:3], s32 offset:272 ; 4-byte Folded Reload
+; GFX9-O0-NEXT: buffer_load_dword v12, off, s[0:3], s32 offset:276 ; 4-byte Folded Reload
+; GFX9-O0-NEXT: buffer_load_dword v13, off, s[0:3], s32 offset:280 ; 4-byte Folded Reload
+; GFX9-O0-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:284 ; 4-byte Folded Reload
+; GFX9-O0-NEXT: buffer_load_dword v3, off, s[0:3], s32 offset:288 ; 4-byte Folded Reload
+; GFX9-O0-NEXT: buffer_load_dword v26, off, s[0:3], s32 offset:292 ; 4-byte Folded Reload
+; GFX9-O0-NEXT: buffer_load_dword v27, off, s[0:3], s32 offset:296 ; 4-byte Folded Reload
; GFX9-O0-NEXT: buffer_load_dword v18, off, s[0:3], s32 offset:52 ; 4-byte Folded Reload
; GFX9-O0-NEXT: buffer_load_dword v19, off, s[0:3], s32 offset:56 ; 4-byte Folded Reload
; GFX9-O0-NEXT: buffer_load_dword v20, off, s[0:3], s32 offset:60 ; 4-byte Folded Reload
; GFX9-O0-NEXT: buffer_load_dword v21, off, s[0:3], s32 offset:64 ; 4-byte Folded Reload
-; GFX9-O0-NEXT: buffer_load_dword v14, off, s[0:3], s32 offset:296 ; 4-byte Folded Reload
-; GFX9-O0-NEXT: buffer_load_dword v15, off, s[0:3], s32 offset:300 ; 4-byte Folded Reload
-; GFX9-O0-NEXT: buffer_load_dword v11, off, s[0:3], s32 offset:304 ; 4-byte Folded Reload
-; GFX9-O0-NEXT: buffer_load_dword v12, off, s[0:3], s32 offset:308 ; 4-byte Folded Reload
+; GFX9-O0-NEXT: buffer_load_dword v22, off, s[0:3], s32 offset:300 ; 4-byte Folded Reload
+; GFX9-O0-NEXT: buffer_load_dword v23, off, s[0:3], s32 offset:304 ; 4-byte Folded Reload
+; GFX9-O0-NEXT: buffer_load_dword v4, off, s[0:3], s32 offset:308 ; 4-byte Folded Reload
+; GFX9-O0-NEXT: buffer_load_dword v5, off, s[0:3], s32 offset:312 ; 4-byte Folded Reload
; GFX9-O0-NEXT: s_mov_b32 s4, 63
-; GFX9-O0-NEXT: s_waitcnt vmcnt(16)
-; GFX9-O0-NEXT: v_lshrrev_b64 v[28:29], s4, v[2:3]
-; GFX9-O0-NEXT: v_mov_b32_e32 v5, v29
+; GFX9-O0-NEXT: s_waitcnt vmcnt(10)
+; GFX9-O0-NEXT: v_lshrrev_b64 v[6:7], s4, v[2:3]
+; GFX9-O0-NEXT: v_mov_b32_e32 v15, v7
; GFX9-O0-NEXT: s_mov_b32 s5, 1
-; GFX9-O0-NEXT: v_lshlrev_b64 v[22:23], s5, v[22:23]
-; GFX9-O0-NEXT: v_mov_b32_e32 v4, v23
-; GFX9-O0-NEXT: v_or_b32_e64 v4, v4, v5
-; GFX9-O0-NEXT: v_mov_b32_e32 v10, v28
-; GFX9-O0-NEXT: v_mov_b32_e32 v5, v22
-; GFX9-O0-NEXT: v_or_b32_e64 v22, v5, v10
-; GFX9-O0-NEXT: ; kill: def $vgpr22 killed $vgpr22 def $vgpr22_vgpr23 killed $exec
-; GFX9-O0-NEXT: v_mov_b32_e32 v23, v4
-; GFX9-O0-NEXT: v_lshlrev_b64 v[28:29], s5, v[2:3]
-; GFX9-O0-NEXT: v_lshrrev_b64 v[4:5], s4, v[6:7]
-; GFX9-O0-NEXT: v_mov_b32_e32 v2, v29
-; GFX9-O0-NEXT: v_mov_b32_e32 v3, v5
+; GFX9-O0-NEXT: s_waitcnt vmcnt(8)
+; GFX9-O0-NEXT: v_lshlrev_b64 v[26:27], s5, v[26:27]
+; GFX9-O0-NEXT: v_mov_b32_e32 v14, v27
+; GFX9-O0-NEXT: v_or_b32_e64 v14, v14, v15
+; GFX9-O0-NEXT: v_mov_b32_e32 v7, v6
+; GFX9-O0-NEXT: v_mov_b32_e32 v6, v26
+; GFX9-O0-NEXT: v_or_b32_e64 v6, v6, v7
+; GFX9-O0-NEXT: ; kill: def $vgpr6 killed $vgpr6 def $vgpr6_vgpr7 killed $exec
+; GFX9-O0-NEXT: v_mov_b32_e32 v7, v14
+; GFX9-O0-NEXT: v_lshlrev_b64 v[26:27], s5, v[2:3]
+; GFX9-O0-NEXT: v_lshrrev_b64 v[14:15], s4, v[10:11]
+; GFX9-O0-NEXT: v_mov_b32_e32 v2, v27
+; GFX9-O0-NEXT: v_mov_b32_e32 v3, v15
; GFX9-O0-NEXT: v_or_b32_e64 v2, v2, v3
-; GFX9-O0-NEXT: v_mov_b32_e32 v3, v28
-; GFX9-O0-NEXT: ; kill: def $vgpr4 killed $vgpr4 killed $vgpr4_vgpr5 killed $exec
-; GFX9-O0-NEXT: v_or_b32_e64 v4, v3, v4
-; GFX9-O0-NEXT: ; kill: def $vgpr4 killed $vgpr4 def $vgpr4_vgpr5 killed $exec
-; GFX9-O0-NEXT: v_mov_b32_e32 v5, v2
+; GFX9-O0-NEXT: v_mov_b32_e32 v3, v26
+; GFX9-O0-NEXT: ; kill: def $vgpr14 killed $vgpr14 killed $vgpr14_vgpr15 killed $exec
+; GFX9-O0-NEXT: v_or_b32_e64 v14, v3, v14
+; GFX9-O0-NEXT: ; kill: def $vgpr14 killed $vgpr14 def $vgpr14_vgpr15 killed $exec
+; GFX9-O0-NEXT: v_mov_b32_e32 v15, v2
; GFX9-O0-NEXT: v_lshlrev_b64 v[2:3], s5, v[0:1]
-; GFX9-O0-NEXT: v_lshlrev_b64 v[28:29], s5, v[6:7]
+; GFX9-O0-NEXT: v_lshlrev_b64 v[26:27], s5, v[10:11]
; GFX9-O0-NEXT: v_lshrrev_b64 v[0:1], s4, v[0:1]
-; GFX9-O0-NEXT: v_mov_b32_e32 v7, v1
-; GFX9-O0-NEXT: v_mov_b32_e32 v6, v29
-; GFX9-O0-NEXT: s_waitcnt vmcnt(10)
+; GFX9-O0-NEXT: v_mov_b32_e32 v11, v1
; GFX9-O0-NEXT: v_mov_b32_e32 v10, v27
-; GFX9-O0-NEXT: v_or3_b32 v6, v6, v7, v10
+; GFX9-O0-NEXT: v_mov_b32_e32 v28, v25
+; GFX9-O0-NEXT: v_or3_b32 v10, v10, v11, v28
; GFX9-O0-NEXT: v_mov_b32_e32 v1, v0
-; GFX9-O0-NEXT: v_mov_b32_e32 v0, v28
-; GFX9-O0-NEXT: v_mov_b32_e32 v7, v26
-; GFX9-O0-NEXT: v_or3_b32 v0, v0, v1, v7
+; GFX9-O0-NEXT: v_mov_b32_e32 v0, v26
+; GFX9-O0-NEXT: v_mov_b32_e32 v11, v24
+; GFX9-O0-NEXT: v_or3_b32 v0, v0, v1, v11
; GFX9-O0-NEXT: ; kill: def $vgpr0 killed $vgpr0 def $vgpr0_vgpr1 killed $exec
-; GFX9-O0-NEXT: v_mov_b32_e32 v1, v6
-; GFX9-O0-NEXT: v_mov_b32_e32 v7, v3
-; GFX9-O0-NEXT: s_waitcnt vmcnt(8)
-; GFX9-O0-NEXT: v_mov_b32_e32 v6, v25
-; GFX9-O0-NEXT: v_or_b32_e64 v6, v6, v7
+; GFX9-O0-NEXT: v_mov_b32_e32 v1, v10
+; GFX9-O0-NEXT: v_mov_b32_e32 v11, v3
+; GFX9-O0-NEXT: v_mov_b32_e32 v10, v17
+; GFX9-O0-NEXT: v_or_b32_e64 v10, v10, v11
; GFX9-O0-NEXT: v_mov_b32_e32 v3, v2
-; GFX9-O0-NEXT: v_mov_b32_e32 v2, v24
+; GFX9-O0-NEXT: v_mov_b32_e32 v2, v16
; GFX9-O0-NEXT: v_or_b32_e64 v2, v2, v3
; GFX9-O0-NEXT: ; kill: def $vgpr2 killed $vgpr2 def $vgpr2_vgpr3 killed $exec
-; GFX9-O0-NEXT: v_mov_b32_e32 v3, v6
-; GFX9-O0-NEXT: v_mov_b32_e32 v6, v4
-; GFX9-O0-NEXT: v_mov_b32_e32 v10, v5
-; GFX9-O0-NEXT: v_mov_b32_e32 v4, v22
-; GFX9-O0-NEXT: v_mov_b32_e32 v5, v23
+; GFX9-O0-NEXT: v_mov_b32_e32 v3, v10
+; GFX9-O0-NEXT: v_mov_b32_e32 v10, v14
+; GFX9-O0-NEXT: v_mov_b32_e32 v16, v15
+; GFX9-O0-NEXT: v_mov_b32_e32 v14, v6
+; GFX9-O0-NEXT: v_mov_b32_e32 v11, v7
; GFX9-O0-NEXT: s_waitcnt vmcnt(1)
-; GFX9-O0-NEXT: v_mov_b32_e32 v13, v11
-; GFX9-O0-NEXT: v_mov_b32_e32 v11, v14
-; GFX9-O0-NEXT: v_mov_b32_e32 v7, v15
-; GFX9-O0-NEXT: v_sub_co_u32_e32 v13, vcc, v13, v6
+; GFX9-O0-NEXT: v_mov_b32_e32 v7, v4
; GFX9-O0-NEXT: s_waitcnt vmcnt(0)
-; GFX9-O0-NEXT: v_subb_co_u32_e32 v12, vcc, v12, v10, vcc
-; GFX9-O0-NEXT: v_subb_co_u32_e32 v11, vcc, v11, v4, vcc
-; GFX9-O0-NEXT: v_subb_co_u32_e32 v7, vcc, v7, v5, vcc
+; GFX9-O0-NEXT: v_mov_b32_e32 v6, v5
+; GFX9-O0-NEXT: v_mov_b32_e32 v4, v22
+; GFX9-O0-NEXT: v_mov_b32_e32 v5, v23
+; GFX9-O0-NEXT: v_sub_co_u32_e32 v7, vcc, v7, v10
+; GFX9-O0-NEXT: v_subb_co_u32_e32 v6, vcc, v6, v16, vcc
+; GFX9-O0-NEXT: v_subb_co_u32_e32 v4, vcc, v4, v14, vcc
+; GFX9-O0-NEXT: v_subb_co_u32_e32 v6, vcc, v5, v11, vcc
; GFX9-O0-NEXT: ; implicit-def: $sgpr5
; GFX9-O0-NEXT: ; implicit-def: $sgpr5
-; GFX9-O0-NEXT: ; kill: def $vgpr11 killed $vgpr11 def $vgpr11_vgpr12 killed $exec
-; GFX9-O0-NEXT: v_mov_b32_e32 v12, v7
-; GFX9-O0-NEXT: v_ashrrev_i64 v[13:14], s4, v[11:12]
-; GFX9-O0-NEXT: v_mov_b32_e32 v7, v14
+; GFX9-O0-NEXT: ; kill: def $vgpr4 killed $vgpr4 def $vgpr4_vgpr5 killed $exec
+; GFX9-O0-NEXT: v_mov_b32_e32 v5, v6
+; GFX9-O0-NEXT: v_ashrrev_i64 v[5:6], s4, v[4:5]
+; GFX9-O0-NEXT: v_mov_b32_e32 v15, v6
; GFX9-O0-NEXT: s_mov_b64 s[4:5], 1
; GFX9-O0-NEXT: s_mov_b32 s8, s5
-; GFX9-O0-NEXT: v_and_b32_e64 v12, v7, s8
-; GFX9-O0-NEXT: v_mov_b32_e32 v11, v13
+; GFX9-O0-NEXT: v_and_b32_e64 v4, v15, s8
+; GFX9-O0-NEXT: v_mov_b32_e32 v17, v5
; GFX9-O0-NEXT: ; kill: def $sgpr4 killed $sgpr4 killed $sgpr4_sgpr5
-; GFX9-O0-NEXT: v_and_b32_e64 v14, v11, s4
-; GFX9-O0-NEXT: ; kill: def $vgpr14 killed $vgpr14 def $vgpr14_vgpr15 killed $exec
-; GFX9-O0-NEXT: v_mov_b32_e32 v15, v12
-; GFX9-O0-NEXT: v_mov_b32_e32 v12, 0
-; GFX9-O0-NEXT: v_mov_b32_e32 v13, 0
+; GFX9-O0-NEXT: v_and_b32_e64 v6, v17, s4
+; GFX9-O0-NEXT: ; kill: def $vgpr6 killed $vgpr6 def $vgpr6_vgpr7 killed $exec
+; GFX9-O0-NEXT: v_mov_b32_e32 v7, v4
+; GFX9-O0-NEXT: v_mov_b32_e32 v4, 0
+; GFX9-O0-NEXT: v_mov_b32_e32 v5, 0
; GFX9-O0-NEXT: v_mov_b32_e32 v22, v21
-; GFX9-O0-NEXT: v_and_b32_e64 v22, v7, v22
-; GFX9-O0-NEXT: v_and_b32_e64 v20, v11, v20
+; GFX9-O0-NEXT: v_and_b32_e64 v22, v15, v22
+; GFX9-O0-NEXT: v_and_b32_e64 v20, v17, v20
; GFX9-O0-NEXT: ; kill: def $vgpr20 killed $vgpr20 def $vgpr20_vgpr21 killed $exec
; GFX9-O0-NEXT: v_mov_b32_e32 v21, v22
; GFX9-O0-NEXT: v_mov_b32_e32 v22, v19
-; GFX9-O0-NEXT: v_and_b32_e64 v7, v7, v22
-; GFX9-O0-NEXT: v_and_b32_e64 v22, v11, v18
-; GFX9-O0-NEXT: ; kill: def $vgpr22 killed $vgpr22 def $vgpr22_vgpr23 killed $exec
-; GFX9-O0-NEXT: v_mov_b32_e32 v23, v7
-; GFX9-O0-NEXT: v_mov_b32_e32 v19, v22
-; GFX9-O0-NEXT: v_mov_b32_e32 v18, v23
-; GFX9-O0-NEXT: v_mov_b32_e32 v11, v20
-; GFX9-O0-NEXT: v_mov_b32_e32 v7, v21
-; GFX9-O0-NEXT: v_sub_co_u32_e32 v6, vcc, v6, v19
-; GFX9-O0-NEXT: v_subb_co_u32_e32 v10, vcc, v10, v18, vcc
-; GFX9-O0-NEXT: v_subb_co_u32_e32 v4, vcc, v4, v11, vcc
-; GFX9-O0-NEXT: v_subb_co_u32_e32 v7, vcc, v5, v7, vcc
+; GFX9-O0-NEXT: v_and_b32_e64 v15, v15, v22
+; GFX9-O0-NEXT: v_and_b32_e64 v17, v17, v18
+; GFX9-O0-NEXT: ; kill: def $vgpr17 killed $vgpr17 def $vgpr17_vgpr18 killed $exec
+; GFX9-O0-NEXT: v_mov_b32_e32 v18, v15
+; GFX9-O0-NEXT: v_mov_b32_e32 v19, v17
+; GFX9-O0-NEXT: ; kill: def $vgpr18 killed $vgpr18 killed $vgpr17_vgpr18 killed $exec
+; GFX9-O0-NEXT: v_mov_b32_e32 v17, v20
+; GFX9-O0-NEXT: v_mov_b32_e32 v15, v21
+; GFX9-O0-NEXT: v_sub_co_u32_e32 v10, vcc, v10, v19
+; GFX9-O0-NEXT: v_subb_co_u32_e32 v16, vcc, v16, v18, vcc
+; GFX9-O0-NEXT: v_subb_co_u32_e32 v14, vcc, v14, v17, vcc
+; GFX9-O0-NEXT: v_subb_co_u32_e32 v11, vcc, v11, v15, vcc
; GFX9-O0-NEXT: ; implicit-def: $sgpr4
; GFX9-O0-NEXT: ; implicit-def: $sgpr4
-; GFX9-O0-NEXT: ; kill: def $vgpr4 killed $vgpr4 def $vgpr4_vgpr5 killed $exec
-; GFX9-O0-NEXT: v_mov_b32_e32 v5, v7
+; GFX9-O0-NEXT: ; kill: def $vgpr14 killed $vgpr14 def $vgpr14_vgpr15 killed $exec
+; GFX9-O0-NEXT: v_mov_b32_e32 v15, v11
; GFX9-O0-NEXT: ; implicit-def: $sgpr4
; GFX9-O0-NEXT: ; implicit-def: $sgpr4
-; GFX9-O0-NEXT: ; kill: def $vgpr6 killed $vgpr6 def $vgpr6_vgpr7 killed $exec
-; GFX9-O0-NEXT: v_mov_b32_e32 v7, v10
+; GFX9-O0-NEXT: ; kill: def $vgpr10 killed $vgpr10 def $vgpr10_vgpr11 killed $exec
+; GFX9-O0-NEXT: v_mov_b32_e32 v11, v16
+; GFX9-O0-NEXT: buffer_store_dword v14, off, s[0:3], s32 offset:292 ; 4-byte Folded Spill
+; GFX9-O0-NEXT: s_nop 0
+; GFX9-O0-NEXT: buffer_store_dword v15, off, s[0:3], s32 offset:296 ; 4-byte Folded Spill
+; GFX9-O0-NEXT: buffer_store_dword v10, off, s[0:3], s32 offset:284 ; 4-byte Folded Spill
+; GFX9-O0-NEXT: s_nop 0
+; GFX9-O0-NEXT: buffer_store_dword v11, off, s[0:3], s32 offset:288 ; 4-byte Folded Spill
; GFX9-O0-NEXT: v_mov_b32_e32 v11, v8
; GFX9-O0-NEXT: ; kill: def $vgpr9 killed $vgpr9 killed $vgpr8_vgpr9 killed $exec
; GFX9-O0-NEXT: s_mov_b64 s[8:9], -1
; GFX9-O0-NEXT: s_mov_b32 s5, s8
; GFX9-O0-NEXT: s_mov_b32 s4, s9
-; GFX9-O0-NEXT: v_mov_b32_e32 v10, v16
-; GFX9-O0-NEXT: v_mov_b32_e32 v8, v17
-; GFX9-O0-NEXT: v_mov_b32_e32 v16, s5
-; GFX9-O0-NEXT: v_add_co_u32_e32 v19, vcc, v11, v16
-; GFX9-O0-NEXT: v_mov_b32_e32 v11, s4
-; GFX9-O0-NEXT: v_addc_co_u32_e32 v9, vcc, v9, v11, vcc
-; GFX9-O0-NEXT: v_mov_b32_e32 v11, s5
-; GFX9-O0-NEXT: v_addc_co_u32_e32 v16, vcc, v10, v11, vcc
-; GFX9-O0-NEXT: v_mov_b32_e32 v10, s4
-; GFX9-O0-NEXT: v_addc_co_u32_e32 v8, vcc, v8, v10, vcc
+; GFX9-O0-NEXT: v_mov_b32_e32 v8, v12
+; GFX9-O0-NEXT: v_mov_b32_e32 v10, v13
+; GFX9-O0-NEXT: v_mov_b32_e32 v12, s5
+; GFX9-O0-NEXT: v_add_co_u32_e32 v11, vcc, v11, v12
+; GFX9-O0-NEXT: v_mov_b32_e32 v12, s4
+; GFX9-O0-NEXT: v_addc_co_u32_e32 v9, vcc, v9, v12, vcc
+; GFX9-O0-NEXT: v_mov_b32_e32 v12, s5
+; GFX9-O0-NEXT: v_addc_co_u32_e32 v8, vcc, v8, v12, vcc
+; GFX9-O0-NEXT: v_mov_b32_e32 v12, s4
+; GFX9-O0-NEXT: v_addc_co_u32_e32 v10, vcc, v10, v12, vcc
; GFX9-O0-NEXT: ; implicit-def: $sgpr4
; GFX9-O0-NEXT: ; implicit-def: $sgpr4
-; GFX9-O0-NEXT: ; kill: def $vgpr19 killed $vgpr19 def $vgpr19_vgpr20 killed $exec
-; GFX9-O0-NEXT: v_mov_b32_e32 v20, v9
+; GFX9-O0-NEXT: ; kill: def $vgpr11 killed $vgpr11 def $vgpr11_vgpr12 killed $exec
+; GFX9-O0-NEXT: v_mov_b32_e32 v12, v9
; GFX9-O0-NEXT: ; implicit-def: $sgpr4
; GFX9-O0-NEXT: ; implicit-def: $sgpr4
-; GFX9-O0-NEXT: ; kill: def $vgpr16 killed $vgpr16 def $vgpr16_vgpr17 killed $exec
-; GFX9-O0-NEXT: v_mov_b32_e32 v17, v8
-; GFX9-O0-NEXT: v_mov_b32_e32 v8, v16
-; GFX9-O0-NEXT: v_mov_b32_e32 v9, v17
-; GFX9-O0-NEXT: v_mov_b32_e32 v10, v19
-; GFX9-O0-NEXT: v_mov_b32_e32 v11, v20
-; GFX9-O0-NEXT: v_mov_b32_e32 v21, v17
-; GFX9-O0-NEXT: v_mov_b32_e32 v18, v20
-; GFX9-O0-NEXT: v_or_b32_e64 v18, v18, v21
-; GFX9-O0-NEXT: v_mov_b32_e32 v17, v16
-; GFX9-O0-NEXT: v_mov_b32_e32 v16, v19
-; GFX9-O0-NEXT: v_or_b32_e64 v16, v16, v17
-; GFX9-O0-NEXT: ; kill: def $vgpr16 killed $vgpr16 def $vgpr16_vgpr17 killed $exec
-; GFX9-O0-NEXT: v_mov_b32_e32 v17, v18
-; GFX9-O0-NEXT: v_cmp_eq_u64_e64 s[4:5], v[16:17], v[12:13]
+; GFX9-O0-NEXT: ; kill: def $vgpr8 killed $vgpr8 def $vgpr8_vgpr9 killed $exec
+; GFX9-O0-NEXT: v_mov_b32_e32 v9, v10
+; GFX9-O0-NEXT: v_mov_b32_e32 v14, v9
+; GFX9-O0-NEXT: v_mov_b32_e32 v13, v8
+; GFX9-O0-NEXT: buffer_store_dword v13, off, s[0:3], s32 offset:276 ; 4-byte Folded Spill
+; GFX9-O0-NEXT: s_nop 0
+; GFX9-O0-NEXT: buffer_store_dword v14, off, s[0:3], s32 offset:280 ; 4-byte Folded Spill
+; GFX9-O0-NEXT: v_mov_b32_e32 v14, v12
+; GFX9-O0-NEXT: v_mov_b32_e32 v13, v11
+; GFX9-O0-NEXT: buffer_store_dword v13, off, s[0:3], s32 offset:268 ; 4-byte Folded Spill
+; GFX9-O0-NEXT: s_nop 0
+; GFX9-O0-NEXT: buffer_store_dword v14, off, s[0:3], s32 offset:272 ; 4-byte Folded Spill
+; GFX9-O0-NEXT: v_mov_b32_e32 v13, v9
+; GFX9-O0-NEXT: v_mov_b32_e32 v10, v12
+; GFX9-O0-NEXT: v_or_b32_e64 v10, v10, v13
+; GFX9-O0-NEXT: v_mov_b32_e32 v9, v8
+; GFX9-O0-NEXT: v_mov_b32_e32 v8, v11
+; GFX9-O0-NEXT: v_or_b32_e64 v8, v8, v9
+; GFX9-O0-NEXT: ; kill: def $vgpr8 killed $vgpr8 def $vgpr8_vgpr9 killed $exec
+; GFX9-O0-NEXT: v_mov_b32_e32 v9, v10
+; GFX9-O0-NEXT: v_cmp_eq_u64_e64 s[4:5], v[8:9], v[4:5]
; GFX9-O0-NEXT: s_or_b64 s[4:5], s[4:5], s[6:7]
-; GFX9-O0-NEXT: v_mov_b32_e32 v17, v3
-; GFX9-O0-NEXT: v_mov_b32_e32 v16, v2
-; GFX9-O0-NEXT: buffer_store_dword v16, off, s[0:3], s32 offset:152 ; 4-byte Folded Spill
-; GFX9-O0-NEXT: s_nop 0
-; GFX9-O0-NEXT: buffer_store_dword v17, off, s[0:3], s32 offset:156 ; 4-byte Folded Spill
-; GFX9-O0-NEXT: v_mov_b32_e32 v17, v1
-; GFX9-O0-NEXT: v_mov_b32_e32 v16, v0
-; GFX9-O0-NEXT: buffer_store_dword v16, off, s[0:3], s32 offset:160 ; 4-byte Folded Spill
-; GFX9-O0-NEXT: s_nop 0
-; GFX9-O0-NEXT: buffer_store_dword v17, off, s[0:3], s32 offset:164 ; 4-byte Folded Spill
-; GFX9-O0-NEXT: v_mov_b32_e32 v17, v15
-; GFX9-O0-NEXT: v_mov_b32_e32 v16, v14
-; GFX9-O0-NEXT: buffer_store_dword v16, off, s[0:3], s32 offset:168 ; 4-byte Folded Spill
-; GFX9-O0-NEXT: s_nop 0
-; GFX9-O0-NEXT: buffer_store_dword v17, off, s[0:3], s32 offset:172 ; 4-byte Folded Spill
-; GFX9-O0-NEXT: v_mov_b32_e32 v17, v13
-; GFX9-O0-NEXT: v_mov_b32_e32 v16, v12
-; GFX9-O0-NEXT: buffer_store_dword v16, off, s[0:3], s32 offset:176 ; 4-byte Folded Spill
-; GFX9-O0-NEXT: s_nop 0
-; GFX9-O0-NEXT: buffer_store_dword v17, off, s[0:3], s32 offset:180 ; 4-byte Folded Spill
-; GFX9-O0-NEXT: s_mov_b64 s[6:7], s[4:5]
-; GFX9-O0-NEXT: v_writelane_b32 v30, s6, 6
-; GFX9-O0-NEXT: v_writelane_b32 v30, s7, 7
-; GFX9-O0-NEXT: s_mov_b64 s[6:7], s[4:5]
-; GFX9-O0-NEXT: v_writelane_b32 v30, s6, 10
-; GFX9-O0-NEXT: v_writelane_b32 v30, s7, 11
-; GFX9-O0-NEXT: s_or_saveexec_b64 s[22:23], -1
-; GFX9-O0-NEXT: buffer_store_dword v30, off, s[0:3], s32 ; 4-byte Folded Spill
-; GFX9-O0-NEXT: s_mov_b64 exec, s[22:23]
-; GFX9-O0-NEXT: buffer_store_dword v14, off, s[0:3], s32 offset:288 ; 4-byte Folded Spill
+; GFX9-O0-NEXT: v_mov_b32_e32 v9, v3
+; GFX9-O0-NEXT: v_mov_b32_e32 v8, v2
+; GFX9-O0-NEXT: buffer_store_dword v8, off, s[0:3], s32 offset:156 ; 4-byte Folded Spill
; GFX9-O0-NEXT: s_nop 0
-; GFX9-O0-NEXT: buffer_store_dword v15, off, s[0:3], s32 offset:292 ; 4-byte Folded Spill
-; GFX9-O0-NEXT: buffer_store_dword v12, off, s[0:3], s32 offset:280 ; 4-byte Folded Spill
+; GFX9-O0-NEXT: buffer_store_dword v9, off, s[0:3], s32 offset:160 ; 4-byte Folded Spill
+; GFX9-O0-NEXT: v_mov_b32_e32 v9, v1
+; GFX9-O0-NEXT: v_mov_b32_e32 v8, v0
+; GFX9-O0-NEXT: buffer_store_dword v8, off, s[0:3], s32 offset:164 ; 4-byte Folded Spill
; GFX9-O0-NEXT: s_nop 0
-; GFX9-O0-NEXT: buffer_store_dword v13, off, s[0:3], s32 offset:284 ; 4-byte Folded Spill
-; GFX9-O0-NEXT: buffer_store_dword v10, off, s[0:3], s32 offset:272 ; 4-byte Folded Spill
+; GFX9-O0-NEXT: buffer_store_dword v9, off, s[0:3], s32 offset:168 ; 4-byte Folded Spill
+; GFX9-O0-NEXT: v_mov_b32_e32 v9, v7
+; GFX9-O0-NEXT: v_mov_b32_e32 v8, v6
+; GFX9-O0-NEXT: buffer_store_dword v8, off, s[0:3], s32 offset:172 ; 4-byte Folded Spill
; GFX9-O0-NEXT: s_nop 0
-; GFX9-O0-NEXT: buffer_store_dword v11, off, s[0:3], s32 offset:276 ; 4-byte Folded Spill
-; GFX9-O0-NEXT: buffer_store_dword v8, off, s[0:3], s32 offset:264 ; 4-byte Folded Spill
+; GFX9-O0-NEXT: buffer_store_dword v9, off, s[0:3], s32 offset:176 ; 4-byte Folded Spill
+; GFX9-O0-NEXT: v_mov_b32_e32 v9, v5
+; GFX9-O0-NEXT: v_mov_b32_e32 v8, v4
+; GFX9-O0-NEXT: buffer_store_dword v8, off, s[0:3], s32 offset:180 ; 4-byte Folded Spill
; GFX9-O0-NEXT: s_nop 0
-; GFX9-O0-NEXT: buffer_store_dword v9, off, s[0:3], s32 offset:268 ; 4-byte Folded Spill
-; GFX9-O0-NEXT: buffer_store_dword v6, off, s[0:3], s32 offset:256 ; 4-byte Folded Spill
+; GFX9-O0-NEXT: buffer_store_dword v9, off, s[0:3], s32 offset:184 ; 4-byte Folded Spill
+; GFX9-O0-NEXT: s_mov_b64 s[6:7], s[4:5]
+; GFX9-O0-NEXT: v_writelane_b32 v29, s6, 6
+; GFX9-O0-NEXT: v_writelane_b32 v29, s7, 7
+; GFX9-O0-NEXT: s_mov_b64 s[6:7], s[4:5]
+; GFX9-O0-NEXT: v_writelane_b32 v29, s6, 10
+; GFX9-O0-NEXT: v_writelane_b32 v29, s7, 11
+; GFX9-O0-NEXT: s_or_saveexec_b64 s[22:23], -1
+; GFX9-O0-NEXT: buffer_store_dword v29, off, s[0:3], s32 ; 4-byte Folded Spill
+; GFX9-O0-NEXT: s_mov_b64 exec, s[22:23]
+; GFX9-O0-NEXT: buffer_store_dword v6, off, s[0:3], s32 offset:260 ; 4-byte Folded Spill
; GFX9-O0-NEXT: s_nop 0
-; GFX9-O0-NEXT: buffer_store_dword v7, off, s[0:3], s32 offset:260 ; 4-byte Folded Spill
-; GFX9-O0-NEXT: buffer_store_dword v4, off, s[0:3], s32 offset:248 ; 4-byte Folded Spill
+; GFX9-O0-NEXT: buffer_store_dword v7, off, s[0:3], s32 offset:264 ; 4-byte Folded Spill
+; GFX9-O0-NEXT: buffer_store_dword v4, off, s[0:3], s32 offset:252 ; 4-byte Folded Spill
; GFX9-O0-NEXT: s_nop 0
-; GFX9-O0-NEXT: buffer_store_dword v5, off, s[0:3], s32 offset:252 ; 4-byte Folded Spill
-; GFX9-O0-NEXT: buffer_store_dword v2, off, s[0:3], s32 offset:240 ; 4-byte Folded Spill
+; GFX9-O0-NEXT: buffer_store_dword v5, off, s[0:3], s32 offset:256 ; 4-byte Folded Spill
+; GFX9-O0-NEXT: buffer_store_dword v2, off, s[0:3], s32 offset:244 ; 4-byte Folded Spill
; GFX9-O0-NEXT: s_nop 0
-; GFX9-O0-NEXT: buffer_store_dword v3, off, s[0:3], s32 offset:244 ; 4-byte Folded Spill
-; GFX9-O0-NEXT: buffer_store_dword v0, off, s[0:3], s32 offset:232 ; 4-byte Folded Spill
+; GFX9-O0-NEXT: buffer_store_dword v3, off, s[0:3], s32 offset:248 ; 4-byte Folded Spill
+; GFX9-O0-NEXT: buffer_store_dword v0, off, s[0:3], s32 offset:236 ; 4-byte Folded Spill
; GFX9-O0-NEXT: s_nop 0
-; GFX9-O0-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:236 ; 4-byte Folded Spill
+; GFX9-O0-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:240 ; 4-byte Folded Spill
; GFX9-O0-NEXT: s_andn2_b64 exec, exec, s[4:5]
; GFX9-O0-NEXT: s_cbranch_execnz .LBB0_6
; GFX9-O0-NEXT: s_branch .LBB0_1
; GFX9-O0-NEXT: .LBB0_7: ; %udiv-preheader
; GFX9-O0-NEXT: s_or_saveexec_b64 s[22:23], -1
-; GFX9-O0-NEXT: buffer_load_dword v30, off, s[0:3], s32 ; 4-byte Folded Reload
+; GFX9-O0-NEXT: buffer_load_dword v29, off, s[0:3], s32 ; 4-byte Folded Reload
; GFX9-O0-NEXT: s_mov_b64 exec, s[22:23]
-; GFX9-O0-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:312 ; 4-byte Folded Reload
-; GFX9-O0-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:316 ; 4-byte Folded Reload
-; GFX9-O0-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:320 ; 4-byte Folded Reload
-; GFX9-O0-NEXT: buffer_load_dword v3, off, s[0:3], s32 offset:324 ; 4-byte Folded Reload
-; GFX9-O0-NEXT: buffer_load_dword v8, off, s[0:3], s32 offset:328 ; 4-byte Folded Reload
-; GFX9-O0-NEXT: buffer_load_dword v9, off, s[0:3], s32 offset:332 ; 4-byte Folded Reload
-; GFX9-O0-NEXT: buffer_load_dword v10, off, s[0:3], s32 offset:336 ; 4-byte Folded Reload
-; GFX9-O0-NEXT: buffer_load_dword v11, off, s[0:3], s32 offset:340 ; 4-byte Folded Reload
-; GFX9-O0-NEXT: buffer_load_dword v16, off, s[0:3], s32 offset:60 ; 4-byte Folded Reload
-; GFX9-O0-NEXT: buffer_load_dword v17, off, s[0:3], s32 offset:64 ; 4-byte Folded Reload
-; GFX9-O0-NEXT: buffer_load_dword v12, off, s[0:3], s32 offset:52 ; 4-byte Folded Reload
-; GFX9-O0-NEXT: buffer_load_dword v13, off, s[0:3], s32 offset:56 ; 4-byte Folded Reload
-; GFX9-O0-NEXT: buffer_load_dword v14, off, s[0:3], s32 offset:44 ; 4-byte Folded Reload
-; GFX9-O0-NEXT: buffer_load_dword v15, off, s[0:3], s32 offset:48 ; 4-byte Folded Reload
+; GFX9-O0-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:316 ; 4-byte Folded Reload
+; GFX9-O0-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:320 ; 4-byte Folded Reload
+; GFX9-O0-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:324 ; 4-byte Folded Reload
+; GFX9-O0-NEXT: buffer_load_dword v3, off, s[0:3], s32 offset:328 ; 4-byte Folded Reload
+; GFX9-O0-NEXT: buffer_load_dword v4, off, s[0:3], s32 offset:332 ; 4-byte Folded Reload
+; GFX9-O0-NEXT: buffer_load_dword v5, off, s[0:3], s32 offset:336 ; 4-byte Folded Reload
+; GFX9-O0-NEXT: buffer_load_dword v6, off, s[0:3], s32 offset:340 ; 4-byte Folded Reload
+; GFX9-O0-NEXT: buffer_load_dword v7, off, s[0:3], s32 offset:344 ; 4-byte Folded Reload
+; GFX9-O0-NEXT: buffer_load_dword v12, off, s[0:3], s32 offset:60 ; 4-byte Folded Reload
+; GFX9-O0-NEXT: buffer_load_dword v13, off, s[0:3], s32 offset:64 ; 4-byte Folded Reload
+; GFX9-O0-NEXT: buffer_load_dword v9, off, s[0:3], s32 offset:52 ; 4-byte Folded Reload
+; GFX9-O0-NEXT: buffer_load_dword v10, off, s[0:3], s32 offset:56 ; 4-byte Folded Reload
+; GFX9-O0-NEXT: buffer_load_dword v16, off, s[0:3], s32 offset:44 ; 4-byte Folded Reload
+; GFX9-O0-NEXT: buffer_load_dword v17, off, s[0:3], s32 offset:48 ; 4-byte Folded Reload
; GFX9-O0-NEXT: buffer_load_dword v18, off, s[0:3], s32 offset:36 ; 4-byte Folded Reload
; GFX9-O0-NEXT: buffer_load_dword v19, off, s[0:3], s32 offset:40 ; 4-byte Folded Reload
; GFX9-O0-NEXT: s_waitcnt vmcnt(9)
-; GFX9-O0-NEXT: v_mov_b32_e32 v4, v10
+; GFX9-O0-NEXT: v_mov_b32_e32 v8, v6
; GFX9-O0-NEXT: s_waitcnt vmcnt(0)
-; GFX9-O0-NEXT: v_lshrrev_b64 v[6:7], v4, v[18:19]
-; GFX9-O0-NEXT: v_mov_b32_e32 v5, v7
+; GFX9-O0-NEXT: v_lshrrev_b64 v[14:15], v8, v[18:19]
+; GFX9-O0-NEXT: v_mov_b32_e32 v11, v15
; GFX9-O0-NEXT: s_mov_b32 s4, 64
-; GFX9-O0-NEXT: v_sub_u32_e64 v20, s4, v4
-; GFX9-O0-NEXT: v_lshlrev_b64 v[20:21], v20, v[14:15]
+; GFX9-O0-NEXT: v_sub_u32_e64 v20, s4, v8
+; GFX9-O0-NEXT: v_lshlrev_b64 v[20:21], v20, v[16:17]
; GFX9-O0-NEXT: v_mov_b32_e32 v22, v21
-; GFX9-O0-NEXT: v_or_b32_e64 v5, v5, v22
-; GFX9-O0-NEXT: ; kill: def $vgpr6 killed $vgpr6 killed $vgpr6_vgpr7 killed $exec
-; GFX9-O0-NEXT: v_mov_b32_e32 v7, v20
-; GFX9-O0-NEXT: v_or_b32_e64 v6, v6, v7
-; GFX9-O0-NEXT: ; kill: def $vgpr6 killed $vgpr6 def $vgpr6_vgpr7 killed $exec
-; GFX9-O0-NEXT: v_mov_b32_e32 v7, v5
-; GFX9-O0-NEXT: v_mov_b32_e32 v22, v7
-; GFX9-O0-NEXT: v_cmp_lt_u32_e64 s[6:7], v4, s4
-; GFX9-O0-NEXT: v_sub_u32_e64 v5, v4, s4
-; GFX9-O0-NEXT: v_lshrrev_b64 v[20:21], v5, v[14:15]
-; GFX9-O0-NEXT: v_mov_b32_e32 v5, v21
-; GFX9-O0-NEXT: v_cndmask_b32_e64 v5, v5, v22, s[6:7]
+; GFX9-O0-NEXT: v_or_b32_e64 v11, v11, v22
+; GFX9-O0-NEXT: ; kill: def $vgpr14 killed $vgpr14 killed $vgpr14_vgpr15 killed $exec
+; GFX9-O0-NEXT: v_mov_b32_e32 v15, v20
+; GFX9-O0-NEXT: v_or_b32_e64 v14, v14, v15
+; GFX9-O0-NEXT: ; kill: def $vgpr14 killed $vgpr14 def $vgpr14_vgpr15 killed $exec
+; GFX9-O0-NEXT: v_mov_b32_e32 v15, v11
+; GFX9-O0-NEXT: v_mov_b32_e32 v22, v15
+; GFX9-O0-NEXT: v_cmp_lt_u32_e64 s[6:7], v8, s4
+; GFX9-O0-NEXT: v_sub_u32_e64 v11, v8, s4
+; GFX9-O0-NEXT: v_lshrrev_b64 v[20:21], v11, v[16:17]
+; GFX9-O0-NEXT: v_mov_b32_e32 v11, v21
+; GFX9-O0-NEXT: v_cndmask_b32_e64 v11, v11, v22, s[6:7]
; GFX9-O0-NEXT: s_mov_b32 s4, 0
-; GFX9-O0-NEXT: v_cmp_eq_u32_e64 s[4:5], v4, s4
+; GFX9-O0-NEXT: v_cmp_eq_u32_e64 s[4:5], v8, s4
; GFX9-O0-NEXT: v_mov_b32_e32 v22, v19
-; GFX9-O0-NEXT: v_cndmask_b32_e64 v5, v5, v22, s[4:5]
-; GFX9-O0-NEXT: v_mov_b32_e32 v7, v6
-; GFX9-O0-NEXT: v_mov_b32_e32 v6, v20
-; GFX9-O0-NEXT: v_cndmask_b32_e64 v6, v6, v7, s[6:7]
-; GFX9-O0-NEXT: v_mov_b32_e32 v7, v18
-; GFX9-O0-NEXT: v_cndmask_b32_e64 v6, v6, v7, s[4:5]
+; GFX9-O0-NEXT: v_cndmask_b32_e64 v11, v11, v22, s[4:5]
+; GFX9-O0-NEXT: v_mov_b32_e32 v15, v14
+; GFX9-O0-NEXT: v_mov_b32_e32 v14, v20
+; GFX9-O0-NEXT: v_cndmask_b32_e64 v14, v14, v15, s[6:7]
+; GFX9-O0-NEXT: v_mov_b32_e32 v15, v18
+; GFX9-O0-NEXT: v_cndmask_b32_e64 v14, v14, v15, s[4:5]
; GFX9-O0-NEXT: ; implicit-def: $sgpr4
; GFX9-O0-NEXT: ; implicit-def: $sgpr4
-; GFX9-O0-NEXT: ; kill: def $vgpr6 killed $vgpr6 def $vgpr6_vgpr7 killed $exec
-; GFX9-O0-NEXT: v_mov_b32_e32 v7, v5
-; GFX9-O0-NEXT: v_lshrrev_b64 v[4:5], v4, v[14:15]
-; GFX9-O0-NEXT: v_mov_b32_e32 v15, v5
+; GFX9-O0-NEXT: ; kill: def $vgpr14 killed $vgpr14 def $vgpr14_vgpr15 killed $exec
+; GFX9-O0-NEXT: v_mov_b32_e32 v15, v11
+; GFX9-O0-NEXT: v_lshrrev_b64 v[16:17], v8, v[16:17]
+; GFX9-O0-NEXT: v_mov_b32_e32 v11, v17
; GFX9-O0-NEXT: s_mov_b64 s[4:5], 0
; GFX9-O0-NEXT: s_mov_b32 s8, s5
-; GFX9-O0-NEXT: v_mov_b32_e32 v14, s8
-; GFX9-O0-NEXT: v_cndmask_b32_e64 v14, v14, v15, s[6:7]
-; GFX9-O0-NEXT: v_mov_b32_e32 v5, v4
+; GFX9-O0-NEXT: v_mov_b32_e32 v8, s8
+; GFX9-O0-NEXT: v_cndmask_b32_e64 v8, v8, v11, s[6:7]
+; GFX9-O0-NEXT: ; kill: def $vgpr16 killed $vgpr16 killed $vgpr16_vgpr17 killed $exec
; GFX9-O0-NEXT: s_mov_b32 s8, s4
-; GFX9-O0-NEXT: v_mov_b32_e32 v4, s8
-; GFX9-O0-NEXT: v_cndmask_b32_e64 v4, v4, v5, s[6:7]
+; GFX9-O0-NEXT: v_mov_b32_e32 v11, s8
+; GFX9-O0-NEXT: v_cndmask_b32_e64 v16, v11, v16, s[6:7]
; GFX9-O0-NEXT: ; implicit-def: $sgpr6
; GFX9-O0-NEXT: ; implicit-def: $sgpr6
-; GFX9-O0-NEXT: ; kill: def $vgpr4 killed $vgpr4 def $vgpr4_vgpr5 killed $exec
-; GFX9-O0-NEXT: v_mov_b32_e32 v5, v14
-; GFX9-O0-NEXT: v_mov_b32_e32 v15, v12
-; GFX9-O0-NEXT: v_mov_b32_e32 v12, v13
+; GFX9-O0-NEXT: ; kill: def $vgpr16 killed $vgpr16 def $vgpr16_vgpr17 killed $exec
+; GFX9-O0-NEXT: v_mov_b32_e32 v17, v8
+; GFX9-O0-NEXT: buffer_store_dword v16, off, s[0:3], s32 offset:292 ; 4-byte Folded Spill
+; GFX9-O0-NEXT: s_nop 0
+; GFX9-O0-NEXT: buffer_store_dword v17, off, s[0:3], s32 offset:296 ; 4-byte Folded Spill
+; GFX9-O0-NEXT: buffer_store_dword v14, off, s[0:3], s32 offset:284 ; 4-byte Folded Spill
+; GFX9-O0-NEXT: s_nop 0
+; GFX9-O0-NEXT: buffer_store_dword v15, off, s[0:3], s32 offset:288 ; 4-byte Folded Spill
+; GFX9-O0-NEXT: v_mov_b32_e32 v8, v9
+; GFX9-O0-NEXT: v_mov_b32_e32 v11, v10
; GFX9-O0-NEXT: s_mov_b64 s[8:9], -1
; GFX9-O0-NEXT: s_mov_b32 s7, s8
; GFX9-O0-NEXT: s_mov_b32 s6, s9
-; GFX9-O0-NEXT: v_mov_b32_e32 v14, v16
-; GFX9-O0-NEXT: v_mov_b32_e32 v13, v17
-; GFX9-O0-NEXT: v_mov_b32_e32 v16, s7
-; GFX9-O0-NEXT: v_add_co_u32_e32 v16, vcc, v15, v16
-; GFX9-O0-NEXT: v_mov_b32_e32 v15, s6
-; GFX9-O0-NEXT: v_addc_co_u32_e32 v12, vcc, v12, v15, vcc
-; GFX9-O0-NEXT: v_mov_b32_e32 v15, s7
-; GFX9-O0-NEXT: v_addc_co_u32_e32 v18, vcc, v14, v15, vcc
-; GFX9-O0-NEXT: v_mov_b32_e32 v14, s6
-; GFX9-O0-NEXT: v_addc_co_u32_e32 v13, vcc, v13, v14, vcc
+; GFX9-O0-NEXT: v_mov_b32_e32 v10, v12
+; GFX9-O0-NEXT: v_mov_b32_e32 v9, v13
+; GFX9-O0-NEXT: v_mov_b32_e32 v12, s7
+; GFX9-O0-NEXT: v_add_co_u32_e32 v8, vcc, v8, v12
+; GFX9-O0-NEXT: v_mov_b32_e32 v12, s6
+; GFX9-O0-NEXT: v_addc_co_u32_e32 v12, vcc, v11, v12, vcc
+; GFX9-O0-NEXT: v_mov_b32_e32 v11, s7
+; GFX9-O0-NEXT: v_addc_co_u32_e32 v10, vcc, v10, v11, vcc
+; GFX9-O0-NEXT: v_mov_b32_e32 v11, s6
+; GFX9-O0-NEXT: v_addc_co_u32_e32 v9, vcc, v9, v11, vcc
; GFX9-O0-NEXT: ; implicit-def: $sgpr6
; GFX9-O0-NEXT: ; implicit-def: $sgpr6
-; GFX9-O0-NEXT: ; kill: def $vgpr18 killed $vgpr18 def $vgpr18_vgpr19 killed $exec
-; GFX9-O0-NEXT: v_mov_b32_e32 v19, v13
+; GFX9-O0-NEXT: ; kill: def $vgpr10 killed $vgpr10 def $vgpr10_vgpr11 killed $exec
+; GFX9-O0-NEXT: v_mov_b32_e32 v11, v9
; GFX9-O0-NEXT: ; implicit-def: $sgpr6
; GFX9-O0-NEXT: ; implicit-def: $sgpr6
-; GFX9-O0-NEXT: ; kill: def $vgpr16 killed $vgpr16 def $vgpr16_vgpr17 killed $exec
-; GFX9-O0-NEXT: v_mov_b32_e32 v17, v12
+; GFX9-O0-NEXT: ; kill: def $vgpr8 killed $vgpr8 def $vgpr8_vgpr9 killed $exec
+; GFX9-O0-NEXT: v_mov_b32_e32 v9, v12
; GFX9-O0-NEXT: v_mov_b32_e32 v13, s5
; GFX9-O0-NEXT: v_mov_b32_e32 v12, s4
-; GFX9-O0-NEXT: v_mov_b32_e32 v15, s5
-; GFX9-O0-NEXT: v_mov_b32_e32 v14, s4
-; GFX9-O0-NEXT: buffer_store_dword v18, off, s[0:3], s32 offset:296 ; 4-byte Folded Spill
-; GFX9-O0-NEXT: s_nop 0
-; GFX9-O0-NEXT: buffer_store_dword v19, off, s[0:3], s32 offset:300 ; 4-byte Folded Spill
-; GFX9-O0-NEXT: buffer_store_dword v16, off, s[0:3], s32 offset:304 ; 4-byte Folded Spill
-; GFX9-O0-NEXT: s_nop 0
-; GFX9-O0-NEXT: buffer_store_dword v17, off, s[0:3], s32 offset:308 ; 4-byte Folded Spill
-; GFX9-O0-NEXT: v_writelane_b32 v30, s4, 10
-; GFX9-O0-NEXT: v_writelane_b32 v30, s5, 11
-; GFX9-O0-NEXT: s_or_saveexec_b64 s[22:23], -1
-; GFX9-O0-NEXT: buffer_store_dword v30, off, s[0:3], s32 ; 4-byte Folded Spill
-; GFX9-O0-NEXT: s_mov_b64 exec, s[22:23]
-; GFX9-O0-NEXT: buffer_store_dword v14, off, s[0:3], s32 offset:288 ; 4-byte Folded Spill
+; GFX9-O0-NEXT: buffer_store_dword v12, off, s[0:3], s32 offset:252 ; 4-byte Folded Spill
; GFX9-O0-NEXT: s_nop 0
-; GFX9-O0-NEXT: buffer_store_dword v15, off, s[0:3], s32 offset:292 ; 4-byte Folded Spill
-; GFX9-O0-NEXT: buffer_store_dword v12, off, s[0:3], s32 offset:280 ; 4-byte Folded Spill
+; GFX9-O0-NEXT: buffer_store_dword v13, off, s[0:3], s32 offset:256 ; 4-byte Folded Spill
+; GFX9-O0-NEXT: v_mov_b32_e32 v13, s5
+; GFX9-O0-NEXT: v_mov_b32_e32 v12, s4
+; GFX9-O0-NEXT: buffer_store_dword v12, off, s[0:3], s32 offset:260 ; 4-byte Folded Spill
; GFX9-O0-NEXT: s_nop 0
-; GFX9-O0-NEXT: buffer_store_dword v13, off, s[0:3], s32 offset:284 ; 4-byte Folded Spill
-; GFX9-O0-NEXT: buffer_store_dword v10, off, s[0:3], s32 offset:272 ; 4-byte Folded Spill
+; GFX9-O0-NEXT: buffer_store_dword v13, off, s[0:3], s32 offset:264 ; 4-byte Folded Spill
+; GFX9-O0-NEXT: buffer_store_dword v10, off, s[0:3], s32 offset:300 ; 4-byte Folded Spill
; GFX9-O0-NEXT: s_nop 0
-; GFX9-O0-NEXT: buffer_store_dword v11, off, s[0:3], s32 offset:276 ; 4-byte Folded Spill
-; GFX9-O0-NEXT: buffer_store_dword v8, off, s[0:3], s32 offset:264 ; 4-byte Folded Spill
+; GFX9-O0-NEXT: buffer_store_dword v11, off, s[0:3], s32 offset:304 ; 4-byte Folded Spill
+; GFX9-O0-NEXT: buffer_store_dword v8, off, s[0:3], s32 offset:308 ; 4-byte Folded Spill
; GFX9-O0-NEXT: s_nop 0
-; GFX9-O0-NEXT: buffer_store_dword v9, off, s[0:3], s32 offset:268 ; 4-byte Folded Spill
-; GFX9-O0-NEXT: buffer_store_dword v6, off, s[0:3], s32 offset:256 ; 4-byte Folded Spill
+; GFX9-O0-NEXT: buffer_store_dword v9, off, s[0:3], s32 offset:312 ; 4-byte Folded Spill
+; GFX9-O0-NEXT: v_writelane_b32 v29, s4, 10
+; GFX9-O0-NEXT: v_writelane_b32 v29, s5, 11
+; GFX9-O0-NEXT: s_or_saveexec_b64 s[22:23], -1
+; GFX9-O0-NEXT: buffer_store_dword v29, off, s[0:3], s32 ; 4-byte Folded Spill
+; GFX9-O0-NEXT: s_mov_b64 exec, s[22:23]
+; GFX9-O0-NEXT: buffer_store_dword v6, off, s[0:3], s32 offset:268 ; 4-byte Folded Spill
; GFX9-O0-NEXT: s_nop 0
-; GFX9-O0-NEXT: buffer_store_dword v7, off, s[0:3], s32 offset:260 ; 4-byte Folded Spill
-; GFX9-O0-NEXT: buffer_store_dword v4, off, s[0:3], s32 offset:248 ; 4-byte Folded Spill
+; GFX9-O0-NEXT: buffer_store_dword v7, off, s[0:3], s32 offset:272 ; 4-byte Folded Spill
+; GFX9-O0-NEXT: buffer_store_dword v4, off, s[0:3], s32 offset:276 ; 4-byte Folded Spill
; GFX9-O0-NEXT: s_nop 0
-; GFX9-O0-NEXT: buffer_store_dword v5, off, s[0:3], s32 offset:252 ; 4-byte Folded Spill
-; GFX9-O0-NEXT: buffer_store_dword v2, off, s[0:3], s32 offset:240 ; 4-byte Folded Spill
+; GFX9-O0-NEXT: buffer_store_dword v5, off, s[0:3], s32 offset:280 ; 4-byte Folded Spill
+; GFX9-O0-NEXT: buffer_store_dword v2, off, s[0:3], s32 offset:244 ; 4-byte Folded Spill
; GFX9-O0-NEXT: s_nop 0
-; GFX9-O0-NEXT: buffer_store_dword v3, off, s[0:3], s32 offset:244 ; 4-byte Folded Spill
-; GFX9-O0-NEXT: buffer_store_dword v0, off, s[0:3], s32 offset:232 ; 4-byte Folded Spill
+; GFX9-O0-NEXT: buffer_store_dword v3, off, s[0:3], s32 offset:248 ; 4-byte Folded Spill
+; GFX9-O0-NEXT: buffer_store_dword v0, off, s[0:3], s32 offset:236 ; 4-byte Folded Spill
; GFX9-O0-NEXT: s_nop 0
-; GFX9-O0-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:236 ; 4-byte Folded Spill
+; GFX9-O0-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:240 ; 4-byte Folded Spill
; GFX9-O0-NEXT: s_branch .LBB0_6
; GFX9-O0-NEXT: .LBB0_8: ; %udiv-bb1
; GFX9-O0-NEXT: s_or_saveexec_b64 s[22:23], -1
-; GFX9-O0-NEXT: buffer_load_dword v30, off, s[0:3], s32 ; 4-byte Folded Reload
+; GFX9-O0-NEXT: buffer_load_dword v29, off, s[0:3], s32 ; 4-byte Folded Reload
; GFX9-O0-NEXT: s_mov_b64 exec, s[22:23]
-; GFX9-O0-NEXT: buffer_load_dword v6, off, s[0:3], s32 offset:36 ; 4-byte Folded Reload
-; GFX9-O0-NEXT: buffer_load_dword v7, off, s[0:3], s32 offset:40 ; 4-byte Folded Reload
-; GFX9-O0-NEXT: buffer_load_dword v10, off, s[0:3], s32 offset:44 ; 4-byte Folded Reload
-; GFX9-O0-NEXT: buffer_load_dword v11, off, s[0:3], s32 offset:48 ; 4-byte Folded Reload
+; GFX9-O0-NEXT: buffer_load_dword v9, off, s[0:3], s32 offset:36 ; 4-byte Folded Reload
+; GFX9-O0-NEXT: buffer_load_dword v10, off, s[0:3], s32 offset:40 ; 4-byte Folded Reload
+; GFX9-O0-NEXT: buffer_load_dword v11, off, s[0:3], s32 offset:44 ; 4-byte Folded Reload
+; GFX9-O0-NEXT: buffer_load_dword v12, off, s[0:3], s32 offset:48 ; 4-byte Folded Reload
; GFX9-O0-NEXT: buffer_load_dword v4, off, s[0:3], s32 offset:20 ; 4-byte Folded Reload
; GFX9-O0-NEXT: buffer_load_dword v5, off, s[0:3], s32 offset:24 ; 4-byte Folded Reload
; GFX9-O0-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:28 ; 4-byte Folded Reload
@@ -1085,118 +1086,118 @@ define i128 @v_srem_i128_vv(i128 %lhs, i128 %rhs) {
; GFX9-O0-NEXT: s_mov_b64 s[6:7], 0
; GFX9-O0-NEXT: s_mov_b32 s8, s6
; GFX9-O0-NEXT: s_mov_b32 s9, s7
-; GFX9-O0-NEXT: v_mov_b32_e32 v0, v4
-; GFX9-O0-NEXT: v_mov_b32_e32 v2, v5
+; GFX9-O0-NEXT: v_mov_b32_e32 v2, v4
+; GFX9-O0-NEXT: v_mov_b32_e32 v0, v5
; GFX9-O0-NEXT: v_mov_b32_e32 v4, s5
-; GFX9-O0-NEXT: v_add_co_u32_e32 v8, vcc, v3, v4
+; GFX9-O0-NEXT: v_add_co_u32_e32 v7, vcc, v3, v4
; GFX9-O0-NEXT: v_mov_b32_e32 v4, s4
; GFX9-O0-NEXT: s_waitcnt vmcnt(0)
; GFX9-O0-NEXT: v_addc_co_u32_e32 v1, vcc, v1, v4, vcc
; GFX9-O0-NEXT: v_mov_b32_e32 v4, s8
-; GFX9-O0-NEXT: v_addc_co_u32_e32 v0, vcc, v0, v4, vcc
-; GFX9-O0-NEXT: v_mov_b32_e32 v4, s9
-; GFX9-O0-NEXT: v_addc_co_u32_e32 v2, vcc, v2, v4, vcc
+; GFX9-O0-NEXT: v_addc_co_u32_e32 v4, vcc, v2, v4, vcc
+; GFX9-O0-NEXT: v_mov_b32_e32 v2, s9
+; GFX9-O0-NEXT: v_addc_co_u32_e32 v0, vcc, v0, v2, vcc
; GFX9-O0-NEXT: ; implicit-def: $sgpr4
; GFX9-O0-NEXT: ; implicit-def: $sgpr4
-; GFX9-O0-NEXT: ; kill: def $vgpr8 killed $vgpr8 def $vgpr8_vgpr9 killed $exec
-; GFX9-O0-NEXT: v_mov_b32_e32 v9, v1
+; GFX9-O0-NEXT: ; kill: def $vgpr7 killed $vgpr7 def $vgpr7_vgpr8 killed $exec
+; GFX9-O0-NEXT: v_mov_b32_e32 v8, v1
; GFX9-O0-NEXT: ; implicit-def: $sgpr4
; GFX9-O0-NEXT: ; implicit-def: $sgpr4
-; GFX9-O0-NEXT: ; kill: def $vgpr0 killed $vgpr0 def $vgpr0_vgpr1 killed $exec
-; GFX9-O0-NEXT: v_mov_b32_e32 v1, v2
-; GFX9-O0-NEXT: v_mov_b32_e32 v5, v1
-; GFX9-O0-NEXT: v_mov_b32_e32 v4, v0
-; GFX9-O0-NEXT: buffer_store_dword v4, off, s[0:3], s32 offset:328 ; 4-byte Folded Spill
+; GFX9-O0-NEXT: ; kill: def $vgpr4 killed $vgpr4 def $vgpr4_vgpr5 killed $exec
+; GFX9-O0-NEXT: v_mov_b32_e32 v5, v0
+; GFX9-O0-NEXT: v_mov_b32_e32 v0, v4
+; GFX9-O0-NEXT: v_mov_b32_e32 v1, v5
+; GFX9-O0-NEXT: buffer_store_dword v0, off, s[0:3], s32 offset:332 ; 4-byte Folded Spill
; GFX9-O0-NEXT: s_nop 0
-; GFX9-O0-NEXT: buffer_store_dword v5, off, s[0:3], s32 offset:332 ; 4-byte Folded Spill
-; GFX9-O0-NEXT: v_mov_b32_e32 v4, v8
-; GFX9-O0-NEXT: v_mov_b32_e32 v5, v9
-; GFX9-O0-NEXT: buffer_store_dword v4, off, s[0:3], s32 offset:336 ; 4-byte Folded Spill
+; GFX9-O0-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:336 ; 4-byte Folded Spill
+; GFX9-O0-NEXT: v_mov_b32_e32 v0, v7
+; GFX9-O0-NEXT: v_mov_b32_e32 v1, v8
+; GFX9-O0-NEXT: buffer_store_dword v0, off, s[0:3], s32 offset:340 ; 4-byte Folded Spill
; GFX9-O0-NEXT: s_nop 0
-; GFX9-O0-NEXT: buffer_store_dword v5, off, s[0:3], s32 offset:340 ; 4-byte Folded Spill
+; GFX9-O0-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:344 ; 4-byte Folded Spill
; GFX9-O0-NEXT: s_mov_b32 s4, 0x7f
; GFX9-O0-NEXT: v_sub_u32_e64 v2, s4, v3
-; GFX9-O0-NEXT: v_lshlrev_b64 v[4:5], v2, v[10:11]
-; GFX9-O0-NEXT: v_mov_b32_e32 v12, v5
+; GFX9-O0-NEXT: v_lshlrev_b64 v[0:1], v2, v[11:12]
+; GFX9-O0-NEXT: v_mov_b32_e32 v6, v1
; GFX9-O0-NEXT: s_mov_b32 s4, 64
; GFX9-O0-NEXT: v_sub_u32_e64 v13, s4, v2
-; GFX9-O0-NEXT: v_lshrrev_b64 v[13:14], v13, v[6:7]
+; GFX9-O0-NEXT: v_lshrrev_b64 v[13:14], v13, v[9:10]
; GFX9-O0-NEXT: v_mov_b32_e32 v15, v14
-; GFX9-O0-NEXT: v_or_b32_e64 v12, v12, v15
-; GFX9-O0-NEXT: ; kill: def $vgpr4 killed $vgpr4 killed $vgpr4_vgpr5 killed $exec
-; GFX9-O0-NEXT: v_mov_b32_e32 v5, v13
-; GFX9-O0-NEXT: v_or_b32_e64 v4, v4, v5
-; GFX9-O0-NEXT: ; kill: def $vgpr4 killed $vgpr4 def $vgpr4_vgpr5 killed $exec
-; GFX9-O0-NEXT: v_mov_b32_e32 v5, v12
-; GFX9-O0-NEXT: v_mov_b32_e32 v14, v5
+; GFX9-O0-NEXT: v_or_b32_e64 v6, v6, v15
+; GFX9-O0-NEXT: ; kill: def $vgpr0 killed $vgpr0 killed $vgpr0_vgpr1 killed $exec
+; GFX9-O0-NEXT: v_mov_b32_e32 v1, v13
+; GFX9-O0-NEXT: v_or_b32_e64 v0, v0, v1
+; GFX9-O0-NEXT: ; kill: def $vgpr0 killed $vgpr0 def $vgpr0_vgpr1 killed $exec
+; GFX9-O0-NEXT: v_mov_b32_e32 v1, v6
+; GFX9-O0-NEXT: v_mov_b32_e32 v6, v1
; GFX9-O0-NEXT: v_cmp_lt_u32_e64 s[4:5], v2, s4
; GFX9-O0-NEXT: s_mov_b32 s10, 63
; GFX9-O0-NEXT: v_sub_u32_e64 v3, s10, v3
-; GFX9-O0-NEXT: v_lshlrev_b64 v[12:13], v3, v[6:7]
-; GFX9-O0-NEXT: v_mov_b32_e32 v3, v13
-; GFX9-O0-NEXT: v_cndmask_b32_e64 v3, v3, v14, s[4:5]
+; GFX9-O0-NEXT: v_lshlrev_b64 v[13:14], v3, v[9:10]
+; GFX9-O0-NEXT: v_mov_b32_e32 v3, v14
+; GFX9-O0-NEXT: v_cndmask_b32_e64 v3, v3, v6, s[4:5]
; GFX9-O0-NEXT: s_mov_b32 s10, 0
; GFX9-O0-NEXT: v_cmp_eq_u32_e64 s[10:11], v2, s10
-; GFX9-O0-NEXT: v_mov_b32_e32 v14, v11
-; GFX9-O0-NEXT: v_cndmask_b32_e64 v3, v3, v14, s[10:11]
-; GFX9-O0-NEXT: v_mov_b32_e32 v5, v4
-; GFX9-O0-NEXT: v_mov_b32_e32 v4, v12
-; GFX9-O0-NEXT: v_cndmask_b32_e64 v4, v4, v5, s[4:5]
-; GFX9-O0-NEXT: v_mov_b32_e32 v5, v10
-; GFX9-O0-NEXT: v_cndmask_b32_e64 v4, v4, v5, s[10:11]
+; GFX9-O0-NEXT: v_mov_b32_e32 v6, v12
+; GFX9-O0-NEXT: v_cndmask_b32_e64 v3, v3, v6, s[10:11]
+; GFX9-O0-NEXT: v_mov_b32_e32 v1, v0
+; GFX9-O0-NEXT: v_mov_b32_e32 v0, v13
+; GFX9-O0-NEXT: v_cndmask_b32_e64 v0, v0, v1, s[4:5]
+; GFX9-O0-NEXT: v_mov_b32_e32 v1, v11
+; GFX9-O0-NEXT: v_cndmask_b32_e64 v0, v0, v1, s[10:11]
; GFX9-O0-NEXT: ; implicit-def: $sgpr10
; GFX9-O0-NEXT: ; implicit-def: $sgpr10
-; GFX9-O0-NEXT: ; kill: def $vgpr4 killed $vgpr4 def $vgpr4_vgpr5 killed $exec
-; GFX9-O0-NEXT: v_mov_b32_e32 v5, v3
-; GFX9-O0-NEXT: v_lshlrev_b64 v[6:7], v2, v[6:7]
-; GFX9-O0-NEXT: v_mov_b32_e32 v3, v7
-; GFX9-O0-NEXT: v_mov_b32_e32 v2, s9
+; GFX9-O0-NEXT: ; kill: def $vgpr0 killed $vgpr0 def $vgpr0_vgpr1 killed $exec
+; GFX9-O0-NEXT: v_mov_b32_e32 v1, v3
+; GFX9-O0-NEXT: v_lshlrev_b64 v[2:3], v2, v[9:10]
+; GFX9-O0-NEXT: v_mov_b32_e32 v9, v3
+; GFX9-O0-NEXT: v_mov_b32_e32 v6, s9
+; GFX9-O0-NEXT: v_cndmask_b32_e64 v6, v6, v9, s[4:5]
+; GFX9-O0-NEXT: v_mov_b32_e32 v3, v2
+; GFX9-O0-NEXT: v_mov_b32_e32 v2, s8
; GFX9-O0-NEXT: v_cndmask_b32_e64 v2, v2, v3, s[4:5]
-; GFX9-O0-NEXT: ; kill: def $vgpr6 killed $vgpr6 killed $vgpr6_vgpr7 killed $exec
-; GFX9-O0-NEXT: v_mov_b32_e32 v3, s8
-; GFX9-O0-NEXT: v_cndmask_b32_e64 v6, v3, v6, s[4:5]
; GFX9-O0-NEXT: ; implicit-def: $sgpr4
; GFX9-O0-NEXT: ; implicit-def: $sgpr4
-; GFX9-O0-NEXT: ; kill: def $vgpr6 killed $vgpr6 def $vgpr6_vgpr7 killed $exec
-; GFX9-O0-NEXT: v_mov_b32_e32 v7, v2
-; GFX9-O0-NEXT: buffer_store_dword v6, off, s[0:3], s32 offset:320 ; 4-byte Folded Spill
+; GFX9-O0-NEXT: ; kill: def $vgpr2 killed $vgpr2 def $vgpr2_vgpr3 killed $exec
+; GFX9-O0-NEXT: v_mov_b32_e32 v3, v6
+; GFX9-O0-NEXT: buffer_store_dword v2, off, s[0:3], s32 offset:324 ; 4-byte Folded Spill
; GFX9-O0-NEXT: s_nop 0
-; GFX9-O0-NEXT: buffer_store_dword v7, off, s[0:3], s32 offset:324 ; 4-byte Folded Spill
-; GFX9-O0-NEXT: buffer_store_dword v4, off, s[0:3], s32 offset:312 ; 4-byte Folded Spill
+; GFX9-O0-NEXT: buffer_store_dword v3, off, s[0:3], s32 offset:328 ; 4-byte Folded Spill
+; GFX9-O0-NEXT: buffer_store_dword v0, off, s[0:3], s32 offset:316 ; 4-byte Folded Spill
; GFX9-O0-NEXT: s_nop 0
-; GFX9-O0-NEXT: buffer_store_dword v5, off, s[0:3], s32 offset:316 ; 4-byte Folded Spill
-; GFX9-O0-NEXT: v_mov_b32_e32 v3, v1
-; GFX9-O0-NEXT: v_mov_b32_e32 v2, v9
-; GFX9-O0-NEXT: v_or_b32_e64 v2, v2, v3
-; GFX9-O0-NEXT: v_mov_b32_e32 v1, v0
-; GFX9-O0-NEXT: v_mov_b32_e32 v0, v8
-; GFX9-O0-NEXT: v_or_b32_e64 v0, v0, v1
-; GFX9-O0-NEXT: ; kill: def $vgpr0 killed $vgpr0 def $vgpr0_vgpr1 killed $exec
-; GFX9-O0-NEXT: v_mov_b32_e32 v1, v2
-; GFX9-O0-NEXT: v_cmp_ne_u64_e64 s[4:5], v[0:1], s[6:7]
-; GFX9-O0-NEXT: v_mov_b32_e32 v0, s6
-; GFX9-O0-NEXT: v_mov_b32_e32 v1, s7
-; GFX9-O0-NEXT: v_mov_b32_e32 v2, s6
-; GFX9-O0-NEXT: v_mov_b32_e32 v3, s7
-; GFX9-O0-NEXT: buffer_store_dword v6, off, s[0:3], s32 offset:144 ; 4-byte Folded Spill
+; GFX9-O0-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:320 ; 4-byte Folded Spill
+; GFX9-O0-NEXT: v_mov_b32_e32 v9, v5
+; GFX9-O0-NEXT: v_mov_b32_e32 v6, v8
+; GFX9-O0-NEXT: v_or_b32_e64 v6, v6, v9
+; GFX9-O0-NEXT: v_mov_b32_e32 v5, v4
+; GFX9-O0-NEXT: v_mov_b32_e32 v4, v7
+; GFX9-O0-NEXT: v_or_b32_e64 v4, v4, v5
+; GFX9-O0-NEXT: ; kill: def $vgpr4 killed $vgpr4 def $vgpr4_vgpr5 killed $exec
+; GFX9-O0-NEXT: v_mov_b32_e32 v5, v6
+; GFX9-O0-NEXT: v_cmp_ne_u64_e64 s[4:5], v[4:5], s[6:7]
+; GFX9-O0-NEXT: v_mov_b32_e32 v4, s6
+; GFX9-O0-NEXT: v_mov_b32_e32 v5, s7
+; GFX9-O0-NEXT: buffer_store_dword v4, off, s[0:3], s32 offset:148 ; 4-byte Folded Spill
; GFX9-O0-NEXT: s_nop 0
-; GFX9-O0-NEXT: buffer_store_dword v7, off, s[0:3], s32 offset:148 ; 4-byte Folded Spill
-; GFX9-O0-NEXT: buffer_store_dword v4, off, s[0:3], s32 offset:136 ; 4-byte Folded Spill
+; GFX9-O0-NEXT: buffer_store_dword v5, off, s[0:3], s32 offset:152 ; 4-byte Folded Spill
+; GFX9-O0-NEXT: v_mov_b32_e32 v4, s6
+; GFX9-O0-NEXT: v_mov_b32_e32 v5, s7
+; GFX9-O0-NEXT: buffer_store_dword v4, off, s[0:3], s32 offset:140 ; 4-byte Folded Spill
; GFX9-O0-NEXT: s_nop 0
-; GFX9-O0-NEXT: buffer_store_dword v5, off, s[0:3], s32 offset:140 ; 4-byte Folded Spill
-; GFX9-O0-NEXT: buffer_store_dword v2, off, s[0:3], s32 offset:128 ; 4-byte Folded Spill
+; GFX9-O0-NEXT: buffer_store_dword v5, off, s[0:3], s32 offset:144 ; 4-byte Folded Spill
+; GFX9-O0-NEXT: buffer_store_dword v2, off, s[0:3], s32 offset:124 ; 4-byte Folded Spill
; GFX9-O0-NEXT: s_nop 0
-; GFX9-O0-NEXT: buffer_store_dword v3, off, s[0:3], s32 offset:132 ; 4-byte Folded Spill
-; GFX9-O0-NEXT: buffer_store_dword v0, off, s[0:3], s32 offset:120 ; 4-byte Folded Spill
+; GFX9-O0-NEXT: buffer_store_dword v3, off, s[0:3], s32 offset:128 ; 4-byte Folded Spill
+; GFX9-O0-NEXT: buffer_store_dword v0, off, s[0:3], s32 offset:132 ; 4-byte Folded Spill
; GFX9-O0-NEXT: s_nop 0
-; GFX9-O0-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:124 ; 4-byte Folded Spill
+; GFX9-O0-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:136 ; 4-byte Folded Spill
; GFX9-O0-NEXT: s_mov_b64 s[6:7], exec
; GFX9-O0-NEXT: s_and_b64 s[4:5], s[6:7], s[4:5]
; GFX9-O0-NEXT: s_xor_b64 s[6:7], s[4:5], s[6:7]
-; GFX9-O0-NEXT: v_writelane_b32 v30, s6, 8
-; GFX9-O0-NEXT: v_writelane_b32 v30, s7, 9
+; GFX9-O0-NEXT: v_writelane_b32 v29, s6, 8
+; GFX9-O0-NEXT: v_writelane_b32 v29, s7, 9
; GFX9-O0-NEXT: s_or_saveexec_b64 s[22:23], -1
-; GFX9-O0-NEXT: buffer_store_dword v30, off, s[0:3], s32 ; 4-byte Folded Spill
+; GFX9-O0-NEXT: buffer_store_dword v29, off, s[0:3], s32 ; 4-byte Folded Spill
; GFX9-O0-NEXT: s_mov_b64 exec, s[22:23]
; GFX9-O0-NEXT: s_mov_b64 exec, s[4:5]
; GFX9-O0-NEXT: s_cbranch_execz .LBB0_5
@@ -1212,10 +1213,10 @@ define i128 @v_srem_i128_vv(i128 %lhs, i128 %rhs) {
; GFX9-O0-NEXT: buffer_load_dword v4, off, s[0:3], s32 offset:88 ; 4-byte Folded Reload
; GFX9-O0-NEXT: buffer_load_dword v18, off, s[0:3], s32 offset:68 ; 4-byte Folded Reload
; GFX9-O0-NEXT: buffer_load_dword v19, off, s[0:3], s32 offset:72 ; 4-byte Folded Reload
-; GFX9-O0-NEXT: buffer_load_dword v14, off, s[0:3], s32 offset:192 ; 4-byte Folded Reload
-; GFX9-O0-NEXT: buffer_load_dword v15, off, s[0:3], s32 offset:196 ; 4-byte Folded Reload
-; GFX9-O0-NEXT: buffer_load_dword v16, off, s[0:3], s32 offset:184 ; 4-byte Folded Reload
-; GFX9-O0-NEXT: buffer_load_dword v17, off, s[0:3], s32 offset:188 ; 4-byte Folded Reload
+; GFX9-O0-NEXT: buffer_load_dword v14, off, s[0:3], s32 offset:196 ; 4-byte Folded Reload
+; GFX9-O0-NEXT: buffer_load_dword v15, off, s[0:3], s32 offset:200 ; 4-byte Folded Reload
+; GFX9-O0-NEXT: buffer_load_dword v16, off, s[0:3], s32 offset:188 ; 4-byte Folded Reload
+; GFX9-O0-NEXT: buffer_load_dword v17, off, s[0:3], s32 offset:192 ; 4-byte Folded Reload
; GFX9-O0-NEXT: buffer_load_dword v20, off, s[0:3], s32 offset:76 ; 4-byte Folded Reload
; GFX9-O0-NEXT: buffer_load_dword v21, off, s[0:3], s32 offset:80 ; 4-byte Folded Reload
; GFX9-O0-NEXT: s_mov_b32 s4, 32
@@ -1489,7 +1490,7 @@ define i128 @v_srem_i128_vv(i128 %lhs, i128 %rhs) {
; GFX9-O0-NEXT: v_lshrrev_b64 v[3:4], s4, v[3:4]
; GFX9-O0-NEXT: ; kill: def $vgpr3 killed $vgpr3 killed $vgpr3_vgpr4 killed $exec
; GFX9-O0-NEXT: s_xor_saveexec_b64 s[4:5], -1
-; GFX9-O0-NEXT: buffer_load_dword v30, off, s[0:3], s32 offset:344 ; 4-byte Folded Reload
+; GFX9-O0-NEXT: buffer_load_dword v29, off, s[0:3], s32 offset:348 ; 4-byte Folded Reload
; GFX9-O0-NEXT: s_mov_b64 exec, s[4:5]
; GFX9-O0-NEXT: s_waitcnt vmcnt(0)
; GFX9-O0-NEXT: s_setpc_b64 s[30:31]
@@ -1703,7 +1704,7 @@ define i128 @v_urem_i128_vv(i128 %lhs, i128 %rhs) {
; GFX9-O0: ; %bb.0: ; %_udiv-special-cases
; GFX9-O0-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX9-O0-NEXT: s_xor_saveexec_b64 s[4:5], -1
-; GFX9-O0-NEXT: buffer_store_dword v30, off, s[0:3], s32 offset:328 ; 4-byte Folded Spill
+; GFX9-O0-NEXT: buffer_store_dword v29, off, s[0:3], s32 offset:328 ; 4-byte Folded Spill
; GFX9-O0-NEXT: s_mov_b64 exec, s[4:5]
; GFX9-O0-NEXT: v_mov_b32_e32 v10, v6
; GFX9-O0-NEXT: buffer_store_dword v4, off, s[0:3], s32 offset:100 ; 4-byte Folded Spill
@@ -1780,16 +1781,16 @@ define i128 @v_urem_i128_vv(i128 %lhs, i128 %rhs) {
; GFX9-O0-NEXT: ; kill: def $vgpr0 killed $vgpr0 def $vgpr0_vgpr1 killed $exec
; GFX9-O0-NEXT: v_mov_b32_e32 v1, v2
; GFX9-O0-NEXT: s_mov_b64 s[6:7], 0
-; GFX9-O0-NEXT: ; implicit-def: $vgpr30 : SGPR spill to VGPR lane
-; GFX9-O0-NEXT: v_writelane_b32 v30, s6, 0
-; GFX9-O0-NEXT: v_writelane_b32 v30, s7, 1
+; GFX9-O0-NEXT: ; implicit-def: $vgpr29 : SGPR spill to VGPR lane
+; GFX9-O0-NEXT: v_writelane_b32 v29, s6, 0
+; GFX9-O0-NEXT: v_writelane_b32 v29, s7, 1
; GFX9-O0-NEXT: v_cmp_eq_u64_e64 s[4:5], v[0:1], s[6:7]
-; GFX9-O0-NEXT: v_mov_b32_e32 v1, v13
-; GFX9-O0-NEXT: v_mov_b32_e32 v3, v9
-; GFX9-O0-NEXT: v_or_b32_e64 v14, v3, v1
-; GFX9-O0-NEXT: v_mov_b32_e32 v0, v12
-; GFX9-O0-NEXT: v_mov_b32_e32 v2, v8
-; GFX9-O0-NEXT: v_or_b32_e64 v8, v2, v0
+; GFX9-O0-NEXT: v_mov_b32_e32 v3, v13
+; GFX9-O0-NEXT: v_mov_b32_e32 v1, v9
+; GFX9-O0-NEXT: v_or_b32_e64 v14, v1, v3
+; GFX9-O0-NEXT: v_mov_b32_e32 v2, v12
+; GFX9-O0-NEXT: v_mov_b32_e32 v0, v8
+; GFX9-O0-NEXT: v_or_b32_e64 v8, v0, v2
; GFX9-O0-NEXT: ; kill: def $vgpr8 killed $vgpr8 def $vgpr8_vgpr9 killed $exec
; GFX9-O0-NEXT: v_mov_b32_e32 v9, v14
; GFX9-O0-NEXT: v_cmp_eq_u64_e64 s[8:9], v[8:9], s[6:7]
@@ -1834,18 +1835,18 @@ define i128 @v_urem_i128_vv(i128 %lhs, i128 %rhs) {
; GFX9-O0-NEXT: ; implicit-def: $sgpr12
; GFX9-O0-NEXT: ; kill: def $vgpr8 killed $vgpr8 def $vgpr8_vgpr9 killed $exec
; GFX9-O0-NEXT: v_mov_b32_e32 v9, v4
-; GFX9-O0-NEXT: v_ffbh_u32_e64 v4, v0
+; GFX9-O0-NEXT: v_ffbh_u32_e64 v4, v2
; GFX9-O0-NEXT: v_add_u32_e64 v4, v4, s9
-; GFX9-O0-NEXT: v_ffbh_u32_e64 v5, v1
+; GFX9-O0-NEXT: v_ffbh_u32_e64 v5, v3
; GFX9-O0-NEXT: v_min_u32_e64 v5, v4, v5
; GFX9-O0-NEXT: ; implicit-def: $sgpr12
; GFX9-O0-NEXT: v_mov_b32_e32 v4, s8
; GFX9-O0-NEXT: ; kill: def $vgpr5 killed $vgpr5 def $vgpr5_vgpr6 killed $exec
; GFX9-O0-NEXT: v_mov_b32_e32 v6, v4
; GFX9-O0-NEXT: v_mov_b32_e32 v7, v6
-; GFX9-O0-NEXT: v_ffbh_u32_e64 v4, v2
+; GFX9-O0-NEXT: v_ffbh_u32_e64 v4, v0
; GFX9-O0-NEXT: v_add_u32_e64 v4, v4, s9
-; GFX9-O0-NEXT: v_ffbh_u32_e64 v10, v3
+; GFX9-O0-NEXT: v_ffbh_u32_e64 v10, v1
; GFX9-O0-NEXT: v_min_u32_e64 v14, v4, v10
; GFX9-O0-NEXT: ; implicit-def: $sgpr9
; GFX9-O0-NEXT: v_mov_b32_e32 v4, s8
@@ -1928,33 +1929,33 @@ define i128 @v_urem_i128_vv(i128 %lhs, i128 %rhs) {
; GFX9-O0-NEXT: v_mov_b32_e32 v5, v6
; GFX9-O0-NEXT: v_cmp_ne_u64_e64 s[6:7], v[4:5], s[6:7]
; GFX9-O0-NEXT: v_mov_b32_e32 v4, s11
-; GFX9-O0-NEXT: v_cndmask_b32_e64 v4, v1, v4, s[8:9]
-; GFX9-O0-NEXT: v_mov_b32_e32 v1, s10
-; GFX9-O0-NEXT: v_cndmask_b32_e64 v0, v0, v1, s[8:9]
-; GFX9-O0-NEXT: ; implicit-def: $sgpr12
-; GFX9-O0-NEXT: ; implicit-def: $sgpr12
-; GFX9-O0-NEXT: ; kill: def $vgpr0 killed $vgpr0 def $vgpr0_vgpr1 killed $exec
-; GFX9-O0-NEXT: v_mov_b32_e32 v1, v4
-; GFX9-O0-NEXT: v_mov_b32_e32 v4, s11
; GFX9-O0-NEXT: v_cndmask_b32_e64 v4, v3, v4, s[8:9]
; GFX9-O0-NEXT: v_mov_b32_e32 v3, s10
; GFX9-O0-NEXT: v_cndmask_b32_e64 v2, v2, v3, s[8:9]
-; GFX9-O0-NEXT: ; implicit-def: $sgpr8
-; GFX9-O0-NEXT: ; implicit-def: $sgpr8
+; GFX9-O0-NEXT: ; implicit-def: $sgpr12
+; GFX9-O0-NEXT: ; implicit-def: $sgpr12
; GFX9-O0-NEXT: ; kill: def $vgpr2 killed $vgpr2 def $vgpr2_vgpr3 killed $exec
; GFX9-O0-NEXT: v_mov_b32_e32 v3, v4
-; GFX9-O0-NEXT: s_and_b64 s[6:7], s[4:5], s[6:7]
+; GFX9-O0-NEXT: v_mov_b32_e32 v4, s11
+; GFX9-O0-NEXT: v_cndmask_b32_e64 v4, v1, v4, s[8:9]
+; GFX9-O0-NEXT: v_mov_b32_e32 v1, s10
+; GFX9-O0-NEXT: v_cndmask_b32_e64 v0, v0, v1, s[8:9]
+; GFX9-O0-NEXT: ; implicit-def: $sgpr8
+; GFX9-O0-NEXT: ; implicit-def: $sgpr8
+; GFX9-O0-NEXT: ; kill: def $vgpr0 killed $vgpr0 def $vgpr0_vgpr1 killed $exec
+; GFX9-O0-NEXT: v_mov_b32_e32 v1, v4
; GFX9-O0-NEXT: buffer_store_dword v2, off, s[0:3], s32 offset:12 ; 4-byte Folded Spill
; GFX9-O0-NEXT: s_nop 0
; GFX9-O0-NEXT: buffer_store_dword v3, off, s[0:3], s32 offset:16 ; 4-byte Folded Spill
; GFX9-O0-NEXT: buffer_store_dword v0, off, s[0:3], s32 offset:4 ; 4-byte Folded Spill
; GFX9-O0-NEXT: s_nop 0
; GFX9-O0-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:8 ; 4-byte Folded Spill
+; GFX9-O0-NEXT: s_and_b64 s[6:7], s[4:5], s[6:7]
; GFX9-O0-NEXT: s_mov_b64 s[4:5], exec
-; GFX9-O0-NEXT: v_writelane_b32 v30, s4, 2
-; GFX9-O0-NEXT: v_writelane_b32 v30, s5, 3
+; GFX9-O0-NEXT: v_writelane_b32 v29, s4, 2
+; GFX9-O0-NEXT: v_writelane_b32 v29, s5, 3
; GFX9-O0-NEXT: s_or_saveexec_b64 s[18:19], -1
-; GFX9-O0-NEXT: buffer_store_dword v30, off, s[0:3], s32 ; 4-byte Folded Spill
+; GFX9-O0-NEXT: buffer_store_dword v29, off, s[0:3], s32 ; 4-byte Folded Spill
; GFX9-O0-NEXT: s_mov_b64 exec, s[18:19]
; GFX9-O0-NEXT: s_and_b64 s[4:5], s[4:5], s[6:7]
; GFX9-O0-NEXT: s_mov_b64 exec, s[4:5]
@@ -1962,50 +1963,47 @@ define i128 @v_urem_i128_vv(i128 %lhs, i128 %rhs) {
; GFX9-O0-NEXT: s_branch .LBB1_8
; GFX9-O0-NEXT: .LBB1_1: ; %Flow
; GFX9-O0-NEXT: s_or_saveexec_b64 s[18:19], -1
-; GFX9-O0-NEXT: buffer_load_dword v30, off, s[0:3], s32 ; 4-byte Folded Reload
+; GFX9-O0-NEXT: buffer_load_dword v29, off, s[0:3], s32 ; 4-byte Folded Reload
; GFX9-O0-NEXT: s_mov_b64 exec, s[18:19]
; GFX9-O0-NEXT: s_waitcnt vmcnt(0)
-; GFX9-O0-NEXT: v_readlane_b32 s4, v30, 4
-; GFX9-O0-NEXT: v_readlane_b32 s5, v30, 5
+; GFX9-O0-NEXT: v_readlane_b32 s4, v29, 4
+; GFX9-O0-NEXT: v_readlane_b32 s5, v29, 5
; GFX9-O0-NEXT: s_or_b64 exec, exec, s[4:5]
; GFX9-O0-NEXT: ; %bb.2: ; %Flow
-; GFX9-O0-NEXT: buffer_load_dword v6, off, s[0:3], s32 offset:136 ; 4-byte Folded Reload
-; GFX9-O0-NEXT: buffer_load_dword v7, off, s[0:3], s32 offset:140 ; 4-byte Folded Reload
-; GFX9-O0-NEXT: buffer_load_dword v4, off, s[0:3], s32 offset:144 ; 4-byte Folded Reload
-; GFX9-O0-NEXT: buffer_load_dword v5, off, s[0:3], s32 offset:148 ; 4-byte Folded Reload
-; GFX9-O0-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:152 ; 4-byte Folded Reload
-; GFX9-O0-NEXT: buffer_load_dword v3, off, s[0:3], s32 offset:156 ; 4-byte Folded Reload
-; GFX9-O0-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:160 ; 4-byte Folded Reload
-; GFX9-O0-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:164 ; 4-byte Folded Reload
-; GFX9-O0-NEXT: s_waitcnt vmcnt(7)
+; GFX9-O0-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:136 ; 4-byte Folded Reload
+; GFX9-O0-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:140 ; 4-byte Folded Reload
+; GFX9-O0-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:144 ; 4-byte Folded Reload
+; GFX9-O0-NEXT: buffer_load_dword v3, off, s[0:3], s32 offset:148 ; 4-byte Folded Reload
+; GFX9-O0-NEXT: buffer_load_dword v4, off, s[0:3], s32 offset:152 ; 4-byte Folded Reload
+; GFX9-O0-NEXT: buffer_load_dword v5, off, s[0:3], s32 offset:156 ; 4-byte Folded Reload
+; GFX9-O0-NEXT: buffer_load_dword v6, off, s[0:3], s32 offset:160 ; 4-byte Folded Reload
+; GFX9-O0-NEXT: buffer_load_dword v7, off, s[0:3], s32 offset:164 ; 4-byte Folded Reload
+; GFX9-O0-NEXT: s_waitcnt vmcnt(1)
; GFX9-O0-NEXT: buffer_store_dword v6, off, s[0:3], s32 offset:128 ; 4-byte Folded Spill
-; GFX9-O0-NEXT: s_waitcnt vmcnt(7)
+; GFX9-O0-NEXT: s_waitcnt vmcnt(1)
; GFX9-O0-NEXT: buffer_store_dword v7, off, s[0:3], s32 offset:132 ; 4-byte Folded Spill
-; GFX9-O0-NEXT: s_waitcnt vmcnt(7)
; GFX9-O0-NEXT: buffer_store_dword v4, off, s[0:3], s32 offset:120 ; 4-byte Folded Spill
-; GFX9-O0-NEXT: s_waitcnt vmcnt(7)
+; GFX9-O0-NEXT: s_nop 0
; GFX9-O0-NEXT: buffer_store_dword v5, off, s[0:3], s32 offset:124 ; 4-byte Folded Spill
-; GFX9-O0-NEXT: s_waitcnt vmcnt(7)
; GFX9-O0-NEXT: buffer_store_dword v2, off, s[0:3], s32 offset:112 ; 4-byte Folded Spill
-; GFX9-O0-NEXT: s_waitcnt vmcnt(7)
+; GFX9-O0-NEXT: s_nop 0
; GFX9-O0-NEXT: buffer_store_dword v3, off, s[0:3], s32 offset:116 ; 4-byte Folded Spill
-; GFX9-O0-NEXT: s_waitcnt vmcnt(7)
; GFX9-O0-NEXT: buffer_store_dword v0, off, s[0:3], s32 offset:104 ; 4-byte Folded Spill
-; GFX9-O0-NEXT: s_waitcnt vmcnt(7)
+; GFX9-O0-NEXT: s_nop 0
; GFX9-O0-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:108 ; 4-byte Folded Spill
; GFX9-O0-NEXT: s_branch .LBB1_5
; GFX9-O0-NEXT: .LBB1_3: ; %Flow2
; GFX9-O0-NEXT: s_or_saveexec_b64 s[18:19], -1
-; GFX9-O0-NEXT: buffer_load_dword v30, off, s[0:3], s32 ; 4-byte Folded Reload
+; GFX9-O0-NEXT: buffer_load_dword v29, off, s[0:3], s32 ; 4-byte Folded Reload
; GFX9-O0-NEXT: s_mov_b64 exec, s[18:19]
; GFX9-O0-NEXT: s_waitcnt vmcnt(0)
-; GFX9-O0-NEXT: v_readlane_b32 s4, v30, 2
-; GFX9-O0-NEXT: v_readlane_b32 s5, v30, 3
+; GFX9-O0-NEXT: v_readlane_b32 s4, v29, 2
+; GFX9-O0-NEXT: v_readlane_b32 s5, v29, 3
; GFX9-O0-NEXT: s_or_b64 exec, exec, s[4:5]
-; GFX9-O0-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:12 ; 4-byte Folded Reload
-; GFX9-O0-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:16 ; 4-byte Folded Reload
-; GFX9-O0-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:4 ; 4-byte Folded Reload
-; GFX9-O0-NEXT: buffer_load_dword v3, off, s[0:3], s32 offset:8 ; 4-byte Folded Reload
+; GFX9-O0-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:4 ; 4-byte Folded Reload
+; GFX9-O0-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:8 ; 4-byte Folded Reload
+; GFX9-O0-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:12 ; 4-byte Folded Reload
+; GFX9-O0-NEXT: buffer_load_dword v3, off, s[0:3], s32 offset:16 ; 4-byte Folded Reload
; GFX9-O0-NEXT: s_waitcnt vmcnt(1)
; GFX9-O0-NEXT: buffer_store_dword v2, off, s[0:3], s32 offset:176 ; 4-byte Folded Spill
; GFX9-O0-NEXT: s_waitcnt vmcnt(1)
@@ -2047,29 +2045,29 @@ define i128 @v_urem_i128_vv(i128 %lhs, i128 %rhs) {
; GFX9-O0-NEXT: v_or_b32_e64 v2, v2, v3
; GFX9-O0-NEXT: ; kill: def $vgpr2 killed $vgpr2 def $vgpr2_vgpr3 killed $exec
; GFX9-O0-NEXT: v_mov_b32_e32 v3, v4
-; GFX9-O0-NEXT: buffer_store_dword v2, off, s[0:3], s32 offset:12 ; 4-byte Folded Spill
+; GFX9-O0-NEXT: buffer_store_dword v2, off, s[0:3], s32 offset:4 ; 4-byte Folded Spill
; GFX9-O0-NEXT: s_nop 0
-; GFX9-O0-NEXT: buffer_store_dword v3, off, s[0:3], s32 offset:16 ; 4-byte Folded Spill
-; GFX9-O0-NEXT: buffer_store_dword v0, off, s[0:3], s32 offset:4 ; 4-byte Folded Spill
+; GFX9-O0-NEXT: buffer_store_dword v3, off, s[0:3], s32 offset:8 ; 4-byte Folded Spill
+; GFX9-O0-NEXT: buffer_store_dword v0, off, s[0:3], s32 offset:12 ; 4-byte Folded Spill
; GFX9-O0-NEXT: s_nop 0
-; GFX9-O0-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:8 ; 4-byte Folded Spill
+; GFX9-O0-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:16 ; 4-byte Folded Spill
; GFX9-O0-NEXT: s_branch .LBB1_3
; GFX9-O0-NEXT: .LBB1_5: ; %Flow1
; GFX9-O0-NEXT: s_or_saveexec_b64 s[18:19], -1
-; GFX9-O0-NEXT: buffer_load_dword v30, off, s[0:3], s32 ; 4-byte Folded Reload
+; GFX9-O0-NEXT: buffer_load_dword v29, off, s[0:3], s32 ; 4-byte Folded Reload
; GFX9-O0-NEXT: s_mov_b64 exec, s[18:19]
; GFX9-O0-NEXT: s_waitcnt vmcnt(0)
-; GFX9-O0-NEXT: v_readlane_b32 s4, v30, 6
-; GFX9-O0-NEXT: v_readlane_b32 s5, v30, 7
+; GFX9-O0-NEXT: v_readlane_b32 s4, v29, 6
+; GFX9-O0-NEXT: v_readlane_b32 s5, v29, 7
; GFX9-O0-NEXT: s_or_b64 exec, exec, s[4:5]
-; GFX9-O0-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:128 ; 4-byte Folded Reload
-; GFX9-O0-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:132 ; 4-byte Folded Reload
-; GFX9-O0-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:120 ; 4-byte Folded Reload
-; GFX9-O0-NEXT: buffer_load_dword v3, off, s[0:3], s32 offset:124 ; 4-byte Folded Reload
-; GFX9-O0-NEXT: buffer_load_dword v4, off, s[0:3], s32 offset:112 ; 4-byte Folded Reload
-; GFX9-O0-NEXT: buffer_load_dword v5, off, s[0:3], s32 offset:116 ; 4-byte Folded Reload
-; GFX9-O0-NEXT: buffer_load_dword v6, off, s[0:3], s32 offset:104 ; 4-byte Folded Reload
-; GFX9-O0-NEXT: buffer_load_dword v7, off, s[0:3], s32 offset:108 ; 4-byte Folded Reload
+; GFX9-O0-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:104 ; 4-byte Folded Reload
+; GFX9-O0-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:108 ; 4-byte Folded Reload
+; GFX9-O0-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:112 ; 4-byte Folded Reload
+; GFX9-O0-NEXT: buffer_load_dword v3, off, s[0:3], s32 offset:116 ; 4-byte Folded Reload
+; GFX9-O0-NEXT: buffer_load_dword v4, off, s[0:3], s32 offset:120 ; 4-byte Folded Reload
+; GFX9-O0-NEXT: buffer_load_dword v5, off, s[0:3], s32 offset:124 ; 4-byte Folded Reload
+; GFX9-O0-NEXT: buffer_load_dword v6, off, s[0:3], s32 offset:128 ; 4-byte Folded Reload
+; GFX9-O0-NEXT: buffer_load_dword v7, off, s[0:3], s32 offset:132 ; 4-byte Folded Reload
; GFX9-O0-NEXT: s_waitcnt vmcnt(1)
; GFX9-O0-NEXT: buffer_store_dword v6, off, s[0:3], s32 offset:192 ; 4-byte Folded Spill
; GFX9-O0-NEXT: s_waitcnt vmcnt(1)
@@ -2087,214 +2085,214 @@ define i128 @v_urem_i128_vv(i128 %lhs, i128 %rhs) {
; GFX9-O0-NEXT: .LBB1_6: ; %udiv-do-while
; GFX9-O0-NEXT: ; =>This Inner Loop Header: Depth=1
; GFX9-O0-NEXT: s_or_saveexec_b64 s[18:19], -1
-; GFX9-O0-NEXT: buffer_load_dword v30, off, s[0:3], s32 ; 4-byte Folded Reload
+; GFX9-O0-NEXT: buffer_load_dword v29, off, s[0:3], s32 ; 4-byte Folded Reload
; GFX9-O0-NEXT: s_mov_b64 exec, s[18:19]
; GFX9-O0-NEXT: s_waitcnt vmcnt(0)
-; GFX9-O0-NEXT: v_readlane_b32 s6, v30, 8
-; GFX9-O0-NEXT: v_readlane_b32 s7, v30, 9
-; GFX9-O0-NEXT: buffer_load_dword v6, off, s[0:3], s32 offset:216 ; 4-byte Folded Reload
-; GFX9-O0-NEXT: buffer_load_dword v7, off, s[0:3], s32 offset:220 ; 4-byte Folded Reload
+; GFX9-O0-NEXT: v_readlane_b32 s6, v29, 8
+; GFX9-O0-NEXT: v_readlane_b32 s7, v29, 9
+; GFX9-O0-NEXT: buffer_load_dword v10, off, s[0:3], s32 offset:216 ; 4-byte Folded Reload
+; GFX9-O0-NEXT: buffer_load_dword v11, off, s[0:3], s32 offset:220 ; 4-byte Folded Reload
; GFX9-O0-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:224 ; 4-byte Folded Reload
; GFX9-O0-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:228 ; 4-byte Folded Reload
-; GFX9-O0-NEXT: buffer_load_dword v22, off, s[0:3], s32 offset:232 ; 4-byte Folded Reload
-; GFX9-O0-NEXT: buffer_load_dword v23, off, s[0:3], s32 offset:236 ; 4-byte Folded Reload
-; GFX9-O0-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:240 ; 4-byte Folded Reload
-; GFX9-O0-NEXT: buffer_load_dword v3, off, s[0:3], s32 offset:244 ; 4-byte Folded Reload
-; GFX9-O0-NEXT: buffer_load_dword v16, off, s[0:3], s32 offset:248 ; 4-byte Folded Reload
-; GFX9-O0-NEXT: buffer_load_dword v17, off, s[0:3], s32 offset:252 ; 4-byte Folded Reload
-; GFX9-O0-NEXT: buffer_load_dword v8, off, s[0:3], s32 offset:256 ; 4-byte Folded Reload
-; GFX9-O0-NEXT: buffer_load_dword v9, off, s[0:3], s32 offset:260 ; 4-byte Folded Reload
-; GFX9-O0-NEXT: buffer_load_dword v26, off, s[0:3], s32 offset:264 ; 4-byte Folded Reload
-; GFX9-O0-NEXT: buffer_load_dword v27, off, s[0:3], s32 offset:268 ; 4-byte Folded Reload
-; GFX9-O0-NEXT: buffer_load_dword v24, off, s[0:3], s32 offset:272 ; 4-byte Folded Reload
-; GFX9-O0-NEXT: buffer_load_dword v25, off, s[0:3], s32 offset:276 ; 4-byte Folded Reload
+; GFX9-O0-NEXT: buffer_load_dword v24, off, s[0:3], s32 offset:232 ; 4-byte Folded Reload
+; GFX9-O0-NEXT: buffer_load_dword v25, off, s[0:3], s32 offset:236 ; 4-byte Folded Reload
+; GFX9-O0-NEXT: buffer_load_dword v16, off, s[0:3], s32 offset:240 ; 4-byte Folded Reload
+; GFX9-O0-NEXT: buffer_load_dword v17, off, s[0:3], s32 offset:244 ; 4-byte Folded Reload
+; GFX9-O0-NEXT: buffer_load_dword v8, off, s[0:3], s32 offset:248 ; 4-byte Folded Reload
+; GFX9-O0-NEXT: buffer_load_dword v9, off, s[0:3], s32 offset:252 ; 4-byte Folded Reload
+; GFX9-O0-NEXT: buffer_load_dword v12, off, s[0:3], s32 offset:256 ; 4-byte Folded Reload
+; GFX9-O0-NEXT: buffer_load_dword v13, off, s[0:3], s32 offset:260 ; 4-byte Folded Reload
+; GFX9-O0-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:264 ; 4-byte Folded Reload
+; GFX9-O0-NEXT: buffer_load_dword v3, off, s[0:3], s32 offset:268 ; 4-byte Folded Reload
+; GFX9-O0-NEXT: buffer_load_dword v26, off, s[0:3], s32 offset:272 ; 4-byte Folded Reload
+; GFX9-O0-NEXT: buffer_load_dword v27, off, s[0:3], s32 offset:276 ; 4-byte Folded Reload
; GFX9-O0-NEXT: buffer_load_dword v18, off, s[0:3], s32 offset:52 ; 4-byte Folded Reload
; GFX9-O0-NEXT: buffer_load_dword v19, off, s[0:3], s32 offset:56 ; 4-byte Folded Reload
; GFX9-O0-NEXT: buffer_load_dword v20, off, s[0:3], s32 offset:60 ; 4-byte Folded Reload
; GFX9-O0-NEXT: buffer_load_dword v21, off, s[0:3], s32 offset:64 ; 4-byte Folded Reload
-; GFX9-O0-NEXT: buffer_load_dword v14, off, s[0:3], s32 offset:280 ; 4-byte Folded Reload
-; GFX9-O0-NEXT: buffer_load_dword v15, off, s[0:3], s32 offset:284 ; 4-byte Folded Reload
-; GFX9-O0-NEXT: buffer_load_dword v11, off, s[0:3], s32 offset:288 ; 4-byte Folded Reload
-; GFX9-O0-NEXT: buffer_load_dword v12, off, s[0:3], s32 offset:292 ; 4-byte Folded Reload
+; GFX9-O0-NEXT: buffer_load_dword v22, off, s[0:3], s32 offset:280 ; 4-byte Folded Reload
+; GFX9-O0-NEXT: buffer_load_dword v23, off, s[0:3], s32 offset:284 ; 4-byte Folded Reload
+; GFX9-O0-NEXT: buffer_load_dword v4, off, s[0:3], s32 offset:288 ; 4-byte Folded Reload
+; GFX9-O0-NEXT: buffer_load_dword v5, off, s[0:3], s32 offset:292 ; 4-byte Folded Reload
; GFX9-O0-NEXT: s_mov_b32 s4, 63
-; GFX9-O0-NEXT: s_waitcnt vmcnt(16)
-; GFX9-O0-NEXT: v_lshrrev_b64 v[28:29], s4, v[2:3]
-; GFX9-O0-NEXT: v_mov_b32_e32 v5, v29
+; GFX9-O0-NEXT: s_waitcnt vmcnt(10)
+; GFX9-O0-NEXT: v_lshrrev_b64 v[6:7], s4, v[2:3]
+; GFX9-O0-NEXT: v_mov_b32_e32 v15, v7
; GFX9-O0-NEXT: s_mov_b32 s5, 1
-; GFX9-O0-NEXT: v_lshlrev_b64 v[22:23], s5, v[22:23]
-; GFX9-O0-NEXT: v_mov_b32_e32 v4, v23
-; GFX9-O0-NEXT: v_or_b32_e64 v4, v4, v5
-; GFX9-O0-NEXT: v_mov_b32_e32 v10, v28
-; GFX9-O0-NEXT: v_mov_b32_e32 v5, v22
-; GFX9-O0-NEXT: v_or_b32_e64 v22, v5, v10
-; GFX9-O0-NEXT: ; kill: def $vgpr22 killed $vgpr22 def $vgpr22_vgpr23 killed $exec
-; GFX9-O0-NEXT: v_mov_b32_e32 v23, v4
-; GFX9-O0-NEXT: v_lshlrev_b64 v[28:29], s5, v[2:3]
-; GFX9-O0-NEXT: v_lshrrev_b64 v[4:5], s4, v[6:7]
-; GFX9-O0-NEXT: v_mov_b32_e32 v2, v29
-; GFX9-O0-NEXT: v_mov_b32_e32 v3, v5
+; GFX9-O0-NEXT: s_waitcnt vmcnt(8)
+; GFX9-O0-NEXT: v_lshlrev_b64 v[26:27], s5, v[26:27]
+; GFX9-O0-NEXT: v_mov_b32_e32 v14, v27
+; GFX9-O0-NEXT: v_or_b32_e64 v14, v14, v15
+; GFX9-O0-NEXT: v_mov_b32_e32 v7, v6
+; GFX9-O0-NEXT: v_mov_b32_e32 v6, v26
+; GFX9-O0-NEXT: v_or_b32_e64 v6, v6, v7
+; GFX9-O0-NEXT: ; kill: def $vgpr6 killed $vgpr6 def $vgpr6_vgpr7 killed $exec
+; GFX9-O0-NEXT: v_mov_b32_e32 v7, v14
+; GFX9-O0-NEXT: v_lshlrev_b64 v[26:27], s5, v[2:3]
+; GFX9-O0-NEXT: v_lshrrev_b64 v[14:15], s4, v[10:11]
+; GFX9-O0-NEXT: v_mov_b32_e32 v2, v27
+; GFX9-O0-NEXT: v_mov_b32_e32 v3, v15
; GFX9-O0-NEXT: v_or_b32_e64 v2, v2, v3
-; GFX9-O0-NEXT: v_mov_b32_e32 v3, v28
-; GFX9-O0-NEXT: ; kill: def $vgpr4 killed $vgpr4 killed $vgpr4_vgpr5 killed $exec
-; GFX9-O0-NEXT: v_or_b32_e64 v4, v3, v4
-; GFX9-O0-NEXT: ; kill: def $vgpr4 killed $vgpr4 def $vgpr4_vgpr5 killed $exec
-; GFX9-O0-NEXT: v_mov_b32_e32 v5, v2
+; GFX9-O0-NEXT: v_mov_b32_e32 v3, v26
+; GFX9-O0-NEXT: ; kill: def $vgpr14 killed $vgpr14 killed $vgpr14_vgpr15 killed $exec
+; GFX9-O0-NEXT: v_or_b32_e64 v14, v3, v14
+; GFX9-O0-NEXT: ; kill: def $vgpr14 killed $vgpr14 def $vgpr14_vgpr15 killed $exec
+; GFX9-O0-NEXT: v_mov_b32_e32 v15, v2
; GFX9-O0-NEXT: v_lshlrev_b64 v[2:3], s5, v[0:1]
-; GFX9-O0-NEXT: v_lshlrev_b64 v[28:29], s5, v[6:7]
+; GFX9-O0-NEXT: v_lshlrev_b64 v[26:27], s5, v[10:11]
; GFX9-O0-NEXT: v_lshrrev_b64 v[0:1], s4, v[0:1]
-; GFX9-O0-NEXT: v_mov_b32_e32 v7, v1
-; GFX9-O0-NEXT: v_mov_b32_e32 v6, v29
-; GFX9-O0-NEXT: s_waitcnt vmcnt(10)
+; GFX9-O0-NEXT: v_mov_b32_e32 v11, v1
; GFX9-O0-NEXT: v_mov_b32_e32 v10, v27
-; GFX9-O0-NEXT: v_or3_b32 v6, v6, v7, v10
+; GFX9-O0-NEXT: v_mov_b32_e32 v28, v25
+; GFX9-O0-NEXT: v_or3_b32 v10, v10, v11, v28
; GFX9-O0-NEXT: v_mov_b32_e32 v1, v0
-; GFX9-O0-NEXT: v_mov_b32_e32 v0, v28
-; GFX9-O0-NEXT: v_mov_b32_e32 v7, v26
-; GFX9-O0-NEXT: v_or3_b32 v0, v0, v1, v7
+; GFX9-O0-NEXT: v_mov_b32_e32 v0, v26
+; GFX9-O0-NEXT: v_mov_b32_e32 v11, v24
+; GFX9-O0-NEXT: v_or3_b32 v0, v0, v1, v11
; GFX9-O0-NEXT: ; kill: def $vgpr0 killed $vgpr0 def $vgpr0_vgpr1 killed $exec
-; GFX9-O0-NEXT: v_mov_b32_e32 v1, v6
-; GFX9-O0-NEXT: v_mov_b32_e32 v7, v3
-; GFX9-O0-NEXT: s_waitcnt vmcnt(8)
-; GFX9-O0-NEXT: v_mov_b32_e32 v6, v25
-; GFX9-O0-NEXT: v_or_b32_e64 v6, v6, v7
+; GFX9-O0-NEXT: v_mov_b32_e32 v1, v10
+; GFX9-O0-NEXT: v_mov_b32_e32 v11, v3
+; GFX9-O0-NEXT: v_mov_b32_e32 v10, v17
+; GFX9-O0-NEXT: v_or_b32_e64 v10, v10, v11
; GFX9-O0-NEXT: v_mov_b32_e32 v3, v2
-; GFX9-O0-NEXT: v_mov_b32_e32 v2, v24
+; GFX9-O0-NEXT: v_mov_b32_e32 v2, v16
; GFX9-O0-NEXT: v_or_b32_e64 v2, v2, v3
; GFX9-O0-NEXT: ; kill: def $vgpr2 killed $vgpr2 def $vgpr2_vgpr3 killed $exec
-; GFX9-O0-NEXT: v_mov_b32_e32 v3, v6
-; GFX9-O0-NEXT: v_mov_b32_e32 v6, v4
-; GFX9-O0-NEXT: v_mov_b32_e32 v10, v5
-; GFX9-O0-NEXT: v_mov_b32_e32 v4, v22
-; GFX9-O0-NEXT: v_mov_b32_e32 v5, v23
+; GFX9-O0-NEXT: v_mov_b32_e32 v3, v10
+; GFX9-O0-NEXT: v_mov_b32_e32 v10, v14
+; GFX9-O0-NEXT: v_mov_b32_e32 v16, v15
+; GFX9-O0-NEXT: v_mov_b32_e32 v14, v6
+; GFX9-O0-NEXT: v_mov_b32_e32 v11, v7
; GFX9-O0-NEXT: s_waitcnt vmcnt(1)
-; GFX9-O0-NEXT: v_mov_b32_e32 v13, v11
-; GFX9-O0-NEXT: v_mov_b32_e32 v11, v14
-; GFX9-O0-NEXT: v_mov_b32_e32 v7, v15
-; GFX9-O0-NEXT: v_sub_co_u32_e32 v13, vcc, v13, v6
+; GFX9-O0-NEXT: v_mov_b32_e32 v7, v4
; GFX9-O0-NEXT: s_waitcnt vmcnt(0)
-; GFX9-O0-NEXT: v_subb_co_u32_e32 v12, vcc, v12, v10, vcc
-; GFX9-O0-NEXT: v_subb_co_u32_e32 v11, vcc, v11, v4, vcc
-; GFX9-O0-NEXT: v_subb_co_u32_e32 v7, vcc, v7, v5, vcc
+; GFX9-O0-NEXT: v_mov_b32_e32 v6, v5
+; GFX9-O0-NEXT: v_mov_b32_e32 v4, v22
+; GFX9-O0-NEXT: v_mov_b32_e32 v5, v23
+; GFX9-O0-NEXT: v_sub_co_u32_e32 v7, vcc, v7, v10
+; GFX9-O0-NEXT: v_subb_co_u32_e32 v6, vcc, v6, v16, vcc
+; GFX9-O0-NEXT: v_subb_co_u32_e32 v4, vcc, v4, v14, vcc
+; GFX9-O0-NEXT: v_subb_co_u32_e32 v6, vcc, v5, v11, vcc
; GFX9-O0-NEXT: ; implicit-def: $sgpr5
; GFX9-O0-NEXT: ; implicit-def: $sgpr5
-; GFX9-O0-NEXT: ; kill: def $vgpr11 killed $vgpr11 def $vgpr11_vgpr12 killed $exec
-; GFX9-O0-NEXT: v_mov_b32_e32 v12, v7
-; GFX9-O0-NEXT: v_ashrrev_i64 v[13:14], s4, v[11:12]
-; GFX9-O0-NEXT: v_mov_b32_e32 v7, v14
+; GFX9-O0-NEXT: ; kill: def $vgpr4 killed $vgpr4 def $vgpr4_vgpr5 killed $exec
+; GFX9-O0-NEXT: v_mov_b32_e32 v5, v6
+; GFX9-O0-NEXT: v_ashrrev_i64 v[5:6], s4, v[4:5]
+; GFX9-O0-NEXT: v_mov_b32_e32 v15, v6
; GFX9-O0-NEXT: s_mov_b64 s[4:5], 1
; GFX9-O0-NEXT: s_mov_b32 s8, s5
-; GFX9-O0-NEXT: v_and_b32_e64 v12, v7, s8
-; GFX9-O0-NEXT: v_mov_b32_e32 v11, v13
+; GFX9-O0-NEXT: v_and_b32_e64 v4, v15, s8
+; GFX9-O0-NEXT: v_mov_b32_e32 v17, v5
; GFX9-O0-NEXT: ; kill: def $sgpr4 killed $sgpr4 killed $sgpr4_sgpr5
-; GFX9-O0-NEXT: v_and_b32_e64 v14, v11, s4
-; GFX9-O0-NEXT: ; kill: def $vgpr14 killed $vgpr14 def $vgpr14_vgpr15 killed $exec
-; GFX9-O0-NEXT: v_mov_b32_e32 v15, v12
-; GFX9-O0-NEXT: v_mov_b32_e32 v12, 0
-; GFX9-O0-NEXT: v_mov_b32_e32 v13, 0
+; GFX9-O0-NEXT: v_and_b32_e64 v6, v17, s4
+; GFX9-O0-NEXT: ; kill: def $vgpr6 killed $vgpr6 def $vgpr6_vgpr7 killed $exec
+; GFX9-O0-NEXT: v_mov_b32_e32 v7, v4
+; GFX9-O0-NEXT: v_mov_b32_e32 v4, 0
+; GFX9-O0-NEXT: v_mov_b32_e32 v5, 0
; GFX9-O0-NEXT: v_mov_b32_e32 v22, v21
-; GFX9-O0-NEXT: v_and_b32_e64 v22, v7, v22
-; GFX9-O0-NEXT: v_and_b32_e64 v20, v11, v20
+; GFX9-O0-NEXT: v_and_b32_e64 v22, v15, v22
+; GFX9-O0-NEXT: v_and_b32_e64 v20, v17, v20
; GFX9-O0-NEXT: ; kill: def $vgpr20 killed $vgpr20 def $vgpr20_vgpr21 killed $exec
; GFX9-O0-NEXT: v_mov_b32_e32 v21, v22
; GFX9-O0-NEXT: v_mov_b32_e32 v22, v19
-; GFX9-O0-NEXT: v_and_b32_e64 v7, v7, v22
-; GFX9-O0-NEXT: v_and_b32_e64 v22, v11, v18
-; GFX9-O0-NEXT: ; kill: def $vgpr22 killed $vgpr22 def $vgpr22_vgpr23 killed $exec
-; GFX9-O0-NEXT: v_mov_b32_e32 v23, v7
-; GFX9-O0-NEXT: v_mov_b32_e32 v19, v22
-; GFX9-O0-NEXT: v_mov_b32_e32 v18, v23
-; GFX9-O0-NEXT: v_mov_b32_e32 v11, v20
-; GFX9-O0-NEXT: v_mov_b32_e32 v7, v21
-; GFX9-O0-NEXT: v_sub_co_u32_e32 v6, vcc, v6, v19
-; GFX9-O0-NEXT: v_subb_co_u32_e32 v10, vcc, v10, v18, vcc
-; GFX9-O0-NEXT: v_subb_co_u32_e32 v4, vcc, v4, v11, vcc
-; GFX9-O0-NEXT: v_subb_co_u32_e32 v7, vcc, v5, v7, vcc
+; GFX9-O0-NEXT: v_and_b32_e64 v15, v15, v22
+; GFX9-O0-NEXT: v_and_b32_e64 v17, v17, v18
+; GFX9-O0-NEXT: ; kill: def $vgpr17 killed $vgpr17 def $vgpr17_vgpr18 killed $exec
+; GFX9-O0-NEXT: v_mov_b32_e32 v18, v15
+; GFX9-O0-NEXT: v_mov_b32_e32 v19, v17
+; GFX9-O0-NEXT: ; kill: def $vgpr18 killed $vgpr18 killed $vgpr17_vgpr18 killed $exec
+; GFX9-O0-NEXT: v_mov_b32_e32 v17, v20
+; GFX9-O0-NEXT: v_mov_b32_e32 v15, v21
+; GFX9-O0-NEXT: v_sub_co_u32_e32 v10, vcc, v10, v19
+; GFX9-O0-NEXT: v_subb_co_u32_e32 v16, vcc, v16, v18, vcc
+; GFX9-O0-NEXT: v_subb_co_u32_e32 v14, vcc, v14, v17, vcc
+; GFX9-O0-NEXT: v_subb_co_u32_e32 v11, vcc, v11, v15, vcc
; GFX9-O0-NEXT: ; implicit-def: $sgpr4
; GFX9-O0-NEXT: ; implicit-def: $sgpr4
-; GFX9-O0-NEXT: ; kill: def $vgpr4 killed $vgpr4 def $vgpr4_vgpr5 killed $exec
-; GFX9-O0-NEXT: v_mov_b32_e32 v5, v7
+; GFX9-O0-NEXT: ; kill: def $vgpr14 killed $vgpr14 def $vgpr14_vgpr15 killed $exec
+; GFX9-O0-NEXT: v_mov_b32_e32 v15, v11
; GFX9-O0-NEXT: ; implicit-def: $sgpr4
; GFX9-O0-NEXT: ; implicit-def: $sgpr4
-; GFX9-O0-NEXT: ; kill: def $vgpr6 killed $vgpr6 def $vgpr6_vgpr7 killed $exec
-; GFX9-O0-NEXT: v_mov_b32_e32 v7, v10
+; GFX9-O0-NEXT: ; kill: def $vgpr10 killed $vgpr10 def $vgpr10_vgpr11 killed $exec
+; GFX9-O0-NEXT: v_mov_b32_e32 v11, v16
+; GFX9-O0-NEXT: buffer_store_dword v14, off, s[0:3], s32 offset:272 ; 4-byte Folded Spill
+; GFX9-O0-NEXT: s_nop 0
+; GFX9-O0-NEXT: buffer_store_dword v15, off, s[0:3], s32 offset:276 ; 4-byte Folded Spill
+; GFX9-O0-NEXT: buffer_store_dword v10, off, s[0:3], s32 offset:264 ; 4-byte Folded Spill
+; GFX9-O0-NEXT: s_nop 0
+; GFX9-O0-NEXT: buffer_store_dword v11, off, s[0:3], s32 offset:268 ; 4-byte Folded Spill
; GFX9-O0-NEXT: v_mov_b32_e32 v11, v8
; GFX9-O0-NEXT: ; kill: def $vgpr9 killed $vgpr9 killed $vgpr8_vgpr9 killed $exec
; GFX9-O0-NEXT: s_mov_b64 s[8:9], -1
; GFX9-O0-NEXT: s_mov_b32 s5, s8
; GFX9-O0-NEXT: s_mov_b32 s4, s9
-; GFX9-O0-NEXT: v_mov_b32_e32 v10, v16
-; GFX9-O0-NEXT: v_mov_b32_e32 v8, v17
-; GFX9-O0-NEXT: v_mov_b32_e32 v16, s5
-; GFX9-O0-NEXT: v_add_co_u32_e32 v19, vcc, v11, v16
-; GFX9-O0-NEXT: v_mov_b32_e32 v11, s4
-; GFX9-O0-NEXT: v_addc_co_u32_e32 v9, vcc, v9, v11, vcc
-; GFX9-O0-NEXT: v_mov_b32_e32 v11, s5
-; GFX9-O0-NEXT: v_addc_co_u32_e32 v16, vcc, v10, v11, vcc
-; GFX9-O0-NEXT: v_mov_b32_e32 v10, s4
-; GFX9-O0-NEXT: v_addc_co_u32_e32 v8, vcc, v8, v10, vcc
+; GFX9-O0-NEXT: v_mov_b32_e32 v8, v12
+; GFX9-O0-NEXT: v_mov_b32_e32 v10, v13
+; GFX9-O0-NEXT: v_mov_b32_e32 v12, s5
+; GFX9-O0-NEXT: v_add_co_u32_e32 v11, vcc, v11, v12
+; GFX9-O0-NEXT: v_mov_b32_e32 v12, s4
+; GFX9-O0-NEXT: v_addc_co_u32_e32 v9, vcc, v9, v12, vcc
+; GFX9-O0-NEXT: v_mov_b32_e32 v12, s5
+; GFX9-O0-NEXT: v_addc_co_u32_e32 v8, vcc, v8, v12, vcc
+; GFX9-O0-NEXT: v_mov_b32_e32 v12, s4
+; GFX9-O0-NEXT: v_addc_co_u32_e32 v10, vcc, v10, v12, vcc
; GFX9-O0-NEXT: ; implicit-def: $sgpr4
; GFX9-O0-NEXT: ; implicit-def: $sgpr4
-; GFX9-O0-NEXT: ; kill: def $vgpr19 killed $vgpr19 def $vgpr19_vgpr20 killed $exec
-; GFX9-O0-NEXT: v_mov_b32_e32 v20, v9
+; GFX9-O0-NEXT: ; kill: def $vgpr11 killed $vgpr11 def $vgpr11_vgpr12 killed $exec
+; GFX9-O0-NEXT: v_mov_b32_e32 v12, v9
; GFX9-O0-NEXT: ; implicit-def: $sgpr4
; GFX9-O0-NEXT: ; implicit-def: $sgpr4
-; GFX9-O0-NEXT: ; kill: def $vgpr16 killed $vgpr16 def $vgpr16_vgpr17 killed $exec
-; GFX9-O0-NEXT: v_mov_b32_e32 v17, v8
-; GFX9-O0-NEXT: v_mov_b32_e32 v8, v16
-; GFX9-O0-NEXT: v_mov_b32_e32 v9, v17
-; GFX9-O0-NEXT: v_mov_b32_e32 v10, v19
-; GFX9-O0-NEXT: v_mov_b32_e32 v11, v20
-; GFX9-O0-NEXT: v_mov_b32_e32 v21, v17
-; GFX9-O0-NEXT: v_mov_b32_e32 v18, v20
-; GFX9-O0-NEXT: v_or_b32_e64 v18, v18, v21
-; GFX9-O0-NEXT: v_mov_b32_e32 v17, v16
-; GFX9-O0-NEXT: v_mov_b32_e32 v16, v19
-; GFX9-O0-NEXT: v_or_b32_e64 v16, v16, v17
-; GFX9-O0-NEXT: ; kill: def $vgpr16 killed $vgpr16 def $vgpr16_vgpr17 killed $exec
-; GFX9-O0-NEXT: v_mov_b32_e32 v17, v18
-; GFX9-O0-NEXT: v_cmp_eq_u64_e64 s[4:5], v[16:17], v[12:13]
+; GFX9-O0-NEXT: ; kill: def $vgpr8 killed $vgpr8 def $vgpr8_vgpr9 killed $exec
+; GFX9-O0-NEXT: v_mov_b32_e32 v9, v10
+; GFX9-O0-NEXT: v_mov_b32_e32 v14, v9
+; GFX9-O0-NEXT: v_mov_b32_e32 v13, v8
+; GFX9-O0-NEXT: buffer_store_dword v13, off, s[0:3], s32 offset:256 ; 4-byte Folded Spill
+; GFX9-O0-NEXT: s_nop 0
+; GFX9-O0-NEXT: buffer_store_dword v14, off, s[0:3], s32 offset:260 ; 4-byte Folded Spill
+; GFX9-O0-NEXT: v_mov_b32_e32 v14, v12
+; GFX9-O0-NEXT: v_mov_b32_e32 v13, v11
+; GFX9-O0-NEXT: buffer_store_dword v13, off, s[0:3], s32 offset:248 ; 4-byte Folded Spill
+; GFX9-O0-NEXT: s_nop 0
+; GFX9-O0-NEXT: buffer_store_dword v14, off, s[0:3], s32 offset:252 ; 4-byte Folded Spill
+; GFX9-O0-NEXT: v_mov_b32_e32 v13, v9
+; GFX9-O0-NEXT: v_mov_b32_e32 v10, v12
+; GFX9-O0-NEXT: v_or_b32_e64 v10, v10, v13
+; GFX9-O0-NEXT: v_mov_b32_e32 v9, v8
+; GFX9-O0-NEXT: v_mov_b32_e32 v8, v11
+; GFX9-O0-NEXT: v_or_b32_e64 v8, v8, v9
+; GFX9-O0-NEXT: ; kill: def $vgpr8 killed $vgpr8 def $vgpr8_vgpr9 killed $exec
+; GFX9-O0-NEXT: v_mov_b32_e32 v9, v10
+; GFX9-O0-NEXT: v_cmp_eq_u64_e64 s[4:5], v[8:9], v[4:5]
; GFX9-O0-NEXT: s_or_b64 s[4:5], s[4:5], s[6:7]
-; GFX9-O0-NEXT: v_mov_b32_e32 v17, v3
-; GFX9-O0-NEXT: v_mov_b32_e32 v16, v2
-; GFX9-O0-NEXT: buffer_store_dword v16, off, s[0:3], s32 offset:136 ; 4-byte Folded Spill
-; GFX9-O0-NEXT: s_nop 0
-; GFX9-O0-NEXT: buffer_store_dword v17, off, s[0:3], s32 offset:140 ; 4-byte Folded Spill
-; GFX9-O0-NEXT: v_mov_b32_e32 v17, v1
-; GFX9-O0-NEXT: v_mov_b32_e32 v16, v0
-; GFX9-O0-NEXT: buffer_store_dword v16, off, s[0:3], s32 offset:144 ; 4-byte Folded Spill
-; GFX9-O0-NEXT: s_nop 0
-; GFX9-O0-NEXT: buffer_store_dword v17, off, s[0:3], s32 offset:148 ; 4-byte Folded Spill
-; GFX9-O0-NEXT: v_mov_b32_e32 v17, v15
-; GFX9-O0-NEXT: v_mov_b32_e32 v16, v14
-; GFX9-O0-NEXT: buffer_store_dword v16, off, s[0:3], s32 offset:152 ; 4-byte Folded Spill
-; GFX9-O0-NEXT: s_nop 0
-; GFX9-O0-NEXT: buffer_store_dword v17, off, s[0:3], s32 offset:156 ; 4-byte Folded Spill
-; GFX9-O0-NEXT: v_mov_b32_e32 v17, v13
-; GFX9-O0-NEXT: v_mov_b32_e32 v16, v12
-; GFX9-O0-NEXT: buffer_store_dword v16, off, s[0:3], s32 offset:160 ; 4-byte Folded Spill
-; GFX9-O0-NEXT: s_nop 0
-; GFX9-O0-NEXT: buffer_store_dword v17, off, s[0:3], s32 offset:164 ; 4-byte Folded Spill
-; GFX9-O0-NEXT: s_mov_b64 s[6:7], s[4:5]
-; GFX9-O0-NEXT: v_writelane_b32 v30, s6, 4
-; GFX9-O0-NEXT: v_writelane_b32 v30, s7, 5
-; GFX9-O0-NEXT: s_mov_b64 s[6:7], s[4:5]
-; GFX9-O0-NEXT: v_writelane_b32 v30, s6, 8
-; GFX9-O0-NEXT: v_writelane_b32 v30, s7, 9
-; GFX9-O0-NEXT: s_or_saveexec_b64 s[18:19], -1
-; GFX9-O0-NEXT: buffer_store_dword v30, off, s[0:3], s32 ; 4-byte Folded Spill
-; GFX9-O0-NEXT: s_mov_b64 exec, s[18:19]
-; GFX9-O0-NEXT: buffer_store_dword v14, off, s[0:3], s32 offset:272 ; 4-byte Folded Spill
+; GFX9-O0-NEXT: v_mov_b32_e32 v9, v3
+; GFX9-O0-NEXT: v_mov_b32_e32 v8, v2
+; GFX9-O0-NEXT: buffer_store_dword v8, off, s[0:3], s32 offset:136 ; 4-byte Folded Spill
; GFX9-O0-NEXT: s_nop 0
-; GFX9-O0-NEXT: buffer_store_dword v15, off, s[0:3], s32 offset:276 ; 4-byte Folded Spill
-; GFX9-O0-NEXT: buffer_store_dword v12, off, s[0:3], s32 offset:264 ; 4-byte Folded Spill
+; GFX9-O0-NEXT: buffer_store_dword v9, off, s[0:3], s32 offset:140 ; 4-byte Folded Spill
+; GFX9-O0-NEXT: v_mov_b32_e32 v9, v1
+; GFX9-O0-NEXT: v_mov_b32_e32 v8, v0
+; GFX9-O0-NEXT: buffer_store_dword v8, off, s[0:3], s32 offset:144 ; 4-byte Folded Spill
; GFX9-O0-NEXT: s_nop 0
-; GFX9-O0-NEXT: buffer_store_dword v13, off, s[0:3], s32 offset:268 ; 4-byte Folded Spill
-; GFX9-O0-NEXT: buffer_store_dword v10, off, s[0:3], s32 offset:256 ; 4-byte Folded Spill
+; GFX9-O0-NEXT: buffer_store_dword v9, off, s[0:3], s32 offset:148 ; 4-byte Folded Spill
+; GFX9-O0-NEXT: v_mov_b32_e32 v9, v7
+; GFX9-O0-NEXT: v_mov_b32_e32 v8, v6
+; GFX9-O0-NEXT: buffer_store_dword v8, off, s[0:3], s32 offset:152 ; 4-byte Folded Spill
; GFX9-O0-NEXT: s_nop 0
-; GFX9-O0-NEXT: buffer_store_dword v11, off, s[0:3], s32 offset:260 ; 4-byte Folded Spill
-; GFX9-O0-NEXT: buffer_store_dword v8, off, s[0:3], s32 offset:248 ; 4-byte Folded Spill
+; GFX9-O0-NEXT: buffer_store_dword v9, off, s[0:3], s32 offset:156 ; 4-byte Folded Spill
+; GFX9-O0-NEXT: v_mov_b32_e32 v9, v5
+; GFX9-O0-NEXT: v_mov_b32_e32 v8, v4
+; GFX9-O0-NEXT: buffer_store_dword v8, off, s[0:3], s32 offset:160 ; 4-byte Folded Spill
; GFX9-O0-NEXT: s_nop 0
-; GFX9-O0-NEXT: buffer_store_dword v9, off, s[0:3], s32 offset:252 ; 4-byte Folded Spill
+; GFX9-O0-NEXT: buffer_store_dword v9, off, s[0:3], s32 offset:164 ; 4-byte Folded Spill
+; GFX9-O0-NEXT: s_mov_b64 s[6:7], s[4:5]
+; GFX9-O0-NEXT: v_writelane_b32 v29, s6, 4
+; GFX9-O0-NEXT: v_writelane_b32 v29, s7, 5
+; GFX9-O0-NEXT: s_mov_b64 s[6:7], s[4:5]
+; GFX9-O0-NEXT: v_writelane_b32 v29, s6, 8
+; GFX9-O0-NEXT: v_writelane_b32 v29, s7, 9
+; GFX9-O0-NEXT: s_or_saveexec_b64 s[18:19], -1
+; GFX9-O0-NEXT: buffer_store_dword v29, off, s[0:3], s32 ; 4-byte Folded Spill
+; GFX9-O0-NEXT: s_mov_b64 exec, s[18:19]
; GFX9-O0-NEXT: buffer_store_dword v6, off, s[0:3], s32 offset:240 ; 4-byte Folded Spill
; GFX9-O0-NEXT: s_nop 0
; GFX9-O0-NEXT: buffer_store_dword v7, off, s[0:3], s32 offset:244 ; 4-byte Folded Spill
@@ -2312,128 +2310,128 @@ define i128 @v_urem_i128_vv(i128 %lhs, i128 %rhs) {
; GFX9-O0-NEXT: s_branch .LBB1_1
; GFX9-O0-NEXT: .LBB1_7: ; %udiv-preheader
; GFX9-O0-NEXT: s_or_saveexec_b64 s[18:19], -1
-; GFX9-O0-NEXT: buffer_load_dword v30, off, s[0:3], s32 ; 4-byte Folded Reload
+; GFX9-O0-NEXT: buffer_load_dword v29, off, s[0:3], s32 ; 4-byte Folded Reload
; GFX9-O0-NEXT: s_mov_b64 exec, s[18:19]
; GFX9-O0-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:296 ; 4-byte Folded Reload
; GFX9-O0-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:300 ; 4-byte Folded Reload
; GFX9-O0-NEXT: buffer_load_dword v2, off, s[0:3], s32 offset:304 ; 4-byte Folded Reload
; GFX9-O0-NEXT: buffer_load_dword v3, off, s[0:3], s32 offset:308 ; 4-byte Folded Reload
-; GFX9-O0-NEXT: buffer_load_dword v8, off, s[0:3], s32 offset:312 ; 4-byte Folded Reload
-; GFX9-O0-NEXT: buffer_load_dword v9, off, s[0:3], s32 offset:316 ; 4-byte Folded Reload
-; GFX9-O0-NEXT: buffer_load_dword v10, off, s[0:3], s32 offset:320 ; 4-byte Folded Reload
-; GFX9-O0-NEXT: buffer_load_dword v11, off, s[0:3], s32 offset:324 ; 4-byte Folded Reload
-; GFX9-O0-NEXT: buffer_load_dword v16, off, s[0:3], s32 offset:60 ; 4-byte Folded Reload
-; GFX9-O0-NEXT: buffer_load_dword v17, off, s[0:3], s32 offset:64 ; 4-byte Folded Reload
-; GFX9-O0-NEXT: buffer_load_dword v12, off, s[0:3], s32 offset:52 ; 4-byte Folded Reload
-; GFX9-O0-NEXT: buffer_load_dword v13, off, s[0:3], s32 offset:56 ; 4-byte Folded Reload
-; GFX9-O0-NEXT: buffer_load_dword v14, off, s[0:3], s32 offset:44 ; 4-byte Folded Reload
-; GFX9-O0-NEXT: buffer_load_dword v15, off, s[0:3], s32 offset:48 ; 4-byte Folded Reload
+; GFX9-O0-NEXT: buffer_load_dword v4, off, s[0:3], s32 offset:312 ; 4-byte Folded Reload
+; GFX9-O0-NEXT: buffer_load_dword v5, off, s[0:3], s32 offset:316 ; 4-byte Folded Reload
+; GFX9-O0-NEXT: buffer_load_dword v6, off, s[0:3], s32 offset:320 ; 4-byte Folded Reload
+; GFX9-O0-NEXT: buffer_load_dword v7, off, s[0:3], s32 offset:324 ; 4-byte Folded Reload
+; GFX9-O0-NEXT: buffer_load_dword v12, off, s[0:3], s32 offset:60 ; 4-byte Folded Reload
+; GFX9-O0-NEXT: buffer_load_dword v13, off, s[0:3], s32 offset:64 ; 4-byte Folded Reload
+; GFX9-O0-NEXT: buffer_load_dword v9, off, s[0:3], s32 offset:52 ; 4-byte Folded Reload
+; GFX9-O0-NEXT: buffer_load_dword v10, off, s[0:3], s32 offset:56 ; 4-byte Folded Reload
+; GFX9-O0-NEXT: buffer_load_dword v16, off, s[0:3], s32 offset:44 ; 4-byte Folded Reload
+; GFX9-O0-NEXT: buffer_load_dword v17, off, s[0:3], s32 offset:48 ; 4-byte Folded Reload
; GFX9-O0-NEXT: buffer_load_dword v18, off, s[0:3], s32 offset:36 ; 4-byte Folded Reload
; GFX9-O0-NEXT: buffer_load_dword v19, off, s[0:3], s32 offset:40 ; 4-byte Folded Reload
; GFX9-O0-NEXT: s_waitcnt vmcnt(9)
-; GFX9-O0-NEXT: v_mov_b32_e32 v4, v10
+; GFX9-O0-NEXT: v_mov_b32_e32 v8, v6
; GFX9-O0-NEXT: s_waitcnt vmcnt(0)
-; GFX9-O0-NEXT: v_lshrrev_b64 v[6:7], v4, v[18:19]
-; GFX9-O0-NEXT: v_mov_b32_e32 v5, v7
+; GFX9-O0-NEXT: v_lshrrev_b64 v[14:15], v8, v[18:19]
+; GFX9-O0-NEXT: v_mov_b32_e32 v11, v15
; GFX9-O0-NEXT: s_mov_b32 s4, 64
-; GFX9-O0-NEXT: v_sub_u32_e64 v20, s4, v4
-; GFX9-O0-NEXT: v_lshlrev_b64 v[20:21], v20, v[14:15]
+; GFX9-O0-NEXT: v_sub_u32_e64 v20, s4, v8
+; GFX9-O0-NEXT: v_lshlrev_b64 v[20:21], v20, v[16:17]
; GFX9-O0-NEXT: v_mov_b32_e32 v22, v21
-; GFX9-O0-NEXT: v_or_b32_e64 v5, v5, v22
-; GFX9-O0-NEXT: ; kill: def $vgpr6 killed $vgpr6 killed $vgpr6_vgpr7 killed $exec
-; GFX9-O0-NEXT: v_mov_b32_e32 v7, v20
-; GFX9-O0-NEXT: v_or_b32_e64 v6, v6, v7
-; GFX9-O0-NEXT: ; kill: def $vgpr6 killed $vgpr6 def $vgpr6_vgpr7 killed $exec
-; GFX9-O0-NEXT: v_mov_b32_e32 v7, v5
-; GFX9-O0-NEXT: v_mov_b32_e32 v22, v7
-; GFX9-O0-NEXT: v_cmp_lt_u32_e64 s[6:7], v4, s4
-; GFX9-O0-NEXT: v_sub_u32_e64 v5, v4, s4
-; GFX9-O0-NEXT: v_lshrrev_b64 v[20:21], v5, v[14:15]
-; GFX9-O0-NEXT: v_mov_b32_e32 v5, v21
-; GFX9-O0-NEXT: v_cndmask_b32_e64 v5, v5, v22, s[6:7]
+; GFX9-O0-NEXT: v_or_b32_e64 v11, v11, v22
+; GFX9-O0-NEXT: ; kill: def $vgpr14 killed $vgpr14 killed $vgpr14_vgpr15 killed $exec
+; GFX9-O0-NEXT: v_mov_b32_e32 v15, v20
+; GFX9-O0-NEXT: v_or_b32_e64 v14, v14, v15
+; GFX9-O0-NEXT: ; kill: def $vgpr14 killed $vgpr14 def $vgpr14_vgpr15 killed $exec
+; GFX9-O0-NEXT: v_mov_b32_e32 v15, v11
+; GFX9-O0-NEXT: v_mov_b32_e32 v22, v15
+; GFX9-O0-NEXT: v_cmp_lt_u32_e64 s[6:7], v8, s4
+; GFX9-O0-NEXT: v_sub_u32_e64 v11, v8, s4
+; GFX9-O0-NEXT: v_lshrrev_b64 v[20:21], v11, v[16:17]
+; GFX9-O0-NEXT: v_mov_b32_e32 v11, v21
+; GFX9-O0-NEXT: v_cndmask_b32_e64 v11, v11, v22, s[6:7]
; GFX9-O0-NEXT: s_mov_b32 s4, 0
-; GFX9-O0-NEXT: v_cmp_eq_u32_e64 s[4:5], v4, s4
+; GFX9-O0-NEXT: v_cmp_eq_u32_e64 s[4:5], v8, s4
; GFX9-O0-NEXT: v_mov_b32_e32 v22, v19
-; GFX9-O0-NEXT: v_cndmask_b32_e64 v5, v5, v22, s[4:5]
-; GFX9-O0-NEXT: v_mov_b32_e32 v7, v6
-; GFX9-O0-NEXT: v_mov_b32_e32 v6, v20
-; GFX9-O0-NEXT: v_cndmask_b32_e64 v6, v6, v7, s[6:7]
-; GFX9-O0-NEXT: v_mov_b32_e32 v7, v18
-; GFX9-O0-NEXT: v_cndmask_b32_e64 v6, v6, v7, s[4:5]
+; GFX9-O0-NEXT: v_cndmask_b32_e64 v11, v11, v22, s[4:5]
+; GFX9-O0-NEXT: v_mov_b32_e32 v15, v14
+; GFX9-O0-NEXT: v_mov_b32_e32 v14, v20
+; GFX9-O0-NEXT: v_cndmask_b32_e64 v14, v14, v15, s[6:7]
+; GFX9-O0-NEXT: v_mov_b32_e32 v15, v18
+; GFX9-O0-NEXT: v_cndmask_b32_e64 v14, v14, v15, s[4:5]
; GFX9-O0-NEXT: ; implicit-def: $sgpr4
; GFX9-O0-NEXT: ; implicit-def: $sgpr4
-; GFX9-O0-NEXT: ; kill: def $vgpr6 killed $vgpr6 def $vgpr6_vgpr7 killed $exec
-; GFX9-O0-NEXT: v_mov_b32_e32 v7, v5
-; GFX9-O0-NEXT: v_lshrrev_b64 v[4:5], v4, v[14:15]
-; GFX9-O0-NEXT: v_mov_b32_e32 v15, v5
+; GFX9-O0-NEXT: ; kill: def $vgpr14 killed $vgpr14 def $vgpr14_vgpr15 killed $exec
+; GFX9-O0-NEXT: v_mov_b32_e32 v15, v11
+; GFX9-O0-NEXT: v_lshrrev_b64 v[16:17], v8, v[16:17]
+; GFX9-O0-NEXT: v_mov_b32_e32 v11, v17
; GFX9-O0-NEXT: s_mov_b64 s[4:5], 0
; GFX9-O0-NEXT: s_mov_b32 s8, s5
-; GFX9-O0-NEXT: v_mov_b32_e32 v14, s8
-; GFX9-O0-NEXT: v_cndmask_b32_e64 v14, v14, v15, s[6:7]
-; GFX9-O0-NEXT: v_mov_b32_e32 v5, v4
+; GFX9-O0-NEXT: v_mov_b32_e32 v8, s8
+; GFX9-O0-NEXT: v_cndmask_b32_e64 v8, v8, v11, s[6:7]
+; GFX9-O0-NEXT: ; kill: def $vgpr16 killed $vgpr16 killed $vgpr16_vgpr17 killed $exec
; GFX9-O0-NEXT: s_mov_b32 s8, s4
-; GFX9-O0-NEXT: v_mov_b32_e32 v4, s8
-; GFX9-O0-NEXT: v_cndmask_b32_e64 v4, v4, v5, s[6:7]
+; GFX9-O0-NEXT: v_mov_b32_e32 v11, s8
+; GFX9-O0-NEXT: v_cndmask_b32_e64 v16, v11, v16, s[6:7]
; GFX9-O0-NEXT: ; implicit-def: $sgpr6
; GFX9-O0-NEXT: ; implicit-def: $sgpr6
-; GFX9-O0-NEXT: ; kill: def $vgpr4 killed $vgpr4 def $vgpr4_vgpr5 killed $exec
-; GFX9-O0-NEXT: v_mov_b32_e32 v5, v14
-; GFX9-O0-NEXT: v_mov_b32_e32 v15, v12
-; GFX9-O0-NEXT: v_mov_b32_e32 v12, v13
+; GFX9-O0-NEXT: ; kill: def $vgpr16 killed $vgpr16 def $vgpr16_vgpr17 killed $exec
+; GFX9-O0-NEXT: v_mov_b32_e32 v17, v8
+; GFX9-O0-NEXT: buffer_store_dword v16, off, s[0:3], s32 offset:272 ; 4-byte Folded Spill
+; GFX9-O0-NEXT: s_nop 0
+; GFX9-O0-NEXT: buffer_store_dword v17, off, s[0:3], s32 offset:276 ; 4-byte Folded Spill
+; GFX9-O0-NEXT: buffer_store_dword v14, off, s[0:3], s32 offset:264 ; 4-byte Folded Spill
+; GFX9-O0-NEXT: s_nop 0
+; GFX9-O0-NEXT: buffer_store_dword v15, off, s[0:3], s32 offset:268 ; 4-byte Folded Spill
+; GFX9-O0-NEXT: v_mov_b32_e32 v8, v9
+; GFX9-O0-NEXT: v_mov_b32_e32 v11, v10
; GFX9-O0-NEXT: s_mov_b64 s[8:9], -1
; GFX9-O0-NEXT: s_mov_b32 s7, s8
; GFX9-O0-NEXT: s_mov_b32 s6, s9
-; GFX9-O0-NEXT: v_mov_b32_e32 v14, v16
-; GFX9-O0-NEXT: v_mov_b32_e32 v13, v17
-; GFX9-O0-NEXT: v_mov_b32_e32 v16, s7
-; GFX9-O0-NEXT: v_add_co_u32_e32 v16, vcc, v15, v16
-; GFX9-O0-NEXT: v_mov_b32_e32 v15, s6
-; GFX9-O0-NEXT: v_addc_co_u32_e32 v12, vcc, v12, v15, vcc
-; GFX9-O0-NEXT: v_mov_b32_e32 v15, s7
-; GFX9-O0-NEXT: v_addc_co_u32_e32 v18, vcc, v14, v15, vcc
-; GFX9-O0-NEXT: v_mov_b32_e32 v14, s6
-; GFX9-O0-NEXT: v_addc_co_u32_e32 v13, vcc, v13, v14, vcc
+; GFX9-O0-NEXT: v_mov_b32_e32 v10, v12
+; GFX9-O0-NEXT: v_mov_b32_e32 v9, v13
+; GFX9-O0-NEXT: v_mov_b32_e32 v12, s7
+; GFX9-O0-NEXT: v_add_co_u32_e32 v8, vcc, v8, v12
+; GFX9-O0-NEXT: v_mov_b32_e32 v12, s6
+; GFX9-O0-NEXT: v_addc_co_u32_e32 v12, vcc, v11, v12, vcc
+; GFX9-O0-NEXT: v_mov_b32_e32 v11, s7
+; GFX9-O0-NEXT: v_addc_co_u32_e32 v10, vcc, v10, v11, vcc
+; GFX9-O0-NEXT: v_mov_b32_e32 v11, s6
+; GFX9-O0-NEXT: v_addc_co_u32_e32 v9, vcc, v9, v11, vcc
; GFX9-O0-NEXT: ; implicit-def: $sgpr6
; GFX9-O0-NEXT: ; implicit-def: $sgpr6
-; GFX9-O0-NEXT: ; kill: def $vgpr18 killed $vgpr18 def $vgpr18_vgpr19 killed $exec
-; GFX9-O0-NEXT: v_mov_b32_e32 v19, v13
+; GFX9-O0-NEXT: ; kill: def $vgpr10 killed $vgpr10 def $vgpr10_vgpr11 killed $exec
+; GFX9-O0-NEXT: v_mov_b32_e32 v11, v9
; GFX9-O0-NEXT: ; implicit-def: $sgpr6
; GFX9-O0-NEXT: ; implicit-def: $sgpr6
-; GFX9-O0-NEXT: ; kill: def $vgpr16 killed $vgpr16 def $vgpr16_vgpr17 killed $exec
-; GFX9-O0-NEXT: v_mov_b32_e32 v17, v12
+; GFX9-O0-NEXT: ; kill: def $vgpr8 killed $vgpr8 def $vgpr8_vgpr9 killed $exec
+; GFX9-O0-NEXT: v_mov_b32_e32 v9, v12
; GFX9-O0-NEXT: v_mov_b32_e32 v13, s5
; GFX9-O0-NEXT: v_mov_b32_e32 v12, s4
-; GFX9-O0-NEXT: v_mov_b32_e32 v15, s5
-; GFX9-O0-NEXT: v_mov_b32_e32 v14, s4
-; GFX9-O0-NEXT: buffer_store_dword v18, off, s[0:3], s32 offset:280 ; 4-byte Folded Spill
-; GFX9-O0-NEXT: s_nop 0
-; GFX9-O0-NEXT: buffer_store_dword v19, off, s[0:3], s32 offset:284 ; 4-byte Folded Spill
-; GFX9-O0-NEXT: buffer_store_dword v16, off, s[0:3], s32 offset:288 ; 4-byte Folded Spill
-; GFX9-O0-NEXT: s_nop 0
-; GFX9-O0-NEXT: buffer_store_dword v17, off, s[0:3], s32 offset:292 ; 4-byte Folded Spill
-; GFX9-O0-NEXT: v_writelane_b32 v30, s4, 8
-; GFX9-O0-NEXT: v_writelane_b32 v30, s5, 9
-; GFX9-O0-NEXT: s_or_saveexec_b64 s[18:19], -1
-; GFX9-O0-NEXT: buffer_store_dword v30, off, s[0:3], s32 ; 4-byte Folded Spill
-; GFX9-O0-NEXT: s_mov_b64 exec, s[18:19]
-; GFX9-O0-NEXT: buffer_store_dword v14, off, s[0:3], s32 offset:272 ; 4-byte Folded Spill
+; GFX9-O0-NEXT: buffer_store_dword v12, off, s[0:3], s32 offset:232 ; 4-byte Folded Spill
; GFX9-O0-NEXT: s_nop 0
-; GFX9-O0-NEXT: buffer_store_dword v15, off, s[0:3], s32 offset:276 ; 4-byte Folded Spill
-; GFX9-O0-NEXT: buffer_store_dword v12, off, s[0:3], s32 offset:264 ; 4-byte Folded Spill
+; GFX9-O0-NEXT: buffer_store_dword v13, off, s[0:3], s32 offset:236 ; 4-byte Folded Spill
+; GFX9-O0-NEXT: v_mov_b32_e32 v13, s5
+; GFX9-O0-NEXT: v_mov_b32_e32 v12, s4
+; GFX9-O0-NEXT: buffer_store_dword v12, off, s[0:3], s32 offset:240 ; 4-byte Folded Spill
; GFX9-O0-NEXT: s_nop 0
-; GFX9-O0-NEXT: buffer_store_dword v13, off, s[0:3], s32 offset:268 ; 4-byte Folded Spill
-; GFX9-O0-NEXT: buffer_store_dword v10, off, s[0:3], s32 offset:256 ; 4-byte Folded Spill
+; GFX9-O0-NEXT: buffer_store_dword v13, off, s[0:3], s32 offset:244 ; 4-byte Folded Spill
+; GFX9-O0-NEXT: buffer_store_dword v10, off, s[0:3], s32 offset:280 ; 4-byte Folded Spill
; GFX9-O0-NEXT: s_nop 0
-; GFX9-O0-NEXT: buffer_store_dword v11, off, s[0:3], s32 offset:260 ; 4-byte Folded Spill
-; GFX9-O0-NEXT: buffer_store_dword v8, off, s[0:3], s32 offset:248 ; 4-byte Folded Spill
+; GFX9-O0-NEXT: buffer_store_dword v11, off, s[0:3], s32 offset:284 ; 4-byte Folded Spill
+; GFX9-O0-NEXT: buffer_store_dword v8, off, s[0:3], s32 offset:288 ; 4-byte Folded Spill
; GFX9-O0-NEXT: s_nop 0
-; GFX9-O0-NEXT: buffer_store_dword v9, off, s[0:3], s32 offset:252 ; 4-byte Folded Spill
-; GFX9-O0-NEXT: buffer_store_dword v6, off, s[0:3], s32 offset:240 ; 4-byte Folded Spill
+; GFX9-O0-NEXT: buffer_store_dword v9, off, s[0:3], s32 offset:292 ; 4-byte Folded Spill
+; GFX9-O0-NEXT: v_writelane_b32 v29, s4, 8
+; GFX9-O0-NEXT: v_writelane_b32 v29, s5, 9
+; GFX9-O0-NEXT: s_or_saveexec_b64 s[18:19], -1
+; GFX9-O0-NEXT: buffer_store_dword v29, off, s[0:3], s32 ; 4-byte Folded Spill
+; GFX9-O0-NEXT: s_mov_b64 exec, s[18:19]
+; GFX9-O0-NEXT: buffer_store_dword v6, off, s[0:3], s32 offset:248 ; 4-byte Folded Spill
; GFX9-O0-NEXT: s_nop 0
-; GFX9-O0-NEXT: buffer_store_dword v7, off, s[0:3], s32 offset:244 ; 4-byte Folded Spill
-; GFX9-O0-NEXT: buffer_store_dword v4, off, s[0:3], s32 offset:232 ; 4-byte Folded Spill
+; GFX9-O0-NEXT: buffer_store_dword v7, off, s[0:3], s32 offset:252 ; 4-byte Folded Spill
+; GFX9-O0-NEXT: buffer_store_dword v4, off, s[0:3], s32 offset:256 ; 4-byte Folded Spill
; GFX9-O0-NEXT: s_nop 0
-; GFX9-O0-NEXT: buffer_store_dword v5, off, s[0:3], s32 offset:236 ; 4-byte Folded Spill
+; GFX9-O0-NEXT: buffer_store_dword v5, off, s[0:3], s32 offset:260 ; 4-byte Folded Spill
; GFX9-O0-NEXT: buffer_store_dword v2, off, s[0:3], s32 offset:224 ; 4-byte Folded Spill
; GFX9-O0-NEXT: s_nop 0
; GFX9-O0-NEXT: buffer_store_dword v3, off, s[0:3], s32 offset:228 ; 4-byte Folded Spill
@@ -2443,12 +2441,12 @@ define i128 @v_urem_i128_vv(i128 %lhs, i128 %rhs) {
; GFX9-O0-NEXT: s_branch .LBB1_6
; GFX9-O0-NEXT: .LBB1_8: ; %udiv-bb1
; GFX9-O0-NEXT: s_or_saveexec_b64 s[18:19], -1
-; GFX9-O0-NEXT: buffer_load_dword v30, off, s[0:3], s32 ; 4-byte Folded Reload
+; GFX9-O0-NEXT: buffer_load_dword v29, off, s[0:3], s32 ; 4-byte Folded Reload
; GFX9-O0-NEXT: s_mov_b64 exec, s[18:19]
-; GFX9-O0-NEXT: buffer_load_dword v6, off, s[0:3], s32 offset:36 ; 4-byte Folded Reload
-; GFX9-O0-NEXT: buffer_load_dword v7, off, s[0:3], s32 offset:40 ; 4-byte Folded Reload
-; GFX9-O0-NEXT: buffer_load_dword v10, off, s[0:3], s32 offset:44 ; 4-byte Folded Reload
-; GFX9-O0-NEXT: buffer_load_dword v11, off, s[0:3], s32 offset:48 ; 4-byte Folded Reload
+; GFX9-O0-NEXT: buffer_load_dword v9, off, s[0:3], s32 offset:36 ; 4-byte Folded Reload
+; GFX9-O0-NEXT: buffer_load_dword v10, off, s[0:3], s32 offset:40 ; 4-byte Folded Reload
+; GFX9-O0-NEXT: buffer_load_dword v11, off, s[0:3], s32 offset:44 ; 4-byte Folded Reload
+; GFX9-O0-NEXT: buffer_load_dword v12, off, s[0:3], s32 offset:48 ; 4-byte Folded Reload
; GFX9-O0-NEXT: buffer_load_dword v4, off, s[0:3], s32 offset:20 ; 4-byte Folded Reload
; GFX9-O0-NEXT: buffer_load_dword v5, off, s[0:3], s32 offset:24 ; 4-byte Folded Reload
; GFX9-O0-NEXT: buffer_load_dword v0, off, s[0:3], s32 offset:28 ; 4-byte Folded Reload
@@ -2461,118 +2459,118 @@ define i128 @v_urem_i128_vv(i128 %lhs, i128 %rhs) {
; GFX9-O0-NEXT: s_mov_b64 s[6:7], 0
; GFX9-O0-NEXT: s_mov_b32 s8, s6
; GFX9-O0-NEXT: s_mov_b32 s9, s7
-; GFX9-O0-NEXT: v_mov_b32_e32 v0, v4
-; GFX9-O0-NEXT: v_mov_b32_e32 v2, v5
+; GFX9-O0-NEXT: v_mov_b32_e32 v2, v4
+; GFX9-O0-NEXT: v_mov_b32_e32 v0, v5
; GFX9-O0-NEXT: v_mov_b32_e32 v4, s5
-; GFX9-O0-NEXT: v_add_co_u32_e32 v8, vcc, v3, v4
+; GFX9-O0-NEXT: v_add_co_u32_e32 v7, vcc, v3, v4
; GFX9-O0-NEXT: v_mov_b32_e32 v4, s4
; GFX9-O0-NEXT: s_waitcnt vmcnt(0)
; GFX9-O0-NEXT: v_addc_co_u32_e32 v1, vcc, v1, v4, vcc
; GFX9-O0-NEXT: v_mov_b32_e32 v4, s8
-; GFX9-O0-NEXT: v_addc_co_u32_e32 v0, vcc, v0, v4, vcc
-; GFX9-O0-NEXT: v_mov_b32_e32 v4, s9
-; GFX9-O0-NEXT: v_addc_co_u32_e32 v2, vcc, v2, v4, vcc
+; GFX9-O0-NEXT: v_addc_co_u32_e32 v4, vcc, v2, v4, vcc
+; GFX9-O0-NEXT: v_mov_b32_e32 v2, s9
+; GFX9-O0-NEXT: v_addc_co_u32_e32 v0, vcc, v0, v2, vcc
; GFX9-O0-NEXT: ; implicit-def: $sgpr4
; GFX9-O0-NEXT: ; implicit-def: $sgpr4
-; GFX9-O0-NEXT: ; kill: def $vgpr8 killed $vgpr8 def $vgpr8_vgpr9 killed $exec
-; GFX9-O0-NEXT: v_mov_b32_e32 v9, v1
+; GFX9-O0-NEXT: ; kill: def $vgpr7 killed $vgpr7 def $vgpr7_vgpr8 killed $exec
+; GFX9-O0-NEXT: v_mov_b32_e32 v8, v1
; GFX9-O0-NEXT: ; implicit-def: $sgpr4
; GFX9-O0-NEXT: ; implicit-def: $sgpr4
-; GFX9-O0-NEXT: ; kill: def $vgpr0 killed $vgpr0 def $vgpr0_vgpr1 killed $exec
-; GFX9-O0-NEXT: v_mov_b32_e32 v1, v2
-; GFX9-O0-NEXT: v_mov_b32_e32 v5, v1
-; GFX9-O0-NEXT: v_mov_b32_e32 v4, v0
-; GFX9-O0-NEXT: buffer_store_dword v4, off, s[0:3], s32 offset:312 ; 4-byte Folded Spill
+; GFX9-O0-NEXT: ; kill: def $vgpr4 killed $vgpr4 def $vgpr4_vgpr5 killed $exec
+; GFX9-O0-NEXT: v_mov_b32_e32 v5, v0
+; GFX9-O0-NEXT: v_mov_b32_e32 v0, v4
+; GFX9-O0-NEXT: v_mov_b32_e32 v1, v5
+; GFX9-O0-NEXT: buffer_store_dword v0, off, s[0:3], s32 offset:312 ; 4-byte Folded Spill
; GFX9-O0-NEXT: s_nop 0
-; GFX9-O0-NEXT: buffer_store_dword v5, off, s[0:3], s32 offset:316 ; 4-byte Folded Spill
-; GFX9-O0-NEXT: v_mov_b32_e32 v4, v8
-; GFX9-O0-NEXT: v_mov_b32_e32 v5, v9
-; GFX9-O0-NEXT: buffer_store_dword v4, off, s[0:3], s32 offset:320 ; 4-byte Folded Spill
+; GFX9-O0-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:316 ; 4-byte Folded Spill
+; GFX9-O0-NEXT: v_mov_b32_e32 v0, v7
+; GFX9-O0-NEXT: v_mov_b32_e32 v1, v8
+; GFX9-O0-NEXT: buffer_store_dword v0, off, s[0:3], s32 offset:320 ; 4-byte Folded Spill
; GFX9-O0-NEXT: s_nop 0
-; GFX9-O0-NEXT: buffer_store_dword v5, off, s[0:3], s32 offset:324 ; 4-byte Folded Spill
+; GFX9-O0-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:324 ; 4-byte Folded Spill
; GFX9-O0-NEXT: s_mov_b32 s4, 0x7f
; GFX9-O0-NEXT: v_sub_u32_e64 v2, s4, v3
-; GFX9-O0-NEXT: v_lshlrev_b64 v[4:5], v2, v[10:11]
-; GFX9-O0-NEXT: v_mov_b32_e32 v12, v5
+; GFX9-O0-NEXT: v_lshlrev_b64 v[0:1], v2, v[11:12]
+; GFX9-O0-NEXT: v_mov_b32_e32 v6, v1
; GFX9-O0-NEXT: s_mov_b32 s4, 64
; GFX9-O0-NEXT: v_sub_u32_e64 v13, s4, v2
-; GFX9-O0-NEXT: v_lshrrev_b64 v[13:14], v13, v[6:7]
+; GFX9-O0-NEXT: v_lshrrev_b64 v[13:14], v13, v[9:10]
; GFX9-O0-NEXT: v_mov_b32_e32 v15, v14
-; GFX9-O0-NEXT: v_or_b32_e64 v12, v12, v15
-; GFX9-O0-NEXT: ; kill: def $vgpr4 killed $vgpr4 killed $vgpr4_vgpr5 killed $exec
-; GFX9-O0-NEXT: v_mov_b32_e32 v5, v13
-; GFX9-O0-NEXT: v_or_b32_e64 v4, v4, v5
-; GFX9-O0-NEXT: ; kill: def $vgpr4 killed $vgpr4 def $vgpr4_vgpr5 killed $exec
-; GFX9-O0-NEXT: v_mov_b32_e32 v5, v12
-; GFX9-O0-NEXT: v_mov_b32_e32 v14, v5
+; GFX9-O0-NEXT: v_or_b32_e64 v6, v6, v15
+; GFX9-O0-NEXT: ; kill: def $vgpr0 killed $vgpr0 killed $vgpr0_vgpr1 killed $exec
+; GFX9-O0-NEXT: v_mov_b32_e32 v1, v13
+; GFX9-O0-NEXT: v_or_b32_e64 v0, v0, v1
+; GFX9-O0-NEXT: ; kill: def $vgpr0 killed $vgpr0 def $vgpr0_vgpr1 killed $exec
+; GFX9-O0-NEXT: v_mov_b32_e32 v1, v6
+; GFX9-O0-NEXT: v_mov_b32_e32 v6, v1
; GFX9-O0-NEXT: v_cmp_lt_u32_e64 s[4:5], v2, s4
; GFX9-O0-NEXT: s_mov_b32 s10, 63
; GFX9-O0-NEXT: v_sub_u32_e64 v3, s10, v3
-; GFX9-O0-NEXT: v_lshlrev_b64 v[12:13], v3, v[6:7]
-; GFX9-O0-NEXT: v_mov_b32_e32 v3, v13
-; GFX9-O0-NEXT: v_cndmask_b32_e64 v3, v3, v14, s[4:5]
+; GFX9-O0-NEXT: v_lshlrev_b64 v[13:14], v3, v[9:10]
+; GFX9-O0-NEXT: v_mov_b32_e32 v3, v14
+; GFX9-O0-NEXT: v_cndmask_b32_e64 v3, v3, v6, s[4:5]
; GFX9-O0-NEXT: s_mov_b32 s10, 0
; GFX9-O0-NEXT: v_cmp_eq_u32_e64 s[10:11], v2, s10
-; GFX9-O0-NEXT: v_mov_b32_e32 v14, v11
-; GFX9-O0-NEXT: v_cndmask_b32_e64 v3, v3, v14, s[10:11]
-; GFX9-O0-NEXT: v_mov_b32_e32 v5, v4
-; GFX9-O0-NEXT: v_mov_b32_e32 v4, v12
-; GFX9-O0-NEXT: v_cndmask_b32_e64 v4, v4, v5, s[4:5]
-; GFX9-O0-NEXT: v_mov_b32_e32 v5, v10
-; GFX9-O0-NEXT: v_cndmask_b32_e64 v4, v4, v5, s[10:11]
+; GFX9-O0-NEXT: v_mov_b32_e32 v6, v12
+; GFX9-O0-NEXT: v_cndmask_b32_e64 v3, v3, v6, s[10:11]
+; GFX9-O0-NEXT: v_mov_b32_e32 v1, v0
+; GFX9-O0-NEXT: v_mov_b32_e32 v0, v13
+; GFX9-O0-NEXT: v_cndmask_b32_e64 v0, v0, v1, s[4:5]
+; GFX9-O0-NEXT: v_mov_b32_e32 v1, v11
+; GFX9-O0-NEXT: v_cndmask_b32_e64 v0, v0, v1, s[10:11]
; GFX9-O0-NEXT: ; implicit-def: $sgpr10
; GFX9-O0-NEXT: ; implicit-def: $sgpr10
-; GFX9-O0-NEXT: ; kill: def $vgpr4 killed $vgpr4 def $vgpr4_vgpr5 killed $exec
-; GFX9-O0-NEXT: v_mov_b32_e32 v5, v3
-; GFX9-O0-NEXT: v_lshlrev_b64 v[6:7], v2, v[6:7]
-; GFX9-O0-NEXT: v_mov_b32_e32 v3, v7
-; GFX9-O0-NEXT: v_mov_b32_e32 v2, s9
+; GFX9-O0-NEXT: ; kill: def $vgpr0 killed $vgpr0 def $vgpr0_vgpr1 killed $exec
+; GFX9-O0-NEXT: v_mov_b32_e32 v1, v3
+; GFX9-O0-NEXT: v_lshlrev_b64 v[2:3], v2, v[9:10]
+; GFX9-O0-NEXT: v_mov_b32_e32 v9, v3
+; GFX9-O0-NEXT: v_mov_b32_e32 v6, s9
+; GFX9-O0-NEXT: v_cndmask_b32_e64 v6, v6, v9, s[4:5]
+; GFX9-O0-NEXT: v_mov_b32_e32 v3, v2
+; GFX9-O0-NEXT: v_mov_b32_e32 v2, s8
; GFX9-O0-NEXT: v_cndmask_b32_e64 v2, v2, v3, s[4:5]
-; GFX9-O0-NEXT: ; kill: def $vgpr6 killed $vgpr6 killed $vgpr6_vgpr7 killed $exec
-; GFX9-O0-NEXT: v_mov_b32_e32 v3, s8
-; GFX9-O0-NEXT: v_cndmask_b32_e64 v6, v3, v6, s[4:5]
; GFX9-O0-NEXT: ; implicit-def: $sgpr4
; GFX9-O0-NEXT: ; implicit-def: $sgpr4
-; GFX9-O0-NEXT: ; kill: def $vgpr6 killed $vgpr6 def $vgpr6_vgpr7 killed $exec
-; GFX9-O0-NEXT: v_mov_b32_e32 v7, v2
-; GFX9-O0-NEXT: buffer_store_dword v6, off, s[0:3], s32 offset:304 ; 4-byte Folded Spill
+; GFX9-O0-NEXT: ; kill: def $vgpr2 killed $vgpr2 def $vgpr2_vgpr3 killed $exec
+; GFX9-O0-NEXT: v_mov_b32_e32 v3, v6
+; GFX9-O0-NEXT: buffer_store_dword v2, off, s[0:3], s32 offset:304 ; 4-byte Folded Spill
; GFX9-O0-NEXT: s_nop 0
-; GFX9-O0-NEXT: buffer_store_dword v7, off, s[0:3], s32 offset:308 ; 4-byte Folded Spill
-; GFX9-O0-NEXT: buffer_store_dword v4, off, s[0:3], s32 offset:296 ; 4-byte Folded Spill
+; GFX9-O0-NEXT: buffer_store_dword v3, off, s[0:3], s32 offset:308 ; 4-byte Folded Spill
+; GFX9-O0-NEXT: buffer_store_dword v0, off, s[0:3], s32 offset:296 ; 4-byte Folded Spill
; GFX9-O0-NEXT: s_nop 0
-; GFX9-O0-NEXT: buffer_store_dword v5, off, s[0:3], s32 offset:300 ; 4-byte Folded Spill
-; GFX9-O0-NEXT: v_mov_b32_e32 v3, v1
-; GFX9-O0-NEXT: v_mov_b32_e32 v2, v9
-; GFX9-O0-NEXT: v_or_b32_e64 v2, v2, v3
-; GFX9-O0-NEXT: v_mov_b32_e32 v1, v0
-; GFX9-O0-NEXT: v_mov_b32_e32 v0, v8
-; GFX9-O0-NEXT: v_or_b32_e64 v0, v0, v1
-; GFX9-O0-NEXT: ; kill: def $vgpr0 killed $vgpr0 def $vgpr0_vgpr1 killed $exec
-; GFX9-O0-NEXT: v_mov_b32_e32 v1, v2
-; GFX9-O0-NEXT: v_cmp_ne_u64_e64 s[4:5], v[0:1], s[6:7]
-; GFX9-O0-NEXT: v_mov_b32_e32 v0, s6
-; GFX9-O0-NEXT: v_mov_b32_e32 v1, s7
-; GFX9-O0-NEXT: v_mov_b32_e32 v2, s6
-; GFX9-O0-NEXT: v_mov_b32_e32 v3, s7
-; GFX9-O0-NEXT: buffer_store_dword v6, off, s[0:3], s32 offset:128 ; 4-byte Folded Spill
-; GFX9-O0-NEXT: s_nop 0
-; GFX9-O0-NEXT: buffer_store_dword v7, off, s[0:3], s32 offset:132 ; 4-byte Folded Spill
+; GFX9-O0-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:300 ; 4-byte Folded Spill
+; GFX9-O0-NEXT: v_mov_b32_e32 v9, v5
+; GFX9-O0-NEXT: v_mov_b32_e32 v6, v8
+; GFX9-O0-NEXT: v_or_b32_e64 v6, v6, v9
+; GFX9-O0-NEXT: v_mov_b32_e32 v5, v4
+; GFX9-O0-NEXT: v_mov_b32_e32 v4, v7
+; GFX9-O0-NEXT: v_or_b32_e64 v4, v4, v5
+; GFX9-O0-NEXT: ; kill: def $vgpr4 killed $vgpr4 def $vgpr4_vgpr5 killed $exec
+; GFX9-O0-NEXT: v_mov_b32_e32 v5, v6
+; GFX9-O0-NEXT: v_cmp_ne_u64_e64 s[4:5], v[4:5], s[6:7]
+; GFX9-O0-NEXT: v_mov_b32_e32 v4, s6
+; GFX9-O0-NEXT: v_mov_b32_e32 v5, s7
+; GFX9-O0-NEXT: buffer_store_dword v4, off, s[0:3], s32 offset:128 ; 4-byte Folded Spill
+; GFX9-O0-NEXT: s_nop 0
+; GFX9-O0-NEXT: buffer_store_dword v5, off, s[0:3], s32 offset:132 ; 4-byte Folded Spill
+; GFX9-O0-NEXT: v_mov_b32_e32 v4, s6
+; GFX9-O0-NEXT: v_mov_b32_e32 v5, s7
; GFX9-O0-NEXT: buffer_store_dword v4, off, s[0:3], s32 offset:120 ; 4-byte Folded Spill
; GFX9-O0-NEXT: s_nop 0
; GFX9-O0-NEXT: buffer_store_dword v5, off, s[0:3], s32 offset:124 ; 4-byte Folded Spill
-; GFX9-O0-NEXT: buffer_store_dword v2, off, s[0:3], s32 offset:112 ; 4-byte Folded Spill
+; GFX9-O0-NEXT: buffer_store_dword v2, off, s[0:3], s32 offset:104 ; 4-byte Folded Spill
; GFX9-O0-NEXT: s_nop 0
-; GFX9-O0-NEXT: buffer_store_dword v3, off, s[0:3], s32 offset:116 ; 4-byte Folded Spill
-; GFX9-O0-NEXT: buffer_store_dword v0, off, s[0:3], s32 offset:104 ; 4-byte Folded Spill
+; GFX9-O0-NEXT: buffer_store_dword v3, off, s[0:3], s32 offset:108 ; 4-byte Folded Spill
+; GFX9-O0-NEXT: buffer_store_dword v0, off, s[0:3], s32 offset:112 ; 4-byte Folded Spill
; GFX9-O0-NEXT: s_nop 0
-; GFX9-O0-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:108 ; 4-byte Folded Spill
+; GFX9-O0-NEXT: buffer_store_dword v1, off, s[0:3], s32 offset:116 ; 4-byte Folded Spill
; GFX9-O0-NEXT: s_mov_b64 s[6:7], exec
; GFX9-O0-NEXT: s_and_b64 s[4:5], s[6:7], s[4:5]
; GFX9-O0-NEXT: s_xor_b64 s[6:7], s[4:5], s[6:7]
-; GFX9-O0-NEXT: v_writelane_b32 v30, s6, 6
-; GFX9-O0-NEXT: v_writelane_b32 v30, s7, 7
+; GFX9-O0-NEXT: v_writelane_b32 v29, s6, 6
+; GFX9-O0-NEXT: v_writelane_b32 v29, s7, 7
; GFX9-O0-NEXT: s_or_saveexec_b64 s[18:19], -1
-; GFX9-O0-NEXT: buffer_store_dword v30, off, s[0:3], s32 ; 4-byte Folded Spill
+; GFX9-O0-NEXT: buffer_store_dword v29, off, s[0:3], s32 ; 4-byte Folded Spill
; GFX9-O0-NEXT: s_mov_b64 exec, s[18:19]
; GFX9-O0-NEXT: s_mov_b64 exec, s[4:5]
; GFX9-O0-NEXT: s_cbranch_execz .LBB1_5
@@ -2829,7 +2827,7 @@ define i128 @v_urem_i128_vv(i128 %lhs, i128 %rhs) {
; GFX9-O0-NEXT: v_lshrrev_b64 v[3:4], s4, v[3:4]
; GFX9-O0-NEXT: ; kill: def $vgpr3 killed $vgpr3 killed $vgpr3_vgpr4 killed $exec
; GFX9-O0-NEXT: s_xor_saveexec_b64 s[4:5], -1
-; GFX9-O0-NEXT: buffer_load_dword v30, off, s[0:3], s32 offset:328 ; 4-byte Folded Reload
+; GFX9-O0-NEXT: buffer_load_dword v29, off, s[0:3], s32 offset:328 ; 4-byte Folded Reload
; GFX9-O0-NEXT: s_mov_b64 exec, s[4:5]
; GFX9-O0-NEXT: s_waitcnt vmcnt(0)
; GFX9-O0-NEXT: s_setpc_b64 s[30:31]
diff --git a/llvm/test/CodeGen/AMDGPU/sdiv64.ll b/llvm/test/CodeGen/AMDGPU/sdiv64.ll
index df496258a2509..0981584598abd 100644
--- a/llvm/test/CodeGen/AMDGPU/sdiv64.ll
+++ b/llvm/test/CodeGen/AMDGPU/sdiv64.ll
@@ -369,12 +369,12 @@ define i64 @v_test_sdiv(i64 %x, i64 %y) {
; GCN-IR-NEXT: v_ffbh_u32_e32 v2, v0
; GCN-IR-NEXT: v_add_i32_e64 v2, s[6:7], 32, v2
; GCN-IR-NEXT: v_ffbh_u32_e32 v3, v1
-; GCN-IR-NEXT: v_min_u32_e32 v10, v2, v3
+; GCN-IR-NEXT: v_min_u32_e32 v18, v2, v3
; GCN-IR-NEXT: v_ffbh_u32_e32 v2, v6
; GCN-IR-NEXT: v_add_i32_e64 v2, s[6:7], 32, v2
; GCN-IR-NEXT: v_ffbh_u32_e32 v3, v7
-; GCN-IR-NEXT: v_min_u32_e32 v11, v2, v3
-; GCN-IR-NEXT: v_sub_i32_e64 v2, s[6:7], v10, v11
+; GCN-IR-NEXT: v_min_u32_e32 v19, v2, v3
+; GCN-IR-NEXT: v_sub_i32_e64 v2, s[6:7], v18, v19
; GCN-IR-NEXT: v_cmp_eq_u64_e32 vcc, 0, v[0:1]
; GCN-IR-NEXT: v_cmp_eq_u64_e64 s[4:5], 0, v[6:7]
; GCN-IR-NEXT: v_subb_u32_e64 v3, s[6:7], 0, 0, s[6:7]
@@ -399,47 +399,47 @@ define i64 @v_test_sdiv(i64 %x, i64 %y) {
; GCN-IR-NEXT: v_lshl_b64 v[2:3], v[6:7], v2
; GCN-IR-NEXT: v_mov_b32_e32 v5, 0
; GCN-IR-NEXT: s_and_saveexec_b64 s[4:5], vcc
-; GCN-IR-NEXT: s_xor_b64 s[8:9], exec, s[4:5]
+; GCN-IR-NEXT: s_xor_b64 s[4:5], exec, s[4:5]
; GCN-IR-NEXT: s_cbranch_execz .LBB1_5
; GCN-IR-NEXT: ; %bb.2: ; %udiv-preheader
; GCN-IR-NEXT: v_add_i32_e32 v16, vcc, -1, v0
; GCN-IR-NEXT: v_addc_u32_e32 v17, vcc, -1, v1, vcc
-; GCN-IR-NEXT: v_not_b32_e32 v4, v10
-; GCN-IR-NEXT: v_lshr_b64 v[8:9], v[6:7], v8
-; GCN-IR-NEXT: v_add_i32_e32 v6, vcc, v4, v11
-; GCN-IR-NEXT: v_mov_b32_e32 v10, 0
-; GCN-IR-NEXT: v_addc_u32_e64 v7, s[4:5], -1, 0, vcc
-; GCN-IR-NEXT: s_mov_b64 s[10:11], 0
-; GCN-IR-NEXT: v_mov_b32_e32 v11, 0
+; GCN-IR-NEXT: v_not_b32_e32 v4, v18
+; GCN-IR-NEXT: v_lshr_b64 v[10:11], v[6:7], v8
+; GCN-IR-NEXT: v_add_i32_e32 v6, vcc, v4, v19
+; GCN-IR-NEXT: v_addc_u32_e64 v7, s[8:9], -1, 0, vcc
+; GCN-IR-NEXT: v_mov_b32_e32 v8, 0
+; GCN-IR-NEXT: s_mov_b64 s[8:9], 0
+; GCN-IR-NEXT: v_mov_b32_e32 v9, 0
; GCN-IR-NEXT: v_mov_b32_e32 v5, 0
; GCN-IR-NEXT: .LBB1_3: ; %udiv-do-while
; GCN-IR-NEXT: ; =>This Inner Loop Header: Depth=1
-; GCN-IR-NEXT: v_lshl_b64 v[8:9], v[8:9], 1
+; GCN-IR-NEXT: v_lshl_b64 v[10:11], v[10:11], 1
; GCN-IR-NEXT: v_lshrrev_b32_e32 v4, 31, v3
-; GCN-IR-NEXT: v_or_b32_e32 v8, v8, v4
-; GCN-IR-NEXT: v_lshl_b64 v[2:3], v[2:3], 1
-; GCN-IR-NEXT: v_sub_i32_e32 v4, vcc, v16, v8
-; GCN-IR-NEXT: v_subb_u32_e32 v4, vcc, v17, v9, vcc
-; GCN-IR-NEXT: v_or_b32_e32 v2, v10, v2
-; GCN-IR-NEXT: v_ashrrev_i32_e32 v10, 31, v4
+; GCN-IR-NEXT: v_or_b32_e32 v10, v10, v4
+; GCN-IR-NEXT: v_sub_i32_e32 v4, vcc, v16, v10
+; GCN-IR-NEXT: v_subb_u32_e32 v4, vcc, v17, v11, vcc
; GCN-IR-NEXT: v_add_i32_e32 v6, vcc, 1, v6
-; GCN-IR-NEXT: v_or_b32_e32 v3, v11, v3
-; GCN-IR-NEXT: v_and_b32_e32 v4, 1, v10
-; GCN-IR-NEXT: v_and_b32_e32 v11, v10, v1
-; GCN-IR-NEXT: v_and_b32_e32 v10, v10, v0
; GCN-IR-NEXT: v_addc_u32_e32 v7, vcc, 0, v7, vcc
+; GCN-IR-NEXT: v_lshl_b64 v[2:3], v[2:3], 1
+; GCN-IR-NEXT: v_ashrrev_i32_e32 v18, 31, v4
; GCN-IR-NEXT: v_cmp_eq_u64_e32 vcc, 0, v[6:7]
-; GCN-IR-NEXT: v_sub_i32_e64 v8, s[4:5], v8, v10
-; GCN-IR-NEXT: v_subb_u32_e64 v9, s[4:5], v9, v11, s[4:5]
-; GCN-IR-NEXT: v_mov_b32_e32 v11, v5
-; GCN-IR-NEXT: s_or_b64 s[10:11], vcc, s[10:11]
-; GCN-IR-NEXT: v_mov_b32_e32 v10, v4
-; GCN-IR-NEXT: s_andn2_b64 exec, exec, s[10:11]
+; GCN-IR-NEXT: v_and_b32_e32 v4, 1, v18
+; GCN-IR-NEXT: v_and_b32_e32 v19, v18, v1
+; GCN-IR-NEXT: v_and_b32_e32 v18, v18, v0
+; GCN-IR-NEXT: v_or_b32_e32 v3, v9, v3
+; GCN-IR-NEXT: v_or_b32_e32 v2, v8, v2
+; GCN-IR-NEXT: v_mov_b32_e32 v9, v5
+; GCN-IR-NEXT: s_or_b64 s[8:9], vcc, s[8:9]
+; GCN-IR-NEXT: v_sub_i32_e32 v10, vcc, v10, v18
+; GCN-IR-NEXT: v_mov_b32_e32 v8, v4
+; GCN-IR-NEXT: v_subb_u32_e32 v11, vcc, v11, v19, vcc
+; GCN-IR-NEXT: s_andn2_b64 exec, exec, s[8:9]
; GCN-IR-NEXT: s_cbranch_execnz .LBB1_3
; GCN-IR-NEXT: ; %bb.4: ; %Flow
-; GCN-IR-NEXT: s_or_b64 exec, exec, s[10:11]
-; GCN-IR-NEXT: .LBB1_5: ; %Flow4
; GCN-IR-NEXT: s_or_b64 exec, exec, s[8:9]
+; GCN-IR-NEXT: .LBB1_5: ; %Flow4
+; GCN-IR-NEXT: s_or_b64 exec, exec, s[4:5]
; GCN-IR-NEXT: v_lshl_b64 v[0:1], v[2:3], 1
; GCN-IR-NEXT: v_or_b32_e32 v5, v5, v1
; GCN-IR-NEXT: v_or_b32_e32 v4, v4, v0
@@ -1420,9 +1420,9 @@ define i64 @v_test_sdiv_k_num_i64(i64 %x) {
; GCN-IR-NEXT: v_ffbh_u32_e32 v2, v0
; GCN-IR-NEXT: v_add_i32_e32 v2, vcc, 32, v2
; GCN-IR-NEXT: v_ffbh_u32_e32 v3, v1
-; GCN-IR-NEXT: v_min_u32_e32 v10, v2, v3
+; GCN-IR-NEXT: v_min_u32_e32 v8, v2, v3
; GCN-IR-NEXT: s_movk_i32 s6, 0xffc5
-; GCN-IR-NEXT: v_add_i32_e32 v2, vcc, s6, v10
+; GCN-IR-NEXT: v_add_i32_e32 v2, vcc, s6, v8
; GCN-IR-NEXT: v_addc_u32_e64 v3, s[6:7], 0, -1, vcc
; GCN-IR-NEXT: v_cmp_eq_u64_e64 s[4:5], 0, v[0:1]
; GCN-IR-NEXT: v_cmp_lt_u64_e32 vcc, 63, v[2:3]
@@ -1444,46 +1444,46 @@ define i64 @v_test_sdiv_k_num_i64(i64 %x) {
; GCN-IR-NEXT: v_lshl_b64 v[2:3], 24, v2
; GCN-IR-NEXT: v_mov_b32_e32 v5, 0
; GCN-IR-NEXT: s_and_saveexec_b64 s[4:5], vcc
-; GCN-IR-NEXT: s_xor_b64 s[8:9], exec, s[4:5]
+; GCN-IR-NEXT: s_xor_b64 s[4:5], exec, s[4:5]
; GCN-IR-NEXT: s_cbranch_execz .LBB11_5
; GCN-IR-NEXT: ; %bb.2: ; %udiv-preheader
; GCN-IR-NEXT: v_add_i32_e32 v14, vcc, -1, v0
; GCN-IR-NEXT: v_addc_u32_e32 v15, vcc, -1, v1, vcc
-; GCN-IR-NEXT: v_lshr_b64 v[8:9], 24, v6
-; GCN-IR-NEXT: v_sub_i32_e32 v6, vcc, 58, v10
-; GCN-IR-NEXT: v_mov_b32_e32 v10, 0
-; GCN-IR-NEXT: v_subb_u32_e64 v7, s[4:5], 0, 0, vcc
-; GCN-IR-NEXT: s_mov_b64 s[10:11], 0
-; GCN-IR-NEXT: v_mov_b32_e32 v11, 0
+; GCN-IR-NEXT: v_lshr_b64 v[10:11], 24, v6
+; GCN-IR-NEXT: v_sub_i32_e32 v6, vcc, 58, v8
+; GCN-IR-NEXT: v_subb_u32_e64 v7, s[8:9], 0, 0, vcc
+; GCN-IR-NEXT: v_mov_b32_e32 v8, 0
+; GCN-IR-NEXT: s_mov_b64 s[8:9], 0
+; GCN-IR-NEXT: v_mov_b32_e32 v9, 0
; GCN-IR-NEXT: v_mov_b32_e32 v5, 0
; GCN-IR-NEXT: .LBB11_3: ; %udiv-do-while
; GCN-IR-NEXT: ; =>This Inner Loop Header: Depth=1
-; GCN-IR-NEXT: v_lshl_b64 v[8:9], v[8:9], 1
+; GCN-IR-NEXT: v_lshl_b64 v[10:11], v[10:11], 1
; GCN-IR-NEXT: v_lshrrev_b32_e32 v4, 31, v3
-; GCN-IR-NEXT: v_or_b32_e32 v8, v8, v4
-; GCN-IR-NEXT: v_lshl_b64 v[2:3], v[2:3], 1
-; GCN-IR-NEXT: v_sub_i32_e32 v4, vcc, v14, v8
-; GCN-IR-NEXT: v_subb_u32_e32 v4, vcc, v15, v9, vcc
-; GCN-IR-NEXT: v_or_b32_e32 v2, v10, v2
-; GCN-IR-NEXT: v_ashrrev_i32_e32 v10, 31, v4
+; GCN-IR-NEXT: v_or_b32_e32 v10, v10, v4
+; GCN-IR-NEXT: v_sub_i32_e32 v4, vcc, v14, v10
+; GCN-IR-NEXT: v_subb_u32_e32 v4, vcc, v15, v11, vcc
; GCN-IR-NEXT: v_add_i32_e32 v6, vcc, 1, v6
-; GCN-IR-NEXT: v_or_b32_e32 v3, v11, v3
-; GCN-IR-NEXT: v_and_b32_e32 v4, 1, v10
-; GCN-IR-NEXT: v_and_b32_e32 v11, v10, v1
-; GCN-IR-NEXT: v_and_b32_e32 v10, v10, v0
; GCN-IR-NEXT: v_addc_u32_e32 v7, vcc, 0, v7, vcc
+; GCN-IR-NEXT: v_lshl_b64 v[2:3], v[2:3], 1
+; GCN-IR-NEXT: v_ashrrev_i32_e32 v16, 31, v4
; GCN-IR-NEXT: v_cmp_eq_u64_e32 vcc, 0, v[6:7]
-; GCN-IR-NEXT: v_sub_i32_e64 v8, s[4:5], v8, v10
-; GCN-IR-NEXT: v_subb_u32_e64 v9, s[4:5], v9, v11, s[4:5]
-; GCN-IR-NEXT: v_mov_b32_e32 v11, v5
-; GCN-IR-NEXT: s_or_b64 s[10:11], vcc, s[10:11]
-; GCN-IR-NEXT: v_mov_b32_e32 v10, v4
-; GCN-IR-NEXT: s_andn2_b64 exec, exec, s[10:11]
+; GCN-IR-NEXT: v_and_b32_e32 v4, 1, v16
+; GCN-IR-NEXT: v_and_b32_e32 v17, v16, v1
+; GCN-IR-NEXT: v_and_b32_e32 v16, v16, v0
+; GCN-IR-NEXT: v_or_b32_e32 v3, v9, v3
+; GCN-IR-NEXT: v_or_b32_e32 v2, v8, v2
+; GCN-IR-NEXT: v_mov_b32_e32 v9, v5
+; GCN-IR-NEXT: s_or_b64 s[8:9], vcc, s[8:9]
+; GCN-IR-NEXT: v_sub_i32_e32 v10, vcc, v10, v16
+; GCN-IR-NEXT: v_mov_b32_e32 v8, v4
+; GCN-IR-NEXT: v_subb_u32_e32 v11, vcc, v11, v17, vcc
+; GCN-IR-NEXT: s_andn2_b64 exec, exec, s[8:9]
; GCN-IR-NEXT: s_cbranch_execnz .LBB11_3
; GCN-IR-NEXT: ; %bb.4: ; %Flow
-; GCN-IR-NEXT: s_or_b64 exec, exec, s[10:11]
-; GCN-IR-NEXT: .LBB11_5: ; %Flow4
; GCN-IR-NEXT: s_or_b64 exec, exec, s[8:9]
+; GCN-IR-NEXT: .LBB11_5: ; %Flow4
+; GCN-IR-NEXT: s_or_b64 exec, exec, s[4:5]
; GCN-IR-NEXT: v_lshl_b64 v[0:1], v[2:3], 1
; GCN-IR-NEXT: v_or_b32_e32 v5, v5, v1
; GCN-IR-NEXT: v_or_b32_e32 v4, v4, v0
@@ -1613,9 +1613,9 @@ define i64 @v_test_sdiv_pow2_k_num_i64(i64 %x) {
; GCN-IR-NEXT: v_ffbh_u32_e32 v2, v0
; GCN-IR-NEXT: v_add_i32_e32 v2, vcc, 32, v2
; GCN-IR-NEXT: v_ffbh_u32_e32 v3, v1
-; GCN-IR-NEXT: v_min_u32_e32 v10, v2, v3
+; GCN-IR-NEXT: v_min_u32_e32 v8, v2, v3
; GCN-IR-NEXT: s_movk_i32 s6, 0xffd0
-; GCN-IR-NEXT: v_add_i32_e32 v2, vcc, s6, v10
+; GCN-IR-NEXT: v_add_i32_e32 v2, vcc, s6, v8
; GCN-IR-NEXT: v_addc_u32_e64 v3, s[6:7], 0, -1, vcc
; GCN-IR-NEXT: v_cmp_eq_u64_e64 s[4:5], 0, v[0:1]
; GCN-IR-NEXT: v_cmp_lt_u64_e32 vcc, 63, v[2:3]
@@ -1631,54 +1631,54 @@ define i64 @v_test_sdiv_pow2_k_num_i64(i64 %x) {
; GCN-IR-NEXT: s_cbranch_execz .LBB12_6
; GCN-IR-NEXT: ; %bb.1: ; %udiv-bb1
; GCN-IR-NEXT: v_add_i32_e32 v6, vcc, 1, v2
-; GCN-IR-NEXT: v_sub_i32_e64 v2, s[4:5], 63, v2
; GCN-IR-NEXT: v_addc_u32_e32 v7, vcc, 0, v3, vcc
-; GCN-IR-NEXT: s_mov_b64 s[4:5], 0x8000
+; GCN-IR-NEXT: v_sub_i32_e64 v2, s[4:5], 63, v2
+; GCN-IR-NEXT: s_mov_b64 s[8:9], 0x8000
; GCN-IR-NEXT: v_mov_b32_e32 v4, 0
; GCN-IR-NEXT: v_cmp_ne_u64_e32 vcc, 0, v[6:7]
-; GCN-IR-NEXT: v_lshl_b64 v[2:3], s[4:5], v2
+; GCN-IR-NEXT: v_lshl_b64 v[2:3], s[8:9], v2
; GCN-IR-NEXT: v_mov_b32_e32 v5, 0
-; GCN-IR-NEXT: s_and_saveexec_b64 s[8:9], vcc
-; GCN-IR-NEXT: s_xor_b64 s[8:9], exec, s[8:9]
+; GCN-IR-NEXT: s_and_saveexec_b64 s[4:5], vcc
+; GCN-IR-NEXT: s_xor_b64 s[4:5], exec, s[4:5]
; GCN-IR-NEXT: s_cbranch_execz .LBB12_5
; GCN-IR-NEXT: ; %bb.2: ; %udiv-preheader
; GCN-IR-NEXT: v_add_i32_e32 v14, vcc, -1, v0
; GCN-IR-NEXT: v_addc_u32_e32 v15, vcc, -1, v1, vcc
-; GCN-IR-NEXT: v_lshr_b64 v[8:9], s[4:5], v6
-; GCN-IR-NEXT: v_sub_i32_e32 v6, vcc, 47, v10
-; GCN-IR-NEXT: v_mov_b32_e32 v10, 0
-; GCN-IR-NEXT: v_subb_u32_e64 v7, s[4:5], 0, 0, vcc
-; GCN-IR-NEXT: s_mov_b64 s[10:11], 0
-; GCN-IR-NEXT: v_mov_b32_e32 v11, 0
+; GCN-IR-NEXT: v_lshr_b64 v[10:11], s[8:9], v6
+; GCN-IR-NEXT: v_sub_i32_e32 v6, vcc, 47, v8
+; GCN-IR-NEXT: v_subb_u32_e64 v7, s[8:9], 0, 0, vcc
+; GCN-IR-NEXT: v_mov_b32_e32 v8, 0
+; GCN-IR-NEXT: s_mov_b64 s[8:9], 0
+; GCN-IR-NEXT: v_mov_b32_e32 v9, 0
; GCN-IR-NEXT: v_mov_b32_e32 v5, 0
; GCN-IR-NEXT: .LBB12_3: ; %udiv-do-while
; GCN-IR-NEXT: ; =>This Inner Loop Header: Depth=1
-; GCN-IR-NEXT: v_lshl_b64 v[8:9], v[8:9], 1
+; GCN-IR-NEXT: v_lshl_b64 v[10:11], v[10:11], 1
; GCN-IR-NEXT: v_lshrrev_b32_e32 v4, 31, v3
-; GCN-IR-NEXT: v_or_b32_e32 v8, v8, v4
-; GCN-IR-NEXT: v_lshl_b64 v[2:3], v[2:3], 1
-; GCN-IR-NEXT: v_sub_i32_e32 v4, vcc, v14, v8
-; GCN-IR-NEXT: v_subb_u32_e32 v4, vcc, v15, v9, vcc
-; GCN-IR-NEXT: v_or_b32_e32 v2, v10, v2
-; GCN-IR-NEXT: v_ashrrev_i32_e32 v10, 31, v4
+; GCN-IR-NEXT: v_or_b32_e32 v10, v10, v4
+; GCN-IR-NEXT: v_sub_i32_e32 v4, vcc, v14, v10
+; GCN-IR-NEXT: v_subb_u32_e32 v4, vcc, v15, v11, vcc
; GCN-IR-NEXT: v_add_i32_e32 v6, vcc, 1, v6
-; GCN-IR-NEXT: v_or_b32_e32 v3, v11, v3
-; GCN-IR-NEXT: v_and_b32_e32 v4, 1, v10
-; GCN-IR-NEXT: v_and_b32_e32 v11, v10, v1
-; GCN-IR-NEXT: v_and_b32_e32 v10, v10, v0
; GCN-IR-NEXT: v_addc_u32_e32 v7, vcc, 0, v7, vcc
+; GCN-IR-NEXT: v_lshl_b64 v[2:3], v[2:3], 1
+; GCN-IR-NEXT: v_ashrrev_i32_e32 v16, 31, v4
; GCN-IR-NEXT: v_cmp_eq_u64_e32 vcc, 0, v[6:7]
-; GCN-IR-NEXT: v_sub_i32_e64 v8, s[4:5], v8, v10
-; GCN-IR-NEXT: v_subb_u32_e64 v9, s[4:5], v9, v11, s[4:5]
-; GCN-IR-NEXT: v_mov_b32_e32 v11, v5
-; GCN-IR-NEXT: s_or_b64 s[10:11], vcc, s[10:11]
-; GCN-IR-NEXT: v_mov_b32_e32 v10, v4
-; GCN-IR-NEXT: s_andn2_b64 exec, exec, s[10:11]
+; GCN-IR-NEXT: v_and_b32_e32 v4, 1, v16
+; GCN-IR-NEXT: v_and_b32_e32 v17, v16, v1
+; GCN-IR-NEXT: v_and_b32_e32 v16, v16, v0
+; GCN-IR-NEXT: v_or_b32_e32 v3, v9, v3
+; GCN-IR-NEXT: v_or_b32_e32 v2, v8, v2
+; GCN-IR-NEXT: v_mov_b32_e32 v9, v5
+; GCN-IR-NEXT: s_or_b64 s[8:9], vcc, s[8:9]
+; GCN-IR-NEXT: v_sub_i32_e32 v10, vcc, v10, v16
+; GCN-IR-NEXT: v_mov_b32_e32 v8, v4
+; GCN-IR-NEXT: v_subb_u32_e32 v11, vcc, v11, v17, vcc
+; GCN-IR-NEXT: s_andn2_b64 exec, exec, s[8:9]
; GCN-IR-NEXT: s_cbranch_execnz .LBB12_3
; GCN-IR-NEXT: ; %bb.4: ; %Flow
-; GCN-IR-NEXT: s_or_b64 exec, exec, s[10:11]
-; GCN-IR-NEXT: .LBB12_5: ; %Flow4
; GCN-IR-NEXT: s_or_b64 exec, exec, s[8:9]
+; GCN-IR-NEXT: .LBB12_5: ; %Flow4
+; GCN-IR-NEXT: s_or_b64 exec, exec, s[4:5]
; GCN-IR-NEXT: v_lshl_b64 v[0:1], v[2:3], 1
; GCN-IR-NEXT: v_or_b32_e32 v5, v5, v1
; GCN-IR-NEXT: v_or_b32_e32 v4, v4, v0
@@ -1715,8 +1715,8 @@ define i64 @v_test_sdiv_pow2_k_den_i64(i64 %x) {
; GCN-IR-NEXT: v_ffbh_u32_e32 v0, v4
; GCN-IR-NEXT: v_add_i32_e64 v0, s[4:5], 32, v0
; GCN-IR-NEXT: v_ffbh_u32_e32 v1, v5
-; GCN-IR-NEXT: v_min_u32_e32 v8, v0, v1
-; GCN-IR-NEXT: v_sub_i32_e64 v0, s[4:5], 48, v8
+; GCN-IR-NEXT: v_min_u32_e32 v12, v0, v1
+; GCN-IR-NEXT: v_sub_i32_e64 v0, s[4:5], 48, v12
; GCN-IR-NEXT: v_subb_u32_e64 v1, s[4:5], 0, 0, s[4:5]
; GCN-IR-NEXT: v_cmp_eq_u64_e32 vcc, 0, v[4:5]
; GCN-IR-NEXT: v_cmp_lt_u64_e64 s[4:5], 63, v[0:1]
@@ -1738,44 +1738,44 @@ define i64 @v_test_sdiv_pow2_k_den_i64(i64 %x) {
; GCN-IR-NEXT: v_lshl_b64 v[0:1], v[4:5], v0
; GCN-IR-NEXT: v_mov_b32_e32 v3, 0
; GCN-IR-NEXT: s_and_saveexec_b64 s[4:5], vcc
-; GCN-IR-NEXT: s_xor_b64 s[8:9], exec, s[4:5]
+; GCN-IR-NEXT: s_xor_b64 s[4:5], exec, s[4:5]
; GCN-IR-NEXT: s_cbranch_execz .LBB13_5
; GCN-IR-NEXT: ; %bb.2: ; %udiv-preheader
-; GCN-IR-NEXT: v_lshr_b64 v[6:7], v[4:5], v6
-; GCN-IR-NEXT: v_add_i32_e32 v4, vcc, 0xffffffcf, v8
-; GCN-IR-NEXT: v_mov_b32_e32 v8, 0
-; GCN-IR-NEXT: v_addc_u32_e64 v5, s[4:5], 0, -1, vcc
-; GCN-IR-NEXT: s_mov_b64 s[10:11], 0
-; GCN-IR-NEXT: v_mov_b32_e32 v9, 0
-; GCN-IR-NEXT: s_movk_i32 s12, 0x7fff
+; GCN-IR-NEXT: v_lshr_b64 v[8:9], v[4:5], v6
+; GCN-IR-NEXT: v_add_i32_e32 v4, vcc, 0xffffffcf, v12
+; GCN-IR-NEXT: v_addc_u32_e64 v5, s[8:9], 0, -1, vcc
+; GCN-IR-NEXT: v_mov_b32_e32 v6, 0
+; GCN-IR-NEXT: s_mov_b64 s[8:9], 0
+; GCN-IR-NEXT: v_mov_b32_e32 v7, 0
+; GCN-IR-NEXT: s_movk_i32 s10, 0x7fff
; GCN-IR-NEXT: v_mov_b32_e32 v3, 0
; GCN-IR-NEXT: .LBB13_3: ; %udiv-do-while
; GCN-IR-NEXT: ; =>This Inner Loop Header: Depth=1
-; GCN-IR-NEXT: v_lshl_b64 v[6:7], v[6:7], 1
+; GCN-IR-NEXT: v_lshl_b64 v[8:9], v[8:9], 1
; GCN-IR-NEXT: v_lshrrev_b32_e32 v2, 31, v1
-; GCN-IR-NEXT: v_or_b32_e32 v6, v6, v2
-; GCN-IR-NEXT: v_sub_i32_e32 v2, vcc, s12, v6
-; GCN-IR-NEXT: v_lshl_b64 v[0:1], v[0:1], 1
-; GCN-IR-NEXT: v_subb_u32_e32 v2, vcc, 0, v7, vcc
+; GCN-IR-NEXT: v_or_b32_e32 v8, v8, v2
+; GCN-IR-NEXT: v_sub_i32_e32 v2, vcc, s10, v8
+; GCN-IR-NEXT: v_subb_u32_e32 v2, vcc, 0, v9, vcc
; GCN-IR-NEXT: v_add_i32_e32 v4, vcc, 1, v4
-; GCN-IR-NEXT: v_or_b32_e32 v0, v8, v0
-; GCN-IR-NEXT: v_ashrrev_i32_e32 v8, 31, v2
; GCN-IR-NEXT: v_addc_u32_e32 v5, vcc, 0, v5, vcc
-; GCN-IR-NEXT: v_and_b32_e32 v2, 1, v8
-; GCN-IR-NEXT: v_and_b32_e32 v8, 0x8000, v8
+; GCN-IR-NEXT: v_lshl_b64 v[0:1], v[0:1], 1
+; GCN-IR-NEXT: v_ashrrev_i32_e32 v12, 31, v2
; GCN-IR-NEXT: v_cmp_eq_u64_e32 vcc, 0, v[4:5]
-; GCN-IR-NEXT: v_or_b32_e32 v1, v9, v1
-; GCN-IR-NEXT: v_sub_i32_e64 v6, s[4:5], v6, v8
-; GCN-IR-NEXT: v_mov_b32_e32 v9, v3
-; GCN-IR-NEXT: v_subbrev_u32_e64 v7, s[4:5], 0, v7, s[4:5]
-; GCN-IR-NEXT: s_or_b64 s[10:11], vcc, s[10:11]
-; GCN-IR-NEXT: v_mov_b32_e32 v8, v2
-; GCN-IR-NEXT: s_andn2_b64 exec, exec, s[10:11]
+; GCN-IR-NEXT: v_and_b32_e32 v2, 1, v12
+; GCN-IR-NEXT: v_and_b32_e32 v12, 0x8000, v12
+; GCN-IR-NEXT: v_or_b32_e32 v1, v7, v1
+; GCN-IR-NEXT: v_or_b32_e32 v0, v6, v0
+; GCN-IR-NEXT: v_mov_b32_e32 v7, v3
+; GCN-IR-NEXT: s_or_b64 s[8:9], vcc, s[8:9]
+; GCN-IR-NEXT: v_sub_i32_e32 v8, vcc, v8, v12
+; GCN-IR-NEXT: v_mov_b32_e32 v6, v2
+; GCN-IR-NEXT: v_subbrev_u32_e32 v9, vcc, 0, v9, vcc
+; GCN-IR-NEXT: s_andn2_b64 exec, exec, s[8:9]
; GCN-IR-NEXT: s_cbranch_execnz .LBB13_3
; GCN-IR-NEXT: ; %bb.4: ; %Flow
-; GCN-IR-NEXT: s_or_b64 exec, exec, s[10:11]
-; GCN-IR-NEXT: .LBB13_5: ; %Flow4
; GCN-IR-NEXT: s_or_b64 exec, exec, s[8:9]
+; GCN-IR-NEXT: .LBB13_5: ; %Flow4
+; GCN-IR-NEXT: s_or_b64 exec, exec, s[4:5]
; GCN-IR-NEXT: v_lshl_b64 v[0:1], v[0:1], 1
; GCN-IR-NEXT: v_or_b32_e32 v3, v3, v1
; GCN-IR-NEXT: v_or_b32_e32 v2, v2, v0
diff --git a/llvm/test/CodeGen/AMDGPU/srem64.ll b/llvm/test/CodeGen/AMDGPU/srem64.ll
index 47dfa9f4fc2d3..a5cb9a507bd89 100644
--- a/llvm/test/CodeGen/AMDGPU/srem64.ll
+++ b/llvm/test/CodeGen/AMDGPU/srem64.ll
@@ -346,12 +346,12 @@ define i64 @v_test_srem(i64 %x, i64 %y) {
; GCN-IR-NEXT: v_ffbh_u32_e32 v4, v2
; GCN-IR-NEXT: v_add_i32_e64 v4, s[6:7], 32, v4
; GCN-IR-NEXT: v_ffbh_u32_e32 v5, v3
-; GCN-IR-NEXT: v_min_u32_e32 v12, v4, v5
+; GCN-IR-NEXT: v_min_u32_e32 v10, v4, v5
; GCN-IR-NEXT: v_ffbh_u32_e32 v4, v0
; GCN-IR-NEXT: v_add_i32_e64 v4, s[6:7], 32, v4
; GCN-IR-NEXT: v_ffbh_u32_e32 v5, v1
-; GCN-IR-NEXT: v_min_u32_e32 v13, v4, v5
-; GCN-IR-NEXT: v_sub_i32_e64 v4, s[6:7], v12, v13
+; GCN-IR-NEXT: v_min_u32_e32 v11, v4, v5
+; GCN-IR-NEXT: v_sub_i32_e64 v4, s[6:7], v10, v11
; GCN-IR-NEXT: v_cmp_eq_u64_e32 vcc, 0, v[2:3]
; GCN-IR-NEXT: v_cmp_eq_u64_e64 s[4:5], 0, v[0:1]
; GCN-IR-NEXT: v_subb_u32_e64 v5, s[6:7], 0, 0, s[6:7]
@@ -375,47 +375,47 @@ define i64 @v_test_srem(i64 %x, i64 %y) {
; GCN-IR-NEXT: v_lshl_b64 v[4:5], v[0:1], v4
; GCN-IR-NEXT: v_mov_b32_e32 v7, 0
; GCN-IR-NEXT: s_and_saveexec_b64 s[4:5], vcc
-; GCN-IR-NEXT: s_xor_b64 s[8:9], exec, s[4:5]
+; GCN-IR-NEXT: s_xor_b64 s[4:5], exec, s[4:5]
; GCN-IR-NEXT: s_cbranch_execz .LBB1_5
; GCN-IR-NEXT: ; %bb.2: ; %udiv-preheader
; GCN-IR-NEXT: v_add_i32_e32 v16, vcc, -1, v2
; GCN-IR-NEXT: v_addc_u32_e32 v17, vcc, -1, v3, vcc
-; GCN-IR-NEXT: v_not_b32_e32 v6, v12
-; GCN-IR-NEXT: v_lshr_b64 v[10:11], v[0:1], v8
-; GCN-IR-NEXT: v_add_i32_e32 v8, vcc, v6, v13
-; GCN-IR-NEXT: v_mov_b32_e32 v12, 0
-; GCN-IR-NEXT: v_addc_u32_e64 v9, s[4:5], -1, 0, vcc
-; GCN-IR-NEXT: s_mov_b64 s[10:11], 0
-; GCN-IR-NEXT: v_mov_b32_e32 v13, 0
+; GCN-IR-NEXT: v_not_b32_e32 v6, v10
+; GCN-IR-NEXT: v_lshr_b64 v[12:13], v[0:1], v8
+; GCN-IR-NEXT: v_add_i32_e32 v8, vcc, v6, v11
+; GCN-IR-NEXT: v_addc_u32_e64 v9, s[8:9], -1, 0, vcc
+; GCN-IR-NEXT: v_mov_b32_e32 v10, 0
+; GCN-IR-NEXT: s_mov_b64 s[8:9], 0
+; GCN-IR-NEXT: v_mov_b32_e32 v11, 0
; GCN-IR-NEXT: v_mov_b32_e32 v7, 0
; GCN-IR-NEXT: .LBB1_3: ; %udiv-do-while
; GCN-IR-NEXT: ; =>This Inner Loop Header: Depth=1
-; GCN-IR-NEXT: v_lshl_b64 v[10:11], v[10:11], 1
+; GCN-IR-NEXT: v_lshl_b64 v[12:13], v[12:13], 1
; GCN-IR-NEXT: v_lshrrev_b32_e32 v6, 31, v5
-; GCN-IR-NEXT: v_or_b32_e32 v10, v10, v6
-; GCN-IR-NEXT: v_lshl_b64 v[4:5], v[4:5], 1
-; GCN-IR-NEXT: v_sub_i32_e32 v6, vcc, v16, v10
-; GCN-IR-NEXT: v_subb_u32_e32 v6, vcc, v17, v11, vcc
-; GCN-IR-NEXT: v_or_b32_e32 v4, v12, v4
-; GCN-IR-NEXT: v_ashrrev_i32_e32 v12, 31, v6
+; GCN-IR-NEXT: v_or_b32_e32 v12, v12, v6
+; GCN-IR-NEXT: v_sub_i32_e32 v6, vcc, v16, v12
+; GCN-IR-NEXT: v_subb_u32_e32 v6, vcc, v17, v13, vcc
; GCN-IR-NEXT: v_add_i32_e32 v8, vcc, 1, v8
-; GCN-IR-NEXT: v_or_b32_e32 v5, v13, v5
-; GCN-IR-NEXT: v_and_b32_e32 v6, 1, v12
-; GCN-IR-NEXT: v_and_b32_e32 v13, v12, v3
-; GCN-IR-NEXT: v_and_b32_e32 v12, v12, v2
; GCN-IR-NEXT: v_addc_u32_e32 v9, vcc, 0, v9, vcc
+; GCN-IR-NEXT: v_lshl_b64 v[4:5], v[4:5], 1
+; GCN-IR-NEXT: v_ashrrev_i32_e32 v18, 31, v6
; GCN-IR-NEXT: v_cmp_eq_u64_e32 vcc, 0, v[8:9]
-; GCN-IR-NEXT: v_sub_i32_e64 v10, s[4:5], v10, v12
-; GCN-IR-NEXT: v_subb_u32_e64 v11, s[4:5], v11, v13, s[4:5]
-; GCN-IR-NEXT: v_mov_b32_e32 v13, v7
-; GCN-IR-NEXT: s_or_b64 s[10:11], vcc, s[10:11]
-; GCN-IR-NEXT: v_mov_b32_e32 v12, v6
-; GCN-IR-NEXT: s_andn2_b64 exec, exec, s[10:11]
+; GCN-IR-NEXT: v_and_b32_e32 v6, 1, v18
+; GCN-IR-NEXT: v_and_b32_e32 v19, v18, v3
+; GCN-IR-NEXT: v_and_b32_e32 v18, v18, v2
+; GCN-IR-NEXT: v_or_b32_e32 v5, v11, v5
+; GCN-IR-NEXT: v_or_b32_e32 v4, v10, v4
+; GCN-IR-NEXT: v_mov_b32_e32 v11, v7
+; GCN-IR-NEXT: s_or_b64 s[8:9], vcc, s[8:9]
+; GCN-IR-NEXT: v_sub_i32_e32 v12, vcc, v12, v18
+; GCN-IR-NEXT: v_mov_b32_e32 v10, v6
+; GCN-IR-NEXT: v_subb_u32_e32 v13, vcc, v13, v19, vcc
+; GCN-IR-NEXT: s_andn2_b64 exec, exec, s[8:9]
; GCN-IR-NEXT: s_cbranch_execnz .LBB1_3
; GCN-IR-NEXT: ; %bb.4: ; %Flow
-; GCN-IR-NEXT: s_or_b64 exec, exec, s[10:11]
-; GCN-IR-NEXT: .LBB1_5: ; %Flow4
; GCN-IR-NEXT: s_or_b64 exec, exec, s[8:9]
+; GCN-IR-NEXT: .LBB1_5: ; %Flow4
+; GCN-IR-NEXT: s_or_b64 exec, exec, s[4:5]
; GCN-IR-NEXT: v_lshl_b64 v[4:5], v[4:5], 1
; GCN-IR-NEXT: v_or_b32_e32 v7, v7, v5
; GCN-IR-NEXT: v_or_b32_e32 v6, v6, v4
@@ -1538,9 +1538,9 @@ define i64 @v_test_srem_k_num_i64(i64 %x) {
; GCN-IR-NEXT: v_ffbh_u32_e32 v2, v0
; GCN-IR-NEXT: v_add_i32_e32 v2, vcc, 32, v2
; GCN-IR-NEXT: v_ffbh_u32_e32 v3, v1
-; GCN-IR-NEXT: v_min_u32_e32 v10, v2, v3
+; GCN-IR-NEXT: v_min_u32_e32 v8, v2, v3
; GCN-IR-NEXT: s_movk_i32 s6, 0xffc5
-; GCN-IR-NEXT: v_add_i32_e32 v2, vcc, s6, v10
+; GCN-IR-NEXT: v_add_i32_e32 v2, vcc, s6, v8
; GCN-IR-NEXT: v_addc_u32_e64 v3, s[6:7], 0, -1, vcc
; GCN-IR-NEXT: v_cmp_eq_u64_e64 s[4:5], 0, v[0:1]
; GCN-IR-NEXT: v_cmp_lt_u64_e32 vcc, 63, v[2:3]
@@ -1561,46 +1561,46 @@ define i64 @v_test_srem_k_num_i64(i64 %x) {
; GCN-IR-NEXT: v_lshl_b64 v[2:3], 24, v2
; GCN-IR-NEXT: v_mov_b32_e32 v5, 0
; GCN-IR-NEXT: s_and_saveexec_b64 s[4:5], vcc
-; GCN-IR-NEXT: s_xor_b64 s[8:9], exec, s[4:5]
+; GCN-IR-NEXT: s_xor_b64 s[4:5], exec, s[4:5]
; GCN-IR-NEXT: s_cbranch_execz .LBB11_5
; GCN-IR-NEXT: ; %bb.2: ; %udiv-preheader
; GCN-IR-NEXT: v_add_i32_e32 v12, vcc, -1, v0
; GCN-IR-NEXT: v_addc_u32_e32 v13, vcc, -1, v1, vcc
-; GCN-IR-NEXT: v_lshr_b64 v[8:9], 24, v6
-; GCN-IR-NEXT: v_sub_i32_e32 v6, vcc, 58, v10
-; GCN-IR-NEXT: v_mov_b32_e32 v10, 0
-; GCN-IR-NEXT: v_subb_u32_e64 v7, s[4:5], 0, 0, vcc
-; GCN-IR-NEXT: s_mov_b64 s[10:11], 0
-; GCN-IR-NEXT: v_mov_b32_e32 v11, 0
+; GCN-IR-NEXT: v_lshr_b64 v[10:11], 24, v6
+; GCN-IR-NEXT: v_sub_i32_e32 v6, vcc, 58, v8
+; GCN-IR-NEXT: v_subb_u32_e64 v7, s[8:9], 0, 0, vcc
+; GCN-IR-NEXT: v_mov_b32_e32 v8, 0
+; GCN-IR-NEXT: s_mov_b64 s[8:9], 0
+; GCN-IR-NEXT: v_mov_b32_e32 v9, 0
; GCN-IR-NEXT: v_mov_b32_e32 v5, 0
; GCN-IR-NEXT: .LBB11_3: ; %udiv-do-while
; GCN-IR-NEXT: ; =>This Inner Loop Header: Depth=1
-; GCN-IR-NEXT: v_lshl_b64 v[8:9], v[8:9], 1
+; GCN-IR-NEXT: v_lshl_b64 v[10:11], v[10:11], 1
; GCN-IR-NEXT: v_lshrrev_b32_e32 v4, 31, v3
-; GCN-IR-NEXT: v_or_b32_e32 v8, v8, v4
-; GCN-IR-NEXT: v_lshl_b64 v[2:3], v[2:3], 1
-; GCN-IR-NEXT: v_sub_i32_e32 v4, vcc, v12, v8
-; GCN-IR-NEXT: v_subb_u32_e32 v4, vcc, v13, v9, vcc
-; GCN-IR-NEXT: v_or_b32_e32 v2, v10, v2
-; GCN-IR-NEXT: v_ashrrev_i32_e32 v10, 31, v4
+; GCN-IR-NEXT: v_or_b32_e32 v10, v10, v4
+; GCN-IR-NEXT: v_sub_i32_e32 v4, vcc, v12, v10
+; GCN-IR-NEXT: v_subb_u32_e32 v4, vcc, v13, v11, vcc
; GCN-IR-NEXT: v_add_i32_e32 v6, vcc, 1, v6
-; GCN-IR-NEXT: v_or_b32_e32 v3, v11, v3
-; GCN-IR-NEXT: v_and_b32_e32 v4, 1, v10
-; GCN-IR-NEXT: v_and_b32_e32 v11, v10, v1
-; GCN-IR-NEXT: v_and_b32_e32 v10, v10, v0
; GCN-IR-NEXT: v_addc_u32_e32 v7, vcc, 0, v7, vcc
+; GCN-IR-NEXT: v_lshl_b64 v[2:3], v[2:3], 1
+; GCN-IR-NEXT: v_ashrrev_i32_e32 v14, 31, v4
; GCN-IR-NEXT: v_cmp_eq_u64_e32 vcc, 0, v[6:7]
-; GCN-IR-NEXT: v_sub_i32_e64 v8, s[4:5], v8, v10
-; GCN-IR-NEXT: v_subb_u32_e64 v9, s[4:5], v9, v11, s[4:5]
-; GCN-IR-NEXT: v_mov_b32_e32 v11, v5
-; GCN-IR-NEXT: s_or_b64 s[10:11], vcc, s[10:11]
-; GCN-IR-NEXT: v_mov_b32_e32 v10, v4
-; GCN-IR-NEXT: s_andn2_b64 exec, exec, s[10:11]
+; GCN-IR-NEXT: v_and_b32_e32 v4, 1, v14
+; GCN-IR-NEXT: v_and_b32_e32 v15, v14, v1
+; GCN-IR-NEXT: v_and_b32_e32 v14, v14, v0
+; GCN-IR-NEXT: v_or_b32_e32 v3, v9, v3
+; GCN-IR-NEXT: v_or_b32_e32 v2, v8, v2
+; GCN-IR-NEXT: v_mov_b32_e32 v9, v5
+; GCN-IR-NEXT: s_or_b64 s[8:9], vcc, s[8:9]
+; GCN-IR-NEXT: v_sub_i32_e32 v10, vcc, v10, v14
+; GCN-IR-NEXT: v_mov_b32_e32 v8, v4
+; GCN-IR-NEXT: v_subb_u32_e32 v11, vcc, v11, v15, vcc
+; GCN-IR-NEXT: s_andn2_b64 exec, exec, s[8:9]
; GCN-IR-NEXT: s_cbranch_execnz .LBB11_3
; GCN-IR-NEXT: ; %bb.4: ; %Flow
-; GCN-IR-NEXT: s_or_b64 exec, exec, s[10:11]
-; GCN-IR-NEXT: .LBB11_5: ; %Flow4
; GCN-IR-NEXT: s_or_b64 exec, exec, s[8:9]
+; GCN-IR-NEXT: .LBB11_5: ; %Flow4
+; GCN-IR-NEXT: s_or_b64 exec, exec, s[4:5]
; GCN-IR-NEXT: v_lshl_b64 v[2:3], v[2:3], 1
; GCN-IR-NEXT: v_or_b32_e32 v5, v5, v3
; GCN-IR-NEXT: v_or_b32_e32 v4, v4, v2
@@ -1729,9 +1729,9 @@ define i64 @v_test_srem_pow2_k_num_i64(i64 %x) {
; GCN-IR-NEXT: v_ffbh_u32_e32 v2, v0
; GCN-IR-NEXT: v_add_i32_e32 v2, vcc, 32, v2
; GCN-IR-NEXT: v_ffbh_u32_e32 v3, v1
-; GCN-IR-NEXT: v_min_u32_e32 v10, v2, v3
+; GCN-IR-NEXT: v_min_u32_e32 v8, v2, v3
; GCN-IR-NEXT: s_movk_i32 s6, 0xffd0
-; GCN-IR-NEXT: v_add_i32_e32 v2, vcc, s6, v10
+; GCN-IR-NEXT: v_add_i32_e32 v2, vcc, s6, v8
; GCN-IR-NEXT: v_addc_u32_e64 v3, s[6:7], 0, -1, vcc
; GCN-IR-NEXT: v_cmp_eq_u64_e64 s[4:5], 0, v[0:1]
; GCN-IR-NEXT: v_cmp_lt_u64_e32 vcc, 63, v[2:3]
@@ -1746,54 +1746,54 @@ define i64 @v_test_srem_pow2_k_num_i64(i64 %x) {
; GCN-IR-NEXT: s_cbranch_execz .LBB12_6
; GCN-IR-NEXT: ; %bb.1: ; %udiv-bb1
; GCN-IR-NEXT: v_add_i32_e32 v6, vcc, 1, v2
-; GCN-IR-NEXT: v_sub_i32_e64 v2, s[4:5], 63, v2
; GCN-IR-NEXT: v_addc_u32_e32 v7, vcc, 0, v3, vcc
-; GCN-IR-NEXT: s_mov_b64 s[4:5], 0x8000
+; GCN-IR-NEXT: v_sub_i32_e64 v2, s[4:5], 63, v2
+; GCN-IR-NEXT: s_mov_b64 s[8:9], 0x8000
; GCN-IR-NEXT: v_mov_b32_e32 v4, 0
; GCN-IR-NEXT: v_cmp_ne_u64_e32 vcc, 0, v[6:7]
-; GCN-IR-NEXT: v_lshl_b64 v[2:3], s[4:5], v2
+; GCN-IR-NEXT: v_lshl_b64 v[2:3], s[8:9], v2
; GCN-IR-NEXT: v_mov_b32_e32 v5, 0
-; GCN-IR-NEXT: s_and_saveexec_b64 s[8:9], vcc
-; GCN-IR-NEXT: s_xor_b64 s[8:9], exec, s[8:9]
+; GCN-IR-NEXT: s_and_saveexec_b64 s[4:5], vcc
+; GCN-IR-NEXT: s_xor_b64 s[4:5], exec, s[4:5]
; GCN-IR-NEXT: s_cbranch_execz .LBB12_5
; GCN-IR-NEXT: ; %bb.2: ; %udiv-preheader
; GCN-IR-NEXT: v_add_i32_e32 v12, vcc, -1, v0
; GCN-IR-NEXT: v_addc_u32_e32 v13, vcc, -1, v1, vcc
-; GCN-IR-NEXT: v_lshr_b64 v[8:9], s[4:5], v6
-; GCN-IR-NEXT: v_sub_i32_e32 v6, vcc, 47, v10
-; GCN-IR-NEXT: v_mov_b32_e32 v10, 0
-; GCN-IR-NEXT: v_subb_u32_e64 v7, s[4:5], 0, 0, vcc
-; GCN-IR-NEXT: s_mov_b64 s[10:11], 0
-; GCN-IR-NEXT: v_mov_b32_e32 v11, 0
+; GCN-IR-NEXT: v_lshr_b64 v[10:11], s[8:9], v6
+; GCN-IR-NEXT: v_sub_i32_e32 v6, vcc, 47, v8
+; GCN-IR-NEXT: v_subb_u32_e64 v7, s[8:9], 0, 0, vcc
+; GCN-IR-NEXT: v_mov_b32_e32 v8, 0
+; GCN-IR-NEXT: s_mov_b64 s[8:9], 0
+; GCN-IR-NEXT: v_mov_b32_e32 v9, 0
; GCN-IR-NEXT: v_mov_b32_e32 v5, 0
; GCN-IR-NEXT: .LBB12_3: ; %udiv-do-while
; GCN-IR-NEXT: ; =>This Inner Loop Header: Depth=1
-; GCN-IR-NEXT: v_lshl_b64 v[8:9], v[8:9], 1
+; GCN-IR-NEXT: v_lshl_b64 v[10:11], v[10:11], 1
; GCN-IR-NEXT: v_lshrrev_b32_e32 v4, 31, v3
-; GCN-IR-NEXT: v_or_b32_e32 v8, v8, v4
-; GCN-IR-NEXT: v_lshl_b64 v[2:3], v[2:3], 1
-; GCN-IR-NEXT: v_sub_i32_e32 v4, vcc, v12, v8
-; GCN-IR-NEXT: v_subb_u32_e32 v4, vcc, v13, v9, vcc
-; GCN-IR-NEXT: v_or_b32_e32 v2, v10, v2
-; GCN-IR-NEXT: v_ashrrev_i32_e32 v10, 31, v4
+; GCN-IR-NEXT: v_or_b32_e32 v10, v10, v4
+; GCN-IR-NEXT: v_sub_i32_e32 v4, vcc, v12, v10
+; GCN-IR-NEXT: v_subb_u32_e32 v4, vcc, v13, v11, vcc
; GCN-IR-NEXT: v_add_i32_e32 v6, vcc, 1, v6
-; GCN-IR-NEXT: v_or_b32_e32 v3, v11, v3
-; GCN-IR-NEXT: v_and_b32_e32 v4, 1, v10
-; GCN-IR-NEXT: v_and_b32_e32 v11, v10, v1
-; GCN-IR-NEXT: v_and_b32_e32 v10, v10, v0
; GCN-IR-NEXT: v_addc_u32_e32 v7, vcc, 0, v7, vcc
+; GCN-IR-NEXT: v_lshl_b64 v[2:3], v[2:3], 1
+; GCN-IR-NEXT: v_ashrrev_i32_e32 v14, 31, v4
; GCN-IR-NEXT: v_cmp_eq_u64_e32 vcc, 0, v[6:7]
-; GCN-IR-NEXT: v_sub_i32_e64 v8, s[4:5], v8, v10
-; GCN-IR-NEXT: v_subb_u32_e64 v9, s[4:5], v9, v11, s[4:5]
-; GCN-IR-NEXT: v_mov_b32_e32 v11, v5
-; GCN-IR-NEXT: s_or_b64 s[10:11], vcc, s[10:11]
-; GCN-IR-NEXT: v_mov_b32_e32 v10, v4
-; GCN-IR-NEXT: s_andn2_b64 exec, exec, s[10:11]
+; GCN-IR-NEXT: v_and_b32_e32 v4, 1, v14
+; GCN-IR-NEXT: v_and_b32_e32 v15, v14, v1
+; GCN-IR-NEXT: v_and_b32_e32 v14, v14, v0
+; GCN-IR-NEXT: v_or_b32_e32 v3, v9, v3
+; GCN-IR-NEXT: v_or_b32_e32 v2, v8, v2
+; GCN-IR-NEXT: v_mov_b32_e32 v9, v5
+; GCN-IR-NEXT: s_or_b64 s[8:9], vcc, s[8:9]
+; GCN-IR-NEXT: v_sub_i32_e32 v10, vcc, v10, v14
+; GCN-IR-NEXT: v_mov_b32_e32 v8, v4
+; GCN-IR-NEXT: v_subb_u32_e32 v11, vcc, v11, v15, vcc
+; GCN-IR-NEXT: s_andn2_b64 exec, exec, s[8:9]
; GCN-IR-NEXT: s_cbranch_execnz .LBB12_3
; GCN-IR-NEXT: ; %bb.4: ; %Flow
-; GCN-IR-NEXT: s_or_b64 exec, exec, s[10:11]
-; GCN-IR-NEXT: .LBB12_5: ; %Flow4
; GCN-IR-NEXT: s_or_b64 exec, exec, s[8:9]
+; GCN-IR-NEXT: .LBB12_5: ; %Flow4
+; GCN-IR-NEXT: s_or_b64 exec, exec, s[4:5]
; GCN-IR-NEXT: v_lshl_b64 v[2:3], v[2:3], 1
; GCN-IR-NEXT: v_or_b32_e32 v5, v5, v3
; GCN-IR-NEXT: v_or_b32_e32 v4, v4, v2
@@ -1836,8 +1836,8 @@ define i64 @v_test_srem_pow2_k_den_i64(i64 %x) {
; GCN-IR-NEXT: v_ffbh_u32_e32 v2, v0
; GCN-IR-NEXT: v_add_i32_e64 v2, s[4:5], 32, v2
; GCN-IR-NEXT: v_ffbh_u32_e32 v3, v1
-; GCN-IR-NEXT: v_min_u32_e32 v10, v2, v3
-; GCN-IR-NEXT: v_sub_i32_e64 v2, s[4:5], 48, v10
+; GCN-IR-NEXT: v_min_u32_e32 v8, v2, v3
+; GCN-IR-NEXT: v_sub_i32_e64 v2, s[4:5], 48, v8
; GCN-IR-NEXT: v_subb_u32_e64 v3, s[4:5], 0, 0, s[4:5]
; GCN-IR-NEXT: v_cmp_eq_u64_e32 vcc, 0, v[0:1]
; GCN-IR-NEXT: v_cmp_lt_u64_e64 s[4:5], 63, v[2:3]
@@ -1859,44 +1859,44 @@ define i64 @v_test_srem_pow2_k_den_i64(i64 %x) {
; GCN-IR-NEXT: v_lshl_b64 v[2:3], v[0:1], v2
; GCN-IR-NEXT: v_mov_b32_e32 v5, 0
; GCN-IR-NEXT: s_and_saveexec_b64 s[4:5], vcc
-; GCN-IR-NEXT: s_xor_b64 s[8:9], exec, s[4:5]
+; GCN-IR-NEXT: s_xor_b64 s[4:5], exec, s[4:5]
; GCN-IR-NEXT: s_cbranch_execz .LBB13_5
; GCN-IR-NEXT: ; %bb.2: ; %udiv-preheader
-; GCN-IR-NEXT: v_lshr_b64 v[8:9], v[0:1], v6
-; GCN-IR-NEXT: v_add_i32_e32 v6, vcc, 0xffffffcf, v10
-; GCN-IR-NEXT: v_mov_b32_e32 v10, 0
-; GCN-IR-NEXT: v_addc_u32_e64 v7, s[4:5], 0, -1, vcc
-; GCN-IR-NEXT: s_mov_b64 s[10:11], 0
-; GCN-IR-NEXT: v_mov_b32_e32 v11, 0
-; GCN-IR-NEXT: s_movk_i32 s12, 0x7fff
+; GCN-IR-NEXT: v_lshr_b64 v[10:11], v[0:1], v6
+; GCN-IR-NEXT: v_add_i32_e32 v6, vcc, 0xffffffcf, v8
+; GCN-IR-NEXT: v_addc_u32_e64 v7, s[8:9], 0, -1, vcc
+; GCN-IR-NEXT: v_mov_b32_e32 v8, 0
+; GCN-IR-NEXT: s_mov_b64 s[8:9], 0
+; GCN-IR-NEXT: v_mov_b32_e32 v9, 0
+; GCN-IR-NEXT: s_movk_i32 s10, 0x7fff
; GCN-IR-NEXT: v_mov_b32_e32 v5, 0
; GCN-IR-NEXT: .LBB13_3: ; %udiv-do-while
; GCN-IR-NEXT: ; =>This Inner Loop Header: Depth=1
-; GCN-IR-NEXT: v_lshl_b64 v[8:9], v[8:9], 1
+; GCN-IR-NEXT: v_lshl_b64 v[10:11], v[10:11], 1
; GCN-IR-NEXT: v_lshrrev_b32_e32 v4, 31, v3
-; GCN-IR-NEXT: v_or_b32_e32 v8, v8, v4
-; GCN-IR-NEXT: v_sub_i32_e32 v4, vcc, s12, v8
-; GCN-IR-NEXT: v_lshl_b64 v[2:3], v[2:3], 1
-; GCN-IR-NEXT: v_subb_u32_e32 v4, vcc, 0, v9, vcc
+; GCN-IR-NEXT: v_or_b32_e32 v10, v10, v4
+; GCN-IR-NEXT: v_sub_i32_e32 v4, vcc, s10, v10
+; GCN-IR-NEXT: v_subb_u32_e32 v4, vcc, 0, v11, vcc
; GCN-IR-NEXT: v_add_i32_e32 v6, vcc, 1, v6
-; GCN-IR-NEXT: v_or_b32_e32 v2, v10, v2
-; GCN-IR-NEXT: v_ashrrev_i32_e32 v10, 31, v4
; GCN-IR-NEXT: v_addc_u32_e32 v7, vcc, 0, v7, vcc
-; GCN-IR-NEXT: v_and_b32_e32 v4, 1, v10
-; GCN-IR-NEXT: v_and_b32_e32 v10, 0x8000, v10
+; GCN-IR-NEXT: v_lshl_b64 v[2:3], v[2:3], 1
+; GCN-IR-NEXT: v_ashrrev_i32_e32 v14, 31, v4
; GCN-IR-NEXT: v_cmp_eq_u64_e32 vcc, 0, v[6:7]
-; GCN-IR-NEXT: v_or_b32_e32 v3, v11, v3
-; GCN-IR-NEXT: v_sub_i32_e64 v8, s[4:5], v8, v10
-; GCN-IR-NEXT: v_mov_b32_e32 v11, v5
-; GCN-IR-NEXT: v_subbrev_u32_e64 v9, s[4:5], 0, v9, s[4:5]
-; GCN-IR-NEXT: s_or_b64 s[10:11], vcc, s[10:11]
-; GCN-IR-NEXT: v_mov_b32_e32 v10, v4
-; GCN-IR-NEXT: s_andn2_b64 exec, exec, s[10:11]
+; GCN-IR-NEXT: v_and_b32_e32 v4, 1, v14
+; GCN-IR-NEXT: v_and_b32_e32 v14, 0x8000, v14
+; GCN-IR-NEXT: v_or_b32_e32 v3, v9, v3
+; GCN-IR-NEXT: v_or_b32_e32 v2, v8, v2
+; GCN-IR-NEXT: v_mov_b32_e32 v9, v5
+; GCN-IR-NEXT: s_or_b64 s[8:9], vcc, s[8:9]
+; GCN-IR-NEXT: v_sub_i32_e32 v10, vcc, v10, v14
+; GCN-IR-NEXT: v_mov_b32_e32 v8, v4
+; GCN-IR-NEXT: v_subbrev_u32_e32 v11, vcc, 0, v11, vcc
+; GCN-IR-NEXT: s_andn2_b64 exec, exec, s[8:9]
; GCN-IR-NEXT: s_cbranch_execnz .LBB13_3
; GCN-IR-NEXT: ; %bb.4: ; %Flow
-; GCN-IR-NEXT: s_or_b64 exec, exec, s[10:11]
-; GCN-IR-NEXT: .LBB13_5: ; %Flow4
; GCN-IR-NEXT: s_or_b64 exec, exec, s[8:9]
+; GCN-IR-NEXT: .LBB13_5: ; %Flow4
+; GCN-IR-NEXT: s_or_b64 exec, exec, s[4:5]
; GCN-IR-NEXT: v_lshl_b64 v[2:3], v[2:3], 1
; GCN-IR-NEXT: v_or_b32_e32 v5, v5, v3
; GCN-IR-NEXT: v_or_b32_e32 v4, v4, v2
diff --git a/llvm/test/CodeGen/AMDGPU/udiv64.ll b/llvm/test/CodeGen/AMDGPU/udiv64.ll
index e9017939f8a4a..f1d7c84836ca7 100644
--- a/llvm/test/CodeGen/AMDGPU/udiv64.ll
+++ b/llvm/test/CodeGen/AMDGPU/udiv64.ll
@@ -341,47 +341,47 @@ define i64 @v_test_udiv_i64(i64 %x, i64 %y) {
; GCN-IR-NEXT: v_lshl_b64 v[4:5], v[0:1], v4
; GCN-IR-NEXT: v_mov_b32_e32 v7, 0
; GCN-IR-NEXT: s_and_saveexec_b64 s[4:5], vcc
-; GCN-IR-NEXT: s_xor_b64 s[8:9], exec, s[4:5]
+; GCN-IR-NEXT: s_xor_b64 s[4:5], exec, s[4:5]
; GCN-IR-NEXT: s_cbranch_execz .LBB1_5
; GCN-IR-NEXT: ; %bb.2: ; %udiv-preheader
; GCN-IR-NEXT: v_add_i32_e32 v12, vcc, -1, v2
-; GCN-IR-NEXT: v_lshr_b64 v[8:9], v[0:1], v10
+; GCN-IR-NEXT: v_lshr_b64 v[10:11], v[0:1], v10
; GCN-IR-NEXT: v_addc_u32_e32 v13, vcc, -1, v3, vcc
; GCN-IR-NEXT: v_not_b32_e32 v0, v14
; GCN-IR-NEXT: v_add_i32_e32 v0, vcc, v0, v15
-; GCN-IR-NEXT: v_mov_b32_e32 v10, 0
-; GCN-IR-NEXT: v_addc_u32_e64 v1, s[4:5], -1, 0, vcc
-; GCN-IR-NEXT: s_mov_b64 s[10:11], 0
-; GCN-IR-NEXT: v_mov_b32_e32 v11, 0
+; GCN-IR-NEXT: v_addc_u32_e64 v1, s[8:9], -1, 0, vcc
+; GCN-IR-NEXT: v_mov_b32_e32 v8, 0
+; GCN-IR-NEXT: s_mov_b64 s[8:9], 0
+; GCN-IR-NEXT: v_mov_b32_e32 v9, 0
; GCN-IR-NEXT: v_mov_b32_e32 v7, 0
; GCN-IR-NEXT: .LBB1_3: ; %udiv-do-while
; GCN-IR-NEXT: ; =>This Inner Loop Header: Depth=1
-; GCN-IR-NEXT: v_lshl_b64 v[8:9], v[8:9], 1
+; GCN-IR-NEXT: v_lshl_b64 v[10:11], v[10:11], 1
; GCN-IR-NEXT: v_lshrrev_b32_e32 v6, 31, v5
-; GCN-IR-NEXT: v_or_b32_e32 v8, v8, v6
-; GCN-IR-NEXT: v_lshl_b64 v[4:5], v[4:5], 1
-; GCN-IR-NEXT: v_sub_i32_e32 v6, vcc, v12, v8
-; GCN-IR-NEXT: v_subb_u32_e32 v6, vcc, v13, v9, vcc
-; GCN-IR-NEXT: v_or_b32_e32 v4, v10, v4
-; GCN-IR-NEXT: v_ashrrev_i32_e32 v10, 31, v6
+; GCN-IR-NEXT: v_or_b32_e32 v10, v10, v6
+; GCN-IR-NEXT: v_sub_i32_e32 v6, vcc, v12, v10
+; GCN-IR-NEXT: v_subb_u32_e32 v6, vcc, v13, v11, vcc
; GCN-IR-NEXT: v_add_i32_e32 v0, vcc, 1, v0
-; GCN-IR-NEXT: v_or_b32_e32 v5, v11, v5
-; GCN-IR-NEXT: v_and_b32_e32 v6, 1, v10
-; GCN-IR-NEXT: v_and_b32_e32 v11, v10, v3
-; GCN-IR-NEXT: v_and_b32_e32 v10, v10, v2
; GCN-IR-NEXT: v_addc_u32_e32 v1, vcc, 0, v1, vcc
+; GCN-IR-NEXT: v_lshl_b64 v[4:5], v[4:5], 1
+; GCN-IR-NEXT: v_ashrrev_i32_e32 v14, 31, v6
; GCN-IR-NEXT: v_cmp_eq_u64_e32 vcc, 0, v[0:1]
-; GCN-IR-NEXT: v_sub_i32_e64 v8, s[4:5], v8, v10
-; GCN-IR-NEXT: v_subb_u32_e64 v9, s[4:5], v9, v11, s[4:5]
-; GCN-IR-NEXT: v_mov_b32_e32 v11, v7
-; GCN-IR-NEXT: s_or_b64 s[10:11], vcc, s[10:11]
-; GCN-IR-NEXT: v_mov_b32_e32 v10, v6
-; GCN-IR-NEXT: s_andn2_b64 exec, exec, s[10:11]
+; GCN-IR-NEXT: v_and_b32_e32 v6, 1, v14
+; GCN-IR-NEXT: v_and_b32_e32 v15, v14, v3
+; GCN-IR-NEXT: v_and_b32_e32 v14, v14, v2
+; GCN-IR-NEXT: v_or_b32_e32 v5, v9, v5
+; GCN-IR-NEXT: v_or_b32_e32 v4, v8, v4
+; GCN-IR-NEXT: v_mov_b32_e32 v9, v7
+; GCN-IR-NEXT: s_or_b64 s[8:9], vcc, s[8:9]
+; GCN-IR-NEXT: v_sub_i32_e32 v10, vcc, v10, v14
+; GCN-IR-NEXT: v_mov_b32_e32 v8, v6
+; GCN-IR-NEXT: v_subb_u32_e32 v11, vcc, v11, v15, vcc
+; GCN-IR-NEXT: s_andn2_b64 exec, exec, s[8:9]
; GCN-IR-NEXT: s_cbranch_execnz .LBB1_3
; GCN-IR-NEXT: ; %bb.4: ; %Flow
-; GCN-IR-NEXT: s_or_b64 exec, exec, s[10:11]
-; GCN-IR-NEXT: .LBB1_5: ; %Flow4
; GCN-IR-NEXT: s_or_b64 exec, exec, s[8:9]
+; GCN-IR-NEXT: .LBB1_5: ; %Flow4
+; GCN-IR-NEXT: s_or_b64 exec, exec, s[4:5]
; GCN-IR-NEXT: v_lshl_b64 v[0:1], v[4:5], 1
; GCN-IR-NEXT: v_or_b32_e32 v4, v7, v1
; GCN-IR-NEXT: v_or_b32_e32 v5, v6, v0
@@ -1065,8 +1065,8 @@ define i64 @v_test_udiv_pow2_k_num_i64(i64 %x) {
; GCN-IR-NEXT: v_ffbh_u32_e32 v2, v0
; GCN-IR-NEXT: v_add_i32_e32 v2, vcc, 32, v2
; GCN-IR-NEXT: v_ffbh_u32_e32 v3, v1
-; GCN-IR-NEXT: v_min_u32_e32 v10, v2, v3
-; GCN-IR-NEXT: v_add_i32_e32 v6, vcc, 0xffffffd0, v10
+; GCN-IR-NEXT: v_min_u32_e32 v14, v2, v3
+; GCN-IR-NEXT: v_add_i32_e32 v6, vcc, 0xffffffd0, v14
; GCN-IR-NEXT: v_addc_u32_e64 v7, s[6:7], 0, -1, vcc
; GCN-IR-NEXT: v_cmp_eq_u64_e64 s[4:5], 0, v[0:1]
; GCN-IR-NEXT: v_cmp_lt_u64_e32 vcc, 63, v[6:7]
@@ -1081,54 +1081,54 @@ define i64 @v_test_udiv_pow2_k_num_i64(i64 %x) {
; GCN-IR-NEXT: s_cbranch_execz .LBB9_6
; GCN-IR-NEXT: ; %bb.1: ; %udiv-bb1
; GCN-IR-NEXT: v_add_i32_e32 v8, vcc, 1, v6
-; GCN-IR-NEXT: v_sub_i32_e64 v2, s[4:5], 63, v6
; GCN-IR-NEXT: v_addc_u32_e32 v9, vcc, 0, v7, vcc
-; GCN-IR-NEXT: s_mov_b64 s[4:5], 0x8000
+; GCN-IR-NEXT: v_sub_i32_e64 v2, s[4:5], 63, v6
+; GCN-IR-NEXT: s_mov_b64 s[8:9], 0x8000
; GCN-IR-NEXT: v_mov_b32_e32 v4, 0
; GCN-IR-NEXT: v_cmp_ne_u64_e32 vcc, 0, v[8:9]
-; GCN-IR-NEXT: v_lshl_b64 v[2:3], s[4:5], v2
+; GCN-IR-NEXT: v_lshl_b64 v[2:3], s[8:9], v2
; GCN-IR-NEXT: v_mov_b32_e32 v5, 0
-; GCN-IR-NEXT: s_and_saveexec_b64 s[8:9], vcc
-; GCN-IR-NEXT: s_xor_b64 s[8:9], exec, s[8:9]
+; GCN-IR-NEXT: s_and_saveexec_b64 s[4:5], vcc
+; GCN-IR-NEXT: s_xor_b64 s[4:5], exec, s[4:5]
; GCN-IR-NEXT: s_cbranch_execz .LBB9_5
; GCN-IR-NEXT: ; %bb.2: ; %udiv-preheader
; GCN-IR-NEXT: v_add_i32_e32 v12, vcc, -1, v0
; GCN-IR-NEXT: v_addc_u32_e32 v13, vcc, -1, v1, vcc
-; GCN-IR-NEXT: v_lshr_b64 v[8:9], s[4:5], v8
-; GCN-IR-NEXT: v_sub_i32_e32 v6, vcc, 47, v10
-; GCN-IR-NEXT: v_mov_b32_e32 v10, 0
-; GCN-IR-NEXT: v_subb_u32_e64 v7, s[4:5], 0, 0, vcc
-; GCN-IR-NEXT: s_mov_b64 s[10:11], 0
-; GCN-IR-NEXT: v_mov_b32_e32 v11, 0
+; GCN-IR-NEXT: v_sub_i32_e32 v6, vcc, 47, v14
+; GCN-IR-NEXT: v_lshr_b64 v[10:11], s[8:9], v8
+; GCN-IR-NEXT: v_subb_u32_e64 v7, s[8:9], 0, 0, vcc
+; GCN-IR-NEXT: v_mov_b32_e32 v8, 0
+; GCN-IR-NEXT: s_mov_b64 s[8:9], 0
+; GCN-IR-NEXT: v_mov_b32_e32 v9, 0
; GCN-IR-NEXT: v_mov_b32_e32 v5, 0
; GCN-IR-NEXT: .LBB9_3: ; %udiv-do-while
; GCN-IR-NEXT: ; =>This Inner Loop Header: Depth=1
-; GCN-IR-NEXT: v_lshl_b64 v[8:9], v[8:9], 1
+; GCN-IR-NEXT: v_lshl_b64 v[10:11], v[10:11], 1
; GCN-IR-NEXT: v_lshrrev_b32_e32 v4, 31, v3
-; GCN-IR-NEXT: v_or_b32_e32 v8, v8, v4
-; GCN-IR-NEXT: v_lshl_b64 v[2:3], v[2:3], 1
-; GCN-IR-NEXT: v_sub_i32_e32 v4, vcc, v12, v8
-; GCN-IR-NEXT: v_subb_u32_e32 v4, vcc, v13, v9, vcc
-; GCN-IR-NEXT: v_or_b32_e32 v2, v10, v2
-; GCN-IR-NEXT: v_ashrrev_i32_e32 v10, 31, v4
+; GCN-IR-NEXT: v_or_b32_e32 v10, v10, v4
+; GCN-IR-NEXT: v_sub_i32_e32 v4, vcc, v12, v10
+; GCN-IR-NEXT: v_subb_u32_e32 v4, vcc, v13, v11, vcc
; GCN-IR-NEXT: v_add_i32_e32 v6, vcc, 1, v6
-; GCN-IR-NEXT: v_or_b32_e32 v3, v11, v3
-; GCN-IR-NEXT: v_and_b32_e32 v4, 1, v10
-; GCN-IR-NEXT: v_and_b32_e32 v11, v10, v1
-; GCN-IR-NEXT: v_and_b32_e32 v10, v10, v0
; GCN-IR-NEXT: v_addc_u32_e32 v7, vcc, 0, v7, vcc
+; GCN-IR-NEXT: v_lshl_b64 v[2:3], v[2:3], 1
+; GCN-IR-NEXT: v_ashrrev_i32_e32 v14, 31, v4
; GCN-IR-NEXT: v_cmp_eq_u64_e32 vcc, 0, v[6:7]
-; GCN-IR-NEXT: v_sub_i32_e64 v8, s[4:5], v8, v10
-; GCN-IR-NEXT: v_subb_u32_e64 v9, s[4:5], v9, v11, s[4:5]
-; GCN-IR-NEXT: v_mov_b32_e32 v11, v5
-; GCN-IR-NEXT: s_or_b64 s[10:11], vcc, s[10:11]
-; GCN-IR-NEXT: v_mov_b32_e32 v10, v4
-; GCN-IR-NEXT: s_andn2_b64 exec, exec, s[10:11]
+; GCN-IR-NEXT: v_and_b32_e32 v4, 1, v14
+; GCN-IR-NEXT: v_and_b32_e32 v15, v14, v1
+; GCN-IR-NEXT: v_and_b32_e32 v14, v14, v0
+; GCN-IR-NEXT: v_or_b32_e32 v3, v9, v3
+; GCN-IR-NEXT: v_or_b32_e32 v2, v8, v2
+; GCN-IR-NEXT: v_mov_b32_e32 v9, v5
+; GCN-IR-NEXT: s_or_b64 s[8:9], vcc, s[8:9]
+; GCN-IR-NEXT: v_sub_i32_e32 v10, vcc, v10, v14
+; GCN-IR-NEXT: v_mov_b32_e32 v8, v4
+; GCN-IR-NEXT: v_subb_u32_e32 v11, vcc, v11, v15, vcc
+; GCN-IR-NEXT: s_andn2_b64 exec, exec, s[8:9]
; GCN-IR-NEXT: s_cbranch_execnz .LBB9_3
; GCN-IR-NEXT: ; %bb.4: ; %Flow
-; GCN-IR-NEXT: s_or_b64 exec, exec, s[10:11]
-; GCN-IR-NEXT: .LBB9_5: ; %Flow4
; GCN-IR-NEXT: s_or_b64 exec, exec, s[8:9]
+; GCN-IR-NEXT: .LBB9_5: ; %Flow4
+; GCN-IR-NEXT: s_or_b64 exec, exec, s[4:5]
; GCN-IR-NEXT: v_lshl_b64 v[0:1], v[2:3], 1
; GCN-IR-NEXT: v_or_b32_e32 v2, v5, v1
; GCN-IR-NEXT: v_or_b32_e32 v3, v4, v0
@@ -1177,44 +1177,44 @@ define i64 @v_test_udiv_pow2_k_den_i64(i64 %x) {
; GCN-IR-NEXT: v_lshl_b64 v[2:3], v[0:1], v2
; GCN-IR-NEXT: v_mov_b32_e32 v5, 0
; GCN-IR-NEXT: s_and_saveexec_b64 s[4:5], vcc
-; GCN-IR-NEXT: s_xor_b64 s[8:9], exec, s[4:5]
+; GCN-IR-NEXT: s_xor_b64 s[4:5], exec, s[4:5]
; GCN-IR-NEXT: s_cbranch_execz .LBB10_5
; GCN-IR-NEXT: ; %bb.2: ; %udiv-preheader
-; GCN-IR-NEXT: v_lshr_b64 v[6:7], v[0:1], v8
+; GCN-IR-NEXT: v_lshr_b64 v[8:9], v[0:1], v8
; GCN-IR-NEXT: v_add_i32_e32 v0, vcc, 0xffffffcf, v10
-; GCN-IR-NEXT: v_mov_b32_e32 v8, 0
-; GCN-IR-NEXT: v_addc_u32_e64 v1, s[4:5], 0, -1, vcc
-; GCN-IR-NEXT: s_mov_b64 s[10:11], 0
-; GCN-IR-NEXT: v_mov_b32_e32 v9, 0
-; GCN-IR-NEXT: s_movk_i32 s12, 0x7fff
+; GCN-IR-NEXT: v_addc_u32_e64 v1, s[8:9], 0, -1, vcc
+; GCN-IR-NEXT: v_mov_b32_e32 v6, 0
+; GCN-IR-NEXT: s_mov_b64 s[8:9], 0
+; GCN-IR-NEXT: v_mov_b32_e32 v7, 0
+; GCN-IR-NEXT: s_movk_i32 s10, 0x7fff
; GCN-IR-NEXT: v_mov_b32_e32 v5, 0
; GCN-IR-NEXT: .LBB10_3: ; %udiv-do-while
; GCN-IR-NEXT: ; =>This Inner Loop Header: Depth=1
-; GCN-IR-NEXT: v_lshl_b64 v[6:7], v[6:7], 1
+; GCN-IR-NEXT: v_lshl_b64 v[8:9], v[8:9], 1
; GCN-IR-NEXT: v_lshrrev_b32_e32 v4, 31, v3
-; GCN-IR-NEXT: v_or_b32_e32 v6, v6, v4
-; GCN-IR-NEXT: v_sub_i32_e32 v4, vcc, s12, v6
-; GCN-IR-NEXT: v_lshl_b64 v[2:3], v[2:3], 1
-; GCN-IR-NEXT: v_subb_u32_e32 v4, vcc, 0, v7, vcc
+; GCN-IR-NEXT: v_or_b32_e32 v8, v8, v4
+; GCN-IR-NEXT: v_sub_i32_e32 v4, vcc, s10, v8
+; GCN-IR-NEXT: v_subb_u32_e32 v4, vcc, 0, v9, vcc
; GCN-IR-NEXT: v_add_i32_e32 v0, vcc, 1, v0
-; GCN-IR-NEXT: v_or_b32_e32 v2, v8, v2
-; GCN-IR-NEXT: v_ashrrev_i32_e32 v8, 31, v4
; GCN-IR-NEXT: v_addc_u32_e32 v1, vcc, 0, v1, vcc
-; GCN-IR-NEXT: v_and_b32_e32 v4, 1, v8
-; GCN-IR-NEXT: v_and_b32_e32 v8, 0x8000, v8
+; GCN-IR-NEXT: v_lshl_b64 v[2:3], v[2:3], 1
+; GCN-IR-NEXT: v_ashrrev_i32_e32 v10, 31, v4
; GCN-IR-NEXT: v_cmp_eq_u64_e32 vcc, 0, v[0:1]
-; GCN-IR-NEXT: v_or_b32_e32 v3, v9, v3
-; GCN-IR-NEXT: v_sub_i32_e64 v6, s[4:5], v6, v8
-; GCN-IR-NEXT: v_mov_b32_e32 v9, v5
-; GCN-IR-NEXT: v_subbrev_u32_e64 v7, s[4:5], 0, v7, s[4:5]
-; GCN-IR-NEXT: s_or_b64 s[10:11], vcc, s[10:11]
-; GCN-IR-NEXT: v_mov_b32_e32 v8, v4
-; GCN-IR-NEXT: s_andn2_b64 exec, exec, s[10:11]
+; GCN-IR-NEXT: v_and_b32_e32 v4, 1, v10
+; GCN-IR-NEXT: v_and_b32_e32 v10, 0x8000, v10
+; GCN-IR-NEXT: v_or_b32_e32 v3, v7, v3
+; GCN-IR-NEXT: v_or_b32_e32 v2, v6, v2
+; GCN-IR-NEXT: v_mov_b32_e32 v7, v5
+; GCN-IR-NEXT: s_or_b64 s[8:9], vcc, s[8:9]
+; GCN-IR-NEXT: v_sub_i32_e32 v8, vcc, v8, v10
+; GCN-IR-NEXT: v_mov_b32_e32 v6, v4
+; GCN-IR-NEXT: v_subbrev_u32_e32 v9, vcc, 0, v9, vcc
+; GCN-IR-NEXT: s_andn2_b64 exec, exec, s[8:9]
; GCN-IR-NEXT: s_cbranch_execnz .LBB10_3
; GCN-IR-NEXT: ; %bb.4: ; %Flow
-; GCN-IR-NEXT: s_or_b64 exec, exec, s[10:11]
-; GCN-IR-NEXT: .LBB10_5: ; %Flow4
; GCN-IR-NEXT: s_or_b64 exec, exec, s[8:9]
+; GCN-IR-NEXT: .LBB10_5: ; %Flow4
+; GCN-IR-NEXT: s_or_b64 exec, exec, s[4:5]
; GCN-IR-NEXT: v_lshl_b64 v[0:1], v[2:3], 1
; GCN-IR-NEXT: v_or_b32_e32 v2, v5, v1
; GCN-IR-NEXT: v_or_b32_e32 v3, v4, v0
@@ -1377,43 +1377,43 @@ define i64 @v_test_udiv_k_den_i64(i64 %x) {
; GCN-IR-NEXT: v_lshl_b64 v[2:3], v[0:1], v2
; GCN-IR-NEXT: v_mov_b32_e32 v5, 0
; GCN-IR-NEXT: s_and_saveexec_b64 s[4:5], vcc
-; GCN-IR-NEXT: s_xor_b64 s[8:9], exec, s[4:5]
+; GCN-IR-NEXT: s_xor_b64 s[4:5], exec, s[4:5]
; GCN-IR-NEXT: s_cbranch_execz .LBB12_5
; GCN-IR-NEXT: ; %bb.2: ; %udiv-preheader
-; GCN-IR-NEXT: v_lshr_b64 v[6:7], v[0:1], v8
+; GCN-IR-NEXT: v_lshr_b64 v[8:9], v[0:1], v8
; GCN-IR-NEXT: v_add_i32_e32 v0, vcc, 0xffffffc4, v10
-; GCN-IR-NEXT: v_mov_b32_e32 v8, 0
-; GCN-IR-NEXT: v_addc_u32_e64 v1, s[4:5], 0, -1, vcc
-; GCN-IR-NEXT: s_mov_b64 s[10:11], 0
-; GCN-IR-NEXT: v_mov_b32_e32 v9, 0
+; GCN-IR-NEXT: v_addc_u32_e64 v1, s[8:9], 0, -1, vcc
+; GCN-IR-NEXT: v_mov_b32_e32 v6, 0
+; GCN-IR-NEXT: s_mov_b64 s[8:9], 0
+; GCN-IR-NEXT: v_mov_b32_e32 v7, 0
; GCN-IR-NEXT: v_mov_b32_e32 v5, 0
; GCN-IR-NEXT: .LBB12_3: ; %udiv-do-while
; GCN-IR-NEXT: ; =>This Inner Loop Header: Depth=1
-; GCN-IR-NEXT: v_lshl_b64 v[6:7], v[6:7], 1
+; GCN-IR-NEXT: v_lshl_b64 v[8:9], v[8:9], 1
; GCN-IR-NEXT: v_lshrrev_b32_e32 v4, 31, v3
-; GCN-IR-NEXT: v_or_b32_e32 v6, v6, v4
-; GCN-IR-NEXT: v_sub_i32_e32 v4, vcc, 23, v6
-; GCN-IR-NEXT: v_lshl_b64 v[2:3], v[2:3], 1
-; GCN-IR-NEXT: v_subb_u32_e32 v4, vcc, 0, v7, vcc
+; GCN-IR-NEXT: v_or_b32_e32 v8, v8, v4
+; GCN-IR-NEXT: v_sub_i32_e32 v4, vcc, 23, v8
+; GCN-IR-NEXT: v_subb_u32_e32 v4, vcc, 0, v9, vcc
; GCN-IR-NEXT: v_add_i32_e32 v0, vcc, 1, v0
-; GCN-IR-NEXT: v_or_b32_e32 v2, v8, v2
-; GCN-IR-NEXT: v_ashrrev_i32_e32 v8, 31, v4
; GCN-IR-NEXT: v_addc_u32_e32 v1, vcc, 0, v1, vcc
-; GCN-IR-NEXT: v_and_b32_e32 v4, 1, v8
-; GCN-IR-NEXT: v_and_b32_e32 v8, 24, v8
+; GCN-IR-NEXT: v_lshl_b64 v[2:3], v[2:3], 1
+; GCN-IR-NEXT: v_ashrrev_i32_e32 v10, 31, v4
; GCN-IR-NEXT: v_cmp_eq_u64_e32 vcc, 0, v[0:1]
-; GCN-IR-NEXT: v_or_b32_e32 v3, v9, v3
-; GCN-IR-NEXT: v_sub_i32_e64 v6, s[4:5], v6, v8
-; GCN-IR-NEXT: v_mov_b32_e32 v9, v5
-; GCN-IR-NEXT: v_subbrev_u32_e64 v7, s[4:5], 0, v7, s[4:5]
-; GCN-IR-NEXT: s_or_b64 s[10:11], vcc, s[10:11]
-; GCN-IR-NEXT: v_mov_b32_e32 v8, v4
-; GCN-IR-NEXT: s_andn2_b64 exec, exec, s[10:11]
+; GCN-IR-NEXT: v_and_b32_e32 v4, 1, v10
+; GCN-IR-NEXT: v_and_b32_e32 v10, 24, v10
+; GCN-IR-NEXT: v_or_b32_e32 v3, v7, v3
+; GCN-IR-NEXT: v_or_b32_e32 v2, v6, v2
+; GCN-IR-NEXT: v_mov_b32_e32 v7, v5
+; GCN-IR-NEXT: s_or_b64 s[8:9], vcc, s[8:9]
+; GCN-IR-NEXT: v_sub_i32_e32 v8, vcc, v8, v10
+; GCN-IR-NEXT: v_mov_b32_e32 v6, v4
+; GCN-IR-NEXT: v_subbrev_u32_e32 v9, vcc, 0, v9, vcc
+; GCN-IR-NEXT: s_andn2_b64 exec, exec, s[8:9]
; GCN-IR-NEXT: s_cbranch_execnz .LBB12_3
; GCN-IR-NEXT: ; %bb.4: ; %Flow
-; GCN-IR-NEXT: s_or_b64 exec, exec, s[10:11]
-; GCN-IR-NEXT: .LBB12_5: ; %Flow4
; GCN-IR-NEXT: s_or_b64 exec, exec, s[8:9]
+; GCN-IR-NEXT: .LBB12_5: ; %Flow4
+; GCN-IR-NEXT: s_or_b64 exec, exec, s[4:5]
; GCN-IR-NEXT: v_lshl_b64 v[0:1], v[2:3], 1
; GCN-IR-NEXT: v_or_b32_e32 v2, v5, v1
; GCN-IR-NEXT: v_or_b32_e32 v3, v4, v0
diff --git a/llvm/test/CodeGen/AMDGPU/urem64.ll b/llvm/test/CodeGen/AMDGPU/urem64.ll
index 6480a88d40f5a..bd742968ba37b 100644
--- a/llvm/test/CodeGen/AMDGPU/urem64.ll
+++ b/llvm/test/CodeGen/AMDGPU/urem64.ll
@@ -322,12 +322,12 @@ define i64 @v_test_urem_i64(i64 %x, i64 %y) {
; GCN-IR-NEXT: v_ffbh_u32_e32 v4, v2
; GCN-IR-NEXT: v_add_i32_e64 v4, s[6:7], 32, v4
; GCN-IR-NEXT: v_ffbh_u32_e32 v5, v3
-; GCN-IR-NEXT: v_min_u32_e32 v12, v4, v5
+; GCN-IR-NEXT: v_min_u32_e32 v10, v4, v5
; GCN-IR-NEXT: v_ffbh_u32_e32 v4, v0
; GCN-IR-NEXT: v_add_i32_e64 v4, s[6:7], 32, v4
; GCN-IR-NEXT: v_ffbh_u32_e32 v5, v1
-; GCN-IR-NEXT: v_min_u32_e32 v13, v4, v5
-; GCN-IR-NEXT: v_sub_i32_e64 v4, s[6:7], v12, v13
+; GCN-IR-NEXT: v_min_u32_e32 v11, v4, v5
+; GCN-IR-NEXT: v_sub_i32_e64 v4, s[6:7], v10, v11
; GCN-IR-NEXT: v_cmp_eq_u64_e32 vcc, 0, v[2:3]
; GCN-IR-NEXT: v_cmp_eq_u64_e64 s[4:5], 0, v[0:1]
; GCN-IR-NEXT: v_subb_u32_e64 v5, s[6:7], 0, 0, s[6:7]
@@ -350,47 +350,47 @@ define i64 @v_test_urem_i64(i64 %x, i64 %y) {
; GCN-IR-NEXT: v_lshl_b64 v[4:5], v[0:1], v4
; GCN-IR-NEXT: v_mov_b32_e32 v7, 0
; GCN-IR-NEXT: s_and_saveexec_b64 s[4:5], vcc
-; GCN-IR-NEXT: s_xor_b64 s[8:9], exec, s[4:5]
+; GCN-IR-NEXT: s_xor_b64 s[4:5], exec, s[4:5]
; GCN-IR-NEXT: s_cbranch_execz .LBB1_5
; GCN-IR-NEXT: ; %bb.2: ; %udiv-preheader
; GCN-IR-NEXT: v_add_i32_e32 v14, vcc, -1, v2
; GCN-IR-NEXT: v_addc_u32_e32 v15, vcc, -1, v3, vcc
-; GCN-IR-NEXT: v_not_b32_e32 v6, v12
-; GCN-IR-NEXT: v_lshr_b64 v[10:11], v[0:1], v8
-; GCN-IR-NEXT: v_add_i32_e32 v8, vcc, v6, v13
-; GCN-IR-NEXT: v_mov_b32_e32 v12, 0
-; GCN-IR-NEXT: v_addc_u32_e64 v9, s[4:5], -1, 0, vcc
-; GCN-IR-NEXT: s_mov_b64 s[10:11], 0
-; GCN-IR-NEXT: v_mov_b32_e32 v13, 0
+; GCN-IR-NEXT: v_not_b32_e32 v6, v10
+; GCN-IR-NEXT: v_lshr_b64 v[12:13], v[0:1], v8
+; GCN-IR-NEXT: v_add_i32_e32 v8, vcc, v6, v11
+; GCN-IR-NEXT: v_addc_u32_e64 v9, s[8:9], -1, 0, vcc
+; GCN-IR-NEXT: v_mov_b32_e32 v10, 0
+; GCN-IR-NEXT: s_mov_b64 s[8:9], 0
+; GCN-IR-NEXT: v_mov_b32_e32 v11, 0
; GCN-IR-NEXT: v_mov_b32_e32 v7, 0
; GCN-IR-NEXT: .LBB1_3: ; %udiv-do-while
; GCN-IR-NEXT: ; =>This Inner Loop Header: Depth=1
-; GCN-IR-NEXT: v_lshl_b64 v[10:11], v[10:11], 1
+; GCN-IR-NEXT: v_lshl_b64 v[12:13], v[12:13], 1
; GCN-IR-NEXT: v_lshrrev_b32_e32 v6, 31, v5
-; GCN-IR-NEXT: v_or_b32_e32 v10, v10, v6
-; GCN-IR-NEXT: v_lshl_b64 v[4:5], v[4:5], 1
-; GCN-IR-NEXT: v_sub_i32_e32 v6, vcc, v14, v10
-; GCN-IR-NEXT: v_subb_u32_e32 v6, vcc, v15, v11, vcc
-; GCN-IR-NEXT: v_or_b32_e32 v4, v12, v4
-; GCN-IR-NEXT: v_ashrrev_i32_e32 v12, 31, v6
+; GCN-IR-NEXT: v_or_b32_e32 v12, v12, v6
+; GCN-IR-NEXT: v_sub_i32_e32 v6, vcc, v14, v12
+; GCN-IR-NEXT: v_subb_u32_e32 v6, vcc, v15, v13, vcc
; GCN-IR-NEXT: v_add_i32_e32 v8, vcc, 1, v8
-; GCN-IR-NEXT: v_or_b32_e32 v5, v13, v5
-; GCN-IR-NEXT: v_and_b32_e32 v6, 1, v12
-; GCN-IR-NEXT: v_and_b32_e32 v13, v12, v3
-; GCN-IR-NEXT: v_and_b32_e32 v12, v12, v2
; GCN-IR-NEXT: v_addc_u32_e32 v9, vcc, 0, v9, vcc
+; GCN-IR-NEXT: v_lshl_b64 v[4:5], v[4:5], 1
+; GCN-IR-NEXT: v_ashrrev_i32_e32 v16, 31, v6
; GCN-IR-NEXT: v_cmp_eq_u64_e32 vcc, 0, v[8:9]
-; GCN-IR-NEXT: v_sub_i32_e64 v10, s[4:5], v10, v12
-; GCN-IR-NEXT: v_subb_u32_e64 v11, s[4:5], v11, v13, s[4:5]
-; GCN-IR-NEXT: v_mov_b32_e32 v13, v7
-; GCN-IR-NEXT: s_or_b64 s[10:11], vcc, s[10:11]
-; GCN-IR-NEXT: v_mov_b32_e32 v12, v6
-; GCN-IR-NEXT: s_andn2_b64 exec, exec, s[10:11]
+; GCN-IR-NEXT: v_and_b32_e32 v6, 1, v16
+; GCN-IR-NEXT: v_and_b32_e32 v17, v16, v3
+; GCN-IR-NEXT: v_and_b32_e32 v16, v16, v2
+; GCN-IR-NEXT: v_or_b32_e32 v5, v11, v5
+; GCN-IR-NEXT: v_or_b32_e32 v4, v10, v4
+; GCN-IR-NEXT: v_mov_b32_e32 v11, v7
+; GCN-IR-NEXT: s_or_b64 s[8:9], vcc, s[8:9]
+; GCN-IR-NEXT: v_sub_i32_e32 v12, vcc, v12, v16
+; GCN-IR-NEXT: v_mov_b32_e32 v10, v6
+; GCN-IR-NEXT: v_subb_u32_e32 v13, vcc, v13, v17, vcc
+; GCN-IR-NEXT: s_andn2_b64 exec, exec, s[8:9]
; GCN-IR-NEXT: s_cbranch_execnz .LBB1_3
; GCN-IR-NEXT: ; %bb.4: ; %Flow
-; GCN-IR-NEXT: s_or_b64 exec, exec, s[10:11]
-; GCN-IR-NEXT: .LBB1_5: ; %Flow4
; GCN-IR-NEXT: s_or_b64 exec, exec, s[8:9]
+; GCN-IR-NEXT: .LBB1_5: ; %Flow4
+; GCN-IR-NEXT: s_or_b64 exec, exec, s[4:5]
; GCN-IR-NEXT: v_lshl_b64 v[4:5], v[4:5], 1
; GCN-IR-NEXT: v_or_b32_e32 v7, v7, v5
; GCN-IR-NEXT: v_or_b32_e32 v6, v6, v4
@@ -1166,8 +1166,8 @@ define i64 @v_test_urem_pow2_k_num_i64(i64 %x) {
; GCN-IR-NEXT: v_ffbh_u32_e32 v2, v0
; GCN-IR-NEXT: v_add_i32_e32 v2, vcc, 32, v2
; GCN-IR-NEXT: v_ffbh_u32_e32 v3, v1
-; GCN-IR-NEXT: v_min_u32_e32 v10, v2, v3
-; GCN-IR-NEXT: v_add_i32_e32 v2, vcc, 0xffffffd0, v10
+; GCN-IR-NEXT: v_min_u32_e32 v8, v2, v3
+; GCN-IR-NEXT: v_add_i32_e32 v2, vcc, 0xffffffd0, v8
; GCN-IR-NEXT: v_addc_u32_e64 v3, s[6:7], 0, -1, vcc
; GCN-IR-NEXT: v_cmp_eq_u64_e64 s[4:5], 0, v[0:1]
; GCN-IR-NEXT: v_cmp_lt_u64_e32 vcc, 63, v[2:3]
@@ -1182,54 +1182,54 @@ define i64 @v_test_urem_pow2_k_num_i64(i64 %x) {
; GCN-IR-NEXT: s_cbranch_execz .LBB8_6
; GCN-IR-NEXT: ; %bb.1: ; %udiv-bb1
; GCN-IR-NEXT: v_add_i32_e32 v6, vcc, 1, v2
-; GCN-IR-NEXT: v_sub_i32_e64 v2, s[4:5], 63, v2
; GCN-IR-NEXT: v_addc_u32_e32 v7, vcc, 0, v3, vcc
-; GCN-IR-NEXT: s_mov_b64 s[4:5], 0x8000
+; GCN-IR-NEXT: v_sub_i32_e64 v2, s[4:5], 63, v2
+; GCN-IR-NEXT: s_mov_b64 s[8:9], 0x8000
; GCN-IR-NEXT: v_mov_b32_e32 v4, 0
; GCN-IR-NEXT: v_cmp_ne_u64_e32 vcc, 0, v[6:7]
-; GCN-IR-NEXT: v_lshl_b64 v[2:3], s[4:5], v2
+; GCN-IR-NEXT: v_lshl_b64 v[2:3], s[8:9], v2
; GCN-IR-NEXT: v_mov_b32_e32 v5, 0
-; GCN-IR-NEXT: s_and_saveexec_b64 s[8:9], vcc
-; GCN-IR-NEXT: s_xor_b64 s[8:9], exec, s[8:9]
+; GCN-IR-NEXT: s_and_saveexec_b64 s[4:5], vcc
+; GCN-IR-NEXT: s_xor_b64 s[4:5], exec, s[4:5]
; GCN-IR-NEXT: s_cbranch_execz .LBB8_5
; GCN-IR-NEXT: ; %bb.2: ; %udiv-preheader
; GCN-IR-NEXT: v_add_i32_e32 v12, vcc, -1, v0
; GCN-IR-NEXT: v_addc_u32_e32 v13, vcc, -1, v1, vcc
-; GCN-IR-NEXT: v_lshr_b64 v[8:9], s[4:5], v6
-; GCN-IR-NEXT: v_sub_i32_e32 v6, vcc, 47, v10
-; GCN-IR-NEXT: v_mov_b32_e32 v10, 0
-; GCN-IR-NEXT: v_subb_u32_e64 v7, s[4:5], 0, 0, vcc
-; GCN-IR-NEXT: s_mov_b64 s[10:11], 0
-; GCN-IR-NEXT: v_mov_b32_e32 v11, 0
+; GCN-IR-NEXT: v_lshr_b64 v[10:11], s[8:9], v6
+; GCN-IR-NEXT: v_sub_i32_e32 v6, vcc, 47, v8
+; GCN-IR-NEXT: v_subb_u32_e64 v7, s[8:9], 0, 0, vcc
+; GCN-IR-NEXT: v_mov_b32_e32 v8, 0
+; GCN-IR-NEXT: s_mov_b64 s[8:9], 0
+; GCN-IR-NEXT: v_mov_b32_e32 v9, 0
; GCN-IR-NEXT: v_mov_b32_e32 v5, 0
; GCN-IR-NEXT: .LBB8_3: ; %udiv-do-while
; GCN-IR-NEXT: ; =>This Inner Loop Header: Depth=1
-; GCN-IR-NEXT: v_lshl_b64 v[8:9], v[8:9], 1
+; GCN-IR-NEXT: v_lshl_b64 v[10:11], v[10:11], 1
; GCN-IR-NEXT: v_lshrrev_b32_e32 v4, 31, v3
-; GCN-IR-NEXT: v_or_b32_e32 v8, v8, v4
-; GCN-IR-NEXT: v_lshl_b64 v[2:3], v[2:3], 1
-; GCN-IR-NEXT: v_sub_i32_e32 v4, vcc, v12, v8
-; GCN-IR-NEXT: v_subb_u32_e32 v4, vcc, v13, v9, vcc
-; GCN-IR-NEXT: v_or_b32_e32 v2, v10, v2
-; GCN-IR-NEXT: v_ashrrev_i32_e32 v10, 31, v4
+; GCN-IR-NEXT: v_or_b32_e32 v10, v10, v4
+; GCN-IR-NEXT: v_sub_i32_e32 v4, vcc, v12, v10
+; GCN-IR-NEXT: v_subb_u32_e32 v4, vcc, v13, v11, vcc
; GCN-IR-NEXT: v_add_i32_e32 v6, vcc, 1, v6
-; GCN-IR-NEXT: v_or_b32_e32 v3, v11, v3
-; GCN-IR-NEXT: v_and_b32_e32 v4, 1, v10
-; GCN-IR-NEXT: v_and_b32_e32 v11, v10, v1
-; GCN-IR-NEXT: v_and_b32_e32 v10, v10, v0
; GCN-IR-NEXT: v_addc_u32_e32 v7, vcc, 0, v7, vcc
+; GCN-IR-NEXT: v_lshl_b64 v[2:3], v[2:3], 1
+; GCN-IR-NEXT: v_ashrrev_i32_e32 v14, 31, v4
; GCN-IR-NEXT: v_cmp_eq_u64_e32 vcc, 0, v[6:7]
-; GCN-IR-NEXT: v_sub_i32_e64 v8, s[4:5], v8, v10
-; GCN-IR-NEXT: v_subb_u32_e64 v9, s[4:5], v9, v11, s[4:5]
-; GCN-IR-NEXT: v_mov_b32_e32 v11, v5
-; GCN-IR-NEXT: s_or_b64 s[10:11], vcc, s[10:11]
-; GCN-IR-NEXT: v_mov_b32_e32 v10, v4
-; GCN-IR-NEXT: s_andn2_b64 exec, exec, s[10:11]
+; GCN-IR-NEXT: v_and_b32_e32 v4, 1, v14
+; GCN-IR-NEXT: v_and_b32_e32 v15, v14, v1
+; GCN-IR-NEXT: v_and_b32_e32 v14, v14, v0
+; GCN-IR-NEXT: v_or_b32_e32 v3, v9, v3
+; GCN-IR-NEXT: v_or_b32_e32 v2, v8, v2
+; GCN-IR-NEXT: v_mov_b32_e32 v9, v5
+; GCN-IR-NEXT: s_or_b64 s[8:9], vcc, s[8:9]
+; GCN-IR-NEXT: v_sub_i32_e32 v10, vcc, v10, v14
+; GCN-IR-NEXT: v_mov_b32_e32 v8, v4
+; GCN-IR-NEXT: v_subb_u32_e32 v11, vcc, v11, v15, vcc
+; GCN-IR-NEXT: s_andn2_b64 exec, exec, s[8:9]
; GCN-IR-NEXT: s_cbranch_execnz .LBB8_3
; GCN-IR-NEXT: ; %bb.4: ; %Flow
-; GCN-IR-NEXT: s_or_b64 exec, exec, s[10:11]
-; GCN-IR-NEXT: .LBB8_5: ; %Flow4
; GCN-IR-NEXT: s_or_b64 exec, exec, s[8:9]
+; GCN-IR-NEXT: .LBB8_5: ; %Flow4
+; GCN-IR-NEXT: s_or_b64 exec, exec, s[4:5]
; GCN-IR-NEXT: v_lshl_b64 v[2:3], v[2:3], 1
; GCN-IR-NEXT: v_or_b32_e32 v5, v5, v3
; GCN-IR-NEXT: v_or_b32_e32 v4, v4, v2
@@ -1262,8 +1262,8 @@ define i64 @v_test_urem_pow2_k_den_i64(i64 %x) {
; GCN-IR-NEXT: v_ffbh_u32_e32 v2, v0
; GCN-IR-NEXT: v_add_i32_e64 v2, s[4:5], 32, v2
; GCN-IR-NEXT: v_ffbh_u32_e32 v3, v1
-; GCN-IR-NEXT: v_min_u32_e32 v10, v2, v3
-; GCN-IR-NEXT: v_sub_i32_e64 v2, s[4:5], 48, v10
+; GCN-IR-NEXT: v_min_u32_e32 v8, v2, v3
+; GCN-IR-NEXT: v_sub_i32_e64 v2, s[4:5], 48, v8
; GCN-IR-NEXT: v_subb_u32_e64 v3, s[4:5], 0, 0, s[4:5]
; GCN-IR-NEXT: v_cmp_eq_u64_e32 vcc, 0, v[0:1]
; GCN-IR-NEXT: v_cmp_lt_u64_e64 s[4:5], 63, v[2:3]
@@ -1284,44 +1284,44 @@ define i64 @v_test_urem_pow2_k_den_i64(i64 %x) {
; GCN-IR-NEXT: v_lshl_b64 v[2:3], v[0:1], v2
; GCN-IR-NEXT: v_mov_b32_e32 v5, 0
; GCN-IR-NEXT: s_and_saveexec_b64 s[4:5], vcc
-; GCN-IR-NEXT: s_xor_b64 s[8:9], exec, s[4:5]
+; GCN-IR-NEXT: s_xor_b64 s[4:5], exec, s[4:5]
; GCN-IR-NEXT: s_cbranch_execz .LBB9_5
; GCN-IR-NEXT: ; %bb.2: ; %udiv-preheader
-; GCN-IR-NEXT: v_lshr_b64 v[8:9], v[0:1], v6
-; GCN-IR-NEXT: v_add_i32_e32 v6, vcc, 0xffffffcf, v10
-; GCN-IR-NEXT: v_mov_b32_e32 v10, 0
-; GCN-IR-NEXT: v_addc_u32_e64 v7, s[4:5], 0, -1, vcc
-; GCN-IR-NEXT: s_mov_b64 s[10:11], 0
-; GCN-IR-NEXT: v_mov_b32_e32 v11, 0
-; GCN-IR-NEXT: s_movk_i32 s12, 0x7fff
+; GCN-IR-NEXT: v_lshr_b64 v[10:11], v[0:1], v6
+; GCN-IR-NEXT: v_add_i32_e32 v6, vcc, 0xffffffcf, v8
+; GCN-IR-NEXT: v_addc_u32_e64 v7, s[8:9], 0, -1, vcc
+; GCN-IR-NEXT: v_mov_b32_e32 v8, 0
+; GCN-IR-NEXT: s_mov_b64 s[8:9], 0
+; GCN-IR-NEXT: v_mov_b32_e32 v9, 0
+; GCN-IR-NEXT: s_movk_i32 s10, 0x7fff
; GCN-IR-NEXT: v_mov_b32_e32 v5, 0
; GCN-IR-NEXT: .LBB9_3: ; %udiv-do-while
; GCN-IR-NEXT: ; =>This Inner Loop Header: Depth=1
-; GCN-IR-NEXT: v_lshl_b64 v[8:9], v[8:9], 1
+; GCN-IR-NEXT: v_lshl_b64 v[10:11], v[10:11], 1
; GCN-IR-NEXT: v_lshrrev_b32_e32 v4, 31, v3
-; GCN-IR-NEXT: v_or_b32_e32 v8, v8, v4
-; GCN-IR-NEXT: v_sub_i32_e32 v4, vcc, s12, v8
-; GCN-IR-NEXT: v_lshl_b64 v[2:3], v[2:3], 1
-; GCN-IR-NEXT: v_subb_u32_e32 v4, vcc, 0, v9, vcc
+; GCN-IR-NEXT: v_or_b32_e32 v10, v10, v4
+; GCN-IR-NEXT: v_sub_i32_e32 v4, vcc, s10, v10
+; GCN-IR-NEXT: v_subb_u32_e32 v4, vcc, 0, v11, vcc
; GCN-IR-NEXT: v_add_i32_e32 v6, vcc, 1, v6
-; GCN-IR-NEXT: v_or_b32_e32 v2, v10, v2
-; GCN-IR-NEXT: v_ashrrev_i32_e32 v10, 31, v4
; GCN-IR-NEXT: v_addc_u32_e32 v7, vcc, 0, v7, vcc
-; GCN-IR-NEXT: v_and_b32_e32 v4, 1, v10
-; GCN-IR-NEXT: v_and_b32_e32 v10, 0x8000, v10
+; GCN-IR-NEXT: v_lshl_b64 v[2:3], v[2:3], 1
+; GCN-IR-NEXT: v_ashrrev_i32_e32 v12, 31, v4
; GCN-IR-NEXT: v_cmp_eq_u64_e32 vcc, 0, v[6:7]
-; GCN-IR-NEXT: v_or_b32_e32 v3, v11, v3
-; GCN-IR-NEXT: v_sub_i32_e64 v8, s[4:5], v8, v10
-; GCN-IR-NEXT: v_mov_b32_e32 v11, v5
-; GCN-IR-NEXT: v_subbrev_u32_e64 v9, s[4:5], 0, v9, s[4:5]
-; GCN-IR-NEXT: s_or_b64 s[10:11], vcc, s[10:11]
-; GCN-IR-NEXT: v_mov_b32_e32 v10, v4
-; GCN-IR-NEXT: s_andn2_b64 exec, exec, s[10:11]
+; GCN-IR-NEXT: v_and_b32_e32 v4, 1, v12
+; GCN-IR-NEXT: v_and_b32_e32 v12, 0x8000, v12
+; GCN-IR-NEXT: v_or_b32_e32 v3, v9, v3
+; GCN-IR-NEXT: v_or_b32_e32 v2, v8, v2
+; GCN-IR-NEXT: v_mov_b32_e32 v9, v5
+; GCN-IR-NEXT: s_or_b64 s[8:9], vcc, s[8:9]
+; GCN-IR-NEXT: v_sub_i32_e32 v10, vcc, v10, v12
+; GCN-IR-NEXT: v_mov_b32_e32 v8, v4
+; GCN-IR-NEXT: v_subbrev_u32_e32 v11, vcc, 0, v11, vcc
+; GCN-IR-NEXT: s_andn2_b64 exec, exec, s[8:9]
; GCN-IR-NEXT: s_cbranch_execnz .LBB9_3
; GCN-IR-NEXT: ; %bb.4: ; %Flow
-; GCN-IR-NEXT: s_or_b64 exec, exec, s[10:11]
-; GCN-IR-NEXT: .LBB9_5: ; %Flow4
; GCN-IR-NEXT: s_or_b64 exec, exec, s[8:9]
+; GCN-IR-NEXT: .LBB9_5: ; %Flow4
+; GCN-IR-NEXT: s_or_b64 exec, exec, s[4:5]
; GCN-IR-NEXT: v_lshl_b64 v[2:3], v[2:3], 1
; GCN-IR-NEXT: v_or_b32_e32 v5, v5, v3
; GCN-IR-NEXT: v_or_b32_e32 v4, v4, v2
diff --git a/llvm/test/CodeGen/AMDGPU/vni8-across-blocks.ll b/llvm/test/CodeGen/AMDGPU/vni8-across-blocks.ll
index a401f989a2507..310f3ad04917b 100644
--- a/llvm/test/CodeGen/AMDGPU/vni8-across-blocks.ll
+++ b/llvm/test/CodeGen/AMDGPU/vni8-across-blocks.ll
@@ -508,12 +508,12 @@ define amdgpu_kernel void @v8i8_phi_const(ptr addrspace(1) %src1, ptr addrspace(
; GFX942: ; %bb.0: ; %entry
; GFX942-NEXT: s_load_dwordx8 s[8:15], s[4:5], 0x24
; GFX942-NEXT: v_and_b32_e32 v16, 0x3ff, v0
-; GFX942-NEXT: v_lshlrev_b32_e32 v3, 3, v16
+; GFX942-NEXT: v_lshlrev_b32_e32 v1, 3, v16
; GFX942-NEXT: v_cmp_lt_u32_e64 s[0:1], 14, v16
; GFX942-NEXT: v_cmp_gt_u32_e32 vcc, 15, v16
; GFX942-NEXT: s_waitcnt lgkmcnt(0)
-; GFX942-NEXT: global_load_dwordx2 v[0:1], v3, s[8:9]
-; GFX942-NEXT: ; implicit-def: $vgpr2
+; GFX942-NEXT: global_load_dwordx2 v[2:3], v1, s[8:9]
+; GFX942-NEXT: ; implicit-def: $vgpr0
; GFX942-NEXT: ; implicit-def: $vgpr12
; GFX942-NEXT: ; implicit-def: $vgpr10
; GFX942-NEXT: ; implicit-def: $vgpr13
@@ -521,74 +521,75 @@ define amdgpu_kernel void @v8i8_phi_const(ptr addrspace(1) %src1, ptr addrspace(
; GFX942-NEXT: ; implicit-def: $vgpr11
; GFX942-NEXT: ; implicit-def: $vgpr15
; GFX942-NEXT: s_waitcnt vmcnt(0)
-; GFX942-NEXT: v_lshrrev_b32_e32 v4, 24, v1
-; GFX942-NEXT: v_lshrrev_b32_e32 v5, 16, v1
-; GFX942-NEXT: v_lshrrev_b32_e32 v6, 8, v1
-; GFX942-NEXT: v_lshrrev_b32_e32 v7, 24, v0
-; GFX942-NEXT: v_lshrrev_b32_e32 v8, 16, v0
-; GFX942-NEXT: v_lshrrev_b32_e32 v9, 8, v0
+; GFX942-NEXT: v_lshrrev_b32_e32 v4, 24, v3
+; GFX942-NEXT: v_lshrrev_b32_e32 v5, 16, v3
+; GFX942-NEXT: v_lshrrev_b32_e32 v6, 8, v3
+; GFX942-NEXT: v_lshrrev_b32_e32 v7, 24, v2
+; GFX942-NEXT: v_lshrrev_b32_e32 v8, 16, v2
+; GFX942-NEXT: v_lshrrev_b32_e32 v9, 8, v2
; GFX942-NEXT: s_and_saveexec_b64 s[2:3], vcc
; GFX942-NEXT: s_cbranch_execz .LBB10_2
; GFX942-NEXT: ; %bb.1: ; %bb.1
-; GFX942-NEXT: global_load_dwordx2 v[2:3], v3, s[10:11]
+; GFX942-NEXT: global_load_dwordx2 v[0:1], v1, s[10:11]
; GFX942-NEXT: v_cmp_gt_u32_e32 vcc, 7, v16
; GFX942-NEXT: s_andn2_b64 s[0:1], s[0:1], exec
; GFX942-NEXT: s_and_b64 s[4:5], vcc, exec
; GFX942-NEXT: v_mov_b32_e32 v4, 8
; GFX942-NEXT: v_mov_b32_e32 v5, 7
; GFX942-NEXT: v_mov_b32_e32 v6, 6
-; GFX942-NEXT: v_mov_b32_e32 v1, 5
+; GFX942-NEXT: v_mov_b32_e32 v3, 5
; GFX942-NEXT: v_mov_b32_e32 v7, 4
; GFX942-NEXT: v_mov_b32_e32 v8, 3
; GFX942-NEXT: v_mov_b32_e32 v9, 2
-; GFX942-NEXT: v_mov_b32_e32 v0, 1
+; GFX942-NEXT: v_mov_b32_e32 v2, 1
; GFX942-NEXT: s_or_b64 s[0:1], s[0:1], s[4:5]
; GFX942-NEXT: s_waitcnt vmcnt(0)
-; GFX942-NEXT: v_lshrrev_b32_e32 v15, 24, v3
-; GFX942-NEXT: v_lshrrev_b32_e32 v11, 16, v3
-; GFX942-NEXT: v_lshrrev_b32_e32 v14, 8, v3
-; GFX942-NEXT: v_lshrrev_b32_e32 v13, 24, v2
-; GFX942-NEXT: v_lshrrev_b32_e32 v10, 16, v2
-; GFX942-NEXT: v_lshrrev_b32_e32 v12, 8, v2
+; GFX942-NEXT: v_lshrrev_b32_e32 v15, 24, v1
+; GFX942-NEXT: v_lshrrev_b32_e32 v11, 16, v1
+; GFX942-NEXT: v_lshrrev_b32_e32 v14, 8, v1
+; GFX942-NEXT: v_lshrrev_b32_e32 v13, 24, v0
+; GFX942-NEXT: v_lshrrev_b32_e32 v10, 16, v0
+; GFX942-NEXT: v_lshrrev_b32_e32 v12, 8, v0
; GFX942-NEXT: .LBB10_2: ; %Flow
; GFX942-NEXT: s_or_b64 exec, exec, s[2:3]
; GFX942-NEXT: s_and_saveexec_b64 s[2:3], s[0:1]
; GFX942-NEXT: s_cbranch_execz .LBB10_4
; GFX942-NEXT: ; %bb.3: ; %bb.2
-; GFX942-NEXT: v_lshlrev_b16_e32 v2, 8, v9
-; GFX942-NEXT: v_lshlrev_b16_e32 v3, 8, v7
-; GFX942-NEXT: v_or_b32_sdwa v2, v0, v2 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
-; GFX942-NEXT: v_or_b32_sdwa v3, v8, v3 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; GFX942-NEXT: v_lshlrev_b16_e32 v0, 8, v9
+; GFX942-NEXT: v_lshlrev_b16_e32 v1, 8, v7
+; GFX942-NEXT: v_or_b32_sdwa v0, v2, v0 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; GFX942-NEXT: v_or_b32_sdwa v1, v8, v1 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
; GFX942-NEXT: v_lshlrev_b16_e32 v11, 8, v4
-; GFX942-NEXT: v_or_b32_sdwa v2, v2, v3 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_0 src1_sel:DWORD
-; GFX942-NEXT: v_lshlrev_b16_e32 v3, 8, v6
-; GFX942-NEXT: v_or_b32_sdwa v3, v1, v3 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; GFX942-NEXT: v_or_b32_sdwa v0, v0, v1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_0 src1_sel:DWORD
+; GFX942-NEXT: v_lshlrev_b16_e32 v1, 8, v6
+; GFX942-NEXT: v_or_b32_sdwa v1, v3, v1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
; GFX942-NEXT: v_or_b32_sdwa v11, v5, v11 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
; GFX942-NEXT: v_mov_b32_e32 v10, 0
-; GFX942-NEXT: v_or_b32_sdwa v3, v3, v11 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_0 src1_sel:DWORD
-; GFX942-NEXT: global_store_dwordx2 v10, v[2:3], s[12:13]
-; GFX942-NEXT: v_mov_b32_e32 v2, v0
+; GFX942-NEXT: v_or_b32_sdwa v1, v1, v11 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_0 src1_sel:DWORD
+; GFX942-NEXT: global_store_dwordx2 v10, v[0:1], s[12:13]
+; GFX942-NEXT: v_mov_b32_e32 v0, v2
; GFX942-NEXT: v_mov_b32_e32 v12, v9
; GFX942-NEXT: v_mov_b32_e32 v10, v8
; GFX942-NEXT: v_mov_b32_e32 v13, v7
-; GFX942-NEXT: v_mov_b32_e32 v3, v1
+; GFX942-NEXT: v_mov_b32_e32 v1, v3
; GFX942-NEXT: v_mov_b32_e32 v14, v6
; GFX942-NEXT: v_mov_b32_e32 v11, v5
; GFX942-NEXT: v_mov_b32_e32 v15, v4
; GFX942-NEXT: .LBB10_4: ; %bb.3
; GFX942-NEXT: s_or_b64 exec, exec, s[2:3]
-; GFX942-NEXT: v_lshlrev_b16_e32 v0, 8, v12
-; GFX942-NEXT: v_lshlrev_b16_e32 v1, 8, v13
-; GFX942-NEXT: v_or_b32_sdwa v0, v2, v0 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
-; GFX942-NEXT: v_or_b32_sdwa v1, v10, v1 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
-; GFX942-NEXT: v_lshlrev_b16_e32 v2, 8, v15
-; GFX942-NEXT: v_or_b32_sdwa v0, v0, v1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_0 src1_sel:DWORD
-; GFX942-NEXT: v_lshlrev_b16_e32 v1, 8, v14
-; GFX942-NEXT: v_or_b32_sdwa v1, v3, v1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
-; GFX942-NEXT: v_or_b32_sdwa v2, v11, v2 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
-; GFX942-NEXT: v_mov_b32_e32 v4, 0
-; GFX942-NEXT: v_or_b32_sdwa v1, v1, v2 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_0 src1_sel:DWORD
-; GFX942-NEXT: global_store_dwordx2 v4, v[0:1], s[14:15]
+; GFX942-NEXT: v_lshlrev_b16_e32 v3, 8, v12
+; GFX942-NEXT: v_or_b32_sdwa v0, v0, v3 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; GFX942-NEXT: v_lshlrev_b16_e32 v3, 8, v13
+; GFX942-NEXT: v_or_b32_sdwa v3, v10, v3 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; GFX942-NEXT: v_mov_b32_e32 v2, 0
+; GFX942-NEXT: v_or_b32_sdwa v0, v0, v3 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_0 src1_sel:DWORD
+; GFX942-NEXT: v_lshlrev_b16_e32 v3, 8, v14
+; GFX942-NEXT: v_or_b32_sdwa v1, v1, v3 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; GFX942-NEXT: v_lshlrev_b16_e32 v3, 8, v15
+; GFX942-NEXT: v_or_b32_sdwa v3, v11, v3 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; GFX942-NEXT: s_nop 0
+; GFX942-NEXT: v_or_b32_sdwa v1, v1, v3 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_0 src1_sel:DWORD
+; GFX942-NEXT: global_store_dwordx2 v2, v[0:1], s[14:15]
; GFX942-NEXT: s_endpgm
entry:
%idx = call i32 @llvm.amdgcn.workitem.id.x()
diff --git a/llvm/test/CodeGen/AMDGPU/wave32.ll b/llvm/test/CodeGen/AMDGPU/wave32.ll
index 4212fd3b35cd8..b7c12854d1115 100644
--- a/llvm/test/CodeGen/AMDGPU/wave32.ll
+++ b/llvm/test/CodeGen/AMDGPU/wave32.ll
@@ -1292,23 +1292,23 @@ define amdgpu_kernel void @test_div_fmas_f32_i1_phi_vcc(ptr addrspace(1) %out, p
; GFX1064: ; %bb.0: ; %entry
; GFX1064-NEXT: s_clause 0x1
; GFX1064-NEXT: s_load_dwordx4 s[8:11], s[4:5], 0x24
-; GFX1064-NEXT: s_load_dwordx2 s[6:7], s[4:5], 0x34
+; GFX1064-NEXT: s_load_dwordx2 s[2:3], s[4:5], 0x34
; GFX1064-NEXT: v_lshlrev_b32_e32 v1, 2, v0
; GFX1064-NEXT: v_cmp_eq_u32_e64 s[0:1], 0, v0
; GFX1064-NEXT: s_mov_b64 vcc, 0
; GFX1064-NEXT: s_waitcnt lgkmcnt(0)
; GFX1064-NEXT: global_load_dwordx3 v[1:3], v1, s[10:11]
-; GFX1064-NEXT: s_and_saveexec_b64 s[2:3], s[0:1]
+; GFX1064-NEXT: s_and_saveexec_b64 s[4:5], s[0:1]
; GFX1064-NEXT: s_cbranch_execz .LBB22_2
; GFX1064-NEXT: ; %bb.1: ; %bb
; GFX1064-NEXT: v_mov_b32_e32 v0, 0
-; GFX1064-NEXT: global_load_dword v0, v0, s[6:7] glc dlc
+; GFX1064-NEXT: global_load_dword v0, v0, s[2:3] glc dlc
; GFX1064-NEXT: s_waitcnt vmcnt(0)
; GFX1064-NEXT: v_cmp_ne_u32_e32 vcc, 0, v0
; GFX1064-NEXT: s_and_b64 vcc, vcc, exec
; GFX1064-NEXT: .LBB22_2: ; %exit
; GFX1064-NEXT: s_waitcnt_depctr 0xffe3
-; GFX1064-NEXT: s_or_b64 exec, exec, s[2:3]
+; GFX1064-NEXT: s_or_b64 exec, exec, s[4:5]
; GFX1064-NEXT: v_mov_b32_e32 v0, 0
; GFX1064-NEXT: s_waitcnt vmcnt(0)
; GFX1064-NEXT: v_div_fmas_f32 v1, v1, v2, v3
diff --git a/llvm/test/CodeGen/ARM/and-cmp0-sink.ll b/llvm/test/CodeGen/ARM/and-cmp0-sink.ll
index fb9139c0d1285..389193f9d2407 100644
--- a/llvm/test/CodeGen/ARM/and-cmp0-sink.ll
+++ b/llvm/test/CodeGen/ARM/and-cmp0-sink.ll
@@ -112,20 +112,20 @@ define void @f(i32 %v, ptr noalias %outp) {
; V6M-NEXT: push {r4, lr}
; V6M-NEXT: movs r2, #0
; V6M-NEXT: str r2, [r1]
-; V6M-NEXT: movs r3, #14
-; V6M-NEXT: ands r3, r0
+; V6M-NEXT: movs r2, #14
+; V6M-NEXT: ands r2, r0
; V6M-NEXT: movs r4, #4
; V6M-NEXT: ands r4, r0
-; V6M-NEXT: movs r2, #2
-; V6M-NEXT: ands r2, r0
+; V6M-NEXT: movs r3, #2
+; V6M-NEXT: ands r3, r0
; V6M-NEXT: lsls r0, r0, #31
; V6M-NEXT: bne .LBB0_5
; V6M-NEXT: @ %bb.1: @ %if.then
; V6M-NEXT: movs r0, #129
-; V6M-NEXT: cmp r2, #0
+; V6M-NEXT: cmp r3, #0
; V6M-NEXT: beq .LBB0_3
; V6M-NEXT: @ %bb.2:
-; V6M-NEXT: lsls r2, r0, #8
+; V6M-NEXT: lsls r3, r0, #8
; V6M-NEXT: .LBB0_3: @ %if.then
; V6M-NEXT: cmp r4, #0
; V6M-NEXT: beq .LBB0_10
@@ -134,22 +134,22 @@ define void @f(i32 %v, ptr noalias %outp) {
; V6M-NEXT: b .LBB0_9
; V6M-NEXT: .LBB0_5: @ %if.else
; V6M-NEXT: movs r0, #129
-; V6M-NEXT: cmp r2, #0
+; V6M-NEXT: cmp r3, #0
; V6M-NEXT: beq .LBB0_7
; V6M-NEXT: @ %bb.6:
-; V6M-NEXT: lsls r2, r0, #6
+; V6M-NEXT: lsls r3, r0, #6
; V6M-NEXT: .LBB0_7: @ %if.else
; V6M-NEXT: cmp r4, #0
; V6M-NEXT: beq .LBB0_10
; V6M-NEXT: @ %bb.8: @ %if.else
; V6M-NEXT: lsls r0, r0, #5
; V6M-NEXT: .LBB0_9: @ %if.else
-; V6M-NEXT: adds r2, r2, r0
+; V6M-NEXT: adds r3, r3, r0
; V6M-NEXT: .LBB0_10: @ %if.else
-; V6M-NEXT: cmp r3, #0
+; V6M-NEXT: cmp r2, #0
; V6M-NEXT: beq .LBB0_12
; V6M-NEXT: @ %bb.11: @ %if.end
-; V6M-NEXT: str r2, [r1]
+; V6M-NEXT: str r3, [r1]
; V6M-NEXT: .LBB0_12: @ %exit
; V6M-NEXT: pop {r4, pc}
entry:
diff --git a/llvm/test/CodeGen/ARM/cttz.ll b/llvm/test/CodeGen/ARM/cttz.ll
index 1146ad64ee709..e1bf4837d6a47 100644
--- a/llvm/test/CodeGen/ARM/cttz.ll
+++ b/llvm/test/CodeGen/ARM/cttz.ll
@@ -229,24 +229,24 @@ define i64 @test_i64(i64 %a) {
; CHECK-6M-NEXT: orrs r0, r3
; CHECK-6M-NEXT: beq .LBB3_6
; CHECK-6M-NEXT: @ %bb.1: @ %cond.false
-; CHECK-6M-NEXT: ldr r6, .LCPI3_0
+; CHECK-6M-NEXT: ldr r5, .LCPI3_0
; CHECK-6M-NEXT: adr r4, .LCPI3_1
; CHECK-6M-NEXT: movs r0, #32
; CHECK-6M-NEXT: cmp r3, #0
-; CHECK-6M-NEXT: mov r5, r0
+; CHECK-6M-NEXT: mov r6, r0
; CHECK-6M-NEXT: beq .LBB3_3
; CHECK-6M-NEXT: @ %bb.2: @ %cond.false
-; CHECK-6M-NEXT: rsbs r5, r3, #0
-; CHECK-6M-NEXT: ands r5, r3
-; CHECK-6M-NEXT: muls r5, r6, r5
-; CHECK-6M-NEXT: lsrs r3, r5, #27
-; CHECK-6M-NEXT: ldrb r5, [r4, r3]
+; CHECK-6M-NEXT: rsbs r6, r3, #0
+; CHECK-6M-NEXT: ands r6, r3
+; CHECK-6M-NEXT: muls r6, r5, r6
+; CHECK-6M-NEXT: lsrs r3, r6, #27
+; CHECK-6M-NEXT: ldrb r6, [r4, r3]
; CHECK-6M-NEXT: .LBB3_3: @ %cond.false
-; CHECK-6M-NEXT: adds r5, #32
+; CHECK-6M-NEXT: adds r6, #32
; CHECK-6M-NEXT: rsbs r3, r2, #0
; CHECK-6M-NEXT: ands r3, r2
-; CHECK-6M-NEXT: muls r6, r3, r6
-; CHECK-6M-NEXT: lsrs r3, r6, #27
+; CHECK-6M-NEXT: muls r5, r3, r5
+; CHECK-6M-NEXT: lsrs r3, r5, #27
; CHECK-6M-NEXT: cmp r2, #0
; CHECK-6M-NEXT: bne .LBB3_7
; CHECK-6M-NEXT: @ %bb.4: @ %cond.false
@@ -260,7 +260,7 @@ define i64 @test_i64(i64 %a) {
; CHECK-6M-NEXT: ldrb r0, [r4, r3]
; CHECK-6M-NEXT: bne .LBB3_5
; CHECK-6M-NEXT: .LBB3_8: @ %cond.false
-; CHECK-6M-NEXT: mov r0, r5
+; CHECK-6M-NEXT: mov r0, r6
; CHECK-6M-NEXT: pop {r4, r5, r6, pc}
; CHECK-6M-NEXT: .p2align 2
; CHECK-6M-NEXT: @ %bb.9:
@@ -279,24 +279,24 @@ define i64 @test_i64(i64 %a) {
; CHECK-8MBASE-NEXT: orrs r0, r3
; CHECK-8MBASE-NEXT: beq .LBB3_6
; CHECK-8MBASE-NEXT: @ %bb.1: @ %cond.false
-; CHECK-8MBASE-NEXT: movw r6, #46385
-; CHECK-8MBASE-NEXT: movt r6, #1916
+; CHECK-8MBASE-NEXT: movw r5, #46385
+; CHECK-8MBASE-NEXT: movt r5, #1916
; CHECK-8MBASE-NEXT: adr r4, .LCPI3_0
; CHECK-8MBASE-NEXT: movs r0, #32
-; CHECK-8MBASE-NEXT: mov r5, r0
+; CHECK-8MBASE-NEXT: mov r6, r0
; CHECK-8MBASE-NEXT: cbz r3, .LBB3_3
; CHECK-8MBASE-NEXT: @ %bb.2: @ %cond.false
-; CHECK-8MBASE-NEXT: rsbs r5, r3, #0
-; CHECK-8MBASE-NEXT: ands r5, r3
-; CHECK-8MBASE-NEXT: muls r5, r6, r5
-; CHECK-8MBASE-NEXT: lsrs r3, r5, #27
-; CHECK-8MBASE-NEXT: ldrb r5, [r4, r3]
+; CHECK-8MBASE-NEXT: rsbs r6, r3, #0
+; CHECK-8MBASE-NEXT: ands r6, r3
+; CHECK-8MBASE-NEXT: muls r6, r5, r6
+; CHECK-8MBASE-NEXT: lsrs r3, r6, #27
+; CHECK-8MBASE-NEXT: ldrb r6, [r4, r3]
; CHECK-8MBASE-NEXT: .LBB3_3: @ %cond.false
-; CHECK-8MBASE-NEXT: adds r5, #32
+; CHECK-8MBASE-NEXT: adds r6, #32
; CHECK-8MBASE-NEXT: rsbs r3, r2, #0
; CHECK-8MBASE-NEXT: ands r3, r2
-; CHECK-8MBASE-NEXT: muls r6, r3, r6
-; CHECK-8MBASE-NEXT: lsrs r3, r6, #27
+; CHECK-8MBASE-NEXT: muls r5, r3, r5
+; CHECK-8MBASE-NEXT: lsrs r3, r5, #27
; CHECK-8MBASE-NEXT: cmp r2, #0
; CHECK-8MBASE-NEXT: bne .LBB3_7
; CHECK-8MBASE-NEXT: @ %bb.4: @ %cond.false
@@ -310,7 +310,7 @@ define i64 @test_i64(i64 %a) {
; CHECK-8MBASE-NEXT: ldrb r0, [r4, r3]
; CHECK-8MBASE-NEXT: bne .LBB3_5
; CHECK-8MBASE-NEXT: .LBB3_8: @ %cond.false
-; CHECK-8MBASE-NEXT: mov r0, r5
+; CHECK-8MBASE-NEXT: mov r0, r6
; CHECK-8MBASE-NEXT: pop {r4, r5, r6, pc}
; CHECK-8MBASE-NEXT: .p2align 2
; CHECK-8MBASE-NEXT: @ %bb.9:
@@ -503,24 +503,24 @@ define i64 @test_i64_zero_undef(i64 %a) {
; CHECK-6M-NEXT: .save {r4, r5, r7, lr}
; CHECK-6M-NEXT: push {r4, r5, r7, lr}
; CHECK-6M-NEXT: mov r2, r0
-; CHECK-6M-NEXT: ldr r5, .LCPI7_0
+; CHECK-6M-NEXT: ldr r4, .LCPI7_0
; CHECK-6M-NEXT: adr r3, .LCPI7_1
; CHECK-6M-NEXT: movs r0, #32
; CHECK-6M-NEXT: cmp r1, #0
-; CHECK-6M-NEXT: mov r4, r0
+; CHECK-6M-NEXT: mov r5, r0
; CHECK-6M-NEXT: beq .LBB7_2
; CHECK-6M-NEXT: @ %bb.1:
-; CHECK-6M-NEXT: rsbs r4, r1, #0
-; CHECK-6M-NEXT: ands r4, r1
-; CHECK-6M-NEXT: muls r4, r5, r4
-; CHECK-6M-NEXT: lsrs r1, r4, #27
-; CHECK-6M-NEXT: ldrb r4, [r3, r1]
+; CHECK-6M-NEXT: rsbs r5, r1, #0
+; CHECK-6M-NEXT: ands r5, r1
+; CHECK-6M-NEXT: muls r5, r4, r5
+; CHECK-6M-NEXT: lsrs r1, r5, #27
+; CHECK-6M-NEXT: ldrb r5, [r3, r1]
; CHECK-6M-NEXT: .LBB7_2:
-; CHECK-6M-NEXT: adds r4, #32
+; CHECK-6M-NEXT: adds r5, #32
; CHECK-6M-NEXT: rsbs r1, r2, #0
; CHECK-6M-NEXT: ands r1, r2
-; CHECK-6M-NEXT: muls r5, r1, r5
-; CHECK-6M-NEXT: lsrs r1, r5, #27
+; CHECK-6M-NEXT: muls r4, r1, r4
+; CHECK-6M-NEXT: lsrs r1, r4, #27
; CHECK-6M-NEXT: cmp r2, #0
; CHECK-6M-NEXT: bne .LBB7_5
; CHECK-6M-NEXT: @ %bb.3:
@@ -532,7 +532,7 @@ define i64 @test_i64_zero_undef(i64 %a) {
; CHECK-6M-NEXT: ldrb r0, [r3, r1]
; CHECK-6M-NEXT: bne .LBB7_4
; CHECK-6M-NEXT: .LBB7_6:
-; CHECK-6M-NEXT: mov r0, r4
+; CHECK-6M-NEXT: mov r0, r5
; CHECK-6M-NEXT: movs r1, #0
; CHECK-6M-NEXT: pop {r4, r5, r7, pc}
; CHECK-6M-NEXT: .p2align 2
@@ -547,24 +547,24 @@ define i64 @test_i64_zero_undef(i64 %a) {
; CHECK-8MBASE-NEXT: .save {r4, r5, r7, lr}
; CHECK-8MBASE-NEXT: push {r4, r5, r7, lr}
; CHECK-8MBASE-NEXT: mov r2, r0
-; CHECK-8MBASE-NEXT: movw r5, #46385
-; CHECK-8MBASE-NEXT: movt r5, #1916
+; CHECK-8MBASE-NEXT: movw r4, #46385
+; CHECK-8MBASE-NEXT: movt r4, #1916
; CHECK-8MBASE-NEXT: adr r3, .LCPI7_0
; CHECK-8MBASE-NEXT: movs r0, #32
-; CHECK-8MBASE-NEXT: mov r4, r0
+; CHECK-8MBASE-NEXT: mov r5, r0
; CHECK-8MBASE-NEXT: cbz r1, .LBB7_2
; CHECK-8MBASE-NEXT: @ %bb.1:
-; CHECK-8MBASE-NEXT: rsbs r4, r1, #0
-; CHECK-8MBASE-NEXT: ands r4, r1
-; CHECK-8MBASE-NEXT: muls r4, r5, r4
-; CHECK-8MBASE-NEXT: lsrs r1, r4, #27
-; CHECK-8MBASE-NEXT: ldrb r4, [r3, r1]
+; CHECK-8MBASE-NEXT: rsbs r5, r1, #0
+; CHECK-8MBASE-NEXT: ands r5, r1
+; CHECK-8MBASE-NEXT: muls r5, r4, r5
+; CHECK-8MBASE-NEXT: lsrs r1, r5, #27
+; CHECK-8MBASE-NEXT: ldrb r5, [r3, r1]
; CHECK-8MBASE-NEXT: .LBB7_2:
-; CHECK-8MBASE-NEXT: adds r4, #32
+; CHECK-8MBASE-NEXT: adds r5, #32
; CHECK-8MBASE-NEXT: rsbs r1, r2, #0
; CHECK-8MBASE-NEXT: ands r1, r2
-; CHECK-8MBASE-NEXT: muls r5, r1, r5
-; CHECK-8MBASE-NEXT: lsrs r1, r5, #27
+; CHECK-8MBASE-NEXT: muls r4, r1, r4
+; CHECK-8MBASE-NEXT: lsrs r1, r4, #27
; CHECK-8MBASE-NEXT: cmp r2, #0
; CHECK-8MBASE-NEXT: bne .LBB7_5
; CHECK-8MBASE-NEXT: @ %bb.3:
@@ -576,7 +576,7 @@ define i64 @test_i64_zero_undef(i64 %a) {
; CHECK-8MBASE-NEXT: ldrb r0, [r3, r1]
; CHECK-8MBASE-NEXT: bne .LBB7_4
; CHECK-8MBASE-NEXT: .LBB7_6:
-; CHECK-8MBASE-NEXT: mov r0, r4
+; CHECK-8MBASE-NEXT: mov r0, r5
; CHECK-8MBASE-NEXT: movs r1, #0
; CHECK-8MBASE-NEXT: pop {r4, r5, r7, pc}
; CHECK-8MBASE-NEXT: .p2align 2
diff --git a/llvm/test/CodeGen/ARM/select-imm.ll b/llvm/test/CodeGen/ARM/select-imm.ll
index 186276b50ceeb..2bef1c83d7969 100644
--- a/llvm/test/CodeGen/ARM/select-imm.ll
+++ b/llvm/test/CodeGen/ARM/select-imm.ll
@@ -455,13 +455,13 @@ define void @t9(ptr %a, i8 %b) {
; ARMT2-NEXT: cmp r0, r0
; ARMT2-NEXT: popne {r4, pc}
; ARMT2-NEXT: .LBB8_1: @ %while.body.preheader
-; ARMT2-NEXT: add r1, r4, #1
-; ARMT2-NEXT: mov r2, r0
+; ARMT2-NEXT: mov r1, r0
+; ARMT2-NEXT: add r2, r4, #1
; ARMT2-NEXT: .LBB8_2: @ %while.body
; ARMT2-NEXT: @ =>This Inner Loop Header: Depth=1
-; ARMT2-NEXT: add r2, r2, #1
; ARMT2-NEXT: add r1, r1, #1
-; ARMT2-NEXT: uxtb r3, r2
+; ARMT2-NEXT: add r2, r2, #1
+; ARMT2-NEXT: uxtb r3, r1
; ARMT2-NEXT: cmp r3, r0
; ARMT2-NEXT: blt .LBB8_2
; ARMT2-NEXT: @ %bb.3: @ %while.end
@@ -503,13 +503,13 @@ define void @t9(ptr %a, i8 %b) {
; THUMB2-NEXT: it ne
; THUMB2-NEXT: popne {r4, pc}
; THUMB2-NEXT: .LBB8_1: @ %while.body.preheader
-; THUMB2-NEXT: adds r1, r4, #1
-; THUMB2-NEXT: mov r2, r0
+; THUMB2-NEXT: mov r1, r0
+; THUMB2-NEXT: adds r2, r4, #1
; THUMB2-NEXT: .LBB8_2: @ %while.body
; THUMB2-NEXT: @ =>This Inner Loop Header: Depth=1
-; THUMB2-NEXT: adds r2, #1
; THUMB2-NEXT: adds r1, #1
-; THUMB2-NEXT: uxtb r3, r2
+; THUMB2-NEXT: adds r2, #1
+; THUMB2-NEXT: uxtb r3, r1
; THUMB2-NEXT: cmp r3, r0
; THUMB2-NEXT: blt .LBB8_2
; THUMB2-NEXT: @ %bb.3: @ %while.end
diff --git a/llvm/test/CodeGen/ARM/struct-byval-loop.ll b/llvm/test/CodeGen/ARM/struct-byval-loop.ll
index 7a38dec2434f7..a90381acf4214 100644
--- a/llvm/test/CodeGen/ARM/struct-byval-loop.ll
+++ b/llvm/test/CodeGen/ARM/struct-byval-loop.ll
@@ -13,13 +13,13 @@ define void @test_80() {
; CHECK-NEXT: .pad #152
; CHECK-NEXT: sub sp, sp, #152
; CHECK-NEXT: add r0, sp, #72
-; CHECK-NEXT: ldr r1, .LCPI0_0
+; CHECK-NEXT: ldr r2, .LCPI0_0
; CHECK-NEXT: add r0, r0, #12
-; CHECK-NEXT: mov r2, sp
+; CHECK-NEXT: mov r1, sp
; CHECK-NEXT: .LBB0_1: @ =>This Inner Loop Header: Depth=1
; CHECK-NEXT: ldr r3, [r0], #4
-; CHECK-NEXT: subs r1, r1, #4
-; CHECK-NEXT: str r3, [r2], #4
+; CHECK-NEXT: subs r2, r2, #4
+; CHECK-NEXT: str r3, [r1], #4
; CHECK-NEXT: bne .LBB0_1
; CHECK-NEXT: @ %bb.2:
; CHECK-NEXT: add r3, sp, #72
@@ -52,14 +52,14 @@ define void @test_4000() {
; CHECK-NEXT: sub sp, sp, #920
; CHECK-NEXT: sub sp, sp, #3072
; CHECK-NEXT: add lr, sp, #3072
-; CHECK-NEXT: ldr r1, .LCPI1_0
+; CHECK-NEXT: ldr r2, .LCPI1_0
; CHECK-NEXT: add r0, lr, #920
-; CHECK-NEXT: mov r2, sp
+; CHECK-NEXT: mov r1, sp
; CHECK-NEXT: add r0, r0, #12
; CHECK-NEXT: .LBB1_1: @ =>This Inner Loop Header: Depth=1
; CHECK-NEXT: ldr r3, [r0], #4
-; CHECK-NEXT: subs r1, r1, #4
-; CHECK-NEXT: str r3, [r2], #4
+; CHECK-NEXT: subs r2, r2, #4
+; CHECK-NEXT: str r3, [r1], #4
; CHECK-NEXT: bne .LBB1_1
; CHECK-NEXT: @ %bb.2:
; CHECK-NEXT: ldr r1, [sp, #3992]
diff --git a/llvm/test/CodeGen/ARM/swifterror.ll b/llvm/test/CodeGen/ARM/swifterror.ll
index f002c54fc60c0..259c20c8c9af6 100644
--- a/llvm/test/CodeGen/ARM/swifterror.ll
+++ b/llvm/test/CodeGen/ARM/swifterror.ll
@@ -79,17 +79,17 @@ define float @caller(ptr %error_ref) {
;
; CHECK-O0-LABEL: caller:
; CHECK-O0: @ %bb.0: @ %entry
-; CHECK-O0-NEXT: push {r7, r8, lr}
-; CHECK-O0-NEXT: add r7, sp, #4
-; CHECK-O0-NEXT: sub sp, sp, #12
+; CHECK-O0-NEXT: push {r7, r8, lr}
+; CHECK-O0-NEXT: add r7, sp, #4
+; CHECK-O0-NEXT: sub sp, sp, #12
; CHECK-O0-NEXT: @ implicit-def: $r1
-; CHECK-O0-NEXT: str r0, [sp] @ 4-byte Spill
-; CHECK-O0-NEXT: mov r8, #0
-; CHECK-O0-NEXT: bl _foo
-; CHECK-O0-NEXT: str r8, [sp, #4] @ 4-byte Spill
-; CHECK-O0-NEXT: movw r0, #0
-; CHECK-O0-NEXT: cmp r8, r0
-; CHECK-O0-NEXT: bne LBB1_2
+; CHECK-O0-NEXT: str r0, [sp] @ 4-byte Spill
+; CHECK-O0-NEXT: mov r8, #0
+; CHECK-O0-NEXT: bl _foo
+; CHECK-O0-NEXT: str r8, [sp, #4] @ 4-byte Spill
+; CHECK-O0-NEXT: movw r0, #0
+; CHECK-O0-NEXT: cmp r8, r0
+; CHECK-O0-NEXT: bne LBB1_2
; CHECK-O0-NEXT: @ %bb.1: @ %cont
; CHECK-O0-NEXT: ldr r1, [sp] @ 4-byte Reload
; CHECK-O0-NEXT: ldr r0, [sp, #4] @ 4-byte Reload
@@ -100,7 +100,7 @@ define float @caller(ptr %error_ref) {
; CHECK-O0-NEXT: bl _free
; CHECK-O0-NEXT: mov r0, #1065353216
; CHECK-O0-NEXT: sub sp, r7, #4
-; CHECK-O0-NEXT: pop {r7, r8, pc}
+; CHECK-O0-NEXT: pop {r7, r8, pc}
;
; CHECK-ANDROID-LABEL: caller:
; CHECK-ANDROID: @ %bb.0: @ %entry
@@ -174,11 +174,11 @@ define float @caller2(ptr %error_ref) {
;
; CHECK-O0-LABEL: caller2:
; CHECK-O0: @ %bb.0: @ %entry
-; CHECK-O0-NEXT: push {r7, r8, lr}
-; CHECK-O0-NEXT: add r7, sp, #4
-; CHECK-O0-NEXT: sub sp, sp, #16
+; CHECK-O0-NEXT: push {r7, r8, lr}
+; CHECK-O0-NEXT: add r7, sp, #4
+; CHECK-O0-NEXT: sub sp, sp, #16
; CHECK-O0-NEXT: @ implicit-def: $r1
-; CHECK-O0-NEXT: str r0, [sp, #8] @ 4-byte Spill
+; CHECK-O0-NEXT: str r0, [sp, #8] @ 4-byte Spill
; CHECK-O0-NEXT: LBB2_1: @ %bb_loop
; CHECK-O0-NEXT: @ =>This Inner Loop Header: Depth=1
; CHECK-O0-NEXT: mov r8, #0
@@ -206,7 +206,7 @@ define float @caller2(ptr %error_ref) {
; CHECK-O0-NEXT: bl _free
; CHECK-O0-NEXT: mov r0, #1065353216
; CHECK-O0-NEXT: sub sp, r7, #4
-; CHECK-O0-NEXT: pop {r7, r8, pc}
+; CHECK-O0-NEXT: pop {r7, r8, pc}
;
; CHECK-ANDROID-LABEL: caller2:
; CHECK-ANDROID: @ %bb.0: @ %entry
@@ -400,35 +400,35 @@ define float @foo_loop(ptr swifterror %error_ptr_ref, i32 %cc, float %cc2) {
; CHECK-O0-NEXT: mov r7, sp
; CHECK-O0-NEXT: sub sp, sp, #20
; CHECK-O0-NEXT: str r0, [sp, #8] @ 4-byte Spill
+; CHECK-O0-NEXT: str r8, [r7, #-8] @ 4-byte Spill
; CHECK-O0-NEXT: vmov s0, r1
-; CHECK-O0-NEXT: vstr s0, [r7, #-8] @ 4-byte Spill
-; CHECK-O0-NEXT: str r8, [r7, #-4] @ 4-byte Spill
+; CHECK-O0-NEXT: vstr s0, [r7, #-4] @ 4-byte Spill
; CHECK-O0-NEXT: b LBB4_1
; CHECK-O0-NEXT: LBB4_1: @ %bb_loop
; CHECK-O0-NEXT: @ =>This Inner Loop Header: Depth=1
-; CHECK-O0-NEXT: ldr r1, [sp, #8] @ 4-byte Reload
-; CHECK-O0-NEXT: ldr r0, [r7, #-4] @ 4-byte Reload
-; CHECK-O0-NEXT: cmp r1, #0
-; CHECK-O0-NEXT: str r0, [sp, #4] @ 4-byte Spill
+; CHECK-O0-NEXT: ldr r0, [sp, #8] @ 4-byte Reload
+; CHECK-O0-NEXT: ldr r1, [r7, #-8] @ 4-byte Reload
+; CHECK-O0-NEXT: str r1, [sp, #4] @ 4-byte Spill
+; CHECK-O0-NEXT: cmp r0, #0
; CHECK-O0-NEXT: beq LBB4_3
; CHECK-O0-NEXT: @ %bb.2: @ %gen_error
; CHECK-O0-NEXT: @ in Loop: Header=BB4_1 Depth=1
; CHECK-O0-NEXT: mov r0, #16
; CHECK-O0-NEXT: mov r1, #0
; CHECK-O0-NEXT: bl _malloc
-; CHECK-O0-NEXT: mov r2, r0
-; CHECK-O0-NEXT: movw r1, #1
-; CHECK-O0-NEXT: strb r1, [r2, #8]
+; CHECK-O0-NEXT: mov r1, r0
; CHECK-O0-NEXT: str r0, [sp, #4] @ 4-byte Spill
+; CHECK-O0-NEXT: movw r0, #1
+; CHECK-O0-NEXT: strb r0, [r1, #8]
; CHECK-O0-NEXT: LBB4_3: @ %bb_cont
; CHECK-O0-NEXT: @ in Loop: Header=BB4_1 Depth=1
-; CHECK-O0-NEXT: vldr s0, [r7, #-8] @ 4-byte Reload
+; CHECK-O0-NEXT: vldr s0, [r7, #-4] @ 4-byte Reload
; CHECK-O0-NEXT: ldr r0, [sp, #4] @ 4-byte Reload
; CHECK-O0-NEXT: str r0, [sp] @ 4-byte Spill
; CHECK-O0-NEXT: vmov.f32 s2, #1.000000e+00
; CHECK-O0-NEXT: vcmp.f32 s0, s2
; CHECK-O0-NEXT: vmrs APSR_nzcv, fpscr
-; CHECK-O0-NEXT: str r0, [r7, #-4] @ 4-byte Spill
+; CHECK-O0-NEXT: str r0, [r7, #-8] @ 4-byte Spill
; CHECK-O0-NEXT: ble LBB4_1
; CHECK-O0-NEXT: @ %bb.4: @ %bb_end
; CHECK-O0-NEXT: ldr r8, [sp] @ 4-byte Reload
@@ -581,20 +581,20 @@ define float @caller3(ptr %error_ref) {
;
; CHECK-O0-LABEL: caller3:
; CHECK-O0: @ %bb.0: @ %entry
-; CHECK-O0-NEXT: push {r7, r8, lr}
-; CHECK-O0-NEXT: add r7, sp, #4
-; CHECK-O0-NEXT: sub sp, sp, #44
-; CHECK-O0-NEXT: bfc sp, #0, #3
+; CHECK-O0-NEXT: push {r7, r8, lr}
+; CHECK-O0-NEXT: add r7, sp, #4
+; CHECK-O0-NEXT: sub sp, sp, #44
+; CHECK-O0-NEXT: bfc sp, #0, #3
; CHECK-O0-NEXT: @ implicit-def: $r1
-; CHECK-O0-NEXT: str r0, [sp, #4] @ 4-byte Spill
-; CHECK-O0-NEXT: mov r8, #0
-; CHECK-O0-NEXT: add r0, sp, #16
-; CHECK-O0-NEXT: mov r1, #1
-; CHECK-O0-NEXT: bl _foo_sret
-; CHECK-O0-NEXT: str r8, [sp, #8] @ 4-byte Spill
-; CHECK-O0-NEXT: movw r0, #0
-; CHECK-O0-NEXT: cmp r8, r0
-; CHECK-O0-NEXT: bne LBB6_2
+; CHECK-O0-NEXT: str r0, [sp, #4] @ 4-byte Spill
+; CHECK-O0-NEXT: mov r8, #0
+; CHECK-O0-NEXT: add r0, sp, #16
+; CHECK-O0-NEXT: mov r1, #1
+; CHECK-O0-NEXT: bl _foo_sret
+; CHECK-O0-NEXT: str r8, [sp, #8] @ 4-byte Spill
+; CHECK-O0-NEXT: movw r0, #0
+; CHECK-O0-NEXT: cmp r8, r0
+; CHECK-O0-NEXT: bne LBB6_2
; CHECK-O0-NEXT: @ %bb.1: @ %cont
; CHECK-O0-NEXT: ldr r1, [sp, #4] @ 4-byte Reload
; CHECK-O0-NEXT: ldr r0, [sp, #8] @ 4-byte Reload
@@ -605,7 +605,7 @@ define float @caller3(ptr %error_ref) {
; CHECK-O0-NEXT: bl _free
; CHECK-O0-NEXT: mov r0, #1065353216
; CHECK-O0-NEXT: sub sp, r7, #4
-; CHECK-O0-NEXT: pop {r7, r8, pc}
+; CHECK-O0-NEXT: pop {r7, r8, pc}
;
; CHECK-ANDROID-LABEL: caller3:
; CHECK-ANDROID: @ %bb.0: @ %entry
@@ -803,26 +803,26 @@ define float @caller4(ptr %error_ref) {
;
; CHECK-O0-LABEL: caller4:
; CHECK-O0: @ %bb.0: @ %entry
-; CHECK-O0-NEXT: push {r7, r8, lr}
-; CHECK-O0-NEXT: add r7, sp, #4
-; CHECK-O0-NEXT: sub sp, sp, #24
+; CHECK-O0-NEXT: push {r7, r8, lr}
+; CHECK-O0-NEXT: add r7, sp, #4
+; CHECK-O0-NEXT: sub sp, sp, #24
; CHECK-O0-NEXT: @ implicit-def: $r1
-; CHECK-O0-NEXT: str r0, [sp] @ 4-byte Spill
-; CHECK-O0-NEXT: mov r8, #0
-; CHECK-O0-NEXT: mov r0, #10
-; CHECK-O0-NEXT: str r0, [r7, #-12]
-; CHECK-O0-NEXT: mov r0, #11
-; CHECK-O0-NEXT: str r0, [sp, #12]
-; CHECK-O0-NEXT: mov r0, #12
-; CHECK-O0-NEXT: str r0, [sp, #8]
-; CHECK-O0-NEXT: ldr r0, [r7, #-12]
-; CHECK-O0-NEXT: ldr r1, [sp, #12]
-; CHECK-O0-NEXT: ldr r2, [sp, #8]
-; CHECK-O0-NEXT: bl _foo_vararg
-; CHECK-O0-NEXT: str r8, [sp, #4] @ 4-byte Spill
-; CHECK-O0-NEXT: movw r0, #0
-; CHECK-O0-NEXT: cmp r8, r0
-; CHECK-O0-NEXT: bne LBB8_2
+; CHECK-O0-NEXT: str r0, [sp] @ 4-byte Spill
+; CHECK-O0-NEXT: mov r8, #0
+; CHECK-O0-NEXT: mov r0, #10
+; CHECK-O0-NEXT: str r0, [r7, #-12]
+; CHECK-O0-NEXT: mov r0, #11
+; CHECK-O0-NEXT: str r0, [sp, #12]
+; CHECK-O0-NEXT: mov r0, #12
+; CHECK-O0-NEXT: str r0, [sp, #8]
+; CHECK-O0-NEXT: ldr r0, [r7, #-12]
+; CHECK-O0-NEXT: ldr r1, [sp, #12]
+; CHECK-O0-NEXT: ldr r2, [sp, #8]
+; CHECK-O0-NEXT: bl _foo_vararg
+; CHECK-O0-NEXT: str r8, [sp, #4] @ 4-byte Spill
+; CHECK-O0-NEXT: movw r0, #0
+; CHECK-O0-NEXT: cmp r8, r0
+; CHECK-O0-NEXT: bne LBB8_2
; CHECK-O0-NEXT: @ %bb.1: @ %cont
; CHECK-O0-NEXT: ldr r1, [sp] @ 4-byte Reload
; CHECK-O0-NEXT: ldr r0, [sp, #4] @ 4-byte Reload
@@ -833,7 +833,7 @@ define float @caller4(ptr %error_ref) {
; CHECK-O0-NEXT: bl _free
; CHECK-O0-NEXT: mov r0, #1065353216
; CHECK-O0-NEXT: sub sp, r7, #4
-; CHECK-O0-NEXT: pop {r7, r8, pc}
+; CHECK-O0-NEXT: pop {r7, r8, pc}
;
; CHECK-ANDROID-LABEL: caller4:
; CHECK-ANDROID: @ %bb.0: @ %entry
@@ -987,12 +987,12 @@ define swiftcc void @swifterror_reg_clobber(ptr nocapture %err) {
;
; CHECK-O0-LABEL: swifterror_reg_clobber:
; CHECK-O0: @ %bb.0:
-; CHECK-O0-NEXT: push {r7, r8, lr}
-; CHECK-O0-NEXT: add r7, sp, #4
+; CHECK-O0-NEXT: push {r7, r8, lr}
+; CHECK-O0-NEXT: add r7, sp, #4
; CHECK-O0-NEXT: @ InlineAsm Start
; CHECK-O0-NEXT: nop
; CHECK-O0-NEXT: @ InlineAsm End
-; CHECK-O0-NEXT: pop {r7, r8, pc}
+; CHECK-O0-NEXT: pop {r7, r8, pc}
;
; CHECK-ANDROID-LABEL: swifterror_reg_clobber:
; CHECK-ANDROID: @ %bb.0:
@@ -1038,34 +1038,34 @@ define swiftcc void @params_in_reg(i32, i32, i32, i32, ptr swiftself, ptr nocapt
;
; CHECK-O0-LABEL: params_in_reg:
; CHECK-O0: @ %bb.0:
-; CHECK-O0-NEXT: push {r7, r10, lr}
-; CHECK-O0-NEXT: add r7, sp, #4
-; CHECK-O0-NEXT: sub sp, sp, #28
-; CHECK-O0-NEXT: bfc sp, #0, #3
-; CHECK-O0-NEXT: str r8, [sp, #20] @ 4-byte Spill
-; CHECK-O0-NEXT: str r10, [sp] @ 4-byte Spill
-; CHECK-O0-NEXT: str r3, [sp, #16] @ 4-byte Spill
-; CHECK-O0-NEXT: str r2, [sp, #12] @ 4-byte Spill
-; CHECK-O0-NEXT: str r1, [sp, #8] @ 4-byte Spill
-; CHECK-O0-NEXT: str r0, [sp, #4] @ 4-byte Spill
+; CHECK-O0-NEXT: push {r7, r10, lr}
+; CHECK-O0-NEXT: add r7, sp, #4
+; CHECK-O0-NEXT: sub sp, sp, #28
+; CHECK-O0-NEXT: bfc sp, #0, #3
+; CHECK-O0-NEXT: str r8, [sp, #20] @ 4-byte Spill
+; CHECK-O0-NEXT: str r10, [sp] @ 4-byte Spill
+; CHECK-O0-NEXT: str r3, [sp, #16] @ 4-byte Spill
+; CHECK-O0-NEXT: str r2, [sp, #12] @ 4-byte Spill
+; CHECK-O0-NEXT: str r1, [sp, #8] @ 4-byte Spill
+; CHECK-O0-NEXT: str r0, [sp, #4] @ 4-byte Spill
; CHECK-O0-NEXT: @ implicit-def: $r0
-; CHECK-O0-NEXT: mov r8, #0
-; CHECK-O0-NEXT: mov r0, #1
-; CHECK-O0-NEXT: mov r1, #2
-; CHECK-O0-NEXT: mov r2, #3
-; CHECK-O0-NEXT: mov r3, #4
-; CHECK-O0-NEXT: mov r10, r8
-; CHECK-O0-NEXT: bl _params_in_reg2
-; CHECK-O0-NEXT: ldr r10, [sp] @ 4-byte Reload
-; CHECK-O0-NEXT: ldr r0, [sp, #4] @ 4-byte Reload
-; CHECK-O0-NEXT: ldr r1, [sp, #8] @ 4-byte Reload
-; CHECK-O0-NEXT: ldr r2, [sp, #12] @ 4-byte Reload
-; CHECK-O0-NEXT: ldr r3, [sp, #16] @ 4-byte Reload
-; CHECK-O0-NEXT: mov r9, r8
-; CHECK-O0-NEXT: ldr r8, [sp, #20] @ 4-byte Reload
-; CHECK-O0-NEXT: bl _params_in_reg2
-; CHECK-O0-NEXT: sub sp, r7, #4
-; CHECK-O0-NEXT: pop {r7, r10, pc}
+; CHECK-O0-NEXT: mov r8, #0
+; CHECK-O0-NEXT: mov r0, #1
+; CHECK-O0-NEXT: mov r1, #2
+; CHECK-O0-NEXT: mov r2, #3
+; CHECK-O0-NEXT: mov r3, #4
+; CHECK-O0-NEXT: mov r10, r8
+; CHECK-O0-NEXT: bl _params_in_reg2
+; CHECK-O0-NEXT: ldr r10, [sp] @ 4-byte Reload
+; CHECK-O0-NEXT: ldr r0, [sp, #4] @ 4-byte Reload
+; CHECK-O0-NEXT: ldr r1, [sp, #8] @ 4-byte Reload
+; CHECK-O0-NEXT: ldr r2, [sp, #12] @ 4-byte Reload
+; CHECK-O0-NEXT: ldr r3, [sp, #16] @ 4-byte Reload
+; CHECK-O0-NEXT: mov r9, r8
+; CHECK-O0-NEXT: ldr r8, [sp, #20] @ 4-byte Reload
+; CHECK-O0-NEXT: bl _params_in_reg2
+; CHECK-O0-NEXT: sub sp, r7, #4
+; CHECK-O0-NEXT: pop {r7, r10, pc}
;
; CHECK-ANDROID-LABEL: params_in_reg:
; CHECK-ANDROID: @ %bb.0:
@@ -1153,63 +1153,63 @@ define swiftcc { i32, i32, i32, i32} @params_and_return_in_reg(i32, i32, i32, i3
;
; CHECK-O0-LABEL: params_and_return_in_reg:
; CHECK-O0: @ %bb.0:
-; CHECK-O0-NEXT: push {r7, r10, lr}
-; CHECK-O0-NEXT: add r7, sp, #4
-; CHECK-O0-NEXT: sub sp, sp, #76
-; CHECK-O0-NEXT: bfc sp, #0, #3
-; CHECK-O0-NEXT: str r8, [sp, #24] @ 4-byte Spill
-; CHECK-O0-NEXT: str r10, [sp, #4] @ 4-byte Spill
-; CHECK-O0-NEXT: str r3, [sp, #20] @ 4-byte Spill
-; CHECK-O0-NEXT: str r2, [sp, #16] @ 4-byte Spill
-; CHECK-O0-NEXT: str r1, [sp, #12] @ 4-byte Spill
-; CHECK-O0-NEXT: str r0, [sp, #8] @ 4-byte Spill
+; CHECK-O0-NEXT: push {r7, r10, lr}
+; CHECK-O0-NEXT: add r7, sp, #4
+; CHECK-O0-NEXT: sub sp, sp, #76
+; CHECK-O0-NEXT: bfc sp, #0, #3
+; CHECK-O0-NEXT: str r8, [sp, #24] @ 4-byte Spill
+; CHECK-O0-NEXT: str r10, [sp, #4] @ 4-byte Spill
+; CHECK-O0-NEXT: str r3, [sp, #20] @ 4-byte Spill
+; CHECK-O0-NEXT: str r2, [sp, #16] @ 4-byte Spill
+; CHECK-O0-NEXT: str r1, [sp, #12] @ 4-byte Spill
+; CHECK-O0-NEXT: str r0, [sp, #8] @ 4-byte Spill
; CHECK-O0-NEXT: @ implicit-def: $r0
-; CHECK-O0-NEXT: mov r8, #0
-; CHECK-O0-NEXT: str r8, [sp, #28] @ 4-byte Spill
-; CHECK-O0-NEXT: mov r0, #1
-; CHECK-O0-NEXT: str r0, [sp, #32] @ 4-byte Spill
-; CHECK-O0-NEXT: mov r1, #2
-; CHECK-O0-NEXT: str r1, [sp, #36] @ 4-byte Spill
-; CHECK-O0-NEXT: mov r2, #3
-; CHECK-O0-NEXT: str r2, [sp, #40] @ 4-byte Spill
-; CHECK-O0-NEXT: mov r3, #4
-; CHECK-O0-NEXT: str r3, [sp, #44] @ 4-byte Spill
-; CHECK-O0-NEXT: mov r10, r8
-; CHECK-O0-NEXT: bl _params_in_reg2
-; CHECK-O0-NEXT: ldr r10, [sp, #4] @ 4-byte Reload
-; CHECK-O0-NEXT: ldr r0, [sp, #8] @ 4-byte Reload
-; CHECK-O0-NEXT: ldr r1, [sp, #12] @ 4-byte Reload
-; CHECK-O0-NEXT: ldr r2, [sp, #16] @ 4-byte Reload
-; CHECK-O0-NEXT: ldr r3, [sp, #20] @ 4-byte Reload
-; CHECK-O0-NEXT: mov r9, r8
-; CHECK-O0-NEXT: ldr r8, [sp, #24] @ 4-byte Reload
-; CHECK-O0-NEXT: str r9, [sp, #48] @ 4-byte Spill
-; CHECK-O0-NEXT: bl _params_and_return_in_reg2
-; CHECK-O0-NEXT: ldr r10, [sp, #28] @ 4-byte Reload
-; CHECK-O0-NEXT: mov r9, r0
-; CHECK-O0-NEXT: ldr r0, [sp, #32] @ 4-byte Reload
-; CHECK-O0-NEXT: str r9, [sp, #52] @ 4-byte Spill
-; CHECK-O0-NEXT: mov r9, r1
-; CHECK-O0-NEXT: ldr r1, [sp, #36] @ 4-byte Reload
-; CHECK-O0-NEXT: str r9, [sp, #56] @ 4-byte Spill
-; CHECK-O0-NEXT: mov r9, r2
-; CHECK-O0-NEXT: ldr r2, [sp, #40] @ 4-byte Reload
-; CHECK-O0-NEXT: str r9, [sp, #60] @ 4-byte Spill
-; CHECK-O0-NEXT: mov r9, r3
-; CHECK-O0-NEXT: ldr r3, [sp, #44] @ 4-byte Reload
-; CHECK-O0-NEXT: str r9, [sp, #64] @ 4-byte Spill
-; CHECK-O0-NEXT: mov r9, r8
-; CHECK-O0-NEXT: ldr r8, [sp, #48] @ 4-byte Reload
-; CHECK-O0-NEXT: str r9, [sp, #68] @ 4-byte Spill
-; CHECK-O0-NEXT: bl _params_in_reg2
-; CHECK-O0-NEXT: ldr r0, [sp, #52] @ 4-byte Reload
-; CHECK-O0-NEXT: ldr r1, [sp, #56] @ 4-byte Reload
-; CHECK-O0-NEXT: ldr r2, [sp, #60] @ 4-byte Reload
-; CHECK-O0-NEXT: ldr r3, [sp, #64] @ 4-byte Reload
-; CHECK-O0-NEXT: mov r9, r8
-; CHECK-O0-NEXT: ldr r8, [sp, #68] @ 4-byte Reload
-; CHECK-O0-NEXT: sub sp, r7, #4
-; CHECK-O0-NEXT: pop {r7, r10, pc}
+; CHECK-O0-NEXT: mov r8, #0
+; CHECK-O0-NEXT: str r8, [sp, #28] @ 4-byte Spill
+; CHECK-O0-NEXT: mov r0, #1
+; CHECK-O0-NEXT: str r0, [sp, #32] @ 4-byte Spill
+; CHECK-O0-NEXT: mov r1, #2
+; CHECK-O0-NEXT: str r1, [sp, #36] @ 4-byte Spill
+; CHECK-O0-NEXT: mov r2, #3
+; CHECK-O0-NEXT: str r2, [sp, #40] @ 4-byte Spill
+; CHECK-O0-NEXT: mov r3, #4
+; CHECK-O0-NEXT: str r3, [sp, #44] @ 4-byte Spill
+; CHECK-O0-NEXT: mov r10, r8
+; CHECK-O0-NEXT: bl _params_in_reg2
+; CHECK-O0-NEXT: ldr r10, [sp, #4] @ 4-byte Reload
+; CHECK-O0-NEXT: ldr r0, [sp, #8] @ 4-byte Reload
+; CHECK-O0-NEXT: ldr r1, [sp, #12] @ 4-byte Reload
+; CHECK-O0-NEXT: ldr r2, [sp, #16] @ 4-byte Reload
+; CHECK-O0-NEXT: ldr r3, [sp, #20] @ 4-byte Reload
+; CHECK-O0-NEXT: mov r9, r8
+; CHECK-O0-NEXT: ldr r8, [sp, #24] @ 4-byte Reload
+; CHECK-O0-NEXT: str r9, [sp, #48] @ 4-byte Spill
+; CHECK-O0-NEXT: bl _params_and_return_in_reg2
+; CHECK-O0-NEXT: ldr r10, [sp, #28] @ 4-byte Reload
+; CHECK-O0-NEXT: mov r9, r0
+; CHECK-O0-NEXT: ldr r0, [sp, #32] @ 4-byte Reload
+; CHECK-O0-NEXT: str r9, [sp, #52] @ 4-byte Spill
+; CHECK-O0-NEXT: mov r9, r1
+; CHECK-O0-NEXT: ldr r1, [sp, #36] @ 4-byte Reload
+; CHECK-O0-NEXT: str r9, [sp, #56] @ 4-byte Spill
+; CHECK-O0-NEXT: mov r9, r2
+; CHECK-O0-NEXT: ldr r2, [sp, #40] @ 4-byte Reload
+; CHECK-O0-NEXT: str r9, [sp, #60] @ 4-byte Spill
+; CHECK-O0-NEXT: mov r9, r3
+; CHECK-O0-NEXT: ldr r3, [sp, #44] @ 4-byte Reload
+; CHECK-O0-NEXT: str r9, [sp, #64] @ 4-byte Spill
+; CHECK-O0-NEXT: mov r9, r8
+; CHECK-O0-NEXT: ldr r8, [sp, #48] @ 4-byte Reload
+; CHECK-O0-NEXT: str r9, [sp, #68] @ 4-byte Spill
+; CHECK-O0-NEXT: bl _params_in_reg2
+; CHECK-O0-NEXT: ldr r0, [sp, #52] @ 4-byte Reload
+; CHECK-O0-NEXT: ldr r1, [sp, #56] @ 4-byte Reload
+; CHECK-O0-NEXT: ldr r2, [sp, #60] @ 4-byte Reload
+; CHECK-O0-NEXT: ldr r3, [sp, #64] @ 4-byte Reload
+; CHECK-O0-NEXT: mov r9, r8
+; CHECK-O0-NEXT: ldr r8, [sp, #68] @ 4-byte Reload
+; CHECK-O0-NEXT: sub sp, r7, #4
+; CHECK-O0-NEXT: pop {r7, r10, pc}
;
; CHECK-ANDROID-LABEL: params_and_return_in_reg:
; CHECK-ANDROID: @ %bb.0:
@@ -1325,17 +1325,17 @@ define swiftcc ptr @testAssign(ptr %error_ref) {
;
; CHECK-O0-LABEL: testAssign:
; CHECK-O0: @ %bb.0: @ %entry
-; CHECK-O0-NEXT: push {r7, r8, lr}
-; CHECK-O0-NEXT: add r7, sp, #4
-; CHECK-O0-NEXT: sub sp, sp, #8
+; CHECK-O0-NEXT: push {r7, r8, lr}
+; CHECK-O0-NEXT: add r7, sp, #4
+; CHECK-O0-NEXT: sub sp, sp, #8
; CHECK-O0-NEXT: @ implicit-def: $r1
-; CHECK-O0-NEXT: mov r8, #0
-; CHECK-O0-NEXT: bl _foo2
-; CHECK-O0-NEXT: str r8, [sp] @ 4-byte Spill
+; CHECK-O0-NEXT: mov r8, #0
+; CHECK-O0-NEXT: bl _foo2
+; CHECK-O0-NEXT: str r8, [sp] @ 4-byte Spill
; CHECK-O0-NEXT: @ %bb.1: @ %a
; CHECK-O0-NEXT: ldr r0, [sp] @ 4-byte Reload
; CHECK-O0-NEXT: sub sp, r7, #4
-; CHECK-O0-NEXT: pop {r7, r8, pc}
+; CHECK-O0-NEXT: pop {r7, r8, pc}
;
; CHECK-ANDROID-LABEL: testAssign:
; CHECK-ANDROID: @ %bb.0: @ %entry
diff --git a/llvm/test/CodeGen/AVR/bug-81911.ll b/llvm/test/CodeGen/AVR/bug-81911.ll
index 2a22666a1ff92..d3436e2da1d3d 100644
--- a/llvm/test/CodeGen/AVR/bug-81911.ll
+++ b/llvm/test/CodeGen/AVR/bug-81911.ll
@@ -41,31 +41,31 @@ define internal i8 @main() {
; CHECK-NEXT: adiw r24, 6
; CHECK-NEXT: std Y+3, r25 ; 2-byte Folded Spill
; CHECK-NEXT: std Y+2, r24 ; 2-byte Folded Spill
-; CHECK-NEXT: movw r8, r16
-; CHECK-NEXT: movw r6, r16
-; CHECK-NEXT: movw r4, r16
; CHECK-NEXT: movw r2, r16
+; CHECK-NEXT: movw r4, r16
+; CHECK-NEXT: movw r6, r16
+; CHECK-NEXT: movw r8, r16
; CHECK-NEXT: rjmp .LBB0_2
; CHECK-NEXT: .LBB0_1: ; %bb1
; CHECK-NEXT: ; in Loop: Header=BB0_2 Depth=1
; CHECK-NEXT: andi r30, 1
; CHECK-NEXT: ldd r31, Y+4 ; 1-byte Folded Reload
; CHECK-NEXT: dec r31
+; CHECK-NEXT: movw r8, r24
+; CHECK-NEXT: movw r6, r22
+; CHECK-NEXT: movw r4, r20
+; CHECK-NEXT: movw r2, r18
; CHECK-NEXT: cpi r30, 0
-; CHECK-NEXT: movw r8, r18
-; CHECK-NEXT: movw r6, r20
-; CHECK-NEXT: movw r4, r22
-; CHECK-NEXT: movw r2, r24
; CHECK-NEXT: mov r18, r31
; CHECK-NEXT: brne .LBB0_2
; CHECK-NEXT: rjmp .LBB0_4
; CHECK-NEXT: .LBB0_2: ; %bb1
; CHECK-NEXT: ; =>This Inner Loop Header: Depth=1
; CHECK-NEXT: std Y+4, r18 ; 1-byte Folded Spill
-; CHECK-NEXT: movw r18, r8
-; CHECK-NEXT: movw r20, r6
-; CHECK-NEXT: movw r22, r4
-; CHECK-NEXT: movw r24, r2
+; CHECK-NEXT: movw r18, r2
+; CHECK-NEXT: movw r20, r4
+; CHECK-NEXT: movw r22, r6
+; CHECK-NEXT: movw r24, r8
; CHECK-NEXT: ldi r26, 10
; CHECK-NEXT: ldi r27, 0
; CHECK-NEXT: movw r10, r26
@@ -85,14 +85,14 @@ define internal i8 @main() {
; CHECK-NEXT: ;APP
; CHECK-NEXT: ;NO_APP
; CHECK-NEXT: ldi r30, 1
-; CHECK-NEXT: cp r8, r1
-; CHECK-NEXT: cpc r9, r1
-; CHECK-NEXT: cpc r6, r16
-; CHECK-NEXT: cpc r7, r17
+; CHECK-NEXT: cp r2, r1
+; CHECK-NEXT: cpc r3, r1
; CHECK-NEXT: cpc r4, r16
; CHECK-NEXT: cpc r5, r17
-; CHECK-NEXT: cpc r2, r16
-; CHECK-NEXT: cpc r3, r17
+; CHECK-NEXT: cpc r6, r16
+; CHECK-NEXT: cpc r7, r17
+; CHECK-NEXT: cpc r8, r16
+; CHECK-NEXT: cpc r9, r17
; CHECK-NEXT: breq .LBB0_3
; CHECK-NEXT: rjmp .LBB0_1
; CHECK-NEXT: .LBB0_3: ; %bb1
diff --git a/llvm/test/CodeGen/LoongArch/machinelicm-address-pseudos.ll b/llvm/test/CodeGen/LoongArch/machinelicm-address-pseudos.ll
index 9142e718e8adc..06edb736e0435 100644
--- a/llvm/test/CodeGen/LoongArch/machinelicm-address-pseudos.ll
+++ b/llvm/test/CodeGen/LoongArch/machinelicm-address-pseudos.ll
@@ -39,17 +39,17 @@ define void @test_la_pcrel(i32 signext %n) {
;
; LA64LARGE-LABEL: test_la_pcrel:
; LA64LARGE: # %bb.0: # %entry
-; LA64LARGE-NEXT: pcalau12i $a1, %pc_hi20(l)
-; LA64LARGE-NEXT: addi.d $a2, $zero, %pc_lo12(l)
-; LA64LARGE-NEXT: lu32i.d $a2, %pc64_lo20(l)
-; LA64LARGE-NEXT: lu52i.d $a2, $a2, %pc64_hi12(l)
-; LA64LARGE-NEXT: move $a3, $zero
+; LA64LARGE-NEXT: move $a1, $zero
+; LA64LARGE-NEXT: pcalau12i $a2, %pc_hi20(l)
+; LA64LARGE-NEXT: addi.d $a3, $zero, %pc_lo12(l)
+; LA64LARGE-NEXT: lu32i.d $a3, %pc64_lo20(l)
+; LA64LARGE-NEXT: lu52i.d $a3, $a3, %pc64_hi12(l)
; LA64LARGE-NEXT: .p2align 4, , 16
; LA64LARGE-NEXT: .LBB0_1: # %loop
; LA64LARGE-NEXT: # =>This Inner Loop Header: Depth=1
-; LA64LARGE-NEXT: ldx.w $zero, $a2, $a1
-; LA64LARGE-NEXT: addi.w $a3, $a3, 1
-; LA64LARGE-NEXT: blt $a3, $a0, .LBB0_1
+; LA64LARGE-NEXT: ldx.w $zero, $a3, $a2
+; LA64LARGE-NEXT: addi.w $a1, $a1, 1
+; LA64LARGE-NEXT: blt $a1, $a0, .LBB0_1
; LA64LARGE-NEXT: # %bb.2: # %ret
; LA64LARGE-NEXT: ret
entry:
@@ -99,18 +99,18 @@ define void @test_la_got(i32 signext %n) {
;
; LA64LARGE-LABEL: test_la_got:
; LA64LARGE: # %bb.0: # %entry
-; LA64LARGE-NEXT: pcalau12i $a1, %got_pc_hi20(g)
-; LA64LARGE-NEXT: addi.d $a2, $zero, %got_pc_lo12(g)
-; LA64LARGE-NEXT: lu32i.d $a2, %got64_pc_lo20(g)
-; LA64LARGE-NEXT: lu52i.d $a2, $a2, %got64_pc_hi12(g)
-; LA64LARGE-NEXT: ldx.d $a1, $a2, $a1
-; LA64LARGE-NEXT: move $a2, $zero
+; LA64LARGE-NEXT: move $a1, $zero
+; LA64LARGE-NEXT: pcalau12i $a2, %got_pc_hi20(g)
+; LA64LARGE-NEXT: addi.d $a3, $zero, %got_pc_lo12(g)
+; LA64LARGE-NEXT: lu32i.d $a3, %got64_pc_lo20(g)
+; LA64LARGE-NEXT: lu52i.d $a3, $a3, %got64_pc_hi12(g)
+; LA64LARGE-NEXT: ldx.d $a2, $a3, $a2
; LA64LARGE-NEXT: .p2align 4, , 16
; LA64LARGE-NEXT: .LBB1_1: # %loop
; LA64LARGE-NEXT: # =>This Inner Loop Header: Depth=1
-; LA64LARGE-NEXT: ld.w $zero, $a1, 0
-; LA64LARGE-NEXT: addi.w $a2, $a2, 1
-; LA64LARGE-NEXT: blt $a2, $a0, .LBB1_1
+; LA64LARGE-NEXT: ld.w $zero, $a2, 0
+; LA64LARGE-NEXT: addi.w $a1, $a1, 1
+; LA64LARGE-NEXT: blt $a1, $a0, .LBB1_1
; LA64LARGE-NEXT: # %bb.2: # %ret
; LA64LARGE-NEXT: ret
entry:
@@ -161,18 +161,18 @@ define void @test_la_tls_ie(i32 signext %n) {
;
; LA64LARGE-LABEL: test_la_tls_ie:
; LA64LARGE: # %bb.0: # %entry
-; LA64LARGE-NEXT: pcalau12i $a1, %ie_pc_hi20(ie)
-; LA64LARGE-NEXT: addi.d $a2, $zero, %ie_pc_lo12(ie)
-; LA64LARGE-NEXT: lu32i.d $a2, %ie64_pc_lo20(ie)
-; LA64LARGE-NEXT: lu52i.d $a2, $a2, %ie64_pc_hi12(ie)
-; LA64LARGE-NEXT: ldx.d $a1, $a2, $a1
-; LA64LARGE-NEXT: move $a2, $zero
+; LA64LARGE-NEXT: move $a1, $zero
+; LA64LARGE-NEXT: pcalau12i $a2, %ie_pc_hi20(ie)
+; LA64LARGE-NEXT: addi.d $a3, $zero, %ie_pc_lo12(ie)
+; LA64LARGE-NEXT: lu32i.d $a3, %ie64_pc_lo20(ie)
+; LA64LARGE-NEXT: lu52i.d $a3, $a3, %ie64_pc_hi12(ie)
+; LA64LARGE-NEXT: ldx.d $a2, $a3, $a2
; LA64LARGE-NEXT: .p2align 4, , 16
; LA64LARGE-NEXT: .LBB2_1: # %loop
; LA64LARGE-NEXT: # =>This Inner Loop Header: Depth=1
-; LA64LARGE-NEXT: ldx.w $zero, $a1, $tp
-; LA64LARGE-NEXT: addi.w $a2, $a2, 1
-; LA64LARGE-NEXT: blt $a2, $a0, .LBB2_1
+; LA64LARGE-NEXT: ldx.w $zero, $a2, $tp
+; LA64LARGE-NEXT: addi.w $a1, $a1, 1
+; LA64LARGE-NEXT: blt $a1, $a0, .LBB2_1
; LA64LARGE-NEXT: # %bb.2: # %ret
; LA64LARGE-NEXT: ret
entry:
@@ -270,11 +270,11 @@ define void @test_la_tls_ld(i32 signext %n) {
; LA64LARGE-NEXT: .cfi_offset 23, -24
; LA64LARGE-NEXT: .cfi_offset 24, -32
; LA64LARGE-NEXT: move $fp, $a0
+; LA64LARGE-NEXT: move $s1, $zero
; LA64LARGE-NEXT: pcalau12i $a0, %ld_pc_hi20(ld)
; LA64LARGE-NEXT: addi.d $a1, $zero, %got_pc_lo12(ld)
; LA64LARGE-NEXT: lu32i.d $a1, %got64_pc_lo20(ld)
; LA64LARGE-NEXT: lu52i.d $a1, $a1, %got64_pc_hi12(ld)
-; LA64LARGE-NEXT: move $s1, $zero
; LA64LARGE-NEXT: add.d $s0, $a1, $a0
; LA64LARGE-NEXT: .p2align 4, , 16
; LA64LARGE-NEXT: .LBB3_1: # %loop
@@ -436,11 +436,11 @@ define void @test_la_tls_gd(i32 signext %n) nounwind {
; LA64LARGE-NEXT: st.d $s0, $sp, 8 # 8-byte Folded Spill
; LA64LARGE-NEXT: st.d $s1, $sp, 0 # 8-byte Folded Spill
; LA64LARGE-NEXT: move $fp, $a0
+; LA64LARGE-NEXT: move $s1, $zero
; LA64LARGE-NEXT: pcalau12i $a0, %gd_pc_hi20(gd)
; LA64LARGE-NEXT: addi.d $a1, $zero, %got_pc_lo12(gd)
; LA64LARGE-NEXT: lu32i.d $a1, %got64_pc_lo20(gd)
; LA64LARGE-NEXT: lu52i.d $a1, $a1, %got64_pc_hi12(gd)
-; LA64LARGE-NEXT: move $s1, $zero
; LA64LARGE-NEXT: add.d $s0, $a1, $a0
; LA64LARGE-NEXT: .p2align 4, , 16
; LA64LARGE-NEXT: .LBB5_1: # %loop
diff --git a/llvm/test/CodeGen/PowerPC/disable-ctr-ppcf128.ll b/llvm/test/CodeGen/PowerPC/disable-ctr-ppcf128.ll
index cd5ea16d4600b..cd2fbdfe71263 100644
--- a/llvm/test/CodeGen/PowerPC/disable-ctr-ppcf128.ll
+++ b/llvm/test/CodeGen/PowerPC/disable-ctr-ppcf128.ll
@@ -51,9 +51,9 @@ define ppc_fp128 @test_ctr0() {
; P9BE-NEXT: .cfi_offset r30, -16
; P9BE-NEXT: li r3, 1
; P9BE-NEXT: std r30, 112(r1) # 8-byte Folded Spill
-; P9BE-NEXT: xxlxor f1, f1, f1
-; P9BE-NEXT: rldic r30, r3, 62, 1
; P9BE-NEXT: xxlxor f2, f2, f2
+; P9BE-NEXT: rldic r30, r3, 62, 1
+; P9BE-NEXT: xxlxor f1, f1, f1
; P9BE-NEXT: .p2align 5
; P9BE-NEXT: .LBB0_1: # %bb6
; P9BE-NEXT: #
@@ -111,8 +111,8 @@ define ppc_fp128 @test_ctr0() {
; P8BE-NEXT: .cfi_offset r30, -16
; P8BE-NEXT: li r3, 1
; P8BE-NEXT: std r30, 112(r1) # 8-byte Folded Spill
-; P8BE-NEXT: xxlxor f1, f1, f1
; P8BE-NEXT: xxlxor f2, f2, f2
+; P8BE-NEXT: xxlxor f1, f1, f1
; P8BE-NEXT: rldic r30, r3, 62, 1
; P8BE-NEXT: .p2align 5
; P8BE-NEXT: .LBB0_1: # %bb6
diff --git a/llvm/test/CodeGen/PowerPC/sms-phi-2.ll b/llvm/test/CodeGen/PowerPC/sms-phi-2.ll
index 4904d11fc8104..0077673292ab3 100644
--- a/llvm/test/CodeGen/PowerPC/sms-phi-2.ll
+++ b/llvm/test/CodeGen/PowerPC/sms-phi-2.ll
@@ -9,7 +9,7 @@ define void @phi2(i32, i32, ptr) local_unnamed_addr {
; CHECK-NEXT: li 5, 55
; CHECK-NEXT: li 6, 48
; CHECK-NEXT: mtctr 3
-; CHECK-NEXT: bdz .LBB0_4
+; CHECK-NEXT: bdz .LBB0_3
; CHECK-NEXT: # %bb.1:
; CHECK-NEXT: divw 9, 8, 4
; CHECK-NEXT: mullw 7, 8, 4
@@ -19,7 +19,7 @@ define void @phi2(i32, i32, ptr) local_unnamed_addr {
; CHECK-NEXT: add 3, 7, 3
; CHECK-NEXT: stbu 3, -1(7)
; CHECK-NEXT: mr 3, 8
-; CHECK-NEXT: bdz .LBB0_3
+; CHECK-NEXT: bdz .LBB0_4
; CHECK-NEXT: .p2align 4
; CHECK-NEXT: .LBB0_2:
; CHECK-NEXT: mr 3, 9
@@ -33,13 +33,12 @@ define void @phi2(i32, i32, ptr) local_unnamed_addr {
; CHECK-NEXT: stbu 8, -1(7)
; CHECK-NEXT: mr 8, 3
; CHECK-NEXT: bdnz .LBB0_2
+; CHECK-NEXT: b .LBB0_4
; CHECK-NEXT: .LBB0_3:
-; CHECK-NEXT: mr 8, 9
-; CHECK-NEXT: b .LBB0_5
-; CHECK-NEXT: .LBB0_4:
; CHECK-NEXT: # implicit-def: $x7
-; CHECK-NEXT: .LBB0_5:
-; CHECK-NEXT: mullw 4, 8, 4
+; CHECK-NEXT: mr 9, 8
+; CHECK-NEXT: .LBB0_4:
+; CHECK-NEXT: mullw 4, 9, 4
; CHECK-NEXT: sub 3, 3, 4
; CHECK-NEXT: cmplwi 3, 10
; CHECK-NEXT: isellt 4, 6, 5
diff --git a/llvm/test/CodeGen/PowerPC/sms-phi-3.ll b/llvm/test/CodeGen/PowerPC/sms-phi-3.ll
index 628822edabf39..2dd8b36389c62 100644
--- a/llvm/test/CodeGen/PowerPC/sms-phi-3.ll
+++ b/llvm/test/CodeGen/PowerPC/sms-phi-3.ll
@@ -19,34 +19,34 @@ define void @phi3(ptr) nounwind {
; CHECK-NEXT: mr 29, 3
; CHECK-NEXT: bl malloc
; CHECK-NEXT: nop
-; CHECK-NEXT: addi 7, 30, -4
+; CHECK-NEXT: addi 6, 30, -4
; CHECK-NEXT: mtctr 3
; CHECK-NEXT: addi 4, 29, -8
; CHECK-NEXT: li 5, 0
-; CHECK-NEXT: lwzu 8, 4(7)
+; CHECK-NEXT: lwzu 8, 4(6)
; CHECK-NEXT: bdz .LBB0_5
; CHECK-NEXT: # %bb.1:
-; CHECK-NEXT: extswsli 6, 5, 5
+; CHECK-NEXT: extswsli 7, 5, 5
; CHECK-NEXT: add 5, 8, 5
-; CHECK-NEXT: lwzu 8, 4(7)
+; CHECK-NEXT: lwzu 8, 4(6)
; CHECK-NEXT: bdz .LBB0_4
; CHECK-NEXT: # %bb.2:
-; CHECK-NEXT: add 6, 3, 6
-; CHECK-NEXT: stdu 6, 8(4)
-; CHECK-NEXT: extswsli 6, 5, 5
+; CHECK-NEXT: add 7, 3, 7
+; CHECK-NEXT: stdu 7, 8(4)
+; CHECK-NEXT: extswsli 7, 5, 5
; CHECK-NEXT: add 5, 8, 5
-; CHECK-NEXT: lwzu 8, 4(7)
+; CHECK-NEXT: lwzu 8, 4(6)
; CHECK-NEXT: bdz .LBB0_4
; CHECK-NEXT: .p2align 5
; CHECK-NEXT: .LBB0_3:
-; CHECK-NEXT: add 9, 3, 6
-; CHECK-NEXT: extswsli 6, 5, 5
+; CHECK-NEXT: add 9, 3, 7
+; CHECK-NEXT: extswsli 7, 5, 5
; CHECK-NEXT: add 5, 8, 5
-; CHECK-NEXT: lwzu 8, 4(7)
+; CHECK-NEXT: lwzu 8, 4(6)
; CHECK-NEXT: stdu 9, 8(4)
; CHECK-NEXT: bdnz .LBB0_3
; CHECK-NEXT: .LBB0_4:
-; CHECK-NEXT: add 6, 3, 6
+; CHECK-NEXT: add 6, 3, 7
; CHECK-NEXT: stdu 6, 8(4)
; CHECK-NEXT: .LBB0_5:
; CHECK-NEXT: extswsli 5, 5, 5
diff --git a/llvm/test/CodeGen/PowerPC/stack-restore-with-setjmp.ll b/llvm/test/CodeGen/PowerPC/stack-restore-with-setjmp.ll
index e225e63980c7f..eec5b4588f7c3 100644
--- a/llvm/test/CodeGen/PowerPC/stack-restore-with-setjmp.ll
+++ b/llvm/test/CodeGen/PowerPC/stack-restore-with-setjmp.ll
@@ -16,13 +16,12 @@ define dso_local signext i32 @main(i32 signext %argc, ptr nocapture readnone %ar
; CHECK-NEXT: stw 12, 8(1)
; CHECK-NEXT: mflr 0
; CHECK-NEXT: stdu 1, -784(1)
-; CHECK-NEXT: # kill: def $r3 killed $r3 killed $x3
-; CHECK-NEXT: cmpwi 2, 3, 2
-; CHECK-NEXT: li 4, 0
-; CHECK-NEXT: # kill: def $r4 killed $r4 killed $x4
-; CHECK-NEXT: mr 3, 4
+; CHECK-NEXT: mr 4, 3
; CHECK-NEXT: std 0, 800(1)
; CHECK-NEXT: mr 31, 1
+; CHECK-NEXT: li 3, 0
+; CHECK-NEXT: # kill: def $r3 killed $r3 killed $x3
+; CHECK-NEXT: cmpwi 2, 4, 2
; CHECK-NEXT: blt 2, .LBB0_3
; CHECK-NEXT: # %bb.1: # %if.end
; CHECK-NEXT: addi 3, 31, 112
@@ -66,7 +65,6 @@ define dso_local signext i32 @main(i32 signext %argc, ptr nocapture readnone %ar
; BE-NEXT: stdu 1, -800(1)
; BE-NEXT: li 4, 0
; BE-NEXT: # kill: def $r3 killed $r3 killed $x3
-; BE-NEXT: # kill: def $r4 killed $r4 killed $x4
; BE-NEXT: cmpwi 2, 3, 2
; BE-NEXT: mr 3, 4
; BE-NEXT: std 0, 816(1)
diff --git a/llvm/test/CodeGen/PowerPC/subreg-postra-2.ll b/llvm/test/CodeGen/PowerPC/subreg-postra-2.ll
index f696745c9d414..10fa8221778f5 100644
--- a/llvm/test/CodeGen/PowerPC/subreg-postra-2.ll
+++ b/llvm/test/CodeGen/PowerPC/subreg-postra-2.ll
@@ -60,27 +60,27 @@ define void @jbd2_journal_commit_transaction(i32 %input1, ptr %input2, ptr %inpu
; CHECK-NO-ISEL-NEXT: bne- 0, .Ltmp0
; CHECK-NO-ISEL-EMPTY:
; CHECK-NO-ISEL-NEXT: #NO_APP
-; CHECK-NO-ISEL-NEXT: std 5, 0(6)
+; CHECK-NO-ISEL-NEXT: std 4, 0(6)
; CHECK-NO-ISEL-NEXT: beq- 5, .LBB0_6
; CHECK-NO-ISEL-NEXT: .LBB0_2: # %while.body392
; CHECK-NO-ISEL-NEXT: #
; CHECK-NO-ISEL-NEXT: bne- 1, .LBB0_5
; CHECK-NO-ISEL-NEXT: # %bb.3: # %wait_on_buffer.exit1319
; CHECK-NO-ISEL-NEXT: #
-; CHECK-NO-ISEL-NEXT: ld 5, 0(6)
-; CHECK-NO-ISEL-NEXT: mr 9, 5
-; CHECK-NO-ISEL-NEXT: ldu 4, -72(9)
-; CHECK-NO-ISEL-NEXT: andi. 4, 4, 1
-; CHECK-NO-ISEL-NEXT: mr 4, 3
+; CHECK-NO-ISEL-NEXT: ld 4, 0(6)
+; CHECK-NO-ISEL-NEXT: mr 9, 4
+; CHECK-NO-ISEL-NEXT: ldu 5, -72(9)
+; CHECK-NO-ISEL-NEXT: andi. 5, 5, 1
+; CHECK-NO-ISEL-NEXT: mr 5, 3
; CHECK-NO-ISEL-NEXT: bc 12, 1, .LBB0_1
; CHECK-NO-ISEL-NEXT: # %bb.4: # %wait_on_buffer.exit1319
; CHECK-NO-ISEL-NEXT: #
-; CHECK-NO-ISEL-NEXT: li 4, -5
+; CHECK-NO-ISEL-NEXT: li 5, -5
; CHECK-NO-ISEL-NEXT: b .LBB0_1
; CHECK-NO-ISEL-NEXT: .LBB0_5:
-; CHECK-NO-ISEL-NEXT: mr 4, 7
+; CHECK-NO-ISEL-NEXT: mr 5, 7
; CHECK-NO-ISEL-NEXT: .LBB0_6: # %while.end418
-; CHECK-NO-ISEL-NEXT: cmplwi 4, 0
+; CHECK-NO-ISEL-NEXT: cmplwi 5, 0
; CHECK-NO-ISEL-NEXT: beq 0, .LBB0_8
; CHECK-NO-ISEL-NEXT: # %bb.7: # %if.then420
; CHECK-NO-ISEL-NEXT: .LBB0_8: # %if.end421
diff --git a/llvm/test/CodeGen/PowerPC/vsx.ll b/llvm/test/CodeGen/PowerPC/vsx.ll
index 14b3d69f8c273..9e0dc87f0ab8b 100644
--- a/llvm/test/CodeGen/PowerPC/vsx.ll
+++ b/llvm/test/CodeGen/PowerPC/vsx.ll
@@ -2488,8 +2488,8 @@ define double @test82(double %a, double %b, double %c, double %d) {
; CHECK-FISL: # %bb.0: # %entry
; CHECK-FISL-NEXT: stfd f2, -16(r1) # 8-byte Folded Spill
; CHECK-FISL-NEXT: fmr f2, f1
-; CHECK-FISL-NEXT: xscmpudp cr0, f3, f4
; CHECK-FISL-NEXT: stfd f2, -8(r1) # 8-byte Folded Spill
+; CHECK-FISL-NEXT: xscmpudp cr0, f3, f4
; CHECK-FISL-NEXT: beq cr0, .LBB67_2
; CHECK-FISL-NEXT: # %bb.1: # %entry
; CHECK-FISL-NEXT: lfd f0, -16(r1) # 8-byte Folded Reload
diff --git a/llvm/test/CodeGen/RISCV/abds.ll b/llvm/test/CodeGen/RISCV/abds.ll
index efb4e1a6f15d6..2133ffeeb19dc 100644
--- a/llvm/test/CodeGen/RISCV/abds.ll
+++ b/llvm/test/CodeGen/RISCV/abds.ll
@@ -543,17 +543,17 @@ define i128 @abd_ext_i128(i128 %a, i128 %b) nounwind {
; RV32I-NEXT: lw t1, 12(a2)
; RV32I-NEXT: lw a2, 4(a2)
; RV32I-NEXT: sltu t0, a6, a5
-; RV32I-NEXT: mv t4, t0
+; RV32I-NEXT: mv t3, t0
; RV32I-NEXT: beq a7, t1, .LBB11_2
; RV32I-NEXT: # %bb.1:
-; RV32I-NEXT: slt t4, t1, a7
+; RV32I-NEXT: slt t3, t1, a7
; RV32I-NEXT: .LBB11_2:
; RV32I-NEXT: sltu t2, a1, a3
; RV32I-NEXT: sltu t5, a2, a4
-; RV32I-NEXT: mv t3, t2
+; RV32I-NEXT: mv t4, t2
; RV32I-NEXT: beq a4, a2, .LBB11_4
; RV32I-NEXT: # %bb.3:
-; RV32I-NEXT: mv t3, t5
+; RV32I-NEXT: mv t4, t5
; RV32I-NEXT: .LBB11_4:
; RV32I-NEXT: addi sp, sp, -16
; RV32I-NEXT: sw s0, 12(sp) # 4-byte Folded Spill
@@ -562,12 +562,12 @@ define i128 @abd_ext_i128(i128 %a, i128 %b) nounwind {
; RV32I-NEXT: or t6, s0, t6
; RV32I-NEXT: beqz t6, .LBB11_6
; RV32I-NEXT: # %bb.5:
-; RV32I-NEXT: mv t3, t4
+; RV32I-NEXT: mv t4, t3
; RV32I-NEXT: .LBB11_6:
-; RV32I-NEXT: mv t4, t2
+; RV32I-NEXT: mv t3, t2
; RV32I-NEXT: beq a2, a4, .LBB11_8
; RV32I-NEXT: # %bb.7:
-; RV32I-NEXT: mv t4, t5
+; RV32I-NEXT: mv t3, t5
; RV32I-NEXT: .LBB11_8:
; RV32I-NEXT: sltu t5, a3, a1
; RV32I-NEXT: mv t6, t5
@@ -575,17 +575,17 @@ define i128 @abd_ext_i128(i128 %a, i128 %b) nounwind {
; RV32I-NEXT: # %bb.9:
; RV32I-NEXT: sltu t6, a4, a2
; RV32I-NEXT: .LBB11_10:
-; RV32I-NEXT: bnez t3, .LBB11_12
+; RV32I-NEXT: bnez t4, .LBB11_12
; RV32I-NEXT: # %bb.11:
; RV32I-NEXT: sub a7, t1, a7
; RV32I-NEXT: sub a5, a6, a5
; RV32I-NEXT: sub a1, a1, a3
; RV32I-NEXT: sub a2, a2, a4
; RV32I-NEXT: sub a4, a7, t0
-; RV32I-NEXT: sltu a6, a5, t4
+; RV32I-NEXT: sltu a6, a5, t3
; RV32I-NEXT: sub a3, a2, t2
; RV32I-NEXT: sub a2, a4, a6
-; RV32I-NEXT: sub a4, a5, t4
+; RV32I-NEXT: sub a4, a5, t3
; RV32I-NEXT: j .LBB11_13
; RV32I-NEXT: .LBB11_12:
; RV32I-NEXT: sltu t0, a5, a6
@@ -639,17 +639,17 @@ define i128 @abd_ext_i128(i128 %a, i128 %b) nounwind {
; RV32ZBB-NEXT: lw t1, 12(a2)
; RV32ZBB-NEXT: lw a2, 4(a2)
; RV32ZBB-NEXT: sltu t0, a6, a5
-; RV32ZBB-NEXT: mv t4, t0
+; RV32ZBB-NEXT: mv t3, t0
; RV32ZBB-NEXT: beq a7, t1, .LBB11_2
; RV32ZBB-NEXT: # %bb.1:
-; RV32ZBB-NEXT: slt t4, t1, a7
+; RV32ZBB-NEXT: slt t3, t1, a7
; RV32ZBB-NEXT: .LBB11_2:
; RV32ZBB-NEXT: sltu t2, a1, a3
; RV32ZBB-NEXT: sltu t5, a2, a4
-; RV32ZBB-NEXT: mv t3, t2
+; RV32ZBB-NEXT: mv t4, t2
; RV32ZBB-NEXT: beq a4, a2, .LBB11_4
; RV32ZBB-NEXT: # %bb.3:
-; RV32ZBB-NEXT: mv t3, t5
+; RV32ZBB-NEXT: mv t4, t5
; RV32ZBB-NEXT: .LBB11_4:
; RV32ZBB-NEXT: addi sp, sp, -16
; RV32ZBB-NEXT: sw s0, 12(sp) # 4-byte Folded Spill
@@ -658,12 +658,12 @@ define i128 @abd_ext_i128(i128 %a, i128 %b) nounwind {
; RV32ZBB-NEXT: or t6, s0, t6
; RV32ZBB-NEXT: beqz t6, .LBB11_6
; RV32ZBB-NEXT: # %bb.5:
-; RV32ZBB-NEXT: mv t3, t4
+; RV32ZBB-NEXT: mv t4, t3
; RV32ZBB-NEXT: .LBB11_6:
-; RV32ZBB-NEXT: mv t4, t2
+; RV32ZBB-NEXT: mv t3, t2
; RV32ZBB-NEXT: beq a2, a4, .LBB11_8
; RV32ZBB-NEXT: # %bb.7:
-; RV32ZBB-NEXT: mv t4, t5
+; RV32ZBB-NEXT: mv t3, t5
; RV32ZBB-NEXT: .LBB11_8:
; RV32ZBB-NEXT: sltu t5, a3, a1
; RV32ZBB-NEXT: mv t6, t5
@@ -671,17 +671,17 @@ define i128 @abd_ext_i128(i128 %a, i128 %b) nounwind {
; RV32ZBB-NEXT: # %bb.9:
; RV32ZBB-NEXT: sltu t6, a4, a2
; RV32ZBB-NEXT: .LBB11_10:
-; RV32ZBB-NEXT: bnez t3, .LBB11_12
+; RV32ZBB-NEXT: bnez t4, .LBB11_12
; RV32ZBB-NEXT: # %bb.11:
; RV32ZBB-NEXT: sub a7, t1, a7
; RV32ZBB-NEXT: sub a5, a6, a5
; RV32ZBB-NEXT: sub a1, a1, a3
; RV32ZBB-NEXT: sub a2, a2, a4
; RV32ZBB-NEXT: sub a4, a7, t0
-; RV32ZBB-NEXT: sltu a6, a5, t4
+; RV32ZBB-NEXT: sltu a6, a5, t3
; RV32ZBB-NEXT: sub a3, a2, t2
; RV32ZBB-NEXT: sub a2, a4, a6
-; RV32ZBB-NEXT: sub a4, a5, t4
+; RV32ZBB-NEXT: sub a4, a5, t3
; RV32ZBB-NEXT: j .LBB11_13
; RV32ZBB-NEXT: .LBB11_12:
; RV32ZBB-NEXT: sltu t0, a5, a6
@@ -743,17 +743,17 @@ define i128 @abd_ext_i128_undef(i128 %a, i128 %b) nounwind {
; RV32I-NEXT: lw t1, 12(a2)
; RV32I-NEXT: lw a2, 4(a2)
; RV32I-NEXT: sltu t0, a6, a5
-; RV32I-NEXT: mv t4, t0
+; RV32I-NEXT: mv t3, t0
; RV32I-NEXT: beq a7, t1, .LBB12_2
; RV32I-NEXT: # %bb.1:
-; RV32I-NEXT: slt t4, t1, a7
+; RV32I-NEXT: slt t3, t1, a7
; RV32I-NEXT: .LBB12_2:
; RV32I-NEXT: sltu t2, a1, a3
; RV32I-NEXT: sltu t5, a2, a4
-; RV32I-NEXT: mv t3, t2
+; RV32I-NEXT: mv t4, t2
; RV32I-NEXT: beq a4, a2, .LBB12_4
; RV32I-NEXT: # %bb.3:
-; RV32I-NEXT: mv t3, t5
+; RV32I-NEXT: mv t4, t5
; RV32I-NEXT: .LBB12_4:
; RV32I-NEXT: addi sp, sp, -16
; RV32I-NEXT: sw s0, 12(sp) # 4-byte Folded Spill
@@ -762,12 +762,12 @@ define i128 @abd_ext_i128_undef(i128 %a, i128 %b) nounwind {
; RV32I-NEXT: or t6, s0, t6
; RV32I-NEXT: beqz t6, .LBB12_6
; RV32I-NEXT: # %bb.5:
-; RV32I-NEXT: mv t3, t4
+; RV32I-NEXT: mv t4, t3
; RV32I-NEXT: .LBB12_6:
-; RV32I-NEXT: mv t4, t2
+; RV32I-NEXT: mv t3, t2
; RV32I-NEXT: beq a2, a4, .LBB12_8
; RV32I-NEXT: # %bb.7:
-; RV32I-NEXT: mv t4, t5
+; RV32I-NEXT: mv t3, t5
; RV32I-NEXT: .LBB12_8:
; RV32I-NEXT: sltu t5, a3, a1
; RV32I-NEXT: mv t6, t5
@@ -775,17 +775,17 @@ define i128 @abd_ext_i128_undef(i128 %a, i128 %b) nounwind {
; RV32I-NEXT: # %bb.9:
; RV32I-NEXT: sltu t6, a4, a2
; RV32I-NEXT: .LBB12_10:
-; RV32I-NEXT: bnez t3, .LBB12_12
+; RV32I-NEXT: bnez t4, .LBB12_12
; RV32I-NEXT: # %bb.11:
; RV32I-NEXT: sub a7, t1, a7
; RV32I-NEXT: sub a5, a6, a5
; RV32I-NEXT: sub a1, a1, a3
; RV32I-NEXT: sub a2, a2, a4
; RV32I-NEXT: sub a4, a7, t0
-; RV32I-NEXT: sltu a6, a5, t4
+; RV32I-NEXT: sltu a6, a5, t3
; RV32I-NEXT: sub a3, a2, t2
; RV32I-NEXT: sub a2, a4, a6
-; RV32I-NEXT: sub a4, a5, t4
+; RV32I-NEXT: sub a4, a5, t3
; RV32I-NEXT: j .LBB12_13
; RV32I-NEXT: .LBB12_12:
; RV32I-NEXT: sltu t0, a5, a6
@@ -839,17 +839,17 @@ define i128 @abd_ext_i128_undef(i128 %a, i128 %b) nounwind {
; RV32ZBB-NEXT: lw t1, 12(a2)
; RV32ZBB-NEXT: lw a2, 4(a2)
; RV32ZBB-NEXT: sltu t0, a6, a5
-; RV32ZBB-NEXT: mv t4, t0
+; RV32ZBB-NEXT: mv t3, t0
; RV32ZBB-NEXT: beq a7, t1, .LBB12_2
; RV32ZBB-NEXT: # %bb.1:
-; RV32ZBB-NEXT: slt t4, t1, a7
+; RV32ZBB-NEXT: slt t3, t1, a7
; RV32ZBB-NEXT: .LBB12_2:
; RV32ZBB-NEXT: sltu t2, a1, a3
; RV32ZBB-NEXT: sltu t5, a2, a4
-; RV32ZBB-NEXT: mv t3, t2
+; RV32ZBB-NEXT: mv t4, t2
; RV32ZBB-NEXT: beq a4, a2, .LBB12_4
; RV32ZBB-NEXT: # %bb.3:
-; RV32ZBB-NEXT: mv t3, t5
+; RV32ZBB-NEXT: mv t4, t5
; RV32ZBB-NEXT: .LBB12_4:
; RV32ZBB-NEXT: addi sp, sp, -16
; RV32ZBB-NEXT: sw s0, 12(sp) # 4-byte Folded Spill
@@ -858,12 +858,12 @@ define i128 @abd_ext_i128_undef(i128 %a, i128 %b) nounwind {
; RV32ZBB-NEXT: or t6, s0, t6
; RV32ZBB-NEXT: beqz t6, .LBB12_6
; RV32ZBB-NEXT: # %bb.5:
-; RV32ZBB-NEXT: mv t3, t4
+; RV32ZBB-NEXT: mv t4, t3
; RV32ZBB-NEXT: .LBB12_6:
-; RV32ZBB-NEXT: mv t4, t2
+; RV32ZBB-NEXT: mv t3, t2
; RV32ZBB-NEXT: beq a2, a4, .LBB12_8
; RV32ZBB-NEXT: # %bb.7:
-; RV32ZBB-NEXT: mv t4, t5
+; RV32ZBB-NEXT: mv t3, t5
; RV32ZBB-NEXT: .LBB12_8:
; RV32ZBB-NEXT: sltu t5, a3, a1
; RV32ZBB-NEXT: mv t6, t5
@@ -871,17 +871,17 @@ define i128 @abd_ext_i128_undef(i128 %a, i128 %b) nounwind {
; RV32ZBB-NEXT: # %bb.9:
; RV32ZBB-NEXT: sltu t6, a4, a2
; RV32ZBB-NEXT: .LBB12_10:
-; RV32ZBB-NEXT: bnez t3, .LBB12_12
+; RV32ZBB-NEXT: bnez t4, .LBB12_12
; RV32ZBB-NEXT: # %bb.11:
; RV32ZBB-NEXT: sub a7, t1, a7
; RV32ZBB-NEXT: sub a5, a6, a5
; RV32ZBB-NEXT: sub a1, a1, a3
; RV32ZBB-NEXT: sub a2, a2, a4
; RV32ZBB-NEXT: sub a4, a7, t0
-; RV32ZBB-NEXT: sltu a6, a5, t4
+; RV32ZBB-NEXT: sltu a6, a5, t3
; RV32ZBB-NEXT: sub a3, a2, t2
; RV32ZBB-NEXT: sub a2, a4, a6
-; RV32ZBB-NEXT: sub a4, a5, t4
+; RV32ZBB-NEXT: sub a4, a5, t3
; RV32ZBB-NEXT: j .LBB12_13
; RV32ZBB-NEXT: .LBB12_12:
; RV32ZBB-NEXT: sltu t0, a5, a6
@@ -1132,17 +1132,17 @@ define i128 @abd_minmax_i128(i128 %a, i128 %b) nounwind {
; RV32I-NEXT: lw t1, 12(a2)
; RV32I-NEXT: lw a2, 4(a2)
; RV32I-NEXT: sltu t0, a6, a5
-; RV32I-NEXT: mv t4, t0
+; RV32I-NEXT: mv t3, t0
; RV32I-NEXT: beq a7, t1, .LBB17_2
; RV32I-NEXT: # %bb.1:
-; RV32I-NEXT: slt t4, t1, a7
+; RV32I-NEXT: slt t3, t1, a7
; RV32I-NEXT: .LBB17_2:
; RV32I-NEXT: sltu t2, a1, a3
; RV32I-NEXT: sltu t5, a2, a4
-; RV32I-NEXT: mv t3, t2
+; RV32I-NEXT: mv t4, t2
; RV32I-NEXT: beq a4, a2, .LBB17_4
; RV32I-NEXT: # %bb.3:
-; RV32I-NEXT: mv t3, t5
+; RV32I-NEXT: mv t4, t5
; RV32I-NEXT: .LBB17_4:
; RV32I-NEXT: addi sp, sp, -16
; RV32I-NEXT: sw s0, 12(sp) # 4-byte Folded Spill
@@ -1151,12 +1151,12 @@ define i128 @abd_minmax_i128(i128 %a, i128 %b) nounwind {
; RV32I-NEXT: or t6, s0, t6
; RV32I-NEXT: beqz t6, .LBB17_6
; RV32I-NEXT: # %bb.5:
-; RV32I-NEXT: mv t3, t4
+; RV32I-NEXT: mv t4, t3
; RV32I-NEXT: .LBB17_6:
-; RV32I-NEXT: mv t4, t2
+; RV32I-NEXT: mv t3, t2
; RV32I-NEXT: beq a2, a4, .LBB17_8
; RV32I-NEXT: # %bb.7:
-; RV32I-NEXT: mv t4, t5
+; RV32I-NEXT: mv t3, t5
; RV32I-NEXT: .LBB17_8:
; RV32I-NEXT: sltu t5, a3, a1
; RV32I-NEXT: mv t6, t5
@@ -1164,17 +1164,17 @@ define i128 @abd_minmax_i128(i128 %a, i128 %b) nounwind {
; RV32I-NEXT: # %bb.9:
; RV32I-NEXT: sltu t6, a4, a2
; RV32I-NEXT: .LBB17_10:
-; RV32I-NEXT: bnez t3, .LBB17_12
+; RV32I-NEXT: bnez t4, .LBB17_12
; RV32I-NEXT: # %bb.11:
; RV32I-NEXT: sub a7, t1, a7
; RV32I-NEXT: sub a5, a6, a5
; RV32I-NEXT: sub a1, a1, a3
; RV32I-NEXT: sub a2, a2, a4
; RV32I-NEXT: sub a4, a7, t0
-; RV32I-NEXT: sltu a6, a5, t4
+; RV32I-NEXT: sltu a6, a5, t3
; RV32I-NEXT: sub a3, a2, t2
; RV32I-NEXT: sub a2, a4, a6
-; RV32I-NEXT: sub a4, a5, t4
+; RV32I-NEXT: sub a4, a5, t3
; RV32I-NEXT: j .LBB17_13
; RV32I-NEXT: .LBB17_12:
; RV32I-NEXT: sltu t0, a5, a6
@@ -1228,17 +1228,17 @@ define i128 @abd_minmax_i128(i128 %a, i128 %b) nounwind {
; RV32ZBB-NEXT: lw t1, 12(a2)
; RV32ZBB-NEXT: lw a2, 4(a2)
; RV32ZBB-NEXT: sltu t0, a6, a5
-; RV32ZBB-NEXT: mv t4, t0
+; RV32ZBB-NEXT: mv t3, t0
; RV32ZBB-NEXT: beq a7, t1, .LBB17_2
; RV32ZBB-NEXT: # %bb.1:
-; RV32ZBB-NEXT: slt t4, t1, a7
+; RV32ZBB-NEXT: slt t3, t1, a7
; RV32ZBB-NEXT: .LBB17_2:
; RV32ZBB-NEXT: sltu t2, a1, a3
; RV32ZBB-NEXT: sltu t5, a2, a4
-; RV32ZBB-NEXT: mv t3, t2
+; RV32ZBB-NEXT: mv t4, t2
; RV32ZBB-NEXT: beq a4, a2, .LBB17_4
; RV32ZBB-NEXT: # %bb.3:
-; RV32ZBB-NEXT: mv t3, t5
+; RV32ZBB-NEXT: mv t4, t5
; RV32ZBB-NEXT: .LBB17_4:
; RV32ZBB-NEXT: addi sp, sp, -16
; RV32ZBB-NEXT: sw s0, 12(sp) # 4-byte Folded Spill
@@ -1247,12 +1247,12 @@ define i128 @abd_minmax_i128(i128 %a, i128 %b) nounwind {
; RV32ZBB-NEXT: or t6, s0, t6
; RV32ZBB-NEXT: beqz t6, .LBB17_6
; RV32ZBB-NEXT: # %bb.5:
-; RV32ZBB-NEXT: mv t3, t4
+; RV32ZBB-NEXT: mv t4, t3
; RV32ZBB-NEXT: .LBB17_6:
-; RV32ZBB-NEXT: mv t4, t2
+; RV32ZBB-NEXT: mv t3, t2
; RV32ZBB-NEXT: beq a2, a4, .LBB17_8
; RV32ZBB-NEXT: # %bb.7:
-; RV32ZBB-NEXT: mv t4, t5
+; RV32ZBB-NEXT: mv t3, t5
; RV32ZBB-NEXT: .LBB17_8:
; RV32ZBB-NEXT: sltu t5, a3, a1
; RV32ZBB-NEXT: mv t6, t5
@@ -1260,17 +1260,17 @@ define i128 @abd_minmax_i128(i128 %a, i128 %b) nounwind {
; RV32ZBB-NEXT: # %bb.9:
; RV32ZBB-NEXT: sltu t6, a4, a2
; RV32ZBB-NEXT: .LBB17_10:
-; RV32ZBB-NEXT: bnez t3, .LBB17_12
+; RV32ZBB-NEXT: bnez t4, .LBB17_12
; RV32ZBB-NEXT: # %bb.11:
; RV32ZBB-NEXT: sub a7, t1, a7
; RV32ZBB-NEXT: sub a5, a6, a5
; RV32ZBB-NEXT: sub a1, a1, a3
; RV32ZBB-NEXT: sub a2, a2, a4
; RV32ZBB-NEXT: sub a4, a7, t0
-; RV32ZBB-NEXT: sltu a6, a5, t4
+; RV32ZBB-NEXT: sltu a6, a5, t3
; RV32ZBB-NEXT: sub a3, a2, t2
; RV32ZBB-NEXT: sub a2, a4, a6
-; RV32ZBB-NEXT: sub a4, a5, t4
+; RV32ZBB-NEXT: sub a4, a5, t3
; RV32ZBB-NEXT: j .LBB17_13
; RV32ZBB-NEXT: .LBB17_12:
; RV32ZBB-NEXT: sltu t0, a5, a6
@@ -1523,17 +1523,17 @@ define i128 @abd_cmp_i128(i128 %a, i128 %b) nounwind {
; RV32I-NEXT: lw t1, 12(a2)
; RV32I-NEXT: lw a2, 4(a2)
; RV32I-NEXT: sltu t0, a6, a5
-; RV32I-NEXT: mv t4, t0
+; RV32I-NEXT: mv t3, t0
; RV32I-NEXT: beq a7, t1, .LBB22_2
; RV32I-NEXT: # %bb.1:
-; RV32I-NEXT: slt t4, t1, a7
+; RV32I-NEXT: slt t3, t1, a7
; RV32I-NEXT: .LBB22_2:
; RV32I-NEXT: sltu t2, a1, a3
; RV32I-NEXT: sltu t5, a2, a4
-; RV32I-NEXT: mv t3, t2
+; RV32I-NEXT: mv t4, t2
; RV32I-NEXT: beq a4, a2, .LBB22_4
; RV32I-NEXT: # %bb.3:
-; RV32I-NEXT: mv t3, t5
+; RV32I-NEXT: mv t4, t5
; RV32I-NEXT: .LBB22_4:
; RV32I-NEXT: addi sp, sp, -16
; RV32I-NEXT: sw s0, 12(sp) # 4-byte Folded Spill
@@ -1542,12 +1542,12 @@ define i128 @abd_cmp_i128(i128 %a, i128 %b) nounwind {
; RV32I-NEXT: or t6, s0, t6
; RV32I-NEXT: beqz t6, .LBB22_6
; RV32I-NEXT: # %bb.5:
-; RV32I-NEXT: mv t3, t4
+; RV32I-NEXT: mv t4, t3
; RV32I-NEXT: .LBB22_6:
-; RV32I-NEXT: mv t4, t2
+; RV32I-NEXT: mv t3, t2
; RV32I-NEXT: beq a2, a4, .LBB22_8
; RV32I-NEXT: # %bb.7:
-; RV32I-NEXT: mv t4, t5
+; RV32I-NEXT: mv t3, t5
; RV32I-NEXT: .LBB22_8:
; RV32I-NEXT: sltu t5, a3, a1
; RV32I-NEXT: mv t6, t5
@@ -1555,17 +1555,17 @@ define i128 @abd_cmp_i128(i128 %a, i128 %b) nounwind {
; RV32I-NEXT: # %bb.9:
; RV32I-NEXT: sltu t6, a4, a2
; RV32I-NEXT: .LBB22_10:
-; RV32I-NEXT: bnez t3, .LBB22_12
+; RV32I-NEXT: bnez t4, .LBB22_12
; RV32I-NEXT: # %bb.11:
; RV32I-NEXT: sub a7, t1, a7
; RV32I-NEXT: sub a5, a6, a5
; RV32I-NEXT: sub a1, a1, a3
; RV32I-NEXT: sub a2, a2, a4
; RV32I-NEXT: sub a4, a7, t0
-; RV32I-NEXT: sltu a6, a5, t4
+; RV32I-NEXT: sltu a6, a5, t3
; RV32I-NEXT: sub a3, a2, t2
; RV32I-NEXT: sub a2, a4, a6
-; RV32I-NEXT: sub a4, a5, t4
+; RV32I-NEXT: sub a4, a5, t3
; RV32I-NEXT: j .LBB22_13
; RV32I-NEXT: .LBB22_12:
; RV32I-NEXT: sltu t0, a5, a6
@@ -1619,17 +1619,17 @@ define i128 @abd_cmp_i128(i128 %a, i128 %b) nounwind {
; RV32ZBB-NEXT: lw t1, 12(a2)
; RV32ZBB-NEXT: lw a2, 4(a2)
; RV32ZBB-NEXT: sltu t0, a6, a5
-; RV32ZBB-NEXT: mv t4, t0
+; RV32ZBB-NEXT: mv t3, t0
; RV32ZBB-NEXT: beq a7, t1, .LBB22_2
; RV32ZBB-NEXT: # %bb.1:
-; RV32ZBB-NEXT: slt t4, t1, a7
+; RV32ZBB-NEXT: slt t3, t1, a7
; RV32ZBB-NEXT: .LBB22_2:
; RV32ZBB-NEXT: sltu t2, a1, a3
; RV32ZBB-NEXT: sltu t5, a2, a4
-; RV32ZBB-NEXT: mv t3, t2
+; RV32ZBB-NEXT: mv t4, t2
; RV32ZBB-NEXT: beq a4, a2, .LBB22_4
; RV32ZBB-NEXT: # %bb.3:
-; RV32ZBB-NEXT: mv t3, t5
+; RV32ZBB-NEXT: mv t4, t5
; RV32ZBB-NEXT: .LBB22_4:
; RV32ZBB-NEXT: addi sp, sp, -16
; RV32ZBB-NEXT: sw s0, 12(sp) # 4-byte Folded Spill
@@ -1638,12 +1638,12 @@ define i128 @abd_cmp_i128(i128 %a, i128 %b) nounwind {
; RV32ZBB-NEXT: or t6, s0, t6
; RV32ZBB-NEXT: beqz t6, .LBB22_6
; RV32ZBB-NEXT: # %bb.5:
-; RV32ZBB-NEXT: mv t3, t4
+; RV32ZBB-NEXT: mv t4, t3
; RV32ZBB-NEXT: .LBB22_6:
-; RV32ZBB-NEXT: mv t4, t2
+; RV32ZBB-NEXT: mv t3, t2
; RV32ZBB-NEXT: beq a2, a4, .LBB22_8
; RV32ZBB-NEXT: # %bb.7:
-; RV32ZBB-NEXT: mv t4, t5
+; RV32ZBB-NEXT: mv t3, t5
; RV32ZBB-NEXT: .LBB22_8:
; RV32ZBB-NEXT: sltu t5, a3, a1
; RV32ZBB-NEXT: mv t6, t5
@@ -1651,17 +1651,17 @@ define i128 @abd_cmp_i128(i128 %a, i128 %b) nounwind {
; RV32ZBB-NEXT: # %bb.9:
; RV32ZBB-NEXT: sltu t6, a4, a2
; RV32ZBB-NEXT: .LBB22_10:
-; RV32ZBB-NEXT: bnez t3, .LBB22_12
+; RV32ZBB-NEXT: bnez t4, .LBB22_12
; RV32ZBB-NEXT: # %bb.11:
; RV32ZBB-NEXT: sub a7, t1, a7
; RV32ZBB-NEXT: sub a5, a6, a5
; RV32ZBB-NEXT: sub a1, a1, a3
; RV32ZBB-NEXT: sub a2, a2, a4
; RV32ZBB-NEXT: sub a4, a7, t0
-; RV32ZBB-NEXT: sltu a6, a5, t4
+; RV32ZBB-NEXT: sltu a6, a5, t3
; RV32ZBB-NEXT: sub a3, a2, t2
; RV32ZBB-NEXT: sub a2, a4, a6
-; RV32ZBB-NEXT: sub a4, a5, t4
+; RV32ZBB-NEXT: sub a4, a5, t3
; RV32ZBB-NEXT: j .LBB22_13
; RV32ZBB-NEXT: .LBB22_12:
; RV32ZBB-NEXT: sltu t0, a5, a6
@@ -2546,17 +2546,17 @@ define i128 @abd_select_i128(i128 %a, i128 %b) nounwind {
; RV32I-NEXT: lw t1, 12(a2)
; RV32I-NEXT: lw a2, 4(a2)
; RV32I-NEXT: sltu t0, a6, a5
-; RV32I-NEXT: mv t4, t0
+; RV32I-NEXT: mv t3, t0
; RV32I-NEXT: beq a7, t1, .LBB38_2
; RV32I-NEXT: # %bb.1:
-; RV32I-NEXT: slt t4, t1, a7
+; RV32I-NEXT: slt t3, t1, a7
; RV32I-NEXT: .LBB38_2:
; RV32I-NEXT: sltu t2, a1, a3
; RV32I-NEXT: sltu t5, a2, a4
-; RV32I-NEXT: mv t3, t2
+; RV32I-NEXT: mv t4, t2
; RV32I-NEXT: beq a4, a2, .LBB38_4
; RV32I-NEXT: # %bb.3:
-; RV32I-NEXT: mv t3, t5
+; RV32I-NEXT: mv t4, t5
; RV32I-NEXT: .LBB38_4:
; RV32I-NEXT: addi sp, sp, -16
; RV32I-NEXT: sw s0, 12(sp) # 4-byte Folded Spill
@@ -2565,12 +2565,12 @@ define i128 @abd_select_i128(i128 %a, i128 %b) nounwind {
; RV32I-NEXT: or t6, s0, t6
; RV32I-NEXT: beqz t6, .LBB38_6
; RV32I-NEXT: # %bb.5:
-; RV32I-NEXT: mv t3, t4
+; RV32I-NEXT: mv t4, t3
; RV32I-NEXT: .LBB38_6:
-; RV32I-NEXT: mv t4, t2
+; RV32I-NEXT: mv t3, t2
; RV32I-NEXT: beq a2, a4, .LBB38_8
; RV32I-NEXT: # %bb.7:
-; RV32I-NEXT: mv t4, t5
+; RV32I-NEXT: mv t3, t5
; RV32I-NEXT: .LBB38_8:
; RV32I-NEXT: sltu t5, a3, a1
; RV32I-NEXT: mv t6, t5
@@ -2578,17 +2578,17 @@ define i128 @abd_select_i128(i128 %a, i128 %b) nounwind {
; RV32I-NEXT: # %bb.9:
; RV32I-NEXT: sltu t6, a4, a2
; RV32I-NEXT: .LBB38_10:
-; RV32I-NEXT: bnez t3, .LBB38_12
+; RV32I-NEXT: bnez t4, .LBB38_12
; RV32I-NEXT: # %bb.11:
; RV32I-NEXT: sub a7, t1, a7
; RV32I-NEXT: sub a5, a6, a5
; RV32I-NEXT: sub a1, a1, a3
; RV32I-NEXT: sub a2, a2, a4
; RV32I-NEXT: sub a4, a7, t0
-; RV32I-NEXT: sltu a6, a5, t4
+; RV32I-NEXT: sltu a6, a5, t3
; RV32I-NEXT: sub a3, a2, t2
; RV32I-NEXT: sub a2, a4, a6
-; RV32I-NEXT: sub a4, a5, t4
+; RV32I-NEXT: sub a4, a5, t3
; RV32I-NEXT: j .LBB38_13
; RV32I-NEXT: .LBB38_12:
; RV32I-NEXT: sltu t0, a5, a6
@@ -2642,17 +2642,17 @@ define i128 @abd_select_i128(i128 %a, i128 %b) nounwind {
; RV32ZBB-NEXT: lw t1, 12(a2)
; RV32ZBB-NEXT: lw a2, 4(a2)
; RV32ZBB-NEXT: sltu t0, a6, a5
-; RV32ZBB-NEXT: mv t4, t0
+; RV32ZBB-NEXT: mv t3, t0
; RV32ZBB-NEXT: beq a7, t1, .LBB38_2
; RV32ZBB-NEXT: # %bb.1:
-; RV32ZBB-NEXT: slt t4, t1, a7
+; RV32ZBB-NEXT: slt t3, t1, a7
; RV32ZBB-NEXT: .LBB38_2:
; RV32ZBB-NEXT: sltu t2, a1, a3
; RV32ZBB-NEXT: sltu t5, a2, a4
-; RV32ZBB-NEXT: mv t3, t2
+; RV32ZBB-NEXT: mv t4, t2
; RV32ZBB-NEXT: beq a4, a2, .LBB38_4
; RV32ZBB-NEXT: # %bb.3:
-; RV32ZBB-NEXT: mv t3, t5
+; RV32ZBB-NEXT: mv t4, t5
; RV32ZBB-NEXT: .LBB38_4:
; RV32ZBB-NEXT: addi sp, sp, -16
; RV32ZBB-NEXT: sw s0, 12(sp) # 4-byte Folded Spill
@@ -2661,12 +2661,12 @@ define i128 @abd_select_i128(i128 %a, i128 %b) nounwind {
; RV32ZBB-NEXT: or t6, s0, t6
; RV32ZBB-NEXT: beqz t6, .LBB38_6
; RV32ZBB-NEXT: # %bb.5:
-; RV32ZBB-NEXT: mv t3, t4
+; RV32ZBB-NEXT: mv t4, t3
; RV32ZBB-NEXT: .LBB38_6:
-; RV32ZBB-NEXT: mv t4, t2
+; RV32ZBB-NEXT: mv t3, t2
; RV32ZBB-NEXT: beq a2, a4, .LBB38_8
; RV32ZBB-NEXT: # %bb.7:
-; RV32ZBB-NEXT: mv t4, t5
+; RV32ZBB-NEXT: mv t3, t5
; RV32ZBB-NEXT: .LBB38_8:
; RV32ZBB-NEXT: sltu t5, a3, a1
; RV32ZBB-NEXT: mv t6, t5
@@ -2674,17 +2674,17 @@ define i128 @abd_select_i128(i128 %a, i128 %b) nounwind {
; RV32ZBB-NEXT: # %bb.9:
; RV32ZBB-NEXT: sltu t6, a4, a2
; RV32ZBB-NEXT: .LBB38_10:
-; RV32ZBB-NEXT: bnez t3, .LBB38_12
+; RV32ZBB-NEXT: bnez t4, .LBB38_12
; RV32ZBB-NEXT: # %bb.11:
; RV32ZBB-NEXT: sub a7, t1, a7
; RV32ZBB-NEXT: sub a5, a6, a5
; RV32ZBB-NEXT: sub a1, a1, a3
; RV32ZBB-NEXT: sub a2, a2, a4
; RV32ZBB-NEXT: sub a4, a7, t0
-; RV32ZBB-NEXT: sltu a6, a5, t4
+; RV32ZBB-NEXT: sltu a6, a5, t3
; RV32ZBB-NEXT: sub a3, a2, t2
; RV32ZBB-NEXT: sub a2, a4, a6
-; RV32ZBB-NEXT: sub a4, a5, t4
+; RV32ZBB-NEXT: sub a4, a5, t3
; RV32ZBB-NEXT: j .LBB38_13
; RV32ZBB-NEXT: .LBB38_12:
; RV32ZBB-NEXT: sltu t0, a5, a6
diff --git a/llvm/test/CodeGen/RISCV/machine-pipeliner.ll b/llvm/test/CodeGen/RISCV/machine-pipeliner.ll
index d250098576687..bf81d03ec1352 100644
--- a/llvm/test/CodeGen/RISCV/machine-pipeliner.ll
+++ b/llvm/test/CodeGen/RISCV/machine-pipeliner.ll
@@ -52,18 +52,18 @@ define void @test_pipelined_1(ptr noalias %in, ptr noalias %out, i32 signext %cn
;
; CHECK-PIPELINED-LABEL: test_pipelined_1:
; CHECK-PIPELINED: # %bb.0: # %entry
-; CHECK-PIPELINED-NEXT: blez a2, .LBB1_6
+; CHECK-PIPELINED-NEXT: blez a2, .LBB1_7
; CHECK-PIPELINED-NEXT: # %bb.1: # %for.body.preheader
; CHECK-PIPELINED-NEXT: lw a4, 0(a1)
; CHECK-PIPELINED-NEXT: addi a2, a2, -1
+; CHECK-PIPELINED-NEXT: addi a3, a0, 4
; CHECK-PIPELINED-NEXT: sh2add.uw a6, a2, a1
-; CHECK-PIPELINED-NEXT: addi a2, a0, 4
; CHECK-PIPELINED-NEXT: addi a1, a1, 4
; CHECK-PIPELINED-NEXT: addi a6, a6, 4
; CHECK-PIPELINED-NEXT: beq a1, a6, .LBB1_5
; CHECK-PIPELINED-NEXT: # %bb.2: # %for.body
; CHECK-PIPELINED-NEXT: lw a5, 0(a1)
-; CHECK-PIPELINED-NEXT: addi a3, a2, 4
+; CHECK-PIPELINED-NEXT: addi a2, a3, 4
; CHECK-PIPELINED-NEXT: addi a4, a4, 1
; CHECK-PIPELINED-NEXT: addi a1, a1, 4
; CHECK-PIPELINED-NEXT: beq a1, a6, .LBB1_4
@@ -72,20 +72,22 @@ define void @test_pipelined_1(ptr noalias %in, ptr noalias %out, i32 signext %cn
; CHECK-PIPELINED-NEXT: sw a4, 0(a0)
; CHECK-PIPELINED-NEXT: mv a4, a5
; CHECK-PIPELINED-NEXT: lw a5, 0(a1)
-; CHECK-PIPELINED-NEXT: mv a0, a2
-; CHECK-PIPELINED-NEXT: mv a2, a3
-; CHECK-PIPELINED-NEXT: addi a3, a3, 4
+; CHECK-PIPELINED-NEXT: mv a0, a3
+; CHECK-PIPELINED-NEXT: mv a3, a2
+; CHECK-PIPELINED-NEXT: addi a2, a2, 4
; CHECK-PIPELINED-NEXT: addi a4, a4, 1
; CHECK-PIPELINED-NEXT: addi a1, a1, 4
; CHECK-PIPELINED-NEXT: bne a1, a6, .LBB1_3
; CHECK-PIPELINED-NEXT: .LBB1_4:
; CHECK-PIPELINED-NEXT: sw a4, 0(a0)
-; CHECK-PIPELINED-NEXT: mv a0, a2
-; CHECK-PIPELINED-NEXT: mv a4, a5
+; CHECK-PIPELINED-NEXT: j .LBB1_6
; CHECK-PIPELINED-NEXT: .LBB1_5:
-; CHECK-PIPELINED-NEXT: addi a4, a4, 1
-; CHECK-PIPELINED-NEXT: sw a4, 0(a0)
-; CHECK-PIPELINED-NEXT: .LBB1_6: # %for.end
+; CHECK-PIPELINED-NEXT: mv a3, a0
+; CHECK-PIPELINED-NEXT: mv a5, a4
+; CHECK-PIPELINED-NEXT: .LBB1_6:
+; CHECK-PIPELINED-NEXT: addi a5, a5, 1
+; CHECK-PIPELINED-NEXT: sw a5, 0(a3)
+; CHECK-PIPELINED-NEXT: .LBB1_7: # %for.end
; CHECK-PIPELINED-NEXT: ret
entry:
%cmp = icmp sgt i32 %cnt, 0
diff --git a/llvm/test/CodeGen/RISCV/rvv/fixed-vectors-masked-gather.ll b/llvm/test/CodeGen/RISCV/rvv/fixed-vectors-masked-gather.ll
index 533b8b6864ebc..7fbe60bc9cb34 100644
--- a/llvm/test/CodeGen/RISCV/rvv/fixed-vectors-masked-gather.ll
+++ b/llvm/test/CodeGen/RISCV/rvv/fixed-vectors-masked-gather.ll
@@ -6571,53 +6571,53 @@ define <8 x i64> @mgather_baseidx_v8i64(ptr %base, <8 x i64> %idxs, <8 x i1> %m,
; RV32ZVE32F: # %bb.0:
; RV32ZVE32F-NEXT: lw a4, 32(a2)
; RV32ZVE32F-NEXT: lw a5, 40(a2)
-; RV32ZVE32F-NEXT: lw a6, 48(a2)
-; RV32ZVE32F-NEXT: lw a7, 56(a2)
-; RV32ZVE32F-NEXT: lw t0, 0(a2)
+; RV32ZVE32F-NEXT: lw a7, 48(a2)
+; RV32ZVE32F-NEXT: lw t0, 56(a2)
+; RV32ZVE32F-NEXT: lw a6, 0(a2)
; RV32ZVE32F-NEXT: lw t1, 8(a2)
; RV32ZVE32F-NEXT: lw t2, 16(a2)
; RV32ZVE32F-NEXT: lw a2, 24(a2)
; RV32ZVE32F-NEXT: vsetivli zero, 8, e32, m2, ta, ma
-; RV32ZVE32F-NEXT: vmv.v.x v8, t0
+; RV32ZVE32F-NEXT: vmv.v.x v8, a6
; RV32ZVE32F-NEXT: vsetvli zero, zero, e8, mf2, ta, ma
-; RV32ZVE32F-NEXT: vmv.x.s t0, v0
+; RV32ZVE32F-NEXT: vmv.x.s a6, v0
; RV32ZVE32F-NEXT: vsetvli zero, zero, e32, m2, ta, ma
; RV32ZVE32F-NEXT: vslide1down.vx v8, v8, t1
; RV32ZVE32F-NEXT: vslide1down.vx v8, v8, t2
; RV32ZVE32F-NEXT: vslide1down.vx v8, v8, a2
; RV32ZVE32F-NEXT: vslide1down.vx v8, v8, a4
; RV32ZVE32F-NEXT: vslide1down.vx v8, v8, a5
-; RV32ZVE32F-NEXT: vslide1down.vx v8, v8, a6
; RV32ZVE32F-NEXT: vslide1down.vx v8, v8, a7
+; RV32ZVE32F-NEXT: vslide1down.vx v8, v8, t0
; RV32ZVE32F-NEXT: vsll.vi v8, v8, 3
-; RV32ZVE32F-NEXT: andi a2, t0, 1
+; RV32ZVE32F-NEXT: andi a2, a6, 1
; RV32ZVE32F-NEXT: vadd.vx v8, v8, a1
; RV32ZVE32F-NEXT: beqz a2, .LBB57_7
; RV32ZVE32F-NEXT: # %bb.1: # %cond.load
; RV32ZVE32F-NEXT: vmv.x.s a2, v8
; RV32ZVE32F-NEXT: lw a1, 0(a2)
; RV32ZVE32F-NEXT: lw a2, 4(a2)
-; RV32ZVE32F-NEXT: andi a4, t0, 2
+; RV32ZVE32F-NEXT: andi a4, a6, 2
; RV32ZVE32F-NEXT: bnez a4, .LBB57_8
; RV32ZVE32F-NEXT: .LBB57_2:
; RV32ZVE32F-NEXT: lw a4, 8(a3)
; RV32ZVE32F-NEXT: lw a5, 12(a3)
-; RV32ZVE32F-NEXT: andi a6, t0, 4
-; RV32ZVE32F-NEXT: bnez a6, .LBB57_9
+; RV32ZVE32F-NEXT: andi a7, a6, 4
+; RV32ZVE32F-NEXT: bnez a7, .LBB57_9
; RV32ZVE32F-NEXT: .LBB57_3:
-; RV32ZVE32F-NEXT: lw a6, 16(a3)
-; RV32ZVE32F-NEXT: lw a7, 20(a3)
-; RV32ZVE32F-NEXT: andi t1, t0, 8
+; RV32ZVE32F-NEXT: lw a7, 16(a3)
+; RV32ZVE32F-NEXT: lw t0, 20(a3)
+; RV32ZVE32F-NEXT: andi t1, a6, 8
; RV32ZVE32F-NEXT: bnez t1, .LBB57_10
; RV32ZVE32F-NEXT: .LBB57_4:
; RV32ZVE32F-NEXT: lw t1, 24(a3)
; RV32ZVE32F-NEXT: lw t2, 28(a3)
-; RV32ZVE32F-NEXT: andi t3, t0, 16
+; RV32ZVE32F-NEXT: andi t3, a6, 16
; RV32ZVE32F-NEXT: bnez t3, .LBB57_11
; RV32ZVE32F-NEXT: .LBB57_5:
; RV32ZVE32F-NEXT: lw t3, 32(a3)
; RV32ZVE32F-NEXT: lw t4, 36(a3)
-; RV32ZVE32F-NEXT: andi t5, t0, 32
+; RV32ZVE32F-NEXT: andi t5, a6, 32
; RV32ZVE32F-NEXT: bnez t5, .LBB57_12
; RV32ZVE32F-NEXT: .LBB57_6:
; RV32ZVE32F-NEXT: lw t5, 40(a3)
@@ -6626,7 +6626,7 @@ define <8 x i64> @mgather_baseidx_v8i64(ptr %base, <8 x i64> %idxs, <8 x i1> %m,
; RV32ZVE32F-NEXT: .LBB57_7:
; RV32ZVE32F-NEXT: lw a1, 0(a3)
; RV32ZVE32F-NEXT: lw a2, 4(a3)
-; RV32ZVE32F-NEXT: andi a4, t0, 2
+; RV32ZVE32F-NEXT: andi a4, a6, 2
; RV32ZVE32F-NEXT: beqz a4, .LBB57_2
; RV32ZVE32F-NEXT: .LBB57_8: # %cond.load1
; RV32ZVE32F-NEXT: vsetivli zero, 1, e32, m1, ta, ma
@@ -6634,15 +6634,15 @@ define <8 x i64> @mgather_baseidx_v8i64(ptr %base, <8 x i64> %idxs, <8 x i1> %m,
; RV32ZVE32F-NEXT: vmv.x.s a5, v10
; RV32ZVE32F-NEXT: lw a4, 0(a5)
; RV32ZVE32F-NEXT: lw a5, 4(a5)
-; RV32ZVE32F-NEXT: andi a6, t0, 4
-; RV32ZVE32F-NEXT: beqz a6, .LBB57_3
+; RV32ZVE32F-NEXT: andi a7, a6, 4
+; RV32ZVE32F-NEXT: beqz a7, .LBB57_3
; RV32ZVE32F-NEXT: .LBB57_9: # %cond.load4
; RV32ZVE32F-NEXT: vsetivli zero, 1, e32, m1, ta, ma
; RV32ZVE32F-NEXT: vslidedown.vi v10, v8, 2
-; RV32ZVE32F-NEXT: vmv.x.s a7, v10
-; RV32ZVE32F-NEXT: lw a6, 0(a7)
-; RV32ZVE32F-NEXT: lw a7, 4(a7)
-; RV32ZVE32F-NEXT: andi t1, t0, 8
+; RV32ZVE32F-NEXT: vmv.x.s t0, v10
+; RV32ZVE32F-NEXT: lw a7, 0(t0)
+; RV32ZVE32F-NEXT: lw t0, 4(t0)
+; RV32ZVE32F-NEXT: andi t1, a6, 8
; RV32ZVE32F-NEXT: beqz t1, .LBB57_4
; RV32ZVE32F-NEXT: .LBB57_10: # %cond.load7
; RV32ZVE32F-NEXT: vsetivli zero, 1, e32, m1, ta, ma
@@ -6650,7 +6650,7 @@ define <8 x i64> @mgather_baseidx_v8i64(ptr %base, <8 x i64> %idxs, <8 x i1> %m,
; RV32ZVE32F-NEXT: vmv.x.s t2, v10
; RV32ZVE32F-NEXT: lw t1, 0(t2)
; RV32ZVE32F-NEXT: lw t2, 4(t2)
-; RV32ZVE32F-NEXT: andi t3, t0, 16
+; RV32ZVE32F-NEXT: andi t3, a6, 16
; RV32ZVE32F-NEXT: beqz t3, .LBB57_5
; RV32ZVE32F-NEXT: .LBB57_11: # %cond.load10
; RV32ZVE32F-NEXT: vsetivli zero, 1, e32, m2, ta, ma
@@ -6658,7 +6658,7 @@ define <8 x i64> @mgather_baseidx_v8i64(ptr %base, <8 x i64> %idxs, <8 x i1> %m,
; RV32ZVE32F-NEXT: vmv.x.s t4, v10
; RV32ZVE32F-NEXT: lw t3, 0(t4)
; RV32ZVE32F-NEXT: lw t4, 4(t4)
-; RV32ZVE32F-NEXT: andi t5, t0, 32
+; RV32ZVE32F-NEXT: andi t5, a6, 32
; RV32ZVE32F-NEXT: beqz t5, .LBB57_6
; RV32ZVE32F-NEXT: .LBB57_12: # %cond.load13
; RV32ZVE32F-NEXT: vsetivli zero, 1, e32, m2, ta, ma
@@ -6673,7 +6673,7 @@ define <8 x i64> @mgather_baseidx_v8i64(ptr %base, <8 x i64> %idxs, <8 x i1> %m,
; RV32ZVE32F-NEXT: sw s1, 8(sp) # 4-byte Folded Spill
; RV32ZVE32F-NEXT: .cfi_offset s0, -4
; RV32ZVE32F-NEXT: .cfi_offset s1, -8
-; RV32ZVE32F-NEXT: andi s0, t0, 64
+; RV32ZVE32F-NEXT: andi s0, a6, 64
; RV32ZVE32F-NEXT: beqz s0, .LBB57_16
; RV32ZVE32F-NEXT: # %bb.14: # %cond.load16
; RV32ZVE32F-NEXT: vsetivli zero, 1, e32, m2, ta, ma
@@ -6681,30 +6681,30 @@ define <8 x i64> @mgather_baseidx_v8i64(ptr %base, <8 x i64> %idxs, <8 x i1> %m,
; RV32ZVE32F-NEXT: vmv.x.s s1, v10
; RV32ZVE32F-NEXT: lw s0, 0(s1)
; RV32ZVE32F-NEXT: lw s1, 4(s1)
-; RV32ZVE32F-NEXT: andi t0, t0, -128
-; RV32ZVE32F-NEXT: bnez t0, .LBB57_17
+; RV32ZVE32F-NEXT: andi a6, a6, -128
+; RV32ZVE32F-NEXT: bnez a6, .LBB57_17
; RV32ZVE32F-NEXT: .LBB57_15:
-; RV32ZVE32F-NEXT: lw t0, 56(a3)
+; RV32ZVE32F-NEXT: lw a6, 56(a3)
; RV32ZVE32F-NEXT: lw a3, 60(a3)
; RV32ZVE32F-NEXT: j .LBB57_18
; RV32ZVE32F-NEXT: .LBB57_16:
; RV32ZVE32F-NEXT: lw s0, 48(a3)
; RV32ZVE32F-NEXT: lw s1, 52(a3)
-; RV32ZVE32F-NEXT: andi t0, t0, -128
-; RV32ZVE32F-NEXT: beqz t0, .LBB57_15
+; RV32ZVE32F-NEXT: andi a6, a6, -128
+; RV32ZVE32F-NEXT: beqz a6, .LBB57_15
; RV32ZVE32F-NEXT: .LBB57_17: # %cond.load19
; RV32ZVE32F-NEXT: vsetivli zero, 1, e32, m2, ta, ma
; RV32ZVE32F-NEXT: vslidedown.vi v8, v8, 7
; RV32ZVE32F-NEXT: vmv.x.s a3, v8
-; RV32ZVE32F-NEXT: lw t0, 0(a3)
+; RV32ZVE32F-NEXT: lw a6, 0(a3)
; RV32ZVE32F-NEXT: lw a3, 4(a3)
; RV32ZVE32F-NEXT: .LBB57_18: # %else20
; RV32ZVE32F-NEXT: sw a1, 0(a0)
; RV32ZVE32F-NEXT: sw a2, 4(a0)
; RV32ZVE32F-NEXT: sw a4, 8(a0)
; RV32ZVE32F-NEXT: sw a5, 12(a0)
-; RV32ZVE32F-NEXT: sw a6, 16(a0)
-; RV32ZVE32F-NEXT: sw a7, 20(a0)
+; RV32ZVE32F-NEXT: sw a7, 16(a0)
+; RV32ZVE32F-NEXT: sw t0, 20(a0)
; RV32ZVE32F-NEXT: sw t1, 24(a0)
; RV32ZVE32F-NEXT: sw t2, 28(a0)
; RV32ZVE32F-NEXT: sw t3, 32(a0)
@@ -6713,7 +6713,7 @@ define <8 x i64> @mgather_baseidx_v8i64(ptr %base, <8 x i64> %idxs, <8 x i1> %m,
; RV32ZVE32F-NEXT: sw t6, 44(a0)
; RV32ZVE32F-NEXT: sw s0, 48(a0)
; RV32ZVE32F-NEXT: sw s1, 52(a0)
-; RV32ZVE32F-NEXT: sw t0, 56(a0)
+; RV32ZVE32F-NEXT: sw a6, 56(a0)
; RV32ZVE32F-NEXT: sw a3, 60(a0)
; RV32ZVE32F-NEXT: lw s0, 12(sp) # 4-byte Folded Reload
; RV32ZVE32F-NEXT: lw s1, 8(sp) # 4-byte Folded Reload
@@ -6726,89 +6726,89 @@ define <8 x i64> @mgather_baseidx_v8i64(ptr %base, <8 x i64> %idxs, <8 x i1> %m,
; RV64ZVE32F-LABEL: mgather_baseidx_v8i64:
; RV64ZVE32F: # %bb.0:
; RV64ZVE32F-NEXT: vsetivli zero, 1, e8, m1, ta, ma
-; RV64ZVE32F-NEXT: vmv.x.s a7, v0
-; RV64ZVE32F-NEXT: andi a4, a7, 1
+; RV64ZVE32F-NEXT: vmv.x.s a6, v0
+; RV64ZVE32F-NEXT: andi a4, a6, 1
; RV64ZVE32F-NEXT: beqz a4, .LBB57_9
; RV64ZVE32F-NEXT: # %bb.1: # %cond.load
; RV64ZVE32F-NEXT: ld a4, 0(a2)
; RV64ZVE32F-NEXT: slli a4, a4, 3
; RV64ZVE32F-NEXT: add a4, a1, a4
; RV64ZVE32F-NEXT: ld a4, 0(a4)
-; RV64ZVE32F-NEXT: andi a5, a7, 2
+; RV64ZVE32F-NEXT: andi a5, a6, 2
; RV64ZVE32F-NEXT: bnez a5, .LBB57_10
; RV64ZVE32F-NEXT: .LBB57_2:
; RV64ZVE32F-NEXT: ld a5, 8(a3)
-; RV64ZVE32F-NEXT: andi a6, a7, 4
-; RV64ZVE32F-NEXT: bnez a6, .LBB57_11
+; RV64ZVE32F-NEXT: andi a7, a6, 4
+; RV64ZVE32F-NEXT: bnez a7, .LBB57_11
; RV64ZVE32F-NEXT: .LBB57_3:
-; RV64ZVE32F-NEXT: ld a6, 16(a3)
-; RV64ZVE32F-NEXT: andi t0, a7, 8
+; RV64ZVE32F-NEXT: ld a7, 16(a3)
+; RV64ZVE32F-NEXT: andi t0, a6, 8
; RV64ZVE32F-NEXT: bnez t0, .LBB57_12
; RV64ZVE32F-NEXT: .LBB57_4:
; RV64ZVE32F-NEXT: ld t0, 24(a3)
-; RV64ZVE32F-NEXT: andi t1, a7, 16
+; RV64ZVE32F-NEXT: andi t1, a6, 16
; RV64ZVE32F-NEXT: bnez t1, .LBB57_13
; RV64ZVE32F-NEXT: .LBB57_5:
; RV64ZVE32F-NEXT: ld t1, 32(a3)
-; RV64ZVE32F-NEXT: andi t2, a7, 32
+; RV64ZVE32F-NEXT: andi t2, a6, 32
; RV64ZVE32F-NEXT: bnez t2, .LBB57_14
; RV64ZVE32F-NEXT: .LBB57_6:
; RV64ZVE32F-NEXT: ld t2, 40(a3)
-; RV64ZVE32F-NEXT: andi t3, a7, 64
+; RV64ZVE32F-NEXT: andi t3, a6, 64
; RV64ZVE32F-NEXT: bnez t3, .LBB57_15
; RV64ZVE32F-NEXT: .LBB57_7:
; RV64ZVE32F-NEXT: ld t3, 48(a3)
-; RV64ZVE32F-NEXT: andi a7, a7, -128
-; RV64ZVE32F-NEXT: bnez a7, .LBB57_16
+; RV64ZVE32F-NEXT: andi a6, a6, -128
+; RV64ZVE32F-NEXT: bnez a6, .LBB57_16
; RV64ZVE32F-NEXT: .LBB57_8:
; RV64ZVE32F-NEXT: ld a1, 56(a3)
; RV64ZVE32F-NEXT: j .LBB57_17
; RV64ZVE32F-NEXT: .LBB57_9:
; RV64ZVE32F-NEXT: ld a4, 0(a3)
-; RV64ZVE32F-NEXT: andi a5, a7, 2
+; RV64ZVE32F-NEXT: andi a5, a6, 2
; RV64ZVE32F-NEXT: beqz a5, .LBB57_2
; RV64ZVE32F-NEXT: .LBB57_10: # %cond.load1
; RV64ZVE32F-NEXT: ld a5, 8(a2)
; RV64ZVE32F-NEXT: slli a5, a5, 3
; RV64ZVE32F-NEXT: add a5, a1, a5
; RV64ZVE32F-NEXT: ld a5, 0(a5)
-; RV64ZVE32F-NEXT: andi a6, a7, 4
-; RV64ZVE32F-NEXT: beqz a6, .LBB57_3
+; RV64ZVE32F-NEXT: andi a7, a6, 4
+; RV64ZVE32F-NEXT: beqz a7, .LBB57_3
; RV64ZVE32F-NEXT: .LBB57_11: # %cond.load4
-; RV64ZVE32F-NEXT: ld a6, 16(a2)
-; RV64ZVE32F-NEXT: slli a6, a6, 3
-; RV64ZVE32F-NEXT: add a6, a1, a6
-; RV64ZVE32F-NEXT: ld a6, 0(a6)
-; RV64ZVE32F-NEXT: andi t0, a7, 8
+; RV64ZVE32F-NEXT: ld a7, 16(a2)
+; RV64ZVE32F-NEXT: slli a7, a7, 3
+; RV64ZVE32F-NEXT: add a7, a1, a7
+; RV64ZVE32F-NEXT: ld a7, 0(a7)
+; RV64ZVE32F-NEXT: andi t0, a6, 8
; RV64ZVE32F-NEXT: beqz t0, .LBB57_4
; RV64ZVE32F-NEXT: .LBB57_12: # %cond.load7
; RV64ZVE32F-NEXT: ld t0, 24(a2)
; RV64ZVE32F-NEXT: slli t0, t0, 3
; RV64ZVE32F-NEXT: add t0, a1, t0
; RV64ZVE32F-NEXT: ld t0, 0(t0)
-; RV64ZVE32F-NEXT: andi t1, a7, 16
+; RV64ZVE32F-NEXT: andi t1, a6, 16
; RV64ZVE32F-NEXT: beqz t1, .LBB57_5
; RV64ZVE32F-NEXT: .LBB57_13: # %cond.load10
; RV64ZVE32F-NEXT: ld t1, 32(a2)
; RV64ZVE32F-NEXT: slli t1, t1, 3
; RV64ZVE32F-NEXT: add t1, a1, t1
; RV64ZVE32F-NEXT: ld t1, 0(t1)
-; RV64ZVE32F-NEXT: andi t2, a7, 32
+; RV64ZVE32F-NEXT: andi t2, a6, 32
; RV64ZVE32F-NEXT: beqz t2, .LBB57_6
; RV64ZVE32F-NEXT: .LBB57_14: # %cond.load13
; RV64ZVE32F-NEXT: ld t2, 40(a2)
; RV64ZVE32F-NEXT: slli t2, t2, 3
; RV64ZVE32F-NEXT: add t2, a1, t2
; RV64ZVE32F-NEXT: ld t2, 0(t2)
-; RV64ZVE32F-NEXT: andi t3, a7, 64
+; RV64ZVE32F-NEXT: andi t3, a6, 64
; RV64ZVE32F-NEXT: beqz t3, .LBB57_7
; RV64ZVE32F-NEXT: .LBB57_15: # %cond.load16
; RV64ZVE32F-NEXT: ld t3, 48(a2)
; RV64ZVE32F-NEXT: slli t3, t3, 3
; RV64ZVE32F-NEXT: add t3, a1, t3
; RV64ZVE32F-NEXT: ld t3, 0(t3)
-; RV64ZVE32F-NEXT: andi a7, a7, -128
-; RV64ZVE32F-NEXT: beqz a7, .LBB57_8
+; RV64ZVE32F-NEXT: andi a6, a6, -128
+; RV64ZVE32F-NEXT: beqz a6, .LBB57_8
; RV64ZVE32F-NEXT: .LBB57_16: # %cond.load19
; RV64ZVE32F-NEXT: ld a2, 56(a2)
; RV64ZVE32F-NEXT: slli a2, a2, 3
@@ -6817,7 +6817,7 @@ define <8 x i64> @mgather_baseidx_v8i64(ptr %base, <8 x i64> %idxs, <8 x i1> %m,
; RV64ZVE32F-NEXT: .LBB57_17: # %else20
; RV64ZVE32F-NEXT: sd a4, 0(a0)
; RV64ZVE32F-NEXT: sd a5, 8(a0)
-; RV64ZVE32F-NEXT: sd a6, 16(a0)
+; RV64ZVE32F-NEXT: sd a7, 16(a0)
; RV64ZVE32F-NEXT: sd t0, 24(a0)
; RV64ZVE32F-NEXT: sd t1, 32(a0)
; RV64ZVE32F-NEXT: sd t2, 40(a0)
diff --git a/llvm/test/CodeGen/RISCV/rvv/fixed-vectors-strided-load-store-asm.ll b/llvm/test/CodeGen/RISCV/rvv/fixed-vectors-strided-load-store-asm.ll
index 07aa05f609c40..55729a680c294 100644
--- a/llvm/test/CodeGen/RISCV/rvv/fixed-vectors-strided-load-store-asm.ll
+++ b/llvm/test/CodeGen/RISCV/rvv/fixed-vectors-strided-load-store-asm.ll
@@ -897,55 +897,54 @@ define void @strided_load_startval_add_with_splat(ptr noalias nocapture %arg, pt
; CHECK-NEXT: beq a2, a3, .LBB14_7
; CHECK-NEXT: # %bb.1: # %bb3
; CHECK-NEXT: li a3, 1023
-; CHECK-NEXT: subw a5, a3, a2
-; CHECK-NEXT: li a6, 31
-; CHECK-NEXT: mv a4, a2
-; CHECK-NEXT: bltu a5, a6, .LBB14_5
+; CHECK-NEXT: subw a4, a3, a2
+; CHECK-NEXT: li a5, 31
+; CHECK-NEXT: bltu a4, a5, .LBB14_5
; CHECK-NEXT: # %bb.2: # %bb9
-; CHECK-NEXT: slli a4, a5, 32
-; CHECK-NEXT: slli t0, a2, 2
-; CHECK-NEXT: add a5, a0, a2
+; CHECK-NEXT: slli a5, a4, 32
+; CHECK-NEXT: slli a7, a2, 2
+; CHECK-NEXT: add a4, a0, a2
; CHECK-NEXT: add a6, a1, a2
; CHECK-NEXT: li t2, 32
-; CHECK-NEXT: srli a4, a4, 32
-; CHECK-NEXT: add t0, a6, t0
-; CHECK-NEXT: addi a6, a4, 1
-; CHECK-NEXT: andi a7, a6, -32
-; CHECK-NEXT: add a4, a7, a2
-; CHECK-NEXT: add a2, a4, a0
+; CHECK-NEXT: srli a5, a5, 32
+; CHECK-NEXT: add a7, a6, a7
+; CHECK-NEXT: addi a5, a5, 1
+; CHECK-NEXT: andi a6, a5, -32
+; CHECK-NEXT: add a2, a6, a2
+; CHECK-NEXT: add t0, a2, a0
; CHECK-NEXT: li t1, 5
; CHECK-NEXT: vsetvli zero, t2, e8, m1, ta, ma
; CHECK-NEXT: .LBB14_3: # %bb15
; CHECK-NEXT: # =>This Inner Loop Header: Depth=1
-; CHECK-NEXT: vlse8.v v8, (t0), t1
-; CHECK-NEXT: vle8.v v9, (a5)
+; CHECK-NEXT: vlse8.v v8, (a7), t1
+; CHECK-NEXT: vle8.v v9, (a4)
; CHECK-NEXT: vadd.vv v8, v9, v8
-; CHECK-NEXT: vse8.v v8, (a5)
-; CHECK-NEXT: addi a5, a5, 32
-; CHECK-NEXT: addi t0, t0, 160
-; CHECK-NEXT: bne a5, a2, .LBB14_3
+; CHECK-NEXT: vse8.v v8, (a4)
+; CHECK-NEXT: addi a4, a4, 32
+; CHECK-NEXT: addi a7, a7, 160
+; CHECK-NEXT: bne a4, t0, .LBB14_3
; CHECK-NEXT: # %bb.4: # %bb30
-; CHECK-NEXT: beq a6, a7, .LBB14_7
+; CHECK-NEXT: beq a5, a6, .LBB14_7
; CHECK-NEXT: .LBB14_5: # %bb32
-; CHECK-NEXT: add a2, a0, a4
-; CHECK-NEXT: slli a5, a4, 2
-; CHECK-NEXT: add a1, a1, a4
-; CHECK-NEXT: subw a3, a3, a4
+; CHECK-NEXT: add a4, a0, a2
+; CHECK-NEXT: slli a5, a2, 2
+; CHECK-NEXT: add a1, a1, a2
+; CHECK-NEXT: subw a3, a3, a2
; CHECK-NEXT: add a1, a1, a5
; CHECK-NEXT: slli a3, a3, 32
; CHECK-NEXT: srli a3, a3, 32
-; CHECK-NEXT: add a0, a4, a0
+; CHECK-NEXT: add a0, a2, a0
; CHECK-NEXT: add a0, a0, a3
; CHECK-NEXT: addi a0, a0, 1
; CHECK-NEXT: .LBB14_6: # %bb35
; CHECK-NEXT: # =>This Inner Loop Header: Depth=1
-; CHECK-NEXT: lbu a3, 0(a1)
-; CHECK-NEXT: lbu a4, 0(a2)
-; CHECK-NEXT: add a3, a4, a3
-; CHECK-NEXT: sb a3, 0(a2)
-; CHECK-NEXT: addi a2, a2, 1
+; CHECK-NEXT: lbu a2, 0(a1)
+; CHECK-NEXT: lbu a3, 0(a4)
+; CHECK-NEXT: add a2, a3, a2
+; CHECK-NEXT: sb a2, 0(a4)
+; CHECK-NEXT: addi a4, a4, 1
; CHECK-NEXT: addi a1, a1, 5
-; CHECK-NEXT: bne a2, a0, .LBB14_6
+; CHECK-NEXT: bne a4, a0, .LBB14_6
; CHECK-NEXT: .LBB14_7: # %bb34
; CHECK-NEXT: ret
bb:
diff --git a/llvm/test/CodeGen/RISCV/rvv/vxrm-insert-out-of-loop.ll b/llvm/test/CodeGen/RISCV/rvv/vxrm-insert-out-of-loop.ll
index 08cab7cd359b9..a9d926b56386a 100644
--- a/llvm/test/CodeGen/RISCV/rvv/vxrm-insert-out-of-loop.ll
+++ b/llvm/test/CodeGen/RISCV/rvv/vxrm-insert-out-of-loop.ll
@@ -29,32 +29,30 @@ define void @test1(ptr nocapture noundef writeonly %dst, i32 noundef signext %i_
; RV32-NEXT: # %bb.3: # %for.cond1.preheader.us.preheader
; RV32-NEXT: li t0, 32
; RV32-NEXT: # %bb.4: # %for.cond1.preheader.us.preheader
-; RV32-NEXT: addi sp, sp, -16
-; RV32-NEXT: .cfi_def_cfa_offset 16
-; RV32-NEXT: sw s0, 12(sp) # 4-byte Folded Spill
-; RV32-NEXT: sw s1, 8(sp) # 4-byte Folded Spill
-; RV32-NEXT: sw s2, 4(sp) # 4-byte Folded Spill
-; RV32-NEXT: .cfi_offset s0, -4
-; RV32-NEXT: .cfi_offset s1, -8
-; RV32-NEXT: .cfi_offset s2, -12
-; RV32-NEXT: .cfi_remember_state
; RV32-NEXT: add t3, a0, t3
; RV32-NEXT: add t4, a2, t4
-; RV32-NEXT: add s0, a4, t5
+; RV32-NEXT: add t5, a4, t5
; RV32-NEXT: bltu t6, t1, .LBB0_6
; RV32-NEXT: # %bb.5: # %for.cond1.preheader.us.preheader
; RV32-NEXT: li t1, 32
; RV32-NEXT: .LBB0_6: # %for.cond1.preheader.us.preheader
; RV32-NEXT: add t3, t3, a6
-; RV32-NEXT: add t5, t4, a6
-; RV32-NEXT: add t4, s0, a6
+; RV32-NEXT: add t6, t4, a6
+; RV32-NEXT: add t4, t5, a6
; RV32-NEXT: j .LBB0_8
; RV32-NEXT: # %bb.7: # %for.cond1.preheader.us.preheader
; RV32-NEXT: mv t1, t0
; RV32-NEXT: .LBB0_8: # %for.cond1.preheader.us.preheader
-; RV32-NEXT: .cfi_restore_state
+; RV32-NEXT: addi sp, sp, -16
+; RV32-NEXT: .cfi_def_cfa_offset 16
+; RV32-NEXT: sw s0, 12(sp) # 4-byte Folded Spill
+; RV32-NEXT: sw s1, 8(sp) # 4-byte Folded Spill
+; RV32-NEXT: sw s2, 4(sp) # 4-byte Folded Spill
+; RV32-NEXT: .cfi_offset s0, -4
+; RV32-NEXT: .cfi_offset s1, -8
+; RV32-NEXT: .cfi_offset s2, -12
; RV32-NEXT: li t0, 0
-; RV32-NEXT: sltu t5, a0, t5
+; RV32-NEXT: sltu t5, a0, t6
; RV32-NEXT: sltu t6, a2, t3
; RV32-NEXT: and t5, t5, t6
; RV32-NEXT: sltu t4, a0, t4
diff --git a/llvm/test/CodeGen/RISCV/xcvbi.ll b/llvm/test/CodeGen/RISCV/xcvbi.ll
index ca2e416e334f0..d5d11585970b0 100644
--- a/llvm/test/CodeGen/RISCV/xcvbi.ll
+++ b/llvm/test/CodeGen/RISCV/xcvbi.ll
@@ -67,14 +67,14 @@ define i32 @select_beqimm_1(i32 %a, i32 %x, i32 %y) {
; CHECK_NOPT: # %bb.0: # %entry
; CHECK_NOPT-NEXT: addi sp, sp, -16
; CHECK_NOPT-NEXT: .cfi_def_cfa_offset 16
-; CHECK_NOPT-NEXT: sw a1, 8(sp) # 4-byte Folded Spill
-; CHECK_NOPT-NEXT: sw a2, 12(sp) # 4-byte Folded Spill
+; CHECK_NOPT-NEXT: sw a2, 8(sp) # 4-byte Folded Spill
+; CHECK_NOPT-NEXT: sw a1, 12(sp) # 4-byte Folded Spill
; CHECK_NOPT-NEXT: cv.beqimm a0, -16, .LBB2_2
; CHECK_NOPT-NEXT: # %bb.1: # %entry
-; CHECK_NOPT-NEXT: lw a0, 8(sp) # 4-byte Folded Reload
-; CHECK_NOPT-NEXT: sw a0, 12(sp) # 4-byte Folded Spill
-; CHECK_NOPT-NEXT: .LBB2_2: # %entry
; CHECK_NOPT-NEXT: lw a0, 12(sp) # 4-byte Folded Reload
+; CHECK_NOPT-NEXT: sw a0, 8(sp) # 4-byte Folded Spill
+; CHECK_NOPT-NEXT: .LBB2_2: # %entry
+; CHECK_NOPT-NEXT: lw a0, 8(sp) # 4-byte Folded Reload
; CHECK_NOPT-NEXT: addi sp, sp, 16
; CHECK_NOPT-NEXT: .cfi_def_cfa_offset 0
; CHECK_NOPT-NEXT: ret
@@ -98,14 +98,14 @@ define i32 @select_beqimm_2(i32 %a, i32 %x, i32 %y) {
; CHECK_NOPT: # %bb.0: # %entry
; CHECK_NOPT-NEXT: addi sp, sp, -16
; CHECK_NOPT-NEXT: .cfi_def_cfa_offset 16
-; CHECK_NOPT-NEXT: sw a1, 8(sp) # 4-byte Folded Spill
-; CHECK_NOPT-NEXT: sw a2, 12(sp) # 4-byte Folded Spill
+; CHECK_NOPT-NEXT: sw a2, 8(sp) # 4-byte Folded Spill
+; CHECK_NOPT-NEXT: sw a1, 12(sp) # 4-byte Folded Spill
; CHECK_NOPT-NEXT: cv.beqimm a0, 0, .LBB3_2
; CHECK_NOPT-NEXT: # %bb.1: # %entry
-; CHECK_NOPT-NEXT: lw a0, 8(sp) # 4-byte Folded Reload
-; CHECK_NOPT-NEXT: sw a0, 12(sp) # 4-byte Folded Spill
-; CHECK_NOPT-NEXT: .LBB3_2: # %entry
; CHECK_NOPT-NEXT: lw a0, 12(sp) # 4-byte Folded Reload
+; CHECK_NOPT-NEXT: sw a0, 8(sp) # 4-byte Folded Spill
+; CHECK_NOPT-NEXT: .LBB3_2: # %entry
+; CHECK_NOPT-NEXT: lw a0, 8(sp) # 4-byte Folded Reload
; CHECK_NOPT-NEXT: addi sp, sp, 16
; CHECK_NOPT-NEXT: .cfi_def_cfa_offset 0
; CHECK_NOPT-NEXT: ret
@@ -129,14 +129,14 @@ define i32 @select_beqimm_3(i32 %a, i32 %x, i32 %y) {
; CHECK_NOPT: # %bb.0: # %entry
; CHECK_NOPT-NEXT: addi sp, sp, -16
; CHECK_NOPT-NEXT: .cfi_def_cfa_offset 16
-; CHECK_NOPT-NEXT: sw a1, 8(sp) # 4-byte Folded Spill
-; CHECK_NOPT-NEXT: sw a2, 12(sp) # 4-byte Folded Spill
+; CHECK_NOPT-NEXT: sw a2, 8(sp) # 4-byte Folded Spill
+; CHECK_NOPT-NEXT: sw a1, 12(sp) # 4-byte Folded Spill
; CHECK_NOPT-NEXT: cv.beqimm a0, 15, .LBB4_2
; CHECK_NOPT-NEXT: # %bb.1: # %entry
-; CHECK_NOPT-NEXT: lw a0, 8(sp) # 4-byte Folded Reload
-; CHECK_NOPT-NEXT: sw a0, 12(sp) # 4-byte Folded Spill
-; CHECK_NOPT-NEXT: .LBB4_2: # %entry
; CHECK_NOPT-NEXT: lw a0, 12(sp) # 4-byte Folded Reload
+; CHECK_NOPT-NEXT: sw a0, 8(sp) # 4-byte Folded Spill
+; CHECK_NOPT-NEXT: .LBB4_2: # %entry
+; CHECK_NOPT-NEXT: lw a0, 8(sp) # 4-byte Folded Reload
; CHECK_NOPT-NEXT: addi sp, sp, 16
; CHECK_NOPT-NEXT: .cfi_def_cfa_offset 0
; CHECK_NOPT-NEXT: ret
@@ -160,15 +160,15 @@ define i32 @select_no_beqimm_1(i32 %a, i32 %x, i32 %y) {
; CHECK_NOPT: # %bb.0: # %entry
; CHECK_NOPT-NEXT: addi sp, sp, -16
; CHECK_NOPT-NEXT: .cfi_def_cfa_offset 16
-; CHECK_NOPT-NEXT: sw a1, 8(sp) # 4-byte Folded Spill
+; CHECK_NOPT-NEXT: sw a2, 8(sp) # 4-byte Folded Spill
+; CHECK_NOPT-NEXT: sw a1, 12(sp) # 4-byte Folded Spill
; CHECK_NOPT-NEXT: li a1, -17
-; CHECK_NOPT-NEXT: sw a2, 12(sp) # 4-byte Folded Spill
; CHECK_NOPT-NEXT: beq a0, a1, .LBB5_2
; CHECK_NOPT-NEXT: # %bb.1: # %entry
-; CHECK_NOPT-NEXT: lw a0, 8(sp) # 4-byte Folded Reload
-; CHECK_NOPT-NEXT: sw a0, 12(sp) # 4-byte Folded Spill
-; CHECK_NOPT-NEXT: .LBB5_2: # %entry
; CHECK_NOPT-NEXT: lw a0, 12(sp) # 4-byte Folded Reload
+; CHECK_NOPT-NEXT: sw a0, 8(sp) # 4-byte Folded Spill
+; CHECK_NOPT-NEXT: .LBB5_2: # %entry
+; CHECK_NOPT-NEXT: lw a0, 8(sp) # 4-byte Folded Reload
; CHECK_NOPT-NEXT: addi sp, sp, 16
; CHECK_NOPT-NEXT: .cfi_def_cfa_offset 0
; CHECK_NOPT-NEXT: ret
@@ -193,15 +193,15 @@ define i32 @select_no_beqimm_2(i32 %a, i32 %x, i32 %y) {
; CHECK_NOPT: # %bb.0: # %entry
; CHECK_NOPT-NEXT: addi sp, sp, -16
; CHECK_NOPT-NEXT: .cfi_def_cfa_offset 16
-; CHECK_NOPT-NEXT: sw a1, 8(sp) # 4-byte Folded Spill
+; CHECK_NOPT-NEXT: sw a2, 8(sp) # 4-byte Folded Spill
+; CHECK_NOPT-NEXT: sw a1, 12(sp) # 4-byte Folded Spill
; CHECK_NOPT-NEXT: li a1, 16
-; CHECK_NOPT-NEXT: sw a2, 12(sp) # 4-byte Folded Spill
; CHECK_NOPT-NEXT: beq a0, a1, .LBB6_2
; CHECK_NOPT-NEXT: # %bb.1: # %entry
-; CHECK_NOPT-NEXT: lw a0, 8(sp) # 4-byte Folded Reload
-; CHECK_NOPT-NEXT: sw a0, 12(sp) # 4-byte Folded Spill
-; CHECK_NOPT-NEXT: .LBB6_2: # %entry
; CHECK_NOPT-NEXT: lw a0, 12(sp) # 4-byte Folded Reload
+; CHECK_NOPT-NEXT: sw a0, 8(sp) # 4-byte Folded Spill
+; CHECK_NOPT-NEXT: .LBB6_2: # %entry
+; CHECK_NOPT-NEXT: lw a0, 8(sp) # 4-byte Folded Reload
; CHECK_NOPT-NEXT: addi sp, sp, 16
; CHECK_NOPT-NEXT: .cfi_def_cfa_offset 0
; CHECK_NOPT-NEXT: ret
@@ -226,14 +226,14 @@ define i32 @select_bneimm_1(i32 %a, i32 %x, i32 %y) {
; CHECK_NOPT: # %bb.0: # %entry
; CHECK_NOPT-NEXT: addi sp, sp, -16
; CHECK_NOPT-NEXT: .cfi_def_cfa_offset 16
-; CHECK_NOPT-NEXT: sw a1, 8(sp) # 4-byte Folded Spill
-; CHECK_NOPT-NEXT: sw a2, 12(sp) # 4-byte Folded Spill
+; CHECK_NOPT-NEXT: sw a2, 8(sp) # 4-byte Folded Spill
+; CHECK_NOPT-NEXT: sw a1, 12(sp) # 4-byte Folded Spill
; CHECK_NOPT-NEXT: cv.bneimm a0, 0, .LBB7_2
; CHECK_NOPT-NEXT: # %bb.1: # %entry
-; CHECK_NOPT-NEXT: lw a0, 8(sp) # 4-byte Folded Reload
-; CHECK_NOPT-NEXT: sw a0, 12(sp) # 4-byte Folded Spill
-; CHECK_NOPT-NEXT: .LBB7_2: # %entry
; CHECK_NOPT-NEXT: lw a0, 12(sp) # 4-byte Folded Reload
+; CHECK_NOPT-NEXT: sw a0, 8(sp) # 4-byte Folded Spill
+; CHECK_NOPT-NEXT: .LBB7_2: # %entry
+; CHECK_NOPT-NEXT: lw a0, 8(sp) # 4-byte Folded Reload
; CHECK_NOPT-NEXT: addi sp, sp, 16
; CHECK_NOPT-NEXT: .cfi_def_cfa_offset 0
; CHECK_NOPT-NEXT: ret
diff --git a/llvm/test/CodeGen/Thumb2/LowOverheadLoops/mve-tail-data-types.ll b/llvm/test/CodeGen/Thumb2/LowOverheadLoops/mve-tail-data-types.ll
index 8a5a15a57912c..08b99c67d9d55 100644
--- a/llvm/test/CodeGen/Thumb2/LowOverheadLoops/mve-tail-data-types.ll
+++ b/llvm/test/CodeGen/Thumb2/LowOverheadLoops/mve-tail-data-types.ll
@@ -328,14 +328,14 @@ define arm_aapcs_vfpcc void @test_vec_mul_scalar_add_char(ptr nocapture readonly
; CHECK-LABEL: test_vec_mul_scalar_add_char:
; CHECK: @ %bb.0: @ %entry
; CHECK-NEXT: push.w {r4, r5, r6, r7, r8, r9, lr}
-; CHECK-NEXT: ldr r4, [sp, #28]
-; CHECK-NEXT: cmp r4, #0
+; CHECK-NEXT: ldr.w r12, [sp, #28]
+; CHECK-NEXT: cmp.w r12, #0
; CHECK-NEXT: beq.w .LBB5_11
; CHECK-NEXT: @ %bb.1: @ %for.body.lr.ph
-; CHECK-NEXT: adds r7, r1, r4
-; CHECK-NEXT: add.w r6, r3, r4, lsl #2
+; CHECK-NEXT: add.w r7, r1, r12
+; CHECK-NEXT: add.w r6, r3, r12, lsl #2
; CHECK-NEXT: cmp r7, r3
-; CHECK-NEXT: add.w r5, r0, r4
+; CHECK-NEXT: add.w r5, r0, r12
; CHECK-NEXT: cset r7, hi
; CHECK-NEXT: cmp r6, r1
; CHECK-NEXT: csel r7, zr, r7, ls
@@ -348,15 +348,15 @@ define arm_aapcs_vfpcc void @test_vec_mul_scalar_add_char(ptr nocapture readonly
; CHECK-NEXT: cmpeq r7, #0
; CHECK-NEXT: beq .LBB5_4
; CHECK-NEXT: @ %bb.2: @ %for.body.preheader
-; CHECK-NEXT: and r12, r4, #3
-; CHECK-NEXT: subs r7, r4, #1
+; CHECK-NEXT: and r8, r12, #3
+; CHECK-NEXT: sub.w r7, r12, #1
; CHECK-NEXT: cmp r7, #3
; CHECK-NEXT: bhs .LBB5_6
; CHECK-NEXT: @ %bb.3:
-; CHECK-NEXT: mov.w r8, #0
+; CHECK-NEXT: mov.w r12, #0
; CHECK-NEXT: b .LBB5_8
; CHECK-NEXT: .LBB5_4: @ %vector.ph
-; CHECK-NEXT: dlstp.32 lr, r4
+; CHECK-NEXT: dlstp.32 lr, r12
; CHECK-NEXT: .LBB5_5: @ %vector.body
; CHECK-NEXT: @ =>This Inner Loop Header: Depth=1
; CHECK-NEXT: vldrb.u32 q0, [r0], #4
@@ -366,18 +366,18 @@ define arm_aapcs_vfpcc void @test_vec_mul_scalar_add_char(ptr nocapture readonly
; CHECK-NEXT: letp lr, .LBB5_5
; CHECK-NEXT: b .LBB5_11
; CHECK-NEXT: .LBB5_6: @ %for.body.preheader.new
-; CHECK-NEXT: bic r7, r4, #3
+; CHECK-NEXT: bic r7, r12, #3
; CHECK-NEXT: movs r6, #1
; CHECK-NEXT: subs r7, #4
; CHECK-NEXT: add.w r5, r3, #8
-; CHECK-NEXT: mov.w r8, #0
+; CHECK-NEXT: mov.w r12, #0
; CHECK-NEXT: add.w lr, r6, r7, lsr #2
; CHECK-NEXT: adds r6, r0, #3
; CHECK-NEXT: adds r7, r1, #1
; CHECK-NEXT: .LBB5_7: @ %for.body
; CHECK-NEXT: @ =>This Inner Loop Header: Depth=1
; CHECK-NEXT: ldrb r9, [r6, #-3]
-; CHECK-NEXT: add.w r8, r8, #4
+; CHECK-NEXT: add.w r12, r12, #4
; CHECK-NEXT: ldrb r4, [r7, #-1]
; CHECK-NEXT: smlabb r4, r4, r9, r2
; CHECK-NEXT: str r4, [r5, #-8]
@@ -396,11 +396,11 @@ define arm_aapcs_vfpcc void @test_vec_mul_scalar_add_char(ptr nocapture readonly
; CHECK-NEXT: adds r5, #16
; CHECK-NEXT: le lr, .LBB5_7
; CHECK-NEXT: .LBB5_8: @ %for.cond.cleanup.loopexit.unr-lcssa
-; CHECK-NEXT: wls lr, r12, .LBB5_11
+; CHECK-NEXT: wls lr, r8, .LBB5_11
; CHECK-NEXT: @ %bb.9: @ %for.body.epil.preheader
-; CHECK-NEXT: add r0, r8
-; CHECK-NEXT: add r1, r8
-; CHECK-NEXT: add.w r3, r3, r8, lsl #2
+; CHECK-NEXT: add r0, r12
+; CHECK-NEXT: add r1, r12
+; CHECK-NEXT: add.w r3, r3, r12, lsl #2
; CHECK-NEXT: .LBB5_10: @ %for.body.epil
; CHECK-NEXT: @ =>This Inner Loop Header: Depth=1
; CHECK-NEXT: ldrb r7, [r0], #1
@@ -604,14 +604,14 @@ define arm_aapcs_vfpcc void @test_vec_mul_scalar_add_uchar(ptr nocapture readonl
; CHECK-LABEL: test_vec_mul_scalar_add_uchar:
; CHECK: @ %bb.0: @ %entry
; CHECK-NEXT: push.w {r4, r5, r6, r7, r8, r9, lr}
-; CHECK-NEXT: ldr r4, [sp, #28]
-; CHECK-NEXT: cmp r4, #0
+; CHECK-NEXT: ldr.w r12, [sp, #28]
+; CHECK-NEXT: cmp.w r12, #0
; CHECK-NEXT: beq.w .LBB7_11
; CHECK-NEXT: @ %bb.1: @ %for.body.lr.ph
-; CHECK-NEXT: adds r7, r1, r4
-; CHECK-NEXT: add.w r6, r3, r4, lsl #2
+; CHECK-NEXT: add.w r7, r1, r12
+; CHECK-NEXT: add.w r6, r3, r12, lsl #2
; CHECK-NEXT: cmp r7, r3
-; CHECK-NEXT: add.w r5, r0, r4
+; CHECK-NEXT: add.w r5, r0, r12
; CHECK-NEXT: cset r7, hi
; CHECK-NEXT: cmp r6, r1
; CHECK-NEXT: csel r7, zr, r7, ls
@@ -624,15 +624,15 @@ define arm_aapcs_vfpcc void @test_vec_mul_scalar_add_uchar(ptr nocapture readonl
; CHECK-NEXT: cmpeq r7, #0
; CHECK-NEXT: beq .LBB7_4
; CHECK-NEXT: @ %bb.2: @ %for.body.preheader
-; CHECK-NEXT: and r12, r4, #3
-; CHECK-NEXT: subs r7, r4, #1
+; CHECK-NEXT: and r8, r12, #3
+; CHECK-NEXT: sub.w r7, r12, #1
; CHECK-NEXT: cmp r7, #3
; CHECK-NEXT: bhs .LBB7_6
; CHECK-NEXT: @ %bb.3:
-; CHECK-NEXT: mov.w r8, #0
+; CHECK-NEXT: mov.w r12, #0
; CHECK-NEXT: b .LBB7_8
; CHECK-NEXT: .LBB7_4: @ %vector.ph
-; CHECK-NEXT: dlstp.32 lr, r4
+; CHECK-NEXT: dlstp.32 lr, r12
; CHECK-NEXT: .LBB7_5: @ %vector.body
; CHECK-NEXT: @ =>This Inner Loop Header: Depth=1
; CHECK-NEXT: vldrb.u32 q0, [r0], #4
@@ -642,18 +642,18 @@ define arm_aapcs_vfpcc void @test_vec_mul_scalar_add_uchar(ptr nocapture readonl
; CHECK-NEXT: letp lr, .LBB7_5
; CHECK-NEXT: b .LBB7_11
; CHECK-NEXT: .LBB7_6: @ %for.body.preheader.new
-; CHECK-NEXT: bic r7, r4, #3
+; CHECK-NEXT: bic r7, r12, #3
; CHECK-NEXT: movs r6, #1
; CHECK-NEXT: subs r7, #4
; CHECK-NEXT: add.w r5, r3, #8
-; CHECK-NEXT: mov.w r8, #0
+; CHECK-NEXT: mov.w r12, #0
; CHECK-NEXT: add.w lr, r6, r7, lsr #2
; CHECK-NEXT: adds r6, r0, #3
; CHECK-NEXT: adds r7, r1, #1
; CHECK-NEXT: .LBB7_7: @ %for.body
; CHECK-NEXT: @ =>This Inner Loop Header: Depth=1
; CHECK-NEXT: ldrb r9, [r6, #-3]
-; CHECK-NEXT: add.w r8, r8, #4
+; CHECK-NEXT: add.w r12, r12, #4
; CHECK-NEXT: ldrb r4, [r7, #-1]
; CHECK-NEXT: smlabb r4, r4, r9, r2
; CHECK-NEXT: str r4, [r5, #-8]
@@ -672,11 +672,11 @@ define arm_aapcs_vfpcc void @test_vec_mul_scalar_add_uchar(ptr nocapture readonl
; CHECK-NEXT: adds r5, #16
; CHECK-NEXT: le lr, .LBB7_7
; CHECK-NEXT: .LBB7_8: @ %for.cond.cleanup.loopexit.unr-lcssa
-; CHECK-NEXT: wls lr, r12, .LBB7_11
+; CHECK-NEXT: wls lr, r8, .LBB7_11
; CHECK-NEXT: @ %bb.9: @ %for.body.epil.preheader
-; CHECK-NEXT: add r0, r8
-; CHECK-NEXT: add r1, r8
-; CHECK-NEXT: add.w r3, r3, r8, lsl #2
+; CHECK-NEXT: add r0, r12
+; CHECK-NEXT: add r1, r12
+; CHECK-NEXT: add.w r3, r3, r12, lsl #2
; CHECK-NEXT: .LBB7_10: @ %for.body.epil
; CHECK-NEXT: @ =>This Inner Loop Header: Depth=1
; CHECK-NEXT: ldrb r7, [r0], #1
@@ -880,14 +880,14 @@ define arm_aapcs_vfpcc void @test_vec_mul_scalar_add_int(ptr nocapture readonly
; CHECK-LABEL: test_vec_mul_scalar_add_int:
; CHECK: @ %bb.0: @ %entry
; CHECK-NEXT: push.w {r4, r5, r6, r7, r8, r9, lr}
-; CHECK-NEXT: ldr r4, [sp, #28]
-; CHECK-NEXT: cmp r4, #0
+; CHECK-NEXT: ldr.w r12, [sp, #28]
+; CHECK-NEXT: cmp.w r12, #0
; CHECK-NEXT: beq.w .LBB9_11
; CHECK-NEXT: @ %bb.1: @ %vector.memcheck
-; CHECK-NEXT: add.w r7, r1, r4, lsl #2
-; CHECK-NEXT: add.w r6, r3, r4, lsl #2
+; CHECK-NEXT: add.w r7, r1, r12, lsl #2
+; CHECK-NEXT: add.w r6, r3, r12, lsl #2
; CHECK-NEXT: cmp r7, r3
-; CHECK-NEXT: add.w r5, r0, r4, lsl #2
+; CHECK-NEXT: add.w r5, r0, r12, lsl #2
; CHECK-NEXT: cset r7, hi
; CHECK-NEXT: cmp r6, r1
; CHECK-NEXT: csel r7, zr, r7, ls
@@ -900,15 +900,15 @@ define arm_aapcs_vfpcc void @test_vec_mul_scalar_add_int(ptr nocapture readonly
; CHECK-NEXT: cmpeq r7, #0
; CHECK-NEXT: beq .LBB9_4
; CHECK-NEXT: @ %bb.2: @ %for.body.preheader
-; CHECK-NEXT: and r12, r4, #3
-; CHECK-NEXT: subs r7, r4, #1
+; CHECK-NEXT: and r8, r12, #3
+; CHECK-NEXT: sub.w r7, r12, #1
; CHECK-NEXT: cmp r7, #3
; CHECK-NEXT: bhs .LBB9_6
; CHECK-NEXT: @ %bb.3:
-; CHECK-NEXT: mov.w r8, #0
+; CHECK-NEXT: mov.w r12, #0
; CHECK-NEXT: b .LBB9_8
; CHECK-NEXT: .LBB9_4: @ %vector.ph
-; CHECK-NEXT: dlstp.32 lr, r4
+; CHECK-NEXT: dlstp.32 lr, r12
; CHECK-NEXT: .LBB9_5: @ %vector.body
; CHECK-NEXT: @ =>This Inner Loop Header: Depth=1
; CHECK-NEXT: vldrw.u32 q0, [r0], #16
@@ -918,18 +918,18 @@ define arm_aapcs_vfpcc void @test_vec_mul_scalar_add_int(ptr nocapture readonly
; CHECK-NEXT: letp lr, .LBB9_5
; CHECK-NEXT: b .LBB9_11
; CHECK-NEXT: .LBB9_6: @ %for.body.preheader.new
-; CHECK-NEXT: bic r7, r4, #3
+; CHECK-NEXT: bic r7, r12, #3
; CHECK-NEXT: movs r6, #1
; CHECK-NEXT: subs r7, #4
; CHECK-NEXT: add.w r5, r3, #8
-; CHECK-NEXT: mov.w r8, #0
+; CHECK-NEXT: mov.w r12, #0
; CHECK-NEXT: add.w lr, r6, r7, lsr #2
; CHECK-NEXT: add.w r6, r0, #8
; CHECK-NEXT: add.w r7, r1, #8
; CHECK-NEXT: .LBB9_7: @ %for.body
; CHECK-NEXT: @ =>This Inner Loop Header: Depth=1
; CHECK-NEXT: ldr r9, [r6, #-8]
-; CHECK-NEXT: add.w r8, r8, #4
+; CHECK-NEXT: add.w r12, r12, #4
; CHECK-NEXT: ldr r4, [r7, #-8]
; CHECK-NEXT: mla r4, r4, r9, r2
; CHECK-NEXT: str r4, [r5, #-8]
@@ -950,11 +950,11 @@ define arm_aapcs_vfpcc void @test_vec_mul_scalar_add_int(ptr nocapture readonly
; CHECK-NEXT: adds r5, #16
; CHECK-NEXT: le lr, .LBB9_7
; CHECK-NEXT: .LBB9_8: @ %for.cond.cleanup.loopexit.unr-lcssa
-; CHECK-NEXT: wls lr, r12, .LBB9_11
+; CHECK-NEXT: wls lr, r8, .LBB9_11
; CHECK-NEXT: @ %bb.9: @ %for.body.epil.preheader
-; CHECK-NEXT: add.w r0, r0, r8, lsl #2
-; CHECK-NEXT: add.w r1, r1, r8, lsl #2
-; CHECK-NEXT: add.w r3, r3, r8, lsl #2
+; CHECK-NEXT: add.w r0, r0, r12, lsl #2
+; CHECK-NEXT: add.w r1, r1, r12, lsl #2
+; CHECK-NEXT: add.w r3, r3, r12, lsl #2
; CHECK-NEXT: .LBB9_10: @ %for.body.epil
; CHECK-NEXT: @ =>This Inner Loop Header: Depth=1
; CHECK-NEXT: ldr r7, [r0], #4
diff --git a/llvm/test/CodeGen/Thumb2/LowOverheadLoops/tail-pred-disabled-in-loloops.ll b/llvm/test/CodeGen/Thumb2/LowOverheadLoops/tail-pred-disabled-in-loloops.ll
index a0e690212d5a4..7acc83343dcb8 100644
--- a/llvm/test/CodeGen/Thumb2/LowOverheadLoops/tail-pred-disabled-in-loloops.ll
+++ b/llvm/test/CodeGen/Thumb2/LowOverheadLoops/tail-pred-disabled-in-loloops.ll
@@ -17,18 +17,18 @@ define dso_local void @check_option(ptr noalias nocapture %A, ptr noalias nocapt
; ENABLED-NEXT: .LBB0_2: @ %vector.ph
; ENABLED-NEXT: @ =>This Loop Header: Depth=1
; ENABLED-NEXT: @ Child Loop BB0_3 Depth 2
-; ENABLED-NEXT: mov r12, r0
-; ENABLED-NEXT: mov r4, r2
-; ENABLED-NEXT: mov r5, r1
-; ENABLED-NEXT: mov r6, r3
-; ENABLED-NEXT: dlstp.32 lr, r6
+; ENABLED-NEXT: mov r4, r0
+; ENABLED-NEXT: mov r5, r2
+; ENABLED-NEXT: mov r6, r1
+; ENABLED-NEXT: mov r7, r3
+; ENABLED-NEXT: dlstp.32 lr, r7
; ENABLED-NEXT: .LBB0_3: @ %vector.body
; ENABLED-NEXT: @ Parent Loop BB0_2 Depth=1
; ENABLED-NEXT: @ => This Inner Loop Header: Depth=2
-; ENABLED-NEXT: vldrw.u32 q0, [r5], #16
-; ENABLED-NEXT: vldrw.u32 q1, [r4], #16
+; ENABLED-NEXT: vldrw.u32 q0, [r6], #16
+; ENABLED-NEXT: vldrw.u32 q1, [r5], #16
; ENABLED-NEXT: vadd.i32 q0, q1, q0
-; ENABLED-NEXT: vstrw.32 q0, [r12], #16
+; ENABLED-NEXT: vstrw.32 q0, [r4], #16
; ENABLED-NEXT: letp lr, .LBB0_3
; ENABLED-NEXT: b .LBB0_2
; ENABLED-NEXT: .LBB0_4: @ %for.cond.cleanup
@@ -44,29 +44,29 @@ define dso_local void @check_option(ptr noalias nocapture %A, ptr noalias nocapt
; DISABLED-NEXT: movs r6, #1
; DISABLED-NEXT: bic r7, r7, #3
; DISABLED-NEXT: subs r7, #4
-; DISABLED-NEXT: add.w r8, r6, r7, lsr #2
+; DISABLED-NEXT: add.w r12, r6, r7, lsr #2
; DISABLED-NEXT: .LBB0_2: @ %vector.ph
; DISABLED-NEXT: @ =>This Loop Header: Depth=1
; DISABLED-NEXT: @ Child Loop BB0_3 Depth 2
-; DISABLED-NEXT: mov r7, r8
-; DISABLED-NEXT: mov r12, r0
-; DISABLED-NEXT: mov r4, r2
-; DISABLED-NEXT: mov r5, r1
-; DISABLED-NEXT: mov r6, r3
-; DISABLED-NEXT: dls lr, r8
+; DISABLED-NEXT: mov r8, r12
+; DISABLED-NEXT: mov r4, r0
+; DISABLED-NEXT: mov r5, r2
+; DISABLED-NEXT: mov r6, r1
+; DISABLED-NEXT: mov r7, r3
+; DISABLED-NEXT: dls lr, r12
; DISABLED-NEXT: .LBB0_3: @ %vector.body
; DISABLED-NEXT: @ Parent Loop BB0_2 Depth=1
; DISABLED-NEXT: @ => This Inner Loop Header: Depth=2
-; DISABLED-NEXT: vctp.32 r6
-; DISABLED-NEXT: mov lr, r7
+; DISABLED-NEXT: vctp.32 r7
+; DISABLED-NEXT: mov lr, r8
; DISABLED-NEXT: vpstt
-; DISABLED-NEXT: vldrwt.u32 q0, [r5], #16
-; DISABLED-NEXT: vldrwt.u32 q1, [r4], #16
-; DISABLED-NEXT: subs r7, #1
-; DISABLED-NEXT: subs r6, #4
+; DISABLED-NEXT: vldrwt.u32 q0, [r6], #16
+; DISABLED-NEXT: vldrwt.u32 q1, [r5], #16
+; DISABLED-NEXT: sub.w r8, r8, #1
+; DISABLED-NEXT: subs r7, #4
; DISABLED-NEXT: vadd.i32 q0, q1, q0
; DISABLED-NEXT: vpst
-; DISABLED-NEXT: vstrwt.32 q0, [r12], #16
+; DISABLED-NEXT: vstrwt.32 q0, [r4], #16
; DISABLED-NEXT: le lr, .LBB0_3
; DISABLED-NEXT: b .LBB0_2
; DISABLED-NEXT: .LBB0_4: @ %for.cond.cleanup
diff --git a/llvm/test/CodeGen/Thumb2/LowOverheadLoops/varying-outer-2d-reduction.ll b/llvm/test/CodeGen/Thumb2/LowOverheadLoops/varying-outer-2d-reduction.ll
index 07c06e10979cd..736d5956b6194 100644
--- a/llvm/test/CodeGen/Thumb2/LowOverheadLoops/varying-outer-2d-reduction.ll
+++ b/llvm/test/CodeGen/Thumb2/LowOverheadLoops/varying-outer-2d-reduction.ll
@@ -29,7 +29,7 @@ define dso_local void @varying_outer_2d_reduction(ptr nocapture readonly %Input,
; ENABLED-NEXT: mov.w r8, #0
; ENABLED-NEXT: mov r9, r12
; ENABLED-NEXT: uxth r0, r0
-; ENABLED-NEXT: rsbs r5, r0, #0
+; ENABLED-NEXT: rsbs r6, r0, #0
; ENABLED-NEXT: b .LBB0_4
; ENABLED-NEXT: .LBB0_2: @ in Loop: Header=BB0_4 Depth=1
; ENABLED-NEXT: movs r0, #0
@@ -52,9 +52,9 @@ define dso_local void @varying_outer_2d_reduction(ptr nocapture readonly %Input,
; ENABLED-NEXT: bic r0, r9, #3
; ENABLED-NEXT: movs r7, #1
; ENABLED-NEXT: subs r0, #4
-; ENABLED-NEXT: sub.w r4, r2, r8
+; ENABLED-NEXT: sub.w r5, r2, r8
; ENABLED-NEXT: vmov.i32 q1, #0x0
-; ENABLED-NEXT: add.w r6, r7, r0, lsr #2
+; ENABLED-NEXT: add.w r4, r7, r0, lsr #2
; ENABLED-NEXT: sub.w r0, r12, r8
; ENABLED-NEXT: bic r0, r0, #3
; ENABLED-NEXT: subs r0, #4
@@ -65,16 +65,16 @@ define dso_local void @varying_outer_2d_reduction(ptr nocapture readonly %Input,
; ENABLED-NEXT: .LBB0_6: @ %vector.body
; ENABLED-NEXT: @ Parent Loop BB0_4 Depth=1
; ENABLED-NEXT: @ => This Inner Loop Header: Depth=2
-; ENABLED-NEXT: vctp.32 r4
+; ENABLED-NEXT: vctp.32 r5
; ENABLED-NEXT: vmov q0, q1
; ENABLED-NEXT: vpstt
; ENABLED-NEXT: vldrht.s32 q1, [r0], #8
; ENABLED-NEXT: vldrht.s32 q2, [r7], #8
-; ENABLED-NEXT: mov lr, r6
-; ENABLED-NEXT: subs r6, #1
+; ENABLED-NEXT: mov lr, r4
+; ENABLED-NEXT: subs r4, #1
; ENABLED-NEXT: vmul.i32 q1, q2, q1
-; ENABLED-NEXT: subs r4, #4
-; ENABLED-NEXT: vshl.s32 q1, r5
+; ENABLED-NEXT: subs r5, #4
+; ENABLED-NEXT: vshl.s32 q1, r6
; ENABLED-NEXT: vadd.i32 q1, q1, q0
; ENABLED-NEXT: le lr, .LBB0_6
; ENABLED-NEXT: @ %bb.7: @ %middle.block
@@ -100,7 +100,7 @@ define dso_local void @varying_outer_2d_reduction(ptr nocapture readonly %Input,
; NOREDUCTIONS-NEXT: mov.w r8, #0
; NOREDUCTIONS-NEXT: mov r9, r12
; NOREDUCTIONS-NEXT: uxth r0, r0
-; NOREDUCTIONS-NEXT: rsbs r5, r0, #0
+; NOREDUCTIONS-NEXT: rsbs r6, r0, #0
; NOREDUCTIONS-NEXT: b .LBB0_4
; NOREDUCTIONS-NEXT: .LBB0_2: @ in Loop: Header=BB0_4 Depth=1
; NOREDUCTIONS-NEXT: movs r0, #0
@@ -123,9 +123,9 @@ define dso_local void @varying_outer_2d_reduction(ptr nocapture readonly %Input,
; NOREDUCTIONS-NEXT: bic r0, r9, #3
; NOREDUCTIONS-NEXT: movs r7, #1
; NOREDUCTIONS-NEXT: subs r0, #4
-; NOREDUCTIONS-NEXT: sub.w r4, r2, r8
+; NOREDUCTIONS-NEXT: sub.w r5, r2, r8
; NOREDUCTIONS-NEXT: vmov.i32 q1, #0x0
-; NOREDUCTIONS-NEXT: add.w r6, r7, r0, lsr #2
+; NOREDUCTIONS-NEXT: add.w r4, r7, r0, lsr #2
; NOREDUCTIONS-NEXT: sub.w r0, r12, r8
; NOREDUCTIONS-NEXT: bic r0, r0, #3
; NOREDUCTIONS-NEXT: subs r0, #4
@@ -136,16 +136,16 @@ define dso_local void @varying_outer_2d_reduction(ptr nocapture readonly %Input,
; NOREDUCTIONS-NEXT: .LBB0_6: @ %vector.body
; NOREDUCTIONS-NEXT: @ Parent Loop BB0_4 Depth=1
; NOREDUCTIONS-NEXT: @ => This Inner Loop Header: Depth=2
-; NOREDUCTIONS-NEXT: vctp.32 r4
+; NOREDUCTIONS-NEXT: vctp.32 r5
; NOREDUCTIONS-NEXT: vmov q0, q1
; NOREDUCTIONS-NEXT: vpstt
; NOREDUCTIONS-NEXT: vldrht.s32 q1, [r0], #8
; NOREDUCTIONS-NEXT: vldrht.s32 q2, [r7], #8
-; NOREDUCTIONS-NEXT: mov lr, r6
-; NOREDUCTIONS-NEXT: subs r6, #1
+; NOREDUCTIONS-NEXT: mov lr, r4
+; NOREDUCTIONS-NEXT: subs r4, #1
; NOREDUCTIONS-NEXT: vmul.i32 q1, q2, q1
-; NOREDUCTIONS-NEXT: subs r4, #4
-; NOREDUCTIONS-NEXT: vshl.s32 q1, r5
+; NOREDUCTIONS-NEXT: subs r5, #4
+; NOREDUCTIONS-NEXT: vshl.s32 q1, r6
; NOREDUCTIONS-NEXT: vadd.i32 q1, q1, q0
; NOREDUCTIONS-NEXT: le lr, .LBB0_6
; NOREDUCTIONS-NEXT: @ %bb.7: @ %middle.block
diff --git a/llvm/test/CodeGen/Thumb2/LowOverheadLoops/while-loops.ll b/llvm/test/CodeGen/Thumb2/LowOverheadLoops/while-loops.ll
index cbcbf1f392ce8..a6a9361050731 100644
--- a/llvm/test/CodeGen/Thumb2/LowOverheadLoops/while-loops.ll
+++ b/llvm/test/CodeGen/Thumb2/LowOverheadLoops/while-loops.ll
@@ -165,74 +165,69 @@ define dso_local i32 @b(ptr %c, i32 %d, i32 %e, ptr %n) "frame-pointer"="all" {
; CHECK-NEXT: sub sp, #16
; CHECK-NEXT: wls lr, r1, .LBB2_3
; CHECK-NEXT: @ %bb.1: @ %while.body.preheader
-; CHECK-NEXT: adds r6, r3, #4
-; CHECK-NEXT: adds r1, r0, #4
-; CHECK-NEXT: mvn r8, #1
-; CHECK-NEXT: @ implicit-def: $r9
-; CHECK-NEXT: @ implicit-def: $r4
+; CHECK-NEXT: add.w r12, r3, #4
+; CHECK-NEXT: add.w r9, r0, #4
+; CHECK-NEXT: mvn r10, #1
+; CHECK-NEXT: @ implicit-def: $r6
+; CHECK-NEXT: @ implicit-def: $r8
; CHECK-NEXT: str r2, [sp] @ 4-byte Spill
; CHECK-NEXT: .LBB2_2: @ %while.body
; CHECK-NEXT: @ =>This Inner Loop Header: Depth=1
-; CHECK-NEXT: str r1, [sp, #12] @ 4-byte Spill
-; CHECK-NEXT: asrs r2, r4, #31
-; CHECK-NEXT: ldr r1, [sp, #12] @ 4-byte Reload
-; CHECK-NEXT: ldr r1, [r1]
+; CHECK-NEXT: ldr.w r1, [r9]
+; CHECK-NEXT: asr.w r2, r8, #31
+; CHECK-NEXT: str r3, [sp, #12] @ 4-byte Spill
; CHECK-NEXT: muls r1, r3, r1
-; CHECK-NEXT: adds r4, r4, r1
+; CHECK-NEXT: adds.w r5, r8, r1
; CHECK-NEXT: adc.w r1, r2, r1, asr #31
-; CHECK-NEXT: adds.w r2, r4, #-2147483648
-; CHECK-NEXT: ldrd r2, r4, [r8]
+; CHECK-NEXT: adds.w r2, r5, #-2147483648
+; CHECK-NEXT: ldrd r2, r5, [r10]
+; CHECK-NEXT: adc r8, r1, #0
+; CHECK-NEXT: asr.w r1, r8, #31
+; CHECK-NEXT: strd r6, r2, [sp, #4] @ 8-byte Folded Spill
+; CHECK-NEXT: smull r5, r6, r5, r6
+; CHECK-NEXT: subs.w r5, r8, r5
+; CHECK-NEXT: sbcs r1, r6
+; CHECK-NEXT: adds.w r6, r5, #-2147483648
; CHECK-NEXT: adc r5, r1, #0
-; CHECK-NEXT: str r2, [sp, #4] @ 4-byte Spill
-; CHECK-NEXT: smull r4, r2, r4, r9
-; CHECK-NEXT: asrs r1, r5, #31
-; CHECK-NEXT: str r5, [sp, #8] @ 4-byte Spill
-; CHECK-NEXT: subs r4, r5, r4
-; CHECK-NEXT: sbcs r1, r2
-; CHECK-NEXT: ldr r2, [sp, #12] @ 4-byte Reload
-; CHECK-NEXT: adds.w r10, r4, #-2147483648
-; CHECK-NEXT: adc r1, r1, #0
-; CHECK-NEXT: ldr r4, [r2, #-4]
-; CHECK-NEXT: muls r4, r3, r4
-; CHECK-NEXT: adds r3, #4
-; CHECK-NEXT: adds.w r12, r4, #-2147483648
-; CHECK-NEXT: asr.w r5, r4, #31
-; CHECK-NEXT: ldr r4, [r6]
-; CHECK-NEXT: adc r5, r5, #0
-; CHECK-NEXT: mul r2, r4, r0
+; CHECK-NEXT: ldr r1, [r9, #-4]
+; CHECK-NEXT: add.w r9, r9, #4
+; CHECK-NEXT: muls r1, r3, r1
+; CHECK-NEXT: adds.w r2, r1, #-2147483648
+; CHECK-NEXT: asr.w r4, r1, #31
+; CHECK-NEXT: ldr.w r1, [r12]
+; CHECK-NEXT: adc r3, r4, #0
+; CHECK-NEXT: mul r4, r1, r0
; CHECK-NEXT: adds r0, #4
-; CHECK-NEXT: add.w r2, r2, #-2147483648
-; CHECK-NEXT: asrl r12, r5, r2
-; CHECK-NEXT: smull r2, r5, r4, r12
-; CHECK-NEXT: lsll r2, r5, #30
-; CHECK-NEXT: ldr r2, [sp, #4] @ 4-byte Reload
-; CHECK-NEXT: asr.w r11, r5, #31
-; CHECK-NEXT: mov r12, r5
-; CHECK-NEXT: lsll r12, r11, r4
-; CHECK-NEXT: mul r2, r2, r9
-; CHECK-NEXT: lsrl r12, r11, #2
-; CHECK-NEXT: adds r2, #2
-; CHECK-NEXT: lsll r12, r11, r2
+; CHECK-NEXT: add.w r4, r4, #-2147483648
+; CHECK-NEXT: asrl r2, r3, r4
+; CHECK-NEXT: smull r2, r3, r1, r2
+; CHECK-NEXT: lsll r2, r3, #30
+; CHECK-NEXT: asr.w r11, r3, #31
+; CHECK-NEXT: mov r4, r3
+; CHECK-NEXT: ldrd r3, r2, [sp, #4] @ 8-byte Folded Reload
+; CHECK-NEXT: lsll r4, r11, r1
+; CHECK-NEXT: lsrl r4, r11, #2
+; CHECK-NEXT: muls r3, r2, r3
+; CHECK-NEXT: adds r3, #2
+; CHECK-NEXT: lsll r4, r11, r3
+; CHECK-NEXT: ldr r3, [sp, #12] @ 4-byte Reload
+; CHECK-NEXT: add.w r2, r4, #-2147483648
+; CHECK-NEXT: asrl r6, r5, r2
+; CHECK-NEXT: movs r2, #2
+; CHECK-NEXT: lsrl r6, r5, #2
+; CHECK-NEXT: adds r3, #4
+; CHECK-NEXT: str r6, [r2]
+; CHECK-NEXT: ldr r2, [r10], #-4
+; CHECK-NEXT: mls r4, r2, r1, r8
; CHECK-NEXT: ldr r2, [sp] @ 4-byte Reload
-; CHECK-NEXT: add.w r5, r12, #-2147483648
-; CHECK-NEXT: asrl r10, r1, r5
-; CHECK-NEXT: ldr r5, [sp, #8] @ 4-byte Reload
-; CHECK-NEXT: lsrl r10, r1, #2
-; CHECK-NEXT: movs r1, #2
-; CHECK-NEXT: mov r9, r10
-; CHECK-NEXT: str.w r10, [r1]
-; CHECK-NEXT: ldr r1, [r8], #-4
-; CHECK-NEXT: mls r5, r1, r4, r5
-; CHECK-NEXT: adds.w r4, r5, #-2147483648
-; CHECK-NEXT: asr.w r1, r5, #31
+; CHECK-NEXT: adds.w r8, r4, #-2147483648
+; CHECK-NEXT: asr.w r1, r4, #31
; CHECK-NEXT: adc r1, r1, #0
-; CHECK-NEXT: lsrl r4, r1, #2
-; CHECK-NEXT: rsbs r1, r4, #0
+; CHECK-NEXT: lsrl r8, r1, #2
+; CHECK-NEXT: rsb.w r1, r8, #0
; CHECK-NEXT: str r1, [r2]
-; CHECK-NEXT: str r1, [r6, #-4]
-; CHECK-NEXT: adds r6, #4
-; CHECK-NEXT: ldr r1, [sp, #12] @ 4-byte Reload
-; CHECK-NEXT: adds r1, #4
+; CHECK-NEXT: str r1, [r12, #-4]
+; CHECK-NEXT: add.w r12, r12, #4
; CHECK-NEXT: le lr, .LBB2_2
; CHECK-NEXT: .LBB2_3: @ %while.end
; CHECK-NEXT: add sp, #16
diff --git a/llvm/test/CodeGen/Thumb2/mve-blockplacement.ll b/llvm/test/CodeGen/Thumb2/mve-blockplacement.ll
index d076cb00ad7e0..edbbbf25aab0a 100644
--- a/llvm/test/CodeGen/Thumb2/mve-blockplacement.ll
+++ b/llvm/test/CodeGen/Thumb2/mve-blockplacement.ll
@@ -355,8 +355,8 @@ define i32 @d(i64 %e, i32 %f, i64 %g, i32 %h) {
; CHECK-NEXT: sub sp, #4
; CHECK-NEXT: .vsave {d8, d9, d10, d11, d12, d13, d14, d15}
; CHECK-NEXT: vpush {d8, d9, d10, d11, d12, d13, d14, d15}
-; CHECK-NEXT: .pad #16
-; CHECK-NEXT: sub sp, #16
+; CHECK-NEXT: .pad #8
+; CHECK-NEXT: sub sp, #8
; CHECK-NEXT: mov lr, r0
; CHECK-NEXT: subs r0, #1
; CHECK-NEXT: sbcs r0, r1, #0
@@ -375,7 +375,7 @@ define i32 @d(i64 %e, i32 %f, i64 %g, i32 %h) {
; CHECK-NEXT: movw r2, #43691
; CHECK-NEXT: adds r1, #2
; CHECK-NEXT: movt r2, #43690
-; CHECK-NEXT: ldr r6, [sp, #128]
+; CHECK-NEXT: ldr r6, [sp, #120]
; CHECK-NEXT: movw r8, :lower16:c
; CHECK-NEXT: umull r1, r2, r1, r2
; CHECK-NEXT: movt r8, :upper16:c
@@ -384,7 +384,6 @@ define i32 @d(i64 %e, i32 %f, i64 %g, i32 %h) {
; CHECK-NEXT: @ implicit-def: $r5
; CHECK-NEXT: @ implicit-def: $r11
; CHECK-NEXT: mov.w r9, #12
-; CHECK-NEXT: str r4, [sp, #12] @ 4-byte Spill
; CHECK-NEXT: add.w r1, r1, r2, lsr #1
; CHECK-NEXT: add.w r0, r0, r2, lsr #1
; CHECK-NEXT: bic r3, r1, #3
@@ -395,7 +394,7 @@ define i32 @d(i64 %e, i32 %f, i64 %g, i32 %h) {
; CHECK-NEXT: vdup.32 q6, r0
; CHECK-NEXT: vadd.i32 q4, q0, r7
; CHECK-NEXT: vdup.32 q7, r0
-; CHECK-NEXT: strd r3, r7, [sp, #4] @ 8-byte Folded Spill
+; CHECK-NEXT: strd r3, r7, [sp] @ 8-byte Folded Spill
; CHECK-NEXT: b .LBB1_6
; CHECK-NEXT: .LBB1_2: @ %for.body6.preheader
; CHECK-NEXT: @ in Loop: Header=BB1_6 Depth=1
@@ -444,21 +443,19 @@ define i32 @d(i64 %e, i32 %f, i64 %g, i32 %h) {
; CHECK-NEXT: bhi .LBB1_17
; CHECK-NEXT: @ %bb.8: @ %for.body6.us.preheader
; CHECK-NEXT: @ in Loop: Header=BB1_6 Depth=1
-; CHECK-NEXT: ldrd r2, r3, [sp, #120]
+; CHECK-NEXT: ldrd r2, r3, [sp, #112]
; CHECK-NEXT: movs r0, #32
; CHECK-NEXT: movs r1, #0
-; CHECK-NEXT: mov r4, r6
; CHECK-NEXT: mov r7, r12
; CHECK-NEXT: mov r6, lr
; CHECK-NEXT: bl __aeabi_ldivmod
; CHECK-NEXT: mov lr, r6
-; CHECK-NEXT: mov r6, r4
; CHECK-NEXT: mov r12, r7
-; CHECK-NEXT: ldr r3, [sp, #4] @ 4-byte Reload
-; CHECK-NEXT: ldr r4, [sp, #12] @ 4-byte Reload
+; CHECK-NEXT: ldr r3, [sp] @ 4-byte Reload
; CHECK-NEXT: vdup.32 q0, r2
-; CHECK-NEXT: ldr r7, [sp, #8] @ 4-byte Reload
+; CHECK-NEXT: ldr r6, [sp, #120]
; CHECK-NEXT: mov r0, r11
+; CHECK-NEXT: ldr r7, [sp, #4] @ 4-byte Reload
; CHECK-NEXT: b .LBB1_10
; CHECK-NEXT: .LBB1_9: @ %for.cond.cleanup17.us
; CHECK-NEXT: @ in Loop: Header=BB1_10 Depth=2
@@ -573,7 +570,7 @@ define i32 @d(i64 %e, i32 %f, i64 %g, i32 %h) {
; CHECK-NEXT: @ =>This Inner Loop Header: Depth=1
; CHECK-NEXT: b .LBB1_27
; CHECK-NEXT: .LBB1_28: @ %for.cond.cleanup
-; CHECK-NEXT: add sp, #16
+; CHECK-NEXT: add sp, #8
; CHECK-NEXT: vpop {d8, d9, d10, d11, d12, d13, d14, d15}
; CHECK-NEXT: add sp, #4
; CHECK-NEXT: pop.w {r4, r5, r6, r7, r8, r9, r10, r11, pc}
diff --git a/llvm/test/CodeGen/Thumb2/mve-float32regloops.ll b/llvm/test/CodeGen/Thumb2/mve-float32regloops.ll
index 28166e455aba2..4c0ded4515b65 100644
--- a/llvm/test/CodeGen/Thumb2/mve-float32regloops.ll
+++ b/llvm/test/CodeGen/Thumb2/mve-float32regloops.ll
@@ -999,7 +999,7 @@ define void @fir(ptr nocapture readonly %S, ptr nocapture readonly %pSrc, ptr no
; CHECK-NEXT: sub sp, #32
; CHECK-NEXT: ldrh r6, [r0]
; CHECK-NEXT: movs r5, #1
-; CHECK-NEXT: ldrd r4, r10, [r0, #4]
+; CHECK-NEXT: ldrd r4, r9, [r0, #4]
; CHECK-NEXT: sub.w r0, r6, #8
; CHECK-NEXT: add.w r3, r0, r0, lsr #29
; CHECK-NEXT: and r0, r0, #7
@@ -1008,10 +1008,11 @@ define void @fir(ptr nocapture readonly %S, ptr nocapture readonly %pSrc, ptr no
; CHECK-NEXT: it gt
; CHECK-NEXT: asrgt r5, r3, #3
; CHECK-NEXT: add.w r3, r4, r6, lsl #2
-; CHECK-NEXT: sub.w r9, r3, #4
+; CHECK-NEXT: subs r3, #4
+; CHECK-NEXT: str r3, [sp, #28] @ 4-byte Spill
; CHECK-NEXT: rsbs r3, r6, #0
; CHECK-NEXT: str r3, [sp, #12] @ 4-byte Spill
-; CHECK-NEXT: add.w r3, r10, #32
+; CHECK-NEXT: add.w r3, r9, #32
; CHECK-NEXT: str r5, [sp, #4] @ 4-byte Spill
; CHECK-NEXT: str r6, [sp, #16] @ 4-byte Spill
; CHECK-NEXT: str r3, [sp, #8] @ 4-byte Spill
@@ -1024,8 +1025,7 @@ define void @fir(ptr nocapture readonly %S, ptr nocapture readonly %pSrc, ptr no
; CHECK-NEXT: b .LBB16_5
; CHECK-NEXT: .LBB16_4: @ %for.end
; CHECK-NEXT: @ in Loop: Header=BB16_6 Depth=1
-; CHECK-NEXT: ldr r1, [sp, #28] @ 4-byte Reload
-; CHECK-NEXT: ldrd r0, r9, [sp, #20] @ 8-byte Folded Reload
+; CHECK-NEXT: ldrd r0, r1, [sp, #20] @ 8-byte Folded Reload
; CHECK-NEXT: wls lr, r0, .LBB16_5
; CHECK-NEXT: b .LBB16_10
; CHECK-NEXT: .LBB16_5: @ %while.end
@@ -1040,15 +1040,17 @@ define void @fir(ptr nocapture readonly %S, ptr nocapture readonly %pSrc, ptr no
; CHECK-NEXT: @ =>This Loop Header: Depth=1
; CHECK-NEXT: @ Child Loop BB16_8 Depth 2
; CHECK-NEXT: @ Child Loop BB16_11 Depth 2
-; CHECK-NEXT: add.w lr, r10, #8
; CHECK-NEXT: vldrw.u32 q0, [r1], #16
-; CHECK-NEXT: ldrd r3, r7, [r10]
-; CHECK-NEXT: ldm.w lr, {r0, r5, r6, lr}
-; CHECK-NEXT: ldrd r11, r8, [r10, #24]
-; CHECK-NEXT: vstrb.8 q0, [r9], #16
+; CHECK-NEXT: add.w lr, r9, #8
+; CHECK-NEXT: ldrd r3, r7, [r9]
+; CHECK-NEXT: str r1, [sp, #24] @ 4-byte Spill
+; CHECK-NEXT: ldr r1, [sp, #28] @ 4-byte Reload
+; CHECK-NEXT: ldm.w lr, {r0, r5, lr}
+; CHECK-NEXT: ldrd r10, r11, [r9, #20]
+; CHECK-NEXT: ldr.w r8, [r9, #28]
+; CHECK-NEXT: vstrb.8 q0, [r1], #16
; CHECK-NEXT: vldrw.u32 q0, [r4], #32
; CHECK-NEXT: str r1, [sp, #28] @ 4-byte Spill
-; CHECK-NEXT: str.w r9, [sp, #24] @ 4-byte Spill
; CHECK-NEXT: vldrw.u32 q1, [r4, #-28]
; CHECK-NEXT: vmul.f32 q0, q0, r3
; CHECK-NEXT: vldrw.u32 q6, [r4, #-24]
@@ -1059,9 +1061,9 @@ define void @fir(ptr nocapture readonly %S, ptr nocapture readonly %pSrc, ptr no
; CHECK-NEXT: vldrw.u32 q2, [r4, #-12]
; CHECK-NEXT: vfma.f32 q0, q4, r5
; CHECK-NEXT: vldrw.u32 q3, [r4, #-8]
-; CHECK-NEXT: vfma.f32 q0, q5, r6
+; CHECK-NEXT: vfma.f32 q0, q5, lr
; CHECK-NEXT: vldrw.u32 q1, [r4, #-4]
-; CHECK-NEXT: vfma.f32 q0, q2, lr
+; CHECK-NEXT: vfma.f32 q0, q2, r10
; CHECK-NEXT: ldr r0, [sp, #16] @ 4-byte Reload
; CHECK-NEXT: vfma.f32 q0, q3, r11
; CHECK-NEXT: vfma.f32 q0, q1, r8
@@ -1075,25 +1077,26 @@ define void @fir(ptr nocapture readonly %S, ptr nocapture readonly %pSrc, ptr no
; CHECK-NEXT: .LBB16_8: @ %for.body
; CHECK-NEXT: @ Parent Loop BB16_6 Depth=1
; CHECK-NEXT: @ => This Inner Loop Header: Depth=2
-; CHECK-NEXT: ldm.w r7, {r0, r3, r5, r6, r8, r11}
+; CHECK-NEXT: ldm.w r7, {r0, r3, r5, r6}
; CHECK-NEXT: vldrw.u32 q1, [r4], #32
+; CHECK-NEXT: add.w r11, r7, #16
; CHECK-NEXT: vldrw.u32 q6, [r4, #-24]
; CHECK-NEXT: vldrw.u32 q4, [r4, #-20]
; CHECK-NEXT: vfma.f32 q0, q1, r0
; CHECK-NEXT: vldrw.u32 q1, [r4, #-28]
+; CHECK-NEXT: ldm.w r11, {r1, r8, r10, r11}
; CHECK-NEXT: vldrw.u32 q5, [r4, #-16]
-; CHECK-NEXT: vldrw.u32 q2, [r4, #-12]
; CHECK-NEXT: vfma.f32 q0, q1, r3
-; CHECK-NEXT: ldrd r9, r1, [r7, #24]
+; CHECK-NEXT: vldrw.u32 q2, [r4, #-12]
; CHECK-NEXT: vfma.f32 q0, q6, r5
; CHECK-NEXT: vldrw.u32 q3, [r4, #-8]
; CHECK-NEXT: vfma.f32 q0, q4, r6
; CHECK-NEXT: vldrw.u32 q1, [r4, #-4]
-; CHECK-NEXT: vfma.f32 q0, q5, r8
+; CHECK-NEXT: vfma.f32 q0, q5, r1
; CHECK-NEXT: adds r7, #32
-; CHECK-NEXT: vfma.f32 q0, q2, r11
-; CHECK-NEXT: vfma.f32 q0, q3, r9
-; CHECK-NEXT: vfma.f32 q0, q1, r1
+; CHECK-NEXT: vfma.f32 q0, q2, r8
+; CHECK-NEXT: vfma.f32 q0, q3, r10
+; CHECK-NEXT: vfma.f32 q0, q1, r11
; CHECK-NEXT: le lr, .LBB16_8
; CHECK-NEXT: b .LBB16_4
; CHECK-NEXT: .LBB16_9: @ in Loop: Header=BB16_6 Depth=1
diff --git a/llvm/test/CodeGen/Thumb2/mve-laneinterleaving-reduct.ll b/llvm/test/CodeGen/Thumb2/mve-laneinterleaving-reduct.ll
index 652d25af02e7c..8fe310bd3d5e3 100644
--- a/llvm/test/CodeGen/Thumb2/mve-laneinterleaving-reduct.ll
+++ b/llvm/test/CodeGen/Thumb2/mve-laneinterleaving-reduct.ll
@@ -180,15 +180,15 @@ define void @correlate(ptr nocapture noundef readonly %ID, ptr nocapture noundef
; CHECK-NEXT: @ in Loop: Header=BB4_4 Depth=1
; CHECK-NEXT: ldr r1, [sp] @ 4-byte Reload
; CHECK-NEXT: add.w r2, r9, r10
-; CHECK-NEXT: add.w r7, r1, r9, lsl #1
+; CHECK-NEXT: add.w r5, r1, r9, lsl #1
; CHECK-NEXT: add.w r2, r1, r2, lsl #1
-; CHECK-NEXT: sub.w r5, r8, r9
-; CHECK-NEXT: dlstp.32 lr, r5
+; CHECK-NEXT: sub.w r7, r8, r9
+; CHECK-NEXT: dlstp.32 lr, r7
; CHECK-NEXT: .LBB4_11: @ %vec.epilog.vector.body
; CHECK-NEXT: @ Parent Loop BB4_4 Depth=1
; CHECK-NEXT: @ => This Inner Loop Header: Depth=2
; CHECK-NEXT: rsbs r4, r3, #0
-; CHECK-NEXT: vldrh.s32 q0, [r7], #8
+; CHECK-NEXT: vldrh.s32 q0, [r5], #8
; CHECK-NEXT: vldrh.s32 q1, [r2], #8
; CHECK-NEXT: vmul.i32 q0, q1, q0
; CHECK-NEXT: vshl.s32 q0, r4
diff --git a/llvm/test/CodeGen/Thumb2/mve-memtp-loop.ll b/llvm/test/CodeGen/Thumb2/mve-memtp-loop.ll
index da59cb259db61..22deb23cad27e 100644
--- a/llvm/test/CodeGen/Thumb2/mve-memtp-loop.ll
+++ b/llvm/test/CodeGen/Thumb2/mve-memtp-loop.ll
@@ -548,44 +548,43 @@ define i32 @reverted(i1 zeroext %b) {
; CHECK-NEXT: letp lr, .LBB19_1
; CHECK-NEXT: .LBB19_2: @ %entry
; CHECK-NEXT: ldr r0, [sp, #8] @ 4-byte Reload
-; CHECK-NEXT: movw r6, :lower16:arr_20
-; CHECK-NEXT: movt r6, :upper16:arr_20
-; CHECK-NEXT: add.w r3, r6, #80
+; CHECK-NEXT: movw r11, :lower16:arr_20
+; CHECK-NEXT: adr r6, .LCPI19_0
+; CHECK-NEXT: movt r11, :upper16:arr_20
; CHECK-NEXT: dls lr, r0
; CHECK-NEXT: movw r0, :lower16:arr_21
; CHECK-NEXT: movt r0, :upper16:arr_21
; CHECK-NEXT: add.w r5, r0, #36
-; CHECK-NEXT: add.w r11, r6, #128
-; CHECK-NEXT: add.w r7, r6, #112
-; CHECK-NEXT: add.w r2, r6, #96
-; CHECK-NEXT: add.w r4, r6, #64
-; CHECK-NEXT: add.w r0, r6, #48
-; CHECK-NEXT: add.w r1, r6, #32
-; CHECK-NEXT: add.w r12, r6, #16
-; CHECK-NEXT: adr r6, .LCPI19_0
-; CHECK-NEXT: vldrw.u32 q0, [r6]
-; CHECK-NEXT: movw r6, :lower16:arr_20
+; CHECK-NEXT: add.w r3, r11, #80
+; CHECK-NEXT: add.w r9, r11, #128
+; CHECK-NEXT: add.w r7, r11, #112
+; CHECK-NEXT: add.w r2, r11, #96
+; CHECK-NEXT: add.w r4, r11, #64
+; CHECK-NEXT: add.w r0, r11, #48
+; CHECK-NEXT: add.w r1, r11, #32
+; CHECK-NEXT: add.w r12, r11, #16
; CHECK-NEXT: mov.w r8, #327685
-; CHECK-NEXT: mov.w r9, #5
+; CHECK-NEXT: vldrw.u32 q0, [r6]
+; CHECK-NEXT: mov r6, r8
+; CHECK-NEXT: mov.w r10, #5
; CHECK-NEXT: vmov.i16 q1, #0x5
-; CHECK-NEXT: mov.w r10, #0
-; CHECK-NEXT: movt r6, :upper16:arr_20
+; CHECK-NEXT: mov.w r8, #0
; CHECK-NEXT: .LBB19_3: @ %for.cond8.preheader
; CHECK-NEXT: @ =>This Inner Loop Header: Depth=1
-; CHECK-NEXT: str r8, [r5, #-4]
+; CHECK-NEXT: str r6, [r5, #-4]
; CHECK-NEXT: vstrh.16 q1, [r5, #-36]
-; CHECK-NEXT: strh.w r9, [r5]
+; CHECK-NEXT: strh.w r10, [r5]
; CHECK-NEXT: vstrh.16 q1, [r5, #-20]
; CHECK-NEXT: vstrw.32 q0, [r3]
; CHECK-NEXT: vstrh.16 q0, [r12], #152
-; CHECK-NEXT: vstrh.16 q0, [r6], #152
+; CHECK-NEXT: vstrh.16 q0, [r11], #152
; CHECK-NEXT: vstrh.16 q0, [r1], #152
; CHECK-NEXT: vstrh.16 q0, [r0], #152
; CHECK-NEXT: vstrh.16 q0, [r4], #152
; CHECK-NEXT: vstrh.16 q0, [r2], #152
; CHECK-NEXT: vstrh.16 q0, [r7], #152
-; CHECK-NEXT: vstrh.16 q0, [r11], #152
-; CHECK-NEXT: strd r9, r10, [r3, #64]
+; CHECK-NEXT: vstrh.16 q0, [r9], #152
+; CHECK-NEXT: strd r10, r8, [r3, #64]
; CHECK-NEXT: adds r5, #38
; CHECK-NEXT: adds r3, #152
; CHECK-NEXT: le lr, .LBB19_3
@@ -601,46 +600,46 @@ define i32 @reverted(i1 zeroext %b) {
; CHECK-NEXT: vstrb.8 q1, [r0], #16
; CHECK-NEXT: letp lr, .LBB19_5
; CHECK-NEXT: .LBB19_6: @ %for.cond.cleanup6
-; CHECK-NEXT: movw r6, :lower16:arr_20
+; CHECK-NEXT: movw r2, :lower16:arr_20
; CHECK-NEXT: movw r0, #7376
-; CHECK-NEXT: movt r6, :upper16:arr_20
-; CHECK-NEXT: adds r3, r6, r0
+; CHECK-NEXT: movt r2, :upper16:arr_20
+; CHECK-NEXT: adds r3, r2, r0
; CHECK-NEXT: movw r0, #7408
-; CHECK-NEXT: add.w r12, r6, r0
+; CHECK-NEXT: add.w r12, r2, r0
; CHECK-NEXT: movw r0, #7344
-; CHECK-NEXT: add.w r9, r6, r0
+; CHECK-NEXT: add.w r11, r2, r0
; CHECK-NEXT: movw r0, #7312
-; CHECK-NEXT: adds r2, r6, r0
+; CHECK-NEXT: add.w r9, r2, r0
; CHECK-NEXT: movw r0, :lower16:arr_21
-; CHECK-NEXT: add.w r1, r6, #7424
-; CHECK-NEXT: add.w r7, r6, #7392
-; CHECK-NEXT: add.w r4, r6, #7360
-; CHECK-NEXT: add.w r5, r6, #7328
-; CHECK-NEXT: add.w r8, r6, #7296
-; CHECK-NEXT: ldr r6, [sp, #8] @ 4-byte Reload
+; CHECK-NEXT: add.w r1, r2, #7424
+; CHECK-NEXT: add.w r7, r2, #7392
+; CHECK-NEXT: add.w r4, r2, #7360
+; CHECK-NEXT: add.w r5, r2, #7328
+; CHECK-NEXT: add.w r6, r2, #7296
+; CHECK-NEXT: ldr r2, [sp, #8] @ 4-byte Reload
; CHECK-NEXT: movt r0, :upper16:arr_21
; CHECK-NEXT: addw r0, r0, #1860
; CHECK-NEXT: mov.w r10, #5
-; CHECK-NEXT: dls lr, r6
-; CHECK-NEXT: mov.w r6, #327685
+; CHECK-NEXT: dls lr, r2
+; CHECK-NEXT: mov.w r2, #327685
; CHECK-NEXT: vmov.i16 q1, #0x5
-; CHECK-NEXT: mov.w r11, #0
+; CHECK-NEXT: mov.w r8, #0
; CHECK-NEXT: .LBB19_7: @ %for.cond8.preheader.1
; CHECK-NEXT: @ =>This Inner Loop Header: Depth=1
-; CHECK-NEXT: str r6, [r0, #-4]
+; CHECK-NEXT: str r2, [r0, #-4]
; CHECK-NEXT: vstrh.16 q1, [r0, #-36]
; CHECK-NEXT: strh.w r10, [r0]
; CHECK-NEXT: vstrh.16 q1, [r0, #-20]
; CHECK-NEXT: vstrw.32 q0, [r3]
-; CHECK-NEXT: vstrh.16 q0, [r2], #152
-; CHECK-NEXT: vstrh.16 q0, [r8], #152
-; CHECK-NEXT: vstrh.16 q0, [r5], #152
; CHECK-NEXT: vstrh.16 q0, [r9], #152
+; CHECK-NEXT: vstrh.16 q0, [r6], #152
+; CHECK-NEXT: vstrh.16 q0, [r5], #152
+; CHECK-NEXT: vstrh.16 q0, [r11], #152
; CHECK-NEXT: vstrh.16 q0, [r4], #152
; CHECK-NEXT: vstrh.16 q0, [r7], #152
; CHECK-NEXT: vstrh.16 q0, [r12], #152
; CHECK-NEXT: vstrh.16 q0, [r1], #152
-; CHECK-NEXT: strd r10, r11, [r3, #64]
+; CHECK-NEXT: strd r10, r8, [r3, #64]
; CHECK-NEXT: adds r0, #38
; CHECK-NEXT: adds r3, #152
; CHECK-NEXT: le lr, .LBB19_7
@@ -663,7 +662,7 @@ define i32 @reverted(i1 zeroext %b) {
; CHECK-NEXT: movw r0, #14704
; CHECK-NEXT: add.w r12, r7, r0
; CHECK-NEXT: movw r0, #14688
-; CHECK-NEXT: add.w r8, r7, r0
+; CHECK-NEXT: add.w r11, r7, r0
; CHECK-NEXT: movw r0, #14640
; CHECK-NEXT: add.w r9, r7, r0
; CHECK-NEXT: movw r0, #14624
@@ -681,7 +680,7 @@ define i32 @reverted(i1 zeroext %b) {
; CHECK-NEXT: dls lr, r7
; CHECK-NEXT: mov.w r7, #327685
; CHECK-NEXT: vmov.i16 q1, #0x5
-; CHECK-NEXT: mov.w r11, #0
+; CHECK-NEXT: mov.w r8, #0
; CHECK-NEXT: .LBB19_11: @ %for.cond8.preheader.2
; CHECK-NEXT: @ =>This Inner Loop Header: Depth=1
; CHECK-NEXT: str r7, [r1, #-4]
@@ -694,10 +693,10 @@ define i32 @reverted(i1 zeroext %b) {
; CHECK-NEXT: vstrh.16 q0, [r2], #152
; CHECK-NEXT: vstrh.16 q0, [r9], #152
; CHECK-NEXT: vstrh.16 q0, [r5], #152
-; CHECK-NEXT: vstrh.16 q0, [r8], #152
+; CHECK-NEXT: vstrh.16 q0, [r11], #152
; CHECK-NEXT: vstrh.16 q0, [r12], #152
; CHECK-NEXT: vstrh.16 q0, [r4], #152
-; CHECK-NEXT: strd r10, r11, [r3, #64]
+; CHECK-NEXT: strd r10, r8, [r3, #64]
; CHECK-NEXT: adds r1, #38
; CHECK-NEXT: adds r3, #152
; CHECK-NEXT: le lr, .LBB19_11
@@ -721,9 +720,9 @@ define i32 @reverted(i1 zeroext %b) {
; CHECK-NEXT: movt r7, :upper16:arr_20
; CHECK-NEXT: add.w r12, r7, r1
; CHECK-NEXT: movw r1, #21984
-; CHECK-NEXT: add.w r8, r7, r1
+; CHECK-NEXT: add.w r10, r7, r1
; CHECK-NEXT: movw r1, #21952
-; CHECK-NEXT: add.w r9, r7, r1
+; CHECK-NEXT: add.w r8, r7, r1
; CHECK-NEXT: movw r1, #21936
; CHECK-NEXT: movw r0, #21968
; CHECK-NEXT: adds r5, r7, r1
@@ -735,7 +734,7 @@ define i32 @reverted(i1 zeroext %b) {
; CHECK-NEXT: add.w r3, r7, #22016
; CHECK-NEXT: add.w r6, r7, #21888
; CHECK-NEXT: ldr r7, [sp, #8] @ 4-byte Reload
-; CHECK-NEXT: mov.w r10, #5
+; CHECK-NEXT: mov.w r9, #5
; CHECK-NEXT: vmov.i16 q1, #0x5
; CHECK-NEXT: mov.w r11, #0
; CHECK-NEXT: dls lr, r7
@@ -744,18 +743,18 @@ define i32 @reverted(i1 zeroext %b) {
; CHECK-NEXT: @ =>This Inner Loop Header: Depth=1
; CHECK-NEXT: str r7, [r2, #-4]
; CHECK-NEXT: vstrh.16 q1, [r2, #-36]
-; CHECK-NEXT: strh.w r10, [r2]
+; CHECK-NEXT: strh.w r9, [r2]
; CHECK-NEXT: vstrh.16 q1, [r2, #-20]
; CHECK-NEXT: vstrw.32 q0, [r0]
; CHECK-NEXT: vstrh.16 q0, [r4], #152
; CHECK-NEXT: vstrh.16 q0, [r6], #152
; CHECK-NEXT: vstrh.16 q0, [r1], #152
; CHECK-NEXT: vstrh.16 q0, [r5], #152
-; CHECK-NEXT: vstrh.16 q0, [r9], #152
; CHECK-NEXT: vstrh.16 q0, [r8], #152
+; CHECK-NEXT: vstrh.16 q0, [r10], #152
; CHECK-NEXT: vstrh.16 q0, [r12], #152
; CHECK-NEXT: vstrh.16 q0, [r3], #152
-; CHECK-NEXT: strd r10, r11, [r0, #64]
+; CHECK-NEXT: strd r9, r11, [r0, #64]
; CHECK-NEXT: adds r2, #38
; CHECK-NEXT: adds r0, #152
; CHECK-NEXT: le lr, .LBB19_15
diff --git a/llvm/test/CodeGen/Thumb2/mve-phireg.ll b/llvm/test/CodeGen/Thumb2/mve-phireg.ll
index dad856c0677a1..14ea3a3713224 100644
--- a/llvm/test/CodeGen/Thumb2/mve-phireg.ll
+++ b/llvm/test/CodeGen/Thumb2/mve-phireg.ll
@@ -14,8 +14,8 @@ define arm_aapcs_vfpcc void @k() {
; CHECK-NEXT: sub sp, #32
; CHECK-NEXT: adr r5, .LCPI0_0
; CHECK-NEXT: adr r4, .LCPI0_1
-; CHECK-NEXT: vldrw.u32 q6, [r5]
-; CHECK-NEXT: vldrw.u32 q5, [r4]
+; CHECK-NEXT: vldrw.u32 q5, [r5]
+; CHECK-NEXT: vldrw.u32 q6, [r4]
; CHECK-NEXT: add r0, sp, #16
; CHECK-NEXT: vmov.i32 q0, #0x1
; CHECK-NEXT: vmov.i8 q1, #0x0
@@ -25,14 +25,14 @@ define arm_aapcs_vfpcc void @k() {
; CHECK-NEXT: movs r1, #0
; CHECK-NEXT: .LBB0_1: @ %vector.body
; CHECK-NEXT: @ =>This Inner Loop Header: Depth=1
-; CHECK-NEXT: vand q5, q5, q0
; CHECK-NEXT: vand q6, q6, q0
-; CHECK-NEXT: vcmp.i32 eq, q5, zr
-; CHECK-NEXT: vpsel q5, q2, q1
+; CHECK-NEXT: vand q5, q5, q0
; CHECK-NEXT: vcmp.i32 eq, q6, zr
; CHECK-NEXT: vpsel q6, q2, q1
-; CHECK-NEXT: vstrh.32 q5, [r0]
-; CHECK-NEXT: vstrh.32 q6, [r0, #8]
+; CHECK-NEXT: vcmp.i32 eq, q5, zr
+; CHECK-NEXT: vpsel q5, q2, q1
+; CHECK-NEXT: vstrh.32 q6, [r0]
+; CHECK-NEXT: vstrh.32 q5, [r0, #8]
; CHECK-NEXT: vldrw.u32 q5, [r0]
; CHECK-NEXT: vcmp.i16 ne, q5, zr
; CHECK-NEXT: vmov.i32 q5, #0x0
diff --git a/llvm/test/CodeGen/Thumb2/mve-pipelineloops.ll b/llvm/test/CodeGen/Thumb2/mve-pipelineloops.ll
index 43ed5eefbf4c7..ff5a27149cb2e 100644
--- a/llvm/test/CodeGen/Thumb2/mve-pipelineloops.ll
+++ b/llvm/test/CodeGen/Thumb2/mve-pipelineloops.ll
@@ -17,16 +17,16 @@ define void @arm_cmplx_dot_prod_q15(ptr noundef %pSrcA, ptr noundef %pSrcB, i32
; CHECK-NEXT: mov.w r5, #0
; CHECK-NEXT: csel r7, r6, r5, hs
; CHECK-NEXT: add.w lr, r7, #1
-; CHECK-NEXT: mov r4, r5
+; CHECK-NEXT: mov r8, r5
; CHECK-NEXT: vldrh.u16 q0, [r0], #32
; CHECK-NEXT: movs r7, #0
-; CHECK-NEXT: mov r8, r5
+; CHECK-NEXT: mov r6, r5
; CHECK-NEXT: vldrh.u16 q1, [r1], #32
-; CHECK-NEXT: vmlsldava.s16 r4, r7, q0, q1
+; CHECK-NEXT: vmlsldava.s16 r8, r7, q0, q1
; CHECK-NEXT: vldrh.u16 q2, [r0, #-16]
-; CHECK-NEXT: vmlaldavax.s16 r8, r5, q0, q1
+; CHECK-NEXT: vmlaldavax.s16 r6, r5, q0, q1
; CHECK-NEXT: vldrh.u16 q3, [r1, #-16]
-; CHECK-NEXT: vmlsldava.s16 r4, r7, q2, q3
+; CHECK-NEXT: vmlsldava.s16 r8, r7, q2, q3
; CHECK-NEXT: vldrh.u16 q0, [r1], #32
; CHECK-NEXT: sub.w lr, lr, #1
; CHECK-NEXT: cmp.w lr, #0
@@ -35,75 +35,72 @@ define void @arm_cmplx_dot_prod_q15(ptr noundef %pSrcA, ptr noundef %pSrcB, i32
; CHECK-NEXT: .p2align 2
; CHECK-NEXT: .LBB0_2: @ %while.body
; CHECK-NEXT: @ =>This Inner Loop Header: Depth=1
-; CHECK-NEXT: vmlaldavax.s16 r8, r5, q2, q3
+; CHECK-NEXT: vmlaldavax.s16 r6, r5, q2, q3
; CHECK-NEXT: vldrh.u16 q3, [r1, #-16]
-; CHECK-NEXT: vmlsldava.s16 r4, r7, q1, q0
+; CHECK-NEXT: vmlsldava.s16 r8, r7, q1, q0
; CHECK-NEXT: vldrh.u16 q2, [r0, #-16]
-; CHECK-NEXT: vmlaldavax.s16 r8, r5, q1, q0
+; CHECK-NEXT: vmlaldavax.s16 r6, r5, q1, q0
; CHECK-NEXT: vldrh.u16 q1, [r0], #32
-; CHECK-NEXT: vmlsldava.s16 r4, r7, q2, q3
+; CHECK-NEXT: vmlsldava.s16 r8, r7, q2, q3
; CHECK-NEXT: vldrh.u16 q0, [r1], #32
; CHECK-NEXT: le lr, .LBB0_2
; CHECK-NEXT: .LBB0_3:
-; CHECK-NEXT: vmlaldavax.s16 r8, r5, q2, q3
-; CHECK-NEXT: movs r6, #14
-; CHECK-NEXT: and.w r2, r6, r2, lsl #1
-; CHECK-NEXT: vmlaldavax.s16 r8, r5, q1, q0
+; CHECK-NEXT: vmlaldavax.s16 r6, r5, q2, q3
+; CHECK-NEXT: movs r4, #14
+; CHECK-NEXT: and.w r2, r4, r2, lsl #1
+; CHECK-NEXT: vmlaldavax.s16 r6, r5, q1, q0
; CHECK-NEXT: vldrh.u16 q2, [r0, #-16]
-; CHECK-NEXT: vmlsldava.s16 r4, r7, q1, q0
+; CHECK-NEXT: vmlsldava.s16 r8, r7, q1, q0
; CHECK-NEXT: vldrh.u16 q0, [r1, #-16]
-; CHECK-NEXT: vmlaldavax.s16 r8, r5, q2, q0
+; CHECK-NEXT: vmlaldavax.s16 r6, r5, q2, q0
; CHECK-NEXT: vctp.16 r2
-; CHECK-NEXT: vmlsldava.s16 r4, r7, q2, q0
+; CHECK-NEXT: vmlsldava.s16 r8, r7, q2, q0
; CHECK-NEXT: vpst
; CHECK-NEXT: vldrht.u16 q1, [r0]
; CHECK-NEXT: cmp r2, #9
; CHECK-NEXT: vpsttt
; CHECK-NEXT: vldrht.u16 q0, [r1]
-; CHECK-NEXT: vmlsldavat.s16 r4, r7, q1, q0
-; CHECK-NEXT: vmlaldavaxt.s16 r8, r5, q1, q0
-; CHECK-NEXT: blo .LBB0_10
+; CHECK-NEXT: vmlsldavat.s16 r8, r7, q1, q0
+; CHECK-NEXT: vmlaldavaxt.s16 r6, r5, q1, q0
+; CHECK-NEXT: blo .LBB0_9
; CHECK-NEXT: @ %bb.4: @ %do.body.1
; CHECK-NEXT: subs r2, #8
; CHECK-NEXT: vctp.16 r2
; CHECK-NEXT: vpstttt
; CHECK-NEXT: vldrht.u16 q0, [r0, #16]
; CHECK-NEXT: vldrht.u16 q1, [r1, #16]
-; CHECK-NEXT: vmlsldavat.s16 r4, r7, q0, q1
-; CHECK-NEXT: vmlaldavaxt.s16 r8, r5, q0, q1
-; CHECK-NEXT: b .LBB0_10
+; CHECK-NEXT: vmlsldavat.s16 r8, r7, q0, q1
+; CHECK-NEXT: vmlaldavaxt.s16 r6, r5, q0, q1
+; CHECK-NEXT: b .LBB0_9
; CHECK-NEXT: .p2align 2
; CHECK-NEXT: .LBB0_5: @ %if.else
-; CHECK-NEXT: mov.w r4, #0
-; CHECK-NEXT: cbz r2, .LBB0_9
+; CHECK-NEXT: mov.w r8, #0
+; CHECK-NEXT: cbz r2, .LBB0_8
; CHECK-NEXT: @ %bb.6: @ %while.body14.preheader
-; CHECK-NEXT: lsls r6, r2, #1
-; CHECK-NEXT: mov r5, r4
-; CHECK-NEXT: mov r7, r4
-; CHECK-NEXT: movs r2, #0
-; CHECK-NEXT: dlstp.16 lr, r6
+; CHECK-NEXT: lsls r4, r2, #1
+; CHECK-NEXT: mov r5, r8
+; CHECK-NEXT: movs r6, #0
+; CHECK-NEXT: mov r7, r8
+; CHECK-NEXT: dlstp.16 lr, r4
; CHECK-NEXT: .p2align 2
; CHECK-NEXT: .LBB0_7: @ %while.body14
; CHECK-NEXT: @ =>This Inner Loop Header: Depth=1
; CHECK-NEXT: vldrh.u16 q0, [r0], #16
; CHECK-NEXT: vldrh.u16 q1, [r1], #16
-; CHECK-NEXT: vmlsldava.s16 r2, r7, q0, q1
-; CHECK-NEXT: vmlaldavax.s16 r4, r5, q0, q1
+; CHECK-NEXT: vmlsldava.s16 r8, r7, q0, q1
+; CHECK-NEXT: vmlaldavax.s16 r6, r5, q0, q1
; CHECK-NEXT: letp lr, .LBB0_7
-; CHECK-NEXT: @ %bb.8: @ %if.end.loopexit177
-; CHECK-NEXT: mov r8, r4
-; CHECK-NEXT: mov r4, r2
-; CHECK-NEXT: b .LBB0_10
+; CHECK-NEXT: b .LBB0_9
; CHECK-NEXT: .p2align 2
-; CHECK-NEXT: .LBB0_9:
-; CHECK-NEXT: mov r7, r4
-; CHECK-NEXT: mov.w r8, #0
-; CHECK-NEXT: mov r5, r4
-; CHECK-NEXT: .LBB0_10: @ %if.end
-; CHECK-NEXT: asrl r4, r7, #6
-; CHECK-NEXT: asrl r8, r5, #6
-; CHECK-NEXT: str r4, [r3]
-; CHECK-NEXT: str.w r8, [r12]
+; CHECK-NEXT: .LBB0_8:
+; CHECK-NEXT: mov r7, r8
+; CHECK-NEXT: movs r6, #0
+; CHECK-NEXT: mov r5, r8
+; CHECK-NEXT: .LBB0_9: @ %if.end
+; CHECK-NEXT: asrl r8, r7, #6
+; CHECK-NEXT: asrl r6, r5, #6
+; CHECK-NEXT: str.w r8, [r3]
+; CHECK-NEXT: str.w r6, [r12]
; CHECK-NEXT: pop.w {r4, r5, r6, r7, r8, pc}
entry:
%cmp = icmp ugt i32 %numSamples, 15
diff --git a/llvm/test/CodeGen/Thumb2/mve-postinc-dct.ll b/llvm/test/CodeGen/Thumb2/mve-postinc-dct.ll
index f90af3cc5ba24..c987c4b537200 100644
--- a/llvm/test/CodeGen/Thumb2/mve-postinc-dct.ll
+++ b/llvm/test/CodeGen/Thumb2/mve-postinc-dct.ll
@@ -845,12 +845,12 @@ define void @DCT_mve6(ptr nocapture readonly %S, ptr nocapture readonly %pIn, pt
; CHECK-NEXT: .LBB5_3: @ %vector.body
; CHECK-NEXT: @ Parent Loop BB5_2 Depth=1
; CHECK-NEXT: @ => This Inner Loop Header: Depth=2
-; CHECK-NEXT: add.w r12, r3, r5
+; CHECK-NEXT: add.w r9, r3, r5
; CHECK-NEXT: vldrw.u32 q6, [r1], #16
; CHECK-NEXT: vldrw.u32 q7, [r3], #16
-; CHECK-NEXT: add.w r10, r12, r5
+; CHECK-NEXT: add.w r10, r9, r5
; CHECK-NEXT: vfma.f32 q4, q7, q6
-; CHECK-NEXT: vldrw.u32 q7, [r12]
+; CHECK-NEXT: vldrw.u32 q7, [r9]
; CHECK-NEXT: add.w r6, r10, r5
; CHECK-NEXT: vfma.f32 q5, q7, q6
; CHECK-NEXT: vldrw.u32 q7, [r10]
@@ -1093,18 +1093,16 @@ define void @DCT_mve7(ptr nocapture readonly %S, ptr nocapture readonly %pIn, pt
; CHECK-NEXT: vfmat.f32 q5, q0, q7
; CHECK-NEXT: vldrwt.u32 q0, [r10]
; CHECK-NEXT: add.w r6, r11, r5
+; CHECK-NEXT: vstrw.32 q5, [sp, #40] @ 16-byte Spill
; CHECK-NEXT: vpstt
; CHECK-NEXT: vfmat.f32 q6, q0, q7
; CHECK-NEXT: vldrwt.u32 q0, [r11]
-; CHECK-NEXT: vstrw.32 q6, [sp, #40] @ 16-byte Spill
-; CHECK-NEXT: vmov q6, q5
-; CHECK-NEXT: vpst
-; CHECK-NEXT: vfmat.f32 q1, q0, q7
; CHECK-NEXT: vmov q5, q4
; CHECK-NEXT: vmov q4, q3
-; CHECK-NEXT: vmov q3, q1
-; CHECK-NEXT: vpst
+; CHECK-NEXT: vpstt
+; CHECK-NEXT: vfmat.f32 q1, q0, q7
; CHECK-NEXT: vldrwt.u32 q0, [r6]
+; CHECK-NEXT: vmov q3, q1
; CHECK-NEXT: vldrw.u32 q1, [sp, #56] @ 16-byte Reload
; CHECK-NEXT: adds r7, r6, r5
; CHECK-NEXT: vpstt
@@ -1122,8 +1120,7 @@ define void @DCT_mve7(ptr nocapture readonly %S, ptr nocapture readonly %pIn, pt
; CHECK-NEXT: vpstt
; CHECK-NEXT: vfmat.f32 q4, q0, q7
; CHECK-NEXT: vldrwt.u32 q0, [r7]
-; CHECK-NEXT: vmov q5, q6
-; CHECK-NEXT: vldrw.u32 q6, [sp, #40] @ 16-byte Reload
+; CHECK-NEXT: vldrw.u32 q5, [sp, #40] @ 16-byte Reload
; CHECK-NEXT: vpst
; CHECK-NEXT: vfmat.f32 q2, q0, q7
; CHECK-NEXT: le lr, .LBB6_3
diff --git a/llvm/test/CodeGen/Thumb2/mve-postinc-distribute.ll b/llvm/test/CodeGen/Thumb2/mve-postinc-distribute.ll
index 096d4382d2c35..bd0e5dabea3cf 100644
--- a/llvm/test/CodeGen/Thumb2/mve-postinc-distribute.ll
+++ b/llvm/test/CodeGen/Thumb2/mve-postinc-distribute.ll
@@ -8,23 +8,24 @@ define i32 @vaddv(ptr nocapture readonly %data, i32 %N) {
; CHECK: @ %bb.0: @ %entry
; CHECK-NEXT: .save {r7, lr}
; CHECK-NEXT: push {r7, lr}
-; CHECK-NEXT: mov lr, r1
; CHECK-NEXT: cmp r1, #1
; CHECK-NEXT: blt .LBB0_4
; CHECK-NEXT: @ %bb.1: @ %for.body.preheader
-; CHECK-NEXT: mov r1, r0
-; CHECK-NEXT: movs r0, #0
+; CHECK-NEXT: dls lr, r1
+; CHECK-NEXT: movs r2, #0
; CHECK-NEXT: .LBB0_2: @ %for.body
; CHECK-NEXT: @ =>This Inner Loop Header: Depth=1
-; CHECK-NEXT: vldrw.u32 q0, [r1], #32
-; CHECK-NEXT: vaddva.s32 r0, q0
-; CHECK-NEXT: vldrw.u32 q0, [r1, #-16]
-; CHECK-NEXT: vaddva.s32 r0, q0
+; CHECK-NEXT: vldrw.u32 q0, [r0], #32
+; CHECK-NEXT: vaddva.s32 r2, q0
+; CHECK-NEXT: vldrw.u32 q0, [r0, #-16]
+; CHECK-NEXT: vaddva.s32 r2, q0
; CHECK-NEXT: le lr, .LBB0_2
; CHECK-NEXT: @ %bb.3: @ %for.cond.cleanup
+; CHECK-NEXT: mov r0, r2
; CHECK-NEXT: pop {r7, pc}
; CHECK-NEXT: .LBB0_4:
-; CHECK-NEXT: movs r0, #0
+; CHECK-NEXT: movs r2, #0
+; CHECK-NEXT: mov r0, r2
; CHECK-NEXT: pop {r7, pc}
entry:
%cmp11 = icmp sgt i32 %N, 0
diff --git a/llvm/test/CodeGen/Thumb2/mve-postinc-lsr.ll b/llvm/test/CodeGen/Thumb2/mve-postinc-lsr.ll
index cba0f9cbba2ca..3e7ed0b096b82 100644
--- a/llvm/test/CodeGen/Thumb2/mve-postinc-lsr.ll
+++ b/llvm/test/CodeGen/Thumb2/mve-postinc-lsr.ll
@@ -212,9 +212,9 @@ define ptr @test(ptr nocapture readonly %input_row, ptr nocapture readonly %inpu
; CHECK-NEXT: beq .LBB2_8
; CHECK-NEXT: @ %bb.2: @ %for.body.lr.ph
; CHECK-NEXT: ldr r3, [sp, #64]
-; CHECK-NEXT: mov.w r9, #0
+; CHECK-NEXT: mov.w r11, #0
; CHECK-NEXT: ldr r1, [sp, #16] @ 4-byte Reload
-; CHECK-NEXT: ldr.w r11, [sp, #56]
+; CHECK-NEXT: ldr.w r9, [sp, #56]
; CHECK-NEXT: add.w r0, r1, r3, lsl #1
; CHECK-NEXT: str r0, [sp, #8] @ 4-byte Spill
; CHECK-NEXT: adds r0, r1, r3
@@ -235,15 +235,15 @@ define ptr @test(ptr nocapture readonly %input_row, ptr nocapture readonly %inpu
; CHECK-NEXT: add.w r1, r8, r10
; CHECK-NEXT: add r1, r6
; CHECK-NEXT: add r1, r12
-; CHECK-NEXT: strb.w r1, [r3, r9]
-; CHECK-NEXT: add.w r9, r9, #1
-; CHECK-NEXT: cmp r9, r2
+; CHECK-NEXT: strb.w r1, [r3, r11]
+; CHECK-NEXT: add.w r11, r11, #1
+; CHECK-NEXT: cmp r11, r2
; CHECK-NEXT: beq .LBB2_8
; CHECK-NEXT: .LBB2_5: @ %for.body
; CHECK-NEXT: @ =>This Loop Header: Depth=1
; CHECK-NEXT: @ Child Loop BB2_7 Depth 2
; CHECK-NEXT: ldr r1, [sp, #68]
-; CHECK-NEXT: ldr.w r12, [r1, r9, lsl #2]
+; CHECK-NEXT: ldr.w r12, [r1, r11, lsl #2]
; CHECK-NEXT: subs r1, r0, r0
; CHECK-NEXT: ble .LBB2_3
; CHECK-NEXT: @ %bb.6: @ %for.body24.preheader
@@ -254,7 +254,7 @@ define ptr @test(ptr nocapture readonly %input_row, ptr nocapture readonly %inpu
; CHECK-NEXT: dls lr, r1
; CHECK-NEXT: ldr r1, [sp, #16] @ 4-byte Reload
; CHECK-NEXT: mov r8, r12
-; CHECK-NEXT: mla r7, r9, r7, r3
+; CHECK-NEXT: mla r7, r11, r7, r3
; CHECK-NEXT: ldr r5, [sp, #8] @ 4-byte Reload
; CHECK-NEXT: ldrd r4, r3, [sp] @ 8-byte Folded Reload
; CHECK-NEXT: mov r10, r12
@@ -262,17 +262,17 @@ define ptr @test(ptr nocapture readonly %input_row, ptr nocapture readonly %inpu
; CHECK-NEXT: @ Parent Loop BB2_5 Depth=1
; CHECK-NEXT: @ => This Inner Loop Header: Depth=2
; CHECK-NEXT: vldrb.s16 q0, [r4], #8
-; CHECK-NEXT: vadd.i16 q1, q0, r11
+; CHECK-NEXT: vadd.i16 q1, q0, r9
; CHECK-NEXT: vldrb.s16 q0, [r7], #8
; CHECK-NEXT: vmlava.s16 r12, q0, q1
; CHECK-NEXT: vldrb.s16 q1, [r5], #8
-; CHECK-NEXT: vadd.i16 q1, q1, r11
+; CHECK-NEXT: vadd.i16 q1, q1, r9
; CHECK-NEXT: vmlava.s16 r6, q0, q1
; CHECK-NEXT: vldrb.s16 q1, [r3], #8
-; CHECK-NEXT: vadd.i16 q1, q1, r11
+; CHECK-NEXT: vadd.i16 q1, q1, r9
; CHECK-NEXT: vmlava.s16 r8, q0, q1
; CHECK-NEXT: vldrb.s16 q1, [r1], #8
-; CHECK-NEXT: vadd.i16 q1, q1, r11
+; CHECK-NEXT: vadd.i16 q1, q1, r9
; CHECK-NEXT: vmlava.s16 r10, q0, q1
; CHECK-NEXT: le lr, .LBB2_7
; CHECK-NEXT: b .LBB2_4
@@ -395,9 +395,9 @@ define ptr @test_optsize(ptr nocapture readonly %input_row, ptr nocapture readon
; CHECK-NEXT: beq .LBB3_8
; CHECK-NEXT: @ %bb.2: @ %for.body.lr.ph
; CHECK-NEXT: ldr r3, [sp, #64]
-; CHECK-NEXT: mov.w r9, #0
+; CHECK-NEXT: mov.w r11, #0
; CHECK-NEXT: ldr r1, [sp, #16] @ 4-byte Reload
-; CHECK-NEXT: ldr.w r11, [sp, #56]
+; CHECK-NEXT: ldr.w r9, [sp, #56]
; CHECK-NEXT: add.w r0, r1, r3, lsl #1
; CHECK-NEXT: str r0, [sp, #8] @ 4-byte Spill
; CHECK-NEXT: adds r0, r1, r3
@@ -411,7 +411,7 @@ define ptr @test_optsize(ptr nocapture readonly %input_row, ptr nocapture readon
; CHECK-NEXT: @ =>This Loop Header: Depth=1
; CHECK-NEXT: @ Child Loop BB3_5 Depth 2
; CHECK-NEXT: ldr r1, [sp, #68]
-; CHECK-NEXT: ldr.w r12, [r1, r9, lsl #2]
+; CHECK-NEXT: ldr.w r12, [r1, r11, lsl #2]
; CHECK-NEXT: subs r1, r0, r0
; CHECK-NEXT: ble .LBB3_6
; CHECK-NEXT: @ %bb.4: @ %for.body24.preheader
@@ -422,7 +422,7 @@ define ptr @test_optsize(ptr nocapture readonly %input_row, ptr nocapture readon
; CHECK-NEXT: dls lr, r1
; CHECK-NEXT: ldr r1, [sp, #16] @ 4-byte Reload
; CHECK-NEXT: mov r8, r12
-; CHECK-NEXT: mla r7, r9, r7, r3
+; CHECK-NEXT: mla r7, r11, r7, r3
; CHECK-NEXT: ldr r5, [sp, #8] @ 4-byte Reload
; CHECK-NEXT: ldrd r4, r3, [sp] @ 8-byte Folded Reload
; CHECK-NEXT: mov r10, r12
@@ -430,17 +430,17 @@ define ptr @test_optsize(ptr nocapture readonly %input_row, ptr nocapture readon
; CHECK-NEXT: @ Parent Loop BB3_3 Depth=1
; CHECK-NEXT: @ => This Inner Loop Header: Depth=2
; CHECK-NEXT: vldrb.s16 q0, [r4], #8
-; CHECK-NEXT: vadd.i16 q1, q0, r11
+; CHECK-NEXT: vadd.i16 q1, q0, r9
; CHECK-NEXT: vldrb.s16 q0, [r7], #8
; CHECK-NEXT: vmlava.s16 r12, q0, q1
; CHECK-NEXT: vldrb.s16 q1, [r5], #8
-; CHECK-NEXT: vadd.i16 q1, q1, r11
+; CHECK-NEXT: vadd.i16 q1, q1, r9
; CHECK-NEXT: vmlava.s16 r6, q0, q1
; CHECK-NEXT: vldrb.s16 q1, [r3], #8
-; CHECK-NEXT: vadd.i16 q1, q1, r11
+; CHECK-NEXT: vadd.i16 q1, q1, r9
; CHECK-NEXT: vmlava.s16 r8, q0, q1
; CHECK-NEXT: vldrb.s16 q1, [r1], #8
-; CHECK-NEXT: vadd.i16 q1, q1, r11
+; CHECK-NEXT: vadd.i16 q1, q1, r9
; CHECK-NEXT: vmlava.s16 r10, q0, q1
; CHECK-NEXT: le lr, .LBB3_5
; CHECK-NEXT: b .LBB3_7
@@ -454,9 +454,9 @@ define ptr @test_optsize(ptr nocapture readonly %input_row, ptr nocapture readon
; CHECK-NEXT: add.w r1, r8, r10
; CHECK-NEXT: add r1, r6
; CHECK-NEXT: add r1, r12
-; CHECK-NEXT: strb.w r1, [r3, r9]
-; CHECK-NEXT: add.w r9, r9, #1
-; CHECK-NEXT: cmp r9, r2
+; CHECK-NEXT: strb.w r1, [r3, r11]
+; CHECK-NEXT: add.w r11, r11, #1
+; CHECK-NEXT: cmp r11, r2
; CHECK-NEXT: bne .LBB3_3
; CHECK-NEXT: .LBB3_8: @ %if.end
; CHECK-NEXT: ldr r0, [sp, #72]
diff --git a/llvm/test/CodeGen/Thumb2/mve-satmul-loops.ll b/llvm/test/CodeGen/Thumb2/mve-satmul-loops.ll
index 29b56639bd769..16e7736886e94 100644
--- a/llvm/test/CodeGen/Thumb2/mve-satmul-loops.ll
+++ b/llvm/test/CodeGen/Thumb2/mve-satmul-loops.ll
@@ -12,12 +12,13 @@ define arm_aapcs_vfpcc void @ssatmul_s_q31(ptr nocapture readonly %pSrcA, ptr no
; CHECK-NEXT: beq.w .LBB0_8
; CHECK-NEXT: @ %bb.1: @ %entry
; CHECK-NEXT: mov r11, r2
+; CHECK-NEXT: mov r8, r1
; CHECK-NEXT: cmp r3, #1
; CHECK-NEXT: bne .LBB0_3
; CHECK-NEXT: @ %bb.2:
; CHECK-NEXT: movs r2, #0
; CHECK-NEXT: mov r12, r0
-; CHECK-NEXT: mov r8, r1
+; CHECK-NEXT: mov r1, r8
; CHECK-NEXT: mov r10, r11
; CHECK-NEXT: b .LBB0_6
; CHECK-NEXT: .LBB0_3: @ %vector.ph
@@ -29,7 +30,7 @@ define arm_aapcs_vfpcc void @ssatmul_s_q31(ptr nocapture readonly %pSrcA, ptr no
; CHECK-NEXT: add.w r10, r11, r2, lsl #2
; CHECK-NEXT: add.w lr, r6, r7, lsr #1
; CHECK-NEXT: str r2, [sp] @ 4-byte Spill
-; CHECK-NEXT: add.w r8, r1, r2, lsl #2
+; CHECK-NEXT: add.w r1, r8, r2, lsl #2
; CHECK-NEXT: add.w r12, r0, r2, lsl #2
; CHECK-NEXT: vldrw.u32 q0, [r4]
; CHECK-NEXT: vmvn.i32 q1, #0x80000000
@@ -37,7 +38,7 @@ define arm_aapcs_vfpcc void @ssatmul_s_q31(ptr nocapture readonly %pSrcA, ptr no
; CHECK-NEXT: @ =>This Inner Loop Header: Depth=1
; CHECK-NEXT: ldrd r4, r2, [r0], #8
; CHECK-NEXT: movs r5, #0
-; CHECK-NEXT: ldrd r7, r6, [r1], #8
+; CHECK-NEXT: ldrd r7, r6, [r8], #8
; CHECK-NEXT: smull r4, r7, r7, r4
; CHECK-NEXT: asrl r4, r7, #31
; CHECK-NEXT: rsbs.w r9, r4, #-2147483648
@@ -80,22 +81,22 @@ define arm_aapcs_vfpcc void @ssatmul_s_q31(ptr nocapture readonly %pSrcA, ptr no
; CHECK-NEXT: .LBB0_6: @ %for.body.preheader
; CHECK-NEXT: sub.w lr, r3, r2
; CHECK-NEXT: mov.w r0, #-1
-; CHECK-NEXT: mov.w r1, #-2147483648
+; CHECK-NEXT: mov.w r2, #-2147483648
; CHECK-NEXT: mvn r3, #-2147483648
; CHECK-NEXT: .LBB0_7: @ %for.body
; CHECK-NEXT: @ =>This Inner Loop Header: Depth=1
-; CHECK-NEXT: ldr r2, [r12], #4
-; CHECK-NEXT: ldr r4, [r8], #4
-; CHECK-NEXT: smull r2, r5, r4, r2
-; CHECK-NEXT: asrl r2, r5, #31
-; CHECK-NEXT: subs r4, r1, r2
-; CHECK-NEXT: sbcs.w r4, r0, r5
-; CHECK-NEXT: csel r2, r2, r1, lt
-; CHECK-NEXT: csel r4, r5, r0, lt
-; CHECK-NEXT: subs r5, r2, r3
-; CHECK-NEXT: sbcs r4, r4, #0
-; CHECK-NEXT: csel r2, r2, r3, lt
-; CHECK-NEXT: str r2, [r10], #4
+; CHECK-NEXT: ldr r4, [r12], #4
+; CHECK-NEXT: ldr r5, [r1], #4
+; CHECK-NEXT: smull r4, r5, r5, r4
+; CHECK-NEXT: asrl r4, r5, #31
+; CHECK-NEXT: subs r6, r2, r4
+; CHECK-NEXT: sbcs.w r6, r0, r5
+; CHECK-NEXT: csel r4, r4, r2, lt
+; CHECK-NEXT: csel r5, r5, r0, lt
+; CHECK-NEXT: subs r6, r4, r3
+; CHECK-NEXT: sbcs r5, r5, #0
+; CHECK-NEXT: csel r4, r4, r3, lt
+; CHECK-NEXT: str r4, [r10], #4
; CHECK-NEXT: le lr, .LBB0_7
; CHECK-NEXT: .LBB0_8: @ %for.cond.cleanup
; CHECK-NEXT: add sp, #8
diff --git a/llvm/test/CodeGen/Thumb2/pr52817.ll b/llvm/test/CodeGen/Thumb2/pr52817.ll
index 87615f0a1f7ef..4cc0960e1f57f 100644
--- a/llvm/test/CodeGen/Thumb2/pr52817.ll
+++ b/llvm/test/CodeGen/Thumb2/pr52817.ll
@@ -18,25 +18,25 @@ define i32 @test(ptr %arg, ptr %arg1, ptr %arg2) #0 !dbg !6 {
; CHECK-NEXT: push {r4, r5, r6, r7, lr}
; CHECK-NEXT: add r7, sp, #12
; CHECK-NEXT: str r8, [sp, #-4]!
-; CHECK-NEXT: mov.w lr, #0
-; CHECK-NEXT: mov.w r9, #1
-; CHECK-NEXT: movw r12, #4100
+; CHECK-NEXT: mov.w r9, #0
; CHECK-NEXT: movs r3, #0
+; CHECK-NEXT: mov.w r12, #1
+; CHECK-NEXT: movw lr, #4100
; CHECK-NEXT: LBB0_1: @ %bb3
; CHECK-NEXT: @ =>This Inner Loop Header: Depth=1
; CHECK-NEXT: adds r5, r3, #1
-; CHECK-NEXT: str.w lr, [r2]
-; CHECK-NEXT: cmp.w lr, #0
+; CHECK-NEXT: str.w r9, [r2]
+; CHECK-NEXT: cmp.w r9, #0
; CHECK-NEXT: add.w r4, r0, r5, lsl #2
-; CHECK-NEXT: add.w r8, r4, r12
-; CHECK-NEXT: lsl.w r4, r9, r3
+; CHECK-NEXT: add.w r8, r4, lr
+; CHECK-NEXT: lsl.w r4, r12, r3
; CHECK-NEXT: and.w r3, r3, r4
; CHECK-NEXT: add.w r4, r1, r5, lsl #2
; CHECK-NEXT: itte ne
; CHECK-NEXT: movne r6, #0
; CHECK-NEXT: Ltmp0:
; CHECK-NEXT: @DEBUG_VALUE: test:this <- [DW_OP_LLVM_arg 0, DW_OP_plus_uconst 135168, DW_OP_LLVM_arg 1, DW_OP_constu 4, DW_OP_mul, DW_OP_plus, DW_OP_plus_uconst 4, DW_OP_stack_value] $r0, $r5
-; CHECK-NEXT: .loc 1 28 24 prologue_end @ test.cpp:28:24
+; CHECK-NEXT: .loc 1 28 24 prologue_end @ test.cpp:28:24 @[ test.cpp:204:23 ]
; CHECK-NEXT: strne.w r6, [r8]
; CHECK-NEXT: moveq r6, #1
; CHECK-NEXT: ldr r4, [r4, #4]
diff --git a/llvm/test/CodeGen/VE/Scalar/br_jt.ll b/llvm/test/CodeGen/VE/Scalar/br_jt.ll
index fd880a7f42912..a418ef4892b33 100644
--- a/llvm/test/CodeGen/VE/Scalar/br_jt.ll
+++ b/llvm/test/CodeGen/VE/Scalar/br_jt.ll
@@ -641,53 +641,53 @@ define signext i32 @br_jt8_m(i32 signext %0, i32 signext %1) {
; PIC: # %bb.0:
; PIC-NEXT: st %s15, 24(, %s11)
; PIC-NEXT: st %s16, 32(, %s11)
-; PIC-NEXT: and %s2, %s0, (32)0
-; PIC-NEXT: adds.w.sx %s0, -1, %s2
-; PIC-NEXT: cmpu.w %s3, 8, %s0
+; PIC-NEXT: and %s0, %s0, (32)0
+; PIC-NEXT: adds.w.sx %s3, -1, %s0
+; PIC-NEXT: cmpu.w %s2, 8, %s3
; PIC-NEXT: lea %s15, _GLOBAL_OFFSET_TABLE_ at pc_lo(-24)
; PIC-NEXT: and %s15, %s15, (32)0
; PIC-NEXT: sic %s16
; PIC-NEXT: lea.sl %s15, _GLOBAL_OFFSET_TABLE_ at pc_hi(%s16, %s15)
-; PIC-NEXT: brgt.w 0, %s3, .LBB7_9
+; PIC-NEXT: brgt.w 0, %s2, .LBB7_9
; PIC-NEXT: # %bb.1:
-; PIC-NEXT: and %s1, %s1, (32)0
-; PIC-NEXT: adds.w.zx %s0, %s0, (0)1
-; PIC-NEXT: sll %s0, %s0, 2
+; PIC-NEXT: and %s2, %s1, (32)0
+; PIC-NEXT: adds.w.zx %s1, %s3, (0)1
+; PIC-NEXT: sll %s1, %s1, 2
; PIC-NEXT: lea %s3, .LJTI7_0 at gotoff_lo
; PIC-NEXT: and %s3, %s3, (32)0
; PIC-NEXT: lea.sl %s3, .LJTI7_0 at gotoff_hi(%s3, %s15)
-; PIC-NEXT: ldl.sx %s0, (%s0, %s3)
+; PIC-NEXT: ldl.sx %s1, (%s1, %s3)
; PIC-NEXT: lea %s3, br_jt8_m at gotoff_lo
; PIC-NEXT: and %s3, %s3, (32)0
; PIC-NEXT: lea.sl %s3, br_jt8_m at gotoff_hi(%s3, %s15)
-; PIC-NEXT: adds.l %s3, %s3, %s0
-; PIC-NEXT: or %s0, 3, (0)1
+; PIC-NEXT: adds.l %s3, %s3, %s1
+; PIC-NEXT: or %s1, 3, (0)1
; PIC-NEXT: b.l.t (, %s3)
; PIC-NEXT: .LBB7_2:
-; PIC-NEXT: or %s0, 0, (0)1
+; PIC-NEXT: or %s1, 0, (0)1
; PIC-NEXT: br.l.t .LBB7_10
; PIC-NEXT: .LBB7_9:
-; PIC-NEXT: or %s0, 0, %s2
+; PIC-NEXT: or %s1, 0, %s0
; PIC-NEXT: br.l.t .LBB7_10
; PIC-NEXT: .LBB7_6:
-; PIC-NEXT: adds.w.sx %s0, -2, %s1
+; PIC-NEXT: adds.w.sx %s1, -2, %s2
; PIC-NEXT: br.l.t .LBB7_10
; PIC-NEXT: .LBB7_8:
-; PIC-NEXT: or %s0, 11, (0)1
+; PIC-NEXT: or %s1, 11, (0)1
; PIC-NEXT: br.l.t .LBB7_10
; PIC-NEXT: .LBB7_7:
-; PIC-NEXT: or %s0, 10, (0)1
+; PIC-NEXT: or %s1, 10, (0)1
; PIC-NEXT: br.l.t .LBB7_10
; PIC-NEXT: .LBB7_3:
-; PIC-NEXT: or %s0, 4, (0)1
+; PIC-NEXT: or %s1, 4, (0)1
; PIC-NEXT: br.l.t .LBB7_10
; PIC-NEXT: .LBB7_4:
-; PIC-NEXT: adds.w.sx %s0, 3, %s1
+; PIC-NEXT: adds.w.sx %s1, 3, %s2
; PIC-NEXT: br.l.t .LBB7_10
; PIC-NEXT: .LBB7_5:
-; PIC-NEXT: adds.w.sx %s0, -5, %s1
+; PIC-NEXT: adds.w.sx %s1, -5, %s2
; PIC-NEXT: .LBB7_10:
-; PIC-NEXT: adds.w.sx %s0, %s0, (0)1
+; PIC-NEXT: adds.w.sx %s0, %s1, (0)1
; PIC-NEXT: ld %s16, 32(, %s11)
; PIC-NEXT: ld %s15, 24(, %s11)
; PIC-NEXT: b.l.t (, %s10)
diff --git a/llvm/test/CodeGen/X86/2012-01-10-UndefExceptionEdge.ll b/llvm/test/CodeGen/X86/2012-01-10-UndefExceptionEdge.ll
index 1962ddebc2115..99338d8f063f5 100644
--- a/llvm/test/CodeGen/X86/2012-01-10-UndefExceptionEdge.ll
+++ b/llvm/test/CodeGen/X86/2012-01-10-UndefExceptionEdge.ll
@@ -126,14 +126,14 @@ define void @f(ptr nocapture %arg, ptr nocapture %arg1, ptr nocapture %arg2, ptr
; CHECK-NEXT: decl {{[-0-9]+}}(%e{{[sb]}}p) ## 4-byte Folded Spill
; CHECK-NEXT: jmp LBB0_8
; CHECK-NEXT: LBB0_18: ## %bb43
-; CHECK-NEXT: Ltmp5:
; CHECK-NEXT: movl %esi, %ebx
+; CHECK-NEXT: Ltmp5:
; CHECK-NEXT: calll _OnOverFlow
; CHECK-NEXT: Ltmp6:
; CHECK-NEXT: jmp LBB0_3
; CHECK-NEXT: LBB0_2: ## %bb29
-; CHECK-NEXT: Ltmp7:
; CHECK-NEXT: movl %esi, %ebx
+; CHECK-NEXT: Ltmp7:
; CHECK-NEXT: calll _OnOverFlow
; CHECK-NEXT: Ltmp8:
; CHECK-NEXT: LBB0_3: ## %bb30
diff --git a/llvm/test/CodeGen/X86/AMX/amx-ldtilecfg-insert.ll b/llvm/test/CodeGen/X86/AMX/amx-ldtilecfg-insert.ll
index 06cf968512db8..8a8e7a3b4df2c 100644
--- a/llvm/test/CodeGen/X86/AMX/amx-ldtilecfg-insert.ll
+++ b/llvm/test/CodeGen/X86/AMX/amx-ldtilecfg-insert.ll
@@ -297,30 +297,30 @@ define dso_local void @test6(i16 signext %0) nounwind {
; CHECK-NEXT: movb $1, -{{[0-9]+}}(%rsp)
; CHECK-NEXT: movb %dil, -{{[0-9]+}}(%rsp)
; CHECK-NEXT: xorl %eax, %eax
-; CHECK-NEXT: movl $buf, %ecx
-; CHECK-NEXT: movl $32, %edx
-; CHECK-NEXT: xorl %esi, %esi
+; CHECK-NEXT: xorl %ecx, %ecx
+; CHECK-NEXT: movl $buf, %edx
+; CHECK-NEXT: movl $32, %esi
; CHECK-NEXT: jmp .LBB5_1
; CHECK-NEXT: .p2align 4
; CHECK-NEXT: .LBB5_3: # %if.false
; CHECK-NEXT: # in Loop: Header=BB5_1 Depth=1
-; CHECK-NEXT: decl %esi
+; CHECK-NEXT: decl %eax
; CHECK-NEXT: .LBB5_4: # %loop.bb2
; CHECK-NEXT: # in Loop: Header=BB5_1 Depth=1
-; CHECK-NEXT: leal (%rdi,%rsi), %r8d
+; CHECK-NEXT: leal (%rdi,%rax), %r8d
; CHECK-NEXT: movw %r8w, -{{[0-9]+}}(%rsp)
-; CHECK-NEXT: cmpw $7, %si
+; CHECK-NEXT: cmpw $7, %ax
; CHECK-NEXT: ldtilecfg -{{[0-9]+}}(%rsp)
; CHECK-NEXT: tilezero %tmm0
-; CHECK-NEXT: tilestored %tmm0, (%rcx,%rdx)
+; CHECK-NEXT: tilestored %tmm0, (%rdx,%rsi)
; CHECK-NEXT: jne .LBB5_5
; CHECK-NEXT: .LBB5_1: # %loop.bb1
; CHECK-NEXT: # =>This Inner Loop Header: Depth=1
-; CHECK-NEXT: testb %al, %al
+; CHECK-NEXT: testb %cl, %cl
; CHECK-NEXT: jne .LBB5_3
; CHECK-NEXT: # %bb.2: # %if.true
; CHECK-NEXT: # in Loop: Header=BB5_1 Depth=1
-; CHECK-NEXT: incl %esi
+; CHECK-NEXT: incl %eax
; CHECK-NEXT: jmp .LBB5_4
; CHECK-NEXT: .LBB5_5: # %exit
; CHECK-NEXT: tilerelease
diff --git a/llvm/test/CodeGen/X86/AMX/amx-spill-merge.ll b/llvm/test/CodeGen/X86/AMX/amx-spill-merge.ll
index 4fb0a4445862f..fda0244cea6eb 100644
--- a/llvm/test/CodeGen/X86/AMX/amx-spill-merge.ll
+++ b/llvm/test/CodeGen/X86/AMX/amx-spill-merge.ll
@@ -242,25 +242,25 @@ define dso_local void @test3(ptr%buf) nounwind {
; CHECK-NEXT: jne .LBB1_3
; CHECK-NEXT: # %bb.1: # %loop.header.preheader
; CHECK-NEXT: movq %rdi, %rbx
-; CHECK-NEXT: movl $32, %r14d
-; CHECK-NEXT: xorl %r15d, %r15d
+; CHECK-NEXT: xorl %r14d, %r14d
+; CHECK-NEXT: movl $32, %r15d
; CHECK-NEXT: .p2align 4
; CHECK-NEXT: .LBB1_2: # %loop.header
; CHECK-NEXT: # =>This Inner Loop Header: Depth=1
-; CHECK-NEXT: tilestored %tmm0, (%rbx,%r14)
+; CHECK-NEXT: tilestored %tmm0, (%rbx,%r15)
; CHECK-NEXT: xorl %eax, %eax
; CHECK-NEXT: tilezero %tmm0
; CHECK-NEXT: vzeroupper
; CHECK-NEXT: callq foo
; CHECK-NEXT: ldtilecfg {{[0-9]+}}(%rsp)
; CHECK-NEXT: tilezero %tmm0
-; CHECK-NEXT: tileloadd (%rbx,%r14), %tmm1
-; CHECK-NEXT: tileloadd (%rbx,%r14), %tmm2
+; CHECK-NEXT: tileloadd (%rbx,%r15), %tmm1
+; CHECK-NEXT: tileloadd (%rbx,%r15), %tmm2
; CHECK-NEXT: tdpbssd %tmm2, %tmm1, %tmm0
-; CHECK-NEXT: tilestored %tmm0, (%rbx,%r14)
+; CHECK-NEXT: tilestored %tmm0, (%rbx,%r15)
; CHECK-NEXT: tilezero %tmm0
-; CHECK-NEXT: incl %r15d
-; CHECK-NEXT: cmpw $100, %r15w
+; CHECK-NEXT: incl %r14d
+; CHECK-NEXT: cmpw $100, %r14w
; CHECK-NEXT: jl .LBB1_2
; CHECK-NEXT: .LBB1_3: # %exit
; CHECK-NEXT: addq $72, %rsp
@@ -297,12 +297,12 @@ define dso_local void @test3(ptr%buf) nounwind {
; EGPR-NEXT: # fixup A - offset: 1, value: .LBB1_3-1, kind: FK_PCRel_1
; EGPR-NEXT: # %bb.1: # %loop.header.preheader
; EGPR-NEXT: movq %rdi, %rbx # encoding: [0x48,0x89,0xfb]
-; EGPR-NEXT: movl $32, %r14d # encoding: [0x41,0xbe,0x20,0x00,0x00,0x00]
-; EGPR-NEXT: xorl %r15d, %r15d # encoding: [0x45,0x31,0xff]
+; EGPR-NEXT: xorl %r14d, %r14d # encoding: [0x45,0x31,0xf6]
+; EGPR-NEXT: movl $32, %r15d # encoding: [0x41,0xbf,0x20,0x00,0x00,0x00]
; EGPR-NEXT: .p2align 4
; EGPR-NEXT: .LBB1_2: # %loop.header
; EGPR-NEXT: # =>This Inner Loop Header: Depth=1
-; EGPR-NEXT: tilestored %tmm0, (%rbx,%r14) # EVEX TO VEX Compression encoding: [0xc4,0xa2,0x7a,0x4b,0x04,0x33]
+; EGPR-NEXT: tilestored %tmm0, (%rbx,%r15) # EVEX TO VEX Compression encoding: [0xc4,0xa2,0x7a,0x4b,0x04,0x3b]
; EGPR-NEXT: xorl %eax, %eax # encoding: [0x31,0xc0]
; EGPR-NEXT: tilezero %tmm0 # encoding: [0xc4,0xe2,0x7b,0x49,0xc0]
; EGPR-NEXT: vzeroupper # encoding: [0xc5,0xf8,0x77]
@@ -310,13 +310,13 @@ define dso_local void @test3(ptr%buf) nounwind {
; EGPR-NEXT: # fixup A - offset: 1, value: foo-4, kind: reloc_branch_4byte_pcrel
; EGPR-NEXT: ldtilecfg {{[0-9]+}}(%rsp) # EVEX TO VEX Compression encoding: [0xc4,0xe2,0x78,0x49,0x44,0x24,0x08]
; EGPR-NEXT: tilezero %tmm0 # encoding: [0xc4,0xe2,0x7b,0x49,0xc0]
-; EGPR-NEXT: tileloadd (%rbx,%r14), %tmm1 # EVEX TO VEX Compression encoding: [0xc4,0xa2,0x7b,0x4b,0x0c,0x33]
-; EGPR-NEXT: tileloadd (%rbx,%r14), %tmm2 # EVEX TO VEX Compression encoding: [0xc4,0xa2,0x7b,0x4b,0x14,0x33]
+; EGPR-NEXT: tileloadd (%rbx,%r15), %tmm1 # EVEX TO VEX Compression encoding: [0xc4,0xa2,0x7b,0x4b,0x0c,0x3b]
+; EGPR-NEXT: tileloadd (%rbx,%r15), %tmm2 # EVEX TO VEX Compression encoding: [0xc4,0xa2,0x7b,0x4b,0x14,0x3b]
; EGPR-NEXT: tdpbssd %tmm2, %tmm1, %tmm0 # encoding: [0xc4,0xe2,0x6b,0x5e,0xc1]
-; EGPR-NEXT: tilestored %tmm0, (%rbx,%r14) # EVEX TO VEX Compression encoding: [0xc4,0xa2,0x7a,0x4b,0x04,0x33]
+; EGPR-NEXT: tilestored %tmm0, (%rbx,%r15) # EVEX TO VEX Compression encoding: [0xc4,0xa2,0x7a,0x4b,0x04,0x3b]
; EGPR-NEXT: tilezero %tmm0 # encoding: [0xc4,0xe2,0x7b,0x49,0xc0]
-; EGPR-NEXT: incl %r15d # encoding: [0x41,0xff,0xc7]
-; EGPR-NEXT: cmpw $100, %r15w # encoding: [0x66,0x41,0x83,0xff,0x64]
+; EGPR-NEXT: incl %r14d # encoding: [0x41,0xff,0xc6]
+; EGPR-NEXT: cmpw $100, %r14w # encoding: [0x66,0x41,0x83,0xfe,0x64]
; EGPR-NEXT: jl .LBB1_2 # encoding: [0x7c,A]
; EGPR-NEXT: # fixup A - offset: 1, value: .LBB1_2-1, kind: FK_PCRel_1
; EGPR-NEXT: .LBB1_3: # %exit
diff --git a/llvm/test/CodeGen/X86/atomic32.ll b/llvm/test/CodeGen/X86/atomic32.ll
index f4666738db7d2..3263b9f61a281 100644
--- a/llvm/test/CodeGen/X86/atomic32.ll
+++ b/llvm/test/CodeGen/X86/atomic32.ll
@@ -228,9 +228,10 @@ define void @atomic_fetch_nand32(i32 %x) nounwind {
; X64-NEXT: andl %edx, %ecx
; X64-NEXT: notl %ecx
; X64-NEXT: lock cmpxchgl %ecx, sc32(%rip)
-; X64-NEXT: sete %cl
-; X64-NEXT: testb $1, %cl
-; X64-NEXT: movl %eax, {{[-0-9]+}}(%r{{[sb]}}p) # 4-byte Spill
+; X64-NEXT: movl %eax, %ecx
+; X64-NEXT: sete %al
+; X64-NEXT: movl %ecx, {{[-0-9]+}}(%r{{[sb]}}p) # 4-byte Spill
+; X64-NEXT: testb $1, %al
; X64-NEXT: jne .LBB5_2
; X64-NEXT: jmp .LBB5_1
; X64-NEXT: .LBB5_2: # %atomicrmw.end
@@ -251,9 +252,10 @@ define void @atomic_fetch_nand32(i32 %x) nounwind {
; X86-NEXT: andl %edx, %ecx
; X86-NEXT: notl %ecx
; X86-NEXT: lock cmpxchgl %ecx, sc32
-; X86-NEXT: sete %cl
-; X86-NEXT: testb $1, %cl
-; X86-NEXT: movl %eax, {{[-0-9]+}}(%e{{[sb]}}p) # 4-byte Spill
+; X86-NEXT: movl %eax, %ecx
+; X86-NEXT: sete %al
+; X86-NEXT: movl %ecx, {{[-0-9]+}}(%e{{[sb]}}p) # 4-byte Spill
+; X86-NEXT: testb $1, %al
; X86-NEXT: jne .LBB5_2
; X86-NEXT: jmp .LBB5_1
; X86-NEXT: .LBB5_2: # %atomicrmw.end
@@ -277,9 +279,10 @@ define void @atomic_fetch_max32(i32 %x) nounwind {
; X64-NEXT: subl %ecx, %edx
; X64-NEXT: cmovgl %eax, %ecx
; X64-NEXT: lock cmpxchgl %ecx, sc32(%rip)
-; X64-NEXT: sete %cl
-; X64-NEXT: testb $1, %cl
-; X64-NEXT: movl %eax, {{[-0-9]+}}(%r{{[sb]}}p) # 4-byte Spill
+; X64-NEXT: movl %eax, %ecx
+; X64-NEXT: sete %al
+; X64-NEXT: movl %ecx, {{[-0-9]+}}(%r{{[sb]}}p) # 4-byte Spill
+; X64-NEXT: testb $1, %al
; X64-NEXT: jne .LBB6_2
; X64-NEXT: jmp .LBB6_1
; X64-NEXT: .LBB6_2: # %atomicrmw.end
@@ -300,9 +303,10 @@ define void @atomic_fetch_max32(i32 %x) nounwind {
; X86-CMOV-NEXT: subl %ecx, %edx
; X86-CMOV-NEXT: cmovgl %eax, %ecx
; X86-CMOV-NEXT: lock cmpxchgl %ecx, sc32
-; X86-CMOV-NEXT: sete %cl
-; X86-CMOV-NEXT: testb $1, %cl
-; X86-CMOV-NEXT: movl %eax, {{[-0-9]+}}(%e{{[sb]}}p) # 4-byte Spill
+; X86-CMOV-NEXT: movl %eax, %ecx
+; X86-CMOV-NEXT: sete %al
+; X86-CMOV-NEXT: movl %ecx, {{[-0-9]+}}(%e{{[sb]}}p) # 4-byte Spill
+; X86-CMOV-NEXT: testb $1, %al
; X86-CMOV-NEXT: jne .LBB6_2
; X86-CMOV-NEXT: jmp .LBB6_1
; X86-CMOV-NEXT: .LBB6_2: # %atomicrmw.end
@@ -334,9 +338,10 @@ define void @atomic_fetch_max32(i32 %x) nounwind {
; X86-NOCMOV-NEXT: movl (%esp), %eax # 4-byte Reload
; X86-NOCMOV-NEXT: movl {{[-0-9]+}}(%e{{[sb]}}p), %ecx # 4-byte Reload
; X86-NOCMOV-NEXT: lock cmpxchgl %ecx, sc32
-; X86-NOCMOV-NEXT: sete %cl
-; X86-NOCMOV-NEXT: testb $1, %cl
-; X86-NOCMOV-NEXT: movl %eax, {{[-0-9]+}}(%e{{[sb]}}p) # 4-byte Spill
+; X86-NOCMOV-NEXT: movl %eax, %ecx
+; X86-NOCMOV-NEXT: sete %al
+; X86-NOCMOV-NEXT: movl %ecx, {{[-0-9]+}}(%e{{[sb]}}p) # 4-byte Spill
+; X86-NOCMOV-NEXT: testb $1, %al
; X86-NOCMOV-NEXT: jne .LBB6_2
; X86-NOCMOV-NEXT: jmp .LBB6_1
; X86-NOCMOV-NEXT: .LBB6_2: # %atomicrmw.end
@@ -368,9 +373,10 @@ define void @atomic_fetch_max32(i32 %x) nounwind {
; X86-NOX87-NEXT: movl (%esp), %eax # 4-byte Reload
; X86-NOX87-NEXT: movl {{[-0-9]+}}(%e{{[sb]}}p), %ecx # 4-byte Reload
; X86-NOX87-NEXT: lock cmpxchgl %ecx, sc32
-; X86-NOX87-NEXT: sete %cl
-; X86-NOX87-NEXT: testb $1, %cl
-; X86-NOX87-NEXT: movl %eax, {{[-0-9]+}}(%e{{[sb]}}p) # 4-byte Spill
+; X86-NOX87-NEXT: movl %eax, %ecx
+; X86-NOX87-NEXT: sete %al
+; X86-NOX87-NEXT: movl %ecx, {{[-0-9]+}}(%e{{[sb]}}p) # 4-byte Spill
+; X86-NOX87-NEXT: testb $1, %al
; X86-NOX87-NEXT: jne .LBB6_2
; X86-NOX87-NEXT: jmp .LBB6_1
; X86-NOX87-NEXT: .LBB6_2: # %atomicrmw.end
@@ -394,9 +400,10 @@ define void @atomic_fetch_min32(i32 %x) nounwind {
; X64-NEXT: subl %ecx, %edx
; X64-NEXT: cmovlel %eax, %ecx
; X64-NEXT: lock cmpxchgl %ecx, sc32(%rip)
-; X64-NEXT: sete %cl
-; X64-NEXT: testb $1, %cl
-; X64-NEXT: movl %eax, {{[-0-9]+}}(%r{{[sb]}}p) # 4-byte Spill
+; X64-NEXT: movl %eax, %ecx
+; X64-NEXT: sete %al
+; X64-NEXT: movl %ecx, {{[-0-9]+}}(%r{{[sb]}}p) # 4-byte Spill
+; X64-NEXT: testb $1, %al
; X64-NEXT: jne .LBB7_2
; X64-NEXT: jmp .LBB7_1
; X64-NEXT: .LBB7_2: # %atomicrmw.end
@@ -417,9 +424,10 @@ define void @atomic_fetch_min32(i32 %x) nounwind {
; X86-CMOV-NEXT: subl %ecx, %edx
; X86-CMOV-NEXT: cmovlel %eax, %ecx
; X86-CMOV-NEXT: lock cmpxchgl %ecx, sc32
-; X86-CMOV-NEXT: sete %cl
-; X86-CMOV-NEXT: testb $1, %cl
-; X86-CMOV-NEXT: movl %eax, {{[-0-9]+}}(%e{{[sb]}}p) # 4-byte Spill
+; X86-CMOV-NEXT: movl %eax, %ecx
+; X86-CMOV-NEXT: sete %al
+; X86-CMOV-NEXT: movl %ecx, {{[-0-9]+}}(%e{{[sb]}}p) # 4-byte Spill
+; X86-CMOV-NEXT: testb $1, %al
; X86-CMOV-NEXT: jne .LBB7_2
; X86-CMOV-NEXT: jmp .LBB7_1
; X86-CMOV-NEXT: .LBB7_2: # %atomicrmw.end
@@ -451,9 +459,10 @@ define void @atomic_fetch_min32(i32 %x) nounwind {
; X86-NOCMOV-NEXT: movl (%esp), %eax # 4-byte Reload
; X86-NOCMOV-NEXT: movl {{[-0-9]+}}(%e{{[sb]}}p), %ecx # 4-byte Reload
; X86-NOCMOV-NEXT: lock cmpxchgl %ecx, sc32
-; X86-NOCMOV-NEXT: sete %cl
-; X86-NOCMOV-NEXT: testb $1, %cl
-; X86-NOCMOV-NEXT: movl %eax, {{[-0-9]+}}(%e{{[sb]}}p) # 4-byte Spill
+; X86-NOCMOV-NEXT: movl %eax, %ecx
+; X86-NOCMOV-NEXT: sete %al
+; X86-NOCMOV-NEXT: movl %ecx, {{[-0-9]+}}(%e{{[sb]}}p) # 4-byte Spill
+; X86-NOCMOV-NEXT: testb $1, %al
; X86-NOCMOV-NEXT: jne .LBB7_2
; X86-NOCMOV-NEXT: jmp .LBB7_1
; X86-NOCMOV-NEXT: .LBB7_2: # %atomicrmw.end
@@ -485,9 +494,10 @@ define void @atomic_fetch_min32(i32 %x) nounwind {
; X86-NOX87-NEXT: movl (%esp), %eax # 4-byte Reload
; X86-NOX87-NEXT: movl {{[-0-9]+}}(%e{{[sb]}}p), %ecx # 4-byte Reload
; X86-NOX87-NEXT: lock cmpxchgl %ecx, sc32
-; X86-NOX87-NEXT: sete %cl
-; X86-NOX87-NEXT: testb $1, %cl
-; X86-NOX87-NEXT: movl %eax, {{[-0-9]+}}(%e{{[sb]}}p) # 4-byte Spill
+; X86-NOX87-NEXT: movl %eax, %ecx
+; X86-NOX87-NEXT: sete %al
+; X86-NOX87-NEXT: movl %ecx, {{[-0-9]+}}(%e{{[sb]}}p) # 4-byte Spill
+; X86-NOX87-NEXT: testb $1, %al
; X86-NOX87-NEXT: jne .LBB7_2
; X86-NOX87-NEXT: jmp .LBB7_1
; X86-NOX87-NEXT: .LBB7_2: # %atomicrmw.end
@@ -511,9 +521,10 @@ define void @atomic_fetch_umax32(i32 %x) nounwind {
; X64-NEXT: subl %ecx, %edx
; X64-NEXT: cmoval %eax, %ecx
; X64-NEXT: lock cmpxchgl %ecx, sc32(%rip)
-; X64-NEXT: sete %cl
-; X64-NEXT: testb $1, %cl
-; X64-NEXT: movl %eax, {{[-0-9]+}}(%r{{[sb]}}p) # 4-byte Spill
+; X64-NEXT: movl %eax, %ecx
+; X64-NEXT: sete %al
+; X64-NEXT: movl %ecx, {{[-0-9]+}}(%r{{[sb]}}p) # 4-byte Spill
+; X64-NEXT: testb $1, %al
; X64-NEXT: jne .LBB8_2
; X64-NEXT: jmp .LBB8_1
; X64-NEXT: .LBB8_2: # %atomicrmw.end
@@ -534,9 +545,10 @@ define void @atomic_fetch_umax32(i32 %x) nounwind {
; X86-CMOV-NEXT: subl %ecx, %edx
; X86-CMOV-NEXT: cmoval %eax, %ecx
; X86-CMOV-NEXT: lock cmpxchgl %ecx, sc32
-; X86-CMOV-NEXT: sete %cl
-; X86-CMOV-NEXT: testb $1, %cl
-; X86-CMOV-NEXT: movl %eax, {{[-0-9]+}}(%e{{[sb]}}p) # 4-byte Spill
+; X86-CMOV-NEXT: movl %eax, %ecx
+; X86-CMOV-NEXT: sete %al
+; X86-CMOV-NEXT: movl %ecx, {{[-0-9]+}}(%e{{[sb]}}p) # 4-byte Spill
+; X86-CMOV-NEXT: testb $1, %al
; X86-CMOV-NEXT: jne .LBB8_2
; X86-CMOV-NEXT: jmp .LBB8_1
; X86-CMOV-NEXT: .LBB8_2: # %atomicrmw.end
@@ -568,9 +580,10 @@ define void @atomic_fetch_umax32(i32 %x) nounwind {
; X86-NOCMOV-NEXT: movl (%esp), %eax # 4-byte Reload
; X86-NOCMOV-NEXT: movl {{[-0-9]+}}(%e{{[sb]}}p), %ecx # 4-byte Reload
; X86-NOCMOV-NEXT: lock cmpxchgl %ecx, sc32
-; X86-NOCMOV-NEXT: sete %cl
-; X86-NOCMOV-NEXT: testb $1, %cl
-; X86-NOCMOV-NEXT: movl %eax, {{[-0-9]+}}(%e{{[sb]}}p) # 4-byte Spill
+; X86-NOCMOV-NEXT: movl %eax, %ecx
+; X86-NOCMOV-NEXT: sete %al
+; X86-NOCMOV-NEXT: movl %ecx, {{[-0-9]+}}(%e{{[sb]}}p) # 4-byte Spill
+; X86-NOCMOV-NEXT: testb $1, %al
; X86-NOCMOV-NEXT: jne .LBB8_2
; X86-NOCMOV-NEXT: jmp .LBB8_1
; X86-NOCMOV-NEXT: .LBB8_2: # %atomicrmw.end
@@ -602,9 +615,10 @@ define void @atomic_fetch_umax32(i32 %x) nounwind {
; X86-NOX87-NEXT: movl (%esp), %eax # 4-byte Reload
; X86-NOX87-NEXT: movl {{[-0-9]+}}(%e{{[sb]}}p), %ecx # 4-byte Reload
; X86-NOX87-NEXT: lock cmpxchgl %ecx, sc32
-; X86-NOX87-NEXT: sete %cl
-; X86-NOX87-NEXT: testb $1, %cl
-; X86-NOX87-NEXT: movl %eax, {{[-0-9]+}}(%e{{[sb]}}p) # 4-byte Spill
+; X86-NOX87-NEXT: movl %eax, %ecx
+; X86-NOX87-NEXT: sete %al
+; X86-NOX87-NEXT: movl %ecx, {{[-0-9]+}}(%e{{[sb]}}p) # 4-byte Spill
+; X86-NOX87-NEXT: testb $1, %al
; X86-NOX87-NEXT: jne .LBB8_2
; X86-NOX87-NEXT: jmp .LBB8_1
; X86-NOX87-NEXT: .LBB8_2: # %atomicrmw.end
@@ -628,9 +642,10 @@ define void @atomic_fetch_umin32(i32 %x) nounwind {
; X64-NEXT: subl %ecx, %edx
; X64-NEXT: cmovbel %eax, %ecx
; X64-NEXT: lock cmpxchgl %ecx, sc32(%rip)
-; X64-NEXT: sete %cl
-; X64-NEXT: testb $1, %cl
-; X64-NEXT: movl %eax, {{[-0-9]+}}(%r{{[sb]}}p) # 4-byte Spill
+; X64-NEXT: movl %eax, %ecx
+; X64-NEXT: sete %al
+; X64-NEXT: movl %ecx, {{[-0-9]+}}(%r{{[sb]}}p) # 4-byte Spill
+; X64-NEXT: testb $1, %al
; X64-NEXT: jne .LBB9_2
; X64-NEXT: jmp .LBB9_1
; X64-NEXT: .LBB9_2: # %atomicrmw.end
@@ -651,9 +666,10 @@ define void @atomic_fetch_umin32(i32 %x) nounwind {
; X86-CMOV-NEXT: subl %ecx, %edx
; X86-CMOV-NEXT: cmovbel %eax, %ecx
; X86-CMOV-NEXT: lock cmpxchgl %ecx, sc32
-; X86-CMOV-NEXT: sete %cl
-; X86-CMOV-NEXT: testb $1, %cl
-; X86-CMOV-NEXT: movl %eax, {{[-0-9]+}}(%e{{[sb]}}p) # 4-byte Spill
+; X86-CMOV-NEXT: movl %eax, %ecx
+; X86-CMOV-NEXT: sete %al
+; X86-CMOV-NEXT: movl %ecx, {{[-0-9]+}}(%e{{[sb]}}p) # 4-byte Spill
+; X86-CMOV-NEXT: testb $1, %al
; X86-CMOV-NEXT: jne .LBB9_2
; X86-CMOV-NEXT: jmp .LBB9_1
; X86-CMOV-NEXT: .LBB9_2: # %atomicrmw.end
@@ -685,9 +701,10 @@ define void @atomic_fetch_umin32(i32 %x) nounwind {
; X86-NOCMOV-NEXT: movl (%esp), %eax # 4-byte Reload
; X86-NOCMOV-NEXT: movl {{[-0-9]+}}(%e{{[sb]}}p), %ecx # 4-byte Reload
; X86-NOCMOV-NEXT: lock cmpxchgl %ecx, sc32
-; X86-NOCMOV-NEXT: sete %cl
-; X86-NOCMOV-NEXT: testb $1, %cl
-; X86-NOCMOV-NEXT: movl %eax, {{[-0-9]+}}(%e{{[sb]}}p) # 4-byte Spill
+; X86-NOCMOV-NEXT: movl %eax, %ecx
+; X86-NOCMOV-NEXT: sete %al
+; X86-NOCMOV-NEXT: movl %ecx, {{[-0-9]+}}(%e{{[sb]}}p) # 4-byte Spill
+; X86-NOCMOV-NEXT: testb $1, %al
; X86-NOCMOV-NEXT: jne .LBB9_2
; X86-NOCMOV-NEXT: jmp .LBB9_1
; X86-NOCMOV-NEXT: .LBB9_2: # %atomicrmw.end
@@ -719,9 +736,10 @@ define void @atomic_fetch_umin32(i32 %x) nounwind {
; X86-NOX87-NEXT: movl (%esp), %eax # 4-byte Reload
; X86-NOX87-NEXT: movl {{[-0-9]+}}(%e{{[sb]}}p), %ecx # 4-byte Reload
; X86-NOX87-NEXT: lock cmpxchgl %ecx, sc32
-; X86-NOX87-NEXT: sete %cl
-; X86-NOX87-NEXT: testb $1, %cl
-; X86-NOX87-NEXT: movl %eax, {{[-0-9]+}}(%e{{[sb]}}p) # 4-byte Spill
+; X86-NOX87-NEXT: movl %eax, %ecx
+; X86-NOX87-NEXT: sete %al
+; X86-NOX87-NEXT: movl %ecx, {{[-0-9]+}}(%e{{[sb]}}p) # 4-byte Spill
+; X86-NOX87-NEXT: testb $1, %al
; X86-NOX87-NEXT: jne .LBB9_2
; X86-NOX87-NEXT: jmp .LBB9_1
; X86-NOX87-NEXT: .LBB9_2: # %atomicrmw.end
diff --git a/llvm/test/CodeGen/X86/atomic64.ll b/llvm/test/CodeGen/X86/atomic64.ll
index 8f4da356e06cb..6fc72bcf67ec5 100644
--- a/llvm/test/CodeGen/X86/atomic64.ll
+++ b/llvm/test/CodeGen/X86/atomic64.ll
@@ -275,9 +275,10 @@ define void @atomic_fetch_nand64(i64 %x) nounwind {
; X64-NEXT: andq %rdx, %rcx
; X64-NEXT: notq %rcx
; X64-NEXT: lock cmpxchgq %rcx, sc64(%rip)
-; X64-NEXT: sete %cl
-; X64-NEXT: testb $1, %cl
-; X64-NEXT: movq %rax, {{[-0-9]+}}(%r{{[sb]}}p) # 8-byte Spill
+; X64-NEXT: movq %rax, %rcx
+; X64-NEXT: sete %al
+; X64-NEXT: movq %rcx, {{[-0-9]+}}(%r{{[sb]}}p) # 8-byte Spill
+; X64-NEXT: testb $1, %al
; X64-NEXT: jne .LBB5_2
; X64-NEXT: jmp .LBB5_1
; X64-NEXT: .LBB5_2: # %atomicrmw.end
@@ -314,9 +315,10 @@ define void @atomic_fetch_max64(i64 %x) nounwind {
; X64-NEXT: subq %rcx, %rdx
; X64-NEXT: cmovgq %rax, %rcx
; X64-NEXT: lock cmpxchgq %rcx, sc64(%rip)
-; X64-NEXT: sete %cl
-; X64-NEXT: testb $1, %cl
-; X64-NEXT: movq %rax, {{[-0-9]+}}(%r{{[sb]}}p) # 8-byte Spill
+; X64-NEXT: movq %rax, %rcx
+; X64-NEXT: sete %al
+; X64-NEXT: movq %rcx, {{[-0-9]+}}(%r{{[sb]}}p) # 8-byte Spill
+; X64-NEXT: testb $1, %al
; X64-NEXT: jne .LBB6_2
; X64-NEXT: jmp .LBB6_1
; X64-NEXT: .LBB6_2: # %atomicrmw.end
@@ -406,9 +408,10 @@ define void @atomic_fetch_min64(i64 %x) nounwind {
; X64-NEXT: subq %rcx, %rdx
; X64-NEXT: cmovleq %rax, %rcx
; X64-NEXT: lock cmpxchgq %rcx, sc64(%rip)
-; X64-NEXT: sete %cl
-; X64-NEXT: testb $1, %cl
-; X64-NEXT: movq %rax, {{[-0-9]+}}(%r{{[sb]}}p) # 8-byte Spill
+; X64-NEXT: movq %rax, %rcx
+; X64-NEXT: sete %al
+; X64-NEXT: movq %rcx, {{[-0-9]+}}(%r{{[sb]}}p) # 8-byte Spill
+; X64-NEXT: testb $1, %al
; X64-NEXT: jne .LBB7_2
; X64-NEXT: jmp .LBB7_1
; X64-NEXT: .LBB7_2: # %atomicrmw.end
@@ -498,9 +501,10 @@ define void @atomic_fetch_umax64(i64 %x) nounwind {
; X64-NEXT: subq %rcx, %rdx
; X64-NEXT: cmovaq %rax, %rcx
; X64-NEXT: lock cmpxchgq %rcx, sc64(%rip)
-; X64-NEXT: sete %cl
-; X64-NEXT: testb $1, %cl
-; X64-NEXT: movq %rax, {{[-0-9]+}}(%r{{[sb]}}p) # 8-byte Spill
+; X64-NEXT: movq %rax, %rcx
+; X64-NEXT: sete %al
+; X64-NEXT: movq %rcx, {{[-0-9]+}}(%r{{[sb]}}p) # 8-byte Spill
+; X64-NEXT: testb $1, %al
; X64-NEXT: jne .LBB8_2
; X64-NEXT: jmp .LBB8_1
; X64-NEXT: .LBB8_2: # %atomicrmw.end
@@ -590,9 +594,10 @@ define void @atomic_fetch_umin64(i64 %x) nounwind {
; X64-NEXT: subq %rcx, %rdx
; X64-NEXT: cmovbeq %rax, %rcx
; X64-NEXT: lock cmpxchgq %rcx, sc64(%rip)
-; X64-NEXT: sete %cl
-; X64-NEXT: testb $1, %cl
-; X64-NEXT: movq %rax, {{[-0-9]+}}(%r{{[sb]}}p) # 8-byte Spill
+; X64-NEXT: movq %rax, %rcx
+; X64-NEXT: sete %al
+; X64-NEXT: movq %rcx, {{[-0-9]+}}(%r{{[sb]}}p) # 8-byte Spill
+; X64-NEXT: testb $1, %al
; X64-NEXT: jne .LBB9_2
; X64-NEXT: jmp .LBB9_1
; X64-NEXT: .LBB9_2: # %atomicrmw.end
diff --git a/llvm/test/CodeGen/X86/atomic6432.ll b/llvm/test/CodeGen/X86/atomic6432.ll
index 8ff5f338e1482..3d0617370a1bb 100644
--- a/llvm/test/CodeGen/X86/atomic6432.ll
+++ b/llvm/test/CodeGen/X86/atomic6432.ll
@@ -16,15 +16,15 @@ define void @atomic_fetch_add64() nounwind {
; X32-NEXT: jmp .LBB0_1
; X32-NEXT: .LBB0_1: # %atomicrmw.start14
; X32-NEXT: # =>This Inner Loop Header: Depth=1
-; X32-NEXT: movl {{[-0-9]+}}(%e{{[sb]}}p), %edx # 4-byte Reload
; X32-NEXT: movl {{[-0-9]+}}(%e{{[sb]}}p), %eax # 4-byte Reload
+; X32-NEXT: movl {{[-0-9]+}}(%e{{[sb]}}p), %edx # 4-byte Reload
; X32-NEXT: movl %eax, %ebx
; X32-NEXT: addl $1, %ebx
; X32-NEXT: movl %edx, %ecx
; X32-NEXT: adcl $0, %ecx
; X32-NEXT: lock cmpxchg8b sc64
-; X32-NEXT: movl %eax, {{[-0-9]+}}(%e{{[sb]}}p) # 4-byte Spill
; X32-NEXT: movl %edx, {{[-0-9]+}}(%e{{[sb]}}p) # 4-byte Spill
+; X32-NEXT: movl %eax, {{[-0-9]+}}(%e{{[sb]}}p) # 4-byte Spill
; X32-NEXT: jne .LBB0_1
; X32-NEXT: jmp .LBB0_2
; X32-NEXT: .LBB0_2: # %atomicrmw.end13
@@ -35,15 +35,15 @@ define void @atomic_fetch_add64() nounwind {
; X32-NEXT: jmp .LBB0_3
; X32-NEXT: .LBB0_3: # %atomicrmw.start8
; X32-NEXT: # =>This Inner Loop Header: Depth=1
-; X32-NEXT: movl {{[-0-9]+}}(%e{{[sb]}}p), %edx # 4-byte Reload
; X32-NEXT: movl {{[-0-9]+}}(%e{{[sb]}}p), %eax # 4-byte Reload
+; X32-NEXT: movl {{[-0-9]+}}(%e{{[sb]}}p), %edx # 4-byte Reload
; X32-NEXT: movl %eax, %ebx
; X32-NEXT: addl $3, %ebx
; X32-NEXT: movl %edx, %ecx
; X32-NEXT: adcl $0, %ecx
; X32-NEXT: lock cmpxchg8b sc64
-; X32-NEXT: movl %eax, {{[-0-9]+}}(%e{{[sb]}}p) # 4-byte Spill
; X32-NEXT: movl %edx, {{[-0-9]+}}(%e{{[sb]}}p) # 4-byte Spill
+; X32-NEXT: movl %eax, {{[-0-9]+}}(%e{{[sb]}}p) # 4-byte Spill
; X32-NEXT: jne .LBB0_3
; X32-NEXT: jmp .LBB0_4
; X32-NEXT: .LBB0_4: # %atomicrmw.end7
@@ -75,8 +75,8 @@ define void @atomic_fetch_add64() nounwind {
; X32-NEXT: jmp .LBB0_7
; X32-NEXT: .LBB0_7: # %atomicrmw.start
; X32-NEXT: # =>This Inner Loop Header: Depth=1
-; X32-NEXT: movl {{[-0-9]+}}(%e{{[sb]}}p), %edx # 4-byte Reload
; X32-NEXT: movl (%esp), %eax # 4-byte Reload
+; X32-NEXT: movl {{[-0-9]+}}(%e{{[sb]}}p), %edx # 4-byte Reload
; X32-NEXT: movl {{[-0-9]+}}(%e{{[sb]}}p), %esi # 4-byte Reload
; X32-NEXT: movl {{[-0-9]+}}(%e{{[sb]}}p), %ecx # 4-byte Reload
; X32-NEXT: movl %eax, %ebx
@@ -84,8 +84,8 @@ define void @atomic_fetch_add64() nounwind {
; X32-NEXT: movl %edx, %ecx
; X32-NEXT: adcl %esi, %ecx
; X32-NEXT: lock cmpxchg8b sc64
-; X32-NEXT: movl %eax, (%esp) # 4-byte Spill
; X32-NEXT: movl %edx, {{[-0-9]+}}(%e{{[sb]}}p) # 4-byte Spill
+; X32-NEXT: movl %eax, (%esp) # 4-byte Spill
; X32-NEXT: jne .LBB0_7
; X32-NEXT: jmp .LBB0_8
; X32-NEXT: .LBB0_8: # %atomicrmw.end
@@ -114,15 +114,15 @@ define void @atomic_fetch_sub64() nounwind {
; X32-NEXT: jmp .LBB1_1
; X32-NEXT: .LBB1_1: # %atomicrmw.start14
; X32-NEXT: # =>This Inner Loop Header: Depth=1
-; X32-NEXT: movl {{[-0-9]+}}(%e{{[sb]}}p), %edx # 4-byte Reload
; X32-NEXT: movl {{[-0-9]+}}(%e{{[sb]}}p), %eax # 4-byte Reload
+; X32-NEXT: movl {{[-0-9]+}}(%e{{[sb]}}p), %edx # 4-byte Reload
; X32-NEXT: movl %eax, %ebx
; X32-NEXT: addl $-1, %ebx
; X32-NEXT: movl %edx, %ecx
; X32-NEXT: adcl $-1, %ecx
; X32-NEXT: lock cmpxchg8b sc64
-; X32-NEXT: movl %eax, {{[-0-9]+}}(%e{{[sb]}}p) # 4-byte Spill
; X32-NEXT: movl %edx, {{[-0-9]+}}(%e{{[sb]}}p) # 4-byte Spill
+; X32-NEXT: movl %eax, {{[-0-9]+}}(%e{{[sb]}}p) # 4-byte Spill
; X32-NEXT: jne .LBB1_1
; X32-NEXT: jmp .LBB1_2
; X32-NEXT: .LBB1_2: # %atomicrmw.end13
@@ -133,15 +133,15 @@ define void @atomic_fetch_sub64() nounwind {
; X32-NEXT: jmp .LBB1_3
; X32-NEXT: .LBB1_3: # %atomicrmw.start8
; X32-NEXT: # =>This Inner Loop Header: Depth=1
-; X32-NEXT: movl {{[-0-9]+}}(%e{{[sb]}}p), %edx # 4-byte Reload
; X32-NEXT: movl {{[-0-9]+}}(%e{{[sb]}}p), %eax # 4-byte Reload
+; X32-NEXT: movl {{[-0-9]+}}(%e{{[sb]}}p), %edx # 4-byte Reload
; X32-NEXT: movl %eax, %ebx
; X32-NEXT: addl $-3, %ebx
; X32-NEXT: movl %edx, %ecx
; X32-NEXT: adcl $-1, %ecx
; X32-NEXT: lock cmpxchg8b sc64
-; X32-NEXT: movl %eax, {{[-0-9]+}}(%e{{[sb]}}p) # 4-byte Spill
; X32-NEXT: movl %edx, {{[-0-9]+}}(%e{{[sb]}}p) # 4-byte Spill
+; X32-NEXT: movl %eax, {{[-0-9]+}}(%e{{[sb]}}p) # 4-byte Spill
; X32-NEXT: jne .LBB1_3
; X32-NEXT: jmp .LBB1_4
; X32-NEXT: .LBB1_4: # %atomicrmw.end7
@@ -173,8 +173,8 @@ define void @atomic_fetch_sub64() nounwind {
; X32-NEXT: jmp .LBB1_7
; X32-NEXT: .LBB1_7: # %atomicrmw.start
; X32-NEXT: # =>This Inner Loop Header: Depth=1
-; X32-NEXT: movl {{[-0-9]+}}(%e{{[sb]}}p), %edx # 4-byte Reload
; X32-NEXT: movl (%esp), %eax # 4-byte Reload
+; X32-NEXT: movl {{[-0-9]+}}(%e{{[sb]}}p), %edx # 4-byte Reload
; X32-NEXT: movl {{[-0-9]+}}(%e{{[sb]}}p), %esi # 4-byte Reload
; X32-NEXT: movl {{[-0-9]+}}(%e{{[sb]}}p), %ecx # 4-byte Reload
; X32-NEXT: movl %eax, %ebx
@@ -182,8 +182,8 @@ define void @atomic_fetch_sub64() nounwind {
; X32-NEXT: movl %edx, %ecx
; X32-NEXT: sbbl %esi, %ecx
; X32-NEXT: lock cmpxchg8b sc64
-; X32-NEXT: movl %eax, (%esp) # 4-byte Spill
; X32-NEXT: movl %edx, {{[-0-9]+}}(%e{{[sb]}}p) # 4-byte Spill
+; X32-NEXT: movl %eax, (%esp) # 4-byte Spill
; X32-NEXT: jne .LBB1_7
; X32-NEXT: jmp .LBB1_8
; X32-NEXT: .LBB1_8: # %atomicrmw.end
@@ -211,14 +211,14 @@ define void @atomic_fetch_and64() nounwind {
; X32-NEXT: jmp .LBB2_1
; X32-NEXT: .LBB2_1: # %atomicrmw.start8
; X32-NEXT: # =>This Inner Loop Header: Depth=1
-; X32-NEXT: movl {{[-0-9]+}}(%e{{[sb]}}p), %edx # 4-byte Reload
; X32-NEXT: movl {{[-0-9]+}}(%e{{[sb]}}p), %eax # 4-byte Reload
+; X32-NEXT: movl {{[-0-9]+}}(%e{{[sb]}}p), %edx # 4-byte Reload
; X32-NEXT: movl %eax, %ebx
; X32-NEXT: andl $3, %ebx
; X32-NEXT: xorl %ecx, %ecx
; X32-NEXT: lock cmpxchg8b sc64
-; X32-NEXT: movl %eax, {{[-0-9]+}}(%e{{[sb]}}p) # 4-byte Spill
; X32-NEXT: movl %edx, {{[-0-9]+}}(%e{{[sb]}}p) # 4-byte Spill
+; X32-NEXT: movl %eax, {{[-0-9]+}}(%e{{[sb]}}p) # 4-byte Spill
; X32-NEXT: jne .LBB2_1
; X32-NEXT: jmp .LBB2_2
; X32-NEXT: .LBB2_2: # %atomicrmw.end7
@@ -250,8 +250,8 @@ define void @atomic_fetch_and64() nounwind {
; X32-NEXT: jmp .LBB2_5
; X32-NEXT: .LBB2_5: # %atomicrmw.start
; X32-NEXT: # =>This Inner Loop Header: Depth=1
-; X32-NEXT: movl {{[-0-9]+}}(%e{{[sb]}}p), %edx # 4-byte Reload
; X32-NEXT: movl (%esp), %eax # 4-byte Reload
+; X32-NEXT: movl {{[-0-9]+}}(%e{{[sb]}}p), %edx # 4-byte Reload
; X32-NEXT: movl {{[-0-9]+}}(%e{{[sb]}}p), %esi # 4-byte Reload
; X32-NEXT: movl {{[-0-9]+}}(%e{{[sb]}}p), %ecx # 4-byte Reload
; X32-NEXT: movl %eax, %ebx
@@ -259,8 +259,8 @@ define void @atomic_fetch_and64() nounwind {
; X32-NEXT: movl %edx, %ecx
; X32-NEXT: andl %esi, %ecx
; X32-NEXT: lock cmpxchg8b sc64
-; X32-NEXT: movl %eax, (%esp) # 4-byte Spill
; X32-NEXT: movl %edx, {{[-0-9]+}}(%e{{[sb]}}p) # 4-byte Spill
+; X32-NEXT: movl %eax, (%esp) # 4-byte Spill
; X32-NEXT: jne .LBB2_5
; X32-NEXT: jmp .LBB2_6
; X32-NEXT: .LBB2_6: # %atomicrmw.end
@@ -287,14 +287,14 @@ define void @atomic_fetch_or64() nounwind {
; X32-NEXT: jmp .LBB3_1
; X32-NEXT: .LBB3_1: # %atomicrmw.start8
; X32-NEXT: # =>This Inner Loop Header: Depth=1
-; X32-NEXT: movl {{[-0-9]+}}(%e{{[sb]}}p), %ecx # 4-byte Reload
; X32-NEXT: movl {{[-0-9]+}}(%e{{[sb]}}p), %eax # 4-byte Reload
+; X32-NEXT: movl {{[-0-9]+}}(%e{{[sb]}}p), %ecx # 4-byte Reload
; X32-NEXT: movl %eax, %ebx
; X32-NEXT: orl $3, %ebx
; X32-NEXT: movl %ecx, %edx
; X32-NEXT: lock cmpxchg8b sc64
-; X32-NEXT: movl %eax, {{[-0-9]+}}(%e{{[sb]}}p) # 4-byte Spill
; X32-NEXT: movl %edx, {{[-0-9]+}}(%e{{[sb]}}p) # 4-byte Spill
+; X32-NEXT: movl %eax, {{[-0-9]+}}(%e{{[sb]}}p) # 4-byte Spill
; X32-NEXT: jne .LBB3_1
; X32-NEXT: jmp .LBB3_2
; X32-NEXT: .LBB3_2: # %atomicrmw.end7
@@ -326,8 +326,8 @@ define void @atomic_fetch_or64() nounwind {
; X32-NEXT: jmp .LBB3_5
; X32-NEXT: .LBB3_5: # %atomicrmw.start
; X32-NEXT: # =>This Inner Loop Header: Depth=1
-; X32-NEXT: movl {{[-0-9]+}}(%e{{[sb]}}p), %edx # 4-byte Reload
; X32-NEXT: movl (%esp), %eax # 4-byte Reload
+; X32-NEXT: movl {{[-0-9]+}}(%e{{[sb]}}p), %edx # 4-byte Reload
; X32-NEXT: movl {{[-0-9]+}}(%e{{[sb]}}p), %esi # 4-byte Reload
; X32-NEXT: movl {{[-0-9]+}}(%e{{[sb]}}p), %ecx # 4-byte Reload
; X32-NEXT: movl %eax, %ebx
@@ -335,8 +335,8 @@ define void @atomic_fetch_or64() nounwind {
; X32-NEXT: movl %edx, %ecx
; X32-NEXT: orl %esi, %ecx
; X32-NEXT: lock cmpxchg8b sc64
-; X32-NEXT: movl %eax, (%esp) # 4-byte Spill
; X32-NEXT: movl %edx, {{[-0-9]+}}(%e{{[sb]}}p) # 4-byte Spill
+; X32-NEXT: movl %eax, (%esp) # 4-byte Spill
; X32-NEXT: jne .LBB3_5
; X32-NEXT: jmp .LBB3_6
; X32-NEXT: .LBB3_6: # %atomicrmw.end
@@ -363,14 +363,14 @@ define void @atomic_fetch_xor64() nounwind {
; X32-NEXT: jmp .LBB4_1
; X32-NEXT: .LBB4_1: # %atomicrmw.start8
; X32-NEXT: # =>This Inner Loop Header: Depth=1
-; X32-NEXT: movl {{[-0-9]+}}(%e{{[sb]}}p), %ecx # 4-byte Reload
; X32-NEXT: movl {{[-0-9]+}}(%e{{[sb]}}p), %eax # 4-byte Reload
+; X32-NEXT: movl {{[-0-9]+}}(%e{{[sb]}}p), %ecx # 4-byte Reload
; X32-NEXT: movl %eax, %ebx
; X32-NEXT: xorl $3, %ebx
; X32-NEXT: movl %ecx, %edx
; X32-NEXT: lock cmpxchg8b sc64
-; X32-NEXT: movl %eax, {{[-0-9]+}}(%e{{[sb]}}p) # 4-byte Spill
; X32-NEXT: movl %edx, {{[-0-9]+}}(%e{{[sb]}}p) # 4-byte Spill
+; X32-NEXT: movl %eax, {{[-0-9]+}}(%e{{[sb]}}p) # 4-byte Spill
; X32-NEXT: jne .LBB4_1
; X32-NEXT: jmp .LBB4_2
; X32-NEXT: .LBB4_2: # %atomicrmw.end7
@@ -402,8 +402,8 @@ define void @atomic_fetch_xor64() nounwind {
; X32-NEXT: jmp .LBB4_5
; X32-NEXT: .LBB4_5: # %atomicrmw.start
; X32-NEXT: # =>This Inner Loop Header: Depth=1
-; X32-NEXT: movl {{[-0-9]+}}(%e{{[sb]}}p), %edx # 4-byte Reload
; X32-NEXT: movl (%esp), %eax # 4-byte Reload
+; X32-NEXT: movl {{[-0-9]+}}(%e{{[sb]}}p), %edx # 4-byte Reload
; X32-NEXT: movl {{[-0-9]+}}(%e{{[sb]}}p), %esi # 4-byte Reload
; X32-NEXT: movl {{[-0-9]+}}(%e{{[sb]}}p), %ecx # 4-byte Reload
; X32-NEXT: movl %eax, %ebx
@@ -411,8 +411,8 @@ define void @atomic_fetch_xor64() nounwind {
; X32-NEXT: movl %edx, %ecx
; X32-NEXT: xorl %esi, %ecx
; X32-NEXT: lock cmpxchg8b sc64
-; X32-NEXT: movl %eax, (%esp) # 4-byte Spill
; X32-NEXT: movl %edx, {{[-0-9]+}}(%e{{[sb]}}p) # 4-byte Spill
+; X32-NEXT: movl %eax, (%esp) # 4-byte Spill
; X32-NEXT: jne .LBB4_5
; X32-NEXT: jmp .LBB4_6
; X32-NEXT: .LBB4_6: # %atomicrmw.end
@@ -444,8 +444,8 @@ define void @atomic_fetch_nand64(i64 %x) nounwind {
; X32-NEXT: jmp .LBB5_1
; X32-NEXT: .LBB5_1: # %atomicrmw.start
; X32-NEXT: # =>This Inner Loop Header: Depth=1
-; X32-NEXT: movl {{[-0-9]+}}(%e{{[sb]}}p), %edx # 4-byte Reload
; X32-NEXT: movl {{[-0-9]+}}(%e{{[sb]}}p), %eax # 4-byte Reload
+; X32-NEXT: movl {{[-0-9]+}}(%e{{[sb]}}p), %edx # 4-byte Reload
; X32-NEXT: movl {{[-0-9]+}}(%e{{[sb]}}p), %esi # 4-byte Reload
; X32-NEXT: movl (%esp), %edi # 4-byte Reload
; X32-NEXT: movl %edx, %ecx
@@ -455,8 +455,8 @@ define void @atomic_fetch_nand64(i64 %x) nounwind {
; X32-NEXT: notl %ebx
; X32-NEXT: notl %ecx
; X32-NEXT: lock cmpxchg8b sc64
-; X32-NEXT: movl %eax, {{[-0-9]+}}(%e{{[sb]}}p) # 4-byte Spill
; X32-NEXT: movl %edx, {{[-0-9]+}}(%e{{[sb]}}p) # 4-byte Spill
+; X32-NEXT: movl %eax, {{[-0-9]+}}(%e{{[sb]}}p) # 4-byte Spill
; X32-NEXT: jne .LBB5_1
; X32-NEXT: jmp .LBB5_2
; X32-NEXT: .LBB5_2: # %atomicrmw.end
@@ -486,8 +486,8 @@ define void @atomic_fetch_max64(i64 %x) nounwind {
; X32-NEXT: jmp .LBB6_1
; X32-NEXT: .LBB6_1: # %atomicrmw.start
; X32-NEXT: # =>This Inner Loop Header: Depth=1
-; X32-NEXT: movl {{[-0-9]+}}(%e{{[sb]}}p), %edx # 4-byte Reload
; X32-NEXT: movl {{[-0-9]+}}(%e{{[sb]}}p), %eax # 4-byte Reload
+; X32-NEXT: movl {{[-0-9]+}}(%e{{[sb]}}p), %edx # 4-byte Reload
; X32-NEXT: movl {{[-0-9]+}}(%e{{[sb]}}p), %ebx # 4-byte Reload
; X32-NEXT: movl (%esp), %ecx # 4-byte Reload
; X32-NEXT: movl %ebx, %esi
@@ -497,8 +497,8 @@ define void @atomic_fetch_max64(i64 %x) nounwind {
; X32-NEXT: cmovll %edx, %ecx
; X32-NEXT: cmovll %eax, %ebx
; X32-NEXT: lock cmpxchg8b sc64
-; X32-NEXT: movl %eax, {{[-0-9]+}}(%e{{[sb]}}p) # 4-byte Spill
; X32-NEXT: movl %edx, {{[-0-9]+}}(%e{{[sb]}}p) # 4-byte Spill
+; X32-NEXT: movl %eax, {{[-0-9]+}}(%e{{[sb]}}p) # 4-byte Spill
; X32-NEXT: jne .LBB6_1
; X32-NEXT: jmp .LBB6_2
; X32-NEXT: .LBB6_2: # %atomicrmw.end
@@ -527,8 +527,8 @@ define void @atomic_fetch_min64(i64 %x) nounwind {
; X32-NEXT: jmp .LBB7_1
; X32-NEXT: .LBB7_1: # %atomicrmw.start
; X32-NEXT: # =>This Inner Loop Header: Depth=1
-; X32-NEXT: movl {{[-0-9]+}}(%e{{[sb]}}p), %edx # 4-byte Reload
; X32-NEXT: movl {{[-0-9]+}}(%e{{[sb]}}p), %eax # 4-byte Reload
+; X32-NEXT: movl {{[-0-9]+}}(%e{{[sb]}}p), %edx # 4-byte Reload
; X32-NEXT: movl {{[-0-9]+}}(%e{{[sb]}}p), %ebx # 4-byte Reload
; X32-NEXT: movl (%esp), %ecx # 4-byte Reload
; X32-NEXT: movl %ebx, %esi
@@ -538,8 +538,8 @@ define void @atomic_fetch_min64(i64 %x) nounwind {
; X32-NEXT: cmovgel %edx, %ecx
; X32-NEXT: cmovgel %eax, %ebx
; X32-NEXT: lock cmpxchg8b sc64
-; X32-NEXT: movl %eax, {{[-0-9]+}}(%e{{[sb]}}p) # 4-byte Spill
; X32-NEXT: movl %edx, {{[-0-9]+}}(%e{{[sb]}}p) # 4-byte Spill
+; X32-NEXT: movl %eax, {{[-0-9]+}}(%e{{[sb]}}p) # 4-byte Spill
; X32-NEXT: jne .LBB7_1
; X32-NEXT: jmp .LBB7_2
; X32-NEXT: .LBB7_2: # %atomicrmw.end
@@ -568,8 +568,8 @@ define void @atomic_fetch_umax64(i64 %x) nounwind {
; X32-NEXT: jmp .LBB8_1
; X32-NEXT: .LBB8_1: # %atomicrmw.start
; X32-NEXT: # =>This Inner Loop Header: Depth=1
-; X32-NEXT: movl {{[-0-9]+}}(%e{{[sb]}}p), %edx # 4-byte Reload
; X32-NEXT: movl {{[-0-9]+}}(%e{{[sb]}}p), %eax # 4-byte Reload
+; X32-NEXT: movl {{[-0-9]+}}(%e{{[sb]}}p), %edx # 4-byte Reload
; X32-NEXT: movl {{[-0-9]+}}(%e{{[sb]}}p), %ebx # 4-byte Reload
; X32-NEXT: movl (%esp), %ecx # 4-byte Reload
; X32-NEXT: movl %ebx, %esi
@@ -579,8 +579,8 @@ define void @atomic_fetch_umax64(i64 %x) nounwind {
; X32-NEXT: cmovbl %edx, %ecx
; X32-NEXT: cmovbl %eax, %ebx
; X32-NEXT: lock cmpxchg8b sc64
-; X32-NEXT: movl %eax, {{[-0-9]+}}(%e{{[sb]}}p) # 4-byte Spill
; X32-NEXT: movl %edx, {{[-0-9]+}}(%e{{[sb]}}p) # 4-byte Spill
+; X32-NEXT: movl %eax, {{[-0-9]+}}(%e{{[sb]}}p) # 4-byte Spill
; X32-NEXT: jne .LBB8_1
; X32-NEXT: jmp .LBB8_2
; X32-NEXT: .LBB8_2: # %atomicrmw.end
@@ -609,8 +609,8 @@ define void @atomic_fetch_umin64(i64 %x) nounwind {
; X32-NEXT: jmp .LBB9_1
; X32-NEXT: .LBB9_1: # %atomicrmw.start
; X32-NEXT: # =>This Inner Loop Header: Depth=1
-; X32-NEXT: movl {{[-0-9]+}}(%e{{[sb]}}p), %edx # 4-byte Reload
; X32-NEXT: movl {{[-0-9]+}}(%e{{[sb]}}p), %eax # 4-byte Reload
+; X32-NEXT: movl {{[-0-9]+}}(%e{{[sb]}}p), %edx # 4-byte Reload
; X32-NEXT: movl {{[-0-9]+}}(%e{{[sb]}}p), %ebx # 4-byte Reload
; X32-NEXT: movl (%esp), %ecx # 4-byte Reload
; X32-NEXT: movl %ebx, %esi
@@ -620,8 +620,8 @@ define void @atomic_fetch_umin64(i64 %x) nounwind {
; X32-NEXT: cmovael %edx, %ecx
; X32-NEXT: cmovael %eax, %ebx
; X32-NEXT: lock cmpxchg8b sc64
-; X32-NEXT: movl %eax, {{[-0-9]+}}(%e{{[sb]}}p) # 4-byte Spill
; X32-NEXT: movl %edx, {{[-0-9]+}}(%e{{[sb]}}p) # 4-byte Spill
+; X32-NEXT: movl %eax, {{[-0-9]+}}(%e{{[sb]}}p) # 4-byte Spill
; X32-NEXT: jne .LBB9_1
; X32-NEXT: jmp .LBB9_2
; X32-NEXT: .LBB9_2: # %atomicrmw.end
@@ -677,13 +677,13 @@ define void @atomic_fetch_swap64(i64 %x) nounwind {
; X32-NEXT: jmp .LBB12_1
; X32-NEXT: .LBB12_1: # %atomicrmw.start
; X32-NEXT: # =>This Inner Loop Header: Depth=1
-; X32-NEXT: movl {{[-0-9]+}}(%e{{[sb]}}p), %edx # 4-byte Reload
; X32-NEXT: movl {{[-0-9]+}}(%e{{[sb]}}p), %eax # 4-byte Reload
+; X32-NEXT: movl {{[-0-9]+}}(%e{{[sb]}}p), %edx # 4-byte Reload
; X32-NEXT: movl {{[-0-9]+}}(%e{{[sb]}}p), %ebx # 4-byte Reload
; X32-NEXT: movl (%esp), %ecx # 4-byte Reload
; X32-NEXT: lock cmpxchg8b sc64
-; X32-NEXT: movl %eax, {{[-0-9]+}}(%e{{[sb]}}p) # 4-byte Spill
; X32-NEXT: movl %edx, {{[-0-9]+}}(%e{{[sb]}}p) # 4-byte Spill
+; X32-NEXT: movl %eax, {{[-0-9]+}}(%e{{[sb]}}p) # 4-byte Spill
; X32-NEXT: jne .LBB12_1
; X32-NEXT: jmp .LBB12_2
; X32-NEXT: .LBB12_2: # %atomicrmw.end
diff --git a/llvm/test/CodeGen/X86/callbr-asm-branch-folding.ll b/llvm/test/CodeGen/X86/callbr-asm-branch-folding.ll
index 3d389523dffb3..95faca819e975 100644
--- a/llvm/test/CodeGen/X86/callbr-asm-branch-folding.ll
+++ b/llvm/test/CodeGen/X86/callbr-asm-branch-folding.ll
@@ -17,7 +17,7 @@ define dso_local void @n(ptr %o, i32 %p, i32 %u) nounwind {
; CHECK-NEXT: pushq %rbx
; CHECK-NEXT: pushq %rax
; CHECK-NEXT: movl %edx, %ebp
-; CHECK-NEXT: movl %esi, %r12d
+; CHECK-NEXT: movl %esi, %r13d
; CHECK-NEXT: movq %rdi, %rbx
; CHECK-NEXT: callq c
; CHECK-NEXT: movl %eax, %r14d
@@ -28,17 +28,17 @@ define dso_local void @n(ptr %o, i32 %p, i32 %u) nounwind {
; CHECK-NEXT: # %bb.1: # %if.end
; CHECK-NEXT: cmpl $0, e(%rip)
; CHECK-NEXT: # implicit-def: $r15d
-; CHECK-NEXT: # implicit-def: $r13d
+; CHECK-NEXT: # implicit-def: $r12d
; CHECK-NEXT: je .LBB0_4
; CHECK-NEXT: # %bb.2: # %if.then4
-; CHECK-NEXT: movslq %r12d, %rdi
+; CHECK-NEXT: movslq %r13d, %rdi
; CHECK-NEXT: callq m
; CHECK-NEXT: # implicit-def: $r15d
; CHECK-NEXT: # implicit-def: $r12d
; CHECK-NEXT: .LBB0_3: # %r
; CHECK-NEXT: callq c
-; CHECK-NEXT: movl %r12d, %r13d
; CHECK-NEXT: .LBB0_4: # %if.end8
+; CHECK-NEXT: movl %r12d, %r13d
; CHECK-NEXT: movl %r15d, %edi
; CHECK-NEXT: callq i
; CHECK-NEXT: movl %eax, %r12d
diff --git a/llvm/test/CodeGen/X86/callbr-asm-kill.mir b/llvm/test/CodeGen/X86/callbr-asm-kill.mir
index 5aabeade52da1..58a9e4de77faf 100644
--- a/llvm/test/CodeGen/X86/callbr-asm-kill.mir
+++ b/llvm/test/CodeGen/X86/callbr-asm-kill.mir
@@ -6,14 +6,12 @@
# subsequent use of [[MOV64rm]] in the INLINEASM_BR instruction which should be
# killed instead.
--- |
- ; ModuleID = '<stdin>'
source_filename = "<stdin>"
target datalayout = "e-m:e-p270:32:32-p271:32:32-p272:64:64-i64:64-f80:128-n8:16:32:64-S128"
target triple = "x86_64-unknown-linux-gnu"
declare void @foo(ptr)
- ; Function Attrs: nounwind
define void @test1(ptr %arg, ptr %mem) #0 {
entry:
br label %loop
@@ -57,18 +55,17 @@ body: |
; CHECK-NEXT: {{ $}}
; CHECK-NEXT: [[COPY:%[0-9]+]]:gr64 = COPY killed $rsi
; CHECK-NEXT: [[COPY1:%[0-9]+]]:gr64 = COPY killed $rdi
- ; CHECK-NEXT: [[COPY2:%[0-9]+]]:gr64 = COPY killed [[COPY1]]
; CHECK-NEXT: {{ $}}
; CHECK-NEXT: bb.1.loop (ir-block-address-taken %ir-block.loop, inlineasm-br-indirect-target):
; CHECK-NEXT: successors: %bb.2(0x80000000), %bb.1(0x00000000)
; CHECK-NEXT: {{ $}}
- ; CHECK-NEXT: [[COPY3:%[0-9]+]]:gr64 = COPY killed [[COPY2]]
+ ; CHECK-NEXT: [[COPY2:%[0-9]+]]:gr64 = COPY killed [[COPY1]]
; CHECK-NEXT: [[MOV64rm:%[0-9]+]]:gr64 = MOV64rm [[COPY]], 1, $noreg, 0, $noreg :: (load (s64) from %ir.mem)
; CHECK-NEXT: ADJCALLSTACKDOWN64 0, 0, 0, implicit-def dead $rsp, implicit-def dead $eflags, implicit-def dead $ssp, implicit $rsp, implicit $ssp
- ; CHECK-NEXT: $rdi = COPY killed [[COPY3]]
+ ; CHECK-NEXT: $rdi = COPY killed [[COPY2]]
; CHECK-NEXT: CALL64pcrel32 target-flags(x86-plt) @foo, csr_64, implicit $rsp, implicit $ssp, implicit killed $rdi, implicit-def $rsp, implicit-def $ssp
; CHECK-NEXT: ADJCALLSTACKUP64 0, 0, implicit-def dead $rsp, implicit-def dead $eflags, implicit-def dead $ssp, implicit $rsp, implicit $ssp
- ; CHECK-NEXT: [[COPY4:%[0-9]+]]:gr64 = COPY [[MOV64rm]]
+ ; CHECK-NEXT: [[COPY1:%[0-9]+]]:gr64 = COPY [[MOV64rm]]
; CHECK-NEXT: INLINEASM_BR &"", 9 /* sideeffect mayload attdialect */, 262190 /* mem:m */, killed [[MOV64rm]], 1, $noreg, 0, $noreg, 13 /* imm */, blockaddress(@test1, %ir-block.loop)
; CHECK-NEXT: JMP_1 %bb.2
; CHECK-NEXT: {{ $}}
diff --git a/llvm/test/CodeGen/X86/coalescer-breaks-subreg-to-reg-liveness-reduced.ll b/llvm/test/CodeGen/X86/coalescer-breaks-subreg-to-reg-liveness-reduced.ll
index e9f529eea7d3f..fe8d131a977ea 100644
--- a/llvm/test/CodeGen/X86/coalescer-breaks-subreg-to-reg-liveness-reduced.ll
+++ b/llvm/test/CodeGen/X86/coalescer-breaks-subreg-to-reg-liveness-reduced.ll
@@ -49,8 +49,8 @@ define void @foo(ptr %arg3, i1 %icmp16) #0 {
; CHECK-NEXT: .p2align 4
; CHECK-NEXT: .LBB0_1: # %bb5
; CHECK-NEXT: # in Loop: Header=BB0_2 Depth=1
-; CHECK-NEXT: orl $1, %r12d
; CHECK-NEXT: movq %r14, %r15
+; CHECK-NEXT: orl $1, %r12d
; CHECK-NEXT: .LBB0_2: # %bb7
; CHECK-NEXT: # =>This Inner Loop Header: Depth=1
; CHECK-NEXT: xorl %eax, %eax
diff --git a/llvm/test/CodeGen/X86/combine-pmuldq.ll b/llvm/test/CodeGen/X86/combine-pmuldq.ll
index 70335f834291d..845c6cd6f2454 100644
--- a/llvm/test/CodeGen/X86/combine-pmuldq.ll
+++ b/llvm/test/CodeGen/X86/combine-pmuldq.ll
@@ -328,9 +328,9 @@ define <8 x i32> @PR49658_zext(ptr %ptr, i32 %mul) {
; SSE: # %bb.0: # %start
; SSE-NEXT: movd %esi, %xmm0
; SSE-NEXT: pshufd {{.*#+}} xmm2 = xmm0[0,1,0,1]
-; SSE-NEXT: pxor %xmm0, %xmm0
-; SSE-NEXT: movq $-2097152, %rax # imm = 0xFFE00000
; SSE-NEXT: pxor %xmm1, %xmm1
+; SSE-NEXT: movq $-2097152, %rax # imm = 0xFFE00000
+; SSE-NEXT: pxor %xmm0, %xmm0
; SSE-NEXT: .p2align 4
; SSE-NEXT: .LBB7_1: # %loop
; SSE-NEXT: # =>This Inner Loop Header: Depth=1
@@ -469,11 +469,11 @@ define <8 x i32> @PR49658_sext(ptr %ptr, i32 %mul) {
; SSE-NEXT: movslq %esi, %rax
; SSE-NEXT: movq %rax, %xmm0
; SSE-NEXT: pshufd {{.*#+}} xmm2 = xmm0[0,1,0,1]
-; SSE-NEXT: pxor %xmm0, %xmm0
+; SSE-NEXT: pxor %xmm1, %xmm1
; SSE-NEXT: movq $-2097152, %rax # imm = 0xFFE00000
; SSE-NEXT: movdqa %xmm2, %xmm3
; SSE-NEXT: psrlq $32, %xmm3
-; SSE-NEXT: pxor %xmm1, %xmm1
+; SSE-NEXT: pxor %xmm0, %xmm0
; SSE-NEXT: .p2align 4
; SSE-NEXT: .LBB8_1: # %loop
; SSE-NEXT: # =>This Inner Loop Header: Depth=1
diff --git a/llvm/test/CodeGen/X86/fp128-select.ll b/llvm/test/CodeGen/X86/fp128-select.ll
index 0486c1c4d28e9..ca1e676e91f35 100644
--- a/llvm/test/CodeGen/X86/fp128-select.ll
+++ b/llvm/test/CodeGen/X86/fp128-select.ll
@@ -14,7 +14,7 @@ define void @test_select(ptr %p, ptr %q, i1 zeroext %c) {
; SSE-NEXT: testl %edx, %edx
; SSE-NEXT: jne .LBB0_1
; SSE-NEXT: # %bb.3:
-; SSE-NEXT: movaps {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm0
+; SSE-NEXT: movaps {{.*#+}} xmm0 = [NaN]
; SSE-NEXT: movaps %xmm0, (%rsi)
; SSE-NEXT: retq
; SSE-NEXT: .LBB0_1:
@@ -61,7 +61,7 @@ define fp128 @test_select_cc(fp128, fp128) {
; SSE-NEXT: xorps %xmm1, %xmm1
; SSE-NEXT: jmp .LBB1_3
; SSE-NEXT: .LBB1_1:
-; SSE-NEXT: movaps {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm1
+; SSE-NEXT: movaps {{.*#+}} xmm1 = [1.0E+0]
; SSE-NEXT: .LBB1_3: # %BB0
; SSE-NEXT: testl %ebx, %ebx
; SSE-NEXT: movaps (%rsp), %xmm0 # 16-byte Reload
@@ -103,17 +103,18 @@ define fp128 @test_select_cc(fp128, fp128) {
; NOSSE-NEXT: movq %r12, %rdx
; NOSSE-NEXT: movq %r15, %rcx
; NOSSE-NEXT: callq __eqtf2 at PLT
-; NOSSE-NEXT: movl %eax, %ecx
-; NOSSE-NEXT: xorl %eax, %eax
-; NOSSE-NEXT: testl %ecx, %ecx
-; NOSSE-NEXT: movabsq $4611404543450677248, %rdx # imm = 0x3FFF000000000000
-; NOSSE-NEXT: cmovneq %rax, %rdx
+; NOSSE-NEXT: xorl %ecx, %ecx
+; NOSSE-NEXT: testl %eax, %eax
+; NOSSE-NEXT: movabsq $4611404543450677248, %rax # imm = 0x3FFF000000000000
+; NOSSE-NEXT: cmovneq %rcx, %rax
; NOSSE-NEXT: testl %ebp, %ebp
-; NOSSE-NEXT: je .LBB1_2
-; NOSSE-NEXT: # %bb.1:
+; NOSSE-NEXT: jne .LBB1_2
+; NOSSE-NEXT: # %bb.1: # %BB1
+; NOSSE-NEXT: xorl %r14d, %r14d
+; NOSSE-NEXT: movq %rax, %rbx
+; NOSSE-NEXT: .LBB1_2: # %BB2
; NOSSE-NEXT: movq %r14, %rax
; NOSSE-NEXT: movq %rbx, %rdx
-; NOSSE-NEXT: .LBB1_2: # %BB2
; NOSSE-NEXT: popq %rbx
; NOSSE-NEXT: .cfi_def_cfa_offset 40
; NOSSE-NEXT: popq %r12
diff --git a/llvm/test/CodeGen/X86/madd.ll b/llvm/test/CodeGen/X86/madd.ll
index bdb7c307a5759..6ca9e77b9a555 100644
--- a/llvm/test/CodeGen/X86/madd.ll
+++ b/llvm/test/CodeGen/X86/madd.ll
@@ -196,10 +196,10 @@ define i32 @_Z10test_shortPsS_i_512(ptr nocapture readonly, ptr nocapture readon
; SSE2-LABEL: _Z10test_shortPsS_i_512:
; SSE2: # %bb.0: # %entry
; SSE2-NEXT: movl %edx, %eax
+; SSE2-NEXT: pxor %xmm1, %xmm1
; SSE2-NEXT: pxor %xmm0, %xmm0
; SSE2-NEXT: xorl %ecx, %ecx
; SSE2-NEXT: pxor %xmm2, %xmm2
-; SSE2-NEXT: pxor %xmm1, %xmm1
; SSE2-NEXT: .p2align 4
; SSE2-NEXT: .LBB2_1: # %vector.body
; SSE2-NEXT: # =>This Inner Loop Header: Depth=1
@@ -210,19 +210,19 @@ define i32 @_Z10test_shortPsS_i_512(ptr nocapture readonly, ptr nocapture readon
; SSE2-NEXT: paddd %xmm5, %xmm2
; SSE2-NEXT: movdqu 16(%rsi,%rcx,2), %xmm3
; SSE2-NEXT: pmaddwd %xmm4, %xmm3
-; SSE2-NEXT: paddd %xmm3, %xmm1
+; SSE2-NEXT: paddd %xmm3, %xmm0
; SSE2-NEXT: addq $16, %rcx
; SSE2-NEXT: cmpq %rcx, %rax
; SSE2-NEXT: jne .LBB2_1
; SSE2-NEXT: # %bb.2: # %middle.block
-; SSE2-NEXT: paddd %xmm0, %xmm2
-; SSE2-NEXT: paddd %xmm0, %xmm1
-; SSE2-NEXT: paddd %xmm2, %xmm1
-; SSE2-NEXT: pshufd {{.*#+}} xmm0 = xmm1[2,3,2,3]
+; SSE2-NEXT: paddd %xmm1, %xmm2
; SSE2-NEXT: paddd %xmm1, %xmm0
-; SSE2-NEXT: pshufd {{.*#+}} xmm1 = xmm0[1,1,1,1]
+; SSE2-NEXT: paddd %xmm2, %xmm0
+; SSE2-NEXT: pshufd {{.*#+}} xmm1 = xmm0[2,3,2,3]
; SSE2-NEXT: paddd %xmm0, %xmm1
-; SSE2-NEXT: movd %xmm1, %eax
+; SSE2-NEXT: pshufd {{.*#+}} xmm0 = xmm1[1,1,1,1]
+; SSE2-NEXT: paddd %xmm1, %xmm0
+; SSE2-NEXT: movd %xmm0, %eax
; SSE2-NEXT: retq
;
; AVX1-LABEL: _Z10test_shortPsS_i_512:
@@ -397,9 +397,9 @@ define i32 @_Z10test_shortPsS_i_1024(ptr nocapture readonly, ptr nocapture reado
; AVX1: # %bb.0: # %entry
; AVX1-NEXT: movl %edx, %eax
; AVX1-NEXT: vpxor %xmm0, %xmm0, %xmm0
+; AVX1-NEXT: vpxor %xmm1, %xmm1, %xmm1
; AVX1-NEXT: xorl %ecx, %ecx
; AVX1-NEXT: vpxor %xmm2, %xmm2, %xmm2
-; AVX1-NEXT: vpxor %xmm1, %xmm1, %xmm1
; AVX1-NEXT: .p2align 4
; AVX1-NEXT: .LBB3_1: # %vector.body
; AVX1-NEXT: # =>This Inner Loop Header: Depth=1
@@ -445,8 +445,8 @@ define i32 @_Z10test_shortPsS_i_1024(ptr nocapture readonly, ptr nocapture reado
; AVX2: # %bb.0: # %entry
; AVX2-NEXT: movl %edx, %eax
; AVX2-NEXT: vpxor %xmm0, %xmm0, %xmm0
-; AVX2-NEXT: xorl %ecx, %ecx
; AVX2-NEXT: vpxor %xmm1, %xmm1, %xmm1
+; AVX2-NEXT: xorl %ecx, %ecx
; AVX2-NEXT: vpxor %xmm2, %xmm2, %xmm2
; AVX2-NEXT: .p2align 4
; AVX2-NEXT: .LBB3_1: # %vector.body
@@ -454,16 +454,16 @@ define i32 @_Z10test_shortPsS_i_1024(ptr nocapture readonly, ptr nocapture reado
; AVX2-NEXT: vmovdqu (%rsi,%rcx,2), %ymm3
; AVX2-NEXT: vmovdqu 32(%rsi,%rcx,2), %ymm4
; AVX2-NEXT: vpmaddwd (%rdi,%rcx,2), %ymm3, %ymm3
-; AVX2-NEXT: vpaddd %ymm1, %ymm3, %ymm1
-; AVX2-NEXT: vpmaddwd 32(%rdi,%rcx,2), %ymm4, %ymm3
; AVX2-NEXT: vpaddd %ymm2, %ymm3, %ymm2
+; AVX2-NEXT: vpmaddwd 32(%rdi,%rcx,2), %ymm4, %ymm3
+; AVX2-NEXT: vpaddd %ymm1, %ymm3, %ymm1
; AVX2-NEXT: addq $16, %rcx
; AVX2-NEXT: cmpq %rcx, %rax
; AVX2-NEXT: jne .LBB3_1
; AVX2-NEXT: # %bb.2: # %middle.block
-; AVX2-NEXT: vpaddd %ymm0, %ymm1, %ymm1
-; AVX2-NEXT: vpaddd %ymm0, %ymm2, %ymm0
+; AVX2-NEXT: vpaddd %ymm0, %ymm2, %ymm2
; AVX2-NEXT: vpaddd %ymm0, %ymm1, %ymm0
+; AVX2-NEXT: vpaddd %ymm0, %ymm2, %ymm0
; AVX2-NEXT: vextracti128 $1, %ymm0, %xmm1
; AVX2-NEXT: vpaddd %xmm1, %xmm0, %xmm0
; AVX2-NEXT: vpshufd {{.*#+}} xmm1 = xmm0[2,3,2,3]
@@ -770,10 +770,10 @@ define i32 @_Z9test_charPcS_i_512(ptr nocapture readonly, ptr nocapture readonly
; SSE2-LABEL: _Z9test_charPcS_i_512:
; SSE2: # %bb.0: # %entry
; SSE2-NEXT: movl %edx, %eax
+; SSE2-NEXT: pxor %xmm1, %xmm1
; SSE2-NEXT: pxor %xmm0, %xmm0
; SSE2-NEXT: xorl %ecx, %ecx
; SSE2-NEXT: pxor %xmm2, %xmm2
-; SSE2-NEXT: pxor %xmm1, %xmm1
; SSE2-NEXT: .p2align 4
; SSE2-NEXT: .LBB6_1: # %vector.body
; SSE2-NEXT: # =>This Inner Loop Header: Depth=1
@@ -790,19 +790,19 @@ define i32 @_Z9test_charPcS_i_512(ptr nocapture readonly, ptr nocapture readonly
; SSE2-NEXT: punpckhbw {{.*#+}} xmm4 = xmm4[8,8,9,9,10,10,11,11,12,12,13,13,14,14,15,15]
; SSE2-NEXT: psraw $8, %xmm4
; SSE2-NEXT: pmaddwd %xmm3, %xmm4
-; SSE2-NEXT: paddd %xmm4, %xmm1
+; SSE2-NEXT: paddd %xmm4, %xmm0
; SSE2-NEXT: addq $16, %rcx
; SSE2-NEXT: cmpq %rcx, %rax
; SSE2-NEXT: jne .LBB6_1
; SSE2-NEXT: # %bb.2: # %middle.block
-; SSE2-NEXT: paddd %xmm0, %xmm2
-; SSE2-NEXT: paddd %xmm0, %xmm1
-; SSE2-NEXT: paddd %xmm2, %xmm1
-; SSE2-NEXT: pshufd {{.*#+}} xmm0 = xmm1[2,3,2,3]
+; SSE2-NEXT: paddd %xmm1, %xmm2
; SSE2-NEXT: paddd %xmm1, %xmm0
-; SSE2-NEXT: pshufd {{.*#+}} xmm1 = xmm0[1,1,1,1]
+; SSE2-NEXT: paddd %xmm2, %xmm0
+; SSE2-NEXT: pshufd {{.*#+}} xmm1 = xmm0[2,3,2,3]
; SSE2-NEXT: paddd %xmm0, %xmm1
-; SSE2-NEXT: movd %xmm1, %eax
+; SSE2-NEXT: pshufd {{.*#+}} xmm0 = xmm1[1,1,1,1]
+; SSE2-NEXT: paddd %xmm1, %xmm0
+; SSE2-NEXT: movd %xmm0, %eax
; SSE2-NEXT: retq
;
; AVX1-LABEL: _Z9test_charPcS_i_512:
@@ -993,9 +993,9 @@ define i32 @_Z9test_charPcS_i_1024(ptr nocapture readonly, ptr nocapture readonl
; AVX1: # %bb.0: # %entry
; AVX1-NEXT: movl %edx, %eax
; AVX1-NEXT: vpxor %xmm0, %xmm0, %xmm0
+; AVX1-NEXT: vpxor %xmm1, %xmm1, %xmm1
; AVX1-NEXT: xorl %ecx, %ecx
; AVX1-NEXT: vpxor %xmm2, %xmm2, %xmm2
-; AVX1-NEXT: vpxor %xmm1, %xmm1, %xmm1
; AVX1-NEXT: .p2align 4
; AVX1-NEXT: .LBB7_1: # %vector.body
; AVX1-NEXT: # =>This Inner Loop Header: Depth=1
@@ -1045,8 +1045,8 @@ define i32 @_Z9test_charPcS_i_1024(ptr nocapture readonly, ptr nocapture readonl
; AVX2: # %bb.0: # %entry
; AVX2-NEXT: movl %edx, %eax
; AVX2-NEXT: vpxor %xmm0, %xmm0, %xmm0
-; AVX2-NEXT: xorl %ecx, %ecx
; AVX2-NEXT: vpxor %xmm1, %xmm1, %xmm1
+; AVX2-NEXT: xorl %ecx, %ecx
; AVX2-NEXT: vpxor %xmm2, %xmm2, %xmm2
; AVX2-NEXT: .p2align 4
; AVX2-NEXT: .LBB7_1: # %vector.body
@@ -1055,17 +1055,17 @@ define i32 @_Z9test_charPcS_i_1024(ptr nocapture readonly, ptr nocapture readonl
; AVX2-NEXT: vpmovsxbw (%rdi,%rcx), %ymm4
; AVX2-NEXT: vpmovsxbw 16(%rsi,%rcx), %ymm5
; AVX2-NEXT: vpmaddwd %ymm3, %ymm5, %ymm3
-; AVX2-NEXT: vpaddd %ymm2, %ymm3, %ymm2
+; AVX2-NEXT: vpaddd %ymm1, %ymm3, %ymm1
; AVX2-NEXT: vpmovsxbw (%rsi,%rcx), %ymm3
; AVX2-NEXT: vpmaddwd %ymm4, %ymm3, %ymm3
-; AVX2-NEXT: vpaddd %ymm1, %ymm3, %ymm1
+; AVX2-NEXT: vpaddd %ymm2, %ymm3, %ymm2
; AVX2-NEXT: addq $32, %rcx
; AVX2-NEXT: cmpq %rcx, %rax
; AVX2-NEXT: jne .LBB7_1
; AVX2-NEXT: # %bb.2: # %middle.block
-; AVX2-NEXT: vpaddd %ymm0, %ymm1, %ymm1
-; AVX2-NEXT: vpaddd %ymm0, %ymm2, %ymm0
+; AVX2-NEXT: vpaddd %ymm0, %ymm2, %ymm2
; AVX2-NEXT: vpaddd %ymm0, %ymm1, %ymm0
+; AVX2-NEXT: vpaddd %ymm0, %ymm2, %ymm0
; AVX2-NEXT: vextracti128 $1, %ymm0, %xmm1
; AVX2-NEXT: vpaddd %xmm1, %xmm0, %xmm0
; AVX2-NEXT: vpshufd {{.*#+}} xmm1 = xmm0[2,3,2,3]
@@ -1271,19 +1271,19 @@ define i32 @test_unsigned_short_256(ptr nocapture readonly, ptr nocapture readon
; SSE2-NEXT: pmullw %xmm2, %xmm3
; SSE2-NEXT: movdqa %xmm3, %xmm2
; SSE2-NEXT: punpcklwd {{.*#+}} xmm2 = xmm2[0],xmm4[0],xmm2[1],xmm4[1],xmm2[2],xmm4[2],xmm2[3],xmm4[3]
-; SSE2-NEXT: paddd %xmm2, %xmm0
+; SSE2-NEXT: paddd %xmm2, %xmm1
; SSE2-NEXT: punpckhwd {{.*#+}} xmm3 = xmm3[4],xmm4[4],xmm3[5],xmm4[5],xmm3[6],xmm4[6],xmm3[7],xmm4[7]
-; SSE2-NEXT: paddd %xmm3, %xmm1
+; SSE2-NEXT: paddd %xmm3, %xmm0
; SSE2-NEXT: addq $16, %rcx
; SSE2-NEXT: cmpq %rcx, %rax
; SSE2-NEXT: jne .LBB9_1
; SSE2-NEXT: # %bb.2: # %middle.block
-; SSE2-NEXT: paddd %xmm1, %xmm0
-; SSE2-NEXT: pshufd {{.*#+}} xmm1 = xmm0[2,3,2,3]
; SSE2-NEXT: paddd %xmm0, %xmm1
-; SSE2-NEXT: pshufd {{.*#+}} xmm0 = xmm1[1,1,1,1]
+; SSE2-NEXT: pshufd {{.*#+}} xmm0 = xmm1[2,3,2,3]
; SSE2-NEXT: paddd %xmm1, %xmm0
-; SSE2-NEXT: movd %xmm0, %eax
+; SSE2-NEXT: pshufd {{.*#+}} xmm1 = xmm0[1,1,1,1]
+; SSE2-NEXT: paddd %xmm0, %xmm1
+; SSE2-NEXT: movd %xmm1, %eax
; SSE2-NEXT: retq
;
; AVX1-LABEL: test_unsigned_short_256:
@@ -1442,22 +1442,22 @@ define i32 @test_unsigned_short_512(ptr nocapture readonly, ptr nocapture readon
; AVX1-NEXT: vpmulld %xmm4, %xmm6, %xmm4
; AVX1-NEXT: vpmovzxwd {{.*#+}} xmm6 = mem[0],zero,mem[1],zero,mem[2],zero,mem[3],zero
; AVX1-NEXT: vpmulld %xmm5, %xmm6, %xmm5
-; AVX1-NEXT: vextractf128 $1, %ymm1, %xmm6
+; AVX1-NEXT: vextractf128 $1, %ymm0, %xmm6
; AVX1-NEXT: vpaddd %xmm6, %xmm2, %xmm2
-; AVX1-NEXT: vpaddd %xmm1, %xmm3, %xmm1
-; AVX1-NEXT: vinsertf128 $1, %xmm2, %ymm1, %ymm1
-; AVX1-NEXT: vextractf128 $1, %ymm0, %xmm2
-; AVX1-NEXT: vpaddd %xmm2, %xmm4, %xmm2
-; AVX1-NEXT: vpaddd %xmm0, %xmm5, %xmm0
+; AVX1-NEXT: vpaddd %xmm0, %xmm3, %xmm0
; AVX1-NEXT: vinsertf128 $1, %xmm2, %ymm0, %ymm0
+; AVX1-NEXT: vextractf128 $1, %ymm1, %xmm2
+; AVX1-NEXT: vpaddd %xmm2, %xmm4, %xmm2
+; AVX1-NEXT: vpaddd %xmm1, %xmm5, %xmm1
+; AVX1-NEXT: vinsertf128 $1, %xmm2, %ymm1, %ymm1
; AVX1-NEXT: addq $16, %rcx
; AVX1-NEXT: cmpq %rcx, %rax
; AVX1-NEXT: jne .LBB10_1
; AVX1-NEXT: # %bb.2: # %middle.block
-; AVX1-NEXT: vextractf128 $1, %ymm0, %xmm2
-; AVX1-NEXT: vextractf128 $1, %ymm1, %xmm3
+; AVX1-NEXT: vextractf128 $1, %ymm1, %xmm2
+; AVX1-NEXT: vextractf128 $1, %ymm0, %xmm3
; AVX1-NEXT: vpaddd %xmm3, %xmm2, %xmm2
-; AVX1-NEXT: vpaddd %xmm1, %xmm0, %xmm0
+; AVX1-NEXT: vpaddd %xmm0, %xmm1, %xmm0
; AVX1-NEXT: vpaddd %xmm2, %xmm0, %xmm0
; AVX1-NEXT: vpshufd {{.*#+}} xmm1 = xmm0[2,3,2,3]
; AVX1-NEXT: vpaddd %xmm1, %xmm0, %xmm0
@@ -1480,15 +1480,15 @@ define i32 @test_unsigned_short_512(ptr nocapture readonly, ptr nocapture readon
; AVX2-NEXT: vpmovzxwd {{.*#+}} ymm3 = mem[0],zero,mem[1],zero,mem[2],zero,mem[3],zero,mem[4],zero,mem[5],zero,mem[6],zero,mem[7],zero
; AVX2-NEXT: vpmovzxwd {{.*#+}} ymm4 = mem[0],zero,mem[1],zero,mem[2],zero,mem[3],zero,mem[4],zero,mem[5],zero,mem[6],zero,mem[7],zero
; AVX2-NEXT: vpmulld %ymm2, %ymm4, %ymm2
-; AVX2-NEXT: vpaddd %ymm1, %ymm2, %ymm1
+; AVX2-NEXT: vpaddd %ymm0, %ymm2, %ymm0
; AVX2-NEXT: vpmovzxwd {{.*#+}} ymm2 = mem[0],zero,mem[1],zero,mem[2],zero,mem[3],zero,mem[4],zero,mem[5],zero,mem[6],zero,mem[7],zero
; AVX2-NEXT: vpmulld %ymm3, %ymm2, %ymm2
-; AVX2-NEXT: vpaddd %ymm0, %ymm2, %ymm0
+; AVX2-NEXT: vpaddd %ymm1, %ymm2, %ymm1
; AVX2-NEXT: addq $16, %rcx
; AVX2-NEXT: cmpq %rcx, %rax
; AVX2-NEXT: jne .LBB10_1
; AVX2-NEXT: # %bb.2: # %middle.block
-; AVX2-NEXT: vpaddd %ymm1, %ymm0, %ymm0
+; AVX2-NEXT: vpaddd %ymm0, %ymm1, %ymm0
; AVX2-NEXT: vextracti128 $1, %ymm0, %xmm1
; AVX2-NEXT: vpaddd %xmm1, %xmm0, %xmm0
; AVX2-NEXT: vpshufd {{.*#+}} xmm1 = xmm0[2,3,2,3]
@@ -1765,15 +1765,15 @@ define i32 @test_unsigned_short_1024(ptr nocapture readonly, ptr nocapture reado
; AVX512-NEXT: vpmovzxwd {{.*#+}} zmm3 = mem[0],zero,mem[1],zero,mem[2],zero,mem[3],zero,mem[4],zero,mem[5],zero,mem[6],zero,mem[7],zero,mem[8],zero,mem[9],zero,mem[10],zero,mem[11],zero,mem[12],zero,mem[13],zero,mem[14],zero,mem[15],zero
; AVX512-NEXT: vpmovzxwd {{.*#+}} zmm4 = mem[0],zero,mem[1],zero,mem[2],zero,mem[3],zero,mem[4],zero,mem[5],zero,mem[6],zero,mem[7],zero,mem[8],zero,mem[9],zero,mem[10],zero,mem[11],zero,mem[12],zero,mem[13],zero,mem[14],zero,mem[15],zero
; AVX512-NEXT: vpmulld %zmm2, %zmm4, %zmm2
-; AVX512-NEXT: vpaddd %zmm1, %zmm2, %zmm1
+; AVX512-NEXT: vpaddd %zmm0, %zmm2, %zmm0
; AVX512-NEXT: vpmovzxwd {{.*#+}} zmm2 = mem[0],zero,mem[1],zero,mem[2],zero,mem[3],zero,mem[4],zero,mem[5],zero,mem[6],zero,mem[7],zero,mem[8],zero,mem[9],zero,mem[10],zero,mem[11],zero,mem[12],zero,mem[13],zero,mem[14],zero,mem[15],zero
; AVX512-NEXT: vpmulld %zmm3, %zmm2, %zmm2
-; AVX512-NEXT: vpaddd %zmm0, %zmm2, %zmm0
+; AVX512-NEXT: vpaddd %zmm1, %zmm2, %zmm1
; AVX512-NEXT: addq $16, %rcx
; AVX512-NEXT: cmpq %rcx, %rax
; AVX512-NEXT: jne .LBB11_1
; AVX512-NEXT: # %bb.2: # %middle.block
-; AVX512-NEXT: vpaddd %zmm1, %zmm0, %zmm0
+; AVX512-NEXT: vpaddd %zmm0, %zmm1, %zmm0
; AVX512-NEXT: vextracti64x4 $1, %zmm0, %ymm1
; AVX512-NEXT: vpaddd %zmm1, %zmm0, %zmm0
; AVX512-NEXT: vextracti128 $1, %ymm0, %xmm1
@@ -2739,9 +2739,9 @@ define i64 @sum_and_sum_of_squares(ptr %a, i32 %n) {
; SSE2: # %bb.0: # %entry
; SSE2-NEXT: movl %esi, %eax
; SSE2-NEXT: pxor %xmm0, %xmm0
+; SSE2-NEXT: pxor %xmm3, %xmm3
; SSE2-NEXT: pxor %xmm1, %xmm1
; SSE2-NEXT: pxor %xmm2, %xmm2
-; SSE2-NEXT: pxor %xmm3, %xmm3
; SSE2-NEXT: .p2align 4
; SSE2-NEXT: .LBB33_1: # %vector.body
; SSE2-NEXT: # =>This Inner Loop Header: Depth=1
@@ -3101,10 +3101,10 @@ define i32 @add_used_by_loop_phi(ptr %a, ptr %b, i64 %offset_a, i64 %offset_b, i
; SSE2: # %bb.0: # %entry
; SSE2-NEXT: addq %rdx, %rdi
; SSE2-NEXT: addq %rcx, %rsi
+; SSE2-NEXT: pxor %xmm1, %xmm1
; SSE2-NEXT: pxor %xmm0, %xmm0
; SSE2-NEXT: xorl %eax, %eax
; SSE2-NEXT: pxor %xmm2, %xmm2
-; SSE2-NEXT: pxor %xmm1, %xmm1
; SSE2-NEXT: .p2align 4
; SSE2-NEXT: .LBB38_1: # %loop
; SSE2-NEXT: # =>This Inner Loop Header: Depth=1
@@ -3121,19 +3121,19 @@ define i32 @add_used_by_loop_phi(ptr %a, ptr %b, i64 %offset_a, i64 %offset_b, i
; SSE2-NEXT: punpckhbw {{.*#+}} xmm3 = xmm3[8,8,9,9,10,10,11,11,12,12,13,13,14,14,15,15]
; SSE2-NEXT: psraw $8, %xmm3
; SSE2-NEXT: pmaddwd %xmm4, %xmm3
-; SSE2-NEXT: paddd %xmm3, %xmm1
+; SSE2-NEXT: paddd %xmm3, %xmm0
; SSE2-NEXT: addq $16, %rax
; SSE2-NEXT: cmpq %r8, %rax
; SSE2-NEXT: jb .LBB38_1
; SSE2-NEXT: # %bb.2: # %afterloop
-; SSE2-NEXT: paddd %xmm0, %xmm2
-; SSE2-NEXT: paddd %xmm0, %xmm1
-; SSE2-NEXT: paddd %xmm2, %xmm1
-; SSE2-NEXT: pshufd {{.*#+}} xmm0 = xmm1[2,3,2,3]
+; SSE2-NEXT: paddd %xmm1, %xmm2
; SSE2-NEXT: paddd %xmm1, %xmm0
-; SSE2-NEXT: pshufd {{.*#+}} xmm1 = xmm0[1,1,1,1]
+; SSE2-NEXT: paddd %xmm2, %xmm0
+; SSE2-NEXT: pshufd {{.*#+}} xmm1 = xmm0[2,3,2,3]
; SSE2-NEXT: paddd %xmm0, %xmm1
-; SSE2-NEXT: movd %xmm1, %eax
+; SSE2-NEXT: pshufd {{.*#+}} xmm0 = xmm1[1,1,1,1]
+; SSE2-NEXT: paddd %xmm1, %xmm0
+; SSE2-NEXT: movd %xmm0, %eax
; SSE2-NEXT: retq
;
; AVX1-LABEL: add_used_by_loop_phi:
diff --git a/llvm/test/CodeGen/X86/masked_load.ll b/llvm/test/CodeGen/X86/masked_load.ll
index 8c4bab99a5b7b..76a94f825266a 100644
--- a/llvm/test/CodeGen/X86/masked_load.ll
+++ b/llvm/test/CodeGen/X86/masked_load.ll
@@ -226,12 +226,11 @@ define <4 x double> @load_v4f64_v4i32(<4 x i32> %trigger, ptr %addr, <4 x double
define <4 x double> @load_v4f64_v4i32_zero(<4 x i32> %trigger, ptr %addr) {
; SSE-LABEL: load_v4f64_v4i32_zero:
; SSE: ## %bb.0:
-; SSE-NEXT: movdqa %xmm0, %xmm1
-; SSE-NEXT: pxor %xmm0, %xmm0
-; SSE-NEXT: pcmpeqd %xmm0, %xmm1
-; SSE-NEXT: movmskps %xmm1, %eax
+; SSE-NEXT: pxor %xmm1, %xmm1
+; SSE-NEXT: pcmpeqd %xmm1, %xmm0
+; SSE-NEXT: movmskps %xmm0, %eax
; SSE-NEXT: testb $1, %al
-; SSE-NEXT: xorps %xmm1, %xmm1
+; SSE-NEXT: xorps %xmm0, %xmm0
; SSE-NEXT: jne LBB3_1
; SSE-NEXT: ## %bb.2: ## %else
; SSE-NEXT: testb $2, %al
@@ -245,7 +244,7 @@ define <4 x double> @load_v4f64_v4i32_zero(<4 x i32> %trigger, ptr %addr) {
; SSE-NEXT: LBB3_8: ## %else8
; SSE-NEXT: retq
; SSE-NEXT: LBB3_1: ## %cond.load
-; SSE-NEXT: movq {{.*#+}} xmm0 = mem[0],zero
+; SSE-NEXT: movsd {{.*#+}} xmm0 = mem[0],zero
; SSE-NEXT: testb $2, %al
; SSE-NEXT: je LBB3_4
; SSE-NEXT: LBB3_3: ## %cond.load1
@@ -1096,9 +1095,9 @@ define <8 x float> @load_v8f32_v8i1_zero(<8 x i1> %mask, ptr %addr) {
; SSE2-NEXT: psllw $15, %xmm0
; SSE2-NEXT: packsswb %xmm0, %xmm0
; SSE2-NEXT: pmovmskb %xmm0, %eax
-; SSE2-NEXT: pxor %xmm0, %xmm0
-; SSE2-NEXT: testb $1, %al
; SSE2-NEXT: xorps %xmm1, %xmm1
+; SSE2-NEXT: testb $1, %al
+; SSE2-NEXT: pxor %xmm0, %xmm0
; SSE2-NEXT: jne LBB10_1
; SSE2-NEXT: ## %bb.2: ## %else
; SSE2-NEXT: testb $2, %al
@@ -1175,9 +1174,9 @@ define <8 x float> @load_v8f32_v8i1_zero(<8 x i1> %mask, ptr %addr) {
; SSE42-NEXT: psllw $15, %xmm0
; SSE42-NEXT: packsswb %xmm0, %xmm0
; SSE42-NEXT: pmovmskb %xmm0, %eax
-; SSE42-NEXT: pxor %xmm0, %xmm0
-; SSE42-NEXT: testb $1, %al
; SSE42-NEXT: xorps %xmm1, %xmm1
+; SSE42-NEXT: testb $1, %al
+; SSE42-NEXT: pxor %xmm0, %xmm0
; SSE42-NEXT: jne LBB10_1
; SSE42-NEXT: ## %bb.2: ## %else
; SSE42-NEXT: testb $2, %al
@@ -2614,9 +2613,9 @@ define <8 x i32> @load_v8i32_v8i1_zero(<8 x i1> %mask, ptr %addr) {
; SSE2-NEXT: psllw $15, %xmm0
; SSE2-NEXT: packsswb %xmm0, %xmm0
; SSE2-NEXT: pmovmskb %xmm0, %eax
-; SSE2-NEXT: pxor %xmm0, %xmm0
-; SSE2-NEXT: testb $1, %al
; SSE2-NEXT: xorps %xmm1, %xmm1
+; SSE2-NEXT: testb $1, %al
+; SSE2-NEXT: pxor %xmm0, %xmm0
; SSE2-NEXT: jne LBB20_1
; SSE2-NEXT: ## %bb.2: ## %else
; SSE2-NEXT: testb $2, %al
@@ -2693,9 +2692,9 @@ define <8 x i32> @load_v8i32_v8i1_zero(<8 x i1> %mask, ptr %addr) {
; SSE42-NEXT: psllw $15, %xmm0
; SSE42-NEXT: packsswb %xmm0, %xmm0
; SSE42-NEXT: pmovmskb %xmm0, %eax
-; SSE42-NEXT: pxor %xmm0, %xmm0
-; SSE42-NEXT: testb $1, %al
; SSE42-NEXT: pxor %xmm1, %xmm1
+; SSE42-NEXT: testb $1, %al
+; SSE42-NEXT: pxor %xmm0, %xmm0
; SSE42-NEXT: jne LBB20_1
; SSE42-NEXT: ## %bb.2: ## %else
; SSE42-NEXT: testb $2, %al
diff --git a/llvm/test/CodeGen/X86/min-legal-vector-width.ll b/llvm/test/CodeGen/X86/min-legal-vector-width.ll
index d752659f94a50..0ebe7d4ea588d 100644
--- a/llvm/test/CodeGen/X86/min-legal-vector-width.ll
+++ b/llvm/test/CodeGen/X86/min-legal-vector-width.ll
@@ -174,8 +174,8 @@ define dso_local i32 @_Z9test_charPcS_i_256(ptr nocapture readonly, ptr nocaptur
; CHECK-SKX: # %bb.0: # %entry
; CHECK-SKX-NEXT: movl %edx, %eax
; CHECK-SKX-NEXT: vpxor %xmm0, %xmm0, %xmm0
-; CHECK-SKX-NEXT: xorl %ecx, %ecx
; CHECK-SKX-NEXT: vpxor %xmm1, %xmm1, %xmm1
+; CHECK-SKX-NEXT: xorl %ecx, %ecx
; CHECK-SKX-NEXT: vpxor %xmm2, %xmm2, %xmm2
; CHECK-SKX-NEXT: .p2align 4
; CHECK-SKX-NEXT: .LBB8_1: # %vector.body
@@ -184,17 +184,17 @@ define dso_local i32 @_Z9test_charPcS_i_256(ptr nocapture readonly, ptr nocaptur
; CHECK-SKX-NEXT: vpmovsxbw (%rdi,%rcx), %ymm4
; CHECK-SKX-NEXT: vpmovsxbw 16(%rsi,%rcx), %ymm5
; CHECK-SKX-NEXT: vpmaddwd %ymm3, %ymm5, %ymm3
-; CHECK-SKX-NEXT: vpaddd %ymm2, %ymm3, %ymm2
+; CHECK-SKX-NEXT: vpaddd %ymm1, %ymm3, %ymm1
; CHECK-SKX-NEXT: vpmovsxbw (%rsi,%rcx), %ymm3
; CHECK-SKX-NEXT: vpmaddwd %ymm4, %ymm3, %ymm3
-; CHECK-SKX-NEXT: vpaddd %ymm1, %ymm3, %ymm1
+; CHECK-SKX-NEXT: vpaddd %ymm2, %ymm3, %ymm2
; CHECK-SKX-NEXT: addq $32, %rcx
; CHECK-SKX-NEXT: cmpq %rcx, %rax
; CHECK-SKX-NEXT: jne .LBB8_1
; CHECK-SKX-NEXT: # %bb.2: # %middle.block
-; CHECK-SKX-NEXT: vpaddd %ymm0, %ymm1, %ymm1
-; CHECK-SKX-NEXT: vpaddd %ymm0, %ymm2, %ymm0
+; CHECK-SKX-NEXT: vpaddd %ymm0, %ymm2, %ymm2
; CHECK-SKX-NEXT: vpaddd %ymm0, %ymm1, %ymm0
+; CHECK-SKX-NEXT: vpaddd %ymm0, %ymm2, %ymm0
; CHECK-SKX-NEXT: vextracti128 $1, %ymm0, %xmm1
; CHECK-SKX-NEXT: vpaddd %xmm1, %xmm0, %xmm0
; CHECK-SKX-NEXT: vpshufd {{.*#+}} xmm1 = xmm0[2,3,2,3]
@@ -209,8 +209,8 @@ define dso_local i32 @_Z9test_charPcS_i_256(ptr nocapture readonly, ptr nocaptur
; CHECK-AVX512: # %bb.0: # %entry
; CHECK-AVX512-NEXT: movl %edx, %eax
; CHECK-AVX512-NEXT: vpxor %xmm0, %xmm0, %xmm0
-; CHECK-AVX512-NEXT: xorl %ecx, %ecx
; CHECK-AVX512-NEXT: vpxor %xmm1, %xmm1, %xmm1
+; CHECK-AVX512-NEXT: xorl %ecx, %ecx
; CHECK-AVX512-NEXT: vpxor %xmm2, %xmm2, %xmm2
; CHECK-AVX512-NEXT: .p2align 4
; CHECK-AVX512-NEXT: .LBB8_1: # %vector.body
@@ -219,17 +219,17 @@ define dso_local i32 @_Z9test_charPcS_i_256(ptr nocapture readonly, ptr nocaptur
; CHECK-AVX512-NEXT: vpmovsxbw (%rdi,%rcx), %ymm4
; CHECK-AVX512-NEXT: vpmovsxbw 16(%rsi,%rcx), %ymm5
; CHECK-AVX512-NEXT: vpmaddwd %ymm3, %ymm5, %ymm3
-; CHECK-AVX512-NEXT: vpaddd %ymm2, %ymm3, %ymm2
+; CHECK-AVX512-NEXT: vpaddd %ymm1, %ymm3, %ymm1
; CHECK-AVX512-NEXT: vpmovsxbw (%rsi,%rcx), %ymm3
; CHECK-AVX512-NEXT: vpmaddwd %ymm4, %ymm3, %ymm3
-; CHECK-AVX512-NEXT: vpaddd %ymm1, %ymm3, %ymm1
+; CHECK-AVX512-NEXT: vpaddd %ymm2, %ymm3, %ymm2
; CHECK-AVX512-NEXT: addq $32, %rcx
; CHECK-AVX512-NEXT: cmpq %rcx, %rax
; CHECK-AVX512-NEXT: jne .LBB8_1
; CHECK-AVX512-NEXT: # %bb.2: # %middle.block
-; CHECK-AVX512-NEXT: vpaddd %ymm0, %ymm1, %ymm1
-; CHECK-AVX512-NEXT: vpaddd %ymm0, %ymm2, %ymm0
+; CHECK-AVX512-NEXT: vpaddd %ymm0, %ymm2, %ymm2
; CHECK-AVX512-NEXT: vpaddd %ymm0, %ymm1, %ymm0
+; CHECK-AVX512-NEXT: vpaddd %ymm0, %ymm2, %ymm0
; CHECK-AVX512-NEXT: vextracti128 $1, %ymm0, %xmm1
; CHECK-AVX512-NEXT: vpaddd %xmm1, %xmm0, %xmm0
; CHECK-AVX512-NEXT: vpshufd {{.*#+}} xmm1 = xmm0[2,3,2,3]
@@ -244,8 +244,8 @@ define dso_local i32 @_Z9test_charPcS_i_256(ptr nocapture readonly, ptr nocaptur
; CHECK-VBMI: # %bb.0: # %entry
; CHECK-VBMI-NEXT: movl %edx, %eax
; CHECK-VBMI-NEXT: vpxor %xmm0, %xmm0, %xmm0
-; CHECK-VBMI-NEXT: xorl %ecx, %ecx
; CHECK-VBMI-NEXT: vpxor %xmm1, %xmm1, %xmm1
+; CHECK-VBMI-NEXT: xorl %ecx, %ecx
; CHECK-VBMI-NEXT: vpxor %xmm2, %xmm2, %xmm2
; CHECK-VBMI-NEXT: .p2align 4
; CHECK-VBMI-NEXT: .LBB8_1: # %vector.body
@@ -254,17 +254,17 @@ define dso_local i32 @_Z9test_charPcS_i_256(ptr nocapture readonly, ptr nocaptur
; CHECK-VBMI-NEXT: vpmovsxbw (%rdi,%rcx), %ymm4
; CHECK-VBMI-NEXT: vpmovsxbw 16(%rsi,%rcx), %ymm5
; CHECK-VBMI-NEXT: vpmaddwd %ymm3, %ymm5, %ymm3
-; CHECK-VBMI-NEXT: vpaddd %ymm2, %ymm3, %ymm2
+; CHECK-VBMI-NEXT: vpaddd %ymm1, %ymm3, %ymm1
; CHECK-VBMI-NEXT: vpmovsxbw (%rsi,%rcx), %ymm3
; CHECK-VBMI-NEXT: vpmaddwd %ymm4, %ymm3, %ymm3
-; CHECK-VBMI-NEXT: vpaddd %ymm1, %ymm3, %ymm1
+; CHECK-VBMI-NEXT: vpaddd %ymm2, %ymm3, %ymm2
; CHECK-VBMI-NEXT: addq $32, %rcx
; CHECK-VBMI-NEXT: cmpq %rcx, %rax
; CHECK-VBMI-NEXT: jne .LBB8_1
; CHECK-VBMI-NEXT: # %bb.2: # %middle.block
-; CHECK-VBMI-NEXT: vpaddd %ymm0, %ymm1, %ymm1
-; CHECK-VBMI-NEXT: vpaddd %ymm0, %ymm2, %ymm0
+; CHECK-VBMI-NEXT: vpaddd %ymm0, %ymm2, %ymm2
; CHECK-VBMI-NEXT: vpaddd %ymm0, %ymm1, %ymm0
+; CHECK-VBMI-NEXT: vpaddd %ymm0, %ymm2, %ymm0
; CHECK-VBMI-NEXT: vextracti128 $1, %ymm0, %xmm1
; CHECK-VBMI-NEXT: vpaddd %xmm1, %xmm0, %xmm0
; CHECK-VBMI-NEXT: vpshufd {{.*#+}} xmm1 = xmm0[2,3,2,3]
diff --git a/llvm/test/CodeGen/X86/pcsections-atomics.ll b/llvm/test/CodeGen/X86/pcsections-atomics.ll
index 672ebc1ec7275..c84bafa612948 100644
--- a/llvm/test/CodeGen/X86/pcsections-atomics.ll
+++ b/llvm/test/CodeGen/X86/pcsections-atomics.ll
@@ -689,11 +689,12 @@ define void @atomic8_nand_monotonic(ptr %a) {
; O0-NEXT: orb $-43, %dl
; O0-NEXT: .Lpcsection19:
; O0-NEXT: lock cmpxchgb %dl, (%rcx)
+; O0-NEXT: movb %al, %cl
; O0-NEXT: .Lpcsection20:
-; O0-NEXT: sete %cl
+; O0-NEXT: sete %al
+; O0-NEXT: movb %cl, {{[-0-9]+}}(%r{{[sb]}}p) # 1-byte Spill
; O0-NEXT: .Lpcsection21:
-; O0-NEXT: testb $1, %cl
-; O0-NEXT: movb %al, {{[-0-9]+}}(%r{{[sb]}}p) # 1-byte Spill
+; O0-NEXT: testb $1, %al
; O0-NEXT: .Lpcsection22:
; O0-NEXT: jne .LBB16_2
; O0-NEXT: jmp .LBB16_1
@@ -1027,11 +1028,12 @@ define void @atomic8_nand_acquire(ptr %a) {
; O0-NEXT: orb $-43, %dl
; O0-NEXT: .Lpcsection32:
; O0-NEXT: lock cmpxchgb %dl, (%rcx)
+; O0-NEXT: movb %al, %cl
; O0-NEXT: .Lpcsection33:
-; O0-NEXT: sete %cl
+; O0-NEXT: sete %al
+; O0-NEXT: movb %cl, {{[-0-9]+}}(%r{{[sb]}}p) # 1-byte Spill
; O0-NEXT: .Lpcsection34:
-; O0-NEXT: testb $1, %cl
-; O0-NEXT: movb %al, {{[-0-9]+}}(%r{{[sb]}}p) # 1-byte Spill
+; O0-NEXT: testb $1, %al
; O0-NEXT: .Lpcsection35:
; O0-NEXT: jne .LBB23_2
; O0-NEXT: jmp .LBB23_1
@@ -1365,11 +1367,12 @@ define void @atomic8_nand_release(ptr %a) {
; O0-NEXT: orb $-43, %dl
; O0-NEXT: .Lpcsection45:
; O0-NEXT: lock cmpxchgb %dl, (%rcx)
+; O0-NEXT: movb %al, %cl
; O0-NEXT: .Lpcsection46:
-; O0-NEXT: sete %cl
+; O0-NEXT: sete %al
+; O0-NEXT: movb %cl, {{[-0-9]+}}(%r{{[sb]}}p) # 1-byte Spill
; O0-NEXT: .Lpcsection47:
-; O0-NEXT: testb $1, %cl
-; O0-NEXT: movb %al, {{[-0-9]+}}(%r{{[sb]}}p) # 1-byte Spill
+; O0-NEXT: testb $1, %al
; O0-NEXT: .Lpcsection48:
; O0-NEXT: jne .LBB30_2
; O0-NEXT: jmp .LBB30_1
@@ -1703,11 +1706,12 @@ define void @atomic8_nand_acq_rel(ptr %a) {
; O0-NEXT: orb $-43, %dl
; O0-NEXT: .Lpcsection58:
; O0-NEXT: lock cmpxchgb %dl, (%rcx)
+; O0-NEXT: movb %al, %cl
; O0-NEXT: .Lpcsection59:
-; O0-NEXT: sete %cl
+; O0-NEXT: sete %al
+; O0-NEXT: movb %cl, {{[-0-9]+}}(%r{{[sb]}}p) # 1-byte Spill
; O0-NEXT: .Lpcsection60:
-; O0-NEXT: testb $1, %cl
-; O0-NEXT: movb %al, {{[-0-9]+}}(%r{{[sb]}}p) # 1-byte Spill
+; O0-NEXT: testb $1, %al
; O0-NEXT: .Lpcsection61:
; O0-NEXT: jne .LBB37_2
; O0-NEXT: jmp .LBB37_1
@@ -2041,11 +2045,12 @@ define void @atomic8_nand_seq_cst(ptr %a) {
; O0-NEXT: orb $-43, %dl
; O0-NEXT: .Lpcsection71:
; O0-NEXT: lock cmpxchgb %dl, (%rcx)
+; O0-NEXT: movb %al, %cl
; O0-NEXT: .Lpcsection72:
-; O0-NEXT: sete %cl
+; O0-NEXT: sete %al
+; O0-NEXT: movb %cl, {{[-0-9]+}}(%r{{[sb]}}p) # 1-byte Spill
; O0-NEXT: .Lpcsection73:
-; O0-NEXT: testb $1, %cl
-; O0-NEXT: movb %al, {{[-0-9]+}}(%r{{[sb]}}p) # 1-byte Spill
+; O0-NEXT: testb $1, %al
; O0-NEXT: .Lpcsection74:
; O0-NEXT: jne .LBB44_2
; O0-NEXT: jmp .LBB44_1
@@ -3134,11 +3139,12 @@ define void @atomic16_nand_monotonic(ptr %a) {
; O0-NEXT: # kill: def $dx killed $dx killed $edx
; O0-NEXT: .Lpcsection119:
; O0-NEXT: lock cmpxchgw %dx, (%rcx)
+; O0-NEXT: movw %ax, %cx
; O0-NEXT: .Lpcsection120:
-; O0-NEXT: sete %cl
+; O0-NEXT: sete %al
+; O0-NEXT: movw %cx, {{[-0-9]+}}(%r{{[sb]}}p) # 2-byte Spill
; O0-NEXT: .Lpcsection121:
-; O0-NEXT: testb $1, %cl
-; O0-NEXT: movw %ax, {{[-0-9]+}}(%r{{[sb]}}p) # 2-byte Spill
+; O0-NEXT: testb $1, %al
; O0-NEXT: .Lpcsection122:
; O0-NEXT: jne .LBB64_2
; O0-NEXT: jmp .LBB64_1
@@ -3488,11 +3494,12 @@ define void @atomic16_nand_acquire(ptr %a) {
; O0-NEXT: # kill: def $dx killed $dx killed $edx
; O0-NEXT: .Lpcsection134:
; O0-NEXT: lock cmpxchgw %dx, (%rcx)
+; O0-NEXT: movw %ax, %cx
; O0-NEXT: .Lpcsection135:
-; O0-NEXT: sete %cl
+; O0-NEXT: sete %al
+; O0-NEXT: movw %cx, {{[-0-9]+}}(%r{{[sb]}}p) # 2-byte Spill
; O0-NEXT: .Lpcsection136:
-; O0-NEXT: testb $1, %cl
-; O0-NEXT: movw %ax, {{[-0-9]+}}(%r{{[sb]}}p) # 2-byte Spill
+; O0-NEXT: testb $1, %al
; O0-NEXT: .Lpcsection137:
; O0-NEXT: jne .LBB71_2
; O0-NEXT: jmp .LBB71_1
@@ -3842,11 +3849,12 @@ define void @atomic16_nand_release(ptr %a) {
; O0-NEXT: # kill: def $dx killed $dx killed $edx
; O0-NEXT: .Lpcsection149:
; O0-NEXT: lock cmpxchgw %dx, (%rcx)
+; O0-NEXT: movw %ax, %cx
; O0-NEXT: .Lpcsection150:
-; O0-NEXT: sete %cl
+; O0-NEXT: sete %al
+; O0-NEXT: movw %cx, {{[-0-9]+}}(%r{{[sb]}}p) # 2-byte Spill
; O0-NEXT: .Lpcsection151:
-; O0-NEXT: testb $1, %cl
-; O0-NEXT: movw %ax, {{[-0-9]+}}(%r{{[sb]}}p) # 2-byte Spill
+; O0-NEXT: testb $1, %al
; O0-NEXT: .Lpcsection152:
; O0-NEXT: jne .LBB78_2
; O0-NEXT: jmp .LBB78_1
@@ -4196,11 +4204,12 @@ define void @atomic16_nand_acq_rel(ptr %a) {
; O0-NEXT: # kill: def $dx killed $dx killed $edx
; O0-NEXT: .Lpcsection164:
; O0-NEXT: lock cmpxchgw %dx, (%rcx)
+; O0-NEXT: movw %ax, %cx
; O0-NEXT: .Lpcsection165:
-; O0-NEXT: sete %cl
+; O0-NEXT: sete %al
+; O0-NEXT: movw %cx, {{[-0-9]+}}(%r{{[sb]}}p) # 2-byte Spill
; O0-NEXT: .Lpcsection166:
-; O0-NEXT: testb $1, %cl
-; O0-NEXT: movw %ax, {{[-0-9]+}}(%r{{[sb]}}p) # 2-byte Spill
+; O0-NEXT: testb $1, %al
; O0-NEXT: .Lpcsection167:
; O0-NEXT: jne .LBB85_2
; O0-NEXT: jmp .LBB85_1
@@ -4550,11 +4559,12 @@ define void @atomic16_nand_seq_cst(ptr %a) {
; O0-NEXT: # kill: def $dx killed $dx killed $edx
; O0-NEXT: .Lpcsection179:
; O0-NEXT: lock cmpxchgw %dx, (%rcx)
+; O0-NEXT: movw %ax, %cx
; O0-NEXT: .Lpcsection180:
-; O0-NEXT: sete %cl
+; O0-NEXT: sete %al
+; O0-NEXT: movw %cx, {{[-0-9]+}}(%r{{[sb]}}p) # 2-byte Spill
; O0-NEXT: .Lpcsection181:
-; O0-NEXT: testb $1, %cl
-; O0-NEXT: movw %ax, {{[-0-9]+}}(%r{{[sb]}}p) # 2-byte Spill
+; O0-NEXT: testb $1, %al
; O0-NEXT: .Lpcsection182:
; O0-NEXT: jne .LBB92_2
; O0-NEXT: jmp .LBB92_1
@@ -5606,11 +5616,12 @@ define void @atomic32_nand_monotonic(ptr %a) {
; O0-NEXT: orl $-43, %edx
; O0-NEXT: .Lpcsection225:
; O0-NEXT: lock cmpxchgl %edx, (%rcx)
+; O0-NEXT: movl %eax, %ecx
; O0-NEXT: .Lpcsection226:
-; O0-NEXT: sete %cl
+; O0-NEXT: sete %al
+; O0-NEXT: movl %ecx, {{[-0-9]+}}(%r{{[sb]}}p) # 4-byte Spill
; O0-NEXT: .Lpcsection227:
-; O0-NEXT: testb $1, %cl
-; O0-NEXT: movl %eax, {{[-0-9]+}}(%r{{[sb]}}p) # 4-byte Spill
+; O0-NEXT: testb $1, %al
; O0-NEXT: .Lpcsection228:
; O0-NEXT: jne .LBB112_2
; O0-NEXT: jmp .LBB112_1
@@ -5944,11 +5955,12 @@ define void @atomic32_nand_acquire(ptr %a) {
; O0-NEXT: orl $-43, %edx
; O0-NEXT: .Lpcsection238:
; O0-NEXT: lock cmpxchgl %edx, (%rcx)
+; O0-NEXT: movl %eax, %ecx
; O0-NEXT: .Lpcsection239:
-; O0-NEXT: sete %cl
+; O0-NEXT: sete %al
+; O0-NEXT: movl %ecx, {{[-0-9]+}}(%r{{[sb]}}p) # 4-byte Spill
; O0-NEXT: .Lpcsection240:
-; O0-NEXT: testb $1, %cl
-; O0-NEXT: movl %eax, {{[-0-9]+}}(%r{{[sb]}}p) # 4-byte Spill
+; O0-NEXT: testb $1, %al
; O0-NEXT: .Lpcsection241:
; O0-NEXT: jne .LBB119_2
; O0-NEXT: jmp .LBB119_1
@@ -6282,11 +6294,12 @@ define void @atomic32_nand_release(ptr %a) {
; O0-NEXT: orl $-43, %edx
; O0-NEXT: .Lpcsection251:
; O0-NEXT: lock cmpxchgl %edx, (%rcx)
+; O0-NEXT: movl %eax, %ecx
; O0-NEXT: .Lpcsection252:
-; O0-NEXT: sete %cl
+; O0-NEXT: sete %al
+; O0-NEXT: movl %ecx, {{[-0-9]+}}(%r{{[sb]}}p) # 4-byte Spill
; O0-NEXT: .Lpcsection253:
-; O0-NEXT: testb $1, %cl
-; O0-NEXT: movl %eax, {{[-0-9]+}}(%r{{[sb]}}p) # 4-byte Spill
+; O0-NEXT: testb $1, %al
; O0-NEXT: .Lpcsection254:
; O0-NEXT: jne .LBB126_2
; O0-NEXT: jmp .LBB126_1
@@ -6620,11 +6633,12 @@ define void @atomic32_nand_acq_rel(ptr %a) {
; O0-NEXT: orl $-43, %edx
; O0-NEXT: .Lpcsection264:
; O0-NEXT: lock cmpxchgl %edx, (%rcx)
+; O0-NEXT: movl %eax, %ecx
; O0-NEXT: .Lpcsection265:
-; O0-NEXT: sete %cl
+; O0-NEXT: sete %al
+; O0-NEXT: movl %ecx, {{[-0-9]+}}(%r{{[sb]}}p) # 4-byte Spill
; O0-NEXT: .Lpcsection266:
-; O0-NEXT: testb $1, %cl
-; O0-NEXT: movl %eax, {{[-0-9]+}}(%r{{[sb]}}p) # 4-byte Spill
+; O0-NEXT: testb $1, %al
; O0-NEXT: .Lpcsection267:
; O0-NEXT: jne .LBB133_2
; O0-NEXT: jmp .LBB133_1
@@ -6958,11 +6972,12 @@ define void @atomic32_nand_seq_cst(ptr %a) {
; O0-NEXT: orl $-43, %edx
; O0-NEXT: .Lpcsection277:
; O0-NEXT: lock cmpxchgl %edx, (%rcx)
+; O0-NEXT: movl %eax, %ecx
; O0-NEXT: .Lpcsection278:
-; O0-NEXT: sete %cl
+; O0-NEXT: sete %al
+; O0-NEXT: movl %ecx, {{[-0-9]+}}(%r{{[sb]}}p) # 4-byte Spill
; O0-NEXT: .Lpcsection279:
-; O0-NEXT: testb $1, %cl
-; O0-NEXT: movl %eax, {{[-0-9]+}}(%r{{[sb]}}p) # 4-byte Spill
+; O0-NEXT: testb $1, %al
; O0-NEXT: .Lpcsection280:
; O0-NEXT: jne .LBB140_2
; O0-NEXT: jmp .LBB140_1
@@ -8128,11 +8143,12 @@ define void @atomic64_nand_monotonic(ptr %a) {
; O0-NEXT: orq $-43, %rdx
; O0-NEXT: .Lpcsection326:
; O0-NEXT: lock cmpxchgq %rdx, (%rcx)
+; O0-NEXT: movq %rax, %rcx
; O0-NEXT: .Lpcsection327:
-; O0-NEXT: sete %cl
+; O0-NEXT: sete %al
+; O0-NEXT: movq %rcx, {{[-0-9]+}}(%r{{[sb]}}p) # 8-byte Spill
; O0-NEXT: .Lpcsection328:
-; O0-NEXT: testb $1, %cl
-; O0-NEXT: movq %rax, {{[-0-9]+}}(%r{{[sb]}}p) # 8-byte Spill
+; O0-NEXT: testb $1, %al
; O0-NEXT: .Lpcsection329:
; O0-NEXT: jne .LBB162_2
; O0-NEXT: jmp .LBB162_1
@@ -8469,11 +8485,12 @@ define void @atomic64_nand_acquire(ptr %a) {
; O0-NEXT: orq $-43, %rdx
; O0-NEXT: .Lpcsection340:
; O0-NEXT: lock cmpxchgq %rdx, (%rcx)
+; O0-NEXT: movq %rax, %rcx
; O0-NEXT: .Lpcsection341:
-; O0-NEXT: sete %cl
+; O0-NEXT: sete %al
+; O0-NEXT: movq %rcx, {{[-0-9]+}}(%r{{[sb]}}p) # 8-byte Spill
; O0-NEXT: .Lpcsection342:
-; O0-NEXT: testb $1, %cl
-; O0-NEXT: movq %rax, {{[-0-9]+}}(%r{{[sb]}}p) # 8-byte Spill
+; O0-NEXT: testb $1, %al
; O0-NEXT: .Lpcsection343:
; O0-NEXT: jne .LBB169_2
; O0-NEXT: jmp .LBB169_1
@@ -8810,11 +8827,12 @@ define void @atomic64_nand_release(ptr %a) {
; O0-NEXT: orq $-43, %rdx
; O0-NEXT: .Lpcsection354:
; O0-NEXT: lock cmpxchgq %rdx, (%rcx)
+; O0-NEXT: movq %rax, %rcx
; O0-NEXT: .Lpcsection355:
-; O0-NEXT: sete %cl
+; O0-NEXT: sete %al
+; O0-NEXT: movq %rcx, {{[-0-9]+}}(%r{{[sb]}}p) # 8-byte Spill
; O0-NEXT: .Lpcsection356:
-; O0-NEXT: testb $1, %cl
-; O0-NEXT: movq %rax, {{[-0-9]+}}(%r{{[sb]}}p) # 8-byte Spill
+; O0-NEXT: testb $1, %al
; O0-NEXT: .Lpcsection357:
; O0-NEXT: jne .LBB176_2
; O0-NEXT: jmp .LBB176_1
@@ -9151,11 +9169,12 @@ define void @atomic64_nand_acq_rel(ptr %a) {
; O0-NEXT: orq $-43, %rdx
; O0-NEXT: .Lpcsection368:
; O0-NEXT: lock cmpxchgq %rdx, (%rcx)
+; O0-NEXT: movq %rax, %rcx
; O0-NEXT: .Lpcsection369:
-; O0-NEXT: sete %cl
+; O0-NEXT: sete %al
+; O0-NEXT: movq %rcx, {{[-0-9]+}}(%r{{[sb]}}p) # 8-byte Spill
; O0-NEXT: .Lpcsection370:
-; O0-NEXT: testb $1, %cl
-; O0-NEXT: movq %rax, {{[-0-9]+}}(%r{{[sb]}}p) # 8-byte Spill
+; O0-NEXT: testb $1, %al
; O0-NEXT: .Lpcsection371:
; O0-NEXT: jne .LBB183_2
; O0-NEXT: jmp .LBB183_1
@@ -9492,11 +9511,12 @@ define void @atomic64_nand_seq_cst(ptr %a) {
; O0-NEXT: orq $-43, %rdx
; O0-NEXT: .Lpcsection382:
; O0-NEXT: lock cmpxchgq %rdx, (%rcx)
+; O0-NEXT: movq %rax, %rcx
; O0-NEXT: .Lpcsection383:
-; O0-NEXT: sete %cl
+; O0-NEXT: sete %al
+; O0-NEXT: movq %rcx, {{[-0-9]+}}(%r{{[sb]}}p) # 8-byte Spill
; O0-NEXT: .Lpcsection384:
-; O0-NEXT: testb $1, %cl
-; O0-NEXT: movq %rax, {{[-0-9]+}}(%r{{[sb]}}p) # 8-byte Spill
+; O0-NEXT: testb $1, %al
; O0-NEXT: .Lpcsection385:
; O0-NEXT: jne .LBB190_2
; O0-NEXT: jmp .LBB190_1
@@ -10527,8 +10547,8 @@ define void @atomic128_store_unordered(ptr %a) {
; O0-NEXT: jmp .LBB203_1
; O0-NEXT: .LBB203_1: # %atomicrmw.start
; O0-NEXT: # =>This Inner Loop Header: Depth=1
-; O0-NEXT: movq {{[-0-9]+}}(%r{{[sb]}}p), %rdx # 8-byte Reload
; O0-NEXT: movq {{[-0-9]+}}(%r{{[sb]}}p), %rax # 8-byte Reload
+; O0-NEXT: movq {{[-0-9]+}}(%r{{[sb]}}p), %rdx # 8-byte Reload
; O0-NEXT: movq {{[-0-9]+}}(%r{{[sb]}}p), %rsi # 8-byte Reload
; O0-NEXT: .Lpcsection426:
; O0-NEXT: xorl %ecx, %ecx
@@ -10538,8 +10558,8 @@ define void @atomic128_store_unordered(ptr %a) {
; O0-NEXT: movl $42, %ebx
; O0-NEXT: .Lpcsection429:
; O0-NEXT: lock cmpxchg16b (%rsi)
-; O0-NEXT: movq %rax, {{[-0-9]+}}(%r{{[sb]}}p) # 8-byte Spill
; O0-NEXT: movq %rdx, {{[-0-9]+}}(%r{{[sb]}}p) # 8-byte Spill
+; O0-NEXT: movq %rax, {{[-0-9]+}}(%r{{[sb]}}p) # 8-byte Spill
; O0-NEXT: .Lpcsection430:
; O0-NEXT: jne .LBB203_1
; O0-NEXT: jmp .LBB203_2
@@ -10654,8 +10674,8 @@ define void @atomic128_store_monotonic(ptr %a) {
; O0-NEXT: jmp .LBB204_1
; O0-NEXT: .LBB204_1: # %atomicrmw.start
; O0-NEXT: # =>This Inner Loop Header: Depth=1
-; O0-NEXT: movq {{[-0-9]+}}(%r{{[sb]}}p), %rdx # 8-byte Reload
; O0-NEXT: movq {{[-0-9]+}}(%r{{[sb]}}p), %rax # 8-byte Reload
+; O0-NEXT: movq {{[-0-9]+}}(%r{{[sb]}}p), %rdx # 8-byte Reload
; O0-NEXT: movq {{[-0-9]+}}(%r{{[sb]}}p), %rsi # 8-byte Reload
; O0-NEXT: .Lpcsection434:
; O0-NEXT: xorl %ecx, %ecx
@@ -10665,8 +10685,8 @@ define void @atomic128_store_monotonic(ptr %a) {
; O0-NEXT: movl $42, %ebx
; O0-NEXT: .Lpcsection437:
; O0-NEXT: lock cmpxchg16b (%rsi)
-; O0-NEXT: movq %rax, {{[-0-9]+}}(%r{{[sb]}}p) # 8-byte Spill
; O0-NEXT: movq %rdx, {{[-0-9]+}}(%r{{[sb]}}p) # 8-byte Spill
+; O0-NEXT: movq %rax, {{[-0-9]+}}(%r{{[sb]}}p) # 8-byte Spill
; O0-NEXT: .Lpcsection438:
; O0-NEXT: jne .LBB204_1
; O0-NEXT: jmp .LBB204_2
@@ -10781,8 +10801,8 @@ define void @atomic128_store_release(ptr %a) {
; O0-NEXT: jmp .LBB205_1
; O0-NEXT: .LBB205_1: # %atomicrmw.start
; O0-NEXT: # =>This Inner Loop Header: Depth=1
-; O0-NEXT: movq {{[-0-9]+}}(%r{{[sb]}}p), %rdx # 8-byte Reload
; O0-NEXT: movq {{[-0-9]+}}(%r{{[sb]}}p), %rax # 8-byte Reload
+; O0-NEXT: movq {{[-0-9]+}}(%r{{[sb]}}p), %rdx # 8-byte Reload
; O0-NEXT: movq {{[-0-9]+}}(%r{{[sb]}}p), %rsi # 8-byte Reload
; O0-NEXT: .Lpcsection442:
; O0-NEXT: xorl %ecx, %ecx
@@ -10792,8 +10812,8 @@ define void @atomic128_store_release(ptr %a) {
; O0-NEXT: movl $42, %ebx
; O0-NEXT: .Lpcsection445:
; O0-NEXT: lock cmpxchg16b (%rsi)
-; O0-NEXT: movq %rax, {{[-0-9]+}}(%r{{[sb]}}p) # 8-byte Spill
; O0-NEXT: movq %rdx, {{[-0-9]+}}(%r{{[sb]}}p) # 8-byte Spill
+; O0-NEXT: movq %rax, {{[-0-9]+}}(%r{{[sb]}}p) # 8-byte Spill
; O0-NEXT: .Lpcsection446:
; O0-NEXT: jne .LBB205_1
; O0-NEXT: jmp .LBB205_2
@@ -10908,8 +10928,8 @@ define void @atomic128_store_seq_cst(ptr %a) {
; O0-NEXT: jmp .LBB206_1
; O0-NEXT: .LBB206_1: # %atomicrmw.start
; O0-NEXT: # =>This Inner Loop Header: Depth=1
-; O0-NEXT: movq {{[-0-9]+}}(%r{{[sb]}}p), %rdx # 8-byte Reload
; O0-NEXT: movq {{[-0-9]+}}(%r{{[sb]}}p), %rax # 8-byte Reload
+; O0-NEXT: movq {{[-0-9]+}}(%r{{[sb]}}p), %rdx # 8-byte Reload
; O0-NEXT: movq {{[-0-9]+}}(%r{{[sb]}}p), %rsi # 8-byte Reload
; O0-NEXT: .Lpcsection450:
; O0-NEXT: xorl %ecx, %ecx
@@ -10919,8 +10939,8 @@ define void @atomic128_store_seq_cst(ptr %a) {
; O0-NEXT: movl $42, %ebx
; O0-NEXT: .Lpcsection453:
; O0-NEXT: lock cmpxchg16b (%rsi)
-; O0-NEXT: movq %rax, {{[-0-9]+}}(%r{{[sb]}}p) # 8-byte Spill
; O0-NEXT: movq %rdx, {{[-0-9]+}}(%r{{[sb]}}p) # 8-byte Spill
+; O0-NEXT: movq %rax, {{[-0-9]+}}(%r{{[sb]}}p) # 8-byte Spill
; O0-NEXT: .Lpcsection454:
; O0-NEXT: jne .LBB206_1
; O0-NEXT: jmp .LBB206_2
@@ -11074,8 +11094,8 @@ define void @atomic128_xchg_monotonic(ptr %a) {
; O0-NEXT: jmp .LBB208_1
; O0-NEXT: .LBB208_1: # %atomicrmw.start
; O0-NEXT: # =>This Inner Loop Header: Depth=1
-; O0-NEXT: movq {{[-0-9]+}}(%r{{[sb]}}p), %rdx # 8-byte Reload
; O0-NEXT: movq {{[-0-9]+}}(%r{{[sb]}}p), %rax # 8-byte Reload
+; O0-NEXT: movq {{[-0-9]+}}(%r{{[sb]}}p), %rdx # 8-byte Reload
; O0-NEXT: movq {{[-0-9]+}}(%r{{[sb]}}p), %rsi # 8-byte Reload
; O0-NEXT: .Lpcsection459:
; O0-NEXT: xorl %ecx, %ecx
@@ -11085,8 +11105,8 @@ define void @atomic128_xchg_monotonic(ptr %a) {
; O0-NEXT: movl $42, %ebx
; O0-NEXT: .Lpcsection462:
; O0-NEXT: lock cmpxchg16b (%rsi)
-; O0-NEXT: movq %rax, {{[-0-9]+}}(%r{{[sb]}}p) # 8-byte Spill
; O0-NEXT: movq %rdx, {{[-0-9]+}}(%r{{[sb]}}p) # 8-byte Spill
+; O0-NEXT: movq %rax, {{[-0-9]+}}(%r{{[sb]}}p) # 8-byte Spill
; O0-NEXT: .Lpcsection463:
; O0-NEXT: jne .LBB208_1
; O0-NEXT: jmp .LBB208_2
@@ -11201,8 +11221,8 @@ define void @atomic128_add_monotonic(ptr %a) {
; O0-NEXT: jmp .LBB209_1
; O0-NEXT: .LBB209_1: # %atomicrmw.start
; O0-NEXT: # =>This Inner Loop Header: Depth=1
-; O0-NEXT: movq {{[-0-9]+}}(%r{{[sb]}}p), %rdx # 8-byte Reload
; O0-NEXT: movq {{[-0-9]+}}(%r{{[sb]}}p), %rax # 8-byte Reload
+; O0-NEXT: movq {{[-0-9]+}}(%r{{[sb]}}p), %rdx # 8-byte Reload
; O0-NEXT: movq {{[-0-9]+}}(%r{{[sb]}}p), %rsi # 8-byte Reload
; O0-NEXT: movq %rax, %rbx
; O0-NEXT: .Lpcsection467:
@@ -11212,8 +11232,8 @@ define void @atomic128_add_monotonic(ptr %a) {
; O0-NEXT: adcq $0, %rcx
; O0-NEXT: .Lpcsection469:
; O0-NEXT: lock cmpxchg16b (%rsi)
-; O0-NEXT: movq %rax, {{[-0-9]+}}(%r{{[sb]}}p) # 8-byte Spill
; O0-NEXT: movq %rdx, {{[-0-9]+}}(%r{{[sb]}}p) # 8-byte Spill
+; O0-NEXT: movq %rax, {{[-0-9]+}}(%r{{[sb]}}p) # 8-byte Spill
; O0-NEXT: .Lpcsection470:
; O0-NEXT: jne .LBB209_1
; O0-NEXT: jmp .LBB209_2
@@ -11334,8 +11354,8 @@ define void @atomic128_sub_monotonic(ptr %a) {
; O0-NEXT: jmp .LBB210_1
; O0-NEXT: .LBB210_1: # %atomicrmw.start
; O0-NEXT: # =>This Inner Loop Header: Depth=1
-; O0-NEXT: movq {{[-0-9]+}}(%r{{[sb]}}p), %rdx # 8-byte Reload
; O0-NEXT: movq {{[-0-9]+}}(%r{{[sb]}}p), %rax # 8-byte Reload
+; O0-NEXT: movq {{[-0-9]+}}(%r{{[sb]}}p), %rdx # 8-byte Reload
; O0-NEXT: movq {{[-0-9]+}}(%r{{[sb]}}p), %rsi # 8-byte Reload
; O0-NEXT: movq %rax, %rbx
; O0-NEXT: .Lpcsection474:
@@ -11345,8 +11365,8 @@ define void @atomic128_sub_monotonic(ptr %a) {
; O0-NEXT: adcq $-1, %rcx
; O0-NEXT: .Lpcsection476:
; O0-NEXT: lock cmpxchg16b (%rsi)
-; O0-NEXT: movq %rax, {{[-0-9]+}}(%r{{[sb]}}p) # 8-byte Spill
; O0-NEXT: movq %rdx, {{[-0-9]+}}(%r{{[sb]}}p) # 8-byte Spill
+; O0-NEXT: movq %rax, {{[-0-9]+}}(%r{{[sb]}}p) # 8-byte Spill
; O0-NEXT: .Lpcsection477:
; O0-NEXT: jne .LBB210_1
; O0-NEXT: jmp .LBB210_2
@@ -11467,8 +11487,8 @@ define void @atomic128_and_monotonic(ptr %a) {
; O0-NEXT: jmp .LBB211_1
; O0-NEXT: .LBB211_1: # %atomicrmw.start
; O0-NEXT: # =>This Inner Loop Header: Depth=1
-; O0-NEXT: movq {{[-0-9]+}}(%r{{[sb]}}p), %rdx # 8-byte Reload
; O0-NEXT: movq {{[-0-9]+}}(%r{{[sb]}}p), %rax # 8-byte Reload
+; O0-NEXT: movq {{[-0-9]+}}(%r{{[sb]}}p), %rdx # 8-byte Reload
; O0-NEXT: movq {{[-0-9]+}}(%r{{[sb]}}p), %rsi # 8-byte Reload
; O0-NEXT: movl %eax, %ecx
; O0-NEXT: .Lpcsection481:
@@ -11480,8 +11500,8 @@ define void @atomic128_and_monotonic(ptr %a) {
; O0-NEXT: # kill: def $rcx killed $ecx
; O0-NEXT: .Lpcsection484:
; O0-NEXT: lock cmpxchg16b (%rsi)
-; O0-NEXT: movq %rax, {{[-0-9]+}}(%r{{[sb]}}p) # 8-byte Spill
; O0-NEXT: movq %rdx, {{[-0-9]+}}(%r{{[sb]}}p) # 8-byte Spill
+; O0-NEXT: movq %rax, {{[-0-9]+}}(%r{{[sb]}}p) # 8-byte Spill
; O0-NEXT: .Lpcsection485:
; O0-NEXT: jne .LBB211_1
; O0-NEXT: jmp .LBB211_2
@@ -11599,8 +11619,8 @@ define void @atomic128_or_monotonic(ptr %a) {
; O0-NEXT: jmp .LBB212_1
; O0-NEXT: .LBB212_1: # %atomicrmw.start
; O0-NEXT: # =>This Inner Loop Header: Depth=1
-; O0-NEXT: movq {{[-0-9]+}}(%r{{[sb]}}p), %rcx # 8-byte Reload
; O0-NEXT: movq {{[-0-9]+}}(%r{{[sb]}}p), %rax # 8-byte Reload
+; O0-NEXT: movq {{[-0-9]+}}(%r{{[sb]}}p), %rcx # 8-byte Reload
; O0-NEXT: movq {{[-0-9]+}}(%r{{[sb]}}p), %rsi # 8-byte Reload
; O0-NEXT: movq %rax, %rbx
; O0-NEXT: .Lpcsection489:
@@ -11608,8 +11628,8 @@ define void @atomic128_or_monotonic(ptr %a) {
; O0-NEXT: movq %rcx, %rdx
; O0-NEXT: .Lpcsection490:
; O0-NEXT: lock cmpxchg16b (%rsi)
-; O0-NEXT: movq %rax, {{[-0-9]+}}(%r{{[sb]}}p) # 8-byte Spill
; O0-NEXT: movq %rdx, {{[-0-9]+}}(%r{{[sb]}}p) # 8-byte Spill
+; O0-NEXT: movq %rax, {{[-0-9]+}}(%r{{[sb]}}p) # 8-byte Spill
; O0-NEXT: .Lpcsection491:
; O0-NEXT: jne .LBB212_1
; O0-NEXT: jmp .LBB212_2
@@ -11724,8 +11744,8 @@ define void @atomic128_xor_monotonic(ptr %a) {
; O0-NEXT: jmp .LBB213_1
; O0-NEXT: .LBB213_1: # %atomicrmw.start
; O0-NEXT: # =>This Inner Loop Header: Depth=1
-; O0-NEXT: movq {{[-0-9]+}}(%r{{[sb]}}p), %rcx # 8-byte Reload
; O0-NEXT: movq {{[-0-9]+}}(%r{{[sb]}}p), %rax # 8-byte Reload
+; O0-NEXT: movq {{[-0-9]+}}(%r{{[sb]}}p), %rcx # 8-byte Reload
; O0-NEXT: movq {{[-0-9]+}}(%r{{[sb]}}p), %rsi # 8-byte Reload
; O0-NEXT: movq %rax, %rbx
; O0-NEXT: .Lpcsection495:
@@ -11733,8 +11753,8 @@ define void @atomic128_xor_monotonic(ptr %a) {
; O0-NEXT: movq %rcx, %rdx
; O0-NEXT: .Lpcsection496:
; O0-NEXT: lock cmpxchg16b (%rsi)
-; O0-NEXT: movq %rax, {{[-0-9]+}}(%r{{[sb]}}p) # 8-byte Spill
; O0-NEXT: movq %rdx, {{[-0-9]+}}(%r{{[sb]}}p) # 8-byte Spill
+; O0-NEXT: movq %rax, {{[-0-9]+}}(%r{{[sb]}}p) # 8-byte Spill
; O0-NEXT: .Lpcsection497:
; O0-NEXT: jne .LBB213_1
; O0-NEXT: jmp .LBB213_2
@@ -11849,8 +11869,8 @@ define void @atomic128_nand_monotonic(ptr %a) {
; O0-NEXT: jmp .LBB214_1
; O0-NEXT: .LBB214_1: # %atomicrmw.start
; O0-NEXT: # =>This Inner Loop Header: Depth=1
-; O0-NEXT: movq {{[-0-9]+}}(%r{{[sb]}}p), %rdx # 8-byte Reload
; O0-NEXT: movq {{[-0-9]+}}(%r{{[sb]}}p), %rax # 8-byte Reload
+; O0-NEXT: movq {{[-0-9]+}}(%r{{[sb]}}p), %rdx # 8-byte Reload
; O0-NEXT: movq {{[-0-9]+}}(%r{{[sb]}}p), %rsi # 8-byte Reload
; O0-NEXT: movl %eax, %ecx
; O0-NEXT: .Lpcsection501:
@@ -11864,8 +11884,8 @@ define void @atomic128_nand_monotonic(ptr %a) {
; O0-NEXT: movq $-1, %rcx
; O0-NEXT: .Lpcsection505:
; O0-NEXT: lock cmpxchg16b (%rsi)
-; O0-NEXT: movq %rax, {{[-0-9]+}}(%r{{[sb]}}p) # 8-byte Spill
; O0-NEXT: movq %rdx, {{[-0-9]+}}(%r{{[sb]}}p) # 8-byte Spill
+; O0-NEXT: movq %rax, {{[-0-9]+}}(%r{{[sb]}}p) # 8-byte Spill
; O0-NEXT: .Lpcsection506:
; O0-NEXT: jne .LBB214_1
; O0-NEXT: jmp .LBB214_2
@@ -11989,8 +12009,8 @@ define void @atomic128_xchg_acquire(ptr %a) {
; O0-NEXT: jmp .LBB215_1
; O0-NEXT: .LBB215_1: # %atomicrmw.start
; O0-NEXT: # =>This Inner Loop Header: Depth=1
-; O0-NEXT: movq {{[-0-9]+}}(%r{{[sb]}}p), %rdx # 8-byte Reload
; O0-NEXT: movq {{[-0-9]+}}(%r{{[sb]}}p), %rax # 8-byte Reload
+; O0-NEXT: movq {{[-0-9]+}}(%r{{[sb]}}p), %rdx # 8-byte Reload
; O0-NEXT: movq {{[-0-9]+}}(%r{{[sb]}}p), %rsi # 8-byte Reload
; O0-NEXT: .Lpcsection510:
; O0-NEXT: xorl %ecx, %ecx
@@ -12000,8 +12020,8 @@ define void @atomic128_xchg_acquire(ptr %a) {
; O0-NEXT: movl $42, %ebx
; O0-NEXT: .Lpcsection513:
; O0-NEXT: lock cmpxchg16b (%rsi)
-; O0-NEXT: movq %rax, {{[-0-9]+}}(%r{{[sb]}}p) # 8-byte Spill
; O0-NEXT: movq %rdx, {{[-0-9]+}}(%r{{[sb]}}p) # 8-byte Spill
+; O0-NEXT: movq %rax, {{[-0-9]+}}(%r{{[sb]}}p) # 8-byte Spill
; O0-NEXT: .Lpcsection514:
; O0-NEXT: jne .LBB215_1
; O0-NEXT: jmp .LBB215_2
@@ -12116,8 +12136,8 @@ define void @atomic128_add_acquire(ptr %a) {
; O0-NEXT: jmp .LBB216_1
; O0-NEXT: .LBB216_1: # %atomicrmw.start
; O0-NEXT: # =>This Inner Loop Header: Depth=1
-; O0-NEXT: movq {{[-0-9]+}}(%r{{[sb]}}p), %rdx # 8-byte Reload
; O0-NEXT: movq {{[-0-9]+}}(%r{{[sb]}}p), %rax # 8-byte Reload
+; O0-NEXT: movq {{[-0-9]+}}(%r{{[sb]}}p), %rdx # 8-byte Reload
; O0-NEXT: movq {{[-0-9]+}}(%r{{[sb]}}p), %rsi # 8-byte Reload
; O0-NEXT: movq %rax, %rbx
; O0-NEXT: .Lpcsection518:
@@ -12127,8 +12147,8 @@ define void @atomic128_add_acquire(ptr %a) {
; O0-NEXT: adcq $0, %rcx
; O0-NEXT: .Lpcsection520:
; O0-NEXT: lock cmpxchg16b (%rsi)
-; O0-NEXT: movq %rax, {{[-0-9]+}}(%r{{[sb]}}p) # 8-byte Spill
; O0-NEXT: movq %rdx, {{[-0-9]+}}(%r{{[sb]}}p) # 8-byte Spill
+; O0-NEXT: movq %rax, {{[-0-9]+}}(%r{{[sb]}}p) # 8-byte Spill
; O0-NEXT: .Lpcsection521:
; O0-NEXT: jne .LBB216_1
; O0-NEXT: jmp .LBB216_2
@@ -12249,8 +12269,8 @@ define void @atomic128_sub_acquire(ptr %a) {
; O0-NEXT: jmp .LBB217_1
; O0-NEXT: .LBB217_1: # %atomicrmw.start
; O0-NEXT: # =>This Inner Loop Header: Depth=1
-; O0-NEXT: movq {{[-0-9]+}}(%r{{[sb]}}p), %rdx # 8-byte Reload
; O0-NEXT: movq {{[-0-9]+}}(%r{{[sb]}}p), %rax # 8-byte Reload
+; O0-NEXT: movq {{[-0-9]+}}(%r{{[sb]}}p), %rdx # 8-byte Reload
; O0-NEXT: movq {{[-0-9]+}}(%r{{[sb]}}p), %rsi # 8-byte Reload
; O0-NEXT: movq %rax, %rbx
; O0-NEXT: .Lpcsection525:
@@ -12260,8 +12280,8 @@ define void @atomic128_sub_acquire(ptr %a) {
; O0-NEXT: adcq $-1, %rcx
; O0-NEXT: .Lpcsection527:
; O0-NEXT: lock cmpxchg16b (%rsi)
-; O0-NEXT: movq %rax, {{[-0-9]+}}(%r{{[sb]}}p) # 8-byte Spill
; O0-NEXT: movq %rdx, {{[-0-9]+}}(%r{{[sb]}}p) # 8-byte Spill
+; O0-NEXT: movq %rax, {{[-0-9]+}}(%r{{[sb]}}p) # 8-byte Spill
; O0-NEXT: .Lpcsection528:
; O0-NEXT: jne .LBB217_1
; O0-NEXT: jmp .LBB217_2
@@ -12382,8 +12402,8 @@ define void @atomic128_and_acquire(ptr %a) {
; O0-NEXT: jmp .LBB218_1
; O0-NEXT: .LBB218_1: # %atomicrmw.start
; O0-NEXT: # =>This Inner Loop Header: Depth=1
-; O0-NEXT: movq {{[-0-9]+}}(%r{{[sb]}}p), %rdx # 8-byte Reload
; O0-NEXT: movq {{[-0-9]+}}(%r{{[sb]}}p), %rax # 8-byte Reload
+; O0-NEXT: movq {{[-0-9]+}}(%r{{[sb]}}p), %rdx # 8-byte Reload
; O0-NEXT: movq {{[-0-9]+}}(%r{{[sb]}}p), %rsi # 8-byte Reload
; O0-NEXT: movl %eax, %ecx
; O0-NEXT: .Lpcsection532:
@@ -12395,8 +12415,8 @@ define void @atomic128_and_acquire(ptr %a) {
; O0-NEXT: # kill: def $rcx killed $ecx
; O0-NEXT: .Lpcsection535:
; O0-NEXT: lock cmpxchg16b (%rsi)
-; O0-NEXT: movq %rax, {{[-0-9]+}}(%r{{[sb]}}p) # 8-byte Spill
; O0-NEXT: movq %rdx, {{[-0-9]+}}(%r{{[sb]}}p) # 8-byte Spill
+; O0-NEXT: movq %rax, {{[-0-9]+}}(%r{{[sb]}}p) # 8-byte Spill
; O0-NEXT: .Lpcsection536:
; O0-NEXT: jne .LBB218_1
; O0-NEXT: jmp .LBB218_2
@@ -12514,8 +12534,8 @@ define void @atomic128_or_acquire(ptr %a) {
; O0-NEXT: jmp .LBB219_1
; O0-NEXT: .LBB219_1: # %atomicrmw.start
; O0-NEXT: # =>This Inner Loop Header: Depth=1
-; O0-NEXT: movq {{[-0-9]+}}(%r{{[sb]}}p), %rcx # 8-byte Reload
; O0-NEXT: movq {{[-0-9]+}}(%r{{[sb]}}p), %rax # 8-byte Reload
+; O0-NEXT: movq {{[-0-9]+}}(%r{{[sb]}}p), %rcx # 8-byte Reload
; O0-NEXT: movq {{[-0-9]+}}(%r{{[sb]}}p), %rsi # 8-byte Reload
; O0-NEXT: movq %rax, %rbx
; O0-NEXT: .Lpcsection540:
@@ -12523,8 +12543,8 @@ define void @atomic128_or_acquire(ptr %a) {
; O0-NEXT: movq %rcx, %rdx
; O0-NEXT: .Lpcsection541:
; O0-NEXT: lock cmpxchg16b (%rsi)
-; O0-NEXT: movq %rax, {{[-0-9]+}}(%r{{[sb]}}p) # 8-byte Spill
; O0-NEXT: movq %rdx, {{[-0-9]+}}(%r{{[sb]}}p) # 8-byte Spill
+; O0-NEXT: movq %rax, {{[-0-9]+}}(%r{{[sb]}}p) # 8-byte Spill
; O0-NEXT: .Lpcsection542:
; O0-NEXT: jne .LBB219_1
; O0-NEXT: jmp .LBB219_2
@@ -12639,8 +12659,8 @@ define void @atomic128_xor_acquire(ptr %a) {
; O0-NEXT: jmp .LBB220_1
; O0-NEXT: .LBB220_1: # %atomicrmw.start
; O0-NEXT: # =>This Inner Loop Header: Depth=1
-; O0-NEXT: movq {{[-0-9]+}}(%r{{[sb]}}p), %rcx # 8-byte Reload
; O0-NEXT: movq {{[-0-9]+}}(%r{{[sb]}}p), %rax # 8-byte Reload
+; O0-NEXT: movq {{[-0-9]+}}(%r{{[sb]}}p), %rcx # 8-byte Reload
; O0-NEXT: movq {{[-0-9]+}}(%r{{[sb]}}p), %rsi # 8-byte Reload
; O0-NEXT: movq %rax, %rbx
; O0-NEXT: .Lpcsection546:
@@ -12648,8 +12668,8 @@ define void @atomic128_xor_acquire(ptr %a) {
; O0-NEXT: movq %rcx, %rdx
; O0-NEXT: .Lpcsection547:
; O0-NEXT: lock cmpxchg16b (%rsi)
-; O0-NEXT: movq %rax, {{[-0-9]+}}(%r{{[sb]}}p) # 8-byte Spill
; O0-NEXT: movq %rdx, {{[-0-9]+}}(%r{{[sb]}}p) # 8-byte Spill
+; O0-NEXT: movq %rax, {{[-0-9]+}}(%r{{[sb]}}p) # 8-byte Spill
; O0-NEXT: .Lpcsection548:
; O0-NEXT: jne .LBB220_1
; O0-NEXT: jmp .LBB220_2
@@ -12764,8 +12784,8 @@ define void @atomic128_nand_acquire(ptr %a) {
; O0-NEXT: jmp .LBB221_1
; O0-NEXT: .LBB221_1: # %atomicrmw.start
; O0-NEXT: # =>This Inner Loop Header: Depth=1
-; O0-NEXT: movq {{[-0-9]+}}(%r{{[sb]}}p), %rdx # 8-byte Reload
; O0-NEXT: movq {{[-0-9]+}}(%r{{[sb]}}p), %rax # 8-byte Reload
+; O0-NEXT: movq {{[-0-9]+}}(%r{{[sb]}}p), %rdx # 8-byte Reload
; O0-NEXT: movq {{[-0-9]+}}(%r{{[sb]}}p), %rsi # 8-byte Reload
; O0-NEXT: movl %eax, %ecx
; O0-NEXT: .Lpcsection552:
@@ -12779,8 +12799,8 @@ define void @atomic128_nand_acquire(ptr %a) {
; O0-NEXT: movq $-1, %rcx
; O0-NEXT: .Lpcsection556:
; O0-NEXT: lock cmpxchg16b (%rsi)
-; O0-NEXT: movq %rax, {{[-0-9]+}}(%r{{[sb]}}p) # 8-byte Spill
; O0-NEXT: movq %rdx, {{[-0-9]+}}(%r{{[sb]}}p) # 8-byte Spill
+; O0-NEXT: movq %rax, {{[-0-9]+}}(%r{{[sb]}}p) # 8-byte Spill
; O0-NEXT: .Lpcsection557:
; O0-NEXT: jne .LBB221_1
; O0-NEXT: jmp .LBB221_2
@@ -12904,8 +12924,8 @@ define void @atomic128_xchg_release(ptr %a) {
; O0-NEXT: jmp .LBB222_1
; O0-NEXT: .LBB222_1: # %atomicrmw.start
; O0-NEXT: # =>This Inner Loop Header: Depth=1
-; O0-NEXT: movq {{[-0-9]+}}(%r{{[sb]}}p), %rdx # 8-byte Reload
; O0-NEXT: movq {{[-0-9]+}}(%r{{[sb]}}p), %rax # 8-byte Reload
+; O0-NEXT: movq {{[-0-9]+}}(%r{{[sb]}}p), %rdx # 8-byte Reload
; O0-NEXT: movq {{[-0-9]+}}(%r{{[sb]}}p), %rsi # 8-byte Reload
; O0-NEXT: .Lpcsection561:
; O0-NEXT: xorl %ecx, %ecx
@@ -12915,8 +12935,8 @@ define void @atomic128_xchg_release(ptr %a) {
; O0-NEXT: movl $42, %ebx
; O0-NEXT: .Lpcsection564:
; O0-NEXT: lock cmpxchg16b (%rsi)
-; O0-NEXT: movq %rax, {{[-0-9]+}}(%r{{[sb]}}p) # 8-byte Spill
; O0-NEXT: movq %rdx, {{[-0-9]+}}(%r{{[sb]}}p) # 8-byte Spill
+; O0-NEXT: movq %rax, {{[-0-9]+}}(%r{{[sb]}}p) # 8-byte Spill
; O0-NEXT: .Lpcsection565:
; O0-NEXT: jne .LBB222_1
; O0-NEXT: jmp .LBB222_2
@@ -13030,8 +13050,8 @@ define void @atomic128_add_release(ptr %a) {
; O0-NEXT: jmp .LBB223_1
; O0-NEXT: .LBB223_1: # %atomicrmw.start
; O0-NEXT: # =>This Inner Loop Header: Depth=1
-; O0-NEXT: movq {{[-0-9]+}}(%r{{[sb]}}p), %rdx # 8-byte Reload
; O0-NEXT: movq {{[-0-9]+}}(%r{{[sb]}}p), %rax # 8-byte Reload
+; O0-NEXT: movq {{[-0-9]+}}(%r{{[sb]}}p), %rdx # 8-byte Reload
; O0-NEXT: movq {{[-0-9]+}}(%r{{[sb]}}p), %rsi # 8-byte Reload
; O0-NEXT: movq %rax, %rbx
; O0-NEXT: .Lpcsection569:
@@ -13041,8 +13061,8 @@ define void @atomic128_add_release(ptr %a) {
; O0-NEXT: adcq $0, %rcx
; O0-NEXT: .Lpcsection571:
; O0-NEXT: lock cmpxchg16b (%rsi)
-; O0-NEXT: movq %rax, {{[-0-9]+}}(%r{{[sb]}}p) # 8-byte Spill
; O0-NEXT: movq %rdx, {{[-0-9]+}}(%r{{[sb]}}p) # 8-byte Spill
+; O0-NEXT: movq %rax, {{[-0-9]+}}(%r{{[sb]}}p) # 8-byte Spill
; O0-NEXT: .Lpcsection572:
; O0-NEXT: jne .LBB223_1
; O0-NEXT: jmp .LBB223_2
@@ -13163,8 +13183,8 @@ define void @atomic128_sub_release(ptr %a) {
; O0-NEXT: jmp .LBB224_1
; O0-NEXT: .LBB224_1: # %atomicrmw.start
; O0-NEXT: # =>This Inner Loop Header: Depth=1
-; O0-NEXT: movq {{[-0-9]+}}(%r{{[sb]}}p), %rdx # 8-byte Reload
; O0-NEXT: movq {{[-0-9]+}}(%r{{[sb]}}p), %rax # 8-byte Reload
+; O0-NEXT: movq {{[-0-9]+}}(%r{{[sb]}}p), %rdx # 8-byte Reload
; O0-NEXT: movq {{[-0-9]+}}(%r{{[sb]}}p), %rsi # 8-byte Reload
; O0-NEXT: movq %rax, %rbx
; O0-NEXT: .Lpcsection576:
@@ -13174,8 +13194,8 @@ define void @atomic128_sub_release(ptr %a) {
; O0-NEXT: adcq $-1, %rcx
; O0-NEXT: .Lpcsection578:
; O0-NEXT: lock cmpxchg16b (%rsi)
-; O0-NEXT: movq %rax, {{[-0-9]+}}(%r{{[sb]}}p) # 8-byte Spill
; O0-NEXT: movq %rdx, {{[-0-9]+}}(%r{{[sb]}}p) # 8-byte Spill
+; O0-NEXT: movq %rax, {{[-0-9]+}}(%r{{[sb]}}p) # 8-byte Spill
; O0-NEXT: .Lpcsection579:
; O0-NEXT: jne .LBB224_1
; O0-NEXT: jmp .LBB224_2
@@ -13296,8 +13316,8 @@ define void @atomic128_and_release(ptr %a) {
; O0-NEXT: jmp .LBB225_1
; O0-NEXT: .LBB225_1: # %atomicrmw.start
; O0-NEXT: # =>This Inner Loop Header: Depth=1
-; O0-NEXT: movq {{[-0-9]+}}(%r{{[sb]}}p), %rdx # 8-byte Reload
; O0-NEXT: movq {{[-0-9]+}}(%r{{[sb]}}p), %rax # 8-byte Reload
+; O0-NEXT: movq {{[-0-9]+}}(%r{{[sb]}}p), %rdx # 8-byte Reload
; O0-NEXT: movq {{[-0-9]+}}(%r{{[sb]}}p), %rsi # 8-byte Reload
; O0-NEXT: movl %eax, %ecx
; O0-NEXT: .Lpcsection583:
@@ -13309,8 +13329,8 @@ define void @atomic128_and_release(ptr %a) {
; O0-NEXT: # kill: def $rcx killed $ecx
; O0-NEXT: .Lpcsection586:
; O0-NEXT: lock cmpxchg16b (%rsi)
-; O0-NEXT: movq %rax, {{[-0-9]+}}(%r{{[sb]}}p) # 8-byte Spill
; O0-NEXT: movq %rdx, {{[-0-9]+}}(%r{{[sb]}}p) # 8-byte Spill
+; O0-NEXT: movq %rax, {{[-0-9]+}}(%r{{[sb]}}p) # 8-byte Spill
; O0-NEXT: .Lpcsection587:
; O0-NEXT: jne .LBB225_1
; O0-NEXT: jmp .LBB225_2
@@ -13428,8 +13448,8 @@ define void @atomic128_or_release(ptr %a) {
; O0-NEXT: jmp .LBB226_1
; O0-NEXT: .LBB226_1: # %atomicrmw.start
; O0-NEXT: # =>This Inner Loop Header: Depth=1
-; O0-NEXT: movq {{[-0-9]+}}(%r{{[sb]}}p), %rcx # 8-byte Reload
; O0-NEXT: movq {{[-0-9]+}}(%r{{[sb]}}p), %rax # 8-byte Reload
+; O0-NEXT: movq {{[-0-9]+}}(%r{{[sb]}}p), %rcx # 8-byte Reload
; O0-NEXT: movq {{[-0-9]+}}(%r{{[sb]}}p), %rsi # 8-byte Reload
; O0-NEXT: movq %rax, %rbx
; O0-NEXT: .Lpcsection591:
@@ -13437,8 +13457,8 @@ define void @atomic128_or_release(ptr %a) {
; O0-NEXT: movq %rcx, %rdx
; O0-NEXT: .Lpcsection592:
; O0-NEXT: lock cmpxchg16b (%rsi)
-; O0-NEXT: movq %rax, {{[-0-9]+}}(%r{{[sb]}}p) # 8-byte Spill
; O0-NEXT: movq %rdx, {{[-0-9]+}}(%r{{[sb]}}p) # 8-byte Spill
+; O0-NEXT: movq %rax, {{[-0-9]+}}(%r{{[sb]}}p) # 8-byte Spill
; O0-NEXT: .Lpcsection593:
; O0-NEXT: jne .LBB226_1
; O0-NEXT: jmp .LBB226_2
@@ -13553,8 +13573,8 @@ define void @atomic128_xor_release(ptr %a) {
; O0-NEXT: jmp .LBB227_1
; O0-NEXT: .LBB227_1: # %atomicrmw.start
; O0-NEXT: # =>This Inner Loop Header: Depth=1
-; O0-NEXT: movq {{[-0-9]+}}(%r{{[sb]}}p), %rcx # 8-byte Reload
; O0-NEXT: movq {{[-0-9]+}}(%r{{[sb]}}p), %rax # 8-byte Reload
+; O0-NEXT: movq {{[-0-9]+}}(%r{{[sb]}}p), %rcx # 8-byte Reload
; O0-NEXT: movq {{[-0-9]+}}(%r{{[sb]}}p), %rsi # 8-byte Reload
; O0-NEXT: movq %rax, %rbx
; O0-NEXT: .Lpcsection597:
@@ -13562,8 +13582,8 @@ define void @atomic128_xor_release(ptr %a) {
; O0-NEXT: movq %rcx, %rdx
; O0-NEXT: .Lpcsection598:
; O0-NEXT: lock cmpxchg16b (%rsi)
-; O0-NEXT: movq %rax, {{[-0-9]+}}(%r{{[sb]}}p) # 8-byte Spill
; O0-NEXT: movq %rdx, {{[-0-9]+}}(%r{{[sb]}}p) # 8-byte Spill
+; O0-NEXT: movq %rax, {{[-0-9]+}}(%r{{[sb]}}p) # 8-byte Spill
; O0-NEXT: .Lpcsection599:
; O0-NEXT: jne .LBB227_1
; O0-NEXT: jmp .LBB227_2
@@ -13678,8 +13698,8 @@ define void @atomic128_nand_release(ptr %a) {
; O0-NEXT: jmp .LBB228_1
; O0-NEXT: .LBB228_1: # %atomicrmw.start
; O0-NEXT: # =>This Inner Loop Header: Depth=1
-; O0-NEXT: movq {{[-0-9]+}}(%r{{[sb]}}p), %rdx # 8-byte Reload
; O0-NEXT: movq {{[-0-9]+}}(%r{{[sb]}}p), %rax # 8-byte Reload
+; O0-NEXT: movq {{[-0-9]+}}(%r{{[sb]}}p), %rdx # 8-byte Reload
; O0-NEXT: movq {{[-0-9]+}}(%r{{[sb]}}p), %rsi # 8-byte Reload
; O0-NEXT: movl %eax, %ecx
; O0-NEXT: .Lpcsection603:
@@ -13693,8 +13713,8 @@ define void @atomic128_nand_release(ptr %a) {
; O0-NEXT: movq $-1, %rcx
; O0-NEXT: .Lpcsection607:
; O0-NEXT: lock cmpxchg16b (%rsi)
-; O0-NEXT: movq %rax, {{[-0-9]+}}(%r{{[sb]}}p) # 8-byte Spill
; O0-NEXT: movq %rdx, {{[-0-9]+}}(%r{{[sb]}}p) # 8-byte Spill
+; O0-NEXT: movq %rax, {{[-0-9]+}}(%r{{[sb]}}p) # 8-byte Spill
; O0-NEXT: .Lpcsection608:
; O0-NEXT: jne .LBB228_1
; O0-NEXT: jmp .LBB228_2
@@ -13818,8 +13838,8 @@ define void @atomic128_xchg_acq_rel(ptr %a) {
; O0-NEXT: jmp .LBB229_1
; O0-NEXT: .LBB229_1: # %atomicrmw.start
; O0-NEXT: # =>This Inner Loop Header: Depth=1
-; O0-NEXT: movq {{[-0-9]+}}(%r{{[sb]}}p), %rdx # 8-byte Reload
; O0-NEXT: movq {{[-0-9]+}}(%r{{[sb]}}p), %rax # 8-byte Reload
+; O0-NEXT: movq {{[-0-9]+}}(%r{{[sb]}}p), %rdx # 8-byte Reload
; O0-NEXT: movq {{[-0-9]+}}(%r{{[sb]}}p), %rsi # 8-byte Reload
; O0-NEXT: .Lpcsection612:
; O0-NEXT: xorl %ecx, %ecx
@@ -13829,8 +13849,8 @@ define void @atomic128_xchg_acq_rel(ptr %a) {
; O0-NEXT: movl $42, %ebx
; O0-NEXT: .Lpcsection615:
; O0-NEXT: lock cmpxchg16b (%rsi)
-; O0-NEXT: movq %rax, {{[-0-9]+}}(%r{{[sb]}}p) # 8-byte Spill
; O0-NEXT: movq %rdx, {{[-0-9]+}}(%r{{[sb]}}p) # 8-byte Spill
+; O0-NEXT: movq %rax, {{[-0-9]+}}(%r{{[sb]}}p) # 8-byte Spill
; O0-NEXT: .Lpcsection616:
; O0-NEXT: jne .LBB229_1
; O0-NEXT: jmp .LBB229_2
@@ -13945,8 +13965,8 @@ define void @atomic128_add_acq_rel(ptr %a) {
; O0-NEXT: jmp .LBB230_1
; O0-NEXT: .LBB230_1: # %atomicrmw.start
; O0-NEXT: # =>This Inner Loop Header: Depth=1
-; O0-NEXT: movq {{[-0-9]+}}(%r{{[sb]}}p), %rdx # 8-byte Reload
; O0-NEXT: movq {{[-0-9]+}}(%r{{[sb]}}p), %rax # 8-byte Reload
+; O0-NEXT: movq {{[-0-9]+}}(%r{{[sb]}}p), %rdx # 8-byte Reload
; O0-NEXT: movq {{[-0-9]+}}(%r{{[sb]}}p), %rsi # 8-byte Reload
; O0-NEXT: movq %rax, %rbx
; O0-NEXT: .Lpcsection620:
@@ -13956,8 +13976,8 @@ define void @atomic128_add_acq_rel(ptr %a) {
; O0-NEXT: adcq $0, %rcx
; O0-NEXT: .Lpcsection622:
; O0-NEXT: lock cmpxchg16b (%rsi)
-; O0-NEXT: movq %rax, {{[-0-9]+}}(%r{{[sb]}}p) # 8-byte Spill
; O0-NEXT: movq %rdx, {{[-0-9]+}}(%r{{[sb]}}p) # 8-byte Spill
+; O0-NEXT: movq %rax, {{[-0-9]+}}(%r{{[sb]}}p) # 8-byte Spill
; O0-NEXT: .Lpcsection623:
; O0-NEXT: jne .LBB230_1
; O0-NEXT: jmp .LBB230_2
@@ -14078,8 +14098,8 @@ define void @atomic128_sub_acq_rel(ptr %a) {
; O0-NEXT: jmp .LBB231_1
; O0-NEXT: .LBB231_1: # %atomicrmw.start
; O0-NEXT: # =>This Inner Loop Header: Depth=1
-; O0-NEXT: movq {{[-0-9]+}}(%r{{[sb]}}p), %rdx # 8-byte Reload
; O0-NEXT: movq {{[-0-9]+}}(%r{{[sb]}}p), %rax # 8-byte Reload
+; O0-NEXT: movq {{[-0-9]+}}(%r{{[sb]}}p), %rdx # 8-byte Reload
; O0-NEXT: movq {{[-0-9]+}}(%r{{[sb]}}p), %rsi # 8-byte Reload
; O0-NEXT: movq %rax, %rbx
; O0-NEXT: .Lpcsection627:
@@ -14089,8 +14109,8 @@ define void @atomic128_sub_acq_rel(ptr %a) {
; O0-NEXT: adcq $-1, %rcx
; O0-NEXT: .Lpcsection629:
; O0-NEXT: lock cmpxchg16b (%rsi)
-; O0-NEXT: movq %rax, {{[-0-9]+}}(%r{{[sb]}}p) # 8-byte Spill
; O0-NEXT: movq %rdx, {{[-0-9]+}}(%r{{[sb]}}p) # 8-byte Spill
+; O0-NEXT: movq %rax, {{[-0-9]+}}(%r{{[sb]}}p) # 8-byte Spill
; O0-NEXT: .Lpcsection630:
; O0-NEXT: jne .LBB231_1
; O0-NEXT: jmp .LBB231_2
@@ -14211,8 +14231,8 @@ define void @atomic128_and_acq_rel(ptr %a) {
; O0-NEXT: jmp .LBB232_1
; O0-NEXT: .LBB232_1: # %atomicrmw.start
; O0-NEXT: # =>This Inner Loop Header: Depth=1
-; O0-NEXT: movq {{[-0-9]+}}(%r{{[sb]}}p), %rdx # 8-byte Reload
; O0-NEXT: movq {{[-0-9]+}}(%r{{[sb]}}p), %rax # 8-byte Reload
+; O0-NEXT: movq {{[-0-9]+}}(%r{{[sb]}}p), %rdx # 8-byte Reload
; O0-NEXT: movq {{[-0-9]+}}(%r{{[sb]}}p), %rsi # 8-byte Reload
; O0-NEXT: movl %eax, %ecx
; O0-NEXT: .Lpcsection634:
@@ -14224,8 +14244,8 @@ define void @atomic128_and_acq_rel(ptr %a) {
; O0-NEXT: # kill: def $rcx killed $ecx
; O0-NEXT: .Lpcsection637:
; O0-NEXT: lock cmpxchg16b (%rsi)
-; O0-NEXT: movq %rax, {{[-0-9]+}}(%r{{[sb]}}p) # 8-byte Spill
; O0-NEXT: movq %rdx, {{[-0-9]+}}(%r{{[sb]}}p) # 8-byte Spill
+; O0-NEXT: movq %rax, {{[-0-9]+}}(%r{{[sb]}}p) # 8-byte Spill
; O0-NEXT: .Lpcsection638:
; O0-NEXT: jne .LBB232_1
; O0-NEXT: jmp .LBB232_2
@@ -14343,8 +14363,8 @@ define void @atomic128_or_acq_rel(ptr %a) {
; O0-NEXT: jmp .LBB233_1
; O0-NEXT: .LBB233_1: # %atomicrmw.start
; O0-NEXT: # =>This Inner Loop Header: Depth=1
-; O0-NEXT: movq {{[-0-9]+}}(%r{{[sb]}}p), %rcx # 8-byte Reload
; O0-NEXT: movq {{[-0-9]+}}(%r{{[sb]}}p), %rax # 8-byte Reload
+; O0-NEXT: movq {{[-0-9]+}}(%r{{[sb]}}p), %rcx # 8-byte Reload
; O0-NEXT: movq {{[-0-9]+}}(%r{{[sb]}}p), %rsi # 8-byte Reload
; O0-NEXT: movq %rax, %rbx
; O0-NEXT: .Lpcsection642:
@@ -14352,8 +14372,8 @@ define void @atomic128_or_acq_rel(ptr %a) {
; O0-NEXT: movq %rcx, %rdx
; O0-NEXT: .Lpcsection643:
; O0-NEXT: lock cmpxchg16b (%rsi)
-; O0-NEXT: movq %rax, {{[-0-9]+}}(%r{{[sb]}}p) # 8-byte Spill
; O0-NEXT: movq %rdx, {{[-0-9]+}}(%r{{[sb]}}p) # 8-byte Spill
+; O0-NEXT: movq %rax, {{[-0-9]+}}(%r{{[sb]}}p) # 8-byte Spill
; O0-NEXT: .Lpcsection644:
; O0-NEXT: jne .LBB233_1
; O0-NEXT: jmp .LBB233_2
@@ -14468,8 +14488,8 @@ define void @atomic128_xor_acq_rel(ptr %a) {
; O0-NEXT: jmp .LBB234_1
; O0-NEXT: .LBB234_1: # %atomicrmw.start
; O0-NEXT: # =>This Inner Loop Header: Depth=1
-; O0-NEXT: movq {{[-0-9]+}}(%r{{[sb]}}p), %rcx # 8-byte Reload
; O0-NEXT: movq {{[-0-9]+}}(%r{{[sb]}}p), %rax # 8-byte Reload
+; O0-NEXT: movq {{[-0-9]+}}(%r{{[sb]}}p), %rcx # 8-byte Reload
; O0-NEXT: movq {{[-0-9]+}}(%r{{[sb]}}p), %rsi # 8-byte Reload
; O0-NEXT: movq %rax, %rbx
; O0-NEXT: .Lpcsection648:
@@ -14477,8 +14497,8 @@ define void @atomic128_xor_acq_rel(ptr %a) {
; O0-NEXT: movq %rcx, %rdx
; O0-NEXT: .Lpcsection649:
; O0-NEXT: lock cmpxchg16b (%rsi)
-; O0-NEXT: movq %rax, {{[-0-9]+}}(%r{{[sb]}}p) # 8-byte Spill
; O0-NEXT: movq %rdx, {{[-0-9]+}}(%r{{[sb]}}p) # 8-byte Spill
+; O0-NEXT: movq %rax, {{[-0-9]+}}(%r{{[sb]}}p) # 8-byte Spill
; O0-NEXT: .Lpcsection650:
; O0-NEXT: jne .LBB234_1
; O0-NEXT: jmp .LBB234_2
@@ -14593,8 +14613,8 @@ define void @atomic128_nand_acq_rel(ptr %a) {
; O0-NEXT: jmp .LBB235_1
; O0-NEXT: .LBB235_1: # %atomicrmw.start
; O0-NEXT: # =>This Inner Loop Header: Depth=1
-; O0-NEXT: movq {{[-0-9]+}}(%r{{[sb]}}p), %rdx # 8-byte Reload
; O0-NEXT: movq {{[-0-9]+}}(%r{{[sb]}}p), %rax # 8-byte Reload
+; O0-NEXT: movq {{[-0-9]+}}(%r{{[sb]}}p), %rdx # 8-byte Reload
; O0-NEXT: movq {{[-0-9]+}}(%r{{[sb]}}p), %rsi # 8-byte Reload
; O0-NEXT: movl %eax, %ecx
; O0-NEXT: .Lpcsection654:
@@ -14608,8 +14628,8 @@ define void @atomic128_nand_acq_rel(ptr %a) {
; O0-NEXT: movq $-1, %rcx
; O0-NEXT: .Lpcsection658:
; O0-NEXT: lock cmpxchg16b (%rsi)
-; O0-NEXT: movq %rax, {{[-0-9]+}}(%r{{[sb]}}p) # 8-byte Spill
; O0-NEXT: movq %rdx, {{[-0-9]+}}(%r{{[sb]}}p) # 8-byte Spill
+; O0-NEXT: movq %rax, {{[-0-9]+}}(%r{{[sb]}}p) # 8-byte Spill
; O0-NEXT: .Lpcsection659:
; O0-NEXT: jne .LBB235_1
; O0-NEXT: jmp .LBB235_2
@@ -14733,8 +14753,8 @@ define void @atomic128_xchg_seq_cst(ptr %a) {
; O0-NEXT: jmp .LBB236_1
; O0-NEXT: .LBB236_1: # %atomicrmw.start
; O0-NEXT: # =>This Inner Loop Header: Depth=1
-; O0-NEXT: movq {{[-0-9]+}}(%r{{[sb]}}p), %rdx # 8-byte Reload
; O0-NEXT: movq {{[-0-9]+}}(%r{{[sb]}}p), %rax # 8-byte Reload
+; O0-NEXT: movq {{[-0-9]+}}(%r{{[sb]}}p), %rdx # 8-byte Reload
; O0-NEXT: movq {{[-0-9]+}}(%r{{[sb]}}p), %rsi # 8-byte Reload
; O0-NEXT: .Lpcsection663:
; O0-NEXT: xorl %ecx, %ecx
@@ -14744,8 +14764,8 @@ define void @atomic128_xchg_seq_cst(ptr %a) {
; O0-NEXT: movl $42, %ebx
; O0-NEXT: .Lpcsection666:
; O0-NEXT: lock cmpxchg16b (%rsi)
-; O0-NEXT: movq %rax, {{[-0-9]+}}(%r{{[sb]}}p) # 8-byte Spill
; O0-NEXT: movq %rdx, {{[-0-9]+}}(%r{{[sb]}}p) # 8-byte Spill
+; O0-NEXT: movq %rax, {{[-0-9]+}}(%r{{[sb]}}p) # 8-byte Spill
; O0-NEXT: .Lpcsection667:
; O0-NEXT: jne .LBB236_1
; O0-NEXT: jmp .LBB236_2
@@ -14860,8 +14880,8 @@ define void @atomic128_add_seq_cst(ptr %a) {
; O0-NEXT: jmp .LBB237_1
; O0-NEXT: .LBB237_1: # %atomicrmw.start
; O0-NEXT: # =>This Inner Loop Header: Depth=1
-; O0-NEXT: movq {{[-0-9]+}}(%r{{[sb]}}p), %rdx # 8-byte Reload
; O0-NEXT: movq {{[-0-9]+}}(%r{{[sb]}}p), %rax # 8-byte Reload
+; O0-NEXT: movq {{[-0-9]+}}(%r{{[sb]}}p), %rdx # 8-byte Reload
; O0-NEXT: movq {{[-0-9]+}}(%r{{[sb]}}p), %rsi # 8-byte Reload
; O0-NEXT: movq %rax, %rbx
; O0-NEXT: .Lpcsection671:
@@ -14871,8 +14891,8 @@ define void @atomic128_add_seq_cst(ptr %a) {
; O0-NEXT: adcq $0, %rcx
; O0-NEXT: .Lpcsection673:
; O0-NEXT: lock cmpxchg16b (%rsi)
-; O0-NEXT: movq %rax, {{[-0-9]+}}(%r{{[sb]}}p) # 8-byte Spill
; O0-NEXT: movq %rdx, {{[-0-9]+}}(%r{{[sb]}}p) # 8-byte Spill
+; O0-NEXT: movq %rax, {{[-0-9]+}}(%r{{[sb]}}p) # 8-byte Spill
; O0-NEXT: .Lpcsection674:
; O0-NEXT: jne .LBB237_1
; O0-NEXT: jmp .LBB237_2
@@ -14993,8 +15013,8 @@ define void @atomic128_sub_seq_cst(ptr %a) {
; O0-NEXT: jmp .LBB238_1
; O0-NEXT: .LBB238_1: # %atomicrmw.start
; O0-NEXT: # =>This Inner Loop Header: Depth=1
-; O0-NEXT: movq {{[-0-9]+}}(%r{{[sb]}}p), %rdx # 8-byte Reload
; O0-NEXT: movq {{[-0-9]+}}(%r{{[sb]}}p), %rax # 8-byte Reload
+; O0-NEXT: movq {{[-0-9]+}}(%r{{[sb]}}p), %rdx # 8-byte Reload
; O0-NEXT: movq {{[-0-9]+}}(%r{{[sb]}}p), %rsi # 8-byte Reload
; O0-NEXT: movq %rax, %rbx
; O0-NEXT: .Lpcsection678:
@@ -15004,8 +15024,8 @@ define void @atomic128_sub_seq_cst(ptr %a) {
; O0-NEXT: adcq $-1, %rcx
; O0-NEXT: .Lpcsection680:
; O0-NEXT: lock cmpxchg16b (%rsi)
-; O0-NEXT: movq %rax, {{[-0-9]+}}(%r{{[sb]}}p) # 8-byte Spill
; O0-NEXT: movq %rdx, {{[-0-9]+}}(%r{{[sb]}}p) # 8-byte Spill
+; O0-NEXT: movq %rax, {{[-0-9]+}}(%r{{[sb]}}p) # 8-byte Spill
; O0-NEXT: .Lpcsection681:
; O0-NEXT: jne .LBB238_1
; O0-NEXT: jmp .LBB238_2
@@ -15126,8 +15146,8 @@ define void @atomic128_and_seq_cst(ptr %a) {
; O0-NEXT: jmp .LBB239_1
; O0-NEXT: .LBB239_1: # %atomicrmw.start
; O0-NEXT: # =>This Inner Loop Header: Depth=1
-; O0-NEXT: movq {{[-0-9]+}}(%r{{[sb]}}p), %rdx # 8-byte Reload
; O0-NEXT: movq {{[-0-9]+}}(%r{{[sb]}}p), %rax # 8-byte Reload
+; O0-NEXT: movq {{[-0-9]+}}(%r{{[sb]}}p), %rdx # 8-byte Reload
; O0-NEXT: movq {{[-0-9]+}}(%r{{[sb]}}p), %rsi # 8-byte Reload
; O0-NEXT: movl %eax, %ecx
; O0-NEXT: .Lpcsection685:
@@ -15139,8 +15159,8 @@ define void @atomic128_and_seq_cst(ptr %a) {
; O0-NEXT: # kill: def $rcx killed $ecx
; O0-NEXT: .Lpcsection688:
; O0-NEXT: lock cmpxchg16b (%rsi)
-; O0-NEXT: movq %rax, {{[-0-9]+}}(%r{{[sb]}}p) # 8-byte Spill
; O0-NEXT: movq %rdx, {{[-0-9]+}}(%r{{[sb]}}p) # 8-byte Spill
+; O0-NEXT: movq %rax, {{[-0-9]+}}(%r{{[sb]}}p) # 8-byte Spill
; O0-NEXT: .Lpcsection689:
; O0-NEXT: jne .LBB239_1
; O0-NEXT: jmp .LBB239_2
@@ -15258,8 +15278,8 @@ define void @atomic128_or_seq_cst(ptr %a) {
; O0-NEXT: jmp .LBB240_1
; O0-NEXT: .LBB240_1: # %atomicrmw.start
; O0-NEXT: # =>This Inner Loop Header: Depth=1
-; O0-NEXT: movq {{[-0-9]+}}(%r{{[sb]}}p), %rcx # 8-byte Reload
; O0-NEXT: movq {{[-0-9]+}}(%r{{[sb]}}p), %rax # 8-byte Reload
+; O0-NEXT: movq {{[-0-9]+}}(%r{{[sb]}}p), %rcx # 8-byte Reload
; O0-NEXT: movq {{[-0-9]+}}(%r{{[sb]}}p), %rsi # 8-byte Reload
; O0-NEXT: movq %rax, %rbx
; O0-NEXT: .Lpcsection693:
@@ -15267,8 +15287,8 @@ define void @atomic128_or_seq_cst(ptr %a) {
; O0-NEXT: movq %rcx, %rdx
; O0-NEXT: .Lpcsection694:
; O0-NEXT: lock cmpxchg16b (%rsi)
-; O0-NEXT: movq %rax, {{[-0-9]+}}(%r{{[sb]}}p) # 8-byte Spill
; O0-NEXT: movq %rdx, {{[-0-9]+}}(%r{{[sb]}}p) # 8-byte Spill
+; O0-NEXT: movq %rax, {{[-0-9]+}}(%r{{[sb]}}p) # 8-byte Spill
; O0-NEXT: .Lpcsection695:
; O0-NEXT: jne .LBB240_1
; O0-NEXT: jmp .LBB240_2
@@ -15383,8 +15403,8 @@ define void @atomic128_xor_seq_cst(ptr %a) {
; O0-NEXT: jmp .LBB241_1
; O0-NEXT: .LBB241_1: # %atomicrmw.start
; O0-NEXT: # =>This Inner Loop Header: Depth=1
-; O0-NEXT: movq {{[-0-9]+}}(%r{{[sb]}}p), %rcx # 8-byte Reload
; O0-NEXT: movq {{[-0-9]+}}(%r{{[sb]}}p), %rax # 8-byte Reload
+; O0-NEXT: movq {{[-0-9]+}}(%r{{[sb]}}p), %rcx # 8-byte Reload
; O0-NEXT: movq {{[-0-9]+}}(%r{{[sb]}}p), %rsi # 8-byte Reload
; O0-NEXT: movq %rax, %rbx
; O0-NEXT: .Lpcsection699:
@@ -15392,8 +15412,8 @@ define void @atomic128_xor_seq_cst(ptr %a) {
; O0-NEXT: movq %rcx, %rdx
; O0-NEXT: .Lpcsection700:
; O0-NEXT: lock cmpxchg16b (%rsi)
-; O0-NEXT: movq %rax, {{[-0-9]+}}(%r{{[sb]}}p) # 8-byte Spill
; O0-NEXT: movq %rdx, {{[-0-9]+}}(%r{{[sb]}}p) # 8-byte Spill
+; O0-NEXT: movq %rax, {{[-0-9]+}}(%r{{[sb]}}p) # 8-byte Spill
; O0-NEXT: .Lpcsection701:
; O0-NEXT: jne .LBB241_1
; O0-NEXT: jmp .LBB241_2
@@ -15508,8 +15528,8 @@ define void @atomic128_nand_seq_cst(ptr %a) {
; O0-NEXT: jmp .LBB242_1
; O0-NEXT: .LBB242_1: # %atomicrmw.start
; O0-NEXT: # =>This Inner Loop Header: Depth=1
-; O0-NEXT: movq {{[-0-9]+}}(%r{{[sb]}}p), %rdx # 8-byte Reload
; O0-NEXT: movq {{[-0-9]+}}(%r{{[sb]}}p), %rax # 8-byte Reload
+; O0-NEXT: movq {{[-0-9]+}}(%r{{[sb]}}p), %rdx # 8-byte Reload
; O0-NEXT: movq {{[-0-9]+}}(%r{{[sb]}}p), %rsi # 8-byte Reload
; O0-NEXT: movl %eax, %ecx
; O0-NEXT: .Lpcsection705:
@@ -15523,8 +15543,8 @@ define void @atomic128_nand_seq_cst(ptr %a) {
; O0-NEXT: movq $-1, %rcx
; O0-NEXT: .Lpcsection709:
; O0-NEXT: lock cmpxchg16b (%rsi)
-; O0-NEXT: movq %rax, {{[-0-9]+}}(%r{{[sb]}}p) # 8-byte Spill
; O0-NEXT: movq %rdx, {{[-0-9]+}}(%r{{[sb]}}p) # 8-byte Spill
+; O0-NEXT: movq %rax, {{[-0-9]+}}(%r{{[sb]}}p) # 8-byte Spill
; O0-NEXT: .Lpcsection710:
; O0-NEXT: jne .LBB242_1
; O0-NEXT: jmp .LBB242_2
diff --git a/llvm/test/CodeGen/X86/pr15705.ll b/llvm/test/CodeGen/X86/pr15705.ll
index 3dd4aabe25b57..bbbeb53764686 100644
--- a/llvm/test/CodeGen/X86/pr15705.ll
+++ b/llvm/test/CodeGen/X86/pr15705.ll
@@ -5,18 +5,19 @@
define i32 @PR15705(i32 %x, i32 %a, i32 %b, i32 %c) #0 {
; X86-LABEL: PR15705:
; X86: # %bb.0: # %entry
-; X86-NEXT: movl {{[0-9]+}}(%esp), %eax
; X86-NEXT: movl {{[0-9]+}}(%esp), %ecx
+; X86-NEXT: movl {{[0-9]+}}(%esp), %eax
; X86-NEXT: movl {{[0-9]+}}(%esp), %edx
-; X86-NEXT: cmpl %ecx, %edx
-; X86-NEXT: je .LBB0_4
-; X86-NEXT: # %bb.1: # %if.end
; X86-NEXT: cmpl %eax, %edx
-; X86-NEXT: jne .LBB0_3
-; X86-NEXT: # %bb.2:
-; X86-NEXT: movl {{[0-9]+}}(%esp), %ecx
-; X86-NEXT: .LBB0_3: # %if.end
+; X86-NEXT: jne .LBB0_2
+; X86-NEXT: # %bb.1:
; X86-NEXT: movl %ecx, %eax
+; X86-NEXT: retl
+; X86-NEXT: .LBB0_2: # %if.end
+; X86-NEXT: cmpl %ecx, %edx
+; X86-NEXT: jne .LBB0_4
+; X86-NEXT: # %bb.3:
+; X86-NEXT: movl {{[0-9]+}}(%esp), %eax
; X86-NEXT: .LBB0_4: # %return
; X86-NEXT: retl
;
diff --git a/llvm/test/CodeGen/X86/pr32256.ll b/llvm/test/CodeGen/X86/pr32256.ll
index 225a3af551a2c..09f7d92c1db2e 100644
--- a/llvm/test/CodeGen/X86/pr32256.ll
+++ b/llvm/test/CodeGen/X86/pr32256.ll
@@ -9,12 +9,12 @@ define void @_Z1av() {
; CHECK: # %bb.0: # %entry
; CHECK-NEXT: subl $2, %esp
; CHECK-NEXT: .cfi_def_cfa_offset 6
-; CHECK-NEXT: movb c, %cl
-; CHECK-NEXT: xorb $-1, %cl
-; CHECK-NEXT: xorl %eax, %eax
-; CHECK-NEXT: # kill: def $al killed $al killed $eax
-; CHECK-NEXT: testb $1, %cl
-; CHECK-NEXT: movb %al, (%esp) # 1-byte Spill
+; CHECK-NEXT: movb c, %al
+; CHECK-NEXT: xorb $-1, %al
+; CHECK-NEXT: xorl %ecx, %ecx
+; CHECK-NEXT: # kill: def $cl killed $cl killed $ecx
+; CHECK-NEXT: movb %cl, (%esp) # 1-byte Spill
+; CHECK-NEXT: testb $1, %al
; CHECK-NEXT: jne .LBB0_1
; CHECK-NEXT: jmp .LBB0_2
; CHECK-NEXT: .LBB0_1: # %land.rhs
diff --git a/llvm/test/CodeGen/X86/pr38795.ll b/llvm/test/CodeGen/X86/pr38795.ll
index c3c96e8228797..02dbe4f545fe5 100644
--- a/llvm/test/CodeGen/X86/pr38795.ll
+++ b/llvm/test/CodeGen/X86/pr38795.ll
@@ -33,6 +33,7 @@ define dso_local void @fn() {
; CHECK-NEXT: .p2align 4
; CHECK-NEXT: .LBB0_15: # %for.inc
; CHECK-NEXT: # in Loop: Header=BB0_1 Depth=1
+; CHECK-NEXT: movl %esi, %ecx
; CHECK-NEXT: movb %dl, {{[-0-9]+}}(%e{{[sb]}}p) # 1-byte Spill
; CHECK-NEXT: .LBB0_1: # %for.cond
; CHECK-NEXT: # =>This Loop Header: Depth=1
@@ -69,7 +70,6 @@ define dso_local void @fn() {
; CHECK-NEXT: movb {{[-0-9]+}}(%e{{[sb]}}p), %dh # 1-byte Reload
; CHECK-NEXT: movb {{[-0-9]+}}(%e{{[sb]}}p), %dl # 1-byte Reload
; CHECK-NEXT: testb %bl, %bl
-; CHECK-NEXT: movl %esi, %ecx
; CHECK-NEXT: # implicit-def: $eax
; CHECK-NEXT: movb %dh, {{[-0-9]+}}(%e{{[sb]}}p) # 1-byte Spill
; CHECK-NEXT: jne .LBB0_15
@@ -119,7 +119,7 @@ define dso_local void @fn() {
; CHECK-NEXT: jne .LBB0_9
; CHECK-NEXT: # %bb.12: # %if.end26
; CHECK-NEXT: # in Loop: Header=BB0_1 Depth=1
-; CHECK-NEXT: xorl %ecx, %ecx
+; CHECK-NEXT: xorl %esi, %esi
; CHECK-NEXT: testb %dh, %dh
; CHECK-NEXT: je .LBB0_15
; CHECK-NEXT: # %bb.13: # %if.end26
@@ -128,7 +128,7 @@ define dso_local void @fn() {
; CHECK-NEXT: jne .LBB0_15
; CHECK-NEXT: # %bb.14: # %if.then31
; CHECK-NEXT: # in Loop: Header=BB0_1 Depth=1
-; CHECK-NEXT: xorl %ecx, %ecx
+; CHECK-NEXT: xorl %esi, %esi
; CHECK-NEXT: xorl %ebp, %ebp
; CHECK-NEXT: jmp .LBB0_15
; CHECK-NEXT: .p2align 4
@@ -279,31 +279,34 @@ define void @verifier_error_reduced_issue38788(i1 %cmp11) {
; CHECK-NEXT: je .LBB1_3
; CHECK-NEXT: # %bb.2: # in Loop: Header=BB1_1 Depth=1
; CHECK-NEXT: xorl %eax, %eax
+; CHECK-NEXT: movl %ebx, %edx
; CHECK-NEXT: jmp .LBB1_5
; CHECK-NEXT: .p2align 4
; CHECK-NEXT: .LBB1_3: # %if.end
; CHECK-NEXT: # in Loop: Header=BB1_1 Depth=1
+; CHECK-NEXT: xorl %edx, %edx
; CHECK-NEXT: testb $1, %al
; CHECK-NEXT: je .LBB1_4
; CHECK-NEXT: # %bb.9: # %if.then13
; CHECK-NEXT: # in Loop: Header=BB1_1 Depth=1
-; CHECK-NEXT: xorl %edx, %edx
; CHECK-NEXT: testb $1, %al
; CHECK-NEXT: movl %ebx, %eax
; CHECK-NEXT: movl $0, %ebx
; CHECK-NEXT: jne .LBB1_8
+; CHECK-NEXT: # %bb.10: # %for.cond35
+; CHECK-NEXT: # in Loop: Header=BB1_1 Depth=1
+; CHECK-NEXT: movl %ebx, %edx
; CHECK-NEXT: jmp .LBB1_5
; CHECK-NEXT: .p2align 4
; CHECK-NEXT: .LBB1_4: # in Loop: Header=BB1_1 Depth=1
; CHECK-NEXT: movl %ebx, %eax
-; CHECK-NEXT: xorl %ebx, %ebx
; CHECK-NEXT: .LBB1_5: # %if.end26
; CHECK-NEXT: # in Loop: Header=BB1_1 Depth=1
; CHECK-NEXT: testb %cl, %cl
; CHECK-NEXT: je .LBB1_7
; CHECK-NEXT: # %bb.6: # %if.end26
; CHECK-NEXT: # in Loop: Header=BB1_1 Depth=1
-; CHECK-NEXT: movl %ebx, %ecx
+; CHECK-NEXT: movl %edx, %ecx
; CHECK-NEXT: jmp .LBB1_7
entry:
br label %for.cond
diff --git a/llvm/test/CodeGen/X86/pr49451.ll b/llvm/test/CodeGen/X86/pr49451.ll
index 173c41140ebef..1a7551f6117e8 100644
--- a/llvm/test/CodeGen/X86/pr49451.ll
+++ b/llvm/test/CodeGen/X86/pr49451.ll
@@ -18,15 +18,15 @@ define void @func_6(i8 %uc_8, i64 %uli_10) nounwind {
; X86-NEXT: .p2align 4
; X86-NEXT: .LBB0_1: # %for.body612
; X86-NEXT: # =>This Inner Loop Header: Depth=1
-; X86-NEXT: testb %dl, %dl
+; X86-NEXT: testb %bl, %bl
; X86-NEXT: je .LBB0_2
; X86-NEXT: # %bb.3: # %if.end1401
; X86-NEXT: # in Loop: Header=BB0_1 Depth=1
; X86-NEXT: addl %eax, %esi
; X86-NEXT: movw %si, s_2
-; X86-NEXT: movw %bx, s_0
+; X86-NEXT: movw %dx, s_0
; X86-NEXT: incl %ecx
-; X86-NEXT: incl %ebx
+; X86-NEXT: incl %edx
; X86-NEXT: cmpw $73, %cx
; X86-NEXT: jl .LBB0_1
; X86-NEXT: # %bb.4: # %for.body1703
diff --git a/llvm/test/CodeGen/X86/pr63108.ll b/llvm/test/CodeGen/X86/pr63108.ll
index b5b80515fc6d9..8df90a935314d 100644
--- a/llvm/test/CodeGen/X86/pr63108.ll
+++ b/llvm/test/CodeGen/X86/pr63108.ll
@@ -21,7 +21,7 @@ define i32 @PR63108() {
; SSE-NEXT: # =>This Inner Loop Header: Depth=1
; SSE-NEXT: movdqa %xmm1, %xmm2
; SSE-NEXT: testb %al, %al
-; SSE-NEXT: pxor %xmm1, %xmm1
+; SSE-NEXT: movdqa %xmm0, %xmm1
; SSE-NEXT: jne .LBB0_3
; SSE-NEXT: # %bb.4: # %middle.block
; SSE-NEXT: pxor %xmm2, %xmm0
diff --git a/llvm/test/CodeGen/X86/sad.ll b/llvm/test/CodeGen/X86/sad.ll
index fe71a16039c19..7cfc88a77dea5 100644
--- a/llvm/test/CodeGen/X86/sad.ll
+++ b/llvm/test/CodeGen/X86/sad.ll
@@ -151,27 +151,27 @@ define dso_local i32 @sad_32i8() nounwind {
; SSE2-LABEL: sad_32i8:
; SSE2: # %bb.0: # %entry
; SSE2-NEXT: pxor %xmm0, %xmm0
-; SSE2-NEXT: movq $-1024, %rax # imm = 0xFC00
; SSE2-NEXT: pxor %xmm1, %xmm1
+; SSE2-NEXT: movq $-1024, %rax # imm = 0xFC00
; SSE2-NEXT: pxor %xmm2, %xmm2
; SSE2-NEXT: .p2align 4
; SSE2-NEXT: .LBB1_1: # %vector.body
; SSE2-NEXT: # =>This Inner Loop Header: Depth=1
; SSE2-NEXT: movdqa a+1024(%rax), %xmm3
; SSE2-NEXT: psadbw b+1024(%rax), %xmm3
-; SSE2-NEXT: paddd %xmm3, %xmm1
+; SSE2-NEXT: paddd %xmm3, %xmm2
; SSE2-NEXT: movdqa a+1040(%rax), %xmm3
; SSE2-NEXT: psadbw b+1040(%rax), %xmm3
-; SSE2-NEXT: paddd %xmm3, %xmm2
+; SSE2-NEXT: paddd %xmm3, %xmm1
; SSE2-NEXT: addq $32, %rax
; SSE2-NEXT: jne .LBB1_1
; SSE2-NEXT: # %bb.2: # %middle.block
-; SSE2-NEXT: paddd %xmm0, %xmm2
; SSE2-NEXT: paddd %xmm0, %xmm1
+; SSE2-NEXT: paddd %xmm0, %xmm2
; SSE2-NEXT: paddd %xmm0, %xmm0
-; SSE2-NEXT: paddd %xmm0, %xmm1
-; SSE2-NEXT: paddd %xmm2, %xmm0
+; SSE2-NEXT: paddd %xmm0, %xmm2
; SSE2-NEXT: paddd %xmm1, %xmm0
+; SSE2-NEXT: paddd %xmm2, %xmm0
; SSE2-NEXT: pshufd {{.*#+}} xmm1 = xmm0[2,3,2,3]
; SSE2-NEXT: paddd %xmm0, %xmm1
; SSE2-NEXT: pshufd {{.*#+}} xmm0 = xmm1[1,1,1,1]
@@ -356,9 +356,9 @@ define dso_local i32 @sad_avx64i8() nounwind {
; AVX1-LABEL: sad_avx64i8:
; AVX1: # %bb.0: # %entry
; AVX1-NEXT: vpxor %xmm0, %xmm0, %xmm0
+; AVX1-NEXT: vpxor %xmm1, %xmm1, %xmm1
; AVX1-NEXT: movq $-1024, %rax # imm = 0xFC00
; AVX1-NEXT: vpxor %xmm2, %xmm2, %xmm2
-; AVX1-NEXT: vpxor %xmm1, %xmm1, %xmm1
; AVX1-NEXT: .p2align 4
; AVX1-NEXT: .LBB2_1: # %vector.body
; AVX1-NEXT: # =>This Inner Loop Header: Depth=1
@@ -408,26 +408,26 @@ define dso_local i32 @sad_avx64i8() nounwind {
; AVX2-LABEL: sad_avx64i8:
; AVX2: # %bb.0: # %entry
; AVX2-NEXT: vpxor %xmm0, %xmm0, %xmm0
-; AVX2-NEXT: movq $-1024, %rax # imm = 0xFC00
; AVX2-NEXT: vpxor %xmm1, %xmm1, %xmm1
+; AVX2-NEXT: movq $-1024, %rax # imm = 0xFC00
; AVX2-NEXT: vpxor %xmm2, %xmm2, %xmm2
; AVX2-NEXT: .p2align 4
; AVX2-NEXT: .LBB2_1: # %vector.body
; AVX2-NEXT: # =>This Inner Loop Header: Depth=1
; AVX2-NEXT: vmovdqa a+1024(%rax), %ymm3
; AVX2-NEXT: vpsadbw b+1024(%rax), %ymm3, %ymm3
-; AVX2-NEXT: vpaddd %ymm1, %ymm3, %ymm1
+; AVX2-NEXT: vpaddd %ymm2, %ymm3, %ymm2
; AVX2-NEXT: vmovdqa a+1056(%rax), %ymm3
; AVX2-NEXT: vpsadbw b+1056(%rax), %ymm3, %ymm3
-; AVX2-NEXT: vpaddd %ymm2, %ymm3, %ymm2
+; AVX2-NEXT: vpaddd %ymm1, %ymm3, %ymm1
; AVX2-NEXT: addq $64, %rax
; AVX2-NEXT: jne .LBB2_1
; AVX2-NEXT: # %bb.2: # %middle.block
-; AVX2-NEXT: vpaddd %ymm0, %ymm2, %ymm2
+; AVX2-NEXT: vpaddd %ymm0, %ymm1, %ymm1
; AVX2-NEXT: vpaddd %ymm0, %ymm0, %ymm3
-; AVX2-NEXT: vpaddd %ymm0, %ymm1, %ymm0
+; AVX2-NEXT: vpaddd %ymm0, %ymm2, %ymm0
; AVX2-NEXT: vpaddd %ymm3, %ymm0, %ymm0
-; AVX2-NEXT: vpaddd %ymm3, %ymm2, %ymm1
+; AVX2-NEXT: vpaddd %ymm3, %ymm1, %ymm1
; AVX2-NEXT: vpaddd %ymm1, %ymm0, %ymm0
; AVX2-NEXT: vextracti128 $1, %ymm0, %xmm1
; AVX2-NEXT: vpaddd %xmm1, %xmm0, %xmm0
diff --git a/llvm/test/CodeGen/X86/sse-scalar-fp-arith.ll b/llvm/test/CodeGen/X86/sse-scalar-fp-arith.ll
index 325f735b09cd9..b4a6960d144e1 100644
--- a/llvm/test/CodeGen/X86/sse-scalar-fp-arith.ll
+++ b/llvm/test/CodeGen/X86/sse-scalar-fp-arith.ll
@@ -1266,13 +1266,12 @@ define <4 x float> @add_ss_mask(<4 x float> %a, <4 x float> %b, <4 x float> %c,
; X86-SSE-LABEL: add_ss_mask:
; X86-SSE: # %bb.0:
; X86-SSE-NEXT: testb $1, {{[0-9]+}}(%esp)
-; X86-SSE-NEXT: jne .LBB70_1
-; X86-SSE-NEXT: # %bb.2:
-; X86-SSE-NEXT: movss {{.*#+}} xmm0 = xmm2[0],xmm0[1,2,3]
-; X86-SSE-NEXT: retl
-; X86-SSE-NEXT: .LBB70_1:
+; X86-SSE-NEXT: je .LBB70_2
+; X86-SSE-NEXT: # %bb.1:
; X86-SSE-NEXT: addss %xmm0, %xmm1
-; X86-SSE-NEXT: movss {{.*#+}} xmm0 = xmm1[0],xmm0[1,2,3]
+; X86-SSE-NEXT: movaps %xmm1, %xmm2
+; X86-SSE-NEXT: .LBB70_2:
+; X86-SSE-NEXT: movss {{.*#+}} xmm0 = xmm2[0],xmm0[1,2,3]
; X86-SSE-NEXT: retl
;
; X86-AVX1-LABEL: add_ss_mask:
@@ -1296,13 +1295,12 @@ define <4 x float> @add_ss_mask(<4 x float> %a, <4 x float> %b, <4 x float> %c,
; X64-SSE-LABEL: add_ss_mask:
; X64-SSE: # %bb.0:
; X64-SSE-NEXT: testb $1, %dil
-; X64-SSE-NEXT: jne .LBB70_1
-; X64-SSE-NEXT: # %bb.2:
-; X64-SSE-NEXT: movss {{.*#+}} xmm0 = xmm2[0],xmm0[1,2,3]
-; X64-SSE-NEXT: retq
-; X64-SSE-NEXT: .LBB70_1:
+; X64-SSE-NEXT: je .LBB70_2
+; X64-SSE-NEXT: # %bb.1:
; X64-SSE-NEXT: addss %xmm0, %xmm1
-; X64-SSE-NEXT: movss {{.*#+}} xmm0 = xmm1[0],xmm0[1,2,3]
+; X64-SSE-NEXT: movaps %xmm1, %xmm2
+; X64-SSE-NEXT: .LBB70_2:
+; X64-SSE-NEXT: movss {{.*#+}} xmm0 = xmm2[0],xmm0[1,2,3]
; X64-SSE-NEXT: retq
;
; X64-AVX1-LABEL: add_ss_mask:
@@ -1336,13 +1334,12 @@ define <2 x double> @add_sd_mask(<2 x double> %a, <2 x double> %b, <2 x double>
; X86-SSE-LABEL: add_sd_mask:
; X86-SSE: # %bb.0:
; X86-SSE-NEXT: testb $1, {{[0-9]+}}(%esp)
-; X86-SSE-NEXT: jne .LBB71_1
-; X86-SSE-NEXT: # %bb.2:
-; X86-SSE-NEXT: movsd {{.*#+}} xmm0 = xmm2[0],xmm0[1]
-; X86-SSE-NEXT: retl
-; X86-SSE-NEXT: .LBB71_1:
+; X86-SSE-NEXT: je .LBB71_2
+; X86-SSE-NEXT: # %bb.1:
; X86-SSE-NEXT: addsd %xmm0, %xmm1
-; X86-SSE-NEXT: movsd {{.*#+}} xmm0 = xmm1[0],xmm0[1]
+; X86-SSE-NEXT: movapd %xmm1, %xmm2
+; X86-SSE-NEXT: .LBB71_2:
+; X86-SSE-NEXT: movsd {{.*#+}} xmm0 = xmm2[0],xmm0[1]
; X86-SSE-NEXT: retl
;
; X86-AVX1-LABEL: add_sd_mask:
@@ -1366,13 +1363,12 @@ define <2 x double> @add_sd_mask(<2 x double> %a, <2 x double> %b, <2 x double>
; X64-SSE-LABEL: add_sd_mask:
; X64-SSE: # %bb.0:
; X64-SSE-NEXT: testb $1, %dil
-; X64-SSE-NEXT: jne .LBB71_1
-; X64-SSE-NEXT: # %bb.2:
-; X64-SSE-NEXT: movsd {{.*#+}} xmm0 = xmm2[0],xmm0[1]
-; X64-SSE-NEXT: retq
-; X64-SSE-NEXT: .LBB71_1:
+; X64-SSE-NEXT: je .LBB71_2
+; X64-SSE-NEXT: # %bb.1:
; X64-SSE-NEXT: addsd %xmm0, %xmm1
-; X64-SSE-NEXT: movsd {{.*#+}} xmm0 = xmm1[0],xmm0[1]
+; X64-SSE-NEXT: movapd %xmm1, %xmm2
+; X64-SSE-NEXT: .LBB71_2:
+; X64-SSE-NEXT: movsd {{.*#+}} xmm0 = xmm2[0],xmm0[1]
; X64-SSE-NEXT: retq
;
; X64-AVX1-LABEL: add_sd_mask:
diff --git a/llvm/test/CodeGen/X86/swifterror.ll b/llvm/test/CodeGen/X86/swifterror.ll
index 5699c447baf41..473cc9d152627 100644
--- a/llvm/test/CodeGen/X86/swifterror.ll
+++ b/llvm/test/CodeGen/X86/swifterror.ll
@@ -426,18 +426,19 @@ define float @foo_loop(ptr swifterror %error_ptr_ref, i32 %cc, float %cc2) {
; CHECK-O0-NEXT: jmp LBB4_1
; CHECK-O0-NEXT: LBB4_1: ## %bb_loop
; CHECK-O0-NEXT: ## =>This Inner Loop Header: Depth=1
-; CHECK-O0-NEXT: movl {{[-0-9]+}}(%r{{[sb]}}p), %ecx ## 4-byte Reload
-; CHECK-O0-NEXT: movq {{[-0-9]+}}(%r{{[sb]}}p), %rax ## 8-byte Reload
-; CHECK-O0-NEXT: cmpl $0, %ecx
-; CHECK-O0-NEXT: movq %rax, {{[-0-9]+}}(%r{{[sb]}}p) ## 8-byte Spill
+; CHECK-O0-NEXT: movl {{[-0-9]+}}(%r{{[sb]}}p), %eax ## 4-byte Reload
+; CHECK-O0-NEXT: movq {{[-0-9]+}}(%r{{[sb]}}p), %rcx ## 8-byte Reload
+; CHECK-O0-NEXT: movq %rcx, {{[-0-9]+}}(%r{{[sb]}}p) ## 8-byte Spill
+; CHECK-O0-NEXT: cmpl $0, %eax
; CHECK-O0-NEXT: je LBB4_3
; CHECK-O0-NEXT: ## %bb.2: ## %gen_error
; CHECK-O0-NEXT: ## in Loop: Header=BB4_1 Depth=1
; CHECK-O0-NEXT: movl $16, %edi
; CHECK-O0-NEXT: callq _malloc
; CHECK-O0-NEXT: movq %rax, %rcx
-; CHECK-O0-NEXT: movb $1, 8(%rcx)
-; CHECK-O0-NEXT: movq %rax, {{[-0-9]+}}(%r{{[sb]}}p) ## 8-byte Spill
+; CHECK-O0-NEXT: movq %rcx, %rax
+; CHECK-O0-NEXT: movq %rcx, {{[-0-9]+}}(%r{{[sb]}}p) ## 8-byte Spill
+; CHECK-O0-NEXT: movb $1, 8(%rax)
; CHECK-O0-NEXT: LBB4_3: ## %bb_cont
; CHECK-O0-NEXT: ## in Loop: Header=BB4_1 Depth=1
; CHECK-O0-NEXT: movss {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 ## 4-byte Reload
@@ -964,17 +965,17 @@ define void @swifterror_isel(ptr) {
; CHECK-O0-NEXT: ## implicit-def: $r12
; CHECK-O0-NEXT: jne LBB8_2
; CHECK-O0-NEXT: LBB8_1: ## =>This Inner Loop Header: Depth=1
-; CHECK-O0-NEXT: movq {{[-0-9]+}}(%r{{[sb]}}p), %r12 ## 8-byte Reload
; CHECK-O0-NEXT: movw {{[-0-9]+}}(%r{{[sb]}}p), %ax ## 2-byte Reload
+; CHECK-O0-NEXT: movq {{[-0-9]+}}(%r{{[sb]}}p), %r12 ## 8-byte Reload
; CHECK-O0-NEXT: movq {{[-0-9]+}}(%r{{[sb]}}p), %r13 ## 8-byte Reload
; CHECK-O0-NEXT: ## implicit-def: $edi
; CHECK-O0-NEXT: movw %ax, %di
; CHECK-O0-NEXT: ## implicit-def: $rax
; CHECK-O0-NEXT: callq *%rax
+; CHECK-O0-NEXT: movq %r12, {{[-0-9]+}}(%r{{[sb]}}p) ## 8-byte Spill
; CHECK-O0-NEXT: ## implicit-def: $rax
; CHECK-O0-NEXT: movw (%rax), %ax
; CHECK-O0-NEXT: movw %ax, {{[-0-9]+}}(%r{{[sb]}}p) ## 2-byte Spill
-; CHECK-O0-NEXT: movq %r12, {{[-0-9]+}}(%r{{[sb]}}p) ## 8-byte Spill
; CHECK-O0-NEXT: jmp LBB8_1
; CHECK-O0-NEXT: LBB8_2:
; CHECK-O0-NEXT: addq $40, %rsp
diff --git a/llvm/test/Transforms/LoopStrengthReduce/RISCV/lsr-drop-solution.ll b/llvm/test/Transforms/LoopStrengthReduce/RISCV/lsr-drop-solution.ll
index 7353acd7228cd..1f7a5b12d16b8 100644
--- a/llvm/test/Transforms/LoopStrengthReduce/RISCV/lsr-drop-solution.ll
+++ b/llvm/test/Transforms/LoopStrengthReduce/RISCV/lsr-drop-solution.ll
@@ -7,27 +7,23 @@ target triple = "riscv64-unknown-linux-gnu"
define ptr @foo(ptr %a0, ptr %a1, i64 %a2) {
; CHECK-LABEL: foo:
; CHECK: # %bb.0: # %entry
+; CHECK-NEXT: mv a3, a0
; CHECK-NEXT: vsetvli a4, a2, e8, m8, ta, ma
-; CHECK-NEXT: bne a4, a2, .LBB0_2
-; CHECK-NEXT: # %bb.1:
-; CHECK-NEXT: vsetvli zero, a2, e8, m8, ta, ma
-; CHECK-NEXT: vle8.v v8, (a1)
-; CHECK-NEXT: vse8.v v8, (a0)
-; CHECK-NEXT: ret
-; CHECK-NEXT: .LBB0_2: # %if.then
+; CHECK-NEXT: beq a4, a2, .LBB0_4
+; CHECK-NEXT: # %bb.1: # %if.then
; CHECK-NEXT: add a2, a0, a2
; CHECK-NEXT: sub a5, a2, a4
-; CHECK-NEXT: mv a3, a0
-; CHECK-NEXT: .LBB0_3: # %do.body
+; CHECK-NEXT: .LBB0_2: # %do.body
; CHECK-NEXT: # =>This Inner Loop Header: Depth=1
; CHECK-NEXT: vle8.v v8, (a1)
; CHECK-NEXT: vse8.v v8, (a3)
; CHECK-NEXT: add a3, a3, a4
; CHECK-NEXT: add a1, a1, a4
-; CHECK-NEXT: bltu a3, a5, .LBB0_3
-; CHECK-NEXT: # %bb.4: # %do.end
+; CHECK-NEXT: bltu a3, a5, .LBB0_2
+; CHECK-NEXT: # %bb.3: # %do.end
; CHECK-NEXT: sub a2, a2, a3
; CHECK-NEXT: vsetvli a2, a2, e8, m8, ta, ma
+; CHECK-NEXT: .LBB0_4: # %if.end
; CHECK-NEXT: vsetvli zero, a2, e8, m8, ta, ma
; CHECK-NEXT: vle8.v v8, (a1)
; CHECK-NEXT: vse8.v v8, (a3)
>From 8cb310722cc63da6e71862c564f4b22f4defbdfc Mon Sep 17 00:00:00 2001
From: Guy David <guyda96 at gmail.com>
Date: Tue, 18 Mar 2025 13:32:38 +0200
Subject: [PATCH 3/3] Manually fixed tests
---
.../CodeGen/Hexagon/swp-conv3x3-nested.ll | 3 +-
llvm/test/CodeGen/Hexagon/swp-epilog-phi7.ll | 4 +--
.../test/CodeGen/Hexagon/swp-matmul-bitext.ll | 2 +-
llvm/test/CodeGen/Hexagon/swp-stages4.ll | 7 ++---
llvm/test/CodeGen/Hexagon/tinycore.ll | 11 +++++--
.../CodeGen/PowerPC/2013-07-01-PHIElimBug.mir | 3 +-
llvm/test/CodeGen/PowerPC/phi-eliminate.mir | 9 ++----
llvm/test/CodeGen/PowerPC/ppcf128-freeze.mir | 30 +++++++++----------
llvm/test/CodeGen/PowerPC/pr116071.ll | 25 +++++++++++-----
llvm/test/CodeGen/SystemZ/swifterror.ll | 4 +--
.../statepoint-cmp-sunk-past-statepoint.ll | 2 +-
.../MIR/InstrRef/phi-regallocd-to-stack.mir | 7 ++---
12 files changed, 57 insertions(+), 50 deletions(-)
diff --git a/llvm/test/CodeGen/Hexagon/swp-conv3x3-nested.ll b/llvm/test/CodeGen/Hexagon/swp-conv3x3-nested.ll
index 006a8b6bfc94a..4d75a733526b0 100644
--- a/llvm/test/CodeGen/Hexagon/swp-conv3x3-nested.ll
+++ b/llvm/test/CodeGen/Hexagon/swp-conv3x3-nested.ll
@@ -1,7 +1,7 @@
; RUN: llc -mtriple=hexagon < %s -pipeliner-experimental-cg=true | FileCheck %s
; This version of the conv3x3 test has both loops. This test checks that the
-; inner loop has 14 packets.
+; inner loop has 13 packets.
; CHECK: loop0(.LBB0_[[LOOP:.]],
; CHECK: .LBB0_[[LOOP]]:
@@ -17,7 +17,6 @@
; CHECK: }
; CHECK: }
; CHECK: }
-; CHECK: }
; CHECK-NOT: }
; CHECK: }{{[ \t]*}}:endloop0
diff --git a/llvm/test/CodeGen/Hexagon/swp-epilog-phi7.ll b/llvm/test/CodeGen/Hexagon/swp-epilog-phi7.ll
index 96a38939dc50e..b50290525002d 100644
--- a/llvm/test/CodeGen/Hexagon/swp-epilog-phi7.ll
+++ b/llvm/test/CodeGen/Hexagon/swp-epilog-phi7.ll
@@ -12,8 +12,8 @@
; CHECK: [[EPLOG]]:
; CHECK: [[VREG1:v([0-9]+)]] = [[VREG]]
; CHECK: [[EPLOG1]]:
-; CHECK: [[VREG2:v[0-9]+]] = [[VREG1]]
-; CHECK: = vlalign([[VREG1]],[[VREG2]],#1)
+; CHECK: [[VREG2:v[0-9]+]] = [[VREG]]
+; CHECK: = vlalign([[VREG2]],[[VREG1]],#1)
; Function Attrs: nounwind
define void @f0(ptr noalias nocapture readonly %a0, i32 %a1, i32 %a2, ptr noalias nocapture readonly %a3, i32 %a4, ptr noalias nocapture %a5, i32 %a6) #0 {
diff --git a/llvm/test/CodeGen/Hexagon/swp-matmul-bitext.ll b/llvm/test/CodeGen/Hexagon/swp-matmul-bitext.ll
index 42efe60b96d48..c4dbbcc5969ca 100644
--- a/llvm/test/CodeGen/Hexagon/swp-matmul-bitext.ll
+++ b/llvm/test/CodeGen/Hexagon/swp-matmul-bitext.ll
@@ -3,7 +3,7 @@
; From coremark. Test that we pipeline the matrix multiplication bitextract
; function. The pipelined code should have two packets.
-; CHECK: loop0(.LBB0_[[LOOP:.]],
+; CHECK: loop0(.LBB0_[[LOOP:[0-9]+]],
; CHECK: .LBB0_[[LOOP]]:
; CHECK: [[REG0:(r[0-9]+)]] = mpyi([[REG1:(r[0-9]+)]],[[REG2:(r[0-9]+)]])
; CHECK: += mpyi
diff --git a/llvm/test/CodeGen/Hexagon/swp-stages4.ll b/llvm/test/CodeGen/Hexagon/swp-stages4.ll
index 0d029dc7d2f2e..bddf9cebe7160 100644
--- a/llvm/test/CodeGen/Hexagon/swp-stages4.ll
+++ b/llvm/test/CodeGen/Hexagon/swp-stages4.ll
@@ -3,11 +3,8 @@
; Test that we rename registers correctly for multiple stages when there is a
; Phi and depends upon another Phi.
-; CHECK: = and
-; CHECK: = and
-; CHECK: r[[REGA:[0-9]+]] = memub(r{{[0-9]+}}+#1)
-; CHECK: = and
-; CHECK: r[[REG0:[0-9]+]] = and(r[[REG1:[0-9]+]],#255)
+; CHECK: jump
+; CHECK-NEXT: r[[REG0:[0-9]+]] = and(r[[REG1:[0-9]+]],#255)
; CHECK-NOT: r[[REG0]] = and(r[[REG1]],#255)
; CHECK: loop0(.LBB0_[[LOOP:.]],
; CHECK: .LBB0_[[LOOP]]:
diff --git a/llvm/test/CodeGen/Hexagon/tinycore.ll b/llvm/test/CodeGen/Hexagon/tinycore.ll
index c44038e767194..b20a7831df4d8 100644
--- a/llvm/test/CodeGen/Hexagon/tinycore.ll
+++ b/llvm/test/CodeGen/Hexagon/tinycore.ll
@@ -8,10 +8,15 @@
; CHECK: .LBB0_[[LOOP]]:
; CHECK: {
; CHECK-NEXT: mpy
-; CHECK-NEXT: combine
-; CHECK-NEXT: memw
-; CHECK-NEXT: }
+; CHECK-NOT: memw
+; CHECK: }
+; CHECK: {
+; CHECK: memw
+; CHECK-NOT: memw
+; CHECK: }
+; CHECK: {
; CHECK: memw
+; CHECK-NOT: memw
; CHECK: } :endloop0
; Test the loop contains a single packet with 4 instructions.
diff --git a/llvm/test/CodeGen/PowerPC/2013-07-01-PHIElimBug.mir b/llvm/test/CodeGen/PowerPC/2013-07-01-PHIElimBug.mir
index 8bdf719f4bb5b..59f1477b5c37f 100644
--- a/llvm/test/CodeGen/PowerPC/2013-07-01-PHIElimBug.mir
+++ b/llvm/test/CodeGen/PowerPC/2013-07-01-PHIElimBug.mir
@@ -80,10 +80,9 @@ body: |
# CHECK-NEXT: %15:g8rc = COPY killed %6
# CHECK: bb.3:
# CHECK: %10:g8rc = COPY killed %15
-# CHECK-NEXT: %9:g8rc = COPY killed %14
+# CHECK-NEXT: %16:g8rc_and_g8rc_nox0 = COPY killed %14
# CHECK-NEXT: %14:g8rc = COPY killed %10
# CHECK-NEXT: %15:g8rc = IMPLICIT_DEF
-# CHECK-NEXT: %16:g8rc_and_g8rc_nox0 = COPY killed %9
# CHECK-NEXT: BCC 68, %7, %bb.3
# CHECK-NEXT: B %bb.4
# CHECK: bb.4:
diff --git a/llvm/test/CodeGen/PowerPC/phi-eliminate.mir b/llvm/test/CodeGen/PowerPC/phi-eliminate.mir
index 72f778286abe4..a4b18e648b7a2 100644
--- a/llvm/test/CodeGen/PowerPC/phi-eliminate.mir
+++ b/llvm/test/CodeGen/PowerPC/phi-eliminate.mir
@@ -195,12 +195,9 @@ body: |
; CHECK: bb.4:
; CHECK: successors: %bb.5(0x80000000)
; CHECK: %44:g8rc_and_g8rc_nox0 = COPY killed %59
- ; CHECK: %43:gprc = COPY killed %57
- ; CHECK: %41:gprc = COPY killed %60
- ; CHECK: %39:g8rc = COPY killed %44
- ; CHECK: %61:gprc = COPY killed %41
- ; CHECK: %62:g8rc_and_g8rc_nox0 = COPY killed %39
- ; CHECK: %63:gprc = COPY killed %43
+ ; CHECK: %63:gprc = COPY killed %57
+ ; CHECK: %61:gprc = COPY killed %60
+ ; CHECK: %62:g8rc_and_g8rc_nox0 = COPY killed %44
; CHECK: bb.5:
; CHECK: successors: %bb.6(0x80000000)
diff --git a/llvm/test/CodeGen/PowerPC/ppcf128-freeze.mir b/llvm/test/CodeGen/PowerPC/ppcf128-freeze.mir
index 474c288bba88b..4cad98eeade77 100644
--- a/llvm/test/CodeGen/PowerPC/ppcf128-freeze.mir
+++ b/llvm/test/CodeGen/PowerPC/ppcf128-freeze.mir
@@ -1,21 +1,10 @@
+# NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py UTC_ARGS: --version 5
# RUN: llc -mtriple powerpc64le-unknown-linux-gnu -start-after=codegenprepare \
# RUN: -o - %s -verify-machineinstrs | FileCheck %s
--- |
define ppc_fp128 @freeze_select(ppc_fp128 %a, ppc_fp128 %b) {
- %sel.frozen = freeze ppc_fp128 %a
- %cmp = fcmp one ppc_fp128 %sel.frozen, 0xM00000000000000000000000000000000
- br i1 %cmp, label %select.end, label %select.false
-
- select.false: ; preds = %0
- br label %select.end
-
- select.end: ; preds = %0, %select.false
- %sel = phi ppc_fp128 [ %a, %0 ], [ %b, %select.false ]
- ret ppc_fp128 %sel
- }
-
- ; CHECK-LABEL: freeze_select
+ ; CHECK-LABEL: freeze_select:
; CHECK: # %bb.0:
; CHECK-NEXT: xxlxor 0, 0, 0
; CHECK-NEXT: fcmpu 1, 2, 2
@@ -28,8 +17,19 @@
; CHECK-NEXT: # %bb.1:
; CHECK-NEXT: crnor 20, 7, 2
; CHECK-NEXT: bclr 12, 20, 0
- ; CHECK-NEXT: # %bb.2: # %select.false
- ; CHECK-NEXT: fmr 1, 3
+ ; CHECK-NEXT: # %bb.2: # %select.false
; CHECK-NEXT: fmr 2, 4
+ ; CHECK-NEXT: fmr 1, 3
; CHECK-NEXT: blr
+ %sel.frozen = freeze ppc_fp128 %a
+ %cmp = fcmp one ppc_fp128 %sel.frozen, 0xM00000000000000000000000000000000
+ br i1 %cmp, label %select.end, label %select.false
+
+ select.false: ; preds = %0
+ br label %select.end
+
+ select.end: ; preds = %0, %select.false
+ %sel = phi ppc_fp128 [ %a, %0 ], [ %b, %select.false ]
+ ret ppc_fp128 %sel
+ }
...
diff --git a/llvm/test/CodeGen/PowerPC/pr116071.ll b/llvm/test/CodeGen/PowerPC/pr116071.ll
index 29f11fc1d3a63..5db84436c22f6 100644
--- a/llvm/test/CodeGen/PowerPC/pr116071.ll
+++ b/llvm/test/CodeGen/PowerPC/pr116071.ll
@@ -1,9 +1,26 @@
+; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py UTC_ARGS: --version 5
; RUN: llc -disable-ppc-vsx-fma-mutation=false -mcpu=pwr10 -verify-machineinstrs \
-; RUN: -ppc-asm-full-reg-names -mtriple powerpc64-ibm-aix7.2.0.0 < %s | FileCheck %s
+; RUN: -ppc-asm-full-reg-names -mtriple powerpc64-ibm-aix7.2.0.0 < %s | FileCheck %s
target datalayout = "E-m:a-Fi64-i64:64-n32:64-S128-v256:256:256-v512:512:512"
define void @initial(<2 x double> %0){
+; CHECK-LABEL: initial:
+; CHECK: # %bb.0: # %entry
+; CHECK-NEXT: xxlxor vs0, vs0, vs0
+; CHECK-NEXT: xxlxor f2, f2, f2
+; CHECK-NEXT: xxlxor f4, f4, f4
+; CHECK-NEXT: xxlxor f3, f3, f3
+; CHECK-NEXT: xvmuldp vs1, vs34, vs0
+; CHECK-NEXT: .align 5
+; CHECK-NEXT: L..BB0_1: # %for.cond251.preheader.lr.ph
+; CHECK-NEXT: #
+; CHECK-NEXT: fmr f5, f3
+; CHECK-NEXT: xsadddp f3, f3, f4
+; CHECK-NEXT: fmr f4, f5
+; CHECK-NEXT: xxmrghd vs3, vs3, vs2
+; CHECK-NEXT: xvmaddmdp vs3, vs0, vs1
+; CHECK-NEXT: b L..BB0_1
entry:
%1 = fmul <2 x double> %0, zeroinitializer
br label %for.cond251.preheader.lr.ph
@@ -18,9 +35,3 @@ for.cond251.preheader.lr.ph: ; preds = %for.cond251.prehead
%7 = extractelement <2 x double> %6, i64 0
br label %for.cond251.preheader.lr.ph
}
-
-; CHECK: xsadddp f4, f3, f4
-; CHECK-NEXT: xxmrghd vs5, vs4, vs2
-; CHECK-NEXT: fmr f4, f3
-; CHECK-NEXT: xvmaddmdp vs5, vs0, vs1
-; CHECK-NEXT: fmr f3, f5
diff --git a/llvm/test/CodeGen/SystemZ/swifterror.ll b/llvm/test/CodeGen/SystemZ/swifterror.ll
index 1b18287cac146..39f0907295ff4 100644
--- a/llvm/test/CodeGen/SystemZ/swifterror.ll
+++ b/llvm/test/CodeGen/SystemZ/swifterror.ll
@@ -162,8 +162,8 @@ define float @foo_loop(ptr swifterror %error_ptr_ref, i32 %cc, float %cc2) {
; CHECK-O0: je
; CHECK-O0: lghi %r2, 16
; CHECK-O0: brasl %r14, malloc
-; CHECK-O0: lgr %r[[REG1:[0-9]+]], %r2
-; CHECK-O0: mvi 8(%r[[REG1]]), 1
+; CHECK-O0: lgr %r{{[0-9]+}}, %r2
+; CHECK-O0: mvi 8(%r2), 1
; CHECK-O0: jnh
; reload from stack
; CHECK-O0: lg %r9, [[OFFS:[0-9]+]](%r15)
diff --git a/llvm/test/CodeGen/X86/statepoint-cmp-sunk-past-statepoint.ll b/llvm/test/CodeGen/X86/statepoint-cmp-sunk-past-statepoint.ll
index 391963de79703..089ec7ae7658e 100644
--- a/llvm/test/CodeGen/X86/statepoint-cmp-sunk-past-statepoint.ll
+++ b/llvm/test/CodeGen/X86/statepoint-cmp-sunk-past-statepoint.ll
@@ -60,7 +60,7 @@ zero:
; CHECK: JMP_1 %bb.4
; CHECK: bb.4
; CHECK: bb.5
-; CHECK: %3:gr64 = COPY %10
+; CHECK: %20:gr64 = COPY %10
; CHECK: %4:gr64 = COPY killed %10
; CHECK: %4:gr64 = nuw ADD64ri32 %4, 8, implicit-def dead $eflags
; CHECK: TEST64rr killed %1, %1, implicit-def $eflags
diff --git a/llvm/test/DebugInfo/MIR/InstrRef/phi-regallocd-to-stack.mir b/llvm/test/DebugInfo/MIR/InstrRef/phi-regallocd-to-stack.mir
index ed04647c84066..31c694ee1e014 100644
--- a/llvm/test/DebugInfo/MIR/InstrRef/phi-regallocd-to-stack.mir
+++ b/llvm/test/DebugInfo/MIR/InstrRef/phi-regallocd-to-stack.mir
@@ -70,8 +70,7 @@ frameInfo:
machineFunctionInfo: {}
body: |
; CHECK-LABEL: bb.0:
- ; CHECK: renamable $ebp = COPY $edi
- ; CHECK: MOV32mr %stack.1, 1, $noreg, 0, $noreg, killed renamable $ebp
+ ; CHECK: MOV32mr %stack.[[SLOT:.+]], 1, $noreg, 0, $noreg, $edi
bb.0:
successors: %bb.2(0x50000000), %bb.1(0x30000000)
liveins: $edi, $esi
@@ -142,8 +141,8 @@ body: |
%64:gr32 = PHI %24, %bb.0, %44, %bb.1, debug-location !18
DBG_INSTR_REF !14, !DIExpression(DW_OP_LLVM_arg, 0), dbg-instr-ref(1, 0), debug-location !12
- ; CHECK: DBG_PHI %stack.1, 1, 32
- ; CHECK: renamable $eax = MOV32rm %stack.1,
+ ; CHECK: DBG_PHI %stack.[[SLOT]], 1, 32
+ ; CHECK: renamable $eax = MOV32rm %stack.[[SLOT]],
; CHECK: DBG_INSTR_REF {{.+}}, dbg-instr-ref(1, 0)
$eax = COPY killed %0, debug-location !19
RET 0, killed $eax, debug-location !19
More information about the llvm-commits
mailing list